2

I am trying to show that $f(x) = \sum_n f_n(x)$ is absolutely continous on $[a,b]$, where $f_n(x)$ is an increasing absolutely continuous function on $[a,b]$ for each $n$ and $\sum_n f_n(x)$ converges.

My thoughts: choose $\delta_k$ as a response to the $\epsilon/2^k$ challenge regarding the criterion for the absolute continuity of $f_k$ on $[a, b]$. Then $\min \{\delta_k\}$ would be the response the $\sum_k \epsilon/2^k = \epsilon$ challenge regarding the criterion for the absolute continuity of $f = \sum_n f_n(x)$ on $[a, b]$. This is certainly true for finite sums, but is this still valid for countable sums? Also I realized that I did not use the condition that each $f_n$ is increasing anythere, which sugguests that this approach might not work.

Can someone give me a hint? Thanks in advance!

Mittens
  • 39,145
  • Thanks for your comment! I just have one question: is $\sum_n f'_n = f'$ from $f = \sum_n f_n$? I don't know if this would be true for countable sums. It looks like what I need to next is state $f(x) = \sum_n f_n(x) = \sum_n f_n(a) + \sum_n \int_a^x f'_n = f(a) + \int_a^x f'$. – InsultedByMathematics Oct 24 '22 at 16:42
  • Yes. I just realized that I don't need $f' = \sum_n f'_n$. I just need $f(x) - f(a)$ to be the indefinite integral of some function. Also, I think $\int \sum f'_n = \sum \int f'_n$ is true not by monotone convergence but rather by Beppo Levi's Lemma since I don't know if $\sum_n f'_n$ converges, right? – InsultedByMathematics Oct 24 '22 at 17:06
  • If you prefer to call it Levy-Beppo theorem, that's fine. you say "zee" I said "zed". – Mittens Oct 24 '22 at 17:15
  • Ok. Thank you very much! – InsultedByMathematics Oct 24 '22 at 17:17

1 Answers1

1

If you are familiar with the fundamental theorem of Calculus (FTC-Lebesgue)in the sense of Lebesgue the problem has a rather simple solution.

The FTC-Lebesgue implies that $f_n$ (by virtue of being absolutely continuous) is differentiable almost surely, $f'_n\in L_1([a,b])$ and $f_n(x)=f(a)+\int^x_af'_n$ for all $a\leq x\leq b$. The monotonicity assumption implies that $f'_n\geq0$ when it exists.

Monotone convergence yields $\sum_n\int_{(a,x]}f'_n=\int_{(a,x]}\sum_n f'_n$. Putting this together $$\sum_n(f_n(x)-f_n(a))=\sum_n\int_{(a,x]}f'_n=\int_{(a,x]}\sum_nf'_n$$ Thus, if $F=\sum_nf_n$ (which by assumption exists on $[a,b]$ then $$F(x)=F(a)-\int_{(a,x]}\sum_nf'_n$$ and so, $F$ is itself ansulutely continuous.

Mittens
  • 39,145