Keyboard shortcuts

Press or to navigate between chapters

Press ? to show this help

Press Esc to hide this help

Maximal Spectral Sharpness Characterises the Support-at-Most-Half Regime

Abstract

Spectral sharpness attains its maximum exactly when the support is at most half the dimension.

Theorem 1.1 (Spectral sharpness is one iff the support is at most half the dimension).

Proof. Machine-checked in Lean as D5/S3/Quantum/Sharpness/SpectralSharpnessSaturation.spectral_sharpness_one_iff_support_le_half (✓ std3). ∎

Source. Repository-derived.

Commentary.

The spectral sharpness of a spectrum r is the total variation between the spectrum and its reversal, sharp(r) = (1/2) sum_i |r_i - r_{rev i}|. For a probability spectrum on n points — antitone, nonnegative, and unit-sum — the sharpness attains its maximum value 1 exactly when the support (the set of nonzero entries) has cardinality at most floor(n/2). In words, freedom is maximally saturated without any pure state: it suffices that the support does not exceed half the dimension. Nonnegativity is a genuine load-bearing hypothesis here, used in the triangle-equality step.

Sharpness 1 is equivalent to mutual singularity of the spectrum and its reversal, that is r_i = 0 or r_{rev i} = 0 for every i: the total variation equals 1 iff every summand saturates the triangle bound |a - b| <= a + b, which for nonnegative entries forces one of r_i, r_{rev i} to vanish. Under antitonicity the support is a downward-closed prefix, so mutual singularity makes the support and its reversed image disjoint, giving 2 times the support size at most n, hence support size at most floor(n/2); conversely a prefix support of size at most floor(n/2) is disjoint from its reversal, which yields mutual singularity.

This records only the saturation clause sharp(r) = 1 iff support at most floor(n/2) of the maximal-sharpness law. The companion clauses of that law — the pure-state capacity, the spectral-pairing closed form of the tunable capacity, the variational supremum realising the sharpness, the median-cut plus-or-minus-one witness, and the minimal-endpoint characterisation sharp(r) = 0 iff the spectrum is uniform — are not covered by this statement.

References