Upload
others
View
10
Download
0
Embed Size (px)
Citation preview
Wide Band Gap Electronic Materials
NATO ASI SeriesAdvanced Science Institutes Series
A Series presenting the results of activities sponsored by the NA TO Science Committee,which aims at the dissemination of advanced scientific and technological knowledge,with a view to strengthening links between scientific communities.
The Series is published by an international board of publishers in conjunction with the NATOScientific Affairs Division
A Life SciencesB Physics
C Mathematical and Physical SciencesD Behavioural and Social SciencesE Applied Sciences
F Computer and Systems SciencesG Ecological SciencesH Cell BiologyI Global Environmental Change
PARTNERSHIP SUB-SERIES
1. Disarmament Technologies2. Environment3. High Technology4. Science and Technology Policy5. Computer Networking
Plenum Publishing CorporationLondon and New York
Kluwer Academic PublishersDordrecht, Boston and London
Springer-VerlagBerlin, Heidelberg, New York, London,Paris and Tokyo
Kluwer Academic PublishersSpringer-VerlagKluwer Academic PublishersKluwer Academic PublishersKluwer Academic Publishers
The Partnership Sub-Series incorporates activities undertaken in collaboration with NATO'sCooperation Partners, the countries of the CIS and Central and Eastern Europe, in Priority Areas ofconcern to those countries.
NATO-PCO-DATA BASE
The electronic index to the NATO ASI Series provides full bibliographical references (with keywordsand/or abstracts) to more than 30000 contributions from intemational scientists published in allsections of the NATO ASI Series.Access to the NATO-PCO-DATA BASE is possible in two ways:
- via online FILE 128 (NATO-PCO-DATA BASE) hosted by ESRIN,Via Galileo Galilei, 1-00044 Frascati, Italy.
- via CD-ROM "NATO-PCO-DATA BASE" with user-friendly retrieval software in English, Frenchand German (© WTV GmbH and DATAWARE Technologies Inc. 1989).
The CD-ROM can be ordered through any member of the Board of Publishers or through NATOPCO, Overijse, Belgium.
3. High Technology - Vol. 1
Wide Band Gap Electronic Materials
edited by
Mark A. Prelas University of Missouri-Columbia, Columbia, Missouri, U.S.A.
Peter Gielisse Florida A&M University/Florida State University, Tallahassee, Florida, U.SA
Galina Popovici Rockford Diamond Technology, Inc., Columbia, Missouri, U.S.A.
Soris V. Spitsyn Russian Academy of Sciences, Moscow, Russia
and
Tina Stacy University of Missouri-Columbia, Columbia, Missouri, U.S.A.
Springer-Science+Business Media, B.V.
Proceedings of the NATO Advanced Research Workshop on Wide Band Gap Electronic Materials - Diamond, Aluminum Nitride and Boron Nitride Minsk, Belarus May 4-6, 1994
A C.I.P. Catalogue record for this book is available from the Library of Congress.
ISBN 978-94-010-4078-5 ISBN 978-94-011-0173-8 (eBook) DOI 10.1007/978-94-011-0173-8
Printed on acid-free paper
AII Rights Reserved © 1995 Springer Science+Business Media Dordrecht Originally published by Kluwer Academic Publishers in 1995 Softcover reprint of the hardcover 1 st edition 1995 No part of the material protected by this copyright notice may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording or by any information storage and retrieval system, without written permission from the copyright owner.
This book is dedicated to Dr. Steven Lin. His effortsin assembling this book were no less than Herculean.
Additionally we wish to thank Dr. Kim Bigelow ofNorton Diamond Film Division, Dr. Paul Blackborowof ASTex Inc., and John Tompkins of Rockford Technology Corp. for supporting this Workshop financially.
CONTENTS
Preface ••••••••..••••.••.•••.•.•.•••.•.•••••.•..•••.•••.•••••.•••••••.•••••••.••.•.••..•••••••.••••.••••••.•.••.••••••..••.•••• vii
DIAMOND
Growth and Doping
Problems of n-type Diamond Doping. Forced Methods ofDoping •..••.•..........•....• 1Galina Popovici, RDT; and M. A. Prelas, UMC.
Diffusion of Boron, Hydrogen, Oxygen and Lithium in Single Crystallineand Polycrystailine Diamond. A Novel Method for the Determination ofthe State of an Impurity: Forced Diffusion of Boron in IA Type NaturalDiamond 15
Galina Popovici, RDT; T. Sung, M. A. Prelas, UMC;R. G. Wilson, HRL; and S. Khasawinah, UMC.
Chemical Aspects ofDiamond Doping •..••.•••••••..••••.•••..••.•••••••..••••.•.•••.•..•..••.•.•........ 31B.V. Spitsyn, RAS.
Diamond Growth by Hot Carbon Filament Chemical Vapor Deposition ....••..•.. 47C. C. Chao, E. J. Charlson, E. M. Charlson, J. Meese,M. A. Prelas and T. Stacy, UMC.
Diamond Particles on Silicon Tips: Preparation, Structure, andField-Emission Properties 53
E.I. Givargizov, A.N. Stepanova, L.L. Aksenova,E.V. Rakova, P.S. Plekhanov, V.V. Zhirnov,and A.N. Kiselev, RAS.
To the Question of the Diamond Nuclei's Formation from the Gas Phase 63A.P. Rudenko, and I.I. Kulakova, MSU.
Characterization and Properties
ElectricaUy and Optically Active Impurities and Defects in Diamond 69A.A. Gippius, RAS.
Prediction of Diamond Film Thermal Conductivity •..••.•..••••.••.•••••.••••.••.•..•.••.•..•. 81N.V. Novikov, T.D. Ositinskaya, A.P. Podoba,and S.V. Shmegerea, UAS.
viii
Spectral Hole-Burning Study of the Defects Created by NeutronIrradiation in a Natural Diamond 89
I. Sildos, G. Zavt, and A. Osvet, EAS.Calculations of Phosphorous Electronic Levels in Diamond 97
V.V. Tokiy and D.L. Savina, DICE.Hydrogen Chemistry on Diamond Surface 105
J. E. Butler, B. D. Thomas, M. McGonial, J.N. Russell, Jr.,and P.E. Pehrsson, NRL.
Surface and Bulk Conductivity of Hydrogen Treated PolycrystaUineDiamond 115
G.A. Sokolina, L.L. Bouilov, A.A. Botev, A.V. Markin, RAS;and M.A. Timofeev, MSU.
Positron Annihilation in Diamond Films 1231.1. Bardyshev, L.L. Bouilov, and B.V. Spitsyn, RAS.
ESR Study of Paramagnetic Defects in Free Standing Diamond Films 129T. A. Karpukhina, RAS; M. A. Prelas, G. Popovici,S. Khasawinah, UMC; and B. V. Spitsyn, RAS.
Applications
Efficient Reduction of Nitride and Nitrate to Ammonia Using B-dopedDiamond Electrodes 137
C. Reuben, E. Galun, R.Tenne, WI; R. Kalish, Tech;Y. Muraki, K. Hashimoto, A. Fujishima, UT;J.M. Butler, NRL; and C. Levy-Clement, CNRS.
Electronic and Sensing Properties ofDiamond 143J.L. Davidson, VU.
Diamond MIS Capacitors With Silicon Dioxide Dielectric 161M.J. Marchywka, D. Moses, and P.E. Pehrsson, NRL.
Diamond Photovoltaics: Characterization of CVD Diamond Film-BasedHeterostructure for Light to Electricity Conversion 171
PJ. Perov, V.1. Polyakov, A.V. Khomich, N.M. Rossukanyi,A.I. Rukovishnikov, V.P. Vamin, and I.G. Teremetskaya, RAS.
Laser Modes in Diamond 187L.-T. S. Lin, M.A. Prelas, UMC; and G. Popovici, RDT.
Advanced Applications ofDiamond Electronics 207C.B. Wallace, BDM.
Laser-assisted Chemical Etching of Diamond Films in Oxygen 219V. G. Ralchenko, K. G. Horotushenko, A. A. Smolinand E. D. Obraztsova, RAS.
Ion Milling of PolycrystaUine Diamond Films 225A. E. Alexenko, RAS; A. F. Belyanin, CRTI; L. L. Bouilov, RAS;A. P. Semenov, SRAS; and B. V. Spitsyn, RAS.
ix
AMORPHOUS AND DIAMOND-LIKE CARBON FILMS
Growth and Doping
Doping ofDiamond-Like Carbon Films •••••••••.•..- 235S. Mitura, TUL; J. Szmidt, and A Sokolowska, WUT.
Unhydrogenated DLC Films Obtained by Magnetron Sputtering 243C. Morosanu, IPTM; N. Tomozeiu, UB; C. Cordos,and T. Stoica, IPTM.
Simulation ofDitJusion in an Amorphous Structure 249A. V. Nazarov, IBMI.
Characterization and Properties
Optical and Electrical Properities of Quantum-dimentional MultilayerStructures Based on Carbon Films 257
V.V. Sleptsov, V.M. Elinson, AM. Baranov,and S.A Tereshin, NPO.
Thermal Stability and Structural Reactions at the Tantalumla-CInterface Under Vacuum Annealing Conditions 265
AP. Novikov, E.A. Shilova, L.D. Buiko,and VA Zaikov, RCEMT.
Extended and Localized Electronic States in Tetrahedral Carbon Films .......... 271V.E. Maschenko, ISA;V.M. Puzikov, UAS;and A.V. Semenov, SRAS.
Optical Properties of Sputtering and Glow Discharge a-C:H Films 285T. Stoica, A. Dragomir, IPTM; M. Gartner, IPCR; C. Morosanu,and G. Pavelescu, IPTM.
Applications
Application ofAmorphous Hydrogenated Carbon Coating toSemiconductor Radiation Detectors 291
I.M. Kotina, T.A Antonova, G.V. Patsekina, V.D. Saveliev,L.M. Tuhkonen, PNPI; 0.1. Konkov, and E.I. Terukov, IP17.
OTHER WIDE BANDGAP SEMICONDUCTORS
Growth and Doping
Device for Growing and Doping in the Growth Process ofThin AIN Films 297
AF. Belyanin, CRTI; A.P. Semenov, SRAS;and B.V. Spitsyn, RAS.
x
Peculiarities of Chemical Vapor Heteroepitaxy of Wide Band Gap111-V Nitrides 305
E.B. Sokolov, G.A. Naida, and N.V. Barovskii, MSIEE.The Peculiarities of Cubic Boron Nitride Formation MechanismUsing Hexa-Ammonicate Boron Hydride ofMagnesium 313
H.I. Polushin,MSSAV; and KP. Burdina,MSV.Investigation of Cubic Boron Nitride Crystallization Processes inthe BN-Li3N-(H~N) System 321
V.B. Shipilo, L.M. Gameza, and A.I. Lukomskii, ASB.Epitaxial Growth ofAIN by Plasma Source Molecular Beam Epitaxy 329
G.W. Auner, T.D. Lenane, F. Ahmad, R Naik, P.K Kuo,and Z. Wu, WSV.
Characterization and Properties
Electronic Structure and Related Properties ofTetrahedrally BondedWide-Band-Gap Materials Containing Early Elements ofthe Periodic Table 335
W.RL. Lambrecht, C.H. Lee, K Kim, A.G. Petukhov,E.A. Albanesi, and B. Segall, CWRV.
Ion·Implantation into Wide Bandgap Semiconductors 373V.S. Vavilov, PNLPI.
Thermodynamic Properties ofBoron Nitride 377V.L. Solozhenko, VAS; and K.S. Gavrichev, RAS.
Electrical Conductivity ofCeramic Based on Different Boron NitrideModifications 393
A.V. Bochko, VAS; G.A. Sokolina, and S.V. Bantsekow, RAS.Cathodoluminescent Investigation of External Factors Influenceon Defective Cubic Boron Nitride Structure 397
V.B. Shipilo, E.M. Shishonok, A.I. Lukomskii,and L.M. Gameza, ASB.
Macro and Micro Structural Factors in Thin Film Growth ofIII-V Compounds 401
PJ. Gielisse and H. Niculescu, FAMV.The Features of the Sintering Process Under High Pressure ofAluminum Nitride Ceramic with High Thermal Conductivity 421
V.B. Shipilo, T.V. Rapinchuk, and N.A. Shishonok, ASB.Reactive Ion Etching ofSilicon Carbide with Fluorine ContainingPlasmas 427
V.E. Sizov, and KV. Vassilevski, CREE.1.54-Jlm Photoluminescence from Er-Implanted AIN & GaN 431
RG. Wilson, R.N. Schwartz, HRL; c.R. Abernathy,S.1. Pearton, VF; N. Newman, M. Rubin, T. Fu, LBL;and J.M. Zavada, ARO.
xi
AES-SIMS Analitical System for Compositional Measurements ofWide Band Gap Semiconductors •.••••.•••..••••.••.....•..•••••....•..••.••.•••••...•••..•.•...•..••.•..••.. 437
A.I. Babanin, IPTI and CREE; and A.A. Lavrentev, SEU.Positron Annihilation in Sintered Boron Nitride 447
1.1. Bardyshev, and A.D. Buravov, RAS.
Applications
Wide Band Gap Electronic Devices 453V.E. Chelnokov, IPTl; K.V. Vassilevski, CREE;and V.A. Dmitriev, CR!.
Wide Band-gap Photovoltaics .....•.••.•...•.••.....••.•........••.•.....•............••••..•.••..•..••..•..•.. 463M. A. Prelas, UMC; G. Popovici, RDT; S. Khasawinah,and T. Sung, UMC.
Considerations in Further Development ofAluminum Nitrides asa Material for Device Applications 475
T. Stacy, B. Y. Liaw, A. H. Khan, and G. Zhao, UMC.Theoretical Aspects ofAluminum Nitride and Diamond in View ofLaser and Photovoltaic Action 487
H. Hora, UR; R. Hopfl, HT; and M. A. Prelas, UMC.
Oral Presentations 511Poster Presentations 513List of Participants 519Affiliations Key 525Author Index 527Key Word Index 529
PREFACE
Minsk, Belarus was the site of the NATO ARW on Wide Band-Gap electronicMaterials May 3 through 6,1994; 143 participants and observers from 15 countries metfor the NATO Advanced Research Workshop on Wide Band-Gap Electronic Materials(NATO ARW). The meeting was marked by a remarkable free exchange between eastand west on these topics by revealing technical achievements not widely known oravailable in the west because of past political climate or present economic realities in
the Newly Independent States.
The topics ranged from electron doping of diamond, n-type diamond, negativeelectron affinity of diamond, applications of aluminum nitride, doping of boron nitride,
wide band gap electronic applications, to nanophase diamond.
Of the many high-lights during the scientific meetings, an energy sub band due todefects in the diamond lattice was described. These defects are responsible for the lightemission from a diamond Light Emitting Diode (LED) which was demonstrated at theNATO ARW. This diamond LED can emit red, green, and blue light. The potential for"high tech" nanostructure electronic devices such as quantum transistors was describedwhich might some day revolutionize electronics. The prospects of aluminum nitride foracusto devices, piezo devices, and electroluminescence devices were discussed.
Other technologies that originated in the Newly Independent States were also presented. The most significant is the mass production of nanophase diamond particles of anarrow size distribution around four nanometer which are created by an explosiveshock wave. Four nanometers is about the length of 40 individual atoms laid side byside. The nanophase diamond process is one of the many fruits of the weapon conversion program in the Newly Independent States. Nanophase diamond particles can beused as seed materials to grow smooth diamond films, as an additive to form new composite materials with superior properties, for nanophase electronic devices, and as abase material for new super lubricants.
Overall, the workshop focused on problems which will have the most impact on thedevelopment of electronic devices from wide band-gap electronic materials. Theseproblems included the production of n-type diamond, heteroepitaxy of diamond films,
xiii
xiv
nanophase diamond electronic devices, doping of aluminum nitride, and growth of large
crystals of boron nitride. The significance of these problems is in the potential of a new
generation of electronic devices.
Mark A. Prelas
University ofMissouri
Columbia, Missouri
November, 1994
NATO Advanced Research Workshop on
WIDE BANDGAP ELECTRONIC MATERIALS
Diamond, Aluminum Nitride and Boron Nitride
May 4-6,1994, Minsk, Belarus
Director: M. A. Prelas, Nuclear Engineering DepartmentUniversity ofMissouri-ColumbiaE2433 Engineering Building East, Columbia, MO 65211, USA.
Co-director: B. V. Spitsyn, Institute of Physical ChemistryRussian Academy of Sciences31 Leninsky Prospekt, Moscow 117915, Russian Federation.
Local Coordinator: V. Varichenko, Semiconductor DepartmentBelarussian State University4 Skarina Prospekt, Minsk 220080, Belarus.
Organizing Committee:
M.A. Prelas, University ofMissouri-Columbia, USA.B.Y. Spitsyn, Institute ofPhysical Chemistry, Russia.A. Zaitsev, Belarussian State University, BelarusT. Stacy, University ofMissouri-Columbia, USAH. Hora, University of Regensberg, Germany
Workshop Sponsors:
NATONorton Diamond Film DivisionASTeXRockford Technology Corporation
The Government of BelarusBelarussian Academy ofSciencesThe University of Missouri, USAThe Florida A&M University/Florida State University, USAS. I. DiamondDiagascrown
Plasmoteg Eng. Ctr.
PROBLEMS OF N-TYPE DIAMOND DOPING.FORCED METHODS OF DOPING
A review
GALINA POPOVICIo * and M. A. PRELAS*
°Rockford Diamond Technology, 501 S. Sixth Street, Champaign,lL61820-5579*Nuclear Engineering Department, University ofMissouri - Columbia,Columbia, MO 65211
AbstractThe experimental and theoretical results on would-be donor
impurities in diamond lattice: N, P, Li, and Na, are reviewed. Newmethods of forced diffusion and ion assisted doping during growth arediscussed. Experimental data on n-type conductivity obtained by forceddiffusion into the diamond CVD films are also presented.
1. Introduction
A major goal of diamond thin film technology research has been thereproducible production of p-n junctions, which are the basic units ofmany electronic devices. While p-type conductivity is relatively easilyattained by boron doping, n-type conductivity has proved much harder toachieve. The experimental and theoretical results on would be donorimpurities are reviewed. In analogy with classical semiconductors, wewill discuss the possibility of obtaining n-type diamond by usingsubstitutional impurity atoms (nitrogen and phosphorus) and interstitialatoms (Li and Na). It is shown that simple hydrogen-like model of theimpurities does not work for n-type diamond doping. Nonordinaryapproaches should be found for diamond.
M.A. Prelas et at. (eds.), Wide Band Gap Electronic Materials, 1-13© 1995 Kluwer Academic Publishers.
2
New methods of forced diffusion and ion assisted doping duringgrowth are discussed. The methods of forcing the introduction ofimpurities into the diamond lattice have an important advantage over theion implantation, widely used for diamond doping. The ion implantationintroduces structural defects (vacancies, vacancy+interstitial, and theircombinations) that is difficult to cure. No additional structural defects,except that inherent to the impurity itself, are introduced by forceddiffusion or ion assisted doping during growth.The experimental data on obtaining n-type conductivity are given in
this paper.
2. Hydrogen-like model of substitutional impurities in Si anddiamond
N- and p-type doping in classical semiconductors are attained usuallyby substituting a fraction of the host lattice atoms by impurity atoms ofneighboring groups of the periodic table of elements. To dope Si andGe, elements of the III group are used as acceptors and those of the Vgroup as donors. The behavior of III and V group atoms in the Si latticeis well described by the simple theoretical model of hydrogen-likeimpurities.
According to the model, a donor atom in the lattice is similar to ahydrogen atom: it is a positive ion keeping, by Coulomb force, its extraelectron, which is not used for covalent bonding by the host lattice.How strong is the interaction between the electron and the ion isdetermined by two factors: the charge carrier effective mass and thescreening ability of the host lattice. This model takes into account onlythe properties of the host lattice. There is a tachit assumption that theimpurity atom is not much different of the host lattice atom,dimensionaly and/or energeticaly. However, some properties of theimpurity itself, like the bonding energy between the impurity nucleus andits electrons, are sometimes quite important. The bonding energy ofelectrons to the impurity atom is strong in atoms with small atomicnumber, where the outer electrons are weakly screened from their nuclei.This is the reason why the hydrogen-like model does not work for thenitrogen impurity in silicon. Nitrogen does not form a shallow donor inthe Si band gap. The energy of its level observed experimentally is 190meV [1-3], whereas the hydrogen-like model gives ilE=25 meV.[4]Indium also does not form a shallow level in the Si band gap (ilE =0.16eV [4]), however due to the other reason. The In atom is big (covalentradius r (In) = 1.62 A comparing to r (Si) = 1.18 A) [5]. Thus, thedistortion of the host Si lattice becomes quite important in the energy
3
balance.
Diamond is unique in its electronic structure, its four outer electronsare bonded much more strongly to the nucleus than those of Si. Indiamond [6] the hydrogen-like model predicts for substitutionalimpurities of the TIl group an acceptor activation energy &3a=0.36 eV
(flf,0=5.7 and mwmo=O.75) [7] and for the V group a donor activation
energy &3d=O.19 eV (meJlllo=O.48) [8]. The covalent radius of carbonatom in diamond is small, only O. 77A. So only boron (r = 0.80 A) andnitrogen (r= 0.6 A) have covalent radii small enough to enter diamondlattice without distorting it. Boron acceptor level has ionization energy of-0.37 eV, close to that predicted by hydrogen like model. However,aluminum is too big an atom ( r= 1.43 A, nearly twice as diamondcovalent radius). Al is not active center in diamond, neither electric, noroptical. The nitrogen and phosphorus behavior in the diamond lattice isdisscussed in the next chapter.
3. Substitutional impurities in diamond
3.1, Nitrogen
Nitrogen can enter in substitutional positions at concentrations up to1019 cm -3. [9] The energy level of the nitrogen substitutional impurityin diamond lattice was calculated by different methods with similarresults. [10-12] Briddon et al.[1O] used an 86 atom cluster with thecentral carbon atom replaced by nitrogen.They found that the resultingelectronic structure consists of two levels: a level 1.2 eV above thevalence band and a level 0.5 eV below the conduction band. A trigonaldistortion of 28% was computed.
Bemholc et al. [11-12] advanced the following model. In thesubstitutional position three nitrogen electrons are bound to carbonatoms. Two remaining electrons form a lone pair bound to the fourthcarbon atom. There is no loosely bound electron to form a hydrogen-likeshallow level. A distortion of the diamond lattice of 25% was computedalong the bond corresponding to the lone pair, due to the chargerepulsion of the lone pair and the carbon electron. The energy level ofthis state was found to be of 1.5 eV. The experimental value of thenitrogen energy level is 1.7-2.1 eV below the bottom of the conductionband. [13-14]
The single-substitutional nitrogen is a very effective recombinationcenter for electron-hole pairs. The intrinsic photocurrent generated by
4
far-UV photons decreases dramatically with increasing nitrogenconcentration [15-16]. The deep nitrogen donor level is an effectivecompensator of acceptor states, reducing by 6 orders of magnitude theconductivity of diamond films at room temperature [17]. Therefore,nitrogen is a poisoning impurity for conductive diamond layers.
3.2. Phosphorus
Different methods of theoretical calculation for the position of thedonor level of the substitutional phosphorus impurity in diamond bandgap gave different results. Theoretical studies by Bernholc, Kajihara etal. [11-12] ,using pseudopotentials and a plane wave basis forimpurities, embedded in a periodically repeated supercell, found theactivation energy of phosphorus impurity centers in substitutional sites tobe of 0.20 eV. However, the equilibrium solubility of phosphorus waspredicted to be very low, even at high temperatures. On the other hand,self consistent local density approximation cluster calculation on C:Nand C:P impurities [18] found the donor levels in both systems to bedeep levels, lying well below the conduction-like states. An activationenergy of 1.09 eV was found for phosphorus and of 0.9 eV for nitrogenimpurities.
Doping by phosphorus during growth of CVD diamond [19-21]gave a deep level with an activation energy of 0.84-1.16 eV, as it waspredicted theoretically by Jackson et al [18]. There also exist inliterature the claimes of the succesful doping of diamond by phosphorus,forming a shallow level. [22] However, the result could not bereproduced by other groups. There have also been unsuccessful attemptsto dope synthetic HTHP diamond by phosphorus. [23] There was noeffect on electrical conductivity and no phosphorus was detected indiamond by secondary-ion mass spectroscopy.
There is also a possible explanation for obtaining a deep level due tophosphorus doping. To explain the results of Electron Spin Resonance(ESR) on the phosphorus impurity in diamond, Tokiy et al. [24]calculated a 29 atom carbon cluster with the central atom replaced byphosphorus. This model did not give a satisfactory explanation of theexperimental results. The authors did find, however, a model whichexplained the experimental ESR results. In that model the paramagneticphosphorus is the second nearest neighbor of a vacancy placed in thecenter of the cluster. This result suggests that phosphorus does not enterin singly dispersed form, but forms impurity-vacancy complexes, whichshould have an activation energy different from that of singly
5
substitutional phosphorus. The fonnation of a phosphorus-vacancycomplex may be a requirement to accommodate the large phosphorusatom.
4. Interstitial impurities
Impurity atoms can enter in interstitial sites of the diamond lattice in
the tetrahedral or hexagonal voids of the lattice. It was showntheoretically [25] that in the diamond lattice for a small ion, for whichthe repulsive energy is small, the hexagonal site is an equilibrium siteand the tetrahedral site is a saddle point. For large ions, the repulsiveenergy will dominate the picture, and the tetrahedral site will be theequilibrium position.
The most promising interstitial dopants for diamond are lithium andsodium. The ionic radius of Li is of 0.060 nm and that of Na is of 0.095nm. But their atomic radii are quite big: r (Li)=1.52 A and r (Na) = 1.66A). If they diffuse in the neutral fonn through the diamond lattice, thelattice distortions will be strong.
The energies of Li and Na impurities in diamond in interstitial andsubstitutional positions have been calculated. [11] As expected, theinterstitial sites were found to be energetically favored. The tetrahedralsites are preferable for both Li and Na, because even the lithium ion in aninterstitial site is not small enough for the diamond lattice. Activationenergies of 0.1 and 0.3 eV below the conduction band minimum for Liand Na, respectively, have been computed. The solubility of thesedopants is expected to be low. The activation energy of diffusion for Liand Na impurities has been predicted to be 0.85 and 1.4 eV, respectively.[11-12] The mobility of lithium in the diamond lattice is predicted to behigh. [12]
Experiments on Li and Na doping have not confinned the theoreticalresults. Doping by lithium was tried by in situ doping [26] , by diffusion
[27] and by ion implantation [20]. None of these methods worked.Lithium fluoride was used as an in situ doping source [26] forhomoepitaxial films grown by RF plasma discharge. SIMS analysisshowed that Li can enter up to - 1021 cm-3 in the most heavily dopedsamples. The films obtained had p-type conductivity and an activationenergy of 0.24 eV, the same as on boron doped films. The conclusion ofthis experiment was that Li is electrically inactive in diamond and theobserved results were due to a boron contamination (the growth systemhad also been used for boron doping). However, there might be otherexplanation. Elements of the seventh group of the periodical table
6
entering interstitial sites should form shallow acceptor levels in diamond.So fluorine from the dopant source (LiF) might be responsible for p-typeconductivity. Another explanation of the failure might be that Li level isdeeper as predicted and it difuses as a neutral atom. In the last case thelattice distortions will be strong and can introduce compensatingacceptor impurities.
The diffusion of lithium into diamond from the vapor phase has beenstudied at temperatures from 400 to 900 OC [29]. The presence oflithium was qualitatively determined by SIMS. However, the change ofelectrical conductivity due to the presence of lithium was small, of only
two orders of magnitude, from 10-11 to 10-9 a-lcm- l . The authorsadvanced the hypothesis of the existence of acceptor-like states close tothe valence-band edge and extended 1-2 eV into the band gap. A highenough density of such states might be responsible for the compensationof lithium donors. The nature of these states is not known. The statesmight be formed due to impurities, defects or localized valence bondanomalies such as carbon-carbon double bonds. No comprehensiveresearch was done so far to verify these hypothesis
Na impurity does also not behave as predicted by theory [11].Jamison et al [28] obtained that Na is a p-type dopant in diamond andforms a shallow level (-0.09 eV). These results are discussed later.
Our work on diffusion of impurities from lithium salts and obtainingof n-type conductivity are discussed in the next chapter.
5. Forced doping methods
The methods of forcing the introduction of impurities into thediamond lattice have an important advantage over the ion implantation,widely used for diamond doping [29]. The ion implantation introducesstructural defects (vacancies, vacancy+interstitial, and theircombinations) that is difficult to cure.[30] No additional structuraldefects are introduced by forced diffusion or ion assisted doping duringgrowth.
5.1. Electric Field Assisted Diffusion
The Debye temperature of diamond is very high, about 2200 K. It isthe highest of all known solids, reflecting the high rigidity of thediamond lattice. [31] For comparison, the Debye temperature of siliconis 635 K. At temperatures lower than the Debye temperature, atomsoscillate with amplitudes small compared to the interatomic distance.
7
The diffusion of substitutional impurities is controlled by the vacancymechanism. The vacancy formation energy of diamond is high, 7.2 eV(for comparison, that of Si is 4.2 eV) [32] Hence, the temperatures thatwould be necessary for effective diffusion of substitutional impurities indiamond are expected to be higher than the maximal temperature whichcan be used for diffusion, which is the temperature at which thegraphitization of diamond films begins (1900 K). However, promisingresults were obtained by using solid BN sources for diffusion of boronand nitrogen in natural diamond. [33] Rapid thermal processing (1400°C for 30 sec in argon) was used. SIMS profiles indicated a boronconcentration of _1019 cm-3 to a depth of 500 A. A diffusioncoefficient of 10-12 cm2s-1 at 1400 0(: was determined.
It is known that fast diffusing interstitial impurities, like Cu, Li, Na,can be introduced in silicon and germanium lattices at relatively lowtemperatures (200-400 0(: for Si, which is much less than the diffusiontemperature for substitutional impurities, of 800-1100 0(:), if electricfield assisted diffusion is used. [34-35] However, to our knowledge, thefield enhanced diffusion in diamond has not been tried yet. We proposeto apply an electric fields to enhance the diffusion of Li in diamond thinfilms. If it is true that the predicted ionization energy of Li level issmall( 0.1 eV ) [11-12], Li atoms will be totally ionized at 1160 K.Therefore, the field enhanced diffusion of Li should be effective.
We studied the diffusion of lithium, chlorine and oxygen underbias.[36] High quality free standing diamond films, of Norton Co, 230/..Im thick polished on both sides have been used. The average crystallitesize was of the order of tens of micrometers. Samples were grownunder the same conditions. The Raman spectra of the samples had thediamond line only, with no lines of graphite and amorphous carbon. Thecathodoluminescence spectra showed a strong free exciton line which isa feature of good crystalline quality diamond films [37]. The films weremounted on a graphite base with an embedded tungsten heater. The basetemperature was monitored by a chromel-alumel thermocouple. Thedopant sources Li2C03 and LiCI04 were of analytical purity (99%).
Thediffusion was performed for 190 min at 1000 0C in argonatmosphere. The experimental arrangement is shown in Fig. 1. Anelectric field of 200 V was applied to the samples. The control samplehad no electric field applied. A large amount of impurities (-4x1019
cm-2) was found to diffuse at relatively low temperatures. On the otherhand, the bias was found not to affect the concentration of the diffused
8
Diamond Plate
ImpI,Jrity
~4 •
Bias
+
Diamond Plates
Fig. 1 Experimental arrangement for the forsed diffusion.
impurities. The last can be explain by the high diffusion coefficientthrough the grain boundaries and other defects inherent to polycrystallinefilms. As the ionization energy of doping atoms in association withother defects of the lattice must be different, usually larger, than theenergy level of the same impurity in the singly dispersed form, thediffusion through the defects on the grain boundaries was probably dueto the neutral atoms and was not influenced by electric field. Thediffusion experiments on the Li doping in a type ITa natural diamond,using the nearly same conditions, showed that Li content is two tothree orders ofmagnitude smaller in the single crystalline diamond, as inthe polycristalline films [38]
The diffusion in polycrystalline films lead to a large change inconductivity (-8 orders of magnitude). An activation energy of -0.25eV was determined from the temperature dependence of the conductivity[39]. The samples showed p-type conductivity, as measured byconventional hot probe. The hot probe measurements under biasshowed n- or p-type conductivity depending on the applied bias. Ahypothesis was advanced, that p-type surface inversion layer was formedon the diffused n-type layer.[40] More research is needed to clarify thisproblem.
The Hall effect measurements on the polycrystalline films diffusedunder bias using other Li salt (LiCI04) were done by two independentgroups. The measurements showed n-type conductivity with electronconcentration about 1015 cm-3 and sheet resistance about 105Ohm/square.[37 ,39] Supposing the thickness of the diffused layer to beof order of 2 J.lm, the mobility of electrons - 50 cm2Ns was estimated.
9
We don't know to which impurity ( or might be to association ofimpurities) n-type conductivity is due in this case. More research isneeded to clarify this problem. However the possibility of n-typediamond doping was demonstrated.
5.2. Ion assisted dopin!: durin!: !:rowth
The basic difficulty with conventional doping during growth is withthe incorporation of dopant atoms into the growing film. Many dopantsmore likely re-evaporate rather than be incorporated into a growing filmsurface, as their bonding energy is small in comparison with the C-Cbond energy (see Table I of ref. 6). Even if a dopant atom isincorporated into the growing film, there may also exist driving forcesdue to the strain energy, related to geometrical factors, which may causethe dopant atom to segregate to the surface and eventually re-evaporate.
Large improvements in the incorporation of In and Sn in Si and GaAsgrown by molecijlar beam epitaxy have been demonstrated [41) usinglow energy (60-300 eV) ionized dopant atoms, which were directed ontothe film during deposition. Ion assisted doping gave also good results forobtaining p-type conductivity in CdTe using As and P ions as dopants[42-44). The ionized dopant atoms are implanted to a small depth andhave a much larger sticking coefficient than dopant atoms in thermalequilibrium with the growing surface, hence they are more likely to beincorporated into the growing film.
This method of doping during growth was tried on diamond byJamison et al.[28) for sodium, rubidium and phosphorus, and proved tobe efficient for embedding the impurities into the growing diamondlayer. Secondary ion mass spectroscopy measurements showed thatphosphorus entered up to 5xlO 18 cm-3, while sodium and rubidium
entered up to 1020 cm-3. Highly conductive p-type layers were obtainedby sodium doping. Energies of activation of 44 and 82 meV wasmeasured for sodium.
The experiments on the influence of deformations on the diamondconductivity might be an explanation of p-type conductivity in diamondcaused by big atoms. The investigation of electrical properties ofdeformed natural diamond [45-47) has shown that dislocations formedby deformations change the electrical, optical and photoelectricalproperties of diamond. An insulating natural diamond of IIa type withinitial electrical resistivity of 1016 n em, after deformation resembledthe semiconducting diamond lib. The resistivity varied between 1013
10
and 102 n em as function of the plastic deformation. If the number ofdislocations is large (-109-1010 cm-2), the electrical resistivity canbecome as low as 102 n em. This change of resistivity was obtainedwithout introducing any impurities in the diamond lattice. The activationenergy of electrically active centers produced by deformation was 0.260.29 eV. The conductivity was of p-type. The results of this experimentsuggest that the introduction in the diamond lattice of impurities withlarge atomic radii can produce not only n-type centers, due to theimpurity itself, but also compensating p-type centers due to latticedeformations. As sodium is a too large atom (atomic radius r = 1.66
A, ionic radius rI+= 0.95 A) [5], the structural defects of the lattice due
to introduction of Na could be too large and might dominate thetransport properties.
6. Conclusions
N-type diamond doping by the would be donor impurities N, P, andNa have been unsuccessful. There might be different explanations forthis behavior. The energy structure of the band gap of diamond filmsgrown in different conditions is not known yet. The existence ofconsiderable densities of acceptor-like states close to the valence-bandedge and extending 1-2 eV into the band gap might be responsible forthe compensation of donors. The nature of these states is not known.The states might be formed due to impurities,and/or structural defects.Another hypothesis is that high densities of structural defects introducedin the lattice by large impurity atoms, might form relatively shallowacceptor level (about 0.3 eV) and might dominate the transportproperties, giving p-type conductivity. No exhaustive experiments havebeen done yet to verify these hypotheses.
The methods of forcing the introduction of impurities into thediamond lattice have an important advantage over the ion implantation,widely used for diamond doping. The ion implantation introducesstructural defects (vacancies, vacancy+interstitial, and theircombinations) that is difficult to cure. No additional structural defectsare introduced by forced diffusion or ion assisted doping during growth.Forcing methods of doping are effective for introduction of large amountof dopants into the diamond lattice.
N-type conductivity in high quality diamond polycrystalline filmswas obtained by diffusion of impurities under bias from Li salts. Themeasurements showed the electron concentration about lOIS cm-3 andsheet resistance about 105 Ohm/square with the estimated electron
11
mobility of - 50 cm2Ns. The possibility of n-type diamond doping wasdemonstrated by this experiment.
References
1. P. V. Pavlov, with John, E. I. Zorin, D. I. Tetelbam and A. F.Khokhlov, (1976) Phys. Stat. Sol.(a) 35 11
2. Y. Tokumaru, H. Okushi, T. Masui, and T. Abe (1982), lap. l.Appl. Phys, 21 L443
3. K. Nauka, M. S. Gorsky, H. C. Gatos, and J. Lagowski, (1985) Appl.Phys. Lett. 47 341
4. S. M. Sze, Physics of Semiconductor Devices (1981), John Willeyand Sons, NY, p. 21
5. Handbook of Chemistry and Physics, 1989-1990, CRC Press BocaRaton, FL, p. FI88-189.
6. B. V. Spitsyn, G. Popovici, and M. A. Prelas, (1993), SecondInternational Conference on the Applications of Diamond Films andRelated Materials, August 25-27, , Editors: M. Yoshikawa, M.Murakawa, Y.Tzeng and W. A. Yarbrough, MYU, Tokio, Japan p.57-64 ;
7. J. E. Field, with John (1992) The Properties of Natural and SyntheticDiamond, ed. J. E. Field, Acad. Press, London, p.865
8. V. S. Vavilov, A. A. Gippius, E. A. Konorova, (1985) Electronic andOptical Processes in Diamond (in Russian), Nauka, Moscow, p.27
9. T. Evans, The Properties of Natural and Synthetic Diamond, (1992)ed. J. E. Field, Acad. Press, London, p.259
10. P. R. Briddon, M. I. Heggie and R. Jones, (1991) New DiamondScience and Technolo~y,MRS Int. Conf. Proc., editors R. Messier, J.T. Glass, J. E. Butler, R. Roy, p. 63.
II. J. Bernholc, S. A. Kajihara, and A. Antonelli, (1991) New DiamondScience and Technolo~y, MRS Int. Conf. Proc., editors R. Messier,J. T. Glass, J. E. Butler, R. Roy, p. 923
12. S. A. Kajihara, A. Antonelli, J. Bernholc, and R. Carr, (1991)Phys. Rev. Lett.., 662010
13. W. J. P. Enkevort and van, E. H. Versteegen, (1992) l. Phys. :Condens. Matter 4 2361
14. R. G. Farrer, Solid State Comm. (1969) 7 68515. R. J. Keddy, T. L. Nam, and R. C. Burns, (1988) Carbon 2634516. M. Seal and W. J. P. van Enkevort, (1988) Diamond Optics (Proc.SPIE 969, Washington, DC, p.144
17. J. Mort, M. Machonkin, and K. Okumura, (1991) Appl. Phys. Lett.59 3148
18. K. Jackson, M. R. Pederson, J. G. Harrison, (1990) Phys. Rev., 41B12641
19. N. Setaka, (1989) Technolo~y Update on Diamond, Extended
12
Abstracts (EA-19), ed. P. P. H. Chang, D. Nelson and A. Hiraki,Materials Research Society
20. M. 1. Landstrass, M. 1., M. A. Plano, D. Moyer, S. P. Smith, and R.G. Wilson, (1991) Diamond Materials, Electrochem. Soc., ed. A. J.Purdes, J. C. Angus, R. F. Davis, B. M. Meyerson, K. E. Spear, andM. Yoder, p.574-579
21. M. Kamo, H. Yarimoto, T. Ando, and Y. Sato, (1991) NewDiamond Science and Technology, MRS Int. Conf.· Proc., editors R.Messier, 1. T. Glass, J. E. Butler, R. Roy, p. 637-641
22. K. Okano, H. Kiyota, T. Iwasaki, T. Kurosu, M. !ida and T.Nakamura, (1991) New Diamond Science and Technology, MRSInt. Conf. Proc.• editors R. Messier. J. T. Glass. J. E. Butler. R. Roy,p.917-922
23. R. M. Chrenko, Phys. Rev. (1973) B74 56024. V. V. Tokiy. N. D. Samsonenko, D. L. Savina, and S. V. Gorban.(1993) Second International Conference on the Applications ofDiamond Films and Related Materials. August 25-27, Editors: M.Yoshikawa. M. Murakawa, Y.Tzeng and W. A. Yarbrough. MYU.Tokio, Japan p. 757-760
25. K. Weiser Phys. Rev. (1962) 126, 142726. G. G. Fountain, R. A. Rudder, D. P. Malta, S. V. Hattangady, R. G.Alley, G. C. Hudson, J. B. Posthill. R. J. Markunas, T. P.Humphreys, R. J. Nemanich. V. Venkatesan and K. Das. (1991)Diamond Materials, Proc. Second Symp. Electochem. Soc. ed. A. J.Purdes, J. C. Angus, R. F. Davis. B. M. Meyerson. K. E. Spear, andM. Yoder. p. 523
27. K. Okumura, J. Mort. and M. Machonkin, (1990) Appl. Phys. Lett.57 1907
28. K. D. Jamison. H. K. Schmidt, D. Eisenmann. and R. P. Hellmer,(1993) MRS Symposium Proc. v.302, MRS. p.251-256
29. J. F. Prins. (1992) Materials Science Reports, 7 27130. A. A. Gippius. Electrically and optically active impurities. thisvolum
31. G. Bums, (1985) Solid State Physics, Academic Press. N. Y., p.354
32. J. Robertson, (1991) Diamond and Diamond-like Films andCoatings. edit. by R. E.Clausing. L. 1. Horton. J. C. Angus, and P.Koidl. Plenum Press NY. p.37-46
33. W. Tsai, M. Delfino. L.-Y. Ching, G. Reynolds, D. Hodul and C. B.Cooper III. (1991) New Diamond Film and Technology, MRS Int.Conf. Proc.• p. 937-941
34. K. A.. Arseni. B. 1. Boltaks, (1969) Soviet Phys. - Solid State 102190
35. Boltaks, B. 1. (1963) Diffusion in semiconductors, Acad. Press,
13
NY, p.222-225.36. G. Popovici, T. Sung, M. A. Prelas, S. Khasawinah, and R. G.Wilson, (1994) MRS Meeting, April 4-8 San Francisco, in press.
37. G. Popovici, M. A. Prelas, T. Sung, S. Khasawinah, A. A.Melnikov, V. S. Varichenko, A. zaitsev, and W. R. Fahrner (1994)to be presented in International Conference, Diamond Films'94,Italy, September 25-29.
38. G. Popovici, T. Sung, M. A. Prelas, S. Khasawinah, and R. G.Wilson, to be published.
39. G. Popovici, T. Sung, S. Khasawinah, M. A. Prelas, GalinaSokolina, M.. G. Ermakov and A. S. Vedeneev, (1994) to bepublished
40. G. Popovici, T. Sung, S. Khasawinah, M. A. Prelas, to bepublished
41 .J. E Greene, S. A. Barnett, A. Rockett, and G. Bajor (1985) Appl.Surf. Sci. 22/23 520
42. A. L Fahrenbruch, A. L, A. Lopez-Otero, K. F. Chien, P. Sharps andR. H. Bube, (1987) Proc. 19th IEEE Photovoltaiv Spec. Conf. p.1309
43. P. Sharps, A. L. Fahrenbruch, A. Lopez-Otero, and R. H. Bube,(1988) Proc. 20th IEEE Photovoltaiv Spec. Conf. p. 1641
44. P. Sharps, A. L. Fahrenbruch, A. Lopez-Otero, M. Shofthaler, andR. H. Bube, (1989) Proc. 21th IEEE Photovoltaiv Spec. Conf. p.493
45. N. D. Samsonenko, V. V. Tokiy, and S. B. Gorban, (1991) Sov.Phys.-Solid State, 33 2496
46. N. D. Samsonenko, V. I. Timchenko, Iu. A. Litvin, and G. B. Bokiy,(1978) Soviet Physics - Doklady 242 826
47. N. D. Samsonenko, V. I. Timchenko, V. A. Emets, and G. B. Bokiy,(1980) Kristallographia (Russian) 25 1300
DIFFUSION OF BORON, HYDROGEN, OXYGEN ANDLITmUM IN SINGLE CRYSTALLINE ANDPOLYCRYSTALLINE DIAMOND. A NOVEL METHODFOR THE DETERMINATION OF THE STATE OF ANIMPURITY: FORCED DIFFUSION OF BORON IN IATYPE NATURAL DIAMOND.
GAUNA POPOVICI,o* T. SUNG,* M. A. PRELAS,*R. G. WILSON," AND S. KHASAWINAH*
°Rockford Diamond Technology, 501 S. Sixth Street,Champaign, Illinois 61820-5579*Nuclear Engineering Department, University of Missouri,Columbia, Missouri 65211"Hughes Research Laboratories, Malibu, California 90265
Abstract
Diffusion of boron, hydrogen, oxygen and lithium in singlecrystalline and polycrystalline diamond is reported. Diffusion under DCelectric field was used to enhance the diffusion rate. SIMS analyses wereused to determine the impurity concentration. A novel method isproposed for the determination of the state of an impurity (donor,acceptor or deep level) in a semiconductor lattice. To demonstrate themethod boron was diffused into Ia type natural diamond under a DCelectric field. The concentration and diffusion profiles of boron wereaffected by the applied field. Boron diffuses as a negative ion since it isan acceptor shallow enough to be partially ionized at the temperature ofdiffusion. The drift velocity of boron ions at the temperature ofdiffusion was also estimated. Diffusion under DC electric field oflithium, oxygen and fluorine in high quality diamond films was alsostudied.
15
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 15-29© 1995 Kluwer Academic Publishers.
16
1. Introduction
Diamond has a great potential as a semiconductor material.[l] It hasa large band gap(5.45 eV), high electron and hole mobility (1900 and
1600 cm2Ns, respectively), the highest breakdown field (107 V/cm)of
any known material, and the highest electrical resistivity ( > 1016 Ohmcm for pure material) [2]. The electrical resistivity may be changed by
boron doping in a wide range (from 1016 to 0.1 Ohm cm). [2] Diamondhas also other unique properties such as the highest hardness, the highestthermal conductivity, the highest nuclear stability and the highestresistance to chemical attacks. These properties are due to a high strengthof the covalent carbon bond and a small interatomic distance of thecarbon atoms in the diamond lattice.On the other hand, the strong covalent bond and small interatomic
distance are also responsible for the highest Debye temperature (2200 K)and high melting point of diamond (>3700 K) which makes itstechnology particularly difficult. Another technological difficulty ismetastability of the diamond lattice. The graphitization of diamond startsat 1800 K, if diamond is heated in an inert atmosphere. In the presence ofoxygen it starts at lower temperatures (-900 K). [2]All these difficulties have considerably hindered the development of
the technology of single crystalline growth and doping of diamond. Sofar only boron is known to form a relatively shallow acceptor level in thediamond band gap. The teoretical calculations and experimental resultson potential donor impurities in the diamond lattice, like phosphorus [38], lithium [3, 6, 9, 10], and sodium[3, 11] are contradictory. Sometheoretical and experimental papers reported that these impurities canform shallow donor levels [3, 4,12] , while other report that theseimpurities form deep levels [5-7], or even acceptor levels ( sodium, forinstance [11]).Diffusion and ion implantation are widely used in the electronic
industry for introduction of impurities in semiconductors to obtaindesired changes in electric and/or optical properties. Diffusion has theimportant advantage over ion implantation of not introducing additionalstructural defects, except for those inherent to the impurity itself.Additional annealing is therefore not required. The ion implantation ofimpurities in diamond has been studied intensively [13], but the resultshave not been encouraging, except for boron doping The usualtemperatures of annealing after ion implantation (1200-1400 0c) are nothigh enough to anneal out radiation defects. Higher temperatures cannotbe used, as diamond is a metastable phase of carbon and begins totransform into graphite at temperatures exceeding 1500 °C [2].In this paper, a study of the diffusion in natural diamond of lithium,
17
boron, oxygen, and hydrogen is reported. Results on the diffusion ofhydrogen in single crystal diamond under conditions specific to CVDdiamond growth are reported for the first time. We also propose a newmethod for the unambigous determination of the state of an impurity inthe semiconductor lattice based on the forced diffusion of impurities.Diffusion of boron under DC electric bias in a single crystal Ia typenatural diamond was chosen to demonstration the method. Diffusionunder bias of lithium, oxygen and hydrogen in polycrystalline diamondfilms is also reported.
2. Experimental
Depth profiles ofB, Li, 0, and H were measured using SIMS. Theprimary beam was 14.5 keV Cs and secondary negative ions weremeasured for H, 0 and N. Li was measured in a separate profile using a8 keY primary 02 beam and positive secondary ions. Quantification was
carried out by using implanted standards for all these elements incrystalline diamond.The SIMS data referring to the first few tens ofnanometers from the surface are within the equilibration distance andshould not be taken into consideration. The limit of detection for lithiumwas estimated at 1014 cm-3. The experimental details of diffusionprocess are described below.
3. Diffusion methods. results and discussions
3.1 DIFFUSION OF BORON, LITHIUM, OXYGEN, ANDHYDROGEN IN IIA TYPE NATURAL DIAMOND.
A IIa type transparent natural diamond crystal was used. The dopantsource of Li2C03 was of analytical purity (99%). The diffusion was
performed in nitrogen atmosphere. The diffusion temperature was of 860°C, the time was of one hour (further called the first step). The samplewas then placed in a hot filament CVD growth reactor and the diffusionwas performed in hydrogen atmosphere from a boron solid source placedon the surface of the sample for two hours (further called the secondstep). The condition of diffusion were those used routinely during CVDgrowth. The pressure was 30 Torr. The filament temperature was 2100°C, as measured by a Mikron two wave length optical pyrometer. Thegraphite support temperature was 860°C. The distance between thefilament and the graphite support was 4 mm. We do not know exactly thetemperature of the diffused diamond surface. The temperature might belarger than that of the graphite support temperature, as the thickness ofthe diamond crystal was - 3mm and the diffused surface of diamond wasat a distance of about one millimeter only from the hot filament. No
18
graphitization of the diamond crystal was observed.The results of SIMS analysis are shown in Figs. 1- 2. Li diffused
into the diamond crystal to -2xlO l6 cm-3 at the depth ofO.5llm (Fig 1).Outdiffusion of lithium during the boron diffusion process could beobserved at the surface. This is the first time that diamond doping bydiffusion of lithium was found to be feasible. This disagrees with theresults of ref. 14-14, where no diffusion of Li was observed on hightemperature treatment up to 1400 °C after ion implantation. Thedifference with our results might be due to the trapping of lithium atomson structural damage sites formed during ion implantation.
The solubilities of boron, hydrogen and oxygen, (Fig. 2, a b, c, d)were found to be much higher than that of lithium. The concentrations ofthese elements were of (l to 4)x1020 cm-3 at a depth of 0.5 Ilm, nearlyfour orders of magnitude larger than that for Li. Our data on thediffusion of oxygen, and boron are in a good qualitative agreement withprevious reports. 16, 17
0.80.60.4
Depth (J.1Itl)
0.21015 L..-,---,---,---'--'---,--,--,--,--,---,--,--,--,--,--",-",-",-~
0.0
,-...1017
1=~
5'.::2
~0g01016U
Fig. 1 SIMS prifile of lithium
19
1021
1018
""" '"D(Oxygen)=4.4xI0-13 cm2/sec '" '"D(Boron)=4.2xI0-13 cm2/sec '"D(Hydrogen)=2.4xlO-13 cm2/sec "
"'\
2.01.51.00.51017 L-L....-JL.-..J--l.--l.--,---,--,---J....---l...-,--,---'---'--...L--,---'---'--.L..-.J0.0
Depth (jlm)
Fig. 2 SIMS profiles of boron, hydrogen, and oxygen. Dashed curvesare the impurities profiles calculated using Eq. (3) Only boron profile canbe described by eq. (3). The diffusion mechanism of oxygen andhydrogen is more complex than that described by eq. (3).
The behavior of hydrogen in diamond is of great interest, andhydrogen can strongly influence the electrical and surface properties ofdiamond 18-22 as CVD films are grown in hydrogen atmosphere. Thehydrogen content in diamond films was reported to be between 10-1 23and 5 percent. 19 It is currently believed that hydrogen is mostlytrapped in diamond films during growth. The diffusion coefficient ofhydrogen was expected to be small. 18 To our knowledge, noexperiment has been done to establish the importance of the diffusion ofhydrogen in diamond under conditions specific to CVD diamond growth.We have found that the diffusion of hydrogen represents a quiteimportant mechanism of diamond doping during CVD growth, thatcannot be neglected.
20
Diffusion processes are described by Fick's laws [24]J = - D (aN/ax) (1)
aN/at = D(a~/ax2) (2)where J is the flux density of diffused atoms, D is the diffusion
coefficient, N is the concentration of diffused atoms. From theexperimental diffusion profiles one can calculate the diffusion coefficientD of the impurities by using the solution of the diffusion laws for a semiinfinite crystal with a constant surface concentration of diffusing atoms[25] :
N(x,t) =NO erfc [-x2/(4Dt)1I2] (3)where N(x,t) is the concentration of the diffusing atoms at the
penetration depth x and time t, NO is the surface concentration, and erfcis the error function complement.
This simple equation assumes that the lattice is perfect and thereis only one mechanism of diffusion. Defects in real crystals (vacancies,dislocations, vacancy complexes, microscopic voids, etc.) can stronglyinfluence the diffusion rate.
Sometimes diffusion is going through more than one channel. Forinstance, in germanium two simultaneously diffusing streams wereobserved for Ag, In, Zn and Te [24]. They followed distinct mechanisms,interstitial and vacancy, respectively. The diffusion rates can also beinfluenced by the presence of other impurities in the lattice, which caninfluence the position of Fermi level and/or act as traps formingcomplexes with diffusing impurities. Non-Fickian diffusion will beobserved in these cases. As an example, a diffusion law of the form exp(-x1L) instead of a complementary error function was observed fordiffusion with trapping. [24]The curves calculated using Eq. (3) are presented by the dashed lines
in Fig. 2. Only boron profile can be described by the Fick law. Thediffusion coefficients of boron is (4.2±O.3)xIO-13 cm2/s. Hydrogenand oxygen profiles have no Fickian shape. These impurities diffuse bytwo different mechanisms at least. Their calculated diffusioncoefficients describe only one of the mechanisms. The diffusioncoefficient of oxygen is 4.4xlO-13 cm2/s, and that of hydrogen is 2.4
xlO-13 cm2/s. This diffusion coefficients are large and most probably thediffusion of these impurities is enhanced by presence of defects in thediamond lattice.As the outdiffusion of lithium in the second step was important,
using the Eq. (3) in order to determine the lithium diffusion coefficient isa very rough approximation. The diffusion coefficient of lithium wasestimated to be more than one order of magnitude lower than thediffusion coefficients of boron, hydrogen and oxygen. The solubility oflithium in diamond was found to be low, as expected 3,4 - nearly four
21
orders of magnitude lower than that of oxygen, as the amount of lithiumand oxygen was nearly the same on the surface (salt Li2CQ:3). This canbe easily understood on account of the differences in atomic radii. Theatomic radius rA of lithium (1.51 A) is significantly larger than that ofcarbon, oxygen, boron, and hydrogen (see Table I). The movement oflithium through the diamond lattice should thus be more difficult.
Table I Geometry and energy parameters of single element-carbon bonds insubstitutional sites in diamond lattice
Element
CB
oH
Covalent E-Cradius bond energy(A) (kcallmol)
0.77 830.78 890.66 660.34 104.8
3.2 A NOVEL METHOD FOR THE DETERMINATION OF THESTATE OF AN IMPURITY
We propose a new method for the unambiguous determination of thestate of an impurity in the semiconductor lattice based on the forceddiffusion of impurities. Diffusion of boron under DC electric bias in asingle crystal Ia type natural diamond was chosen to demonstration themethod.
In a semiconductor lattice impurity atoms can be in a neutral state(usually a deep level), positively charged (ionized donors) or negativelycharged (ionized acceptors). If an impurity atom can be ionizedthermally at room temperature (at least partially), it is said that to form ashallow level in the semiconductor band gap [26]. An electrical fieldapplied to a semiconductor during diffusion can influence themovement of dopant atoms, provided these atoms are ionized. Attemperatures of practical interest for the technology of wide band gapsemiconductors only the diffusion of shallow acceptors and donors canbe influenced by electric fields. Conversely, if electric fields do notaffect diffusion, this means that the corresponding impurity can not beionized thermally.Diffusion of thermally ionized impurities under DC electric field has
22
been studied for Si and Ge [24]. It has been used mainly for the fastdiffusing impurities like Cu and Li that can be diffused at relatively lowtemperatures ( as low as 400°C). The voltage used was of order of a fewvolts. In diamond and other wide band gap semiconductors the electricalconductivity may remain low even at high temperatures. This allows theapplication of relatively high voltage (hundreds of volts) to increase therate of drift of ions in the lattice.To demonstrate a novel method for determining the state of an
impurity in the diamond lattice, the diffusion of boron under DC bias inIa type natural diamond was investigated. Mechanically polished parallelplates (labeled ND3.1 and ND3.3) cut from a crystal of natural Ia typediamond were used. The plates were mounted on a graphite base with anembedded tungsten heater.
0.8
ND3.1 Boron
ND3.1 Niitrogen
ND~'lPositive Bias
LhC03
ND3.3 've Bias
0.2 0.3 0.4 0.5 0.6 0.7
Depth (J.lrn)
ND3.3 Nitrogen
1.e+15 -1---.,r-----r---,...----r---,...-----r---.-----f
0.0 0.1
1.e+22
1.e+21
..-... 1.e+20§........,
§ 1.e+19.-~
~ 1.e+18(I)
~0
U1.e+17
1.e+16
Fig.3 SIMS profiles for boron (B) and nitrogen (N) of the diamondsamples ND3.1 and ND3.2 after diffusion under electric bias of 230 V at1040 K for 5.5 hours. Inset:Experimental arrangement for the diffusion
23
The samples were arranged as shown in Fig. 3 (inset). A lithium salt(Li2C03) was placed between the plates to assure an electric contact. A
DC voltage of 230 V was applied during the diffusion process with thenegative contact on the sample ND3.3 and positive contact on thesample ND3.1. The applied field should have accelerated the diffusionof positive ions into the sample ND3.3 and of negative ions into thesample ND3.1. In this arrangement the doped surfaces have nearly thesame temperature and the surface conditions. the only difference is thesign of the applied electric field. Therefore, the influence of electricfield can be easily seen.Diffusion of boron was from the gas phase, a small quantity of boron
having been placed on the graphite support near the samples. Thediffusion was performed in an argon atmosphere. The base temperaturewas monitored by a chromel-alumel thermocouple. The diffusiontemperature was 1040 K, the time was 5.5 hours.
The concentration of lithium was under the detection limit of 1014
cm-3 for both samples. A constant concentration of nitrogen of 4X1020
cm-3 was found in both samples (Fig. 4), as expected for Ia type naturaldiamond, usually containing nitrogen. No outdiffusion of nitrogen wasobserved, implying a low diffusion coefficient of nitrogen in diamond atthe diffusion temperature. The nitrogen profile does not depend on theapplied voltage, because nitrogen forms a deep level in the diamondband gap.[2] A small difference in the concentration might be due tothe inhomogenities in the nitrogen doping of the initial crystal.However, the applied field did affect the boron concentration (see
Fig. 4). In the positively biased sample boron concentration is larger upto the depth about 0.6 Ilm. As boron is a relatively shallow acceptor inthe diamond band gap ( ionization energy of boron level LlE=O.37 eV [2])the negatively charged boron ions were forced by the electric field intothe sample ND3.1 that was biased positively. The concentration profileof boron diffused into the sample ND3.1 has a greater dependence ondepth than for the samples with the reverse field direction.The method also permits estimating of the ion drift velocity in the
lattice under the DC electric bias. Diffusion in an external field can bedescribed by the modified Fick;s laws [24]J =JD + Jv =-D (aN/ax) + Nv (4)
aN/at =D (a2N/ax2) - v(aN/ax) (5)where J is the sum of the flux densities of the diffused atoms, JD is
the flux density of the freely diffused atoms determined by the sign and
magnitude of concentration gradient, Jv is is the flux density of atoms
diffused under the influence of electric field, D is the diffusion
24
coefficient, and v is the drift velocity imposed on the ions by the appliedfield, N is the concentration of the diffusing atoms.By solving Eqs. (4) and (5) for the diffusion from a constant source
into a semi-infinite body, one obtains:N_ = N+ exp(-vxID) (6)
where N+ is the concentration of impurities for a direction of the
applied field that pushes the ionized impurities into the sample. N_ is the
concentration for the opposite direction of the field. By taking N+ =
1018 cm-3, N- =1016 cm-3 at the depth x=O.l Ilm (Fig. 4), one finds
from Eq. (6) that vlD = 4.6x105 cm-3.As the concentration of boron becomes equal in both samples at the
depth d= 0.6 Ilm, the average velocity v of ions can be calculated, by
supposing d = vt, where t is the diffusion time. For t = 2><104 s (5.5
hours), one find v = dlt = 3xlO-9 cm/s. Evaluation of the diffusioncoefficient from equation (3) gives a reasonable value for the diffusion
coefficient D '" 6x 10-15 cm 2 /s .
3.3 DIFFUSION UNDER ELECTRICAL BIAS OF LITHIUM,FLUORINE AND OXYGEN IN CVD DIAMOND FILMS
Free standing diamond films, of Norton Co, 230 Ilm thick polishedon both sides have been used in this work. The average crystallite sizewas of the order of tens of micrometers. All samples were grown underthe same conditions. Their Raman spectra had the diamond line only,with no lines of graphite and amorphous carbon. Thecathodoluminescence spectra presented a strong exciton line, which is afeature of good crystalline quality diamond films. [27] The films weremounted on a graphite base with an imbedded tungsten heater. The basetemperature was monitored by a chromel-alumel thermocouple. TheLi2C03 dopant source was of analytical purity (99%). The diffusion
was performed 190 min at 1000 0C in argon atmosphere. An electricfield of 200 V was applied. The negative terminal of the source wasapplied to the graphite base. The applied field should have pushed thepositive ions into sample B4 and the negative ions into sample B5. Thecurrent during diffusion was 0.5 rnA. The control sample B6 had noelectric field applied.The results of SIMS analysis for samples B6 are shown in Fig. 4.
The results are shown only for sample B6, as there was no difference inthe doping profiles for all B samples ( except for fluorine, see Fig. 5).Fluorine was present as an impurity in the diffusion source, as its purity
25
was of only 99%. Oxygen, hydrogen, lithium, and fluorine were foundto be present in the diamond lattice. Oxygen and hydrogen
concentrations of approximately 3x1019 cm-3 were found. We did notcheck the presence of hydrogen in the sample before the diffusion;however, the presence of hydrogen of the order of 0.1 at% in CVDdiamond films was reported earlier. [23] We should also not exclude thepossibility of hydrogen diffusion from graphite, as earlier experimentshave been done in hydrogen atmosphere.
2.0
Oxygen
1.0
Depth (l1JlI)
0.51019 lL..---'-.........--'-....L............."'---'-"~--'- ........-'-..l.-.-.l-....---o.--'---'-...J0.0
Fig. 4 SIMS analyses of the concentration depth of the impurities.No difference in the depth profile of Li, 0 and H was observedfor the samples B4, B5, and B6.
The diffusion of lithium and oxygen into single crystalline diamondsample was performed in approximately the same conditions as for thesample B6. The results of SIMS analysis showed that Li entered5xlOl7 cm-3 at 0.1 J..l1ll and 2x1016 cm-3 at 0.5 J.l.m. Tree orders ofmagnitude difference for polycrystalline sample (_3x1019 cm-3 ) isprobably due to the faster diffusion on the grain boundaries, ascrystallites forming film were of high quality, as determined by Ramanand cathodoluminescence measurements. The oxygen entered into single
26
crystalline IIA type natural diamond (see 3.1, diffusion in the singlecrystal diamond) approximately in the same quantity as inpolycrystalline film, implying that oxygen moves through the diamondlattice as quickly as through the grain boundaries. The last result can beunderstood taking into account that atomic rA radii of the lithium (rA =1.51 A) is larger than that of oxygen, the last is smaller than atomicradius of carbon (see Table I).
1co'P 1018
~goo
_""---_B6
2.01.51.0
Depth (J.Ull)
0.51017 L..-J'---'........--'--L..................""'"----'--.l..-........--o.........--L. .................-.......J
0.0
Fig. 5 Depth profiles of fluorine in samples B4, B5, B6
As we discussed before the theory predicts a shallow level in thediamond band gap for the interstitial lithium atoms. Lithium complexeswith other lattice defects formed on the grain boundaries are notexpected to give also shallow levels. As diffusion dominates on thegrain boundaries and other defects inherent to the polycrystalline films ,this can explain the independence of lithium diffusion on the appliedfield.Only the fluorine concentration was found to be possibly influenced
by the electric field applied during diffusion (Fig.6). The quantity offluorine that entered the lattice is smaller for the negatively biased andlarger for positively biased sample as compared with the unbiased one.Fluorine appears to diffuse at least partially as negative ions. In the
unbiased B6 sample, the fluorine concentration was -2.0 x 1017 cm-3 (at
27
biased B4 sample the concentration is -1.5XI017 cm-3. In order to bepartially ionized at 1300 K, fluorine should have in the diamond bandgap a shallow level with energy of the order of 2kT "" 0.22 eV. Asatomic radius of fluorine is small (rA = 0.7 A) [26] we can expect nottoo different diffusion rate through the crystallites as through the grainboundaries.
Conclusions.
We have presented results on diffusion of boron, hydrogen, oxygenand lithium in type ITa natural diamond at 860°C. After diffusion, the
concentration of Li was of the order of 2XIOI6 cm-1 at a depth of 0.5/-lm, while those of oxygen, nitrogen and boron were found to be of the
order of 1020 cm-3 at the same depth. The concentration of hydrogenwas about 1018 cm-3 at a depth of 0.5 /-lm. The diffusion coefficients of
boron was estimated to be (4.2±O.3»xlO-13 cm2/s, while the diffusioncoefficient of lithium was rougWy estimated to be about one order ofmagnitude lower.
A novel method for determination of the state of impuntles in asemiconductor lattice (donor, acceptor or deep level) was proposed inthis paper. Diffusion of boron in natural type Ia diamond wasperformed under an electric field to check the method. Boron candiffuse through the diamond lattice as negative ion, as it is a relativelyshallow acceptor in the diamond band gap. The boron concentrationsand diffusion profile were affected by the applied field.
The diffusion of lithium, oxygen and fluorine under bias in CVDdiamond films was performed at 1000 0c. Li2C03 was used as adiffusion source. After diffusion, the concentrations of Li and ° in thediamond films were found to be of the order of (3-4)x1019 cm-3. Nodependence of the impurity concentration of lithium on the applied biaswas observed, as diffusion of lithium atoms was dominated by diffusionthrough the grain boundaries. Fluorine diffusion had possibly beeninfluenced by the electric field. Its concentration dependence on theelectric field indicates that fluorine may have formed a shallow level inthe diamond band gap.
28
References
1. M. N. Yoder Applications of Diamond films and related Materials,edit. Y. Tseng, M. Yoshikawa, M. murakawa, A. Feldman, ElsevierScience Publishers, Ny 1991, p.287-293
2. J. E. Field, Tables of Properties, in The properties of Natural andSynthetic Diamond, edit. Field, Acad. Press, London, 1992, p. 667699
3. Bernholc, J., S. A. Kajihara, and A. Antonelli, (1991) New DiamondScience and Technology, MRS Int. Com. Proc., editors R. Messier, J.T. Glass, J. E. Butler, R. Roy, p. 923
4. Kajihara,S. A., and A. Antonelli, J. Bernholc, and R. Carr, (1991)Phys. Rev. Lett.., 662010
5. Setaka, N. (1989) Technology Update on Diamond. ExtendedAbstracts (EA-19), ed. P. P. H. Chang, D. Nelson and A. Hiraki,Materials Research Society
6. Landstrass, M. I., M. A. Plano, D. Moyer, S. P. Smith, and R. G.Wilson, (1991) Diamond Materials, Electrochem. Soc., ed. A. J.Purdes, J. C. Angus, R. F. Davis, B. M. Meyerson, K. E. Spear, andM. Yoder, p.574-579
7. Kamo, M., H. Yarimoto, T. Ando, and Y. Sato, (1991) New DiamondScience and Technology, MRS Int. Com. Proc., editors R. Messier, J.T. Glass, J. E. Butler, R. Roy, p. 637-641
8. Jackson, K., M. R. Pederson, J. G. Harrison, (1990) Phys. Rev., 41B12641
9. Fountain, G. G., R. A. Rudder, D. P. Malta, S. V. Hattangady, R. G.Alley, G. C. Hudson, J. B. Posthill, R. J. Markunas, T. P.Humphreys, R. J. Nemanich, V. Venkatesan and K. Das, (1991)Diamond Materials, Proc. Second Symp. Electochem. Soc. ed. A. J.Purdes, J. C. Angus, R. F. Davis, B. M. Meyerson, K. E. Spear, andM. Yoder, p. 523
10. Okumura, K., J. Mort, and M. Machonkin, (1990) Appl. Phys. Lett.57 1907
11. Jamison, K. D., H. K. Schmidt, D. Eisenmann, and R. P. Hellmer,(1993) MRS Symposium Proc. v.302, MRS, p.251-256
12. K. Okano, K., H. Kiyota, T. Iwasaki, T. Kurosu, M. !ida and T.Nakamura, (1991) New Diamond Science and Technology, MRSInt. Com. Proc., editors R. Messier, J. T. Glass, J. E. Butler, R. Roy,p.917-922
13. J. F. Prins, Materials Science Reports, 7(1992) 27114. C. Cytermann, R. Brener, R. Kalish, Diamond and Related Materials,3,667 (1994)
15. E. A. Konorova, V. F. Sergienko, S. D. Tkachenko, A. A.Shulzhenko, A. G. Gontar, and A. A. Bocheka, J. Superhard Mater.
29
6, 1 (1984)16. Y. Mori, N. Eimori, H. Kozuka, Y. Tokota, J. Moon, J. S. Ma, T. Ito,A. Hiraki, Appl. Phys. Lett., 60, 47 (1992)
17. D. Narducci, and J. J. Cuomo, J. Appl. Phys 68,1184 (1990)18. G. Davies, Properties and Growth of Diamond, edit G. Davies,lSPEC, lEE, London, 1994 p. 127-140
18. D. Narducci, and J. J. Cuomo, J. Appl. Phys 68,1184 (1990)19. F. G. Celli, A. J. Purdes, B. E. Gnade, and D. L. Weathers, NewDiamond Science and Technolol:Y, 1991 MRS lnt. Conf. Proc.,editors R. Messier, J. T. Glass, J. E. Butler, R. Roy, p. 631
20. S. Albin and L.Watkins, IEEE Electron Device Lett., 11, 159 (1990)21. M. 1. Landstrass and K. V. Ravi, Appl. Phys. Lett., 55,1391 (1989)22. T. Sung, S. Khasawinah, G. Popovici, M. A. Prelas, B. V. Spitsyn,G. Mannig, S. Loyalka, R. Tompson, Diamond, SiC and NitrideWide-bandgap Semiconductors, MRS Proc. v. 339, San Francisco,1994 p.
23. S. Khasawinah, T. Sung, B. Spitsyn, W. H. Miller, G. Popovici, M.A. Prelas, E. J. Charlson, E. M. Charlson, J. M. Meese, T. Stacy, G.Mennig, S. Loyalka , B Tompson, J. Chamberlain, and H. White,Diamond Materials, ed. J. P. Dismukes and K. V. Ravi,Electrochemical Society Proc. v.93-17, 1993, p. 1032-1035
24. B. 1. Boltaks, Diffusion in semiconductors, edit. H. J. Goldsmid,Academic Press, NY,1963 p.93-128; S. M. Hu, Diffusion in Siliconand Germanium, in Atomic Piffusion in Semiconductors, edit D.Shaw, Plenum Press, NY 1973, p. 217-350
25. Handbook of Chemistry and Physics, CRC Press, 1989-1990, p. F189 to F-213
26. S. M. Sze, Physics of Semiconductor devices, John Wi1ey&Sons,NY, 1981 p.7-61
27. G. Popovici, M. A. Prelas, S. Khasavinah, T. Sung, A. A. Melnikov,V. S. Varichenko, A. M. Zaitsev, and W. A. Fahrner, to be publishedin Diamond and Related Materials
CHEMICAL ASPECTS OF DIAMOND DOPING
B.V. SPITSYN
Institute ofPhysical Chemistry, Russian Academy ofSciences,
31 Leninsky Prospekt, Moscow 117915, Russia
Abstract
Doping of diamond is an interesting problem in basic and
experimental science, and unless real progress in its solution is attained,
advancement in the development of many new kinds of techniques and
the spread and proliferation of high technologies will be critically
restricted. In this paper, the prognostic value of chemical and
physicochemical approaches to diamond doping has been discussed.
Consideration is mostly made on diamond crystallization from an
activated vapor phase as well as its importance in other diamond doping
techniques. Data on the constitution and energetics of prototype
molecules containing dopant atoms, chemically bonded with 2, 3 or 4
carbon atoms are considered from the point of view of the probability for
noncarbon atoms to be accommodated in the diamond lattice. The
approach may be helpful for consideration of equilibrium diamond.doping
by boron, phosphorus, or sulphur. The distribution. coefficients between
diamond film and gaseous crystallization media for boron or phosphorus
have· been evaluated and their reasonable agreement with experimental
data has been demonstrated. Some obstacles, such as compensative
action of dislocation related acceptors and/or bonded hydrogen, which
must be eliminated for effective n-type diamond synthesis, have also been
discussed.
31
M.A Prelas et al. (eds.), Wide Band Gap Electronic Materials, 31-45© 1995 Kluwer Academic Publishers.
32
1. Introduction
Growth of pure (non-intentionally doped) diamond film still
provides interesting and valuable opportunities for new applications of
diamond in science and industry. These opportunities could be extended
significantly by intentionally doping of the diamond lattice with impurities
with the aim of changing the electrical, optical and chemical ,properties
ofdiamond for use as a future electronic and engineering material[l].
2. Goals and benefits of diamond doping
Actually, with the progress in the understanding of and
experience with diamond and, especially, diamond film doping, the
following opportunities have arisen:
- study of kinetics of homogeneous and heterogeneous reactions on
growing diamond surface,
establishment of crystal growth mechanism (layer-by-Iayer, normal,
etc.),
- change of the optimal range of parameters (e.g., crystallization
temperature) for diamond and diamond film growth,
- achievement of equilibrium or non-equilibrium doping,
- use ofnew chemical reactions for diamond CVD,
- control of dopant concentration, and chemical and electronic states
immediately before the supply ofdopant to the growing diamond surface,
-effect of intrinsic defect concentration,
growth of ultrathin (sub-micrometer thickness) uniformly doped
diamond films,
- deposition of diamond films with desirable spatial dopant distribution
in the course ofsynthesis of designed multilayer diamond structures.
33
3. Basic requirements of doping process
The problem of controlled doping with impurity atoms of a
known concentration in distinct crystallographic positions and at the same
time in a specific electronic (valence) state is one of the most difficult in
modem solid material synthesis. The solution to this problem depends
particularly on equilibrium impurity solubility which depends, in its tum,
on such variables as pressure and temperature of the process and impurity
concentration in the feeding crystallization medium.
Impurities are a kind of defect in an ideal crystals and by their
doping, interaction with non-impurity defects (vacancies, interstitials,.
dislocations, stacking faults or grain boundaries) can take place. It is
possible in principle to consider it by this interaction as extra intrinsic
defect generation and from another aspects as a promising method for the./
neutralization of intrinsic (especially low-dimensional) defects by impurity
gettering. Hence the concentration of some kind of intrinsic defect may
either be increased or decreased in the doping process. This phenomenon
significantly influences the mechanism and the ultimate result of doping
itself. For instance, the artificial generation of a dislocation network in
diamond crystal leads to the emergence of sub-bands in the forbidden
energy gap of diamond. The sub-band is situated 0.2-0.3 eV higher than
the top of the valence band induces intrinsic (non-impurity) p-type
conductivity ofdiamond on the order of 10-2 (Ohm.cm)-l [2].
Such acceptor centers generated in the course of growth or by
postgrowth treatment may compensate donor centers produced by
supplying a source ofelectron-type conductivity to the diamond crystal.
4. Doping techniques
Methods of diamond doping can be subdivided (Table 1) into in-
34
TABLE 1. Basic characteristics of diamond doping techniques(efficiency: ++ high, + middle, 0 low, - no data)
Doping related characteristics
hnpurity Technique Defect Average Spacial Crystallo- . Control ofsupply generation concentration concentration graphic position impurity
control control ofimouritv valence state
In the HPHT + + + + +
growth DiamondCVD 0 ++ ++ + +
b-implantation + ++ ++ + -Post- Diffusion + 0 - + -growth Forced diffusion - + + - -
doping Ion implantati~n ++ + + + -Transmutation ++ + + 0 -dopin~·
• combination of in-growth and post-growth techniques
growth doping and those that exploit impurity doping in already grown
diamond crystal or film. The known techniques include the recently
proposed transmutation doping [3] and forced diffusion [4] indicated in
Table 1. Conventional high-pressure high-temperature diamond synthesis
(HPHT) proceeds under conditions close to those of equilibrium, but
supplying dopant to crystallization zone is not as easy as in diamond CVD.
In the classification of doping techniques presented in Table 1, it
is of importance to note that the diffusion coefficient of most impurities
in diamond is extremely low. If the temperature in the doping process is
substantially different from that adopted in studies or in some
applications of doped diamond, no impurity redistribution takes place.
35
Therefore, under practical conditions solid solution of impunties should
be in non-equilibrium as a rule. However this does not lead to
simultaneous substantial change of spatial distribution of the impurities
because of the extremely low diffusion coefficient value.
One of the important prevalences of ion implantation [5] is
forced introduction of an impurity element at arbitrary concentration.
However the middle ion penetration length for ions with energy equal to
100 keY is of the order of 0.1 J.U1l. At this characteristic length, local
impurity concentration changes several-fold. Moreover, decelerating
Ions generate, in implanted crystal, defects such as vacancies and
interstitials, and after some critical dose, an amorphization of the crystal
takes place. This shortcoming of conventional ion implantation can be
eliminated to a significant degree in "delta-doping"[6,7] of diamond films
which has been beneficially applied to growing thin films of gallium
arsenide and other A3-B5 semiconductors. Such techniques as
transmutation doping[3] are hybrid ones which provide opportunities for
both in-growth and post-growth change ofdiamond chemical composition.
Referring once more to Table 1, one can state that among the
different doping techniques, the method of supplying impurity under
diamond CVD conditions has the most potential, because for low pressure
diamond film growth, activated CVD using this technique provides
additional opportunities for effective control of impurity concentration
and also its valence state.
5. Equilibrium doping
For precalculation of the equilibrium doping of diamond, one
must have data on thermodynamic functions of solid solutions of impurity
elements under consideration. Since, as a rule, such data are not available,
36
it is reasonable to estimate the formation energy of related solid solutions
based on known bond energies[8].
To estimate the possibility of introducing a light element as the
substitution impurity, the use of the value of the single-bond length
between a carbon atom and the light-element atom, as well as the
value of the corresponding bond energy, has been proposed[8]. The
corresponding values and their relative difference in comparison with
geometric and energetic parameters of carbon atoms in a diamond lattice
are listed in Table 2.
In using Table 2, one must keep in mind that the accumulated
data have been obtained· on the basis of geometric and energetic
parameters of organoelemental-molecules such as B(CH3h, N(CH3h,
P(CH3h and S(CH3h with ordinary element-carbon sigma bonds.
However, by adopting a light-element atom in the substitution position
of the diamond lattice, two serious peculiarities arise. First the
coordination number of impurity atom, according to its situation in the
substitutional position, should be equal to four. Second, as a
consequence of the first peculiarity, the atom may exhibit a donor
(nitrogen, phosphorus, sulfur) or acceptor (boron) behavior relation with
the surrounding diamond lattice.
6. Prototype molecules
Because of non-isovalent substitution of a carbon atom by a non
carbon one (Table 2) in the diamond lattice the non-carbon atom becomes
"carbonlike" and is able to accept, from the diamond lattice as a whole, an
electron (boron), or to donate excess electrons (phosphorus,sulphur) to it.
The in-the-Iattice chemical state of bonded impurity atoms could be
characterized from the substantial change of geometric (tetrahedral radii)
37
TABLE 2. Geometric and energetic parameters of single elementcarbon bonds
Covalent Relative radii Element~ E-C to C-CElement radius difference carbon bond bond energy
energy difference
(A) RE-Rcx 100 (kcaVmol) (kcaVmol)
Rc(%)
C 0.77 0 83 0
B 0.85 +10.4 89 +6
N 0.70 -9.1 73 -10
P 1.10 +43 62 -21
S 1.04 +35 63 -20
and energetic (moderate energy of carbon atom to that of impurity atom
bond) parameters.
From the organometallic chemistry viewpoint, one can propose
the following series of prototype compounds which, with increasing
complexity, will more closely approximate the corresponding state of the
impurity atom in the diamond lattice.
Because the set of elements favorable for doping of diamond in
an atomically dispersive form is restricted to a few light elements [8], we
will consider only the prototype molecule series for boron, phosphorus
and sulphur.
For boron, B(CH3h, B(C2H5h, and Na[B(CH3)4] (sodium
tetra-methyl borate), alpha- and beta-boronadamantane, their heavy
analogies have a boron "atom" which should, as in alkaline tetramethyl
borates, be in the tetravalent and negatively charged (electron-deficient
bonding) state. Such a state of boron atom in the diamond lattice of p-
38
type semiconductor diamond can be classified as "intrinsic salt".
For phosphorus, P(CH3h, P(C2HSh, [P(CH3)4]I (tetramethyl
phosphonium iodide), and other phosphorusorganics with tetra
coordinated phosphorus should be considered.
For sulphur at the present time, it is possible to investigate only
rather remote prototypes:linear S(CH3h and S(C2HSh, and heterocyclic
S=(CH2)S, where sulphur is two-coordinated. Closer molecular analogues
for the substitutional sulphur position in the diamond lattice must be sought
among tetra-coordinated sulphurorganics with four single sulphur-carbon
bonds to one sulphur atom, if such compounds indeed exist. In line with
the basic theories of coordinative compound chemistry, one may assume
that as energetic, and also geometric parameters of boron, phosphorus
and sulphur introduction to the diamond lattice should be more favored
than can be supposed from consideration ofonly Table 2 data.
7. Nonequilibrium doping
The doping of impurity into crystal in the form of a non
equilibrium solid solution is possible if one or more stages of subsequent
homogeneous or heterogeneous reactions with participation of impurity
atom are in nonequilibrium. This clearly occurred during boron doping of
epitaxial diamond films grown on the (111) face of natural diamond in the
course of the chemical transport reaction at crystallization temperature
below 8000 C[3]. At higher temperatures, the doping was in equilibrium.
However, postgrowth cooling of this film, as was mentioned· earlier,
should transform it to a nonequilibrium one.
8. Equilibrium doping
Restricting now our consideration to equilibrium doping, we
600
_1400E~ 1200-E 1000;;::
'U 800EEco'Uc:
.. Nishimura etal.[9]• Sato et al.[1 0]• Flemish et al.[11]
500 1000 1500EtC in gas phase (ppm)
39
Fig 1. Relationship between content ofdoping element in diamond film and invapor phase for boron [9] and phosphorus [10,11].
According to theoretical [14] and experimental [8] data, the
phosphorus-related level of energy in the band gap of diamond must lJe
0.1 to 0.2 eV below the bottom of the conductivity band. This is the
reason why phosphorus remains as one of the promising impurities in the
research and development of effective n-type diamond semiconductor
with a shallow position of the donor level.
The up-to-date information indicates that donor centers in
diamond can compensate for intrinsic defects such as the dislocation
acceptor sub-band [2]. M.Kamo, YSato et al. in ref.10 proposed
possible compensation action in phosphorus-doped diamond films by
bonded hydrogen [10].
Only strictly directed and simultaneously multidisciplinary
research can provide the answer to the vital question of how to obtain an
effectively doped diamond semiconductor with donor conductivity.
40
have no choice but to use bond energy data in Table 2 for the calc;ulation
of equilibrium distribution coefficients between the vapor phase and the
grown diamond film (OF).
The qualitative ratioKE = (E / C)DF / (E / C)vapor phase
should be dependent on'the C-E to C-C bond energy difference, and on
temperature.
Thermochemical calculation at 1200 K gives distribution
coefficient values. For Band P, they are equal to
KB .. exp (6000 / 1.98.1200) = 12.49and
Kp .. exp (- 21000 / 1.98. 1200) = 1.45 x 10-4 .The above results are in reasonable agreement with the
experimental ones [8-11]. According to data from experiments on doping
performed with different CVD reactors and with different dopant
precursors [8 -12] the doping level for a constant set of experimental
parameters is usually directly proportional to the dopant precursor
concentration in the vapor phase (Fig. I).
Data in Fig. I were also in reasonable qualitative agreement with
the distribution coefficient values for boron and phosphorus calculated
above.
A very low experimental value of the distribution coefficient (on
the order of 10-3 to 10-4) was obtained for phosphorus doping of
epitaxial diamond films grown by the chemical transport reaction [8] and
in a microwave reactor[1l]. These data are in quantitative agreement
with the above-calculated distribution coefficient for phosphorus.
As demonstrated in refs.8 and 10, phosphorus concentration in
diamond films anddiamondlike carbon film [13] may be fairly high.
41
9. Doping level and doping efficiency
The equilibrium, and hence the nonequilibrium, doping level IS
dependent on the kind of diamond CVD process, composition of vapor
phase [12] and intensity of vapor-phase activation and of chemical
constitution of dopant precursor, and it concentration in vapor phase (Fig.
1). The temperature of the growing diamond surface and its
crystallographic orientation [15] are also dominant parameters. These and
other parameters probably govern the trap and compensative defect
concentration responsible, for instance, for carrier concentration and
mobility.
TABLE 3. Some characteristics ofdiamond films doped during thechemical transport reaction
Dopant Specific Film Ea, Conductivity
Dopant concentration, resistivity, thickness, type
atom.cm-3 Ohm.cm urn eV
B 1.7x1019- 106 - 0.1-2 0.37- P4.5xl021 2xlO-3 0.05
P 5xl018- 105 - 101 0.05-1 0.11- n
5xlQ20 0.04
S -lx1020 105 - 103 0.05-1 0.17 n*
*) according to high nonlinear resistance ofelectrical contacts to sulfur-dopeddiamond film.
,42
The method of CVD ofdiamond remains the simplest, and as far
as is known was applied, for the first time, to the thick diamond filmgrowth--chemical transport reaction (16-18]. Boron impurity in the
method was supplied in the form of its simplest carboranes. Phosphorus
and sulphur dopants were provided to the surface of the growing diamond
film in the form of pure vapor. The doped diamond, films were grown on
the (111) natural diamond face. Dopant concentration in diamond films
has been measured by spark mass spectrometry.
Basic results of the doping experiments are summarized in Table 3.
As has become known recently [19], n-type doping ofdiamond film
by sulphur has also been achieved in a microwave reactor and confirmed
from the sign of the Seebeck coefficient measured at high temperature.
10. Conclusions
The following conclusions were obtained:
- diamond doping offers many challenging goals and promising benefits
for basic science and new technology,
-the most flexible technique for controllable doping is activated CVD of
diamond with possible combinations with shallow [6,7] and conventional
[5,20] ion implantation and other techniques, to change the chemical
composition and properties ofdiamond materials,
- doping ofdiamond films by donors as acceptors is possible,
-the prototype molecule approach appears helpful for the prediction of
impurity atom accommodation by the diamond crystal lattice,
- numerical estimations of distribution coefficients between diamond film
containing boron and phosphorus, and relative crystallization media are in
reasonable agreement with experimental data,
-donor doping by phosphorus obviously is very sensitive to the presence
43
of intrinsic and impurity (e.g., bonded hydrogen) defects in diamond
material,
-the effective and comprehensive solution of problems related to diamond
doping requires interdisciplinary and international cooperation.
11. Acknowledgments
The author greatly appreciates A. E. Alexenko and I. V. Galkina
for participating in doping experiments. The help of N. Ohashi and M.
Fukutomi in the preparation of the manuscript is gratefully acknowledged.
12. References
1 Yoder,M.N. (1994) The vision of diamond as an engineered material,
in K.E.Spear and 1.P.Dismukes (eds), Synthetic Diamond: Emerging
CVD Science and Technology, John Wiley and Sons, Inc.,N.Y.,3-17.
2 Samsonenko,N.D. (1985) Electronic properties of diamond,Doctor
of Sciences thesis, Institute of Crystallography, RAS, Moscow(in
Russian).
3 Spitsyn, B.V.,G.Popovici,and M.A.Prelas(1993) Problems of
diamond film doping,in M.Yoshikawa, Y.Tzeng, M.Murakawa, and
W.Yarbrough(eds), Applications of Diamond Films and Related
Materials,Elsevier,57-64.
4 Popovici,G. and M.A.Prelas (1994) Forced methods of diamond
doping, in Abstracts of 2-nd International Symposium on Diamond
Films and NATO Workshop on Advanced Wide Bandgap Electronic
Materials (Minsk,Belarus,3-6 May 1994) 16.
5 Vavilov,V.S. (1975) Semiconducting diamond, Phys. Status Solidi, A
31,11-26.
44
6 Kobayashi,T., T.Ariki, M.Iwabuchi, T.Maki, S.Shikawa ,and
S.Suzuki(1993) Analitical study on multiple delta-doping in diamond
thin films for efficient hole excitation and conductivity enhancement,
I Appl. Phys. 76,1977-1979..
7 Jamison,K.D., R.P.Helmer and H.K.Schmidt (1994) Optical and
electrical properties ofdoped diamond films, in[4], p.35.
8 Spitsyn,B.V. and AE.Alexenko (1986) Physico-chemical basis for
doping of diamond from the gas phase, Arch. Nauki Mater.,7(2),201
205.
9 Nishimura,K., K.Das, IT.Glass, K.Kobashi, and R.J.Nemanich
(1990) Electrical properties of B doped CVD grown polycrystalline
diamond films, in R.Freer(ed.), The Physics and Chemistry of
Carbides, Nitrides and Borides, 183-194.
10 Kamo M.,H.Yurimoto, T.Ando, and Y.Sato (1990) SIMS analYsis of
epitaxially grown CVD diamond, in R.Messier, IT.Glass, IE.Butter,
R.Roy, eds., New Diamond Sci. Technol., Proc.2-nd Int. Coof,
Washington, DC, MRS, 637-641.
11 Flemish,IR., S.N.Schauer, R.Wittstruck, M.I.Landstrass, and
M.APlano(1994) Growth and characterization of phosphorus doped
diamond films, Diamond and Related Materials 3,672-676
12 Kawarada,H., H.Matsuyama, Y.Yokota, T.Sogi, AYamaguchi, and
AHiraki(1993) Exitonic recombination in undoped and boron-doped
chemical-vapor-deposited diamonds,Phys.Rev.B 47,3633-3637.
13 Veerasamy, V.S., G.AIAmaratungra, C.ADavis, AE.Timbs,
W.I.Milne and D.R.McKenzie(1993)N-type doping of highly
tetrahedral diamond-like amorphous carbon,J.Ph'ys., Condens.Matter
5, LI69-LI74.
14 Kajihara,S.A, AAntonelly, IBemholc,and R.Car(1991)Nitrogen and
potential n-type dopants in diamond,Phys.Rev.Lett.66,2010-2013.
45
15 Spitsyn,B.V. (1990) Diamond films: synthesis,properties and some
fields ofapplications, in Science and Technology ofNew Diamond, in
S.Saito, a.Fukunaga, and M.Yoshikawa (eds), KTK Scientific
Publishing Company,Tokyo,I-7.
16 Spitsyn,B.V. and A V.Smol'yaninov(1971) The technique of growing
diamond layers, USSR Author's Certificate No. 987912, appl.
1533329/23-26, filed April 21,1971.
17 Spitsyn,B.V. (1973) Chemical crystallization of diamond,Doctor of
Sciences thesis,Institute ofPhysical Chemistry, RAS, Moscow.
18 Spitsyn,B.V. (1991)The state of the art in studies of diamond
synthesis from the gaseous phase and some unsolved problems,in
Applications of Diamond Films and Related Materials, Y.Tzeng,
M.Yoshikawa, M.Murakawa, and AFeldman(eds),Materials Science
Monographs, 73, Elsevier, Amsterdam, 475-482.
19 Fujimori,N.(1994,March) Personal communication.
20 Wilson,R.G., M.I.Landstrass,and S.P.Smith(1994) Quantified
impurity analisys of diamond films using SIMS, electrical
characterization of boron implants, and a boron-implanted FET,
fabricated and characterized in diamond, in V.D.Zhitkovsky,
B.V.Spitsyn,AF.Belyanin,L.L.Bouilov,andV.G.Glotov(eds)
Diamond.Diamond Films. Proc. I-st Int.Seminar on Diamond Films
(Ulan-Ude and Baikal, June 30-July 6,1991), CNITI, Tekhnika
sredstv svyasi. Nauchno-tekhnicheskiy sbornik.Seriya tekhnologiya
proizvodstva i oborudovaniye. Vypusk 4, Moscow, 110-115.
DIAMOND GROWTH BY HOT CARBON FILAMENT CHEMICAL VAPORDEPOSITION
C. H. CHAO", E. 1. CHARLSON", E. M. CHARLSON", J. MEESE",M. A. PRELASb and T. STACY"aDepartment ofElectrical and Computer Engineeringb Department ofNuclear EngineeringUniversity ofMissouri-Columbia, Columbia, MO 65211, USA
1. Abstract
Diamond depositions from hydrogen-methane-acetone mixtures using hot carbonfilaments as a thermal source have been performed. Depositions on both the carbon filaments and the silicon substrates have been observed. Stratic pyrolytic-graphite (PG)coatings were observed on the carbon filaments at temperature above 1800°C. Additionof oxygen from acetone enabled diamond growth on silicon substrates by pitting the PGcoatings. Diamond growth rate on substrates is about 0.1-0.2 Ilffi per hour resulting from amajor hydrocarbon deposition on the carbon filaments. The diamond particles depositedon the substrates are about 1 Ilm in size and have cubo-octahedron shape consisting ofsmooth {100} and rough {Ill} faces. The rough {Ill} faces consist of triangular tilesand pits, resulting from the low carbon concentration near the growth planes. This methodprovides a contamination-free, PG-coated and durable carbon filament for diamondgrowth and for bias enhanced diamond growth.
2. Introduction
Hot filament chemical vapor deposition (HFCVD) is a promising method for depositing diamond films because of the possibility of upscaling, easy control of process parameters and low cost setup. Diamond growth rates are normally about 1-2 J.Ull per hour.The filaments are carbide-forming refractory metals such as tungsten and often set undertension to ensure a certain filament to substrate distance. The filaments become brittle after their conversion into carbide and are easily fractured under coolinglheating cycles.Due to the low growth rate of HFCVD, investigators usually combine DC bias or othertechniques to boost the diamond growth rate. The consequence is often tungsten contamination from the filament.
It has been found that in the tungsten filament carburization process tungsten(W) is transformed to W2C then WC, and then a pyrocarbon coating is formed on thefilament [1]. In HFCVD, diamond growth can still proceed after the tungsten filament is
47
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 47-51© 1995 Kluwer Academic Publishers.
48
carburized and coated with pyrocarbon. It has also been found that a non-carbide-formingrefractory metal filament, such as platinum, produces only graphitic depositions [2],which means that such filament can not interact with hydrocarbon species and providethem appropriate activation energies. Thus the purpose of hot filament is to provide interaction for the activation of the reactant gases. In this study, the use of carbon filament forCVD ensures a contamination-free deposition and a durable PO-coated carbon filamentsfor multiple diamond growths.
3. Experimental
Diamond films were grown in a typical bell jar hot filament chemical vapordeposition system. Silicon substrates were scratched with 1 micron diamond paste, followedby organic cleaning in ultrasonic bath and thorough rinsing in de-ionized water.Trimethyloxylborate, (CH30hB, dissolved in acetone was used as the doping source. Two6-inch long, I/16-inch in diameter, evaporation grade carbon rods carrying 80-100ampere current were used as hot filaments. The diamond films were grown at substratetemperatures of about 850°C and at filament temperatures about 2000°C, measured by atwo-wavelength infrared pyrometer. The gas mixture consisted of 97% Hz, 1.5% CRt, and1.5% of acetone vapor carrying dopant. The deposition pressure and the filament-substratedistance were maintained at about 35 torr and 2-4 rom, respectively.
4. Results and Discussion
Depositions on both carbon filament and silicon substrates have been observed.A high ratio of oxygen-containing hydrocarbon gases and a shorter filament-substratedistance were required for diamond depositions on substrates. No diamond growth onsubstrates but outgrowth of pyrolytic graphite on filaments has been observed without theadditions of oxygen-containing reactant gases. A higher hydrocarbon content (with acetone) resulted in ball-like nanocrystalline diamond particles on silicon substrates. The factindicates a possible diamond growth mechanism from C-O contained radicals. Diamonddepositions on silicon substrates were characterized by x-ray diffraction (XRD), scanningelectron microscopy (SEM), energy dispersive spectroscopy (EDS) and Raman spectroscopy. XRD produced d-spacings of the diamond depositions of 2.053A (Ill), 1.256A(220), and 1.078A (311) which compares to 2.060A, 1.26IA and 1.075A respectivelyfrom ASTM 6-675.
SEM micrographs of a diamond deposition on silicon at 850°C for 11 hours isshown in Figure 1. A relatively low growth rate of about 0.1-0.2 11m per hour were observed. The diamond particles are about 1 11m in size and have cubo-octahedron shapeswith smooth {100} and rough {Ill} faces. The rough {Ill} face consists of triangulartiles and pits, as shown in Figure lb. We speculate that the growth structure observed isdue to the low hydrocarbon concentration near the growth plane. Similar triangular
49
structures on {Ill} surfaces of diamond crystals grown by HFCVD using 0.5 % methanein hydrogen were reported under high resolution SEM [3]. The EDS for the diamonds under SEM indicated that no contaminants with atomic number over 10, except silicon fromsubstrates, was found. Raman scattering spectrum of the diamond film exhibits the characteristic diamond signature at 1336 cm-! and the graphite peak at around 1550 cm-! asshown in Figure 2.
Figure 1 (a) SEM micrograph of a diamond deposition on silicon at 850 DC for 11 hours with a relatively lowgrowth rate of about 0.1-0.2 11m per hour. (b) The diamond particles have cubo-octahedron shapes consistingof smooth {loo} faces and rough {III} faces.
600 800 1000 1200 1400
Raman shift (cm-1)1600 1800
Figure 2 Raman scattering spectrum of the diamond film exhibits the characteristic diamond signature at 1336cm- I and the graphitic phase at around 1550 cm- I .
50
The stratic pyrolytic-graphite coatings were observed on all the carbon filamentsat temperature above 1800 °C, as shown in Figure 3a-b, while substantial etching on carbon filament was observed at lower filament temperature in the oxygen-containing reactant gas mixture. Some scattered etching pits were also observed on carbon filaments athigh filament temperature. Possible diamond particles can be observed on the boundariesand summits of grains on the surface of the PO coating, as shown by the bright spots inFigure 3c and by micro-Raman spectrum of Figure 4.
Figure 3 (a) The pyrolytic-graphite coating on the carbon filament about 20 11m in thickness. (b) the stratic POcoating on the carbon filament. (c) surface structure of the PO coating. The bright spots observed on theboundaries and summits of grains on the surface of the PG coating can possiblly be diamond particles.
1590
700 900 1100 1300 1500Raman shift (cm-1.
1700 1900
FiMure 4 Micro- Raman spectrum of the carbon filament contains both peaks for disorder graphite at 1366 em"and 1590 em,l, The peak marked 'd' at 1332 cm'l is most probably due to diamond particle.
51
The laser for micro-Raman scattering with an 8 Ilm beam size was focused onthe largest bright spot, about 4 Ilm in diameter, on the PG coating as shown in Figure 3c.The spectrum of carbon filament, shown in Figure 4, contains peaks for disorderedgraphite at 1366 cm- I and 1590 cm,l. The peak marked 'd' at 1332 cm- I is probably due todiamond particle. The micro-Raman spectra of the pyrolytic graphite depositions on thetungsten, carbon, and graphite filaments at higher temperature suggested ordered graphiteat 1590 cm- I . The fact that the structure ofpyrolytic graphite improves as the temperatureincreases is consistent with an XRD study on the CVD PG structure as a function ofdeposition temperature [4].
5. Conclusion
In hot carbon filament chemical vapor deposition, diamond deposition on siliconsubstrates and pyrolytic-graphite deposition on carbon filaments have been observed. Astratic pyrolytic-graphite (PG) coating was observed on the carbon filament at temperatures above 1800°C. Diamond growth rate on substrates is about 0.1-0.2 Ilm per hour resulting from a major hydrocarbon deposition on the carbon filaments. No diamondgrowths on substrates have been observed without the additions of oxygen-containing reactant gases. The diamond particles have cubo-octahedron shape with smooth {IOO} andrough {III} faces containing triangle tiles and pits due to low carbon concentration neargrowth planes. This method provides a contamination-free, PG-coated and durable carbonfilament for diamond growth and for hybrid techniques to increase the diamond growthrate. Further studies are under investigation.
6. Reference
I. Kharalyan, S. L., Chatilyan, A. A., and Merzhanov, A. G. (1990) Kinetics of tungsten - methane interaction, Soviet Journal of Chemical Physics 6, 404-419.
2. Singh, 8., Ark Y., Levine, A. W., and Mesker, O. R. (1988) Effect of filament and reactor wall materials inlow-pressure chemical vapor deposition synthesis of diamond, Applied Physics Letters, 52, 451-452.
3. Hirabayashi, K. and Kurihara, N. I. (1990) Triangular structures on {III} surfaces of diamond crystalssynthesized by the hot-filament CVD method. Japanese Journal ofApplied Physics, 29, L1901-L1903.
4. Galasso, F. S. (1991), Chemical Vapor Deposited Materials, CRC Press, Boca Raton, pp.114-122.
DIAMOND PARTICLES ON SILICON TIPS:PREPARATION, STRUCTURE, AND FIELD EMISSION PROPERTIES
E.I.GIVARGIZOV, A.N.STEPANOVA, L.L.AKSENOVA, E.V.RAKOVA,P.S.PLEKHANOV, V.V.ZHIRNOV, AND A.N.KISELEV
Institute of Crystallography, Russian Academy of Sciences,Moscow 117333, Russia, FAX (095) 135-1011
Diamond particles with sizes in the micrometerrange were deposited onto sharpened Si tips. A marked tendency for deposition of the particles on veryend of the tips was observed. Field emission studiesof the tips with the particles have shown that theemitters have a decreased work function.
1. INTRODUCTION
Negative electron affinity (NEA) of diamond as ar?roperty inherent in the material is known for a time Ll].Quite recently, the property has attracted a strong inte·rest for applications in vacuum microelectronics and inother fields of modern science and technology il]. Tentative applications in field-emission displays (FED) isprobably one of the most popular.It is clear that preparation of field emitters from
single-crystalline diamond is not realistic due to extremely high price of the material.Another way consists in deposition of diamond coatings
on specially formed substrates: it is far easier and, inaddition, allows to prepare objects, e.g., with higheraspect ratio.In this paper, a technique is described where diamond
(as separate particles or continuous coatings) is deposited onto a very end of sharpened Si tips. The objectsprepared, as well as their field-emission properties,were studied.
2. EXPERIMENTAL TECHNIQUES
Si tips were prepared from Si whiskers qrown accordingto the vapor-liquid-sOlid (VLS) technique l~. Gold wasused as a liquid-forming agent, and solidified hemispherical Si-Au alloy remained on very end of the whiskers.
53
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 53-62© 1995 Kluwer Academic Publishers.
54
If Au particles were deposited in a regular manner, arrayof whiskers was formed (fig. 1).
Fig.l. A regular array of Si whiskers on (lll)-Si substrate.
Then, the hemispherical alloy cap was removed by chemical etching, some sharpening occuring durin~ the process.Finally, oxidation-dissolution procedure [4Jensured further sharpening [5]. In such a way, arrays of very sharpSi tips were prepared (fig. 2). Typical heights of thetips were 50 to 100 pm.Next, diamond particles were deposited onto end of the
tips by a hot-filament ~rocess from a methane-hydrogenmixture at 750-9000 C ~J. No seeding was used to facilitate the deposition of the diamond particles.Microstructure of the particles was studied by electron
diffraction and by transmission electron microscopy (TEM)technique.Finally, the tips with the particles were tested for
field emission in high vacuum.Some experiments on ion milling of diamond particles
deposited on plane Si substrates were made, too.
3. RESULTS AND DISCUSSION
3.1. Morphology and microstructure of diamond particles
The principal result of the deposition is the fact thatdiamond particles were formed on very end of the tips oron side faces, preferentially close to the ends, or on
Fig.2. A regular array of verysharp 8i tips.
sharp edges of the faces. Typical pictures are shown infigs 3 to 5.
Fig.3. Diamond particles on 8itips.
55
56
Fig.4. Almost continuous diamondcoating of tips.
Fig.5. A regular array of Si tipswith diamond particles onthem.
57
Investigations of the particles at relatively high SEMmagnifications revealed a fine-grain surface microstructure (fig. 6).
Fig.6. Fine grains of diamond areseen in the particles atrelatively high magnifications.
Microstructural analysis of such particles by reflection electron diffraction has shown that they representa fine-crystalline diamond. Judging from the sharpness ofdiffraction rings, a conclusion was made that typicalsizes of diamond crystallites in the particles were about10 nm. TEM studies confirmed this conclusion (fig. 7).As is seen, one diamond microparticle (shown by the arrow)was deposited onto another.
3.2. Deposition mechanism
These results can be explained as follows.The hot-filament process is based on decomposition of
methane, typically at 700 to 10000 C (but also at lowertemperatures), in the presence of atomic hyd~o$en, thedeposition rate increasing with temperature L~. A feature of the process is that the atomic hydrogen participates actively in it supressing formation of graphitephase [~' the hot filament serving as a source of theatomic hydrogen.
58
Fig.? TEM of diamond microparticles (MP), depositedon a Si tip.
In our apparatus, the hot filament was placed abovethe substrate at a distance 10 to 20 mm, the temperatureof ends of the tips was not strongly different from thatof the substrate. Distances between tips were 30 to 50 um.Each of the tips had a hemisphere from which a given tipwas supplied by both methane and atomic hydrogen.Atomic hydrogen can recombine at very end of the tip
evolving a large amount of energy. Due to low thermalconductivity of Si, and to a high aspect ratio of thetips, thermal sink to the substrate is not large. Forthese reasons, a marked increase of the local temperatureof the very end of the tips can be assumed. Accordingly,the ends and the edges are places for preferential deposition of diamond.Earlier, a preferential deposition of diamond particles
on cone/pyramidal projections of Si ~ubstrates, preparedby cQemical etching, was observed by Denning and Stevenson l.9].
3.3. Field emission from the diamond particles
For field-emission studies, all tips, except of one,were removed mechanically (e.g., by a needle) from thesubstrate, and measurements were made with the remainingsingle tip.
1 (j -e
-, 10 _7
«
59
Typical result of the measurements in I-V coordinatesis given in Fig. 8, while that in the Fowler-Nordheim (FN)coordinates in Fig. 9.
q,-7/Y
,.t'.n~·
,.tl
?c: 10 _8
1
' lClJ I
et. 10 -9 tP
10 -,0j l10 -11 ILlLI" LLlLI"L1L111"'L1"L111'L1""L111',L1L1j' '"'=,
;,00 400 600 8GO 1000 'I ZOO 1400 1800\I0 Ito ge ( V )
Fig.8. I-V characteristic offield emission from adiamond tip.
1C' -14
10 -15
0'9
\'0
\ll
'\\\~
'\rnn,""""""".""""",,,,,,,,,,,,,,,,,,,,,,,,•• ,,,,,,,,,,,,~,,,,,,,0,000 0.001 0.00! 0.002 0.002
1 /'1
Fig.9. I-V characteristic inF-N coordinates.
60
A specificity of our emitters is that the diamond particles are relatively large and consist of a lot of tinycrystallites having different coordinates, while only(lll)-faces exhibit NEA. This means that only a part ofthe surface is actively emitted one.Using a computer program developed for the F-N formula
and fitting experimental data to the calculations, we determined values of effective work function. They werefound to vary between 0.3 and 1.2 eVe
3.4. Ion milling of diamond particles
Some experiments on ion milling of diamond particlesdeposited onto plane Si substrate were made with the aimto sharpen them. Argon ions Ar+ with energy 30 keV at doses "'" 1018 em -2 were used in the experiments.The result is given in Fig. 10.
Fig.lO. Microcones formed fromdiamond particles deposited on a plane Sisubstrate.
61
4. CONCLUSION
Diamond particles with sizes in the micrometer rangewere deposited onto sharpened Si tips. A marked tendencyto deposition of the particles on very end of the tipswas observed. Field-emission studies of the tips with theparticles have shown that the emitters have a decreasedwork function.
ACKNOWLEDGEMENTS
The authors thank Dr. E.S.Mashkova from the MoscowUniversity for experiments with ion milling, L.N.Obolenskaya for preparation of Si tips, O.B.Volskaya and N.N.Chechulina for assistance in preparation of the manuscript.
REFERENCES
1. Himpsel,F.J., Knapp,J.A., Van Vechten,J.A., andEastman,D.E. (1979) Quantum photoyield of diamond(111) - a stable negative affinity emitter, Phys.Rev. B20, 624-627.
2. Geis,M.W., Twichel,J.C., Bozler,C.O., Rathman,D.D.,Efremov,N.N., Krohn,K.E., Hollis,M.A., Uttaro,R.,Lyszcard,T.M., and Kordesh,M.(1993) Diamond fieldemission cathodes, Paper at 6th Intern. Conf. Vacuum Microelectr., Newport, RI, USA.
3. Givargizov,E.I.(1993) Ultrasharp tips for fieldemission applications prepared by the vapor-liquid-solid growth technique, J. Vac. Sci. Technol. B11,449-453.
4. Marcus,R.B., Ravi,T.S., Gmitter,T., Chen,K., Liu,D.,Orvis,W.J., Ciarlo,D.R., Hunt,C.E., and TrujillO,J.(1990) Formation of silicon tips with < 1nm radius,Appl. Phys. Lett. 56, 236-238.
5. Givargizov,E.I., Kiselev,A.N., Obolenskaya,L.N.,and Stepanova,A.N. (1993) Nanometric tips for scanning probe devices, Appl. Surf. Sci. 67, 73-81.
6. Givargizov,E.I., Zhirnov,V.V., Kuznetsov,A.V., andPlekhanov,P.S. (1993) Growth of diamond particleson sharpened silicon tips, Mater. Lett. 18, 61-63.
7. Yamaguchi,A., Ihara,M., and Komiyama,H. (1994) Temperature dependence of growth rate for diamondsgrown using a hot filament assisted chemical vapordeposition method at low substrate temperatures,Appl. Phys. Lett. 64, 1306-1308.
62
8. Spitsyn,B.V., Bouilov,L.L., and Deryagin,B.V.(1988)Diamond and diamond-like films: deposition fromvapour phase, structure, and properties, Progr.Cryst. Growth Charact. 17, 79-170.
9. Dennig,P.A. and Stevenson,D.A.(1991) Influence ofsubstrate topography on the nucleation of diamondthin films, Appl. Phys. Lett. 59, 1562-1564.
TO THE QUESTION OF THE DIAMOND NUCLEI'S FORMATIONFROM THE GAS PHASE
A.P. RUDENKO, I.I. KULAKOVAChemical Department. Moscow State University, Leninskiye Gory,119899 Moscow, Russia.
One can't think that the formation of polymorphic carbon forms (diamond, graphite,charbon) having different chemical nature of carbon-carbon bonds is a simple physiC<'l1crystallization of carbon atoms. Undoubtedly in this case chemical factors have adecisive imporlance and the above-mentioned polymorphic forms of carbon areproducts in different chemical reactions. As it has been shown in [1,2] they arecondensing products which form in polycondesation processes of the type
(I)
where A represents a carbon-containing molecule, that is a polycondansation
{K
monomer; An} -a polymeric condensing product adsorbed on the catalyst; n - the(Ids
degree of the molecular condensing; {An - mL} ~ds - remaining on the catalyst solid
condensing products with C-C-bonds of one or other crystalline modification ofcarbon or of different polymeric forms of one; L • light molecules removing during thepolycondensation process.The direction in which the molecular polycondensation takes place is determined
primarily by the kinetic conditions in so far as the thermodynamic conditions areequally favorable for all these processes.
