Communications in Algebra

ISSN: (Print) (Online) Journal homepage: [Link]

On normal modules

Chillumuntala Jayaram, Ünsal Tekir, Suat Koç & Seçil Çeken

To cite this article: Chillumuntala Jayaram, Ünsal Tekir, Suat Koç & Seçil Çeken (2022): On
normal modules, Communications in Algebra, DOI: 10.1080/00927872.2022.2137519
To link to this article: [Link]

Published online: 26 Oct 2022.

Submit your article to this journal

Article views: 107

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

[Link]
COMMUNICATIONS IN ALGEBRA®
[Link]

On normal modules
Chillumuntala Jayarama , Ünsal Tekirb , Suat Koçc , and Seçil Çekend
a
Department of CMP, The University of the West Indies, Bridgetown, Barbados; b Department of Mathematics, Marmara
University, Ziverbey, Goztepe, Istanbul, Turkey; c Department of Mathematics, Faculty of Engineering and Natural Sciences,
Istanbul Medeniyet University, Istanbul, Turkey; d Department of Mathematics, Trakya University, Edirne, Turkey

ABSTRACT ARTICLE HISTORY

Recall that a commutative ring R is said to be a normal ring if it is reduced and Received 03 January 2022
every two distinct minimal prime ideals are comaximal. A finitely generated Revised 12 October 2022
reduced R-module M is said to be a normal module if every two distinct minimal Communicated by Toma
prime submodules are comaximal. The concepts of normal modules and locally Albu
torsion free modules are different, whereas they are equal in theory of commu- KEYWORDS
tative rings. We give many properties and examples of normal modules, we use Baer modules; locally
them to characterize locally torsion free modules and Baer modules. Also, we torsion-free modules; normal
give the topological characterizations of normal modules. modules; quasi-regular
modules
2020 MATHEMATICS
SUBJECT CLASSIFICATION
16E50; 13A15

1. Introduction
Throughout this article, we focus only on commutative rings with a nonzero identity and nonzero unital
modules. R will always represent such a ring and M will represent such an R-module. We denote the
set of all maximal ideals and the set of all prime ideals of R by Max(R) and Spec(R), respectively. Recall
that a commutative ring R is said to be a locally integral domain if RP is an integral domain for each
P ∈ Spec(R). It is easy to see that every integral domain is also a locally integral domain but the converse
is not true in general. For instance, take a von Neumann regular ring R that is not a field. Then, R is
clearly a locally integral domain but not an integral domain. Let M be an R-module. The set of torsion
elements of M is denoted by T(M) = {m ∈ M : rm = 0 for some nonzero r ∈ R} [3]. An R-module M is
said to be a torsion free module if T(M) = 0. M is said to be a torsion module if T(M) = M. Otherwise,
we say that M is a non-torsion module, that is, M is a module with a non-torsion element m ∈ M. In
a recent paper, Jayaram et al. studied the concept of locally torsion free modules as a generalization of
torsion free modules, and used them to characterize Baer modules and von Neumann regular rings.
Recall from [16] that an R-module M is said to be a locally torsion free module if RP -module MP is a
torsion free module for every P ∈ Spec(R). Also, recall that a reduced ring (without nilpotent elements)
R is said to be a normal ring if every two distinct minimal prime ideals P1 , P2 of R are comaximal, that
is, P1 + P2 = R [11]. Our aim in this paper is to extend the concept of normal rings to modules and to
use them in characterizing locally torsion free modules and Baer modules. For the sake of completeness,
we give some notions and notations which will be used in the sequel. Let M be an R-module. Suppose
that N is a submodule of M, K is a nonempty subset of M and L a nonempty subset of R. The residuals
of N by K and L are denoted by
(N :R K) = {a ∈ R : aK ⊆ N}
(N :M L) = {m ∈ M : Lm ⊆ N}.

CONTACT Seçil Çeken cekensecil@[Link] Department of Mathematics, Trakya University, Edirne, Turkey.
© 2022 Taylor & Francis Group, LLC
2 C. JAYARAM ET AL.

In particular, if N = 0, we will use annR (K) and annM (L) instead of (0 :R K) and (0 :R L), respectively.
If annR (M) = 0, then M is said to be a faithful module. Also, recall from [5], an R-module M is said
to be a multiplication module if every submodule N of M has the form IM for some ideal I of R. For
more details on multiplication modules, the reader may consult [2] and [7]. M is said to be a reduced
module if whenever a2 m = 0 for some a ∈ R, m ∈ M, then am = 0 [18]. A finitely generated
reduced R-module M is said to be a normal module if every two distinct minimal prime submodules
are comaximal. By [11, Theorem 8] and [16, Lemma 2.1], we know that the classes of normal rings and
locally integral domains coincide. So, one can naturally ask whether the classes of normal modules and
locally torsion free modules are equal or not. However, we show that these are different concepts (See,
Examples 2.4 and 2.5). Also, we determine the conditions under which locally torsion free modules
and normal modules are equal (See, Propositions 2.14, 2.15 and Theorem 2.16). Among the other
things in this paper, we characterize normal modules in terms of normal rings (See, Lemma 2.11 and
Proposition 2.12). Let S be a multiplicatively closed set of R and denote the set {m ∈ M : sm = 0
for some s ∈ S} by O(S). Note that O(S) is always a submodule of M. In particular, we prefer OP to
denote O(R − P) for every P ∈ Spec(R). Recall from [15] that a submodule N of M is said to be a ∗-
submodule if N = O(S) for some multiplicatively closed set S of R. Also, recall from [14] that a submodule
N of M is said to be a σ -submodule if for each m ∈ N, we have annR (m) + (N :R M) = R. The
set of all σ -submodules of M is denoted by σ (M). N is called a Baer submodule if m ∈ N implies that
annM (annR (m)) ⊆ N [14]. The authors in [14] and [15] studied some basic properties of ∗-submodules,
σ -submodules and Baer submodules and they used them to characterize Baer modules, von Neumann
regular modules, locally torsion free modules and weak quasi-regular modules. Here, we investigate
further properties of ∗-submodules, σ -submodules and Baer submodules (See, Lemmas 3.1, 3.5, 3.7,
3.9, 3.13, Propositions 3.11, 3.12, 3.15, 3.16, 3.17 and Theorem 3.14). Also, we use ∗-submodules, σ -
submodules and Baer submodules to give some new characterizations of Baer modules, locally torsion-
free modules and normal modules (See, Theorems 3.2, 3.10 and 3.18). Finally, in Section 3, we study
the Zariski topology of modules and give the topological characterizations of normal modules (See,
Theorem 4.4).

2. Characterizations of normal modules

Let M be an R-module and N, K be two submodules of M. Recall that N and K are said to be comaximal
if N + K = M. In particular, any two ideals I and J are comaximal if they are comaximal submodules of
R-module R.
Recall that an element e ∈ R is said to be a weak idempotent if e − e2 ∈ annR (M), or equivalently,
em = e2 m for each m ∈ M [13]. Jayaram and Tekir, introduced von Neumann regular modules as
a generalization of von Neumann regular rings and they gave many properties and characterizations
of von Neumann regular modules in terms of weak idempotent elements. Recall from [13] that an R-
module M is said to be a von Neumann regular module if for each m ∈ M, there exists a ∈ R such
that Rm = aM = a2 M. The authors in [13, Theorem 1] showed that a finitely generated multiplication
module M is von Neumann regular module if Rm = eM for some weak idempotent e ∈ R. Recently,
Jayaram et al. introduced the class of Baer modules which is a generalization of finitely generated von
Neumann regular modules. Recall from [14] that an R-module M is said to be a Baer module if for each
m ∈ M, there exists a weak idempotent element e ∈ R such that annR (m)M = eM. We warn the reader
that there are other Baer module definitions in the literature (see for example [6], [23]). In this paper,
we study the concept of Baer module in the sense of [14].
Now, we begin the study by giving a new property of minimal prime submodules in finitely generated
Baer modules.

Proposition 2.1. Let M be a finitely generated module. If M is a Baer module, then any two distinct minimal
prime submodules are comaximal.
 COMMUNICATIONS IN ALGEBRA® 3

Proof. From [19], we know that maximal submodules, prime submodules and minimal prime submod-
ules exist in finitely generated modules. Let N and K be two distinct minimal prime submodules of a
finitely generated Baer module M. Then, (N :R M) and (K :R M) are prime ideals containing annR (M).
Thus, (N :R M) contains P1 and (K :R M) contains P2 , where P1 and P2 are prime ideals which are
minimal over annR (M). This implies that P1 /annR (M) and P2 /annR (M) are minimal prime ideals of
R/annR (M). On the other hand, by [14, Proposition 2.12], R/annR (M) is a Baer ring. Then we have
P1 /annR (M) + P2 /annR (M) = R/ann(M), which implies that P1 + P2 = R. Then we conclude that
(N :R M) + (K :R M) = R and so N + K = M which completes the proof.

Definition 2.2. A finitely generated reduced module M is said to be a normal module if any two distinct
minimal prime submodules are comaximal.

Remark 2.3. Note that, by [14, Proposition 2.7] and Proposition 2.1, every finitely generated Baer
module is an example of a normal module. Also it can be easily shown that a finitely generated reduced
module M is a normal module if and only if every prime submodule contains a unique minimal prime
submodule.

Example 2.4 (A normal module that is not locally torsion free). Consider the Z-module M = Z2 ×
Z2 . Then P = {(0, 0)} is a prime submodule so is unique minimal prime. Since M is finitely generated
reduced and has a unique minimal prime, M is a normal module. Let a = 2 and m = (0, 1). Then
am = (0, 0) but annZ (a) + annZ (m) = 2Z = Z. Therefore, M is not a locally torsion free module by
[16, Lemma 1].

Example 2.5 (A locally torsion free module that is not normal). Consider the Z-module M =
Z[X]. Since M is not finitely generated, M is not a normal module. On the other hand, since M is a
torsion free module, by [16, Example 2], M is a locally torsion free module.

Proposition 2.6. Let n > 1 be an integer. Then Zn is a normal Z-module if and only if n = p1 p2 · · · pr for
some distinct prime numbers p1 , p2 , . . . , pr .

Proof. Suppose that Zn is a normal Z-module. Since it is reduced, one can easily see that n is square
free integer, that is, n = p1 p2 · · · pr for some distinct prime numbers p1 , p2 , . . . , pr . For the converse,
assume that n = p1 p2 · · · pr for some distinct prime numbers p1 , p2 , . . . , pr . If r = 1, then Zp1 has a
unique minimal prime submodule which is also maximal. So, Zp1 is a trivially normal module. Assume
that r ≥ 2. It can be easily seen that every prime submodule of Zn is of the form (pi ) which is also
minimal prime. Let i = j. Since gcd(pi , pj ) = 1, we have (pi ) + (pj ) = Zn . Therefore, Zn is a normal
Z-module.
It is easy to see that if M is a reduced R-module, then R/annR (M) is a reduced ring. But the converse
is not always true.

∞
Example 2.7. Let p be a prime number. Consider R = Z-module M = Zpn . Then it is clear that
n=1
annR (M) = 0 and so R/annR (M) = Z is a reduced ring. Now take m = (1, 1, 0, 0, ...). Since p2 m =
0M but pm = 0M , it follows that M is not a reduced module.

Definition 2.8. Let M be a finitely generated R-module. Then according to [15],

(i) M is said to satisfy the condition (#) if K is a minimal  prime submodule, then K = (K :R M)M,
(ii) M is said to satisfy the condition (P) if (PM) = ( P)M for all prime ideals P minimal over
annR (M),
(iii) M is said to satisfy the condition (# #) if it satisfies the condition (#) and (P).
4 C. JAYARAM ET AL.

The proof the following two propositions can be found in [15, Proposition 2.16] and [15, Proposition
2.17]. Since they will be frequently used in the sequel, we will give them for the sake of completeness.

Proposition 2.9. Let M be a finitely generated module satisfying the condition (#) and K a submodule of
M. Then,

(i) If P is a minimal prime ideal over annR (M), then PM is a minimal prime submodule of M.
(ii) If K is a minimal prime submodule of M, then (K :R M) is a minimal prime ideal over annR (M).

Proposition 2.10. Let M be a finitely generated module and I be an ideal containing annR (M). Assume
that N = (N :R M)M for each prime submodule N of M minimal over IM. Then,

(i) If P is a prime ideal minimal over I, then PM is a prime submodule minimal over IM.
(ii) If K is a prime submodule minimal over IM, then (K :R M) is a prime ideal minimal over I.

Lemma 2.11. Suppose M is a finitely generated module satisfying the condition (# #). Then M is a reduced
module if and only if R/annR (M) is a reduced ring.

Proof. The "if " part is clear. Now we√will show that the "only if " part is also true.√Suppose that
R/annR (M) is a reduced ring, that is, annR (M) = annR (M).
√ Let a m = 0. If a ∈ annR (M) =
2

annR (M), there is nothing to prove. Suppose that a ∈ / annR (M). Take a minimal prime P of
annR (M). Then N = PM is a minimal prime submodule of M by Proposition 2.9. Assume that am ∈ /
PM = N. Since a(am) = 0 ∈ N and N is a prime submodule, we conclude that a ∈ (PM :R M) = P and
thus am ∈ aM ⊆ PM which is a contradiction. Since M satisfies (# #) condition, we have
 
0 = annR (M)M = ( P)M = (PM).
P∈Min(annR (M)) P∈Min(annR (M))


Then we get am ∈ (PM) = 0 and so am = 0 which completes the proof.
P∈Min(annR (M))

Proposition 2.12. Let M be a finitely generated R-module satisfying (# #). Then M is a normal module if
and only if R/annR (M) is a normal ring.

Proof. (⇒) : Suppose that M is a normal module. Since M is a reduced module, so is R/annR (M) by
Lemma 2.11. Take two minimal prime ideals P1 /annR (M) and P2 /annR (M) of R/annR (M). Then P1 and
P2 are minimal prime ideals of annR (M). By Proposition 2.9, P1 M, P2 M are minimal prime submodules
of M and thus P1 M + P2 M = (P1 + P2 )M = M. Since M is finitely generated and P1 + P2 ⊇ ann(M), by
[4, Corollary 2.5], P1 +P2 = R. This implies that P1 /annR (M)+P2 /annR (M) = R/annR (M). Therefore,
R/annR (M) is a normal ring.
(⇐) : It is similar to the proof of Proposition 2.1.

Lemma 2.13. ([16, Lemma 2.1]) Let M be an R-module. The following statements are equivalent.

(i) M is a locally torsion free module.

(ii) am = 0 implies annR (a) + annR (m) = R for each a ∈ R and m ∈ M.
(iii) annR (am) = annR (a) + annR (m) for each a ∈ R and m ∈ M.
(iv) RP -module MP is a torsion free module for each P ∈ Max(R).

Proposition 2.14. Let M be a finitely generated faithful reduced module satisfying the condition (#).
Further assume that for any m in M, annR (m) = annR (a) for some a in R. If M is a normal module,
then M is a locally torsion-free module.
 COMMUNICATIONS IN ALGEBRA® 5

Proof. Suppose rm = 0 for some r in R and m in M. Assume that annR (r) + annR (m) contained in
P for some prime ideal P of R. By assumption, we have annR (m) = annR (a) for some a in R. Then
annR (r), annR (a) are contained in P, so there exist prime ideals P0 and P1 minimal over annR (r), annR (a)
and contained in P. Since R is a reduced ring, by [9, Lemma 2.8], P0 and P1 are minimal prime ideals. If
P0 = P1 , then r and a are not in P0 , a contradiction. So by Proposition 18 and Proposition 8 of [20], P0 M
and P1 M are distinct minimal prime submodules. Since M is normal P0 M +P1 M = M and so PM = M,
a contradiction. Therefore, annR (r) + annR (m) = R, that is, M is a locally torsion-free module.

Proposition 2.15. Let M be a finitely generated module satisfying (#) property. If M is a locally torsion-free
module, then M is a normal module.

Proof. Let N1 , N2 be two minimal prime submodules of M. Then by Proposition 2.9, (N1 :R M), (N2 :R
M) are distinct two minimal prime ideals over annR (M). Take r ∈ (N1 :R M) − (N2 :R M). Then it is
easy to see that annR (r) is a subset of (N2 :R M). Since r ∈ (N1 :R M), we get yrn ∈ annR (M) for some
y ∈ R − (N1 :R M) and n ∈ N. Then we get yrn M = 0. Since M is a locally torsion-free module, it is
a reduced module and so r(yM) = 0. Let m ∈ M − N1 . Since N1 is a prime submodule and y is not
in (N1 :R M), we have ym is not in N1 . Thus we have annR (ym ) is a subset of (N1 :R M). Note that
r(ym ) = 0. Then by Lemma 2.13, we have annR (r) + annR (ym ) = R. Since annR (r) + annR (ym ) ⊆
(N2 :R M) + (N1 :R M), we have N1 + N2 = M which completes the proof.

Theorem 2.16. Let M be a finitely generated faithful reduced module satisfying (#) property. Further
assume that for any m in M, annR (m) = annR (a) for some a in R. Then M is a normal module if and
only if M is a locally torsion-free module.

Proof. Follows from Propositions 2.14 and 2.15.


n
Theorem 2.17. Let Mi be an Ri -module for each i = 1, 2, . . . , n. Suppose that R = Ri and M =
i=1

n
Mi . The following statements are equivalent.
i=1

(i) M is a normal R-module.

(ii) Mi is a normal Ri -module for each i = 1, 2, . . . , n.

Proof. First note that M is a finitely generated reduced module if and only if Mi is a finitely generated
reduced Ri -module for each i = 1, 2, . . . , n.
(⇒) : Suppose that M is a normal module and choose i ∈ {1, 2, . . . , n}. Let Pi,1 , Pi,2 be two distinct
minimal prime submodules of Mi . Then N1 = M1 × M2 × · · · × Mi−1 × Pi,1 × Mi+1 × · · · × Mn and
N2 = M1 × M2 × · · · × Mi−1 × Pi,2 × Mi+1 × · · · × Mn are two minimal prime submodules of M. Since
M is normal, we have N1 + N2 = M which implies that Pi,1 + Pi,2 = Mi . Therefore, Mi is a normal
Ri -module.
(⇐) : Suppose that Mi is a normal Ri -module for each i = 1, 2, . . . , n. Let P, K be two minimal prime
submodules of M. Then P = M1 × M2 × · · · × Pt × · · · × Mn and K = M1 × M2 × · · · × Kj × · · · × Mn
for some minimal prime submodules Pt of Mt and Kj of Mj . If j = t, we have P + K = M. So assume
that j = t. Since K = P and Mt is normal, we have Pt = Kt so that Pt + Kt = Mt which implies that
P + K = M. Therefore, M is a normal module.

3. Baer modules, locally torsion-free modules and normal modules

We begin this section by giving a characterization of locally torsion-free modules in terms of (∗)-
submodules and σ -submodules.
6 C. JAYARAM ET AL.

Lemma 3.1. (i) Let M √ be a reduced module and N = O(S) be a proper submodule for some multiplicatively
closed set S of R. Then (N :R M) = (N :R M). In particular, (N :R M) ∩ S = ∅.
(ii) Let M be a finitely generated faithful reduced module satisfying the condition (#). Then minimal
prime submodules are ∗-submodules.

Proof. (i) : It is clear.

(ii) : Let N be minimal prime submodule and let (N :R M) = P. By Proposition 2.9, P is a minimal
prime ideal of R. Denote OP = {m ∈ M : rm = 0 for some r ∈ / P}. First note that, by condition (#),
N = PM. Let m ∈ N. Then m = r1 m1 + · · · + rn mn for some ri ∈ P and mi ∈ M. Since R is a reduced
ring and P is a minimal prime ideal, we get yi ri = 0 for some yi ∈/ P. Now take y = y1 · · · yn . Then
y∈/ P and ym = 0. This yields that m ∈ OP . Also it can be easily seen that OP ⊆ N. i.e., N = OP is a
∗-submodule.

Theorem 3.2. Let M be a finitely generated faithful reduced module satisfying the condition (#). Then M
is a locally torsion-free module if and only if ∗-submodules are σ -submodules if and only if minimal prime
submodules are σ -submodules.

Proof. Suppose M is a locally torsion-free module and let N = O(S) be a ∗-submodule. Let m ∈ N.
Then am = 0 for some a ∈ S. By hypothesis, annR (a) + annR (m) = R. Clearly, annR (a) is contained in
(N :R M), so N is a σ -submodule. By Lemma 3.1 (ii), minimal prime submodules are σ -submodules.
Now, assume that minimal prime submodules are σ -submodules. Suppose rm = 0 and annR (r) +
annR (m) is contained in P for some prime ideal P of R. Then annR (r) is contained in P0 ⊆ P for some
minimal prime ideal P0 of R. Note that r does not belong to P0 . By Proposition 2.9, P0 M is a minimal
prime submodule, so m is in P0 M. Therefore annR (m) + (P0 M :R M) = R. Then by Proposition
8 of [20], (P0 M :R M) = P0 . So P = R, a contradiction. Therefore, M is a locally torsion-free
module.

Definition 3.3. Let N be a submodule of M. Then

(i) N is called an m-submodule if N = (N :R M)M.

(ii) An m-submodule N is said to be a strong m-submodule if all prime submodules minimal over
N are m-submodules.

Lemma 3.4. Suppose that M is a finitely generated reduced R-module. Let N be a strong m-submodule and
N = O(S) for some multiplicatively closed subset S of R. If N is a prime submodule minimal over N, then
N is a minimal prime submodule of M.

Proof. By Proposition 2.10, (N :R M) is a minimal prime ideal over (N :R M). We claim that (N :R
M)∩S
√ = ∅. Suppose that d ∈ (N :R M)∩S. Then there exists d ∈
/ (N :R M) such that dd ∈ (N :R M) as
(N :R M) = (N :R M) and (N :R M) is minimal over (N :R M). Thus dd M ⊆ O(S). It can be easily
shown that d M ⊆ O(S) = N, so d ∈ (N :R M) ⊆ (N :R M) which is a contradiction. Therefore (N :R
M) ∩ S = ∅. We claim that (N :R M) is minimal over annR (M). As M is a finitely generated module,
n n
M = Rmi . So annR (M) = annR (mi ). It is enough, if we show that for any a ∈ (N :R M) there exists
i=1 i=1
d∈/ (N :R M) such that ad ∈ annR (M). Let a ∈ (N :R M). Then there exists b ∈
/ (N :R M) such that
n
ab ∈ (N :R M) and so abM ⊆ N. Since M = Rmi , abmi ∈ O(S) for each i. Then for each i, there exists
i=1
ci ∈ S such that abci mi = 0. Put c = c1 c2 · · · cn . Then c ∈ S and abcmi = 0 for i = 1, 2, . . . , n. Therefore
abc ∈ annR (M). Since (N :R M) ∩ S = ∅, it follows that c ∈ / (N :R M) and so bc ∈ / (N :R M). Put
d = bc. Thus d ∈ / (N :R M) such that ad ∈ annR (M). So (N :R M) is minimal over annR (M).
Consequently, by Proposition 2.10, N is a minimal prime submodule of M.
 COMMUNICATIONS IN ALGEBRA® 7

Lemma 3.5. Let M be a finitely generated reduced module satisfying the condition (P) and N = OP =
{m ∈ M : am = 0 for some a ∈  P ∈ Spec(R), i.e, N is a ∗-submodule. Assume that N is a strong
/ P}, where
m-submodule of M. Then N = OP = Nα , where each Nα is a minimal prime submodule of M.
α∈

Proof. By Lemma 3.1 and [15, Lemma 2.20], radN = N. Then by Lemma 3.4, N = Nα , where each
α∈
Nα is a minimal prime submodule containing N.

Theorem 3.6. Suppose that M is a finitely generated faithful reduced module satisfying the condition (P).
Suppose for each P ∈ Spec(R), OP is a strong m-submodule. Consider the following conditions:

(i) M is a locally torsion-free module.

(ii) OP is a prime submodule for each P ∈ Spec(R).
(iii) Any two distinct minimal prime submodules are comaximal.

Then (i) ⇒ (ii) ⇒ (iii) and (iii) ⇒ (i) if for any m ∈ M, annR (m) = annR (r) for some r ∈ R.

Proof. (i) ⇒ (ii) : Suppose (i) holds. Suppose am ∈ OP with m ∈ / OP , where P is prime ideal. Then
a am = 0 for some a ∈ / P. By (i), annR (a) + annR (a m) = R. Note that annR (a m) ⊆ P, so annR (a) 
P, and thus ra = 0 for some r ∈ / P. Clearly a ∈ (OP :R M). Thus, OP is a prime submodule.
(ii) ⇒ (iii) : Suppose (ii) holds. By Lemma 3.4, for each prime ideal P of R, OP is a minimal prime
submodule. Again for each prime submodule K of M, O(K:R M) is the only minimal prime submodule
contained in K, so any two distinct minimal prime submodules are comaximal.
(iii) ⇒ (i) : Suppose that P is a minimal prime submodule of M. Then P is minimal over O(P:M) . Since
O(P:M) is a strong m-submodule, P is an m-submodule, that is, P = (P :R M)M. Thus, M satisfies the
condition (#). So by Proposition 2.14, M is a locally torsion-free module.
√
Lemma 3.7. Let M be a reduced module and N be a Baer submodule. Then (N :R M) = (N :R M).
√
Proof. Let a ∈ (N :R M). Then ak ∈ (N :R M) for some k ∈ N. This implies that ak m ∈ N for each
m ∈ M. Since N is a Baer submodule and annR (ak m) ⊆ annR (am), we get am ∈ annM (annR (am)) ⊆
annM (annR (ak m)) ⊆ N, and so a ∈ (N :R M).

Definition 3.8. Let M be an R-module and N a proper σ -submodule of M.

(i) N is called a maximal σ -submodule if N is a maximal element in σ (M).

(ii) N is called a prime σ -submodule if N1 ∩ N2 ⊆ N implies that either N1 ⊆ N or N2 ⊆ N for every
N1 , N2 ∈ σ (M).
Note that a maximal σ -submodule is a prime σ -submodule and if M is finitely generated, then every
σ -submodule is contained in a maximal σ -submodule.
Lemma 3.9. Let M be a finitely generated reduced R-module. Consider the following cases:

(i) Every minimal prime submodule is a σ -submodule.

(ii) Every minimal prime submodule is a maximal σ -submodule.
(iii) Every maximal submodule contains a unique minimal prime submodule.
(iv) M is a normal module, i.e, every prime submodule contains a unique minimal prime submodule.

Then (i) ⇒ (ii) ⇒ (iii) ⇒ (iv).

Proof. (i) ⇒ (ii) : Let N be a minimal prime submodule. Then by (i), N is a σ -submodule. Since
M is finitely generated and N is a proper submodule, there exists a proper maximal σ -submodule
8 C. JAYARAM ET AL.

K containing N. Let m ∈ K. Assume that m ∈ / N. This implies that annR (m) ⊆ (N :R M) since N is a
prime submodule. As K is a σ -submodule and m ∈ K, it follows that annR (m) + (K :R M) = R and this
gives (N :R M) + (K :R M) = (K :R M) = R and so K = M which is a contradiction. So that N = K, i.e,
N is a maximal σ -submodule.
(ii) ⇒ (iii) : Let Q be a maximal submodule of M. Assume that Q contains two distinct minimal
prime submodules K, N of M. By assumption K, N are maximal σ -submodules. Since K + N is also a
σ -submodule, we have K + N = M ⊆ Q, a contradiction.
(iii) ⇒ (iv) : It is clear.

Theorem 3.10. Let M be a finitely generated faithful reduced module satisfying (#) property. Further
assume that for any m in M, annR (m) = annR (a) for some a in R. Then the following statements are
equivalent:

(i) Every maximal submodule contains a unique minimal prime submodule.

(ii) M is a normal module.
(iii) M is a locally torsion-free module.
(iv) Every ∗-submodule is a σ -submodule.
(v) Every minimal prime submodule is a σ -submodule.
(vi) Every minimal prime submodule is a maximal σ -submodule.

Proof. Follows from Lemmas 3.9, 2.13, Theorems 3.2 and 2.16.
Now, we characterize σ -submodules in terms of ∗-submodules.
Proposition
 3.11. Let M be an R-module. Then a submodule N of M is a σ -submodule if and only if
N = OQ , where Q denotes any maximal ideal containing (N :R M).
Q

Proof. ⇒: Let m ∈ N. Since N is a σ -submodule, it follows that annR (m) + (N

:R M) = R. Let P be a
maximal ideal contaning (N :R M). Then annR (m)  P and so m ∈ OP , i.e, m ∈ OQ , where Q denotes
 Q
any maximal ideal containing (N :R M). Conversely, assume that m ∈ OQ , where Q is any maximal
Q
ideal containing (N :R M). Let P be a maximal ideal of R. If P contains (N :R M), then m ∈ OP and so
rm = 0 for some r ∈/ P. This implies that annR (m)P = RP . Otherwise, we would have (N :R M)  P and
so (N :R M)P = RP . In both cases we have (annR (m) + (N :R M))P = RP for all maximal ideal P of
R and so annR (m) + (N :R M) = R. This gives x + y = 1 for some x ∈ annR (m) and y ∈ (N :R M) and
so m = xm + ym = ym ∈ N. Thus N = OQ .
 Q
⇐: Assume that N = OQ , where Q denotes any maximal ideal containing (N :R M). Now we show
Q
that N is a σ -submodule of M. Let m ∈ N. Assume that annR (m) + (N :R M)
 = R. Then there exists a
maximal ideal Q of R such that annR (m)+(N :R M) ⊆ Q . Since m ∈ N = OQ , we have m ∈ OQ and
Q
so r m = 0 for some r ∈
/ Q , i.e, annR (m)  Q which is a contradiction.

Recall from [13], that a submodule N of M is a complemented submodule if there exists a submodule
K of M such that N + K = M and N ∩ K = 0. We say that a submodule N of M is a locally complemented
if NQ is a complemented submodule of RQ -module MQ for each Q ∈ Max(R).

Proposition 3.12. Let N be a σ -submodule of M. Then N is a locally complemented submodule.


Proof. Assume that N is a σ -submodule of M. Then by previous proposition, N = OQ , where
Q
Q denotes any maximal ideal containing (N :R M). Let P be a maximal ideal of R. If (N :R M)  P, then
 COMMUNICATIONS IN ALGEBRA® 9

NP = MP . Assume that (N :R M) ⊆ P. Let m ∈ N, then m ∈ OP and so rm = 0 for some r ∈

/ P, i.e,
m
1 = 0. Thus NP = 0P .

Recall that an R-module M is called a colon distributive module if (N + K :R M) = (N :R M) +

(K :R M) for each N, K ∈ L(M). In fact, colon distributive module was first studied by Smith in [24]
as μ-module. Note that a finitely generated module M is a colon distributive module if and only if M is
a multiplication module [24, Theorem 3.8]. We call an R-module M a locally faithful (locally colon
distributive) module if RQ -module MQ is a faithful (colon distributive) for all maximal ideal Q of R. Note
that if M is a finitely generated module, then M is locally faithful (locally colon distributive) module if
and only if it is a faithful module (colon distributive).

Lemma 3.13. Let M be a locally faithful and locally colon distributive R-module. If N is a locally
complemented submodule, then N is a σ -submodule.

Proof. Let N be a locally complemented submodule of a locally faithful and locally colon distributive
R-module M. Assume that N is not a σ -submodule. Then there exists m ∈ N such that m ∈ / OQ for
some maximal ideal Q ⊇ (N :R M). This implies that for all r ∈ / Q, rm = 0 and so m1 = 0 and so
NQ = 0. Since (N :R M) ⊆ Q, we have NQ = MQ . As N is a locally complemented and MQ is a colon
distributive, by [13, Lemma 3], NQ = es MQ for some weak idempotent es ∈ RQ . Since MQ is faithful, then
e
s is idempotent in RQ . Note that only idempotents are 0Q , 1Q in RQ and so NQ = MQ or NQ = 0Q , a
contradiction.

Theorem 3.14. Let M be a locally faithful and locally colon distributive module. Then the following
statements are equivalent for any submodule N of M:

(i) N is a locally complemented submodule.

(ii) N is aσ -submodule.
(iii) N = OQ , where Q denotes any maximal ideal containing (N :R M).
Q

Proof. Follows from Lemma 3.13 and Proposition 3.12.

Proposition 3.15. Let M be a faithful colon distributive normal module. Further assume that for any m
in M, annR (m) = annR (a) for some a in R. Then a submodule N of M is a maximal σ -submodule if and
only if N is a minimal prime submodule.

Proof. Let N be a minimal prime submodule. Then by Theorem 3.10, N is a maximal σ -submodule.
Let N be a maximal σ -submodule of M. First note that M is a multiplication module since it is
colon distributive. Take a maximal submodule K containing N. Then Q = (K :R M) is a maximal
ideal containing (N :R M). Since N is a σ -submodule, NQ is a complemented in RQ -module MQ by
Proposition 3.12 and so by the proof of Lemma 3.13, NQ = 0. This implies that N ⊆ OQ . Since OQ is
a ∗-submodule and M is a normal module, so OQ is a σ -submodule of M by Theorem 3.10. By the
maximality of N, N = OQ . Since MQ is torsion-free, it follows that N = OQ is a prime submodule.
Let N be a minimal prime submodule of M which is contained in N. Then N is a σ -submodule by
Theorem 3.10. Let m ∈ N − N . Then annR (m) ⊆ (N :R M) and also annR (m) + (N :R M) = R =
(N :R M) + (N :R M) = (N :R M) a contradiction so that N = N .

Proposition 3.16. Let M be a finitely generated R-module. Then the following statements are satisfied:

(i) If N is a prime submodule and also a σ -submodule, then N is a prime σ -submodule.

(ii) If M = T(M) and N is a prime submodule which is also a σ -submodule, then (N :R M) is a prime
σ -ideal.
10 C. JAYARAM ET AL.

Proof. (i) : Let N be a prime submodule which is also a σ -submodule. Since M is finitely generated, there
exists a maximal σ -submodule K containing N. Then (N :R M) ⊆ (K :R M). Let m ∈ K − N. Since N is
a prime submodule, it follows that annR (m) ⊆ (N :R M). Also note that annR (m) + (K :R M) = R since
K is a σ -submodule. Thus we have (N :R M) + (K :R M) = (K :R M) = R which is a contradiction.
Hence N = K is a maximal σ -submodule so that it is a prime σ -submodule.
(ii) : By [14, Proposition 3.18], (N :R M) is a σ -ideal. Since N is a prime submodule, (N :R M) is a
prime ideal. By similar argument in (i), (N :R M) is a maximal σ -ideal so that it is a prime σ -ideal.

Proposition 3.17. Let M be a finitely generated non-torsion module satisfying (#) property and R is a
reduced ring. Then every minimal prime submodule is a Baer submodule.

Proof. Suppose that N is a minimal prime submodule. By Proposition 2.9, (N :R M) is a minimal prime
ideal. So (N :R M) is a ∗-ideal. Since M is a non-torsion module and N is a prime m-submodule, by [15,
Lemma 2.5], N is a ∗-submodule. Then by [15, Proposition 2.7], N is a Baer submodule.
We now characterize Baer modules in terms of ∗-submodules and σ -submodules.
Theorem 3.18. Let M be a finitely generated non-torsion reduced multiplication module. Then M is a Baer
module if and only if every Baer submodule is a ∗-submodule and every ∗-submodule is a σ -submodule.

⇒: Suppose that M is a Baer module. Let N be a Baer submodule. Then by [15,

Proof.   Lemma 2.23],
N= Nα , where Nα ’s are prime Baer submodules. This implies that (N :R M) = (Nα :R M). By
α∈ α∈
[15, Lemma 2.24], (Nα :R M)’s are prime Baer ideals. Since M is finitely generated faithful Baer module,
by [14, Proposition 2.12], R is a Baer ring so is a quasi-regular ring. Then by [12, Theorem 3], (Nα :R M)’s
are minimal prime ideals. Again by [12, Corollary 3], (N :R M) is a ∗-ideal. Suppose that (N :R M) =
−−→
O(S) for some multiplicatively closed subset S of R. Then (N :R M) ∩ S = ∅. Let O(S) = {m ∈ M :
−−→ n
rm = 0 for some r ∈ S}. Now we will show that N = O(S). Let m ∈ N. Then m = ai mi , where
i=1
−−→
ai ∈ (N :R M). So ai si = 0 for some si ∈ S. Put s = s1 s2 · · · sn . Then sm = 0. So that N ⊆ O(S). Note
−−→
that by [16, Theorem 3], M is a locally torsion-free module. Suppose that m ∈ O(S). Then rm = 0 for
some r ∈ S. This implies that annR (r) + annR (m) = R and so 1 = s + r for some s ∈ annR (r) and
r ∈ annR (m). Thus m = sm. Since sr = 0, s ∈ O(S) = (N :R M). So m ∈ (N :R M)M = N. Then we
−−→
have N = O(S).Thus N is a ∗-submodule. By Theorem 3.10, every ∗-submodule is a σ -submodule. The
converse part follows from [14, Theorem 3.6].

4. Topological characterizations of normal modules

Let R be a ring and I be an ideal of R. The variety of I, denoted by VR (I), is defined by VR (I) = {p ∈
Spec(R) : I ⊆ p}. Then the collection ζ (R) = {VR (I) : I ≤ R} satisfies the axioms for closed sets
in a topological space. The resulting topology is known as the Zariski topology on Spec(R). For any
element r ∈ R, the set Dr = Spec(R) − VR (rR) is open in Spec(R) and it is well-known that the family
F = {Dr : r ∈ R} forms a base for the Zariski topology on Spec(R).
The set of minimal prime ideals of a ring R will be denoted by Min(R). We consider Min(R) as a
subspace of Spec(R) with the Zariski topology. For an ideal I of R, we denote the set VR (I) ∩ Min(R) by
VRo (I). Clearly, {VRo (I) : I ≤ R} is the family of closed sets of Min(R) and the family {Dor = Dr ∩ Min(R) :
r ∈ R} forms a base for the Zariski topology on Min(R). More information about the space of minimal
prime ideals can be found in [1] and [9].
The set of prime submodules of an R-module M is called the prime spectrum of M and denoted by
Spec(M). The Zariski topology on Spec(M) is described by taking the set {V(N) : N ≤ M} as the family
of closed sets of X = Spec(M) where V(N) = {P ∈ Spec(M) : (N :R M) ⊆ (P :R M)}.
 COMMUNICATIONS IN ALGEBRA® 11

Let M be an R-module. The ring R/annR (M) will be denoted by R and I will denote the ideal
I/annR (M) where I is an ideal of R containing annR (M). The map ψ : Spec(M) −→ Spec(R), defined
by ψ(P) = (P :R M), is called the natural map of Spec(M) [21]. If M is a finitely generated R-module,
then the natural map ψ is surjective by [21, Proposition 3.5].

Lemma 4.1. Let M be an R-module and ψ : Spec(M) −→ Spec(R) be the natural map.

(1) [21, Proposition 3.1] ψ is continuous for the Zariski topologies; more precisely, ψ −1 (VR (I)) = V(IM)
for every ideal I of R containing annR (M).
(2) [21, Theorem 3.6] If ψ is surjective, then ψ is both open and closed map; more precisely, for every
N ≤ M, ψ(V(N)) = VR ((N :R M)) and ψ(Spec(M) − V(N)) = Spec(R) − VR ((N :R M)).
(3) [21, Proposition 4.3] Let M be an R-module and Xr = Spec(M) − V(rM) for any r ∈ R. Then the set
{Xr : r ∈ R} forms a base for the Zariski topology on Spec(M).
(4) [21, Proposition 4.1] Let M be an R-module, r ∈ R and ψ : Spec(M) −→ Spec(R) be the natural map.
Then

(a) ψ −1 (Dr ) = Xr
(b) ψ(Xr ) ⊆ Dr . If ψ is surjective, then ψ(Xr ) = Dr .

The set of all minimal prime submodules of an R-module M will be denoted by Min(M). Consider
Min(M) as a subspace of Spec(M) with the Zariski topology. For a submodule N of M, we denote the
set V(N) ∩ Min(M) by V o (N). Clearly, {V o (N) : N ≤ M} is the family of closed sets of Min(M).
Since the family {Xr : r ∈ R} is a base for the Zariski topology on Spec(M), the subspace Min(M) has
{Xro = Xr ∩ Min(M) : r ∈ R} as a base of open sets.
Let M be a finitely generated R-module satisfying condition (#). Then, by Proposition 3, we can define
the map ψ o : Min(M) −→ Min(R) by ψ o (P) = (P :R M). Clearly, ψ o is the restriction map of the
natural map ψ on Min(M). Also, ψ o is surjective by Proposition 3.

Lemma 4.2. Let M be a finitely generated R-module satisfying condition (#). Consider the map ψ o :
Min(M) −→ Min(R) defined as above. Then the following statements are true.

(1) The map ψ o is continuous.

(2) The map ψ o is both closed and open.
(3) (ψ o )−1 (Dor ) = Xro for any r ∈ R.
(4) ψ o (Xro ) = Dor for any r ∈ R.
(5) Assume that R is a reduced ring. Then, Xro = V o (IM) where I/annR (M) = annR (r). In particular,
V o (rM) and Xro are disjoint open and closed sets.

Proof. (1) This is clear since ψ o is the restriction of the continuous map ψ : Spec(M) −→ Spec(R).
(2) Let N ≤ M. Then we have ψ o (V(N) ∩ Min(M)) = ψ(V(N) ∩ Min(M)) = ψ(V(N)) ∩
ψ(Min(M)) = VR ((N :R M)) ∩ Min(R) by Lemma 4.1-(2). Thus ψ o is a closed map. Similarly,
ψ o (Min(M) − (Min(M) ∩ V(N))) = ψ(Min(M) − (Min(M) ∩ V(N))) = ψ(ψ −1 (Min(R)) −
ψ −1 (Min(R) ∩ VR ((N :R M)))) = ψψ −1 (Min(R) − (Min(R) ∩ VR ((N :R M)))) = Min(R) − (Min(R) ∩
VR ((N :R M))) since ψ is surjective. Thus ψ o is an open set.
(3) (ψ o )−1 (Dor ) = ψ −1 (Dr ∩ Min(R)) = ψ −1 (Dr ) ∩ ψ −1 (Min(R)) = Xr ∩ Min(M) = Xro by
Lemma 4.1-(4).
(4) Since ψ o is surjective, this follows from part (3).
(5) By part (3), we have Xro = (ψ o )−1 (Dor ). By [9, Theorem 2.3], Dor = VRo (annR (r)). There exists
an ideal I of R containing annR (M) such that annR (r) = I/annR (M). Thus Dor = VRo (annR (r)) =
VR (annR (r)) ∩ Min(R) = VR (I) ∩ Min(R). It follows that Xro = (ψ o )−1 (Dor ) = (ψ o )−1 (VR (I) ∩
12 C. JAYARAM ET AL.

Min(R)) = ψ −1 (VR (I) ∩ Min(R))) = ψ −1 (VR (I)) ∩ ψ −1 (Min(R)) = V(IM) ∩ Min(M) = V o (IM) by
Lemma 4.1-(1).

Corollary 4.3. Let M be a finitely generated R-module satisfying condition (#) and R be a reduced ring.
Then Min(M) is a Hausdoff space with a base of clopen sets.

Proof. Let P and P be distinct minimal prime submodules of M. Since M satisfies the condition (#),
(P :R M) = (P :R M). Let a ∈ (P :R M) − (P :R M). Then, by Lemma 4.2-(5), V o (aM) and Xao are
disjoint open sets containing P and P , respectively. Hence Min(M) is a Hausdorff space.

Theorem 4.4. Let M be a finitely generated multiplication R-module. Then the following are equivalent.

(1) M is a Baer module.

(2) M is a normal module and Min(M) is a compact space.
(3) R is a reduced ring and there exists a continuous function δ from Spec(M) onto Min(R) which is the
identity on Min(M).

Proof. (1) ⇒ (2) M is a normal module by Remark 1. R is a Baer ring by [14, Proposition 2.12].
Since M is a reduced R-module, R is a reduced ring. According to [1, Theorem on p. 83], Min(R) is a
compact space. Since {Xro : r ∈ R} is a base for the Zariski topology on Min(M), for any open cover of
Min(M), there is a family {rλ : λ ∈ } of elements of R such that Min(M) ⊆ ∪λ∈ Xroλ . By Lemma 4.2-
(4), ψ o (Min(M)) = Min(R) ⊆ ∪λ∈ ψ o (Xroλ ) = ∪λ∈ Dorλ . Since Min(R) is a compact space, there
exists a finite subset of such that Min(R) ⊆ ∪λ∈ Dorλ whence (ψ o )−1 (Min(R)) = Min(M) ⊆
o −1
∪λ∈ (ψ ) (Drλ ) = ∪λ∈ Xroλ by Lemma 4.2. Thus Min(M) is a compact space.
o

(2) ⇒ (3) By Proposition 2.12, R is a normal ring, hence it is reduced and every prime ideal of R
contains a unique minimal prime ideal. Also, Min(R) is the image of the compact space Min(M) under
the continuous function ψ o . So Min(R) is also a compact space. By [1, Theorem on p. 83], there exists a
continuous function ϕ from Spec(R) onto Min(R) which is the identity on Min(R). Consider the natural
map ψ : Spec(M) −→ Spec(R). Since ψ is continuous, the function ϕoψ is a continuous map from
Spec(M) onto Min(R) which is the identity on Min(M).
(3) ⇒ (1) Define the map φ : Spec(R) −→ Spec(M) by φ(P) = PM. Then it can be easily seen
that φ −1 (V(N)) = VR ((N :R M)) for any submodule N of M. Thus φ is a continuous map. By (3), there
is a continuous function δ : Spec(M) −→ Min(R) which is the identity on Min(M). It follows that
δoφ : Spec(R) −→ Min(R) is a continuous map which is the identity on Min(R). By [1, Theorem on p.
83], R is a Baer ring. Hence, M is a Baer module by [14, Theorem 2.14].

Acknowledgments
We would like to thank the referee for his/her valuable comments that improved the paper.

References
[1] Artico, G., Marconi, U. (1976). On the compactness of minimal spectrum. Rend. Sem. Mat. Univ. Padova 56:79–84.
[2] Anderson, D. D., Arabaci, T., Tekir, Ü., Koç, S. (2020). On S-multiplication modules. Commun. Algebra 48(8):3398–
3407.
[3] Anderson, D. D., Chun, S. (2014). The set of torsion elements of a module. Commun. Algebra 42(4):1835–1843.
[4] Atiyah, M. F., Macdonald, I. G. (1969). Introduction to Commutative Algebra. Boston: Addison Wesley Publishing
Company. Inc.
[5] Barnard, A. (1981). Multiplication modules. J. Algebra. 71(1):174–178.
[6] Birkenmeier, G. F., Kim, J. Y., Park, J. K. (2001). Principally quasi-Baer rings. Commun. Algebra 29(2):639–660.
[7] El-Bast, Z. A., Smith, P. P. (1988). Multiplication modules. Commun. Algebra 16(4):755–779.
[8] Evans, M. (1972). On commutative PP rings. Pacific J. Math. 41(3):687–697.
 COMMUNICATIONS IN ALGEBRA® 13

[9] Henriksen, M., Jerison, M. (1965). The space of minimal prime ideals of a commutative ring. Trans. Amer. Math.
Soc. 115:110–130.
[10] Huckaba, J. A. (1990). Commutative Rings with Zero Divisors. New York Basel: Marcel Dekker.
[11] Huijsmans, C. B. (1974). Some analogies between commutative rings, Riesz spaces and distributive lattices with
smallest element. Indag. Math. (Proceedings) 77(2):132–147.
[12] Jayaram, C. (1984). Baer ideals in commutative semiprime rings. Indian J. Pure Appl. Math 15(8):855–864.
[13] Jayaram, C., Tekir, Ü. (2018). von Neumann regular modules. Commun. Algebra 46(5):2205–2217.
[14] Jayaram, C., Tekir, Ü., Koç , S. (2022). On Baer modules. Rev. Union Mat. Argent 63(1):109–128.
[15] Jayaram, C., Tekir, Ü., Koç , S. (2021). On Quasi regular modules and trivial extension. Hacettepe J. Math. Stat.
50(1):120–134.
[16] Jayaram, C., Uğurlu, E. A., Tekir, Ü., Koç , S. (2022). Locally torsion-free modules. J. Algebra Appl. 2350103.
[17] Kist, J. (1963). Minimal prime ideals in commutative semigroups. Proc. London Math. Soc. 3(1):31–50.
[18] Lee, T. K., Zhou, Y. (2004). Reduced modules. In: Facchini, A., Houston, E., Salce, L., eds. Rings, Modules, Algebras
and Abelian Groups, Vol. 236, Boca Raton, FL: CRC Press, pp. 365–377.
[19] Lu, C. P. (1990). M-radicals of submodules in modules II. Math. Japonica 35:991–1001.
[20] Lu, C. P. (1984). Prime submodules of modules. Rikkyo Daigaku sugaku zasshi 33(1):61–69.
[21] Lu, C. P. (1999). The Zariski topology on the prime spectrum of a module. Houston J. Math. 25(3):417–425.
[22] Northcott, D. G. (2008). Lessons on Rings, Modules and Multiplicities. Cambridge: Cambridge University Press.
[23] Rizvi S. T., Roman,C. S. (2004). Baer and quasi-Baer modules. Commun. Algebra 32(1):103–123.
[24] Smith, P. F. (2014). Mappings between module lattices. Int. Eletron. J. Algebra 15:173–195.
[25] Von Neumann, J. (1936). On regular rings. Proc. Natl. Acad. Sci. 22(12):707–713.