1
$\begingroup$

Let $(\mathbb{R}^d,\mathcal{B}(\mathbb{R}^d),\mu)$ be a $\sigma$-finite Borel measure on $d$-dimensional Euclidean space. Can one always construct a sequence of finite equivalent measures $\left\{\mu_n\right\}_{n \in \mathbb{N}}$ on $(\mathbb{R}^d,\mathcal{B}(\mathbb{R}^d))$ such that $$ \lim_{n \rightarrow \infty} \frac{d\mu_n}{d\mu} \uparrow 1. $$

Motivation: This is true for both the Lebesgue measure and the counting measure on $\mathbb{R}^d$ by taking $\frac{d\mu_n}{d\mu}(x)\triangleq \begin{cases} e^{-q_n\min\{\|x\|,\frac1{\|x\|}\}}&:\, x\neq 0\\ 1 &: \, x =0 \end{cases} $, where $q_n$ runs over $\mathbb{Q}\cap(0,\infty)$.

$\endgroup$

2 Answers 2

3
$\begingroup$

Since $\mu$ is $\sigma$-finite there is a $\mu$-integrable function $f$ with $0<f\le 1$. For an increasing sequence $A_n$ with $\bigcup_{n\in\mathbb N} A_n=\mathbb R^d$ define densities $f_n=I_{A_n}+fI_{A_n^c}$. Then $\mu_n=f_n\cdot \mu$ are equivalent finite measures whose densities converge to $1$.

$\endgroup$
4
  • $\begingroup$ But how to show that such a function exists (ie the bounded away from 0 part)? $\endgroup$
    – ABIM
    Jan 10, 2020 at 10:07
  • 1
    $\begingroup$ If $A_n$ are as in the answer set $B_1=A_1$, $B_n=A_n\setminus A_{n-1}$, and $f=\sum\limits_{n=1}^\infty c_n 1_{B_n}$ with suitable $c_n>0$, so that $\int f d \mu= \sum\limits_{n=1}^\infty c_n \mu(B_n)<\infty$. $\endgroup$ Jan 10, 2020 at 10:17
  • 1
    $\begingroup$ Note that this has nothing to do with the topology of $\mathbb R^d$. It holds for every $\sigma$-finite measure space. $\endgroup$ Jan 10, 2020 at 10:18
  • $\begingroup$ Yes I had noticed that; thanks :) $\endgroup$
    – ABIM
    Jan 10, 2020 at 10:20
2
$\begingroup$

Let $(E,\cal{E},\mu)$ be a measure space and $\mu$ a not finite $\sigma$-finite measure on $\cal{E}$. Then there is a sequence $E_n$ of $E_n \in \cal{E}$ with $E_n \uparrow E$ and $0 < \mu(E_n) < \mu(E_{n+1}) < \infty$. Let $a_1 := \mu(E_1)$ and $a_n := \mu(E_n \setminus E_{n-1})$ for $n > 1$. Let $\mu_n(A) := \mu(A \cap E_n) + \sum_{k=n+1}^\infty 2^{-k} / \mu(E_k \setminus E_{k-1}) \cdot \mu(A \cap (E_k \setminus E_{k-1}))$ for $A \in \cal{E}$. Then each $\mu_n$ is finite, equivalent to $\mu$ and $\mu_n \uparrow \mu$.

$\endgroup$
1
  • $\begingroup$ Edited after reading Jochen Wengenroths answer. I ommited the topological assumtions. $\endgroup$ Jan 10, 2020 at 12:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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