Let $f\in L_1(-\infty,\infty)$ be a Lebesgue-summable function on $\mathbb{R}$. I read that the function$$\varphi(x)=\int_{[\xi_0,\xi]}f(x+t)d\mu_t$$is absolutely continuous on any real closed interval. I knew that if $g\in L_1[a,b]$ then $\int_{[a,x]}g(t)d\mu_t$ is absolutely continuous as a function of $x$, but I cannot prove the absolute continuity of $\varphi$ to myself.

Moreover, it is said that it is Lebesgue-integrable on $\mathbb{R}$.

How can we prove those two statements?

As to the Lebesgue-integrability, I see that it would follow, because of Fubini's theorem, from the measurability of $\int_{\mathbb{R}}|f(x+t)|d\mu_x$ as a function of $t$, but I have no idea of how to prove it, if it is true.

I $\infty$-ly thank you all!!!

  • 1
    $\begingroup$ $\mu$ is the Lebesgue measure on $\mathbb{R}$? $\endgroup$ – Daniel Fischer Nov 19 '14 at 13:50
  • $\begingroup$ Thank you for the comment! Yes, it is. $\endgroup$ – Self-teaching worker Nov 19 '14 at 13:51

Regarding the integrability of $\varphi$, rather than showing the measurability of

$$t \mapsto \int_\mathbb{R} \lvert f(x+t)\rvert\,d\mu_x,$$

use the mesurability of

$$(x,t) \mapsto f(x+t)$$

which is for example shown here. Then we can compute

\begin{align} \int_\mathbb{R} \lvert \varphi(x)\rvert\,d\mu_x &= \int_\mathbb{R} \left\lvert \int_{[\xi_0,\xi]} f(x+t)\,d\mu_t\right\rvert\,d\mu_x\\ &\leqslant \int_\mathbb{R} \int_{[\xi_0,\xi]} \lvert f(x+t)\rvert\,d\mu_t\,d\mu_x\\ &= \int_{[\xi_0,\xi]}\int_{\mathbb{R}} \lvert f(x+t)\rvert\,d\mu_x\,d\mu_t\\ &= \int_{[\xi_0,\xi]} \lVert f\rVert_{L^1(\mathbb{R})}\,d\mu_t\\ &= (\xi-\xi_0)\lVert f\rVert_{L^1(\mathbb{R})}\\ &< +\infty \end{align}

to see that $\varphi \in L^1(\mathbb{R})$.

For the absolute continuity of $\varphi$, write (for $x < y$ to avoid dealing with signs)

\begin{align} \varphi(y) - \varphi(x) &= \int_{[\xi_0,\xi]} f(y+t)\,d\mu_t - \int_{[\xi_0,\xi]} f(x+t)\,d\mu_t\\ &= \int_{y+\xi_0}^{y+\xi} f(u)\,d\mu_u - \int_{x+\xi_0}^{x+\xi} f(u)\,d\mu_u\\ &= \int_{x+\xi}^{y+\xi} f(u)\,d\mu_u - \int_{x+\xi_0}^{y+\xi_0} f(u)\,d\mu_u\\ &= \int_{x}^y f(\xi+t)\,d\mu_t - \int_x^y f(\xi_0+t)\,d\mu_t\\ &= \int_x^y f(\xi+t) - f(\xi_0+t)\,d\mu_t. \end{align}

Thus we see that $\varphi$ is the integral of the integrable function

$$t \mapsto f(\xi + t) - f(\xi_0+t),$$

and hence $\varphi$ is absolutely continuous.

  • $\begingroup$ I heartily thank you! As always, your asnwers are deep and precious. Only one thing isn't clear to me in the proof of $(x,y)\mapsto f(x\pm t)$: why all measurable sets of $\mathbb{R}\times[\xi_0,\xi]$ are in the form $G_\delta \cup N$ with $N$ of null measure? If you consider it more proper, I could open a new thread... Thank you again!!!!! $\endgroup$ – Self-teaching worker Nov 19 '14 at 20:09
  • 1
    $\begingroup$ They are $G_\delta \setminus N$ (or, equivalently, $F_\sigma \cup N$). The Lebesgue measure is regular, so for every measurable set $M$ we have $\lambda(M) = \inf \{ \lambda(U) : U\text{ open and } M\subset U\}$. Take a sequence $(U_n)$ of open sets with $M\subset U_n$ and $\lambda(U_n) < \lambda(M) + 2^{-n}$. Then $A =\bigcap U_n$ is a $G_\delta$ with $M\subset A$, and $\lambda(A) \leqslant \lambda(M) + 2^{-n}$ for all $n$, hence $\lambda(A) = \lambda(M)$, so $A\setminus M$ is a null set [and $M = A\setminus (A\setminus M)$]. $\endgroup$ – Daniel Fischer Nov 19 '14 at 20:23
  • $\begingroup$ Ouch, I spent the afternoon trying to use the sets $U_n$ such that $\lambda(A\triangle U_n)< \varepsilon_n$ defining the Lebesgue measurable set. I heartily thank you again!!!!! $\endgroup$ – Self-teaching worker Nov 19 '14 at 21:03
  • $\begingroup$ If I'm not bothering you too much: I see that the same holds for any topological subspace of $\mathbb{R}^n$ endowed with Lebesgue "ordinary" measure $\lambda$ (constructed by using the volume of $n$-parallelepipeds): am I right? Thank you so much again!!! $\endgroup$ – Self-teaching worker Nov 19 '14 at 21:51
  • 1
    $\begingroup$ What "the same"? The $G_\delta \setminus N$? Yes, that holds for all Lebesgue-measurable subspaces of $\mathbb{R}^n$. [In non-Lebesgue-measurable subspaces, we have the question what we call measurable subsets, if Lebesgue-measurable subsets of $\mathbb{R}^n$ which happen to be contained in our subspace, the same holds.] $\endgroup$ – Daniel Fischer Nov 19 '14 at 22:08

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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