1: \documentstyle[aps,floats,graphicx]{revtex}
2:
3: \begin{document}
4:
5: \newcommand{\bec}{\begin{center}}
6: \newcommand{\ec}{\end{center}}
7: \newcommand{\be}{\begin{equation}}
8: \newcommand{\ee}{\end{equation}}
9: \newcommand{\beqn}{\begin{eqnarray}}
10: \newcommand{\eeqn}{\end{eqnarray}}
11: \newcommand{\bet}{\begin{table}}
12: \newcommand{\ent}{\end{table}}
13: \newcommand{\bib}{\bibitem}
14:
15:
16: %\baselineskip 4.2mm
17:
18:
19: \wideabs{
20:
21: \title{
22: Simulated ion-sputtering and Auger electron spectroscopy depth profiling study of intermixing
23: in Cu/Co
24: }
25:
26:
27: \author{M. Menyh\'ard, P. S\"ule}
28: \address{Research Institute for Technical Physics and Materials Science,\\
29: Konkoly Thege u. 29-33, Budapest, Hungary,sule@mfa.kfki.hu,www.mfa.kfki.hu/$\sim$sule\\
30: %Accelerator Lab., Helsinki, Finland
31: }
32: %\email{sule@mfa.kfki.hu}
33:
34: \date{\today}
35:
36: \maketitle
37:
38: \begin{abstract}
39: The ion-bombardment induced evolution of intermixing is studied by molecular dynamics simulations
40: and by Auger electron spectroscopy depth profiling analysis (AESD) in Cu/Co multilayer.
41: It has been shown that
42: from AESD we can derive
43: the low-energy mixing rate and which can be compared with
44: the simulated values obtained by molecular dynamics (MD) simulations.
45: The overall agreement is reasonably good hence MD can hopefully be used to estimate
46: the rate of intermixing in various interface systems.
47:
48: %{\em PACS numbers:} 61.80.Jh, 61.82.-d, 66.30.-h\\
49: %{\em Keywords:} ion-sputtering, sputter removal, depth profiling, atomistic Computer simulations, thin film, multilayer, ion-solid interaction, molecular dynamics, Auger electron spectroscopy, Cu/Co
50:
51: \end{abstract}
52: }
53:
54:
55: %\scriptsize{
56: %61.80.Az Theory and models of radiation effects,
57: %61.80.Jh Ion radiation effects,
58: %61.82.-d Radiation effects on specific materials,
59: %61.82.Bg Metals and alloys,
60: %66.30.-h Diffusion in solids (for surface and interface diffusion, see 68.35.Fx)\\
61: %68. Surfaces and interfaces; thin films and low-dimensional systems (structure and nonelectronic properties) (for surface and interface chemistry, see 82.65.+r, for surface magnetism, see 75.70.Rf)
62: %68.35.Bs Structure of clean surfaces (reconstruction)\\
63: %68.35.-p Solid surfaces and solid-solid interfaces: Structure and energetics
64: %68.55.-a Thin film structure and morphology (for methods of thin film deposition, film growth and epitaxy, see 81.15.-z)
65: %79.20.Rf Atomic, molecular, and ion beam impact and interactions with surfaces\\
66: %81.40.Wx Radiation treatment (particle and electromagnetic) (see also 61.80.-x Physical radiation effects, radiation damage)\\
67: %61.46.+w Nanoscale materials: clusters, nanoparticles, nanotubes, and nanocrystals (see also 36.40.-c Atomic and molecular clusters; for fabrication and characterization of nanoscale materials, see 81.07.-b in materials science)
68: %\\
69: %79.20.Ap Theory of impact phenomena; numerical simulation\\
70: %34.50.Dy Interactions of atoms and molecules with surfaces; photon and electron emission; neutralization of ions (for surface characterization by particle-surface scattering, see 68.49.-h in surfaces, interfaces, thin films, and low-dimensional structures)\\
71: %34.10.+x General theories and models of atomic and molecular collisions and interactions (including statistical theories, transition state, stochastic and trajectory models, etc.)\\
72: %96.35.Gt Surface features, cratering, and topography\\
73: %}
74:
75:
76:
77: %\section{Introduction}
78:
79: Low-energy ion beams are commonly used in surface analysis and for film growth \cite{Donnelly,IAFG}.
