Thm 4.9 in Gilbarg-Trudinger's book states that :

  1. if $B$ is a ball in $\mathbb{R}^n$ centred at $x_0$ and
  2. $f\in C^{\alpha}(B): \sup_{x\in B} (\text{dist}(x, \partial B))^{2-\beta}\vert f(x)\vert \leq N<\infty$ for some $\beta\in (0, 1)$.

Then there is a unique function $u\in C^{2}(B)\cap C(\overline{B})$, satisfying $\Delta u=f$ in $B$ and $u=0$ on $\partial B$. Furthermore $u$ satisfies an estimate: \begin{equation} \sup_{x\in B} \ \text{dist}(x, \partial B)^{-\beta}\vert u(x)\vert\leq CN, \end{equation} where $C=C(\beta)$.

I am following the proof of this theorem and I am happy with everything upto the point they start to prove the existence of $u$.

To show the existence of $u$ they first let \begin{equation} f_m=\left\{\begin{aligned} m &\quad\text{if } f\geq m\\ f &\quad\text{if }\vert f\vert\leq m\\ -m &\quad\text{if } f\leq -m\end{aligned}\right. \end{equation}

and let $\{B_k\}_{k=n}^{\infty}$, $n=\lceil \inf_{x\in B} \vert f\vert\rceil$, be a collection of concentric balls about $x_0$ such that: \begin{equation} B_k\subset\subset B_{k+1}, \quad x\in B_{k}\Rightarrow \vert f(x)\vert\leq k\quad\text{and} \quad \bigcup_k B_{k}=B. \end{equation}

With this setup they define the set $\{u_m\}$ to be the set of functions that satisfy: \begin{align*} \Delta u_m &=f_m\quad\text{in } B\\ u_m &=0\quad\text{on } \partial B. \end{align*}

By the estimate provided above (this was proved first in their argument) we know that $u_m$ is uniformly bounded and that for $m\geq k$ \begin{equation} \Delta u_m=f\quad\text{in } B_k. \end{equation}

Now they say that if we apply Corollary 4.7 successively to the balls $B_k$ then we get a subsequence that converges to a function $u\in C^{2}(B)$ that satisfies $\Delta u=f$ in $B$.

Corollary 4.7 says that if $u_n$ is a uniformly bounded sequence of solution to Poisson's equation, $\Delta u=f$ in a domain $\Omega$ with $f\in C^{\alpha}(\Omega)$ then for any $\Omega'\subset\subset \Omega$ there is a subsequence that converges uniformly to a solution.

I tried to apply this Corollary 4.7 in this way, however, I find myself taking subsequences infinitely often as follows:

If $B_k\subset\subset B_{k+1}$ then: \begin{equation} \Delta u_m=f\quad\text{in } B_{k+1}\ \forall m\geq k+1\Rightarrow \exists\ u_{m_j}\rightrightarrows u: \Delta u=f \ \text{in } B_{k}. \end{equation}

So certainly this subsequence converges to a $u\in C^{2}(B_k)$ that satisfies Poisson's equation in $B_{k}$ and if I repeat the process for $B_{k+1}\subset\subset B_{k+2}$ then I take a subsequence of $u_{m_j}$ to get the result for $B_{k+1}$ and so on.

I don't understand how this process will stop such that I find one subsequence that works throughout $B$ as claimed by [G-T]. If we are to use Corollary 4.7 in this way then shouldn't we find ourselves stopping this procedure once $B$ becomes compactly contained in some bigger set in which Poisson's equation is satisfied by $u_m$? This is clearly not the case.


We have that for $n\leq m$, $\Delta u_m=f$ in $B_n$. Hence, you can extract a subsequence $u_{1,j}$ of $u_m$, such that $u_{1,j}$ converges to some $u_1$ in $B_n$.

Now, you hve that for big $j$, $\Delta u_{1,j}=f$ in $B_{n+1}$. Hence, you can extract a subsequence $u_{2,j}$ of $u_{1,j}$, such that $u_{2,j}$ converges to some $u_2$ in $B_{n+1}$.


Now, you hve that for big $j$, $\Delta u_{k,j}=f$ in $B_{n+k}$. Hence, you can extract a subsequence $u_{k+1,j}$ of $u_{k,j}$, such that $u_{k+1,j}$ converges to some $u_{k+1}$ in $B_{n+k}$.

Now try to use a "Cantor Diagonal" argument.


Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.