Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Definitions for Second-order coding rates for pure-loss bosonic channels

Definition
QAlgorithms_Papers_Independent_Intakenext100_P76

by Gohan · Oct 31, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

The objects the formalized statements of Second-order coding rates for pure-loss bosonic channels (arXiv:1408.5328) are stated in. Each was checked against the paper by a blind read-back and a moderator.

  • IsUnitCoefficientVector: PDF p.4: unit-vector coefficients on the full countable orthonormal basis. Square summability is explicit; the zero vector is excluded.
  • coefficientInner: PDF p.4: conjugate-linear first argument, as in Mathlib's Hilbert inner product.
  • meanOutput: PDF pp.4,6, (4.1): the density is computed from the channel and input law. Coordinate measurability of pure projectors is the trace-norm Borel convention on this normalized separable-output domain. No finite-support restriction.
  • HasSpectralLaw: PDF p.4: a genuine countable spectral decomposition, allowing zero weights and zero vectors at those weights. The operator is reconstructed, not free data. All nonzero eigenvectors are normalized and mutually orthogonal.
  • lowEigenvalueMass: PDF p.4, (3.3)–(3.6): retained mass of eigenvalues at most 2^(-gamma). The weak inequality is taken from the spectral projector (3.3); eigenvalues at the cutoff are retained, as required by (3.7).
  • spectralInfEntropy: PDF p.4: spectral inf-entropy. On a probability spectrum and 0<=e<1, the defining set is nonempty and bounded above, so the real supremum is valid.
  • selfInformationMean: PDF p.5, (3.12): spectral mean self-information, used only with summability.
  • selfInformationVariance: PDF p.5, (3.13): variance computed independently of the claimed bosonic formula.
  • kernelEntropyBits: Spectral entropy of an actual positive trace-one kernel, in bits, including infinity. Spectral resolutions determine the same value. An empty set of resolutions gives infinity, not zero. PDF pp.4–5, (3.2),(3.12).
  • PureCQCode: PDF p.5: finite nonempty uniform messages, deterministic encoder and one joint POVM on the entire output system. There is no oracle-dependent decoder.
  • pureBornProbability: PDF p.5: Born probability for a normalized vector and bounded effect. Inner row sums and the outer Hilbert pairing converge absolutely on that domain.
  • pureCQAverageError: PDF p.5, (3.15): average, not maximal, message error.
  • logMaxCodeSize: PDF p.5, (3.16): supremum of achievable log cardinalities. EReal preserves an infinite unrestricted code size; all witnesses remain actual finite codes.
  • HasFiniteCode: An actual deterministic finite code attaining a real lower threshold.
  • productPureCQChannel: PDF p.9, (5.1): exact tensor-product pure outputs, in the full product basis.
  • productSpectrum: PDF p.9: iid eigenvalue law of rho tensor n, with no finite truncation.
  • thermalPhotonProbability: PDF p.10, (6.5): complete number-diagonal thermal spectrum, including vacuum.
  • thermalDispersion: PDF pp.3,10: scalar thermal dispersion, explicitly extended to vacuum. This is distinct from selfInformationVariance, to which the theorem equates it.
  • heterodyneDispersion: PDF p.11, (6.12), bits squared: log2(e)=1/ln(2).
  • coherentChannel: PDF p.10, (6.2): identity-transmissivity coherent-state cq channel.
  • photonLeakage: PDF p.11: leakage outside the total-photon subspace, averaged over uniform messages; the cutoff is ceil(n NS), not a separate cutoff on each mode.
  • bpskAverage: PDF p.13: actual equiprobable BPSK average density, not its entropy formula.
  • bpskHolevoBits: PDF p.13, (6.14): scalar BPSK Holevo expression in bits.
  • bpskHolevoNats: PDF p.13: same scalar expression converted to nats.
  • thermalEntropyNats: PDF p.13, (6.15): unconstrained-modulation Holevo expression in nats.
  • dolinarCrossover: PDF p.13, (6.17): exact binary symmetric crossover probability.
  • dolinarNats: PDF p.13: scalar BPSK/Dolinar capacity in nats, independently of the expansion.
Definition code
import Mathlib
import Definitions.Def_QAlgorithms_Defs_Information
import Definitions.Def_QAlgorithms_Papers_Independent_Information100_P63
import Definitions.Def_QAlgorithms_Papers_Independent_Information100_P54
import Definitions.Def_QAlgorithms_Papers_Independent_Dependency50_P39
import Definitions.Def_QAlgorithms_Papers_Independent_Next46_P32

/-!
arXiv:1408.5328v3. Countable coordinates describe completed separable Hilbert
spaces, including finite dimensions. Infinite matrices below are coefficient
arrays, never finite cutoffs. Positive finite forms and POVM completeness imply
bounded effects on l2; all Born pairings are iterated sums, not an unordered
non-absolutely-convergent double series. Distinct new definitions are paper-local;
accepted definitions are imported as explicitly required by the repair instruction.
The three source deferrals and the printed sign correction are documented in
the unit artifacts. Compilation is pending the batch handoff.
-/

noncomputable section

namespace QAlgorithms.Papers.Independent.Intakenext100.P76

open MeasureTheory Filter
open QAlgorithms.Papers.Independent.Information100.P63 (logTwo normalCDF normalQuantile)
open QAlgorithms.Papers.Independent.Information100.P54 (thermalEntropy)
open QAlgorithms.Papers.Independent.Dependency50.P39 (coherentCoefficient)
open QAlgorithms.Papers.Independent.Next46.P32 (PositiveKernel)
open scoped BigOperators ENNReal Topology

attribute [local instance] Classical.propDecidable
attribute [local instance] Classical.decEq

/-- PDF p.4: unit-vector coefficients on the full countable orthonormal basis.
Square summability is explicit; the zero vector is excluded. -/
def IsUnitCoefficientVector {I : Type} (u : I → ℂ) : Prop :=
  Summable (fun i => Complex.normSq (u i)) ∧ ∑' i, Complex.normSq (u i) = 1

/-- PDF p.4: conjugate-linear first argument, as in Mathlib's Hilbert inner product. -/
def coefficientInner {I : Type} (u v : I → ℂ) : ℂ :=
  ∑' i, star (u i) * v i

/-- PDF pp.4,6, (4.1): the density is computed from the channel and input law.
Coordinate measurability of pure projectors is the trace-norm Borel convention
on this normalized separable-output domain. No finite-support restriction. -/
def meanOutput {X : Type*} {I : Type} [MeasurableSpace X]
    (W : X → I → ℂ) (mu : Measure X) : Matrix I I ℂ :=
  fun a b => ∫ x, Matrix.vecMulVec (W x) (star (W x)) a b ∂mu

/-- PDF p.4: a genuine countable spectral decomposition, allowing zero weights
and zero vectors at those weights. The operator is reconstructed, not free data.
All nonzero eigenvectors are normalized and mutually orthogonal. -/
def HasSpectralLaw {I : Type} {J : Type*} (rho : Matrix I I ℂ) (p : J → ℝ) : Prop :=
  (∀ j, 0 ≤ p j) ∧ Summable p ∧ (∑' j, p j) = 1 ∧
    ∃ u : J → I → ℂ,
      (∀ j, 0 < p j → IsUnitCoefficientVector (u j)) ∧
      (∀ j k, j ≠ k → 0 < p j → 0 < p k → coefficientInner (u j) (u k) = 0) ∧
      ∀ a b, rho a b = ∑' j, (p j : ℂ) * Matrix.vecMulVec (u j) (star (u j)) a b

/-- PDF p.4, (3.3)–(3.6): retained mass of eigenvalues at most 2^(-gamma).
The weak inequality is taken from the spectral projector (3.3); eigenvalues
at the cutoff are retained, as required by (3.7). -/
def lowEigenvalueMass {J : Type*} (p : J → ℝ) (gamma : ℝ) : ℝ :=
  ∑' j, if p j ≤ (2 : ℝ) ^ (-gamma) then p j else 0

/-- PDF p.4: spectral inf-entropy. On a probability spectrum and 0<=e<1,
the defining set is nonempty and bounded above, so the real supremum is valid. -/
def spectralInfEntropy {J : Type*} (p : J → ℝ) (e : ℝ) : ℝ :=
  sSup {gamma : ℝ | 1 - e ≤ lowEigenvalueMass p gamma}

/-- PDF p.5, (3.12): spectral mean self-information, used only with summability. -/
def selfInformationMean {J : Type*} (p : J → ℝ) : ℝ :=
  ∑' j, p j * (-logTwo (p j))

/-- PDF p.5, (3.13): variance computed independently of the claimed bosonic formula. -/
def selfInformationVariance {J : Type*} (p : J → ℝ) : ℝ :=
  ∑' j, p j * (-logTwo (p j) - selfInformationMean p) ^ 2

/-- Spectral entropy of an actual positive trace-one kernel, in bits, including
infinity. Spectral resolutions determine the same value. An empty set of
resolutions gives infinity, not zero. PDF pp.4–5, (3.2),(3.12). -/
def kernelEntropyBits {I : Type} (rho : Matrix I I ℂ) : ℝ≥0∞ :=
  sInf {z : ℝ≥0∞ | ∃ p : ℕ → ℝ, HasSpectralLaw rho p ∧
    z = ∑' j, ENNReal.ofReal (-p j * logTwo (p j))}

/-- PDF p.5: finite nonempty uniform messages, deterministic encoder and one
joint POVM on the entire output system. There is no oracle-dependent decoder. -/
structure PureCQCode (X : Type*) (I : Type) (M : ℕ) where
  nonempty : 0 < M
  encode : Fin M → X
  effect : Fin M → Matrix I I ℂ
  positive : ∀ m, PositiveKernel (effect m)
  complete : ∀ a b, (∑ m, effect m a b) = if a = b then 1 else 0

/-- PDF p.5: Born probability for a normalized vector and bounded effect.
Inner row sums and the outer Hilbert pairing converge absolutely on that domain. -/
def pureBornProbability {I : Type} (u : I → ℂ) (E : Matrix I I ℂ) : ℝ :=
  (∑' a, star (u a) * (∑' b, E a b * u b)).re

/-- PDF p.5, (3.15): average, not maximal, message error. -/
def pureCQAverageError {X : Type*} {I : Type} {M : ℕ} (W : X → I → ℂ) (c : PureCQCode X I M) : ℝ :=
  1 - (∑ m, pureBornProbability (W (c.encode m)) (c.effect m)) / (M : ℝ)

/-- PDF p.5, (3.16): supremum of achievable log cardinalities. EReal preserves
an infinite unrestricted code size; all witnesses remain actual finite codes. -/
def logMaxCodeSize {X : Type*} {I : Type} (W : X → I → ℂ) (e : ℝ) : EReal :=
  sSup {r : EReal | ∃ (M : ℕ) (c : PureCQCode X I M),
    pureCQAverageError W c ≤ e ∧ r = (logTwo (M : ℝ) : EReal)}

/-- An actual deterministic finite code attaining a real lower threshold. -/
def HasFiniteCode {X : Type*} {I : Type} (W : X → I → ℂ) (e R : ℝ) : Prop :=
  ∃ (M : ℕ) (c : PureCQCode X I M), pureCQAverageError W c ≤ e ∧ R ≤ logTwo (M : ℝ)

/-- PDF p.9, (5.1): exact tensor-product pure outputs, in the full product basis. -/
def productPureCQChannel {X : Type*} {I : Type} (W : X → I → ℂ) (n : ℕ) :
    (Fin n → X) → (Fin n → I) → ℂ :=
  fun x a => ∏ i, W (x i) (a i)

/-- PDF p.9: iid eigenvalue law of rho tensor n, with no finite truncation. -/
def productSpectrum {J : Type*} (p : J → ℝ) (n : ℕ) : (Fin n → J) → ℝ :=
  fun a => ∏ i, p (a i)

/-- PDF p.10, (6.5): complete number-diagonal thermal spectrum, including vacuum. -/
def thermalPhotonProbability (x : ℝ) (k : ℕ) : ℝ := x ^ k / (x + 1) ^ (k + 1)

/-- PDF pp.3,10: scalar thermal dispersion, explicitly extended to vacuum.
This is distinct from selfInformationVariance, to which the theorem equates it. -/
def thermalDispersion (x : ℝ) : ℝ :=
  if x = 0 then 0 else x * (x + 1) * (logTwo (x + 1) - logTwo x) ^ 2

/-- PDF p.11, (6.12), bits squared: log2(e)=1/ln(2). -/
def heterodyneDispersion (x : ℝ) : ℝ :=
  x * (x + 2) * (1 / Real.log 2) ^ 2 / (x + 1) ^ 2

/-- PDF p.10, (6.2): identity-transmissivity coherent-state cq channel. -/
def coherentChannel (n : ℕ) : (Fin n → ℂ) → (Fin n → ℕ) → ℂ :=
  productPureCQChannel coherentCoefficient n

/-- PDF p.11: leakage outside the total-photon subspace, averaged over uniform
messages; the cutoff is ceil(n NS), not a separate cutoff on each mode. -/
def photonLeakage {n M : ℕ} (c : PureCQCode (Fin n → ℂ) (Fin n → ℕ) M) (NS : ℝ) : ℝ :=
  (∑ m, ∑' a : Fin n → ℕ,
    if ⌈(n : ℝ) * NS⌉₊ < ∑ i, a i then
      Complex.normSq (coherentChannel n (c.encode m) a) else 0) / (M : ℝ)

/-- PDF p.13: actual equiprobable BPSK average density, not its entropy formula. -/
def bpskAverage (alpha : ℂ) : Matrix ℕ ℕ ℂ :=
  fun a b => (Matrix.vecMulVec (coherentCoefficient alpha) (star (coherentCoefficient alpha)) a b +
    Matrix.vecMulVec (coherentCoefficient (-alpha)) (star (coherentCoefficient (-alpha))) a b) / 2

/-- PDF p.13, (6.14): scalar BPSK Holevo expression in bits. -/
def bpskHolevoBits (x : ℝ) : ℝ :=
  QAlgorithms.binEntropy2 ((1 - Real.exp (-2 * x)) / 2)

/-- PDF p.13: same scalar expression converted to nats. -/
def bpskHolevoNats (x : ℝ) : ℝ := Real.log 2 * bpskHolevoBits x

/-- PDF p.13, (6.15): unconstrained-modulation Holevo expression in nats. -/
def thermalEntropyNats (x : ℝ) : ℝ := Real.log 2 * thermalEntropy x

/-- PDF p.13, (6.17): exact binary symmetric crossover probability. -/
def dolinarCrossover (x : ℝ) : ℝ :=
  (1 - Real.sqrt (1 - Real.exp (-4 * x))) / 2

/-- PDF p.13: scalar BPSK/Dolinar capacity in nats, independently of the expansion. -/
def dolinarNats (x : ℝ) : ℝ :=
  Real.log 2 * (1 - QAlgorithms.binEntropy2 (dolinarCrossover x))

/-! Sanity checks on zero weights, zero energy, tensor products and normalization. -/

example : logTwo 1 = 0 := by simp [logTwo]

example : ¬ IsUnitCoefficientVector (fun _ : ℕ => (0 : ℂ)) := by simp [IsUnitCoefficientVector]

example : Matrix.vecMulVec (fun _ : ℕ => (0 : ℂ))
    (star (fun _ : ℕ => (0 : ℂ))) = 0 := by ext a b; simp [Matrix.vecMulVec]

example : IsUnitCoefficientVector (fun k : ℕ => if k = 0 then (1 : ℂ) else 0) := by
  constructor
  · apply summable_of_ne_finset_zero (s := {0})
    intro k hk
    have hk0 : k ≠ 0 := by simpa using hk
    simp [hk0]
  · simp

example : pureBornProbability (fun _ : Fin 1 => (1 : ℂ)) (fun _ _ => (1 : ℂ)) = 1 := by
  simp [pureBornProbability]

example : thermalPhotonProbability 0 0 = 1 := by simp [thermalPhotonProbability]

example (k : ℕ) : thermalPhotonProbability 0 (k + 1) = 0 := by simp [thermalPhotonProbability]

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

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

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

example : coherentCoefficient 0 0 = 1 := by simp [coherentCoefficient]

example : bpskHolevoBits 0 = 0 := by
  simp [bpskHolevoBits, QAlgorithms.binEntropy2]

example : dolinarCrossover 0 = 1 / 2 := by simp [dolinarCrossover]

example : dolinarNats 0 = 0 := by
  rw [dolinarNats, show dolinarCrossover 0 = 1 / 2 by simp [dolinarCrossover]]
  have hb : QAlgorithms.binEntropy2 (1 / 2) = 1 := by
    rw [QAlgorithms.binEntropy2, show (1 / 2 : ℝ) = 2⁻¹ by norm_num,
      Real.binEntropy_two_inv]
    exact div_self (Real.log_pos (by norm_num : (1 : ℝ) < 2)).ne'
  rw [hb]
  simp

example {J : Type*} (p : J → ℝ) (a : Fin 0 → J) : productSpectrum p 0 a = 1 := by
  simp [productSpectrum]

example {X : Type*} {I : Type} (W : X → I → ℂ) (x : Fin 0 → X) (a : Fin 0 → I) :
    productPureCQChannel W 0 x a = 1 := by simp [productPureCQChannel]

end QAlgorithms.Papers.Independent.Intakenext100.P76
Source
Second-order coding rates for pure-loss bosonic channels, arXiv:1408.5328v3, https://arxiv.org/abs/1408.5328v3
Read-back

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

QAlgorithms.Papers.Independent.Intakenext100.P76.HasFiniteCode

For arbitrary types XXX and III, W:X→(I→C)W:X\to(I\to\mathbb C)W:X→(I→C), and real numbers e,Re,Re,R, this is the proposition that there exist a natural number M>0M>0M>0, encoded inputs cm∈Xc_m\in Xcm​∈X for m=0,…,M−1m=0,\ldots,M-1m=0,…,M−1, and complex matrices EmE_mEm​ indexed by III, such that

1−1M∑m=0M−1Re⁡ ⁣(∑a:IW(cm)(a)‾(∑b:IEm(a,b)W(cm)(b)))≤e,R≤ln⁡Mln⁡2.1-\frac1M\sum_{m=0}^{M-1} \operatorname{Re}\!\left(\sum_{a:I}\overline{W(c_m)(a)} \left(\sum_{b:I}E_m(a,b)W(c_m)(b)\right)\right)\le e, \qquad R\le\frac{\ln M}{\ln2}.1−M1​m=0∑M−1​Re(a:I∑​W(cm​)(a)​(b:I∑​Em​(a,b)W(cm​)(b)))≤e,R≤ln2lnM​.

The matrices must satisfy ∑mEm(a,b)=1a=b\sum_m E_m(a,b)=\mathbf1_{a=b}∑m​Em​(a,b)=1a=b​, and every finite quadratic expression ∑a,b∈sz(a)‾Em(a,b)z(b)\sum_{a,b\in s}\overline{z(a)}E_m(a,b)z(b)∑a,b∈s​z(a)​Em​(a,b)z(b), for finite s⊆Is\subseteq Is⊆I and z:I→Cz:I\to\mathbb Cz:I→C, must have zero imaginary part and nonnegative real part. There is no normalization assumption on WWW, no restriction on eee or RRR, and no summability assumption on the nested infinite sums; a nonsummable family has total infinite sum zero.

QAlgorithms.Papers.Independent.Intakenext100.P76.HasSpectralLaw

For arbitrary index types I,JI,JI,J, a complex matrix ρ\rhoρ indexed by III, and a function p:J→Rp:J\to\mathbb Rp:J→R, this predicate requires pj≥0p_j\ge0pj​≥0 for every jjj, summability of ppp, and ∑jpj=1\sum_jp_j=1∑j​pj​=1. It further requires the existence of functions uj:I→Cu_j:I\to\mathbb Cuj​:I→C such that, whenever pj>0p_j>0pj​>0, the family (∣uj(a)∣2)a∈I(|u_j(a)|^2)_{a\in I}(∣uj​(a)∣2)a∈I​ is summable with sum one; whenever j≠kj\ne kj=k and both pj,pk>0p_j,p_k>0pj​,pk​>0, ∑auj(a)‾uk(a)=0\sum_a\overline{u_j(a)}u_k(a)=0∑a​uj​(a)​uk​(a)=0; and, for every a,b∈Ia,b\in Ia,b∈I, ρab=∑jpjuj(a)uj(b)‾\rho_{ab}=\sum_jp_j u_j(a)\overline{u_j(b)}ρab​=∑j​pj​uj​(a)uj​(b)​. Vectors at zero-weight indices have no normalization or orthogonality requirement. No finiteness or countability assumption is imposed on either index type, and no separate summability condition is stated for the matrix-entry expansions; the complex infinite sums use the total convention of zero for nonsummable families. An empty JJJ cannot satisfy the required total weight.

QAlgorithms.Papers.Independent.Intakenext100.P76.IsUnitCoefficientVector

For any index type III and function u:I→Cu:I\to\mathbb Cu:I→C, this predicate means that the family (∣ui∣2)i∈I(|u_i|^2)_{i\in I}(∣ui​∣2)i∈I​ is summable and ∑i∈I∣ui∣2=1\sum_{i\in I}|u_i|^2=1∑i∈I​∣ui​∣2=1. No finiteness or countability assumption on III is imposed. If III is empty, the predicate cannot hold because its sum is zero.

QAlgorithms.Papers.Independent.Intakenext100.P76.PureCQCode

For arbitrary types X,IX,IX,I and a natural number MMM, this structure consists of a proof that M>0M>0M>0, an encoding function from {0,…,M−1}\{0,\ldots,M-1\}{0,…,M−1} to XXX, and a complex matrix EmE_mEm​ indexed by III for each message mmm. Each EmE_mEm​ must satisfy finite quadratic-form positivity: for every finite subset s⊆Is\subseteq Is⊆I and every function z:I→Cz:I\to\mathbb Cz:I→C, the number ∑a,b∈sza‾(Em)abzb\sum_{a,b\in s}\overline{z_a}(E_m)_{ab}z_b∑a,b∈s​za​​(Em​)ab​zb​ has imaginary part zero and nonnegative real part. Completeness is required entrywise: ∑m=0M−1(Em)ab\sum_{m=0}^{M-1}(E_m)_{ab}∑m=0M−1​(Em​)ab​ equals one if a=ba=ba=b and zero otherwise, for every a,b∈Ia,b\in Ia,b∈I. Neither index type is required to be finite or countable, and the encoder need not be injective. No structure exists for M=0M=0M=0 or empty XXX; empty III is allowed, in which case the matrix conditions are vacuous.

QAlgorithms.Papers.Independent.Intakenext100.P76.bpskAverage

For every complex α\alphaα, this defines the complex matrix indexed by a,b∈Na,b\in\mathbb Na,b∈N with entries

12(cα(a) cα(b)‾+c−α(a) c−α(b)‾),cγ(k)=exp⁡(−∣γ∣2/2) γkk!.\frac12\left( c_\alpha(a)\,\overline{c_\alpha(b)} +c_{-\alpha}(a)\,\overline{c_{-\alpha}(b)} \right), \qquad c_\gamma(k)= \frac{\exp(-|\gamma|^2/2)\,\gamma^k}{\sqrt{k!}}.21​(cα​(a)cα​(b)​+c−α​(a)c−α​(b)​),cγ​(k)=k!​exp(−∣γ∣2/2)γk​.

The parameter α\alphaα is unrestricted, including α=0\alpha=0α=0; powers use the convention 00=10^0=100=1.

QAlgorithms.Papers.Independent.Intakenext100.P76.bpskHolevoBits

For every real xxx, set p=(1−exp⁡(−2x))/2p=(1-\exp(-2x))/2p=(1−exp(−2x))/2; this defines

−pln⁡p−(1−p)ln⁡(1−p)ln⁡2.\frac{-p\ln p-(1-p)\ln(1-p)}{\ln2}.ln2−plnp−(1−p)ln(1−p)​.

No assumption that x≥0x\ge0x≥0, or that p∈[0,1]p\in[0,1]p∈[0,1], is imposed. The logarithms are total real logarithms, with ln⁡0=0\ln0=0ln0=0 and ln⁡t=ln⁡∣t∣\ln t=\ln|t|lnt=ln∣t∣ for t<0t<0t<0, so the formula remains defined outside the usual probability range; its value at x=0x=0x=0 is zero.

QAlgorithms.Papers.Independent.Intakenext100.P76.bpskHolevoNats

For every real xxx, with p=(1−exp⁡(−2x))/2p=(1-\exp(-2x))/2p=(1−exp(−2x))/2, this defines

(ln⁡2) −pln⁡p−(1−p)ln⁡(1−p)ln⁡2.(\ln2)\,\frac{-p\ln p-(1-p)\ln(1-p)}{\ln2}.(ln2)ln2−plnp−(1−p)ln(1−p)​.

There is no nonnegativity restriction on xxx or probability-range restriction on ppp. Real logarithms are total, with ln⁡0=0\ln0=0ln0=0 and ln⁡t=ln⁡∣t∣\ln t=\ln|t|lnt=ln∣t∣ for negative ttt; in particular, the value at x=0x=0x=0 is zero.

QAlgorithms.Papers.Independent.Intakenext100.P76.coefficientInner

For any index type III and functions u,v:I→Cu,v:I\to\mathbb Cu,v:I→C, this defines the complex number ∑i∈Iui‾vi\sum_{i\in I}\overline{u_i}v_i∑i∈I​ui​​vi​. Neither normalization nor summability is assumed; the sum is the total infinite-sum operation, whose value is zero when the family is not summable. For empty III, its value is also zero.

QAlgorithms.Papers.Independent.Intakenext100.P76.coherentChannel

For every natural number nnn, this defines a function taking α=(α0,…,αn−1)∈Cn\alpha=(\alpha_0,\ldots,\alpha_{n-1})\in\mathbb C^nα=(α0​,…,αn−1​)∈Cn and a=(a0,…,an−1)∈Nna=(a_0,\ldots,a_{n-1})\in\mathbb N^na=(a0​,…,an−1​)∈Nn to

∏i=0n−1exp⁡(−∣αi∣2/2) αiaiai!.\prod_{i=0}^{n-1} \frac{\exp(-|\alpha_i|^2/2)\,\alpha_i^{a_i}}{\sqrt{a_i!}}.i=0∏n−1​ai​!​exp(−∣αi​∣2/2)αiai​​​.

All real factors are regarded as complex numbers. The amplitudes αi\alpha_iαi​ are unrestricted; when n=0n=0n=0, the value on the unique pair of empty tuples is 111.

QAlgorithms.Papers.Independent.Intakenext100.P76.dolinarCrossover

For every real number xxx, this defines

r(x)=dolinarCrossover⁡(x)=1−1−e−4x2.r(x)=\operatorname{dolinarCrossover}(x) =\frac{1-\sqrt{1-e^{-4x}}}{2}.r(x)=dolinarCrossover(x)=21−1−e−4x​​.

The real square root is defined to be 000 on negative arguments, so the definition includes x<0x<0x<0, for which r(x)=1/2r(x)=1/2r(x)=1/2, as well as x=0x=0x=0, for which r(0)=1/2r(0)=1/2r(0)=1/2.

QAlgorithms.Papers.Independent.Intakenext100.P76.dolinarNats

For every real number xxx, this defines

dolinarNats⁡(x)=ln⁡2 (1−H2(r(x))),r(x)=1−1−e−4x2,\operatorname{dolinarNats}(x)=\ln 2\,\bigl(1-H_2(r(x))\bigr), \qquad r(x)=\frac{1-\sqrt{1-e^{-4x}}}{2},dolinarNats(x)=ln2(1−H2​(r(x))),r(x)=21−1−e−4x​​,

where

H2(t)=−tln⁡t−(1−t)ln⁡(1−t)ln⁡2.H_2(t)=\frac{-t\ln t-(1-t)\ln(1-t)}{\ln2}.H2​(t)=ln2−tlnt−(1−t)ln(1−t)​.

The logarithm and square root are total real functions, with ln⁡0=0\ln0=0ln0=0 and t=0\sqrt t=0t​=0 for t<0t<0t<0. No nonnegativity hypothesis on xxx is imposed; for x≤0x\le0x≤0, the defined value is 000.

QAlgorithms.Papers.Independent.Intakenext100.P76.heterodyneDispersion

For every real xxx, this defines

x(x+2)(x+1)2(1ln⁡2)2.\frac{x(x+2)}{(x+1)^2}\left(\frac1{\ln2}\right)^2.(x+1)2x(x+2)​(ln21​)2.

There is no assumption that xxx is nonnegative. Division by zero is defined to be zero, so the value at x=−1x=-1x=−1 is zero.

QAlgorithms.Papers.Independent.Intakenext100.P76.kernelEntropyBits

For any index type III and complex matrix ρ\rhoρ indexed by III, this defines an element of [0,+∞][0,+\infty][0,+∞] as the infimum of ∑j∈Nmax⁡{−pjlog⁡2pj,0}\sum_{j\in\mathbb N}\max\{-p_j\log_2p_j,0\}∑j∈N​max{−pj​log2​pj​,0} over all real sequences ppp satisfying the following conditions: pj≥0p_j\ge0pj​≥0 for every jjj, ppp is summable with sum one, and there exist functions uj:I→Cu_j:I\to\mathbb Cuj​:I→C such that every positive-weight uju_juj​ has summable squared moduli totaling one, distinct positive-weight vectors satisfy ∑auj(a)‾uk(a)=0\sum_a\overline{u_j(a)}u_k(a)=0∑a​uj​(a)​uk​(a)=0, and ρab=∑jpjuj(a)uj(b)‾\rho_{ab}=\sum_jp_j u_j(a)\overline{u_j(b)}ρab​=∑j​pj​uj​(a)uj​(b)​ for every a,b∈Ia,b\in Ia,b∈I. Here log⁡2x=ln⁡x/ln⁡2\log_2x=\ln x/\ln2log2​x=lnx/ln2, with ln⁡0=0\ln0=0ln0=0; the maximum denotes conversion into the nonnegative extended reals, and the entropy sum may be infinite. Zero-weight vectors are unrestricted, the matrix-entry sums are total complex infinite sums, and no independent density assumption on ρ\rhoρ is supplied. If no such sequence and vectors exist, the infimum is +∞+\infty+∞.

QAlgorithms.Papers.Independent.Intakenext100.P76.logMaxCodeSize

For arbitrary types XXX and III, a function W:X→(I→C)W:X\to(I\to\mathbb C)W:X→(I→C), and e∈Re\in\mathbb Re∈R, this defines the extended real number

sup⁡{ln⁡Mln⁡2: M∈N, ∃c, 1−1M∑m=0M−1Re⁡ ⁣(∑a:IW(cm)(a)‾(∑b:IEm(a,b)W(cm)(b)))≤e}.\sup\left\{\frac{\ln M}{\ln 2}:\ M\in\mathbb N,\ \exists c,\ 1-\frac1M\sum_{m=0}^{M-1} \operatorname{Re}\!\left(\sum_{a:I}\overline{W(c_m)(a)} \left(\sum_{b:I}E_m(a,b)W(c_m)(b)\right)\right)\le e\right\}.sup{ln2lnM​: M∈N, ∃c, 1−M1​m=0∑M−1​Re(a:I∑​W(cm​)(a)​(b:I∑​Em​(a,b)W(cm​)(b)))≤e}.

Here a code ccc consists of M>0M>0M>0, an encoding m↦cm∈Xm\mapsto c_m\in Xm↦cm​∈X, and complex matrices EmE_mEm​ indexed by III, satisfying ∑mEm(a,b)=1a=b\sum_m E_m(a,b)=\mathbf1_{a=b}∑m​Em​(a,b)=1a=b​ for all a,ba,ba,b, and, for every mmm, every finite s⊆Is\subseteq Is⊆I, and every z:I→Cz:I\to\mathbb Cz:I→C, the quadratic expression ∑a,b∈sz(a)‾Em(a,b)z(b)\sum_{a,b\in s}\overline{z(a)}E_m(a,b)z(b)∑a,b∈s​z(a)​Em​(a,b)z(b) has zero imaginary part and nonnegative real part. No normalization of W(x)W(x)W(x), finiteness of III, or convergence of the displayed infinite sums is assumed; these sums use the total infinite-sum operation, which returns zero for a nonsummable family. The supremum is taken in the extended reals, so an empty admissible set gives −∞-\infty−∞, and a set unbounded above gives +∞+\infty+∞.

QAlgorithms.Papers.Independent.Intakenext100.P76.lowEigenvalueMass

For any index type JJJ, function p:J→Rp:J\to\mathbb Rp:J→R, and real number γ\gammaγ, this defines ∑j∈J,  pj≤2−γpj\sum_{j\in J,\;p_j\le 2^{-\gamma}}p_j∑j∈J,pj​≤2−γ​pj​, with equality included at the threshold. No nonnegativity, normalization, summability, or spectral interpretation of ppp is required. The sum is zero if the selected family is nonsummable, and is also zero when JJJ is empty.

QAlgorithms.Papers.Independent.Intakenext100.P76.meanOutput

For any measurable space XXX, index type III, function W:X→(I→C)W:X\to(I\to\mathbb C)W:X→(I→C), and measure μ\muμ on XXX, this defines the matrix indexed by III with entries (meanOutput⁡(W,μ))ab=∫XW(x)a W(x)b‾ dμ(x)(\operatorname{meanOutput}(W,\mu))_{ab}=\int_X W(x)_a\,\overline{W(x)_b}\,d\mu(x)(meanOutput(W,μ))ab​=∫X​W(x)a​W(x)b​​dμ(x) for every a,b∈Ia,b\in Ia,b∈I. There is no finiteness assumption on III, no normalization condition on W(x)W(x)W(x), and no requirement that μ\muμ be a probability measure or that the integrands be measurable or integrable. Each integral is the total Bochner integral, which is zero when its integrand is not integrable.

QAlgorithms.Papers.Independent.Intakenext100.P76.photonLeakage

For natural numbers n,Mn,Mn,M, a code ccc of the specified kind, and any real NSN_SNS​, this defines

1M∑m=0M−1∑a∈Nn1{⌈nNS⌉+<∑i=0n−1ai}∣∏i=0n−1exp⁡(−∣αm,i∣2/2) αm,iaiai!∣2,\frac1M\sum_{m=0}^{M-1} \sum_{a\in\mathbb N^n} \mathbf1_{\{\lceil nN_S\rceil_+<\sum_{i=0}^{n-1}a_i\}} \left| \prod_{i=0}^{n-1} \frac{\exp(-|\alpha_{m,i}|^2/2)\,\alpha_{m,i}^{a_i}}{\sqrt{a_i!}} \right|^2,M1​m=0∑M−1​a∈Nn∑​1{⌈nNS​⌉+​<∑i=0n−1​ai​}​​i=0∏n−1​ai​!​exp(−∣αm,i​∣2/2)αm,iai​​​​2,

where αm=c.encode(m)∈Cn\alpha_m=c.\mathrm{encode}(m)\in\mathbb C^nαm​=c.encode(m)∈Cn, and ⌈t⌉+\lceil t\rceil_+⌈t⌉+​ is the natural-number ceiling, equal to zero for t≤0t\le0t≤0. The code includes M>0M>0M>0 and matrices EmE_mEm​ indexed by Nn\mathbb N^nNn whose sum is the identity kernel and whose quadratic expressions on every finite subset have zero imaginary part and nonnegative real part; these matrices do not enter this formula. There is no restriction on NSN_SNS​ or the encoded amplitudes. The inner sum is the total real infinite sum, returning zero if nonsummable. For n=0n=0n=0, the strict threshold condition is false and the defined value is zero.

QAlgorithms.Papers.Independent.Intakenext100.P76.productPureCQChannel

For arbitrary types X,IX,IX,I, any function W:X→(I→C)W:X\to(I\to\mathbb C)W:X→(I→C), and every natural number nnn, this defines a function taking tuples x=(x0,…,xn−1)∈Xnx=(x_0,\ldots,x_{n-1})\in X^nx=(x0​,…,xn−1​)∈Xn and a=(a0,…,an−1)∈Ina=(a_0,\ldots,a_{n-1})\in I^na=(a0​,…,an−1​)∈In to

∏i=0n−1W(xi)(ai).\prod_{i=0}^{n-1}W(x_i)(a_i).i=0∏n−1​W(xi​)(ai​).

No normalization or other condition on WWW is required. When n=0n=0n=0, both tuples are the unique empty tuples and the value is 111.

QAlgorithms.Papers.Independent.Intakenext100.P76.productSpectrum

For an arbitrary type JJJ, any real-valued function p:J→Rp:J\to\mathbb Rp:J→R, and every natural number nnn, this defines the function on tuples a∈Jna\in J^na∈Jn by

a⟼∏i=0n−1p(ai).a\longmapsto\prod_{i=0}^{n-1}p(a_i).a⟼i=0∏n−1​p(ai​).

There is no nonnegativity, summability, or normalization requirement on ppp. For n=0n=0n=0, the value on the unique empty tuple is 111, including when JJJ is empty.

QAlgorithms.Papers.Independent.Intakenext100.P76.pureBornProbability

For any index type III, function u:I→Cu:I\to\mathbb Cu:I→C, and complex matrix EEE indexed by III, this defines Re⁡ ⁣(∑aua‾(∑bEabub))\operatorname{Re}\!\left(\sum_a\overline{u_a}\left(\sum_bE_{ab}u_b\right)\right)Re(∑a​ua​​(∑b​Eab​ub​)), with the inner sum over bbb taken separately for each aaa, followed by the outer sum over aaa. No normalization of uuu, positivity or completeness condition on EEE, or summability assumption is imposed. Each complex sum is the total infinite-sum operation, returning zero for a nonsummable family; the definition does not itself assert that the result lies between zero and one. For empty III, the result is zero.

QAlgorithms.Papers.Independent.Intakenext100.P76.pureCQAverageError

For arbitrary types X,IX,IX,I, a natural number MMM, a function W:X→(I→C)W:X\to(I\to\mathbb C)W:X→(I→C), and a code consisting of M>0M>0M>0, an encoder c:{0,…,M−1}→Xc:\{0,\ldots,M-1\}\to Xc:{0,…,M−1}→X, and matrices EmE_mEm​ indexed by III, this defines 1−1M∑m=0M−1Re⁡ ⁣(∑aW(c(m))a‾(∑b(Em)abW(c(m))b))1-\frac1M\sum_{m=0}^{M-1}\operatorname{Re}\!\left(\sum_a\overline{W(c(m))_a}\left(\sum_b(E_m)_{ab}W(c(m))_b\right)\right)1−M1​∑m=0M−1​Re(∑a​W(c(m))a​​(∑b​(Em​)ab​W(c(m))b​)). The code requires each finite quadratic form ∑a,b∈sza‾(Em)abzb\sum_{a,b\in s}\overline{z_a}(E_m)_{ab}z_b∑a,b∈s​za​​(Em​)ab​zb​ to be real and nonnegative for every finite s⊆Is\subseteq Is⊆I and function z:I→Cz:I\to\mathbb Cz:I→C, and requires ∑m(Em)ab\sum_m(E_m)_{ab}∑m​(Em​)ab​ to equal one when a=ba=ba=b and zero otherwise. No normalization or summability condition is imposed on WWW, and the nested complex sums use zero for nonsummable families, so the definition does not itself assert an error between zero and one. The denominator is nonzero because the code requires M>0M>0M>0; if III is empty, the defined error is one.

QAlgorithms.Papers.Independent.Intakenext100.P76.selfInformationMean

For any index type JJJ and function p:J→Rp:J\to\mathbb Rp:J→R, this defines ∑jpj(−log⁡2pj)\sum_jp_j\bigl(-\log_2 p_j\bigr)∑j​pj​(−log2​pj​), where log⁡2x=ln⁡x/ln⁡2\log_2 x=\ln x/\ln2log2​x=lnx/ln2. No nonnegativity, normalization, or summability hypothesis is imposed. The real logarithm is total, with ln⁡0=0\ln0=0ln0=0 and ln⁡x=ln⁡∣x∣\ln x=\ln|x|lnx=ln∣x∣ for negative xxx, so zero weights contribute zero and negative weights are permitted. The result is zero when the summand family is nonsummable, and also when JJJ is empty.

QAlgorithms.Papers.Independent.Intakenext100.P76.selfInformationVariance

For any index type JJJ and function p:J→Rp:J\to\mathbb Rp:J→R, put m=∑kpk(−log⁡2pk)m=\sum_kp_k(-\log_2p_k)m=∑k​pk​(−log2​pk​), where log⁡2x=ln⁡x/ln⁡2\log_2x=\ln x/\ln2log2​x=lnx/ln2. This defines ∑jpj(−log⁡2pj−m)2\sum_jp_j(-\log_2p_j-m)^2∑j​pj​(−log2​pj​−m)2. No nonnegativity, normalization, or summability hypothesis is imposed, so the definition does not itself require a nonnegative result. The real logarithm uses ln⁡0=0\ln0=0ln0=0 and ln⁡x=ln⁡∣x∣\ln x=\ln|x|lnx=ln∣x∣ for negative xxx; each real infinite sum is zero if its family is nonsummable. For empty JJJ, the result is zero.

QAlgorithms.Papers.Independent.Intakenext100.P76.spectralInfEntropy

For any index type JJJ, function p:J→Rp:J\to\mathbb Rp:J→R, and real number eee, this defines the real supremum sup⁡{γ∈R:1−e≤∑j∈J,  pj≤2−γpj}\sup\{\gamma\in\mathbb R:1-e\le\sum_{j\in J,\;p_j\le2^{-\gamma}}p_j\}sup{γ∈R:1−e≤∑j∈J,pj​≤2−γ​pj​}. There are no restrictions on eee or assumptions that ppp is nonnegative, normalized, summable, or a spectral law. Each sum uses the total real infinite-sum convention, returning zero for a nonsummable family. The real supremum operation returns zero if the displayed set is empty or unbounded above, rather than an extended-real value.

QAlgorithms.Papers.Independent.Intakenext100.P76.thermalDispersion

For every real xxx, this defines

{0,x=0,x(x+1)(ln⁡(x+1)ln⁡2−ln⁡xln⁡2)2,x≠0.\begin{cases} 0,&x=0,\\[2mm] x(x+1)\left(\dfrac{\ln(x+1)}{\ln2}-\dfrac{\ln x}{\ln2}\right)^2,&x\ne0. \end{cases}⎩⎨⎧​0,x(x+1)(ln2ln(x+1)​−ln2lnx​)2,​x=0,x=0.​

No positivity assumption on xxx is imposed. The real logarithm is total: ln⁡0=0\ln0=0ln0=0, and for negative arguments it equals the logarithm of their absolute value; consequently the expression is also defined for negative xxx.

QAlgorithms.Papers.Independent.Intakenext100.P76.thermalEntropyNats

For every real number xxx, this defines

thermalEntropyNats⁡(x)=ln⁡2  (x+1)ln⁡(x+1)−xln⁡xln⁡2=(x+1)ln⁡(x+1)−xln⁡x.\operatorname{thermalEntropyNats}(x) =\ln 2\;\frac{(x+1)\ln(x+1)-x\ln x}{\ln 2} =(x+1)\ln(x+1)-x\ln x.thermalEntropyNats(x)=ln2ln2(x+1)ln(x+1)−xlnx​=(x+1)ln(x+1)−xlnx.

There is no assumption that x≥0x\ge0x≥0. The logarithm is the total real logarithm, with ln⁡0=0\ln 0=0ln0=0 and ln⁡t=ln⁡∣t∣\ln t=\ln|t|lnt=ln∣t∣ for negative ttt; in particular, the value at x=0x=0x=0 is 000.

QAlgorithms.Papers.Independent.Intakenext100.P76.thermalPhotonProbability

For every real xxx and natural number kkk, this defines

xk(x+1)k+1.\frac{x^k}{(x+1)^{k+1}}.(x+1)k+1xk​.

There is no assumption that x≥0x\ge0x≥0. Division is total, with division by zero equal to zero, so the value is zero at x=−1x=-1x=−1. At x=0x=0x=0, the value is 111 for k=0k=0k=0 and zero for k>0k>0k>0, using 00=10^0=100=1.

Human review
  • Endorsed by Shuze Chen · Oct 31, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Gohan · Oct 31, 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