1: %\documentclass[aps,prb,preprint,superscriptaddress,showpacs]{revtex4}
2: \documentclass[aps,prb,twocolumn,superscriptaddress,showpacs]{revtex4}
3: \usepackage{graphicx}% Include figure files
4: %\usepackage[bookmarks=false]{hyperref}
5: %\hypersetup{pdfstartview=FitBH}
6:
7: \begin{document}
8:
9: \title{Longitudinal versus transverse spheroidal vibrational
10: modes of an elastic sphere}
11:
12: \author{Lucien Saviot}
13: \affiliation{Laboratoire de Recherche sur la R\'eactivit\'e des Solides,
14: UMR 5613 CNRS - Universit\'e de Bourgogne\\
15: 9 avenue A. Savary, BP 47870 - 21078 Dijon - France}
16: \email{lucien.saviot@u-bourgogne.fr}
17:
18: \author{Daniel B. Murray}
19: \affiliation{Department of Physics and Astronomy,
20: The University of British Columbia Okanagan, 3333 University Way,
21: Kelowna, British Columbia, Canada V1V 1V7}
22: \email{daniel.murray@ubc.ca}
23:
24: \date{\today}
25:
26: \begin{abstract}
27: Analysis of the spheroidal modes of vibration of a free
28: continuum elastic sphere show that they can be qualitatively grouped
29: into two categories: primarily longitudinal and primarily
30: transverse. This is not a sharp distinction. However,
31: there is a relatively stark contrast between the two kinds
32: of modes. Primarily transverse modes have a small divergence
33: and have frequencies that are almost functionally independent
34: of the longitudinal speed of sound. Analysis of inelastic
35: light scattering intensity from confined acoustic phonons
36: in nanoparticles requires an understanding of this qualitative
37: distinction between different spheroidal modes. In
38: addition, some common misconceptions about spheroidal modes
39: are corrected.
40: \end{abstract}
41:
42: \pacs{62.20.-x,43.20.Ks,62.25.+g}
43: \maketitle
44:
45: \section{Introduction}
46: \label{Introduction}
47:
48: With the explosion of interest in the optical properties of
49: nanoparticles, the classic elastic mechanical problem of
50: the vibrational modes of a free continuum sphere has found
51: a new context for application. The problem was formally
52: and numerically solved back in 1882.\cite{lamb1882}
53: Nanoparticles, \textit{i.e.} spherical clusters of atoms ranging
54: in diameter from 1~nm to 100~nm, have sufficiently few atoms
55: that the continuum approximation can be
56: questioned.\cite{ChengPRB05,ErratumChengPRB05}
57: Even so, it is acceptable to ignore the effects of the discrete
58: crystal lattice for the few vibrational modes with lowest
59: frequency as long as the nanoparticle diameter exceeds
60: several nanometers.
61:
62: Inelastic light scattering of a continuous laser beam shining on
63: the nanoparticle permits detection of the mechanical vibrations
64: since the changing size and shape of the nanoparticle modulates
65: the polarizability of the nanoparticle, so that the monochromatic
66: incident light turns into scattered light with sidebands shifted
67: up and down by the frequency of the vibrations. This can be
68: seen using experimental setups of the Raman and Brillouin type.
69:
70: For theoretical convenience the material is assumed to be
71: homogeneous, isotropic and linear. The outer surface of the
72: sphere is free from externally imposed stresses and this
73: situation will be referred to as the ``free sphere model'' (FSM).
74: The original paper by Lamb\cite{lamb1882} classified the FSM
75: modes of vibration into two classes, now called ``torsional''
76: (TOR) and ``spheroidal'' (SPH). The distinctive feature of
77: torsional modes is that the material density does not vary. In
78: other words, the divergence of the displacement field is zero.
79: Furthermore, the spherical symmetry permits classification of
80: the modes by angular momentum number $\ell \geq 0$.
81: (However, later on we will show that there is value
82: in considering $\ell$ to be a continuous variable.) There is no
83: dependence of the frequency on the $z$ angular momentum $m$.
84: Beyond this, the modes are indexed by $n \geq 0$.
85: It is convenient to let $p$ denote either SPH or TOR,
86: to indicate individual modes by $(p,\ell,n)$ and their
87: frequencies by $\omega_{p \ell n}$.
88:
89: What we explore in this paper is an additional classification of
90: the SPH modes beyond that employed since Lamb. In particular,
91: SPH modes can be classified (albeit approximately and
92: subjectively) as either being primarily longitudinal (SPH$_L$)
93: or primarily transverse (SPH$_T$) in nature. The specific meaning
94: of this will be explained further on. This is not a sharp
95: division, and actual modes fall somewhere in between the two
96: ideals. However, the contrast is sufficiently sharp that this
97: new distinction among SPH FSM modes as SPH$_L$ or SPH$_T$ is a
98: very important tool.
99:
100: In a recent theoretical paper, G.~Bachelier and
101: A.~Mlayah\cite{bachelierPRB04} predicted that
102: (SPH,$\ell=2$,$n$) modes with differing values of $n$
103: contribute to the Raman spectrum in a highly non-uniform way.
104: In this paper we will show that this can be explained using
105: the previously mentionned distinction between SPH modes.
106: They pointed out that there are two separate mechanisms that
107: couple (SPH,$\ell=2$) acoustic vibrations to the surface plasmon
108: resonance and in turn lead to Raman scattering. First, change
109: of the particle shape and second, modulation of the density
110: leading to change of optical properties through the deformation
111: potential.
112:
113: Section \ref{Freesphere} reprises the formalism necessary for the
114: FSM solution. In Section \ref{Longitudinality}, we show explicitly
115: what we mean by SPH$_L$ and SPH$_T$. In Section \ref{Highfreq}, we
116: illustrate the natural appearance of SPH$_L$ and SPH$_T$ modes in the
117: high frequency limit. Section \ref{Discussion} discusses these results
118: and their connection with inelastic light scattering experiments.
119:
120: \section{The Free Sphere Model}
121: \label{Freesphere}
122:
123: Vibrational modes of a free linear elastic continuum homogeneous
124: isotropic sphere were found by Lamb in 1882.\cite{lamb1882}
125:
126: For a mode with angular frequency $\omega$, the displacement of
127: material point $\vec{r}$ from its equilibrium position is
128: $\vec{u}(\vec{r}) cos(\omega t)$. For a $m=0$ TOR mode,
129: $\vec{u}$ =
130: A $\nabla \times ( \vec{r} j_{\ell}(k_T r) P_{\ell}(cos \theta) )$,
131: where $j_{\ell}$ are spherical Bessel functions of the first kind
132: and $P_{\ell}$ are Legendre polynomials. For a $m=0$ SPH mode,
133: $\vec{u}$ = $\vec{u}_L$ + $\vec{u}_T$ where
134: \begin{equation}
135: \vec{u}_L(r,\theta) = B \nabla j_{\ell}(k_L r) P_{\ell}(cos \theta)
136: \label{Bterm}
137: \end{equation}
138: and
139: \begin{equation}
140: \vec{u}_T(r,\theta) = C \nabla \times \nabla \times ( \vec{r} j_{\ell}(k_T r) P_{\ell}(cos \theta) )
141: \label{Cterm}
142: \end{equation}
143: where $A$, $B$ and $C$ are real coefficients,
144: $v_L k_L$ = $v_T k_T$ = $\omega$,
145: and $v_T$ and $v_L$ are the transverse and longitudinal speeds
146: of sound.
147:
148: Modes with $z$ angular momentum $m \ne 0$ have a different
149: functional form.
150:
151: $R$ is the nanoparticle radius. If $\sigma_{ij}$ is the stress
152: tensor, the boundary conditions at $r$ = $R$ are $\sigma_{rr} $
153: = $\sigma_{r \theta}$ = 0. It is convenient to introduce
154: dimensionless frequencies $\eta$ = $k_T R$ and $\xi$ = $k_L R$.
155: Following Eringen,\cite{eringen} application of these boundary
156: conditions determines the allowed SPH vibrational frequencies as
157: zeroes of a 2 $\times$ 2 determinant for $\ell>0$.
158: \begin{equation}
159: \Delta_{\ell} =
160: \begin{array}{|cc|}
161: T_{11} & T_{13}\\
162: T_{41} & T_{43}
163: \end{array}
164: \label{twobytwo}
165: \end{equation}
166: where
167: \begin{eqnarray}
168: \nonumber T_{11} & = &
169: \left(\ell^2 - \ell - \frac{\eta^2}2\right) j_{\ell}(\xi) + 2 \xi j_{\ell+1}(\xi)\\
170: \nonumber T_{13} & = &
171: \ell (\ell + 1)\left\{ (\ell - 1) j_{\ell}(\eta) - \eta
172: j_{\ell+1}(\eta)\right\}\\
173: \nonumber T_{41}& = &
174: (\ell - 1) j_{\ell}(\xi) - \xi j_{\ell+1}(\xi)\\
175: \nonumber T_{43} & = &
176: \left(\ell^2 - 1 - \frac{\eta^2}2 \right) j_{\ell}(\eta) + \eta j_{\ell+1}(\eta)
177: \end{eqnarray}
178:
179: For $\ell=0$, the allowed vibrational frequencies are the zeroes
180: of $T_{11}$.
181:
182: Noting that the displacement fields are real-valued, it is
183: appropriate to use the following inner product between two
184: displacement fields $u_A$ and $u_B$:\cite{murrayPRB04}
185:
186: \begin{equation}
187: (u_A | u_B) =
188: \frac{\int_{r<R} \vec{u_A}(\vec r) \cdot \vec u_B(\vec r) \rho d^3\vec{r}}
189: {\int_{r<R} \rho d^3\vec{r}}
190: \label{norm}
191: \end{equation}
192:
193: A normalization condition (such as ($u$$|$$u$) = 1) would
194: typically determine the final values of $B$ and $C$. But the
195: details of the condition do not affect the results reported
196: here. The displacement field $\vec{u}(\vec{r})$ for some
197: selected modes are depicted in Fig.~\ref{prf12}.
198:
199: \begin{figure}[!ht]
200: \includegraphics[width=\columnwidth]{prf1a}
201: \includegraphics[width=\columnwidth]{prf1b}
202: \caption{\label{prf12}(Color online) Displacement fields
203: $\vec{u}(\vec{r})$ of selected SPH $\ell$=2 modes. As explained in the
204: text, the first three are primarly transverse (\textit{i.e.} SPH$_T$).
205: (SPH,2,3) is primarly longitudinal (\textit{i.e.} SPH$_L$). The
206: equilibrium surface of the nanoparticle and the $z$-axis are shown as
207: dotted lines. The solid (red online) line shows the distorted surface.
208: Note that the (SPH,2,1) mode does not change the nanoparticle shape.}
209: \end{figure}
210:
211: \section{Spheroidal Mode Longitudinality}
212: \label{Longitudinality}
213:
214: Isotropic elastic materials differ in their Poisson ratio,
215: $\nu$, which is related to $x = v_T/v_L$ through
216: $x = \sqrt{(1 - 2 \nu)/(2 - 2 \nu)}$. Figure~\ref{SPH2} shows how
217: the dimensionless frequency, $\eta$, of the SPH $\ell = 2$ FSM
218: modes varies with $v_T/v_L$. It is quite apparent that some
219: modes keep the same $\eta$ as $v_T/v_L$ is varied. However,
220: other modes change frequency as $v_T/v_L$ changes. There
221: are transition points where a given mode changes from
222: being constant to varying with $v_T/v_L$.
223:
224: \begin{figure}[!ht]
225: \includegraphics[width=\columnwidth]{SPH2}
226: \caption{\label{SPH2}(Color online) Dimensionless mode frequency $\eta$
227: as a function of $v_T/v_L$ for (SPH,$\ell=2$) modes. Vertical lines
228: (blue online) mark $v_T/v_L$ for Au, Ag, Si and Ge from left to right.}
229: \end{figure}
230:
231: This pattern visible in Fig.~\ref{SPH2} motivates the search
232: for a numerical criterion to permit this contrast among modes to
233: be quantified. We adopt the starting point that in some
234: sense some modes are more transverse in nature (SPH$_T$) while others
235: are more longitudinal (SPH$_L$). We then coin the term ``longitudinality'',
236: denoted by $L$, for a quantity that varies on a scale from 0 to
237: 1 with 0 being purely transverse and 1 being purely longitudinal.
238: There is no single obvious way of doing this. Rather, we have
239: evaluated a number of quantities as candidates for the best
240: measure of longitudinality, of which we present four which
241: work well. These will be denoted $L1$, $L2$, $L3$, and $L4$.
242:
243: Consider a particular SPH mode with indices $\ell$ and $n$. Its
244: frequency is $\omega(v_L,v_T)$. Define $L1$ by
245: \begin{equation}
246: L1 = \frac{v_L}{\omega} \frac{\partial \omega}{\partial v_L}
247: = - \frac{x}{\eta} \frac{d \eta}{d x}
248: = 1 - \frac{v_T}{\omega} \frac{\partial \omega}{\partial v_T}
249: \label{eqL}
250: \end{equation}
251:
252: Noting that
253: $\vec{u}(\vec{r}) = \vec{u}_T(\vec{r}) + \vec{u}_L(\vec{r})$,
254: we define
255: $L2$ as $(u_L|u_L) / (u|u)$,
256: and $L3$ as $1 - \left((u_T|u_T) / (u|u)\right)$.
257: But note also that
258: $(u_L|u_L) + (u_T|u_T) \ne (u|u)$
259: since $(u_L|u_T) \ne 0$.
260:
261: Given a fixed value of $v_T/v_L$ and $n$, $\eta$ may be considered
262: to be a continuous function of $\ell$, as in Fig.~\ref{etavsl}.
263: In terms of this $\eta(\ell)$, define
264: \begin{equation}
265: L4 = \frac{v_T}{v_L-v_T}
266: \left( \frac2\pi \frac{d\eta}{d\ell} - 1 \right)
267: \end{equation}
268:
269: \begin{figure}[!ht]
270: \includegraphics[width=\columnwidth]{ell}
271: \caption{\label{etavsl}(Color online) Variation of the SPH mode
272: frequency with $\ell$ for a material with $v_T/v_L$ = 0.5. Full
273: circles are exact FSM frequencies and are connected with curves
274: calculated for non-integer $\ell$ (red online). Lines with crosses
275: are roots of $T_{11}$ (blue online) which approximate SPH$_L$ modes.
276: Lines with empty squares are roots of $T_{43}$ (green online) which
277: approximate SPH$_T$ modes.}
278: \end{figure}
279:
280: Let $<..>_V$ and $<..>_S$ denote averages over the nanoparticle
281: volume and surface, respectively. In particular,
282: $(u|u) = < u_r^2 + u_\theta^2 + u_\phi^2 >_V$.
283: Some other measures of interest are as follows:
284: URV = $< u_r^2 >_V/(u|u)$,
285: URS = $< u_r^2 >_S/(u|u)$,
286: UTS = $< u_\theta^2 + u_\phi^2 >_S/(u|u)$,
287: and U2S = URS + UTS.
288:
289: Note that all of these quantities are defined in such a
290: way as to be independent of $m$.
291:
292: \begin{figure}[!ht]
293: \includegraphics[width=\columnwidth]{Poisson2}
294: \caption{\label{Poisson2}(Color online) Dimensionless mode frequency
295: $\eta$ as a function of longitudinality measures $L1$ (circles (red
296: online)) and $L2$ (crosses (green online)) for SPH modes of a material
297: with $v_T/v_L=0.5$. $\ell = 1 \ldots 4$ from left to right. SPH$_T$
298: modes have L1 and L2 near zero, while for SPH$_L$ modes they are close
299: to 1.}
300: \end{figure}
301:
302: \begin{table}
303: \caption{\label{AllL}Longitudinality measures for a material with
304: $v_T$/$v_L$ = 0.5.}
305: \begin{tabular}{|c|c|c|c|c|c|c|c|c|c|c|}
306: \toprule
307: $\ell$&$n$&$\eta$ & L1 & L2 & L3 & L4\\
308: \colrule
309: 0 & 0 & 5.49 & 1.36 & 1.00 & 1.00 & 0.86\\
310: 0 & 1 & 12.23 & 1.06 & 1.00 & 1.00 & 0.84\\
311: 0 & 2 & 18.63 & 1.02 & 1.00 & 1.00 & 0.89\\
312: \colrule
313: 1 & 0 & 3.60 & 0.24 & 0.32 &-0.27 &-0.12\\
314: 1 & 1 & 7.24 & 0.19 & 0.13 & 0.10 &-0.09\\
315: 1 & 2 & 8.55 & 0.94 & 0.84 & 0.84 & 0.66\\
316: 1 & 3 & 10.73 & 0.02 & 0.03 & 0.02 &-0.03\\
317: 1 & 4 & 13.89 & 0.03 & 0.03 & 0.03 &-0.06\\
318: 1 & 5 & 15.19 & 0.99 & 0.96 & 0.96 & 0.79\\
319: 1 & 6 & 17.11 & 0.01 & 0.01 & 0.01 &-0.03\\
320: \colrule
321: 2 & 0 & 2.65 & 0.02 & 0.14 &-0.85 &-0.06\\
322: 2 & 1 & 5.10 & 0.23 & 0.41 & 0.02 & 0.01\\
323: 2 & 2 & 8.63 & 0.13 & 0.09 & 0.04 &-0.13\\
324: 2 & 3 & 10.99 & 0.85 & 0.77 & 0.75 & 0.42\\
325: 2 & 4 & 12.29 & 0.14 & 0.15 & 0.15 & 0.05\\
326: 2 & 5 & 15.35 & 0.03 & 0.03 & 0.02 &-0.07\\
327: 2 & 6 & 17.85 & 0.87 & 0.84 & 0.84 & 0.57\\
328: 2 & 7 & 18.68 & 0.13 & 0.14 & 0.14 & 0.07\\
329: \colrule
330: 3 & 0 & 3.95 & 0.04 & 0.27 &-1.15 &-0.25\\
331: 3 & 1 & 6.71 & 0.19 & 0.37 & 0.24 & 0.03\\
332: 3 & 2 & 9.98 & 0.12 & 0.10 & 0.04 &-0.14\\
333: 3 & 3 & 12.95 & 0.48 & 0.41 & 0.38 & 0.08\\
334: 3 & 4 & 14.12 & 0.50 & 0.51 & 0.51 & 0.28\\
335: 3 & 5 & 16.80 & 0.04 & 0.04 & 0.04 &-0.07\\
336: 3 & 6 & 19.84 & 0.20 & 0.19 & 0.18 & 0.01\\
337: \colrule
338: 4 & 0 & 5.07 & 0.06 & 0.36 &-1.30 &-0.31\\
339: 4 & 1 & 8.31 & 0.14 & 0.29 & 0.33 & 0.01\\
340: 4 & 2 & 11.33 & 0.12 & 0.12 & 0.06 &-0.14\\
341: 4 & 3 & 14.49 & 0.24 & 0.19 & 0.16 &-0.07\\
342: 4 & 4 & 16.21 & 0.71 & 0.69 & 0.69 & 0.33\\
343: 4 & 5 & 18.26 & 0.08 & 0.09 & 0.09 &-0.05\\
344: \botrule
345: \end{tabular}
346: \end{table}
347:
348: Except at low $\eta$, Fig.~\ref{Poisson2} shows that L1
349: and L2 are in close agreement. L3 and L4 are not plotted, but
350: also agree closely except at low $\eta$ (see Table~\ref{AllL}).
351: Generally, a given mode either
352: has all of L1, L2, L3, and L4 low, or else all high.
353: It is thus possible
354: to classify modes as SPH$_L$ or SPH$_T$. Rarely, there are
355: cases where the values of L1, L2, L3, and L4 are in the intermediate
356: range, such as in Fig.~\ref{Poisson2} for $\ell = 3$ for
357: the two modes near $\eta = 13$. Such modes which are neither
358: clearly SPH$_L$ nor SPH$_T$ always occur in pairs.
359: The reason for this is explained in Section~\ref{Highfreq}.
360:
361: Table~\ref{def} provides additional information about the modes.
362: The dimensionless frequency $\eta$ is provided for convenience,
363: as is the ratio of coefficients $B$ and $C$.
364:
365: In principle, $C/B$ could be expected to provide useful
366: information about whether a mode is SPH$_L$ or SPH$_T$. In the
367: extreme case that $C = 0$, the mode is evidently SPH$_L$,
368: and likewise when $B = 0$ the mode is SPH$_T$. But the
369: $C/B$ values do not exhibit an informative pattern.
370:
371: There is a strong contrast in the values of URS for
372: different modes. However, it does not correlate to whether the
373: mode is SPH$_L$ or SPH$_T$ except at high enough $\eta$.
374: URS is an interesting quantity
375: because it is the one we have to monitor for the surface
376: deformation mechanism except for $\ell=0$ modes.
377:
378: \begin{table}
379: \caption{\label{def}Other features of SPH modes for a material with
380: $v_T$/$v_L$ = 0.5.}
381: \begin{tabular}{|c|c|c|c|c|c|c|c|c|c|c|}
382: \toprule
383: $\ell$&$n$&$\eta$ &$C/B$&URV & URS & UTS & U2S & Class &$n_L$&$n_T$\\
384: \colrule
385: 0 & 0 & 5.49 & 0.00 & 1.00 & 0.91 & 0.00 & 0.91 & SPH$_L$ & 0 & \\
386: 0 & 1 & 12.23 & 0.00 & 1.00 & 0.71 & 0.00 & 0.71 & SPH$_L$ & 1 & \\
387: 0 & 2 & 18.63 & 0.00 & 1.00 & 0.68 & 0.00 & 0.68 & SPH$_L$ & 2 & \\
388: \colrule
389: 1 & 0 & 3.60 & -0.99 & 0.37 & 0.05 & 1.26 & 1.31 & SPH$_T$ & & 0 \\
390: 1 & 1 & 7.24 & 1.65 & 0.43 & 0.14 & 0.66 & 0.81 & SPH$_T$ & & 1 \\
391: 1 & 2 & 8.55 & -0.29 & 0.43 & 0.70 & 0.03 & 0.72 & SPH$_L$ & 0 & \\
392: 1 & 3 & 10.73 & 4.37 & 0.21 & 0.00 & 0.69 & 0.69 & SPH$_T$ & & 2 \\
393: 1 & 4 & 13.89 & -3.93 & 0.14 & 0.02 & 0.68 & 0.69 & SPH$_T$ & & 3 \\
394: \colrule
395: 2 & 0 & 2.65 & -0.44 & 0.59 & 0.66 & 0.20 & 0.86 & SPH$_T$ & & 0 \\
396: 2 & 1 & 5.10 & -0.38 & 0.26 & 0.00 & 1.78 & 1.78 & SPH$_T$ & & 1 \\
397: 2 & 2 & 8.63 & 1.09 & 0.50 & 0.08 & 0.85 & 0.92 & SPH$_T$ & & 2 \\
398: 2 & 3 & 10.99 & -0.22 & 0.35 & 0.71 & 0.04 & 0.75 & SPH$_L$ & 0 & \\
399: 2 & 4 & 12.29 & 0.96 & 0.34 & 0.07 & 0.64 & 0.71 & SPH$_T$ & & 3\\
400: \colrule
401: 3 & 0 & 3.95 & -0.17 & 0.74 & 0.89 & 0.03 & 0.91 & SPH$_T$ & & 0 \\
402: 3 & 1 & 6.71 & -0.21 & 0.22 & 0.04 & 1.91 & 1.95 & SPH$_T$ & & 1 \\
403: 3 & 2 & 9.98 & 0.70 & 0.53 & 0.04 & 1.05 & 1.09 & SPH$_T$ & & 2 \\
404: 3 & 3 & 12.95 & -0.33 & 0.35 & 0.43 & 0.41 & 0.84 & mix & 0 & 3 \\
405: 3 & 4 & 14.12 & 0.27 & 0.37 & 0.39 & 0.29 & 0.68 & mix & 0 & 3 \\
406: \colrule
407: 4 & 0 & 5.07 & -0.08 & 0.81 & 1.05 & 0.00 & 1.05 & SPH$_T$ & & 0 \\
408: 4 & 1 & 8.31 & -0.15 & 0.23 & 0.14 & 1.78 & 1.92 & SPH$_T$ & & 1 \\
409: 4 & 2 & 11.33 & 0.46 & 0.54 & 0.01 & 1.25 & 1.27 & SPH$_T$ & & 2 \\
410: 4 & 3 & 14.49 & -0.41 & 0.41 & 0.21 & 0.75 & 0.95 & SPH$_T$ & & 3 \\
411: 4 & 4 & 16.21 & 0.14 & 0.35 & 0.65 & 0.03 & 0.68 & SPH$_L$ & 0 & \\
412: \botrule
413: \end{tabular}
414: \end{table}
415:
416: Group theoretical arguments\cite{duval92} show that only
417: SPH modes with $\ell$=0 and $\ell$=2 are Raman active.
