If $\{f_n:[0,1]\to\mathbb{R}\}_{n\ge 1}$ is a sequence of Riemann-integrable functions that converge pointwise to the zero function and $\{f_n\}_{n\ge 1}$ is uniformly bounded by $M$ can we prove that $\int_{[0,1]}f_n\to0$ ?

It seems difficult both to prove or disprove it. For a possible counterexample I came up with the sequence $$f_1(x)=1-|2x-1|$$ $$f_{n+1}=1-2|f_n(x)-1/2|$$ which is the sequence of "spikes" at $x=j/2^n$ for $j=1,...,2^{n-1}$ so that $\int_{[0,1]}f_n=1/2$ for every $n\in\mathbb{N}$ but $f_n(x)$ it doesn't seem to converge to $0$ at some values like $x=1/3$. While we can get some $j/2^n$ arbitrarily close to every value, the closer we get, the crazier the spikes get.

For a possible proof, the set $A_{n,\varepsilon}:=\{\,x:|f_n(x)|\ge\varepsilon\,\}$ should in some sense be small so that we can separate the integral (or at least every lower/upper sum) of $f_n$ into the $A_{n,\varepsilon}$ with "large but bounded" $f_n$ but small interval width and the $[0,1]\setminus A_{n,\varepsilon}$ part with small $f_n$ and whatever interval. So let $P_m:=\{I_1,...,I_m\}$ be the partition of $[0,1]$ with $I_j=[\frac{j-1}{m},\frac{j}{m}]$. Now $J_1,...,J_r$ the $I_j$'s that contain points of $A_{n,\varepsilon}$ and $K_1,...,K_s$ the remaining $I_j$'s. $$U(f_n,P_m)=\sum_{i=1}^rM_{J_i}\frac{1}{m}+\sum_{i=1}^sM_{K_i}\frac{1}{m}\le\frac{rK}{m}+\varepsilon$$ We at least know the limit $r/m$ exists as $m\to\infty$ and depends on $n$ because $f_n$ is integrable. Do you know of any proof or counterexample?



It is true, by Lebesgue's dominated convergence theorem. Just take $M>0$ such that$$(\forall n\in\mathbb N)\bigl(\forall x\in[0,1]\bigr):\bigl\lvert f_n(x)\bigr\rvert\leqslant M.$$Then, if you define $g(x)=M$ for each $x\in[0,1]$, $g$ is integrable and$$(\forall n\in\mathbb N)\bigl(\forall x\in[0,1]\bigr):\bigl\lvert f_n(x)\bigr\rvert\leqslant g(x).$$

  • $\begingroup$ Is there any proof with Riemann-integration techniques? Thanks anyway $\endgroup$ – Pedro May 16 at 16:49
  • 2
    $\begingroup$ @Pedro Yes. Look up a proof for Arzela's dominated convergence theorem (which statement is exactly your question). $\endgroup$ – C_M May 16 at 16:51
  • 1
    $\begingroup$ You will find such a proof here. $\endgroup$ – José Carlos Santos May 16 at 16:55

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.