On isospectral potentials on tori (Q808319): Difference between revisions

From author's summary: ``Given a square integrable function q on a flat torus \(T={\mathbb{R}}^ d/{\mathcal L}\), denote by spec(q) the eigenvalue spectrum of the Schrödinger operator \(-\Delta+q\) and set \(Iso(q)=\{q'\in L^ 2(T): spec(q)=spec(q')\}\), the isospectral set of q. In case \(d=1\) (i.e., T is a circle), all the isospectral sets are well known; generically, they are infinite-dimensional tori. In dimension \(d\geq 2\) there is evidence that the isospectral sets are small. However, the only isospectral sets known are those of the constant potentials. They are uniquely determined by their spectra. A natural starting point in studying dimensions \(d\geq 2\) is to consider Iso(q) for q a completely separable potential on a rectangular torus \(T={\mathbb{R}}^ d/(a_ 1{\mathbb{Z}}\times...\times a_ d{\mathbb{Z}})\). We say q is completely separable if q can be written in the form \(q=\sum^{d}_{i=1}q_ i(x_ i)\), where \(q_ i\in L^ 2({\mathbb{R}}/a_ i{\mathbb{Z}})\). For such potentials q, \(\sum^{d}_{i=1}Iso(q_ i)\subseteq Iso(q)\) where \(Iso(q_ i)\) is the isospectral set of the potential \(q_ i\) in \(L^ 2({\mathbb{R}}/a_ i{\mathbb{Z}})\). Thus in this case Iso(q) being small means that the above inclusion is in fact an equality. Therefore, we ask (Q) does \(Iso(q)=\sum^{d}_{i=1}Iso(q_ i)?\) Question (Q) is equivalent to the following two questions. (Q1) If q is completely separable and \(spec(\tilde q)=spec(q)\), is \(\tilde q\) completely separable? (Q2) If q and \(\tilde q\) are isospectral, completely separable potentials, are the one-dimensional potentials \(q_ i\) and \(\tilde q_ i\) isospectral, \(1\leq i\leq d\) (up to possible permutations)? Our interest in this problem is motivated by a paper of \textit{G. Eskin}, \textit{J. Ralston} and \textit{E. Trubowitz} [Commun. Pure Appl. Math. 37, 647-676 and 715-753 (1984; Zbl 0574.35021 and Zbl 0582.35031)] which addresses question (Q1). They answer positively question (Q1) for generic rectangular tori by constructing certain spectral invariants involving a decomposition of q into a sum of one-dimensional potentials. One of our main results is an affirmative answer to (Q) in dimension \(d=2,3\) for a large class of rectangular tori, including all rational tori. The extra hypothesis on the tori is used only for question (Q2) and can be dropped completely in case \(d=2\) if we impose a mild regularity condition on the potentials.''