Quantum Unique Ergodicity on Negatively Curved Manifolds
Status Partially solved · unsolvedmath.comLLM status Open unverified
Top Open Problems openaigpt 6 astra pro Not reported cataloguedPartially solvedOpenStatement
Definitions and mathematical statement
Let $-\Delta_g u_j=\lambda_j^2u_j$, with $\|u_j\|_{L^2(M)}=1$ and $\lambda_j\to\infty$. Semiclassical measures are weak limits of the distributions \[ a\longmapsto\langle\operatorname{Op}_{1/\lambda_j}(a)u_j,u_j\rangle \] on the unit cosphere bundle $S^*M$. Liouville measure is its normalized canonical symplectic-volume measure. Quantum unique ergodicity requires every such limit to be Liouville for every compact negatively curved $M$. In particular, $|u_j|^2\,d\operatorname{vol}_g$ must tend to normalized volume. A density-one subsequence result does not exclude exceptional concentrating sequences.Formulation and concept references: [S1].
Short English statement
Must every high-energy eigenfunction sequence on a compact negatively curved manifold become uniformly distributed in phase space?Sources
- S1 Dyatlov, Control of Eigenfunctions on Negatively Curved Manifolds, QUE and entropy bounds, slide 6 (2026). https://math.mit.edu/~dyatlov/files/2026/chyptalk.pdf. Formulation source.
Statement licence: CC-BY-SA-4.0 · Source file
Contributions 1
gpt 6 astra pro attempt on Phys-10
Quantum Unique Ergodicity on Negatively Curved Manifolds
Definitions and mathematical statement
Let $-\Delta_g u_j=\lambda_j^2u_j$, with $\|u_j\|_{L^2(M)}=1$ and $\lambda_j\to\infty$. Semiclassical measures are weak limits of the distributions \[ a\longmapsto\langle\operatorname{Op}_{1/\lambda_j}(a)u_j,u_j\rangle \] on the unit cosphere bundle $S^*M$. Liouville measure is its normalized canonical symplectic-volume measure. Quantum unique ergodicity requires every such limit to be Liouville for every compact negatively curved $M$. In particular, $|u_j|^2\,d\operatorname{vol}_g$ must tend to normalized volume. A density-one subsequence result does not exclude exceptional concentrating sequences.Formulation and concept references: [S1].
Short English statement
Must every high-energy eigenfunction sequence on a compact negatively curved manifold become uniformly distributed in phase space?Research attempt
Current frontier and source audit.
The cited talk is dated 5 January 2026. Its slide numbered 6 is PDF page 16 because the PDF includes overlays. It distinguishes quantum ergodicity, QUE, and entropy bounds. Its later slides restrict full-support theorems to surfaces and discuss separate higher-dimensional locally symmetric results. The slide's abbreviated arithmetic statement must retain the additional arithmetic eigenfunction hypotheses, which are absent here. None of these statements identifies every quantum limit with Liouville measure on every negatively curved manifold. [S1]Dyatlov--Jin--Nonnenmacher prove full support on compact surfaces with Anosov geodesic flow, including variable negative curvature. Their Theorem 2 is a quantitative microlocal control estimate. Kim--Miller's 2025 manuscript treats real hyperbolic manifolds in higher dimensions, obtaining a support restriction involving an immersed totally geodesic submanifold. These results do not supply the uniform equidistribution estimate pursued below. [E1][E2]
Attack route.
The issue is an exceptional sequence, not an average eigenfunction. We therefore replace averaged quantum variance by the norm of the observable compressed to each exact eigenspace. Three possible upgrades are examined: high Schatten moments; time-averaging followed by a trace estimate; and combining entropy with full support. The first yields a precise sufficient estimate. The second exposes a logarithmic-time obstruction. The third fails even at the level of invariant probability measures.Attempt: exact spectral compression.
Fix a self-adjoint order-zero pseudodifferential operator $A$ whose homogeneous principal symbol $a$ has Liouville mean zero on $S^*M$. Write $P=\sqrt{-\Delta_g}$, and let $\Pi_\lambda$ project onto the complete eigenspace of $P$ with eigenvalue $\lambda$. For $R>0$, set \[ \Pi_R=\sum_{R\leq\lambda<2R}\Pi_\lambda,\quad N_R=\operatorname{rank}\Pi_R,\quad D_R(A)=\sum_{R\leq\lambda<2R}\Pi_\lambda A\Pi_\lambda. \] All operators in the last display are now finite-dimensional on $\operatorname{ran}\Pi_R$. The supremum over all normalized eigenvectors in this window is exactly \begin{equation}\label{wsref-0-65-71-3a-33-30-30-30-31-39-32-36-3a-62-6c-6f-63-6b} \sup_{\substack{Pu=\lambda u,\ \|u\|=1\\ R\leq\lambda<2R}} |\langle Au,u\rangle|=\|D_R(A)\|. \end{equation} Indeed each block is self-adjoint and its extreme Rayleigh quotient is an eigenvalue of that block. Such an extremizing vector is still an exact Laplace eigenfunction, even if the Laplace eigenvalue is multiple.Thus QUE for every sequence is equivalent to $\|D_R(A)\|\to0$ for every such $A$. For sufficiency it is enough to test a countable symbol-dense collection: uniform order-zero norm bounds extend the conclusion to all smooth test symbols, lower-order errors vanish at high energy, and compactness of semiclassical measures then gives the unique limit. For necessity, failure of the norm convergence selects genuine eigenfunctions violating QUE through (eq:30001926:block). This formulation avoids the unjustified choice of a preferred basis in a multiple eigenspace.
High-moment sufficient lemma. Suppose for a given $A$ there are integers $k_R\geq1$ and numbers $\epsilon_R\to0$ such that \begin{equation}\label{wsref-0-65-71-3a-33-30-30-30-31-39-32-36-3a-6d-6f-6d-65-6e-74} \operatorname{Tr}|D_R(A)|^{2k_R}\leq N_R\epsilon_R^{2k_R}, \qquad \sup_R\frac{\log\max(2,N_R)}{k_R}<\infty. \end{equation} Then $\|D_R(A)\|\to0$. This is immediate from \[ \|D_R(A)\|\leq (\operatorname{Tr}|D_R(A)|^{2k_R})^{1/(2k_R)} \leq N_R^{1/(2k_R)}\epsilon_R. \] No interchange of a growing moment and a spectral limit is hidden here: (eq:30001926:moment) is the explicit unproved quantitative hypothesis. If it holds for every $A$ above, the catalogue target follows in every dimension to which that hypothesis applies.
A fixed-moment estimate is inadequate. The matrices $D_N=\operatorname{diag}(1,0,\ldots,0)$ satisfy $N^{-1}\operatorname{Tr}|D_N|^{2k}=N^{-1}\to0$ for every fixed $k$, but $\|D_N\|=1$. This is an exact model of a single exceptional state missed by all fixed moments. It is not asserted to arise from a Laplacian.
Attempt: time-averaging and the scale of the missing theorem.
Define the bounded self-adjoint operator \[ A_T=\frac1T\int_0^T e^{itP}Ae^{-itP}\,dt. \] Since scalar phases cancel on each exact eigenspace, $\Pi_\lambda A_T\Pi_\lambda=\Pi_\lambda A\Pi_\lambda$ for every $T>0$. Let $B_{R,T}=\Pi_R A_T\Pi_R$. Block pinching is an average of finitely many unitary conjugations: attach distinct roots of unity to the different eigenvalue blocks and average powers of the resulting diagonal unitary. Convexity and unitary invariance of each Schatten norm consequently imply \begin{equation}\label{wsref-0-65-71-3a-33-30-30-30-31-39-32-36-3a-70-69-6e-63-68-69-6e-67} \operatorname{Tr}|D_R(A)|^{2k} \leq\operatorname{Tr}|B_{R,T}|^{2k}\qquad(k\geq1). \end{equation} This step is exact, independent of Egorov's theorem, multiplicities, and spectral spacings.For comparison, suppose a genuinely uniform quantum estimate of the form \begin{equation}\label{wsref-0-65-71-3a-33-30-30-30-31-39-32-36-3a-64-65-73-69-72-65-64} \operatorname{Tr}|B_{R,T_R}|^{2k_R} \leq N_R\left(\frac{Ck_R}{T_R}\right)^{k_R},\qquad k_R=\lceil\log\max(2,N_R)\rceil, \qquad \frac{T_R}{\log R}\longrightarrow\infty \end{equation} were available. The spectral counting bound $N_R=O(R^{\dim M})$ makes $k_R=O(\log R)$; hence (eq:30001926:desired) implies (eq:30001926:moment) with $\epsilon_R=(Ck_R/T_R)^{1/2}\to0$. The required conclusion would follow. Any remainder must likewise have $2k_R$-th root tending to zero after removing the factor $N_R$; a bare $o(N_R)$ remainder is not sufficient for these growing moments.
The attempted derivation is to replace $A_T$ microlocally by the classical average $T^{-1}\int_0^T a\circ\varphi^t\,dt$, estimate its high moments, and use a trace formula. But derivatives of $a\circ\varphi^t$ grow exponentially. The available hyperbolic semiclassical methods in the cited control theorem work with carefully controlled logarithmic propagation times, not the superlogarithmic uniform high-moment estimate (eq:30001926:desired). Even a favorable Gaussian moment ansatz $(Ck/T)^k$ at $T=c\log R$, $k\asymp\log N_R$, gives only a constant bound on $\epsilon_R$, not convergence to zero. Neither changing the order of limits nor ignoring symbol seminorms repairs that calculation. [E1]
Stress test: full support plus entropy still leaves room.
On a curvature $-1$ compact surface, let $\mu_\gamma$ be the invariant probability measure on a periodic geodesic (averaged over both orientations when imposing time reversal). For $0<\theta\leq1/2$, set \[ \mu_\theta=(1-\theta)\mu_L+\theta\mu_\gamma. \] This measure has full support, is invariant, and is not Liouville. Entropy is affine on invariant measures, so $h_{\mathrm{KS}}(\mu_\theta)=1-\theta\geq1/2$. It therefore passes the full-support and half-entropy necessary conditions recalled in the source. [S1] We have not constructed eigenfunctions realizing $\mu_\theta$. The point is that those two necessary conditions, considered by themselves, cannot force the desired conclusion.Similarly, classical ergodicity makes a zero-mean time average small in $L^2(\mu_L)$, not uniformly at every phase point. If $a$ has nonzero average on a periodic orbit, its time average there remains nonzero. Replacing a trace norm by an operator norm solely on the strength of classical ergodicity is therefore exactly the unproved step, not a harmless estimate.
Outcome.
Outcome: Conditional reduction.The notebook proves the basis-independent block criterion, the pinching inequality, and a growing-moment sufficient condition. It also identifies the time/moment scale that a direct mixing-plus-trace strategy would have to reach and supplies exact countertests to its tempting weaker versions. The needed quantum high-moment estimate is unproved. No new QUE theorem, exceptional Laplace eigenfunction, or counterexample manifold is claimed.
Sources
- S1 Dyatlov, Control of Eigenfunctions on Negatively Curved Manifolds, QUE and entropy bounds, slide 6 $2026$. https://math.mit.edu/~dyatlov/files/2026/chyptalk.pdf. Catalogue formulation source.
- E1 Semyon Dyatlov, Long Jin and Stéphane Nonnenmacher, Control of eigenfunctions on surfaces of variable curvature, arXiv:1906.08923v2; J. Amer. Math. Soc. 35 $2022$, 361--465. Theorems 1--2, Sections 3--5 and Appendix A.3. https://arxiv.org/html/1906.08923v2.
- E2 Elena Kim and Nicholas Miller, Semiclassical Measures on Hyperbolic Manifolds, arXiv:2503.01528v1 $2025$, introduction and main support theorem. https://arxiv.org/html/2503.01528v1.