cond-mat0604243/text.tex
1: \documentclass[prb,twocolumn,showpacs,superscriptaddress]{revtex4}
2: % \documentclass[prb,preprint,showpacs,superscriptaddress]{revtex4}
3:  \usepackage{graphicx}
4:  \usepackage{dcolumn}
5: 
6:  \begin{document}
7: 
8:  \title{Effects of deposition dynamics on epitaxial growth}
9: 
10:  \author{Jikeun Seo}
11:  \affiliation{Department of Ophthalmic Optics, Chodang University,
12:  Muan 534-701, Korea }
13:  \author{Hye-Young Kim}
14:  \affiliation{ Department of Physics, The Pennsylvania State University,
15:  University  Park, PA 16802}
16:  \author{J.-S. Kim}
17:  \affiliation{Department of Physics, Sook-Myung Women's University,
18:  Seoul, 140-742,  Republic of Korea}
19: 
20:  \date{\today}
21: 
22:  \begin{abstract}
23:    The dynamic effects, such as the steering and the screening effects during deposition,
24:  on an epitaxial growth (Cu/Cu(001)), is studied by kinetic Monte
25:  Carlo simulation that incorporates molecular
26:  dynamic simulation to rigorously take the interaction of the
27:  deposited atom with the substrate atoms into account.
28:  We find three characteristic features of the surface morphology
29:  developed by grazing angle deposition:
30:  (1) enhanced surface roughness, (2) asymmetric mound, and (3)
31:  asymmetric slopes of mound sides.
32:   Regarding their dependence on both deposition angle and substrate
33:  temperature, a reasonable agreement of the simulated results
34:  with the previous experimental ones is found.
35:  The characteristic growth features by grazing angle deposition
36:  are mainly caused by the inhomogeneous deposition flux due to
37:  the steering and screening effects, where the steering effects play
38:  the major role rather than the screening effects.
39:  Newly observed in the present simulation is that the
40:  side of mound  in each direction is composed of various facets
41:  instead of all being in one selected mound angle even if the slope
42:  selection is attained, and that the slope selection does not
43:  necessarily mean the facet selection.
44:  \end{abstract}
45: 
46:   \pacs{PACS numbers:68.35.-p,68.37.-d }
47:   \maketitle
48: 
49:   \section{Introduction}
50: 
51:    For the tailored growth of a structure on a substrate, careful consideration of
52:  energetic parameters, such as surface and interface energies, and
53:  kinetic parameters, such as diffusion barriers, is necessary.
54:  However, the parameters related to the dynamics of the deposited
55:  atoms, other than the deposition flux, have not been seriously
56:  considered. Recently, Dijken, Jorritsma, and Poelsema\cite{Dijken1}
57:  has observed by spot profile analysis low energy electron
58:  diffraction (SPA-LEED)  that the islands formed on Cu (001) show
59:  rectangular symmetry in contrast to the square symmetry of the
60:  substrate when the 0.5 monolayer (ML) of Cu is deposited at
61:  deposition angle ($\theta$) of 80$^o$ from the surface normal. They
62:  suggest a model in which the interaction between deposited atom and
63:  substrate atoms steers the deposited atom and thus results in the
64:  inhomogeneous deposition flux, so called the steering effect. Our
65:  previous study on the thin film growth \cite{Seo1}, simulating
66:  dynamics of the deposited atoms by molecular dynamics (MD), has
67:  confirmed this hypothetical model. These studies demonstrate that
68:  the dynamic parameters involved in the deposition process, which
69:  have been ignored in most of the previous studies, exert notable
70:  effects on the thin film growth. Since the steering effect is
71:  unavoidable during deposition process, it influences all the thin
72:  film growth by atomic deposition in some degree. Even in the case of
73:  the normal deposition, the steering effect is found to affect the
74:  thin film growth.\cite{Montalenti,Montalenti2,Yu} It has been also
75:  shown that the deposition dynamics is a cause of the unstable growth
76:  of thin film on a vicinal surface.\cite{Seo2} \par
77: 
78:  Dijken, Jorritsma and Poelsema \cite{Dijken2} has observed, in the
79:  thicker film (40 ML) growth with $\theta =80^o$ that the surface
80:   is rougher  than that with $\theta = 0^o$. It has
81:  also been observed that the slope of the mound facing the deposition
82:  direction is much steeper than that of the side shadowed from the
83:  deposition.\cite{Dijken2} Although a qualitative model is suggested
84:  relating the experimental result with the inhomogeneous deposition
85:  flux due to the steering and screening effects,\cite{Wormeester}
86:  there has not been any realistic growth study or simulation work
87:  which confirms such speculation.
