A question about finite fields

Background

Cyclic groups

For each positive integer m let's write Zm for the (commutative) ring Z/mZ. Here are some quick facts:

We also have very precise information about the entire lattice of subgroups of a finite cyclic group:

Subgroups of finite cyclic groups

Suppose G is a finite cyclic group of order m, i.e, Gβ‰…Zm. Then for every subgroup H≀G we have:

  • H is cyclic; and
  • |H| divides |G|.

Conversely, for each divisor d of m there is a unique subgroup H≀G of order d. This subgroup consists precisely of those elements g∈G that have order dividing d. (If we use multiplicative notation in G, then these are the elements g that satisfy gd=1.)

Moreover, if H,K≀G are subgroups of orders d and e, respectively, then H≀Kexactly when d divides e.

In other words, the subgroups of Zm look exactly like Zd for the divisors d of n, and Zd≀Ze exactly when d divides e.

Finite fields

For each finite field F, we recall the following basic facts:

Quick summary of the above facts: Every finite field F has order pn for some prime p and positive integer n, and up to isomorphism there is exactly one field of each such order.

The big result we need from field theory

We need the following foundational result from field theory, which is sneakier to prove than it might seem:

Groups of units of fields

Suppose F is any field and FΓ— is its group of (multiplicative) units. Then every finite subgroup of FΓ— is cyclic.

Corollary

Suppose F is a finite field of order pn. Then FΓ—β‰…Zpnβˆ’1.

So, if we're wondering about the subgroups of FΓ— for a finite field, we just need to understand the subgroups of the cyclic group Zpnβˆ’1. Fortunately, the result above tells us the entire lattice structure for the subgroups of Zpnβˆ’1: there is a unique subgroup H of order d for each divisor d of pnβˆ’1, and for two such subgroups H,K≀FΓ— we have H≀K exactly when |H| divides |K|.


The subgroup of kth powers of FΓ—

Now we wish to understand the subgroup K≀FΓ— consisting of elements that are kth powers of elements in FΓ—. This subgroup is exactly the image of the group morphism Ο•:FΓ—β†’FΓ— defined by Ο•(Ξ±)=Ξ±k. What is the size of this subgroup? One way to answer this question is to use the First Isomorphism Theorem for groups, which tells us that

FΓ—/ker⁑(Ο•)β‰…im(Ο•).

In other words,

|K|=|im(Ο•)|=[FΓ—:ker⁑(Ο•)]=|FΓ—||ker⁑(Ο•)|=pnβˆ’1|ker⁑(Ο•)|.

Can we compute ker⁑(Ο•)? This is the subgroup of FΓ— consisting of those elements Ξ± that satisfy Ξ±k=1, i.e., they are the elements of the field F that are the roots of the polynomial f(x)=xkβˆ’1. Or, strictly from the point of view of group theory, they are the elements in the group FΓ— of order dividing k.

From the description of such Ξ± as the roots of f(x)=xkβˆ’1, you might at first think there are exactly k such elements. But remember that we're working in positive characteristic, so we can easily be duped by repeated roots. For example, if p=2 and k=4, then we have f(x)=x4βˆ’1=(xβˆ’1)4, which only has one root!

We can make things easier on ourself by imposing some conditions on k. For example, if we restrict ourselves to those k that divide pnβˆ’1, then by our fact about finite cyclic groups above, the set of elements of order dividing k is exactly the subgroup of FΓ— of order k. In other words, in this case |ker⁑(Ο•)|=k and hence |K|=pnβˆ’1k.

Now we can answer the following question:

A question about kth powers in finite fields

Suppose F is a finite field of order pn and k is a positive integer that divides pnβˆ’1. Let H=ZpΓ— and let K≀FΓ— be the subgroup of kth powers in FΓ—. When do we have H≀K?

In this case, we have |H|=pβˆ’1 and (based on our work above) |K|=pnβˆ’1k. Using our basic theory about subgroups of the finite cyclic group FΓ—, we can answer the question: we have H≀K exactly when pβˆ’1 divides pnβˆ’1k.

If desired, we can rewrite the divisibility condition equivalently as: H≀K exactly when k divides pnβˆ’1pβˆ’1.