418: This assumes that the nanoparticle is perfectly spherical
419: in shape and spherically symmetric in all of its properties.
420: The basic nature of the $\ell=0$ modes is much more clear
421: because of their simplicity and symmetry. Consequently,
422: the modes (SPH,$\ell=2$,$n$) are of primary interest when
423: trying to understand Raman intensities.
424:
425: From the value of L2 $\simeq$ 0.14 in Tab.~\ref{def}, the displacement
426: of (SPH,2,0) is mostly due to the $u_T$ term and not the $u_L$ term.
427: Its squared displacement due to the $u_L$ term alone is
428: just 14\% of the total. The $u_T$ term has zero divergence. Therefore,
429: (SPH,2,0) doesn't have much divergence. So the effect of changing
430: density on the dielectric constant through the deformation potential
431: may not be significant to the overall Raman intensity.
432:
433: On the other hand, based on its URS of $\simeq$ 0.66 and UTS
434: of $\simeq$ 0.20, the surface displacement of (SPH,2,0)
435: is strongly along $r$ and only weakly along $\theta$ as
436: Fig.~\ref{prf12} illustrates. Note that, $r$ surface
437: displacement changes the nanoparticle shape, while $\theta$
438: displacement does not.
439:
440: The (SPH,2,1) mode differs from (SPH,2,0) in several ways. From the
441: $L1$ value of 0.2281 in Fig.~\ref{Poisson2}, we can see that the
442: frequency of (SPH,2,1) depends more on $v_L$. Also, $L2 \simeq 0.416$
443: in Fig.~\ref{Poisson2} shows that (SPH,2,1) has more of a $u_L$
444: component, even if it is still weaker than the $u_T$ part. But this
445: means that (SPH,2,1) can have much more divergence than (SPH,2,0).
446: So the deformation potential mechanism can modulate the dielectric
447: constant. But it is very interesting to notice from the URS value of
448: $\simeq$ 0.00 in Tab.~\ref{def} (more precisely, 0.0003) that (SPH,2,1)
449: causes negligible radial movement of the surface. So (SPH,2,1) barely
450: changes the shape of the nanoparticle, as Fig.~\ref{prf12} shows.
451:
452: (SPH,2,3) has strong $v_L$
453: dependence ($L1 \simeq 0.8475$) in Fig.~\ref{Poisson2} as well
454: as a strong $u_L$ component ($L2 \simeq 0.766$). So it is clear
455: that it is SPH$_L$. It's surface displacement is mostly along
456: $r$ and not $\theta$ from its URS value of 0.71 and UTS $\simeq$
457: 0.04. So (SPH,2,3) will strongly affect the shape of the
458: nanoparticle surface, as shown in Fig.~\ref{prf12}.
459:
460: $u_L$ and $u_T$ take on simpler forms as $\eta$ becomes
461: larger. For large $\eta$, the $u_L$ term has primarily
462: radial displacement, while the $u_T$ term corresponds
463: to displacement in the $\theta$ direction.
464:
465: For the lowest modes, the situation is qualitatively different.
466: Consider (SPH,2,0) with $v_T/v_L = 0.5$. Suppose to simplify
467: this discussion we normalize the displacement field so that
468: $(u|u) = 1$. Then $(u_L|u_L) \simeq 0.14$. However,
469: $(u_T|u_T) \simeq 1.85$. So L3 for (SPH,2,0) is actually $\simeq$ -0.85,
470: making it ``ultra transverse'' by that measure. It seems quite
471: odd that the $u_T$ term alone has a magnitude much greater than
472: that of the overall motion. The resolution of this puzzle is
473: that $u_L$ and $u_T$ are not mutually orthogonal with respect to
474: the inner product of Eq.~\ref{norm}. In fact, $(u_T|u_L)$
475: $\simeq$ -0.50. According to the usual vector relation,
476: $\vec{a} \cdot \vec{b}$ = $\| \vec{a} \| \| \vec{b} \| cos \theta_{ab}$,
477: the ``angle'' between $u_L$ and $u_T$ is $\simeq$ 165 degrees
478: for the (SPH,2,0) mode. This angle is nearly unchanged as
479: $v_T/v_L$ varies. Thus, $u_L$ and $u_T$ are nearly antiparallel
480: vectors in the function space of vector fields within the
481: nanoparticle interior. It can be said, then, that the functional
482: forms of $u_L$ and $u_T$ are actually relatively similar. This
483: is a bit of a surprise since one is curl-free while the other is
484: divergence free. This angle between $u_L$ and $u_T$ rapidly
485: approaches 90 degrees as $\eta$ increases (\textit{i.e.} for modes with
486: higher $n$).