88: 
89:  In the present work, we perform kinetic Monte Carlo (KMC) simulation
90:  in conjunction with MD simulation that is designed to probe the
91:  effects of the dynamic processes on the thick film growth in atomic
92:  level. The main results of our simulation are as follows; (1) the
93:  roughness increases with increasing deposition angle, (2) the mound
94:  formed in the thick film growth has rectangular symmetry with sides
95:  elongated in the direction perpendicular to the deposition
96:  direction, and (3) the slopes of the illuminated and shadowed sides
97:  of the mound significantly differ, which is consistent with the experimental
98:  results.\cite{Dijken2} The aforementioned three characteristic
99:  morphological features are mainly caused by the inhomogeneous
100:  deposition flux on the top terrace of the mound mainly due to the steering
101:  effects rather than the screening effects. In the present study, in
102:  addition, it is found that the side of mound in each direction is
103:  formed of various local facets instead of all being in one selected
104:  mound angle, even if the slope selection is attained,  and that the
105:  experimentally observed mound slope actually corresponds to the mean
106:  slope of various local facets coexisting on each mound side. Also
107:  found is that the dependence of the mound slope on growth condition
108:  is due to the variation of the relative population of the facets. \par
109: 
110: 
111:  \section{Simulation Schemes}
112: 
113:     Kinetic Monte Carlo (KMC) simulation is utilized to study the thin film
114:  growth by deposition. In most of the previous KMC simulations,
115:  the deposition process is treated by randomly or uniformly
116:  positioning atoms at arbitrary adsorption sites. In the present
117:  study, when the deposition process is selected during usual KMC
118:  routine, an MD routine is called to simulate the trajectory of a
119:  deposited atom by fully considering the interaction between the
120:  deposited atom and substrate atoms.\par
121: 
122:    The details of the simulation are as follows. The substrate, Cu(001),
123:  lies on the xy-plane (at z=0) with x-axis lying parallel to the
124:  [110] direction. In the deposition process, the deposition starts at
125:  the height of 11-28 $a_z$ above the substrate, and deposited atoms
126:  are incident at an angle in the direction of x-axis. Here, $a_z$ is
127:  the interlayer spacing of Cu(001). The interaction between a
128:  deposited atom and substrate atoms is calculated by summing pairwise
129:  Lennard-Jones potentials, $U(r)=4D[(\sigma/r)^{12}-(\sigma/r)^{6}]$.
130:  Here, $D=$ 0.4093 eV, $\sigma =$ 2.338 \AA,\cite{Sanders1} and $r$
131:  is the distance between two atoms. The initial kinetic energy of the
132:  deposited atom is set to 0.15 eV, which corresponds to the melting
133:  temperature of copper. During each deposition process, all the
134:  substrate atoms are assumed to be frozen in their positions. Verlet
135:  algorithm is used in the MD simulation.\par
136: 
137:   In between two sequential deposition processes, KMC simulation is
138:  performed to simulate the diffusion processes of atoms on the
139:  substrate. In the KMC, only the diffusion into empty lattice sites
140:  is allowed and the exchange diffusion is not allowed. Also the
141:  overhang is not allowed during both deposition and diffusion
142:  processes. The simulation system is composed of 400$\times$400
143:  atomic lattice sites in fcc (001) surface and a vacuum region on top
144:  of the substrate with height of 28 atomic layers. \par
145: 
146:  %%%%%%%%Table I%%%%%%%%%%%%%%%%%%%
147:   \begin{table}
148:   \caption{ diffusion barriers and
149:   parameters adopted in  our simulation}
150:   \begin{ruledtabular}
151:   \begin{tabular}{cc}
152:   type of diffusion & diffusion barrier \\
153:   \tableline
154:   single adatom hopping(E1) & 0.48 eV\\
155:    step edge diffusion (E2) & 0.44 eV \\
156:    dimer lateral bond breaking (E3) & 0.46 eV \\
157:    re-estbilishing of a NN bond (E6) & 0.18 eV \\
158:    ES barrier (ES) & 0.10 eV\\
159:    ES barrier (kink site) & 0.05 eV\\
160:   \tableline
161:   jump frequency($\nu_{0}$) & $2.4 \times 10^{13} $ \\
162:   deposition rate ($F_{0}$) & 0.00416 ML/s \\
163:   \end{tabular}
164:   \end{ruledtabular}
165:   \end{table}
166:   %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
167: 
168:    Values of diffusion coefficients and diffusion barriers are adopted from
169:  those used by Furman and coworkers,\cite{Furman,Mehl} which
170:  reproduced the surface morphology of Cu islands on Cu(001) very
171:  well; the step Ehrlich-Schwoebel (ES) barrier is 0.10 eV and the
172:  kink ES barrier is 0.05 eV. In total, eleven kinds of diffusion
173:  barriers (including the ES barriers) are used in the KMC simulation
174:  and some of the important diffusion barriers are listed in Table I.
175:  Note that the barrier (E2) for the diffusion along step edge is
176:  0.44 eV in the present study,
177:  which is much larger than the generally accepted values of 0.2
178:  $\sim$0.3 eV, to save the computation time for the very frequent
179:  diffusions back and forth along steps.
180:  We examine the dependence of the surface morphology on E2,
181:  and find that  that the morphology
182:  does not show any noticeable dependence on E2 down to 0.34 eV,
183:  the lowest tested value of E2.
184:  In addition, the simulation with the E2 fairly reproduces the real
185:  growth mode. Hence, we anticipate that the high E2 value would not
186:  seriously limit the validity of the present simulation.\par
187: 
188:   The surface roughness is determined by the root-mean-square
189:   fluctuation of surface height around the mean height.
190:   The mound radius is determined as the radius (r) that makes the first
191:  zero of the height-height correlation function $G({\vec r})= <h({\vec
192:  r})h(0)> - <h(0)>^{2} $, and the mound radii along x- and y-axis are
193:  calculated separately. All the results presented are
194:  obtained from the average of 20 simulations under identical
195:  conditions. Unless mentioned otherwise, the distance in
196:  a plane is in unit of a$_{nn}$,   and that in the vertical direction is
197:  in unit of a$_z$. Here, the nearest neighbor distance, a$_{nn}$,
198:   is $a_{nn}=a/\sqrt{2}$, and the interlayer distance is $a_{z}=a/2$,
199:   where $a$ is the lattice constant of Cu. \par
200: 
201: 
202:  \section{RESULTS}
203: 
204:  \subsection{Roughness}
205: 
206:  %%%%%% Fig. 1 roughness%%%%%%%%%%%%
207:   \begin{figure}
208:   \includegraphics[width=0.45\textwidth]{fig1}
209:   \caption{Surface roughness as a function of coverage at 250 K. Inset
210:  shows the roughness as function of the deposition angle at 10 ML
211:  coverage. The solid line in the inset is the simulation result of random
212:  deposition without considering the steering or screening effect (NSS: No
213:  Steering or Screening effect). }
214:  \end{figure}
215:  %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
216: 
217:   In Fig. 1,  the surface roughness is presented
218:  as a function of the coverage ($\Theta$)
219:   when Cu atoms are deposited at various deposition angles on Cu(001) at 250K.
220:   The most notable feature is that, even at the same
221:  coverage, the surface becomes much rougher as the deposition angle
222:  increases. Also, as $\Theta$ increases, the difference in roughness
223:  between the grazing angle deposition and the normal deposition
224:  ($\theta = 0^o$) multiplies. At $\Theta =$ 10 ML, the roughness with
225:  $\theta = 80^o$, comes to be twice larger than that with the normal
226:  deposition. (Inset of Fig. 1) The presently observed dependence of
227:  the roughness on the deposition angle is expected to originate from
228:  the deposition dynamics rather than the diffusion kinetics of the
229:  atoms on the substrate on the following grounds; (1)  such angle
230:  dependence of the roughness is observed irrespective of the
231:  substrate temperature ($T$) (Fig. 2), and (2) both steering and
232:  screening effects become more effective as $\theta$ increases, as
233:  revealed by the simulation.(Fig.1 Inset) \par
234: 
235: %%%%%%% Fig. 2 %%%%%%%%%%%%%%%%%%%%%
236:   \begin{figure}
237:   \includegraphics[width=0.45\textwidth]{fig2}
238:   \caption{Surface roughness as a function of temperature at coverage of
239:   10 ML for two deposition angles of 80$^o$ and 0$^o$ (normal
240:  incidence).  The temperature dependence is more distinct for 80$^o$ case
241:  than 0$^o$ case. }
242:   \end{figure}
243:   %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
244: 
245:  In Fig. 2,  shown is the dependence of the roughness on $T$; the
246:  roughness tends to decrease when $T$ becomes too high or too low.
247:  The $bell-shape$ curve is well-explained by the $T$-dependence of
248:  the destabilizing current by the ES barrier.\cite{Amar} Thus, such
249:  $T$-dependence is caused by the kinetics of deposited atoms and is
250:  irrelevant to the deposition dynamics. When the deposition is made
251:  at $\theta = 80^o$, such $bell-shape$ $T$-dependence is also
252:  observed, but now in an amplified form. This illustrates that the
253:  surface roughness is determined $synergetically$ by the diffusion
254:  kinetics and the deposition dynamics.\par
255: 
256:  \subsection{Mound radius}
257: 
258:  %%%%%%Fig. 3  (Mound radius) %%%%%%%%%%%%
259:   \begin{figure}
260:   \includegraphics[width=0.45\textwidth]{fig3}
261:   \caption{Mound radius as a function of coverage. Results for
262:  $\theta = 80^o$ are shown in solid and broken curve for
263:  radius along x- and y-axis, respectively. Result for the normal deposition
264:  ($\theta = 0^o$) is also shown in gray broken curve. The substrate
265:  temperature is 250 K for all the cases. Inset: the mound radius as a
266:  function of the deposition angle at the coverage of 10 ML.
267:  Open squares and
268:  closed circles signify the radii along x- and y-axis, respectively. }
269:   \end{figure}
270:   %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
271: 
272:     The mound radius as a function of the coverage and the deposition angle
273:  is shown in Fig. 3. When atoms are deposited in normal direction
274:  ($\theta = 0^o$), square mounds form with the same four-fold
275:  symmetry as that of the substrate. However, when the atoms are
276:  deposited at grazing angles, rectangular mounds form with evidently
277:  elongated sides along y-axis. (Here, x(y)-axis is parallel
278:  (perpendicular) to the deposition direction.) It is conspicuous,
279:  from Fig. 3 and its inset, that the difference between mound radii
280:  in x- and y-axis increases as does the coverage or the deposition
281:  angle.  This prediction agrees well with the experimental results of
282:  Dijken {\it et al.}\cite{Dijken2} and Lu {\it et al.}\cite{Lu},
283:  where  such side way growth of the mound with high aspect ratio has
284:  been observed with the deposition at grazing angle. In
285:  regards to Fig. 3, the anisotropy in the shape of mound is caused
286:  mainly because the growth of the mound along x-axis slows down once
287:  it reaches a certain length, $\simeq$13 $a_{nn}$. \par
288: 
289:  %%%%%%Fig. 4 %%%%%%%%%%%%
290:   \begin{figure}
291:   \includegraphics[width=0.45\textwidth]{fig4}
292:   \caption{Mound radius as a function of temperature for the deposition angle of
293:  80$^o$ and the coverage of 10 ML. Open and closed circles are for
294:  the radius along x- and y-axis, respectively. Inset: the aspect
295:  ratio of the mound radii as a function of temperature.}
296:   \end{figure}
297:   %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
298: 
299:   Fig. 4 shows $T$-dependence of the mound radii in two
300:  directions when 10 ML film is grown at $\theta = 80^o$.
301:  It shows that both mound radii increase as does $T$
302:  due to the increased atomic diffusion length. At both
303:  high and low $T$ regimes, the mound shape is very
304:  asymmetric. The inset of Fig. 4 shows the aspect ratio as a function
305:  of $T$ at $\Theta =$ 10 ML,  and  the difference between the 
306:  two mound radii as high  as 40 \%. (Here, the aspect ratio is defined
307:  as the ratio of the mound radius along y-axis to that along  x-axis.)
308:   The complicated dependence of the aspect ratio on the
309:  temperature suggests that the diffusion kinetics also plays a
310:  substantial role in the determination of the mound shape in
311:  conjunction with the deposition dynamics. The asymmetric mound shape
312:   is, however, found over the whole $T$-range, which illustrates
313:  that the effect of deposition dynamics on the mound shape is never
314:  wiped out by the diffusion kinetics over the examined $T$-range.  \par
315: 
316:  \subsection{Mound slope}
317: 
318:    For the characterization of the slopes of the mound, we
319:  investigate the local slope
320:  at each step on each side of the mound that is defined as
321:  the step height divided by the width of the adjacent
322:  lower terrace. Most steps in the sides of the mounds are of
323:  one-atomic-layer height. Thus, if the width of a lower terrace
324:  adjacent to a step is 0.5 a$_{nn}$, then, the local slope is
325:  a$_z$/0.5 a$_{nn}$ which corresponds to the slope of \{1,1,1\}-facet
326:  on the fcc (001) surface. For the steps with the terrace widths, 1.5
327:   and 2.5 $a_{nn}$, the corresponding local slopes are those
328:  of \{1,1,3\}- and \{1,1,5\}-facets, respectively.
329: 
330:  %%%%%%  Fig. 5 (slope coverage dependence)%%%%%%%%%%
331:   \begin{figure} 
332:    \includegraphics[width=0.45\textwidth]{fig5}
333:   \caption{Distribution of the facets that compose the sides of the
334:  mounds formed with the normal deposition (0$^o$) at  250 K. Inset:
335:  the average slope of the mounds as a function of the coverage. }
336:   \end{figure}
337:   %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
338: 
339:  In our simulation, we find that various kinds
340:  of steps coexist on each side of mound.
341:  In Fig. 5,  the distribution of the various steps with different
342:  local slopes is presented as a function of the coverage for the thin
343:  films grown by the deposition at $\theta = 0^o$ and $T =$ 250 K. As
344:  $\Theta$ increases, so does the mean slope. (Inset of Fig. 5) At
345:  $\Theta =$ 100 ML, the mean terrace width becomes 1.67 $a_{nn}$ that
346:  is close to that of the \{1,1,3\}-facet. However, at the coverage,
347:  the relative population of the step with the local slope of
348:  \{1,1,j\}-facet (from now on, referred as \{1,1,j\}-step) is 21 \%
349:  (j=1), 53 \% (j=3), 18 \% (j=5), 6 \% (j=7), and 2 \% (j=9).
350:  (The portion of steps with  their slopes less than that of \{1,1,11\}-facet
351:  is negligible.)   Therefore, even though the mean slope converges to that of
352:  \{1,1,3\}-facet, the actual relative population of steps with the
353:  mean slope is only 50 \% and the rest of the mounds are composed of
354:  the steps with relatively wide range of the local slopes.  \par
355: 
356:  %%%%%% Fig. 6 (slope angle dep)%%%%%%%%%%
357:   \begin{figure}
358:   \includegraphics[width=0.45\textwidth]{fig6}
359:   \caption{Distribution of the facets that compose the sides of the
360:  mounds formed after depositing 10 ML at 250 K at various deposition angles.
361:  (a) P-side: Two sides of mound facing in the direction perpendicular
362:  to the deposition direction, (b) IL-side: the side facing the deposition direction,
363:  and (c) SH-side: the shadowed (or back) side of  mound in the deposition direction.}
364:   \end{figure}
365:   %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
366: 
367:  %%%%%%Fig. 7 %%%%%%%%%%%%
368:   \begin{figure}
369:   \includegraphics[width=0.45\textwidth]{fig7a}
370:   \includegraphics[width=0.45\textwidth]{fig7b}
371:   \caption{Relative population of various facets and average facet slope
372:  at the coverage of 10 ML  as
373:  a function of temperature for both deposition angles of (a) 0$^o$  and (b-d) 80$^o$.
374:  (e)Mean slopes of the three sides of the mound
375:  as a function of the deposition temperature for the coverage of 10 ML. }
376:   \end{figure}
377:  %%%%%%%%%%%%%%%%%%%%%%%
378: 
379:  Now, we examine the dependence of the distribution of slopes or
380:  steps on the deposition angle for three inequivalent sides of the
381:  mounds, illuminated (IL), shadowed (SH), and perpendicular (P)
382:  sides. (See the caption of Fig. 6 for the description of the sides.)
383:  In Fig. 6 shown is the distribution of local steps on each side
384:  after depositing 10 ML at 250 K at various deposition angles. From
385:  the figure, we find that the distributions of the steps are similar
386:  for these three sides of the mounds for $\theta < 50^o$. As the
387:  deposition angle further increases, the distributions start to
388:  change and become quite distinct from each other at $\theta = 80^o$.
389:  For $\Theta =$ 10 ML and $\theta = 80^o$, 54 \% of the IL-sides are
390:  composed of \{1,1,1\}-step. On the other hand, only 30 \% of SH-side
391:  is composed of the \{1,1,1\}-step,  and thus its mean slope is much
392:  less steep  than that of the IL-side. In the SPA-LEED experiment of
393:  Cu/Cu(001), Dijken {\it et al.} \cite{Dijken2} reported that the
394:  IL-side is made of \{1,1,1\}-step and SH- and P-sides are of
395:  \{1,1,3\}-step. These experimental results nearly match the
396:  average behaviors observed in our simulation. \par
397: 
398:    We also investigate the distribution of the steps
399:    as a function of the substrate temperature (Fig. 7).
400:    For the case of grazing angle
401:  deposition, the step distributions (Fig. 7 (b)-(d)) and the mean
402:  slopes(Fig. 7 (e)) on the three sides become similar around 220 K.
403:  Further, they vary little, as the substrate temperature is lowered
404:  below 220K. These observations suggest that the step distribution
405:  for $T < 220 K$ represents the limiting behavior of the slope
406:  formation or the steepest slope reachable by deposition, during
407:  which some less steep steps such as \{113\}-steps other than the
408:  steepest \{111\}-steps occurs due to statistical fluctuation.\par
409: 
410:  As $T$ increases, the portion of the less steep steps increases and
411:  that of the steeper steps such as \{111\}-step decreases. That is,
412:  the mound becomes smooth as $T$ increases. The way of smoothing,
413:  however, differs depending on the mound side: As $T$ increases, the
414:  P-side (Fig. 7(b)) and the SH-side (Fig. 7(d)) rapidly become
415:  smooth, while the change in the step distribution on the IL-side
416:  occurs in the smaller scale (Fig.
417:  7(c)). This can also be seen clearly in the $T$ dependence of the
418:  mean slopes in Fig. 7(e). \par
419: 
420: 
421:  \section{DISCUSSION}
422: 
423: 
424:  \subsection{Effects of the deposition dynamics on the morphology of the film}
425: 
426:  In the previous section, we observe that the roughness (Inset of
427:  Fig. 1 and 2), the mound shape (Fig. 3 and 4) and the mound slope
428:  (Fig. 5-7) depend on the non-kinetic variable such as deposition
429:  angle. Such angular dependence can be attributed to the deposition
430:  dynamics that includes (1) the steering of the trajectory of the
431:  deposited atom due to  its interaction  with substrate
432:  atoms and (2) the screening effects of the deposited atoms due to
433:  the geometrical structure already formed on the substrate. Both
434:  effects cause inhomogeneous deposition flux
435:  depending  on the deposition condition and make the growth of thin films
436:  sensitive to the deposition condition. \par
437: 
438:  %%%%%% Fig 8  flux %%%%%%%%%%%%
439:   \begin{figure}
440:   \includegraphics[width=0.45\textwidth]{fig8}
441:   \caption{Deposition flux calculated by MD simulation.
442:  Atoms are deposited at grazing deposition angle of 80$^o$ on an 8-layer
443:  high mound surrounded by (1,1,5)-facets. (a) The local deposition flux in
444:  gray scale, where the brighter color indicates the higher local flux. (b) The local
445:  deposition flux along the line crossing the center of the mound along x-axis.
446:  The ordinate is the deposition flux in percentage
447:  relative to the average deposition flux over the total area. The deposition
448:  flux at normal (0$^o$) deposition is shown with + symbol as a
449:  reference and the mound is depicted by gray circles at the bottom. }
450:  \end{figure}
451:  %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
452: 
453: 
454:  To study the effects of the inhomogeneous deposition flux on the
455:  growth of thin film, we investigate the flux on an 8-layer high
456:  mound surrounded with \{1,1,5\}-facets by MD simulation. Fig. 8(a)
457:  shows the deposition flux over the mound in gray scale for $\theta =
458:  80^o$. Fig. 8(b) shows the deposition flux along a line passing
459:  through the center of the mound along the x-axis, which shows strong
460:  asymmetry, i.e., the deposition flux on the IL-side is 2 to 4 times
461:  larger than the average deposition flux and that on SH-side is only
462:  about 10 to 50 \% of the average. The simulation with the mounds
463:  surrounded by the different facets show the same trend, $i.e.$, the
464:  enhanced deposition flux on the IL-side and the reduced deposition
465:  flux on the SH-side.\par
466: 
467:  Such inhomogeneous deposition flux gives rise to different growth
468:  speeds on each side, and thus the film is rougher than the film
469:  grown under the normal deposition condition (Figs.1 and 2).
470:  Especially, the enhanced deposition flux near the front edge of the
471:  top terrace increases the destabilizing current toward ascending
472:  step edge by the ES barrier, and critically roughens the
473:  surface.\cite{Amar}(See the following section for the detailed
474:  description.) \par
475: 
476:  The  asymmetric shape of mound formed during grazing angle
477:  deposition (Fig. 3 and 4) is attributed to the decrease of the
478:  overall deposition flux along x-direction (i.e., over both the IL-
479:  and SH-sides), and the simultaneous increase of the effective
480:  deposition flux over the P-sides (Fig. 8) due to the following two
481:  reasons; the first one is the net mass transfer from the region near
482:  the rear edge of the mound to the top of the mound due to the
483:  steering effects, reducing the deposition flux along
484:  x-axis.\cite{Dijken1,Seo1} Some of the transferred mass on the top
485:  of the mound is in the long run redistributed equally to the four
486:  sides.\cite{Seo1} It effectively increases the deposition flux along
487:  the y-axis and {\it vice versa}. The second one is the increased
488:  deposition flux over the P-sides due to the attraction of the
489:  deposited atoms moving along the edges of the P-sides toward the
490:  sides.\cite{Seo1} Such imbalance of the effective deposition flux
491:  results in the faster growth speed in the y-direction than in the
492:  x-direction, giving birth to the asymmetric mounds elongated 
493:  along y-axis. \par
494: 
495:  The observed asymmetric slopes of the mounds observed for $\theta >
496:  50^o$ (Fig. 6) can also be explained by the inhomogeneous deposition
497:  flux on the following two grounds. Firstly, the higher deposition
498:  flux on IL-side than that on SH-side offers growth
499:  environment effectively equivalent to the lower growth temperature
500:  on IL-side than that on SH-side. Thus, the terrace size in
501:  IL-side is narrower than that in SH-side or, equivalently, the slope
502:  in IL-side is steeper than that in SH-side as observed in both the
503:  previous experiment\cite{Dijken2} and the present simulation. \par
504: 
505:  In addition, the flux distribution on the top terrace strengthens
506:  the asymmetric slope formation; A notable feature of the deposition
507:  flux in Fig. 8(b) is that the flux near the edge toward IL-side is
508:  much higher than that near the opposite edge, which should result in
509:  the increased density of islands near to the edge of IL-side
510:  on the top terrace. Sequential formation of islands on the top terrace
511:  preferentially close to the edge makes the mound have steps
512:  with the narrower terrace width or the steeper slope on IL-side than
513:  those on SH-side. \par
514: 
515:  In Fig.7, we observe the dependence of the smoothing kinetics on the
516:  side of the mound; as $T$ increases, relatively rapid smoothing
517:  occurs on both the P- and SH- sides, while it is retarded at the
518:  IL-side. This is because the deposition flux is larger on the
519:  IL-side than those on the other sides (Fig. 8); The mean capture
520:  length of the deposited atoms to form islands on the IL-side is
521:  shorter than those on the other sides. It means that the effective
522:  temperature felt by the deposited atoms on the IL-side is lower than
523:  that on the other sides. As a result, the smoothing proceeds
524:  relatively slowly on the IL-side, as $T$ increases.  \par
525: 
526:  In summary, both (1) the inhomogeneous deposition flux over the
527:  sides of the mound due to the steering and screening effects and (2)
528:  the enhanced deposition flux near the front edge of the top terrace
529:  of the mound due to the steering effect cooperatively give rise to
530:  the asymmetric mound formation with different lengths and slopes on
531:  each side,  and also accelerate the roughening of surface during
532:  deposition at grazing angle.\par
533: 
534: 
535:  \subsection{The steering effects v.s. the screening effects}
536: 
537:  %%%%%% Fig 9  Screening v.s. Steering %%%%%%%%%%%%
538:   \begin{figure}
539:  \includegraphics[width=0.45\textwidth]{fig9ab}
540:  \includegraphics[width=0.45\textwidth]{fig9cde}
541:    \caption{Simulation results with the screening effects alone
542:  with $\theta = 80^o$ and $\Theta = 10$ ML. As a  reference, the
543:  simulation results considering both screening and steering effects
544:  are also presented. (a) Roughness as a function of temperature: open
545:  circles for deposition considering screening effects alone, + for
546:  random deposition (NSS), and closed circles for deposition
547:  considering both screening and steering effects. (b) Mean slopes for
548:  the SH-side (circle) and the IL-side (square). Open (closed) symbols
549:  for deposition including screening effects alone (both screening and
550:  steering effects). (c-e) Mound radius as a function of coverage.
