No label defined (Q4454213)

Let \(T\in(0,\infty)\) and \(x(t),\) \(t\geq0,\) be a real-valued function on \([0,T].\) Define \(Q=\{(t,x)\colon\;t\in(0,T),\;x>x(t)\}\) and set \(\partial'Q\) for the lateral boundary of \(Q:\) \(\partial'Q=\partial Q\setminus \{(0,x)\colon\;x>x(0)\}.\) For a constant \(\nu>0\) and function \(g\) consider the problem \[ u_t(t,x)+ \tfrac12 \nu^2 u_{xx}(t,x)=0,\quad (t,x)\in Q,\qquad u=g, \quad \text{on}\;\partial Q. \tag{1\(_{\nu,g}\)} \] The main result of the paper can be summarized as follows: If a point \((t_0,x(t_0))\in \partial'Q,\) \(t_0\in[0,T),\) is ``Hölder'' regular for the problem \((1_{\nu,\,g})\) with at least one \(\nu>0\) then it is regular for \((1_{\nu,\,g})\) with any \(\nu>0.\) Precisely, let \(\nu,\nu_0>0,\) \(t_0\in[0,T),\) and let \(v\) be a probabilistic or Perron's solution of \((1_{\nu_0,\,g_0})\) where \(g_0\) is a nonnegative bounded continuous function such that \(g_0(T,x)>0\) for some \(x>x(T).\) Suppose \[ v(t_0,x+x(t_0))\leq x^\lambda \] for \(\lambda\in(0,1]\) and all \(x\in(0,T].\) Let \(u\) be a probabilistic or Perron's solution of \((1_{\nu,\,g})\) with a bounded and continuous function \(g.\) Then \[ \lim_{\substack{(t,x)\to (t_0,x_0)\\ (t,x)\in Q}} u(t,x)=g(t_0,x(t_0)). \] The proof is based on two ingredients. The first one is similar to the sufficient condition for regularity of boundary points for two-dimensional elliptic equations with measurable coefficients (exterior cone condition). The second ingredient is provided by a sort of ``square root law'', asserting that there is a constant \(c_0\in(0,\infty)\) depending only on \(\nu_0\) and \(\lambda,\) such that \[ \limsup_{h\downarrow 0} {{x(t_0+h)-x(t_0)}\over {\sqrt{h}}} \geq -c_0. \] It is worth noting that results of the type considered at the paper are already known in case of \((1/2+)\) Hölder boundaries \((\lambda\in(1/2,1]),\) whereas a typical situation, appearing in the theory of stochastic partial differential equations, is when \(x(t)\) is a trajectory of a Brownian motion and therefore is only \((1/2)\)-Hölder continuous. This makes Krylov's paper highly appreciable. For the entire collection see [Zbl 0998.00008].