1: \begin{proof}

2: We have 

3: $$\underline\mu_{1}=\inf_{(x,\xi)\in\R^2}\mu_{1}(x,\xi)\leq \inf_{x\in\R}\mu_{1}(x,0).$$

4: We use a finite element method, with the Finite Element Library {\sc M\'elina} (see \cite{Melina}), on $[-10,10]$ with Dirichlet condition on the artificial boundary, with 1000 elements $\mathbb P_{2}$. The discretization space for the finite element method is included in the form domain of the operator and thus the computed eigenvalue provides a rigorous upper-bound (see \cite[Section 2]{BDMV07} and \cite[Section 5.1]{BDPR11}).

5: For any $x$, these computations give a upper-bound of $\mu_{1}(x,0)$.

6: Numerical computations and Proposition~\ref{conjecture-mont} give

7: $$\inf_{x\in\R}\mu_{1}(x,0)\leq 0.33227<0.5<\nu_{\Mont},$$

8: In fact, numerical simulations suggest that $\inf_{x\in\R}\mu_{1}(x,0)\simeq 0.33227$ which is an approximation of the first eigenvalue for $x=0.827$.

9: \end{proof}

10: