1-s2.0-004060909390012E-main.pdf

Embed Size (px)

Citation preview

  • 8/9/2019 1-s2.0-004060909390012E-main.pdf

    1/13

    Thin Solid Fil ms, 232 (1993) 215-227 215

    Transport properties of low-resistance ohmic contacts to InP

    Thomas Clausen and Otto LeistikoM ik roelekt roni k Centeret, Technical U niv ersit y of Denmark, Bldg. 345E, 2800 Ly ngby (Denmark)

    Ib Chorkendorff and John LarsenDepartm ent of Physics, Technical Uni versity of Denmark, Bldg. 307, 2800 Ly ngby (D enmark )

    (Received February 3, 1993; accepted April 29, 1993)

    Abstract

    The transport properties of conventional Au-based low-resistance ohmic contacts to n- and p-type InP have beeninvestigated. For n-type InP, a good agreement between the minimum specific contact resistance and the bulkdoping density is observed in accordance with an inverse bulk doping density dependence for the specific contactreistance. Drift and diffusion across a thermodynamically stable metalphosphide-InP junction is found to be

    the rate-limiting step of the low-resistance contacts to n-type InP. Tunneling is not observed, but rather asignificantly effective Schottky barrier lowering is initially responsible for the low-resistance contacts. SignificantSchottky barrier lowering is also observed for ohmic contacts to p-type InP, but no total transition frompredominately thermionic emission to drift and diffusion is observed for low-resistance contacts to p-type InP,indicating that the contacts are not fully developed, and that further reduction of the specific contact resistancecan be expected.

    1 Introduction

    The understanding of metal contacts to III-V com-pound semiconductors is probably one of the most

    fundamental and challenging problems with respect todevice fabrication using these materials. From a pro-cessing point of view it is preferable to use simplemetal-semiconductor (MS) contacts of high quality fora low-leakage diode and ohmic contact applications infield effect transistors (FETs) and simple photodiodestructures. For many III-V compound semiconductors,however, this is not achievable, and more complicatedprocessing steps are needed in order to obtain thedesired properties of the device. For InP, Schottkybarriers on n-type InP tend to pin around 0.4-0.5 eV,independent of the metal used, owing to a high density

    of surface states [ 11. This barrier height is too low forsimple MS diodes to be realised because of excessiveleakage currents. Very low Schottky barriers [2, 31 ton-type InP and high barriers up to 0.9 eV [4, 51 forn-type InP have been reported in the literature, indicat-ing that the high density of surface states can be alteredand that it is possible to modulate the Schottky barrierheight over a wide range of values.

    A particularly interesting and important group ofMS contacts for III-V compound semiconductors arethe standard alloyed Au-based ohmic contacts, whichform very low-resistance paths for current transport. Infact, these contacts have provided the lowest measured

    specific contact resistances in comparison with othernon-gold based contacts [6, 71, although great efforts toreplace them have been made [8]. One of the mainproblems in using Au-based metallizations has been the

    rather poor morphology of the contacts after alloying.This problem has been overcome for both GaAs [6] andInP [7] using an optimized annealing sequence, whichimproves the surface morphology of the contacts, and itseems that Au-based alloys are still the best choice forobtaining high quality, low-resistance ohmic contacts toIII-V compound semiconductors. For further reduc-tion of the specific contact resistance it is necessary toobtain an experimental knowledge of the dominanttransport mechanisms at play in the ohmic contacts.

    In this paper we present a thorough investigation ofthe dominant transport properties of Au-based alloyed

    ohmic contacts for both n- and p-type InP. This isachieved by measuring the ambient temperature depen-dence of the specific contact resistance for differentdoping levels of InP. For n-type InP, three differentmetallization schemes were examined: Au/Ni/AuGe,Au/Cr/AuGe and Au/N/Au. For p-type InP, fourdifferent metallization schemes were examined: AuZn,AuZn/Au, Au/Ni/AuZn and AuZn/Cr. For the metal-lization schemes of n-type InP, compositional proper-ties have been investigated using Auger electronspectroscopy (AES). For the metallization schemes ofp-type InP results regarding structural properties havebeen published elsewhere [7].

    0040-6090/93/$6.00 0 1993 - Elsevier Sequoia. All rights reserved

  • 8/9/2019 1-s2.0-004060909390012E-main.pdf

    2/13

    2 6 T. Clausen et al. 1 Transport properties of low-resistance ohmic contacts to InP

    2 Theory

    The Au-based alloying technique for III-V com-pound semiconductors relies upon gold containing adopant (Ge,Sn for the n-type and Zn,Mg for thep-type), from where the dopant is believed to diffuse

    into the semiconductor during annealing, replacing ei-ther substitutional III- or V-atoms in the lattice anddegenerately doping the semiconductor. Through heavydoping of the near-surface semiconductor region thepinned barrier width decreases, enhancing the tunnelingprobability of carriers through the barrier, forming alow-resistance path between the metal and the semicon-ductor.

    This picture is too simple to describe the resultingcomplex interface in alloyed MS contacts, since ex-tended reactions and phase formation involving allatomic species of both the metallization and the semi-

    conductor are likely to occur based on thermodynamicconsiderations. The metallization scheme and the semi-conductor normally involve a total of 4-5 differentatomic species layered in a complex multilayer struc-ture. Therefore, it is difficult to predict the reactionpath as a function of the annealing temperature. Withrespect to the electrical properties of the contact we cansimplify the complex interface of an alloyed contact byconsidering the different interfacial structures as local-ized diodes in parallel. The diode structures are charac-terized by different Schottky barrier heights anddifferent local semiconductor doping densities, and the

    diode current path with the lowest resistance will domi-nate the measured specific contact resistance, providedthat the surface coverage of that particular interfacialstructure is large enough. Otherwise, the specific con-tact resistance will be a weighted average with respectto surface coverage of all the diodes in parallel.

    The transport properties of carriers in a MS contactdepends mainly on the ambient temperature and thedoping density of the semiconductor. Four differentdominating modes, according to classical transport the-ory for MS contacts [ 1, 91, can be formulated as afunction of the doping density. These are thermoionic

    emission (TE) and/or drift and diffusion (DD) for lowdoping densities, thermoionic field emission (TFE) forintermediate densities and field emission (FE) for veryhigh doping densities. In the following we present thetransport equations for each of the classical transportmechanisms for alloyed ohmic contacts on n-type semi-conductors with emphasis on the functional dependenceof ambient temperature and bulk doping density, bear-ing in mind that the equations equally well apply to ap-type semiconductor with suitable replacements of in-dices and numbers. Also, we present important non-classical transport mechanisms for ohmic contactswhich have been developed during the last ten years or

    so, in order to explain both the experimentally observedlow specific contact resistances and the experimentallyobserved inverse bulk doping density dependence forohmic contacts to a variety of semiconductors [lo- 171.

    Thermionic emission relies upon the emission of car-riers across the top of a barrier, the Schottky barrier &,,

    and is very dependent on the ambient temperature. Thespecific contact resistance, r of an ideal MS contactin which thermionic emission is dominant (i.e. thebackground doping density is low), is best described by

    [l> 9

    kr,(TE) = -

    qA*T

    where k is the Boltzmann constant, A* is Richardsonsconstant, q is the elementary charge and T is theambient temperature. As discussed above, this equationdoes not hold for an alloyed ohmic contact, since it

    does not take into account the fractional surface cover-age S,, of the dominating low-resistance phase. This canbe accounted for by introducing a microstructural fac-torj&S,,) as discussed by Chu et al. [ 181, which is thenprimarily a measure of surface coverage of the dominat-ing low-resistance phase. Furthermore, the alloyed con-tact forms a number of parallel diodes with differentSchottky barrier heights. An effective Schottky barrier&,eff forms, which is a weighted value with respect tosurface coverage of all the different Schottky barriers inthe contact [19]. For an alloyed MS contact, assumingthermoionic emission, we can then write

    r,(TE) = (4

    Note that the Richardson constant has also been incor-porated into the microstructural factor.

    The effects of drift and diffusion are often neglectedwith respect to the transport of carriers across a MSinterface by assuming infinite mobility in the semicon-ductor and constancy of the Fermi potential 4 (i.e. thedriving force for diffusion dy/dx = 0) throughout thedepletion region of the MS contact. These are fairlygood approximations for high mobility semiconductors

    and reasonably high barrier MS contacts. However, forlow-mobility semiconductors (say p-type InP) and lowbarrier MS contacts (for instance, metal contacts ton-InAs), drift and diffusion becomes important andmust be taken into account. The specific contact resis-tance of an alloyed MS contact in which drift anddiffusion is dominant, is best described by [9]

    rc(DD) =k&(T)T

    j,,& $Jq p,( T)N,( T)N, exp(3)

    where E(T) is the temperature dependent dielectric con-stant (a,~~) of the semiconductor, ,uJT) is the tempera-ture dependent mobility of the carriers, N,(T) is the

  • 8/9/2019 1-s2.0-004060909390012E-main.pdf

    3/13

    T. Clausen et al. / Transport properti es of low -resistance ohmi c contacts to InP 217

    temperature dependent effective density of carriers inthe conduction band of the semiconductor, Nd is thebulk doping density and Jr,,,(&) is the correspondingmicrostructural factor.

    The field emission of carriers relies on a very thintunneling barrier between the metal and the semicon-

    ductor, where carrier emission directly from the Fermilevels through the tunneling barrier is highly probable.The specific contact resistance of an alloyed MS contactin which field emission is dominant (i.e. the surfaceand/or bulk doping density is high), is best described byW

    rc(FE) = GJFE(WG

    where C, and C, are virtually temperature independentconstants, 4, is the tunneling barrier and jr,( ) is thecorresponding microstructural factor, which again is a

    measure of fractional surface coverage S,, of the domi-nating low-resistance phase.

    Thermoionic field emission ties the two extremes ofthermoionic emission and field emission together in theintermediate region of doping densities. This region isimportant because, as recieved commercial highly-doped InP wafers, both n- and p-type InP have dopingdensities in this region. For an as-deposited ideal, non-alloyed, MS contact on these wafers, the carriers tunnelacross the thin top part of the Schottky barrier, but stillhave to be thermally activated in order to reach the toppart of the barrier. We are interested in alloyed ohmic

    contacts and, therefore, we only consider the functionaldependence of the ambient temperature and the bulkdoping density for the two extremes, and only discusspossible deviations with respect to the transition case ofthermoionic field emission.

    Comparing the basic equations for thermoionic emis-sion, drift and diffusion and field emission (eqns. (2),(3) and (4) respectively) the major differences reside inthe ambient temperature dependence and the depen-dence of the doping density. While transport of carriersvia both thermoionic emission and drift and diffusionare very temperature sensitive, transport via field emis-

    sion is virtually temperature insensitive. Transport ofcarriers via thermoionic emission is virtually insensitiveto the doping density of the semiconductor, apart frombarrier lowering due to image forces, while both fieldemission and drift and diffusion are sensitive to thedoping density. In fact, the transport of carriers viadrift and diffusion can qualitatively explain the ob-served l/N dependence for ohmic contacts for both n-and p-type Si and III-V semiconductors [lo- 171.

    None of the transport equations above can fullyexplain both the very low specific contact resistancesand the observed inverse dependence of doping densityfor ohmic contacts, unless the Schottky barrier of the

    MS junction is lowered significantly during the alloyingsequence. Several publications have dealt with the prob-lem of obtaining low specific contact resistances with aninverse dependence of bulk doping density, and there isa general agreement that the rate-limiting step in anoptimum alloyed ohmic contact is an n+ -n step

    formed during alloying. In this widely accepted modelthe carriers first surmount the MS junction via a field-assisted mechanism (tunneling) and then surmount thebarrier established between the alloyed highly dopedn +-layer and the bulk n-layer. The specific contactresistance is then a series combination of the MS junc-tion resistance according to field emission theory andthe n+ -n junction resistance. The main controversyhas therefore concerned how the carriers actually sur-mount the n + -n step. Both diffusion-assisted [ 12, 131and thermoionic-emission-assisted [ 14- 171 transport ofcarriers across the step are postulated, and each of them

    can account for the l/N, dependence quite satisfacto-rily. For the thermoionic emission approach, eqn. (1) isusually applied to the step with the Schottky barrierreplaced by the step barrier (i.e. the difference betweenthe conduction band minima and the Fermi potential).Thus, the step barrier is dependent on the bulk dopingdensity, and it can be shown that [ 141

    r c,n+ -.(TE) = gTzor

    r c,n+ -.VE) =j TEc;,)qTdif eqn. (2) is applied more correctly to the step. SinceN o= T. rc,n+_,(TE) cc To., quite differently from theGassical case of thermoionic emission, wherer,(TE) cc l/T.

    The diffusion approach to the problem has recievedlittle attention, mainly because the predicted minimumvalues for the specific contact resistance as a function ofthe bulk doping density seems to be too high comparedwith corresponding measured values [ 14- 171. However,most of the experimental values for the specific contact

    resistance presented in refs. lo- 17 are measured byeither the transmission line method (TLM) or theShockley method for contacts on epitaxial or implantedlayers. Values of the specific contact resistances mea-sured using these methods should only be quoted withcaution, since both methods can result in values whichare more than an order of magnitude too low [20] if aproper value of the sheet resistance of the semiconduc-tor underneath the alloyed region is not utilized. Themain difference between the thermoionic emission ap-proach and the diffusion approach is the pre-exponen-tial factors in front of the step barrier exponential,since the diffusion approach also simply replaces the

  • 8/9/2019 1-s2.0-004060909390012E-main.pdf

    4/13

    218 T. Clausen et al. / Transport properties of low-resistance ohmic contacts to InP

    Schottky barrier with the step barrier formed betweenthe Fermi level and conduction band minima of ann+ -n step. According to Dingfen and coworkers [ 12,131

    c,,+-.(D) =2(27~)~~(m*kT)

    rq2h 3N~cN,K~

    (7)

    where N,,, is the donor concentration in the highlydoped alloyed region and KB is a proportionality con-stant between the electron concentration level in thealloyed region and the donor concentration, which isassumed to have a simple linear relationship. From eqn.(7) diffusion-assisted transport across the rate-limitingn+ -n step is more temperature sensitive than isthe thermoionic emission assisted transport, sincer C,n+m. D) x T .

    For each of the transport equations (eqns. (l)-(7))there is a different dependence on the ambient tempera-

    ture and bulk doping density with respect to the others.Therefore, by measuring the ambient temperature de-pendence for ohmic contacts to differently doped sam-ples, the actual transport mechanism of alloyed ohmiccontacts with low specific contact resistance can beresolved. If, for example, thermoionic emission acrossan effective Schottky barrier is dominant, then eqn. (2)applies and a plot of ln(r,T) US. 1000/T will yield astraight line with the slope directly proportional to theeffective Schottky barrier height and the intercept onthe r,T axis proportional to the surface coverage. Devi-ations from linearity indicate the presence of othertransport mechanisms, such as drift and diffusion,which have a different ambient temperature dependencewith respect to the specific contact resistance (eqn. (3))or field emission related processes which are virtuallytemperature independent (eqn. (4)).

    3. Experimental procedure

    The InP wafers employed were ( 100) liquid encapsu-lated Czochralski (LEC) grown n+ -1nP (S, 6 x lOIcmm3), n- -1nP (undoped, 5 x lOI cm-) and p+-InP

    (Zn, 5 x lOI cm p3). The contact metal dots weredefined by standard optical lithography in a sputter-deposited SiO, layer passivating the InP surface. Thepassivation, and, especially the associated high vacuumtreatment of at least 2-3 h at background pressuresbetter than 1O-6 Torr, where water vapour fromthe surface of the semiconductor is removed [21], isvery important for assuring a high quality ohmic con-tact. Remaining surface oxides were removed inNH,OH:H,O/NH,OH:H,O,:H,O prior to the deposi-tion of the metal layers. The metal layers were de-posited by resistive evaporation using tungsten boatsand RF sputtering at background pressures of 10 ~ to

    10 - 6 Torr. The metallization schemes investigated wereAu/Ni/AuGe (500/250/1000 A), Au/Cr/AuGe (500/250/1000 A) and Au/Ni/Au (500/250/ 1000 A) for n-typeInP and AuZn (1000 A), AuZn/Au (10001200 A), Au/Ni/AuZn (500/250/1000 A) and AuZn/Cr ( 1000/70 A)for p-type InP. AuGe (1000 A) for n-type InP and

    AuZn/Au (1000/200 A) for p-type InP were depositedat the backside of the wafers for reliable, large-areabackside contacts. The concentration of Ge in theAuGe alloys was approximately 12 wt.% while theconcentration of Zn in the AuZn alloys was about 10wt.%. After metal layer deposition, each wafer wassliced into 6-8 samples for subsequent annealing in aHeatpulse 410 rapid thermal annealing (RTA) system.The temperature was varied from 300-500 C at aconstant annealing time of 20 s. As a final step, Au/Crmetal pads in contact with the dots were patterned ontop of the SiO, layer for probing purposes.

    I- v measurements for the ohmic contacts were per-formed using a Hewlett-Packard HP4145B parameteranalyzer and a probing station, where the samples weremounted on an alumininum chuck. Vacuum on thebackside of the samples ensured a good backside con-tact to the chuck. Furthermore, it was possible to heatthe chuck between room temperature and 175 C byilluminating the backside of the chuck, which wascoated with an antireflection coating, with an intensehalogen lamp. Temperature measurements were madedirectly underneath the samples by means of a standardthermocouple. I- V measurements were performed bysweeping an appropriate current range (from - 0.1 Ato 0.1 A for the best contacts, i.e. r, < lo- 0 cm) thenmeasuring the associated total voltage drop from theprobing pad on top of the SiO, layer across the metal-semiconductor interface to the backside of the chuck.Parasitic voltage drops are present in the measuredtotal voltage drop, because of spreading resistance inthe bulk of the samples and additional resistances inmetal wires, backside contacts and chuck resistance, butthey can be accounted for by using a well-tested methodfor calculating the specific contact resistance, Y,.

    The measurement of the specific contact resistance Y,was performed using the Cox and Strack technique[22]. In general each sample contained 8- 10 chipssuitable for measurements. Each chip consisted of 15-20 dots with 5-8 different radii suitable for Cox andStrack measurements. The radii of the dots were variedfrom 5 to 100 pm. Thus, a total of about 150-200independent measurements were performed for eachsample, and the measured total resistance for each dotarea on a specific sample were averaged over around20- 30 independent measurements, thereby ensuringgood statistics. The spreading resistance contributionwas calculated for the different dot areas based on theresistivity and the thickness of the InP wafers [22]. For

  • 8/9/2019 1-s2.0-004060909390012E-main.pdf

    5/13

    T. Clausen et al. / Transport pr operties of low -resistance ohmic contacts t o InP 219

    some of the lowest values of r,, the main contributionto the total resistance for the large dot areas was thespreading resistance, but for the smaller dot areas, itremained a fraction of the total resistance.

    A PHI 590A scanning Auger microprobe (SAM)with a base pressure of less than 2 x lo- mBar,

    combined with in situ argon ion milling, was used tostudy the compositional properties of the alloyed ohmiccontacts to n-type InP. Auger spectra and depth profi-les were recorded using an angled sample holder withrespect to the primary electron beam. The Auger spec-tra were obtained with a 5 keV primary beam. Ar+sputtering was performed with a PHI 04-303 sputtergun. The Ar + ion energy was 2 keV and the fluencywas 25 PA cm-2. During the various sputtering cycles,the operating pressure was increased from about2 x 10 - lo to 2 x 10 -7 Torr by leaking argon gas intothe working chamber.

    In order to obtain atomic concentration profiles forthe AES spectra, the sensitivity factors for each of theelements are necessary. Using a Au-foil and an InPcrystal, we calculated the important sensitivity factorsof Au, In and P by measuring the peak-to-peak heightsof the signals and comparing them with the peak-to-peak height of a standard Ag foil, which is defined tohave a sensitivity factor of 1. From the peak-to-peakheights in the differentiated mode, sensitivity factors forgold, indium and phosphorous using a 5 keV primarybeam were calculated to be 0.402, 0.952 and 0.541,respectively. For Ni and Cr, sensitivity factors of 0.27

    and 0.19 were assumed. Carbon and oxygen were alsopresent in the metallizations as deduced from the Augermeasurements. However, these signals were onlypresent in the near-surface region of the metallizations,and, therefore, for the sake of clarity, they were omittedfrom the calculated concentration depth profiles.

    4. Results

    4.1. Electrical measurements at room temperatureThe as-measured room temperature values of the

    specific contact resistance r, for different metallizationschemes to highly doped n-type InP are shown in Fig. 1and listed in Table 1 for the important values. Thelowest value of rc, z 7 x 10 _ * R cm2, is obtained at anannealing temperature of 420-450 C using a AuGeNilayer combination. In fact, the AuGeNi layer combina-tion provides the lowest value of rC for temperaturesabove 370 C annealing, and the lowest values of rC forall the other metal layer combinations, i.e. AuGeCr,AuGeNiCr and AuNi, are all at least one order ofmagnitude higher than for the AuGeNi combinationabove this annealing temperature. Note that there is asignificant hump in r, around 410 C annealing for the

    lo; 60 340 400 460 520

    Ann. temp. (C)

    Fig. 1. Measured specific contact resistance US. annealing temperaturefor ohmic contacts to n + -1nP. -, Au/Ni/AuGe; , Au/O/AuGe; - - -, Au/Ni/Au; *-+*, Au/Cr/Ni/AuGe.

    AuGeNi combination, which is also present for boththe AuGeCr and AuGeNiCr combinations, althoughless pronounced, but not for the AuNi combination.

    The as-measured room temperature values of the

    specific contact resistance for AuGeNi and AuGeCrcombinations to lightly doped n-type InP are shown inFig. 2 and listed in Table 1. As can be seen from thisfigure and Fig. 1, the curve appearences coincide in thelow-resistance region with a shift in the value of thespecific contact resistance of three orders of magnitude,corresponding quite closely to the shift in bulk dopingdensity, which is also three orders of magnitude. Infact, the lowest value of r, for the AuGeNi combinationto lightly doped n-type InP, z 7 x 10e5 R cm, isalmost exactly three orders of magnitude larger thanthe corresponding value for highly doped InP. Thus,the data presented here closely follow the inverse bulkdoping density dependence of minimum specific contactresistance observed by many different authors [lo- 171.For the AuGeCr combination, the value of r, for500 C annealing is unexpectedly low, since we wouldexpect it to be higher than the value at 440 C anneal-ing, if it were to follow the behaviour of AuGeCr onhighly doped InP.

    The as-measured room temperature values of thespecific contact resistance for different metallizationschemes on highly doped p-type InP are shown in Fig.

    lo SO1340 400 460 520

    Ann. Temp. (C)

    Fig. 2. Measured specific contact resistance OS. annealing temperaturefor ohmic contacts to n--1nP. -, Au/Ni/AuGe; - - -, Au/Q/AuGe.

  • 8/9/2019 1-s2.0-004060909390012E-main.pdf

    6/13

    220 T. Clausen et al. / Transport properti es of low -resistance ohmic contacts t o InP

    TABLE 1. List of rc values and the associated values of the mean standard deviation for ohmic contacts to n-type InP as function of themetallization layer combination and the annealing sequence

    Metallization layer combination Annealing sequence(C s-1) cm2)

    Remarks

    Au/Ni/AuGe/n--1nP

    Au/Ni/Au/n+-InP

    Au/Cr/AuGe/n+-InP

    Au/Ni/AuGe/n -1nP

    Au/Cr/AuGe/n-InP

    320/20370/20400/20420/20440/20500/20

    300/20350/20400/20440/20500/20

    320/20370/20400/20420/20440/20

    500/20

    300/20350/20400/20420/20440/20465120500/20

    300/20350/20400/20440/20500/20

    3.7 x 1o-5 * 1 x 10-h6.2 x IO-* 1.0 x lo-1.1 x 1o-7 * 1 x 10-s7.3 x 10-s * 5 x 10-g7.4 x 10-s + 8 x 10-g1.5 x lo- f 1 x 10-s

    6.4 x 1O-3 + 2 x IO-1.39 x 1o-2 f 5 x 10-41.62 x 10m4 + 9 x lO-67.1 x lo-5+ 1 x 10-h1.5 x lo-4+ 2 x 10-s

    2.3 x 10-s 2 1 x 1O-61.40 x 10-e+ 5 x 10-s1.42 x 10mh k 3 x 1O-88.5 x lo- + 7 x 10-s2.5 x 10m6 f 3 x lo-

    1.5 x 10-s* 1 x 10-e

    3.4 x 10-h + 3 x 10-T1.6 x 10m6+ 1 x lO-74.8 x lo-+ 1 x 10-x1.2 x 10-e * 2 x 10-T9.3 x lo-7 & 1.2 x IO-1.11 x 10-b+ 3 x 10-s4.0 x 10--h* 1 x 10-7

    2.28 x 10m3 k 7 x 10m58.0 x 10m4 k 4 x lo--3.9 x 1om4 + 1 x 10-z3.5 x 1om4 + 4 x 10-51.6 x lo- It 2 x lo-

    Ni(70 nm)/AuGe(300 nm)/n+-InP (8 x 10 cm-), 400 C/2 min:rc = 1.2 x 10e6 R cm [23]

    Au( 100 nm)/Ni(7.5 nm)/AuGe(SO nm)/n+-InP (10 cmm3), 450 C/25 s:r, = 8 x lo- Q cm2 [21]

    Ni(25 nm)/AuGe( 125 nm)/n+-InP (10 cm-l), 375 C/3 min:r, = 8 x 10e6 Q cm [24]

    Ni(30 nm)/AuGe(lSO nm)/n+-InP (8 x 10 cm-), 460 C/30 s:r,=6.5 x 10~sQcm [25]

    A~(400 nm)/Ni,P(lOO nm)/n+-InP (3 x lOI* cme3), 300 C/30 min:r, = 3 x 10m6 R cm2 [26]

    Layer deposition by thermal evaporation

    TABLE 2. List of rc values and the associated values of the mean standard deviation for ohmic contacts to p-type InP as function of themetallization layer combination and the annealing sequence

    Metallization layer combination Annealing sequence r,(C s-1) (a cm2)

    Remarks

    AuZn/p+-InP

    AuZn/Au/p+-InP

    Electroplating layer deposition.

    Au/Ni/AuZn/p+-InP

    350/20 13.7 * 0.5385/20 10.4* 1 x 10-X400/20 1.9 x 10-z+ 1 x 10-s415/20 3.5 x 1o-3 * 5 x lo-*425120 3.7 x lo-h&9 x 10-s440/20 1.4 x 10-s* 1 x 10-h450/20 2.6 x lo- + 3 x lO-6

    400/20 4.4 & 1.1430/20 5.1 x lo-s*4 x 10--e440/20 7.0 x 1O-6 + 6 x lo-465120 4.2 x 10m5 *4 x lo-500/20 8.2 x 10-s + 3 x 10-G

    410/20 1.14 x lo-4+5 x 10-C415/20 2.16 x 10m4 + 5 x 1O-6430/20 2.7 x 10m5+4 x lo-440/20 6.2 x 1O-4 + 2 x lo-450/20 8.0 x 10-d & 3 x 10-z465120 8.4 x 1O-4 + 6 x lO-S

    AuZn(70&100 nm)/p+-InP(108cm-3),400 C/2 s: rc = 7.2 x 10m5 R cm2 [27]

    AuZn/Au(200 nm)/p+-InP(7 x 10 cm-),420 C/45 s: r= = 4 x 10-s Q cm [28]

    A~(250 nm)/ZN(40 nm)/Au(40 nm)/p-InP(5.5 x 10 cm-), 400 C/180 s:r, = lo- R cm [291b

    A~(250 nm)/Cr(25 nm)/Ni( 15 nm)/Au(10 nm)/AuZn(50 nm)/p-InP(10scmm3), 410 C/30 s:r, = 2 x 10-s Q cm2 [301b

    bLayer deposition by thermal evaporation.

  • 8/9/2019 1-s2.0-004060909390012E-main.pdf

    7/13

    T. Clausen et al. / Transport pr opert ies of low -resistance ohmi c contacts t o InP 221

    350 380 410 440 470 500

    Ann. temp. (C)

    Fig. 3. Measured specific contact resistance vs. annealing temperaturefor ohmic contacts to p + -1nP: -, AuZn; , AuZn/Au; - -,Au/Ni/AuZn.

    3 and listed in Table 2 for important values. The lowestvalue of r,, = 7 x 10m6 0 cm2, is obtained at anannealing temperature of 440 C using a single AuZnlayer. The minimum is very narrow, and there is a very

    abrupt transition from a non-ohmic behaviour to anohmic behaviour around 400 C. The addition of a thinAu layer, seperating the AuZn layer and the InP,originally introduced in order to improve the adhesionof the contact metallization layers, does not signifi-cantly alter the rc W. annealing temperature depen-dence. The onset of ohmic behaviour is loweredapproximately 10 C, but the minimum specific contactresistance in the narrow 440 C annealing range is twotimes larger than for the single AuZn layer case. Theaddition of a Ni layer significantly alters the r, us.annealing temperature dependence as is evident from

    Fig. 3. First of all, the onset of ohmic behaviour islowered significantly with respect to the single AuZnlayer case. Secondly there is hump around 415 C, aswas the case for AuGeNi on highly-doped n-type InP.Finally the value of rC ncreases very abruptly by nearlytwo orders of magnitude at annealing temperaturesabove 440 C. The minimum specific contact resistancefor AuZnNi metallization, = 2.7 x 1O-5 Q cm, occursat 430 C annealing. Evaporating a thin Cr layer (70 A)prior to the deposition of the AuZn layer altered the r,us. annealing temperature dependence even more drasti-cally, since the contacts were highly non-ohmic up to500 C annealing, and then only a high value of r, of3.3 x 10m3 R cm* was measured.

    From Table 1 and 2 it is seen that the standard meandeviations for the specific contact resistance are small(in general less than 15% of the value of the specificcontact resistance), indicating that the experiments havebeen well controlled and that the metallization havereacted uniformly across most samples. Also, in Tables1 and 2, comparative values of the specific contactresistance from the work of other authors on the samemetallization systems are listed. From this it is seen thatour measured values for the specific contact resistanceare indeed among the lowest ever reported [21, 23-301.

    4.2. Electrical measurements at elevated temperaturesThe ambient temperature dependence of the specific

    contact resistance, represented in semilogarithmic r,Tus. 1000/T plots, are shown for different annealingtemperatures in Figs. 4, 5 and 6 for AuGeNi, AuGeCrand AuNi to highly doped n-type InP. From Fig. 4 the

    linearity of the plots are quite good and the slopes arepositive for annealing temperatures below 370 C, cor-responding to the positive effective Schottky barrierassuming thermoionic emission (eqn. (2)) of carriersacross the MS interface. For annealing temperaturesabove 370 C the slopes are negative and the linearity ofthe plots is not good. This behaviour is not expected ifclassical thermoionic emission (eqn. (2)) across the MSinterface is dominant, since a negative effective Schot-tky barrier is not expected. Rather a different transport

    I........lO Z.0 25 3.0 3.5 40

    1000/T (K-l)

    Fig. 4. Temperature dependence of specific contact resistance forAu/Ni/AuGe/n + -1nP.

    I

    52i

    lo--/As-dep.

    E0 to-

    1000/T (K-l)

    Fig. 5. Temperature dependence of specific contact resistance forAu/Cr/AuGe/n + -TnP.

    ; ;:,imo1000/T (K-l)

    Fig. 6. Temperature dependence of specific contact resistance forAu/Ni/Au/n + -1nP.

  • 8/9/2019 1-s2.0-004060909390012E-main.pdf

    8/13

    222 T. Cl ausen et al. 1 Transport pr opert ies of low -resistance ohmi c contacts t o InP

    mechanism is dominant, which is not proportional tol/T for the pre-exponential factor, but rather it isproportional to T, where x is positive and dependenton which of the other transport mechanisms is domi-nant. For classical drift and diffusion (eqn. (3)) x isdependent on the temperature dependence of the mobil-

    ity, which is different for highly-doped and lightly-doped samples. For non-classical thermoionic emission(eqn. (6)) x z 0.5, independent of the doping density ofsamples, and for non-classical drift and diffusion (eqn.(7)), x z T2, also independent of the doping density.For field emission x x 0, as is not the case since Y,increases with increasing ambient temperature. Thus,field emission can be ruled out as the rate-limiting stepof MS ohmic contacts using Au-based metallizations.For AuGeNi at 440 C annealing, the most fully devel-oped contact with respect to low values of r. we findthat rc cc T*, which is different from non-classical ther-

    moionic emission, but in accordance with non-classicaldrift and diffusion.For lightly-doped n-type InP, semilogarithmic plots

    of r,T vs. 1000/T are shown in Figs. 7 and 8 forAuGeNi and AuGeCr, respectively. For AuGeNi (Fig.7), there is a similar transition from predominatelythermoionic emission according to eqn. (2) to anotherdominating transport mechanism in accordance withFig. 4 for AuGeNi on highly-doped n-type InP. For afully developed AuGeNi ohmic contact to lightly-dopedn-type InP, i.e. annealing at 440 C, rc cc T3.. Thus,

    PI.lo i .4 2.6 3.2 3.6 4.0

    1000/T (K-l)

    Fig. 7. Temperature dependence of specific contact resistance forAu/Ni/AuGe/n - -1nP.

    I0 2.41.......1.1.....12.8 3.2 3.6 4.0

    1000/T (K-l) 1000/T K-l)

    Fig. 8. Temperature dependence of specific contact resistance for Fig. 10. Temperature dependence of specific contact resistance forAu/Cr/AuGe/n _ -1nP. Au/Ni/AuZn/p + -1nP (--) and AuZn/p + -1nP (- ~ -).

    there is a significant difference in the ambient tempera-ture dependence for fully developed ohmic contacts onhighly-doped n-type InP and lightly-doped n-type InP.This difference cannot be explained by non-classicaldrift and diffusion, since non-classical drift and diffu-sion predicts an identical ambient temperature depen-

    dence regardless of doping density. However, drift anddiffusion according to eqn. (3), predicts a differentambient temperature dependence of the pre-exponentialfactor, owing to the difference in ambient temperaturedependence of the mobility. From the measurements ofthe ambient temperature dependence for fully devel-oped ohmic contacts to different doping densities ofn-type InP, we therefore conclude that the rate-limitingstep is classical drift and diffusion across a MS junctionwith no effective Schottky barrier.

    For highly-doped p-type InP, semilogarithmic plotsof r,T us. 1000/T are shown in Figs. 9 and 10 for

    AuZnAu and AuZnNi( AuZn), respectively. For theseplots, thermoionic emission across an effective Schottkybarrier height, according to eqn. (2) with the effectivebarrier height deduced from the slopes and linearity ofthe plots, satisfactorily explains the observed ambienttemperature dependence of r,. Comparing Figs. 9, 10and 3, the reduction in the specific contact resistancetowards 440 C annealing for AuZnAu closely matchesthe reduction in the effective Schottky barrier height,and the effective barrier heights for both AuZn andAuZnAu are comparable. Furthermore, for AuZnNi,

    10 ;

    1000/T K-l)

    Fig. 9. Temperature dependence of specific contact resistance forAulAuZnlp + -1nP.

  • 8/9/2019 1-s2.0-004060909390012E-main.pdf

    9/13

    T. Clausen et al. / Transport properties of low-resistance ohmic contacts to InP 223

    the observed increased value of r, for annealing temper-atures above 440 C (Fig. 3) is also apparent from Fig.10, where an increase in the effective Schottky barrierheight is observed for annealing temperatures above440 C. For the lowest effective Schottky barriers, i.e.annealing at 440 C for AuZn and AuZnAu, the total

    resistance of the contact decreased only for the smallercontact dots as the ambient temperature was raised,indicating that thermoionic emission, according to eqn.(2), is dominant. For the larger contact dots, the totalresistance of the dots increased with increasing ambienttemperature, indicating the presence of another domi-nant transport mechanism such as drift and diffusion.Thus, the transport properties of MS ohmic contacts top-type InP resemble that of n-type InP, apart from theimportant fact that the contacts to p-type InP are notfully developed, meaning that there is no transistionfrom predominately thermoionic emission according to

    eqn. (2) to predominately drift and diffusion accordingto eqn. (3), as expected and as is the case for fullydeveloped MS ohmic contacts to n-type InP.

    4.3. Auger electron spectroscopy measurementsThe Auger sputter depth profiling of AuGeNi,

    AuGeCr and AuNi metallizations are shown in Figs.11, 12 and 13, respectively. In all figures the depth isgiven in sputter time since the yield will change withcomposition. From the figures, a general path of reac-tions leading to the observed low values of rC for themetallizations can be deduced. On comparing AuGeNi

    (Fig. 11) and AuGeCr (Fig. 12) combinations, an im-portant difference in the phosphorous profiles of the

    0 1200 2400Sputtering time (set)

    b) 1.0 +Y 320C 1

    Sputtering time (set)

    d

    d)

    4

    I9

    1.0

    gj 0.8

    E 0.6

    ,; 0.4

    2 0.2

    0.0

    1.0

    2 0.8

    E 0.6

    -; 0.4

    2 0.2

    0.0

    1.0

    ; 0.8

    E 0.6

    .zE 0.4

    2 0.2

    0.0

    0 1200 2400

    Sputtering time set)

    Sputtering (set)

    I 42OC

    Sputtering time (set)

    Sputtering time set)

    1.0500C

    B 0.8E 0.6

    ,; 0.4

    2 0.2

    Fig. 11. Sputter Auger profiles for alloyed samples of Au/Ni/AuGe/InP: a) as-deposited; b) 320 C; c) 370 C; d) 400 C; e) 420 C; f)440 C; g) 500 C.

    0.00 1200 2400 3600

    Sputtering time (set)

  • 8/9/2019 1-s2.0-004060909390012E-main.pdf

    10/13

    224 T. Clausen et al. 1 Transport pr opert ies of low -resistan ce ohmic contacts to InP

    a) 1.0

    0.8

    0.6

    0.4

    0.2

    0.0

    b) 1.0

    ; 0.8

    8 0.6

    .g 0.4

    2 0.2

    0.0

    cl 1.0

    2 0.8

    E 0.6

    j 0.4

    2 0.2

    0.0

    d) 1.0

    0.8

    8 0.6

    .i 0.4

    2 0.2

    0.0

    9 1.0

    i 0.8

    E? 0.6

    .IE 0.4

    2 0.2

    0.0

    Cr As-dep

    0 900 1800

    Sputtering time (set)

    300

    _ Au

    0 1200 2400 3600

    Sputtering time (set)

    I 350C

    0 1200 2400 3600Sputtering time (set)

    4o 0

    0 1200 2400 3600 4800

    Sputtering time (set)

    I 42OC 1

    Sputtering time (set)

    0 1.0 440c2 0.8 _

    E 0.6 -

    Sputtering time set)

    g) 1.00.8

    : 0.6

    .; 0.4

    2 0.2

    0.0

    Sputtering time (set)

    Sputtering time (set)

    Fig. 12. Sputter Auger profiles for alloyed samples of Au/Cr/AuGe/InP: a) as-deposited; b) 300 C; c) 350 C; d) 400 C; e ) 420 C; f)440 C; g) 465 C; h) 500 C.

    two combinations are observed. The phosphorousprofile between the Cr-P layer and the bulk InP dropsto a value close to its background level for the AuGeCrcombination at annealing temperatures above 420 C.For the AuGeNi combination, there is a continousphosphorous signal between the Ni-(Ge) -P [31] layerand bulk InP for all annealing temperatures above370 C, the onset of very low specific contact resistance.In fact, at 500 C annealing, a Ni-(Ge) -P layer hasformed adjacent to the InP surface with a Au-In layerlying above for AuGeNi, while a Au-In layer formsadjacent to the InP surface with a Cr-P layer lyingabove with little contact to the InP surface for AuGeCr.For annealing temperatures above 370 C, there is al-ways a larger amount of phosphorus in contact with thebulk InP for the AuGeNi combination as comparedwith AuGeCr. Other differences are, of course, present.The reactions between the metal constituents in themetallization and the InP has developed to quite a large

