泛函分析4——希尔伯特空间

它只是为了凑个标题

3 Hilbert Spaces

希尔伯特空间是有限维欧拉空间的推广。

3.1Introduction

Definition

Let $X$ be a linear space over a field $\mathbb{F}$. An inner product on $X$ is a scalar-valued function $⟨\cdot ,\cdot ⟩:X\times X\to \mathbb{F}$ such that for all $x,y,z\in X$ and for all $\alpha ,\beta \in \mathbb{F},$ we have

IP1. $⟨x,x⟩\ge 0$

IP2. $⟨x,x⟩=0⟺x=0$

IP3. $⟨x,y⟩=\overline{⟨y,x⟩}$ (The bar denotes complex conjugation.);

IP4. $⟨\alpha x,y⟩=\alpha ⟨x,y⟩$

IP5. $⟨x+y,z⟩=⟨x,z⟩+⟨y,z⟩$

An inner product space $(X,⟨\cdot ,\cdot ⟩)$ is a linear space $X$ together with an inner $⟨\cdot ,\cdot ⟩$ product defined on it. $An$ inner product space is also called pre-Hilbert space.

Example

[1] Fix a positive integer $n.$ Let $X={\mathbb{F}}^n.$ For $x=\left(x_1,x_2,\dots ,x_n\right)$ and $y=\left(y_1,y_2,\dots ,y_n\right)$ in $X$, define
$$⟨x,y⟩=\sum _{i=1}^n{x}_i\overline{y_i}$$
The space ${\mathbb{R}}^n$ (resp. ${\mathbb{C}}^n$ ) with this inner product is called the Euclidean $n$ -space (resp. unitary $n$ -space) and will be denoted by ${\ell }_2(n)$.

Theorem

[1] (Cauchy-Bunyakowsky-Schwarz Inequality). Let $(X,⟨\cdot ,\cdot ⟩)$ be an inner product space over a field $\mathbb{F}$. Then for all $x,y\in X$,
$$|⟨x,y⟩|\le \sqrt{⟨x,x⟩}\sqrt{⟨y,y⟩}$$
Moreover, given any $x,y\in X,$ the equality
$$|⟨x,y⟩|=\sqrt{⟨x,x⟩}\sqrt{⟨y,y⟩}$$
holds if and only if $x$ and $y$ are linearly dependent.

[2] Let $(X,⟨\cdot ,\cdot ⟩)$ be an inner product space over a field $\mathbb{F}$. For each $x\in X,$ define
$$\Vert x\Vert :=\sqrt{⟨x,x⟩}$$
Then $\Vert \cdot \Vert$ defines a norm on $X.$ That is, $(X,\Vert \cdot \Vert )$ is a normed linear space over $\mathbb{F}$.

任何内积空间都可以通过上面的方式定义一个范数来形成赋范空间，该范数又称为内积范数。

以下例子中的范数都是内积范数inner product norm

[3] Theorem (Polarization Identity). Let $(X,⟨\cdot ,\cdot ⟩)$ be an inner product space over a field $\mathbb{F}$. Then for all $x,y\in X$,
$$⟨x,y⟩=\frac{\Vert x+y{\Vert }^2}{4}-\frac{\Vert x-y{\Vert }^2}{4}\text{ if }\mathbb{F}=\mathbb{R},\text{ and }$$

$$⟨x,y⟩=\frac{\Vert x+y{\Vert }^2}{4}-\frac{\Vert x-y{\Vert }^2}{4}+i\left(\frac{\Vert x+yi{\Vert }^2}{4}-\frac{\Vert x-yi{\Vert }^2}{4}\right)\text{ if }\mathbb{F}=\mathbb{C}$$

[4] Theorem (Parallelogram Identity). Let $(X,⟨\cdot ,\cdot ⟩)$ be an inner product space over a field $\mathbb{F}$. Then for all $x,y\in X$,
$$\Vert x-y{\Vert }^2+\Vert x+y{\Vert }^2=2\Vert x{\Vert }^2+2\Vert y{\Vert }^2$$

平行四边形恒等式表示，对角线平方的和等于四条边平方的和。

[5] A normed linear space $X$ over a field $\mathbb{F}$ is an inner product space if and only if the Parallelogram Identity
$$\Vert x-y{\Vert }^2+\Vert x+y{\Vert }^2=2\Vert x{\Vert }^2+2\Vert y{\Vert }^2$$
holds for all $x,y\in X$.

对于一个赋范线性空间，如果其中的所有元素都满足平行四边形恒等式，那么该赋范线性空间就是一个内积空间。

Corollary

Let $(X,\Vert \cdot \Vert )$ be a normed linear space over a field $\mathbb{F}$. If every two-dimensional linear subspace of $X$ is an inner product space over $\mathbb{F},$ then $X$ is an inner product space.

范数函数对应关系是 $X\to \mathbb{F}$, 是X中的一个元素对应一个复数

内积函数对应关系是 $X\times X\to \mathbb{F}$,是X中的一个状态对对应一个复数

3.2 Hilbert Spaces Definition

Let $(X,⟨\cdot ,\cdot ⟩)$ be an inner product space. If $X$ is complete with respect to the norm induced by the inner product $⟨\cdot ,\cdot ⟩,$ then we say that $X$ is a Hilbert space.

Hilbert space = Banach space : an complete inner product space with inner product norm

3.3 Orthogonality

Definition

Two elements $x$ and $y$ in an inner product space $(X,⟨\cdot ,\cdot ⟩)$ are said to be orthogonal, denoted by $x\perp y$, if
$$⟨x,y⟩=0$$
The set $M\subset X$ is called orthogonal if it consists of non-zero pairwise orthogonal elements. If $M$ is a subset of $X$ such that $⟨x,m⟩=0$ for all $m\in M,$ then we say that $x$ is orthogonal to $M$ and write $x\perp M.$ We shall denote by
M ⊥ = { x ∈ X : ⟨ x , m ⟩ = 0 ∀ m ∈ M } M^{\perp}=\{x \in X:\langle x, m\rangle=0 \forall m \in M\} M⊥={x∈X:⟨x,m⟩=0∀m∈M}
the set of all elements in $X$ that are orthogonal to $M.$ The set $M^{\perp }$ is called the orthogonal complement of $M$

M的正交补：X中所有与M正交的元素的集合。

Proposition

Let $M$ and $N$ be subsets of an inner product space $(X,⟨\cdot ,\cdot ⟩)$. Then
[1] { 0 } ⊥ = X \{0\}^{\perp}=X {0}⊥=X and X ⊥ = { 0 } X^{\perp}=\{0\} X⊥={0}
[2] $M^{\perp }$ is a closed linear subspace of $X$;
$[3]M\subset {\left(M^{\perp }\right)}^{\perp }=M^{\perp \perp }$
[4] If $M$ is a linear subspace, then M ∩ M ⊥ = { 0 } M \cap M^{\perp}=\{0\} M∩M⊥={0};
[5] If $M\subset N,$ then $N^{\perp }\subset M^{\perp }$
[6] $M^{\perp }=(\mathrm{lin}M{)}^{\perp }=(\overline{\mathrm{lin}}M{)}^{\perp }$

Theorem

[1] (Pythagoras). Let $(X,⟨\cdot ,\cdot ⟩)$ be an inner product space over a field $\mathbb{F}$ and let $x,y\in X$
[a] If $\mathbb{F}=\mathbb{R},$ then $x\perp y$ if and only if
$$\Vert x+y{\Vert }^2=\Vert x{\Vert }^2+\Vert y{\Vert }^2$$
​ [b] If $\mathbb{F}=\mathbb{C},$ then $x\perp y$ if and only if
$$\Vert x+y{\Vert }^2=\Vert x{\Vert }^2+\Vert y{\Vert }^2\text{ and }\Vert x+iy{\Vert }^2=\Vert x{\Vert }^2+\Vert y{\Vert }^2.$$

Corollary

If M = { x 1 , x 2 , … , x n } M=\left\{x_{1}, x_{2}, \ldots, x_{n}\right\} M={x1​,x2​,…,xn​} is an orthogonal set in an inner product space $(X,⟨\cdot ,\cdot ⟩)$ then
$${\Vert \sum _{i=1}^n{x}_i\Vert }^2=\sum _{i=1}^n{\Vert x_i\Vert }^2$$

3.4 Best Approximation in Hilbert Spaces

Definition

[1] Let $K$ be a closed subset of an inner product space $(X,⟨\cdot ,\cdot ⟩).$ For a given $x\in X\backslash K,$ a best approximation or nearest point to $x$ from $K$ is any element $y_0\in K$ such that
$$\Vert x-y_0\Vert \le \Vert x-y\Vert \text{ for all }y\in K$$
Equivalently, $y_0\in K$ is a best approximation to $x$ from $K$ if
$$\Vert x-y_0\Vert =\underset{y\in K}{\inf }\Vert x-y\Vert =d(x,K)$$

nearest point：顾名思义，是K中距离K的补集中某点x的最近的一些点。最佳近似点来自于K集。

[2] The (possibly empty) set of all best approximations to $x$ from $K$ is denoted by $P_K(x).$ That is,
P K ( x ) = { y ∈ K : ∥ x − y ∥ = d ( x , K ) } P_{K}(x)=\{y \in K:\|x-y\|=d(x, K)\} PK​(x)={y∈K:∥x−y∥=d(x,K)}
The (generally set-valued) map $P_K$ which associates each $x$ in $X$ with its best approximations in $K$ is called the metric projection or the nearest point map. The set $K$ is called

$P_K$: K中离x最近的点的映射。

[a] proximinal if each $x\in X$ has a best approximation in $K$; i.e., $P_K(x)\ne \varnothing$ for each $x\in X$;

存在最近点，什么情况下会不存在呢？K是开集的情况下？

[b] Chebyshev if each $x\in X$ has a unique best approximation in $K$; i.e., the set $P_K(x)$ consists of a single point.

只存在一个最近点。

The following important result asserts that if $K$ is a complete convex subset of an inner product space $(X,⟨\cdot ,\cdot ⟩),$ then each $x\in X$ has one and only one element of best approximation in $K$.

Theorem

[1] Every nonempty complete convex subset $K$ of an inner product space $(X,⟨\cdot ,\cdot ⟩)$ is a Chebyshev set.

[2] Let $K$ be a nonempty closed convex subset of a Hilbert space $(\mathcal{H},⟨\cdot ,\cdot ⟩),x\in \mathcal{H}\backslash K$ and $y_0\in K.$ Then $y_0$ is the best approximation to $x$ from $K$ if and only if
$${\Re }_{\mathrm{e}}⟨x-y_0,y-y_0⟩\le 0\text{ for all }y\in K$$
[3] (Projection Theorem). Let $\mathcal{H}$ be a Hilbert space, $M$ a closed subspace of $\mathcal{H}.$ Then
$[a]\mathcal{H}=M\oplus M^{\perp }.$ That is, each $x\in \mathcal{H}$ can be uniquely decomposed in the form
$$x=y+z\text{ with }y\in M\text{ and }z\in M^{\perp }$$
​ $[b]M=M^{\perp \perp }$

Corollary

[1] Every nonempty closed convex subset of a Hilbert space is Chebyshev.

[2] (Characterization of Best Approximations from closed subspaces). Let $M$ be a closed subspace of a Hilbert space $\mathcal{H}$ and let $x\in \mathcal{H}\backslash M$. Then an element $y_0\in M$ is the best approximation to $x$ from $M$ if and only if $⟨x-y_0,y⟩=0$ for all $y\in M$ (i.e., $x-y_0\in M^{\perp })$.

so map $P_M:x\to P_M(x)$ is also called the orthogonal projection of $\mathcal{H}$ onto $M$

[3] If $M$ is a closed subspace of a Hilbert space $\mathcal{H},$ and if $M\ne \mathcal{H},$ then there exists z ∈ H \ { 0 } z \in \mathcal{H} \backslash\{0\} z∈H\{0} such that $z\perp M$
Proof. Let $x\in \mathcal{H}\backslash M$. Then by the Projection Theorem,
$$x=y+z,\text{ where }y\in M\text{ and }z\in M^{\perp }$$
Hence $z\ne 0$ and $z\perp M$.

Proposition

Let $S$ be a nonempty subset of a Hilbert space $\mathcal{H}$. Then
$[1]S^{\perp \perp }=\overline{\mathrm{lin}}S$
[ 2 ] S ⊥ = { 0 } [2] S^{\perp}=\{0\} [2]S⊥={0} if and only if $\overline{\mathrm{lin}}S=\mathcal{H}$

3.5 Orthonormal Sets and Orthonormal Bases

Definition

[1] Let $(X,⟨\cdot ,\cdot ⟩)$ be an inner product space over $\mathbb{F}.A$ set S = { x α : α ∈ Λ } S=\left\{x_{\alpha}: \alpha \in \Lambda\right\} S={xα​:α∈Λ} of elements of $X$ is called an orthonormal set if
(a) $⟨x_{\alpha },x_{\beta }⟩=0$ for all $\alpha \ne \beta$ (i.e., $S$ is an orthogonal set), and
(b) $\Vert x_{\alpha }\Vert =1$ for all $\alpha \in \Lambda$.
If S = { x α : α ∈ Λ } S=\left\{x_{\alpha}: \alpha \in \Lambda\right\} S={xα​:α∈Λ} is an orthonormal set and $x\in X,$ then the numbers $⟨x,x_{\alpha }⟩$ are called the Fourier coefficients of $x$ with respect to $S$ and the formal series ${\sum }_{\alpha \in \Lambda }⟨x,x_{\alpha }⟩x_{\alpha }$ the Fourier series of $x$.

$\Lambda$表示S是X的规范正交系

[2] An orthonormal set $S$ in an inner product space $(X,⟨\cdot ,\cdot ⟩)$ is said to be complete in $X$ if $S\subset T$ and $T$ is an orthonormal set in $X,$ then $S=T$.

Simply put, a complete orthonormal set $S$ in an inner product space is an orthonormal set that is not properly contained in any other orthonormal set in $X;$ in other words, $S$ is complete if it is a maximal orthonormal set in $X$.

完备正交集是X中最大的正交集。

It is easy exercise to show that a set $S$ is complete in an inner product $(X,⟨\cdot ,\cdot ⟩)$ if and only if S ⊥ = { 0 } S^{\perp}=\{0\} S⊥={0}.

[3] Let $(X,⟨\cdot ,\cdot ⟩)$ be an inner product space over $\mathbb{F}$. An orthonormal set { x n } \left\{x_{n}\right\} {xn​} is called an orthonormal basis for $X$ if for each $x\in X$
$$x=\sum _{k=1}^{\infty }⟨x,x_k⟩x_k$$

在无限维的空间中定义的。

That is, the sequence of partial sums $\left(s_n\right),$ where $s_n={\sum }_{k=1}^n⟨x,x_k⟩x_k,$ converges to $x$.

[4] isomorphic(同构)

Two linear spaces $X$ and $Y$ over the same field $\mathbb{F}$ are said to be isomorphic if there is a one-to-one $\mathrm{map}T$ from $X$ onto $Y$ such that for all $x_1,x_2\in X$ and all $\alpha ,\beta \in \mathbb{F},$
$$T\left(\alpha x_1+\beta x_2\right)=\alpha T\left(x_1\right)+\beta T\left(x_2\right)$$

the result of that T is applied to $x\in X$ is $y\in Y$. 同构的用途是将一些领域的定理能够迁移到另一个同构的领域。

[5] isometry(等距)

Let $(X,\Vert \cdot \Vert )$ and $(Y,\Vert \cdot \Vert )$ be two normed linear spaces and $T:X\to Y$. Then $T$ is called an isometry if
$$\Vert Tx\Vert =\Vert x\Vert \text{ for all }x\in X.$$
Simply put, an isometry is a map that preserves lengths.

等距映射，其中左边的范数是Y中的范数，右边的范数是X中的范数。为了简洁，都写成一样的。

Normed linear spaces that are isometrically isomorphic are essentially identical

等距同构的赋范线性空间基本相同。

Theorem

[1] An orthonormal set $S$ in a separable inner product space $(X,⟨\cdot ,\cdot ⟩)$ is at most countable.

可分离的内积空间的正交集合最多。

Let $(X,⟨\cdot ,\cdot ⟩)$ be a separable inner product space over $\mathbb{F}$.
[2] (Best Fit). If { x 1 , x 2 , … , x n } \left\{x_{1}, x_{2}, \ldots, x_{n}\right\} {x1​,x2​,…,xn​} is a finite orthonormal set in $X$ and M = lin ⁡ { x 1 , x 2 , … , x n } , M=\operatorname{lin}\left\{x_{1}, x_{2}, \ldots, x_{n}\right\}, M=lin{x1​,x2​,…,xn​}, then for each $x\in X$ there exists $y_0\in M$ such that
$$
\begin{array}{l}
\qquad\left|x-y_{0}\right|=d(x, M) \

\end{array}
$$
$\text{ In fact, }{y}_0={\sum }_{k=1}^n⟨x,x_k⟩x_k$

[3] (Bessel’s Inequality). Let ${\left(x_n\right)}_{n=1}^{\infty }$ be an orthonormal sequence in $X.$ Then for any $x\in X$,
$$\sum _{k=1}^{\infty }{\left|⟨x,x_k⟩\right|}^2\le \Vert x{\Vert }^2$$
In particular, $⟨x,x_k⟩\to 0$ as $k\to \infty$.

[4] (Riesz-Fischer Theorem). Let ${\left(x_n\right)}_1^{\infty }$ be an orthonormal sequence in a separable Hilbert space $\mathcal{H}$ and let ${\left(c_n\right)}_1^{\infty }$ be a sequence of scalars. Then the series ${\sum }_{k=1}^{\infty }{c}_k{x}_k$ converges in $\mathcal{H}$ if and only if $c={\left(c_n\right)}_1^{\infty }\in {\ell }_2.$ In this case,
$$\Vert \sum _{k=1}^{\infty }{c}_k{x}_k\Vert ={\left(\sum _{k=1}^{\infty }{\left|c_k\right|}^2\right)}^{\frac{1}{2}}$$

[5] Let $\mathcal{H}$ be a separable infinite-dimensional Hilbert space and assume that S = { x n } S=\left\{x_{n}\right\} S={xn​} is an orthonormal set in$\mathcal{H}$. Then the following statements are equivalent:
[a] $S$ is complete in $\mathcal{H};$ i.e., S ⊥ = { 0 } S^{\perp}=\{0\} S⊥={0}.
[b] $\overline{\mathrm{lin}}S=\mathcal{H};$ i.e., the linear span of $S$ is norm-dense in $\mathcal{H}$.
[c] (Fourier Series Expansion.) For any $x\in \mathcal{H}$, we have
$$x=\sum _{i=1}^{\infty }⟨x,x_i⟩x_i$$
That is, $S$ is an orthonormal basis for $\mathcal{H}$.
[d] (Parseval’s Identity.) For all $x,y\in \mathcal{H}$,
$$⟨x,y⟩=\sum _{k=1}^{\infty }⟨x,x_k⟩\overline{⟨y,x_k⟩}$$
​ [e] For any $x\in \mathcal{H}$,
$$\Vert x{\Vert }^2=\sum _{k=1}^{\infty }{\left|⟨x,x_k⟩\right|}^2$$
[6] (Gram-Schmidt Orthonormalisation Procedure). If { x k } 1 ∞ \left\{x_{k}\right\}_{1}^{\infty} {xk​}1∞​ is a linearly independent set in an inner product space $(X,⟨\cdot ,\cdot ⟩)$ then there exists an orthonormal set { e k } 1 ∞ \left\{e_{k}\right\}_{1}^{\infty} {ek​}1∞​ in $X$ such that
lin ⁡ { x 1 , x 2 , … , x n } = lin ⁡ { e 1 , e 2 , … , e n }  for all  \operatorname{lin}\left\{x_{1}, x_{2}, \ldots, x_{n}\right\}=\operatorname{lin}\left\{e_{1}, e_{2}, \ldots, e_{n}\right\} \quad \text { for all } lin{x1​,x2​,…,xn​}=lin{e1​,e2​,…,en​} for all 
[7] Every separable Hilbert space $\mathcal{H}$ has a countable orthonormal basis.

[8] Every separable infinite-dimensional Hilbert space $\mathcal{H}$ is isometrically isomorphic to ${\ell }_2$.

每个可分的无限维希尔伯特空间和赋范空间${\ell }_2$都是等距同构的。

Lemma

Let M = lin ⁡ { x n } M=\operatorname{lin}\left\{x_{n}\right\} M=lin{xn​} be a linear subspace of $X.$ Then there exists a subsequence { x n k } \left\{x_{n_{k}}\right\} {xnk​​} of { x n } \left\{x_{n}\right\} {xn​} which has the following properties:
(i) lin ⁡ { x n k } = M \operatorname{lin}\left\{x_{n_{k}}\right\}=M lin{xnk​​}=M
(ii) { x n k } \left\{x_{n_{k}}\right\} {xnk​​} is linearly independent.