Interchange of limit and sum justification

986 Views Asked by At

From the book I have been reading, it seems the following result is implicitly used:

If $f_{a}(x):=\sum_{n=0}^{\infty} s_{n}(a,x)$ converges uniformly with respect to (large) $a$, for example $\forall |x|<|a|$, and $\lim_{a \rightarrow \infty} s_{n}(a,x) = s_{n}(x)$ $\forall n$ then, \begin{eqnarray} \lim_{a \rightarrow \infty} \sum_{n=0}^{\infty} s_{n}(a,x) = \sum_{n=0}^{\infty} s_{n}(x). \end{eqnarray} I have not been able to find a reference for such a Theorem that justifies this kind of interchange with limit and sum and I could not think of a proof. Is there a Theorem known for this? Or, any idea how to prove the result?

Of course, the result does seem very reasonable. Especially in the example I am working with (I haven't written it to minimise details and keep focus on the main part of my problem).

Many thanks for all help.

1

There are 1 best solutions below

3
On BEST ANSWER

The result is not true without extra assumptions.
If $X= \{ x\}$, then the result would state that we can interchange limits and integration whenever the limit of the integrals exists, which is simply not true.
However, even for arbitrary $X$, it is still not true. Let us define \begin{align*} s_n(a,x) = \begin{cases} 0 & a \not\in (n-1 , n] \\ 1 & a \in (n-1,n] \end{cases}. \end{align*} Then we have \begin{align*} f_a(x) = \sum_{n=1}^\infty s_n(a,x) = s_{\lceil a \rceil }(a,x) = 1 \end{align*} for all $a$. So in particular $f_a(x)$ converges uniformly to $1$. However \begin{align*} s_n(x) = \lim_{a \to \infty } s_n(a,x) = \lim_{ \substack{a \to \infty \\ a > n}} s_n(a,x) =0 \end{align*} for all $n \in \mathbb{N}$.
Thus we find \begin{align*} 1 = \lim_{a \to \infty} \sum_{n=1}^\infty s_n(a,x) \neq \sum_{n=1}^\infty s_n(x) = 0. \end{align*}