Journal of Mathematical Sciences, Vol. 123, No.

2, 2004

MULTIPLICATION MODULES
A. A. Tuganbaev UDC 512.553

1. Preliminaries
All rings are assumed to be associative and with nonzero identity element. Unless otherwise stated all
modules are unital right modules. Expressions such as “a Noetherian ring” mean that the corresponding
right and left conditions hold. For a ring A, a right A-module M is called a multiplication module if for
every submodule N of M , there exists an ideal B of A such that N = M B.
There are many works containing results on multiplication modules over commutative rings; for
example, see [1–3, 5–7, 10–23, 27, 29–32, 35–46, 48–65, 68–74, 76, 78, 79, 81, 84, 96, 97].
This work is primarily a survey of results on multiplication modules over not necessarily commutative
rings. In the noncommutative case, there exist several works devoted to multiplication modules and
noncommutative generalizations of commutative rings whose ideals are multiplication modules (see [77,
80, 85, 86, 88, 89, 91–95]).
In this work, we often consider multiplication modules over invariant rings. (A ring A is said to be
right invariant (resp. left invariant) if each of its right (resp. left) ideals is an ideal of A.) For example,
if A is a right invariant ring, then each of its cyclic right modules is a multiplication module (Proposition
3.1), and each of its ideals generated by idempotents is a multiplication right (left) module (Proposition
3.4). In addition, every invertible ideal of an invariant ring A is a finitely generated multiplication right
(left) A-module by Proposition 3.14. If {Fi }i∈I is an infinite set of fields and A is the direct product of
all fields Fi , then we note that the ideal ⊕i∈I Fi of A is a multiplication A-module that is not finitely
generated.
In the case of noncommutative invariant rings, several additional difficulties arise. One of the diffi-
culties is that an invariant ring A with a maximal ideal M does not necessarily have the right localization
AM with respect to M , although A\M is a multiplicatively closed set, and for any two elements a ∈ A and
s ∈ A \ M , there exist elements b ∈ A and t ∈ A \ M such that at = sb. In Proposition 6.13, we construct
an invariant semilocal distributive Bezout ring A with a maximal ideal M such that the right localization
AM does not exist, although all finitely generated right (resp. left) ideals of A are multiplication right
(resp. left) ideals. (A right ideal B of a ring A is called a multiplication right ideal if B is a multiplication
right A-module. A module is called a Bezout module if each of its finitely generated submodules is a
cyclic module.

A
module is

said to be distributive if the lattice of its submodules is distributive, i.e.,
(X + Y ) Z = X Z + Y Z for any of its submodules X, Y , and Z.)
In the case of modules over commutative rings, many main results of the presented work turn into
modifications of the results proved in [3–5,17,78]. In the case of modules over noncommutative rings, the
main results of the presented work are proved in [88] and [89].
For a module M , we denote by J(M ), End(M ), Soc(M ), Lat(M ), and max(M ) the Jacobson radical,
the endomorphism ring, the socle, the lattice of all submodules, and the set of all its maximal submodules
of M , respectively.
The following Notes 1.1, 1.2, and 1.3 are directly verified.
1.1. Note. (1) Every simple module is a multiplication module.
(2) Every nonzero multiplication module over a simple ring is a simple module.
Translated from Itogi Nauki i Tekhniki, Seriya Sovremennaya Matematika i Ee Prilozheniya. Tematicheskie Obzory.
Vol. 114, Algebra, 2002.

1072–3374/04/1232–3839 
c 2004 Plenum Publishing Corporation 3839
 For a subset X of a right module M over a ring A, the right ideal {a ∈ A | Xa = 0} of A is called
the right annihilator of X in A; it is denoted by rA (X) or r(X).
1.2. Note. For a right module M over a ring A, the following conditions are equivalent.
(1) M is a multiplication module.
(2) For every ideal B of A such that B ⊆ r(M ), the A/B-module M is a multiplication module.
(3) There exists an ideal B of A such that B ⊆ r(M ) and M is a multiplication A/B-module.
For two subsets X and Y of a right module M over a ring A, the subset {a ∈ A | Xa ⊆ Y } of A is
denoted by (Y : X). If Y is a submodule of M , then it is directly verified that for any subset X of M ,
the set (Y : X) is a right ideal of A. For any two submodules X and Y of M , it is directly verified that
(Y : X) is an ideal of A.
1.3. Note. For a right module M over a ring A, the following conditions are equivalent.
(1) M is a multiplication module.
(2) N ⊆ M (N : M ) for every submodule N of M .
(3) N = M (N : M ) = M r(M/N ) for every submodule N of M .
1.4. Note. Every homomorphic image of a multiplication module is a multiplication module.
Proof. Let M be a multiplication right module over a ring A, h : M → M be an epimorphism and N be
a submodule of M . Then there exists a submodule N of M with h(N ) = N . By assumption, there exists
an ideal B of the ring A such that N = M B. Then N = h(N ) = h(M B) = h(M )B = M B and M is a
multiplication module.

1.5. Note. For a multiplication right module over a ring A, the following assertions hold.
(1) Every submodule of M is a fully invariant
submodule of M .
(2) If N is a submodule of M such that N M B = N B for every ideal B of A, then N is a
multiplication module.
Proof. (1) Let X be a submodule of M and let f be an endomorphism of M . There exists an ideal B of
A such that X = M B. Then f (X) = f (M B) = f (M )B ⊆ M B = X.
(2) Let Y be a submodule of N . Since

M is a multiplication module, there exists an ideal B of A
such that Y = M B. By assumption, N M B = N B. Therefore,
 
Y = MB = Y MB ⊆ N M B = N B ⊆ M B = Y.
Then Y = N B and N is a multiplication module.

1.6. Note. Every endomorphic image of a multiplication module is a fully invariant multiplication
submodule of the module.
Proof. The proof follows from Note 1.4 and Note 1.5(1).

1.7. Note. Every direct summand of a multiplication module is a fully invariant multiplication submodule
of the module.
Proof. Since every direct summand of a module is an endomorphic image of the module, the assertion
follows from Note 1.6.

1.8. Note. In the endomorphism ring of every homomorphic image of any multiplication module, all
idempotents are central.
Proof. Let M be a multiplication right module over a ring A, M be a homomorphic image of M ,
R = End(M ), and f be an idempotent of the ring R. By Note 1.4, M is a multiplication module. By
Note 1.5(1),
 
(1 − f )Rf (M) ⊆ f (M ) (1 − f )(M ) = 0 and f R(1 − f )(M ) ⊆ (1 − f )(M ) f (M) = 0.
Therefore, (1 − f )Rf = f R(1 − f ) = 0 and f is a central idempotent of R.

3840
1.9. Note. If M is a multiplication right module over a ring A and P is an ideal of A such that
M = M P , then there exists a cyclic submodule X of M such that P does not contain the annihilator of
the module M/X.
Proof. Since M = M P , there exists a cyclic submodule X of M that is not contained in the module
M P . Since M is a multiplication module, there exists an ideal B of A such that X = M B. Then B ⊆ P ,
since X ⊆ M P . Since X = M B, we have B ⊆ r(M/X); therefore, P does not contain r(M/X).

1.10. Note. Let M be a multiplication right module over a ring A, and let P be a maximal ideal of A.
(1) If M = M P , then the module M/M P is simple and there exists a cyclic submodule X of M such
that A = P + r(M/X).
(2) M/M P is a cyclic module with at most two submodules and either M = M P or M P is a maximal
submodule of M .
Proof. (1) By Note 1.4, M/M P is a multiplication module over the simple ring A/P . Since M/M P =
0, it follows from Note 1.1(2) that the module M/M P is simple. By Note 1.9, there exists a cyclic
submodule X of M such that P does not contain r(M/X); in addition, P is a maximal ideal. Therefore,
A = P + r(M/X).
(2) By Note 1.4, M/M P is a multiplication module over the simple ring A/P . If M/M P = 0, then
the zero module M/M P is a cyclic module with exactly one submodule. If M/M P = 0, then by (1), M P
is a maximal submodule of M and M/M P is a cyclic module with exactly two submodules.

1.11. Note. Let A be a ring such that the lattice

of two-sided

ideals of A is distributive, and let M be a
multiplication right A-module such that M B M C = M (B C) for any two ideals B and C of A. Then
M is a distributive module.
Proof. Let X, Y , and Z be three submodules of M . Since M is a multiplication module, there exist
ideals B, C, and D of A such that X = M B, Y = M C, and Z = M D. Then
(X + Y ) ∩ Z = (M B + M C) ∩ M D = (M (B + C)) ∩ (M D) = M ((B + C) ∩ D)
= M (B ∩ D + C ∩ D) = M (B ∩ D) + M (C ∩ D) = M B ∩ M D + M C ∩ M D = X ∩ Z + Y ∩ Z.

1.12. Note. Let A be a ring with commutative multiplication of ideals, M be a multiplication right
A-module, and B be an ideal of A such that M = M B. Then the following assertions hold.
(1) N = N B for every submodule N of M .
(2) For every element m of M , there exists an element b of the ideal B such that m(1 − b) = 0.
Proof. (1) By assumption, M = M B. Since M is a multiplication module, there exists an ideal C of A
such that N = M C = M BC. In addition, BC = CB, since A is a ring with a commutative multiplication
of ideals. Therefore N = M CB = N B.
(2) For any element m ∈ M , it follows from (1) that mA = mAB = mB. Therefore, there exists an
element b of B such that m(1 − b) = 0.

1.13. Note. Let A be a ring with commutative multiplication of ideals, M be a multiplication right
A-module, and P be a maximal ideal of A. Then the following conditions are equivalent.
(1) M = M P .
(2) N = N B for every submodule N of M .
(3) X = XP for every cyclic submodule of M .
(4) P does not contain the annihilator of any cyclic submodule of M .
Proof. The implication (1)=⇒(2) follows from Note 1.12(1).
The implications (2)=⇒(3) and (3)=⇒(1) are obvious.
(3)=⇒(4) Assume that P contains the annihilator of some cyclic submodule X of M . By assumption,
X = XP = 0. Then A = r(X) ⊆ P ; this is a contradiction.

3841
 (4)=⇒(3) Let X be a cyclic submodule of M . Since r(X) ⊆ P and the ideal P is maximal, A =
P + r(X). Therefore,
X = XA = X(P + r(X)) = XP + Xr(X) = XP.

1.14. Note. Let M be a multiplication module over a commutative ring A. Then the endomorphism ring
End(M ) of the A-module M is a commutative ring naturally containing the ring A/r(M ), and the lattice
of all submodules of the A-module M coincides with the lattice of all submodules of the End(M )-module
M.
Proof. Let f and g be two endomorphisms of M , and let m be an element of M . By Note 1.12(1),
f (m), g(m) ∈ mA. Therefore, there exist elements a, b ∈ A such that f (m) = ma and g(m) = mb.
Therefore,
(f g − gf )(m) = f (g(m)) − g(f (m)) = f (m)b − g(m)a = mab − mba = 0
and End(M ) is a commutative ring. For every element a of the commutative ring A, the mapping ϕa : m →
ma is an endomorphism of the A-module M . It is directly verified that the mapping ϕ : A → End(MA )
defined by ϕ(a) = ϕa is a ring homomorphism with kernel r(M ). Therefore, every End(M )-submodule of
M is an A-submodule of M . By Note 1.12(1), every A-submodule of M is an End(M )-submodule of M .


We present the necessary notation and definitions. For any module homomorphism f : Q → M and
each submodule N of M , we denote by f −1(N ) the submodule {q ∈ Q | f (q) ∈ N } of M . For a module
M and any two submodules X and Y of M with Y ⊆ X, the factor module X/Y is called a subfactor of
M.
A ring A is said to be right quasi-invariant if each of its maximal right ideals is an ideal of A. A
module M is called an invariant (resp. quasi-invariant) if each of its submodules (resp. each of its maximal
submodules) is a fully invariant submodule of M . It is directly verified that a ring A is right invariant
(resp. right quasi-invariant) if and only if A is an invariant (resp. quasi-invariant) right A-module.
A right A-module M is said to be faithful if r(M ) = 0.
A module is said to be uniform if any two of its nonzero submodules have a nonzero intersection.
A module is said to be uniserial if any two of its submodules are comparable with respect to inclusion.
A module is said to be finite-dimensional if it does not contain a submodule that is a direct sum of
infinitely many nonzero modules.
A module M is said to be regular if each of its cyclic submodules is a direct summand of M . A ring
A is said to be regular if a ∈ aAa for every element a of A. It is well known that a ring A is regular ⇐⇒
A is a regular right A-module ⇐⇒ A is a regular left A-module. A ring A is said to be strongly regular if
a ∈ a2 A for every element a of A.
A ring is said to be semiprime (resp. reduced) if each of its nilpotent ideals (resp. elements) is equal
to zero. A ring is said to be normal if all its idempotents are central.
A module is said to be semi-Noetherian if every its nonzero submodule has a maximal submodule.
Let M and N be two right modules over a ring A. The module M is said to be projective with respect
to N (or N -projective) if for every epimorphism h : N → N and each homomorphism f : M → N , there
exists a homomorphism f : M → N such that f = hf . A module is called quasiprojective if it is projective
with respect to itself. The module M is said to be projective if M is projective with respect to every right
A-module. It is well known that M is projective if and only if M is isomorphic to a direct summand of a
direct sum of some set of isomorphic copies of the module AA .
A module is said to be semihereditary (resp. hereditary) if all of its finitely generated submodules
(resp. all its submodules) are projective.

A submodule N of a module M is said to be essential if N X = 0 for every nonzero submodule X
of M . In this case, M is called an essential extension of N .

3842
 A submodule N of a module M is called a superfluous submodule if N + X = M for every proper
submodule X of M .
A ring A is called a prime ring if BC = 0 for any two nonzero ideals B and C of A. A proper ideal
P of a ring A is called prime if the factor ring A/P is a prime ring (i.e., BC ⊆ P for any two ideals
B and C of A such that B ⊆ P and C ⊆ P ). A nonzero right module M is called a prime module if
r(K) = r(M ) for every nonzero submodule K of M . A proper submodule N of a nonzero right module
M is called prime if M/N is a prime module (i.e., r(K/N ) = r(M/N ) for every nonzero submodule K/N
of M/N ). For a module M , a minimal element of the set of prime submodules of M is called a minimal
prime submodule of M . For a module M , the intersection of all prime submodules of M is called the
prime radical of M ; it is denoted by rad(M ). For a module M and a proper submodule N of M , the
intersection of all prime submodules of M that contain N is called the N -prime radical of M ; it is denoted
by N -rad(M ). Thus, rad(M ) coincides with 0-rad(M ).
A ring is called a domain if it does not have nonzero divisors of zero. A proper ideal P of a ring A is
called completely prime if the factor ring A/P is a domain (i.e., bc ∈ / P for any two elements b, c ∈ A \ P ).
A nonzero right module M is called a completely prime module if r(m) = r(M ) for every nonzero element
m of M . A proper submodule N of a nonzero right module M is called a completely prime submodule if
M/N is a completely prime module (i.e., (N : m) = (N : M ) for every element m ∈ M \ N ).
For a proper ideal P of a ring A, it is directly verified that P is a prime (resp. completely prime)
ideal ⇐⇒ P is a prime (resp. completely prime) submodule of the right A-module A ⇐⇒ P is a prime
(resp. completely prime) submodule of the left A-module A. In addition, if A is a right-invariant ring,
then a right A-module is a prime module if and only if it is a completely prime module.
A ring A is said to be semilocal if the factor ring A/J(A) is an Artinian ring.
An element a of a ring A is said to be right regular (resp. left regular) if r(a) = 0 (resp. (a) = 0). A
right and left regular element is called a regular element. A right module M over a ring A is said to be
torsion-free if ma = 0 for every nonzero element m ∈ M and each regular element a ∈ A.
A unitary subring A of a ring Q is called a right order (resp. left order) in Q if every regular element
of A is invertible in the ring Q, and for every element q of Q, there exist elements a, s ∈ A such that s is
a regular element of A and q = as−1 (resp. q = s−1 a). In this case, the ring Q is called the right classical
ring of fractions of the ring A.
A subset S of a ring A is said to be right permutable if for any two elements a ∈ A and s ∈ S, there
exist two elements b ∈ A and t ∈ S with at = sb. If A is a right invariant ring, then every subset S of A
is right permutable, since as ∈ sA for any elements a ∈ A and s ∈ S. It is well known that a ring A has
the right classical ring of fractions if and only if the set of all regular elements of A is a right permutable
set. Thus, every right (resp. right and left) invariant ring A is a right (resp. right and left) order in some
ring.
For a (right and left) order A in a ring Q, a subbimodule M of the A-A-bimodule Q is said to be
invertible if there exists a subbimodule M ∗ of A-A-bimodule Q such that M M ∗ = M ∗ M = A.
Let A be a ring. A subset S of A is called multiplicative if 1 ∈ S, 0 ∈ / S, and S is multiplicatively
closed in A. A subset S of A is called right reversible if for any two elements a ∈ A and s ∈ S with
sa = 0, there exists an element t ∈ S with at = 0. A subset S of A is called a right denominator set if
the following two equivalent conditions hold (see [82, p. 51]): (1) S is a right permutable, right reversible
multiplicative set; (2) there exist a ring AS −1 and a ring homomorphism f : A → AS −1 such that all
elements of the set f (S) are invertible in AS −1 ,

AS −1 = {f (a)f (t)−1 | a ∈ A, t ∈ S}, and Ker(f ) = {a ∈ A | S rA (a) = ∅}.

In this case, the ring AS −1 is called the ring of fractions of A with respect to S, f is called the canonical
ring homomorphism, and for any right A-module M , the right AS −1 -module M ⊗A (AS −1 ) is called
the module of fractions of M with respect to S (this AS −1 -module is denoted by M S −1 ). The mapping
g : M → M ⊗A A ⊆ M S −1 defined by g(m) = m ⊗ 1 is an A-module homomorphism from MA into

3843
M S −1 ; this homomorphism is called the canonical module homomorphism. If P is a right ideal of A and
A \ P is a right denominator set, then the ring A(A \ P )−1 is called the right localization with respect to
P and we write AP and MP instead of A(A \ P )−1 and M (A \ P )−1 , respectively. A ring A is called right
localizable if for each maximal right ideal P of A, the right localization AP exists.

2. General Properties of Multiplication Modules

2.1. Proposition. For a right module M over a ring A, the following conditions are equivalent.
(1) M is a multiplication module.
(2) For every cyclic submodule X of M , there exists a right ideal B of the ring A such that X = M B.
(3) For every submodule X of M , there exist a set {Xi }i∈I of submodules of X and a set {Bi }i∈I of
ideals of A such that X = i∈I Xi and Xi = M Bi for each i ∈ I.
Proof. The implication (1)=⇒(2) is obvious.
(2)=⇒(3) Let X be a submodule of M , {Xi }i∈I be the set of all cyclic  submodules of X, and
Bi = (Xi : M ) (i ∈ I). By assumption, Xi ⊆ M Bi ⊆ Xi for all i. Since X = i∈I Xi , we have that {Xi }
and {Bi } are the required sets.
(3)=⇒(1) Let X be a submodule of M . By assumption,
 there exist a set {Xi }i∈I of submodules of
X and a set {Bi }i∈I of ideals of A such that X = i∈I Xi and Xi = M Bi for each i ∈ I. We denote by
B the ideal i∈I Bi of A. Then
 
  
X= Xi = M Bi = M Bi = M B,
i∈I i∈I i∈I

and M is a multiplication module.

2.2. Theorem. Let M be a right module over a ring A and let M = ⊕i∈I Mi . Then the following
conditions are equivalent.
(1) M is a multiplication module.
(2) Every submodule of M is fully invariant in M , and all modules Mi are multiplication modules
such that there exist ideals

Bi of A with Mi = M Bi (i ∈ I).
(3) N = ⊕i∈I (N Mi ) for every submodule N of M , and all modules Mi are multiplication modules
such that there exist ideals Bi of A with Mi = M Bi (i ∈ I).
(4) For every finite subset J of I, the module ⊕j∈J Mj is a multiplication module such that ⊕j∈J Mj =
M BJ for some ideal BJ of A.
Proof. The implication (1)=⇒(2) follows from Note 1.5(1) and Note 1.7.
(2)=⇒(3) Let N be a submodule of M and let ϕi : M → Mi be natural projections. Since N is a
fully invariant submodule of M , we have ϕi (N ) ⊆ N for all i ∈ I. Therefore,
N ⊆ ⊕i∈I ϕi (N ) ⊆ N.
Thus,

N = ⊕i∈I ϕi (N ) and N = ⊕i∈I N Mi .


(3)=⇒(1) Let N be a submodule of M and let Ni = N Mi (i ∈ I). By assumption, all modules
Mi are multiplication modules, N = ⊕i∈I Ni , and for every i ∈ I,
there exist ideals Bi and Ci of A such
that Mi = M Bi and Ni = Mi Ci . Since Ni = M Bi Ci and N = i∈I Ni , it follows from Proposition 2.1
that M is a multiplication module.
(1)=⇒(4) Since M is a multiplication module, ⊕j∈J Mj = M BJ for some ideal BJ of A. By Note
1.7, the direct summand ⊕j∈J Mj of the multiplication module M is a multiplication module.
(4)=⇒(1) Let N be a cyclic submodule of M . There exists a finite subset J of I such that N ⊆
⊕j∈J Mj . By assumption, ⊕j∈J Mj is a multiplication module such that ⊕j∈J Mj = M BJ for some ideal BJ

3844
of A. Since ⊕j∈J Mj is a multiplication module, there exists an ideal CJ of A such that (⊕j∈J Mj )CJ = N .
Then N = M BJ CJ . By Proposition 2.1, M is a multiplication module.

2.3. Proposition. Let M be a multiplication right module over a ring A and let h : A → A/r(M ) be the
natural ring epimorphism. Then the following assertions hold.
(1) If B is a right ideal of A such that B ⊇ r(M ) and h(B) is an essential right ideal of h(A), then
M B is an essential submodule of the A-module M .
(2) If the ring h(A) does not contain ideals that are direct sums of infinitely many nonzero ideals of
h(A), then M is a finite-dimensional A-module.
(3) If any two nonzero ideals of the ring h(A) have nonzero intersection, then M is a uniform
A-module.
(4) If r(M ) is a prime ideal of A, then M is a uniform prime A-module.


Proof. (1) Let N be a
submodule of M
such that (M B) N =
0. There exists an ideal C
of A with
N = M C. Then M (B C) ⊆ (M B) N = 0. Therefore, B C ⊆ r(M ). Since h(B) h(C) = 0
and h(B) is an essential right ideal of the ring h(A), we have h(C) = 0. Therefore, C ⊆ r(M ) and
N = M C = 0. It follows that M B is an essential submodule of M .
(2) Let X = ⊕∞ i=1 Xi be a submodule of M . Since M is a multiplication module, there exists a set
∞
{Bi }i=1 of ideals of the ring A such that for every i, we have Bi ⊇ r(M ) and Xi = M Bi . Further,
  
M (Bi Bj ) ⊆ M Bi M ( Bj ) = 0
j=i j=i
∞
for every i. Therefore, the sum i=1 h(Bi ) is a direct sum of the ideals h(Bi ) of the ring h(A) = A/r(M ).
By assumption, there exists a positive integer n such that h(Bi ) = 0 for all i > n. Then Xi = M Bi = 0
for all i > n; therefore, M is a finite-dimensional module.

(3) Let X and Y be two submodules of M such that X Y = 0. There exist ideals B and C of A
such that X = M B and Y = M C. Then
  
M (B C) ⊆ M B N C = X Y = 0.
Therefore,
 
B C ⊆ r(M ) and h(B) h(C).
Since any two nonzero ideals of the ring h(A) have a nonzero intersection, we can assume that h(B) = 0.
Then X = M B = 0 and M is a uniform A-module.
(4) Since any two nonzero ideals of the prime ring h(A) have nonzero intersection, it follows from (3)
that M is a uniform A-module. Let N be a nonzero submodule of M . Then r(M ) ⊆ r(N )
A and there
exists an ideal B of A such that N = M B and h(B) = 0. Since Br(N ) ⊆ r(M ), we have h(B)h(r(N )) = 0;
in addition, h(A) is a prime ring and h(B) = 0. Therefore, h(r(N )) = 0 and r(N ) ⊆ r(M ) ⊆ r(N ).

2.4. Lemma. For a ring A, the following assertions hold.

(1) If B1
, . . . , Bu , C1 , . . . , Cv are ideals of A such that A = Bs + Ct for all s and t, then A =
( us=1 Bi ) + ( vy=1 Cj ).
(2) If B and C are ideals of A and M is a right A-module such that M/M B and M/M C are finitely
generated modules, then M/M (BC) is a finitely generated module.
(3) If B1 , . . . , Bn are ideals of A and M is a
right
A-module such that
all modules

M/M Bi are
finitely generated, then M/M (B1 ·· · · ·Bn ), M/M (B1 · · · Bn ), and M/(M B1 · · · M Bn ) are finitely
generated modules.



Proof. (1) We denote by B and C the ideals us=1 Bs and vt=1 Ct of A, respectively. For any Bs , we
have
A = (Bs + C1 ) · · · (Bs + Cv ) ⊆ Bs + C1 · · · Cv ⊆ Bs + C ⊆ A.

3845
Therefore, A = C + Bs for all Bs . Then
A = (C + B1 ) · · · (C + Bv ) ⊆ C + B1 · · · Bv ⊆ C + B ⊆ A.
(2) Since M/M B and M/M C are finitely generated modules, there exist elements
x1 , . . . , xm , y1 , . . . , yn ∈ M such that

m 
n
M= xi A + M B = yj A + M C.
i=1 j=1
m
Therefore, M C = i=1 xi C + (M B)C and

n 
n 
m 
n 
m
M= yj A + M C = yj A + xi C + M (BC) ⊆ yj A + xi A + M BC ⊆ M.
j=1 j=1 i=1 j=1 i=1
m 
Let h : M → M/M (BC) be a natural epimorphism. Then h(M ) = i=1 h(xi )A + nj=1 h(yj )A.
(3) It follows from (2) and
the induction

on n that

M/M
(B1 · · · · · Bn ) is a finitely generated


module.
Since M (B
1 · · ·

· · Bn ) ⊆ M (B1 · · · Bn ) ⊆ M B1 · · · M Bn , the modules M/M (B1 · · · Bn ) and
M/(M B1 · · · M Bn ) are
isomorphic

to factor modules

of
the finitely generated module M/M (B1 · · · · ·
Bn ). Therefore, M/M (B1 · · · Bn ) and M/(M B1 · · · M Bn ) are finitely generated modules.

2.5. Proposition. Let M be a multiplication right module over a ring A, and let M = ⊕i∈I Mi with
card(I) ≥ 2. Then the following assertions hold.
(1) If M is a multiplication module, then all modules Mi are multiplication modules and Mi =
Mi r(⊕j=i Mj ) for every i ∈ I.
(2) If all modules Mi are multiplication modules and A = r(Mi ) + r(⊕j=i Mj ) for every i ∈ I, then
M is a multiplication module.
(3) Assume that the set I is finite, all modules Mi are multiplication modules, and A = r(Mi )+r(Mj )
for all i = j. Then M is a multiplication module.
(4) Assume that all modules Mi are multiplication modules, A = r(Mi ) + r(Mj ) for all i = j, and
for every finite subset J of I, there exists an ideal BJ of A such that ⊕i∈J Mi = M BJ . Then M is a
multiplication module.
Proof. (1) By Note 1.7, every direct summand Mi of the multiplication module M is a multiplication
module. Since M is a multiplication module, there exist ideals Bi of A such that Mi = M Bi for every i.
In addition,  
(⊕j=i Mj )Bi ⊆ M Bi (⊕j=i Mj ) = Mi (⊕j=i Mj ) = 0.
For every i ∈ I, we have
Mi = M Bi = (Mi ⊕ (⊕j=i Mj )) Bi = Mi Bi ⊕ (⊕j=i Mj ) Bi = Mi Bi .
(2) Let N be a submodule of M , Bi = r(⊕j=i Mj ), and hi : M → Mi be the projection with kernel
⊕j=i Mj (i ∈ I). By assumption, A = r(Mi ) + Bi for every i ∈ I. Therefore,

Mi = Mi A = Mi (r(Mi ) + Bi ) = Mi Bi and N Bi ⊆ N Mi .


Let i ∈ I. Since hi (N ) ⊆ Mi and N Bi ⊆ N Mi , we have
 
hi (N ) = hi (N )(r(Mi ) + Bi ) = hi (N )Bi = hi (N Bi ) ⊆ hi (N Mi ) = N Mi ⊆ N.


Therefore, N = ⊕i∈I (N Mi ). By Theorem 2.2, M is a multiplication module.
Let i ∈ I. By assumption, A = r(Mi ) + r(Mj ) for every j = i. By Lemma 2.4(1), A =
(3)

r(Mi ) + j=i r(Mj ) = r(Mi ) + r(⊕j=i Mj ). By (2), M is a multiplication module.
(4) Let J be a finite subset of the set I. By (3) applied to the module ⊕j∈J Mj , this module is a
multiplication module. By Theorem 2.2, M is a multiplication module.

3846
2.6. Proposition. Let M be a right module over a ring A such that
 
 
(M Bi ) = M (Bi + r(M ))
i∈I i∈I

for every nonempty set of ideals Bi (i ∈ I) of A. Then the following conditions are equivalent.
(1) M is a multiplication module.
(2) For any submodule N of M and each ideal C of A with N
M C, there exists an ideal B of A
such that B
C and N = M B.
(3) For any submodule N of M and each ideal C of A with N
M C, there exists an ideal B of A
such that B
C and N ⊆ M B.
Proof. (1)=⇒(2) Let N be a submodule of M and let C be
an ideal of A such that N
M C. Since M
is a multiplication module, N = M (N : M ). We set B = C (N : M ). Since N = M (N : M )
M C, we
have C ⊆ (N : M ). Therefore, B
C. Since r(M ) ⊆ (N : M ), we have
 
(C + r(M )) (N : M ) = C (N : M ) + r(M ) = B + r(M )


by the modular law. Further, N = M C M (N : M ), and by assumption,
  
M [(C + r(M )) (N : M )] = M (C + r(M )) M (N : M ) = M C M D = N.
Therefore,

N = M [(C + r(M )) (N : M )] = M (B + r(M )) = M B.
The implication (2)=⇒(3) is obvious.
(3)=⇒(1) Let N be a submodule of M and let
D = {D | D is an ideal of A and N ⊆ M D}.


Clearly A ∈ D. Let Bi (i ∈ I) be any nonempty collection of ideals in D. By assumption, i∈I Bi ∈ D.
By the Zorn lemma D has a minimum member C. Then N ⊆ M C. Assume that N = M C. By (3), there
exists an ideal B with B
C and N ⊆ M B. In this case, B ∈ D, contradicting the choice of C. Thus,
N = M C. It follows that M is a multiplication module.

In connection with Proposition 2.6, we present the following example.
2.7. Example ([17, p. 760]). Let Z be the ring of integers, Q be the field of rational numbers, A be the
commutative ring of all 2 × 2 matrices of the form
z q
0 z
with z ∈ Z and q ∈ Q, and M be the ideal of A consisting of all such matrices with z = 0. It is directly
verified that M is not a multiplication A-module but for any submodule N of MA and each ideal C of A
with N
M C, there exists an ideal B of A such that B
C and N ⊆ M B.
2.8. Proposition ([9]). For a right module M over a ring A, the following conditions are equivalent.
(1) Every factor module of M is a finite-dimensional module.
(2) Every submodule X of M contains a finitely generated submodule Y such that the module X/Y
does not have maximal submodules.
Proof. (1)=⇒(2) Assume the contrary. Then there exists a submodule X of M such that every finitely
generated submodule of X is contained in a maximal submodule of X. Let M1 be an arbitrary maximal
submodule of X. We take an element x1 of the nonempty set X \ M1 . Assume that we have chosen
maximal submodules M1 , . . . , Mn of X and elements x1 , . . . , xn ∈ X such that xi ∈
/ Mi but xi ∈ Mj for
all j > i. There exists a maximal submodule Mn+1 of X containing x1 A + · · · + xn A. We take an element

3847


h : X → X/( ∞
xn+1 of the nonempty set X
\ Mn+1 .
Let
i=1 Mi ) be the natural epimorphism and let n be
a positive integer. Then h( ni=1 Mi ) h( ∞i=n+1 Mi ) = 0. We have a properly descending chain
∞
 ∞
  ∞
  
h( Mi ) ⊃ h( Mi ) h(Mn ) ⊃ h( Mi ) h(Mn ) h(Mn+1 ) ⊃ . . . .
i=n+1 i=n+1 i=n+1


n
Therefore, h( ∞i=n+1 Mi ) has the same composition length as the module h(X)/h( i=1 Mi ). Therefore,

n ∞

h( Mi ) ⊕ h( Mi ) = h(X),
i=1 i=n+1


so that h(X) contains direct sums of arbitrary size. Therefore the factor module M/( ∞ i=1 Mi ) is not
finite-dimensional; this is a contradiction.
(2)=⇒(1) Assume that M has a factor module that is not finite-dimensional. Since every nonzero
module has a factor module with simple submodule, there exist submodules X and K of M such that
K ⊂ X and X/K = ⊕∞ i=1 X i , where all the modules X i are simple. By assumption, the module X contains
a finitely generated submodule Y such that the module X/Y does not have maximal submodules. Then
the module X/(K + Y ) does not have maximal submodules; in addition, the module X/(K + Y ) is a
homomorphic image of the semisimple module X/K. Therefore, X/(K + Y ) = 0 and X = K + Y . Then

⊕∞ ∼
i=1 X i = (K + Y )/K = Y /(Y K)
is a finitely generated module; this is a contradiction.

2.9. Theorem. Let A be a ring such that every factor ring of A does not contain an ideal that is an
infinite direct sum of nonzero ideals of the factor ring, and let M be a multiplication right A-module.
Then the following assertions hold.
(1) Every factor module of M is a finite-dimensional module.
(2) M is a Noetherian module ⇐⇒ every factor module of M is a semi-Noetherian module.
Proof. (1) Let M be a factor module of M . By assumption, the ring A/r(M ) does not contain an ideal
that is an infinite direct sum of nonzero ideals of A/r(M ). By Proposition 2.3(2), M is a finite-dimensional
module.
(2) The implication =⇒ is true for every Noetherian module over any ring.
⇐= By (1), every factor module of M is a finite-dimensional module. By Proposition 2.8, every
submodule X of M contains a finitely generated submodule Y such that the module X/Y does not have
maximal submodules. By assumption, every nonzero submodule of any factor module of M has a maximal
submodule. Therefore, X = Y , X is a finitely generated module, and M is a Noetherian module.

2.10. Proposition. Let A be a ring such that all right primitive factor rings of A are simple rings (this
is the case if A is a ring with polynomial identity). For a multiplication right A-module M , the following
assertions hold.
(1) A submodule N of M is a maximal submodule of M if and only if N = M P = M for some
maximal ideal P of A.
(2) J(M ) = M JM , where JM is the intersection of all maximal ideals P of A with M = M P .
Proof. (1) If P is a maximal ideal of A and N = M P = M , then N is a maximal submodule by Note
1.10(2).
If N is a maximal submodule of M , then N = M P for some right primitive ideal P of A, and P is
a maximal ideal by assumption.
(2) The proof follows from (1).

2.11. Proposition. For a multiplication right module M over a ring A, the following assertions hold.
(1) If A = B + r(M ) for every ideal B of A with M = M B, then M is a finitely generated module.

3848
 (2) If M = M P for every maximal ideal P of A with r(M ) ⊆ P , then M is a finitely generated
module.
(3) If M
1, . . .

, Mn are submodules of M such that every factor module M/Mi is finitely generated,
then M/(M1 · · · Mn ) is a finitely generated module.
(4) If there exist

submodules

M1 , . . . , Mn of M such that every factor module M/Mi is finitely
generated and M1 · · · Mn = 0, then M is a finitely generated module.
(5) If there exist

submodules

M1 , . . . , Mn of M such that every factor module M/Mi is finitely
generated and M1 · · · Mn is a superfluous submodule of M , then M is a finitely generated module.
Proof. (1) Let {Xi }i∈I be the set of all cyclic submodules of M . Since M is a multiplication module,
we have that for any Xi , there exists an ideal Bi of A such that Xi = M Bi . We denote by B the ideal

i∈I Bi of A. Then  
M= Xi = M Bi = M B.
i∈I i∈I
Since M = M B, we have A = B + r(M ) by assumption.  Since 1 ∈ B + r(M ), there exists a finite subset
{B1 , . . . , Bn } of the set {Bi }i∈I such that 1 ∈ ni=1 Bi + r(M ). Therefore,
n 
n 
n
M = M · 1 = M( Bi + r(M )) = M Bi = Xi ,
i=1 i=1 i=1
so that M is a finitely generated module.
(2) Let {Pi }i∈I be the set of all maximal ideals of the ring A with P ⊇ r(M ). By assumption,
M = M Pi for every Pi . Therefore, there exists a nonzero cyclic submodule Xi of M such that Xi ⊆ M Pi
(i ∈ I). We denote by B the ideal i∈I r(M/Xi ) of A. Then B ⊇ r(M/Xi ) ⊇ r(M ).
Assume that A = B. Then there exists a finite subset J of I such that A = j∈J r(M/Xj ). Since
M is a multiplication module, Xj = M r(M/Xj ) for every j ∈ J. Therefore,
   
  
M = MA = M  r(M/Xj ) =  M r(M/Xj ) = Xj .
j∈J j∈J j∈J

Since M = j∈J Xj , the module M is finitely generated.
Assume that A = B. Then there exists a maximal ideal P of A such that P ⊇ B ⊇ r(M ).
Therefore, P = Pi for some i ∈ I. Then r(M/Xi ) ⊆ B ⊆ Pi . Since M is a multiplication module,
Xi = M r(M/Xi ) ⊆ M Pi ; this is a contradiction.
(3) Since M is a multiplication module, there exist ideals B1 , . . . , Bn of A such that Mi = M Bi
for every i. Then


every factor module M/M Bi is finitely generated.


By Lemma
2.4(3),
the mod-
ule M/M (B1 · · ·
Bn
) is finitely generated. Since M (B1 · · · Bn ) ⊆ M B1 · · · M Bn , the
module M/(M


B1 · · · M Bn ) is isomorphic
to
a factor module of the finitely generated module
M/M
(B1
· · · Bn ). Therefore,


M/(M B 1 · · · M Bn ) is a finitely generated module; in addition,
M B1 · · · M Bn = M1 · · · Mn .
(4) The proof follows from (3).
(5)

It follows

from (3) that

there

exists a finitely generated submodule N of M such that M =
N + M1 · · · Mn . Since M1 · · · Mn is a superfluous submodule of M , we have M = N .

2.12. Lemma. Let A be a ring, M be a multiplication right A-module, and M = X ⊕ Y . Then the
following assertions

hold.

(1) Z = X Z ⊕ Y Z for any submodule Z of the module M .
(2) If there exist a module P and epimorphisms α : P → X and β : P → Y , then the homomorphism
α + β : P → M is an epimorphism.
(3) If the modules X and Y are cyclic modules, then M is a cyclic module.
Proof. (1) The proof follows from Theorem 2.2.

3849
 (2) We denote by Z the submodule (α + β)(P ) of M = X ⊕ Y . Let πX : M → X and let πY : M → Y
be natural projections. We have
πX (Z) = α(P ) = X and πY (Z) = β(P ) = Y ;



in addition, Z = X Z ⊕ Y Z by (1). Therefore,
 
X = πX (Z) = X Z ⊆ Z and Y = πY (Z) = Y Z ⊆ Z.
Therefore, M = X ⊕ Y = Z and α + β is an epimorphism.
(3) Since X and Y are cyclic modules, there exist epimorphisms AA → X and AA → Y . By (2),
there exists an epimorphism AA → M ; therefore, M is a cyclic module.

2.13. Proposition. If M is a multiplication right module that is a direct sum of finitely many cyclic
modules, then M is a cyclic module.
Proof. The proof follows from Lemma 2.12(3).

2.14. Theorem. For an Artinian multiplication module M , the following assertions hold.
(1) The factor module M/J(M ) is a cyclic module.
(2) If J(M ) is a superfluous submodule of M , then M is a cyclic module.
(3) If M is a finitely generated module, then M is a cyclic module.
Proof. (1) The homomorphic image M/J(M ) of the multiplication module M is a multiplication mod-
ule by Note 1.4. Since M/J(M ) is a semiprimitive Artinian module, M/J(M ) is a finitely generated
semisimple module. By Proposition 2.13, the factor module M/J(M ) is a cyclic module.
(2) Since M/J(M ) is a cyclic A-module, there exists a cyclic submodule X of M such that M =
X + J(M ). By assumption, J(M ) is a superfluous submodule of M . Therefore, M = X.
(3) Since M is a finitely generated module, J(M ) is a superfluous submodule of M . By (2), M is a
cyclic module.

2.15. Proposition. Let A be a ring such that J(A)B ⊆ BJ(A) for every ideal B of A. Then the
following assertions hold.
(1) M = M J(A) for any nonzero multiplication right A-module M .
(2) For any multiplication right A-module M , the module M J(A) is a superfluous submodule of M .
Proof. We set J = J(A).
(1) Let X be a nonzero cyclic submodule of M . Then X = XJ. Since M is a multiplication module,
there exists an ideal B of A such that M B = X. Assume that M = M J. Since JB ⊆ BJ by assumption,
we have
X = M B = M JB ⊆ M BJ = XJ ⊆ X.
Therefore, X = XJ; this is a contradiction.
(2) Let M be a multiplication right A-module, N be a submodule of M with M = N + M J, and
M = M/N . By Note 1.4, the homomorphic image M of the multiplication module M is a multiplication
module; in addition, M = M J. It follows from (1) that M = 0. Therefore, N = M and M J is a
superfluous submodule of M .

2.16. Lemma. Let A be a ring, M be a multiplication right A-module, and M = X ⊕ Y . Then the
following assertions hold.
(1) f (X) ⊆ X for any homomorphism f : X → M .
(2) Hom(X, Y ) = Hom(Y, X) = 0.
(3) The natural ring homomorphism End(X) × End(Y ) → End(M ) is an isomorphism.
Proof. (1) There exists an endomorphism g of the module M such that g(x + y) = f (x) for all elements
x ∈ X and y ∈ Y . By Note 1.5(1), all submodules of the multiplication module M are fully invariant
submodules of M . Therefore, X ⊇ g(X) = f (X).

(2) If f ∈ Hom(X, Y ), then by (1), f (X) ⊆ X Y = 0.

3850
 (3) The proof follows from (2).

2.17. Lemma ([8, 83]). For a right module M over a ring A, the following conditions are equivalent.
(1) M is a distributive module.
(2) Every 2-generated

submodule of
M is a distributive module.
(3) (x + y)A = xA (x + y)A + yA (x + y)A for any two elements x and y of M .
(4) For any two elements x and y of M , there exists an element a of the ring A such that xa ∈ yA
and y(1 − a) ∈ xA.
(5) A = (yA : x) + (xA : y) for any two elements x and y of M .
(6) For any two isomorphic simple right A-modules X and Y , the module X ⊕Y is not a homomorphic
image of any 2-generated submodule of M .
Proof. The implications (1)=⇒(2)
and (2)=⇒(3) follow from the definition of a distributive module and
the relation (x + y)A = (xA + yA) (x + y)A.
(3)=⇒(4) Since  
(x + y)A = yA (x + y)A + xA (x + y)A,
there exist elements a, b, c, d ∈ A such that
 
x + y = (x + y)a + (x + y)b, (x + y)a = yc ∈ yA (x + y)A, and (x + y)b = xd ∈ xA (x + y)A.
Then
xa = yc − ya ∈ yA and (x + y)(1 − a) = (x + y)b = xd.
Therefore,
y(1 − a) = (x + y)(1 − a) − x(1 − a) = xd − x(1 − a) ∈ xA.
The equivalence of conditions (4) and (5) is directly verified.
(4)=⇒(3) Since xa ∈ yA and y(1 − a) ∈ xA, we have
 
(x + y)a = xa + ya ∈ yA (x + y)A and (x + y)(1 − a) = x(1 − a) + y(1 − a) ∈ xA (x + y)A.
Therefore,
 
(x + y)A ⊆ (x + y)aA + (x + y)(1 − a)A ⊆ xA (x + y)A + yA (x + y)A ⊆ (x + y)A.
(3)=⇒(1) Assume
that X, Y , and Z are three submodules of the module M and z is an element of
the module (X + Y ) Z. There exist elements x ∈ X and y ∈ Y such that z = x + y. By assumption,
   
zA = xA zA + yA zA ⊆ X Z +Y Z.
Therefore,    
(X + Y ) Z⊆X Z +Y Z ⊆ (X + Y ) Z.
(1)=⇒(6) Assume the contrary. Then the module
X ⊕ Y is
distributive. Since X ∼
= Y , there exists
∼
a submodule Z of X ⊕ Y such that Z = X = 0 and Z X = Z Y = 0. Then
  
Z = Z (X ⊕ Y ) = Z X ⊕Z Y = 0;
this is a contradiction.
(6)=⇒(5) Let x and y be two elements of M . Assume that A = (yA : x) + (xA : y). There exists a
maximal right ideal P of A such that (yA : x) + (xA : y) ⊆ P . Let h : xA + yA → (xA + yA)/(xP + yP )
be the natural epimorphism, X = h(x)A, and Y = h(y)A. We have X ⊆ Y , since otherwise x ∈ xP + yP .
There exists an element p of P such that x(1 − p) ∈ yP . Thus 1 = p + (1 − p) ∈ P + (yA : x) = P ; this is
a contradiction. Therefore, X = 0. Similarly, Y = 0. Since h(x)P = h(0) = h(y)P , we have that X and
Y are nonzero homomorphic images of the simple right A-module A/P . Therefore, X ∼ = (A/P
)A ∼
= Y ; in
addition, X and Y are simple modules and X + Y = h(xA + yA). It follows from (3) that X Y = 0; in
addition, X and Y are simple modules. Therefore, X = Y ; this contradicts the relation X ⊆ Y .

3851
2.18. Lemma ([83]). For a ring A and a distributive right A-module
M , the following assertions hold.
(1) If x and y are two elements of the module M such that xA yA = 0, then there exists an element
a of the ring A such that xa = y(1 − a) = 0 and xA ⊕ yA is a cyclic module with generator x + y.
(2) If X is a submodule of M that is a finite direct sum
of cyclic modules, then X is a cyclic module.
(3) If X and Y are two submodules of M such that X Y = 0, then Hom(X, Y ) = Hom(Y, X) = 0.
(4) If M = X ⊕ Y , then f (X) ⊆ X for every homomorphism f : X → M .
Proof. (1) It follows from Lemma 2.17 that there exists an element a of A such that xa = y(1 − a) = 0.
In addition,
x = x(1 − a) + y(1 − a) = (x + y)(1 − a) ∈ (x + y)A and y = xa + ya ∈ (x + y)A.
(2) The proof follows from (1).
(3) Let f : X → be a homomorphism, x be an element of the module X, and y = f (x) ∈ Y . By (1),
there exists an element a of A such that xa = y(1 − a) = 0. Then
f (x) = ya + y(1 − a) = ya = f (xa) = f (0) = 0.
Therefore, Hom(X, Y ) = 0.
(4) The proof follows from (3).

2.19. Proposition. (1) M is a semisimple distributive module if and only if M = ⊕i∈I Mi , where all the
modules Mi are simple and Mi ∼ = Mj for all i = j.
(2) Every semisimple multiplication module is a distributive module.
(3) Every finitely generated semisimple multiplication module is a cyclic distributive module.
(4) If M is a finitely generated semisimple distributive right module over a right quasi-invariant ring
A, then M is a multiplication module.
Proof. (1) The assertion is verified with the use of Lemma 2.18(3).
(2) Let M be a multiplication module such that M = ⊕i∈I , where all modules Mi are simple. If
i, j ∈ I and i = j, then Mi ⊕ Mj is a direct summand of the multiplication module M ; therefore, Mi ⊕ Mj
is a multiplication module by Note 1.7. By Lemma 2.16(2), Mi ∼ = Mj for all i = j. By (1), M is a
distributive module.
(3) Let M be a finitely generated semisimple multiplication module. By (2), M is a distributive
module. Since M is a finite direct sum of cyclic modules, M is a cyclic module by Proposition 2.13.
(4) Let M = ⊕ni=1 Mi , where all modules Mi are simple. Let i, j ∈ {1, . . . , n}, and let i = j. Since A
is a right quasi-invariant ring, there exist maximal ideals Bi and Bj of A such that the factor rings A/Bi
and A/Bj are division rings,
Mi ∼
= (A/Bi )A , Mj ∼ = (A/Bj )A , Bi = r(Mi ), and Bj = r(Mj ).
Since M is a distributive module, it follows from (1) that (A/Bi )A ∼ = (A/Bj )A . Therefore, Bi = Bj .
Then A = Bi + Bj , since Bi and Bj are maximal ideals. We note that all modules Mi are multiplication
modules and A = r(Mi ) + r(Mj ) for all i = j. By Proposition 2.5(3), M is a multiplication module.

In connection with Proposition 2.19, we consider the following example.
∞
2.20. Example. Let A be a right quasi-invariant ring that
∞has an infinite set {Bi }i=1 of distinct maximal
ideals Bi of A such that the ideal B1 contains the ideal i=2 Bi . (For example, we can take A = Z and
Bi = pi Z, where {pi }∞ ∞
i=1 is the set of all distinct prime integers.) Then the right A-module ⊕i=1 (A/Bi )A
is a semisimple distributive module that is not a multiplication module.
Proof. Let M = ⊕∞ i=1 (A/Bi )A and let M

i = (A/Bi )A (i = 1, 2, . . . ). By Proposition 2.19(1), M is a
semisimple distributive module. Let B = ∞ ∞
i=2 Bi . Then B = r(⊕i=2 Mi ). By assumption, B ⊆ B1 =
r(M1 )
Assume that M is a multiplication module. By Proposition 2.5(1), M1 = M1 r(⊕∞ i=2 Mi ). Thus,
M1 = M1 B ⊆ M1 r(M1 ) = 0; this is a contradiction.

3852
2.21. Proposition. Every multiplication right module over a direct product of finitely many simple rings
is a cyclic semisimple distributive module.
Proof. Let A be a direct product of finitely many simple rings A1 , . . . , An , and let M be a multiplication
right A-module. It is directly verified that MA = M A1 ⊕ · · · ⊕ M An , and for every i, we have that M Ai
is a multiplication right module over the simple ring Ai . Without loss of generality, we can assume that
M Ai = 0 for all i. By Note 1.1(2), M Ai is a simple Ai -module. Therefore, M Ai is a simple A-module,
and M is a finitely generated semisimple multiplication module. By Proposition 2.19(3), M is a cyclic
distributive module.

2.22. Lemma. Let A be a semilocal ring with Jacobson radical J and let M be a multiplication right
A-module. Then the following assertions hold.
(1) M/M J is a cyclic distributive semisimple A-module.
(2) If M J is a superfluous submodule of M , then M is a cyclic module.
Proof. (1) The homomorphic image M/M J of the multiplication module M is a multiplication module
by Note 1.4. Therefore, M/M J is a multiplication A-module. Since the semisimple Artinian ring A/J is
a direct product of finitely many simple rings, M/M J is a cyclic distributive semisimple A/J-module by
Proposition 2.21. Therefore, M/M J is a cyclic distributive semisimple A-module.
(2) By (1), M/M J is a cyclic A-module. Therefore, there exists a cyclic submodule X of MA such
that M = X + M J. By assumption, M J is a superfluous submodule of M . Therefore, M = X.

2.23. Theorem. Every finitely generated multiplication module over a semilocal ring is a cyclic module.
Proof. If A is a semilocal ring with the Jacobson radical J and M is a finitely generated right A-module,
then M J is a superfluous submodule of M . Therefore, the assertion follows from Lemma 2.22(2).

2.24. Theorem. Let A be a semilocal ring such that J(A)B ⊆ BJ(A) for every ideal B of A. Then
every multiplication right A-module is a cyclic module.
Proof. Let M be a multiplication right A-module. By Proposition 2.15(2), the module M J(A) is a
superfluous submodule of M . By Lemma 2.22(2), M is a cyclic module.

2.25. Proposition. For a right module M over a ring A, the following assertions hold.
(1) If X and Y are two submodules of the module M such that X + Y is a multiplication module,
then   
(X + Y ) Z = X Z +Y Z
for any submodule Z of M .
(2) If all 2-generated submodules of M are multiplication modules, then M is a distributive module.
Proof.
(1) Since X + Y is a multiplication module, there exists an ideal B of the ring A such that
(X + Y ) Z = (X + Y )B. In addition,
 
XB ⊆ X Z and Y B ⊆ Y Z.
Therefore,
   
(X + Y ) Z = (X + Y )B = XB + Y B ⊆ X Z +Y Z ⊆ (X + Y ) Z.

(2) Let x and

y be two elements

of M . By assumption, xA + yA is a multiplication module. By (1),
(x + y)A = xA (x + y)A + yA (x + y)A. By Lemma 2.17, M is a distributive module.

2.26. Proposition. Let M be a right module over a ring A, and let X and Y be two multiplication
submodules of M . Then the following conditions are equivalent.
(1) X + Y is a multiplication module.

3853
 (2)

X = (X + Y )(X : (X + Y )) = (X + Y )(X : Y ) = X(X : Y ) + X Y,

Y = (X + Y )(Y : (X + Y )) = (X + Y )(Y : X) = Y (Y : X) + X Y,
and for any submodule Z of M , we have
  
(X + Y ) Z=X Z +Y Z.



(3) Z = X Z + Y Z for any submodule Z of M , and there exist ideals B and C of the ring A
such that X = (X + Y )B and Y = (X + Y )C.




Proof. (1)=⇒(2) By Proposition 2.25(1), (X + Y ) Z = X Z + Y Z for any submodule
Z of M .
Since X + Y and Y are multiplication modules, X = (X + Y )(X : (X + Y )) and X Y = Y (X : Y )
by Proposition 2.1. In addition, (X : (X + Y )) = (X : Y ). Therefore,

X = (X + Y )(X : Y ) = X(X : Y ) + Y (X : Y ) = X(X : Y ) + X Y.
Similarly, 
Y = (X + Y )(Y : (X + Y )) = (X + Y )(Y : X) = Y (Y : X) + X Y.
The implication (2)=⇒(3) is obvious.
(3)=⇒(1) Let Z be a submodule of the module X + Y . We have
to prove

that there exists an ideal
H of the ring A such that Z = (X + Y )H. By assumption, Z = X
Z + Y Z. Since
X and Y are
multiplication modules, there exist ideals D and E of A such that Z X = XD and Z Y = Y E. By
assumption, there exist ideals B and C of the ring A such that X = (X + Y )B and Y = (X + Y )C.
Therefore,
 
Z=X Z +Y Z = XD + Y E = (X + Y )BD + (X + Y )CE = (X + Y )(BD + CE),
and BD + CE is the required ideal H of A.

2.27. Corollary. Let M be a right module over a ring A, and let X and Y be two multiplication
submodules of M such that A = (X : Y ) + (Y : X). Then X + Y is a multiplication module.



Proof. Let Z
∈ Lat(X + Y ). Then
Z =
((X + Y ) Z)((X : Y ) + (Y : X)) = ((X + Y ) Z)(X :
Y ) + ((X + Y ) Z)(Y : X) ⊆ X Z + Y Z ⊆ Z. By Proposition 2.26, it is sufficient to prove that
there exist ideals B and C

of the ring A such that X = (X + Y )B and Y = (X + Y )C. Since X and Y are
multiplication modules, X Y = X(Y : X) = Y (X : Y ) by Proposition 2.1. Since A = (X : Y )+(Y : X),
we have
X = X((X : Y ) + (Y : X)) = X(X : Y ) + X(Y : X) = X(X : Y ) + Y (X : Y ) = (X + Y )(X : Y ).
Similarly, Y = (X + Y )(Y : X) and we can set B = (X : Y ) and C = (Y : X).

2.28. Proposition. Let M be a multiplication right module over a ring A, and let N be a proper
submodule of M . Then the following assertions hold.
(1) N is a prime submodule of M ⇐⇒
r(M/N ) is a prime ideal of A.
(2) If N is a maximal submodule of M , then N is a prime submodule of M and N = M P for some
right primitive ideal P of A.
(3) If A is a ring with commutative multiplication of ideals, then for every proper submodule N of
M , there exists a right primitive ideal P of A such that M = M P + N and M P is a prime submodule of
M.
Proof. (1) We prove the implication =⇒. Let B and C be two ideals of A such that BC ⊆ r(M/N ) and
B ⊆ r(M/N ). Then (M/N )B is a nonzero submodule of M/N and C ⊆ r((M/N )B). Since N is a prime
submodule of M , we have C ⊆ r(M/N ). Therefore, r(M/N ) is a prime ideal of A.

3854
 ⇐= We set P = r(M/N ). Since M is a multiplication module, N = M P . Let K/N be a nonzero
submodule of the module M/N , where N = M P
K ⊆ M . Since M P
K and M is a multiplication
module, there exists an ideal B of A such that K = M B and P
B. In addition, M Br(K/N ) ⊆ N .
Therefore, Br(K/N ) ⊆ P ; in addition, P is a prime ideal of A. Therefore,
r(K/N ) ⊆ P = r(M/N ) ⊆ r(M/N ) and r(K/N ) = r(M/N ).
(2) Let N be a maximal submodule of the multiplication module M . Then N = M r(M/N ) and
r(M/N ) is a right primitive ideal of A, since M/N is a simple module. Since the right primitive ideal
r(M/N ) of A is a prime ideal and N = M r(M/N ), it follows from (1) that N is a prime submodule of
M.
(3) Let M be the nonzero multiplication module M/N , and let X be a nonzero cyclic submodule of
M . Since the nonzero cyclic module X has a maximal submodule, there exists a right primitive ideal P
of A such that X = XP . By Note 1.12(1), M = M P . Therefore, M = M P + N . By (1), M P is a prime
submodule of M .

2.29. Proposition. Let A be a ring with commutative multiplication of ideals such that all right primitive
factor rings of A are simple rings. For a multiplication right A-module M , the following assertions hold.
(1) For every proper submodule N of M , there exists a maximal ideal P of A such that M P is a
maximal submodule of M , M P is a prime submodule of M , and M P contains N .
(2) Every factor module M of M is a semi-Noetherian module such that J(M ) = M J(A) and J(M )
is a superfluous submodule of M .
(3) If every factor ring of A does not contain an ideal that is an infinite direct sum of nonzero ideals
of the factor ring, then M is a Noetherian module.
Proof. (1) Let N be a proper submodule of M . By Proposition 2.28(3), there exists a right primitive
ideal P of A such that M = M P + N and M P is a prime submodule of M . Since all right primitive factor
rings of A are simple rings, P is a maximal ideal of A. By Note 1.10(2), M P is a maximal submodule of
M ; in addition, M = M P + N . Therefore, N ⊆ M . We have P ⊆ r(M/M P )
A.
(2) By (1), every factor module M of M is a semi-Noetherian module. Let N be a submodule of
M such that M = N + J(M ). We have to prove that M = N . If N is a proper submodule of the
multiplication module M , then we obtain a contradiction with (1).
(3) The proof follows from (2) and Theorem 2.9(2).

In the case of modules over commutative rings, the main results of this section turn into modifications
of the results proved in [3], [4], [17], and [78]. In the case of modules over noncommutative rings, the
main results of the presented work are proved in [88] and [89].

3. Multiplication Modules over Invariant Rings

3.1. Proposition. For a ring A, the following conditions are equivalent.

(1) A is a right invariant ring.
(2) All cyclic right A-modules are multiplication modules.
(3) The free principal right ideal A of A is a multiplication right ideal.
Proof. (1)=⇒(2) Let M be a cyclic right A-module with generator m, N be a submodule of M = mA,
and B = (N : m). Since N ⊆ mA, we have N ⊆ m(N : m) ⊆ N ; therefore, N = mB. In addition, B is
an ideal of A, since the ring A is right invariant. Therefore, N = mB = m(AB) = (mA)B = M B and M
is a multiplication module.
The implication (2)=⇒(3) is obvious.
(3)=⇒(1) Let N be a right ideal of the ring A. Since AA is a multiplication module, there exists an
ideal B of A such that N = AB. Therefore, N = AB = A(AB) = AN and N is an ideal of the ring A.

3855
3.2. Theorem. For a right module M over a right invariant ring A, the following conditions are
equivalent.
(1) M is a multiplication module that is a direct sum of finitely many cyclic modules.
(2) M is a cyclic module.
(3) There exist elements m1 , . . . , mn of M such that M = ⊕ni=1 mi A and A = r(mi ) + r(mj ) for all
i = j.
Proof. The implication (1)=⇒(2) follows from Proposition 2.13.
The implication (2)=⇒(1) follows from Proposition 3.1.
(2)=⇒(3) Let M = m1 A. We set m2 = 0. Then M = m1 A ⊕ m2 A and A = r(m1 ) + r(m2 ).
(3)=⇒(2) We set m = m1 + . . . + mn ∈ M . Let i ∈ {1, . . . , n}. Since r(mi ) is an ideal

of the right
invariant ring A and A = r(mi ) + r(mj ) for all j = i, Lemma 2.4(1) implies A = r(mi ) + j=i r(mj ).


Therefore, there exists an element ai ∈ r(mi ) such that 1 − ai ∈ j=i r(mj ). Therefore,
m(1 − ai ) = (m1 + . . . + mn )(1 − ai ) = mi (1 − ai ) = mi .
Thus mi A ⊆ mA for every i. Therefore, M = mA.

3.3. Note. If M is a Bezout right module over a right invariant ring, then M is a distributive module
and all finitely generated submodules of M are multiplication modules.
Proof. By Proposition 3.1, all finitely generated submodules of M are multiplication modules. Therefore,
M is a distributive module by Proposition 2.25(2).

3.4. Proposition. If A is a right invariant ring, then every ideal of A generated by some set of
idempotents of A is a multiplication right ideal.

Proof. Let {ei }i∈I be a set of idempotents of the ring A, M = i∈I ei A, and B be a right ideal of A
contained in M . Since A is a right invariant ring, we have M B ⊆ B. Therefore, it is sufficient to prove
the relation B ⊆ M B.  Let b be an element of B. There exists a finite subset {e1 , . . . , en } of the set
{ei }i∈I such that b = ni=1 ei ai for some elements a1 , . . . , an of A. Since all right ideals ei A and (1 − ei )A
are ideals of A, all idempotents ei are central. Therefore, (1 − e1 ) · · · · · (1 − en )b = 0 and
 n 

b = (1 − (1 − e1 ) · · · · · (1 − en ))b ∈ ei A b ⊆ M B.
i=1
Therefore, B ⊆ M B.

3.5. Lemma. Let A be a right invariant ring and let M be a right A-module. Then the following
assertions hold.
(1) For any cyclic submodule X of M with generator x and each submodule Y of M , we have
(Y : x) = (Y : X); consequently, (Y : x) is an ideal of the ring A.
(2) If m is an element of M and P is a maximal ideal of A such that r(mA) ⊆ P , then r(m(1−p)A) ⊆
P for every element p of P .
(3) If

X and
Y are two
submodules of M such that X + Y is a finitely generated module and
(X + Y ) Z = (X Z) + (Y Z) for all submodules Z of M , then A = (X : Y ) + (Y : X).
(4) If M is a distributive module and X and Y are two submodules of M such that X + Y is a finitely
generated module, then A = (X : Y ) + (Y : X).
Proof. (1) Since (Y : x) is a right ideal of A and the ring A is right invariant, (Y : x) is an ideal of A.
Therefore,
(xA)(Y : x) = x(A(Y : x)) = x(Y : x) ⊆ Y and (Y : x) ⊆ (Y : xA) ⊆ (Y : x).
(2) Assume that r(m(1 − p)A) ⊆ P for some element p of P . Since P is a maximal ideal, A =
P + r(m(1 − p)A). Therefore, there exists an element p1 of P such that m(1 − p)A(1 − p1 ) = 0. We denote

3856
by p2 the element 1 − (1 − p)(1 − p1 ) of the ideal P . Since A is a right invariant ring,
mA(1 − p2 ) = mA(1 − p)(1 − p1 ) ⊆ m(1 − p)A(1 − p1 ) = 0.
Therefore, 1 − p2 ∈ r(mA) ⊆ P and 1 = p2 + (1 − p2 ) ∈ P ; this is a contradiction.
(3) Since X + Y is a finitely generated module, there exist a positive integer n and elements xi ∈ X,
yi ∈ Y (1

≤ i ≤ n)
such that
X + Y = (x1 + y1 )A + · · · + (xn + yn )A. Let y ∈ Y . By assumption,
(X + Y ) Z = (X Z) + (Y Z) for all submodules Z of M . For any 1 ≤ i ≤ n, we have
  
(xi + y)A = (xi + y)A (X + Y ) = [(xi + y)A X] + ((xi + y)A Y ].
Therefore, there exist elements a ∈ A and z ∈ Y such that
xi + y = (xi + y)a + z and (xi + y)a ∈ X.
Thus, xi (1 − a) ∈ Y and ya ∈ X. It follows that
A = (X : yA) + (Y : xi A) (1 ≤ i ≤ n).
Therefore,  
A = (X : yA) + [(Y : x1 A) ··· (Y : xn A)] = (X : yA) + (Y : X).
In particular,
A = (X : yi A) + (Y : X) (1 ≤ i ≤ n).
Therefore,  
A = [(X : y1 A) ··· (X : yn A)] + (Y : X) = (X : Y ) + (Y : X).
(4) The proof follows from (3).

3.6. Lemma. For a right module M over a right invariant ring A, the following conditions are equivalent.
(1) M is a distributive module.
(2) For any two cyclic submodules X and Y of M , there exists an element a of A such that Xa ⊆ Y
and Y (1 − a) ⊆ X.
(3) A = (X : Y ) + (Y : X) for any two cyclic submodules X and Y of M .
Proof. The proof follows from Lemma 2.17 and Lemma 3.5(1).

3.7. Lemma. For a right module M over a right invariant ring A, the following conditions are equivalent.
(1) M is a distributive module.
(2) For any two finitely generated submodules X and Y of M , there exists an element a of A such
that Xa ⊆ Y and Y (1 − a) ⊆ X.
(3) A = (X : Y ) + (Y : X) for any two finitely generated submodules X and Y of M .
Proof. (1)=⇒(2) Let X = X1 + · · · + Xm and Y = Y1 + · · · + Yn , where all the modules Xi and Yj are
cyclic. By Lemma 3.6, A = (Xi : Yj ) + (Yj : Xi ) for all i and j. It follows from Lemma 2.4(1) that

n 
m
A=( (Xi : Yj )) + ( (Yj : Xi )) ⊆ (Xi : Y ) + (Yj : X) ⊆ (X : Y ) + (Y : X) ⊆ A.
j=1 i=1

The equivalence of conditions (2) and (3) is directly verified.

The implication (2)=⇒(1) follows from Lemma 3.6.

3.8. Theorem. For every distributive right module over a right invariant ring, all finitely generated
submodules of the module are multiplication modules.
Proof. Let A be a right invariant ring. Since every submodule of a distributive module is distributive,
it is sufficient to prove that for any positive integer n, every distributive n-generated right A-module is
a multiplication module. We use induction on n. By Proposition 3.1, all cyclic right modules over the
right invariant ring A are multiplication modules. Therefore, the assertion holds for n = 1. Assume that
n ≥ 2 and every (n − 1)-generated distributive right A-module is a multiplication module. Let M be a

3857
distributive n-generated right A-module and let Z be a submodule of M . The n-generated module M has
submodules X and Y such that M = X + Y , X is a cyclic module, and Y is an (n − 1)-generated module.
By the induction hypothesis, X and Y are multiplication modules. By Lemma 3.7, A = (X : Y )+(Y : X).
By Corollary 2.27, M is a multiplication module.

3.9. Lemma. For a right invariant ring A, the following assertions hold.
(1) B(1 − b) ⊆ (1 − b)B for every right ideal B of A and each element b of B.
(2) If B is an ideal of A and M is a cyclic right A-module such that M = M B, then A = B + r(M ).
(3) Let n be a positive integer ≥2, M be an n-generated right A-module with generators m1 , . . . , mn ,

N = n−1 i=1 mi A, and B be an ideal of A such that M = M B. Then A = B + (N : mn A) and N = N B.
. . . , an are elements of A such that 1 = ni=1 ai , then there exist elements b1 , . . . , bn of A
(4) If a1 , 
such that 1 = ni=1 a2i bi .
Proof. (1) Let b1 be an element of B. Since A is a right invariant ring, (1 − b)A is an ideal of A.
Therefore, there exists an element a of A such that b1 (1 − b) = (1 − b)a. Then
a = (1 − b)a + ba = b1 (1 − b) + ba ∈ B and B(1 − b) ⊆ (1 − b)B.
(2) Let M = mA. Since A is a right invariant ring, r(M ) = r(m) and mA = mAB = mB by
assumption. Therefore, there exists an element b of B such that m(1 − b) = 0. Therefore, A = B + r(m) =
A + r(M ).
(3) We set 
0 if n = 2
Y = n−1 .
i=2 mi A if n≥3
For every mi , we have
 n
mi ∈ M = M B = mj B,
j=1

and there exists an element bi of the ideal B such that mi (1 − bi ) ∈ j=i mj B. Therefore,
mn A(1 − bn ) ⊆ N B = m1 B + Y B and m1 A(1 − b1 ) ∈ Y B + mn B,
since A is a right invariant ring; in addition, B(1 − b1 ) ⊆ (1 − b1 )B by (1). Thus A = B + (N : mn A) and
we have
m1 A(1 − b1 )(1 − bn ) ⊆ (Y B + mn B)(1 − bn ) ⊆ Y B + mn B(1 − bn )
⊆ Y B + mn (1 − bn )B ⊆ Y B + m1 B = N B.
Therefore, there exists an element b∗ of B such that
m1 (1 − b1 )(1 − bn ) − m1 b∗ ∈ Y B and m1 [(1 − b1 )(1 − bn ) − b∗ ] ∈ Y B.
We set b = 1 − [(1 − b1 )(1 − bn ) − b∗ ] ∈ B. Then m1 = m1 b + m1 (1 − b) = m1 b + m1 [(1 − b1 )(1 − bn ) − b∗ ] ∈
m1 B + Y B = N B. Therefore, m1 A ⊆ N B. Similarly, we can prove that mi A ⊆ N B for i = 2, . . . , n − 1.
Therefore, N = N B. 
(4) It is sufficient to prove that the ideal ni=1 a2i A of the right invariant ring A coincides with A.
Assume the contrary. Then there exists a maximal ideal P of A that contains B. Therefore, every element
a2i is contained in P .Then every element ai is contained in P , since the factor ring A/P is a division
ring. Therefore, P  ni=1 ai = 1 ∈ / P ; this is a contradiction.

3.10. Proposition. For a right invariant ring A, the following assertions hold.
(1) Let M be a finitely generated right A-module and let B be an ideal of A. Then M = M B if and
only if A = B + r(M ).
(2) If B is a finitely generated ideal of A and B = B 2 , then there exists a central idempotent e of A
such that B = eA.
Proof. (1) If A = B + r(M ), then M = M (B + r(M )) = M B + M r(M ) = M B.

3858
 
Assume that M = ni=1 mi A = M B. We use induction on n. For n = 1, the assertion follows from
Lemma 3.9(2).
Assume
 that n ≥ 2 and the assertion holds for all positive integers that are less than n. Let
N = n−1i=1 mi A. By Lemma 3.9(3), A = B + (N : mn A) and N = N B. By the induction hypothesis,
A = B + r(N ). Then
M (N : mn A)r(N ) = (N + mn A)(N : mn A)r(N )
= N (N : mn A)r(N ) + mn A(N : mn A)r(N ) ⊆ N r(N ) = 0.
Therefore, (N : mn A)r(N ) ⊆ r(M ) and
A = A · A = (B + (N : mn A))(B + r(N )) ⊆ B + (N : mn A)r(N ) ⊆ B + r(M ) ⊆ A.
(2) Since B is a finitely generated ideal of the right invariant ring A, the ideal B is a finitely generated
right ideal of A. By (1), A = B + r(B). Therefore, there exists an element e of B such that B(1 − e) = 0.
In particular, e(1 − e) = 0 and e is an idempotent of A. Since eA and (1 − e)A are ideals of the right
invariant ring A, the idempotent e is central.

3.11. Theorem. Let A be a right invariant ring, and let M be a nonzero multiplication right A-module.
Then the following conditions are equivalent.
(1) M is a finitely generated module.
(2) M = M B for every proper ideal B of A with r(M ) ⊆ B.
(3) M = M P for every maximal ideal P of A with r(M ) ⊆ P . 
(4) There exists a finite set {m1 , . . . , mn } of elements of M such that A = ni=1 (mi A : M ).
(5) There exist elements a1 , . . . , an , b1 , . . . , bn of A and a finite set {m1 , . . . , mn } of elements of M
such that 1 = ni=1 a2i bi and M ai ⊆ mi A for i = 1, . . . , n.
Proof. (1)=⇒(2) Assume that there exists a proper ideal B of A such that r(M ) ⊆ B and M = M B.
By Proposition 3.10(1), A = B + r(M ) = B; this is a contradiction.
The implication (2)=⇒(3) is obvious.
The implication (3)=⇒(1) follows from Proposition 2.11(2).
(4)=⇒(1) Since M = M A, we have
 
n n 
n
M = M r(M ) + (mi A : M ) = M r(M ) + (M (mi A : M )) = mi A.
i=1 i=1 i=1
n n
(1)=⇒(4) Let M = i=1 mi A, and let B be the ideal i=1 (mi A : M ) of A. Then

n 
n
MB = (M (mi A : M )) = mi A = M.
i=1 i=1

Since M B = M and M is a finitely generated module, A = r(M ) + B by Proposition 3.10(1). In addition,

r(M ) ⊆ (mi A : M ) for every
ni. Therefore, A = B. 
(4)=⇒(5) Since A = ( i=1 (mi A : M )), there exist elements a1 , . . . , an+1 of A such that 1 = ni=1 ai
M ai ⊆ mi A for i = 1, . . . , n. By Lemma 3.9(4), there exist elements b1 , . . . , bn of A such that
and 
1 = ni=1 a2i bi .
nai bi for i = 1, . . . , n. Since M ai ⊆ mi A, we have M ci ⊆ mi A. Since
(5)=⇒(4) We set ci = 2
 n
1 = i=1 ci , we have A = i=1 (mi A : M ).

3.12. Theorem. Let M be a right module over a right invariant ring A, M = ⊕i∈I Mi , card(I) ≥ 2, and
let all modules Mi be finitely generated. Then the following conditions are equivalent.
(1) M is a multiplication module.
(2) All modules Mi are multiplication modules and A = r(Mi ) + r(⊕j=i Mj ) for every i ∈ I.
(3) All modules Mi are multiplication modules, A = r(Mi ) + r(Mj ) for all i = j, and for every finite
subset J of I, there exists an ideal BJ of A such that ⊕i∈J Mi = M BJ .

3859
Proof. (1)=⇒(2) By Proposition 2.5(1), all modules Mi are multiplication modules and Mi =
Mi r(⊕j=i Mj ) for every i ∈ I. By Proposition 3.10(1), A = r(Mi ) + r(⊕j=i Mj ) for every i ∈ I.
(2)=⇒(3) By (2), all modules Mi are multiplication modules and A = r(Mi ) + r(⊕k=i Mk ) for every
i ∈ I. In addition, if j = i, then r(⊕k=i Mk ) ⊆ r(Mj ). Therefore, A = r(Mi ) + r(Mj ) for all i = j. By
Proposition 2.5(2), M is a multiplication module. It follows from Theorem 2.2 that for every finite subset
J of I, there exists an ideal BJ of A such that ⊕i∈J Mi = M BJ .
The implication (3)=⇒(1) follows from Proposition 2.5(4).

For a right module M over a ring A, we denote by Sing(M ) the set of all elements m ∈ M such that
r(m) is an essential right ideal of A. The set Sing(M ) is a fully invariant submodule of M ; it is called the
singular submodule of M . Thus, Sing(AA ) is an ideal of A; it is called the right singular ideal of A.
3.13. Proposition. For a faithful multiplication right module M over a right invariant ring A, the
following assertions hold.
(1) Sing(M ) = M Sing(AA ).
(2)
{a ∈ A | ma = 0 for some 0 = m ∈ M } = {a ∈ A | ba = 0 for some 0 = b ∈ A}.
(3) M is a torsion-free module.
Proof. (1) It is directly verified that the relation M Sing(AA ) ⊆ Sing(M ) holds for every right module
M over any ring A.
Now assume that M is a multiplication right module over a right invariant ring A. Let m be an
element of the module Sing(M ). Since M is a multiplication module, there exists an ideal B of A such
that mA = M B. Since A is a right invariant ring, r(m) is an ideal of A. Therefore, r(m) = r(mA) and
M Br(m) = mAr(m) = mAr(mA) = 0.
Therefore, Br(m) = 0, since M is a faithful module. Since r(m) is an essential right ideal of A and
Br(m) = 0, we have B ⊆ Sing(AA ). Therefore, m ∈ M B ⊆ M Sing(AA ) and Sing(M ) ⊆ M Sing(AA ).
(2) We set
z(M ) = {a ∈ A | ma = 0 for some 0 = m ∈ M },
z(A) = {a ∈ A | ba = 0 for some 0 = b ∈ A}.
Let a ∈ z(A). There exists a nonzero element b of A such that ba = 0. Since M is faithful, it follows that
M b = 0 and (M b)a = 0, so that a ∈ z(M ). Thus z(A) ⊆ z(M ).
Conversely, let a ∈ Z(M ). There exists a nonzero element m of M such that ma = 0. There exists
an ideal B of A such that mA = M B. Therefore, M Ba = 0 and Ba = 0, since M is faithful. Then B = 0
and a ∈ z(A). Thus, z(M ) ⊆ z(A).
(3) The proof follows from (2).

In connection with Proposition 3.14, we note that every invariant ring A has the (two-sided) classical
ring of fractions, since as ∈ sA and sa ∈ As for any elements a, s ∈ A.
3.14. Proposition. Let A be an invariant ring, Q be the classical ring of fractions of A, and M be an
ideal of A. Then the following conditions are equivalent.
(1) M is an invertible ideal.
(2) MA and A M are finitely generated multiplication modules and M contains a regular element s of
A.
(3) MA and A M are multiplication modules and M contains a regular element s of A.
Proof. (1)=⇒(2) By assumption, there exists a subbimodule M ∗ of the A-A-bimodule Q such that
M M ∗ = M ∗ M = A. Let N be a submodule of the right or left A-module M . Since A is an invariant
ring, N is an ideal of A. We denote by B and C the ideals M ∗ N and N M ∗ of A, respectively. We have
N = AN = M M ∗ N = M B and N = N A = N M ∗ M = CN.

3860
Therefore, MA and A M are multiplication modules. 
Since M M ∗ = A, there exist elements m1 , . . . , mn ∈ M and q1 , . . . , qn ∈ M ∗ such that ni=1 mi qi =
1. There exist elements a1 , . . . , an ∈ A and a regular element s of A such that qi = ai s−1 for i = 1, . . . , n.
Then  n 
  n
s =1·s = mi ai s−1 s = mi ai ∈ M
i=1 i=1
and M contains the regular element s. In addition,
 n 
 
n 
n
M =1·M = mi qi M = mi (qi M ) ⊆ mi A ⊆ M.
i=1 i=1 i=1
Therefore, M is a finitely generated right ideal of A. Since A is an invariant ring, M is a finitely generated
left ideal of A.
The implication (2)=⇒(3) is obvious.
(3)=⇒(1) Since MA and A M are multiplication modules, there exist ideals B and C of A such that
   
M B = sA = As = CM ; therefore, M Bs−1 = A = s−1 C M.
Since A is an invariant ring, s−1 as ∈ A and sas−1 ∈ A for any element a of A; in addition, B and C are
ideals of the invariant ring A. Therefore, it is directly verified that Bs−1 and s−1 C are subbimodules of
the A-A-bimodule Q. In addition,
   
Bs−1 = ABs−1 = s−1 C M Bs−1 = s−1 C M Bs−1 = s−1 CA = s−1C.
Therefore, we can set M ∗ = Bs−1 = s−1 C. Thus M M ∗ = M ∗ M = A and M is an invertible ideal.

3.15. Proposition. Let A be a right invariant ring, M be a finitely generated right A-module, and X
and Y be two submodules of M such that X + Y = M . Then the following assertions hold.
(1) If M is a multiplication module, then A = (X : Y )

+ (Y : X).


(2) If X and Y are multiplication modules and Z = X Z + Y Z for any submodule Z of M , then
M is a multiplication module.
Proof. (1) We denote by B the ideal (X : Y ) + (Y : X) of A. We have
(X : Y ) = (X : (X + Y )) = (X : M ) and (Y : X) = (Y : (X + Y )) = (Y : M ).
Then
M B = M ((X : M ) + (Y : M )) = M (X : M ) + M (Y : M ) = X + Y = M.
Since M B = M and M is a finitely generated module, A = r(M ) + B by Proposition 3.10(1). In addition,
r(M ) ⊆ (X : Y ) ⊆ B for every i. Therefore, A = B.
(2) By Lemma 3.5(3), A = (X : Y ) + (Y : X). By Corollary 2.27, M is a multiplication module.

Let A be a ring, S be a right denominator set in A, f : A → AS −1 be the canonical ring homomor-
phism, M be a right A-module, g : M → N S −1 be the canonical module homomorphism, and ∼ be a
relation on the Cartesian product M × S such that
(m, s) ∼ (n, t) ⇐⇒ ∃a, b ∈ A : ma = nb, sa = tb ∈ S.
It is directly verified that ∼ is an equivalence relation on M × S. We denote the factor set (M × S)/ ∼
by M [S −1 ]. Let h be a natural surjective mapping M × S → M [S −1 ] and let v be the mapping from M
into M [S −1 ] defined by v(m) = h(m, 1). In M [S −1 ], we define multiplication by elements of AS −1 and
addition as follows:
h((m, s)) · f (a)f (t)−1 = h((mb, tu)), (a, b ∈ A, t, u ∈ S, sb = au),
h((m, s)) + h((n, t)) = h((ma + nb, u)) (u = sa = tb ∈ S).
It is directly verified that M [S −1 ] is a right AS −1 -module. There exists a mapping w : M S −1 → M [S −1 ]
such that w(m ⊗ x) = h((m, 1))x. It is directly verified that v is an A-module homomorphism, the

3861
mapping w : M S −1 → M [S −1 ] is an AS −1 -module isomorphism, and v = wg. Therefore, we can identify
M S −1 with M [S −1 ].
In Lemmas 3.16, 3.17, and 3.18, we gather several known properties of rings and modules of fractions.
3.16. Lemma. Let A be a ring, M be a right A-module, S be a right denominator set in A, Q be
the right ring of fractions of A with respect to S, f : A → Q be the canonical ring homomorphism, and
g : M → M S −1 be the canonical module homomorphism. Then the following assertions hold.
(1) Ker(g) = {m ∈ M | ms = 0} for some s ∈ S.
(2) For any elements q1 , . . . , qn of Q, there exist elements a1 , . . . , an ∈ A and s ∈ S such that
qi = f (ai )f (s)−1 for i = 1, . . . , n.
(3) For every submodule X of M , we have
g(X)Q = {g(x)f (s)−1 | x ∈ X, s ∈ S}.
(4) For every submodule X of M and each right ideal B of A, we have
g(XB)Q = g(X)f (B)Q = {g(x)f (s)−1 | x ∈ XB, s ∈ S}.
(5) For every submodule X of the Q-module M S −1 , there exists a submodule X of M such that
Ker(g) ⊆ X and g(X)Q = X.
(6) If m is an element of M and X is a submodule of M , then g(m) ∈ g(X)Q if and only if there
exists an element s of S such that ms ∈ X.
(7) If m is an element of M and X is a submodule of M , then f ((X : m))Q is contained in the ideal
(g(X)Q : g(mA)Q) of Q.
 (8) If X and Y  are two submodules
of M and {Xi }
is some set of submodules of M , then
i∈I (g(X i )Q) = g( X
i∈I i )Q and g(X)Q g(Y )Q = g(X Y )Q. Consequently, if MA = ⊕i∈I Xi ,
then (g(M )Q)Q = ⊕i∈I (g(Xi )Q).
(9) If the set A \ S is closed under addition, then Q is a local ring.
(10) If q is an
element of Q, then qQ = Q if and only if there exists an element a of A such that
qQ = f (a)Q and S aA = ∅.
(11) If the right A-module M is free, then the right Q-module g(M )Q is free.
(12) If the right A-module M is projective, then the right Q-module g(M )Q is projective.
Proof. (1) The assertion is proved in Corollary 3.3 of [82], p. 57.
(2) By induction on n, it is sufficient to consider the case n = 2. Assume that qi = f (di )f (si )−1 ,
where di ∈ A and si ∈ S (i = 1, 2). Since S is right permutable, there exist t1 ∈ S and t2 ∈ A such
that s1 t1 = s2 t2 = s ∈ S. We set ai = di ti (i = 1, 2). Then f (si )−1 = f (ti )f (s)−1 and qi = f (ai )f (s)−1
(i = 1, 2).
 x be an element of g(X)Q. There exist elements x1 , . . . , xn ∈ X and q1 , . . . , qn ∈ Q such
(3) Let
−1
that x = ni=1 g(xi )qi . By (2), nthere exist elements a1 , . . . , an ∈ A and s ∈ S such that qi = f (ai )f (s)
for i = 1, . . . , n. We set x = i=1 xi ai ∈ X. Then
 n 
n 
x= g(xi )f (ai )f (s)−1 = g(xi ai ) f (s)−1 = g(x)f (s)−1 .
i=1 i=1

(4) The relations g(XB)Q = g(X)f (B)Q and g(XB)Q ⊇ {g(x)f (s)−1 | x ∈ XB, s ∈ S} are
directly verified. Let y be an element  of g(XB)Q. There exist elements x1 , . . . , xn ∈ X, b1 , . . . , bn ∈ B,
and q1 , . . . , qn ∈ Q such that y = ni=1 g(xi bi )qi . By (2), there exist elements
 a1 , . . . , an ∈ A and s ∈ S
such that qi = f (ai )f (s)−1 for i = 1, . . . , n. We denote by x the element ni=1 xi bi ai of XB. Then
 n 
n 
y= g(xi bi )qi = g( xi bi ai ) f (s)−1 = g(x)f (s)−1 .
i=1 i=1

3862
 (5) For every element y of X, we take any element x(y) of M such that y = g(x(y))f
 (s)−1 for some
element s of S; thus yQ = g(x(y))Q. We denote by X the submodule Ker(g) + y∈X yA of M . Then
Ker(g) ⊆ X and g(X)Q = X.
(6) If there exists an element s of S such that ms ∈ X, then g(m)f (s) ∈ g(X) and g(m) ∈
g(X)f (s)−q ⊆ g(X)Q.
Now assume that g(m) ∈ g(X)Q. By (3), there exist elements x ∈ X and t ∈ S such that g(m) =
g(x)f (t)−1 . Then g(mt) = g(x) and mt − x ∈ Ker(g). By (1), there exists an element u ∈ S such that
(mt − x)u = 0. We set s = tu ∈ S. Then ms = mtu = xu ∈ X.
(7) Let y be an element of (g(X)Q : g(m)Q). There exist elements a ∈ A and s ∈ S such that
y = f (a)f (s)−1 . Then
g(ma) = g(m)f (a)f (s)−1 f (s) ∈ g(m)yQ ⊆ g(X)Q.
By (6), there exists an element t of S such that mat ∈ X. Then at ∈ (X : m) and y = f (at)f (t)−1 f (s)−1 ∈
f ((X : m))Q.  


(8) The relations i∈I (g(Xi )Q) =
g( i∈I Xi )Q and g(X)Q g(Y )Q ⊇ g(X Y )Q are directly
verified. Let z be an element of g(X)Q g(Y )Q. By (3), there exist elements t, u ∈ S, x ∈ X, and
y ∈ Y such that z = g(x)f (t)−1 = g(y)f (u)−1 . By (2), there exist elements a, b ∈ A and s ∈ S such
that f (t)−1 = f (a)f (s)−1 and f (u)−1 = f (b)f (t)−1 . Therefore, z = g(xa)f (s)−1 = g(yb)f (s)−1 . Then

= g(yb) and xa − yb ∈ Ker(g). Therefore, there exists an element v ∈ S such that xav = ybv ∈
g(xa)
X Y . Then
  
z = g(xav)f (v)−1 f (s)−1 ∈ g(X Y )Q and g(X)Q g(Y )Q ⊆ g(X Y )Q.

(9) Let q be an element of Q such that qQ = Q. It is sufficient to prove that (1 − q)Q = Q. There
exist elements a ∈ A and s ∈ A such that q = f (a)f (s)−1 . Since qQ = Q, we have a ∈ A \ S. Since
s = (s − a) + a ∈ S and the set A \ S is closed under addition, it follows from a ∈ A \ S that s − a ∈ S.
Therefore,
Q = f (s − a)Q = f (s − a)f (s)−1 Q = (f (s) − f (a))f (s)−1 Q = (1 − q)Q.
(10) Let q be an element of Q. If there exist elements a, b ∈ A such that qQ = f (a)Q and ab ∈ S,
then the element f (ab) is invertible in Q and Q = f (ab)Q ⊆ f (a)Q = qQ ⊆ Q.
Now assume that qQ = Q. There exist elements a ∈ A and s ∈ S such that q = f (a)f (s)−1 .
Therefore, Q = qQ ⊆ f (a)Q ⊆ Q, f (a)Q = Q, and there exist elements c ∈ A and s ∈ S such
that f (a)f (c)f (s)−1 = f (1). Then

ac − s ∈ t ∈ Ker(f ) and there exists an element t ∈ S such that
(ac − s)t = 0. Then st = act ∈ S aA.
(11) The proof follows from (8).
(12) The proof follows from (8) and (11).

3.17. Lemma. Let A be a ring, M be a right A-module, P be an ideal of A such that A \ P is a right
denominator set in A, Q be the right ring of fractions of A with respect to A \ P , f : A → Q be the
canonical ring homomorphism, and g : M → MP be the canonical A-module homomorphism from M into
the module of fractions MP with respect to A \ P .


(1) Q is a local ring with Jacobson radical PP and XP YP = (X Y )P for any two submodules X
and Y of M .
(2) If m is an element of M and X is a submodule of M , then g(m) ∈ XP if and only if the right
ideal (X : m) of A is not contained in P .
(3) Assume that the ideal P of A is maximal. If m is an element of M and X is a submodule of M ,
then g(m) ∈ XP if and only if A = P + (X : m). In particular, g(m) = 0 if and only if m(1 − p) = 0 for
some element p of P .
(4) If M is a nonzero projective right A-module, then the right Q-module g(M )Q is free.
Proof. We set S = A \ P .

3863



(1) By Lemma 3.16(8), XP YP = (X Y )P for any two submodules X and Y of M . Since
P = A \ S) and P is closed under addition, it follows from Lemma 3.16(9) that Q is a local ring. Let p be
an element of f (P )Q. By Lemma 3.16(3), there exist elements p ∈ P and s ∈ S such that p = f (p)f (s)−1 .
Then
(1 − p)Q = (f (s) − f (p))f (s)−1 Q = f (s − p)Q.
Since p ∈ P and s ∈ / P , we have s − p ∈ / P . By Lemma 3.16(10), (1 − p)Q = Q; in addition, p is an
arbitrary element of the right ideal f (P )Q of Q. Therefore, f (P )Q ⊆ J(Q).
Now we take an element q ∈ Q \ f (P )Q. It follows from Lemma 3.16(3) that there exist elements
a, s ∈ S such that q = f (a)f (s)−1 . Therefore, every element of the set Q \ f (P )Q is invertible in Q.
Therefore, J(Q) ⊆ f (P )Q ⊆ J(Q).

(2) By Lemma 3.16(6), g(m) ∈ g(X)Q if and only if there exists an element s ∈ (A \ P ) (X : m).
(3) The proof follows from (2).
(4) By Lemma 3.16(12), the right Q-module g(M )Q is projective. By (1), Q is a local ring. Every
nonzero projective module over a local ring is free [33].

3.18. Lemma. Let A be a right localizable ring and let M be a right A-module. For every maximal right
ideal P of A, we denote by AP , MP , fP , and gP the right localization of A with respect to P , the module
of fractions of M with respect to P , the canonical ring homomorphism A → AP , and the canonical module
homomorphism M → MP , respectively. Then the following assertions hold.
(1) If X is a submodule of M and m is an element of M such that gP (m) ∈ XP for each maximal
right ideal P of A, then m ∈ X.
(2) If m is an element of M such that gP (m) = 0 for each maximal right ideal P of A, then m = 0.
(3) If X and Y are submodules of M such that XP = YP for each maximal right ideal P of A, then
X =Y.
(4) Assume that for every maximal right ideal P of A, the right localization AP is a right invariant
ring; then A is a right invariant ring.
(5) If M is a multiplication module and A is a ring with commutative multiplication of ideals, then
MP = 0P for every maximal right ideal P of A such that P ⊇ m∈M (mA : M ).
Proof. (1) If (X : m) = A, then m ∈ X. Assume that (X : m) = A. Then there exists a maximal right
ideal P of A such that the right ideal (X : m) of A is contained in P . Since gP (m) ∈ XP , it follows from
Lemma 3.16(6) that there exists an element s of A \ P such that ms ∈ X. Then
 
s ∈ (X : m) (A \ P ) ⊆ P (A \ P ) = ∅;
this is a contradiction.
(2) and (3) The assertions follow from (1).
(4) Let B be a right ideal of A and let P be a maximal right ideal of A. Since AP is a right invariant
ring, we have
BP = AP BP ⊇ (AB)P ⊇ BP .
By (3), B = AB and A is a right invariant ring.
(5) We set B = m∈M (mA : M ). We have
  
M= mA = M (mA : M ) = M ( (mA : M )) = M B.
m∈M m∈M m∈M
Let P be a maximal ideal of A with P ⊇ B and let m be an element of M . Then
mA = M (mA : M ) = M B(mA : M ) = M (mA : M )B = mAB ⊆ mP.
Therefore, there exists an element p of P such that m(1 − p) = 0. Therefore, m ∈ Ker(gP ) and MP = 0P .


3.19. Lemma. Let A be a ring, M be a right A-module, S be a right denominator set in A such that
the right ring of fractions Q of A with respect to S is a right invariant ring, f : A → Q be the canonical

3864
ring homomorphism, and g : M → M S −1 be the canonical module homomorphism. Then the following
assertions hold.
(1) g(XB)Q = g(X)Qf (B)Q = {g(x)f (s)−1 | x ∈ XB, s ∈ S} for every submodule X of M and
each right ideal B of A.
(2) f ((X : Y ))Q is contained in the ideal (g(X)Q : g(Y )Q) of Q.
(3) If X is a submodule of M and m is an element of M such that (X : m) = (X : mA), then
f ((X : mA))Q coincides with the ideal (g(X)Q : g(mA)Q) of Q.
Proof. (1) By Lemma 3.16(4),
g(XB)Q = g(X)f (B)Q = {g(x)f (s)−1 | x ∈ XB, s ∈ S}.
Since Q is a right invariant ring, f (B)Q = Qf (B)Q and g(X)f (B)Q = g(X)Qf (B)Q.
(2) Let z be an element of f ((X : Y ))Q. By Lemma 3.16(3), there exist elements a ∈ (X : Y ) and
s ∈ S such that z = f (a)f (s)−1 . Since Q is a right invariant ring, Qz ⊆ zQ. Therefore,
f (Y )Qz ⊆ f (Y )zQ = f (Y )f (a)f (s)−1 Q ⊆ f (Y a)Q ⊆ f (X)Q
and f ((X : Y ))Q ⊆ (g(X)Q : g(Y )Q).
(3) By (2), f ((X : mA))Q ⊆ (g(X)Q : g(mA)Q). By Lemma 3.16(7), f ((X : m))Q ⊆ (g(X)Q :
g(mA)Q). By assumption, (X : m) = (X : mA).

3.20. Lemma. Let A be a right invariant ring and let S be a multiplicative subset of A. Then the
following assertions hold.
(1) S is a right permutable set.
(2) Assume that for any elements x ∈ A and s ∈ S such that x2 = 0 and sx = 0, there exists an
element t ∈ S with xt = 0. Then S is a right denominator set in A.
(3) Assume that P is a maximal right ideal of A and for any elements x ∈ A and s ∈ A \ P such
that x2 = 0 and sx = 0, there exists an element t ∈ A \ P with xt = 0. Then the right localization A \ P
exists.
(4) If every element x ∈ A with x2 = 0 is contained in the center of A, then S is a right denominator
set and A is a right localizable ring.
(5) If A is a reduced ring, then S is a right denominator set and A is a right localizable ring.
(6) If A is a ring with commutative multiplication of ideals, then S is a right denominator set, AS −1
is a right invariant ring with commutative multiplication of ideals, and A is a right localizable ring.
(7) If A is a ring with maximum condition on right annihilators of elements of A, then S is a right
denominator set and A is a right localizable ring.
Proof. (1) If a ∈ A and s ∈ S, then sA is an ideal of the right invariant ring A and there exists an
element b ∈ A such that as = sb. Therefore, S is a right permutable set.
(2) By (1), S is a right permutable set. Therefore, it is sufficient to prove that S is a right reversible
set. Let a ∈ A, s ∈ S, and sa = 0. We set x = as. Then x2 = a(sa)s = 0 and sx = (sa)s = 0. By
assumption, there exists an element t ∈ S with xt = 0. Since st ∈ S and ast = xt = 0, the set S is right
reversible.
(3) Let P be a maximal right ideal of A. Then P is an ideal of the right invariant ring A and the
factor ring A/M is a division ring. Therefore, A \ P is a multiplicative set. By (2) applied to A \ P , this
set is a right denominator set.
(4) The proof follows from (2) and (3).
(5) The proof follows from (4).
(6) Let x ∈ A, s ∈ S, and sx = 0. Since A is a right invariant ring with commutative multiplication
of ideals,
xs ∈ (AxA)(AsA) = (AsA)(AxA) = AsxA = 0.
By (2), S is a right denominator set. By (3) and the above argument, A is a right localizable ring. Let
Q = AS −1 , X and Y be two right ideals of Q, and f : A → Q be the canonical ring homomorphism. By

3865
Lemma 3.16(5), there exist right ideals X and Y of A such that X = f (X)Q and Y = f (Y )Q. Since A
is a right invariant ring with commutative multiplication of ideals,
XY = (AX)(AY ) = (AY )(AX) = Y X.
This implies that for any elements x ∈ X and s ∈ S, we have
xs ∈ xAsA = sAxA = sxA and
f (s)−1 f (x) ∈ f (x)f (A)f (s−1) ⊆ f (X)Q = X.
Therefore,
QX ⊆ f (A)f (X)Q = f (AX)Q = f (X)Q = X and
Y X = f (Y )Q(X) = f (Y )X = f (Y X)Q = f (XY )Q ⊆ XY .
Similarly, we have XY ⊆ Y X. Therefore, Q is a ring with commutative multiplication of right ideals and
QX ⊆ X. Therefore, Q is a right invariant ring.
(7) Let x ∈ A, s ∈ S, and sx = 0. Then r(si ) ⊆ r(si+1 ) for i = 1, 2, . . . . By assumption, there exists
a positive integer n such that r(sn ) = r(sn+1 ). By (1), S is a right permutable set. Therefore, there exist
elements b ∈ A and t ∈ S such that sn b = xt. Since sn+1 b = sxt = 0, we have b ∈ r(sn+1 ) = r(sn ).
Therefore, at = sn b = 0. By (2), S is a right denominator set. By (3) and the above argument, A is a
right localizable ring.

3.21. Proposition. Let A be a ring, M be a multiplication right A-module, and S be a right denominator
set in A such that AS −1 is a right invariant ring. Then the following assertions hold.
(1) The right AS −1 -module M S −1 is a multiplication module.
(2) If the set A \ S is closed under addition and M is a finitely generated module, then the right
AS −1 -module M S −1 is a cyclic multiplication module.
Proof. We set Q = AS −1 .
(1) Let g : M → M S −1 be the canonical A-module homomorphism and let X be a submodule of
the right Q-module M S −1 . By Lemma 3.16(5), there exists a submodule X of M such that g(X)Q = X.
Since M is a multiplication A-module, there exists an ideal B of A such that X = M B. In addition,
f (B)Q = Qf (B)Q, since Q is a right invariant ring. Therefore,
X = g(X)Q = g(M B)Q = g(M )f (B)Q = g(M )Qf (B)Q = M S −1 (Qf (B)Q);
thus, M S −1 is a multiplication Q-module.
(2) By Lemma 3.17(1), Q is a local ring. Since M is a finitely generated A-module, M S −1 is a finitely
generated Q-module. By (1), M S −1 is a finitely generated multiplication module over the local ring Q.
By Theorem 2.23, M is a cyclic module.

3.22. Proposition. Let A be a ring, M be a multiplication right A-module, and S be a right denominator
set in A such that AS −1 is a right invariant semilocal ring with commutative multiplication of ideals. Then
M S −1 is a cyclic multiplication right AS −1 -module.
Proof. By Proposition 3.21(1), M S −1 is a multiplication right AS −1 -module. By Theorem 2.24, M S −1
is a cyclic AS −1 -module.

3.23. Proposition. Let A be a right invariant ring, M be a multiplication right A-module, and h : A →
A/r(M ) be the natural ring epimorphism. Then the following assertions hold.
(1) For every finitely generated direct summand X of M , there exists a central idempotent e of the
ring h(A) such that X = M e and rh(A) (X) = (1h(A) − e)h(A).
(2) For every cyclic direct summand X of M , there exists a central idempotent e of the ring h(A)
such that the right h(A)-module X is isomorphic to the projective direct summand eh(A) of h(A)h(A) .

3866
Proof. (1) Let M = X ⊕Y , B = (X : M ) ⊇ r(M ), and C = (Y : M ) ⊇ r(M ). Since M is a multiplication
module, X = M B and Y = M C. We have
M r(X) = (X ⊕ Y )r(X) = Y r(X) ⊆ Y, r(X) ⊆ C,

XC ⊆ X Y = 0, and C ⊆ r(X).
Therefore, C = r(X). Similarly, we obtain B = r(Y ). We have

X = M B = (X ⊕ Y )B = XB ⊕ Y B = X (XB ⊕ Y B) = XB.
Since X = XB and X is a finitely generated right module over the right invariant ring A, it follows from
Proposition 3.10 that
A = B + r(X) = B + C and h(A) = h(B) + h(C).
In addition, we have
    
M (B C) ⊆ M B M C = X Y = 0, B C = r(M ), and h(B) h(C) = h(0).
Therefore, there exists a central idempotent e of the ring h(A) such that h(B) = eh(A) and h(C) =
(1h(A) − e)h(A). Then X = M e and rh(A) (X) = (1h(A) − e)h(A).
(2) Since X is a cyclic module over the right invariant ring h(A), there exists an element x of X such
that X = xh(A) and rh(A) (x) = rh(A) (X). By (1), there exists a central idempotent e of the ring h(A)
such that rh(A) (x) = rh(A) (X) = (1h(A) − e)h(A). Therefore,
Xh(A) ∼= h(A)/rh(A) (x) = h(A)/(1h(A) − e)h(A) ∼= eh(A).

3.24. Lemma ([85, p. 170], [86, p. 143]). Let M be a right module
over a right
distributive, right
invariant ring A and let B and C be two right ideals of A. Then M B M C = M (B C).
 

Proof. Let m = si=1 mi bi = tj=1 nj cj ∈ M B M C, where mi , nj ∈ M , bi ∈ B, and cj ∈ C. We
 
denote by X and Y the finitely generated right ideals si=1 bi A and tj=1 cj A of A, respectively. Then
X and Y are finitely generated right ideals of the right distributive, right invariant ring A. By Lemma


3.7, there exists an element a of A such that all the elements bi a and cj (1 − a) are contained in X Y .
 


Then m = ma + m(1 − a) = si=1 mi bi a + tj=1 nj cj (1 − a) ∈ M (X Y ) ⊆ M (B C). Therefore,




M B M C ⊆ M (B C) ⊆ M B M C.

3.25. Theorem. Every multiplication right module over a right distributive, right invariant ring is a
distributive module.
Proof. The proof follows from Lemma 3.24 and Note 1.11.

In the case of modules over commutative rings, the main results of this section turn into modifications
of the results proved in [3], [4], [17], and [78]. In the case of modules over noncommutative rings, the
main results of the presented work are proved in [88] and [89].

4. Multiplication Modules over Rings with Commutative Multiplication of Ideals

4.1. Theorem. Let A be a right invariant ring with commutative multiplication of ideals. For a right
A-module M , the following conditions are equivalent.
(1) M is a multiplication module.
(2) For any maximal ideal P of A, either M = M P or M = M P and there exists a cyclic submodule
X of M such that A = P + r(M/X).
(3) For any maximal ideal P of A, either P does not contain the annihilator of any cyclic submodule
of M or there exists a cyclic submodule X of M such that P does not contain r(M/X).
(4) For any maximal ideal P of A, either P does not contain the annihilator of any element of M or
there exist elements p ∈ P and x ∈ M such that M (1 − p) ⊆ xA.

3867
Proof. (1)=⇒(2) First, assume that M = M P . By Note 1.10(1), there exists a cyclic submodule X of
M such that A = P + r(M/X).
Now assume that M = M P . Let X be a cyclic submodule of M . Since M = M P , we have X = XP
by Note 1.12(1). Therefore, P does not contain r(X).
The implication (2)=⇒(3) follows from Note 1.13.
(3)=⇒(4) Assume that there exists an element m of M such that r(m) ⊆ P . Since A is a right
invariant ring, P contains the annihilator of the cyclic submodule mA of M . It follows from (3) that there
exists a cyclic submodule X of M such that the maximal ideal P does not contain r(M/X). Therefore,
A = P + r(M/X) and there exists an element p ∈ P such that M (1 − p) ⊆ X.
(4)=⇒(1) Let Y be a cyclic submodule of M . By Proposition 2.1, it is sufficient to prove that
Y ⊆ M (Y : M ). If (M (Y : M ) : Y ) = A, then Y = Y (M (Y : M ) : Y ) ⊆ M (Y : M ).
Assume that (M (Y : M ) : Y ) = A. Then there exists a maximal ideal P of A such that (M (Y :
M ) : Y ) ⊆ P . By assumption, either P does not contain the annihilator of any cyclic submodule of M
or there exists a cyclic submodule X of M such that P does not contain r(M/X). Since r(Y ) ⊆ (M (Y :
M ) : Y ) ⊆ P , it follows from the assumption that there exist an element p ∈ P and a cyclic submodule
X of M such that M (1 − p) ⊆ X. It follows that Y (1 − p)A is a submodule of the cyclic module X over
the right invariant ring A. Therefore, there exists an ideal D of A such that Y (1 − p)A = XD. We have
M D(1 − p)A = M (1 − p)D ⊆ XD ⊆ Y.
Therefore, D(1 − p)A ⊆ (Y : M ). It follows that
Y (1 − p)2 ⊆ Y (1 − p)A(1 − p)A = XD(1 − p)A ⊆ M (Y : M ).
Therefore,
(1 − p)2 ∈ (M (Y : M ) : Y ) ⊆ P and 1 ∈ pA + P = P ;
this is a contradiction.

4.2. Proposition. Let A be a right invariant ring with commutative multiplication of ideals. For a right
A-module M , the following conditions are equivalent.
(1) M is a multiplication module.
(2) For any maximal ideal P of A, either P does not contain the annihilator of any element of M or
there exist an element p of P and a multiplication submodule Y of M such that M (1 − p) ⊆ Y .
Proof. (1)=⇒(2) Let P be a maximal ideal of A. Since 0 ∈ P and M (1 − 0) is contained in the
multiplication submodule M of M , condition (2) holds.
(2)=⇒(1) Let P be a maximal ideal of A. Assume that P contains the annihilator of some element
m of M . Since A is a right invariant ring, r(mA) = r(m) ⊆ P . By Theorem 4.1, it is sufficient to prove
that there exist an element p of P and a cyclic submodule X of M with M (1 − p) ⊆ X. Since P contains
r(mA), it follows from (2) that there exist an element p1 of P and a multiplication submodule Y of M
such that M (1 − p1 ) ⊆ Y . Since r(mA) ⊆ P , it follows from Lemma 3.5(2) that the ideal P contains the
annihilator of the cyclic submodule m(1 − p1 )A of the multiplication module Y . By Theorem 4.1 applied
to the multiplication module Y , there exist an element p2 of P and a cyclic submodule X of Y such that
Y (1 − p2 ) ⊆ X. Then M (1 − p1 )(1 − p2 ) ⊆ X and M (1 − p) ⊆ X, where p = 1 − (1 − p1 )(1 − p2 ) ∈ P .

