Imagine we have a nonlinear PDE, e.g. some velocity model

$$\begin{aligned} -\Delta v + (v \cdot \nabla) v + \nabla p = f \\ \text{div} \, v = 0 \end{aligned}$$

which has a weak solution for some initial & boundary data if we are in the stationary case. In the existence proof via Galerkin we can use compact embeddings of the kind $H^1 \hookrightarrow \hookrightarrow L^2$ to get strong convergence and we can do the limit process in the nonlinear term.

Now, let us couple that PDE to some time-dependent PDE so that the right-hand side $f$ now depends on time, e.g. we couple it to a convective reaction-diffusion equation as follows $$\begin{aligned}-\Delta v + (v \cdot \nabla) v + \nabla p &= \nabla u \\ \text{div} \, v &= 0 \\ \partial_t u - \Delta u+ \nabla u \cdot v &= 0 \end{aligned}$$ Hence our velocity model got instationary and $v$ will depend on time.

In my opinion it is now impossible to show existence of a weak solution via Galerkin's method. We'd need strong convergence, but for this we'd need a bound on the (fractional) time derivative of $v$. Note that the embedding $L^2(0,T;H^1) \hookrightarrow L^2(0,T;L^2)$ is not compact. Instead we want to use something like $L^2(0,T;H^1) \cap H^1(0,T;(H^1)') \hookrightarrow \hookrightarrow L^2(0,T;L^2).$ This is not directly available. Of course, adding a term $\partial_t v$ to the velocity model would save the day but we don't want to do that. Maybe adding some $\varepsilon \partial_t$ and later letting $\varepsilon \to 0$ would help.

I can hardly imagine that there is theorem that if a weak solution exist then one can prove it with Galerkin's method. So maybe some fixed point theorem could help. But here in that case I can't believe it that there exists a weak solution. It seems so far away.

So, can we prove the non-existence of weak solution in such a case?

  • $\begingroup$ Is there some kind of continuous or bounded (mapping bounded sets into bounded sets) solution mapping for the stationary equation? $\endgroup$ – Hannes Nov 8 '18 at 15:36
  • $\begingroup$ What PDE do you want to couple with? Everything depends on that. The equation you wrote down doesn't touch $t$ at all, so it's not parabolic in $(x,t)$, and I see no reason to expect Galerkin to work. $\endgroup$ – user126920 Nov 8 '18 at 15:54
  • $\begingroup$ @StanleySnelson Let us say that $f(x,t)=\nabla u(x,t)$ where $u$ is the weak solution to the coupled PDE $\partial_t u - \Delta u + \nabla u \cdot v = 0$ or something. So the equation will get the $t$ through $f$, that's the point. $\endgroup$ – Fritz Nov 8 '18 at 16:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.