551:  Gray (black) curves correspond to the deposition considering
552:  screening effects (both screening and steering effects). Solid and
553:  broken curves are radii along x- and y-axis, respectively. }
554:  \end{figure}
555:  %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
556: 
557:   As the thickness of the film become thick, in addition to the
558:  steering effects, the geometric screening effects play an important
559:  role. To study the contribution of the individual dynamic effect to
560:  thin film growth, we carry out another set of growth simulation
561:  which takes only the geometric screening effects into account. In
562:  this simulation, the trajectory of the deposited atom is a straight
563:  line determined by the initial position and velocity of the
564:  atom.\par
565: 
566:  In Fig. 9(a) shown is the roughness evolution as a function of the
567:  temperature for three different cases; (1) the case assuming no
568:  steering and screening effects (NSS) or random deposition, (2) the
569:  case taking only the geometric screening effects into account, and
570:  (3) the usual simulation taking all the dynamic effects, i.e., both
571:  steering and screening effects. At low temperature (e.g., 190 K),
572:  the screening effects contribute to the roughening almost as much as
573:  the steering effects. Here, the contribution of the screening
574:  effects to the roughening is estimated as the difference between the
575:  roughness obtained with the screening effects included and that of
576:  NSS. The contribution of the steering effects is estimated similarly
577:  as the difference between the roughness with both effects considered
578:  and that with only the screening effects. The roughening due to the
579:  screening effects, however, decreases gradually as temperature
580:  increases and becomes negligible at temperatures higher than 260 K.
581:  On the other hand, the roughness due to the steering effect  keeps
582:  increasing up to 240 K and starts to decrease gradually at the
583:  higher temperatures. This suggests that the effects of the
584:  inhomogeneous deposition flux due to the screening effect alone
585:  relaxes through diffusion kinetics at a temperature lower than that
586:  due to the steering effects.\par
587: 
588:  %%%%%% Fig 10  flux %%%%%%%%%%%%
589:   \begin{figure}
590:   \includegraphics[width=0.45\textwidth]{fig10}
591:   \caption{Deposition flux calculated by considering only the screening
592:  effects. Atoms are deposited at grazing deposition angle of 80$^o$
593:  on an 8-layer high mound surrounded by (1,1,5)-facets.
594:  (a) The local deposition flux in
595:  gray scale, where the brighter color indicates the higher local flux.
596:  (b) The local
597:  deposition flux on the line crossing the center of the mound along
598:  x-axis is presented; the ordinate is the relative amount (\%) of
599:  deposition flux relative to the average deposition flux over the
600:   whole surface. The mound is shown with gray circles in the upper region.}
601:  \end{figure}
602:  %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
603: 
604:  The origin of the difference in the roughness and the relaxation
605:  temperature between these   two cases lies in their different
606:  flux distributions: In Fig. 10 shown is the flux distribution for the same mound as
607:  that in Fig. 8, but only with the screening effects taken into
608:  account. The flux near SH-side is totally depleted and about the same
609:  amount is added to IL-side. The most notable difference between
610:   two cases is found in the flux on the top terrace. The flux distribution
611:  taking both dynamic effects in Fig. 8 shows pronounced enhancement
612:  of the flux near the front edge of the top terrace, while no such enhancement of flux
613:  is found in Fig. 10. The steering effect causes
614:  vertical mass redistribution, displacing the flux
615:  on SH-side expected  for the random deposition to the top terrace,
616:  accelerating the roughening of the surface. In contrast, the
617:  screening effect just redistributes
618:  the depleted flux from SH-side to IL-side in the same plane.
619:  Thus, for the relaxation of the screening effects,
620:  only the diffusions in the same plane via terrace
621:  diffusion or diffusion along the edges of mound
622:  need to be activated.   For the relaxation of the steering effects, however,
623:   interlayer diffusion across the ES barrier that requires much higher activation energy
624:  than that for the in-plane diffusion should be accompanied. In fact,
625:  the $bell-shape$ dependence of the roughness on the temperature in
626:  Fig. 9(a), shown for the case when the steering effect is also
627:  considered, is reminiscent of the growth mode where the limiting
628:  process is the diffusion over ES barrier.\cite{Amar} Hence, the
629:  relaxation of the steering effects takes place at temperature
630:  higher than that for the relaxation of the screening effects. \par
631: 
632:    In Fig. 9(b), the mean slopes for IL- and SH-sides are shown as a function
633:  of growth temperature also for the aforementioned two cases. The
634:  mean slope of the IL- side is steeper than that of the SH-side over
635:  the whole temperature range for both cases. If only the screening
636:  effect is considered, the difference in their slopes dwindles as $T$
637:  increases, and finally becomes almost negligible around 270 K. At
638:  250 K, the slopes of both sides are slightly lower than that of
639:  \{113\} facet. In the previous experiment depositing 40 ML of Cu on
640:  Cu(001) at 250 K\cite{Dijken2}, the slopes of the IL- and SH-side is
641:  found to be those of \{111\} and \{113\}, respectively, in
642:  contradiction with the results of the aforementioned simulation
643:  considering the screening effects alone.\par
644: 
645:  The inclusion of the steering effect makes the slopes steeper,
646:  especially  on the IL-side, and the difference in
647:  slopes becomes evident as observed in the experiment. (Fig. 9(b))
648:  Further,  the mean slope of the IL-side lies in between \{111\} and
649:  \{113\}, and that of the SH- side corresponds to that of \{113\}.
650:  These slopes now approach the ones observed in the previous
651:  experiment.\cite{Dijken2}  \par
652: 
653:  Regarding the mound radius, the screening effects alone make
654:  the mounds of almost symmetric shape (Figs. 9 (c)-(e))
655:  which strikingly differs
656:  from what has been observed in the experiments.\cite{Dijken2,Lu}
657:  This is understood by the fact that the flux
658:  blocked by the screening effect on the SH-side
659:  results in the increase of the flux on the IL-side (Fig. 10),
660:  which looks as if  the blocked flux is simply displaced
661:  to the IL side.  Therefore,  the reduced lateral growth speed on the SH-side is
662:  almost compensated by the increased growth speed
663:  on the IL-side.
