Two-state common-unitary mixtures and the two marginal conditions
DefinitionRandomDimReduction_Channel_Supplement_Jointp2o-batch-stoc27sp2o-gran-per-chapterp2o-plan-paperp2o-v1quantum-informationrepresentation-theory
For copies in each group, the data register is and the purifying register is . For any Borel probability law on define
One unitary is shared by the two states and all copies. The marginal property of a complex-linear map says: for every four density operators and common isometry with and , there exists a probability measure such that, for ,
The law may depend on the entire compatible input tuple. This defines only the two equations, with neither CPTP validity nor impossibility built in, and with no rank-sum restriction.
Definition code
import Definitions.Def_RandomDimReduction_Channel_Tensor
set_option autoImplicit false
noncomputable section
open Matrix MeasureTheory
open scoped Kronecker Matrix ComplexConjugate Matrix.Norms.Elementwise
namespace RandomDimReduction.Channel.Supplement
abbrev Data (n d : ℕ) := Word n d × Word n d
abbrev Purify (n k : ℕ) := Word n k × Word n k
/-- The same sampled U acts in both tensor factors and in all copies. -/
def jointMixture {k : ℕ} (n : ℕ) (nu : Measure (Matrix.unitaryGroup (Fin k) ℂ))
(rho sigma : Op (Fin k)) : Op (Purify n k) :=
∫ U : Matrix.unitaryGroup (Fin k) ℂ,
(tensor n ((U : Op (Fin k)) * rho * (U : Op (Fin k))ᴴ) ⊗ₖ
tensor n ((U : Op (Fin k)) * sigma * (U : Op (Fin k))ᴴ)) ∂nu
/-- Exactly the two desired marginal equations of Eq.(9).
The probability law may depend on every input and the common embedding. -/
def JointMarginals (n k d : ℕ)
(P : Op (Data n d) →ₗ[ℂ] Op (Data n d × Purify n k)) : Prop :=
∀ (rho sigma : Op (Fin d)) (rho0 sigma0 : Op (Fin k)),
MixedCloning.Optimal.IsState rho → MixedCloning.Optimal.IsState sigma →
MixedCloning.Optimal.IsState rho0 → MixedCloning.Optimal.IsState sigma0 →
∀ W : Matrix (Fin d) (Fin k) ℂ, Wᴴ * W = 1 →
W * rho0 * Wᴴ = rho → W * sigma0 * Wᴴ = sigma →
∃ nu : Measure (Matrix.unitaryGroup (Fin k) ℂ), IsProbabilityMeasure nu ∧
WildeQIT.partialTraceRight (P (tensor n rho ⊗ₖ tensor n sigma)) =
tensor n rho ⊗ₖ tensor n sigma ∧
WildeQIT.partialTraceLeft (P (tensor n rho ⊗ₖ tensor n sigma)) =
jointMixture n nu rho0 sigma0
end RandomDimReduction.Channel.Supplement
Source
Angus Lowe, Xinyu Tan, Random dimension reduction and learning symmetric properties of quantum states, STOC 2027 submission #994 (October 2026), https://github.com/sigact-stoc/stoc2027/releases/download/papers-2/994.pdf, §3.3, pp.11–12, Eq. (9)