cond-mat0611356/pap.tex
1: % ****** Start of file prb_temp.tex ******
2: %
3: %                    Physical Review B Latex Template
4: %
5: %   This file is a modification of the APS sample file in the REVTeX 4 
6: %   distribution (Version 4.0 of REVTeX, August 2001). 
7: %
8: %   Modified by:    Emmanuel Yewande 
9: %
10: %   See the REVTeX 4 README file for restrictions and more information.
11: %
12: % TeX'ing this file requires that you have AMS-LaTeX 2.0 installed
13: % as well as the rest of the prerequisites for REVTeX 4.0
14: %
15: % It also requires running BibTeX. The commands are as follows:
16: %
17: %  1)  latex apssamp.tex
18: %  2)  bibtex apssamp
19: %  3)  latex apssamp.tex
20: %  4)  latex apssamp.tex
21: %
22: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb,prb,floatfix]{revtex4}
23: %Use this initially, to see 2 column output
24: 
25: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
26: %Use this for the version to be submitted to a preprint archive
27: 
28: % Some other (several out of many) possibilities
29: %\documentclass[preprint,aps]{revtex4}
30: %\documentclass[preprint,aps,draft]{revtex4}
31: 
32: %\documentclass[prb]{revtex4}% Use this for submission to Physical Review B
33: 
34: \usepackage{graphicx}% Include figure files
35: \usepackage{dcolumn}% Align table columns on decimal point
36: \usepackage{bm}% bold math
37: \usepackage{natbib}
38: %\nofiles
39: 
40: \newcommand{\change}[1]{{\large \bf #1}}
41: %% for highlighting changes made on the paper. Usage: \change{added text}
42: 
43: \begin{document}
44: 
45: \preprint{}
46: 
47: \title{Numerical analysis of the quantum dots on off-normal incidence ion
48:   sputtered surfaces}% Force line
49:                                 % breaks with \\ 
50: 
51: \author{Emmanuel O. Yewande}
52: \email{e.yewande@mmu.ac.uk}
53: \thanks{present address: Dept. of Computing \& Mathematics, 
54: MMU, John Dalton Building, Manchester M1 5GD, United Kingdom.}
55: \author{Reiner Kree}%
56: \email{kree@theorie.physik.uni-goettingen.de}
57: \author{Alexander K. Hartmann}%
58: \email{hartmann@physik.uni-goe.de}
59: \affiliation{%
60: Institut f\"ur Theoretische Physik, Friedrich-Hund Platz 1, D-37077
61: G\"ottingen, Germany.
62: }%
63: 
64: \date{\today}% It is always \today, today,
65:              %  but any date may be explicitly specified
66: 
67: \begin{abstract}
68: We implement substrate rotation in a 2+1 dimensional solid-on-solid 
69: model of ion beam sputtering of solid surfaces. With this extension of
70: the model, we study the effect of concurrent rotation, as the surface
71: is sputtered, on possible topographic regions of surface patterns. In
72: particular we perform a detailed numerical analysis of the time
73: evolution of dots obtained from our Monte Carlo simulations at
74: off-normal-incidence sputter erosion. We found the same power-law
75: scaling exponents 
76: of the dot characteristics for two different sets of ion-material
77: combinations, without and with substrate rotation. 
78: \end{abstract}
79: 
80: \pacs{05.10.-a,68.35.-p,79.20.-m}% PACS, the Physics and Astronomy
81:                              % Classification Scheme.
82: %\keywords{Materials, Surfaces/Interfaces, Statistical Mechanics}%Use
83:                                 %showkeys class option if keyword 
84:                               %display desired
85: \maketitle
86: 
87: \section{\label{sec:}INTRODUCTION}
88: The size-tunable atomic-like properties of (e.g.\ II-VI and III-V)
89: semiconductor nanocrystals have diverse applications. These properties  
90: arise from the quantum confinement of electrons or holes in the
91: quantum dots (QDs) to a region on the order of the electrons' de Broglie
92: wavelength. Examples of the applications can be found in solid-state quantum
93: computation \cite{Gershenfeld98} and in 
94: electronic and opto-electronic devices like   
95: diode lasers, amplifiers, biological sensors,              
96: electrolumniscent displays, and photovoltaic cells. \cite{Colvin94, ORegan91}
97: 
98: A more recent method of fabrication is the sputtering of
99: semiconductor surfaces with a beam of energetic ions impinging at an
100: angle $\theta$ with respect to the direction perpendicular to the surface. \cite{Facsko99}
101: This has been 
102: shown to be a cost-effective and more efficient means of producing
103: uniform high-density semiconductor nanocrystals,
104: %include more properties of dots obtained
105:                           %from sputtering
106: \cite{Facsko99, Frost00, Facsko01, Gago01} in contrast to previous
107: methods such as epitaxy, lithographic techniques, colloidal
108: synthesis, electrochemical techniques, and pyrolytic synthesis.      
109:   
110: Using the continuum theory, it was shown that QD formation by $\theta
111: =$ 0 sputtering is restricted 
112: to a very narrow region of the parameter space. \cite{Kahng01,
113:   Frost02} It has also been shown 
114: that dot formation is possible for $\theta >$ 0, within a broader
115: region, under concurrent
116: sputtering and sample rotation.\cite{Frost02} Using a simple discrete
117: solid-on-solid  
118: model, which includes the competing processes of surface roughening via
119: sputtering and surface relaxation via thermal diffusion, we recently
120: found that for $\theta >$ 0, without sample
121: rotation, a dot topography is only one among other possible
122: topographies which may 
123: arise. The type of the emerging topography 
124:  depends on the longitudinal and lateral straggle of the
125: impinging ion as it dissipates its kinetic energy via collision
126: cascades with atoms within the material. \cite{Yewande06}
127: 
128: In this study, we implement sample rotation in the simulation model
129: along the lines of Refs.\ \onlinecite{Bradley_Cirlin96} and
130: \onlinecite{Bradley96}. 
131: For varying values of the ion parameters we find different kinds of surface
132: patters, including dots. 
133: The dots are similar to those obtained without sample rotation, but
134: the underlying oriented ripple structures are lacking. We study the time
135: evolution of dots emerging from oblique ion incidence in more 
136: detail, performing simulations for two different sets of parameters,
137: corresponding to two different materials (GaAs sputtered with Ar and
138: Si sputtered with Ne) with and without rotation. We found
139: that without rotation 
140: the average number of dots decreases with increasing fluence, but stays
141: approximately constant for a rotated sample. 
142: Furthermore the uniformity of dots is
143: greatly enhanced by sample rotation. Remarkably, both materials exhibit
144: the same scaling of the dot characteristics with sputter time.
145: 
146: The rest of the paper is organized as follows: In section \ref{sec:CT}
147: we shall briefly review the continuum theory of sputtered amorphous
148: surfaces. In section \ref{sec:sim} we shall give a brief description
149: of the discrete simulation models applied in this 
150: work. In section \ref{sec:rot_effect} we study the effect of rotation
151: on the possible topographies reported in Ref.\ \onlinecite{Yewande06};
152: expecially off-normal incidence dot formation. 
153: Finally, in section \ref{sec:analysis}, we shall present and discuss our
154: results of the analysis of dot topographies.       
155: 
156: \section{\label{sec:CT}CONTINUUM THEORY}  
157: In the seminal theory of P. Sigmund on ion-beam sputtering of amorphous 
158: and poly-crystalline targets, it was shown that the spatial energy
159: distribution  
160: $E({\bf r})$ of an impinging ion may be approximated by a two-dimensional 
161: Gaussian of widths 
162: $\sigma$
163: and $\mu$, parallel ($z'$-axis) and perpendicular 
164: ($x',y'$-axes) to the ion beam direction, respectively
165: \cite{Sigmund69, Bradley88},   
166: \begin{equation}
167: \label{eq:GE} %%Gaussian Energy
168: E({\bf r}')=\frac{\epsilon}{(2\pi)^{3/2}\sigma\mu^2}\exp\biggl(
169: -\frac{[z'+a]^2}{2\sigma^2}-\frac{x'^2+y'^2}{2\mu^2}\biggr)\,,    
170: \end{equation}
171: where $\epsilon$ is the total energy of an impinging ion and 
172: $a$ the average penetration depth. 
173: 
174: When describing the surface by a two-dimensional height field $h({\bf r}, t)$
175: (solid-on-solid description),
176: the normal erosion
177: velocity   $v=-[1+(\nabla h)^2]^{-1/2}\partial_t h$  
178: at the point ${\bf r}=(0,0,0)$ is proportional to the total power
179: transported to this point by
180: the ions of the impinging ion beam. It may
181: be expressed as  
182: \cite{Cuerno95, Makeev02} 
183: \begin{equation}
184: \label{eq:erov} %erosion velocity
185: v=p_c\int_Rd{\bf r}\Phi({\bf r})E({\bf r}).  
186: \end{equation}
187: The integral is taken over the region $R$ containing all the points at
188: which the deposited energy contributes to the total power at ${\bf r}=0$
189: and $p_c$ is a proportionality constant.\cite{Makeev02}
190:  $\Phi({\bf r})$ is a local correction to the uniform flux $J_f$. Note
191:  that shadowing effects among neighboring points and redeposition of
192:  eroded material are ignored. 
193: 
194: By standard arguments, an equation for the evolution of the surface height
195: field $h(x,y,t)$ due to sputtering can be derived from Eq.\ (\ref{eq:erov}), which
196: takes on the form
197: \cite{Bradley88, Cuerno95}
198:  \begin{eqnarray}
199: \label{eq:hS}
200: (\partial_th)_S \approx -v_f +
201: v_f^\prime\partial_xh+\nu_x\partial^2_xh+\nu_y\partial^2_yh 
202: \nonumber
203: \\
204: +\frac{\chi_x}{2}(\partial_xh)^2+\frac{\chi_y}{2}(\partial_yh)^2 +
205: \eta_S({\bf r}, t),
206: \end{eqnarray}
207: where the $x$-axis is parallel to the ion beam direction, $v_f$ is the
208: erosion velocity of a flat surface and $\nu_x(\theta, \sigma, \mu)$ 
209: [$\nu_y(\theta, \sigma, \mu)$] is the surface-tension 
210: coefficient along [perpendicular to] the ion beam
211: direction. Depending on the parameters $\theta,
212: \sigma, \mu$,  the quantity $\nu_x$ can exhibit positive as well as negative values, whereas $\nu_y$ is
213: always negative. $\chi_x$ and $\chi_y$ are the coupling constants of
214: the dominant non-linearities along the
215: respective directions. $\eta_S$ is an uncorrelated noise term, with zero
216: mean. This term represents the random arrival of the ions onto the
217: surface.     
218: 
219: The surface height also evolves due to surface particles hopping from
220: one point to the other. On a coarse-grained
221: level this can be described  by the continuity equation  
222: for the conserved particle current  
223: \begin{equation}
224: \label{eq:hD}
225: (\partial_t h)_D=-\nabla\cdot {\bf j}+\eta_D({\bf
226:   r}, t),
227: \end{equation}  
228: where the local current density  ${\bf j}$ is a function of the
229: derivatives of $h$; i.e, ${\bf j}$ = ${\bf j}[\nabla^m h, (\nabla
230: h)^n]$ ($m$ and $n$ are integers), and the acceptable
231: functional form is subject to symmetry constraints. $\eta_D$ reflects the
232: randomness inherent in the surface-diffusion process.                  
233: A commonly studied example of a surface migration model is the
234: Mullins-Herring model, 
235: which leads to the particularly simple form
236:  $\nabla\cdot {\bf j}=D\nabla^4h$ for the current density.\cite{Kim94}
237: 
238: Thus, considering Eqs.\ (\ref{eq:hS}) and (\ref{eq:hD}), the time evolution
239: of a sputtered surface is governed by a Kuramoto-Sivashinsky type
240: equation 
241: \begin{eqnarray}
242: \label{eq:KS}
243:  \partial_th = -v_f +
244: v_f^\prime\partial_xh+\nu_x\partial^2_xh+\nu_y\partial^2_yh 
245: +\frac{\chi_x}{2}(\partial_xh)^2 
246: \nonumber
247: \\
248: +\frac{\chi_y}{2}(\partial_yh)^2
249: -D\nabla^4h +
250: \eta({\bf r}, t).
251: \end{eqnarray} 
252: 
253: The exact form of each coefficient of
254: Eq.\ \ref{eq:KS} is given in Ref.\ \onlinecite{Makeev02}
255: and will be used later in the analysis of our
256: results (see Fig.\ (\ref{fig:iso_coeff}) in section\ (\ref{sec:sim})).
257: 
258: The anisotropy ($\nu_x\ne\nu_y$, $\chi_x\ne\chi_y$) arises from the
259: oblique incidence which leads to different erosion rates (or different
260: rates of maximizing the exposed area) along
261: parallel and perpendicular directions relative to the ion beam
262: direction. 
263: From the linearized Eq.\ (\ref{eq:KS}) it is easy to see that periodic ripples are growing  with wavevector ${\bf
264: k}=\sqrt{\max(|\nu_{x/y})|/2D}$ driven by the dominant negative
265: surface tension. These ripples are stabilized by the surface diffusion term.
266: With prolonged
267: sputtering surface slopes become too big to neglect the nonlinear terms and the ripples are
268: destroyed, after which a new rotated ripple structure may
269: emerge. \cite{Park99} 
270: 
271: If the sputtered substrate is rotated with sufficiently large angular
272: velocity, \cite{Thesis} 
273: or if $\theta=$ 0, no externally induced 
274: relevant anisotropy is present. 
275: In this case
276: ripples do not develop for the class of materials we study here 
277: and the relevant evolution equation is the
278: isotropic version of Eq.\ (\ref{eq:KS}), which predicts cellular
279: structures of mean width $\sim (D/|\nu|)^{1/2}$ and mean height $\sim
280: |\nu|/|\chi|$.\cite{Bradley96} According to Ref.\
281: \onlinecite{Kahng01}, 
282: these cellular structures eventually evolve to either a dot 
283: topography ($\chi>$0) or a hole topography ($\chi<$0); which explains
284: the normal incidence dot\cite{Facsko99, Gago01, Facsko01} and
285: hole\cite{Rusponi98, Rusponi99} topographies found in 
286: previous experiments.    
287: 
288: Thus, one would expect oblique incidence dot formation to result from 
289: isotropic sputtering induced by substrate rotation. However,
290: in our recent simulations \cite{Yewande06} 
291: we found dot topographies without sample rotation,
292: though with some 
293: underlying (anisotropic) ripples oriented parallel to the
294: ion beam direction. We found that the oblique incidence dots arise
295: from a dominance of the ripples oriented parallel to the ion beam
296: direction, over those with a perpendicular orientation; the dots being the
297: remains of such perpendicular ripples (see Fig.\
298: \ref{fig:phase_profiles} below) after long times.  
299: 
300: In the next chapter
301: we will outline the simulation method we have used in 
302: Ref.\ \onlinecite{Yewande06} and which we have extended to the case
303: with rotation as presented in this work.
304: 
305:      
306:      
307: \section{\label{sec:sim}MONTE CARLO SIMULATION}
308: 
309: To simulate the competing processes discussed in section \ref{sec:CT}
310: on a discrete system, we use the 
311: Monte Carlo model of sputter erosion introduced in Ref.\
312: \onlinecite{Hartmann02} (HKGK model). To make the paper self-contained,
313: we provide some details here. 
314: We simulate the sputtering process on an initially flat surface of size $L^2$ with periodic
315: boundary conditions, by starting an ion at
316: a random position in a plane parallel to  the initial 
317: surface, and projecting it along a straight trajectory inclined at
318: angle $\theta$ to the direction perpendicular to the averaged surface
319: configuration and at an azimuthal angle
320: $\phi$. An ion penetrates the solid through a
321: depth $a$ and releases its energy according to Eq.\ (\ref{eq:GE}). 
322: An atom at a position $(x,
323: y, h)$ is eroded (see Fig.\ \ref{fig:Model}) with probability
324: proportional to $E({\bf r'})$ 
325: given in Eq.\ (\ref{eq:GE}). It should be noted that,
326: in accordance with the assumptions of the theoretical models 
327: \cite{Makeev02, Bradley88, Cuerno95}, this
328: sputtering model neglects evaporation, redeposition of eroded
329: material and preferential sputtering of surface material at the point of
330: penetration. The surface is defined by a single valued, discrete time
331: dependent SOS height function $h({\bf r},t)=h(x, y, t)$.  
332: The time t is measured in terms of the
333: ion fluence; i.e, the number of incident ions per two-dimensional 
334: lattice site $(x,y)$. The choice of the parameters
335: $\sigma,\mu$ and $\theta$ is discussed below.
336: Substrate rotation, at constant angular
337: velocity $\omega\ge \nu^2/D$, is implemented by keeping the substrate static and
338: then rotating  
339: the ion beam instead, which is equivalent to keeping the ion beam direction fixed 
340: and rotating the substrate \cite{Bradley_Cirlin96}. In our
341: simulations we choose
342: $\phi$ randomly 
343: from the interval $0 \le \phi < 2\pi$, since, in the large $\omega$ limit the solid appears  
344: to be sputtered from all angles $\phi$.\cite{Bradley96}
345: 
346: 
347: \begin{figure}[htb]
348:  \begin{center}
349:  \includegraphics[width=0.8\columnwidth]{model.eps}
350:  \end{center}
351:  \caption{
352:  \label{fig:Model}
353: The model consists of a square field  of discrete height
354:   variables $h(x,y)$, corresponding to piles of $h(x,y)$ particles at position
355:   $(x,y)$. Here a projection to the $h-x$ is shown. 
356: {\em Left:} Each ion impact is modeled by an distribution describing
357:   the energy deposited by the ion. Atoms on the surface are removed
358:   with a probability proportional to the energy. {\em Right:} Surface
359:   diffusion; by decreasing height differences the energy is decreased.
360: }
361:  \end{figure}
362: 
363: 
364: Our model of the sputtering mechanism sets the time scale of the
365: simulation in a way, which allows direct comparison with experiments. 
366: Any relevant surface diffusion
367: mechanism may be combined with this sputtering model. \cite{Thesis}
368: Here, we use a realistic solid-on-solid model of thermally activated surface
369: diffusion \cite{Smilauer93}. 
370: Surface diffusion is simulated as a nearest neighbor
371: hopping process with an Arrhenius hopping rate
372: \begin{equation} 
373: \label{eq:arrhenius}
374: R(E, T)=R_0\exp(-E/k_BT),
375: \end{equation}
376: where $R_0=k_BT/\hbar$, and $T$ is the {\em effective} substrate
377: temperature (see below).
378:  The 
379: energy barrier $E=E_{vn}+n_nE_{ln}+E_{se}$ consists of a
380: substrate term ($E_{vn}=$0.75 eV), a nearest neighbor term
381: ($E_{ln}=$0.18 eV), and a step barrier term ($E_{se}=$0.15 eV), which
382: only contributes in the vicinity of a step edge, and is zero
383: otherwise.  
384: In each diffusion sweep all
385: surface particles are considered, and those that are not fully
386: coordinated may hop to a neighboring site. Note
387: that we have to use a higher effective temperature \cite{Yewande05}
388: in our simulation
389: in order to account for the greatly enhanced surface diffusion due to {\em thermal spikes}. 
390: A thermal spike is a series of sharp peaks and
391: minima in the spatio-temporal distribution of the surface temperature,
392: arising from the occurence of local heating of the surface right after every
393: ion impact, followed by rapid cooling. Hence, we have used a higher
394: effective temperature $k_BT=$0.1 eV, which was estimated in our
395: previous work and which roughly corresponds to experimental sputtering at room
396: temperature. 
397: 
398: The model captures the essential features of sputtered surfaces at
399: nanometer lengthscales; especially nonlinear effects
400: \cite{Hartmann02, Yewande05}.    
401: Although a direct mapping of this model to the continuum equations is
402: unavailable, the model is expected to be consistent with variants of the KS
403: equation. \cite{Yewande06, Lauritsen96} 
404: 
405: \section{\label{sec:rot_effect}EFFECT OF ROTATION ON THE 
406: TOPOGRAPHIES} 
407: 
408: 
409: In this study, a lattice of linear size  $L$ =
410: 128 is used with a lattice spacing corresponding roughly to a distance
411: of 0.5 nm. The simulated time is set such that
412: 1 ion/atom corresponds to an ion fluence of
413: 3$\times$10$^{14}$ ions/cm$^2$.  
414: We use a sputter yield of about 7 surface atoms/ion, as
415: compared to 5 SiO$_2$ molecules/ion (reduced to 0.1 SiO$_2$
416: molecules/ion with H ion) and 0.3-0.5 molecules/ion in the experiments
417: of Refs.\ \onlinecite{Mayer94} and \onlinecite{Habenicht99}
418: respectively; this may result in lengthscales differing from
419: those of the cited experiments, but we found in our previous studies 
420: \cite{Hartmann02,Yewande05,Yewande06} that predicted  
421: universal features exist, which are directly comparable to
422: experimental results. We will subsequently discuss such features for
423: rotated samples.  
424: 
425: In Ref.\ \onlinecite{Yewande06} six regions with different topographies
426: (including e.g, smooth, hole, and dot regions) were shown to emerge for
427: ion collisional parameters $a$ = 6, 0 $\le \sigma \le$ 5, 0 $\le \mu
428: \le$ 5 (all measured in units of lattice spacings), 
429: at time $t$ = 3 ions/atom. In this section we study
430: the effects of sample rotation 
431: on these topographies, in particular  effects on the region with dot topography.
432: The differences in topographies between rotated and unrotated samples
433: are visualized in Fig.\
434: \ref{fig:phase_profiles} and Fig.\
435: \ref{fig:phase_profiles_rot}, for $a$ = 6, $t$ = 3, and $\theta$ =
436: 50$^\circ$. 
437: 
438: \begin{figure}[!htbp]
439: \begin{center}
440: \includegraphics[width=0.48\textwidth]{phase.eps}
441: \caption[Phase profiles,
442: t=3]{\label{fig:phase_profiles}  Possible topographies of the model,
443:   within the experimental constraints considered in Ref.\
444:   \onlinecite{Yewande06}. $t =$ 3, $a =$ 6, $\theta =$
445:   50$^\circ$. Left - right columns: $\sigma$ = 1, 3, and 5,
446:   respectively.   
447: Bottom row - top row: $\mu$ = 0.5, 1, 1.5, 2, and 5, respectively. 
448: The last two profiles of the top row belong to the dot
449: region (region V) of Ref.\ \onlinecite{Yewande06}.}
450: \end{center}
451: \end{figure}
452: 
453: \begin{figure}[!htbp]
454: \begin{center}
455: \includegraphics[width=0.48\textwidth]{phase_rot.eps}  
456: \caption[Concurrent sputtering and rotation (PSSR),
457: t=3]{\label{fig:phase_profiles_rot}  Profiles obtained from simultaneous 
458:   sputtering and rotation, using the same parameters as in Fig.\
459:   \ref{fig:phase_profiles}. $t =$ 3, $a =$ 6, $\theta =$ 50$^\circ$.  
460: Left - right columns: $\sigma$ = 1, 3, and 5,
461: respectively.   
462: Bottom row - top row: $\mu$ = 0.5, 1, 1.5, 2, and 5, respectively. 
463: The last two profiles of the top row belong to the dot
464: region (region V) of Ref.\ \onlinecite{Yewande06}. Structure factors
465: of the lettered profiles are provided in Fig.\ \ref{fig:Sk_rot} [S -
466: (relatively) smooth; H - hole; N - non-oriented structures; D - dot].} 
467: \end{center}
468: \end{figure}
469: 
470: As seen from Fig.\ \ref{fig:phase_profiles_rot} (and Fig.\
471: \ref{fig:Sk_rot}, see below), no anisotropy can be found with substrate 
472: rotation, as expected from the continuum theory. The ripple structures obtained
473: for $\mu \le$ 2 (Fig.\ \ref{fig:phase_profiles}) do not appear for rotated
474: substrates. The underlying parallel ripples of the dot
475: region (topmost row of Fig.\ \ref{fig:phase_profiles}) are also
476: absent for rotated substrates. However, hole formation is not 
477: suppressed,  
478: we get holes  with as without rotation as is visible  in Fig.\
479: \ref{fig:phase_profiles}.
480: This fact can be understood from the continuum theory,
481: which predicts roughly equal 
482: erosion rates along both directions for parameters in the hole
483: region,\cite{Yewande06} hence there is no anisotropy to be 
484: destroyed. Furthermore, 
485: ripple patterns perpendicular with resepct to the ion beam direction are replaced by  
486: non-oriented structures, and the ordered parallel ripples are 
487: no longer present if the substrate is rotated.  
