Orthonormal Spectral Cluster Bounds on Manifolds with Nonpositive Curvature

Technical Summary: Orthonormal Spectral Cluster Bounds on Manifolds with Nonpositive Curvature

Problem Statement
The paper addresses the problem of establishing sharp spectral cluster bounds for orthonormal systems of eigenfunctions on closed Riemannian manifolds $(M, g)$ with nonpositive sectional curvature. Specifically, it investigates the behavior of the spectral projector $\Pi_{\lambda, \varepsilon(\lambda)}$ associated with the Laplace–Beltrami operator $-\Delta_g$ in the "supercritical" range of exponents $q > q_c$, where $q_c = 2(n+1)/(n-1)$.

While universal bounds for single eigenfunctions (Sogge's bounds) and orthonormal systems (Frank–Sabin bounds) are well-established for window sizes $\varepsilon(\lambda) = 1$, the authors focus on the case where the spectral window is logarithmically small, $\varepsilon(\lambda) = (\log \lambda)^{-1}$. In this setting, Hassell and Tacy previously proved sharp logarithmically improved $L^q$ estimates for single functions on manifolds with nonpositive curvature. The central question is whether these improvements extend to orthonormal systems, specifically determining the optimal dependence on the coefficients $\nu = (\nu_j)$ in the $\ell^\beta$ norm.

Methodology
The authors employ a synthesis of three primary analytical tools to derive their results:

1. Universal Orthonormal Bounds: They utilize the general framework established by Frank and Sabin, which provides bounds for orthonormal systems based on the spectral projector properties. This serves as the foundational input for handling the orthogonality of the system.
2. Bérard-Type Kernel Estimates: Leveraging the nonpositive curvature assumption, the authors incorporate pointwise kernel estimates for the wave operator (specifically related to the global part of the spectral multiplier). These estimates, rooted in the work of Bérard and Hassell–Tacy, provide the necessary decay to achieve logarithmic improvements over the universal bounds.
3. Generalized Multiplier Estimates: A core technical innovation is the generalization of the Bourgain–Shao–Sogge–Yao multiplier estimate to the orthonormal setting. The authors prove a spectral multiplier theorem (Theorem 2.1) that allows them to control operators of the form $\sum \nu_j |m(P)u_j|^2$ using a specific norm $K_q(m)$ of the multiplier function $m$. This theorem is derived via duality principles and interpolation techniques.

The proof of the main theorem (Theorem 1.1) proceeds by decomposing the spectral multiplier into a "local" part (near the frequency $\lambda$) and a "global" part (away from $\lambda$). The local part is handled using the generalized multiplier theorem and integration by parts, while the global part relies on the nonpositive curvature to establish pointwise kernel bounds, followed by bilinear complex interpolation to reach the desired Schatten norm estimates.

Key Contributions and Results
The primary result is Theorem 1.1, which establishes sharp, logarithmically improved spectral cluster bounds for orthonormal systems on manifolds with nonpositive sectional curvature.

• The Bound: For an orthonormal system $(u_j) \subset E_{\lambda, (\log \lambda)^{-1}}$ and coefficients $\nu$, the paper proves:
  $$\left\| \sum_{j \in J} \nu_j |u_j|^2 \right\|_{L^{q/2}(M)} \lesssim_{q, \beta} (\log \lambda)^{-1} \lambda^{n(1-2/q)-1} \|\nu\|_{\ell^\beta}$$
  for supercritical exponents $q > q_c$ and $1 \le \beta < \frac{q(n-1)}{2n}$.

• Optimality: The authors demonstrate that this result is optimal in two distinct senses:

  1. The factor $(\log \lambda)^{-1} \lambda^{n(1-2/q)-1}$ cannot be improved, as it reduces to the known sharp bound for single functions (Hassell–Tacy) when the system size is 1.
  2. The condition on the exponent $\beta$ is sharp. If $\beta \ge \frac{q(n-1)}{2n}$ (for finite $q$), the inequality fails. This is shown by testing the inequality against an orthonormal basis of the eigenspace with constant coefficients, utilizing the Weyl law and the dimension of the spectral window.

• Spectral Multiplier Theorem: As a byproduct, the paper establishes Theorem 2.1, a general spectral multiplier estimate for orthonormal systems that extends the Bourgain–Shao–Sogge–Yao lemma. This result is presented as potentially of independent interest, providing a unified framework for handling spectral multipliers in the orthonormal setting without relying on specific geometric assumptions beyond the universal bounds.

Significance and Claims
The paper claims to successfully generalize the logarithmically improved $L^q$ estimates of Hassell and Tacy from single eigenfunctions to orthonormal systems. By combining universal orthonormal bounds with geometric kernel estimates, the authors close a gap in the understanding of spectral clusters on manifolds with nonpositive curvature.

The significance lies in the precise characterization of the "Lieb–Thirring philosophy" in this context: the paper quantifies how orthogonality "needs space" to improve bounds, specifically showing that the improvement factor $(\log \lambda)^{-1}$ persists for orthonormal systems but is constrained by a specific threshold for the coefficient norm $\beta$. The work does not propose new applications or experimental setups but rather refines the theoretical bounds within the existing framework of spectral geometry and harmonic analysis on manifolds.