Definitions for Quantum Illumination
DefinitionQAlgorithms_Papers_Independent_Intakenext100_P96The objects the formalized statements of Quantum Illumination (arXiv:0803.2022) are stated in. Each was checked against the paper by a blind read-back and a moderator.
SignalIndex: PDF pp.2--3: a vacuum outcome and d orthogonal single-photon modes.ModelParameters: PDF pp.2--3: parameters for the exact retained finite model. The physical approximation db << 1 is not a claim of full-channel exactness.background: PDF p.3, (1.0): the retained thermal background, including the vacuum.signalState: PDF p.3: embed the normalized single-photon input with zero vacuum amplitude.signalTest: PDF p.3: the yes projector of the unentangled test.signalPresent: PDF p.3, (1.1): returned state when the object is present.pairState: PDF pp.5--6: Phi_d = d^(-1/2) sum_k |k>_S |k>_A; no vacuum component.pairTest: PDF pp.6--7: the ideal joint yes projector onto Phi_d.ancillaMixed: PDF p.6: the perfectly retained ancilla's completely mixed marginal I_A/d.pairAbsent: PDF p.6, retained body of (8.0): rho_SA0 = rho_0 tensor I_A/d.pairPresent: PDF p.6, retained body of (8.1): only the signal traverses the noisy path.ebits: PDF p.9, (14): logarithmic mode count, identified with marginal entropy in the theorem.binaryError: PDF pp.3--4: equal-prior error of the binary effect whose yes answer means present.iidEqualPriorError: PDF p.4: optimal equal-prior error over all joint n-copy binary POVMs. Reuse the accepted tensor copies and total effect-error infimum, dividing by two for equal priors. On density inputs that extended infimum is finite and in [0,1], so toReal never uses its infinity default.countMass: PDF pp.4--5: probability of k yes outcomes in n independent Bernoulli trials.countingError: PDF pp.4--5,8: minimum error after counting yes outcomes and optimally deciding from the entire count. This includes both hypotheses with prior probability 1/2.errorFactor: PDF p.4, (3): exponential error factor, indexed using n+1 copies so that all root exponents have a strictly positive denominator.illuminationChernoffFactor: PDF p.4, (3), p.7, (11): the one-copy scalar Chernoff factor. Reuse the accepted Petz overlap with the arguments reversed: cyclicity of trace identifies Tr(rho1^s rho0^(1-s)) with the source Tr(rho0^(1-s) rho1^s). Using 0<s<1 includes one-sided endpoint limits without kernel 0^0 conventions. For density inputs this is a nonempty, bounded-below real infimum. It differs from the accepted oscillator N-copy quantumChernoff bound.binaryChernoff: PDF pp.4--5,8: Chernoff factor for the two Bernoulli yes probabilities.entangledFactor: PDF p.7, first line of (11), with its printed parentheses and denominator.
import Mathlib
import Definitions.Def_QAlgorithms_Defs_Information
import Definitions.Def_QAlgorithms_Defs_Ensemble
import Definitions.Def_QAlgorithms_Defs_FiniteRegisters
import Definitions.Def_QAlgorithms_Defs_BipartiteSeparability
import Definitions.Def_QAlgorithms_Papers_Independent_Claimed50_P15
import Definitions.Def_QAlgorithms_Papers_Independent_Information100_P60
/-!
# Quantum Illumination, arXiv:0803.2022v2
All states below belong to the retained vacuum/single-photon model, (1.0), (1.1),
(8.0), (8.1). No equality with the full optical thermal channel is asserted.
Page numbers are PDF page numbers. The ancillary photon is retained perfectly.
All new definitions are local to this paper. Compilation is pending batch handoff.
-/
noncomputable section
open Filter
open scoped BigOperators Topology ComplexConjugate ComplexOrder Matrix.Norms.L2Operator
namespace QAlgorithms.Papers.Independent.Intakenext100.P96
attribute [local instance] Classical.propDecidable Classical.decEq
/-- PDF pp.2--3: a vacuum outcome and d orthogonal single-photon modes. -/
abbrev SignalIndex (d : ℕ) := Option (Fin d)
/-- PDF pp.2--3: parameters for the exact retained finite model.
The physical approximation db << 1 is not a claim of full-channel exactness. -/
structure ModelParameters where
d : ℕ
d_pos : 0 < d
b : ℝ
b_nonneg : 0 ≤ b
db_le_one : (d : ℝ) * b ≤ 1
eta : ℝ
eta_nonneg : 0 ≤ eta
eta_le_one : eta ≤ 1
/-- PDF p.3, (1.0): the retained thermal background, including the vacuum. -/
def background (d : ℕ) (b : ℝ) : Matrix (SignalIndex d) (SignalIndex d) ℂ :=
Matrix.diagonal fun i => match i with
| none => ((1 - (d : ℝ) * b : ℝ) : ℂ)
| some _ => (b : ℂ)
/-- PDF p.3: embed the normalized single-photon input with zero vacuum amplitude. -/
def signalState {d : ℕ} (psi : EuclideanSpace ℂ (Fin d)) :
EuclideanSpace ℂ (SignalIndex d) :=
WithLp.toLp 2 (fun i => match i with | none => 0 | some k => psi k)
/-- PDF p.3: the yes projector of the unentangled test. -/
def signalTest {d : ℕ} (psi : EuclideanSpace ℂ (Fin d)) :
Matrix (SignalIndex d) (SignalIndex d) ℂ :=
pureDensity (signalState psi)
/-- PDF p.3, (1.1): returned state when the object is present. -/
def signalPresent {d : ℕ} (b eta : ℝ) (psi : EuclideanSpace ℂ (Fin d)) :
Matrix (SignalIndex d) (SignalIndex d) ℂ :=
((1 - eta : ℝ) : ℂ) • background d b + (eta : ℂ) • signalTest psi
/-- PDF pp.5--6: Phi_d = d^(-1/2) sum_k |k>_S |k>_A; no vacuum component. -/
def pairState (d : ℕ) : EuclideanSpace ℂ (SignalIndex d × Fin d) :=
if hd : 0 < d then
letI : Nonempty (Fin d) := ⟨⟨0, hd⟩⟩
WithLp.toLp 2 (fun i => match i.1 with
| none => 0
| some k => finiteMaxEntangled (Fin d) (k, i.2))
else 0
/-- PDF pp.6--7: the ideal joint yes projector onto Phi_d. -/
def pairTest (d : ℕ) : Matrix (SignalIndex d × Fin d) (SignalIndex d × Fin d) ℂ :=
pureDensity (pairState d)
/-- PDF p.6: the perfectly retained ancilla's completely mixed marginal I_A/d. -/
def ancillaMixed (d : ℕ) : Matrix (Fin d) (Fin d) ℂ :=
(((d : ℝ)⁻¹ : ℝ) : ℂ) • 1
/-- PDF p.6, retained body of (8.0): rho_SA0 = rho_0 tensor I_A/d. -/
def pairAbsent (d : ℕ) (b : ℝ) :
Matrix (SignalIndex d × Fin d) (SignalIndex d × Fin d) ℂ :=
Matrix.kronecker (background d b) (ancillaMixed d)
/-- PDF p.6, retained body of (8.1): only the signal traverses the noisy path. -/
def pairPresent (d : ℕ) (b eta : ℝ) :
Matrix (SignalIndex d × Fin d) (SignalIndex d × Fin d) ℂ :=
((1 - eta : ℝ) : ℂ) • pairAbsent d b + (eta : ℂ) • pairTest d
/-- PDF p.9, (14): logarithmic mode count, identified with marginal entropy in the theorem. -/
def ebits (d : ℕ) : ℝ := Real.log (d : ℝ) / Real.log 2
/-- PDF pp.3--4: equal-prior error of the binary effect whose yes answer means present. -/
def binaryError {I : Type*} [Fintype I] [DecidableEq I]
(rho0 rho1 M : Matrix I I ℂ) : ℝ :=
(effectProb M rho0 + effectProb (1 - M) rho1) / 2
/-- PDF p.4: optimal equal-prior error over all joint n-copy binary POVMs.
Reuse the accepted tensor copies and total effect-error infimum, dividing by two
for equal priors. On density inputs that extended infimum is finite and in [0,1],
so toReal never uses its infinity default. -/
def iidEqualPriorError {I : Type*} [Fintype I] [DecidableEq I]
(rho0 rho1 : Matrix I I ℂ) (n : ℕ) : ℝ :=
(QAlgorithms.Papers.Independent.Claimed50.P15.optimalError true
(QAlgorithms.Papers.Independent.Information100.P60.copies n rho0)
(QAlgorithms.Papers.Independent.Information100.P60.copies n rho1)).toReal / 2
/-- PDF pp.4--5: probability of k yes outcomes in n independent Bernoulli trials. -/
def countMass (p : ℝ) (n k : ℕ) : ℝ :=
(n.choose k : ℝ) * p ^ k * (1 - p) ^ (n - k)
/-- PDF pp.4--5,8: minimum error after counting yes outcomes and optimally deciding
from the entire count. This includes both hypotheses with prior probability 1/2. -/
def countingError (p0 p1 : ℝ) (n : ℕ) : ℝ :=
(∑ k : Fin (n + 1), min (countMass p0 n k.val) (countMass p1 n k.val)) / 2
/-- PDF p.4, (3): exponential error factor, indexed using n+1 copies so that
all root exponents have a strictly positive denominator. -/
def errorFactor (err : ℕ → ℝ) (n : ℕ) : ℝ :=
(2 * err (n + 1)) ^ (((n + 1 : ℕ) : ℝ)⁻¹)
/-- PDF p.4, (3), p.7, (11): the one-copy scalar Chernoff factor.
Reuse the accepted Petz overlap with the arguments reversed: cyclicity of trace
identifies Tr(rho1^s rho0^(1-s)) with the source Tr(rho0^(1-s) rho1^s).
Using 0<s<1 includes one-sided endpoint limits without kernel 0^0 conventions.
For density inputs this is a nonempty, bounded-below real infimum. It differs
from the accepted oscillator N-copy quantumChernoff bound. -/
def illuminationChernoffFactor {I : Type*} [Fintype I] [DecidableEq I]
(rho0 rho1 : Matrix I I ℂ) : ℝ :=
sInf ((fun s : ℝ =>
QAlgorithms.Papers.Independent.Claimed50.P15.petzOverlap rho1 rho0 s) '' Set.Ioo 0 1)
/-- PDF pp.4--5,8: Chernoff factor for the two Bernoulli yes probabilities. -/
def binaryChernoff (p0 p1 : ℝ) : ℝ :=
sInf ((fun s : ℝ => (1 - p0) ^ (1 - s) * (1 - p1) ^ s +
p0 ^ (1 - s) * p1 ^ s) '' Set.Ioo 0 1)
/-- PDF p.7, first line of (11), with its printed parentheses and denominator. -/
def entangledFactor (d : ℕ) (b eta s : ℝ) : ℝ :=
(1 - eta) ^ s * (1 + b / (d : ℝ) *
(-1 + (1 + eta * (d : ℝ) / ((1 - eta) * b)) ^ s))
/-! Small definition-level sanity checks; theorem proofs are intentionally omitted. -/
example : background 1 (1 / 4) none none = (3 / 4 : ℂ) := by
norm_num [background]
example {d : ℕ} (psi : EuclideanSpace ℂ (Fin d)) : signalState psi none = 0 := rfl
example {d : ℕ} (b : ℝ) (psi : EuclideanSpace ℂ (Fin d)) :
signalPresent b 0 psi = background d b := by
simp [signalPresent]
example : pairState 1 (some (0 : Fin 1), (0 : Fin 1)) = 1 := by
norm_num [pairState, finiteMaxEntangled]
example (d : ℕ) (b : ℝ) : pairPresent d b 0 = pairAbsent d b := by
simp [pairPresent]
example {I : Type*} (rho : Matrix I I ℂ) (x y : Fin 0 → I) :
QAlgorithms.Papers.Independent.Information100.P60.copies 0 rho x y = 1 := by
simp [QAlgorithms.Papers.Independent.Information100.P60.copies]
example (p0 p1 : ℝ) : countingError p0 p1 0 = 1 / 2 := by
simp [countingError, countMass]
example : ebits 1 = 0 := by simp [ebits]
example (err : ℕ → ℝ) : errorFactor err 0 = 2 * err 1 := by
simp [errorFactor]
end QAlgorithms.Papers.Independent.Intakenext100.P96
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
QAlgorithms.Papers.Independent.Intakenext100.P96.ModelParameters
A model parameter record consists of a natural number , real numbers and , and the requirements
Thus is excluded, while equality is permitted in each of the non-strict inequalities.
QAlgorithms.Papers.Independent.Intakenext100.P96.SignalIndex
For every natural number , the index set is the disjoint union of a distinguished element, denoted , and the indices . It has elements; when , it consists only of .
QAlgorithms.Papers.Independent.Intakenext100.P96.ancillaMixed
For every natural number , this defines the complex matrix indexed by : its diagonal entries are the real number regarded as complex, and its off-diagonal entries are zero. The inverse is total, with ; when , the matrix has an empty index set.
QAlgorithms.Papers.Independent.Intakenext100.P96.background
For every natural number and every real number , this defines a complex diagonal matrix indexed by . Its diagonal entry at is , its diagonal entry at every ordinary index is , and all off-diagonal entries are zero. No sign or size restriction on is imposed. When , the matrix is the one-by-one matrix with entry .
QAlgorithms.Papers.Independent.Intakenext100.P96.binaryChernoff
For arbitrary real numbers , this defines
All powers use the total real-power operation, and neither parameter is required to lie in . The endpoints are excluded; throughout the indexing interval both exponents are positive, so a zero base contributes zero. This is a real infimum, which is defined to be zero if the displayed set is unbounded below.
QAlgorithms.Papers.Independent.Intakenext100.P96.binaryError
For any finite index type equipped with decidable equality, and any three complex matrices indexed by , this defines the real number
There is no assumption that either is a density matrix or that is an effect, so the definition does not require this number to be a probability. The index type may be empty, in which case the traces and the defined value are zero.
QAlgorithms.Papers.Independent.Intakenext100.P96.countMass
For every real number and natural numbers , this defines
where is subtraction in the natural numbers. There is no restriction , so these values need not be nonnegative. If , the binomial coefficient is zero and the defined value is zero; powers with exponent zero, including , equal . In particular, the value for is .
QAlgorithms.Papers.Independent.Intakenext100.P96.countingError
For arbitrary real numbers and any natural number , this defines
Both parameters are unrestricted real numbers; no probability assumptions are imposed. The sum includes both endpoints , with every zeroth power interpreted as , including . For , the value is .
QAlgorithms.Papers.Independent.Intakenext100.P96.ebits
For every natural number , this defines the real number , where is the real natural logarithm. The definition includes and uses the total-function convention , so its value is both at and at .
QAlgorithms.Papers.Independent.Intakenext100.P96.entangledFactor
For any natural number and arbitrary real numbers , this defines
Here is cast to a real number, every power uses the total real-power operation, and division by zero is defined to be zero. No assumptions exclude , , , negative bases, or arbitrary exponents. In particular, if or , the expression reduces to .
QAlgorithms.Papers.Independent.Intakenext100.P96.errorFactor
For any function and any natural number , this defines
using the total real-power operation. No nonnegativity or other condition is imposed on , so the base may be negative. The denominator is always positive, the value is never used, and at the result is .
QAlgorithms.Papers.Independent.Intakenext100.P96.iidEqualPriorError
For any finite index type equipped with decidable equality, any complex matrices indexed by , and any natural number , define matrices , indexed by functions , by . This declaration defines
where the infimum is taken in the extended real numbers and preserves finite real values but sends both infinities to zero. No positivity or normalization is assumed for the input matrices, and may be empty. When , the indexing set consists of the unique empty function, both matrices have their sole entry equal to the empty product , and the defined error is .
QAlgorithms.Papers.Independent.Intakenext100.P96.illuminationChernoffFactor
For any finite index type equipped with decidable equality and any complex matrices indexed by , this defines
where denotes the continuous functional calculus applied to with the indicated real function . No Hermitian, positivity, or trace-normalization assumptions are imposed, so the total functional-calculus operation is used even outside those assumptions. The endpoints are excluded, and the infimum is a real infimum, with value zero if the displayed set is unbounded below. The index type may be empty, in which case the value is zero.
QAlgorithms.Papers.Independent.Intakenext100.P96.pairAbsent
For every natural number and real number , let and for . This defines the complex matrix indexed by with entries
where each equals for equal indices and otherwise. No restriction on is assumed. The convention is , and when the matrix has an empty index set.
QAlgorithms.Papers.Independent.Intakenext100.P96.pairPresent
For every natural number and real numbers , this defines a complex matrix indexed by . For , put , , and let when is the ordinary index , and zero otherwise. Its entries are
where the factors indicate equality of their indices. No bounds on or are assumed. When , the matrix has an empty index set.
QAlgorithms.Papers.Independent.Intakenext100.P96.pairState
For every natural number , this defines a complex vector indexed by . When , its coordinates are
In particular, every coordinate with first index is zero. When , the definition returns the zero vector on the empty index set.
QAlgorithms.Papers.Independent.Intakenext100.P96.pairTest
For every natural number , this defines a complex matrix indexed by pairs with and . When , its entries are
These entries are the products , where on the indicated matching indices and is zero elsewhere. When , the matrix has an empty index set.
QAlgorithms.Papers.Independent.Intakenext100.P96.signalPresent
For every natural number , real numbers , and vector , define , for ordinary indices, and define , . The resulting complex matrix, indexed by , has entries
where is when and otherwise. No bounds on or , and no normalization of , are assumed. When , this is the one-by-one matrix with entry .
QAlgorithms.Papers.Independent.Intakenext100.P96.signalState
For every natural number and every vector , this defines the vector indexed by through
There is no normalization assumption on . When , the resulting vector has its sole coordinate equal to zero.
QAlgorithms.Papers.Independent.Intakenext100.P96.signalTest
For every natural number and every vector , this defines a complex matrix indexed by whose entries at ordinary indices are
and whose entries are zero whenever either index is . No normalization assumption on is required. When , this is the one-by-one zero matrix.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.