487:
488: As Fig.~\ref{SPH2} shows, the starkness of the contrast between
489: SPH$_L$ and SPH$_T$ modes is at its best when $v_T/v_L$ is lower. For
490: materials with high $v_T/v_L$ such as Si and Ge, FSM modes tend more to
491: be mixtures of SPH$_L$ and SPH$_T$, especially at low $\eta$. But
492: the concept of longitudinality is quite applicable to materials such as
493: Au and Ag.
494:
495: \section{High frequency mode classification}
496: \label{Highfreq}
497:
498: The reason for the dichotomy of SPH modes as SPH$_T$ and SPH$_L$ can be
499: simply explained in the high frequency limit. Consider $\Delta_{\ell}$,
500: the 2 $\times$ 2 determinant in Eq.~\ref{twobytwo}, and its four matrix
501: elements. Note that $\xi / \eta$ = $v_T/v_L$. So at high frequency,
502: both $\eta$ and $\xi$ are large. In that case, $T_{11}$ and $T_{43}$
503: will be much larger than $T_{13}$ and $T_{41}$ because of their terms
504: including factors of $\eta^2$. Consequently, $\Delta_{\ell}$ is very
505: well approximated by $T_{11} T_{43}$. Since normal modes correspond to
506: zeroes of $\Delta_{\ell}$, it is clear that there will be two sets of
507: modes: those which are approximately zeroes of $T_{11}$ and $T_{43}$
508: respectively. The first group are SPH$_L$ and the second group are
509: SPH$_T$.
510:
511: The roots of $T_{11}$ and $T_{43}$ are plotted
512: versus $\ell$ in Fig.~\ref{etavsl} with lines with crosses for
513: SPH$_L$ modes and lines with empty squares for SPH$_T$ modes.
514: The lines with full circles are the exact FSM mode frequencies.
515:
516: There are three kinds of situations where we don't
517: expect this approximation to be valid: (1) for low $\eta$
518: (2) when it is not true that $\eta \gg \ell$
519: and (3) where longitudinal
520: ($T_{11}$) and transverse ($T_{43}$) modes for a given $\ell$
521: are close -- \textit{i.e.} when the associated curves cross. Except
522: in the previously mentioned places, the agreement between FSM
523: and our approximation is quite good. The low $\eta$
524: situation corresponds specifically to the similar prefactors
525: of $j_{\ell}$ for T$_{11}$ and T$_{43}$ not being large. It
526: is apparent that T$_{11}$ and T$_{43}$ can only be useful
527: as estimators of SPH$_L$ and SPH$_T$ mode frequencies when
528: $\eta \gg \ell$. This is confirmed from inspection of the
529: lower right portion of Fig.~\ref{etavsl}.
530:
531: For large $x$, $j_{\ell}(x) \simeq \sin(x-\ell\frac \pi 2) / x$.
532: Therefore, for large $\xi$ and $\eta$, the roots of $T_{11}$ can
533: be approximated by $\xi \simeq \ell \frac\pi 2 + (1+n_L) \pi$ and the
534: roots of $T_{43}$ by $\eta \simeq \ell \frac\pi 2 + n_T \pi$ where
535: $n_L \geq 0$ and $n_T \geq 0$ are integers.
536: These lead to remarkably compact approximate expressions for SPH$_L$
537: and SPH$_T$ FSM frequencies in Hertz, respectively:
538: \begin{equation}
539: f \simeq \frac{v_L}{d} ( \frac{\ell}{2} + n_L + 1 )
540: \end{equation}
541: \begin{equation}
542: f \simeq \frac{v_T}{d} ( \frac{\ell}{2} + n_T )
543: \end{equation}
544: where $d$ = $2 R$. These expressions are very suggestive
545: of the formula for acoustic standing waves in a one dimensional
546: system of length $d$.
547: Table~\ref{def} shows the value of either
548: $n_L$ or $n_T$ for each mode.
549:
550: The behaviour observed in Fig.~\ref{SPH2} becomes simple to
551: explain. To a good approximation, SPH FSM modes are either
552: SPH$_L$ or SPH$_T$. This approximation is considered here to be
553: good because it gives the right number of vibrational modes and
554: it predicts their frequency with a reasonable accuracy.
555:
556: ``Anti-crossing'' is observed in Fig.~\ref{SPH2} each time the
557: variation of the frequency of a SPH$_L$ mode crosses the one of
558: a SPH$_T$ mode. In Fig.~\ref{SPH2} there are two kinds of curves:
559: horizontal lines for SPH$_T$ modes and descending curves for
560: SPH$_L$ ones. Then, each time these curves come together, an
561: anti-crossing pattern appears for the FSM solutions. In
562: Fig.~\ref{etavsl}, FSM frequencies $\eta$ are plotted versus
563: $\ell$ for a sphere made of a material which has $v_T/v_L$ = 0.5.
564: Because the SPH$_L$ and SPH$_T$ approximation curves are plotted,
565: the anti-crossing patterns are clearly revealed. The
566: continuation of Bessel functions to non-integer $\ell$ permits
567: relationships among modes for different integer $\ell$ to be
568: clearly seen. This is preferable to the common practice of
569: joining modes on such a graph with hand-drawn straight lines.
570:
571:
572: \section{Discussion:}
573: \label{Discussion}
574:
575: Normal elastic waves in a solid have a longitudinal acoustic
576: (LA) branch and two transverse acoustic (TA) branches. However,
577: for FSM it seemed there are just two kinds: SPH and TOR. By
578: classifying SPH modes into two kinds (\textit{i.e.} SPH$_L$ and
579: SPH$_T$), there are now three categories of modes, as we would
580: expect.
581:
582: We plot in Fig.~\ref{AgSPH2} the mean squared radial surface
583: displacement (URS) at the surface of a 5~nm diameter silver
584: nanoparticle for all SPH $\ell=2$ modes.
585: The magnitude of URS is in good agreement with the calculated
586: Raman intensities\cite{bachelierPRB04}. (These calculations took into
587: account the non-linear dispersion of acoustic phonons in silver. As a
588: result, the calculated vibration wavenumbers do not match.) As discussed
589: before, (the SPH,2,0) mode is quite special even if we class it as a
590: SPH$_T$ mode. It changes the surface shape and therefore contributes
591: significantly to inelastic light scattering. Other harmonics contribute
592: significantly only when their URS is large and this in turn is very well
593: correlated to their SPH$_L$ nature as can be seen in Fig.~\ref{SPH2}.
594:
595: \begin{figure}[!ht]
596: \includegraphics[width=\columnwidth]{URS2_Ag}
597: \caption{\label{AgSPH2}Mean squared radial surface displacement
598: (URS) as a function of wavenumber for (SPH,$\ell=2$) modes of
599: a 5 nm diameter silver nanoparticle ($v_T/v_L$ = 0.464).}
600: \end{figure}
601:
602: Many experiments have observed peaks in Raman spectra attributed
603: to acoustic phonon vibrations of
604: silver
605: \cite{portales01a,portales01b,duval01,fujii91,portalesthesis,courty02,NeletASS04}
606: silicon
607: \cite{fujii96,saviotPRB03}
608: and CdS$_x$Se$_{1-x}$.
609: \cite{verma99,saviot96,ivanda03,irmer00,saviot98}
610: These studies have regularly succeeded in observing the (SPH,2,0) mode
611: and the (SPH,0,0) mode. A number of studies have seen (SPH,0,$n$)
612: with $n$ up to 4\cite{NeletASS04}.
613: However, there has never been a clear indication of
614: Raman scattering from (SPH,2,1) even though there have been determined
615: efforts to see it.
616:
617: At the same time, (SPH$_L$,2,$n_L=0$) seems like a strong
618: candidate to have noticeable Raman scattering, since it has
619: strong radial surface motion as well as a strong $u_L$ component
620: that will give it stronger divergence in its interior.
621:
622: It should be noted that $\ell = 0$ modes are always SPH$_L$.
623: That is why no full circles are plotted in Fig.~\ref{etavsl}
624: on the $T_{43}$ root curves at $\ell = 0$. This has been the
625: source of many erroneous calculations in the past\cite{saviotPRE04}.
626:
627: It is often
628: claimed\cite{tanaka93,tamura82,tamura83,ovsyuk96}
629: that modes with $n$ = 0 are ``surface
630: modes'' while modes with $n > 0$ are ``inner modes''.
631: The values of U2S in Tab.~\ref{def} show that
632: this is a misconception. While this is true for $\ell$
633: = 0 and 1, for $\ell$ = 2, 3 and 4 it can be seen that
634: (SPH,$\ell$,1) has the strongest surface motion relative
635: to all (SPH,$\ell$,$n$).
636:
637: Although Tab.~\ref{def} shows URS to be zero for (SPH,2,1),
638: the more precise value of $v_T/v_L$ where URS is zero is
639: 0.488. URS for (SPH,2,1) is only near zero for $v_T/v_L$
640: close to 0.488. However, URS remains small for (SPH,2,1)
641: for materials whose Poisson ratio is close to $\frac13$
642: which is true of many common materials. This contradicts
643: a widespread
644: misconception\cite{ohno00,lomnitz05,heyliger92}
645: that SPH FSM modes always have a radial displacement
646: component at the surface.
647:
648: \begin{acknowledgments}
649: D. B. M. acknowledges support from the Natural Sciences and Engineering
650: Research Council of Canada, the Okanagan University College
651: Grant-in-Aid Fund and National Sun Yat-Sen University and thanks L. M.
652: L. Murray and A. S. Laarakker for valuable scientific suggestions.
653: \end{acknowledgments}
654:
655: %\bibliographystyle{utphys}
656: \bibliography{pr}
657: \end{document}
658: