One-dimensional Projective Structures, Convex Curves and the Ovals of Benguria and Loss

Appendix A On extending the conjecture of Benguria and Loss

Benguria and Loss’s conjecture concerns closed curves. In light of the present paper, specifically the symmetry of Theorem 1.2, it is tempting to think that their conjecture can be extended to degree-one curves with more than two balance points. However, this is not the case.

Lemma A.1

For every \(N\in \mathbb{N}\), there is a \(\sigma\in \mathcal{D}^\infty_+\) so that \(\phi_\sigma\in \mathrm{ BDiff}_+^\infty(\mathbb{S}^1,N)\) and \(\mathcal{E}_S[\sigma,f]<0\) for some function \(f\in C^\infty(\mathbb{S}^1)\).

Proof. Consider \(\sigma_\tau\) to be the curve in \(\mathcal{D}_+^{2,1}\) which has \(\sigma_\tau(\mathbf{e}_1)=\mathbf{e}_1\), \(L(\sigma_\tau)=2\pi\) and induced diffeomorphism \(\phi_{\sigma_\tau}=\psi_\tau\) where \(\psi_\tau\) is given by (4.2). Note, that for \(\tau\neq 0\), \(\sigma_\tau\) is not closed. One computes that for \(f_\tau=\kappa_{\sigma_\tau}^{-1/2}\in C^{0,1}(\mathbb{S}^1)\subset H^1(\mathbb{S}^1)\), that \(\mathcal{E}_S[\sigma_\tau,f_\tau]=\mathcal{E}_G[\sigma_\tau]=0\). However, \[L_{\sigma_\tau}f_\tau=-f_\tau'' +\kappa_{\sigma_\tau}^2 f_\tau = f_\tau +C(\tau)\delta_{\mathbf{e}_2}-C(\tau)\delta_{-\mathbf{e}_2},\] distributionally and the constant \(C(\tau)\neq 0\) if and only if \(\tau\neq 0\). Hence, for \(\tau\neq 0\), \(f_\tau\) is not an eigenfunction and so there must be a \(\hat{f}_\tau\in C^2(\mathbb{S}^1)\) with \(\mathcal{E}_S[\sigma_\tau,\hat{f}_\tau]<0\). Consider the elements \(\psi_{\tau}^\lambda\in \mathrm{ Diff}_+^{1,1}(\mathbb{S}^1)\) given by (4.3) and pick \(\sigma_{\tau}^\lambda\in \mathcal{D}_+^{2,1}\) so that \(\sigma_\tau^\lambda(\mathbf{e}_1)=\mathbf{e}_1\), \(L(\sigma_\tau^\lambda)=2\pi\) and the induced diffeomorphism is \(\psi_{\tau}^\lambda\). Clearly, \(\sigma_{\tau}^\lambda\to \sigma_\tau\) as \(\lambda\to 1\) in the \(C^2\) topology. Hence, \(\mathcal{E}_S[\sigma_{\tau}^{\lambda},\hat{f}_\tau]\to \mathcal{E}_S[\sigma_{\tau},\hat{f}_\tau]\) as \(\lambda\to 1\). Hence, for \(\tau\neq 0\) and \(\lambda>1\) sufficiently close to \(1\), we obtain a \(\sigma\in \mathcal{D}_+^\infty\) with \(n_B(\sigma)=\infty\) and \(\mathcal{E}_S[\sigma, \hat{f}_\tau]<0\) by smoothing out \(\sigma_{\tau}^\lambda\) as in Lemma 4.7. Smoothing out \(\hat{f}_\tau\) gives \(f\) so that \(\mathcal{E}_S[\sigma,f]<0\).