664:  Hence, the overall shape of mound remains square symmetric,\cite{square} if only the
665:  screening effects are taken into account in the growth.
666:   The asymmetric mounds with
667:  longer side along y-axis (Fig. 9)  form as observed in the
668:  experiment,\cite{Dijken2} only when the steering effect is
669:  added.  \par
670: 
671:    In short, in the thin film growth at grazing deposition angle, the
672:  asymmetric mound shape, the surface roughness, and the slope
673:  difference on the IL- and SH-sides of mounds
674:  result mainly due to
675:  the steering effect rather than the screening effect. (Fig. 9).
676:  Remembering that the main difference of the steering effects from
677:  the screening effects is the higher deposition flux near the front
678:  edge of the top terrace, the observed characteristics of the films
679:  grown at grazing deposition angle
680:   is largely determined by the growth characteristics of the
681:  top layer as the growth front.  \par
682: 
683:  \section{SUMMARY AND CONCLUSION}
684: 
685:   We perform KMC simulation to study the thin film growth by deposition at
686:  grazing angle. We observe (1) the notable increase of the surface
687:  roughness and (2) the asymmetry in both mound shape and slopes, as
688:  compared with the thin film grown by the normal deposition. Such
689:  results are in good agreement with the previous experimental
690:  observations. We find that the aforementioned structural features of
691:  the films grown by grazing angle deposition are mainly attributed to
692:  the steering effect rather than the screening effect. Especially,
693:  the inhomogeneous deposition flux on the top terrace induced by the
694:  steering effects is the most influential factor. \par
695: 
696:  We also make an additional interesting observation that the mound
697:  side is not composed of one kind of facet, even when the slope selection is attained.
698:   Instead, we find that there coexist
699:  variety of local facets and the selected mound slope observed in
700:  experiment represents only the mean slope of those.
701:  Therefore, the slope selection does not mean the facet selection.  \par
702: 
703:  \begin{acknowledgements}
704: 
705: 
706:  \end{acknowledgements}
707: 
708:  \begin{references}
709: 
710:  \bibitem{Dijken1} S. V. Dijken, L. C. Jorritsma, and B. Poelsema, Phys.
711:  Rev. Lett. {\bf 82},4038 (1999).
712: 
713:  \bibitem{Seo1} J.  Seo, S.-M.  Kwon,  H.-Y. Kim,  and J.-S.
714:  Kim, Phys. Rev. B {\bf 67}, R121402 (2003).
715: 
716:  \bibitem{Montalenti} F. Montalenti, M. R. Sorensen,  and A. F.  Voter,
717:  Phys. Rev. Lett. {\bf 87}, 126101 (2001).
718: 
719:  \bibitem{Montalenti2} F. Montalenti  and A. F.  Voter,
720:  Phys.  Rev. B {\bf 64}, R081401 (2001).
721: 
722:  \bibitem{Yu} J. Yu and J. G. Amar, Phys. Rev. Lett. {\bf 89}, 286103
723:  (2002).
724: 
725:  \bibitem{Seo2} J. Seo, H.-Y. Kim, and J.-S. Kim, Phys. Rev. B {\bf 71}
726:  075414 (2005).
727: 
728:  \bibitem{Dijken2} S.V. Dijken, L.C. Jorritsma, and B.
729:  Poelsema, Phys. Rev. B {\bf 61}, 14047 (2000).
730: 
731:  \bibitem{Wormeester} H. Wormeester and B. Poelsema, Phys. Rev. {\bf B
732:  66}, 165406 (2002).
733: 
734:   \bibitem{Sanders1}D. E. Sanders and A. E. DePristo, Surf. Sci.{\bf 254},
735:  341(1991).
736: 
737:  \bibitem{Sanders2} D.E. Sanders D.M. Halstead, and A.E. DePristo,
738:  J. Vac. Sci. Technol. A {\bf 10(4)}, 1986 (1992).
739: 
740:  \bibitem{Furman} I. Furman, O. Biham, Jiang-Kai Zuo, A. K. Swan,
741:  and John F. Wendelken, Phys. Rev. B {\bf 62}, R10649 (2000).
742: 
743:  \bibitem{Mehl} H. Mehl, O. Biham, I. Furman, and M. Karimi, Phys. Rev.
744:   {\bf B}, 2106 (1999).
745: 
746:  \bibitem{Amar} J.G. Amar and F. Family, Phys. Rev. B {\bf 54}, 14742
747:  (1996).
748: 
749:  \bibitem{Lu} T.-M. Lu, D.-X. Ye, T. Karabacak, and G.-C. Wang,
750:  Mater. Res. Soc. Symp. Proc. {\bf 849}, KK8.4.1 (2005).
751: 
752:  \bibitem{square} The slightly longer mound length in the y-axis
753:  than in the x-axis may be
754:  caused by the net mass transfer through diffusion from the IL-side
755:  to the P-side due to the large adatom density gradient resulting from
756:  the large difference in the deposition flux between the two sides.
757: 
758:   \end{references}
759: 
760: 
761: 
762: 
763:   \end{document}
764: