# Conditions for the existence of a solution to a semilinear second-order PDE with a-priori bounds

Consider the general semilinear elliptic second-order PDE $$u_t-\mathcal L u=f\left(t,x,u,\nabla u\right)$$ where $$\mathcal L$$ is an elliptic linear operator (like minus the Laplace operator), $$t \in [0,T],$$ $$x$$ is in a bounded smooth domain $$\Omega,$$ and the boundary conditions are $$u(T,x)=c, \quad \forall x\in\Omega$$ where $$c$$ is a constant. $$f$$ can be nonlinear in $$u$$ and $$\nabla u$$ with at most quadratic growth in both of them, and has no singularity in both of them and $$x$$. For example, $$f\left(t,x,u,\nabla u\right) = x^\alpha u^2 + ∇u \cdot ∇u$$ is a possible form. Mind that this is a Cauchy problem with final data, a situation which fits with the "backward" heat operator $$\partial_t+\Delta$$.

Suppose we know from the original problem that the solution to this PDE is bounded by two smooth $$\mathcal C^2(\Omega)$$ functions: $$\underline u \leq u \leq \overline u, \quad \forall (t,x)\in[0,T]\times \Omega.$$ My question is: is there a result (with possible conditions on $$f$$) to provide an existence result for a (weak) solution to this PDE ? any reference would be greatly appreciated.

• Please check your "boundary condition" on $\nabla u$. Apr 12 at 15:45
• Hi Giorgio, thanks for the comment, I edited the question.
– FDR
Apr 12 at 15:47
• Do you really want that the full gradient vanishes on the boundary? Apr 12 at 15:58
• Have a look at Section 4, Chapter 7 of the book of Friedmann. Theorem 10 deals with small time existence. Concerning global existence when there is no blow up in the sup norm, the basic idea is explained before theorem 8. From boundedness in the sup norm, one deduces boundeness in the Holder norm to apply a fix point theorem. It is far from being immediate. Apr 12 at 20:51
• Maybe you could have a look also at Krylov: lectures on elliptic and parabolic equations in Holder spaces. See exercise 4.3.9 in the elliptic case. See also 8.9.5 where however the growth is subquadratic Apr 12 at 21:44

Basically, the (classical) results there shows that for any bounded domain $$\Omega$$ for your space variables, if you have lower and upper solutions for a quasi-linear / semi-linear PDE, and that your nonlinear operator in the PDE has at most quadratic growth in the gradient, then you can have (in this order) bounds on the solution and on the gradient in your bounded domain. I recommend reading it for simple scalar PDEs.