Worked solution: The Gauss–Wantzel theorem: constructible regular polygons via cyclotomic fields
Nothing in the argument for was really about the number itself — it only used that satisfies a cyclotomic equation of degree , a power of , which let the roots be peeled apart by repeated halving. The exact same test can be run for any : compute , and check whether it is a power of .
Euler's totient function counts how many whole numbers from to share no common factor with ; it is exactly the degree of the algebra a regular -gon boils down to.
A regular -gon is constructible with straightedge and compass exactly when the primitive -th root of unity is a constructible complex number, equivalently when is constructible. The minimal polynomial of over is the -th cyclotomic polynomial, of degree (Euler's totient function, counting integers from to coprime to ), so .
The argument used for generalizes without change: whenever is a power of , the multiplicative group (which controls how the roots of unity permute among themselves) has an order that is a power of , and groups of order always have a chain of subgroups each half the size of the last. This chain of subgroups produces exactly the kind of nested periods used in Steps 2–3, giving a tower of quadratic extensions from up to , and hence constructibility.
So the question 'which regular polygons are constructible?' reduces entirely to a question about a single arithmetic function: for which is a power of ? The next step answers this by factoring .
- Euler's totient function
- The number of integers from to that share no common factor with ; for example since is prime, and (namely ).