7.

Biharmonic equation solid mechanics and creeping ow

The biharmonic equation is the \square of the Laplace equation", is minimised, subject to taking given values ( j j ) = j at a nite number of points (at least
u u X ;Y U

3 of them), when is a so-called thin-plate spline |an interpolating function of the simple form
u

( )= + 1 + 2 +
X 2 log

2 = 0
u ; (1) u x; y a b x b y

j
cj rj rj ;

Pj j = Pj j j =
where
= 2 21 +


+ 2 2n is the Laplacian operator. Like the Laplace equation, the
@ =@x @ =@x
Pj j jj2==0(to ensure
where
c Y
r x
2
j ) +(
X
2
j ) whose later coecients j satisfy the conditions
y Y ; c

that ( ) is nite. The thin-plate spline is a standard device for constructing

I u
c c X

biharmonic equation is elliptic, but, being of order four rather than two, it requires two boundary a smooth function through data given at points arbitrarily distributed
conditions rather than one to dene a unique solution. In 2D, it is the equation satised, to a good P in the plane. The technique
can be extended to 3D by adding in a term 3 and a condition j j j = 0 and replacing the 2D
b z c Z
approximation, by a small transverse de ection of a thin at elastic plate. fundamental solution 2 log by the corresponding 3D solution . Extension to more than three
r r r

dimensions is possible, but one then needs to replace the biharmonic equation by a polyharmonic
The eigenvalues and eigenfunctions of the biharmonic equation
r = 0 with 2. Generalisations of these data tting methods based on functions
operator
2 , with suitable homogeneous boundary
u r >

other than solutions of the biharmonic equation are the business of the eld of radial basis functions.
conditions, give the modes of transverse vibration of
such a plate. The physicist and astronomer Ernst The biharmonic equation also arises in the
Chladni (1756{1827) carried out a famous series of ex- theory of steady Stokes (i.e., speed  0)
periments using particles of sand to locate the nodal ow of viscous uids, where it is the equa-
curves of a plate clamped at its centre and excited in tion satised by the stream function. For
697 Hz 1.295 kHz
various modes. The resulting patterns are known as example, it has been shown that in the
Chladni gures and some results from an experiment vicinity of a right-angle corner, the eigen-
of this kind are shown in Figure 1. One can try the functions of the biharmonic operator oscil-
experiment oneself at some science museums. late innitely often in sign, with each suc-
cessive region of oscillation being 16 56743
The biharmonic equation has two independent funda-
:

times smaller in size and 36267 55 times

mental solutions (spherically symmetric and singular
:

smaller in amplitude than the last. This

at the origin), one of which is the fundamental so- result can be interpreted as the statement
lution of the Laplace equation. These are log and r
that a plate distorted in a certain way will
2 log in 2D, 1 and in 3D, 2 and log in 4D,
1.409 kHz 1.538 kHz
r r r r r r
and 2 d and 4 d in dimensions  5. In the 2D
r r d
uid mo- Fig. 2: A thin-plate spline, interpolating
case, just as any harmonic function in the ( )-plane
is the real part of an analytic function ( ), where
in principle bend back and forth innitely
often near a corner, or that a
x; y
tion at Reynolds number 0 will in princi- function values given at 9 scattered points
f z
ple exhibit an innite sequence of counter-
z = + i , so any biharmonic function is the real
x y
rotating \Moatt vortices ".
part of a function of the form ( ) + ( ), where f z zg z f

and are analytic and = i is the complex con-

g z x y

jugate of . For instance, 2 log = <( ( )) where

z r r zg z

( ) = log . The functions and (which are not

g z z z f g

uniquely determined) are the Goursat functions of the

1.737 kHz 1.824 kHz problem. References
Fig. 1: Chladni gures In two dimensions, the function ( ) minimising
N. Aronszajn, T. M. Creese, and L. J. Lipkin, Polyharmonic functions, Oxford, 1983.
for a square plate the value of the integral
u x; y
P. E. Bjrstad and B. P. Tjstheim, High precision solutions of two fourth order eigenvalue problems, Computing
63 (1999), 97{107.
ZZ n o L. V. Kantorovich and V. I. Krylov, Approximate methods of higher analysis, Interscience, 1958.
( )=
I u ( uxx )2 + 2 ( uxy )2 + ( uyy
2
) d d x y A. E. H. Love, A treatise on the mathematical theory of elasticity, Cambridge, 4th ed., 1934.

H. Ockendon and J. R. Ockendon, Viscous ow, Cambridge University Press, 1995.

over a given domain
, subject to suitable conditions on the boundary, can be shown to satisfy M. J. D. Powell, A review of methods for multivariable interpolation at scattered data points, in I. S. Du and G.
the biharmonic equation on
. This is an analogue for the biharmonic equation of the Dirichlet A. Watson, eds., The state of the art in numerical analysis, Clarendon Press, 1997.
integral for the Laplace equation (! ref ). In particular, if
is the whole of IR, then this integral W. L. Wendland, Elliptic systems in the plane, Pitman, 1979.

28 February 2001: David Handscomb http://www.ph.unimelb.edu.au/lecdem/wb5.htm

http://www.regencynet.com/modal/chladni.htm
c 1999