# Perfect field

In algebra, a field k is perfect if any one of the following equivalent conditions holds:

Otherwise, k is called imperfect.

In particular, all fields of characteristic zero and all finite fields are perfect.

Perfect fields are significant because Galois theory over these fields becomes simpler, since the general Galois assumption of field extensions being separable is automatically satisfied over these fields (see third condition above).

Another important property of perfect fields is that they admit Witt vectors.

More generally, a ring of characteristic p (p a prime) is called perfect if the Frobenius endomorphism is an automorphism.[1] (When restricted to integral domains, this is equivalent to the above condition "every element of k is a pth power".)

## Examples

Examples of perfect fields are:

• every field of characteristic zero, so $\displaystyle{ \mathbb{Q} }$ and every finite extension, and $\displaystyle{ \mathbb{C} }$;[2]
• every finite field $\displaystyle{ \mathbb{F}_q }$;[3]
• every algebraically closed field;
• the union of a set of perfect fields totally ordered by extension;
• fields algebraic over a perfect field.

Most fields that are encountered in practice are perfect. The imperfect case arises mainly in algebraic geometry in characteristic p > 0. Every imperfect field is necessarily transcendental over its prime subfield (the minimal subfield), because the latter is perfect. An example of an imperfect field is the field $\displaystyle{ \mathbf{F}_q(x) }$, since the Frobenius sends $\displaystyle{ x \mapsto x^p }$ and therefore it is not surjective. It embeds into the perfect field

$\displaystyle{ \mathbf{F}_q(x,x^{1/p},x^{1/p^2},\ldots) }$

called its perfection. Imperfect fields cause technical difficulties because irreducible polynomials can become reducible in the algebraic closure of the base field. For example,[4] consider $\displaystyle{ f(x,y) = x^p + ay^p \in k[x,y] }$ for $\displaystyle{ k }$ an imperfect field of characteristic $\displaystyle{ p }$ and a not a p-th power in f. Then in its algebraic closure $\displaystyle{ k^{\operatorname{alg}}[x,y] }$, the following equality holds:

$\displaystyle{ f(x,y) = (x + b y)^p , }$

where bp = a and such b exists in this algebraic closure. Geometrically, this means that $\displaystyle{ f }$ does not define an affine plane curve in $\displaystyle{ k[x,y] }$.

## Field extension over a perfect field

Any finitely generated field extension K over a perfect field k is separably generated, i.e. admits a separating transcendence base, that is, a transcendence base Γ such that K is separably algebraic over k(Γ).[5]

## Perfect closure and perfection

One of the equivalent conditions says that, in characteristic p, a field adjoined with all pr-th roots (r ≥ 1) is perfect; it is called the perfect closure of k and usually denoted by $\displaystyle{ k^{p^{-\infty}} }$.

The perfect closure can be used in a test for separability. More precisely, a commutative k-algebra A is separable if and only if $\displaystyle{ A \otimes_k k^{p^{-\infty}} }$ is reduced.[6]

In terms of universal properties, the perfect closure of a ring A of characteristic p is a perfect ring Ap of characteristic p together with a ring homomorphism u : AAp such that for any other perfect ring B of characteristic p with a homomorphism v : AB there is a unique homomorphism f : ApB such that v factors through u (i.e. v = fu). The perfect closure always exists; the proof involves "adjoining p-th roots of elements of A", similar to the case of fields.[7]

The perfection of a ring A of characteristic p is the dual notion (though this term is sometimes used for the perfect closure). In other words, the perfection R(A) of A is a perfect ring of characteristic p together with a map θ : R(A) → A such that for any perfect ring B of characteristic p equipped with a map φ : BA, there is a unique map f : BR(A) such that φ factors through θ (i.e. φ = θf). The perfection of A may be constructed as follows. Consider the projective system

$\displaystyle{ \cdots\rightarrow A\rightarrow A\rightarrow A\rightarrow\cdots }$

where the transition maps are the Frobenius endomorphism. The inverse limit of this system is R(A) and consists of sequences (x0, x1, ... ) of elements of A such that $\displaystyle{ x_{i+1}^p=x_i }$ for all i. The map θ : R(A) → A sends (xi) to x0.[8]

3. Any finite field of order q may be denoted $\displaystyle{ \mathbf{F}_{q} }$, where q = pk for some prime p and positive integer k.