Nonclassical solutions are non-existent for the heat equation and rare for nonlinear diffusion (Q1922995)

If a system \((F)\) of partial differential equations for an unknown function \(u=u(x_1,\dots,x_n)\) admits an infinitesimal Lie transformation \(Z=\sum X_i\partial/\partial x_i+U\partial/\partial u\), the invariant solutions satisfy the invariant surface condition \((S)\): \(\sum X_i\partial u/\partial x_i=U\) which permits to reduce the primary system \((F)\) and, in favourable cases, to determine some particular explicit solutions. However, the true sense of the method consists only in the compatibility of \((F,S)\) and the Lie group invariance is in principle needless. This results in the so-called nonclassical method. The authors however prove that this improvement does not bring any new solutions for all second order linear equations with constant coefficients, and very little for the nonlinear equation \(f(u)u_t= \partial^2u/\partial x^2+\partial^2 u/\partial y^2\).