Tibetan Singing Bowls - Denis Terwagne and John W.M. Bush
Tibetan Singing Bowls - Denis Terwagne and John W.M. Bush
Denis Terwagne
Department of Mathematics,
Massachusetts Institute of Technology, 02139 Cambridge, USA
July 1, 2011
Abstract
We present the results of an experimental investigation of the acoustics and
uid dynamics of Tibetan singing bowls. Their acoustic behavior is rationalized
in terms of the related dynamics of standing bells and wine glasses. Striking
or rubbing a uid-lled bowl excites wall vibrations, and concomitant waves at
the uid surface. Acoustic excitation of the bowls natural vibrational modes
allows for a controlled study in which the evolution of the surface waves with
increasing forcing amplitude is detailed. Particular attention is given to ratio-
nalizing the observed criteria for the onset of edge-induced Faraday waves and
droplet generation via surface fracture. Our study indicates that drops may be
levitated on the uid surface, induced to bounce on or skip across the vibrating
uid surface.
1 Introduction
Tibetan singing bowls are thought to have originated from Himalayan re cults
of the 5th century BC and have since been used in various religious ceremonies,
including shamanic journeying and meditation. The Tibetan singing bowl (see
Fig. 1) is a type of standing bell played by striking or rubbing its rim with a
wooden or leather-wrapped mallet. This excitation causes the sides and rim
of the bowl to vibrate and produces a rich sound. Tibetan bowls are hand
made and their precise composition is unknown, but generally they are made of
a bronze alloy that can include copper, tin, zinc, iron, silver, gold and nickel.
When the bowl is lled with water, excitation can cause ripples on the water
surface. More vigorous forcing generates progressively more complex surface
1
a
r
X
i
v
:
1
1
0
6
.
6
3
4
8
v
1
[
p
h
y
s
i
c
s
.
f
l
u
-
d
y
n
]
2
1
J
u
n
2
0
1
1
wave patterns and ultimately the creation of droplets via wave breaking. We
here quantify this evolution, and demonstrate the means by which the Tibetan
singing bowl can levitate droplets.
a) b)
c)
d)
Figure 1: Our Tibetan singing bowls: a) Tibet 1, b) Tibet 2, c) Tibet 3 and d)
Tibet 4.
Related phenomena are known to appear in other vessels, including the
Chinese singing bowl, on which vibrations are generated by rubbing the vessels
handles with moistened hands. The more familiar vibration of a wine glass is
produced by rubbing its rim with a moist nger. The dependence of the glasss
vibration frequency on its material properties, geometry, and characteristics of
the contained uid was elucidated by French [1] and subsequent investigators [2,
3, 4, 5]. The coupling between two singing wine glasses has been investigated by
Arane et al. [6]. Apfel [7] demonstrated experimentally that wine glass vibration
generates capillary waves near the walls of a uid-lled glass. Moreover, he made
the connection between these waves and the acoustic whispering gallery modes
elucidated by Rayleigh [8]. By studying the deformation and the sound spectrum
produced by a single wine glass, Rossing [9] elucidated the mechanism of the
glass harmonica [10], an instrument designed by Benjamin Franklin. Joubert et
al. [11] provided a theoretical rationale for observations of standing waves in a
singing wine glass. The Tibetan singing bowl has to date received relatively little
attention. Inacio et al. examined experimentally the acoustic response of bowls
excited by impact and rubbing [12], and proposed a dynamical formulation of
the bowl and presented some numerical simulations [13]. The hydrodynamics
of a uid-lled Tibetan bowl will be the focus of our investigation.
In 1831, Faraday [14] demonstrated that when a horizontal uid layer is
vibrated vertically, its interface remains at until a critical acceleration is ex-
ceeded. Above this threshold, a eld of waves appears at the interface, para-
metric standing waves oscillating with half the forcing frequency [15, 16, 17, 18,
19, 20]. The form of such Faraday waves depends on the container geometry;
however boundary eects can be minimized by using a large container. The
2
Faraday waves have a wavelength prescribed by the relative importance of sur-
face tension and gravity, being capillary and gravity waves in the short and long
wavelength limits, respectively. As the forcing acceleration is increased, progres-
sively more complex wave patterns arise, and the interfacial dynamics become
chaotic [20, 21, 19]. Ultimately, large amplitude forcing leads to surface fracture
and the ejection of droplets from the uid bath. A recent study on the breaking
of Faraday waves in a vertically shaken bath has been performed in both the
capillary and gravity wave limits by Puthenveettil et al. [22]. Goodridge et al.
[23] studied the drop ejection threshold of capillary waves in a glycerine-water
solution for frequencies up to 100 Hz.
Faraday [14] reported that such parametric waves can also be emitted by a
vertical plate plunged into a liquid bath and oscillating horizontally: along both
sides of the plate, waves aligned perpendicular to the plate oscillate at half the
forcing frequency. These so-called cross-waves, or edge-induced Faraday waves,
are typically produced by a wave-maker, and have received considerable atten-
tion [24, 25, 26, 27, 28]. Hsieh [29] examined theoretically wave generation in
a vibrating circular elastic vessel, specically the axisymmetric capillary waves
and circumferential crispations that appear in an inviscid uid subject to radial
wall displacement. These studies have demonstrated that the excitation of the
cross waves is due to a parametric resonance. The complexity of this problem
lies in the nonlinear interactions between the motion of the oscillating rim and
the resulting wave eld.
Droplets ejected on the liquid surface by breaking Faraday waves may bounce,
skid and roll before coalescing. A number of recent studies have examined
droplets bouncing on a vertically vibrating liquid bath below the Faraday thresh-
old [30, 31, 32]. The air lm between the drop and the liquid surface is squeezed
and regenerated at each successive bounce, its sustenance precluding coalescence
and enabling droplet levitation. A similar eect arises on a soap lm, a system
more readily characterized theoretically [35]. The bouncing periodicity depends
on the size of the drop and the vertical forcing acceleration of the bath [33, 34].
Couder et al. [30] have shown that, through the waves emitted at previous
bounces, some droplets can walk horizontally across the liquid surface. Several
factors are needed to sustain a so-called walker [36]. First, the drop must
bounce at half the forcing frequency, so that it resonates with the resulting
Faraday wave eld. Second, the bath must be close to the Faraday instability
threshold so that Faraday waves of large amplitude and spatial extent can be
excited by the drop impacts. The droplet bounces on the slope of the wave
emitted at the previous bounce and so receives an impulsive force in a specic
direction, along which it walks with a constant speed. Such walkers have both
wave and particle components, and have been shown to exhibit quantum-like
dynamical behaviour previously thought to be peculiar to the microscopic realm
[37, 38, 39, 40]. Might such modern physics arise in our ancient bowls?
The manuscript is divided into two main parts. In 2, we examine the
acoustics of the Tibetan singing bowls. We begin in 2.1 by reviewing the
related dynamics and theoretical description of the wine glass [1]. In 2.2, our
bowls are presented and their deformation spectra analyzed. Then, by adapting
3
the theoretical description of the vibrating wine glass, we infer the Youngs
modulus of the alloy comprising our bowls. In 3, we examine the dynamics of
ows generated within liquid-lled vibrating bowls. A review of Faraday waves
and droplet ejection on a vertically shaken bath is presented in 3.1. In 3.2,
our experimental technique is detailed. In 3.3, we analyze the surface waves
generated on the liquid bath, and their relation to Faraday waves. In 3.4, we
examine the limit of large amplitude forcing, in which droplets are ejected by
breaking Faraday waves. Comparisons are made with experiments performed
on a vertically shaken liquid bath. Droplet levitation is considered in 3.5,
where particular attention is given to developing criteria for droplet bouncing
and exploring the possibility of walking droplets. A summary of our results is
presented in 4.
2 Acoustics
2.1 Background
Both the wine glass and the Tibetan bowl can be excited by either tapping or
rubbing its rim. We denote by (n,m) the vibrational mode with n complete
nodal meridians and m nodal parallels. Tapping excites a number of vibra-
tional modes [10], while rubbing excites primarily the (2,0) fundamental mode
(see Fig. 2b). Entirely exural motion implies radial and tangential displace-
ments proportional to nsin n and cos n, respectively, being the azimuthal
coordinate [41]. For the (2,0) mode the maximum tangential motion is neces-
sarily half the maximum normal motion.
A leather mallet can excite bowl vibrations via a stick-slip process, as does
a nger moving on a wineglass. The moving mallet forces the rim to follow the
mallet during the stick phase; during the slip phase, the bowl rim relaxes to its
equilibrium position. This rubbing results in a sound composed of a fundamental
frequency plus a number of harmonics. While the mallet is in contact with the
bowl, one of the nodes follows the point of contact [11], imparting angular
momentum to the bound liquid.
To simplify the acoustic analysis, one can approximate the glass or bowl by
a cylindrical shell with a rigid base and an open top (Fig. 2a). The system
can then be described in terms of 7 physical variables, the radius R, height
H
0
, thickness a, Youngs modulus Y and density
s
of the cylindrical shell, and
the frequency f and amplitude of its oscillating rim. The system can thus
be described in terms of 4 independent dimensionless groups, which we take
to be R/H, /a, /R and a Cauchy number Ca =
s
f
2
2
/Y that indicates
the relative magnitudes of the inertial and the elastic forces experienced by the
vibrating rim.
The sound is emitted by bending waves that deform the rim transversely as
they propagate. The speed V
b
of bending waves on a two-dimensional plate of
4
R
Y
s
a
f,
H
0
H
a) b)
Figure 2: a) Schematic illustration of a vibrating vessel lled with liquid. Rele-
vant parameters are the height H
0
and the radius R of the vessel, the thickness
of the rim a, the liquid level H, the frequency f and the amplitude of the
oscillating rim. b) Top view of a vibrating vessel in its fundamental mode (2,0),
characterized by its 4 nodes and 4 antinodes.
thickness a is given by [42]
V
b
=
_
V
L
fa
3
_
1/2
, (1)
where V
L
is the longitudinal wave speed in the solid (approximately 5200 m/s
in glass). In order for the bending wave to traverse the perimeter in an integer
multiple of the period, we require
1
f
2R
V
b
. (2)
Thus, since V
b
fa, we have f a/R
2
: the frequency increases with rim
thickness, but decreases with radius.
A more complete theoretical analysis of the wine glass acoustics [1] can
be applied to our Tibetan bowls. An ideal cylinder xed at the bottom is
considered (Fig. 2a), its wall vibrating with largest amplitude at its free edge or
rim. The rims kinetic energy and elastic energy of bending in the mode (2,0)
are calculated by assuming that the radial position is proportional to cos 2,
with being the azimuthal coordinate. By considering conservation of total
energy (kinetic plus elastic bending energy), an expression for the frequency of
this mode can be deduced:
f
0
=
1
2
_
3Y
5
s
_
1/2
a
R
2
_
1 +
4
3
_
R
H
0
_
4
_
1/2
. (3)
5
When the bowl is partially lled with liquid to a depth H (Fig. 2a), the frequency
decreases. French [1] captured this eect by considering the kinetic energy of
the liquid near the rim, and so deduced the frequency of the fundamental mode:
_
f
0
f
H
_
2
1 +
5
l
R
s
a
_
H
H
0
_
4
, (4)
where
l
is the liquid density and 1.25 is a constant indicating the coupling
eciency between the rim and uid displacements. Similarly, frequencies of
higher modes can be calculated by considering a radial position proportional to
cos n and with m nodal parallels [1]:
f
(n,m)
=
1
12
_
3Y
s
_
1/2
a
R
2
_
(n
2
1)
2
+ (mR/H
0
)
4
1 + 1/n
2
_
1/2
(5)
2.2 Tibetan bowls
Four dierent antique bowls of dierent sizes have been studied (Fig. 1). They
are referred to as Tibet 1, 2, 3 and 4 and their physical characteristics are
reported in Table 1.
Bowl name Tibet 1 Tibet 2 Tibet 3 Tibet 4
Symbol
f
(2,0)
(empty) (Hz) 236 187 347 428
Radius R (cm) 7.5 8.9 6 5.9
Thickness a (cm) 0.34 0.38 0.31 0.37
Mass (kg) 0.690 0.814 0.306 0.312
Volume (cm
3
) 76 97 35 33
Density (kg/m
3
) 9079 8366 8754 9372
Table 1: Physical properties of the four Tibetan bowls used in our study.
When a bowl is struck or rubbed, the sound emitted by the resulting bowl
vibrations is recorded with a microphone and a fast Fourier transform per-
formed on the signal. Dierent peaks clearly appear in the frequency spectrum,
corresponding to the bowls dierent vibrational modes. Fig. 3a indicates the
frequency spectrum generated by striking the empty bowl Tibet 4. When the
bowl is rubbed with a leather wrapped mallet, the lowest mode is excited along
with its harmonics, an eect known as a mode lock in [43]. The frequency
spectrum of Tibet 4 when rubbed by a leather mallet is presented in Fig. 3b.
Due to the bowl asymmetry, two peaks separated by several Hz arise and a
beating mode is heard. This split is highlighted in a magnication of the rst
peak f
(2,0)
in the inset of the gure. The deformation shapes are the same with
both frequencies but there is horizontal angular shift observed to be /4 be-
tween them for the fundamental modes (2, 0) and /2n for other (n, 0) modes.
6
Finally, we note that, owing to the relative squatness of the bowls and the as-
sociated high energetic penalty of modes with m = 0, only modes (n, 0) were
excited; henceforth, such modes are denoted simply by n.
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0 1000 2000 3000 4000 5000 6000
r
e
l
a
t
i
v
e
m
a
g
n
i
t
u
d
e
f(Hz)
(2,0)
(3,0)
(4,0)
(5,0)
(6,0)
a)
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0 1000 2000 3000 4000 5000 6000
r
e
l
a
t
i
v
e
m
a
g
n
i
t
u
d
e
f(Hz)
(2,0)
b)
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
410 420 430 440 450 460
(2,0)
Figure 3: a) Frequencies excited in the bowl Tibet 4 when struck with a wooden
mallet. The dierent peaks correspond to the natural resonant frequencies of the
bowl, and the associated deformation modes (n, m) are indicated. b) Excited
frequencies of the bowl Tibet 4 when rubbed with a leather mallet. The rst
peak corresponds to the deformation mode (2,0) and subsequent peaks to the
harmonics induced by mode lock-in. In the inset, a magnication of the rst
peak provides evidence of its splitting due to the asymmetry of the bowl.
We can also nd a relation between the dierent mode frequencies. Since the
speed of the bending wave is proportional to the square root of the frequency,
v
f and since = v/f, we expect 1/
l
R
sa
. In Fig. 5b, we
present the dependence of (1+1/n
2
)(R
2
/a)
2
f
2
on (n
2
1)
2
for the deformation
modes n =2 through 6 of the bowls Tibet 1, 2, 3 and 4. According to (5),
the slope should be equal to
1
48
2
Y
s
. For each value of the abscissa there are
4 measurements corresponding to the 4 Tibetan bowls. Data points from the
7
0
1000
2000
3000
4000
5000
6000
7000
0 2 4 6 8 10 12
Tibet 2
Tibet 1
Tibet 3
Tibet 4
f
(
H
z
)
n
a)
100
1000
10
4
2 3 4 5 6 7 8 9 10
Tibet 2
Tibet 1
Tibet 3
Tibet 4
all
f
R
2
/
a
(
m
/
s
)
n
b)
Figure 4: a) Measurements of the resonant frequencies of the dierent deforma-
tion modes n for the 4 dierent Tibetan bowls. b) Characteristic speed fR
2
/a
as a function of deformation mode n for the 4 dierent bowls. Data are t by
a power 2 curve indicating that f
n
n
2
a/R
2
, consistent with Eq. 6.
dierent Tibetan bowls overlie each other, especially at low n, indicating that
all bowls have the same value of ratio Y/
s
, and so are likely made of the same
material. A linear t gives Y/
s
= 8.65 10
6
Pa.m
3
/kg.
0.9
1
1.1
1.2
1.3
1.4
1.5
1.6
0 0.2 0.4 0.6 0.8 1
(
f
0
/
f
H
)
2
(H/H
0
)
4
a)
0
5 10
6
1 10
7
1.5 10
7
2 10
7
2.5 10
7
0 200 400 600 800 1000 1200 1400
Tibet 1
Tibet 2
Tibet 3
Tibet 4
B
(
1
+
1
/
n
2
)
R
4
/
a
2
f
2
(n
2
-1)
2
b)
190
200
210
220
230
240
250
0 0.2 0.4 0.6 0.8 1
f
0
(
H
z
)
H/H
0
Figure 5: a) The fundamental frequency of Tibet 1s mode (2,0) as a function of
the water height H/H
0
. According to Eq. 4 , the slope is equal to
5
l
R
sa
. b) The
dependence of the resonant frequencies f of all the bowls on the deformation
mode n may be used to infer the ratio Y/
s
. According to (5), the slope of the
line shown is
1
48
2
Y
s
, from which we infer Y/
s
= 8.65 10
6
Pa.m
3
/kg.
We can simply estimate the density for each Tibetan bowl by measuring
its mass and volume. Masses are measured by a weight scale with an error of
8
2 g while bowl volumes are deduced by uid displacement during submersion.
The error made on the volume with this method is estimated to be no more
than 5%. All such density measurements are reported in Table ??. Taking the
mean value of the densities, equal to 8893 kg/m
3
, we calculate a Young modulus
Y = 77 6% GPa. This value is in the Youngs modulus range of glasses and
lower than typical brass, copper or bronze alloys for which Y > 100 GPa.
3 Fluid dynamics
3.1 Faraday waves
Consider an inviscid uid of density and surface tension in a horizontal layer
of uniform depth h in the presence of a gravitational acceleration g. The layer
is oscillated vertically in a sinusoidal fashion at a forcing frequency f
0
=
0
/2,
amplitude and acceleration g =
2
. Above a critical forcing acceleration,
standing Faraday waves appear on the surface. The associated surface deforma-
tion a(x, y, t) can be expressed in terms of the containers eigenmodes S
m
(x, y)
as
a(x, y, t) =
m
a
m
(t)S
m
(x, y), (7)
with a
m
(t) being the oscillating amplitude of the eigenmode m. Benjamin and
Ursell [16] demonstrate that the coecients a
m
(t) satisfy:
a
m
+
2
m
(1 2 cos
0
t)a
m
= 0, (8)
where
2
m
=
_
gk
m
+
k
3
m
_
tanh (k
m
h) (9)
represents the classic gravity-capillary wave dispersion relation,
=
2(1 + Bo
1
)
(10)
is the dimensionless forcing acceleration, and Bo =
g
k
2
is the Bond number.
k
m
= 2/
m
is the wave number of the mode m, and
m
is the corresponding
wavelength.
In the absence of forcing, = 0, (8) describes a simple harmonic oscillation
with frequency prescribed by (9) corresponding to the free surface vibrations
of the liquid. When > 0, (8) describes a parametric oscillator as the forcing
term depends on time. The resulting Faraday waves can be either capillary
or gravity waves according to the magnitude of Bo; specically, the former and
latter arise in the respective limits Bo 1 and Bo 1. Equation 8 is known as
the Mathieu equation and cannot be solved analytically since one of the terms
is time dependent. However, as the forcing is periodic, Floquet theory can be
applied to show that an inviscid uid layer is always unstable to Faraday waves
with a frequency
F
that is half the forcing frequency
0
= 2
F
[16]. In the
9
deep water (kh 1), capillary wave (Bo 1) limit, we can infer from (9) a
Faraday wavelength :
F
= (2)
1/3
(/)
1/3
(f
0
/2)
2/3
. (11)
To incorporate the inuence of the uid viscosity, one can add to (8) a
phenomenological dissipation term proportional to the velocity [18]:
a
m
+ 2 a
m
+
2
m
(1 2 cos
0
t)a
m
= 0, (12)
where is the dissipation rate. This dissipation term leads to an acceleration
threshold for the Faraday instability. Assuming capillary waves in an unbounded
and innite depth liquid, the critical acceleration needed to induce parametric
instability is given by
F
1
g
(/)
1/3
5/3
0
, (13)
where is the kinematic viscosity of the uid [19, 44].
As the forcing amplitude is further increased, the Faraday wave amplitude
increases progressively until the waves become chaotic. Ultimately, the waves
break and droplets are ejected from the surface. Since drops will be ejected by
the breaking of Faraday waves, we expect their diameter to scale as
d
m
F
(/)
1/3
f
2/3
0
(14)
in the capillary wave limit. Droplet ejection arises when the destabilizing inertial
driving force mg (with m
3
F
) exceeds the stabilizing surface tension force
F
. This implies, via (14), a threshold acceleration that scales as:
d
1
g
(/)
1/3
f
4/3
0
(15)
The range of validity of these scalings will be investigated in our experimental
study.
3.2 Experimental technique
The experimental setup used for studying the surface waves generated within
the Tibetan bowls is presented in Fig. 6. A loudspeaker is set in front of the
bowl, its signal received from a signal function generator then amplied. When
the applied signal frequency is close to one of the bowls resonant frequencies,
it oscillates in the corresponding deformation mode. Conveniently, with this
method, we can select a single deformation mode. Recall that by striking or
rubbing the bowls, we excited several modes (Fig. 3): moreover, rubbing induced
a rotational motion that followed the mallet. We can now examine the vibration-
induced ows in a controlled fashion.
In order to extend the range of natural frequencies, we consider Tibetan
bowls, wineglasses and soda cans (with their tops removed), whose resonant
10
Accelerometer
Loudspeaker
Figure 6: Experimental setup. The bowl deformation is excited by a loudspeaker
emitting sound at a frequency corresponding to a particular vibrational mode of
the bowl. The deformation of the bowl is measured by an accelerometer glued
to the bowls outer wall at the height of the liquid surface.
frequencies span a broad range from 50 to 750 Hz. For each, we can vary
the resonant frequency by lling it with liquid. The sound emitted by the
loudspeaker was not powerful enough to induce signiant oscillations of the
soda can rim, a problem we eliminated by directly connecting the vibrating
membrane of the loudspeaker to the rim of the soda can with a rigid rod.
We measure the acceleration of the rim at an antinode by gluing a lightweight
accelerometer (PCB-Piezotronics, 352C65) on the bowls rim at the level of the
liquid surface. In the following, we characterize the sinusoidal rim oscillation
by the dimensionless acceleration dened as the maximal acceleration of the
rim normalized by the gravitational acceleration =
2
/g. For wineglasses,
the accelerometer cannot be used since it dramatically alters the resonant fre-
quency. We thus used a light weight strain gauge, whose eect on the resonance
frequency is negligible.
The strain gauge system provides a measurement of the local extension
length of the rim at an antinode. The length variation of the strain gauge
is deduced by measuring its electrical resistance with a Wheatstone bridge. To
deduce the acceleration of the radial rim movement, we deduce a relation be-
tween the longitudinal extension and the radial amplitude of the rim . Then,
the acceleration can be readily deduced. The validity of this indirect method
was tested on Tibet 1. An accelerometer was glued next to a strain gauge at
an antinode of the bowl Tibet 1. From the strain gauge measurement, we de-
duced an acceleration that we compared to the direct measurement from the
accelerometer. These two independent measurements match well for a wide
range of accelerations.
French [1] gives a relation for the convex change of length l of a curved
segment of thickness a as a function of the initial mean radius of curvature
R and the deformed radius of curvature r
c
:
l
= l
0
a
2
_
1
rc
1
R
_
where l
0
is
11
the initial segment length. Since the deformation is small, we can approximate:
1
rc
1
R
+
3x
R
2
where x is the radial displacement of the wall [1]. The strain gauge
gives the measurement of the convex longitudinal extension of the rim = l/l
0
.
We can thus deduce the maximal radial amplitude of the rim deformation [1]:
x
m
=
2
3
R
2
a
.
3.3 Surface waves
Two dierent liquids were used in our experiment, pure water with density
w
= 1000 kg/m
3
, viscosity = 1 cSt and surface tension = 72 mN/m, and
Dow Corning silicone oil with
o
= 820 kg/m
3
, = 1 cSt and = 17.4 mN/m.
The uid depth and resulting natural deformation frequencies of the bowl were
varied. Specic deformation modes of the bowls were excited acoustically. We
proceed by reporting the form of the ow induced, specically, the evolution of
the free surface with increasing rim forcing.
In Fig. 7, we present snapshots of the bowl Tibet 1 resonating in its fun-
damental deformation mode with dierent when it is completely lled with
water. The loudspeaker produces a sinusoidal sound at a frequency f
0
= 188 Hz
that corresponds to the mode (2,0) with four associated nodes and antinodes.
The vibration of the water surface is forced by the horizontal oscillation of the
rim. When the amplitude of the rim oscillation is small, axisymmetric progres-
sive capillary waves with frequency commensurate with the excitation frequency
appear on the liquid surface. Though almost invisible to the naked eye, they can
be readily detected by appropriate lighting of the liquid surface (Fig. 7a). When
is further increased, relatively large amplitude circumferential standing waves
appear at the waters edge (Fig. 7b). These standing ripples, aligned perpen-
dicular to the bowls edge, are spaced by approximately a Faraday wavelength
F
, as dened in (11). Moreover, their frequency is half that of the axial waves,
indicating that these waves correspond to classic cross waves or, equivalently,
edge-induced Faraday waves [14]. More complicated wave modes appear at
higher excitation amplitude (Fig. 7c). At suciently high , the Faraday waves
break, and water droplets are ejected from the edge of the vessel (Fig. 7d),
specically from the anti-nodes of the oscillating wall. The ejected droplets
may bounce, slide, and roll on the water surface before eventually coalescing.
One of our bowls (Tibet 2) resonates readily in both modes (2,0) and (3,0).
When completely lled with water, the resonant frequencies of its (2,0) and
(3,0) modes are f = 144 Hz and f = 524 Hz, respectively. In Fig. 8, we observe
the progression of the surface waves with increasing amplitude for each of these
modes. Note that for the mode (3,0), since the frequency is higher, the wave-
lengths are shorter. Moreover, the sound amplitude needed to produce surface
waves is necessarily higher for mode (3,0) than (2,0).
The transition from axisymmetric capillary waves to Faraday waves arises at
a critical acceleration
F
readily measured by the accelerometer. This threshold
was measured as a function of the forcing frequency, the latter having been
tuned to excite the fundamental deformation modes of the bowls with dierent
liquid levels. We also measured this acceleration threshold for the deformation
12
Figure 7: Evolution of the surface waves in Tibet 1 bowl lled with water and
excited with a frequency f = 188 Hz corresponding to its fundamental mode
(2,0). The amplitude of deformation is increasing from left to right: a) = 1.8,
b) = 2.8, c) = 6.2, d) = 16.2.
Figure 8: Evolution of the surface waves in Tibet 2 bowl lled with water. a-d)
Mode (2,0) is excited with a frequency f = 144 Hz. e-h) Mode (3,0) is excited
with f = 524 Hz. The amplitude of deformation is increasing from left to right.
For the deformation mode n = 2, a) = 1.0, b) = 1.7, c) = 5.0, d)
= 12.7. For the deformation mode n = 3, e) = 5.7, f) = 7.7, g) = 15.1,
h) = 33.8.
mode (3,0) of the bowl Tibet 2. Higher frequencies were explored with three
dierent wineglasses lled to dierent levels using the strain gauge system. All
the measurements with silicone oil of viscosity 1 cSt are presented in Fig. 9
(lower curve). Consistent with (13), the data suggest a dependence
F
f
5/3
.
In Fig. 10a, we report our measurements of
F
as a function of frequency for
both silicone oil and distilled water. Each data set is t by a 5/3 power law.
Prefactors of 3.5 10
4
for water and 1.7 10
4
for 1 cSt silicone oil were
inferred.
13
3.4 Surface fracture
When the Faraday waves become suciently large, they break, leading to droplet
ejection. A second critical acceleration can thus be measured,
d
, above which
droplets are ejected from the surface. The droplet ejection starts with very
few droplets ejected, then the ejection rate increases with forcing amplitude.
Our criterion for onset is that at least two droplets are ejected in a 15 second
time interval. The dependence of
d
on f is presented in Fig. 9 (upper curve)
for bowls and wine glasses lled with dierent levels of 1 cSt silicone oil. The
drop ejection threshold scales as
d
f
4/3
, which is consistent with the scaling
presented in (15).
10
-1
10
0
10
1
10
2
10
2
10
3
!
f(Hz)
Faraday waves
+
jumping droplets
Faraday
waves
Capillary waves
Figure 9: Phase diagram indicating the critical acceleration
F
=
2
/g above
which axial capillary waves give way to Faraday waves, and
d
above which the
Faraday waves break. The , , and symbols correspond, respectively, to
the measurements made with the bowls Tibet 1, 2, 3 4 and the to wineglasses
lled with dierent levels of 1 cSt silicone oil. The rst acceleration threshold
(lower curve) scales as
F
f
5/3
and the second (upper curve) as
d
f
4/3
.
In Fig. 10b, our measurements of the dependence of
d
on f are reported for
bowls lled with dierent levels of either 1 cSt silicone oil or water. The two
d
curves collapse when we use the scaling law (15) expected to apply for vertical
forcing. Specically, we nd
d
0.23(/)
1/3
f
4/3
. (16)
14
The droplet ejection acceleration threshold is thus in accord with measurements
of Goodridge [23] and Puthenveettil [22], even though our forcing is horizontal
rather than vertical. Moreover, our prefactor is consistent with the results of
both, who reported values between 0.2 and 0.3.
10
-1
10
0
10
1
10
2
10
2
10
3
!
F
f(Hz)
Capillary waves
Faraday waves
a)
10
2
10
3
10
4
10
2
10
3
!
d
(
"
/
#
)
1
/
3
f(Hz)
Faraday waves
+
jumping droplets
Faraday waves
0.2
0.3
0.23
b)
Figure 10: Dependence of acceleration thresholds on the frequency f for our
vessels lled with dierent levels of pure water (blue symbols) and silicone oil
of viscosity = 1 cSt (red symbols). The , , and symbols correspond,
respectively, to the measurements made with the bowls Tibet 1, 2, 3 4 and the
to wineglasses. a) Each data set of Faraday threshold measurements,
F
, is
t by a 5/3 power law consistent with (13). b) The droplet ejection threshold,
d
, is consistent with the scaling (15).
The diameter of the ejected droplets was measured for several forcing fre-
quencies. The Tibetan bowls were fully lled with liquid, either 1 cSt silicone
oil or pure water. Resonant deformation modes were then excited by the loud-
speaker emitting sinusoidal signals at the appropriate resonant frequency. The
level of sound was adjusted so that the acceleration of the rim at the antin-
odes of the bowl were just above the threshold for droplet ejection,
d
. A high
speed video camera (Phantom) was used to record the ejected droplets, from
which drop size measurements were taken. We used a liquid-lled glass in or-
der to extend the frequency range to 720 Hz. Soda cans have very thin walls,
and very low resonant frequencies. We were thus able to measure droplet sizes
for the two liquids at a frequency of 98 Hz. Three cumulative distributions of
ejected droplet sizes are presented on Fig. 11a for the bowls Tibet 1, 3 and 4
resonating in their fundamental deformation modes (2,0). Assuming that these
distributions are Gaussian, appropriate ts to the cumulative distribution func-
tions yield the parameters of the Gaussian distribution functions plotted in the
inset of Fig. 11b. The dependencies of the mean droplet size on the forcing fre-
quency for the two liquids are presented in Fig. 11b. Equation (14) adequately
collapses our data, provided we choose a prefactor of 0.87 . By way of compar-
ison, Puthenveettil found 0.92 for their experiments with water, and 1.01 with
15
Peruoro-compound FC-72 liquid. Donnelly et al. [45] found a prefactor of 0.98
from his measurements of aerosol water droplets.
0
0.2
0.4
0.6
0.8
1
0 0.5 1 1.5 2
185 Hz
281 Hz
354 Hz
C
D
F
d (mm)
a)
10
-2
10
-1
10
2
10
3
Pure water
1cSt Silicone oil
B
B
d
m
(
!
/
"
)
1
/
3
f(Hz)
2
3
1
0.87
b)
0
0.5
1
1.5
2
2.5
3
0 0.5 1 1.5 2
P
D
F
d (mm)
354 Hz 281 Hz
185 Hz
Figure 11: a) Three cumulative distribution functions of ejected droplet sizes
for the bowls Tibet 1, 3 and 4, fully lled with pure water and resonating in
their fundamental deformation modes (2,0). The lines are the curves t by an
equation of a cumulative distribution function for the Gaussian distribution. b)
Mean size of the ejected droplets as a function of the frequency of the oscillating
vessel, either a Tibetan bowl, wineglass or soda can. The two leftmost data
points are from the measurements made with the soda cans, the two rightmost
from the wineglass. Error bars are the size of our symbols. The black curve
indicates a power law dependence with slope -2/3 and prefactor 0.87. The
dotted line has the same slope, but a prefactor of 1. Inset: The corresponding
Gaussian functions inferred from the cumulative distribution functions.
3.5 Bouncing droplets
With a more viscous uid (e.g. 10 cSt silicone oil), the waves are less pro-
nounced, and the uid is more strongly coupled to the vibrating sidewalls;
specically, more of the surface oscillates up and down near the walls anti-
node. When a droplet of the same liquid is deposited on the surface, it may
bounce, levitated by the underlying wave eld. Such sustained levitation was
not observed in the Tibetan singing bowl with liquid viscosities lower than 10
cSt.
In Fig. 12a, we present a still image of a drop of diameter 0.5 mm bouncing
on the liquid surface inside the bowl (Tibet 1) resonating at a frequency of
f
0
= 188 Hz. The drop has been made by dipping then extracting a syringe
needle, on the tip of which a capillary bridge forms and breaks, leaving a drop
that bounces near the oscillating rim. A movie of the bouncing was recorded,
then vertical slices of each image through the droplet centerline juxtaposed. We
can thus construct an image illustrating the dynamics of the droplet (Fig. 12b).
16
In this case, the drop experience two bounces of slightly dierent amplitude
while the liquid surface (and the rim) oscillates twice. When a smaller droplet
(of diameter 0.35 mm) is placed on the oscillating surface, the bouncing motion
can be more complex. In Fig. 12c, we see that the droplet bounces only once
during three oscillations of the surface. Fig. 12d illustrates the corresponding
trajectory. We note that the bouncing motion of suciently small drops can
become chaotic.
