1: \documentclass[12pt]{article}
2: \textwidth 16cm
3: \textheight 21.5cm
4: \oddsidemargin 4mm
5: \evensidemargin 4mm
6: \parindent 1cm
7: %%%%\renewcommand{\baselinestretch}{1.5}
8: \usepackage{epsfig}
9: \usepackage{graphicx}
10: %%%%
11: \newcommand{\bm}[1]{\mbox{\boldmath $#1$}}
12: \newcommand{\be}{\begin{equation}}
13: \newcommand{\bE}{{\bf e}}
14: \newcommand{\ee}{\end{equation}}
15: \newcommand{\sq}{S({\bf q})}
16: \newcommand{\bx}{{\bf x}}
17: \newcommand{\bj}{{\bf j}}
18: \newcommand{\bq}{{\bf q}}
19: \newcommand{\bu}{{\bf u}}
20: \newcommand{\bk}{{\bf k}}
21: \newcommand{\bF}{{\bf f}}
22: \newcommand{\bFF}{{\bf F}}
23: \newcommand{\eq}[1]{(\ref{#1})}
24: \newcommand{\fig}[1]{Fig.~\ref{#1}}
25: \newcommand{\hpm}{h^{\prime}}
26: \newcommand{\hpdm}{h^{\prime \prime}}
27: \newcommand{\cit}[1]{[\ref{#1}]}
28:
29: \renewcommand{\baselinestretch}{1.5}
30: %\setlength{\topmargin}{10.5cm}
31:
32: \begin{document}
33:
34: \title{\bf Strain induced stabilization of stepped Si and Ge surfaces near (001)}
35:
36: \author{ V.~B.~Shenoy, C.~V.~Ciobanu and L.~B.~Freund \\
37: Division of Engineering, Brown University, Providence, RI
38: 02912}
39:
40: \date{\small \today}
41:
42: \maketitle
43:
44: \begin{abstract}
45:
46: We report on calculations of the formation energies of several [100] and [110]
47: oriented step structures on biaxially stressed Si and Ge (001) surfaces. It is
48: shown that a novel rebonded [100] oriented single-height step is strongly
49: stabilized by compressive strain compared to most well-known step structures. We
50: propose that the side walls of ``hut''-shaped quantum dots observed in recent
51: experiments on SiGe/Si films are made up of these steps. Our calculations
52: provide an explanation for the nucleationless growth of shallow mounds, with
53: steps along the [100] and [110] directions in low- and high-misfit films,
54: respectively, and for the stability of the (105) facets under compressive strain.
55: \end{abstract}
56:
57: %\pacs{46.65.+g, 62.30.+d}
58: \newpage
59: Strain induced self-assembly is actively being pursued as a technique for the
60: fabrication of nanoscale electronic devices and memories that have the potential
61: to bring higher speed to information processing and higher areal and volumetric
62: capacity to information storage. In the past few years, significant advances
63: have been made on both the technological and fundamental aspects of self
64: assembly. On the technological side, it has become possible to prepare regular
65: spatial arrays of nanostructures.
66:
67: Several basic issues concerning the
68: physical mechanisms involved in the different stages of the formation and
69: morphological evolution of the nanostructures are only now becoming clear. The initial stages of
70: epitaxial nanostructure formation remains least well understood. Of particular interest here, recent experimental work [\ref{sutter2000}-\ref{rastelli2001}] has
71: revealed that (105) oriented quantum dots in SiGe/Si films grow from shallow
72: precursor mounds whose sidewalls are made up of widely spaced steps that are
73: oriented in the [100] or [110] directions for low or high Ge compositions,
74: respectively. These observations cannot be understood on the basis of a
75: competition between surface energy increase involved in creating the walls of the
76: dots and the strain energy reduction through elastic relaxation. Such a competition
77: would lead to a nucleation barrier for the formation of the islands, while the
78: experiments clearly show that the stepped mounds emerge as a natural instability
79: without any such barrier [\ref{sutter2000},\ref{tromp2000}].
