Whitney topologies

In mathematics, and especially differential topology, functional analysis and singularity theory, the Whitney topologies are a countably infinite family of topologies defined on the set of smooth mappings between two smooth manifolds. They are named after the American mathematician Hassler Whitney.

Construction
Let M and N be two real, smooth manifolds. Furthermore, let C∞(M,N) denote the space of smooth mappings between M and N. The notation C∞ means that the mappings are infinitely differentiable, i.e. partial derivatives of all orders exist and are continuous.

Whitney Ck-topology
For some integer k ≥ 0, let Jk(M,N) denote the k-jet space of mappings between M and N. The jet space can be endowed with a smooth structure (i.e. a structure as a C∞ manifold) which make it into a topological space. This topology is used to define a topology on C∞(M,N).

For a fixed integer k ≥ 0 consider an open subset U ⊂ Jk(M,N), and denote by Sk(U) the following:
 * $$ S^k(U) = \{ f \in C^{\infty}(M,N) : (J^kf)(M) \subseteq U \} . $$

The sets Sk(U) form a basis for the Whitney Ck-topology on C∞(M,N).

Whitney C∞-topology
For each choice of k &ge; 0, the Whitney Ck-topology gives a topology for C∞(M,N); in other words the Whitney Ck-topology tells us which subsets of C∞(M,N) are open sets. Let us denote by Wk the set of open subsets of C∞(M,N) with respect to the Whitney Ck-topology. Then the Whitney C∞-topology is defined to be the topology whose basis is given by W, where:
 * $$ W = \bigcup_{k=0}^{\infty} W^k . $$

Dimensionality
Notice that C∞(M,N) has infinite dimension, whereas Jk(M,N) has finite dimension. In fact, Jk(M,N) is a real, finite-dimensional manifold. To see this, let ℝk[x1,...,xm] denote the space of polynomials, with real coefficients, in m variables of order at most k and with zero as the constant term. This is a real vector space with dimension
 * $$ \dim\left\{\R^k[x_1,\ldots,x_m]\right\} = \sum_{i=1}^k \frac{(m+i-1)!}{(m-1)! \cdot i!} = \left( \frac{(m+k)!}{m!\cdot k!} - 1 \right) . $$

Writing a = dim{ℝk[x1,...,xm]} then, by the standard theory of vector spaces ℝk[x1,...,xm] ≅ ℝa, and so is a real, finite-dimensional manifold. Next, define:
 * $$B_{m,n}^k = \bigoplus_{i=1}^n \R^k[x_1,\ldots,x_m], \implies \dim\left\{B_{m,n}^k\right\} = n \dim \left\{ A_m^k \right\} = n \left( \frac{(m+k)!}{m!\cdot k!} - 1 \right) . $$

Using b to denote the dimension Bkm,n, we see that Bkm,n ≅ ℝb, and so is a real, finite-dimensional manifold.

In fact, if M and N have dimension m and n respectively then:
 * $$ \dim\!\left\{J^k(M,N)\right\} = m + n + \dim \!\left\{B_{n,m}^k\right\} = m + n\left( \frac{(m+k)!}{m!\cdot k!}\right). $$

Topology
Given the Whitney C∞-topology, the space C∞(M,N) is a Baire space, i.e. every residual set is dense.