A field extension is a pair of fields and , together with an inclusion of into , written . The operations on agree with those inherited from . Familiar examples include the real numbers extending the rational numbers, and the complex numbers extending the real numbers. More generally, an injective field homomorphism allows its domain to be identified with a subfield of its codomain, giving an extension even when no literal inclusion was initially specified.
Field extensions provide a systematic way to enlarge a number system so that equations acquire solutions. They also reveal how those solutions are related, how much information an enlargement adds, and which constructions can be performed within it. Their central invariants include degree, algebraicity, separability, and groups of automorphisms. These ideas connect the classical theory of equations with number theory, algebraic geometry, and the arithmetic of finite fields.
Subfields, embeddings, and adjoining elements
A field is a system in which addition, subtraction, multiplication, and division by nonzero elements are defined and satisfy the usual field axioms. A subfield must contain the same multiplicative identity as the larger field and be closed under all these operations. Consequently, every field extension preserves characteristic: if , then the least positive integer satisfying , when such an integer exists, is the same in both fields. The notation denotes an extension, not a quotient of sets or algebraic structures.
Every field contains a smallest subfield, called its prime subfield. In characteristic zero this is isomorphic to ; in characteristic it is isomorphic to , where is a prime number. Thus every field can be regarded as an extension of one of these basic fields. The characteristic-zero and positive-characteristic settings share much of their formal theory, but differ sharply in the behavior of polynomial derivatives and repeated roots.
If is a subset of an extension field , the notation means the smallest subfield of containing both and . It consists of rational expressions in finitely many elements of , with coefficients in , whose denominators are nonzero. For a single element , the field is called a simple extension. For example,
This set is closed under inversion because, for a nonzero element,
The denominator cannot vanish unless .
A -embedding between extension fields is an injective field homomorphism that fixes every element of . These maps preserve all polynomial relations over . For instance, any -embedding of into must send to either or , since its image must still satisfy . Both choices define embeddings. Embeddings distinguish the abstract structure of an extension from a particular realization of it inside a larger field, a distinction essential when comparing fields generated by different roots of the same polynomial.
Algebraic elements and minimal polynomials
An element is algebraic over if some nonzero polynomial satisfies . Otherwise it is transcendental over . An extension is algebraic if every one of its elements is algebraic over the base field. These properties are relative to the base: a transcendental number over is nevertheless algebraic over any field containing it, because it satisfies the linear polynomial there.
For algebraic , there is a unique monic polynomial of least positive degree over that vanishes at . This is its minimal polynomial, denoted . It is irreducible: a factorization into two polynomials of smaller positive degree would force one factor to vanish at . Conversely, an irreducible polynomial having as a root becomes the minimal polynomial after division by its leading coefficient. Polynomial division shows that every polynomial over vanishing at is divisible by .
The evaluation homomorphism
has kernel generated by the minimal polynomial. It therefore induces an isomorphism
Because an irreducible polynomial generates a maximal ideal in , this quotient is a field. Hence : polynomial expressions already suffice, even though the definition of adjoining an element permits rational expressions. If the minimal polynomial has degree , every element has a unique expression
For example, is irreducible over : a reducible cubic over a field has a root in that field, while the rational root test excludes any rational root here. Thus has degree three over , and has basis . The quotient construction also supplies an abstract root without choosing a pre-existing ambient field: for any irreducible , the residue class of in is a root of . This construction is the fundamental mechanism for creating algebraic extensions.
Degree, bases, and towers
An extension field is a vector space over its subfield , with field multiplication supplying scalar multiplication. Its degree is
The extension is finite if this dimension is finite, and infinite otherwise. A basis therefore measures the extension in an additive, linear sense, even though the multiplication of basis elements records additional structure. For example, has degree two with basis , while has degree four with basis .
Every finite extension is algebraic. If , the elements are linearly dependent for each , producing a polynomial equation over . The converse fails: the field of all algebraic numbers over is algebraic but has infinite degree. A useful exact statement is that an algebraic extension generated by finitely many elements is finite. Equivalently, every algebraic extension is the union of its finite intermediate extensions.
For a tower , the tower law states
For finite degrees, if is a -basis of and is an -basis of , then the products form a -basis of . The same principle gives cardinal multiplication for infinite dimensions under the usual choice-based framework. In particular, the degree of every intermediate field in a finite extension divides the total degree.
The tower law also explains why adjoining independent-looking generators does not always multiply degrees. In
the second step has degree two because , so the total degree is four. By contrast, adjoining after adds nothing, since . More generally, the degree of an element can decrease after the base field is enlarged, because its minimal polynomial may factor over the larger base.
Algebraicity is transitive along towers. If and are algebraic, then is algebraic. For an element , only finitely many coefficients from occur in an equation satisfied by . Those coefficients generate a finite extension of , over which also generates a finite extension; the tower law completes the argument.
Transcendental extensions and transcendence degree
If is transcendental over , no nonzero polynomial relation over holds among its powers. The evaluation map from the polynomial ring to is therefore injective, and adjoining inverses gives
the field of rational functions in one indeterminate. Unlike a simple algebraic extension, has infinite degree: the elements are linearly independent over . A transcendental element can be an actual number or a formal symbol; the extension-theoretic behavior is identical.
A family is algebraically independent over if no nonzero polynomial over in finitely many distinct members of vanishes upon substitution. A maximal algebraically independent family is a transcendence basis. Its defining properties are that its members are algebraically independent and that is algebraic over . Transcendence bases exist using a maximality argument, and all have the same cardinality. This cardinality is the transcendence degree, written .
Transcendence degree and extension degree measure different kinds of enlargement. The field has transcendence degree one but infinite vector-space dimension over . An algebraic extension, however large, has transcendence degree zero. A finitely generated extension can always be separated into these two components: if , a maximal algebraically independent subfamily can be chosen from the generators, and is finite algebraic over the rational function field generated by that subfamily. Such an extension is called purely transcendental only when it is itself generated by algebraically independent elements, with no additional algebraic enlargement required.
A central geometric example is the function field of an irreducible algebraic curve. Over a field , consider the relation
and the fraction field of , assuming characteristic different from two. Here is transcendental over , while is algebraic of degree two over . The resulting function field therefore has transcendence degree one. More generally, for an integral algebraic variety of finite type over a field, the transcendence degree of its function field equals its dimension. This connects polynomial dependence among functions with the geometric number of independent parameters.
Splitting fields and algebraic closures
A splitting field of a polynomial is an extension generated over by all the roots of , in which factors completely into linear factors. It can be constructed by adjoining one root at a time and factoring the remaining polynomial over each successive field. Because only finitely many algebraic elements are adjoined, a splitting field of a single polynomial has finite degree over . Any two splitting fields of the same polynomial are isomorphic by an isomorphism fixing , although the isomorphism need not be unique.
For over , adjoining the real root does not produce a splitting field. The other roots are and , where is a primitive cube root of unity. The splitting field is
It has degree six over : the real cubic field does not contain the nonreal number , whose polynomial remains irreducible over that field. This example separates “having a root” from “containing every root.”
A field is algebraically closed if every nonconstant polynomial over it has a root in it, equivalently if every such polynomial splits completely. An algebraic closure of is an algebraic extension of that is algebraically closed. Every field has an algebraic closure, unique up to a -isomorphism. The qualification “up to isomorphism” matters: there is generally no distinguished choice of a closure or of an isomorphism between two choices.
The fundamental theorem of algebra makes an algebraic closure of , since . It does not make an algebraic closure of , because most complex numbers are transcendental over . Instead, the subfield consisting of all numbers algebraic over is an algebraic closure of . It is countable: there are countably many rational polynomials, and each nonzero polynomial has finitely many roots. Its algebraic closedness follows because roots of polynomials with algebraic coefficients remain algebraic over , by transitivity of algebraicity.
Separability and inseparability
An algebraic element is separable over if its minimal polynomial has distinct roots in an algebraic closure. An algebraic extension is separable if all its elements are separable. The practical test uses the formal derivative: a polynomial has no repeated roots exactly when
For an irreducible polynomial, repeated roots can occur only if , since otherwise irreducibility makes relatively prime to its lower-degree derivative. In characteristic zero, the derivative of a nonconstant irreducible polynomial is nonzero, so every algebraic extension is separable.
Positive characteristic permits a different behavior. In characteristic , the derivative of is zero. Let be transcendental over , and compare
The element has degree over , but its minimal polynomial has only one distinct root in an algebraic closure, since
Thus the extension has degree while admitting only one -embedding into an algebraic closure.
An algebraic extension is purely inseparable if every has some power in . Such an element has only one possible image under a -embedding into an algebraic closure. For a finite extension, the number of distinct -embeddings into an algebraic closure is its separable degree, denoted . The total degree factors as
where the inseparable degree is a power of in characteristic , and is one in characteristic zero.
A perfect field is a field over which every algebraic extension is separable. All characteristic-zero fields are perfect. In characteristic , perfection is equivalent to surjectivity of the map , called the Frobenius map. Finite fields are perfect because this map is injective and therefore surjective on a finite set; is not perfect.
Separability also controls how extensions can be generated. The primitive element theorem states that every finite separable extension has the form . Thus a field presented using many algebraic generators can often be encoded by one element and one minimal polynomial. Without separability, finite extensions need not be simple.
Normal extensions and Galois correspondence
An algebraic extension is normal if every irreducible polynomial over with a root in splits completely in . For finite extensions, normality is equivalent to being the splitting field of a polynomial over . It is also equivalent to the requirement that every -embedding of into an algebraic closure sends onto itself. Normality concerns the presence of all conjugate roots, whereas separability concerns whether those roots are distinct; neither property substitutes for the other.
An extension is Galois if it is both normal and separable. Its Galois group
consists of the field automorphisms of fixing . For a finite extension, the number of such automorphisms never exceeds , and equality holds exactly when the extension is Galois. For example, has two automorphisms, determined by . The real cubic extension has only the identity automorphism: its other conjugate roots do not lie in that real field.
The fundamental theorem of Galois theory gives an inclusion-reversing correspondence for a finite Galois extension:
where , is a subgroup of , and is the set of elements fixed by every member of . If and , then
Moreover, is Galois exactly when is a normal subgroup of , in which case its Galois group is the quotient group .
For the splitting field of , the Galois group is the symmetric group . Its subgroup of order three fixes the quadratic field ; its three subgroups of order two fix three conjugate cubic fields. The group structure therefore encodes the complete intermediate-field structure.
Infinite Galois extensions require a topological refinement. Their automorphism groups carry the Krull topology, and intermediate fields correspond to closed subgroups, not to arbitrary subgroups. This makes infinite Galois groups profinite and connects field theory with topology as well as group theory.
Extensions of finite fields
A finite field has elements for some prime and positive integer . Its characteristic is , and it is an -dimensional vector space over . Conversely, for every prime power , a field with elements exists and is unique up to isomorphism. It may be realized inside an algebraic closure of as the set of roots of
These roots are distinct because the derivative is , and they form a field because raising to the -th power preserves addition and multiplication in characteristic .
Fix a field . For every positive integer , there is, up to -isomorphism, a unique extension of degree , namely . Every finite extension of a finite field is Galois. Its Galois group is cyclic of order , generated by the -power Frobenius automorphism
The fixed elements of are exactly , and is the identity on .
The intermediate fields are completely determined by divisibility. Inside , a subfield containing has elements for a divisor , and there is exactly one such subfield for each divisor. Its elements satisfy . Thus can be embedded over into exactly when . This explicit classification is an especially transparent instance of Galois correspondence.
For computation, finite extensions are commonly represented as
where is irreducible of degree . Elements are coefficient vectors of length ; multiplication is polynomial multiplication followed by reduction modulo , and inversion can be performed with the extended Euclidean algorithm. Different irreducible polynomials give different coordinate representations of isomorphic fields.
Finite-field extensions are used in error-correcting codes and cryptographic constructions. Reed–Solomon codes evaluate polynomials at distinct field elements, so enlarging the field increases the available evaluation points. The Advanced Encryption Standard performs its byte arithmetic in a field with elements. These applications depend on exact finite-field operations, not on approximating real or complex arithmetic.
Trace, norm, and changes of base field
A finite extension converts multiplication into linear algebra. For each , multiplication by defines a -linear map
The field trace and field norm are
They lie in and are independent of the basis chosen to represent as a matrix. The trace is additive and -linear; the norm is multiplicative.
For a finite separable extension, these invariants can be expressed through its embeddings into an algebraic closure:
For , with a nonsquare rational number, this gives
These formulas explain the familiar conjugate-based inversion rule in quadratic fields. They also turn arithmetic questions into equations over the smaller field, including norm equations related to Pell-type equations.
Trace and norm behave transitively in a tower of finite extensions:
The trace pairing is nondegenerate exactly when is separable. For a basis , the determinant
is its discriminant. In number theory, discriminants connect extension structure to integral bases and the primes at which arithmetic behaves exceptionally.
Changing the base field is another important operation. If and lie in a common overfield and both contain , their compositum is the smallest field containing both. For finite ,
Equality holds when and are linearly disjoint over , meaning that -linearly independent elements of remain linearly independent over . This condition can be expressed through the injectivity of the multiplication map , using the tensor product. Merely having does not guarantee linear disjointness in general, although it does when is finite Galois. This distinction matters when combining extensions or extending the coefficient field of an equation.
Historical development and major applications
The origins of extension theory lie in the problem of solving polynomial equations. Sixteenth-century formulas for cubic and quartic equations involved successive adjunctions of radicals, though the abstract language of fields came much later. Joseph-Louis Lagrange's work on permutations of roots in 1770–1771 helped shift attention from formulas alone to the structural relationships among roots. Évariste Galois, in work from 1830–1832, connected the possibility of solving equations by radicals with groups of permutations preserving algebraic relations. His principal mathematical manuscripts were published by Joseph Liouville in 1846.
The abstract setting emerged during the nineteenth and early twentieth centuries. Richard Dedekind developed the language of fields in algebraic number theory, while Leopold Kronecker and Heinrich Weber contributed to the systematic treatment of algebraic quantities and abstract fields. Ernst Steinitz's 1910 paper Algebraische Theorie der Körper organized field theory around characteristic, algebraic and transcendental extensions, and separability. This framework made it possible to treat number fields, finite fields, and function fields as instances of a common theory rather than as unrelated constructions.
One classical application concerns straightedge-and-compass constructions. Starting from rational coordinates, each new intersection of lines and circles requires solving equations of degree at most two. A constructible real number therefore lies in a tower of quadratic extensions, and its degree over must be a power of two. The number , of degree three, cannot be constructed, proving the impossibility of doubling the cube. The angle of is not constructible: satisfies the irreducible cubic , ruling out a general straightedge-and-compass procedure for trisecting angles. Being of power-of-two degree is necessary but not sufficient by itself; the quadratic-tower condition is the stronger structural requirement.
Cyclotomic extensions explain regular polygon construction. If is a primitive -th root of unity, then has degree and Galois group isomorphic to the invertible residue classes modulo . Carl Friedrich Gauss established the constructibility of the regular 17-gon in 1796. The Gauss–Wantzel theorem characterizes constructible regular polygons: their numbers of sides are precisely times a product of distinct Fermat primes.
Modern applications place different emphasis on the same structures. Algebraic number theory studies finite extensions of through their rings of integers, prime decomposition, and local completions. Algebraic geometry uses function-field extensions to encode dominant rational maps, with finite extensions describing generically finite maps. Arithmetic geometry studies the action of absolute Galois groups on objects such as elliptic curves. Computational algebra uses explicit bases, minimal polynomials, and polynomial factorization algorithms to represent extensions and calculate their automorphisms and intermediate fields.