80: %In this letter, we
81: %will consider the influence of strain on the formation of steps on (001) Si and
82: %Ge surfaces to understand the stabilization of different types of stepped
83: %surfaces during the early stages of strained heteroepitaxy.
84:
85: Recently we have demonstrated that nucleationless growth of stepped mounds can
86: be understood by including the physics of surface steps, in particular the
87: dependence of their formation energy on strain and their interactions
88: \cit{shenoy2002}. In this letter, we calculate the parameters that characterize step
89: formation and interactions in order to understand the stabilization of [100] or
90: [110] steps on (2$\times$1) reconstructed Si and Ge (001) surfaces at different levels of
91: biaxial strain. The strain dependence of the step formation energy is
92: determined by the disposition of the atomic bonds in the vicinity of the
93: step-edge; if there is significant rebonding such that these bonds are stretched
94: (compressed) relative to the bonds in the bulk material, a compressive (tensile)
95: mismatch stress tends to lower their formation energy \cit{xie1994}. The
96: interactions between steps include the repulsive dipolar interactions
97: \cit{marchenko1980}, which depend on the local step density, and the non-local
98: attractive interactions due to the force monopoles induced by the mismatch
99: strain \cit{tersoff1995}. The surface energy of a stepped surface that is
100: oriented at a small angle $\theta$ with respect to the (001) direction can be
101: written by including the step formation and dipolar interaction energies as
102: \be
103: \gamma(\theta,\epsilon) = \gamma_0(\epsilon)\cos\theta + (\beta_1 + \tilde{\beta}_1\epsilon)|\sin\theta| + \beta_3\frac{|\sin\theta|^3}{\cos^2\theta},
104: \label{se}
105: \ee
106: where $\epsilon$ is the biaxial surface strain, $\gamma_0$ is the surface
107: energy of the (001) surface, $\beta_1$ is the formation energy of a step,
108: $\tilde{\beta}_1$ is a measure of the sensitivity of the formation energy of a
109: step to strain and $\beta_3$ characterizes the strength of the step-step dipolar
110: interactions. Since the attractive monopole induced interactions favor a
111: stepped surface over a flat surface, stepped mounds can grow without any
112: nucleation barrier if the surface energy of their sidewalls becomes lower than
113: that of the (001) surface \cit{shenoy2002}. It can be seen from \eq{se} that this indeed happens
114: for stepped surfaces with orientations in the range $|\theta| < \theta_c$,
115: where $\theta_c \approx \sqrt{-\beta(\epsilon)/\beta_3}$, if the condition
116: $\beta(\epsilon) \equiv \beta_1 + \tilde{\beta}_1\epsilon < 0$ is satisfied.
117: In the present work we show by means of atomistic simulations
118: that this condition is satisfied for the mismatch strains of interest.
119:
120: In the [110] direction, we will only consider the double-height
121: B-type (DB) steps \cit{chadi1987}; these steps are known to be lower in energy
122: than the combination of alternating single-height SA and SB steps for vicinal
123: angles that are more than about 1$^0$ - 2$^0$. In the [100] direction we consider
124: two kinds of steps: (1) zig-zag SA+SB$^{[100]}$ steps proposed by Chadi
125: \cit{chadi1987} and experimentally observed by Wu {\em et.al.} \cit{wu1993} and
126: (2) straight steps on which the dimer rows of the adjoining terraces arrive at the
127: steps at
128: an angle of 45$^0$. If we focus attention on the family of (10n) surfaces to
129: which the (105) surface belongs, there are two distinct low-energy step
130: structures that can be obtained by removing every other atom from the step-edge
131: in order to reduce the number of dangling bonds, as shown in \fig{steps}. The key
132: difference between the two structures is that in one of them, which we call
133: single-height rebonded (SR), there are atoms at the edge of the step
134: bonded to atoms on the adjoining terraces, while this type of rebonding is absent
135: in the single-unrebonded (SU) step (refer to \fig{steps}).
