Let $\Omega \subset \mathbb{R}^{N}$ be an open set with boundary of class $C^{\infty}$ and let $\{\lambda_{k}\}$ and $\{v_{k}\}$ be the eigenvalues and eigenvectors of -$\Delta$ with Dirichlet condition in $\Omega$, i.e., \begin{equation*} -\Delta v_{k} = \lambda_{k} v_k \end{equation*} in which $\Delta$ is the Laplacian Operator. We assume that $\{v_{k}\}$ are normalized in $L^{2}(\Omega)$, so that $\{v_{k}\}$ is a Hilbert Basis for $C^{\infty}(\Omega)$, i.e., \begin{equation*} u = \sum_{k=1}^{\infty} \langle u,v_{k} \rangle v_{k}(x) \ \ \ \ \text{ (in } L^{2} \text{)} \end{equation*} for all $u \in C^{\infty}(\Omega)$ and \begin{equation*} \langle v_{k},v_{j} \rangle = \begin{cases} 0, & \text{ if } k \neq j \\ 1, & \text{ if } k = j \end{cases} \end{equation*} in which $\langle u,v \rangle = \int_{\Omega} u(x) v(x) \ dx$ is the inner product of $L^{2}(\Omega)$ (we suppose that $u$ and $v_{k}$ are real-valued functions).

Now, let $u \in C^{1}(\bar{\Omega})$, $u(x) = 0$ for $x \in \partial \Omega$. Show that there exists a constant $C > 0$ such that \begin{equation*} \Bigg\lvert \int_{\Omega} u(x) v_{k}(x) \ dx \Bigg\rvert \leq \frac{C}{\lambda_{k}^{1/2}} \end{equation*}

From First Greens's Identity, for example, we get that \begin{equation*} \lambda_{k} \Bigg\lvert \int_{\Omega} u(x) v_{k}(x) \ dx \Bigg\rvert \leq \Bigg\lvert \int_{\Omega} \langle\nabla u(x),\nabla v_{k}(x)\rangle \ dx \Bigg\rvert \end{equation*} but I cannot figure out how to proceed from here.

EDIT: The result follows by applying the Cauchy-Schwarz inequality \begin{equation*} \lambda_{k} \Bigg\lvert \int_{\Omega} u(x) v_{k}(x) \ dx \Bigg\rvert \leq \int_{\Omega} \lvert\langle\nabla u(x),\nabla v_{k}(x)\rangle \rvert\ dx \leq \int_{\Omega} |\nabla u(x)||\nabla v_{k}(x)| dx \leq \lVert (|\nabla u(x)|) \rVert_{L^{2}} \lVert(|\nabla v_{k}(x)|) \rVert_{L^{2}} \end{equation*} and First Green's identity \begin{equation*} \int_{\Omega} v_{k}(x) \Delta v_{k}(x) dx = -\int_{\Omega} |\nabla v_{k}(x)|^{2} dx \implies \lVert(|\nabla v_{k}(x)|) \rVert_{L^{2}} = \sqrt{\lambda_{k}} \end{equation*} so the result follows taking $C = \lVert (|\nabla u(x)|) \rVert_{L^{2}} < \infty$. Notice the difference between the norm $| \cdot |$ of vectors in $\mathbb{R}^{N}$ and the norm $\lVert \cdot \rVert_{L^{2}}$ of $L^{2}$.


From the inequality you have obtained: $$\begin{equation*} \lambda_{k} \Bigg\lvert \int_{\Omega} u(x) v_{k}(x) \ dx \Bigg\rvert \leq \Bigg\lvert \int_{\Omega} \nabla u(x) \cdot \nabla v_{k}(x) \ dx \Bigg\rvert \end{equation*}$$ you can first use Cauchy-Swharz: $$\lambda_{k} \Bigg\lvert \int_{\Omega} u(x) v_{k}(x) \ dx \Bigg\rvert \leq \|\nabla u\|_{L^2} \|\nabla v_k \|_{L^2}$$ then you can use the crucial property:

$$\|\nabla v_k\|_{L^2}= \sqrt{\lambda_k}$$

as by Green: $$\|\nabla v_k\|_{L^2}^2=\int_\Omega \nabla v_k \cdot \nabla v_kdx = -\int_\Omega v_k \nabla\cdot(\nabla v_k)dx= -\int_\Omega v_k \Delta v_kdx=\lambda_k \int_\Omega v_k v_k dx =\lambda_k \|v_k\|_{L^2}^2=\lambda_k$$

So finally: $$\lambda_{k} \Bigg\lvert \int_{\Omega} u(x) v_{k}(x) \ dx \Bigg\rvert \leq \sqrt{\lambda_k} \|\nabla u\|_{L^2}$$ which is the expected result (as $u \in C^1$, the quantity $\|\nabla u\|_{L^2}$ is finite).

  • $\begingroup$ Thanks. I had applied the Cauchy-Swharz ineuqality but have not been able to bound the norm of the gradient of $v_{k}$. $\endgroup$ – marcstyle Oct 27 '18 at 18:50
  • $\begingroup$ @marcstyle to bound the norm of the gradient of $v_k$ the idea is to notice that by integration by parts (or Green): $$\int \nabla v_k \nabla v_k = -\int \nabla \cdot (\nabla v_k) v_k=-\int \Delta v_k v_k$$ and $-\delta v_k =\lambda_k v_k$. $\endgroup$ – Delta-u Oct 28 '18 at 12:47
  • $\begingroup$ Indeed, but it just seem that it is missing a vector norm. In your answer I believe we have to show that $\lVert (|\nabla v_{k}|) \rVert_{L^{2}} = \sqrt{\lambda_{k}}$, i.e., the $L^{2}$ norm of the gradient vector norm equals the square root of the eigenvalue. In fact, $\lVert \nabla v_{k} \rVert_{L^{2}}$ does not make sense as $\nabla v_{k}$ is a vector in $\mathbb{R}^{N}$ and not a function in $L^{2}$. I have incorporated your answer to my question with this correction. $\endgroup$ – marcstyle Oct 30 '18 at 19:24
  • $\begingroup$ By the $L^2$ norm of the vector field I mean:$$\left(\int_\Omega \nabla v_k \cdot \nabla v_k \right)^\frac{1}{2}$$ where $\cdot$ is the inner product of $\mathbb{R}^N$. If you need additional explanations I can edit my answer :-). $\endgroup$ – Delta-u Oct 30 '18 at 20:32
  • $\begingroup$ Yeah, I figured, thanks. $\endgroup$ – marcstyle Oct 30 '18 at 23:00

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.