PhyslibAlpha.AlgebraicFramework.WStarAlgebra.TracePairingSurjectivity
The Riesz half of trace-pairing surjectivity
Ported from `unbounded-alpha-public`'s `WStarAlgebra/TracePairingSurj.lean`. This module contains the functional-analytic core of the converse trace-pairing theorem. A continuous functional on the trace class is converted into a bounded sesquilinear form by testing it on rank-one operators, and the Riesz theorem (`BoundedSesquilinearForm.lean`) turns that form into a bounded operator.
The density of the linear span of rank-one trace-class operators in the trace class is proved directly below (`rankOneSpan_dense`) via a Hilbert–Schmidt truncation argument, rather than kept as an external hypothesis: this repo's `TraceClass` files already supply every ingredient the argument needs (`isHilbertSchmidt_polarFactor_mul_sqrt_abs_and_sqrt_abs`, `traceNorm_mul_le_of_isHilbertSchmidt`).
Hilbert–Schmidt finite truncations
The density theorem is obtained by truncating a Hilbert–Schmidt factor in a Hilbert basis. The following definitions and identities are kept separate from the final topological argument so they can also be reused for other trace-ideal approximation results.
17 declarations
Rank-one operators are trace-class
Let be a complex Hilbert space. For any vectors , the rank-one operator (also denoted as ), defined by , is a trace-class operator.
Rank-one operator
Let be a complex Hilbert space. For any vectors , this definition constructs the rank-one operator (defined by ) as an element of the trace-class Banach space .
Let be a complex Hilbert space. For any bounded linear operator on and vectors , the trace pairing of with the rank-one operator (defined as the operator ) is equal to the matrix coefficient :
-span of rank-one operators in
Let be a complex Hilbert space and be the Banach space of trace-class operators on . The `rankOneSpan` is the -submodule of generated by the set of all rank-one operators. Specifically, it is the set of all finite -linear combinations of operators of the form for .
Density of Rank-One Operators Implies Surjectivity of the Trace Pairing
Let be a complex Hilbert space. Let denote the space of trace-class operators on equipped with the trace norm , and let denote the space of bounded linear operators on . If the -linear span of rank-one operators is dense in , then the trace pairing map , which sends an operator to the continuous linear functional , is surjective. That is, every continuous linear functional can be represented as for some .
Hilbert-Schmidt partial sum
Given a bounded linear operator on a Hilbert space , a Hilbert basis of , and a finite subset of indices , the Hilbert-Schmidt partial sum is the bounded linear operator defined by which maps a vector to .
Action of the Hilbert–Schmidt partial sum operator on a vector
Let be a complex Hilbert space, be a bounded linear operator, and be a Hilbert basis for . For any finite subset of indices and any vector , the Hilbert–Schmidt partial sum operator, defined as , maps the vector to the sum where denotes the complex inner product.
Action of the Hilbert-Schmidt partial sum on basis vectors
Let be a Hilbert space and be a bounded linear operator. Let be a Hilbert basis for , and let be a finite subset of indices. The Hilbert-Schmidt partial sum operator , defined by acts on a basis vector (where ) as follows:
Action of on a Hilbert Basis is if and otherwise
Let be a Hilbert space and be a bounded linear operator. Let be a Hilbert basis for . For any finite subset of indices , let be the Hilbert–Schmidt partial sum operator defined by . Then for any index , the action of the operator on the basis vector is given by:
If is Hilbert-Schmidt, then is Hilbert-Schmidt
Let be a complex Hilbert space and be a Hilbert-Schmidt operator. Given a Hilbert basis of and a finite subset of indices , let be the Hilbert-Schmidt partial sum operator defined by . Then the operator is also a Hilbert-Schmidt operator.
for Hilbert–Schmidt operators
Let be a complex Hilbert space with a Hilbert basis . Let be a Hilbert–Schmidt operator and be a finite subset of indices. Let be the Hilbert–Schmidt partial sum operator defined by (which maps to ). Then the sum over the basis indices of the squared norms of is equal to the sum of the squared norms of for all indices not in :
Trace-class partial sum
Let be a complex Hilbert space with a Hilbert basis . Given two bounded linear operators and a finite subset of indices , this definition constructs a trace-class operator in defined by the partial sum: where denotes the rank-one operator .
Trace-Class Partial Sum equals times Hilbert-Schmidt Partial Sum
Let be a complex Hilbert space. For any bounded linear operators , any Hilbert basis of , and any finite set of indices , the operator corresponding to the trace-class partial sum is equal to the composition of with the Hilbert-Schmidt partial sum . That is,
Let be a complex Hilbert space with a Hilbert basis . For any bounded linear operators and any finite subset , the trace-class operator defined by the partial sum belongs to the -linear span of rank-one operators in the trace-class space . Here, denotes the rank-one operator mapping .
Trace Norm Error Bound for Hilbert–Schmidt Truncations
Let be a complex Hilbert space. Let be bounded linear operators on such that is a trace-class operator, and are Hilbert–Schmidt operators, and . Let be a Hilbert basis of and be a finite subset of indices. Let be the trace-class partial sum , where denotes the rank-one operator . Let be the Hilbert-Schmidt partial sum . Then the trace norm of the difference between and satisfies:
Density of rank-one operators in
Let be a complex Hilbert space and be the Banach space of trace-class operators on equipped with the trace norm . The -linear span of rank-one operators is dense in .
Surjectivity of the Trace Pairing Map
Let be a complex Hilbert space. Let denote the Banach space of trace-class operators on equipped with the trace norm , and let denote the space of bounded linear operators on . The trace pairing map , which maps each operator to the continuous linear functional for , is surjective. That is, for every continuous linear functional , there exists a bounded linear operator such that for all .
