Vacuum 65 (2002) 147154

Inuence of high power densities on the composition ofpulsed magnetron plasmas

A.P. Ehiasariana,*, R. Newa, W.-D. M .unza, L. Hultmanb, U. Helmerssonb,V. Kouznetsovc

aMaterials Research Institute, Sheffield Hallam University, Howard Street, Sheffield S1 1WB, UKbThin Film Physics Division, Department of Physics, Link .oping University, SE-581 83 Link .oping, Sweden

cChemfilt R&D AB, Kumla G (ardsv .agen 28, SE-145 63 Norsborg, Sweden

Received 15 October 2001; accepted 5 November 2001

Abstract

The application of high power pulses with peak voltage of 2 kV and peak power density of 3 kWcm2 to magnetronplasma sources is a new development in sputtering technology. The high power is applied to ordinary magnetron

cathodes in pulses with short duration of typically some tens of microseconds in order to avoid a glow-to-arc transition.

High plasma densities are obtained which have been predicted to initiate self-sputtering. This study concerns Cr and Ti

cathodes and presents evidence of multiply charged metal ions as well as of Ar ions in the dense plasma region of the

high power pulsed magnetron discharge and a substantially increased metal ion production compared to continuous

magnetron sputtering. The average degree of ionisation of the Cr metal deposition ux generated in the plasma source

was 30% at a distance of 50 cm. Deposition rates were maintained comparable to conventional magnetron sputtering

due to the low pressure of operation of the pulsed dischargeFtypically 0.4 Pa (3mTorr) of Ar pressure was used.Observations of the currentvoltage characteristics of the discharge conrmed two modes of operation of the plasma

source representing conventional pulsed sputtering at low powers (0.2 kWcm2) and pulsed self-sputtering at higher

powers (3 kWcm2). The optical emission from the various species in the plasma showed an increase in metal ion-to-

neutral ratio with increasing power. The time evolution within a pulse of the optical emission from Ar0, Cr0, Cr1+, and

Cr2+ showed that at low powers Cr and Ar excitation develops simultaneously. However, at higher powers a distinct

transition from Ar to Cr plasma within the duration of the pulse was observed. The time evolution of the discharge at

higher powers is discussed. r 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Pulsed magnetron sputtering; Time evolution; Ionised sputtering; Highly charged metal ions; High density plasma; Ionised

metal plasma

1. Introduction

Magnetron sputtering is a well-establishedphysical vapour deposition technique. It is used

to deposit metallic as well as compound thin lmsfor a wide range of applications. In most cases, it isbenecial to produce dense, defect-free coatings,which often requires low-energy ion bombardmentof the condensing lm surface to increase themobility of adatoms [13] to achieve void free thinlms. To obtain considerable ion bombardment a

*Corresponding author. Fax: +44-114-221-3053.

E-mail address: a.ehiasarian@shu.ac.uk (A.P. Ehiasarian).

0042-207X/02/$ - see front matter r 2002 Elsevier Science Ltd. All rights reserved.

PII: S 0 0 4 2 - 2 0 7 X ( 0 1 ) 0 0 4 7 5 - 4

sufcient ion-to-neutral ratio in the condensingux needs to be achieved. Conventional magne-tron plasma sources produce mainly sputteringgas ions [4] reaching a typical charge carrierdensity of 109 cm3. Arc evaporators easilyreach densities of 1018 cm3 producing highlyionised metal ion uxes, however, their maindisadvantage is that they produce a dropletof the target material, which deteriorates thequality of deposited coatings. Several attemptshave been made to enhance the plasma densityin the vicinity of the substrates using sputteringtechniques. An effective method, suggested byWindow et al. [5], was to unbalance the magneticeld, thus guiding the Ar plasma far awayfrom the magnetron. In the work of Rossnageland Hopwood [6], an additional ionisationsource comprising an inductively coupled radiofrequency (RF) antenna was introduced in themagnetron discharge producing increased plasmadensity as well as ionisation of the sputtered metalux. Recently, Kouznetsov et al. [7] reported amodied operational mode of the conventionalmagnetron utilising high power density pulsedmagnetron sputtering (HIPIMS). Due to a highplasma density, they reported that the metal ion-to-neutral ratio in the deposition ux reached asmuch as 70% in the case of Cu.The present paper reports on the transition

between conventional pulsed magnetron sputter-ing to HIPIMS as the power dissipated in thedischarge is increased. Cr and Ti cathodes wereused as model systems. The discharge parametersare compared to conventional magnetron sputter-ing. High peak plasma densities of the order of1013 cm3 containing multiply ionised metal spe-cies as observed by optical emission spectroscopy(OES). The effect of power on the temporalevolution of the plasma composition is discussed.

2. Experimental

The experiments concentrated on characterisingthe pulsed magnetron plasma, by measuringcurrentvoltage characteristics, carrying out opti-cal emission and probe diagnostics and determin-ing the metal ion-to-neutral ratio in the deposition

ux using a biasing/deposition technique. Thechamber, power supply characteristics and dis-charge current and voltage measurements were asdescribed previously [8]; however, a target with adiameter of 50mm was used.The plasma composition was analysed by OES

utilising a Jobin Yvonne Triax 320 scanningmonochromator with resolution 0.12 nm and agrating groove density of 1200 grooves/mm. Timeevolution of the OES signals was studied using thesame monochromator, with the output of itsphoto-multiplier tube (PMT) connected to groundvia a 100 kO resistor. The voltage across theresistor was proportional to the current signal inthe PMT and was monitored using a Tektronix520 digitising oscilloscope with a 10MO probe,triggering on the target voltage signal. In order toobtain the temporal evolution of a certain emis-sion line, the monochromator was set to transmitthe wavelength of interest and the voltage acrossthe resistor was recorded with time. The opticalemission was collected in vacuo by an optical brebundle equipped with a collimating tube(f 4mm, l 50mm) positioned at a distanceof 1 cm parallel to the target surface. The completeoptical set-up was sensitive to emission in thewavelength range 200950 nm. Macak et al. [8]were rst to report a study of the plasmacomposition and temporal evolution of the HI-PIMS using OES. Similar measurements presentedin this work go beyond previous investigations interms of an enhanced spectral range, covering theultraviolet region, an improved resolution redu-cing the signal-to-noise ratio and a vastly superiorsensitivity of PMT over charge coupled devicedetectors.Ion saturation current measurements were

performed with a at probe with a diameter of2 cm equipped with a guard ring and positioned3.5 cm from the target surface. The deposition ratewas measured using a quartz crystal microbalancesituated at a distance of 6.5 cm from the targetsurface. The metal ion-to-neutral ratio in thedeposition ux was calculated by depositing for aknown time on a positively biased and a oatingmetal plate located 50 cm from the target. Thedifference in mass between the plates was mea-sured and the metal ion and neutral ux were

A.P. Ehiasarian et al. / Vacuum 65 (2002) 147154148

determined for a medium peak power density of1.5 kWcm2.

3. Results and discussion

3.1. IU characteristic at p=const

It is well documented that conventional magne-tron I2U characteristics follow a power law I kUn . The exponent n is typically in the range 515[9] reaching low values when the dischargeoperates at very low pressures or is conned by aweak magnetic eld. Fig. 1 shows the I2Ucharacteristics of a magnetron discharge operatedon a typical industrially sized rectangular cathodeand the high power density pulsed magnetrondischarge of the present work operated on acylindrical cathode, 50mm diameter. Both dis-charges were operated at a pressure of 0.4 Pa(3mTorr). The exponent calculated from the slopeof the relation for the conventional magnetron is 8,which is well within the dened range. The pulseddischarge exhibits two slopes depending on thetarget current. At low currents (o600mAcm2),

the exponent is approximately 7 indicating normalmagnetron operation. At higher currents, theexponent changes to 1, in a mode of operation inwhich the increase of discharge voltage to veryhigh values of 1.6 kV is not accompanied by alarge increase in the discharge current. Onepossible explanation of this mode of operation isthat the secondary electrons accelerated in thesheath to 1.6 keV cannot be trapped by themagnetic eld and the probability of ionisingcollisions per electron is decreased. Nevertheless,as shown in Fig. 2, at the higher powers the plasmadensity at a distance of 3.5 cm from the targetincreases faster than at low powers. The peakdensity calculated from the ion saturation currentand assuming low ion energy was estimated to be1013 cm3 at a target power density of 3 kWcm2.The reasons for the steeper increase of saturationcurrent at higher powers are not fully understood.It is possible to attribute it to the escapeof the plasma from the target due to poor electricand magnetic eld connement at high targetvoltages.

Fig. 1. I2U characteristics of the pulsed discharge. The

exponent n of the power law I kUn is indicated. Cr sputteringin Ar atmosphere at a pressure of 0.4 Pa (3mTorr).

Fig. 2. Ion saturation current at a distance of 3.5 cm with the

probe facing the target (circles) and with the probe perpendi-

cular to the target surface (squares). The at probe was biased

to 150V. Maximum plasma densities derived from the currentdensities are in the range 1013 cm3.

A.P. Ehiasarian et al. / Vacuum 65 (2002) 147154 149

3.2. Detection and study of metal ions in the plasma

at high target currents

It is well known that the probability ofionisation of species depends on their dwell timein the plasma and the energy of ionising species. Intypical magnetrons, the plasma comprises mainlyresidual gas ions [4]. Electrons conned in an ExBdrift loop in the magnetron collide with the almoststationary gas atoms and the ionisation probabil-ity per electron is quite high. On the other hand,sputtered metal atoms leave the target surface withrelatively high energies, typically a few eV,traversing the dense plasma region in a fewmicroseconds. In typical magnetron discharges,the plasma density is such that the traverse time istoo short for ionising collisions to occur and only anegligible proportion of metal ions is created. Inthe high power pulsed discharge, although thesputtering voltage can be at least two times higher,the sputtered atoms leave the target with the sameenergy of a few eV. However, they cannotpenetrate the dense highly energetic electron cloudformed in the sheath due to the high dischargepower without collisions and the probability ofionisation of the pulse-sputtered ux is high. Themaximum tangential magnetic eld of the magne-tron (B500G) is not sufcient to magnetise the

ionised metal and it readily escapes from the denseplasma region and reaches long distances wheredeposition samples can be placed. On the otherhand, a portion of the metal ions is acceleratedback to the target surface, giving rise to self-sputtering and decreasing the sputtering rate dueto the low self-sputtering yield. Preliminary resultsshow a clear trend between self-sputtering yieldand deposition rate [10]. In the present work, thedeposition ux at a distance of 50 cm from a Crtarget surface was found to contain 30% metalions for peak target power of 1.5 kWcm2.OES studies conrm an enhanced proportion of

ions of both sputtered species and residual gas inthe HIPIMS discharge. The plasma compositionduring sputtering of Cr and Ti was investigatedusing OES. The spectrum during Cr sputtering at apeak power of 3 kWcm2 is shown in Fig. 3. One-and two-fold ionised metal species were detected inthe dense plasma region of the magnetron.Up to two-fold ionised metal ions were found

also in a Ti pulsed discharge. The optical emissionspectra in Fig. 4 show three kinds of magnetronsputtering discharges. Pulsed magnetron sputter-ing plasma was realised with peak power1.5 kWcm2 and average power 300W. A con-ventional magnetron was operated at an averagepower of 300W. A magnetron with enhanced

Fig. 3. Optical emission spectrum of high power pulsed sputtering of Cr in an argon atmosphere. Peak target power=3000Wcm2,

PAr=0.4 Pa (3mTorr), 1 cm from target. Two-fold and one-fold ionised Cr was observed in the dense plasma region of the magnetron.

A.P. Ehiasarian et al. / Vacuum 65 (2002) 147154150

ionisation by RF coil.1 The Ti ion to neutralspectral line ratio is negligible in the conventionalmagnetron sputtering, and increases from 0.9 inthe RF coil enhanced discharge to 5 in the pulseddischarge. The numbers quoted above are re-garded only as a qualitative comparison of thedegree of ionisation in the three methods ofsputtering.The above evidence for signicant ionisation is

further supported by the dense, smooth lmsgenerated in subsequent coating runs which willbe described in detail elsewhere [12]. The lms hadan excellent adhesion due to the pre-treatment ofthe substrates by metal ion containing HIPIMSplasma bombardment under a bias of 1200V.

