First we show that every g in Sn can be written as a product of disjoint cycles (the existence). Then we prove the uniqueness of this product. Both parts are proved by induction.
If the support of g is empty, then g is e, the identity element, a 0-cycle. We regard this as an empty product of cycles.
Now assume that for some number k > 0 any element g with |supp(g)| < k can be written as a product of disjoint cycles. Let g be an element with k elements in its support. Fix an element x in supp(g). We try to `split off' a cycle containing x.
We set a1 = x and ai + 1 = g(ai) for i > 0. Let m denote the smallest positive integer for which am + 1 = x and consider the cycle c = (a1, ..., am). Its support is a subset of supp(g). So the permutation h = gc-1 fixes all points of fix(g) as well as the points ai, 0 < i < m + 1. Indeed, h(ai) = gc-1 (a i) = g(ai - 1) = ai, where we set a0 = am. This implies that the support of h is contained in supp(g)\ {a1, ..., am}. By the induction assumption we may write h as a product of disjoint cycles c1, c2, ... , ct. The support of these cycles is contained in supp(h) and therefore disjoint from {a1, ..., am}. But then
is a product of disjoint cycles.
By induction we have finished the first part of the proof.
Uniqueness
Assume that g is the product of the
disjoint cycles c1, ..., ct
and at the same time of the disjoint cycles
d1, ..., ds, all of length at least 2.
We prove the uniqueness
by induction on t.
The case t = 0
is trivial.
So assume that t > 0. Then supp(g)
is not empty and we can find an element x in supp(g).
As x is not fixed by g,
there exist cycles ci and dj
which do not fix x. Without loss of generality we may suppose that
x
supp(ct) and
x
supp(ds). For every
m
N, we have
In particular
ct = ds.
But then also c1 ···
ct - 1 =
gct - 1 =
gds - 1 =
d1 ··· ds - 1. The induction hypothesis
yields that t - 1 = s - 1 and, possibly after renumbering of the indices,
ci = di for all i
from 0 to t. This proves the proposition.