Home A generatingfunctionological proof of the binomial theorem
Post
Cancel

A generatingfunctionological proof of the binomial theorem

Make sure you have read the last post on generating functions first, else proceed at your own risk!

Recall that for \(k,n \in \mathbb{N}\), the binomial coefficient is defined by

\[\binom{k}{n} = \frac{k!}{n!(k - n)!} = \frac{\overbrace{k(k - 1)(k - 2)\dotsb(k - n + 1)}^{n\ \text{terms}}}{n!}.\]

You may remember from high school that for some \(k \in \mathbb{N}\),

\[(x + y)^k = \sum_{n = 0}^k \binom{k}{n} x^n y^{k - n} \implies (1 + x)^k = \sum_{n = 0}^k \binom{k}{n} x^n = \underbrace{\binom{k}{0}}_{= 1} + \underbrace{\binom{k}{1}}_{= k} x + \underbrace{\binom{k}{2}}_{= \frac{k(k - 1)}{2}} x^2 + \dotsb + \underbrace{\binom{k}{k}}_{= 1}x^k.\]

This is the binomial theorem.

The generalised binomial theorem: redefining binomial coefficients

But what if we do not restrict $k$ to be a natural number above? This leads us to Newton’s generalised binomial theorem, involving the binomial series:

Theorem 1 (binomial theorem). If \(\alpha \in \mathbb{R}\), then

\[(1 + x)^\alpha = \sum_{n \geq 0} \binom{\alpha}{n} x^n \overset{\text{ops}}{\leftrightarrow} \left(\binom{\alpha}{n}\right)_n;\]

this is an equality in the formal sense, where we don’t care about convergence. (Note that the series converges when \(\lvert x\rvert < 1\) in the analytic sense.)

What exactly does \(\binom{\alpha}{n}\) mean for \(\alpha \not\in \mathbb{N}\)? The factorials are no longer well-defined (even if we consider the Gamma function \(\Gamma\) where \(\Gamma(n) = (n - 1)!\) and enforce that \(\alpha! = \Gamma(\alpha + 1)\), this still fails for negative integers since \(\Gamma\) is not defined there); so we instead use this definition:

Definition 1. The binomial coefficient \(\binom{\alpha}{n}\), where \(\alpha \in \mathbb{R}\), is defined by

\[\binom{\alpha}{n} := \frac{\overbrace{\alpha(\alpha - 1)(\alpha - 2)\dotsb(\alpha - n + 1)}^{n\ \text{terms}}}{n!}.\]

In the case that \(\alpha \in \mathbb{N}\), this obviously agrees with the usual definition, as noted above. Let’s prove the binomial theorem using generating functions! The proof turns out to be quite interesting, and involves solving a differential equation!

A generatingfunctionological proof of the binomial theorem

Fix \(\alpha \in \mathbb{R}\) and let \(A \overset{\text{ops}}{\leftrightarrow} \left(\binom{\alpha}{n}\right)_n\); this is the generating function of the binomial coefficients. Note that when \(\alpha \in \mathbb{N}\), these terms eventually become 0, but usually we get infinitely many nonzero terms. Our goal is to show that \(A = (1 + x)^\alpha\).

Note that

\[\frac{\binom{\alpha}{n}}{\binom{\alpha}{n + 1}} = \frac{\alpha(\alpha - 1)\dotsb(\alpha - n + 1)}{\alpha(\alpha - 1)\dotsb(\alpha - n + 1)(\alpha - n)} \frac{(n + 1)!}{n!} = \frac{n + 1}{\alpha - n} \implies (n + 1)\binom{\alpha}{n + 1} = (\alpha - n)\binom{\alpha}{n};\]

this is a recurrence for the binomial coefficients that holds for \(n \geq 0\), with \(\binom{\alpha}{0} = \frac{1}{0!} = 1\), since 1 is the empty product.

Now recall that if \(A \overset{\text{ops}}{\leftrightarrow} (a_n)\), then \(DA \overset{\text{ops}}{\leftrightarrow} ((n + 1)a_{n + 1})\); this is precisely the left hand side of our recurrence! In summary, and using rule 2 (on polynomials), we see

\[\left((n + 1)\binom{\alpha}{n + 1}\right)_n \overset{\text{ops}}{\leftrightarrow} DA \quad \text{and} \quad \left((\alpha - n)\binom{\alpha}{n}\right)_n \overset{\text{ops}}{\leftrightarrow} (\alpha - xD)A.\]

So \(DA = (\alpha - xD)A \implies (1 + x)DA = \alpha A\). This is a differential equation that the generating function for the binomial coefficients satisfy! We can use the theory of differential equations, in particular the integrating factor method (this is a first order linear ODE), to write this as

\[\left(\frac{1}{1 + x}\right)^\alpha DA - \alpha \left(\frac{1}{1 + x}\right)^{\alpha + 1} A = 0\]

and observe that this gives

\[D\left(\left(\frac{1}{1 + x}\right)^\alpha A\right) = 0 \implies \left(\frac{1}{1 + x}\right)^\alpha A = 1\]

where we use the initial condition \(A(0) = \binom{\alpha}{0} = 1\). Then rearranging, we see that

\[A = \left(\frac{1}{1 + x}\right)^{-\alpha} = (1 + x)^\alpha,\]

and we are done – we have proved that

\[(1 + x)^\alpha = \sum_{n \geq 0} \binom{\alpha}{n} x^n \overset{\text{ops}}{\leftrightarrow} \left(\binom{\alpha}{n}\right)_n!\]

Applying the binomial theorem to the Catalan numbers

Recall from the last post that Catalan numbers count, for instance, the number of ways to have a balanced string of parentheses of length \(2n\), e.g. something like (()(()))(()), but not ())((()). Then it turns out that \(C_n = \frac{1}{n + 1} \binom{2n}{n}\), and their generating function is

\[C = \frac{1 - \sqrt{1 - 4x}}{2x}.\]

Let’s use the binomial theorem with \(\alpha = 1/2\), to recover the general formula for \(C_n\). This is useful, because for instance in the balanced parentheses problem, we can use a combinatorial argument to show that the generating function is \(C\), but a priori, we don’t know an explicit formula for \(C_n\) (or some similar problem).

First recognise that

\[-C = \frac{\sqrt{1 - 4x} - 1}{2x} = \frac{1}{x}\left(\frac{\sqrt{1 - 4x}}{2} - \frac{1}{2}\right);\]

by the binomial theorem,

\[\sqrt{1 - 4x} = (1 - 4x)^{1/2} = \sum_{n \geq 0} \binom{1/2}{n} (-4x)^n = \sum_{n \geq 0} \binom{1/2}{n} (-4)^n x^n\]

where \(\frac{1}{2}\binom{1/2}{0}(-4)^0 = \frac{1}{2}\), so we recognise

\[-C \overset{\text{ops}}{\leftrightarrow} \left(\frac{1}{2}\binom{1/2}{n + 1}(-4)^{n + 1}\right) \implies C \overset{\text{ops}}{\leftrightarrow} \left(\binom{1/2}{n + 1}(-1)^n 2^{2n + 1}\right)\]

using rule 1 (with \(m = 1\)). Then it remains to check

\[\begin{align*} C_n = [x^n]C &= \binom{1/2}{n + 1}(-1)^n 2^{2n + 1} = \frac{\overbrace{(1/2)(-1/2)(-3/2)\dotsb(-(2n-1)/2)}^{n\ \text{terms}}}{(n + 1)!}(-1)^n 2^{2n + 1} \\ &= \frac{1 \cdot 3 \cdot 5 \dotsb (2n - 1)}{2^{n + 1} \cdot (n + 1)!}2^{2n + 1} = \underbrace{\frac{2 \cdot 4 \cdot 6 \dotsb 2n}{2^n \cdot n!}}_{= 1} \cdot \frac{1 \cdot 3 \cdot 5 \dotsb (2n - 1)}{(n + 1)!} 2^n \\ &= \frac{1}{n + 1} \cdot \frac{(2n)!}{(n!)^2} = \frac{1}{n + 1}\binom{2n}{n}, \end{align*}\]

which is the claimed formula for the Catalan numbers! OK, that’s enough for now; see you next year probably with my current rate of writing posts :P

This post is licensed under CC BY 4.0 by the author.