3.3. Plasma density and composition as a function

of peak target current

As can be seen from Fig. 2, the peak density ofthe magnetron plasma increased monotonically byapproximately two orders of magnitude as thepeak target current increased by a factor of 30. It isinteresting to note that when high powers aredissipated in the discharge, the discharge voltageincreases simultaneously with the current with an

exponent close to 1 (see Fig. 1). The increaseddischarge voltage is probably responsible forincreasing, for example, the hot electron tempera-ture, which, combined with the increased plasmadensity, is a pre-requisite for an improved ionisa-tion efciency of the sputtered Cr atoms.OES measurements of the Cr2+ line at

232.03 nm, Cr1+ line at 205.55 nm and the Cr0

line at 399.11 nm were carried out in order toinvestigate the degree of ionisation of the sputteredmetal ux. These particular emission lines havebeen chosen because they are formed by transi-tions for which the upper level energies are similar:5.4, 6.0 and 5.66 eV for Cr2+, Cr1+, and Cr0,respectively. Assuming a Corona model for theplasma [13], the line emission intensity ratio can beused as a qualitative measure of the density ratioof the species because of the similarity in theirexcitation energies. It should be noted that, sinceonly Cr neutrals are sputtered from the target allCr ions are created in the discharge by a process ofelectron impact ionisation, which depends stronglyon the electron temperature. The intensity ratio ofsuch optical emission lines depends directly on thedensity of the species and is inuenced indirectly

Fig. 4. Optical emission spectra of plasmas produced by

discharges in a conventional DC magnetron (bottom), a RF

coil enhanced magnetron (middle) and high power pulsed

magnetron.

Fig. 5. Ion-to-neutral ratio vs. peak target current for Cr form

optical emission line intensities I(Cr2+)/I(Cr0) (triangles) and

I(Cr1+)/I(Cr0) (circles). Also shown is the ratio of the ion

current density and deposition rate (squares).

1The discharge was operated with DC magnetron power of

300W, RF coil power of 450W, and PAr=4Pa (30mTorr).

The experimental details are described in Ref. [11].

A.P. Ehiasarian et al. / Vacuum 65 (2002) 147154 151

by changes in electron temperature. The ratio ofthe spectral line intensities of Cr2+ and Cr0

indicated in Fig. 5 with triangles shows anincreasing ionisation with target current. Twoother diagnostic results that indicate an increasedion-to-neutral ratio are also shown in Fig. 5,namely the spectral line intensity ratio I (Cr1+)/I(Cr0) recorded during early conditioning runs.The ratio of saturation current to deposition rate isalso given as a qualitative measure of the ion-to-neutral arrival rate.

3.4. Time evolution of the plasma composition

The temporal evolution of the plasma was foundto be strongly inuenced by the power of thedischarge pulse. Fig. 6a shows a typical temporalevolution of a pulsed magnetron discharge atrelatively low power. The target current and theoptical emission from Ar and Cr neutrals occuralmost simultaneously. This kind of relationship isroutinely observed in the widely used pulsedmagnetron discharges operating at lower powers

Fig. 6. Typical temporal evolution of the plasma. (a) OES and target current at low power (peak target voltage=500V), (b) OES andion saturation current Js at high power, (c) target voltage and current at high power (the apparent uctuations in the current signal

after 80ms are principally caused by digitisation noise). The species shown are: Cr0 (399 nm), Cr1+ (232 nm), Ar0 (811 nm), Ar1+

(440 nm).

A.P. Ehiasarian et al. / Vacuum 65 (2002) 147154152

than presented here [14]. However, when the peakpower density is increased to a few kWcm2 thetemporal evolution of the discharge plasma isquite different as is illustrated in Fig. 6b. The OESmaxima appear in a clearly differentiated sequencebeginning with Ar neutral and ending in Crneutral. Such a separation in time between theoptical emission of Ar and Ti neutral lines at highpowers was rst reported by Macak et al. [8]. Theplasma model proposed by them assumed that theplasma developed from being Ar dominated tometal ion dominated due to the gas rarefactioneffects discussed by Rossnagel et al. [15]. Theseparation for high powers is also conrmed for Crin the present work.In detail, the evolution of the optical emission

signals shown in Fig. 6b can be separated intothree stages as follows. A high power and highplasma density discharge is produced during therst stage, from 0 to 40 ms. The rst emission isobserved from Ar atoms, present in the back-ground gas when vacuum breakdown occurs. AsAr is ionised the plasma density and target currentincrease, and Ar1+ emission is developed. With adelay of a few microseconds after the Ar lines, theCr0 emission lines are detected. After a further 3 msemission from Cr ion lines is detected, indicatingthat the sputtered Cr ux is being ionised in thehigh density plasma. The Cr1+ lines peak atapproximately 30 ms, approximately 10 ms afterthe ion saturation current Js has reached itsmaximum value. The rapid fall in Ar emission atthis time may indicate that the Ar pressure is beinglocally reduced by the momentum transfer fromsputtered Cr atoms [15,9]. The Cr ion emissionsfall over the next 40 ms.At 40 ms the discharge voltage and current have

fallen to levels which indicate the start of a secondstage. At this time the voltage is approximately750V and the discharge is leaving the high powerregime as shown in Fig. 1 and entering a regime ofnormal magnetron operation with a high targetcurrent. During this phase the Cr0 emission isstronger and the Cr1+ emission is weaker than inthe high power stage, and the target current shownin Fig. 6c starts to drop quickly for a smalldecrease in voltage as is typical of a normalmagnetron discharge. The magnetic connement

eld in these circumstances is very efcient andworks in favour of decreasing the rate of decay ofthe plasma density peak at 20 ms.The nal stage of the discharge occurs at about

80 ms when a typical magnetron glow discharge isinitiated with voltage of 500V and currentdensity some tens of mA cm2. The subsequentemission from Ar neutral probably indicates thereturn of the discharge voltage and current totypical magnetron levels and the restoration of theAr pressure, respectively.

4. Conclusions

High power pulsed magnetron sputtering withpeak powers of 3 kWcm2 were used to producemagnetron glow discharges with very high plasmadensities. By optical emission, doubly and singlycharged metal ions were detected in the denseplasma region for both Cr and Ti targets. Thedegree of ionisation of the sputtered ux that wasmeasured for the Cr reached a value of 30%. Theplasma conditions were strongly inuenced by thepeak power applied to the target as seen fromI2U characteristic and temporal evolution of thecomposition of the discharge and a thresholdcurrent when the discharge switched from typicalmagnetron sputtering to ionised sputtering wasobserved.

Acknowledgements

The work presented in this paper was carriedout with the nancial support of EPSRC GrantRef: GR/R32420/01, The Swedish Foundation forStrategic Research Program on Low TemperatureThin Film Synthesis and The Swedish ResearchCouncil.

References

[1] Petrov I, Hultman L, Sundgren JE, Greene JE. J Vac Sci

Technol A 1992;10(2):26572.

[2] Hultman L, Helmersson U, Barnett SA, Sundgren JE,

Greene JE. J Appl Phys 1987;61(2):5525.

A.P. Ehiasarian et al. / Vacuum 65 (2002) 147154 153

[3] Helmersson U, Hentzell HTG, Hultman L, Hibbs MK,

Sundgren JE. AIP Conf Proc 1986;149:7994.

[4] Petrov I, Myers A, Greene JE, Abelson JR. J Vac Sci

Technol A 1994;12(5):284654.

[5] Window B, Sharples F, Savvides N. J Vac Sci Technol A

1985;3(6):236872.

[6] Rossnagel SM, Hopwood J. J Vac Sci Technol 1994;B

12(1):44953.

[7] Kouznetsov V, Macak K, Schneider JM, Helmersson U,

Petrov I. Surf Coat Technol 1999;122(23):2903.

[8] Macak K, Kouznetsov V, Schneider JM, Helmersson U,

Petrov I, J Vac Sci Technol A (Parts 1 and 2)

2000;18(40):15337.

[9] Rossnagel SM, Kaufman HR. J Vac Sci Technol A

1988;6(2):2239.

[10] Alami J, Music D, Helmersson U, unpublished results.

[11] Nouvellon C, Konstantinidis S, Dauchot JP, Wautelet M,

Jouan PY, Ricard A, Hecq M. J Appl Phys, in press.

[12] Ehiasarian AP, et al. manuscript in preparation.

[13] McWhirter RWP. In: Huddlestone RG, Leonard SL,

editors. Plasma diagnostic techniques. New York: Aca-

demic Press, 1965. p. 208.

[14] List M, Krausse U, W .unsche T. Paper presented at

the 43rd Annual Technical Conference of the

Society of Vacuum Coaters, April 2000, Philadelphia,

USA.

[15] Rossnagel SM. J Vac Sci Technol A (Part 2) 1988;

6(3):18216.

A.P. Ehiasarian et al. / Vacuum 65 (2002) 147154154

Influence of high power densities on the composition of pulsed magnetron plasmasIntroductionExperimentalResults and discussionI-U characteristic at p=constDetection and study of metal ions in the plasma at high target currentsPlasma density and composition as a function of peak target currentTime evolution of the plasma composition

ConclusionsAcknowledgementsReferences