80: The use of ion-sputtering in the controllable production of nanostructures and self-assembled nanoppaterns have also become one of the most important fields
81: in materials science \cite{IAFG,Chen,Persson,Timm,Facsko,Gomez}.
82: %However, our knowledge on the nanoscale design of materials is far from being complete.
83: %Auger electron spectroscopy depth profiling analysis (AESD) \cite{Hofmann},
84: %however, this leads
85: %to further intermixing and to other detrimental effects at interfaces (surface and interface roughening, defect creation, etc.) of thin multilayers \cite{Adamowich}.
86:
87: The future of nanotechnology ultimately rests on the controllable fabrication, integration, and stability of nanoscale devices.
88: However, the understanding of the fundamental phenomena leading to the formation, stability, and morphological evolution of nanoscale features is lacking \cite{Schukin}.
89: As the dimensions of the surface features is reduced to the nanoscale, many classical macroscopic (continuum and mesoscopic) models for morphological evolution lose their validity.
90: Therefore the understanding of the driving forces and laws governing mass transport involved in the synthesis and organisation of nanoscale features in solid state materials
91: is inevitable.
92: Unfortunately the fundamental understanding and the nanoscale control of interdiffusion
93: is not available yet \cite{Adamowich,Ladwig}.
94: In order to get more insights in the atomic relocation processes during postgrowth
95: low-energy ion-sputtering, it is important to measure and to calculate accurately the rate of
96: intermixing at the interface.
97:
98:
99:
100:
101: In this Letter we will show that using a newly developed code for simulated ion-sputtering based on molecular dynamics
102: we are able to get mixing rates.
103: We will also show that it is possible to extract the mixing rate data from AESD.
104: It turns out that the agreement of the two methods is reasonably good.
105: In this way we also could check the reliability of molecular dynamics simulations.
106:
107: The Auger electron spectroscopy depth profiling analysis (AESD) \cite{Hofmann,Barna} measurements were carried out on a Cu/Co multilayer system. The sample was made by sputter deposition on polished single-crystal (111) silicon substrates in a plasma beam sputter deposition system. It was characterized by XTEM and RBS and flat interfaces have been found \cite{Barna}. AESD depth profiling was carried out using a dedicated device \cite{Barna2} by applying Ar$^+$ ions of energy of 1 keV and angle of incidence of $10^{\circ}$ with respect to the surface of the crystal (grazing angle of incidence). The sample was rotated during ion bombardment.
108:
109: The depth profiles were measured as a function of the sputtering time keeping the bombarding ion current constant. A STAIB DESA 100 pre-retarded CMA with fixed energy resolution was used to record the AES spectra. The following AES peaks were detected Cu (60 eV), Cu (920 eV) and Co (656 eV, to avoid overlapping). A part of a typical depth profile is shown in Fig 1. For clarity only the copper is shown in Fig 1. To calculate the concentration the intensity of the copper Auger current was normalized to that of the pure copper. The depth scale was calculated from the known thickness of the sample \cite{Barna}.
110:
111: It is generally supposed that during ion mixing the atomic movements are similar to that of the usual diffusion and thus the occurring broadening can be described by the same equations \cite{Bolse}. Accordingly in case of a bilayer system the concentration distribution formed due to ion bombardment of the interface by a given fluence can be described by the $erf$ function. The variance, $\sigma^2$, of the $erf$ function determines the extent of ion mixing.
112: In many cases $\sigma^2$ is linearly dependent on the bombarding ion fluence $\Phi$. In these cases the $\sigma^2/\Phi$ ratio called as mixing rate characterizes the mixing for a given ion bombarding condition. Important
113: advantage of using the term mixing rate is that it can be in principle directly derived from the experiments.
114:
115: %------------------------------------------------------
116: \begin{figure}[hbtp]
117: \begin{center}
118: \includegraphics*[height=5cm,width=6cm,angle=0.]{fig1.eps}
119: \caption[]{
120: The measured depth profile: projectile Ar$^+$, ion energy 1 keV,
121: angle of incidence 84$^{\circ}$.
122: The concentration of Cu is given as a function of the removed
123: layer thickness (measured from the interface in $\hbox{\AA}$).
124: The diamonds denote the points at which $\sigma^2$ values are determined
125: by AESD.
126: %The dashed line is obtained by MD simulations using the error function
127: %composition profile.
128: {\em Inset figure}:
129: The broadening ($\sigma^2$) of the interface as a function of the fluence (ion/
130: $\hbox{\AA}^2$).
