Exact nonnegative solutions to the multidimensional nonlinear diffusion equation (Q1288115)

The authors develop and generalize, to the multidimensional case, their results [Sib. Mat. Zh. 38, No. 5, 1130-1139 (1997; Zbl 0880.35054)] on constructing exact nonnegative solutions to the nonlinear diffusion equation \[ u_t = \nabla\cdot (u^{\lambda} \nabla u),\qquad x\in\Omega, \] where \(\Omega\subset\mathbb R^{n}\) is a bounded domain; \(u\geq 0\), and \(\lambda\in \mathbb R\), \(\lambda\neq 0\). The equation is rewritten in the form of an equivalent system \[ u_t + \nabla\cdot \big(uf(x,t)\big) =0, \qquad f(x,t) = -u^{\lambda - 1}\nabla u, \] overdetermined with respect to the temperature \(u\). The first equation with a given vector-valued function \(f\) is the Liouville equation in the temperature \(u\) for the system of ordinary differential equations \(\dot{x}=f(x,t)\), \(x \in \mathbb R^n\). The functions \[ Z_k(x,t) = \tfrac{1}{2}\big(x, A_k(t)x\big) + \big(x, B_k(t)\big) + C_k(t),\qquad k=1, 2 \] are considered, where \(A_k(t)=[a_{kij}(t)]\) are \((n\times n)\)-matrices; \(B_k(t) = (b_{k1}(t),\dots, b_{kn}(t))'\) are vectors; \(C_k\) are scalar functions, and \(a_{kij}\), \(b_{k1}\), \(C_k\in C^1(\mathbb R^+)\); \(i, j = 1, 2,\dots ,n\); \((\cdot , \cdot)\) is the inner product in \(\mathbb R^n\). The first result reads: If \(f(x,t)\) in the equivalent system is given by the equality \(f(x,t)=-(pZ_1^{p-1}\nabla Z_1 + \nabla Z_2)\) then the function \[ u(x,t)=[\lambda [\tfrac{1}{2}(x, A_1(t)x) + (x, B_1(t)) + C_k(t)]^p_+ +\lambda [\tfrac{1}{2}(x, A_2(t)x)+ (x, B_2(t)) + C_2(t)]]^{1/\lambda}_+ \] is an exact nonnegative solution to the nonlinear diffusion equation. Moreover, \(A_k(t)\), \(B_k(t)\), and \(C_k(t)\) are defined from the resolving system for the nonlinear diffusion equation which is analyzed in detail. The properties of solutions to the resolving system enable us to expose some new exact solutions to nonlinear diffusion equations and, in particular, exemplify the application of the obtained results to the equations \(u_t = \Delta\ln u\) and \(u_t = \nabla\cdot(u^{\lambda} \nabla u) + \alpha u^{1-\lambda}\).