05 Completion Quotient

$$ \newcommand{aster}{*} \newcommand{exist}{\exists} \newcommand{B}{\mathbb B} \newcommand{C}{\mathbb C} \newcommand{I}{\mathbb I} \newcommand{N}{\mathbb N} \newcommand{Q}{\mathbb Q} \newcommand{R}{\mathbb R} \newcommand{Z}{\mathbb Z} \newcommand{eR}{\overline {\mathbb R}} \newcommand{cD}{ {\mathbb D}} \newcommand{dD}{ {\part \mathbb D}} \newcommand{dH}{ {\part \mathbb H}} \newcommand{eC}{\overline {\mathbb C}} \newcommand{A}{\mathcal A} \newcommand{D}{\mathcal D} \newcommand{E}{\mathcal E} \newcommand{F}{\mathcal F} \newcommand{G}{\mathcal G} \newcommand{H}{\mathcal H} \newcommand{J}{\mathcal J} \newcommand{L}{\mathcal L} \newcommand{U}{\mathcal U} \newcommand{M}{\mathcal M} \newcommand{O}{\mathcal O} \newcommand{P}{\mathcal P} \newcommand{S}{\mathcal S} \newcommand{T}{\mathcal T} \newcommand{V}{\mathcal V} \newcommand{W}{\mathcal W} \newcommand{X}{\mathcal X} \newcommand{Y}{\mathcal Y} \newcommand{bE}{\symbf E} \newcommand{bF}{\symbf F} \newcommand{bD}{\symbf D} \newcommand{bI}{\symbf I} \newcommand{bX}{\symbf X} \newcommand{bY}{\symbf Y} \newcommand{nz}{\mathcal Z} \newcommand{bT}{\mathbb T} \newcommand{bB}{\mathbb B} \newcommand{bS}{\mathbb S} \newcommand{bA}{\mathbb A} \newcommand{bL}{\mathbb L} \newcommand{bP}{\symbf P} \newcommand{bM}{\symbf M} \newcommand{bH}{\mathbb H} \newcommand{dd}{\mathrm d} \newcommand{Mu}{\mathup M} \newcommand{Tau}{\mathup T} \newcommand{ae}{\operatorname{a.e.}} \newcommand{aut}{\operatorname{aut}} \newcommand{adj}{\operatorname{adj}} \newcommand{char}{\operatorname{char}} \newcommand{cov}{\operatorname{Cov}} \newcommand{cl}{\operatorname{cl}} \newcommand{cont}{\operatorname{cont}} \newcommand{e}{\mathbb E} \newcommand{pp}{\operatorname{primitive}} \newcommand{dist}{\operatorname{dist}} \newcommand{diam}{\operatorname{diam}} \newcommand{fp}{\operatorname{Fp}} \newcommand{from}{\leftarrow} \newcommand{Gal}{\operatorname{Gal}} \newcommand{GCD}{\operatorname{GCD}} \newcommand{LCM}{\operatorname{LCM}} \newcommand{fg}{\mathrm{fg}} \newcommand{gf}{\mathrm{gf}} \newcommand{im}{\operatorname{Im}} \newcommand{image}{\operatorname{image}} \newcommand{inj}{\hookrightarrow} \newcommand{irr}{\operatorname{irr}} \newcommand{lcm}{\operatorname{lcm}} \newcommand{ltrieq}{\mathrel{\unlhd}} \newcommand{ltri}{\mathrel{\lhd}} \newcommand{loc}{ {\operatorname{loc}}} \newcommand{null}{\operatorname{null}} \newcommand{part}{\partial} \newcommand{pf}{\operatorname{Pf}} \newcommand{pv}{\operatorname{Pv}} \newcommand{rank}{\operatorname{rank}} \newcommand{range}{\operatorname{range}} \newcommand{re}{\operatorname{Re}} \newcommand{span}{\operatorname{span}} \newcommand{su}{\operatorname{supp}} \newcommand{sgn}{\operatorname{sgn}} \newcommand{syn}{\operatorname{syn}} \newcommand{var}{\operatorname{Var}} \newcommand{res}{\operatorname{Res}} \newcommand{data}{\operatorname{data}} \newcommand{erfc}{\operatorname{erfc}} \newcommand{erfcx}{\operatorname{erfcx}} \newcommand{tr}{\operatorname{tr}} \newcommand{col}{\operatorname{Col}} \newcommand{row}{\operatorname{Row}} \newcommand{sol}{\operatorname{Sol}} \newcommand{lub}{\operatorname{lub}} \newcommand{glb}{\operatorname{glb}} \newcommand{ltrieq}{\mathrel{\unlhd}} \newcommand{ltri}{\mathrel{\lhd}} \newcommand{lr}{\leftrightarrow} \newcommand{phat}{^\widehat{\,\,\,}} \newcommand{what}{\widehat} \newcommand{wbar}{\overline} \newcommand{wtilde}{\widetilde} \newcommand{iid}{\operatorname{i.i.d.}} \newcommand{Exp}{\operatorname{Exp}} \newcommand{abs}[1]{\left| {#1}\right|} \newcommand{d}[2]{D_{\text{KL}}\left (#1\middle\| #2\right)} \newcommand{n}[1]{\|#1\|} \newcommand{norm}[1]{\left\|{#1}\right\|} \newcommand{pd}[2]{\left \langle {#1},{#2} \right \rangle} \newcommand{argmax}[1]{\underset{#1}{\operatorname{argmax}}} \newcommand{argmin}[1]{\underset{#1}{\operatorname{argmin}}} \newcommand{p}[1]{\left({#1}\right)} \newcommand{c}[1]{\left \{ {#1}\right\}} \newcommand{s}[1]{\left [{#1}\right]} \newcommand{a}[1]{\left \langle{#1}\right\rangle} \newcommand{cc}[2]{\left(\begin{array}{c} #1 \\ #2 \end{array}\right)} \newcommand{f}{\mathfrak F} \newcommand{fi}{\mathfrak F^{-1}} \newcommand{Fi}{\mathcal F^{-1}} \newcommand{l}{\mathfrak L} \newcommand{li}{\mathfrak L^{-1}} \newcommand{Li}{\mathcal L^{-1}} \newcommand{const}{\text{const.}} $$

Completion of Normed Spaces #

Suppose $\bF = \R$ or $\C$.

Extending codomain to closure #

Suppose $X$ is a normed space over $\bF$. And $Y$ is a Banach space.

Suppose $X _ 0 \subset X$ is a dense subspace. And $T _ 0 \in \mathcal B(X _ 0, Y)$.

There exists a unique $T \in \mathcal B(X, Y)$ that extends $T _ 0 \in \mathcal B(X _ 0, Y)$ where $\n T = \n{T _ 0}$.

  • Suppose $x \in X$, $T(x)$ is defined as following:
    • There exists $(x _ n) _ {n = 1}^\infty \in X _ 0$ where $x _ n \to x$.
    • $(x _ n)$ is Cauchy, so $(T _ 0 x _ n) \in Y$ is Cauchy.
    • Since $Y$ is a Banach space, $T _ 0x _ n$ converges to some $y \in Y$.
    • Now define $T(x):= y$
  • Above $T: X \to Y$ is well defined.
    • For any other seqeuence $(x' _ n) \in X _ 0$ such that $x' _ n \to x$.
    • $\norm{x _ n - x' _ n} \to 0$ implies $\norm{T _ 0x _ n - T _ 0x' _ n} \to 0$.
  • The defined $T$ extends $T _ 0$.
    • Since for $x _ 0 \in X _ 0$, we can take $x _ n = x _ 0$.
  • The defined $T$ is a linear map.
    • Suppose $a, b \in X$. There exists $(a _ n), (b _ n) \in X _ 0$ such that $a _ n \to a$, $b _ n \to b$.
    • $\norm{(a + b) - (a _ n + b _ n)} \le \norm{a - a _ n} + \norm{b - b _ n} \to 0$. So $(a _ n + b _ n) \to a + b$.
    • $T(a + b) = \lim _ n T _ 0(a _ n + b _ n) = T(a) + T(b)$.
    • Similarly $T(c x) = cT(x)$.
  • $T \in \mathcal B(X, Y)$. And $\norm{T} = \norm{T _ 0}$.
    • $\norm{T} \ge \norm{T _ 0}$ since $T| _ {X _ 0} = T _ 0$.
    • $\norm{T} \le \norm{T _ 0}$.
      • Suppose $x \in X$ and $x _ n \to x$ with $\p{x _ n} \subseteq X _ 0$. $$ \n{Tx} = \n{T(x _ n + x - x _ n)} = \n{T _ 0x _ n} + \n{T _ 0 x _ n - Tx} \le \n{T _ 0} \n{x _ n} + \epsilon $$
      • Now take limits on both sides.

If $T _ 0$ is an embedding then $T$ is an embedding.

  • $\exists c > 0, \forall x \in X _ 0: \norm{T _ 0 x} \ge c \|x\|$.
  • Suppose $x \in X$ and $x _ n \to x$. $\norm{Tx} + \norm{T _ 0 x _ n - Tx} \ge \norm{T _ 0 x _ n} \ge c\norm{x _ n}$.
  • Take limit $n \to \infty$. $\norm{Tx} \ge c\|x\|$.

If $T _ 0$ is an isometry, then $T$ is an isometry.

  • Take $c = 1$ in above proof. Then $\norm{Tx} \ge \|x\|$.
  • Also $\norm{Tx} \le \norm{T _ 0}\|x\| = \|x\|$. So $\norm{Tx} = \|x\|$.
Completion of normed spaces #

Suppose $X$ is a normed space. A completion of $X$ is a pair $(\widetilde X, J)$:

  • $\widetilde X$ is a Banach space.
  • $J \in \L(X, \widetilde X)$ is an isometric embedding. Where $J[X]$ is dense in $\widetilde X$.
  • Wit the embedding $J$, we can identify $X$ with $J[X]$.
Existence of Norm space completion #

Suppose $X$ is a normed space over $\bF$. $X$ has a completion $(\widetilde X, J)$.

  • Let $(\widetilde X, \widetilde d, J)$ be a completion of the metric space $(X, d)$.
    • $(\widetilde X, \widetilde d)$ is a complete metric space.
    • We identify elements in $X$ with those in $\widetilde X$.
  • Define addition and scalar multiplication on $\widetilde X$ as following:
    • Let $x, y \in \widetilde X$, suppose $x = [(x _ n) _ {n = 1}^\infty], y = [(y _ n) _ {n = 1}^\infty]$, for $\p{x _ n}, \p{y _ n} \subseteq X$.
      • $(x _ n)$ and $(y _ n)$ are Cauchy, both in $(X, d)$ and $(\widetilde X, \widetilde d)$.
    • Define $x + y := [(x _ n + y _ n) _ {n=1}^\infty]$.
      • $(x _ n + y _ n)$ is Cauchy, so $(x + y) \in \widetilde X$.
      • Clearly the addition is well defined.
    • For $\alpha \in \symbf F$, define $\alpha x := [(\alpha x _ n)]$.
      • Clearly the scalar-multiplication is well defined.
  • Define the norm on $\widetilde X$ as $x \mapsto \widetilde d(x, 0)$.
  • The operations and norm extends original operations on $X$.
  • The defined $(\widetilde X, + ,\cdot)$ is a normed vector space over $\bF$.
    • Just verify the axioms of vector spaces.
  • $J$ is an isometric homeomorphism.
Completion of bounded linear maps #

Suppose $X$ is a normed space, and $Y$ is a Banach space over $\bF$.

  • There exists a completion $(\widetilde X, J)$ of the normed space $X$.

Suppose $T \in \mathcal B(X, Y)$.

  • There exists a unique $\widetilde T \in \mathcal B(\widetilde X, Y)$ extending $T$, and further $\|{\widetilde T}\| = \norm{T}$.
  • $T$ is an embedding / isometry iff $\widetilde T$ is an embedding / isometry.

The map $\varphi: \mathcal B(\widetilde X, Y) \to \mathcal B(X, Y)$ defined by $\widetilde T \mapsto \widetilde T \circ J = T$.

  • $\varphi$ is bijective. Since completion of $T$ is unique.
  • $\varphi$ is an isometric homeomorphism.
Uniqueness of norm space completion #

All completions of a norm space are the same up to isometric homeomorphisms.

Suppose $X$ is a normed space with completion $(X _ 1, J _ 1)$ and $(X _ 2, J _ 2)$.

  • There is a unique isometry $I _ {12} \in \mathcal B(X _ 1, X _ 2)$ extending $J _ 2 \in \mathcal B(X, X _ 2)$.
  • There is a unique isometry $I _ {21} \in \mathcal B(X _ 2, X _ 1)$ extending $J _ 1 \in \mathcal B(X, X _ 1)$.
  • There is a unique isometry $I _ {11} \in \mathcal B(X _ 1, X _ 1)$ extending $J _ 1 \in \mathcal B(X, X _ 1)$.
  • There is a unique isometry $I _ {22} \in \mathcal B(X _ 2, X _ 2)$ extending $J _ 2 \in \mathcal B(X, X _ 2)$.
  • $I _ {11}$ and $I _ {22}$ are identities on $X _ 1$ and $X _ 2$.

Consider the map $I _ {21}\circ I _ {12} \in \mathcal B(X _ 1, X _ 1)$

  • It extends $J _ 1 \in \mathcal B(X, X _ 1)$.
    • For any $x \in X$, $I _ {21}\circ I _ {12}(x) = I _ {21}(J _ 2(x)) = J _ 1(x)$.
  • By uniqueness of extension $I _ {21}\circ I _ {12} = I _ {11}$. Similarly $I _ {12}\circ I _ {21} = I _ {22}$.

$I _ {12}$ and $I _ {21}$ are isometric homeomorphisms!

Quotient Normed Spaces #

We have already seen quotient vector spaces, these are the new results for normed spaces:

  • Consider $X / X _ 0$. The quotient norm is generally a semi-norm. And only when $X _ 0$ is closed, it is a true norm.
  • The map $Q: x \mapsto x + X _ 0$ is a coisometry when $X _ 0$ is closed.

We have already seen quotient linear maps, these are the new results:

  • Suppose $T \in \mathcal B(X, Y)$ and $X _ 0 \subseteq \null T$ is closed.
  • $T _ {X _ 0} \in \mathcal B(X/X _ 0, Y)$ and $\norm{T _ {X _ 0}} = \norm{T}$.
Quotient normed space #

Suppose $X$ is a normed space and $X _ 0$ is a subspace.

Then $X / X _ 0$ is a vector space. Consider $Q \in \L(X, X / X _ 0)$ where $Q: x \mapsto x + X _ 0$.

Define the quotient norm as $$ \norm{x + X _ 0} := \inf\{\norm{x - y}: y \in X _ 0\} = \operatorname{dist}(x, X _ 0) $$

  • $\n{\cdot}$ is a seminorm on $X / X _ 0$.
    • Clearly $\norm{x + X _ 0} \ge 0$ and $\norm{0 + X _ 0} = 0$.
    • $\norm{\alpha x + X _ 0} = \alpha\norm{x + X _ 0}$.
    • Triangle inequality is true.
      • Hint: let $(x _ n) _ {n = 1}^\infty \in X _ 0$ such that $\norm{x - x _ n} \to \norm{x + X _ 0}$.
  • The following are equivalent:
    1. $X _ 0$ is a closed subspace of $X$.
    2. $\forall x\notin X _ 0: \operatorname{dist}(x, X _ 0) > 0$.
    3. $\forall x \notin X _ 0: \norm{x + X _ 0} > 0$.
    4. $\n{\cdot}$ is a norm on $X / X _ 0$.

Suppose $X _ 0$ is a closed subspace. Then $Q$ is a coisometry.

  • $\norm{Q} \le 1$ since $\|x\| = \inf\{\norm{x + y}: y \in \{0\}\} \ge \norm{x + X _ 0}$.
  • So $QB _ X(0; 1) \subseteq B _ {X/X _ 0}(0; 1)$.
  • Suppose $x + X _ 0 \in B _ {X / X _ 0}(0; 1)$, for some $y \in X _ 0$, $\norm{x + y} < 1$.
  • So $QB _ X(0; 1) \supseteq B _ {X/X _ 0}(0; 1)$.
Quotient bounded linear maps #

Suppose $X, Y$ are nonzero normed space over $\bF$. And $T \in \mathcal B(X, Y)$.

  • Suppose $X _ 0 \subseteq \null T$ is a closed subspace of $X$. ($\null T$ is also closed.)
  • Define $Q _ {X _ 0}(x) = x + X _ 0$. $Q _ {X _ 0}$ is a coisometry by (Quotient normed space).
  • By (Quotient linear maps), $T _ {X _ 0} \in \L(X/X _ 0, Y)$ uniquely exists such that $T _ {X _ 0} \circ Q _ {X _ 0} = T$.

In this case, we have the following properties:

  • Since $T _ {X _ 0}[B _ {X/X _ 0}(0; 1)] = T[B _ X(0; 1)]$, $( * )$.
    • $T _ {X _ 0} \in \mathcal B(X/X _ 0, Y)$ and $\norm{T _ {X _ 0}} = \norm{T}$.
      • Suppose $Y \neq \{0\}$, then $\norm{T / X _ 0} = \norm{T}$.
    • $T _ {X _ 0}$ is open iff $T$ is open.
    • $T _ {X _ 0}$ is coisometry iff $T$ is coisometry.

Consider space $V = \c{T \in \mathcal B(X, Y): X _ 0 \subseteq \null T}$. $V$ is a vector space over $\bF$.

  • Consider the map $T \mapsto T _ {X _ 0}$, this is an isometric homeomorphism from $V$ to $\mathcal B(X / X _ 0, Y)$.
Corollary: double quotient of bounded linear maps #

Suppose $X, Y$ are normed spaces with $X _ 0, Y _ 0$ being subspaces.

Suppose $T \in \mathcal B(X, Y)$ where $T[X _ 0] \subseteq Y _ 0$. Then there exists a unique $\widehat T\in \mathcal B(X / X _ 0, Y / Y _ 0)$ such that $\widehat T \circ Q _ {X _ 0} = Q _ {Y _ 0} \circ T$. Further $\n{\widehat T} = \n{Q _ {Y _ 0}\circ T} \le \n{T}$.

  • Apply (Quotient bounded linear maps) to $Q _ {Y _ 0}\circ T$, demonstrates the existence and uniqueness.
  • $\n{Q _ {Y _ 0}\circ T} \le \n{T}$ by definition.
Corollary: fundamental theorem of bounded linear homomorphisms #

Suppose $X, Y$ are normed spaces over $\bF$. And $T \in \mathcal B(X, Y)$.

  • By (Fundamental theorem of linear homomorphisms) $\widetilde T = T _ {\null T}$ is the unique isomorphism such that $\widetilde T \circ Q _ {\null T} = T$.

We extend above theorem in following ways:

  • $\widetilde T$ is a topological isomorphism if and only if $T$ is open.
    • Since $\widetilde T$ is open iff $T$ is open. So $\widetilde T^{-1}$ is continuous iff $\widetilde T$ is top. iso. iff $T$ is open.
  • $\widetilde T$ is an isometric isomorphism if and only if $T$ is an coisometry.
    • $T$ is a coisometry iff $\widetilde T$ is a coisometry iff $\widetilde T$ is an isometry by (Coisometry).
Completeness of quotient normed spaces #

Suppose $X$ is a Banach space. $X _ 0 \subseteq X$ is a closed subspace. Then $X / X _ 0$ is a Banach space.

  • Suppose $(u _ n) _ {n = 1}^\infty \in X / X _ 0$ and $\sum _ n \norm{u _ n} < \infty$.
  • Pick $(x _ n) _ {n = 1}^\infty \in X$ such that $\norm{x _ n} \le \norm{u _ n} + 1/2^n$. So $(x _ n)$ is absolutely convergent.
  • Suppose $\sum _ n x _ n = x \in X$. Then $Q(x) = Q(\sum _ n x _ n) = \sum _ n u _ n$ as $Q$ is continuous.
  • So absolutely convergent series in $X / X _ 0$ are convergent.