1: \documentstyle[preprint,aps]{revtex}
2: \tightenlines
3:
4: \begin{document}
5: \draft
6:
7: \title{Scarring in vibrational modes of thin metal plates}
8:
9: \author{Andre {\it J}. Starobin and Stephen W. Teitsworth}
10:
11: \address{Duke University, Department of Physics, Box 90305, Durham, NC 27708-0305}
12:
13: \date{\today}
14: \maketitle
15:
16: \begin{abstract}
17: We report the first direct experimental observation of scarring
18: phenomenon in transverse vibrational modes of a thin metal
19: plate. The plate has the shape of a full stadium and clamped boundary
20: conditions.
21: Normal modes are imaged using time-averaged holographic
22: interferometry, and modes corresponding to ``bouncing ball'' and higher
23: order periodic trajectories are found.
24: An eikonal approximation of the
25: solution along classical trajectories of the stadium including
26: nontrivial phase shifts at clamped boundaries yields a useful quantization
27: condition for the observed modes.
28: \end{abstract}
29: \pacs{PACS numbers: 05.45.Mt, 03.65.Sq, 46.40.-f}
30:
31:
32: \maketitle
33:
34: Scarring refers to the build-up of amplitude along unstable periodic
35: trajectories in high order modes of certain wave systems. The term is
36: generally reserved for wave problems with domains which correspond to
37: classically chaotic billiards. The phenomenon was first reported in
38: numerical studies of high order, short wavelength modes of the Helmholtz
39: equation inside a domain known as the Bunimovich stadium [1]; an
40: explanation of scarring relies on methods of quantum chaos [1,2]. The
41: Helmholtz equation serves, for example, as the time-independent
42: Schr\"{o}dinger equation for a particle in a box, and also as the equation
43: governing the transverse vibration of elastic membranes. In addition to
44: numerical studies of the Helmholtz equation, several analogue experiments
45: have reported quantum chaotic spectral and spatial properties including
46: the electromagnetic field modes of quasi-two-dimensional microwave
47: cavities [3-5], stationary capillary waves on water [6], three-dimensional
48: acoustic resonances in water-filled cavities [7], and vibrational modes of
49: drumheads [8].
50:
51: A possible extension of quantum chaos methods to the high frequency limit
52: of wave problems with modes that are not described by the Helmholtz
53: equation was noted by Berry [9]. The statistical properties of spectra as
54: predicted by random matrix theory have since been confirmed in experiments
55: on elastic waves in three-dimensional metal blocks [10-13], and on studies
56: of electromagnetic modes in three-dimensional microwave cavities [14,15].
57: In both of these cases the vectorial nature of the modes destroys the
58: exact analogy with the time-independent Schr\"{o}dinger equation. More
59: recently, scar-like phenomena have been reported for the electromagnetic
60: modes of three-dimensional microwave cavities, although the role of
61: underlying periodic orbits is unclear [16].
62:
63:
64: The transverse vibrational waves of a thin plate provide another important
65: example of a wave system for which the stationary modes are not described
66: by the Helmholtz equation. In a recent numerical study [17], the computed
67: asymptotic spectrum for a fully clamped thin plate was shown to possess
68: statistical properties predicted by random matrix theory; furthermore,
69: the authors found strong evidence of scarring in vibrational amplitude
70: plots for some of the high frequency modes. More recently, a general
71: theory of scarring and spectral properties has been developed by Bogomolny
72: and Hughes for transverse vibrations in thin plates [18]. In this paper,
73: we present experimental results on the high frequency properties of
74: vibrating metal plates, with particular attention given to the
75: distinguishing spatial properties of individual high order modes. To our
76: knowledge, this is the first report of scarring phenomena in an
77: experimental vibrating plate. We also present a quantization criterion
78: for the eigenfrequencies of modes that are scarred by particular classical
79: periodic orbits. This criterion includes the effects of angle-dependent
80: phase shifts associated with reflection of periodic orbits at the fully
81: clamped boundary of a thin plate.
