spectral-positivity
Perron-Frobenius theory, Jentzsch's theorem, and matrix/operator positivity in Lean 4. A shared library used by:
- pphi2 — transfer matrix spectral gap
- graphops-qft — heat kernel positivity on graphops
- markov-semigroups — positivity of Markov semigroups
What this package provides
Matrix level (Matrix/)
- Nonneg matrix powers:
(proved) - Metzler matrix exponential:
nonneg off-diagonal for (the heat kernel on a graph has nonneg entries) - Decomposition:
with , so - M-matrix inverse positivity: if
is a nonsingular M-matrix (a -matrix with positive principal minors / equivalently with entrywise and ), then entrywise. Theorem Matrix.MMatrix.inverse_posinMatrix/MMatrixInverse.lean. Used by markov-semigroups for the diamagnetic-inequality chain and by graphops-qft for resolvent positivity. - M-matrix inverse non-negativity: the same conclusion without the
irreducibility hypothesis:
PD with for entrywise. Theorem Matrix.MMatrix.inverse_nonneginMatrix/MMatrixInverse.lean. Used by markov-semigroups (Matrix/LaplaceTransform.leanre-exports it asm_matrix_inverse_nonneg, the previously axiomatized form) for the diamagnetic chain consumed by pphi2N's QHJ thimble proofs.
Operator level (Operator/)
- Positivity-preserving:
(on ) - Positivity-improving:
, a.e. - Jentzsch's theorem: For compact self-adjoint positivity-improving
: - Simple ground eigenvalue
- Ground eigenfunction
a.e. - Spectral gap:
for all other eigenvalues
- Simple ground eigenvalue
Origin
The Jentzsch proof was originally developed in pphi2 (1082 lines, fully proved) for the P(Phi)_2 transfer matrix. This package extracts and generalizes it to work on any measure space, making it reusable.
The matrix positivity was developed in graphops-qft for heat kernels on graphop semigroups.
File structure
SpectralPositivity/
Matrix/
NonnegPower.lean -- M ≥ 0 ⟹ M^k ≥ 0, truncated exp nonneg
MetzlerExp.lean -- Metzler decomposition, e^{tL} ≥ 0
MMatrixInverse.lean -- M-matrix ⟹ M⁻¹ > 0 entrywise
Operator/
PositivityPreserving.lean -- IsPositivityPreserving, IsPositivityImproving
Jentzsch.lean -- Jentzsch theorem (spectral gap)
Building
lake update
lake build
References
- Reed and Simon, Methods of Modern Mathematical Physics IV, 1978, Theorems XIII.43–44.
- Horn and Johnson, Matrix Analysis, Cambridge, 2013, Thm 8.5.5.
- Berman and Plemmons, Nonnegative Matrices in the Mathematical Sciences, SIAM, 1994.
License
Copyright (c) 2026 Michael R. Douglas. Released under the Apache 2.0 license.