Real Time Imaging of Surface Acoustic Waves on Crystals and Microstructures

Oliver B. WRIGHT
+
, Osamu MATSUDA and Yoshihiro SUGAWARA
1
Department of Applied Physics, Graduate School of Engineering, Hokkaido University, Sapporo 060-8628, Japan
1
School of Physics and Astronomy, University of Southampton, SO17 1BJ, Southampton, UK
(Received November 12, 2004; accepted February 13, 2005; published June 24, 2005)
The use of acoustic pulses to image materials is well-known in echography or sonar applications. We are extending this eld
by generating point-excited sound pulses on solid surfaces with ultrashort laser pulses and imaging the resulting surface wave
propagation in real time. To see the tiny vibrations of the surface, smaller in amplitude than the dimensions of a single atom,
we use another set of laser pulses for scanned probing. The typical surface phonon wavelength is of the order of a few
microns, corresponding to frequencies up to 1 GHz. With such a system we can watch coherent acoustic wave packets in two
dimensions rippling across crystal surfaces and microscopic landscapes. [DOI: 10.1143/JJAP.44.4292]
KEYWORDS: laser ultrasonics, picosecond, interferometer, surface acoustic waves, imaging, crystal, microstructure
1. Introduction
The human ear as an acoustic instrument is well adapted
for the spectral resolution of dierent tones, but not for
visualizing the individual waves of sound in space. By
casting a stone into a pond one can immediately appreciate
the spreading of circular water ripples. This prompts the
question: how could one watch acoustic ripples on solids?
Such ripples, best known for their role in earthquakes, were
rst explained by Lord Rayleigh in 1881.
1)
Just like water
ripples, the surfaces of solids provide a guiding mechanism
for acoustic waves. Rather than resorting to the radar
mapping of earthquakes from satellites,
2)
we present in this
paper a more practical means for watching ripples on solids,
or surface acoustic waves.
3)
Surface acoustic waves in anisotropic solids can exhibit
complex propagation phenomena owing to the nature of the
fourth-order elastic constant tensor. Even in a homogeneous
crystal a point acoustic source can lead to singularities in
acoustic ux in certain directions owing to the angular
dependence of the phase and group velocities of the acoustic
waves.
4)
With the propagation vectors conned to two
dimensions the acoustic energy is concentrated in certain
directions giving rise to cusps, a manifestation of caustic
behaviour. The real-time visualisation of surface acoustic
waves thus oers the possibility for studying the temporal
evolution of cuspidal structures. Apart from the aesthetic
appeal of such phonon focusing patterns, interest in surface
acoustic waves has recently intensied because surface
acoustic wave lters, made up of microscopic interdigital
electrodes on piezoelectric crystal substrates, are widely
used in communication systems and cellular phones. Such
lters handle gigahertz electronic signals at micron acoustic
wavelengths; for a given frequency, the wavelength of sound
in crystals is 10
5
times smaller than that of electro-
magnetic waves, allowing the convenient manipulation of
electronic signals through micron-scale engineering. In
addition, surface acoustic waves penetrate into the support-
ing solid to a depth of the order of the acoustic wavelength,
and are sensitive to the presence of thin lms. This allows
surface acoustic wave devices such as gas sensors to be
designed or lm thicknesses to be measured with nanometer
resolution from an analysis of the propagation of surface
acoustic wave packets of micron order in wavelength.
Imaging methods for surface acoustic waves were
developed over the same period as those for bulk phonons.
4)
A variety of methods have been proposed,
511)
but until
recently
3,12)
none of these allowed two-dimensional imaging
of surface acoustic waves on anisotropic substrates in real
time. We review here recent progress we have made in a
non-contact technique for imaging surface acoustic waves in
real-time using ultrashort optical pulses for excitation and
detection.
3,1316)
Surface acoustic waves are impulsively
generated at frequencies up to 1 GHz at a point source on the
surface of opaque materials. Acoustic wave fronts with an
omnidirectional acoustic wave vector distribution are
tracked with picosecond temporal and micron spatial
resolution. After outlining the experimental technique we
present results from representative samples with isotropic
and anisotropic substrates, including microstructures, and
describe how the acoustic dispersion relations can be
extracted from the data.
2. Experimental Technique
Experiments are carried out with metallic thin lms and
microstructures on isotropic and anisotropic substrates. The
basic setup for transparent substrates is shown in Fig. 1,
making use of an ultrashort optical pulses for the generation
and detection of surface acoustic waves with picosecond
time resolution.
3)
Recently a similar method using nano-
second laser pulses for generation and a continuous wave
laser for detection at frequencies up to 1 MHz was
developed.
12)
In our setup the surface acoustic waves are
excited by blue pump pulses of duration 1 ps, repetition
Fig. 1. Basic imaging apparatus for samples with transparent substrates.
Inset: imaging apparatus for samples with opaque substrates.
+
E-mail address: assp@kino-ap.eng.hokudai.ac.jp
Japanese Journal of Applied Physics
Vol. 44, No. 6B, 2005, pp. 42924296
#2005 The Japan Society of Applied Physics
4292
Review Paper
rate 80 MHz (one pulse every 12.5 ns), wavelength 415 nm,
and incident energy per pulse 0.10.3 nJ, producing a
maximum transient temperature rise 100 K. These optical
pulses are derived from the second harmonic output of a
mode-locked Ti:sapphire laser. The pump light is focused at
normal incidence through a 50 or 100 long-working-
distance microscope objective lens from the transparent-
substrate side of the sample to a spot of diameter 1{3 mm.
The generation in our metallic samples is governed by the
thermoelastic eect, producing surface acoustic waves of
wavelength 3{30 mm and frequencies 100 MHz1 GHz.
The dominant wavelength of the thermoelastically generated
surface acoustic waves, typically 10 mm, is determined by
a combination of the optical spot size, the sample geometry
such as lm thickness, and diusion processes. The theory
of thermoelastic generation has not yet been fully worked
out for a metal lm on either isotropic or anisotropic
substrates.
17,18)
The surface acoustic wave detection is done interfero-
metrically by two 1 ps optical probe pulses in the infrared
of wavelength ' = 830 nm (derived from the same laser)
that are temporally separated by 300500 ps from one
another, and focused at normal incidence through another
objective lens onto the front surface of the lm. We use a
highly stable Sagnac common-path interferometer, described
in detail elsewhere,
14,19)
that allows a direct measurement of
the optical phase change (o) caused by the sample surface
displacement (oz = 'o,4). A signal proportional to the
phase dierence between the two probe pulses is obtained,
related to the normal surface displacement of the sample
(10 pm). Because the signal is proportional to the dier-
ence in phase between two probe pulses, we are essentially
imaging the out-of-plane component of the velocity of the
surface.
For samples with opaque substrates we use the apparatus
shown in the inset of Fig. 1. The p-polarized pump light is
focused onto the samples at an angle of 60

through a 20
long-working-distance microscope objective lens. Cylindri-
cal lenses are used to correct the optical spatial prole to
produce a pump beam spot with axial symmetry (approx-
imately Gaussian) on the sample.
13)
By the use of a variable optical delay between the pump
and probe and by raster scanning the probe beam laterally
over the sample, animations of the surface acoustic wave
propagation can be obtained with micron lateral spatial
resolution (using typically 1020 images). The typical
magnitude of the detected signals, expressed as a dierence
in surface displacement as measured by the two probe
pulses, is 5 pm. The typical noise level corresponding to
the above image acquisition rates is 0.3 pm (peak-to-peak),
limited by laser intensity uctuations and beam pointing
stability. The high resolution for measurement of displace-
ment using optical interferometry is enabled because the
interference process converts mechanical motion into an
intensity change. Our ability to measure small intensity
changes implies that the interferometer resolution is much
smaller than the optical wavelength ', in our case
',3000000.
3. Results for Homogeneous Samples
Isotropic samples produce circular wave fronts. Figure 2
shows an experimental surface acoustic wave image for a
200 mm200 mm region of a sample of crown-glass of
thickness 1 mm coated by thermal evaporation with a 70 nm
lm of polycrystalline gold. Gold has the advantange of a
relatively strong optical absorption in the blue at the
excitation wavelength and a high reectivity in the infrared
at the detection wavelength. The concentric series of rings
arises because of the 12.5 ns period for the arrival of pump
optical pulses at the centre of the pattern. The circular shape
is expected because of the polycrystalline nature of the lm
and the isotropic substrate. (We expect the lm to show at
least axial symmetry owing to the symmetry of the
deposition process.). These images are supercially similar
to that of the surface of a liquid disturbed at a point. Like
ripples on a liquid surface, surface acoustic waves in thin
lms are subject to dispersion, and this tends to increase the
width of the pulse as it propagates outwards. The smaller
(wavelength) ripples, more inuenced by the denser gold
lm, travel more slowly than the larger ripples, leading to a
dispersion opposite to that of ripples on water. The analysis
of the dispersion of these ripples is possible using the
general methods described below for the anisotropic case.
The sound velocity is approximately 3 kms
1
in this sample.
Anisotropic samples produce complex wave fronts. Let us
consider a simple example of a crystal with cubic symmetry:
Figure 3 shows two consecutive 100 mm100 mm exper-
imental surface acoustic wave images separated by a time
interval of 1.56 ns for the (100) surface of a transparent slab
of LiF of thickness 1 mm coated with a 50 nm polycrystal-
line gold lm. The x and y axes correspond to the [010] and
[001] directions. LiF is cubic crystal which, from a
consideration of ratios of its elastic constants, is expected
to have a folded group velocity surface with 4-fold
symmetry and to show surface phonon focusing. The
experimental wave fronts are rounded square shapes.
Because of the impulsive optical excitation the surface
acoustic waves travel outwards from the centre of the pattern
as coherent wave packets. It is the group velocity that
determines the speed of the acoustic wave packet. The
direction-dependent group velocities for surface acoustic
waves (SAW), exhibiting characteristic cusps, and pseudo-
surface acoustic waves (PSAW) obtained by analytical
calculation in the absence of lm loading
3)
are also shown in
Fig. 3; by comparing the bright ring of diameter 50 mm in
Fig. 2. Experimental image of surface acoustic wave fronts for a crown
glass substrate coated with a 70 nm gold lm.
Jpn. J. Appl. Phys., Vol. 44, No. 6B (2005) O. B. WRIGHT et al. 4293
the experimental images with the form of the predicted SAW
and PSAW wave fronts one can deduce that the SAW
component is dominant for angles within 25

to the [010]
directions, whereas the PSAW component is dominant for
angles within 20

to the [011] directions. The typical sound

velocity of the SAW is 4 kms
1
for [010] propagation.
The anisotropy factor of cubic LiF, A = 2c
44
,(c
11
c
12
) -
1.8 (c
i j
the elastic constants),
3)
is suciently high (A = 1.0
for isotropic materials) for cusps to appear in the group
velocity surfaces. In spite of this, it is interesting that the out-
of-plane detection (combined with the circular symmetry of
the laser excitation) tends to smooth the wave fronts. It is in
fact well known for propagation on the (100) plane of cubic
crystals for A > 1 that in the [011] directions the SAW
becomes a pure shear horizontal bulk mode with displace-
ment vector parallel to the surface.
20)
We therefore expect
the contribution from SAW to disappear in these and nearby
directions for our out-of-plane displacement detection, as
observed. In contrast the PSAW for the [011] directions has
a pure sagittal plane displacement component (being in fact
a non-leaky wave in these special directions), and can
therefore be detected, whereas the PSAW for the [010]
directions is expected to have a relatively low intensity
because of a reduced phonon focusing eect.
21)
The
calculated group velocity for quasilongitudinal bulk waves
(LBW-that travel close to the surface) is also shown in
Fig. 3. Wave fronts for LBW are visible in the experimental
images (see the isolated darker ring of diameter 80 mm). As
was the case for the glass substrate, LiF coated with gold
also shows wave-packet broadening caused by acoustic
dispersion associated with the nite lm thickness.
We have also imaged a transparent slab of TeO
2
,
3)
a
tetragonal crystal with a much stronger anisotropy. Figure 4
shows two consecutive 100 mm100 mm surface acoustic
wave images separated by a time interval of 1.56 ns for the
(001) surface of single-crystal TeO
2
of thickness 1 mm
coated with a 40 nm polycrystalline gold lm. (Note that the
sign of the SAW signals for TeO
2
in refs. 3 and 15 is in
error.) The x and y axes correspond to the [100] and [010]
directions. The pattern shows the expected 4-fold symmetry
for this cut.
22)
The complicated wave fronts in experiment
arise from the strong anisotropy and resulting caustic
behavior evident in the group velocity predictions also
shown in Fig. 4 for the case of negligible lm loading (and
ignoring small piezoelectric eects). The PSAW and SAW
wave fronts not observed correspond in this gure to the
crossed curves labeled PSAW (in the 0 to 17

range
measured from [100]) and the diagonal curves labelled
SAW, respectively. The square wave fronts (joining the four
vertices of the star shape) correspond to LBW. These are
also faintly resolved in experiment (see also refs. 3 and 15).
Representative sound velocities estimated from experiment
for SAW and LBW are 2.5 and 3.2 kms
1
for [100].
These surface phonon focusing patterns are particularly
striking because of the extreme anisotropy of TeO
2
; in fact
this crystal possesses the highest known anisotropy among
tetragonal materials.
23)
Dispersion analysis for such data is best carried out from a
full animation of 1020 images covering the 12.5 ns time
period between consecutive laser pulses. We take a dual
spatial and temporal Fourier transform of the data to directly
obtain the acoustic dispersion relation in o-k space.
15)
Instead of directly working with the dual Fourier transform
of the full set of data f (r, t), we add to f (r, t) an inverted
100 m 100 m
[100]
[
0
1
0
]
G
R
O
U
P

V
E
L
O
C
I
T
Y


(
k
m
/
s
e
c
)
0.5
0.5
0.0
4
4
0
4 4 0
LBW
PSAW
SAW
GROUP VELOCITY (km/sec)
Fig. 4. Top: two consecutive experimental images temporally separated
by 1.56 ns of surface acoustic wave fronts for a TeO
2
(001) surface coated
with a 40 nm gold lm. Bottom: theoretical group velocity surfaces for
TeO
2
(001), ignoring lm loading, for surface acoustic waves (SAW),
pseudo-surface acoustic waves (PSAW), and quasilongitudinal acoustic
waves (LBW).
100 m 100 m
[010]
[
0
0
1
]
G
R
O
U
P

V
E
L
O
C
I
T
Y


(
k
m
/
s
e
c
)

LBW
SAW
PSAW
0.4
0.4
0.0
7
7
0
7 7 0
GROUP VELOCITY (km/sec)
Fig. 3. Top: two consecutive experimental images temporally separated
by 1.56 ns of surface acoustic wave fronts for a LiF (100) surface coated
with a 50 nm gold lm. Bottom: theoretical group velocity surfaces for
LiF (100), ignoring lm loading, for surface acoustic waves (SAW),
pseudo-surface acoustic waves (PSAW), and quasilongitudinal acoustic
waves (LBW).
4294 Jpn. J. Appl. Phys., Vol. 44, No. 6B (2005) O. B. WRIGHT et al.
time-reversed component f (r, t) representing incoming
wave fronts, matching the two sets of data to produce a
combination of converging and diverging acoustic waves,
with no acoustic sources. The sign inversion of f stems from
an approximation to experimental observations.
3,15)
We
therefore determine a function g(r, t) = f (r, t) f (r, t)
and its dual Fourier transform [G(k, o)[ for image analysis.
This processing has the advantage of improving the sharp-
ness of the Fourier transform in this case, although it is not
essential.
Figure 5(a) shows an example for TeO
2
(001) of a section
of the experimentally determined function [G(k, o)[ for
o,2 = 570 MHz. This corresponds to a constant frequency
surface in wave vector space. Figure 5(b) shows the
corresponding result of theoretical calculations for SAW,
PSAW, and LBW on the same scale, that is the slowness
surfaces for these modes with the conventional axes k
x
,o
and k
y
,o multiplied by the constant value of o to give k
x
and
k
y
. (We have ignored small lm loading and piezoelectric
eects in the calculation.) The slowness surface can be
considered as the reciprocal of the phase velocity surface,
and is considered to be of most interest in characterizing the
elastic properties of an anisotropic material. The lines traced
out by the experimental [G(k, o)[ are reproduced well by the
theory. Analogously to the results for group velocity, the
combined eects of phonon focusing and the selective
detection of the out-of-plane motion prevent some SAW
branches being detected [of the type marked with the star in
Fig. 5]. One obvious advantage of using the slowness
surface is that this surface is never folded like the group
velocity surface, allowing simpler interpretation. It is also
possible to derive the o-k relations from this analysis.
15)
For
the thin gold lms used in these experiments on crystal
substrates and in contrast to the case for the glass substrate,
the dispersion for the LiF and TeO
2
samples is not
signicant for the imaged areas and frequency range
involved.
If one acoustic mode only is signicantly excited in a
particular direction, it is also possible to obtain the
dispersion relation from an analysis of just three images,
rather than a whole series covering the laser repetition
period. This method was previously outlined in ref. 3, but is
restricted to homogeneous solids with simple phonon
focusing patterns, and in particular to the case when the
LBW and ultrasonic attenuation can be neglected. In
principle it is possible to determine the lm thickness by
studying the curvature of the dispersion relation.
3)
Having now discussed homogeneous samples we shall
now briey describe an application to the probing of
microstructures.
4. Results for Microstructures
The imaging of surface acoustic waves on complex
microstructures is of particular interest because of the
diculty of carrying out simulations for samples with
complex three-dimensional features. We fabricated a sample
consisting of an array of pyramids as shown in Figs. 6(a) and
6(b). The pyramids are suciently spaced so that the surface
acoustic wave propagation through a single pyramid could
be studied. Figure 6(c) shows an optical reectivity image
(at the probe wavelength) and Fig. 6(d) shows surface
acoustic wave images for a 110 mm110 mm region of the
sample. The pump optical pulses are incident at the top right.
To make the sample a 140 nm polycrystalline chromium lm
was prepared by electron beam deposition on a 1 mm crown-
glass substrate. Polycrystalline gold of thickness 1 mm was
then thermally evaporated through a mask. The dierence
in surface acoustic wave propagation velocities in gold
(1.13 kms
1
), in Cr (3.66 kms
1
) and in crown glass
(3.1 kms
1
) is mainly responsible for the refraction at the
pyramid boundaries, producing an inversion of the wave
front curvature inside the pyramid. The surface acoustic
waves are thus concentrated and are strongly scattered on
exiting the pyramid.
5. Conclusions
We have demonstrated a method for real time imaging of
surface acoustic waves at frequencies in the 100 MHz
1 GHz range with picosecond temporal and micron spatial
resolutions using an ultrafast optical pump and probe
LBW
SAW
PSAW
Fig. 5. (a) Experimentally determined dispersion of (001) TeO
2
: constant
frequency surface of the experimentally-determined function [G(k, o)[ for
o,2 = 570 MHz. (b) The result of theoretical calculations of the
slowness surfaces for surface acoustic waves (SAW), pseudo-surface
acoustic waves (PSAW), and quasilongitudinal acoustic waves (LBW) for
the same frequency as in (a). The star shows branches not observed in
experiment.
Fig. 6. (a), (b) Geometry of a microstructure consisting of a crown glass
substrate coated with a 140 nm chromium lm and gold pyramids of
thickness 1 mm. (a): surface. (b): cross section. (c) Optical reectivity
image of the sample surface. (d) Experimental surface acoustic wave
images, at the dierent delay times shown, obtained for the same region
as (c). The pump pulses are incident at the top right.
Jpn. J. Appl. Phys., Vol. 44, No. 6B (2005) O. B. WRIGHT et al. 4295
technique. We have also demonstrated a technique for
obtaining the dispersion from a set of images making up a
complete spatiotemporal animation. This allows the indi-
vidual modes excited in the sample to be distinguished.
Wave fronts in complex nanostructures can also be mapped.
This method shows great promise for the study of the
physics of surface acoustic waves in very complicated
geometries. One example is in the study of phononic
crystals, the acoustic analogue of photonic crystals. Another
is in the study of acoustic resonators. Moreover, in principle
there are exciting prospects for extension with near-eld
optical techniques to frequencies up to 10 GHz or above.
Acknowledgments
We thank S. Tamura, Y. Tanaka, V. E. Gusev and D. H.
Hurley for experimental support and valuable discussions on
the theory of acoustic propagation and dispersion. This work
was partially supported by the 21st Century Center of
Excellence (COE) program entitled Topological Science and
Technology.
1) Lord Rayleigh: Proc. London Math. Soc. 17 (1885) 4.
2) T. J. Wright: Phil. Trans. Roy. Soc. (London) Ser. A 360 (2002) 2873.
3) Y. Sugawara, O. B. Wright, O. Matsuda, M. Takigahira, Y. Tanaka, S.
Tamura and V. E. Gusev: Phys. Rev. Lett. 88 (2002) 185504.
4) J. P. Wolfe: Imaging Phonons (Cambridge University Press, Cam-
bridge, U.K., 1998).
5) A. A. Kolomenskii and A. A. Maznev: Phys. Rev. B 48 (1993) 14502.
6) R. E. Vines, S. Tamura and J. P. Wolfe: Phys. Rev. Lett. 74 (1995)
2729.
7) D. Shilo and E. Zolotoyabko: Phys. Rev. Lett. 91 (2003) 115506.
8) G. Eberharter and H. P. Feuerbaum: Appl. Phys. Lett. 37 (1980) 698.
9) T. Hesjedal and G. Behme: Appl. Phys. Lett. 78 (2001) 1948.
10) K. Nakano, K. Hane, S. Okuma and T. Eguchi: Opt. Rev. 4 (1997)
265.
11) S. Sharples, M. Clark and M. Somekh: Ultrasonics 41 (2003) 295.
12) J. A. Scales and A. E. Malcolm: Phys. Rev. E 67 (2003) 046618.
13) Y. Sugawara, O. B. Wright and O. Matsuda: Rev. Sci. Inst. 74 (2003)
519.
14) Y. Sugawara, O. B. Wright, O. Matsuda and V. E. Gusev: Ultrasonics
40 (2002) 55.
15) Y. Sugawara, O. B. Wright and O. Matsuda: Appl. Phys. Lett. 83
(2003) 1340.
16) O. B. Wright, Y. Sugawara, O. Matsuda, M. Takigahira, Y. Tanaka, S.
Tamura and V. E. Gusev: Physica B 316317 (2002) 29.
17) J. K. Chen, W. P. Latham and J. E. Beraun: Num. Heat Transfer B 42
(2002) 1.
18) J. A. Hildebrand: J. Acoust. Soc. Am. 79 (1986) 1457.
19) D. H. Hurley and O. B. Wright: Opt. Lett. 24 (1999) 1305.
20) G. W. Farnell: Physical Acoustics, eds. W. P. Mason and R. N.
Thurston (Academic Press, New York, 1970) Vol. 6, p. 109.
21) A. A. Maznev, A. M. Lomonosov, P. Hess and Al. A. Kolomenskii:
Eur. Phys. J. B 35 (2003) 429.
22) Y. Tanaka, M. Takigahira and S. Tamura: Phys. Rev. B 66 (2002)
075409.
23) A. G. Every: Phys. Rev. B 37 (1988) 9964.
4296 Jpn. J. Appl. Phys., Vol. 44, No. 6B (2005) O. B. WRIGHT et al.