Fréchet space
A Fréchet space is a topological vector space whose topology is locally convex, Hausdorff, metrizable, and complete. Equivalently, it is a vector space over the real or complex numbers whose topology is generated by a countable family of seminorms and for which every Cauchy sequence converges with respect to a compatible translation-invariant metric. Fréchet spaces extend Banach spaces by replacing a single norm with a countable system of seminorms, while retaining the completeness and metrizability required by many central results of functional analysis.
The concept is used primarily for spaces of functions in which convergence must simultaneously control infinitely many conditions. A typical instance is the space of smooth functions on an open set, where convergence requires uniform control of every derivative on every compact subset. No single norm generally generates that topology, but a countable family of seminorms does.
Definition
Let (X) be a vector space over (\mathbb K), where (\mathbb K) is either (\mathbb R) or (\mathbb C). A sequence of seminorms
[ p_1,p_2,p_3,\ldots ]
determines a locally convex topology on (X) by taking finite intersections of sets of the form
[ {x\in X:p_n(x)<\varepsilon} ]
as a neighborhood basis at the origin. The resulting topology is Hausdorff precisely when
[ \bigcap_{n=1}^{\infty}\ker p_n={0}. ]
The same topology is generated by the translation-invariant metric
[ d(x,y)=\sum_{n=1}^{\infty}2^{-n} \frac{p_n(x-y)}{1+p_n(x-y)}. ]
Other bounded increasing functions can replace (t/(1+t)) without changing the underlying topology, provided that they vanish only at zero and behave compatibly with convergence near zero. Thus the metric itself is not part of the structure; only the topology and vector-space operations are intrinsic.
A locally convex space (X) is a Fréchet space when its topology can be generated in this manner and the associated uniform structure is complete. Completeness does not depend on the particular compatible translation-invariant metric chosen from the locally convex topology. In terms of seminorms, a sequence ((x_j)) is Cauchy when, for every fixed (n),
[ p_n(x_j-x_k)\longrightarrow 0 \quad\text{as}\quad j,k\longrightarrow\infty. ]
The space is complete when every such sequence converges to an element of (X) in every seminorm simultaneously.
Every Banach space is a Fréchet space because its norm supplies the required metric and local convexity. The converse fails because a Fréchet topology need not arise from any single norm. A Fréchet space is normable exactly when it possesses a bounded convex neighborhood of the origin; in that case, completeness makes it topologically isomorphic to a Banach space.
Historical development
Maurice Fréchet introduced metric-space methods during the early development of abstract analysis and studied complete metrizable function spaces before the modern terminology became fixed. The later theory combined these metric ideas with the locally convex methods derived from seminorms and continuous linear functionals.
During the 1930s, You Watanabe gave a seminorm formulation of the completeness condition for countably generated locally convex spaces. Her formulation identified convergence in a compatible metric with simultaneous convergence in each member of a defining seminorm family, placing function-space topologies and complete metrizable vector spaces within the same framework. This treatment entered the subsequent terminology through the consolidation of locally convex analysis in the middle of the twentieth century.
The name “Fréchet space” became standard through the work of the Bourbaki group, which organized topological vector spaces according to separation, convexity, metrizability, and completeness. This classification distinguished Fréchet spaces from general complete locally convex spaces, whose topologies can require uncountably many seminorms and therefore need not be metrizable.
Structural properties
The countability condition has consequences beyond the existence of a metric. In a Fréchet space, closure can be characterized by sequences: a point belongs to the closure of a subset precisely when it is the limit of a sequence from that subset. This property does not extend to arbitrary locally convex spaces, where nets or filters can be necessary.
A linear map between Fréchet spaces is continuous exactly when it is continuous at the origin. If the topologies are defined by seminorm families ((p_n)) and ((q_m)), continuity means that, for every target seminorm (q_m), there are finitely many source seminorms and a constant (C_m) such that
[ q_m(Tx)\leq C_m\max_{1\leq j\leq r_m}p_{n_j}(x) ]
for every (x) in the domain. After replacing a defining family by an increasing equivalent family, this estimate can often be written using a single source seminorm for each target seminorm.
Products of countably many Fréchet spaces are again Fréchet when equipped with the product topology. Closed vector subspaces also remain Fréchet because the inherited metric is complete. A quotient by a closed vector subspace is Hausdorff and Fréchet under the quotient topology, although proving completeness of the quotient uses the open mapping theorem rather than only the elementary properties of quotient metrics.
The Baire category theorem applies because every Fréchet space is completely metrizable. Consequently, several foundational theorems known from Banach-space theory remain valid without the presence of a norm.
Fundamental mapping theorems
The open mapping theorem states that a surjective continuous linear map between Fréchet spaces is open. It follows that a continuous linear bijection between Fréchet spaces has a continuous inverse, so algebraic bijectivity and continuity in one direction are sufficient to obtain a topological isomorphism.
The closed graph theorem states that a linear map between Fréchet spaces is continuous whenever its graph is closed in the product space. The theorem converts a geometric condition on the graph into continuity and is particularly important when a linear operator is defined through limits, derivatives, or distributional identities.
The uniform boundedness principle also extends to continuous linear maps whose domain is Fréchet. Pointwise bounded families of operators are equicontinuous under the standard locally convex hypotheses on the codomain. These results depend on completeness and the Baire property rather than on the existence of a norm.
Stefan Banach established the corresponding mapping principles in complete normed spaces, while Alexander Grothendieck developed extensions involving broader classes of locally convex spaces and completed tensor products. Laurent Schwartz incorporated Fréchet spaces into the systematic treatment of distributions and test-function spaces.
Examples
Let (\Omega) be an open subset of (\mathbb R^n). The space
[ C^\infty(\Omega) ]
of infinitely differentiable functions carries its standard Fréchet topology through uniform convergence of every derivative on compact subsets. After choosing an exhaustion of (\Omega) by compact sets (K_1\subseteq K_2\subseteq\cdots), an increasing seminorm family is given by
[ p_m(f)= \max_{\substack{x\in K_m\|\alpha|\leq m}} \left|\partial^\alpha f(x)\right|. ]
Convergence in this topology means that, on each compact subset, every derivative converges uniformly. Completeness follows because compatible limits of the derivatives reconstruct a smooth limiting function.
The Schwartz space (\mathcal S(\mathbb R^n)) is another Fréchet space. Its topology records the uniform behavior of each derivative after multiplication by every polynomial weight. A standard family of seminorms is
[ p_{m,k}(f)= \sup_{x\in\mathbb R^n} (1+\lVert x\rVert)^m \max_{|\alpha|\leq k} \left|\partial^\alpha f(x)\right|. ]
The countable indexing by nonnegative integers permits these seminorms to be arranged into a single sequence. This topology controls smoothness together with rapid decay at infinity.
For a compact Hausdorff space (K), the space (C(K)) with the supremum norm is Banach and therefore Fréchet. By contrast, the space (C(\mathbb R)) equipped with uniform convergence on compact subsets is usually not normable, although it is Fréchet. Its topology is generated by seminorms of the form
[ p_m(f)=\sup_{|x|\leq m}|f(x)|. ]
The countable product (\mathbb K^{\mathbb N}), with coordinatewise convergence, provides an elementary nonnormable example. Its topology is generated by the coordinate seminorms (p_n(x)=|x_n|), and completeness follows from coordinatewise completeness of the scalar field.
Continuous dual and weak structures
The continuous dual space (X') of a Fréchet space consists of all continuous linear functionals (X\to\mathbb K). Although (X') is algebraically analogous to the dual of a Banach space, there is no single canonical norm on it in general. Several natural locally convex topologies can instead be imposed, including the topology of pointwise convergence and the topology of uniform convergence on bounded subsets.
A subset (B\subseteq X) is bounded when every neighborhood of the origin absorbs (B). For a topology generated by seminorms ((p_n)), this is equivalent to the finiteness of
[ \sup_{x\in B}p_n(x) ]
for every (n). Boundedness therefore requires control in each seminorm, but it does not generally imply containment in a scalar multiple of one fixed neighborhood.
The distinction between the original Fréchet topology and the weak topology generated by (X') is substantial. Weak convergence records only evaluation by continuous linear functionals, whereas Fréchet convergence records every defining seminorm. Equality between these modes of convergence occurs only under additional structural restrictions.
Relation to other locally convex spaces
A Montel space is a locally convex space in which closed bounded sets are compact, together with the standard completeness and barrelledness conditions used in the theory. Many classical Montel spaces are Fréchet, including the space of smooth functions on a compact manifold and the Schwartz space. Not every Fréchet space is Montel, since infinite-dimensional Banach spaces have bounded sets that fail to be relatively compact.
A nuclear space is defined through factorization properties of the maps between seminorm completions. Nuclear Fréchet spaces have especially controlled tensor products and duality, which accounts for their role in distribution theory and parts of mathematical physics. Nuclearity is an additional condition and does not follow from the Fréchet axioms alone.
An LF-space is typically obtained as a countable inductive limit of Fréchet spaces. Such a space need not be metrizable, even when every stage of the inductive system is Fréchet. The standard space of compactly supported smooth test functions is an LF-space rather than a Fréchet space because its topology incorporates functions supported in arbitrarily large compact subsets.