An equation may have no solution in the field in which its coefficients lie. Rather than simply declaring a new symbol to be a root, we build a field containing that root and ask how much larger it is. Linear algebra measures the enlargement, while polynomial factorization tells us which construction is valid.
Why enlarge a field
Three small examples show the phenomenon this unit studies. The polynomial has no root in , because squares of real numbers are nonnegative. The polynomial has no root in , because is not the square of a rational number. The polynomial has no root in the two-element field , because its values at and are both . In each case we would like a larger field in which the polynomial does have a root, and we would like to describe that larger field completely enough to compute in it.
Two questions organize everything that follows. The first is existence: given a polynomial over , can we construct a field containing a root, and what does look like? The second is measurement: how large is over ? The answers are tied together by a single idea. A root of a polynomial over satisfies an algebraic relation, and the lowest-degree relation, the minimal polynomial, controls both the arithmetic of and its dimension over .
Every claim below depends on the base field. A polynomial that is irreducible over one field may factor over a larger one, and an element that is algebraic of degree over one field may have degree over another. We therefore always state the base field explicitly.
Algebraic elements and minimal polynomials
Definition 1.1.1
(Extension and degree)
An extension consists of a field containing as a subfield with the same identity. Its degree is the dimension of as a vector space over ; it may be infinite. For , is the smallest subfield of containing and . The element is algebraic over if a nonzero polynomial in vanishes at ; otherwise it is transcendental.
Definition 1.1.2
(Irreducible polynomial)
A polynomial of positive degree is irreducible over if it cannot be written as with both of positive degree. Irreducibility is a statement about a specific field : the polynomial is irreducible over but factors as over .
When is algebraic, choose a monic polynomial of least positive degree that vanishes at it, denoted . Division by shows that every polynomial vanishing at is a multiple of it. Consequently is unique and irreducible: a nontrivial factorization would force a lower-degree factor to vanish at . Conversely, a monic irreducible polynomial vanishing at must equal .
The last sentence is the working tool of the whole unit. To find a minimal polynomial we do not need to search through all polynomials of lower degree. We exhibit one monic polynomial that vanishes at and prove that it is irreducible over the base field; it is then automatically the minimal polynomial. Indeed, divides it, and an irreducible monic polynomial has no monic divisors other than and itself.
Theorem 1.1.3
(A simple algebraic extension)
If has degree , evaluation induces an isomorphism , and is an -basis. Hence .
Proof
(Proof)
The evaluation homomorphism has kernel . Because is irreducible, the quotient is a field: Bézout's identity gives an inverse modulo to each polynomial not divisible by it. Its image is therefore a field containing and , and is exactly . Division gives representatives of degree below ; their values span. A nontrivial linear relation among them would be a smaller-degree polynomial vanishing at , so they are independent.
The same argument constructs a root from an irreducible : the residue class of in is a root. Irreducibility matters; if with positive-degree factors then their residue classes are nonzero zero divisors, so the quotient is not a field.
Worked example 1.1.4
(An explicit inverse in a quadratic extension)
Over , the polynomial is irreducible because is irrational. To invert , observe the polynomial identity
Evaluation at gives
This is precisely the Bézout construction with . The inverse remains a polynomial expression in the adjoined element.
Constructing fields from irreducible polynomials
The construction deserves to be stated as a recipe. Let be irreducible of degree and put and . The map sending to the class of the constant polynomial is injective, because a nonzero constant is not a multiple of a polynomial of positive degree, so we may regard as a subfield of . Then in , the minimal polynomial divides , so irreducibility makes the monic associate of , and by the theorem above with basis . Computation in means reducing polynomials in using the single relation .
Worked example 1.1.5
(A field with eight elements)
Over consider . It has degree , so it is irreducible exactly when it has no root in . We compute and in . Neither value is zero, so is irreducible, and is a field with basis over . Since each coordinate is or , has exactly elements.
The relation is , because in characteristic . We compute successive powers: , , , , and . The displayed reduced forms are pairwise distinct, hence are all seven nonzero elements.
Inverses can be read off. From we get , which agrees with . To invert , write with . Expanding and using gives . Setting this equal to forces , and , so . Hence . Check: .
The example illustrates why irreducibility, and not merely the absence of a convenient root, is the correct hypothesis. In the same ring construction applied to over we have , so the class of is nonzero but squares to zero, and the quotient is not even a domain.
Theorem 1.1.6
(Kronecker's theorem: every polynomial acquires a root)
Let have positive degree. There is an extension with in which has a root.
Proof
(Pass to an irreducible factor)
The monic divisors of of positive degree form a nonempty set, since it contains divided by its leading coefficient. Choose one, , of least degree. It is irreducible: a factorization with both factors of positive degree would make , divided by its leading coefficient, a monic positive-degree divisor of of smaller degree. The recipe above makes a field with , containing the root of . Since divides , also .
Applying the theorem again to the quotient over , and repeating, produces an extension in which splits into linear factors; this is the starting point of the splitting-field unit. The construction is also unique in the following sense. If and , in two extensions of , have the same minimal polynomial over , then the simple-extension theorem gives isomorphisms that fix and send to . The embeddings unit develops this observation.
Proposition 1.1.7
(A finite-dimensional domain is a field)
Let be a commutative ring that contains the field as a subring with the same identity, is finite-dimensional as an -vector space, and has no zero divisors. Then is a field. In particular, if is algebraic over , the ring of polynomial expressions in is already the field .
Proof
(Multiplication by a nonzero element is a bijection)
Let be nonzero. The map is -linear, and it is injective because forces in a ring without zero divisors. An injective linear map from a finite-dimensional space to itself is surjective, so some satisfies . For the second claim, lies inside a field, so it has no zero divisors, and it is spanned by with , because division by reduces every polynomial expression. Hence is a field containing and . Every subfield containing and contains , so .
This proposition gives a second proof that is a field, without Bézout's identity. The hypothesis "finite-dimensional" cannot be dropped: the polynomial ring has no zero divisors but is not a field, since has no inverse. The next theorem shows that this is exactly what happens for a transcendental element.
Theorem 1.1.8
(Simple transcendental extensions)
If is transcendental over , evaluation gives an isomorphism , which extends to an isomorphism from the rational function field onto . In this case and .
Proof
(Proof)
The evaluation homomorphism has zero kernel by the definition of transcendence, so it is an isomorphism onto its image . Send , where , to ; the denominator is nonzero because would make algebraic. If , then , and evaluating at shows that the two images agree, so the map is well defined. It preserves sums and products because evaluation does, it sends to , and therefore it is injective, as every field homomorphism is. Its image is a subfield containing and , and every such subfield contains all quotients ; so the image is . Next, : otherwise for some , and would be a nonzero polynomial vanishing at . Finally, the powers are linearly independent over , again by transcendence, so the degree is infinite.
For instance, Lindemann proved in 1882 that is transcendental over , so . Thus every simple extension is, up to isomorphism, either for a monic irreducible or the rational function field .
Towers and practical tests
Theorem 1.1.9
(Tower law)
For fields with both and finite, .
Proof
(Proof of the tower law)
Let be an -basis of and a -basis of . Expanding coefficients twice shows that the products span over . If an -linear combination of them is zero, collect terms by ; independence of the gives a zero -linear combination of the for each , so all coefficients vanish.
Proof X-ray
Why the product basis is independent
The tower proof uses independence twice, at different levels. A relation among first groups into coefficients in ; independence of the forces each grouped coefficient to vanish in . Only then may independence of the over force the original coefficients to vanish. Expanding the coefficients proves spanning, but it does not by itself prove independence.
The tower law also holds with infinite degrees, reading a product that involves as : is finite if and only if both and are finite. Indeed, if is finite, then is an -subspace of , so , and a finite -basis of also spans over the larger field , so . The converse is the theorem itself.
A polynomial of degree two or three over a field is irreducible precisely when it has no root in that field. This criterion does not extend to degree four: has no real root but is reducible over . For rational polynomials, the rational-root test is useful in degrees two and three; in higher degrees one must also rule out factors of intermediate degree. Eisenstein's criterion gives another test: for a primitive integer polynomial, a prime dividing every nonleading coefficient, dividing the constant coefficient only once, and not dividing the leading coefficient proves irreducibility over .
Worked example 1.1.10
(Two square roots)
The polynomial has no rational root, so . The element is not in : if with , squaring gives and . If , would be a rational square; if , would be a rational square. Both contradict unique prime factorization. Thus remains irreducible over , and the tower law yields degree for . A basis is .
Common mistake
(Degree depends on the base field)
The minimal polynomial of is over , but is over . Likewise, adjoining two quadratic roots does not automatically multiply the degree by four: adjoining after adds nothing.
Why the practical tests are correct
The tests quoted above are used constantly, so we record why they hold and exactly which facts they depend on.
Theorem 1.1.11
(Proposition: low degrees and the rational-root test)
Let have degree or . Then is irreducible over if and only if has no root in . Moreover, if with and is a root in lowest terms, then divides and divides .
Proof
(Proof of the proposition)
If has a root , division by gives with , so is reducible. Conversely suppose with and . Then one factor has degree exactly , because two factors of degree at least would give total degree at least . A degree-one polynomial with has the root , which is also a root of .
For the rational-root test, multiply by to get . Every term except is divisible by , so divides ; since , divides . Every term except is divisible by , so divides and hence .
Eisenstein's criterion is more delicate because it passes between and . We use Gauss's lemma, proved in the Gauss's lemma unit: if a primitive polynomial in factors in into two polynomials of positive degree, then it factors in into two polynomials of the same two degrees. We also use that has unique factorization, which follows from the division algorithm over a field. The same unit proves the criterion over any unique factorization domain, such as .
Theorem 1.1.12
(Eisenstein's criterion)
Let be primitive with . Suppose a prime satisfies , for , and . Then is irreducible over .
Proof
(Proof of Eisenstein's criterion)
Suppose is reducible over . By Gauss's lemma, with of degrees and . Reduce modulo . The hypotheses give with , so in . By unique factorization, each of is a nonzero constant times a power of . The leading coefficients of and multiply to , which does not divide, so neither leading coefficient vanishes mod and the degrees are preserved: and . Since , the constant terms and are divisible by , hence divides . This contradicts the hypothesis.
Proof X-ray
Where each Eisenstein hypothesis is used
Gauss's lemma moves a rational factorization into , where reduction modulo makes sense. The condition for makes a single monomial, and unique factorization in then forces and to be monomials too. The condition keeps their degrees equal to , so both reductions have zero constant term; without it a factor could collapse to a constant modulo . Finally closes the argument, because two constant terms each divisible by would put into .
A related sufficient test is reduction modulo a prime. If is monic and its reduction modulo is irreducible over , then is irreducible over . Indeed, a factorization over yields, by Gauss's lemma, a factorization into monic integer polynomials of the same degrees; monic factors keep their degrees modulo , producing a nontrivial factorization of the reduction. For example, is irreducible over : it has no root since its values at and equal , and a quartic without roots could only factor as a product of two irreducible quadratics. The only irreducible quadratic over is , whose square is . Hence is irreducible over as well. The converse of this test is false, as the counterexamples below show.
Common mistake
(The reduction test needs the degree to survive)
The reduction test is stated for monic polynomials, or more generally for leading coefficients not divisible by . Without this, it fails: is reducible over , but its reduction modulo is , which is irreducible over . The factor became the constant , so the reduction no longer sees the factorization.
Degrees of elements inside a finite extension
The tower law has an immediate and powerful consequence, which also supplies the second main theorem of this unit.
Theorem 1.1.13
(Proposition: degrees divide the ambient degree)
Let be a finite extension and . Then is algebraic over , and the degree of its minimal polynomial divides . Consequently, if is prime and , then .
Proof
(Proof of the divisibility statement)
Algebraicity follows from a dimension count: if , the powers are linearly dependent over . Now , and both steps are finite because is an -subspace of the finite-dimensional space and is finite-dimensional over its subfield as well. By the simple-extension theorem, . The tower law gives , so divides . If is prime and , then (a degree-one minimal polynomial would put in ), so , giving . A subspace of a finite-dimensional space with the same dimension is the whole space, so .
This proposition converts arithmetic about degrees into non-membership statements with no computation. The root has degree over by Eisenstein's criterion at applied to . It therefore cannot lie in , whose degree over is , nor in any extension of of degree , because does not divide . Similarly does not lie in , since does not divide . Compare the earlier worked example, where proving required a hands-on argument because both degrees equal and divisibility gave no information.
A primitive element computation
Sometimes an extension generated by two elements is generated by a single one. Here we verify this for and find the minimal polynomial explicitly.
Worked example 1.1.14
(The element )
Let . Squaring gives , so , that is, . Thus is a root of the monic polynomial .
To determine its degree we show that equals . One inclusion is clear. For the other, compute , using and . Then , so and . Hence , which has degree over by the earlier worked example.
By the simple-extension theorem, . The monic quartic vanishes at , so it is a monic multiple of of the same degree, and therefore equals . In particular is irreducible over , even though it has no rational root and, as a polynomial in , looks as if it might factor. This is a case where irreducibility was deduced from a degree computation rather than used to obtain one.
Counterexamples and the limits of the criteria
Each criterion above has hypotheses that cannot be dropped, and it is worth seeing precisely how each fails when they are.
Counterexample mode
The root test stops at degree three
Claim: a polynomial with no root in a field is irreducible over that field. Over take . For real we have , so it has no rational root. Yet is a product of two rational quadratics; over the simpler behaves the same way. The proof of the low-degree test used that a factorization of total degree two or three has a linear factor; in degree four two quadratic factors avoid this. The correct statement is the proposition as stated, for degrees two and three; from degree four onward one must also rule out factors of intermediate degree.
-
Eisenstein is sufficient, not necessary, and may need a substitution. Eisenstein does not apply directly to because its middle coefficients are zero but the constant term is , not divisible by any prime. Yet satisfies the criterion at , so is irreducible over . The failure of a direct application therefore proves nothing about reducibility.
-
Reduction modulo gives a one-way test. The polynomial is irreducible over since is not a rational square. Modulo we have , so is reducible over . Irreducibility of the reduction implies irreducibility over for monic polynomials, but not conversely.
-
Irreducibility depends on the field. The polynomial is irreducible over and splits over . The degree-one factors are exactly the minimal polynomials of over that larger base.
-
Transcendental elements and infinite degree. Let be the field of rational functions in an indeterminate over . The element is transcendental over : a nonzero polynomial evaluates at to the same expression viewed in , which is a nonzero polynomial in . Consequently are linearly independent over , and . The theorem on simple algebraic extensions does not apply to such , and no minimal polynomial exists.
Practice and solutions
Quick check
(Find the degree and a basis of .)
Solution · Answer
Eisenstein at proves irreducible over , so it is the minimal polynomial of . The degree is , with basis .
Quick check
(Why does a finite extension contain only elements algebraic over ?)
Solution · Answer
If , the elements are linearly dependent over for every . A nontrivial relation among them is a nonzero polynomial over vanishing at .
Quick check
(Find the minimal polynomial of over .)
Solution · Answer
Put . Then , so is a root of . The prime divides , does not divide the leading coefficient , and does not divide the constant term . By Eisenstein, is irreducible over ; being monic and vanishing at , it is , so has degree .
Quick check
(Is irreducible over ? Compare with .)
Solution · Answer
Over , , so is a root and the cubic is reducible: , since the right side expands to and . Over the same cubic has values at and , so it is irreducible there, as in the eight-element example. Irreducibility depends on the base field.
Quick check
(Prove that .)
Solution · Answer
If , then with and . The tower law gives , impossible because does not divide .
Quick check
(Compute the inverse of in as a rational combination of .)
Solution · Answer
Write and seek . Using , the product is , so , and . Then , and , giving . Check: .
Quick check
(Let be the class of in . Show that this ring is a field with four elements, and find and .)
Solution · Answer
The quadratic has value at both and , so it has no root in and is irreducible. The quotient is a field with basis , hence with the four elements , and . Since , we get and .
Quick check
(Let be prime and let . Show that and that has degree .)
Solution · Answer
Since , , and this degree divides by the divisibility proposition, so it equals . An -subspace of of full finite dimension is all of , so , and by the simple-extension theorem.
Quick check
(Is , the ring of polynomial expressions in with rational coefficients, a field?)
Solution · Answer
No. Since is transcendental over , the theorem on simple transcendental extensions identifies with the polynomial ring , and shows that . The field generated by is the strictly larger . Compare , which is already a field because is algebraic.
Exercise 1.1.15
(Find and give a -basis of this field.)
Core
Hint(1)
Two subfields of known degree give a lower bound by divisibility; one quadratic equation gives an upper bound.
Solution(3)
- 1
A lower bound from divisibility
Key step
Put . It contains , of degree , and , of degree . By the tower law both numbers divide , so divides it.
- 2
An upper bound from one equation
, and is a root of over . Its minimal polynomial over that field divides , so and .
- 3
Conclude and build a basis
Hence and , with basis over . The proof of the tower law turns this and the basis of into the -basis .
Exercise 1.1.16
(Show that is irreducible over but reducible over . If , find .)
Core
Hint(1)
Complete the square in a different way: .
Solution(4)
- 1
Irreducibility over the rationals
The counterexample list above shows, by Eisenstein's criterion after the substitution , that is irreducible over . Hence , since .
- 2
A factorization over the larger field
Key step
From we get over .
- 3
The degree of over
Since and , we have . The tower gives .
- 4
Conclude
is a root of the first quadratic factor, which is therefore its minimal polynomial over . The same quartic is irreducible over one base and a product of two quadratics over the other.