Mathematical analysis demonstrates that reconstructing fields and curvature in closed spatial geometries requires multiple radial shells, revealing sharp limits on single-shell observations.
This work develops a finite-dimensional theory of field reconstruction and curvature identifiability in a closed spatial geometry, using scalar fields with an exact harmonic cutoff 0 ≤ n ≤ L on a round three-sphere S³. The central distinction is between three inverse problems that are often conflated: reconstructing a spatial field from angular observations, recovering that field from a radially resolved region, and identifying the curvature radius itself. For q concentric spherical maps, the analysis determines the lost field subspace exactly: ker B_q = P_q(cos θ) H≤L−q, with dimension NL−q, where N_j = (j+1)(j+2)(2j+3)/6. This saturates a general rank bound for q rotationally equivariant scalar channels, rank T ≤ N_L − NL−q, showing that increasing angular resolution cannot compensate for missing radial information. In particular, a single spherical map contains only (L+1)² independent field directions, whereas the full three-dimensional band contains N_L ∼ L³/3. Conversely, L+1 appropriately chosen shells are sufficient for exact reconstruction, and a Gauss–Chebyshev shell configuration provides an exact isometry in which every singular value equals one. The situation changes when a spherical surface is replaced by a radially resolved layer of nonzero half-width ε. Such data are mathematically injective for every ε > 0, but reconstruction becomes increasingly ill-conditioned as the layer collapses. The squared singular values separate into a hierarchy λℓk(ε) = Θ(ε²ᵏ), with the ε²ᵏ branch occurring with multiplicity (L−k+1)². The weakest mode scales as λ_min(ε) ∼ [sin²ᴸχ₀ h_L(w) / J_L(cos χ₀)] ε²ᴸ, implying an inverse amplification ||A_ε⁺|| ∼ ε⁻ᴸ. Thus uniqueness and practical recoverability are sharply different notions: an arbitrarily thin resolved volume formally determines the finite-band field, while the precision required for reconstruction diverges rapidly in the zero-width limit. The statistical analysis reveals an equally important distinction. For a single spherical map of an SO(4)-isotropic Gaussian field with unrestricted positive band powers P_n, changes in angular radius χ can be compensated exactly by changes in the spectrum, P(χ) = M(χ)⁻¹M(χ₀)P⁽⁰⁾. Consequently, after profiling over the P_n, the map carries zero local Fisher information about χ, and therefore about R when χ = d/R with fixed d. A second calibrated shell can break this degeneracy without reconstructing the entire field. In the highest angular sector ℓ = L, only the spatial degree n = L contributes, so the same realization appears on both shells with a purely geometric relative amplitude: b₂ = g(R)b₁, where g(R) = [sin(d₂/R)/sin(d₁/R)]ᴸ. For independently known distances d₁ and d₂ and R > d₂/π, this ratio is monotonic and determines a unique curvature radius whenever 0 < g¹/ᴸ < d₂/d₁. For d₂ = 2d₁, the inversion becomes R = d₁ / arccos(g¹/ᴸ/2). The result eliminates both the unknown realization amplitude and the unknown highest-band power P_L. The paper further derives the corresponding noise information, the nearly flat K = R⁻² limit, and a quantitative contamination bound for violations of the exact spectral cutoff. These results are conditional observability theorems, not a detection of cosmic curvature: they identify precisely which information about a closed geometry is recoverable from specified observation operators, and which additional transfer physics, calibration, spectral assumptions, or radial information would be required for an observational cosmological test.
No takes yet. Share an insight, caveat, or question.
Batenin et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: