Finite dimensional topological vector spaces over complete fields #
Let π
be a complete nontrivially normed field, and E
a topological vector space (TVS) over
π
(i.e we have [AddCommGroup E] [Module π E] [TopologicalSpace E] [TopologicalAddGroup E]
and [ContinuousSMul π E]
).
If E
is finite dimensional and Hausdorff, then all linear maps from E
to any other TVS are
continuous.
When E
is a normed space, this gets us the equivalence of norms in finite dimension.
Main results : #
LinearMap.continuous_iff_isClosed_ker
: a linear form is continuous if and only if its kernel is closed.LinearMap.continuous_of_finiteDimensional
: a linear map on a finite-dimensional Hausdorff space over a complete field is continuous.
TODO #
Generalize more of Mathlib.Analysis.NormedSpace.FiniteDimension
to general TVSs.
Implementation detail #
The main result from which everything follows is the fact that, if ΞΎ : ΞΉ β E
is a finite basis,
then ΞΎ.equivFun : E ββ (ΞΉ β π)
is continuous. However, for technical reasons, it is easier to
prove this when ΞΉ
and E
live in the same universe. So we start by doing that as a private
lemma, then we deduce LinearMap.continuous_of_finiteDimensional
from it, and then the general
result follows as continuous_equivFun_basis
.
The space of continuous linear maps between finite-dimensional spaces is finite-dimensional.
Equations
- (_ : FiniteDimensional π (E βL[π] F)) = (_ : FiniteDimensional π (E βL[π] F))
If π
is a nontrivially normed field, any T2 topology on π
which makes it a topological
vector space over itself (with the norm topology) is equal to the norm topology.
Any linear form on a topological vector space over a nontrivially normed field is continuous if its kernel is closed.
Any linear form on a topological vector space over a nontrivially normed field is continuous if and only if its kernel is closed.
Over a nontrivially normed field, any linear form which is nonzero on a nonempty open set is automatically continuous.
Any linear map on a finite dimensional space over a complete field is continuous.
Equations
- (_ : ContinuousLinearMapClass (E ββ[π] F') π E F') = (_ : ContinuousSemilinearMapClass (E ββ[π] F') (RingHom.id π) E F')
In finite dimensions over a non-discrete complete normed field, the canonical identification
(in terms of a basis) with π^n
(endowed with the product topology) is continuous.
This is the key fact which makes all linear maps from a T2 finite dimensional TVS over such a field
continuous (see LinearMap.continuous_of_finiteDimensional
), which in turn implies that all
norms are equivalent in finite dimensions.
The continuous linear map induced by a linear map on a finite dimensional space
Equations
- One or more equations did not get rendered due to their size.
Instances For
A surjective linear map f
with finite dimensional codomain is an open map.
Equations
- One or more equations did not get rendered due to their size.
The continuous linear equivalence induced by a linear equivalence on a finite dimensional space.
Equations
Instances For
Equations
- One or more equations did not get rendered due to their size.
Two finite-dimensional topological vector spaces over a complete normed field are continuously linearly equivalent if they have the same (finite) dimension.
Two finite-dimensional topological vector spaces over a complete normed field are continuously linearly equivalent if and only if they have the same (finite) dimension.
A continuous linear equivalence between two finite-dimensional topological vector spaces over a complete normed field of the same (finite) dimension.
Equations
Instances For
Construct a continuous linear map given the value at a finite basis.
Equations
- Basis.constrL v f = LinearMap.toContinuousLinearMap ((Basis.constr v π) f)
Instances For
The continuous linear equivalence between a vector space over π
with a finite basis and
functions from its basis indexing type to π
.
Equations
- Basis.equivFunL v = let src := Basis.equivFun v; ContinuousLinearEquiv.mk src
Instances For
Builds a continuous linear equivalence from a continuous linear map on a finite-dimensional vector space whose determinant is nonzero.
Equations
Instances For
A finite-dimensional subspace is complete.
A finite-dimensional subspace is closed.
An injective linear map with finite-dimensional domain is a closed embedding.