  • 8/9/2019 1-s2.0-004060909390012E-main.pdf

    11/13

    T. Clausen et al. 1 Transport properties of low-resistance ohmic contact s to InP 225

    a)2 0.8

    8 0.6

    .; 0.4

    2 0.2

    b)2 0.8

    0.6

    900 1800

    Sputtering time (set)

    320C 1

    1200 2400

    Sputtering time (set)

    0.8

    0.6

    0.4

    0.2

    0.0

    Sputtering time (set)

    Fig. 13. Sputter Auger profiles for alloyed samples of Au/Ni/Au/InP:a) as-deposited; b) 320 C; c) 370 C.

    extent at 300 C and 320 C annealing for bothAuGeCr (Fig. 12(b)) and AuNi (Fig. 13(b)), while onlyGe has moved into the Ni layer for AuGeNi and nosignificant interaction of metallization and bulk InP canbe deduced (Fig. 1 (b)). For AuGeCr there is only asmall exchange between the Au layers as a function ofannealing temperature for annealing temperaturesabove 300 C (Figs. 12(b) - 12(h)), indicating that the

    E,

    Fig. 14. Schematics of the origin of different dominant transportmechanisms in alloyed ohmic contacts as deduced from Figs. l- 13.The resistances of thermoionic emmission and drift and diffusion arein series.

    Cr-P layer effectively acts as a diffusion barrier. ForAuGeNi (Figs. 1 (c) - 1 (g)), the Au signal smears outrelatively fast with no indications of the original place-ments of the Au layers.

    The separation of the Cr-P layer and bulk InP,leaving predominately Au and Ge and only a low

    concentration of the Cr-P phase adjacent to the inter-face, constitutes the greatest difference between theAuGeCr combination and the AuGeNi combination,where the Ni-(Ge)-P layers are in better contact withthe bulk InP. Since from the discussion of the ambienttemperature dependence of the MS contacts above, thetransport mechanisms of both AuGeCr and AuGeNiare not too different, the difference in interface concen-tration of the two combinations explains the differencein the values of the specific contact resistance for thetwo combinations quite satisfactorily, provided that themetal-phosphide phases constitute the low-resistance

    paths of current flow.

    5. Discussion

    From the results of the as-measured and ambienttemperature-dependent values of the specific contactresistance (Figs. l-lo), Auger electron analysis forAu-based metallizations on n-type InP (Figs. 11- 13)and the results from ref. 5, it is evident that low specificcontact resistance is associated with interfacial reactionsand thermodynamically stable phase formation between

    the metallization and InP. Especially, growth of metalphosphides adjacent to the InP surface is important forobtaining low values of r,. In contrast to the theoriesdeveloped in order to account for the inverse bulkdoping density dependence of r,, however, we havefound that the rate limiting steps in alloyed Au-basedmetallizations to InP are thermoionic emission, accord-ing to eqn. (2), across a small effective Schottky barrierand drift and diffusion, according to eqn. (3), across ametal phosphide/InP junction, forming the step junc-tion height that gives the inverse bulk doping densitydependence of r,. The Schottky barrier lowering processtowards zero effective barrier heights are shown in Fig.14, where it is also seen that the thermoionic emissionmechanism and drift and diffusion mechanism are effec-tively in series. The total resistance is dominated bywhich of them has the highest resistance, but since thedrift and diffusion part of the resistance is alwayspresent, this is the true rate limiting step for minimumspecific contact resistance, as also indicated in Fig. 14.

    Metal phosphides also form at the interface for theAu/Ni/Au metallization of the n-type, yet the specificcontact resistance for this combination is still at least anorder of magnitude higher than the specific contactresistance minimum for a similar metallization where

  • 8/9/2019 1-s2.0-004060909390012E-main.pdf

    12/13

    226 T. Clausen et al. / Transport properties of low-resistance ohmic contacts to InP

    the only difference is that the Au layer adjacent to theInP has been replaced by a AuGe layer. Of course, thesimplest answer to this problem concerns the dopingcapabilities of Ge in InP, postulating that a net surfacedoping of InP from Ge is responsible for the observeddifference in Y,. This is not in accordance with the

    measured temperature dependence of rc since the AuNiand AuGeNi combinations have a similar transitionfrom predominantly thermoionic emission across aneffective Schottky barrier to predominantly drift anddiffusion across a step junction between the metal phos-phides and bulk InP. Instead, Ge acts as an effectivebarrier towards In incorporation into the importantmetal-phosphide layer, since it is probable that bothNi-In and Ni-In-P phases form at the interface in theabsence of Ge [7, 321 even if Au, which is known as ahost for In incorporation [33], is present as is the casefor the Au/Ni/Au metallization. Thus, relatively small

    amounts of In altering the surface coverage of thephase dominating the electrical properties of the con-tact, significantly alter the value of r .

    For p-type InP, no direct transition from thermoionicemission to drift and diffusion is evident from Fig. 9 orFig. 10, respectively. Further improvements of the spe-cific contact resistance for ohmic contacts to p-type InPcan therefore be expected by increasing the surfacecoverage of the dominating barrier-lowering phase,since the small effective Schottky barrier would thenprobably disappear, similar to the case in Fig. 14, andthe onset of drift and diffusion would constitute the rate

    limiting step for the transport of carriers across themetal phosphide/InP junction. Assuming that the mo-bility of the carriers is the most important differencebetween the minimum obtainable specific contact resis-tance for identical doping densities of n- and p-typeInP, a minimum specific contact for p-type InP which isapproximately 20-30 times the minimum specific con-tact resistance for n-type is expected. By taking theminimum specific contact resistance for AuGeNi onn-type InP, 7 x 10e8 R cm2, at least a three-fold oreven a five-fold improvement can then be expected foran optimized ohmic contact metallization to p-type InP.

    Ni- and Cr-based metallizations to p-type InP do notlower the specific contact resistance compared withmetallizations without, but instead it is raised signifi-cantly. This is in accordance with the reverse situationon n-type InP, where Cr-P- and Ni-P-based intermetal-lit phases lower the effective Schottky barrier height. Forp-type InP, the effective Schottky barrier height, and,thus, the specific contact resistance, is raised duringstable Cr-P- and Ni-P-based phases. Similarly, highvalues of Y, have been reported using Au/Cr metalliza-tion to p-type InP [34], where transmission electronmicroscopy (TEM) showed that stable Cr-P layersformed adjacent to the InP surface. Both in ref. 34 and

    in our experiment the Cr layer was applied to the InPsurface first, forming an effective diffusion barrier to-wards interdiffusion of other elements in the metalliza-tion and InP, and only by annealing at relative highannealing temperatures or long annealing times, can thediffusion barrier be by-passed. An optimum metalliza-

    tion scheme for p-type InP using Au-based metalliza-tions thus cannot involve Cr adjacent to the InP surfaceor Ni, since these will eventually raise the effectiveSchottky barrier height. Cr can be incorporated into anoptimum metallization scheme for p-type InP if it is notplaced adjacent to the InP surface, since the resultingCr-P layer can act as a diffusion barrier placed wellaway from the important MS interface.

    6. Conclusion

    From the examination of the ambient temperaturedependence, bulk doping density dependence of theminimum specific contact resistance, and structuralexamination using Auger electron spectroscopy, thedominant transport mechanism and the dominant low-resistance intermetallic phase of alloyed Au-based ohmiccontact metallizations to n- and p-type InP has beenestablished. The dominant transport mechanism in low-resistance ohmic contacts is drift and diffusion across ametal phospide/InP junction, where the effective Schot-tky barrier between the metal-phosphide phase and InPhas disappeared. Thus, the rate limiting step in the

    ohmic contact is the junction formed between the Fermilevel and the conduction band minima and results in aninverse bulk doping density dependence for the mini-mum specific contact resistance, which has also beenaccounted for by the drift and diffusion theory, andobserved in our ohmic contacts to n-type InP. Ohmiccontacts to n-type InP therefore seems fully developedwith only small improvements ahead.

    Ohmic contacts to p-type InP do not exhibit drift anddiffusion and are not fully developed, and it is expectedthat the ohmic contact metallization to p-type InP canbe further optimized, leading to a further reduction ofthe specific contact resistance and the onset of drift anddiffusion as the only rate limiting step. Optimizationcannot be achieved using Cr adjacent to the p-type InPsurface, because immobile Cr-P phases increase theeffective Schottky barrier height. Similarly for Ni, Ni-Pphases increase the effective Schottky barrier height,but Ni seems to be more mobile than Cr (Figs. 11 and12), and Ni-P phases form adjacent to the InP surfaceirrespective of the original location of the Ni layer. Nishould therefore be avoided in any scheme of metalliza-tion to p-type InP, while Cr can be considered inmetallizations to p-type InP if it is placed well awayfrom the InP surface.

  • 8/9/2019 1-s2.0-004060909390012E-main.pdf

    13/13

    T. Clausen et al. / Transport properties of low-resistance ohmic contacts to InP 221

    References

    1 G. Y. Robinson, Schottky diodes and ohmic contacts, in Car1 W.Wilmsen (ed.), Physics and Chemistry of III& V Compound Semi-conductor Interfaces, Plenum Press, New York, 1985.

    2 L. J. Brillson, C. F. Brucker, A. D. Katnani, N. G. Stoffel and G.Margaritondo, Appl. Phys. Lett., 38 (10) (1981) 784-786.

    3 A. Appelbaum, M. Robbins and F. Schrey, IEEE, 34 (5) (1987)1026-1031.

    4 C. J. Sa and L. G. Meiners, Appl. Phys. Lett., 48 (26) (1986)1796-1798.

    5 2. Q. Shi, R. L. Wallace and W. A. Anderson, Appl. Phys. Lett.,59 (4) (1991) 446-448.

    6 R. K. Ball, Thin Solid Films, 176 (1989) 55-68.7 T. Clausen and 0. Leistiko, Microelectron. Eng., 18 (1992) 305-

    325.8 M. Murakami, Mater. Sci. Rep., 5 (6) (1990) 273-317.9 E. H. Rhoderick, Metal-semiconductor Conracts, Oxford Univer-

    sity Press, Oxford, 1978.10 N. Braslau, J. Vat. Sci. Technol., 19 (3) (1981) 803-807.11 T. C. Shen, G. B. Gao and H. Morkoc, J. Vat. Sci. Technol. B, IO

    (5) (1992) 2113-2133.12 W. Dingfen and K. Heime, Electron. Lert., 18(22) (1982) 940-

    941.13 W. Dingfen, W. Dening and K. Heime, Solid-State Electron., 29

    (5) (1986) 489-494.14 R. P. Gupta and W. S. Khokle, IEEE, 6 (6) (1985) 300-303.15 G. Brezeanu, C. Cabuz, D. Dascalu and P. A. Dan, Solid-State

    Electron., 30 (5) (1987)527-532.16 R. K. Kupka and W. Anderson, J. Appl. Phys., 69 (6) (1991)

    3623-3632.17 P. J. McNally, Solid-Sta te Electron ., 35 (12) (1992) 1705-1708.18 S. N. G. Chu, A. Katz, T. Boone, P. M. Thomas, V. G. Riggs,

    W. C. Dautremont-Smithand and W. D. Johnston, Jr., J. Appl.Phys., 67 (8) (1990) 3754-3760.

    19 I. Ohdomari and K. N. Tu, J. Appl. Phys., 51 (7) (1980) 373 5%3739.

    20 G. K. Reeves and H. B. Harrison, IEEE, 3 (5) (1982) 111-113.21 H. Morkoc, T. J. Drummond and C. M. Stanchak, IEEE, 28 (1)

    (1981) 1-5.

    22 R. H. Cox and H. Strack, Solid-State Electron., IO (1967) 1213-1218.

    23 E. Kuphal, Solid-Sta te Electron., 24 (1981) 69-78.24 K. P. Pande, E. Martin, D. Gutierrez and 0. Aina, Solid-State

    Electron., 30 (3) (1987) 253-258.25 J. A. Del Alamo and T. Mizutani, Solid-Sta te Electron., 31 (11)

    (1988) 1635-1639.26 A. Appelbaum, M. Robbins and F. Schrey, IEEE, 34 (5) (1987)

    1026-1031.

    31 R. J. Graham and J. W. Steeds, Insr. Phys. Conf: Ser., 67 (1983)507-512.

    27 H. H. Wehmann, S. Aytac and A. Schlachetzki, Solid-StateElectron., 26 (2) (1983) 149-153.

    28 K. Tabatabaie-Alavi, A. N. N. M. Choudhury, N. J. Slater and C.G. Fonstad, Appl. Phys. Lett., 40 (5) (1982) 398-400.

    29 E. Kaminska, A. Piotrowska, A. Barcz, J. Adamczewska and A.Turos, Solid-State Electron., 29 (3) (1986) 219-286.

    30 J. B. Boos and W. Kruppa, J. Vat. Sci . Technol. B, 7 (3) (1989)502-504.

    32 T. Sands, C. C. Chang, A. S. Kaplan, V. G. Keramidas, K. M.Krishnan and J. Washburn, Appl. Phys. Lett., 50 (19) (1987)1346-1348.

    33 N. S. Fatemi and V. G. Weizer, J. Appl. Phys., 67 (4) (1990)1934-1939.

    34 D. G. Ivey, R. Bruce and G. R. Piercy, Solid-Sta te Electron., 31(8) (1988) 1251-1258.