1: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
2: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
3:
4: % Some other (several out of many) possibilities
5: %\documentclass[preprint,aps]{revtex4}
6: %\documentclass[preprint,aps,draft]{revtex4}
7: %\documentclass[prb]{revtex4}% Physical Review B
8:
9: %\usepackage{graphicx}% Include figure files
10: %\usepackage{dcolumn}% Align table columns on decimal point
11: %\usepackage{bm}% bold math
12:
13: \begin{document}
14: \title {Linking Vibrational Dynamics to Local Electronic Structure:\\
15: Local Analysis of Dynamics of the relaxed Si$_{87}$ Cluster}
16:
17: \author{Lei Liu}
18: \author{C. S. Jayanthi}
19: \author{Shi-Yu Wu}
20: \affiliation{Department of Physics, University of Louisville,
21: Louisville, KY 40292}
22:
23: \date{\today}% It is always \today, today,
24: % but any date may be explicitly specified
25:
26: \begin{abstract}
27: A flexible scheme for decomposing the vibrational density of states in terms of pair
28: vibrational density of states is presented. This scheme provides the linkage between the
29: site vibrational density of states and pair vibrational density of states so that vibrational
30: modes, in particular localized modes, can be conveniently examined in terms of the
31: correlation between the vibration at a given site and those at its neighboring sites.
32: Furthermore, within the framework of a total energy vibrational spectrum study, this
33: scheme allows the analysis of vibrational modes in terms of their electronic origin. A
34: case study of the vibrational dynamics of the relaxed Si$_{87}$ cluster is carried out to
35: demonstrate the flexibility of the scheme in analyzing the properties of vibrational
36: modes, particularly for complex systems with reduced or no symmetry.
37: \end{abstract}
38:
39: \pacs{63.10.+a, 63.22.+m}% PACS, the Physics and Astronomy
40: % Classification Scheme.
41: %\keywords{Suggested keywords}%Use showkeys class option if keyword
42: %display desired
43: \maketitle
44:
45:
46: \section{\label{sec:level1}Introduction}
47: In recent years, total-energy-vibrational-spectrum studies have been carried out
48: for extended periodic systems as well as complex systems with reduced symmetry
49: because of the development of efficient and reliable methods of calculating the total
50: energy\cite{Kim95,Sti95,Sti96}. These studies start with the calculation of the total energy using semi-empirical
51: tight-binding (TB) approaches\cite{Kim95} or first principles methods\cite{Sti95,Sti96} such as those based on the
52: density functional theory (DFT). The force constant matrix between a pair of atoms are
53: then calculated as the appropriate second partial derivatives of the total energy with
54: respect to the displacements of the atoms about their equilibrium positions. While it is
55: indeed more satisfying to calculate the vibrational properties using force constant
56: matrices determined from the total energy based on quantum mechanical calculations
57: rather than from empirically-fitted classical potential, a key advantage of the total energy
58: approaches over the approaches based on the classical potentials is that the former allows
59: the possibility of linking the interesting dynamical features, in particular vibrational
60: modes associated with surfaces, interfaces, defects, impuritites, etc., directly to the local
61: electronic structure. However, this issue has not yet been systematically addressed. We
62: present here a scheme that provides a framework through which dynamic features of a
63: vibrating system can be related to the electronic structure. We also demonstrate the
64: application of this scheme using the vibration dynamics of the relaxed Si$_{87}$ cluster as the
65: working example. We will show how all the interesting dynamical features of the relaxed
66: Si$_{87}$ clusters can be traced to their electronic origins, including the local electronic
67: structure such as bond charges, dangling bonds, etc.
68:
69:
70: \section{\label{sec:level1}Methods}
71: The dynamics of a system is described by the equation of motion
72: \begin{equation}
73: \sum_{j\beta} H_{i\alpha,j\beta} c_{j\beta}^\lambda = \omega_\lambda^2 c_{i\alpha}^\lambda
74: \end{equation}
75: %\eqn(1)
76: where $H_{i\alpha,j\beta}=\phi_{i\alpha,j\beta}/\sqrt{m_im_j}$,
77: $c_{i\alpha}^\lambda=\sqrt{m_i}u_{i\alpha}^\lambda$,
78: with $\phi_{i\alpha,j\beta}$ being
79: the force constant matrix
80: element between atom $i$ of mass $m_i$ along the direction $\alpha$ ($x,y,$ or $z$) and atom $j$ of mass
81: $m_j$ along the direction $\beta$, and $u^\lambda_{i\alpha}$ the eigen-displacement of the $\lambda$-mode (with the eigen-frequency $\omega_\lambda$) at site $i$ along the $\alpha$-direction. Traditionally, the vibrational density of
82: states (VDOS) projected on sites (individual atoms) is the tool to analyze the nature and
83: properties of a defect or impurity mode. This is usually accomplished by introducing the
84: sum rule $\sum_{i\alpha}(c^\lambda_{i\alpha})^2=1$ into the defining equation of the VDOS such that
85: \begin{equation}
86: \rho (\omega) = \sum_{\lambda} \delta (\omega - \omega_\lambda)
87: =\sum_{\lambda i\alpha} (c_{i\alpha}^\lambda)^2 \delta (\omega - \omega_\lambda)
88: \end{equation}
89: %\eqn(2)
90: In this way,
91: \begin{equation}
92: \rho (\omega) = \sum_{i\alpha} \rho_{i\alpha} (\omega)
93: \end{equation}
94: %\eqn(3)
95: where
96: \begin{equation}
97: \rho_{i\alpha} (\omega) =\sum_{\lambda} (c_{i\alpha}^\lambda)^2 \delta (\omega - \omega_\lambda)
98: \end{equation}
99: %\eqn(4)
100: is the local VDOS at site $i$ along the $\alpha$ polarization. However, since the force field
101: characterizing a particular vibrational mode is determined by the interaction of the atom
102: at a given site with its neighbors, it is more illuminating if, in addition, the VDOS
103: projected on pairs of atoms is used in combination with the site VDOS to analyze the
104: mode. The interactions of a certain atom with its neighbors are described by the force
105: constant matrices. Thus a sum rule that involves the force constant matrices would be an
106: appropriate vehicle to construct the VDOS projected onto pairs of atoms. Such a scheme
107: can be set up as follows.
108:
109: Multiplying the equation of motion with $c^\lambda_{i\alpha}$ (Eq. (1)), we have\cite{Jay97a,Jay97b}
110: \begin{eqnarray}
111: %\begin{equation}
112: (c_{i\alpha}^\lambda)^2 &=& \sum_{j\beta} \frac{H_{i\alpha,j\beta} c_{i\alpha}^\lambda c_{j\beta}^\lambda}{\omega_\lambda^2}
113: \nonumber \\ &=&\sum_{j\ne i,\beta}
114: \frac{(u_{i\alpha}^\lambda u_{j\beta}^\lambda-u_{i\alpha}^\lambda u_{i\beta}^\lambda)\phi_{i\alpha,j\beta}}
115: {\omega_\lambda^2}
116: %\end{equation}
117: \end{eqnarray}
118: %\eqn(5)
119: In arriving at Eq. (5), the relation $\phi_{i\alpha,i\beta}=-\sum_{j\ne i}\phi_{i\alpha,j\beta}$ has been used. Equation (5) can
120: serve two useful functions. First, the summation of Eq. (5) over both $i$ and $\alpha$ leads to
121: \begin{equation}
122: \sum_{i\alpha,j\ne i,\beta}