We sought to sustain walking droplets in our system. Once the horizontal
forcing amplitude is just above the Faraday wave threshold
F
, circumferential
Faraday waves are sustained at the edge of the vessel. Then, when is increased,
Faraday waves propagate progressively towards the center of the vessel, their
amplitude damped by viscosity. Beyond these waves, the liquid surface was
quiescent unless perturbed by a bouncing droplet, in which case it could sustain
a eld of Faraday waves. This phenomenon was observed in the bowl Tibet
2 almost fully lled with 10 cSt silicone oil (f
0
=140 Hz). Just beyond the
Faraday waves, a bouncing droplet of diameter 500 m was made such that
its bouncing frequency corresponded to the Faraday wave frequency, that is,
half the frequency of the vibrating rim. Such droplets were unable to excite
suciently large Faraday waves to enable them to walk. We note that the usual
range of walker diameters is between 650 m and 850 m [33, 34]: their mass
is thus 4 times larger than that of our drops.
4 Conclusion
We have presented the results of an experimental investigation of the Tibetan
singing bowl, its acoustics and hydrodynamics. Its acoustical properties are
similar to those of a wine glass, but its relatively low vibration frequency makes it
a more ecient generator of edge-induced Faraday waves and droplet generation
via surface fracture.
Our observations of the bowl acoustics have been rationalized by adapting
Frenchs [1] theory of the singing wine glass. This model allowed us to charac-
terize the bowl acoustics and infer the Young modulus of the alloy constituting
our antique bowls. The value we found, Y = 77 6% GPa is in the range
of glass, somewhat smaller than typical brass, copper or bronze alloys. This
low value of Y and associated low resonant frequency is a critical component
in the hydrodynamic behavior of singing bowls: bowls with high fundamental
frequencies are, like the wine glass, relatively inecient generators of droplets.
Particular attention has been given to the Faraday waves produced when a
critical acceleration of the horizontal rim oscillation is exceeded. These have
been shown to be due to a destabilization of the axial capillary waves similar
to those observed and studied theoretically [29, 24]. The acceleration thresh-
old for droplet ejection has also been investigated and rationalized by simple
scaling arguments. Droplet size was shown to be proportional to the Faraday
wavelength, and our measurements were consistent with those on a vertically
shaken liquid surface [22, 23, 45].
17
Figure 12: a) Still image of a droplet of 0.5 mm diameter bouncing on a surface
oscillating at 188 Hz, the frequency of the rim. Both droplet and bath are 10 cSt
Silicon oil. Spatio-temporal diagrams indicate the vertical trajectory of droplets
of diameter b) 0.5 mm and c) 0.35 mm. Time elapses from left to right and the
drops reection is apparent. d) A single bounce of a droplet of diameter 0.35
mm is illustrated by an image sequence. Images are spaced by 1 ms.
We have demonstrated that, following their creation via surface fracture,
droplets may skip across or roll along the surface of uid contained within a
singing bowl. Moreover, careful choice of uid properties and droplet position
introduces the possibility of stable bouncing states reminiscent of those on a
vertically driven free surface [30]. However, stable walking droplets and their
concomitant quantum behaviour were not observed. Nevertheless, in develop-
ing hydrodynamic analogues of quantum systems, the edge-forcing examined
here may be valuable in presenting a lateral gradient in proximity to Faraday
18
threshold.
Acknowledgements
The authors thank Rosie Warburton for bringing this problem to our atten-
tion, and for supplying the bowls for our study. We are grateful to Jim Bales (of
MITs Edgerton Center), Barbara Hughey and Stephane Dorbolo for assistance
and fruitful discussions. Denis Terwagne thanks the University of Li`ege and
the GRASP for nancial support. This research was supported by the National
Science Foundation through grant CBET-0966452.
References
[1] A. French, In vino veritas - a study of wineglass acoustics, Am. J. Phys.
51, 688 (1983).
[2] K. Chen, C. Wang, C. Lu and Y. Chen, Variations on a theme by a singing
wineglass, Europhys. Lett. 70, 334 (2005).
[3] Y.-Y. Chen, Why does water change the pitch of a singing wineglass the
way it does?, Am. J. Phys. 73, 1045 (2005).
[4] G. Jundt, A. Radu, E. Fort, J. Duda, H. Vach and N. Fletcher, Vibrational
modes of partly lled wine glasses, J. Acoust. Soc. Am. 119, 3793 (2006).
[5] M. Courtois, B. Guirao and E. Fort, Tuning the pitch of a wine glass by
playing with the liquid inside, Eur. J. Phys. 29, 303 (2008).
[6] T. Arane, A. Musalem and M. Fridman, Coupling between two singing
wineglasses, Am. J. Phys. 77, 1066 (2009).
[7] R. Apfel, Whispering waves in a wineglass, Am. J. Phys. 53, 1070 (1985).
[8] L. Rayleigh, Theory of sound (Cambridge University Press, 1945).
[9] T. Rossing, Wine glasses, bell modes and lord rayleigh, Phys. Teach. 28,
582 (1990).
[10] T. Rossing, Acoustics of the glass harmonica, J. Acoust. Soc. Am. 95, 1106
(1993).
[11] S. V. Joubert, T. H. Fay and E. L. Voges, A storm in a wineglass, Am. J.
Phys. 75, 647 (2007).
[12] O. Inacio, L. Henrique and J. Antunes, The dynamics of tibetan singing
bowls, Acta Acust United Ac 92, 637 (2006).
[13] O. Inacio, L. Henrique and J. Antunes, The physics of tibetan singing bowls
(2004), Revista de Ac ustica. 35, 33 (2004).
19
[14] M. Faraday, On a peculiar class of acoustical gures; and on certain forms
assumed by groups of particles upon vibrating elastic surfaces, Philos. Trans.
R. Soc. London 121, 299 (1831).
[15] L. Rayleigh, On the crispations of uid resting upon a vibrating support.,
Philos. Mag. 16, 50 (1883).
[16] T.B. Benjamin and F. Ursell, The stability of the place free surface of a
liquid in vertical periodic motion, Proc. R. Soc. Lond. A 255, 505 (1954).
[17] J. Miles and D. Henderson, Parametrically forced surface-waves, Annu Rev
Fluid Mech 22, 143 (1990).
[18] K. Kumar and L. S. Tuckerman, Parametric instability of the interface
between two uids, J Fluid Mech 279, 49 (1994).
[19] W. Edwards and S. Fauve, Patterns and quasi-patterns in the faraday ex-
periment, J Fluid Mech 278, 123 (1994).
[20] A. Kudrolli and J. Gollub, Patterns and spatiotemporal chaos in parametri-
cally forced surface waves: A systematic survey at large aspect ratio, Phys-
ica D 97, 133 (1996).
[21] P. Chen and J. Vinals, Pattern selection in faraday waves., Phys. Rev. Lett.
79, 2670 (1997).
[22] B. A. Puthenveettil and E. J. Hopnger, Evolution and breaking of para-
metrically forced capillary waves in a circular cylinder, J Fluid Mech 633,
355 (2009).
[23] C. Goodridge, W. T. Shi, H. Hentschel and D. Lathrop, Viscous eects in
droplet-ejecting capillary waves, Physical Review E 56, 472 (1997).
[24] C. Garrett, On cross-waves, Journal of Fluid Mechanics 41, 837 (1970).
[25] B. J. S. Barnard and W. G. Pritchard, Cross-waves. part 2. experiments,
J. Fluid Mech. 55, 245 (1972).
[26] J. J. Mahony, Cross-waves. part 1. theory, J. Fluid Mech. 55, 229 (1972).
[27] J. Miles and J. Becker, Parametrically excited, standing cross-waves, J.
Fluid Mech. 186, 119 (1988).
[28] J. Becker and J. Miles, Standing radial cross-waves, J. Fluid Mech. 222,
471 (1991).
[29] D. Hsieh, Theory of water waves in an elastic vessel, Acta. Mech. Sinica
16, 97 (2000).
[30] Y. Couder, E. Fort, C.-H. Gautier and A. Boudaoud, From bouncing to
oating: Noncoalescence of drops on a uid bath., Phys. Rev. Lett. 94,
177801 (2005).
20
[31] T. Gilet, D. Terwagne, N. Vandewalle and S. Dorbolo, Dynamics of a bounc-
ing droplet onto a vertically vibrated interface, Phys. Rev. Lett. 100, 167802
(2008).
[32] S. Dorbolo, D. Terwagne, N. Vandewalle and T. Gilet, Resonant and rolling
droplet, New J. Phys. 10, 113021 (2008).
[33] S. Protiere, Y. Couder, E. Fort and A. Boudaoud, The self-organization
of capillary wave sources, Journal of Physics-Condensed Matter 17, S3529
(2005).
[34] A. Eddi, D. Terwagne, E. Fort and Y. Couder, Wave propelled ratchets and
drifting rafts, Eur. Phys. Lett. 82, 44001 (2008).
[35] T. Gilet and J. W. M. Bush, The uid trampoline: droplets bouncing on a
soap lm, J. Fluid Mech. 625, 167 (2009).
[36] A. Eddi, E. Sultan, J. Moukhtar, E. Fort, M. Rossi and Y. Couder, In-
formation stored in faraday waves: the origin of a path memory, J. Fluid
Mech. 674, 433 (2011).
[37] Y. Couder and E. Fort, Single-particle diraction and interference at a
macroscopic scale, Physical Review Letters 97, 154101 (2006).
[38] A. Eddi, E. Fort, F. Moisy and Y. Couder, Unpredictable tunneling of a
classical wave-particle association, Physical Review Letters 102, 240401
(2009).
[39] E. Fort, A. Eddi, A. Boudaoud, J. Moukhtar and Y. Couder, Path-memory
induced quantization of classical orbits, Proc. Natl. Acad. Sci. 107, 17515
(2010).
[40] J. W. M. Bush, Quantum mechanics writ large, Proc. Natl. Acad. Sci. 107,
17455 (2010).
[41] R. Perrin, T. Charcley, H. Banut and T. D. Rossing, Chladnis law and the
modern english church bell, Journal of Sound and Vibration 102, 11 (1985).
[42] L. Kinsler, A. Frey, A. Coppens and J. Sanders, Fundamentals of acoustics
(3rd edition) (Wiley, 1982).
[43] A. Akay, Acoustics of friction, J. Acoust. Soc. Am. 111, 1525 (2002).
[44] S. Douady, Experimental study of the faraday instability, J. Fluid Mech.
221, 383 (1990).
[45] T. Donnelly, J. Hogan, A. Mugler, N. Schommer, M. Schubmehl, A. Berno
and B. Forrest, An experimental study of micron-scale droplet aerosols pro-
duced via ultrasonic atomization, Physics of Fluids 16, 2843 (2004).
21
Tibetan singing bowls
Citation
Terwagne, Denis, and John W M Bush. Tibetan Singing Bowls.
Nonlinearity 24.8 (2011): R51R66. Web. 11 Apr. 2012.
As Published
http://dx.doi.org/10.1088/0951-7715/24/8/r01
Publisher
IOP Publishing
Version
Accessed
Citable Link
Terms of Use
Detailed Terms
Author's final manuscript
http://hdl.handle.net/1721.1/69991
Creative Commons Attribution-Noncommercial-Share Alike 3.0
http://creativecommons.org/licenses/by-nc-sa/3.0/
The MIT Faculty has made this article openly available. Please share
how this access benefits you. Your story matters.