I don't understand how this works, I understand the idea behind it. Similar to the squeeze theorem it is pretty logical and easy to visually see.

So then what is happening when I want to evaluate $$\sum_{n \geq 1} \frac{1}{n}.$$

So my $b_n$ is then $\frac{1}{n^2}$

My $\{b_n\}$ converges but my $\{a_n\}$ doesn't which I already know because it is drilled into you. I guesss I could prove it with a $p$-series test or something like that, but why does the comparison test fail here and where else does it fail that isn't in specified?

  • $\begingroup$ For the harmonic series, use the condensation test. $\endgroup$ – Hagen von Eitzen Jul 21 '13 at 14:40
  • $\begingroup$ But why, it isn't specified as a short coming of the comparison test. $\endgroup$ – Paul the Pirate Jul 21 '13 at 14:43

The comparison test does not fail here as the hypothesis is not satisfied. Note that

$$ \frac{1}{n} > \frac{1}{n^{2}}$$ and not the other way round. Hence, the convergence of the series

$$\sum\limits_{n=1}^{\infty} \frac{1}{n^{2}}$$

gives no information about the convergence/divergence of $\sum\frac{1}{n}$.

By the way, to prove that $\sum\frac{1}{n}$, you can show that the sequence of partial sums is not Cauchy.

EDIT: The comparison test says that if $(a_{n})$ and $(b_{n})$ are two sequences such that $a_{n} \leq b_{n}$ and that $\sum b_{n}$ converges, then $\sum a_{n}$ converges.

  • $\begingroup$ I thought the $a_n$ and $b_n$ were referring to the part being squared so I was just replacing the function part of it. Like that weird function replacement stuff. I don't know what "Cauchy" means. $\endgroup$ – Paul the Pirate Jul 21 '13 at 14:46
  • $\begingroup$ See this for the test of whether the partials sums are a Cauchy sequence. In particular, if the series were to converges, then $\sum_n^{2n} a_n$ should go to $0$ (taking $p=n$); for the harmonic series, you have $$\sum_n^{2n} \frac{1}{n} \geq n\cdot \frac{1}{2n} = \frac12$$ for all $n$. $\endgroup$ – Clement C. Jul 21 '13 at 15:00

What is written below is intended to help a bit with the intuition. All series discussed below have non-negative terms.

Call a series $\sum x_n$ bad if it diverges, and good if it converges.

Comparison Test 1 says that if $\sum b_n$ is good, and $0\le a_n\le b_n$ for all $n$, then $\sum a_n$ is good.

In cruder terms, if $\sum b_n$ doesn't blow up, and the $a_n$ are smaller, then $\sum a_n$ doesn't blow up.

Now suppose $\sum a_n$ is bad, and $a_n\le b_n$ for all $n$. Then $\sum b_n$ cannot be good. For if it were, then by Comparison Test 1, $\sum a_n$ would be good. But it isn't. So we have obtained:

Comparison Test 2: If $\sum a_n$ is bad, and $0\le a_n\le b_n$ for all $n$, then $\sum b_n$ is bad. If $\sum a_n$ blows up, and the $b_n$ are bigger than the $a_n$, then $\sum b_n$ blows up.

There are useful variants of the comparison tests. For one thing, what happens for "small" $n$ doesn't matter. If $a_n\le b_n$ for large enough $n$, and $\sum b_n$ is good, then $\sum a_n$ is good. Also, if $\sum b_n$ is good, and there is a positive constant $c$ such that $0\le a_n\le cb_n$, then $\sum a_n$ is good.

In order to apply the Comparison Tests successfully, we need to have a basic stock of series whose convergence/divergence we know about.

That basic stock need not be large. In practice, most comparisons are done with geometric series, or with $p$-series $\sum\frac{1}{n^p}$.

We do have to be careful about the direction of the inequalities. Suppose we know already that $\sum \frac{1}{n^2}$ is good. That doesn't help at all with $\sum \frac{1}{n}$. For we have $\frac{1}{n^2}\le \frac{1}{n}$. So the terms of $\sum \frac{1}{n}$ are bigger than the terms of a good series. Not useful: A series bigger than a good series can easily blow up.

Very informally, a series of positive terms converges (is good) if the terms approach $0$ "fast enough." It turns out that $\frac{1}{n^2}$ approaches $0$ fast enough, but $\frac{1}{n}$ does not approach $0$ fast enough.

Interestingly, $\frac{1}{n}$ approaches $0$ almost fast enough. For example, $\frac{1}{n^{1.01}}$, which approaches $0$ only a little faster than $\frac{1}{n}$, is "fast enough."

  • $\begingroup$ So how is this limited? If I have $\sum \frac{1}{47^n - 44^n$ can I compare it to the larger function $\frac{1}{2^n}$ and state it converges? $\endgroup$ – Paul the Pirate Jul 29 '13 at 19:54
  • $\begingroup$ This one is somewhat harder. Maybe I would do this. The denominator is $44^n((47/44)^n -1)$. So the denominator is $\ge 44^n (3/44)$. Now by comparison with a geometric series, we have convergence. Probably easier for most students is the limit comparison test, using $\sum \frac{1}{47^n}$. $\endgroup$ – André Nicolas Jul 29 '13 at 20:05
  • $\begingroup$ I suppose what I am getting at is why can't I just do what I did above? What specifically is wrong with that claim. $\endgroup$ – Paul the Pirate Jul 29 '13 at 23:09
  • $\begingroup$ It is not enough to have a good intuition (though that is enormously useful). One also needs to be able to give detailed verifications. $\endgroup$ – André Nicolas Jul 29 '13 at 23:25
  • $\begingroup$ I guess the idea of "good" is kind of fuzzy. How good is good enough is what I don't get. $\endgroup$ – Paul the Pirate Jul 30 '13 at 1:13

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.