136: %As we shall see, rebonding stretches the
137: % bonds at the step edge and hence the formation energy of this step is very
138: %sensitive to the presence of a mismatch strain.
139: The structure of the (105)
140: facet made up of SU and SR steps correspond to the reconstructions for this
141: surface proposed by Mo {\em et. al.} \cit{mo1990} and Kohr and Das Sarma
142: \cit{khor1997}, respectively.
143: %In order to understand the preference for
144: %[110] ([100]) oriented steps during early stages of island growth in Ge
145: %rich (poor) films, we consider the strain dependence of step formation energies
146: %in both Si and Ge using the Stillinger-Weber (S-W) \cit{stillinger1985} and
147: %Tersoff (T3) \cit{tersoff1988} potentials.
148:
149: We determine the step formation and interaction energies using the following
150: procedure: the surface energies of various stepped surfaces are first
151: determined as a function of applied biaxial strain for both the (10n) and (11n)
152: families as shown in \fig{gamma}. The step formation and interaction energies
153: are then extracted from the surface energy $\gamma(\epsilon,\theta)$ using
154: \eq{se}. The strain dependent step formation energies for various step
155: structures calculated using the Tersoff (T3) \cit{tersoff1988} potential are shown in \fig{gamma} for Si and
156: Ge \cit{monopole}. In all cases, the step formation energies can be reasonably
157: approximated by a linear relation assumed in \eq{se} with small deviations that
158: can be attributed to non-linear effects, particularly at larger strains.
159: Calculation of step energies using the Stillinger-Weber (S-W) \cit{stillinger1985} potential also showed almost linear
160: dependence of the formation energies with strain. The parameters $\beta_1$ and
161: $\tilde{\beta}_1$, obtained by using a linear fit to $\beta(\epsilon)$,
162: along with the dipolar interaction strength $\beta_3$ are given in Table I.
163:
164: If we focus attention on the [100] oriented steps in Si, we find that the
165: formation energies of the SR steps are lower than those for the SU and
166: SA+SB$^{[100]}$ steps for compressive mismatch strains. This can be understood
167: by analyzing the bond lengths at the step-edge (refer to \fig{steps}); since
168: there are several stretched bonds in the SR structure, we expect its energy to
169: be lower in compressive strains than the SU structure. Although the zig-zag
170: SA+SB$^{[100]}$ steps are energetically lower than the SR steps in the absence
171: of strain, \fig{gamma} shows that a modest amount of compressive strain (0.3 \% in Si)
172: stabilizes the SR steps over these steps. Table I also shows that while the
173: formation energies of the SR steps on an unstrained surface are larger that the
174: corresponding values for the DB steps, the strain sensitivity $\tilde{\beta}_1$
175: for the SR steps is about 50\% larger in both Si and Ge. This implies that the
176: SR steps are preferred over the DB steps at large values of compressive strain,
177: with one key difference between Si and Ge. Using both the S-W and T3 potentials,
178: we find that the formation energies of the SR steps fall below the energies for DB
179: steps when strains become more than about 1\% and about 4\% in Si and Ge,
180: respectively (refer to the insets in \fig{gamma} and Table I).
181:
182: The above observations have important implications for the growth of stepped
183: islands observed in recent experiments [\ref{sutter2000}-\ref{rastelli2001}]. If
184: we assume that the results for Si are indicative of the trends in low-misfit
185: films ($\epsilon$ $\approx$ 1\%-2\%), it is clear that the [100] SR steps should
186: be observed during early stages of island growth since their formation energies
187: are lowest among all the cases that we have considered. While we are not aware
188: of a direct experimental observation confirming this prediction, our picture is
189: consistent with the experiments of Sutter and Lagally \cit{sutter2000} who find
190: that in SiGe films with 25\% Ge content, the stepped mounds are made up of [100]
191: oriented monolayer steps. Experiments that can identify the surface structure of
192: these stepped mounds will be invaluable in resolving the issue. In distinct
193: contrast to the low-misfit films, experiments on pure Ge \cit{vailionis2000} and
194: Si-capped Ge \cit{rastelli2001} islands show that the stepped mounds are made up
195: of [110] oriented steps \cit{scount}. These observations can also be understood
196: on the basis of the formation energy of the SR and DB steps in Ge shown in
197: \fig{gamma}. Here, the SR steps do not become favorable until mismatch strain
198: becomes close to 4\% and 5\% in the T3 and S-W potentials respectively; since
199: the strains in these mounds should be less than the mismatch between Si and Ge
200: (4.2\%), the observation of [110] oriented steps are in agreement with our
201: calculations.
202:
203: While a negative step creation energy would lead to formation of surfaces with
204: closely spaced steps, repulsive interactions increase as the spacing between
205: the steps become smaller. This competition leads to an optimum slope, which
206: can be determined using \eq{se} as $\theta^* \approx
207: \sqrt{-\beta(\epsilon)/3\beta_3}$ \cit{shenoy2002}. The existence of such a slope is indeed
208: confirmed by the surface energies for Ge shown in \fig{gamma}, where at 4\%
209: strain the (105) and (115) surfaces are seen to be the optimum facets in the
210: (10n) and (11n) directions respectively. In the case of the (11n) surfaces, the
211: (115) surface corresponds to closest spacing of DB steps (the (114) and (113)
212: surfaces are reconstructed so that the steps lose their identity). In the
213: presence of compressive strains, the (105) surface with SR steps has a
214: smaller surface energy than both the high index (10n) surfaces shown in
215: \fig{gamma} and the low index (103) surface. It can also be seen from \fig{gamma}
216: that the surface energy of the (105) surface in Ge lies below the energy of
217: the (115) surface throughout the range of interest, while it falls below the
218: energy of the (117) at a strain of about 3\%. In the case of Si, we find a
219: similar trend with the (105) surface falling below the (117) surface close to
220: 1\% . These observations indicate that the (105) facet made up of SR steps is
221: stabilized by compressive strains in both low- and high-misfit SiGe films. On
222: the other hand, the (105) surface with the SU steps proposed by Mo {\em et.al.} is
223: stabilized by tensile rather than compressive stresses since the bonds at the
224: step-edge are in compression (refer to \fig{steps}). While the results in the
225: present work are obtained using empirical potentials , we have verified by means
226: of {\em ab initio} simulations that the atoms at the edges of the SR steps in
227: the (105) surface are indeed in a stretched state. The details of these calculations
228: will be published elsewhere.
229:
230: In summary, the nucleationless growth of stepped mounds can be understood on the
231: basis of strain dependence of [100] and [110] oriented steps on (001) surfaces.
232: The rebonding at the edge plays an important role in stabilizing stepped
233: surfaces in the presence of strain. The competition between repulsive step
234: interactions and strain induced lowering of step formation energies leads to
235: optimum low-energy orientations such as (105). Further work that analyzes the effect of
236: surface segregation on step formation energies will be invaluable in gaining deeper insight into early stages of quantum dot formation in SiGe and other alloys.
237:
238:
239: \vspace{1cm}
240: \noindent{\em Acknowledgments :}{\small \setlength{\baselineskip}{11pt}{ The research support of the National
241: Science Foundation through grant CMS-0093714 and the Brown University MRSEC
242: Program, under award DMR-0079964, is gratefully acknowledged. }}
243:
244: \vspace{1cm}
245: \newpage
246: \noindent{\large{\bf{References}}}
247:
248: \begin{enumerate}
249:
250: \item \label{sutter2000}
251: P.~Sutter and M.~G.~Lagally, {\it Phys. Rev. Lett.} {\bf 84}, 4637 (2000).
