Problem (Laplacian Invariance by Rotations). Let $U \subset \mathbb{R}^{m}$ be an open set. For any function $f: U \to \mathbb{R}$ twice differentiable, the Laplacian of $f$ is the function $\triangle f: U \to \mathbb{R}$, defined by $$\displaystyle \triangle f = \frac{\partial^{2} f}{\partial x^{2}_{1}} + \cdots + \frac{\partial^{2} f}{\partial x^{2}_{m}}.$$ Prove that if $T: \mathbb{R}^{m} \to \mathbb{R}^{n}$ is an orthogonal linear transformation, then $$\triangle (f \circ T) = (\triangle f)\circ T: V \to \mathbb{R},$$ where $V = T^{-1}(U)$.

I tried to solve this using the Hessian matrice and writing the partial derivatives as inner product, but I didn't get something promising. I appreciate any hint!


Let $e_1, \ldots, e_m$ be the standard $\mathbb R^m$ basis. Then we have $$Te_j = \sum_{i}a_{ij}e_i$$ where $a_{ij} = \langle Te_j , e_i \rangle$. Let $T_i(x)$ be the $i$-th coordinate of $T(x)$.

So we have $$T_i(x) = \sum_{j} a_{i,j}x_j.$$

Lets denote $\partial / \partial x_i$ by $\partial_i$. We have $$\partial_i [f(T(x))] = \sum_j \partial_i T_j(x)\partial_j(f)(T(x)) $$ and since $ \partial_i T_j(x) = a_{ij}$ we obtain $$\partial_i [f(T(x))] = \sum_j a_{ij}\partial_j(f)(T(x)).$$

Let's now compute the second derivative: $$\partial_i^2 [f(T(x))] = \sum_j a_{ij}\partial_i[\partial_j(f)(T(x))].$$

By the above calculation, putting $\partial_j(f)(T(x))$ in the place of $f(T(x))$, we obtain $$\partial_i[\partial_j(f)(T(x))] = \sum_k a_{ik}\partial_k\partial_j(f)(T(x))$$ and therefore $$\partial_i^2 [f(T(x))] = \sum_{j,k} a_{ij}a_{ik}\partial_k\partial_j(f)(T(x)).$$

Since $T$ is ortonormal we have $$\sum_{i}a_{ij}a_{ik} =\langle Te_j,Te_k\rangle = \langle e_j, e_k \rangle = \delta_{jk}.$$

Therefore \begin{align*} \Delta[f\circ T(x)] &= \sum_{i}\partial_i^2 [f(T(x))] =\sum_i\sum_{j,k}a_{ij}a_{ik}\partial_k\partial_j(f)(T(x)) \\&= \sum_{j,k} \partial_k\partial_j(f)(T(x)) \sum_i a_{ij}a_{ik} = \sum_{j,k} \partial_k\partial_j(f)(T(x)) \delta_{jk} \\&= \sum_k \partial_k^2(f)(T(x)) = \Delta(f) \circ T(x). \end{align*}

And that ends the proof.

  • $\begingroup$ A more elegant proof can be done using the Fourier transform and that it commutes with ortogonal transformations. $\endgroup$ – Hugocito Jun 10 '18 at 0:51
  • 1
    $\begingroup$ Nice proof! Thank you for the support! $\endgroup$ – Lucas Corrêa Jun 10 '18 at 4:07

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.