Sharp derivative bounds for solutions of degenerate semi-linear partial differential equations (Q1762334)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Sharp derivative bounds for solutions of degenerate semi-linear partial differential equations
scientific article

    Statements

    Sharp derivative bounds for solutions of degenerate semi-linear partial differential equations (English)
    0 references
    0 references
    0 references
    23 November 2012
    0 references
    The authors consider semilinear partial differential equation of the form \[ \partial_{t}u(t,x)={1\over 2}\sum\limits_{i=1}^{N}V_{i}^{2}u(t,x)+V_0u(t,x)+f\left(t,x,u(t,x),\left(Vu(t,x)\right)^{T}\right),\;(t,x)\in(0,\infty)\times R^{d}, \] with initial condition \(u(0,x)=h(x), x\in \mathbb R^{d}\), where \(Vu(t,x)=(V_1u(t,x),\ldots, V_{N}u(t,x))\), \(V_1,\ldots,V_{N}\) are first-order differential operators. The main results of the present paper are summarized in the following statement. Assume that vector fields \(\{V_{i},i=\overline{0,N}\}\) satisfy the UFG condition of order \(m\). Then, if \(h\) is of polynomial grows and continuous and if \(f\) satisfies additional conditions that are specified in the paper, the considered semilinear partial differential equation is uniquely solvable in a suitable space of classical solutions and the solution is differentiable in any direction \(V\in {\mathcal W}\). Moreover, if \(h\) is a Lipschitz continuous function, then, for any \(T>0\), there exists a constant \(C\) such that, for all \((t,x)\in (0,T]\times \mathbb R^{d}\), \(| V_{[\alpha_1]}\cdots V_{[\alpha_{n}]}u(t,x)|\leq Ct^{(1-\|\alpha\|)/2}\), \(n\leq K-m-1\), with \(\alpha=(\alpha_1,\ldots,\alpha_{n})\in [{\mathcal A}_0(m)]^{n}\). If \(h\) is a continuous function of polynomial growth, but not necessary Lipschitz, then there exists a constant \(C\) such that, for all \((t,x)\in (0,T]\times \mathbb R^{d}\), \(| V_{[\alpha_1]}\cdots V_{[\alpha_{n}]}u(t,x)|\leq Ct^{-\|\alpha\|/2}\), if \(n\leq 2\), or \(n=3\) and \(\min\{\| \alpha_{i}\|, i=1,2,3\}=1\). However, if \(3\leq n\leq K-m-1\), then for any \(\delta>0\), there exists a constant \(C(\delta)\) such that, for all \((t,x)\in (0,T]\times \mathbb R^{d}\), \[ | V_{[\alpha_1]}\cdots V_{[\alpha_{n}]}u(t,x)|\leq C(\delta)t^{-\|\alpha\|/2}\left[1+t^{-n/2+1+\min[1/\|\alpha_{(1)}\|,1/2+1/(2\| \alpha_{(2)}\|)]-\delta}\right], \] where \(\|\alpha_{(1)}\|\leq\|\alpha_{(2)}\|\) stand for the two smallest elements among \(\|\alpha_1\|,\ldots,\|\alpha_{n}\|\). If \(h\) is of polynomial growth and measurable only, the considered equation is uniquely solvable as well, but in a suitable space of generalized solutions. The solution admits generalized derivatives in any direction \(V\in{\mathcal W}\) and satisfies two last bounds almost-everywhere.
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    probabilistic methods
    0 references
    UFG condition
    0 references
    backward stochastic differential equations
    0 references
    Malliavin calculus
    0 references
    0 references