2
$\begingroup$

I'm looking for a reference for the following. Take the obstacle problem: $$\int_\Omega \nabla u \nabla (v-u) \leq \int_\Omega f(v-u)$$ for a function $u \in K:=\{v \in H^1_0(\Omega) : v \geq \varphi\}$ for a sufficiently nice given function $\varphi$, and holding for all test functions $v \in K$. Say $\Omega$ is a bounded smooth domain.

Denote $u=\sigma(f)$ as the solution map.

I read, some time ago, that depending on the nature of the coincidence set $\{u = \varphi\}$ and the obstacle $\varphi$, the solution map $\sigma$ can be Frechet or directionally differentiable or not differentiable at all (but I may be misremembering some aspects of this). Unfortunately I can't find a reference to these issues. I would appreciate any help. Thanks.

$\endgroup$
3
  • $\begingroup$ I'd suggest starting with the seminal work of Mignot here where it is proven that the solution mapping is only directionally differentiable (provided you're capable of reading mathematics in French). If you are looking for a bigger picture, it might be worthwile to check out more modern papers which cite this one. $\endgroup$
    – Hannes
    Commented Jan 4, 2017 at 13:13
  • $\begingroup$ @Hannes I just checked the citations.. unfortunately nothing good. I'm familiar with the paper, but I think we can expect a bit more on the differentiability given sufficient conditions on the data. Incidentally, do you know any references for differentiability of the obstacle to solution mapping? $\endgroup$
    – M.L
    Commented Jan 4, 2017 at 15:59
  • $\begingroup$ Maybe the introduction here helps some more? But I'm not really an expert in this field myself, sorry - I just happened to look through Mignot's paper a few days ago. $\endgroup$
    – Hannes
    Commented Jan 4, 2017 at 17:22

0

You must log in to answer this question.