Term-wise Differentiation of an Infinite Series

464 Views Asked by At

One way to prove if an infinite sum is to use the following theorem

Theorem Suppose $\{f_n\}$ is a sequence of functions, differentiable on $[a, b]$ and such that $f_n(x_0)$ converges for some point $x_0$ on $[a,b]$. If $f_n^\prime$ converges uniformly on $[a,b]$, then $f_n$ converges uniformly on $[a, b]$, to a function $f$, and $$ f^\prime(x) = \lim_{n\to\infty} f_n^\prime(x) \qquad (a \leq x \leq b) $$

Does this theorem have a converse?


An example of what I mean:

In Elementary Classical Analysis (Marsden) there is the following exercise (p. 316, #52):

Can the series $$x = \sum_{k=1}^\infty \frac{x^k}{k} - \frac{x^{k+1}}{k+1} \quad 0 \leq x \leq 1 $$ be differentiated term by term?

Letting $S_n = \sum_{k=1}^n \frac{x^k}{k} - \frac{x^{k+1}}{k+1} = x-\frac{x^{n+1}}{n+1} $, then $S_n' = 1 - x^{n+1}$. This converges uniformly to $1$ on $[0,r], r < 1$.

So, on $[0,r], r < 1$, we can differentiate term by term by the theorem. (term-by-term is an odd way of expressing the point of this exercise I feel. I think it is phrased this way because $S_n$ are differentiated term by term.)

In this case, we know that the sum is equal to $x$, so the derivative is equal to $1$. The term-wise differentiation yields $0$ at $x=1$, so term-wise differentiation is not possible at $x=1$.

But is there a general method to disproving this? (i.e., if the series was not equal to something that is easy to differentiate, can we conclude anything about it's differentiability from this theorem?

1

There are 1 best solutions below

0
On

The problem is you could have a sequence $f_n$ that converges to $0$ everywhere on $[0,1]$, but with a very large bump that gets closer and closer to $0$. A classical example is something like $f_n(x) = nxe^{-nx^2}$.

The problem is to also obtain $f_n'(x) \to 0$ everywhere on $[0,1]$ which the preceding example does not satisfy. A little experimentation with desmos led me to $$f_n(x) = nx^2e^{-nx^3}.$$

It isn't hard to show that $f_n(x) \to 0$ and $f_n'(x) \to 0$ for all $x \in [0,1]$, but the convergence is highly non-uniform.