488: (see Fig.\ \ref{fig:phase_profiles_rot}), 
489: 
490: For a closer inspection, we calculate the 
491: structure factors, $S({\bf k}, t)=|h({\bf k}, t)|^2$  from the Fourier
492: transform $h({\bf k}, t)$ of the height field $h({\bf r},t)$. In
493: particular we consider four prototypical topographies marked by
494: letters  S, H, N, D in Fig.\ 
495: \ref{fig:phase_profiles_rot}.  S stands for ``relatively smooth'', H for ``hole'', N
496: for ``non-oriented structures'', and D for ``dots''. The results are shown in Fig.\
497: \ref{fig:Sk_rot}. As can
498: be seen from this figure, and as expected, there is no anisotropy
499: visible in
500: all cases. In the 
501: case of the relatively smooth surface S, there is also no 
502: characteristic lengthscale. For the hole
503: topography, H, there is still no specific lengthscale but there now
504: exists an upper bound $k_{ub}$ on $|{\bf k}|$ due to the presence of
505: the holes. On the surface with non-oriented
506: structures (N) a well defined lengthscale with $k_{ub}$ as well as a lower
507: bound $k_{lb}$ can be found. And finally, in the case of the dot topography (D), we
508: also have a characteristic lengthscale, but $k_{lb}$ is shorter here
509: than for the N topography, which implies
510: that the average separation of the dots is larger than that of the
511: non-oriented structures, as expected from Fig.\
512: \ref{fig:phase_profiles_rot}.
513: 
514: \begin{figure}[!htbp]
515: \begin{center}
516: \includegraphics[width=0.5\textwidth]{2DSk_rot.eps} 
517: \caption[Structure factor]{\label{fig:Sk_rot}  Structure factor
518:   of the lettered surface profiles in Fig.\
519:   \ref{fig:phase_profiles_rot}. S $\Rightarrow$ 
520: (relatively) smooth; H $\Rightarrow$ hole; N $\Rightarrow$
521: non-oriented structures; D $\Rightarrow$ dot. ${\bf k}={\bf 0}$ at 
522: the centre; and ${\bf k}=\frac{2\pi}{8}(\pm 1, \pm 1)$ at the corners.}
523: \end{center}
524: \end{figure}
525: 
526: 
527: \begin{figure}[!htbp]
528: \begin{center}
529: \includegraphics[width=0.99\columnwidth]{iso_coeff.eps} 
530: \caption[Coefficients of the iKS equation for the rotated
531: case]{\label{fig:iso_coeff}  The coefficients, $\nu_{av}$ and 
532:   $\chi_{av}$, of the isotropic version of Eq. \ref{eq:KS}, 
533:   \cite{Bradley96} for the  
534: rotated case, as functions of $\mu$. (a) $\sigma =$
535:  1.0, (b) $\sigma =$ 3.0, and (c) $\sigma =$ 5.0.} 
536: \end{center}
537: \end{figure}
538:  
539: According to the continuum theory, there exists a single effective surface tension 
540: coefficient $\nu_{av}$ = $\nu_x$ + $\nu_y$ and a single nonlinear
541: coupling\cite{Bradley96} $\chi_{av}$  for all directions  in rotated samples,
542: since there is no anisotropy left in the system. 
543: Fig.\ \ref{fig:iso_coeff} shows a plot of these coefficients based
544: upon the explicit expressions in Ref.\ \onlinecite{Makeev02}. 
545: As can be seen from Fig.\ \ref{fig:iso_coeff} and Eq.\ (\ref{eq:KS}) 
546: the surface roughens with
547: time, with smaller  
548: $\nu_{av}$ ($\nu_{av}<$0) corresponding to higher roughness. 
549: Surfaces in the parameter range for which 0$> \nu_{av} \gg$ -1, are
550: relatively smooth (S) topographies.
551: 
552: \begin{figure}[!htbp]
553: \begin{center}
554: \includegraphics[width=0.4\textwidth]{roughness_nodot.eps} 
555: \caption[Roughness as a function of collision cascade
556: parameters]{\label{fig:roughness}  Surface roughness, $W$, with dots
557:   excluded. Main plot: $W$ as a function of $\mu$ for $\sigma=$ 5. 
558:   Inset: $W$ as a function of $\sigma$ for $\mu=$ 5. $t=$ 3 ions/atom.   
559: } 
560: \end{center}
561: \end{figure}
562: 
563: Considering Fig.\ \ref{fig:iso_coeff}, one sees that  
564: $|\nu_{av}|$ first increases, in accordance with
565: the increasing height difference on the  
566: greyscale charts on the profiles of Fig.\
567: \ref{fig:phase_profiles_rot} and then it decreases as we
568: tend towards the dot region. These changes 
569: in the roughness are not visible on the greyscale charts due to the
570: appearance of the dots, which are considerably higher
571: than an average surface protrusion in the dot-free profiles. 
572: Therefore we have also studied the surface roughness, with the dots
573: excluded, as shown in Fig.\ \ref{fig:roughness} (Details of our 
574: dot-isolation method are discussed in the next section).  In
575: this figure, the 
576: roughness $W$ as a function of $\mu$ ($\sigma=$ 5) is 
577: shown in the main 
578: plot, where the roughness first increases and than decreases again,
579: in accordance with the continuum theory. The inset shows a plot of $W$ versus
580: $\sigma$, for $\mu=$ 5. 
581:  
582: \begin{figure}[!htbp]
583: \begin{center}
584: \includegraphics[width=0.3\textwidth]{rough_tevol_rot.eps} 
585: \caption[Time evolution of rough surface with sample
586: rotation]{\label{fig:rough_tevol_rot}  Time evolution of the
587:   relatively smooth topography of Fig.\ \ref{fig:phase_profiles} 
588:   with rotation. $\sigma =$ 3, $\mu =$ 0.5. 
589: (a)-(d): $t =$ 3, 40, 90, and 150, respectively.} 
590: \end{center}
591: \end{figure}
592: 
593:  When the local surface slopes become significant (with prolonged
594: sputtering), nonlinearities become relevant. It has been shown that for $\theta =$ 0 crossover to the nonlinear regime either gives
595: rise to dot formation (if $\chi_{av} >$ 0) or hole formation (in case
596: $\chi_{av} <$ 0)   if 
597: ion-induced effective surface diffusion is the dominant relaxation
598: mechanism  \cite{Kahng01}. This is consistent with
599: our results for parameters at which
600: $\chi_{av} >$ 0 in Fig.\ \ref{fig:iso_coeff} (i.e, for $\mu \gtrsim$
601: 3). We also found holes for $\chi_{av} <$ 0 (``H'' region), but hole formation for long times is not as widespread as
602: Fig.\ \ref{fig:iso_coeff} seems to indicate.
603: In particular, the hole topography eventually evolves into cellular
604: structures similar to those shown in
605: Fig.\ \ref{fig:rough_tevol_rot}  at long times. 
606: %In 
607: %Ref.\ \onlinecite{Yewande06} the relatively smooth surfaces with same
608: %parameters as in Fig.\ \ref{fig:rough_tevol_rot} were shown (without
609: %rotation) to evolve to a hole topography, which is consistent with
610: %this evolution of the hole topography found with rotation into similar  
611: %cellular structures as in Fig.\ \ref{fig:rough_tevol_rot}.
612: Since $\nu_{av} \ne$ 0, the surface roughening is not 
613: wavelength independent, which explains the presence of the non-oriented protrusions. 
614: 
615: For very small longitudinal and lateral straggle, $\sigma\le$
616: 1, $\mu\le$ 0.5, i.e. in the ``S'' region, 
617: we did not find any structure up to the longest simulation times. 
618: Note that with increasing $\sigma$, the $|\nu_{av}| \approx$ 0
619: interval is reduced.
620: 
621: For non-oriented structures ``N'', simulations at longer times reveal
622: only slight  changes in the structures; no dot formation (see 
623: Fig.\ \ref{fig:ripple_tevol_rot}) appears.
624: Hence, we have only observed dots with rotation wherever they can be
625: found without rotation.
626: 
627:  \begin{figure}[!htbp]
628: \begin{center}
629: \includegraphics[width=0.3\textwidth]{ripple_tevol_rot.eps} 
630: \caption[Time evolution a non-oriented structure with sample
631: rotation]{\label{fig:ripple_tevol_rot}  Time evolution of the 
632:   non-oriented structures  arising from rotation of rippled region 
633:   of Fig.\ \ref{fig:phase_profiles}. $\sigma$ = 3, $\mu$ = 1.5. (a) -
634:   (d): $t$ = 3, 40, 
635:   90, and 150, respectively.} 
636: \end{center}
637: \end{figure}
638: 
639: %But this figure (i.e, the positive
640: %$\nu_{av}$ in Fig.\  
641: %\ref{fig:iso_coeff}) also imply that the sputtering process itself no
642: %longer maximizes the  
643: %exposed surface area, but instead, tends to minimize it.
644: 
645:  \begin{figure}[!htbp]
646: \begin{center}
647: \includegraphics[width=0.5\textwidth]{tasd_a6_sig1_mu2_tm3diff_theta_rot.eps} 
648: \caption[Topographic transitions for $\sigma=$ 1, $\mu=$ 2 and
649: 5]{\label{fig:diff_theta}  Topography changes as a result of
650:   changing the angle of incidence $\theta$, for $\sigma=$ 1; $\mu=$ 2
651:   (top row), and 5 (bottom row). Top row, L-R: $\theta=$ 10, 30, 40,
652:   50, 80. Bottom row, L-R: $\theta=$ 5, 10, 30, 50, 80.} 
653: \end{center}
654: \end{figure}
655: 
656: When considering the emerging topographies of Fig.\
657: \ref{fig:phase_profiles_rot} at other angles of incidence, we found
658: notable changes with $\theta$, as illustrated in Fig.\ \ref{fig:diff_theta} for $\sigma=$ 1, $\mu=$ 2 and $\sigma=1$, $\mu=5$). 
659: That is, changes from
660: smooth $\to$ 
661: hole $\to$ non-oriented structure, and back to smooth (no structure)
662: topographies appear. These changes imply that the roughness 
663: initially increases and then starts to decrease with 
664: increasing $\theta$ as shown in
665: Fig.\ \ref{fig:rough_diff_theta}. In this figure $\sigma$ = 1 and data
666: for $\mu$ = 2 and $\mu=5$ are represented by  
667:   (black) circle and (red) square symbols, respectively. The 
668:   behavior of the roughness with 
669: varying $\theta$ shown in Fig.\ \ref{fig:rough_diff_theta} is in
670: agreement with the experiments reported for Ar-ion 
671: sputtered rotating InP surfaces in Ref.\
672: \onlinecite{Frost00}. But note that the roughness data reported in this
673: experiment \cite{Frost00} were obtained for an exposure time of 1200
674: secs., i.e.\, in the steady state regime. To use such steady state
675: values in Fig.\ \ref{fig:rough_diff_theta} would require simulation
676: times beyond our computational time constraints.
677: 
678: \begin{figure}[!htbp]
679: \begin{center}
680: \includegraphics[width=0.4\textwidth]{rough_theta_sig1mu2_mu5.eps} 
681: %{roughness_nodot.eps}
682: \caption[Roughness as a function of collision cascade
683: parameters]{\label{fig:rough_diff_theta}  (Color online) Surface
684:   roughness      
685:   $W$ as a function of $\theta$ for profiles shown in Fig.\
686:   \ref{fig:diff_theta}. Data for $\mu$ = 2 and 5 are represented by
687:   circle and square symbols, respectively. In both cases $\sigma$ = 1,
688:   $t$ = 3 ions/atom. Error bars are included, but they are much smaller than
689:   symbol size except at $\theta=10^{\circ}$ ($\mu=5$). For this case
690:   for about half of the runs the structure had already evolved to a
691:   non-oriented structure pattern with a considerable roughness, 
692: while the other half exhibited still  a rather smooth surface. Hence,
693: there seems to be a sharp transition in time from smooth to rough,
694: where the transition time of finite samples fluctuates strongly.
695: } 
696: \end{center}
697: \end{figure}
698: 
699: \section{\label{sec:analysis}NUMERICAL ANALYSIS OF THE DOTS}
700: 
701: \begin{table}
702: \caption{\label{tab:SRIM}SRIM results for materials bombarded with
703:   noble gas ions.}
704: \begin{ruledtabular}
705: \begin{tabular}{llllll}
706: Ion & Material & E/keV & $a$ \footnotemark[1]\footnotetext[1]{in
707:   lattice units} & $\sigma$ \footnotemark[1] & $\mu$ \footnotemark[1] \\
708: Ar & GaAs & 1.6 & 6 & 5 & 3.8 \\
709: Ar & Ge & 1.6 & 6 & 5 & 3.8 \\
710: Ne & Si & 0.65 & 6 & 4.4 & 3.2 \\
711: \end{tabular}
712: \end{ruledtabular}
713: \end{table}
714: 
715: Focussing on topographic region ``D'' where dots have
716: been found also without 
717: rotation, we obtained the experimental parameters corresponding to
718: this region from SRIM simulations \cite{srim} as shown in Table
719: \ref{tab:SRIM}. We have performed two sets of simulations using the
720: SRIM results: (i) 1.6 keV Ar ion sputtering of GaAs surfaces; where  
721: $a$ = 6, $\sigma$ = 5, $\mu$ = 3.8. (ii) 650
722: eV Ne ion sputtering of Silicon surfaces; where $a$ = 6, $\sigma$ =
723: 4.4, $\mu$ = 3.2. In both cases we have used 
724: an ion beam inclination of $\theta=$50$^\circ$ to the vertical. 
725: 
726: In order to study the time evolution of dot characteristics
727: (e.g. dot density, area, and height), we have used a similar clustering
728: approach as in Ref.\ \onlinecite{Yewande05}. A dot is defined as a
729: cluster of points of local height maxima.  This is done in two steps
730: \begin{itemize}
731: \item For any given time
732: $t$ (we do not mention the time-dependence explicitly here)
733: we consider the set of points 
734: \begin{equation}
735: M\equiv \{(x, y)|x, y \in (1, \cdots, L); h(x,y)\ge h_c \},
736: \end{equation}    
737: where $L$ is the linear size of the system,  $h_c$
738: is a cutoff height which the surface height at a point must equal/exceed for
739: the point to be counted as (or part of) a dot. \cite{Yewande05} 
740: We define $h_c$ to be a function of the average surface protrusion, which
741: has the form: $h_c=h_{min}+p(\langle h\rangle-h_{min})$. Where
742: $h_{min}$ is the lowest surface height, $\langle h\rangle$ is the average
743: surface height, and $p$ is a fixed percentage. 
744: \item  We call two
745:   points in $M$  {\em neighbors} if their distance is smaller than a given
746:   threshold $d_c$, we use $d_c=1$ here. Then, two points are
747:   called {\em connected}, if there exists a path from the 
748: first point to the other
749:   point such that
750: all consecutive points along the path are neighbors.
751: Now, each dot $D$ is a subset of $M$ (non-overlapping with
752:   any other dot) of maximum size such that all points in $D$ are mutually
753:   connected. Hence, dots are
754: the transitive closures of the neighbor relation on $M$.
755: \end{itemize}
756: 
757: 
758: We start our
759: simulations with a dot configuration obtained from a topography at $t=$ 3 ions/atom and we choose a
760: value $p=p_o$ that yields the highest number of sampled dots $N_o$ (see
761: Table \ref{tab:clustering}).         
762: The initial clusters are shown in Fig.\ \ref{fig:clusters} for simulation
763:  of Ar-ion sputtering of GaAs 
764: without substrate rotation  and with substrate rotation, respectively. 
765:  The dot cross-section $A(D)$ is defined $A(D)=|D|$; i.e the cardinality of
766: $D$. The average height $h_d$ of a single dot $D$ is defined as  
767: $h_d (D) = \frac{1}{|D|}\sum_{(x,y)\in D} h_{x, y}$.  
768: Note that, in order to exclude non-dot surface  protrusions (see top row of
769: Fig.\ \ref{fig:dotwt}) from our analysis, we have included an upper
770: boundary of 50 cluster points (i.e $A(D)\le$ 50) to our definition.
771: In our 
772: analysis we only consider dots defined by these clusters.
773: % Hence, it should
774: %be noted when interpreting the results that the actual initial number
775: %of dots is not important, instead, the main focus is on how the number
776: %of sampled dots vary with sputter time/fluence.
777:  The results reported
778: here are obtained from an average of 100 independent runs.       
779: 
780:   \begin{table}
781: \caption{\label{tab:clustering} Optimal parameters used to determine
782:   the cutoff height $h_c$}
783: \begin{ruledtabular}
784: \begin{tabular}{llll}
785: Simulation & $p_o$ (\%) & $N_o$ & State \\
786: Ar$^+$ on GaAs & 50 & 69 & unrotated \\
787: Ar$^+$ on GaAs & 10 & 149 & rotated \\
788: Ne$^+$ on Si & 60 & 76 & unrotated \\
789: Ne$^+$ on Si & 20 & 139 & rotated \\
790: \end{tabular}
791: \end{ruledtabular}
792: \end{table}
793: 
794: \begin{figure}[!htbp]
795: \begin{center}
796: \includegraphics[width=0.4\textwidth]
797: {cls_tp0.1size128_a6.0_sig5.0_mu3.8_tm003.0.out.eps}
798: \caption[clusters for the non-rotated and the rotated
799: case]{\label{fig:clusters}  (Color online) 
800: Sample surface profiles for the
801:   unrotated case (top), and the rotated case (bottom) at t =
802:   3 ions/atom. In the figures on
803:   the right the clusters formed from the corresponding profile on the
804:   left (as defined in the text) are printed on top of the profile. 
805:   The scales indicate the surface height measured 
806:   from the lowest height. 
807: }
808: \end{center}
809: \end{figure}
810: 
811: 
812: In Fig.\ \ref{fig:dotwt}, we show sample surface profiles without
813: rotation (top row), and with rotation (bottom row), for simulation of
814: 1.6 keV Ar-ion sputtering of GaAs. As can be seen from the top row of
815: this figure, the ripples (with parallel orientation to ion beam
816: direction) that coexist with the dots become more ordered with time,
817: whereas the dots decrease in number with time; more analysis of these
818: dots is provided below. On the other hand (bottom row of Fig.\
819: \ref{fig:dotwt}), these ripples do not exist when the substrate is
820: subjected to concurrent rotation, and the density of the dots is more
821: uniform. 
822: 
823:  In Fig.\ \ref{fig:dotcharNH}, we present results of the average
824: number of sampled dots N$_c$, and the average dot height
825: H$_c=\overline{h_d}$, where the average is taken over all dots and all
826: independent runs. The
827: results of the average area of cross-section of the dots is presented
828: in Fig.\ \ref{fig:dotcharA}. 
829:  In Figs.\
830: \ref{fig:dotcharNH} and \ref{fig:dotcharA}, open and closed circle
831: symbols represent data obtained from Ar-GaAs with and without rotation
832: respectively; while open and closed triangle symbols represent data
833: obtained from Ne-Si with and without rotation 
834: respectively.  
835: The main result is that we found
836: power law scaling of the dot characteristics with time; with or
837: without rotation. In addition, we found the same scaling behavior for
838: the two sets of simulation performed; i.e, 1.6 keV Ar ion sputtering
839: of GaAs (Ar-GaAs), and 650 eV Ne ion sputtering of Si (Ne-Si).
840: 
841: \begin{figure}[!htbp]
842: \begin{center}
843: \includegraphics[width=0.5\textwidth]
844: {dots_difft_tp0.1size128_a6.0_sig5.0_mu3.8.eps} 
845: \caption[time evolution of the dots the non-rotated and the rotated
846: case]{\label{fig:dotwt}  Surface profiles for the
847:   non-rotated (top row) and the rotated (bottom row) case. L - R, t =
848:   10, 20, and 30 ions/atom respectively. The bar indicates the ion
849:   beam direction, and the scales indicate the surface height measured
850:   from the lowest height.  
851: }
852: \end{center}
853: \end{figure}
854: 
855: To be more specific, 
856: as can be seen from Fig.\ \ref{fig:dotcharNH} (a), the average number
857: of sampled dots 
858: decreases with time as N$_c\sim t^{-\psi}$ for non-rotated samples, 
859: where $\psi=$ 0.583$\pm$0.007, while it stays approximately constant
860: for rotated samples ($\psi\approx 0$). 
861: This indicates that the number of dots becomes insignificant without substrate
862: rotation, so that one should use
863: rotation if dot creation is the main purpose of the sputtering.
864: Note that this result is already visible in 
865: Fig.\ \ref{fig:dotwt} in qualitative form.
866: %, where
867: %the time evolution of the surface profiles for Ar-GaAs, without (top
868: %row) and with (bottom row) substrate rotation, is shown; for simulation
869: %times 10, 20, and 30 ions/atom (note that the profile for $t$ = 3
870: %ions/atom has been shown in Fig. \ref{fig:clusters}).     
871:    
872: \begin{figure}[!htbp]
873: \begin{center}
874: \includegraphics[width=0.4\textwidth]
875: {clusterNH100r_tp0.1size128hco2.eps}
876: \caption[time evolution of the dot
877: characteristics]{\label{fig:dotcharNH}  (Color online)  
878: Time evolution of (a) the average number N$_c$, and, (b) the average
879: height H$_c$ (both in lattice units), of 
880: the sampled dots with time (in ions/atom). Symbols: open and closed
881: circle (triangle) symbols denote data obtained 
882: with and without substrate rotation respectively, for Ar-GaAs (Ne-Si).
883: }
884: \end{center}
885: \end{figure}
886: 
887: On the other hand, as shown in
888: Fig.\ \ref{fig:dotcharNH} (b), the average height
889: H$_c$ increases with time as H$_c\sim t^\varrho$, the increase is more
890: rapid for the unrotated case ($\varrho=$0.409$\pm$0.004) than for the
891: rotated case ($\varrho=$0.159$\pm$0.007). As expected, the dot height
892: is lower with sample rotation due to the enhanced smoothening effect
893: of the rotation. \cite{Bradley_Cirlin96, Bradley96} This smoothening
894: also explains the higher initial number of dots for the rotated case
895: [Fig.\ \ref{fig:dotcharNH} (a)]: 
896: The dots that are
897: not ``visible'' in the unrotated case, since they do not surmount the ripples,
898: become ``visible'' in the rotated case since the rotation
899: prevents ripple formation.   
900: 
901: The mean cross sectional area A$_c$ finally becomes almost time independent for
902: both cases. However, as shown in Fig. \ref{fig:dotcharA} , an initial
903: power law scaling A$_c\sim t^{-\varsigma}$ occurs, which is more
904: significant for the rotated case (where $\varsigma$ = 0.113$\pm$0.006)  
905: than for the unrotated case (where $\varsigma\approx$ 0). 
906: 
907: \begin{figure}[!htbp]
908: \begin{center}
909: \includegraphics[width=0.4\textwidth]
910: {clusterA100r_tp0.1size128hco2.eps}
911: \caption[time evolution of the dot area]{\label{fig:dotcharA} 
912:   (Color online)  
913: Time evolution of the average area of crosssection A$_c$ of the sampled dots
914: with time for (a) the unrotated case, and (b) the rotated
915: case. Symbols: open and closed circle (triangle) symbols denote data obtained 
916: with and without substrate rotation respectively, for Ar-GaAs (Ne-Si).
917: }
918: \end{center}
919: \end{figure} 
920: 
921: These results indicate that while no new dots are created with
922: rotation (for $\theta >$ 0$^\circ$), the uniformity of the density of
923: the existing dots and the stability of the dot height are greatly
924: enhanced with substrate rotation. Our results of the analysis of the
925: dots are in agreement with previous experiments on dots
926: obtained from normal incidence sputtering of semiconductors (Si),
927: \cite{Gago01} 
928: semiconductor alloys (GaSb, InSb); \cite{Facsko99, Facsko01} as well
929: as oblique incidence dots 
930: obtained from simultaneous sputtering and sample rotation
931: (InP). \cite{Frost00} In these experiments, the dot height has been
932: reported to increase with time; and the average dot size have
933: been reported to become constant with time.  
934: 
935: To summarize, we have implemented substrate rotation for a
936: solid-on-solid model of surface sputtering and used it to study the
937: effects of concurrent rotation on the different possible
938: topographies. In particular, we have studied the effect of rotation on
939: the dot region as well as a detailed analysis of the time evolution of
940: the dot characteristics (number, cross sectional area, and height)
941: with and without substrate rotation. We found that different materials
942: whose sputtering parameters fall within this region exhibit the same
943: scaling behavior. The number of dots $N_c$ formed in the absence of
944: substrate rotation decrease with time as $N_c\sim t^{-0.58}$, whereas
945: $N_c$ is roughly constant with substrate rotation. Both with and without rotation
946: the dot cross section $A$ finally becomes independent of time, however, it
947: initially decreases according to $A\sim t^{-0.11}$ with rotation.    
948: 
949: Additionally, for other choices of the sputtering conditions, we find
950: different patterns which have not been observed experimentally so
951: far. In particular we found transitions in time from one kind of
952: surface structures (e.g. smooth, or holes) to other structures (like
953: non-oriented structures), which can be explained only by the presence
954: of non-linear effects. Hence, more sputtering experiments with different
955: ion/substrate types and varying parameters are needed to
956: verify whether the structures we predict by simulations can indeed be
957: found experimentally.
958: 
959: \begin{acknowledgments}
960: This work was funded by the German research association, the Deutsche 
961: Forchungsgemeinschaft (DFG), within the Sonderforchungsbereich (SFB)
962: 602: {\it Complex Structures in Condensed Matter from Atomic to
963:   Mesoscopic Scales}. A. K. H acknowledges financial support from the
964: Volkswagenstiftung within the program ``Nachwuchsgruppen an
965: Universit\"aten".  
966: \end{acknowledgments}
967: 
968: 
969: 
970: 
971: \bibliography{pap}% Produces the bibliography via BibTeX.
972: 
973: \end{document}
974: 
975: %
976: % ****** End of file apssamp.tex ******
977: