Which integers live inside a field?
In a field , the symbol usually means . This notation hides a choice: in some fields that sum is zero. Before working with derivatives, finite fields or irreducibility, we need to know what integer coefficients mean in the chosen field. The answer is controlled by the smallest subfield of , called its prime subfield.
Throughout, a field has and a field homomorphism preserves . Write for repeated addition if , its additive inverse for negative , and for . The map
preserves addition and multiplication. It need not be injective: in , for instance, .
Definition 1.4.1
(Characteristic and prime subfield)
The characteristic of is the least positive integer for which , if one exists; otherwise it is zero. The prime subfield is the intersection of all subfields of .
The intersection really is a subfield: it contains , and sums, differences, products and inverses of its nonzero members belong to every subfield in the family. The family is nonempty because it contains itself.
Why a positive characteristic is prime
Theorem 1.4.2
(Only zero or a prime can occur)
The characteristic of a field is either zero or a prime number . More precisely, the kernel of is in characteristic zero and in characteristic .
Proof
(Use division of integers and the absence of zero divisors)
If no positive integer maps to zero, neither does a negative integer, so the kernel is zero. Otherwise let be the least positive integer in the kernel. Since , we have . Divide any integer in the kernel by : , with . Then ; minimality forces . Thus the kernel is .
If with , then . A field has no zero divisors, so one factor is zero, contradicting minimality of . Hence is prime.
The field hypothesis matters.
Counterexample mode
Zero divisors allow a composite characteristic
Claim: in any commutative ring with , the least positive with , when it exists, is prime. In the multiples of are , so the least such is six. The proof breaks exactly at the zero-divisor step: although and . This ring is not a field. The repaired statement needs only that the ring has no zero divisors, so it holds in every field.
Notice also that implies for every , since .
The smallest subfield
Theorem 1.4.3
(Classification of prime subfields)
If , the prime subfield of is canonically isomorphic to . If , it is canonically isomorphic to .
Proof
(Construct the subfield and prove minimality)
In characteristic , the kernel calculation gives an injective map , sending to . The image is closed under addition and multiplication and contains . For , the integer is not divisible by the prime , so . The Bézout identity gives ; its image is therefore closed under inverses of nonzero elements. Thus the image is a subfield. Every subfield of contains , all its integer multiples, and therefore the image. Hence the image is precisely the intersection of all subfields.
In characteristic zero, define for integers and nonzero . The denominator is nonzero. If , then , and cross multiplication in shows the images agree. For example, the image of is the sum of the images, because the nonzero elements and have inverses and their product is . The multiplication and unit identities follow similarly. A zero image forces , hence ; thus the map is injective. Its image is a subfield. Every subfield contains all integer multiples of and inverses of their nonzero members, so it contains that image. Again the image is the prime subfield.
"Canonical" means these maps are forced by the requirement that map to ; they do not require a choice of generator. Indeed, preservation of addition fixes every integer image, and preservation of multiplicative inverses fixes every rational image when the characteristic is zero. In positive characteristic, every residue class is represented by an integer, so its image is forced as well. Prime subfields of fields of the same characteristic are canonically isomorphic by the maps fixing ; this identifies them without treating elements of different ambient fields as literally equal. Field homomorphisms preserve characteristic: if is such a map, it is injective because its kernel is a proper ideal in a field. Consequently exactly when .
Proof X-ray
Do not divide by a vanished integer
In characteristic , the coefficient cannot be interpreted in : has no inverse. Nevertheless, can be interpreted when , by multiplying by the inverse of . For example becomes in , but has no such interpretation. There is no unital field homomorphism .
Proposition 1.4.4
(Every homomorphism fixes the prime subfield)
Let be a field homomorphism. Then for every integer , and in characteristic zero for all integers and . In particular, every automorphism of is the identity on the prime subfield of , and is therefore linear over it. The fields and have no automorphism other than the identity.
Proof
(Additivity fixes the integers, inverses fix the fractions)
Since and preserves sums, induction gives for ; preservation of additive inverses extends this to negative . In characteristic zero, for , and a field homomorphism sends the inverse of a nonzero element to the inverse of , because . Hence sends each quotient to the corresponding quotient in . When , these formulas say that fixes every element of the prime subfield, as described by the classification theorem. Then for every in the prime subfield, which is linearity. The prime subfield of or of is the whole field, so its only automorphism is the identity.
Worked example 1.4.5
(The field has only the identity automorphism)
Let be the real cube root of and . By the simple-extension theorem of the introductory unit, every element of is a polynomial expression in with rational coefficients. An automorphism of fixes by the proposition, so it is determined by . Applying to gives . But lies in , and is strictly increasing on , so is the only real cube root of . Hence , and is the identity. We assumed only that is a field automorphism; that it fixes was automatic.
Characteristic does not determine size
Worked example 1.4.6
(An infinite field of positive characteristic)
The rational function field consists of fractions with and . It has characteristic two because . Its elements are distinct: their differences are nonzero polynomials, and the polynomial ring embeds in its fraction field. Thus positive characteristic does not imply finiteness.
On the other hand, if is finite of characteristic , it is a vector space over its prime subfield . Choose a basis ; every element has a unique expression with . There are exactly independent choices for each of the coordinates, and uniqueness prevents two choices from representing the same element. Consequently . This is a necessary shape for the size, not by itself a construction of a field of each such size. The companion finite-field chapter supplies existence and uniqueness.
Worked example 1.4.7
(Equal characteristic is necessary, not sufficient, for an embedding)
There is no unital embedding , although both fields have characteristic two. If one existed, would be a finite-dimensional vector space over its image, so its size would be for some positive integer . But is not a power of . The prime subfield embeds in both; sharing it does not force either larger field to embed in the other.
Concrete arithmetic in a quadratic finite field
Worked example 1.4.8
(A field with nine elements and an inverse)
Over , the polynomial has no root: the squares of are , respectively. A reducible quadratic over a field has a linear factor and hence a root, so is irreducible. Put and let be the residue of . The quotient by an irreducible polynomial is a field, and . Division by gives each element a unique representative with . Therefore is a basis over , and the field has elements.
For a concrete division, , so . The field has nine elements but characteristic three: , and neither nor is zero. Its prime subfield consists of , not all nine elements.
The calculation uses two distinct ideas. Irreducibility makes the quotient a field; counting representatives then gives its size. A quotient by an arbitrary quadratic need not be a field. For instance, in the nonzero classes of and multiply to zero.
Common mistake
(The nine-element field is not )
The ring has nine elements, but with , so it has zero divisors and is not a field; consistently, the additive order of its is nine, which is not prime. The field above has characteristic three and is a two-dimensional vector space over . A finite field of order with is never the ring of integers modulo .
What changes when a coefficient vanishes?
We will use one direction of the repeated-root criterion: if is a repeated root of , write over a field containing . The product rule gives , so . Therefore, finding no common root of and rules out repeated roots, in any characteristic.
Worked example 1.4.9
(A polynomial and its derivative in two characteristics)
Over , the polynomial has derivative . Its roots in a splitting field are distinct: a common root of and would have to be zero, but . Over , the same integer coefficients give
All three roots counted with multiplicity are now the single element . There is no contradiction: the coefficients have been interpreted in different fields. In particular, the formal derivative of a nonconstant polynomial can be zero in positive characteristic.
The same issue explains why dividing by an integer requires care. In characteristic , a denominator has an inverse exactly when its residue modulo is nonzero. One may reduce a fraction presented with such a denominator, but there is no unital ring homomorphism : it would send to . A fraction with a denominator divisible by may admit another presentation after cancellation—for example —so failure of one presentation is not a statement about every possible presentation of the same rational number.
When is reduction of a fraction well defined?
Suppose has characteristic . Consider a rational number that has a presentation with . If it also has a presentation with , then in the integers. The same equality holds after mapping the integers into , and both denominators have nonzero images. Multiplying by their inverses shows that the two presentations give the same element of . Thus the reduction does not depend on which permitted presentation we choose.
To decide whether a rational number has a permitted presentation, first write it in lowest terms . If , that presentation works. If , no presentation with can represent it: would imply because and , contradicting . For instance, can be reduced modulo five and gives , even though its first displayed denominator vanishes modulo five. In contrast, has no permitted presentation. This is a reduction rule on certain rational numbers, not a homomorphism defined on all of .
The Frobenius map: injective need not mean onto
Let have characteristic . The binomial theorem explains a new kind of arithmetic.
Theorem 1.4.10
(The Frobenius endomorphism)
Let be a field of characteristic . For all ,
Consequently
is an injective unital field homomorphism, called the Frobenius map, and it is the identity on the prime subfield.
Proof
(The middle binomial coefficients vanish)
For , the prime divides : from and , Euclid's lemma gives the divisibility. Thus each middle coefficient is zero in , which proves the displayed identity. Together with and , it shows that is a unital field homomorphism. It is injective: if , then , and a field has no nonzero nilpotent elements. On the prime subfield is the identity, since it fixes and hence every sum of copies of , and in characteristic these sums make up the prime subfield.
Common mistake
(The collapse happens for the exponent , not for every exponent)
In characteristic the identity holds for , and by iterating Frobenius for , but not for other exponents. In , , while . The middle coefficient is not divisible by three, so it does not vanish.
Worked example 1.4.11
(An injective map that misses an element)
In , Frobenius is not surjective: is not a th power. Suppose instead that , with nonzero . Clearing denominators gives . Taking polynomial degrees gives , impossible modulo . Thus injectivity alone does not make an endomorphism of an infinite field an automorphism.
By contrast, if is finite, an injective map from to itself is also surjective: the distinct images of all elements fill the entire finite set. Frobenius is then an automorphism. This conclusion uses finiteness, not merely positive characteristic.
Proof X-ray
Zero derivative and powers of the variable
For , its formal derivative is . Distinct exponents give distinct derivative terms, so exactly when every nonzero coefficient occurs at an exponent divisible by . Hence is a polynomial in . It does not follow in an arbitrary field that is a th power in : the coefficients would also need th roots in . For example, has zero derivative over , but cannot be a th power there, since its constant coefficient has no th root. Indeed, if , then , including when , contradicting the example. In a finite field every coefficient does have a th root, by surjectivity of Frobenius.
Practice with solutions
Quick check
(Determine the characteristic and prime subfield of .)
Solution · Answer
The field contains with its usual , so no positive integer multiple of is zero. Its characteristic is zero and its prime subfield is ; the element does not belong to that smallest subfield.
Exercise 1.4.12
(Prove that no field contains a subfield with six elements, without assuming a classification of finite fields.)
CoreProof
Hint(1)
A subfield is itself a field. Find its characteristic, then count by a basis.
Solution(3)
- 1
Suppose a six-element field exists
A subfield with six elements would be such a field. Since is finite, the map from cannot be injective, so is positive and hence a prime .
- 2
Count elements by a basis over the prime subfield
Key step
is a vector space over its prime subfield , finite-dimensional because is finite. With a basis of elements, unique coordinates give .
- 3
Compare with six
∎Six is , not a prime power, so no such exists, and no field can contain one as a subfield.
Quick check
(In a field of characteristic seven, compute the image of and explain why cannot be interpreted using the same rule.)
Solution · Answer
The image of is , and in because , so maps to . But , which has no inverse, and is already in lowest terms. The calculation takes place in the prime subfield of the given field.
Quick check
(Suppose is a finite extension, , and . Find and its prime subfield.)
Solution · Answer
A basis of over has two elements, giving elements. The common identity gives characteristic three, and its prime subfield is , not .
Quick check
(Explain why a unital homomorphism would contradict a dimension count.)
Solution · Answer
A unital field homomorphism is injective. Its image would make a vector space over a nine-element field, forcing for a positive integer , which is impossible. The shared characteristic is not enough.
Quick check
(In the nine-element field above, compute and .)
Solution · Answer
Using , we obtain . The product proves that the inverse is .
Quick check
(In characteristic five, expand and compute its formal derivative. Why is the zero derivative not evidence that the polynomial is constant?)
Solution · Answer
Each middle binomial coefficient with is divisible by five, so . Its derivative is . Nevertheless has degree five; in characteristic a zero derivative only says that the polynomial is a polynomial in .
Quick check
(A student proposes sending to and to under a unital field homomorphism . Are those two values individually compatible with inversion? Does the map exist?)
Solution · Answer
Yes: and in . But a unital homomorphism on all of would also have to send to an inverse of , which is impossible. Compatibility for selected denominators does not give a map on the entire field.
Exercise 1.4.13
(In characteristic seven, reduce if possible, and decide whether can be reduced by choosing another presentation.)
Core
Hint(1)
Write each rational number in lowest terms first.
Solution(4)
- 1
Put in lowest terms
, and , so this presentation is permitted even though vanishes modulo seven.
- 2
Invert the denominator in
, so , and the image is .
- 3
Test every presentation of
Key step
If with , then , so and hence by Euclid's lemma, a contradiction.
- 4
Conclude
reduces to in characteristic seven, while has no permitted presentation and cannot be reduced.
Quick check
(In characteristic , show that for all , including when .)
Solution · Answer
Additivity of Frobenius gives . For odd , . For , , and because . In both cases .
Exercise 1.4.14
(Let be a field of characteristic . Prove that is exactly the prime subfield of , and deduce Fermat's little theorem.)
CoreProof
Hint(1)
One inclusion is part of the Frobenius theorem. For the other, count the roots of .
Solution(3)
- 1
Every element of the prime subfield is a solution
Key step
The Frobenius map is the identity on the prime subfield, so each of its elements satisfies .
- 2
There are no other solutions
The solutions are the roots in of , a nonzero polynomial of degree . Over a field, a nonzero polynomial of degree has at most roots: each root splits off a factor by division, and the quotient, of degree , contains every other root because a field has no zero divisors. The prime subfield already supplies distinct roots, so there is no room for more.
- 3
Conclude
∎The solution set is the prime subfield. Taking gives for every integer , that is, , which is Fermat's little theorem.
Quick check
(Show that every automorphism of fixes the elements , and that the Frobenius map is an automorphism of different from the identity.)
Solution · Answer
The elements form the prime subfield , which every automorphism fixes by the proposition on homomorphisms. Frobenius is an automorphism of the finite field by the injectivity argument above. In the model with , it sends to , and because . So it is not the identity, although it fixes .