The diamond fonnation from the gas phase is a chemical process ofpolycondensation of different light carbon-containing substances, for example
CH4 +CO2 ;::! 2Cdi + 2H20
63
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 63-68© 1995 Kluwer Academic Publishers.
(2)
(3)
(4)
64
etc.
(5)
(6)
(7)
At the same time non-diamond carbon can deposit equally as diamond.The original non-dissociated molecules if they have properties of the
polycondansation monomer or forming from them more light molecules, ions andradicals take part in the process in dependence on the gas phase activationtemperature. But in each case the process has a polycondensation character and takesplace at the support temperature and with its directing and increasing participation.
The diamond formation consists of two stages distinguishing by the kineticconditions: the formation of critical diamond nuclei and the growth of them to thedefinite macrostructural forms (films, crystals).As it has been shown [3) diamond nuclei can form under static conditions at the
equilibrium. But the growth of macroscopic diamond crystals needs conditions ofnon-equilibrium open catalytic systems:
{nA -+ Cd; +mL}
A-+ L -+LK· C","'W
(8)
These conditions provide the constant feed of the original light carbon-eontainingmolecules A and the removal of products L. This leads to the displacement of theequilibrium in the system.The diamond formation process in the closed systems is possible only before the
equilibrium in reactions (2)-(7) is established. In this time equilibrium diamondnuclei fonns. These nuclei can represent epitaxial diamond films of several carbonatomic layers.
It was interesting to define the specific parameters of these equilibrium criticaldiamond nuclei obtained in static conditions.The presence of starting substance A and the establishment of definite ratio
between the rates of formation (for.) and destruction (des.) of diamond andnon-diamond carbon are
(9)
65
necessary for the critical diamond nuclei's formation in one from reactions (2)-(7) inclosed local volume. The presence of the support showing a crystal-structuralcorrespondence with the arrangement of the carbon atoms on the diamond faces andbeing capable to initiate of its formation is the important part of the favorable kineticconditions for this process.In according with the Gibbs-Curie's principle of minimum of free surface energy
and the crystal-growth theories of Wollf. Volmer and other authors the work of thefonnation of a such two-dimensional nucleus of a new phase (6G). its critical size (I)and the rate of the formation (1) are equal:
(10)4yhAGv
AG= __G__ . J= fa ex (_ AG)2yhAG
v• '\ kT
where !: is the specific boundary free energy; f· the area of the nucleus; y - acoefficient which takes into consideration the digression the form of the nucleus fromsquare; h - the thickness of the nucleus; 6Gv= (kT/n)ln(PlPcquil) - the free energy ofthe medium; P and Pcquil - the partial pressure and the equilibrium of one of the gas;n- the volume of one carbon atom in the diamond crystal; a - a coefficient whichdepends on the frequency of the molecular collisions and pressure of the gas.This approach has been used in [4] for the estimation of the indicated parameters.
Without knowing the values of a y and !: it is difficult to calculate the specificparameters of the nucleus and the rate of the nucleus formation.As we have shown [1,2) the thickness calculation of the equilibrium epitaxial
diamond film at definite sizes of the two-dimensional nucleus (d=2r) and the localvolume containing the reaction's components, temperature (T) and pressure of the gas(P) can be carried out more simply on the basis of the equilibrium degree of atransfonnation (YT) into diamond in any reactions from (2)-(7). The value of YT isthe thermodynamic character of the corresponding reaction.The thickness of the equilibrium epitaxial diamond film forming on the support is
equal according [1]
(11)
where 0 is the thickness expressed as the number of carbon atomic layers; Cloc - thenumber of carbon atoms in the volume of the local equilibrium; Cl - the number ofcarbon atoms in one monolayer; /3 - the relative volume concentration of A in thevolume of the local equilibrium (if the volume is occupied by a solution ofA inthe melt or by a solid, /3;el; ifit is occupied by a gas, /3=1).In a particular case when the epitaxial film forms on the support limited by
hemispheric range of the local equilibrium with a diameter d (Figure) the thickness ofthis film is equal:
66
(12)
(cr is the area occupied by one carbon atom on the face of diamond, for the {Ill} facecr=5.5A2
; NA - Avogadro's number; R - the gas constant). From Eqn.(I2) it followsthat the value of 0 is proportional to the pressure as it increases from normal value tothe critical one (Per), and it is independent of pressure for P>Per.The results of the 0 calculations for different temperatures of the equilibrium of
reactions (2) and (3) at f3=1 and P=I atm are given in table I.Table shows that the thickness of the epitaxial film- the critical nucleus of diamond
may be greater, the lower the temperature of the polycondensation. And at -thetemperature less than 400-600 K it is determined only by the value of Cloc as the valueof YT reaches the maximum. The value of 0 is the greater the greater the size of thenucleus (d=2r) and the volume of the local equilibrium. In the limit the ratio of thefilm thickness (h=O*2.06A) to its diameter is independent of the diameter of thenucleus and is equal about 2x 10-4 •
The obtained results show that the equilibrium epitaxial nuclei of diamond crystalshave the thickness of a several carbon atomic layers and the great ratio d/o (about5000). The formation the conservation of these films are possible even in conditionsof the equilibrium of closed static systems.
67
If the whole support at the range of the local equilibrium does not appears active inthe diamond forming process then the epitaxial film will grow only on its separateregions. This may result in the formation of the isometric and two-dimensionaldiamond microcrystals with the greater number of atomic layers than it follows fromEqn.(12).In the calculations by Eqn.(12) it may be taken into consideration the change of
the form of the local equilibrium region up the transition to the open system whenthere are no restrictions connected with the quantity of carbon-containing moleculesin the gas phase. In this case the formation of macroscopic films and macrocrystalsof diamond becomes possible.
References
1. A.P. Rudenko, 1.1. Kulakova, V.L. Skvortsova (1993) Russian Chemical Reviews62(2) 87-104.
2. A.P. Rudenko, 1.1. Kulakova (1993) Vestn. Mosk. Univ. Ser.2 Khim. 34(6)523-548.
3. A.P. Rudenko, 1.1. Kulakova (1989) Geokhimiya (7) 261-272.
4. B.V. Deryagin, D.V. Fedogeev (1977) Rost Almaza i Grafita iz Gasovoi Fazy(Moscow: Nauka).
68
a
(a)
(b)
,(c)
Figure. The equilibrium epitaxial diamond film on a support showing a crystal-structural correspondencewith the {III} face. (a) View of a crystal nucleus from above: the arrows indicate the free valances ofcarbon atoms, lying at an angle of 90· to the (III} sUlface and localized by the functional groups; 1,2- carbon atoms of the upper and lower sheets respectively. (b) View of the critical nucleus from the side:1- three layer epitaxial film of diamond; 2 - the face of the support. (c) Model of the region of the localequilibrium on the surface of the support bringing about the epitaxial growth of the diamond film: I - theepitaxial film with thickness Il and diameter d =2r; 2 - the support; 3 - a hemisphere bounding the rangeof the local equilibrium of reaction.
ELECTRICALLY AND OPTICALLY ACTIVE IMPURITIES
AND DEFECTS IN DIAMOND
A.A. GIPPIUSP.N. Lebedev Physical Instituteof the Russian Academy of Sciences,Leninsky prospect 53, Moscow 117924, Russia
Properties of impurities and defects affecting electrical and optical parameters ofdiamond are reviewed briefly with the emphasis on their possible use in diamondbased electronic devices. Special attention is given to doping diamond by ionimplantation with relevant problems of defect control discussed on the exampleof boron-implanted p-type layers. Luminescent centres due to impurity-defectcomplexes are considered in view of their possible use in diamond based lightemitting devices.
1. Introduction
Diamond is distinguished from all other semiconducting materials in thatpractically all its properties were determined and even several types ofdevices produced using natural crystals. But it is the progress in growinglarge semiconducting diamonds and in particular the explosive increase,over the last 15 years, of the research and growth of CVD diamond filmsthat led to a renewed interest in the use of diamond for semiconductingdevices.Once the main physical parameters of diamond were established [1]
it became considered as a material possessing many attractive properties for the fabrication of high-power, high-frequency, high-temperaturesemiconducting devices. This notion was based upon the values of thermal conductivity (20 W'cm-t, which is about 5 times that of copper atroom temperature), breakdown field (107 V·cm- I ), saturated electron ve-
69
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 69-80© 1995 Kluwer Academic Publishers.
70
locity (2.7.107 cm's- I ), hole mobility in natural semiconducting diamond(2000 cm2y- I s- I ). Johnson figure of merit for the power and frequencyperformance and Keyes figure of merit for the speed, based upon theseparameters, were found to be respectively 8206 and 32.2 times higher fordiamond than they are for silicon [2]. These estimates, combined with theseemingly encouraging results on prototype diamond devices, have led to anover-optimistic view of the prospects of diamond as a material for electronics. Some authors claimed that, basing upon CYD technology, one couldproduce high power microwave amplifiers [3] and microcircuits with highpackaging density [4], capable of working at higher temperature [3] andfrequencies than the best that can be achieved using GaAs [5].In contrast to these enthusiastic views Collins [6] presented a critical
appraisal of prospects of diamond applications, based upon the analysis ofelectronic properties of diamond (in particular, taking into account the temperature dependence of device related parameters) and realistic estimatesof relevant technological problems. His conclusion was that "the likelihoodof diamond device outperforming Si or GaAs in other than a few esotericapplications seems extremely remote" [6]. This view, which left little hopefor diamond to become a material of XXI century intended to substitute Siof GaAs, does not exclude its use in active devices of electronics in general.Certainly, in the competitive world the area of application is the area ofsuperiority. It is the unique combination of properties (stability at high temperature, low sensitivity to nuclear radiation, chemical inertness and thementioned electronic parameters) that makes diamond a really promisingmaterial for high-power (but not necessarily very high-frequency) deviceswith the improved radiation resistance, intended to be used at high temperature and in hostile environment.Applications of diamond in optoelectronics should also be considered
seriously in view of its transparency in wide spectral range and rich varietyof optical centres, potentially suited for light emitting devices.The development of diamond-based semiconductor technology has just
begun. The specific properties of diamond determine both the advantagesand disadvantages of this technology. Since pure diamond is a perfect insulator, the problem of insulation of elements and devices on a wafer looksrelatively simple, if proper combination of undoped and doped CYD orion-implanted layers are used. On the other hand, the absence of a stablenative oxide and the very chemical inertness of diamond present definitedifficulties in patterning device structures.Production of semiconductor grade material is evidently the vital com
ponent of diamond electronic technology. It implies development of techniques of doping and defect control. In the following sections we shall discussthese two interrelated subjects.
71
In section 2 theoretical and experimental data on the properties of acceptors and donors in diamond are briefly discussed. Section 3 deals with theproblems of defect control, with the emphasis on ion implantation dopingwhere these problems are particularly important, in view of great amountof damage introduced by ion implantation. In section 4 some optical centrespotentially interesting for the use in light-emitting devices are consideredalong with the relevant problems of defect control and luminescence excitation mechanisms.
2. p- and n-type doping of diamond
The material in this section will be presented not in chronological orderof relevant works, but following the sequence: theory-experiment-applications.Diamond is a classic covalent solid with the lattice parameter of cubic
unit cell 3.567 A, space group O~ and interatomic distance between tetrahedrally coordinated Sp3 hybridized carbon atoms 1.54 A. The diamondstructure is loosely packed with 4 octants (out of 8, to which the unit cellcan be divided) filled with tetrahedra and 4 are empty. Due to small tetrahedral radius of carbon (0.77 A) almost all impurity atoms (with notableexception of Band N) are considerably larger and their incorporation intodiamond implies deformation of the lattice.Diamond belongs to the elements of the fourth group which includes
semiconductors Si and Ge. It was natural to choose dopants for diamondbasing upon the analogy with Si and Ge. On this way a number of groupIII, group V and group I elements have been tried both experimentally andtheoretically as candidates for dopants for diamond.Starting with theory we should mention the work [7] devoted mostly to
donors but with some reference to boron acceptor. The calculations basedupon local density theory and pseudopotential formalism were performedfor impurities at the substitutional and interstitial sites. As expected, Pprefers the substitutional site, whereas Li and Na prefer tetrahedral interstitial position. It is instructive to compare calculated formation energiesfor neutral impurities given in [7]:B: -0.5 eVj N: -0.6 eV; P: +10.4 eV; Li: +5.5 eV; Na: +15.3 eV.Note, that negative formation energies of substitutional Band N pro
vides a thermodynamic force for their incorporation into diamond. "Experimental proof" of these calculations is provided by occurrence of naturaldiamonds containing isolated N (type Ib) and B (type lIb) impurities.As far as P, Li" and Na are concerned, the theory predicts that they
introduce levels respectively 0.2, 0.1 and 0.3 eV below the conduction band,which makes them good candidates for shallow donors. However, due to
72
the relatively high formation energies, the solubility of all these donorsmust be very low. The authors conclude, that for doping purposes the highformation energies of these impurities would require kinetic trapping ofthe atoms during growth or ion implantation, since their low solubilitiesprevent significant incorporation via in-diffusion.
Available experiment~l data confirm the predictions of theory. Boronwas found to be the only successful dopant which gives rise to p-typeelectrical conductivity in diamond. Properties of boron as a classical acceptor with an ionization energy E; = 0.37 eV are well established intype IIb natural diamond [8], where the temperature dependence of thehole concentration at low temperatures (below about 330 K) is dominatedby the exponential factor exp( -E;fkT). In boron-doped CVD films thetemperature dependence of resistivity was found to follow the Arrheniusequation p = const exp(W/ kT) with the activation energy W between0.1 and 0.37 eV, depending on boron concentration [9,10], the reductionof the activation energy for heavily doped films being attributed to theformation of the impurity band. Low activation energies found in heavilyboron-implanted layers were explained in terms of variable-range hoppingconduction.
The problem of obtaining "electronic grade" doped diamond is seriouseven for boron doping. Boron can be introduced into large synthetic crystalsgrown by the temperature-gradient [11]. In this case the rate, at whichboron is incorporated, depends strongly on the orientation of the growthplane and is higher in the octahedral growth sectors than in the cubicgrowth sectors. The same is true for boron doped CVD films (which areobtained introducing B2H6 into the plasma gas mixture [12]) at least forhigh boron concentration. The inhomogeneity of doping is aggravated forCVD films by the general problem of obtaining CVD homoepitaxiallayersof quality comparable to semiconductor grade silicon at realistic growthrates. During last several years much attention was given to developmentof effective routes of doping diamond by ion implantation. In this case themost serious problem is that of defect control, to which we shall refer inthe next section.Experimental data on donors in diamond are scarce. Isolated substitu
tional nitrogen is known to be a donor with an ionization energy of about1.7 eV [13]. It might be responsible for partial compensation of boron acceptors in type IIb diamonds. Nitrogen is known from EPR data to distortin the (111) direction [14]. The first principle forces calculations [7] showed,that the energetically favourable distortion with total gain in energy 0.76 eVincludes full lattice relaxation around N atom and lengthening of C-N distance along (111) by 25% (experimental values range from 10 to 36%). Theenergy level of the ground (nondegenerate) state drops from E e - 0.7 eV
73
for undistorted substitutional N to E c - 1.5 eV in fair agreement with theexperimental value of E c - 1.7 eV. Nitrogen as deep donor is of no practicalimportance for n-doping of diamond.
No effective donor type atom has been found that will substitute forcarbon during high-pressure synthesis of diamond [11]. Attempts to dopeCVD films by admixing PHa were partially successful [15] in the sensethat, with n-type conductivity definitely established by Hall-effect measurements, the resistivity of phosphorus-doped films was usually too highfor electronic applications. With the activation energy in the range 0.081.16 eV, depending upon growth and doping conditions, and theoreticallypredicted ionization energy of 0.2 eV it is definite that the contributionof lattice damage to the conductivity of phosphorus doped films must beconsiderable.
As far as lithium and sodium are concerned, the only doping techniquewhich have been used so far is ion implantation. Early results of LebedevInstitute (Moscow) group [16,17] which demonstrated n-type conductivity(with the activation energy of 0.1 eV and with relatively high electron mobility of about 1000 cm2V- 1c- 1 ) of Li-implanted natural diamond, were laterconfirmed by the studies made at Harwell [18] and Technion (Haifa) [19]. Inthe latter work some general view is presented of the results referring to Liand Na implantation. The authors replotted the data of temperature dependence of conductivity of Li and Na implanted diamond (usually analyzedin terms of activated conductivity process with multiple activation energiesdeduced from Arrhenius, Le. 10g(R)vs(I/T), plots) as log(R)vs(I/T)1/4and found that both their data and many of the data of other authors (inparticular, those of [18]) "exhibit a linear dependency with a single slopecovering the entire temperature range, consistent with a variable range hopping mechanism". This conduction mechanism, being somewhat differentfrom what was expected of Li or Na doping, does not exclude the possibleuse of these dopants as donors. This was confirmed by successful fabricationof p-n junction in diamond, using Li-doped (n-type) layer [18].
3. Doping and defect control
For any technological process the final properties of solid state structuresare the outcome of reactions on the surface, interfaces and in the crystalvolume between defects and impurities, both background and introducedby a given operation. In the case of doping the goal, in principle, is to introduce a given impurity (or a given combination of impurities) to obtainmaterial, whose properties are dominated by dopants and not by defects orbackground impurities. In this section we shall consider problems of defectcontrol as applied to ion implantation, where these problems are particu-
74
larly important due to strong radiation damage introduced by implantation.The studies of ion implantation in diamond started at the end of the six
ties at Lebedev Institute (Moscow), their summary is presented in a reviewof Vavilov [20]. Since then considerable amount of work has been done inthis field, with important contribution of group of University of Witwatersrand (Johannesburg) summarized in a review paper of Prins [21]. We shalltouch here some problems of defect control related to ion doping, leavingaside "amorphization" or "graphitization" which occur at high fluences ofions.One of the key features of ion implantation is the very big amount of
radiation damage, the number of vacancies and interstitials in the implantedlayer exceeding the number of implanted impurity atoms by two orders ofmagnitude for ion energies around 100 keV. We deliberately use the word"amount" instead of "density" since we believe that it is the inhomogeneityof damage and, in particular, high local density of defects within disorderedregions produced by ions, that strongly affect the processes of impuritydefect interactions within implanted layers.Any efficient scheme of ion doping implies that an implanted atom
should be activated, i.e. placed at the correct lattice position (for instance,substitutional for boron and interstitial for lithium) and the effects oflatticedisorder introduced by implantation should be minimized. Since the conditions necessary to activate substitutional or interstitial impurities must bedifferent, we shall consider first the case of boron.
It should be noted that quantitative data on the efficiency of activation of implanted boron (i.e. the ratio of amount of acceptor centres tothat of implanted atoms) are scarce. The value of sheet resistance determined in conditions where its temperature dependence at least contains acomponent with the activation energy corresponding to boron ionizationenergy, is often taken as a measure of efficiency of activation. To increasethis efficiency various conditions of implantation and annealing have beentried, the most extensive experimental and theoretical studies have beenperformed by Prins and co-workers [21]. The "cold implantation-rapidannealing" sequence, suggested by Prins, produced so far the best resultsin boron implantation doping, but it is the physical approach used thatneeds some comment in relation to general problems of defect control.The assumptions made by Prince and co-workers in their theoretical
treatment and experimental studies are the following: a) "frozen" defectproduced by cold implantation are evenly mixed ("a fair, macroscopic approximation" [21]); b) "dopant interstitials" will diffuse and interact withvacancies in, approximately, the same way as self-interstitials during annealing; c) spacial distribution of implanted ions is more symmetric andpeaks deeper below the surface than the vacancy distribution.
75
In the search for an effective "implantation-annealing" sequence it isargued, that the difference between the profiles of implanted atoms andradiation damage limits the efficiency of activation of substitutional impurities, since dopant interstitials "compete at a disadvantage for vacancieswith self-interstitials" which are "more intimately mixed with vacancies".It was concluded that it should be helpful to introduce additional damage(for example, by C+ co-implantation) to provide more vacancies needed toactivate the implanted boron atoms.
Let us note first of all that simple comparison of distributions of vacancies and implanted atoms seems somewhat misleading, if we take intoaccount that integrated profile area of vacancies is two orders of magnitude larger than that of impurities. Additional damage (proq.uced by coimplantation) which overlaps the impurity profile, introduces equal amountsof vacancies and self-interstitials and should not improve the situation asfar as competition for vacancies is concerned.
Even more important is the inhomogeneity of damage which was nottaken into account in [21]. It is known that in ion-implanted layers considerable fraction of defects is concentrated within disorder regions producedby displacement cascades [22]. Due to dynamics of the cascade formationthe inner part of a disordered region is vacancy-rich, while a large fraction of interstitials is created predominantly at the periphery of the cascade [23]. It is this defect distribution which is frozen in cold implantation experiments. Locally the situation can hardly be described simply as"dopant-interstitial" even for comparatively light ions. As far as heavierions are concerned, it was shown in our experiments, that the defect structure surrounding an implanted Ni atom should be understood not in termsof point defects but rather in terms of gross lattice disorder, which survivesthe annealing temperatures close to the limit of stability of diamond atnormal pressure (1650°C) [24]. Kalish et al found that implanted In atomsare surrounded by strongly damaged region up to the annealing temperature of 1800°C [25]. These data probably explain the failure of attemptsto effectively dope diamond by implantation of relatively heavy atoms andstress the importance of light dopants. In the case of light implants B, Li, Na(?) the local damage is not that severe, so that the lattice canbe restored by proper annealing. It should be noted, that the mentionedstructure of disordered regions implies that there must be enough vacanciesin the closest proximity of an implanted atom to activate a substitutionalimpurity. Moreover, with vacancy-rich nucleus of a disordered region andinterstitials at its periphery the competition for vacancies can hardly be toa disadvantage of an implanted atom.
As far as radiation damage is concerned it can be stated in general(without discussing a controversial concept of vacloids, introduced by Prins
76
[21] as a model for donor-type residual damage) that its contribution tothe observed conductivity is considerable, which is evidenced by multiplicity of activation energies, deduced from temperature dependencies of sheetresistance of diamond doped with boron under various conditions of implantation and annealing. One noticeable exception is cold implantation rapid annealing regime, which gives the temperature dependence of sheetresistance with single activation energy close to the ionization energy ofboron and with reasonable (after additional annealing at 1500°C) value ofresistivity, comparable with the values found in natural type lIb diamonds.We believe that this result is largely due to processes which occur within orclose to disordered regions, where considerable fraction of radiation damageis frozen and where annihilation of its components can be most effective, iffast transition from low to high temperature is realized.
Interstitial impurities, such as Li, require conditions for activation different from those for substitutional boron. In this case the presence of greatamount of vacancies, particularly their high density within disordered regions, is a negative factor. That is why the efficiency of Li activation underusual implantation conditions must be inherently low. To be activated, Liatoms should be somehow separated from the region of high density of vacancies. This is confirmed by the experimental data presented by LebedevInstitute and Technion groups [16,19]. In both studies the n-conductivitydue to Li donors could be found only in the tail of the implant distribution(in some cases up to about 8 times deeper than the damage range [16]).In view of these data it might be promising to perform implantation intoaligned samples to reduce displacement of host atoms, particularly at lowtemperature, which may allow the impurities to remain interstitial [7].
4. Optically active centres
Wide band gap, high transparency and rich variety of luminescent centresmake diamond a challenging material for optoelectronics. The prospectsof fabrication of workable diamond based light-emitting devices are determined by: a) spectral characteristics of relevant luminescent centres, b}availability of technology, capable of introducing into diamond a given combination of luminescent centres, c} availability of efficient mechanisms ofexcitation in electroluminescent devices. In the following we shall brieflydiscuss these aspects.
More than 100 luminescent centres have been documented for diamond.Some of them are present in natural stones or synthetic material producedeither by high-pressure synthesis or by CVD technique. Some specific centres can be produced by radiation damage or ion implantation. The luminescent bands of the centres range from ultraviolet (230 nm, "edge emis-
77
sion") to near infrared (rv 1000 nm, titanium luminescent centre). We shallenumerate some of the best studied centres or luminescent bands, followingthe list given by Collins [26], with some omissions to concentrate on thosecentres, whose structure has been (at least to some extent) established.
1. Edge emission around 5.3 eV (230 nm): radiative recombination offreeand bound excitonsj can be observed only in diamond relatively freeof defects.
2. Band A: a broad (about 100 nm wide) band with maximum at (400-;500) nm most commonly observed practically in all natural and synthetic diamonds. The former attribution to donor-acceptor recombination at present does not seem adequate.
3. The 5RL system, a combination of sharp lines and a broad band inthe range (3.6 -;- 4.6) eV, Le. (270 -;- 380) nm with zero-phonon line at4.582 eV (270.5 nm). The centre involves the carbon interstitial.
4. The 3.188 eV (388.8 nm) system: the center containing a carbon interstitial and a single nitrogen atom.
5. The N3 system, 2.985 eV (415.2 nm): one ofthe most studied vibronicbands in natural diamond. The centre is believed to be three nearestneighbour nitrogen atoms on a (111) plane bonded to a common vacancy.
6. The H3 and H4 systems: 2.463 and 2.499 eV (503.2 and 496.0 nm).The centres consist of a vacancy trapped at the A form and at the Bform of nitrogen, respectively.
7. The 484 nm system: a nickel related centre with zero-phonon line at2.56 eV (484 nm).
8. The 575 nm system (zero-phonon line at 2.156 eV, Le. 574.9 nm): thecentre is composed of a single nitrogen atom and a vacancy.
9. The 1.945 eV (637.2 nm) system. One of the major centres observedin photoluminescence, but not in cathodoluminescence. Consists of avacancy trapped at a single substitutional nitrogen atom.
10. The 1.681 eV (737 nm) system: a silicon related centre.11. The GR1 system (1.673 eV, 741 nm): the neutral vacancy.12. The 1.40 eV (885 nm) system: a nickel related centre.13. The 1.249 eV (992 nm) system: a titanium related centre.
We see that the components of these centres are impurities (nitrogen,by far the most common impurity in diamond; boron, responsible for semiconducting properties of diamond; transition elements nickel and titanium;silicon) or/and lattice defects (vacancies and interstitials).It is evident that for device applications the problem of production of
a desirable combination of luminescent centres is, in general, more difficultthan that of doping with electrically active impurities (which is also notsimple, as was mentioned in section 3 and 4). To introduce the needed
78
amount of, say, three or four component centres (H3 or N3 respectively), togrant that they dominate in luminescence spectrum and that their emissionis not absorbed within the crystal by other optically active centres, one hasto perform some sophisticated defect engineering.The potentials for optoelectronics of some of the luminescent centres
are encouraging. It refers first of all to nitrogen related centres, in particular to, probably, the most promising H3 centre ( laser action under opticalexcitation has already been demonstrated in diamond containing these centres [27]). One of the advantages of nitrogen related centres is that theirmicrostructure is well established and considerable amount of data are accumulated referring to association-dissociation reactions involving nitrogenatoms and vacancies [28).Transition elements (nickel and titanium) and silicon related centres,
emitting in the visible and near infrared regions, might also be promisingcandidates for the use in light emitting devices, but at present few data ontheir structure are available. In the case of nickel we deal probably with afamily of centres containing nickel and nitrogen in various configurations[29). Titanium also forms at least three luminescent centres incorporatingsome background impurities or defects [30). Little is known about the structure of 1.681 eV silicon related centre. Certainly a lot of work should bedone to reach a level of understanding of these centres at least comparableto that of nitrogen related centres.Optoelectronic applications of diamond imply the development of effi
cient mechanisms of electroluminescence. These mechanisms can be basedeither on generation of electron-hole pairs with subsequent transfer of theexcitation to luminescent centres or on the direct excitation of the centresby electron impact. In view of difficulties with n-doping of diamond, theprospects of fabrication of p-n diodes with low resistance (necessary foreffective minority carriers injection) looks dim. Schottky diods can hardlybe a viable solution since it is difficult to generate a high concentration ofelectron-hole pairs at a Schottky barrier. More promising are some versionsof high electric field devices working in the regime of either avalanche breakdown (electron-hole pairs generation) or impact excitation of luminescentcentres.Diamond-based electroluminescent devices, emitting in the visible range
have been fabricated, using Schottky barrier made on boron doped CVDfilms [31), p-n junction produced by carbon implantation into natural ptype semiconductors [32) and the structures similar to thin film electroluminescence devices working in a breakdown mode [33].Perhaps the most encouraging results were obtained using p-i-p struc
tures fabricated on synthetic diamond crystals [34]. The structure consistedof two semiconducting p-type (implanted boron concentration 1020 cm- 3 )
79
stripes (1000 /Lm X 50 /Lm) separated by a narrow (2 /Lm) undoped insulatingregion. Depending upon the properties of the substrates, the electroluminescence, observed under applied voltage of about 102 V, was dominatedby A-band (blue), 575 system (orange) and H3 system (green). The bestresults with light output 0.1 CdxA-1 and efficiency 0.1% were reportedfor green (H3) emission. The excitation mechanism is not quite clear. Itis believed to involve injection of holes into insulating area and their acceleration in the high electric field, resulting in either impact excitation ofluminescent centres or generation of electron-hole pairs.
5. Conclusion
Realization of potentials of diamond as a semiconducting material requiresdevelopment of methods of doping and defect control. Due to the intrinsicparameters of diamond and properties of electrically and optically activeimpurities, ion implantation doping, which is standard technique in siliconsemiconductor industry, is all the more important for diamond. However,in view of diamond's metastability, the ion implantation route is bound tobe more complicated, since remelting and resolidification or solid state epitaxial regrouth of ion-damaged layers by means of rapid thermal annealingcan not be applied to diamond."Defect engineering" , important not only in ion implantation route, but
in a broad sense, in any growth process, is based upon accumulated dataon properties of impurities and defects and their interaction. In spite ofdefinite progress in this field a lot of work is still ahead before we approacha level of understanding comparable to that of silicon or III-V compounds.
References
1. Field, J.E. (ed) (1979) The Properties of Diamond, Academic Press, London.2. Davies, R.F., Sitar, Z., Williams, B.E., Kong, H.S., Kim, H.J., Palmour, J.W.,Edmond, J.A., Ryu, J., Glass, J.T., and Carter, Jr.C.H. (1988) Mater.Sci.Eng. Bl,77.
3. Brown, A.S. (1987) Aerospace Am. 25(No.ll), 12.4. Yoder, M.N. (1987) Nav.Res.Rev. (NARRA) 39(No.2), 27.5. Simpson, M. (1988) New Scientist 117(N0.1603), 50.6. Collins, A.T. (1992) Mater.Sci.Eng. Bll, 257.7. Kajihara, S.A., Antonelli, A. and Bernholc, J. (1993) Physica B185, 144.8. Collins, A.T., Lightowlers, E.C., in [1], chapter 3.9. Spitsyn, B.V., Bouilov, L.L. and Derjaguin B.V. (1981) J.Cryst.Growth 52, 219.10. Fujimori, N., Imai, T. and Doi, A. (1986) Vacuum 36, 99.11. Bundy, F.P., Strong, H.M. and Wentorf (1973) Chem.Phys.Carbon 10, 213.12. Shiomi, H., Nakahata, H., Imai, T., Nishibayashi, Y. and Fujimori, N. (1989)
Jap.J.Appl.Phys. 28, 758.13. Vermeulen, L.A. and Farrer, R.G. (1975) Diamond Research 1975, 18.14. Ammerlaan, C.A.J. and Burgemeister, E.A. (1981) Phys.Rev.Let. 47, 954.15. Spitsyn, B.V. and Aleksenko, A.E. (1986) Arch.Nauki 0 Materialach 7, 201.
80
16. Vavilov, V.S., Gukasyan, M.A., Guseva, M.1. and Konorova, E.A. (1972) SovietPhys.Semicond. 6, 741.
17. Vavilov, V.S., Konorova, E.A., Stepanova, E.B. and Trukhan E.M. (1979) SovietPhys.Semicond. 13, 635.
18. Buckley-Golder, I.M., Bullough, R., Hayns, M.R., Willis, J.R., Piller, R.C.,Blamires, N.G., Gard, G. and Stephen, J. (1991) Diamond Relat.Mater. 1, 43.
19. Prawer, S., Uzan-Saguy, C., Braunstein, G. and Kalish, R. (1993) Appl.Phys.Lett.63, 2502.
20. Vavilov, V.S. (1975) Phys.Stat.Sol.(a) 31, 11.21. Prins, J.F. (1992) Mater.Sci.Rep. 7, 271.22. Kimerling, L.C. and Poate, J.M. (1975) Inst.Phys.Conf.Ser. 23, 126.23. Nelson, R.S. (1973) Inst.Phys. Conf.Ser. 16, 140.24. Gippius, A.A. (1992) Mater.Sci.For. 83-87, 1219.25. Kalish, R., Deicher, M., Recknagel, E. and Wichert, T. (1979) J.Appl.Phys. 50,
6870.26. Collins, A.T. (1992) Diam. Rei. Mater. 1, 457.27. Rang, S.C. and De Shazer, L.G. (1985) Optics Lett. 10, 481.28. Woods, G.S. (1986) Proc.R.Soc.(London), A 470, 219.29. Lawson, S.C. and Kanda, H. (1993) Diam.Rel.Mat. 2, 130.30. Gippius, A.A. (1993) Diam.Rel.Mat. 2, 640.31. Fujimori, N., Nishibayashi, Y. and Shiomi, H. (1991) Jap.J. Appl.Phys. 30,1728.32. Prins, J.F. (1989) Europ.Pat.Appl.j Pub. No.:0413435A2.33. Kim, S.B. and Wagner, J.F. (1988) Appl.Phys.Lett. 53, 1880.34. Burchard, B., Zaitsev, A.M., Fahrner, W.R., Melnikov, A.A., Denisenko, A.V. and
Varichenko, V.S. (1993) Preprint of Fern Universitii.t Gesamthochschule in Hagen.
PREDICTION OF DIAMOND FILM THERMAL CONDUCTIVITY
N.V. Novikov, T.D. Ositinskaya, A.P. Podoba and S.V.ShmegeraInstitute for Superhard Materials, Ukrainian Academy ofScience,2, Avtozavodskaya Str., Kiev, 254074, Ukraine
Abstract
A new mechanism for the scattering of heat-carrying phonons has been suggested,based on the assumption of the activation of normal processes by isotopes. The attemptis made to prediction of diamond film thermal conductivity to isotropic impurity concentration and grit sizes in temperature region O-IOOOK.
1. Introduction
A high sensitivity of diamond lattice thermal conductivity to isotopic impurity therein,experimentally established in [I], has stimulated an interest in a more detail study of theisotope effect on the thermal conductivity of both mono- and polycrystalline structures,including diamond films.
Earlier [2] the authors of the present work have suggested anew mechanism of the effectof an isotopic impurity on the lattice thermal conductivity. The suggested mechanismhas been tested with the help ofCallaway's known model [3]. The results obtained agreerather well with the experimental data, which evidently is indicative of the correctchoice of factors forming the basis of the above mechanism.
2. Theoretical Results
The mechanism of the isotopic effect on the lattice thermal conductivity is based on theassumption that the isotopic impurity present in the lattice activates normal phonon scattering processes, and as a result redistribute the energy heat-carrying phonons betweenpassive (those which do not carry the heat) modes of the lattice vibrations. In this case,the intensity of such forced normal three-phonon processes turns out to depend on theisotope concentration.
Qualitatively, the mechanism for N-process activation is as follows: phonon scattering
81
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 81-88© 1995 Kluwer Academic Publishers.
82
from isotopes leads to the increase of a long-wavelength fraction in a phonon spectrum,that in turn increases the probability of N- process occurrence. N-processes are knownto occur with a contribution of both long-wavelength and short wavelength phonons.Short-wavelength phonons appearing in result of N-processes are able to contributeeffectively to the different resistive scattering processes, the latter are virtually responsible for the value of the lattice thermal conductivity. Figure 1 presents a diagram illustrating one of the most probable channels of phonon scattering in the suggestedmechanism. In this case, the heat transfer is assumed to occur, mainly (about 90%), byphonons with a cross-polarization [4]. It is known that the main contribution to the normal scattering from such phonons are made by the following type processes:
(1)
where g is the wave vector, and t- and I-indices refer to phonons with cross- and longitudinal polarization, respectively.
The diagram in Fig. 1 corresponds to the mass M (a 13cisotope) scattering from a gtheat-carrying phonon with across-polarization. The gt' phonon, which appear in thiscase, can participate in the normal (N) process of interaction with a longitudinal shortwavelength gl' if the gt phonon wavelength is big enough, or with a long-wavelength gl',if the gt phonon wavelength is not sufficient for an independent initiation of the normalprocess. As a result of the normal (N) process, a short-wavelength phonon gl' is produced with a longitudinal polarization, that does not make an essential contribution intoa heat-transfer. Besides, such phonons are very efficient in isotopic mass (M) scatteringor in Umklapp-processes (U).
The suggested mechanism of the isotopic effect on the diamond lattice thermal conductivity has been considered in the framework of Callaway's model using Debye'sapproach. Callaway's expression for the lattice thermal conductivity looks like [3]
(2)
(2.1)
83
(2.2)
where k is Boltzman's constant; Ii is Planck's constant; T is the temperature; eis Debye'stemperature; v is the mean phonon velocity; x = lico/kT; co is the phonon frequency; 'tR isthe total relaxation time resulting from all kinds of resistive scattering; 'tN is the relaxation time ofN processes.
If the lattice contains an isotopic impurity only, three types of resistive processes canoccur: Umklapp-processes (U); isotope scattering; crystal boundaries scattering andintergranular scattering in polycrystalline structures. Then'tR can be expressed as [5]
(3)
where A, b, m, n - som~ consR\nts, 1i isoto~ relative concentration, V0- volume of alattice unit cell, ~M = l3e - 12e, l3e and MI2e being mass of 13C isotope and 1ZCcarbon atom, d} and dz- characteristic crystal dimensions.
The relaxation time ofN-processes scattering for csoss polarization phonons and allowing for the assumption 0- their activation by isotopic impurity can be represented as
(4)
where B - some constant value, fCC) - to be determined isotope concentration function.The inclusion of relaxation time concentration dependence represents an attempt to takeinto account the isotopic impurity initiating role in N-processes as well as their indiced
84
nature. The f(C) function clear type has been chosen from the analysis of the data onisotope impact on the lattice thermal conductivity. The most agreeable has turned to bethe following
f(C)
1
(C (1- C» 2: (5)
Theoretical and experimental data [l,6,7] agree at the following values of the constantscontained in (3) and (4) for e= 1900 K and u= 1.31*104 mls [8], A = 3.6*1O-23sfK2, B= 4.4*10 -11 K-3, b = 8.36, m= 2, n= 2.
FIgure 2 shows the results of calculation of diamond lattice thermal conductivity as afunction of an isotopic impurity content. Some known experimental data [1] are alsogiven here. For isotopic impurity concentration 80 <C<20 %, the calculation resultsagree well with the experimental data.
A considerable contribution into the formation of efficient thermal conductivity of polycrystalline diamond films with grit sizes equal or below 10 Jlm is made by the processesof phonon boundary scattering (Fig. 3). As is seen from the Fig.3 the boundary scattering becomes dominating for the characteristic grit sizes:::;l Jlffi, in this case the isotopicscattering virtually does not affect the thermal conductivity.
Figures 4-6 show the obtained temperature dependences of thermal conductivity of diamond films with different grit sizes and different BC content. As it has been expected,with decrease in grit sizes, the thermal conductivity peak shifts to the room and highertemperature region.
3. Conclusions
It should be noted, that the thermal conductivity values have been calculated for CVDdiamond polycrystals. The CVD method, unlike a high pressure-high temperature sintering, permits virtually porous-free structures to be made. The above calculations donot allow for the anisotropy of diamond film thermal conductivity (the heat flow isassumed to be normal to the film surfaces).
Our calculations permit the prediction of thermal conductivity values of diamond filmsof the approximate structure of perfect polycrystals. The influence of other defects(vacancies, graphite inclusions, etc.) have not been taken into account. The dataobtained make it possible to estimate approximately the limit value of thermal conductivity in relation to the film temperature and grit size.
85
For diamond porous-free coatings with thickness in excess onO-lOO J..Ul1, the thermalconductivity is affected by the de~ee of isotopic purification and increases essentiallywhen crystallites contain 99.5 % 1 Cor 13C atoms
References
1. T.R. Anthony, W.F. Banholzer Diamond and Related Materials,l (1992) 717-7262. T.D.Ositinskaya, A.P.Podoba and S.V.Shmegera Diamond and Related Materials, 2 (lgg3) 1500-15043. J.Callaway,Phys.Rev.113,(1959)1046.4. R.A.H.Hamilton,I.E.Parrott,Phys.Rev.,178 (1969) 1284.5. P.G.Klemens, Solid State Phys.,7 (1958) 596. W.Banho1zer, T.Anthony and R.Gilmore, Proc. Int. Coni. on New Diamond Science and Technology, Materials Research Society, Pittsburgh, PA, (1991) 857
7. Lanhua Wei,P.K.Kuo,R.L.Thomas,et al.,Proc.Int.Conf. on New Diamond Science and Technology, Materials Research Sosiety, Pittsburgh, PA, (1991) 875
8. T.D.Ositinskaya,A.P.Podoba,S.V.Shmegera,Sverkhtverdye Materialy2 (1991) 19-23.
Figures
b)
c)
Figure 1. Scattering of a gt heat-carrying phonon.
86
50
::l£
E40
Q
"~.i-30:~....~-t)g20()
CE...1 10l- ~ A
o I Ii I ii, iii I iii I I I I i I Ii i Ii I i I I .J' I. Iii I i II Ii' . I , I t Io 20 40 60 80 100
Concentration C13, sg
Figure 2. Curves-theory in accordance with our model, (Ll)-experimental data [l].
"2-3
10 100 1000 10000Thickness, d· 10 - 8 m
100000
Figure 3. Phonon boundary scattering, curves 1,2,3,4 for concentration l3C 0%, 0.1%, 1%, 10% correspondingly.
87
4-ClO 600Temperoture. K
200
II
o
5000"";Jj
~ 3-i
~4000~~ j>. ~~ 3000 ~:;:; ~
g ~
§ 2000j I
<.) 3~
v ~
J'~IlL:::","',,,,,,,,'~,--:;:'':::::1"",'--i~- Ii Ii i
500 lCOO
Figure 4. Concentration 13C in diamond fihns 0%; curves 1,2,3 for grit sizes lO~, 1~, 0.1~ correspondingly.
2
1
4000
~ ]
;300°1:;:; ::lg 2000-j
] ~- ~ /
1'00°1 1 3
~ 0 I,/~-,I -,I ,-,i,-:-,I ,-":2,:::::::,,i~I'i~:"" ,o 200 4-00 600 500 i 000
Temperature. K
Figure 5. Concentration 13C in diamond fihns 1%; curves I, 2, 3 for grit sizes 10 ~, 1~, 0.1~ correspondingly.
88
1500
200 400 600Temperature, K
800 1000
Figure 6. Concentration l3c in diamond films 10%; cuIVes 1,2,3 for grit sizes 10 J.U11, 1 Jlm, 0.1 J.U11 correspondingly.
SPECTRAL HOLE-BURNING STUDY OF THE DEFECTSCREATED BY NEUTRON IRRADIATION IN A NATURALDIAMOND
I. SILDOS, G. ZAVT and A. OSVETInstitute ofPhysics, Estonian Acad. Sci.Riia 142, EE2400 Tartu, ESTONIA
Abstract
In this work a neutron-irradiated natural laB-type diamond has beeninvestigated as a prospective material for optical information storagebased on persistent spectral hole-burning phenomenon. The opticalspectra of the sample revealed several new lines both in luminescence andabsorption, peaking at 644, 649, 655 and 723 nm. Two lines at 774 and813 nm with an extralarge inhomogeneous broadening were observed inthe luminescence spectrum. Temperature behaviour of the lines has beeninvestigated and polarized luminescence experimements have been carriedthrough. The hole-burning method has been applied to the spectral lines toinvestigate the homogeneous widths of zero-phonon lines. Spectrallyselective phototransformation of defects at room temperature has beendemonstrated.
1. Introduction
The development of artificial hard materials and wide-band-gapsemiconductors is one of the most significant fields ofmaterial research.Diamond, boron nitride (BN), carbon nitride (C-N), silicon carbide (SiC),aluminium nitride, gallium nitride, etc. show superb mechanical andelectrical properties [1]. Especially diamond due to its unique propertiesand the advances in low-pressure deposition techniques is a perspectivematerial for electronics and optoelectronics. Among the large number of
89
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 89-96© 1995 Kluwer Academic Publishers.
90
appreciable properties are an exceptional thermal conductivity, a largeband gap, low dielectric constant and high charge carrier mobilities areconducive to a number of passive and active electronic applications ofdiamond . Potential applications include electronic devices capable ofoperation at high (>600°C ) temperatures, high powers and highfrequencies [2]. In optoelectronics short-wavelength devices could befabricated on its basis.In this work diamond as a potential material for optical information
storage and processing has been studied. The main idea of the presentinvestigation is to combine the unique properties of diamond and highlyselective spectroscopic method of spectral hole burning (SHB). SHB is amethod in which holes are burnt in inhomogeneously broadened spectrallines by a selective excitation of optical centres, followed byphotochemical (-physical) transformation processes [3]. Optical highdensity information storage could be one of the most attractiveapplications of PSHB. However, the application of the method is greatlyhindered by the absence ofmaterials suitable for room-temperature (or atleast liquid nitrogen temperature) hole burning. Fairly good results havebeen gained by using Sm2+-doped composite crystals ( Sm2+: SrFClo.s;Sm2+:Sro.sM&.sFClo.sBro.s) [4]. High thermostability of the spectral holesburnt at liquid helium temperature (LHeT) has been observed in neutronirradiated sapphire in which the spectral holes can be restored even afterwarming the sample up to 670 K and cycling it back to LHeT [5]. Indiamond hole burning has been performed at low temperatures in thezero-phonon lines (ZPL) belonging to the well-known N-V (peaking at637 om), GR1 (741 om), H4 (496 om) centres [6]. Persistent SHB hasbeen demonstrated in the 774 om luminescence line in the neutronirradiated diamond. In this case an interesting phenomenon has beenobserved: a shift has been detected between the positions of the holes inthe excitation and the luminescence spectrum, which indicates relaxationprocesses in the surroundings of the defect following its optical excitation[7].For a successful application of PSHB a material is needed in which
the spectral lines of optical defects/impurities would be observable at RTand in which selective phototransformation processes are possible. Theexcitation should preferably be a two-step process to realize a photongated hole burning [5]. Also, the ratio of the homogeneous width of theelectron transition to the inhomogeneous width of the spectral line mustbe as small as possible.
91
The prospects of dimond for room-temperature hole burning aregood because of the high Debye temperature which enables theobservation of several ZPL-s of defects and impurities at roomtemperature. At 300 K the crystal remains sufficiently "cold", i.e. fewthermal vibration are activated.We have performed spectroscopic measurements on a piece of
natural laB-type diamond . This complicated crystal could provide aselection of defects with large inhomogeneous distribution. The nitrogenrich sample contains the A (pairs of adjacent substitutional nitrogenatoms), B1 (larger aggregates of nitrogen, presumably composed of fouratoms), B2 (platelets) and N3 defects. Neutron irradiation has been usedto create new optical defects and defect aggregates and, on the otherhand, to increase the inhomogeneous broadening of the spectral lines bycreating disorder in the crystal. It should be mentioned that neutronirradiation causes not only the occurrence of single vacancies but alsovacancy-rich disordered regions with the diameter of 10 - 20 A [8].Annealing at temperatures higher than 6000C causes an aggregation ofradiation defects with impurities (nitrogen), partial recombination ofvacancies and interstitials and the coalescence of disordered regions intolarger ones.
2. Experimental
2.1. SAMPLE AND EXPERIMENTAL SETUP
A piece of natural diamond (linear dimensions ~3 mm ) was cut andpolished to enable optical measurements in the (100) and (110) directions.The crystal was subjected to irradiation with high-energy (> 1 MeV)neutrons at temperature 340 K with a dose of 1-1019 neutrons/cm2.Subsequently the sample was annealed in vacuum at T= 950°C for 30minutes.
In order to determine the contents ofnitrogen the infrared absorptionspecta of the sample were measured by using an Interspectrum PFS-2000Fourier transform spectrometer in the region of 500 - 2500 cm-1•
Although the one-phonon absorption in a perfect diamond is stronglyprohibited by the selection rules associated with the cubic symmetry of thelattice, there are still absorption bands in the one-phonon region of the
92
spectrum induced by impurities. The latter destroy locally the centre ofsymmetry of the diamond structure and permit an electromagneticcoupling to fundamental optical vibrations [9]. The shape and theintensities· of the bands, mainly induced by different nitrogen impurities,allow the determination of the concentration of nitrogen in differentaggregates. The relations between absorption coefficients and the amountof nitrogen [10] have been used and, as a result, the content of nitrogenin the A (pairs) and B1 form (larger aggregates) was found to be 4'1019
cm-3 and 8'1019 cm-3, respectively. The concentration of B2 defects(platelets) is ::::::2'1015 cm-3. The 1366 cm-} infrared absorption lineconnected with the presence of platelets is very sensitive to crystal strains.According to Sobolev [11] its peak position is correlated to thedimensions of the platelets and in our case their diameter is of the order of200 A.Absorption measurements were made by using a halogene
incandescent lamp in combination with a double monochromator DFS 24. An Art-ion-Iaser-pumped Coherent CR-599 single-frequency dyelaser or XeCI-excimer-laser-pumped pulsed dye laser were used for holeburning. Luminescence was excited either with an Art-ion laser or an Artion-laser-pumped dye laser focussed on a spot of ::::::0.5 mm. Theexperiments in the temperature range of 5 K to 300 K were performedwith the sample mounted in the flow ofcold He vapours.
2.2. RESULTS
As a result of irradiation and annealing a large number of spectrallines appeared in the luminescence and absorption spectra, the latter onesbeing situated on top of a strong continuous background (increasingtowards the blue end of the spectrum and not depending on temperature).The most intensive lines at T= 5 K in absorption have peak wavelengths594, 644.1, 649.5, 651.4, 655, 681, 723, and 732 om; in luminescencerelatively strong lines appeared at 594, 649.5, 676, 681, 712, 724, 731.7,733, 774 and 813 om ( excitation at 514 om ands 704 om). Figure 1depicts the low-temperature luminescence in the 710 - 840 om region atthe excitation of 704 om.Temperature dependence measurements show that the 774 and 813
om lines are clearly detectable even at RT without broadening, whichindicates a small homogeneous linewidth of the ZPL compared to the
93
inhomogeneous one. Also, the 681 nm absorption line can be detected atRT. In this case the maximum wavelength is shifted to lower energiesproportionally with T4, which is in agreement with the classical pointdefect model [12].
In order to determine the orientation of the electric dipole momentsof transitions polarized photoluminescence measurements wereundertaken in the spectral region of 720 - 780 nm. An analyses of theexperimental values of polarization allows the determination of the dipoleoientation [13]. Measurements were carried out in the (100) and (110)directions of the crystal and polarization of luminescence was measuredfor different orientations of the E-vector of the exciting laser light.Theoretical values of polarization were calculated for the dipole momentsorientated along the 2-nd, 3-rd and 4-th-order axes in the cubic crystalunder different excitation conditions and the calculated values werecompared with the experimental results. It was concluded that the electricdipole moments of transitions in the defect giving rise to the 723 nmspectral line are orintated along the (111) axis. As for the other lines inthe spectral region under investigation, the orientation could not bereliably established.
T= 7 K
750 800WAVElENGTH. run
Figure 1. Luminescence spectrum ofn-irradiated diamond at 7 Kexcitation at 704 nIn.
Persistent spectral hole burning is possible in the lines peaking at649.5, 655, 681, 723, 731.7, 734, 774 and 813 nm. Spectral holes burnt
94
at 5 K in the 655 om line appeared to be very thennostable and wereobservable even at 200 K. In the 649.5 and the 681 om line the holes werepartly detectable on wanning the sample up to 100 K . After cycling thetemperature back to "5 K the holes were restored with the same width asbefore cycling but a 50% decrease ofthe depth was observed.A special case is the 774 and 813 om line the intensity of which
depends weakly on temperature. The unusually broad inhomogeneouswidths ( 6 om) of these lines exceed the homogeneous ones by 4 timeseven at room temperature. Spectral hole burning is perfonnable in theselines at RT, but the holes are broad (2 om) and burning efficiency is quitelow (30 minutes of irradiation with the intensity of8 W/cm2 yielded 15%deep holes).
on0.2
OJ
b
645 650WAVElENGTH, nrn
T= 7 K
"Figure 2. Absorption spectrum of n-irradiated diamond; (a) the initialspectrum. (b) the spectrum after burning with cw laser at 648.3 and 649.2nIn. Burning conditions: intensity 2 W/em? , laser linewidth 0.3 em- l andduration 2 min.
The homogeneous widths of the ZPL-s at LHeT were estimated fromthe SHB experiments with narrow-band (2 MHz) cw dye laser. In case ofthe 649.5 and 681 om lines the approximation to zero doses yielded thehomogeneous widths of 0.7 and 2.0 GHz, respectively. These valuescorrespond to the excited state lifetimes of215 and 80 ps, which indicatesthe presence of radiationless channels of relaxation. The narrowestholewidths in the 774 and 813 om lines at LHeT were 5 GHz.
95
An investigation of the phototransformation processes in the groupof lines around 650 run revealed photoproducts locating at 644.1 and651.4 run . A light-induced defect site distortion could be an explanationfor the mechanism of hole burning. Figure 2 shows the low-temperatureholes in 649.5 run line and the location ofphotoproduct.Photochromic properties of the luminescence line at 731.7 run were
used to cast light on the connection between the lines in the 720 -850 runregion. The reduction of the intensity of 731.7 run line caused aproportional decrease of the 774 run and 813 run line intensities. Wepropose that these lines are originated by one and the same defect.
3. Summary
An irradiation ofa natural laB-type diamond crystal containing nitrogen indifferent aggregates caused the appearance ofa number of spectral lines inthe visible and the near infra-red spectral region. These lines with theexception of the well-known 594 run line (nitrogen + vacancy) and 681and 724 run lines, which have been mentioned earlier are completelyunidentified. The data obtained from the spectroscopic measurements arenot sufficient to determine the nature of the defects, yet it is supposedthat they are combinations of nitrogen and radiation defects formed in thecourse ofannealing.Through neutron irradiation and the subsequent annealing a material
has been created which possesses useful properties from the point ofviewof SHB. Fairly broad spectral lines are present at room temperature, andthere are mechanisms for PSHB (e.g. electron traps are created byradiation). Still ways have to be found for further increasing of theinhomogeneous widths of the spectral lines and investigations have to becarried out to gain knowledge ofthe structure of the defects.
References
1. Edgar, 1. H. (1991) Prospects for device implementation ofwide band gapsemiconductors, J. Mater. Res. 7,235 - 252.
2. Ravi, K. V. (1992) Low pressure diamond synthesis for electronic applications,Materials Science and Engineering B19, 203 - 227.
3. Moemer, W.E. (00) (1988) Persistent Spectral Hole Burning:Science andApplications, Springer, Berlin.
96
4. Jaaoiso, R, Hagemann, H., Kobel, F. and Bill, H. (1992) Memebers of thePbFCI-Type Family: Possible Candidates for Room-Temperature PhotochemicalHole Burning, Chimimia 46, 133 - 137.
5. Aizengendler, M., Bogner, U., Dolindo, I., Kikas, 1. and Sildos, I. (1991) Photon-gated thermoresistant spectral hole burning in a neutron-irradiatedsapphire, Chem. Phys. Lett. 183,245.
6. Harley, R T., Henderson, M. 1. and Macfarlane, R M. (1984) Persistent spectralhole burning ofcolour centres in diamond, J. Phys. C: Solid State Phys. 17, L233 L236.
7. Sildos, I. and Osvet, A (1994) Spectral hole burning study ofa neutron-irradiatedtype laB natural diamond, Diamond and RelatedMaterials 3, 725 - 727.
8. Morelli, D. T., Perry, T. A, Vandersande, 1. W. and Uher, C. (1993) GI~slike
thermal transport in heavily irradiated diamond, Phys. Rev. B48, 3038 - 3041.9. Lax, M. and Burstein, E. (1955) Infrared Lattice Absorption in Ionic andHomopolar Crystals, Phys. Rev. 97,39 - 52.
10.Woods, G. S., Purser, G. C. and Collins, A T. (1990) The Nitrogen Content ofType Ia Natural Diamond, J. Phys. Chem. Solids 51, 1191 - 1197.
11. Sobolev, E. V., Lisoivan, V. I. and Lenskaya, S. V. (1968) Zhumal StruktumoiKhim;; (in russian) 9, 1029.
12. Rebane, K. K. (1970) Impurity spectra ofsolids, Plenum Press, New York.13.Feofilov, P. P. (1954) Zh. Eksp. Teor. Fiz.(in russian) 26, 609 - 623.
CALCULATIONS OF PHOSPHOROUS ELECTRONIC LEVELS IN DIAMOND
VALENTINE V. TOKIY AND DIANA L. SAVINAResearch Group ofDiarrwnd Electronics and Coatings,Donbass Institute ofConstructions Engineering of,339023 Makeyevka, Ukraine
The present paper is dedicated to the simulation of phosphorus entering into diamondand its influence upon the vacancy in diamond, using the theory of shallow donor statesand the tight-binding theory (TET).
Comparison of the experimental and theoretical data on ESR of the paramagnetic centers in diamond has been made.Phosphorus forms neither shallow nor deep levels in thediamond band gap. Its levels are in the conduction band of diamond.
ESR (paramagnetism) of phosphorus in diamond connected with unpaired electron inbroken bond of nearest neighbors of phosphorus.
1. Introduction
The problem of describing the electronic states of defects and impurities in diamond isone of great importance, both from a fundamental and practical point of view. Thedevelopment of an electronic technology based on diamond requires solving the problem of reproducible doping by donor and acceptor impurities to obtain p-n junctions,which form an essential part of many devices. It is well known that p-type conductivitycan easily be obtained. Obtaining n-type conductivity is a more difficult problem [1,2].Active electronics is in need of n-type semiconductor diamond. Both the mechanism offormation of different defects and their properties in diamond have been unknown untilnow. Both these aspects of the problem of defects in diamond are closely related, andsuccess in one field enables success in the other one. Recently, the superfine structure ofimpurity of impure phosphorus was studied by [3-5]. In particular, the phosphorus spectrum exhibits a doublet of low intensity, but with clear lines on the edges of the centralcomponent of the nitric triplet, and has the following parameters: g = 2.0025+-0.0002,Apar = 23.2+-0.2 Gs, Aper = 19.6+-0.2 Gs. But the position of the phosphorus atom inthe diamond lattice is obscure.
The present paper is dedicated to the simulation of phosphorus entering into diamond
97
M.A. Prelas et at. (eds.), Wide Band Gap Electronic Materials. 97-103© 1995 Kluwer Academic Publishers.
98
and its influence upon the vacancy in diamond, using the theory of shallow donor statesand the tight-binding theory (TB1).
Usually the calculation to predict the structure is possible with several methods in whichthe structure is determined as that having the lowest energy (without any experiment).The possible comparison of the sole calculation result with experiment allow to determine quality of calculation method in that case. The purpose of the semiempirical calculation of present work is to determine the structure by the ESR. The necessary compariof a number of calculation results with experiment allow to determine the position of theparamagnetic phosphorus in the diamond.
The impurity center which results when a carbon atom is replaced by an element fromthe fifth column of the periodic table (e. g. nitrogen, phosphorus, arsenic, antimony) iscalled a donor. Four of the valence electron of this impurity center form covalent bondswith their nearest carbon neighbors and therefor do not exhibit a paramagnetic resonance. It is the fifth electron that we will be concerned with in this paper. We will startby reviewing, briefly, the results of the theory of shallow donor states in diamond lattice.
2. The Shallow Donor (the hydrogen-like model)
In this model we suppose that fifth electron moves in a Coulomb-like potential of thedonor nucleus to which it is only loosely bound.
2.1. ENERGY LEVELS
Energy levels of a donor electron can be obtained by solving an effective mass Schrodinger equation for a hydrogen-like model
En = Ec - [13.5*(met)]/ [(eps)*(eps)*m*n*n] eVwhere m is mass of electron, (met) is the effective mass, (eps) is the dielectric constantof diamond, n =1,2... For the ground state (at n =1) of donor electron we obtain
E1 = Ec - [13.5*(met)]/ [(eps)*(eps)*m] eVIt is approximately 0,08 eV below the conduction-band edge, when mef = 0,2 m and eps= 5.7 for diamond.
2.2. HYPERFINE INTERACfION
The superfine structure of the ESR spectra is connected to the interaction of unpairedelectrons of the donor atoms with the magnetic moment of the nucleus. Isotropic interactions, which are Fermi contact interactions, are characterized by the constant A. Thisparameter is a function of the density of the unpaired electron at the nucleus of phosphorus with the nuclear spin J=l/2. Which is function of the multiplicity of the conductionband edge, the atom s volume, and the nuclear charge [3]. Therefore, for phosphorus asshallow donor (in hydrogen-like model) the isotropic constant Ashd is approximately 35
99
Gs and anisotropic constant Bshd = 0 because of spherical symmetry wave function.
2.3. CONCLUSION
Contradiction between experimental (A = (Apar + 2Aper) / 3 = 20.8 Gs and B = (AparAper)/3 = 1.2 Gs) and theoretical results (Ashd and Bshd) leads to conclusion, that inconsider case phosphorus does not form a shallow donor in the diamond band gap.
3. TBT Calculation in Absence of Defect
3.1. THE TIGHT-BINDING CLUSTER MODEL
We use the tight-binding theory [6,7]. In this theory of SP-bonded systems the electronic eigenstates are written in terms of a basis set consisting of a single S state andthree P states on each atom. The corresponding one-electron eigenvalues and eigenfunctions are then obtained by diagonalizing an N*N Hamiltonian matrix based uponthese N orbitals.
In the given work, the basis system is limited by SP3-hybrydized orbitals made of oneS-orbital and three Px-, Py- and pz- orbitals of the valence shell of each atom. These are2s- and 2p- atom wave functions of carbon. We utilize parameters consisting values fordiagonal terms (eps)s= -19.5 eV, (eps)p= -10.7 eV and forms Vss,sg= -5.1 eV Vsp,sg=2.55 eV, VPP,sg= 5.1 eV, Vpp,pi=-1.7 eV for interactions between nearest-neighbororbitals from [8]. The first two subscripts indicate the orbital coupled and the last indicates the component of angular momentum around the internuclear axis.
Our consideration is based upon simulation of the diamond crystal by means of a largegroup of atoms (cluster). For calculations, we have used a group of thirty-five carbonatoms, inclusive the central unit (0,0,0) plus all its first-(1,l,l), second-(2,2,0), third(3,1,1) and fourth-(4,O,0) nearest neighbors in diamond lattice. During simulation thephosphorus impurity is placed in omit the cluster's central unit, and, if the central unit isvacant, as the first-, the second- and the third-nearest neighbors to the vacancy, in succession.
3.2. EFFECT OF CLUSTER SIZE
In Fig.1 (1-5) we show the calculated molecular-orbital energies as a function of the sizeof the cluster, starting from a central carbon atom and then adding successive shells. Wesee immediately that as the size of the cluster increases, the "band structure" begins toemerge, with the bonding MO's grouped into what can be identified as a valence band,the antibonding MO's forming a conduction band, and with a forbidden gap between.The width of the valence band (measured as the difference between the highest and lowest energy states) is 18.1 eV, which agrees well with the calculated band structure value
100
of 20.7 eV [9]. The band gap is 8.2 eV, which is to be compared with the experimentallyobserved value of 5.5 eV and theoretically calculated value of 9.5 eV [9] for the 35atom cluster. However, the calculated band gap decreases, as the size of the cluster isincreased suggesting that the agreement here is also satisfactory.
4. Substitutional Phosphorus in Diamond
4.1. ENERGY LEVELS OF PHOSPHORUS IN DIAMOND
In Fig.1 (6-10) and Fig.2 (1,2) we show the one-electron molecular- orbital energies thatresult from the calculation when the central atom of the cluster is replaced by phosphorus (ionization potentials of 18.7 eV and 10.3 eV for 3S- and 3P- orbitals, respectively[8]). In Fig.1 (6-10) we show energies as a function of the size of the cluster, startingfrom a central phosphorus atom (10) and then adding successive shells of carbon atoms(9-6).
E eUl
-1.00
-14.00
-21.00
-28.00
-=
-----
-3S.00 L----Z--J--q---==S=--=-(;-1 8 9 10
Figurel. Energies of the one-electron MO's for a diamond cluster as a function of cluster size, startingwith central atom (1) and adding successive shells of neighbors up to and including the first(2), second (3), third (4) and fourth (5)- nearest neighbor shell. Shown are clusters containing acentral phosphorus (10) and (9) one, (8) two, (1) three, and (6) four neighbor shells. Thus itshows that (1) is C, (2) is 5C, (3) is 17C, (4) is 29C, (5) is 35C, (6) is P+ 34C, (7) is P+ 28C, (8)is P+ 16C, (9) is P+ 4C and (10) is P.
101
E eV
--7.20
~
(.02)
<'00) <.03) <.06) <.OG>
<.E!:I - <."') (.07)-7.30 <'00)
<.00)
<.t~)
-7.40
<'00)-1.50
<.00)
-7.(,031 2 4 5 (, 7 8
Figure 2. States near the bottom of the conduction-band edge calculated for diamond clusters. Shownare clusters containing a central phosphorus (1,2) a central unit vacancy (3-6) and (2-7) three (1-8) fourneighbor shells. The phosphorus impurity is placed in the first-(3). the second-(4) and the third-(5) nearest neighbors to the vacancy. Thus it shows that (1) is P+34C. (2) is V+27C+PI. (4) is V+27C+PIl. (5) isV+27C+Pill. (6) is V+28C, (7) is 29C and (8) is 35C.The numbers shown in-brackets the sum of themolecular wave-function coefficients on phosphorus and serve as wmeasure of the localization of thestate.
From Fig. 1 (6-10) we see that triply degenerate orbital appears in the forbidden gap onlyfor 17-atom cluster (8). For greater clusters (6,7) a donor nondegenerate state of spherical symmetry is in the conduction band above the forbidden gap. It is approximately 0.2eV above the conduction band edge of the 35-atom cluster, see Fig.2 (1,2). The molecular-wave-function coefficients reveal that it is highly localized on the phosphorus and itsimmediate carbon neighbors.
4.2. HYPERFINE INTERACTION ON PHOSPHORUS IN DIAMOND
Electron-spin resonance (ESR) is one of the most powerful tools for microscopic identification of the paramagnetic centers in diamond, particularly when the nuclei P onwhich the unpaired spin resides have naturally magnetic isotope.
The diagonal elements of the hyperfine tensor are written asApar =RsA* + 2RpB*Aper =RsA* - RpB*
102
Where: A* and B* are the constants of the hyperfine interaction for the free phosphorusatom P which usually cannot be identified experimentally but can be calculated with asufficient accuracy from the wave functions of the free phosphorus atom. In the givenwork we have used A*= 3640 Gs and B*= 103 Gs from [10]. Rs - the probability of theunpaired electron residence on the orbital 3S of the carbon P. Rp - the probability of theunpaired electron residence on the 3P orbital of the carbon P.
Where we assume also that the defect wave function can be written in the form of thesum of the items of 2S- and 2P-orbitals of each atom of carbon and separately of phosphorus P. By diagonalizing the Hamiltonian matrix and by finding coefficients of thewave function, using the formula (3) and (4), we can find Apar and Aper. The results ofcalculation for 29-atom cluster are shown in Table 1.
4.3. CONCLUSION
Comparing the calculation for central atom of phosphorus with experimental values inTable 1 we may conclude that fifth electron of the phosphorus is not localized on thephosphorus.
S. Phosphorus as Neighbor to the Vacancy
5.1. ENERGY LEVELS AND HYPERFINE PARAMETERS
In Fig.2 (3-8) we show the one-electron-molecular-orbital energies near the bottom ofthe conduction-band edge that result from the calculation when one simply removes thecentral atom of the 29-atom cluster and then one of first (3) or second or third (5) nearestneighbors of vacancy is replaced by phosphorus.
TABLE 1. Hyperfine parameters and wave function coefficients of phosphorus in diamond.
System P site Rs Rp Apar (Gs) Aper(Gs)
28C+P (0,0,0) 0.236 0.000 858 858
V+27C+P 0,1,1) 0.116 0.570 538 362
V+27C+P (2,2,0) 0.005 0.021 24.2 17.8
V+27C+P (3,1,1) 0.000 0.014 2.9 -1.5
experiment 0.006 0.012 23.2 19.6
103
5.2. CONCLUSIONS
Comparing calculation with experimental values in Table 1, we may conclude that paramagnetic phosphorus is second-nearest neighbor to the vacancy in diamond.
6. General Conclusions
Phosphorus forms neither shallow nor deep levels in the diamond band gap. Its levelsare in the conduction band of diamond ESR (paramagnetism) of phosphorus in diamondconnected with unpaired electron in broken bond of nearest neighbors of phosphorus.
7. References
1. Popovici G. and Prelas M.A.(1994) Forced methods of diamond doping. Second Int. Symp. on DiamondFilms, May 3-5,1994, Minsk, Belarus, 16.
2. Spitsyn B. (1994) Chemical problems of diamond doping. NAlD Advanced Workshop on Wide BandgapElectronic Materials,May 4-6, Minsk, Belarus, 164.
3. Samsonenko N.,Tokiy V.,Gorban S.and Tunchenko V. (1991) Electron spin resonance spectroscopy ofimpure and pure structural defects in diamond.Surf.and Coat.Technol. 47,618-622.
4. Samsonenko N.,Tokiy V. and Gorban S. (1991) Electron paramagnetic resonance of phosphorus in diamond. Sov.Phys. Solid State. 33,1409-1411.
5. Tokiy V.,Samsonenko N. and Gorban S. (1991) The electronic structure of phosphorus in diamond.AbstrXm AIRAPT Int. Conf.on High Science and Technol. ,Bangalor, p.A-10.
6. Harrison W.A.,(1989) Electronic Structure and the Properties of Solids ,Freeman,San Francisco,1980,re-printed by Dover, New York, 1989.
7. Harrison W.A. (1990) Interatomic interactions covalent and ionic solids. Phys.Rev., B 41, 6008-6019.8. Levin A. (1974) Solid State Quantum Chemistry, McGraw-Hill, New York.9. Messmer, R.P. and Watkins, G.D. (1973) Molecular Orbital Treatment for Deep Levels in Semiconductors:Substitutional Nitrogen and the Lattice Vacancy in Diamond. Phys.Rev. B7, 2568-2590.
10. Atkins P.W., Symons M.c.R. (1967) The Structure of Inorganic Radicals, Elsevier, Amsterdam.
HYDROGEN CHEMISTRY ON DIAMOND SURFACES
James E. Butler, Brian D. Thoms, Marianne McGonigal,John N. Russell, Jr., and Pehr E. PehrssonGas/Surface Dynamics Section, Code 6174Naval Research Laboratory, Washington, DC 20375 USA
1. Abstract
Surface chemical reactions of hydrogen with diamond are importantevents in the complex processes leading to diamond chemical vapor deposition(CVD). Our ongoing research addresses the surface reactions related todiamond growth. Studies were performed on single crystals, C(100) andCOlO), and large grained polycrystalline CVD diamond films using a diversearray ofultra-high vacuum surface analytical techniques. Surface vibrationalspectroscopies, high resolution electron energy loss spectroscopy (HREELS)and multiple internal reflection infrared spectroscopy (MIRIRS), wereprincipal probes of surface structure, while HREELS and temperatureprogrammed desorption (TPD) were used to investigate hydrogen adsorption,abstraction, and desorption. We found: 0) the COOO), C(llO) andpolycrystalline diamond surfaces are terminated primarily with CH species;(2) the ratio of H atom abstraction to adsorption is 0.05 ± 0.01; (3) hydrogendesorbs molecularly from C(llO) peaking at 892 C and exhibits first orderdesorption kinetics with an activation energy of 75 kcaVmole; and (4) bothclean and hydrogen-saturated C(100) surfaces display a 2x1 surfacereconstruction.
2. Introduction
The chemical vapor deposition (CVD) of diamond at low temperaturesand pressures raises many scientific questions and provides the opportunityfor diverse technological applications. Recent reviews [1-4], books [5,6], andconference proceedings [7] offer excellent summaries of the reasons forinterest in the CVD of diamond. Molecular and atomic hydrogen [8] in thegrowth environment are important for diamond deposition. Many roles havebeen proposed for hydrogen atoms in CVD diamond film growth [9,10]. Forexample, gas phase hydrogen atoms produce condensable carbon radicals byreaction with hydrocarbons [4]. Hydrogen atoms incident on the surface
The U.S. Governmental right to retain a non-exclusive, royalty free licence in and to anycopyright is acknowledged.
105
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 105··114© 1995 Kluwer Academic Publishers.
106
abstract hydrogen to produce vacant sites [3] and surface radicals [10-12],refill vacant sites by adsorption [10,13), and etch surface graphite [14].Adsorbed hydrogen stabilizes the diamond surface structure [9,15]. Hence,the hydrogen-diamond system has received considerable attention.
We used multiple techniques to investigate the structure and bondingof bare and hydrogen-terminated diamond surfaces. Hydrogen adsorption,desorption, and abstraction, all important processes during diamond growth,were examined. In this paper, several of these experiments [16-20] aredescribed and a broader picture of the role of hydrogen on the diamondsurface is developed.
3. Experimental Description
The experimental apparatus and procedures are described only brieflysince detailed descriptions of individual chambers, sample mounts,temperature control and measurement, gas dosing, and sample preparationand characterization are reported elsewhere [16-20].
All experiments were performed in stainless steel ultra-high vacuum(UHV) chambers with base pressures ofless than 5 x 10-10 Torr. The HREELspectra were acquired with a 1270 cylindrical monochromator and analyzer(LK Instruments, Model LK2000). The incident electron beam had an energybetween 10 and 12 eV and a direction 60· from the surface normal. Spectrawere collected in the specular direction. The MIRIR spectra were collectedwith a Nicolet 800 FTIR spectrometer using an InSb detector. The singlebeam infrared spectra for the data reported here are an average of 4000 scanstaken at 8 cm'l resolution. The samples were maintained near roomtemperature during acquisition of all HREEL and MIRIR spectra. A randomflux shielded (2 mm diameter sampling aperture), differentially pumped,computer controlled, UTI quadrupole mass spectrometer (QMS) was used formonitoring line-of-site thermal desorption of up to 10 different masses fromthe C(llO) crystal surface as a function of surface temperature.
Atomic hydrogen (deuterium) was exposed to the surface byintroducing molecular hydrogen (deuterium) into the chamber through a leakvalve and positioning the sample in line-of-sight of a W wire resistivelyheated between 1400-1800 C. While the flux of atomic hydrogen is unknown,it is assumed to be proportional to the measured molecular flux of hydrogen.Thus, all exposures (atomic and molecular) were measured in Langmuirs (L)of molecular hydrogen, where 1 L = 10'6 Torr sec. For simplicity, we refer tohydrogen exposures in the presence of a hot filament as H (D) or atomichydrogen (deuterium) exposures and exposures without a hot filament as H 2
(D2) or molecular hydrogen (deuterium) exposures. Absolute inter-apparatusexposure comparisons are not possible due to dosing geometry and filamenttemperature effects. However, exposures within a given experiment weredirectly comparable since the dosing geometry and filament temperature were
107
not changed.Four different diamond samples were studied. A free-standing B-doped
polycrystalline diamond film, 7 mm x 7 mm x 0.25 mm [21a], and a B-dopedhigh pressure, high temperature grown synthetic crystal with a 3 mm x 3mm C(100) face [21b], were used in HREELS studies. Scanning' electronmicroscopy (SEM) showed the polycrystalline sample to have numerous 10-50f.Lm triangular facets and Raman spectroscopy revealed it to be high qualitydiamond. A 3.88mm x 4.75mm x 1.75mm type 1A natural diamond crystal[21c], polished to within 20 of the (110) face, was used in TPD studies. Forthe MIRIR experiments, a trapezoidal shaped, natural, type IIa stone,measuring 11mm x 5.5 mm x 0.71 mm [21d], yielded 15 internal reflectionsat 45 0
• The large parallel faces were oriented approximately 70 off of the(110) plane. The entrance and exit faces were nominally of (100) surfaceorientation. Each sample was boiled for 15 minutes in 3 HCl: 1 HNOa and for15 minutes in 3 H2S04 : 2 HNOa followed by rinses in water, and/or organicsolvents prior to vacuum insertion. This treatment resulted in an oxygensurface termination which was removed by heating in vacuum to >1000 C.The samples were examined with Auger electron (AES), energy loss (ELS),and/or x-ray photoelectron spectroscopy (XPS) to verify the surfaces were freeof Sp2 bonded (graphitic) carbon and other surface contaminants.
4. Results
4.1. STRUCTURE AND BONDING OF H TERMINATED DIAMONDSURFACES.
The chemical structure of the hydrogen-terminated polycrystallinediamond, C(100), and C(110) surfaces was investigated using HREELS andMIRIRS [16-18]. Figure 1 shows several HREEL spectra of the polycrystalline diamond surface [16]. The top one was acquired after heating for3 minutes to 1050 C. It has a weak feature at 2920 cm'l and somewhat largerfeatures at 720 and 1220 cm'l. After exposing the room temperature surfaceto several thousand L ofmolecular hydrogen, no changes were observed in theHREEL spectrum. However, peaks at 1220 and 2920 cm'l increased inintensity and shifted to 1250 and 2900 cm'I, respectively, upon exposure toatomic hydrogen, as seen the middle spectrum. In addition, a broad featureappeared near 2400 cm'l and the intensity of the peak at 720 cm'l decreased.The HREELS spectrum stopped changing after about a 200 L H exposure.Once the sample was heated above 900 C, the HREELS spectra showedreduced intensity in the features at 1250, 2400, and 2900 cm'l and a peakappeared at 720 cm'l after heating above 1000 C. After heating to 1050 C,HREELS showed peaks at 720, 1220 and 2920 cm'I, similar to the top spectrain Figure 1. A HREEL spectrum was also acquired after a saturation Dexposure and displays new peaks at 900 and 2140 cm'I, as shown in the
108
102
2920 o exposure
IHeated c:-oto '0;
1050 ·C '"'E 101.9.~ '"<:
'" l!!<:CD f-]; C'C Iiell.~
..,iii rr.u
E ....." 101.80 Iz
rr.
101.7
3050 2950 2850 27500 1000 2000 3000 Frequency (em-I)
Energy Loss (em'l)
Figure 1. HREEL spectra from B-dopedpolycrystalline diamond after heating to 1050C, exposing to H atoms, and exposing to Datoms.
Figure 2. Comparison ofMIRIR spectra from Dand H saturation coverages on C(llO).
lowest spectrum of Figure 1.A sharp two domain 2x1 pattern was observed in low energy electron
diffraction of a hydrogen-saturated C(100) surface [17]. Spots were visible atincident electron energies as low as 30 eV, indicating a smooth, well orderedreconstructed surface. In addition, HREELS reveals vibrational peaks at2915 and 1250 cm-\ similar to the spectrum from hydrogenatedpolycrystalline diamond. Mter heating to near 1100 C, a sharp two domain2x1 LEED pattern was observed again. However, the intensity of the peakat 2915 cm- l was substantially decreased.
Hydrogen on C(llO) was examined with MIRIRS [18]. A clean(reference) surface was prepared by heating the crystal to 1000 C anddesorbing the hydrogen surface termination. The crystal was subsequentlyheld at 400 C and exposed to 1050 t H, a saturation exposure. The resultinginfrared spectrum, taken at room temperature, is shown in Figure 2. Asymmetric, broad peak (FWHM - 75 cm-l ) at 2880 cm- l was observed. Thisfeature disappeared from the infrared spectrum after heating the crystal tobetween 800 and 1000 C. Exposing the clean surface to either H2 or a hot
109
filament without any hydrogen did not produce a feature at 2880 cm-l .
Exposure to 4000 L of atomic deuterium resulted in a spectrum, also shownin Figure 2, which has no feature at 2880 cm-\ as expected. Unfortunately,no modes could be observed in the range of 1800-2600 cm-l since the 2-phononmode overlaps that part of the spectrum.
4.2. HYDROGEN SURFACE REACTION PROCESSES: ADSORPTION,ABSTRACTION AND DESORPTION.
In order to investigate the rates of hydrogen abstraction andadsorption, two important reactions during diamond growth, a series ofHREEL spectra were acquired from a polycrystalline diamond surface as afunction of deuterium exposure [19]. Exposure of a hydrogen-saturatedpolycrystalline diamond surface to atomic deuterium decreased the intensityof the CH stretching vibration and increased the intensity of the CDstretching vibration. These changes were interpreted as due to abstractionof surface hydrogen by atomic deuterium. Exposure of an annealed surfaceto atomic deuterium produced increasing intensity of the CD stretchingvibration. This was interpreted as indicating adsorption of atomic deuteriumonto a bare diamond surface.
To quantify the rates of adsorption and abstraction on this surface,peaks in HREEL spectra were normalized to the intensity of the elastic peak,fit to gaussian lineshapes with a linear background, and the area under thegaussian peak was integrated. Figure 3 is a plot of the integrated intensityof the CD stretching vibration as a function of exposure to D at surfacetemperatures of 80 and 600 C. Similarly, Figure 4 contains the integratedintensity of the CH stretching vibration versus exposure to D for anabstraction series at surface temperatures of 80 and 600 C. The data are fitassuming first order dependence of the reaction rates on the populations ofreactive sites and on a constant atom flux for both adsorption and abstraction.These fits yield a ratio of abstraction probability to adsorption probability,PaJPad' of 0.05 ± 0.01 at both 80 and 600 C.
TPD was used to determine the desorbing species, the temperaturerange for the desorption, and the kinetics of the desorption process fromC(llO) [20]. In Figure 5, Dz TPD spectra are shown for a series of Dexposures to C(llO) at room temperature. The desorption spectra, andconsequently the surface D coverage, saturate at -lOOL exposures. Moleculardeuterium desorbs between 227 C and 1027 C with an asymmetic desorptionpeak and a desorption peak maximum at 892 C which is independent ofdeuterium coverage. No Dz desorption is observed after exposing the surfaceto Dz.
110
b
70 ,----,----.--,----.-----,
<>
SIu1ace Temperatyre
<> 80 "c• 600't;
60
Hydrogen Abstraction
10
U>§ 50
.e~~ 40.~
~.<: 30
IG20
Surface Temperatyre
<> 80°C
• 600°C
Deuterium Adsorption
a50 ,---,--,--,--,---,-------,
40
10
200 400 600 800 1000
D2
exposure (ll
0'----'-_--'-__.1.-_---'-_---'o10 20 30 40 50 60
D2 exposure (ll
0'--_1--_1--_.1.-_.1.-_.1.------'o
Figure 3. Deuterium adsorption onto an Figure 4. Surface hydrogen abstraction by'hydrogen free' surface. atomic deuterium.
5. Discussion
5.1. STRUCTURE AND BONDING OF H-TERMINATED DIAMONDSURFACES.
Peaks at 1250, 2400 and 2900 cm-1 in HREELS from a H-exposedpolycrystalline diamond surface are assigned to CH bending, bendingovertone, and stretching vibrations from monohydride terminated diamond,respectively [16]. The vibrational peaks at 900 and 2140 cm-I, seen after Dexposure, are assigned to the bending and stretching vibrations of CD,respectively [16]. Overall, HREELS spectra in Fig. 1 reveal only twofundamental vibrational modes for hydrogen on a polycrystalline diamondsurface and an absence of modes in the range of 1350-1450 cm-I, suggestingthat CH2 or CHa species are not present after exposure to H at roomtemperature. Similarly, HREEL spectra from H-exposed COOO) indicate CHspecies and not CH2 or CHa are present [17]. The observation of a two
111
8
Figure 5. Temperature programmed desorptionspectra ofD2 from D terminated C(llO).
domain 2xl LEED pattern andvibrational spectra consistent withmonohydride termination togetherimply that C(100) undergoes amonohydride dimer rowreconstruction when hydrogensaturated. The reduced intensityof CH vibrations and thepersistence of a two domain 2xlLEED pattern following annealingis consistent with a reconstructionof C(lOO) to rows of pi-bondeddimers following removal ofhydrogen. On the C(llO) surface,the 2880 cm-1 absorption band isassigned to a C-H stretching modeon an Sp3 hybridized surfacecarbon [18]. The ideal terminationof C(llO) with H should bemonohydride and the presence of asingle symmetric peak in the CHstretching region suggests thatthis is the case. Overall, HREELSand MIRIRS indicatesmonohydride termination of Hexposed polycrystalline diamond,C(lOO), and C(llO).
85
60
30
25
6
400
200
Exposure(LaID,)
Temperature (DC)
T"",=25°CdT/dt = 2 °Cls
o 200 400 600 800 1000 1200
o
5.2. HYDROGEN SURFACE REACTION PROCESSES: ADSORPTION,ABSTRACTION AND DESORPTION.
The HREELS, MIRIRS, and TPD data consistently show H/D2) doesnot dissociatively chemisorb on bare diamond surfaces [16-18,20]. Thedissociative chemisorption process is highly activated. In contrast, atomic Hreadily reacts with the bare surface, resulting in a hydrogen-terminatedsurface at room temperature.
In the case of hydrogen abstractions from hydrocarbons, activationbarriers of -5-7 kcaVmol have been determined [22]. Recent calculations byBrenner et at. [23] for H abstraction on C(lll) indicate an activation barrierof approximately 10 kcaVmol. They calculate adsorption and abstractionprobabilities per collision per site at 1200 K of 0.44 and 0.028, respectively,yielding a value for PaJPad of 0.064 [23], in good agreement with the value of0.05 ± 0.01 measured in this work. The lack of any significant surface
112
temperature dependence between 80 and 600 C for either adsorption orabstraction indicates that the H atoms accommodate very poorly with thesurface. This suggests the process follows a generalized Eley-Ridealmechanism [24], where the gas phase reactant does not equilibrate with thesurface prior to entering the transition state for the reaction.
The intensity of HREELS peaks assigned to CH bending andstretching vibrational modes decreased between 950 and 1000 C onpolycrystalline diamond and below 1100 C on C(lOO), indicating a reductionin surface hydrogen coverage. A reduced absorption for the C-H stretchingmode was observed for MIRIRS on C(110) upon heating the crystal to between800 and 1000 C. The TPD results indicate molecular hydrogen(deuterium)desorbs from C(110) up to 1037 C with a desorption peak maximum at 892 C.A common temperature range is observed for hydrogen desorption fromvarious diamond surfaces using various techniques. In addition, theasymmetry and the coverage independence of the D2 desorption peakmaximum are characteristic of desorption kinetics which are first order inadsorbate concentration [25]. The appearance of first order recombinativedesorption kinetics implies the rate-limiting step for the overall measuredkinetics is not recombinative desorption, but another step in the sequence ofevents leading to the desorption of the molecule. This rate-limiting step couldbe the generation of the reactive site, the mobility of the surface species, oranother event which is linearly dependent on the surface hydrogenconcentration. By assuming a kinetic prefactor of 1 x 1013 /sec [25], wecalculate the activation energy of the overall desorption process is about 75kcal/mole. First order desorption kinetics were also observed for molecularhydrogen desorption from C(lOO), and polycrystalline diamond films [26].
6. Summary
This paper elucidates several aspects of the interaction of hydrogenwith diamond surfaces. The bare diamond surface is unreactive to molecularhydrogen. HREELS, MIRIRS, and TPD indicate atomic hydrogen reacts withthe diamond surface. Molecular hydrogen desorbs from the diamond surfacefollowing first order desorption kinetics with an activation energy of about 75kcal/mole. The diamond surface is principally terminated by CH species.Hydrogen atoms can abstract adsorbed H with an abstraction to adsorptionefficiency of 0.05 ± 0.01.
7. Acknowledgements
The authors wish to thank Dan Vestyck for performing micro-Raman,and Prof. Sellschop (University of Witswatersrand), Prof. Maguire (TrentPolytechnic), Dr. Cliff Spiro (General Electric Co.), and Dr. Richard Woodin
113
(Norton Diamond Films) for providing the samples used in these studies. Thesupport of ONR and ARPA are gratefully acknowledged.
8. References
1. DeVries, R C. (1987) Synthesis ofdiamond under metastable conditions,Ann. Rev. Mat.Sci. 17, 161-187.
2. Angus, J. C. and Hayman, C. C. (1988) Low pressure, metastable growth of diamondand "diamondlike" phases, Science 241, 913-921.
3. Spear, K. E. (1989) Diamond - ceraminc coating of the future, J. Am. Ceram. Soc. 72,171-191.
4. Celii, F. G. and Butler, J. E. (1991) Diamond chemical vapor deposition, Ann. Rev. Phys.Chem. 42, 643-684.
5. Davis, R F. (ed.) (1993) Diamond Films and Coatings: Development, Properties, andApplications, Noyes Publications, Park Ridge, NJ.
6. Spear, K. E. and Dismukes, J. P. (eds.), (1993) Synthetic Diamond: Emerging CVDScience and Technology, John Wiley, New York.
7.a. Messier, R, Glass, J. T., Butler, J. E. and Roy, R (eds.) (1990) Proceedings of the 2ndInternational Conference on New Diamond Science and Technology, Materials ResearchSociety, Pittsburgh.
b. Tzeng, Y., Yoshikawa, M., Murakawa, M. and Feldman A. (eds.) (1991) Applications ofDiamond Films and Related Materials, Elsevier, Amsterdam.
c. Bachmann, P. K and Matthews, A.,(eds.) (1991) Diamond, Diamondlike and RelatedCoatings, Elsevier, Amsterdam.
8.a. Celii, F. G. and Butler, J. E. (1989) Hydrogen atom detection in the filament-assisteddiamond deposition environment, Appl. Phys. Lett. 54, 1031-1033.
b. Celii, F. G., Thorsheim, H. R, Butler, J. E., Plano, L. S. and Pinneo, J. M. (1990)Detection of ground-state atomic hydrogen in a dc plasma using third-harmonicgeneration, J. Appl. Phys. 68, 3814-3817.
9. Anthony, T. R (1990) Metastable synthesis of diamond, Vacuum 41, 1356-1359.10. Butler, J. E. and Woodin, R L., (1993) Thin film diamond growth mechanisms, Phil.
Trans. R. Soc. Lond. A 342, 209-224.11. Page, M. and Brenner, D. W., (1991) Hydrogen abstraction from a diamond surface. Ab
initio quantum chemical'study with constrained isobutane as a model, J. Am. Chem.Soc. 113,3270-3274.
12. Huang, D. and Frenklach, M. (1992) Energetics of surface reactions on (100) diamondplane, J. Phys. Chem. 96, 1868-1875.
13. Pate, B. B., Hecht, M. H., Binns, C., Lindau, I. and Spicer, W. E., (1982) Photoemissionand photon-stimulated ion desorption studies of diamond(111): Hydrogen, J. Vac. Sci.Tech. 21, 364-367.
14. Angus, J. C., Will, H. A. and Stanko, W. S., (1968) Growth of diamond seed crystals byvapor deposition, J. Appl. Phys. 39, 2915-2922.
15. Yarbrough, W. A. and Messier, R (1990) Current issues and problems in the chemicalvapor deposition of diamond, Science 247, 688-696.
16. Thoms, B. D., Pehrsson, P. E. and Butler, J. E. (1994) A vibrational study of theadsorption and desorption of hydrogen on polycrystalline diamond, J. Appl. Phys. 75,1804-1810.
17. Thoms, B. D., Owens, M. S., Butler, J. E. and Spiro, C.(in preparation).18. McGonigal, M., Russell, Jr., J. N., Pehrsson, P. E., Maguire, H. and Butler, J. E.
(submitted). Multiple internal reflection infrared spectroscopy ofhydrogen adsorbed ondiamond(110), J. Appl. Phys.
19. Thoms, B. D., Russell, Jr., J. N., Pehrsson, P. E. and Butler, J. E. (1994) Adsorption and
114
abstraction of hydrogen on polycrystalline diamond, J. Chem. Phys. 100,8425-8431.20. Russell, Jr., J. N., Sellschop, J. P. F. and Butler, J. E. (in preparation) Thermal
desorption from hydrogenated diamond(llO).21. Samples supplied by a) Norton, b) General Electric, c) Prof. Sellschop, and d) Prof.
Maguire.22. Westbrook, C. K., Warnatz, J. and Pitz, W. J. (1988), A detailed chemical kinetic
reaction mechanism for the oxidation of iso-octane and n-heptane over an extendedtemperature range and its application to analysis of engine knock, in Twenty-secondSymposium (International) on Combustion ,The Combustion Institute, Seattle, WA,p.893-901.
23. Brenner, D., private communication.24. Harris, J. and Kasemo, B., (1981) On precursor mechanisms for surface reactions,
Surface Sci. 105, L281-L287.25. Redhead, P. A., (1962) Thermal desorption of gases, Vacuum 12, 203-211.26.a. Thomas, R. E., Rudder, R. A., Markunas, R. J. (1992) Thermal desorption from
hydrogenated and oxygenated diamond (100) surfaces, J. Vac. Sci. Techno!. A 10, 24512457.
b. Hamza, A. V., Kubiak, G. D. and Stulen, R. H. (1990) Hydrogen chemisorption and thestructure of the diamond C(100)-(2xl) surface, Surface Sci. 237, 35-52.
c. Kubiak, G. D., Schulberg, M. T. and Stulen, R. H. (1992) Erratum to "Hydrogenchemisorption and the structure of the diamond C(100)-(2xl) surface" Surface Sci. 277,234.
d. Schulberg, M. T., Kubiak, G. D. and Stulen, R. H. (1992) Temperature programmeddesorption ofhydrogen and deuterium from CVD diamond samples, in C. L. Renschler,J. J. Pouch, and D. M. Cox (eds), Novel forms ofcarbon, Mat. Res. Soc. Symp. Proc. 207,Materials Research Society, Pittsburgh, PA, pp. 401-406.
SURFACE AND BULK CONDUCTMlY OF HYDROGEN TREATEDPOLYCRYSTALLINE DIAMOND
GASOKOLINA, L.L.BOUILOV, AABOTEV, A.V.MARKINInstitute of Physical Chemistry, Russian Academy of SciencesLeninsky Prospect 31, Moscow 117915, RussiaMATIMOFEEVMoscow State UniversityLeninskiye Gory, Moscow 119899, Russia
1. Introduction
The great attention attracted at present to the problem of hydrogenation ofa diamond is caused by the fact that CVD diamond grows from the gas phasecontaining more than 90 % of hydrogen, which is able to penetrate into thestructure of growing film. In opinion of a number of authors [1,2,3,4] hydrogeninfluences upon conductivity of CVD diamond fIlms (DF) and single diamondcrystals, which opens perspectives to modify the electrophysical properties of CVDdiamond by hydrogen treatment.
A diamond exposed to hydrogen plasma was studied in above works. Nodata were found in literature concerning the affect of molecular hydrogen ondiamond conductivity G. Moreover the effective value of conductivity have beenmeasured in ref.[1,2,3,4], the contributions of surface conductivity Gs and bulkconductivity Gb being unknown. There are also no experiments on measuring thetemperature dependencies G(1) of hydrogenated DF which complicates theinterpretation of the results.
The temperature dependencies of Gs and Gb of DF treated in molecularhydrogen have been studied in this paper.
2. Experimental
The objects of investigation were free diamond plates of 0.5 mm thickness,series C and B, presented by Norton Company (USA) and also the plates of0.07-0.10 mm thickness, series W, obtained in Laboratory of Crystallization ofdiamond coveringof the Institute of Physical Chemistry RAS (Russia) by meansof crystallization from methane-hydrogen mixture activated electrically by arcdischarge.
The 6x8 mm platelets were used. The central aquadag electrodes of 2 mmdiameter were deposited on the both sides of platelets. The guard ring of 4 mm
115
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 115-121© 1995 Kluwer Academic Publishers.
116
diameter was deposited on the growth side. G(T) measurements were conductedin specially constructed vacuum set with radiation heating of the samples in 10-2-10-3
Torr vacuum. The reactor designed for synthesis of OF was used to elaborate OFin molecular hydrogen at 120 torr pressure. The temperature Th was varied from800 K to 1200 K, the processing time th being varied from 15 to 120 min.
3. Results and discussion
The analysis made on the basis of GiT) and GbeT) data showed that thesamples do not differ significantly and have high dielectric properties at roomtemperature. The activation energy was 0,5-1 eV for bulk conductivity and of0,3-0,9 eV for surface conductivity. Experiments on the affect of temperature Th ofhydrogen treatment on Gs and Gb have shown that reliably determinable changesin conductivity values are observed at Th > 1000 K Fig.1 presents the dependencieson T ofGs and Gb measured at room temperature. It can be seen that the rise ofGs begins at some what lower values ofTh compared toG b, in the range ofTh from1150 K the rate ofGs rise decreases and Gs enters the plateau at 10-
5 - 10-4 Ohm-I.Fig.2 illustrates the influence of hydrogen treatment on conductivity of DF
and on the character of temperature dependencies ofGs and G b. The typical curves19 Gs(l/T) and 19 Gb(l/T) represent measurements made before and afterprocessing in hydrogen. One can see that hydrogen treatment not only increases theOF conductivity but also changes drastically the character of temperature Gs andGb dependencies. So while before the treatment Gs andGb exhibit distinct activationcharacter, after heating in hydrogen there are insignificant changes in the values ofGs and Gb in the range 300-400 K,at T > 400 K decrease of Gs and Gb has beenobserved and only at T > 500 K there is a rise of conductivity.
After the heating up to 850-900 K in vacuum the course ofGs (T) and Gb(T)dependencies on cooling differs from that measured on heating: the minimum inGiT) graph disappears and a returning to the values ofGs(T) close to initial onestakes place.
The influence of Th on GiT) is clearly seen in Fig.3 presenting 19 Gs(l/T)graphs for the sample C after annealing at Th = 1000, 1050 and 1150 K It followsfrom the figure that the value ofGs increases on rising of Th , yet the character oftemperature dependencies of conductivity does not change.
Fig.4 shows a similar dependence 19 Gi1/T) for the sample W after H2treatment at 1200 K for 1 hour. The bulk conductivity practically did not changedas compared to the initial value.
Comparing the data on H2 annealing influence on the conductivity ofNorton's and IPC's samples, one can conclude that the samples synthesized in IPCreveal the sensibility by several orders lesser than that for Norton's samples, whichmay testify higher stability of IPC's fllms against surface graphitization.
The joint data on H2 influence upon surface conductivity of vapour growndiamond may be explained on the assumption of surface reconstruction of OFcrystalline taking place under hydrogen action on diamond surface after above H2elaboration. This was confrrmed also by data on change of contact angles onwetting the diamond fllm by water: after H2 treatment diamond surfacehydrophobicity increased.
117
G Ohm-11.0B~4 '
1.0B~5
1.0B~6
1.0B~7
1.0B~8
1.0E~9
1.0E-l0
1.0B-ll
1.0B-12
1.0E-13
1.0B-14
120011001000900
1.0B-15~====::::;:=====:------r-----.600
Temperature, KFigure 1. Dependencies of surface G, and bulk Gb conductivities on the temperature of hydrogentreatment.
The reconstruction of surface which in turn results in the change ofelectronic structure of the surface, gives an opportunity to form a band of surfacestates, and to change conductivity of DF. Practically non-activation character ofconductivity in some of hydrogenated DF in 300-400 K range may be indicative ofconductivity through partially filled surface band.
Conductivity decrease with temperature in the 400-500 K range may becaused by partial hydrogen desorption and/or DF surface interaction with oxygenof residual gas. The growth of G on temperature increase from 500 to 900 K maybe connectedwith transition to activation mechanism of conductivity, typical for DFin the above mentioned temperature range. On cooling of samples to roomtemperature and on repeated heating theGs(T) dependence has no maximum.Thisindicates the change of adsorption layer content and also probable partial DF
118
surface "graphitization" after heating up to 900 K in low vacuum. Such partialsurface "graphitization" is observed in our experiments only for samples, heatedpreliminary in molecular hydrogen. As was noted before the analogous backing ofDF, not elaborated in hydrogen, gave no irreversible changes in conductivitycharacter. It is possible to suppose that the partial "graphitization" arouse from jointaction of chemisorbed hydrogen and residual gas (oxygen, water vapour) on heatingat 10-2 - 10-3 Torr up to 900 K. The distinct mechanism of this phenomenon is notclear yet.
G Ohm-ll.OE4l5 '
l.OE4l6
l.OE4l7
l.OE4l8
l.OE4l9
l.OE-lO
1.0E-ll
l.OE-l2
1.0E-l3
l.OE-142
1
3.532.521.5
l.OE-15 l-----,-----r------;r------,-----'--,
1
1000/T,KFigure 2. Temperature dependencies of surfaceGs and bulkGb conductivities of the DF sample (seriesC). 1,2- before, 3,4,5,6- after hydrogen treatment at 1050K. The arrows show heating and cooling.
119
Data on influence of hydrogen elaboration on bulk diamond conductivity (G b
increase after annealing cycles in H2, returning of conductivity value to that typicalfor nontreated in H2 samples) are not adaptable to above discussed model of DF.It can be supposed, that hydrogen introduced by annealing cycles to bulk of fIlm orsituated at boundaries is able to compensate trapping centers of charge carriers andto rise DF conductivity. Vacuum annealing removes hydrogen and gives rise to Gdiminishing.
1.OE-05
l.OE-oa
l.OE-07
l.OE-oa
1.OE-09
l.OE-10
l.OE-ll
3
2
8.532.521.5l.OE-12 '-------'------'-----~-----'-----'
1
lOOO/T,KFigure 3. Temperature dependencies of surface conductivity of the DF sample (series C) afterhydrogen treatment at 1000K (1), 1050K (2), 1150K (3).
120
Gs ,Ohm-1
lE-9
IE-ll
IE-IS
lE-15 1------,..---...,..----.-------,..-----.I 1.5 2 2.5 3 3.5
lOOO/T,KFigure 4. Temperature dependencies of surface conductivity of the DF sample (series W). 1- before,2,3- after hydrogen treatment at 1200K. The arrows show heating and cooling.
4. Conclusions
1. It has been found that annealing of vapour grown diamond films in molecularhydrogen at Th > 1000 K causes the increase of surface conductivity of DF. AtTh > 1050 K the rise of bulk conductivity takes place for Norton's samples.2. The samples synthesized in IPC reveal the sensibility by several orders lesserthan that for Norton's samples, which may testify higher stability of IPC's filmsagainst surface graphitization.3. It has been shown that after vacuum annealing at 800-900K the bulk conductivityof Norton's samples returns to initial value.
121
4. Molecular hydrogen treatment of vapour grown films can be used to obtainconducting surface layers on high-ohmic DF without doping by impurities such asboron and phosphorus.
5. References
1. Landstrass, M.L. and Ravi, K V. (1989) Hydrogen passivation of electrically active defects indiamond, AppL Phys. Lett.SS(14), 1391-1393.2. Landstrass, M.L. and Ravi, K V. (1989) Resistivity of chemical vapour deposited diamond films,AppL Phys. Lett.SS(10), 975-977.3. Albin, S. and Watkins, Z. (1990) Current-voltage characteristics of thin films and bulk diamondtreated in hydrogen plasma, IEEE Electron Device Lett.ll(4), 159-161.4. Mori, Y. et al (1993) Characterization of surface conductive diamond layer grown by microwaveplasma chemical vapour deposition, lpn. 1. AppL Phys.32L, 987-989.
POSITRON ANNIHILATION IN DIAMOND FILMS
I.I. BARDYSHEV, L.L. BOUILOV, B.V. SPITSYNThe Institute of Physical Chemistry Russian Academy of Sciences31 Leninsky Prospect, Moscow 117915, Russia
Positron annihilation spectroscopy (PAS) is known as a powerful nondestructivetechnique for research in solids. Because positrons annihilate mostly in localized sites,their annihilation characteristics reflect the electronic structure of such sites. This isthe basis for the applications of positron annihilation in the studies of various kinds oflattice defects in metals and semiconductors [1]. Due to the positive charge ofpositrons the annihilation technique is selectively sensitive to negatively charged andneutral vacancy-type defects. The electron density and average electron momentum ina vacancy are lower than those in the interstitial regions of the perfect lattice.Therefore, the annihilation characteristics: mean lifetime and angular correlation ofannihilation radiation of positrons trapped at vacancy differ from those for positronsthat annihilate in the perfect crystal. The sensitivity of the PAS technique toopen-volume defects is excellent: in semiconductors defect concentrations of the orderof 0.1 ppm or less can be detected by this method.We have studied vacancy defects in CVD diamond films on silicon substrate using
angular correlation of annihilation radiation (ACAR) technique (long-slit geometry).ACAR curves have been measured in diamond films of various thickness (12-25mcm)deposited under temperature from 875 to 1085· C and in natural single crystal diamond.Diamond is known to present the record annihilation characteristics: the shortest
positron lifetime and the broadest ACAR curves allover the substances studied before.The experimental ACAR curves for [100] orientation natural single crystal diamond,
for silicon substrate and for system "diamond film of 35 mcm thickness on siliconsubstrate", represented in Figure 1. All curves are normalized.One can see that ACAR curve for system "film on substrate" is much narrower for
both crystal diamond and silicon substrate. It means that CVD film contains moredefects than diamond crystal.The contribution of silicon substrate in experimental ACAR curve can be subtracted
to obtain the ACAR curve only for diamond film. The scheme of positrons fromradioactive source passing and slowing-down in the system "film on substrate" isshown in Figure 2. Positrons can be stopped in film, can run through the film andannihilate into substrate or reflect from substrate surface and return to the film.Positrons which goes out of the film will be detected and will not contribute in ACAR
123
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 123-127© 1995 Kluwer Academic Publishers.
124
N(e)
25
20
15
10
5
o 2
silicon substrate
l21
10 129, I'IRAD
Figure 1. Normalized ACAR curves for natural single crystal diamond (1). silicon substrate (2)and system" diamond film 35 mcm on silicon substrate" (3).
curve.The probability of positron annihilation in the film I,. can be calculated as:
1=(1+ I-n )-1a [1-exp(-aL)][exp(aL) +n]
where a=positron absorption coefficient of the film, L=the film thickness,n= positron reflection coefficient of substrate surface.
(1)
The appropriate calculations of I,. for all systems under investigationhave been doneand after substraction of silicon substrate contribution the ACAR curves only for CVDdiamond films were obtained.The result of such calculation as comparison ofACAR curve for 35 mcm CVD film
with the ACAR curve for single crystal diamond is shown in Figure 3. Curves arenormalized. The ACAR cure for diamond film is extremely narrow as compared withACAR curve for crystal diamond.
125
f.llm}· ~ (0o
substrate
o
Figure 2. Scheme of positron and slowing-down in system "diamond film on silicon substrate"
N(e)
40
35
50
25
20
15
10
5
0 2
film
crystal diamond
10 12e, MRAJ)
Figure 3.Normalized ACAR curves for natural single crystal diamond (I) and for35 ncm diamond film (2).
126
This dramatic difference may be explained in terms of positron capture andannihilation in microvoids (vacancy clusters). In this case narrow component of ACARcurve connected with para-positronium atoms self-annihilation captured by microvoidswhen diffuse in the film.Using positron capture model [2] from width and intensity of narrow component
obtained after substraction of the broad component (when angles more than 6 mrad thetails of ACAR curves for diamond film and crystal diamond should be the same) themean radius and concentration of microvoids may be calculated.Mean radius of microvoids - positronium traps is [1]:
R,,=16.6/FWHM-1.66(0. Inm)
where FWHM=full width on half maximum of narrow component.
Concentration of microvoids:
(2)
(3)
where In=intensity of narrow component, D=positron diffusivity ( in diamondD=4 cm2/s ), t=positron mean lifetime ( t = 0.13 ns).
The results for all CVD films under investigation represented in TABLE 1 .
TABLE 1 . Mean radius 1(", concentration Ilv of microvoidsand free volume Vf as sum of spherical voids for CVDdiamond films of thickness L deposited under varioustemperature T
875 18 0.721 2.5 1.56 1.07910 20 0.683 2.9 1.12 1.20915 35 0.714 2.9 1.29 1.39940 12 0.767 2.9 1.71 1.841085 22 0.379 2.2 0.42 0.19
127
References
1. Positrons in Solids, Edited by Hautojarvi, P. (1979) Topics in Current Physics,12, Springer, Heidelberg.
2. Brandt, W. (1974) Positron Dynamics in Solids, Appl.Phys. 5, 1-23.
ESR STUDY OF THE PARAMAGNETIC DEFECTS IN FREE STANDING
DIAMOND FILMS
T.A.KARPUKHINA, M.A.PRELAS*, G.POPOVICI*,
S.KHASAWINAH*, B.V.SPITSYN
Institute of Physical Chemistry, Russian Academy of
Sciences, Leninsky Prospekt 31, Moscow 117915,
Russia
*)Nuclear Engineering, University of Missouri,
Columbia, MO 65211, USA
1. Introduction
Chemical vapor deposition (CVD) is widely used to growdiamond polycrystalline thin films. However at present thefilms synthesized by CVD contain a fairly large amount ofdefects. Electron Spin Resonance (ESR) has been used forthe study paramagnetic impurities and defects, predeterminethe physicochemical properties of synthetic diamond. Thedefects may present an important obstacle if the films areused as semiconducting materials. Correlations between thenumber of broken carbon bonds and conductivity in diamond[1], and diamond-like films [2] are being found. Overrecent years the ESR study of diamond layers, grown bymicrowave plasma CVD [3,4] has been reported.
The number of paramagnetictheir distribution in the filmby the growth conditions.
defects, peculiarities ofcomposition are determined
129
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials. 129-135© 1995 Kluwer Academic Publishers.
130
In this study we have investigated by ESR the effect ofthe crystallization temperature Tc on formation of defectswith unpaired bonds in diamond films obtained by CVD in hotfilament reactor. Similar investigation has been carriedout on polycrystalline diamond films grown on the tungstensubstrates by chemical crystallization at Tc in the range800-1200 °c [5].
2. Experimental
In this work diamond films were synthesized by hot filamentCVD at the temperature of filament T1 1900 °c, at the distancebetween substrate and the filament of 4 mm, at the pressurein the reactor of 30 Torr. The temperature of the substratevaried between 640 and 1000 °C.
The ESR spectra registered using an X-bandRE-1306 at a frequency of 9,5 GHz at a roomwith a 100 KHz field modulation.
3. Results and Discussion
Spectrometertemperature
center peak (Fig.1: a,b,c)that the spectrum is composedbroad lines and having almost
Figure 1 shows the ESR spectra of the films prepared at thedifferent crystallization temperatures. Those films exhibitESR single line signals with the g value 2,002, apeak-to-peak linewidth Hpp =2,6-3,7 G and a spin density Ns- 1015_1016 spin/g.
On both sides of theshoulders appear, suggestingof two lines with narrow andthe same g-value.
The spectra of all films studied have been deconvolutedinto two lines, narrow and broad by several ways:
- in two Lorentzian lines;
- in two Gaussian lines;
131
- in narrow Gaussian line and broad Lorentzian line sincefor the wings of the curve are fairly well comparable withthe Lorentzian shape.
,.
b
c
d
,II
/1j/ \
\.•...•.._._.-
.'
\/
/'1 I,.. I I'__-~..... III1 ..,..--'I /---V
Figure 1. ESR spectra for the films depositedat crystallization temperature T : a - 660 °Cj
b -900 °Cj c - 950 °Cj d - 1000 °c
The best correlation of the simulated and originalspectra, registrated at Tc 900-1000 °c is reached whenspectra are deconvoluted into two Lorentzian lines (Fig.2).The spectra of the films deposited at Tc 640-660 °c may bedeconvoluted into two lines with less accuracy and probablyinvolve third overlapping components with sligtly appearedhyperfine structure. But this is not to be due to anitrogen impurity. with our ESR technique sensivity (1012
spin/G) we have not revealed paramagnetic nitrogen defectsin diamond films.
132
Figure 2. The example of the spectrum separating
into two Lorentzian lines (Tc = 1000 °e): a andb narrow and broad Lorentzian lines accordingly;
c - simulated spectrum; points - original data
The results of the spectra resolving are-presented in
Tables 1 and 2. The line width Hgp of the original spectraand the line widths Hppnand Hpp {the narrow and broad
Lorentzian components accordingly} are given in Table 1,
and the corresponding spin densities N s ' N sn and N s
b - inTable 2.
Table 1. The widths of the originalESR spectra Hpp and those Lorentziancomponents Hppn and Hpp
b (narrow and
broad accordigly) at the different
crystallization temperature Tc
Hpp,G nHpp ,GbHpp ,G
640 3,3 4,0 8.0
666 3,7 4,1 8,2
900 2,6 3,0 6,0
950 2,9 3,0 6,0
1000 3,1 3,4 6,8
133
The Lorentzian line is typical for liquid sistems.TheLorentzian line in solid sistems appears as the result ofexchange interactions between neighbouring spins. Thelatter are possible when the distance between spins is notlarger then 4-5 A. Consequently, the structure defectsdistribution in volume of diamond films is not homogeneous,and local spin densities in studied films must be reachedby 1019_1020 spin/g, when average values of Ns were inrange of 1015_1016 spin/g (Table 2).
Table 2. Spin densities Ns of the original spectra,their Lorentzian components N s
n and N sb (narrow
and broad accordingly) and the ratios of Nsb and Ns
n
at the different Tc
Nsn .1015 ,spin/g
NSb .1016 ,spin/g
N biN ns s
640 3,5 8,6 2,1 2,5660 3,0 8,0 1,9 2,4900 2,3 6,6 1,7 2,5950 1,7 4,9 1,1 2,31000 2,0 7,5 1,3 1,7
The narrow line, which is usually observed in diamond,is attributed to carbon dangling bonds on defects of thediamond structure [4,5]. As it was shown in [4], thespectra with g=2.002 and Hpp =3,1 G were corresponded tothe perfect structure of the crystalline particlesrevealing the intense line at 1330 cm-1 in Raman spectra.Maximal value Ns
n (8,6 1015 spin/g) is observed at 640°C.
The wide line with Hpp =7-8 G may be originated bydifferent way. It may be due to the broken C-C bonds at thesites, such as the boundaries of joins between crystallites
134
[4,5], where the degree of disordering is greater than thatin the fine diamond film, or it may be attributed to thedefects, produced during the formation of a-carbon inthe synthesired film composition [3]. Further studies ofthe films, using Raman spectroscopy and X-ray diffractiontechniques might be useful to identify the defects that areunder consideration. Maximal spin density for theseparamagnetic centers reaches 2_1016 spin/g at Tc 640°c, decreasing in two times at 950 °c (Table 2) .
The growth rate curve versus Tc is shown at the Fig. 3
d
Co
II/ II I
/ I_____..-r// \'
-----/.~.r~.<
.,/,Illl (, I'... I
/ p'-------- 1,!( ~-.,'-~----, /
VFigure 3. DF growth rate vs.crystallizationtemperature, T2; T1=1900
oC, h=4 rom, P=30 Torr
The comparison of the spin density values with the dataof the growth rate shows, that the Integral intensities ofthe spectra decrease with increasing Tc up to 900 °c andpass through minimum at 950 °c, when the growthe rate riseup to 850 °c and begin to decrease at Tc > 850 °c. Theminimal number of paramagnetic defects Ns
n and Nsb are
formed at Tc =950 °c, when the growth rate of the films isminimal.
The line width decreases as well in the range Tc 640
Semenovich, V.A. Sozin, Yu.J.,and Torishny, V.J. (1987), Sverkhtverdye
135
950 C and pass through minimum at 950 C. At this time thecontribution of the narrow component increases sharply atT =1000 C (Table 2). The Ns increasing at 1000 C issurmised to derive from a larger amount of carbon danglingbonds, which are left without bonding after hydrogendelivery with the substrate temperature raising.
4. References
1. Avdreev, V.D.,Nachalnaya, T.A.Mater., 6, 19-23.
2. Adel,M.E., Kalish, R., and Prawer, S., (1987) J.Appl.Phys., 62, 4096-4099
3. Watanabe, I., and Sugata, K., (1988) ESR in diamond thinfilms synthesized by microwave plasma chemical vapordeposition, Jpn. J. Appl.Phys.27, 1808-1811.
4. Mori, Y., Show, Y., Deguchi, M., Yagi, H., Yagyu, H.,Eimori,N., Okada, T., Hatta, A., and Nishimura, K. (1993)Characterization of surface conductive diamond layergrown by microwave plasma chemical vapor deposition,Jpn. J. Appl. Phys. 32, L 987-L 989
5. Karpukhina, T.A., Bouilov, L.L.,Electron spin resonance studypolycrystalline diamond films,Technology, 47, 538-545
and Chuvaev, V. F. (1991 )of the defects in
Surface and Coatings
EmCIENT REDUCTION OF NITRIDE AND NITRATE TO AMMONIAUSING B-DOPED DIAMOND ELECTRODES
C.REUBEN,E.GALUN,R.TENNEWeizmann Institute, Rehovot 76100, Israel.R. KAI1SHTechnion, Haifa 32000, Israel.Y.NITilUUQ,K.HASHIMOTO,A.FUllSHmMAThe University of Tokyo, 7-3-1, Hongo, Bunkyo-ku Tokyo 113, Japan.J.M. BUTLERNaval Research Laboratories, Washington DC 20375-5000, USA.C. LEVY-CLEMENTCNRS Bellevue, Meudon 92195, France.
1. Introduction
Diamond possesses many outstanding properties such as its chemical inertness,mechanical hardness and high thermal conductivity. It is a good electrical insulatorin nature but its resistivity can be controlled by doping. These redeeming propertiesrender diamond suitable for many new applications. Although many reports haveappeared on the growth and applications of diamond, very few papers have beenpublished dealing with its chemical and electrochemical behavior.
Reduction of effluent gases, like CCh , NOx , (or NOx- solutions), and SCh , isa process of major importance for the control of the environment and for thetreatment of radioactive wastes.
Diamond electrodes were shown to exhibit slow release of hydrogen undercathodic polarization [1], which suggests that this kind of electrode could befavorably utilized for demandive (reductive) electrochemical processes of differentkinds. In keeping with this idea, relatively efficient reduction of nitrate to ammoniawas demonstrated recently, using chemically vapor deposited (CVD) thin diamondelectrode [2].
Diamond thin films are deposited on Si substrate, in general, and therefore therole of the substrate in the electrochemical reaction can not be overlooked.Furthermore, since diamond is a very hard material, the binding of any twocrystallites along the grain boundary is very weak and hence open channels areavailable for the diffusion of the reactant through the grain boundaries down to thesubstrate. This mode of reaction is possibly very relevant to the electrochemicalsystem being investigated. The solubility of NOx in aqueous solutions of high pH is
137
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 137-142© 1995 Kluwer Academic Publishers.
138
very high and consequently our study is essentially equivalent to the study of thereduction of effluent NOx gas. On the other hand the electrochemical reduction ofbicarbonate solutions, is much more difficult than the reduction of COz and hence itcan not serve as a good model system for the reduction of COz. Therefore we haveadjourned the study of the reduction of COz gas to later stages of the project andfocused our attention to the reduction of nitrate and nitrite solutions to ammonia.
Here we report the efficient reduction of nitrate and nitrite solutions intoammonia on diamond thin film electrodes with Faradeic efficiencies (FE) [3] as highas 7. We find that this process scales with the boron doping level of the diamondfilm. It is further shown that the contribution of the Si substrate to the FE ofammonia reduction is appreciable, even for the thickest diamond films (ca. 30 mm).A detailed account of the status of this project is described in what follows.
2. Results and Discussion
Fig.l shows the time (charge passed) dependence of the FE of ammonia productionfrom nitrate (a) and nitrite (b) solutions. Remarkably, the FE of ammonia productionscales with the formal level of boron doping (the relative concentration of boron inthe reaction mixture) in the diamond film. High FE were obtained with sampleshaving high boron concentration. After a few hour of electrolysis, a strong ammoniaodor spreaded-off from the reaction vessel. The time decay of the process could beaccounted for by the deposition of SiOz film from the substrate on the surface of thediamond electrode. Short immersion of samples in a concentrated HF solution,which dissolves the SiOz, restored the high FE of the process. X-ray photoelectronspectroscopy (XPS) confirmed the presence of a thin SiOx film (10m) on the frontsurface of the diamond film after lOOOC/cm2• It also confirmed the removal of theoxide film after immersion in HF solution.
Inductive coupled plasma (lCP) measurements of boron in the reactionsolution were used to refute any corrosion of the diamond film itself during theelectrolysis. Aposteriory XPS analysis of the front surface revealed a less than amonolayer of bound (to diamond) nitrogen which remained almost intact uponheating the sample to 500°C in vacuum.
Fig.2 shows the dependence of the FE of nitrite reduction into ammonia as afunction of the potential (vs. SCE). One notes that the FE increases with the cathodicbias up to -2V, wherefrom the FE falls-off as a result of the increase in hydrogenproduction on the diamond electrode. Under -1.5V no ammonia could be detected.The same trend was observed for nitrate reduction to ammonia. The dependence ofthe FE as a function of the nitrite concentration is shown in Fig.3. The FE increaseswith nitrite concentration up to 2.0M and then drops. Similar dependence wasobserved for the KOH concentration (Fig.4). The reaction exhibits a quasi- firstorder dependence on both nitrite and KOH solutions. To try to eliminate theinfluence of the substrate on the electroreduction process, a diamond film (9mm,
139
Processes At -2V, KOH 2M, NaND3
2M12
>U10ZEJU 8Jo-<u..u..o:J 6
S:2<Co 4<C~
« 2u..
........ 10000 ppm Boron
•••••••• - .. - 10 ppm Boron- -. -. 1000 ppm Boron
••••••••••••••••••••~ 80'00 f'"
.._--.-.-.- ..-..
-------1----- * * ..1000
oL.------------:-----....JlOa CHARGE (C/cm2)
Fig. la
Processes At -2V, KOH 2M, NaN022M
• •_..---•-t::"_
7r-------;===========~......... - .. - 1a ppm Boron
-Boron free". --.--1000 ppm Boron••••• _....-. 10000 ppm Boron
............'.'.....
......._----- .. -.:.;....- .......... ~
>--U 6
§ 5
~o:J 4
U~ 3o<C 2~<Cu.. 1
1000
Fig. Ib
OL...------------------JlOa CHARGE (C/cm2)
140
processes after 166.67 C/Area, nitrite and KOH 2M3.5
.............,1 ; i ~ -1
~ 3 ~\
r ~~I\-~ 1.51-.................,:.............. ,.......................... ., + \x
;
-1-2-4 -3POTENTIAL (V)
-5o.sL.-o............--..L......---...........l.....o--............I...................-.l-.......---.......
-6Fig. 2
processes after 166.67 C/Area, -2V, KOH 2M3.5
>-0 31-· .. ···········+············+···············;·········.....•:A .U
a'3!E 2.5
UJU 21- .. ·..········+·······:;;~············ i············· t··············+······..···· * -I
:;:Cl< 1.sl-r..······..·+··········· ; + ; , + -1
0:::<U. 1""· ..············+··················..;.··············,··· , ; ; -/
3.50.5 1 1.5 2 2.5 3
CONe. OF NITRITE (2M)
o.sw..... ...o
Fig. 3
141
process after 166.67 C/Area, nitrite 2M, -2V3.5
i)
V ~--'
//'
...........•..
VJ........
1o
Fig. 40.5 1 1.5 2 2.5
CONe. OF KOH (2M)3
700300 400 500 600CHARGE (C/CM2)
200
51- ....-..
"'"""~
""~!
~1-............ ~
~1
~~
-x
Fig. 5
process at -2V, KOH 2M, Nitrate 2M1.6
>U 1.ZUJ 1.4
~...... 1.3UJU 1.2..-<Cl 1.
<~
<u.. 0.90.1800
142
lOppm B) was lifted-off from the substrate by etching in HF:NH03 mixture (1:1).The lifted diamond film was carefully placed on Ti substrate. Fig.5 shows the resultsof a preliminary experiment using this film in nitrate solution. The FE decreasedalmost 50% compared with the same electrode on Si substrate, but was> 1, whichsuggests an electrocatalytic chain reaction. Unfortunately the FE decreases withtime, the reason for that decay is not clear at this moment. Further experiments withsuch electrodes are planned to investigate the reduction processes mentioned abovewith a substrate-free diamond film.
For a reaction of high FE to take place on diamond electrode a catalytic model,which is self-sustained, must be invoked. We suggest that the bound nitrogen, whichis less than a monolayer thick, plays the role of a catalyst. The detailed structure ofthis catalyst and its function are currently investigated using a synchrotron radiation.The involvement of NOx and NOxOH intermediates ofvarious kinds in this reactionis quite likely.
3. Conclusion
We have shown that nitrate and nitrite can be reduced into ammonia effectively. Thehigh FE suggests that there is some kind of a radical mechanism (chain reaction)involved. Yet we can not rule out the possibility that the silicon substrate participatesin the process during dissolution. However, this requires a fast electrolyte diffusionthrough 30mm diamond film. Initial experiments using stand alone (Ti supported)diamond films are reported.
4. References
1. a. Pleskov, Y.V., Sakharova, A.Ya., Krotova, M., Bouilov, L.L. and Spitsyn,B.V. (1987)J. Electroanal. Chern., 228,19; b. Patel, K., Hashimoto, K. andFujishima, A. (1992) Denki Kagaku, 60, 659.
2. Tenne, R., Patel, K., Hashimoto, K. and Fujishima, A. (1993) J. Electroanal.Chern., 347, 409.
3. FE=nm/QF, where n is the change in valency of nitrogen in the reaction, m is thenumber of moles of ammonia, Q is the charge passed through the cathode and F isthe Faraday constant.
ELECTRONIC AND SENSING PROPERTIES OF DIAMOND
J. L. DAVIDSONVanderbilt University. Department ofApplied and Engineering SciencesBox 99-B, Nashville. TN 37235.
Abstract
Diamond has attractive properties as an advanced electronic material. Itscombination of high mobility, breakdown and thermal conductivity results in thelargest Johnson's and Keyes' figures of merit by far. For example, the cutofffrequency of diamond transistors, as governed by the semiconductor material,suggests that if diamond devices could be realized, the enhancement in electronicperformance would be extensive. Trew's calculations of the frequencyperformance of realistically scaled diamond MESFETs noted "performancesignificantly better than possible with GaAs MESFETs."The high pressure diamond fabrication processes developed in the mid 1950'ssynthesize particulate diamond. The direct deposition of diamond layers fromchemical vapor processes, pioneered by B. V. Derjaguin, B. V. Spitsyn and J.Angus, provide new vistas for diamond applications. These efforts have clearlydemonstrated that diamond film with electronic potential can be routinely formed.The realization of that potential is another matter. There are material problems thatcurtail exploitation for circuits, principal among these is that the deposited filmsare often polycrystalline and hence contain grain boundaries, twins, stacking faultsand other lineage and area defects which reduce the mobility and minority carrierlifetime. Although the potential of diamond has been demonstrated and activedevices have been made, they were typically achieved by using homoepitax onnatural or synthetic high pressure diamond substrates. To date there have been nocorroborated observations of a means of achieving heteroepitax (single crystaldiamond grown on a non diamond substrate) and, therefore, no practical means ofachieving diamond devices. Also, the excellent mobility values for carriersmeasured in natural single crystal diamond have not been routinely achieved insynthetic diamond film.Nevertheless, polycrystalline diamond films (pDF) possess the same basicsemiconductor material properties which are desired for use in electronics andsensors, such as a wide band gap (5.45 eV), low thermal coefficient of expansion(-I x 10-6 fOC) and high thermal conductivity (20 W(cmKr l ). Because of theseproperties, diamond devices could potentially be used in high-power, hightemperature environments. The development of a PDF passive device, such as aresistor, discussed in this presentation, facilitates the refinement of techniques
143
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 143-160© 1995 Kluwer Academic Publishers.
144
required for the development of diamond active devices. These techniques are, inpart, adapted from standard microelectronic technologies. Furthermore, there is agrowing interest in the utilization of polycrystalline diamond films as a basicmaterial for sensor application, and with good cause. The feasibility of using thenew diamond technology as a piezoresistor structure for strain sensing applicationshas been established and demonstrated by the author. The use of PDF resistorstructures for strain sensing applications brings together many desirablecharacteristics, such as, high temperature operation, relatively high gauge factor,small size, good mechanical, thermal, and chemical properties, electrical stability,and compatibility with hostile environments.
In this paper we review some of the reported achievements related to diamond asan electronic and sensor material. The state of heteroepitaxial films andpolycrystalline diamond active electronics will be covered; negative electronaffinity is briefly covered, and the piezoresistive effect in diamond and itsrelevance to sensing is described.
1. ELECTRONIC DIAMOND
The properties of diamond, including high carrier mobility, breakdown, andthermal conductivity, clearly indicate that diamond is a material for superiorelectronic performance. For example, the cutoff frequency of transistors l , asgoverned by the semiconductor material, Figure I, suggests that ifdiamond devicescould be realized, the enhancement in electronic performance would be extensive.Trew2, as part of an in-depth comparison of wide band gap semiconductors,calculated the hypothetical frequency performance of realistically scaledMESFETs, Figure 2 noting the "performance is significantly better than possiblewith GaAs MESFETs."Collins3 presents a view of avoiding overly optimistic projections for diamond,pointing out that the mobility and carrier densities in diamond may not be nearly ashigh in real material as they are theorized or measured in natural single. Forexample, Figure 3 compares the conductance with temperature and goes on to note,for example at 300 °c the frequency response of the GaAs device is more thantwice as high as that of the diamond device, due in part to the rapid decline of holemobility in diamond as a function oftemperature4 .The high pressure diamond fabrication processes developed since the mid 1950'ssynthesize particulate diamond. The direct deposition of diamond layers fromchemical vapor processes, pioneered by 1. Angus and B. V. Spitsyn, provide newvistas for diamond applications. A listing of deposition techniques nowdemonstrated for diamond includes various plasma assisted chemical vapordepositions at microwave, RF, and DC frequencies, and heated filaments. Also,electron, UV and laser assisted CVD, ion beam, and halogen-carbon thermal CVDprocesses have achieved diamond films. For a review of deposition techniques see,for example, Zhus.
145
These efforts have clearly demonstrated that diamond film with electronicpotential can be routinely formed. However, there are material problems thatcurtail exploitation for circuits, principal among these is that the deposited filmsare often polycrystalline and hence contain grain boundaries, twins, stacking faultsand other lineage and area defects which reduce the mobility and minority carrierlifetime. Although the potential of diamond has been demonstrated and activedevices have been made, they were typically achieved by using homoepitax onnatural or synthetic high pressure diamond substrates. To date there have been nocorroborated observations of a means of achieving heteroepitax (single crystaldiamond grown on a non diamond substrate) and, therefore, no practical means ofachieving diamond devices. Also, the excellent mobility values for carriersmeasured in natural single crystal diamond have not been routinely achieved insynthetic diamond film.
1.1 SINGLECRYSTAL DIAMOND
There are many examples of active semiconductor diamond structures derived inpre-existing single crystal diamond substrates by implantation and epitax. Thedevices are typically crude by modern IC standards, but illustrative of diamondpotential.Historically, Prins6 achieved bipolar transistor behavior (but not current gain witha - .1) in natural p type diamond crystals, employing ion implantation and a novelshadowing technique to mask the implant to form n type regions. Otherinvestigators implanted lithium?, or arsenic8, into p-type diamond, subjected it to ahigh temperature anneal and ascertained that p-n junctions had been formed.Tzeng9 fabricated diodes and transistors in natural diamond with arsenic implantedin p-type natural single crystal substrate, achieving a npn transistor with currentgain of 0.8. Geis20 reported forming point contact transistors on synthetic p-typediamond, exhibiting power gain of 1.3 to 4.5 at T > 5000C. Geis11 also fabricateda permeable base transistor in natural p-type diamond with the aid of ion beamassisted etching to produce the grating, but had a transconductance of only 30mS/rom due to high (- 104 ohm-cm) diamond resistance. In work thatcharacterized the metal-Si02-single diamond interface, it was demonstratedl2, withproper cleaning and surface preparation, that surface carrier modulation of dopeddiamond is feasible.More recently, Tsai13 demonstrated a ten-fold modulation of the channel currentin a metal semiconductor field effect transistor (MESFET) operating in the linearregion. An ultra shallow p-doped channel of less than 50 urn was formed by rapidthermal diffusion with solid cubic boron nitride as the dopant source in directcontact with the diamond surface. The drain current was modulated by applyinggate voltages between -2 and +5v and pinchoff was observed at high positive gatebias. Scho~ gate current-voltage characteristics indicated reverse bias leakage of- 3.0(l0-13)a at +5 V. 1.6 rna/rom was obtained for the open channel current at 10V drain bias. The calculated tranconductance was only 0.7 mS/rom, the maximumfield being only 140 V/cm.
146
Selective growth of boron doped homoepitaxial diamond films l4 was achievedusing sputtered Si02 as a masking layer. The hole mobility of selectively grownfilms varied between 210 and 290 cm2Ns for hole concentrations between 1.0(1014) and 6.9(1014) /cm3. The technique was used to fabricate a thin filmdiamond field effect transistor that functioned at 300 °C. The channel resistance ofthe device was an exponential function of temperature. The high temperatureoperation of the device was made possible by incorporating an insulating gate.Since heavily doped source and drain regions were not formed, the device suffersfrom excessive series resistance which remains a problem for most diamonddevices, at least at 300K. Another problem was the strong temperature dependenceof the channel resistance. This results from the fact that at 300 °c diamond is stillin the carrier freeze out regime (per ionization energy ofboron at 0.37 eV in lightlydoped material I 5.Grot l6 fabricated mesa isolated recessed gate field effect transistors. Etching theboron doped homoepitaxial diamond films with electron cyclotron resonance ECRplasma, the transistors operated at temperatures up to 350 0c. The roomtemperature hole concentration was reported to be 1.2 (10 13) /cm3 and the mobilitywas 280 cm2Ns. Device operation characteristics for this diamond MOSFETdevice operating at 350 °c is shown in Figure 4. The maximum transconductancewas 87 mS/rom at 200 °C. Figure 5 illustrates the fabrication sequences involvedin achieving this device.These examples and others, not cited here for lack of space, illustrate that naturalor high pressure synthesized single crystal diamond can be processed withtechniques similar or identical to semiconductor technology and achieve the activeelectronic functions expected of a semiconductor material, namely: resistivity(doping), pn junctions, Schottky barrier contacts, MIS surface modulation andvarious three terminal bipolar structures. A few of these examples suggested thepotential ofdiamond by "performing" at temperatures in excess of 300 to 5000 C.From these examples we can see how much more robust diamond devices can beusing the modem CVD processes for diamond layers. However, practicality is stillfar removed because a single crystal "mother substrate" was required in allinstances, severely curtailing substrate availability even for research purposes.Furthermore, as observed by Gildenblatl4 "to summarize, at present the feasibilityof diamond electronics has been demonstrated using a variety of simpleexperimental devices, while the development of real diamond based semiconductortechnology has just begun. In particular the potential of semiconductor diamondfor high speed or high power operation has not been demonstrated experimentallyas of today."
1.2 HETEROEPITAXY
The ideal scientific and technological achievement for practical diamondsemiconductor devices would be diamond heteroepitaxy. The published activitiesin this area are few. The challenge of heteroepitaxy is believed to be dependent onmatching the lattice parameter of diamond and/or the thermal coefficient of
147
expansion to the host substrate. Other factors, such as the role of hydrogen,physical chemistry of the carbon species, and process conditions, are likewiserelevant, see for example, Angusl? and Bachmannl8 .Narayanl9 combined implantation and laser processing to synthesize "continuousdiamond films on non-diamond substrates". By implanting carbon, energy from 60to 120 KeV, dose 1(1018) to 2(IOI8)/cm2 , into copper followed by nanosecondexcimer laser pulses, they reported that diamond film layers were formed at thesurface ofthe copper as observed by SEM, TEM, X-ray and Raman analysis.
In another approach to achieve large area single crystal diamond without the useof a "self' substrate, Geis20 has demonstrated that small particulate single crystaldiamond powder can be settled into small pockets ofa tray-like structure to providea planar array of commonly oriented particles. Then, by applying processes similarto those achieving homoepitax, the growth on each diamond particle proceedslaterally as well as vertically, joining with the other particles in close proximity,forming a continuous layer of diamond. Films over several cm2 in area have beencreated by this process named "MOSAIC".
1.3 POLYCRYSTALLINE DIAMOND
Considering polycrystalline diamond (pCD) films, techniques such as hot filamentand microwave plasma assisted CVD readily achieve these PCD films across planarsurfaces in excess of 25 cm diameter and thickness from a fraction of microns toover 5 mm. Many substrates have been found to accommodate PCD, the mostcommon being silicon. The science and technology in forming PCD is welldeveloped and covered elsewhere, see, for example Zhu5.Pan21has observed mobilities of over 1000 cm2Ns, in PCD films, "propertiescomparable to natural IIa diamond", in PCD films recently fabricated bymicrowave plasma assisted CVD. Interestingly, a thickness dependency wasobserved, higher mobilities in thicker layers, explained on the basis that thick filmshave less grain boundary volume. lmprovements in carrier lifetimes were alsoreported but those values were lower than natural diamond. Hall mobility in borondoped PCD was found to be < 1 cm2Ns in moderate to heavily doped films22 .These films, deposited on alumina, were 1.4 microns thick. A tendency towarddegeneracy was noted as boron concentration exceeded - 5(10 20)/cm3. Alsoinferred from the Hall data was an increasing percentage of carrier activation suchthat plNA approached unity as NA became greater than -5(l020)/cm
3.Discrete electrical resistor components of boron doped patterned PCD werefabricated on aluminum nitride substrates23, Figure 6. Using boron oxide wafersolid source in situ doping, resistivities < 0.1 ohm-cm and sheet resistance < 200ohm-cm were observed. A silver based thick film metalization formed ohmiccontacts.Okan024 described a diamond pn junction diode in PCD. The diamond film was
deposited by hot filament CVD method with P205 and B203 as doping sources forthe n and p t~ diamond layer respectively. The p n junction diode displayeddistinct rectification at 300 and 370 K. Glesener25 fabricated a thin film Schottky
148
diode from flame grown boron doped PCD. Using a titanium dot as the topcontact, the rectification ratio at 5 volts was found to be greater than 300. Aninteresting MIS PCD diode was constructed by Miyata26. They observed diodetype behavior similar to the structures in homoepitaxial diamond. The undopedlayer was estimated to be - 100 urn thick and to contain an "unintentional" dopingboron concentration of 2(l016)/cm3. The intentionally doped layer had boronconcentrations of4(1017)/cm3.Tessmer27 has fabricated a PCD field effect transistor. This device, whosestructure is shown in Figure 7, embodied several fabrication techniques to achievea field effect transistor which demonstrates channel modulation. The PCD wasgrown by microwave plasma enhanced CVD. A 15 micron undoped diamond layerwas grown on the silicon substrate. A 0.5 micron thick active layer was depositedon that intrinsic diamond layer using diborane to achieve p-tr doping. SIMSanalysis indicated the boron concentration to be - 7(l016)/cm . The FET's werefabricated using the concentric ring structure to reduce parasitic conduction pathsbetween the source and drain. Gate length was 2 microns, gate width 314 microns.Source and drain contacts were titanium/gold, the gate electrode was gold on top of75 nm ofSi02' Current levels were low, but three terminal characteristics did showcurrent modulation with varying gate voltage. Device performance at temperatureis shown in Figure 8. At an applied gate bias in excess of 32 volts, the activechannel was pinched off. The peak transconductance at 150 °c was 6.4 nS/mm. Itwas reported that the device failed at a temperature of 200 °C due to failure of thegate oxide.
It would appear that interesting active electronic devices may be achieved inPCD. The mobility, lifetime and other carrier properties adversely affected by grainboundaries, stacking faults, dislocations and other imperfections associated withpolycrystaIline material may preclude performance approaching theoreticalpredictions. Nevertheless, as with polycrystalline silicon, there may be applicationsinvolving rugged environments where PCD devices will prove useful.
2.0 SENSOR DIAMOND
We have been considering diamond basically as a conducting or semiconductingelement to achieve electronic functions in a "classical" sense, such as resistors,diodes, capacitors, and transistors. The other area where the rugged properties ofdiamond are still relevant to achieving electronic function include vacuum fieldeffect transistors and sensors. Most of the known solid state electronic phenomenacurrently put to practical use in other semiconductor material have beendemonstrated in diamond. A sensitive high temperature thermistor28, a lightsensitive switch29, an optical radiation detector30, an opto-electronic switch31 and afast room temperature IR detector in diamond32 are earlier examples of thepotential for diamond "functionality as a robust electronic material". Interestingmodem examples of the radiation detectors are the work of Kania33 which
149
demonstrated with PCD the measurement of single minimum ionizing particles,Figure 9 and UV photodetection34, Figure 10.
2.1 NEGATIVE ELECTRON AFFINITY COLD CATHODE
By generating and modulating electrons in free space (micro-space) similar tovacuum tubes and CRTs, but on an integrated circuit dimensional scale, carriers nolonger interact with the solid and temperature insensitivity, high speed, powerefficiency and resistance to radiation damage are greatly enhanced. To achieveuseful free carrier microdevices, emission efficiency is critical. The emission canbe considered for practical purposes when band gap conditions are created withselected materials and interfaces such that electrons excited into the conductionband of a p or n type semiconductor at low temperatures concurrently reachvacuum energy level (are emitted). This is called the negative electron affinity coldcathode (NEACC).Wide band gap materials that have their conduction band close to the vacuumenergy level such as diamond have been recognized as NEA candidates. Geis35
characterized the metal - Si02 - diamond (MOD) interface for (100) and (Ill)diamond. From C- V measurements on MOD substrate of type II b (boron p-type)natural diamond, they found the conduction band of diamond to be 2.3 ± .4 eVbelow the vacuum level for the (100) orientation surface and 0.7 ± .5 eV abovethe vacuum level for the (III) surface of diamond. Figure II illustrates theessence of that finding, which is in general agreement with earlier reports ofNEA36,37.
Subsequently, they constructed a diamond cold cathode38, see Figure 12. Byforming a junction diode from carbon implanted into p-type substrate and mesaetching (a process developed earlier by Geis), diode isolation and junction exposurewas achieved. With an aluminum electrode, the junction was biased and emissionto vacuum observed. Current densities of 0.1 - la/cm2 were estimated for a diodecurrent of 10 mao The cathode efficiency, that is, the emitted current: diodecurrent, varied from 2(10-4) to 4(10-1O)A and was dependent on the ambient,increasing with the addition of 10-2 torr partial pressure of oxygen into the vacuumsystem.The material properties of diamond films are also advantageous for NEACe. Forexample the hardness39 and Young's modulus40 of diamond films has beendetermined to be nearly that of bulk diamond, indicating its capability to physicallyachieve the NEA structure. An additional benefit realizable from diamond will beits very high heat conduction. Diamond cathode microstructures would conductheat very efficiently, a property important to NEA devices.Field enhanced electron emission from CVD diamond films at room temperaturehas been observed in emission electron microscopy41. These polycrystallinediamond films, grown on molybdenum and silicon substrates, were imaged in theaccelerating field of an emission electron microscope. The total emission currentwas of the order of .01 a/cm2. By improving cathode design and having an ultrahigh vacuum environment, they believe that higher current densities and
150
efficiencies could be obtained. Apparently there may be NEA possibilities forPCD. The occurrence of a (Ill) texture is commonly observed in PCD films andadditional efforts to observe emission are proceeding. The observed low field coldelectron emission from polycrystalline diamond films demonstrates possibilities forelectron emission arrays.
2.2 MICROELECTROMECHANICAL DIAMOND SENSORS
It has been recently established that boron doped diamond films havepiezoresistance (PZR). Aslam et al.42 reported the PZR effect in PDF exhibitedgauge factors between 6 to 17, see Fig. 13. An increase in gauge factor wasobserved with increasing temperature. Dorsch et al.43measured the gauge factor ofPDF in a double-layer diamond structure using a cantilever beam method. Gaugefactors were reported between 2.3 and 5.4 for resistivities between 0.01 ohm-cmand 2 ohm-cm boron doped PDF at room temperature., see Fig. 14. An increase inthe gauge factor from 5.4 to 13.7 was measured as the temperature increased from27 to 60 'C, see Fig. 15.The piezoresistance effect in diamond could be very useful for high temperaturesensors because it appears that the effect is preserved to high temperatures, whichis not the case for the presently pervasive silicon technology. PZR, coupled withthe ability to micromachine diamond films into miniature beams and membranes44
can achieve microstructures for sensors such as accelerometers and pressuresensors. Pressure sensing is also of interest because of diamond's unsurpassedstiffness.The phenomena of PZR effect in semiconductors was first investigated in siliconand germanium in 195445. The resistivity change of n-type material under strainwas described as a band minima shift resulting in the transfer of carriers.Subsequent research applied this PZR effect in semiconductors to strain sensingapplications46,47. For p-type semiconducting material, the PZR effect wasattributed to the band structure of the valence band. The upper P312 state of thevalence band consists of a pair of degenerate bands at k=O. These two degeneratebands have different curvature, hence different effective mass and mobilities notedas light and heavy hole bands. As stress is applied, the two bands split andredistribution of holes occur. A resistivity change arises from both the masschanges and hole transfer48.
For polycrystalline films, the film structure consists of grains and therefore PZRis affected by the texture of the films. Theoretical models for the PZR effect ofpolycrystalline films with preferential orientations were described49,50.Thermionic emission and diffusion explain carrier transport across the grainboun<tary47. The grain interior behaves as crystalline material and hence valenceband warpage can give rise to PZR.Macroscopically, piezoresistance is the phenomenon whereby a resistor
changes resistance in response to an applied strain: i.e., a resistance change will
result when there is an increment change in the piezoresistor length, ~LIL,
151
produced by a reversible (elastic) applied strain. Gauge factor, an index of thesensitivity of a piezoresistive sensor device, is defined as the ratio of relativechange in resistance to the relative applied strain e:
GF=M/R=M/R&/L c
(I)
For a homogeneous material, with isotropic elastic properties, the gauge factor isdefined by
GF = 1+2V + IIp I pc (2)
where V is Poisson's ratio, E: is strain. The tenn 1+2V, nominally less than 5, ismainly from the geometrical piezoresistance effect which is the dominant effect for
the metal film strain gaugeSl . The tenn (IlR/R)/E:, arising from the resistivitychange, is the dominant tenn for semiconducting material and can be > 150 for ptype crystalline Si4s.
An interesting example now considered of applying the diamond PZR effect to amicrostructure is the pressure sensor. PZR based pressure sensors can be derivedfrom micromechanical diaphragm structures. This configuration can be achieved indiamond because PDF piezoresistors can be directly patterned on undoped diamonddiaphragmss2. There are two advantages in using PDF as diaphragm material. First,PDF is chemically inert and can serve as an etch-stop with large etching timetolerance. Second, PDF can be deposited with control by adjusting deposition timeand other growth parameters resulting in selectability of diaphragm thickness.Therefore, the full-scale pressure range can be varied with PDF as diaphragmmaterial. The sensitivity of diaphragm sensors depends on a variety of parameters,such as the diaphragm size, the impurity doping concentration and the gaugelocation. Resistor placement on a diaphragm can be chosen to achieve maximumstrain and optimize the PZR effect.The stress pattern due to pressure difference across a diaphragm is knownS3 . Theradial strain in the diaphragm is
3p(I-vl) 2 2
crr = 8Et2 (Ro - 3r ) (3)
where Ro is the radius of the diaphragm, P is the pressure difference, t is thickness,n is Poisson's ratio which is I to 2 for diamond. The tangential strain can beexpressed as
co _ 3P(1-vl) 2 2<:-88 - 8Et2 (Ro - r ) (4)
152
The distribution of strain over a diaphragm is shown in Figure 16. The radial andtangential strains have the same magnitude in the center. There are tensile stressesat the edges and compressive stress in the center.A diamond membrane type sensor was designed52, Figure 17. The dashed circlerepresents the edge of the diaphragm. Four resistors (width = 0.5 mm) arepositioned on top of the diaphragm. Two resistors (RI, R2) have their lengthoriented radially to examine longitudinal PZR, and the other two resistors (R3, R4)have their length orthogonal to the radius to examine transverse PZR. Theconductor pads are off the diaphragm to avoid interference. Two other resistors areplaced offdiaphragm, on opposite comers, for reference purposes.Undoped diamond diaphragms were fabricated with the process described below
with ten hours deposition time to achieve thickness of 5 microns on a 2 cm x 2 cmsilicon substrate which had been pre-deposition polished with 0.1 micron diamondpowder. Films were annealed for one minute at 850 ·C in argon to removehydrogen.The p-type boron doped PDF piezoresistors were fabricated on the undopeddiamond film via in situ boron solid source doping method. A lift-off maskcomprised of a printed dielectric was used to achieve selective deposition of thedoped diamond resistors. The resistors were doped during deposition with boronfrom a solid wafer boron compound source23 . The dielectric masking layer wasthen lifted off (etched away) by an HF etch. Terminal metalization was achieved byusing silver based thick film conductor paste printed and fired at 500·C. Thediamond diaphragm was accomplished with an isotropic silicon etch of the backside of the substrate. The process flow to achieve the diamond pressure sensor isdepicted in Figure 18.The doped diamond resistors directly deposited on the diamond diaphragm,exhibited excellent adherence. The diaphragm thickness was 5 microns anddiameter 0.92 cm. Pictures of top and back view of the all-diamond pressuresensor are shown in Figure 19. The diaphragm was higWy insulative (> 1 G ohm)as determined by I-V measurements at room temperature. Figure 20 shows thecross-section and top view of a doped diamond resistor (3 microns thick) on top ofthe undoped diamond diaphragm (5 microns).Pressure differential was produced on a special test fixture to examine the PZR ofthese diamond resistors. Strain was applied to the diaphragm in the range of 0 to250 mm Hg by applying positive and negative air pressure against the sealeddiaphragm. Diaphragm rupture stress (in compression) was observed to beapproximately 350 mm Hg in two sample diaphragms with thickness of 5 mm. Theresistance of the four resistors were taken individually at varying pressure. Figure21 shows the resistance change versus pressure of these resistors from a typicalsample of the several fabricated at this time. It can be seen that the resistancechange of the two longitudinal PZR, identified as RI and R2, is larger than the twotransverse resistors, R3 and R4. RI has piezoresistance of 0.1% per 100 mm Hgand R2 has a slightly lower response. This is due to the slight asymmetry of thealignment (see Figure 19), where more length of Rl is on the diaphragm than R2.R3 and R4 have a diminished PZR, as they are in the transverse configuration. The
153
resistivity of these resistors is 4 to 5 ohm-cm and the gauge factor was estimated tobe approximately 7.
3.0 SUMMARY
Polycrystalline diamond films (pDF) possess semiconductor material propertieswhich may be suitable for use in electronics, such as a wide band gap (5.45 eV),low thermal coefficient of expansion (1.1 x 10-6 JOC) and high thermalconductivity (20 W(cmKr1 ). Because of these properties, diamond devices couldpotentially be used in high-power, high-temperature environments. Thedevelopment of a PDF passive device, such as a resistor, facilitates the refinementof techniques required for the development of diamond active devices. Thesetechniques are, in part, adapted from standard thick-film and thin-filmtechnologies.There is a growing interest in the utilization of polycrystalline diamond films as abasic material for sensor application, and with good cause. The feasibility of usingthe new diamond technology as a resistor structure for strain sensing applicationshas been established and demonstrated. The use of PDF resistor structures forstrain sensing applications brings together many desirable characteristics. Theseinclude high temperature operation, relatively high gauge factor, small size, goodmechanical, thermal, and chemical properties, electrical stability, and compatibilitywith hostile environments.
154
FIGURES
o:.o>
, ....., --".__...>L__..>...._,.l..-:- , ...L........:>-_,-J..
CUTOf'F fftEQUEHCY ,~ ~tUJ
Figure I. Theoretical curve of the maximum
operational voltage of a transistor as a
function of the cut off frequency FT for
several semiconductors I.
200 \
'00 SIC \ Dtamon<J
50
_..,\ ,(I'<odlc<od)
\ ,20 \ \
\ \
~'0
\ ,, \
\ \a." \ ,
\ \, \\,
0.5 \,0.2
0.' , I. 20 50 '00
Frequency (GHz)
ooo'---'---~,--_+---'.....J
Q30IT 1«41
Figure 2. RF power performance versusfrequency for diamond, SiC, and GaAsMESFETs 2.
--- .....( -----'
Figure 4. Drain current versus drain voltagefor recessed gate MOSFET operating at350'C 16
·w___ "'""10'
"
f'ICIUIlolU",,"".........."""STaATt
ooo..':~=:-----F====l<-0_"'"..,m",," ----- =====~;~
Figure 3. Conductance per square centimetercalculated for one-sided diodes made fromdiamond, GaAs, and silicon 3.
Figure 5. Fabrication sequence for recessedgate diamond MOSFET 16.
Fipre 6. Pattemed boroa-dopedpolycryslal1ine diamoad resistors witb silverterminals on I" by I" A1N substrate 23.
155
o·lS ·10 ·15 .10 ·sYd. Ofain 10 Source Vola"e ( V )
vc·ov
<'.s
§u
~~ -JO9<
;S -40
-".~
·)0
OevK:e Layoul
In...!flnSK: ~~nd _ _ Gale
SIlicon Subsltate
SChematIC Cross SecllOfi
Soulce Au Gate source----+I'l~.2~\ ! g~~~.~ IOra E\ectrode
on lamo
Figure 7. Cross-section and top view of
concentric electrode diamond
FET structure 27.
Figure 8. 150'C operation ofP-typepolycrystalline diamond FET. This is the firstreported PCD-FET to show saturation andpinch-off 27
•.I!.v
Figure 10. Quantum efficiencies ofMSMdetectors fabricated in type Ib syntheticdiamond and polycrystalline diamond. Thelines are drawn as a visual guide. The
.---'--- Ec applied bias for all samples was IOOV 34.O.i'.V
Figure 9. Sample data demonstrating thedetection of single high energy electrons(MIPS) in a CVD diamond detector. Thelower trace is from a reference silicondetector and the upper trace is the CVDdiamond pulse 33
YACf.JUM LEVEL----..L.,..---=:.:...-:=-----j·····-·-·1·---····Ec
Figure 11. Energy diagram ofa metal-SiOrdiamond structure for (Ill )-oriented
substrates 35.
Figure 12. Schematic drawing of the mesa
etched cold cathode diode 38.
156
Figure 13. Gauge factor ofP0!rcl)'stallinediamond measured at 300 K 4 .
2£-3o£-oE
(OM'!l£S$1'Vt STItA.l,. TENSUJ: sn"Uf
6£-3.------r-----~
-4£-1
-6£-3-'-·--------_-----1-2£-3
r-1"'t"1l 0 0
t'B~,&>... ...,..;rCOUf'tl(SS()M
...,----------:-:-c:-:---, ..."Ol 1[HS()N .. '"ell) ~.~.g)
_______ '1'?!Jo-~ Val
• ~ ,II).'·CleM._) ...
.'".Ol
lQIQI_ DofltClO< ·1.,•..Ol
.»m .1)(1)...... ·.·a:. .... lIlcJQ;,
SllW< ("<oosl''''''')
I'S ...
!i 'Ill-OlD~ ....0:\ .•.,...,·..Ol
15,-----------~
·~.. !>(".. :e~~·1oI-. .. iS6
°2.:;O:----4~O:----6.,..O.,.....------160T rOC]
Figure 14. Relative change in resistancenormalized to the resistance for e = 0 for thelonj(itudinal effect 43.
Tangential strain
Radial strain I
·······ffi··"~riEnII.
I-~I I
I I II I I
Figure 16. The distribution ofstrain over adiaphragm 53 .
Figure 15. Temperature dependence of thelongitudinal gauge factor under tensile strainfor two doping concentrations 43
C====:=J Silicon wafers polished wilhdiamond powder
L eVl) diamond dcpo.ilion(undo(lcd)
AS 5b
Resislor lin offmasl.:
[hron·doped diamondresislor deposition
Terminal metaliution
Figure 17. Design pattern ofa diamondpressure sensor 52.
Figure 18. Diamond pressure sensor process
flow overview 52.
Figure 20. SEM ofcross section view ofa
doped diamond resistor on undoped PDF,
3,OOOX 52.
157
Figure 19. Photograph ofall-diamond pressure
sensor.Top picture: Top view ofdevice withmetalization and resistors. Bottom picture:Backside, showing diamond membrane52.
0.003S
0.003
ARJR 0.002S
0.002
0.0015
0.001
0.0005
100 Iso 200 2SO
P",ssu-e (mmHg)
Figure 21. MVR vs. pressure
(at 2S °C. no pressure. Rl=1731ill,R2=14SIill, R3=I711ill, R4=24Sk.Q.) 52 .
158
REFERENCES
1. A. Johnson, RCA Review, 26, 163, 1963.2. R 1. Trew, 1. Yan and P. M. Mock, Proceedings of the IEEE, 79 (5), 602,1991.3. A. T. Collins, Materials Science and Engineering, 811, 257-263, 1992.4. A. T. Collins, Semicond. Sci. Technol., 4, 60S, 1989.5. W. Zhu, B. R Stoner, B. E. Williams and 1.T. Glass, Proceedings ofthe IEEE,79 (5), 621-646,1991.6. J. F. Prins, Appl. Phys. Lett., 41, 950-952, 1982.7. V. S. Vavilov, E. A. Konorova, E. B. Stepanova and E. M. Trukhan, SOl'. Phys.-Semicond, 13, 635-638, 1979.8. Y. Tzeng, T. H. Lin, J. L. Davidson and L. S. Lan, Proc. Seventh BiennialUniversity/Government/Industry Microelectronics Symposium, IEEE, Rochester,NY,1987.9. ibid 8.10. M. W. Geis, D. D. Rathman, D. J. Ehrlich and R A. Mtrrphy, IEEE Elec. Del'.Lett., 8 (8), 341-343,1987.11. M. W. Geis, N. N. Efremow and D. D. Rathman, J Vac. Sci. Techno/. , A6 (3),1953-1954, 1988.12. M. W. Geis, N. N. Efremow, J. D. Woodhouse, M. D. McAleese, M.Marchywka, D. G. Socker and J. F. Hochedez, IEEE Electron Device Letters, 12(8),456-459, 1991.13. W. Tsai, M. Delfino, D. Hokul, M. Raiziat, L. Y. Ching, G. Reynolds and C.B. Cooper, IEEE Electron Device Letters, 12 (4), 157, 1991.14. G. S. Gildenblat, S. A. Grot and A. R. Badzian, Proc. IEEE, 79 (5), 647-668,1991.15. G. H. Glover, Solid-State Electron., 16,973-983,1973.16. S. A. Grot, G. S. Gildenblat, C. W. Hatfield, C. R Wronski, A. R. Badzian, T.Badzian and R Messier, IEEE Electron Device Letters, 11, 100, 1990.17. J. C. Angus and C. C. Hayman, Science, 241,913-921,1988.18. P. K. Bachmann, D. Leers and H. Lydtin, Diamond and Related Materials, 1,1-12,1991.19. J. Narayan, V. P. Godbole, C. W. White, Science, 252,416-418,1991.20. M. W. Geis and J. C. Angus, Scientific American, October, 84-89, 1992.21. L. S. Pan, D. R Kania, S. Han, 1. W. Agerii, M. Landstrass, O. L. Landen andP. Pianetta, Science, 255, 830-833, 1992.22. K. Nishimura, K. Das and 1. T. Glass,J Appl. Phys., 69 (5), 3142-3148,1991.23. L. M. Edwards and J. L. Davidson, Third Int. Coni on the New Diamond Sci.and Tech., Heidelberg, Germany, September, 1992.24. K. Okano, H. Kiyota, T. Iwasaki and Y. Nakamura, Solid State Electronics,34, 139-141, 1991.25. J. W. G1esener, A. A. Morrish and K. A. Snail, Applications ofDiamond Filmsand Related Materials, edited by Y. Tzeng, M. Yoshikawa, M. Murakawa and A.Feldman, Elsevier Science Publishers, B. V., 347, 1991.
159
26. K. Miyata, D. L. Dreifus, K. Das, 1. T. Glass and K. Kobashi, SecondInternational Symposium on Diamond Materials, The Electrochemical SocietyMeeting, Washington, DC, 1991.27. A. J. Tessmer, L. S. Plano and D. L. Dreifus, Proceedings of the FiftiethAnnual Device Research Conference, MIT, Cambridge, MA, 1992.28. G. B. Rodgers and F. A. Rall, Rev. Sci. Instrum., 31, 663-664, 1960.29. L. A. Venneulen and A. 1. Harris, Diamond Res.. Suppl. to Ind. Diam. Rev., 37[1977],25-26, 1977.30. L. A. Venneulen and A. 1. Harris,J Appl. Phys., 49, 913-916, 1978.31. P. T. Ho and C. H. Lee, Opt. Commun, 46, 202-204,1983.32. J. F. Young, L. A. Venneulen and H. M. van Oriel, Proc. Int. Con! Lasers,110-112,1981.33. D. R. Kania, M. I. Landstrass, M. A. Plano, L. S. Pan and S. Han, Diamondand Related Materials, vol. 2, 1012-1019, 1993.34. S. C. Binari, M. Marchywka, D. A. Koolbeck, H. B. Dietrich and D. Moses,Diamond and Related Materials, vol. 2, 1020-1023, 1993.35. M. W. Geis, 1. A. Gregory and B. B. Pate, IEEE Transactions on ElectronDevices, 38 (3), 619-626, 1991.36. W. F. Wei and W. J. Leivo, Carbon, 13,425-430, 1975.37. F. J. Himpsel, 1. A. Knapp, J. A. Van Vechten and D. E. Eastman, Phys. Rev"B20, 624-627, 1979.38. ibid 35.39. C. P. Beetz and S. H. Tan, SPIE Proc. Vol. 1759: Diamond Optics V, SanDiego, CA, 1992.40. J. L. Davidson and X. Cao, Second International Symposium on DiamondMaterials, 179th Meeting of the Electrochemical Society, Washington, D.C., May5-10, 1991.41. C. Wang, A. Garcia, D. C. Ingram, M. Lake and M. E. Kordesch, ElectronicLett., 27, 1459, 1991.42. M. Aslam, I. Taber, A. Masood, M. A. Tamor and T. 1. Potter, Appl. Phys.Lett., 60(23), 2923-2925, 1992.43. M. Werner, K. Holzner, O. Dorsch, E. Obenneier, R. E. Harper, C. Johnson, P.R. Chalker and I. M. Buckley-Golder, Diamond and Related Materials. 2, 10961099, 1993.44. J. L. Davidson, R. Ramesham and C. Ellis, J Electrochem. Soc., 137. 3206,1990.45. C. S. Smith, Phys. Rev., 94, 42-49, 195446. J. Y. W. Seta, J Appl. Phys., 47(11), 4780-4783, 1976.47. P. J. French and A. G. R. Evans, Solid-State Electronics, 32(1), 1-10, 1989.48. Y. Kanda, Sensors and Actuators A, 28,83-91, 1991.49. V. Mosser, J. Suski and 1. Goss, Sensors and Actuators A, 28, 113-132, 1991.50. G. Zhao and M. Bao, Chin. J Semi., 10, (9), 692-701, 1989.51. S. M. Chitale and C. Y. D. Huang, Proceedings 7th InternationalMicroelectronics Conference, 561-570, 1992.
160
52. D. Wur, 1. L. Davidson, W. P. Kang and D. L. Kinser, Transducers. 93,Yokohama, Japan, 1993.53. from Design Considerations for Diaphragm Pressure transducers, MicroMeasurements Tech. Note, 129, 1968.
DIAMOND MIS CAPACITORS WITH
SILICON DIOXIDE DIELECTRIC
M. J. MARCHYWKA AND D. MOSESE.O. Hulburt Center for Space Sciences
S. C. BINARIElectronics Science and Technology Division
AND
P. E. PEHRSSONChemistry Division
Naval Research LabomtoryWashington DC 20375-5000Internet: [email protected]
1. Introduction
Diamond Metal-Insulator-Semiconductor (MIS) devices hold great potential for applications requiring long storage times and superlative mechanical, electrical, and chemical properties. In particular, diamond is the idealmaterial for constructing imaging devices for Vacuum Ultraviolet, X-ray,and energetic particle fluxes. Results with metal-semiconductor-metal [1]devices and diodes [11] demonstrate that diamond has favorable spectralresponse characteristics for a variety of detector applications. Diamond'slack of native oxide is particularly helpful for ultraviolet imagers since native oxides tend to have uncontrolled properties and strong absorption inthe far UV. [2][18][9][7][19].The development of diamond MIS charge storage devices faces several
obstacles. The lack of a native oxide makes it difficult to fabricate the highquality insulator-diamond interface that would be required for a chargetransfer device such as a CCD. Diamond is metastable and its surface willgraphitize under probable processing conditions, further complicating MISdevice fabrication. Both n-type and p-type dopants are non-ideal. Moreover,
161
M.A. Prelas et at. (eds.), Wide Band Gap Electronic Materials, 161-170© 1995 Kluwer Academic Publishers.
162
natural diamond contains electrically active defects and impurities near theppm level [5].
We have fabricated and characterized diamond MIS capacitors to determine their utility as photodetectors and the mechanisms which dominatetheir performance. Photocapacitance (PC) and current-voltage(IV) measurements were performed on a variety of these devices. Photocapacitancemeasures the change in device capacitance that occurs when light is used togenerate free charge via intrinsic ( electron-hole pair generation), extrinsic (trap levels), and surface mechanisms. This technique has been well studiedfor evaluating silicon and medium band gap materials [13][16][8][15][14] [3].PC is especially suited for the characterization of deep defects in wider gapmaterials because of the long thermal emission times expected from thesestates. It is also a fairly direct measurement of the device performance asan integrating photodetector. IV characteristics indicate the properties ofthe insulator and interface. If a MIS device is to function with a surfaceinversion layer, it is critical that a barrier to electron flow from the semiconductor exists, and, therefore, that current flow be minimal in depletionand inversion.
Our results indicate that integrating, UV photodetectors can be madein diamond and that surface channel CCD's can potentially be fabricated.We observed PC transients that exhibit strong wavelength dependence andcharacteristics of free charge generation at 3eV(400nm), which is well below the band gap of diamond. We also found that devices fabricated onproperly prepared diamond substrates exhibited low current flow at moderate biases while devices fabricated on diamond that had been damagedby ion implantation had uniformly higher current flow and did not exhibitsignificant transient PC. This result highlights the importance of diamondsurface preparation prior to device fabrication.
2. MIS Overview
We investigated the MIS capacitor because it is the simplest device structure which can integrate photogenerated charge. This structure forms thebasic pixel element in silicon CCD imagers and is the first choice for a chargeintegrating device in an imager design. It has three states: accumulation,depletion, and inversion. Accumulation on p-type substrates is achieved byapplying negative bias to the metal electrode. This pushes electrons awayfrom the semiconductor-oxide interface into the bulk substrate. Depletionis achieved by applying positive bias which repels holes from the interfaceleaving a region beneath the gate electrode free of mobile carriers. Withsufficient positive bias and a supply of electrons, electrons will collect atthe interface to form an inversion layer. The device is evaluated as an in-
163
tegrating photodetector by first biasing into accumulation, then into deepdepletion, and finally populating the inversion layer with photogeneratedelectrons. There is no direct way to measure the amount of photogeneratedcharge collected in the depletion region of such a device. However, it canbe estimated by observing the device capacitance as a function of illumination conditions. In deep depletion, device capacitance will increase dueto inversion layer formation and hole emission in the depletion region.We made several MIS devices as described in [10] using a CVD Si02
dielectric on electrochemically cleaned, p-type natural diamond and evaluated their ability to act as integrating photodetectors. Devices were characterized by their equivalent parallel capacitance as a function of bias andillumination. We summarize results that are presented in greater detail in[12].
Illuminated and Dark CV Curves15 rr......,.......,.......rr-.:r:;...:..r-;,:..r:;.=.,::=.--T::..:,..:..:;r-..=;...::r'T'-r....r.;r--r:;.::,:-i-'ic.:y...,.......T"'T......,.-.-.
10 ~~!!!IIa:lIE!'I::::-<+----~-----+------1
.....a.
cs::U
2010obios( volts )
-10OL..L...................-'-'.........'-'-.............................................L..L......................................'-'-L..L...................-'-'.........
-20
Figure 1. Typical CV curves for MIS capacitors in the dark and illuminated by adeuterium lamp.
Typical capacitance characteristics for a device are shown' in figure 1as a function of voltage with and without illumination. Experiments andanalyses described below indicate that the device response is dominated bythe collection offree electrons with illumination above about 3eV and thatnegligible charge is thermally generated ( dark current) at room temperature.
3. Modelling
We modelled tlie device response by accounting for changes in net dopinglevel in the depletion region(Nb) and inversion layer population(N1)[8)[20]
164
[4][13] [17]. Both of these change with time due to thermal and opticalprocesses which occur after the device is driven into deep depletion. Theapplied gate voltage, Vg , falls across the insulator and semiconductor depletion region according to
where W, the depletion depth, is
with
C is the device capacitance per area, Tox is the oxide thickness, f.ox is theinsulator permittivity, and f.s is the diamond permittivity. Differentiatingwith the gate bias held constant, we get dG/dt in terms of inversion layerand effective doping changes:
We took N[ as 9 * W = 9 * f.s (1/G - l/Gin!) where Gin! is the asymptotic value of C. This accounts for electron emission in the depletion regionbut ignores any diffusion from the bulk. This is expected to be a reasonable approximation when charge is generated from deep levels and wouldquickly recombine in the absence of an electric field to separate it fromthe trap. With constant photon flux, electron emission from states in thedepletion region will decrease with time as these states are emptied unless they emit holes at a significant rate. The approximation of a constant"g" ignores this effect. This expression should poorly fit generation due tosurface states since this rate will not be directly effected by the depletionlayer thickness. The net doping is attributed to the sum of a fixed acceptorconcentration (NA) and other levels emitting charge at a rate characterizedby a concentration(NT) and time constant(T):
Nb = NA - NTexp(-t/T).
The overall doping change may be attributable to multiple levels or a singlelevel with markedly different time constants for electron and hole emission.To fit the results obtained here, we have invoked up to two such levels:
Nt, =92 * exp( -t/T2) +93 * exp( -t/T3)
165
~./
~v
~~
4.0
3.~
3.0
G5:2.~8=2.0
1.~
1.0o 2 4 6
tlrne(hour_>e 10
(a)
4.0
3.~
3.0
$; 2.~
ts2.0
1.~
~1.0 """''''--'---''_L-..............--''_L-..............---'_L-...............---'_L-...............---'---'o 2 4 6 IS 10
tlrne(hour_>
(b)
0.060.02 0.04time (hour_>
1 .0 ""-_""--_........_ ......._ ......._-'-_........_--'-_-.L_---'_~'___&..._---'0.00
1.~I__--___:::';lIIII'......--+-------_t----------::I
3.~1---------+-----'=='-7.c----::"""'---------I
3.01---------+-----+.."""''----+--------1
2.01__---------.,'--------_t----------::I
"CS.. 2.51---------+---=""------+--------11f
(c)
Figure 2. Capacitance transients( curves) and model fits( points) from back illuminateddevices. Illumination wavelengths are (a)400,(b)640, and (c)200 and 400nm.
where gi is NTi!Ti. The initial dopant level was taken arbitrarily as lxl016cm-3
although it could be extracted from CV data.Model parameters were adjusted to fit the capacitance transients at
different wavelengths. Representative results are shown in 2 with fitting
166
A (nm) g (e- Icm3 Is) g2(e- Icm3 Is) g3(e- Icm3 Is) T2(S) T3 (s) N:;N~(O) I640 I.lle12 9.17ell 0 25200 1.21 I560 6.lle12 1.25e12 0 14400 4.88 I520 9.02e12 1.3ge12 0 3600 6.48 I500 6.94e12 5.56ell 0 14400 12.5 I400 7.5e12 -5ell 6.94ell 14400 3600 38.6 I400 fast 1.25e15 0 0 I
TABLE 1. Parameters used to fit the model to the data.
parameters summarized in table 1. It was observed that transient shape isnearly independent of light intensity but is very dependent on wavelength.At wavelengths longer than 400nm, the response is dominated by bulk doping changes and is everywhere concave downward. At shorter wavelengths,the transient is nearly described by the Zerbst relation, indicative of thecollection of free electrons at the oxide-diamond interface. The appearanceof this change at 400nm rather than near the band edge at 220nm, is indicative of the large number of bulk traps that may emit electrons. It willbe interesting to see if the below-gap response can be decreased with theuse of high-quality synthetic material as it becomes available.
4. Photodetector Performance
The utility of these devices for integrating photodetection was further examined by measuring the dark current and response to illumination impulses.A typical dark result is shown in figure 3. The device was biased into deepdepletion and allowed to integrate charge for 1.5 days. A measurable capacitance increase occurred due to inversion layer formation by thermalelectron emission from bulk traps and hole emission from surface and bulkstates. Hole emission components would not be relevant to the performanceof a diamond CCD imager. The total of all of these effects, however, did notsaturate the device in 36 hours. For comparison, a CCD pixel made withtoday's highest quality silicon would have saturated in several seconds orminutes at room temperature.The response to transient illumination with a broadband light source
demonstrates integrating photodetection. In figure 4, the device is biasedinto deep depletion at T=O minutes. At T=8 minutes, a short light impulseis applied. The device capacitance increases and remains constant untilanother impulse is applied at T=15 minutes. The process can be repeated
167
Dark Res anse of Diamond MIS2.50
2.40
2.30-a-IS:u 2.20
2.10&-----,""'-------1------------3
4020time (hours)
2.001l...l:_.........__-'-_---''--_.....L..__........_---'__........_--..=J
o
Figure 3. Dark Current integrated over 1.5 days.
with various light doses until the device saturates at about 4.2pF. Then,pulsing into accumulation and back to deep depletion removes the storedcharge and apparently repopulates bulk traps that were emptied during theprevious exposure. This last step demonstrates the repeatable nature of theresponse.
605020 30 40time (minutes)
10
IS PhotoreSDonse
E<- rl_from aee mulotlon horae St DS Inverslo reset
r f-J_
I.
.• ~
.1o
2
4
5
~3a
U
Figure 4. Stepwise inversion layer population with illumination impulses.
168
5. Discussion
It is a little surprising to observe the generation of inversion layer chargeat the long wavelengths investigated here. However, if we plot the relativesurface and bulk components as a function of wavelength and extrapolateto longer wavelengths, we see that there should be no inversion layer chargegeneration with photon energies less than 1.7eV. This is consistent with theelectrons being emitted from donor levels that are known to exist 1.7eVbelow the diamond conduction band[6].
SftActral DeftAndence af Surface and Bulk Phenomena
.
...
..
8
2
o1.1, 2.0 2.2 2.4 2.8
photon ener'9Y2.8 3.0 3.2
Figure 5. Relative importance of surface and bulk mechanisms as a function of wavelength.
169
6. Conclusion
Diamond MIS capacitors have been fabricated that function as integrating,UV photodetectors. Device characteristics indicate that integrating photoresponse is associated with low leakage and proper surface preparation.Photocapacitance data differentiates bulk from surface generation mechanisms and suggests that a surface inversion layer can be formed via electronemission from bulk trap levels. Correlation of this work with previous photoconductivity[2] results on diodes, further supports this conclusion. Noevidence of an interfacial SiC layer was observed although further investigation of the interface properties is certainly warranted.The outlook for making a solar blind, surface channel CCD in diamond
is good. Response below 5.5 eV, which apparently is due to bulk traps, canbe reduced with synthetic diamond. Reduction in surface state density tomake a useful charge transfer device should prove to be an interesting areafor further work.
7. Acknowledgments
We would like to thank Jim Butler at NRL for many helpful discussionsand support. This work was sponsored by NASA under a grant for thedevelopment of diamond UV detectors for astronomy.
References
1. S C Binari, M Marchywka, D A Koolbeck, H B Dietrich, and D Moses. DiamondMSM UV photodetectors. Jrnl. of Diamond and Related Materials, 2(2):1020-1023,February 1993.
2. L R Canfield, J Kerner, and R Korde. Stability and QE performance of siliconphotodiode detectors in the far ultraviolet. Applied Optics, 28(18):3940, September1989.
3. S H Chiao and G A Antypas. Photocapacitance effects in deep traps in n-type InP.J. Appl. Phys., 49(1):466-468, January 1978.
4. T W Collins and J N Churchill. Exact modelling of the transient response of anMOS capacitor. IEEE Trans. ED, ED-22(3):90-101, March 1975.
5. G Davies. The Optical Properties of Diamonds, volume 13. Marcell Decker Inc.,New York, 1977.
6. R G Farrer. On the substitutional nitrogen donor in diamond. Solid State Communications, 7(9):685-687, 1969.
7. J R Janesick, D Campbell, T Elliot, and T Daud. Flash technology for CCD imagingin the UV. Optical Engineering, 26(9):852, September 1987.
8. E Kamieniecki. Low temperature photocapacity measurement in MOS structure.Solid State Electronics, 16:1487-1493, 1973.
9. Raj Korde and Jon Geist. Quantum efficiency stability of silicon photodiodes.Applied Optics, 26(24):5284, December 1987.
10. M Marchywka, S C Binari, and D Moses. Diamond MIS capacitors for integratingradiation detection. In Proceedings of the 2nd Interntaional Conference on the Applications of Diamond Films and Related Materials, Saitama, Japan, August 1993.
170
11. M Marchywka, JF Hochedez, M W Geis, D G Socker, D Moses, and R T Goldberg. Ultraviolet photoresponse characteristics of diamond diodes. Applied Optics,30(34):5011-5013, December 1991.
12. M J Marchywka and D Moses. Photo-characterization of diamond MIS capacitors.IEEE Electron Device Transactions, July 1994.
13. E H Nicollian and J R Brews. MOS Physics and Technology. John Wiley and Sons,New York, 1982.
14. R F Pierret and W M Au. Photo-accelerated MOS-C C-t transient measurements.Solid State Electronics, 30(9):983-987, 1987.
15. R F Pierret and C T Sah. Quantitative analysis of the effects of steady-stateillumination on the MOS-capacitor I. Solid State Electronics, 13:269-288, 1970.
16. C T Sah, L Forbes, L L Rosier, and A F Tasch. Thermal and optical emissionand capture rates and cross sections of electrons and holes at imperfection centersin semiconductors from photo and dark junction current capacitance experiments.Solid State Electronics, 13:759-788, 1970.
17. C T Sah, L L Rosier, and L Forbes. Direct observation of the multiplicity ofimpurity charge states in semiconductors from low-temperature high-frequency photocapacitance. Applied Physics Letters, 15(10):316-318, November 1969.
18. N S Saks, D McCarthy, M C Peckerar, and DJ Michels. Deep depletion CCD's forUV and X-ray imaging for astronomy. Proc. Custom Integrated Circuits Conj., page124, 1985.
19. R A Stern, R C Catura, R Kimble, A F Davidsen, M Winzenread, MM Blouke,R Hayes, DM Walton, and JL Culhane. Ultraviolet and extreme ultraviolet responseof CCD detectors. Optical Engineering, 26(9):875, September 1987.
20. M Zerbst. Relaxation effects in MIS devices. Z. angew. Phys., 2(22):30-33, October1966.
DIAMOND PHOTOVOLTAICS: CHARACTERIZATION OF CVDDIAMOND FILM-BASED HETEROSTRUCTURES FOR LIGHT TOELECTRICITY CONVERSION.
P.I. PEROV, V.I. POLYAKOV, A.V. KHOMICH, N.M. ROSSUKANYI,A.I. RUKOVISHNIKOVInstitute ofRadio Engineering and Electronics ofRussian Academy ofSciences, 11 Mohovaya Str., Moscow, 103907, Russia;V.P. VARNIN, I.G. TEREMETSKAYAInstitute of Physical Chemistry ofRussian Academy of Sciences, 31
Leninski Prospect, Moscow, 117915, Russia.
1. Introduction
For many years, diamond was considered to be very perspective for development ofdevices for radiation detection. Having such excellent parameters as mechanicalhardness, radiation, chemical, and wear resistance, wide transparency window fromultraviolet to middle infrared, this material nevertheless has not found a broadapplication field until diamond film technology has been introduced in practice. Thiswas connected with limited sizes of natural and synthetic diamond crystals, lowreproducibility of their parameters, and lack of the possibility to control the parametersto get the optimal ones for given application.With new technique developed for deposition diamond films using chemical vapor
deposition on different substrates [1,2], wide perspectives have been opened for widescope of applications of diamond including electronics and photovoltaics. However,during the first period main attention was paid to diamond film growth processes to getthe possibility for deposition of high quality films with desired and controlled parametersuch as crystal structure, grain size, dopant type and concentration, optical absorbance,reflectance, and scattering, etc. Recently, radiation detectors have been developedsuccessfully using CVD diamond films [3-5]. These detectors were just photoresistorsand consisted from photoconductive diamond film placed between metal electrodes. Insuch a case the mechanism of action of the detector is controlling the current throughthe diamond film from some external current supply by means of generating thephotocarriers by the irradiation flux.Another type of phototransformers deals with the structures having internal electric
fields which lead to separation of photogenerated carriers and to generation ofphotovoltage in such structures - p-n junction, heterojunction, Schottky barrier aresome examples. Such devices unlike the photoresistors can be used not only forradiation detection but for radiation energy transformation into electrical one as well.This direction is under especially intense development in respect to solar energyconversion problem, mainly on the base of silicon and gallium arsenide type materials.
171
M.A. Prelas et at. (eds.), Wide Band Gap Electronic Materials, 171-185© 1995 Kluwer Academic Publishers.
172
Another very important field in the energy conversion is transforming nuclearreaction energy into electricity. Prelas et al. described [6] a possible process of nuclearenergy conversion with a wide band gap material based photovoltaic element which wascalled as Photovoltaic Energy conversion of Nuclear energy System (PENS). In thissystem, nuclear energy is transported to a fluorescer like Krypton-85 or some otherwhich emits ultraviolet radiation after several intermediate steps. Using a properphototransformer, this radiation can be converted into electricity. Therefore a veryessential element of PENS is a photovoltaic structure which converts this ultravioletradiation into electrical power.Diamond film photovoltaic structures are among few under consideration for such
application. At present, undoped and p-doped diamond film technology is mostlyavailable, and n-doping diamond films is much more difficult to produce. Therefore,main attention was paid in this work to unipolar barrier structures, like Schottky barrierjunctions, and to heterojunctions between diamond film and another n- or p-typesemiconductor. In particular, it was shown that rather good rectifying contacts to CVDdiamond films can be fabricate using different metals (see the review paper byGildenblat el al. [7] and references in it). High temperature resistant, rectifying contactsto boron-doped polycrystalline diamond films were obtained by deposition of TiC film[8]. To fabricate photovoltaic structures with adoptable parameters, one needs tochoose properly the transparent rectifying contacts, doping level, thickness of the film,growth and post-growth treatment regimes.
In this work, polycrystalline diamond films were grown by CVD method ondifferent substrates; photovoltaic structures were prepared on the base of these films; theelectrical, optical and photoelectric measurements were used for characterization ofdiamond films and diamond film photovoltaic structures.
2. Preparation of polycrystalline boron-doped diamond films
The methods and techniques of Chemical Vapor Deposition (CVD) allows to producediamond materials [1] in the form of thin films, thick coatings and free-standing platesand open a way to the large applications of CVD-diamond. However there are severalproblems to be solved connected with grain structure, defects and non-diamondinclusions giving rise to light scattering losses and additional absorption, and toelectrical parameter changes as well. Surface roughness at the grow surface depends onthe growth conditions and, first of all for thin films, on the density of the growthcenters. The growth of a film starts from appearing separate nucleus, and continuouscovering of the substrate surface is reached at the thickness of the order of the averagedistance between the nucleus. As a rule, the surface roughness increases with thethickness. For photovoltaic applications, thin diamond films with the thickness up toseveral microns are the most appropriate. To obtain thin films with a smooth uppersurface and reduced light scattering the substrates were preceded with ultra-fine diamond
powder. The diamond nucleation density up to 109_1010 cm-2 was obtained.Diamond films were grown by HF CVD from a mixture of 0.5-1.5 % methane or
acetone and hydrogen on Mo or single crystal silicon substrate. For boron doping thefilm, the trimethilborate was added to the mixture. The boron/carbon ratio was variedin given experiments from 30 to 2000 ppm. The substrate temperature as measured by
173
optical pyrometer was varied from 870 to 920 'C, the temperature deviations during18h experiments were less than lO'C.
3. Samples
To be used as light to electricity converters, photosensitive elements of different kindcontaining internal electric field regions were prepared and investigated. The structuresand their schematic band diagrams are show in Fig. la and lb. For better efficiency,antireflection coatings might be useful. However, at this stage the samples withoutsuch coatings were studied. Silicon/diamond/metal heterostructures contained twointernal field regions - at the silicon/diamond and the metal/diamond interfaces.Therefore the photocarriers generated by the light absorbed in both diamond and siliconcould be separated in these regions giving rise to photo voltage appearance between themetal electrode on diamond and the ohmic contact to silicon. For the simple model ofan interface between two materials (without interface built-in charge), band bendingvalues were obtained using known work function and electron affinity values, as shownin Fig. 1. For the metal/p-diamond/metal structures, band bending are possible at eachinterface depending on the contact type, rectifying or ohmic one.For part of the samples, the silicon substrate was removed by etching in the acid
mixtures to form diamond windows with area ranging from about 0.2 to 1.0 cm2. Thesurface of the window facing the substrate was as smooth as the starting surface of thesilicon wafer. The morphology of the upper surface gets more rough as the filmthickness increases.Semitransparent Ni electrodes were deposited by vacuum evaporation onto the
diamond film surface for electrical and photoelectric measurements; ohmic contacts tothe silicon substrates were made of indium. The annealing in air atmosphere wasperformed in an open silica container inside a furnace at the temperatures from 200 to680'C.
4. Electrical characteristics
Current- and capacitance-voltage characteristics and deep level transient spectroscopy fordifferent conditions (bias voltage, temperature, parameters of the testing signal) werestudied. Current-voltage characteristics for the most of the as-grown samples werelinear or slightly non-linear. Depending on the boron concentration and on theannealing regimes, it was possible to get nonlinear rectifying I-V curves, Fig. 2a.Capacitance-voltage characteristics (Fig. 2b) showed very weak capacitance dependenceon the bias applied, so the depletion region was spread almost through all the thicknessof the diamond film - the case the most useful for phototranducer mode of operation ofthe barrier structure under study.Investigating the annealing effects on deep levels in diamond films, the isothermal
Q-DLTS technique [9] seems to be the most appropriate. In the Q-DLTSmeasurements, one uses cyclic bias pulses to change the charge state of the centers justas in the case of the capacitance DLTS [10], but unlike the last method, the measuredvalue is not the capacitance but the charge emitted from the traps. During the first part
174
of the cycle, the centers are filled with charge carriers by applying a forward bias pulseto the sample. In the next part of the cycle, the traps emit the charge carriers afterchanging the bias on the sample to zero.The charge emitted during the emission process is collected on a capacitor by using
an integrator circuit shown in Fig.3. The integrator circuit combines a high-speedoperational amplifier Ml with a current amplifier in the feedback loop. Output of theintegrator is a voltage signal V (t) = Q (t)/GC where Q (t) is the charge on theintegrating capacitor C and G is again of the current amplifier. The charging time
constant of integration is <10-7s. In order to obtain the Q-DLTS signal, the output ofthe integrator is applied to a follow-and-hold circuit through switches Ke3 and Ke4 asshown in Fig.3. The measured value of the Q-DLTS signal can be written as
~ Q = Q (t2) - Q (tt) = GC [V (t2) - V (tt)] (1)
where tl and t2 are the time intervals from the discharge beginning.Assuming that the charge emission from the deep level varies exponentially with
time and the integration time constant is much smaller than the trap emission time
constant (en (p)f1, the signal at the integrator output represents the trapped chargeemission Q (t)=Qo [l-exp (-en (p) t]. The Q-DLTS signal is given by
~Q = Qo [exp (-en (p) q) - exp (-en (P) t2)] (2)
where Qo is the whole charge emitted from the traps of given type.The cyclic DLTS algorithm used in this work is different from that of Lang [10].
In Lang's algorithm, the rate window 'tm = (t2 -tl) I In (t2 Iq) is kept fixed while thesample temperature is scanned to obtain the DLTS spectrum. The alternative algorithmused in the present work obtains the spectrum by scanning the rate window 'tm whilekeeping the temperature of the sample fixed. If we keep the ratio t2 I q = a constantand vary 'tm then a maximum in ~Q occurs at the rate window equal to the emissionrate of the trap at temperature T, Le. Ina/(a-l)q = en(p). The maximum value of theDLTS signal
(3)
In these measurements, a=2 value was selected such that en(p) =ln2/q and ~Qmax =QO/4.In comparison with widely used capacitance based C-DLTS [10], our methodics
give one a possibility to investigate the structures which tend to be depleted at room orlow temperature and high frequency capacitance to be not dependent on the charge stateof the interface and bulk traps.
For the most of the structures investigated, the high frequency capacitancedependence on the applied voltage and annealing was increased by about 1-2 orders dueto the annealing even at rather low temperatures, about 200·C. Therefore, using theisothermal charge transient spectroscopy technique was very essential in obtaining theinterface bulk level spectra as dependent on the annealing regimes.
175
The Q-DLTS spectra of the boron-doped CVD diamond films taken at roomtemperature are shown in Fig. 4a and 4b as dependent on the charging time andannealing temperature. It is clearly seen from Fig. 4a that the peak designated as EA2was increased in amplitude with increasing the charging time but its position wasunchanged. On the contrary, the EAl peak was shifting to higher window rate values.The same behavior was observed for the charging time case but with increasing thecharging voltage pulse amplitude. From such a different behavior one can to concludethat EA2 peak is related to the bulk traps and EAl originates from the interface deeplevels [11]. Measuring Q-DLTS spectra at several fixed temperatures and at fixedvalues of applied voltage pulse duration and amplitude, two activation energies, 0.2 eVand 0.4 eV were obtained for these boron-doped diamond films (Fig. 5). The 0.4 eVactivation energy obtained was very close to the 0.37 eV activation energy known forhole emission form boron acceptors in diamond [7]. As to 0.2 eV activation energy, itdepended on the applied charging pulse parameters and could be related to some meanenergy for the continuous distribution of interface states which are recharged by givenpulse.
It was observed for the first time that annealing the samples resulted in redistribution of the Q-DLTS spectral density in favor of deeper levels as it is shown inFig.4b. However, the activation energy value for the EA2 peak remained practicallythe same, see Fig. 5. This activation energy could be connected with boron acceptors;in this case the increase of the ~Q peak as the result of annealing could be caused bythe thickness increase of that part of the depletion layer which contain rechargingacceptors for given applied biasing pulse. However, the capacitance of the sample andhence the depletion layer thickness did not change so much.Another possible reason for such re-distribution of the deep level densities with
annealing could be the hydrogen removal from the diamond polycrystalline film whichleads to higher concentration of dangling bonds and hence higher concentration of moredeep levels. In such a case, the boron acceptor peak is superimposed on the tail of thepeak (or peaks) not seen separately at room temperature and connected with thesedangling bonds. The increase of the dangling bond concentration with annealing shouldlead to higher amplitude of this peak and hence to higher ~Q values in the region ofEA2peak.
5. Optical characteristics of diamond films
Optical absorption measurements were performed with undoped and doped CVDdiamond films in the 0.19-50 mm range using Specord M400 UV-VIS and SpecordM80 IR spectrophotometers. The threshold for electronic interband absorption wasobserved for all the samples as a sharp changing in the absorption (optical density) near230 nm, Fig.6a.Hydrogen in polycrystalline diamond is a subject of a great current interest since it
plays an important role in the growth of diamond films by vapor deposition and inpassivation of dangling bonds in diamond films. Information on the content and thestate of hydrogen in different parts of the diamond films was obtained from the IR
absorption features at 2800-3000 cm-1. A typical absorption band shape in this
176
spectral region is shown in Fig.7 for undoped diamond film grown with methane asactive gas. Hydrogen content was obtained from the overall area under thisCH-stretching band envelope. Deconvolution of this band showed the presence of CH2
CH3, CH sp3-groups, and the band at 2836 cm-1 with half width near 10 cm-1 The
close similarity of 2836 cm-1 band to one observed on a fully relaxed diamond(111)(lxl) surface by infrared-visible sum frequency generation [12] and also thecharacter of changes of the ratio between amplitudes of this band and the bands
connected with sp3 groups when the growth conditions change testify this absorptionband to originate from the hydrogen located at (111) diamond grain surfaces. The realshapes of the CH-stretching envelopes differ from sample to sample, however thespectral positions of absorption bands remain constant.Post-deposition annealing the CVD diamond films in the oxygen atmosphere is a
valuable tool for obtaining information about the location of non-diamond carbonphases [13]. At the first stage (400-550'C) the concentration of C-H bonds lowered byfew percent due to changing near-surface layers. At higher temperatures, annealing inthe oxygen containing atmosphere leads to selective, non-uniform removal of thenon-diamond carbon and defect diamond from the volume of the CVD diamond films.Fig.8 shows the diamond film IR spectra transformation as resulted from annealing atdifferent temperatures. It is clearly seen that interference maxima are well observed inthe spectra after annealing at the temperatures as high as 620-660'C indicating that thesurface of the film is sufficiently smooth to allow interference between beams reflectedinternally. The average transmittance in the far IR grew from 69% up to 85-90% andeven higher.The optical constants of such annealed diamond film were simulated using an
effective medium approximation (14], in which we assume that the film is a compositeheterogeneous medium consisting of diamond and void components. A calculation ofthe dielectric functions based on this model showed that the volume fraction of thediamond in such porous films was decreased down to 40%. This value corresponds tothe effective refractive index of the film equal to 1.5. The same results we obtainedfrom the interference fringes analysis.The same changes in hydrogen concentration in diamond films were observed for
boron-doped diamond films grown with acetone as an active gas. The difference was inmuch lower hydrogen concentrations for such films, however the integral change of thehydrogen contents due to annealing could be detected. In as grown boron-doped films,
the hydrogen content was found to be about 1020.1021 cm-3. The annealing resulted inhydrogen content decreasing, like for undoped diamond films. For boron doped filmsthe same changes in volume fractions of diamond and voids should occur as in undopedfilms. In such a case, the conductivity over intergrain barriers will prevail over theconductivity along the interface nondiamond phase. This could be the reason forchanging the shape of current-voltage characteristics from linear to symmetricalnonlinear, which is usual for the bicrystals, see Fig.2a.The concentration of noncompensated boron acceptors was determined from the
one-phonon absorption band intensity centered at 1290 cm-1(Fig.6b) which ischaracteristic for llb diamond.
177
6. Photoelectric characteristics
The photoresponse measurements were made for open and short circuit conditions.Illuminating one of two semitransparent metal electrodes deposited onto the diamondfilm (see Fig.l) with a light pulse and measuring the photovoltage between the twoelectrodes, the unipolar photoresponse with the shape similar to the light pulse wasobtained, the polarity of the response to be dependent on which of the electrodes wasilluminated (Fig.9, the middle and bottom curves). The photoresponse for thestructures with one contact to the diamond film and another to the silicon substrate hadcomplicated shape with changing the polarity of the signal (upper curve). Such a shapeof the photovoltaic signal might result from the nonequlibrium photogenerated chargecarrier separation in two regions in the heterostructure-near the metal/diamond anddiamond/silicon interfaces (Fig.1).At high intensity pulsed illumination, the photoresponse increased in time sharply
following the beginning the light pulse and reached the maximum value, Fig.lOa.After ending the light pulse, the photoresponce decreased with a time constants whichvary from sample to sample and to be dependent on the light pulse intensity, seeFig. lOb. This time dependence was defined by the photogenerated carrier trapping,recombination, diffusion, and drift processes. One can see from Fig.lOb that thephotovoltage pulse was longer in a time scale for higher intensities.The amplitude of the photoresponse in an open circuit drastically depended on the
growth conditions and the post-growth annealing procedures, see Fig lOa. For someheterostructures it reached 0.48 V. It is interesting to note that even for diamond/metaljunctions with linear current-voltage characteristics, photogenerated carrier separationwas observed resulting in photovoltage appearing between the electrodes. In ouropinion, this fact might be understood as the result of a potential barrier existence atthe metal/diamond grain interface. The non-diamond carbon layers between the grainsmay have much higher conductivity which masks the barrier behavior in current-voltagecharacteristics, but in photoresponse measurements the conductivity along the interfacesbypasses the barrier and just leads to decreasing the photoresponce. The annealingresult in decreasing the conductivity of this phase and allow photovoltage to reachhigher values depending on the regimes.
In the case the resistivity of the intergrain phase become too high, the serialresistance might result in decreasing the photovoltage observed. This model issupported by the existence of optimal conditions of annealing to obtain highphotoresponce amplitudes, depending on the diamond film doping level. The bestresults in respect to the photovoltage were obtained with diamond films grown at B/Cconcentration in the range from 100 to 700 ppm. It was observed also that in the caseof heavily doped films, the photovoltage could be drastically increased by deposition ofa thin undoped diamond layer between the doped film and metal electrodes.
7. Conclusion
Diamond film photovoltaic elements were prepared and their electrical, optical, andphotoelectric characteristics were investigated. Boron-doped CVD polycrystallinediamond films were deposited onto silicon and metal substrates. It was shown that
178
even for non-rectifying junctions, photoresponse for open circuit conditions still exists.The effect of annealing on photoelectric, optical, and electrical characteristics wasstudied also. Interface and bulk: deep levels were studied as influenced by the annealingat different temperatures. It was shown that deep level spectra undergo to the strongchanges depending on the annealing regimes. The photovoltage ofmetal/polycrystalline diamond film junctions in the open circuit was assumed to bedepended on the conductivity and hydrogen content of the intergrain non-diamond phase.In such an assumption, the dependencies of the photovoltage, current-voltagecharacteristics, and deep level concentrations on the annealing regimes were discussed.The photovoltaic parameters of the diamond/metal barrier structures were shown to beadjustable by the proper annealing treatments.
8. Acknowledgments
Authors acknowledge supporting this work from the Department of Energy and theUniversity of Missouri, Columbia, and NATO for the travel grant for participating theNATO Advanced Workshop on Wide Bandgap Semiconductors in Minsk, Belorussia,1994.
9. References
1. Spitsyn, B.V., Bouilov. L.L., Deljaguin. B. V. (1981) Vapor growth ofdiamond on diamond and other surfaces, 1. Cryst. Growth 52,219-226.
2. Kamo, M., Sato, Y., Matsumoto, S. (1983) Diamond synthesis from the gasphase in microwave plasma, J. Cryst. Growth 62, 642-644.
3. Kania, D.R.. , Landstrass, M.I., Plano, M.A., Pan, L.S., Han, S. (1993)Diamond radiation detectors; Diamond and Related Materials 2, 1012-1019.
4. Binari, S.c., Marchuwka, M., Koolbeck, D.A., Dietrich,H.B., and Moses, D.(1993) Diamond metal-semiconductor-meta1 ultraviolet photodetectors,Diamond and Related, Materials 2,1020-1023.
5. Vaitkus, R., Inushima, T., Yamazaki, S. (1993) Diamond thin films foreffective ultraviolet and visible radiation detection, 2nd InternationalConference on, Application ofDiamond and Related Materials, MYU,Tokyo, 95-100.
6. Prelas, M.A, Charlson, EJ., Charlson, E.M., Meese, 1., Popovici, G, andStacy, T. (1993) Diamond Pholovoltaics in Energy Conversion, 2ndInternational Conference on Application ofDiamond and RelatedMaterials, MYU, Tokyo, 329-334.
7. Gildenblat, G.Sh., Badzian, A. (1991) The electrical properties and deviceapplications of homoepitaxial and polycrystalline diamond films,Proceedings of the IEEE 79,647-668.
8. Tachibana, T., Glass, J.T., Thompson, D.G. (1993) Titanium carbiderectifying contacts on boron-doped po1ycrystalline diamond, Diamond andRelated Materials 2,37-40.
9. Polyakov, V.I., Perov, P.I., Rukovishnikov, A.I., Ermakova, O.N.,
179
Aleksandrov, A.L., Jgnalov, B.G. (1987) Multi-functional device formeasuring the parameters of the levels in semiconductor structures,Mikroelektronika 16, 326-333.
10. Lang, D.V. (1974) New method to characterize traps, J. Appl. Phys. 15,3023-3032.
11. Yamasaki, K., Yoshida, M., Sugano, T. (1979) Deep level transientSpectroscopy of bulk traps and interface states in Si MOS diodes, JapaneseJ urn. Appl. Phys. 18, 113-122.
12. Chin, R.P., Huang, J.Y., Shen, Y.R., Chuang, TJ., Seki, H., Buck, M.(1992) Vibrational spectra of hydrogen on diamond C(111-(lxl), Phys. Rev.B, 45, 1522-1524.
13. Bachmann, P.K., Leers, D., and Wiechert, D.D. (1993) Post-depositionalDiamond etching, Diamond and Related Materials 2, 683-693.
14. Wehman, I., Jorter, J., Cohen, M.H. (1977) Theory of optical andmicrowave properties of microscopically inhomogeneous materials, Phys.Rev. B, 15, 5712-5718.
180
(a)
(b)
E~~~ ~ It h""IIS_SclIotillycontaets (AI,)
HJI" - poly-dIamond (a-doped)
4001'"' - Si (p OR f1 type)c_OAmic contact (7n)
I'::--=--L__ Ev
~El
If.
Ii E"SfeY
w~,J----L~L~
NetLl p-bl8.MonJ
Figure 1. Sample structures and band diagrams_ a) Diamond/siliconheterostructure; b) Schottky barrier structure.
181
2·1 0
6ias, V
---~---~---~----I I II I II j I
" ---,----t---i---I I II I II I I
is -2.-QS 0 o.SB,dS, V
...c:
I------=t""'-------j'<;; 6
120
100
80
60
... '10
j!20
I 0u- 20
-'10
-60
-80
-/0
-f..S -f
(a) (b)Figure 2. (a) Current-voltage (b) capacitance-voltage characterictics of Ni/diamondmmjMo structure. 1 - Ni electrode was deposited onto as grown polycrystallinediamond film; 2 - before the electrode deposition, the mm was annealed in an airatmosphere at 470 'c.
Figure 3. Electronic circuit for Q-DLTS measurements.
182
fO-r--------....,..------,
T' 293K~I -u
<: •
~"'-~8
fA!
tEAl00-'-_-'--L I
7 I 2 3 4 6
lo9(r..) (r~,j'J3 4
f.og(rm) (rm,)'S)
(a) (b)Figure 4. (a) Q-DLTS spectra of the HF CVD boron doped diamond fIlms takenby the rate window scan method. The charging voltage pulse amplitude equa12V,:duration 2 ms(l), lms(2), 0.5 ms(3). (b) Q-DLTS spectra of the HF CVDdiamond film before annealing(3) and after annealing at 200'C (2) 350=C (1)measured at 2V pulse and charging duration 1 ms.
•J
-v -
-7
-,-$
- (0 L----z-'-.s=--·---~,,:-------:JLS-----'
r-~/oJ
Figure 5. Arrenius plot of the deep levels detected
183
2000
C-If S1rr/,A
3000
10
20
1;%40
0,2 0.8
A,rn
(a) (b)Figure 6. Transmission spectra of 1.6j.lm boron-doped diamond film at (a) UVVIS and (b) JR.
fO
30
20
'U/Jo-
---"I --- I
ooo~. sp'. 'Ctlle'"
-' '-
qOOJ
0.007.
Qoot -
... -- ------_._-, ------,sp:c-' ,
JOOO 2900 2800
W, em-'
Figure 7. C-H stretching band of the undoped 1.8j.lm thick CVD diamond film (las deposited, 2- annealed at 350T, 3- at 470·C, 4- at 580·C)
184
1;7
gO
60
40
20
,.,.,..;",,.. .,.",/
10004000 3000 2000
ro cm- l
Figure 8. IR transmission spectra of undoped 10 11m thick CVD diamond film(solid line- before annealing, dash line- after annealing at S90'C, dash-dot line- at61S'C, dotted- at 640'C).
o
liurll hurl Ia - baa=>
R (t) R
cf~'
Schema-tic measWling diagRam
V,k.v IIa~c
~l
-0,1
0. Z limt,trlS
o.t
'''..~II
0.3 a-'lO,z
QI
0. 0 j {imt,ms
\W II~
:'\/ 1\a~~
~ll
J\ ~
Figure 9. Photovoltage time dependencies for the diamond- silicon heterostructurewhen illuminating different electrodes by the flash lamp light pulse.
185
o
o
Q3 '-~I---------,----.----,
II
1 I I I II~ 1 L L L _
- - I I I 11 1 1 11 1 1 1I 1 I 1 I1 I 1_..j ..J 1_ - - -,--I I I I1 I 1 I 1
I 1 II 2 I II I I I 1____L -----
I I II 1 1I I II I I
7:p 1-a(L--l----+---.J----l---t------lO( 02 03
time, ms(a)
I
- t
I
I-t- --
II
I
I I---I --,.-
06 10
lime, ms
I1I
.. --If-1;l
., '.J .. -_.,
o 02
~~-~-~,---------"~----,-,---,
I L I-, I I
I II
II I-
I
i
anI
o
009
007
"
~-
003
(b)
Figure 10. a) Photovoltage kinetics of the Ni/diamond film barrier structurebefore (2) and after (1) annealing at 650"C; b) Photovoltage time dependencies atdifferent flash lamp pulse intensities I. 11<12<13<4, 'tp is the flash lamp pulseduration.
LASER MODES IN DIAMOND
A Review and Prospects
L.-T. S. LIN and M. A. PRELASUniversity ofMissouri-ColwnbiaE2433 Engineering Building EastColumbia, M065211
GAUNA POPOVICIRockford Diamond Technology, Inc.Professional Arts Building501 South Sixth StreetChampaign,IL 61820-5579
Abstract
Diamond color center lasers have several attractive features such as high power, highefficiency, a wide tunability and room temperature operation. Several successful demonstrations of H3 color center lasers in natural and synthetic diamonds established the scientific feasibility of diamond lasers. Possible diamond color centers for laser action arediscussed. The observation of the luminescence of the color centers in the CVD diamond films implies the suitability of CVD diamond films for making optoelectronicdevices. A UV laser system using excitonic recombination radiation is introduced. Theconfigurations of tunable diamond lasers and diamond photon-emitting devices aredescribed. A brief review of laser physics indicates that laser action usually occurs witha large stimulated emission cross section and a longer laser upper level lifetime.
1. Introduction
Diamond has been recognized as a most promising material for optical, mechanical, andelectronic applications due to its unique properties. With the development of vapor deposition techniques, diamond films have the capability to be the new generation of semiconductors for microelectronic devices. Diamond can also be the active medium forhigh power lasers. Due to its high thermal conductivity and low thermal expansion, heatdissipation will not be a problem and beam quality can be ensured. Moreover, optoelectronic devices, such as light emitting diodes (LEDs) and laser diodes, can be made of
187
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 187-206© 1995 Kluwer Academic Publishers.
188
diamond films from chemical vapor deposition (CVD). With different color center emissions in red, green, and blue regions, diamond LEDs can be applied as the basic units inflat displays and high definition televisions [1]. The advanced diamond lasers in all thevisible wavelengths can be further realized.
The luminescence in diamond is produced mainly by de-exciting processes of color centers. Color centers are electron (or hole) trapping defects which produce optical absorption and emission bands in insulating crystals [2]. Due to the surrounding crystalpotential. the electronic states of color centers are tightly coupled to the crystal phonons.Thus. the optical transitions in color centers give a wide vibrationally broadened band(av - 1000 cm- I ) and usually represent an extreme case of homogenous broadening. Allof the centers are able to contribute energy to a given laser mode and also can bepumped by a laser at a single frequency within the pump band. Furthermore, the fundamental (lowest energy) transition usually has a large oscillator strength (f ~ 0.1) whichmeans that the transitions in the emission band are almost fully allowed. Therefore. alarge absorption cross section (cr - 3 x 10-16 cm2) can be obtained corresponding to thebandwidth and the oscillator strength and in turn leads to a large single pass gain incrystals of only 1 or 2 mm thick and containing only modest;center densities (N _ 1017
cm-3).
Most color center lasers are generated by optically pumping alkali halide crystals wherethe centers (F-related centers) are most easily created. Color center lasers are efficient(-50%) due to very little self-absorption occurring with the fundamental transition.Wide range of tunability can be achieved by color center lasers (e.g. alkali halide colorcenter lasers in the infrared exhibit range between 0.8 and 4 /lm). Moreover. due to thelow saturated intensity, color center lasers are easily pumped by low to modest powers(the threshold is often only a few tens of mW) and operated in cw mode. Ultrafastpulsed operation (as short as a few tens of femtoseconds) for color center lasers can alsobe achieved by mode-locking in the soliton laser. However. alkali halide color centerlasers suffer the disadvantages of short shelf-life. low-power optical bleaching of thecenters, and cryogenic operation [3][4]. Additionally, the blue-green spectral region ofinterest in many applications (e.g. optical storage. flat displays and optical communications) can not be covered by F centers in alkali halide crystals. Thus, diamond color center lasers are promising comparing with alkali halide crystals. especially in the visibleregion.
In this paper. the fundamental parameters for laser action are defined by a brief description in next section. The diamond color centers and their laser actions are reviewed inSection 3. The configuration of diamond laser systems is described in Section 4. Furthermore, a four-level exciton transition is introduced, in Section 5, to produce a tunableUV laser. Diamond. with UV to near infrared luminescence, can thus be utilized to produce efficient tunable solid state lasers operable at room temperature.
189
x·o
b:3n ELAX
3 2W 't21
ASS. LUM.
~~~~~~Jr'S~-_~~ 1,0
0:
oI
00Figure 1. a. Configuration coordinate diagram. showing ground and first-excited electronic states, and
associated vibronic levels and wavefunctiOlls. b. Simplified four-level diagram, showing onlythe most commonly excited vibronic levels of the normal and relaxed configurations. [2]
2. Fundamentals of Lasers
In this section, a brief review of laser physics [5] is given with the fundamental parameters for selecting a potential laser medium.
Figure 1 is a typical vibronic transition for color centers. Consider a four-level laserwith a pumping rate W from the ground state 0 to the excited state 3 and the lifetimes't32' 't21' and 't1O for states 3,2, and 1 respectively. For laser operation, the first requirement is the population inversion of the laser levels. That is to say,
(1)
where N1 and N2 are the population densities and gi and g2 are the degeneracies ofstates 1 and 2 respectively. Therefore, it is preferred to have fast relaxation in the transitions of 3~ 2 andl ~ 0, i.e. short lifetime 't32 and 't1O' By this condition, we may onlyconsider the laser levels, the upper level 2 and the lower level!. The laser-gain equationis written as
(2)
where yo(v) is the gain coefficient (em-I), A.o is the center wavelength of the emissionspectrum, and n is the refractive index. The line shape g(v) for homogenous broadening(usually for color centers) is
190
g (v)l'iv
(3)
where l'iv is the linewidth which is related to the lifetime of state 2, 't21' For a non-zerolifetime of state 1, the width of the transition is .
1 1 1l'iv = -(-+-)
21t 't21 't 1O(4)
Usually, the linewidth can be measured from the spectrum of luminescence experimentally. The stimulated emission cross section crSE in equation 2 can be written in terms ofthe absorption cross section crAB as
(5)
where crAB is measurable by the absorption coefficient and the density of color centers.The Einstein A coefficient is
(6)
where c is the speed of light, e is the unit electric charge, m is the electron mass, and f12is the absorption oscillator strength, which is the fraction of the active electron participated in the absorption. All of the parameters discussed above effect the gain coefficientof the laser medium and therefore are related to the possibility of the realization of diamond lasers.
The intrinsic optical gain may not ensure the laser action since there is some loss occurring in the photon path. If the length of the laser medium is 19' the small signal gain canbe defined as
(7)
However, the gain may be eliminated by the loss in the laser medium or in the opticalcavity by a factor of exp(oog) where a is defined as a single-pass loss coefficient. Thus,the threshold condition for the laser action is
191
(8)
Then, the net gain is
(9)
(11)
In addition to gain, another consideration is the pumping power for lasers. Usually,lasers are operated at or near the saturated intensity which is defined as
hvI = (10)S CJSE'tZ
It is easier to extract laser energy from a system with a lower saturated intensity. In otherwords, the system with a larger stimulated emission cross section and a longer laserupper level lifetime is more likely to produce lasers. The pumping power density at thesaturated intensity can also be estimated by
Yo (v) Isp= ---
g(v)
where
g (v)
6.v z(-)2
(12)
3. Color Centers in Diamond and Their Laser Actions
Since Rand and DeShazer [3] demonstrated the first diamond laser by using H3 colorcenters, very few results for successful diamond lasers have been reported and no paperabout CVD diamond lasers is found so far. Several types of color centers other than theH3 center are candidates for diamond lasers though most of their laser actions have notyet been demonstrated yet.
3.1 H3 COLOR CENlERS
Diamonds contain a variety of color centers due to the existence of impurities andvacancies. Nitrogen is the most common impurity in natural and synthetic diamonds.
192
H3 color centers consist of a vacancy and two substitutional nitrogen atoms in C2v symmetry [4]. H3 color centers are easily produced by electron beam (1-2 MeV) irradiationwith a following annealing at temperatures greater than 773 K. The structure ofH3 center is pictured in Figure 2a with its absorption and emission spectra. The wide absorption band (M - 50 nm) from 420 to 520 nm gives a possibility for flashlamp pumping.For example, lithium excimer at 459 nm may provide an·efficient pumping [6]. Thewide emission band from 480 to 630 nm provides the tunability with a range of -100nm. Figure 2b is the energy level diagram for the H3 center, which shows the possiblelaser transitions.
_ \.5 r----.--r--,---r---r----r-..,
5,.:~Q1.0
~
8z00.5
~
~«
(a)
20.19119.99519.865
...1IIIIIII1I
(b)
}vrBAONfCLEI/ELS
Figure 2. a. Absorption (solid line) and emission (dash line) spectra and strucmral model of the H3 center.Dash circle, vacancy. b. Energy levels of the H3 center. [41
Crossfield et al. [7] measured the decay time of H3 luminescence to be 16.7 ± 0.5 nswhich they considered to be close to the radiative lifetime of H3 centers. They also suggested that there was a nonradiative transfer of energy from the H3 center to the Aaggregate due to an electric dipole-quadrupole coupling. The nonradiative lifetime wasestimated to be close to _10-1 sec. The diamond samples they used were irradiated by1023 m-2 2 MeV electrons and annealed at 1073 K for three hours. The decay timebecame smaller as sample nitrogen concentration increased. A linear relationshipbetween the decay time and the zero-phonon linewidth was observed in gem-qualitysamples, in which nitrogen was predominantly present in the A form. Rand and DeShazer [4] concluded the properties of H3 color centers and estimated the gain coefficient as 0.201 cm- l . Table I is a comparison of H3, N3 and N-V centers in diamond. N3and N-V centers will be discussed in detail later.
The first diamond laser was demonstrated by Rand and DeShazer [3]. A green diamond,2-mm-thick and 7 mm in diameter, with polished parallel optical faces, produced a laserpeaked at 540 nm by optically pumping with a 10 Hz pulsed dye laser at 494 nm withpulse energy up to 5 mI. The laser output was from 18% Fresnel reflections of the dia-
193
TABLE I. Comparison of ID, N3 and N-V color centers In diamond. [4][7][8][9]
Parameters H3 Center N3 Center N-VCenter
Zero Phonon Line (eV) 2.463 2.985 1.945
Refractive Index, n 2.4262 2.4468 2.4065
Temperature, T (K) 295 295 295
Measured Decay TIme, to (ns) 16.7 ± 0.5 41± 1 13.3 ±0.2
Radiative Lifetime, t" (ns) 16±2 157 ± 16 NA
Nonradiative Lifetime, t. (ns) 10.5 (3.3 ± 0.3)xl0·7 NA
Quannun Efficiency, 1\ 0.95 0.29 0.99
Linewidth, AV (s·l) 5.270 X 1013 8.40 X 1013 6.53 X 1013
Center Wave length, Ao (nm) 531 445 697
Gain Coefficient, y(cm'l) 0.201 0.010 0.366
mond faces. No bleaching of the laser-active center was observed at excitation levels ashigh as 70 MW/cm2• They also indicated the tuning range could be as high as 100 omby applying the external mirrors and tuning elements.
Following the first success, Rand and DeShazer [4] demonstrated another diamond laseraction by H3 color centers. The diamond sample with 1.85 mm thick and a measurablewedge angle of 0.36 degree on the polished parallel surfaces was pumped at 490 nmwith intensities exceeding 30 MW/cm2. Laser action occurred at 530 om with efficiencyof 13.5% from 18% Fresnel reflections of diamond faces. A minimum pumping intensity was found as low as 8 MW/cm2 estimated from a calculated inversion density (AN=4.2 x 1017 cm-3).
A successful H3 diamond laser, demonstrated by the group at the ltami Research Laboratories of Sumitomo Electric in 1988, was revealed recently and cited in [10]. A naturaldiamond plate with the dimensions of 3 x 2 x 2 mm3 was side-pumped by a pulse dyelaser at 490 om (pulsewidth of 500 ns) with a low-threshold external optical cavity,which consisted of a 100% end mirror and a 97% output mirror with curvatures of 400mm. Laser action occurred in the yellow region at room temperature with a threshold of3 MW/cm2 without auxiliary cooling.
They also used a synthetic diamond plate to generate a H3 diamond laser [11]. The sample was a type Ib diamond prepared by a radiation of 2 MeV electron beam with a dos-
194
age of 1018 cm-2, followed by a 50-hour annealing at 1700°C under 5 GPa. Itwas foundthat boron doping enhanced the density of the H3 center and reduced the production ofother centers (N-V and H2 centers). The diamond plate was 1.6 mm thick with highlypolished and almost parallel (a wedge angle of 4') surfaces. A laser occurring at -530nm with an efficiency of 0.1% was pumped by a pulse dye laser at 480 nm, with a pulseduration of 500 ns and a pulse energy of 10 mJ. Accordin! to the curve of excitationintensity V.s. output intensity, the threshold was -40 kWfcm .
A cw laser action with a remarkable tunability was demonstrated by Taylor and Henderson at the University of Strathclyde, Scotland in 1988 (cited in [10]) with a cw argon ionlaser operating at 488 nm. Two synthetic diamond crystals from De Beers (JH5, RD1)were used and cooled down to 77 K in a 3-mirror cavity. However, the efficiency wasvery low (6.6 x 10-4%) by the slope of the output V.s. input power curves due to theabsorption of other centers and impurities.
3.2 N3 COLOR CENTERS
The N3 color center is another well known nitrogen-related defect. Usually, N3 centersare present in many natural crystals. Unlike H3 centers, the N3 center is difficult to produce artificially [4]. N3 color centers consist of a vacancy and three substitutional nitrogen atoms with a C3v symmetry. Figure 3a pictures the structure of a N3 center with theabsorption and emission spectra. A comparison with H3 centers shows that both of themhave broad bands in either absorption or emission but N3 spectra shifts to higher energy(-100 nm lower). The absorption band of N3 centers extends from 350 to 420 nm andthe emission band from 410 to 520 nm. It is interesting to note that the combination ofH3 and N3 emission spectra covers all of the visible range from 400 to 630 nm, whichmeans diamond lasers might be tunable through all visible wavelengths. Fi~ure 3bshows the energy level diagram of N3 centers where the transitions of 2A1 ~ A2 areforbidden and the transitions of 2A1~ 2E are allowed [8].
Rand and Deshazer had tried to make N3 color center laser action in diamond [4]. Theyprepared two identical diamond samples: one with only N3 centers with a concentrationof 0.03% (no measurable H3 center concentration) and the other with only H3 centers(described in previous section). However, there was no successful demonstration for N3color center lasers. Thomaz and Davis [8] studied the decay time of N3 luminescenceand found the intrinsic room temperature decay time was 41 ± 1 ns. They also examinedthe wavelength, the specimen, and the temperature dependencies of the decay time. Inthe analysis of wavelength dependence, there was no gross variation in decay timeacross the entire N3 band. The dependence of specimens was actually due to the concentration of nitrogen aggregates. A good monotonic relation between the effectivedecay time and the concentration of A aggregates (high concentration of A aggregatesgave short decay time) was found but no meaningful correlation existed for the decaytime with the concentrations of other types of defects, such as the B aggregates, theplatelets or natural H3 centers. The decay time of N3 luminescence was approximately
195
constant up to a temperature of 430 K. A rapid decrease occurred in decay time abovethis temperature. The ~roperties of N3 color centers is included in Table I with a gaincoefficient of 0.01 cm- , about 20 times less than the gain coefficient of H3 centers.
====IN41VIBRONIC
} LEVELS20.9'019.753
} VIBRON1CLEVELS
\
! 2A
I12A
IIII
:
75.40024.800
24.0702E
(b)
Figure 3. a. Absorption (solid line) and emission (dash line) spectra and structural model of the N3 center.Filled circles, nitrogen; open circles, carbon. b. Energy levels of the N3 center. [4]
3.3 N-V COLOR CEN1ERS
The high theoretical gain coefficient of N-V color centers (almost twice of H3 center'sgain coefficient in Table I) indicates that the N-V center is an excellent laser candidate[10]. N-V color centers, consisting of a substitutional nitrogen adjacent to a vacancywith a C3v symmetry, exhibit a zero-phonon line at 1.945 eV (638 nm) [12]. The energylevels and the absorption spectra ofN-V centers are shown in Figure 4 [9]. The luminescence spectrum was expected to be the mirror image of the absorption spectrum in thezero phonon line [13]. Due to the short decay time of the 3E state (13.3 ns), two processes have been indicated by Redman et al. [9] for the relaxation of N-V centers: intersystem crossing from the metastable state 1A to the ground state ('t = 265.3 ± 0.6 ms)and the spin-lattice relaxation between the 3A ground-state components ('t = 1.170 ±0.003 ms) for low-intensity saturation. However, the vibronic transition of 3E ~ 3A canstill be the channel for the laser action regardless of these two extremely slow processes.Furthermore, high concentration ofN-V color centers can be easily generated by irradiation and annealing processes in type Ib diamond. The absorptive loss in the emissionband is also very low [10]. These conditions make N-V color centers suitable for diamond lasers in red to near-infrared region.
196
•uc.,.0...o•,g0( I=~1A"'0___ 3
A
fOO ISO toO .50
W8V.'.nllth (nm)
Figure 4. Absorption spectra of the diamond samples with 3.1 and 7.7 x 1018 N-V/cm3 and the energy leveldiagram ofN-V centers. [9]
3.4 GRI COLOR CENTERS
The GRI defect is another common center produced by radiation damage. The GRIcenter is attributed to a neutral vacancy with Td symmetry and has a zero phonon line at1.673 eV resulting from the E to T transition [12]. The temperature and wavelengthdependence of decay time for the GRI luminescence was measured by Davies et al.[14]. The decay time decreases from 2.55 ± 0.1 ns at 15 K and 1.4 ± 0.2 ns at room temperature to 0.4 ns at 473 K. This decrease with increasing temperature indicated that thenon-radiative processes were important channels for de-excitation. The true radiativelifetime was calculated as 182 ns, by which the quantum efficiency of GRI centers wasestimated as only about 1.5%. Thus, GRI centers might not be a good laser mediumthough they are easily produced. A recent study [15] showed the non-linear absorptionfor GRI centers, which can be applied as a Q-switch for a laser system.
3.5 OTHER COLOR CENTERS
More than fifty types of color centers have been indicated in diamond [12]. Most of theproperties of color centers are unknown. At this point, the research of the properties andoptical parameters for diamond color centers is becoming of prime importance.
Color centers in diamond are naturally present or are created by artificial methods suchas ion implantation. Ion implantation can introduce various elements into diamond andgenerate a large variety of color centers. Especially in optoelectronic or microelectronicdevices, ion implantation provides a simple tool for device processing. Figure 5 [16]shows the luminescence spectra of ion implanted diamonds. These spectra provide awide selection of required luminescence in visible to near-infrared region and the
197
choices of spectral widths (sharp or broad) depending on the applications. As oneshould note, nitrogen related centers covered almost all of the wavelengths from 390 nmto 800 nm with broad bands including H3 and N3 centers mentioned previously. However, the luminescence of nitrogen-related centers were not strong enough by intentionalnitrogen ion doping except H3 centers [16]. For the consideration oflaser tunabilitY, thebroad band is necessary. Besides nitrogen-related centers, the impurities of Ni in480-600 nm and Ne in 710-870 nm are the potential candidates for widely tunable diamond lasers.
Another wide band emission in the blue-green region has been identified as band-Aemission [17], which exists in both natural and synthetic diamonds. Although the originof band-A emission has not been clearly understood, it is believed that the emission isfrom the donor-acceptor pairs localized around dislocations by the electron-hole recombination. Band-A emission can be obtained by the cathodoluminescence and the electroluminescence [17]. The cathodoluminescence utilizes the direct impact of electronbeam. The electroluminescence uses carrier injection to a diamond device (e.g. theSchottky diode of boron-doped diamond in [17)). An electroluminescent device made ofdiamond has been demonstrated with band-A emission spectrum extending roughlyfrom 400 to 800 nm [17][18]. By applying a properly designed optical cavity, band-Aemission with the widest band is very promising for diamond lasers with carrier injection. Khong et al. [19] studied the decay time of band-A emission and the weaker linesstructure in the region from 2.3 to 3.0 eV using the time-resolved spectroscopy. Theyfound that band-A emission had two components, one with the decay time of 5.7 - 13.6ns and the other with 21 - U5 ns, when the measurements were taken at around 2.8 eVfor all five samples. This result provides a support for a band-A laser by the componentwith a longer decay time.
Pereira and Santos [20] observed laser emission from a red band (2.052 and/or 2.082 eV[21)) that occurs in some natural brown diamonds. This red band presented a strongemission and a large spectral width. Pereira and Santos also proposed S3 color centers(with a zero-phonon line at 2.496 eV) as another possible laser agent with a tunability inthe range of 517 - 590 nm (2.4 - 2.1 eV) due to the similarities to the red band [20].However, preliminary results indicated that excited state absorption competing withlaser action may reduce the possibility of the S3 color center laser.
3.6 COLOR CENTERS IN CVD DIAMOND
Most of luminescence discussed above are from natural and HPHT synthetic diamond.Actually, similar defects and the same color centers have also been identified in theCVD diamond films.