82:
83: For our experiment, we have studied a stainless steel plate of thickness
84: $h = 0.305$ mm. The plate is in the shape of a full stadium and consists
85: of a square central section of side $8.00$ cm and two semi-disks on each side
86: of radius $R = 4.00$ cm. Contour plots of vibrational amplitude for
87: individual
88: modes are obtained using the technique of time-averaged holographic
89: interferometry [19]. Characteristic vibrational amplitudes are of the
90: same order as the wavelength of the 1 mW frequency-stabilized HeNe laser
91: used for imaging. To achieve a fully clamped boundary condition (i.e.,
92: both amplitude and its normal derivative tend to zero at the boundary),
93: the plate edges were carefully epoxied to a massive aluminum support. The
94: interferograms of all imaged modes show that the vibrational amplitudes
95: and amplitude gradients at the boundary are very small, thus confirming
96: the effectiveness of the clamping procedure. To drive the plate
97: vibrations harmonically, we use a modified audio speaker (frequency
98: range from 50 Hz to 12 kHz) which is coupled to the plate via a thin steel
99: rod which acts as the driving rod. The rod is firmly attached to the
100: speaker voice coil at one end and lightly epoxied to the plate at the
101: other end. A scannable phonograph cartridge stylus is used to monitor the
102: amplitude response of the plate at various positions. The signal is
103: filtered through a low-noise pre-amplifier before being fed to an HP 3561A
104: signal analyzer which is used to measure the resonant frequency and
105: quality factor ($Q$) for each detectable mode. Significant damping
106: due to coupling of the plate to the support structure and the driving
107: rod sets in at approximately 12 kHz, at which point
108: the quality factors of the system become too low to resolve individual
109: modes. The fundamental frequency of the plate is approximately 200 Hz.
110:
111: In the frequency range between seven and eight kilohertz we find four
112: vibrational resonances at $7272 Hz$, $7442 Hz$, $7622 Hz$ and at $7874 Hz$.
113: The wave patterns excited at $7442 Hz$ and at $7622 Hz$ reveal prominent scarring
114: in the imaged amplitude distributions. Figure 1a shows the holographic
115: interferogram, effectively an amplitude contour plot,
116: for the $7442$ Hz frequency of excitation. The large, connected bright areas correspond to
117: places with
118: very small vibrational amplitude, while darker areas and their interiors
119: correspond to regions with large amplitude.
120: Clearly visible in Fig. 1a
121: is scarring by a ``bouncing ball" trajectory across
122: the width of the stadium plate; furthermore, we can see that there are
123: 7 antinode pairs as one traverses the complete periodic trajectory.
124: The excited wave pattern has a broken symmetry with respect to
125: $\beta$ symmetry axis of the plate, with the symmetry axis of the plate
126: and the position of the driving rod-to-plate contact designated in Figure 1b.
127: The symmetry with respect to $\alpha$ axis is partially preserved on the left side of the plate.
128: Further,
129: while the left side of the plate displays a clear build-up of amplitude
130: along the ``bouncing ball" orbit and has regular radial amplitude features
131: in left semi-disk, features characteristic of the (6,3) modes of a pure disk,
132: the amplitude distribution in the right half of the plate does not appear regular
133: and does not show any obvious scarring.
134:
135: Figure 2 shows the interferogram
136: for the $7622$ Hz mode. In this case, scarring on the
137: right side of the plate clearly follows a rectangular orbit
138: producing a ``whispering gallery" type mode. The number of antinodes
139: traversed
140: in going around the right half of the orbit is 16, which suggests that the
141: total
142: number of antinode pairs for the full orbit is 16. This resonant wave pattern
143: is almost perfectly localized on the right-hand side of the plate, the same side
144: where the speaker rod is glued to the plate. The semi-rectangular
145: scar feature is confined fully to the right half of the plate and is
146: essentially symmetric with respect to the $\alpha$ axis. Immediately to the left
147: of the $\beta$ axis we find clustering of amplitude along a vertical located
148: almost at the same position as the ``bouncing ball'' scar of the 7442Hz wave pattern.
149: As we move further to the left, the excitation amplitude decays, with
150: the first bright fringe barely resolved.
151:
152: The broken symmetry of the excited wave patterns we observe clearly indicates
153: that they are not pure modes of the underlying unperturbed plate.
154: In the 7-8 KHz range the experimental Q-factors, measured as the ratio of
155: the resonance peak position to the 3dB width of the resonance peak, are between 50 and 90.
156: This corresponds approximately
157: to a minimum 75Hz level spacing that we can resolve experimentally. The observed
158: resonance spacing in this frequency range is 200Hz which compares well with
159: the mean Weyl level spacing for this plate of 300Hz. With such quality factors in the
160: experiment we can conclude that at each excitation frequency two-to-three pure modes
161: of the system are excited.
162:
163: The 2nd Newton's law written for small amplitude flexural vibration of a
164: thin plate and the general form of the solution have the following form:
165:
166: \begin{eqnarray}
167: & -D\Delta^2\psi + F_{dephase} + F_{dr} = \rho\partial_{tt}\psi \\
168: \psi({\bf r}, t; f) = & \sum_n \{ \alpha_n \cos(2\pi ft + \gamma)\phi_n({\bf r})
169: + \beta_n\sin(2\pi ft + \gamma)\phi_n({\bf r}) \}
170: \end{eqnarray}
171: $D$ is the flexural rigidity of the plate and $\rho$ is the mass
172: per unit area of the plate. $F_{dephase}$ and $F_{dr}$ are dephasing and driving
173: forces respectively. $f$ is the excitation frequency and $\gamma$ is a fixed
174: time-independent phase shift.
175: The driving force in our experiment is harmonic and is localized at the
176: position where the speaker needle is glued to the plate. In the absence
177: of dephasing (with a proper choice of $\gamma$) $\alpha_n=\delta(f-f_n)$
178: and $\beta_n=0$. With a dephasing damping force both elastic and inelastic
179: response is present and $\alpha_n, \beta_n$ are frequency dependent coefficients.
180: In this experiment damping occurs via plate-speaker and plate-air coupling, the
181: former being the primary damping mechanism. Below, we write down expressions for
182: imaged resonant wave patterns in terms of pure modes of the system and argue that these are also
183: scarred.
184:
185: The localization of wave pattern on the right at 7622 Hz excitation
186: strongly suggests that we excite two closely spaced symmetry related modes:
187: $\phi^{-+}_{i+1}$ and $\phi^{--}_{i+1}$, with the first (second) sign in the superscript
188: indicating odd (even) symmetry with respect to $\alpha$ ($\beta$) axis. The
189: $\phi^{-+}_{i+1}$ and $\phi^{--}_{i+1}$ narrowly split mode pair
190: located near 7622Hz and the $\phi_{i}$ mode slightly lower than 7442Hz are marked on the
191: level diagram in figure 3. The opposite symmetry of the two $\phi_{i+1}$ modes
192: with respect to $\beta$ axis is
193: needed to produce the observed localization. The odd symmetry with respect to
194: the $\alpha$ axis of $\phi^{-+}_{i+1}$ and $\phi^{--}_{i+1}$
195: is supported by the observed narrowness of the nodal
196: line that runs along the $\alpha$ axis, clearly passing through the right semi-disk center
197: (cf. fig. 2). For even symmetry one would expect
198: a wider bright white region reflecting a wider interval of small excitation amplitudes.
199: A much wider bright region is always seen near the clamped edges where the derivative
200: vanishes (cf. fig 1 and 2) and the amplitude increases slower from zero than near the
201: $\alpha$ axis of the 7622Hz excitation pattern. The placement of $\phi^{-+}_{i+1}$ mode
202: on the level diagram higher than $\phi^{--}_{i+1}$ mode is motivated by a similar
203: consideration. The odd-even mode picks up extra strain energy compared to the odd-odd mode
204: along the symmetry line where its curvature is non-zero.
205:
206: With the level diagram of figure 3, the three mode expansion of the
207: wave pattern excited at $7622Hz$ has the following
208: form:
209: \begin{eqnarray*}
210: \psi(7622) = \sin(2\pi f_{exc}t)\{\beta^{-+}_{i+1}\phi^{-+}_{i+1} +
211: \beta^{--}_{i+1}\phi^{--}_{i+1} + \beta_{i}\phi_{i} \} +& \\
212: \cos(2\pi f_{exc}t)\{\alpha_i\phi_i + \alpha^{-+}_{i+1}\phi^{-+}_{i+1} +
213: \alpha^{-+}_{i+1}\phi^{-+}_{i+1} \}
214: \end{eqnarray*}
215: Since the two symmetry related modes are closely spaced with a spacing much smaller that the
216: mean level spacing,
217: $\beta^{-+}_{i+1} \approx \beta^{--}_{i+1}$ and with
218: $\beta_{i} \ll \alpha_i \approx \alpha_{i+1} \ll \beta_{i+1}$ at 7622Hz,
219: the excited wave pattern can be simplified to:
220: \begin{equation}
221: \psi(7622) = \sin(2\pi f_{exc}t)\{\beta_{i+1}\phi^{R}_{i+1}\} +
222: \cos(2\pi f_{exc}t)\{\alpha_i\phi_i + \alpha_{i+1}\phi^{R}_{i+1} \}
223: \end{equation}
224:
225: where $\phi^{R}_{i+1}$ is the sum(difference) of the two symmetry related modes with the
226: sign chosen so to produce a wave pattern localized on the right side of the plate as is
227: seen in the experiment. With the main contribution to this wave pattern localized on
228: the right we expect to find the signature of the pure $ith$ mode of the plate on the left.
229: This we believe is the origin of the amplitude clustering along the vertical just off to the left
230: of the $\beta$ symmetry axis.
231:
232: Similarly, we can now write the wave pattern excited at 7442Hz as :
233:
234: \begin{equation}
235: \psi(7442) = \sin(2\pi f_{exc}t)\{\beta_i\phi_i \} +
236: \cos(2\pi f_{exc}t)\{\alpha_{i+1}\phi^{R}_{i+1} + \alpha_{i}\phi_{i} \}
237: \end{equation}
238:
239: The perturbing term is fully localized on the right hand side of the plate.
240: We than expect to find a signature of the $ith$ mode of an unperturbed plate to the left
241: of the $\beta$
242: symmetry axis and a wave pattern arising due to superposition of the $ith$ and
243: the $(i+1)st$ doublet on the right. Indeed, this is what we observe. The
244: left side of the pattern in figure 1a is scarred by a ``bouncing ball'' orbit, while
245: the right side exhibits no visible scarring, or regularity.
246:
247: In an earlier work by Sridhar and Heller a localization effect similar to
248: ours has been observed [9]. In that work resonant wave patterns of a MW cavity slightly
249: perturbed from Sinai billiard shape were studied.
250: There the authors of [9] were able
251: to mix closely spaced symmetry related scarred modes of a Sinai shaped MW cavity, by slightly
252: displacing the disk scatterer in the middle of the plate, with localization occurring on the
253: side away from the displacement. When the symmetry of the cavity was reestablished the
254: two mixed modes were split again. In our experiment we did not
255: attempt to resolve individual symmetry related modes due to relatively low
256: Q-factors of the excitation system.
257:
258: To understand the scarring observed in high order modes within the simplest version
259: of the PO formalism, we assume that
260: the time-independent part of the solution along the scarred trajectories has a
261: form $\displaystyle u = A_\Gamma e^{i S_\Gamma}$, where
262: $A_\Gamma$ is
263: assumed to be a single valued function of position and is assumed to be
264: nearly constant in the neighborhood of a classical trajectory $\Gamma$[18].
265: $S_\Gamma(q_{\perp}, q_{\parallel})$ is a multivalued function written in
266: terms
267: of local coordinates $q_{\perp}$ and $q_{\parallel}$, along and
268: perpendicular to
269: the trajectory.
270:
271: In the high frequency limit the form of $S_\Gamma(q_{\perp},
272: q_{\parallel})$ is
273: restricted. Far from the boundaries, sources, and caustics, $S_\Gamma$ is
274: nearly
275: linear in both $q_{\perp}$ and $q_{\parallel}$. Moreover, $S_\Gamma$ is a
276: rapidly growing function of $q_{\perp}$ and $q_{\parallel}$ in the limit
277: of high order modes. In the regime where $S_\Gamma(q_{\perp}, q_{\parallel})$
278: is nearly linear, it is well described by an eikonal equation
279: for a thin plate:
280: \begin{equation}
281: (\nabla S_\Gamma)^4 = \frac{12\rho(1-\sigma^2)}{Eh^2} {\omega_0}^2,
282: \end{equation}
283: where $\omega_0$ denotes the frequency of the orbit. Unlike the eikonal
284: equation for an ideal membrane, Eq. $(6)$ is
285: only valid up to frequencies for which volume deformations become
286: appreciable, i.e., such that $\displaystyle\frac{\Delta^2 u({\bf
287: r})}{\lambda^3}
288: \sim \frac{h^3}{\lambda^3} \sim 1$, where $u({\bf r})$ is the amplitude
289: function for the mode and $\lambda$ is the effective wavelength.
