Counting quintic fields with genus number one

Kevin J. McGown Frank Thorne Amanda Tucker

arXiv:2006.12991v2 [math.NT] 27 Jul 2023

July 28, 2023

1 Introduction
The genus field of a number field K is defined to be the maximal extension K ∗ of K that is unramified at
all finite primes and is a compositum of the form Kk ∗ where k ∗ is absolutely abelian. The genus number is
defined as gK = [K ∗ : K]. It follows immediately from class field theory that gK divides the narrow class
number h+ K . See [MT23] for a slightly more detailed introduction and [Ish76] for a comprehensive account
of genus fields.
One may consider the density of genus number one fields among all number fields of a fixed degree
and signature, ordered by their discriminants. It essentially follows from a classical theorem of Gauss on
quadratic forms that 0% of quadratic fields have genus number one.1 On the other hand, McGown and
Tucker ([MT23]) proved that a positive proportion (roughly 96.23%) of cubic fields have genus number one.
Due to subtleties that arise in the quartic case, we will temporarily put degree four fields aside and consider
this problem in the quintic case.
Let F denote the collection of all quintic fields K, and G denote the collection of all quintic fields with
gK = 1. For i = 0, 1, 2, write F (i) and G (i) to denote the subsets of F and G, respectively, consisting of
fields with precisely i pairs of complex embeddings. Define N (i) (X) = #{K ∈ F (i) : | Disc(K)| ≤ X} and
(i)
Ng (X) = #{K ∈ G (i) : | Disc(K)| ≤ X}. Bhargava proved (see Theorem 1 of [Bha10], pg. 1559) that
N (X) ∼ C (i) X where
(i)

1
Y  240 if r = 0

(i) −2 −4 −5 1
C = (1 + p − p − p ) · 24 if r = 1
p 1

16 if r = 2 .
Our main result concerning counting genus number one fields is the following:
Theorem 1.
 
506874 Y p4 + p3 + 2p2 + 2p  1
Ng(i) (X) = C (i) 4 3 2
X + O(X 1− 400 +ε ) .
506875 p + p + 2p + 2p + 1
p≡1 (mod 5)

Corollary 2. The proportion of quintic fields with i pairs of complex embeddings having genus number one
equals
506874 Y p4 + p3 + 2p2 + 2p
.
506875 p + p3 + 2p2 + 2p + 1
4
p≡1 (mod 5)

Although the number above (which is independent of i and approximately equal to 0.999935) is quite
close to 1, this implies that a positive proportion of quintic fields have class number divisible by 5. (Note
that since gK divides the narrow class number and gK is a power of 5, we also have that gK divides the class
number.) In fact, our methods yield the following stronger result:
Gauss proves (in the language of quadratic forms) that when K is quadratic, gK = 2t−1 where t is the number
1 Indeed,

of prime divisors of Disc(K), and moreover, that gK in fact equals the 2-part of h+ K in that situation. It was Hasse who first
reproved Gauss’ result in the language of class field theory, by considering the genus field in the quadratic setting.

1
Theorem 3. Given any k ≥ 0, there is a positive proportion of quintic fields with genus number equal to
5k , and hence with class number divisible by 5k .
We also prove the following:
Theorem 4. The average of the genus number taken over all quintic fields of a given signature is finite. In
particular, we have
P
K∈G (i) gK
| Disc(K)|≤X 506879 Y p4 + p3 + 2p2 + 2p + 5
lim P = .
X→∞
K∈G (i) 1 506875 p4 + p3 + 2p2 + 2p + 1
p≡1 (mod 5)
| Disc(K)|≤X

Numerically, this constant is 1.00026 . . . , but the main point is that the limit exists.
Late in the preparation of this paper, we became aware of the work of Kim (see [Kim20]) on statistical
questions concerning the genus number in cyclic and dihedral extensions of prime degree, in which he says
“it seems very difficult to compute genus numbers of [S5 ] fields”. All of the results in this paper hold, with
essentially identical proofs, when restricted to S5 -quintic fields only. As such, our results establish that a
variety of statistics can indeed be computed for genus numbers of such fields.
Finally, we also show the following:
Theorem 5. A positive proportion of quintic fields with genus number one fail to be norm-Euclidean.

2 Counting quintic fields with specified local completions

Our proofs will apply two results on counting quintic fields. The first, essentially due to Ellenberg, Pierce,
and Wood [EPW17, Theorem 5.1], and building on Bhargava’s work [Bha10], is a result counting quintic
fields with a finite specified set of local conditions.
By a local condition Σp at a prime p, we mean some fixed subset Σp of the quintic étale algebras over
Qp , and we say that a quintic (number) field K satisfies Σp if K ⊗ Qp ∈ Σp . For squarefree e, by a set of
local conditions Σ (mod e) = (Σp )p|e , we mean a choice of local condition Σp for each prime p | e, and we
say that K satisfies Σ if it satisfies each of the Σp .
To avoid a technical complication, we assume, for each p | e other than p = 5, that Σp corresponds to a
‘splitting type’ – i.e., that Σp consists of those algebras Kv1 × · · · × Kvg for which g, each e(Kvi |Qp ), and
each f (Kvi |Qp ) take a prescribed value. We consider such an algebra to be ramified if any of the Kvi is
ramified, and we write e = e1 e2 or e = 5e1 e2 , where e1 and e2 are the products of those primes p 6= 5 for
which Σp consists of unramified or ramified algebras, respectively.
Define the function
m(p) = 1 + p−1 + 2p−2 + 2p−3 + p−4 ,
which is the total mass of all quintic étale extensions of Qp . Indeed, Bhargava (see Theorem 1.1 of [Bha07])
proves a generalization of Serre’s mass formula (Théorème 2 of [Ser78]) which, specialized to the quintic
case, yields
X 1 1
= m(p) .
Discp (K) # Aut(K)
[K:Qp ]=5 étale

For the condition that p not totally ramify, i.e.,

Σp = {[F : Qp ] = 5 étale | e(F/Qp ) < 5} ,

we have X 1 1
= 1 + p−1 + 2p−2 + 2p−3 ,
Discp (F ) # Aut(F )
F ∈Σp

whereas for the condition that p totally ramify the same sum is equal to p−4 , as originally proved by Serre.
The following theorem is adapted from a result of Ellenberg–Pierce–Wood (see Theorem 2.4 of [EPW17]).

2
Theorem 6. Let N (i) (X; Σ) be the number of quintic fields K with | Disc(K)| < X having i pairs of complex
embeddings, and satisfying a set of local conditions Σ (mod e) with the restriction described above. Then,
we have
 
1/2
N (i) (X; Σ) = C (i) (Σ)X + O e1 e2 X 79/80+ε + X 199/200+ε ,

where
Y
C (i) (Σ) = C (i) Cp (Σp ),
p
X 1 1
Cp (Σp ) = m(p)−1 ,
Discp (F ) # Aut(F )
F ∈Σp
−1
m(p) = 1 + p + 2p−2 + 2p−3 + p−4 .
Proof. The proof is essentially identical to that of [EPW17, Theorem 5.1]. The explanation given there is
quite thorough, and we will indicate only what needs to be changed.
As in [EPW17], the proof applies an inclusion-exclusion argument to points in the lattice VZ = Z40
studied in [Bha10]. For each squarefree integer q coprime to 5e, let Wq,e ⊂ VZ denote the set of elements
corresponding to quintic rings that are: nonmaximal at each prime dividing q; maximal and which satisfy
the desired local conditions at primes dividing 5e.
Write e1 for the product of primes p | e, excluding 5, for which Σp consists of unramified algebras, and
write e2 for the analogous product where Σp consists of ramified algebras.
Then, Wq,e is defined by congruence conditions modulo q 2 e1 e22 5k , for some positive integer k. This
contrasts to the situation in [EPW17], where no special conditions modulo 5 were imposed, and where the
conditions modulo e were all unramified. In the ramified case, the splitting type (15 ) is defined (mod e),
as described in [Bha10, Section 12], but the maximality condition is no longer automatic and this is defined
only (mod e2 ).
The remainder of the analysis remains essentially unchanged. The error term of [EPW17, (5.3)] includes
an error term q 2 δP e, and e must be replaced by e1 e22 as described above. (The 5k term contributes to the
implied constant, and we may ignore this contribution.) We now have δP ≪ e−1 2 , reflecting the fact that the
fields being counted are rare. Finally, in invoking [Bha10, (27)], it must be assumed that q 2 e1 e22 ≪ X 1/40 .
The ensuing analysis then remains valid, with identical or improved bounds on the error terms E1 , E2 , E3 , E4 .
−1/2
The restriction that q 2 e1 e22 ≪ X 1/40 proves to be the bottleneck, and choosing Q = X 1/80 e1 e−1 2 completes
the proof.
Remark 7. For a general set of local conditions Σ (mod e), analogous results hold with an undetermined
e-dependence in the error term. As the proof shows, this dependence can be computed in terms of a modulus
for which Wq,e can be defined by congruence conditions.
We will also apply the following complementary ‘tail estimate’, which will be contained in the forthcoming
article [BCT] by Bhargava, Cojocaru, and the second author:
Theorem 8. For any Y > 1, define
N5 (X, Y ) := # K quintic : | Disc(K)| ≤ X, q 2 | Disc(K) for some squarefree q > Y .

(1)
Then
39 X
N5 (X, Y ) ≪ε X 40 +ε + .
Y 1−ε

3 The genus theory of quintic fields

Let K be a non-cyclic quintic field. For p 6= 5, write k(p) to denote the unique quintic subfield of Q(ζp ) if
such a field exists and Q otherwise. Ishida proves that Kk(p)/K is a nontrivial unramified extension if and
only if p is totally ramified in K and p ≡ 1 (mod 5).

3
 We now consider the case of p = 5. Write k(5) to denote the unique quintic subfield of Q(ζ25 ). Ishida
shows that Kk(5)/K is a nontrivial unramified extension if and only if 5 is totally ramified and N (γ)4 ≡ 1
(mod 25) for all γ ∈ OK coprime to 5. We will refer to this latter congruence condition as condition (⋆).
Ultimately, Ishida proves the following result (see Equation 5.9 of [Ish76, Chapter 5], pg. 65).
Theorem 9 (Ishida). Let L be a quintic field. Let t denote the number of primes p such that p is totally
ramified in L and p ≡ 1 (mod 5), and add +1 to t if 5 is totally ramified in L and N (γ)4 ≡ 1 (mod 25) for
all γ ∈ OL coprime to 5. Then we have:
(
5t−1 if L is cyclic
gK =
5t if L is not cyclic

We wish to reformulate condition (⋆) in a manner suitable for our calculations. Suppose 5 is totally
ramified in K. Let p be the unique prime in K above 5. It is plain that the congruence condition above
involving the norm can be checked in the quintic extension Kp /Q5 of local fields. It is equivalent to requiring
that N (u)4 ≡ 1 (mod 25) for all u ∈ Op× . Since every element of Op× is a 4-th root of unity times an element
of the principal units U (1) = 1 + p, it suffices to check the congruence for all u ∈ U (1) .
Since 5 is totally ramified in Kp it is possible to choose a generating polynomial f (x) = x5 +a4 x4 +a3 x3 +
a2 x2 + a1 x + a0 that is Eisenstein at the prime 5. A simple calculation (given on pages 57–59 of [Ish76])
shows that the condition N (u)4 ≡ 1 (mod 25) for all u ∈ U (1) is equivalent to

a1 ≡ a2 ≡ a3 ≡ a4 + a0 ≡ 0 (mod 52 ) .

(In truth, Ishida gives one direction of this claim, but the argument is easily reversible.) There are 25 quintic
ramified extensions of Q5 . See [JR06, Ama71]. For each of these, we use the generating polynomial given in
the Jones–Roberts database (now incorporated into the LMFDB [LMF23]) to check condition (⋆). We find
that precisely 5 of these 25 extensions satisfy the condition. Generating polynomials for these extensions are
given by x5 − 5x4 + 5(1 + 5a) for 0 ≤ a ≤ 4. Moreover, these are precisely the Galois extensions. We mention
in passing that all of these extensions have discriminant 58 . We have thus proved the following two results.
Lemma 10. Suppose K is a non-cyclic quintic field where 5 is totally ramified. Let k(5) be the unique
quintic subfield of Q(ζ25 ). We have that Kk(5)/K is unramified if and only if K ⊗ Q5 is a Galois extension
of Q5 .
Lemma 11. The mass of all totally ramified quintic Galois extensions of Q5 is equal to
X 1 1 1
= 8.
Disc5 (K) # Aut(K) 5
[K:Q5 ]=5 Galois,
totally ramified

4 Proofs of Theorems 1, 3, 4, 5.
Recall that F denotes the collection of all quintic fields K, and that G denotes the collection of all quintic
fields with gK = 1.
Proof of Theorem 1. Theorem 9 and Lemma 10 establish that G, the set of non-cyclic quintic fields K with
genus number gK = 1, consists precisely of those K satisfying the following local conditions:
• No prime p ≡ 1 (mod 5) is totally ramified in K.
• Either 5 is not totally ramified in K, or K ⊗ Q5 is not a Galois field extension of Q5 .
We count these fields (and hence prove Theorem 1) by means of an inclusion-exclusion sieve, adapting
the approach of Belabas, Bhargava, and Pomerance in [BBP10]. Throughout this section, we ignore the
cyclic quintic fields; by Theorem 1.1 of [CDyDO02] there are ∼ cX 1/4 of them with discriminant bounded

4
by X, and hence they do not contribute to any of our asymptotics. Alternatively, one may exclude all of the
non-S5 quintic fields, of which there are ≪ X 39/40+ǫ by [BCT].
For each quintic field K, let f (K) denote the product of the primes p ≡ 0, 1 (mod 5) that are totally
ramified in K, with the additional condition for p = 5 that K ⊗ Q5 be a Galois field extension of Q5 . Then
(i)
G (i) is the set of all K in F (i) for which f (K) = 1. Recall that Ng (X) denotes the counting function for
(i)
G .
Let T denote the collection of all positive squarefree f whose prime divisors p all satisfy p ≡ 0, 1 (mod 5).
(i)
For each f ∈ T , write Σf for the set of local conditions specifying that f | f (K). Let Af be the set of fields
in G (i) satisfying the conditions Σf , and observe that
[
G (i) = F (i) \ Ap(i) .
p≡0,1 (mod 5)

Applying inclusion-exclusion it follows, for an arbitrary parameter Y < X, that

X
Ng(i) (X) = µ(f )N (i) (X; Σf ) (2)
f ∈T
X X
= µ(f )N (i) (X; Σf ) + µ(f )N (i) (X; Σf ).
f ∈T f ∈T
f ≤Y f >Y

As f 4 | Disc(K) for each K satisfying Σf , the sum over f > Y is handled by the tail estimate of Theorem
8. Each field K is counted with multiplicity at most d(Disc(K)) = O(X ε ) in this sum (where d(n) is the
number of positive divisors of n), so that we have
X
µ(f )N (i) (X; Σf ) ≪ X 39/40+ε + X 1+ε /Y. (3)
f ∈T
f >Y

Define m∗ (p) to be m(p) for p 6= 5, with m∗ (5) = 54 m(5). For each f ≤ Y , upon applying Theorem 6
with Lemma 11 we obtain
 
Y
N (i) (X; Σf ) = C (i) m∗ (p)−1 p−4  · X + O(X 199/200+ε ) ,
p|f
199 79 3
provided that Y ≤ X 200 − 80 =X 400 , so that
 
X X Y  
µ(f )N (i) (X; Σf ) = C (i) X  −m∗ (p)−1 p−4  + O Y X 199/200+ε . (4)
f ∈T f ∈T p|f
f ≤Y f ≤Y

In the main term we extend the sum over f ≤ Y to all f ∈ T , at the expense of an error term ≪ XY −3 ,
since each product over p | f has absolute value ≤ f −4 . We thus have that
 
X X Y  
µ(f )N (i) (X; Σf ) = C (i) X  −m∗ (p)−1 p−4  + O Y X 199/200+ε + XY −3 .
f ∈T f ∈T p|f
f ≤Y
1 1
Choosing Y = X 400 , we see that the error terms above and in (3) are all O(X 1− 400 +ε ). We therefore
have that
 
X Y  1

Ng(i) (X) = C (i) X  −m∗ (p)−1 p−4  + O X 1− 400 +ε

f ∈T p|f
 
1 Y  1

= C (i) X · 1 − 1 − m(p)−1 p−4 + O X 1− 400 +ε ,


8
5 m(5)
p≡1 (mod 5)

5
which is what we wanted to prove.
Proof of Theorem 3. For a parameter Z, let U denote any set of k primes p ≡ 0, 1 (mod 5) with p ≤ Z, and
write T (U ) for the set of squarefree f whose prime factors p all satisfy p ≡ 0, 1 (mod 5) and p 6∈ U .
We modify the proof of Theorem 1 by replacing T with T (U ), and by adding the condition throughout
that each p ∈ U totally ramify and that K ⊗ Q5 be Galois if 5 ∈ U . All such fields have genus number
gK = 5k , and by an identical argument the number of such fields K with | Disc(K)| ≤ X is
   
Y Y  1

C (i)  1 − m∗ (p)−1 p−4  ·  m∗ (p)−1 p−4  X + OZ X 1− 400 +ε .

p∈T (U) p∈U

(i) 1
Adding over all choices of U with each p ≤ Z, we obtain an expression of the form Ck (Z)X +OZ X 1− 400 +ε ,


(i)
for a sequence Ck (Z) which increases with Z and is bounded above by C (i) , and which therefore converges
(i)
to a fixed constant Ck .
This counts all quintic fields with | Disc(K)| < X, with the specified signature, and with gK = 5k – with
the exception of those for which any prime p > Z is totally ramified. By Theorem 8, the number of such
39
is ≪ X 40 +ε + X/Z 1−ǫ . Therefore, letting Z → ∞, we see that the total number of fields being counted is
(i)
Ck X + o(X).
Proof of Theorem 4. Let ω(n) denote the number of prime divisors of n, and again let T denote the collection
of all positive squarefree f whose prime divisors p all satisfy p ≡ 0, 1 (mod 5).
By construction we have gK = 5ω(f (K)) , with f (K) defined as before, and hence also
X
gK = 4ω(f ) ,
f |f (K)

so that in analogy with (2) we have

X X
gK = 4ω(f ) N (i) (X; Σf ).
K∈G (i) f ∈T
| Disc(K)|≤X

This sum is evaluated exactly as in the proof of Theorem 1, with −m∗ (p) replaced by 4m∗ (p) at every
occurrence. All of the error terms satisfy identical bounds up to a factor of O(X ǫ ) since 4ω(f ) ≪ X ǫ for
f ≤ X. We therefore conclude that
 
X
(i) 4 Y  1

1 + 4m(p)−1 p−4 + O X 1− 400 +ε ,


gK = C 1+ 8 X (5)
(i)
5 m(5)
K∈G p≡1 (mod 5)
| Disc(K)|≤X

as desired.
Proof of Theorem 5. Let Σ denote the local conditions that 2 is inert, 5 is inert, 7 is totally ramified, and no
prime p ≡ 1 (mod 5) is totally ramified. Observe that any quintic field K satisfying the conditions Σ must
have genus number one. By way of contradiction, suppose K is norm-Euclidean and satisfies the conditions
Σ. Let p denote the unique prime lying over 7. Then there exists α ∈ OK such that 4 ≡ α (mod p) with
|N (α)| < |N (p)| = 7. It follows that 2 ≡ 45 ≡ N (α) (mod 7) and therefore N (α) ∈ {2, −5}. We are forced
to conclude that either 2 or 5 is not inert, a contradiction. The result now follows from techniques similar
to the proof of Theorem 1.

6
Acknowledgments
We would like to thank Arul Shankar, Melanie Matchett Wood, and several anonymous referees for helpful
comments.
This work was supported by a grant from the Simons Foundation (No. 586594, F.T.) and an internal
research grant through California State University, Chico.

References
[Ama71] Shigeru Amano. Eisenstein equations of degree p in a p-adic field. J. Fac. Sci. Univ. Tokyo
Sect. IA Math., 18:1–21, 1971.
[BBP10] Karim Belabas, Manjul Bhargava, and Carl Pomerance. Error estimates for the Davenport-
Heilbronn theorems. Duke Math. J., 153(1):173–210, 2010.
[BCT] Manjul Bhargava, Alina Cojocaru, and Frank Thorne. Non-S5 quintic extensions of bounded
discriminant. In preparation.
[Bha07] Manjul Bhargava. Mass formulae for extensions of local fields, and conjectures on the density
of number field discriminants. Int. Math. Res. Not. IMRN, (17):Art. ID rnm052, 20, 2007.
[Bha10] Manjul Bhargava. The density of discriminants of quintic rings and fields. Ann. of Math. (2),
172(3):1559–1591, 2010.
[CDyDO02] Henri Cohen, Francisco Diaz y Diaz, and Michel Olivier. On the density of discriminants of
cyclic extensions of prime degree. J. Reine Angew. Math., 550:169–209, 2002.
[EPW17] Jordan Ellenberg, Lillian B. Pierce, and Melanie Matchett Wood. On ℓ-torsion in class groups
of number fields. Algebra Number Theory, 11(8):1739–1778, 2017.
[Ish76] Makoto Ishida. The genus fields of algebraic number fields. Lecture Notes in Mathematics, Vol.
555. Springer-Verlag, Berlin-New York, 1976.
[JR06] John W. Jones and David P. Roberts. A database of local fields. J. Symbolic Comput., 41(1):80–
97, 2006.
[Kim20] Henry H. Kim. Genus numbers of cyclic and dihedral extensions of prime degree. Acta Arith.,
192(3):289–300, 2020.
[LMF23] The LMFDB Collaboration. The L-functions and Modular Forms database.
http://www.lmfdb.org, 2023.
[MT23] Kevin McGown and Amanda Tucker. Statistics of genus numbers of cubic fields. Ann. Inst.
Fourier (Grenoble), to appear, 2023.
[Ser78] Jean-Pierre Serre. Une “formule de masse” pour les extensions totalement ramifiées de degré
donné d’un corps local. C. R. Acad. Sci. Paris Sér. A-B, 286(22):A1031–A1036, 1978.

Kevin J. McGown
Department of Mathematics and Statistics
California State University, Chico
Chico, CA 95929
E-mail address: kmcgown@csuchico.edu

Frank Thorne
Department of Mathematics
University of South Carolina

7
Columbia, SC 29208
E-mail address: thorne@math.sc.edu

Amanda Tucker
Department of Mathematics
University of Rochester
Rochester, NY 14627
E-mail address: amanda.tucker@rochester.edu