Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Theorem: Let $F$ be a continuous linear functional on the Hilbert space $H$, then $\exists !$ (exists one and only one) $y \in H$ such that $F(x) = (x,y)$ for $x\in H$.


  1. Uniqueness: $$F(x)=(x,y)=(x,y') \Rightarrow (x,y-y')=0 \space \forall x \in H \Rightarrow ||y-y'||^2=0$$

  2. Existence: if for any $x\in H, \space F(x) =(x,y)= 0 $ we'll set $y=0$. Let that $F \neq 0. $ and define a closed subspace $H_0 = \{x \in H: F(x)=0\}$. We know that $H \neq H_0$ because $H_0 = F^{-1}(\{0\})$.

Q1: "$H_0$ is a subspace by the linearity of $F$, closed by continuity of $F$." What does this mean? I understand what's a subspace, but don't understand what does it have to do with linearity and continuity.

$\exists z\in H$ such that $||z||=1$. For $u \in H_0, \space F(u)=(u,z) =0$.

Q2: Does this imply that $u \perp z \space \forall u \in H_0, \forall z \in H$?

Let's define $u=F(x)z - F(z)x, \space x \in H$

$$F(u)=(F(x)z - F(z)x,z)=F(x)F(z)-F(z)F(x)=0=F(x)(z,z)-F(z)(x,z)$$ Since we know that $||z|| = \sqrt{(z,z)}=1$, then $$0=F(x)(z,z)-F(z)(x,z) \Rightarrow F(x) =F(z)(x,z) = (x,\bar{F(z)z)} $$ By defining $y= \bar{F(z)}z$ we've completed the proof.

By the way, the bar on the top denotes the complex conjugate, as is common in Hilbert spaces. i.e. $\alpha (x,y)= (\alpha x,y)=(x,\bar{\alpha} y) $

Corollary: If $H=L^2(\mu)$ and $F$ is a continuous linear functional on $L^2(\mu)$, then $\exists ! \space g \in L^2(\mu)$ such that $$F(f)=\int_X f \bar{g} d\mu = (f,g)$$

In the context of the Radon-Nikodym theorem, $g$ is the R-N derivative.

share|cite|improve this question
Since $H_0\subset H$ how can it be that $u\perp z$ for all $u\in H_0$ and all $z\in H$? This however implies that $u\perp z$ for any $u\in H_0$ and a fixed $z$ which have been chosen on the previous step. On the other hand, $a\perp b$ if and only if $(a,b) = 0$ by definition. – Ilya May 29 '13 at 8:36
so should it be $\forall z \in H_0^{\perp} \subset H$? – shimee May 29 '13 at 8:40
sure, if you recall the definition of $H^\perp_0$. – Ilya May 29 '13 at 8:41
W.r.t. your new question: I understand what's a subspace - so what is it in your understanding? Set $A\subset H$ is a subspace of a Hilbert $H$ space if and only if $A$ is closed under linear operations (that is, it is a linear space by itself) and it is a closed subset of $H$. " – Ilya May 29 '13 at 9:05
Also, would you consider asking all of your question in the first version of OP, rather than adding them more and more? – Ilya May 29 '13 at 9:06
up vote 0 down vote accepted

Question 1: You defined $H_0 = F^{-1}(\{0\})$, where $F \in H^*$ (the continuous linear functionals on $H$). In order for $H_0$ to be a subspace, you have to check that $\lambda x + \mu y \in H_0$ for all $x,y \in H_0$, $\lambda, \mu \in \mathbb K$, right? But $$ \lambda x + \mu y \in H_0 \iff F(\lambda x + \mu y) = 0 $$ As $F$ is linear (and $x,y \in H_0$), we have $$ F(\lambda x + \mu y) = \lambda F(x) + \mu F(y) = 0 $$ So $H_0$ being a subspace follows from the linearity of $F$. In order for $H_0$ to be closed, we must show that for all sequences $(x_n)$ in $H_0$ converging to some $x \in H$ we must have $x\in H_0$. But, as $F$ is continuous, we have that $x_n \to x$, implies $F(x_n) \to F(x)$. As $F(x_n) = 0$ for every $n$, we get $F(x) = 0$, so $x \in H_0$. So the closedness of $H_0$ follows from the continuity of $F$.

Question 2: No, it doesn't. As $H_0$ is a proper closed subspace of $H$ ($H_0 \subsetneq H$ as $F \ne 0$), there is some $z \in H$, with $z \mathrel\bot H_0$ and $\|z\| = 1$. This means that for this $z$, not for all we have $$ (z, u) = 0, \quad \text{all } u \in H_0. $$

share|cite|improve this answer

1) As with any other (linear) vector space, the kernel of a linear transformation is always a vector subspace of the definition domain vector space, and in this case we have that $\;H_0:=\ker F\;$ , with $\,F:H\to\Bbb F\;,\;\;\Bbb F=\Bbb R\;\,\;\Bbb C\,$ . This is also denoted by $\,F\in H^*:=$ the dual space of $\,H\,$

The conditions about closedness and being a subspace of Hilbert space then follow from conditions other than the linear ones (i.e., continuity of a linear functional), which hopefully are already clear.

(2) About your second question: the answer's no, since the only elements that is orthogonal to all the elements in the space is the zero vector.

What is written there follows from the fact that $\,H_0\neq H\,$ (in fact, since $\,F\,$ is a non-trivial functional we have that $\;H_0\;$ is a maximal proper subspace of $\,H\,$ . This much is true also in any general vector space ,whether dinite or infinite dimensional over its definition field) . and then $\,H_0^\perp\neq0\;$ , and from here that there exists such $\,z\in H\,$ as implied just before your second question

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.