252:
253: \item \label{tromp2000}
254: R.~M.~Tromp, F.~M.~Ross and M.~C.~Reuter, {\it Phys. Rev. Lett.} {\bf 84}, 4641 (2000).
255:
256: \item \label{vailionis2000}
257: A.~Vailionis, B.~Cho, G.~Glass, P.~Desjardins, D.~G.~Cahill and J.~E.~Greene, {\it Phys. Rev. Lett.} {\bf 85}, 3672 (2000).
258:
259: \item \label{rastelli2001}
260: A.~Rastelli, M.~Kummer and H.~von Kanel, {\em Phys. Rev. Lett.} {\bf 87}, 6101 (2001).
261:
262: \item \label{shenoy2002}
263: V.~B.~Shenoy and L.~B.~Freund, A continuum description of the energetics and evolution of stepped surfaces in strained nanostructures, {\em J.~Mech.~Phys.~Sols.} (in press); A copy of this article can be found at {\em http://arXiv.org/abs/cond-mat/0203514}.
264:
265: \item \label{xie1994}
266: Y.~H.~Xie, G.~H.~Gilmer, C.~Roland, P.~J.~Silverman, S.~K.~Buratto, J.~Y.~Cheng, E.~A.~Fitzgerald, A.~R.~Kortan, S.~Schuppler, M.~A.~Marcus and P.~H.~Citrin, {\bf 73} 3006 (1994); C.~Roland, {\it MRS Bulletin} {\bf 21}, 27 (1996).
267:
268:
269: \item \label{marchenko1980}
270: V.~I.~Marchenko and A.~Ya.~Parshin, {\it Sov. Phys. JETP} {\bf 52}, 129 (1980).
271:
272: \item \label{tersoff1995}
273: J.~Tersoff, Y.~H.~Phang, Z.~Y.~Zhang and M.~G.~Lagally, {\it Phys. Rev. Lett.} {\bf 75}, 2730 (1995).
274:
275: \item \label{chadi1987}
276: D.~J.~Chadi, {\it Phys. Rev. Lett.} {\bf 59}, 1691 (1987).
277:
278: \item \label{wu1993}
279: F.~Wu, S.~G.~Jaloviar, D.~E.~Savage and M.~G.~Lagally, {\it Phys. Rev. Lett.} {\bf 71}, 4190 (1993).
280:
281:
282: \item \label{mo1990}
283: Y.~W.~Mo, D.~E.~Savage, B.~S.~Swartzentruber and M.~G.~Lagally, {\it Phys. Rev. Lett.} {\bf 65}, 1020 (1990).
284:
285: \item \label{khor1997}
286: K.~E.~Khor and S.~Das Sarma, {\em J.~Vac.~Sci.~Technol.~B}, {\bf 15} 1051 (1997).
287:
288:
289: \item \label{tersoff1988}
290: J.~Tersoff, {\em Phys. Rev. B}, {\bf 38}, 9902 (1988); {\em Phys. Rev. B}, {\bf 39} 5566 (1989).
291:
292: \item \label{stillinger1985}
293: F.~H.~Stillinger and T.~A.~Weber, {\em Phys. Rev. B}, {\bf 31}, 5262 (1985); K.~Ding and H.~C.~Andersen, {\em Phys. Rev. B}, {\bf 34} 6987 (1986).
294:
295: \item \label{monopole}
296: In the case of SR steps there should be an additional term in \eq{se} that accounts for interaction of the monopoles
297: due to the discontinuity of surface stress at the step-edge. We have ignored this term in our analysis of vicinal surfaces with $\theta>$ 3$^0$ since the magnitude of this term becomes significant for small vicinal angles ($\theta<$ 1$^0$).
298:
299: \item \label{scount}
300: From the STM images of the stepped mounds in \cit{rastelli2001} we have inferred that the [110] oriented steps are of the double-height type.
301:
302: \end{enumerate}
303:
304: \newpage
305: \begin{table}[h]
306: {Table I: Step formation energies $\beta_1$, their derivatives with respect to
307: strain $\tilde{\beta}_1$, and
308: the step-step dipolar interaction coefficient $\beta_3$
309: for Si and Ge (with values for Ge enclosed in parentheses) calculated using the Stillinger-Weber (S-W) and Tersoff (T3) potentials. In order to allow a direct comparison of the formation energies of different step structures, $\beta_1$ is measured relative to the formation energy of the DB step calculated using the T3 potential. All quantities are expressed in meV/ \AA$^2$.}
310: \begin{center}
311: \begin{tabular}{|l| l r|| l r|| c|| c | }
312: \hline \hline
313: &\ \ \ \ \ \ \ \ \ \ \ SR& &\ \ \ \ \ \ \ \ \ \ \ DB & & SU & SA+SB$^{[100]}$ \\
314: & S-W & T3 & S-W & T3 & T3 & T3 \\
315: \hline \hline
316:
317: $\beta_1$ &19.9(29.9)&5.8(10.1)&12.1(14.8)&0.0(0.0)& 46.3(38.6)& 5.98(0.96) \\
318: $\tilde{\beta}_1$ &741(738) &1046(781)&440(459) &667(501)& -93(-89)& 523(385)\\
319: $\beta_3$ &169(186) & 217(143)&239(189) &307(229)& 2(11)& 585(382) \\
320: \hline \hline
321: \end{tabular}
322: \end{center}
323: \label{fulltable}
324: \end{table}
325: \newpage
326: \vspace{-6cm}
327: \begin{figure}
328: \center{\bf{Figure Captions}}
329: \vspace{1cm}
330: \caption{The structure of the [100] oriented single-height rebonded (SR) step (top) and the unrebonded (SU) step (bottom) on a (2$\times$1) reconstructed (001) surface. The dashed line denotes the edge of the step, where the dimer orientation undergoes a 90$^0$ rotation. Atoms are colored according to the number of dangling bonds(b) {\em before} surface and step-edge reconstructions: red=2b, green=1b, blue=0b.
331: The numbers represent the stretching (in \% ) of some of the bonds relative to the bond-length in the bulk for the Si(109) surface relaxed using the T3 potential. Note that the rebonding of the atoms across the step edge leads to stretching of several bonds in the
332: SR step. }
333: \label{steps}
334: \end{figure}
335: \begin{figure}
336: \caption{Surface energy of stepped
337: surfaces (in meV/ \AA$^2$) consisting of [100]SR(red) and
338: [110]DB(black) steps as a function of the biaxial strain
339: $\epsilon$ for Ge, computed with the T3 potential. The vicinal
340: angles and surface orientations are indicated in the figure.
341: The insets show the strain dependent formation
342: energy $\beta(\epsilon)$ (in meV/\AA$^2$)
343: for the three types of steps, DB(black squares), SR(red circles) and
344: SA+SB$^{[100]}$ (blue triangles) in Si and Ge.}
345: \label{gamma}
346: \end{figure}
347: \newpage
348: \begin{figure}
349: \begin{center}
350: \bf {Figure 1: Shenoy, Ciobanu and Freund}
351: \vspace{2cm}
352: \includegraphics{scf_105_fig1.eps}
353: \end{center}
354: \end{figure}
355: \newpage
356: \begin{figure}
357: \begin{center}
358: \bf {Figure 2: Shenoy, Ciobanu and Freund}
359: \vspace{2cm}
360: %\centerwmf{10cm}{8cm}{Aplfig2.wmf}
361: \includegraphics{scf_105_fig2.eps}
362: \end{center}
363: \end{figure}
364:
365: \end{document}
366:
367:
368:
369:
370:
371:
372:
373:
374:
375:
376:
377:
378:
379: