Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Definitions for Adaptation of the Alicki-Fannes-Winter method for the set of states with bounded energy and its use

Definition
QAlgorithms_Papers_Independent_Support50_P18

by Goku · Oct 23, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

The objects the formalized statements of Adaptation of the Alicki-Fannes-Winter method for the set of states with bounded energy and its use (arXiv:1609.07044) are stated in. Each was checked against the paper by a blind read-back and a moderator.

  • logPowerLevel: PDF p.8, Lemma 2: n+1 enumerates k=1,2,... . The first term is finite.
  • logPowerPartition: PDF p.8, Lemma 2: a divergent partition sum is infinity, not zero.
  • OscillatorParameters: Section 4.2, PDF p.9: positive physical parameters and at least one mode. This is a parameter record, not Frontier30.P07.Oscillator (the one-mode Hilbert space).
  • OscillatorParameters.quantum: PDF p.9: the positive mode energy quanta hbar omega_i.
  • OscillatorParameters.level: PDF p.9: eigenvalues of the actual Fock Hamiltonian, including the vacuum.
  • OscillatorParameters.domain: PDF p.9: maximal dense domain of the self-adjoint diagonal Hamiltonian.
  • OscillatorParameters.apply: PDF p.9: the Hamiltonian acts by its number-basis eigenvalues.
  • OscillatorParameters.ground: PDF p.9: E0=(1/2) sum_i hbar omega_i.
  • OscillatorParameters.geometricMean: PDF p.9, (23): geometric mean of the mode energy quanta.
  • OscillatorParameters.meanEnergy: PDF pp.5,9: extended mean energy on positive trace-class arrays. The nonnegative series equals the increasing spectral-cutoff supremum.
  • OscillatorParameters.entropyMaximum: PDF pp.7,9, (18): supremum of actual entropies over the energy ball.
  • OscillatorParameters.entropyUpper: PDF p.9, (23): explicit oscillator entropy majorant.
  • OscillatorParameters.allocations: PDF p.9: all admissible allocations, with equality of total energy.
  • OscillatorParameters.allocationEntropy: PDF p.9: sum of single-mode entropies for a fixed allocation.
  • AFWConstants: PDF pp.7--9: the four state-independent nonnegative constants (16),(17).
  • OscillatorParameters.finiteMarginalEnergy: PDF p.9: finite energy of the actual oscillator marginal; R is arbitrary.
  • OscillatorParameters.EntropyControlled: PDF pp.7,9: both inequalities (16),(17), exactly on finite-energy states. f accepts coefficient arrays, with values outside the source domain unconstrained. Actual matrix mixtures are used; normalization is not an additional premise.
  • OscillatorParameters.continuityBound: PDF p.9, (24): ell and the logarithm are OUTSIDE sqrt(2 epsilon).
  • nuclearStateDistance: PDF pp.5,9: normalized-state adapter to the imported full trace distance. The frozen QAlgorithms.traceDistance requires finite matrices and returns a real; the imported extendedHalfTraceDistance covers arbitrary nuclear coefficients.
  • SubsystemBasis: PDF pp.9--10: complete basis for a selected multipartite subsystem.
  • ComplementBasis: PDF pp.9--10: complete basis for the complementary subsystem.
  • joinBasis: PDF pp.9--10: full basis coordinate reconstructed from its two factors.
  • subsystemMarginal: PDF p.10: actual subsystem marginal, using the accepted completed partial trace.
  • entropyCombination: PDF pp.9--10, (25): the prescribed finite real combination, on its domain of finite marginal entropies with nonzero coefficients. No infinite value is projected in the theorem.
  • spectralGround: PDF p.5: infimum of the unit-vector energies, in a complete eigenbasis.
  • Condition22: PDF p.8, (22): extended partition trace and a right-hand limit. Neither summability nor finite entropy is assumed separately.
  • TraceUniformContinuousOn: Trace-distance uniform continuity on a specified state set (PDF pp.5,17--18). This epsilon-delta definition avoids imposing a coefficient uniformity.
  • marginalMutualInformation: PDF p.4: genuine mutual information of a joint state, defined by relative entropy.
  • extendedConditionalEntropy: PDF p.11, (28): finite-A-entropy conditional entropy, without infinity subtraction.
  • outputMutualInformation: PDF p.17, (48): mutual information after the actual reference-amplified channel.
  • entropyModulus: PDF p.8, (21): the real modulus, used only with proved finite F_H.
  • EnergyContinuity: PDF pp.17--18: finite-valued uniform continuity and the quantitative bound, on one FIXED energy ball. This is a conclusion package, never a hypothesis.
  • countableRelativeEntanglementNats: PDF p.14, (38): relative entropy distance to the actual trace-norm closed convex hull of product states. Empty infima retain infinity. The frozen P54 relativeEntanglement is finite-dimensional and in bits; this interface is in nats on arbitrary countable systems and cannot use that finite-dimensional function.
  • bipartiteTensorPower: PDF pp.14,16: n copies, regrouped as A^n:B^n, without restricting other states of that tensor-power system to be products.
  • regularizedCountableEntanglementNats: PDF p.14, (39): regularization as the extended infimum of ER(rho^n)/n, equivalent to the actual limit by subadditivity. Only positive block lengths are used; a divergent value remains infinity. The frozen P54 regularization is finite-dimensional and in bits, so it does not represent this domain.
  • additiveEnergy: PDF p.16: additive tensor-power Hamiltonian eigenvalues.
Definition code
import Mathlib
import Definitions.Def_QAlgorithms_Papers_Independent_Frontier18_P01
import Definitions.Def_QAlgorithms_Papers_Independent_Frontier18_P07
import Definitions.Def_QAlgorithms_Papers_Independent_Claimed50_P39
import Definitions.Def_QAlgorithms_Papers_Independent_Renew46_P08

/-!
Statements from Shirokov, arXiv:1609.07044v5. Pinned accepted nuclear states,
trace norm and extended spectral entropies are reused. Product coordinates
represent completed Hilbert tensor products. General Hamiltonians, including
continuous spectra, are deferred rather than replaced by diagonal observables.
The oscillator uses its full Fock basis and includes zero-point energy.
Compilation is pending the separate batch handoff.
-/

noncomputable section

open Filter
open scoped ENNReal NNReal BigOperators Topology

attribute [local instance] Classical.propDecidable Classical.decEq

namespace QAlgorithms.Papers.Independent.Support50.P18

open QAlgorithms.Papers.Independent.Dependency50.P33 (State PureState traceSize rankOne)
open QAlgorithms.Papers.Independent.Frontier18.P01 (convexOperatorMixture traceSecond)
open QAlgorithms.Papers.Independent.Frontier18.P07
  (extendedEntropyNats extendedRelativeEntropyNats)
open QAlgorithms.RealRegisters (Hilbert)
open QAlgorithms.Papers.Independent.Claimed50.P39 (g)
open QAlgorithms.Papers.Independent.Renew46.P08
  (coefficientEnergy entropySupremum extendedHalfTraceDistance)

/-! The imported g is (x+1) log(x+1)-x log x (PDF p.9), equal to
(1+x) h2(x/(1+x)) for x>=0. Every use below has a nonnegative argument. -/

example : g 0 = 0 := by
  simp [QAlgorithms.Papers.Independent.Claimed50.P39.g]

/-- PDF p.8, Lemma 2: n+1 enumerates k=1,2,... . The first term is finite. -/
def logPowerLevel (q : ℝ) (n : ℕ) : ℝ := (Real.log (n + 1 : ℝ)) ^ q

/-- PDF p.8, Lemma 2: a divergent partition sum is infinity, not zero. -/
def logPowerPartition (q lambda : ℝ) : ℝ≥0∞ :=
  ∑' n : ℕ, ENNReal.ofReal (Real.exp (-lambda * logPowerLevel q n))

example : logPowerLevel 2 0 = 0 := by simp [logPowerLevel]

example (q lambda : ℝ) : logPowerPartition q lambda =
    ∑' n : ℕ, ENNReal.ofReal (Real.exp (-lambda * logPowerLevel q n)) := rfl

/-- Section 4.2, PDF p.9: positive physical parameters and at least one mode.
This is a parameter record, not Frontier30.P07.Oscillator (the one-mode Hilbert space). -/
structure OscillatorParameters (ell : ℕ) where
  modes_pos : 0 < ell
  hbar : ℝ≥0
  hbar_pos : 0 < hbar
  omega : Fin ell → ℝ≥0
  omega_pos : ∀ i, 0 < omega i

/-- PDF p.9: the positive mode energy quanta hbar omega_i. -/
def OscillatorParameters.quantum {ell : ℕ} (O : OscillatorParameters ell) (i : Fin ell) : ℝ≥0 :=
  O.hbar * O.omega i

/-- PDF p.9: eigenvalues of the actual Fock Hamiltonian, including the vacuum. -/
def OscillatorParameters.level {ell : ℕ} (O : OscillatorParameters ell) (n : Fin ell → ℕ) : ℝ≥0 :=
  ∑ i, O.quantum i * ((n i : ℝ≥0) + 1 / 2)

/-- PDF p.9: maximal dense domain of the self-adjoint diagonal Hamiltonian. -/
def OscillatorParameters.domain {ell : ℕ} (O : OscillatorParameters ell) : Set (Hilbert (Fin ell → ℕ)) :=
  {u | Memℓp (fun n => (O.level n : ℂ) * u n) 2}

/-- PDF p.9: the Hamiltonian acts by its number-basis eigenvalues. -/
def OscillatorParameters.apply {ell : ℕ} (O : OscillatorParameters ell) (u : O.domain) :
    Hilbert (Fin ell → ℕ) :=
  ⟨fun n => (O.level n : ℂ) * u.val n, u.property⟩

/-- PDF p.9: E0=(1/2) sum_i hbar omega_i. -/
def OscillatorParameters.ground {ell : ℕ} (O : OscillatorParameters ell) : ℝ :=
  (1 / 2 : ℝ) * ∑ i, (O.quantum i : ℝ)

/-- PDF p.9, (23): geometric mean of the mode energy quanta. -/
def OscillatorParameters.geometricMean {ell : ℕ} (O : OscillatorParameters ell) : ℝ :=
  (∏ i, (O.quantum i : ℝ)) ^ (1 / (ell : ℝ))

/-- PDF pp.5,9: extended mean energy on positive trace-class arrays.
The nonnegative series equals the increasing spectral-cutoff supremum. -/
def OscillatorParameters.meanEnergy {ell : ℕ} (O : OscillatorParameters ell)
    (X : Matrix (Fin ell → ℕ) (Fin ell → ℕ) ℂ) : ℝ≥0∞ :=
  coefficientEnergy O.level X

/-- PDF pp.7,9, (18): supremum of actual entropies over the energy ball. -/
def OscillatorParameters.entropyMaximum {ell : ℕ} (O : OscillatorParameters ell) (E : ℝ) : ℝ≥0∞ :=
  ⨆ rho : {rho : State (Fin ell → ℕ) //
    O.meanEnergy rho.coefficients ≤ ENNReal.ofReal E}, extendedEntropyNats rho.val.coefficients

/-- PDF p.9, (23): explicit oscillator entropy majorant. -/
def OscillatorParameters.entropyUpper {ell : ℕ} (O : OscillatorParameters ell) (E : ℝ) : ℝ :=
  (ell : ℝ) * Real.log ((E + O.ground) / ((ell : ℝ) * O.geometricMean)) + ell

/-- PDF p.9: all admissible allocations, with equality of total energy. -/
def OscillatorParameters.allocations {ell : ℕ} (O : OscillatorParameters ell) (E : ℝ) : Set (Fin ell → ℝ) :=
  {e | (∑ i, e i) = E ∧ ∀ i, (O.quantum i : ℝ) / 2 ≤ e i}

/-- PDF p.9: sum of single-mode entropies for a fixed allocation. -/
def OscillatorParameters.allocationEntropy {ell : ℕ} (O : OscillatorParameters ell) (e : Fin ell → ℝ) : ℝ :=
  ∑ i, g (e i / (O.quantum i : ℝ) - 1 / 2)

example {ell : ℕ} (O : OscillatorParameters ell) (u : O.domain) (n : Fin ell → ℕ) :
    O.apply u n = (O.level n : ℂ) * u.val n := rfl

example {ell : ℕ} (O : OscillatorParameters ell) :
    O.meanEnergy (0 : Matrix (Fin ell → ℕ) (Fin ell → ℕ) ℂ) = 0 := by
  simp [OscillatorParameters.meanEnergy, coefficientEnergy]

example {ell : ℕ} (O : OscillatorParameters ell) (e : Fin ell → ℝ) (E : ℝ) :
    e ∈ O.allocations E ↔ (∑ i, e i) = E ∧ ∀ i, (O.quantum i : ℝ) / 2 ≤ e i := Iff.rfl

/-- PDF pp.7--9: the four state-independent nonnegative constants (16),(17). -/
structure AFWConstants where
  a : ℝ≥0
  b : ℝ≥0
  cMinus : ℝ≥0
  cPlus : ℝ≥0

/-- PDF p.9: finite energy of the actual oscillator marginal; R is arbitrary. -/
def OscillatorParameters.finiteMarginalEnergy {ell : ℕ} {R : Type} (O : OscillatorParameters ell)
    (rho : State ((Fin ell → ℕ) × R)) : Prop :=
  O.meanEnergy (traceSecond rho.coefficients) < ⊤

/-- PDF pp.7,9: both inequalities (16),(17), exactly on finite-energy states.
f accepts coefficient arrays, with values outside the source domain unconstrained.
Actual matrix mixtures are used; normalization is not an additional premise. -/
def OscillatorParameters.EntropyControlled {ell : ℕ} {R : Type} (O : OscillatorParameters ell)
    (C : AFWConstants) (f : Matrix ((Fin ell → ℕ) × R) ((Fin ell → ℕ) × R) ℂ → ℝ) : Prop :=
  (∀ rho sigma : State ((Fin ell → ℕ) × R),
    O.finiteMarginalEnergy rho → O.finiteMarginalEnergy sigma →
    ∀ p : ℝ, 0 < p → p < 1 →
      -(C.a : ℝ) * Real.binEntropy p ≤
        f (convexOperatorMixture p rho.coefficients sigma.coefficients) -
          p * f rho.coefficients - (1 - p) * f sigma.coefficients ∧
      f (convexOperatorMixture p rho.coefficients sigma.coefficients) -
          p * f rho.coefficients - (1 - p) * f sigma.coefficients ≤
        (C.b : ℝ) * Real.binEntropy p) ∧
  (∀ rho : State ((Fin ell → ℕ) × R), O.finiteMarginalEnergy rho →
    -(((C.cMinus : ℝ≥0∞) * extendedEntropyNats (traceSecond rho.coefficients)).toEReal) ≤
        (f rho.coefficients : EReal) ∧
      (f rho.coefficients : EReal) ≤
        ((C.cPlus : ℝ≥0∞) * extendedEntropyNats (traceSecond rho.coefficients)).toEReal)

/-- PDF p.9, (24): ell and the logarithm are OUTSIDE sqrt(2 epsilon). -/
def OscillatorParameters.continuityBound {ell : ℕ} (O : OscillatorParameters ell)
    (C : AFWConstants) (E eps : ℝ) : ℝ :=
  ((C.cMinus : ℝ) + C.cPlus) * Real.sqrt (2 * eps) *
    (ell : ℝ) * (Real.log ((E / eps + O.ground) /
      ((ell : ℝ) * O.geometricMean)) + 1) +
    ((C.a : ℝ) + C.b) * g (Real.sqrt (2 * eps))

/-- PDF pp.5,9: normalized-state adapter to the imported full trace distance.
The frozen QAlgorithms.traceDistance requires finite matrices and returns a real;
the imported extendedHalfTraceDistance covers arbitrary nuclear coefficients. -/
def nuclearStateDistance {I : Type} (rho sigma : State I) : ℝ≥0∞ :=
  extendedHalfTraceDistance rho.coefficients sigma.coefficients

example {I : Type} (rho sigma : State I) : nuclearStateDistance rho sigma =
    traceSize (rho.coefficients - sigma.coefficients) / 2 := rfl

example : (AFWConstants.mk 0 0 0 0).a = 0 := rfl

/-- PDF pp.9--10: complete basis for a selected multipartite subsystem. -/
abbrev SubsystemBasis {n : ℕ} (I : Fin n → Type) (S : Finset (Fin n)) :=
  (i : {i : Fin n // i ∈ S}) → I i.val

/-- PDF pp.9--10: complete basis for the complementary subsystem. -/
abbrev ComplementBasis {n : ℕ} (I : Fin n → Type) (S : Finset (Fin n)) :=
  (i : {i : Fin n // i ∉ S}) → I i.val

/-- PDF pp.9--10: full basis coordinate reconstructed from its two factors. -/
def joinBasis {n : ℕ} {I : Fin n → Type} {S : Finset (Fin n)}
    (x : SubsystemBasis I S) (r : ComplementBasis I S) : (i : Fin n) → I i :=
  fun i => if h : i ∈ S then x ⟨i, h⟩ else r ⟨i, h⟩

/-- PDF p.10: actual subsystem marginal, using the accepted completed partial trace. -/
def subsystemMarginal {n : ℕ} {I : Fin n → Type} (S : Finset (Fin n))
    (X : Matrix ((i : Fin n) → I i) ((i : Fin n) → I i) ℂ) :
    Matrix (SubsystemBasis I S) (SubsystemBasis I S) ℂ :=
  traceSecond (fun a b : SubsystemBasis I S × ComplementBasis I S =>
    X (joinBasis a.1 a.2) (joinBasis b.1 b.2))

/-- PDF pp.9--10, (25): the prescribed finite real combination, on its domain
of finite marginal entropies with nonzero coefficients. No infinite value is projected in the theorem. -/
def entropyCombination {n m : ℕ} {I : Fin n → Type}
    (S : Fin m → Finset (Fin n)) (c : Fin m → ℝ)
    (X : Matrix ((i : Fin n) → I i) ((i : Fin n) → I i) ℂ) : ℝ :=
  ∑ k, c k * (extendedEntropyNats (subsystemMarginal (S k) X)).toReal

example {n : ℕ} {I : Fin n → Type} (S : Fin 0 → Finset (Fin n))
    (c : Fin 0 → ℝ) (X : Matrix ((i : Fin n) → I i) ((i : Fin n) → I i) ℂ) :
    entropyCombination S c X = 0 := by simp [entropyCombination]

example {n : ℕ} {I : Fin n → Type} {S : Finset (Fin n)}
    (x : SubsystemBasis I S) (r : ComplementBasis I S)
    (i : Fin n) (h : i ∈ S) : joinBasis x r i = x ⟨i, h⟩ := by
  simp [joinBasis, h]

/-- PDF p.5: infimum of the unit-vector energies, in a complete eigenbasis. -/
def spectralGround {I : Type} (energy : I → ℝ≥0) : ℝ :=
  sInf (Set.range (fun i => (energy i : ℝ)))

/-- PDF p.8, (22): extended partition trace and a right-hand limit.
Neither summability nor finite entropy is assumed separately. -/
def Condition22 {I : Type} (energy : I → ℝ≥0) : Prop :=
  Tendsto (fun lambda : ℝ =>
    (∑' i, ENNReal.ofReal (Real.exp (-lambda * (energy i : ℝ)))) ^ lambda)
    (nhdsWithin 0 (Set.Ioi 0)) (nhds (1 : ℝ≥0∞))

/-- Trace-distance uniform continuity on a specified state set (PDF pp.5,17--18).
This epsilon-delta definition avoids imposing a coefficient uniformity. -/
def TraceUniformContinuousOn {I : Type} (A : Set (State I)) (f : State I → ℝ) : Prop :=
  ∀ eta : ℝ, 0 < eta → ∃ delta : ℝ, 0 < delta ∧
    ∀ rho ∈ A, ∀ sigma ∈ A, nuclearStateDistance rho sigma < ENNReal.ofReal delta →
      |f rho - f sigma| < eta

/-- PDF p.4: genuine mutual information of a joint state, defined by relative entropy. -/
def marginalMutualInformation {I J : Type}
    (X : Matrix (I × J) (I × J) ℂ) : ℝ≥0∞ :=
  extendedRelativeEntropyNats X
    (QAlgorithms.Papers.Independent.Dependency50.P33.productCoeff
      (traceSecond X) (QAlgorithms.Papers.Independent.Frontier18.P01.traceFirst X))

/-- PDF p.11, (28): finite-A-entropy conditional entropy, without infinity subtraction. -/
def extendedConditionalEntropy {I J : Type}
    (X : Matrix (I × J) (I × J) ℂ) : ℝ :=
  (extendedEntropyNats (traceSecond X)).toReal - (marginalMutualInformation X).toReal

/-- PDF p.17, (48): mutual information after the actual reference-amplified channel. -/
def outputMutualInformation {I J R : Type}
    (Phi : QAlgorithms.Papers.Independent.Frontier18.P01.Channel I J)
    (X : Matrix (I × R) (I × R) ℂ) : ℝ≥0∞ :=
  marginalMutualInformation
    (QAlgorithms.Papers.Independent.Dependency50.P33.tensorApply Phi.val X)

/-- PDF p.8, (21): the real modulus, used only with proved finite F_H. -/
def entropyModulus (F : ℝ → ℝ≥0∞) (C D E eps : ℝ) : ℝ :=
  C * Real.sqrt (2 * eps) * (F (E / eps)).toReal + D * g (Real.sqrt (2 * eps))

/-- PDF pp.17--18: finite-valued uniform continuity and the quantitative bound,
on one FIXED energy ball. This is a conclusion package, never a hypothesis. -/
def EnergyContinuity {I : Type} (energyOn : Matrix I I ℂ → ℝ≥0∞)
    (f : Matrix I I ℂ → ℝ≥0∞) (E : ℝ) (bound : ℝ → ℝ) : Prop :=
  (∀ rho : State I, energyOn rho.coefficients ≤ ENNReal.ofReal E → f rho.coefficients < ⊤) ∧
  TraceUniformContinuousOn {rho | energyOn rho.coefficients ≤ ENNReal.ofReal E}
    (fun rho => (f rho.coefficients).toReal) ∧
  ∀ (rho sigma : State I) (eps : ℝ),
    energyOn rho.coefficients ≤ ENNReal.ofReal E →
    energyOn sigma.coefficients ≤ ENNReal.ofReal E →
    0 < eps → eps ≤ 1 / 2 → nuclearStateDistance rho sigma ≤ ENNReal.ofReal eps →
      |(f rho.coefficients).toReal - (f sigma.coefficients).toReal| ≤ bound eps

example {I : Type} (energy : I → ℝ≥0) :
    coefficientEnergy energy (0 : Matrix I I ℂ) = 0 := by simp [coefficientEnergy]

example {I : Type} (f : State I → ℝ) : TraceUniformContinuousOn ∅ f := by
  intro eta heta
  exact ⟨1, by norm_num, by simp⟩

example {I : Type} (energy : I → ℝ≥0) (E : ℝ) : entropySupremum energy E =
    ⨆ rho : {rho : State I // coefficientEnergy energy rho.coefficients ≤ ENNReal.ofReal E},
      extendedEntropyNats rho.val.coefficients := rfl

/-- PDF p.14, (38): relative entropy distance to the actual trace-norm closed
convex hull of product states. Empty infima retain infinity. The frozen P54
relativeEntanglement is finite-dimensional and in bits; this interface is in nats
on arbitrary countable systems and cannot use that finite-dimensional function. -/
def countableRelativeEntanglementNats {I J : Type}
    (X : Matrix (I × J) (I × J) ℂ) : ℝ≥0∞ :=
  ⨅ omega : {omega : State (I × J) //
    QAlgorithms.Papers.Independent.Dependency50.P33.IsSeparable omega.coefficients},
    extendedRelativeEntropyNats X omega.val.coefficients

/-- PDF pp.14,16: n copies, regrouped as A^n:B^n, without restricting other
states of that tensor-power system to be products. -/
def bipartiteTensorPower {I J : Type} (n : ℕ)
    (X : Matrix (I × J) (I × J) ℂ) :
    Matrix ((Fin n → I) × (Fin n → J)) ((Fin n → I) × (Fin n → J)) ℂ :=
  fun a b => ∏ k, X (a.1 k, a.2 k) (b.1 k, b.2 k)

/-- PDF p.14, (39): regularization as the extended infimum of ER(rho^n)/n,
equivalent to the actual limit by subadditivity. Only positive block lengths
are used; a divergent value remains infinity. The frozen P54 regularization
is finite-dimensional and in bits, so it does not represent this domain. -/
def regularizedCountableEntanglementNats {I J : Type}
    (X : Matrix (I × J) (I × J) ℂ) : ℝ≥0∞ :=
  ⨅ n : ℕ, countableRelativeEntanglementNats (bipartiteTensorPower (n + 1) X) / (n + 1 : ℝ≥0∞)

/-- PDF p.16: additive tensor-power Hamiltonian eigenvalues. -/
def additiveEnergy {I : Type} (energy : I → ℝ≥0) (n : ℕ) (a : Fin n → I) : ℝ≥0 :=
  ∑ k, energy (a k)

example {I J : Type} (X : Matrix (I × J) (I × J) ℂ)
    (a b : (Fin 0 → I) × (Fin 0 → J)) : bipartiteTensorPower 0 X a b = 1 := by
  simp [bipartiteTensorPower]

example {I : Type} (energy : I → ℝ≥0) (a : Fin 0 → I) : additiveEnergy energy 0 a = 0 := by
  simp [additiveEnergy]

end QAlgorithms.Papers.Independent.Support50.P18
Source
Adaptation of the Alicki-Fannes-Winter method for the set of states with bounded energy and its use, arXiv:1609.07044v5, https://arxiv.org/abs/1609.07044v5
Read-back

What the Lean code literally says, in plain math · gpt-6.1-sol

QAlgorithms.Papers.Independent.Support50.P18.AFWConstants

This structure consists of four independently specified nonnegative real numbers aaa, bbb, c−c_-c−​, and c+c_+c+​. Each may be zero; the structure imposes no further equations, inequalities, or positivity requirements.

QAlgorithms.Papers.Independent.Support50.P18.ComplementBasis

For every natural number nnn, every family of types (Ii)i=0n−1(I_i)_{i=0}^{n-1}(Ii​)i=0n−1​, and every finite subset S⊆{0,…,n−1}S\subseteq\{0,\ldots,n-1\}S⊆{0,…,n−1}, this abbreviates the type of assignments rrr that choose an element ri∈Iir_i\in I_iri​∈Ii​ for each index i∉Si\notin Si∈/S, namely ∏i∈{0,…,n−1}∖SIi\prod_{i\in\{0,\ldots,n-1\}\setminus S}I_i∏i∈{0,…,n−1}∖S​Ii​. If SSS contains every index, this type has exactly one element, the empty assignment, including when n=0n=0n=0; if some type IiI_iIi​ outside SSS is empty, there are no such assignments.

QAlgorithms.Papers.Independent.Support50.P18.Condition22

For any type III and nonnegative real energies eie_iei​, the condition means precisely

lim⁡λ→0+(∑i∈Ie−λei)λ=1in [0,∞].\lim_{\lambda\to0^+} \left(\sum_{i\in I}e^{-\lambda e_i}\right)^\lambda=1 \quad\text{in }[0,\infty].λ→0+lim​(i∈I∑​e−λei​)λ=1in [0,∞].

Each exponential is embedded into [0,∞][0,\infty][0,∞] using ofReal⁡(a)=max⁡(a,0)\operatorname{ofReal}(a)=\max(a,0)ofReal(a)=max(a,0), the sum is the extended nonnegative sum, and the exponent is the extended nonnegative real-power operation. No countability assumption or separate requirement that the sum be finite for every positive λ\lambdaλ appears in this condition. For empty III, the expression is zero for positive λ\lambdaλ, so the condition does not hold.

QAlgorithms.Papers.Independent.Support50.P18.EnergyContinuity

For any index type III, any functions e,f:CI×I→[0,∞]e,f:\mathbb C^{I\times I}\to[0,\infty]e,f:CI×I→[0,∞], any real EEE, and any function b:R→Rb:\mathbb R\to\mathbb Rb:R→R, this defines the following conjunction. Here a state is a matrix XXX admitting a representation Xij=∑n=0∞un(i)vn(j)‾X_{ij}=\sum_{n=0}^{\infty}u_n(i)\overline{v_n(j)}Xij​=∑n=0∞​un​(i)vn​(j)​, where un,vn∈ℓ2(I;C)u_n,v_n\in\ell^2(I;\mathbb C)un​,vn​∈ℓ2(I;C) and ∑n∥un∥∥vn∥<∞\sum_n\|u_n\|\|v_n\|<\infty∑n​∥un​∥∥vn​∥<∞, whose quadratic forms on every finite coordinate set are real and nonnegative, and whose real diagonal entries form a summable family with sum 111. Write τ(Z)\tau(Z)τ(Z) for the infimum, over all such representations of ZZZ, of ∑n∥un∥∥vn∥\sum_n\|u_n\|\|v_n\|∑n​∥un​∥∥vn​∥, with value +∞+\infty+∞ if no representation exists, and put d(X,Y)=τ(X−Y)/2d(X,Y)=\tau(X-Y)/2d(X,Y)=τ(X−Y)/2. For an extended nonnegative number aaa, write a♭a^\flata♭ for its real value when finite and 000 when a=+∞a=+\inftya=+∞, and put E+=max⁡(E,0)E_+=\max(E,0)E+​=max(E,0). The conjunction requires: f(X)<+∞f(X)<+\inftyf(X)<+∞ for every state XXX with e(X)≤E+e(X)\le E_+e(X)≤E+​; for every real η>0\eta>0η>0, there exists a real δ>0\delta>0δ>0 such that all states X,YX,YX,Y satisfying e(X),e(Y)≤E+e(X),e(Y)\le E_+e(X),e(Y)≤E+​ and d(X,Y)<δd(X,Y)<\deltad(X,Y)<δ satisfy ∣f(X)♭−f(Y)♭∣<η|f(X)^\flat-f(Y)^\flat|<\eta∣f(X)♭−f(Y)♭∣<η; and all states X,YX,YX,Y and real 0<ε≤120<\varepsilon\le\tfrac120<ε≤21​ satisfying e(X),e(Y)≤E+e(X),e(Y)\le E_+e(X),e(Y)≤E+​ and d(X,Y)≤εd(X,Y)\le\varepsilond(X,Y)≤ε satisfy ∣f(X)♭−f(Y)♭∣≤b(ε)|f(X)^\flat-f(Y)^\flat|\le b(\varepsilon)∣f(X)♭−f(Y)♭∣≤b(ε). No countability, nonemptiness, positivity, or regularity assumptions are imposed on I,e,f,bI,e,f,bI,e,f,b; a negative EEE gives the cutoff 000, and if there are no states satisfying the cutoff, all three requirements hold vacuously.

QAlgorithms.Papers.Independent.Support50.P18.OscillatorParameters.EntropyControlled

For every natural number ℓ\ellℓ, arbitrary type RRR, oscillator parameters OOO with ℓ>0\ell>0ℓ>0, ℏ>0\hbar>0ℏ>0, and ωi>0\omega_i>0ωi​>0, four nonnegative real constants C=(a,b,c−,c+)C=(a,b,c_-,c_+)C=(a,b,c−​,c+​), and any real-valued function fff on all complex matrices indexed by A×RA\times RA×R, where A=NℓA=\mathbb N^\ellA=Nℓ, this predicate imposes the following two conditions. A state ρ\rhoρ means a matrix admitting ρxy=∑nun(x)vn(y)‾\rho_{xy}=\sum_nu_n(x)\overline{v_n(y)}ρxy​=∑n​un​(x)vn​(y)​ for square-summable complex vectors with ∑n∥un∥∥vn∥<∞\sum_n\|u_n\|\|v_n\|<\infty∑n​∥un​∥∥vn​∥<∞, having real nonnegative quadratic forms on every finite coordinate set, and having summable real diagonal entries with sum one; write ρA(a,a′)=∑r∈Rρ(a,r),(a′,r)\rho_A(a,a')=\sum_{r\in R}\rho_{(a,r),(a',r)}ρA​(a,a′)=∑r∈R​ρ(a,r),(a′,r)​, and say its marginal energy is finite when ∑a∈A[∑iℏωi(ai+1/2)]max⁡(0,Re⁡(ρA)aa)<∞\sum_{a\in A}[\sum_i\hbar\omega_i(a_i+1/2)]\max(0,\operatorname{Re}(\rho_A)_{aa})<\infty∑a∈A​[∑i​ℏωi​(ai​+1/2)]max(0,Re(ρA​)aa​)<∞. First, for every two such states ρ,σ\rho,\sigmaρ,σ of finite marginal energy and every real 0<p<10<p<10<p<1,

−a h(p)≤f(pρ+(1−p)σ)−pf(ρ)−(1−p)f(σ)≤b h(p),h(p)=−plog⁡p−(1−p)log⁡(1−p).-a\,h(p)\le f\bigl(p\rho+(1-p)\sigma\bigr)-p f(\rho)-(1-p)f(\sigma) \le b\,h(p), \qquad h(p)=-p\log p-(1-p)\log(1-p).−ah(p)≤f(pρ+(1−p)σ)−pf(ρ)−(1−p)f(σ)≤bh(p),h(p)=−plogp−(1−p)log(1−p).

Second, every state ρ\rhoρ of finite marginal energy satisfies, in the extended real numbers,

−c−S(ρA)≤f(ρ)≤c+S(ρA),-c_-S(\rho_A)\le f(\rho)\le c_+S(\rho_A),−c−​S(ρA​)≤f(ρ)≤c+​S(ρA​),

where, for a matrix XXX on AAA,

S(X)=sup⁡s⊆A finitemax⁡ ⁣(0, Re⁡tr⁡(Xs)log⁡ ⁣(Re⁡tr⁡(Xs))−Re⁡tr⁡ ⁣(Xs(log⁡)cfc(Xs))).S(X)=\sup_{s\subseteq A\ {\rm finite}} \max\!\left(0,\, \operatorname{Re}\operatorname{tr}(X_s)\log\!\bigl(\operatorname{Re}\operatorname{tr}(X_s)\bigr) -\operatorname{Re}\operatorname{tr}\!\bigl(X_s(\log)_{\rm cfc}(X_s)\bigr)\right).S(X)=s⊆A finitesup​max(0,Retr(Xs​)log(Retr(Xs​))−Retr(Xs​(log)cfc​(Xs​))).

Here XsX_sXs​ is the coordinate compression, (log⁡)cfc(\log)_{\rm cfc}(log)cfc​ denotes continuous functional calculus of the total real logarithm, and the products c±Sc_\pm Sc±​S are formed in [0,∞][0,\infty][0,∞] before passage to extended real numbers, with 0⋅∞=00\cdot\infty=00⋅∞=0. The predicate itself does not require these entropies to be finite. There are no assumptions on fff away from the displayed conditions, and p=0,1p=0,1p=0,1 are excluded. No countability or nonemptiness assumption is placed on RRR; for empty RRR both conditions are vacuous because there are no states.

QAlgorithms.Papers.Independent.Support50.P18.OscillatorParameters.allocationEntropy

For every natural number ℓ\ellℓ, oscillator parameters OOO with ℓ>0\ell>0ℓ>0, ℏ>0\hbar>0ℏ>0, and ωi>0\omega_i>0ωi​>0, and every real vector e=(ei)i=0ℓ−1e=(e_i)_{i=0}^{\ell-1}e=(ei​)i=0ℓ−1​, put xi=ei/(ℏωi)−1/2x_i=e_i/(\hbar\omega_i)-1/2xi​=ei​/(ℏωi​)−1/2. This defines

O.allocationEntropy(e)=∑i=0ℓ−1((xi+1)log⁡(xi+1)−xilog⁡xi).O.\mathrm{allocationEntropy}(e) =\sum_{i=0}^{\ell-1}\bigl((x_i+1)\log(x_i+1)-x_i\log x_i\bigr).O.allocationEntropy(e)=i=0∑ℓ−1​((xi​+1)log(xi​+1)−xi​logxi​).

No condition that eee belong to an allocation set is imposed, so negative xix_ixi​ are included. The formula uses the total real logarithm, with log⁡0=0\log0=0log0=0 and log⁡x=log⁡∣x∣\log x=\log|x|logx=log∣x∣ for negative xxx; consequently the expression is defined also at xi=0x_i=0xi​=0 and xi=−1x_i=-1xi​=−1. No oscillator parameters exist for ℓ=0\ell=0ℓ=0.

QAlgorithms.Papers.Independent.Support50.P18.OscillatorParameters.allocations

For every natural number ℓ\ellℓ, oscillator parameters OOO consisting of ℓ>0\ell>0ℓ>0, ℏ>0\hbar>0ℏ>0, and ωi>0\omega_i>0ωi​>0 for i=0,…,ℓ−1i=0,\ldots,\ell-1i=0,…,ℓ−1, and every real EEE, this defines the set of real vectors e=(ei)i=0ℓ−1e=(e_i)_{i=0}^{\ell-1}e=(ei​)i=0ℓ−1​ satisfying

∑i=0ℓ−1ei=E,ei≥ℏωi2for every i.\sum_{i=0}^{\ell-1}e_i=E, \qquad e_i\ge\frac{\hbar\omega_i}{2}\quad\text{for every }i.i=0∑ℓ−1​ei​=E,ei​≥2ℏωi​​for every i.

Thus the set is empty if E<12∑iℏωiE<\frac12\sum_i\hbar\omega_iE<21​∑i​ℏωi​; at equality it consists of the vector ei=ℏωi/2e_i=\hbar\omega_i/2ei​=ℏωi​/2. The oscillator parameters exclude ℓ=0\ell=0ℓ=0.

QAlgorithms.Papers.Independent.Support50.P18.OscillatorParameters.apply

For every natural number ℓ\ellℓ, parameter object OOO consisting of ℓ>0\ell>0ℓ>0, ℏ>0\hbar>0ℏ>0, and positive nonnegative real frequencies ωi\omega_iωi​, and vector u∈ℓ2(Nℓ)u\in\ell^2(\mathbb N^\ell)u∈ℓ2(Nℓ) satisfying square summability of n↦LO(n)u(n)n\mapsto L_O(n)u(n)n↦LO​(n)u(n), where LO(n)=∑i<ℓℏωi(ni+12)L_O(n)=\sum_{i<\ell}\hbar\omega_i(n_i+\tfrac12)LO​(n)=∑i<ℓ​ℏωi​(ni​+21​), this defines a vector in ℓ2(Nℓ)\ell^2(\mathbb N^\ell)ℓ2(Nℓ) by

(Ou)(n)=(∑i=0ℓ−1ℏωi(ni+12))u(n).(Ou)(n)=\left(\sum_{i=0}^{\ell-1}\hbar\omega_i\left(n_i+\frac12\right)\right)u(n).(Ou)(n)=(i=0∑ℓ−1​ℏωi​(ni​+21​))u(n).

Its square summability is supplied by the domain condition on uuu; uuu need not be normalized.

QAlgorithms.Papers.Independent.Support50.P18.OscillatorParameters.continuityBound

For every natural number ℓ\ellℓ, oscillator parameters OOO with ℓ>0\ell>0ℓ>0, ℏ>0\hbar>0ℏ>0, and ωi>0\omega_i>0ωi​>0, nonnegative real constants C=(a,b,c−,c+)C=(a,b,c_-,c_+)C=(a,b,c−​,c+​), and arbitrary real E,εE,\varepsilonE,ε, put G=12∑iℏωiG=\frac12\sum_i\hbar\omega_iG=21​∑i​ℏωi​, Q=(∏iℏωi)1/ℓQ=(\prod_i\hbar\omega_i)^{1/\ell}Q=(∏i​ℏωi​)1/ℓ, and g(x)=(x+1)log⁡(x+1)−xlog⁡xg(x)=(x+1)\log(x+1)-x\log xg(x)=(x+1)log(x+1)−xlogx. This defines the real number

O.continuityBound(C,E,ε)=(c−+c+)2ε ℓ(log⁡ ⁣E/ε+GℓQ+1)+(a+b) g(2ε).O.\mathrm{continuityBound}(C,E,\varepsilon) =(c_-+c_+)\sqrt{2\varepsilon}\,\ell \left(\log\!\frac{E/\varepsilon+G}{\ell Q}+1\right) +(a+b)\,g(\sqrt{2\varepsilon}).O.continuityBound(C,E,ε)=(c−​+c+​)2ε​ℓ(logℓQE/ε+G​+1)+(a+b)g(2ε​).

No energy constraint or sign condition on EEE or ε\varepsilonε is imposed. Division by zero is zero, the real square root of a nonpositive number is zero, and the total real logarithm has log⁡0=0\log0=0log0=0 and log⁡x=log⁡∣x∣\log x=\log|x|logx=log∣x∣ for negative xxx; in particular the displayed bound equals zero whenever ε≤0\varepsilon\le0ε≤0. The parameters exclude ℓ=0\ell=0ℓ=0.

QAlgorithms.Papers.Independent.Support50.P18.OscillatorParameters.domain

For every natural number ℓ\ellℓ and parameter object OOO consisting of ℓ>0\ell>0ℓ>0, ℏ>0\hbar>0ℏ>0, and positive nonnegative real frequencies ωi\omega_iωi​, this defines a subset of the complex Hilbert space ℓ2(Nℓ)\ell^2(\mathbb N^\ell)ℓ2(Nℓ):

DO={u∈ℓ2(Nℓ):∑n∈Nℓ∣ (∑i=0ℓ−1ℏωi(ni+12))u(n)∣2<∞}.\mathcal D_O= \left\{u\in\ell^2(\mathbb N^\ell): \sum_{n\in\mathbb N^\ell} \left|\,\left(\sum_{i=0}^{\ell-1}\hbar\omega_i\left(n_i+\frac12\right)\right)u(n)\right|^2<\infty \right\}.DO​=⎩⎨⎧​u∈ℓ2(Nℓ):n∈Nℓ∑​​(i=0∑ℓ−1​ℏωi​(ni​+21​))u(n)​2<∞⎭⎬⎫​.

No normalization or nonzero condition is imposed on uuu; in particular, the zero vector belongs to this set.

QAlgorithms.Papers.Independent.Support50.P18.OscillatorParameters.entropyMaximum

For every natural number ℓ\ellℓ, parameter object OOO consisting of ℓ>0\ell>0ℓ>0, ℏ>0\hbar>0ℏ>0, and positive nonnegative real frequencies ωi\omega_iωi​, and every real number EEE, this defines

SO(E)=sup⁡X a state on Nℓ∑nLO(n)max⁡{Re⁡Xnn,0}≤max⁡{E,0} sup⁡s⊆Nℓs finitemax⁡ ⁣{tslog⁡ts−Re⁡tr⁡ ⁣(Xslog⁡(Xs)), 0}∈[0,+∞],S_O(E)= \sup_{\substack{X\text{ a state on }\mathbb N^\ell\\ \sum_n L_O(n)\max\{\operatorname{Re}X_{nn},0\}\le\max\{E,0\}}} \ \sup_{\substack{s\subseteq\mathbb N^\ell\\s\text{ finite}}} \max\!\left\{ t_s\log t_s-\operatorname{Re}\operatorname{tr}\!\left(X_s\log(X_s)\right),\,0 \right\} \quad\in[0,+\infty],SO​(E)=X a state on Nℓ∑n​LO​(n)max{ReXnn​,0}≤max{E,0}​sup​ s⊆Nℓs finite​sup​max{ts​logts​−Retr(Xs​log(Xs​)),0}∈[0,+∞],

where LO(n)=∑i<ℓℏωi(ni+12)L_O(n)=\sum_{i<\ell}\hbar\omega_i(n_i+\tfrac12)LO​(n)=∑i<ℓ​ℏωi​(ni​+21​), Xs=(Xab)a,b∈sX_s=(X_{ab})_{a,b\in s}Xs​=(Xab​)a,b∈s​, ts=Re⁡tr⁡Xst_s=\operatorname{Re}\operatorname{tr}X_sts​=RetrXs​, and log⁡(Xs)\log(X_s)log(Xs​) is defined by continuous functional calculus using the total natural logarithm, with log⁡0=0\log0=0log0=0. Here a state means a complex matrix XXX admitting sequences ur,vr∈ℓ2(Nℓ)u_r,v_r\in\ell^2(\mathbb N^\ell)ur​,vr​∈ℓ2(Nℓ) with ∑r∥ur∥∥vr∥<∞\sum_r\|u_r\|\|v_r\|<\infty∑r​∥ur​∥∥vr​∥<∞ and Xab=∑rur(a)vr(b)‾X_{ab}=\sum_r u_r(a)\overline{v_r(b)}Xab​=∑r​ur​(a)vr​(b)​, such that every finite quadratic form ∑a,b∈sza‾Xabzb\sum_{a,b\in s}\overline{z_a}X_{ab}z_b∑a,b∈s​za​​Xab​zb​ is real and nonnegative for every finite sss and every complex family zzz, and such that ∑aRe⁡Xaa\sum_a\operatorname{Re}X_{aa}∑a​ReXaa​ is summable and equals 111. Both suprema are taken in the extended nonnegative reals, with an empty supremum equal to zero; the finite set sss may be empty. Since LO(n)≥12∑iℏωi>0L_O(n)\ge\frac12\sum_i\hbar\omega_i>0LO​(n)≥21​∑i​ℏωi​>0, the admissible-state set is empty whenever E<12∑iℏωiE<\frac12\sum_i\hbar\omega_iE<21​∑i​ℏωi​, including every E≤0E\le0E≤0.

QAlgorithms.Papers.Independent.Support50.P18.OscillatorParameters.entropyUpper

For a natural number ℓ\ellℓ and oscillator parameters OOO, which include ℓ>0\ell>0ℓ>0, a positive nonnegative real number ℏ\hbarℏ, and positive nonnegative real frequencies ωi\omega_iωi​ for i=0,…,ℓ−1i=0,\ldots,\ell-1i=0,…,ℓ−1, put qi=ℏωiq_i=\hbar\omega_iqi​=ℏωi​, G=12∑iqiG=\frac12\sum_i q_iG=21​∑i​qi​, and Q=(∏iqi)1/ℓQ=(\prod_i q_i)^{1/\ell}Q=(∏i​qi​)1/ℓ. For every real EEE, the definition assigns the real number

O.entropyUpper(E)=ℓlog⁡ ⁣(E+GℓQ)+ℓ.O.\mathrm{entropyUpper}(E)=\ell\log\!\left(\frac{E+G}{\ell Q}\right)+\ell.O.entropyUpper(E)=ℓlog(ℓQE+G​)+ℓ.

There is no assumption that E≥GE\ge GE≥G or that E+G>0E+G>0E+G>0. The logarithm is the total real logarithm used in the code: it is zero at zero and equals log⁡∣x∣\log|x|log∣x∣ for negative xxx. Although ℓ\ellℓ is implicitly quantified over all natural numbers, an OOO cannot exist when ℓ=0\ell=0ℓ=0.

QAlgorithms.Papers.Independent.Support50.P18.OscillatorParameters.finiteMarginalEnergy

For every natural number ℓ\ellℓ, arbitrary type RRR, oscillator parameters OOO with ℓ>0\ell>0ℓ>0, ℏ>0\hbar>0ℏ>0, and ωi>0\omega_i>0ωi​>0, and state ρ\rhoρ on A×RA\times RA×R, where A=NℓA=\mathbb N^\ellA=Nℓ, this predicate means

∑a∈A(∑i=0ℓ−1ℏωi(ai+1/2))max⁡ ⁣(0,Re⁡∑r∈Rρ(a,r),(a,r))<∞.\sum_{a\in A} \left(\sum_{i=0}^{\ell-1}\hbar\omega_i(a_i+1/2)\right) \max\!\left(0,\operatorname{Re}\sum_{r\in R}\rho_{(a,r),(a,r)}\right)<\infty.a∈A∑​(i=0∑ℓ−1​ℏωi​(ai​+1/2))max(0,Rer∈R∑​ρ(a,r),(a,r)​)<∞.

The inner sum is the diagonal of the matrix obtained by summing over the second subsystem. A state here is a complex matrix having an expansion ρxy=∑nun(x)vn(y)‾\rho_{xy}=\sum_nu_n(x)\overline{v_n(y)}ρxy​=∑n​un​(x)vn​(y)​ with square-summable complex vectors un,vnu_n,v_nun​,vn​ and ∑n∥un∥∥vn∥<∞\sum_n\|u_n\|\|v_n\|<\infty∑n​∥un​∥∥vn​∥<∞, having real nonnegative quadratic forms on every finite coordinate set, and having summable real diagonal entries with sum one. No countability or nonemptiness assumption is made on RRR; if RRR is empty, no such state exists. The sum over RRR uses the total complex infinite sum, which is zero if its summands are not summable.

QAlgorithms.Papers.Independent.Support50.P18.OscillatorParameters.geometricMean

For every natural number ℓ\ellℓ and parameter object OOO consisting of ℓ>0\ell>0ℓ>0, ℏ>0\hbar>0ℏ>0, and positive nonnegative real frequencies ωi\omega_iωi​, this defines the real number

GO=(∏i=0ℓ−1ℏωi)1/ℓ,G_O=\left(\prod_{i=0}^{\ell-1}\hbar\omega_i\right)^{1/\ell},GO​=(i=0∏ℓ−1​ℏωi​)1/ℓ,

using real exponentiation. The parameter object requires ℓ>0\ell>0ℓ>0, so the reciprocal exponent does not involve division by zero.

QAlgorithms.Papers.Independent.Support50.P18.OscillatorParameters.ground

For every natural number ℓ\ellℓ and parameter object OOO consisting of ℓ>0\ell>0ℓ>0, ℏ>0\hbar>0ℏ>0, and positive nonnegative real frequencies ωi\omega_iωi​, this defines the real number

EO 0=12∑i=0ℓ−1ℏωi.E_O^{\,0}=\frac12\sum_{i=0}^{\ell-1}\hbar\omega_i.EO0​=21​i=0∑ℓ−1​ℏωi​.

QAlgorithms.Papers.Independent.Support50.P18.OscillatorParameters.level

For every natural number ℓ\ellℓ, parameter object OOO consisting of ℓ>0\ell>0ℓ>0, ℏ>0\hbar>0ℏ>0, and positive nonnegative real frequencies ωi\omega_iωi​, and every occupation tuple n=(ni)i<ℓ∈Nℓn=(n_i)_{i<\ell}\in\mathbb N^\elln=(ni​)i<ℓ​∈Nℓ, this defines the nonnegative real number

LO(n)=∑i=0ℓ−1ℏωi(ni+12).L_O(n)=\sum_{i=0}^{\ell-1}\hbar\omega_i\left(n_i+\frac12\right).LO​(n)=i=0∑ℓ−1​ℏωi​(ni​+21​).

The tuple with every ni=0n_i=0ni​=0 is included.

QAlgorithms.Papers.Independent.Support50.P18.OscillatorParameters

For each natural number ℓ\ellℓ, this structure consists of a proof that ℓ>0\ell>0ℓ>0, a nonnegative real number ℏ\hbarℏ with a proof that ℏ>0\hbar>0ℏ>0, and a family of nonnegative real numbers ωi\omega_iωi​, indexed by i=0,…,ℓ−1i=0,\ldots,\ell-1i=0,…,ℓ−1, with proofs that every ωi>0\omega_i>0ωi​>0. Thus no such parameter object exists when ℓ=0\ell=0ℓ=0.

QAlgorithms.Papers.Independent.Support50.P18.OscillatorParameters.meanEnergy

For every natural number ℓ\ellℓ, parameter object OOO consisting of ℓ>0\ell>0ℓ>0, ℏ>0\hbar>0ℏ>0, and positive nonnegative real frequencies ωi\omega_iωi​, and arbitrary complex matrix XXX indexed by Nℓ×Nℓ\mathbb N^\ell\times\mathbb N^\ellNℓ×Nℓ, this defines

EO(X)=∑n∈Nℓ(∑i=0ℓ−1ℏωi(ni+12))max⁡{Re⁡Xnn,0}∈[0,+∞].\mathcal E_O(X)= \sum_{n\in\mathbb N^\ell} \left(\sum_{i=0}^{\ell-1}\hbar\omega_i\left(n_i+\frac12\right)\right) \max\{\operatorname{Re}X_{nn},0\} \quad\in[0,+\infty].EO​(X)=n∈Nℓ∑​(i=0∑ℓ−1​ℏωi​(ni​+21​))max{ReXnn​,0}∈[0,+∞].

The matrix is not required to be positive, normalized, or trace class. Negative real diagonal entries contribute zero, and the sum may be infinite.

QAlgorithms.Papers.Independent.Support50.P18.OscillatorParameters.quantum

For every natural number ℓ\ellℓ, parameter object OOO consisting of ℓ>0\ell>0ℓ>0, ℏ>0\hbar>0ℏ>0, and positive nonnegative real frequencies ω0,…,ωℓ−1\omega_0,\ldots,\omega_{\ell-1}ω0​,…,ωℓ−1​, and every index i<ℓi<\elli<ℓ, this defines the nonnegative real number Qi=ℏωiQ_i=\hbar\omega_iQi​=ℏωi​.

QAlgorithms.Papers.Independent.Support50.P18.SubsystemBasis

For every natural number nnn, every family of types (Ii)i=0n−1(I_i)_{i=0}^{n-1}(Ii​)i=0n−1​, and every finite subset S⊆{0,…,n−1}S\subseteq\{0,\ldots,n-1\}S⊆{0,…,n−1}, this abbreviates the type of assignments xxx that choose an element xi∈Iix_i\in I_ixi​∈Ii​ for each index i∈Si\in Si∈S, namely the dependent product ∏i∈SIi\prod_{i\in S}I_i∏i∈S​Ii​. If SSS is empty, this type has exactly one element, the empty assignment, including when n=0n=0n=0; if some selected type IiI_iIi​ is empty, there are no such assignments.

QAlgorithms.Papers.Independent.Support50.P18.TraceUniformContinuousOn

For any type III, set AAA of states indexed by III, and real-valued function fff on those states, this property says that for every η>0\eta>0η>0 there is a δ>0\delta>0δ>0 such that, for all ρ,σ∈A\rho,\sigma\in Aρ,σ∈A,

τ(ρ−σ)2<ofReal⁡(δ)⟹∣f(ρ)−f(σ)∣<η.\frac{\tau(\rho-\sigma)}2<\operatorname{ofReal}(\delta) \quad\Longrightarrow\quad |f(\rho)-f(\sigma)|<\eta.2τ(ρ−σ)​<ofReal(δ)⟹∣f(ρ)−f(σ)∣<η.

Here a state is a matrix with an expansion Xij=∑a≥0ua(i)va(j)‾X_{ij}=\sum_{a\ge0}u_a(i)\overline{v_a(j)}Xij​=∑a≥0​ua​(i)va​(j)​, where ua,va∈ℓ2(I;C)u_a,v_a\in\ell^2(I;\mathbb C)ua​,va​∈ℓ2(I;C) and ∑a∥ua∥∥va∥<∞\sum_a\|u_a\|\|v_a\|<\infty∑a​∥ua​∥∥va​∥<∞, whose finite quadratic forms are real and nonnegative and whose real diagonal entries are summable with sum one. The quantity τ(X)\tau(X)τ(X) is the infimum, over all such expansions of XXX, of ofReal⁡(∑a∥ua∥∥va∥)\operatorname{ofReal}(\sum_a\|u_a\|\|v_a\|)ofReal(∑a​∥ua​∥∥va​∥); an empty family of expansions gives +∞+\infty+∞. Both the distance and its comparison are in [0,∞][0,\infty][0,∞], and ofReal⁡(a)=max⁡(a,0)\operatorname{ofReal}(a)=\max(a,0)ofReal(a)=max(a,0). No countability assumption is imposed, and the property holds vacuously when AAA is empty.

QAlgorithms.Papers.Independent.Support50.P18.additiveEnergy

For any index type III, any function e:I→[0,∞)e:I\to[0,\infty)e:I→[0,∞), any natural number nnn, and any tuple a:{0,…,n−1}→Ia:\{0,\ldots,n-1\}\to Ia:{0,…,n−1}→I, this defines its additive energy, as a finite nonnegative real number, by

e(n)(a)=∑k=0n−1e(ak).e^{(n)}(a)=\sum_{k=0}^{n-1}e(a_k).e(n)(a)=k=0∑n−1​e(ak​).

No countability or nonemptiness assumption is imposed on III. For n=0n=0n=0, the unique empty tuple has energy 000.

QAlgorithms.Papers.Independent.Support50.P18.bipartiteTensorPower

For arbitrary index types I,JI,JI,J, any natural number nnn, and any complex matrix XXX indexed by I×JI\times JI×J, this defines a matrix indexed by In×JnI^n\times J^nIn×Jn by

X(a,c),(b,d)[n]=∏k=0n−1X(ak,ck),(bk,dk),a,b:{0,…,n−1}→I,c,d:{0,…,n−1}→J.X^{[n]}_{(a,c),(b,d)} =\prod_{k=0}^{n-1}X_{(a_k,c_k),(b_k,d_k)}, \qquad a,b:\{0,\ldots,n-1\}\to I,\quad c,d:\{0,\ldots,n-1\}\to J.X(a,c),(b,d)[n]​=k=0∏n−1​X(ak​,ck​),(bk​,dk​)​,a,b:{0,…,n−1}→I,c,d:{0,…,n−1}→J.

There are no state, positivity, normalization, finiteness, or countability assumptions. For n=0n=0n=0, both function spaces are singletons and the empty product makes this the one by one matrix with entry 111, regardless of XXX and even if III or JJJ is empty.

QAlgorithms.Papers.Independent.Support50.P18.countableRelativeEntanglementNats

For arbitrary index types I,JI,JI,J and an arbitrary complex matrix XXX indexed by I×JI\times JI×J, this defines

ER(X)=inf⁡ωD(X∥ω),E_{\mathrm R}(X)=\inf_{\omega}D(X\|\omega),ER​(X)=ωinf​D(X∥ω),

where the infimum ranges over the separable states described below; no countability, positivity, normalization, or trace class assumption is imposed on XXX. On any index set LLL, a trace class expansion is Aij=∑nun(i)vn(j)‾A_{ij}=\sum_nu_n(i)\overline{v_n(j)}Aij​=∑n​un​(i)vn​(j)​, where un,vn∈ℓ2(L;C)u_n,v_n\in\ell^2(L;\mathbb C)un​,vn​∈ℓ2(L;C) and ∑n∥un∥∥vn∥<∞\sum_n\|u_n\|\|v_n\|<\infty∑n​∥un​∥∥vn​∥<∞; let τ(A)\tau(A)τ(A) be the infimum of these costs, or +∞+\infty+∞ if there is no expansion. A state is a matrix with such an expansion, real nonnegative quadratic forms on every finite coordinate set, and summable real diagonal entries of sum 111. A state ω\omegaω on I×JI\times JI×J is separable here precisely when, for every real δ>0\delta>0δ>0, there is a finite real convex combination Y=∑r=1mpr(Ar⊗Br)Y=\sum_{r=1}^{m}p_r(A_r\otimes B_r)Y=∑r=1m​pr​(Ar​⊗Br​), with pr≥0p_r\ge0pr​≥0, ∑rpr=1\sum_rp_r=1∑r​pr​=1, and states Ar,BrA_r,B_rAr​,Br​ on I,JI,JI,J, such that τ(ω−Y)<δ\tau(\omega-Y)<\deltaτ(ω−Y)<δ; product entries are (Ar⊗Br)(i,a),(j,b)=(Ar)ij(Br)ab(A_r\otimes B_r)_{(i,a),(j,b)}=(A_r)_{ij}(B_r)_{ab}(Ar​⊗Br​)(i,a),(j,b)​=(Ar​)ij​(Br​)ab​. The relative quantity is

D(X∥ω)=sup⁡s⊆I×Js finite{[Re⁡tr⁡(Xslog⁡Xs−Xslog⁡ωs+ωs−Xs)]+,ker⁡ωs⊆ker⁡Xs,+∞,otherwise,D(X\|\omega)=\sup_{\substack{s\subseteq I\times J\\s\ \mathrm{finite}}} \begin{cases} [\operatorname{Re}\operatorname{tr}(X_s\log X_s-X_s\log\omega_s+\omega_s-X_s)]_+, &\ker\omega_s\subseteq\ker X_s,\\ +\infty,&\text{otherwise}, \end{cases}D(X∥ω)=s⊆I×Js finite​sup​{[Retr(Xs​logXs​−Xs​logωs​+ωs​−Xs​)]+​,+∞,​kerωs​⊆kerXs​,otherwise,​

where compressions are to the same coordinate set sss, logarithms use continuous functional calculus with log⁡0=0\log0=0log0=0, and [a]+=max⁡(a,0)[a]_+=\max(a,0)[a]+​=max(a,0). The supremum includes s=∅s=\varnothings=∅. If there is no separable state, the defining infimum is +∞+\infty+∞; in particular this occurs when III or JJJ is empty.

QAlgorithms.Papers.Independent.Support50.P18.entropyCombination

For all natural numbers n,mn,mn,m, types (Ii)i<n(I_i)_{i<n}(Ii​)i<n​, subsets Sk⊆{0,…,n−1}S_k\subseteq\{0,\ldots,n-1\}Sk​⊆{0,…,n−1} and real coefficients ckc_kck​ for k<mk<mk<m, and an arbitrary complex matrix XXX indexed by ∏i<nIi\prod_{i<n}I_i∏i<n​Ii​, define

ES,c(X)=∑k<mck r⁡ ⁣(H(XSk)).\mathcal E_{S,c}(X)=\sum_{k<m}c_k\,\operatorname{r}\!\left(H(X_{S_k})\right).ES,c​(X)=k<m∑​ck​r(H(XSk​​)).

Here XS(x,y)=∑rX(x⊔Sr,y⊔Sr)X_S(x,y)=\sum_r X(x\sqcup_S r,y\sqcup_S r)XS​(x,y)=∑r​X(x⊔S​r,y⊔S​r), with rrr ranging over complementary assignments and joining performed coordinatewise; a nonsummable complex sum has value zero. For any matrix AAA indexed by a type LLL,

H(A)=sup⁡s⊆Ls finiteofReal⁡ ⁣(Re⁡tr⁡(As) log⁡ ⁣(Re⁡tr⁡(As))−Re⁡tr⁡ ⁣(Aslog⁡fcAs)),H(A)= \sup_{\substack{s\subseteq L\\s\text{ finite}}} \operatorname{ofReal}\!\left( \operatorname{Re}\operatorname{tr}(A_s)\, \log\!\bigl(\operatorname{Re}\operatorname{tr}(A_s)\bigr) -\operatorname{Re}\operatorname{tr}\!\bigl(A_s\log_{\mathrm{fc}}A_s\bigr) \right),H(A)=s⊆Ls finite​sup​ofReal(Retr(As​)log(Retr(As​))−Retr(As​logfc​As​)),

where AsA_sAs​ is its restriction to s×ss\times ss×s, log⁡fc\log_{\mathrm{fc}}logfc​ is the continuous functional calculus applied to the real logarithm, and ofReal⁡(a)=max⁡(a,0)\operatorname{ofReal}(a)=\max(a,0)ofReal(a)=max(a,0), viewed in [0,∞][0,\infty][0,∞]. The conversion r⁡\operatorname{r}r returns the corresponding real number for a finite extended value and returns zero at +∞+\infty+∞. Thus this definition does not require any entropy to be finite or any matrix to be a density matrix; coefficients may have either sign, and m=0m=0m=0 gives zero.

QAlgorithms.Papers.Independent.Support50.P18.entropyModulus

For any function F:R→[0,∞]F:\mathbb R\to[0,\infty]F:R→[0,∞] and arbitrary real numbers C,D,E,εC,D,E,\varepsilonC,D,E,ε, put s=2εs=\sqrt{2\varepsilon}s=2ε​. The entropy modulus is the real number

C s r⁡ ⁣(F(E/ε))+D((s+1)log⁡(s+1)−slog⁡s),C\,s\,\operatorname{r}\!\left(F(E/\varepsilon)\right) +D\bigl((s+1)\log(s+1)-s\log s\bigr),Csr(F(E/ε))+D((s+1)log(s+1)−slogs),

where r⁡\operatorname{r}r converts finite extended values to their real values and converts +∞+\infty+∞ to zero. No sign, finiteness, or regularity assumptions are imposed on the arguments. Division is total, so E/0=0E/0=0E/0=0, the real logarithm satisfies log⁡0=0\log0=0log0=0, and the real square root is zero on negative arguments; consequently the modulus is zero whenever ε≤0\varepsilon\le0ε≤0.

QAlgorithms.Papers.Independent.Support50.P18.extendedConditionalEntropy

For any types I,JI,JI,J and arbitrary complex matrix XXX indexed by I×JI\times JI×J, define its marginals Aij=∑r∈JX(i,r),(j,r)A_{ij}=\sum_{r\in J}X_{(i,r),(j,r)}Aij​=∑r∈J​X(i,r),(j,r)​ and Bab=∑r∈IX(r,a),(r,b)B_{ab}=\sum_{r\in I}X_{(r,a),(r,b)}Bab​=∑r∈I​X(r,a),(r,b)​, and put Z(i,a),(j,b)=AijBabZ_{(i,a),(j,b)}=A_{ij}B_{ab}Z(i,a),(j,b)​=Aij​Bab​. The extended conditional entropy is the real number

r⁡(H(A))−r⁡(D(X∥Z)),\operatorname{r}(H(A))-\operatorname{r}(D(X\|Z)),r(H(A))−r(D(X∥Z)),

where

H(U)=sup⁡s finiteofReal⁡ ⁣(Re⁡tr⁡(Us)log⁡ ⁣(Re⁡tr⁡(Us))−Re⁡tr⁡(Uslog⁡fcUs))H(U)=\sup_{s\text{ finite}} \operatorname{ofReal}\!\left( \operatorname{Re}\operatorname{tr}(U_s)\log\!\bigl(\operatorname{Re}\operatorname{tr}(U_s)\bigr) -\operatorname{Re}\operatorname{tr}(U_s\log_{\mathrm{fc}}U_s) \right)H(U)=s finitesup​ofReal(Retr(Us​)log(Retr(Us​))−Retr(Us​logfc​Us​))

and

D(U∥V)=sup⁡s finite{ofReal⁡ ⁣(Re⁡tr⁡(Uslog⁡fcUs−Uslog⁡fcVs+Vs−Us)),ker⁡Vs⊆ker⁡Us,+∞,otherwise.D(U\|V)=\sup_{s\text{ finite}} \begin{cases} \operatorname{ofReal}\!\left( \operatorname{Re}\operatorname{tr} (U_s\log_{\mathrm{fc}}U_s-U_s\log_{\mathrm{fc}}V_s+V_s-U_s) \right),&\ker V_s\subseteq\ker U_s,\\ +\infty,&\text{otherwise}. \end{cases}D(U∥V)=s finitesup​{ofReal(Retr(Us​logfc​Us​−Us​logfc​Vs​+Vs​−Us​)),+∞,​kerVs​⊆kerUs​,otherwise.​

Finite coordinate restrictions are denoted by subscripts, log⁡fc\log_{\mathrm{fc}}logfc​ denotes continuous functional calculus of the real logarithm, and ofReal⁡(a)=max⁡(a,0)∈[0,∞]\operatorname{ofReal}(a)=\max(a,0)\in[0,\infty]ofReal(a)=max(a,0)∈[0,∞]. The conversion r⁡\operatorname{r}r sends finite extended values to their real values and sends +∞+\infty+∞ to zero, separately in both terms. No state or summability assumption is imposed, and nonsummable complex marginal sums are assigned zero.

QAlgorithms.Papers.Independent.Support50.P18.joinBasis

For every natural number nnn, family of types (Ii)i<n(I_i)_{i<n}(Ii​)i<n​, and subset S⊆{0,…,n−1}S\subseteq\{0,\ldots,n-1\}S⊆{0,…,n−1}, let xxx assign an element of IiI_iIi​ to each i∈Si\in Si∈S, and let rrr assign an element of IiI_iIi​ to each i∉Si\notin Si∈/S. Their joined assignment x⊔Sr∈∏i<nIix\sqcup_S r\in\prod_{i<n}I_ix⊔S​r∈∏i<n​Ii​ is defined coordinatewise by

(x⊔Sr)i={xi,i∈S,ri,i∉S.(x\sqcup_S r)_i= \begin{cases} x_i,&i\in S,\\ r_i,&i\notin S. \end{cases}(x⊔S​r)i​={xi​,ri​,​i∈S,i∈/S.​

When SSS is empty this is the assignment supplied by rrr; when SSS contains every index it is the assignment supplied by xxx. The definition includes n=0n=0n=0, when the resulting assignment is the unique empty assignment.

QAlgorithms.Papers.Independent.Support50.P18.logPowerLevel

For every real number qqq and natural number nnn, this defines the real number Lq(n)=(log⁡(n+1))qL_q(n)=(\log(n+1))^qLq​(n)=(log(n+1))q, where log⁡\loglog is the natural logarithm and the exponent denotes real power. The index n=0n=0n=0 is included: L0(0)=1L_0(0)=1L0​(0)=1, while Lq(0)=0L_q(0)=0Lq​(0)=0 for every q≠0q\ne0q=0, including negative qqq, under the total real-power convention.

QAlgorithms.Papers.Independent.Support50.P18.logPowerPartition

For every pair of real numbers q,λq,\lambdaq,λ, this defines the extended nonnegative real number

Zq(λ)=∑n=0∞exp⁡ ⁣(−λ(log⁡(n+1))q).Z_q(\lambda)=\sum_{n=0}^{\infty}\exp\!\left(-\lambda(\log(n+1))^q\right).Zq​(λ)=n=0∑∞​exp(−λ(log(n+1))q).

The sum may equal +∞+\infty+∞; neither convergence nor positivity of λ\lambdaλ is required. Real powers use the convention 00=10^0=100=1 and 0q=00^q=00q=0 for q≠0q\ne0q=0, so the n=0n=0n=0 summand is e−λe^{-\lambda}e−λ when q=0q=0q=0 and 111 otherwise.

QAlgorithms.Papers.Independent.Support50.P18.marginalMutualInformation

For any types I,JI,JI,J and arbitrary complex matrix XXX indexed by I×JI\times JI×J, define

Aij=∑r∈JX(i,r),(j,r),Bab=∑r∈IX(r,a),(r,b),Z(i,a),(j,b)=AijBab.A_{ij}=\sum_{r\in J}X_{(i,r),(j,r)},\qquad B_{ab}=\sum_{r\in I}X_{(r,a),(r,b)},\qquad Z_{(i,a),(j,b)}=A_{ij}B_{ab}.Aij​=r∈J∑​X(i,r),(j,r)​,Bab​=r∈I∑​X(r,a),(r,b)​,Z(i,a),(j,b)​=Aij​Bab​.

The marginal mutual information is D(X∥Z)D(X\|Z)D(X∥Z), where

D(U∥V)=sup⁡s finite{ofReal⁡ ⁣(Re⁡tr⁡(Uslog⁡fcUs−Uslog⁡fcVs+Vs−Us)),ker⁡Vs⊆ker⁡Us,+∞,otherwise.D(U\|V)= \sup_{s\text{ finite}} \begin{cases} \operatorname{ofReal}\!\left( \operatorname{Re}\operatorname{tr} (U_s\log_{\mathrm{fc}}U_s-U_s\log_{\mathrm{fc}}V_s+V_s-U_s) \right),&\ker V_s\subseteq\ker U_s,\\ +\infty,&\text{otherwise}. \end{cases}D(U∥V)=s finitesup​{ofReal(Retr(Us​logfc​Us​−Us​logfc​Vs​+Vs​−Us​)),+∞,​kerVs​⊆kerUs​,otherwise.​

Here sss ranges over finite subsets of the common matrix index type, the subscript denotes coordinate restriction, log⁡fc\log_{\mathrm{fc}}logfc​ is continuous functional calculus of the real logarithm, and ofReal⁡(a)=max⁡(a,0)∈[0,∞]\operatorname{ofReal}(a)=\max(a,0)\in[0,\infty]ofReal(a)=max(a,0)∈[0,∞]. No countability, positivity, normalization, or summability assumption is required; the complex marginal sums return zero when they are not summable, including when their indexing type is empty.

QAlgorithms.Papers.Independent.Support50.P18.nuclearStateDistance

For every type III and states ρ,σ\rho,\sigmaρ,σ on III, this defines the extended nonnegative number

d(ρ,σ)=12 T(ρ−σ),d(\rho,\sigma)=\frac12\,\mathcal T(\rho-\sigma),d(ρ,σ)=21​T(ρ−σ),

where

T(X)=inf⁡{∑n=0∞∥un∥∥vn∥:un,vn∈ℓ2(I;C), ∑n∥un∥∥vn∥<∞, Xij=∑nun(i)vn(j)‾ for all i,j}.\mathcal T(X)= \inf\left\{ \sum_{n=0}^{\infty}\|u_n\|\|v_n\|: u_n,v_n\in\ell^2(I;\mathbb C),\ \sum_n\|u_n\|\|v_n\|<\infty,\ X_{ij}=\sum_nu_n(i)\overline{v_n(j)}\ \text{for all }i,j \right\}.T(X)=inf{n=0∑∞​∥un​∥∥vn​∥:un​,vn​∈ℓ2(I;C), n∑​∥un​∥∥vn​∥<∞, Xij​=n∑​un​(i)vn​(j)​ for all i,j}.

Each real cost is embedded in [0,∞][0,\infty][0,∞] by taking its nonnegative part, and the infimum of an empty family is ∞\infty∞. A state means a matrix admitting at least one expansion of this form, whose quadratic form on every finite coordinate set is real and nonnegative, and whose real diagonal entries are summable with sum one. No countability or nonemptiness assumption is imposed on III; when III is empty there are no states to supply as arguments.

QAlgorithms.Papers.Independent.Support50.P18.outputMutualInformation

For any types I,J,RI,J,RI,J,R, channel Φ\PhiΦ from matrices indexed by III to matrices indexed by JJJ, and arbitrary complex matrix XXX indexed by I×RI\times RI×R, first define

Y(j,r),(j′,r′)=Φ ⁣((X(i,r),(i′,r′))i,i′∈I)j,j′.Y_{(j,r),(j',r')}= \Phi\!\left((X_{(i,r),(i',r')})_{i,i'\in I}\right)_{j,j'}.Y(j,r),(j′,r′)​=Φ((X(i,r),(i′,r′)​)i,i′∈I​)j,j′​.

The output mutual information is D(Y∥Z)D(Y\|Z)D(Y∥Z), where Ajj′=∑r∈RY(j,r),(j′,r)A_{jj'}=\sum_{r\in R}Y_{(j,r),(j',r)}Ajj′​=∑r∈R​Y(j,r),(j′,r)​, Brr′=∑j∈JY(j,r),(j,r′)B_{rr'}=\sum_{j\in J}Y_{(j,r),(j,r')}Brr′​=∑j∈J​Y(j,r),(j,r′)​, and Z(j,r),(j′,r′)=Ajj′Brr′Z_{(j,r),(j',r')}=A_{jj'}B_{rr'}Z(j,r),(j′,r′)​=Ajj′​Brr′​. For matrices U,VU,VU,V, the quantity D(U∥V)D(U\|V)D(U∥V) is the supremum over finite coordinate sets sss of

{ofReal⁡ ⁣(Re⁡tr⁡(Uslog⁡fcUs−Uslog⁡fcVs+Vs−Us)),ker⁡Vs⊆ker⁡Us,+∞,otherwise,\begin{cases} \operatorname{ofReal}\!\left( \operatorname{Re}\operatorname{tr} (U_s\log_{\mathrm{fc}}U_s-U_s\log_{\mathrm{fc}}V_s+V_s-U_s) \right),&\ker V_s\subseteq\ker U_s,\\ +\infty,&\text{otherwise}, \end{cases}{ofReal(Retr(Us​logfc​Us​−Us​logfc​Vs​+Vs​−Us​)),+∞,​kerVs​⊆kerUs​,otherwise,​

with log⁡fc\log_{\mathrm{fc}}logfc​ denoting continuous functional calculus of the real logarithm and ofReal⁡(a)=max⁡(a,0)∈[0,∞]\operatorname{ofReal}(a)=\max(a,0)\in[0,\infty]ofReal(a)=max(a,0)∈[0,∞]. The channel assumption means that Φ\PhiΦ preserves trace-class matrices, is additive on pairs of trace-class matrices and complex homogeneous on trace-class matrices, and, for every natural number qqq, its corresponding slice action sends every density matrix indexed by I×{0,…,q−1}I\times\{0,\ldots,q-1\}I×{0,…,q−1} to a density matrix indexed by J×{0,…,q−1}J\times\{0,\ldots,q-1\}J×{0,…,q−1}. Here trace class means admitting an expansion Uab=∑ℓ≥0uℓ(a)vℓ(b)‾U_{ab}=\sum_{\ell\ge0}u_\ell(a)\overline{v_\ell(b)}Uab​=∑ℓ≥0​uℓ​(a)vℓ​(b)​ with square-summable vectors and ∑ℓ∥uℓ∥∥vℓ∥<∞\sum_\ell\|u_\ell\|\|v_\ell\|<\infty∑ℓ​∥uℓ​∥∥vℓ​∥<∞; a density matrix additionally has real, nonnegative finite quadratic forms and summable real diagonal entries summing to one. No countability or state assumption is imposed on XXX, and nonsummable complex marginal sums have value zero.

QAlgorithms.Papers.Independent.Support50.P18.regularizedCountableEntanglementNats

For arbitrary index types I,JI,JI,J and any complex matrix XXX indexed by I×JI\times JI×J, this defines

ERreg(X)=inf⁡m≥1ER(X[m])m,X(a,c),(b,d)[m]=∏k=0m−1X(ak,ck),(bk,dk),E_{\mathrm R}^{\mathrm{reg}}(X) =\inf_{m\ge1}\frac{E_{\mathrm R}(X^{[m]})}{m}, \qquad X^{[m]}_{(a,c),(b,d)} =\prod_{k=0}^{m-1}X_{(a_k,c_k),(b_k,d_k)},ERreg​(X)=m≥1inf​mER​(X[m])​,X(a,c),(b,d)[m]​=k=0∏m−1​X(ak​,ck​),(bk​,dk​)​,

with a,b∈Ima,b\in I^ma,b∈Im, c,d∈Jmc,d\in J^mc,d∈Jm, and arithmetic in [0,∞][0,\infty][0,∞]. Here ER(A)=inf⁡ωD(A∥ω)E_{\mathrm R}(A)=\inf_\omega D(A\|\omega)ER​(A)=infω​D(A∥ω), over separable states on the relevant product index set. To specify these terms, a trace class matrix on LLL admits an expansion Aij=∑nun(i)vn(j)‾A_{ij}=\sum_nu_n(i)\overline{v_n(j)}Aij​=∑n​un​(i)vn​(j)​ with un,vn∈ℓ2(L;C)u_n,v_n\in\ell^2(L;\mathbb C)un​,vn​∈ℓ2(L;C) and ∑n∥un∥∥vn∥<∞\sum_n\|u_n\|\|v_n\|<\infty∑n​∥un​∥∥vn​∥<∞; its size τ(A)\tau(A)τ(A) is the infimum of expansion costs, with empty infimum +∞+\infty+∞. A state is such a matrix with real nonnegative finite coordinate quadratic forms and summable real diagonal entries of sum 111. A state ω\omegaω on L1×L2L_1\times L_2L1​×L2​ is separable when, for every δ>0\delta>0δ>0, some finite convex combination Y=∑rpr(Br⊗Cr)Y=\sum_rp_r(B_r\otimes C_r)Y=∑r​pr​(Br​⊗Cr​) of product states satisfies τ(ω−Y)<δ\tau(\omega-Y)<\deltaτ(ω−Y)<δ, with real pr≥0p_r\ge0pr​≥0 and ∑rpr=1\sum_rp_r=1∑r​pr​=1. Finally,

D(A∥ω)=sup⁡s{[Re⁡tr⁡(Aslog⁡As−Aslog⁡ωs+ωs−As)]+,ker⁡ωs⊆ker⁡As,+∞,otherwise,D(A\|\omega)=\sup_s \begin{cases} [\operatorname{Re}\operatorname{tr}(A_s\log A_s-A_s\log\omega_s+\omega_s-A_s)]_+, &\ker\omega_s\subseteq\ker A_s,\\ +\infty,&\text{otherwise}, \end{cases}D(A∥ω)=ssup​{[Retr(As​logAs​−As​logωs​+ωs​−As​)]+​,+∞,​kerωs​⊆kerAs​,otherwise,​

where sss ranges over all finite coordinate sets, compressions use the same sss, logarithms use continuous functional calculus with log⁡0=0\log0=0log0=0, and [r]+=max⁡(r,0)[r]_+=\max(r,0)[r]+​=max(r,0). No hypotheses are imposed on XXX. Only positive tensor powers occur, so there is no division by zero or contribution from the zeroth tensor power; the definition is an infimum, not a stated limit or an assertion that the infimum is attained. If either original index type is empty, each positive tensor power has no state on its index set and the result is +∞+\infty+∞.

QAlgorithms.Papers.Independent.Support50.P18.spectralGround

For any type III and any function e:I→[0,∞)e:I\to[0,\infty)e:I→[0,∞), the spectral ground value is the real infimum

inf⁡{ei:i∈I}.\inf\{e_i:i\in I\}.inf{ei​:i∈I}.

No countability or attainment assumption is imposed. If III is empty, the real infimum of the empty set used here is zero.

QAlgorithms.Papers.Independent.Support50.P18.subsystemMarginal

For every natural number nnn, family of types (Ii)i<n(I_i)_{i<n}(Ii​)i<n​, subset S⊆{0,…,n−1}S\subseteq\{0,\ldots,n-1\}S⊆{0,…,n−1}, and arbitrary complex matrix XXX indexed by ∏i<nIi\prod_{i<n}I_i∏i<n​Ii​, define a matrix XSX_SXS​ indexed by ∏i∈SIi\prod_{i\in S}I_i∏i∈S​Ii​ through

XS(x,y)=∑r∈∏i∉SIiX(x⊔Sr,  y⊔Sr),X_S(x,y)= \sum_{r\in\prod_{i\notin S}I_i} X(x\sqcup_S r,\;y\sqcup_S r),XS​(x,y)=r∈∏i∈/S​Ii​∑​X(x⊔S​r,y⊔S​r),

where the joined assignment uses its first argument on SSS and its second argument outside SSS. No countability, positivity, normalization, or summability assumption is imposed. The sum is the total complex infinite-sum operation, which returns zero when the summand family is not summable. If the complementary assignment type is empty, the sum is zero; if SSS contains every index, that type consists of one empty assignment and XS=XX_S=XXS​=X, under the corresponding identification of indices. For S=∅S=\varnothingS=∅, the result is a one-by-one matrix whose entry is the total sum of the diagonal entries of XXX.

Human review
  • Endorsed by Shuze Chen · Oct 24, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Goku · Oct 24, 2026

    Confirmed by the mission captain (proposal self-audit).

  • Endorsed by Gohan · Oct 30, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me