MathLabs
TheoremProved

Riesz representation theorem (Hilbert space case)

Statement

Let HH be a Hilbert space and φ\varphi a bounded (continuous) linear functional on HH. Then there is a unique y∈Hy\in H such that φ(x)=⟨x,y⟩\varphi(x)=\langle x,y\rangle for every x∈Hx\in H, and moreover ∥φ∥=∥y∥\|\varphi\|=\|y\|.

Why is it true?

It says every way of assigning a number to each vector linearly and continuously is secretly just "taking the inner product with some fixed vector" — abstract functionals are no more general than the most concrete ones you already know.

Proof sketch

If φ=0\varphi=0, take y=0y=0 and we are done. Otherwise let N=ker⁡φ={x∈H:φ(x)=0}N=\ker\varphi=\{x\in H:\varphi(x)=0\}; since φ\varphi is continuous and linear, NN is a closed proper subspace of HH. Because HH is a Hilbert space, the projection theorem gives an orthogonal decomposition H=N⊕N⊥H=N\oplus N^{\perp}, and since NN is a proper subspace, N⊥N^{\perp} contains some z≠0z\ne0.

For any x∈Hx\in H, consider the vector u=φ(x)z−φ(z)xu=\varphi(x)z-\varphi(z)x. Applying φ\varphi gives φ(u)=φ(x)φ(z)−φ(z)φ(x)=0\varphi(u)=\varphi(x)\varphi(z)-\varphi(z)\varphi(x)=0, so u∈Nu\in N. Since z∈N⊥z\in N^{\perp}, we get ⟨u,z⟩=0\langle u,z\rangle=0, i.e. φ(x)⟨z,z⟩−φ(z)⟨x,z⟩=0\varphi(x)\langle z,z\rangle-\varphi(z)\langle x,z\rangle=0. Solving for φ(x)\varphi(x) gives φ(x)=φ(z)∥z∥2⟨x,z⟩=⟨x,φ(z)‾∥z∥2z⟩\varphi(x)=\dfrac{\varphi(z)}{\|z\|^2}\langle x,z\rangle=\left\langle x,\dfrac{\overline{\varphi(z)}}{\|z\|^2}z\right\rangle, so y=φ(z)‾∥z∥2zy=\dfrac{\overline{\varphi(z)}}{\|z\|^2}z represents φ\varphi.

For uniqueness, if ⟨x,y1⟩=⟨x,y2⟩\langle x,y_1\rangle=\langle x,y_2\rangle for all xx, take x=y1−y2x=y_1-y_2 to get ∥y1−y2∥2=0\|y_1-y_2\|^2=0, so y1=y2y_1=y_2. For the norm identity, Cauchy–Schwarz gives ∣φ(x)∣=∣⟨x,y⟩∣≤∥y∥∥x∥|\varphi(x)|=|\langle x,y\rangle|\le\|y\|\|x\|, so ∥φ∥≤∥y∥\|\varphi\|\le\|y\|; and plugging in x=yx=y gives φ(y)=∥y∥2\varphi(y)=\|y\|^2, so ∥φ∥≥∣φ(y)∣/∥y∥=∥y∥\|\varphi\|\ge|\varphi(y)|/\|y\|=\|y\|. Together, ∥φ∥=∥y∥\|\varphi\|=\|y\|.

Topics that use this theorem

Step-by-step proofs

No step-by-step proof yet for this theorem.

References

  1. Walter Rudin (1991). Functional Analysis
  2. John B. Conway (2007). A Course in Functional Analysis
  3. Assaf Naor (2012). An introduction to the Ribe program · arXiv:1205.5993