I am considering a problem in Browder's Mathematical Analysis:

Let $(a_n)_{n=1}^{\infty}$ be a sequence in $\mathbb{R}$.

(a) Show that if $\lim_{n \rightarrow \infty} a_n = A$ exists, then $$\lim_{n \rightarrow \infty} \frac{a_1 + \dots + a_n}{n} = A.$$

Proof: Without loss of generality, suppose $a_n \rightarrow 0$. Fix $\epsilon > 0$. Then there exists an $N \in \mathbb{N}:\ \forall n \geq N$, $|a_n| < \epsilon.$ Set $S_N = \sum_{k=1}^N |a_k|.$ Choose $\tilde{N} > N:\ \forall n \geq \tilde{N}$, $$\frac{S_N}{n} < \epsilon.$$ Then if $n \geq \tilde{N} > N$, \begin{align*} \left|\frac{a_1 + \dots + a_n}{n}\right| & \leq \frac{1}{n}\left(|a_1| + \dots + |a_n|\right) \\ & = \frac{S_N}{n} + \frac{|a_{N+1}| + \dots + |a_n|}{n} \\ & < \epsilon + \frac{(n-(N+1))\epsilon}{n} \\ & \leq 2 \epsilon. \end{align*}

(b) Show that if $\sum a_n$ converges, then $$\lim_{n \rightarrow \infty} \frac{1}{n} \sum_{k=1}^n ka_k = 0.$$

Unfortunately I am having trouble proving part (b). I believe that the assumption that the series $\sum a_n$ converges tells us that the sequence $na_n \rightarrow 0$, i.e. that the sequence $a_n$ converges fairly quickly to zero. If this is the case, I can invoke the result in part (a) and conclude but I am not sure how to proceed. I've seen an extremely similar question asked here: Series converges implies $\lim{n a_n} = 0$ but this question does not place any assumptions on the monotonicity on the sequence $a_n$.

  • $\begingroup$ I think that part b) can be solved with summation by parts (not totally sure). The derived sequence of part a) is named Césaro sum. $\endgroup$ – Masacroso Apr 30 '17 at 0:28

For part (b), assume $S_n = \sum_{j=1}^n a_j \to S $ as $n \to \infty.$

We have

$$\sum_{j=1}^nS_j = \sum_{j=1}^n\sum_{k=1}^ja_k = \sum_{j=1}^n\sum_{k=1}^na_k 1_{(k \leqslant j)}= \sum_{k=1}^n\sum_{j=1}^na_k 1_{(k \leqslant j)}$$

where we use the indicator $1_{(k \leqslant j)} = 1 (\,\,\text{if} \,\,k \leqslant j), = 0 (\,\,\text{if} \,\,k > j)$ and switch the sums.


$$\sum_{j=1}^nS_j = \sum_{k=1}^n\sum_{j=1}^na_k 1_{(k \leqslant j)} = \sum_{k=1}^n\sum_{j=k}^na_k = \sum_{k=1}^n a_k (n - k +1) \\ = (n+1) \sum_{k=1}^n a_k - \sum_{k=1}^n k a_k,$$


$$\tag{*}\frac{1}{n} \sum_{k=1}^n k a_k = (1 + 1/n)S_n - \frac{1}{n} \sum_{j=1}^n S_j.$$

Since $S_n \to S$ it follows that $\frac{1}{n} \sum_{j=1}^n S_j \to S$ by the result of part (a), and taking the limit of (*) we get

$$\lim_{n \to \infty} \frac{1}{n} \sum_{k=1}^n k a_k = 1 \cdot S - S = 0.$$

Alternatively, summing by parts with $S_0 := 0$ we get

$$\begin{align}\sum_{k=1}^n k a_k &= \sum_{k=1}^n k(S_k - S_{k-1}) \\ &= nS_n+ \sum_{k=1}^{n-1}k S_k - \sum_{k=2}^nkS_{k-1} \\ &= nS_n + \sum_{k=1}^{n-1}k S_k - \sum_{k=1}^{n-1}(k+1)S_{k} \\ &= nS_n - \sum_{k=1}^{n-1} S_k \\ &= (n+1)S_n - \sum_{k=1}^n S_k \end{align}$$

Now proceed as before.

  • $\begingroup$ This is very slick thank you! $\endgroup$ – xk3 Apr 30 '17 at 2:19
  • $\begingroup$ You're welcome. The summation by parts approach that I added is the typical way to go, but the trick in the first approach should prove useful for other types of problems. $\endgroup$ – RRL Apr 30 '17 at 2:22

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.