Constructing Dedekind Domains with

Prescribed Prime Factorizations and Class

Groups
George Lowther
January 8, 2011

This note is in response to a question asked by Pete L. Clark on Math-

Overflow. Actually, he asked two related questions, but I only consider the
first one here. In particular, given any countable abelian group G, I give a
construction of a Dedekind domain with finite residue fields and class group
G.

1 Two questions about finiteness of ideal classes

in abstract number rings
The following is as asked by Pete L. Clark:
Let us say that an abstract number ring is an integral domain R which is
not a field, and which has the “finite norms” property: for any nonzero ideal
Iof R, the quotient R/I is finite.
(I have taken to calling such rings abstract number rings and have some
vague ambitions of extending the usual algebraic number theory to this class
of rings. Note that they include the two basic rings Z and Fp [t] and are
closed under: localization, passage to an overring – i.e., a ring intermediate
between R and its field of fractions – completion, and taking integral closure
in a finite degree extension of the fraction field. In order to answer my
questions affirmatively one would have to know something about abstract
number rings which are not obtained from the two basic rings via any of the
above processes – if any!)

1
 Note that such a ring is necessarily Noetherian of dimension one, so it is
a Dedekind domain iff it is normal, and in any case its integral closure is a
Dedekind abstract number ring.

Question 1. Does there exist an integrally closed abstract number ring with
infinite Picard (= ideal class, here) group?

Question 2. Let R be a not-necessarilly integrally closed abstract number

ring with integral closure R̃. Suppose that the ideal class group of R̃ is fi-
nite. Consider the ideal class monoid ICM(R) of R, i.e., the quotient of the
monoid of nonzero ideals of R by the submonoid of principal ideals. (Note
that the group of units of ICM(R) is precisely the Picard group, but if R is not
integrally closed it will necessarily have non-invertible ideals so that Pic(R)
will not be all of ICM(R).) Can it be that ICM(R) is infinite?

2 An answer to Question 1
To answer Question 1: Yes, there do exist integrally closed abstract number
rings with infinite class group.
By factorization of ideals, for R to be an abstract number ring it is enough
that it is a Dedekind domain with finite residue field R/p at each prime
p. Theorem B of the paper mentioned by Hagen Knaf in his answer actu-
ally gives what you ask for (R. C. HEITMANN, PIDS WITH SPECIFIED
RESIDUE FIELDS, Duke Math. J. Volume 41, Number 3 (1974), 565-582).

Theorem B. Let G be a countable abelian torsion group. Then there is a

countable Dedekind domain of characteristic 0 whose class group is G, and
whose residue fields are those of the integers (i.e. one copy of Z/pZ for each
prime p).

As such rings have finite residue fields, this gives an integrally closed
abstract number ring with class group any countable torsion group you like.
We can do much better than this though. After thinking about your question
for a bit, I see how we can construct the following, so that all countable
abelian groups occur as the class group of such rings.

Theorem 1. For any countable abelian group G, there is a Dedekind domain

Z[X] ⊆ R ⊆ Q(X) with finite residue fields and class group isomorphic to
G.

2
 The condition that G is countable is necessary. If R is any Dedekind
domain then, for each K > 0, there can only be finitely many ideals a with
|R/a| ≤ K. In zero characteristic, this is because any such ideal must factor
into prime ideals lying over a rational prime p < K and, as there are only
finitely many prime ideals lying over p, there will only be finitely many such
ideals a. The same argument holds in non-zero characteristic q by looking
at the primes in Fq [X] instead of the rational primes. So, there can only be
countably many ideals a with a finite quotient R/a. If all residue fields of
R are finite, then it can only have countably many ideals and must have a
countable class group.
So, Theorem 1 shows that we need only look at over-rings of Z[X] to
find any feasible class group we like for Dedekind domains with finite residue
fields. It might be a bit surprising that this is the case but, in fact, Z[X]
is quite general. Given any characteristic zero Dedekind domain with finite
residue fields, we can exactly replicate its structure of ideals and principal
ideals in a Dedekind over-ring of Z[X]. We can also replicate the completions
at localizations of the primes. Let me write Id(R) for the set of fractional
ideals in a ring R.

Theorem 2. Let R be a characteristic zero Dedekind domain with finite

residue fields. Then, there is a Dedekind domain R0 with Z[X] ⊆ R0 ⊆ Q(X)
and a bijection π : Id(R) → Id(R0 ) satisfying

1. π(ab) = π(a)π(b).

2. π(a) is prime if and only if a is.

3. π(a) is principal if and only if a is.

4. If p ⊆ R is a nonzero prime then R̄p ∼ 0

= R̄π(p) .

In particular, the class groups are isomorphic, Cl(R) ∼

= Cl(R0 ).

Here, I am using Rp to denote the localization at prime ideal p, which

is a discrete valuation ring (DVR). Then, R̄p denotes it’s completion. As a
complete DVR with finite residue field, R̄p is compact.
Knowing the set of prime ideals p of a Dedekind domain together with
the completions R̄p up to isomorphism tells exactly how the rational primes
factor in that ring. If p is a prime ideal containing rational prime p, let e

3
be the ramification index of p in R̄p . Then, we can factor p as a product of
primes, Y
pR = pe(p) .
p3p

We can construct over-rings of Z[X] with any desired set of prime ideals p
and completions R̄p that we like, and with any principal ideal structure that
is possible in any Dedekind domain (restricting to the zero characteristic
case). In the following, all valuations are expressed additively (i.e., v(xy) =
v(x) + v(y)) and assumed to be normalized so that Im(v) = Z ∪ {∞}. I
denote the set of principal fractional ideals of a ring R by Prin(R).
Theorem 3. Let R1 , R2 , R3 , . . . be a sequence of compact DVRs with valua-
tions vi . Let H be a subgroup of G ≡ ⊕∞ i=1 Z. The following are equivalent.

1. There exists a characteristic zero Dedekind domain R whose nonzero

primes are pi (in some order), such that R̄pi ∼
= Ri and
( )
Y h(i)
Prin(R) = pi : h ∈ H . (1)
i

2. There exists a Dedekind domain R satisfying 1 with Z[X] ⊆ R ⊆ Q(X).

3. For each f ∈ Z[X], every finite sequence x1 ∈ R1 , . . . , xn ∈ Rn with
f (xi ) 6= 0 extends to an infinite sequence xi ∈ Ri (i ∈ N) such that
⊕i vi (f (xi )) is in H.
Theorem 2 follows directly from this, and Theorem 1 also follows, as I’ll
show in a bit. I’ll give a proof of Theorem 3. It is immediate that 1 follows
from 2. Also, it is not hard to show that 3 follows from 1.
Proof of 1 implies 3 in Theorem 3. I’ll regard R as a subset of its pi -adic
completion Ri . Suppose that f ∈ Z[X] and x1 ∈ R1 , . . . , xn ∈ Rn are such
that f (xi ) 6= 0. By independence of pi -adic valuations, there exists a sequence
rj ∈ R such that limj rj = xi for each i ≤ n (w.r.t. vi ). Taking x = rj for
some large enough j gives vi (f (x)) = vi (f (xi )) for i ≤ n. Set xi = x for all
i > n (regarded as an element of ki ).
Then, Rf (x) ∈ Prin(R) so, by 1, there is an h ∈ H with h(i) = vi (f (x)).
However, vi (f (x)) = vi (f (xi )) for each i.
It remains to show that 2 follows from 3 in Theorem 3, which I’ll do next.

4
3 Construction of the Dedekind domains
I’ll give a construction of an over-ring R of Z[X] with ideals satisfying state-
ment 3 of Theorem 3. The idea is that every Dedekind domain R can be
written as the intersection of its localizations Rp . Given a discrete valua-
tion v on a field k, I’ll use kv ≡ {x ∈ k : v(x) ≥ 0} to denote its DVR and
pv ≡ {x ∈ k : v(x) > 0} to denote its unique maximal ideal.
The idea is to construct R as the intersection of DVRs kvi for some set
of discrete valuations {vi : i ∈ I} on a field k. In general, the intersection of
DVRs need not be Dedekind but, using the condition of finite residue fields,
it does hold under reasonably mild conditions.
Lemma 4. Let {vi : i ∈ I} be a set of discrete valuations on a field k with
finite residue fields. Suppose that each nonzero x ∈ k satisfies vi (x) = 0 for
all but finitely many i. Then, the intersection of the valuation rings
\
R= k vi
i

is a Dedekind domain with finite residue fields. Supposing that R has field of
fractions k, the nonzero prime ideals of R are

pi = {x ∈ R : vi (x) > 0} = R ∩ pvi ,

and these are distinct if no two of the valuations are equivalent.

Proof. I’ll first show that the quotient R/a is a finite ring for each nonzero
deal a. Choosing a nonzero a ∈ a, the map R/Ra → R/a is onto, so it is
enough to show that R/Ra is finite. By the hypothesis of the T lemma, there
is a finite J ⊆ I with vi (a) = 0 for all i ∈ I \ J. Let RJ = j∈J kvj which,
being an intersection of finitely many DVRs, is a principal ideal domain
and, by construction, has finite residue fields. Consider the canonical map
R → RJ /RJ a. This has kernel Ra, so R/Ra is isomorphic to a subset of the
finite ring R/RJ a, and must be finite.
Given any nonzero ideal a of R, the map b → b/a ⊆ R/a gives a one-
to-one map between the ideals b of R containing a and the ideals of the
finite ring R/a. Hence, a is only contained in finitely many ideals, and R is
Noetherian. Also, for a nonzero prime p, R/p is a finite integral domain and
hence a field, so p is maximal. Finally, being an intersection of integrally
closed rings, R is integrally closed. So, R is Dedekind.

5
 Next, suppose that that R has field of fractions k. Then, the valuations
are determined by their restriction to R, so pi 6= pj whenever vi , vj are
not equivalent. As the valuations are non-trivial, pi is a nonzero prime.
Conversely, suppose that p is a nonzero prime ideal of R. Then, every
S nonzero
x ∈ p satisfies vi (x) > 0 for some i, as it is not a unit. So, p ⊆ i pi and, by
prime avoidance, p ⊆ pi for some i. Finally, as nonzero primes are maximal,
p = pi .
The idea behind constructing Z[X] ⊆ R ⊆ Q(X) is then as follows. Find
an infinite sequence of distinct discrete valuations vi such that Q̄(X)vi ∼
= Ri
and such that each nonzero f ∈ Q(X) satisfies vi (f ) = 0 for all but finitely
many i. And, constructing valuations on Q(X) is easy. Consider a compact
DVR R with field of fractions k and valuation v. Choosing x ∈ R, the
morphism θ : Q(X) → R taking X to x defines a valuation u = v ◦ θ. For θ
to be a well-defined morphism requires x to be transcendental over Q, and
for the completion Q̄(X)v to be isomorphic to R requires Q(x) to be dense
in R. There are infinitely (in fact, uncountably many) x ∈ R for which this
holds.
Lemma 5. Let R be a compact DVR of characteristic 0 and with field of
fractions k. Then, the set of x ∈ R transcendental over Q and such that
Q(x) is dense in k is itself dense in R.
Proof. Choose any x ∈ R. Let p be the characteristic of the residue field of
R and k be its field of fractions. Then k is a finite separable and normal
extension of the p-adic numbers, so that k = Qp (y) for some y ∈ R. This
implies that Q(y) is dense in k. Without loss of generality, it can be assumed
that y is not algebraic over Q(x), otherwise we could replace it by y + z for
some z ∈ Zp which is not algebraic over Q(x).
Choosing any x ∈ R, the sequence xr ≡ x + pr y ∈ R tends to x, so it
only needs to be shown that Q(xr ) is dense in R and xr is transcendental for
infinitely many r.
As there are only finitely many subextensions of Qp → k, we have
Qp (xr ) = Qp (xs ) for some arbitrarily large s > r. In that case, y =
(xr − xs )/(pr − ps ) is in Q(xr ). So, this is not an algebraic extension, and xr
is transcendental over Q. Also, Q(xr ) ⊇ Q(y) is dense in k.
I’ll now move on to the construction of the Dedekind over-ring R of Z[X]
satisfying condition 2 of Theorem 3. The idea is simple enough. We induc-
tively choose xi ∈ Ri defining valuations on Q(X). This should be done in

6
such a way that, for any nonzero f ∈ Z[X], we have vi (f ) = hi for some
h ∈ H. At the same time, we need to make sure that for any h ∈ H, there
will exist f Q(X) with vi (f ) = hi . The details of this construction do get
rather tricky though, so I’ll split it up into a few lemmas.
For the remainder of this section, let R1 , R2 , . . . be a sequence of compact
DVRs with fields of fractions k1 , k2 , . . .. Denote the valuation on Ri by vi
and the characteristic of its residue field by pi . Also, let H be a subgroup of
G ≡ ⊕∞ i=1 Z satisfying condition 3 of Theorem 3. The following lemma will
enable us to force every f ∈ Z[X] to satisfy vi (f ) = hi for some h ∈ H.
Q
Lemma 6. Suppose that x ∈ i Ri and g ∈ Z[X] satisfy vi (g(xi )) = 0 for all
but finitely many i. Then, given any finite S ⊆ N and irreducible Q f ∈ Z[X]
not dividing g and satisfying f (xi ) 6= 0 for i ∈ S, we can find y ∈ i Ri and
h ∈ H with the following properties.
1. yi = xi for all i ∈ S.
2. vi (g(yi )) = vi (g(xi )) for all i.
3. vi (f (yi )) = hi for all i.
Proof. Choose an integer a with f (a) 6= 0 and g(a) 6= 0. As f and g are
coprime, there exist b1 , b2 ∈ Z[X] with

(X − a)f b1 + gb2 = c (2)

for a nonzero integer c. Choose n ∈ N large enough that all i > n satisfy
i 6∈ S, vi (g(xi )) = 0 and pi - cf (a). For any i ≤ n with i 6∈ S, we are
free to replace xi by xi + pri for sufficiently large exponent r that vi (g(xi )) is
unchanged and, so, we may assume that f (xi ) 6= 0 for all i ≤ n.
Q condition 3 of Theorem 3 implies the existence of an h ∈ H and
Then
y ∈ i Ri with yi = xi for i ≤ n and vi (f (yi )) = hi for all i. Note that, for
any i > n, if vi (f (yi )) > 0 then (2) implies that vi (g(yi )) = 0 as required.
Also, for i > n, we have pi - f (a)g(a) so vi (f (a)) = vi (g(a)) = 0. So, if
vi (f (yi )) = 0 then we can replace yi by a to ensure that vi (g(yi )) = 0.
In the construction of R, we will also need to make sure that for every
h ∈ H there does exist some f = f1 /f2 with f1 , f2 ∈ Z[X] with vi (f (xi )) =
hi . This will require forcing the numerator and denominator f1 , f2 to be
irreducible and that ⊕i vi (f1 (xi )) and ⊕i vi (f2 (xi )) are in H. This is a bit
tricky, so I first prove the following lemma.

7
 Q
Lemma 7. Suppose that x ∈ i Ri and g ∈ Z[X] satisfy vi (g(xi )) = 0 for
all but finitely many i. Also, choose a finite set P of rational primes and
f ∈ Z[X] satisfying f (xi ) 6= 0 whenever pi ∈ P.
Then, there exists an irreducible f˜ ∈ Z[X] and y ∈ i Ri such that
Q

1. yi = xi and vi (f˜(xi )) = vi (f (xi )) for pi ∈ P.

2. vi (f˜(yi )) = 1 for pi 6∈ P.

3. vi (g(yi )) = vi (g(xi )) for all i.

Proof. Start by choosing an integer a > 0 such that g(a) and g(a + 1) are
both non-zero, and let P 0 ⊇ P be a finite set of primes containing every pi
for which vi (g(xi )) 6= 0 and all the primes dividing ag(a)g(a + 1).
Now choose a prime number q not in P 0 and not dividing a. The Chinese
remainder theorem enables us to construct a polynomial f1 of degree d > 0
with the following properties. First, f1 − f is a multiple of sufficiently high
powers of pi for pi ∈ P that vi (f1 (xi )) = vi (f (xi )). Next, f1 ≡ 1 mod pi for
pi ∈ P 0 \P, so that vi (f1 (xi )) = 0. Also, f1 ≡ X d +q mod q 2 . By Eisenstein’s
criterion, the polynomial

f˜ = uf1 + vq 2 (X − a)

is irreducible for any pair of integers u, v with q not dividing u. I’ll choose u
to be the product of the primes dividing f1 (a) and not in P 0 . Choosing v to
be the product of sufficiently high powers of primes in P 0 gives vi (f˜(xi )) =
vi (f1 (xi )) for pi ∈ P 0 . So, we can take yi = xi whenever pi ∈ P 0 .
Now, consider pi 6∈ P 0 , so that vi (g(xi )) = 0. As pi does not divide
g(a)g(a + 1), we also have vi (g(a)) = vi (g(a + 1)) = 0. Consider the case
where pi 6= q. If pi | f1 (a) then vi (f˜(a)) = vi (uf1 (a)) = 0 and we can
take yi = a. If pi divides f1 (a) then it also divides u by construction and
vi (f (a+1)) = vi (vq 2 ) = 0 and we can take yi = a+1. Lastly, consider pi = q.
Then f˜(a) ≡ uf1 (a) ≡ uan mod pi and as q - ua, we have vi (f˜(a)) = 0, so
we take yi = a.
Now, we can prove the following, which will enable us to construct R
in such a way that every h ∈ H can be written as hi = vi (f (xi )) for some
f ∈ Q(X).

8
Lemma 8. Suppose that xi ∈ Ri (i ∈ N) are such that xi is transcendental
over Q, Q(xi ) is dense in ki and the valuations f 7→ vi (f (xi )) on Q(X) are
distinct. Suppose also that g ∈ Z[X] satisfies vi (g(xi )) = 0 for all but finitely
many i. Q
Then, given any finite S ⊆ N and h ∈ H, we can find y ∈ i Ri , irre-
ducible f1 , f2 ∈ Z[X] and h1 , h2 ∈ H with the following properties.
1. yi = xi for all i ∈ S.

2. vi (g(yi )) = vi (g(xi )) for all i.

3. vi (f1 (xi )) = h1i and vi (f2 (xi )) = h2i for all i.

4. h = h1 − h2 .
Proof. Let P be the finite set of primes {pi } for i ∈ S. As Q(xi ) is dense in ki ,
the discrete valuation ui (f ) = vi (f (xi )) is normalized so that Im(ui ) = Z ∪
{∞}. By independence of valuations, there exists f = f˜1 /f˜2 for f˜1 , f˜2 ∈ Z[X]
with ui (f ) = hi for all pi ∈ P. Choosing any prime q not in P we can use
the Chinese remainder theorem to find f1 ∈ Z[X] of degree d > deg(g) such
that ui (f1 ) = ui (f˜1 ) for all i ∈ S and f1 ≡ X d + q mod q 2 . By Eisenstein’s Q
criterion, f1 is irreducible. Lemma 6 implies the existence of ỹ ∈ i Ri and
h1 ∈ H with ỹi = xi for pi ∈ P and vi (g(ỹi )) = vi (g(xi )), vi (f1 (ỹi )) = h1i for
all i.
Now let P 0 ⊇ P be a finite set of primes including all pi for which h1i 6= 0.
Setting f˜3 = af1 + bf˜2 for a a product of sufficiently high powers of primes
in P and b a product of sufficiently high powers of primes in P 0 \ P gives
vi (f˜3 (ỹi )) = vi (f˜2 (ỹi )) for piQ
∈ P and vi (f˜3 (ỹi )) = vi (f1 (ỹi )) for pi ∈ P 0 \ P.
Lemma 7 gives us y ∈ i Ri and irreducible f2 ∈ Z[X] such that yi =
ỹi , vi (f2 (yi )) = vi (f˜3 (ỹi )) for pi ∈ P 0 , vi (f2 (yi )) = 0 for pi 6∈ P 0 , and
vi (g(ỹi )f1 (ỹi )) = vi (g(xi )f1 (xi )) for all i. In particular, for pi 6∈ P 0 , we
have vi (g(ỹi )) = vi (f1 (ỹi )) = 0.
If we set h2i = vi (f2 (yi )) then this construction gives h2i = vi (f1 (ỹi )) = h1i
for pi 6∈ P and h2i = vi (f˜2 (xi )) = h1i − vi (f (xi )). In any case, we have
h2i = h1i − hi so h2 = h1 − h ∈ H as required.
Finally, we can put Lemmas 6 and 6 to give a construction of R. This
procedes inductively choosing xni ∈ Ri . At each step, we fix a finite set of
polynomials Sn for which the values of vi (f (xni )) (f ∈ Sn ) are unchanged by
any further steps of the construction, and are in H.

9
Proof of 3 implies 2 in Theorem 3. Let {g1 , g2 , . . .} be the set of irreducible
polynomials in Z[X] and {h1 , h2Q
, . . .} be a generating set for H.
Start by choosing any x0 ∈ i Ri and set S0 = ∅. Inductively apply the
n
Q
following steps. Suppose that we have constructed x ∈ i Ri and a finite
set of irreducible polynomials Sn ⊆ Z[X] such that i 7→ vi (f (xni )) is in H
for all f ∈ Sn . We also suppose that xi are transcendental over Q, Q(xi )
is dense in ki andQthat the valuations on Q(X) given by f 7→ vi (f (xi )) are
distinct. Let g = f ∈S f and apply the following steps.
1. If n = 2m is even,Q let r be minimal such that n+1 gr 6∈ Sn . Use Lemma 6 to
n+1
choose x ∈ i Ri and h ∈ H such that xi = xni whenever i ≤ n
or vi (g(xi )) 6= 0, and vi (g(xn+1
n
i )) = vi (g(xni )) = 0 otherwise, and such
that vi (gr (xn+1
i )) = hi for all i. It follows that vi (f (xn+1i )) = vi (f (xni ))
for all f ∈ Sn and all i, so we can take Sn+1 = Sn ∪ {gr }.

2. If n = 2m − 1 is odd, use Lemma 8 to choose xn+1 ∈ i Ri , h̃1 , h̃2 ∈ H

Q

with hm = h̃1 − h̃2 and irreducible f1 , f2 ∈ Z[X] with vi (f1 (xn+1 i )) = h̃1 ,
n+1 n+1 n
vi (f2 (xi )) = h̃2 . This can be done such that xi = xi whenever
n+1
i ≤ n or vi (g(xi )) 6= 0 and such that vi (g(xi )) = vi (g(xni )) for all
n

i. It follows that vi (f (xn+1

i )) = vi (f (xni )) for all f ∈ Sn and all i, so
we can take Sn+1 = Sn ∪ {f˜1 , f˜2 }. Then, hm is in the subgroup of H
generated by i 7→ vi (f (xn+1
i )) for f ∈ Sn+1 .
After applying either of these steps, for i > n we are free to replace xn+1 i by
n+1 n+1
any close enough approximation x̃i so that vi (f (xi )) is unchanged for all
f ∈ Sn+1 . By Lemma 5, we can do this to ensure that xn+1 i is transcendental
over Q and Q(xn+1 i ) is dense in k i . Similarly, replacing x n+1
i by xn+1
i + pri for
large enough exponent r ensures that the valuations f 7→ vi (f (xn+1 i )) remain
distinct.
Now, applying the Q induction above, we have xn+1 i = xni for all i ≤ n,
so we can define x ∈ i Ri by xi = xni for all n ≥ i. By construction,
the valuations ui on Q(X) defined by ui (f ) = vi (f (xi )) are all distinct and
Q̄(X)ui ∼ = Ri . Also, if f ∈ Sn for any n, then f ∈ Sm for all m ≥ n giving
ui (f ) = vi (f (xni )). T
We define the ring R = i Q(X)ui , which contains Z[X]. By Lemma
4, R is a Dedekind domain with distinct prime ideals pi = R ∩ pui , so that
ui is the pi -adic valuation. If f ∈ Z[X] then the first step of the induction
above Q implies that f ∈ Sn for some n, so that hi ≡ ui (f ) is in H and
Rf = i phi i . Conversely, if h ∈ H, the second step of the induction implies

10
that
Q hthere exists f1 , f2 ∈ Sn for some n such that ui (f1 ) − ui (f2 ) = hi . Then,
i pi = Rf1 /f2 is principal, as required.
i

4 Construction of domains with specified class

group
I’ll now use Theorem 3 to give a proof of Theorem 1 showing that every
countable abelian group occurs as the class group of an over-ring of Z[X].

Lemma 9. Let p1 , p2 , . . . be rational primes, so that each individual prime

only occurs finitely often. Let e1 , e2 , . . . be non-negative integers.
Then, there exist compact DVRs R1 , R2 , . . . where Ri contains pi and
has ramification index ei , such that each non-constant irreducible polynomial
f ∈ Z[X] has a root in the residue field of Ri for all large i.
Suppose that H is any subgroup of G ≡ ⊕i⊕N Z satisfying the following.

1. For each rational prime p, there is an element h ∈ H with hi = ei

whenever pi = p and hi = 0 otherwise.

2. For any g : N → N and finite set S ⊆ N, there is an h ∈ H with

h|S = g|S .

Then, H satisfies condition 3 of Theorem 3 and, in particular, the Dedekind

domain Z[X] ⊆ R ⊆ Q(X) given by condition 2 has class group Cl(R) ∼ =
G/H.

Proof. Choose non-negative integers fi and let Ri be a finite extension of

Qpi of degree ei fi and ramification index ei , so that the residue field of Ri
is a finite extension of the finite field Fpi of degree fi . As long as fi grows
fast enough, any non-constant irreducible polynomial f ∈ Z[X] has a non-
repeated root in the residue field of Ri for all large enough i.
We need to show that H satisfies condition 3 of Theorem 3. If f ∈ Z[X]
is constant and nonzero then vi (f ) = ci ei where ci is the number of factors
of pi dividing f , and the first condition above implies that ⊕i vi (f ) is in H.
So, suppose that f is non-constant and we have a finite sequence x1 ∈
R1 , . . . , xn ∈ Rn satisfying f (xi ) 6= 0. Choose any integer a with f (a)
nonzero. Also, as f is irreducible, we can write uf + vf 0 = c for some
u, v ∈ Z[X] and nonzero integer c. Let P be a finite set of primes containing

11
{pi : i ≤ n} and primes dividing cf (a) as well as all pi for which f does not
have a non-repeated root in the residue field of Ri .
Next, choose xi ∈ Ri for pi P and i > n such that f (xi ) 6= 0. By the
second condition above, there is an h ∈ H with hi = vi (f (xi ) for pi ∈ P. By
adding terms of the form h̃i = 1{pi =p} ei (which is in H by the first condition)
to h, we can assume that hi ≥ 0 for all i.
Now, for each pi 6∈ P, we have vi (f (a)) = 0 so, if hi = 0, we can take
xi = a. On the other hand, if hi > 0, there is a non-repeated root of f in
the residue field of Ri , which will lift to a root x of f in Ri . Choosing y ∈ Ri
with vi (y) = hi gives

vi (f (x + y)) = vi (f 0 (x)y) = hi ,

so we can take xi = x + y. This extends to an infinite sequence xi ∈ Ri with

vi (f (xi )) = hi as required.
Finally, the
Q fact that Cl(R) ∼
= G/H follows immediately from the isomor-
gi
phism g 7→ i pi between G and the fractional ideals of R which, by (1),
gives a bijection between H and the principal ideals.
Finally, let us construct a Dedekind over-ring of Z[X] whose class group
is isomorphic to any given countable abelian group. Let q1 , q2 , . . . be an
distinct sequence of rational primes and set p2i = p2i−1 = qi for each i ∈ N.
By Lemma 9 there exists a sequence of DVRs R1 , R2 , . . . such that Ri is a
finite unramified extension of Qpi .
Now, in any ring R with prime ideals pi satisfying R̄pi ∼
= Ri , the rational
primes decompose as
qi R = p2i p2i−1 .
Let θ : N2 → N and define H0 to be the subgroup of G ≡ ⊕i∈N Z given by
those h ∈ G satisfying
X

h2θ(i,j) − h2θ(i,j)−1 = 0
i

for all j ∈ N. It can be seen that this satisfies the conditions of Lemma 9
and, furthermore, that G/H0 ∼ = ⊕i∈N Z. In fact, G/H0 is freely generated by
the gj ∈ G defined by gji = 1{i=θ(1,j)} .
Now, any countable abelian group A is isomorphic to a quotient of G/H0 ∼
=
∼
⊕i∈N Z and, therefore, A = G/H for a subgroup H of G containing H0 .

12
Then, H will satisfy the requirements of Lemma 9, and the Dedekind do-
main Z[X] ⊆ R ⊆ Q(X) given by condition 2 of Theorem 3 has class group
Cl(R) ∼
= G/H ∼= A.

13