The Genius in Abel’s Theorem

Abel’s theorem is usually taught immediately after introducing the convergence of power series on an interval. By the time a student encounters it, they know that a power series converges on an open disk and that within that disk the function it defines is well-behaved. The natural question, which Abel asked, is as follows: what happens at the boundary? If the series also converges at the endpoint, do we arrive there smoothly?

This is one of those theorems that really screams “given all we know, how can I extend this by a marginal step?” The question itself might occur to anyone with good enough intuition. But what blows your mind is not the statement — it is the execution. The proof squeezes everything possible out of a single assumption, and in the process it reveals just how much load a solid understanding of prior concepts can bear.

So let me start by stating the theorem.

The Statement

We have one assumption: the series converges at the endpoint. That’s it. Whatever we are trying to prove, every claim we make will have to follow from that single fact. The proof that follows is an exercise in making sure nothing is wasted.

The Two Tricks

The first genius move is asking: how do I relate the limit K to the power series f(x)? The geometric series gives us the bridge. Since 1/(1−x) = ∑xⁿ for 0 < x < 1, we can multiply both sides by K and suddenly the limit has a power series representation that lives in the same world as f(x).

The second move is even more striking: express each coefficient a_n as the difference of partial sums, a_n = S_n − S_n−1. This rewrites the entire problem in terms of partial sums, which is exactly where the convergence assumption lives.

Harmonizing Indices

Before any progress can be made, the indices must match. This is the unglamorous but essential bookkeeping that makes the algebra work. Once everything is expressed over the same summation range, the (1−x) factor emerges naturally, and the expression collapses into something we can bound.

Split, Bound, Conquer

Now comes the classic analysis move: split the sum into a tail and a head. The tail is controlled by the convergence assumption — since S_n → K, we can make |S_n − K| as small as we like for large enough N. The geometric series collapses ∑xⁿ to 1/(1−x), which turns (1−x) · 1/(1−x) into 1, giving us ε/2.

The head is a finite sum — a constant. And here the proof uses the fact that (1−x) → 0 as x → 1⁻. A finite constant times something going to zero goes to zero. Choose δ = ε/(2(C+1)) and the head is bounded by ε/2.

The Finish

ε/2 + ε/2 = ε. That’s it. The proof is complete.

A final note on how this all comes out to ε. You might look at the proof and wonder why ε/2 was chosen from the start, as if Abel knew in advance that the sum would split into exactly two pieces. He didn’t. The ε/2 was chosen after the structure of the proof revealed itself. First you discover that the expression splits into a tail and a head. Then you realize each can be bounded independently. Only then do you go back to the beginning and write “let ε > 0” knowing that you will divide your budget into two equal halves.

This is the nature of analysis proofs: you discover them backwards and write them forwards. The final product reads as though every step was inevitable, but the actual work happened in reverse — from the conclusion back to the assumptions. The ε/2 is not foresight. It is hindsight, carefully disguised.