A question involving the proof that the instanton function minimize the Sobolev constant

35 Views Asked by At

I'm studying the Willem's book Minimax Theorems. And there is the following result (My question is on below):

Theorem 1.42(Aubin, Talenti, 1976) The instanton $$U(x):=\dfrac{[N(N-2)]^{(N-2)/4}}{[1+|x|^2]^{(N-2)/2}}$$

is a minimizer for $S:=\inf_{u\in D^{1,2}(R^N),|u|^{2^*}=1}|\nabla u|_2^2>0$.

Proof. 1) By the preceding theorem, there exists a minimizer $u\in D^{1,2}(R^N)$ for $S$. By theorem C.4, $u$ is radially symmetric. Replacing $u$ by $|u|$, we may assume that $u$ is non-negative.

  1. It follows from Lagrange multiplier rule that, for some $\lambda>0$, $u$ is a solution of $$-\Delta u=\lambda u^{\frac{N+2}{N-2}}$$.

By the argument of Lemma 1.30, $u\in C^2(R^N).$ The strong maximum principle implies that $u$ is positive.

  1. After scaling, we may assume $$-\Delta u= u^{\frac{N+2}{N-2}}$$

Moreover we can choose $\varepsilon>0$ such that $$U_\varepsilon(x):=\varepsilon^{(2-N)/2}U(x/\varepsilon)$$

satisfies $$U_{\varepsilon}(0)=u(0)$$.

But then $u$ and $U_{\varepsilon}$ are solutions of the problem $$\partial_r(r^{N-1}\partial_r v)=r^{N-1}v^{\frac{N+2}{N-2}}, r>0$$ $$v(0)=u(0), \partial_rv(0)=0$$.

It follows easily that $u=U_{\varepsilon}$. By invariance, $U$ is a minimizer for $S$.

So I'm stuck in proving that the function $u$ in fact satisfies the ODE $\partial_r(r^{N-1}\partial_r v)=r^{N-1}v^{\frac{N+2}{N-2}}$. Maybe I even do not know what that means. I think that since $u$ is radially symmetric then we can see $u$ as a function of $r$ where $r$ is the lenght of a vector ($\|x\|=r$). But how to differentiate such a function, using the hypotesis on $u$ and in fact show that such ODE is satisfied? How to interpret such ODE? Thanks in advance.