The cathodoluminescence study of diamond films formed by microwave and ECRplasma-assisted CVD showed the blue and green luminescence between 2.4-2.8 eV[17][22][23]. It was recognized as A-band emission. Yokota et al. [24] investigated the
198
I~.I
.....'!J.:.:...:::9+=--l· -"=-1
6
~:_SOD
I
7 0I
8 0
&00
9 0 1 0
SOD 600 700 800wavelength (run)
Figure 5. The luminescence spectra of different impurity related centers in diamond. [16]
formation of color centers in microwave-plasma CVD diamond by elecIron and neuIronirradiation followed with one-hour annealing at 900°C and obtained the 5RL center(4.58 eV) and the color centers at 3.19,2.16, and L94 eV. Freitas et al. [25] studied thephotoluminescence of polycrystalline diamond films by filament-assisted CVD andoxygen-acetylene flame technique. A defect band with energy of 1.68 eV, correspondingto GRI emission band, was observed in filament-assisted CVD diamond. From oxygenacetylene flame diamond, they observed the luminescence of 1.95 and 2.16 eV centers
199
which had been assigned to nitrogen-vacancy complexes. Popovici et al. [26] identifiedthe defects with peaks at 1.12, 1.786 and 1.82 eV in Raman measurements with different excitation light (1.16. 2.96 and 2.41 eV respectively) for hot-filament CVD diamond. Lin et al. [27] observed the cathodoluminescence with peaks at 1.94.2.16.2.47,2.92 and 3.19 eV in hot-filament CVD diamond films. Robins et al. [28] identified thedefects in single crystals and polycrystalline diamond films grown by hot-filament CVDas 2.156. 2.326. 2.82. 3.188. and 1.673 eV. Collins et al. [29] compared the absorptionand cathodoluminescence spectra of single crystals and polycrystalline diamond filmsgrown by microwave-assisted CVD. A-band luminescence was observed in both cases.In addition. H3 centers were observed in the single crystal diamond and GRI in polycrystalline film. At this point. we may conclude that color centers occurring in natural orRPIIT synthetic diamond are also existent in the CVD diamond films.
4. Configuration of Diamond Lasers
4.1 PUMPING SOURCES
Usually, color center luminescence can be produced by electrons or photons. Both ofthem can serve as the pumping sources for diamond lasers. As reviewing the diamonddefects [12], it is noted that some color centers have both absorption and luminescence,such as H3. N-V and GR-I centers etc. Photon pumping can be applied to these centersfor making lasers. There are two types of photon sources: coherent (lasers) and incoherent (flashlamps). providing different benefits and drawbacks. A laser source provideshigh intensity with extremely narrow pumping band; therefore. the pumping efficiencyis high. The commercialized alkali-halide color center lasers utilized laser pumping forefficient operation. If a color center can be pumped at low saturation intensity (Is) andhas a broad absorption band. an economic and efficient diamond laser can be obtainedby flashlamp pumping. A high-intensity microwave excimer lamp with narrow-bandphoton output is the most suitable and efficient source for flashlamp-driven lasers [30].Some color centers without the absorption band. such as band-A emission. can bepumped by electron bealn (cathodoluminescence) or carrier injection (electroluminescence). Especially. the latter pump method is most suitable for active optoelectronicdevices.
4.2 OPTICAL CAVITY AND TUNABILITY
Figure 6 is a schematic of the first commercial color center lasers designed by K. German [2]. For optically pumped diamond lasers, the similar configuration can be applied.The left part is the crystal chamber with optical cavity and the right part is the tuningarm with the tuning components, a diffraction grating and an etalon. The most commontuning grating is used as a first-order restroreflector with a high selectivity of only a fewcm-} [31]. The etalon, a pair of partially transmissive plane-parallel mirrors. is oftenused to increase the overall selectivity. Moreover, other tuning elements (such as prisms
200
and birefringence plates) and techniques (e.g., mode locking) are also used for laser tuning [32]. Additionally, diamond lasers can be operated at room temperature without thecryogenic chamber. Hence, a compact system can be obtained.
Out: IR andVisible Tracer Beam
Coupling Window
PiezoelectricTunable Elalon
Tuning Arm
Crystal Chamber
Figure 6. Structure of the first commercial color center laser. [2]
4.3 LEDS AND LASER DIODES
A patent application of Burchard et al. [1] described the designs of diamond basedLEDs and laser diodes. Basically, there are three regions in such diamond photon emitting devices. First, there is an optically transparent semiconducting (p+- or n+- type)region which is created by using ion implantation with electrically active impurity andsuccessive annealing and etching, or by deposition of diamond films doped with electrically active impurities during growth on a diamond substrate. An intermediate region isnext defined by the diamond substrate (layer). And finally there is a collector regionwhich is created by ion implantation or metallization at a distance d from the semiconductor region exceeding the mean free path of charge carriers in the diamond substrate.As described in Figure 7, 1 indicates the optically transparent semiconducting region,which may be produced by ion implantation (a) or CVD methods (b and c), while 2indicates the intermediate diamond region and 3 the collector. Arrow 7 indicates thedirection of cathodoluminescence output and arrow 10 is the laser output from two polished opposite lateral surfaces with a totally and a partially reflecting mirror. Recently,developments of vertical-cavity surface emitting lasers (VCSELs) based on the III-Vsystems provide another configuration for diamond lasers. By high reflectivity of lasermirrors and an effective electron confinement in the vertical cavity, the threshold of surface emitting lasers in a room temperature cw operation can be reduced significantly[33]. Furthermore, a narrow circular Gaussian beam can be achieved by VCSELs. However, the difficulty with the electron (or hole) confinement of diamond devices has to besolved first.
201
3
@
2
3
11 -"d~_-3
8
Figure 7. A schematic illustration of possible light emitting structures. [1]
5. Excitons as a LaserMedium
Excitonic recombination radiation (ERR), ranging from 4.9 to 5.4 eV, provides the highest photon energy close to the energy gap in diamond [34][35]. An exciton is a boundelectron-hole pair due to coulomb force. It can travel through the crystal known as freeexcitons. If there is an impurity, a bound exciton can be formed with the hole (or electron) of an acceptor (or donor).
ERR usually occurs in high-purity diamond and has been used to examine the quality ofCVD diamond by cathodoluminescence [36]. Figure 8 illustrates a four-level exciton
202
transition. First, an electron in the valence band can be excited by a high-energy electronor photon from the valence band (level 0) to the conduction band (level 3) and decaysnon-radiatively to the exciton state (level 2). Then, a photon is produced due to therecombination of the electron-hole pair with subsequent relaxation by emitting phononsfrom level 1 to level O.
3
IE&
0--L
---- • E. --2.__ 't Uc1lun
Figure 8. The exciton transition and its possible four-level laser system. [36)
The interaction of phonons with the lower state (level 1) of the exciton transition causesthe multiple peaks in the exciton emission band. Figure 9 shows the ERR spectra forfour diamond samples. The notation A and B are assigned as free exciton (intrinsic)recombination associated with the transverse acoustic (TA) and the transverse optic(TO) phonons respectively. The notation D is assigned as bound exciton (extrinsic)recombination associated with TO phonons. The subscript number indicates the numberof phonons emitted in the relaxation of level 1. From the bandwidths and the distribution of these peaks, the laser action by exciton recombination may be tunable in a narrow range (2-3 nm) within these peaks extending from 4.9 to 5.4 eV.
Excitons can be formed in every insulating crystal. In most semiconductors, excitonsexist only in low temperatures due to their low binding energy. The situation for diamond is different since the binding energy of diamond excitons, 80 meV for free excitons and 140 meV for bound excitons, is high enough for them to survive at roomtemperature [34][37]. Moreover, ERR has been observed even in polycrystalline diamond films [36][37], which means high-quality polycrystalline diamond is sufficient toform excitons.
The other emission competing with the ERR in most high-quality diamond is the bandA emission. One can enhance the ERR by doping to form bound excitons or optimizing
203
Photon En~r9Y (~V)
5.0 5.2 5.4
PhOton Energy I~V)
SO 52 54
0,8,
~
2:>
~Ciii
~c..
'.. 0, £c..- (a) 8, lh ......= v 0, CI>.....v
(c) x I
0, Do(b) 8,
(d)x 1
250 240 230 250 240 230Wov~l~ngth (nm) Wov~l~ngth (rm)
Figure 9. Excitonic recombination radiation spectra in diamond at 90 K by cathodoluminescence. a. Natural type lIb (semiconducting) diamond; b. undoped CVD diamond particles with a diameterof 5 J.Un; c. boron-doped CVD diamond film with a concentration of 1000 ppm (B/C); d.boron-doped CVD diamond film with 100 ppm. [35]
Photon Energy (eV)2 3 4 5
::IF/A=15.c (d)..
C02=1.5"'"'-'>.....
(c)III 26c C02=1.0"~....c.-...::I (b)c..:>
C02=0.5" 16"0<VN
'"(a) 3.3
• C02=0"..0:z
600 500 400 300 200Wavelength (nm)
Figure 10. The intense free exciton emission of diamond particles forming by microwave plasma CVDwith CO (10%) and CO2 diluted with hydrogen. [38]
204
the growth parameters for the dominance of free exciton recombination (as shown inFigure 10) [36][38]. Actually, band-A emission is good for laser operation since a combination of ERR and band-A emission provides a tunability from UV to visible region.
6. Conclusions
Diamond lasers are one of the most promising applications for CVD diamond. Diamondcolor center lasers present several attractive features such as high power, high efficiency,a wide tunability and room temperature operation. The laser actions of diamond H3color centers are reviewed to demonstrate the scientific feasibility of diamond lasers.Moreover, many potential color centers in natural or synthetic diamond can also befound in CVD diamond films but no laser emission from CVD diamond films has yetbeen reported. High quality diamond having excitons can be used as a laser medium fora room-temperature UV laser to extend the diamond laser spectrum from UV to near IR.A brief review of laser physics indicates that laser actions usually occur with a largestimulated emission cross sections and a longer laser upper level lifetime. However, little understanding of excitons and most diamond color centers impedes the progress ofdiamond lasers.
7. Acknowledgment
The authors are grateful to Dr. S. C. Rand and Dr. H. Kawarada for providing manyimportant papers about diamond lasers and excitonic recombination radiation respectively. In addition, the figures used in the paper are with the permission of Dr. L. F. Mollenauer, Dr. S. C. Rand, Dr. A. M. Zaitsev, Dr. B. Burchard and Dr. H. Kawarada. Theircourtesies are acknowledged.
8. References
1. Burchard, B., Denisenko, A., Fahme, W., Melnikov, A., Varichenko, V., and Zaitsev, A., Diamond diodestructure, European Patent Application, Int. C.: HOIL 33/00, HOIS 3/18.
2. Mollenauer, L. F. (1985) Color center lasers, in M. L. Stitch and M. Bass (eds), Laser Handbook, 4, pp.143-228.
3. Rand, S. C. and DeShazer, L. G. (1985) Laser action of H3 color center in diamond, Tunable Solid StateLasers, Springer-Verlag, pp. 199-201.
4. Rand, S. C. and DeShazer, L. G. (1985) Visible color-center Laser in diamond, Optics Letters, 10(10),481-483.
5. Verdeyen, J. T. (1989), Laser Electronics, 2nd ed., Prentice-Hall International.6. Lin, L., Prelas, M. A., He, Z., Bahns, J. T., Stwalley, W. C., Miley, G. H., Batyrbekov, E. G., Shaban, Y.R., and Petra, M. (1995) Design of an rCF plant using a nuclear-driven solid state laser, to be publishedin Lasers and Particle Beams, 13(1).
7. Crossfield, M. D., Davis, G., Collins, A. T. and Lightowlers, E. C. (1974) The role of defect interactionsin reducing the decay time of H3 luminescence in diamond, J. Phys. C: Solid State Physics, 7, 1909-
205
1917.8. Thomaz, M. R and Davis, G. (1978) The decay time of N3 luminescence in natural diamond, Proc. R.
Soc. Lond. A, 362, 405-419.9. Redman, D., Shu, Q., Lenef, A. and Rand, S. C. (1992) Two-beam coupling by nitrogen-vacancy centersin diamond, Opt. Lett., 17(3),175-177.
10. Rand, S. R. (1994) Diamond lasers, in G. Davies (ed), Properties and Growth ofDiamond, INSPEC, theInstitute o(E1ectrical Engineers, London, United Kingdom, pp. 235-239.
11. Nakashima, T. and Yazu, S. (1990) Optical properties and laser action of H3 center in synthetic diamond, Diamond Optics Ill, SPill vol. 1325, 10-16.
12. Bridges, E, Davis, G., Robertson, J. and Stoneham, A. M. (1990) The spectroscopy of crystal defects: Acompendium of defect nomenclature, J. Phys.: Condens. Matter, 2, 2875-2929.
13. Davies, G. and Hamer, M. R (1976) Optical studies of the 1.945 eV vibronic band in diamond, Proc. R.Soc. Lond. A, 348, 285-298.
14. Davies, G., Thomaz, M. R, Nazare, M. H., Martin,M. M. and Shaw, D. (1987) The radiative decay timeof luminescence from the vacancy in diamond,J. Phys. C., 20, L13-L17.
15. Mironov, V. P., Martinovich, E. E, and Grigorov, V. A. (1994) Laser materials based on diamond withGRI centers,Diamond and Related Materials 3, 936-938.
16. Zaitsev, A. M., Varichenko, V. S., Melnikov, A. A., Denisenko, A. V., Burchard, B. B. and Fahrner, W. R.Luminescence of ion implanted diamond, unpublished paper, Belarussian State University, Minsk,Belarus, and University of Hagen, Germany.
17. Kawarada, H., Yokota, Y, Mori, Y., Nishimura, K. and Hiraki, A. (1990) Cathodoluminescence andelectroluminescence of undoped and boron-doped diamond formed by plasma chemical vapor deposition, J. Appl. Phys., 67(2), 983-989.
18. Fujimori, N., Nishibayashi, Y and Shiomi, H. (1991) E1ectroluminecsent device made of diamond,Japan. J. Appl. Phys. Part 1,30(8),1728-1730.
19. Khong, Y L., Collins, A. T. and Allers, L. (1994) Luminescence decay time and time-resolved cathodoluminescence spectroscopy of CVD diamond, Diamond and Related Materials, 3, 1023-1027.
20. Pereira, E. and Santos, L (1990) Dynamics of S3 luminescence in diamond," J. Luminescence, 45, 454457.
21. Pereira, E. and Santos, L. (1987) Long lived red luminescence in diamond, J. Luminescence, 38, 181183.
22. Kawarada, H., Nishimura, K., Ito, T., Suzuki, J., Mar, K. and Yokota, Y (1988) Blue and green cathodoluminescence of synthesized diamond films formed by plasma-assisted chemical vapor deposition,Japan. J. Appl. Phys. Part 1/,27(4), L683-L686.
23. Yacobi, B. G., Lebens, J., Vahala, K. J., Badzian, A. R. and Badzian, T. (1993) Optical characterizationof preferential incorporation of defects in monocrystalline diamond films, Diamond and Related Materials 2, 92-99.
24. Yokota, Y, Kotsuka, H., Sogi, T., Ma, J. S., Hiraki, A., Kawarada, H., Matsuda, K. and Hatada, M.(1992) Formation of optical centers in CVD diamond by electron and neutron irradiation, Diamond andRelated Materials 1, 470-477.
25. Freitas, J. A., Jr., Butler, J. E. and Strom, U. (1990) Photoluminescence studies of polycrystalline diamond films," J. Mater. Res., 5(11), 2502-2506.
26. Popovici, G., Khasawinah, S., Sung, T., Prelas, M. A., Spitsyn, B. V., Loyalka, S., Tompson, R., Chamberlain, J. and White, H. (1994) Raman scattering characterization of (100) and (111) oriented diamondfilms grown in the same run by hot filament chemical vapor deposition, J. Mater. Res., 9(11), 1·6.
27. Lin, L.-T. S., Popovici, G., Prelas, M. A., Spitsyn, B. V., Khasawinah, S., Sung, T., Y Mori and A.Hiraki (1995), to be published.
28. Robins, L. H., Cook, L. P., Farabaugh, E. N. and Feldman, A. (1989) Cathodoluminescence of defects indiamond films and particles grown by hot-filament chemical vapor deposition, Phys. Rev B, 39(18),13367-13377.
29. Collins, A. T., Kamo, M. and Sato, Y. (1990) A spectroscopic study of optical centers in diamond grownby microwave-assisted chemical vapor deposition, J. Mater. Res., 5(11), 2507-2514.
30. Lin, L.-T. S. (1994) Microwave and Nuclear Excitation of Alkali Metal Vapors, Ph.D. Dissertation,
206
Nuclear Engineering Program, University ofMissouri-Columbia.31. Mollenauer, L. F. and White J. C. (1992) General principles and some common features, in L F. Mol
lenauer, J. C. White and C. R. Pollock (eds), Tunable Lasers, Springer-Verlag, pp. 1-18.32. Mollenauer, L F. (1992) Color center lasers, in L F. Mollenauer, J. C. White and C. R. Pollock (oos),
Tunable Lasers, Springer-Verlag, pp. 225-277.33. Iga, K. and Koyama, F. (1993) "Venical-cavity surface emitting lasers and arrays, in G. A. Evans and J.
M. Hammer (eds), Surface Emitting Semiconductor Lasers and Arrays, Academic Press, Inc., pp. 71117.
34. Dean, P. J., Lightowlers, E. C. and Wight, D. R. (1965) Intrinsic and extrinsic recombination radiationfrom natural and synthetic aluminum-doped diamond, Phys. Rev., 140, A352-A368.
35. Kawarada, H., Yokota, Y. and Hiraki, A. (1990) Intrinsic and extrinsic recombination radiation fromundoped and boron-doped diamnods formed by plasma chemical vapor deposition, Appl. Phys. Lett.,57(18), 1889-1891.
36. Kawarada, H. and Yamaguchi, A. (1993) Excitonic recombination radiation as characterization of diamonds using cathodoluminescence,Diamond and Related Materials 2, 100-105.
37. Kawarada, H., Matsuyama, H., Yokota, Y., Sogi, T., Yamaguchi, A. and Hiraki, A. (1993) Excitonicrecombination radiation in undoped and boron-doped chemical-vapor-deposition diamonds, Phys. Rev.B, 47(7), 3633-3637.
38. Kawarada, H, Tsutsumi, T., Hirayama, H. and Yamaguchi, A. (1994) Dominant free-exciton recombination radiation in chemical vapor deposted diamonds, Appl. phys. Lett., 64(4), 451-453.
ADVANCED APPLICATIONS OF DIAMOND ELECTRONICS
Christopher B. WallaceAdvanced ResearchBDM Federal, Inc.1801 Randolph Rd SEAlbuquerque, NM 87106
Recent advances in the ability to generate high mobility diamond films coupled with thedevelopment of nanophase diamond materials creates opportunities for advanceddiamond electronic device applications. The present paper describes new electronicsapplications resulting from these innovations which include: diamond nanotechnology,high bandwidth, high power RF switches, and high speed analog to digital electronicciIcuits.
1. Introduction
Interest in diamond as a semiconductor material stems from its well known potential toconstruct high power, high speed, high temperature, optical and microelectronic devicesresistant to environmental damage. Ideally, diamond is - 8200 times better thansilicon, - 1200 times better than gallium arsenide, and - 500 times better than indiumphosphide for high power, high frequency operation. Diamond transistors will alsoperhaps be capable of speeds - 32 times that of silicon, - 70 times that of galliumarsenide, and - 53 times that of indium phosphide [1,2]. Semiconductor devicetechnology continues to advance so rapidly it is important to overcome those factorswhich limit widespread application of new technology; for example, the achievement ofmaximum packing densities, speeds and frequencies of operation and the powerhandling capability for a given device. The extreme thermal, mechanical, chemical,and electrical properties of diamond allow the types and numbers of current andconceivable semiconductor device applications to be extended in new directions.As a simple example of the technical benefit, consider the maximum integrated
circuit packing fraction on a given chip. Two factors determine achievableperformance for a given chip size. One, how small can the fabricated circuit elementbe? This is a function of the writing technology, Le. the smallest circuit element ordevice that can be etched into the semiconductor and its associated transport properties.Two, how many circuit elements can be written in a given chip volume before jouleheating from the resistive electrical power loss damages the semiconductor? This is afunction of the thermal conductivity and heat capacity of the semiconductor. Diamondis approximately 8200 times better than silicon and 500 times better than indiumphosphide at handling high power, or high thermal stress. The integrated ciIcuitpacking fraction scales directly with these ratios while the computational speed of agiven integrated circuit scales with the packing fraction.
207
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 207-217© 1995 Kluwer Academic Publishers.
208
Recent advances in the ability to generate high mobility diamond films [3] coupledwith the development of nanophase diamond materials [4] creates opportunities foradvanced diamond electronic device applications. The present paper describes newelectronics applications resulting from these innovations which include: diamondnanotechnology. high bandwidth. high power RF switches. and high speed analog todigital electronic circuits.
2. Diamond Nanotechnology
"At any rate, it seems that the laws ofphysics present no barrier to reducing the size ofcomputers until bits are the size ofatoms and quantum behavior holds dominant sway."
R.P. Feynmann
Nanometer scale devices operate on phase coherent principles if the fabrication size of agiven structure is on the order of. or smaller, than the phase breaking length of the
electrons, L~ = ..jD~iIt. It is extremely important to be able to experimentally
characterize both quantities under the radical. The diffusion coefficient is readilyestimated via the Einstein relations for the hole and/or carrier mobilities.Unfortunately, a review of the literature yields no report of measured values for theinelastic relaxation time. Consequently, it is important for future developments thatresearchers involved in the growth and characterization of diamond fIlms measure theabove variables in tandem.
As an example. if the phase breaking length is equal to, say, the 4 nm scale size of ananophase diamond particle, the carrier or hole wavefunctions will be coherent overthis scale for D - 0.0016 m2s-1 and't - 10-14 s. This value for the diffusion coefficientis easily achievable while the choice of inelastic scattering time constant is believed tobe conservative.STM measurements have been successfully performed on boron doped diamond for
B/C ratios - 4 ppm or greater [5]. Consequently the presence of a few boron atoms in aseveral cubic nm volume is more than sufficient to generate a useful hole mobility and,in fact, the addition of a dilute impurity introduces some interesting physics.To illustrate, consider a nanophase diamond particle 10 nm or so in diameter which
has one boron atom as a dopant. Assume the impurity has been activated (e.g.optically) and there is now a single bound electron-hole pair. or Wannier exciton ,interacting with the nanophase diamond particle and its external environment. Theexciton possesses some interesting properties. First, the exciton is, of course, analogousto the hydrogen atom in a quantum mechanical sense providing the effective electronand hole masses are substituted into the usual expression for the hydrogen atom reducedmass. The bound state radii are give by [6]
2 ma" =n E,--aom*
and the associated energy levels are
(I)
209
e2
E =-- (2)" 2erQ"
where n=1,2,3,... and ao is the Bohr radius. The effective exciton mass, for diamond, is- 0.11 m [7], where m is the free electron mass. Consequently the ground stateexcitonic radius is - 2.7 nm with an associated binding energy - 47 meV and effectivecapacitance - 1.7xl0-18 F. This binding energy and capacitance are sufficient topermit the ground state to easily persist at significant temperatures. The fIrst excitedstate radius would lie outside the particle and the binding energy would drop to - 11.6meV. In actuality, once the exciton acquired an appreciable probability density at aradius on the order of the particle radius, the dielectric screening will vanish and theeffective mass will increase thereby causing the excitonic wavefunction to be localizednear the particle surface. The transition energy would be - 36 meV and the remainingbinding energy - 11 meV. Note that if the diamond particle size were increased and thesystem cooled to prevent thermal lattice vibrations from dissociating the exciton, theorbital radii could become quite large. In fact, as a matter of basic principle, highlyexcited "Rydberg" excitons could be created and studied in the laboratory.The technological applications associated with the ability to create single excitons in
nanophase diamond particles are, in the author's opinion, extremely attractive.Consider the confIguration shown below in Figure 1.
OPTICAL WAVEGUIDE
NANOPHASE DIAMOND~
~
~~
~
~
~
~,INSULATINGSUBSTRATE,,,
"CONDUCTING SUBSTRATE
FIGURE 1. Simple Illustration of an Opto-electronic Transistor (Switch)UsingSTM, Optical Waveguide, and Nanophase Diamond.
210
A nanophase diamond particle with a single acceptor impurity rests on an insulatingsubstrate over a conducting backplane. A STM and optical waveguide tip reside within- 0.5 nm or so of the nanophase diamond particle surface. The distance between theparticle and the conducting substrate is similar. When the impurity is not activated, theSTM tip detects no current. The optical waveguide provides both the means foractivating the impurity and inducing transistor (switching) action. Two light sources,one at 360 meV, the other at 36 meV, activate the impurity (create the exciton) andcause the transition to the fIrst excited state. The optical wave guide tip is several nm insize hence the principle of operation is similar to that used in a Near Field ScanningOptical Microscopy (NSOM) [8]. The photon wavelengths required can be provided byInSbAsP and PbSnSeTe tunable diode lasers [9].Operationally, the concept is straightforward. In the unactivated state and the
excitonic ground state the tunneling current essentially vanishes. Recall that thetunneling current is proportional to exp[lCd], where d is the barrier width and 1C is thedecay constant (negative) for the carrier wave function in the barrier. When theimpurity is unactivated, the barrier width is essentially the particle radius and the STMtip-particle distance. The barrier height is the impurity ionization energy measuredfrom the top of the valence band plus the effective local work function. The excitonicground state is characterized by a barrier height - 47 meV plus the effective local workfunction and the barrier width is - (rp-an)+O.S [nm]. The decay constant within theparticle can be estimated as
1C - 21 1 (1,(r2 1Vt(rt)r Vt(r)
while that outside the particle (assume vacuum)
1C -li-I.v2m<p .
(3)
(4)
Here, m is the free electron mass and ell is the effective local work function. Currentresearch suggests that the work function for diamond can be zero, or negative,depending on the surface orientation of the diamond [10]. Even though the electronaffInity of nanophase diamond has yet to be determined, it does no violence to thisanalysis if we take it to be zero. For now, we take the normalized radial wavefunctionsto be the same as for hydrogen except that the Bohr radius is replaced by the excitonicvalue
(S)
(6)
where Sex=al and equations S and 6 show the ground and fIrst excited excitonic statesfor zero orbital angular momentum. A more detailed calculation which takes intoaccount the different boundary conditions between the diamond particle and theexternal environment will be published elsewhere [11].Asymptotically, for large r, the decay constant tends to - const/Sex for the ground
and first excited exciton states. Once d becomes larger than the exciton "radius", the
211
tunneling probability drops significantly; hence, the ground state exciton has a strongtendency to remain localized within the diamond particle. The frrst excited state,however, drives the probability for tunneling to near unity. This is because theeffective surface work function is quite small and the exciton "radius" becomeslocalized at the diamond particle surface. In essence, the transition from the groundstate to the frrst excited state causes the electron partner in the exciton to tunnel throughthe gap between the STM tip and the diamond particle. Similarly, to enforce chargeneutrality, a replacement electron should tunnel from the conducting substrate to thediamond particle. Consequently, the configuration illustrated in figure I functions as atrue opto-electronic transfer resistor, or switch, with essentially no impedance in the"on" state.The large phase breaking length for electrons and holes in diamond, the ability to
synthesize nanophase diamond particles, and, in particular, the existence of stableexciton structures at greater than room temperatures in doped diamond creates a newopportunity to develop a true electronic nanotechnology without expensive cryogenicsand costly apparatus.
3. High Bandwidth, High Power, Optically Triggered Switches
The previous section explored the application of diamond technology in the realm ofthe small. We now focus on an application area where the macroscopic advantages ofdiamond CVD technology can play an important, enabling role.Pulsed electrical systems play an important role in many areas of modem science
and technology. A key factor which limits the applicability of extremely high powerpuIsed electrical technology for a variety of applications is the switch used to transferthe electrical energy from the pulse fonning network to the load. Searching for theperfect switch is an unsatisfiable quest, however, the growing field of research intooptically activated semiconductor switching promises a significant step toward theresolution of that quest. Ultimately, it may make the potential benefits of pulsed powertechnology available outside the confines of the research laboratory.The recent book by Rosen and Zutavem [12) provides an excellent introduction and
review of the state of the art in high power optically activated switches. Presently, themost interesting solid state switch technology utilizes optically triggered GaAs althoughresearchers have investigated optically triggered natural diamond and have also injectedan electron beam directly into the diamond to close the switch.The advantage to using diamond as the switching material is straightforward. Given
a fIXed laser power, the power delivered to the load is maximized if the load impedanceis adjusted to half the switch resistsance when the switch is on. The maximum power tothe load is then [13)
V 2 ( )2P _ b _ /l'r Vb PIoad,max - 8R
s- 81iro / e d las., (7)
where J.1 is the charge mobility, 't is the carrier lifetime, e is the electronic charge, d isthe separation distance between the electrodes, and Vb is the bias voltage across theswitch. The limit on the bias voltage is set by the surface breakdown voltage of theswitch material and can be quite large in the case of diamond.
212
While the GaAs work appears promising, several problems must be solved in orderfor the technology to become widespread. First, and foremost among these is thedeleterious effect of semiconductor plasma instability on reliable switch performance.Second, the poor thermal conductivity coupled with the residual resistivity of thetriggered switch limits the amount of voltage and current which may be transferred bya single switch. After a brief review of the plasma instability problem, we present anew configuration for a high power optically triggered diamond switch.Plasma instabilities in semiconductors are well known [14] and are a manifestation
of negative differential conductivity (ndc). Generally, two different ndc behaviors areknown to occur. These are illustrated in Figure 2.
j2
jl
J
[J~j ...
E el E e2
Figure 2. S and N j/E Characteristics Cause Filamentation and Domain Instabilities.
The left-hand S characteristic favors formation of current filaments while the right-handN type characteristic gives rise to electric field domain (virtual cathode) formation. Ineither case the semiconductor behavior splits spatially into two different regions, eachof which possesses its own subsequent dynamics. The figure emphasizes the relationbetween j and E since it is rare that IN characteristics serve as experimental signpostsof this unstable behavior.The appearance of instabilities in semiconductor electric current flow is associated
with nonlinearities in the charge carrier transport process, particularly in narrowbandgap semiconductors. Charge carrier densities which change in the presence of ahigh bias voltage are easily generated for current density amplitudes below thatrequired for simple Wunsch-Bell thermal failure. The charge carriers in the conductionband undergo recombination and autocatalytic generation interactions with the groundand excited states of shallow impurity centers, therefore the nonlinear nature of the
213
instability depends on the doping profile of the device and. perhaps. subtle details of themanufacturing process.The primary instability manifestation in the GaAs switches is the creation of current
filaments where the current density becomes so large that Wunsch-Bell thermal failurecomes into play; the semiconductor melts along the filament path and the switch isdestroyed. Virtual cathode. or electric field domain instabilities have also beenobserved. but only rarely.Filamentation results from the unstable nature of the attractive interaction between
current elements of like sign in circumstances where the total current is neutralized[15]. Transverse perturbations caused by microscopic irregularities in the surfaceelectrodes create a situation where the uniform current density flow across the switchcollapses into one or more channels of intense localized charge flow. The residualsemiconductor resistivity provides the mechanism whereby the local temperature risesdramatically in the current filament. thermal runaway occurs. and the semiconductormelts. and/or scores. a permanent channel thereby destroying the switch.A configuration which solves the above problem utilizing diamond technology is
shown in Figure 3 below.
hv - 0.38 eV -> InSbAsP Diode Laser
OHMIC CONTACT OHMIC CONTACT
-,,,#
Lightly Doped
Photon AbJorbtion Depth
Moderately Doped
Figure 3. Differentially Doped Diamond Optically Activated Semiconductor Switch.
214
The classical method which prevents filamentation formation is to control thecurrent neutralization fraction. Given that no background magnetic field permeates theswitch, the current neutralization fraction must be less than a geometry dependentconstant in order that the self field prevent the instability from forming. This isaccomplished by differentially doping the switch regions as shown in the above figure.A primary cause of the initial perturbation which creates subsequent filamentation
problems is the ohmic contact mask which provides the connection with the pulsepower circuit. The lateral contact interface between the metal contacts and thesemiconductor is irregular on a mesoscopic scale due to the nature of the depositionprocess. The irregularities seed the instability since the underlying semiconductor isnormally uniform.Differential doping with diamond as the base wide band gap semiconductor
efficiently resolves several technical problems. The conductance of the twodimensional well structure is modulated by varying the mobility. The mobility dependson the scattering rates of electrons which depends explicitly on the wavefunctionswithin one coherence length of the boundary. The external bias voltage applies theexternal electric field needed to push the electronic wavefunctions into the differentiallydoped region which changes the scattering rates and mobility. The high scattering ratein the moderately doped region suppresses the formation of coherent instability near theboundary between the two reions while the higher mobility within the lightly dopedregion provides for lower ohmic loss (lower switch on resistance) and space chargeneutralization. Most importantly, the switch current is not completely neutralized,therefore the filamentation instability is further suppressed. Also, overlaying the ohmiccontacts on a moderately doped region (created by differential thermal and/or electricfield assisted diffusion) which is slightly larger than the contact area reduces thelikelyhood of spatial irregularities which can act as instability sources.Choosing diamond as the base semiconductor reduces the possibility of surface
flashover and allows the switch to stand off a higher voltage in a given geometry.Diamond's higher resistivity reduces the possibility of thermal runaway.
4. High Speed Analog to Digital Electronic, ,Circuits
The final advanced electronic device application to be considered in this paper is a highspeed analog to digital electronic circuit. The basic design was explored byBandyopadhyay and co-workers (16) and is schematically shown in figure 4 below.The device relies on 2D ballistic transport, ensemble averaging over the electron's
transverse wavevector. Charge carriers created by the source diffract through the slitand are collected on the "drain" consisting of the closely spaced fingers. Thediffraction pattern (and the current collected by the fingers) is changed by modulatingthe slit width. The analog voltage signal from the external environment is applied tothe gate pads which constrict the slit by modulating the associated depletion layer. Asthe slit width changes, the diffraction patter shifts and the current amplitudes in thefmgers are modified. Fixing the source current and voltage allows the current level inthe fingers at zero gate voltage to represent one "bit" of signal, hence the deviceactively transduces the analog signal into a bit pattern, or "digital" signal at the drain.The beauty of the entire device lies in replacing an entire analog to digital converterelectronic set by a single transistor.The QUADFET is inherently two dimensional and because it requires ballistic phase
coherent charge transport, the temperature at which it can operate, and the material
215
from which it is fabricated will play a significant role in the eventual commercialacceptance of the technology it represents. Presently, the best available material forQUADFET fabrication is GaAs. Given a typical 2D carrier concentration - 1011 cm-2,the operating temperature is -77 K. Consequently, the device is of limited commercialinterest. QUADFETs made with diamond, however, offer the potential for operation atroom temperature. The issue is the ability to resolve successive diffraction minima.Diamond is unique in this regard since the phase coherence length can be quite large
at room temperature. GaAs, however, is plagued by such a high scattering rate fromdefects (even in ultrapure intrinsic material) that the relationship between the slit width,
ak . 1a, and the wavelength which enters the well known formula, - sm q>=m +- for
21r 2the location of the diffraction minima is the Fermi wavelength for the given twodimensional carrier concentration. Therefore, diamond offers the potential to be far lessrestrictive in the lithography limits imposed by the need to resolve the diffraction and tobe able to operate at room temperature thereby enabling the development of acommercially attractive single transistor NO converter.
FINGERS
GATE
SOURCE
Figure 4. Quantum Diffraction Field Effect Transistor (QUADFEl)
216
5. Conclusions
Recent progress in fabricating nanophase diamond and controlling purity encouragesthe investigation of diamond devices in the realm of the very small. These deviceswould operate on the principle of quantum ballistic transport at room temperature. Theadvances reported in these proceedings suggest that diamond may make the nascentdiscipline of electronic nanotechnology practical, possible, and commercially attractive,thereby engendering a new revolution in electronics for the 21st century.
6. References
1. E.O. Johnson, (1965) Physical Limitations on Frequency and Power Parametersof Transistors, RCA Rev., 26, 163-177
2. R.F. Davis, Z. Sitar, B.E. Williams. H.S. Kong. HJ. Kim. J.W. Palmour. J.A.Edmond. J. Ryu. J.T. Glass. C.H. Carter Jr.• (1988) Critical evaluation of theStatus of the Areas for Future Research Regarding the Wide Band GapSemiconductors Diamond, Gallium Nitride. and Silicon Carbide. Materials
Science & Engr.• Bl. 77-104
3. M.A. Prelas and B.V. Spitsyn. private communication
4. M.A. Prelas. private communication
5. X.H. Wang. G.-H.M. Ma. Wei Zhu. J.T. Glass. L. Bergman. K.F. Turner. RJ.Nemanich. (1992) Effects of Boron Doping on the Surface Morphology andStructural Imperfections of Diamond Films.Diamond and
Related Materials.l. 828-835
6. H. Haug and S.W. Koch. (1993) Quantum Theory ofthe Optical and ElectronicProperties ofSemiconductors, 2nd Edition, Wodd Scientific. Singapore
7. S.M. Sze, (1981) Physics ofSemiconductor Devices. John Wiley & Sons. NewYork
8. E. Betzig and J .K. Trautman. (1992) Near Field Optics: Microscopy.Spectroscopy. and Surface Modification Beyond the Diffraction Limit.Science. 257. 189-195
9. A.I. Nadezhdinskii and A.M. Prokhorov. (1992) Modem Trends in Diode LaserSpectroscopy, SPIE Vol. 1724: Tunable Diode Laser Applications. 1724, 2-13
10. J.L. Davidson, (1994) Diamond Electrical Properties and Electronic DeviceBehavior. in K.E. Spear and J.P. Dismukes (eds), Synthetic Diamond: Emerging
CVD Science and Technology, pp. 355-399; see also M.W. Geis. (1991)Diamond Transistor Performance and Fabrication, Proc./EEE, 79(5), 669-676;F.J. Himpsel, J.A. Knapp. J.A. VanVechten. D.E. Eastman, (1979) QuantumPhotoyield ofDiamond (Ill) A Stable Negative Affinity Emitter, Phys. Rev. B,20.624-627; and B.B. Pate, (1986) The Diamond Surface: Atomic and
217
Electronic Structure, Surface Sci., 165, 83-142
11. C.B. Wallace, to be published
12. A. Rosen and F. Zutavem (1994) High-Power Optically Activated Solid StateSwitches, Artech House, Boston
13. P.T. Ho, J. Goldhar, (1994) Photoconductive Switching Using Diamond andZinc Selenide, in A. Rosen and F. Zutavem (eds), High-Power OpticallyActivated Solid State Switches, pp. 81-93, Artech House, Boston
14. J. Pozhela, (1981) Plasma and Current Instabilities in Semiconductors.Pergammon, Oxford
15. R.C. Davidson, (1990) Physics ofNoneutral Plasmas. Addison-Wesley,Redwood City, CA
16. S. Badnyopadhyay, G.H. Berstein, W. Porod, (1989) Quantum Devices Basedon Phase Coherent Lateral Quantum Transport, in M.A. Reed and W.P. Kirk(eds),Nanostructure Physics and Fabrication, pp. 183-188. Academic Press,Boston
LASER-ASSISTED CHEMICAL ETCHING OF DIAMOND FILMSIN OXYGEN
V.G. RALCHENKO, K.G. KOROTUSHENKO, A.A. SMOLINand E.D. OBRAZTSOVA
Institute of General Physics, Russian Academy of Sciences,ul. Vavilova 38, Moscow 117942, Russia
E.N. LOUBNIN
Institute of Physical Chemistry, Russian Academy of Sciences,Leninsky prosp.31, Moscow 117915, Russia
ABSTRACT. The patterning of diamond films is required in manyapplications, in particular, in fabrication of diamond-based electronicdevices. We report on the use of low power (:=:::2 W) continuous wave Ar+laser for etching (engraving) of diamond films via laser-induced localoxidation reaction. Smooth fine-grained diamond films of about 10 f...£mthickness have been grown on Mo and Si substrates in a DC arc dischargeusing CH4 /Hz gas mixtures. The sharply focussed laser beam was scannedunder a computer control over the film surface causing the diamondoxidation in air or pure oxygen atmosphere. Various trenches and holes of afew microns size have been produced at high etch rate.
1. Introduction
Fabrication of diamond-based sensors, transistors and other electronicdevices involves a procedure of fine patterning of diamond films. Several dryetching techniques, including reactive ion etching [1,2], etching bycatalytically active metals [3,4], and laser ablation [5-7] have been developedfor this purpose. Ablation of diamond proceeds through a fast heating ofmaterial by ultraviolet laser pulse and successive vaporization of carbon withrates of up to hundreds nanometers per pulse. In the present paper wedescribe the applications of a low power continuous wave Ar+ laser forengraving and microdrilling of CVD diamond films via diamond burninginduced within small laser spot. The direct writing technique is fast andflexible, no photomask is needed for patterning, as opposed to pulsed laserprojection method. Previously we demonstrated this technique for patterningof diamond-like carbon films on silicon wafers [8].
219
M.A. Prelas et at. (eds.), Wide Band Gap Electronic Materials, 219-224© 1995 Kluwer Academic Publishers.
220
2. Experimental details
Fine grained diamond films with crystallite dimensions smaller than 1micron were grown using a DC arc discharge system [9] in CH4/H2 mixtures.Typical methane content in reaction gas was 5%. Before deposition themolybdenum and silicon substrates of 10xl0xl mm3 dimensions were seededwith 5 nm diamond particles by treatment in ultrasonic bath with ultrafinediamond powder to enhance nucleation density to N::::I01O cm-2 [10]. Thediamond films of 10 J..Lm thickness were obtained in 2 hours deposition time.For post-growth etching the diamond samples were placed in a miniaturevacuum chamber, that could be filled by oxygen, air or argon gas. A finelyfocused beam of an Ar + laser operated at all green lines (457-514 nmwavelengths, beam spot 2 J..Lm, laser power up to 2 W) was introduced intothe chamber and scanned over the diamond surface under computer controlat a velocity of up to 4 mm/s. The process of the direct laser writing wasobserved with an optical microscope. The as-deposited and patterned filmswere examined with scanning electron microscopy and Raman spectroscopy.
3. Results
When the irradiation was carried out in air or pure 02 atmosphere the lasertriggered oxidation occurred, resulting to local etching of diamond, while noetching effect was observed upon laser processing in vacuum or inert gasatmosphere. For this reason the majority of experiments was carried out inair. Various trenches of a few microns width were produced with etch rateof up to hundreds microns per second. As an example a brunch of zigzagchannels with width less than 10 J..Lm etched in the film is shown in Fig.I. Aseries of narrow trenches of different depth have been fabricated bychanging the scanning velocity of laser beam as shown in Fig. 2. The mostdeep trench has been etched through the entire film at slow scanning rate, sothat the molybdenum substrate has been exposed after laser treatment. Thewalls of the trenches were smooth, their width was constant along the trace.The depth of channels depends on laser power, beam scanning velocity
and number of passes along the same trace. The depth d increasesapproximately linearly with laser power, and decreases with scanning rate.The dependence of d on scanning velocity V is shown in Fig. 3. After 16laser passes the 5 J..Lm deep trenches were obtained at V=0.06 mm/s, whileshallow trenches (d=O.5 J..Lm) were observed for fast scanning at V=4 mm/s.Raman spectroscopy revealed a certain graphitization of laser-treated
diamond. Raman spectra were taken from a trench bottom by focussing theprobe beam of an Ar+ laser (488 nm wavelength) at a spot of 2 J..Lm indiameter. The Raman spectra for diamond film before and after laseretching are shown in Fig. 4. As-grown film displays four main peaks typicalfor fine-grained CVD diamond. The peak of diamond at 1332 cm-1 is
221
imposed on two broad peaks around 1350 cm-1 and 1550 cm-1 assigned toamorphous carbon. Additional peak at 1150 cm-1 is characteristic fornanocrystalline diamond.
Figure 1. A brunch of narrow zigzag channels etched in diamond film.
Figure 2. A series of 5 1J.m wide trenches etched at different scanningvelocities of laser beam. Scanning velocity increases from right to left.
222
7
6
!Laser power 0.75 W
5 !16 passes
§-L. 4
-B 3 !c.
!Q,l
~2 ! !1 ,00.01 0.1 1 10
Scanning velocity, mm/s
Figure 3. The dependence of trench depth on beam scanning velocity. Eachtrench has been etched by 16 laser passes along the same trace at laserpower 0.75 W.
1000 1200 1400 1600Raman shift, cm-1
1800
Figure 4. Raman spectra of diamond film before and after laser etching.
223
Diamond peaks are absent for laser etched film, only peaks at 1350 cm-1
and 1580 cm-1 known for glassy carbon with graphitic structure are observed.We conclude that diamond etching proceeds through a surfacegraphitization, similar to the case of pulsed laser ablation [6]. It isinteresting, however, that only weak graphitization took place for one specialsample (not discussed here) with a morphology intermediate betweennanocrystalline and polycrystalline ones.
Figure 5. An array of 1 J.Lm holes burnt off in diamond.
A vanatiOn of residence time of the beam at a given site allows theetching of more complex patterns with a sharp and/or gradual change in thedepth. A simplest example of this sort is represented in Fig. 5, where aregular array of 1 J.Lm holes burnt off in diamond is shown. This techniquecould be applied for fabrication of vias in diamond-based multichip modules.
4. Conclusions
A direct laser writing technique has been used for patterning of CVDdiamond films. The etching mechanism is the intensive oxidation reactioninduced locally by micron-sized beam of continuous wave argon ion laser.No etching effect was observed upon irradiation in vacuum or an inert gasatmosphere. Etch rate dependence on laser parameters was measured in the
224
experiments. Various trenches and holes of a few microns SIze werefabricated in 10 J.1.m thick diamond film on Si and Mo substrates.
5. References
1. Werner, M., Schlichting, V., and Obermeier, E. (1992) Etching ofpolycrystalline diamond and amorphous carbon films by RIE, Diamondand Relat. Mater. 1,277-280.
2. Shikata, S., Nishibayashi, Y., Tomikawa, T., Toda, N., and Fujimor~ N.(1993) Microfabrication technique for diamond devices, in M. Yoshikawa,M.Murakawa, Y. Tzeng and W.A. Yarbrough (eds), Proc. 2nd Int. Con! onApplications of Diamond Films and Related Materials, MID, Tokyo, pp. 377380.
3. Ralchenko, V.G., Kononenko, T.V., Pimenov, S.M., Chernenko, N.V.,Loubnin, E.N., Armeyev, V.Yu., and Zlobin, A.Yu. (1993) Catalyticinteraction of Fe, Ni" and Pt with diamond films: patterning applications,Diamond and Relat. Mater. 2,904-909.
4. Jin, S., Graebner, J.E., Tiefel, T.H., and Kammlott, G.W. (1993) Thinningand patterning of CVD diamond films by diffusional reaction, Diamondand Relat. Mater. 2, 1038-1042.
5. Ageev, V.P., Bouilov, L.L., Konov, V.I., Kuzmichev, A.V., Pimenov, S.M.,Prokhorov, A.M., Ralchenko, V.G., Spitsyn, B.V., and Chapliev, N.1. (1988)Laser interaction with diamond films, Soviet Physics-Doklady 33, 840-842.
6. Pimenov, S.M., Smolin, A.A., Ralchenko, V.G., and Konov, V.1. (1993)Excimer laser processing of diamond films, Diamond Films and Technol. 2,201-214.
7. Johnston, c., Chalker, P.R., Buckley-Golder, I.M., Marsden, PJ., andWilliams, S.W. Diamond device delineation via excimer laser patterning(1993) Diamond and Relat. Mater. 2, 829-834.
8. Armeev, V.Yu., Chapliev, N.J., Lubnin, E.N., Mikhailov, V.I., Ralchenko,V.G. and Strelnitsky, V.E. (1991) Ar+ laser annealing and etching ofhydrogenated amorphous carbon films, Surface Coat. Technol. 47, 279-286.
9. Chapliev, N.I., Konov, V.I., Pimenov, S.M., Prokhorov, A.M., and Smolin,A.A. (1991) Laser-assisted selective area deposition of diamond films, inY. Tzeng, M. Yoshikawa, M. Murakawa and A. Feldman (eds.),Applications of Diamond Films and Related Materials, Elsevier SciencePublishers, Amsterdam, pp. 417-421.
10. Smolin, A.A., Ralchenko, V.G., Pimenov, S.M., Kononenko, T.V., andLoubnin, E.N. (1993) Optical monitoring of nucleation and growth ofdiamond films, Appl. Phys. Lett. 62, 3449-3451.
ION MILLING OF POLYCRYSTALLINE DIAMOND FILMS
AE. ALEXENtO, AF. BELYANIN*, L.L. BOUILOV,AP.SEMENOv**, B.V. SPITSYN
Institute ofPhysical Chemistry ofthe RussianAcademy ofSciences.Russia, 117915, Moscow, Leninskiypr., 31.
•Central Research Technological Institute.Russia, 121355, Moscow, Ivan Franko str., 4.••Buryat Institute ofNatural Sciences ofthe Siberian Division
ofthe Russian Academy ofSciences.Russia, 670042, U/an-Ude, Sakhjanova str., 6.
One of the methods of the diamond processing is the ionmilling. By ion bombardment one can obtain the elements of surfacetopography ofmicron and submicron dimensions, the sharp edges of the~rooves having absolutely vertical walls and optically smooth surfacesfonned on the removal ofa diamo)ld layer.
The present paper gives consideration to various methods ofmechanical treatment of diamond fihns (milling, polishing, drilling)by means of the ion beam. The ion beams of Ar+ or Ar++0+ were
extracted from the plasma of a gas-discharge source with cold hollowcathode [I,2].
The set for ion milling.
The work was conducted using the standard vacuum set (VUP-4)
equipped by a plasma ion source. The ion source contains a hollow
225
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 225-234© 1995 Kluwer Academic Publishers.
226
4'---1-"'?lI'=~
\,-t-~---12
lOmm'"-'
Fig.I. Construction ofthe ion source [1,2].
cathode I (Fig.I), an emitter cathode 2 and a cylindric anode 3. Thecathode I has a cavity 4 mm in diameter and 40 mm in length. In the-cathode 2 there is an emission opening 4 18 mm in diameter. Thecathodes 1 and 2 are made ofmagnetic steel and seNe as pole pieces ofaring permanent magnet (SmCos) 5. The magnet provides thelongitudinal magnetic field of 0, I T in the anode cavity. The cathodesare connected mechanically and electrically by a non-magneticmetal yoke 6, fixed by the nut 7 on the high- voltage caprolone insulator8. The anode 3 is insulated from the cathode I by teflon washers. Theinsulator 8 is mounted on a cover 9 where the cable inputs 10 and waterandgas ifiIets -~placed.The casiniIlis connected with the insulator 8 bymeans of a joint vacuum seal. The ions are extracted from thedischarge plasma by an accelerating electrode 12 with the aperture of20 mm in diameter. Owing· to the thread joint of the acceleratins
227
electrode 12 with the face part of the casing 11, the regulation of a gapbetween the emitter cathode 2 and the accelerating electrode 12 whichdetermines the length of the accelerating inteIVal is simplified. A flexiblegas line 13 is placed into a tube electric screen 14. The initiation of thedischarge in the gas line is prevented by filling it with quartz sand. Theheat is abstracted from the hollow cathode and the magnet through anoil heat-transfer agent to a heat-sink 15 plunged into the oil andcooled by running water. The anode 3 is.connected with a high-voltagecable through a metal-ceramic vacuum cUrrent input 16 welded to thehollow cathode. The construction permits the ion source to be mountedon the vacuwn camera in such a way that it can have any spaceorientation. The coupling nut 17 makes the source· constructionseparable easily from the vacuum camera~ The ion source starts bymeans of two stabilized rectifiers with a smooth adjustment of the outputvoltage 0-1,5 kVand 0-10 kV. The working gas leaks in the cathodespace through a narrow duct. Due to the wide aperture of the cathode 2the longitudinal differential pressure between the acce1ex:ating inteIValand the space of the anode cylinder is absent. On giving the voltagefrom the power source 0-1,5 kV the Penning-discharge is initiatedbetween the hollow, circular cathodes and the anode. If the currentexceeds 10 rnA, the Penning-discharge plasma penetrates into thecavity due to the breakage of the cathode ion sheath in front of theaperture of the cavity. A discharge with hollow catode is initiated. The-pressure in the cavity ofthe anode cylinder is 0,13 Pa. The extraction ofthe ions is realized through the opening 18 mm in diameter by meansofthe accelerating electrode 12 having a hole 20 mm in diameter. Thelength of the accelerating inteIVal is 3 mm. The accelerating voltage isadjusted in the range 0-10 kV.
The scheme of the vacuum set with the ion source is presented inFig.2. The ion beam I is extracted from the source 2 to the vacuumchamber 3 by the electrode 4. To cany out the ion beam processing ofa film. the substrate is" placed on a water-cooled holder 5. The~between the nonnal to the substrate surface anel. the axis ofthe"ion
228
Fig.2. Set for the ionm.illing:
I-ion beam; 2-plasma ion source; 3-vacuum chamber; 4-extractingelectrode; 5-substrate holder; 6-discharge current source;
7-high-voltage rectifier; 8-gas vessel.
beam (<9) can be changedwithin 0-900
• The distance from the extractingelectrode 4 to the substrate fixed on the holder 5 is 40-70 mm.
The parameters of the ion processing of the diamond fIlms arepresented in the table 1.
229
TABLE 1. The generalized parameters ofthe ionmilling ofdiamond fIlms.
[4]
[3]
[3]
[6]
Literature[3-5][3-5]
[5][3-6]
[3][3][3][3]
[5,6][5]
0-86
5-6
1,551,251,880,31
25-81
0,032-0,783-4
ParametersIon current, rnAIon energy, keVPressure, Pa
Value5-104-52,7' 10-2
Ions Ar+,O+,Ar++O+
Angle (e) ofincidenceofions, deg.Thickness ofdiamondfilm, J.L1l1Roughness ofgrowth
surface, Rz, fJJ1lIonmilling rate, J.L1l1/he = 0° (Ar)e = 45° (Ar)e = 45° (Ar:02 =1:1)e = 86° (Ar)
Speed ofdrilling ofopening2mmin diameter (e = 0°), J.L1l1/h 1,43Roughness ofpolished
swface, J.LrtlRotation speed, s -1
Ion milling, polishing.
The diamond fIlms used in the work were grown on W substratesfrom the activated gas phase (H2+CH4 ) [7]. All the diamond layerssubjected to ion milling were polycristalline and consisted of the twocomponent axial texture <110> and <111>. The growth swface ofdiamond layers was developed strongly. The size of grains of the ionmilled samples was from 10 to 50 fJJ1l. The roughness (Rz) of the
230
swface layers of the diamond depended upon the layer thickness and
for 50 J.Ull thick layer it was 6 J.Ull'Not only the growth swface of the diamond but also a
mechanically polished one was treated by ion milling,' Themechanical polishing was carried out by means of a cutting machine [8],The removal rate on mechanical polishing the diamond by the cuttingmachine was 0,2-3 J.Ull/h, The rate of the removal of the material onion milling (0,3-2 J.Ull/h) and the roughness of the ion milled swfacedepended on the density of the ion current, the ion energy, the angle ofincidence ofthe ions onto the swface and some other factors.
,The results of the ion processing of the growth surface ofdiamond layers are presented in the table 2.
The surface roughness profiles of polycrystalline diamond ftlmsafter ion milling at various regimes are shown in Fig.3. Under theaction of the ion beam directed perpendicularly or at 45
0
, thesmoothing of acute asperities of the surface relief takes place, thecontours of the grains and the surface roughness (FigAa-c) stayingthe same. On working with masks or with photolithogniphy methodsthe absolute verticality ofthe walls ofthe etch pits is obtained (Fig.3a-c).
TABLE 2. Data on ionmilling ofdiamond films
Thickness ofdiamond films, J.Ull 55 25 52 '47,8Slope ofion beam, deg. ° 45 45 86Working gas Ar Ar Ar+02 (1:1) AIIonmilling time, min 388 371 297 387Removed layer, J.Ull 10,0 7,7 9,2 2,0Ionmilling rate, J.LlllIh 1,55 1,25 1,86 0,31Surface roughness, .' J.Ullinitial 5,83 3,0 3,1 0,032after ionmillinglongitudinal measuring 5,80 4,0 5,4 0,032transverse mea.sllQng 5,80 2,96 4,32 0,032
231
x 100
§t--.......---1on Z.. r--'-"'---I
o
<:><:>oN
><
x 100x 100
0 C'
0 0 CC C
C~
on0 ..on.. ..
oFig.3. Profile records of polycrystalline surface diamond layers, and also of astep at the boundary ofgrowth surface and of the diamond surface milled at
following conditions: a) Ax+ ions, '8=00 , diamond removal rate of 1,55 J.l.IIl/h;b) Ax+ ,8=45
0, 1,25 J.l.IIl/h; c) Ax++ 0+ , 8 =45
0, 1,86 J.l.IIl/h; d) ion
milling (Ax+) of mechanically polished (curve 1) diamond surface; curve 2
8=860
, 2,0 J.l.IIl thick layer is removed; curve 3 - the two-step ionmilling: 1
8 =450
, 7,3 J.l.IIl thick layer is removed; 2 - 8=860
, 2,0 J.l.IIl thick layer isremoved.
IJ.ffi . ., ., __....._.- ~
~~l :: .. ············..·i········· Growth surface uf th~ diamund filens1 l' ~ ':I0:: Sul1lM:c ofthe diamond after ion mi1liDg ....•. " ~ ~ .~
t :': ~.I I 11 tfl:r:J'i:G\;~····~···~:.·· "'~~;)" wl )1~r···.··.·.··,··...::.:.:~ :L Jto z;o .$(jO_.._~-r$l.-----_J. J.l.IIl
Fig.4. Profile record of81 J.l.IIl thick diamond polycrystallinefilmmilled byAx+ ion beam for 9 hours.
232
On processing the diamond fllm in (Ar+02) plasma the etching ofthe diamond layer is accelerated by a chemical interaction of oxygenions with carbon atoms, which results in a more active etching of grainboundaries, thus increasing the roughness of the sunace (Table 2 andFig.3c).
On the ion milling of polycrystalline fJ1ms, the grains withdifferent orientation to the incidence line of the ion beam have different
sputtering yields. This makes it impo~ible to conserve the initialsmoothness of mechanically polished sUrface after the removal of adiamond layer from it by the ion beam, directed perpendicularly or ata small angle to the nonnal of the ftlm sunace. Acting by an ion beamwhich is almost parallel to the sunace of the diamond layer (6 = 8587°), it is possible to remove the diamond material, preserving thesmoothness of the polished sunace of the diamond (Rz = 0,032 ~before and after the ion processing), (Fig.3d). One can obtain theprescnoed roughness parameters on polishing, though at low etchingrates, if the angle e is close to 90. The roughness of ion milleddiamond layers depends on the direction relating to the course of thebeam in which measurement is carried out (longitudinally or~ersely to the action, see Table 2). So to obtain a unifonnsmoothing of the surface in all directions, it is necessary to rotatecontinuously the substrate with the ftlm on it. The rotation speed is 3-4revolutions per second.
Fig.4 presents the profIle record of 81 J.LIIl thick diamond ftlmafter ion polishing during 9 hours. The smoothing of the growthswface relieffrom the initial roughness R.z = 3,575 JLm to R.z = 0,78 JLmis observed. At constant conditions of polishing the value of theresulting roughness depends only on the time ofprocessing.
Drilling.
The diamond plates were drilled by the ion beam. Theseparation of the diamond layer from the substrate was carried out bychemical dissolving ofthe substrate. The scheme of the set for drilling the
233
I IrlR! Di~mond1ilm
Spacer
Spril1£
Collection ofions
IntUlalor
Fig.S. Scheme ofthe set for dri11.ing diamond fiJms bythe ion beam.
diamond plates is shown in Fig.S. The completeness of drilling a hole is
detennined by the ion current in the collector.
References.
1. SemenovAP. (1993) Pnb. i Tech. Eksp. N5, pp.128-133.2. Semenov AP., Batuiev B.-Sh.Ch. (1991) POD. i Tech. Eksp. NIpp.177-179.3. Alexenko AE., Belyanin AF., Bouilov L.L., Semenov AP.,Spitsyn B.V. (1991) in: Materialy 2 Konf. "Tonkiye Plyonki vElektronike", Moskva, Izhevsk, pp.74-79.4. Semenov AP., Smimyagina N.N., Belyanin AF., Alexenko AE.,Bouilov L.L., Spitsyn B.V. (1992) in: Materialy 3 Konf.'TonkiyePlyonki v Electronike", Moskva, Yoshkar-Ola, 2, pp.154-155.5. Belyanin AF., Semenov AP., Alexenko AE., .Bouilov L.L.,Spitsyn B.V. (1993) in: Materialy 4 Konf. 'Tonkiye Plyonki vElectronike", Moskva, Ulan-Ude, pp.166-168.6. Semenov AP., Belyanin AF., Alexenko AE., Bouilov L.L.,Spitsyn B.V. (1991) in: Almazniye Plyonki (1 Intemat. Seminar, UlanUde, Baikalskiye Volny) Moskva, Ulan-Ude, p.39.7. Bouilov L.L., Alexenko AE.,Botev AA,Spitsyn B.V.(1986) Dold.AN SSSR, vo1.287, N4, pp.888-891.
234
8. Spitsyn B.V., Belyanin AF., Bulyonkov N.A, Rivilis V.M. (1987)in: Sb. Technika Sredstv Svyazi, Ser.TPO, Vyp.l, pp.61-70.
DOPING OF DIAMOND-LIKE CARBON FILMS
STANISLAW MITURAInstitute ofMaterials Science and Engineering, Technical University ofLOdi, ul. Stefanowskiego J, 90-924 Lodz, Poland
JAN SZMIDTInstitute ofMicroelectronics and Optoelectronics. Warsaw University ofTechnology. ul. Koszykowa 75, 00-662 Warsaw, Poland
ALEKSANDRA SOKOLOWSKAInstitute ofMaterials Science and Engineering, Warsaw University ofTechnology. ul. Narbutta 85, 02-524 Warsaw, Poland
Abstract
In this paper electrical properties of heterojunction silicon-AI doped DLC wereanalysed. The AI doped DLC films (DLC:AI) were obtained by the RF decompositionof methane with in situ magnetron sputtering of doping materials. The obtained p-njunction characteristics and TEM observations indicated that aluminium could beintroduced into DLC. An inexpensive new method for manufacturing doped-DLCmaterials on a very large surface is offered. One should underline that this dopingmethod dispenses with the use of poisonous substances.
t. Introduction
The potential use of diamond as a semiconducting electronic material has beendiscussed by several authors [I-11]. Thin-film diamond deposition technology is nowsufficiently advanced for such devices as high-temperature Schottky diodes [1], metalsemiconductor field-effect transistors (MESFET) [2], permeable base transistors [3],point-contact transistors [4], thermistors [5] and polycrystalline diamond field-effecttransistors [6].The doped-diamond films were deposited by microwave plasma CVD techniques [7,8]or by the hot filament plasma CVD methods [9]. The diamond cold cathodes [10] wereproduced by forming diodes in p-type diamond using carbon ion implantation into thediamond monocrystal [11].There is another group of materials, i.e. diamond-like carbon (DLC), whose namederives from partial similarity to diamond. One should, however, emphasize thedifference between the diamond film and DLC film [12].
235
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 235-242© 1995 Kluwer Academic Publishers.
236
Doped-DLC films can be obtained as a very cheap material in comparison to dopeddiamond crystals and doped-diamond films. The introduction of additional materialsinto DLC films can be achieved by either incorporation of required atoms during thefilm growth [13-23] or by a post-growth implantation of ions [24]. The metalcontaining DLC films can be produced by the use of the following plasma assistedchemical vapour deposition (PA CVD) methods• Decomposition of hydrocarbon with simultaneous sputtering ofa metal [13-18].• Decomposition of hydrocarbon with simultaneous evaporation ofa metal [21, 22].• Decomposition of organometallic vapour as a single-source precursors of a metaland a carbon matrix [23].
• In situ doping ofDLC layers [25]
2. Experimental
The films were deposited by the RF decomposition of methane [26]. Additionally, inthe same reactor chamber, a magnetron sputtering source [27] with aluminium targetparallel to the RF electrode was used. A schematic diagram of the apparatus is shownin Figure 1.
GFMC
PS
Figure 1. The apparatus for amorphous DLC:A1 film synthesis by RF decomposition ofCH4 with additionalmagnetron sputtering ofdoped materials method; RF, 27.12 MHz generator; MU, matching unit; GFMC, gas
feeder with a microcomputer control; PS, power supply for magnetron; V, vacuum gauge
237
The experimental details for the investigation are given in Table 1. Silicon substrateswere placed on an RF-powered negatively self-biased electrode.
TABLE 1. The main parameters and the growth conditionsfor the RF CVD process with additional magnetron sputtering
SubstratesGasPressureRF generator- negative self-bias voltage
ofRF-powered electrodeMagnetron- negative bias voltageElectrode-to-electrode distance
Si, n-type, 0.2-0.5 nemMethane55 Pa27.12 MHz, 2 kW
150 VDC,lkW200 V150mm
The properties of the silicon wafers are as follows: "n" (type), 300J.l.m (thickness),<110> (crystallographic orientation), 0.2+0.5 Ocm (resistivity). Before deposition offilms, the reverse side of the silicon wafers was chemically cleaned and thenaluminium electrodes were evaporated onto these surfaces. The AI electrodes wereevaporated onto the surfaces. The AI electrodes were thermally fused to produce anohmic contact.The DLC films, aluminium-doped (DLC:AI) in situ by magnetron sputtering, weredeposited onto the obverse side of the polished and ion-etched silicon wafers.To enable the electrical characteristics of the Si-C heterojunction to be measured, afterthe plasma deposition process, aluminium electrodes were evaporated onto theDLC:AI films. The current-voltage characteristics of the heterojunction were measured.Then, the DLC:AI films were analysed by transmission electron microscope (TEM),Auger electron spectroscope (AES) and ellipsometer.
3. Results and Discussion
The typical current-voltage characteristic of the Si-C heterojunction is presented inFig. 2. The fact that aluminium could be introduced into the DLC film indicated a p-njunction characteristic curve, as shown in Fig. 2. When the AI-doped DLC film wasreversely biased, the leakage current at IOV was 10 pA. When the AI-doped DLC filmwas forward biased, an increase in current was observed. The current was greater than5 rnA at 10V. Conductivity of the AI-doped DLC films was increased in comparison toDLC films by seven orders of magnitude. The t-V characteristics for pure DLC filmsobtained by the same growth conditions but without magnetron sputtering, areextremely different. When the applied voltage was changed from -20 V to +20 V, thecurrent was lower than I nA. The shape of characteristic of p-n heterojunction proves
238
that AI atoms satisfy all conditions that are satisfied by the acceptor impurity insemiconductor materials.
I [fJA]
600
400
200
o
- V
- _J_- DLC:Al
-n - Si
-
- I-
-
-I I I I I I
-8 -6 -4 -2 o 2 4 6 8U [V]
Figure 2. The current-voltage characteristic ofSi-DLC:AJ heterojunction
The DLC:AI films are predominantly amorphous. The presence of aluminium in theDLC film was proved by AES examination. The films are without AI droplets. Theexamples offilm structure that has been obtained in the plasma process, as in Table I,are shown in Figure 3.One should underline that these films are chemically homogeneous. On the other hand,however, the repeatability of the obtained results is not satisfactory. At the present stateof the art it is hard to explain the reason of this phenomenon.
239
Figure 3. TEM micrographs ofamorphous AI-doped DLe film: (a) bright field image, (b) electron diffractionpattern
240
The ellipsometric measurements of the refractive index for D, DLC and DLC:Al filmsare as follows:• the refractive index n =2.4 for the diamond films [28]• the refractive index's mean value n = 2.1 for DLC films• the refractive index n = 1.9 for the DLC fIlms containing aluminium (DLC:Al).The DLC:Al films are transparent in the region similar to DLC films [29].
4. Conclusions
A possible production of doped-DLC fIlms by means of RF decomposition of methanewith additional magnetron sputtering of doped materials has been presented. Theelectrical properties of doped-DLC films make it possible to apply these materials inelectronics to produce semiconducting devices.The RF decomposition of methane with in situ magnetron sputtering of dopedmaterials seems to be a convenient method of the production of doped structures. Thisdoping method dispenses with the use of poisonous substances.
Acknowledgements
This work was supported by Grant no. 3-3602-91-02 of the Polish State Committeefor Scientific Research.
References
1. Gildenblat, G.Sh., Grot, S., Wronski, C.R, Badzian, AR, Badzian, T., andMessier, C.R. (1988) Electrical characteristics of Schottky diodes fabricated usingplasma-assisted CVD diamond films, Appl. Phys. Lett. 53 (7), 586-588.
2. Shiomi, H., Nishibayashi, I., and Fujimori, N. (1989) Field-effect transistors usingboron-doped diamond epitaxial films, Jap. J. Appl. Phys. 28 L 2, 153-154.
3. Geis, M.W., Efremov, N.N., and Rathman, D.D. (1988) Device applications ofdiamonds, in G.H.Johnson, M.Geis and ABadzian (eds.), Diamond and Diamondlike Material Science and Engineering Study. Materials Research Society,Pittsburgh, PA
4. Geis, M.W., Rathman, D.D., Ehrlich, D.l, Murphy, RA, and Lindley, W.T.(1987) High temperature point contact transistor and Schottky diodes formed onsynthetic boron-doped diamond, IEEE Elect. Dev. Lett. EDL-8,341-343.
5. Werner, M., Schlichting, V. and Obermeier, E. (1992) Thermistor based on dopedpolycrystalline diamond thin films, Diamond and Related Materials 1, 669-672.
6. Tessmer, Al, Das, K., Dreifus, D.L. (1992) Polycrystalline diamond field-effecttransistors, Diamond and RelatedMaterials 1, 89-92.
7. Fujimori, N., Imai, T., and Doi, A (1986) Characterization of conducting diamondfilms, Vacuum 36, 99-102.
241
8. Nishimura,K., Das, K., and Glass, IT. (1991) Material and electriccharacterization of polycrystalline boron-doped diamond films grown by microwaveplasma chemical vapor deposition, J. Appl. Phys. 69,3142-3148.
9. Okano, K., Naruki, H., Akiba, Y., Kurosu, T., Iida, M., Hirose Y. andNakamura, T., (1989) Characterization of boron-doped diamond film, Jpn. J. Appl.Phys. 28,1066-1071.
10. Geis, M.W., Efremov, N.N., Woodhouse, lD. and McAleese, M.D. (1991)Diamond cold cathodes, in Y. Tzeng, M. Yoshikawa, M. Murakawa and AFeldman (eds.), Applied of Diamond Films and Related Materials, Elsevier Sci.Publ. B.V., Amsterdam, p.309-31O.
11. Prins, IF. (1982) Bipolar transistor action in ion implanted diamond, Appl. Phys.Lett. 41, 950-952.
12. Spear, K. (1989) Diamond - ceramic coating of the future, J. Amer. Ceram. Soc. 72,171-191.
13. Meyerson, 8., and Smith, F.W. (1980) Chemmical modyfication of the elctricalproperties of hydrogenated amorphous carbon films, Solid State Commun. 34,531-534.
14. Meyerson, B., and Smith, F.W. (1982) Thermopower of doped semiconductinghydrogenated amorphous carbon films, Solid State Commun. 41,23-27.
15. Jones, D.I., and Stewart, A (1982) Properties of hydrogenated amorphous carbonfilms and the effects of doping, Philosophical Mag. B46,423-434.
16. Dimigen, H., Hubsch, H., and Memming, R (1987) Tribological and electricalproperties of metal-containing hydrogenated carbon films, Appl. Phys. Lett. 50,1056-1058.
17. Chen, P.A (1989) Characteristics of copper-incorporated amorphous carbon film,Thin Solid Films 182, 261-263.
18. Demichelis, F., Kamiadakis, G., Mpawenayo, P., Perino, M.A, Tagliaferro, A,Tresso, E., Rava, P., Della Mea, G., and Vallino, M. (1987) Structure and opticalproperties of hydrogenated amorphous carbon-tin alloys prepared using the sputterassisted plasma chemical deposition technique, Thin Solid Films 150, 189.
19. Despax, B., Flouttard, lL. (1989) Synthesis of gold-carbon composites bysimultaneous sputtering and plasma polymerization of propane in RF. capacitivelycoupled diode system (13.56 MHz), Thin Solid Films 168, 81- 88.
20. Gerstenberg, K.W., and Grischke, M. (1991) Thermal gas evolution studies on aC:H:Ta films, J. Appl. Phys. 69,736739.
21. Biederman, H., and Martinu, L. (1990) in R d'Agostino (ed.), Plasma Deposition,Treatment and Etching ofPolymers, Academic Press, Boston, MA, Chapter 4.
22. Biederman, H., Chudacek, I., Slavinska, D., Martinu, L., David, J., and Nespurek,S. (1989) Physical properties ofmetal/a-C:H composite, Vacuum 39,13.
23. Wrobel, M., Czeremuszkin, G., Szymanowski, H., Szur, H., Klemberg-Sapieha, lE.and Wetheimer, M.R. (1992) Plasma CVD of iron -containing hydrogenated carbonfilms, J. Chem. Vapor Deposition 1,41-58.
24. Amir, O. and Kalish, R (1992) Doping of amorphous-hydrogenated carbon filmsby ion implantation, Diamond and Related Materials 1, 364-368.
25. Szmidt, l, Olszyna, A, Sokolowska, A, Mitura, S. (1994) In-situ doping of DLClayers, ISCDF-94, Minsk, 2-5 May, Paper No.5-I.
242
26. Has, Z., Mitura, S., Clapa, M. and Szmidt, 1. (1986) Electrical properties of thincarbon films obtained by RF methane decomposition on RF-powered negativelyself-biased electrode, Thin Solid Films 136, 161-165.
27. Mitura, S., Has, Z. and Gorokhovsky, V. (1991) The system for depositing harddiamond-like films onto complex-shaped machine elements in an r.f. arc plasma,Surf Coatings Technol. 47,106-112.
28. Mitura, S. (1992) Radio-frequency hot-filament CVD of diamond, Diamond andRelatedMaterials 1, 239-242.
29. Staryga, E., Lipinski, A., and Mitura, S. (1986) Electrical conductivity and opticalabsorption of carbon films produced by RF decomposition of hydrocarbon, ThinSolid Films 145,17-21.
UNHYDROGENATED DLC FILMS OBTAINED BY MAGNETRON SPUTTERING
C.MOROSANU, +N.TOMOZEIU, C.CORDOS and T.STOICA
Inst. of Phys. & Tech.of Mat.,Magurele,P.D.Box Mg7, Bucharest, Romania+Faculty of Physics, University of Bucharest, Magurele,P.D.Box Mgll, Bucharest, Romania
The paper deals with the properties of unhydrogenated carbon layerswith high optical gap. RF magnetron sputtering in an argon atmospherewas used in order to grow high optical gap DLC layers on Corning glasssubstrates at low deposition temperature «150 C). For 2.2eV optical gaplayers, the coniuctivity measurements at 100 C show a very highresistivity of 10 Oem and an activation energy of the order of 1.2eV.Light intensity and spectral dependencies of the photoconductivity werealso measured.
1. Introduction
Magnetron sputtering within high argon pressure (>2Pa) region wasrecently used to obtain high optical gap DLC layers /1-2/. For DCmognetron sputtering /1/, the increase of optical gap is accompanied bythe increase of the electrical resistivity and of voids fraction, but alsoby a decrease of the hardness. From 1 Pa up to 8 P~, the resistivityincreases by about four orders of magnitude (30 : >10 Oem) when theoptical gap varies from 0.5 to 1.2eV /1/. At 1 Pa, similar results areobtained by decreasing the deposition power /3/. Compatibility betweenoptical gap and electrical gap was found.
In this paper, high pressure RF magnetron sputtered layers areinvestigated through electrical and photoelectrical measurements.Information on smoothness of the films was obtained by both optical andSEM investigations.
2. Experimental details
RF magnetron sputtering in an argon atmosphere was used in orderto grow different optical gap DLC layers on Corning glass substrates atlow deposition temperature. The carbon sputtering source was a target ofspectral graphite. The deposition conditions are reported in/2/.
The resistance of a-C coplanar samples of high optical gap laners(>2eV) was found greater than the range of our instrumentation (>10 0).
243
M.A. Pre/as et al. (eds.), Wide Band Gap Electronic Materials. 243-248© 1995 Kluwer Academic Publishers.
244
For electrical and photoelectric measurements, only sandwich structures(Corning glass substrate/ ITO electrode/ a-C/ semitransparent Auelectrode) have been successfully used. Spectral and integral lightphotoconductivity measurements were performed using calibratedinterferential and neutral filters with a Xenon lamp.
J. Results and discussions
When the deposition pressure increases from 1.3 to 13 Pa, the layerstype changes gradually from graphitic, for optical gap smaller than 0.5eV,to diamond-like for optical gap as high as 2.2eV (Fig.i). Optical gap wasobtained using Tauc formulae and absorption coefficient given bytransmission measurements in visible and UV range.
S;2~--t---+---::::t-t""'::=-+-----'!---4•-oWUI~----t-+-+--+--+-----'!---4
iRF - +-c Iayer~
10 Q.Ii Iip(Pa) -
7.66
CJ.IIL-_--i__--i.__~__..1...__ ____JL-_ ___'
o
Fig.l Optical gap Eo of RF sputtered layers ondeposition pressure P.
For optical application, the smoothness of the layer is veryimportant. The polycrystalline diamond films have a roughness of theorder of the thickness of the layer /4/.
The amorphous DLC films have smooth surfaces depending on thegrowth conditions. SEM image of a transversal section for a high gap DLClayer (Fig.2) shows a columnar structure with domains smaller than O.I~m.
This has no effect on the optical scattering as can be seen in fig.3. Inthe same figure, transmission curves for a single crystal type IIA naturaldiamond and three polycrystalline CVD diamond with different surfaceroughness are shown comparatively. The transmission throughpolycrystalline films is drastically reduced due to the scattering losse¥.Unfortunately, for our layers a small value of hardness (HV - 50 kgf mm")was found. This behavior makes our high gap (>2eV) a-C layers notsuitable for optical applications if high hardness is also required.
Besides the experimentaly observed high resistivity of high gapa-C layers, an additional slow transient current makes the electrical
245
Fig.2 SEN image on DLe 2.2eV gap layer.
measurements difficult. As can be seen from FigA, the transient currenthas two relaxation times (about lOs and 3.5min) nearly independent ontemperature. In our opinion, the observed transient current is due to thetime variation of internal polarization within the a-C layer. This relaxationinvolves electron transitions between localized states near Fermi level.
-lI'eG I- fc: I
leG IE
f•1 40
~
lIlI II
00 - eGO _ llIOO _
X(rm)-
Pig.3 Transmission spectra of (-) our DLC fil.comparatively with(---) three CVD di8lllOnd fillllS wi th different surfaceroughness Ra, and ( .•. ) a natural diaaond /3/.
246
I (pA)
T '" 1(XPC; U"'10V
10
~ T '" 25 CPu=10VJ------ -.:...-:~~
1.1122.1133.644..1111
t (min)
Fig.4 Transient dark currents at two temperaturevalues, after lOY bias.
In our case, a-C layers are characterized by a large amount of voids /5/which may be the source of a high density of localized states chains.
The temperature dependence of the steady state dark conductivityin Fig.S shows activation Fnergy of Eo 1.2eV and exponential prefactor 0
0of the order of 0.1 (Qcmf .
.• ~___ += 77rrtNIscm.~----
u
Fig.S Conductivity temperature dependence for dark andilluminated sample ( • is the white light intensity;30V - measurement voltage; 1.2~ - thickness of a-Clayer) .
Twice the activation energy 2.4eV is approximating the optical gap,that means the Fermi level is flose to the middle of the gap. A carriermobility value of ~bout3 0.1 cm /Vs results, if we consider the effectiveband density of 10 cm- and Fermi level departure from conduction bandequal with Ed'
247
Fig.6 Photo to dark conductivity ratio VB.white light intenaity.
Fig.? Spectral collection efficiency of 30Vbiased sample.
o
USA ....
Epb(eV)-•u ..... u
" ,"......
value of I!-'t product of about 3 10-11
T ~ 100\;.. aWrk~ t.1101~oanf1
ITO/a-c/Au strootll'a2.2aV optical gop
... -I----.---r-....-r,..Q,rr---.---r-....-r".."".---r~..................0..
1E-610.,.----,---,---,--..........-.,.----,---,---,---,
I I I ... ~_~:".;. .' .,~.:,J.11 • .. · ..t··t... , , ""!
c e ~.. . j.....+..... ~ ~ ~i ~ ~ ········~·~·..~~r• ········l·..···t·····-+ :.ot;~·;;t·t~· ..··I· .. ··t-·· ..···trot~=1i~lXe• _""T-"--+-----;;--~~r----+-~-+~- ...~--i!,........ ......~
In the same Fig.S, theconductivity of the a-C sampleilluminated by white light ofvarious intensities values isshown. The temperaturedependence gets slower when lightintensity is increased. The lightcomes to the a-C through thesemitransparent Au electrode. Theratio value of 20 at 100 C(Figs.S-6) betwee~
photoconductivity at 100 mW/cmand dark conductivity increases atlower temperature, reaching morethan three orders of magnitude atroom temperature. The lineardependence (Fig.6) of thephotoconductivity 0p~ shows amonomolecular type ofp hotocarriers recom bina tionthrough localized states.
Monochromatic light obtainedthrough the interferential filterswas calibrated by using anIn ternation al Lig h t I LSOOinstrument. The quantum collectionefficiency defined as the ratio ofexternal circuit collected carriersnumber and the incident photonsn~ber is shown in Fig.7 for 2.510 V/cm electrical field. We cannotice that the collection efficiencyincreases significantly at photonenergy higher than optical gapvalue (t.2eV). The maximum valueof 28 10- corresponds to an estimatedcm/V.
Conclusions
High optical gap (2.2eV) unhydrogenated DLC layers obtained by RFmagnetron sputtering sho~ high VIS transmitance, but small value ofmicrohardness (SO kgf/mm). Stationary dark and photo-currents weremeasured at higher than room temperature. The dark conductivity is verysmall and can be measured only on sandwich samples and ~t hig~
temperatures (above 60 C). At 100 C, the dark conductivity of 10-1 (Ocm)"and a ratio of dark and photoconductivity of 20 at 100mW/cm white lightwere obtained.lIFror spectral photoconductivity dependence, a value of I!-'tproduct of 10- cm /V was estimated at 100 C.
248
Acknowledgements
The authors wish to thank C. Popescu formeasurement system donated by the Alexanderis gratefully acknowledged.
helpful discussions. Thevon Humboldt Foundation
ReferencesI. Clarke G.A and Parsons R.R.(1993) Characterization of magnetron-sputtered diamond-likethin films for optical coatings in IR. Thin Solid Films 236. 67-71.
2. Morosanu C., StoicaT., De Martino C., Demichelis F. and Tagliaferro A. (1994) High gapsputtered DLC layers, Diamond and Related Materials, 3, 814-816.
3. Fang T. - Proc. SPIE, 159, 1146 (1989).4. Savvides N. (1986) Optical constants and associated functions of metastable diamond likeamorphous carbon films in the energy range 0.5-7.3 eV, J.Appl.Phys. 59, 4133-4145.
5. Stoica T., Dragomir A., Gartner M., Morosanu C. and PaveIescu G. (1994) Opticalproperties of sputtering and glow-discharge a-C:H films, presented at "NATO AdvancedWorkshop on Wide Bandgap Electronic Materials", Minsk, Belarus, May 4-6 1994.
SIMULATION OF DIFFUSION IN AN AMORPHOUS STRUCTURE
A. V. NazarovDepartment of Metal Physics. I.PBardin Central Research Institute ofFerrous Metallurgy, Moscow 107005, Russia
Abstract
Universal single-atomic mechanism ofdiffusion suggested by the author is applied to thedescription of the diffusion in amorphous structures. This mechanism is the only one thatgives possibility to control changes of the local atom environment during the time intervals comparable to mean time between two jumps of atoms. This enables us to take intoconsideration the dependence of the mutual atoms arrangements on the jump frequencyspectrum of the system. Mathematical models of diffusion in an amorphous structurebased on this mechanism are suggested. In this framework the influence of different factors distinguishing amorphous and crystalline structures was studied by computer simulation. The results of simulation shows that within these simplified models theexperimental data on diffusion of atoms in amorphous alloys can be quantitatively explained.
1. Introduction
A characteristic feature of amorphous materials is the occurrence of structural-relaxationphenomena in their mechanical, magnetic, and electronic properties. Structural relaxation processes are realized by diffusion processes. The analysis of some experimentaldata [l - 3] allows to determine the following distinctive features of diffusion in amorphous structures: high diffusion mobility of metal atoms (some orders of magnitudehigher than volume diffusion in crystals), compliance oflaboratory data to Arrhenius lawin many cases, comparable (within one - two orders) mobility ofmetal and non-metal atoms, strong dependence in same cases of diffusion coefficients on structure relaxationcaused by isothermal exposure a temperature lower than Tg (glass transition temperature). There has not been given yet any satisfactory explanation of these factors based ona unified theoretical approach. A universal single-atomic (USIA) mechanism of diffusion is suggested in the present paper. A mathematical model of diffusion in amorphousstructures has been developed in the framework of this mechanism and the influence ofspecific features of amorphous alloys structure on diffusion mobility has been studiedusing simulation methods. As distinguished from others models [4-5] this one gives pos-
249
M.A. Prelas et at. (eds.), Wide Band Gap Electronic Materials, 249-256© 1995 Kluwer Academic Publishers.
250
sibility to take into consideration changes of the local atom environment.
2. Model for Diffusion Process
It is supposed that as well as in most cases in crystals the diffusion in metallic glass follows single atomic mechanism. Elementary diffusion jump is characterized by the following factors which distinguish amorphous structure from crystal one:
1. Spread in values of activation barriers Qi2. Different length of atom jumps3. Different density of substance4. Local character of spectrum of atom oscillations frequency5. Alteration of structure within time.
The analysis of thc above mentioned factors based on the general theory ofdiffusion processes [6] allows to come to a conclusion that correlation effects playa great role in diffusion in such systems. Therefore by simulation it is necessary the displacement ofparticles to be much higher than the specific unhomogeneity length in the system. Uniparticle approximation proved it efficiency for solving diffusion problems. For investigating the diffusion of atoms in amorphous structures we suggested a simulation method[7], which is based on following the movement ofone thermalized particle in the field ofremaining frozen atoms. The movement of particles is described by Newton equationsand thermalization is achieved in such way: the velocity of a particle takes one of the random values according to Maxwell distribution in equal time intervals 'tT - l!roD (roD Debye frequency). one has the usual equations.
dx·1- = v·dt 1
dv.1
-dt1-P-m 1
aup. =--
I aXii = 1, n (1)
where x, v are the coordinates and particle velocity; U - field created by surrounding atoms, n - space dimension.In the approximation of central differences [8] we have:
(2)
It is not difficult to simulate the movement of particles using the given equations if theinitial conditions and potential field are set. Having determined the shift of a particle ineach run within migration time 'tm , it is possible to calculate the diffusion coefficient according to relation D = <R2 >/2n'tm • averaging being performed on the number of runs.
The suggested method has been tested on some simple models. For example, in case of
251
one-dimensional structure the potential field takes the form U = 1 - B*Cos(x). The testhas shown that the law <R2> -'tm and Arrhenius law (DCr = Do
Cr exp(-Q / (kT)) are implemented for one-dimensional and two-dimensional crystalline structures with periodicfield (Fig. 1). It can be easily seen that in the suggested simulation method the first fourfactors specific for amorphous structures are taken into account if quasiperiodic potentialfield is set in a corresponding way (In case of one-dimensional structure, for example,the field had the form:U = 1 - Bk * COS(Uk*x), where Bk had any value from (Bk) values, the average valuebeing equal to 1; a - normally distributed random value, <Uk> =1, k - site number).
enD-2
-4
-8
-.8
- fa
-12
-14
-16
Figure 1. Arrhenius plot of the diffusion coefficient on an amorphous structure (full line) with Qmin=0.7,Ll.Q = 0.2, <Qi> = I, and on Ctystalline structure (dash line) with Q = 1. (1) - one-dimensionalstructure, (2) - two-dimensional structure.
Now we discuss in more detail the problem connected with changing the structure withinthe time. In papers [2,3] it has been shown that at isothermal annealing at a temperaturelower than Tg the diffusion coefficient decreases, activation energy being increased. Thissuggests an idea that local surroundings of each atom and activation barriers values varywithin time. Generally speaking we come across similar situation at vacancy diffusionmechanism. However in different descriptions of this mechanism the indicated situationis not obviously taken into account. Fig.2 schematically shows time variation of the valueof potential barrier which the atom has to overcome in order to move to an adjacent position. Such variation is specific for diffusion of interstitial atoms in solid solution, forvacancy mechanism and for diffusion of atoms in amorphous structures. In general theprobability of atom jump in any direction within time unit can be presented as following:
252
]--- (j)
(2)
"bv (5._~__t
Figure 2. The variation of potential barrier as a function of the time: (1) - interstitial mechanism, (Z) vacancies mechanism, (3)- amoxphous structure.
r =sL/ (Qi' t) V (Qi)"ti
i (3)
where f - probability for the potential barrier of the atom jump to a position s in a momentt is equal Qi' "ti - interim, in the course of which the barrier equals Qi. v - correspondingjump frequenc.y. The probability f is conditioned by the jump frequency spectrum andthe mutual atoms arrangements.
In case of interstitial mechanism we have one action barrier
r = v(Q) (4)
In case of vacancy mechanism in one component system the activation barrier may taketwo values: high value if there is no vacancy near and low value if there is a vacancy nearthe atom. It is obvious that contribution to the sum corresponding to the first case is negligibly small because the frequency of jump is about zero and for the second case f(Qz)- Cv and r =Cv v(Qz).
253
After examining this vacancy mechanism we can say that it necessary to consider changes in local surroundings within time in simulating the movement of one isolated atom. Inthe simplest case it was done as follows. If a particle got into interstice where it was earlier, then the values of activation barriers were determined again randomly, except forthe barrier which has been just overcome by the atom.
Simulation in one- and two-dimensional structures showed [7] that the second and fourthfactors did not influence much on the diffusion coefficient. Spread of activation barriervalues play the main role in diffusion in amorphous structures. This gave an opportunityto use the realization ofMonte-Carlo method which is similar to one described in [91 The results obtained by the fIrst and second methods for quasiperiodic field which is characterized by spread only of activation barrier values coincide within the experiment accuracy.
3. Results and Discussion
Simulation showed (Fig. 1-3) that diffusion migration of particles in quasi-periodic potential fIeld with random variables ofbarrierQk was much faster at low temperature thanin periodic fIeld with activation barrier Q = <Qi>' And at a low temperature (lower thansome specillc temperature) there is a correspondence to Arrhenius law and activation energy is equal to the lowest of the values of activation barriers, i.e. DAm = DoAmexp(-Q/(kT)) (Fig. 3). As it is seen from the figures, Do
Am « DoCr which can be explained bystrong correlation effects by diffusion in amorphous structures.
/0 20 r- I
Q.",,,, dQ. tV0)5 0,1 'I (2)OJ )
0,1 ~2 'I (Ir)0,1 ,6 2 (5)0,6 IH' j (6)
5
-/0
-/5
Figure 3. Anhenius plot of the diffusion coefficient on an amorphous structure (full line) with a differentparameters of the activation energy spectrum and on crystalline structure (dashed line) with Q = 1.
254
We have investigated the influence of spectrum parameters of the barriers on diffusionmovement in one-dimensional structures. The results are indicated in Fig. 3. See also [7].
More detailed investigation of the properties of the above mentioned model showed thatlocal surroundings of the atoms changed too often and this aspect is to be specified. Having this in mind we consider the vacancy mechanism. In this case the following simulation scheme is possible. At first one of the possible directions of atom. jump is randomlyselected, then according to f(Q) barrier value in this direction is selected then a randomvalue is compared with the probability of jump in the adjacent site and we determinewhether there was ajump of atom or not. If to consider that a barrier is retained for a certain time after the jump, then the correlation effects will be taken into account to certaindegree There are many "empty" events in the proposed scheme. Therefore we developedanother model with an average time ofchanging local surroundings of the atom 'tl' whichis the parameter of simulation. When simulating amorphous structures it is supposed thatthe change of local surroundings occur only at adjacent atom jumps. 't = 1/vmax is takenas a time unit where vmax - exp(-QmiJkT).
Fig.4 shows the results of simulation. These results confirm the main role of changinglocal surroundings. If there is no change (tl =00) the ratio DAm/DCr at a low temperatureis less than 1. Besides, when decreasing the temperature from a certain level the migration of atoms in such systems has no longer a diffusion character (law <R2> - t is notsatisfied). In other cases atoms move quicker in amorphous structures than in crystalline.And the less is tt the higher is the ratio DAmlDCr' Qualitative regularities of the influenceof specific features of activation energy spectrum on the diffusion mobility is of the samenature as of the model investigated earlier (Fig.3).
The kinetics of the diffusion process is determined by the relation between thc time required for the transition through the barriers at a given local surroundings and the timerequired for changing local surroundings. If the period of time for changing local surrounding is short enough, the diffusion jumps of atoms will occur mainly through thelowest of the barriers available in the spectrum for the atoms of given species, i.e. satisfied for metal atoms; and the change of local surroundings occur due to "fast" diffusionof more mobile components of the alloy.
Another limiting case (the time period of changing local surroundings is long) is realizedat diffusion of light impurity atoms. In this case the change of local surroundings of atoms is of no substantial importance and it is expected that Arrehnius law will not be satisfied. The diffusion is also possible to be retarded as compared with crystalline state atcertain parameters values characterizing the spectrum of activation barriers.
The above given results were obtained when simulating two-dimensional structures Theinfluence of changing dimension on diffusion mobility is of great importance. Thereforea model was simulated and a program developed for investigating diffusion in three-dimensional structures. Fig.S shows the result of simulation. As it is seen from the figures
255
~ ...---------------.--------,'lle-c
2 ~---------+---J~----I------+-+-',L-----~
1+-...,.01!!~-----l -.L. -"::::'.....-..{
5 10
Figure 4. Ration of diffusion coefficients on a amorphous structure and on a crystalline one as a function ofIff, for different values 1:1'
7)A~---------'--------'-----------'
lJc~
2
Figure 5. Ration of diffusion coefficients on a amorphous structure and on a crystalline one as a function ofIff, for different values 1:1 in cases: two-dimensional structure (full line) and three-dimensionalone (dashed line).
256
the diffusion mobility in three-dimensional structures does not differ much from two-dimensional structures. Therefore all the above given conclusions refer to the three-dimensional structures as well.
It should be noted that the calculation of changing local surroundings allows to suggesta unified diffusion mechanism for atoms of both metals and non-metals in amorphousmetal alloys and hence the diffusivities of any kind of atoms should not differ greatly asfor example in diffusion of light atoms and metal atoms in alloys having crystalline structure.
Processes of self-diffusion in diamond were examined by J. Bemholc and co-workers[10]. They concluded that self-diffusion in diamond is dominated by vacancies. Theabove given results allow suggest USIA mechanism of diffusion for atoms in amorphousdiamond-like structures. Spread in values of activation barriers depends on structures butin any case thc activation energy is lower in diamond-like structures than in diamond.
Thus the simulation shows that within the universal single-atom mechanism which isbased on diffusion controlled change of local atom surroundings the experimental dataon diffusion of atoms in amorphous structures can be qualitatively explained.
References
1. Y.Limoge. G.Brebec and Y.Adda. in DIMETA 82. K.FJ.Kedves and D.L.Beke. Editors. Diffusion and Defect Monograph Ser.7. p.285. Trans. Tech. Pub!. 1983.
2. B.Cantor and RW. Cahn, in Amorphous Metallic Alloys. F.E.Luborsky. Editor. p 487, Butterworths Mono-graphs in Materials 1983.
3. H.Mehrer and W.Domer. in Defect and Diffusion Forum. 66-69 (1989) 189.4. R.Kirchheim, Acta Met.• v35. 2. p 271, (1987).5. Y.Limoge and J.L.Bocquet. in Defect and Diffusion Forum. 66-69 (1989) 269.6. J.P.Stark. Solid state diffilsion. AWiley-Interscience Publication. N.Y.• 1976.7. A.V.Nazarov. in Fizika Amopfnich Splavov. p 157. Igevsk 1984.8. J.R.Beeler,Jr.• Adv. Mater. Res., v4, p 295, (1970).9. A.V.Nazarov. V.T.Borisov, in Problemimetallovedenija i fiziki metallov, N 6, p 35. M."Metallurgija". 1980.10. J.Bernholc,A.Antonelli,T.M.Del Sole, Y.Bar-Yam and S.T.Pantelides. Phys. Rev. Lett.• v61, p2689. 1988.
OPTICAL AND ELECTRICAL PROPERTIES OF QUANTUM-DIMENSIONALMULTILAYER STRUCTURES BASED ON CARBON FLLMS
V.V.Sleptsov, V.M.Elinson, A.M.Baranov, and S.A.TereshinNPO "Vacuummashpribor", Nagorny pro 7,113105 Moscow, Russia
Abstract
Quantum-dimensional multilayer structures based on thin carbon films with differentband gaps have been obtained. Optical properties, electrical properties of heterostructures containing such-structures and electroluminescence in visible spectrum range atroom temperatures have been investigated. It is shown that a blue shift in the optical bandgap with decreasing well layer thickness corresponds to energy level quantization in thecase of one-dimensional periodic potential. Investigation of x-rays properties of multilayer carbon structures have approved the presence of structures periodicity. Currentvoltage characteristics in dark and under illumination have demonstrated a resonant tunneling injection of minority carriers in heterostructures monocrystalline Si-multilayerstructure. Electroluminescence in visible spectrum range has been observed in such heterostructures at room temperatures.
1. Introduction
In the last few years, periodic multilayer structures have attracted considerable attentionbecause of their potential utilization as "new perspective" materials. The application ofmultilayer structures (MSs) allows the development of new devices. The development ofamorphous semiconductor superlattices is considered to be one of the main trends in thisfield.
Various combinations of materials, e.g. a-Ge:H/a-Si:H; a-Si:H/a-SiN and a-Si:H/aSiCx:H, and modular doped a-Si:H, can be used to from layers in amorphous semiconductor superlattices in which quantum-dimensional effects take place [1,2]. However,the thermodynamic instability caused by steep concentration gradients and rapid diffusion processes leads to degradation of the structure parameters. An attempt to synthesizeamorphous superlattices from unusual combination of materials (a-Si:H and a-C:H) withshort-range order of the diamond type [3] did not permit the problem of thermodynamicstability to be resolved.
257
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 257-264© 1995 Kluwer Academic Publishers.
258
The logical subsequent development of this trend is to attempt the synthesis of amorphous superlattices based on carbon layers. Carbon is the origin of a whole class of materials, making possible the formation of amorphous superlattices based solely on carboncompounds, and with these it is hoped to produce MSs with better thermodynamic stability and minimized diffusion processes between layers. Since the band gap (Eg) ofamorphous carbon films may be varied continuously from 0.4 to 3 eV, a-C:H depositionconditions have an advantage over a-Si:H films which must be deposited with practicallyconstant Eg values [4-6]. This makes it possible to use a-C:H films with different bandgaps as layers of superlattices.
2. Experimental Details
Amorphous semiconductor multilayer structures were synthesized from alternating aC:H layers with optical band gaps ofEgl = 0.55 eV (a-C:H1) andEgz = 1.5 eV (a-C:Hz).
The carbon layers were obtained by DC magnetron sputtering of graphite target in argon(a-C:H1) and by plasma-ion beam deposition in C6H1z (a-C:Hz). During deposition, pressure and substrate temperatures were in the range (1-2)*10-3 torr and 330-350 K, respectively. Carbon layers were deposited on fused quartz wafers and silicon substrates. Thedeposition growth rates were about 50 A/min for a-C:HI and about 25 A/min for a-C:Hz.The multilayer structures contained from 5 to 19 layers. The layer thicknesses were varied from 10 A up to 200 A and from 30 A up to 50 A for a-C:H} and a-C:Hzlayers, respectively.
Properties of carbon layers depend on deposition conditions and thicknesses, and havebeen described in previous papers [7].
3. Results and discussion
At the first stage x-ray properties of multilayer carbon structures were measured. Fig. 1gives the results of reflecting property of the 31 layers structure with different opticalband gap and the thickness of an individual layer of 45 A. It is seen that the reflectioncoefficient value for Bragg maximum of the ftrst order is 6% and for diffraction peak ofthe third order-O.l%. The peak full with at half maximum of the first order is 0.065°. Theangular position of Bragg reflection peaks correspond exactly the calculated thickness ofMIS period, while absence of the second order diffraction peak shows thickness equalityof the neighboring layers in all structure periods.
Then the influence of layer thickness on the change of Eg a-C:HI layers in superlatticeswas investigated.
For this purpose we performed the transmission and reflection spectra of the structures
259
20 .-----------------,
i5
R., %J
fO t "2
5 I .30
00.1 06 f.f
8, de!
Figure 1. X -ray diffraction pattern of multilayer carbon structure with equal thickness of layers inperiod.
1.61.'10.6o
OJI
300S
~250
~~
~.
200I
~
'"~ J.SO~.:t
fOO
SO
1.0
f,eVFigure 2. Dependence of (<x*E)I12 on E for multilayer structures with thickness of a-C:Hzlayers (5.0
nm) and a-C:H1layers: (1) d = 8.0 nm; (2) d = 3.5 om; (3) d = 2.0 nm; (4) d = 1.5 nm; (5) d =1.0 om and (6) the separate a-C:Hz films.
260
in the energy range 0.8-1.5 eV. On the basis of these spectra, the dispersive dependencesof the absorption coefficient (a) were calculated. Optical band gaps of multilayer structures were determined by Tauc's equation [8]
(a*E)112 =A*(E-Eg)
where A is a constant and E is the photon energy.
Figure 2 shows a Tauc plot ofdependence ex upon E for multilayer structures with a-C:HInarrow band gap layers with thicknesses (dl) in the range from 10 A to 80 A (curves 15) and thickness of wide band gap layer with dz =50 A as well as dependence a(E) fora-C:Hz film (curve 6).
It can be seen that these graphs are linear over a wide range of photon energy values, andabsorption in the a-C:Hz layers is small. The optical band gap of the superlattice that isEgi narrow band gap can be found by linear extrapolation to a = 0 dependence (a*E)112onE.
The dependence ofEg on a-C:HI thicknesses for multilayer structures is presented in Fig.3(a). It is seen (Fig. 3(a» that the experimental curve may be divided into two ranges.While the thickness of the a-C:HI layers were more than 35 A, the Eg of the structureswas 0.5 eV and did not change with thickness. This value is in good agreement with Eg= 0.55 eV of the bulk narrow band gap materials. However, the Eg value increased from0.5 to 0.9 eV with decreasing thickness of a-C:HI from 35 A to 10 A. Such an increasein Eg may be interpreted by confinement of electron wave function in well.
For a simple quantum well model, the increase in Eg is proportional dl-z(me-l+mh-l)where me and mh are the effective masses for electrons and holes, respectively. The observed experimental dependence ofEg on dl agrees well with the above-mentioned prediction (Fig. 3(b».
Current-voltage characteristics of the structures with a-C:H quantum wells and barriersdepended on the relative thicknesses of inner and outer barriers. For samples with thesame thicknesses of all barriers, experimental curves showed no clear evidence of quantization effects and were similar to those for MIS structures with tunnel-thin insulator.Following [9], we assume coherent resonant tunnelling to be suppressed with strongerlateral effects due to the great discrepancy in tunnel transparencies ofouter and inner barriers. As a result, the bias voltage is applied mainly to outer barriers. It was shown [10]that lateral effects may be weakened to a great extent by changing the relation betweenthe outer and inner barrier thicknesses.
Experiments with a-C:H MQW-structures deposited on p-Si substrates with lateral barriers being thinner than inner ones, approved the model [10] and revealed sharp kinks inreverse branches of the I-V characteristics. Figure 4 shows dark and light I-V characteristics for the MQW structures composed of two potential wells 5 nm thick and three barriers: the inner one 5 nm thick, and the outer twice as thin.
261
1.0 a
0.8E~
reV] 0.6
10.050 7.5d,[lI-m]
2.50. 'I L.-_----l__--'-__--L__-..J
o
1.0
0.8E8
reV] 0.6
O.lIo 0.2 0.'1 0.6 Qg
d-~ [fl.m- 2j
1.0
Figure 3. Dependence of the optical band gap on the well thicknesses: (a) Eg V.s. d· t/2 and (b) for multilayer structures.
The dark I-V curve in Fig. 4(a) has a kink at VI =-1 V with corresponding growth of thedifferential conductivity O'D in this region. Under illumination O'D grows further, moreover, the second region of O'D changes sharply at V2 = -5 V (Fig. 4(b». Temperature lowering results in O'D non-monotony growth due to background scattering decreasing.
A tunneling current should reach maxima when resonant tunnelling conditions are fulfilled and the bias applied equalizes the energy levels in quantum wells (Fig. 5). SmallerO'D values at low reverse biases (S; 1V) may be a result of the built-in electric field causedby the charge near the a-C:H-p-Si interface. Irradiation of the structures with the light absorbed in Si substrates increases the concentration of minority carrier which take part inresonant tunnelling and results in amplification of the effects observed. Similar resonant-
262
20 8
~ 10 ~ ",-,",
~I
.~ 0 0ClQ~I
~. ~lU ~
~ ~~~ -10 -f~
-8
-z 0 2SIASfv}
~ t
~ 2 1 ,:::-...I
~t--.~I
~ ~~ 0 a "-.J
ClC ~q~:::) -2 -1~
-I( -2
-6 -4 -2 2BI-lS(V)
Figure 4. Current-voltage and differential conductance aD characteristics of MQW structure shown inFig. 5: (a) in darkness, and (b) under illumination
263
a-C:H c E~"
Q- C:i-I c E~1Figure 5. The conduction band diagram of the MQW structure investigated. E t
l , El and E l2, E2
2• thefirst and the second energy levels in the first and the second quantum wells respectively.
tunnelling pecularities were also revealed in the I-V characteristics of these structures[11].
Electroluminescence in visible spectrum range has been observed in multilayer thin filmheterostructures at room ternperature. Heterostructures used were the alternative thin carbon layers with a variable band gap; these alternative thin carbon layers were depositedonto monocrystalline p-doped silicon surface by different ion-plasma deposition methods. A peculiarity of these luminescent structures is a moderate thickness of carbon layers (-25 AO) comparable with de Broglie wavelength for electron. Upper metalsemitransparent electrode was deposited by thermal vacuum evaporation method.
Electroluminescence was occurred in metal (Au)-multilayer carbon film - semiconductor(Si) structures. Voltage-current characteristic is of a rectifying character. Luminescencewas generated on inverse branch of voltage-current characteristic at bias voltage of about10-15 V. The generation of luminescence on inverse branch was in agreement with steeprise in the current. Electroluminescence was detected by visually inspection as viewedfrom semitransparent metal contact and this electroluminescence manifested itself in thestable green or yellow-orange luminescence that is uniform in area and which is reproducible to the point of heterostructure breakdown. Integrated intensity of electroluminescence increased proportionally with the steady leakage current passed through thestructure.
264
Physical mechanism of observed effects is apparently dictated by the radiative recombination in quantum-dimensional thin carbon layers; the nonradiative recombination at layer-layer interface was suppressed by a great quantity of hydrogen atoms inheterostructure. Obtained results have opened up fresh opportunities for creating noveloptoelectronic devices based on these heterostructures, in particular, the visible lightsources with a variable wave length of light being radiated. Such sources are necessaryfor the advent of the screen planes and they can be used for development of optical systems of information processing (optical computer) in the future.
4. Conclusion
In conclusion it has been shown that it is possible to synthesize amorphous superlatticeson the basis of carbon layers with different optical properties in which the effect of optical absorption quantization may be observed. X-ray study has proved the presence ofstructure periodicity. A complex study of heterostructures with a-C:H layers has discovered some features in electrical and photoelectrical characteristics due to resonant tunneling effects. Electroluminescence in visible range at room temperatures may be alsoconnected with radiative recombination in quantum dimensional thin carbon layers.
Taking into account a simple and cheap technology of amorphous a-C:H heterostructures, using such structures or quantum effect devices fabrication in micro- and optoelectronics looks very promising.
References
1. S.Miyazaki, Y.Thara and M.Hirose. Phys. Rev. Lett., N 1 (1987), 1252. I.Pereyra and M.P.Carreno. J.Non-Cryst. Solids, 110 (1989),1753. Z.M.Chen, J.N.Wag, X.Y.Mei and G.L.Kong. Solid State Commun, 58(6) (1986), 3794. G.F.Ivanovsky, V.V.S1eptsov, V.M.Elinson and P.E.Kondrashov. Electron. Promyshl., N 2 (1989) 265. P.Conders and lCatherine. Thin Solid Films, 146 (1987), N 1,936. J.Robertson and E.P.O'Reilly. Phys. Rev., B35 (1988) 29467. V.V.S1eptsov, A.A.Kuzin, G.F.Ivanovsky, V.M.Elinson, S.S.Gerasimovich, A.M.Baranov and P.E.Kon-drashov. J.Non.-Cryst Solids, 136 (1991), 53-59
8. J.Tauc, R.Grigorovici and A.Wancee. Phys. Stat. Sol., 15 (1966),6279. Y.I.Jiang and H.L.Mwang. Jap J.Appl. Phys, 27 (1988) L243410. L.V.Iogunsen. Len. J.Techn. Phys., 13 (1987) 114311. V.I.Polyakov, P.1Perov, M.G.Ermakov, O.N.Ermakova, V.M.Elinson and V.V. Sleptsov. Thin SolidFilms, 212 (1992), 226-231.
THERMAL STABILITY AND STRUCTURAL REACTIONS AT THETANTALUM/a-CINTERFACEUNDERVACUUMANNEALINGCONDITIONS
AP. NOVIKOV, E.A. SHILOVA, L.D. BUlKO, and V.A ZAIKOVResearch Center of Electronic Materials and TechnologiesP. O. Box 66 Minsk, 220029 Belarus
Abstract
Thermal stability, structural changes, and reactions at the tantalum/amorphouscarbon(Ta/a-C) interface in the annealing temperature range from 100 to l000°C havebeen investigated by AES, RBS, TEM, and Raman spectroscopy. It is found that800°C is the threshold temperature at which carbon atoms begin to diffuse into thetantalum film and form inclusions of a carbide phase.
1. Introduction
Recently much attention has been given to development of electron devices based onthe crystals of diamond and diamond-like materials [1,2]. It owes not only to thelimiting characteristics of diamond devices but also to their wide potentialities incircumstances when traditional semiconductor device fail to operate. On creation ofany semiconductor device the formation of ohmic contacts is one of the importantstages since just the quality of a contact specifies the characteristics and the reliableperformance of the whole device. At present many works are known concerned withvarious aspects of the formation of ohmic and barrier contacts to diamonds anddiamond like materials. Structural and phase changes at the AI/diamond, Ti/diamond,Ni/diamond interfaces at different annealing temperatures have been investigated in [3].Some works are devoted to the Ta/diamond system [4]. It is established that duringthermal annealing the corresponding carbides are formed in a metal film. It is shownthat the thin films of carbide-forming metals may be used for formation of satisfactorycontacts to semiconductor diamonds under annealing conditions.The present work deals with investigation of structural and phase changes and
reactions at the Ta/a-C interface exposed to steady-state vacuum annealing for an hour.This system, being interesting in its own right, may be also employed as a model forforecasting of the Ta/diamond interface Properties.
265
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 265-270© 1995 Kluwer Academic Publishers.
266
2. Experimental
Carbon and tantalum films were deposited by a d.c. magnetron sputterer. A diameterof flat targets prepared from pyrolitic carbon and tantalum of spectral purity was 110nm. Prior to deposition the target was obligatorily "trained". For this, the magnetronsputterer was operated at the maximum power for 30 min for the carbon target and for10 min for the metal target. As substrates, we used 100 nm dia silicon wafers. Thedeposition time was 90 min for carbon and 2 min for tantalum.The preliminary studies have allowed us to choose the following deposition
conditions for carbon (tantalum) films:The discharge chamber pressure 0.3 Pa (0.3 Pa), the discharge power 25 W (480 W),the bias potential -60 V (-100 V).The Ta/a-C/Si thin films obtained were annealed for 1 hour in a vacuum furnace
with a residual gas pressure of 10-5 Pa in the temperature range from 100 to lOoo°Cwith a subinterval of 100°C.Structural and phase changes at the Ta/a-C interface were studied using such
experimental methods as AES, RBS, TEM, and Raman spectroscopy (RS).A structure of carbon films of the initial sample and its change in the course of
thermal annealing was analyzed by the RS technique at an operating laser wavelengthof 488 nm. Changes in the structure of tantalum films and phase formation processeswere detected by TEM at the accelerating voltage 200 kV. Redistribution of carbonand tantalum atoms under thermal annealing conditions was controlled by measuringthe Auger-electron spectra of carbon and tantalum atoms with successive removal ofthin layers of the tested samples by the sputtering (Ar+) technique.All-round analysis of the interface processes provides obtaining of the necessary
information on the atom diffusion, structural and phase changes in the Ta/a-C system.
3. Results and discussions
3.1. STRUCTURE OF CARBON FILMS
In the majority of cases a structure of carbon materials is reliably identified by theirvibration spectra using the RS technique. For instance, natural diamonds arecharacterized by the presence of the vibrational mode at 1331 cm-), for diamond-likefilms a vibration spectrum is localized at 1540±20 cm-', while for graphite at 1581cm-' [5,6]. According to experimental data (Fig. La) the RS spectrum of the initialcarbon film is indicative of its amorphous structure. This is confirmed by two diffusepeak 1350 cm-' and 1550 cm-). It is commonly considered that the peak 1350 cm-' isdue to the sp3-bonds while the peak 1550 cm-! to the sp2-bonds in the carbon film.An amorphous nature of the initial carbon films is also confirmed by electronmicroscope studies that have detected the presence ofbroad diffuse rings on diffractionpatterns and by analysis of film microstructure under ooסס20 magnification.
;::l
nj
;>.. a+-'.r-!7;1)
~Q)
+-'~
b.....
c
267
1700 1500 1300 1100
Wavenumber, cm-'
Figure 1. Raman spectrum of carbon films: a) before annealing; b) Tann=400°C;c.Tann=8OO°C.
Vacuum annealing of the films in the temperature range from 100 to l()()()OCchanges an intensity and a half width of the peak 1550 cm-1• In the high-temperaturerange the peak bifurcates and forms the extra peak at '1585 cm-I(Fig. lb,c). Such astructure of the films is known as "bridge graphite" or "diamite" [5].The results obtained suggest that structural changes in the initial carbon films must
be taken into consideration when analyzing the interface processes of a system underinvestigation.
3.2. STRUCTURE OF TANTALUM FILMS
As judged from the electron microscope studies at 200000 magnification, the initialtantalum films hl\ve a fine-dispersed polycrystalline structure with a grain size of 5-7om. The TEMdata point to inhomogeneity of the metallic films throughout theirvolume and allow us to separate two regions which differ in a phase composition. The
268
diffraction patterns of a near-surface layer demonstrate characteristic rings with radiicorresponding to the Bragg reflection from the (110), (200), (211), (220), (310), (321)planes of tantalum. Metallic tantalum possesses a body-centered cubic (bcc) latticewith its constant a=0.331 nm. Analysis of the diffraction patterns of deeper layers ofthe metal film has revealed the presence of extra signals which, in our opinion, pertainto the Bragg reflection from the (002), (004), (110), (114) planes ofgraphite. Graphitepossesses a hexagonal lattice with the parameters a=0.364 nm and 0=39°30'. Theindicated changes in the film structures may be attributed to the grafitization processesinitiated by carbon atoms (up to 5% at) present in a volume of the metal film.Graphite microinclusions may be responsible for a decrease of a mean grain size in thetantalum film. By the annealing temperature 400°C mean grain sizes of the metalattain 2-3 nm. A fixed increase of dispersion may be also caused by relaxation ofstresses developing in the metal film after deposition.A further increase of the annealing temperature entails a decrease of dispersion of
the tantalum film, mean sizes of its grains attain 15-20 nm at l000°C. In this case,inclusions of tantalum, not graphite, carbide are detected in the unreacted film.
3.3. DIFFUSION AND PHASE FORMATION IN THE Ta/a-C SYSTEM
Analysis of the TEM and AES data have allowed us to single out two temperatureregions in which the Ta/a-C system has its special features of a phase composition andstructural properties.(a) T< 800°C. The spectrum ofAuger-electrons coincides with the initial spectrum
thus indicating that the Ta/a-C interface has not changed (Fig. 2a). The electronmicroscope data fix the signals on diffraction patterns that pertain to metallic tantalum(the bcc lattice with its parameter a=0.331 nm) and to graphite (the hexagonal latticewith the parameters a=0.364 nm and 0=39°30').(b) T~800°C. For the tested Ta/a-C system subjected to one hour thermal
annealing, 800°C is a threshold temperature that is characterized by diffusion of carbonatoms into the tantalum film and by the onset of solid-phase reactions at the Ta/a-Cinterface. This is evidenced by the appearance of special "step" on Auger-electronspectra (Fig. 2b). At l000°C a series of diffraction rings has been detected ondiffraction patterns which testifies to formation of inclusions of a new phase in themetallic tantalum film. A comparison of the calculated interplane distances of asubstance of the new phase with those known for tantalum compounds allow us toconclude that the appearance of the extra reflection signals on the diffraction patternsowes to formation of polycrystalline tantalum carbide TazC possessing the hexagonal lattice with the parameters a=0.31O nm, c=0.494 nm. The disappearan;;e of reflectionfrom the corresponding planes of graphite is, probably, due to its interaction withtantalum atoms to form carbide inclusions in the film.The AES data indicate that this temperature interval is characterized by
enhancement of carbon atom diffusion in the metal film and formation of an interlayerof the carbide phase. A growth process is accomplished by diffusion of carbon atoms
269
a
3
c
21o
(20
60
100
20
0 1 2 3
100 C+Jro b
:--.: 60
c0'--i+J 20ro~ 0 1 oj j+J L~<L' 100 (',-'c0 cu
60
Sputter time. mIn
Figure 2. Auger-electron distribution profiles of elements in Ta/a-C: a) prior to annealing;b) Tann=800°C; c) Tann-lOOO°C.
through the formed carbide layer to the Ta/a-C boundary.After the one-hour annealing at lOOO°C the process of carbide formation in the
Ta/a-C system throughout the tantalum film has not been completed as confirmed by
270
the TEM data and by the published AES spectra of carbon and tantalum atoms(Fig.2c). In [4], where the possibility of forming the ohmic contacts to naturalsemiconductor diamonds of the II type has been investigated, the authors are of theopinion that the tantalum-diamond substrate interaction is completely over afterone-hour thermal annealing at 885°C and that the products of this interaction showstability during extra one-hour annealing. Our data disagree with this statement aboutcomplete cessation of the interaction processes in the Ta/diamond system. However,this may be explained by a considerable difference in a thickness of the initial tantalumfilms. In [4] the metal film was deposited by the electron-beam sputtering and itsthickness was 8 nm, while in our studies a thickness of the tantalum film deposited bya magnetron sputterer attained 150 nm.
4. Concluding Remarks
The studies performed have revealed the following special features of the interphaseinteraction at the Ta/a-C interface:
-the temperature 800°C characterizes the onset of diffusion of carbon atomsin the metal film which brings about the inclusions of a carbide phase;-a further increase of the annealing temperature is accompanied byenhancement of the carbon atom diffusion from the Ta/a-C interface towardsthe metal film surface and is characterized by formation of a carbideinterlayer;-after one-hour annealing of the Ta/a-C system at lOOO°C the carbideformation process in the metal film has not been completed yet.
5. References
1. Gildenblat G.Sh., Grot S.A., Wronski C.R., Badzian A.R., Badzian T., andMessier R. (1988) Electrical characteristics of Schottky diodes fabricated usingPlasma assisted chemical vapor deposited diamond films, Appl. Phys. Lett. 53,586-588.
2. Himpsel F.J., Heimann P., and Eastman D.E. (1980) Schottky barriers ondiamond (111), Solid state Commun. 36,631-633.
3. Lurie P.G., and Wilson J.M. (1977) The diamond surface, Surf. Science 65,453-474.
4. Moazed K.L., Nguyen R., and Zeidler J.R. (1988) Ohmic Contacts toSemiconducting Diamond, IEEE Electr. Dev. lett. EDL-9, 350-351.
5. Huong P.V. (1991) structural studies of diamond films and ultrahard materialsby Raman and micro-Raman spectroscopies, Diamond Relat. Mater. 1,33-40.
6. Robertson J. (1986) Amorphous carbon, Advances in Physics 35,317-339.
EXTENDED AND WCALIZED ELECfRONIC STATES IN TETRAHEDRALCARBON FILMS
V.E.MASCHENK01, V.M.PUZIKOv2, A.V.SEMENOv3Institute ofSteel and Alloys1,Institute ofSingle Crystals ofthe Ukrainian Academx ofScience;Moscow 117936,' Russia1, Kharkov 310141, Ukrain~
KEYWORDS/ABSTRACf: tetrahedral carbon films/ absoprtionluminescence/electronic states/ Urbach tails/ hexagonal cubic sp3-carbon/electronic disorder
The disordered tetrahedral carbon films have been grown by ion beamdeposition at -120 eV. An absorption maximum 6.15-5.80 eV (IcY and a threshold4.8-4.5 eV were resolved and tentatively attributed to ~5 - rSand ts-K:ztransitions in lonsdaleite. The cubic sp3-carbon manifested as maxima at 5.45-5.50eV (1"25 _t:...x) and 7.0 eV (rb - r15). Urbach tails of different steepness wereresolved. A 6.508 eV maximum of 0.52 eV half-width at 90K was resolved. Thespectroscopic data on absorpsion and luminescence of sp3-bonded carbon fllms ofdifferent allotropies are discussed.
1. Introduction
Great interest was created in growing and characterization of tetrahedrallycoordinated carbon films without long range ordering [1-3]. The fllms were grownat low temperatures on numerous substrates by several processes [1-6]. In the caseof low energy carbon beam deposition the prevailing content (up to 90%) of thesp3-carbon orbitals is due to a high compressive stress generated in the growinglayer by particles with the energy of 70-150 eV [3]. In the subsurface layer with ahigh concentration of the electronically and vibrationally excited of "subplanted"carbon [5], local pressures and temperature may exceed the critical PT-valuescorresponding to the diamond phase stability [1].
The properties of unhydrogenated tetrahedral carbon fIlms are expected tobe similar to amorphous diamond [2, 3] whose structural and electronic disordermay be close to noncrystalline group IV compounds [7]. According to [3] sp3_bonded carbon fIlms grown by low-energy ion beam process have displayed the
271
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 271-283© 1995 Kluwer Academic Publishers.
272
structural disorder which was close to disorder in amorphous germanium.Alternatively the electronography data simulation [8] gave a high content of sp2_bonded carbon in this type of films. The electron diffraction data [9] for filmsdeposited from 20-300 eV carbon ion beam indicated the dominant content of sp3_carbon in the form of small hexagonal crystallites. The optical properties involvingdelocalized and localized states of low energy ion beam deposited carbon films arediscussed in [10-12]. The maxima~,~ with exponential low energy tails prove tobe resolved in the absorption spectra at 90K - 300K The intrisic maxima were dueto cubic and tentatively hexagonal sp3-carbon disordered allotropes.
This results were confirmed the growing of two forms of sp3-carbonsimultaneously. However, in some spectra the only lfa maximum was resolved.According to [10] this maximum is proton bombardment stimulated. The spectrawith the only intrinsic absorption maxima of cubic disordered diamond phase wereobserved too.
It is instructive to determine the deposition parameters responsible for thegrowth of sp3-carbon allotropes. We have studied the absorption and secondaryemission in disordered diamond fIlms of different allotropic forms. The absorptionspectroscopy at 6.4-1.5 eV, the secondary emission under N2-laser or tunable Arlaser excitation, Auger spectroscopy, SIMS, as well as above mentionedelectronography results were used to estimate the properties of the films and todetermine the deposition conditions for sp3_films of different structuralmodifications. New data have been obtained on the fine structure of fundamentalabsorption edge and character of electronic disorder in tetrahedral carbon films.
2. Experimental procedure
The design and characteristics of ion-beam system and the data on thestructure and composition of sp3-carbon ftlms were discussed in [13]. Carbon ionswere generated by arc discharge using evaporative :athode made of high-puritygraphite. The introduction of methane or argon into the discharge chamberallowed ion current to be increased markedly. The optimal conditions [14] for ftlmgrowing include the use of - 120 eV ions and substrate temperature of about50oC. The pressure was maintained at a 10-2 Pa. The diffraction data show that theconcentration of hybridized sp3-carbon atoms in the ftlms was))O%. The SIMSgives -3% hydrogen and -1% oxygen concentrations. The techniques used torecord the optical spectra and its computer processing to determine the Urbach tailslopes were described elsewhere [10, 11]. Luminescence was excited by an N2-laserup to the pumping rates which give rise to ftlm ablation «1028 cm-3 s-l). Atunable Ar-Iaser which may supposedly excite (via stepwise local transitions)delocalized states was employed.
273
3. Results and discussion
3.1. Absorption
Weakly temperature depended Ito, E~ maxima of 0.6-0.7 eV halfwidth withextensive exponential tails were described in [10-12]. These feature may beinterpreted as evidence for strong electronic disorder in carbon films. Theestimated well depth at the band boundaries may amount 1.0-0.9 eV. A key featurewas a higher Ifo contrast compared to rather diffusive~ maximum. EO maximumexhibits a substructure of overlapped equidistant submaxima. Energy changes of~ maximum (due to internal stresses [3, 5]) were also observed.
The spectra shown in Fig.1a,b illustrate the feasibility of depositing the ftlmswith two types of their intrinsic maxima above 5.5 eV and exponential tails below5.4 and 4.8 eV respectively. The survey spectrum 2a contains the overlapped 5.63eV maximum 0fcJ, indirect r2s --4 transition in disordered cubic sp3-carbon [10,12], the 5.2-4.6 eV extrinsic plateau [10] and low energy Urbach tail. In thespectrum 1a Eo maximum at 6.03 eV and low intensity exponential tail below 5.4eV were resolved [10, 14]. The low-energy tails of ~ maxima exhibited anexponential shape of 03-0.7 eV steepness. Spectra of 3, 4 on ftg.1b illustrate inmore details the shapes of two intrinsic absorption maxima and it's low energyexctrinsic maximum (spectrum 3) and Urbach tail (spectrum 4). The absorptioncoefficients 25 cm-1 at 1.92 eV and 40 em-I at 1.7 eV were measurecl in spectrum 4and 3 respectively. According to energy band calculations [15], the :Efo maximum of0.43-0.90 eV half-width at 6.19-6.00 eV may be identifted tentatively to be a directtransition in disordered hexagWlal carbon [10, 11, 14]. We have discussedqualitative relationships of the EO maximum to excitons localized on compositionfluctuations or to self-localized excitons produced by the electrons and holes of r~and rSstates which are coupled to phonons in partly-ordered hexagonal carbonlattice [14]. Extension of the measurements to higher energies revealed anadditional maximum at -7.0 eV snd gave more information on the shape of~maximum. The spectrum of 900 A thickness fIlm on KCI (Fig.2) deposited at 0.5rnA em-2 ion current density 180 eV and 2000C with CR4 added to the reactorexhibit only the -7.0 eV and 5.45 eV components with their Urbach tails of 0.180.20 eV slopes below 5.40-5.36 eV. The position and high energy tail of 7.0 eVmaximum are instrumentally distored. According to [16] and to the energy-bandcalculations [15] the maximum near 7.0 eV may be due to a direct transition in thedisordered cubic sp3-carbon lattice. These features indicate that the 'l-transitionsare dominated in absorption of this ftlm.
The diffusition and weakening of the~ maximum, as well as a relative
274
7So432.62i
-I1I
'5.3095.73I.. l....
.JviJ:) 5011872 ....j
:::I I
iI
39Ell0.71 -1.31 G22.77
2511&';6
t99~,2.r::2
...._-- --.......
-'.,
...............
/
.~.•."".
t
:::.
eV
'.1454.3j : .-.------------------ ._-----------_..
"'.
300K
3.......
'-.
.-........
/ ..•..
"V"T", / f \\--1-+ r.-'v', ....JrS- 5/\ r2S- Ax
I I \
I ./ '.1-,/ "l '-.1 '.l '.
j \\~ \
....j~; \\, " \. \
'----__ It \--------- \,
~ ----------~
10000
1.3~.J3.b4
'.~'82.51.3
~1:':"'O.10t1
7585.775
7943.282
1047128
8317.G37
7244.359
6.2 5.6 5.4 e,V 54.6 4.2 3.8
Figure 10, b. Survey absorption spectra of two carbon films (1, 2) and spectra of films in the region of
intensive intrinsic maxima (3, 4).
275
weakness of the r~ - IS transitions were discussed elsewhere [14] allowance forthe oscillator strengths of direct and indirect transitions and for a marked overlapof energy ranges in the ris -~ and r! -IS indirect transitions. The absorptioninvolving the localized state is another process which smeared the extrema.
-90K20
12 ,. ---- 300K':;:E
,/-. ,
~~<.>
r~i~x'e 8 ,,~
x -' en, :zen " <C(J) '.- 10 !=-< 4 E'"0
.1
eV 62 4.6 3.0
Figure 2. Absorption and transmission spectra of cubic sp3-carbon film of 900:' thickness.
In spectra of some films growed at high currents~ absorption maximum of0.40-0.50 eV halfwidth exhibits the fine structure of submaxima at 300K Theenergy interval between neighboring submaxima is clRse to TA-phonon energy(-0.102 eV) at the X point of diamond [16]. The EO maximum decreases inintensivity and widenes, while its fme structure disappears at 90K
The subband series on the B8 contour may be a spectroscopic evidence forthe strong contribution of electron-phonon interaction to the shape of themaximum. Its Gaussian shape and temperature behavior indicate that theelectronic excitation may be localized in the disordered sp3-carbon medium.
Singularities in the region of the :r! -IS indirect transition 4.7-4,8 eV wereresolved in the spectra of the fllms with intensive Ifo maxima and weak Urbach tailabsorption below 5.5 eV-2.2 eV [14]. Fig.3 shows the spectra from the neigboringareas of a fIlm with an intensive 6.1 eV extremum (~ • rs> and a 4.7 eV step. ThefIlm is of different thicknesses because of radial nonuniformity of ion beam. Theintensive step in the region of the I!-IS transition expected from calculasion [15]is well pronounced in the fIlm areas of greater thicknesses (spectrum 2).
In spectra of some fUms the UV-structure parameters are characterized by astrong temperature dependence. FigA shows the temperature dependence ofoptical density and transmission of the film on KCl of 900 ~ thickness. The doublet6.141 eV and 5041 clearly resolved in spectra at 90K smeared out into 5.952 eVmaximum accompanied by extended Urbach tail up to 1.90 eV at 300K There is agood probability than two maxima at 90K are corresponded to disordered
276
hexagonal and cubic sp3-bonded carbon. We have obtained spectroscopic dataindicating a definite ordering in carbon films. In some spectra the structured withabove mentioned energy interval contrast and gaussian-shaped Eb maximum of 0.4eV half-width and a plateau - 5.0 - 4.8 eV were resolved. The absorption in a weakUrbach tail with a 1.0 eV slope reached <102 em-1 at 4.0 eV. The maximum waswidened and the intensity of exponential tail strongly built up at room temperature.
14 r:-,...--;---:-:-----:----;--;---;-----_-.-_-9-0-K--_Fa==-.-': 300K 1
~~j 10 asj ~J ~
4.6 3.0
Figure 3. Temperature dependence of optical spectra with intensiveEgmaximum.
6.2 ~.6 3.0 .... eY20
lStK2, ,, -90K,, ....-300K,t.O ,,
,,,,,
::>->!-::;:;= in:=:: 0.5 ---.......-==.......;::: S-=-C:J,
50 ~o 30 20-10 3CM-i
Figure 4. Absorption and transmission spectra of a sp3-carbon films of 700Athickness.
277
Adjustment of the ion source parameters permitted the films with theirabsorption edges above the E~ maximum to be deposited [12, 14]. The spectraexhibit a temperature-independent intensity maximum at 6.508 eV (a 0.52 eV haHwidth) and 6.417 eV (a 0.72 eV half-width) at 90K and 3OOK, respectively, as wellas a weak Urbach tail of -0.7-1.0 eV steepness below 5.95 eV. Simultaneously,some spectra contain the overlapped E~ and 6.508 eV maxima. We have resolvedthe spectra with an anomalously-shaped and strongly cool-quenched part between5.9 and 6.4 eV. Fig.5 shows the spectra of three films which include the newmaximum (1) the ifc> structured maximum (2) and a steep step at 6.013 eV (3)quenched strongly at 90K All the maxima are accompanied by exponential tails ofdifferent slopes. The insert illustrates in more details the temperature dependenceof the parameters of transmission extreme. In spectra 1 and 3, the high-energywings of their maxima are instrumentally distorted. At the same time; anomaloustemperature bleaching of transmission between the steps at 6.01 and '6.40 eV isquite evident.
~.
2:23.03.84.65;4eV 6.2
~v; 6.6 S.B 5.0, II I /1 ,. r-'------ 20
i~ i\ t, /l ./3E- 'I; If i :/. \": ./
"~-'j, :\ .: { j'-. .'," , ;./"
i 2 :, ;I i \ : ;"'("':r. .: \; . i .; J'. ••.. \' /' : i.. .;::::90K: \'- ...:~.."", . N/ \ : ~ ': )! ; I .=-:=300K1 " 2~ '. ''''. ~! :' "'. ". 3 '~..~ 1,1,
'-\ __ ,~;;;~~;",:,~:::: __ a0.0
:a! 0.0C-J
~0.3
Figure 5. Temperature dependence of optical density and transmission spectra of three carbon
allotropies.
3.2. Luminescence
The experiments carried out at 90K under different levels of excitation onfl1ms characterized by the absorption spectra shown in Fig.I-3,5 have demonstratedthe occurrence of strongly overlapped 2.407 eV and 2.680 eV luminescence maxima
278
of 0.48-0.64 eV full half-width [12, 14]. The energy, the overall halfwidth and therelative intensities of two maxima were sample-dependent and were changed withlevels of excitation. Fig.6 shows the temperature (10, b, c) levels of excitation (10, a,b) and N2-laser spot «50 JllII1) position (2-4, lOa) on the excited surfacedependencies of emission spectra for the film with extended Urbach tail. Theabove mentioned peculiarities of the doublet emission band are well evidenced.The value of full halfwidth of the doublet changes upon excitation at 90K in theinterval 0.85-0.60 eV. The fine structure of the doublet is smeared out and thehalfwidth of Gaussian diminished at 3001<.
390 450 510 570 nM
602591
=-90K
---- 300K
./,-,-.-.-." ..../ '\ ". 3
1/··..... '\ ".I' "-.---, . 2-;" ./ ',\ '\ .
/ / "'---" '. " ....../ /' .' '\. ". ·····4.....' ..' ." .
./ ./ / '. "-. ......- /' .. ' '-.,
../ ...,." ··'_~··_ll.lO/' .. ' """
-"," )
eV 3,0 2,8 2,6 2.4 2,2 2,0
Figure 6. Position and temperature dependence of doublet luminescence band for the carbon film
with intensive exponential tail.
No correlations between the position and the shape of the fundamentalabsorption extrema and the given emission spectrum were discovered. Data onparameter dependencies of two-component band emission at 90K upon level ofexcitation presents Fig.? b, C, d. The absorption of this film is presented by
279
spectrum 2a, on Fig.1. The intensity of the doublet enhanced sublinearly as afunction of excitation. Index of proportionality being 0.4-1.1 depending onexcitation intensity' and radiation recombination yield of the film under study.
Despite of scattering of experimental points on curves ~) the high-energyshift of submaxima upon excitation is evident. The halfwidth of the band emissiondiminishes from 0.68 eV to 0.59 eV under two order of magnitude strengthening oflaser intensity.
400 450 500 400 450 500 550 nM
a.
"
, ."
........ '.
6111.1
.=:: 90K
•••• lOOK
. .
eV 2.8 2.6
'.
2,4 2.2
eV 3.0 2.8 2.6 2.40.7 d
c 00 I) 0° 0 -°°0 0 . :J. 0.6 ci
10' ~
0.5Cll
>2.6eu > ' .. <:
,,,,4',A 6.. ,,4 6 .,5i2.6 :S
.. c... 0L
,,'&A,AAAAAA6 "0 . '"~24 '3 .. '"UJ .. . E
;;;;- UJ
c 10'~
10' 10' 10' 10' 10' 10'
Exilalion. a.u. Exilalion.a.u
Figure 7. Pumping parameters dependence of luminescence spectra of sp3-carbon film, 13, 7a - two
different points of the surface with "blue" emission
Moving the laser beam on the mm surface reveals bright blue spots withsufficiently another spectrum. In spectra of Fig.8 well shaped maxima 3.031, 2.849,2.678, 2.600 and 2.480 eV are resolved. A 6-9 meV high energy shift of the maximaupon excitation was obtained. The energy interval between three first maxima0.182 eV and 0.171 eV is not coincides with the energy of phonons in diamondlattice. This values are close to the energy of local modes in impure diamond [18]and data for CVD-fl1ms [19].
Fig.7 shows the dependence of "blue" spots emission spectra upon a positionof laser beam on the film surface (spectra 7a, 2a) and the temperature (3a). The
280
above mentioned sample depended variation of the energy of the ~ absorptionmaximum is observed in spectra of "blue" emission too.
Another spectrum of luminescence was obtained for films with absorptionspectrum 3 on FigS. Luminescence maximum 2.638 eV of 0.18 eV halfwidth wasregistered. The narrow 2.84 eV maximum and low intensity 3.024 eV maximumand 2.10 eV band were appeared under laser beam moving on surface of the films.Some areas of these f1lms possessed the spectrum of Fig.B.
The doublet 2.680-2.407 eV nearly coincides with extrinsic band A observedin natural and synthetic diamonds [20] which was attributed to donor-acceptor pairrecombination.
Very recently rich sharp lines structure of distinct donor-acceptor pairrecombination were observed in this region for CVD-diamond [19]. Our data onthe excitation dependence of the energy of two-component band confIrmed thisconception. More investigations must be done for identillcation of this doublet. Thesame is true for the "blue" emission spectrum.
We have observed the intrinsic emission in the energy interval 6.2-4.0 eVunder strong 488.0 and 514.5 nm laser fxcitation of some carbon films with diffuseemission maximum in the region of EO absorption extremum. The results will bepublished in a separate paper.
The studies of absorption and luminescence, as well as the preliminaryresults of electronography supplemented with the Auger spectroscopy, SIMS [9, 13Jmade it possible to identify the deposition conditions for monophase noncrystallinesp3-carbon films of three allotropic forms. For other equal in magnitude and
281
optimal parameters the decisive factor for deposition of sp3-carbon films of distinctmodification was the ion-beam current.
Two of the three optically identified allotropes may be associated tentativelyto be the disordered cubic and hexagonal tetrahedral carbon. To get anunambiguous identification, microscopic calculations for tetrahedral carbonallotropes are needed. This is of particular importance for the new carbon phasewith intrinsic high-energy absorption maxima and anomalous temperaturequenching of its spectra.
The films of different types show strictly-individual properties in theintrinsic regions and are characterized by theoretically predicted [21] exponentialabsorption tails in the transparency region. The parameters of the tails indicate thatthe sp3-phases are disordered and, therefore, there exists a random relief of theedges of the allowed electron and hole energies with the macrofluctuation welldepths which are an order as great as those in amorphous Si and Ge [7].
Tetrahedral carbon fllms showed a lot of extrinsic luminescence bands andliries under strong N2-laser excitation of Urbach tail localized states. Now onlyqualitative identification of 2.408-2.640 eV emission band is possible.
In conclusion let us discuss some expected details in optical spectra oflonsdaleite carbon films. The E~ maximum identification is based on the calculatedand known experimental energies of the r~ -r15 and ris -4 transitions in cubicdiamond, as well as on energy band structure of lonsdaleite [15]. The resolution ofthis maximum in spectra may be due to a high oscillator strength of the directtransition and to the spectral features of the calculated density state function [15].The data for the region of the indirect transition r~ - IS are less definite. Thistransition has a smaller oscillator strength, corresponds to the region of the strongUrbach tail and may show itself as a diffuse structure.
The single-electron calculation neglects spin-orbit and crystal-field splittingswhich transform the degenerate r25 state into the r 9 ' r 7 and r 7 states. This causethe fine structure and well known selection rules for optical transition in wurtzite.The fine structure may be violated by macrofluctuating electric fields wich smearr 9 - r 7 ' r 7 - r 7 and r 7 - r 7 transitions in optical spectrum of lonsdaleitecrystalline fllm. This make the situation more sophisticated. The single-electroncalculations [21, 22] yield the exponential tails of the fundamental edge boundaryfor amorphous semiconductor. For the uncorrelated short-range disorder potential,the optical spectrum of the fundamental edge is defmed by the generalized statedensity function. In the £2 (E) spectrum, the long-range correlated fluctuationsprovide for the dominating role of the transition matrix element with its extremumbetween the mobility boundary and the energy gap of perfect crystal [21]. Thecalculations [21, 22] do not make allowance for excitons which are distroyed inrandom fields of an amorphous media. The ODMR investigations [23] oU-Si:Hand amorphous glasses confirm the generation of excitons by light involving theelectrons and holes which belong to the state density tails.
282
The concept of excitons allowing for their C(.xtrinsic polarizability makes itpossible to explain the presence of the Gaussian EO in the spectra. Its substantialand temperature-independent width is indicative of strong exciton-phononinteractions which may result in exciton self-localization. The localization ofelectron excitations in potential wells near the energy band boundaries [24] mayalso turn out to be effective, if considerable macrofluctuation depths are allowedfor.
The author1 is grateful to DIAGASKRON for partial support of the work.
4. References
1. Angus J.C. and Hayman C.C., Science, 241(1988)913.2. Davanloo F., Juengermaan E.M. et aI., JAppI.Phys., 67(1991)2081.3. McKenzie D.R., Muller D. and Pailthorpe BA., Phys.Rev.Lett., 67(1991)m4. Aisenberg S. and Chabot RJ., J.AppI.Phys., 42(1971)2953.5. Lifshitz Ye., S.R.Kasi, J.W.Rabalais, Phys.Rev.Lett., 62(1989)1290.6. Aksenov 1.1., Belous VA., Padalka V.G. and Khoroshikh V.M., J.Plasma
Phys.,4(1978)425.7. Molt, N.F., Davis, EA.: Electronic Processes in Noncrystalline Materials,
Oxford, 1979, Molt, N.F., J.Phys. C: Solid St. Phys., 13(1980)5433.8. Jungnikll G., KuhI M., Deutschmann S. et aI. Diamond end Related Mater.,
1994 (in press)9. Puzikov, V.M., Semenov, AV., Surf. and Coat. Technol., 47(1991)445.10. Elyutin, V.P., Maschenko, V.E., Sumskoy, EA. et aI., Dok. Akad. Nauk
SSSR, 311(1990)1118: Maschenko, V.E., Soloviev, G.G., Proc.C-MRS'9OInt.Conf. Thin Films and Beam-Solid Inter., vA, Amst.l99O.
11. Maschenko, V.E., Puzikov, V.M., Semenov, AV., Sukhorada, E.P.: PreprintISC-9O-20, Inst. Single Cryst., Kharkov, 1990, Maschenko, V.E., Puzikov,V.M., Semenov, AV., Soloviev, G.G.: Preprint ISC-92-4, Inst. Single Cryst.,Kharkov 1992
12. Maschenko V.E., PllZikov V.M., Semenov AV.: Thin Films eds. G.Hecht,F.Richter, J.Hahn, Proc TATF/HVITF94 Intern. Conf. Verlag, Germany,1994, pA94.; ibid Phys. stat sol (a) 146/2 (1994) in press.
13. Puzikov, V.M., Semenov, A.V., Zosirn, D.I., In: Abstracts DIAMOND-92,Aug.31-Sept.4,1992, Heidelberg, Abstr. 8.7, ibid Preprint ISC-92-4, Inst.ofSigle Cryst., Kharkov 1993.
14. Maschenko, V.E., PllZikov, V.M. and Semenov, AV.: In: AbstractsDIAMOND-92, Aug.31-SeptA,1992, Heidelberg, Abstr. 13.6, ibid PreprintISC-93-4, Inst.of Sigle Cryst., Kharkov 1993, Abstracts DIAMOND
283
FILMS'93, Sept.20-24, 1993, Albufeira, abstr.12-072.15. Salephour, M.R, Sutputny, S., Phys.Rev. B, 41(1990)3048.16. Philip, H.R, Taft, EA.: Phys.Rev., 127(1962)159.17. Nohnson, Progr. in Semicond., 6(1963)180.18. Angress J.F., Goodwin, Smith S.D., Proc.Roy.SocA., 308(1968)11119. Dischler, B., Rothemund, W., Maier.K, Wild, C., Biebl, H., Koidl, P.,In:
Abstracts DIAMOND FILMS'93, Sept.20-24, 1993, Albufeira, abst.12-078.20. Dean, PJ., Lightowlers, E.C., Wight, D.R, Phys.Rev., 140(1963)552,
Denham, P., Lightowlers, E.C. and Dean, PJ. ibid., 161(1967)762.21. Derch, U., Gruewald, M., Overhof, H., Thomas, P., J.Phys. C., 20(1987)121.22. Abe S., Toyozawa Y., J.Phys.SocJpn., 50 (1981) 2185.23. Cavenett. B.C., Adv. in Phys., 30 (1981) 475.24. Baranovsky S.D., Efros A.U., Fiz.Tekhn.Polupr., 12 (1978) 2223.
OPTICAL PROPERTIES OF SPUTTERING AND GLOW DISCHARGE a-C:H FILMS
T.STOICA, A.DRAGOMIR, +M.GARTNER, C.MOROSANU, andG.PAVELESCU
Institute of Physics and Technology of Materials,Magurele, POB Mg7, Bucharest, Romania.+Institute of Physical Chemistry of Romanian Academy,Spl. Ind.202, Bucharest, Romania.
Amorphous carbon layers with high optical gap have been investigatedwithin UV-visible light range by optical transmission andspectroellipsometry methods. Samples were obtained by RF magnetronsputtering of carbon target, or by glow discharge decomposition ofmethane. A fit of spectral dependence of optical constants to thetransmission and ellipsometric measurements is given. The experimentalresults on optical constants are used in order to evaluate by EMAprocedure the diamond, graphitic, polymeric and voids composition of aC:H films. A large fraction of voids was found within all high optical gapsputtering layers. By addition of up to 5%H to the sputtering atmosphere,the optical gap increases and graphitic component decreases. Above 5%H,the optical gap decreases and a significant polymeric component wasobserved. By glow discharge technique at low temperature, a-C:H films ofa dominant polymeric type are obtained.
1. Introduction
Hydrogenated amorphous carbon films (a-C:H) have many potentialapplications such as hard, VIS-IR optical layers, wear-resistant coatingsand also as a novel semiconducting material /1/. Their properties appearto be caused by an elevated sp3/sp2 ratio in C-C bonds /2/,/3/. Hydrogenincorporation within the carbon films plays an important role in thestabilization of the tetrahedral coordination of carbon atoms and in theincrease of sp3/sp2 ratio /4/-/5/. The optical and electrical properties ofhigh gap un hydrogenated a-C films deposited by DC and RF magnetronsputtering were recently studied /6/,/7/.
In this paper we report data about the optical properties of the aC:H films deposited on glass substrates, in two different ways: by glowdischarge in methane and argon and by RF magnetron sputtering in argonwith various hydrogen concentrations. The films have been studied bytransmission measurements and by spectroellipsometry in the UV-VISrange.
285
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 285-290© 1995 Kluwer Academic Publishers.
286
2. Experimental details
The a-C:H films were prepared on fused silica and BK7 glasssubstrates by two different techniques: the glow discharge decompositionof methane 10% in Ar atmosphere (GD-samples), and RF magnetronsputtering of a graphite target in Ar or Ar with H atmosphere (SPsamples) /7/. For both methods, the deposition temperature was less than150C; the deposition pressure of the order of 10Pa. The thickness of thelayers was varied within 0.5 - 1.2 mm range. The deposition rate waswithin 0.4-1.0 A/s and 0.7-3.0 A/s ranges for glow discharge andsputtering methods, respectively.
UV-VIS transmission data from a Beckmann Model DUspectrophotometer have been used to evaluate by a procedure publishedin /8/, the absorption coefficient and optical gap as a function ofdeposition parameters.
For a more complete optical characterization of the films, a nullellipsometer /9/ was used. The back side of· glass substrates wereroughened in order to minimize back surface reflection. The values of therefractive index n and extinction coefficient k were obtained from themeasured ellipsometric ., A and incidence angles values, according tohomogeneous thin layer optical theory /10/.
3. Results and discussion
Complex reflectance was estimated from. and A ellipsometric datawithin the spectral region 300:700nm. Fit of theoretical curves of real andimaginary reflectance, to the experimental ones was performed using theknown values of the optical constants for the glass substrate.
0A6 • • • •I ! Rs. hory 1
i:--'-'-'~-'0. ~ ~ _ ~ -: .
:~:~J==..,
o,'"
oL----.:~wmt~a:J .....~_--.J1 U » u • up(,.-' -
Fig.la Fit examples of real (R.re) andi ....ginary (R. ia) parts of complexreflectance ratio for a sputtering (SPOll") s .....ple.
Fig.lb Fit ex....ples of real (R.re) andi ....ginary (R. im) parts of complexreflectance ratio for .. glow discharges_ple.
For the fit, analytical functions to describe refractive index andextinction coefficient were supposed. The function parameters and the
287
thickness were adjusted to minimize the mean square error, on all usedspectral range.
An extinction coefficient function on wave length 1, K=KO+K t(l-1/1g)was supposed. This correspond to a constant KO added to a spectraldependence function of a Tauc absorption type (K is a parameterproportional with the band density of states, and 1 correspond to theoptical gap). For the refractive coefficient functioh a parabolic typedependence on wave number was considered.
Figs. la,lb show examples of experimental and theoretical curves forsputtering and glow discharge samples. The obtained msq errors are ofthe order of 0.05. Spectral dependencies of refractive index correspondingto the best fit for various samples are shown in Fig.2.
1.711
Ij 1.7
.1 1•51.=:: ...,
t!1.
1.61 til
Fig.2 - Spectral dependence of refractive index forsputtering sample of various H percent and -. glowdischarge one.
For RF sputtering samples, thE' hydrogen concentration in argon(CH) was varied within 0- ~tl% range. The values of absorption coefficientevaluated from the ellipsometric data are in agreement with thoseestimated from transmission measurements. Fig.3 shows Tauc plotting ofoptical absorption coefficient spectra within 1.5 - 5 eV photon energyrange, for samples obtained through glow discharge (GD) and sputtering(SP) deposition methods. The absorption data are obtained using theprocedure described in 18/. The corresponding curve for glow dischargesample shows a higher optical gap with respect to sputtering samples, anda lower slope of linear dependence region of sqrt(a,,) function on". Thissignificant change of absorption coefficient, between the two type ofsamples, suggests quite different structures of the films.
As can be seen from Fig.3, the hydrogen addition to the sputteringatmosphere results in a decreasing of the absorption coefficient.
The maximum increase of the film transparency was obtained within3-5% H concentration range. The optical gap as a function of Cu has alsoa maximum value in 3-5% region. Fig.4 shows the dependence of the
288
optical gap on hydrogen concentration. The maximum optical gap ofsputtering samples, of the order of 2.SeV, is significantly lower than the4.4eV gap value for glow-discharge films. The higher Eg value for thefilms deposited by glow discharge technique denote a polymericpredominant component due to the possible incorporation of morehydrogen into the films /11/.
:I2Uiol.-_.L---.l.~--'----''---L.l.-...L-----'-----',
1r-----:::::::-------------,
2
Fig.3 - Tauc plot of absorption coefficient a on wavenumber ¥, for sputtering (SP) and glow-discharge (GD)samples (<11 - the hydrogen concentration in argonsputtering atmosphere).
SP - S8fTl)Ies
1.6 toCtfltJ-
6
,.8I.... --'- "'- ....... .J
o
Fig.4 - Dependence of optical gap (Bg) of a-C:Hsputtering layers on concentration of hydrogen (Cit) inthe sputtering atmosphere.
The results on refractive index and absorption coefficient can beused in order to estimate the component fractions of diamond, graphite,polymer and voids components /2/. A convenient way to compare theoptical properties of a-C:H films with those of supposed components, is to
289
Fig.5 - Spectral dependence of real Elpart of dielectric constant for sputtering(SP) and glow-discharge (OD) layers. Forcomparison, data /2/ are included fordiaaond, graphite and polymer.
Graphll
t..
--~------~---------_..
to
Il'
IE'2
Fig.5b - Spectral dependence of iaaginaryE2 part of dielectric constant forsputtering (SP) and glow-discharge (OD)layers. Data /2/ are included for diBDOnd,graphite and polymer.
use the spectral dependence of realand imaginary parts of the dielectricconstants, E L and E2' respectively.Based on deduced El and E2 values(Figs. Sa and 5b ), an effectivemedium approximation procedure(EMA) was used to evaluate thecomponent fractions. The data on oursamples are compared with the dataextracted from literature for diamond,graphite and polymer materials. Thespectral dependencies of E1 and E2from literature for crystallinediamond are smoothed for amorphouscase as described in /2/. Foramorphous polymer, we consider E
close to zero within our experimentalwave range.
The values obtained fordiamond, graphite, polymer and voidscomponent fractions are displayed inthe Tab.l. High percents of voids(above 45% ) are obtained for the SPsamples and only 10% for GD ones. Aslow decrease of diamond content,from 33% to 29%, is obtained throughhydrogenation of SP films. The GDsample has a small amount of diamondand graphite component. Thegraphitic fraction is minimum for theSP 5%H sample, in agreement with thehighest optical transparency of thissample. The GD sample ischaracterized by a high polymericpercent of 82%, while SP sampleshave only 10% at more than 5% H.The large voids fraction on sputtering sl,lmples is in agreement with thesmall value of microhardness (50 Kgf/mm; and high value of localized
Tab.1 BNA fit results on sputtering (SP - the 8 percent indicates the8 content of the deposition atmosphere) and glow discharge samples. Thedensity is computed on the base of component fractions.
S....ple Diaaond Graphite Polymer Voids Density
(lIO) (lIO) (lIO) (lIO) (g/c..3)
SP OllO8 33.7 17. I 0.0 49.2 1. 56SP 5l108 32.4 9.6 11.3 46.7 1.46SP IOl108 29.2 II. 7 11.4 47.7 1.39GD 4.6 3.3 82.0 10.1 0.99
290
states near the Fermi levelphotoconductivity measurements ona-C layers 6/,/7/.
4. Conclusions
resulting from conductivity andhigh pressure magnetron sputtered
The analysis of a-C:H films deposited by the two different methodsdenotes a strong dependence of the structure of the films on thedeposition conditions. Coherent description of diamond, graphite, polymerand voids content was obtained by modeling of ellipsometric and opticaltransmission behaviors of the layers.
For sputtered layers, the maximum optical gap around 5%H withindeposition atmosphere is correlated with a small graphitic content. Thediamond fraction decreases slowly by hydrogen addition. Voids fractionhigher than 45% values are obtained for high optical gap sputteringsamples. High polymeric content of 82% is obtained for low temperature aC:H layers.
Acknowledgements
The authors wish to thank C. Popescu for helpful discussions. Themeasurement system donated by the Alexander von Humboldt Foundationis gratefully acknowledged.
References
I. Mort,J. and Jansen,F.(1986)Plasms Deposited Thin Films, C.R.C.Press, Boca Raton, F.L.2. Smith,F.W.(1983) Optical constants of a hydrogenated amorphous carbon
film,J.Appl.Phys., 55, 764-771.3. Dasgupta D., Demichelis F., Piri C.F. and Tagliaferro A. (1991) U-bands and gap states
from optical absortion and electron-spin-resonance studies on amorphous carbon andamorphous hydrogenated carbon films, Phys.Rev B 43, 2131-2135.
4. Malshe A.M. ,Kanetkar S.M. ,Ogale S.B. and Kshirsagar S.T. (1990) Pulsed laser depositionof diamond like hydrogenated amorphous carbon films, J.App.Phys. 69, 5648-5651.
5. Robertson J. and O'Reilly E.P. (1987) Electronic and atomic structure of amorphouphouscarbon, Phys.Rev.B 35, 2946-2957.
6. Clarke G.A and Parsons R.R.(1993) Characterization of magnetron-sputtered diamond-likethin films for optical coatings in IR, Thin Solid Films 236, 67-71.
7. Morosanu C., Stoica T" De Martino C., Demichelis F. and Tagliaferro A. (1994) High gapsputtered DLC layers, Diamond and Related Materials, 3, 814-816.
8. Grigorovici R., Stoica T. and Vancu A. (1982) Evaluation of the optical constants andthicknesses of weakly absorbing non-uniform thin films, Thin Solid Films 97, 173-185.
9. Gartner M., Parlog C. and Osiceanu P. (1993) Spectroellipsometric characterization oflanthanides-doped Ti0 2 films obtained via sol-gel, Thin Solid Films 234, 561-565.
10.Azzam R.M.A. and Bashara N.M. (1987) Ellipsometry and Polarized Light, North-Holland,Amsterdam.
II.Kaplan S., Jansen F. and Machonkin M. (1985) Characterization of amorphous carbonhydrogen films by solid-state nuclear magnetic resonance, Appl. Phys. Lett. 47, 7.
APPLICATION OF AMORPHOUS HYDROGENATED CARBON COATINGTO SEMICONDUCTOR RADIATION DETECTORS.
I.M.Kotina, T.AAntonova, G.VPatsekina, VD.Saveliev,L.M.Tuhkonen.PNPI, Gatchina, Leningrad district, 188350, RussiaO.I.Konkov, E.I.Terukov.AF.Ioffe PTI, S.Petersburg, Russia.
1. Abstract.
The real surfaces of the high-resistivity p-type silicon and surfaces coated by a-C:Hfilms have been investigated. It has been found that the natural oxide behaves as tunnelisolator stored charge near the surface. Just owing to this the surface of p-type Si canbecome a strong n-type semiconductor at the liquid nitrogen temperature. Theresistivity of this n-channel depends on the cool condition.
It has been shown that the a-C:H coating can be to produce stable semiconductorradiation detectors.
2. Introduction.
The protection of the surface of the semiconductor radiation detectors is a problemwhich differs in many respects from the passivation problem in the silicon transistor orintegrated circuit technology. The Si02 coating used for latter cannot be applied toradiation detectors for several reasons. The main one is that the high-purity (HP)crystal material must not be heated to the temperatures required for silicon dioxideformation to avoid contamination.
The effect of the surface condition on the characteristics of semiconductordetectors has been the subject of many investigations [1-3]. The conclusion has beendone that in p-type Si and Ge the conventional etching process by HF-HN03 mixedsolution generates the positive charge at the surface states. This surface charge iscompensated for by an equal amount of charge of opposite sign within the bulk of thesemiconductor thus causing the energy bands to band downwards as shown in Fig.l.Thanks to this there is the electron concentration near the silicon surface. It is definedthe following
(1)
291
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 291-296© 1995 Kluwer Academic Publishers.
292
Oxidelayer
------ -------------------.-.----- E,
where ns - surface electronconcentration;nb • bulk electron concentration;Fs - surface potential.
In the case when high resistivityp-type silicon is used for radiationdetector fabrication the surface electronconcentration can exceed theconcentration of majority carriers.Then the semiconductor shows an n-
Fig. 1 Energy bond of p.type silicon. type conducting surface channel(inversion layer) which results in the
increasing reverse currents. Thus the surface must to generate a small positive surfacecharge of surface channels are to be avoided. To be truly satisfactory the surface coatingmust be adjustable to produce that band condition or to stabilize the position of theFermi level at the surface.
The effect of evaporated dielectrics (Si02, SiO, Ti02, Ah03) 'on real surfaces hasbeen investigated in the work [4] using metal-insulator-semiconductor transistorarrangement from high purity p-type Ge. It has been found that some dielectricsintroduce a large number of surface states near the center of the band gap. Allmeasurements were taken at 80°K. However the author did not take into account theeffect of temperature on the electron concentration in the inversion layer. We supposethat the surface state concentration was wrong determined on this particular reason.
3. Experimental details.
The studying of the effect of evaporated a-Si:H films on real Si surfaces carried out at77~. Configurations of the structures under investigation are shown in Fig. 2.
Pd or Au eledrode~
1:Z;;:=:1~ AI electrode
Au electrode a-C:H
----::" r-1:;;:;:::1~ AI electrode
Fig. 2 Structure configuration.
P-type silicon wafers with p= 40kn*cm were etched with mixed solution ofHF:HN03(1:3) and quenched with deionized water. A-C:H layer was deposited by theRF glow discharge decomposition process from a 10% methane-argon gas mixture onone of chemically etched surfaces. Front contact was formed by Au or Pd thermalevaporation. Back contact was formed by allowing with AI at the temperature a 350°C.Such contact has not oxide layer between AI and Si.
293
C-V (capacity-voltage) characteristics at frequency 20 kHz and I-V(current-volt-) h .. d f' I fi th 239.-. 238Pu 233Uage c aractenstics an energy spectra 0 a-partIc es rom e source .t"u- -
were measured.
4. Results and discussion.
-2
,-...".oXW
IPo-.-8
wIU
o 2 4 6 8Volt<l~e (volts:)
10 12
(2)
Fig. 3 c-2-v characteristics for Pd-Si(p)-structure;I - without application forward bias;2 - with preliminary forward bias.
In Fig.3 (curve 1), C-2 measurement of Pd-P-Si contact as a function of reversebias is shown. From the intercept voltage ({Jb is given by [5].
k·T({Jb =~nl + E f +-- - ~ If/
qwhere Ef is the Fermi level measured from the top of the valence band and ~'I'
image force lowering Ef ( Psi = 40k.O*cm)=O.leV at the liquid nitrogen (LN)temperature, ~If/ can be neglected since impurity concentration is very low. In this casewe receive that ({Jb= 0.6 eY.
A simplified dependence of the barrier high <J>b against <Pm for contact metalsemiconductor at the presence of an interfacial layer and surface states in p-Si is givenby [5]:
({Jb =y·(Eg+X-({Jm)+(l-Y)·({Jo (3)
where Eg is the band gap of Si, x- the electron affinity of Si, rpm - metal workfunction and rpo - the neutral level for surface states measured from the bottom of theconduction band. The constant r -is defined by
c:.Y - ' (3a)- C+q2· 0 ·D
, s
294
where 0 and 4' - are the thickness and absolute permittivity of interfacial layer,D s - the density of the surface states per unit area and energy. Using the values,lsi= 4.0S eV, 'Pm (Pd) =S.2 eV, 'Po = 1/3Eg [S] and y= 0.13 [3] we receive that 'Ph mustbe equal 0.31 eV
Such discrepancy between the theoretical and experimental data is obviouslyrelated to the presence of thick interfacial layer between the metal and thesemiconductor [6]. It has been found that this natural oxide layer behaves as thermaloxide and stored the electron charge near the metal-semiconductor interface. Indeed, inaccording to (1) the equilibrium electron concentration near the surface at 770K isequal
(4)
Using the theoretical value of 'Ph = 0.31eV, we receive ns =7.S-105.However experiment involving the application offorward bias (curve 2) showed
that the surface electron charge was enough large. The experiment was carried out inthe following way. The sample was cooled from room temperature down to 77"1<. bydropping into liquid nitrogen. Then C-V characteristics in reverse bias were measured(curve 1). After that forward voltage was applied and then C-V characteristic in reversebias was measured again (curve 2). As seen from comparison curve 1 and 2 there is theincrease of the capacity following forward bias. This effect was not observed on thesample has been heated at the 80°C. Therefore it was not associated with the repopulating of the impurity centers in the bulk of semiconductor. We believe the above experiment shows that oxide layer is responsible for high electron concentration. To be trulysatisfactory native oxide layer behave as tunnel dielectric similar the thermal oxide.
V-I characteristics of Au-Si(p) structures without forward bias (curve 1) andafter forward bias (curve 2) is shown in Fig 4.
90 I,(A'! 0°)
70
50
30
-50 -JO
2
-1~
-25
0.2
Fig. 4 Current - Voltage characteristics ofAu - Si(p) -structures:I - without application forward bias;2 - with preliminary forward bias.
295
It is seen the electron charge near the surface causes the increasing of leakagecurrent.
2
7
-....0
x 5tv
Il:E..
-Btv
IU 3
-2 o 2 468Volt6~e (volts)
10 12
Fig. 5 C-2_V characteristics for Au-a-C:H-Si(p)-structure;1 - without application forward bias;2 - with preliminary forward bias.
Fig. 5 represents C 2 - V characteristics of the Au-a-C:H-Si(p) structures,obtained without application of the forward bias (curve I) and after application offorward bias (curve 2). From a comparison of the C-V characteristics in figures 3 and 5it is obvious the a-C:H coating does not change the surface charge. It should be pointedout the impurity concentration reduced from the gradient ofC2closely agrees with thatof the c-Si in the case of Au-Si(p) and Au-a-C:H-Si(p) structures. This indicates thedepletion layer spreads in the c-Si for both structures.
Energy spectra of a. - particles from the source with 239pu(5.15 MeV), 238pU(5.5MeV) and 233U (4.8MeV) measured by Pd-Si(p) structure and Au-a-C:H-Si(p)structure at the V=O are illustrated in Fig. 6 and Fig. 7.M m
$13 1025 1~7 2:149 2561 JO'73 3$85 4097
Fig. 6 a - spectra e39pu +238PU+233U) measured byPd - Si(p) structure
Fig. 7 a - spectra (239Pu +238pU+233U) measured byAu - a - C:H - Si(p) structure
296
This spectra was obtained after application forward bias, i.e., without theelectron charge near surface. It can be seen that charge collection in the case of PdSi(P) structure is more worse. So the energy resolution of 5.5 MeV peak is about 138keY for the Au-a-C:H-Si(p) structure and about 360 keY for the Pd-a-Si(p). We believethat the surface effects are responsible for this effect.
In addition repeated cyclings from room to cryogenic temperatures have notproduced any change in the characteristics for the Au-a-C:H-Si(p) structures and havedone the characteristics for Pd-Si(p) structure worse.
5. Conclusion.
Our results have demonstrated that a-C:H films can be successfully employed to stableSi(p) surface. But additional work is required to understand the effect of a-C:H films onthe electronic properties of the Si surface.
6. References.
I. J.L1acer Surface effects in silicon radiation detectors IEEE Tr. Ns-II, 221 (1964).
2. Dinger R.J. Dead layers at the surface ofp-i-n detectors IEEE Tr. Ns-20, 135 (1975).
3. Takami Y. and Shiraishi F. Surface chemical treatment effects in ultra-high purity p-type Si detectors IEEE Tr.
Ns-30, 376 (1983).
4. Dinger R.J. Effect of evaporated dielectric materials on the surface of high purity germanium,
J.Electrochem.Soc.:Solid State science and technology, September, 1398 (1976).
5. A1vilm, Goadman Metal- semiconductor barrier high measurement by the differential capacitance method,
J.Appl.Phys.34, 329 (1963).
6. Gowley A.M. Depletion capacitance and diffusion potential of Gallium Phosphide schottky-barrier diodes,
J.AppI.Phys. 37, 3024 (1966).
7. Acknowledgments.
We would like to thank Dr. A.Kh.Khusainov for encouraging this work sincerely.Thanks are also to Pavlova N.J. for help in the preparation of samples and VolkhonskyaE.V. and Patsekin v.P. for their editorial assistance.
This work was supported in part by US Department ofDefence.
DEVICE FOR GROWING AND DOPING IN THE GROWTH
PROCESS OF THIN AIN FILMS.
* **AF.BELYANIN, AP.SEMENOV ,B.V.SPITSYNCentral Research Technological Institute.
Russia, 121355, Moscow, Ivan Franko str., 4.
*Buryat Institute ofNatural Scienses ofthe Siberian Division
ofRussian Academy ofSciences.
Russia, 670042, Ulan-Ude, Sakhjanova str., 6.
**Institute ofPhysical Chemistry ofRussian
Academy ofSciences.
Russia, 117915, Moscow, Leninskiy pr., 31.
Among the known methods of growing thin AlN fIlms by ionsputtering a defmite place belongs to growth regimes realized by planarmagnetrons [I J and gas discharge ion sources [2J. The possibility tocombine growth processes and many new technological methods ofgrowing fIlms open the way to improve the equipment on the base ofmatched action of magnetron sputtering and sputtering by ion beam,which enables to activate the growth surface and the growing AlN fIlm bya beam of slow electrons in a unified deposition device. Such anapproach was considered for the first time in [3], being developed furtherin [4-6J. In the present paper attention is payed to a new constructionof sputtering device and to more complete utilization of the advantagesacquired by combining planar magnetron with ion sources.
The scheme of the set is presented in the figure. Magnetronconsists of a ring target I, a ring permanent magnet 2, a pole piece 3,
297
M.A. Prelas et at. (eds.), Wide Band Gap Electronic Materials, 297-303© 1995 Kluwer Academic Publishers.
298
7
I Gas
electrically connected central anode 4 and screen 5, the earthed screenembracing coaxially the target 1. The magnet 2 and the pole piece 3 forma magnetic circuit which closes through the target, maintaining at itssurface completed radial magnetic field with the magnitude of inductionof 2,7.10-2 T. The planar magnetron is made with a central anodePI, essentially new properties of magnetron being obtained bycombination of electrodes with "open" anode [5]. The central anode 4has a conic hole 34 mm in diameter at the base and 4 mm at the top.Anode with the hole gets a~ditional functions: it passes the flow ofaccelerated ions which fall on the growth surface of a ftlm' and, at acorresponding geometrical form, extracts the ions from cathodeplasma of the ion source, for which sake it is placed in line with emissionchannel 6 inside the emitter cathode of g-as discharge source 7 [81.Parallel to the target I at a distance of 50-70 mm there are mounted aradiation heated plate 8 with fixed substrates 9 and a shutter 10. At theperiphery of magnetron an additional target 11 is hinged which lightens
299
to adjust the slope of target with regard to the sputtering ion beam 12.The ion source 13 [8] is connected with the frame 14 ofvacuum chamberby means of tube bellows 15. The bellows maintains a swinging motion ofthe ion source, which permits the ion beam to be transferred for a shortwhile from the additional target to the circular one without breaking thevacuum. Transversal dimensions of ion beams are regulated by focusingof plasma. A narrow slightly diverging ion beam 12 sputters the
additional target 11 (withdrawal voltage 5-10 kV, discharge currents 0,050,1 A), strongly diverging beam 16 affects only the substrates 9(withdrawal voltage 0,3-1 kV, discharge currents 0,2-0,5 A). Electricpower of the magnetron is applied to the earthed anode (positivepotential) and to the isolated target (negative potential). Plasma
generating gas leaks separately in ion sources and in the magnetron,through a hole in the side wall [9].
The anomalous glow discharge is initiated in crossedelectrical and magnetic fields. Magnetic field keeps the plasma ofdischarge close to the target 1, thus promoting the rise of plasma densityand of the ion current. The discharge is glowing stationarily in thegas pressure range 2·1Q-C 6 Pa [9]. However, at these pressures there isno guaranteed vacuum regime (l<A) of the magnetron operation (l is thedistance between the target and substrates). The mean free path A ofparticles leaving the cathode is noticeably lesser than the minimum
length (lmin) permitted for the transfer intervals of the particles sputteringfrom the target to the growth surface of substrates. In this conditionsthe difficulties arise due to the uncontrolled mutual influence of thedischarge plasma and the growing film. The thickness nonuniformity ischaracteristic of the fIlms grown. Besides, the transfer of sputteredparticles hinders the growth of fIlms with perfect structure and theattainment of recquired accuracy of reproduction of growth regimes. Thecondition lrnin<l<A fulfils at the pressures p<2·10-1 Pa. At low pressures astrong dependency of ignition voltage on the pressure is observed [9].The discharge is initiated at voltages >I kV, and the situation whenignition voltage exceeds the voltage of the power source is not rare. One
300
can lower the ignition voltage and expand the range of working
pressures in low-pressure region by means of external factors evoking
appearance of electrons, which ionize gas in a break-down gap. One of
such factors is ion-electron emission excited by the ion beam. The ion
beam is a rational means of lowering the ignition voltage of the low
pressure discharge, which allows to realize the process of physical
sputtering of additional target accompanying the electron emission.
The flow of knocked out particles is directed to fall on the growth
surface of substrates, thus leading to a regulated introduction of
admixture into the thin films growing by magnetron sputtering. The
conditions of the ignition of a low-pressure discharge (_8.10-2 Pa) the
in planar magnetron by means of the ion beam were considered for the
first time in [4], and the principles of growing thin fIlms by ion beam
sputtering were analyzed in [10, 11 ]. The break-down potential at
which the anomalous discharge lighted up, was determined by the
energy and the current of the ion beam. A threshold-like character of
the experimental dependency of the ignition voltage on the current of
the ion beam is an evidence that there exists a lower limiting value of the
ignition voltage (in the experiments -0,42 kY). The low-pressure
discharge could not be lighted by the usual method (li=O), though the
voltage of magnetron electrodes was rised gradually to comparatively
high values 1,5 kV (Ii is the current of the ion beam). Both regimes, the
ignition of magnetron discharge and the sputtering of the additional
target, are realized by the ion beam 12 due to the swinging motion of
the ion source 13, when the ion beam, after the ignition of the magnetron
discharge, returns to the additional target and fulftls the main function
connected with the driving out the particles, whose flow falls then on the
growing surface of the substrates. During the ignition of the dischargeand the bringing of the magnetrone to the stationary regime of the
sputtering of the ring target, the substrates 9 are screened by the shutter
10.Along with the widened possibilities of controlling the
elemental and chemical composition of growing fIlms, the set gives an
301
opportunity to fulfIl pre-growing operations, connected with preparationof the substrate surface. It is necessary for the ion source 7 to generate thediverging beam of sputtering ions (1-2 keY), cleaning the growth surface.By lowering the energy of ions falling on the substrates to 0, 1-0,5 keV andby regulating the ion current density one can achieve the surfacesputtering regime of the ion activation of the films growing by magnetronsputtering or by ion beam sputtering. The delicate regulating of sharecorrelation of the admixture introduced by the sputtering of theadditional target by fast ions and the activation of the growing ftlmby relatively slow ions make it possible to: rise the adhesion of the films;lower the temperature of epitaxial growth; create the internalcompressing or stretching tensions; obtain high-oriented films withsmooth surface; affect in a needed direction the inner structure and phasecomposition of the ftlms [10-12]. The recquired homogeneousdistribution of added dope and the thickness uniformity of thegrowing films are achieved rather reliably by the rotation of the heatedholder plate 8 with the removably mounted substrates, the rotationspeed being 2-3 S-1 (the plate rotates by electric motor 17).
The typical for the considered device conditions of growing ofthin piezoelectric AlN ftlms, the structure, properties and generalizedelectrophysical characteristics of acoustoelectronic converters (delaylines) are brought together in the table.
TABLEParameter
Pressure, PaWorking gas
Sputtering power, WGlowing voltage, VIon beams current, rnA
Value
o8.10-1,Ar, N2 (N/Ar=3/2)
900350-4001-50
302
Ions energy, keVAngle of incidence, degr.Growth rate, f.lm/h
Target substrate distance, mmSubstrates
Temperature of substrates, KRotation speed, s -1
Thickness of films, f.lmGrowth time, hRougness, f.lm
Adhesion, kg/mm2
Lattice constant "c", nmSize of regions of coherent
scattering, nm
Crystallization degree, %Inclination angle of<000I>
texture axis, gradGrain disorientation, grad
TargetDoping impurity
Working frequency, MHzInsertion loss, dB
Electromechanical couplingcoefficientDelay time, f.lS
Mechanical tensionsDoping dose, at.%
0,1-100-451-450-70
fused quartz; a-~03(0001,01T2)573-7232-31-40,4-2>0,03>5
0,49826-0,49933
40-6060
0,2-11-2
Al(A-99)
SmF3, E~03' EuF3, TbF3, ErF341,6; 149,9; 504,8
34-37
0,1-0,120,465; 0,9861,5-4,9up to 5
303
References.1. Be1yanin AF., Bu1yonkov N.A, Bogomo1ov AB., Ba1akirev Y.G.
(1990) Technika sredstv svyazi. Ser.TPO. Vyp.3. pp.4-24.2. Semenov AP., Be1yanin AF., Haltanova Y.M., Ter-Markaryan
AA (1988) Technika sredstv svyazi. Ser.TPO. Vyp.1. pp.25-31.3. Be1yanin AF., Pashchenko P.V., Semenov AP., Besogonov Y.Y.,
Soldatenkov AV. (1990) Technika sredstv svyazi. Ser.TPO. Vyp.5. pp.42
49.4. Semenov AP. (199 3) Izvestiya SORAN. Sibirskij fIziko
tehnitcheskij zhumal. N6. pp.68-72.5. Semenov AP., Batuyev B.-Sh.Tch.A s.1832134.MKI S
23S1435.6. Semenov AP. (1992) Tez.dokl. 3 Mezhregionalnogo
soveshchaniya "Tonkiye p1yonki v e1ektronike". Yoshkar-01a. Tch.1.pp.34-38.7. Danilin B.S., Syrtchin V.K. (1982) Magnetronniye
raspylitelniye sistemy. M.: Radio i svyaz.8. Semenov AP. (1993) Pribory i tehnika eksperimenta. N5. pp.128
133.9. Semenov AP., Batuyev B.-Sh.Tch.A (1991) Pribory i tehnika
eksperimenta. N5. pp.192-195.10. Semenov AP. (1990) Pribory i tehnika eksperimenta. N4. pp.26
42.11. Semenov AP. (1993) Pribory i tehnika eksperimenta. N2. pp.ll
27.
12. P1eshivtzev N.N., Semashko N.N. (1989) Itogi nauki i tehniki.Ser.Fizitcheskiye osnovy 1azemoy i putchkovoy tehno1ogii. T.5.Ionno-putchkovaya tehno1ogiya. pp.55-112.
PECULIARITIES OF CHEMICAL VAPOR HETEROEPITAXY OF WIDEBAND GAP ill-V NITRIDES.
E.B.SOKOLOY. G.A.NAIDA, N.Y.BAROYSKIISpecial Materials ofMicroelectronic Department, Moscow State InstituteofElectronic Engineering, 103498, Moscow, Russia.
Abstract
Epitaxial aluminium nitride and gallium nitride films on sapphire substrate (T012) wereprepared by pyrolysis of complex compounds of ammonia and halide of AI or Ga. Thequality of heteroepitaxial structures (11'26) AIN / (TO12) Al20 3 depends not only on theconditions of nitride deposition but on disorientation and atmosphere and heat regimesof the substrate as well. Nitride deposition by pyrolysis requires that the sapphire substrate (T012) be heat treated in the reduction atmosphere and disorientation of the<'2110>.
1. Introduction.
The study of materials having a perspective future as microelectronics materials is ofvital importance. In this respect semiconductor wide gap materials and the heteroepitaxial structures are in the focus of attention. Among the materials nitrides of AI and Gahave great potential capacities. Presently heterostructures of AlN and GaN have foundthe niche in acousto- and opto-electronic devices.
Comparison of properties of nitrides reveals both common physical and chemical regularities and suprising differences, besides in many ways they supplement each other.
By now definite approaches to the problem of nitride heterostructure synthesis haveevolved resulting in cubic GaN. multilayer nitride heterostructure compositions andtheir solid solution /1/. Chemical vapor deposition exhibits the most complete array ofprocess peculiarities characteristic of nitride heteroepitaxy namely structural, orientational and morphological inhomogeneity of heterostructures, the properties being determined by the cumulative technological conditions of deposition.
It was found that nitride heteroepitaxiallayers on a sapphire substrate can contain crystallites of azimuth orientation (2,3/. this leading under certain conditions to twinning. It
305
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 305-311© 1995 Kluwer Academic Publishers.
306
is evident that such inhomogeneities in AlN and GaN heterostructures restrict thedevices working conditions, decrease functional capabilities of the materials. Now thereis a capacious experimental evidence of the effect of the substrate structure on the AlNand GaN film growth and orientation.
In the present study, we have examined the mechanism of twinning, the influence ofcrystallographic disorientation and heat treatment of sapphire substrates on the structural and morphological properties of nitride heterostructures.
2. Preparation of Helerostructures.
AIN and GaN films were prepared by pyrolysis of complex compounds of ammonia andhalide of AI or Ga (AICI3·NH3 and GaC13·NH3). The process is characterized by highoutput, the possibility to obtain high oversaturation in deposition zone and high growthrates. Of all the drawbacks it is worth mentioning a substantial spontaneous crystallization upon the inner structures and walls of the reactor. The technique has not found anypractical application though its main attraction is the simplicity of the setup and sufficient technologiCal security.
The deposition carried out in an installation with a horizontally fixed quartz reactor inthe flow of inert gas - argon with the dew point of (55-65°C). A three-zone resistiveheating oven was used as a heater. In the deposition zone, the temperature varied from1050 K to initial complex compounds aluminium chloride-ammonia or gallium-chlorideammonia were subjected to additional reduction and rectification.
The sapphire substrates (T012) had disorientation along definite crystallographic directions. Primarily they were annealed at high temperature in vacuum. Right before thedeposition, vapor etching was used for prior-to-epitaxy substrate treatment.
The surface structure of nitride films, sapphire substrates and the nitride-sapphire interface area was analyzed by reflective electronography. A 3MP-IOO electronograph wasused with accelerating voltages of 25,50 and 100 kV. More accurate data on the surfacearea were obtained by X-ray diffraction technique in the sliding Bragg-Laue geometry /4/. The basic method of assessing the quality heteroepitaxial nitride layers was a rocking curve diffraction technique. The measurements were carried out on a two-crystalDRON-I diffractometer.Crystallographic disorientation value in heterostructures wasperformed on X-ray diffractograms /5/.
3. Results and Discussion.
The effect of the (TOI2) sapphire substrate dislocation and the layer growth rate on thestructure and morphology of nitrides have been studied. The epitaxial layers of nitrides
307
(11'16) were found to contain various azimuth orientation phases. i.e. twins.(11'16)[0001] projection AlN II (T012)['111O]Al20 3 /1/.(11'16)[0001] projection AlN II (TO12)[2ITO]Al20 3 /2/.
This phenomenon is also typical for GaN films (11'16) on a sapphire substrate. Thisstructural uniqueness is sufficiently affected by the disorientation of a sapphire substratefrom (f012). The disorientation vector acan be found by the technique suggested inworks /5,6/. The disorientation vec~or components ax and ay to the directions ['1110] and[Oln] correspondingly were 0.1-2 .
Itwas found that a small disorientation angle deposition results in layers with oppositelyoriented twins of [0001] AlN. The azimuth orientation [0001] pr AlN of the preferentially growing nitride phase is determined by the sign of the component ax in the directions [2ITO] or ['2110] (Figure 1).
The quantitative assessment of the oppositely oriented [0001] pr AIN crystallites wasperformed by the reflection intensity of the X-ray CUKa emission from (0001) AlNplanes at azimuth positions of the samples of cp =900
and 2700
•
In Figure 2 the rocking curves of (11'16) AlN layer from (0001) AlN plane at cp = 900
and 2700
(ax < 0) are shown. The value of the intensity depended on the value of (ax)and the layer growth rates. The results are presented in Figures 3 and 4. As follows fromthese data, the number of twins in the layers rises when the component ax decreases andthe growth rate goes up.
The analogous dependencies can be observed with the growth heteroepitaxiallayers ofGaN on sapphire (T 012) from a vapor phase. It was also revealed that the vector direction of the azimuth orientation [0001] AlN in heteroepitaxial (11''26) AlN layers isaffected by heat treatment of substrates in reduction atmosphere (hydrogen, ammonia,Al vapor) right before the nitride deposition. The heat treatment turns the azimuth orientation [0001] AlN in nitride alternate procedures. The conversion completeness of theazimuth orientation depends on the time, temperature and concentration of reagents. Inall the cases the temperature increases cuts the time of substrate heat treatment necessary for the full turnabout of [0001] AlN thus excluding twinning in the growing layer.It is typical for both the vapor phase and vacuum processes. When ammonia is involvedlow pressure heat treatment is more effective. The electronographic study (T012) Al20 3substrate surface after the heat treatment in ammonia revealed the formation of AlNgrains on the surface. It was also verified by the examination of the surface layer composition by Auger technique. Figure 5 presents nitrogen distribution profiles in the substrate surface layer after the heat treatment in ammonia at 1475 K for 30 min for vaporphase and vacuum processes.
It is evident that the nitridization efficacy is markedly higher in vacuum than that atthere. The higher heat treatment time increases the thickness of the AlN surface layerwith interdiffusion nitrogen and oxygen taking place at the same time. In AlN layers
308
grown on substrates treated by ammonia in vacuum treatment at 1535 K for 20 sec, 60%of crystallites were oppositely oriented and after 1 min treatment - all the 100 %. Theoptimum conditions are those which make nitridization of microsteps {TlOl} on thesurface possible. In small angle disoriented AlN layers grown on ammonia treated substrates, no complete turnabout of the azimuth orientation (0001) AlN is observed. Thelayers contained crystallites of opposite direction projection of (0001) AIN which deteriorates acoustic parameters of the structures due to the sound wave propagation loss.Apparently, in this case one of the main causes of the simultaneous growth of crystals ofopposite direction of projection (0001) AlN is twinning in the surface film due to thenitridization of the substrate with ammonia.
As seen from the data presented in Table 1 during the process of heat treatment hydrogen and Al vapor affect the orientation of [0001] AlN in the similar way.
Table 1. Substrate treaunent dependence of the level A of the azimuth orientation (0001) AIN.
N laxl, heat treatment reagent treatment pressure, A,%degree temperature,K time, min P
0.10 1473 NH3 30 HOS 38
2 0.10 1473 Hz 30 l·lOS 40
3 0.43 1473 NH3 30 l·lOS 99
4 0.43 1473 Hz 30 I·IOS 99
5 0.25 1293 NH3 5 3·IO'z 0.5
6 0.10 1393 NH3 5 3·10'z 12
7 0.10 1453 NH3 30 3·IO'z 45
8 0.05 1453 NH3 15 3·IO'z 12
9 0.30 1553 NH3 3·IO'z 40
10 0.50 1553 NH3 3·\o'z 100
11 0.89 1543 1\'H3 0.3 3·\o'z 60
12 1.8 1473 Alv 0.3 5·\0,3 30
13 0.03 1533 Alv 0.15 5·\0,3 75
14 0.30 1523 Alv 0.15 5·\0,3 98
One can suppose that in reduction atmosphere the surface of the substrate is stripped ofoxygen atoms and the uncovered structure of surface AI atoms then bond with nitrognatoms, Le. the growth goes along the AI-N bond.
The data of Table 1 show that this processes is more effective at temperature higher than1473 K in disoriented substrates. With disorientation to the direction <7110> of primary
309
importance is the surface skeleton structure of AI atoms on planes (lIOl) and (lOTI)confining microsteps.
In inert argon atmosphere or at low temperatures the surface skeleton structure of oxygen atoms is being filled with AI atoms, i.e. the growth goes along the O-Al bond. Insuch samples the azimuth orientation of crystallographic directions of the substrate andlayer is analogous to the one shown in figure 1.
4. Conclusions.
As it is seen from above observations the quality of heteroepitaxial structures (11'16)AlN I (TO12) AI20 3 depends not only on the conditions of nitride deposition but on disorientation and atmosphere and heat regimes of the substrate as well. The disorientationas one of the ways of altering the energy distribution of the surface greatly influencesthe rates of surface processes and consequently the structure and morphology of films.
Nitride deposition by pyrolysis requires that the sapphire substrate (T012) be heattreated in the reduction atmosphere and the disorientation of the substrate <'1110> making microsteps of plans {Ill0) possible.
The analogous twinning is observed in structures (11'10) AIN I (T012) Al20 3 . Indirectlythe availability of twins is proved by the appearance of a double peak on Roentgen rocking curves from the AlN plane (11'10). Though the lack of methods for determiningpolar directions in AlN structure by X ray diffraction cannot draw a simple conclusion.
The AlN structures with twins had excessively big in coming losses (> 60 dB) whenbeing the basis of SAW devices.
References.
I. Davis, R.E(1991) ill·V Nitrides for electronic and optoelectronic applications,Proceedings of the IEEE 79,702-712.
2. Sokolov, E.B., Maljukov, B.A., Kudakov, U.D., Naida, G.A. (1983) Orientation relationships of aluminiumnitride on sapphire, Crystal Research and Technology 8, 53-59.
3. Dryburgh, P.M. (1989) Factors affecting the growth of aluminium nitride layers on sapphire by the reactionsof nitrogen with aluminium monoselenide, J.Crystal Growth 94, 23-33.
4. Afanasjev, A.M., Alexandrov, P.A., Imamov, P.M. (1986) Rentgenovskaya Structrumaja Diagnostica vIssledovanii Pripoverhnostnih Sioev Monokristallov, Nayka, Moskva.
5. BOlnev, A.V., Kuznetsov, A.V., Maljukov, B.A., Sotnikova, 0.S.(1985) Metodikaopredelenija uglov cristallograficheskoj razorientatsii v geteroepitaksialnih stryctyrah,Zavodskaya laboratorija 3, 29-32.
6. KU711etsov, A.V., Maljukov, B.A., Sotnikova, O.S., Chaplygin, G.V.(1989) Iavlenie mnogopozicionnosty vgeteroepitaksialnih structurah GaN/Al20 3.
310
W~(j 'P:2 0<,
.'l(jljf'd~. . !o
...02 ..0 ." 0 1 [2i 'ojfll.!L 3/10
~.,
. .'.,
f--+---+--,+----l---I---1
t-_+-_-+-_-o-lo..L--i_.L--!---!"jd"
'm'.' ., .\ o. • , b lolJ?Li:?3!.,+--+--+--1·1
.,~_--L._---'-_---'
Figure 1. Substrate disorientation dependence of azimuth orientation of (0001) A1N in (1116)/1iOl2jlflt.1J3
so 100
4l lID
:J30 = lID
a ag aa1.1 20
1.1 «II
10 3D
075 75S 76 7b5 71 75S 76 7b5 71
Figure 2. X-ray reflection intensity from (0001) A1N versus azimuth position of the sample.<p=90e. <p=270e.
050.40.3
__TI373K
A103.NHl
~TI300KAlV-tNHl
0.20.1
o.l-_-~--~--~""-~--~
o
0.1 ~90/1270
0.05
lax' degree
Figure 3. Ratio of X-ray reflection intensity from twin oppositely oriented direction [0001) prA1N phases versus substrate disorientation to the direction <2110>.
1a.xI=O·02190/1270
0.11
0.1
0.09
0.08
0.07
0.06
0.000 1 2 3 4 s 6
V.l1m/min
311
Figure 4. AlN growth rate versus X-ray reflection intensity from twin oppositely
oriented (0001) pr AlN phases.
IdN,AE(., ......) I:=~I I.N~E
--- AI (o.~._.)
-+-0
10 IS 20 10 20
.pent' U.r._1S
'0 '"
Figure 5. Auger analysis of sapphire substrate surface area after heat treatment in N H J .
THE PECULIARITIES OF CUBIC BORON NITRIDE FORMAnONMECHANISM USING HEXA-AMMONIACATE BORON HYDRIDE OFMAGNESIUM
H.I. POLUSHINAssistant Prof.High Temperature Materials DepartmentMoscow State Steel and Alloys UniversityLeninsky pr., 4, Moscow, 117936, Russia
K.P. BURDINAResearch ScientistHigh Pressure Chemistry and Physics DepartmentMoscow State UniversityLeninskye Gory, Moscow, Russia
Abstract
In this article pyrolysis of hexa-ammoniacate boron hydride of magnesiumMg(BH4h . 6NH3 (HABH Mg), which is a representative of a whole group ofsimilar compounds, is examined. P,T region of cubic boron nitride (~BN) formationwhile Mg(BH4h • 6NH3 decomposition was determined. The obtained resultsallow to assume that ~BN formation under condition of HABH Mg slow heating athigh pressure, when the system is in the state near to stability, has the followingstages: a) HABH Mg decomposition together with formation of amorphoussuperdisperse BN; b) crystallization (to a certain degree) of the synthesized aBN; c)formation of ~BN nucleus (a certain degree of aBN crystal perfection is evidentlynecessary for this process); d) growth of ~BN crystals in the presence of a gas phase.
Among the methods of cubic boron nitride (~BN) formation those where a gas phaseis formed in the process of synthesis are of the most practical and theoretical interest.A number of authors describe the use of hydrogen- and nitrogen-containing materialsas aBN->~BN transformation catalysts. However, the reported data are rathercontradictory.
313
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 313-319© 1995 Kluwer Academic Publishers.
314
It should be observed that the formation mechanism of J3BN still remains to beelucidated. There are many vague points at the stage of the J3BN nucleation phase aswell as at the stage of the growth of the formed nucleus. The task is complicated bythe existence of a starting graphite-like boron nitride (aBN) of different purity degreesand different degrees of crystal perfection.The point of view taken here is that the most perspective way of J3BN
crystallization is J3BN crystallization directly of atoms synthesized duringdecomposition of chemical compound under appropriate P-T conditions in thepresence of a gas phase, which is an element of a crystal-forming environment. Basedon this position, we have made an attempt to investigate the J3BN formationmechanism using polyammoniacate boron hydrides of metals as starting materials. Thespecified compounds have a common chemical formula:
Me(BH4)n • mNH3'where n is determined by metal valency, m=1+8. These compounds contain boron,hydrogen, nitrogen and metal. Moreover, nitrogen is contained in a great abundanceregarding its content in the BN molecule. It is possible to obtain the specifiedmaterials in a higWy pure state. While heating in an atmosphere without oxygen theyare decomposed with formation of nitride of metal, nitride of boron, ammonia andhydrogen. What is more, hard products of pyrolysis are obtained in amorphous state.A complex study of pyrolysis at high pressures and high temperatures was carried
out for a series of similar compounds. These compounds have the following chemicalformulae:
Li BH4 • 2NH3'Mg(BH4>2 • 6NH3,Zn(BH4>2 • 4NH3'Al(BH4)3 . 4NH3'AI(BH4)3 . 6NH3'Sc(BH4)3 . 6NH3'La(BH4)3 • 6NH3,Zr(BH4)4 • 4NH3'Zr(BH4)4 • 8NH3'Cr(BH4)3 • 6NH3,Co(BH4)3 • 6NH3.
The present article describes the results of experiment using hexa-ammoniacate boronhydride of magnesium (HABH Mg):
Mg(BH4>2 • 6NH3'The experiments were conducted using a .. toroid"-anvil type high-pressure
apparatus. A sample of the examined material was enveloped in a niobium capsule,which was placed in a high-pressure cell. The capsule was isolated from the graphiteheater with a layer of graphite-like boron nitride, which passes on pressure quite well[1] and at the same time is a dielectric. In fig. 1 the high-pressure cell assembly ispresented. The capsule diameter is 0.3 of the heater diameter, the capsule height is
315
0.25 of the heater height. This scheme of the examined material location in the highpressure apparatus was elaborated for the following purposes:1) to isolate HABH Mg from air penetration as well as from container material andheater material penetration;2) to preserve in the capsule the gas atmosphere, which is produced duringdecomposition of examined material.
Figure 1. The scheme of high-pressure cell assembly.
1 - container; 2 - graphite heater; 3 - isolation of aBN; 4 - capsule of Nb; 5 - UABU Mg.
Relative sizes of the reaction cell details were chosen experimentally to providehomogeneity of the temperature and pressure field in the sample. This allows toobtain the same degree of transformation in the whole volume of the capsule and, as aresult, to treat kinetic data correctly.The pressure was calibrated based on the phase transitions of Bi (2.5 and 7.7
GPa), Yb (4.0 GPa) and Ba (5.5 GPa) at room temperature. The temperature wasestimated from the calibrating graph constructed in accordance with the thermocouplereadings. The containers assembly was conducted in a dry cell in an argonatmosphere. The synthesized patterns were examined using X-ray diffraction (XRD)method on CuKa radiation. Quantitative measuring was carried out using weightmethod and also by a calibrating graph construction in accordance with thecorrelation of peaks intensities in diffraction patterns of standard aBN and pBNmixtures. Subsequently the degree of aBN->pBN transformation in quenchedsamples was determined from the calibrating curve. pBN crystals were examinedwith the help of optical scanning electronic microscope JSM-35.Heating of HABH Mg at the pressure less than 4.2 GPa led to the starting
material decomposition according to reaction:3Mg(BH4>2 • 6NH3->Mg3N2+6BN+ 10NH3 + 24H2'
316
A large amount of MgO was also discovered in the reaction products. The cause ofMgO appearance in the products of decomposition is the subject of fuIil)er study.To determine certain properties of the synthesized uBN a series of experiments
was carried out at the consistent pressure (4.0 GPa). Heating up to 673 K resulted inthe complete decomposition of HABH Mg with amorphous hard reaction productsformation. At the temperature of above 873 K the crystal structure regulation processstarted.Table 1 shows the results of determination of the obtained uBN lattice
parameters and c and a axes sizes of its crystallites.
TABLE 1. Crystallographic parameters of the obtained uBN
P, GPa T, K
4.0 873
4.0 1273
4.0 1673
Standard values for uBN
uBN lattice Axes sizes
parameters of crystallites
c, AO a,Ao Lc' AO La,Ao
3.3580 2.507 186 460
3.3467 2.508 349 745
3.3367 2.509 369 937
3.3305 2.504
As it can be seen from the data presented, the lattice parameter c decreases withheating and approaches to the table value. The lattice parameter a increases slightlywith increasing temperature. We have not definitely revealed the cause of thisphenomenon. With increasing temperature rather intensive growing of uBNcrystallites sizes took place.Heating HABH Mg over 973-1173 K at the pressure of 4.2-6.0 GPa led to the
appearance of yellow-orange-co10ured pBN crystals (according to XRD phaseanalysis) in obtained patterns. An experimental P,T region of pBN formation wasdetermined (line 2 in the fig. 2). Line 3 in the fig. 2 was marked in accordance withthe data of [2] and corresponds to the theoretical thermodynamical uBN-PBNequilibrium. The minimum of synthesis pressure is 4.2 GPa at 1173 K. At smallvalues of P,T parameters a certain amount of uBN was observed in the reactionproducts. The amount of uBN decreased with increasing P and T.
317
""6
\\
\
\
-' .. _ - - -1-- - - - • - - - - 1- - -
z
150014001300120011001000
z !r,.---~----r-----.,....------"-:';'...,..---, Sl!I!'
T.K
Figure 2. P,T-diagram of PBN fonnation while pyrolysis ofHABH Mg at high pressure.
We have examined the kinetics of HABH Mg pyrolysis process at high pressure.The experiments time range did not exceed 10 min.The data of synthesis kinetics examination are shown in the fig. 3. It can be seen
that the process rate strongly depends on P and T. There are latent periods of J3BNformation process on kinetic curves 1, 2, 3, 4. The duration of latent periods wasestimated to be from 10 sec at 5.5 GPa and 1273 K to 2.5 min at 4.7 GPa and 1173K. Quenched patterns of the latent period contain aBN of different degrees of crystalperfection depending on temperature. Curve 5 describes the kinetics of J3BN synthesisprocess at the pressure of 6.0 GPa and at the temperature of 1573 K. Under these P,Tconditions the latent period was not detected. In case of heating for T = 1 sec 20% ofboron nitride being synthesized in the process of pyrolysis at high pressure aretransformed to J3BN. The complete transformation takes place in T = 30 sec.
318
- .,""
........
..iI
rr, mLn
:&•••.•••....
......... ············2
1.........
.......................
...~.•
3
.......//
/
.....-...........
.'
/
// .
.I-'~" .. -
./!!
if
IIIIIIII,II
./.'
O..J..'li--.,'--_r_....---,---or---------,----,----,----,.----,.-'--;;......
20
10
Cfoj3BN1100 ••-,'
i 90 : 5I :
I
80 - :I,
I
o 2 3 4 6 7 8 9 10
Figure 3. The kinetics of J3BN synthesis using HABH Mg.
1· P=4.7 GPa, t= 1170 K;
2 - P=4.7 GPa, t= 1270 K;
3· P=4.7 GPa, t= 1370 K;
4 - P=5.5 GPa, t= 1270 K;
5· P=6.0 GPa, t= 1570 K.
All the obtained crystals of J3BN have facets. However, crystals synthesized nearthe aBN-J3BN equilibrium line have a small amount of defects, and crystalssynthesized inside the region of J3BN thennodynamical stability possess a largeamount of defects and are frequently growing as dendrites. Minimum size of thesynthesized crystals came to lOO-150 J.1m.The data of kinetics examination show that in case of short periods of
experiments poor-structured (possibly, amorphous as well) boron nitride issynthesized first, then the degree of aBN crystal perfection increases , and only thenthe J3BN fonnation occurs. It is natural to assume that J3BN fonnation undercondition of HABH Mg slow heating at high pressure, when the system is in thestate near to stability, has the following stages:a) HABH Mg decomposition together with fonnation of amorphous superdisperseBN;b) crystallization (to a certain degree) of the synthesized aBN;c) fonnation of J3BN nucleus (a certain degree of aBN crystal perfection isevidently necessary for this process);d) growth of J3BN crystals in the presence of a gas phase.
319
However, the presented conclusions are of a tentative character. Further examinationof different ammoniacate boron hydrides of metals will make it possible to conduct amore detailed study of J3BN formation mechanism in similar systems.
References
1. Litvin, Y.A. (1978) Comparison of different environments regarded to the ability of the pressure
creating in a a "toroid"-anvil type high-pressure apparatus in a range of 2.0-14.0 GPa, in
Experiment and Techniques ofHigh Gas and Solid Phase Pressures, "Nauka", Moscow, pp. 153
164.
2. Bundy, F.P. and Wentorf, R.H. (1963) Direct transformation of hexagonal boron nitride to
denser forms, Journal of Chemical Physics, 38, 5, 1144-1149.
INVESTIGATION OF CUBIC BORON NITRIDECRYSTALLIZATION PROCESSES IN THE BN-LiJN(H,N) SYSTEM
V.B.SHIPILO, L.M.GAMEZA, AND A.I.LUKOMSKIIInstitute of Solid State and Semiconductor Physics, Academy ofSciences of Belarus, Minsk
In studying the crystallization processes characterized mainly by the rates of theformation and growth of crystals, the thermodynamic approach proves to beinadequate. To gain insight into these processes, kinetic investigations are needed.Earlier, we reported [1] on kinetic dependences of the rates of nucleation andgrowth of the cubic boron nitride nuclei (BNsph) in the BN-LbN system. Thissystem is favored over others [2], since it allows one to grow crystals of higherstructural quality. In recent times, attention is concentrated on the systems inwhich diamond or BNsph crystallization occurs from a melt-fluid saturated with agas phase [3]. Therefore, the studies of the kinetics of BNsph crystallization fromsuch melts are very timely at present.The kinetics of the crystallizalion of BNsph in the BN-LbN-<H, N) system was
investigated at temperatures of 1700, 1790, and 1870 K and pressures of 4.2 and4.4 GPa (the pressure was determined at room temperature). The additives werehydrogen- and nitrogen-containing compounds (H, N), and BN was used asgraphite-like boron nitride (BNgr). It had the following characteristics: latticeparameters a "" 0.2504 nm, c 0= 0.6658 nm; the degree of three-dimensionalordering P3 0= 0.85; graphitization index G = 1.45. The details of the experimentalprocedure are given elsewhere [1].The kinetic dependence of the degree ( a 0= msph 'pgrlmgr'psph, where IDsph,
psph, mgr and pgr are the mass and density of BNsph and BNgr) and rate (C ""6.al6.t) of the conversion BNgr ...BNsph at temperatures of 1700, 1790, and 1870K and pressures of 4.2 and 4.4 GPa are presented in Figs. 1 and 2. From Fig. 1 itis seen that the function a(t) at the indicated temperatures of the crystallizationmelt and pressure of 4.2 GPa has a S-like character: a region of smooth increaseat the start of the process, a region of the most drastic change and a saturationregion. As the temperature rises, the degree of conversion increases but theinduction period reduces from 60 sec (1700 K) to 30 sec (1870 K). For the timeof synthesis equal to 600 sec, a1700K'" 0.22; al790K 0= 0.26 and a1870K'" 0.30.
321
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 321-327© 1995 Kluwer Academic Publishers.
322
C'104
S-l
0.3 15
0.2 fO
0.1 5
180 JOO 420 540 t.sFigure I. Kinetic relationships for the degree (a) and rate (C) or conversion of boron nitride(BN) in the BN-Li3N-(l-I, N) system at temperatures of 1700 (I). 1790 (2), 1870K(3) andpressure of 4.2 GPa.
The kinetic dependences of the conversion rates reveal extremal character withthe maxima at 165 sec (CI700K '" 14'10-4 sec-I), 135 sec (CI790K '" 15.5'10-4
sec-I), and 105 sec (C1870K '" 16.7'10-4 sec-I). Thereafter, the conversion ratesreduce to the values 1.8.10-4 sec-I (t'" 360 sec) and then change insignificantlyto 1.0.10-4 sec- I (t =570 sec).The rise in the pressure up to 4.4 GPa (Fig. 2) brought about a decrease by
about 15 sec in the induction period and a marked increase in the degree ofconversion. Thus, at t = 600 sec, al700K '" 0.24, al790K =0.28, and al870K '" 0.32.In this case, the conversion rates also increased (CI7ooK'" 15.7 '10-4 sec-I, CI790K'" 22.8 '10-4 sec-I, and CI870K = 26.6 '10-4 sec-I) and the time decreased at whichthe above-mentioned maximum values were recorded. For comparison, in the timeinterval t = 360-570 sec the mean conversion rates decreased to (1.3 - 0.3) '10-4
sec-I. Irrespective of pressure at the second stage of crystallization (t > 360 sec)the conversion rates are almost insensitive to temperature.The behavior of the nucleation rates (J '" l1N/(Vgr 'At),where AN is the
number of crystals formed for the time At in a volume Vgr of graphite-like boronnitride) at 4.2 and 4.4 GPa (Figs. 3 and 4) closely resembles the correspondingdependences for conversion rates (Figs. 1 and 2). The maxima of J (Fig. 3) arerecorded at about the same times as those of C (Fig. 1) and with the temperaturerise they also increase in intensity and shift to the side of smaller times. When t> 360 sec, the nucleation rates are practically insensitive to temperature (J '"3.6'109 sec-Icm-J). A rise of pressure (Fig. 4) results in the situation that themaximal values of J, just as of C, are observed at the outset of the synthesis. Then,the J values begin to fall at a markedly higher rate than at P - 4.2 GPa and when
ex:
0,5 J
0,42
0.1
323
C·ro*sot '
25
20
J2 15
110
5
60 180 300 420 540 t.sFigure 2. Kinetic relationships ror the degree (a) and rate (C) or BN conversion in theBN-Ll3N-CH. N) syslem at temperatures of 1700 (I). 1790 (2), 1870 K (3) and pressureof 4.4 GPa.
t ~ 180 sec they become independent of the temperature and time of exposure (J
... 2.2'109 sec·1cm-3).The nucleation rate depends on the work of the formation of critical nuclei (A)
and on the activation energy of the growth of single crystals (E) [4]:
J ... Jo exp(-A/kT) exp(-E/kT) ... Jo exp(-Aerr/kT),
where Acrr = A + E == A (since for the melt A » E) is the effective work of theformation of a crytical nucleus. At the initial stage of the formation of BNsph crystals(Fig. 3) this work turned out to be equal to 1.7 '10-19 J at 4.2 GPa and to 1.4'10-19
J at 4.4 GPa (in the BN-Li3N system Aerr ... 2.0.10-19 J. In other words, thenucleation rate is mainly governed by the supersaturation and viscosity of thecrystallization melt [2].Figures 5 and 6 represent the kinetic dependences of the largyst dimension 1
and linear growth rate v (v'"M/ ~t) of BNsph crystals at the indicated parametersof the synthesis.The increase in the temperature and time of the synthesis facilitates the
formation of largest crystals. During 600 sec the crystals of size 140-160,um growat 4.2 GPa and 1700-1870 K to the size of 300-400,um (Fig. 5), Le. they increaseon the average 2.6 times. The BNsph nuclei grow at the highest rate in the initialperiod (VI700K'" 8.6.10-7, VI790K'" 13.0.10-7, V1870K'" 19.9'10-7 m'sec-1), then,if the synthesis lasts for 300 sec, the growth rate falls to 0.5-3.5) .10-7 m 'sec-1,and thereafter the crystals grow at a relatively small decrease ofv to 0.0-3.0) '10-7
324
~nN ---:14
~13121110
:nJ24
~232221
60 180 JOO 420 540 t,sFigure 3. Kinetic relationships for the number of crystallization centers (N) and nucleationrate (J) In the BN-U3N-(H, N) system at temperatures of 1700 (1),1790 (2).1870 K (3)and pressure of 4.2 GPa.
tnN16 J15 !p' 2
11413
tnJ26
24
22
60 180 JOO 420 540 t,sFigure 4. Kinetic relationships for the number of crystallization centers (N) and nucleationrate (J) in the BN-Li3N-(H. N) system at temperatures of 1700 (I). 1790 (2). 1870 (3)and pressure of 4.4 GPa.
m·sec- I . It should be noted that when t ::: 300 sec the rates of conversion andnucleation are also not large.As the pressure rises (P == 4.4 GPa), the spontaneously formed nuclei grow at
higher rates (VI7ooK == 14.0.10-7, VI790K == 18.6.10-7, V1870K == 34.0'10-7 m'sec-1)for the first 30-60 sec. As a result, larger crystals are formed (160-200 ,um). Thenthe growth rates fall rapidly and when t ::: 240 sec the rates of growth are thesmallest (v == 0.3 .10-7 m ·sec-I). Lower growth rates than those in Fig. 5 at thesecond stage of the growth of the crystals result in a situation that after a time
325
1: 106 V·107
,m/s
500 J
400 220
JOO 1
200 10
100
60 180 300 '120 540 t,SFigure 5. Kinetic relationships lor the largest size 01 BNsph crystals (~nd growth rate (v)at temperatures 011700 (I), 1790 (2), 1870 K (3) and pressure 01 4.2 GPa.
lapse of 600 sec finer grains (280-370,um) are formed, i.e. the crystals increasedin size 1.8 times on the average.From the similarity of the kinetic curves of nucleation rates and their growth,
as well as the rates of conversion, a conclusion can be drawn that the indicatedrelationships are caused in the main by the change in the supersaturation andviscosity of the crystallization melt at different stages of crystallization.Especially noteworthy is the fact that at 4.2 GPa the growth rates at the initial
stages of the process have maximum values (Fig. 5), whereas the nucleation rates(Fig. 3) are minimal. This indicates that the supersaturation of the mass crystallization of BNsph specified by the parameters p and T is preceded by a periodduring which a limited number of rapidly growing nuclei is formed. At the timescorresponding to the maximum conversion rates (Fig. I), such a critical mass ofBNsph is attained which substantially reduces the supersaturation of the melt andthus increases its viscosity. With the supersaturation and viscosity varyingcontinually, the nucleation and growth rates decrease. Pressure exerts anessential effect on the behavior of kinetic functions C(t), J (t) and v(t) at the initialstages of the crystallization process (Figs. 2, 4 and 6). The increase of the meltsupersaturation, caused by the rise in pressure, promotes the formation andgrowth of critical nuclei leading to a decrease in the induction period and to highervalues of C, J and v. In contrast to the above-considered process at P = 4.2 GPa,in this case the formation and growth of the BNsph nuclei occur simultaneously(Figs. 4 and 6). The rise of the supersaturation rate (Fig. 2) at the first stage ofcrystallization increases to a greater extent the critical mass of BNsph. This leadsto a greater decrease in the melt supersaturation and, consequently, to thereduction of C, J and v. When t ~ 180 sec, the supersaturation of the system
326
t'10~ V'T07,
m m/s600 3D
J 220
1 10
60 180 JOO 420 540 t.sFigure 6. Kinetic relationships for the largest size of BNsph crystals and growth rate (v) attemperatures of 1700 (l), 1790 (2). 1870 K (3) and pressure of 4.4 GPa.
reaches such a level which does not exert any marked effect on the rates ofnucleation and growth of BNsph.
It is well known [2] that the kinetic dependence of the degree of conversionBNgr -+ BNsph during catalytic synthesis can be described by the expression
a(t) = 1- exp (_Ktm),
where K = 4/3 Jr.] (t) ·v3 (t) is the reaction rate constant,] and v are the rates ofnucleation and nuclei growth for BNsph. The kinetic parameter m under ratherequilibrium cOQditions of the crystallization of BNsph in the BN-LbN system isequal to four [I ].Since the a(t) curves have a pronounced S-like shape at P - 4.2 GPa, this
indicates that in the system considered there are also rather equilibriumconditions for crystallization. Therefore the K value was estimated from the slopeof the linear dependences In (I-a) - t4• On the other hand, the temperaturedependence of the reaction rate constant is expressed by the Arrhenius equationK = Ko exp (-U IkT). The slope of the curve In K - I IT gave the activation energyU of the process of BNsph formation at the initial stage of conversion which turnedout to be equal to 49 kJ/mole.The decrease in the activation energy of the process of BNsph formation in the
system considered as compared to the BN-LbN system (60 kJ/mole) can beattributed to the decrease in the work of the formation of critical nuclei (1.7, 1019
] against 2.0' 10-19 J for the BN-Li3N system). Thus, the introduction of easilydecomposable hydrogen- and nitrogen-containing compounds into the BN-Li3Nsystem should lead to the decrease of the surface energy 0 of the BNsph nucleusmelt interface, since the work of the formation of homogeneous nuclei isproportional to 0
3 [5]. More equilibrium conditions of crystallization in the system
327
investigated and, consequently, the higher quality of the material produced areevidenced by the improved strength and highly polished surfaces of the majorityof the powders of single crystals. Thus, for graininess 160/125, the compressionstrength of the powders obtained in the BN-LbN-<H, N) system was 14-16 N,while for BNsph powders of the same graininess, but grown in the BN-LbNsystem, the strength was 8-10 N.Thus, the crystallization of BNsph in the BN-Li3N-<H, N) system at the given
parameters of synthesis (P"" 4.2 GPa, T = 1700-1870 K) occurs under sufficientlyequilibrium conditions characterized by the presence of the induction period andthe kinetic parameter m equal to 4, S-like shape of the kinetic dependence of thedegree of conversion BNgr .... BNsph, extended in time range extrema of nucleationrates and conversion at different temperatures and by a much lower work of theformation of critical nuclei as compared to the BN-Li3N system which is associatedwith the decrease in the surface energy of the BNsph nuclei-melt interface.The present work was financed by the Fundamental Research Fund of the
Republic of Belarus.
References
1. Shipilo, V.B., Gameza, L.M., Semashko, N.V., and Bartnitskaya, T.S.(1989) The processes of nucleation and growth of the crystals of sphaleriteboron nitride in the BN-Li3N system, Zh. Fiz. Khim. 63, 1509-1602.
2. Shipilo, V.B., Gameza, L.M., and Smolarenko, E.M. (1988) The processesof nucleation and growth of the crystals of the sphalerite modification ofboron nitride, Poroshk. Metallurg. No.1, 73-79.
3. Belyavina, N.N., Ignateva, LYu., Markov, V.N., et a1. (1993) Investigationof the real structure of cubic boron nitride single crystals grown in differentmedia, in Effect of High Pressures on Materials, Kiev, pp. 25-28.
4. Galperin, N.L and Nosov, G.A.(1975) Fundamentals of the Melt Crystallization Techniques, Khimiya, Moscow.
5. Fedoseev, D.V., Deryagin, B.V., et a1. (1984). Diamond Crystallization,Science, Moscow.
EPITAXIAL GROWTH OF A1N BY PLASMA SOURCE MOLBCULARBBAM EPITAXY
G.W. AUNER, T.O. LENANE, F. AHMAD, Department ofElectrical andComputerEngineering, R.NAIK,P.K. KUOandZ. WU, Department ofPhysics, Wayne State University, Detroit, MI, 48202
1. AbstractAlN films were grown on Si (III), Si (100), and Al203 (0001), and Al203 (1102)substrates by plasma source molecular beam epitaxy (pSMBE). This depositiontechnique uses a magnetically enhanced hollow cathode, lined with an aluminum targetmaterial. A nitrogen plasma is fonned within the hollow cathode. A low energy flux ofsputtered aluminum and nitrogen (partially dissociated) stream out of the cathode to thesubstrate. The energy of the source ions can be controlled by a bias ring. AlN filmquality is correlated with substrate temperature and bias effects. The substrates werecontinuously rotated during growth and the temperature was varied from 4000C to800OC. The acceleration bias was varied between OV to -20 V dc. Epitaxial growth ofAlN on Si(lll) occurred with growth temperatures above 5OO0C. A buffer layer of AlNgrown at 4000C on AI203 followed by growth at temperatures higher than 5000Cresulted in high quality AlN films with preferred orientation. The crystalline quality andcomposition of the deposited films were characterized by x-ray diffraction, atomicresolution TEM, and atomic force microscope. The thennal conductivity of the filmswere characterized by mirage thennal characterization technique. Optical properties wereanalyzed by spectroscopic ellipsometry.
2. Background
Aluminum nitride is a promising system for semiconductor optoelectronic applications. Ithas a very high chemical and thennal stability, good thermal conductivity, and a directband gap at 6.2 eV and a very fast Rayleigh velocity. As such, AlN has potentialapplications for ultraviolet photonic detectors and surface acoustic wave sensors. It canbe also used as a dielectric coating for e~.hancementof Kerr rotation in optical storagemedia and as an insulator for silicon cwoide based devices. Furthennore, the wurtzitepolytypes of InN, GaN, and AlN fonn a continuous alloy system whose direct band gapsrange from 1.9 eV for InN to 6.2 eV for AlN. Thus, AlN in combination with othernitrides could potentially be fabricated into optical devices which are active at opticalwavelengths ranging from red well into the UV.
329
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 329-334© 1995 Kluwer Academic Publishers.
330
Fonnation of nibide semiconductors for device applications depends on many factors.These factors include: (i) achieving the correct stoichiometry, (ii) inducing the correctenergy to fonn a highly crystalline matrix, (iii) maintaining high purity, and (iv) matchingthe lattice parameter of the semiconductor and substrate. Much effort was expended togrow and characterize GaN, AlN, and InN in the 1960's and 1970's. Those researchersencountered significant difficulties in obtaining high-quality material. More recentresearch has developed higher quality nibides. However, all AlN, GaN and InN had highn-type background carrier concentrations resulting from native defects commonly thoughtto be nittogen vacancies. Furthennore, oxygen contamination was a major problem inInN growth. Several approaches have been employed in an attempt to grow thin nittidesemiconductor films. Currently thin film AlN is commonly prepared by magnetronsputtering, chemical vapor deposition, ion beam sputtering, and ion beam assisteddeposition[l-8]. All of these methods need to operate at elevated temperatures andgenerally don't have epitaxial growth. All of these were successful in producingpolycrystalline nibide films. However, to date these methods have not succeeded insynthesizing electronic grade single crystal films. Here we present a new and uniquetechnique for growing wide bandgap semiconductors such as AlN, InN and cBN. Thistechnique uses a magnetically enhanced hollow cathode source for plasma sourcemolecular beam epitaxy (pSMBE).
3. Experimental DescriptionThe PSMBE deposition system consists of a UHV chamber pumped by a 500 Vs ionpump for a base UHV and a throttled en cryopump for use under dynamic gas flowconditions, and a sample load lock system to maintain vacuum chamber integrity (FigureI). The PSMBE system has a new and innovative MBE deposition source- a hollowcathode plasma deposition source. This deposition technique uses a magneticallyenhanced hollow cathode lined with the target material. An argon plasma is fonnedwithin the hollow cathode. A low energy flux of target atoms stteam out of the cathode tothe substrate. This source allows very wide ranging parameter conttol such as the fluxenergy (from thennal energy to high energy by adding bias) of the depositing species.The flux rate of each material can be independently controlled, allowing precisecomposition control. This deposition technique also allows very high target utilizationefficiency compared to other deposition methods and can be run in the low 10-5 Torrrange. The substrate is rotated during deposition to insure very unifonn thickness. Largesubstrates (up to 3 inch wafers) may be deposited in this manner. Ultra-high purityprocess gases are conttolled by MKS mass flow controllers and additional gaspurification is perfonned by a titanium and copper gettering furnace for argon andnitrogen, respectively. Gas compositions are monitored and controlled by a differentiallypumped UTI mass spectrometer connected for feedback control to a MKS mass flowcontroller. This experimental setup enables precise control of deposition parameters(such as gas flow and partial pressures, substrate temperatures, plasma densities, d. targetand substrate bias, and plasma source induced ion-solid interactions) for synthesis ofalmost any thin film material.
We deposit AlN mms on four kinds of substrates, Si(lll), Si(IOO), sapphire(OOOI) and
331
sample TransferLoad Lock
IL uto-matehlng
r.f. Power SUpply I'--two_rk_---'
c:::!I!::::I---f High TemperawreRotating SUbstrale Holder
Mass Flow Control System
Cathode
ollow Cathode Off-Axis Sputtering Source
Gu Inlet(Nilrogm)
--!!II
MagnelS
Figure 1- IDuslrlllion of the Plasma SourceMoleal1sr Beam EpilaXy (PSMBE) System.
sapphire(lI02). The Si(lll) and Si(lOO) substrates are prepared by fIrst cleaning in anultrasonic bath of acetone, followed by methanol. Subsequent to the ultrasonic cleaning,the silicon substrates are dipped in 10% (by volume) hydrofluoric acid for 30 seconds.The substrates are then pull dried and mounted in a vacuum load lock chamber. The pulldried substrates are hydrogen terminated, which passivates silicon and changes the
332
surface free energy, allowing enhanced epitaxial growth of AIN. The sapphire substratesare similarly ultrasonically cleaned and etched in a 3:1 mixture of sulfuric acid andphosphoric acid, followed by methanol cleaning. The substrates are transferred from theload lock to the main chamber within 10 minutes of the cleaning processes. Thedeposition Parameters are outlined in Table 1.
Table 1. Outline of the deposition Parameters for growth of AIN films onsilicon and sapphire substrates
Pressure (Torr) Nitrogen Argon Flow Temperature Bias (de)Flow(SCCM) (SCCM) (CC)
7 x 10-4 5 • 15 20 400 - 700 OV to ·20V
4. Results and Discussion
The AlN samples varied in morphology depending on the parameters duringdeposition. Three dimensional growth of individual crystals approximately 10 to 100f.lm in diameter on a seed layer of amorphous AlxNI-x mixture were observed forgrowth under high temperature and high bias deposition conditions. Bias conditions offrom OVdc to -IOVdc at temperatures of 5000C - 6500C showed high quality epitaxialgrowth of AIN (0002) columns on Si (III) and sapphire (0001). The XRD patternshows only an AlN (0002) reflection peak for AlN. The absence of other reflectionsindicate complete film texture with AlN[OOO1] II Sir111]. Cross-sectional transmissionelectron micrograph (Figure 2) reveals epitaxy of AlN on Si(III) with orientation
Figure2 - TEM micrograph indicates epilaXial groWlh of AlN on Si(lll) wi1h AIN[2110] II Si[Oli] orientation.
333
relationships of AlN [2110] II Si [011]. Furthennore HREM micrograph of the AlN /Si(lll) interface shows parallel lattice fringes of Si(lll) and AlN(OOOI) extending upto the interface, indicating sharpness and the absence of interfacial diffusion. Althoughit is commonly believed that below 6OO0C polycrystalline growth occurs, ourobservations clearly indicate epitaxial growth in the temperature range of 4()()OC to6000C. Low temperature epitaxial growth of AIN probably occurs due to acombination of the controlled kinetic energy of the depositing species from the PSMBEsource and the change in surface energx of the hydrogen tenninated silicon substrate.In contrast, AlN grown on sapphire (1102) substrates gave a highly textured a-planegrowth. The quality of all of the AlN films improved at low temperature growth(4000C) with a bias of approximately -lOVdc. A summary of the growth mode isillustrated in Figure 3.
Column GrowthTemp = 600-700 °cr.f. Power = 200 WBias = OV - -15V
3-Dim. Growth onamorphous layerTemp = 650 - 800 °cr.f. Power = 200WBias ~-20V
Smooth CrystalTemp=4000Cr.f. Power = 300WBias = -10V
Figure 3- Growth mode of AIN under different power, temperature and bias conditions.
Ellipsometry studies show that the AlN samples have an index of refraction ofapproximately 2.2. The thenna1 conductivity values of AlN range from 0.48 to 25.2 Wm1K-l. These values should be com~edwith the bulk value of AlN at room temperature,which ranges from 1.76 Wm-1K- [9] to 50 Wm-1K-l [10], depending on the specificsample preparation. The theoretical value is 320 Wm-1K-l for pure single crystal AlN.
334
The best value of our deposited film compares quite favorably. taking into account thefact that the thermal conductivity values of thin films are typically orders of magnitudelower than the corresponding bulk materials [11-14].
S. Acknowledgments
This work has been supported by the Institute for Manufacturing Research in Wayne StateUniversity.
6. References
1. Y.G. Roman and A.P.M. Adriaansen. Thin Solid Films. Aluminum nitride filmsmade by low pressure chemical vapour deposition: preparation and properties 169.241(1989).
2. G.A. Cox. D.O. Cummins. K. Kawabe. and T. Tredgold. J. Phys. Chern. Solids. 28.On the preparation. optical properties and electrical behavior of an aluminumnitride. 543(1967).
3. S.Strite and H.Morkoc. J. V8/;. Sci. Technol. B. GaN. AlN. andlnN: A review10. 1237 (1992).
4. H. Itoh. M. Kato and K. Sugiyama. Thin Solid Films, 146, Plasma-enhancedchemical vapor deposition ofAlN coatings on graphite substrates. 255(1987).
5. K.Kubota et aI.• J. Appl. Phys. 66. Preparation and properties ofJII-V nitridethinfilms 2984 (1989).
6. H. Windischmann. J. Appl. Phys. 62, An intrinsic stress scaling law forpolycrystalline thin films prepared by ion beam sputtering 1800(1987)
7. H.T.G. Hentzell. J.M.E. Haper. and JJ. Cuomo. Mater. Res. Symp. Proc. 27,Structure ofAI-Nfilms deposited by a dual ion beam process 519(1984).
8. J. S. Morgan. W.A. Bryden. TJ. Kistenmacher, S.A. Ecelberger. and T.O.Poehler. J. Mater. Res. 5. Single-phase aluminum nitride films by dc-magnetronsputtering 2677(1990).
9. Y.S. Touloukian, R.W. Powell. C.Y. Ho, and P.K. Klemens. Thermal Conductivity:Nonmetallic Solids. IFIIPLENUM, New York-Washington,1970, p.653.
10. G.A. Slack, J. Phy. Chern. Solids 34, Nonmate1lic crystals with high thermalconductivity 321 (1973).
11. A.H. Guenther and McIver, Thin Solid Films, The role of thermal conductivity inpulsed laser damage sensitivity ofoptical thin films 163 (1988) 203, and referencestherein.
12. J.C. Lambropoulos, M.R. Jolly, C.A. Amsden. S.E. Gilman. M.J. Sinicropi, D.Diakomihalis. and S.D. Jacobs, J. Appl. Phys. 66, Thermal Conductivity ofa thindielectric film 4230(1989).
13. C.H. Henager, Jr. and W.T. Pawlewicz. Appl. OpL 32. Thennal conductivities ofthin. sputtered optical films 91(1993).
14. Z.L. Wu. P.K. Kuo, Lanhua Wei, SL. Gu and R.L. Thomas. Thin Solid Films. 236.Photothennal characterization ofoptical thin films 191(1993).
ELECTRONIC STRUCTURE AND RELATED PROPERTIES
OF TETRAHEDRALLY BONDED WIDE-BAND-GAP MATE
RIALS CONTAINING EARLY ELEMENTS OF THE PERIO
DIC TABLE
W. R. L. LAMBRECHT, C. H. LEE, K. KIM, A. G. PETUKHOV,E. A. ALBANESI AND B. SEGALLDepar'tment of PhysicsCase Western Reserve Univer'sityCleveland, OH-44106-7079
1. Introduction
The tetrahedrally bonded materials involving early elements of the periodictable, such as Be, B, C, and N have rather extreme properties comparedto the conventional tetrahedrally bonded semiconductors. These uniqueproperties make them interesting for a variety of applications, includingtheir use as hard coatings for mechanical tools, metal/ceramic composites, heat-sinks, and electronics. The latter are the most demanding type ofapplications because they require semiconductor grade purity single crystalline material in order to realize the superior performance these materialspromise to have. But their extreme properties also makes them difficultmaterials to grow in the form of pure single crystals. Their properties arethus not yet very well known. In order to make a rational choice of materialfor specific applications, it is necessary to understand the interrelationshipsbetween the intrinsic materials properties and the trends in these propertieswith atomic number, crystal structure, and so on. Virtually all materialsproperties can in ultimate instance be related to the underlying electronicstructure. The purpose of this paper is to describe the basic trends in electronic structure of these materials and some of their related properties.The results presented here are based on a systematic study of the electronic structure of these materials carried out over a number of years inour research group. Although many of the results discussed here were pre-
335
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 335-372© 1995 Kluwer Academic Publishers.
336
sented elsewhere they were not included in an overall description of thetrends as presented here. Some new results are included as well. Extensivereferences are provided as a guide to our previous work not covered here indetail.We first describe some of the relationships between intrinsic properties
and their importance for various types of applications. Next, we briefly describe our computational approach. Then, we present calculated results forthe groundstate properties such as the equilibrium lattice constants, bondstrength, elastic constants, as well as the underlying electronic structure in particular the band gaps - for a number of materials systems.We classify the materials as follows:
materials containing only elements from the second row of the periodictable: diamond, BN, BeOrelated Be compounds: BeCN2 and Be2Cpolytypes of SiC;B-compounds: BN, BP, BAsthe group III-Nitrides: BN, AIN, GaN, InN;related II-IV-N2 materials;alloys among III-N's and heterovalent alloys.
2. Relations among intrinsic properties
\-Vide-band-gap semiconductors have a number of advantages for electronicapplications over conventional semiconductors. Among these are largerbreakdown fields, resistance to harsh environments, ability to operate athigh-temperatures, and high-thermal conductivity. These various properties are important for a variety of applications where conventional semiconductors fail. For example, in conventional semiconductors, intrinsic carrier concentration at a few 100°C wipe out the effects of doping and makesemiconductor devices inoperable. Obviously, since intrinsic carrier generation varies exponentially with the gap, wide band-gap materials have animportant advantage in this respect. Intrinsic carrier generation also playsan important role in the time-decay of space charges induced by opticalexcitation of e.g. an MIS or Schottky diode. This may enable new kinds ofsemipermanent memory and UV detetectors.At high temperatures, the formation of various types of defects also
start to degrade the material. For example, dislocation movement proceedsby kink formation and migration. Vacancy and interstitial production alsohave a thermal activation energy. All these processes require bond breakingand thus the stronger the covalent bond the more resistant the materialsare to high-temperature degradation of the crystalline quality. The wideband-gap materials we will consider here derive their wide gaps from strong
337
covalent bonding and not exclusively from high ionicity. They are thusalso strongly bonded. This is also what gives these materials resistance toradiation damage.The above effects are not only important for high-temperature operation
but also for high-power and high-frequency applications. In fact, both ofthese lead to generation of heat. The thermal conductivity is thus anotherimportant performance parameter for a semiconductor. The wide-bandgap tetrahedrally bonded materials have a high phonon-mediated thermalconductivity because of their high sound velocities, which in turn are closelyrelated to the stiffness of the bonds. The latter can best be described interms of bond strengths, elastic constants and the force constants involvedin certain phonon modes.Another important parameter for high-sp~ed electronics is the static
dielectric constant. This is because in a very general sense, electronic devicesare based on the transfer of charges from one spatial area to another andthus involve capacitance. One clearly wants as Iowa dielectric constant aspossible. Also good insulators with low dielectric constant are required forisolating various active parts of a device from each other. Thus, as one startsto use wider gap semiconductors, one also needs even wider gap insulatorsfor packaging applications. The electronic dielectric constant at frequenciesabove vibrational frequencies f oo is roughly inversally proportional to theband gap. In partially ionic materials, the static dielectric constant involvesalso a contribution from the ionic motions. Thus ionicity is an importantparameter. Ionicity can be defined in terms of the ratio of the atomic energylevel difference to the effective gap between bonding and antibonding states.Phillips' ionicity Ii [1) is related to Harrison's [2) polarity ap by Ii = a}with
(1)
Here Ec and Ea are cation and anion atomic energy levels and 13 is thecovalent interaction between sp3 orbitals pointing towards each other. Thequantity Ea = v'IEc - Ea l2 + 4132can be interpreted as an average effectivegap between cation and anion levels. Effective values for these levels canbe extracted from our first-principles calculations [3] and define an ionicityscale. The ionicities of some of the materials discussed here are given inTable 1.The underlying reason for the unique properties of the materials involv
ing early elements of the periodic table can be understood on the basisof the simple molecular bond picture embodied in Eq.(l). Because of theabsence of lower energy p-like orbitals in the core, the 2p levels are anomalously deep compared to 3p levels. Also, 2s levels feel the nucleus morestrongly than 3s electrons because the nucleus is less effectively screenedby the small core consisting only of Is electrons. This relatively compact
338
TABLE 1. Polarities calculated fromLMTO-ASA calculations for equal cation and anion sphere sizes
SiC 0.473
BN 0.475 AlN 0.807
BP 0.018 CaN 0.771
BAs 0.019 InN 0.792
BeO 0.818 BeC" 0.588 BeN" 0.701
" Individual bond polarities in BeCN2.
nature of the 2s and 2p orbitals is responsible for the small lattice constantand strong covalent bonding. The strong covalent interaction (3 leads to alarge effective gap Ea.When atoms of the second row are combined with atoms of the third
row, there is also an anomalous difference in atomic energy levels. Thisleads to a strong ionic component to the bonding for SiC (even though thelatter is a IV-IV compound) and AIN and a surprisingly low ionicity forBP. The latter occurs because the 2p level of B is sufficiently deep thatits position approaches that of the P 3p level. By Eq.(l) that implies lowionicity. In fad, taking into account that the effective charge on the Velement is given by q = 1 - 4ap, one finds that it .actually has positivecharge when ap < 1/4. This is the case for BP and BAs which as a resulthave inverted ion character: P or As behave as cation while B behaves asanion despite the fact that normally the group-III element behaves as cationand the group-V element as anion. On the other hand, the enhancementof ionicity due to the deep 2s and 2p levels for SiC and III-N compoundscontributes significantly to their large gaps.The relative extent of 2p versus 2s orbitals compared to those of 3s
and 3p has another important effect. It implies that there is less advantage in promoting a 2s electron to a 2p orbital to form directed bonds. Infact, strong bonds can also be realized by means of 11" bonding between 2porbitals directly. This means that for elements from the second row of theperiodic table there tends to be a competition between tetrahedral bondingand planar sp2 plus 11" bonding. Thus, C and BN have competing layeredstructures: graphite and h-BN. BeO does not exhibit such structures [4]which is related to the higher ionicity of this II-VI compound. Indeed ionicity tends to favor higher coordination.Ionicity also plays a role in the relative stability of cubic versus hexag
onal stacking of tetrahedrally bonded layers [5]. Thus, the relatively lowionicity compounds BP, BAs and BN and (zero ionicity ) elemental C fa-
339
vor the cubic structure while the III-nitrides have wurtzitic structures andSiC exhibits polytypism. lonicity plays also an important role in electronictransport. The polar mode scattering increases with increasing effectivecharges.Wide band gaps have another important application in optoelectronics
where they extend the range of light emission frequencies into the blue andUV. In this context, it is important whether the gap is direct or indirect.Other important parameters for electronic applications are effective masses,deformation potentials, and other band structure details. Some exampleswill be discussed below.Finally, an important aspect of potential semiconductor materials is the
possibility of doping the material nand p-type. Diamond has proved to bequite difficult to dope n-type. GaN was difficult to dope p-type althougheffective p-type doping with Mg was recently achieved using post-dopingannealing and low energy electron beam irradiation (LEEBI) treatments[6]. AIN has sofar resisted effective doping of either type. BN has beenreported to be both nand p-type dopable [7]. Although this is an importantissue, we will not discuss it further because we have not done studies of thisaspect in our group.In summary, we have here given some examples of how the electronic,
structural and bonding properties are interrelated and how they impactdevice applications. In the remainder of the paper we focus on the electronicstructure.
3. Computational Method
The theoretical framework of our first-principles calculations is densityfunctional theory [8]. This approach basically reduces the many-body problem of interacting electrons to that of independent electrons in an effectivepotential with the same ground state electron density as the real system.The total energy of the system is expressed in terms of the Coulombic energy of the electron charge density and the positively charged nuclei, thekinetic energy of the independent model electron system and a remainderwhich is called the exhange-correlation energy. The method is in principlerestricted to the study of groundstate properties. The local density approximation (LDA) allows one to incorporate effects of exchange and correlationin a mean-field approximation. It assumes the same functional form for theexchange-correlation energy as in the homogeneous electron gas but applies it locally to each volume element of the real inhomogeneous system:Exc = JExc [n(r)]n(r)d3
1'. We use the parameterization of the exchange andcorrelation functionals by Hedin and Lundqvuist [9].An important point to realize about this theory is that the eigenvalues
340
Enk of the effective one-electron Schrodinger equation
(2)
where VH is the Hartree potential combining the electron and nuclear electrostatic potential, and Vxc is the exchange-correlation potential, do nothave the mea,ning of single particle excitations. For a periodic solid theEnk constitute the band structure. In other words the band structures wecalculate in this theory form a useful construct to understand total energyproperties and bonding but do not represent the energies required to extract electrons from the system as measured for example by photoemissionor as involved in optical excitations (excluding excitonic effects). The truequasiparticle excitations are given be the related equation:
with I;xc(r, r', Enk) the non-local and energy dependent exchangecorrelation self-energy operator. In practice, however, the wavefunctionsof the two equations are quite close to each other and the effects of replacing the Vxc by I;xc can be treated by perturbation theory. The self-energyoperator can be approximated by Hedin's GW approximation [10]. Carrying out this calculation in practice is rather cumbersome and has beendone for only a few systems [11, 12, 13, 14]. The main finding, however, isthat the spectrum of Enk and Enk mainly differ by a gap correction whichis roughly k-independent and energy independent. The latter can actuallybe estimated easily by means of an approximate treatment of the abovetheory proposed by Bechstedt and Del Sole [15].The fact that LDA band structures underestimate the gap is well known
as the "gap problem". It is actually not a failure of LDA but a questionof appropriately interpreting the theory. The GW approximation predictsthe correction to be only weakly dependent on crystal structure. Thus LDAband structures Enk can be used to investigate changes in band structurewith crystal structure. The GW calculations actually also predict a smallrigid shift of the bands with respect to the average electrostatic potential.This is irrelevant for most purposes, because one is usually interested onlyin interband differences, but it does have an effect on the band-alignment ata heterojunction. The magnitude of this shift appears to depend somewhaton the actual choice of parametrization of the LDA exchange correlationpotential and to that extent is not yet well established theoretically [12].While for wide bands the only important correction appears to be the abovedescribed gap correction, narrow bands such as core and semi-core levels
341
are subject to additional self-energy corrections. This is e.g. important forGa 3d states in GaN [16].The method we use to calculate the energy bands, i.e. to solve Eq.(2) and
obtain the self-consistent potential, is the linear muffin-tin orbital method[17). Most of the results reported here were obtained in the so-called atomicsphere approximation (ASA) in which Wigner-Seitz spheres replace the actual Wigller-Seitz cells and a spherical approximation is made to the chargedensity and potential inside each sphere. This method is in very good agreement with the so-called full-potential calculations which do not make thisapproximation except for quantities which involve small energy differencesor in cases where significant distortions of the local spherical symmetryoccur. For example, cohesive energies, lattice constants, bulk moduli andband structures are usually given quite accurately by the ASA. Surfaceenergies and workfunctions, however, are not because of the strong nonspherical dipole potential near a surface. In fact, the main problem withthe ASA is in the spherical approximation to the charge density rather thanto the potential. The errors induced by using slightly overlapping sphericalcharge densities can to a good extent be corrected by means of the so-calledEwald correction [17] in which one corrects for the overlap by redistributing the overlap charge into a homogeneous background charge. We haveused this approach successfully in our structural energy minimizations forthe chalcopyrite II-IV-N2 compounds [18). For the calculation of opticalphonon frequencies and elastic constants, which req~ire the calculation ofvery small energy differences in order to stay within the linear responseregime with respect to the strain or phonon distortion perturbation, weneed full-potential calculations. The latter were done using the approach ofMethfessel [19).Our approach for dealing with disorder in alloys and other specifics will
be described along with the results in the appropriate sections.
4. Results
4.1. DIAMOND, BN, AND BEO.
In Table 2, we present calculated lattice constants, bulk moduli, bondstrengths (defined as 1/4 ofthe cohesive energy per zincblende unit cell withrespect to free (spin-polarized) atoms), and band gaps for diamond, BN,and BeO. We include calculations both for the cubic and hexagonal crystal structures: i.e. cubic diamond and lonsdaleite diamond and zincblendeand wurtzite for the other binary compounds. Our results for hexagonaldiamond and BN are are in good agreement with those of Salehpour andSatapathy [21] and Gorczyca and Christensen [22] respectively, but aremore complete. (Not all quantities given here are provided in the above ref-
342
erences.) We note that the groundstate equilibrium properties are similar inthe hexagonal and cubic forms. The present ASA calculations, however, arenot sufficiently accurate to determine the relative stability of the wurtziteand zincblende structures. In addition, special care has to be taken in using equivalent k-point sets and in relaxing the internal degree of freedomto establish this small energy difference. That would require full-potentialcalculations. The structure determination was not the primary purpose ofthe present study. We also note that typical errors for bulk moduli are 10%. The f'V10 %difference in bulk modulus between cubic and hexagonal diamond and c-BN is probably overestimated in part because we did not relaxthe structural degrees of freedom, i.e. the relative position of the two sublattices and the cia ratio in hexagonal structure, which are not determinedby symmetry. The overestimate of the bulk modulus is consistent with theunderestimate of the lattice constant and overestimate of the bond energy.The band gaps of the hexagonal crystals, however are expected to be quiteaccurate.
TABLE 2. Properties of diamond, BN, BeO and BeCN2: lattice constants (ae for cubic, ahand cIa for hexagonal) (A), bulk modulus B (GPa), bond energy Eb (eV), and minimum bandgap E g (eV).
cubic hexagonal
theory expt. theory exptDiamond ae (A) 3.58 3.567 ah (A) 2.46 2.522"
cia 1.633"B (GPa) 434 442 B (GPa) 506Eb (eV) 4.2 3.7 Eb (eV) 4.7E g (eV) 4.4 5.5 f-c. E g 3.6 (4.7)b f-K
BN ae (A) 3.62 3.616 ah (A) 2.51 2.56cIa 1.633"
B (GPa) 381 290-465 B (GPa) 425Eb (eV) 3.8 3.3 Eb (eV) 4.2E g (eV) 4.5 6.4 f-X E g (eV) 5.7 (7.6)b f-K
BeO ae (A) 3.80 3.79 ah (A) 2.64 2.664cia 1.633" 1.623
B (GPa) 224 B (GPa) 248 224-249Eb (eV) 3.4 3.0 Eb (eV) 3.7 3.0E g (eV) 6.9 (9.3 )b f-X Eg (eV) 8.3 ]0.7 f-f
" assuming ideal cIa = .,j873b estimate assuming gap correction to be independent of structure
343
The band gaps depend significantly on the structure. Even the k-pointlocation of the conduction band minimum changes. In Fig. 1 we show howsome of the conduction-band eigenvalues relative to the valence band maximum change across the series. One may note that the s-like r 1 states inboth cubic and hexagonal structures come down in the series while ther6 state of hexagonal and r15 state of cubic move up. Mh and J(h statesof the hexagonal structure also move up toward BeO. One can relate thek-points of the cubic fcc and hexagonal hcp Brillouin zones by assumingthe [lll]c direction to be equivalent to the [OOOl]h direction and the [llO]cdirection to be equivalent to the [1120]h direction [23, 24]. The cubic Xpoint corresponds to the point 2/3 along M-Lh and is different from J(h.We note that in all cases going from cubic to hexagonal the X conduction band minimum moves up in energy and the J( minimum comes down.This is thus a structure related effect. In the case of BeO, however, ther minimum is lower in energy than J(h and thus leads to a direct gapmaterial. We also note that the gap in hexagonal diamond is lower thanin cubic while the inverse is true for BN and BeO. The bond energy andbulk moduli decrease across the series diamond-BN-BeO while the latticeconstant increases. Both hexagonal diamond and wurtzitic BN appear tobe metastable phases. It is not clear whether there is a stability regionin the phase dia.gram or whether their occurence is due only to kinetics.The latter appears to be the case since hexagonal diamond has sofaI' onlybeen formed under shock compression conditions. BeO is only known in thewurtzitic phase. This is consistent with its high ionicity. The cubic phaseis significantly higher in energy (0.14 eV latom according to Chang et al.[25]) and is unlikely to be formed.
Relation to graphitic structures We note that diamond and c-BN havecompeting layered structures graphite and h-BN, which are actually theequilibrium phases at ambient pressures. As shown by Fahy et a1. [28,29] theenergy differences between the layered forms and the tetrahedrally bondedforms are small but there is a large energy barrier which impedes the transition from one phase to the other. In their models, the transition pathfrom rhombohedral graphite to diamond, or h-BN to wurtzite BN corresponds to a gradual decrease of the distance between the layers which thenstart to buckle. Layered phases of BeO have been shown to be unstable byWentzcovitch et a1. [4].Diamond tetrahedral bonding, however, ca·n also be stabilized by sur
face hydrogen. Hydrogenation of aromatic molecules yields buckled tetrahedrally bonded rings. Similarly, graphite sheets start to form buckled ringswhen hydrogenated from the edges inwards. This leads to a curling or bending of the planes [30] and to the formation of a diamond-like structure on
344
20
C BN
~.15
r 1
>~
>- 10C.9a:wzw
5
BeO
......<J r 15..........
....
OL-_~_-J'---~-----'--~--'---~----'
3.5 3.6 3.7 3.8effective cubic lattice constant (A)
Figure 1. Conduction band eigenvalues of diamond, BN and ReO versus effective cubiclattice constant. Full lines and symmetry notation on the left refer to the hexagonalstructure, dotted lines and symmetry notation on the right refer to the cubic structure.
the prism planes of graphite. This is suggested by the open circles in Fig. 2which shows the calculated structure at the prism plane interface betweengraphite and diamond. That relaxed structure was calculated [31J using thesemi-empirical Tersoff potential [32J. It was conjectured by Lambrecht eta.l. [31J to form a critical step in the nucleation of diamond.
Interfaces We have also studied interfaces between some of these materials: diamond/c-BN [20J, diamond/BeO [33J and diamond/Cu [34J. We foundthat the diamond/c-BN offset is of type II, Le. the valence and conductionband of diamond are both above those of c-BN. This also has importancefor alloys, as will be discussed below. The bonding at these interfaces isquite strong because there are no dangling bonds and the valence bands ofthe materials overlap each other in a wide energy range. In contrast, thebonding at Cu/diamond interfaces is much weaker because the Cu3d-bandis narrow and essentially only the Cu4s bands form bonds with the diamonddangling bonds. An important quantity is the adhesion energy W a , which
a
a
(0001)
L[11001
(111)
[1211 ...-1C
B
A
C
B
A
345
FigUl'e 2, Struct.ure of t.he prism plane interface between graphite and diamond. Threelayers of diamond almost perfectly match t.wo layers of graphit.e. Hydrogen atoms (smallcircles) sat.isfy the dangling bonds of t.he "non-bonded" diamond layers. The st.ructureat, the int.erface (open circles) indicates that the curling of buckled rings forms an initialdiamond nucleus.
is given by
(4)
where /'A, /'B a.re the surface energies of the materials brought together atthe interface and /'i is the interface energy. For diamond/Cu, the adhesionis much weaker tha.n for the interfaces of diamond with c-BN and BeO,mainly because the Cu surface energy is so much smaller than that ofthese tetrahedrally bonded materials. Surface energies basically scale withcohesive energies because they involve bond breaking. Table 3 gives thecalculated surface energies and adhesion energies. Details can be found in[20,33,34].
The suitability of c-BN and BeO as substrates for heteroepitaxia.l diamond growth was confirmed by several experimental studies [3.5, 36, 37].While fully developed epitaxial films could be grown on c-BN [36], only oriented growth of particles of size up to about 10 Jlm in diameter and withslight misorientations of a few degrees has been realized sofaI' on BeO. Thisis consistent with the better lattice match and adhesion energy to c-BN.
346
TABLE 3. Surface and adhesion energies (in J/m2 ).
diamond ideal (110) surface energy 6.5c-BN ideal (110) surface energy 5.2zincblende BeO ideal (110) surface energy 4.8
diamond/c-BN adhesion energy 5.4
diamond/BeO adhesion energy 4.6diamond 2 x 1 (111) surfa.ce energy 5.5
Cu (111) surface energy 2.0diamond/Cu adhesion energy 2.3
4.2. RELATED BE COMPOUNDS: BE2C, BE3N2 AND BECN2
Besides the elemental IV, 111-V and 11-VI materials of the second row of theperiodic table, three other compounds with tetrahedral or closely relatedbonding are of interest: BeCN2, Be2C, and Be3N2' The main propertiesobtained in our calculations for the first two of these are given in Table 4.
TABLE 4. Properties of BeCN2 and Be2 C.
StructureLattice constant a (A)ciaub
Bulk modulus B (GPa)Cohesive energy (eV/atom)Band gap (eV)
BeCN2
chalcopyrite3.711.96
0.303337.7
4.2 (6.3)"direct r - r
Be2 Cantifluorite4.29 (4.342)"
2136.0
1.15 (2.15)"indirect r - X
a expt. value in parenthesesb 11 is the internal displacement parameter of the chalcopyrite structure [18]
C With estimated gap correction based on interpolation for related materials.
Beryllium carbonitride BeCN2 is todate a hypothetical material [26, 18),which is here assumed to have the chalcopyrite structure in analogy withother II-IV-V2 compounds such as ZnGeP2' One may think of it as beingderived from BN by replacing every other B (group-III) by a Be (group-II)or a C (group-IV) atom. This substitution maintains the local stoichiometrybecause each N atom is surrounded by two Be and two C atoms. The
•BeCN2
a •
347
c
•• 0Be C N
Figw'e 3. Chalcopyrit.e structure assumed for the hypothetical material BeCN2 . Thest.ruct.ure represented here is idealized. Bond lengt.hs relax by a relative mot.ion of the Nsllblattice t.owards C and away from Be. If Be and C atoms are both replaced by B weobt.ain t.he zincblende stuct.ure of c-BN.
structure is shown in Fig. 3. The bonding in this nlaterial is tetrahedral.There are several interesting things about this material.
It has a lattice constant intermediate between those of c-BN and BeOand may thus form a good bufferlayer for diamond or c-BN growth.We note that it is more ionic than c-BN and may thus, as BeO, bemore likely to avoid layered sp2 bonded structures.It has a pseudodirect band gap in the sense that the "directness" results from the fact that the zincblende X point is folded onto r inthe chalcopyrite structure. A strong mixing of r and X states, however, is expected to occur due to the significant difference between Beand C potentials and, in addition, due to the lattice distortions. Thisshould lead to a significant dipole matrix element and thus stronglight-emission.Related materials such as ZnGeP2 are promising for non-linear opticsapplications [27] due to the occurence of IV-V bonds. If, as is expected,similar non-linear optical advantages ofthe C-N bonds occur here, theywould be of particular interest because the band gap is already in thelTV.
348
Be2C A material with bonding closely related to tetrahedral bonding isBe2C. It has the cubic antifluorite structure. The latter consists of a fccsublattice of carbon atoms with Be atoms occupying all tetrahedral interstitices. In the zincblende structure by contrast, only half of the tetrahedralinterstices of the fcc lattice are filled with a counter ion. We have calculatedthe ground state properties and band structure of this material [38]. Themain results are given in Table 4. The material is an indirect semiconductorwith band gap and lattice constant close to those of 3C-SiC.
Formation energies Under atmospheric conditions Be2C is unstable towards formation of Be(OHh by reaction with water vapor, or BeO withO2, This indicates that Be2C is less stable than Be(OHh or BeO. Nevertheless, a small amount of Be2C formation was detected in growth exprimentsof diamond on BeO [37]. This indicates that in spite of the much smallerformation energy of Be2C than that of BeO, Be2C can form under the veryreducing conditions of diamond growth. This is significant for the prospectsof synthesising BeCN2. Our calculations predict a much higher energy offormation for this material, close to that of BeO. The energies of formation,defined with respect to the elements in their standard state and being positive in the case of exothermic reactions, Le. when the compound has lowerenergy than the sum of the energies of the elements, are given in Table 5.They were calculated by combining our LDA calculated cohesive energiesof the respective solids with LDA binding energies of the molecules O2 andN2 given by Jones and Gunnarsson [39]. In this way, the systematic LDAoverestimate of bonding is cancelled out. The energy of carbon was takento be that of diamond instead of graphite, the latter being almost equal inenergy. Zero point motion corrections were not included in our cohesive energies because for the present purpose, we only need a qualitative estimate.We note that a large part of the stabilization energy of BeCN2 is due tothe effective relaxation energy of 1.2 eVlatom. The related material Be3N2is included in this table. Its structure is an ordered vacancy structure ofthe antifluorite structure. One may see that per equivalent amount of Beatoms, i.e. 1/3 mole of Be3N2' its energy of 47 kcal is also much smallerthan that of BeCN2. This indicates that under reaction conditions whereeither Be2C plus N2, or Be3N2 plus graphite, or BeCN2 could form, thelast is most likely in a thermodynamic sense. It indicates that when Beis exposed to C and N simultaneously, BeCN2 may form exothermically.This calculation, being strictly thermodynamic, of course, does not tell usanything about possible reaction barriers.
349
TABLE 5. Exothermic energies offormation (in kcal/mole)of some Be compounds.
theory expt.
BeO 158 1451/2 Be2C 14.5 141/3 Be3N2 47BeCN2 156
4.3. SILICON CARBIDE
Silicon carbide is unique among the wide-band-gap semiconductors in thesense that it exhibits polytypism. Over a hundred different crystal struchues called polytypes have been discovered [40]. All are tetrahedrallybonded but differ in the stacking of what are the {111} planes in cubicSiC (called 3C-SiC or f3-SiC) and the basal planes {OOOl} in hexagonal orrhombohedral SiC, sometimes collectively denoted et-SiC. The polytypescan most easily be described as a periodic structure of twin boundaries incubic SiC. When a twin boundary occurs every 3 (2) layers, we obtain (3)«(2) which is the Zhdanov notation [41] for 6H-SiC (4H-SiC). When theyoccur every layer we obtain the wurtzite structure which is purely hexagonal and denoted 2H-SiC or (1). Rhomobohedral polytypes exhibit morecomplex patterns such as (32) or 15R.The relative stability of the various polytypes has been studied exten
sively by Heine et 301. [42]. Some of the main conclusions of their work isthat polytypes with bands of three and two layers between twin boundaries(e.g. 6H, 4H, 15R) are most stable, followed by 3C SiC, while least stableis 2H-SiC. On the basis of their total energy calculations, they proposed amodel in terms of interlayer interactions. The origin of the polytypism inthis context is that the first nearest neighbor layer interactions prefer tostack cubically with respect to each other while the second nearest neighbor interaction favors hexagonal stacking and has almost exactly half themagnitude of the first nearest neighbor interaction. The third and higherneighbor layer interactions are found to be negligible. Their model explainsthe preference for polytypes with bands of two and three successive cubically stacked layers. It also explains why in surface growth, if complete 3Dequilibrium is not reached, there is a tendency to form cubic stacking. Thisoccurs because for the surface layer there is only one next neighbor layerinteraction to be considered instead of two in the bulk. Finally,· small differences in the phonon contribution to the free energy were claimed to be
350
the origin of different temperature stability regions for 4H and 6H. Experimentally, it is found that polytype selection can be influenced by a numberof kinetic factors besides thermodynamic equilibrium. Growth on vicinalsurfaces tends to stabilize the same polytype as the substrate on which it isgrown through a ledge growth mechanism. Imperfections on large terracestend to promote independent nucleation and cubic growth. Lower temperatures also tend to lead to cubic growth. All of this is consistent with thenotion that there must be sufficient mobility of species on the surface tomove to ledges to continue the substrate polytype and that otherwise cubicnuclei tend to form. In case of cubic growth in the [111] orientation one is often faced with the problem of so-called double position boundaries (DPB).The latter correspond to the two equivalent orientations of a cubic nucleuswith respect to the substrate. Stress can induce polytpic transformations[43]. Also, ion implantation can influence polytypism after recrystallizationof the damaged area [44]. Finally, doping and growth rates may have aninfluence on which polytypes form during growth [44].Our own work has focused on the relationship between polytypism and
properties. Integrated properties such as elastic constants [45] are onlyweakly dependent on polytype while band structure depends dramaticallyon polytype. This is because band structure essentially corresponds to theformation of electronic standing waves and is thus very sensitive to thestructure [24]. The changes in band structure also lead to important changesin optical reflectivity [46].Fig. 4 shows the dependence of the minimum gap and some specific indi
rect gaps on polytype. The polytypes are here ordered in order of increasinghexa.gonality, Le. the proportion of layers that are stacked hexagonally. Onemay see that changes in the position of the conduction band minimum occur. While the low hexagonality polytypes have their minimum along thehexagonal M - L-axis, including the cubic X-minimum, 2H-SiC has itsminimum at J{ similar to hexagonal diamond and BN. One may also notice that the specific gaps do not vary monotonously with hexagonality.Details on how the band structures are related to each other can be foundin [24, 23].Fig. 5 shows the calculated UV-reflectivity spectra of some SiC poly
types compared to experimental data. We note that a single k-independentand polytype independent gap correction of 1 eV as applied. The latternot only provides good agreement for the various peak positions in the reflectivity spectra but also for the minimum gaps given above. This lendssupport to our statements in the section on computational method that thegap corrections beyond the LDA are relatively structure independent.The changes in band structure between the polytypes leads to different
effective masses, and hence are important for transport [24]. For example,
3C5.0
~ 4.0>~a..«CJClz«co 3.0
6H 15R 4H
r-Kgap
~ r-D gap
2H
351
X 3C,2.0 o 20 40 60
HEXAGONALITY (%)
80 100
Figtll'e 4. Band gaps of SiC polytypes. Squares: experiment, solid circle; thoereticalminimum gap: diamonds, upward and downward triangles, specific k-point gaps as illClicat.ed; dashed line, nearly linear r - U gap along U == AI - L line of hexagonal Brillouinzone.
it is found that 6H SiC has twelve equivalent minima along the M - L a,xesclose to ]v!, which are elongated in the c-direction and with only small barrier between them. This means that at larger temperature the two minimadisplaced along the c-axis can effectively be thought of as one minimumwith a very large mass anisotropy. This is consistent with recent measurements of a large anisotropy in mobilities in 6H-SiC [47].
We have also studied several interfaces involving SiC. SiC/AlN andSiC/BP interfaces [48] will be briefly discussed in the section on alloys because of the close relationship among these topics and the new results wepresent here on the corresponding alloys. We also studied SiC/TiC interfaces which are of interest for metallization of SiC and in connection withceramic reinforcement of metals [49]. Finally, we studied so-called inversiondomain boundaries in SiC [50]. These are planar defects that occur oftenduring growth on stepped Si {001}.
352
C: .
........
B
B
D
D
.•.:
/.....\ :\ C:..
A ....... .,.....f' : .
.........'
....
4H
0.5 2H
0.4
0.3
0.2 \-H-....,.:.:j-++-t-+-l-++++!-+-+-M>-l-+-+-++4-+-+-+-+-l
0.5
0.4
0.3
0.2 i-+-t'..,:+.f-+-Mt-+-l-++++!-+-++->-l-+-+-++4-+-+-+-+-l
0.4
0.3
~:> 0.5tw...Ju..wa:
15Rc
....f\ ..:···
.... ."., .: .
D' /..... C
A ···
3C
6H
0.2 \-H--A-+-++-t-+-l-++++!-+-++->-l-+-+-++4-+-+-+-+-l
0.5
0.4
0.3
0.2 1-+-t-=-1-++++j1-+++-+-1-H-+-+-1-+++-H--+-I-+-+-l
0.5
0.4
0.3
0.2 !...o=:!i::O.....................""-'-'............................J...................J....o..............J
4 5 6 7 B 9 10PHOTON ENERGY (eV)
FigUl'e 5. Reflectivity of SiC polytypes. Solid lines experiment, dotted lines theory.
4.4. BORON COMPOUNDS: BN, BP AND BAS
Table 6 presents an overview of boron compounds c-BN, BP and BAs as
353
TABLE 6. Properties of tetrahedrally bonded boron compounds: alattice constant, B bulk modulus, Eb bond energy, E g minimum bandgap, and conduction-band minimum location kmin.
a (A) B (GPa) Eb (eV) E g (eV) kmin
BN present 3.62 381 3.8 4.5 Xppb 3.606 367 3.6 4.2 XExpt. a,b 3.615 290-465 3.3 6.4
BP present 4.51 172 2.8 1.35 Appb 4.558 165 2.9 1.2 AExpt.a,b 4.538 173-265 2.6 2.4
BAs present 4.79 130 2.3 1.06 Appc 4.777 145 2.6 1.2 AExpt.a,c 4.777 2.1 d
a Experimental values from Landolt-Bornstein Tables [53].b Pseudopotential calculation by Wentzcovitch et al. [51]c Pseudopotential calculation by Wentzcovitch et al. [52]
d Our estimate assuming correction same as for BP. Expt. values are uncertain.
calculated using the ASA-LMTO method. We have included data on BN,which given above to facilitate the comparison to BP and BAs. The reasonwhy BP in the table below has a slightly underestimated lattice constantwhile the others have sligthly overestimated values is that the calculationof [48J for BP did not include the so-called combined correction. For thepresent purposes this is not relevant. Our results are in good agreementwith the previous calculations by Wentzcovitch et al. [51,52], although ourvalues for the gaps differ somewhat. We note the usual "softening" as we goto heavier elements. Due to the very low ionicity mentioned earlier, the BPand BAs compounds have a band structure somewhat similar to Si withthe minimum along the A axis close to X. Experimental band gaps arenot well known~ We estimate a gap correction of ",,1 eV for both BP andBAs on the basis of their similarity to SiC and Bechstedt and del Sole'smodel [15J. We emphasize that the low ionicity is expected to reduce polarscattering which may give these materials an advantage in mobilities.
4.5. GROUP-III NITRIDES
We have performed total energy and band structure calculations for the entire series of Group-III nitrides in both the zincblende and wurtzite structure. The two forms are close in energy and can both be stabilized undersuitable epitaxial growth conditions. This has already been realized for GaNand InN but not for AIN. At high pressure the nitrides exhibit a phase tran-
354
sition to the rocksalt structure. This is a well known phenomenon due tothe increase in ionicity under compression and the tendency towards highercoordination for ionic compounds. It has been studied theoretically by Gorczyca and Christensen [22] and by Munoz and Kunc [54] and experimentallyby Perlin et al. [55], Volstadt et al. [.56] and Deno et al. [57]. Our calculatedgroundstate properties can be found in [23] while full details on the bandstructures are given in [.58].
15 .--------,----,----,------,-----,
r"1r "1
-----
r v
10
>~
>- 5<.9a:wzw
0r ...v..............
1 ~.,. ~ ~_--------..:...s:::~
r~===~~~---- ---6
\
r6c
-...:
r~\I \
K"c "Xc.............. "1 ........ \
',-\ ----------\\
-5 l...-__---L ----'- .l..-.-__---L --l
BN AIN GaN InN
Figlwe 6. Trends in selected band states of III-nitrides with atomic number in zincblendeand wurtzite structure.
Trends in the band stru.ctures. The trends of some eigenvalues with cationin the series of nitrides are shown in Fig. 6. It includes the important eigenvalues near the gap, Le. the valence band maximum f]'s for ZB or f]' and f6for wurtzite, the conduction band state fi and K 2for WZ and Xf for ZBwhich "compete" for the minumum. The conduction band fi states in bothwurtzite and zincblende are mostly s-like (actually purely s-like in ZB) and,being antibondillg states, have important cation-s components. One maysee that these states show the largest variation with atomic number, and,in particular, are monotonically decreasing with increasing atomic number.
355
Of course, the above statement requires that we compare the different bandstructures with respect to the same reference level. The natural referencelevel used here is the ASA zero of energy [59] which coresponds to the average of the point charge electrostatic potential and is close to the averagepotential in the interstitial region. This is not sufficiently accurate to determine band-offsets because that requires consideration of charge transfer.But it is adequate for our present qualitative discussion. The reason forthe downward trend with atomic number is well known and also ocurs forexample in the series C-Si-Ge-Sn. It is simply due to the fact that s-orbitalshave a non-zero value at the nucleus and as such feel the -Z/1' potential.Thus, heavier elements have lower s-levels with respect to correspondingp-levels. Since the r 1 state is s-like and the Xi-level in zincblende (or diamond) is a mixture of sand p-like components and the ](~ state in wurtziteor lonsdaleite is purely p-like, this explains why diamond, SiC and BN areindirect gap materials in either zincblende or wurtzite structure while GaNand InN are direct in both. AlN turns out to be indirect in the zincblendestructure and direct in the wurtzite. Even in WZ, however, the indirectand direct gap in AlN are close to each other in energy. Fig. 6 shows thatthe variation of rt is linear from Al to In but has a different slope betweenB and Ai. This emphasizes again the peculiar role of the second row elements. In addition, the rts state is lower in c-BN than the rj state. Thisagaiu is a manifestation of the relative energies of p alid s-levels and thusa result, once again of the absence of a p-like core f9r 2p wave functions.We note that the levels shown are LDA levels which are underestimated forthe conduction band states as discussed above. Again, however, the LDA isadequate for the study of the structurally induced changes in band gap. Itis noteworthy that for zb-InN the LDA gap is actually negative while thereal gap is 1.9 eV. The gap in this material is thus almost entirely due tothe self-energy correction.
Ionicity explains why the bandgap variation is so much stronger in theIII-N series than in the III-As or III-P series. The reason for this is that thehigh ionicity makes the cation component of the rj state relatively moreimportant. The lower ionicity of GaP compared to GaN, for example, isconsitent with GaP having an indirect gap. Indeed, the rt state havinga smaller Ga (heavy element) component is less sensitive to the nuclearpotential and lies relatively higher with respect to the Gap-Pslike Xi state.By the time we get to GaAs, the anion is also a heavier element and thus italso tends to bring the rt state down with respect to p-like states. Ionicityalso explains why SiC is indirect in both structures while AlN is only directin wurtzite. The reason is that SiC being less ionic has a relatively smallercomponent of the Si (Le. "heavier" element) in the conduction band rtstate. We note that even for Si, the p-like rts state lies below the s-like r 1
356
state.
Elastic constants. We have recently also performed full-potential calculations of the zincblende III-nitrides including uniaxial strain distortions.This allowed us to obtain the full set of cubic elastic constants. From thelatter, the hexagonal elastic constants can be obtained by a simple tensortransformation operation [60] under the assumption that the bonding issimilar in both materials. The changes due to the twinning of the tetrahedrons can be incorporated in this transformation [60].
TABLE 7. Elastic constants for III-nitrides in GPa. (expt.values in parentheses)
AIN GaN InN
B 202 201 139
cubic C'll 304 296 184C'12 152 154 116C'44 199 206 177
hexagonal C'll 398 (345)G 396 (296)b 271 (190)b
C'12 140 (125) 144 (130) 124 (104)C'13 70 (120) 64 (158) 21 (121)
C'33 468 (395) 476 (267) 375 (182)C'44 96 (118) 91 (24) 46 (99)
G Tsubouchi and Mikoshiba [61]b Savastenko and Sheleg [62]
Table 7 gives our calculated elastic constants for cubic and hexagonalIII-nitrides [63]. The deviations from the experimental values are of theorder of 20 % for AIN - except for the C13 which is difficult to measure- and worse for the other two materials. Particularly disturbing is thedeviation by almost a factor 2 for the Cll and C33 in GaN and InN. Theselarge elastic constants are closely related to the Youngs moduli in the cplane and perpendicular to it and should be more easy to determine thanthe shear moduli. These discrepancies are much worse than typical for LDAresults and suggest problems with the experimental values. We note that forSi [19], diamond [19], and SiC [45], the elastic constants obtained by thismethod are to within 10 %of the experimental values. Only for AIN are theexperimental values given here based on sound velocity measurements. Inthe other cases, they are based on analyses of X-ray linewdiths in powders,which is a rather indirect and probably inaccurate method for obtaining
357
elastic constants. The poor quality and small sizes of crystals available atthe time of the measurements is probably responsible for these experimentalproblems.
Ga3d and In4d semicore states A noteworthy aspect of the GaN and InNbandstructures lis the effects of the Ga3d and In4d semicorelevels. Becausethe lattice constants of these materials are fairly small, the Ga or In atomsare close to each other and their semicore d states have significant interactions. Also, they are close in energy to the deep N2s states. Indirectinteractions through hybridization with N2s thus further broadens thesebands. The LDA eigenvalues of Ga3d and In4d actually overlap with theN2s bands. This means that the Ga3d and In4d participate in the bonding.It is found necessary to include these d-states as valence bands in order toobtain accurate values for the ground state properties [16, 23, 64]. On theother hand, in X-ray photoelectron spectroscopy (XPS), the Ga 3d levelsare found to be situated well below the N2s. This is another manifestationof the difference between LDA eigenvalues and the true quasiparticle excitations of the system. These effects are more pronounced for narrow baudsand increase with increasing energy separation from the Fermi level. Wefound that the XPS spectrum of Ga3d can be well accounted for by meansof the so-called ~SCF approach [16].
Interfaces We have calculated band-offsets among III-N semiconductors[65,66]. The calculations were performed for the zincblende crystals and forthe {llO} interface. Since the valence-band maxima are similar in nature inboth strucures, one may expect that valence band offsets would be similarfor the wurtzite. We can then obtain the conduction band offsets by addingthe experimental gaps.
TABLE 8. Valence-bandt::.E. = E;; - E: (eV) and conduction-band t::.Ec = E~ - E! offsetswith t::.E~ for zincblende and t::.E':for wurtzite.
t::.E. t::.E~ t::.E':
AINjGaN 0.85 0.81 1.78GaNjInN 0.51 1.07 1.09AINjlnN 1.09 2.15 3.14
358
The valence-band offset of AlN IGaN measured by XPS [67] is 0.8 ±0.3eV in good agreement with our results. Interface orientation effects need tobe studied separately and are expected to be of the order of a few 0.1 eV.
0.5 ZBGaN
0.4
0.3
0.2
0.1
0.0
~0.5 WZ
:> 0.4 It= " Ellc() 0.3 1\w ,-' 0.2 \
u.W \a: 0.1 E.lc ,,
'i" --0.0
0.5 WZ
0.4
0.3
0.2 ....""
0.1",
0.00 5 10 15 20 25 30
PHOTON ENERGY (eV)
Figure 7. UV-reflectivity of GaN. The bottom panel shows theory and experiment forwurtzite and E 1. Cj the second panel shows the theoretically predicted polarizationdependence for wurtzite: dashed line E II c, full line E 1. Cj the top panel shows thecalculated reflectivity for zincblende.
Optical p1'Operties We have calculated optical response functions for AlNand GaN. Our results for GaN are shown in Fig. 7. As for SiC polytypes,we find that good agreement with all major peak positions in experimentalreflectivity can be achieved with a single constant gap correction or fV1.0
359
eV. There is, however, a significant discrepancy in absolute intensity whichincreases with increasing energy. A possible explanation for this is diffusescattering in the experiment. Our calculations predict significant differencesin spectra between zincblende and wurtzite and significant anisotropy in thecase of wurtzite. These remain to be confirmed by experiment. For wurtziteGaN, our calculated results and the experiments by Rife et al. [68], shownhere a,re in good agreement with those of Olson et al. [69]. A more detailedanalysis of these spectra will be given elsewhere.
0.5
WZ-AIN0.4
~> 0.3i=()wLi 0.2wa:
0.1
0.0 o 5 10 15PHOTON ENERGY (eV)
20 25
Figtll'e 8. Calculal,ed refletivity of wurtzite AIN: full line, E ..L c, dashed line, Ell c.AnLDA gap correction of 1.72 eV was included.
Our calculated reflectivity for wurtzite AIN is shown in Fig. 8. Theresults are in good agreement with those of Loughin et al. [70] as far aspeak positions and general shape of the spectrum is concerned in the energyregion below"" 10 eV. As for GaN, our calculations predict a significantlystronger reflectivity at high energies than observed in the experiment.
4.6. 1I-IV-N2 MATERIALS
An interesting dass of wide-band-gap semiconductors, which has sofar received very little attention, is formed by the II-IV-V2 materials. These arederived from the corresponding III-V compound by substituting half of theIII elements by a group II and half by a group IV. This does not completelyspecify the crystal structure. We have performed studies of thes materials
360
asstlltling the chalcopyrite structure because the latter is a simple superstructure of zincblende and is the equilibrium form of II-IV-V2 materialsother than nitrides such as ZnGeP2' ZnGeAs2, MgSiP2. In this structure,each group-V atom is tetrahedrally coordinated with two group-II atomsand two group-IV atoms. We have studied the compounds with II=Be orMg and IV= Si and C and also MgSiP2 [18]. We have already briefly discussed BeCN2 along with diamond, c-BN and BeO. Among these nitrides,only the silico-nitrides with IV=Si have sofaI' been synthesized. They havea crystal structure that is derived from wurtzite in a manner similar to howchalcopyrite is derived from zincblende. We have sofaI' only studied the simpler but related chalcopyrite structure. Our main results are summarizedin Table 9.
TABLE 9. Properties of II-IV-V2 materials: lattice constant a, bond lengths d,bond energ)' Eb, bulk modulus B, band gap Eg •
a (A) dIV-V (A) dII-V (A) Eb (eV) B (GPa) E g (eV)
BeCN2 3.n 1.50 1.71 3.8 333 4.62 (6.3)a
MgCN2 4.11 1.58 1.92 2.5 210 3.69 (5.0)BeSiN2 4.10 1.70 1.88 3.0 240 4.47 (5.7)
1.76b 1.76b
MgSiN2 4.44 1.77 2.08 2.8 195 4.03 (5.3)1.76b 2.08b
MgSiP2 5.64 2.26 2.59 1.9 75 1.77 (2.5)
a Including estimated correction, see Petukhov et al. [18]b Experimental values, see Petukhov et al. [18] for original refs.
Generally speaking (except BeCN2), these materials have higher gapsthan the corresponding III-nitrides, which is probably related to theirhigher ionicity. The folding effects of the chalcopyrite lead to conductionband minima at r, except for MgSiP2. Except for BeCN2, however, thevalence-band maxima are displaced from r. Although the gaps are thusindirect, except in BeCN2, they are very close to the corresponding directgaps. In Fig. 9 we show the gaps versus lattice constant along with thoseof the related nitrides, diamond, and BeO. The figure also includes a calculated gap for AlBN2 in the (idealized, i.e. unrelaxed) chalcopyrite structure.The lines are only guides for the eye, and should not be interpreted as gapsof a.Iloys since the latter may exhibit non-linear behavior with lattice constant. One may apprecia,te from this figure that these materials all havegaps similar to those of diamond and c-BN but extend over a lattice constant range where no binary tetrahedrally bonded compounds occur. Theythus fill the gap between the lattice constants of the tetrahedrally bonded
361
12
0BeO
10
S-O
z-BeO~a.« 8(!)
w-~N0z«CD
AINBeCN2c-BN 0
60
doC
z·AIN
43.4 3.6 3.8 4.0 4.2 4.4
a (Angstrom)
Figlll'e 9. Band gaps ofII-IV-Nz materials and related binary compounds versus effectivecubic la.t.tice constant, a (A). The squares indicate wurtzitic compounds, the circles cubicor closely related structures. Note tbat for BeO and AIN tbe equilibrium phase is wurtzitewhile the zincblende phases indicated by z- are hypotbetical.
materials made from elements of the second row (B, C, N) only and thosemade from mixtures of the elements of the second and third row (SiC, BP,AIN). This makes them possibly useful as bufferlayers or substrates for thegrowth of c-BN and diamond. The figure suggests that even higher gapscan be expected for the corresponding wurtzitic compounds.An interesting question is whether the carbonitrides which sofaI' have
not been synthesized can be stabilized. From the above cohesive energiesand our calculated cohesive energy of SiC and diamond it follows that thereactions
MgSiN2 + 2CBeSiN 2 + 2C
MgCN2 + SiC,BeCN2 + SiC,
D.H = 115kcal/moleD.H = -84kcal/mole
(5)
«3)
are respectively endothermic and exotherimc. The relative unstability ofMgCN2 with respect to MgSiN2 follows from its larger mismatch in atomic
362
radii. Similarly BeCN2 has less strain to overcome than BeSiN2. The synthesis of BeCN2, which is the most interesting of this class of materialsbecause its properties are closest related to those of diamond and becauseit is a direct gap material in contrat to its parent compound c-BN, wouldseem to be synthesizable by the substitution of Si by C in the existing compound BeSiN2 • It remains an open question whether alternative 1I"-bondedcrystal structures are possible for these materials. Their higher ionicitywhich favors higher coordination, however, disfavors such structures. Thishas been argued earlier in this paper to be the reason why such structuresdo not appear for BeO. Since BeCN2 is intermediate in properties betweenc-BN and BeO, we are presently unable to predict whether it will behavelike BN and exhibit 11" bonding, or whether it will behave like BeO and notshow 11"-bonding.
4.7. ALLOYS
Several alloy systems among the wide-band-gap semiconductors are of interest. The III-nitrides are known to form continuous alloys. Of these, wehave sofaI' only studied the GaxAlt_xN system [71].
Gaa,All_xN alloys The band gaps were found to behave nearly linearlywith concentration and a cross-over takes place between direct and indirectgaps in the zincblende alloys at about 40 %GaN. Th~ relatively low energyof formation (6.Ef ~ 0.02 eV/ atom and hence the good miscibility we foundfor this alloy system is expected on the basis of the chemical similarity ofAl and Ga and the close lattice match.
Ala:BI_xN alloys Alloys with of III-nitrides with c-BN suffer from a muchlarger atomic size mismatch. We have not yet performed a systematic studyof them. Only the 50 % compound with the chalcopyrite structure wasinvestigated in connection with the related II-IV-N2 compounds. Our calculated cohesive energies for AlBN2 , c-BN and AIN yield a formation energy of 6.Ef ~ 2 eV/ atom, Le. two orders of magnitude higher than forAlxGal_xN. The cohesive energy used here does not include relaxation ofthe bond lengths and is thus an overestimate. Nevertheless, it is clear thatthe AlxBI-xN aHoy system will have a much lower solubility. The bandgap we obtained for AlBN2 is significantly less than the average of those ofz-AIN and c-BN as can be seen in Fig. 9. The bandgap bowing is conventionally described by means of a bowing parameter b defined by
Eg(x) = E g + 6.Eg(x - 1/2) - bx(l - x), (7)
363
with Eg = [Eg(O) + Eg(1)]j2 the average gap and !:lEg = Eg(1) - Eg(O)the gap difference between the endpoints A and B. Using our LDA gapsfor c-BN, chalcopyrite AlBN 2 and z-AlN of respectively 4.73,3.71 and 3.44eV, we find b = 1..5 eV. Including the estimated gap corrections, we obtainfor these numbers respectively: 6.4, 5.0, 4.7, b = 2.2 eV. Bond length relaxations are expected to influence this bowing strongly. The present numbersare thus only an indication that bowing is large in this system. A similarsituation is expected to hold for BxGal_xN alloys.
Heterovalent alloys The heterovalent alloys among group IV compounds,or, elemental solids, and III-V compounds present an even more challengingproblem. We have carried out a study of diamond-c-BN alloys [72] and ofthe alloys of SiC with AIN and BP. All of these are reasonably latticematched alloys systems. We have thus neglected bond-length rf:!laxationssofar. Details of our calculations for diamond-c-BN were presented in [72].Our results on the SiC-AIN and SiC-BP were reported at the MRS SpringMeeting 1993 but have not yet been published, except for a brief mentionof the SiC-AlN results in [73, 58].First, we discuss the short-range order in these systems and our compu
tational approach. Because of the donor and acceptor character of IV-V andIV-III bonds respectively, there is a significant energetic advantage in compensating the two types of cation-anion bonds by placing them in close proximity and by incorporating them in equal amounts. 'Charge transfer fromone to the other will satisfy each bond with two electrons and thus yield again in band-structure energy [74]. The charging of the bonds that accompa.nies this, however, leads to a mutual attraction. We thus assume thatlow-energy configurations have nearest neighbor IV-V and IV-III bonds.For a similar reason, we exclude cation-cation and anion-anion "wrong"bonds completely as these nearest-neighbor interactions would lead to highenergies. We are then left with a simple pseudobinary alloy problem inwhich we consider only disorder on one of the sublattices and place theaccompanying anions so as to satisfy the above short-range order chargeneutrality or compensation rules. In the case of SiC jBP, it is necessary totake into account the fact that in BP, B behaves as anion and P as cation.Thus we assume Si-P and C-B bonds. The other structures were found tohave sufficiently high energies that they could be ruled out [48].This does not completely specify the relevant local structures but guides
the choice of physically plausible structures. We then choose the L12 ordering for 25 % or 705 % and L 10 ordering for SO %cation mixtures on the fccsublattice of the zincblende structure. These form the basis of the so-calledConnolly-Williams method [75] in which one considers fcc nearest neighbortetrahedrons of the type An B4 - n with A and B the two types of cations and
364
n ranging from 0 to 4. Next, we perform first-principles calculations of thetotal energies and band structures of these mixed ordered compounds. Wethen assume that the properties of the disordered system can be obtainedas a suitable average over these "basis-states". In principle, one should obtain the distribution of clusters in the alloy at a given concentration andtemperature by the minimization of the free energy. The latter includes theenergy of formation, the configurational entropy, and eventually the vibrational free energy and elastic strain energy contributions. This approach isknown as the cluster variation method.For the alloys considered, we find that the energy of formation is nearly
a parabolic function of the concentration and is high. For 50 % SiC-AlNit is about 0.25 eVlatom. This is high and implies that there is onlya very limited thermodynamic miscibility. The near parabolicity impliesthat the Connolly-Williams cluster-expansion model can be mapped ontothe ferromagnetic Ising model [72]. This allows as to make an immediate and simple estimate of the miscibility temperature, which is given by0.816 x 2~Ef(1/2)lkB with kB the Boltzman constant and ~Ef(1/2) theenergy of formation of the 50 % compound. This has a value of "'5000 Kin the present case. Lattice relaxation effects and inclusion of longe-rangecorrelations may easily reduce this number by a factor of two. The presentcalculation thus only gives a rough estimate. The experimental phase diagram [76] of the SiC-AlN alloy system indicates a miscibility temperatureof "'2000°C. Furthermore, it indicates that alloys preferentially form in thewurtzite structure except near the SiC rich end whel:e other polytypes arestabilized. In any case, for the present purposes, it suffices to point out thatthese heterovalent alloys are likely to be metastable towards phase separation up to rather high temperatures. Nevertheless, once they have beenformed by non-equilibrium processes, e.g. from a melt or gas phase mixture, the phase separation process may be slow because it would requiremassive diffusion. The overall conclusion of this is that thermodynamicequilbirum considerations may be rather irrelevant to the actual structureof these alloys. Thus, instead of calculating the thermodynamic equilibriumdistributions of the various types of clusters, we can equally well assume arandom distribution,
(8)
of the 5 types of basic tetrahedron clusters. We use this procedure to obtaina first approximation to the band-gap behavior in the disordered alloys.Before presenting those results, we note that inclusion of configurations ofthe type we ruled out from the start would drive up the formation energyeven higher and reinforce our conclusions. Although such configurations
365
will probably appear in small amounts due to random errors during growth,they can be considered as defects instead of the average structure. In somesense, they are similar to anitisite defects in a simple binary compound.Our calculations for the interfaces [48] which considered "wrong bonds" ofthe type Si-AI, C-N, showed that these give rise to defect states in the gap.
Figs. 10 and 11 show our calculated band structures of the ordered compounds of SiC-AlN and SiC-BP respectively. The Brillouin zone notationfollows Bradley and Cracknell [77]. Figs. 12 and 13 show the band gaps ofthe ordered compounds we used as "basis" for the cluster expansion as wellas the disordered alloy average obtained using a random probability distribution of the clusters. The bowing coefficients are included in the figure.We give both the LDA values and the corrected gaps for the ordered compounds. We assume here that the correction can be linearly interpolatedbetween the known corrections for the end compounds.
We note that the x zb point where the conduction band-minimum ocursfor SiC and AIN is folded onto r in the ordered intermediate compounds.For the 50 % and 75 % AIN compounds, however, the valence-band maximum does not ocur at r. This is related to the very flat valence-bandmaximum along the r - X direction in the tetragonal Ll0 structure whichcorreponds to r - !vI in the simple cubic L12 structure. The valence bandmaximum in both cases occurs at R. Both the minimum gap and the smallest direct gap are indicated in Fig. 12. The smallest direct gap, in any caseis only pseudodirect because it corresponds to a trans·ition form the highestvalence state at r to the folded X zb minimum. Since the perturbation ofthe parent band structures is strong in these materials, the matrix elementsfor these pseudodirect transitions may be appreciable. The curve for thedisordered alloy gap is based on the minimum gaps. In SiC-BP, the alloyswith 50 %or more BP have negative LDA gaps.We note that in both cases a very large bowing is present. The band
gaps in SiC-BP are predicted to be smaller than the almost equal gaps atthe end points over almost the entire composition range. As in diamondc-BN, [72] this is related to the type-II offset in this system. Indeed, theband-offset at heterojunctions among these materials is shown in Fig. 14and indicates that the gap of the overall SiC-BP heterojucntion system issmaller than that of the two pure compounds. This is thus proposed to bethe prima.ry cause of the large bowing. SiC-AlN has a type-I offset, Le. anoffset with the larger gap enclosing the smaller gap. Nevertheless, a strongbowing is also apparent here. In both cases the origin of this phenomenonma.y be argued to be the large charge transfers taking place between theconstituent semiconductors. If one thinks of the alloy as a molecular levelmixture of regions of BP (or AIN) and SiC, it is clear that charge transferprocesses similar to those found at the heterojunction must occur. To what
366
(SiCh(AIN)
'>~
>-~QlC
W
-18'-------'---'---L.-----'r fA
(SiC)(AIN)
r z A
(SiC)(Al!\
-16
-18 '-------'------'----'-----'r
Figure 10. Band structures of ordered compounds of (SiCh(AIN), (SiC)(AIN) and(SiC)(AINh. The gap is indicated by shading.
(SiCh(BP)
'>~>-~QlC
W
-16
_18l--l-....L--'------'r
(SiC)(BP) (SiC)(BPh
-16_18l-.....l._--L__'--_---'
r
Figw'e 11. Band structures of ordered compounds of (SiCh(BP), (SiC)(BP) and(SiC)(BPh ,The gap is indicated by shading,
367
5.5
5 SiC-AlN
4.5;;- 4
Q) b=3.7 eV'-' 3.5 c
~0 3
~ 2.5 c
•<t: 2c •
~
1.5
1
0.50 0.25 0.5 0.75
x
Figw'e 12. Band gap versus concentration in zincblende (SiCh-x(AlN)x allo.vs: triangles, LDA minimnm gaps for ordered compounds; pluses, LDA pseudodirect gaps ofthe ordered compounds; squares, corrected gaps for the ordered compounds; solid line,corrected gaps for the disordered alloys. b is the bowing parameter.
2.5
SiC-BP2
~ 5~ 1.'--'
b=4.5 eV
1
0.5
o
0.750.25-0.5
o 0.5X
Figtl1'e 13. Band gap versus concentration in zincblende (SiCh-x(BP)x alloys. triangles,LDA minimum gaps of the ordered compounds; squares, corrected gaps for the orderedcompounds; solid line, corrected gap for the disordered alloys. b is the bowing parameter.
extent one can transfer this idea to the homogeneous alloys considered inour calculations is not dear. Qualitatively, however, it is dear that similarB-Si (N-Si) and P-C (C-AI) bonds occur and that the balance betweenthe charge transfers in these bonds determines the electrostatic dipole offsetting the potentials in BP (AIN) units from the ones in SiC units.We note that for SiC-AlN alloys of wurtzitic structure, a large bowing
368
4
3
....-> 2Q)-->-CJa: 0wZ -1W
-2
-3
Ec
BP
Ev SiC
AIN
Figw'e 14, Calculated band offsets at SiC/AIN and SiC/BP zincblende heterojunctiolls,The valence band maximum Ev of SiC is arbitrarily chosen as zero, If transitivity isassumed the figure also gives all estimate of the AlN/BP offset,
has indeed been found to exist [78] both for the direct r - r gap and theindirect r - J{ gap. Although our present calculations are for zincblendederived a.lloys, corresponding charge transfer effects are expected for htewurtzite alloys. Thus the experimental results provide indirect confirmationof the validity of our results. Work on the wurtzite alloys is in progress.
5. Conclusions
In this paper, we have reviewed the electronic structure and related properties of a wide range of tetrahedrally bonded semiconductors whose commonfeature is that they include early elements of the periodic table, i.e. Be, B,C, Nand O. This survey includes diamond, SiC, the binary compoundsof formula III-N with the group-III element including B, AI, Ga, In; thecompounds of formula. B-V with the group-V element being N, P, and As;and the II-IV-N2 compounds with IV being Si or C and II being Be a.ndMg. It also includes the II-VI compound of the second row: BeO. The Be2Cand Be3N2 compounds, which have a bonding type that is closely relatedto tetrahedral bonding, were also included as results for them are relevant
369
for the discussion of the synthesis of BeCN2. Bandgaps, lattice constants,bond energies or cohesive energies, bulk moduli and, in the case of the nitI'ides, other elastic constants were presented. The strong covalent bondingand the deep N2p levels were shown to be the basic reason for the wideband gaps occuring in these materials. The interrelationship between various properties, and, in particular, between crystal structure and electronicstructure was discussed in detail. Special attention was given to SiC whichforms in a large number of polytypes. The optical reflectivity of some of themost important polytypes of SiC and of GaN and AIN were discussed. Theyprovide an important experimental check of the calculated band structures.We have revealed the trends with atomic number and ionicity. Related issues such as the electronic structure of alloys and interfaces were discussed.Besides an extensive discussion of previously published material, the paperincludes new results on SiC-BP and SiC-AlN alloys, new results on hexagonal modifications of diamond, BN and BeO, and on the related materialBe2C, We have analyzed the prospects for synthesis of a new matrial BeCN2on the basis of our calculated energies of formation.
Acknowledgements
We thank M. Alouani, M. van Schilfgaarde and M. Methfessel for providing computer programs, J. C. Angus, A. Argoitia for collaborations ondiamond, W. J. Choyke and R. P. Devaty and coworkers for collaborationson SiC reflectivity, and J. Rife, W. R. Hunter and D. Ie Wickenden for collaborations on GaN reflectivity. The nitride work was supported by NSFDMR-92-22387, the diamond work by the NSF-MRG DMR-89-03.527 andONR N-00014-89-J-1631 and the SiC work by Wright-Laboratories (Contract No. F33615-93-C-5347). Part of the computations were done a theOhio Supercomputer Center.
References
1. J. C. Phillips, Bonds and Bands ill Semiconductors, (Academic, New York, 1973).2. W. A. Harrison, Electronic Structure and Properties of Solids, (Dover, New York1989).
3. W. R. 1. Lambrecht and B. Segall, Phys. Rev. B 41, 2832 (1990).4. R. M. Wentzcovitch, A. Continenza, and A. J. Freeman, in Diamond, Silicon Car
bide and Related Wide Balldgap Semiconductors, ed. J. T. Glass, R. Messier, and N.Fujimori, Mater. Res. Soc. Symp. Proc. Vol. 162, (MRS, Pittsburgh 1990), p. 611.
5. C.-Y. Yeh, Z. W. Lu, S. Froyen and A. Zunger Ph.ys. Rev. B 46, 10086 (1992).6. H. Amano, M. Kito, K HiramatslI, Jpn. J. Appl. Pllys. 28, L2112 (1989).7. R. H. Wentorf, Jr., J. Chem. Pllys. 26, 956 (1957); ibid. 34,809 (1961); ibid. 36,1990 (1962).
8. P. Hohenherg and W. Kohn, PllYS. Rev. 136, B864 (1964); W. Kohn and L. J. Shamibid. 140, A1133 (1965).
9. 1. Hedin and B. I. Lundqvist, J. Pllys. C 4, 2064 (1971).
370
10. L. Hedin, PIIYs. Rev. 139, A796, (1965).11. M. S. Hybertsen and S. G. Louie, Phys. Rev. B 34, 5390 (1986).12. R. W. Godby, M. Schluter, and L. J. Sham, Phys. Rev. B 37, 10159 (1988).13. M. Rohlfing, P. Kruger, and J. Pollmann, Phys. Rev. B 48, 17791 (1993).14. M. Palummo, L. Reining, R. W. Godby, and C. M. Bertoni in Proc. 21 st Int. Conf
on the Physics of Semiconductors, held in Beijing, Aug. 1992, ed. Ping Jiang, andHou-Zhi Zheng, (World Scientific, Singapore 1993), p. 89.
15. F. Bechstedt and R. Del Sole, Phys. Rev. B 38, 7710 (1988).16. W. R. L. Lambrecht, S. Strite, G. Martin, A. Agarwal, H. Morko<;, and A. Rockett,
PIIYs. Rev. B (1994), submitted.17. O. K. Andersen, PllYs. Rev. B 123060 (1975); O. K. Andersen, O. Jepsen, andM. Sob, in Electronic Band StructUl'e and its Applications, edited by M. Yussouff,(Springer, Heidelberg, 1987).
]8. A. G. Petukhov, W. R. 1. Lambrecht. and B. Segall, Phys. Rev. B 49,4549 (1993).19. M. Met.llfessel, PIIYs. Rev. B 38, 1537 (1988); M. Methfessel, C. O. Rodriguez, andO. K. Andersen, PllYs. Rev. B 40, 2009 (1989); H. M. Polatoglou and M. Methfessel,ibid. 41, 5898 (1990).
20. W. R. L. Lambrecht and B. Segall, PllYs. Rev. B 40, 9909 (1989); 41, 5409 (1990).21. M. R. Salehpour and S. Satpathy, Phys. Rev. B 41, 3048 (1990).22. 1. Gorczyca and N. E. Christensen, Physica B 185, 410 (1993).23. W. R. 1. Lambrecht and B. Segall, in Wide Band Gap Semiconductors, ed. T. D.Moustakas, J. 1. Pankove, and Y. Hamakawa, Mater. Res. Soc. Symp. Proc. Vol. 242(MRS, Pittsburg, 1992) p. 367.
24. W. R. L. Lambrecht in Diamond, Silicon Carbide and Nitride Wide-Bandgap Semiconductors, ed. C. H. Carter, Jr., G. Gildenblat, S. Nakamura, and R. J. Nemanich,Mater. Res. Soc. Symp. Proe. Vol. 339 (1994), to be published.
25. K. J. Chang, S. Froyen, and M. 1. Cohen, J. Phys. C 16,3475 (1983); K. J. Changand M. L. Cohen, Solid State Commun. 50, 487 (1984).
26. W. R. L. Lambrecht and B. Segall, Phys. Rev. B 45, 1485 (1992).27. B. F. Levine, Pbys.Rev. B 7, 2600 (1973).28. S. Fahy, S. G. Louie, and M. L. Cohen, Phys. Rev. B 34, 1192 (1986); ibid. 35, 7623(1987).
29. R. Wentzcovitch, S. Fahy, M. L. Cohen, and S. G. Louie, Pbys. Rev. B 38, 6191(1988).
30. S. P. Mehandru, A. B. Anderson, and J. C. Angus, J. Pbys. Chem. 96, 10978 (1992).31. W. R. L. Lambrecht, C. H. Lee, B. Segall, J. C. Angus, Z. Li, and M. Sunkara,NatllJ'e 364, 607 (]993).
32. J. Tersoff, PllYs. Rev. Lett. 61, 2879 (1988).33. W. R. L. Lambrecht and B. Segall, J. Mater. Res. 7, 696 (1992).34. W. R. 1. Lambrecht, Pbysica B 185, 512 (1993).35. S. Koizumi, T. Murakami, T. Inuzuka, and K. Suzuki, AppJ. Phys. Lett. 57, 563(1990).
36. A. Argoitia, J. C. Angus, J. S. Ma; L. Wang, P. Pirouz, and W. R. L. Lambrecht,J. Matel·. Res. (1994), in press; L. Wang, P. Pirouz, A. Argoitia, J. S. Ma, and J. C.Angus, Appl. Phys. Lett. 63 1336 (1993).
37. A. Argoitia, J. C. Angus, L. Wang, X. 1. Ning, and P. Pirouz, J. Appl. Pbys., 73,4305 (1993).
38. C. H. Lee, W. R. L. Lambrecht, and B. Segall, unpublished.39. R. O. Jones and O. Gunnarsson, Rev. Mod. Phys. 61, 689 (1989).40. A. R. Verma and P. Krishna, Polymolpllism and Polytypislll in Crystals, (Wiley,New York, 1966).
41. G. S. Zhdanov, C. R. Acad. SCi. USSR, 48,43 (1945).42. V. Heine, C. Cheng, G. E. Engel, and R. J. Needs, in Wide Band Gap Semicolldcu
tors, edited by T. D. Moustakas, J. 1. Pankove, Y. Hamakawa, Mater. Res. Soc. Symp.
371
Proc. Vol. 242 (MRS, Pittsburgh, 1992) p. 507; C. Cheng, R. J. Needs, and V. Heine,l. PlIys. C 21, 1049 (1988); J. J. A. Shaw and V. Heine, l. PlIys.: Condens. Matter24351 (1990); C. Cheng, V. Heine and 1. L. Jones, ibid., 5097; C. Cheng, V. Heineand R. J. Needs, ibid. 5U5. N. Churcher, 1<. Kunc, and V. Heine, l. Phys. C 19,4413(1986).
43. P. Pirouz, Scripta Met. 23, 401 (1989); P. Pirouz and J. W. Wang, Ultramicroscopy51, 189 (1993); P. Pirouz, Inst. of Physics Conf. Ser. No. 104, p. 49 (1989).
44. J. Pezoldt, A. A. Kalnin, D. R. Moskwina, and W. D. Savelyev, Nucl. Instr. andMeth. iII Physics Res., B80/81, 943 (1993). A. A. Kalnin, F. Neubert, J. Pezoldt,in Pl'Oc. 4th European Conference on Diamond, Diamond-like and Related Materials,Albufeira, Portugal, Sepember 20-24, 1993, ed. P. K. Bachman, 1. M. Buckley-Golder,J. T. Glass, M. Kamo, (Elsevier Sequoia, Lausanne 1994), p. 346; also published inDiamond a.nd Related Materials Vol. 3, Nos.4-6.
45. W. R. L. Lambrecht, B. Segall, M. Methfessel, and M. van Schilfgaarde, Phys. Rev.B 44, 3685 (1991).
46. W. R. 1. Lambrecht, B. Segall, W. Suttrop, M. Yonagathan, R. P. Devaty, W. J.Choyke, J. A. Edmond, J. A. Powell, and M. Alouani, Appl. Phys. Lett. 63, 2747(1993); and unpublished.
47. W. J. Schaffer, G. H. Negley, K. G. Irvine, and J. W. Palmour, in Diamond, Silicon Carbide and Nitride Wide-Bandgap Semicollductors, ed. C. H. Carter, Jr., G.Gildenblat, S. Nakamura, and R. J. Nemanich, Mater. Res. Soc. Symp. Proc. Vol. 339(1994), to be published.
48. W. R. L. Lambrecht and B. Segall, PI,ys. Rev. B 43 7070 (1991).49. W. R. L. Lambrecht and B. Segall, Acta metal. mater. 40, S17 (1992).50. W. R. L. Lambrecht and B. Segall, PllYs. Rev. B 41, 8353 (1991); ibid. 42, 1462(1990); W. R. 1. Lambrecht, C. H. Lee and B. Segall, in Atomic Scale Structure ofInterfaces, edited by R. D. Bringans, R. M. Feenstra and J. M. Gibson, Mater. Res.Soc. Symp. Proc., Vol. 159 (MRS, Pittsburgh 1990), p. 377-82; W. R. L. Lambrecht,C. H. Lee, M. Methfessel, M. van Schilfgaarde, C. Amador, B. Segall, in Defects inMatel'ials, Mater. Res. Soc. Symp. Proc.,Vol. 209, edited .by P. D. Bristowe, J. E.Epperson, J. E. Griffith and Z. Liliental-Weber, (MRS, Pittsburgh 1991), p. 667.
51. R. M. Wentzcovitch, K. J. Chang, and M. L. Cohen, Phys. rev. B 34, 1071 (1986).52. R. M. Wentzcovitch, and M. L. Cohen, l. PllYs. C 19, 6791 (1986).53. Landolt-Bornstein: Numerical Data and Functional Relationships in Science and
Tec1lllOlogy, New Series, Group III, Vol. 17a, edited by O. Madelung, (Springer, Berlin1982).
54. A. Munoz and K. I(unc, Pllysica B 185, 422, (1993).55. P. Perlin, C. Jaubert.llie-Carillon, J. P. Ithie, A. San Miguel, 1. Grzegory, and A.Polian, PII,Vs. Rev. B 45, 83 (1992).
56. H. Volstiidt, E. Ito, M. Akaishi, S. Akimoto, and O. Fukunara, Proc. lpn. Acad. B667(1991).
57. M. Ueno, M. Yoshida, A. Onodera, O. Shimomura, and K. Takemura, in Proc. 5thInt. Conf. High pressure in Semiconductor Pllysics, Kyoto (1992), lpn. J. Appl. PllYs.32, Suppl. 32-1,42 (1993).
58. W. R. L. Lambrecht and B. Segall, in Propel' ties of the Group-III Nitrides, editedby J. H. Edgar, EMIS Data Review Series, (lEE, Stevenage Herts, UK 1994), in press.
59. W. R. 1. Lambrecht, B. Segall, and O. K. Andersen, Phys. Rev. B 41, 2813 (1990).60. R. M. Martin, PllYs. Rev. B 6, 4546 (1972).61. K. Tsubouchi and N. Mikoshiba, IEEE Trans. Sonics Ultrason. SU-32, 634 (1985).62. V. A. Savastenko and A. U. Sheleg, Phys. Stat. Sol. (a) 48 K135 (1978); A. U.Sheleg and V. A. Savastenko, Inorg. Mater. 151257 (1979).
63. K. Kim, W. R. L. Lambrecht, and B. Segall, Plrys. Rev. B (1994), in press.64. V. Fiorentini, M. Methfessel, and M. Scheffler, Phys. Rev. B 47,13353 (1993).65. E. A. Albanesi, W. R. L. Lambrecht, and B. Segall, J. Vac. Sci. Tec1lI1oI. B (1994),in press (PCSI-21 Proceedings).
372
66. E. A. Albanesi, W. R. L. Lambrecht, and B. Segall, in Diamond, Silicon Carbideand Nitride Wide-Bandgap Semiconductors, ed. C. H. Carter, Jr., G. Gildenblat, S.Nakamura, and R. J. Nemanich, Mater. Res. Soc. Symp. Proc. Vol. 339 (1994), to bepublished.
67. G. Martin, S. Strit.e, A. Bot.chkarev, A. S. Agarwal, A. Rockett, W. R. L. Lambrecht,B. Segall, and H. Morko<;;, Appl. Phys. Lett., (1994), to be published.
68. J. Rife, W. R. Hunter, W. R. L. Lambrecht, B. Segall, and D. K. Wickenden, Bull.Am. PllYs. Soc, 39, 211 (1994); and nnpublished.
69. C. G. Olson, D. W. Lynch, and A. Zehe, Phys. Rev. B 24, 4629 (1981).70. S. Loughin, R. H. French, W. Y. Ching, Y. N. Xu, and G. A. Slack, Appl. Phys. Lett.63, 1182 (1993); and S. Loughin, Ph. D. Thesis, Univ. of Pennsylvania, Philadelphia1992.
71. E. A. Albanesi, W. R. L. Lambrecht, B. Segall, Phys. Rev. B 48, 17841 (1993).72. W. R. L. Lambrecht and B. Segall, Ph.ys. Rev. B 47 , 9289 (1993)73. W. R. L. Lambrecht, E. A. Albanesi, and B. Segall, in Proc. Internatiollal Con[erellce
011 Silicon Carbide and Related Materials, Washington DC, November 1993, in press.74. W. R. L. Lambrecht, C. Amador, and B. Segall, Phys. Rev. Lett. 68, 1363 (1992).75. J. W. D. Connolly and A. R. Williams, Phys. Rev. B 27,5169 (1983).7f}. A. Zangvil and R. Ruh, J. Am. Ceram. Soc. 71, 884 (1988).77. W. C. Bradley and A. P. Cracknell, The Mathematical theory o[Symmetry ill Solids:
Representation Theory [or Point Gl'OUpS and Space Groups (Clarendon Press, Oxford1972).
78. G. K. Safaraliev, Yu M. Tairov, and V. F. Tsvetkov, S'ov. Pliys. S'emicond. 25, 865(1991).
ION IMPLANTATION INTO WIDE BAND GAP SEMICONDUCTORS
VICTOR S.VAVILOVPNLebedev Physics Institute, 117924 Moscow, Russia
Abstract
There are discussed the major peculiarities which must be taken into account when theaccelerated ions are used for the modification of the properties of solids with a widebandgap (over -2 eV). Such solids are interesting for both the fundamental research andfor such applications as optoelectronics and high-temperature devices. The materialsconsidered below are usually called wide band gap semiconductors (yVS).
1. Introduction
Recently the author has published a review[l] in "Uspekhi Fizicheskikh Nauk" (TheProgress of Physical Sciences) with the aim to draw attention of the readers to theprocesses especially typical for WS, which have so far been explained onlyqualitatively. It is proper to mention that in the initial period of the development ofthe concepts of light absorption by solids as quantum phenomena, Pohl and Guddenhave used diamond and other WS to prove that the photoionization quantum yield isnear to unity. Somewhat later in the remarkable experiments of O. Losev in Russia[2],electroluminescence was discovered in silicon carbide SiC, as well as the generation andamplification of high frequency electrical currents. A high level of technology ofvacuum tubes as generators and amplifiers in twenties, as it seems, has detractedattention of physicists, technologists and governments from the possibilities of solidstate electronics, and silicon carbide and even more diamond were regarded as exoticobjects. In our days these two unique materials, especially in the form of epitaxialfilms have opened a new way both as superhard coatings for a variety of specialapplications. However, both the diamond and SiC, as well as other members of a widefamily of WS, such as cadmium and zinc sulfides behave as capricious, individualisticand unpredictable materials, quite hard to handle technologically.During the first years after the invention of transistor by Bardeen, Brattain and
Shockley, as I well remember from personal experience, the unpredictable behavior wasoften typical for single crystals of undoped and doped Ge and Si single crystals. Theearly stage of transistor electronics has been brilliantly described by H.Queisser in hisbook "Kristallene Krisen" [3] still not existing in Russian translation.
373
M.A. Pre/as et al. (eds.), Wide Band Gap Electronic Materials, 373-376© 1995 Kluwer Academic Publishers.
374
2. The ion-beam implantation of impurities
In optoelectronic and other devices based on the use of WS at present and later on theprocesses of electrical injection must be used. During these processes thenonequilibrium charge carriers are generated as a result of transitions through p-njunctions, heterojunctions or Schottky barriers, Le. semiconductor-metal contacts. Thestructures exhibiting electrical injection with nearly ideal properties have been analyzedby many authors. Unfortunately, at present in the majority of WS, except GaP andSiC, the interfaces used in the structures to produce the injection always contain insideand near the interface numerous centers of the carriers localization. An additional,technically harmful peculiarity is due to practical absence of effective donors andacceptors with sufficiently small ionization energies. As a result, at room temperaturethe majority of these centers do not supply carriers into the conductance and valencebands. Thus, in semiconducting diamond the main "shallow" acceptor impurity, i.e.boron B has the ionization energy of 0.37 eV and for the donor impurities, according totheoretical estimates must be close to -0.15 eV for the interstitial Li, -0.2 eV for Na,and 0.1 eV for P.A technical problem of WS doping by impurities with shallow energy levels is still
far from the solution; however, intensive efforts are evident [5, 8, 12].A real success has been mentioned above for GaP and, to some extent, to SiC.
However, the thermal diffusion of B and AI into SiC is effective only at temperaturesabout 2000 C. The WS of A2B6 family have excellent luminescent parameters, andsome of them can be rather easily grown from a vapor phase. However, their stabilityseems to be far below that of Si, Ge and diamond.At early stages of semiconductor technology a highly nonequilibrium way of the
properties modification was suggested by Shockley in the USA [4] and, independently,used in Moscow and Leningrad [5]. This technique is now called ion implantation.During a rather long period most technologists have been severely worried by inevitablecomplications due to the radiation-induced damage surrounding the location ofimplanted atom. Now it is well known that an adequate annealing of such materials asGe and Si, particularly, pulsed local annealing fIrst developed in Russia, has been quitesuccessful. Presently several thousand specially designed implanter accelerators areconstantly used in technically developed countries. In a long chain of operations usedin the production of integrated circuits (IC) based on silicon, the ion implantation hasbecome indispensable. This example is of great importance, as in the beginning, evenin early 70-ies, the use of ion implantation seemed to be too romantic and expensive.Figure 1 presents the data on the ranges and straggling of various ions in silicon [6].Qualitative data for various WS must be similar to it. There exist dependable tableswhich permit one to predict with a 10% error the results of the implantationexperiments [7].Thus, for an optimist there exists a direct way to modify the properties of WS,
limited only by the ions ranges. For most types of technical equipment the ion energydoes not exceed -500 keY, that is sufficient for silicon IC production [8]. Second, avery important limit is due to the time necessary for the desired fluence that, in manyinteresting cases, requires a prolonged running of the implanter, especially in the caseswhen one has to produce a new compound or to build a "buried layer" inside thesample.
375
1
fO-2:--.;;..:",.-=::_:-_---------~~
li""1
Mass50 100o
................... _ ------L......,;L-1...,-'"',-'---'-:--'-.l-....JL.-J.--1..,-'---l--.....-_-..i..-.::L=< 10-3
H Be At At" CaGa Nb Tn Xe 'CdHe 0 P Tio As Ag Sb
Figure 1. The dependence of projected ranges Rp (solid lines) and of straggling~Rp (broken lines) in nonoriented crystals of Si on the mass and energy ofimplanted ions [6].
Thus, in the experiments conducted by the author and his colleagues, for thesynthesisof SiC in diamond or silicon, the necessary fluence was achieved only after some hours.Nevertheless, ion implantation as the way of precision doping of WS by electrically
active impurities, by light emitting centers or by centers that interact actively with"genetic" defects and uncontrolled chemical impurities has become probably the mosteasy and predictable way to control the properties of WS, and, of course, of othermaterials [8].
3. General properties of centers with deep levels in WS (CDL)
The wide band gap semiconductors, and in the first case, the diamond and SiC presentthe most interesting object of theoretical and experimental research of the centers of
376
strong localization of the charge carriers. Already at the end of 50-ies E.WJ. Mitchellpresented a highly valuable data on the "color centers" in diamond [9] that were obtainedat Reading University in England, mostly by optical and luminescence methods. Lateron the microscopic structure of these centers, first of all their geometry and their nature,was identified by the use of electron paramagnetic resonance (EPR). In fundamentalmonographs, particularly in [10], edited by J.Field in 1992, often called a "Bible ofDiamond", and in the papers of Russian authors [11] numerous data on CDL indiamond are presented. According to the author's opinion, the data on other WS havebeen generally of more qualitative nature [12].To conclude my remarks, one should draw attention of those engaged in this field,
that, along with the romanticism and interesting details, this field presents the maindifficulty to be overcome ifWS are going to become a really valuable and controllablematerial of future. Lately, the extremely important achievements seem to be reached inthe technology of 6H and 4H SIC single crystals that makes them the major materialfor the high-temperature and radiation-resistant devices [13]. In my opinion, SiC anddiamond, first of all the diamond films, present a special aim for applied research, andthe technology of their growth and doping deserves a substantial support.
4. References
1. Vavilov, V.S. (1994) Peculiarities of wide band gap semiconductors, Sov.Phys. Uspekhi 164, 287-296.
2. Lossev, a.v. (1923), Telegraphy and Telephones 18, 61.Lossev, a.v. (1928), Phil.Mag. 7, 1024.
3. Queisser, H.(1985) ,Kristallene Krisen, Munchen, Piper, p.350.4. Shockley, W. US Patent No 2-787564, Class 148-1,5-1954.5. Vavilov, V.S. (1985), Some physical aspects of ion implantation, SOV145(2), 329-346.
6. Shulz, M. (1974) J.Appl.Phys. 4, 91.7. Burenkov, A.F. et. al. (1980) Tables of special redistribution of ion implantatedimpurities, Belarus St. Univ. edition, Minsk.
8. Ion beam modification of materials IBMM92 Heidelberg, Proc. of Int. Conf.Ed. by S.Kalbitzer. a.Meyer, G.K.Wolf, North Holland, 1993, p.1538 (2volumes).
9. E.W.J.Mitchell, In: "Physics and Chemistry of Solids", Ed. Pergamon Press,New York, vol.8, 1959, pp.444-449.
10. The properties of natural and synthetic diamond, EdJ.Field, Academic Press1992, London, p.677.
11. Gippius, A.A., and Vavilov, V. S. (1985) Proc. 7th Int. Conf. on ionimplantation in semiconductor, Vilnus. Belarus State Univ. Edition, p.59.
12. Goorgobiani, A.N. and Scheiynin, M.K. (1985), Physics ofA2B6Compounds, Eds.,Moscow, Nauka.
13. J.L.Lewis, Cree Research Inc., 6H Silicon Carbide Data Sheet4H SIlicon Carbide Data SheetEffective 8. 1993
2810 Meridian Parkway, Durham NC 27713, USA; Fax (919) 361-4630.
THERMODYNAMIC PROPERTIES OF BORON NITRIDE
V.L.Solozhenko* and K.S.Gavrichev*** BakuI' Institute for Superhard Materials, Ukrainian AcademyofSciences, Avtozavodskaya ul. 2, Kiev, 254153 Ukraine.** Kurnakov Institute ofGeneral and Inorganic Chemistry,Russian Academy ofSciences, Leninskii pro 31, Moscow, 117907Russia.
I. Introduction
Boron nitride occupies a prominent place among the most important inorganic materialsand serves as a basis for many advanced technologies. Its versatility is due to a widerange of properties presented by four polymorphic modifications of BN: two graphitelike (hexagonal and rhombohedral) and two close-packed (cubic and wurtzite).
Over the past several years, there has been a stable tendency towards a rise in the production of materials based on boron nitride. Given that the most attention is paid to hightechnology products, the development ofscientific foundations is vital for producing materials possessing a given combination of physicochemical properties. This, in tum, implies the necessity of reliable data on the thermodynamic properties of BN.
In this paper, we present a brief review of the calorimetric studies dealing with differentcrystalline forms of boron nitride.
2. Structures of the Polymorphic Forms ofBoron Nitride
The structure of hexagonal graphite-like BN (hBN) is made up of parallel layers of hexagonal rings where each atom of one sort has three nearest-neighbor atoms of the othersort. The rings from different layers are arranged strictly above one another in such a waythat boron and nitrogen atoms alternate along the threefold axis. Correspondingly, thecrystal structure of hBN is represented by the stacking sequence A'AA'A. The in-planeB--N distance is 0.1446 om, whereas the interlayerdistance is 0.33306 nm [I]. The identity period along the c axis is twice the layer thickness, c =0.66612 om; the in-plane parameter a =0.2504 nm. The X-ray density ofhBN is 2.279 g1cm3.
The structure of rhombohedral BN (rBN) is characterized by the stacking sequence AB
377
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 377-392© 1995 Kluwer Academic Publishers.
378
CABC, so that the identity period along the c axis is three times the layer thickness: c =1.0000(2) nm [2]. As in hBN, the in-plane boron--nitrogen distance is 0.1446 nm and a= 0.2504 nm. The X-ray density of rBN is 2.276 g/cm3•
The cubic form ofboron nitride (cBN) has the sphalerite-type structure. The in-plane parameter a =0.36160(3) nm [3] and the B--N interatomic distance is 0.1565 nm. The Xray density of cBN is 3.486 g/cm3.
The wurtzite form ofBN (wBN) has a hexagonal wurtzite-type structure where both theB--N interatomic distance and the X-ray density are nearly identical to those of cBN(0.1568 nm and 3.470 g/cm3, respectively). The lattice parameters of wBN are a =0.2550(2) and c = 0.4213(1) nm [4].
One more distinct form is boron nitride with a turbostratic (fully orientation-disordered)structure (tEN). Random in-plane translations or rotations of layers, accompanied by disordering of graphite-like forms of BN, were revealed in [5] and were called turbostraticstacking faults. These defects enhance the interlayer spacing d002' which amounts to0.356 nm at a three-dimensional order parameter (P3) close to zero.
3. Low-temperature Heat Capacity of Boron Nitride
hBN. The low-temperature heat capacity of hBN was studied in [6 - 9]. Cp(T) measurements by adiabatic calorimetry in the range 20 - 300 K were reported in [6]. The sampleof hBN contained 0.15 wt% iron oxide and boric anhydride. Later, the heat capacity ofhBN was measured in [7]; however, the sample studied (97.2% purity) contained muchlarger amounts of impurities, and the experimental data showed a large scatter (3 - 10%).Cp(T) data for hBN from 7.5 to 305 K were obtained in [8] in an adiabatic calorimeter.The content of boron oxide in the sample did not exceed 0.15 wt%. In calculating thethermodynamic properties, the influence of impurities was taken into account. A comparison of these three measurements shows agreement within a few percent. A featurepresent in all of the experimental curves is worth noting: in the range 60 - 140 K, thereis a very weak hump (Fig. 1). The heat capacity of graphite shows no similar anomaliesin this temperature range [10]. In order to elucidate the origin of this anomaly and to assess the thermodynamic properties of highly ordered hBN, the low-temperature heat capacity ofa sample with P3 =0.98 was measured in [9]. The results showed that the abovementioned anomaly in heat capacity in the range 60 - 140 K is absent for ordered hBN,whereas below and above this temperature range, there is close agreement between bothcurves (Fig. 1).
rBN. The rhombohedral form of boron nitride occurs mainly as a minority phase in thegraphite-like hexagonal form, and, as a consequence, its properties are not studied in sufficient detail. Recently, a technique has been proposed in [11] to produce rBN by reactionof boron with ammonia at high pressures and temperatures with a yield of the main phase
379
10
----0E 8
~'--"...........J
6
0...U
4
2
00 50
h/.
//
/
//
/./.
b
100 150
2
200
Fig. 1. Heat capacity as a function of temperaturefor graphite-like hBN in the region of flat anomaly:data from (1) [8) and (2) [9)
25.,-----------------------,
T,K
21
300
543
5
o+-~~..,...,==-.___--.___--.___--..___-___..__-1o 100 200
Fig.2.Heat capacity as a function of temperaturefor the four polymorphs of BN and tBN: (1) cBN,(2) wBN. (3) hBN, (4) rBN. and (5) tBN.
-----0E 20
~'-"
"--:>15
a..U
10
380
as high as 98 wt%. The low-temperature heat capacity of rBN thus prepared was measured by adiabatic calorimetry in [12]. The sample with P3 =0.93 contained about 2 wt%hBN and 0.8 wt% of impurities such as boron oxide, boron carbide, and magnesium nitride. In the calculation of the thermodynamic quantities, appropriate corrections for thepresence of impurities were introduced. No anomaly similar to that in the heat capacityof hBN was detected for rBN.
cBN. The low-temperature heat capacity of cubic boron nitride was measured in [13 17]. A sample of cBN studied in [13] contained up to 4 wt% of impurities, whereas samples studied in [14 - 17] were 99.8 to 99.9%-pure (although, the contents of carbon andoxygen were not specified in [17]). Moreover, the technique used in heat-capacity measurements in [13] did not ensure sufficient accuracy. It should be noted that the authorsof [17] report thermodynamic properties only for a standard state without giving the experimental and smoothed dependences in the temperature range examined. The values at298 K, reported in [14 - 17], agree within 1%, whereas the values given in [13] are substantially lower (by tens percent). Therefore, in determining the thermodynamic properties of cBN, one should consider the data reported in [14 - 17] as more reliable. Note thatthe authors of [14, 16] observed poorly defined flat anomalies in heat capacity below 25K, which were attributed to the impurity effect. The cumulative contribution from theseanomalies to the changes in enthalpy and entropy did not exceed 0.003 and 0.05% of thecorresponding standard values.
wBN. The low-temperature heat capacity of the wurtzite form of boron nitride was measured in [2,8, 18]. The sample studied in [18] had a purity as low as 92%, and the scatterin the heat capacity measurements was up to 5%. Adiabatic calorimetry measurementsof the heat capacity of a 99.7%-pure wBN sample in the range 6 - 307 K were reportedin [8]. The Cp(T) dependence for wBN showed a weak anomaly peaked at 21 K. A studyreported in [2] dealt with two samples of the wurtzite form ofboron nitride, one of themsubjected to a special treatment to remove structural defects. After this heat treatment, aweak anomaly in heat capacity centered at 16 K [LUI = 0.69 J/mol, ~s = 0.042J/(mol K)),which was observed for the initial wBN sample, did not appear. It was assumed [2] thatthis anomaly is associated with the presence of ordered point defects in the structure ofthe initial sample. A comparison of the available data shows that the values reported in[18] are substantially higher (by as much as 25%) than those reported in [2,8] and eventhan the corresponding values for hBN, which indicates there low reliability.
tBN. The low-temperature heat capacity of turbostratic boron nitride was measured in[9]. The sample studied contained carbon and oxygen (0.1 wt% each) and under 0.02wt% of other impurities. At T < 300 K, the heat capacity of tBN is higher than that ofhBN and rBN (Fig. 2).
Numerous measurements of heat capacity for different polymorphs of boron nitride havedemonstrated the extent to which the structural perfection of the sample, its phase composition, and purity are critical. The presence of anomalies associated with these factors
381
may substantially alter the thermodynamic properties of a sample. In our opinion, the origin of the flat anomaly in the Cp(l) dependence of hBN is still not thoroughly understood. As shown in [9], it is insensitive to the particle size in the sample under study orto the presence of intercalated impurities. It was assumed [9] that this anomaly resultsfrom partial layer disordering, which gives rise to interlayer vibrations at a frequency of-140 cm- l ,Another possible hypothesis is thathBN contains a small amount (3 to 5 wt%,as estimated from heat capacity data) of defective material in the form of rolled up structures (something like fullerens or tubulens) whose translational vibrations give rise to theflat anomaly in heat capacity in the temperature range specified (Fig. 3).
4. High-temperature Heat Capacity of Boron Nitride
hBN. The heat capacity of graphite-like hexagonal BN in the range 1300 - 2000 K wasstudied in [19] by comparing the cooling rates of hBN and graphite; however, given theinsufficient accuracy of this technique, the results can be considered only as rough estimates.
cBN. The true heat capacity of cubic boron nitride single crystals in the temperaturerange 300 - 1100 K was measured in [20] by scanning adiabatic calorimetry in a continuous heating mode. The data points were fitted with the following equation:
wBN. The heat capacity of the wurtzite form of boron nitride, obtained by thermal stabilization at 1300 K, was measured in the range 420 - 980 K by differential scanning calorimetry during stepwise heating. The data points were fitted with the followingequation:
The experimental data on the high-temperature heat capacity of the dense modificationsof boron nitride are presented in Fig. 4.
5. High-temperature Enthalpy of Boron Nitride
hBN. The high-temperature enthalpy of the graphite-like hexagonal form was measuredin [21,22] by drop calorimetry; however, no data on the degree of three-dimensional order in the hBN sample studied were reported in these works. The enthalpy of highly ordered hBN (P3 = 0.98), measured by inverse drop calorimetry from 300 to 1700 K, wasreported in [23]. rBN. The only measurement of the enthalpy of rBN was reported in[23]. The investigation in the range 300 - 1700 K by inverse drop calorimetry showedthat the high-temperature thermodynamic properties of the two graphite-like forms of
382
12
,.......0 10E~'--''-.... 8J
a. ~
() 6 "/"/
//
14 // 2·
"/~
2
o-J------..------,----.-------,----jo 50 100 150 200 250
Fig.3. (1) Heat capacity as a function of temperature T, Kfar pure, highly ordered hBN and. (2) for analogoussample containing 4 wt:5ll of admixture.
~
0 450
E~'-"~J
35.0D-
U
250+ cBN [20J
o wBN [2J
15 0 +rrTrrTTTTTTrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrTTTCrrnro1250 450 650 850 1Q50 1250
F!g.4. Data points and smoothed curves of thehIgh-temperature heat capacit.{' for dense BNpolymorphs: (1) cBN [20]. (2) wBN £21.
383
BN are nearly identical.
cBN. The high-temperature enthalpy of cubic boron nitride was reported in [22,24 - 27]for polycrystalIine samples and in [27,28] for single crystals. The measurements reported in [22,24,25] were carried out on sintered polycrystalline samples that contained highcontents (-4%) of unidentified impurities and noticeable amounts of hBN; therefore,these data should be considered only as an approximate assessment.
wBN. The enthalpy of the wurtzite form of boron nitride was measured in the range 300- 1200 K by drop calorimetry [22] and in the range 300 - 1300 K by inverse drop calorimetry [27,29]. Given that wBN studied in the latter two works was thermally stabilizedand contained a low level of impurities, these results appear to be more accurate. Themost reliable experimental data on the enthalpy of the graphite-like forms and wBN aregiven in Fig. 5; those for cBN are shown in Fig. 6.
6. Temperature Dependences of Thermodynamic Functions of Boron Nitride
Based on comparative analysis of the available data on the low-temperature heat capacities of the cubic, wurtzite-like, and rhombohedral forms, as well as on the heat capacitiesof the ordered and fully orientation-disordered graphite-like hexagonal forms, we propose the values of the thermodynamic functions, listed in Table 1, as the best standardreference data for the crystalline forms of boron nitride.
In processing the most reliable experimental data on enthalpy, we followed Shomate[31]; to calculate Shomate's function, we used the standard heat capacities of BN polymorphs, determined by adiabatic calorimetry (Table 1).
Shomate's function that ensures the best fit of the experimental variation of enthalpy inthe temperature range examined is given by:
W(T) -W(298.15 K) =aT + bT2 + cT3 +dr + 3eT-I + f, 1. marl (3)
The coefficients of (3) for the four polymOlphs ofboron nitride are listed in Table 2.
The thermodynamic functions of the four polymorphs of BN, calculated from these data(Tables 3 - 6), can be recommended as reference data.
7. Enthalpies of Formation of Boron Nitride Polymorphs
The determination of boron nitride polymorphs enthalpies of formation cannot be performed by the usual methods owing to chemical passivity of these substances and that iswhy the method of fluorine calorimetry was used to solve this problem.
384
60.0 ...,-------------------,-
oE"-...J
.::L 40.0
co:Ol·N20.0
'. hBNo rBNA wBN
7CO i 100 .1500T,K
Fig.5.Data points and smoothed curves of theenthalpy for the grafite-like and wurtziteboron nitride: (1) hBN, (2) rBN, (3) wBN.
bU.U ...,-------------------,
oE
"-...J
.::L 40.0
co·Ol'C\! 20.0
o 1o 2A 3• 4
1800T,K
13008000.0 -f<T........-r....,..--,.--,,.--,,.....,-,.....,-,.....,-,.....,-,.....,-,,rrl300
Fig.6.Data points and smoothed curves of the enthalpyfor (1,2) sln<;tle-crystal and (3(4) palycrystalline cBN:data from (1) [28]. (2) [27J. 3) [26] • ~4) [21] .
385
Enthalpy of formation of hexagonal graphite-like boron nitride modification was adduced in [32,33], in which the close values (250.5 ± 1.5 and 250.6 ± 2.1 kJ mor l , respectively) was obtained. Enthalpy of formation of wurtzitic boron nitride was studied in[33]. Obtained in [33] value (.1fHO(BN, w, 298.15 K) = 255.8 ± 2.0) was recalculated in[34] taking into account the content of admixtures and became equal 263.6±2.3 kJ mor l .
Enthalpy of formation of cubic boron nitride was investigated in [34]. The comparisonof obtained value (266.8 ± 2.2) with one for wBN shown that these quantities are veryclose.
8. Phase Diagram of Boron Nitride
Phase p,T-diagram of boron nitride have been constructed by Corrigan and Bundy [35]from the data on cathalytic transformation of hexagonal phase into cubic one, and takinginto account the analogy of carbon and boron nitride phase diagram. According withthere diagram the graphite-like hexagonal modification of boron nitride is stable at normal pressure. But in [36] the thermodynamic aspect of transformation of graphite-likemodification to cubic was examined and was shown that at normal conditions in the temperature range 0-1570 K the cubic boron nitride is stable. In [2] the detail thermodynamicanalysis of boron nitride polymorphism was carried out and new phase p,T-diagram ofboron nitride was constructed (Fig. 7).
12 p, GPo
10
8
II
6/ / \
cBN / \ L/ / \
/ \/ / I
6
4
2
././
//
/
hBN
4000300020001000O-l---------.---L..---..--- -+-_'-- --.o
Fig. 7. The p - T phose diagram of boron nitride; T,Kfrom [35] -dashed lines and [2J -solid lines.
386
TABLES
Table 1. Thermodynamic properties of the BN polymorphs and tBN
at standard conditions.
C~(298.15 K), 5°(298.15 K), I HO(298.15 K)-Ho(O',
1-----------------------------------1--------------------J K- 1mol- 1 J mol- 1
hBN I 19.85±0.06 14.80±0.10 2632±25------------------------------------------------------------------rBN I 20.63±0.06 15.83±0.10 2805±25------------------------------------------------------------------tBN I 21.58±0.06 17.54±0.10 2963±25------------------------------------------------------------------cBN I 15.95±0.10 6.71±0.06 1457±15------------------------------------------------------------------wBN I 16.45±0.05 7.34±0.05 1541±15
Table 2. Coefficients of equation (3) used to fit the temperatu~e
dependent enthalpy of the four polymorphic forms of boron nitrice.
--------------~-------~-~~2-1--~-~~5-1--~-~~9-1-~-~~:5-I---~------
hBN[23]
rBN[23]
I 2.0579 I 4.1285 1-1.6977 i 2.7542 I 2.2932 / -4621.SC5
/-3.4955 I 4.7683 1-2.0330 i 3.4232 I 0.6754 1 -2455.:£9
cBN[27,28] I 6.2507 1 4.0024 1-1.9353 I 3.7362 I 8.4850 / -7781.E~7
wBN[29] /-17.992 I 7.9652 1-4.8917 I 11.823 I 1.1095 I -882.693
387
Table 3. Thermodynamic properties of hBN.
---;-------~~(;)----------;O(;)----------;O(;)----I-~O(;)-=-~O(~)-
---;-------------------;-;:1~~~:1----------------------;-~~~:1----
16202530354045506070809010011012013014015016017018019020021022023024025026027028029030040050060070080090010001100120013001400150016001700
0.16030.25010.39140.56320.76350.99061.2421.5172.12942.8133.5564.2214.9005.6026.3237.0607.8098.5679.33210.1010.8811.6512.4313.2013.9714.7415.5016.2617.0217.7718.5119.2519.9927.2532.4636.1938.9340.9942.5943.8644.8845.7246.4247.0147.5247.9648.34
0.05300.09810.16850.25450.35600.47240.60330.74821.0771.4561.8792.3362.8153.3153.8334.3684.9185.4836.0606.6497.2487.8578.4749.0999.73110.3711. 0111.6612.3212.9713.6314.2914.9621.7228.3934.6540.4545.7950.7155.2759.8063.4467.1370.5973.8576.9379.85
0.01290.02490.04580.07280.10550.14390.18700.23550.3:4 630.47610.62360.78780.96601.1561.3571.5681.7882.0152.2492.4912.7382.9923.2503.5143.7824.0544.3314.6114.8945.1815.4715.7646.0639.09712.2915.5018.6621.7224.6727.5130.2332.8335.3337.7340.0342.2444.37
0.6411. 4623.0665.4528.76913.1518.7425.6343.8668.57100.4139.3184.9237.4297.0363.9438.3520.2609.7706.8811.7924.41045117213081452160'317621929210322842473266950488048114901525019250234302776032200367304134046010507405551060330
388
Table 4. Theraodynarnic functions of rBN.
T I C~(T) SeCT) $o(T) I HO(T) - HO(O)---~--I----------------;-~:i~~~:i----------------------;-~~~:i----
16202530354045506070809010011012013014015016017018019020021022023024025026027028029030040050060070080090010001100120013001400150016001700
0.14710.25180.H'J80.62010.85161.1121.3951.7012.3613.0703.8074.5465.2666.0006.7507.5148.2889.0709.85810.6511.4412.2413.0413.8314.6215.4416.1916.9717.7418.5119.2720.0320.7827.4932.3835.9438.6140.6642.2943.6044.6745.5746.3J46.9847.5448.0348.46
0.04900.09260.16600.25960.37220.50260.64950.81211.1791.5962.0542.5453.0613.5974.1514.7225.3075.9056.5167.1377.7688.4089.0569.71210.3711.0411.7112.3713.0713.7614.4415.1315.8322.7629.4535.6741. 4446.7351.6256.1560.3564.2867.9671.4274.6877.7680.69
0.01220.02330.04340.07090.10520.14620.19350.24640.37060.51510.67820.85791.0521.2591.4771.7041.9412.1852.4362.6952.9593.2293.5043.7844.0684.3574.6494.9455.2455.5475.8536.1616.4719.68312.9716.2519.4422.5325.4928.3331.0633.6636.1638.5640.8643.0745.20
0.5881.3863.0655.6659.34414.2520.5228.2648.5475.67110.0151.8200.9257.2320.9392.4471.3558.1652.7755.3865.7984.11110124413871537169518612035221624042601280552328240116701540019370235202781032230367404134046000507305551060330
Table 5. Thermodynamic functions of cBN.
389
T
4681015202530354045506070809010011012013014015016017018019020022024026028030040050060070080090010001100120013001400150016001700
C~(T)
0.000050.000650.002280.003250.005630.007820.01580.02150.03060.04560.06440.08730.14870.24120.37490.55780.80211.1041. 4741.9132.4152.9773.6034.2915.0345.8236.6558.42510.2912.1914.0915.9724.4230.3834.5337.4339.5041. 0042.1142.9643.6144.1344.5444.8745.1445.37
0.0000250.0001010.0005270.001140.002890.004920.007360.010800.01470.01970.02610.03410.05500.08440.12490.17900.25000.34010.45150.58640.74610.93151. 1431.3821. 64 81.9412.2612.9773.7904.6885.6616.69712.5418.6624.5930.1535.2940.0344.4148.4652.2355.7459.0362.1165.0267.76
0.0000030.0000180.0000830.0002240.0008230.001600.002480.003560.004870.006360.008170.010400.015910.023500.033540.046670.063340.084200.11000.14130.17860.22220.27300.33140.39690.46990.55200.73880.95921.2111. 4931.8043.7426.1148.70411.3814.0516.6719.2321.7124.1026.4028.6130.7432.8034.77
0.000090.000500.003550.009160.03100.06640.12210.21730.34420.53350.80691.1852.3444.2627.31111. 9218.6728.1540.9857.8679.45106.4139.2178.6225.2279.5341. 8492.4679.4904.1116714683517627595331314016990210202518029430337603815042590470605156056080
390
Table 6. Thermodynamic functions of wBN.
T
610152025303540455060708090100110120130140150160170180190200210220230.24025026027028029030040050060070080090010001100120013001400
C~(T)
0.00070.00650.02780.07200.07230.07500.08870.10610.13030.16320.25490.38730.55250.75951. 0191.3401.7312.1952.7293.3324.0054.7265.4906.3137.1698.0558.9669.89610.8511.8112.7813.7514.7215.6816.6324.5630.3834.4837.3639.4440.9742.1343.0243.7144.2744.72
0.000230.001410.006420.02080.03820.05140.06390.07700.09080.10620.14380.19310.25520.33170.42460.53630.66910.82551.0071.2161. 4521. 7162.0082.3262.6723.0433.4393.8584.2994.7615.2435.7446.2616.7947.34213.2519.4025.3230.8636.0040.7J45.1149.1752.9556.4759.77
0.000060.000210.001200.004400.009500.015400.021700.027810.034040.040390.054430.070480.089610.11210.13860.16950.20540.24700.29460.34890.41040.47930.55600.64060.73340.83450.94381.0611.1871. 3211.4621.6111.7681. 9322.1034.1426.5809.21411.9214.6117.2519.8222.3124.7127.0229.24
0.00110.01200.07760.3270.7151.0781. 4851.9712.5593.2895.3668.58313.2519.7628.6140.3555.6475.2299.78130.0166.7210.3261.3320.3387.7463.8548.8643.1746.8860.1983.111161258141015723644640896631326017110211302529029550338903829042740
391
References
1. Pease, R.S., (1952) An X-ray Study of Boron Nitride, Acta Crystallogr., 5, 356 - 361.2. Solozhenko, V.L., Thermodynamic Aspect of Polymorphism of Boron Nitride, Dr. Sci. (Chern.) Dissertation, Moscow: Moscow State Univ., 1993.
3. Solozhenko, V.L., Chernyshev, V.V., Fetisov, G.V., Rybakov, V.B., and Petrusha, I.A., (1990) StructureAnalysis of the Cubic Boron Nitride Crystals, J. Phys. Chern. Solids, 1990, 51, 1011 - 1012.
4. Solozhenko, V.L., Kurdyumov, A.V., Petrusha, I.A., and Zelyavsky, W.B., (1993) Thermal Phase Stabilization of Wurtzite Boron Nitride, J. Hard Mater., 4, 107 - Ill.
5. Tomas, J., Weston, N.E., and O'Connor, T.E., (1963) Turbostratic Boron Nitride, Thermal Transformationto Ordered-Layer-Lattice Boron Nitride, J. Am. Chern. Soc., 84, no. 24,4619 - 4622 .
6. Dworkin, A.S., Sasmor, D.J., and Van Artsdalen, J., (1954) The Thermodynamics of Boron Nitride, LowTemperature Heat Capacity and Entropy; Heats of Combustion and Fonnation, J. Chern. Phys., 22, 837 842.
7. Sirota, N.N., Kofman N.A., and Petrova, Zh. K., (1975) Temperature Dependence of Heat Capacity for Hexagonal Boron Nitride in the Range 5 - 300 K, Izv. Akad. Nank BSSR, Ser. Fiz. -Mat. Nauk, no. 6,75 -78 .
8. Gorbunov, V.E., Gavrichev, K.S., Totrova, G.A., Bochko, A.A., and Lazarev, V.B., (1988) ThermodynamicProperties of lX- and y-Fonns of Boron Nitride at Low Temperatures, Zh. Fiz. Khim., 62, no. I, 18 - 24.
9. Gavrichev, K.S., Solozhenko, V.L., Gorbunov, V.E., Golushina, L.N., Totrova, G.A., and Lazarev, V.B.,(1993) Low Temperature Heat Capacity and Thermodynamic Properties of Four Boron Nitride Modif ications, Thermochim. Acta, 217, 77 - 89.
