cond-mat0102166/ter.tex
1: \documentstyle[aps,prl,amssymb,twocolumn,latexsym]{revtex}
2: %\usepackage{latexsym}
3: 
4: \newcommand	{\beq}		 {\begin {equation}}
5: \newcommand	{\eeq}		 {\end	 {equation}}
6: \newcommand	{\ds} 	         {\displaystyle}
7: \newcommand	{\klm}[1]	  {{\, \left(\, {#1} \,\right) }}
8: \newcommand	{\ekl}[1]     {{\, \left[\, {#1} \,\right] }}
9: \newcommand	{\gkl}[1]     {{\, \left\{\, {#1} \,\right\}}}
10: \newcommand	{\skl}[1]     {{\left\langle {#1} \right\rangle}}
11: \newcommand	{\abs}[1]     {{\, \left| \,{#1}\,\right|}}
12: \newcommand{\lsed}{\mbox{$l_{\mbox{\scriptsize SED}}$}}
13: \newcommand{\ld}{\mbox{$l_{\mbox{\scriptsize D}}$}}
14: \newcommand{\lt}{\mbox{$l$}}
15: \newcommand{\lc}{\mbox{$l_c$}}
16: \newcommand{\lr}{\mbox{$l_r$}}
17: \newcommand{\li}{\mbox{$l_\ell$}}
18: %\newcommand{\lr}{\mbox{$l_{\mbox{\scriptsize r}}$}}
19: %\newcommand{\li}{\mbox{$l_{\mbox{\scriptsize l}}$}}
20: \newcommand{\tw}{\widetilde{w}}
21: 
22: \begin{document}
23: 
24: \twocolumn[\hsize\textwidth\columnwidth\hsize\csname@twocolumnfalse\endcsname
25: 
26: \title{\bf Particle currents and the distribution of terrace sizes in unstable epitaxial 
27: growth}
28: \author{{\sc M. Biehl$^{1,2}$, M. Ahr$^{1}$, M. Kinne$^{3}$, W. 
29: Kinzel$^{1}$, 
30:    and S. Schinzer$^{1,2}$}} 
31:  \noindent
32: \address{
33:             $^{1}$Institut f\"ur Theoretische Physik und Astrophysik,
34:             $^{2}$Sonderforschungsbereich 410\\ 
35:             Julius--Maximilians--Universit\"at W\"urzburg\\
36:             Am Hubland, 97074 W\"urzburg, Germany\\
37:             $^{3}$Lehrstuhl f\"ur Physikalische Chemie II\\
38:             Friedrich--Alexander--Universit\"at Erlangen--N\"urnberg \\
39:             Egerlandstr. 3, 91058 Erlangen, Germany 
40:        }
41: 
42: 
43: \maketitle
44: 
45: \begin{abstract}
46: A solid--on--solid model of epitaxial growth in $1+1$ dimensions
47: is investigated in which slope dependent upward and downward
48: particle currents compete on the surface. The
49: microscopic mechanisms which give rise to these currents 
50: are the smoothening
51: incorporation of particles upon deposition and
52: an Ehrlich--Schwoebel barrier which hinders inter--layer transport 
53: at step edges.  We calculate the distribution of 
54: terrace sizes and the resulting currents on a stepped surface with  
55: a given inclination angle.  The cancellation of the competing effects
56: leads to the selection of a stable {\sl magic slope\/}. 
57: Simulation results are in very good agreement with the theoretical findings. 
58: \end{abstract}
59: {PACS numbers: 81.10.Aj, 05.70.Ln, 68.55.-a}  \ \\   ]
60: 
61: Epitaxial growth has become a standard method for the production
62: of high--quality crystals and films, needed for  e.g.\ semiconductor devices.
63: An overview of experimental techniques  can be found in \cite{Krim}, for instance.
64: Significant effort has been devoted to a theoretical 
65: understanding of the many morphologies and scaling behaviors 
66: that can be observed in epitaxial growth, see e.g.\ \cite{pimpinelli} for a review of 
67: theoretical approaches. 
68: Here we address the frequently observed phenomenon
69: of mounds in unstable growth  which has
70: attracted considerable interest, see e.g. 
71: \cite{rostkrug,siegert,tang,epl,sed,gitter}.
72: Specifically, we consider situations in which competing
73: smoothening and steepening effects control the surface morphology
74: and lead to the selection of a stable  slope
75: in the system.
76: 
77: We discuss potential microscopic mechanisms which result in the
78: emergence of mounds and slope selection in the frame of
79: a discrete $(1+1)$--dimensional model. 
80: The net particle currents as well as the distribution of terrace 
81: sizes on a surface of a given slope can be worked out and this 
82: allows then to evaluate the magic slope  as well as
83: the complete statistical properties of the emerging surface.
84: The analysis complements previous theoretical investigations 
85: which address the mean terrace size only or neglect 
86: fluctuations explicitly \cite{georg} in the spirit of Burton Cabrera Frank (BCF) 
87: theory \cite{pimpinelli,BCF}.
88: We demonstrate that the full distribution of terrace sizes
89: carries relevant information that should be taken into account. 
90: Our results suggest, for example,  that it should be possible
91: to identify relevant microscopic mechanisms  from 
92: experimental data. 
93: 
94: Our $(1+1)$--dimensional model obeys the solid--on--solid  (SOS)
95: restriction, 
96: i.e.\ the surface can be described by an integer array of height 
97: variables $h_k$. Single particles are  deposited at 
98: randomly chosen  sites $k \in \{1,2,\ldots,L\}$.
99: Upon arrival, an incorporation process moves a particle to
100: the lowest available site  within a neighborhood of $\pm 
101: R$  lattice  constants, i.e.\ the site $j$ with 
102: $h_j = \min \left\{h_{k-R},\ldots,h_k,\ldots
103:  h_{k+R} \right\} $. 
104: In case of a tie, the site closest to the deposition is  
105: chosen. Only if this is still ambiguous, an additional random 
106: selection is performed. Such a smoothening mechanism, in absence 
107: of further effects, is commonly associated with the Edwards--Wilkinson 
108: universality class of
109: growth \cite{pimpinelli}. 
110: The parameter $R={\cal O}(1)$ (in lattice constants) 
111: is termed the incorporation radius and sets the 
112: typical length
113: scale of the process. Various interpretations of incorporation have
114: been considered,  including downward funneling 
115: on non--trivial
116: lattices and knock--out--processes due to the          
117: momentum of incoming particles, see e.g.\  \cite{evans,yue}.
118: 
119: 
120: A particle which is, after deposition and possibly incorporation, 
121: not yet bound to a lateral neighbor diffuses on the surface by 
122: performing a random walk (RW) until it 
123: reaches an additional binding partner and
124: becomes immobile or until it collides with
125: another moving adatom and forms an island
126: nucleus. 
127: In a density of diffusing particles, this nucleation process
128:  would result in a typical collision free path 
129: $\ld$. In the case of irreversible aggregation on a flat
130: substrate, and if islands
131: of two or more adatoms are considered immobile, it has been
132: shown that  $\ld \propto \left(d/f\right)^{1/4}$ in $(1+1)$ 
133: dimensions \cite{pimpinelli,villain}. Here,  $d$ is the diffusion constant
134: and $f$ the incoming flux, with all 
135: lengths in dimensionless lattice constants. 
136: 
137: 
138: In  \cite{epl,sed,gitter}  nucleation is represented in an 
139: effective single particle picture and $\ld$ fixes 
140: the typical distance of island nuclei in the first layers 
141: on a flat
142: substrate. 
143: After the formation of mounds, terrace sizes 
144: are much smaller than $\ld$, typically.
145: \iffalse
146: Throughout the following we will assume that typical terrace sizes
147: $\lt$ on the stepped surface are much smaller than the diffusion 
148: length
149: $\ld$.
150: \fi 
151: Hence we will completely neglect nucleation on the stepped surface
152: which is justified
153: for small incoming flux or fast diffusion, respectively. 
154: 
155: 
156: Here, the RW ends whenever the particle sticks irreversibly to a lateral 
157: neighbor, i.e. when it reaches a terrace step. Attachment can occur 
158: from below and above, in principle, but this symmetry is broken due
159: to the so--called Ehrlich--Schwoebel (ES) effect \cite{pimpinelli}: 
160: An additional energy barrier $E_{es}$ at step edges hinders downward 
161: moves of diffusing adatoms as this would involve loosely bound 
162: intermediate positions.  Our model takes the ES effect into account by
163: assigning  a probability $ p_{es} \, \propto \exp\ekl{-E_{es}/(k T)}$ 
164: to
165: downward moves. 
166: 
167: The ES effect results in an  uphill current of adatoms
168: because particles will stick
169: to upper terraces preferably for any $p_{es} < 1$. 
170: On the other hand, incorporation 
171: constitutes a downhill
172: current. Both effects are slope dependent and their cancellation 
173: gives rise to
174: the formation of mounds with a well--defined inclination angle.
175: Once the structures have built up the {\it magic slope}, a coarsening 
176: process begins
177: which decreases the number of mounds, see \cite{epl,sed,gitter} for details.  
178:   
179: We will first work out the distribution of terrace sizes  
180: which emerges in our model on surfaces with a given, fixed 
181: inclination.
182: Further, we  will calculate 
183: the mean displacement of a particle's final position from
184: its deposition site. The latter corresponds to the net particle  
185: current on the growing surface and its zero as a function of the 
186: inclination angle determines the stable slope.
187: 
188: 
189: We consider a triple of terraces on an inclined surface
190: with a central terrace $c$ of width $l_{c}$ (lattice sites 
191: $j=1,2,\ldots l_c$)
192: and its neighbors $\ell$ to the left ($r$ to the right) with size $\li$ $(l_r)$.
193: Without loss of generality, we assume
194: that the surface height decreases to the right. 
195: For a particle deposited on a site $k$ of terrace $c$ with $l_c > R$ 
196: we have to distinguish the following cases:  \\[-6mm]
197: \begin{itemize} 
198: \item[(a)] $k > l_c - R$:~ the incorporation process places the 
199: particle  
200:     at its final position at site $l_c +1$, attaching to the lower 
201: terrace end.\\[-6mm]
202: \item[(b)] $1\leq k \leq l_c -R$:~ the particle performs a random 
203: walk until
204:  it reaches one of the {\it trap sites}  $j=1$ or $j=l_c+1$ where it 
205: comes to rest.  This includes deposition at site $k=1$ without 
206: subsequent 
207: diffusion. \\[-6mm]
208: \end{itemize}
209: A diffusing particle located at a site $j$  with $2\!\leq\!j\!\leq\!l_c\!-\!1$ moves
210: to one of the neighboring positions $j\pm1$ with equal probability 
211: $1/2$. 
212: The asymmetry of the RW (b) is due to the ES--barrier 
213: present for jumps from site $j=l_c$.  We denote 
214: with 
215: $p_{es}/2$ the
216: probability for a downward move to site $l_c+1$.
217: With probability 
218: $(1-p_{es})/2$ the move is
219: rejected and the particle remains at $j=l_c$ for the next time step, whereas with 
220: probability $1/2$ it jumps to $l_c -1$. 
221: 
222: A straightforward exercise
223: yields the probability
224: for  a RW  initiated at site $k$ to end in $j=\lc\!+\!1$ by downward 
225: diffusion:
226: $q(k,l_c) = \left. \klm{(k\!-\!1)\, p_{es}} \right/ \klm{1\!+\!(l_c 
227: \!-\!1)  p_{es}} $
228: which obviously satisfies $q(1,l_c)=0$ for all $p_{es}$.
229: For similar problems of this type see for instance
230: \cite{Shehawey}. 
231: 
232: Hence, the total probability $d(l_c)$ for a deposition event 
233: to occur on terrace $c$  with 
234: subsequent downward diffusion is $ d(\lc) = \Delta(\lc)/L$ 
235:  with   \\[-4mm]
236: \beq 
237:  \ds \label{deltadef}
238:  \Delta(\lc) = \sum_{k=1}^{l_c\!-\!R} q(k,\lc) \!=\! 
239:  \frac{\klm{\lc\!-\!R}\klm{\lc\!-\!R-1} p_{es}}
240:  {2 + 2 \klm{\lc-1} p_{es}} \eeq
241:  if $\lc > R$ and $\Delta(\lc) =0 $
242:   else.
243:  The second case accounts for the fact that any particle directly 
244:  deposited  onto a  terrace of width $\lc \leq R$ will be 
245: incorporated 
246:   without performing  diffusion. 
247: 
248: 
249:  The quantity $\Delta(l_c)$ can be interpreted as the effective number of
250:  deposition sites which contribute to downward diffusion from terrace $c$. 
251:  The prefactor $1/L$ of $d(\lc)$ is simply the constant probability for deposition
252:  on any of the sites in the system.
253: 
254:   
255:  Now we can work out the probability $\tw(\lc\!\to\!\lc\!-\!1)$ for the  
256:  central terrace $c$ to be shortened by the next deposition event. 
257:  For $\lc >  R$ one finds 
258:   \begin{eqnarray}
259:    \tw (\lc\!\to\!\lc\!-\!1) &=& 
260:  \klm{R\!+\! [\lc\!-\!R\!-\!\Delta(\lc)] \!\
261:   +\!\Delta(\li)} / L  \nonumber \\
262:    &=&  \klm{\lc \!-\! \Delta(\lc) \!+\! \Delta(\li)} /L.
263: \label{shorten}
264:   \end{eqnarray}
265:   The first contribution, $R$, represents deposition on any of the $R$ sites
266:   left of terrace $c$. Note that the outcome of the subsequent 
267:   incorporation process is completely independent of the surface 
268:   configuration, in particular of the left neighbor terrace width 
269:   $\li$.  The second term $\ekl{\ldots}$ accounts for deposition on 
270:   $c$ with final attachment to the upper terrace. Finally 
271:   $\Delta(\li)/L$ is the probability for the shortening of $c$ through 
272:   diffusion from terrace $\ell$. 
273:   
274:  If $0<\lc<R$, only two processes can shorten the central terrace:
275:  incorporation from exactly $\lc$ sites left of $c$ and downward 
276:  diffusion from terrace $\ell$. One obtains $\tw (\lc\!\to\!\lc\!-\!1) =
277:  (\lc\!+\!\Delta(\li))/L$ and we finally observe
278:  that Eq.\ (\ref{shorten})
279:  is valid for any $\lc \geq 1$ since $\Delta(\lc)=0$ for all $\lc 
280:  \leq R$. Obviously, $\tw (\lc \to \lc-1) = 0$ if $\lc=0$ already. 
281:  
282:  
283:  Since $\lc$ can only increase at the cost of shortening $\lr$ at the 
284:  same time, one obtains immediately the result
285:  \beq \label{lengthen}
286:    \tw (\lc\!\to\!\lc\!+\!1) = \left\{ \begin{array}{cl}
287:    \frac{1}{L} \klm{\lr\!-\!\Delta(\lr) \!+\!\Delta(\lc)} & \mbox{if~}
288:     \lr>0 \\
289:      0 & \mbox{if~} \lr=0. \end{array} \right.
290:    \eeq 
291:   We proceed by assuming that in a {\it population} of terraces the
292:  distribution of their sizes factorizes, i.e.\ that a single, 
293: identical
294:   $p(l)$ is sufficient to describe their statistics. 
295:  For the limiting case of an infinite ES barrier ($p_{es}=0$) the 
296:  evolution of a terrace is independent of the entire configuration 
297:  left of it, and the above property can be shown to hold true.
298:  In general, neighbor terraces clearly 
299:  interact. Nevertheless, our simulations show that the assumption of
300:  identically distributed, independent terrace sizes yields excellent
301:  approximations, at the very least. Figure \ref{strom} shows, for instance, that
302:   the correlation coefficient of neighboring terrace sizes in a 
303:   system with $L=1000$ and $p_{es} = 0.2$ vanishes within error bars.
304: 
305: 
306:  The analysis simplifies significantly if terrace sizes are considered
307:  to be uncorrelated.  The above expressions 
308: (\ref{shorten},\ref{lengthen}) were obtained for a given
309: triple of terraces $\gkl{\ell,c,r}$. By averaging
310: $\tw (\lc\!\to\!\lc\!\pm\!1)$ 
311:   over $p(\li)$ and
312: $p(\lr)$, respectively, one obtains the mean probabilities
313: $w(\lc\!\to\!\lc\!\pm\!1)$:
314: \begin{eqnarray}
315:  \label{lengthenmean}
316:   w(\lc\!\to\!\lc\!-\!1) & = & 
317:  %   \sum_{\lr =0}^{\infty} p(\lr) \tw(\lc\!\to\!\lc\!+\!1) = 
318:    \left\{ \begin{array}{cl} 
319:    \frac{1}{L} \klm{\lc\!-\!\Delta(\lc) \!+\!\skl{\Delta}} 
320:     & \mbox{for~~} \lc >0 \\
321:     0 & \mbox{for~~} \lc=0 \end{array} \right. \\
322:    \label{shortenmean}
323:    w (\lc\!\to\!\lc\!+\!1) & = &
324:      % \sum_{\li =0}^{\infty} \, p(\li) \tw (\lc \to \lc-1) \, = 
325:      \klm{\skl{l} \!-\!\skl{\Delta} \!+\! [1\!-\!p(0)]
326:     \Delta(\lc) } / L
327: \end{eqnarray}
328: where the r.h.s.\ involve only the width of the considered terrace itself,
329: the frequency $p(0)$ of vanishing terrace sizes and the mean values
330:  $\skl{l}=\sum_{j=0}^{\infty} \, j p(j)$ and $ \skl{\Delta} = 
331:  \sum_{j=0}^{\infty} \, \Delta(j) p(j)$, see Eq.\ (\ref{deltadef}).  
332: 
333: 
334: 
335: 
336:  The evolution of terraces according to 
337:  (\ref{lengthenmean},\ref{shortenmean}) produces a stationary distribution $p(l)$, 
338:  if $ p(l\!+\!1) w(l\!+\!1\!\to\!l) = p(l)  w(l\!\to\!l\!+\!1) $,
339:  hence 
340:  \begin{equation} \label{recursion}
341:    p(l+1)  =  p(l) \, \frac{\skl{l} - \skl{\Delta} + [1\!-\!p(0)] \Delta(l)}
342:                            {l+1 - \Delta(l+1) + \skl{\Delta}} 
343:  \end{equation}
344:  for $l\geq0$. This relation is implicit, since all $p(l)$ have to be known
345:  for the evaluation of the averages on the r.h.s. 
346:  In the  particular case of an infinite ES barrier, Eq.\ (\ref{recursion})
347:  reads $ p(l\!+\!1)=p(l) \skl{l}/(l\!+\!1)$ which is  satisfied
348:  by the Poissonian $p(l) = \lambda^l \, e^{-\lambda} / l!$ with mean 
349:  $\skl{l}=\lambda$.   
350: 
351:  In order to obtain the stationary $p(l)$ for $p_{es}>0$ on a 
352:  surface with a given
353:  mean terrace size $\lambda$,  we  replace $\skl{l}$ with $\lambda$  
354:  and $\skl{\Delta}$ with an adjustable parameter $D$ in Eq.\ (\ref{recursion}). 
355:  The quantities $p(0)$ and $D$ are determined such 
356:  that $\sum_l p(l)=1$ is satisfied and $\skl{l}=\lambda$ is reproduced self--consistently.
357:  Note that then, by construction,
358:  (\ref{recursion})  guarantees $\skl{\Delta}=D$ as well. 
359:  In the numerical treatment, sums are truncated at a  value $l_{max}$, 
360:  with the resulting $p(l_{\max})$
361:  small enough to justify the truncation a posteriori.
362: 
363: 
364:  A particle deposited at, say, lattice site $i$ will become immobile at 
365:  a final position $j\neq i$ after incorporation and diffusion, in general. 
366:  The expected displacement $ \delta = \skl{j-i} $ depends on the distribution of
367:  terrace sizes in the system. Taking into account all possible displacement processes  
368:  and their corresponding probabilities one finds  
369:   \begin{eqnarray}  \label{totaldelta}
370:  \delta & = &
371:     \frac{1}{2} R (R+1) - \frac{1}{2} \sum_{l=0}^{R\!-\!1} p(l) \ekl{(R-l)(R-l+1)}  +  \\
372:   & & \sum_{l=R\!+\!1}^{\infty}  p(l) \klm{ -l(l\!-\!R\!-\!\Delta(l))\!+\!\frac{1}{2} l 
373:   (l\!+\!1) - \frac{1}{2} R(R\!+\!1)}, \nonumber 
374:   \end{eqnarray}
375:  where the $p(l)$ obtained from (\ref{recursion}) have to be inserted for a 
376:  given $\skl{l}=\lambda$. 
377:  Here, the first line corresponds to the expected (positive, downward)
378:  effect of incorporation 
379:  and the second represents the total (negative, upward) contribution of diffusion. 
380:  In the limiting case $p_{es}=0$ Eq. (\ref{totaldelta}) reduces to
381:  $\delta = R \skl{l} - \skl{l}^2/2$, exploiting the fact that $\Delta(l)=0$ in this case
382:  and $\skl{l^2} - \skl{l}^2 = \skl{l}$ for the Poisson distribution.
383: 
384:  Figure \ref{strom} shows the result of Monte Carlo simulations of 
385:  the growth process in a system of size $L=1000$ for the model with $R=2$ and
386:  different values of $p_{es}$, where  boundary conditions were used to fix  
387:  $\skl{l}=\lambda$.  We have displayed the particle current $\delta/\lambda$
388:  on the surface as a function of $\lambda$.
389:  Note that  $\delta$ in Eq. (\ref{totaldelta}) was obtained 
390:  for the normalization $\sum_l p(l)=1$
391:  which corresponds to a fixed number of terraces. 
392:  In systems with a fixed number $L$ of 
393:  lattice sites, an additional factor has to be introduced as the number of 
394:  terraces grows like $1/\lambda$.
395: 
396:  For very 
397:  steep surfaces, $\lambda\to0$, the displacement approaches
398:  the limiting value $R$,
399:  representing the fact that
400: \begin{figure}[t]
401: \begin{center}
402: \setlength{\unitlength}{2pt}
403: \begin{picture}(160,130)(0,0)
404: \put(0,0){\makebox(160,125)
405:           {\special{psfile=r2strom.ps
406:                     angle=270 hoffset=-160 voffset=170 hscale=32 vscale=32}}}
407:  \put(1,125){\large $\delta/\lambda$}
408:  \put(65,54){\large$\lambda$}
409: \caption{The mean particle displacement $\delta/\lambda$ vs.\ $\lambda=\skl{l}$, cf.\ Eq.\ 
410:          (\ref{totaldelta}), shown for the model with $R=2$
411:          and different ES barriers.  Symbols represent
412:          the result of simulations with $L=1000$ for
413:          $p_{es}= 0 (\mbox{\Large$\bullet$}), 0.1 (\blacksquare), 0.2 
414:          (\mbox{\large $\blacktriangledown$}), 
415:          0.4
416:          (\mbox{\large $\blacktriangle$})$.
417:          Averages were performed over 100 runs and standard error bars
418:          would be smaller than the symbols, where not shown. 
419:          In addition, the correlation coefficient $\rho\!=\!(\skl{l_a l_b}\!-\!
420:          \skl{l}^2)/(\skl{l^2}\!-\!\skl{l}^2)$ for neighboring terraces $a$ and $b$
421:          is displayed for the case $p_{es}=0.2$;
422:          crosses and error bars correspond to $10\cdot\rho$.}
423: \label{strom}
424: \end{picture}
425: \end{center}
426: \end{figure}
427: \noindent
428:  every deposited particle is shifted by $R$ lattice
429:  constant in the incorporation and then comes to rest.  In the limit of vanishing
430:  slope, our model yields a diverging negative ``upward'' current. 
431:  This is an artifact of completely neglecting nucleation, which inevitably 
432:  becomes 
433:  important as $\lambda\to\infty$ and  imposes a maximal displacement on the order of $\ld$.
434: 
435:  On mounded surfaces, bottom and top terraces  limit the extension of
436: inclined flanks. Any slope that  results in a net uphill current according to 
437:  Eq.\ (\ref{totaldelta})  will steepen
438:  this portion of the surface and
439: vice versa. Accordingly, a mean terrace width $\lambda_o$ will be stabilized which corresponds
440:  to the zero of $\delta (\lambda)$. 
441: In the presence of an infinite ES--barrier we find the exact relation $\lambda_o=2R$.
442: A naive and not quite correct argument was used  in \cite{epl} to obtain
443: the same result.
444: It is instructive to check that the magic slope cannot be obtained
445: from the condition that the mean displacement vanishes on a particular terrace of size
446: $\hat{\lambda}$. This would correspond to setting $p(l)=\delta_{l,\hat{\lambda}}$ 
447: in Eq. (\ref{totaldelta})
448: and ~ gives results ~ analogous to the BCF--like 
449: treatment in \cite{georg} 
450: which does not account for fluctuating terrace sizes. 
451: For $p_{es}=0$ one obtains, e.g., $\hat{\lambda}=2R\!+\!1\!=\!\lambda_o\!+\!1$.     
452: 
453: Fig. \ref{distri} shows the frequency of terrace sizes as observed in two different
454: systems which both stabilize the mean $\lambda_o=6$. 
455: In one we have set $p_{es}=0$
456: and $R=3$, the second \mbox{example corresponds to $R=2$ and $p_{es}=0.258$.}
457: Computer simulations
458: show excellent agreement. 
459: Systems with a very pronounced ES--effect produce a narrow distribution
460: with a very low frequency $p(0)$ of {\sl step\/}
461: {\sl bunching\/}, i.e. zero terrace sizes.
462: As a limiting case one
463: finds $p(0)=e^{-\lambda_o}=e^{-2R}$ for infinite ES--barrier $(p_{es}=0)$. 
464: \begin{figure}[t]
465: \begin{center}
466: \setlength{\unitlength}{2pt}
467: \begin{picture}(160,108)(0,0)
468: \put(0,0){\makebox(160,105)
469:           {\special{psfile=vgl.ps
470:                     angle=270 hoffset=-150 voffset=150 hscale=32 vscale=32}}}
471:  \put(2,105){\large $p(l)$}
472:  \put(80,36){\large$l$}
473: \caption{Frequency $p(l)$ of terrace sizes in two cases with 
474: $\skl{l}=\lambda_o=6$. The dashed line shows the theoretical
475:  prediction for $R=3, p_{es}=0$, triangles represent
476: simulations (as in Fig.\ \ref{strom}); the solid line and bullets
477: correspond to $R=2, p_{es}=0.258$.
478: The dot--dashed curve displays a geometric distribution with $\skl{l}=6$
479: for comparison.}
480: \label{distri}
481: \end{picture}
482: \end{center}
483: \end{figure}
484: \noindent
485: On the
486: contrary, step bunching is observed with a much larger frequency in cases
487: with a weaker ES--effect where the distribution $p(l)$ is much broader.
488: Fig.\ \ref{mvsp} displays $\lambda_o$ and the variance $\sigma^2$ of the 
489: terrace size distribution as functions of $p_{es}$. Note that $\sigma^2$ 
490: grows drastically with increasing $p_{es}$, indicating 
491: significant deviations from the Poissonian for
492: infinite ES--effect.  
493: 
494: The analysis of experimental data is frequently based on the
495: simple assumption of random, non--interacting terrace sizes on 
496: vicinal surfaces. This leads to the geometric distribution 
497: $p_g(l) = (1-1/\lambda)^{l-1} / \lambda $ with $\skl{l}=\lambda$, 
498: see e.g.\ \cite{wollschlaeger}  for a discussion. 
499: Note that $p_g(l)$ differs significantly from the type of statistics 
500: that we find in our model, cf.\ Fig.\ \ref{distri}. 
501: In particular, step bunching is much more frequent in this
502: simple picture: $p_g(0)=(\lambda\!-\!1)^{-1}$.
503: 
504: 
505: In summary we have presented a  microscopic model of unstable 
506: epitaxial growth in which it is possible to derive the net 
507: particle currents on surfaces of a given inclination.  For the
508: first time it is possible to work out the full distribution of 
509: terrace sizes in such a system. Further,
510: we were able to calculate the stable mean terrace 
511: size and the corresponding statistical properties of the surface. 
512: We have restricted ourselves to the 
513: analysis of $(1+1)$--dimensional growth in this work. 
514: However, our results should carry over to a more 
515: realistic $(2+1)$--dimensional picture to a large extent, 
516: whenever terrace edges do not meander significantly.
517: 
518: Our findings allow for a qualitative interpretation of experimental results
519: in systems which display slope selection: frequent step bunching
520: and a broad distribution  hint at a relatively weak ES--barrier.  
521: Narrow distributions with little or no step bunching indicate that
522: a significant ES--effect is present but is compensated for 
523: by smoothening effects like downhill funneling. 
524: 
525: Extensions of this work  will concern  desorption and its
526: \begin{figure}[t]
527: \begin{center}
528: \setlength{\unitlength}{2pt}
529: \begin{picture}(160,108)(0,0)
530: \put(0,0){\makebox(160,105)
531:           {\special{psfile=mvsp.ps
532:                     angle=270 hoffset=-145 voffset=150 hscale=32 vscale=32}}}
533:  \put(1,100){\large $\lambda_o,\sigma^2$}
534:  \put(68,34){\large$p_{es}$}
535: \caption{The selected mean $\lambda_o$ (solid lines) and $\sigma^2 =
536: \skl{l^2}-\lambda_o^2$ (dashed) vs.\ $p_{es}$.
537: Pairs of curves correspond to (from below) $R=1,2,3$.
538: Note that for $p_{es}=0$ we find $\lambda_o=\sigma=2R$. The symbols represent  two choices
539: of $(p_{es},R)$ which result in $\lambda_o=6 $, cf.\ Fig.\ \ref{distri}.}
540: \label{mvsp}
541: \end{picture}
542: \end{center}
543: \end{figure}
544: \noindent
545:  influence on the growth process. Preliminary results indicate that
546: a significant desorption rate can trigger a transition from slope selection
547: to rough growth
548: and we expect  non--trivial 
549: effects in the statistics of terrace sizes. 
550: 
551: This work was supported by the Deutsche Forschungsgemeinschaft
552: through a grant (M. Ahr) and through the 
553: Sonderforschungsberich 410 (S. Schinzer). We thank R. Metzler
554:  for a critical reading of the manuscript.
555: 
556:  
557: \ \\[-1.0cm]
558: \begin{thebibliography}{99}
559: \bibitem{Krim} J.\ Krim, G.\ $\!\!$ Palasantzas, Int.\ $\!\!$ J.\ $\!\!$ Mod.\ $\!\!$ Phys.\ B
560:  {\bf 9}, $599, \!\! 1995$
561: \bibitem{pimpinelli} A.\ Pimpinelli,  J.\ Villain, {\it Physics
562:  of Crystal Growth\/}, Cambridge University Press, 1998
563: \bibitem{rostkrug} M.\ Rost, J.\ Krug, Phys.\ Rev.\ E {\bf 55}, 3952, 1997
564: \bibitem{siegert} M.\ Siegert, Phys.\ Rev.\ Lett.\ {\bf 81}, 5481, 1998
565: \bibitem{tang} L.-H.\ Tang, P.\ Smilauer, D.D. Vvedensky, Eur.\ Phys.\ J.\ B {\bf 2}, 409,
566: 1998 
567: \bibitem{epl} M.\ Biehl, W.\ Kinzel, S.\ Schinzer, Europhys.\ Lett.\ 
568: {\bf 41}, 443, 1998
569: \bibitem{sed} S.\ Schinzer, M.\ Kinne, M.\ Biehl, W.\ Kinzel, Surf.\ Sci,\ {\bf 439}, 191, 1999
570: 
571: \bibitem{gitter} M.\ Ahr, M.\ Biehl, Surf.\ Sci.\ {\bf 465}, 339, 2000
572: 
573:  \bibitem{georg}  S.\ Schinzer, S.\ K\"ohler, G.\ Reents, Eur.\ Phys.\ J.\ B {\bf 15}, 161,
574: 2000
575: 
576: \bibitem{BCF} W.K.\ Burton, N.\ Carbrera, F.C.\ Frank, Philos.\ Trans.\ R.\ Soc.\
577:  London A {\bf 243}, 299, 1951 
578: 
579: \bibitem{evans} J.W.\ Evans, D.E.\ Sanders, P.A.\ Thiel, A.E.\ DePristo, 
580: Phys.\ Rev.\ B {\bf 41}, 5410, 1990
581: 
582: \bibitem{yue} Y.\  Yue, Y.K.\ Ho, Z.Y.\ Dan Phys.\ Rev.\ B {\bf 57}, 
583: 6685, 1998
584: 
585: \bibitem{villain} J.\ Villain, A.\ Pimpinelli, L.\ Tang, D.\ Wolf, J.\
586: Phys.\ I (France) {\bf 2}, 2107, 1992
587: 
588: \bibitem{Shehawey} M.A.\ El--Shehawey, J.\ Phys.\ A {\bf 33}, 9005, 2000
589: 
590: 
591: \bibitem{wollschlaeger} J.\ Wollschl\"ager, F.\ Sch\"afer, K.M.\ Schr\"oder,
592:  Surf.\ Sci.\ {\bf 396}, 94, 1997 
593: 
594: 
595: 
596: \end{thebibliography}
597: \end{document}
598: