Corollary 10
OpenQAlgorithms.Papers.Independent.Gap48.P18.finite_constrained_channel_capacity_compact_imageLet H and K be nonzero separable complex Hilbert spaces, Phi a quantum channel from H to K, and A a convex set of normalized input states. Let Cbar(Phi,A) be the supremum of output Holevo information over finite input ensembles whose average lies in A. If Cbar(Phi,A)<infinity, then Phi(A) is relatively compact in trace norm. The explicit convexity guard corrects the printed arbitrary-subset domain.
import Definitions.Def_QAlgorithms_Papers_Independent_Gap48_P18
noncomputable section
open MeasureTheory Filter
open scoped ENNReal NNReal Topology Matrix.Norms.L2Operator
open QAlgorithms.Papers.Independent.Dependency50.P33
(State PureState traceSize traceClosure IsTraceClassChannel productCoeff)
open QAlgorithms.Papers.Independent.Frontier18.P07
(GeneralizedEnsemble HasFiniteSupport HasAverage averageCoefficients
extendedEntropyNats extendedRelativeEntropyNats
finiteEnsembleCapacity pureStates)
open QAlgorithms.Papers.Independent.Frontier18.P01 (traceFirst traceSecond)
namespace QAlgorithms.Papers.Independent.Gap48.P18
attribute [local instance] Classical.propDecidable Classical.decEq
/-- Corollary 10, arXiv:quant-ph/0510073v2, PDF p.36, corrected domain.
Printed: arbitrary input subset A with finite average-constrained capacity.
Changed: explicitly require A convex; retain the same capacity and conclusion.
Why: for the identity channel, the noncompact set of orthogonal pure states has
average-constrained capacity zero. Convexity makes every finite ensemble of
signals in A admissible, which is exactly what the printed proof's inequality needs.
The correction is nonvacuous, including convex singletons and energy sublevels. -/
theorem finite_constrained_channel_capacity_compact_image {I J : Type}
[Countable I] [Countable J] [Nonempty I] [Nonempty J]
(Phi : StateChannel I J) (A : Set (State I))
(hconvex : CoefficientConvexStates A)
(hfinite : finiteEnsembleCapacity Phi.map A < ⊤) :
IsCompact (closure (Phi.action '' A)) := by sorry
end QAlgorithms.Papers.Independent.Gap48.P18
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
Let be countable, nonempty types, allowing either finite or infinite types. For any label type , call a complex matrix on trace class when it has an expansion , with and , and define as the infimum, over all such expansions, of , viewed in ; an infimum over no expansions is . A state on is a trace-class matrix such that every finite quadratic form , for every finite and every , has zero imaginary part and nonnegative real part, and such that is summable and equals . Equip states with the topology generated by the sets , for states and real , and with its Borel measurable structure. Suppose consists of a matrix map and an action on states whose coefficient matrix at each state is . Assume this matrix map preserves trace class, is additive on pairs of trace-class matrices, is complex homogeneous on trace-class matrices, and, for every natural number and every state on , produces a state coefficient matrix on by the entrywise rule
Here is included, although its state quantification is empty. Let be any set of states on whose coefficient matrices form a real convex set. For finite matrices of the same size, define
where matrix logarithms use continuous functional calculus with the total real logarithm, including , and the empty compression is included. For a Borel probability measure on input states, put , and call a state its average when every coefficient function is integrable and . Assume the following extended nonnegative supremum is strictly less than :
The support here is topological support, and admissibility requires an average in , without requiring the support to lie in . Then the closure of is compact in the state topology on . The set is allowed to be empty; in that case the displayed supremum is and the concluded compact set is empty.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.