290: At such short wavelengths and high frequencies the biharmonic wave equation
291: itself
292: breaks down. For the experimental results reported here we never reach
293: this regime
294: because $\displaystyle\frac{h^3}{\lambda^3} \sim 10^{-3} \ll 1$.
295:
296: It can be seen that there are two types of solution of Eq. (6). $S_\Gamma$ is
297: either a real or complex function of position. Real $S_\Gamma$ gives rise to
298: propagating wave solutions which are plane wave-like in a homogeneous
299: plate, exactly as for a membrane. Complex $S_\Gamma$ yields exponentially
300: decaying, or rising solutions of the wave equation. Each mode of the plate
301: must be bounded, and hence the exponentially growing solutions cannot be
302: present far from the boundary. On the other hand, the decaying solutions
303: decay to zero for distances significantly larger than a wavelength. This
304: can be seen already in the one-dimensional clamped rod problem which
305: serves as a useful guide in understanding various aspects of the
306: two-dimensional
307: plate [20,22].
308:
309: It is straightforward to show that a plane wave propagating in a thin
310: plate incident at an angle $\alpha$ with respect to the normal onto a
311: clamped plate edge, is reflected with a phase shift given by:
312: \begin{equation}
313: \delta = -2\arccos\left(\sqrt{\frac{1 +
314: \sin^{2} \alpha}{2}}\,\right).
315: \end{equation}
316: This expression can be obtained in the standard way by matching incoming
317: and outgoing plane wave solutions under the condition that the wavelength
318: is much smaller than the local boundary curvature [18]. Then, a simple
319: quantization condition may be written for a trajectory $\Gamma$ with $N$
320: boundary reflections in the form:
321: \begin{equation}
322: k_{\parallel}L - 2\sum\limits_{i=1}^{N}\arccos\left(\sqrt{\frac{1 +
323: \sin^{2} {\alpha}_{i}}{2}}\,\right) = 2\pi n,
324: \end{equation}
325: where $k_{\parallel}$ is a wavevector component along the periodic
326: trajectory, $L$ is the total orbit length, and $n$ is an integer
327: ``quantum'' number.
328:
329: For the $7442$ Hz mode, the primary orbit of interest is the
330: ``bouncing ball" one.
331: The amplitude build-up along the
332: ``bouncing ball" orbit in Fig. 1 (a) is clearly evident, and, by
333: inspection, we expect
334: the quantum number associated with this orbit to be $n = 7$. Substituting
335: this
336: into Eq. (8) along with the appropriate $\alpha$ values for the orbit (i.e.,
337: $\alpha_1 = \alpha_2 = 0$), we obtain
338: $k_{\parallel} = 295$ m$^{-1}$. Then, we may use Eq. (1) to estimate a
339: frequency contribution of $f_{\parallel} = 6675$ Hz. Of course, the total
340: frequency
341: depends on $k_{\parallel}^{2} + k_{\perp}^{2}$, so we use the difference of
342: the
343: observed $f$ and calculated $f_{\parallel}$ to estimate an effective
344: $k_{\perp} =
345: 100$ m$^{-1}$.
346: This corresponds to an effective ``perpendicular" wavelength of
347: $\lambda_{\perp} = 6.30$ cm,
348: which is comparable to the full width of the straight section of the plate.
349: This observation is consistent with numerical studies of the Helmholtz
350: equation, in which wavefunction scars of
351: ``bouncing ball" trajectories have an effective perpendicular wavelength which
352: is slightly less than the full width of the straight section of the stadium
353: boundary [1].
354:
355: For the rectangular orbit that scars the $7622$ Hz mode, there are four
356: bounces
357: each with $\alpha_i = \pi/4$. Using a quantum number of $n = 16$, we
358: determine that
359: $k_{\parallel} = 271$ m$^{-1}$ and $f_{\parallel} = 5644$ Hz, considerably
360: less than
361: the measured frequency. In this case, the
362: estimated perpendicular wavelength is rather small, $\lambda_{\perp} =
363: 3.90$ cm.
364: This can be understood with the following argument.
365: First, we note that the amplitude maxima of the rectangular orbit along the
366: top straight edge of the plate
367: should occur very near to the classical trajectory which lies a distance
368: $R(1-\frac{1}{\sqrt{2}}) = 1.17$ cm
369: below the plate edge in Fig. 2. Next, we consider the related problem of
370: a long rectangular strip of plate which is
371: clamped on one side and simply-supported on the other side (i.e., the amplitude
372: and its second normal derivative vanish). Furthermore, the width $l$ of
373: the strip
374: is selected such that the amplitude maximum of the lowest
375: order vibrational mode occurs along the scarring orbit. This
376: problem is easily solved using standard methods [20, 22], and we find
377: $l = 2.02$ cm. We expect that the length $l$ should be close to the
378: above estimate based on the quantization condition
379: for one-half of the perpendicular wavelength, $\lambda_{\perp}/2$.
380: We find $\lambda_{\perp}/2 = 1.95$ cm, in good agreement.
381:
382: We have observed qualitatively similar scarring effects in clamped metal
383: plates that have non-stadium boundary shapes, but which are predicted to be
384: chaotic
385: for the Helmholtz equation,
386: e.g., one-quarter of a ``bow-tie" shape [24]. We also
387: studied the spatial properties of vibrational modes in clamped plates with
388: a circular boundary; in this case no significant scarring was found.
389: We attempted to measure spectral statistics, but the present experimental
390: set-up is not suitable for obtaining accurate results. The point
391: driving source does not allow detection of those modes with
392: nodal lines close to the driving point. Furthermore, if one
393: moves the driving point to another location the eigenfrequencies
394: shift due to coupling between the plate and the driving rod [8,24],
395: so that one cannot combine spectra taken for different
396: drive positions in any analysis of spectral statistics.
397:
398: We have presented direct experimental evidence for scarring in the vibrational
399: modes of a thin metal plate in the shape of a stadium. We have also introduced
400: a novel quantization condition
401: which is useful for
402: estimating the
403: experimental resonant frequencies of those modes that correspond to unstable
404: periodic orbits of the underlying classical billiard system.
405: It seems remarkable that
406: scarring effects - predicted on the basis of semiclassical theories
407: that include no frictional
408: terms whatever - are clearly visible in this relatively low $Q$ system.
409: Further, it is surprising that of the three wave patterns we have imaged
410: two were scarred by short periodic orbits. The imaging was done only at those
411: excitation frequencies where the response of the plate was especially strong, since at the
412: lower amplitudes the quality of interferograms is poor. At the 7442Hz and especially at 7622Hz
413: excitation frequency one could actually hear the response of the plate. The Q-factors
414: at these excitation frequencies were also largest detected. It is possible that due to
415: reduced overlap between scarred and neighboring non-scarred modes the leakage of energy
416: is suppressed and the Q-factors are enhanced. If true, such an effect would be generic
417: for wave systems exhibiting scarring, with relatively small,
418: but non-negligible losses. Further
419: experiments on bimorph quartz piezoelectric plates which would allow a better coupling
420: scheme are being planned to explore this possibility
421: and to improve on already existing measurements.
422:
423: We thank Prof. K. Matveev for helpful discussions, and thank
424: Lester Chen, Dean Moyar, and Thomas Hund for help
425: with the experimental set-up.
426:
427: \begin{references}
428:
429: \item E. J. Heller, Phys. Rev. Lett., {\bf 53},
430: 1515 (1984).
431: \item E. B. Bogomolny, Physica D, {\bf 31},169 (1988).
432: \item H.-J. St\"{o}ckmann and J. Stein , Phys. Lett. {\bf 64}, 2215
433: (1990).
434: \item S. Sridhar, Phys. Rev. Lett. {\bf 67}, 785 (1991).
435: \item J. Stein and H.-J. St\"{o}ckmann, Phys. Lett. {\bf 68}, 2867
436: (1992).
437: \item R. Blumel, I. H. Davidson, W. P. Reinhardt, H. Lin,
438: and M. Sharnoff, Phys. Rev. A {\bf 45}, 2641 (1992).
439: \item P. A. Chinnery and V. F. Humphrey, Phys. Rev. E {\bf 53}, 272
440: (1996).
441: \item S. W. Teitsworth, to appear in {\it Proceedings of the 16th
442: Sitges
443: Conference} (Springer, Berlin, 1999).
444: \item S. Sridhar, E.J. Heller, Phys. Rev. E {\bf 46}, 1728, 1992
445: \item M. V. Berry, Proc. R. Soc. London, Ser.
446: A, {\bf 413}, 183 (1987).
447: \item R. Weaver,
448: J. Acoust. Soc. Am. {\bf 85}, 1005 (1989).
449: \item O. Bohigas, O.
450: Legrand, C. Schmit, D. Sornette, J. Acoust.
451: Soc. Am. {\bf 89}, 1456 (1991).
452: \item D. Delande,
453: D. Sornette, R. Weaver,
454: J. Acoust. Soc. Am. {\bf 96}, 1873 (1994).
455: \item C. Ellegard, T. Guhr, K. Lindemann, H. Q. Lorensen, J.
456: Nygaard, and
457: M. Oxborrow, Phys. Rev. Lett. {\bf 75}, 1546 (1995).
458: \item S. Deus, P. Koch, and L. Sirko, Phys. Rev. E, {\bf 52}, 1146
459: (1995).
460: \item H. Alt {\it et al.}, Phys. Rev. E {\bf 54}, 2303 (1996).
461: \item U. Dorr, H.-J. Stockmann, M. Barth, and U. Kuhl, Phys. Rev.
462: Lett.
463: {\bf 75}, 1030 (1998).
464: \item O. Legrand, C. Schmitt,
465: D. Sornette, Euorophys. Lett. {\bf 18}, 101 (1992).
466: \item E. Bogomolny and E. Hughes, Phys. Rev. E {\bf 57}, 5404 (1998).
467: \item Charles M. Vest, {\it Holographic interferometry} (Wiley,
468: New York, 1979).
469: \item L. D. Landau and E. M. Lifshitz, {\it Theory of Elasticity},
470: 2nd ed. (Pergammon, London, 1970).
471: \item Philip M. Morse and K. U. Ingard, {\it Theoretical Acoustics},
472: (Princeton University Press, Princeton, 1986).
473: \item A. Starobin, undergraduate thesis, Duke University (1997).
474: \item A. Starobin and S. Teitsworth, unpublished (1997).
475: \item T. M. Antonsen, E. Ott, Q. Chen, and R. N. Oerter, Phys. Rev. E
476: {\bf 51}, 111 (1995).
477:
478: \end{references}
479:
480: \newpage
481:
482: \begin{figure}
483: \caption{ (a) Holographic interferogram
484: for the $7442$ Hz mode.
485: The white dashed line superimposed on the hologram is a
486: guide to the eye. (b) Plate schematic. $\alpha$ and $\beta$ are the symmetry axis. $R_s$ is
487: the position where the speaker needle was glued to the plate }
488: \end{figure}
489:
490: \begin{figure}
491: \caption{ Holographic interferogram
492: for the $7622$ Hz mode.
493: A rectangular orbit is superimposed on the right half of the image
494: as a guide to the eye. }
495: \end{figure}
496:
497: \begin{figure}
498: \caption{the proposed level diagram of an underlying pure system
499: unperturbed by the speaker-plate coupling and air damping. Each mode
500: of an unperturbed plate is marked by the orbit that scars it. (-/+, -/+)
501: notation is used to show the symmetry class of a given mode. The first
502: index shows the symmetry with respect to the $\alpha$ axis, the second with respect to the $\beta$
503: axis. Minus sign stands for odd symmetry, plus for even. }
504:
505: \end{figure}
506:
507: \end{document}
508: