7

I am self-learning Real Analysis from the text Understanding Analysis by Stephen Abbott. I would like someone to verify if my proof for the below exercise problem on a continuous nowhere differentiable function is correct.

[Abbott 6.4.3] (a) Show that:

\begin{equation*} g( x) =\sum _{n=0}^{\infty }\frac{\cos\left( 2^{n} x\right)}{2^{n}} \end{equation*} is continuous on all of $\displaystyle \mathbf{R}$.

Proof.

Since, $\displaystyle ( \exists M_{n})$ such that

\begin{equation*} |g_{n}( x) |=\left| \frac{\cos\left( 2^{n} x\right)}{2^{n}}\right| \leq \frac{1}{2^{n}} =M_{n} \end{equation*} and $\displaystyle \sum _{n=1}^{\infty } M_{n}$ converges, by the Weierstrass M-Test, $\displaystyle \sum _{n=1}^{\infty } g_{n}$ converges uniformly on $\displaystyle \mathbf{R}$.

Since each $\displaystyle g_{n}( x)$ is continuous on $\displaystyle \mathbf{R}$, by the term-by-term Continuity theorem, $\displaystyle \sum _{n=1}^{\infty } g_{n}( x)$ is continuous on $\displaystyle \mathbf{R}$.

(b) The function $\displaystyle g$ was cited in section 5.4 as an example of a continuous nowhere differentiable function. What happens if we try to use the theorem 6.4.3 to explore whether $\displaystyle g$ is differentiable?

Proof.

Let

\begin{equation*} g_{n}( x) =\frac{\cos\left( 2^{n} x\right)}{2^{n}} \end{equation*} So,

\begin{equation*} g_{n} '( x) =-\sin\left( 2^{n} x\right) \end{equation*} Since $\displaystyle g$ is nowhere differentiable, by the contrapositive of the term-by-term differentiability theorem, we have that:

If $\displaystyle g$ is not differentiable on $\displaystyle A$, then either $\displaystyle \sum g_{n}( x)$ converges NOT pointwise for all $\displaystyle x\in A$, or $\displaystyle \sum g_{n} '( x)$ converges NOT uniformly on $\displaystyle A$.

Since $\displaystyle \sum g_{n}$ converges uniformly on $\displaystyle \mathbf{R}$, the only possibility is that $\displaystyle \sum _{n=1}^{\infty } g_{n} '=\sum -\sin\left( 2^{n} x\right)$ does not converge uniformly on $\displaystyle \mathbf{R}$.

The question I have is, are we actually required prove this result? And how to go about proving it anyway? Does the below check out?

If a function $\sum h_n$ is uniformly convergent on $A$, it is uniformly convergent $(\forall S) \subseteq A$.

If $(\exists S \subseteq A)$ where $\sum h_n$ converges NOT uniformly, then $\sum h_n$ converges NOT uniformly on $A$.

Consider the point $x_0 = \frac{\pi}{3}$. The sequence $g_n'(x_0)=(-1)^n \frac{\sqrt{3}}{2}$. Thus, $\lim g_n'(x_0) \neq 0$. Consequently, by the $n$th term test, $\sum g_n'(x_0)$ does not converge pointwise on any interval containing $x_0 = \frac{\pi}{3}$.

Therefore, $\sum g_n'$ converges NOT uniformly on $\mathbf{R}$.

Quasar
  • 5,644
  • 2
    I don't have any examples in mind, but there's probably some series $\sum f_n$ such that $\sum f_n$ converges uniformly to $f$ and $f$ is differentiable but such that $\sum f'_n$ does not converge (maybe not even pointwise? Who knows). The derivability theorem is only an implication, we may not have any reciprocal (?). So my guess here would be to go back to the definition of derivative and show that $\frac{g(x+h) - g(x)}{h}$ doesn't have a finite limit when $h$ goes to $0$? – Bruno B Nov 23 '22 at 07:27
  • 1
    I'd like to go back on this comment: I read the question too fast and thought you were trying to prove that the function is nowhere differentiable (which is something that is not very straightforward). What I said holds for that, but that was not your intent. Your proofs and reasonings seem correct, good job! – Bruno B Nov 23 '22 at 11:47
  • The continuous case is interesting. – Claude Leibovici May 07 '24 at 10:58

0 Answers0