Find all harmonic functions $u$ defined in $\mathbb R^d$ ($d\geq 2$) s.t. $$\int_{\mathbb R^d}|D^2 u|^2<\infty .$$

I recall that $$|D^2 u|^2=\sum_{i=1}^d\sum_{j=1}^d \left(\frac{\partial^2 u}{\partial x_i\partial x_j}\right)^2.$$

My attempts

My idea was to set $h_i=\frac{\partial u}{\partial x_i}$ and remark that $$|D^2 u|^2=\sum_{i=1}^d|\nabla h_i|^2=\sum_{i=1}^d\nabla h_i\cdot \nabla h_i=\sum_{i=1}^d \text{div}(h_i\nabla h_i)-\sum_{i=1}^n (h_i \Delta h_i).$$ Since $u$ is harmonic, then $u\in \mathcal C^\infty $ and thus $\Delta h_i$ is well defined, but I can't conclude. I wanted to use divergence theorem, but since $\mathbb R^d$ is not bounded, it's not possible.

A second idea was to sue the fact that $u$ is harmonic, and thus $$|D^2u|^2=2\sum_{i=1}^{n-2}\sum_{j=1}^{n-1}\left(\frac{\partial ^2 u}{\partial x_i\partial x_j}\right)^2,$$ and get $$\int_{\mathbb R^d}|D^2 u|^2<\infty \iff 2\sum_{i=1}^{n-2}\sum_{j=i+1}^{n-1}\int_{\mathbb R^d}\left(\frac{\partial ^2u}{\partial x_i\partial x_j}\right),$$ but still, I can't continue.


Suppose that $u$ is a harmonic function on $\mathbb{R}^n$. Then by the mean-value property we know that for each $x \in \mathbb{R}^n$ and $r>0$ we have that $$ u(x) = \frac{1}{\omega_n r^{n}} \int_{ B(x,r)} u(y) dy $$ where $\omega_n$ is the volume of the unit ball in $\mathbb{R}^n$. From this and Cauchy-Schwarz we get the estimate $$ |u(x)| \le \frac{1}{\omega_n r^n} \int_{B(x,r)} |u(y)| dy \le \frac{1}{\omega_n r^n} \left( \int_{B(x,r)} |u(y)|^2 dy \right)^{1/2} |B(x,r)|^{1/2} \\ =\frac{1}{\sqrt{\omega_n} r^{n/2}} \left( \int_{B(x,r)} |u(y)|^2 dy \right)^{1/2} $$ Now, if $u$ is such that $$ \int_{\mathbb{R}^n} | u(y)|^2 dy = M^2 < \infty, $$ then we in turn find that $$ |u(x)| \le \frac{1}{\sqrt{\omega_n} r^{n/2}} \left( \int_{B(x,r)} |u(y)|^2 dy \right)^{1/2} \le \frac{1}{\sqrt{\omega_n} r^{n/2}} \left( \int_{\mathbb{R}^n} |u(y)|^2 dy \right)^{1/2} = \frac{M}{\sqrt{\omega_n} r^{n/2}}. $$ Then for any fixed $x \in \mathbb{R}^n$ we can send $r \to \infty$ to get that $$ |u(x)| \le \lim_{r \to \infty}\frac{M}{\sqrt{\omega_n} r^{n/2}} =0, $$ and hence we find that $u(x) =0$ for all $x$. Thus the only harmonic function that is square-integrable over all of $\mathbb{R}^n$ is $0$.

Having established this, we can now answer your question. Suppose now that $u$ is harmonic on $\mathbb{R}^n$ and $D^2 u$ is square-integrable. Since $u$ is harmonic, it's smooth, and so we can apply arbitrarily many derivatives to the PDE $\Delta u =0$. In particular we find that if $|\alpha|=2$ then $$ 0 = \partial^\alpha \Delta u = \Delta \partial^\alpha u, $$ which means that all of the second-order partial derivatives are also harmonic. Since $$ \int_{\mathbb{R}^n} |\partial^\alpha u |^2 \le \int_{\mathbb{R}^n} |D^2 u |^2 < \infty $$ we find that $\partial^\alpha u$ is a harmonic function that is square-integrable. By the above analysis we then know that $\partial^\alpha u =0$ for all multi-indices with $|\alpha |=2$. In other words, we know that $D^2 u =0$, and so basic calculus tells us that $u$ is linear, i.e. $u(x) = a + b \cdot x$ for some $a \in \mathbb{R}$ and $b\in \mathbb{R}^n$. It turns out that all such functions are trivially harmonic.

Thus a harmonic functions has $D^2 u$ square integrable if and only if it's linear.

  • $\begingroup$ Very nice answer, I will stud it and be sure I understand every thing before validate it, but don't worry, I will (or ask you question if something is not clear). $\endgroup$ – MathBeginner Jan 7 '17 at 17:15
  • $\begingroup$ Note: I just fixed a typo in the mean-value property (the first eqn). I accidentally wrote a derivative on the right side. Nothing else changes, though. $\endgroup$ – Glitch Jan 7 '17 at 17:22
  • $\begingroup$ Just fixed another typo in the same equation... the integral was supposed to be over the full ball, not the boundary. $\endgroup$ – Glitch Jan 7 '17 at 17:27

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.