10. De Sorbo, W. and Tyler, W.W., (1953) The Specific Heat of Graphite from 13 to 300 K, J. Chern. Phys.,21, no. 10, 1660 - 1663.
11. Solozhenko, V.L., Mukhanov, V.A., and Novikov, N.V., (1990) Reaction of Ammonia and Nitrogen withBoron and Its Compounds at High Pressures and Temperatures, Dokl. Akad. Nauk SSSR, 312, no. 3,630633.
12. Gavrichev, K.S., Gorbunov, V.E., Solozhenko, V.L., Golushina, L.N., and Totrova, G.A., (1992) Heat Capacity and Thermodynamic Functions of the Rhombohedral Form of Boron Nitride in the TemperatureRange 15 - 300 K, Zh. Fiz. Khim., 66, no. 10,28242828.
13. Sirota, N.N. and Kofman N.A., (1975) Temperature Dependence of the Heat Capacity of Cubic Boron Nitride in the Range 5 - 300 K, Dok!. Akad. Nauk, 225, no. 6,1316 - 1318.
14. Solozhenko, V.L., Yachmenev, V.E., Vil'kovskii, V.A., Sokolov, A.N., and Shul'zhenko, A.A., (1987) HeatCapacity and Thermodynamic Functions of Single Crystals of Cubic Boron Nitride in the Range 4 - 300 K,Zh. Fiz. Khim., 61, no. 10,2816 - 2818.
IS. Gorbunov, V.E., Gavrichev, K.S., Totrova, G.A., Bochko, A.A., and Lazarev, V.B., (1987) Thermodynamic Properties of ,t--BN at Low Temperatures, Zh. Fiz. Khim., 61, no. 12,3357 - 3360.
16. Solozhenko, V.L., Yachmenev, V.E., Vil'kovskii, V.A., and Petrusha, I.A., (1989) Heat Capacity and Thermodynamic Functions of Polycrystalline Cubic Boron Nitride in the Temperature Range 4 - 300 K, Izv.Akad. Nauk SSSR, Neorg. Mater., 25, no. 1,160-162.
17. Atake, T., Honda, A., Saito, Y., and Saito, K., (1990) Low-Temperature Heat Capacity of Cubic Boron Nitride, Jpn. J. Appl. Phys., 29, no. 10, 1869 - 1870.
18. Sirota, N.N. and Kofman N.A., (1976) Thermodynamic Functions in the Range 5 - 320 K for the WurtziteForm of Boron Nitride, Dok!. Akad Nauk SSSR, 230, no. 1,82 - 85.
19. Prophet, H. and Stull, D.R., (1963) Heat Capacities of Boron Nitride and Aluminium Oxide Using an ArcImage Furnace, J. Chern. Eng. Data, 8,78 - 81.
20. Lyustemik, V.E. and Solozhenko, V.L., (1992) Heat Capacity and Thermodynamic Functions in the Range300 - 1100 K for Single Crystals of the Sphalerite Form of Boron Nitride, ZIt. Fiz. Khim., 66, no. 5, 11861191.
21. McDonald, R.A., and Stull, D.R., (1961) The Heat Content and Heat Capacity of Boron Nitride from 298to 1689 K, J. Phys. Chern., 65 , no. 10, 1918.
22. Agoshkov, V.M. and Bogdanova, S.V., (1990) Thermodynamic Properties of Polymorphic Forms of BoronNitride in the Temperature Range from 298 to 1200 K, Sverkhtverd. Mater., no. 1,26 - 30.
392
23. Solozhenko. V.L.. (1993) Thennodynamic Properties of Graphite- Like Fonns of Boron Nitride in theRange 300 - 1700 K. Zh. Fiz. Khim.• 67. no. 8. 1580 - 1582 .
24. Mezaki. R.. Tilleux. E.W.• Barnes. D.W.• and Margrave. J.L.. (1962) High-Temperature ThennodynamicProperties of Some Refractory Borides. in Thennodynamics ofNuclearMaterials. Vienna: IAEA. pp. 775 788.
25. Kiseleva. I.A., Mel'chakova. L.V.• and Topor. N.D.• (1973) Experimental Measurement of the High-Ternperature Heat Capacity of .13-BN, Izv. Akad. Nauk SSSR. Neorg. Hater.• 9. no. 3, 494 - 495 .
26. Solozhenko. V.L.. Chaikovskaya. I.Ya.• and Petrusha. I.A.• (1989) Thennodynamic Properties of Polycrystalline Cubic Boron Nitride in the Temperature Range 300 - 1600K.. Izv. Akad. Nauk SSSR. Neorg. Mater.•25. no. 10, 1672 - 1675.
27. Solozhenko. V.L.. (1993) Thennodynamics of Dense Boron Nitride Modifications and a New Phase P. TDiagram for BN. Thennochim. Acta. 218. 221 - 227.
28. Solozhenko, V.L., Chaikovskaya. lYa., Sokolov, A.N., and Shul'zhenko. A.A., (1987) ThennodynamicProperties of Cubic Boron Nitride in the Range 298 - 930 K. Zh. Fiz. Khim.• 61. no. 3. 801 - 803.
29. Solozhenko. V.L., (1993) Inverse Drop-Calorimetry. A Study ofMetastable and Non-Equilibrium Phases,Thennochim. Acta. 218. 395 - 400.
30. Shomate, C.H.• (1944) High-Temperature Heat Contents of Magnesium Nitrite. Calcium Nitrite. and Barium Nitrite. J. Am. Chern. Soc.• 66. no. 6.928 - 929.
31. Glushko V.P. (ed), (1981) Thennodynamic Properties of Indivi- dual Substances. A Handbook. V .3,part1, Moscow, Nauka,
32. Leonidov, V.Ya., Timofeev. lV.• and Solozhenko, V.L.. (1988) Bomb Calorimetry of Reactions with Fluorine. Enthalpies of Graphite-like and Wurtzitic Boron Nitride. In Thennodynamics of chemical substances,Gorkii. pp. 16-18.
33. Leonidov, V.Ya., Timofeev. lV.• Lazarev. V.B., and Bochko. A.V., (1986) Enthalpy of Fonnation ofWurtzitic Boron Nitride. Proceedings of XI All-Union Conference on Calorimetry and Chemical Thennodynamics. Part 1. Novosibirsk. NlOKH SO AN SSSR, p.7-8.
34. Leonidov. V.Ya., Timofeev, 1V.• and Solo2henko. V.L., (1987) Enthalpy of Fonnation ofCubic Boron Nitride. Zh. Fiz. Khim.• 61. N 10.2851-2852.
35. Corrigan. P.R., and Bundy F.P. (1975) Phase Diagram of Boron Nitride, J. Chern. Phys.• 63, N 9, 38123820.
36. Solozhenko, V.L., Leonidov, V.Ya.• (1988) On Thennodynamics of Graphite-like to Cubic Boron NitrideFonn Transfonnation, Zh. Fiz. Khim., 62, NIl, 3145-3146.
ELECTRICAL CONDUCTIVITY OF CERAMICS BASED ON DIFFERENTBORON NITRIDE MODIFICATIONS
A.V.BOCHKOInstitute of Materials Science Problems, Ukraine Academy of Sciences,Krzhyzhanovsky str. 3, Kiev, 252680, UkraineG.A. SOKOLINA, S.V. BANTSEKOVInstitute of Physical Chemistry, Russian Academy of Sciences, LeninskyProspect 31, Moscow, 117915, Russia
1. Introduction
Boron nitride, along with diamond, is one of those rare high-ohmic materials which possesses high heat conductivity. The important peculiarity of boron nitride as compared tocarbon is that all its modifications: c-BN - cubic, w-BN - wurtzite and h-BN - hexagonal(graphite-like) - have high resistance in contrast with hexagonal modification of carbongraphite, which is a material with high conductivity[l,2]. The difficulties in obtainingthe .diamond ceramics are connected in a great measure with possible appearance ofgraphite on baking of diamond powders under high pressures and temperatures.
One can assume that BN ceramics produced by baking ofdiamond powders, which contains no conducting phases, will possess high resistance and heat conductivity, whichpermits to use it as heat sink.
2. Experimental
The ceramics was obtained by sintering at high temperatures and pressures of powderwurtzite modification of boron nitride, purified from all alloying impurities. After baking a part of w-BN transforms into c-BN, which gives as a result two-phase materialcomposed of wurtzite and cubic boron nitride (Hexanit-R), containing h-BN as impurity(1-3%).
For Hexanit-R samples were studied temperature dependencies of heat conductivityA(f), electrical conductivity G(T), dielectric loss tangent tg B(f) and dielectric permeability e(f). The temperature dependencies of conductivity were measured for HexanitR samples having 508 rom in diameter and 1-3 rom in height. The contacts were madeof aquadag or by sprayning of platinum. Prior to G(f) measurements volt-ampere characteristics of the samples I(V) were obtained. The data on G(T) were obtained for the
393
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 393-396© 1995 Kluwer Academic Publishers.
394
linear region of I(T).
The stationary conductivity of samples in the range of temperatures 300<T<1000 K wasmeasured by two-electrode circuit in the special vacuum set with radiation heating[3].The values of tg 5(T) and £(T) were measured in air using I MHz frequency. The phasecontent of Hexanit-R samples was determined by means ofX-ray diffracion. G(T) measurements were made for BN ceramic samples composed of different phases. The content of c-BN varied from 15 to 90%. and that ofw-BN changed in the range 10-80 %. Inall samples the content ofh-BN was 1-3 %.
3. Results and Discussion
Experiments showed. that conductivity of Hexanit-R ceramic samples. measured afterthe baking is influenced by preliminary heating in air at 450°C for 30 minutes. Fig.lshows a typical graph of 19 G(lff) dependence. measured in vacuum before and afterpreliminary heating in air.
6,OhM- 4'CM-1
10-10
fO- 13 2 '""""""----~
20)
Figure 1. Temperature dependence of Hexanlte-R conductivity prior (l) and after (2) preliminary heating in
air.
It follows from G(T) data obtained for Hexanit-R samples of various phase content thatthe changing of phase composition does not influence the magnitude of conductivity as
395
well as of activation energy E of Hexanit-R ceramics. For all of the samples E is 0.95+0.1 eV in the range 400<T<1000 K, and the magnitude of conductivity at 300 K is lessthan 10-14 Ohm-1cm-1.
g
6 -
4
2
400 600 800
i,8
i ,6
1,4
i,D
Figure 2. Temperature dependance of dielectric losses tangent and dielectric permeabilities ratio for the sam
ple of Hexanite-R containing 60 Wl.% c-BN.
Fig.2 presents the curves tg B(T)and £(T) for the sample containing 60% c-BN. Thevalue ofdielectric permeability measured at room temperature was =6.8.
Fig.3 shows temperature dependencies of heat conductivity for two Hexanit-R samples.One can see that the value of heat conductivity is practically the same for the samplesstudied.
Analysis of experimental data on measuring electrophysical and thermo-physical properties of Hexanit-R gives evidence that Hexanit-R ceramics can be a promising materialfor heat sinks in microradioelectronics.
396
p-)WjM'K
i50
iDa
50
1A__
1
__6
~-_---c:r;----'i----100---- ~
400 GOO 800Figure 3. Temperature dependence of thermal conductivity coefficient for w-BN - based ceramics as com
pared with that for c-BN: 1 - Hexanlt-R with 20 wt.% c-BN; 2 - Hexanlt-R with 100 wI.% c-BN.
References
l. Samsonov G.V. Nonmetallic nltrides.- Moscow: Metallurgy.- 1969.- 265 p.2. VL. PrImachuk, A.V.Bochko and A.O.Avetisyan. Thermophyslcal properties of various modifications ofboron nltrlde.- J.Sov. Powder Metall Ceramics, 1983, v.22, N8(248), p. 664-666.
3. Sokollna G.A., Ippolltov S.A., Bantsekov S.V., Fedoseev D.V. Temperature and frequency dependencies ofdiamond ceramics electrical conductivity, Nonorganlc Materials, v.15, w.3, p. 444-448, 1979.
CATHODOLUMINESCENT INVESTIGATION OFEXTERNAL FACTORS INFLUENCE ON DEFECTIVECUBIC BORON NITRIDE STRUCTURE
V.B.SHIPILO, E.M.SHISHONOK, A.I.LUKOMSKII,L.M.GAMEZAInstitute of Solid State and Semiconductor Physics, Academy ofSciences ofBelarus, 17 P.Brovka str., 220072 Minsk, Belarus
A cathodoluminescent (CL> method has been employed to study a defect stuctureof cBN monocrystals as well as possibilities of its purposeful modification bythermal treatments under pressure (in high-pressure apparatures) and in anitrogen atmosphere.Astudy has beenmade of transparent, light-yellow, flattened cBN monocrystals
synthesized in the Li-B-N-H system. Cathodoluminescencewas excited on the face(Ill) of monocrystals. An excitation energy amounted to 35 keV and a detectiontemperature was 77 and 300 K.From the treatment temperature dependence of the parameter of the zero
phonon line (ZPL) of the GC-2 center [l] of boron-vacancy nature it has beenestablished that changes of internal stresses in the cBN lattice at the temperaturesof up to tann :::::: 1270 K independently of a treatment pressure (P= 5-6 GPa) , a gaspressure (p = 10-3 GPa) of a treatment medium (atomic nitrogen), and ofstoichiometry are of oscillating character with peaks at definite temperatures(Fig. I ). This phenomenon is assosiated with various processes of defect formationon some sections of the annealing temperature curve. The influence of variousannealing conditions on cBN (a nitrogen atmosphere or pressure) is revealed whentann >700 DC and lies in the fact that, in the spectra of single crystals annealed in anitrogen atmosphere over a temperature range 700 -900 DC, while the C-bandincreases in intensity, the intensities of ZPL at 2.515 eV (the S center) and itsphonon wing (the band at 530 nm) reduce practically to complete disappearance(Fig.2, b, curve l). The center ZPL rises dramatically in intensity, narrowing attann = 1000 DC (Fig.2, b, curve 2). Such dependence of the ZPL intensity couldtestify, by analogy with diamond [2], to the formation in the indicated temperaturerange of the defects, annealed at tann = 1000 DC, onto which energy can betransferred from the excited S centers, for instance, that scattering nonradiatively.
397
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 397-400© 1995 Kluwer Academic Publishers.
398
a
:>~ 10
..cS-6
'r;l:l0l:l0 b GC-2.<:l0<
l:<)
N 10
5
200 600 1000 t °can.
Fig.l Half-width of GC-2and at 2.515 eV centers as a function of temperature forcBN singlecrystals annealed in nitrogen atmosphere (a); under pressure (b)
e~.;;3"2""._2.".- 2,62
\
\ ." \')\649f.! .,~/\.2\.._
•...•.... \\ GC~2
I-J~-~~--..•. ----~
oj.E...:Iu
800 600 400 ).. nm
Fig.2 CL spectra of initial cBN single crystals (a) annealed in nitrogen atmosphere att-900oC (lb); t= 10000C (2b); t> 1270 0C (c)
The costant half-width and the invariable relative ZPL intensity of S center inthe spectra of samples annealed under pressure with tann < 900°C (Fig.l,b)evidence that the above-mentioned defects are not formed after the indicatedtreatment.A simultaneous absence of the C-band from the spectra (Fig.3,a) points to a
possible connection of these defects with the luminescence centers inducing thisC-band and, by analogy with diamond [2], to their geometric closeness.
399
.£;g~.5....1u
800 600 1+00 A. nmFig.3 CL spectra of cBN single crystals annealed under pressure at t = 900 °c (a) and t >1000 °C(b) and in vacuum at t > 1270 °C(c)
Earlier [3] we assigned the existence of the C-band in the CL spectra to thepresense of dislocations in the material, however, the nature of the above-mentinedluminescence center was not discussed. These can be free broken bonds ofdislocations (from one to three per outer atom of an extra plane of A3BScompounds), disrupted valence bonds, point defects settled on them ,etc. It wouldbe logical to assume that thermal pressure treatment on cBN leads to plasticdeformation of the material, thus resulting in an increase in the dislocation density(Fig.3,b) in it at tann > 1000 °c and also in an increase in the C-band intensity.However, the increase in the intensity of the dislocation band under thermaltreatment of cBN at tann = 700°C is unlikely to be related to its plastic deformation,but rather to the variation in the concentration of the defects that are radiationcenters on the bonds of dislocation cores. These defects could be centers on whichthe energy trasport from the S centers is effected.The enhancement of the C-bandintensity in the spectra of the samples annealed under pressure could be connectedwith the increase in density of the dislocations per se, even under conditions ofincomplete saturation of the bonds with radiation centers.In support of the foregoing, it is found in our study that microstresses of the cBN
crystal lattice caused by the material annealing in a nitrogen atmosphere underpressure and resulting in the formation of the C-band in the spectra are of differentqualitative character and are traced easily by the multiplet structure of ZPLs fixedin the C-band (600-800 / nm) of the spectrum.
It is established [4] that the luminescent doublet ZPL at 1.883-1.884eV,registered in cubic boron nitride spectra simultaneouselywith the C-band and PC-3centers [3] after thermal treatment in nitrogen atmosphere, is generated by
400
the ~A (d)-+2E(d) type transitions on a defect which is the Mn+4 ion implanted intooxygen octahedr~n of Al203 microinclusions in cBN, being an analog of the PC-3center generated by the Cr ion.ZPL at 1.814 eV, always registered apart from theindicated doublets simultaneousely with the C-band in the spectra of samplesannealed under pressure, is suppousedly, generated, by additional electronictransitions to the Mn ion [5]. This attests that the state of the cBN defect structurewhich determines the stress field of thematerial crystal lattice distorting the oxygenoctahedron with the Mn+4 ions after thermal treatment of the samplesunder pressure is different from that formed after their annealing in a nitrogenatmosphere. This fact confirms the above suppositions regarding the nature oflumenescence in the C-band range occuring in CL spectra after various thermaltreatments. When tann > 1270 K, changes occur in the defect cBN structure due tothe transformation of electronic states in the forbidden zone. The luminescencecenters in the spectra of the samples thermally treated under pressure are registeredat 2.52, 2.32 and 2.27 eV, and of those treated in a nitrogen atmosphere, at 2.62,2.52,2.49,2.315 and 2.27 eV. These centers have been detected for the first time.
References
1. Shipilo, V.B., Zaitsev, A.M., Shishonok, E.M., and Melnikov, A.A. (986) Theeffect of Annealing on IR Spectra of the CBN Reflection, J.Prikl.Spectr.,45,601-605
2. Davies,G.(970) No Phonon Lineshapes and Crystal Strain Fields inDiamonds, J.Phys.C.Solid St.Phys.,3,2474-2486
3. Shipilo, V.B., Shishonok, E.M., Zaitsev, A.M., and Melnikov, A.A.(988)Influence of High Pressure on Cathodoluminiscence of Cubic Boron Nitride,Phisica Status Solidi, 108,431-436
4. Sviridov, D.T., Sviridova, R.K., and Smirnov, Yu.F.(976) Optical Spectra ofthe Ions of Transition Metals in a Crystal, Nauka Press, Moscow
5. Shishonok, E.M., Shipilo, V.B., Lukomskii, AJ., and Rapinchuk, T.V.(989)The nature of some luminescence centers in cubic boron nitride, PhisicaStatus Solidi(a),1l5,N2,237-242
MACRO AND MICRO STRUCTURAL FACTORSIN TlllN FILM GROWTH OF ill-V COMPOUNDS
PETER J. GIELISSE AND HALINA NICULESCUFAMU/FSU College of Engineering2525 Pottsdamer StreetTallahassee, FL 32310, USA
Abstract
This paper reviews some of the phase equilibria and structural concepts through whichthe various members of the high density nitrides are related.In some of its aspects this information is felt to be significant, in reduced pressuresynthesis and processing of the wide bandgap nitrides.We also discuss some of the areas that need further emphasis if realistic applicationsin the electronic and mechanical areas are to be realized in the near future.
1. Introduction
Interest in synthesis and development of single crystal, polycrystalline or amorphousmembers of the III-V nitrides is founded primarily in the measured or calculated valuesof their physical, chemical, thermal, mechanical and optical properties. No one familyof materials offers, at least in principle, so much promise in as many different areas.Potential applications in the active and passive device fields, in thermal managementand high speed circuitry, as well as in wear and chemical resistant areas, abound. Yetprogress towards actual product or process realizations has lagged behind developmentin other material families such as diamond, carbides, oxides, and phosphides. Much ofthis is due to the complexities generated when one switches from single component(e.g. diamond) to binary or ternary systems which, for one, require the maintenanceof chemical equilibrium and atomic level structural control in the growth interface toa much higher degree than in unary systems. These and other factors - amply supportedby current experimental results - may make one conclude that the availability of widebandgap III-V nitride materials for device and system applications is currently still inits infancy.The synthesis of each member of the III-V nitride family is further complicated bythe occurrence of several polymorphs and polytypes , for which the exact pressuretemperature domain boundaries, equilibrium and non-equilibrium, are not yet wellknown.
401
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 401-420© 1995 Kluwer Academic Publishers.
402
The efficient synthesis of high quality mononitrides also suffers from theoveremphasis on the "similarity" of these nitrides to group IV elemental structuressuch as diamond and the monophosphides of boron, aluminum and gallium.
It is often said that c-BN is a diamond "analogue". At first sight this argument canbe easily understood when specific physical properties are superficially evaluated andcompared. The cubic form of BN is indeed the hardest substance next to diamond. Ithas a large bandgap (6.1 eV) and both occur in the zincblende structure. This is wherethe supposed commonality ends. In fact, there are no singular or direct corollaries ofmerit. Diamond occurs naturally and c-BN is entirely synthetic. Their oxidationcharacteristics are drastically different. Diamond is the archetypical example of thecovalency concept, whereas c-BN is a much more heteropolar material. BP (zincblende)is also often compared to diamond, based primarily on its substantially 100% covalentcharacter. In fact, from a scientific and research point of view, c-BN should beconsidered the end member of the binary nitrides, just as BP is for the binaryphosphides. They are simply the "diamond-like end members" of the nitrides andphosphides. Any direct comparison with diamond should be with the carbides such asSiC, Al4C and B4C for which diamond is the "carbon carbide". The recent introductionof the concept of a nitride of carbon [1], possibly as a defect structure mononitride oras a Si3N4-like material, will not violate the rule that the high density mononitrides BN,AlN, GaN and InN as well as the phosphides and carbides of similar structures, needbe compared to members within, not outside, their respective families.Figures 1 - 4, in which the important physical properties bandgap size and hardness
are compared to the Debye temperature, clearly illustrate the close relationship amongthe members of each family as well as the lack of parallels from group to group. Thevalues for GaN and InN in Figs 1 and 2 might need adjustment when either hardnessfigures (GaN) or hardness and Debye temperature values become more closely defined.The bandgap values for BP, Fig.3, have been reported between limits of 2-6 eV. Ourown absorption edge measurements [2] on some of the first small grain sizepolycrystalline BP samples, gave much less variation in the gap value as indicated bythe cross hatched area of Fig.3. The Debye temperatures used in Figs I - 4, werecalculated from the pertinent frequency values of the respective nitride vibrationalspectra [2-4].The physical properties of the wide bandgap nitrides have been gathered in Table 1.It will be noted that information is given for all the crystalline structures, three for BNand AIN and two for GaN and InN. For comparison purposes we also present similarinformation for the monophosphides, group IV elements and the semiconductorcompound SiC in Tables 2 and 3.The physical property values have been taken from several recent survey articles[5-7], while other data were specifically calculated for this presentation as indicatedabove. The recent experimental determination of the zincblende (CN =4) to rocksalt(eN =6) structural conversion pressure in AlN [8-9] together with our earlier derived
6
o
403
500 1000
Debye Temperature (K)
1500 2000
Figure 1. Bandgap values as a function of Debye temperature for high density nitrides.
404
4500
4000
NE2000~
~~<Jlon
"c:.."1;3 1500::r0..0g~
1000
500
o500 1000
Debye Temperature (K)
1500 2000
Figure 2. Hardness as a function of Debye temperature for the nitride family.
405
Range of values from absorption measurements
/
/
BP
/ IIII
/1Reported IValues~
/
6
5.4
5
4.6
4
>~0- 3'"0"0c:'"a:I
2
o
o 200 400 600 800 1000
Debye Temperature (K)
Figure 3. Bandgap values vs. Debye temperature for the phosphides.
AlP
SP
406
4000
3500
3000
('IE 2500
~~
'" 2000'"'l).§1;;:I: 15000-00c::>t: 1000
500
t InP
0
300 400
GaP
500 600 700 800 900 1000
Debye Temperature (K)
Figure 4. Hardness as a function of Debye temperature for high density phosphides.
TABLEI.Physicalpropertiesufwid..:bandgapnitrides
Property
BN
AIN
GaN
InN
Structure
c-BN
h-BN
R-BNc-AlN
h-AIN
R-AINc-GaN
h-GaN
c-lnNh-lnN
zw
rz
wr
zw
zw
0a=3.4934.045a=3.11
a=3.19
a=3.548
LatticeConstant(A)
3.615
(calc.)4:D~
4.5-4.55
4.98
c=5.760
c=4.98
c=5.19
Density(g/cm3 )
3.488
3.28
6.11
425
215
195
BulkModulus(GPa)
367
(calc.)(calc.)(calc.)
CTE(xIO-6
K-l
)4.2ta l
5.59
~al
3.0
5.3c
3.17c
3.7
4.5
ThermalConductivity
10-13
1.75-32
1.3
.56
(W/c
mK
)eo=8.5
e o=9.5
DielectricConstant
4.3
e.=4.68
e.=5.35
-4.85
IndexofRefraction
2.117
2.15
233-267
285-3.05
Resistivity(Hem)
J014_ld6
}(j!_
d11016
5.11
(~}
3.2-3.3
""1.lP-195
BandGap(eV)
3.4(0)
_.-6.4(1)
(C~f(6.28(0)
(D)
(calc.)
(0)
4.04I
1310
1022
DebyeTemp
(K)
-1900
(calc.)
(calc.)
z-zincblende,w-wurtzite,r-rocksalt
..,. o -.J
TABLE2.Physicalpropertiesofsemiconductorphosphides
Property
BP
AlP
GaP
InP
Structure
zr
zz
z
0
5.45
5.45
LaniceConstant(A)
4.538
4.339
5.87
Density(g/cm3 )
2.42
4.13
4.787
BulkModulus(GPa)
166
155
8689
71
CTE(x10-6K-l)
5.3
4.6
ThennalConductivity
3.6
.92
.752
.8(W/cmK)
DielectricConstant
11.1
12.4
IndexofRdraction
Resistivity(Ocm)
BandGap(eV)
2.18(1)
2.45(1)
2.24
1.27
3.0(D)
DebycTemp(K)
985
588
446
321
z-zincblendc,w-wurtzite,r-rocksalt
~ o 00
TABLE3.Physicalpropt,'rtiesofgroupIVmaterials
Property
Diamond
SiGe
o-SiC
Structure
zw
rz
zz
0
LatticeConstant(A)
3.57
4.436
5.43
5.66
4.36
Density(glcm3 )
3.52
2.3
5.32
3.21
BulkModulus(GPa)
443
411
9876
212
CTE(xI0-6K-1
)1.2
3.59
4.7
ThermalConductivity
201.4
.64.9-5.0
(W/cmK)
DielectricConstant
5.5-5.7
11.7
15.8
40
IndexofRefraction
2.41
3.49
4.0
Resistivity(Oem)
1016
2.3xl05
1.5x102
BandGap(eV)
5.47(()
3.3(1)
2.5(1)
1.1(D)
.67
3(D)
DebyeTemp(K)
2340
645
374
1130
z-zincblende,w-wurtzite,r-rocksalt
... ~
410
formulation for the calculation of such transition pressures from thermodynamic data[10],
P,=Po+K-1ln[Klvo . (Go- G,)+I]allowed the determination of the significant physical properties of the zincblendephases of c-BN and AlN as shown in Table 4 and Fig. 5. In the formula K is thecompressibility and (Go- G,) the work energy difference, i.e the work energy term forthe transition.Physical property values are not determined by major chemistry, structure and
bonding type only, a fact which complicates the development of a variety of promissingapplication areas in the mononitrides. A case in point is the value of the thermalconductivity of aluminum nitride, which, depending on minor oxygen content, variationin density and grain boundary thickness, can vary from 10 W/m·K to its maximumvalue in single crystals of 319 W/m·K. A quick overview of the range of theseproperty dependent values is given in Fig.6. AIN can, with reference to this propertybe compared with that of diamond, Fig.7. Both materials are currently used assubstrates in high power high-frequency electronic devices, particularly muitichipmodules (MCM's). When c-BN becomes available in sizes and with propertiescommensurate with cost efficient microelectronic substrate properties, it will clearlytake its place among AlN and diamond. Information on c-BN is not yet sufficient towarrant the development of charts of the type of Figs 6 and 7.
2. Structure
As Tables 1-3 indicate, the four fold coordinated nitrides c-BN and AlN both occur inthree structural modifications while GaN and InN appear, at least so far, in only twopolymorphic forms. The four members of the nitride family consist, therefore, of someten separate and distinct solids, each with specific physical properties. This constitutescomplications with reference to synthesis in as much as several of the phases can bemetastably retained under standard conditions, while processing methodologies spanpressure-temperature regimes covering the primary phase stability fields of severalphases. This situation is well illustrated by the phase occurrence(s) of c-BN under staticor high temperature and pressure conditions versus those under dynamic or shock typeloading, see Fig. 8. We start here with the two low density, layered, soft polymorphsof BN, the hexagonal h-BN type and the closely related rhombohedral r-BN type asstarting materials in the conversion from sixfold coordinated layer structures into thewide bandgap four fold coordinated structures.The layered structures are very similar [11] and differ only in the arrangement of
their covalently bonded individual layers which are stacked, in the c- direction, in anABC---fashion for h-BN and as ABAB---- in the r-BN case. This may not appear asvery significant at first glance. The impact of reaction kinetics and the occurrence ofspecific structures under prevailing synthesis conditions of pressure and temperature is,however, significant.In shock compression h-BN transforms to w-BN. Kinetics prevents simultaneous
formation of z-BN. Similarly r-BN is found to yield only z-BN in as much as the shock
TABLE 4. Parameters in CN-4 to CN-6 pressure transition
411
BN AIN
IonicCharacter, % 22.11 43
Bulk Modulus,GPa 367 195
Volume Change
tN, cm3/mole6.6 7.25
TransitionPressure, GPa 6O? 11.5-16.5
Free Energy-(Go-Gtl, Kcallmole 94 33.33-37.14
Entropy Change
As, caUmole'K 1.738 1.249
412
3
~jl
+0E:::.os~ 2.5<Il<I
~c:os..c:
U....0.. 2gc:UJ
1.5
500 1000
Debye Temperature (K)
1500 2000
Figure 5. Entropy-Debye temperature relationship for the high density nitrides.
413
60
(nm)
11066
(wt.%)CalculatedValue (Max.)
o L- _
50
350
319ALUMINUM
300 NITRIDE285
.034
250
QE§:
200.c:~
Single Oxygen175
tl 5:::l
" Crystal Contentc0
150l;
(ij
E Grain::.>
Density BoundaryoJ::r-
Thickness100
Figure 6. Thermal conductivity values for various types of AIN (wurtzite).
414
3500
CurrentValues
(%)
C 'al 1500ommerCl •CVDDiamond
1200
DIAMOND
1200
CH4
Content(%)
CVDDianond
600
2190
NitrogenContent(at%)
2600
NaturalDiamond
12230
IsotopeContent13C(%)
33200.07
SyntheticSingleCrystal
0.5
500
o
3000
1000
2000
1500
2500
Figure 7. Thermal conductivity values for various types of diamond.
HEXAGONAL
(~>-----(.:;:::: :~1,-::-t A
, , . ,...., )
;C:~L~_w-~--t-ln' --<~) 1""1 -+-+-oIY:-:i--r.' B
415
RHOMBOHEDRAL
DYNAMICCONDITIONS
B
B
WURTZITE
C
B
A
DYNAMICCONDITIONS
ZINCBLENDE
Figure 8. Structural changes from layered to high density nitrides for static and dynamicsynthesis conditions.
416
level (energy) is insufficient to reach the stability condition for w-BN. We predict,therefore, that w-BN could form from r-BN but at considerably higher energy(pressure) levels. Under static conditions, which are more favorable towards reachingequilibrium conditions and where pressures are lower than under shock conditions, bothh-BN and r-BN can form z-BN.Under static high pressure-high temperature conditions, in which growth takes place
over time periods of multiples of minutes, z-BN as a starting material tends to formw-BN and h-BN generates c-BN. Highly dynamic conditions, as experienced in shockformation, produces w-BN from h-BN and c-BN from r-BN. In these latterpolymorphic transitions the very short transition times allow for only a martensitic typetransformation, in which the dense fourfold coordinated structures retain the axialsymmetry and the layering sequence ABAB---- and ABCABC----, depending on thestarting material. The only actual change is a topical rearrangement within the layers,manifested by adjustments in the inter and intra (layer) bond lengths and the formationof a "corrugated" or "puckered" layer geometry. Alternatively, HPHT synthesisprocedures and conditions, allowing for diffusion-like processes to act as the majorstructural modifiers, produce the less ordered w-BN from r-BN and the densezincblende form c-BN preferentially from h-BN. The ultra short reaction timesexperienced during dynamic processing allows a new phase to be formed only viaminor restructuring within the layers, without three dimensional rearrangement. A trulyvolumetric rearrangement becomes possible only when, under static conditions, thestarting structure is broken down, i.e. is dissolved into the "catalyst" and reforms asthe thermodynamically stable form under the prevailing pressure and temperatureconditions. It will be noted that the retention of the layering order in the h-BN to w-BNconversion is along the c-axis direction, while for the r-BN to z-BN case, this is onlyobserved along a totally different, (I I 1) direction, as shown in Figs 9 and 10.In formation from the vapor [12] (low or reduced pressure conditions) all three
phases, h-BN, z-BN and w-BN can occur simultaneously.
3. Discussion
It is realized that successful reduced pressure synthesis of well crystallized andmorphologically well developed c-BN and AIN has sofar proven to be very difficult.One of the current drawbacks is the fact that the phase relationships in the P-T domainfor the many polymorphs and polytypes have not yet been definitively determined. Themixed phase nature of most of the products points to the highly non-equilibriumconditions at the growth interface(s) with reference to both nucleation and growthprocesses. A much better understanding of the many parameters that influence theseprocesses need be developed.The areas in greatest need of process development appear to be the refinement of thegrowth platform or substrate, the delineation of the optimal growth environment andways and means of process control. We are not likely to generate high quality (single)crystal films until the appropriate substrates (chemistry, structure, finish) are developed.Specially prepared or engineered substrate surfaces are likely to be more appropriate
417
A
_;-----------rv-/~..:::}./1' __ V',fIi ----- I d/! B
/ I -----+---- II I I
! I i I
! I Ii Ii I
I
I• C-(OOOll
I
o~/~r--I I ii I I
i: !:: i. ( !
i ~, I
Figure 9 S. tructure 0 f the .stacking s zmcblende type h' hequence . Ig density . .perpendicular to (000 I) mtndes. Note the ABAB
418
I (1) I)
+IIII
C
B
A
Figure 10. Structure of the wurtizite type high density nitrides. Note that ABCstacking occurs along the octahedral (III) direction as distinct fromthe zincblende case.
419
than those chosen on the basis of their availability. High energy density and particularyuniform energy density across the growth interface seems to be required for singlephase development. From both a research and future device application point of viewit is hard to imagine processing systems without in-situ environment and product qualitysensing capability.This article was written with the intent to bring together information on the wide
bandgap nitrides, which is likely to impact on the synthesis and processing of thesematerials and to place the nitrides, as a family, in the proper structural and phaseequilibrium perspective. Recent review articles have amply treated other aspects andto a certain extent applications. Specific references that, due to space limitations, couldnot be given here, may be obtained from the author.
4. References
I. Cohen, M.L. (1994) Harder than diamonds?, The Sciences May-June, 26-30.2. Gielisse, P.L, Mitra, S.S., Plendl, J.N., Griffis, R.D., Mansur, L.C., Marshall, R. and Pascoe,E.A. (1967) Lattice infrared spectra of boron nitride and boron monophosphide, Phys. Rev. 155, 1039-1045.
3. Olszyna, A., Kowalski, B. and Szmidt J. (1994) Optical properties of E-BN, Diamond and RelatedMaterials 3, 840-843.
4. Chrenko, R.M. (1974), Solid State Commun. 14, 511.5. Stacy, T., Liaw, B.Y., Khan, A.H. and Zhao, G. (1995) Considerations in further development ofaluminum nitride as a material for device applications, in M. Prelas, G. Popovici and PJ. Gielisse(eds) Proc. NATO Advanced Research Workshop on Wide Bandgap Electronic Materials, KluwerAcademic Publishers, Dordrechl.
6. Davis, R.F. (1989) Current status of the research on ill-V mononitride thin films for electronic andoptoelectronic applications, in R. Freer (ed) Proc. of NATO Advanced Research Workshop, KluwerAcademic Publishers, Dordrecht, pp. 653-660.
7. Davis, R.F., (1991) III-V nitrides for electronic and opto-electronic applications, Proc. IEEE 79,702712.
8. Volstaedt, H., Ito, E., Akaishi, M., Akimoto, S. and Fukunaga, O. (1990) High pressure synthesis ofrocksait type AlN, Proc. Japan Acad. B 66, 7-9.
9. Van Camp, P.E., Van Doren, V.E. and DeVreese, LT. (1991) High pressure properties of wurtzite androcksalt-type aluminum nitride, Phys. Rev. B 44,9056-9059.
10. Gielisse, PJ. and Yu, W.C. (1972) Polymorphism of solids under pressure, Mat. Res. Bull. 7,1151-1164.
II. Lam, P.K., Wentzcovitch, R.M. and Cohen, M.L. (1990) High density phases of BN, in J. J.Pouch and S.A Alterovitz (eds), Synthesis and Properties of Boron Nitride, Trans Tech Publications,Zuerich, pp.165-192.
12. Karim, M.Z., Cameron, D.C. and Hasami, M.S.J. (1992) Effect of deposition parameters on theformation of cubic BN ftlms deposited by plasma-assisted chemical vapour deposition from non-toxicmaterial, Surface and Coatings Technology 54/55, 355-359.
Acknowledgement
This work was made possible through support from the Defense Research ProjectAgency under contract N00174-93-0066 and the Army Research Laboratory undercontract DAALO1-94-K-3424.
mE FEATURES OF mE SINTERING PROCESSUNDER HIGH PRESSURE OF ALUMINIUM NITRIDE CERAMICWIm HIGH mERMAL CONDUCTMlY
V.B.SHIPILO,T.Y.RAPINCHUK,N.A.SHISHONOKInstitute of Solid State and Semiconductor Physics, Academy ofSciences ofBelarus, 17P. Brovka str. ; 220072 Minsk, Belarus
Due to its high dielectric and thenno-physical characteristics, aluminium nitridecan be regarded as an advanced thennoconductive material for semiconductordevices.In the present paper we study the influence of temperature and pressure on the
process of sintering of aluminium nitride.in the temperature range 1920-2220K.For the starting powder, we used aluminium nitride obtained by the
carbothennic recovery-nitrification of aluminium oxide. The content of nitrogenin the powder A1N was 33-33.5 weight %, that of oxygen was 0.8-1.0 weight %,
/and that of carbon didn't exceed 0.2 weight %. The size of particles of A1N was:d=I-6 l!m. The specific surface area was < 4.3-4.5 m2jg.The sintering of powder A1N was carried out in containers of calcite under thepressure 4 GPa and at the temperatures 1920, 2020, 2120, 2220K. Due to thethennal pressure increment, the actual pressure in the chamber was about 5 OPa[I]. To prevent diffusion of impurities out of the container, the tablets of A1N ofsize IOx6 mm were surrounded by the tantalum shell of thickness 0.1 mm. Thetechnique of measuring the Young's moduls and the internal friction is describedin [2]. The dielectric constant, tan (for the (requency I kHz), density, thennaland electric conductivity were measured by standard methods.As seen in fig. I , the powder of aluminium nitride sinters most intensely in thebeginning of the process. The effective time of sintering for achieving maximumdensity, was 40,30,20 and IS sec, for the temperatures 1920,2020 and 2220K,respectively. It should be noted that apparent density (fig. 1), as well as thennalconductivity (fig.2) and electric resistance (fig.3) of ceramics sintered at 2220Kand t > 15-20 sec., decrease with time. Approximately in the same time interval,we observed decrement of the elasticity modulus, and of the dielectric losses, forthe samples sintered at the temperature 2120K.Decrement of the elasticity modulus and growth of the internal friction can becorrelated to the irregularities in the stochiometric content of aluminium nitrideresulting from its dissociation and removal of the vaporous component, nitrogen.
421
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 421-426© 1995 Kluwer Academic Publishers.
422
Created therein point defects, primarily, nitrogen vacancies, interacting withinterstitial atoms, form complexes which concentration increases with time. Suchcomplexes, as had been shown earlier for the case of a close analogue, cubicboron nitride, may have dipole moment [3] and contribute to polarisation of thematerial, which accounts for growth of the dielectric constant. Apparently, thesame reason brings to increment of angle of dielectric losses, and also todecrement of the micro-hardness and the cracking toughness (Klc), with theduration of thermobaric processing of AlN [4].Building upon the cited data, one can conclude that aluminium nitride sinteringunder high pressure (5GPa) within the range of optimal temperatures for sinteringin the nitric medium, reduced considerably the effective time for sintering thismaterial. The kinetics of sintering aluminium nitride under isothermal conditioncan be described by the equation [5J:
ex = I - exp(-ktn) (I)
where ex = (PI - P2)J(P3 - PI) and PI, P2' P3 are the densities of compressed,sintered and completely sintered samples of AlN.It was established that in the initial segments of the curves 1-4 (fig.l),
In[-In( 1-ex)] depends linearly on In t. This enables us to determine the exponent nfor each temperature, and also the constant of the rate of sintering K described bythe Arrenius equation (K=Ko exp-Ea/kT). Calculation show that the energy ofactivation of sintering Ea which quantaively characterises the activity of processunder investigation, is equal to 34 kllmol, and n grows from 0.4 to 0.9 astemperature rises (1920-2220K).
It should be noted that equation (I) also describes the kinetic of crystallisationprocess of cubic boron nitride [I]. In the later case, the value of n ranges from Ito 4. Low value of n characterises unbalance in the system. Assuming that theexponent n in equation (I) has the same physical meaning, we may conclude thatin the initial stage of the process of sintering the powder of aluminium nitrideunder high pressure runs in strongly unbalanced and changing with temperature,condition which is typical of the material under going plastic deformation.The value of elasticity modulus of the samples obtained via processing the
powder AlN by pressure without heating, testified that the formation of intergranular ties in ceramics takes place already at this stage.
It is known[6] that sintering of the ultra-dispersed powder of aluminiumnitride is performed at the initial stage mainly on account of grain-boundaryslipping and is accompanied by the grains growth. The described mechanism ischaracterised by low activation energy ( about 20-60 kllmol) which was shownfor the case of thermobaric sintering of titanium carbide. Since sintering of AlNunder high pressure is also characterised by low value of Ea (34kJ/mol), we may
423
A g/cm 3
3.25
I3.20
3.15 t10 20 30 60 t. S
Fig 1. Dependence of density (p) of samples from AIN on the duration (t) ofisothermal processing lUlder the temperature of sintering, K: 1-1920, 2-2020, 3-2120,4-2220. The pressure is 5GPa.
;;e. W/mK
160
140
120
'0 20 30 60 t, S
Fig 2. Alteration of specific thermal conductivity of ceramic samples from AlN,depending on the duration of isothermal processing (t) under different temperaturesof sintering, K: 1-1920, 2-2020, 3-2140, 4-2220. The pressure is 5GPa.
R· JOIO
• Ohm 'em
25
20
15
10
5
10 20 30 60 t. S
Fig 3. Alteration of specific electric resistance(R) of ceramic samples from AlNdepending on the duration of isothermal processing under different tempatures ofsintering, K: 1-1920, 2-2020, 3-2120, 4-2220. The pressure is 5GPa.
424
E,GPa Q-I· 103
32516
30012
275 28
0
250
20 40 60 80 t, s
Fig 4. Dependence ofYmmg modulus E (I) and ofintemal friction Q-I(2) of ceramic samples AIN on the duration of isotermic processing under thetemperature of sintering 2120K. The pressure is 5 GPa.
tg6
12
11
10
9
0.06
0.04
0.02
I
20 40 60 80 t, s
Fig 5. Dependence of dielectric constant and of tg 0 for ceramic samples from AINon the duration of isotermic processing under the temperature of sintering 2120K.The pressure is 5 GPa.
suppose that in this case mass-transition is carried out by means of the boundaryslipping as the dominant mechanism of plastic defonnation too. The fast growthofYoung modulus and the fall of internal friction at the initial stage of sinteringunder high pressure are caused by two simultaneous processes, namely intensivecontinuation of fonnation of integranular ties and of annealing unbalancedstructural defects created by plastic defonnation. The existence of annealing stage
425
in the barometric processing of the nitride ceramics had been discovered earlier inthe investigation of sintering the ceramics of cubic boron nitride.From this point of view, the growth of thermal conductivity with prolongation
and the temperature rise in sintering (fig.2) will be determined by alteration of notonly porosity and the sizes of grains due to recrystallisation, but also of the defectstructure of the material the reduction of concentration of defects and thelowering of elastic energy of the lattice distortion due to annihilation of vacancies,the formation of admixture-vacancy complexes and the intervacation ofdislocation with point defects. Apparently, the size of grains has by far lesserinfluence on phonon scattering than point defects, since the calculated value ofthermal conductivity of aluminium nitride 320Wjm K [9] differs less from itsactual value obtained for relatively pure ceramic samples 260Wjm K [IOJ thanfor the samples with higher content of oxygen 50Wjm K (2*10 cm of oxygen)[11].On the basis of similarities between the electronegativities and the covalent
radii of nitrogen and oxygen atoms, and also taking into account the fact thatthermal conductivity of solids is sensitive mainly to the impurities located in thesites of the crystal lattice rather than to the ones in the interstices, we maysuppose that in aluminium nitride oxygen replaces, first and foremost, the atomsof nitrogen.The alteration of electric resistance of ceramic samples of AlN depending on
the duration of sintering at different temperatures is correlated to the kinetics ofthe powder compression (fig. I ), the higher electric resistance being the case forthe samples sintered at lower temperatures and during shorter time intervals. So,taking into account the above-stated, we may conclude that the observedalteration of electric resistance in AlN is determined mainly by the total area ofthe intergranular contacts which increases with the growth of parameters ofsintering under the given pressure, by the growth of grain and by the changes inthe spectrum of energy levels responsible for electric resistance, since the energylevels in the prohibitive zone of AlN are created by a certain type of defects in thecrystal lattice (vacancies, interstitial atoms, impurities, complexes, dislocationets.), the shape of which changes while sintering.The tendency of the observed correlation's t>20, there corresponds a certain
level of porosity anddeffectiveness of aluminium nitride under investigation whichcan be reduced by sintering the powder at optimal temperature which is 2120K forthe pressure 5GPa. 'The results of the our investigation' allow us to conclude that the sintering of
aluminium nitride under high pressure is based on the plastic deformationfollowed by post deformation annealing. Similarities in the process of sinteringaluminium nitride and in that of cubic boron nitride indicate the common natureof formation of the nitride ceramics under high pressure.
426
REFERENCES
1. Shipilo,VB., Gameza,L.M. and Smoljarenko,E.M.(1988) Nucleation andgrowth of cubic boron nitride crystals processes, Poroshkovaja me!allurgia1, pp. 73-79
2. Mazovko,A.V, Shipilo,V.B., Shishonok,N.A. and Chobot,AN. (1990)Resonant method ofYoung's modulus determination of superhard materials,Sverkhtverdye materialy, 1, pp. 18-21.
3. Shipilo, VB., Shishonok, E.M., Akimov, AI. and Shishonok, N.A. (1985)Investigation of dielectric properties of cubic boron nitride, Doklady Akad.Nauk Belarus ,XXiX, pp. 604-606.
4. Barashkov,G.A., Nespor,VS., Berdikov,VF. et al. (1987) Micromechanicalcharacteristics of AL 0 -TiN ceramic made in high pressure chamber,Poroshkovaja metallyrgia, 3, pp. 88-90.
5. Kisly,P.S. and Kusenkova, M.A (1980) Sintering ofrefractory compounds,Nauk. Dumka, Kiev.
6. Kusenkova, M.A.,Kisly,P.S., Makarenko, G.N. et al. (1978) Sintering ofaluminium nitride synthesised in low temperature plasma, Poroshkovajametallurgia, 4, pp. 25-29.
7. Stasjuk, eF. and Kaidash, O.N. (1983) Investigation of cinetics oftitaniumcarbide hot sintering under high pressure, Poroshkovaja metallurgia, 3, 30-32.
8. Shishonok, N.A., Shipilo, VB. and Anichenko, N.G. (1990) Influence ofsintering temperature on the elastic modulus and internal friction ofcubicboron nitride, Influence ofhigh pressure on material properies, Publisher ofInstitute Problem materialovedenia, Kiev, pp. 36-42.
9. Borom, M.P. Slack, G.A. and Szymadzek, J.W. (1972) Thermal conductivityof commercial aluminium nitride, American Ceram. Soc. Bull., 52 , pp.852856.
10. Takamidzawa, H. (1986) Technology of substrates manufacture of NikonDenky Company aluminium nitride, Electron. ceramics, 17, pp. 28-32.
II. Slack, G.A (1973) Nonmetallic crystals with high thermal conductivity,J. Phys. Chem. Solids, 34, pp-321-335
REACTIVE ION ETCHING OF SILICON CARBIDE WITH FLUORINECONTAINING PLASMAS
V.E. SIZOV, K.V. VASSILEVSKICree Research EED26, Politechnicheskaya Str., St. Petersburg, 194021, Russia
ABSTRACT: The reactive ion etching of silicon carbide with fluorine containingplasmas have been studied. The fast etch rates of SiC up to 0.45 ~m/min were obtained.The same etch rates were obtained for SiC and Si in SF6/02 mixture.
1. Introduction
Recently a significant progress has been achieved in the development of silicon carbidedevices for optoelectronics (light emitting diodes, photo diodes) and active electronics(field effect transistors, bipolar transistors, thyristors) [1,2]. The problem of structurefabricating is very important for SiC electronics because of its very high chemicalstability. Plasma etching in fluorinated gases is usually used to form SiC surface reliefwith reasonable fast etch rates [2]. In this paper we report on the reactive ion etching(RIE) of single crystal SiC in SF6, SF6/02, and CCI2F2 plasmas and characteristics ofSiC surface plasma treated.
2. Experimental results
The plasma was generated by radio frequency (13.56 MHz) discharge in a reactor withinduction excitation (H-mode) and maximum RF power of 5 kW. Evaporatedaluminium was used as a mask. Optimum etch conditions have been developed forvarious gas mixtures using pressures from 3 up to 50 mTorr and flow rates of 12-27sccm.
Dry etch experiments have been performed on 6H-SiC and 15R-SiC Lelycrystals. After the etch process the surfaces of the samples were investigated by electronspectroscopy for chemical analysis (ESCA), scanning electron microscopy (SEM), andreflection high-energy electron diffraction (RHEED).
The fastest etch rate of 6H-SiC in pure SF6 plasma was 0.45 ~m/min atpressure 40 mTorr and gas flow rate 12 sccm (Figure 1). The same etch rates wereobserved for Si- and C- face regardless of SiC polytype and conductivity. Decreasing ofetch rate during the process time was not observed (Figure 2a). The saturation of thesurface by fluorine at 11 at.percent level was registered by ESCA. Carbon enrichment
427
M.A. Pre/as et al. (eds.), Wide Band Gap Electronic Materials, 427-430© 1995 Kluwer Academic Publishers.
428
l'I·s 0.4S::l. 0.3
i~ 0.2
.a.s~ 0.1
a
/
.0 0
00
o
l'I 0.4§"-[ 0.3
oje0.2 0
.aB 0.1~
o
0·Ck.6 2.6 3.0 3.2 3.4 3.6
input power. arb.unit3.6 0.0 O!:-'--o-l'::-O~~20~--='30o-"-,.'40o-"-5="0,..-'--""60
pressure, mTorr
Figure J. Dependence of6H-SiC etch rate(a) on input power in pure SF6 plasma at a gas flow of 12 sccm (<1 <1 <1 <1 - p-SiC at pressure 3 mTorr,o 0 0 - n-SiC at pressure 3 mTorr, 0 0 0 - n-SiC at pressure 50 mTorr);(b) on gas pressure in pure SF
6plasma at a gas flow of 12 sccm, and input power equalled 3.4 arb. units.
of surface was not observed (Figure 2b).The RHEED pattern in Figure 3 demonstrates the single crystal state of the
surface of 6H-SiC substrate etched on depth of 5 J-lm.Mesa structures up to 20 J-lm in height were formed using pure SF6 plasma.
Surfaces with significant roughing on silicon and carbon faces were obtained at low gaspressure, and smooth surfaces were obtained at high pressure (Figure 4).
,,,I , I I I
60 b--:----l----1----:----:---\ I ;-' <!l. I
I I I I Alo, , , I I
40 - - - - ~ - - - - ~ - - - - ~ - - - - ~ - - - - ~ - --I I I I 9: + ~ ? :
20 - - - -~ - - - - ~ - - - - ~ - - - - ~ - - - - ~ - --
* ~ ~ W QqJ T *
o0!'='-""':5~~1':.'0~~15::=~2~0:;=~2~5:;=~30
Time. min
I I I I II I I I 1r
5 ----~----~----~----~--- 1---I r I I Ia: : : I :----,----r----r--- ----r---I I I ,
I , I I II I I I I----;----r---- ----r----r---I I I'I • I I ,
I r I I I----,---- ----,----,----,---, t , I I, I I I II I I I ,
--- ----~----~----~----~---I 1 I •I I I I
0o!u-u~~~Lu'~~''"_'_'_""''"'~~'~.=J5 10 15 20 25 30
Time, min
48:t3
Figure 2. Dependencies of (a) 6H-SiC etch depth and (b) surface composition on etch time in pure SF6plasma at a gas flow of 12 sccm and total pressure 3 mTorr (000 - Silicon, <1 <1 <1 <1 - Carbon, 0 0 0 Fluorine, * * * -Oxygen).
429
Figure 3. RHEED pattern of6H-SiC surface after treatment in pure SF6 plasma at a gas flow of 12 seem andpressure of3 mTorr for 25 min.
The SF6/02 mixtures were used for etching Si and· 6H-SiC. The same etchrates equalled to 0.2 ~m/min were obtained in SF6/02 mixture at flow ratio 1 andpressure 5 mTorr. .
In addition we have investigated the etch process of 6H-SiC in pure CC12F2plasma. This is due to the facts that chlorine-based discharge chemistry is useful fortreatment of nitrides, such as GaN and A1N, and silicon carbide is usually used as asubstrate for epitaxial growth of these materials. The etch rates of 6H-SiC in pureCCI2F2 were 0.03-0.05 ~m/min at pressures from 5 up to 50 mTorr and gas flow 12-27sccm.
3. Conclusion
In summary, we have found that RIE of SiC in SF6 discharge in reactor with inductionexcitation provided very fast etch rates and smooth surface without residuesindependently of the polytype and orientation of SiC substrates. Also it was found thatSiC was slowly etched in pure CCI2F2 with practically constant rate over a wide rangeof the etch conditions.
430
(a)
Figure 4. SEM patterns of 6H-SiC mesa-structures 80 J.lm in diameter and 20 J.lm in height formed bytreatment in pure SF6 plasma at a gas flow of 12 sccm, (a) pressure 3 mTorr and (b) pressure 50 mTorr.
4. References
I. Palmour,J.W., Edmond,J.A., Kong,H.S., and Carter,C.H.,Jr.(1992) Applications for 6H-Silicon CarbideDevices, in c.Y.Yang, M.M.Rahman, and GLHarris (eds), Amourphous and Crystalline Silicon CarbideIV, Springer Proceedings in Physics 71, Springer, Berlin, pp.289-297
2. Ivanov,P.A., and Chelnokov,V.E.(I992) Recent developments in SiC single-crystal electronics,Semicond.Sci. Technol. 7, 863-880
1.54-J,lm PHOTOLUMINESCENCE FROM Er-IMPLANTED AIN & GaN
R.G. WILSON and R.N. SCHWARTZ, Hughes Research Laboratories, MalibuCA 90265
c.R. ABERNATHY and SJ. PEARTON, University of Florida, GainesvilleFL 32611
N. NEWMAN, M. RUBIN, and T. FU, Lawrence Berkeley Laboratory, BerkeleyCA 94720
I.M. ZAVADA, Army Research Office, Research Triangle Park NC27709
The study of rare-earth-doped TIl-V semiconductors has generated interest becauseof the potential for producing efficient, room temperature, electrically excited, sharp intra4f rare-earth emissions for optoelectronic device applications such as lasers and lightemitting diodes. TIl-V nitrides and in particular GaN, AIN, and their ternary alloys withInN, are attractive for fabricating optoelectronics devices that could operate in the uv,visible, and infrared wavelength ranges. Recent advances in growth techniques for GaN,AIN, and ternary films have led to the availability of quality material that is suitable as ahost for rare-earth ion incorporation. We discuss the observation of the 1.54-J,lmluminescence of optically excited Er3+ in ion implanted epitaxially grown films thatinclude GaN and AIN grown on GaAs or on sapphire. We propose energy level diagramsto account for essentially all of the observed wavelengths/energies for two different GaNstructures.
1. Introduction
AIN, GaN, and their ternary compounds are of interest for optoelectronics andhigh temperature electronics applications because of their wide bandgaps, 6.2 eV for AINand 3.4 eV for GaN, their relatively high temperature stability, and their compatibilitywith other TIl-V semiconductor systems. They are also of interest as passivation layerson other III-V materilas sytems. We grew AIN or GaN films on GaAs or sapphiresubstrates using MOMBE or ion-heam-assisted MBE. These materials have either cubicor hexagonal structure. We implanted the films with lanthanide rare earth elements,especially Er, at 300 keY and fluences from 0.5 to 5xlO14 cm-2. We additionallyimplanted oxygen to provide a negative ligand to enhance the efficiency of opticallystimulated emission from among the levels of the 4113/2 and the 4115/2 field split Starkmanifolds of Er3+ in the GaN or AIN crystals. These transitions give rise to opticalemission around 1.54J.lffi wavelength, which is of interest for optical communicationsystems. Upconversion can produce emission in the visible and uv wavelength ranges.We have observed optically stimulated emission near 1.54J,lm that consists of manyindividual wavelenghts, the details of which depend on annealing and measurementtemperatures. We have observed emission of these individual lines at room temperaturewith an intensity almost as great as the intensity measured at 6 and 17K, which isconsistent with the large bandgap of these materials.
431
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 431-435© 1995 Kluwer Academic Publishers.
432
2. Experimental Techniques
Films of AlN and GaN about l~m thick were grown on GaAs using MOMBEand an ECR source for atomic nitrogen, and films of GaN of similar thickness weregrown on sapphire using low energy ion-beam-assisted MBE. Er was implanted intothese films at 300 keY and f1uences from 0.5 to 5xlOl4 cm-2 using a 400-kY custompost-acceleration mass separation mass spectrometer/implanter. Ions were separated at.full energy using a large double focusing magnet, and the implants were performed atroom teperature and 7° off the normal. Implanted materials were broken into pieces andannealed at temperatures between 500 and 800°C in a quartz furnace in flowing dry N2.
Secondary ion mass spectrometry (SIMS) measurements were performed using CAMECA4f and 5f instruments and an 02 primary beam and positive secondary ions. SIMSquantification was performed using the carefully measured f1uences of the implantedelements. Photoluminescence (PL) emission from samples of Er-implanted AlN andGaN was stimulated by incident light of 457.9nm wavelength from an Ar laser. Theincident power was about 200mW and the power density was about lW/crn2. PL spectrawere measured at 6, 77, and 300 K, using a 0.85-m Spex 1404 double monochromatorwith a 6OO-mm grating blazed at 1.6~m, a cooled Ge detector, and a two-phase lock-inamplifier. The spectral resolution of the system is about 1 nm. The FWHM of theintense lines of from these samples was about 2 nm.
3. Results and Discussion
Depth profiles of Er and 0 were measured using SIMS. Depth profiles for Erimplanted in GaN are shown in Fig. 1. The 0 profile has a somewhat different shape butthe peak of the distributon is near that of the Er. Er does not redistribute in GaN or AINfor annealing up to 800°C, as it does in Si, GaAs, GaP, InP, and other semiconductors.Optically stimulated PL measurements using the 457.9-nm line of an Ar laser were madeat 6K for samples annealed at 500, 600, 650, 700, and 800°C. For GaN grown onsapphire, we observed the emission lines to be narrow in width and to increase inintensity with annealing temperature from 600 to 650 to 700, reaching best resolution atabout 700°C. The intensity of the PL lines decreased for 800°C annealing, consistentwithg observations reported by Elsaesser et al. l Pomrenke et al.2 reported that themaximum emission intensity for Er in other III -Y materials occurred between 625 and700°C. The most intense emission was observed for 650 and 700°C, the intensity beingless for 500, 600, and 800°C. Many individual emission lines were observed between 1.5and 1.65~m for GaN grown on sapphire. The emission spectra were subsequentlymeasured at 77 and 300 K for the 700°C annealing condition. PL spectra from Erimplanted hexagonal GaN on sapphire measured at 6, 77, and 300 K are shown in Fig.2.Shown in Fig. 3 is the PL spectrum measured at 6 K for Er-implanted AIN grown onGaAs. Because the pump excitation energy is less than the bandgap of both GaN andAlN, we suggest that the observed PL of Er 3+ is caused by direct optical excitation.For the GaN/sapphire sample, resolved transitions among the levels of the Er3+ Starkmanifold of the 4113/2 first excited state and the 4115/2 ground state are seen for 6 K; the
Figure 2. Photoluminescence spectra from Erimplanted GaN/sapphire annealed at 700°C
433
1ri"'c----r---.,..---,---,
Er In~:IOO.v 2X1014 .....2
101l0l'O:---D.2:i-=----0A.f:----:o.a=---~o.a
DEPTII (jim)
Figure 1. Depth distribution of Er implantedin GaN/sapphire at 300 keY and 2xl014
cm-2 annealed at 700°C (02 primary ionsand positive secondary ions ). As-implantedand an-nealed up to 800°C are the same profiles.
1580 1600
Figure 3. Photoluminescence spectra from Er-implanted AlN/GaAs annealed at 650°C andmeasured at 6 K.
434
energies of the lines are the same but the intensities of the lines are somewhat differentfor 77 and 300 K. Even at 300 K the transitions between individual Stark levels areresolved. Comparison of the PL spectra for the GaN/sapphire sample at 6, 77, and 300 Kshows that there are more higher energy levels populated in the first excited state Starkmanifold at 77 K than at 6 K, and even more at 300 K, and that there are more transitionsamong the higher energy levels ("hot lines") of both states. Comparison of the areasunder the PL spectra measured at 6, 77, and 300 K shows that there is negligible loss oftotal PL intensity when the temperature is raised from 6 to 77 K, and at most a decreaseof a factor of two when the temperature is raised to room temperature, in agreement withthe prediction for a material with a bandgap of 3.4 eV, and in agreement with the plots ofEr PL intensity vs temperature with host bandgap as parameter shown in reference 3. ForGaN and AIN on GaAs, the Er3+ PL emission spectra are superimposed on a broademission spectrum as previousy observed for GaAs and other narrower bandgapmaterials. l The spectrum for GaN/GaAs is similar to the one measured from identicallyprocessed Si (cubic site symmetry), except that for Si they are not on top of the broadbackground described above that is characteristic of GaAs, indicating that the spectrum forGaN/GaAs is probably charactereistic of Er at sites of cubic (Dh) symmetry.
The various PL emission spectra for GaN were studied in detail using energydifference systematics, and energy levels schemes were constructed for the Stark energylevels of the 13/2 first excited state and the 15/2 ground state of Er3+. Examples of theseproposed level schemes are shown in Figs. 4 and 5.
cnr"1 cm-1mz ,.. ........ --'''''-AlL-.. 2DI
.. ~.-115 1571312
__,:E.7 2DI
8S71
6-
Ito 1:._114 1:-&0
to 1:.608Z 1:.-502fl 1:.1380 1';.61
Figure 4. Proposed Stark manifold energylevel scheme for Er3+ in GaN/sapphireannealed at 700°C and measured at 6 K.
221 t.121. :E.'184 t.10
11' t.13
~ ==111::=:::::.1:= i:~52fl--:~-===-~=:::;I==== 1:. 60.50" LaM
15/2
Figure 5. Proposed Stark energy level schemefor Er3+ in GaN/sapphire annealed at 700°Cand measured at room temperature.
435
4. Conclusions
We have measured optically stimulated photoluminescence from GaN and AlNgrown on sapphire and GaAs substrates at about 1.54 J.UIl at temperatures of 6, 77, and300 K. The emission from GaN grown on sapphire consists of more than 20 discretelines whose relative intensities vary with measurement temperature and whose totalintensity decreases only slightly at room temperature. We propose Stark energy levelschemes for the 13/2 first excited state and the 15/2 ground state of Er3+ in the GaNcrystal. The depth distribution of Er implanted in GaN and AIN does not change withannealing up to 800°C, as it does in other materials of narrower bandgap.
5. References
1. Elsaesser, D.W., Colon, J.E., Yeo, Y.K., Hengehold, R.L., and Pomrenke, G.S., Mat. Res.Soc. Symp. Proc. 301, 251 (1993)
2. Pornrenke, G.S., Ennen. H., and Haydl, W., J. Appl. Phys. 59, 601 (1986)3. Favennec, P.N., L'Haridon, H., Salvi, M., Moutonnet, D. and Guilion, Y., Electron. Lett.
25, 718 (1989)
AES-SIMS ANALITICAL SYSTEM FOR COMPOSITION MEASUREMENTSOF WIDE BAND GAP SEMICONDUCTORS.
A.I. BABANIN,A.F. Ioffe Physical TechnicalInstitute and CREE Res.EED,26, Polytechnicheskaya Str., St.Petersburg,I94021, Russia.A.A. LAVRENT'EV,State Electrotechical University,5, Popova Str. ,St. Petersburg,197376, Russia.
Characteristics of wide bandgap semiconductors are determined by depth andsurface distribution of the chemical elements [1-5]. Auger electron spectroscopy (AES)is widely used for composition analysis of element concentrations higher than I at.%[6-7]. Due to its high sensitivity, secondary ion mass spectroscopy (SIMS) is usuallyused to determine impurity concentrations in the range 10-6 - 10-1 at.% [8]. Theanalytical setup we will describe in this report combines AES and SIMS methods andhas the ability to investigate a correlation between concentration and distribution ofbasic components and impurity content in the epitaxial layers.
AES-SIMS analytical setup was mounted on the base of an ultra high vacuumchamber (10- 11 Torr) [Fig 1-2]. For SIMS measurements ions are mass filtered andocused by an objective system, which subsequently allows independent adjustment ofthe ion beam diameter and density. The ion beam scanning size is 500x500 micron, ion
current density is 0.2 mA/cm2. High transmission secondary ion optics with a narrowband pass energy filter allows one to obtain mass resolution of the quadruple filter(3M) with a dynamic range of six orders of magnitude. Weak draw field of secondaryions made it possible to make simultaneous AES and SIMS analysis.
For AES measurements cylindrical mirror analyzers with energy resolution of0.5% and coaxial electron gun with energy of 3 keY and electron beam size of 5micron were used. The electron beam is matched with the center of the crater of ionetching. The specific procedure of surface cleaning during the measurements wasdeveloped to reduce an oxygen reabsorbtion effect.
We are presenting AES-SIMS results on GaN, SiC-AIN and GaN-AIN alloysgrown on SiC substrates [Fig.3-7].
This work was completed with samples, produced at CREE Researh, Inc.(USA).We are indebted to Dr.V.Dmitriev for samples and helpful discussion.
437
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 437-445© 1995 Kluwer Academic Publishers.
438
I. Dmitriev V.A.,lvanov P.A.,Morozenko Ya.V.,Chelnokov V.E. and Cherenkov A.E. (1991). Proc. 1stIntern. Con! on Applications ofDiamond Films and Related Materials (Auburn, Alabama, USA), p.769.,2. Davis R.F.,Kelner G.,Shur M.,Palrnour J.W.,and Edmond J.A. (I991).Proc.IEEE 79,p.677.3. Karmann S.,Suttrop W.,Schooner A.,Schadt M.,Haperstron C.,Engelbrecht F.,Helbig R., Pensl G.,SteinR.A., and Leibenzeder.(l992).Chemical vapor deposition and characterizationof undoped and nitrogendoped single crystalline 6H-SiCJ.Appl.Phis.V.72, N.lI,p.5437.4. Dmitriev VA (1993).Silicon carbide and SiC-AIN solid solution pen structures grown by liquid-phaseepitaxy. Physica B.,V.ISS, p.440-452.5. Asif Khan M.,Kuzhia J.N.,Olson D.T.,George T.and Pike W.T. (1993), GaN/AIN digital alloy shortperiod superlattices by switched atomic Iyer metallorganic chemical vapor deposition. Appl.Phys.LeIt.,v.63., N.25.6. Davis L., Mac Donald N., (1976), Handbook ofAES. Physical Electronics lndastries. Minnesota.7. Suzuki A.,Matsunami H.,and Tanaka T., (1990) Auger Electron Spectroscopy Analysis ofThermalOxide Layers ofSilicon Carbide, J.Appl.Phys.,V.67,N.l, p. 1896.8. Yamaguchi N., Horma E., Kashiwakura I., Koiki K. (1984), SIMS quantitative analysis of gallium insilicon by using ion implantated samples for standarts. Secondary ion masspectroscopy .Berlin : Springer,p.llO.
AES-SIMS ~.'t~icol ci.t&",
o
439
Fig. I. The schematic diagram ofAES - SIMS Analitical System.
440
i::'"iiic.2c
441
i::'"iiic.2c
~'iiic.2c
rt~---_..-.~-_·_.....
(5",,"'
. .~
jlt·--'t'\,--'lI__-----~r-..,..--........l iOoN")
1\,"+
N1<.ll.
I --......------.-l,-,....~r-- ...,---·...-- Jr.
(20....,..)I\r+
a 500 1000 1500 2000
Electron energy (eV)
Fig. 3..-\uger spectra ofSiC-AIN alloy (iOat%) on 6H-SiC.* - surface cleaning during the measurement.
In ( ) shown Ar+ sputtering time.
".--
....
:J L --- >.. ......, (/)
C (1)
......, C
.j:>.
.j:>. tv
NGQ
o500
1000
1500
2000
Ele
ctro
ne
ne
rgy
(eV
)
Fig.4AugerspectrumofGaNlayeron6H-SiC
SIMS DEPTH PROFILE
---.. stu:J~
'-.../
>. 4/c;(1-+-'(J)
cQ)
-+-'c
443
o ·5 10 15 20· 25SPUTIER TIME (min.)
Fig.5. SIMS depth profile of GaN layers.doping Mg. on 6-SiC.
30
444
oooN
00
-.J L{)a:: ,- u
~::I:.-----.. 'CJ
> c:0
Q)~'----" :at:S
<:!> 0 >-.. ~010 L.. ~0 Q) 0C .....Q) 0
ec .5
u
0 ~L.. ....
+-' "0 00:l
Q) -«0 0 'CJ
W oil0 ~
Z L{)
o
445
AUGER DEPTH PROFILE
51 234SPUTIER TIME (min.)
o
.....-.. N.~ C.10-
......."
>. Al~.-(j)
cQ)~
c
Fig.7..-\uger depth protile of GaN-AIN alloy on 6H-SiC.
POSITRON ANNIHILATION IN SINTERED BORON NITRIDE
I.I.BARDYSHEV, A.D.BURAVOVThe Institute ofPhysical Chemistry, Russian Academy ofSciences31 Leninsky Prospect, Moscow 117915, Russia
Positron annihilation spectroscopy is a very sensitive technique to investigate theelectronic structure and defects in solids 1. The most common mode of positron annihilation with the electron in a solid is two gamma rays to be emitted in coincidence inapproximately opposite directions. If the center of mass of the annihilating electronpositron pair is at rest, all of the rest mass energy of the pair (2mc2) would transform totwo gamma rays of 0.511 Mev (mc2) emitted at 180° to each other. However, since theactual annihilating pair has some momentum, the two gamma rays are emitted in directions which deviate from 180° by an angle 8 given by
8=pmc (1)
where p is the linear momentum of the annihilating pair and c is the speed of light. Thetwo-gamrna-ray angular correlation technique (ACAR) measures the angular deviationof Eqn.(I) and provides information on the electron momentum distribution.
On the surface of solids electrons have less momenta than electrons of bulk. Therefore ACAR curves for positrons annihilating on the surface are more narrow thanACAR curves for massive solids, and this phenomenon gives opportunity to study thesurface structure.
In this work the ACAR technique was used to study the sintering process of activeboron nitride turbostratic powders in nitrogen atmosphere under temperature from 900to 1400°C.
The experimental ACAR curves for sintered (1 h) boron nitride ceramics are shownin Figure 1. One can see the full width on half maximum (FWHM) of ACAR curvesincreases with the sintering temperature. The experimental ACAR curves was decomposed into two components: the broad Gaussian component connected with positronannihilation in the bulk of boron nitride particles and narrow component from positronannihilation on the particles surface.
447
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 447-452© 1995 Kluwer Academic Publishers.
448
The values of ACAR curves FWHM, calculated intensity of narrow component Iand specific surface S of ceramics (measured by BAT technique) are listed inTABLE 1.
TABLE 1. Specific surface S, FWHM of ACAR curves, intensityof narrow component I of ceramics sintered
under different temperature T.
T,OC S,m2/g FWHM,mrad 1,%
900 201 6.2 7.0
1000 198 4.7 11.7
lloo 98 5.7 10.5
1200 74 6.6 6.6
1300 29 8.1 1.2
1400 24 8.5 0.9
The narrow component intensity is found to decrease when specific surface decreaseswith sintering temperature.
The intensity of narrow component I connected with the probability for diffusingpositron to reach the surface of boron nitride particle. When sintering occurs positrondiffusing inside the particle may pass from one particle to other through the contact area(Fig.2).
The electron microscopy shows the boron nitride powder particles are the thinplates (Fig.2). The yield rate of positrons to the surface of a thin plate is2:
k = 3.142 D2L
where D - positron diffusion coefficient, L - thickness of the plate.
The plate thickness L may be calculated from specific surface S:
2L=
Sd
where d - boron nitride density.
(2)
(3)
On the other hand the rate k connected with the intensity of narrow ACAR compo-
k = 1t (1- I)
where t = 155 ps - the mean lifetime of positrons in boron nitride3.
449
(4)
By comparison of eqn. 2, 3, and 4 one can obtain the correlation between the experimental intensity of narrow ACAR component 1 and specific surface of ceramics S:
(5)
The theoretical curve (Eqn.5) and experimental values of I(S) shown in Fig.3. Themaximum agreement between theory and eXl?eriment for the boron nitride ceramics sintered under temperature from 1100 to I 400 C provides the value of the positron diffusion coefficient D= 6.6*10.5 cm2 s-l,
The large deflexion for 900 and I OOO°C ceramics from theoretical curve can beexplained by inside structure rebuilding of the boron nitride particles at this temperature.
The value of positron diffusion coefficient obtained in this work is the most important new result for further positron annihilation investigations of boron nitride structuredefects and surface.
References
1. Positron Solid State Physics, Ed. by W.Brandt and A.Dupasquier, North Holland, Amsterdam, 19832. W.Brandt, Positron Dynamics in Solids, Appl.Phys.,5, 1-23 -1974)3. H.Murakami and T.Endo, Positron Annihilation in Boron Nitride, J.Phys.:Condens. Matter 1 (1989) SA131-SA134
JGOOl'~
C(@);a
-.tJ..
IltOO
O
12.0
00
/00
80
00
sqop
qoOO
-6-I
t-2
0'i
'a
iO£
2.1
H
6,M~D
oj::.
VI o
Fig
ure1.Angularcorrelationofannihilationradiation(ACAR)curvesforboronnitridepowdersinteredunderdifferenttemperature.
451
//11//1'//1/ (((1// II / / 1'1'1'1 1/ I' It /
o l- . ---1
Figure 2. Scheme of positron diffusion and annihilation in the bulk of boron nitride particles and on thesurface.
452
'"<u<
'"'cQ)
coCoeo"
14
12
10
6
2
a 2 6 a 10 12 1"1 16 1S 20
Specific surface. S'10"·5 (cm-2/g)
Figure 3. The ACAR narrow component intensity via specific surface of sintered boron nitride and theoretical curve corresponding Fig. 2 model.
WIDE BAND GAP ELECTRONIC DEVICES
V.E. CHELNOKOV,A.F. Ioffe Institute26 Polytechnicheskaya Street, St. Petersburg, 194021 RussiaK.V. VASSILEVSKICree Research EED26 Polytechnicheskaya Street, St. Petersburg, 194021 RussiaV.A. DMITRIEVCree Research, Inc.2810 Meridian Parkway, Durham, NC 27713 USA
ABSTRACT: This paper summaries the recent experimental results on siliconcarbide and gallium nitride electronic devices. These semiconductors are far ahead interms of device fabrication than other wide band gap materials because of progress inepitaxial growth, doping, p-n junction fabrication and contact development.Optoelectronic devices, as well as high-temperature, high-power and microwave devicesare reviewed.
1. Introduction
Main applications for wide band-gap semiconductor devices are optoelectronics, hightemperature, high-power and high-frequency electronics. Today the best characteristicson wide band gap semiconductor devices are achieved on silicon carbide and galliumnitride. The key to success of the fabrication of SiC devices was (1) the development ofa high quality SiC pn junction and (2) growth of SiC bulk crystals, making up to 2"diameter SiC substrates available for homoepitaxy. Both SiC devices and wafers areavailable commercially.Recent progress in (I) GaN epitaxial growth, (2) p- and n-type GaN doping and (3)
fabrication of contacts to GaN resulted in the development of GaN devices. GaNsubstrates are not available, and a possible way around this is to use SiC substrates forGaN heteroepitaxy. The following paper briefly summaries recent developments on SiCand GaN electronic devices. For detailed information about predicted and achievedcharacteristics of the devices on wide-band gap materials the readers should see Refs.[1-3].
2. Optoelectronic devices
The two primary optoelectronic devices in SiC are light emitting diodes (LEDs) andphotodetectors [4]. The peak wavelength of SiC LEDs may be controlled both by (I)the band gap of different SiC polytypes (Eg 3C-SiC - 2.3 eV, Eg 6H-SiC - 3.1 eV, Eg4H-SiC - 3.4 eV) and (2) the type of recombination level in the band gap (Fig. I). For
453
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 453-461© 1995 Kluwer Academic Publishers.
454
example, LEOs based on 4H-SiC at room temperature may have a peak wavelength of 396 nm (exiton recombination) [5], - 430 nm (AI-N donor-acceptor pair recombination)[6], - 480 nm (so called "defect" luminescence) [7], - 530 nm (recombination via Blevel) [8]. On the other hand, the peak wavelength of Al-N donor-acceptor pairrecombination differs for different polytypes: 4H-SiC - 425 nm, 6H-SiC - 470 nm, 3CSiC - 600 nm.Blue 6H-SiC LEOs are by far the largest market for silicon carbide devices [4]. The
optical power output for blue 6H-SiC LEOs is typically between 20 and 35 /lW at 20rnA and 3.4 V [9]. Important feature of SiC LEOs is their light output stability. Thetypical degradation after 10,000 hours is - 10 - 15 %.
100
80
== 60~-.;;;= 4041-=...
..:l 20"'l
0350 450 550
Wavelength, nm
Figure 1. Electroluminescence spectrum of SiC LEOs (300 K):I, 2 - 4H-SiC LEOs, 3.4 - 6H-SiC LEOs.
Photodetectors fabricated on 6H-SiC show peak quantum efficiencies in the rangefrom 250-280 nm [4]. Quantum efficiency is in the range 80-100 %. The extremelylow leakage current of SiC pn junctions results in a high photodetector sensitivity.Recently high efficiency LEDs were reported on GaN [10], and heterostructures of
GaN/lnGaN [II] and AlGaN/lnGaN [12]. For InGaN/AIGaN LEOs (A.max - 450 nm) aluminous intensity greater than I cd and external quantum efficiency of 2.7 % wasmeasured at 20 rnA forward current (300 K). This luminous intensity is the highest everreported for blue LEDs. Structures were grown on insulating sapphire substrate whichrequired both electrodes to be mounted on one side of the structure. Also high GaNsapphire crystal lattice mismatch (>10 %) may cause degradation of these devices.Silicon carbide substrates are promising for GaN LEDs because of its low latticemismatch (-3 %) and high electrical conductivity, making vertical device geometrypossible.
455
3. High temperature devices
The first high temperature SiC diodes and transistors were fabricated in the 60s [13].High temperature gas sensors based on MOS SiC structures operating up to 800°C werefabricated by Spetz and co-workers [14]. High temperature 6H-SiC blue LEOs werereported in Ref. [15]. Current-voltage characteristics and electroluminescence of theLEOs were investigated in the temperature range 20 - 500°C [15]. Operatingtemperature of silicon carbide pn junctions is higher than 700°C, and currenttemperature limit of SiC devices is determined by the maximum package temperature(-4000C). So far there are qo reports on GaN devices with operating temperature higherthan 175°C.
3.1 DIODES
Au/6H-SiC Schottky barrier diodes with breakdown voltage greater than 1100 V werefabricated [16]. The diodes operated at temperature up to 400°C.Properties of 6H-SiC pn diodes were investigated in a wide temperature range [17
21]. 6H-SiC pn junction diode with a breakdown voltage of 710 V was investigated inthe temperature range from 20°C up to 400°C [20]. It was found that for currentdensities lower than 10-3Ncm2, forward current is determined by recombination processin a depletion region. Reverse current density at 700 V was less than 10-3 Ncm2 at400°C. The forward current of 500 rnA was measured at 2.9 V and 2.49 V at roomtemperature and 400°C, respectively. Successful life testing measurements were alsoconducted.Temperature dependence of reverse current-voltage characteristics was investigated in
[17,18] at low current density. For high current densities 0::;60 kNcm2) thetemperature coefficient of avalanche breakdown voltage of 6H-SiC(0001) pn diode wasstudied in the temperature range 20 - 630°C [21].
3.2. TRANSISTORS AND THYRISTORS
Several studies have been conducted on high temperature SiC pn junction gate fieldeffect transistors (JFET) [22,23]. Dohnke et al [22] reported 6H-SiC high temperatureJFET operating in the temperature range 20 - 600°C. A maximum transconductance of3 mS/mm was measured at 600°C. In the whole temperature range complete pinch offof current-voltage characteristics was observed and source-drain current showed goodsaturation. Characteristics of a normally off 6H-SiC JFET were investigated up to450°C by Rupp et al [23]. The saturation current at room temperature was - 1.2 rnAand decreased to 0.8 rnA at 300°C.A high temperature depletion mode 3C-SiC metal-oxide-semiconductor field effect
transistor [MOSFET] was fabricated by Palmour et al [24]. 3C-SiC pn structure utilizedin this MOSFET was grown on 6H-SiC substrate. Stable transistor operation wasreported at 650°C. The maximum transconductance measured at 650°C was - 12mS/mm. High temperature planar depletion mode 6H-SiC MOSFET was fabricatedrecently by Krishnamurthy et al [25]. At 350°C the threshold voltage andtransconductance (Vg =- 2 V) was - -5.5 V and 0.1 mS/mm, respectively. High
456
temperature 6H-SiC MOSFETs were also fabricated and characterized by Palmour et al[26]. The devices showed good characteristics at temperatures as high as 6500c.4H-SiC MOSFETs were fabricated by Palmour et al [27] based on UMOS design
utilizing n+source fingers ion implanted into an epitaxially grown p-type channel layer(Fig. 2). The active area of the device was 6.7xlO-4 cm2 and the gate periphery was 4rom. The highest current devices blocked 80 V on the drain and could withstand currentdensities as high as 550 Ncm2 and power densities greater than 10 kW/cm2.A 3C-SiC metal-semiconductor field effect transistor (MESFET) operating up to
650°C was made [28]. 6H-SiC MESFET characteristics were measured at temperaturesup to 500°C [26].SiC p-n-p-n structures were reported by several research teams [29,30]. A 6H-SiC
thyristor operated at temperatures up to 400°C was fabricated by Palmour et al [30](Fig. 3,4). The thyristor had a 160 V breakover voltage. At a trigger current of 200 J.LAthe breakover voltage was 6 V. In on-state a current density of 100 Ncm2 (I=105 rnA)was achieved at 3 V.The first SiC operational amplifier IC was demonstrated by Brown et al [31].
Amplifier gain remained constant (-50 dB) in the temperature range from 25 to 3000c.
so )00
300K J Gate Voltage =18 V
< 6j)
~ <(
573Ke e1: 1:~ 40 ~ 14V= =
U U aoc:: c::-e 20 14V -i!
Q Q ~ IOV
lOY OV
0 O.OY
0 20 40 60 80 0 %0 ~ 60 10 lell
Drain Voltage, V Drain Voltage, V
Figure 2. Current-voltage characteristics of a 4H-SiC vertical UMOS MOSFET structure [27J.
4. High power devices
A high current 6H-SiC pn diode was fabricated by Anikin et al [32]. A forward currentof I A (current density - 200 Ncm2) was measured at voltage of 4 V. The breakdownvoltage was in excess of 300 V. So far it is the most powerful SiC diode. Avalanchebreakdown at high current densities was studied by Vassilevski et al [21]. A breakdowncurrent density of 60 kNcm2 and a power density of 9 MW/cm2 was reached at a 60 nscurrent pulse length.
-
n-type 6H·SiC
p' 6H·SiC
n+ 6H-5iC subslnlle
rZZZ/ZI?ZI?ZZZ????????????)t
CadlodeCoIUICt
Figure 3. Cross-sectional view of a n-p-n-p 6H-SiC thyristor device design [27].
·S "'"""...,(....--"""j-"""I---------I-.I- T.nJK
457
<!
......
<U
-4
·3
·2
-I
CATIiODE VOLTAGE (V)
--
-
----
-80 -100
Figure 4. Current-voltage characteristics of a 6H-SiC n-p-n-p thyristor (27].
458
High voltage 4H-SiC diodes were_made by Palmour et al (30]. The pn structures
were grown on both (OOOI)Si and (000 1)C faces of 4H-SiC substrates. The maximumreverse voltage was 1130 V. The maximum breakdown electric field ranged from2.2xlO6 V/cm to 5.2xlO6 V/cm for an Nct-Na concentration in the diode base region of8xlO15 - 2xlO18 em-3.2000 V 6H-SiC pn diode was made by Neudeck et al [33] based on CVD grown
structure. The diode structure consists of n+SiC substrate. n+-epilayer (8 ~m thick.Nct>1018 cm-3), n--epilayer (24 ~m thick, Nd -3xlO15 cm-3). p+-epilayer (l ~m thick,Na>1018 cm·3). Mesa structures with 7xlO-6 - 4xlO·4 em-2 area were made by dryetching. Reverse IV characteristics were measured in Fluorinert to prevent surfacebreakdown. The diode series resistance was of -0.3 Q cm2 at a forward voltage above2.5 V.6H-SiC buried-gate JFET was reported by Neudeck et al [34]. The JFET
demonstrated stable operation for a duration of 30 hours in an 6000C air environment at1.2 AllOOV. Vertical power SiC MOSFETs [27,30] are discribed in section 3.2.
5. Microwave devices
The first SiC diodes operating at 140 GHz was fabricated by Vassilevski et al [35].These diodes showed avalanche breakdown at 140 V. Diode capacitance at zero bias fordifferent diodes was in the range from 2.5 to 3.0 pF. Capacitance ratio Cmax/Cmin was- 8. The diode was used as a varactor to tune in the frequency of a silicon IMPATToscillator. A frequency modulation about 0.1 GHz was obtained at operating frequencyof 140GHz.Microwave power MESFETs were reported by Sriram et al [36] and Palmour et al
[37]. With 4H-SiC MESFET fmax of 12.9 GHz and a small-signal power gain of 2.2dB at 10 GHz has been achieved [37] (fig. 5).
30
o
! 1 I III ! I! II---l i II II : . ill
."}-, I ,: iii
i i 't I ' , iii ! : I: i i ! ! !!I : i I
"i ! III "--I I::, i I!
""'-',I III !' Ilill , I Ii
; ~i I MSG/MAGI Ii
I r H21 I : : I!
I I I ",-j I I ! IIi : l!i l"- I IIII III I 1\ I Ii !i
I 10 100
Frequency. GHz
Figure 5. Small-signal parameters. MSG/MAG and H21 • as a function of frequencyfor a 4H-SiC MESFET [37].
459
AIGaN/GaN MESFET for microwave applications was fabricated by Khan et al[38]. The heterostructure consists of n-GaN channel (n - Ix 1017 cm-3, ~n - 300cm2Nsec) and 250 A thick A10.13Gao.87N barrier. The maximum transconductance of25 mS/mm and maximum drain current of 50 mA/mm at room temperature wasreported. For a device with a 0.25 micron gate length (gate width 150 microns) thecutoff frequency of 11 GHz and the maximum oscillation frequency of 35 GHz wasmeasured.
6. Conclusion
Today, commercially available electronic devices in wide band gap semiconductors arerealized only on silicon carbide and gallium nitride. These devices are developed due toprogress achieved in epitaxial growth, doping control and device structure processing. Itis very important that silicon carbide wafers are available for homoepitaxy. To enterinto the market, the same problems have to be resolved for other wide band gapsemiconductors.
7. References
I. Davis, R.F., Kelner, G., Shur, M., Palmour. l.W.• and Edmond. l.A. (1991) Thinfilm deposition and microelectronic and optoelectronic device fabrication andcharacterization in monocrystalline alpha and beta silicon carbide, Proceedings ofthe IEEE, 79, 677-70l.
2. Ivanov, P.A. and Chelnokov, V.E. (l992) Recent development in SiC single-crystalelectronics, Semicond. Sci. Technol.7, 863-880.
3. Spencer, M.G., Devaty, R. P., Edmond, l.A., Khan, M.A., and Rahman, M. (eds)(l994) Institute of Physics Conference Series Number /37, Proceedings of the FifthConference on Silicon Carbide and Related Materials, Institute of PhysicsPublishing. Bristol.
4. Palmour, l.W., Edmond, l.A., Kong, H.S., and Carter, C.H.Jr. (1992) Applicationsfor 6H-Silicon Carbide Devices, in C.Y. Yang, M.M. Rachman and G.L. Harris (eds),Amorphous and Crystalline Silicon Carbide IV, Springer, Berlin, pp. 289-296.
5. Dmitriev. V.A., Kogan, L.M., Morozenko. Ya.V, Chelnokov, V.E. andCherenkov, A.E. (1990) Light-emitting diode with "-max =398 nm.Sov.Phys. Tech. Lett. , 16,828.
6. Dmitriev, V.A., Ivanov, P.A., Levin ,V.I., Popov, LV., Strelchuk. A.M.• Tairov,Yu.M., Tsvetkov, V.F., and Chelnokov, V.E. (1987) Fabrication of epitaxial SiCp-n structures on substrates obtained from bulk SiC crystals, SOy. Tech. Phys. Lett.13. 489-490.
7. Vodakov, Yu.A., Girka, A.I., Konstantinov, A.a., Mokhov, E.N., Roenkov, A.D.,Svirida, S.V., Semenov, V.V., Sokolov, V.I., and Shishkin, A.V. (1992) The lightemitting diodes on the basis of fast electron irradiated silicon carbide, in Ref. [4],pp. 374-380.
8. Barash, A.S., Vodakov, Yu.A., Koltsova. E.N., Maltsev. A.A., Mokhov, E.N., andRoenkov, A.D. (1988) Light emitting diodes for green spectrum region based onheteroepitaxial layers of silicon carbide 4H polytype, Pisma v Jurnal Techn. Fiziki14, 2222-2225.
9. Edmond, l .. Kong, H., Dmitriev, V., Bulman, G. and Carter, C.lr., (1994) BluelUVemitters from SiC and its alloys, in Ref. [3], pp. 515-518.
lO. Nakamura, S., Mukai. T.. and Senoh. M. (1992) High-Power GaN P-N Junction BlueLight-Emitting Diodes. Japanese Journal of Appl. Phys. 30, Ll998-200l.
460
II. Nakamura, S., Senoh, M., and Mukai. T. (1993) High-power InGaN/GaN doubleheterostructure violet light emitting diode, Appl. Phys. Lett. 62, 2390-2392.
12. Nakamura, S., Mukai, T., and Senoh, M. (1994) Candela-class high brightnessInGaN/AIGaN double-heterostructure blue-light-emitting diodes, Appl. Phys. Lett.64, 1687-1689.
13. Campbell, R.B. and Berman, H.S. (1969) Electrical properties of SiC devices.Mater. Res. Bull. 4, 211-222.
14. Spetz, A, Arbab, A., and Lundstrom, I. (1994) High-temperature gas sensors basedon metal oxide silicon carbide (MOSIC) devices, in Ref. [3], pp. 629-632.
15. Dmitriev, V.A, Linkov, I.Yu., Morozenko, Ya.V., Chelnokov, V.E. (1992) Hightemperature blue light-emitting diode, SOy. Tech. Phys. Lett. 18, 67.
16. Urushidani, T., Kobayashi. S., Kimoto, T., and Matsunami, H. (1994) High-VOltageAu/6H-SiC Schottky barrier diodes, in Ref. [3], pp. 471-474.
17. Anikin, M.M., Levinshtein, M.E., Strelchuk, A.M., and Syrkin. A.L. (1992)Breakdown in Silicon Carbide pn Junctions, in G.L. Harris, M.G. Spencer. and C.Y.Yang (eds), Amorphous and Crystalline Silicon Carbide Ill, Springer, Berlin. pp.283-285
18. Konstantinov, A.O. (1992) The Temperature Dependence of Impact Ionization inSilicon Carbide, and Related Effects, in: Ref. [17], pp. 213-219.
19. Strelchuk, A.M., Syrkin, A.L., Chelnokov, V.E., Cherenkov, AE. and Dmitriev,V.A. (1994) The current, electroluminescence and recombination parameters of SiCpn structures produced by container-free liquid-phase epitaxy, in Ref. [3], pp. 549552.
20. Edmond. J.A., Waltz, D.G.• Brueckner, S., Kong, H., Palmour, J.W., and Carter,C.H.Jr. (1991) High temperature rectifiers in 6H-silicon carbide, in D.B. King andF.V. Thome (eds), Proceedings of the First International High TemperatureElectronics Conference, Albuquerque, NM, pp. 500-505.
21. Vassilevski, K.V., Dmitriev, V.A., and Zorenko, A.V. (1993) Silicon carbide diodeoperating at avalanche breakdown current density of 60 kAlcm2, J.AppI.Phys. 74,7612-7614.
22. Dohnke, K., Rupp, R.. Peters, D., Volkl. J., and Stephani, D., (1994) 6H-SiCjunction field effect transistor for high-temperature applications, in Ref. [3],pp. 625-627.
23. Rupp, R., Dohnke. K.. Volkl J., and Stephani, D. (1994) Normally off 6H-SiCJFET and its high-temperature, in Ref. [31, pp. 503-506.
24. Palmour. J.W., Kong, H.S., and Davis, R.F. (1987) High-temperature depletionmode metal-oxide-semiconductor field-effect transistors in beta-SiC thin filmsAppl. Phys. Lett. 51. 2028-2030.
25. Krishnamurthy, V., Brown, D.M., Ghezzo, M., Kretchmer, J., Hennessy, W.,Downey, E., and Michon, G. (1994) Planar depletion-mode 6H-SiC MOSFETs, inRef. [3]. pp. 483-486.
26. Palmour, J.W., Kong, H., Waltz, D.G., Edmond, J.A., and Carter, C.H.Jr. (1994)6H-silicon carbide transistors for high temperature operation, in Ref. [201. pp. 511518.
27. Palmour, J.W. and Lipkin, L.A. (1994) High temperature power devices in siliconcarbide, in: Transactions of Second International High Temperature ElectronicsConference, Charlotte 1, pp. XI-3 - XI-8.
28. Davis, R.F.( 1989) Epitaxial growth and doping of and device development inmonocrystalline I3-SiC semiconductor thin films, Thin Solid Films 181, 1-15.
29. Dmitriev, V.A.• Levinshtein, M.E., Vainshtain, S.N., and Chelnokov, V.E. (1988)First SiC Dynistor, Electronics Letters 24, 1031-1033.
30. Palmour. J.W.• Edmond, J.A., Kong, H.S., and Carter, C.H.Jr. (1994) Vertical powerdevices in silicon carbide, in Ref. [3], pp. 499-502.
461
31. Brown. D.M., Ghezzo. M.• Kretchmer, J.• Krishnamurthy, V., Michon, G., Gati. G.,(1994) High temperature silicon carbide planar IC technology and first monolithicSiC operational amplifier IC, in Ref. [27J ,pp.XI-17 - XI-22.
32. Anikin, M.M., Lebedev, A.A., Popov, I.V., Rastegaev, V.P ., Strelchuk. A.M.,Syrkin, A.L.• Tairov, Yu.M., Tsvetkov, V.F., and Chelnokov, V.E. (1988)Electrical characteristics of epitaxial p+-0-0+ structures made of 6H polytype ofsilicon carbide, SOy. Phys. Semicond. 22, 181-183.
33. Neudeck, P.G.• Larkin, D.l., Salupo, CS., Powell, 1.A.• and Matus, L.G. (1994)2000 V 6H-SiC pn junction diode. in Ref. [3J, pp. 475-478.
34. Neudeck. P.G., Petit. 1.B., and Salupo, CS. (1994) Silicon carbide buried-gatejunction field-effect transistors for high-temperature power electronic applications.in Ref. [31 J, pp. X-23 - X-28.
35. Vassilevski. K.V.• Zorenko. A.V., Dmitriev, V.A. (1993) Presented at the FifthConference on Silicon Carbide and Related Materials, Novernber 1993,Washington DC.
36. Sriram. S., Clarke. R.C.• Hanes, M.H.• McMullin, P.G., Brandt, C.D., Smith, T.l .•Burk. A.A., Hobgood. H.M., Barrett, D.L.. Hopkins, R.H. (1994) SiC microwavepower MESFETs, in Ref. [3], pp. 491-494.
37. Palmour, 1.W., Weitzel. CE., Nordquist, K., and Carter, CH.lr. (1994) Siliconcarbide microwave FETs, in Ref. [3], pp. 495-498.
38. Khan, M.A.• Kuznia. 1.N., Olson, D.T., Schaff, W.J., Burm, J.W., Shur, M.S.,(1994) Deep Submicron AIGaN/GaN Heterostructure Field Effect Transistors forMicrowave and High Temperature Applications, 52nd Annual Device ResearchConference, Colorado, p. VIB-4.
Wide Band-Gap Photovoltaics
M. A. Prelasa, G. Popovicia,b, Salim Khasawinaha, and Jeff Sunga
a) College of Engineering, University ofMissouri, Columbia, MO 65211
b) Rockford Diamond Technology, Professional Arts Bldg., S. 6th Street, suite 101,
Champaign, IL 61801
Abstract
Wide bandgap materials will have many applications as coatings and as electronic devices. This paper describes an electronic application for wide bandgap materials in energyproduction. A specific portable power technology which converts the energy emitted fromnuclear reactions to electrical energy using wide bandgap photovoltaic cells without intermediate thermalization is described in this paper. The potential efficiency for the photovoltaic process is 35%, nuclear energy to electrical energy. And, ifcombined with hightemperature thermionic conversion the nuclear to electrical energy conversion efficiencyis 41% while the overall size of the system remains small. The key to the process is to firstconvert the high-grade ion energy to photon energy, which can then be directly convertedto electrical energy. This process is also usable as an advanced topping cycle for largescale energy production in conjunction with fusion power, as well as fission power. In addition to improved efficiency, the process also promises advantages in smaller volumes,smaller mass, and lower cost of the energy conversion hardware.
Introduction
The use ofwide bandgap photovoltaics (e.g., diamond and aluminum nitride) in fusionenergy conversion was discussed in 1981 [1] and in fission energy conversion was discussed in 1984 [2]. The focus of this discussion will be in the use and implementation ofradioisotopes. In this paper the process of nuclear energy conversion with wide bandgapphotovoltaics will be called the ~hotovoltaic ~nergy conversion of ~uclearenergy fu'stern (PENS). PENS can use both gaseous or solid nuclear fuels for power production. Asummary of how solid fuels can be introduced into PENS is given in References 3 and 4.The important underlying principle is to introduce the solid into the PENS so that it opti-
463
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 463-474© 1995 Kluwer Academic Publishers.
464
cally thin. This can be achieved by introducing the solid as an aerosol, thin fibers, thin films,or encapsulating the individual atoms of the solid in Fullerenes [3,4].
In the first step of the PENS, the nuclear energy is transported to a fluorescer which convertsit into photons. (The fluorescer could be a solid, liquid, or gas. This paper will focus on the
use ofa gas.) Then, in the second step of the process, the photons are transported out of theactive region to high bandgap photovoltaic cells which efficiently convertthe photon energyto electricity.
Figure l. Schematic Diagram ofthe PENS.
The efficiency of the two-step PENS process, while inherently less efficient than one-stepdirect energy conversion has two major advantages over thermal energy conversion, whichis a many step process. These advantages are: I) that it is a direct process producing a usefulenergy form from high grade energy and thus avoiding the Carnot cycle efficiency limitsimposed by thermalization and 2) that it is much simpler, potentially leading to more compact, more reliable, and less expensive energy conversion systems.
The advantage of the PENS process over a one-step direct energy conversion process is thatof feasibility. The scale length for the transport of the primary high-grade energy mustmatch the geometrical scale of the energy converter. Energetic ions have a transport lengthof micrometers while useful energy converters, on the other hand, have a scale length offractions ofmeters. For this reason direct conversion of nuclear energy has not previouslybeen possible. What was required was the concept ofan intermediate high-level energy converter that can be intermingled with nuclear material on a micrometer scale-length but produces an energy form that can be transported to meter scale-length direct converters
465
producing useful output - a sort of"impedance matching" for scale length of energy fonns.With PENS that scale length matching medium is a fluorescing gas, the nuclear-driven fluorescer. The photons it produces can be transported great distances, making it possible tocouple them to various energy conversion processes. Also, some conversion processes require greater power densities than the primary energy sources can provide. A PENS intermediate photon flux can be absorbed in a smaller volume than that in which it is produced,effectively concentrating the flux, enabling achievement of the high threshold power density for such conversion processes.
Thin Rectangular Film
Charged Particles
Small Aerosol Particles
Thin Cylindrical Fiber
Figure 2. An illustration of the use of thin solid geometries which allow reaction productsto escape the solid matrix into a surrounding gas.
The RECS concept of high efficiency production of light from radioisotopes makes thisconcept useful for remote power applications.
The choice of the fluorescer is important to the process. What is desired is that the fluoresceremits the energy deposited in it efficiently at a single wavelength without self absorption.The closest thing to a perfect fluorescer in nature is an excimer or an exciplex. If an atom(or molecule) is exited (So -> SI*) the atom or molecule can migrate and bond with an unexcited molecule (SOSI*). If the atoms or molecules are identical then the resulting molecule is called and excimer. if the atoms or molecules are different then the resultingmolecule is called an exciplex. The excimer/exciplex rapidly falls apart after the emissionof narrow band fluorescence since the ground state is unbound. Excimers/exciplexes havebeen experimentally shown to be efficient (up to 50%), narrow band (+/- 10 nm), non selfabsorbing fluorescers [see review in reference 3].
466
Some of the most efficient excimers fluoresce in the vacuum ultraviolet (He2' - 80 nm- 50%
efficient, Ar2*-129 nm- 50% efficient, Kr2*- 147nm- 47% efficient, Xe2*- 172nm- 48% efficient).
Thus, to convert these excimer's narrow band fluorescence to electrical energy requireswide bandgap photovoltaics. At this time the technology for wide bandgap photovoltaics isnot mature. Several promising materials - such as SiC, C (diamond), and AIN - do exist.Progress has been made in fabricating simple devices such as Schottky barrier diodes ondiamond and platinum silicide [5,6], a primitive p-njunction in diamond [7], n-type diamond material has been developed [8], and p-type and n-type aluminum nitride have beenclaimed [9,10]. Progress is being made on the fabrication of wide bandgap photovoltaics.
The use of portable power sources using radioisotopes have been reported [11] and couldhave immediate applications in space exploration. The current generation oflong lived portable power supplies are based upon the isotope Pu238 which is used to power the Radioisotope Thermion Generator (RTG) for missions such as the Voyager. In the Voyager, threeRTG units produced 7,200 watts of thermal power and 540 watts of electricity regulated to30 volts from 12,900 grams of the isotope.
Fluorescers
The Ion Source
An example of the PENS is shown in Figure 3 where the ion source is from the decay ofradioisotopes dispersed within a fluorescer gas. Effective dispersal is essential so that theions produced by the isotope decay deposit most of their kinetic energy in the excimer gasrather than in the radioisotope material. There are at least four methods of achieving the desired dispersal: gaseous radioisotopes, radioisotopes embedded in thin films, radioisotopesembedded in thin fibers, or microscopic aerosol of radioisotopes. The efficiency of transport of the ion energy from the radioisotope to the fluorescer medium varies with the scalelength of the thin film, fibers or aerosol, the chemical form of the radioisotope, and the uniformity of the radioisotope density. The variation of ion energy transport efficiency from amicrosphere, to the fluorescer medium, with thin films and microspheres are discussed inReference 12. Energy transport efficiencies are about 50% for reasonably designed thinfilms, 62% for reasonably designed fibers, and 70% for reasonably designed microspheres.The average atomic density in the medium must be on the order of lx1019 particles cm-3,enough to achieve reasonable power densities but not so great as to significantly degradethe transport ofthe fluorescence through the aerosol. Combining the constraints of efficiency, and optical transparency determines scale length of the thin film, fibers, or microspheresand number density. For example, a microsphere diameter of 5 11m and number density ofIx1()6 cm-3, which should not create significant absorption of the fluorescence [12], resultsin a fuel density of 0.63 mg cm-3, quite reasonable dimensions, and good number densities(3,9xI019 atoms cm-3). However, we believe that fibers would be an improvement. As dis-
467
cussed in Reference 12, a diameter of 5 f!m results in an ion transport efficiency of about60%.
Flu
PressureVessel
Figure 3. Schematic Diagram of the PENS which uses a radioisotope fuel in aerosol form.
The structure of a fission fueled PENS is discussed in Reference 12, and a fusion fueledPENS in Reference 13.
Excimer Fluorescers
Excimer fluorescers are the most efficient radiators known and, because of their unboundlower levels, do not self absorb. They radiate in the single, relatively narrow, band ofwavelengths required for efficient photovoltaic energy conversion [12]. The intrinsic fluorescence efficiencies of rare-gas and rare-gas halide excimers, based on standard W-valuetheory [12], are listed in Table 1. Achievable efficiencies should be near the intrinsic valuesat the power and electron densities characteristic of nuclear reactions.
In fact one group has reported measuring a nuclear-driven rare-gas excimer fluorescenceefficiency higher than that predicted by W-value theory [12]. Measurements of actual fluorescence efficiencies at various laboratories, including Lawrence Livermore NationalLaboratory, have demonstrated high fluorescence efficiencies for excimers. Experimentswith a variety ofexcitation sources (e.g. electrons, fission fragments, protons) and particledensities have given fluorescence efficiency values ranging from a few percent to as highas 68% (see review paper by Prelas et al. [12]). The most efficient excimer fluorescers arethe rare-gas excimers. Ion-Driven Fluorescers are discussed in much greater detail in thepaper "Nuclear-Driven Flashlamps" [12].
468
Table 1: Theoretical Maximum Intrinsic Photovoltaic, Tlpv' and Ion-to-Electric, Tlie'Efficiencies for Selected Rare-Gas and Rare-Gas Halide Excimer Fluorescers with
Matched High-Bandgap Photovoltaic Materials.
Energy PhotovoltaicBandgap 'I1i= 'I1pvx 'I1f'I1f Energy T1pV= EglEIExcimer (eV) Material (eV) Efficiency
Ar2* 0.50 9.6 AIN 6.2 0.65 0.33
Kr2* 0.47 8.4 AIN 6.2 0.74 0.35
Diamond 5.5 0.65 0.31
F2* 0.44 7.8 AIN 6.2 0.79 0.35
Diamond 5.5 0.71 0.31
Xe2* 0.48 7.2 AIN 6.2 0.85 0.41
Diamond 5.5 0.76 0.37
ArF* 0.35 6.4 AIN 6.0 0.94 0.33
Diamond 5.5 0.86 0.30
KrBr* 0.33 6.0 Diamond 5.5 0.92 0.30
KrCI* 0.31 5.6 Diamond 5.5 0.98 0.30
Na2* 0.46 2.84 ZnSe 2.7 0.95 0.44
SiC 2.4 0.845 0.39
Li2* 0.42 2.7 CuAISe2 2.6 0.96 0.40
SiC 2.4 0.89 0.37
Hg2* 0.21 2.58 GaS 2.5 0.97 0.20
SiC 2.4 0.93 0.19
ArO* 0.11 2.27 GaP 2.2 0.97 0.105
GaAIAs 2.2 0.97 0.105
KrO* 0.13 2.27 GaP 2.2 0.97 0.125
GaAlAs 2.2 0.97 0.125
XeO* 0.15 2.27 GaP 2.2 0.97 0.145
GaAIAs 2.2 0.97 0.145
469
The Photon Energy Converter
The key to the feasibility ofthe PENS is the photovoltaic Photon Energy Converter. Thecommon impression of photovoltaics is that they cannot be very efficient. This misunderstanding comes from the fact that photovoltaics are most commonly employed as "solarcells." And solar cells are not very efficient, ranging from 10-20% for commercial units andreaching as high as about 25% for laboratory cells. However the low efficiency is more dueto the characteristics of the solar spectrum than to the photovoltaics devices themselves, especially for the laboratory units with efficiencies of -25%. The problemwith the solar spectrum is that it is very broadband - its ratio of the average photon energy to the width(FWHM) of the spectrum (Em...,l~E) is about 1. This is good for color vision but quite badfor efficient energy conversion.For excimers, however, this ratio is greater than 10. Underthese conditions photovoltaics have intrinsic efficiencies of 75-95%.
Photovo1taic cells for use in photon-intermediate direct energy conversion of electricitywill require the development of a doped semiconductor material with a bandgap that matches the UV radiation. With such photovoltaic cells, a system efficiency of56% for fusion iondriven fluorescence has been projected [13]. Studies of fission ion driven fluorescence indicate that system efficiencies of about 40% are possible [12].
Photovoltaic Conversion of Narrowband fluorescence
For Xe2*, Em...,!~ =14,compared to a corresponding value of 1.3 for the AM2 solar spectrum. For a narrow distribution one can have E/Emean-1 and still have l1EG-1. A narrowbandspectrum will consequently have the highest intrinsic efficiency.
For a given spectrum, the efficiency of conversion is basically determined by the variationof the irradiance with photon energy and by the substrate bandgap energy, Eg, of the photovoltaic converter. Complete conversion (l00%) is not possible because of the width ofthe solar spectrum. This leads to two competing effects on the efficiency. The first effect isthat the energy of all photons with quantum energy hV<Eg is lost because they do not havesufficient energy to excite electrons from the valence band to the conduction band. Thepower density lost in this case is given by
E,
Plost = fW (E) dEo
(eq.1)
where W(E) is the irradiance in W/cm2/eY. Thus, the lower the bandgap of the photovoltaicconverter, the larger the fraction of the total spectrumconverted. Competing with this effecthowever is the fact that, for the photons with quantum energy hV>Eg that do contribute, thephoton energy in excess of the bandgap energy is lost. Thus the maximum intrinsic efficien-
470
cy for photovoltaic conversion is assuming an ideal collection device is given in equation 2.
..,. =In
- Ef (W(E) ;dE)E,
fW(E)dE
E,
(eq.2)
Typically the effect of the details of the solar radiation spectrum on calculating overall conversion efficiency is translated into a photon flux density, which then relates to an idealshort circuit current density. This is convenient because it is a good assumption that eachphoton absorbed and collected effectively causes one electron to move around the circuit.Also after each electron thermalizes, that is, gives off energy in excess of Eg to the lattice,it contributes maximally a constant Eg in energy to the overall process [13].
Figure 4. The theoretical maximum conversion efficiency of photovoltaic cells with various bandgap energies using either a solar spectrum or a Xe2* spectrum. Bandgap energies associated with Si, GaAs, diamond, and AIN are shown [13].
This contribution is conveniently modeled in the photovoltaic device using the ideal Shock-
471
ley model for the p-n junction. Using these concepts the intrinsic conversion efficiency can
be written in the following terms [13],
JNph (E)EdE£,
=~N(E>Eg )
Emean N tot(sq. 3)
where Npb is the photon flux density, in #/s-cm2-eV, N(E>Eg) is the photon flux in the interval E>Eg, Nto• is the total photon flux, and Tleg is the fraction of photons with E>Eg•
Figure 4 is a plot of maximum efficiencies for a p-n junction converter versus bandgap energy of the converter substrate material. Two plots are shown, one for an AM2 solar spectrum and one for a Xe2' spectrum. The previously derived equations were used to calculatethese curves. Also shown are vertical lines representing the bandgap energies of the twomaterials theoretically predicted to maximally convert these two spectra. Lines representing Si and diamond are shown for comparison. The approximately 30% maximum isthought to be an upper bound on the ability of a single material junction to convert the solarspectrum. The highest known conversion efficiency for silicon, to date, has been 26% obtained with a highly optimized MIS solar cell.
In contrast to the relatively low values for conversion of the solar spectrum, it can be seenfrom Figure 4 that efficiencies as high as 80% can theoretically be obtained using a p-njunction and converting the X~' spectrum. Although it is still speculative about whether ornot high quality p-n junctions can be made in materials with bandgaps above 4 eV, it is encouraging to note that high conversion efficiencies are possible.
Wide Bandgap Photovoltaic Materials
Table I lists several potential wide bandgap materials. Table I matches the more efficientand desirable fluorescers to materials with appropriate bandgaps. The theoretical maximumintrinsic photovoltaic efficiency (the ratio of the bandgap to the mean photon energy rangesfrom 75% to 95%) while the corresponding theoretical maximum efficiency for conversionof ion energy to electrical energy (the product of the photovoltaic efficiency and the fluorescence efficiency) ranges from 30% to 45%. If the most optimistic reported values of thefluorescence efficiency were used, the maximum ion-to-electric efficiency would increaseto 56%. The outlook for such cells is hopeful [11]. Rare-gas halide excimers have lowerphoton energies (3.5 eV for XeF*, 5.0 eV for KrF·, and 6.4 eV for ArF*) and, while theirfluorescence efficiency may be lower than that of the rare-gas excimers, their photon energy falls in the range of well known semiconductor materials.
Radiation damage to the photovoltaics from X-rays and neutrons is a concern. However, it
472
is well known that ionizing radiation appears to enhance fluorescence absorption efficiencyin continuous-wave insulating crystal lasers. Additionally, if a radiation damaged crystal isthermally annealed, the damage disappears. Potential crystalline photovoltaic materialswhich may experience the positive effects described above exist (Diamond and AluminumNitride).
Integration of Ion Source to Photovoltaics
Ion Source
There are many potential ion sources which can be used for the conversion method. For example, it is possible to use fission reactions (e.g., U233(n,ffl)ffh' U23\n,ff1)ffh, or Pu
239(n,_ff1)ffh), fusion reactions (when such sources become feasible), or radioisotopes (e.g, A~9,Kr85, Sr90, Po2IO, Pu238, etc.). We have chosen in this discussion to focus on the use ofradioisotopes for scaling estimates (see Figure 5).
In these studies the parameter which influenced the systems scale and power source lifetime was the radioisotope halflife (Kr85_10.76 yrs, Sr90-29 yrs, Po21O_0.38 yrs, and Pu238_
87.74 yes).
Both the scale size and the gamma ray emitted from the reaction influenced the system'smass. We were conservative in our estimate of a personnel radiation shield by requiring thatthe contact radiation be less than 2.5 millirems per hour (e.g., Kr85_O.514 MeV yO.38% ofdecays, Sc9°-no y, Po2IO_O.802 MeV yO.OO11% of decays, and Pu238_0.567 MeV y 5x1O5% of decays).
Coupling of Ion Energy to the Fluorescence Source
Based upon the above criteria, we believe that compact power sources can be made fromradioisotopes. The radioisotope ion source can be gaseous (e.g., Kr85 with a half life of10.76 years which emits a 0.67 MeV beta 100% ofthe decays and a 0.514 MeV gamma0.38% of the decays), or it can be in the form of a solid (e.g., a thin film, embedded in afiber, or as an aeroso1--See Figure 1). The types of solid radioisotopes which can be usedin the mobile power system are: Sr90 with a half life of 29 years which emits only a 0.67MeV beta; Po2lO with a halflife of 138.4 days which emits a 5.305 MeV alpha 100% of thedecays and a 0.803 MeV gamma 0.0011% of the decays; and Pu238 with a half life of 86.4years which emits a 5.5 MeV alpha 100% of the decays and a 0.77 MeV gamma 1x1O-5%ofthe decays. The Pu238 source was used to power the RTG used in the Voyager spacecraft.
Three RTG units produced 7200 watts of thermal power and 540 watts of electricity regulated to 30 volts from 12,900 grams of the isotope. Using the PENS concept, 12,900 gramsof Pu238 could produce 2,616 Watts of electrical power.
473
Kr-852.5
15 ye20
usefullife 7.5
15cm2.5
10
7.5 0.1 0.2 0.3 0.4
2
o
8
6
4
Electric Power kWPo-211 year 14useful 12lifecm 10
Sr-9000
50
20 year 8025useful
00 life 60
75 c 40
5020
25
Electric Power (kW)75 Pu-238 0
50 100 yr. 0
25 useful00 life75
50
Figure 5. Estimation of PENS mass, scale, and decay are shown for Kr85, Sr90, Po21O, andPu238. The geometry is assumed to be spherical with a diameter equal to the system scale estimation. Photovoltaic cells are assumed to surround the fluorescermedia and the vessel is shielded with lead.
474
Making an integrated system which produces a significant power density and remains optically thin is a challenge. The use of thin films, fibers, or aerosols along with thin aluminumor silver coatings have been examined in Prelas, Boody, Kunze, and Miley 1988. Withsolid materials, an average atomic density of approximately 5xl019 atoms cm-3 can beachieved for a reasonably optically thin system. The optical transport properties of films,fibers, or aerosols can be enhanced by a thin coating of reflective material [12]. Chargedparticles can penetrate the thin reflective coating without losing significant energy or significantly effecting the coating [12].
References
1)M. A. Prelas, "A Potential UV Fusion Light bulb for Energy Conversion", Bult. of theAm. Phys. Soc., 26(1),1045, 1981: See also Inside R&D, Vol. 10, Number 41 (Oct.14,1981).
2) M. A. Prelas, F. P. Boody, M. Zediker and M. Rowe, "A Direct Energy ConversionTechnique Based on an Aerosol Core Reactor Concept", 1984 IEEE Int. Conf. on Plasma Sci., IEEE Publication Number: 84CHI958-8, 38, 1984.
3) M. A. Prelas, E. J. Charlson, F. P. Boody, and G. H. Miley, Prog. In Nuclear Energy, 23(3), pp. 223-240 (1990).
4) D. J. Mencin and M. A. Prelas, "Low Temperature Gaseous Core Reactors Using Uranium Particles Trapped in C60 Cages," Proceedings ofNuclear Technologies for SpaceExploration," Sponsored by the Idaho Section of the American Nuclear Society, SnowKing Resort, Jackson Hole Wyoming, August 16-19, 1992.
5) Zhou K, Charlson E. M., Charlson E. J., Meese J., Stacy T., Popovici G., and Prelas M,Appl. Phys. Lett., 61 (9),1119-1121 (1992).
6) C. K. Chen, B. Nechay and B. Tsaur, IEEE Trans. on Electron Dev. 38,1094 (1991).
7) T. Stacy, et. aI., "Rectifying Contact Formation with Indium on Polycrystalline p-typeHot Filament CVD Diamond Utilizing Molecular Ion Implantation," Accepted forPublication in Journal of Applied Physics (February, 1993).
8) Galina Popovici, T. Sung, M. Prelas, and S. Khasawinah, "Evidence for n-type Diamond," Gordon Conference on Diamond Films, June 1994.
9) W. M Yim, E. J. Stofko, P. J. Zanzucchi, J. I. Pankove, M. Ettenberg, and S. L. Bilbert,J. Appl. Phys., 44, 29 (1973)
10) R. F. Davis, Proc. IEEE, 79, No.5, 702 (1991).
11) M. Prelas, E. J. Charlson, E. M. Charlson, J. Meese, G. Popovici, and T. Stacy, "Diamond Photovoltaic in Energy Conversion," Second International Conference on theApplications ofDiamond Films and Related Materials, M. Yoshikawa, M. Murakawa,Y. Tzeng and W. A. Yarbrough editors, MYU, Tokyo (ISBN 4-943995-07-1), 329-334(1993)
12) M. A. Prelas, F. P. Boody, G. H. Miley, and J. Kunze, "Nuclear-Driven Flashlamps",Lasers and Particle Beams, 6(1), 25, 1988; also, M. A. Prelas, and S. Loyalka, Progressin Nuclear Energy, 8, 35-52, 1981
13) M. A. Prelas, E. J. Charlson, E. M. Charlson, J. Meese, G. Popovici, and T. Stacy, "Diamond and Diamond Like Substrates as First Wall Materials in Inertial ConfinementFusion," Lasers and Particle Beams, 11(1),65-79 (1993).
CONSIDERATIONS IN FURTHER DEVELOPMENT OF ALUMINUMNITRIDE AS A MATERIAL FOR DEVICE APPLICATIONS
T.STACY,B:Y.LLAW,A.H.KHAN,G.ZHAO·Department of Electrical and Computer Engineering, University of Missouri, Columbia,MO 65211, USA* R&D Center, ENDEVCO, 355 N. Pastoria Ave., Surmyvale, CA 94086, USA
Abstract
Development of aluminum nitride as a wide bandgap semiconductor is in itsinitial stages. An overview of this unique material is presented and some of the majorissues in materials deposition, characterization, as well as device applications arediscussed. Consideration of defects, dopants, and initial results on development ofrectifying junctions are presented as well.
1. Introduction
Interest in aluminum nitride for potential device applications has existed forseveral decades. However, advances in materials growth and processing techniqueswithin the last few years have resulted in a renewed interest originating from severalperspectives. Aluminum nitride naturally crystallizes into the wurtzite structure witha large direct bandgap (Eg=6.2 eV). It is at the high bandgap binary end of acontinuum of materials that could potentially result from combinations with two otherdirect bandgap wurtzite structure III-V nitrides, InN (Eg=1.9 eV) and GaN (Eg=3.4eV). Therefore ternary and quaternary structures of the III-V nitrides can be utilizedfor optical devices operating in wavelengths ranging from red to well into theultraviolet region of the spectrum. Review articles on the III-V nitride materials giveboth a historical perspective and overview of more recent developments in thesematerials [1-3]. Zincblende forms of these nitrides have been studied theoretically andexperimentally to various degrees and material properties are beginning to bemeasured. The cubic structures represent new materials and properties may varysignificantly from the wurtzite forms. For example, cubic AIN is expected to be anindirect bandgap material with a bandgap of 5.1 eV and lattice constant of 433 A[4].Zincblende polytypes of these nitrides, as well as cubic boron nitride (c-BN, indirectEg=6-6.5 eV) are metastable and require non-equilibrium processing conditions.However, the zinc-blend crystal structures are expected to be more amenable to dopingin both conductivity types [5]. Further interest in AIN also has impetus from the
475
M.A. Prelas et at. (eds.), Wide Band Gap Electronic Materials, 475-486© 1995 Kluwer Academic Publishers.
476
success that has occurred in development of other high bandgap materials for deviceapplications, including SiC ( cubic 3C Eg,~=Z.2 eV, hexagonal 6H Eg,a=Z.86 eV) anddiamond (Eg=55 eV). At this time, little work has been done to develop AlN as asemiconductor for general electronic and opto-electronic applications. However,solution of growth and processing problems in other materials systems suggest thatAlN development as a potential semiconductor is worth pursuing.
2.0 Properties
Wurtzite aluminum nitride has several properties that make it desirable forelectronic, electro-optic, and acoustical device applications. It has a high electricalresistivity (>1011 O-cm), large dielectric constant (10
0"",85, 10,,,,,,4.7), and index of
refraction n "",Z.15 [2]. AlN has excellent acoustical properties, with a high surfaceacoustical velocity (Raleigh VR=6-6.Z kIn/s, longitudinal VL=1l-12 kIn/s) [1]. Inaddition to the large bandgap, it is a mechanically hard, chemically inert, and thermallystable material that is therefore well suited for high temperature, high pressure, harshenvironments. Table 1 contains a comparison of some AlN properties with those ofcubic boron nitride and diamond [1,2,6,7]. These are the two large bandgap materialsalso suitable for some of the same types of severe applications. As seen from Table1, in some properties, the other materials are superior to AlN. The major advantagesof AlN, notwithstanding doping and other materials development issues that will bediscussed later, are in acoustical and optical device applications. Additionally, singlecrystal growth ofAlN and c-BN have been achieved, therefore large area single crystalstructures appear possible. The issue of how severe an environment AlN can toleratehas not been entirely determined. One study irradiated several silicates and severalnitrides, including AlN, with 1019_l(fl Krypton (3MeV) ions/m2 [8]. Transmissionelectron microscopy and x-ray diffraction data were used to conclude that grossstructural changes to crystalline structure occurred with the silicates, but not with thenitrides. Recent experimental studies and theoretical calculations support that a highpressure phase transition to a cubic rocksalt structure (indirect Eg=8.9, a=3.982 A)occurs at 16-25 GPa [9].
3.0 Material Growth
Growth of wurtzite AlN (a=3.11ZA, c=4.98ZA) occurs on most substrates,even cubic substrates, under most processing conditions. From work that has beendone on cubic GaN, it appears that the nitride zincblende structures form underspecialized growth conditions and with specific substrate conditioning [10-17]. Forexample, Mizuta et. al. showed that pyrolytic CVD of hydrazine and trimethyl galliumon (111) GaAs resulted in zincblende GaN growth [10]. More recently, Miyoshi andothers obtained excellent single crystal zincblende GaN grown on (100) GaAs withmetalorganic chemical vapor deposition (MOCVD) utilizing dimethylhydrazine as thenitrogen source [11]. In another study, zincblende GaN growth occurred in a
477
microwave plasma assisted molecular beam epitaxy (MBE) system with a two-stepgrowth process that includes low temperature GaN monocrystalline buffer layer growthonto (001) Si prior to high temperature GaN epitaxy [13]. Investigations of zincblende
Table 1. Comparison of wurtzite AlN properties to Diamond and cubic BN(from references 1,2,6, and 7).
w-AlN diamond c-BN
Bandgap (eV) direct, 63 indirect, 5.5 indirect, 6.4
Knoop hardness 1,800 5,700-10,400 4,500(kg/mm2
)
Thermal 3.2-3.7 20, single crystal, 13conductivity 25%-50% of this (theoretical)(W/cm-°C) value for
polycrystalline
Melting Point >2275 K Graphitization: 898 Converts toK in oxygen, h-BN,1573-1793 K in vacuum or 1773 Kinert gas
Thermal Expansion 4.5 avg, 0.8 (300 K), 1.2 (400- ----Coefficient c: 53, a: 4.2 1200 K)(xlO-6rC)
Intrinsic Resistivity >1011 >106 (as-grown >1010
undoped,processingdependent)
AlN growth on cubic SiC have resulted in wurtzite AlN, although lattice mismatch isless than 1% [18]. Epitaxial zincblende AlN has been synthesized by a solid-statereaction between Al and TiN deposited on MgO [19]. The layers were deposited bysputtering of Al and Ti targets with argon and nitrogen, respectively. Depositionfollowed by furnace annealing resulted in zincblende AIN and Al3Ti layer formationbetween the Al and TiN layers.
Wurtzite AlN has been grown on a variety of substrates including variousorientations of (sapphire) a-Alp3' Si, GaAs, MgO, and a(6H)-SiC. Film morphology
478
varies from amorphous to monocrystalline, depending on the substrate and processingconditions. Despite the lattice mismatch to silicon @d sapphire, single crystal AIN wasreported by Yoshida and others on (0001) and (0112) sapphire and Si (Ill) betweenlOOO°C and 12OO°C [20]. The transparent films were 0.1-4 JLm thick and had electricalresistivity higher than 1013 O-em. Of these, the best lattice match with AIN occurs onhexagonal SiC , where the lattice mismatch is less than 1%. Monocrystalline AIN(0001) of high quality has recently been grown on hexagonal SiC (0001) by plasmaassisted gas-source MBE at temperatures ranging from 1050-1200 °C [21]. The AI wasevaporated from an effusion cell and an electron cyclotron resonance plasma sourceproduced active nitrogen for film formation.
A variety of methods have been used for aluminum nitride growth includingmolecular beam epitaxy (MBE), metalorganic chemical vapor deposition (MOCYD),and chemical vapor deposition (CYD). These methods have been used in combinationwith other assisted growth methods, including reactive plasma and hot-filament assistedgrowth. Growth with MBE has included a standard effusion cell for aluminumevaporation and nitrogen sources of NH3 or an ECR plasma reactive nitrogen beam[20,22,23]. The ECR plasma source designed specifically for MBE deposition systems,has given excellent results in deposition of GaN/AIN superlattice layers on a-SiC [24].Most MOCYD of AIN occurs at high temperatures (~ 1000 0c) withtrimethylaluminum (TMA) and ammonia in the presence of hydrogen [25,26]. Morerecent efforts have included photoassisted, plasma activated, and hot-filament assistedreactions and the use of other metalorganic precursors in order to permit lowering ofdeposition temperature [27-29]. The effort to lower process temperature is directedat reducing nitrogen vacancies. Monocrystal1ine AlN films have resulted using CYDwith NH3 and AlCI3 as sources. One study used (0112) and (0001) sapphire substratesto grow single crystal AIN at temperatures of 1000-1100 °C [30]. Another groupachieved single crystal growth on hexagonal SiC at temperatures ranging 1200-1250 °C[31]. Studies using the same reactants and (Ill) Si substrates showed that polycrystalsize increaseswith temperature in the range 800-1200 °C [32]. Other methods that haverecently been investigated include reactive sputter deposition [33,34].
Recently, a number of devices have been made using the alloy AlxGa1.)J andGaN. A notable success is p-type doping in wurtzite [35] and zincblende [36] GaNusing Mg. Prior to this achievement, there was doubt that p-type doping could beachieved because nitrogen vacancies in GaN cause n-type conductivity with typicalbackground concentrations of ]018 cm-3• The p-type doping of the AlGaN has also beenachieved with Mg in low aluminum fraction layers ( < 0.2) [37]. The devices fabricatedhave had many accomplishments and excel1ent results. However, we will not considerthese here since we are specifically focused on AIN.
4.0 Semiconductor Development
Little work has been done to develop AIN as a wide bandgap semiconducting
479
material. Sparse information exists on properties of metal contacts or on dopants.Aluminum nitride has mostly been investigated for device applications in two majorareas: as an insulating film for MIS structures or as passivating layers [38], and forsurface acoustic wave applications [34,39,40]. More recently, its use as a buffer layerin GaN/AlGaN devices has also been considered. To our knowledge, the earliestreport of use of AlN as a semiconducting layer in a device was in anelectroluminescent device that emitted from 215 nm in the ultraviolet into the blue endof the spectrum at applied voltages ranging from 17V to 150V [41]. The AlN wasgrown by a pyrolytic vapor transport technique on either tungsten or sapphire, bothwith a nonluminescent AlN layer which had been deposited by reactive sputtering.Substrates were placed face down on a polycrystalline sintered AlN wafer and thesewere heated to 1850 °C in forming gas. Electrical contact were made by etching thetungsten and sputtered AlN layer and alloying Nb and Al contacts. For the sapphiresubstrates, topside alloyed aluminum contacts were formed. The sintered AlNappeared blue in color and was highly resistive, while the grown layers were amber orclear with resistivities ranging from 400-1000 O-cm. Sample conductivity was n-type,as measured with a hot-probe, and x-ray analysis indicated strained single crystal AlNwas grown. The source of sample conductivity was not identified. The color andresistivity relationship found by Rutz differed from an earlier study by Edwards andothers [42]. In that work colorless crystals were found to have resistivities> 1012 0cm, while blue crystals had resistivities of 104 O-cm and had p-type conductivity. Theyattributed the blue color to the presence of Al2OC. A Hall mobility of 14 cm
2/V-smeasured at room temperature was viewed with caution due to small sample size andscatter in the data taken. Another early report of doping in AlN used Hg and H2Se toachieve p-type and n-type doping, respectively [31]. More recently, others havesuggested dopants for AlN: C for p-type and 0 for n-type doping [21].
Several workers have considered both defect and dopant atoms in InN, GaN,and AlN and theoretical determination was made of location within the energy gap [43,44]. Native defects include the antisite defects, N on a cation site (NAl) and a cationon the N site (AlN), and the nitrogen and aluminum vacancies, VaN and VaAl'respectively. These are neutral native defects described as being "s-like" (with twodegenerate spin states) and "p-like" (with six degenerate spin states, assuming the Px'Py' and pz splitting is small). In addition to the s-like and p-like character, theoccupancy and energy bandgap location of the defects was calculated. This informationcan then be used as a guide on whether a defect is a trap level for electrons or holes,or whether they contribute to material conductivity. Their treatment predicts theobserved natural n-type conductivity of InN and GaN and confirms that the cause ofthe conductivity is due to a nitrogen vacancy. For AlN, the nitrogen vacancy occursas a p-like level within the gap (::::::0.5-0.8 eV below Ec) making the neutral N vacancya deep electron trap. The VAl acts as a hole and electron trap at approximately 13-1.5eV above the valence band. The antisite defects similarly do not contribute to materialconductivity, with levels in the bandgap far from the band edges and levels in theconduction and valence band that do not contribute to conductivity. Comparison ofthe theoretical study to experimentally observed luminescence [44] shows a degree of
480
agreement, but is not perfectly correlated.
The same analysis was used to predict energy levels and occupancies ofdifferent impurities at different lattice locations: neutral impurities from columns I,II, and IV at the AI site, neutral column IV and column VI impurities at the N site,and isoelectronic impurities at both the AI and N site. The results from this study arepresented in Table 2. From the theoretical study it appears that several candidates forn-type and p-type dopants exist. Whether these can actually be incorporated aspredicted by the study remains to be determined.
Table 2. Summary of theoretical study of various impurities in the AIand N site of aluminum nitride.
Column of impurity AI site: results N site: resultsatom
I shallow (double)acceptor
II shallow acceptor
III inert
IV shallow donor deep e- & h+ trap(p-Iike)
V hole traps
VI donor ( and S,se,Te haveoccupied p-like levels @band-edge)
The effect of oxygen on AlN films has been shown to have a significant effecton film properties. Discrepancies in early measurements of aluminum nitride bandgaphave been attributed to oxygen contamination [2J. More recently, light inducedchanges to aluminum nitride have been studied [45-47J. Luminescence data ofaluminum nitride has demonstrated a variation in both intensity and wavelength of thepeak intensity as a function of oxygen concentration. The luminescent peak positionincreased in wavelength from 305 nm to approximately 395 nm as oxygen contentincreased from 0.1 to 0.75 atomic percent. Further increase in oxygen content did notaffect the peak position at 375 nm. Moreover, as the oxygen content increased beyond0.75 % , the luminescence intensity also increased. The model that has been used to
481
explain this effect is one in which the AIN cell volume contracts as oxygen is dissolvedup to 0.75 %, and then re-expands beyond that concentration. These studies have alsoshown that oxygen defects are responsible for AIN visibly darkening when exposed toultraviolet light.
5.0 Rectifying Junctions
We are studying several aspects of development of AIN as a semiconductingmaterial, including study of as-grown film conduction [48J and AIN surfacemodification [49], which we will discuss here. Our films are oriented polycrystallinematerial grown in a tubular CVD reactor by the reaction of AlCI3 and NH3 in thepresence of hydrogen at substrate temperatures of 700-800 °C and pressure of 35-40Torr. More detailed processing conditions have been described elsewhere [50]. Wehave grown AIN on single crystal (111) n-type (3-15 O-cm) to thickness ranging from200nm to several microns. Current-voltage measurements made on MIS junctionsindicate our films have bulk resistivities of 101°_1011 O-cm when AIN thickness is 200nmand "'" 1013 O-cm when AIN thickness is Itlm.
We conducted a study of the effects of low energy ion bombardment on anAIN layer prior to MIS junction formation. A d.c. corona discharge was used to ionizemethane and bombard the as-grown A1N. Samples were partially masked with acleaned and scribed silicon sample so that comparisons could be made on adjacentsample areas. A total of 31 gold contacts were sputtered onto the aluminum nitride,and a back contact was made by thermal evaporation of Al onto a roughened Sisurface. Of the 31 front contacts, 19 were placed on areas previously bombarded withmethane, and the remainder were placed on the previously masked A1N' Currentvoltage measurements show rectifying behaviorwith forward bias when positive voltageis applied to the A1N. Although not all contacts were rectifying, good rectifyingbehavior was only observed on bombarded areas. Quality of the rectifying junction canbe quantified by graphing log J vs. V and taking the ratio of forward to reversecurrent, IF/IR• Forward to reverse current ratios varies by 5 orders of magnitude ondifferent sample positions,ranging from values of 20 to 106• When low current densityvalues are eliminated from the calculation, the range is 500 to 106• Midrange values(500-1000) and higher current densities were observed on several positions. Weattribute the rectifying behavior to defects created by ion bombardment. Rectifyingbehavior with sputtered gold contacts has also been observed when samples arebombarded with other species, including nitrogen and argon [51J. Additional studiesare being conducted to further understand and characterize these experimental results.
Acknowledgements: We would like to acknowledge funding from the Department ofEnergy under grant DE FG02-91ER12107 and from the University of MissouriResearch Board under grant RB 94-076.
482
References
1. Davis , R F. (1991) III-V nitrides for electronic and optoelectronicapplications., Proc. IEEE, 79, 702-712, and references therein.
2. S. Strite and H. Morkoc (1992) GaN, AlN, and InN: A review, J. Vac. Sci.Techno/. B, 10, 1237-1266, and references therein.
3. J. H. Edgar (1992) Prosects for device implementation of wide bandgapsemiconductors, J. Mater. Res., 7, 235-252.
4. W. R L. Lambrecht and B. Segall (1991) Electronic structure and bonding atSiC/AlN and SiC/BP interfaces, Phys. Rev. B, 43, 7070-7085, and W. R L.Lambrecht and B. Segall (1992) A comparison of the wurtzite and zincblendeband structures for SiC, AlN, and GaN, Mat. Res. Soc. Symp. Proc., 242, 367372.
5. J. I. Pankove (1990) Perspective on Gallium Nitride, Mat. Res. Soc. Symp.Proc., 162,515-524.
6. M. N. Yoder, (1993) Diamond Properties and Applications, in R F. Davis(ed), Diamond Films and Coatings, Noyes Publishers, Parkridge, N. J., pp.l30.
7. O. Madelung and K.-H. Hellwege, eds. (1982) Landolt-Bomstein Tables, 17a,Springer-Verlag, New York.
8. L. Cartz, F. G. Karioris, and R A. Foumelle (1981) Heavy ion bombardmentof silicates and nitrides, Rad. Effeds, 54, 57-64 (1981).
9. R Pandey, A. Sutjianto, M. Seel, and J. E. Jaffe (1993), Electronic structureof high pressure phase of AlN, J. Mater. Res., 8, 1922, and references therein.
10. M. Mizuta, S. Fujieda, Y. Matsumoto, and T. Kawamura (1986) Lowtemperature growth of GaN and AlN on GaAs utilizing metalorganics andhydrazine, Jap. J. Appl. Phys., L945-L948, (1986).
11. S. Miyoshi, K. Onabe, N. Ohkouchi, H. Yaguchi, R Ito, S. Fukatsu, and Y.Shiraki (1992) MOVPE growth of cubic GaN on GaAs usingdimethylhydrazine, J. Crystal Growth, 124, 439-442.
483
12. S. Strite, J. Ruan, Z. Li, A. Salvador, H. Chen, D. J. Smith, W. J. Choyke, H.Morkoc (1991) An investigation of the properties of cubic GaN grown onGaAs by plasma-assisted molecular-beam epitaxy,J. Vac. Sci. and Tec1mol. 89,1924-1929.
13. T. Lei, M. Fanciulli, R. J. Molnar, T. D. Moustakas, R. J. Graham, J. Scanlon(1991) Epitaxial growth of zinc blende and wurtzitic gallium nitride thin filmson (001) silicon, Appl. Phys. Lett., 59,944-946.
14. M. J. Paisley, Z. Sitar, J. B. Posthill, and R. F. Davis (1989) Growth of cubicphase gallium nitride by modified molecular-beam epitaxy, J. Vac. Sci.Technol., A7, 701-705 (1989).
15. H. Liu, A. C. Frenkel, J. G. Kim, and R. M. Park (1993) Growth of zincblende-GaN on I3-SiC coated (001) Si by molecular beam epitaxy using a radiofrequency plasma discharge, nitrogen free-radical source, J. Appl. Phys., 74,6124-6127.
16. T. Lei, T. D. Moustakas, R. J. Graham, Y. He, and S. J. Berkowitz (1992),Epitaxial growth and characterization of zinc-blende gallium nitride on (001)silicon, J. Appl. Phys., 4933-4943.
17. R. C. Powell, G. A. Tomasch, Y. W. Kim, J. A. Thornton, and J. E. Greene(1990), Growth of high-resistivity wurtzite and zincblende structure singlecrystal GaN by reactive-ion Molecular beam epitaxy, Mat. Res. Soc. Symp.Proc., 162, 525-530.
18. B. S. Sywe, Z. J. Yu, and J. H. Edgar (1992) Epitaxial growth of AlN on 3CSiC and Al20 3 substrates, Mater. Res. Soc. Symp. Proc., 242, 463-467.
19. I. Petrov, E. Mojab, R. C. Powell, J. E. Greene, L. Hultman, and J. E.Sundgren (1992) Synthesis of metastable epitaxial zinc-blende structure AlNby solid-state reaction, Appl. Phys. Lett., 60, 2491-2493 (1992).
20. S. Yoshida, S. Misawa, Y. Fujii, S. Takada, H. Hayakawa, S. Gonda, and A,Hoh, (1979) Reactive molecular beam epitaxy of aluminum nitride,J. Vac. Sci.Tec1mol., 16, 990-993.
21. L. B. Rowland, R. S. Kern, S. Tanaka, and R. F. Davis (1993) Epitaxialgrowth ofAlN by plasma-assisted,gas-source molecular beam epitaxy, J. Mater.Res., 8, 2310-2314.
484
22. Z Sitar, M. J. Paisley, B. Yan, J. Ruan, W. J. Choyke, and R. F. Davis (1990)Growth ofAIN IGaN layered structures by gas source molecular-beam epitaxy,J. Vac. Sci. and Technol., 88, 316-322.
23. H. U. Baier and W. Monch (1992) Growth of AIN on GaAs (110) by reactivemolecular beam deposition, J. Vac. Sci. and Technol., 810, 1735-1739.
24. Z. Sitar, M. J. Paisley, D. K. Smith, and R. F. Davis (1990) Design andperformance of an electron cyclotron resonance plasma source for standardmolecular beam epitaxy equipment, Rev. Sci. Instrum." 61, 2407-2411.
25. M. Morita, S. lsogai, N. Shimizu, K. Tsubouchi, and N. Mikoshiba (1981)Aluminum Nitride Epitaxially Grown on Silicon: Orientation Relationships,Jap. J. Appl. Phys., 20, Ll73-Ll75.
26. U. Rensch and G. Eichhorn (1983) Investigations on the structure ofMOCVDALN layers on silicon, Phys. Stat. Sol., 77, 195-199.
27. T. Y. Sheng, Z. Q. Yu, and G. J. Collins (1988) Disk hydrogen plasmaassisted chemical vapor deposition of aluminum nitride, Appl. Phys. Lett., 52,576-578.
28. K.-L. Ho, K. F. Jensen, S. A. Hanson, J. F. Evans, D. C. Boyd, W. L.Gladfelter (1990) MOCVD of wide bandgap III-V semiconductors by usingnovel precursors, Mat. Res. Soc. Symp. Proc., 162,605-610.
29. J. L. Dupuie and E. Gulari (1992) The low temperature catalyzed chemicalvapor deposition and characterization of aluminum nitride thin films, J. Vac.Sci. and Technol., A 10, 18-28.
30. W. M. Yim, E. J. Stofko, P. J. Zanzucchi, J. I. Pankove, M. Ettenberg, S. L.Gilbert (1973) Epitaxially grown AlN and its optical bandgap, J. Appl. Phys.,44, 292-295.
31. T. L. Chu, D. W. lng, and A. J. Noreika (1967) Epitaxial growth of aluminumnitride, Solid-State Electron., 10, 1023-1027.
32. T. L. Chu and R. W. KeIrn (1975) The preparation and properties ofaluminum nitride films, J. Electrochem. Soc., 122,995-1000.
33. W. J. Meng, J. A. Sell, T. A. Perry, L. E. Rehn, and P. M. Baldo (1994)Growth of aluminum nitride thin films on Si(111) and Si(OOl): Structuralcharacteristics and development of intrinsic stress, J. Appl. Phys., 75, 34463455.
485
34. H. Okano, N. Tanaka, Y. Takahashi, T. Tanaka, K. SHibata, and S. Nakano(1994) Preparation of aluminum nitride thin films by reactive sputtering andtheir applications to GHz-band surface acoustic wave devices, Appl. Phys.Lett., 64, 166-168.
35. H. Amano, M. Kito, K. Hiramatsu, and I. Akasaki (1989) P-type conductionin Mg-doped GaN treated with low-energy electron beam irradiation(LEEBI), Jap. J. Appl. Phys., 12, L2112-L2114.
36. M. E. Lin, G. Xue, G. L. Zhou, J. E. Greene, and H. Morkoc (1993), P-typezinc-blende GaN on GaAs substrates, Appl. Phys. Lett., 63, 932-933.
37. I. Akasaki and H. Amano (1992) Conductivity control of AlGaN, fabricationof AlGaN/GaN multi-heterostructure and their application to UV/blue lightemitting devices, Mater. Res. Soc. Symp. Proc., 242, 383-394.
38. M. Morita, S. Isogai, K. Tsubouchi, and N. Mikoshiba (1981) Characteristicsof the metal insulator semiconductor structure: AlN/Si, Appl. Phys. Lett., 38,50-52.
39. J. K. Liu, K. M. Lakin, and K. L. Wang (1975) Growth morphology andsurface-acoustic-wave measurements of AlN films on sapphire, J. Appl. Phys.,46, 3703-3706.
40. G. D. O'Clock (1973) Acoustic surface wave properties of epitaxiaUy grownaluminum nitride and gallium nitride on sapphire, Appl. Phys. Lett., 55-56.
41. R. F. Rutz (1976) Ultraviolet electroluminescence in AlN, Appl. Phys. Lett.,
28 (7), 379-381.
42. J. Edwards, K. Kawabe, G. Stevens, and R. H. Tredgold (1965) Space chargeconduction and electrical behaviour of aluminum nitride single crystals, SolidState wmm., 3,99-100.
43. D. W. Jenkins and J. D. Dow (1989) Electronic structures and doping of InN,InxGa1.,.N, and InxAll.,.N, Phys Rev. B, 39,3317-3329.
44. T. L. Tansley and R. J. Egan (1992) Optical and electronic properties of thenitrides of indium, gallium, and aluminum and the influence of native defects,Mat. Res. Soc. Symp. Proc., 242, 395-407.
45. J. H. Harris and R. A Youngman (1992) An investigation of light induceddefects in aluminum nitride ceramics,Mat. Res. Soc. Symp Proc., 242, 451-456.
486
46. J. H. Harris and R. A. Youngman (1993) Light-induced defects in aluminumnitride ceramics, J. Mater. Res., 8, 154-162.
47. J. H. Harris and R. C. Enck (1993) UV induced dielectric loss in ALNceramics, J. Mater. Res., 8,2734-2740.
48. A. H. Khan, J. M. Meese, T. Stacy, E. M. Charlson, G. Zhao, G. Popovici,and M. A. Prelas (1994) Electrical characterization of aluminum nitride filmson silicon grown by chemical vapor deposition, presented at the Spring MRSmeeting.
49. T. Stacy, B. Y. Liaw, A. H. Khan, G. Zhao, E. M. Charlson, E. J. Charlson,J. M Meese, M. Prelas, J. L. Wragg, J. E. Chamberlain, and H. W. White(1994) Ion beam surface modification for achieving rectification in goldaluminum nitride-silicon junctions, Mat. Res. Soc. Symp. Proc., 316, 355-360.
50. A. H. Khan, M. F. Odeh, J. M. Meese, E. M. Charlson, E. J. Charlson, T.Stacy, G. Popovici, M. A. Prelas, and J. L.Wragg (1994) Growth of orientedaluminum nitride films on silicon by chemical vapour deposition, J. Mater. Sci.,29, 4314-4318.
51. T. Stacy, et. aI., to be published.
NATO Advanced Workshop WIDE BANDGAP ELECTRONIC MATERIALS Minsk 3-5 May 1994
THEORETICAL ASPECTS OF ALUMINIUM NITRIDE ANDDIAMOND IN VIEW OF LASER AND PHOTOVOLTAlC
ACTION
HEINRICH HORA *, REINHARD HOPFL, AND MARK A. PRELAS **
Facuity of Electrical Engineering
Hochschule fUr Technik 93049, Regensburg, Germany
After the technological preparation of compact single crystals of aluminium nitride
and of layers of diamond of any n- or p-semiconducting property has been
demonstrated, the use of these wide band gap semiconductors for lasers and
photovoltaic cells is of interest. The very wide band gap of AIN should provide laser
action in the far ultraviolet spectrum (photon energies higher than 6 eV). Since the
threshold conditions are much higher for shorter wave lengths it is interesting to
evaluate the thresholds for an electron beam excited laser of this type. The numbers
calculated show that the first solid state laser for the vacuum-UV should be possible, on
top having a high gain as usual for this type of lasers. The same laser operation of
diamond requires a direct band semiconductor. Based on the work of Pickett and Mehl
and a quantum model of compressibility, we calculate the conditions for diamond
similar to the well known results from silicon. Highly p-doped diamond with a direct
band structure should be available after strong ion implantation causing stress energy in
the range of I eV per atom. The resulting diamond laser thresholds with electron beam
pumping are in a range comparable to lasers of this type for longer wave lengths. A
further application of wide band gap photovoltaic cells with very high conversion
efficency ofUV-light into electricity is described using the ganuna-free beta emitter Kr
85 of 10.65 years half life for a very compact source of electricity, light and heat.
487
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials, 487-509© 1995 Kluwer Academic Publishers.
488
I. Introduction
It was typical for chemistry of the past to look into processes occuring close to
thermodynamic equilibrium under which conditions practically all technological
processes in chemistry are performed and which led to the most impressive world of
todays chemical industry. In this sense also the question how to produce the diamond
modifaction of carbon against the graphite modification was considered to go through
the very high pressure (60000 atmospheres) and high temperature (1500 degrees
Celsius) conditions where diamond is the stable phase against graphite. This is possible
with carbon dissolved in molten transition metals which work as catalyst and as solvent
such that diamond nuclei are formed and crystals are growing [I]. This technology at
General Electric (a one year earlier sucess of this technique was performed at the
Swedish Allmanna Svenska Elektriska Aktiebolaget (ASEA) ) was developed to
produce today about 100 tons of diamond per year.
Any low pressure diamond sysnthesis was regarded as "equivlent to transmuting lead
into gold - a vioation of fundamental principles" since graphite is more stable than
diamond at low pressures and "workers who attempted it put their careers in jeopardy"
[2]. The fundamental principles of thermodynamics away from euqilibrium - just away
as a small perturbation - was just being developed during this time which ran into
numerous difficulties of which at least some points of solutions were honoured with the
Nobel Prize to IvanPrigogine.
For physicists working with extreme nonequilibrium conditions as given with intense
electron beams [3] as well as with growth of crystals using epitaxy including the thermal
or electron collision produced dissociation of gaseous molecules of carbon compounds,
it was a systematic consequence that diamond epitaxy should be possible at low
pressures if such extreme nonequilibrium conditions are used. This scheme of 1962,
developed at IBM, resulted in a disclosure of patents [4]. These views - though as a
result of systematic sciences - were played down as "patent literature in which some
claims bordered on the ludicrous".
An intuitively invented method of having a diamond seed crystal next to a high
temperature filament within a gaseous carbon compound at the just mentioned extreme
non-equilibrium conditions led to the CHEMICAL VAPOUR DEPOSITION (CVD) of
low pressure growth of diamond crystals. This was experimentally approached by G.
Eversole at Union Carbide in 1952 [5] whose methods were proposed also by Boris V.
489
Spitsyn 1956 at the Moscow University [6] and by John C. Angus 1959 at the University
Michigan who in 1967 succeeded following the Eversole's method to produce syntesized
diamond as proved - despite the small growth rate - from the blue colour when adding
diborane to the gas while the convincing proof for this slow growing of synthesized
diamond films is dated 1971 [2].
From then on the wide range development of synthetic diamond films and thicker
monocrystalline layers even on amorphous substrates was developed on a nearly
industrial volume. This can be seen in several reviews [7-9]. The strategic importance of
diamond surfaces of materials was not only with respect to hardening of surfaces against
direct radiation weapons in the time of SOl but even in important other technologies.
E.g. the covering of crucibles for melting glass simply produced high quality laser glass
[10] while the use of platinum crucibles without the diamond film coverage resulted in
glass with inclusions of platinum metal particles of about micrometer diameter. Such
glass - after completely installed - could not be used in the very big NOVA neodymium
glass laser at Livermore and had to be replaced, costing many million dollars.
The following contribution is devoted to some theoretical aspects how the diamond
and similar larger size crystals like aluminium nitride produced by the same CVD
method may be used as lasers or as photovoltaic materials.
2. Efficient Electron Beam Pumped AIN Laser: first Solid State Laser for the
Vacuum Ultraviolet
Most of the solid state lasers emit in the red or infrared optical range since the levels
for laser action are mainly those of impurity atoms in otherwise wide bandgap host
crystals. A step to shorter wave length solid state lasers was possible by using
semiconductor crystals with pumping by injected electron beams of about 50 keY
energy. As it was theoretically discussed [II] and clarified in details [12], one can
achieve the practical working conditions only with highly p- doped crystals and not with
n-doped material. This semiconductor condition for polar crystals could not be extended
to heteropolar crystals like LiF or aluminium oxide or quartz and glasses as entire lasers
following similar schemes. The first successful laser emission of highly p-doped polar
semiconductor lasers was achieved with GaAs [13] with infrared photons of 1.47 eV
energy (corresponding to 840 nm wave length).
490
This type of laser could then be used for shorter wavelengths in the visible {CdSxSeI_x
with 1.79 eV (690 nm) [14] ; GaAsxPl-x with 2.024 eV (6I2nm) [15]; CdS with 2.5 eV
(490 nm) [16] and even in the ultraviolet range { ZnS with 3.7 eV (330 nm) [l7]}. Since
this type of solid state laser works with high efficiencies, e.g. 35% for the blue laser of
CdS[ 16], it is especially interesting what possibilities will exist for using highly p
conducting materials with a sufficient degree of polar binding.
A candidate for this purpose is aluminium nitride [18] which crystals are light blue and
otherwise transparent in the visible range. Aluminium nitride, although similar to
gallium nitride, has demonstrated some promising results for the formation of p-type
direct bandgap material using an in-situ growing process[l9] . Sufficiently high levels of
dopant concentrations makes interesting the process of electron beam pumped lasers.
Such a laser would have a band gap corresponding to 6.25 eV photon energy (198 nm).
This value includes the upshift of the band gap at intense electron beam irradiation in
analogy to the case of silicon [20].
We calculate now the treshold conditions for optimized dimensions of these electron
beam pumped lasers following the early calculations [21] which predicted numbers [12]
were experimentally reproduced [13]. The calculated optimum values were very close to
those of the very first experiments. There are, however, several difficulties in the
prediction for AIN since the effective masses of the electrons and the holes are not
available - which results in a rather minor error in the following estimations- and since
the complicated spontaneous absorption has not jet been measured. For the last case we
shall evaluate a wide range of parameters and look what conditions for the minimum
current density of electron beams are required.
Because of the high dielectric constant the losses by diffraction will be low for the
thickness d of the laser cavity of up to several micrometer. We use the same assumptions
for electron-hole pair production as before [21] with respect to the agreement of the
results [13] where the effective number of produced pairs per incident 50 keY electrons
is just compensated by the assumption of unmodified exit reflectivity of the cavity. We
find then for the minimum elctron beam current density the threshold
(I)
where CI is given for a wave length of I98nm. The difference to the wave length of
GaAs alone increases the threshold current density by the factor of 17.9. In Eq. (I), n is
(2)
491
the optical refractive index for the laser wave length of 198 nm, R is the reflectivity of
the ends of the cavity, I is the length of the cavity andy is the band width for the laser
operation determined by the sum of the quasi-Fermi levels of the conduction and of the
valence band determined by the hole density np for a temperature of degeneracy. Since
the effective masses are not known for AIN we use the vacuum mass, m, of electrons
h (3np)~Llv=- -2m 81t
where h is Plank's constant.
Since the number of pairs produced per 50 keY electron should be larger than one, we
introduce a factor N which represents the number of pairs produced per 50keV electron.
We assume that the maximum number of pairs per electron is,
(3) N =E.E;
where Ee is the energy of the incident electron and Ei is the pair production energy. The
value Ee is higher than the band gap taking into account the larger difference in the
energy bands for higher momentum vectors /s... In this case Ee=50keV, and Ei=8eV. Eq.(1) can be rewritten.
. -8 1 Llvd(4) j>3.03·10 N(l-R)-/-L.
Here the loss by usual optical absorption in the crystal, La is defined as in reference [12]
(5)
where a. is the absorption coefficient. The absorption coefficient is written in the form of
[12],
(6)
which is the empirically derived expression for polar semiconductors [22].
The constant C2 is not known for aluminium nitride but in a chemically similar material
GaAs it is 1x10-21 [12] which will be used in the following.
492
Eq. (4) becomes,
(7)
Using Eqs. (6) and (7) and a value for the reflectivity of the one end laser cavity, R of
0.995, and assuming a value for d of 2 /lm, the threshold current density for an e-beam
pumped aluminium nitride laser can be found as a function of I and np' This result is
shown in Figure I.
The resulting threshold current densities even for optimized cavity lengths and hole
densities are rather high. This is not unexpected since the current density goes with the
inverse square of the wavelength [12]. Since electron beam currents of a few hundred
A/cm2 were easily tolerated when using GaAs [15] without destruction or damage of the
crystals, we can assume that the stronger chemical bound of the AIN crystals may easily
permit a ten times higher electron beam current density. A similarity is given by the fact
that the laser damage threshold for diamond is about ten times higher than for materials
with metallic or heteropolar bounds.
One example for a laser condition with the 50 keY electrons would be a current
density j=1.44 kA/cm2 for a cavity length of 6 mm using a hole density of 2xlO 18cm-3
For such a hole density the Fermi temperature is 67 K and since the operation has to be
performed in the degenerate state, the working temperature has to be less than half of
this value.
One of the uncertain points in this calculation was the use of the constant C2 in Eq. (6)
though it was reasonable to keep the semi-empirical relation of the absorption constant
depending on the hole density as known for polar semiconductors. For estimating the
worse case of a three times higher constant C2 for AIN, a plot of this case of the
minimum electron beam current densities for laser action is shown in Fig. 2. We see that
the current densities are then higher and the minima depending on the cavity length are
shifted to shorter values than in Fig. I which indicates that both changes are due to
stronger absorption. In order to see dependance of laser threshold on the absorption the
evaluation in Fig. 3 is given.
493
Current Density For AIN Laser
Figure I. SOkeV electron beam current density threshold for laser action in p-AIN
depending on the hole density np and cavity length without any correction of reflectivity
and an assumed absorption similar to that of GaAs.
Current Density For AIN Laser
~
~.::..~....~
~....z...
106<X
B~
10400
z0
~
~
Figure 2. Same as Figure I for three times higher absorption
494
Current Density For AIN laser
1~inzwC...Zwcr~u
~z
~u~w
ABSORPTION em·'
'0
Figure 3. Threshold electron beam current density depending on the absorption for
various hole densities np.
Summarizing the threshold of 5 kNcm2 at d = 4 mm for a hole density of 1.5xlO18
cm-3 is not critical. A further result is that the laser threshold current density decreases
to 140 A cm-2 for a cavity length of I cm at a hole density of 2xlO 17 cm-3 which
corresponds to a Fermi temperature of 14 K. The necessary lower temperature can be
achieved by using liquid helium as a coolant having the advantage because of the thin
active layer and the high thermal conductivity of AIN.
3. High Internal Stress in Diamond for Direct Band Conductivity
For the laser emission of an electron beam excited semiconductor it is not only
necessary that it is p-conducting with a hole concentration above 1017 cm-3 as it was
concluded theoretically [12, 21] and as it can be realized in AIN [18], it is further a
necessary condition that the semiconductor is a direct band conductor as it is the case in
GaAs under normal conditions. This is not the case for germanium, silicon and diamond
under normal pressure, as pure and clean crystals without defects. The normal crystals
495
are typical indirect band semiconductors which is an anomalous property and whichwas necessary for the unnaturally long life time of minority carriers as the basis of thetransistor effect for the polar transistor of Shockley. Bardeen and Brattain [23].
The question how to change diamond into a direct band semiconductor was studiednumerically by putting the diamond crystal under high strain [24]. Fig. 4 shows theresult of these studies if a diamond crystal is under strain of 4%. It was confirmed thatthe diamond is changing from the normal indirect band semiconducting properties intothe direct band properties.
The question is then how can one put diamond under such stress. Any macroscopicstrain may be very difficult though one may imagine a CVD film on a steel band ofcertain thickness and subsequent bending of the band such that the diamond gets thestrain of the necessary 4%.
Working then with the input of electron beams for laser excitation of such a stresseddiamond film should permit the desired laser emission.
Instead of such linear strain mechanisms. however. one may take advantage fromthe knowledge of the very high internal stress known from silicon single crystals aftervery strong ion implantation - just before the crystals are braking and arriving at adilatation of up to 10% [25]. The analogy of the diamond to the silicon case may beexplained here in more details.
1.8
8
1.66
>: 1.4 , --- 4 :>!!:. ~>- >-~
2 ~II: II:w wz 1.2 zw w
0
1.0 -2<001 > < 100>
0.8 -4
L' r L'
Fig. 4 E-k-Band diagram of diamond at normal condition (straight line; multivalleybands) and at 4% strain along the <100> direction according to Pickett and Mehl [24]
496
While the compressibility of crystalline silicon has been studied carefully [26],leading to an experimental value of 1.022xl0-12 cm2/dyn, the low level ofinformations on the mechanical properties of amorphous silicon requires an extensionfrom the crystalline values on the basis of an appropriate theory. Though Madelung'spotential theory was developed for polar crystals only, the use for the heteropolarsilicon crystal was possible using higher-order perturbation [21] where theexperimental value was reproduced within 19% accuracy. The concept of theMadelung potential is to use a potential u
(8) e2 /3u = -0.2905-+-r r9
where e is the electron charge and r the distance of the atoms. The first term is thephysically determined electrostatic attraction modified by a factor taking into accountthe ions of various sign in the neighborhood of an atom in a polar crystal, while thesecond term expresses the expansion force with the ad-hoc assumption of an exponent.The value /3 = 0.2905 e2ro8.9 contains the equilibrium radius ro and therefore thedensity n of the solid. The resulting compressibility
(9)
depends then on the density n and fairly agrees with the experimental values. Thewhole physics of solids is then transferred to the derivation of n based on forces of thevalences, cohesion, van-der-Waals forces, etc., a problem which hitherto has not yetbeen elucidated quantum mechanically, The essential process for the compression isthe ad-hoe assumption of the high power nine of a potential in (8). The significancewas the factor of the mutual interaction of the different atoms of the neighborhood.This question can be investigated, e.g., by the theory of lattice sums and advancedconvergence models [28].
Contrary to this Madelung model of the compressibility based on ad-hoc assumptionof a power law with fitting the right exponents, a theory was elaborated [29; and p. 6 ofRef. 30] based on the physical facts of the quantum pressure of the centermostelectrons ia molecules, crystals or metals in which last case it emerges into the FermiDirac pressure.
497
In order to demonstrate that the quantum (Fermi) pressure really acts as a pressure
even if the temperature of the electrons is very much lower thaneF, we mention here as
an example of its action the compressibility of solids by the quantum pressure. For the
ideal equation of state, the compressibility (using V=l/n) is determined by the pressure P
(10)I oV
K=---.V oP
For solids there was the theory of Madelung (1918) which was modified slightly later
[31]. Contrary to this highly hypothetical relation, we can explain the compressibilities
of solids with a better agreement with the experimental values of the compressibility
over three orders of magnitude in dependence on the electron density, if we take the
quantum pressure of the valence electrons (with one or two electrons in the outermost
shell). Taking eF as the energy per volume V of the electrons in an atom (P=eFIV) and
calculating
(II)I OV 6moK-------
- V oP - h2nX
where the balance of the quantum energy with the electrostatic energy (see p.28 of Ref.
[32]) was included. Without this inclusion, the value OfK is half.
In Fig 5, the result for one and two valence electrons are drawn and compared with the
Madelung value. The Quantum pressure theory of compressibility [29; 30] reproduces
the experimental values to a good approximation. Experimental deviations from the
plots are due to the effective mass in the solids Eq. (II) where the vacuum electron only
was used; for an alternative derivation of the SchrOdinger equation for the effective
mass, see Appendix I of (Ref. [32]), and due to additional electrostatic energy densities
of attraction and repulsion by electron-electron, electron-ion or ion-ion interaction.
The result is that solids cannot be compressed easily because of the quantum (Fermi)
pressure which is nearly 10eV per atom (corresponding to about 100000K temperature)
but it acts as pressure only, not as temperature of the ideal equation of state. This follows
for metals as well as for insulators. It demonstrates that the Fermi pressure is still present
even in the unionized condensed material. The pressure of a substantial energy in the
quantum (Fermi) energy can also be seen for nuclear reactions at very low temperatures
498
of O.leV or less if the density is about 105 times the solid state or higher [33; 34] the
quantum energy of protons (with one particle per energy level) of mass mp
(12)
c
~ 10"11~E
.:::,,.,.,.E'5a.c
810·"
Fig. 5 Compressibility of solid materials compared with Madelung's theory [30] (dashed
line) and with the quantum pressure including the counteracting electrostatic energy
density (upper line); without the electrostatic energy, the lower line will follow where an
interpretation of spin coupling for two valence electrons is possible per quantum state.
is then 17 keV for a density of 1029 cm-3, just enough for nuclear fusion reactions with
a very low rate. Without following up the sophisticated theory for these reactions [34],
Eq. (12) indicates that the ions even at low temperature do have enough quantum energy
for performing the nuclear reactions.
499
What results for silicon after bombardment with the usual 1015 ions of around halfMeV energy is a state of up to 10% higher volume and a decrease of tablerecrystallization energy from 4.7 eV to 2.7 eY. It was estimated [25] that a stress of 1to 2 eV per atom, corresponding to a pressure of 1.4xl05 atm equivalent to 1.47xlOll
dynlcm2 is the result. How these intercrystalline ion implanted atoms are thenincorporated into the crystal at annealing is well known if one uses annealing stovesand several hours duration. The use of laser pulses led to voids, visible with theelectron microscope [35]. These voids reduced the thermal conductivity by a factor upto 30 [36].
Converting all these results to diamond and using the compressibility as derivedfrom the quantum pressure mode, Fig. 5, we find that a strain of 4% results in aninternal pressure of 3.076xlOll dynlcm2. This can easily be compared with the strainin diamond of 1 eV per atom resulting in 2.7xIOll dynlcm2 similar to the case ofsilicon. Therefore we can conclude that instead of stressing the diamond for thetransition into the direct band semiconductor state, we simply put the crystal under ionimplantation of a similar rate per volume as mentioned before for silicon. The crystalwill then increase its volume by several percents and will then without any outsidestress or strain have an internal strain such that the semiconductor band structurechanges from the indirect band structure into the direct band structure.
4. Electron Beam Pumped Diamond Laser
The possibility for the consideration of diamond as a laser was mentioned [37]where at least the results of luminescence at irradiation by electron beams (cathodoluminescence) [38; 39] was used as a basis. Since the reported production of diamondsingle crystals on amorphous surfaces was possible for the first time [18; 40; 4] usingthe chemical vapor deposition CVD [2; 4] it seems to be interesting to discuss underwhat conditions a diamond laser could be produced.
Diamond is an indirect band gap material. The possibility of producing directbandgap diamond under stress explained in the preceding Section 3.
Electron beam excited lasers are solid state lasers which may emit frequencies up tothe far ultraviolet range. There may be a limitation that the single crystals have to bemade of a polar semiconductor like GaAs or at least with a material having a strongpolar component such as CdS or ZnS. An efficiency of35% was measured for CdS[14;
500
16; 17]. Diamond has basically such a property similar to AIN which is a r--')spect for an
electron beam excited laser in the far UV, see section 2.
One condition for such a laser is that it is p-conducting with a rather high hole
concentration. It seems that this can be fulfilled experimentally [18; 41]. The more
critical difficulty for diamond is that sufficient strain is required to undergo a transition
to a direct bandgap [24]. Examined is a method suggested by Prawer [42] based on ion
implantation as we discussed in the preceeding section 3.
The stress energies per atom are in a reasonable range, e.g. 1 eV per atom corresponds
to a pressure in silicon which values are fully in the range to produce a strain of more
than 3% without damaging the crystal structure. As known from the theory and
numerical studies of the diamond band model [24], this should be just sufficient for the
conversion of the indirect band into the necessary direct band diamond structure.
Following the band energy in the stressed diamond crystal we base the following
calculation on this value of 4 eV for the band gap which corresponds to a wave length of
310 nm in the ultraviolet range of the laser emission. It cannot be easily predicted
whether the stress will shift this energy to lower or to higher values. It is remarkable at
least that in the case of silicon, the irreversible effects of electron beam irradiation
caused an increase of the fundamental band absorprion by up to 0.2 eV before the silicon
structure was destroyed by brittling [21].
We calculate now the threshold conditions for optimized dimensions of these electron
beam pumped lasers following the preceeding calculations in Section 2. There are,
however, several difficulties in the prediction for diamond since the effective masses of
the electrons and the holes are not known - which results in a rather minor error in the
following estimations - and since the complicated spontaneous absorption has is not
available for this wave length. For the last case we shall evaluate a wide range of
parameters and look at what conditions the minimum current density of electron beams
are required.
Because of the high dielectric constant the losses by diffraction will be low for the
thickness d of the laser cavity of up to several micrometer. We use the same assumptions
for electron-hole pair production as before [12] where the effective number of produced
pairs per incident SO keV electrons is just compensated by the assumption of unmodified
exit reflectivity of the cavity. We find then for the minimum electron beam current
density the threshold of Eq. (1) where C1 is given now for a wave length of 310 nm. The
difference to the wave length of GaAs alone increases the threshold current density by a
501
factor of 5.7. Since the effective masses are not know for diamond we use the vacuum
mass, m, of electrons as in Eq. (2) and we use the number of pairs produced per 50 keY
electron as given in Eq. (3). Equation (1) results then in
Here, La is defined as before [12; 21], see Eq. (5). The absorption coefficienta. is
written in the form of Eq. (6) which is the empirically derived expression for polar
semiconductors [22].
The constant C2 is not known for diamond but in a chemically similar material GaAs
is 1x10-21 (12; 21] which will be used in the following. Equation (13) becomes,
(14)
Using Eqs. (6) and (7) value R of 0.995, and assuming a value for d of 2 !lm, the
threshold current density for an e-beam pumped diamond laser can be found as a
function of I and np' This result is shown in figure 6.
The resulting threshold current densities even for optimzed cavity lengths and hole
densities are rather high. This is not unexpected since the current density goes with the
inverse square of the wavelength. Since electron beam currents of a few hundred Ncm2
were easily tolerated when using GaAs [15] without destruction or damage of the
crystals, we can assume that the stronger chemical bound of the diamond crystals may
easily permit a ten times higher electron beam current density. A similarity is given by
the fact that the laser damage threshold for diamond is about ten times higher than for
materials with metallic or heteropolar bounds.
One example for a laser condition with the 50 keY electrons would be a current
density j=0.39 kNcm2 for a cavity length of 5 mm using a hole density or' 2x 1018 cm-3
For such a hole density the Fermi temperature is 67 K and since the operation has to be
performed in the degenerate state, the working temperature has to be less than half of
this value.
One of the uncertain points in this calculation was the use of the constant C2 in Eq. (6)
though it was reasonable to keep the semi-empirical relation of the absorption constant
depending on the hole density as known for polar semiconductors. For estimating the
502
worse case of a three times higher constant C2 for diamond, a plot of this case of the
minimum electron beam current densities for laser action is shown in Fig. 7. We see that
the current densities are then higher and the minima depending on the cavity length are
shifted to shorter values than in Fig. (6) which indicates that both changes are due to
stronger absorption. Nevertheless, the threshold of 0.7 kA/cm2 at d=4 /-lm for a hole
density of I.5xlO l8 cm-3 is not critical. A further result is that the laser threshold
current density decreases to 34 Acm-2 for a cavity length of I cm at a hole density of 2x
1017cm-3 which corresponds to a Fermi temperature of 14 K. The lower temperature
can be achieved by using liquid helium as a coolant because of the thin active layer and
the high thermal conductivity of diamond.
We conclude that the presented calculations indicate the possibiltity of electron beam
pumped diamond lasers for the ultra violet spectral range.
Current Density For Diamond Laser
N
G3:~...;:;;~Q...~103'"'"i3
~a:l
Z
~
Fig. 6 50 keY electron beam current density threshold for laser action in p-diamond
depending on the hole density np and cavity length and an absorption similar to that of
GaAs.
503
Current Density For Diamond Laser
~
3~>
V>
~Q
~
~'" 104'"13
~%
~
~
Fig. 7 Same as Figure! for three times higher absorption.
5. Compact Nuclear Battery with Wide Bandgap Photovoltaics
After having discussed electron beam excited wide bandgap lasers, the use ofphotovoltaic cells of similar materials is considered. When light in the UV (e.g.Krypton) with emission at about 250 nm wavelength is photoelectrically converted bywide bandgap p-n junction cells, energy efficiencies of 50% and much more can beexpected.
The conversion of nuclear energy into electricity has a number of well knownsolutions apart form the globally used nuclear reactors. There are thermoelectric, orphotoelectric arrangements in use or such using thermionic emission. It is known alsothat electricity can be generated by direct conversion of the electron energy of betaradiation by movement of the electrons radially from a central beta-emitter against apositively biased surrounding hollow sphere (Moseley battery) which is well knownand practically free of radioactivity if strontium-90 is used as a beta source. Thecurrents, however are very small and losses are given alone by the fact that the betaspectrum is non-monochromatic.
504
Compared with all known nuclear batteries the following described battery is more
compact, has higher efficiencies, is completely free from gamma radiation, and has a
minimum risk with regard to the rare gas isotope. The used ultraviolet radiation and any
low energetic bremsstrahlung from the betas is harmless because of the hermetic sealing
of the nuclear reaction vessel.
All these advantages are being achieved if according to the following description a
photovoltaic conversion of the radiation of the betas in the rare gas crypton - or with
some additional gases of appropriate optical excitation levels - is done in broadband
semiconductors as diamond, aluminium nitride and analogous materials having a high
opto-electronic efficiency.
Fig. 8 describes an example of the battery. A hermetically closed metallic or quartz
container preferably welded by electron beams, contains a diamond or analogous layer
which has been prepared as a photovoltaic cell either with a Schottky barrier or with a
specially wide distance changing p-n junction. As an electrical contact a metal is used
which is so thin that it is highly transparent to ultraviolet light. This metal layer is
connected to a contact which electrically isolated is going to the outside of the container.
Further is there a contact with the container such that between both contacts electrical
power P from the operation of the photovoltaic cell can be used.
In the inner part of the vessel is a gaseous Krypton-85 isotope, as it is available from
radioactive waste for nuclear reactors in huge amounts for relative low cost. This gas can
- but not necessarily - contain further mixtures of other gases for improvement of the
spectral range of the beta ray induced emissions of ultraviolet light. The Krypton-85 is a
beta ray emitter with 10.76 years half life, 290 keY decay energy and a gamma emission
of less than the 10-10 ths part of the decay energy, i. e. it is as good as free from nuclear
radiation because all beta energy is going completely into optical, electrical and thermal
energy.
The high nuclear-electric efficiency is given by the fact that all emitted beta particles
even at very low energies are - apart from the walls - converted into a narrow spectral
line UV radiation (more than 50% of them are in the spectral range of 145 nm± 5 nm)
for which the diamond photocell has an efficiency of 60% and more for conversion into
electric energy.
505
high pressure container (steel or quartz glass)
for electric power
generation
cooling system
quartz
or steel
luminescent layer
electric contacts converting UV-lightfor Pholocel1(e.g. Sn02) lo visible
or 100m gold layers
Thermionic or thermoelectricelement for generating
electric power
Fig 8 Nuclear Battery with wide bandgap photovoltaic cells
Since the remaning energy of about 50% will be released as thermal energy, either a
global cooling system is neccessary with heat exchangers using e. g. water at the outer
part of the container or sufficient thermal radiation emission is needed. The temperature
of the apparatus can because of the use of the high temperature operation of diamond
photocells without reduction in the electrical efficiency, be operated at high
temperatures, e. g. at several hundred degrees centigrade. There can cooling ripples be
applied also which cause an emission of radiation to the outside and which are at a
definitely lower temperature than the container. Between the container and an outer
mantle can also be a thermoelectric generator which additionally contributes to the
generation of electricity.
The high efficiency, compactness and radioactive harmlessness of the nuclear battery
can be seen from the fact, that an apparatus of the kind of Fig. 8 but with a spherical
mantle of the container of 30 cm diameter and a total weight of less than 200 kg can
produce an electric power of 20 kW. For comparison with other nuclear batteries we can
506
mention the following cases; 20 kW batteries with Sr-90 have a diameter of 175 cm and
a weight of 1750kg, that with Plutonium-238 a diameter of 250 cm and a weight of 6000
kg. Only the Polonium-210 generator with 40 cm diameter and 350 kg system weight is
close to the invention of the Kr-85-diamond battery but there is its radioactive decay
time a disadvantage, as it reduces its power after 1.2 months to 80 %, and after 12
months it is down to 18 % while the Krypton-95 diamond battery decays to 80% only
after 3.5 years.
For operation at very high powers, e. g. at 50 kW or higher, the battery cooling can be
designed to be used as optical black body radiator. The surface of the metal container (of
sufficient high melting point, using e. g. titanium, tantalum or even tungsten) may be at a
temperature of 1000 degrees centigrade. While producing electrical energy, the surface
radiates then mostly visible radiation and can be used for illumination. As an example, if
the container is a sphere with 40 cm diameter and if the temperature is 727 degrees
centigrade (1000 degrees Kelvin), the emission of black body radiation is 28 kW.
One can produce batteries with another type of optical radiation which is of interest
e.g. for the operation in the few kW range. If the container is not of a metal but of an
optically transparent material, e. g. quartz glass with a then necessary additional
electricity production, one can add a luminescent material between this last mentioned
contact ahd the inner part of the quartz container, see Fig. 8. The radiation which is not
converted into electricity in the photovoltaic cell is then converted used to be converted
into visible radiation in the luminescent material and radiates then like an incandescent
lamp. Permanentely for many years working light sources in the range of kW
simultaneously producing electricity and heat can be generated. For the heat exchange
one may use a quartz or hard glass container with half spherical and half parabolical or
rotationally elliptical shape. Then the latter surface there may be of a highly reflecting
coating such that the optical emission is then highly directed to the other side, e. g. for
illumination of football fields without needing any electrical installation of power
support. The coated layer is connected to a heat exchange e. g. a liquid one, for
generating heat and cooling the battery.
507
6. Acknowledgement
Assistance by Dip!. Ing. A. Schonberger is gratefully acknowledged. The work was
supported in part by a Fullbright Senior Scholarship and the Deutsche Akademische
Austauschdienst (DAAD) in Bonn.
* On leave from University of New South Wales, Sydney 2053, Australia,** University of Missouri, Columbia, MO, USA
[1] F. P Bundy, H. T. Hall, H. M. Strong, R. H. Wentorf, Nature 176, 81 (1955)
[2] M. W. Geis and 1 C Angus, Scientific American, October 1992 p.64
[3] H. Hora, Chemische Rundschau (Basel) H, 393 (1961)[4] H. Hora, Japanese Patent 472771 ( 16 July 1965); Brit. Patent 1,0011,308
(11Aug. 1965);
[5] W. G. Eversole, U. S. Patents 3,030,187 and 3,030,188 (1962), see A. D. Kiffer,
Synthesis of Diamond from Carbon Monoxide, Report from the Tonawanda
Laboratories, Linde Air Products Co, Tonawanda NY, 6 June 1956.
[6] B. V. Spitsyn, 4th USSR Meeting Crystal Growth, Erevan University Pub!.
House, Erevan 1972 (pp. 676 to 679)
[7] 1 C. Angus and C C Hayman, Science 241913 (1988)
[8] M. W. Geiss, Proc. IEEE, 79, 669 (1991)
[9] G. Popovici and M. A. Prelas; Phys. Stat. Solidi, A132, 233 (1992)
[10] A. M. Prokhorov, private communication 1983
[11] N. G. Basov, and O. N. Krokhin, Soviet Phys. JETP, li, 1508 (1964)
[12] H. Hora, Public Lecture, Siemens Munich April 1964; Zschr. Ang. Math. Phys.
(ZAMP), lQ, 95 (1965); H. Hora, Phys. Stat. Solidi,~, 197 (1965)
[13] CR. Hurwitz and R. 1 Keyes, App!. Phys. Lett., 2, 139 (1964)[14] C R. Hurwitz, Appl. Phys. Lett.,~, 243 (1966)
[15] N. Holonyak, et. aI., 1 App!.Phys., 44,5517 (1973)
[16] C. R. Hurwitz, Appl. Phys., Lett.,:l, 420 (1966)
[17] C. R. Hurwitz, App!. Phys. Lett., 2, 116 (1966)[18] M. A. Prelas, E. 1 Charlson, E. M. Charlson, and 1 Meese, Contract Report,
"A Study of Potential High Band-Gap Photovoltaic Materials for a Two Step
Photon Intermediate Technique in Fission Energy Conversion," Advanced
508
Energy Projects, Department of Energy, (Aug. 1992)
[19] J. Meese, Personal Communication, Univ. of Missouri, Columbia, MO65211, (Aug. 1992)
[20] H. Hora, Naturwissensch., 48, 641 (1961)
[21] H. Hora, Zeitschrift f., Naturforschung, 20A, 543 (1962)
[22] I. Kudman, and T. Seidel, J. Appl. Phys., 11. 771(1962)[23] Frank: Herrmann, Proc. IRE 43,1703 (1955); Rev. Mod. Phys. 30,102 (1958)
[24] W.E. Pickett, and M.J. Mehl, SPIE Vol. 877, Micro-Optoelectronic Materials(1988), p. 64
[25] H. Hora, Appl. Phys. A32, 217 (1983)
[26] H.B. Huntington, Solid State Physics Vol. 7. p. 213 (Academic Press, NewYork,1958)
[27] T. Soma, Phys. Stat. Solidi B76, 753 (1976)
[28] S. Mathias, Phys. Stat. Solidi B74, 69 (1976)
[29] H. Hora, and R. Romatka, Naturwissenschaften 69,399 (1982)
[30] S. Eliezer, A.K. Ghatak, H. Hora and E. Teller, Equations of State (Cambridge
University Press 1986)
[31] G. Joos, Theoretische Physik, (Akad. Verlagses.Wiesbaden, 1976) p. 537
[32] H. Hora, Plasmas at High Temperature and Density (Springer, Heidelberg,1991)
[33] E. Teller, in Laser Interaction and Related Plasma Phenomena, AlP Proceedingsof the 11th conference, Monterey Oct 1993, G.H. Miley ed. (Am. Inst.Phys. New York 1994)
[34] S. G. Brush, Progr. High Temp. Phys. Chern., c.A. Rouse Ed., Vol 1, p. 1(1967)
[35] E. F. Krimmel, H. Oppolzer, H. Runge, and W. Wondrak, Phys. Stat. Solidi A66,565 (1981)
[36] H.J. Goldsmid, H. Hora, and G.L. Paul, Phys. Stat. Solidi A81, K127 (1984)
[37] A. Feldman, and L.F, Robins, Applications of Diamond Films, Y. Tseng, M.Yoshikawa, M. Murakawa, and A. Feldman, eds. (Elsevier SciencesPublishers, BV, 1991) p. 181
[38] O. Davies, Chemistry and Physics of Carbon, P.L. Walker and P.A. Throwereds (Marcel Dekker, 1977) Vol. 13 p. 1
[39] J,Walker, Rep. Progr. Phys., 42, 1605 (1979)
[40] G. Zhao, T. Stacy, E. 1. Charlson, E. M. Charlson, C. H. Chao, M. Hajsaid, J.
Meese, G. Popovici, and M.A. Prelas Appl. Phys. Letters §l, 1119 (1992)
[41] X. Jiang, and c.P. Klages, Fraunhofer Inst. Schicht und Oberflacherntech.,
Hamburg, International Diamond Conf. Heidelberg, Aug. 1992
[42] S. Prawer, Personal Communication, School of Physics, Melbourne University
(Nov. 1992)
509
ORAL PRESENTATIONS
Diamond Session I: C. Wallace, A. Zaitsev.
Chemistry ofWide Bandgap Electronic Materials Growth, W. Lambrecht, Case Western
University, USA.
Chemical Problems of Diamond Doping, B.y. Spitsyn, Institute of Physical Chemistry,
Russia.
Structural Control and Properties of CVD Diamond, C. Wild, Fraunhofer-Institute fur
Angewandte Festkorperphysik, Germany.
Sub-band Structure of Diamond, N. Samsonenko, Institute of Constructions Engineer
ing ofDonbass, Ukraine.
Problems of Diffusion Doping of Wide Bandgap Electronic Materials, G. Popovici,Rockford Diamond Technology, USA.
Negative Electron Affinity Properties of Diamond, B. Pate, Washington State Univer
sity, USA.
Diamond Session II: B. Pate, A. Gontar.
Doping of Diamond-like Films, S. Mitum, Technical Universityof Lodz, Poland.
Device Properties of Diamond Crystals, A. Gontar, Institute for Superhard Materials,
Ukrainian Academy ofSciences, Ukraine.
Diamond Based Electronics, 1. Davidson, Vanderbilt University, USA.
Diamond Photovoltaics, P. I. Perov, Institute ofRadioengineering and Electronics, Rus
sian Academy ofSciences, Russia.
Calculation ofPhosphorous Electronic Levels in Diamond, V. Tokiy, Department of
Physics, Donbass Institute ofConstruction Engineering, Ukraine.
511
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials. 511© 1995 Kluwer Academic Publishers.
512
AINIBN Session: P. Gielisse, P. Mazurenko.
Deposition and Structure of Aluminum Nitride Films, A.F. Belyanin, Central Techology
Institute. Russia.
ECR Growth ofAIN and Cubic Boron Nitride. G. Auner. Wayne State University, USA.
CVD ofAlN, V. Sokolov, Institute ofElectron Techniques, Russia.
AIN Electronic Devices, T. Stacy. University ofMissouri-Columbia. USA.
Macro and Micro Structural Factors Effecting Thin Film Growth of Ill-V Compounds,
P. Gielisse, Florida A&M/Florida State University, USA.
Wide Band Gap Electronic Devices, V. Chelnokov, Ioffe Physical Technical Institute.
Russia.
Boron Nitride Electronics, A. Zaitsev, Belarussian State University, Belarus.
Doping of BN, V. Shipilo, Institute ofSolid and Semiconductor Physics, Belarus.
Theoretical Issues and Applications Session: C. WI1d, V. Koleshko.
Theoretical Issues of Wide Band-Gap Electronics, H. Hora, University of Regensberg.
Germany.
Photovoltaic Applications of Wide Band Gap Electronics. M.A. Prelas, University of
Missouri-Columbia, USA.
Advanced Applications of Diamond Electronics, C. Wallace, BDM International, USA.
Present Notations of BN Phase Diagram, V. L. Solozhenko,
Electrical and Optical Impurity Centers in Diamond, A. Gippius, Lebedev Physical
Institute, Russia.
Panel Discussion:
Problems in the Development ofWide Band Gap Materials for Electronics, W. Lambre
cht, B. Spitsyn. P. Gielisse and A. Zaitsev.
Doping ofWide Band-Gap Materials, V. Shipilo, G. Popovici, A. Gippius and C. WI1d.
Scientic Challenges and Technical Barriers for Wide Band-Gap Electronic Materials, H.
Hora, E. Tochitsky, T. Stacy and V. Vavilov.
POSTER PRESENTATIONS
Research of Dehydrogenation of Thin BN~:H Layers, Z.L. Akkerman, M.L. Kosinova,N.L Fainer, Yu. M. Rumyantsev, Inst. Inor. Chem., Russian AS, Novosibirsk, Russia.
AES-SIMS Analytical System for Composition Measurements of Wide Band GapSemiconductors, A.I. Babanin, Ioffe Inst. and CREE EED, St-Peterburg, Russia, andA.A. Lavrent'ev, SPbETU, St-Peterburg, Russia.
AlN Thin Film Deposition by Reactive Ion Beam Sputtering, A.F. Belyanin, Centr.
Research Technological Institute (Moscow), A.P. Semenov, and V.M. Khailanova,
Buryat Inst. Nat. Sci. Siber. Div., Russian AS (Ulan-Ude).
Electrical Conductivity of Ceramics Based on Different Boron Nitride Modifications,
A.V. Bochko, Inst. Mater. Sci. Prabl., Ukraine AS, Kiev, and G.A. Sokolina, Inst. Phys.
Chem., Russian AS, Moscow.
Structure and Thinning in CVD Epitaxial Layers of Aluminum Nitride on Sapphire,
N.V. Borovsky, A.V. Dobrynin, G.A. Naida, E.B. Sokolov,Moscow State Technical University, Russia.
The Structure of Cubic Boron Nitride Polycrystals, V.F Britun, A.V. Kurdyumov, I.A.Petrusha. Inst. Prabl. Mater. Sci., Inst. Superhard Mater., Ukrainian AS.
Hydrogen Chemistry of Diamond Surfaces, J.E. Butler, J.N. Russell, Jr., P.E. Pehrsson,
B.D. Thoms, Gas/Surface Dynamics Section Nav.Res Lab. Washington, DC, USA.
Structure and Property Degradation in Irradiated Boron Nitride, V.S. Dedkov, A.V. Kab
yabev, V.V. Lopatina, Yu.P. Surov,High Voltage Inst., Tomsk Polytechn. University, Rus
sia.
The Growth of Pvrolitic Boron Nitride by Pyrolysis of Borazine, V.N. Demin, A.I.Basov, Z.L. Akkerman, Inst. Inorg. Chem. RAS, Novosibirsk.
Optical Properties of Sputtering and Glow Discharge a-C:H Films, A. Dragornir, M.
Gartner*, C. Morosanu, G. Pavelescu, T. Stoica, Inst. Phys. Chem. Romanian Acad.Bucharest.
513
M.A. Prelas et at. (eds.), Wide Band Gap Electronic Materials,© 1995 Kluwer Academic Publishers.
514
Diamond Particles on Silicon Tips: Preparation. Structure and Field Emission Properties, E.I. Givargizov, L.v. Aksenova, E.V.Rakova, AN.Stepanova, P.S. Plekhanov, V.V.Zhirnov, Inst. Crystallog. Russian AS, Moscow.
Thermodvnamic Analysis of the Chemical Vapour Deposition of Boron Nitride From
Borazine, AN. Golubenko, V.N. Demin, M.L. Kosinova, Inst. Inorg. Chem. RussianAS,
Novosibirsk.
Electronic Structure of Perfect and Defect Hexagonal and Rombohedral Boron Nitride,
S.N. Grinyaev, V.V. Lopatin, High Voltage Inst. Tomsk Polytechn. University, Russia.
Diamond and Aluminum Nitride Laser with Electron Beam Excitation, H. Hora, M.A.Prelas*, University of NSW Kensington, Australia, *Fulbright Professor University of
Missouri, Columbia, MO, USA.
Energetic Levels of Defects in Graphitelike and Rhombohedral Boron Nitride, F.V.Konusov, V.V. Lopatin, High Voltage Inst. Tomsk Polytech. University, Russia.
Structure Characteristics and Formation Mechanisms of Diamond and Diamond-like
Boron Nitride Modifications at High Pressure, AV. Kurdyumov, v.F. Britun, N.F.Ostrovskaya, Inst. Probl.Mater.Sci. Ukrainian AS.
XSP Plasmon Loss Spectroscopy of CVD Diamond, R. McEwen, British Aerospace
Ltd, Filton, Bristol, England.
Cathodoluminescent Investigation of External Factors Influence on Defective CubicBoron Nitride Structure, V.B. Shipilo, T.W. Rapinchuk, N.A. Shishonok, Institut of
PhysicsofSolids and Semicond, Minsk.
Reactive Ion Etching of Silicon Carbide with Fluorine Containing Plasmas, V.E. Sizov,
K.v. Vassilevski, Cree Res. EED, St. Petersburg, Russia.
Present Notion of BN Phase Diagram, V.I. SolozhenkoJnst. Superhard Mater. Ukrainin
AS, Kiev.
Kinetics Threshold for Growth of Cubic Boron Nitride, C. Taylor, S. Kidner, R. Clarke,Univ. ofMichigan, Randall Laboratory ofPhysics, Ann Arbor; MI, USA.
Dislocation for Photovoltaic Energy Conversion, V.1. Timchenko, v.v. Tokly, N.D.
Samsonnenko, Res. Group Diamond Coatings, Donbass Inst. Constr. Eng., Makeyevka,
Ukraine; L.L. Bullov, G.A Sokolina, Inst. Phys. Chem Russian AS, Moscow.
The Influence of High Pressures and Temperatures on Properties of Aluminum Nitride,
V.S. Urbanovich, Inst. Solid State and Semic. Phys. Belarus AS, Minsk.
515
1.54-mm Photoluminescence from Er-implanted GaN, AlN, and InAlN, ecl., R.G. Wil
son, R.N. Schwartz, Hughes Res. Labs.• Malibu. CA. USA, C.A. Abernathy, SJ.Pearton, AT&T Bell Labs.• Murray Hill. NJ. USA, N.Newman, Lawrence Berkeley Lab..
Berkeley, CA, USA, J.M. Zavada, Army Res. Office. Res. Triang. Park. NC. USA.
Diamond Growth by Hot Carbon Filament Chemical Vapor Deposition, C.c. Chao, EJ.Charlson, E.M. Charlson, J. Meese, M.A. Prelas, and T. Stacy, University of Missouri
Columbia, USA.
To the Question of the Diamond Nuclei's Formation from the Gas Phase, A.P. Rudenko,
and 1.1. Kulakova, Moscow State University. Moscow. Russia.
Prediction of Diamond Film Thermal Conductivity, N.V. Novikov, T.D. Ositinskaya,
A.P. Podoba, and S.V. Shmegerea, Ukrainian Academy ofSciences. Kiev. Ukraine.
Spectral Hole-Burning Study of the Defects Created by Neutron Irradiation in a Natural
Diamond, I. Sildos, G. Zavt, and A. Osvet, Estonian Academy of Sciences. T. Artu. Es
tonia.
Surface and Bulk Conductivity of Hydrogen Treated Polycrystalline Diamond, G.A
Sokolina, L.L. Bouilov, AA Botev, AV. Markin, Russian Academy ofSciences. Mos
cow. Russia, and M.A. Timofeev, Moscow State University. Moscow. Russia.
Positron Annihilation in Diamond Films, I.I. Bardyshev, L.L. Bouilov, and B.V. Spitsyn,
Russian Academy ofSciences. Moscow. Russia.
ESR Study of Paramagnetic Defects in Free Standing Diamond Films, T. A Karpukhina,
Russian Academy of Sciences. Moscow, Russia,M. A Prelas, G. Popovici, S. Khasawi
nah, University ofMissouri-Columbia. USA, and B. V. Spitsyn,Russian Academy ofSciences. Moscow. Russia.
Efficient Reduction of Nitride and Nitrate to Ammonia Using B-doped Diamond Elec
trodes, C. Reuben, E. Galun, R.Tenne, Weitzman Institute. Rehovot. Isreal, R. Kalish,
Technion. Haifa. Isreal, Y. Muraki, K. Hashimoto, A. Fujishima, University of Tokyo.
Tokyo. Japan, J.M. Butler, Naval Research Laboratories. Washington DC, USA, and C.
Levy-Clement, CNRS Bellevue, Meudon, France.
Diamond MIS Capacitors With Silicon Dioxide Dielectric, MJ. Marchywka, D. Moses,and P.E. Pehrsson, Naval Research Laboratories.Washington DC. USA.
Laser Modes in Diamond, L.-T. S. Lin, M.A. Prelas, University of Missouri-Columbia.USA, and G. Popovici, Rockford Diamond Technology, Inc., Columbia, Missouri, USA.
51S
Laser-assisted Chemical Etching of Diamond Films in Oxygen, V. G. Ralchenko, K. G.Horotushenko, A. A. Smolin and E. D. Obraztsova,Russian Academy ofSciences, Mos
cow, Russia.
Ion Milling of Polycrvstalline Diamond Films, A. E. Alexenko, RAS; A. F. Belyanin,Central Research Technology Institute, Moscow, Russia, L. L. Bouilov, Russian Acade
my ofSciences, Moscow. Russia, A. P. Semenov, Siberian Branch, Russian Academy of
Sciences. Kranoyarsk, Russia, and B. V. Spitsyn, Russian Academy of Sciences, Mos
cow, Russia.
Unhydrogenated DLC Films Obtained by Magnetron Sputtering, C. Morosanu, Institute
Physics and Technology ofMaterials. Bucharest, Romania, N. Tomozeiu, University of
Bucharest, Romania, C. Cordos, and T. Stoica, Institute Physics and Technology ofMa
terials. Bucharest. Romania.
Simulation ofDiffusion in an Amorphous Structure, A. V. Nazarov,I. Bardin Metallurgy
Institute, Moscow. Russia.
Optical and Electrical Properities of Quantum-dimentional Multilayer Structures Basedon Carbon Films, V.V. Sleptov, V.M. Elinson, A.M. Baranov, and S.A. Tereshin, NPO
"Vacuummasphribor". Moscow, Russia.
Thermal Stability and Structural Reactions at the Tantalum/a-C Interface Under Vacuum
Annealing Conditions, A.P. Novikov, E.A. Shilova, L.D. Buiko, and V.A. Zaikov,Rrsearch Centre ofElectronic Materials and Technology. Minsk, Belarus.
Extended and LocalizedElectronic States in Tetrahedral Carbon Films, V.E. Maschenko,
Institute ofSteel and Alloys, Moscow. Russia, V.M. Puzikov, Ukrainian Academy ofSciences. Kiev. Ukraine, and A.V. Semenov, Siberian Branch, Russian Academy ofScienc
es. Kranoyarsk. Russia.
Application of Amorphous Hydrogenated Carbon Coating to Semiconductor RadiationDetectors, I.M. Kotina, T.A. Antonova, G.V. Patsekina, V.D. Saveliev, L.M. Tuhkonen,PNPI. Leningrad district, Russia, 0.1. Konkov, and E.I. Terukov, Ioffe Physcal Techni
cal Institute. St. Petersburg, Russia.
Peculiarities of Chemical Vapor Heteroepitaxy of Wide Band Gap III-V Nitrides, E.B.Sokolov, G.A. Naida, and N.V. Barovskii, Moscow State Institute of Electronic Engi
neering. Moscow, Russia.
Ion Implantation intoWide Bandgap Semiconductors, V.S. Vavilov, P.N. Lebedev Phys
ics Institute. Moscow. Russia.
517
Thermodynamic Properties of Boron Nitride, V.L. Solozhenko, Ukrainian Academy of
Sciences. Kiev. Ukraine, and K.S. Gavrichev, Russian Academy of Sciences. Moscow.
Russia.
Aluminum Nitride Ceramic with High Thermal Conductivity, V.B. Shipilo, T.V. Rapin
chuk, and N.A. Shishonok, Academy ofSciences ofBelarus, Minsk, Belarus.
Reactive Ion Etching of Silicon Carbide with Fluorine Containing Plasmas, V.E. Sizov,
and K.V. Vassilevski, CREE EED. St. Ptetersburg. Russia.
Positron Annihilation in Sintered Boron Nitride, 1.1. Bardyshev, and A.D. Buravov,Russian Academy ofSciences. Moscow, Russia.
LIST OF PARTICIPANTS
(a) Directors
M. A. Prelas
B. V. Spitsyn
Nuclear Engineering Dept., University ofMissouri-Columbia,E2435 Engr. Bldg. East, Columbia, MO 65211, USA
Diamond Film Crystallization Laboratory,Institute of Physical Chemistry,31 Leninsky Prospect, Moscow 117915, Russia
(b) Key Speakers (listed alphabetically by the participants' countries and then names)
A. Zaitsev
H. HoraC. Wl1d
S. Mitura
A. F. Belyanin
V. Chelnokov
A. Gippius
E. SokolovV. Vavilov
N. Samsonenko
HELL&FD Laboratory,Belarussian State University, Minsk, Belarus
University of Regensberg, GermanyFraunhofer-Institute fur Angewandte Festkorperphysik,Thllastabe 72, D-7800, Freiburg, Germany
Institute ofMaterials Science, Technical University of Lodz,Stefanowskiego Str. 1, Lodz 90924, Poland
Central Techology Inst.,121355 Iv. Franko Str. 4, Moscow, Russialoffe Physical Techical Institute,21 Street Polytechnicheskya, RussiaP.N. Lebedev Institute ofPhysics, Russian Academy of Sciences,Leninsky Pr. 53, Moscow 117924, RussiaInstitute ofElectron Techniques, RussiaPN Lebedev Institute of Physics, Russian Academy of Sciences,Leninsky Pr. 53, Moscow 117924, Russia
Institute ofConstructions Engineering ofDonbass,339023, Makeyevka, Ukraine
519
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials,© 1995 Kluwer Academic Publishers.
520
P. Gielisse
W. Lambrecht
B. pate
G. Popovici
T. Stacy
College ofEngineering,Florida A&M University, Florida State University,2525 Pottsdamer St., Tallahassee, FL 32310, USADept. ofPhysics, Case Western Reserve University,Cleveland, OH 44106-4671, USADept. ofPhysics, Washington State University,Pullman, WA 99164-2814, USARockford Diamond Technology,P. O. Box 30675, Columbia, MO 65205;also Nuclear Engineering Program, University ofMissouri,Columbia, MO 65211, USAElectrical Engineering Dept., University ofMissouri,Columbia, MO 65211, USA
(c) Other Participants (listed alphabetically by the participants' countries and then names)
N. BerezhnovL. Gamezak
A. GarejeraV. GusakovV.LomovoyA. MelnikovA. NovikovO. SelifanowV. Shipilo
E. Shishonok
N. Shishonok
T. ShperovaA. StanishevskyDr. StelmachE. Tochitsky
V. Uglov
V. Urbanovich
V. Varichenko
R. McEwen
Belarussian State University, Minsk, BelarusBelarussian State UniversityFrolikova Str., 5-71, Minsk 220037, BelarusBelarussian State University, Fr. Skarina Pr. 4, Minsk, BelarusBelarussian State University, Fr. Skarina Pr. 4, Minsk, BelarusBelarus Academy of Science, Minsk, BelarusBelarussian State University, Fr. Skarina Pr. 4, Minsk, BelarusBelarussian State University, Minsk, BelarusInstitute ofElectronics, Minsk, BelarusInstitute of Solid and Semiconductor Physics,P. Brovka Str. 17, Minsk 220072, BelarusInstitute of Solid and Semiconductor Physics,P. Brovka Str. 17, Minsk 220072, BelarusInstitute of Solid and Semiconductor Physics,P. Brovka Str. 17, Minsk 220072, BelarusBelarus Academy of Science, Minsk, BelarusBelarus Academy of Science, Minsk, BelarusBelarussian State University, Minsk, BelarusBelarussian State University,Skarina Prospect 4, Minsk 220100, BelarusBelarussian State University,F. Skariny Pr. 4, Minsk 220090, BelarusTechnical University of Belarus,SkarinaPr. 131-2, Apt. 87, Minsk 220114, BelarusBelarussian State University, Minsk, Belarus
British Aerospace, Sowerby Research CentreP.O.Box 5, FPC, Filton, Bristol, BS12 7QW, England
W. Kulisch
C. Reuben
M. Kamo
Y. Sato
J. Szmidt
M. Gartner
O. Morosanu
T. Stoica
A. Alexenko
A. Arkhidov
A. Babanin
P. Belobrov
N. BuilovaI. Galushko
V. Klujev
I. KotinaA. Nazarov
V. NepshaP. Perov
V. Ralchenko
Institute ofTechnical Physics, University of Kassel,Henrich Plett Str. 40, Kassel 34109, Germany
Department ofMaterials and Inter Face Weitzman Institute,Levin Building, Rehovot 76100, Israel
National Institute for Research in Inorganic Materials,1-1 Namiki, Tsukuba, Ibaraki 305, JapanNational Institute for Research in Inorganic Materials,1-1 Namiki, Tsukuba, Ibaraki 305, Japan
Institute ofMicroelectronics & Optoelectronics,Koszykowa Str. 75, Warsaw 00-662, Poland
Institute ofPhysical Chemistry,Spl. Independentei 202, Bucharest, RomaniaInstitute Physics and Technology ofMaterials,Magurele, P.O.Box Mg 7, Bucharest, RomaniaInstitute Physics and Technology ofMaterials,Magurele, P.O.Box Mg 7, Bucharest, Romania
Institute ofPhysical Chemistry,Leninsky Pro 31, Moscow 117915, RussiaInstitute ofPhysical Chemistry,Leninsky Pr. 31, Moscow 117915, RussiaCREEEED,Polytechnicheskaya Str. 26, St. Petersburg 194021, RussiaInst. Biophys. Siberian Branch RAS,Krasnoyarsk 660036, RussiaVINm, Usievicha 20 A, Moscow 125219, RussiaInstitute ofPhysical Chemistry,Leninsky Pr. 31, Moscow 117915, RussiaInstitute ofPhysical Chemistry,Leninsky Pr. 31, Moscow 117915, RussiaNuclear Physics Institute, St. Petersburg, RussiaI. Bardin Met. Inst., Deptartment ofMetal Physics,Baumanskaya 2, Moscow 107005, RussiaDiamond Institue Vaiialmaz, Moscow, RussiaInstitute ofRadioengineering and Electronics,Russian Academy of Sciences,Frazino, Vvedensky Sqr. 1, Moscow 141120, RussiaGeneral Physics Institute,Moscow, Russia
521
522
A. Semenov
V. Sizov
G. Sokolina
A. Soroko
A. Stepanova
N. SuetinV. Vamin
K. Vassilevski
A. Vasutina
S. Voronina
A. Bochko
A. Gontar
V. Malogolovets
A. Podoba
N. Samsonenko
V. Timchenko
V. Tokiy
G. Auner
Institute of Natural Sciences, Siberian Division,Russian Academy of Sciences,Sakhyanova Str. 6, Ulan-Ude 670042, RussiaCREEEED,Polytechnicheskaya Str. 26, St. Petersburg 194021, Russia
Institute ofPhysical Chemistry,Leninsky Pr. 31, Moscow 117915, RussiaInstitute for Radioelectronics and Electronics,Russian Academy of Sciences, Mohovaya, RussiaInstitute of Crystallography (RAS),Leninsky Pr. 59, Moscow 117333, RussiaMoscow State University, Moscow, RussiaInstitute of Physical Chemistry,Leninsky Pr. 31, Moscow 117915, RussiaCREEEED,Polytechnicheskaya Str. 26, St. Petersburg 194021, RussiaInstitute ofPhysical Chemistry,Leninsky Pr. 31, Moscow 117915, RussiaInstitute ofPhysical Chemistry,Leninsky Pr. 31, Moscow 117915, Russia
Institute for Materials Sciences,Ukraine Academy of Sciences,Kzhizhanovsky Str. 3, Kiev 252680, UkraineInstitute for Superhard Materials,Ukrainian Academy ofSciences,Avtozavodskaya Str. 2, Kiev 254153, UkraineInstitute for Superhard Materials,Ukrainian Academy ofSciences,Avtozavodskaya Str. 2, Kiev 254153, UkraineInstitute for Superhard Materials,Ukrainian Academy of Sciences,Avtozavodskaya Str. 2, Kiev 254153, UkraineDonbass Institute ofConstruction Engineering,Leninsky Pr. I Apt 44, Donetsk-l02 34102, UkraineDonbass Institute ofConstructions,Shekspira Str. 1-41, Donetsk 340050, UkraineDepartment of Physics,Donbass Institute of Construction Engineering,Schorsa Str. 25, apt. 24, Donetsk 340050, Ukraine
Wayne State University,5050 Anthony Wayne Dr., Detroit, MI 48202, USA
A. Badzian
J. Butler
J. Davidson
V. Dmitriev
G. KelnerS. Khasawinah
H. Niculescu
C. Wallace
R. Wilson
523
Materials Research Laboratory. Pennsylvania State University.University Park. PA USANaval Research Laboratories.Code 6147. Washington DC 20375. USAVanderbilt University.P.O.Box 1555. Station B. Nashville 1N 37235. USACREE Research.2810 Meridian Parkway. Suite 176. Durham. NC 27713. USANaval Research Laboratory. Washington DC 20375. USANuclear Engineering Dept.. University of Missouri-Columbia.E2435 Engr. Bldg. East. Columbia. MO 65211. USAFlorida A&M Univiversity/Florida State University.College of Engineering.2525 Pottsdamer Street. Tallahassee. FL 32310. USABDM International USA.1801 Raldolph Rd. SEt Albuquerque. NM 87106. USAHughes Research Laboratories,Canyon Rd.• Malibu. CA 90265. USA
AFFILIATIONS KEy
ARO
ASB
BDM
BSU
CNRS
CREE
CRI
CRTI
CWRU
DICE
EAS
FAMU
HRL
HT
ffiMIIMO
IPCR
IPTI
IPTM
ISA
Anny Research Office
Academy of Sciences of Belarus
BDM Federal, Inc
Belarussian State University
CNRS Bellevue
CREEEED
Cree Research, Inc.
Central Research Technological Institute
Case Western Reaserve University
Donbass Institute ofConstruction Engineering
Estonian Academy of Science
FAMUIFSU College of Engineering
Hughes Research Laboratories
Hochschule filr Technik
I. Bardin Met. Inst.Institute of Microelectronics & Optoelectronics
Institute of Physical Chemistry ofRomania Academy
loffe Physical Technical Institute
Institute Physics and Technology ofMaterials
Institute of Steel and Alloys
525
Research Triangle Park, NC27709, USA
Minsk, Belarus
Albuquerque, NM 87106,USA
Minsk, Belarus
Meudon 92195, France
St. Petersburg 194021, Russia
Durham, NC 27713, USA
Moscow, Russia
Cleveland, OH 44106-4671,USA
Donetsk 340050, Ukraine
T. Artu, Estonia
Tallahassee, FL 32310, USA
Malibu, CA 90265, USA
Regensberg, Gennany
Moscow 107005, Russia
Warsaw 00-662, Poland
Bucharest, Romania
St. Petersburg, Russia
Bucharest, Romania
Moscow, Russia
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials,© 1995 Kluwer Academic Publishers.
526
ISSP
LBL
MSIEE
MSSAU
MSU
NPI
NPO
NRL
PNLPI
PNPI
RAS
RCEMT
RDT
SEU
SRAS
Tech
TUL
UAS
UB
UFUMC
UR
UT
VU
WI
WSU
WUT
Institute of Solid and SemiconductorPhysics
Lawrence Berkeley Laboratory
Moscow State Institute of ElectronicEngineering
Moscow State Steel and Alloys University
Moscow State University
Nuclear Physical Institute
NPO "Vacuummasphribor"
Naval Reserarch Laboratories
P.N. Lebedev Physics Institute
PNPI
Russian Academy of Sciences
Research Centre of Electronic Materials and Technology
Rockford Diamopnd Technology
State Electrotechical University
Siberian Branch RAS
Technion
Technical University ofLodz
Ukrainian Academy of Sciances
University ofBucharest
University of Florida
University ofMissouri
University of Regensberg
The University of Tokyo
Vanderbilt University
Weitzman Institute
Wayne State University
Warsaw University of Technology
Minsk 220072, Belarus
Berkeley, CA 94720, USA
Moscow 103498, Russia
Moscow, 117936, Russia
Moscow, Russia
St.Ptetersburg, Russia
Moscow 113105, Russia
Washington DC 20375, USA
Moscow 117924, Russia
Leningrad district 188350,Russia
Moscow, Russia
Minsk 220029, Belarus
Columbia, MO, USA
St. Petersburg 197376, Russia
Krasnoyarsk 660036, Russia
Haifa 32000, Israel
Lodz 90924, Poland
Kiev 252680, Ukraine
Bucharest, Romania
Gainesville, FL 32611 USA
Columbia, MO 65211, USA
Regensberg, Germany
Tokyo 113, Japan
Nashville TN 37235, USA
Rehovot 76100, Israel
Detroit, MI 48202, USA
Warsaw, Poland
AUTHOR INDEX
(A)Abernathy, C.R., UF,431Ahmad, F., MSU, 329Aksenova, L.L., RAS, 53Albanesl, E.A., CWRU, 335Alexenlco, A.E., RAS, 225Antonova. T.A.• PNPf, 291Auner. G.W., WSU, 329
(B)Babanln. A.l., fPTl and CREE, 437Bantsekow, S.V.• RAS, 393Baranov. A.M.• NPO. 257Bardyshev. 1.1.. RAS, 123,447Barovskii. N.V.•MSfEE, 305Belyanln, A.F., C[([,f. 225, 297Bochko, A.V., UAS, 393Botev, A.A.. RAS, 115Bouilov, LL., RAS. 115, 123,225Bulko, V.A.• RCEMT. 265Buravov, A.D.• RAS, 447Burdlna, KP.,MSU, 313Butler, I.E., NRL, 105Butler. I.M., NRL, 137
(C)
Chao, C.C.• UMC, 47Charlson, E.J., UMC. 47Charlson, E.M., UMC, 47Chelnokov, V.E.• fPTf. 453Cordos, C., fPTM. 243
(D)Davidson, I.L., VU, 143Dmltriev, V.A., CRf. 453Dragomir. A., fPTM, 285
(E)Ellnson, V.M., NPO. 257
(F)
Fu. T., LBL, 431FUjlshima, A., UT, 137
(G)Galun. E., WI, 137Gameza, L.M.• ASB, 321, 397Gartner, M., fPCR, 285Gavrlchev. K.S., RAS. 362Glelisse, P.I., FAMU. 401Glpplus. A.P.. RAS, 69Glvarglwv, E.l., RAS, 53
(H)Hashimoto. K., UT, 137Hllpfl, R., lIT, 487Hora. H., UR,487Horoyushenko, KG., RAS, 219
(K)
Kalish. R., Tech, 137Karpukhlna, T.A., RAS, 129Khan. A.H., UMC, 475Khasawlnab, S.• UMC, 15, 129.463Khomlch, A.V., RAS,I71Kim, K, CWRU, 335Kiselev, A.N., RAS, 53Konkov, O.l.,fPTf, 291Kotlna, I.M., PNPf, 291Kulakova. 1.1.,MSU. 63Kuo. P.K.•MSU. 329
(L)
Lambrecht. W.R.L.• CWRU. 335Lavrentev. A.A.• SEU. 437Lee, C.H., CWRU. 335Lenane. T.D.,MSU. 329Levy-Clement. C .• CNRS. 137Llaw, B.Y., UMC, 475Lin, L.-T. S., UMC, 187Lukornskil, A.I., ASB, 321, 397
(M)Marchywka. M.I., NRL, 161Markin. A.V.• RAS. 115Maschenko. V.E., fSA. 271McGonlal, M.• NRL, 105Meese. I., UMC, 47
527
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials,© 1995 Kluwer Academic Publishers.
528
(M)Mitura, S., TUL, 235Morosanu, C.• IPTM, 243. 285Moses, D.• NRL. 161Muraki, Y., err. 137
(N)
Nalda, G.A., MSIEE. 305Nalk, R., MSU. 329Nazarov, A.V., IBMI. 249Nicuiescu, R. FAMU. 401Novikov, AP., RCEMT, 265Novikov. N.V•• UAS, 81Newman"N., LBL,431
(0)Obraztsova, E.D., RAS, 219Ositinskaya, T.D., UAS, 81Osvet, A., E4.S, 89
(P)Patseldna, G.V., PNPI,291Pavelescu, G., IPTM, 285Pearton, S.J., UFo 431Pehrsson, P.E., NRL, 105, 161Perov, P.I., RAS, 171Petukhov, AG., CWRU. 335Plekhanov, P'S., RAS, 53Podoba, A.P., UAS, 81Polushin, H.I.,MSSAU, 313Polyakov, V.I., RAS, 171PopOvici, G., RDTand UMC, I, 15, 129, 187,463Prelas, MA., UMC, I, 15,47,129.187,463,487Puzikov, V.M., UAS, 271
(R)Rakova, E.V., RAS, 53Ralchenko, Y.G., RAS, 219Rapinchuk, T.Y., ASB, 421Reuben, C., WI, 137Rossukanyi, N.M.,RAS,I71Rubin, M., LBL, 431Rudenko, A.P., MSU, 63Rukovishnlkov, A.I., RAS, 171Russell, J.N.• NRL, 105
(S)Saveliev, V.D., PNPI, 291Savina, D.L., DICE, 97Schwartz, R.N., HRL, 431Segall. B., CWRU. 335Semenov, AP., SRAS, 225, 297
(8)Semenov, A.V., SRAS, 271Shilova, E.A., RCEMT, 265Shipllo, V.B., ASB, 321, 397, 421Shishonok, E.M., ASB, 397Shishonok, N.A., ASB, 421Shmegerea, S.V.• UAS, 81Sildos, I., E4.S, 89Sizov, V.E., CREE,427Sleptsov, V.V.• NPO, 257Smolin. A.A, RAS. 219Sokolina, G.A., RAS, lIS, 393Sokolov, E.B.•MSIEE. 305Sokolowska, A., WUT, 235Solozhenko. V.L., UAS, 362Spitsyn, B.V., RAS. 31,123,129,225,297Stacy, T, UMC, 47, 475Stepanova, A.N., RAS, 53Stoica, T., IPTM, 243, 285Sung, T., UMC, 15,463Szmidt, I., WUT, 235
(T)Tenne, R., WI, 137Teremetskaya, I.G., RAS, 171Tereshin, S.A., NPO, 257Terukov, E.I., IPTI, 291Thoms, B.D.,NRL, 105TImofeev. M.A., MSU, 115Toldy, V.V., DICE, 97Tomozeiu, N.• UB,243Thhkonen, L.M., PNPI, 291
(V)
Vamin, V.P.• RAS. 171Vassllevsld, K,V., CREE, 427, 453Vavilov, V.S., PNLPI, 373
(W)Wallace, C.B., BMD, 207Wilson, R.G., HRL, 15,431Wu. Z., MSU, 329
(Z)Zaikov. Y.A.• RCEMT, 265Zavada, I.M., ARD, 431Zavt. G., E4.S, 89Zhao, G., UMC, 475Zhirnov, V.V., RAS, 53
KEy WORD INDEX
(A)a-particle. 291absoprtion luminescence. 271acceptors. 69acetone, 47acoustoelectronic converter. 297activation energy. 321active electronics. 97AFM,329AI doped OLC (DLC:AI). 235alloys. 335ALN.329alwnlnum, 329aluminium nitride. 305,401aluminium nitride ceramic, 421ammonia, 137amorphous carbon, 243, 285amorphous hydrogenated carbon, 291analog to digital electronic circult. 207Arrhenius law. 249
(B)band gaps, 401boron, 15boron nitride, 313, 362. 393.401.447cubic, 321
boron phosphide, 401
(C)carbon filament, 47cathodolwnlnescence. 397. 487chemical vapor deposition. 105. 129.305hot filament CVD, 47. 53
chemisorption. 115cluster modeling, 97color center, 373color center laser. 187complex compound AIQ3.NH3 , 305composition measurement. 437conductivity, 115dark, 243photo-conductivity. 243
crystallization kinetics'3, 321crystallization temperature. 129crystallogrnphicorlen~t1on.305
crystal structures. 401current-vol~ge measurement, 161
(D)
dangling bonds. 129DC magnetron sputtering, 257Debye temperature, 401deep level, 97.171defects, 123defect control. 69defects of diamond, 97device application. 475diamond, I, 105,115,123.137,143,401CVD,219film, 129, 143. 171nanophase. 207particle. 53polycrystalline, 15,225single crystalline, 15
diamond electronic devices, 69electronics. 143microelectronics, 143micromechanical devices, 143microsensors, 143sensor, 143technology. 143
diamond film doping. 31by boron, 31by phosphorus. 31by sulfer. 31
diamond growth, 47diamond laser, 187diamond-like. 243, 285diamond-like carbon (DLC). 235diamond mechanical treatment, 225diamond metal-Insulator-semiconductor (MIS)device. 161MIS capacitor, 161
dielectric cons~t, 421diffusion, 15. 265coefficient. 249forced diffusion, 1. 15In amorphous structure. 249
disordered regions, 69distribution coefficient, 31donor impurities, 1donors. 69doping, 1, 97, 373
529
M.A. Prelas et al. (eds.), Wide Band Gap Electronic Materials,© 1995 Kluwer Academic Publishers.
530
(E)
elastic constants, 335electrical conductivity, 393electrochemical reduction, 131electroluminescence, 251electron beam-pumped laser, 481electronic disorder, 211electronics states, 211electronics states of defects and impurities, 91electronic structure, 335energy conversion, 111energy conversion efficiency, 463enthalpy,362entropy, 362epitaxial growth, 329erbium (Er), 431ESR,91,129excimer, 463excitation of luminescence, 69exciton, 2m
(F)
field emission, 53film,.115film structure pyrolysis, 305f1uorescer, 463fluorine containing plasma, 421free energy,surface, 63boundary, 63
(G)gallium nitride, 453Gibbs-Curie's principle, 63glow discharge, 285growth rate, 321
(10heat capacity, 362heat treatment, 305heha-arnmoniacate boron hydride of magnesium, 313heteroepitaxy, 305heterostructures, 111hexagonal cubic sp3-carbon, 211high optical gap, 243, 285high-power semiconductor devices, 453high pressure. 321high pressure synthesis. 401high temperature, 321high-temperture semiconductor devices,453hydrogen,15,41,115hydrogenation, 285hyperfine interaction, 91hypothetical compounds, 335
(I)
impurities, 69, 91interfaces, 335ionicity, 335ion implantation, 69, 313ion milling, 53, 225ion sputtering, 291isotopic impurity, 81
(K)
krypton-85,481
(L)laser etching, 219lattice thermal conductivity, 81lithium. I. 15lithium nitride, 321Lorentzian line, 129luminescence, 181luminescence center. 391
(M)
Madelung potential theory, 481magnetron sputtering, 235mesa structure, 421methane. 41microstructura pararnagentic center. 91microwave semiconductor devices, 453molecular beam eptaxy (MBE), 329MOMBE, 431lon-beam-assisted,431
monocristal. 391multi-quantum well (MQW) structure, 251
(N)
n-type,1n-type doping, 31n-type semiconductor, 91negative electron affinity, 53neutron irradiated laB-type dimond, 89nitrate, 131nitrite, 131, 329nitrogen, Initrogen defects in diamond. 89normal (N) process. 81nuclear battery, 481nucleation rate. 321
(0)optical band gap. 251optical centres, 69optoelectronic applications, 69optoelectronic device, 181
(0)optoelectronic semiconductor devices, 453oxidation, 219oxygen, 15, 115
(P)p-type silicon, 291paramagnatic defects, 129patterning, 219phase diagram, 362phase formation, 265phase transition, 313phonon scattering, 81phosphorus, 1,31,97photocapacitance (PC), 161photoehromic effects, 89photodetector, 161photoelectronics, 171photoluminescence, 431Photovoltaic Energy conversion of Nuclear energySystem (PENS), 463photovoltaics, 171, 463, 487physical properties, 401plasma soure, 329polymorphs, 362polymorphism, 401polytypes, 335, 401polycondensation process, 63positron annihilation, 123,447prototype molecules, 31pyrolysis at high pressue, 313
(R)
radiation defects, 69radiation detector, 291reactive ion etching, 427rectifying juction, 475resonant tunneling effect, 257RF decomposition ofmethane, 235room-temperature spectral holes, 89
(S)sapphire substrate, 305Schottky barriers, 171semiconducting diamond, 69shallow donor, 97silicon, 427silicon carbide, 427, 453simulation, 97sintering activation energy, 421sodium, 1spectral hole burning, 89spectroellipsometry, 285
531
(S)spectroscopyAuger electron spectroscopy (AES), 437high resolution electron energy ion spectroscopy,105
infrared, 105mass, 105secondary ion mass spectroscopy (SIMS), 431,437
spindensity, 129sputtering,243,285Stark energy level, 431stressed diamond film, 487structural relaxation process, 249sulphur, 31surface, 447surface chemistry, 105surface inversion layer, 161switchoptically activated, 207RF,207
synthesiz of cubic boron nitrid, 313
(T)tantalum carbide, 265Tauc's equation, 257TEM,329tetrahedraI carbon films, 271thermal conductivity, 401thermodynamic properties, 362threshold condition, 487tight-binding theory, 97treatment pressure, 397tunabl1ity, 187
(U)ultraviolet phototransformers, 171universal single-atom (USIA) mechanism, 249Urbach tails, 271UV-Vis and IR transmission, 171
(W)wide band gap, 329wide band gap nitrides, 401wide band gap semiconductors, 453wurtzite structure, 475
(X)X-ray diffraction, 329
(Y)Young's modulus, 421
(Z)zero-phonon line, 397