Theorem E.7 — active-spectrum reduction: ,
OpenSpectralFeedback.Supplement.theorem_E_7_active_spectrum_reductionlinear-algebramatrix-analysisoptimizationp2o-gran-per-chapterp2o-plan-paperp2o-v1quantum-information
Let act on , and . Put , , and in any decreasing spectral resolution of define
Then
where the truncation is not renormalized. On the maximum is attained by atoms each of rank at most , and adding to any one of these atoms attains the same optimum for with the same symbols.
This is the instance of the base mission's Lemma H.1. It needs the continuity and monotonicity of , which the lemma takes as hypotheses.
Differences from the source:
- encoding: the value G_{m,r}(ρ) (a maximum, (C.6)) → IsGreatest of the published rewardSet with F = χ_r; reason: the base mission's convention; IsGreatest asserts the maximum exists and equals the value, never a junk sSup
- encoding: ρ_[s] in any decreasing spectral resolution → the published activePart ρ U (min d (m·r)) for every decreasing spectral basis U (IsSortedEigenbasis); reason: the statement holds for every such U
- reading: adding ρ − ρ_[s] to one of these atoms → to any atom h₀ of the rank-≤ r active optimizer; reason: 'one of these atoms' does not name the atom; the stronger reading (every choice works) is stated
Preamble
import Mathlib import Definitions.Def_SpectralFeedback_rewardSet import Definitions.Def_SpectralFeedback_pairAtom import Definitions.Def_SpectralFeedback_Supplement_chi open scoped ComplexOrder
Formal statement
namespace SpectralFeedback
namespace Supplement
/-- **Theorem E.7** (Active-spectrum reduction; Qi, *Spectral Structure and Exact Arithmetic of
Finite-Feedback Measurement Discrimination*, manuscript (Nov. 2026), p.82 (PDF p.84)). Let `ρ ⪰ 0`
act on `ℂ^d`, let `1 ≤ r ≤ d`, and let `m ≥ 1`. Put `p = mr`, `s = min(d, p)`, and, in any
decreasing spectral resolution of `ρ`, define `ρ_{[s]} = ∑_{i=1}^s λ_i(ρ) e_i e_i*`. Then
`G_{m,r}(ρ) = G_{m,r}(ρ_{[s]})` (E.12). The truncation is not renormalized. On `ρ_{[s]}`, the maximum
can be attained with every atom of rank at most `r`. Adding `ρ − ρ_{[s]}` to one of these atoms
attains the same optimum for `ρ`, using at most the same `m` symbols.
**Formalization Note** `G_{m,r}(·) = v` is `IsGreatest (rewardSet r (chi r) m ·) v`; `ρ_{[s]}` is the
published `activePart ρ U (min d (m * r))` for an arbitrary decreasing spectral basis `U`
(`IsSortedEigenbasis`), and the statement holds for every such `U`. "One of these atoms" is read as
any atom `h₀` of the rank-`≤ r` active optimizer. -/
theorem theorem_E_7_active_spectrum_reduction (m r : ℕ) (hm : 1 ≤ m) {d : ℕ} (hr : 1 ≤ r)
(hrd : r ≤ d) (ρ : Matrix (Fin d) (Fin d) ℂ) (hρ : ρ.PosSemidef)
(U : Matrix (Fin d) (Fin d) ℂ) (hU : IsSortedEigenbasis ρ U) :
∃ v : ℝ,
IsGreatest (rewardSet r (chi r) m ρ) v ∧
IsGreatest (rewardSet r (chi r) m (activePart ρ U (min d (m * r)))) v ∧
∃ A : Fin m → Matrix (Fin d) (Fin d) ℂ, (∀ h, (A h).PosSemidef) ∧
∑ h, A h = activePart ρ U (min d (m * r)) ∧ ∑ h, reward r (chi r) (A h) = v ∧
(∀ h, (A h).rank ≤ r) ∧
∀ h₀ : Fin m,
(∀ h, (Function.update A h₀ (A h₀ + (ρ - activePart ρ U (min d (m * r)))) h).PosSemidef) ∧
∑ h, Function.update A h₀ (A h₀ + (ρ - activePart ρ U (min d (m * r)))) h = ρ ∧
∑ h, reward r (chi r)
(Function.update A h₀ (A h₀ + (ρ - activePart ρ U (min d (m * r)))) h) = v := by sorry
end Supplement
end SpectralFeedback
Source
Qi, Spectral Structure and Exact Arithmetic of Finite-Feedback Measurement Discrimination, manuscript (Nov. 2026; PDF sha256 a5b5cf40…eabde9cf), Theorem E.7, (E.12), p.82 (PDF p.84)