Exponential decay of the solutions of quasilinear second-order equations and Pohozaev identities (Q1587635): Difference between revisions

The paper is devoted to a rather detailed study of the pointwise exponential decay of the solutions of quasilinear second-order equations, \[ -\sum\limits _{\alpha ,\beta =1}^{N}a_{\alpha \beta }(x,u,\nabla u)\partial_{\alpha }\partial_{\beta }u+b(x,u,\nabla u)=f \] on \(\mathbb{R}^N\) under general assumptions about the coefficients \(a_{\alpha \beta }\) and \(b\) and provided that \(f\) exhibits appropriate exponential decay at infinity. The authors study the exponential decay of the solutions of this equation independently of their existence and they also examine the behaviour of the first and second derivatives of the solutions. The main result is that every solution in \(C^1( \mathbb{R}^N)\cap W_{\text{loc}}^{2,N}( \mathbb{R}^N)\) tending uniformly to zero at infinity along with its gradient must tend to zero exponentially. This is even true for \(W_{\text{loc}}^{2,N}\) solutions tending uniformly to zero at infinity, with no assumption made about their gradients, when the coefficients \(a_{\alpha\beta }\) are independent of \(\nabla u\) and when \(b\) depends linearly upon \(\nabla u\). It is shown that the \(W^{2,p}\) solutions of the equation under consideration are eventually of class \(C^2\) and in \(W^{2,q}\) for every \(1\leq q\leq \infty \), i.e., this is true in the complement of some closed ball depending upon the solution. The exponential decay of the derivatives is more delicate to extend to general unbounded domains and, in fact, this extension requires introducing some geometric limitations.