MULTISCALE SIMULATIONS OF THE RF DIODESPUTTERING OF COPPER

H. N. G. WADLEY, W. ZOU, X. W. ZHOU, J. F. GROVESIntelligent Processing of Materials Laboratory

Department of Materials science and EngineeringUniversity of Virginia

Charlottesville, VA 22903S. DESA, R. KOSUT, E. ABRAHAMSON, S. GHOSAL, A. KOZAK

SC SolutionsSanta Clara, CA 95054

D. X. WANGNonvolatile Electronics, Inc.

Eden Prairie, MN 55344-3617

ABSTRACT

The morphology and microstructure of RF diode sputter deposited materials is acomplicated function of many parameters of the reactor operating conditions. Using acombination of computational fluid dynamics (CFD), RF plasma, molecular dynamics (MD)sputter, and direct simulation Monte Carlo (DSMC) transport models, a multiscale approach hasbeen used to analyze the RF diode sputtering of copper. The CFD model predicts the velocityand pressure distribution of the working gas flows in the deposition chamber. The plasma modeluses these CFD results to compute ion energies and fluxes at the target and substrate. The MDmodel of sputtering is used to determine the initial energy distribution of sputtered atoms andreflected neutral working gas atoms and both of their angular distributions. A DSMC transportmodel then deduces the target atom deposition efficiency, the spatial distribution of the filmthickness, the target and reflected neutral atoms energy and impact angle distributions givenreactor operating input conditions such as background pressure, temperature, gas type, togetherwith the reactor geometry. These results can then be used in atomistic growth models to begin asystematic evaluation of surface morphology, nanoscale structure, and defects dependencesupon the reactor design and its operating conditions.

INTRODUCTION

Physical vapor deposition processes are being used to produce increasingly complex deviceswhose performance is critically dependent upon atomic scale features of their structure. A goodexample is giant magnetoresistive (GMR) metal multilayers (e.g., NiFe/Cu/NiFe) which exhibitlarge drops in their electrical resistance when a magnetic field is applied [1-3]. These materialscan be used for making new magnetic field sensors [4-6], read heads for disk drives, and magneticrandom access memories (MRAM) [5,6]. GMR-based MRAM has many potentially attractivefeatures, such as non-volatility, radiation hardness, low power consumption, high memory densi-ties (comparable to those of dynamic random access memory), and high access speed. The keyissue for this technology is to design an economical deposition process that can produce thermallystable GMR multilayers with high GMR ratios (defined as the maximum resistance changedivided by the resistance at magnetic saturation) at low magnetic field. Both theoretical and

323Mat. Res. Soc. Symp. Proc. Vol. 538 0 1999 Materials Research Society

experimental work indicated that the best GMR properties are achieved when the atomic scaleinterfacial roughness and interlayer chemical mixing are both minimized. Molecular dynamicssimulation of GMR multilayer deposition identified that adatom incident energies in the rangebetween 0.1 to 5.0 eV are needed to minimize both interfacial roughness and intermixing [7].Sputter deposition methods can result in incident adatom energies in the range of 0.1 to 20 eV, andhave been widely explored for GMR multilayer deposition [8-23]. However, because the experi-mental atomic scale characterization is difficult, and the dimensions of the processing parameterspace is large, the experimental search for an optimized sputter deposition process for synthesis ofGMR multilayers has been prolonged. A multiscale reactor simulation tool relating the morphol-ogy and microstructure of multilayers to the processing conditions of a diode sputter depositionsystem has been developed to help the optimization of the process.

