The Weierstrass M-Test states that if you have a sequence of functions $(f_k)_{k\epsilon\mathbb N}$ where $f_k:A\mapsto\mathbb R$, and suppose that $\forall k \epsilon \mathbb N \exists M_k >0 $ such that

$$|f_k(x)|\leq M_k \forall x \epsilon A$$ and $$\sum_{k=1}^{\infty}M_k <\infty$$

Then the series $\sum_{k}f_k$ converges uniformly on A.

However, the Weierstrass M-test is not applicable to series that converge uniformly but not absolutely.

Therefore, when attempting to prove a series converges uniformly using the M test, do I need to show first of all that the series converges absolutely?

  • $\begingroup$ If you've shown $|f_k|\le M_k$ and $\sum M_k<\infty$, then you've already shown it: the series converges absolutely everywhere by the comparison test. $\endgroup$
    – zhw.
    Dec 19, 2017 at 18:53

1 Answer 1


Not really. The conclusion of the usual statement of the M test (cf Wikipedia) is that the series converges uniformly and absolutely. So obviously if it does not converge absolutely you will never be able to use the M test, but that will be reflected in the premises of the M test not holding.

In particular, if $f_n(x)$ did not converge absolutely for some $x,$ and the premises of the m test held we would have $$\infty=\sum_n |f_n(x)| \leq \sum_n M_n < \infty,$$ a clear contradiction.

  • $\begingroup$ So if I have a series of functions and I have to decide whether it converges uniformly or not, I could first try using the M-test. If the M-test conditions are not satisfied, this does not necessarily mean the series does not uniformly converge. However I don't need to check whether the series converges absolutely as if it doesn't, it would fail the conditions for the M-test and I would have to use another test to decide whether it is uniformly convergent. Is this correct? $\endgroup$ Dec 19, 2017 at 18:06
  • $\begingroup$ Yes that's right. For instance I saw this question the other day about proving uniform convergence for a series that didn't converge absolutely math.stackexchange.com/questions/2567371/… . They needed an different approach than the M test. $\endgroup$ Dec 19, 2017 at 18:19

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.