131: The dotted lines show the two extreme of the possible slopes (mixing rates).
132: }
133: \label{fig1}
134: \end{center}
135: \end{figure}
136: %------------------------------------------------------
137:
138: In the case of AESD we measure the Auger current of the elements present. The measured Auger current depends on the in-depth distribution as follows:
139: $I_j=i_j(1)+i_j(2) \kappa_j+i_j(3) \kappa_j^2+...$, where $I_j$ is the measurable Auger current of element $j$, $i_j(k)$ is the Auger current of element $j$ emitted by the atomic layer $k$, and $\kappa_j$ gives the attenuation of the Auger current crossing an atomic plane. $i_j(k) \approx X_j(k)N(k)$, where $X_j(k)$ is the concentration of element $j$ in layer $k$, while $N(k)$ is the number of atoms of the $k$-th atomic plane. It is evident that in general from a single measured $I_j$ one cannot determine the $i_j(k)$ values. On the other hand this equation can be used to simulate the measured Auger current during the depth profiling procedure if we assume an in-depth distribution.
140:
141:
142: We do not know any experimental measurement of the in-depth distributions formed during AESD applying low ($0.2-2$ keV) ion energy. It seems, however, that dynamic TRIM simulation can reliably be used to describe AESD \cite{Trim}. This calculation also provides the in-depth distribution during the procedure. It turns out that at the beginning of the depth profiling procedure, when the interface is still far from the surface the in-depth distribution can be approximated by the $erf$ function. For the evaluation of the
143: experimentally determined depth profile we will suppose that
144: (i) the interface has an intrinsic surface roughness, which can also be approximated by an $erf$ function of $\sigma_0$.
145: (ii) ion mixing is the only process contributing to the broadening of the interface at least in the beginning part of the depth profiling, which will be studied.
146: Thus for any measured copper Auger current we should find $x_0$ (the distance of the interface from the surface) and $\sigma_m$ (the measured variance) values of the erf function. Then the variance due to the ion bombardment induced mixing is $\sigma^2= (\sigma_m^2-\sigma_0^2)$.
147: We derived $\sigma^2$ at three depths (indicated in Fig 1.) to be $16 \pm 4$ $\hbox{\AA}^2$, $72 \pm 16$ $\hbox{\AA}^2$, and $135 \pm 35$ $\hbox{\AA}^2$. To proceed we must know the number of ions causing the broadening at the interface. In our experimental arrangement we cannot measure the ion fluence. On the other hand we can measure accurately the removed layer thickness. Taking the sputtering yield from the literature to be $Y \approx 1.2$ \cite{Barna}, we can derive the curve $\sigma^2$ vs. fluence $\Phi$ which is shown in inset Fig. 1. Inset Fig. 1 also
148: %------------------------------------------------------
149: \begin{figure}[hbtp]
150: \begin{center}
151: \includegraphics*[height=5cm,width=6cm,angle=0.]{fig2.eps}
152: \caption[]{
153: The simulated filtered $\langle R^2 \rangle$ (SAD) as a function of the number of the ion impacts (number fluence)
154: at 1 keV ion energy at grazing angle of incidence using simulated ion-sputtering (molecular
155: dynamics).
156: Dotted line is a linear fit to the $\langle R^2 \rangle$ curves.
157: $\langle R^2 \rangle$ is simulated with ion impact points at the free surface (9 ML above the interface) and also at $4$ ML above the interface.
158: %{\em Inset figure}:
159: %The simulated unfiltered $\langle R^2 \rangle$.
160: }
161: \label{fig2}
162: \end{center}
163: \end{figure}
164: %------------------------------------------------------
165: shows the two limiting slopes; thus we can derive the mixing rate being
166: $250 \pm 150$ $\hbox{\AA}^4$.
167:
168:
169:
170: Classical constant volume molecular dynamics simulations were also used to simulate the ion-solid interaction
171: (ion-sputtering) using the PARCAS code \cite{Nordlund_ref}.
172: The computer animations can be seen in our web page \cite{web}.
173: Further details are given in ref.
174: \cite{Nordlund_ref,Sule_PRB05,Sule_NIMB04}.
175: We irradiate the bilayer Cu/Co (9 monolayers, (ML) film/substrate)
176: with 1 keV Ar$^+$ ions repeatedly with a time interval of 5-20 ps between each of
177: the ion-impacts at 300 K
178: which we find
179: sufficiently long time for the termination of interdiffusion, such
180: as sputtering induced intermixing (ion-beam mixing) \cite{Sule_NIMB04}.
181: The initial velocity direction of the
182: impinging atom was $10^{\circ}$ with respect to the surface of the crystal (grazing angle of incidence)
183: to avoid channeling directions and to simulate the conditions applied during ion-sputtering.
184:
185: To describe homo- and heteronuclear interaction of Cu and Co, the Levanov's \cite{Levanov} tight-binding
186: potentials are used \cite{CR}.
187: %The first neighbour distance $r_0$ is calculated as a harmonic mean for the crosspotential
188: %on the basis of the Tersoff's scheme \cite{Tersoff}.
189: The cutoff radius $r_c$ is taken as the second neighbor distance.
190:
191: We randomly varied the impact position and the azimuth angle $\phi$.
192: In order to approach the real sputtering limit a large number of ion irradiation are
193: employed using automatized simulations conducted subsequently together with analyzing
194: the history files (movie files) in each irradiation steps.
195: In this article we present results up to 100 ion irradiation which we find suitable for
196: comparing with low to medium fluence experiments. 100 ions are randomly distributed
197: over a $40 \times 40$ \hbox{\AA}$^2$ area.
198: The size of the simulation cell is $100 \times 100 \times 75$ $\hbox{\AA}^3$ including
199: 62000 atoms.
200: %In order
201: %to ensure similar conditions to that of sputter removal,
202: %we also initialize the ion from below the surface (nearly above the interface by $10$ \hbox{\AA} (4 ML),
203: %embedded simulated ion-sputtering, ESIS).
204: %------------------------------------------------------
205: \begin{figure}[hbtp]
206: \begin{center}
207: \includegraphics*[height=5cm,width=6cm,angle=0.]{fig3.eps}
208: \caption[]{
209: The crossectional view of the ion-sputtered Cu/Co bilayer at 1 keV
210: Ar$^+$ ion energy (ions are initialized at -4ML) obtained by simulated ion-sputtering. The red circles are the incorporated
211: ions.
212: }
213: \label{fig3}
214: \end{center}
215: \end{figure}
216: %------------------------------------------------------
217: Chanelling recoils are left to move outside the cell
218: and in the next step these energetic and the sputtered particles are deleted.
219: %In this way
220: %we can concentrate directly on the intermixing phenomenon avoiding
221: %many other processes occur at the surface (surface roughening, sputter erosion,
222: %ion-induced surface diffusion, cluster ejection, etc.) which weaken energy deposition at the
223: %interface.
224:
225:
226:
227: Unfortunately the realistic simulation of the layer-by-layer sputter-removal is beyond
228: the performance of the computers available. Just to give a hint of the difficulties,
229: the simulation of few thousands of ion impacts and subsequent relaxations should be treated.
230: To reduce the computational demand,
231: we carried out
232: two types of simulations. In the first case ions are
233: initialized from the surface (9 ML far from the interface).
234: %We find the sputtering yield of $Y \approx 1.6$ ion/atom.
235: $100$ ions corresponds to $\Phi \approx 0.063$ ion/$\hbox{\AA}^2$ (the removal of
236: $\sim 0.35$ ML).
237: If the ion is initialized e.g. 5 ML below the free surface (-4 ML far from the interface), than
238: we can simulate high dose experiments, when nearly 4 ML is sputter-removed.
239: Although, the removal of the upper 4 ML has not been done,
240: we expect that this kind of an artificial setup of ion-sputtering
241: can simulate the real experiment.
242: This is because, we expect that the main source of interfacial mixing
243: is the progressive enhancement of energy deposition at the interface with respect
244: to the depth position of the ion-impacts.
245: %Using our simulated $Y \approx 1.7$, 1 ML removal is $\sim 0.3$ ion/$\hbox{\AA}^2$.
246: %Hence 4 ML removal leads to $\Phi \approx 1.2$ ion/$\hbox{\AA}^2$.
247: %the simulated $\Phi$ varies in the range of $1.2-1.6 ion/\hbox{\AA}^2$ up to
248: %100 ions.
249:
250:
251:
252: In Fig 2 the evolution of the sum of the square atomic displacements (SAD) along the depth direction of all intermixing atoms $\langle R^2 \rangle= \sum_i^N [z_i(t)-z_i(t=0)]^2$, where $z_i(t)$ is the depth position of atom $i$ at time $t$, can be followed as a function of the
253: number of ions.
254: Using the relation
255: $\sigma^2 = \langle R^2 \rangle$ it is possible to calculate the mixing rate
256: ($k=\langle R^2 \rangle/\Phi$) from the slope of the fitted straigh line
257: on the $\langle R^2 \rangle$ vs. $\Phi$ curve in Fig 2.
258: The contribution of intermixed atoms to $\langle R^2 \rangle$ is excluded in a given layer
259: if the in-layer concentration of them is less then $\sim 5$ \%.
260: AESD can not measure intermixed atoms which has very low concentration
261: and which are below the treshold sensitivity of AESD.
262: In the 2nd and 3rd Co layers below the interface we find less then $5$ and
263: $1$ \% Cu which are below the sensitivity of AESD measurements.
264: Ion-sputtering in the 1st (-9 ML) and in the 5th (-4 ML) layers
265: $k_{-9 ML} \approx 317$ $\hbox{\AA}^4$ and $k_{-4 ML} \approx 397$ $\hbox{\AA}^4$, are obtained, respectively.
266: The corresponding fluences are $\Phi \approx 0.063$ ion/$\hbox{\AA}^2$ and
267: $\Phi \approx 1.26$ ion/$\hbox{\AA}^2$ (the artificial removal of 4 ML corresponds to $\Phi \approx 1.2$ ion/$\hbox{\AA}^2$ using the $Y \approx 1.2$ ion/atom), respectively.
268: It must also be noted that Cai {\em et al.} obtained
269: $k \approx 400$ $\hbox{\AA}^4$ using 1 MeV Si$^+$ ions and x-rax scattering techniques
270: in Co/Cu multilayer \cite{Cai}.
271: Since the ion mixing occur at the low energy end of the cascade process, e.g.
272: the high energy collisional cascades split into low energy subcascades (see e.g. refs. in \cite{Sule_PRB05}),
273: an agreement is expected for.
274: We also give the measured and calculated mixing efficiencies of $\xi =k/F_D$, where
275: $F_D$ is the deposited ion-energy/depth.
276: We get $\xi
277: \approx 10 \pm 2$ $\hbox{\AA}^5$/eV both by experiment and by simulations.
278: On the basis of this value we can characterize low-energy ion-sputtering induced intermixing in Cu/Co
279: as a ballistic interdiffusion process.
280:
281: The crossectional view of the ion-sputtered system can be seen in Fig 3 as obtained
282: by MD simulations. The interface is only weakly intermixed.
283: The incorporated ions are also shown in Fig 3.
284: The simulation can also be seen as an animation \cite{web}.
285:
286: In this Letter we have presented
287: that the combination of atomistic simulations with
288: Auger electron spectroscopy depth profiling might be a new efficient method
289: to depth profiling analysis of multilayered materials.
290: Also, the reasonably good agreement between experiment and simulations
291: provides us the possibility of predicting interdiffusion properties for various multilayers
292: for which no experimental results are available.
293:
294:
295:
296: \section{acknowledgement}
297: {
298: \scriptsize
299: This work is supported by the OTKA grant F037710
300: from the Hungarian Academy of Sciences.
301: %I wish to thank to K. Nordlund for
302: %helpful discussions and constant help.
303: The work has been performed partly under the project
304: HPC-EUROPA (RII3-CT-2003-506079) with the support of
305: the European Community using the supercomputing
306: facility at CINECA in Bologna.
307: The help of the NKFP project of
308: 3A/071/2004 is also acknowledged.
309: }
310:
311:
312:
313: \begin{thebibliography}{99}
314:
315: \scriptsize{
316:
317: \bib{Donnelly}
318: S. E. Donnelly, R. C. Birtcher, and K. Nordlund, {\em Engineering Thin Films and Nanostructures with Ion Beams}, edited by E. J. Knystautas (Marcel Dekker, New York, 2005)
319:
320:
321: \bib{IAFG}
322: A. R. Gonz\'ales-Elipe, F. Yubero, J. M. Sanz,
323: {\em Low Energy Ion Assisted Film Growth}, (Imperial College Press,
324: Singapore, 2003).
325:
326: \bib{Chen}
327: H. H. Chen, O. A. Urquidez, S. Ichim, L. H. Rodriquez, M. P. Brenner, and M. J. Aziz,
328: Science {\bf 310}, 294 (2005).
329:
330: \bib{Persson}
331: A. I. Persson, {\em et al.}, Nature Materials, {\bf 3}, 677 (2004).
332:
333: \bib{Timm}
334: R. Timm, {\em et al.}, Appl. Phys. Lett., {\bf 85}, 5890 (2004).
335:
336: \bib{Facsko}
337: S. Facsko, T. Dekorsy, C. Koerdt, C. Trappe, H. Kurz, A. Vogt, and H. L. Hartnagel, Science,
338: {\bf 285}, 1551 (1999).
339:
340: \bib{Gomez}
341: L. G\'omez, C. Slutzky, J. Ferr\'on, Phys. Rev. {\bf B71}, 233402 (2005).
342:
343:
344: %\bib{Egelhoff}
345: %W. F. Egelhoff, Jr., P. J. Chen, C. J. Powell, R. D. McMichael, and M. D. Stiles,
346: %Prog. Surf. Sci., {\bf 67}, 355 (2001).
347: %M. T. Kief, W. F. Egelhoff, Jr.,
348: %Phys. Rev., {\bf B47}, 10785 (1993).
349:
350: \bib{Schukin}
351: V.A. Schukin, N. N. Ledentsov, D. Bimberg, {\em Epitaxy of Nanostructures},
352: Springer (2004).
353:
354:
355: \bib{Adamowich}
356: D. Adamowich, {\em et al.}, Appl. Phys. Lett., {\bf 86}, 211915 (2005).
357:
358: \bib{Ladwig}
359: P. F. Ladwig, {\em et al.}, Appl. Phys. Lett., {\bf 87}, 121912-1 (2005).
360:
361: \bib{Hofmann}
362: S. Hofmann,
363: Prog. Surf. Sci., {\bf 36}, 35 (1991).
364:
365: \bib{Barna}
366: A. Barna, M. Menyhard, G. Zsolt, A. Koos, A. Zalar and P. Panjan, J. Vac.Sci. Tech. {\bf A21}, 553 (2003).
367:
368: %\bib{Allen}
369: %M. P. Allen, D. J. Tildesley, {\em Comupter Simulation of Liquids}, (Oxford Science Publications, Oxford 1989),
370: %S. Yip (Ed.), {\em Handbook of Materials Modeling}, Springer, (2005).
371:
372: \bib{Barna2}
373: A. Barna and M. Menyhard, Phys.Stat Sol. (a) {\bf 145}, 263 (1994).
374:
375: \bib{Bolse}
376: W. Bolse, Mat. Sci. Eng. {\bf A253}, 194 (1998).
377:
378: \bib{Trim}
379: M. Menyhard, Surf. Interface Anal. {\bf 26}, 100 (1998).
380:
381: \bib{Nordlund_ref}
382: K. Nordlund, Comput. Mater. Sci, {\bf 3}, 448. (1995).
383:
384: \bib{web}
385: P. S\"ule, http://www.mfa.kfki.hu/$\sim$sule/animations/cuco.htm.
386:
387:
388: \bib{Sule_PRB05}
389: P. S\"ule, M. Menyh\'ard, Phys. Rev., {\bf B71}, 113413 (2005).
390:
391: \bib{Sule_NIMB04}
392: P. S\"ule, M. Menyh\'ard, K. Nordlund, Nucl Instr. and Meth. in Phys. Res.,
393: {\bf B226}, 517 (2004).
394:
395:
396: %\bib{Wadley}
397: %H. N. G. Wadley, X. Zhou, R. A. Johnson, M. Neurock, Prog. in Material. Sci., {\bf 46}, 329 (2001).
398:
399: \bib{Levanov}
400: N. Levanov, V. S. Stepanyuk, W. Hergert, O. S. Trushin, K. Kokko, Surf. Sci., {\bf 400}, 54 (1998).
401:
402: \bib{CR}
403: F. Cleri and V. Rosato, Phys. Rev. {\bf B48}, 22. (1993),
404:
405: %\bib{Tersoff}
406: %J. Tersoff, Phys. Rev. {\bf B39}, 5566 (1989).
407:
408: \bib{Cai}
409: M. Cai, {\em et al.}, J. Appl. Phys., {\bf 95}, 1996 (2004).
410:
411: }
412:
413:
414: \end{thebibliography}
415:
416:
417:
418:
419:
420: \end{document}
421: