You are on page 1of 8

CHINESE JOURNAL OF PHYSICS VOL. 38, NO.

2-I APRIL 2000

On the Low Frequency Elastic Response of a Spherical Particle

Zhen Ye
Department of Physics, National Central University,
Chung-Li, Taiwan 320, R.O.C.
(Received October 19, 1999)

This article reconsiders the vibration of a freely suspended elastic sphere. For the first
time, equations determining general vibration eigen-modes are obtained in the closed forms.
The previously published results are also re-derived, and examined in detail. In the procedure,
validity conditions are explicitly inspected, and necessary approximations are elucidated. It
is found that although tempting, the previous formulas for the eigen-frequencies of various
vibrations are only valid for a limited number of multipole oscillations.

PACS. 78.30.Cp – Liquids.

I. Introduction

The vibration of a freely suspended solid sphere has been a subject of thorough research.
The formulation of the problem is well documented in the literature. A detailed account was given
by Love in his classic treatise on the theory of elasticity [1]. In general, the vibrations can be
classified into two categories: class one and class two. Modes of the vibrations of the first class
are associated with the rotatory motions of the sphere, i.e. the torsional modes, while modes of
the second class are related to the change of shape due to vibrations in the radial direction, such
as the pulsating and the spheroidal modes. The former have also been called the torsional modes,
whereas the latter are called the spheroidal modes [2]. Both vibrations can be readily computed
through standard modal series expansion (e.g. Nishiguchi and Sakuma [3]). Recently, Bastrukov
has attempted to consider the problem further [4]. He proposed a continuum approach to describe
the vibration of a freely suspended elastic sphere. The vibration is interpreted in terms of the long
wavelength normal modes of a perfect elastic globe. The eigen-frequencies of rotatory and radial
oscillations are derived in a closed form as functions of the sphere radius and multipole degree n.
The frequencies of spheroidal vibrations were found to be

!R;n = !0 [2(n ¡ 1)(2n + 1)]1=2 (n > 1); (1)

whereas the eigen-modes of the rotatory or torsional vibrations are

!T ;n = !0 [(n ¡ 1)(2n + 3)]1=2 (n > 1): (2)


p
Here !0 = ¹=(½a2 ), where ¹ and ½ are the shear modulus and the bulk density respectively.
Though the results are simple and attractive, their range of validity was not discussed. The
derivation of Bastrukov tempted one to believe that the formulas are valid for any integers greater

103 °c 2000 THE PHYSICAL SOCIETY


OF THE REPUBLIC OF CHINA
104 ON LOW FREQUENCY ELASTIC RESPONSE OF ¢¢¢ VOL. 38

than 1. Moreover, the physical insight is obscured by the complication involved in Bastrukov’s
approach. Because of this and also because similar results based on such an obscuring approach
continue to appear in the literature using this approach (e.g. Bastrukov et al., Physica A 250,435,
1998), it is necessary to further investigate the problem and examine the validity of these formulas
to see under exactly what conditions both formula are usable.
The present work provides a careful examination of the results of Bastrukov. We found that
the usage of his results is in fact very limited and in some cases it is essentially invalid. To show
our points, we follow the well-known exact procedures in such textbooks as e.g. Ref. [1, 2, 5] and
the formulation given in Ref. [3]. We explore under what genuine approximations the results of
Bastrukov follow. We will inspect when the results become inconsistent with the approximations.
For the convenience of the general readers, we purposely provide the detailed steps, although some
could be found in standard textbooks. It is shown that though tempting, the expressions (1) and
(2) are very imprecise and are basically unusable. This is seen by comparing the approximate
results to exact results.

II. The formulation

The mechanical vibration of an elastic medium is governed by a set of standard Navier-Stoke


equations, stemming from the mass conservation and the Newton law [1, 2]. For the convenience
of the general readers, we summarize the results as follows.
The equation of motion is given by

@ ~u 2
(¸ + 2¹ )r(r ¢~u) ¡ 2¹r £ -~ + ½f~ = ½ 2 ; (3)
@t

where ~u is the displacement vector field, f~ refers to the body force, - = 21 r £ ~u, and ¸ and
¹ are respectively the two Lamé constants characterizing the elastic properties; the later is the
usual shear modulus, and the bulk modulus B is related to ¸ and ¹ as B = ¸ + 2¹=3. Note that
the body force may refer to forces such as gravitational or Coulomb force; however, in the rest
discussion, this force term will be dropped.
The stress tensor at a point is expressed in terms of the strain tensor in the rectangular
coordinates

¿ij = ¸r ¢~u±ij + 2¹ ²ij ; (4)

where ±ij is the usual Kronecker tensor, and


µ ¶
1 @ui @uj
²ij = + : (5)
2 @xj @xi

It is shown [6] that a vector field can always be separated into the gradient of a scalar
and the curl of a zero-divergence vector. The equation can be much simplified by introducing the
scalar and vector potentials such that

~
~u ´ r© + r £ A; ~ = 0:
r ¢A (6)
VOL. 38 ZHEN YE 105

With this notion, Eq. (3) yields

@ 2©
(¸ + 2¹ )r2 © = ½ ; (7)
@t2
and

@2A~
~=½
¹r2 A : (8)
@t2
q q
The phase speeds for the compressional and shear waves are therefore cl = ¸ +2¹ ½ and cs = ¹½
respectively.
It is clear that once the scalar and vector fields are known, the displacement field and the
stress tensor can be subsequently calculated. To solve for © and A, ~ proper boundary conditions
need be considered.
Now we consider the vibration of a free solid sphere. Take the vibration frequency as !.
Eqs. (7) and (8) become
!
(r2 + kl2 )© = 0; kl ´ ; (9)
cl
and
~ = 0; !
(r2 + ks2 )A ks ´ : (10)
cs
These two equations can be solved by separation of variables for the spherical geometry. The
solution for the scalar field of multipole mode n can be written in terms of the spherical coordinates
as

© = An jn (kl r)Pn (cos µ); n = 0; 1; 2; : : : ; (11)

where An is the unknown coefficient to be determined by the boundary conditions, jn (¢) is the
spherical Bessel functions of the first kind, and Pn (¢) is the Legendre polynomials of the first
kind. Here we consider that the field must remain Qfinite at the origin. The solution for the vector
field can be constructed in terms of a scalar field ,
Y
~ = r £ (~r ):
A (12)
Q
Function satisfies the following equation,
Y
(r2 + ks2 ) = 0: (13)

The solution can also be written in the same form as Eq. (11),
Y
= Bn jn (ks r)Pn (cos µ); (14)

where Bn is the coefficient to be determined by the boundary condition.


106 ON LOW FREQUENCY ELASTIC RESPONSE OF ¢¢¢ VOL. 38

Substituting the solutions in Eqs. (11) and (14) into the displacement field expression, and
subsequently into the stress tensor formula, we invoke the stress-free boundary condition

¿rr jr=a = 0; ¿rµ jr=a = 0: (15)

Except for an arbitrary constant, the coefficients An and Bn can be determined by these equations.
In the next section we consider the results.

III. The results

The solution for the vibration of a free solid sphere can be obtained from Eqs. (11), (14)
and the boundary condition (15) with Eqs. (6) and (12). Two vibrational modes can be identified
~ = 0, and the other is the torsional modes appearing when
[4]: one is the spheroidal modes, i.e. A
© = 0.

III-1. Spheroidal vibration


In the present case, by solving the boundary value problem, the eigen-frequencies are shown
to be determined by the following equation in general,
¯ ¯
¯ an bn ¯
det ¯ ¯
¯ cn dn ¯ = 0; (16)

where
(¡ 1)n ® ¡ n¡ 2 £ 2 ¤
an = ¯ jn (® ) ¡ 2(n ¡ 1)® jn¡ 1 (® ) ;
2n + 1
· ¸
(¡ 1)n+1 ® ¡ n ¯ 2 2(n + 2)
bn = jn (® ) ¡ j( n + 1)(® ) ;
2n + 1 ®2 ®
µ ¶
¡1 n 2
cn = [¯ jn (¯ ) ¡ 2(n ¡ 1)jn¡ 1 (¯ )]
¯
µ ¶
¡1 n n
dn = [¯ 2 jn (¯ ) ¡ 2(n + 2)¯ jn+1 (¯ )];
¯ N +1
with

® = kl a; ¯ = ks a:

Eigen-equation (16) is solvable numerically. However, under certain approximations, an


analytic solution may be obtained. The first is in line with Bastrukov by taking the incom-
pressibility approximation by taking ® ! 0, i.e. cl ! 1. In this limit, the eigen-equation
yields
n
[¯ 2 ¡ 2(n ¡ 1)(2n + 1)][¯ 2 jn (¯ ) ¡ 2(n + 2)¯ jn+1 (¯ )]
n+1 (17)
= ¡ ¯ 2 [¯ 2 jn (¯ ) ¡ 2(n ¡ 1)¯ jn¡ 1 (¯ )]:
VOL. 38 ZHEN YE 107

This is the exact equation determing the eigen spheroidal modes and can be solved numerically.
For each n, there actually will be an infinite number of eigen-frequencies, due to the oscillatory
feature of the Bessel function. This is not seen in Bastrukov’s approach.
When further taking the approximation that
r
4(2n + 1)(2n + 3)
¯¿ ; (18)
n+1
we have
¯n ¯ 2n + 1
jn (¯ ) ¼ ; jn+1 (¯ ) ¼ jn (¯ ); and jn¡ 1 (¯ ) ¼ jn (¯ ): (19)
(2n + 1)!! 2n + 3 ¯
Then the solution for ¯ from Eq. (17) is obtained as
p
¯n ¼ 2(n ¡ 1)(2n + 1); (20)
p
which is reduced to, by noting cs = ¹=½,
p p p
!n = ¹=(½a2 ) 2(n ¡ 1)(2N = 1) = !0 2(n ¡ 1)(2n + 1); with
p (21)
!0 = ¹=(½a2 ):

This is essentially the result in Eq. (1). Eq. (21) only gives the lowest eigen-frequency for
each multipole n. Therefore we see that in order to get the result of Bastrukov one has to
invoke the following
q approximations. (1) The incompressibility: cl ! 1. (2) Low frequency
4(2n+1)(2n+3)
limit: ¯ ¿ n+1 . Note that in [4], the second condition is replaced by the low
frequency limit, i. e. ks a = ¯ ¿ 1. This is not true, as the result is against this requirement,
because (ks a)n ´ ¯ n is always bigger than one. (3) The higher-order eigen-frequencies associated
with each n are negligible. However, for reasons below, some of these approximations are not
appropriate.
Taking the solution (21) into the requiring condition in Eq. (18), the number of actually
existing modes can be estimated from
r
p 4(2n + 1)(2n + 3)
2(n ¡ 1)(2n + 1) ¿ ; (22)
n+1
which gives

n2 ¡ 4n ¡ 7 ¿ 0: (23)

It follows that n ¿ 5. One may conclude that formula (1) at most predicts the lowest eigen-
frequencies for multipole mode n less than 5. However, even for n = 2 we find large errors: the
numerical computation gives ¯ = 2:13 while Eq. (1) gives 3.162. The more detailed comparison
between the exact results for the lowest mode from Eq. (17) and the approximate results from
Eq. (1) is presented in Table I. Here it shows that there are significant discrepancies between the
exact and approximate results. The approximate formula steadily overestimate the results.
108 ON LOW FREQUENCY ELASTIC RESPONSE OF ¢¢¢ VOL. 38

TABLE I. The comparison between the approximate results from Eq. (1) and the exact results
from Eq. (17).

n Exact Approximate Discrepancy

2 2.13 3.16 48%


3 3.35 5.25 57%
4 4.45 7.35 65%
5 5.53 9.38 70%

As mentioned earlier, there is an infinite number of modes corresponding to each n. For


example, the frequency of the second order mode for n = 2 is at about ¯ = 4:90, which is
in fact lower than the frequency of the lowest order mode for n = 3, thus possibly leading to
mis-interpretation of data if the approximate formula is to be used. This has not been discussed
in [4] .
We further note on the cases with the monopole n = 0 and dipole n = 1. When n = 0,
we arrive at
tan(® ) 1
= 2 ; (24)
® 1 ¡ ¯4

which is just the result given in Love [1]. When ® ! 0, ¯ = 0, there thus will be no pulsating
mode, as expected because the solid is incompressible. For the dipole situation, we derived from
Eq. (17)
1
¯ j1 (¯ ) = j2 (¯ ); (25)
2
which gives rise, to the lowest mode ¯ = 3:051. The consideration of these two cases are
incomplete in Bastrukov’s paper.

III-2. Torsional vibration


In this situation, without approximation the eigen-frequency equation is

(n ¡ 1)jn (¯ ) ¡ ¯ jn+1 (¯ ) = 0: (26)

This equation can be genuinely solved numerically. An infinite number of eigen-frequencies will
be associated with each multipole n. For n = 1, we have


tan(¯ ) = ; (27)
3 ¡ ¯2

which is in agreement with that in [1].


VOL. 38 ZHEN YE 109

TABLE II. The comparison between the approximate results from Eq. (2) and the exact results
from Eq. (26).

n Exact Approximate Discrepancy

2 2.30 2.64 15%


3 3.61 4.24 17%
4 4.81 5.74 19%
5 5.96 7.21 21%
6 7.09 8.66 22%
7 8.19 10.1 23%
8 9.28 11.5 24%

In order to obtain formula (2), we again take the approximations in Eqs. (18) and (19), and
then get

¯ n2 ¼ (n ¡ 1)(2n + 3): (28)

This is just the result in Eq. (2). This solution should satisfy Eq. (18), leading to

n2 ¡ 8n ¡ 5 ¿ 0:

It follows n ¿ 8, the maximum possible number of torsional modes that can be described by
Eq. (2). However, we find that the results from Eq. (2) are not accurate. We compare the results
from this equation and the results from the rigorous Eq. (26) for the lowest order modes in Table
II.
Here, again it shows that though smaller than the case with the spheroidal modes, the
differences between the exact and approximate results are substantial and the approximate formula
constantly overestimate the results.
Due to the oscillatory behavior of the Bessel function, there is an infinite number of eigen-
frequencies associated with each n. For example, for n = 2, the frequency for the second mode
is at about ¯ = 6:55. This important information has also been ignored by Bastrukov.

IV. Summary

In summary, this paper reconsidered the vibration of a free solid sphere in general, and
inspected the results published in [4]. The exact eigen-frequency equations were deduced from
the standard textbook approach. It is shown that certain approximations must be used to recover
Bastrukov’s formulas. The necessary conditions for the validity of the results are unfolded, thus
providing a useful guideline in inspecting the previous results. It was shown that the validity of
the previous analytic results is questionable. By comparing to the exact numerical results, we see
large deviations in the approximate results, indicating that the approach taken by Bastrukov misses
110 ON LOW FREQUENCY ELASTIC RESPONSE OF ¢¢¢ VOL. 38

important information about the vibration of a free solid sphere. From our derivation, we conclude
that the mistake in Bastrukov’s result lies in that he applied the long wavelength approximation
to situations in which this approximation is not valid. Finally we note that the consideration here
has not included the azimuthal dependence of vibration; if it is to be included, degeneracy will
appear, which is beyond the scope of the present study.

Acknowledgments

Discussion with fellow members at NCU is much appreciated. Communication with Dr.
Bastrukov is also acknowledged, not suggesting he agrees or does not agree with the paper . In
fact, Dr. Bastrukov does not believe that he had made invalid approximations. The work received
support from the National Science Council.

References
[1] A. E. H. Love, A Treatise on the Mathematical Theory of Elasticity, (Dover, New York, 1944).
[2] B. A. Auld, Acoustic Fields and Waves in Solids, (John Wiley & Sons, New York, 1973).
[3] N. Nishiguchi and T. Sakuma, Sol. Stat. Comm. 38, 1073 (1981).
[4] S. I. Bastrukov, Phys. Rev. E49, 3166 (1994).
[5] F. K. Graff, Wave Motion in Elastic Solid, (Ohio University Press, Ohio, 1994).
[6] P. M. Morse and H. Feshbach, Method of Theoretical Physics, (McGraw-Hill, Inc., New York, 1953).

You might also like