Proof. Each term in the expansion of
$(a+b)^n$ is obtained by making
n decisions of whether to use
a or
b as a
factor. Moreover, any
sequence of
n such decisions yields a term in the expansion. So the expandsion of
$(a+b)^n$ is precisely the
sum of all the
ab-words of length
n, where each word appears exactly once.
Since a and b commute, we can reduce each term via rewrite rules of the form $b^na\mapsto ab^n$ to a term in which the a factors precede all the b factors. This produces a term of the form $a^k b^{n-k}$ for some k, where we use the expressions $a^0 b^n$ and $a^n b^0$ to denote $b^n$ and $a^n$ respectively. For example, reducing the word $babab^2aba$ yields $a^4 b^5$ , via the following reduction. $$ babab^2aba \mapsto a ba ba
b^2a b \mapsto a^2babab^3 \mapsto a^3bab^4 \mapsto a^4b^5. $$ After performing this rewriting process, we collect like terms. Let us illustrate this with the case n = 3.
To determine the
coefficient of a
reduced term, it suffices to determine how many
ab-words have that reduction. Since reducing a term only changes the positions of
as and
bs and not their
number, all the
ab-words where
k of the letters are
bs and
n-k are
as, for
$0\le k\le n$ , have the same normalization. But there are exactly
$\binom{n}{k}$ such
ab-words, since there are
$\binom{n}{k}$ ways to select
k positions out of
n to place
as in an
ab-word of length
n. This shows that the coefficient of the
$a^n$ term is
$\binom{n}{0}=1$ , the coefficient of the
$b^n$ term is
$\binom{n}{n}=1$ , and that the coefficient of the
$a^k b^{n-k}$ term is
$\binom{n}{k}$ .
