Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Definitions for Quantum trade-off coding for bosonic communication

Definition
QAlgorithms_Papers_Independent_Gap48_P44

by Goku · Oct 22, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

The objects the formalized statements of Quantum trade-off coding for bosonic communication (arXiv:1105.0119) are stated in. Each was checked against the paper by a blind read-back and a moderator.

  • cqFamily: PDF pp.6,14, (40)-(41), (81)-(82): nonnegative CQ coding family. The three arguments are the unconditional receiver entropy and the conditional receiver/complement entropies in bits, as functions of photon sharing lambda. This is a scalar inner family, with no assertion of operational achievability in its definition.
  • ceFamily: PDF pp.7,14, (42)-(43) and the CE display: positive entanglement consumption. The coordinates are classical bits and consumed ebits per use.
  • cqTimeSharing: PDF pp.6-8, Figure 2(a): same-budget endpoint time sharing, with rate disposal. Coordinates are (classical, quantum); t is the quantum-code time fraction.
  • ceTimeSharing: PDF pp.7-8, Figure 2(b): same-budget CE endpoint mixtures. Extra consumption is allowed, while generated classical bits can be discarded.
  • lossCQ: PDF pp.6-8: the displayed pure-loss CQ family at the stated photon budget.
  • lossCE: PDF pp.7-8: the displayed pure-loss CE family, with consumption positive.
  • thermalD: PDF p.14, (80): D is the nonnegative square root, rather than an arbitrary root.
  • thermalEnvironment: PDF p.14, (80): entropy in bits of the TRUE complementary output. Both symplectic modes, including the environment purifier, contribute.
  • amplifierCQ: PDF pp.16-17, Figure 6(a): vacuum-amplifier CQ family, at N_S=200.
  • amplifierCE: PDF pp.16-17, Figure 6(b): vacuum-amplifier CE family, at N_S=200.
  • fockParity: PDF p.5, (27): parity implements the environmental output phase in the source beam splitter relative to the accepted opposite-sign output convention.
  • lossComplement: PDF p.5, (26)-(27): the fixed pure-loss complement on arbitrary block inputs. The imported receiver has precisely (26). Its unused port has the opposite sign to (27); parity on both matrix indices fixes that sign without changing entropy.
  • BlockEnsemble: PDF pp.2,11, (60)-(63): a discrete block ensemble, allowing countably many nonzero weights and arbitrary intra-block entanglement. PMF labels may have zero mass.
  • BlockEnsemble.average: PDF pp.4,11: the ensemble's actual unconditional density coefficients. Absolute convergence follows from trace-one states and normalized probabilities.
  • BlockEnsemble.energy: PDF pp.2,11-12: averaged TOTAL input photon number. Sum the imported modeEnergy over modes, then average over the ensemble. By positivity this is the energy of the unconditional density.
  • blockEntropy: PDF pp.2,11, (60)-(63): bit entropy, keeping infinity until after weighting.
  • BlockEnsemble.entropyAverage: PDF p.11: weighted entropy is formed in the extended nonnegative reals. In particular a zero-probability label with infinite entropy contributes zero.
  • lossQuantumFamily: PDF p.5, (28)-(30): all three faces of the signed pure-loss inner family. This scalar set is not used as the definition of any operational capacity.
  • BlockEnsemble.dynamicEntropyRegion: PDF pp.3-4,11: the computed block-ensemble entropy region, scaled per use. Only finite-energy ensembles at positive blocklength enter the optimization below. Each entropy average is weighted BEFORE conversion to real; zero times infinity is zero. Input energy bounds imply finiteness of every converted average and the unconditional output entropy. Signed net rates are allowed in all coordinates.
  • gaussianDynamicOptimization: PDF pp.11-13: regularized dynamic ENTROPY optimization restricted to Gaussian conditional input states, not to Gaussian gates or operational codes. All finite label sets and all positive blocklengths are included, followed by closure. Conditional states may have arbitrary Gaussian correlations among the n modes; their unconditional mixture is not required to be Gaussian. Finite alphabets with closure retain the continuous displaced-thermal ensemble through discretization.
Definition code
import Mathlib
import Definitions.Def_QAlgorithms_Papers_Independent_Claimed50_P39
import Definitions.Def_QAlgorithms_Papers_Independent_Information100_P54
import Definitions.Def_QAlgorithms_Papers_Independent_Next46_P32

/-!
Quantum trade-off coding for bosonic communication, arXiv:1105.0119v4.
Scalar coding-family comparisons and the common-parameter block entropy bounds.
Operational results lacking a full Fock-space code interface are recorded separately
in HARD.md and GAPS.json; no scalar family is defined to be an operational capacity.
Compilation and independent moderation are pending the separate batch stages.
-/

noncomputable section

namespace QAlgorithms.Papers.Independent.Gap48.P44

open Filter
open scoped ENNReal Topology
open QAlgorithms.Papers.Independent.Claimed50.P39 (g)
open QAlgorithms.Papers.Independent.Information100.P54 (thermalEntropy)

namespace Fock
export QAlgorithms.Papers.Independent.Next46.P32
  (NumberString FockKernel pureLoss jointMix charlieMarginal vacuum
   modeEnergy spectralEntropyNats StrongConjecture2 Gaussian)
export QAlgorithms.Papers.Independent.Dependency50.P33 (State)
end Fock

/-- PDF pp.6,14, (40)-(41), (81)-(82): nonnegative CQ coding family.
The three arguments are the unconditional receiver entropy and the conditional
receiver/complement entropies in bits, as functions of photon sharing lambda.
This is a scalar inner family, with no assertion of operational achievability in its definition. -/
def cqFamily (output : ℝ) (conditional environment : ℝ → ℝ) : Set (ℝ × ℝ) :=
  {r | 0 ≤ r.1 ∧ 0 ≤ r.2 ∧ ∃ share ∈ Set.Icc (0 : ℝ) 1,
    r.2 ≤ conditional share - environment share ∧
    r.1 + r.2 ≤ output - environment share}

/-- PDF pp.7,14, (42)-(43) and the CE display: positive entanglement consumption.
The coordinates are classical bits and consumed ebits per use. -/
def ceFamily (budget output : ℝ) (environment : ℝ → ℝ) : Set (ℝ × ℝ) :=
  {r | 0 ≤ r.1 ∧ 0 ≤ r.2 ∧ ∃ share ∈ Set.Icc (0 : ℝ) 1,
    r.1 ≤ thermalEntropy (share * budget) + output - environment share ∧
    r.1 ≤ output - environment share + r.2}

/-- PDF pp.6-8, Figure 2(a): same-budget endpoint time sharing, with rate disposal.
Coordinates are (classical, quantum); t is the quantum-code time fraction. -/
def cqTimeSharing (classical quantum : ℝ) : Set (ℝ × ℝ) :=
  {r | 0 ≤ r.1 ∧ 0 ≤ r.2 ∧ ∃ t ∈ Set.Icc (0 : ℝ) 1,
    r.1 ≤ (1 - t) * classical ∧ r.2 ≤ t * quantum}

/-- PDF pp.7-8, Figure 2(b): same-budget CE endpoint mixtures.
Extra consumption is allowed, while generated classical bits can be discarded. -/
def ceTimeSharing (unassisted assisted consumption : ℝ) : Set (ℝ × ℝ) :=
  {r | 0 ≤ r.1 ∧ 0 ≤ r.2 ∧ ∃ t ∈ Set.Icc (0 : ℝ) 1,
    r.1 ≤ (1 - t) * unassisted + t * assisted ∧ t * consumption ≤ r.2}

/-- PDF pp.6-8: the displayed pure-loss CQ family at the stated photon budget. -/
def lossCQ (eta budget : ℝ) : Set (ℝ × ℝ) :=
  cqFamily (thermalEntropy (eta * budget))
    (fun share => thermalEntropy (eta * share * budget))
    (fun share => thermalEntropy ((1 - eta) * share * budget))

/-- PDF pp.7-8: the displayed pure-loss CE family, with consumption positive. -/
def lossCE (eta budget : ℝ) : Set (ℝ × ℝ) :=
  ceFamily budget (thermalEntropy (eta * budget))
    (fun share => thermalEntropy ((1 - eta) * share * budget))

/-- PDF p.14, (80): D is the nonnegative square root, rather than an arbitrary root. -/
def thermalD (eta share budget noise : ℝ) : ℝ :=
  Real.sqrt ((share * (1 + eta) * budget + (1 - eta) * noise + 1) ^ 2 -
    4 * eta * share * budget * (share * budget + 1))

/-- PDF p.14, (80): entropy in bits of the TRUE complementary output.
Both symplectic modes, including the environment purifier, contribute. -/
def thermalEnvironment (eta share budget noise : ℝ) : ℝ :=
  thermalEntropy ((thermalD eta share budget noise +
    share * (1 - eta) * budget - (1 - eta) * noise - 1) / 2) +
  thermalEntropy ((thermalD eta share budget noise -
    share * (1 - eta) * budget + (1 - eta) * noise - 1) / 2)

/-- PDF pp.16-17, Figure 6(a): vacuum-amplifier CQ family, at N_S=200. -/
def amplifierCQ (gain : ℝ) : Set (ℝ × ℝ) :=
  cqFamily (thermalEntropy (gain * 200 + gain - 1))
    (fun share => thermalEntropy (gain * share * 200 + gain - 1))
    (fun share => thermalEntropy ((gain - 1) * (share * 200 + 1)))

/-- PDF pp.16-17, Figure 6(b): vacuum-amplifier CE family, at N_S=200. -/
def amplifierCE (gain : ℝ) : Set (ℝ × ℝ) :=
  ceFamily 200 (thermalEntropy (gain * 200 + gain - 1))
    (fun share => thermalEntropy ((gain - 1) * (share * 200 + 1)))

/-- PDF p.5, (27): parity implements the environmental output phase in the
source beam splitter relative to the accepted opposite-sign output convention. -/
def fockParity {n : ℕ} (a : Fock.NumberString n) : ℂ :=
  (-1 : ℂ) ^ (∑ i, a i)

/-- PDF p.5, (26)-(27): the fixed pure-loss complement on arbitrary block inputs.
The imported receiver has precisely (26). Its unused port has the opposite sign
to (27); parity on both matrix indices fixes that sign without changing entropy. -/
def lossComplement (n : ℕ) (eta : ℝ) (rho : Fock.FockKernel n) :
    Fock.FockKernel n :=
  fun a b => fockParity a *
    Fock.charlieMarginal (Fock.jointMix n eta rho (Fock.vacuum n)) a b * star (fockParity b)

/-- PDF pp.2,11, (60)-(63): a discrete block ensemble, allowing countably many
nonzero weights and arbitrary intra-block entanglement. PMF labels may have zero mass. -/
structure BlockEnsemble (X : Type*) (n : ℕ) where
  probability : PMF X
  state : X → Fock.State (Fock.NumberString n)

/-- PDF pp.4,11: the ensemble's actual unconditional density coefficients.
Absolute convergence follows from trace-one states and normalized probabilities. -/
def BlockEnsemble.average {X : Type*} {n : ℕ} (e : BlockEnsemble X n) :
    Fock.FockKernel n :=
  fun a b => ∑' x, ((e.probability x).toReal : ℂ) * (e.state x).coefficients a b

/-- PDF pp.2,11-12: averaged TOTAL input photon number.
Sum the imported modeEnergy over modes, then average over the ensemble.
By positivity this is the energy of the unconditional density. -/
def BlockEnsemble.energy {X : Type*} {n : ℕ} (e : BlockEnsemble X n) : ℝ≥0∞ :=
  ∑' x, e.probability x * (∑ i : Fin n, Fock.modeEnergy (e.state x).coefficients i)

/-- PDF pp.2,11, (60)-(63): bit entropy, keeping infinity until after weighting. -/
def blockEntropy {n : ℕ} (rho : Fock.FockKernel n) : ℝ≥0∞ :=
  Fock.spectralEntropyNats rho / ENNReal.ofReal (Real.log 2)

/-- PDF p.11: weighted entropy is formed in the extended nonnegative reals.
In particular a zero-probability label with infinite entropy contributes zero. -/
def BlockEnsemble.entropyAverage {X : Type*} {n : ℕ} (e : BlockEnsemble X n)
    (channel : Fock.FockKernel n → Fock.FockKernel n) : ℝ≥0∞ :=
  ∑' x, e.probability x * blockEntropy (channel (e.state x).coefficients)

/-- PDF p.5, (28)-(30): all three faces of the signed pure-loss inner family.
This scalar set is not used as the definition of any operational capacity. -/
def lossQuantumFamily (eta budget : ℝ) : Set (Fin 3 → ℝ) :=
  {r | ∃ share ∈ Set.Icc (0 : ℝ) 1,
    r 0 + 2 * r 1 ≤ thermalEntropy (share * budget) + thermalEntropy (eta * budget) -
      thermalEntropy ((1 - eta) * share * budget) ∧
    r 1 + r 2 ≤ thermalEntropy (eta * share * budget) -
      thermalEntropy ((1 - eta) * share * budget) ∧
    r 0 + r 1 + r 2 ≤ thermalEntropy (eta * budget) -
      thermalEntropy ((1 - eta) * share * budget)}

/-- PDF pp.3-4,11: the computed block-ensemble entropy region, scaled per use.
Only finite-energy ensembles at positive blocklength enter the optimization below.
Each entropy average is weighted BEFORE conversion to real; zero times infinity
is zero. Input energy bounds imply finiteness of every converted average and the
unconditional output entropy. Signed net rates are allowed in all coordinates. -/
def BlockEnsemble.dynamicEntropyRegion {X : Type*} {n : ℕ}
    (e : BlockEnsemble X n) (eta : ℝ) : Set (Fin 3 → ℝ) :=
  let input := (e.entropyAverage id).toReal
  let output := (blockEntropy (Fock.pureLoss n eta e.average)).toReal
  let conditional := (e.entropyAverage (Fock.pureLoss n eta)).toReal
  let environment := (e.entropyAverage (lossComplement n eta)).toReal
  {r | (n : ℝ) * (r 0 + 2 * r 1) ≤ input + output - environment ∧
    (n : ℝ) * (r 1 + r 2) ≤ conditional - environment ∧
    (n : ℝ) * (r 0 + r 1 + r 2) ≤ output - environment}

/-- PDF pp.11-13: regularized dynamic ENTROPY optimization restricted to
Gaussian conditional input states, not to Gaussian gates or operational codes.
All finite label sets and all positive blocklengths are included, followed by closure.
Conditional states may have arbitrary Gaussian correlations among the n modes;
their unconditional mixture is not required to be Gaussian. Finite alphabets with
closure retain the continuous displaced-thermal ensemble through discretization. -/
def gaussianDynamicOptimization (eta budget : ℝ) : Set (Fin 3 → ℝ) :=
  closure {r | ∃ n : ℕ, 0 < n ∧ ∃ k : ℕ, 0 < k ∧
    ∃ e : BlockEnsemble (Fin k) n,
      (∀ x, Fock.Gaussian (e.state x).coefficients) ∧
      e.energy ≤ ENNReal.ofReal ((n : ℝ) * budget) ∧
      r ∈ e.dynamicEntropyRegion eta}

example : g 0 = 0 := by simp [g]

example : thermalEntropy 0 = 0 := by simp [thermalEntropy]

example (N : ℝ) :
    (-1 : ℝ) ^ (0 + 2) / (((0 + 1 : ℕ) : ℝ) * ((0 + 2 : ℕ) : ℝ) * N ^ (0 + 1)) =
      1 / (2 * N) := by simp

example : (0, 0) ∈ cqFamily 0 (fun _ => 0) (fun _ => 0) := by
  refine ⟨le_rfl, le_rfl, 0, ⟨le_rfl, zero_le_one⟩, ?_, ?_⟩ <;> simp

example : (0, 0) ∈ ceFamily 0 0 (fun _ => 0) := by
  refine ⟨le_rfl, le_rfl, 0, ⟨le_rfl, zero_le_one⟩, ?_, ?_⟩ <;>
    simp [thermalEntropy]

example : (0, 0) ∈ cqTimeSharing 0 0 := by
  exact ⟨le_rfl, le_rfl, 0, ⟨le_rfl, zero_le_one⟩, by simp, by simp⟩

example : (0, 3) ∈ ceTimeSharing 0 0 0 := by
  refine ⟨le_rfl, by norm_num, 0, ⟨le_rfl, zero_le_one⟩, ?_, ?_⟩ <;> norm_num

example (eta share : ℝ) : thermalD eta share 0 0 = 1 := by simp [thermalD]

example (eta share : ℝ) : thermalEnvironment eta share 0 0 = 0 := by
  simp [thermalEnvironment, thermalD, thermalEntropy]

example : fockParity (fun _ : Fin 1 => 1) = -1 := by simp [fockParity]

example (rho : Fock.FockKernel 0) : (∑ i : Fin 0, Fock.modeEnergy rho i) = 0 := by simp

example : (0 : ℝ≥0∞) * (⊤ : ℝ≥0∞) = 0 := by simp

example (eta : ℝ) : (0, 0) ∈ lossCQ eta 0 := by
  refine ⟨le_rfl, le_rfl, 0, ⟨le_rfl, zero_le_one⟩, ?_, ?_⟩ <;>
    simp [thermalEntropy]

example (eta : ℝ) : (0, 0) ∈ lossCE eta 0 := by
  refine ⟨le_rfl, le_rfl, 0, ⟨le_rfl, zero_le_one⟩, ?_, ?_⟩ <;>
    simp [thermalEntropy]

example : amplifierCQ 1 = lossCQ 1 200 := by
  simp [amplifierCQ, lossCQ, thermalEntropy]

example : amplifierCE 1 = lossCE 1 200 := by
  simp [amplifierCE, lossCE, thermalEntropy]

example (n : ℕ) (rho : Fock.State (Fock.NumberString n)) :
    (BlockEnsemble.average
      (⟨PMF.pure (), fun _ : Unit => rho⟩ : BlockEnsemble Unit n)) = rho.coefficients := by
  classical
  ext a b
  simp [BlockEnsemble.average, PMF.pure_apply]

example (n : ℕ) (rho : Fock.State (Fock.NumberString n)) :
    (BlockEnsemble.energy
      (⟨PMF.pure (), fun _ : Unit => rho⟩ : BlockEnsemble Unit n)) =
      (∑ i : Fin n, Fock.modeEnergy rho.coefficients i) := by
  simp [BlockEnsemble.energy, PMF.pure_apply]

example (n : ℕ) (rho : Fock.State (Fock.NumberString n)) :
    (BlockEnsemble.entropyAverage
      (⟨PMF.pure (), fun _ : Unit => rho⟩ : BlockEnsemble Unit n) id) =
      blockEntropy rho.coefficients := by
  simp [BlockEnsemble.entropyAverage, PMF.pure_apply]

example (eta : ℝ) : (fun _ : Fin 3 => (0 : ℝ)) ∈ lossQuantumFamily eta 0 := by
  refine ⟨0, ⟨le_rfl, zero_le_one⟩, ?_, ?_, ?_⟩ <;> simp [thermalEntropy]

end QAlgorithms.Papers.Independent.Gap48.P44
Source
Quantum trade-off coding for bosonic communication, arXiv:1105.0119v4, https://arxiv.org/abs/1105.0119v4
Read-back

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

QAlgorithms.Papers.Independent.Gap48.P44.BlockEnsemble.average

For any type XXX, any natural number nnn, including zero, and an ensemble eee with probability masses p(x)∈[0,∞]p(x)\in[0,\infty]p(x)∈[0,∞], ∑xp(x)=1\sum_xp(x)=1∑x​p(x)=1, and density kernels ρx\rho_xρx​ on Nn\mathbb N^nNn, its average is the complex kernel defined entrywise by

ρ‾ab=∑xp(x)R(ρx)ab,a,b∈Nn,\overline{\rho}_{ab}=\sum_x p(x)_{\mathbb R}(\rho_x)_{ab}, \qquad a,b\in\mathbb N^n,ρ​ab​=x∑​p(x)R​(ρx​)ab​,a,b∈Nn,

where p(x)Rp(x)_{\mathbb R}p(x)R​ is the real value of the probability mass, subsequently regarded as a complex scalar. Each density kernel here has a representation (ρx)ab=∑juj(a)vj(b)‾(\rho_x)_{ab}=\sum_j u_j(a)\overline{v_j(b)}(ρx​)ab​=∑j​uj​(a)vj​(b)​ with uj,vj∈ℓ2(Nn;C)u_j,v_j\in\ell^2(\mathbb N^n;\mathbb C)uj​,vj​∈ℓ2(Nn;C) and summable ∥uj∥∥vj∥\|u_j\|\|v_j\|∥uj​∥∥vj​∥, has real nonnegative finite quadratic forms ∑a,b∈Sz(a)‾(ρx)abz(b)\sum_{a,b\in S}\overline{z(a)}(\rho_x)_{ab}z(b)∑a,b∈S​z(a)​(ρx​)ab​z(b), and has summable real diagonal with total one. The displayed entrywise sum uses the total complex-series convention, which assigns zero to a nonsummable family; this definition itself supplies no separate summability or density assertion about the resulting kernel.

QAlgorithms.Papers.Independent.Gap48.P44.BlockEnsemble.dynamicEntropyRegion

For any type XXX, natural number nnn, ensemble e=(p,ρ)e=(p,\rho)e=(p,ρ) on occupation labels N=Nn\mathcal N=\mathbb N^nN=Nn, and real η\etaη, this region consists of all real triples rrr satisfying

n(r0+2r1)≤I+O−E,n(r1+r2)≤Q−E,n(r0+r1+r2)≤O−E,n(r_0+2r_1)\le I+O-E,\qquad n(r_1+r_2)\le Q-E,\qquad n(r_0+r_1+r_2)\le O-E,n(r0​+2r1​)≤I+O−E,n(r1​+r2​)≤Q−E,n(r0​+r1​+r2​)≤O−E,

where

I=τ ⁣(∑xp(x)H(ρx)),O=τ ⁣(H(Lη(ρˉ))),Q=τ ⁣(∑xp(x)H(Lη(ρx))),E=τ ⁣(∑xp(x)H(Lη c(ρx))).I=\tau\!\left(\sum_xp(x)H(\rho_x)\right),\quad O=\tau\!\left(H(L_\eta(\bar\rho))\right),\quad Q=\tau\!\left(\sum_xp(x)H(L_\eta(\rho_x))\right),\quad E=\tau\!\left(\sum_xp(x)H(L_\eta^{\,c}(\rho_x))\right).I=τ(x∑​p(x)H(ρx​)),O=τ(H(Lη​(ρˉ​))),Q=τ(x∑​p(x)H(Lη​(ρx​))),E=τ(x∑​p(x)H(Lηc​(ρx​))).

Here τ:[0,∞]→R\tau:[0,\infty]\to\mathbb Rτ:[0,∞]→R keeps finite values and sends +∞+\infty+∞ to zero, ρˉab=∑xp(x)R(ρx)ab\bar\rho_{ab}=\sum_xp(x)_{\mathbb R}(\rho_x)_{ab}ρˉ​ab​=∑x​p(x)R​(ρx​)ab​, and ppp is a probability mass function. Each ρx\rho_xρx​ admits an expansion (ρx)ab=∑juj(a)vj(b)‾(\rho_x)_{ab}=\sum_j u_j(a)\overline{v_j(b)}(ρx​)ab​=∑j​uj​(a)vj​(b)​ with uj,vj∈ℓ2(N;C)u_j,v_j\in\ell^2(\mathcal N;\mathbb C)uj​,vj​∈ℓ2(N;C) and ∑j∥uj∥∥vj∥<∞\sum_j\|u_j\|\|v_j\|<\infty∑j​∥uj​∥∥vj​∥<∞, has real nonnegative finite quadratic forms, and has summable real diagonal of total one. To specify the two maps literally, put Vab=1V_{ab}=1Vab​=1 if a=b=0a=b=0a=b=0 and Vab=0V_{ab}=0Vab​=0 otherwise, and define

bη(m,ℓ;p,q)=1m+ℓ=p+qm!ℓ!p!q!∑u=0p1u≤m, m−u≤q(pu)(qm−u)(η)u(1−η)p−u+(m−u)(−η)q−(m−u),Bη((a,e),(c,d))=∏i=0n−1bη(ai,ei;ci,di),Jη(Z)uv=∑a∈N×N∑b∈N×NBη(u,a) Za1b1Va2b2 Bη(v,b)‾,Lη(Z)ab=∑e∈NJη(Z)(a,e),(b,e),Lη c(Z)ab=(−1)∑iai(∑e∈NJη(Z)(e,a),(e,b))(−1)∑ibi‾.\begin{aligned} b_\eta(m,\ell;p,q) &=\mathbf1_{m+\ell=p+q} \sqrt{\frac{m!\ell!}{p!q!}} \sum_{u=0}^{p} \mathbf1_{u\le m,\ m-u\le q} \binom pu\binom q{m-u} (\sqrt\eta)^u (\sqrt{1-\eta})^{p-u+(m-u)} (-\sqrt\eta)^{q-(m-u)},\\ B_\eta((a,e),(c,d)) &=\prod_{i=0}^{n-1}b_\eta(a_i,e_i;c_i,d_i),\\ J_\eta(Z)_{uv} &=\sum_{a\in\mathcal N\times\mathcal N} \sum_{b\in\mathcal N\times\mathcal N} B_\eta(u,a)\,Z_{a_1b_1}V_{a_2b_2}\, \overline{B_\eta(v,b)},\\ L_\eta(Z)_{ab}&=\sum_{e\in\mathcal N}J_\eta(Z)_{(a,e),(b,e)},\\ L_\eta^{\,c}(Z)_{ab} &=(-1)^{\sum_i a_i} \left(\sum_{e\in\mathcal N}J_\eta(Z)_{(e,a),(e,b)}\right) \overline{(-1)^{\sum_i b_i}}. \end{aligned}bη​(m,ℓ;p,q)Bη​((a,e),(c,d))Jη​(Z)uv​Lη​(Z)ab​Lηc​(Z)ab​​=1m+ℓ=p+q​p!q!m!ℓ!​​u=0∑p​1u≤m, m−u≤q​(up​)(m−uq​)(η​)u(1−η​)p−u+(m−u)(−η​)q−(m−u),=i=0∏n−1​bη​(ai​,ei​;ci​,di​),=a∈N×N∑​b∈N×N∑​Bη​(u,a)Za1​b1​​Va2​b2​​Bη​(v,b)​,=e∈N∑​Jη​(Z)(a,e),(b,e)​,=(−1)∑i​ai​(e∈N∑​Jη​(Z)(e,a),(e,b)​)(−1)∑i​bi​​.​

The real beam coefficients are regarded as complex numbers in these kernel formulas. For any kernel ZZZ, H(Z)H(Z)H(Z) is 1/ln⁡21/\ln21/ln2 times the infimum of ∑j(−wjln⁡wj)+\sum_j(-w_j\ln w_j)_+∑j​(−wj​lnwj​)+​ over summable nonnegative weights totaling one and coefficient resolutions Zab=∑jwjhj(a)hj(b)‾Z_{ab}=\sum_jw_jh_j(a)\overline{h_j(b)}Zab​=∑j​wj​hj​(a)hj​(b)​, with hj∈ℓ2(N;C)h_j\in\ell^2(\mathcal N;\mathbb C)hj​∈ℓ2(N;C) norm one and mutually orthogonal wherever the weights are positive; an empty infimum is +∞+\infty+∞. Here t+=max⁡(t,0)t_+=\max(t,0)t+​=max(t,0), ln⁡0=0\ln0=0ln0=0, and 0⋅∞=00\cdot\infty=00⋅∞=0 in the extended nonnegative averages. Complex infinite sums use the total sum, equal to zero for a nonsummable family; natural-number subtractions in the beam coefficients are truncated at zero, and the real square root of a negative number is zero. There is no restriction on η\etaη, no nonnegativity restriction on rrr, and no finiteness assumption on the entropies; n=0n=0n=0 is included and makes all three left sides zero.

QAlgorithms.Papers.Independent.Gap48.P44.BlockEnsemble.energy

For any type XXX, any natural number nnn, including zero, and an ensemble with probability masses p(x)p(x)p(x) summing to one and density kernels ρx\rho_xρx​ on the occupation labels a∈Nna\in\mathbb N^na∈Nn, its energy is the extended nonnegative number

E(e)=∑xp(x)(∑i=0n−1∑a∈Nn(aiRe⁡(ρx)aa)+),E(e)=\sum_xp(x)\left(\sum_{i=0}^{n-1}\sum_{a\in\mathbb N^n} \bigl(a_i\operatorname{Re}(\rho_x)_{aa}\bigr)_+\right),E(e)=x∑​p(x)(i=0∑n−1​a∈Nn∑​(ai​Re(ρx​)aa​)+​),

where t+=max⁡(t,0)t_+=\max(t,0)t+​=max(t,0) is regarded as an element of [0,∞][0,\infty][0,∞]. A density kernel means a kernel admitting an expansion (ρx)ab=∑juj(a)vj(b)‾(\rho_x)_{ab}=\sum_j u_j(a)\overline{v_j(b)}(ρx​)ab​=∑j​uj​(a)vj​(b)​, with uj,vj∈ℓ2(Nn;C)u_j,v_j\in\ell^2(\mathbb N^n;\mathbb C)uj​,vj​∈ℓ2(Nn;C) and ∑j∥uj∥∥vj∥<∞\sum_j\|u_j\|\|v_j\|<\infty∑j​∥uj​∥∥vj​∥<∞, whose finite quadratic forms are real and nonnegative and whose real diagonal is summable with total one. The energy is not divided by nnn, may be infinite, and equals zero when n=0n=0n=0; extended nonnegative multiplication uses 0⋅∞=00\cdot\infty=00⋅∞=0.

QAlgorithms.Papers.Independent.Gap48.P44.BlockEnsemble.entropyAverage

For any type XXX, any natural number nnn, including zero, an ensemble e=(p,ρ)e=(p,\rho)e=(p,ρ), and any function CCC from complex kernels on Nn\mathbb N^nNn to such kernels, the entropy average is

Ae(C)=∑xp(x)H(C(ρx))∈[0,∞].A_e(C)=\sum_xp(x)H(C(\rho_x))\quad\in[0,\infty].Ae​(C)=x∑​p(x)H(C(ρx​))∈[0,∞].

The masses satisfy ∑xp(x)=1\sum_xp(x)=1∑x​p(x)=1, and every ρx\rho_xρx​, including those of mass zero, admits a coefficient expansion (ρx)ab=∑juj(a)vj(b)‾(\rho_x)_{ab}=\sum_j u_j(a)\overline{v_j(b)}(ρx​)ab​=∑j​uj​(a)vj​(b)​ with uj,vj∈ℓ2(Nn;C)u_j,v_j\in\ell^2(\mathbb N^n;\mathbb C)uj​,vj​∈ℓ2(Nn;C) and summable ∥uj∥∥vj∥\|u_j\|\|v_j\|∥uj​∥∥vj​∥, has real nonnegative finite quadratic forms, and has summable real diagonal of total one. For any kernel ZZZ, H(Z)H(Z)H(Z) is 1/ln⁡21/\ln21/ln2 times the infimum of ∑j(−wjln⁡wj)+\sum_j(-w_j\ln w_j)_+∑j​(−wj​lnwj​)+​ over summable nonnegative weights of total one and vectors hj∈ℓ2(Nn;C)h_j\in\ell^2(\mathbb N^n;\mathbb C)hj​∈ℓ2(Nn;C) satisfying Zab=∑jwjhj(a)hj(b)‾Z_{ab}=\sum_jw_jh_j(a)\overline{h_j(b)}Zab​=∑j​wj​hj​(a)hj​(b)​, with norm-one, mutually orthogonal vectors at the positive weights; the infimum is +∞+\infty+∞ if there is no such resolution, and ln⁡0=0\ln0=0ln0=0. No linearity, positivity, continuity, or normalization condition is imposed on CCC. The average may be infinite, and zero probability times infinite entropy is zero.

QAlgorithms.Papers.Independent.Gap48.P44.BlockEnsemble

For any type XXX, with no finiteness or countability assumption, and any natural number nnn, including n=0n=0n=0, a block ensemble consists of a probability mass function p:X→[0,∞]p:X\to[0,\infty]p:X→[0,∞], satisfying ∑xp(x)=1\sum_x p(x)=1∑x​p(x)=1, together with a density kernel ρx\rho_xρx​ indexed by Nn×Nn\mathbb N^n\times\mathbb N^nNn×Nn for every x∈Xx\in Xx∈X, including labels of probability zero. Here Nn\mathbb N^nNn means the functions {0,…,n−1}→N\{0,\ldots,n-1\}\to\mathbb N{0,…,n−1}→N. Saying that ρx\rho_xρx​ is a density kernel means that there exist sequences uj,vj∈ℓ2(Nn;C)u_j,v_j\in\ell^2(\mathbb N^n;\mathbb C)uj​,vj​∈ℓ2(Nn;C) with ∑j∥uj∥∥vj∥<∞\sum_j\|u_j\|\|v_j\|<\infty∑j​∥uj​∥∥vj​∥<∞ and (ρx)ab=∑juj(a)vj(b)‾(\rho_x)_{ab}=\sum_j u_j(a)\overline{v_j(b)}(ρx​)ab​=∑j​uj​(a)vj​(b)​; for every finite S⊆NnS\subseteq\mathbb N^nS⊆Nn and every z:Nn→Cz:\mathbb N^n\to\mathbb Cz:Nn→C, the number ∑a∈S∑b∈Sz(a)‾(ρx)abz(b)\sum_{a\in S}\sum_{b\in S}\overline{z(a)}(\rho_x)_{ab}z(b)∑a∈S​∑b∈S​z(a)​(ρx​)ab​z(b) has zero imaginary part and nonnegative real part; and the real diagonal entries form a summable family with ∑aRe⁡(ρx)aa=1\sum_a\operatorname{Re}(\rho_x)_{aa}=1∑a​Re(ρx​)aa​=1. No ensemble exists when XXX is empty; when n=0n=0n=0, the occupation-label set N0\mathbb N^0N0 has one element.

QAlgorithms.Papers.Independent.Gap48.P44.amplifierCE

For any real number GGG, this defines the set of pairs (r1,r2)∈R2(r_1,r_2)\in\mathbb R^2(r1​,r2​)∈R2 with r1,r2≥0r_1,r_2\ge0r1​,r2​≥0 for which some s∈[0,1]s\in[0,1]s∈[0,1] satisfies

r1≤H(200s)+H(200G+G−1)−H((G−1)(200s+1)),r_1\le H(200s)+H(200G+G-1)-H((G-1)(200s+1)),r1​≤H(200s)+H(200G+G−1)−H((G−1)(200s+1)), r1≤H(200G+G−1)−H((G−1)(200s+1))+r2,r_1\le H(200G+G-1)-H((G-1)(200s+1))+r_2,r1​≤H(200G+G−1)−H((G−1)(200s+1))+r2​,

where

H(x)=(x+1)log⁡(x+1)−xlog⁡xlog⁡2,H(x)=\frac{(x+1)\log(x+1)-x\log x}{\log2},H(x)=log2(x+1)log(x+1)−xlogx​,

using 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. The constant 200200200 is fixed, s=0,1s=0,1s=0,1 are allowed, and GGG is unrestricted.

QAlgorithms.Papers.Independent.Gap48.P44.amplifierCQ

For any real number GGG, this defines the set of pairs (r1,r2)∈R2(r_1,r_2)\in\mathbb R^2(r1​,r2​)∈R2 with r1,r2≥0r_1,r_2\ge0r1​,r2​≥0 for which some s∈[0,1]s\in[0,1]s∈[0,1] satisfies

r2≤H(200Gs+G−1)−H((G−1)(200s+1)),r_2\le H(200Gs+G-1)-H((G-1)(200s+1)),r2​≤H(200Gs+G−1)−H((G−1)(200s+1)), r1+r2≤H(200G+G−1)−H((G−1)(200s+1)),r_1+r_2\le H(200G+G-1)-H((G-1)(200s+1)),r1​+r2​≤H(200G+G−1)−H((G−1)(200s+1)),

where

H(x)=(x+1)log⁡(x+1)−xlog⁡xlog⁡2.H(x)=\frac{(x+1)\log(x+1)-x\log x}{\log2}.H(x)=log2(x+1)log(x+1)−xlogx​.

The real logarithm has value 000 at 000 and value log⁡∣x∣\log|x|log∣x∣ at negative arguments. The constant 200200200 is fixed, both endpoints for sss are included, and no lower bound or other restriction is imposed on GGG.

QAlgorithms.Papers.Independent.Gap48.P44.blockEntropy

For any natural number nnn, including zero, and any complex kernel ρ\rhoρ on Nn×Nn\mathbb N^n\times\mathbb N^nNn×Nn, with no density assumption, its block entropy is

H(ρ)=1ln⁡2inf⁡(w,u)∑j=0∞(−wjln⁡wj)+∈[0,∞].H(\rho)=\frac{1}{\ln2}\inf_{(w,u)} \sum_{j=0}^{\infty}\bigl(-w_j\ln w_j\bigr)_+ \quad\in[0,\infty].H(ρ)=ln21​(w,u)inf​j=0∑∞​(−wj​lnwj​)+​∈[0,∞].

Here t+=max⁡(t,0)t_+=\max(t,0)t+​=max(t,0); the infimum ranges over sequences of nonnegative real weights wjw_jwj​ and vectors uj∈ℓ2(Nn;C)u_j\in\ell^2(\mathbb N^n;\mathbb C)uj​∈ℓ2(Nn;C) such that the weights are summable with ∑jwj=1\sum_jw_j=1∑j​wj​=1, every uju_juj​ with wj>0w_j>0wj​>0 has norm one, distinct vectors with positive weights are orthogonal, and ρab=∑jwjuj(a)uj(b)‾\rho_{ab}=\sum_jw_ju_j(a)\overline{u_j(b)}ρab​=∑j​wj​uj​(a)uj​(b)​ for all a,ba,ba,b. Vectors of weight zero have no normalization or orthogonality requirement. The convention is ln⁡0=0\ln0=0ln0=0, and the infimum is +∞+\infty+∞ if no such resolution exists; thus the definition remains meaningful for arbitrary kernels.

QAlgorithms.Papers.Independent.Gap48.P44.ceFamily

For any real numbers B,oB,oB,o and arbitrary function e:R→Re:\mathbb R\to\mathbb Re:R→R, this defines the set of pairs (r1,r2)∈R2(r_1,r_2)\in\mathbb R^2(r1​,r2​)∈R2 with r1,r2≥0r_1,r_2\ge0r1​,r2​≥0 for which some s∈[0,1]s\in[0,1]s∈[0,1] satisfies

r1≤H(sB)+o−e(s),r1≤o−e(s)+r2,r_1\le H(sB)+o-e(s),\qquad r_1\le o-e(s)+r_2,r1​≤H(sB)+o−e(s),r1​≤o−e(s)+r2​,

where

H(x)=(x+1)log⁡(x+1)−xlog⁡xlog⁡2.H(x)=\frac{(x+1)\log(x+1)-x\log x}{\log2}.H(x)=log2(x+1)log(x+1)−xlogx​.

Here the real logarithm is total: log⁡0=0\log0=0log0=0, and its value at a negative argument is the logarithm of that argument’s absolute value. Both endpoints of the interval for sss are allowed, and there are no restrictions on B,o,eB,o,eB,o,e.

QAlgorithms.Papers.Independent.Gap48.P44.ceTimeSharing

For arbitrary real numbers U,A,cU,A,cU,A,c, this defines the set of pairs (r1,r2)∈R2(r_1,r_2)\in\mathbb R^2(r1​,r2​)∈R2 satisfying r1,r2≥0r_1,r_2\ge0r1​,r2​≥0 and admitting some t∈[0,1]t\in[0,1]t∈[0,1] such that

r1≤(1−t)U+tA,tc≤r2.r_1\le(1-t)U+tA,\qquad tc\le r_2.r1​≤(1−t)U+tA,tc≤r2​.

Both endpoints for ttt are included; no sign restrictions are imposed on U,A,cU,A,cU,A,c, and the second displayed inequality is a lower bound on r2r_2r2​.

QAlgorithms.Papers.Independent.Gap48.P44.cqFamily

For any real number ooo and arbitrary functions c,e:R→Rc,e:\mathbb R\to\mathbb Rc,e:R→R, this defines the set of pairs (r1,r2)∈R2(r_1,r_2)\in\mathbb R^2(r1​,r2​)∈R2 satisfying r1≥0r_1\ge0r1​≥0, r2≥0r_2\ge0r2​≥0, and, for at least one s∈[0,1]s\in[0,1]s∈[0,1],

r2≤c(s)−e(s),r1+r2≤o−e(s).r_2\le c(s)-e(s),\qquad r_1+r_2\le o-e(s).r2​≤c(s)−e(s),r1​+r2​≤o−e(s).

The endpoints s=0,1s=0,1s=0,1 are included; no sign, continuity, or other assumptions are imposed on o,c,eo,c,eo,c,e, and the resulting set may be empty.

QAlgorithms.Papers.Independent.Gap48.P44.cqTimeSharing

For arbitrary real numbers C,QC,QC,Q, this defines the set of pairs (r1,r2)∈R2(r_1,r_2)\in\mathbb R^2(r1​,r2​)∈R2 satisfying r1,r2≥0r_1,r_2\ge0r1​,r2​≥0 and admitting some t∈[0,1]t\in[0,1]t∈[0,1] such that

r1≤(1−t)C,r2≤tQ.r_1\le(1-t)C,\qquad r_2\le tQ.r1​≤(1−t)C,r2​≤tQ.

The values t=0,1t=0,1t=0,1 are included, and C,QC,QC,Q are not assumed nonnegative.

QAlgorithms.Papers.Independent.Gap48.P44.fockParity

For every natural number nnn and every function a:{0,…,n−1}→Na:\{0,\ldots,n-1\}\to\mathbb Na:{0,…,n−1}→N, this defines the complex number

(−1)∑i=0n−1ai.(-1)^{\sum_{i=0}^{n-1}a_i}.(−1)∑i=0n−1​ai​.

Thus its value is 111 when the sum is even and −1-1−1 when it is odd. The case n=0n=0n=0 is included: the sum is empty and the value is 111.

QAlgorithms.Papers.Independent.Gap48.P44.gaussianDynamicOptimization

For arbitrary real η\etaη and BBB, this is the ordinary Euclidean closure in R3\mathbb R^3R3 of the triples rrr obtainable as follows: choose positive natural numbers nnn and kkk, a probability vector pxp_xpx​ on x=0,…,k−1x=0,\ldots,k-1x=0,…,k−1, and Gaussian density kernels ρx\rho_xρx​ on N=Nn\mathcal N=\mathbb N^nN=Nn, such that

∑x=0k−1px∑i=0n−1∑a∈N(aiRe⁡(ρx)aa)+≤(nB)+,\sum_{x=0}^{k-1}p_x \sum_{i=0}^{n-1}\sum_{a\in\mathcal N} \bigl(a_i\operatorname{Re}(\rho_x)_{aa}\bigr)_+ \le(nB)_+,x=0∑k−1​px​i=0∑n−1​a∈N∑​(ai​Re(ρx​)aa​)+​≤(nB)+​,

and require

n(r0+2r1)≤I+O−E,n(r1+r2)≤Q−E,n(r0+r1+r2)≤O−E.n(r_0+2r_1)\le I+O-E,\qquad n(r_1+r_2)\le Q-E,\qquad n(r_0+r_1+r_2)\le O-E.n(r0​+2r1​)≤I+O−E,n(r1​+r2​)≤Q−E,n(r0​+r1​+r2​)≤O−E.

Every ρx\rho_xρx​, even at px=0p_x=0px​=0, must be a density kernel: it has an expansion (ρx)ab=∑juj(a)vj(b)‾(\rho_x)_{ab}=\sum_j u_j(a)\overline{v_j(b)}(ρx​)ab​=∑j​uj​(a)vj​(b)​ with uj,vj∈ℓ2(N;C)u_j,v_j\in\ell^2(\mathcal N;\mathbb C)uj​,vj​∈ℓ2(N;C) and summable ∥uj∥∥vj∥\|u_j\|\|v_j\|∥uj​∥∥vj​∥, all finite quadratic forms are real and nonnegative, and its real diagonal is summable with total one. The Gaussian requirement means that there exist μ∈Cn\mu\in\mathbb C^nμ∈Cn, a symmetric positive-definite real matrix AAA indexed by {0,…,n−1}×{0,1}\{0,\ldots,n-1\}\times\{0,1\}{0,…,n−1}×{0,1}, and c>0c>0c>0, possibly different for each xxx, such that for every α∈Cn\alpha\in\mathbb C^nα∈Cn,

1πnRe⁡(∑a∈N∑b∈N∏icαi(ai)‾(ρx)ab∏icαi(bi))=cexp⁡ ⁣(−∑i,jviAijvj),cγ(t)=e−∣γ∣2/2γtt!,\frac{1}{\pi^n}\operatorname{Re} \left(\sum_{a\in\mathcal N}\sum_{b\in\mathcal N} \overline{\prod_i c_{\alpha_i}(a_i)} (\rho_x)_{ab} \prod_i c_{\alpha_i}(b_i)\right) =c\exp\!\left(-\sum_{i,j}v_iA_{ij}v_j\right), \quad c_\gamma(t)=\frac{e^{-|\gamma|^2/2}\gamma^t}{\sqrt{t!}},πn1​Re(a∈N∑​b∈N∑​i∏​cαi​​(ai​)​(ρx​)ab​i∏​cαi​​(bi​))=cexp(−i,j∑​vi​Aij​vj​),cγ​(t)=t!​e−∣γ∣2/2γt​,

where v(i,0)=Re⁡(αi−μi)v_{(i,0)}=\operatorname{Re}(\alpha_i-\mu_i)v(i,0)​=Re(αi​−μi​) and v(i,1)=Im⁡(αi−μi)v_{(i,1)}=\operatorname{Im}(\alpha_i-\mu_i)v(i,1)​=Im(αi​−μi​). In the three inequalities, writing ρˉab=∑xpx(ρx)ab\bar\rho_{ab}=\sum_xp_x(\rho_x)_{ab}ρˉ​ab​=∑x​px​(ρx​)ab​ and letting τ\tauτ preserve finite extended nonnegative values and send +∞+\infty+∞ to zero,

I=τ ⁣(∑xpxH(ρx)),O=τ(H(Lη(ρˉ))),Q=τ ⁣(∑xpxH(Lη(ρx))),E=τ ⁣(∑xpxH(Lη c(ρx))).I=\tau\!\left(\sum_xp_xH(\rho_x)\right),\quad O=\tau(H(L_\eta(\bar\rho))),\quad Q=\tau\!\left(\sum_xp_xH(L_\eta(\rho_x))\right),\quad E=\tau\!\left(\sum_xp_xH(L_\eta^{\,c}(\rho_x))\right).I=τ(x∑​px​H(ρx​)),O=τ(H(Lη​(ρˉ​))),Q=τ(x∑​px​H(Lη​(ρx​))),E=τ(x∑​px​H(Lηc​(ρx​))).

For any kernel ZZZ, H(Z)H(Z)H(Z) is 1/ln⁡21/\ln21/ln2 times the infimum of ∑j(−wjln⁡wj)+\sum_j(-w_j\ln w_j)_+∑j​(−wj​lnwj​)+​ over nonnegative summable weights of total one and resolutions Zab=∑jwjhj(a)hj(b)‾Z_{ab}=\sum_jw_jh_j(a)\overline{h_j(b)}Zab​=∑j​wj​hj​(a)hj​(b)​, with hj∈ℓ2(N;C)h_j\in\ell^2(\mathcal N;\mathbb C)hj​∈ℓ2(N;C) normalized and mutually orthogonal at positive weights; the infimum is +∞+\infty+∞ when there is no resolution. The maps in these formulas are given by

bη(m,ℓ;p,q)=1m+ℓ=p+qm!ℓ!p!q!∑u=0p1u≤m, m−u≤q(pu)(qm−u)(η)u(1−η)p−u+(m−u)(−η)q−(m−u),Bη(u,a)=∏ibη(u1,i,u2,i;a1,i,a2,i),Jη(Z)uv=∑a∈N×N∑b∈N×NBη(u,a)Za1b11a2=0, b2=0Bη(v,b)‾,Lη(Z)ab=∑eJη(Z)(a,e),(b,e),Lη c(Z)ab=(−1)∑iai(∑eJη(Z)(e,a),(e,b))(−1)∑ibi‾.\begin{aligned} b_\eta(m,\ell;p,q) &=\mathbf1_{m+\ell=p+q}\sqrt{\frac{m!\ell!}{p!q!}} \sum_{u=0}^{p}\mathbf1_{u\le m,\ m-u\le q} \binom pu\binom q{m-u} (\sqrt\eta)^u (\sqrt{1-\eta})^{p-u+(m-u)} (-\sqrt\eta)^{q-(m-u)},\\ B_\eta(u,a)&=\prod_i b_\eta(u_{1,i},u_{2,i};a_{1,i},a_{2,i}),\\ J_\eta(Z)_{uv} &=\sum_{a\in\mathcal N\times\mathcal N} \sum_{b\in\mathcal N\times\mathcal N} B_\eta(u,a)Z_{a_1b_1} \mathbf1_{a_2=0,\ b_2=0}\overline{B_\eta(v,b)},\\ L_\eta(Z)_{ab}&=\sum_eJ_\eta(Z)_{(a,e),(b,e)},\\ L_\eta^{\,c}(Z)_{ab} &=(-1)^{\sum_i a_i} \left(\sum_eJ_\eta(Z)_{(e,a),(e,b)}\right) \overline{(-1)^{\sum_i b_i}}. \end{aligned}bη​(m,ℓ;p,q)Bη​(u,a)Jη​(Z)uv​Lη​(Z)ab​Lηc​(Z)ab​​=1m+ℓ=p+q​p!q!m!ℓ!​​u=0∑p​1u≤m, m−u≤q​(up​)(m−uq​)(η​)u(1−η​)p−u+(m−u)(−η​)q−(m−u),=i∏​bη​(u1,i​,u2,i​;a1,i​,a2,i​),=a∈N×N∑​b∈N×N∑​Bη​(u,a)Za1​b1​​1a2​=0, b2​=0​Bη​(v,b)​,=e∑​Jη​(Z)(a,e),(b,e)​,=(−1)∑i​ai​(e∑​Jη​(Z)(e,a),(e,b)​)(−1)∑i​bi​​.​

Here t+=max⁡(t,0)t_+=\max(t,0)t+​=max(t,0) is embedded in [0,∞][0,\infty][0,∞], ln⁡0=0\ln0=0ln0=0, and 0⋅∞=00\cdot\infty=00⋅∞=0; complex sums are total sums, returning zero for nonsummable families. Natural subtractions are truncated at zero, and real square roots of negative numbers are zero. The parameters η,B\eta,Bη,B and the coordinates of rrr have no additional sign restrictions; in particular, the energy bound uses zero on its right when B≤0B\le0B≤0. The closure includes limits of triples whose choices of n,k,p,ρn,k,p,\rhon,k,p,ρ may vary.

QAlgorithms.Papers.Independent.Gap48.P44.lossCE

For any real numbers η,B\eta,Bη,B, this defines the set of pairs (r1,r2)∈R2(r_1,r_2)\in\mathbb R^2(r1​,r2​)∈R2 with r1,r2≥0r_1,r_2\ge0r1​,r2​≥0 for which some s∈[0,1]s\in[0,1]s∈[0,1] satisfies

r1≤H(sB)+H(ηB)−H((1−η)sB),r1≤H(ηB)−H((1−η)sB)+r2,r_1\le H(sB)+H(\eta B)-H((1-\eta)sB),\qquad r_1\le H(\eta B)-H((1-\eta)sB)+r_2,r1​≤H(sB)+H(ηB)−H((1−η)sB),r1​≤H(ηB)−H((1−η)sB)+r2​,

where

H(x)=(x+1)log⁡(x+1)−xlog⁡xlog⁡2,H(x)=\frac{(x+1)\log(x+1)-x\log x}{\log2},H(x)=log2(x+1)log(x+1)−xlogx​,

using the total real logarithm, with log⁡0=0\log0=0log0=0 and log⁡x=log⁡∣x∣\log x=\log|x|logx=log∣x∣ for x<0x<0x<0. Both endpoints for sss are included, and there are no restrictions on η\etaη or BBB.

QAlgorithms.Papers.Independent.Gap48.P44.lossCQ

For any real numbers η,B\eta,Bη,B, this defines the set of pairs (r1,r2)∈R2(r_1,r_2)\in\mathbb R^2(r1​,r2​)∈R2 with r1,r2≥0r_1,r_2\ge0r1​,r2​≥0 for which some s∈[0,1]s\in[0,1]s∈[0,1] satisfies

r2≤H(ηsB)−H((1−η)sB),r1+r2≤H(ηB)−H((1−η)sB),r_2\le H(\eta sB)-H((1-\eta)sB),\qquad r_1+r_2\le H(\eta B)-H((1-\eta)sB),r2​≤H(ηsB)−H((1−η)sB),r1​+r2​≤H(ηB)−H((1−η)sB),

where

H(x)=(x+1)log⁡(x+1)−xlog⁡xlog⁡2.H(x)=\frac{(x+1)\log(x+1)-x\log x}{\log2}.H(x)=log2(x+1)log(x+1)−xlogx​.

The real logarithm has value 000 at 000 and value log⁡∣x∣\log|x|log∣x∣ at negative xxx. The share sss may equal either endpoint; η\etaη and BBB are unrestricted real numbers, so in particular neither 0≤η≤10\le\eta\le10≤η≤1 nor B≥0B\ge0B≥0 is required.

QAlgorithms.Papers.Independent.Gap48.P44.lossComplement

For every natural number nnn, real number η\etaη, and arbitrary complex kernel ρ:In×In→C\rho:I_n\times I_n\to\mathbb Cρ:In​×In​→C, where In={0,…,n−1}→NI_n=\{0,\ldots,n-1\}\to\mathbb NIn​={0,…,n−1}→N, this defines a kernel whose entry at a,b∈Ina,b\in I_na,b∈In​ is

(−1)∑iai(∑e∈In ∑u∈In×In ∑v∈In×InBη((e,a),u) ρ(u1,v1) Vn(u2,v2) Bη((e,b),v)‾)(−1)∑ibi‾,(-1)^{\sum_i a_i} \left( \sum_{e\in I_n}\ \sum_{u\in I_n\times I_n}\ \sum_{v\in I_n\times I_n} B_\eta((e,a),u)\, \rho(u_1,v_1)\, V_n(u_2,v_2)\, \overline{B_\eta((e,b),v)} \right) \overline{(-1)^{\sum_i b_i}},(−1)∑i​ai​(e∈In​∑​ u∈In​×In​∑​ v∈In​×In​∑​Bη​((e,a),u)ρ(u1​,v1​)Vn​(u2​,v2​)Bη​((e,b),v)​)(−1)∑i​bi​​,

where Vn(c,d)=1V_n(c,d)=1Vn​(c,d)=1 if both ccc and ddd are the zero function and Vn(c,d)=0V_n(c,d)=0Vn​(c,d)=0 otherwise, and

Bη((x,y),(p,q))=∏i=0n−1βη(xi,yi,pi,qi).B_\eta((x,y),(p,q)) =\prod_{i=0}^{n-1}\beta_\eta(x_i,y_i,p_i,q_i).Bη​((x,y),(p,q))=i=0∏n−1​βη​(xi​,yi​,pi​,qi​).

The real coefficient βη(m,ℓ,p,q)\beta_\eta(m,\ell,p,q)βη​(m,ℓ,p,q) is 000 unless m+ℓ=p+qm+\ell=p+qm+ℓ=p+q; when that equality holds, it is

m! ℓ!p! q!∑u=0p{(pu)(qm−u)(η)u(1−η)p−u+(m−u)(−η)q−(m−u),u≤m and m−u≤q,0,otherwise.\sqrt{\frac{m!\,\ell!}{p!\,q!}} \sum_{u=0}^{p} \begin{cases} \displaystyle \binom pu\binom q{m-u} (\sqrt\eta)^u (\sqrt{1-\eta})^{p-u+(m-u)} (-\sqrt\eta)^{q-(m-u)}, &u\le m\ \text{and }m-u\le q,\\ 0,&\text{otherwise}. \end{cases}p!q!m!ℓ!​​u=0∑p​⎩⎨⎧​(up​)(m−uq​)(η​)u(1−η​)p−u+(m−u)(−η​)q−(m−u),0,​u≤m and m−u≤q,otherwise.​

The nested infinite sums are taken in the displayed order; each complex infinite sum is the total sum operation, which gives 000 if its summand family is not summable. No positivity, normalization, trace-class, or convergence assumption is imposed on ρ\rhoρ, and η\etaη is unrestricted; real square roots of negative arguments equal 000, natural-number subtractions truncate at 000, and zeroth powers equal 111, including 000^000. For n=0n=0n=0, the products are empty products equal to 111, and the resulting kernel equals ρ\rhoρ.

QAlgorithms.Papers.Independent.Gap48.P44.lossQuantumFamily

For arbitrary real numbers η\etaη and BBB, the loss quantum family is the set of all real triples r=(r0,r1,r2)r=(r_0,r_1,r_2)r=(r0​,r1​,r2​) for which some real s∈[0,1]s\in[0,1]s∈[0,1] satisfies

r0+2r1≤h(sB)+h(ηB)−h((1−η)sB),r1+r2≤h(ηsB)−h((1−η)sB),r0+r1+r2≤h(ηB)−h((1−η)sB),h(x)=(x+1)ln⁡(x+1)−xln⁡xln⁡2.\begin{aligned} r_0+2r_1&\le h(sB)+h(\eta B)-h((1-\eta)sB),\\ r_1+r_2&\le h(\eta sB)-h((1-\eta)sB),\\ r_0+r_1+r_2&\le h(\eta B)-h((1-\eta)sB), \end{aligned} \qquad h(x)=\frac{(x+1)\ln(x+1)-x\ln x}{\ln2}.r0​+2r1​r1​+r2​r0​+r1​+r2​​≤h(sB)+h(ηB)−h((1−η)sB),≤h(ηsB)−h((1−η)sB),≤h(ηB)−h((1−η)sB),​h(x)=ln2(x+1)ln(x+1)−xlnx​.

There are no nonnegativity requirements on the three coordinates and no restrictions on η\etaη or BBB. Consequently the formula also includes negative arguments of hhh: the real logarithm used here has ln⁡0=0\ln0=0ln0=0 and, for nonzero real xxx, equals ln⁡∣x∣\ln|x|ln∣x∣.

QAlgorithms.Papers.Independent.Gap48.P44.thermalD

For arbitrary real numbers η,s,B,N\eta,s,B,Nη,s,B,N, this defines the real number

(s(1+η)B+(1−η)N+1)2−4ηsB(sB+1).\sqrt{\bigl(s(1+\eta)B+(1-\eta)N+1\bigr)^2 -4\eta sB(sB+1)}.(s(1+η)B+(1−η)N+1)2−4ηsB(sB+1)​.

The square root is the total real square root and equals 000 when its argument is negative; no assumptions on the parameters or the radicand are required.

QAlgorithms.Papers.Independent.Gap48.P44.thermalEnvironment

For arbitrary real numbers η,s,B,N\eta,s,B,Nη,s,B,N, let

D=(s(1+η)B+(1−η)N+1)2−4ηsB(sB+1),H(x)=(x+1)log⁡(x+1)−xlog⁡xlog⁡2.D=\sqrt{\bigl(s(1+\eta)B+(1-\eta)N+1\bigr)^2-4\eta sB(sB+1)}, \qquad H(x)=\frac{(x+1)\log(x+1)-x\log x}{\log2}.D=(s(1+η)B+(1−η)N+1)2−4ηsB(sB+1)​,H(x)=log2(x+1)log(x+1)−xlogx​.

This defines the real number

H ⁣(D+s(1−η)B−(1−η)N−12)+H ⁣(D−s(1−η)B+(1−η)N−12).H\!\left(\frac{D+s(1-\eta)B-(1-\eta)N-1}{2}\right) + H\!\left(\frac{D-s(1-\eta)B+(1-\eta)N-1}{2}\right).H(2D+s(1−η)B−(1−η)N−1​)+H(2D−s(1−η)B+(1−η)N−1​).

There are no parameter restrictions: a negative radicand gives D=0D=0D=0, and the real logarithm used in HHH has log⁡0=0\log0=0log0=0 and log⁡x=log⁡∣x∣\log x=\log|x|logx=log∣x∣ for negative xxx.

Human review
  • Endorsed by Shuze Chen · Oct 22, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Goku · Oct 22, 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