1: \begin{proof}[Comments] 

2: A look at a simplified problem is instructive. Consider a positive measure $\sigma$ with 

3: $$ \int_{\R^2} \sigma = N,$$

4: the particle number, minimizing a continuous/mean-field version of the energy~\eqref{eq:class hamil 2}:

5: \begin{multline}\label{eq:class MF}

6:  \frac{\pi}{2}  \int_{\R^2} |x| ^2 \sigma (x) dx - \frac{1}{2} \iint_{\R^2 \times \R^2} \sigma (x) \log |x-y| \sigma (y) dx dy \\

7:  + N^{-N}\int_{\R^{2N}} \cW (x_1,\ldots,x_N) \sigma(x_1) \ldots \sigma(x_N) dx_1 \ldots dx_N. 

8: \end{multline}

9: The Euler-Lagrange for this problem says that, on the support of $\sigma$, 

10: $$ \frac{\pi}{2} |x| ^2 - \log |\,.\,| \star \sigma (x)+ N^{1-N} \int_{\R^{2(N-1)}} \cW (x,x_2,\ldots,x_N) \sigma(x_2) \ldots \sigma(x_N) dx_2 \ldots dx_N = \mu,$$ 

11: a constant (Lagrange multiplier associated with the mass constraint). Taking the Laplacian of the above equation and using~\eqref{eq:superharm} immediately gives 

12: $$ \sigma \leq 1.$$

13: The issue is that, for a general genuine many-body potential $\cW$, it does not seem feasible to reduce the minimization of~\eqref{eq:class hamil 2} for point configurations to that of~\eqref{eq:class MF} for measures. For particular $\cW$ containing only few-particle interactions one can pass rigorously~\cite{RouYng-14} from~\eqref{eq:class hamil 2} to~\eqref{eq:class MF}, which gives a particular, weaker, case of the Theorem and its applications. The proof of the general case, sketched below, does not use the continuum/mean-field approximation.

14: \end{proof}

15: