ENERGETIC INERT GAS ATOM IMPACT EFFECTSDURING ION BEAM MULTILAYER DEPOSITION

X. W. ZHOU, W. ZOU, H. N. G. WADLEYDepartment of Materials Science and EngineeringUniversity of Virginia, Charlottesville, VA 22903

ABSTRACT

New magnetic field sensors and non-volatile magnetic random access memories can bebuilt from giant magnetoresistive multilayers with nanoscale thickness. The performance of thesedevices is enhanced by decreasing the atomic scale interfacial roughness and interlayer mixing ofthe multilayers. During ion beam sputtering, inert gas neutrals with energies between 50 and 200eV impact the growth surface. A molecular dynamics method has been used to study the effectsof these impacts on the surface roughness and interlayer mixing of model nickel/coppermultilayers. The results indicate that impacts with energy above 50 eV cause mixing due to theexchange of Ni atoms with Cu atoms in an underlying Cu crystal. The extent of the mixingincreases with impact energy, but decreases as the number of the Ni monolayers above the Cucrystal increases. While Xe and Ar impacts have a similar mixing effect at low energies, heavierXe ions/neutrals induce more significant mixing at high energies.

INTRODUCTION

Many nanostructured materials exhibit technologically interesting properties that are notpossessed by their bulk constituents. For instance, if a -20A thick conductive layer is sandwichedbetween -10-100 A thick ferromagnetic layers, the metallic multilayer stack can exhibit giantmagnetoresistance (GMR) manifested as a large (5-50%) change in electrical resistance upon theapplication of an external magnetic field [1]. GMR materials have been used to build the readheads for hard drives that can greatly increase the storage capacity of computers [2]. They are alsobeing investigated for a new class of low cost, nonvolatile magnetic random access memories [2].These applications require materials with a high magnetoresistance. This can be achieved if themultilayers have flat interfaces with minimal chemical mixing between the layers [3]. Experimen-tal data indicate that deposition at low temperatures is required to minimize thermally activatedinterlayer diffusion [4]. A recent molecular dynamics (MD) simulation of the deposition of themodel Ni/Cu/Ni multilayers [5,6] showed that a moderate (1 eV) adatom energy can be used toflatten the interfaces of the multilayers grown under low temperature conditions without causingextensive interfacial mixing. This may account for recent observations that hyperthermal metalatoms created by sputtering processes such as RF diode (or magnetron) sputtering and ion beamsputtering produce better GMR multilayers than their thermal energy (e.g., those used in molecu-lar beam epitaxy) counterparts [7]. MD simulation also revealed that very high incident energiescould promote interlayer mixing by an atomic exchange mechanism at the Ni on Cu interface[5,6]. This is consistent with other observations that magnetron sputtering gives rise to the bestGMR multilayers under intermediate energy deposition conditions [8]. These observations haveled to a proposal that further reductions of both interfacial roughness and interlayer mixing couldbe obtained by a modulated adatom energy deposition, in which the first half of a new layer wasdeposited with a thermalized flux and the remainder with a hyperthermal flux [5,6].

103

Mat. Res. Soc. Symp. Proc. Vol. 616 © 2000 Materials Research Society

However, other fluxes are incident upon a growth surface during metal sputtering deposi-tions. For ion beam sputter deposition, energetic ions generated in the ion beam gun are used forthe target sputtering. These ions resulted in a reflected neutral flux that can reach the substratewith energies in excess of 100 eV [9]. This may potentially cause significant atomic exchange andrelated mixing, especially on the Ni on Cu surface [10]. To improve growth of GMR materials, abetter understanding of the effects of these high energy particle impacts needs to be developed.Here, a molecular dynamics simulation method is used to investigate the effects of high energyxenon and argon neutral impacts on model (111) Ni on Cu surfaces.

COMPUTATIONAL METHODS

Since the embedded atom method (EAM) potential for the Cu-Ni system [11] captures thelocal environment dependence of the potential and realistically describes the energetics near sur-faces and interfaces, it was used to define the interactions between Cu-Cu atoms, Ni-Ni atoms,and Cu-Ni atoms. A universal potential that was fitted to a significant amount of surface bom-bardment experimental data [12] was employed to calculate the interactions between the xenon/argon particles and the surface metal atoms. Cu crystals containing 84(112) planes in the x direc-tion, 11 (111) planes in the y direction, and 48 (110) planes in the z direction were created byassigning the atoms to the equilibrium lattice sites (see Figure 1). 1 to 7 Ni (111) planes were thenplaced on top of the surface based upon an epitaxial lattice extension. Inert particles with a givenkinetic energy were injected from random locations to impact the top (y) surface at a normalangle. Periodic boundary conditions were applied in the x and z directions to extend the crystaldimensions. The bottom two Cu planes were fixed at their equilibrium positions to prevent theshift of the crystal. A thermostat algorithm [13] was used to maintain a fixed substrate tempera-ture of 300 K. The surface evolution during impact were then simulated by tracking the positionsof atoms and inert neutrals using Newton's equations of motion.

RESULTS

Impacts on the surface composed of one Ni monolayer above a Cu crystal were first simu-lated at different impact energies between 50 and 200 eV. Representative atomic surfaces after 10xenon impacts at xenon energies of 50 eV and 200 eV are shown in Figures 1 (a) and 1 (b) respec-tively. In Figure 1, dark and light spheres are used to distinguish Ni and Cu atoms. Figure I indi-cates that xenon impacts caused atomic exchanges between Ni atoms in the perfect Ni layer andunderlying Cu atoms even at the low energy of 50 eV. The degree of this impact induced mixingincreased with the impact energy. In addition, the impacts caused pits and surface roughness byejecting in-plane atoms onto the top of the surface. The extent of this type of surface roughnessalso increased with the impact energy.

The effect of impact induced mixing should depend on the distance between the surface andthe interface. Impacts on the Ni on Cu surfaces that contained more Ni monolayers were henceexplored. Figure 2 shows the atomic structure of a surface containing 7 Ni monolayers above a Cucrystal after 10 Xe impacts at an ion energy of 200 eV. To examine mixing at the Ni-Cu interface,some of the surface layers are displaced in the vertical (y) direction. It can be seen that no mixingoccurred during ion impact when the Ni on Cu interface was 7 Ni monolayers below the surface.

104

Y[111] )ArorXe# ( Normal incidence

Ni

CL

[112]ore impacts

100 to oo'**,*0# to to00 #0 to #

00 tot #00 W#toto 0,0oft *0 # V* to 0# V*

00 Vto V# it'00 Wo # v 0000 0i

to V# #004 0000 to 0# Or*to to W#0# _##__#

Fig. 1. The structure of the Ni monolayer above a Cu crystalenergy of (a) 50 eV and (b) 200 eV.

after 10 Xe ion impacts at an ion

To more clearly explore the impact effect trend, a mixing parameter was defined as the aver-age number of exchanged Cu atoms per impact. The mixing parameter obtained at differentimpact energies and impact species is plotted in Figure 3 as a function of the number of Ni mono-layers above the Cu crystal. Clearly, the mixing parameter rapidly decreased as the number of Nimonolayers increased. Otherwise, the mixing parameter increased with the impact energy. At thehigh energy of 200 eV, Xe impacts generally caused more significant mixing than Ar impacts.However, this difference became much smaller at 100 eV, and diminished at 50 eV.

105

(a) Exe = 50 eV

LýA

J

Exe = 200 eV

y [111] Aror XeNormal incidence

A1 [X [11r]/z [ 110 ] Before impacts

Fig. 2. The structure of 7 Ni monolayer/Cu crystal after 10 Xe ion impacts at an ionenergy of 200 eV.

To quantify the impact induced surface pitting and roughening, the number of atoms ejectedonto the top of the surface per impact was also calculated at different impact energies and impactspecies. The results of this calculation are plotted in Figure 4 as a function of the number of Nimonolayers above the Cu crystal. It can be seen from Figure 4 that the pitting and the rougheninginduced by impacts are insensitive to the number of the Ni monolayers. While no big difference isobserved for the Ar and Xe impacts, increases in the impact energy increased the roughness of anoriginally flat Ni surface. At high deposition temperatures or low deposition rates, the density ofthese pits and roughness regions may be reduced due to thermally activated diffusion.

106

5

4

3

2

1

01 2 3 4 5 6

Number of Ni monolayers above Cu

Fig. 3. Mixing as a function of Ni monolayers

20

161

12

8

4

01 2 3 4 5 6

Number of Ni monolayers above Cu

Fig. 4. Surface roughness as a function of Ni monol

C.2_0n

E0

X

E

Cu

C.)

Z0

x

..0Ez=

layer of Co between the Nis1Fejg/Cu interface can significantly improve the GMR properties ofthe Nis1FeI9/Cu/Ni81Fe,9 multilayers [141. This may be helpful to growth because mixing betweenthe Co and Nis1Fej9 alloy is not expected to affect the magnetic properties of the layer, while theimpact induced mixing between the thermodynamically immiscible Co and Cu can be subse-quently reduced by annealing. Nevertheless, the recognition of this high energy impact effectemphasize the need to reduce the high energy particle flux during a further exploration of growthmethods for GMR technology. A number of ways can be used to do this. For the ion beam sputterdeposition process, this may involve refining the position of the substrate to avoid the high fluxreflected neutral zone, or switching the working gas from Ar to Xe because the density and energyof reflected Xe neutrals are both much lower than those of reflected Ar [9].

107

DISCUSSION

pis Nanostructured materialsXe - such as GMR multilayers con-Ar+ tain different material layers,Xe+Ar+ each with very thin thicknessle+ (in the scale of several atomicr layers). These materials are

often required to have very lowinterfacial roughness and inter-layer chemical mixing. Becauselow temperature must be usedto deposit these nonequilibrium

- - materials and prevent thermal7 diffusion induced mixing, flat

surfaces (and interfaces) can-not be achieved under thermalenergy conditions. Hyperther-mal energy sputtering deposi-

J tion methods such as RF diodeor ion beam sputtering deposi-tion techniques need to be used

8+÷ to promote energetic impactr+ induced surfaces flattening.

However, inert gas particleswith energies between 50 and200 eV are likely to impact thesubstrate during either the RFdiode or the ion beam sputterdeposition processes. The cal-culations here indicated that

7 these impacts can cause signifi-cant interlayer mixing in themultilayers.

layers. Experiments have shownthat the insertion of 1-2 mono-

.2

0

0)

ca

SI I I

Symbolso 200 eV X* 200 eV A'5100 eV X

100 eVA50 eV Xe

* 50 eV Arl

0ý

CONCLUSIONS

Molecular dynamics simulations of high energy inert gas impacts with Ni on Cu surfaceindicated that:

* 50-200 eV impacts caused significant mixing between surface Ni and the underlying Cu.The degree of the mixing increased with the impact energy. At high impact energies, Xeimpacts caused more mixing than Ar impacts, but the difference became smaller at lowimpact energies.

0 The degree of mixing rapidly decreased as the number of Ni monolayers on top of the Cucrystal increased. It is hence important to reduce the probability of high energy impacts atthe substrate surface during deposition of the first few monolayers of Ni on to a predepos-ited Cu surface.

* High energy impacts also induced pits and roughness on an originally flat surface. Theextent of the roughness increased with the impact energy, but was insensitive to impactingspecies and the number of the Ni monolayers above Cu.

ACKNOWLEDGEMENTS

We are grateful to the Defense Advanced Research Projects Agency (A. Tsao and S. Wolf,Program Managers) and the National Aeronautics and Space Administration for support of thiswork through NASA grants Nos. NAGW 1692 and NAG- 1-1964.

REFERENCES

1. S. Honda, S. Ohmoto, R. Imada and M. Nawate, J. Magn. Magn. Mater 126, 419 (1993).2. G. A. Prinz, Science 282, 1660 (1998).3. W. H. Butler (private communication).4. H. Kano, K. Kagawa, A. Suzuki, A. Okabe, K. Hayashi and K. Aso, Appl. Phys. Lett. 63,

2839 (1993).5. X. W. Zhou and H. N. G. Wadley, J. Appl. Phys. 84, 2301 (1998).6. X. W. Zhou and H. N. G. Wadley, J. Appl. Phys. 87, 553 (2000).7. X. W. Zhou and H. N. G. Wadley, in preparation.8. J. C. S. Kools, J. Appl. Phys. 77, 2993 (1995).9. X. W. Zhou, S. Sainathan, H. N. G. Wadley, in preparation.10. X. W. Zhou and H. N. G. Wadley, Mat. Res. Soc. Symp. Proc., (1999), in press11. S. M. Foiles, Phys. Rev. B 32, 7685 (1985).12. R. E. Johnson, in Energetic Charged-Particle Interactions with Atmospheres and Surfaces,

Physics and Chemistry in Space Planetology, Vol. 19, ed. L. J. Lanzerotti, M. Hill and D.Stoffler, Springer-Verlag, Berlin, p. 68, 1990.

13. X. W. Zhou, R. A. Johnson and H. N. G. Wadley, Acta Mater. 45, 1513 (1997).14. S. S. P. Parkin, Appl. Phys. Lett. 61, 1358 (1992).

108