I have a problem with determining whether these series are convergent/divergent at the endpoints of their radii of convergence. None of the tests or approaches I know seems to by applicable here...



It is immediate that the radii of convergence are 1. The intervals of convergence are supposedly $\langle x_0-1,x_0+1)$, resp. $(x_0-1,x_0+1\rangle$, which doesn't seem correct, as $\sum_{n=1}^{\infty}\frac{(-1)^{\lfloor\log_2n\rfloor+1}}{\lfloor\log_2n\rfloor+1}$ and $\sum_{n=1}^{\infty}\frac{(-1)^{\lfloor\log_2n\rfloor}}{\lfloor\log_2n\rfloor+1}$ either both converge or diverge.

But the question is: how to determine whether these series convergent/divergent?




There are two different problems here, $x-x_0=1$ and $x-x_0=-1$. We look first at $x-x_0=1$.

So we are interested in the series $$\sum_1^\infty \frac{(-1)^{\lfloor \log_2 n\rfloor}}{\lfloor \log_2 n\rfloor +1 }.$$

Look at the sum of the terms from $n=2^m$ to $n=2^{m+1}-1$, where $m$ is large. Over this interval $\lfloor\log_2 n\rfloor$ does not change. Let us imagine (it makes no real difference) that $m$ is even.

Then the sum of the terms from $n=2^m$ to $n=2^{m+1}-1$ is equal to $\frac{2^m}{m+1}$, which is large. So the partial sums do not have a limit, and therefore the series diverges.

The case $x-x_0=-1$, despite the forbidding appearance, turns out to be pleasantly trivial. Again consider the sum from $n=2^m$ to $n=2^{m+1}-1$. For $m\ge 1$, there is an even number of terms of equal absolute value, and alternating in sign! They cancel. Since the terms have limit $0$, we have convergence.

  • $\begingroup$ I have seen this before and realized that. What confused me is that multiplying by $-1$ we get convergent series. Interesting! What was really helpful was pointing out the Leibniz test. I thought it is not applicable and discarded it, without thinking about modifying it, thanks! $\endgroup$
    – mirgee
    Jul 9 '14 at 16:00
  • $\begingroup$ Another modification of Leibniz would prove convergence of $\sum_{n=1}^{\infty}\frac{(-1)^{\lfloor\log_2n\rfloor+n}}{\lfloor\log_2n\rfloor+1}$ I guess? $\endgroup$
    – mirgee
    Jul 9 '14 at 16:10
  • 1
    $\begingroup$ You are welcome. For the case of the denominator we used, everything is trivial, since there is complete cancellation. The "Leibniz" argument I mentioned is not necessary, though it would be for denominators that did not have such a simple form. $\endgroup$ Jul 9 '14 at 16:13

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.