This article was downloaded by: [Nipissing University]

On: 09 October 2014, At: 00:32

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Communications in Algebra
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/lagb20

A Class of Atomic Rings

a b
Sangmin Chun & D. D. Anderson
a
Department of Mathematics , Seoul National University , Seoul , South Korea
b
Department of Mathematics , The University of Iowa , Iowa City , Iowa , USA
Published online: 29 Feb 2012.

To cite this article: Sangmin Chun & D. D. Anderson (2012) A Class of Atomic Rings, Communications in Algebra, 40:3,
1086-1095, DOI: 10.1080/00927872.2010.544698

To link to this article: http://dx.doi.org/10.1080/00927872.2010.544698

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
 Communications in Algebra® , 40: 1086–1095, 2012
Copyright © Taylor & Francis Group, LLC
ISSN: 0092-7872 print/1532-4125 online
DOI: 10.1080/00927872.2010.544698

A CLASS OF ATOMIC RINGS

Sangmin Chun1 and D. D. Anderson2

1
Department of Mathematics, Seoul National University, Seoul, South Korea
2
Department of Mathematics, The University of Iowa, Iowa City, Iowa, USA

Let R be a commutative ring with identity. A nonunit element a ∈ R is an atom if

a = bc
b c ∈ R implies
a =
b or
a =
c, and R is atomic if each nonunit of
Downloaded by [Nipissing University] at 00:32 09 October 2014

R is a finite product of atoms. A ring R is a (weak) CK ring if R is atomic and

(each maximal ideal of) R contains only finitely many nonassociate atoms. We show
that the following are equivalent: (1) R is a weak CK ring, (2) each (prime) ideal
of R is a finite union of principal ideals, (3) R is atomic and each maximal ideal is a
finite union of principal ideals, and (4) R is a finite direct product of finite local rings,
special principal ideal rings (SPIRs), and (one-dimensional) Noetherian domains in
which every maximal ideal is a finite union of principal ideals (or equivalently, weak
CK domains). We show that a domain R is a weak CK domain if and only if R is
a Noetherian domain with RM a CK domain for each maximal ideal M of R and
Pic
R = 0.

Key Words: Atomic ring; Cohen–Kaplansky ring; Weakly factorial domain.

2000 Mathematics Subject Classification: 13A05; 13E99; 13F99.

Throughout this note, all rings will be commutative with identity. Let R be a
commutative ring. A nonunit a ∈ R is called an atom or said to be irreducible
if whenever a = bc, b
c ∈ R, then a = b or a = c, and R is said to be
atomic if each nonunit of R is a finite product of atoms. For a detailed study
of factorization in commutative rings with zero divisors, the reader is referred to
[7]. Finally, R is called a (weak) Cohen–Kaplansky ring, or a (weak) CK ring for
short, if R is atomic and (each maximal ideal of) R contains only finitely many
nonassociate atoms. CK rings, there called f.p.rings, where studied in [1]. It was
shown that the following conditions are equivalent: (1) there exists a finite set of
ideals of R so that every ideal of R is a product of powers of these ideals, (2) R
is a CK ring, (3) R is semiquasilocal and each (prime) ideal of R is a finite union
of principal ideals, and (4) R is a finite direct product of special principal ideal
rings (SPIRs), finite local rings, and one-dimensional semilocal integral domains
D with the property that for each nonprincipal maximal ideal M of D, DM is
analytically irreducible, and D/M is finite. (Recall that an SPIR is a principal

Received September 14, 2010; Revised November 11, 2010. Communicated by I. Swanson.
Address correspondence to Dr. Dan D. Anderson, Department of Mathematics, The University
of Iowa, Iowa City, IA 52242, USA; E-mail: dan-anderson@uiowa.edu

1086
 A CLASS OF ATOMIC RINGS 1087

ideal ring with one prime ideal and that prime ideal is nilpotent. Also, the terms
“local” and “semilocal” carry the Noetherian hypothesis.) Thus CK domains are
characterized as the domains in (4). For a detailed study of CK domains, see [4].
In this paper we study weak CK rings. We show that the following conditions
are equivalent: (1) R is a weak CK ring, (2) every (prime) ideal of R is a finite
union of principal ideals, (3) R is atomic and every maximal ideal of R is a finite
union of principal ideals, (4) R is a finite direct product of SPIRs, finite local
rings and (one-dimensional) Noetherian domains in which every maximal ideal is
a finite union of principal ideals, or equivalently, weak CK domains. The last
equivalence effectively reduces the study of weak CK rings to weak CK domains.
It is shown that an integral domain D is a weak CK domain if and only if D is
Noetherian, for each maximal ideal M of D, DM is a CK domain, and PicD =
0. Thus a Dedekind domain is a weak CK domain if and only if it is a principal
Downloaded by [Nipissing University] at 00:32 09 October 2014

ideal domain (PID). An important property of weak CK domains is that each atom
is primary.
First note that, if R is an atomic ring, then for a prime ideal P of R, P =
∪  a  a ∈ P is an atom. Indeed, if x ∈ P is a product of atoms x = a1 · · · an ,
then some ai ∈ P and hence x ∈ ai  ⊆ P. Thus if P contains only finitely many
nonassociate atoms, then P is a finite union of principal ideals. In the case of a
domain, it is easily seen that the converse is true. It follows from Theorem 2 that if
every prime ideal is a finite union of principal ideals, then every ideal is. This can
easily be shown directly. (First show via Zorn’s Lemma that an ideal that is not a
finite union of principal ideals is contained in an ideal maximal with respect to this
property. Then a modification of the proof of Cohen’s Theorem [13, Theorem 7]
shows that an ideal maximal with respect to not being a finite union of principal
ideals is prime. Recall that Cohen’s Theorem states that a ring is Noetherian if each
prime ideal is finitely generated.) However, since there exist valuation domains of
rank greater than one with principal maximal ideals, we see that in general every
maximal ideal being a finite union of principal ideals does not give that every ideal is
a finite union of principal ideals. Under certain finiteness conditions such as R being
Noetherian, or more generally atomic, every maximal ideal being a finite union of
principal ideals does give that every ideal is. This is shown in Theorem 2, which
gives a number of conditions equivalent to being a weak CK ring. But first, we state
the following useful proposition due to McCoy. (The last sentence of Proposition 1
while not stated by McCoy is an immediate consequence of his theorem.)

Proposition 1 ([14, Theorem 1]). Let R be a commutative ring and I an ideal of R.

Suppose that I = a1  ∪ · · · ∪ an  is an irredundant union (i.e., no ai  can be deleted).
Then for some natural number k, I k ⊆ a1  ∩ · · · ∩ an . Hence if I is a maximal ideal,
each ai  is I-primary.

Theorem 2. For a commutative ring R the following conditions are equivalent:

(1) Every ideal of R is a finite union of principal ideals;

(2) Every prime ideal of R is a finite union of principal ideals;
(3) R is Noetherian and every maximal ideal of R is a finite union of principal ideals;
(4) R is atomic and every maximal ideal of R is a finite union of principal ideals;
 1088 ANDERSON AND CHUN

(5) R is a finite direct product of finite local rings, SPIRs, and one-dimensional
Noetherian domains in which every maximal ideal is a finite union of principal
ideals;
(6) R is a weak CK ring.
Hence a weak CK ring R has dim R ≤ 1.

Proof. (1) ⇒ (2) Clear. (2) ⇒ (3) Since every prime ideal of R is a finite union of
principal ideals, every prime ideal of R is finitely generated. By Cohen’s Theorem,
R is Noetherian. (3) ⇒ (4) This follows since a Noetherian ring, or more generally,
a ring satisfying ACCP is atomic [7, Theorem 3.2]. (4) ⇒ (5) Since R is atomic, R is
a finite direct product of indecomposable atomic rings [7, Theorem 3.3]. Moreover,
since each of these indecomposable factors of R is a homomorphic image of R,
each of these factors inherits the property that each maximal ideal is a finite union
Downloaded by [Nipissing University] at 00:32 09 October 2014

of principal ideals. Thus we may assume that R is indecomposable. First, suppose

that dim R = 0. Then every prime ideal of R is finitely generated; so by Cohen’s
Theorem, R is Noetherian. Hence being an indecomposable Artinian ring, R is local.
By [1, Theorem 1], R is finite or an SPIR. So suppose that dim R ≥ 1. Let P be
a nonmaximal prime ideal of R and let a ∈ P be irreducible, we show that a
is idempotent and hence 0. Then since R is atomic, P = 0; and thus R is a one-
dimensional domain. Again by Cohen’s Theorem, R is Noetherian. Let M be a
maximal ideal of R. If a M, then aM = RM = a2M . So suppose that a ∈ M. Let
M = a1  ∪ · · · ∪ an  be an irredundant union where each ai is irreducible. Now
a ⊆ ai  for some i, say a ⊆ a1 . By Proposition 1, there is a k ≥ 1 with M k ⊆
a1  ∩ · · · ∩ an . Suppose that a = a1 . Then M k ⊆ a1  = a ⊆ P; so M ⊆ P, a
contradiction. Thus a
a1 . Let a = ra1 where r ∈ R. Since a is irreducible, r =
r a for some r ∈ R. Thus a = ra1 = r aa1 . So 1 − r a1 a = 0 and hence aM =
0M since 1 − r a1 M. Thus aM = 0M = a2M . So a = a2 locally and hence
globally. (5) ⇒ (1) It is easily checked that if rings S and T each have the property
that each ideal is a finite union of principal ideals, then S × T also has this property.
Thus it suffices to show that each of the three types of rings in (5) has this property.
For finite rings and SPIRs this is clear. So suppose that R is a Noetherian domain in
which every maximal ideal is a finite union of principal ideals. Let M be a maximal
ideal of R. If M = a1  ∪ · · · ∪ an  where the union is irredundant, then each ai  is
M-primary by Proposition 1; so ht M = 1. Thus dim R = 1. Hence every prime ideal
of R is a finite union of principal ideals. Suppose that some ideal I of R is not a finite
union of principal ideals. Let J ⊇ I be maximal with respect to not being a finite
union of principal ideals. An easy modification of the proof of Cohen’s theorem (see
the proof given in [13, Theorem 7]) shows that J is prime. But J is a finite union
of principal ideals, a contradiction. (6) ⇒ (4) Clear. (5) ⇒ (6) R is Noetherian and
hence atomic. Each indecomposable direct factor of R has the property that each
maximal ideal contains only finitely many nonassociate atoms. Since an atom of a
direct product R1 × · · · × Rn has the form u1
  
ui−1
ai
ui+1
  
un  where ai is
an atom of Ri and the other coordinates are units [7, Theorem 2.15], it is easy to
see that the direct product has the property that each maximal ideal contains only
finitely many nonassociate atoms. 

Note that (5) of Theorem 2 can be restated as (5 ) R is a finite direct product of

finite local rings, SPIRs, and weak CK domains. So Theorem 2 effectively reduces
 A CLASS OF ATOMIC RINGS 1089

the study of weak CK rings to weak CK domains. We next wish to show that if R is
a weak CK ring and M is a maximal ideal of R, then RM is a CK ring. We actually
prove a stronger result.

Theorem 3. Let R be an atomic ring. If M is a maximal ideal of R that is a finite

union of principal ideals, then RM is a CK ring.

Proof. Let M = a1  ∪ · · · ∪ an  be an irredundant union. Then each ai is an

atom and by Proposition 1, each ai  is M-primary. Now M = a1  ∪ · · · ∪ an 
gives that MM = a1 M ∪ · · · ∪ an M . So by Theorem 2 it suffices to show that RM
is Noetherian. First suppose ht M = 0. Then MM is finitely generated; so RM is
Noetherian by Cohen’s Theorem. So suppose that P
M is a prime ideal. Let a ∈ P
be an atom. Now a ⊆ ai  for some i and we must have a
ai  as ai  is M-
primary. Hence as in the proof of (4) ⇒ (5) of Theorem 2, we have PM = 0M . So
Downloaded by [Nipissing University] at 00:32 09 October 2014

RM is a one-dimensional quasilocal domain with MM finitely generated. Again by

Cohen’s Theorem, RM is Noetherian. 

Corollary 4. If R is a weak CK ring and M is a maximal ideal of R, then RM is a CK

ring.

So a weak CK ring is locally a CK ring. The converse is false. For if R is an

almost Dedekind domain (i.e., R is locally a discrete valuation ring (DVR)) that is
not Dedekind, then R is locally a CK domain, but R is not a weak CK domain as
it is not even Noetherian. With a little more work, we get that the converse is false
even if R is Noetherian.

Proposition 5. Let R be a commutative ring and M a maximal ideal of R that is a

finite union of principal ideals. Then M is locally principal if and only if it is principal.

Proof. (⇐) Clear. (⇒) Let M = a1  ∪ · · · ∪ an  be an irredundant union. So by

Proposition 1 each ai  is M-primary. Now MM = a1 M ∪ · · · ∪ an M is principal;
so MM = ai M for some i. But ai  is M-primary, so M = ai  is principal. 

Corollary 6. A Dedekind domain is a weak CK domain if and only if it is a PID.

So a Noetherian domain that is locally a CK domain need not be weak CK

domain. Our next goal is to generalize Corollary 6 to: an integral domain D is a
weak CK domain if and only if D is Noetherian, locally a CK domain, and PicD =
0. To do this we need the notion of a weakly factorial domain. Recall that an
integral domain D is weakly factorial [3] if each nonzero nonunit of D is a product
of primary elements. The following result from [3] is needed.

Proposition 7 [3, Theorem 12]. For a one-dimensional Noetherian domain R, the

following statements are equivalent:
(1) R is weakly factorial;
(2) Every irreducible element of R is primary;
(3) PicR = 0.
 1090 ANDERSON AND CHUN

If R is a one-dimensional weakly factorial Noetherian domain, then the integral

closure R of R is a PID.

Proposition 8. Let R be a weak CK domain. Then every irreducible element of R is

primary. Hence R is weakly factorial and Pic R = 0. Moreover, R is a PID.

Proof. Let a be a nonzero irreducible element of R. Let a ∈ M, a maximal ideal.

Now M = a1  ∪ · · · ∪ an  where each ai is irreducible. So a ⊆ ai  and since a is
irreducible, a is a maximal element in the set of proper principal ideals. Thus a =
ai . By Proposition 1, ai , and hence a, is M-primary. Since R is a one-dimensional
Noetherian domain, R is weakly factorial and Pic R = 0 by Proposition 7. (Since
R is atomic, we really do not need to appeal to Proposition 7 to get that R is weakly
factorial.) Again by Proposition 7, R is a PID. 
Downloaded by [Nipissing University] at 00:32 09 October 2014

Theorem 9. For an integral domain R, the following statements are equivalent:

(1) R is a weak CK domain;
(2) R is Noetherian, RM is a CK domain for each maximal ideal M of R, and
PicR = 0;
(3) R is Noetherian, for each nonprincipal maximal ideal M of R, R/M is finite and
RM is a one-dimensional analytically irreducible local domain, and PicR = 0;
(4) R is of finite character, RM is a CK domain for each maximal ideal M of R, and
PicR = 0;
(5) R is weakly factorial and RM is a CK domain for each maximal ideal M of R.

Proof. (1) ⇒ (2) R is Noetherian by Theorem 2 and R is locally a CK domain

by Corollary 4. By Proposition 8, PicR = 0. (2) ⇒ (3) This follows since for a
maximal ideal M of R, RM is a CK domain if and only if either RM is a DVR
(equivalently, M is invertible and hence principal since PicR = 0) or RM /MM ≈
R/M is finite and RM is a one-dimensional analytically irreducible local domain. (3)
⇒ (4) R is a one-dimensional Noetherian domain and hence has finite character.
And as in the proof of (2) ⇒ (3), R is locally a CK domain. (4) ⇒ (5) For each
maximal ideal M of R, RM is a CK domain and hence ht M = ht MM = dim RM =
1. So dim R = 1. Since R has finite character and R is locally Noetherian, R is
Noetherian ([12, Lemma 37.3] or [17, Remark 1.I]). So R is a one-dimensional
Noetherian domain with Pic R = 0. By Proposition 7, R is weakly factorial. (5) ⇒
(1) We first note that R is Noetherian and hence atomic. Since R is locally a CK
domain, dim R = 1. Since R is weakly factorial, by [3, Theorem 13] each nonzero
nonunit of R is contained in only finitely many height-one prime ideals (or in this
case, maximal ideals). Since R is locally Noetherian, R is Noetherian. Let M be a
maximal ideal of R, and let x ∈ M be M-primary. We next note that x is an atom of
R if and only if x is an atom of RM . Now if x is not an atom of R, then x = yz where
y and z are nonunits of R. It is easily checked that y and z are also M-primary. So
y
z ∈ M and hence x = yz is not an atom in RM . Conversely, suppose that x is not
an atom in RM . So in RM , we have x = y/sz/t where y
z ∈ M and s
t ∈ R − M.
Moreover, we can choose y and z to be M-primary. Since a product of M-primary
elements is again M-primary, stx = yz is M-primary. Thus the factors s and t of stx
are either units or M-primary. But s
t M, so s and t are units in R. Thus x is not
 A CLASS OF ATOMIC RINGS 1091

an atom of R. Now as R is weakly factorial, each irreducible element of R is primary

and thus an irreducible element contained in M is M-primary. Let q1
  
qn be a
complete set of nonassociate atoms for RM where we can take each qi to be an M-
primary element of R. Then q1
  
qn is a complete set of nonassociate M-primary
atoms of R. So M contains only finitely many nonassociate atoms. Thus R is a weak
CK domain. 

Let R be a weak CK domain. For a nonzero nonunit x ∈ R, x has a primary

decomposition x = q1  ∩ · · · ∩ qn  = q1  · · · qn  where qi  is Mi -primary and
M1
  
Mn  is the set of maximal ideals containing x. Moreover, qi  = xMi ∩ R is
uniquely determined. Thus an atomic factorization of x is obtained by factoring the
q1
  
qn into atoms. Since an Mi -primary element q is an atom in R if and only if
it is an atom of RMi (proof of (5) ⇒ (1) of Theorem 9), the factorization properties
of R are largely determined by the factorization properties of the localizations of R.
Downloaded by [Nipissing University] at 00:32 09 October 2014

Two such factorization properties that will concern us are now recalled. An integral
domain R is a half factorial domain (HFD) if R is atomic and if x1 · · · xn = y1 · · · ym
are two atomic factorizations, then n = m. A number of equivalent conditions for
a local CK domain to be half factorial are given in [4, Theorem 6.3]. Also, an
integral domain R is a finite factorization domain (FFD) if one the following three
equivalent conditions holds: (1) R is atomic and each nonzero nonunit of R has
only finitely many nonassociate irreducible divisors, (2) each nonzero nonunit of R
has only finitely many nonassociate divisors, and (3) each nonzero nonunit of R has
only finitely many factorizations up to order and associates. For results concerning
FFDs, see [6].

Theorem 10. Let R be a weak CK domain.

(1) R is a UFD, or equivalently a PID, if and only if RM is a DVR for each maximal
ideal M of R.
(2) R is an HFD if and only if RM is an HFD for each maximal ideal M of R.
(3) R is always an FFD.

Proof. (1) Since R is a one-dimensional domain, it is a UFD if and only if it is a

PID. Since PicR = 0, R is a PID if and only if it is a Dedekind domain. But R is
a Dedekind domain if and only if RM is a DVR for each maximal ideal M.
(2) Let x be a nonzero nonunit of R. As mentioned in the paragraph
preceding Theorem 10, x = q1 · · · qn , where qi  is Mi -primary for distinct maximal
ideals M1
  
Mn , and this factorization is unique up to order and associates. Now
an atomic factorization of x is obtained by taking atomic factorizations for each qi .
Since an Mi -primary element of R is an atom of R if and only if it is an atom of
RMi , any two atomic factorizations of qi have the same length if and only if any two
atomic factorizations of qi in RMi have the same length. So R is an HFD if and only
if each localization RM is an HFD.
(3) Let x be a nonzero nonunit of R. Now any irreducible divisor of x
is M-primary for some maximal ideal M containing x. Since x is contained in
only finitely many maximal ideals and since each of these maximal ideals contains
only finitely many nonassociate (M-primary) atoms, x has only finitely many
 1092 ANDERSON AND CHUN

nonassociate irreducible divisors. Hence R is an FFD. Alternatively, R is a locally

finite intersection of FFDs and hence an FFD [6, Theorem 2]. 

Recall that a ring extension R ⊆ S is called a root extension if for each

s ∈ S, there exists an n = ns with sn ∈ R. So a root extension is an integral
extension. Moreover, a strong version of Lying Over holds: for each prime ideal
P of R there exists a unique prime ideal Q of S with Q ∩ R = S, namely, Q = s ∈
S  sn ∈ P for some n ≥ 1; see [9, Theorem 2.1]. Hence if R ⊆ R is a root extension,
MaxR = MaxR. If R is a CK domain, then there exists an n ≥ 1 so that xn ∈
R for each x ∈ R [4, Lemma 4.1]. This result is generalized in Theorem 13.
A generalized CK domain is an atomic domain in which almost all atoms are
prime. These domains were introduced in [2] where it was shown that a generalized
CK domain is weakly factorial; in fact the stronger result was proved: an atomic
domain R in which almost all atoms are primary has every atom primary, that is, R
Downloaded by [Nipissing University] at 00:32 09 October 2014

is weakly factorial [2, Theorem 4]. The main result of [2] was Theorem 6: an integral
domain R is a generalized CK domain if and only if (1) R is a UFD, (2) R ⊆ R is a
root extension, (3) R  R is a principal ideal of R, and (4) R/R  R is finite. While
the implication ⇒ is correct, unfortunately the proof of ⇐ has a gap as pointed
out by M. Picavet-L’Hermitte [16] who also provides a counterexample to ⇐. The
error is in the paragraph on the top of page 1458 of [2] showing that vRM = PM
where P ∈ X 1 R. The proof of the case M ⊇ R  R inadvertently uses P ⊆ M to
conclude that vRM = PM . She corrects the proof by adding either of the equivalent
conditions (5) each P ∈ X 1 R such that P ⊇ R  R is principal or (5 ) Clt R = 0.
Here Clt R is the t-class group of R, that is, the group of t-invertible t-ideals modulo
the subgroup of nonzero principal (fractional) ideals. We next give a correction of
Theorem 6 showing that (5 ) can be replaced by the weaker condition Pic R = 0.

Theorem 11. Let R be an integral domain. Then R is a generalized CK domain if and

only if (1) R is a UFD, (2) R ⊆ R is a root extension, (3) R  R is a principal ideal
of R, (4) R/R  R is finite, and (5) PicR = 0, or equivalently, Clt R = 0.

Proof. ⇒ The proof given in [2] shows that (1)–(4) hold. Since R is weakly
factorial, Clt R = 0 [8] and hence PicR = 0. ⇐ It suffices to show that if P ∈
X 1 R with P ⊃ R  R , then P is principal. Let M ⊇ P be a maximal ideal. Then
M ⊇ R  R because a maximal ideal containing R  R has height one, so PRM =
P RM is a height-one prime in the UFD RM and hence is principal. Since R =
∩P∈X1 R RP is locally finite, an appeal to [5, Corollary 2.4] gives that P is invertible.
Thus the hypothesis (5) PicR = 0 gives that P is principal. 

We next relate weak CK domains and generalized CK domains.

Theorem 12. For an integral domain R, the following are equivalent:

(1) R is a weak CK domain with R  R = 0;

(2) R is a weak CK domain with R a finitely generated R-module;
(3) R is a weak CK domain and a generalized CK domain;
(4) R is a one-dimensional generalized CK domain;
(5) R is a PID, R ⊆ R is a root extension, R/R  R is finite, and PicR = 0.
 A CLASS OF ATOMIC RINGS 1093

Proof. (1) ⇔ (2) A weak CK domain is Noetherian and for a Noetherian domain
R  R = 0 if and only if R is a finitely generated R-module. (1) ⇒ (3) Let M be a
maximal ideal of R. If M ⊇ R  R , then RM = RM is a DVR because R is a PID
(Proposition 8). So M is locally principal and hence invertible as R is Noetherian.
Since PicR = 0 (Theorem 9), M is principal. Since R  R is contained in only
finitely many maximal ideals of R, almost all maximal ideals of R are principal: So
almost all atoms of R are prime because a maximal ideal containing R  R has
only finitely many atoms, that is, R is a generalized CK domain. (3) ⇒ (1) Here
R  R = 0 by Theorem 11. (4) ⇔ (5) This follows from Theorem 11 since a one-
dimensional domain is a UFD if and only if it is a PID. (3) ⇒ (4) This follows
since a weak CK domain is one-dimensional. (4) ⇒ (3) A generalized CK domain
is weakly factorial [2, Corollary 5] and a one-dimensional generalized CK domain
is locally a CK domain. By Theorem 9, R is a weak CK domain. 
Downloaded by [Nipissing University] at 00:32 09 October 2014

We next show that if R is a weak CK domain, then R ⊆ R is a root extension.

Theorem 13. Let R be a weak CK domain or a generalized CK domain. Then R ⊆ R

is a root extension. If R  R = 0 (e.g., R is a generalized CK domain), then there is a
fixed n ≥ 1 so that xn ∈ R for each x ∈ R.

Proof. Let x ∈ R. Since R is either a weak CK domain or a generalized CK

domain, R = ∩P∈X1 R RP is locally finite. Moreover, for P ∈ X 1 R either RP = RP
is a DVR or P is a maximal ideal with P ⊇ R  R and RP is a CK domain. Now x is
in almost all RP since the intersection is locally finite. Let P1
  
Ps ∈ X 1 R with
x RPi . Now RPi is a CK domain, so RPi ⊆ RPi is a root extension with RPi a DVR.
In fact, there exists an ni ≥ with zni ∈ RPi for all z ∈ RPi . Take n = n1 · · · ns . Then
zn ∈ RPi for each i. Hence xn ∈ ∩P∈X1 R RP = R. If R  R = 0, then x ∈ RP = RP for
all P ⊇ R  R . As above, there exists a fixed n not depending on x so that xn ∈ RQ
for the finite set of height-one primes containing R  R . Thus for each x ∈ R, xn ∈
∩P∈X1 R RP = R. 

Recall that an integral domain R is an almost Bezout domain or AB-domain

(resp., almost principal ideal domain or API-domain) if for a
b ∈ R (resp., a  ⊆
R), there exists an n ≥ 1 with an
bn  (resp., an ) principal. These domains were
introduced in [9]. A domain R is an AB-domain if and only if R is a Prüfer domain
with torsion class group and R ⊆ R is a root extension [9, Corollary 4.8]. Now if R
is an API-domain, then R is an AB-domain; so R ⊆ R is a root extension. Suppose
that R is a domain with the property that there exists a fixed n ≥ 1 so that x ∈ R
implies xn ∈ R. Then R is an API-domain if and only if R is a Dedekind domain
with torsion class group [9, Corollary 4.13]. Thus a weak CK domain is an AB-
domain (Propositions 8 and 7 and Theorem 13) and a one-dimensional generalized
CK domain is an API-domain (Theorems 12 and 13).
We end by considering the existence of weak CK domains. Of course
a PID is a weak CK domain. Let be a cardinal number, k a field, and
X  ∈ a set of indeterminates over k with   = . Then kX  ∈ S , where
S = f1 · · · fn fi ∈ kX  ∈ is irreducible with fi  X for any , is a PID with
nonassociate principal primes. A weak CK domain is semilocal if and only if it is
a CK domain. A structure theory for CK domains is given in [4, Theorem 4.4],
 1094 ANDERSON AND CHUN

basically, they are a special type of pullback of a semilocal PID. Coykendall and
Spicer [11] show, assuming a slight strengthening of the Goldbach Conjecture, that
for each n ≥ 3, there exists a CK domain with exactly n nonassociate atoms, each
of which is not prime. Picavet-L’Hermitte [16] investigates nonmaximal orders that
are generalized CK domains, or equivalently, weak CK domains with nonzero √
conductor. One such example which is neither a CK domain nor a PID is
 −3 .
One searches for a weak CK domain R with zero conductor. By Theorem 12,
R cannot be a generalized CK domain. Note that such an R cannot contain an
infinite field, for then each residue field is infinite and hence R is a PID. Also, R is
not a finitely generated R-module while RM is a finitely generated RM -module for
each maximal ideal M. Now Nagata [15, Example 8, p. 211] does give an example
of a one-dimensional Noetherian domain with this property, but unfortunately R
contains an infinite field and hence cannot be a weak CK domain. Next we can
consider the A + XBX construction.
Downloaded by [Nipissing University] at 00:32 09 October 2014

Theorem 14. Let B be an integral domain with quotient field K and let A be a subring
of B. Let R = A + XBX . Then the following conditions are equivalent:
(1) R is a one-dimensional generalized CK domain;
(2) R is a weak CK domain;
(3) A = B = K or B = K is a finite field.

Proof. (1) ⇒ (2) Theorem 12. (2) ⇒ (3) Suppose that R is a weak CK
domain. Since dim R = 1, A ⊆ B = K are fields. Here A is a field since dim A =
dim R/XBX = 0 and B is a field since 1 = dim R ≥ dim RX −1 = dim BX
X −1 >
dim B. Suppose A  B; so R is not a PID since the ideal XBX of R is not principal.
Now R is Noetherian, so B  A < . Hence R = BX and R  R = XBX . So R
is a generalized CK domain and hence B = R/R  R is finite. (3) ⇒ (1) If A = B =
K, R = KX is a PID. Suppose that K is finite. Then R = KX is a PID, R ⊆ R
is a root extension, R/R  R = K is finite, and PicR = 0 [10, Example 1.10]. By
Theorems 11 and 12, R is a one-dimensional generalized CK domain. 

Finally, suppose that R is a weak CK domain with char R = 0. Let p >

0 be prime. Since p is in only finitely many maximal ideals of R, only finitely
many residue fields of R have characteristic p > 0. Now if a residue field R/M has
char R/M = 0, then R/M is infinite and hence M is principal. We conclude that if R
is a weak CK domain with char R = 0, then the set of nonprincipal maximal ideals
of R is countable.

REFERENCES
[1] Anderson, D. D. (1978). Some finiteness conditions on a commutative ring. Houston J.
Math. 4:289–299.
[2] Anderson, D. D., Anderson, D. F., Zafrullah, M. (1992). Atomic domains in which
almost all atoms are prime. Comm. Algebra 20:1447–1462.
[3] Anderson, D. D., Mahaney, L. A. (1988). On primary factorizations. J. Pure Appl.
Algebra 54:141–154.
[4] Anderson, D. D., Mott, J. L. (1992). Cohen–Kaplansky domains: Integral domains
with a finite number of irreducible elements. J. Algebra 148:17–41.
 A CLASS OF ATOMIC RINGS 1095

[5] Anderson, D. D., Mott, J. L., Zafrullah, M. (1992). Finite character representations
for integral domains. Bollettino U. M. I. (7) 6-B:613–630.
[6] Anderson, D. D., Mullins, B. (1996). Finite factorization domains. Proc. Amer. Math.
Soc. 124:389–396.
[7] Anderson, D. D., Valdes-Leon, S. (1996). Factorization in commutative rings with zero
divisors. Rocky Mountain J. Math. 26:439–480.
[8] Anderson, D. D., Zafrullah, M. (1990). Weakly factorial domains and groups of
divisibility. Proc. Amer. Math. Soc. 109:907–913.
[9] Anderson, D. D., Zafrullah, M. (1991). Almost Bézout domains. J. Algebra
142:285–309.
[10] Bouvier, A., Zafrullah, M. (1988). On some class groups of an integral domain. Bull.
Greek Math. Soc. 29:45–59.
[11] Coykendall, J., Spicer, C. C. Cohen–Kaplansky domains and the Goldbach conjecture.
Proc. Amer. Math. Soc. (to appear).
[12] Gilmer, R. (1992). Multiplicative Ideal Theory. Queen’s Papers in Pure and Appl. Math.
Downloaded by [Nipissing University] at 00:32 09 October 2014

Vol. 90. Ontario: Queen’s University, Kingston.

[13] Kaplansky, I. (1994). Commutative Rings. Revised Edition, Washington, NJ:
Polygonal Press.
[14] McCoy, N. (1957). A note on finite unions of ideals and subgroups. Proc. Amer. Math.
Soc. 8:633–637.
[15] Nagata, M. (1962). Local Rings. New York: Interscience.
[16] Picavet-L’Hermitte, M. (2000). Factorization in Some Orders with a PID as Integral
Closure, Algebraic Number Theory and Diophantine Analysis. Berlin: de Gruyter,
pp. 365–390.
[17] Zafrullah, M. (1977). On a result of Gilmer. J. London Math. Soc. (2) 16:19–20.