1
$\begingroup$

Cesàro's Lemma: Suppose that ($b_n$) is a monotonic increasing sequence of strictly positive real numbers such that $b_n \to \infty$. Let $x_n \to x < \infty$, then: $$\frac{1}{b_n}\sum_{k=1}^{n}{(b_k-b_{k-1})x_k}\to x \quad (n \to \infty).$$

Proof is here: proof.

However, I think that the condition of $b_n$, which is a positive monotonic increasing sequence, is too strict, all the proof need is $b_n \to +\infty$.

So is the result still right if I change the condition to $\lim b_n = +\infty$, that is,

$\forall G>0, \exists N > 0$, such that $b_n > G, \forall n > N.$

1 Answers 1

0

Merge the two sequences $(a_n)=(n)$ and $(c_n)=(2^n)$ as one being the even and the other being the odd terms of a new sequence $(b_n)_n$. Set $x_n$ to be a positive constant for all $n$. What result do you obtain?

In the proof, the monotonicity is assumed in the first inequality where the liminf is involved as you need each term of the sum to be positive.

  • 0
    Oh! I ignore the fact that the inequality holds only if both sides multiply a positive number. Thanks!! But in your example, if $x_n = C$, then $\frac{1}{b_n}\sum_{k=1}^{n}{(b_k-b_{k-1})C}=C,\forall n$. So $x_n$ should not be a constant number.2017-01-28
  • 0
    @JyChen At first sight, I did not see it either. It came once I constructed a counterexample. And the counterexample was kind of an educated guess: I considered two sequences of different speed of growth and it worked. Also there is no need to thank me :-)2017-01-28
  • 0
    Thanks for your reminding :-)2017-01-28