1: \begin{proof}

2: Set $q_\alpha:=\coeff{Q}{\alpha,z_0}$ for $\alpha=0,\ldots,N$. 

3: Since 

4: $$

5: \coeff{Q\mcS}{\alpha,z_0}=\sum_{n=0}^\alpha q_n\coeff{\mcS}{\alpha-n,z_0}

6: =\sum_{n=0}^N q_n\coeff{\mcS}{\alpha-n,z_0}

7: $$ 

8: according to our convention that $\coeff{\mcS}{\beta,z_0}=0$ for $\beta<0$, we have

9: \begin{align*}

10: 	\nonumber

11: \jfunbar{Q}^2= & \sum_{\alpha=M+1}^E

12: 	\dualw{ \coeff{Q\mcS}{\alpha,z_0} } { \coeff{Q\mcS}{\alpha,z_0} }

13: 	\rho^{2\alpha}\\

14: 	\nonumber

15: =& \sum_{\alpha=M+1}^E

16: 	\dualw{ \sum_{j=0}^{N} q_j \coeff{\mcS}{\alpha-j,z_0} }

17: 	{ \sum_{i=0}^{N} q_i \coeff{\mcS}{\alpha-i,z_0} }

18: 	\rho^{2\alpha}\\

19: 	\nonumber

20: =& \sum_{\alpha=M+1}^E \sum_{i,j=0}^{N}

21: 	q_i^* q_j \dualw{ \coeff{\mcS}{\alpha-j,z_0} }{ \coeff{\mcS}{\alpha-i,z_0} }

22: 	\rho^{2\alpha}\\

23: = & \sum_{i,j=0}^N q_i^* q_j \sum_{\alpha=M+1}^E 

24: 	\dualw{ \coeff{\mcS}{\alpha-j,z_0} }{ \coeff{\mcS}{\alpha-i,z_0} }

25: 	\rho^{2\alpha}\\

26: = & \qq^\star \G \qq,

27: \end{align*}

28: where $\G\in\C^{(N+1)\times(N+1)}$ is defined in~\eqref{eq:gram_matrix}, and $\qq=(q_0,\ldots,q_N)^T$. 

29: By definition, $\G$ is Hermitian. Moreover, definition~\eqref{eq:jbar} implies that $\G$ is positive-semidefinite, so that all its eigenvalues are real non-negative. 

30: Finally, observe that the constraint $\sum_{\alpha=0}^N\abs{\coeff{Q}{\alpha,z_0}}^2=1$ is equivalent to the condition $\norm{\qq}{2}=1$. Hence, we conclude that the constrained minimization problem~\eqref{eq:min_Q} is equivalent to the identification of the (normalized) eigenvector corresponding to the smallest eigenvalue $\G$.

31: Finally, we observe that equation~\eqref{eq:coeff_S} is obtained by repeated differentiation of equation~\eqref{eq:helmholtz_weak}; see~\cite{Bonizzoni2016} for a rigorous derivation.

32: \end{proof}

33: