Phase-invariant Haar exponential-mean inequality on U(d)
ProvedHaarPhaseInvariantUnitary.logMeanBoundconcentrationhaar-measurelog-sobolevprobability
Let , and equip with normalized Haar probability and Hilbert–Schmidt distance. Let be continuous and -Lipschitz for . Suppose that whenever . Then
This phase-invariant specialization retains the special-unitary logarithmic Sobolev constant when passing to the full unitary Haar group. It supplies reusable exponential-moment bounds for conjugation observables, including normalized Pauli trace correlations.
Preamble
import Mathlib import Definitions.Def_QAlgorithms_Defs_SchurUnitaryEnsembles open Matrix MeasureTheory ProbabilityTheory
Formal statement
namespace HaarPhaseInvariantUnitary
theorem logMeanBound (d:ℕ) (hd:1 ≤ d)
(F:Matrix.unitaryGroup (Fin d) ℂ→ℝ) (hF:Continuous F)
(L:ℝ) (hL:0 ≤ L)
(hLip:∀U V:Matrix.unitaryGroup (Fin d) ℂ,|F U-F V| ≤
L*Real.sqrt (∑x:Fin d,∑y:Fin d,
‖(U:Matrix (Fin d) (Fin d) ℂ) x y-(V:Matrix (Fin d) (Fin d) ℂ) x y‖^2))
(hphase:∀z:Circle,∀U V:Matrix.unitaryGroup (Fin d) ℂ,
(U:Matrix (Fin d) (Fin d) ℂ)=(z:ℂ) • (V:Matrix (Fin d) (Fin d) ℂ)→F U=F V) :
Real.log (∫U,Real.exp (F U) ∂QAlgorithms.Schur.unitaryHaar d) ≤
(∫U,F U ∂QAlgorithms.Schur.unitaryHaar d)+L^2/d := by sorry
end HaarPhaseInvariantUnitarySource
Meckes–Meckes, Spectral measures of powers of random matrices, arXiv:1210.2681v3, Appendix pp.11–13, proof of Theorem15 (SU(N) log-Sobolev constant 2/N) and Lemma16; phase-invariant specialization via Herbst. Application: STOC2027 #157, Lemma4.13, printed p.30. https://arxiv.org/pdf/1210.2681