Skip to content

Hilbert Space

We first introduce inner product in abstract space.

Inner Product

Assume U is a linear space on number field K (i.e. R or C). A mapping

(,):U×UK

is said to be a Inner Product, if it satisfies that, for all x,y,zU, for all αK,

(i) Homogeneous on the first element. (αx,y)=α(x,y).

(ii) Linearity on the first element. (x+z,y)=(x,y)+(z,y).

(iii) Conjugate inverse. (x,y)=(y,x).

(iv) Positive-definite. (x,x)0 and (x,x)=0x=θ.

And A linear space U with inner product is called Inner Product Space.

With the above definition, we have the following properties of inner product.

Properties of inner product

Assume U is an inner product space on number field K. Then x,y,zU, for all αK,

(i) Conjugate Homogeneous on the second element. (x,αy)=α(x,y).

(ii) Linearity on the second element. (x,y+z)=(x,y)+(x,z).

(iii) Zero for null vector. (x,θ)=(x,0x)=0(x,x)=0.

(iv) Normed linear space. Define x=(x,x), then it is a norm.

(v) (x,y) is a continuous function with respect to both x,y.

Easy to show that (i), (ii), (iii).

We prove for (iv) and the triangle inequation. This is based on Schwarz Inequation.

|(x,y)|2x2y2

We still make use of λK

0x+λy2=(x+λy,x+λy)=x2+λ(x,y)+λ(y,x)+|λ|2y2=x2+λ(y,x)+λ[(x,y)+λy2]

Assume (y,y)0, and choose λ=(x,y)y2, we have

0x2(x,y)y2(y,x)|(x,y)|2x2y2

With Schwarz inequation, we have

x+y2=|(x+y,x+y)|=|(x+y,y)+(x+y,x)||(x+y,y)|+|(x+y,x)|x+yy+x+yxSchwarz Inequation

and we are done.

Polarization Identity

(i) For K=R, we have

(x,y)=14(x+y2xy2).

(ii) For K=C, we have

(x,y)=14(x+y2xy2+ix+iy2ixiy2).

(i) Just expand the righthand item in inner product.

(ii) Still the same logic, but

x+y2xy2=2(x,y)+2(x,y)=4Re[(x,y)].

We notice Re[(x,iy)]=Re[i(x,y)]=Im[(x,y)], so apply y=iy in the above equation and we have

4Im[(x,y)]=4Re[(x,iy)]=x+iy2xiy2.

Hilbert Space

If a inner product space U is complete, then it is called Hilbert Space.

We have known that a inner product space must be a normed linear space, then how about the inverse? Generally speaking, a normed linear space could not be a inner product space, i.e. use the inner product to deduce a norm, unless we add the following condition.

Parallelogram Law

If a norm satisfies

x+y2+xy2=2(x2+y2)

then we could define a inner product using polarization identity.

Inversely speaking, a norm introduced by a norm satisfy the above condition, which is called Parallelogram Law.

  • Sufficient.

Just expand lefthand of the equation using inner product and we have the result.

  • Necessary.

We first show that the inner product defined by Polar equation satisfies linearity.

(x,z)+(y,z)=14(x+z2xz2+ix+iz2ixiz2)+14(y+z2yz2+iy+iz2iyiz2)=18(x+y+2z2x+y2z2+ix+y+i2z2ix+yi2z2)=12(x+y2+z2x+y2z2+ix+y2+iz2ix+y2iz2)=2(x+y2,z)=12(x+y,2z)this is not good.

Let y=θ, we have (x,z)=2(x2,z). Choose x=x+y, we have (x+y,z)=2(x+y2,z). Comparing with before and we have

(x,z)+(y,z)=(x+y,z).

So let y=yx, we have

(x,z)+(yx,z)=(y,z)(yx,z)=(y,z)(x,z)

which means "+" and "-" satisfy.

This is a little tricky, and we have to show a lemma.

For a continuous function f(α) defined on R, which satisfies

f(α1+α2)=f(α1)+f(α2),α1,α2R,

then αR, we have

f(α)=αf(1).

Let

f(α)=(αx,z).

it is continuous because

|f(α)f(α0)|=|((αα0)x,z)|subtraction

For αα0, (αα0)x0x=θ, so the lefthand of the above equation tends to 0.

And

f(α1+α2)=((α1+α2)x,z)=(α1x+α2x,z)=(α1x,z)+(α2x,z),

so we have

(αx,z)=α(x,z),αR,

which is homogeneity.

As for conjugate property and positive-definitivity, just follow the definition.

Orthogonal System

Orthogomal System

Assume U is a inner product space, x,yU. If (x,y)=0, then we call x is orthogamal to y, denoted by xy. If a subset MU, yM, xy, then we call x is orthogonal to M, denoted by xM. If NU, xN, xM, then we call N is orthogonal to M, denoted by NM.

An orthogonal complement of M is denoted by

M={xU:xM}.

Property of Orthogomal

(i) Pythagorean Theorem. For x1,,xn which are mutually orthogonal, then x:=x1++xn, we have

x2=i=1nxi2.

(ii) Dense set with Zero Vector. Assume LU is a dense subset, if xU is orthogonal to L, then x=0.

(iii) For MU, M is a closed subset.

Orthogomal Projection

We introudce orthogonal projection using the best approximation element.

Definition of best Approximation element

Assume MU, xU, yU is called the best approximation element of x in M, if

xy=infzMxz

Here we give a sufficient condition for the existence and uniqueness of best approximation element.

Theorem for existence and uniqueness of best approximation element

Assume MU, xU. If M is a closed convex set, then there exists a unique best approximation element y of x in M.

This proof leverages the convex property.

From the above theorem, A subspace is naturally convex, thus adding closedness could satisfy the above condition. So

Orthogomal decomposition and projection

Assume MU is a closed subspace, then xU, !yM,zM, such that

x=y+z.

the above equation is called orthogonal decomposition, and y is called the orthogonal projection of x in M.

Normalized Orthogomal System

Definition of Normalized Orthogomal System

System {en}n1U is called Normalized Orthogomal System, if it satisfies

(en,em)={1,n=m,0,nm.

For a element xU, define

cn=(x,en)

to be the Fourier coefficient of x. They also form a system {cn}.

Now we talk a little deeper into this one.

Explicite expression of orthogonal decomposition

Assume n is given, {e1,,en} is a finite-number normalized orthogonal system. Let M=span(e1,,en), then xU, we have the orthogonal projection of x on M

x=i=1n(x,en)en

First use Orthogomal decomposition and projection to show that

x=y+z,yspace(e1,,en).

So by property of base, y=i=1nanen. Making a inner product with ej on both sides gives

aj=(y,ej)

So we are done.

Bessel Inequation

For a normalized orthogonal system {en}U, xU with its Fourier coefficients {cn}, we have

n=1(x,en)2x2.

or {en}l2.

Use finite many elements and then let n.

Inversely speaking, given a sequence in {cn}l2, could we find a element in xU such that cn is its fourier coefficient? It turns out that we need to add another condition.

Riesz-Fischer Theorem

Assume U is a Hilbert space, {en} are its normalized orthogonal system, number sequence {cn}l2. Then there exists xU, such that for each n, we have

cn=(x,en)

and Parseval equation holds

i=1cn2=i=1(x,en)2=x2

Now we pause here. It is interesting to see that Bessel inequation holds for all inner product space, but Parseval equation does not also hold. We dig deeper into this and this is actually related to the number of {en}. For a Hilbert space, its number is enough to let "=" hold.

Completeness

Definition for completeness of orthogonal system

Assume U is a inner product space. An normalized orthogonal system is said to be complete, if for all xU, Parseval equation holds; it is said to be total, if it is dense in U.

Relationship

Assume U is a Hilbert space. {en} is a normalized orthogonal system. Then we have the following equivalence.

(i) {en} is complete.

(ii) xU, series i=1n(x,ei)ei converges to x, i.e.

x=n=1(x,en)en.

(iii) x,yU,

(x,y)=n=1(x,en)(y,en).

(iv) {en} is total.

(i) and (ii) are equivalent because

xn=1N(x,en)en2=x2n=1N|(x,en)|2

Using equation from (ii) we only get infinite series. We start from finite-numbers. Define

xn=i=1n(x,ei)ei,yn=j=1n(y,ej)ej,

then

(xn,yn)=(i=1n(x,ei)ei,j=1n(y,ej)ej)=i=1n(x,en)(y,en).

Let n, since by hölder inequation

(i=1|(x,en)(y,en)|)2(i=1|(x,en)|2)(i=1|(y,en)|2)

by (ii) the above converges. By continuity of inner product, we are done.

Assume xU, (x,en)=0. So yU

(x,y)=n=1(x,en)(y,en)=0.

Choose y=x and (x,x)=0 implies x=θ.

xU, by Bessel ineuqation, n=1|(x,en)|2<. For number sequence {(x,en)}l2, by Riesz-Fischer Theorem, yU, such that (x,en)=(y,en) for each n. So (xy,en)=0 for each n, by (iv) assumption, we have xy=θ, i.e. x=y and x satisfies

x2=y2=i=1|(x,en)|2.