Two-sided spectral conditioning in the intermediate separation regime

Determine whether arbitrary-cloud spherical Fourier Gram matrices admit uniform two-sided spectral conditioning when the separation satisfies an intermediate scale between (s^{1/3}) and (s^{2/3}), thereby resolving the gap between the known worst-case lower- and upper-frame thresholds.

Background

The paper establishes a sufficient arbitrary-cloud separation condition of order κδXs2/3\kappa\delta_X\gtrsim s^{2/3} for two-sided frame bounds on the spherical Fourier atoms, using an absolute coherence row-sum estimate and Gershgorin's theorem. It also constructs grid-restricted counterexamples showing that, in the worst case, the lower-frame bound requires at least the s1/6s^{1/6} scale and the upper-frame bound requires at least the s1/3s^{1/3} scale.

Consequently, the known results leave an intermediate range between exponents $1/3$ and $2/3$. The unresolved issue is whether uniform two-sided spectral conditioning can be established somewhere in this range, or whether stronger obstructions exist. The authors explicitly distinguish this question from the sharpness of the absolute-row-sum proof method, whose exponent $2/3$ is separately shown to be optimal.

References

Therefore, a two-sided arbitrary-cloud spectral theorem needs at least the s{1/3} scale in the worst case. The interval between exponents 1/3 and 2/3 remains open for two-sided spectral conditioning.

Certified Spherical MUSIC for 3D Localization under Adversarial Subspace Perturbations  (2609.03264 - Fannjiang et al., 3 Sep 2026) in Section 3, 'Frame bounds,' Remark following Theorem \ref{thm:sparse-spherical-frame}

Our second objective is to establish a more complete theory for the resolution limit of super-resolution from random measurements. Such a theory should provide both lower and upper bounds for the resolution limit, as well as quantitative stability estimates for the recovered locations and amplitudes. It should also reveal how randomization affects these different regimes and whether random sampling can achieve nearly the same resolution as complete Fourier measurements with substantially fewer measurements.

Generalized Hankel/Toeplitz matrix for array signal processing  (2609.03325 - Huang et al., 3 Sep 2026) in Section 6, Conclusion and future work