DIODE SPUTTER DEPOSITION

1Atomistic simulations of vaporVacuum chamber at pressure P Working gas deposited films have shown that film

Water cooling (Ar or Xe) structure is a function of the key deposi-Ftion conditions including substrate tem-

Am pat ttered -perature, deposition rate, adatom incidentep a , a R F energy, and incident angle [7,24]. These

S .... power key deposition conditions are in turn con-Ar+ supply trolled by many process parameters. To

SPatered illustrate, a diode sputter deposition sys-I tem is schematically shown in Figure 1.

l An inert gas plasma is initiated and main-tained between target and substrate by anRF power. The inert gas ions created in theplasma are accelerated to the metal targetunder the bias voltage. The high energy

Vacuum pumps bombardment of inert gas ions on the tar-get surface results in the sputtering of

Figure 1. Schematic of A Diode Sputter System metal atoms from the target. These sput-tered atoms are then transported to the

substrate and deposited on the substrate surface. Generally, the system geometry and the RFpower determine the densities, the energies, and the angles of the atoms emitted from the target.Scattering with the lower energy working gas modifies these quantities during the transportationof the sputtered metal atoms to the substrate. The final distributions of these quantities depend onthe target-substrate distance, the working gas temperature, and pressure. Some inert gas ions areneutralized after their bombardment of the target and can be reflected toward the substrate. Thesereflected neutral particles can also modify the morphology and the structure of a growth film. Todevelop a better understanding of the process and to be able to predict the uniformity, morphol-ogy, and structure of thin films, it is essential to model the deposition efficiency (defined by theratio of the deposited material at the substrate to the sputtered materials at the target), the distribu-tions of density, energy, and angle of depositing atoms, as well as inert ions at the substrate, all asa function of reactor scale pressure, temperature, system geometry, power, etc.

324

EXPERIMENTS

Vapor deposited copper films were grown on silicon wafer using a Randex Model 2400-6Jdiode sputter deposition system at fixed Ar pressure and target-substrate distance but withdifferent plasma power. The surface morphology was characterized using a Pico SPM MS 300atomic force microscope. The morphology of two typical samples are shown in Figure 2.

Pressure = 20 mTorr, target-substrate distance = 1.5 in, Thickness = 2000 A

Power = 50 W Power = 350 W

Figure 2. Atomic Force Microscope of Sputter Deposited Copper Films

Figure 2 indicates that an increase of the power from 50 to 350 W dramatically increasesthe grain size and reduces the surface roughness.

SIMULATION METHODOLOGYModel Inputs Model model Outputs

The multiscale reactor model is illustrated in RamFigure 3. A CFD finite element model was used to r 350e WvsrM drama ly icalculate the velocity and pressure distribution ofinert gas flow in the chamber; a steady-state plasmamodel was used to simulate the density and energy Pa

of Ar ions striking the target and substrate; a ....molecular dynamics (MD) model was used todetermine the sputtering yield of target by the inertions, the energy and angular distribution of the W mmsputtered atoms at the target; and a DirectSimulation Monte Carlo (DSMC) model was usedto trace the change of density, energy, and angle ofthe sputtered atoms as they transported through the * MWlow-pressure inert gas to the deposition substrate.The individual models for gas flow, plasma Figure 3. Reactor Scale Integrated Modeldischarge, sputtering, and atom transport were then

325

integrated to create a detailed, steady-state, input-output model capable of predicting incidentenergy, incident angle, deposition-rate, and uniformity as a function of the process inputvariables: power, pressure, gas temperature, and electrode spacing. These results can in turnaccount for the morphology and microstructure of vapor deposited films [7,24].

RESULTS

The multiscale model was used to explore the effects of power during deposition of copperby argon ion sputtering at fixed pressure of 20 mTorr and fixed target-substrate distance of 1.5inches. The energy and current density for argon ions impacting the target as a function ofpower are given in Figures 4 and 5. The deposition rate as a function of power is shown inFigure 6. The energy and current density of depositing fluxes at the substrate as a function ofpower are drawn in Figures 7 and 8. It can be seen that increasing the power from 50 to 350 Wincreases the average ion energy from 0 to 600 eV, and the ion current density from 0 to 7x10 19

atoms/m 2. Higher ion energy at the target leads to higher energy sputtered atoms and a highersputtering yield [25]. As a result, increase of power not only increases the deposition rate, Figure6, but also increases the incident energy of the atoms deposited at the substrate [26], Figure 7.

600 1 . . . . . . . 12 . 17.50x 1019

Pressure=2OmTorr * Pressure=2OmTorr500 - Target-substrate N 10 Target-substrate - 6.25E o --agtsusrt - -62

distance=1.5 inches distance=1.5 inches "oSt 400 - 8 -- 5.00 E

21 E(D 300 -~ 6- 3.5 .

(C 3.7 St

C 'o200 -4 - - 2.50

100 .9 2 1.25

0 0' ' i i 0 i 0.000 50 100 150 200 250 300 350 400 0 50 1 100 1 0O 200 250 300 350 400

Input power (W) Input power (W)

Figure 4. Effect of Power on Inert Figure 5. Effect of Power on InertIon Energy at Target Ion Current Density at Target

It can be seen from Figure 6 that the deposition rate is almost linearly related to the power.The simulated results are in good agreement with experimental measurements. The resultsshown in Figure 7 indicate that during RF diode sputter deposition at a pressure of 20 mTorr anda working distance of 1.5 inches, the scattering with the background working gas almostcompletely thermalize the metal flux. A net increase of power of 350 W results in only a 0.04eV increase in the depositing energy of copper. This is unlikely to significantly change the thinfilm morphology and cannot account for the results shown in Figure 2. Interestingly, Figure 8indicates that the density of the reflected inert gas atoms at the substrate is comparable to thedensity of the depositing flux. On the other hand, Figure 7 reveals that increasing input powersignificantly increases the energy of the reflected inert flux to 100 eV range. Moleculardynamics simulations indicated that 100 eV inert atom bombardment on the substrate can causea significant transient local heating that results in surface atom athermal diffusion and a

326

flattening of a growth surface. It appears that the neutral flux is a powerful means for modifyingthe surface morphology. However, it may be undesirable for growing thin metal multilayersused for GMR devices because it induces interlayer mixing. The results obtained by theintegrated model allow us to use atomistic simulations [7,24] to quantify the microstructure andmorphology of deposited films as a function of reactor scale processing parameters.

E

ca

000.

0

40

30

2C

10

Pressure =2rn Torroo ' It-subsra• te

distance=t.5 nches

00 Simidati

E rpeiment0"'"

00 50 100 150 200 250 300 350

Input power (W)

Figure 6. Effect of Power on Deposition Rate

CONCLUSIONS

A multiscale reactor model has beendeveloped to simulate the effects of processingconditions (power, pressure, temperature, target-substrate distance, etc.) on the deposition rate, thedensity and the energy of depositing atoms, andreflected inert atoms. These results can be in turnused in atomistic growth models to simulatesurface morphology and microstructure ofdeposited films. Our results indicate that:

1. Increasing the RF power increases the inertgas ion density and energy at the target.

2. Increasing power linearly increases the metalflux density at the substrate and therefore the

St

CD

CC

E

.2

Cuatoms

0.12 --

0.08 -- 'L Ar ions

0.04 o Pressure=2OmTorr

Target-substratedistance= 1.5 inches

n nn400 0 50 100 150 200 250 300 350

Input power (W)

160

120

80

40

ot

E5

0

-0400

Figure 7. Effect of Power onFlux Energies at Substrate

CS0o

cc

U)

cc

1.6

1.2 Cu

0.8 - 9- Ar ions

0.4 Pressure=2OmTorr. o Target-substrate

distance= 1.5 inches0:

0 50 100 150 200 250 300 350 40

Input power (W))0

Figure 8. Effect of Power on FluxCurrent Densities at Substrate

deposition rate, but causes little change in metal incident energy at pressure of 20 mTorr.

3. The flux of reflected inert gas neutral flux is comparable to that of the depositing flux. Theneutral energy increases rapidly with power.

4. The energetic bombardment of the substrate by reflected neutral Ar is the apparent origin ofthe smoother surfaces observed after higher power deposition. However, this may lead tomultilayer intermixing of metal like those of interest for GMR devices.

327

Fd)(3 ....... NIR

ACKNOWLEDGEMENTS

We are grateful to the Defence Advanced Research Projects Agency (A. Tsao, ProgramManager) and the National Aeronautics and Space Administration for support of this workthrough NASA grants NAGW 1692 and NAG- 1-1964.

