Let $$P:=\sum_{j=1}^{n}\left(\frac{\partial}{\partial x_{j}}\right)^{m}$$

Then I want to show that $P$ is an elliptic operator on $\mathbb{R}^{n}$ when $m$ is even, but not if $m$ is odd.

Definition: $P$ is elliptic if the top order symbol has no real zeros except $\xi=0$, or equivalently if the full symbol satisfies $|p(\xi)|\ge c|\xi|^{m}$ for $|\xi|\ge A$ for $c,A$ positive constants.

We see that

$$(Pu)^{\widehat{}}(\xi)=\left(-i\sum_{j=1}^{n}\xi_{j}\right)^{m}\hat{u}(\xi),$$ so that $$p(\xi)=\left(-i\sum_{j=1}^{n}\xi_{j}\right)^{m}=\begin{cases}\sum_{j=1}^{n}\xi^{m}_{j}&\mbox{if }m\text{ is even},\\ -i\sum_{j=1}^{n}\xi_{j}^{m}&\mbox{if }m\text{ is odd}.\end{cases}$$

However, $|p(\xi)|$ is the same for $m$ even and odd, and we get that

$$|p(\xi)|=\sum_{j=1}^{n}|\xi_{j}|^{m}\ge |\xi|^{m}$$

which means that $P$ is elliptic; but this isn't what I wanted. Where did I go wrong?


Indeed, the factor of $-i$ makes no difference: it's all about $\sum \xi_j^m$ being zero or not. Your proof goes wrong when you "distribute" the absolute value over a sum.

When $m$ is even, each $\xi_j^m$ is nonnegative, so the only way for their sum to be $0$ is if every term is $0$.

When $m$ is odd, the above no longer applies, and counterexamples such as $\xi=(1,-1,0,\dots,0)$ present themselves.


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.