The injective seminorm on a binary tensor product #
A tensor u : E β[π] F defines a bounded operator StrongDual π E βL[π] F, sending f to
β i, f (x i) β’ y i for any decomposition u = β i, x i ββ y i. The injective seminorm of u
is the operator norm of this map. Over β and β it is the supremum of
ββ i, f (x i) * g (y i)β over continuous linear functionals f on E and g on F of norm
at most one, which is the classical definition of the injective norm.
This file proves that the injective seminorm is bounded by the projective seminorm. Over β
and β, both agree with βxβ * βyβ on a pure tensor x ββ y and are norms when E and F
are normed spaces. Definiteness follows, via HahnβBanach, from the injectivity of the map from
tensors to operators on the dual.
The type TauCeti.InjectiveTensor π E F is a copy of E β[π] F carrying the injective
seminorm, and the identity of the underlying tensors is a contraction from
TauCeti.ProjectiveTensor π E F to it.
Mathlib's PiTensorProduct.injectiveSeminorm is a different seminorm on finite
families of factors, defined by duality against all normed target spaces rather than against the
scalar field; it is not the classical injective norm used here.
Main definitions #
TensorProduct.contractStrongDual π E F: the bounded operator on the dual ofEinduced by a tensor.TensorProduct.injectiveSeminorm π E F: the injective seminorm onE β[π] F.TauCeti.InjectiveTensor π E F: the algebraic tensor product with the injective seminorm.TauCeti.ProjectiveTensor.toInjectiveTensor: the contraction from the projective to the injective tensor product.
Main results #
TensorProduct.injectiveSeminorm_le_projectiveSeminorm:Ξ΅ β€ Ο.TensorProduct.injectiveSeminorm_le_iff_forall_norm_le_one: the classical description of the injective seminorm overβandβ.TensorProduct.injectiveSeminorm_le_iff_forall_norm_contractStrongDual_le: overβandβ, the injective seminorm is the supremum ofβcontractStrongDual π E F u fβover the closed unit ball of the dual ofE.TensorProduct.injectiveSeminorm_tmulandTensorProduct.projectiveSeminorm_tmul: both seminorms are cross norms.TensorProduct.injectiveSeminorm_eq_norm_lid: onπ β[π] E, the injective seminorm is the norm of the image inE, over any nontrivially normed field.TensorProduct.injectiveSeminorm_commandTensorProduct.injectiveSeminorm_eq_norm_rid: overβandβ, the injective seminorm is preserved byTensorProduct.comm, and onE β[π] πit is the norm of the image inE.TensorProduct.contractStrongDual_injective: overβandβ, a tensor is determined by the operator it induces on the dual of a normed left factor.TensorProduct.injectiveSeminorm_eq_zero_iffandTensorProduct.projectiveSeminorm_eq_zero_iff: both seminorms are definite.
References #
- R. A. Ryan, Introduction to Tensor Products of Banach Spaces, Springer, 2002, Β§Β§2.1, 3.1.
The bounded operator StrongDual π E βL[π] F induced by a tensor in E β[π] F: the pure
tensor x ββ y gives f β¦ f x β’ y.
Instances For
The operator induced by x ββ y sends f to f x β’ y.
The operator induced by a tensor sends f to the image of the tensor under f β id,
followed by π β[π] F β F.
The injective seminorm on E β[π] F: the operator norm of the operator on the dual of
E induced by a tensor. Over β and β it is the supremum of ββ i, f (x i) * g (y i)β over
continuous linear functionals f and g of norm at most one, for any decomposition
u = β i, x i ββ y i; see injectiveSeminorm_le_iff_forall_norm_le_one.
Instances For
The injective seminorm is the operator norm of the induced operator on the dual.
The operator induced by a tensor is bounded by its injective seminorm.
The linear map induced on algebraic tensors by the bilinear map (x, y) β¦ f x β’ g y sends a
tensor u to g applied to the operator induced by u on the dual of E, evaluated at f.
The injective seminorm of a pure tensor is at most the product of the norms.
The injective seminorm is bounded by the projective seminorm.
Evaluating the operator induced by u : π β[π] E at the identity functional of π gives
the image of u under a ββ x β¦ a β’ x.
On π β[π] E, the injective seminorm is the norm of the image under a ββ x β¦ a β’ x.
The operators induced by u and by its swap TensorProduct.comm π E F u pair the same way
with functionals on both factors: both values are β i, f (x i) * g (y i) for any decomposition
u = β i, x i ββ y i.
Over β and β, the injective seminorm is the supremum of
βg (contractStrongDual π E F u f)β = ββ i, f (x i) * g (y i)β over the closed unit balls of
the duals of E and F.
The injective seminorm is at most C iff the operator induced on the dual of the left factor
maps the closed unit ball into the closed ball of radius C.
The injective seminorm is a cross norm: its value on x ββ y is βxβ * βyβ.
The projective seminorm is a cross norm: its value on x ββ y is βxβ * βyβ.
Over β and β, the injective seminorm is symmetric: swapping the factors preserves it.
Over β and β, on E β[π] π the injective seminorm is the norm of the image under
x ββ a β¦ a β’ x.
Over β and β, a tensor whose left factor is a normed space is determined by the operator
it induces on the dual of that factor.
Over β and β, the injective seminorm on a tensor product of normed spaces is definite.
Over β and β, the projective seminorm on a tensor product of normed spaces is
definite.
The algebraic tensor product E β[π] F carrying the injective seminorm.
This wrapper distinguishes the injective seminorm from other norms on the same algebraic tensor product, such as the projective norm or the Hilbert-space norm of inner product spaces.
- toTensorProduct : TensorProduct π E F
The underlying element of the algebraic tensor product.
Instances For
Forget the injective seminorm.
Instances For
Forget the injective seminorm and retain the additive structure.
Instances For
The injective tensor product is linearly equivalent to the algebraic tensor product.
Instances For
The equivalence with E β[π] F forgets the wrapper.
The inverse equivalence wraps an algebraic tensor.
The additive equivalence with E β[π] F forgets the wrapper.
The inverse additive equivalence wraps an algebraic tensor.
The linear equivalence with E β[π] F forgets the wrapper.
The inverse linear equivalence wraps an algebraic tensor.
The underlying tensor of 0 is 0.
The underlying tensor of a sum is the sum of the underlying tensors.
The underlying tensor of a negation is the negated underlying tensor.
The underlying tensor of a difference is the difference of the underlying tensors.
The underlying tensor of a scalar multiple is the scalar multiple of the underlying tensor.
Wrapping 0 gives 0.
Wrapping a sum gives the sum of the wrapped tensors.
Wrapping a negation gives the negation of the wrapped tensor.
Wrapping a difference gives the difference of the wrapped tensors.
Wrapping a scalar multiple gives the scalar multiple of the wrapped tensor.
Two injective tensors are equal when their underlying algebraic tensors are equal.
The norm of an injective tensor is the injective seminorm of its underlying tensor.
Two bounded operators out of the injective tensor product agree when they agree on pure tensors.
The identity of the underlying tensors, from the projective to the injective tensor product. It is a contraction because the injective seminorm is bounded by the projective seminorm.
Instances For
The contraction to the injective tensor product keeps the underlying tensor.
The identity from the projective to the injective tensor product is a contraction.
Over β and β, the norm of a pure projective tensor is the product of the norms.
Over β and β, the norm of a pure injective tensor is the product of the norms.