1
$\begingroup$

This is an exercise of K.L. Chung's book.

Suppose $\{a_n\}$ is an arbitrary enumeration of all rationals,$b_n\ge 0,$ and $\displaystyle \sum_{n=1}^\infty b_n$ converges to a finite number. Let $$f(x)=\sum_{n: a_n\le x}b_n.$$ show that $f(-\infty)\triangleq\lim_{x\to -\infty}f(x)=0$ and $f(\infty)\triangleq\lim_{x\to \infty}f(x)=\sum_{n=1}^\infty b_n$.


It's intuitive that $f(-\infty)=0$ and $f(\infty)=\sum\limits_{n=1}^\infty b_n$, since as $x$ becomes smaller, fewer terms of $\{b_n\}$ are included in the summation, and vice versa. However, I'm not so sure how to prove it rigorously. I list what I tried below, and would appreciate it if someone can take a look and point out any flaws or a better proof.


Suppose $\displaystyle \sum_{n=1}^\infty b_n=c.$ So $\forall \epsilon>0, \exists N$ s.t. $\forall K\ge N$ we have $\displaystyle \sum_{n=1}^K b_n>c-\epsilon.$ Now consider $A\triangleq\{a_1, a_2, ..., a_N\}$, and let $a_{\max}=\max A.$ Then $\forall x>a_{\max}$, $f(x)\ge \displaystyle \sum_{n=1}^N b_n$, and hence $|f(x)-c|<\epsilon,$ since $f$ is clearly monotonically increasing. This shows $f(x)\to c$, as $x\to \infty.$

Similarly, let $a_{\min}=\min A.$ Then $\forall x<a_{\min},$ it follows that $f(x)\le \displaystyle \sum_{n=N+1}^{\infty} b_n<\epsilon$. Since $f(x)\ge 0$ obviously, this implies that $\displaystyle \lim_{x\to-\infty} f(x)=0.$


Are the arguments above correct? Are there flaws or better proofs?

$\endgroup$
3
$\begingroup$

Your proof is quite legit. Nice job!


We can give a shorter and intuitively simpler proof if some big theorems are available to you, although thinking that this example appears at the very beginning of K.L.Chung's A Course in Probability Theory I guess this is probably not the case.

Anyway here is a solution using the dominated convergence theorem: We begin by writing

$$ f(x) = \sum_{n=1}^{\infty} b_n \mathbf{1}_{\{a_n \leq x\}},$$

where $\mathbf{1}_{\{a_n \leq x\}}$ is the indicator function of $a_n \leq x$, i.e., it takes value $1$ when $a_n \leq x$ holds and $0$ otherwise. Then by the dominated convergence theorem, we can pass limit inside the sum. So we have

$$ \lim_{x\to\infty} f(x) = \sum_{n=1}^{\infty} \lim_{x\to\infty} b_n \mathbf{1}_{\{a_n \leq x\}} = \sum_{n=1}^{\infty} b_n $$

and similarly

$$ \lim_{x\to-\infty} f(x) = \sum_{n=1}^{\infty} \lim_{x\to-\infty} b_n \mathbf{1}_{\{a_n \leq x\}} = \sum_{n=1}^{\infty} 0 = 0. $$


A less transparent but still shorter proof is available once we have the following easy-to-prove lemma at our hands:

Lemma. Let $A$ and $B$ be non-empty sets and $f : A\times B \to \mathbb{R}$ be a function. Then

$$ \sup_{a\in A}\sup_{b\in B}f(a, b) = \sup_{(a,b)\in A\times B}f(a, b) = \sup_{b\in B}\sup_{a\in A}f(a, b). $$

Now by the monotonicity,

$$ f(\infty) = \sup_{x\in\mathbb{R}} \sup_{N\geq 1} \sum_{n=1}^{N} b_n \mathbf{1}_{\{a_n \leq x\}} = \sup_{N\geq 1} \sup_{x\in\mathbb{R}} \sum_{n=1}^{N} b_n \mathbf{1}_{\{a_n \leq x\}} = \sup_{N\geq 1} \sum_{n=1}^{N} b_n = \sum_{n=1}^{\infty} b_n. $$

Similarly, if we consider $g(x) = \sum_{n=1}^{\infty} b_n - f(x) = \sum_{n=1}^{\infty} b_n \mathbf{1}_{\{a_n > x\}}$, then

$$ g(-\infty) = \sup_{x\in\mathbb{R}} \sup_{N\geq 1} \sum_{n=1}^{N} b_n \mathbf{1}_{\{a_n > x\}} = \sup_{N\geq 1} \sup_{x\in\mathbb{R}} \sum_{n=1}^{N} b_n \mathbf{1}_{\{a_n > x\}} = \sup_{N\geq 1} \sum_{n=1}^{N} b_n = \sum_{n=1}^{\infty} b_n $$

and hence $f(-\infty) = 0$.

$\endgroup$
  • $\begingroup$ Thanks a lot for the confirmation and providing 2 alternative proofs! They're very helpful, and I really appreciate it. $\endgroup$ – syeh_106 Sep 6 '17 at 2:21
1
$\begingroup$

Suppose $\sum_{n=1}^\infty b_n=c.$ So $\forall \epsilon>0, \exists N$ s.t. $\forall K\ge N$ we have $\displaystyle \sum_{n=1}^K b_n>c-\epsilon.$ (True though, the following argument is for a fixed $\epsilon>0$ . It is a bit strange to use the quantifier $\forall$ this way.)

Now consider $A\triangleq\{a_1, a_2, ..., a_N\}$, and let $a_{\max}=\max A.$ Then $\forall x>a_{\max}$, $f(x)\ge \sum_{n=1}^N b_n$ (by the definition of f), and hence $|f(x)-c|<\epsilon$ (OK, $c-\epsilon<f(x)<c+\epsilon$ and the second inequality is from the definition of $f$ ), since $f$ is clearly monotonically increasing (how is the monotonicity of $f$ relevant here?). This shows $f(x)\to c$, as $x\to \infty.$

Similarly, let $a_{\min}=\min A.$ Then $\forall x<a_{\min},$ it follows that (again by the definition of $f$ ) $f(x)\le \sum_{n=N+1}^{\infty} b_n<\epsilon$. Since $f(x)\ge 0$ obviously, this implies that $\displaystyle \lim_{x\to-\infty} f(x)=0.$


Are the arguments above correct? (Yes.)

$\endgroup$
  • $\begingroup$ You're right. Instead of saying "since $f$ is clearly monotonically increasing", I should've said "since clearly $f(x)\le c.$" Thanks for pointing this out, and other comments! $\endgroup$ – syeh_106 Sep 6 '17 at 2:02

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.