Injective tensor product

In mathematics, the injective tensor product of two topological vector spaces (TVSs) was introduced by Alexander Grothendieck and was used by him to define nuclear spaces. An injective tensor product is in general not necessarily complete, so its completion is called the completed injective tensor products. Injective tensor products have applications outside of nuclear spaces. In particular, as described below, up to TVS-isomorphism, many TVSs that are defined for real or complex valued functions, for instance, the Schwartz space or the space of continuously differentiable functions, can be immediately extended to functions valued in a Hausdorff locally convex TVS Y without any need to extend definitions (such as "differentiable at a point") from real/complex-valued functions to Y-valued functions.

Preliminaries and notation

Throughout let X,Y, and Z be topological vector spaces and ${\displaystyle L:X\to Y}$ be a linear map.

• ${\displaystyle L:X\to Y}$ is a topological homomorphism or homomorphism, if it is linear, continuous, and ${\displaystyle L:X\to \operatorname {Im} L}$ is an open map, where ${\displaystyle \operatorname {Im} L}$, the image of L, has the subspace topology induced by Y.
  • If S is a subspace of X then both the quotient map ${\displaystyle X\to X/S}$ and the canonical injection ${\displaystyle S\to X}$ are homomorphisms. In particular, any linear map ${\displaystyle L:X\to Y}$ can be canonically decomposed as follows: ${\displaystyle X\to X/\ker L{\overset {L_{0}}{\rightarrow }}\operatorname {Im} L\to Y}$ where ${\displaystyle L_{0}(x+\ker L):=L(x)}$ defines a bijection.
• The set of continuous linear maps ${\displaystyle X\to Z}$ (resp. continuous bilinear maps ${\displaystyle X\times Y\to Z}$) will be denoted by L(X, Z) (resp. B(X, Y; Z)) where if Z is the scalar field then we may instead write L(X) (resp. B(X, Y)).
• The set of separately continuous bilinear maps ${\displaystyle X\times Y\to Z}$ (i.e. continuous in each variable when the other variable is fixed) will be denoted by ${\displaystyle {\mathcal {B}}(X,Y;Z)}$ where if ${\displaystyle Z}$ is the scalar field then we may instead write ${\displaystyle {\mathcal {B}}(X,Y)}$.
• We will denote the continuous dual space of X by X* or ${\displaystyle X^{\prime }}$ and the algebraic dual space (which is the vector space of all linear functionals on X, whether continuous or not) by ${\displaystyle X^{\#}}$.
  • To increase the clarity of the exposition, we use the common convention of writing elements of ${\displaystyle X^{\prime }}$ with a prime following the symbol (e.g. ${\displaystyle x^{\prime }}$ denotes an element of ${\displaystyle X^{\prime }}$ and not, say, a derivative and the variables x and ${\displaystyle x^{\prime }}$ need not be related in any way).

Notation for topologies

• σ(X, X′) denotes the coarsest topology on X making every map in X′ continuous and ${\displaystyle X_{\sigma \left(X,X^{\prime }\right)}}$ or ${\displaystyle X_{\sigma }}$ denotes X endowed with this topology.
• σ(X′, X) denotes weak-* topology on X* and ${\displaystyle X_{\sigma \left(X^{\prime },X\right)}}$ or ${\displaystyle X_{\sigma }^{\prime }}$ denotes X′ endowed with this topology.
  • Note that every ${\displaystyle x_{0}\in X}$ induces a map ${\displaystyle X^{\prime }\to \mathbb {R} }$ defined by ${\displaystyle \lambda \mapsto \lambda \left(x_{0}\right)}$. σ(X′, X) is the coarsest topology on X′ making all such maps continuous.
• b(X, X′) denotes the topology of bounded convergence on X and ${\displaystyle X_{b\left(X,X^{\prime }\right)}}$ or ${\displaystyle X_{b}}$ denotes X endowed with this topology.
• b(X′, X) denotes the topology of bounded convergence on X′ or the strong dual topology on X′ and ${\displaystyle X_{b\left(X^{\prime },X\right)}}$ or ${\displaystyle X_{b}^{\prime }}$ denotes X′ endowed with this topology.
  • As usual, if X* is considered as a topological vector space but it has not been made clear what topology it is endowed with, then the topology will be assumed to be b(X′, X).
• 𝜏(X, X′) denotes the Mackey topology on X or the topology of uniform convergence on the convex balanced weakly compact subsets of X′ and ${\displaystyle X_{\tau \left(X,X^{\prime }\right)}}$ or ${\displaystyle X_{\tau }}$ denotes X endowed with this topology. ${\displaystyle \tau (X,X^{\prime })}$ is the finest locally convex TVS topology on X whose continuous dual space is equal to ${\displaystyle X^{\prime }}$.
• 𝜏(X′, X) denotes the Mackey topology on X′ or the topology of uniform convergence on the convex balanced weakly compact subsets of X and ${\displaystyle X_{\tau \left(X^{\prime },X\right)}}$ or ${\displaystyle X_{\tau }^{\prime }}$ denotes X endowed with this topology.
  • Note that ${\displaystyle \tau \left(X^{\prime },X\right)\subseteq b\left(X^{\prime },X\right)\subseteq \tau \left(X^{\prime },X^{\prime \prime }\right)}$.[1]
• ε(X, X′) denotes the topology of uniform convergence on equicontinuous subsets of X′ and ${\displaystyle X_{\varepsilon \left(X,X^{\prime }\right)}}$ or ${\displaystyle X_{\varepsilon }}$ denotes X endowed with this topology.
  • If H is a set of linear mappings ${\displaystyle X\to Y}$ then H is equicontinuous if and only if it is equicontinuous at the origin; that is, if and only if for every neighborhood V of 0 in Y, there exists a neighborhood U of 0 in X such that ${\displaystyle \lambda (U)\subseteq V}$ for every ${\displaystyle \lambda \in H}$.
• A set H of linear maps from X to Y is called equicontinuous if for every neighborhood V of 0 in Y, there exists a neighborhood U of 0 in X such that ${\displaystyle h(U)\subseteq V}$ for all ${\displaystyle h\in H}$.[2]

Definition

Throughout let X and Y be topological vector spaces with continuous dual spaces ${\displaystyle X^{*}}$ and ${\displaystyle Y^{*}}$. Note that almost all results described are independent of whether these vector spaces are over ${\displaystyle \mathbb {R} }$ or ${\displaystyle \mathbb {C} }$ but to simplify the exposition we will assume that they are over the field ${\displaystyle \mathbb {C} }$.

Continuous bilinear maps as a tensor product

Note that the question of whether or not one vector space is a tensor product of two other vector spaces is a purely algebraic one (i.e. the answer does not depend on the topologies of X or Y). Nevertheless, the vector space ${\displaystyle B\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)}$ of continuous bilinear functionals is always a tensor product of X and Y, as we now describe.

For every ${\displaystyle (x,y)\in X\times Y}$ we now define bilinear form, denoted by the symbol x ⊗ y, from ${\displaystyle X^{\prime }\times Y^{\prime }}$ into the underlying field (i.e. ${\displaystyle x\otimes y:X^{\prime }\times Y^{\prime }\to \mathbb {C} }$) by

${\displaystyle \left(x\otimes y\right)\left(x^{\prime },y^{\prime }\right):=x^{\prime }(x)y^{\prime }(y)}$.

This induces a canonical map ${\displaystyle \cdot \otimes \cdot :X\times Y\to {\mathcal {B}}\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)}$ defined by sending ${\displaystyle (x,y)\in X\times Y}$ to the bilinear form ${\displaystyle x\otimes y}$. The span of the range of this map is ${\displaystyle B\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)}$. The following theorem may be used to verify that ${\displaystyle B\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)}$ together with the above map ⊗ is a tensor product of X and Y.

Theorem Let X, Y, and Z be vector spaces and let ${\displaystyle T:X\times Y\to Z}$ be a bilinear map. Then the following are equivalent:[3]

1. (Z, T) is a tensor product of X and Y;
2. (a) the image of T spans all of Z, and (b) X and Y are T-linearly disjoint (this means that for all positive integers n and all elements ${\displaystyle x_{1},\ldots ,x_{n}\in X}$ and ${\displaystyle y_{1},\ldots ,y_{n}\in Y}$ such that ${\displaystyle \sum _{i=1}^{n}T(x_{i},y_{i})=0}$, (i) if all ${\displaystyle x_{1},\ldots ,x_{n}}$ are linearly independent then all ${\displaystyle y_{i}}$ are 0, and (ii) if all ${\displaystyle y_{1},\ldots ,y_{n}}$ are linearly independent then all ${\displaystyle x_{i}}$ are 0).

• Equivalently, X and Y are T-linearly disjoint if and only if for all linearly independent sequences ${\displaystyle x_{1},\ldots ,x_{m}}$ in X and all linearly independent sequences ${\displaystyle y_{1},\ldots ,y_{n}}$ in Y, the vectors ${\displaystyle \left\{T(x_{i},y_{j}):1\leq i\leq m,1\leq j\leq n\right\}}$ are linearly independent.

Topology

Henceforth, all topological vector spaces considered will be assumed to be locally convex. If Z is any locally convex topological vector space, then for any equicontinuous subsets ${\displaystyle G\subseteq X^{\prime }}$ and ${\displaystyle H\subseteq Y^{\prime }}$, and any neighborhood ${\displaystyle N}$ in Z, define

${\displaystyle {\mathcal {U}}(G,H,N)=\{b\in {\mathcal {B}}\left(X_{b}^{\prime },Y_{b}^{\prime };Z\right):b(G,H)\subseteq N\}}$

It can be shown that every set ${\displaystyle b(G,H)}$ is bounded, which is necessary and sufficient for the collection of all such ${\displaystyle {\mathcal {U}}(G,H,N)}$ to form a locally convex TVS topology on ${\displaystyle {\mathcal {B}}\left(X_{b}^{\prime },Y_{b}^{\prime };Z\right)}$ called the ε-topology. When ${\displaystyle {\mathcal {B}}\left(X_{b}^{\prime },Y_{b}^{\prime };Z\right)}$ (resp. ${\displaystyle {\mathcal {B}}\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime };Z\right)}$, ${\displaystyle B\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime };Z\right)}$, etc.) is endowed with the ε-topology, it will be denoted by ${\displaystyle {\mathcal {B}}_{\varepsilon }\left(X_{b}^{\prime },Y_{b}^{\prime };Z\right)}$ (resp. ${\displaystyle {\mathcal {B}}_{\varepsilon }\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime };Z\right)}$, ${\displaystyle B_{\varepsilon }\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime };Z\right)}$, etc.). (Note that ${\displaystyle B\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime };Z\right)\subseteq B\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime };Z\right)\subseteq {\mathcal {B}}\left(X_{b}^{\prime },Y_{b}^{\prime };Z\right)}$).

In particular, when Z is the underlying scalar field then since ${\displaystyle B\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)=X\otimes Y}$, we will denote ${\displaystyle B_{\varepsilon }\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)}$ by ${\displaystyle X\otimes _{\varepsilon }Y}$, which is called the injective tensor product of X and Y. The space ${\displaystyle B\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)}$ is not necessarily complete so we denote its completion by ${\displaystyle X{\widehat {\otimes }}_{\varepsilon }Y}$. The space ${\displaystyle {\mathcal {B}}\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)}$ is complete if and only if both X and Y are complete, in which case the completion of ${\displaystyle B\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)}$ is a subvector space (denoted by ${\displaystyle X{\widehat {\otimes }}_{\varepsilon }Y}$) of ${\displaystyle {\mathcal {B}}\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)}$. If X and Y are normed then so is ${\displaystyle {\mathcal {B}}_{\varepsilon }\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)}$. And ${\displaystyle {\mathcal {B}}_{\varepsilon }\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)}$ is a Banach space if and only if both X and Y are Banach spaces.[4]

Equicontinuous sets

One reason for converging on equicontinuous subsets (of all possibilities) is the following important fact:

A set of continuous linear functionals H on a TVS ${\displaystyle X}$ (not necessarily Hausdorff or locally convex) is equicontinuous if and only if it is contained in the polar of some neighborhood U of ${\displaystyle 0}$ in ${\displaystyle X}$ (i.e. ${\displaystyle H\subseteq U^{\circ }}$).

A TVS's topology is completely determined by the open neighborhoods of the origin. This fact together with the bipolar theorem means that via the operation of taking the polar of a subset, the collection of all equicontinuous subsets of ${\displaystyle X^{\prime }}$ "encodes" all information about ${\displaystyle X}$'s given topology. Specifically, distinct TVS topologies on ${\displaystyle X}$ produce distinct collections of equicontinuous subsets and conversely, given any such collection of equicontinuous sets, the TVS's original topology can be recovered by taking the polar of every (equicontinuous) set in the collection. Thus through this identification, uniform convergence on the collection of equicontinuous subsets is essentially uniform convergence on the very topology of the TVS; this allows one to directly relate the injective topology with the given topologies of ${\displaystyle X}$ and ${\displaystyle Y.}$ Furthermore, the topology of a locally convex Hausdorff space ${\displaystyle X}$ is identical to the topology of uniform convergence on the equicontinuous subsets of ${\displaystyle X^{\prime }.}$[5]

For this reason, the article now lists some properties of equicontinuous sets that are relevant for dealing with the injective tensor product. Throughout ${\displaystyle X}$ and ${\displaystyle Y}$ are arbitrary TVSs and ${\displaystyle H}$ is a collection of linear maps from ${\displaystyle X}$ into ${\displaystyle Y.}$

• If ${\displaystyle H\subseteq L(X;Y)}$ is equicontinuous then the subspace topologies that ${\displaystyle H}$ inherits from the following topologies on ${\displaystyle L(X;Y)}$ are identical:[6]

1. the topology of precompact convergence;
2. the topology of compact convergence;
3. the topology of pointwise convergence;
4. the topology of pointwise convergence on a given dense subset of ${\displaystyle X.}$

• An equicontinuous set ${\displaystyle H\subseteq L(X;Y)}$ is bounded in the topology of bounded convergence (i.e. bounded in ${\displaystyle L_{b}(X;Y)}$).[6] So in particular, ${\displaystyle H}$ will also bounded in every TVS topology that is coarser than the topology of bounded convergence.
• If ${\displaystyle X}$ is a barrelled space and ${\displaystyle Y}$ is locally convex then for any subset ${\displaystyle H\subseteq L(X;Y),}$ the following are equivalent:

1. ${\displaystyle H}$ is equicontinuous;
2. ${\displaystyle H}$ is bounded in the topology of pointwise convergence (i.e. bounded in ${\displaystyle L_{\sigma }(X;Y)}$);
3. ${\displaystyle H}$ is bounded in the topology of bounded convergence (i.e. bounded in ${\displaystyle L_{b}(X;Y)}$).

In particular, to show that a set ${\displaystyle H}$ is equicontinuous it suffices to show that it is bounded in the topology of pointwise converge.[7]

• If ${\displaystyle X}$ is a Baire space then any subset ${\displaystyle H\subseteq L(X;Y)}$ that is bounded in ${\displaystyle L_{\sigma }(X;Y)}$ is necessarily equicontinuous.[7]
• If ${\displaystyle X}$ is separable, ${\displaystyle Y}$ is metrizable, and ${\displaystyle D}$ is a dense subset of X, then the topology of pointwise convergence on ${\displaystyle D}$ makes ${\displaystyle L(X;Y)}$ metrizable so that in particular, the subspace topology that any equicontinuous subset ${\displaystyle H\subseteq L(X;Y)}$ inherits from ${\displaystyle L_{\sigma }(X;Y)}$ is metrizable.[6]

We now restrict our attention to properties of equicontinuous subsets of the continuous dual space ${\displaystyle X^{\prime }}$ (where Y is now the underlying scalar field of ${\displaystyle X}$).

• The weak closure of an equicontinuous set of linear functionals on ${\displaystyle X}$ is a compact subspace of ${\displaystyle X_{\sigma }^{\prime }<math>H}$</math>.[6]
• If ${\displaystyle X}$ is separable then every weakly closed equicontinuous subset of ${\displaystyle X_{\sigma }^{\prime }}$ is a metrizable compact space when it is given the weak topology (i.e. the subspace topology inherited from ${\displaystyle X_{\sigma }^{\prime }}$).[6]
• If ${\displaystyle X}$ is a normable space then a subset ${\displaystyle H\subseteq X^{\prime }}$ is equicontinuous if and only if it is strongly bounded (i.e. bounded in ${\displaystyle X_{b}^{\prime }}$).[6]
• If ${\displaystyle X}$ is a barrelled space then for any subset ${\displaystyle H\subseteq X^{\prime }}$, the following are equivalent:[7]

1. ${\displaystyle H}$ is equicontinuous;
2. ${\displaystyle H}$ is relatively compact in the weak dual topology;
3. ${\displaystyle H}$ is weakly bounded;
4. ${\displaystyle H}$ is strongly bounded.

We mention some additional important basic properties relevant to the injective tensor product:

• Suppose that ${\displaystyle B:X_{1}\times X_{2}\to Y}$ is a bilinear map where ${\displaystyle X_{1}}$ is a Fréchet space, ${\displaystyle X_{2}}$ is metrizable, and ${\displaystyle Y}$ is locally convex. If ${\displaystyle B}$ is separately continuous then it is continuous.[8]

Canonical identification of separately continuous bilinear maps with linear maps

One may show that the set equality ${\displaystyle L\left(X_{\sigma }^{\prime };Y_{\sigma }\right)=L\left(X_{\tau }^{\prime };Y\right)}$ holds (i.e. if ${\displaystyle u:X^{\prime }\to Y}$ is a linear map, then ${\displaystyle u:X_{\sigma \left(X^{\prime },X\right)}^{\prime }\to Y_{\sigma \left(Y,Y^{\prime }\right)}}$ is continuous if and only if ${\displaystyle u:X_{\tau \left(X^{\prime },X\right)}^{\prime }\to Y}$ is continuous, where here Y has its original topology).[9]

There is also a canonical vector space isomorphism ${\displaystyle J:{\mathcal {B}}\left(X_{\sigma \left(X^{\prime },X\right)}^{\prime },Y_{\sigma \left(Y^{\prime },Y\right)}^{\prime }\right)\to L\left(X_{\sigma \left(X^{\prime },X\right)}^{\prime };Y_{\sigma \left(Y,Y^{\prime }\right)}\right)}$.[9] To define it, for every separately continuous bilinear form ${\displaystyle B}$ defined on ${\displaystyle X_{\sigma \left(X^{\prime },X\right)}^{\prime }\times Y_{\sigma \left(Y^{\prime },Y\right)}^{\prime }}$ and every ${\displaystyle x^{\prime }\in X^{\prime }}$, let ${\displaystyle B_{x^{\prime }}\in \left(Y_{\sigma }^{\prime }\right)^{\prime }}$ be defined by ${\displaystyle B_{x^{\prime }}\left(y^{\prime }\right):=B\left(x^{\prime },y^{\prime }\right)}$. Since ${\displaystyle \left(Y_{\sigma }^{\prime }\right)^{\prime }}$ is canonically vector space-isomorphic to Y (via the canonical map ${\displaystyle y\mapsto }$ value at y), we will identify ${\displaystyle B_{x^{\prime }}}$ as an element of Y, which we'll denote by ${\displaystyle {\tilde {B}}_{x^{\prime }}\in Y}$. This defines a map ${\displaystyle {\tilde {B}}:X^{\prime }\to Y}$ given by ${\displaystyle x^{\prime }\mapsto {\tilde {B}}_{x^{\prime }}}$ and so the canonical isomorphism is of course defined by ${\displaystyle J(B):={\tilde {B}}}$.

When ${\displaystyle L\left(X_{\sigma }^{\sigma };Y_{\sigma }\right)}$ is given the topology of uniform convergence on equicontinous subsets of X′, the canonical map becomes a TVS-isomorphism ${\displaystyle J:{\mathcal {B}}_{\varepsilon }\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)\to L_{\varepsilon }\left(X_{\tau }^{\prime };Y\right)}$.[9] In particular, ${\displaystyle X\otimes _{\varepsilon }Y=B_{\varepsilon }\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)}$ can be canonically TVS-embedded into ${\displaystyle L_{\varepsilon }\left(X_{\tau }^{\prime };Y\right)}$; furthermore the image in ${\displaystyle L\left(X_{\sigma }^{\prime };Y_{\sigma }\right)}$ of ${\displaystyle X\otimes _{\varepsilon }Y=B_{\varepsilon }\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)}$ under the canonical map J consists exactly of the space of continuous linear maps ${\displaystyle X_{\sigma \left(X^{\prime },X\right)}^{\prime }\to Y}$ whose image is finite dimensional.[4]

One always has ${\displaystyle L\left(X_{\tau }^{\prime };Y\right)\subseteq L\left(X_{b}^{\prime };Y\right)}$. If X is normed then ${\displaystyle L_{\varepsilon }\left(X_{\tau }^{\prime };Y\right)}$ is in fact a topological vector subspace of ${\displaystyle L_{b}\left(X_{b}^{\prime };Y\right)}$. And if in addition Y is Banach then so is ${\displaystyle L_{b}\left(X_{b}^{\prime };Y\right)}$ (even if X is not complete).[4]

Properties

• The canonical map ${\displaystyle \cdot \otimes \cdot :X\times Y\to {\mathcal {B}}\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)}$ is continuous.[10]
• If X and Y are normed then ${\displaystyle X\otimes _{\varepsilon }Y}$ is normable.
• ${\displaystyle X\otimes _{\varepsilon }Y}$ is Hausdorff if and only if both X and Y are Hausdorff.[10]
• Suppose that ${\displaystyle u:X_{1}\to Y_{1}}$ and ${\displaystyle v:X_{2}\to Y_{2}}$ are two linear maps between locally convex spaces. If both u and v are continuous then so is their tensor product ${\displaystyle u\otimes v:X_{1}\otimes _{\varepsilon }X_{2}\to Y_{1}\otimes _{\varepsilon }Y_{2}}$.[11]
  • If u and v are both TVS-embeddings then so is ${\displaystyle u{\widehat {\otimes }}_{\varepsilon }v:X_{1}{\widehat {\otimes }}_{\varepsilon }X_{2}\to Y_{1}{\widehat {\otimes }}_{\varepsilon }Y_{2}}$.[12]
  • If ${\displaystyle X_{1}}$ (resp. ${\displaystyle Y_{1}}$) is a linear subspace of ${\displaystyle X_{2}}$ (resp. ${\displaystyle Y_{2}}$) then ${\displaystyle X_{1}\otimes _{\varepsilon }Y_{1}}$ is canonically isomorphic to a linear subspace of ${\displaystyle X_{2}\otimes _{\varepsilon }Y_{2}}$ and ${\displaystyle X_{1}{\widehat {\otimes }}_{\varepsilon }Y_{1}}$ is canonically isomorphic to a linear subspace of ${\displaystyle X_{2}{\widehat {\otimes }}_{\varepsilon }Y_{2}}$.[13]
  • There are examples of u and v such that both u and v are surjective homomorphisms but ${\displaystyle u{\widehat {\otimes }}_{\varepsilon }v:X_{1}{\widehat {\otimes }}_{\varepsilon }X_{2}\to Y_{1}{\widehat {\otimes }}_{\varepsilon }Y_{2}}$ is not a homomorphism.[14]
  • If all four spaces are normed then ${\displaystyle \|u\otimes v\|_{\varepsilon }=\|u\|\|v\|}$.[15]
• The ε-topology is always finer than the π-topology and coarser than the inductive topology (which is the finest locally convex TVS topology making ${\displaystyle X\times Y\to X\otimes Y}$ separately continuous).
• If X and Y are normed then for all ${\displaystyle \theta \in X\otimes Y}$, ${\displaystyle \|\theta \|_{\varepsilon }\leq \|\theta \|_{\pi }}$.[15]

Relation to projective tensor product and nuclear spaces

The strongest locally convex topology on ${\displaystyle B\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)=X\otimes Y}$ making the canonical map ${\displaystyle \cdot \otimes \cdot :X\times Y\to B\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)}$ (defined by sending ${\displaystyle (x,y)\in X\times Y}$ to the bilinear form ${\displaystyle x\otimes y}$) continuous is called the projective topology or the π-topology. When ${\displaystyle B\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)=X\otimes Y}$ is endowed with this topology then it will be denoted by ${\displaystyle X\otimes _{\pi }Y}$ and called the projective tensor product of X and Y.

The following definition was used by Grothendieck to define nuclear spaces.[16]

Definition 0: Let X be a locally convex topological vector space. Then X is nuclear if for any locally convex space Y, the canonical vector space embedding ${\displaystyle X\otimes _{\pi }Y\to {\mathcal {B}}_{\varepsilon }\left(X_{\sigma }^{\prime },Y_{\sigma }^{\prime }\right)}$ is an embedding of TVSs whose image is dense in the codomain.

Canonical identifications of bilinear and linear maps

In this section we describe canonical identifications between spaces of bilinear and linear maps. These identifications will be used to define important subspaces and topologies (particularly those that relate to nuclear operators and nuclear spaces).

Dual spaces of the injective tensor product and its completion

Suppose that ${\displaystyle \operatorname {In} :X\otimes _{\varepsilon }Y\to X{\widehat {\otimes }}_{\varepsilon }Y}$ denotes the TVS-embedding of ${\displaystyle X\otimes _{\varepsilon }Y}$ into its completion and let ${\displaystyle {}^{t}\operatorname {In}$ :\left(X{\widehat {\otimes }}_{\varepsilon }Y\right)_{b}^{\prime }\to \left(X\otimes _{\varepsilon }Y\right)_{b}^{\prime }} be its transpose, which is a vector space-isomorphism. This identifies the continuous dual space of ${\displaystyle X\otimes _{\varepsilon }Y}$ as being identical to the continuous dual space of ${\displaystyle X{\widehat {\otimes }}_{\varepsilon }Y}$.

The identity map ${\displaystyle \operatorname {Id} _{X\otimes Y}:X\otimes _{\pi }Y\to X\otimes _{\varepsilon }Y}$ is continuous (by definition of the π-topology) so there exists a unique continuous linear extension ${\displaystyle {\hat {I}}:X{\widehat {\otimes }}_{\pi }Y\to {\widehat {\otimes }}_{\varepsilon }Y}$. If X and Y are Hilbert spaces then ${\displaystyle {\hat {I}}:X{\widehat {\otimes }}_{\pi }Y\to X{\widehat {\otimes }}_{\varepsilon }Y}$ is injective and the dual of ${\displaystyle X{\widehat {\otimes }}_{\varepsilon }Y}$ is canonically isometrically isomorphic to the vector space ${\displaystyle L^{1}\left(X;Y^{\prime }\right)}$ of nuclear operators from X into Y (with the trace norm).

Injective tensor product of Hilbert spaces

There is a canonical map ${\displaystyle K:X\otimes Y\to L\left(X^{\prime };Y\right)}$ that sends ${\displaystyle z=\sum _{i=1}^{n}x_{i}\otimes y_{i}}$ to the linear map ${\displaystyle K(z):X^{\prime }\to Y}$ defined by ${\displaystyle K(z)\left(x^{\prime }\right):=\sum _{i=1}^{n}x^{\prime }(x_{i})y_{i}\in Y}$, where it may be shown that the definition of ${\displaystyle K(z):X\to Y}$ does not depend on the particular choice of representation ${\displaystyle \sum _{i=1}^{n}x_{i}\otimes y_{i}}$ of z. The map ${\displaystyle K:X\otimes _{\varepsilon }Y\to L_{b}\left(X_{b}^{\prime };Y\right)}$ is continuous and when ${\displaystyle L_{b}\left(X_{b}^{\prime };Y\right)}$ is complete, it has a continuous extension ${\displaystyle {\hat {K}}:X{\widehat {\otimes }}_{\varepsilon }Y\to L_{b}\left(X_{b}^{\prime };Y\right)}$.

When X and Y are Hilbert spaces then ${\displaystyle {\hat {K}}:X{\widehat {\otimes }}_{\varepsilon }Y\to L_{b}\left(X_{b}^{\prime };Y\right)}$ is a TVS-embedding and isometry (when the spaces are given their usual norms) whose range is the space of all compact linear operators from X into Y (which is a closed vector subspace of ${\displaystyle L_{b}\left(X^{\prime };Y\right)}$. Hence ${\displaystyle X{\widehat {\otimes }}_{\varepsilon }Y}$ is identical to space of compact operators from ${\displaystyle X^{\prime }}$ into Y (note the prime on X). The space of compact linear operators between any two Banach spaces (which includes Hilbert spaces) X and Y is a closed subset of ${\displaystyle L_{b}\left(X;Y\right)}$.[17]

Furthermore, the canonical map ${\displaystyle X{\widehat {\otimes }}_{\pi }Y\to X{\widehat {\otimes }}_{\varepsilon }Y}$ is injective when X and Y are Hilbert spaces. [17]

Integral forms and operators

Integral bilinear forms

Let ${\displaystyle \operatorname {Id} :X\otimes _{\pi }Y\to X\otimes _{\varepsilon }Y}$ denote the identity map and ${\displaystyle {}^{t}\operatorname {Id}$ :\left(X\otimes _{\varepsilon }Y\right)_{b}^{\prime }\to \left(X\otimes _{\pi }Y\right)_{b}^{\prime }} denote its transpose, which is a continuous injection. Recall that ${\displaystyle \left(X\otimes _{\pi }Y\right)^{\prime }}$ is canonically identified with ${\displaystyle B(X,Y)}$, the space of continuous bilinear maps on ${\displaystyle X\times Y}$. In this way, the continuous dual space of ${\displaystyle X\otimes _{\varepsilon }Y}$ can be canonically identified as a subvector space of ${\displaystyle B(X,Y)}$, denoted by ${\displaystyle J(X,Y)}$. The elements of ${\displaystyle J(X,Y)}$ are called integral (bilinear) forms on ${\displaystyle X\times Y}$. The following theorem justifies the word integral.

Theorem[18][19] The dual J(X, Y) of ${\displaystyle X{\widehat {\otimes }}_{\varepsilon }Y}$ consists of exactly those continuous bilinear forms v on ${\displaystyle X\times Y}$ that can be represented in the form of a map

${\displaystyle b\in B(X,Y)\mapsto v(b)=\int _{S\times T}b{\big \vert }_{S\times T}\left(x^{\prime },y^{\prime }\right)\operatorname {d} \mu \left(x^{\prime },y^{\prime }\right)}$

where S and T are some closed, equicontinuous subsets of ${\displaystyle X_{\sigma }^{\prime }}$ and ${\displaystyle Y_{\sigma }^{\prime }}$, respectively, and ${\displaystyle \mu }$ is a positive Radon measure on the compact set ${\displaystyle S\times T}$ with total mass ${\displaystyle \leq 1}$. Furthermore, if A is an equicontinuous subset of J(X, Y) then the elements ${\displaystyle v\in A}$ can be represented with ${\displaystyle S\times T}$ fixed and ${\displaystyle \mu }$ running through a norm bounded subset of the space of Radon measures on ${\displaystyle S\times T}$.

Integral linear operators

Given a linear map ${\displaystyle \Lambda :X\to Y}$, one can define a canonical bilinear form ${\displaystyle B_{\Lambda }\in Bi\left(X,Y^{\prime }\right)}$, called the associated bilinear form on ${\displaystyle X\times Y^{\prime }}$, by ${\displaystyle B_{\Lambda }\left(x,y^{\prime }\right):=\left(y^{\prime }\circ \Lambda \right)\left(x\right)}$. A continuous map ${\displaystyle \Lambda :X\to Y}$ is called integral if its associated bilinear form is an integral bilinear form.[20] An integral map ${\displaystyle \Lambda :X\to Y}$ is of the form, for every ${\displaystyle x\in X}$ and ${\displaystyle y^{\prime }\in Y^{\prime }}$:

${\displaystyle \left\langle y^{\prime },\Lambda (x)\right\rangle =\int _{A^{\prime }\times B^{\prime \prime }}\left\langle x^{\prime },x\right\rangle \left\langle y^{\prime \prime },y^{\prime }\right\rangle \operatorname {d} \mu \left(x^{\prime },y^{\prime \prime }\right)}$

for suitable weakly closed and equicontinuous aubsets ${\displaystyle A^{\prime }}$ and ${\displaystyle B^{\prime \prime }}$ of ${\displaystyle X^{\prime }}$ and ${\displaystyle Y^{\prime \prime }}$, respectively, and some positive Radon measure ${\displaystyle \mu }$ of total mass ${\displaystyle \leq 1}$.

Canonical map into L(X; Y)

There is a canonical map ${\displaystyle K:X^{\prime }\otimes Y\to L(X;Y)}$ that sends ${\displaystyle z=\sum _{i=1}^{n}x_{i}^{\prime }\otimes y_{i}}$ to the linear map ${\displaystyle K(z):X\to Y}$ defined by ${\displaystyle K(z)(x):=\sum _{i=1}^{n}x_{i}^{\prime }(x)y_{i}\in Y}$, where it may be shown that the definition of ${\displaystyle K(z):X\to Y}$ does not depend on the particular choice of representation ${\displaystyle \sum _{i=1}^{n}x_{i}^{\prime }\otimes y_{i}}$ of z.

Examples

Space of summable families

Throughout this section we fix some arbitrary (possibly uncountable) set A, a TVS X, and we let ${\displaystyle {\mathcal {F}}(A)}$ be the directed set of all finite subsets of A directed by inclusion ${\displaystyle \subseteq }$.

Let ${\displaystyle \left(x_{\alpha }\right)_{\alpha \in A}}$ be a family of elements in a TVS X and for every finite subset H of A, let ${\displaystyle x_{H}:=\sum _{i\in H}x_{i}}$. We call ${\displaystyle \left(x_{\alpha }\right)_{\alpha \in A}}$ summable in X if the limit ${\displaystyle \lim _{H\in {\mathcal {F}}(A)}x_{H}}$ of the net ${\displaystyle \left(x_{H}\right)_{H\in {\mathcal {F}}(A)}}$ converges in X to some element (any such element is called its sum). The set of all such summable families is a vector subspace of ${\displaystyle X^{A}}$ denoted by ${\displaystyle S}$.

We now define a topology on S in a very natural way. This topology turns out to be the injective topology taken from ${\displaystyle l^{1}(A){\widehat {\otimes }}_{\varepsilon }X}$ and transferred to S via a canonical vector space isomorphism (the obvious one). This is a common occurrence when studying the injective and projective tensor products of function/sequence spaces and TVSs: the "natural way" in which one would define (from scratch) a topology on such a tensor product is frequently equivalent to the injective or projective tensor product topology.

Let ${\displaystyle {\mathfrak {U}}}$ denote a base of convex balanced neighborhoods of 0 in X and for each ${\displaystyle U\in {\mathfrak {U}}}$, let ${\displaystyle \mu _{U}:X\to \mathbb {R} }$ denote its Minkowski functional. For any such U and any ${\displaystyle x=\left(x_{\alpha }\right)_{\alpha \in A}\in S}$, let

${\displaystyle q_{U}(x):=\sup _{x^{\prime }\in U^{\circ }}\sum _{\alpha \in A}\left|\left\langle x^{\prime },x_{\alpha }\right\rangle \right|}$

where ${\displaystyle q_{U}}$ defines a seminorm on S. The family of seminorms ${\displaystyle \{q_{U}:U\in {\mathfrak {U}}\}}$ generates a topology making S into a locally convex space. The vector space S endowed with this topology will be denoted by ${\displaystyle l^{1}(A,X)}$.[21] The special case where X is the scalar field will be denoted by ${\displaystyle l^{1}(A)}$.

There is a canonical embedding of vector spaces ${\displaystyle l^{1}(A)\otimes X\to l^{1}(A,E)}$ defined by linearizing the bilinear map ${\displaystyle l^{1}(A)\times X\to l^{1}(A,E)}$ defined by ${\displaystyle \left(\left(r_{\alpha }\right)_{\alpha \in A},x\right)\mapsto \left(r_{\alpha }x\right)_{\alpha \in A}}$.[21]

Theorem:[21] The canonical embedding (of vector spaces) ${\displaystyle l^{1}(A)\otimes X\to l^{1}(A,E)}$ becomes an embedding of topological vector spaces ${\displaystyle l^{1}(A)\otimes _{\varepsilon }X\to l^{1}(A,E)}$ when ${\displaystyle l^{1}(A)\otimes X}$ is given the injective topology and furthermore, its range is dense in its codomain. If ${\displaystyle {\hat {X}}}$ is a completion of X then the continuous extension ${\displaystyle l^{1}(A){\widehat {\otimes }}_{\varepsilon }X\to l^{1}\left(A,{\hat {X}}\right)}$ of this embedding ${\displaystyle l^{1}(A)\otimes _{\varepsilon }X\to l^{1}\left(A,X\right)\subseteq l^{1}\left(A,{\hat {X}}\right)}$ is an isomorphism of TVSs. So in particular, if X is complete then ${\displaystyle l^{1}(A){\widehat {\otimes }}_{\varepsilon }X}$ is canonically isomorphic to ${\displaystyle l^{1}(A,E)}$.

Space of continuously differentiable vector-valued functions

Throughout, let ${\displaystyle \Omega }$ be an open subset of ${\displaystyle \mathbb {R} ^{n}}$, where ${\displaystyle n\geq 1}$ is an integer and let ${\displaystyle Y}$ be a locally convex topological vector space (TVS).

Definition[22] Suppose ${\displaystyle p^{0}=\left(p_{1}^{0},\ldots ,p_{n}^{0}\right)\in \Omega }$ and ${\displaystyle f:\operatorname {Dom} f\to Y}$ is a function such that ${\displaystyle p^{0}\in \operatorname {Dom} f}$ with ${\displaystyle p^{0}}$ a limit point of ${\displaystyle \operatorname {Dom} f}$. Then we say that f is differentiable at ${\displaystyle p^{0}}$ if there exist n vectors ${\displaystyle e_{1},\ldots ,e_{n}}$ in Y, called the partial derivatives of f, such that

${\displaystyle \lim _{p\to p^{0},p\in \operatorname {Dom} f}{\frac {f(p)-f\left(p^{0}\right)-\sum _{i=1}^{n}\left(p_{i}-p_{i}^{0}\right)e_{i}}{\left\|p-p^{0}\right\|_{2}}}=0}$ in Y

where ${\displaystyle p=\left(p_{1},\ldots ,p_{n}\right)}$.

One may naturally extend the notion of continuously differentiable function to Y-valued functions defined on ${\displaystyle \Omega }$. For any ${\displaystyle k=0,1,\ldots ,\infty }$, let ${\displaystyle C^{k}\left(\Omega ;Y\right)}$ denote the vector space of all ${\displaystyle C^{k}}$ Y-valued maps defined on ${\displaystyle \Omega }$ and let ${\displaystyle C_{c}^{k}\left(\Omega ;Y\right)}$ denote the vector subspace of ${\displaystyle C^{k}\left(\Omega ;Y\right)}$ consisting of all maps in ${\displaystyle C^{k}\left(\Omega ;Y\right)}$ that have compact support.

One may then define topologies on ${\displaystyle C^{k}\left(\Omega ;Y\right)}$ and ${\displaystyle C_{c}^{k}\left(\Omega ;Y\right)}$ in the same manner as the topologies on ${\displaystyle C^{k}\left(\Omega \right)}$ and ${\displaystyle C_{c}^{k}\left(\Omega \right)}$ are defined for the space of distributions and test functions (see the article: Differentiable vector-valued functions from Euclidean space). All of this work in extending the definition of differentiability and various topologies turns out to be exactly equivalent to simply taking the completed injective tensor product:

Theorem[23] If Y is a complete Hausdorff locally convex space, then ${\displaystyle C^{k}\left(\Omega ;Y\right)}$ is canonically isomorphic to the injective tensor product ${\displaystyle C^{k}\left(\Omega \right){\widehat {\otimes }}_{\varepsilon }Y}$.

Spaces of continuous maps from a compact space

If Y is a normed space and if K is a compact set, then the ${\displaystyle \varepsilon }$-norm on ${\displaystyle C(K)\otimes Y}$ is equal to ${\displaystyle \|f\|_{\varepsilon }=\sup _{x\in K}\|f(x)\|}$.[23] If H and K are two compact spaces, then ${\displaystyle C\left(H\times K\right)\cong C\left(H\right){\widehat {\otimes }}_{\varepsilon }C\left(K\right)}$, where this canonical map is an isomorphism of Banach spaces.[23]

Spaces of sequences converging to 0

If Y is a normed space, then let ${\displaystyle l_{\infty }(Y)}$ denote the space of all sequences ${\displaystyle \left(y_{i}\right)_{i=1}^{\infty }}$ in Y that converge to the origin and give this space the norm ${\displaystyle \left\|\left(y_{i}\right)_{i=1}^{\infty }\right\|:=\sup _{i\in \mathbb {N} }\left\|y_{i}\right\|}$. Let ${\displaystyle l_{\infty }}$ denote ${\displaystyle l_{\infty }\left(\mathbb {C} \right)}$. Then for any Banach space Y, ${\displaystyle l_{\infty }{\widehat {\otimes }}_{\varepsilon }Y}$ is canonically isometrically isomorphic to ${\displaystyle l_{\infty }(Y)}$.[23]

Schwartz space of functions

We will now generalize the Schwartz space to functions valued in a TVS. Let ${\displaystyle {\mathcal {L}}\left(\mathbb {R} ^{n};Y\right)}$ be the space of all ${\displaystyle f\in C^{\infty }\left(\mathbb {R} ^{n};Y\right)}$ such that for all pairs of polynomials P and Q in n variables, ${\displaystyle \left\{P(x)Q\left(\partial /\partial x\right)f(x):x\in \mathbb {R} ^{n}\right\}}$ is a bounded subset of Y. To generalize the topology of the Schwartz space to ${\displaystyle {\mathcal {L}}\left(\mathbb {R} ^{n};Y\right)}$, we give ${\displaystyle {\mathcal {L}}\left(\mathbb {R} ^{n};Y\right)}$ the topology of uniform convergence over ${\displaystyle \mathbb {R} ^{n}}$ of the functions ${\displaystyle P(x)Q\left(\partial /\partial x\right)f(x)}$, as P and Q vary over all possible pairs of polynomials in n variables.[23]

Theorem:[23] If Y is a complete locally convex space, then ${\displaystyle {\mathcal {L}}\left(\mathbb {R} ^{n};Y\right)}$ is canonically isomorphic to ${\displaystyle {\mathcal {L}}\left(\mathbb {R} ^{n}\right){\widehat {\otimes }}_{\varepsilon }Y}$.

See also

• Auxiliary normed spaces
• Final topology
• Inductive tensor product
• Integral maps
• Nuclear operators
• Nuclear spaces
• Projective tensor product
• Topological tensor product

References

1. Trèves 2006, pp. 432-434.
2. Trèves 2006, pp. 338-345.
3. Trèves 2006, pp. 403-404.
4. Trèves 2006, pp. 432-433.
5. Trèves 2006, pp. 368-370.
6. Trèves 2006, pp. 338-343.
7. Trèves 2006, pp. 347-350.
8. Trèves 2006, pp. 351-354.
9. Trèves 2006, pp. 428-430.
10. Trèves 2006, p. 434.
11. Trèves 2006, p. 439.
12. Trèves 2006, p. 440.
13. Trèves 2006, p. 441.
14. Trèves 2006, p. 442.
15. Trèves 2006, p. 444.
16. Schaefer & Wolff 1999, p. 170.
17. Trèves 2006, p. 494.
18. Schaefer & Wolff 1999, p. 168.
19. Trèves 2006, pp. 500-502.
20. Trèves 2006, pp. 502-505.
21. Schaefer & Wolff 1999, pp. 179-184.
22. Trèves 2006, pp. 412-419.
23. Trèves 2006, pp. 446-451.

Bibliography

• Diestel, Joe (2008). The metric theory of tensor products : Grothendieck's résumé revisited. Providence, R.I: American Mathematical Society. ISBN 978-0-8218-4440-3. OCLC 185095773.
• Dubinsky, Ed (1979). The structure of nuclear Fréchet spaces. Berlin New York: Springer-Verlag. ISBN 3-540-09504-7. OCLC 5126156.
• Grothendieck, Grothendieck (1966). Produits tensoriels topologiques et espaces nucléaires (in French). Providence: American Mathematical Society. ISBN 0-8218-1216-5. OCLC 1315788.
• Husain, Taqdir (1978). Barrelledness in topological and ordered vector spaces. Berlin New York: Springer-Verlag. ISBN 3-540-09096-7. OCLC 4493665.
• Khaleelulla, S. M. (1982). Counterexamples in Topological Vector Spaces. Lecture Notes in Mathematics. 936. Berlin, Heidelberg, New York: Springer-Verlag. ISBN 978-3-540-11565-6. OCLC 8588370.
• Narici, Lawrence; Beckenstein, Edward (2011). Topological Vector Spaces. Pure and applied mathematics (Second ed.). Boca Raton, FL: CRC Press. ISBN 978-1584888666. OCLC 144216834.
• Nlend, H (1977). Bornologies and functional analysis : introductory course on the theory of duality topology-bornology and its use in functional analysis. Amsterdam New York New York: North-Holland Pub. Co. Sole distributors for the U.S.A. and Canada, Elsevier-North Holland. ISBN 0-7204-0712-5. OCLC 2798822.
• Nlend, H (1981). Nuclear and conuclear spaces : introductory courses on nuclear and conuclear spaces in the light of the duality. Amsterdam New York New York, N.Y: North-Holland Pub. Co. Sole distributors for the U.S.A. and Canada, Elsevier North-Holland. ISBN 0-444-86207-2. OCLC 7553061.
• Pietsch, Albrecht (1972). Nuclear locally convex spaces. Berlin,New York: Springer-Verlag. ISBN 0-387-05644-0. OCLC 539541.
• Robertson, A. P. (1973). Topological vector spaces. Cambridge England: University Press. ISBN 0-521-29882-2. OCLC 589250.
• Ryan, Raymond (2002). Introduction to tensor products of Banach spaces. London New York: Springer. ISBN 1-85233-437-1. OCLC 48092184.
• Schaefer, Helmut H.; Wolff, Manfred P. (1999). Topological Vector Spaces. GTM. 8 (Second ed.). New York, NY: Springer New York Imprint Springer. ISBN 978-1-4612-7155-0. OCLC 840278135.
• Trèves, François (2006) [1967]. Topological Vector Spaces, Distributions and Kernels. Mineola, N.Y.: Dover Publications. ISBN 978-0-486-45352-1. OCLC 853623322.
• Wong (1979). Schwartz spaces, nuclear spaces, and tensor products. Berlin New York: Springer-Verlag. ISBN 3-540-09513-6. OCLC 5126158.

External links

• Nuclear space at ncatlab