Let G be a locally compact group, and let $M(G)$ be the space of complex Radon measures on G. I identify the function f with the measure $f(x) \rm dx$ . but How do I prove this inclusion?؟ . .
-
if $f\neq 0$ in $L^1(G)$ then $\int_G|f(x)|,dx>0$, and $\int_G|f(x)|,dx=\int_Gu(x)f(x),dx$ where $u$ is a bounded measurable function – user8268 May 12 '14 at 09:08
-
What must be proven for to prove f∈ M(G)? – user147972 May 12 '14 at 09:17
-
can you explain about relationship of equality – user147972 May 12 '14 at 09:23
-
I edited: What must be illustrate for to prove f∈ M(G)? – – user147972 May 12 '14 at 09:30
-
You have to show that $f(x) ,dx$ is Radon, see e.g. http://math.stackexchange.com/questions/215235/reconciling-several-different-definitions-of-radon-measures . This means (depending on the exact definition that you take) that $f(x) ,dx$ is inner regular on open sets and outer regular on all Borel sets (and finite on compact sets, but that is trivial, because $f(x) ,dx$ is a finite measure for $f \in L^1$). My answer below gives detailed instructions for proving this. – PhoemueX Jun 20 '14 at 20:02
2 Answers
The precise answer to this question depends on your definition of a (complex) Radon measure. I assume in the following that this means that the total variation $|\mu|$ is Radon. I also assume that you know $|f(x) \,dx| = |f(x)| \,dx$, where the absolute value on the left-hand side denotes the total variation. Thus, we can assume $f\geq 0$ in the following.
Let us first show that $f(x) \,dx$ is indeed Radon.
Let $f \in L^1(G)$, $f \geq 0$. By approximating $f$ with simple functions, you can show that for every $\varepsilon > 0$, there is some $\delta > 0$ such that
$$ \int_E f(x) \, dx < \frac{\varepsilon}{2} $$
holds for all measurable $E \subset G$ with $|E| < \delta$.
Furthermore, there is a countable sequence $E_n \subset G$ of sets of finite measure with $f \equiv 0$ on the complement of $\bigcup_n E_n$, take e.g. $E_n = \{x \in G \mid |f(x)| \geq 1/n\}$. We can assume w.l.o.g. that $E_n \subset E_{n+1}$ for all $n$.
These are the main ingredients for showing that $f(x) \,dx \in M(G)$ (i.e. is Radon).
Indeed, let $F \subset G$ be measurable. Then
$$ \int_{F\cap E_n} f(x) \,dx \rightarrow \int_F f(x) \,dx, $$
(for example by monotone convergence), i.e. $\int_{F \cap E_n} f(x) \,dx > \int_F f(x) \,dx - \frac{\varepsilon}{2}$ for $n$ large enough. By inner regularity of $dx$ on sets of finite measure, there is a compact set $K \subset F \cap E_n \subset F$ with $|(F \cap E_n) \setminus K| < \delta$, where $\delta$ is chosen as above. This yields
$$ \int_K f(x) \,dx \leq \int_F f(x) \,dx < \int_{F \cap E_n} f(x) \,dx + \frac{\varepsilon}{2} = \int_K f(x) \,dx + \int_{(F\cap E_n) \setminus K} f(x) \,dx + \frac{\varepsilon}{2} < \varepsilon. $$
This establishes inner regularity of $f(x) \,dx$.
For outer regularity, let again $F \subset G$ be measurable. As shown above, there is a compact set $K \subset F^c$ with
$$ \int_K f(x) \,dx \leq \int_{F^c} f(x) \,dx < \int_K f(x) \,dx + \varepsilon. $$
This entails that $U := K^c \supset F$ is open and
$$ \begin{eqnarray*} \int_{F}f\left(x\right)\, dx & \leq & \int_{U}f\left(x\right)\, dx\\ & = & \int_{G}f\left(x\right)\, dx-\int_{U^{c}}f\left(x\right)\, dx\\ & = & \int_{G}f\left(x\right)\, dx-\int_{K}f\left(x\right)\, dx\\ & < & \int_{G}f\left(x\right)\, dx-\left[\int_{F^{c}}f\left(x\right)\, dx-\varepsilon\right]\\ & = & \int_{F}f\left(x\right)\, dx+\varepsilon \end{eqnarray*} $$
This also proves outer regularity of $f(x) \,dx$.
Now, if you know (as used above) $|f(x) \,dx| = |f(x)|\,dx$ for the total variation, you see
$$ \Vert f(x) \,dx \Vert_{M(G)} = (|f(x)| \, dx )(G) = \int_G |f(x)| \,dx = \Vert f \Vert_1, $$
so that the (by the above) well-defined map
$$ L^1(G) \rightarrow M(G), f \mapsto f(x) \,dx $$
is isometric, and hence injective. It is clearly linear, completing the proof.

- 35,087
-
Sorry for reviving this post again, but why does for every $\epsilon>0$ exist some $\delta>0$ such that $\int_E f(x) dx<\epsilon/2$ for all measurable $E$ with $\vert E\vert<\delta$? I can't exactly envision the argument with simple functions. – Sha Vuklia Jan 04 '24 at 22:58
-
@ShaVuklia: See for instance here: https://math.stackexchange.com/questions/880246/uniform-integrability-of-a-function-in-l1 – PhoemueX Jan 05 '24 at 10:54
-
Gotta love SE to receive a reply on a 10 year old post - thank you! – Sha Vuklia Jan 05 '24 at 13:34
If one takes a look at the proof of the construction of a Haar measure $m$ for $G$, one finds that $m$ is a Radon measure ($m$ arrises from a linear functional on $C_c(G)$ by invoking the Riesz representation theorem). The following post proves that $f\mu$ where $\mu$ is a Radon measure is again a Radon measure. By Theorem C.8 in Conway's Functional Analysis, we know that $\vert f m\vert=\vert f\vert m$, and hence $\Vert fm\Vert=\Vert f\Vert_1$ (details spelled out by PhoemueX already). Therefore we can consider $L^1(G)$ as a closed subspace of $M(G)$.

- 3,960
- 4
- 19
- 37