10.12 Spaces of Linear Maps

Definition 10.12.1 (Space of Bounded Linear Maps).label Let $E, F$ be TVSs over $K \in \RC$, $\sigma \subset 2^{E}$ be an ideal, and $k \in \nat$. The space $B_{\sigma}^{k}(E; F)$ is the set of all $k$-linear maps $T: E^{k} \to F$ with $T(S^{k}) \in \mathfrak{B}(F)$ for all $S \in \sigma$, equipped with the $\bracsn{S^k| S \in \sigma}$-uniform topology.

Let $\fB \subset 2^{E}$ be the collection of all bounded subsets of $E$, then $B_{\sigma}(E; F) = B(E; F)$ is the space of bounded linear maps from $E$ to $F$.

Proposition 10.12.2 ([III.3.3, SW99]).label Let $E, F$ be TVSs over $K \in \RC$, $\sigma \subset 2^{E}$ be an ideal, and $A \subset B_{\sigma}(E; F)$, then the following are equivalent:

  1. (1)

    $A \subset B_{\sigma}(E; F)$ is bounded with respect to the $\sigma$-uniform topology.

  2. (2)

    For each $V \in \cn_{F}(0)$, $\bigcap_{T \in A}T^{-1}(V)$ absorbs every $S \in \sigma$.

  3. (3)

    For every $S \in \sigma$, $\bigcup_{T \in A}T(A)$ is bounded in $F$.

Proposition 10.12.3.label Let $E, F$ be TVSs over $K \in \RC$, $\sigma \subset 2^{E}$ be a covering ideal, and $k \in \natp$, then

  1. (1)

    The map

    \[I: B_{\sigma}^{k}(E; B_{\sigma}(E; F)) \to B^{k+1}_{\sigma}(E; F)\]

    defined by

    \[(IT)(x_{1}, \cdots, x_{k+1}) = T(x_{1}, \cdots, x_{k})(x_{k+1})\]

    is an isomorphism.

  2. (2)

    The map

    \[I: \underbrace{B_{\sigma}(E; B_{\sigma}(E; \cdots)))}_{k \text{ times}}\to B^{k}_{\sigma}(E; F)\]

    defined by

    \[IT(x_{1}, \cdots, x_{k}) = T(x_{1})\cdots (x_{k})\]

    is an isomorphism.

which allows the identification

\[\underbrace{B_{\sigma}(E; B_{\sigma}(E; \cdots)))}_{k \text{ times}}= B^{k}_{\sigma}(E; F)\]

under the map $I$ in (2).

Proof. (1): To see that $I$ is surjective, let $T \in B_{\sigma}^{k+1}(E; F)$ and

\[I^{-1}T: E \to B_{\sigma}(E; F) \quad x \mapsto T(x, \cdot)\]

Let $(x_{1}, \cdots, x_{k}) \in E^{k}$ and $S \in \sigma$. Since $\sigma$ is a covering ideal, assume without loss of generality that $\bracsn{x_j}_{1}^{k} \subset S$. In which case,

\[T(x_{1}, \cdots, x_{k}, S) \subset T(S^{k+1}) \in \mathfrak{B}(F)\]

by assumption. Thus $I^{-1}T(x_{1}, \cdots, x_{k}) \in B_{\sigma}(E; F)$.

In addition, for any $S_{1} \in \sigma$ and entourage $E(S_{2}, U)$ of $B_{\sigma}(E; F)$ where $S_{2} \in \sigma$ and $U$ is an entourage of $F$, there exists $S \in \sigma$ with $S \supset S_{1} \cup S_{2}$. Given that $T(S^{k+1}) \in \mathfrak{B}(F)$, there exists $\lambda > 0$ such that $T(S^{k+1}) \subset \lambda U(0)$. In which case, $I^{-1}T(S^{k}) \subset \lambda E(S, U)(0)$ and $I^{-1}T(S^{k}) \in B(B_{\sigma}(E; F))$. Thus $I^{-1}T \in B^{k}_{\sigma}(E; B_{\sigma}(E; F))$.

It remains to show that $I$ and $I^{-1}$ is continuous. To this end, let $S \in \sigma$ and $U$ be an entourage of $F$, then for any $T \in E(S^{k}, E(S, U))(0)$, $IT \in E(S^{k+1}, U)(0)$, so $I$ is continuous.

On the other hand, let $S_{1} \in \sigma$ and $E(S_{2}, U)$ be an entourage of $B_{\sigma}(E; F)$ where $S_{2} \in \sigma$ and $U$ is an entourage of $F$. Let $S \in \sigma$ with $S \supset S_{1} \cup S_{2}$, then for any $T \in E(S^{k+1}, U)(0)$, $I^{-1}T \in E(S^{k}, E(S, U))(0)$. Thus $I^{-1}$ is continuous as well.

(2): The case for $k = 2$ is given by (1). If the proposition holds for $k \in \natp$, then

\[\underbrace{B_{\sigma}(E; B_{\sigma}(E; \cdots)))}_{k+1 \text{ times}}= B^{k}_{\sigma}(E; B_{\sigma}(E; F)) = B^{k+1}_{\sigma}(E; F)\]

Thus (2) holds for all $k \in \natp$.$\square$

Definition 10.12.4 (Strong Operator Topology).label Let $E, F$ be TVSs over $K \in \RC$, $\fF \subset 2^{E}$ be the collection of finite subsets of $E$, then the $\fF$-uniform topology on $F^{E}$ is the strong operator topology.

The space $L_{s}(E; F)$ denotes $L(E; F)$ equipped with the strong operator topology.

Proposition 10.12.5.label Let $E, F$ be TVSs over $K \in \RC$ and $\net{T}\subset L(E; F)$ and $T \in L_{s}(E; F)$. If

  1. (a)

    There exists a dense subset $S \subset E$ such that $T_{\alpha} x \to Tx$ strongly for all $x \in S$.

  2. (b)

    $\bracs{T_\alpha|\alpha \in A}$ is uniformly equicontinuous.

then $T_{\alpha} \to T$ in $L_{s}(E; F)$.

Proof. Let $x \in E$, $U \in \cn_{F}(Tx)$, and $V \in \cn_{F}(Tx)$ be balanced such that $V + V + V \subset U$. By (b), there exists a balanced neighbourhood $W \in \cn_{E}(0)$ such that $T(W) \cup \bigcup_{\alpha \in A}T_{\alpha}(W) \subset V$. By (a), there exists $y \in S \cap (x + W)$ and $\alpha_{0} \in A$ such that for all $\alpha \ge \alpha_{0}$, $T_{\alpha} y - Ty \in V$. In which case, for any $\alpha \ge \alpha_{0}$,

\[T_{\alpha} x - Tx = \underbrace{T_\alpha x - T_\alpha y}_{\in V}+ \underbrace{T_\alpha y - Ty}_{\in V}+ \underbrace{Ty - Tx}_{\in V}\in U\]

$\square$

Definition 10.12.6 (Weak Operator Topology).label Let $E, F$ be TVSs over $K \in \RC$, $\fF \subset 2^{E}$ be the collection of finite subsets of $E$, then the $\fF$-uniform topology on $F_{w}^{E}$ is the weak operator topology.

The space $L_{w}(E; F) = L_{s}(E; F_{w})$ denotes $L(E; F)$ equipped with the weak operator topology.

Definition 10.12.7 (Bounded Convergence Topology).label Let $E, F$ be TVSs over $K \in \RC$, $\fB \subset 2^{E}$ be the collection of bounded subsets of $E$, then the $\fB$-uniform topology on $L(E; F)$ is the topology of bounded convergence.

The space $L_{b}(E; F)$ denotes $L(E; F)$ equipped with the topology of bounded convergence.

Proposition 10.12.8.label Let $E, F$ be TVSs over $K \in \RC$ with $F$ being separated, then:

  1. (1)

    $\hom(E; F)$ is a closed subspace of $F^{E}$ with respect to the product topology.

  2. (2)

    $B(E; F)$ is a closed subspace of $F^{E}$ with respect to the topology of bounded convergence. In particular, if $F$ is complete, then so is $B(E; F)$.

Proof. (1): For each $x, y \in E$ and $\lambda \in K$, the mappings

\[\phi_{x, y}: F^{E} \to F \quad T \mapsto Tx + Ty - T(x + y)\]

and

\[\psi_{x, \lambda}: F^{E} \to F \quad T \mapsto T(\lambda x) - \lambda Tx\]

are continuous with respect to the product topology. Since

\[\hom(E; F) = \bigcap_{x, y \in E}\bracsn{\phi_{x, y} = 0}\cap \bigcap_{\substack{x \in E \\ \lambda \in K}}\bracsn{\psi_{x, \lambda} = 0}\]

and $\bracs{0}$ is closed in $F$, $\hom(E; F)$ is a closed subspace of $F^{E}$.

(2): By Definition 10.11.2 and (1), the space of bounded functions and the space of linear functions from $E$ to $F$ are closed subspaces of $F^{E}$ with respect to the topology of bounded convergence. Therefore $B(E; F)$ is also a closed subspace.$\square$