The reciprocal of Abel's Theorem for Limits

124 Views Asked by At

Let $\sum_{n=0}^{\infty} a_n x^n$ be a series of powers with a convergence radius $r \in (0, \infty)$, $a_n \ge 0$ for all $n \in \mathbb Z_ +$ and

$$f(x)=\sum_{n=0}^{\infty} a_n x^n, \quad |x| \lt r.$$

Supposing further that there is $$\lim_{x\to r^-} f(x) = s \in \mathbb R. $$ Show that $$f(x)=\sum_{n=0}^{\infty} a_nr^n = s. $$

Let's do it in two steps

$r= 1$ : Let's take $x=ry$.

$$\sum_{n=0}^{\infty} a_n x^n = \sum_{n=0}^{\infty} a_n r^ny^n = \sum_{n=0}^{\infty} b_n y^n; \quad b_n=a_nr^n$$

This is a power series with a convergence ray

$$r' = \frac{1}{\lim_{n \to \infty}|b_n|^{\frac{1}{n}}} = \frac{1}{\lim_{n \to \infty}|a_nr^n|^{\frac{1}{n}}} = \frac{1}{\lim_{n \to \infty}|a_n|^{\frac{1}{n}}r}=1 $$

So, if we have given any power series with radius of convergence $r$, we can transform it into another power series with radius of convergence $1$. So we can take $r = 1$.

Let $\sum_{n=0}^{\infty} a_n x^n$ be a power series with radius of convergence 1.

$$\sum_{n=0}^{\infty} a_n \quad \text{then} \quad \lim_{x \to 1^-} f(x) = \sum_{n=0}^{\infty} a_n $$

Let $s_n= a_0 + a_1 + \cdots +a_n$, $ s_{-1} = 0$, and $\sum_{n=0}^{\infty} a_n = s $.

\begin{align*} \sum_{n=0}^{m} a_n x^n &= \sum_{n=0}^{m}(s_n - s_{n-1}x^n) \\ &= \sum_{n=0}^{m} s_nx^n - \sum_{n=0}^{m} s_{n-1}x^n \\ &= \sum_{n=0}^{m-1} s_nx^n + s_mx^m - x \sum_{n=0}^{m} s_{n-1}x^{n-1} \\ &= \sum_{n=0}^{m-1} s_nx^n + s_mx^m - x\sum_{n=1}^{m} s_{n-1}x^{n-1} \\ &= \sum_{n=0}^{m-1} s_nx^n + s_mx^m - x \sum_{n=0}^{m-1} s_nx^n, \end{align*}

and so,

$$\sum_{n=0}^{m} a_nx^n = (1-x)\sum_{n=0}^{m-1} s_nx^n + s_mx^m .$$

$|x| \lt 1$, $x^m \to 0 $ as $m \to \infty$ and $s_m \to s$. Therefore

$$\lim_{m \to \infty} \sum_{n=0}^{m} a_nx^n = \lim_{m \to \infty} (1-x) \sum_{n=0}^{m-1} s_nx^n + \lim_{m \to \infty} s_mx^m,$$

and

$$ f(x) = \sum_{n=0}^{\infty} (1-x) s_nx^n \quad\text{for} \quad 0 \lt x \lt 1.$$

Since $s_n \to s$, therefore for $\epsilon \gt 0$ then there is an integer $N$ such that $|s_n -s| \lt \frac{\epsilon}{2}$ for all $n \ge N $

$$(1-x)\sum_{n=0}^{\infty} x^n = (1-x)\frac{1}{1-x}$$ for $ |x| \lt 1$

$$(1-x)\sum_{n=0}^{\infty} x^n = 1$$

Therefore, for $ n \ge N$ and $0 \lt x \lt 1 $,

\begin{align*} |f(x) -s| &= \left|(1-x)\sum_{n=0}^{\infty} s_nx^n - s \cdot 1\right| \\ &= \left|(1-x)\sum_{n=0}^{\infty} s_nx^n - (1-x)\sum_{n=0}^{\infty} sx^n \right| \\ &= \left|(1-x)\sum_{n=0}^{\infty} (s_n - s)x^n \right| \\ &\le (1-x)\sum_{n=0}^{N} |s_n - s|x^n + \frac{\epsilon}{2}(1-x)\sum_{n=N+1}^{\infty} x^n \\ &\le (1-x)\sum_{n=0}^{N} |s_n -n|x^n + \frac{\epsilon}{2} \cdot 1 \end{align*}

Now, for a fixed $N$,

$$(1-x)\sum_{n=0}^{N}|s_n - s|x = g(x).$$

$g(x) = 0$ at $ x = 1 \to g(1) = 0$. For definition of continuity, given $\epsilon \gt 0 $ there exist $\delta \gt 0$ s.t $|g(x) - g(1)| \lt \frac{\epsilon}{2}$ whenever $1-\delta \lt x \lt 1$,

$$(1-x)\sum_{n=0}^N |s_n - s| x^n \lt \epsilon$$

So, $|f(x) - s| \lt \frac{\epsilon}{2} + \frac{\epsilon}{2} = \epsilon$ whenever $1-\delta \lt x \lt 1$. Therefore

$$\lim_{x \to 1^-} f(x) = s = \sum_{n=0}^{\infty}a_n$$

That's all I could do, how bad is it?

Thanks in advance for any help.

1

There are 1 best solutions below

3
On BEST ANSWER

1. The most significant gap in your proof is:

Q. How do you know that $s_m \to s$?

Recall that $s$ is defined as $\displaystyle \lim_{x \to 1^-} \sum_{n=0}^{\infty} b_n x^n$. So, whether $s$ is equal to $\displaystyle \sum_{n=0}^{\infty} b_n = \lim_{n\to\infty} s_n$ is not known in advance. And most of all, this is precisely what we want to prove. So your argument is suffering from a circular reasoning.

Edit: Upon a second reading, it seems that you redefined $s$ to be the $\sum_{n=0}^{\infty} a_n$ in the intermediate step of your proof. Then there is only one minor issue remaining: How do you know that the sum $\sum_{n=0}^{\infty} a_n$ is finite?


2. So, how one would prove the claim? I will present two proofs.

1st Proof. Fix $N$, then for each $x \in (0, r)$, the non-negativity of $a_n$'s implies that

$$ \sum_{n=0}^{N} a_n x^n \leq f(x) \leq \sum_{n=0}^{\infty} a_n r^n. $$

Letting $x \to r^-$, we get

$$ \sum_{n=0}^{N} a_n r^n \leq s \leq \sum_{n=0}^{\infty} a_n r^n. $$

Now letting $N \to \infty$ proves the desired equality.

2nd Proof. For any function $f : A \times B \to \mathbb{R}$, we can prove that

$$ \sup_{x \in A} \biggl( \sup_{y \in B} f(x, y) \biggr) = \sup_{(x,y)\in A\times B} f(x, y) = \sup_{y \in B} \biggl( \sup_{x \in A} f(x, y) \biggr) $$

in $\overline{\mathbb{R}} = \mathbb{R}\cup\{-\infty,+\infty\}$. Then by using the monotonicity of $\sum_{n=0}^{N} a_n x^n$ both in $N \in \mathbb{Z}_{\geq 0}$ and in $x \in [0, \infty)$, we get

\begin{align*} s &= \lim_{x \to r^-} \lim_{N\to\infty} \sum_{n=0}^{N} a_n x^n \\ &= \sup_{x \in (0, r)} \biggl( \sup_{N\in\mathbb{Z}_{\geq 0}} \sum_{n=0}^{N} a_n x^n \biggr) \\ &= \sup_{N\in\mathbb{Z}_{\geq 0}} \biggl( \sup_{x \in (0, r)} \sum_{n=0}^{N} a_n x^n \biggr) \\ &= \lim_{N\to\infty} \lim_{x \to r^-} \sum_{n=0}^{N} a_n x^n \\ &= \sum_{n=0}^{\infty} a_n r^n. \end{align*}

Note that this in fact holds regardless of whether both sides are finite or $+\infty$.