123: \frac{(u_{i\alpha}^\lambda u_{j\beta}^\lambda-u_{i\alpha}^\lambda u_{i\beta}^\lambda)\phi_{i\alpha,j\beta}}
124: {\omega_\lambda^2}
125: = \sum_{i\alpha} (c_{i\alpha}^\lambda)^2 =1
126: \end{equation}
127: %\eqn(6)
128: This is precisely the sum rule needed to project the VDOS onto pairs of atoms $i$ and $j$ as
129: Eq. (6) also involves the force constant matrix elements $\phi_{i\alpha,j\beta}$. Thus, when Eq. (6) is
130: inserted into the defining equation for the VDOS (Eq. (2)), we obtain
131: \begin{eqnarray}
132: %\begin{equation}
133: \rho (\omega) &=& \sum_{\lambda,i\alpha,j\ne i,\beta}
134: \frac{(u_{i\alpha}^\lambda u_{j\beta}^\lambda-u_{i\alpha}^\lambda u_{i\beta}^\lambda)\phi_{i\alpha,j\beta}}
135: {\omega_\lambda^2}
136: \delta (\omega - \omega_\lambda) \nonumber \\
137: &=&\sum_{i\alpha,j\ne i,\beta} \rho_{i\alpha,j\beta} (\omega)
138: %\end{equation}
139: \end{eqnarray}
140: %\eqn(7)
141: where $\rho_{i\alpha,j\beta} (\omega)$, the pair VDOS of atom $i$ along the $\alpha$-direction and atom $j$ along the
142: $\beta$-direction linked by the force constant matrix element $\phi_{i\alpha,j\beta}$, is given by
143: \begin{equation}
144: \rho_{i\alpha,j\beta} (\omega)=\sum_{\lambda}
145: \frac{(u_{i\alpha}^\lambda u_{j\beta}^\lambda-u_{i\alpha}^\lambda u_{i\beta}^\lambda)\phi_{i\alpha,j\beta}}
146: {\omega_\lambda^2}
147: \delta (\omega - \omega_\lambda)
148: \end{equation}
149: %\eqn(8)
150: The second useful function of Eq. (5) is that it provides the framework to link the local
151: VDOS at a particular site to the interactions with its neighbors. In this way, the
152: vibrational mode localized in the vicinity of a given site can be correlated with the
153: vibrational motions of its interacting neighbors. Thus, from Eqs. (4) and (8), we have
154: \begin{equation}
155: \rho_{i\alpha} (\omega) = \sum_{j\ne i,\beta}\rho_{i\alpha,j\beta} (\omega)
156: \end{equation}
157: %\eqn(9)
158: Equation (5) can also be rewritten as
159: \begin{equation}
160: (c_{i\alpha}^\lambda)^2 =
161: \frac{\sum_{\beta\ne\alpha}\phi_{i\alpha,i\beta}u_{i\alpha}^\lambda u_{i\beta}^\lambda
162: +\sum_{j\ne i,\beta}\phi_{i\alpha,j\beta}u_{i\alpha}^\lambda u_{j\beta}^\lambda}
163: {\omega_\lambda^2-H_{i\alpha,i\alpha}}
164: \end{equation}
165: %\eqn(10)
166: $\sum_{i\alpha}(c_{i\alpha}^\lambda)^2 =1$ then leads to
167: \begin{equation}
168: \rho (\omega) = \sum_{i\alpha,j\beta}\bar{\rho}_{i\alpha,j\beta} (\omega)
169: \end{equation}
170: %\eqn(11)
171: with
172: \begin{equation}
173: \bar{\rho}_{i\alpha,j\beta} (\omega) = \sum_{\lambda}
174: \frac{\phi_{i\alpha,j\beta}u_{i\alpha}^\lambda u_{j\beta}^\lambda
175: -\phi_{i\alpha,i\alpha}(u_{i\alpha}^\lambda)^2\delta_{ij}\delta_{\alpha\beta}}
176: {\omega_\lambda^2-H_{i\alpha,i\alpha}}
177: \delta (\omega - \omega_\lambda)
178: \end{equation}
179: %\eqn(12)
180: Equations (10) and (12) also yield
181: \begin{equation}
182: \rho_{i\alpha} (\omega) = \sum_{j\beta}\bar{\rho}_{i\alpha,j\beta} (\omega)
183: \end{equation}
184: %\eqn(13)
185: Equation (13) allows the decomposition of the site VDOS in terms of on-site as well as
186: off-site contributions.
187:
188: While the scheme based on Eq. (5) provides the vehicle to decompose the VDOS
189: in terms of contributions from pairs of atoms directly interacting with each other via
190: $\phi_{i\alpha,j\beta}$, one may also devise a scheme to project the VDOS onto pairs of atoms interacting
191: directly as well as those not interacting directly. Such a scheme is useful for examining
192: collective (extended) as well as more localized vibrational modes. To develop such a
193: scheme, we note that
194: \begin{equation}
195: \sum_{j} c_{j\beta}^\lambda = 0
196: \end{equation}
197: %\eqn(14)
198: Multiplying Eq. (14) by $c_{i\alpha}$ and summing over $i, \alpha$ and $\beta$ lead to
199: \begin{equation}
200: \sum_{i\alpha,j\beta} c_{i\alpha}^\lambda c_{j\beta}^\lambda =
201: \sum_{i\alpha} (c_{i\alpha}^\lambda)^2 +
202: \sum_{i\alpha,j\beta} \{c_{i\alpha}^\lambda c_{j\beta}^\lambda
203: -(c_{i\alpha}^\lambda)^2\delta_{ij}\delta_{\alpha\beta}\}=0
204: \end{equation}
205: %\eqn(15)
206: Hence,
207: \begin{equation}
208: -\sum_{i\alpha,j\beta} \{c_{i\alpha}^\lambda c_{j\beta}^\lambda
209: -(c_{i\alpha}^\lambda)^2\delta_{ij}\delta_{\alpha\beta}\}
210: =\sum_{i\alpha} (c_{i\alpha}^\lambda)^2=1
211: \end{equation}
212: %\eqn(16)
213: Equation (16) then leads to
214: \begin{equation}
215: \rho (\omega) = \sum_{i\alpha,j\beta}\hat{\rho}_{i\alpha,j\beta} (\omega)
216: \end{equation}
217: %\eqn(17)
218: with
219: \begin{equation}
220: \hat{\rho}_{i\alpha,j\beta} (\omega) =
221: -\sum_{\lambda} \{c_{i\alpha}^\lambda c_{j\beta}^\lambda
222: -(c_{i\alpha}^\lambda)^2\delta_{ij}\delta_{\alpha\beta}\}
223: \delta (\omega - \omega_\lambda)
224: \end{equation}
225: %\eqn(18)
226: According to Eq.(16), we can also write
227: \begin{equation}
228: \rho_{i\alpha} (\omega) = \sum_{j\beta}\hat{\rho}_{i\alpha,j\beta} (\omega)
229: \end{equation}
230: %\eqn(19)
231: This equation then provides the link of the site VDOS to the pair VDOS $\hat{\rho}_{i\alpha,j\beta}$.
232: Similarly $\sum_{j} c_{j\beta}^\lambda = 0$ leads to
233: \begin{equation}
234: \sum_{i\alpha,j} c_{i\alpha}^\lambda c_{j\alpha}^\lambda =
235: \sum_{i\alpha} (c_{i\alpha}^\lambda)^2 +
236: \sum_{i\alpha,j\ne i} c_{i\alpha}^\lambda c_{j\alpha}^\lambda =0
237: \end{equation}
238: %\eqn(20)
239: or
240: \begin{equation}
241: \rho (\omega) = \sum_{i\alpha,j\ne i}\tilde{\rho}_{i\alpha,j\alpha} (\omega)
242: \end{equation}
243: %\eqn(21)
244: with
245: \begin{equation}
246: \tilde{\rho}_{i\alpha,j\alpha} (\omega)= -\sum_{\lambda} c_{i\alpha}^\lambda c_{j\alpha}^\lambda
247: \delta (\omega - \omega_\lambda)
248: \end{equation}
249: %\eqn(22)
250: Equation (20) also leads to
251: \begin{equation}
252: \rho_{i\alpha} (\omega) = \sum_{j}\tilde{\rho}_{i\alpha,j\alpha} (\omega)
253: \end{equation}
254: %\eqn(23)
255:
256: Equations (5) through (23) provide some examples of projecting the VDOS onto
257: pairs of atom and relating site VDOS to pair VDOS for the purpose of studying the
258: correlation between the vibrations at different localities for localized as well as more
259: extended vibrational modes.
260:
261: Within the framework of a total-energy-vibrational-spectrum approach, the
262: bonding configurations of atoms in the system, such as the bond charges, the nature of
263: the bonds, and the number of bonds associated with each atom, are already determined\cite{Alf99}.
264: Therefore, it is straightforward to relate a particular vibrational mode to its electronic
265: origin via either the site VDOS or the pair VDOS. For example, the VDOS can be
266: partitioned according to the contributions from different ranges of the total bond charge
267: $q_i=(1/2)\sum_j\sigma_{ij}$ associated with the atom at site $i$ in the system where $\sigma_{ij}$ is the bond
268: charge between the pair of atoms $i$ and $j$. This can be accomplished by summing the site
269: VDOS $\rho_i=\sum_\alpha\rho_{i\alpha}$ over those sites $i$ where $q_i$ falls within the range $[a,b]$, i.e.,
270: \begin{equation}
271: \rho (\omega ; a\le q_i \le b) = \sum_{a\le q_i \le b} \rho_i (\omega)
272: \end{equation}
273: %\eqn(24)
274: Since the bond charge is closely related to the strength of the bond, such a partitioning
275: can shed light on whether a particular mode is associated with the strengthening or
276: softening of the bonds. Similarly, the VDOS can also be grouped in terms of
277: contributions from atoms possessing a certain number of bonds, namely
278: \begin{equation}
279: \rho (\omega ; n_i=k) = \sum_{n_i=k} \rho_i (\omega)
280: \end{equation}
281: %\eqn(25)
282: where $n_i$ denotes the number of bonds possessed by the particle $i$. This presentation of
283: the vibrational spectrum provides the means to identify the modes associated with
284: dangling bonds and those associated with distorted bonding configurations. To examine
285: how an individual bond affects a particular mode, one may group the pair VDOSs
286: $\rho_{ij}=\sum_{\alpha\beta}\rho_{i\alpha,j\beta}$ according to the range of the bond charge $\sigma_{ij}$ such that
287: \begin{equation}
288: \rho (\omega ; a\le \sigma_{ij} \le b) = \sum_{a\le \sigma_{ij} \le b} \rho_{ij} (\omega)
289: \end{equation}
290: %\eqn(26)
291:
292: The characteristics of the bonds is usually described by molecular orbitals such as
293: $\tau=ss\sigma$, $sp\sigma$, $pp\sigma$, or $pp\pi$. To determine the nature of the bonds responsible for the
294: vibrational modes, one may classify the contribution to a particular mode by a certain
295: bonding type $\tau$ using
296: \begin{equation}
297: \rho (\omega ; \tau ) = \sum_{i, j\ne i} \frac{\sigma_{ij}^\tau}{\sigma_{ij}}\rho_{ij} (\omega)
298: \end{equation}
299: %\eqn(27)
300: where
301: \begin{equation}
302: \sigma_{ij} = \sum_{\tau} \sigma_{ij}^\tau
303: \end{equation}
304: %\eqn(28)
305:
306: The VDOS can also be decomposed in terms of quantities such as the distance,
307: the bond charge, etc. via $\bar{\rho}_{i\alpha,j\beta}$, or $\hat{\rho}_{i\alpha,j\beta}$ or $\tilde{\rho}_{i\alpha,j\alpha}$, following similar procedure as
308: described in Eq. (26). Thus, a very flexible framework that allows the convenient
309: decomposition of the VDOS to pair VDOS via $\rho_{i\alpha,j\beta}$, or $\bar{\rho}_{i\alpha,j\beta}$, or $\hat{\rho}_{i\alpha,j\beta}$, or $\tilde{\rho}_{i\alpha,j\alpha}$, and
310: the linking of the site VDOS to these pair VDOS (see Eqs. (9), (13), (19), and (23)) can
311: be set up. This framework can then be profitably used to analyze interesting dynamical
312: features of vibrations, in particular how these features are correlated with the local
313: electronic structure, of complex systems with reduced or no symmetry.
314:
315: \section{\label{sec:level1}Results and Discussion}
316: To demonstrate the application of the scheme outlined in Eqs. (5) to (27) to
317: examine the properties of the vibrational modes and to trace these modes to their
318: electronic origins, we use the dynamics of a relaxed Si$_{87}$ as the working example. The
319: initial configuration of the cluster is an 87-atom tetrahedral network with an atom at the
320: central site. The $x$-, $y$-, and $z$-axis are along the cube edges of the cube-unit cell with the
321: central site at a corner of the cell. The equilibrium configuration of this cluster (see Fig.
322: 1) was first determined by the MD scheme using the non-orthogonal tight-binding
323: Hamiltonian developed by Mennon and Subbaswamy\cite{Men97}. The force constant matrices of the
324: relaxed cluster were then numerically calculated from the total energy.
325:
326: In Fig. 2, the VDOS of the relaxed cluster, together with contributions to the
327: VDOS from those sites with the total bond charge in a certain range (see Eq. (24)), is
328: shown. We use the bonding configuration of the bulk Si as the reference to choose the
329: ranges of the total bond charge because Si-based systems favor the tetrahedral $sp^3$
330: bonding configuration. In bulk Si, each Si atom contributes 0.25$e$ to one of the four
331: equivalent bonds, giving rise to a total bond charge of 1$e$ for every atom. Thus, we have
332: selected the following four ranges, namely $q_i>1e$, $1e>q_i>0.8e$, $0.8e>q_i>0.6e$,
333: and $0.6e>q_i$, to examine the effect of relaxation on the bonding configuration, and hence on
334: the vibrational spectrum of Si$_{87}$. From Fig. 2, it can be seen that the majority of the
335: contributions to the VDOS originate from those sites with $1e>q_i>0.8e$ or $0.8e>q_i>0.6e$,
336: indicating that the relaxation has probably caused a general softening of
337: the bonds and thus resulting in an overall shift of the VDOS towards lower frequencies
338: compared to the bulk VDOS. Figure 2 also shows that the mode with the highest
339: frequency is a defect mode originating from the range $q_i>1e$, thus a strengthened
340: bonding configuration, while the mode with the lowest frequency originates from the
341: range $0.6e>q_i$, a softened bonding configuration or a configuration with dangling bonds.
342: In Fig. 3, the contribution to the VDOS is decomposed according to the number of bonds
343: associated with each atom (see Eq. (25)). The existence of a bond between a pair of
344: atoms is established by the criterion $\sigma_{ij}>0.04e$ rather than relying on the distance\cite{Alf99}.
345: Figure. 3 shows that the majority of the contributions still originate from atoms with 4
346: bonds, with a significant minority contribution from atoms with 3 bonds. The defect
347: mode at the highest frequency can be seen as associated with an atom with 4 bonds.
348: Together with the evidence in Fig. 2, it can be concluded that this mode results from the
349: strengthened bonding configuration. The lowest (non-zero) frequency mode, on the other
350: hand, is seen to originate from an atom with 2 bonds. The evidences provided by Figs. 2
351: and 3 thus indicate that this mode is a dangling-bond mode, but with remaining bonds
352: strengthened. To further explore this issue, we present in Fig. 4 the decomposition of the
353: VDOS in terms of the ranges of bond charge $\sigma_{ij}$ (see Eq. (26)). It can be seen that indeed
354: both the highest- and the lowest-frequency mode are associated with $\sigma_{ij}>0.5e$, indicating
355: that the bonds involved in these vibrations are strengthened bonds (compared to the bulk
356: bond). Figure 4 also shows that the majority of the contribution to the VDOS come from
357: the range $0.4e<\sigma_{ij}<0.5e$, the range with bond charge not too much deviated from the bulk
358: bonding configuration, thus reinforcing the observation provided by Figs. 2 and 3.
359: Another interesting feature in Fig. 4 is the negative contribution in the low frequency
360: VDOS from the range $\sigma_{ij}<10^{-3}e$. The negative contribution is an indication that for
361: contributions to a certain vibrational mode by pairs of atoms $i$ and $j$ interacting via
362: $\phi_{i\alpha,j\beta}$, some of those with no chemical bond formed between them may have the force
363: constant too ``soft'' to support the vibration at that frequency. In Fig. 5, we show the
364: decomposition of the VDOS in terms of bonding types (see Eq. (27)). It can be clearly
365: seen that the main contribution to all the vibrational modes comes from $pp\sigma$-bonds, with
366: a substantial minority contribution from $sp\sigma$-bonds. The directional characteristics of the
367: $pp\sigma$-bonds indicates the dominance of the stretching vibrations as is expected for a
368: Si-based structure.
369:
370: Figure 6 presents the decomposition of the VDOS according to the ranges of the
371: distance between atoms via $\hat{\rho}_{i\alpha,j\beta}$ (see Eqs. (18)). Since $\hat{\rho}_{i\alpha,j\beta}$ only involves $c^\lambda_{i\alpha}$ and $c^\lambda_{j\beta}$,
372: it provides an useful vehicle to analyze contributions to a certain vibrational mode from
373: the on-site environment, pair of atoms directly interacting (via $\phi_{i\alpha,j\beta}$) with each other,
374: and pairs of atoms not directly interacting. The three ranges of distances chosen for the
375: decomposition in Fig. 6 correspond roughly to 1st nearest-neighbor distances ($r_{ij}<3.10\AA$),
376: 2nd nearest-neighbor distances ($3.10\AA <r_{ij}<4.15\AA$), and 3rd nearest-neighbor distances
377: and beyond ($r_{ij}>4.15\AA$) respectively. Figure 6 shows that there are contributions to the
378: VDOS from all three ranges to most of the low- to intermediate-frequency modes and the
379: contributions are consistent with the number of pairs of atoms in each range. This is an
380: indication that these modes are extended modes. On the other hand, the dominant
381: contribution to the high-frequency modes mainly originates from the range $r_{ij}<3.10\AA$.
382: Considering the fact that for any given atom in the cluster, there are many more
383: neighbors at farther distances than close distances, the dominance of the contribution to
384: VDOS from $r_{ij}<3.10\AA$ for high-frequency modes suggests that these modes are more
385: localized. For example, the highest-frequency mode appears to be a localized mode as
386: contributions to this mode from second nearest neighbors and beyond are out-weighted
387: by the contribution from the nearest neighbors, even though there are many more farther
388: neighbors than nearest neighbors for every atom in the cluster. The inset of Fig. 6 shows
389: in more detail the contributions to the non-zero lowest-frequency mode. The dominant
390: contribution to this mode is apparently from the on-site environment. To gain more
391: insight to these two interesting modes, one may first plot $\sum_\alpha (c^\lambda_{i\alpha})^2$ for the mode
392: $\omega_\lambda$ as a function of site index $i$ to determine where the mode is localized and then
393: decompose the site VDOS for that site according to distances (1st-, 2nd-, etc., based on
394: Eqs. (8) and (9)) to shed light on the correlation between the vibration of the site and
395: those of its neighbors. The upper inset of Fig. 7 is the plot of $\sum_\alpha (c^\lambda_{i\alpha})^2$ for the
396: highest-frequency mode (~108 THz) as a function of the site index. The mode is obviously
397: localized in the immediate neighborhood of site 1 (the central site). The main panel of
398: Fig. 7 shows the site VDOS $\rho_1=\sum_\alpha\rho_{1\alpha}(\omega)$ decomposed in terms of 1st-, 2nd-, etc.,
399: nearest-neighbor distance using Eqs. (8) and (9). It can be seen that the contribution to
400: this mode is almost entirely from the coupling of the central site (site 1) with its 4 nearest
401: neighbors. In the lower inset of Fig. 7, the nature of the coupling is examined by
402: decomposing $\rho_{1}$ in terms of $\sum_{j\ne 1}\rho_{1\alpha,j\beta}$. It is seen that the majority of the contribution are
403: from $xx$-, $yy$-, and $zz$-nearest neighbor coupling. It is also seen that the mode is actually
404: a composite of three almost degenerate modes (107.2, 108.0, and 107.5 THz), an
405: indication of the distortion of the immediate environment of the central site from the
406: perfect tetrahedral configuration.
407:
408: The analysis of the non-zero lowest-frequency mode is shown in Fig. 8. The
409: upper inset shows that the mode is an on-site mode at site 82 located on the surface of the
410: cluster (see Fig. 1). The site VDOS $\rho_{82}$, decomposed according to Eq. (13), confirms this
411: picture. The lower inset of Fig. 8 is the decomposition of $\rho_{82}$ according to Eq. (13) so
412: that the on-site coupling can be examined. It shows that two almost equivalent on-site $xy$-
413: and $yx$-coupling are the main contributors to this mode, indicating that this is mainly a
414: wagging mode in the $xy$ plane anchored by the two strengthened bonds of the atom at site
415: 82 (a site with two dangling bonds, see Fig. 1).
416:
417: The relaxed Si$_{87}$ cluster is a system with no symmetry. The demonstration
418: discussed in this section shows the flexibility of the methodologies developed in Sec. II
419: to carry out detailed analysis of the dynamics of such a system and to trace interesting
420: dynamical features to their electronic origin. Hence the framework described in Sec. II is
421: expected to be extremely useful for the study and the analysis of vibrational dynamics of
422: complex systems with low or no symmetry.
423:
424:
425: \begin{acknowledgments}
426: This work was supported by the NSF (DMR-011284)
427: and the DOE (DE-FG02-00ER45832) grants.
428: \end{acknowledgments}
429: \newpage
430:
431: \begin{references}
432:
433: \bibitem{Kim95} J. Kim, M.-L. Yeh, F.S. Khan, and J.W. Wilkins, Phys. Rev. B{\bf 52}, 14709 (1995).
434: \bibitem{Sti95} I. Stich, K. Terakura, and B. E. Larson, Phys. Rev. Lett. {\bf 74}, 4491 (1995).
435: \bibitem{Sti96} I. Stich, J. Kohanoff, and K. Terakura, Phys. Rev. B{\bf 54}, 2642 (1996).
436: \bibitem{Jay97a}C.S. Jayanthi, M. Tang, S.Y. Wu, J.A. Cocks, and S. Yip, {\it Computer-aided Design of
437: High Temperature Materials}, ed. A. Pechenik, R.K. Kalida, and P. Vashishta, Oxford
438: University Press (1997).
439: \bibitem{Jay97b} C.S. Jayanthi, M. Tang, S.Y. Wu, J.A. Cocks, and S. Yip, Phys. Rev. Lett. {\bf 79}, 4601
440: (1997).
441: \bibitem{Alf99}D. R. Alfonso, S.Y. Wu, C.S. Jayanthi, and E. Kaxiras, Phys. Rev. B{\bf 59}, 7745 (1999).
442: \bibitem{Men97} M. Menon and K. R. Subbaswamy, Phys. Rev. B{\bf 55}, 9231 (1997).
443:
444: \end{references}
445:
446:
447: \newpage
448:
449: \noindent{\bf FIGURES}
450:
451: \vskip 0.1in
452:
453: \noindent {FIG. 1 The equilibrium configuration of the relaxed Si$_{87}$ cluster. The $z$-axis is the
454: vertical axis and the $x$-axis is the horizontal axis. Atom 82 (see the result and discussion
455: in the text) is highlighted by the dark ball.}
456: \vskip 0.1in
457: \noindent {FIG. 2 The decomposition of the VDOS in terms of the range of the bond charge $q_i$
458: associated with the atoms in the cluster.}
459: \vskip 0.1in
460: \noindent{FIG.3 The decomposition of the VDOS in terms of the number of bonds associated
461: with the atoms in the cluster.}
462: \vskip 0.1in
463: \noindent{FIG.4 The decomposition of the VDOS in terms of the bond charge $\sigma_{ij}$ between pairs
464: of atoms in the cluster.}
465: \vskip 0.1 in
466: \noindent{FIG.5 The decomposition of the VDOS according to the bond types.}
467: \vskip 0.1 in
468: \noindent{FIG.6 The decomposition of the VDOS in terms of the distance between pairs of
469: atoms according to $\hat{\rho}_{i\alpha,j\beta}$ (Eq. (17)). The inset is the enlargement of the curves in the
470: immediate neighborhood of the mode with the lowest non-zero frequency.}
471: \vskip 0.1 in
472: \noindent {Fig. 7 Upper inset shows the plot of $\sum_\alpha (c^\lambda_{i\alpha})^2$ vs. the site index $i$ for the highest frequency
473: mode (~108 THz). The main panel shows the site VDOS for site 1 (the central site), $\rho_1$,
474: decomposed according to 1st nearest neighbor, 2nd nearest neighbor, etc., distance using
475: Eqs. (8) and (9). The lower inset shows the decomposition of $\rho_{1}$ in terms of $\sum_{j\ne 1}\rho_{1\alpha,j\beta}$
476: in the
477: immediate neighborhood of the mode with the highest frequency.}
478: \vskip 0.1 in
479: \noindent {Fig. 8 Upper inset shows the plot of $\sum_\alpha (c^\lambda_{i\alpha})^2$ vs. the site index for the lowest non-zero
480: frequency mode. It can be seen that this mode is confined at the site 82 (a surface site, see
481: Fig. 1). The main panel shows the site VDOS $\rho_{82}$ and its decomposition according to Eq.
482: (13). The lower inset shows the decomposition of $\rho_{82}$ in terms of $\sum_j\bar{\rho}_{82\alpha,j\beta}$ in the
483: immediate neighborhood of the mode with the lowest non-zero frequency.}
484:
485: \end{document}
486:
487:
488:
489:
490:
491:
492: