2
$\begingroup$

let $\Omega$ be (locally) compact Hausdorff space.

I Denote by $C(\Omega)$ normed space of all continuous (compactly supported) functions with sup norm $$\| f \| = \sup_{x \in \Omega} f(x) $$ amd by $\mathfrak{M}(\Omega)$ space of finite (or regular) signed measures .

When from Riesz–Markov–Kakutani representation theorem it follows that where is a linear isomorphism $I : \mathfrak{M}(\Omega) \to C^*(\Omega)$ defined by the rule $$ I(\mu)(f) = \int_\Omega f \, \mathrm{d}\mu. $$ I know that in case $\Omega \subset \mathbb{R}$ I know that (local) bounded variation norm on $\mathfrak{M}(\Omega)$ will make $I$ into isometry of normed spaces. The obvious norm for general case would be push-forward $$ \| \mu \| = \| I(\mu) \| $$

But aren't there a more meaningful norm on $\mathfrak{M}(\Omega)$ with the same property?
Why total variation $\| \mu \| = |\mu |(\Omega)$ norm won't work?

Secondly it's well known that signed measure can be represented as $$ \mu = \mu^+ - \mu^-, $$ where $\mu^+$ and $\mu^-$ are normal measures. When, as $I$ maps measures into positive functionals by her structure, every functional $F$ can be written for some signed measure $\mu$ as $$ F = I(\mu) = I(\mu^+) - I(\mu^-) = F^+ - F^-, $$

where $F^+ $ and $F^{-}$ are positive functionals.

What is the way to define $(F^+,F^-)$ just in terms of $F$ without any reference to $I$?
Basicly, I want to be able to migrate this definition to other functional spaces which are not exactly $C(\Omega).$

It would be nice to know some direct references concerning topic of positive-negative decomposition of functionals too. As it feels like quiet a standard topic. (I Just want to know what part of general functional analysis texts speaks about it is generally presented, because I haven't seen it yet).

Thanks for your time and attention.

--Addendum

The construction of $F^+$ for abstract normed space is obvious if positive cone $V_+$ spans $V$. As positive cone $V_+$ spans $V$ there exists a basis $\{e_a | a \in A \}$ of exclusively positive vectors for some (uncountable) set $A$ (This needs axiom of choice). Then every vector $v$ can be written as $$ v = \sum_{a \in A}v_ae_a $$ with only finite number of non-zero $v_a$ .Then put $$ F^+(v) = \sum_{a \in A} v_a (f(e_a))^+ $$ and $$F^-(v) = \sum_{a \in A} v_a (f(e_a))^- $$ From properties of Cones it follows that for every $v \in V$ coordinates $v_a > 0$ so this functionals are indeed positive. However, It seems that this functions are bounded functionals only if norm in $V$ is invariant to a change of signs of coordinates.

This properties hold for any finite-dimensional space $\mathbb{R}^n$ trivialy and for every inner product space which admits Schauder basis of positive elements. So if $H$ is such inner product space and where exists an embedding $E:V \to H$ which preserves positivity of an element the proposed construction will work ( extend pull-back of $F E^{-1}$ from $\mathrm{im} E$ to $H$ by Hahn-Banach theorem as $G$ . Then define $F^+ = G^+ E$ and $F^{-} = G^{-} E$ ).

For $V = C(\Omega)$ role of $H$ can be played by Hilbert space $L_2(\Omega,\mu)$ where $\mu$ is a finite measure which is non-zero on every open set (Existence of such measure is another question).

This construction is bad for two reasons:

Firstly, it depends on "coordinates".

Secondly, It still relies on some external space when $V$ is not simple enough.

  • 0
    Conway's *A Course in Functional Analysis* has a couple of sections on positivity in an ordered vector space. Then towards the end, positive linear functionals on $C^*$-algebras plays a large role in the representation theory of the algebras.2017-01-17
  • 0
    Aweygan , Thanks. I will check it out.2017-01-17
  • 0
    Just "it requires the axiom of choice" does not merit the tag. The tag [axiom-of-choice] is meant for questions explicitly discussing the role of the axiom in definitions and proofs.2017-01-18

1 Answers 1

0

I found correct decomposition in this real analysis post by Terence Tao in Lemma 9. Jordan decomposition of functions.

It seems that this construction will work in every ordered normed vector space $V$ such that $V = V_+ + V_+$ which holds if $V_+$ is cofinal.