Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

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

Suppose $g\in L^1([0,\infty))$ and $\int_0^\infty g(x)\, dx=1$.

How to prove that if $f:[0,\infty) \rightarrow \mathbb{R}$ is a continuous function then $n \int_0^1 f(x+t) g(nt)\, dt \rightarrow f(x)$ as $n\rightarrow \infty$ for $x\in \mathbb{R}$?

Say $nt=u$ in the integral then we get $\int_0^n f(x+u/n) g(u) \,du$. Estimating this integral from above and letting $f_n(x)= f(x+u/n) \chi_{[0,n]} $ gives nothing since we don't know whether $f$, being pointwise limit of this sequence, is integrable on $\mathbb{R}$. How can I use the continuity of $f$? Somehow Lebesgue Dominated Convergence Theorem is gonna play a role but I cannot get it, could you help?


share|cite|improve this question
Hint: Consider $|\int_0^{\infty}f(x+u/n) g(u) \,du-\int_0^{\infty}f(x) g(u) \,du|\leq|\int_0^{A}M\epsilon\,du|+|\int_A^{\infty}\epsilon g(u)\,du|$ – 89085731 Apr 17 '12 at 0:51
Why is $g \leq \epsilon$ on the interval $[0,A] $? Do you mean this? – Ragnar Apr 17 '12 at 1:22
Sorry I need to make a correction.See below$|\int_0^{\infty}f(x+u/n) g(u) \,du-\int_0^{\infty}f(x) g(u) \,du|\leq|\int_0^{A}M_1\epsilon|+|\int_A^{\infty}M_2 g(u)|$ – 89085731 Apr 17 '12 at 1:26
$\epsilon$ is for f(x+u/n)-f(x) in a finite interval – 89085731 Apr 17 '12 at 1:27
up vote 3 down vote accepted

I can't help but see these questions in terms of probability.

Let $T$ be a non-negative random variable with density $g(t)$; then $ng(nt)$ is the density of $T/n$, and so $$n\int_0^1f(x+t)g(nt)\,dt=\int f(x+T/n)\,1_{[0,1]}(T/n)\,dP.$$

The random variables $f(x+T/n)\,1_{[0,1]}(T/n)$ are uniformly bounded by $M:=\sup_{x\leq y\leq x+1}|f(y)|$ and converge to the constant $f(x)$ so by dominated convergence we have $$ \int f(x+T/n)\,1_{[0,1]}(T/n)\,dP\to \int f(x)\,dP=f(x).$$

share|cite|improve this answer

This is just Byron's answer again:

Let $g_n(t) = f(x+t/n)\chi_{[0,n]}g(t)$. Note that for $t\in[0,n]$ we have $ x+t/n\in[x,x+1]$. Let $M$ be an upper bound for $|f|$ on $[x,x+1]$. We then have $|g_n(t)|\le M |g(t)|$ for all $t\ge0$. Also $\lim\limits_{n\rightarrow\infty} g_n(t)= f(x)g(t)$ for all $t\ge0$. Thus, by the Dominated Convergence Theorem: $$ \int_0^n f(x+t/n) g(t)\,dt=\int_0^\infty g_n(t)\,dt\ \ \buildrel{n\rightarrow\infty} \over\longrightarrow\ \ \int_0^\infty f(x)g(t)\,dt=f(x). $$

share|cite|improve this answer

You can easily prove the statement for linear combinations of indicator functions. Take $g=\frac{1}{b-a}\chi_{[b-a]}$: $$ n\int_0^1 f(x+t)g(nt)dt=\frac{n}{b-a}\int_{a/n}^{b/n}f(x+t)dt=f(x+c), $$ for a $c\in[a/n,b/n]$ (this is the meanvalue theorem). For $n\rightarrow\infty$ $c$ converges to $0$ and so $f(x+c)\rightarrow f(x)$. Now approximate $g$ by such functions to get the result.

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.