Documentation

TauCeti.Analysis.Normed.Operator.SeparatingSpace

The separating space of a linear map #

For a linear map T : E →ₗ[R] F between topological modules, the separating space T.separatingSpace is the set of y : F such that (0, y) lies in the closure of the graph of T. When E and F are first countable these are exactly the limits of T xₙ along sequences xₙ → 0. The separating space measures the failure of continuity of T, and it is the basic object of automatic continuity theory for Banach algebras.

The separating space is a closed submodule, and the closure of the graph of T is described fibrewise by it: (x, y) lies in the closure exactly when y - T x ∈ T.separatingSpace. Hence the graph of T is closed exactly when the separating space vanishes. Between Banach spaces the closed graph theorem turns this into the criterion Continuous T ↔ T.separatingSpace = ⊥, and more precisely the separating space is the smallest closed subspace modulo which T becomes continuous.

Sinclair's stability lemma controls separating spaces along a sequence of composites. Given Banach spaces E n and F, continuous linear maps R n : E (n + 1) → E n, and linear maps T n : E n → F with T (n + 1) = T n ∘ R n, the separating spaces of the T n form a decreasing sequence that is eventually constant. This is the analytic input to the theory of separating ideals in automatic continuity. Applied to continuous linear maps that intertwine T up to continuous errors, it shows that the closures of the images of the separating space under the successive products stabilize.

Main definitions #

Main results #

References #

The separating space of a linear map T : E →ₗ[R] F between topological modules: the vertical fibre over 0 of the closure of the graph of T. For first countable spaces it consists of the limits of T xₙ along sequences xₙ → 0 (LinearMap.mem_separatingSpace_iff_exists_seq).

Instances For

    The separating space is the vertical fibre over 0 of the closure of the graph.

    y lies in the separating space of T exactly when it is a cluster point of T at 0.

    For first countable spaces, y lies in the separating space of T exactly when T xₙ → y for some sequence xₙ → 0.

    The closure of the graph of T is the graph of T thickened by the separating space.

    The graph of a linear map is closed if and only if its separating space is trivial.

    A continuous linear map into a Hausdorff space has trivial separating space.

    A continuous linear map U sends the separating space of T into that of U ∘ T.

    If U ∘ T is continuous for a continuous linear map U into a Hausdorff space, then U vanishes on the separating space of T.

    Precomposing with a continuous linear map can only shrink the separating space.

    @[simp]

    The separating space is unchanged by adding a continuous linear map.

    Two linear maps differing by a continuous linear map have the same separating space.

    If U ∘ T - T' ∘ V is continuous for continuous linear maps U and V, then U maps the separating space of T into that of T'. Taking T' = T and U, V the left or right actions of an algebra on F and E, this says that the separating space of a map whose commutators with the actions are continuous is a submodule.

    The graph of T becomes closed after passing to the quotient by its separating space.

    A linear map between Banach spaces is continuous if and only if its separating space is trivial.

    A linear map between Banach spaces becomes continuous modulo its separating space.

    For a linear map T between Banach spaces and a continuous linear map W into a Hausdorff space, W ∘ T is continuous if and only if W vanishes on the separating space of T.

    For a linear map T between Banach spaces and a continuous linear map U into a topological vector space, the separating space of U ∘ T is the closure of the image under U of the separating space of T.

    theorem LinearMap.exists_norm_le_le_norm_mkQ_apply {𝕜 : Type u_1} {E : Type u_2} {F : Type u_3} [NontriviallyNormedField 𝕜] [NormedAddCommGroup E] [NormedSpace 𝕜 E] [NormedAddCommGroup F] [NormedSpace 𝕜 F] (T : E →ₗ[𝕜] F) {N : Submodule 𝕜 F} [IsClosed ↑N] (hT : ¬T.separatingSpace ≤ N) {δ : ℝ} (hδ : 0 < δ) (c : ℝ) :
    ∃ (x : E), ‖x‖ ≤ δ ∧ c ≤ ‖N.mkQ (T x)‖

    If the separating space of a linear map T between normed spaces is not contained in a closed subspace N, then modulo N the map T sends arbitrarily small vectors to arbitrarily large ones.

    The stability lemma #

    theorem TauCeti.antitone_separatingSpace {𝕜 : Type u_1} {F : Type u_2} [Ring 𝕜] [AddCommGroup F] [Module 𝕜 F] [TopologicalSpace F] [IsTopologicalAddGroup F] [ContinuousConstSMul 𝕜 F] {E : ℕ → Type u_3} [(n : ℕ) → AddCommGroup (E n)] [(n : ℕ) → Module 𝕜 (E n)] [(n : ℕ) → TopologicalSpace (E n)] [∀ (n : ℕ), IsTopologicalAddGroup (E n)] [∀ (n : ℕ), ContinuousConstSMul 𝕜 (E n)] {T : (n : ℕ) → E n →ₗ[𝕜] F} {R : (n : ℕ) → E (n + 1) →L[𝕜] E n} (hT : ∀ (n : ℕ), T (n + 1) = T n ∘ₗ ↑(R n)) :
    Antitone fun (n : ℕ) => (T n).separatingSpace

    For linear maps T n : E n →ₗ F with T (n + 1) = T n ∘ R n for continuous linear maps R n : E (n + 1) → E n, the separating spaces of the T n decrease.

    theorem TauCeti.eventuallyConst_separatingSpace {𝕜 : Type u_1} {F : Type u_2} [NontriviallyNormedField 𝕜] [NormedAddCommGroup F] [NormedSpace 𝕜 F] {E : ℕ → Type u_3} [(n : ℕ) → NormedAddCommGroup (E n)] [(n : ℕ) → NormedSpace 𝕜 (E n)] {R : (n : ℕ) → E (n + 1) →L[𝕜] E n} [CompleteSpace F] [∀ (n : ℕ), CompleteSpace (E n)] {T : (n : ℕ) → E n →ₗ[𝕜] F} (hT : ∀ (n : ℕ), T (n + 1) = T n ∘ₗ ↑(R n)) :

    Sinclair's stability lemma. Let E n and F be Banach spaces, let R n : E (n + 1) → E n be continuous linear maps, and let T n : E n →ₗ F be linear maps, not assumed continuous, with T (n + 1) = T n ∘ R n, so that T n = T 0 ∘ R 0 ∘ ⋯ ∘ R (n - 1). Then the decreasing sequence of separating spaces of the T n is eventually constant.

    Stability of the images of the separating space #

    theorem LinearMap.eventuallyConst_topologicalClosure_map_separatingSpace {𝕜 : Type u_1} {E : Type u_2} {F : Type u_3} [NontriviallyNormedField 𝕜] [NormedAddCommGroup E] [NormedSpace 𝕜 E] [CompleteSpace E] [NormedAddCommGroup F] [NormedSpace 𝕜 F] [CompleteSpace F] (T : E →ₗ[𝕜] F) (R : ℕ → E →L[𝕜] E) (S : ℕ → F →L[𝕜] F) (hRS : ∀ (n : ℕ), Continuous ⇑(↑(S n) ∘ₗ T - T ∘ₗ ↑(R n))) {V : ℕ → F →L[𝕜] F} (hV0 : V 0 = ContinuousLinearMap.id 𝕜 F) (hV : ∀ (n : ℕ), V (n + 1) = V n ∘SL S n) :

    Stability of the images of the separating space. Let T : E → F be a linear map between Banach spaces, and let R n on E and S n on F be continuous linear maps such that each S n ∘ T - T ∘ R n is continuous. If V n = S 0 ∘ ⋯ ∘ S (n - 1), then the closures of the images of the separating space of T under V n are eventually constant.