Convex Hamilton-Jacobi equations under superlinear growth conditions on data (Q538471)

The main result of this paper is to propose a new way to deal with the PDE \[ \begin{aligned} \frac{\partial u}{\partial t}+H(x,t,Du,D^2u)= 0 \quad &\text{in }\mathbb R^N\times (0,T),\\ u(x,0)= \psi(x)\quad &\text{in }\mathbb R^N,\end{aligned}\tag{1} \] where \[ H(x,t,q,X)=\sup_{\alpha\in A} \big\{-(b(x,t,\alpha),q)-l(x,t,\alpha)-\text{Trace}[\sigma (x,t,\alpha)\sigma^T(x,t,\alpha)X]\big\}\tag{2} \] is convex with respect to the gradient variable, with diffusion matrix \(\sigma\) which is unbounded with respect to the control. The strategy consists in noticing that one can express the definition of viscosity solutions of (1) without writing the ``sup'' in (2). By this way, one can handle functions \(u\) which are not in the domain of \(H\). The authors then obtain comparison, existence and uniqueness for viscosity solutions in the class of functions growing at most like \(o(1+|x|^p)\) for \(p>1\). The second issue of this paper is to extend some old results for \(p\)-growth type conditions on the data and the solutions and for more general equations with an additional nonlinearity \(f\) which is convex with respect to the gradient and depends on \(u\). Let us mention that the convexity of the operator with respect to the gradient is crucial in proofs.