Scalar entropy losses for the finite block certificate
DefinitionNonadditivity_BlockScalarsholevo-additivity-gap-v100quantum-channelsquantum-informationreal-analysis
The Collins–Youn purity factor gives a logarithmic loss at most for and . Choosing realizes any positive additive tolerance . The interface also computes the logarithm of the output dimension and the exact cancellation between twice that logarithm and the Bell entropy upper estimate, leaving . These identities use natural logarithms.
Definition code
import Definitions.Def_Nonadditivity_Net
import Mathlib.Analysis.Complex.ExponentialBounds
import Mathlib.Analysis.Normed.Module.FiniteDimension
import Mathlib.Analysis.SpecialFunctions.Log.Basic
import Mathlib.Data.Real.Sqrt
import Mathlib.MeasureTheory.Measure.Typeclasses.Probability
import Mathlib.Tactic.Abel
import Mathlib.Tactic.FieldSimp
import Mathlib.Tactic.Linarith
import Mathlib.Tactic.Positivity
import Mathlib.Tactic.Ring
/-
Copyright (c) 2026 the Nonadditivity project contributors.
All rights reserved. See COPYRIGHT.md for licensing and attribution.
-/
/-!
# Scalar consequences of the actual block norm certificate
These lemmas insert the exact Collins--Youn constant and output dimension
into the channel entropy bounds. Entropies here use natural logarithms.
The tolerance can be chosen arbitrarily small using `κ = exp (η/2)`.
-/
noncomputable section
namespace Nonadditivity.BlockScalars
/-- The exact logarithmic loss after inserting `c_n²`. -/
theorem log_purity_factor_le {K n : ℕ} {κ : ℝ} (hK : 2 ≤ K) (hκ : 1 ≤ κ) :
Real.log (1 + (K : ℝ) ^ n * (κ * collinsYounConstant K n) ^ 2) ≤
(n : ℝ) * Real.log (1 + 9 / (K : ℝ)) + 2 * Real.log κ := by
have hKreal : (2 : ℝ) ≤ K := by exact_mod_cast hK
have hKpos : (0 : ℝ) < K := by linarith
have hκpos : 0 < κ := by linarith
have hleft : 0 < 1 + (K : ℝ) ^ n * (κ * collinsYounConstant K n) ^ 2 := by positivity
have hb : 0 < 1 + 9 / (K : ℝ) := by positivity
have hp := collinsYoun_purity_factor_le (n := n) hK hκ
have hl := Real.log_le_log hleft hp
rw [Real.log_mul (pow_ne_zero _ hκpos.ne') (pow_ne_zero _ hb.ne'),
Real.log_pow, Real.log_pow] at hl
norm_num at hl
linarith
def amplification (η : ℝ) : ℝ := Real.exp (η / 2)
theorem amplification_gt_one {η : ℝ} (hη : 0 < η) : 1 < amplification η := by
dsimp [amplification]
exact Real.one_lt_exp_iff.mpr (by linarith)
theorem amplification_log (η : ℝ) : 2 * Real.log (amplification η) = η := by
rw [amplification, Real.log_exp]
ring
theorem amplification_times_constant_pos {K n : ℕ} {η : ℝ}
(hK : 2 ≤ K) (hn : 1 ≤ n) :
0 < amplification η * collinsYounConstant K n :=
mul_pos (Real.exp_pos _) (collinsYounConstant_pos hK hn)
/-- Arbitrarily small additive entropy loss from an arbitrarily accurate
finite-dimensional norm approximation. -/
theorem log_purity_factor_le_eta {K n : ℕ} {η : ℝ}
(hK : 2 ≤ K) (hη : 0 < η) :
Real.log (1 + (K : ℝ) ^ n *
(amplification η * collinsYounConstant K n) ^ 2) ≤
(n : ℝ) * Real.log (1 + 9 / (K : ℝ)) + η := by
have hh := log_purity_factor_le (n := n) hK (amplification_gt_one hη).le
rw [amplification_log] at hh
exact hh
theorem log_output_dimension {K n : ℕ} :
Real.log ((K ^ n : ℕ) : ℝ) = (n : ℝ) * Real.log (K : ℝ) := by
rw [Nat.cast_pow, Real.log_pow]
/-- Exact cancellation of the two-copy output dimension with the Bell entropy
upper bound; no asymptotic remainder is introduced. -/
theorem two_use_entropy_cancellation (K n : ℕ) :
2 * Real.log ((K ^ n : ℕ) : ℝ) -
(n : ℝ) * (2 * Real.log (K : ℝ) - Real.log (K : ℝ) / K) =
(n : ℝ) * Real.log (K : ℝ) / K := by
rw [log_output_dimension]
ring
end Nonadditivity.BlockScalars
Source
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.