3

Suppose that $\lim_{n\to\infty}\frac{a_n}{b_n}$=1. And $\sum^{\infty}_{n=1}a_n$ converges, $\sum^{\infty}_{n=1}b_n$ diverges. Are such sequences $(a_n)$, $(b_n)$ exist?

Sandra West
  • 854
  • 6
  • 18

3 Answers3

6

You may consider, for example, $a_n=\frac{(-1)^{n-1}}{\sqrt{n}}$ and $b_n=a_n+\frac{1}{n}$.

23rd
  • 16,234
  • 44
  • 70
2

If we assume that $\{a_k\}$ and $\{b_k\}$ are positive, no such sequences exist:

To see why, suppose that

$$\lim_{n \to \infty} \frac{a_n}{b_n}=1,$$

we may choose $N>0$ sufficiently large so that $b_n < 2a_n$ for all $n \geq N$. Then the tails of our series satisfy $$\sum_{k=N}^\infty b_k<\sum_{k=N}^\infty 2a_k,$$ so that $\sum b_k$ converges by the comparison test.

Wikipedia has more information here.

awwalker
  • 6,924
1

For $n \ge 3$, let $a_n= -\frac{1}{\log n}$ if $n$ is odd, and $\frac{1}{\log(n-1)}$ if $n$ is even. The series converges to $0$, consecutive terms cancel.

Let $b_n=-\frac{2}{\log 2n+\log n}$ if $n$ is odd, and $b_n=a_n$ if $n$ is even.

The sum of two consecutive terms, the first of which has odd index, of the series $\sum b_n$ is $\frac{1}{\log(n-1)}-\frac{2}{\log(2n)+\log(n)}$. This simplifies to $$\frac{\log(2n)+\log n-2\log(n-1)}{(\log(2n)+\log n)\log(n-1)}.$$

The top has limit $\log 2$. The bottom is slow-growing, so the series diverges.

Note that $\lim_{n\to\infty}\frac{a_n}{b_n}=1$.

Remark: The example by Landscape is much nicer.

André Nicolas
  • 507,029