Certain results on a parabolic type Monge-Ampère equation (Q1191756)

Let \(D\) be an open convex bounded set of \(\mathbb{R}^ d\), \(T>0\) and \(u\in C_ T=C({\mathcal D}\times(0,T))\) such that \(u(x,t)\) is convex in \(D\) for each \(t\in[0,T]\). The Monge-Ampère measure corresponding to the convex function \(u(\cdot,t)\) will be denoted by \(Mu(t)\) [see e.g. \textit{R. T. Rockafellar}, Convex Analysis (1970; Zbl 0193.184)]. For every \(\varphi\in C_ 0({\mathcal D}\times(0,T))\) the formula \(Qu(\varphi)=\int_ 0^ T dt\int_{\mathcal D}\varphi(x,t)Mu(t)(dx)\) defines a positive Radon measure on \({\mathcal D}\times(0,T)\) which is not a.c. in general. But if \(u\in C^ 2({\mathcal D}\times(0,T))\) then \(dQu=\text{det}[{\mathcal D}_{ij}u]dx dt\) where \({\mathcal D}_{ij}=\partial^ 2/\partial x_ i\partial x_ j\). Furthermore given a \(d\)-dimensional Brownian motion \((\Omega,F,F_ s,P,B_ s)\), \(s\geq 0\), \(S_ N\) denotes the set of \(F_ s\)- progressively measurable stochastic processes \((\sigma,b)\) such that \(\sigma:[0,\infty]\times\Omega\to R^{d^ 2}\), \(b:[0,\infty)\times\Omega\to\mathbb{R}\), \(\sigma\sigma^*>0\), \(b>0\), \(\text{Tr} \sigma\sigma^*+b\leq N\), \(b\text{det}\sigma\sigma^*\geq 1\), \(S=\bigcup_ N S_ N\) and \(X_ s^ \sigma=\int_ 0^ s\sigma_ \eta dB\eta\), \(Y_ s^ b=\int_ 0^ s b_ r dr\). Theorem. Let \(f\in C^ 2(\mathbb{R}^ d\times[0,\infty))\), such that: \(f\geq 0\), \(\| f\|_ \infty+\|\nabla f\|_ \infty+\|\Delta f\|_ \infty<\infty\), \(f=0\) outside of \(K\times[0,T]\), where \(K\) is a compact subset of \({\mathcal D}\). Then there exists a function \(u\in C_ T\) such that (i) \(u=0\) on \((\partial{\mathcal D}\times(0,T))\setminus\{(x,0):\;x\in D\}\equiv\Gamma\), (ii) there exists the right derivative \({\mathcal D}_ t^ + u\) bounded on \({\mathcal D}\times(0,T)\), (iii) \({\mathcal D}_ t^ + u dQu=f^{d+1}dx dt\) on \({\mathcal D}\times(0,T)\) (this is the generalized Monge-Ampère equation), (iv) \({\mathcal D}_ t u\in L^ \infty({\mathcal D}\times (0,T))\) and \({\mathcal D}_ t u dQu^ a=f^{n+1}dx dt\) on \({\mathcal D}\times(0,T)\) where \(Qu^ a\) is the a.c. part of \(Qu\). This measure is a.c. on \(\{{\mathcal D}_ t^ + u\geq 0\}\) and the function \(u\) can be represented on \({\mathcal D}\times(0,T)\) as \[ u(x,t)=\inf_{(\sigma,b)\in S}E\{-\int_ 0^ \tau{1\over 2}(d+1)2^{1/d+1}f(x+X_ s^ \sigma,t+Y_ s^ b)ds\}, \] where \(E\{\dots\}\) denotes the expectation with respect to the measure \(P\) and \(\tau\) is the time when the process \((x+X_ s^ \sigma,t+Y_ s^ b)\) hits \(\Gamma\).