Compacton Equations: Approximate Solitons With Compact Support
Compacton Equations: Approximate Solitons With Compact Support
Compacton equations
The KdV equation ut = (u2 )x + uxxx (! ref ) mixes nonlinear convection with linear dispersion. Its solutions include the famous solitary waves, which pass through one another as solitons with no lasting e ect other than a phase shift. These solitary waves are localised in space in the sense that they decay exponentially. In the early 1990s Rosenau and Hyman, motivated by the formation of patterns in liquid drops (! ref ), investigated generalisations of the KdV equation in which the dispersion too is nonlinear. For m > 0 and 1 < n 3, they de ned the compacton equation K (m; n) by 2u 1 PSfrag replacements 0 60 40
= 0:
(1)
The equation K (2; 2), for example, is ut + (u2 )x + (u2 )xxx = 0. Qualitatively, we can guess what the e ect of the restriction n > 1 will be. We now have amplitude-dependent dispersion which shuts o as juj ! 0. This will tend to eliminate exponentially decaying tails, for in such a tail, the dispersion term that makes the signal spread would be negligible. Thus the possibility arises of traveling wave solutions with compact support , i.e., exactly zero value outside a bounded interval. To calculate a traveling wave solution of (1), we make the substitution u(x; t) = U (s) with s = x ct for some wave velocity c. After some manipulations the PDE reduces to U m 1 cU CU 2 = D (Us )2 + m +2 3 for arbitrary constants C and D. If we choose C = D = 0 we nd that there are solutions with compact support for any c 2 R, known as compactons . Here are three of the simplest cases:
50
t
Fig. 1:
25 20 0 50 25 0
K (2; 2) : U (s) = (4c=3) cos2 (s=4); K (3; 2) : U (s) = (37:5c s2)=30; K (3; 3) : U (s) =
jsj 2 jsj p
3 2
37:5c
The pictures are suggestive, but in fact, Rosenau and Hyman found that compacton interactions are not mathematically perfect. Though it is not visible in our gure, a small residue consisting of low amplitude and hence low-velocity \compacton-anticompacton pairs" is left behind after a compacton interaction; the behaviour of compactons as particles is only approximate. This is probably related to the fact that unlike the KdV equation, compacton equations apparently do not possess an in nite set of conservation laws. Compacton equations are among the hardest one-dimensional PDEs to solve numerically, because they contain terms that are simultaneously nonlinear and of high order. Nevertheless, most of what is known about them is based on numerical experiments; there is little mathematical theory.
Note that as with a KdV soliton, the amplitude of a compacton varies with its velocity. In cases (2,2) and (3,3), however, the width is xed. The use of the word \compacton" suggests that these traveling wave solutions to K (m; n) behave like solitons, interacting as particles in the sense that the long-term e ect of an interaction is a phase shift and nothing more. Approximately speaking, this is what Rosenau and Hyman discovered numerically. For certain values of m and n, a fast compacton will overtake a slow one, pass through it, and emerge approximately unchanged. More general initial data typically separates into a train of two or more compactons that travel at the appropriate individual speeds. In Figure 1, for example, note how the initial signal, with global support, separates into at least three waves of di erent amplitudes but the same widths, each with compact support.
4 March 2001: David Kay, Marc Maestracci, Nick Trefethen
J. de Frutos, M. A. Lopez-Marcus and J. M. Sanz-Serna, A nite di erence scheme for the (2 2) compacton equation, J. Comp. Phys., 120 (1995), 248{252. P. Rosenau and J. M. Hyman, Compactons: solitons with nite wavelength, Phys. Rev. Lett., 70 (1993), 564{567.
K ; c 1999
References