On the convexity of the KdV Hamiltonian (Q739571)

This paper deals with the convexity properties of the nonlinear part of the KdV Hamiltonian expressed in action variables. It contains two results. The first result (Theorem 1, p. 193) proves a (generalization of) conjecture of \textit{E. L. Korotyaev} and \textit{S. Kuksin} [J. Dyn. Control Syst. 22, No. 4, 661--682 (2016; Zbl 1351.37243)] and states that the nonlinear part \(H^{\ast}\) \[ H^{\ast}(I)=H^{kdv}(I)-\sum_{n\geq1}8\pi^{3}n^{3}I_{n} \] of the KdV Hamiltonian \(H^{kdv}(u)=\int_0^1\left(\frac{1}{2}\left(\partial_{x}u\right)^2+u^3\right) dx\), when expressed in the action coordinates \(\left( I_{n}\right) _{n\geq1}\), admits a real analytic extension to a complex neighborhood \(V\subset\ell_{\mathbb{C}}^{2}\left(\mathbb{N}\right) \) of the positive quadrant \(\ell_{+}^{2}\left(\mathbb{N}\right) \) of \(\ell^{2}\left(\mathbb{N}\right) \) and that it is strictly concave at \(0\). The second result (Theorem 2, page 194) states that for any \(-\frac{1}{s}\leq s\leq0\) and and \(2\leq p<\infty\) the resctriction of the Birkhoff map \(\Phi(u)=\left( z_{n}\left( u\right) _{n\in\mathbb{Z}}\right) \) (for details see [the author et al., Sel. Math., New Ser. 11, No. 1, 37--98 (2005; Zbl 1089.37042)]) to the Fourier-Lebesgue space \(\mathcal{FL} _{0}^{s,p}\) takes values in \(\ell_{0}^{s+1/2,p}\) and that \(\Phi:\mathcal{FL} _{0}^{s,p}\rightarrow\ell_{0}^{s+1/2,p}\) is a real analytic diffeomorphism thus providing global Birkhoff coordinates on \(\mathcal{FL}_{0}^{s,p}\). An interesting consequence of Theorem 2 is Theorem 3 (page 194), which extends the result of [\textit{T. Kappeler} and \textit{P. Topalov}, Duke Math. J. 135, No. 2, 327--360 (2006; Zbl 1106.35081)] and states that the initial value problem of the Korteveg-de Vries equation \(\partial_{t}u=-\partial_{x}^{3}u+6u\partial_{x}u\) is globally in time \(C^{0}\)-well-posed on \(\mathcal{FL}_{0}^{s,p}\) with \(-\frac{1}{s}\leq s\leq0\) and and \(2\leq p<\infty\). The line of the proof of Theorem 1 uses the spectral methods developed in [Korotyaev and Kuksin, loc. cit.] while proof of Theorem 2 uses asymptotic estimates of periodic and Dirichlet eigenvalues of the (Lax) operator \(-\partial_{x}^{2}+u\) for \(u\in\) \(\mathcal{FL}_{0}^{s,p}\), to some extent present already in [the author et al., loc. cit.].