Sampling and numerical methods

Exact Spectra of Generalized Cubic Subdivision Matrices

ORCID

Preprintv1.0.0

Abstract

The spectrum of a subdivision matrix controls local convergence near an extraordinary mesh element. For every surface valence n ≥ 3, Dietz's generalized cubic B-spline construction produces an initial-element matrix

S_n = exp((log 2)(L̃_n − I)) exp((log 2)(R̃_n − I)),

and a released double-ring matrix Ŝ_n assembled from S_n and regular cubic stencils. The source established two independent 1/2-eigenvectors but left open whether 1/2 is subdominant, has exactly the expected multiplicity, and is semisimple. We resolve all three questions and give the complete spectra of both released surface matrices. A discrete Fourier transform reduces S_n to one 3 × 3 block and n − 1 blocks of size 2 × 2; a phase-free block-triangular form gives the outer spectrum

{0^[24n], (1/64)^[n], (1/32)^[n], (1/16)^[n], (1/8)^[n]},

where brackets denote algebraic multiplicity. Thus both matrices have a simple dominant eigenvalue 1, a semisimple subdominant eigenvalue 1/2 of algebraic and geometric multiplicity two, and no other eigenvalue of modulus at least 1/2. For n ≥ 5, the subsubdominant eigenvalue satisfies

1/2 − 3π²/(4n²) + π⁴/(2n⁴) + O(n⁻⁶)

as n → ∞, explaining the vanishing spectral gap. A proved tensor identity gives the expected multiplicity three for the semi-regular initial-volume family. The analogous double-shell conclusion remains conditional on the tensor identity stated with a proof sketch in Dietz's source. Exact-identity checks, independent reconstructions, and Lean certificates accompany the paper. Fully irregular three-dimensional matrices and a complete C¹ theorem are not covered.

Contribution

Derives exact spectra of released cubic subdivision matrices and identifies the dominant and subdominant eigenvalue structure.

Formalization scope

partial

What Lean checks

All 28 outer roles per sector and exact rational entries of the released outer matrix Q_n; sector maps and the explicit block equivalence; equality to block normal form; strict upper triangularity of the 24n zero block; and the full characteristic polynomial with multiplicities 24n,n,n,n,n for every n ≥ 3.

What it does not check

Lean does not parse or evaluate MATLAB or cryptographically authenticate the source file; its row-by-row definition was audited against the pinned file. The central S_n Fourier decomposition, Corollary 7.3, tensor-product volume results, fully irregular volume matrices, and a complete C¹ theorem are not formalized.

This scope is reconciled with the hosted manuscript and pinned project README. The mathematical statements and assumptions in the source determine what the formalization establishes; manuscript revisions may clarify the interpretation of the same proof artifact.

Toolchain
leanprover/lean4:v4.30.0
Reproduce the formalization
git clone https://github.com/lennrt/palomar-formalizations.git
cd palomar-formalizations
git checkout 9437c5e5ba09d6ba95285eaac4532414a84923e1
cd zenodo-21925578/palomar
lake exe cache get
lake build

See the project README for the complete setup, dependencies, and selected declarations.

Citation & version

Lennart Rudolph. (2026). Exact Spectra of Generalized Cubic Subdivision Matrices. v1.0.0. Zenodo. https://doi.org/10.5281/zenodo.21925578.

View and copy BibTeX
Download .bib

Hosted manuscript
v1.0.0 ·
All versions DOI
10.5281/zenodo.21925577
Manuscript file integrity

SHA-256 of this exact hosted PDF:

bd6035626bfe4f53b4e1fda289bf3f1d3f4cfb0a8db00914bf0ba23411c2eb46