1: \begin{proof}

2: Since $\lambda$ is an eigenvalue of $D$, we have $D|\lambda\>=\lambda|\lambda\>$. As a result, we have\footnote{This proof is not fully rigorous as $\delta(0)$ is ill-defined. However, $\delta$ can be regarded as the limit (in the sense of distributions) of the sequence of zero-centered normal distributions $\delta_{\sigma}(x)=\frac{1}{|\sigma|\sqrt{\pi}}e^{-(x/\sigma)^{2}}$ as $\sigma\to 0$. Then the LHS of \eqn{lem-spectrum-D-1} is replaced by $|\lambda|\cdot\frac{1}{|\sigma|\sqrt{\pi}}$ and the RHS of \eqn{lem-spectrum-D-2} is replaced by $\frac{1}{|\sigma|\sqrt{\pi}}$, so $|\lambda|\leq 1$, and this also holds in the limit $\sigma\to 0$. For convenience we use similar arguments (regarding $\delta(0)$ as a positive real number) in this section, but keep in mind that rigorous proofs can be given by limit arguments.}

3: \begin{align}

4: |\lambda|\delta(0)=|\lambda|\<\lambda|\lambda\>&=|\<\lambda|D|\lambda\>| \label{eqn:lem-spectrum-D-1} \\

5: &=\Big|\int_{\Omega}\int_{\Omega}\d x\,\d y\,\sqrt{p_{y\to x}p_{x\to y}}\<\lambda|x\>\<y|\lambda\>\Big| \\

6: \text{(by Cauchy-Schwarz)}\quad&\leq\sqrt{\Big(\int_{\Omega}\int_{\Omega}\d x\,\d y\,p_{y\to x}|\<y|\lambda\>|^{2}\Big)\Big(\int_{\Omega}\int_{\Omega}\d x\,\d y\,p_{x\to y}|\<\lambda|x\>|^{2}\Big)} \\

7: \text{(by $\int_{\Omega}\d y\,p_{x\to y}=1$)}\quad&\leq\sqrt{\Big(\int_{\Omega}\d y\,|\<y|\lambda\>|^{2}\Big)\Big(\int_{\Omega}\d x\,|\<\lambda|x\>|^{2}\Big)} \\

8: &=\int_{\Omega}\d x\,\<\lambda|x\>\<x|\lambda\> \\

9: \text{(by \eqn{identity})}\quad&=\<\lambda|\Big(\int_{\Omega}\d x\,|x\>\<x|\Big)|\lambda\> \\

10: &=\delta(0). \label{eqn:lem-spectrum-D-2}

11: \end{align}

12: Hence the result follows.

13: \end{proof}

14: