A diffusion problem involving two free boundaries (Q807849)

The author considers the boundary value problem \[ (*)\quad L[u]:=(\phi (u_ x))_ x-(c(u))_ t=0\text{ in } Q_ T=(0,1)\times (0,T], \] \[ u(0,t)=f_ 1(t),\quad u(1,t)=f_ 2(t),\quad u(x,0)=u_ 0(x). \] In addition to differentiability conditions it is assumed \(that\) c\({}'(x)>0\), \(\phi '(x)>0\) for \(x\neq 0\); \(f_ 1>0\), \(f_ 2<0\) and that there exists a pair of numbers \(0<a<b<1\) such \(that\) u\({}_ 0(x)>0\) on [0,a], \(u_ 0(x)<0\) on [b,1] and \(u'_ 0(x)\leq 0\) on [a,b]. The mathematical difficulty for problem (*) lies in the fact that the equation may degenerate, where the nature of degeneracy is characterized by the behavior of the functions \(c'(x)\) and \(\phi '(x)\) as \(x\to 0\). Here it is allowed that \(c'(x)\), \(\phi '(x)\) may oscillate near zero, may be unbounded above and may not be bounded away from zero as \(x\to 0\). The author proves the existence of a weak solution of problem (*), where the weak solution is constructed as the limit of a sequence of classical solutions of the approximating problems. Further some regularity properties of the weak solution are obtained. The author remarks, that the uniqueness of a weak solution under precisely those assumptions on the data which are needed for the existence is still an open problem.