Quasilinear elliptic equations with quadratic growth in the gradient (Q1188227): Difference between revisions

The authors wish to establish existence of a solution \(u:\Omega\to\mathbb{R}\) \((\Omega\) open bounded subset of \(\mathbb{R}^ n)\) such that: \(u\in H^ 1_ 0(\Omega)\cap L^ \infty(\Omega)\) and satisfies \[ Au(x)=-\sum^ n_{i,j=1}\partial_ i(a_{ij}(x))\partial_ ju=H(x,u,Du) \tag{1} \] where \(A\) is elliptic, \(a_{ij}(x)\) are measureable functions satisfying a.e. in \(\Omega\) \(\|\xi\|^ 2\leq\sum a_{ij}(x)\xi_ i\xi_ j\leq\Lambda\|\xi\|^ 2\) for any \(\xi\in\mathbb{R}^ n\) and for some \(\Lambda\geq 1\), with the usual Caratheodory assumptions. They also require a quadratic behavior of \(H(x,u,\xi)\) with respect to \(\xi\). Thus the problem (1) is well defined in the weak sense. The author's quadratic assumption may be restated as follows: There exist nondecreasing functions \(K\), \(\tilde K\): \([0,+\infty)\to[0,+\infty)\) and measurable functions: \(f\), \(\tilde f\in L^ p(\Omega)\), \(p>n/2\) such that \[ \tilde f(x)-\tilde K(| z|)\|\xi\|^ 2\leq H(x,z,\xi)\leq f(x)+K(| z|)\|\xi\|^ 2. \] Previous papers contained assumptions implying boundedness of \(K\), \(\tilde K\) [see \textit{H. Amann} and \textit{M. G. Crandall}, Indiana Univ. Math. J. 27, 779-790 (1978; Zbl 0391.35030)]. In other papers [see e.g. \textit{L. Boccardo}, \textit{F. Murat} and \textit{J. P. Puel}, Ann. Sc. Norm. Super. Pisa; Cl. Sci, IV. Ser. 11, 213-235 (1984; Zbl 0557.35051)] the existence of sub- and supersolutions was assumed. The authors observe by a simple estimate that if one disregards the sign of \(f\) then it is not possible to establish the existence of solutions without making an additional assumption that the measure of \(\Omega\) is ``small'' compared to the norm of \(f\). Here the authors prove the existence of solutions without making assumptions of boundedness of \(K\), \(\tilde K\), existence of sub- and supersolutions, or the so-called ``one-sided'' conditions of Raskotoson. Their technique is best explained by considering the model equation \[ \Delta u+K(| u|)(Du)^ 2+f(x)=0. \] Let us denote \({\mathcal K}(t)=\int^ t_ 0K(| s|)ds\) and \(v=E(u)=\int^ u_ 0\exp({\mathcal K}/t))dt\), \(q=\exp({\mathcal K}E^{-1})\), \(W(t)=E(t)/E'(t)\). Then \(v\) satisfies: \(\Delta v+f(x)q(v)=0\). Comparing the solutions of the Dirichlet problem (1) with solutions of \(- \Delta V=f^*_ +\) in \(\Omega\), \(V=0\) on \(\partial\Omega\), they derive a new maximum principle for \(W(u)\). Creating a family of Dirichlet problems depending on a parameter they use the contraction mapping theorem to prove the existence of solutions.