The Weierstrass M-Test and a Tricky Series
Define the function
(a) Prove that is continuous on .
(b) Prove that is differentiable nowhere.
Wait — actually, part (b) is false! The series is differentiable (in fact ) everywhere.
Here is the real question:
True or False, and prove it: The function is continuous on .
And the follow-up:
Can you construct a series that is continuous but not differentiable anywhere? What is the threshold on the decay rate of ?
Focus on: Is well-defined and continuous? Prove your answer carefully.
Answer: Weierstrass Series: Continuity via Abel Summation
Key Idea / Intuition
The key question is whether the series converges uniformly. The Weierstrass M-test says: if and , then the series converges uniformly, hence the limit is continuous. For , the coefficients are , and diverges. So the M-test fails. But failing the M-test doesn't immediately mean is discontinuous or ill-defined — we need to think more carefully about pointwise convergence first.
The punchline: is a Weierstrass-type function — it converges conditionally for each fixed (by Dirichlet's test applied to partial sums of ), but the convergence is not uniform, and in fact is continuous but nowhere differentiable — a genuine Weierstrass-type example.
Below we focus on the precise claim the question asks: is well-defined and continuous.
Formal Proof / Solution
Step 1: Pointwise convergence via Dirichlet's test
For fixed not a multiple of , the partial sums are bounded (this follows from the fact that has bounded partial sums when is irrational, and can be checked directly for rational multiples of ).
More precisely, for any fixed , one can show (by geometric series / exponential sum estimates) that
for some constant depending on .
By Dirichlet's test for series: if has bounded partial sums and monotonically, then converges. Here and . So converges pointwise for all .
Step 2: Uniform convergence on compact sets (and continuity)
To show continuity, we use a more refined tool: Abel's summation (summation by parts).
Write . The Weierstrass-type estimate gives:
Actually, let us instead invoke the cleaner uniform version. The key estimate is:
For any , on , the partial sums are uniformly bounded by a constant .
Using Abel summation:
where are the partial sums of the series.
Since uniformly on , and
we get
uniformly on .
This means the series converges uniformly on compact subsets of (and by periodicity, on all of ). Since each partial sum is continuous, and the convergence is uniform, is continuous.
Step 3: The threshold
The general Weierstrass-type function :
- : M-test applies, uniformly convergent, .
- : Converges uniformly (M-test partially), continuous.
- : Converges conditionally (Dirichlet), continuous but nowhere differentiable — a Weierstrass phenomenon.
Our has , which sits exactly at the boundary of the nowhere-differentiable regime.
Summary
The proof uses Dirichlet's test for pointwise convergence and Abel summation + uniform boundedness of partial sums of to upgrade to uniform convergence on compact sets.