REFERENCES

1. P. M. Levy, J. Magn. Magn. Mater., 140-144, 485(1995).2. M. D. Stiles, Phys. Rev., B48, 7238(1993).3. W. H. Butler, X. G. Zhang, D. M. C. Nicholson and J. M. MacLaren, Phys. Rev., 52B,

13399(1995).4. J. Daughton, J. Brown, E. Chen, R. Beech, A. Pohm and W. Kude, IEEE Trans. Magn., 30,

4608(1994).5. C. Tsang, R. E. Fontana, T. Lin, D. E. Heim, V. S. Speriosu, B. A. Gurney and M. L. Will-

iams, IEEE Trans. Magn., 30, 3801(1994).6. J. L. Simonds, Phys. Today, April, 26(1995).7. X. W. Zhou and H. N. G. Wadley, J. Appl. Phys. 84, 2301(1998).8. E. E. Fullerton, D. M. Kelly, J. Guimpel and I. K. Schuller, Phys. Rev. Lett., 68, 859(1992).9. M. L. Yan, W. Y. Lai, Y. Z. Wang, S. X. Li and C. T. Yu, J. Appl. Phys., 77, 1816(1995).10. T. L. Hylton, K. R. Coffey, M. A. Parker and J. K. Howard, J. Appl. Phys., 75, 7058(1994).II. S. Honda, S. Ohmoto, R. Imada and M. Nawate, J. Magn. Magn. Mater., 126, 419(1993).12. R. J. Pollard, M. J. Wilson and P. J. Grundy, J. Magn. Magn. Mater., 146, L1(1995).13. H. Zhang, R. W. Cochrane, Y. Huai, M. Mao, X. Bian and W. B. Muir, J. Appl. Phys., 75,

6534(1994).14. T. R. McGuire, J. M. Harper, C. Cabral Jr. and T. S. Plaskett, J. Appl. Phys., 76, 6601(1994).15. J. D. Kim, A. K. Petford-Long, J. P. Jakubovics, J. E. Evetts and R. Somekh, J. Appl. Phys.,

76, 6513(1994).16. D. H. Mosca, F. Petroff, A. Fert, P. A. Schroeder, W. P. Pratt Jr. and R. Laloee, J. Magn.

Magn. Mater., 94, L1(1991).17. S. S. P. Parkin, Z. G. Li and D. J. Smith, Appl. Phys. Lett., 58, 2710(1991).18. H. Sato, Y. Kobayashi, Y. Aoki, R. Loloee and W. P. Pratt Jr., J. Magn. Magn. Mater., 140-

144, 567(1995).19. R. J. Highmore, W. C. Shih, R. E. Somekh and J. E. Evetts, J. Magn. Magn. Mater., 116,

249(1992).20. K. Kagawa, H. Kano, A. Okabe, A. Suzuki and K. Hayashi, J. Appl. Phys., 75, 6540(1994).21. K. Meguro, S. Hirano, M. Jimbo, S. Tsunashima and S. Uchiyama, J. Magn. Magn. Mater,

140-144, 601(1995).22. J. C. S. Kools, J. Appl. Phys., 77, 2993(1995).23. T. C. Anthony, J. A. Brug and S. Zhang, IEEE Trans. Magn., 30, 3819(1994).24. Y. Yang, R. A. Johnson and H. N. G. Wadley, Acta Mater., 45, 1455(1997).25. H. H. Andersen, H. L. Bay and H. E. Roosendaal, in Topics in Applied Physics, Sputtering by

Particle Bombardment I, edited by R. Behrisch (Springer-Verlag, Berlin Heidelberg, 1981),Vol. 47, p. 145; 219.

26. W. 0. Hofer, in Topics in Applied Physics, Sputtering by Particle Bombardment III, edited byR. Behrisch and K. Wittmaack (Springer-Verlag, Berlin Heidelberg, 1991), Vol. 64, p. 15.

328