Definitions for On some additivity problems in quantum information theory
DefinitionQAlgorithms_Papers_Independent_Support50_P31The objects the formalized statements of On some additivity problems in quantum information theory (arXiv:math-ph/0003002) are stated in. Each was checked against the paper by a blind read-back and a moderator.
positiveRealTracePower: PDF pp.3–4: trace power on positive operators, with the full spectrum.pPurity: PDF p.3: Schatten purity for p at least one, including operator norm at infinity.nuP: PDF p.3, (5): maximum over ALL density inputs. For channels on a nonempty finite space this range is nonempty, bounded, and its supremum is attained.nuH: PDF p.3, (4): minimum output entropy; the same channel and dimension guards make this a nonempty bounded optimization with attainment.nuMinus: PDF p.3, (6): minimum least output eigenvalue INCLUDING zero. Continuous spectral extension at singular outputs, never a support-only inverse.maxPower: PDF p.4: the maximal trace power in the displayed right-limit identity.nuFlat: PDF p.6, (12): maximal survival probability over norm-one vectors.weakNoise: PDF p.7, (14): the actual weak-noise convex mixture.depolarizingProduct: PDF p.10: tensor product of the frozen depolarizing maps on all matrices.subsetExpectation: PDF p.10: replace factors in L by normalized traces, leave the others alone.depolarizingSubsetCoefficient: PDF p.10: membership form of the coefficient in (20); no Nat subtraction.zeroRight: PDF pp.7–11: limit through strictly positive parameters.ComparableScale: PDF p.10: comparable small parameters, without exact rays or equality. All restrictions need hold only eventually near zero from the right.
import Mathlib
import Definitions.Def_QAlgorithms_Defs_Channel
import Definitions.Def_QAlgorithms_Defs_Hamiltonian
import Definitions.Def_QAlgorithms_Defs_QuditChannelFidelities
import Definitions.Def_QAlgorithms_Papers_Independent_Gap48_P03
import Definitions.Def_QAlgorithms_Papers_Independent_Claimed50_P38
import Definitions.Def_QAlgorithms_Papers_Independent_Dependency50_P44
/-!
Amosov, Holevo and Werner, arXiv:math-ph/0003002v2, PDF pp.2–12.
Natural entropy units; all density inputs, including entangled ones.
The false minus-infinity branch of (9) and literal quantitative remainder on p.8
are deferred. The final theorem explicitly corrects the scalar-factor edge.
-/
noncomputable section
open Filter Asymptotics
open scoped ComplexConjugate ENNReal
namespace QAlgorithms.Papers.Independent.Support50.P31
attribute [local instance] Classical.propDecidable Classical.decEq
open QAlgorithms.Papers.Independent.Information100.P56 (entropy)
open QAlgorithms.Papers.Independent.Information100.P63 (schatten)
open QAlgorithms.Papers.Independent.Gap48.P03 (holevoQuantity)
open QAlgorithms.Papers.Independent.Claimed50.P38 (tensorSuperop)
open QAlgorithms.Papers.Independent.Dependency50.P44 (tensorChannels)
/-- PDF pp.3–4: trace power on positive operators, with the full spectrum. -/
def positiveRealTracePower {I : Type} [Fintype I] [DecidableEq I] (p : ℝ)
(X : Matrix I I ℂ) : ℝ :=
(cfc (fun x : ℝ => x ^ p) X).trace.re
/-- PDF p.3: Schatten purity for p at least one, including operator norm at infinity. -/
def pPurity {I : Type} [Fintype I] [DecidableEq I] (p : ℝ≥0∞)
(X : Matrix I I ℂ) : ℝ :=
if p = ⊤ then specNorm X else schatten p.toReal X
/-- PDF p.3, (5): maximum over ALL density inputs. For channels on a nonempty
finite space this range is nonempty, bounded, and its supremum is attained. -/
def nuP {I : Type} [Fintype I] [DecidableEq I] (p : ℝ≥0∞)
(Phi : Superop I I) : ℝ :=
if p = ⊤ then sSup (Set.range fun X : {X : Matrix I I ℂ // IsDensityMatrix X} => specNorm (Phi X.val))
else QAlgorithms.Papers.Independent.Gap48.P03.maximumOutputPNorm Phi p.toReal
/-- PDF p.3, (4): minimum output entropy; the same channel and dimension guards
make this a nonempty bounded optimization with attainment. -/
def nuH {I : Type} [Fintype I] [DecidableEq I] (Phi : Superop I I) : ℝ :=
sInf (Set.range fun X : {X : Matrix I I ℂ // IsDensityMatrix X} => entropy (Phi X.val))
/-- PDF p.3, (6): minimum least output eigenvalue INCLUDING zero.
Continuous spectral extension at singular outputs, never a support-only inverse. -/
def nuMinus {I : Type} [Fintype I] [DecidableEq I] (Phi : Superop I I) : ℝ :=
sInf (Set.range fun X : {X : Matrix I I ℂ // IsDensityMatrix X} => minEigenvalue (Phi X.val))
/-- PDF p.4: the maximal trace power in the displayed right-limit identity. -/
def maxPower {I : Type} [Fintype I] [DecidableEq I] (p : ℝ)
(Phi : Superop I I) : ℝ :=
sSup (Set.range fun X : {X : Matrix I I ℂ // IsDensityMatrix X} => positiveRealTracePower p (Phi X.val))
/-- PDF p.6, (12): maximal survival probability over norm-one vectors. -/
def nuFlat {I : Type} [Fintype I] [DecidableEq I] (Phi : Superop I I) : ℝ :=
sSup (Set.range fun v : {v : EuclideanSpace ℂ I // ‖v‖ = 1} =>
effectProb (pureDensity v.val) (Phi (pureDensity v.val)))
/-- PDF p.7, (14): the actual weak-noise convex mixture. -/
def weakNoise {I : Type} (eps : ℝ) (Phi : Superop I I) : Superop I I :=
fun X => ((1 - eps : ℝ) : ℂ) • X + (eps : ℂ) • Phi X
/-- PDF p.10: tensor product of the frozen depolarizing maps on all matrices. -/
def depolarizingProduct {n : ℕ} (d : Fin n → ℕ) (hd : ∀ i, 0 < d i)
(p : Fin n → ℝ) :=
tensorChannels (fun i => QuditFidelity.depolarizingMap (d i) (hd i) (p i))
/-- PDF p.10: replace factors in L by normalized traces, leave the others alone. -/
def subsetExpectation {n : ℕ} (d : Fin n → ℕ) (hd : ∀ i, 0 < d i)
(L : Finset (Fin n)) :=
tensorChannels (fun i =>
if i ∈ L then QuditFidelity.depolarizingMap (d i) (hd i) 1 else id)
/-- PDF p.10: membership form of the coefficient in (20); no Nat subtraction. -/
def depolarizingSubsetCoefficient {n : ℕ} (p : Fin n → ℝ) (L : Finset (Fin n)) : ℝ :=
∏ i, if i ∈ L then p i else 1 - p i
/-- PDF pp.7–11: limit through strictly positive parameters. -/
abbrev zeroRight : Filter ℝ := nhdsWithin 0 (Set.Ioi 0)
/-- PDF p.10: comparable small parameters, without exact rays or equality.
All restrictions need hold only eventually near zero from the right. -/
def ComparableScale {n : ℕ} (q : ℝ → Fin n → ℝ) : Prop :=
∃ a b : ℝ, 0 < a ∧ a ≤ b ∧
∀ᶠ t in zeroRight, ∀ i, 0 < q t i ∧ q t i < 1 ∧ a * t ≤ q t i ∧ q t i ≤ b * t
/-- Sanity: entropy of the scalar density is zero. -/
example : entropy (1 : Matrix (Fin 1) (Fin 1) ℂ) = 0 := by
simp [entropy, cfc_one]
/-- Sanity: trace power of the scalar density is one. -/
example (p : ℝ) : positiveRealTracePower p (1 : Matrix (Fin 1) (Fin 1) ℂ) = 1 := by
simp [positiveRealTracePower, cfc_one]
/-- Sanity: infinity selects exactly the L2 operator norm. -/
example {I : Type} [Fintype I] [DecidableEq I] (X : Matrix I I ℂ) :
pPurity ⊤ X = specNorm X := by
simp [pPurity]
/-- Sanity: zero noise is the identity map. -/
example {I : Type} (Phi : Superop I I) : weakNoise 0 Phi = id := by
funext X
simp [weakNoise]
/-- Sanity: the tensor extension preserves the zero operator. -/
example {n : ℕ} (d : Fin n → ℕ)
(Phi : (i : Fin n) → Superop (Fin (d i)) (Fin (d i))) :
tensorChannels Phi 0 = 0 := by
ext u v
simp [tensorChannels]
/-- Sanity: the accepted Holevo score of a singleton scalar ensemble is zero. -/
example : QAlgorithms.Papers.Independent.Information100.P56.holevo
(fun _ : Fin 1 => (1 : ℝ)) (fun _ : Fin 1 => (1 : Matrix (Fin 1) (Fin 1) ℂ)) = 0 := by
simp [QAlgorithms.Papers.Independent.Information100.P56.holevo,
QAlgorithms.Papers.Independent.Information100.P56.average, entropy, cfc_one]
/-- Sanity: the empty tensor coefficient is one. -/
example (p : Fin 0 → ℝ) : depolarizingSubsetCoefficient p ∅ = 1 := by
simp [depolarizingSubsetCoefficient]
/-- Sanity: a constant density output has exactly the specified entropy. -/
example {I : Type} [Fintype I] [DecidableEq I] (rho : Matrix I I ℂ)
(hrho : IsDensityMatrix rho) : nuH (fun _ => rho) = entropy rho := by
letI : Nonempty ({X : Matrix I I ℂ // IsDensityMatrix X}) := ⟨⟨rho, hrho⟩⟩
simp [nuH]
/-- Sanity: zero eigenvalues are retained by a constant-output objective. -/
example {I : Type} [Fintype I] [DecidableEq I] (rho : Matrix I I ℂ)
(hrho : IsDensityMatrix rho) : nuMinus (fun _ => rho) = minEigenvalue rho := by
letI : Nonempty ({X : Matrix I I ℂ // IsDensityMatrix X}) := ⟨⟨rho, hrho⟩⟩
simp [nuMinus]
/-- Sanity: a constant-output map has its output operator norm as its maximum. -/
example {I : Type} [Fintype I] [DecidableEq I] (rho : Matrix I I ℂ)
(hrho : IsDensityMatrix rho) : nuP ⊤ (fun _ => rho) = specNorm rho := by
letI : Nonempty ({X : Matrix I I ℂ // IsDensityMatrix X}) := ⟨⟨rho, hrho⟩⟩
simp [nuP]
/-- Sanity: the trace-power wrapper also preserves constant-output objectives. -/
example {I : Type} [Fintype I] [DecidableEq I] (rho : Matrix I I ℂ)
(hrho : IsDensityMatrix rho) (p : ℝ) :
maxPower p (fun _ => rho) = positiveRealTracePower p rho := by
letI : Nonempty ({X : Matrix I I ℂ // IsDensityMatrix X}) := ⟨⟨rho, hrho⟩⟩
simp [maxPower]
/-- Sanity: a zero output has zero survival objective on a nonempty unit sphere. -/
example {I : Type} [Fintype I] [DecidableEq I] (v : EuclideanSpace ℂ I)
(hv : ‖v‖ = 1) : nuFlat (fun _ : Matrix I I ℂ => 0) = 0 := by
letI : Nonempty {w : EuclideanSpace ℂ I // ‖w‖ = 1} := ⟨⟨v, hv⟩⟩
simp [nuFlat, effectProb]
/-- Sanity: at unit noise the weak-noise map is exactly the given map. -/
example {I : Type} (Phi : Superop I I) : weakNoise 1 Phi = Phi := by
funext X
simp [weakNoise]
end QAlgorithms.Papers.Independent.Support50.P31
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
QAlgorithms.Papers.Independent.Support50.P31.ComparableScale
For a natural number , including , and a function , this property means that there exist real constants and such that and, for every and every index ,
Thus the same constants and neighborhood of zero apply to every coordinate. When , all coordinate conditions are vacuous, so every such function satisfies the property.
QAlgorithms.Papers.Independent.Support50.P31.depolarizingProduct
For any natural number , any natural dimensions for every , and any real parameters , this defines a function on complex matrices indexed by . Define the local functions
The resulting function sends to the matrix whose entries are
where has a single in position . The parameters have no range restrictions. The case is included: is a singleton, the product is , and the resulting function is the identity on the corresponding one-dimensional matrix space.
QAlgorithms.Papers.Independent.Support50.P31.depolarizingSubsetCoefficient
For any natural number , any real parameters indexed by , and any finite subset of those indices, this defines
There are no sign or range assumptions on the . Empty and full are included, and for the empty product equals .
QAlgorithms.Papers.Independent.Support50.P31.maxPower
For any finite type with decidable equality, any real exponent , and any function from complex matrices to themselves, this defines
using the total continuous functional calculus for real scalar functions and the total real power function. The exponent may be zero or negative, and no channel, linearity, continuity, or positivity-preservation assumption is imposed on . The supremum is the total real supremum, returning zero for an empty set or a set unbounded above; empty gives zero.
QAlgorithms.Papers.Independent.Support50.P31.nuFlat
For any finite type with decidable equality and any function from complex matrices to themselves, this defines
where . No channel, linearity, continuity, or positivity-preservation assumption is imposed on . This is a total real supremum, returning zero if the set is empty or unbounded above; when is empty, there are no unit vectors and the value is zero.
QAlgorithms.Papers.Independent.Support50.P31.nuH
For any finite type with decidable equality and any function from complex matrices to themselves, this defines
where is the total natural logarithm and is evaluated by the total real continuous functional calculus. No channel, linearity, positivity-preservation, or continuity assumption is imposed on . The infimum is the total real infimum, returning zero if its set is empty or unbounded below; empty therefore gives zero.
QAlgorithms.Papers.Independent.Support50.P31.nuMinus
For any finite type with decidable equality and any function from complex matrices to themselves, this defines the real infimum, over all positive semidefinite matrices with trace , of the following value: the smallest eigenvalue of when is nonempty and is Hermitian, and zero when is empty or is non-Hermitian. No channel, linearity, continuity, or positivity-preservation assumption is imposed on . The total real infimum is zero for an empty set or a set unbounded below; in particular, empty gives zero.
QAlgorithms.Papers.Independent.Support50.P31.nuP
For any finite type with decidable equality, any extended nonnegative real , and any function from complex matrices to complex matrices, this defines
Here the norm is the Euclidean operator norm; the are the descending eigenvalues of Hermitian , with multiplicities, and are all zero for non-Hermitian . No linearity, continuity, or channel assumption is imposed on . The finite branch includes , where each displayed power expression equals . These are total real suprema, which return zero for an empty set or a set unbounded above; in particular, empty gives zero because there are no trace-one density matrices.
QAlgorithms.Papers.Independent.Support50.P31.pPurity
For any finite type with decidable equality, any extended nonnegative real , and any complex matrix , this defines
where is the Euclidean operator norm and are the eigenvalues in descending order, with multiplicities, when is Hermitian, and are all defined to be zero otherwise. The powers and division are total real operations; in particular, at , and the finite branch equals . No Hermitian or positivity assumption is imposed. Empty is allowed: the value is at and zero otherwise.
QAlgorithms.Papers.Independent.Support50.P31.positiveRealTracePower
For any finite type with decidable equality, any real exponent , and any complex matrix indexed by , this defines the real number
where is evaluated by the total continuous functional calculus for real scalar functions. No Hermitian, positivity, or normalization assumption is imposed on , and may be zero or negative; the expression uses the total real power function, including at zero. If is empty, the trace and hence the defined value are zero.
QAlgorithms.Papers.Independent.Support50.P31.subsetExpectation
For any natural number , any natural dimensions for , and any finite subset of those indices, this defines a function on complex matrices indexed by . Put
The function sends to the matrix with entries
where is the single-entry matrix unit. Empty is allowed and gives the identity function; is also allowed and gives the identity on the singleton-indexed matrix space.
QAlgorithms.Papers.Independent.Support50.P31.weakNoise
For any type , any real , and any function from complex matrices to themselves, this defines another such function by
with the real coefficients embedded into . Neither finiteness of nor any linearity or channel property of is assumed, and is unrestricted, including values outside .
QAlgorithms.Papers.Independent.Support50.P31.zeroRight
This abbreviates the filter of neighborhoods of restricted to the strictly positive real numbers . Equivalently, a set belongs to this filter exactly when there exists such that ; membership of itself in is not required.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.