SIMPLE RADICAL EXTENSIONS

KEITH CONRAD

1. Introduction
A field extension L/K is called simple radical if L = K(α) where√αn = a √ for some n ≥ 1
×
and a ∈ K .√ Examples of simple radical extensions of Q are Q( 2), Q( 3 2), and more
√
generally Q( n√2). A root of T n − a will
√ be denoted n a, so a simple radical extension of K
looks like K( n a), but the notation n a in general fields is ambiguous: different nth roots
of a can generate different extensions of K, and they could even be nonisomorphic (e.g.,
have different degrees over K) if T n − a is reducible in K[T ].
Example 1.1. In C the three roots of T 3 − 8 are 2, 2ω, and 2ω 2 , where ω is a nontrivial
cube root of unity; note ω 2 = 1/ω and ω is a root of (T 3 − 1)/(T − 1) = T 2 + T + 1.
While Q(2) =√Q, the extension Q(2ω) = Q(ω) = Q(2/ω) has degree 2 over √Q, so when
the notation 3 8 denotes any √ of the roots
√ of T 3 − 8 over Q √
then the field Q( 3 8) has two
different meanings and R( 8) is R if 8 = 2 and it is C if 3 8 is 2ω or 2ω 2 .
3 3

√ √ √ √
Example 1.2. In the field Q( 5) the number 2 + 5 is a cube: 2 + 5 = ( 1+2 5 )3 . The
√ √
polynomial T 3 − (2 + 5) factors over Q( 5) as
√ ! √ √ !
3
√ 1+ 5 2 1+ 5 3+ 5
T − (2 + 5) = T − T + T+
2 2 2
√
and the second factor is irreducible over Q( 5) since it is irreducible over the larger field R
√ p3
√ √
(it is a quadratic with negative discriminant −3(3 + 5)/2). If 2 + 5 means (1 + 5)/2
p
3
√ √ √ p
3
√
then Q( 2 + 5) = Q((1 + 5)/2) =p Q( 5), and if 2 + 5 is a root of the quadratic
√ 3
√ √
factor of T 3 − (2 + 5) above then Q( 2 + 5) is a quadratic extension of Q( 5).
√
We will focus here on the degree [K( n a) : K] and irreducibility relations
√ for T n − a
n
among different values of n, and intermediate fields between K and K( a).
√
2. Basic properties of T n − a and n a
√
Theorem 2.1. The degree [K( n a) : K] is at most √ n, and it equals n if and only if T n − a
is irreducible over
√ K, in which case the field K( n a) up to isomorphism is independent of
n
the choice of a.
√ √
Proof. Since n a is a root of T√n − a, which is in K[T ], the minimal polynomial of n a over
K is at most
√ n, and thus [K( n a) : K] ≤ n. √
If [K( n a) : K] = n then the minimal polynomial of n a over √ K has degree n, so it must
be T n − a since that polynomial has degree n in K[T ] with n a as a root. As a minimal
polynomial in K[T ] for some number, T n − a is irreducible over K.
1
2 KEITH CONRAD

√
Conversely, assume T n − a is irreducible over K. Then n a has minimal polynomial
T n − a over K (the minimal polynomial of a number over√K is the unique monic irreducible
polynomial in K[T ] with that number as a root), √ so [K( n a) : K] = deg(T n − a) = n.
When T − a√is irreducible over K, the field K( a) is isomorphic to K[T ]/(T n − a)√using
n n

evaluation at n a and thus, up √ to isomorphism (not up to equality!), the field K( n a) is

n
independent of the choice of a. 
√
Example
√ 2.2. The√ polynomial T 3 − 2 is irreducible over Q √ and the three fields Q( 3 2),
3 3 2 3
Q( 2ω), and Q( 2ω ) are isomorphic to each other, where 2 is the real cube root of 2
(or any cube root of 2 in characteristic 0) and ω is a nontrivial cube root of unity. This is
no longer true if we replace Q by R, since T 3 − 2 has one root in R.
√
Theorem 2.3. The roots of T n − a in a splitting field over K are numbers of the form ζ n a
where ζ is an nth root of unity (ζ n = 1) in K.
√
Proof. Set α = n a, which is a fixed choice of root of T n − a over K. If β is another root
in a splitting field
√ of T n − a over K then β n = a = αn , so (β/α)n = 1. Set ζ = β/α ∈ K,
so β = ζα = ζ n a and ζ n = (β/α)n = 1. √ √
Conversely, if ζ n = 1 and ζ ∈ K then (ζ n a)n = ζ n a = a, so ζ n a is a root of T n − a in
K. 

3. Prime exponents
In degree greater than 3, lack of roots ordinarily does not imply irreducibility. Consider
(T 2 −2)(T 2 −3) in Q[T ]. The polynomial T p −a, where the exponent is prime, is a surprising
counterexample: for these polynomials lack of a root is equivalent to irreducibility.
Theorem 3.1. For an arbitrary field K and prime number p, and a ∈ K × , T p − a is
irreducible in K[T ] if and only if it has no root in K. Equivalently, T p − a is reducible in
K[T ] if and only if it has a root in K.
Proof. Clearly if T p − a is irreducible in K[T ] then it has no root in K (since its degree is
greater than 1).
In order to prove that T p − a not having a root in K implies it is irreducible we will prove
the contrapositive: if T p − a is reducible in K[T ] then it has a root in K.
Write T p − a = g(T )h(T ) in K[T ] where m = deg g satisfies 1 ≤ m ≤ p − 1. Since T p − a
is monic the leading coefficients of g and h multiply to 1, so by rescaling (which doesn’t
change degrees) we may assume g is monic and thus h is √ monic.
Let L be a splitting field of T p − a over K and α = p a be one root of T p − a in L. Its
other roots in L are ζα where ζ p = 1 (Theorem 2.3), so in L[T ]
T p − a = (T − ζ1 α)(T − ζ2 α) · · · (T − ζp α)
where ζip = 1. (Possibly ζi = ζj when i 6= j; whether or not this happens doesn’t matter.)
By unique factorization in L[T ], every monic factor of T p − a in L[T ] is a product of some
number of (T − ζi α)’s. Therefore
(3.1) g(T ) = (T − ζi1 α)(T − ζi2 α) · · · (T − ζim α)
for some pth roots of unity ζi1 , . . . , ζim .
Now let’s look at the constant terms in (3.1). Set c = g(0), so
c = (−1)m (ζi1 · · · ζim )αm .
 SIMPLE RADICAL EXTENSIONS 3

Since g(T ) ∈ K[T ], c ∈ K and c 6= 0 on account of g(0)h(0) = 0p − a = −a. Therefore

(3.2) c = (−1)m (ζi1 · · · ζim )αm ∈ K × .
We want to replace αm with α, and will do this by raising αm to an additional power to
make the exponent on α congruent to 1 mod p.
Because p is prime and 1 ≤ m ≤ p − 1, m and p are relatively prime: we can write
mx + py = 1 for some x and y in Z. Raise the product in (3.2) to the x-power to make the
exponent on α equal to mx = 1 − py:
cx = (−1)mx (ζi1 · · · ζim )x αmx
= (−1)mx (ζi1 · · · ζim )x α1−py
α
= (−1)mx (ζi1 · · · ζim )x p y
(α )
α
= (−1)mx (ζi1 · · · ζim )x y ,
a
so
(ζi1 · · · ζim )x α = ay (−1)mx cx ∈ K ×
and the left side has the form ζα where ζ p = 1, so K contains a root of T p − a. 
Remark 3.2. For an odd prime p and any field K, the irreducibility of Tp − a over K
r
implies irreducibility of T p − a for all r ≥ 1, which is not obvious! And this doesn’t quite
r
work when p = 2: irreducibility of T 4 − a implies irreducibility of T 2 − a for all r ≥ 2
(again, not obvious!), but irreducibility of T 2 − a need not imply irreducibility of T 4 − a. A
basic example is that T 2 + 4 is irreducible in Q[T ] but T 4 + 4 = (T 2 + 2T + 2)(T 2 − 2T + 2).
See [2, pp. 297–298] for a precise irreducibility criterion for T n − a over any field, which
is due to Vahlen [4] in 1895 for K = Q, Capelli [1] in 1897 for K of characteristic 0, and
Rédei [3] in 1959 for positive characteristic.

4. Irreducibility relations among T n − a for different exponents

Theorem 4.1. Let K be a field, a ∈ K × , and assume T √
n − a is irreducible over K. If d | n
d
then T − a is irreducible over K. Equivalently, if [K( n a) : K] = n for some nth root of a
√
over K then for all d | n we have [K( a) : K] = d for every dth root of a.
d

Proof. We prove irreducibility of T n −a implies irreducibility of T d −a in two ways: working

with polynomials and working with field extensions.
Polynomials: assume T d − a is reducible over K, so T d − a = g(T )h(T ) where 0 <
deg g(T ) < d. Replacing T with T n/d in this equation, we get T n − a = g(T n/d )h(T n/d )
where deg g(T n/d ) = (n/d)
√ deg g < (n/d)d = n and clearly deg g(T √
n/d ) > 0.
n n
Field extensions: let a be an nth root of a over K, so [K( a) : K] = n by Theorem
√ √ n/d √ d √ n/d √ n
2.1. Define d a = n a . This is a root of T d − a since √
d
a = ( n a )d = n a = a.√To
prove T d − a is irreducible
√ over K √we will prove [K(√ d
a) : K] = d using that√ of d a.
choice √
In the tower K ⊂ K( a) ⊂ K( a), we have [K( a) : K] ≤ d and [K( a) : K( a)] ≤
d n d n d
√ √ √
n/d√by Theorem 2.1, since d a is a root of T d − a ∈ K[T ] and n a is a root of T n/d − d a ∈
K( d a)[T ]. We have
√ √ √ √
[K( n a) : K] = [K( n a) : K( d a)][K( d a) : K]
and our irreducibility hypothesis implies the left side is n, so it follows that our upper
√ bounds
n/d and d for the factors on the right must be equalities. In particular, [K( d a) : K] = d
so T d − a is irreducible over K (it has a root with degree d over K). 
4 KEITH CONRAD

There was an important calculation in this proof that we will use repeatedly below: if
√ √ √ √ n/d
d | n then K( n a) contains K( d a), where
√
d
a := n
a . This is a root of T d − a, so the
notation is reasonable, but
√ note that d a is not an arbitrary dth root of a: it depends on
n
the choice made first of a.
By Theorem 4.1 and Remark 3.2, for odd primes p irreducibility of T p − a is equivalent
r
to irreducibility of T p − a for any single r ≥ 1, and for the prime 2 irreducibility of T 4 − a
r
is equivalent to irreducibility of T 2 − a for any single r ≥ 2.
Theorem 4.2. For relatively prime positive integers m and n, T mn − a is irreducible over
K if and only if T m −a and T n −a are each √irreducible over K. Equivalently, if√m and n are
relatively√prime positive integers then [K( mn a) : K] = mn if and only if [K( m a) : K] = m
and [K( n a) : K] = n.
Proof. That irreducibility of T mn − a over K implies irreducibility of T m − a and T n − a
over K follows from Theorem 4.1.
To prove irreducibility of T m − a and T n − a over K implies irreducibility of T mn − a
over K we will work with roots of these polynomials. It is convenient √ to select mth, nth,
and mnth roots√of a in√a multiplicatively
√ compatible
√ m way:
√ fix a root mn
a of T mn −√a over
n m
m
K and define a := mn n
a and a := mn
a . Then a is a root of T − a and n a is a
m

root n − a, so we have the √

√ of T √
m n mn
following field diagram, where the containments are due to
a and a being powers of a.

√
K( mn a)
≤n ≤m

√ √
K( m a) K( n a)

m n

K
The bottom field degree values come from T m n
√−a and T −a beingn irreducible
√ over
√ K, and
the top√field degree
√ upper bounds come√ from mn
a being a root of T − m
a ∈ K( m
a)[T ] and
m
T − a ∈ K( a)[T ]. Let d = [K( a) : K], so by reading the field diagram along either
n n mn

the left or right we have d ≤ mn . Also d is divisible by m and by n since field degrees are
multiplicative in towers, so from relative primality of m and n we get m | d, n | d =⇒ mn | d,
√
so mn ≤ d . Thus d = mn, so T mn − a is the minimal polynomial of mn a over K and thus
is irreducible over K. 
Corollary 4.3. For an integer N > 1 with prime factorization pe11 · · · pekk , T N − a is irre-
ei
ducible over K if and only if each T pi − a is irreducible over K.
Proof. Use Theorem 4.2 with the factorization N = pe11 (pe22 · · · pekk ) to see irreducibility of
e1 e2 ek
T N − a over K is equivalent to irreducibility of T p1 − a and T p2 ···pk − a over K, and
then by induction on the number of different prime powers in the degree, irreducibility of
e2 ek ei
T p2 ···pk − a over K is equivalent to irreducibility of T pi − a over K for i = 2, . . . , k. 
Example 4.4. Irreducibility of T 90 − a over K is equivalent to irreducibility of T 2 − a,
T 9 − a, and T 5 − a over K.
 SIMPLE RADICAL EXTENSIONS 5

Remark 4.5. By Remark 3.2, if N is odd then irreducibility of T N − a over K is equivalent

to irreducibility of T pi − a over K as pi runs over the prime factors of N (the multiplicities
ei don’t matter!), and for these we know the story for irreducibility by Theorem 3.1: it’s
the same thing as T pi − a not having a root in K for each pi .
Example 4.6. Irreducibility of T 75 − a over K is equivalent to a not having a cube root
or fifth root in K.

5. Intermediate fields in a simple radical extension

√ √ √ n/d
For
√ a choice of nth root n a and a factor d | n, d a := n a is a root of T d − a in
n
K( a), so we have the following field diagram.
√
K( n a)

√
K( d a)

K
√ √
It’s natural to ask if every field between K and K( n a) is K( d a) for some d dividing n.
The simplest setting to study this is when T n −√ a is irreducible over K
√ (and thus also T d − a
d d
is irreducible over K, by Theorem
√ 4.1), so [K( a) : K] = d. Is K( a) the only extension
of K of degree d inside K( n a)? This is not always true.
√ √
Example 5.1. Let K = Q and consider the field Q( 4 −1). Set α = 4 −1, so α4 + 1 = 0.
The polynomial T 4 + 1 is irreducible
√ over Q because it becomes Eisenstein at 2 when√ T is
replaced with T + 1. Since [Q( 4 −1) :√Q] = 4, any field strictly between Q and Q( 4 −1) is
quadratic over Q. One of these is Q( −1), but it is not the only one.
√
Q( 4 −1)

√ √ √
Q( 2) Q( −1) Q( −2)

If α4 = −1 then (α + 1/α)2 = α2 + 2 + 1/α√ 2 = (α4 + 1)/α2 + 2 = 2 and (α − 1/α)2 =

√ √
2 2 4 2
α − 2 + 1/α = (α√+ 1)/α − 2√= −2, so Q( 4 −1) contains Q( 2) and Q( −2). None of
the fields Q(i), Q( 2), and Q( −2) are the same, so we have√ at least three (and in fact
there are just these three) quadratic extensions of Q in Q( 4 −1).
In the above
√ example, the “reason”
√ for the appearance of more intermediate√
fields between
4 4
Q and Q( −1) than √ just Q( −1) is that there are 4th roots of unity in Q( −1) that are
not in Q, namely ± −1. The following theorem shows we get no such unexpected fields if
all nth roots of unity in the top field are actually in the base field.
6 KEITH CONRAD

Theorem 5.2. Let K be a√field, a ∈ K × , and assume T n − a is irreducible over K. If

all nth roots of unity in K( n a) are in K then for each d | n the only field between K and
√ √ √ √ n/d
K( n a) of degree d over K is K( d a), where d a := n a .
√
Proof. Every field between K and K( n √ a) has degree over K that divides n. For√d | n
suppose L is a field
√ with K ⊂ L ⊂ K( n
K] = d. To prove L = K(√d a), it
a) and [L : √
suffices to show a ∈ L, since that would give us K( d a) ⊂ L and we know K( d a) has
d
√
degree d over K, so the containment K( d a) ⊂ L would have to be an equality.
√
K( n a)
n/d

L
d

K
√
Let f (T ) be the minimal polynomial of √n
a over L, so f (T ) | (T n − a) and deg f = n/d.
We can write any other root of f (T ) as ζ n a for some nth root of unity Q ζ. (Theorem √
2.3). In a splitting field of T n − a over K, the factorization of f (T ) is i∈I (T − ζi n a)
for some nth roots of unity ζi (I is just an index set). The constant term of f (T ) is in
Q √ n/d Q √ n/d √ Q √
L, so ( i∈I ζi ) n a ∈ L. Therefore ( i∈I ζi ) n a ∈ K( n a), so ζ i ∈ K( n
a). The
√ Q i∈I
only nth roots of unity in K( a) are, by hypothesis, in K, so i∈I ζi ∈ K ⊂ L. Therefore
n

√n n/d √
a = d a is in L, so we’re done. 
n √
Example 5.3. If K = Q, a > 0, and T − a is irreducible over Q then Q( a) is isomorphic
n

to a subfield√of R (using the real positive nth root of a), which implies the only roots of
unity in Q( n a) are ±1 and those both lie Q. For example, the only fields between Q and
√ √ √ √ n/d
Q( n 2) are Q( d 2) where d | n and d 2 = n 2 .
Example 5.4. Let F be a field and K = F (u), the rational functions over F in one
indeterminate.
√ The polynomial T n − u is irreducible over F (u) since it is Eisenstein at u.
n
We let u denote one root of T n − u, so K(√ n
u) = F (
√n
u) has degree n over
√ √ F (u). All
roots of unity in F ( n u) – not just nth roots of unity – are √ in F , because F ( n u) is itself
a rational function field in one indeterminate over F (since n u is transcendental over F )
and all elements of a rational function field in one indeterminate over F that are not in F
are transcendental over F and √ thus can’t√ be a root of unity. Therefore by Theorem 5.2, the
fields between F (u) and F ( n u) are F ( d u) for d | n.
√
Example 5.5. An example where the hypothesis that all nth roots of unity in K( n a) are
in K is false, yet the conclusion
√ of Theorem 5.2 is true, is K = Q(i), a = 2, and n√= 8: it
8
can be√ shown that [Q(i, 2) : Q(i)] = 8 and the only fields between
√ Q(i) and Q(i, 8 2) are
Q(i, 2) for d = 1, 2, 4, 8 while 1+i
d 8
√ is an 8th root of unity in Q( 2, i) that is not in Q(i).
2

References
[1] A. Capelli, Sulla riduttibilità delle equazioni algebriche, Nota prima, Rend. Accad. Sci. Fis. Mat. Soc.
Napoli 3 (1897), 243–252.
[2] S. Lang, “Algebra,” 3rd revised ed., Springer-Verlag, New York, 2002.
[3] L. Rédei, “Algebra,” Erster Teil, Akad. Verlagsges, Geest & Portig K.-G., Leipzig, 1959.
[4] K. T. Vahlen, Über reductible Binome, Acta Math. 19 (1895), 195–198.