See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/43095063

Variations on a theme by a singing wineglass

Article  in  EPL (Europhysics Letters) · May 2005

DOI: 10.1209/epl/i2004-10493-9

CITATIONS READS

12 4,222

4 authors, including:

Kuan-Wen Chen
The National High Magnetic Field Laboratory (National MagLab)
37 PUBLICATIONS   256 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Kuan-Wen Chen on 15 November 2015.

The user has requested enhancement of the downloaded file.

HOME | SEARCH | PACS & MSC | JOURNALS | ABOUT | CONTACT US

Variations on a theme by a singing wineglass

This article has been downloaded from IOPscience. Please scroll down to see the full text article.

2005 Europhys. Lett. 70 334

(http://iopscience.iop.org/0295-5075/70/3/334)

The Table of Contents and more related content is available

Download details:
IP Address: 140.112.113.225
The article was downloaded on 01/07/2009 at 10:01

Please note that terms and conditions apply.

EUROPHYSICS LETTERS 1 May 2005
Europhys. Lett., 70 (3), pp. 334–340 (2005)
DOI: 10.1209/epl/i2004-10493-9

Variations on a theme by a singing wineglass

Kuan-Wen Chen 1 , Chih-Kai Wang 1 , Chin-Ling Lu 1 and Yih-Yuh Chen 2,3,4
1
Hong-wen Senior High School - Taichung, Taiwan, ROC
2
Physics Department, National Taiwan University - Taipei, Taiwan, ROC
3
Institute of Astrophysics, National Taiwan University - Taipei, Taiwan, ROC
4
National Center of Theoretical Sciences at Taipei - Taipei, Taiwan, ROC

received 12 October 2004; accepted in final form 14 March 2005

published online 30 March 2005

PACS. 43.20.+g – General linear acoustics.

PACS. 62.30.+d – Mechanical and elastic waves; vibrations.

Abstract. – A “singing wineglass” is a glass which generates a more or less pure tone when
one rubs its rim with a wet finger. The frequency of this tone is lowered when water or wine is
poured into the glass. One usually associates this downward frequency shift with an effective
mass contributed by the liquid to the vibrating wineglass. We present experimental work
and simplified theoretical models to show that this idea requires modifications for a couple of
counter-intuitive phenomena.

Undoubtedly, using a wet finger to rub the rim of a wineglass to produce a sound of more or
less pure tone must have been a fun and rewarding experience for many party guests. So much
so that Benjamin Franklin even modified the idea and invented “armonica” for one’s listening
enjoyment [1]. Even scientists could not resist the charm and had also studied the sound
generated by such a “singing wineglass” in some length [1–4]. These previous works suggest
that the tone one hears consists mainly of the fundamental mode modulated by the rubbing
motion of the wet finger [1]. Thus, simply changing the rubbing speed or the magnitude of
the applied force does not alter the pitch one hears, though volume modulation can be easily
perceived. Despite its popularity, a singing wineglass might exhibit interesting phenomena
that go quite contrary to one’s intuition. Nice and easy accounts of this aspect can be found
in [1, 2]. Added to the list of counter-intuitive behavior, a singing wineglass enjoys another
phenomenon we recently encountered: The unexpected frequency shift associated with the
tone when water is present in a glass.
To begin, we first note that adding water to a wineglass readily lowers its frequency. This
phenomenon is hardly surprising if one recalls that the added water is forced by the wineglass
to participate in the vibration, thus contributing an effective mass to the vibrating glass. If
this were the whole story, then we would expect the following scenario to hold: Dipping a solid
object into a glass already filled with water to its rim must make the frequency higher, because
the amount of water inside the glass is made less by the presence of the foreign object. Wrong!
We observed a lowering in the tone when a solid column is dipped into the center of the glass
while keeping the water level intact. We also measured the frequency of an empty wineglass
completely surrounded by water put in a laterally large bucket. This frequency turned out
c EDP Sciences


Kuan-Wen Chen et al.: Variations on a theme by a singing wineglass 335

Fig. 1 – A comparison of frequencies. (a) From left to right, the wineglass is empty, filled outside,
filled inside, and filled both in and out to the same water level, respectively. (b) The frequency f2
decreases to f4 when a rigid column is dipped into the center while keeping the water level intact.

to be slightly higher than that for a glass filled with water inside. Figure 1 summarizes
schematically the ordering of the frequencies for various configurations. In table I we also
listed the data collected in the table-top experiments we performed, using a piezoelectric
microphone for sound sampling and both an oscilloscope and Adobe Audition (a waveform
analyzing software) for frequency analysis.
For both cases described above, a larger amount of water clearly contributes less to the
effective mass of the vibration. So what is going on here?
This puzzle can be most easily resolved if we forget about the true geometry of the glass
and focus on a simplified model problem. The one we start with is a two-dimensional model
in which a layer of water of thickness s is in contact with a thin slab of glass of thickness τg set
into vibration. (See fig. 2.) Setups similar to this, such as a pair of parallel plates sandwiching
a layer of different elastic material, appear common in some acoustical applications [5, 6]. In
what follows, a subscript l (g) will be used to denote a variable for the liquid (glass). Suppose
the displacement of a liquid/glass element from its equilibrium position is denoted by ξ,  then
the linearized forms of F = ma for the liquid and the glass are

∂ 2 ξl
ρl = −∇P, (1)
∂t2
∂ 2 ξg

ρg τg 2 = F̂ ξg + P n̂, (2)
∂t
where eq. (1) is Euler’s equation for fluid motion [7], with P being the pressure deviation of

Table I – Wineglass frequencies in hertz for the various water filling configurations depicted in fig. 1.
Glasses 1 and 2 are regular ones purchased in a store whereas Glass 3 is a custom-made right circular
cylindrical glass filled at different water levels for comparison.

Glass 1 Glass 2 Glass 3 Glass 3 (lower water level)

f1 1020 1280 1030 1030
f2 930 928 831 917
f3 918 857 829 915
f4 836 692 714 814
thickness (mm) 1.72 1.86 2.4 2.4
336 EUROPHYSICS LETTERS

Fig. 2 – A simplified two-dimensional model for the wineglass problem: A layer of liquid of thickness
s is sandwiched between a thin slab of glass and a solid wall. Nothing varies in the z-direction. (a) 3D
view, (b) top view.

the liquid from its equilibrium configuration, n̂ is the outward normal vector of the fluid at an
appropriate interface, ρ denotes the mass density, and F̂ is the restoring force per unit area
whose detailed form is not important for our purpose [8]. The pressure P is to be solved using
the incompressibility condition
∇ · ξl = 0. (3)

(Water can be treated as incompressible in our model because the sound speed in water, about
1500 m/s, is much greater than the flexural wave speed on the glass rim, which is estimated
to be at most 200 m/s in our experiments.)
These variables also satisfy the following boundary conditions:

ξl · n̂ = ξg · n̂ (at the liquid-glass interface), (4)

ξl · n̂ = 0 (at the solid wall y = s), (5)

which simply say that the liquid cannot penetrate either the glass or the solid wall. (But notice
that there can and in general will be slipping motion between the liquid and the wall/glass.)
Assuming a sinusoidal time dependence, we can verify that the displacement vector ξg for
the flat glass at hand can be approximated as

ξg = (0, A sin kx sin ωt) (6)

for some wave number k and small amplitude A. The reason there is no displacement in the
x-direction is that glass is inextensible for our case [2], and this form guarantees that the glass
wall has a constant length in the linear approximation when undergoing flexural wave motion.
To solve for the fluid displacement ξl , we can take the divergence of eq. (1) and make use
of eq. (3) to yield ∇2 P = 0. With the assumption that the liquid-glass interface is located
at y = 0 and the solid wall is at y = s, we can quickly derive the following solution using
Kuan-Wen Chen et al.: Variations on a theme by a singing wineglass 337

separation of variables:
 
 cosh k(s − y) sinh k(s − y)
ξl = −A cos kx sin ωt, A sin kx sin ωt , (7)
sinh ks sinh ks
ω2 cosh k(s − y)
P = −ρl A sin kx sin ωt. (8)
k sinh ks
At the liquid-glass interface y = 0 we therefore have

ω 2 cosh ks 
P n̂ = ρl ξg , (9)
k sinh ks
which reduces eq. (2) to
 
ρl

−ω 2 ρg τg 1 + coth ks ξg = F̂ ξg . (10)
ρg τg k

On the other hand, the original frequency ω0 for the case when the liquid is absent obviously
satisfies

−ω02 ρg τg ξg,0 = F̂ ξg,0 . (11)
Hence, a comparison of these two equations tells us that
 
2 ρl
ω 1+ coth ks = ω02 . (12)
ρg τg k

Because coth ks is a monotonically decreasing function of s, the inevitable conclusion is that

the frequency is lower when s is smaller, that is, when there is less water.
Having arrived at this result, we see why the idea of an effective mass contributed by the
fluid may be formally justified. This is all embedded in eq. (10), where influence of liquid
explicitly enters the scene. Tracing back, we know it arises as a direct consequence of eq. (9).
Thus, a clear understanding of what this equation means will help solve the puzzle.
One feature of this equation is that it predicts a positive pressure when the glass bulges
out. This is certainly in line with our intuition, because the pressure is then always counter-
acting the restoring force of the deformed glass, thus contributing a positive effective mass.
The presence of the factor coth ks does deserve further discussion, though, since it becomes
extremely large for small s, making the pressure increase without bound. The way we un-
derstand this feature is as follows. When a fluid element driven by the vibrating glass rushes
towards the solid wall at the other end, it simply has to stop in view of the impermeability
condition (eq. (5)). But a small stopping distance s implies a large velocity gradient in the
y-direction, which in turn means there must be a compensating flow in the x-direction because
the fluid is incompressible. This lateral fluid motion has a large velocity, which can only be
driven by a large pressure gradient. Hence, the pressure necessarily rises when s is small.
This is the physical mechanism behind eq. (9). Of course, one can also directly verify this
argument by looking at eq. (7), paying special attention to the relative orders of magnitude
of the two components of the displacement vector ξl at y = 0 for the case when s ≈ 0. In this
sense, however, we see that the concept of an added effective mass from the liquid should not
be taken too literally, because the physically existing small amount of water simply cannot
account for the large effective mass.
One can generalize the above model to make it slightly more realistic by considering a
layer of fluid sandwiched between two concentric right circular cylinders of radius r1 and
338 EUROPHYSICS LETTERS

r2 , respectively. We will assume that the outer one having a radius r2 is made of glass
and the inner one of rigid material. If we adopt a cylindrical coordinate system and write
ξ = ξθ êθ +ξr êr , then it is straightforward to show that the area enclosed by the glass exhibiting
flexural motion remains a constant in the first-order approximation of the displacement. In
other words, our setup of the problem is self-consistent, and we do not have to work with the
fully three-dimensional problem at this stage. Resorting to separation of variables, we find
the solution to be
 
r12m
ξr = A rm−1 − m+1 cos mθ sin ωt, (13)
r
 
r12m
ξθ = −A rm−1 + m+1 sin mθ sin ωt, (14)
r
 
ω2 r2m
P = ρl A rm + 1m cos mθ sin ωt (15)
m r
for some integer m > 1 and some number A. (Previous work shows that one has m = 2 as the
lowest mode; see [2, 4].) Before going on, we should remark that, for the case when r1 = 0,
ref. [4] contains one mistake: In order to satisfy the incompressibility condition, the radial and
the tangential components of the velocity of the bulk fluid must have the same magnitudes,
as opposed to the 2 : 1 ratio claimed in sect. VI of this reference. However, the fact that the
glass in our problem only sustains flexural wave does imply that this ratio is 2 : 1 for the glass
on the rim. (See ref. [2] and p. 579 of ref. [9].) This then implies that actually there is a
slipping motion between the fluid and the glass, as was already mentioned before.
The results above allow one to quickly derive
 
2m
ω2 1 + (r1 /r2 )
P = ρl ξr (16)
(m/r2 ) 1 − (r1 /r2 )2m

at the liquid-glass interface. This equation clearly is the cylindrical counterpart of eq. (9),
with the obvious identification of k = m/r2 . And once again, we know this implies that the
liquid-filled cylindrical glass vibrates at a frequency higher than that for one having a solid
placed at its center.
This calculation also implies that the frequency satisfies
  
2m
2 ρl 1 + (r1 /r2 )
ω 1+ = ω02 , (17)
ρg τg k 1 − (r1 /r2 )2m

which closely resembles eq. (12). For the case when r1 = 0 we then have
 
ρl
ω2 1 + = ω02 . (18)
ρg τg k
Of course, we can just as well solve the exterior problem by placing the fluid outside the
cylindrical glass. Interestingly, the algebra shows that the formula works out exactly like
eq. (18). Thus, for a very thin glass, the interior and the exterior problems turn out to
produce the same tone! But for thicker glass wall, our argument above does suggest a higher
frequency for the exterior problem, which has been experimentally verified in table I. (The
fact that f2 and f3 differ by only 2 Hz for Glass 3, whose geometry conforms much better to
that investigated in our model, suggests that the thin-glass approximation is a valid one, and
that 2.4 mm can be considered as thin enough for this purpose.)
Kuan-Wen Chen et al.: Variations on a theme by a singing wineglass 339

Fig. 3 Fig. 4

Fig. 3 – Water level dependence of the wineglass frequency. The slope 3.1 is close to the predicted
value of n + 1, with n = 2.

Fig. 4 – Liquid density dependence of the wineglass frequency. The liquids used (in increasing density)
are: alcohol, linseed oil, water, vinegar, and brine, respectively.

For a right circular cylindrical glass of height H filled with liquid up to a height h, one
will have to solve the full three-dimensional problem, because now there is an extra restoring
force coming from the bending of the glass wall from its upright position. However, since
we are dealing with a linear system here, we know this force must be proportional to the
pressure of the fluid in an averaged sense. Thus, under certain approximations, we may
absorb this force into the pressure term so as to reduce the problem into an equivalent two-
h h h
dimensional model. Specifically, defining P  ≡ 0 P dz/h, ξlr  ≡ 0 ξlr dz/h = 0 ξgr dz/h,
H
and ξgr  ≡ 0 ξgr dz/H, and in view of eqs. (1) and (2), we now have
−ρl ω 2 ξlr  = −êr · ∇P , (19)

 h
−ρg τg ω 2 ξgr  = êr · F̂ ξg + γ P n̂, (20)
H
h
ξgr dz H
ξlr  =  0H ξgr , (21)
0
ξgr dz h
where γ is some numerical factor. Approximating the radial displacement of the glass by that
of a vertical bar clamped at its base and free at the top end, one has [10]
ξgr ∝ (cosh α + cos α) (cosh αz/H − cos αz/H) − (sinh α − sin α)(sinh αz/H − sin αz/H),
where α ≈ 1.8751. Due to cancellations in the next higher-order terms in the Taylor expansion
of the hyperbolic and sinusoidal functions, the above is well approximated by
(αz/H)3
ξgr ∝ (cosh α + cos α) (αz/H)2 − (sinh α − sin α) .
3
For h/H
0.5, we may retain only the lowest-order term and take ξgr ∝ z n for n = 2; and for
higher water level it is found that n = 1.5 is a good approximation [4]. At any rate, eqs. (16)
and (21) suggest   
2m n
ω2 1 + (r1 /r2 ) h
P  ≈ γρl 2m ξgr ,
(m/r2 ) 1 − (r1 /r2 ) H
340 EUROPHYSICS LETTERS

which immediately yields

   n+1 
2m
ρl 1 + (r1 /r2 ) h
ω2 1+γ 2m = ω02 , (22)
ρg τg k 1 − (r1 /r2 ) H

in analogy with our previous two-dimensional analysis. With the previously mentioned custom-
made Pyrex glass having H = 11.74 cm, r2 = 4.02 cm, τg = 0.24 cm and ρg = 2.23 gm/cm3 ,
we see in fig. 3 that the agreement is quite good if we take n = 2. This supplements the result
of [4] when the water level is not high. In addition, we have also verified the ρl -dependence
using various liquids. This is shown in fig. 4.
Concluding, our experiments show that the finite-size effect of the liquid on the tone of the
singing wineglass appears counter-intuitive if one attempts to explain it in terms of the “effec-
tive fluid mass” driven by the glass. But using simplified models, we were able to derive quanti-
tative relations and support them via experiments. In particular, we show that a more coherent
way of viewing the whole situation is to place the emphasis on how the fluid pressure varies
as the system geometry is changed; and an intuitive picture explaining it is also proposed.

∗∗∗

The first three authors were supported by a grant from the National Taiwan Science
Education Center. YYC was supported by the National Science Council of the Republic of
China under grant NSC92-2112-M-002-020.

REFERENCES

[1] Rossing T. D., J. Acoust. Soc. Am., 95 (1994) 1106.

[2] Rossing T. D., The Physics Teacher, 28 (1990) 582.
[3] Strutt J. W. (Lord Rayleigh), The Theory of Sound, 2nd edition, Vol. I (Macmillan, Lon-
don) 1894, pp. 383-389, reprinted by Dover, New York, 1945.
[4] French A. P., Am. J. Phys., 51 (1983) 688.
[5] Cremer L. and Heckl M., (translated and revised by Ungar E. E.), Structure-Borne Sound,
2nd edition (Springer-Verlag, Berlin) 1988, pp. 450-462.
[6] Junger M. C. and Feit D., Sound, Structures, and Their Interaction, 2nd edition (The MIT
Press, Cambridge) 1986, Chapts. 8 and 9.
[7] Acheson D. J., Elementary Fluid Dynamics (Clarendon Press, Oxford) 1990, p. 8.
[8] Readers interested in its detailed form can consult pp. 168-194 of ref. [5] or Chapt. 7 of ref. [6].
[9] Fletcher N. H. and Rossing T. D., The Physics of Musical Instruments (Springer-Verlag,
New York Inc.) 1991, pp. 65-88.
[10] See pp. 276-279 of ref. [3].

View publication stats