Definition

Let EE and FF be over the same field. A continuous linear map T:EFT:E\to F is a that is continuous for the given topologies. Linearity makes continuity at one point equivalent to continuity everywhere. In particular, TT is continuous exactly when, for every VV of 00 in FF, there is a neighborhood UU of 00 in EE such that

T(U)V.T(U)\subseteq V.

Thus continuity compares the chosen topologies on the source and target; it is not an algebraic property of the underlying linear map.

Normed-space specialization

If EE and FF are , continuity is equivalent to the existence of C0C\geq 0 such that

TxFCxE(xE).\lVert Tx\rVert_F\leq C\lVert x\rVert_E \qquad (x\in E).

This familiar bounded-operator criterion is a special feature of norm topologies. For general topological vector spaces, continuity is expressed with zero-neighborhoods or families of seminorms rather than one .

Locally convex criterion

Suppose EE and FF are . For every continuous qq on FF, continuity of TT implies that there are continuous seminorms p1,,pnp_1,\ldots,p_n on EE and C>0C>0 with

q(Tx)Cmax1jnpj(x).q(Tx)\leq C\max_{1\leq j\leq n}p_j(x).

Conversely, such estimates for a defining family of seminorms on FF imply continuity. This is the standard form used for spaces of smooth functions and distributions.

Categorical role

Identity maps and composites of continuous linear maps are continuous and linear. Topological vector spaces with these maps therefore form a category. A linear bijection need not be an isomorphism in this category: its inverse must also be continuous.

The zero-neighborhood criterion and the theory of spaces of continuous linear mappings are treated in Bourbaki, Chapters I and III.

References
  1. Nicolas Bourbaki, Topological Vector Spaces: Chapters 1–5, Springer, 2003. Springer DOI record. Relevant: Chapters I and III.
  2. Helmut H. Schaefer and Manfred P. Wolff, Topological Vector Spaces, 2nd ed., Springer, 1999. Springer DOI record. Relevant: Chapter III, “Linear Mappings.”