4.3. Theorem. Let A be a right invariant ring with commutative multiplication of ideals. For a right
A-module M , the following conditions are equivalent.
(1) M is a multiplication module.
(2) For every nonempty collection of right ideals Bi (i ∈ I) of A, we have
 
 
(M Bi ) = M (Bi + r(M )) ,
i∈I i∈I

3868
and for any submodule N of M and each right ideal C of A with N
M C, there exists an ideal B of A
such that B
C and N ⊆ M B.
Proof. The implication (2)=⇒(1) follows from Proposition 2.6.
(1)=⇒(2)

Let Bi (i ∈ I) be any nonempty collection of right ideals of
the right invariant ring A.
Let B =
i∈I (Bi + r(M )). By

Proposition 2.6, it is sufficient to prove that i∈I (M Bi ) = M B. Clearly
M B ⊆ i∈I (M Bi ). Let y ∈ i∈I (M Bi ) and let K = {a ∈ A | ya ∈ M B}. Assume that K = A. Then
there exists a maximal ideal P of A such that K ⊆ P . Since r(yA) = r(y) ⊆ K ⊆ P , it follows from
Theorem 4.1 that there exists a cyclic submodule mA of M such that A = P + r(M/mA). There exists
an element p of P such that M (1 − p) ⊆ mA. Then
  
y(1 − p) ∈ (M Bi A(1 − p)A) = (M (1 − p)Bi ) ⊆ (mBi ).
i∈I i∈I i∈I

For each i ∈ I, there exists an element bi ∈ Bi such that y(1 − p) = mbi . We choose j ∈ I. For each i ∈ I,
we have mbj = mbi so that mA(bj − bi ) ⊆ m(bj − bi )A = 0. Now
M (bj − bi )A(1 − p)A = M (1 − p)A(bj − bi )A ⊆ mA(bj − bi ) = 0.
Therefore,
(bj − bi )(1 − p) ∈ r(M ) and bj (1 − p)A ⊆ bi (1 − p)A + r(M ) ⊆ Bi + r(M ) (i ∈ I).
Therefore, bj (1 − p) ∈ B. Thus
y(1 − p)2 = mbi (1 − p) = mbj (1 − p) ∈ M B.
It follows that (1 − p)2 ∈ K ⊆ P
. Therefore, 1 ∈ (1 − p)2 A + P = P ; this is a contradiction. Thus K = A
and y ∈ M B. This shows that i∈I (M Bi ) ⊆ M B.

In connection with Theorem 4.3, we present the following example.
4.4. Example (cf. [17, p. 760]). Let F = Z/2Z be the residue field consisting of two elements, x and
y be two cyclic groups of order 2, G be the direct product x × y, A be the group ring F [G], and M
be the ideal (1 − x)A + (1 − y)A of A. It is directly verified that A is a commutative finite local ring with
Jacobson radical M , M 2 = 0, the ideal M is not principal, and
 
 
(M Bi ) = M (Bi + r(M ))
i∈I i∈I

for every nonempty collection of ideals Bi (i ∈ I) of A. We note that M is not a multiplication module,
since (1 − x)A = M B for any ideal B of A.
4.5. Theorem. Let A be a right invariant ring with commutative multiplication of ideals,  M be a right
A-module, and {Mi }i∈I be a collection of multiplication submodules of M such that M = i∈I Mi . Then
the following conditions are equivalent.
(1) M is a multiplication module.
(2) Mi = M (Mi  : M ) for all i ∈ I.
(3) A = r(m) + i∈I (Mi : M ) for every element m of M .
(4) A = r(X) + i∈I (Mi : M ) for every finitely generated submodule X of M .
(5) For every maximal ideal Pof A, either P does not contain the annihilator of any cyclic submodule
of M or there exist elements x ∈ i∈I Mi and p ∈ P such that M (1 − p) ⊆ xA.
Proof. (1)=⇒(2) If M is amultiplication module, then Mi = M (Mi : M ) (i ∈ I) by Note 1.3.
(2)=⇒(3) We set C = i∈I (Mi : M ) ⊆ A. Let m be an element of M . Assume that r(m) + C = A.
Then there exists a maximal ideal Q of A such that r(m) + C ⊆ Q. Since C ⊆ Q, we have Mi ⊆ M Q

3869
(i ∈ I) by 
(2); therefore, M = M Q. There exist a finite subset J of I and elements xj ∈ Mj (j ∈ J) such
that m = j∈J xj . Let j ∈ J. There exists an ideal B of A such that xj A = Mj B. Therefore,

xj A = Mj B = M (Mj : M )B = M Q(Mj : M )B = M (Mj : M )BQ = Mj BQ = xj Q.


Thus, xj (1 − qj ) = 0 for some element qj of Q. Let q = 1 − j∈J (1 − qj ) ∈ Q. Since A is a right invariant


ring, 1 − q = j∈J (1 − qj ) ∈ j∈J (1 − qj )A. Therefore, m(1 − q) ∈ j∈J xj (1 − qj )A = 0. Therefore,
1 − q ∈ r(m) ⊆ Q; this is a contradiction.
(3)=⇒(5) Let P be a maximal ideal of A. Without  loss of generality, we can assume that there exists
an element m of M such that r(mA) ⊆ P . By (3), i∈I (Mi : M ) ⊆ P . Therefore, there exists i ∈ I such
that (Mi : M ) ⊆ P . Thus there exists an element p1 of P such that M (1 − p1 ) ⊆ Mi . By Lemma 3.5(2),
the ideal P contains the annihilator of the cyclic submodule (m(1 − p1 )A) of the multiplication module
Mi . By Theorem 4.1 applied to the multiplication module Mi , there exist elements x ∈ Mi and p2 ∈ P
such that Mi (1 − p2 ) ⊆ xA. We denote by p the element 1 − (1 − p1 )(1 − p2 ) of the ideal P . Then

M (1 − p) = M (1 − p1 )(1 − p2 ) ⊆ Mi (1 − p2 ) ⊆ xA,

as required.
The implication (5)=⇒(1) follows from Theorem 4.1.
(4)=⇒(3) Since A isa right invariant ring,  r(m) = r(mA). This implies (3).
(3)=⇒(4) Let X = nj=1 xj A and let C = i∈I (Mi : M ). Since r(xj ) = r(xj A), it follows from (3)


that A = C + r(xj A) for j = 1, . . . , n. Therefore, A = C + nj=1 = C + r(X).

4.6. Corollary. Let A be a right invariant ring with commutative multiplication of ideals and let M be
a right A-module. Then the following assertions hold. 
(1) If there exists a collection {Mi }i∈I of multiplication submodules of M such that A = i∈I (Mi :
M ), then M is amultiplication module.
(2) If M = i∈I mi A for some elements mi∈I (i ∈ I), then M is a multiplication module if and only
if there exist ideals Bi of A such that mi A = M Bi for all i ∈ I.

Proof. (1) We have

 
M = M A = M ( (Mi : M )) ⊆ Mi ⊆ M.
i∈I i∈I

By Theorem 4.5, M is a multiplication module.

(2) By Proposition 3.1, every cyclic module mi A is a multiplication module. Therefore, the assertion
follows from Theorem 4.5.

4.7. Proposition. Let A be a right invariant ring with commutative multiplication of ideals, M be a
right A-module,
 and {Mi }i∈I be a collection of finitely generated multiplication submodules of M such that
M = i∈I Mi . Then the following conditions are equivalent.
(1) M is amultiplication
 module.
(2) A = r( j∈J Mj ) + k∈I (Mk : M ) for every finite subset J of I.

(3) A = r(Mi ) + k∈I (Mk : M ) for every i ∈ I.

Proof. The implication (1)=⇒(2) follows from Theorem 4.5 and the fact that j∈J Mj is a finitely
generated module.
The implication (2)=⇒(3) is obvious. 
(3)=⇒(2) Since J is a finite set and A = r(Mj ) + k∈I (Mk : M ) (j ∈ J), we have
   
A= r(Mj ) + (Mk : M ) = r( Mj ) + (Mk : M ).
j∈J k∈I j∈J k∈I

3870
 
(2)=⇒(1) Let m be an element of M . There exists a finite subset J of I such that mA ⊆ j∈J Mj .
Then   
A = r( Mj ) + (Mk : M ) ⊆ r(m) + (Mk : M ) ⊆ A.
j∈J k∈I k∈I
By Theorem 4.5, M is a multiplication module.

4.8. Proposition. Let A be a right invariant ring with commutative multiplication of ideals, M be
A-module, and {Mi }i∈I be a set of finitely generated multiplication submodules of M such that
a right
M = i∈I Mi . Then the following conditions are equivalent.
(1) M is a finitely generated multiplication module. 
(2) There exists a finite subset J of I such that A = j∈J (Mj : M ).
Proof.
 (1)=⇒(2) Since M is a finitely generated module, there exists a finite subset J of I such that
M = j∈J Mj . By Proposition 4.7,
   
A = r( Mj ) + (Mj : M ) = r(M ) + (Mj : M ) = (Mj : M ),
j∈J j∈J j∈J j∈J

since r(M ) ⊆ (Mj : M ) for any j ∈ J.

(2)=⇒(1) By Theorem 4.5, M is a multiplication module. Moreover
 
M = MA = M (Mj : M ) = Mj ⊆ M.
j∈J i∈J

Therefore, M = j∈J Mi so that M is a finitely generated module.

4.9. Theorem. Let A be a right invariant ring with commutative multiplication of ideals. For a finitely
generated right A-module M , the following conditions are equivalent.
(1) M is a multiplication module.
(2) M/M J(A) is a multiplication module.
(3) For every maximal ideal P of A, the module M/M P is cyclic.
(4) For every maximal ideal P of A, there exists a cyclic submodule X of M such that P does not
contain r(M/X).
Proof. The implication (1)=⇒(2) follows from Note 1.4.
(2)=⇒(3) All maximal ideals of A contain J(A). Therefore, without loss of generality, we can assume
that J(A) = 0. In this case, the assertion follows from Note 1.10(2).
(3)=⇒(4) Since the module M/M P is cyclic, there exists a cyclic submodule X of M such that
M = X + M P . We set M = M/X. By Note 1.4, M is a multiplication module; in addition, M = M P .
By Proposition 3.10(1), A = P + r(M ). Therefore, P does not contain r(M ) = r(M/X).
The implication (4)=⇒(1) follows from Theorem 4.1.

4.10. Theorem. Let A be a right invariant ring with commutative multiplication of ideals and let M be
a multiplication right A-module. Then the following conditions are equivalent.
(1) M is a finitely generated module.
(2) A = B + r(M ) for any ideal B of A such that M = M B.

(3) If B and C are ideals of A such that M B ⊆ M C and r(M ) ⊆ B C, then B ⊆ C.
(4) For each submodule N of M there exists a unique ideal B of A such that N = M B and r(M ) ⊆ B.
(5) There exists a lattice isomorphism ϕ from the lattice of all ideals of the factor ring A/r(M ) onto
the lattice of all submodules of the module M such that ϕ(B) = M B for every ideal B of A/r(M ).
Proof. The implication (1)=⇒(2) follows from Proposition 3.10(1).
The implication (2)=⇒(1) follows from Proposition 2.11(1).
The implication (4)=⇒(5) follows from Theorem 4.3.
The implication (5)=⇒(4) is obvious.

3871
 (1)=⇒(3) Let M be a finitely

generated multiplication module and let B and C be ideals of A such
that M B ⊆ M C and r(M ) ⊆ B C. Let b ∈ B. Assume that (C : bA) = A. Then there exists a maximal
ideal P of A such that r(M ) ⊆ (C : bA) ⊆ P . By Theorem 3.11, M = M P . By Theorem 4.1, there exist
elements x ∈ M and p ∈ P such that M (1 − p) ⊆ xA = X. In particular,
xb(1 − p) ∈ M B(1 − p) ⊆ M C(A(1 − p)A) = M (A(1 − p)A)C ⊆ xC,
so that there exists an element c of C such that b(1 − p) − c ∈ r(x) = r(X). But
M r(X)(1 − p) ⊆ M r(X)(A(1 − p)A) = M (A(1 − p)A)r(X) ⊆ Xr(X) = 0.
Therefore, r(X)(1 − p) ⊆ r(M ) ⊆ C. Thus, [b(1 − p) − c](1 − p) ⊆ C and this implies that b(1 − p)2 ∈ C so
that (1 − p)2 ∈ (C : b) = (C : bA) ⊆ P . Then 1 ∈ (1 − p)2 A + P = P ; this is a contradiction. Therefore,
(C : bA) = A and hence b ∈ C. It follows that B ⊆ C.
(3)=⇒(4) Let N be a submodule of M . Since M is a multiplication module, N = M (N : M ), where
the ideal (N

: M ) of A contains r(M ). Let B and C be two ideals of A such that N = M B = M C and
r(M ) ⊆ B C. By (3), B ⊆ C ⊆ B.
(4)=⇒(1) Let P be a maximal ideal of A with r(M ) ⊆ P . By Proposition 2.11(2), it is sufficient to
prove that M = M P . Assume that M = M P . Since M = M A and r(M ) ⊆ A, it follows from (4) that
P = A; this is a contradiction.

4.11. Theorem. Let A be a right invariant ring with commutative multiplication of ideals, M be a
multiplication right A-module, and N be a proper submodule of M . Then the following conditions are
equivalent.
(1) N is a prime submodule of M .
(2) N is a completely prime submodule of M .
(3) N = M P for some prime ideal P of A such that r(M ) ⊆ P .
(4) N = M P for some completely prime ideal P of A such that r(M ) ⊆ P .
(5) There exists a completely prime ideal P of A such that r(M ) ⊆ P , N = M P , and M/N is a
nonzero torsion-free multiplication module over the domain A/P .
(6) N = M P for some prime ideal P of A with r(M ) ⊆ P , and for every prime ideal Q of A such that
r(M ) ⊆ Q and N = M Q, we have that Q is a completely prime ideal and M/N is a nonzero torsion-free
multiplication module over the domain A/Q.
Proof. Since A is a right invariant ring, prime right A-modules coincide with completely prime right
A-modules. Therefore, we obtain the equivalences (1)⇐⇒(2) and (3)⇐⇒(4).
(1)=⇒(3) Since M is a multiplication module, N = M r(M/N ). By Proposition 2.28(1), r(M/N ) is
a prime ideal of A.
(4)=⇒(2) Assume that there exists a proper completely prime ideal P of A such that N = M P and
r(M ) ⊆ P . Let x be an element of M , X = xA, and a be an element of A such that xa ∈ M P . Assume
that a ∈/ P . It is sufficient to prove that x ∈ M P . Assume the contrary. Then X ⊆ M P and the ideal
(M P : X) of A is proper. Therefore, there exists a maximal ideal Q of A such that r(X) ⊆ (M P : X) ⊆ Q.
By Theorem 4.1, either M = M Q and Q does not contain the annihilator of any cyclic submodule of
M , or M = M Q and there exists a cyclic submodule Y of M such that A = Q + r(M/Y ). In addition
r(X) ⊆ Q. Therefore, M = M Q and there exists a cyclic submodule Y of M such that A = Q + r(M/Y ).
Thus, M (1 − q) ⊆ Y for some element q of Q. Then M (1 − q)r(Y ) = 0 and (1 − q)r(Y ) ⊆ r(M ) ⊆ P . The
element 1 − q is not contained in the completely prime ideal P , since P ⊆ Q and 1 − q ∈ / Q. In addition
(1 − q)r(Y ) ⊆ P . Therefore, r(Y ) ⊆ P . Let y be a generator of Y . Since A is a right invariant ring,
r(y) = r(Y ) ⊆ P . We have x(1 − q) ∈ yA and
x(1 − q)a ∈ X(A(1 − q)A)(M P : X) = X(M P : X)(A(1 − q)A)
⊆ M P (A(1 − q)A) = M (A(1 − q)A)P ⊆ Y P = yP.

3872
Therefore, there exist elements b ∈ A and p ∈ P such that x(1 − q) = yb and x(1 − q)a = yp. Then
yba = x(1 − q)a = yp, ba − p ∈ r(y) ⊆ P , and ba = p + (ba − p) ∈ P . Thus, ba is an element
of the completely prime ideal P and a ∈ / P . Therefore, b ∈ P . Then x(1 − q) = yb ∈ M P and
1 − q ∈ (M P : x) = (M P : X) ⊆ Q. Therefore, 1 = q + (1 − q) ∈ Q; this is a contradiction.
The implications (6)=⇒(5)

and (5)=⇒(4) are obvious.
The implication ((2) (4)) =⇒ (6) is directly verified.

4.12. Proposition. Let A be a right invariant ring with commutative multiplication of ideals, M be a
multiplication right A-module, h : A → A/r(M ) be the
natural ring epimorphism, E be the set of all ideals
E of A such that M = M E and r(M ) ⊆ E, and T = E∈E E. Then the following assertions hold.
(1) J(M ) = M JM = M JM ∗ , where J
M is the intersection of all maximal ideals P of A with M = M P
∗
and JM is the intersection of all maximal ideals Q of A with r(M ) ⊆ Q.
(2) M = M T , r(T ) ⊆ r(M ), T is an essential right ideal of A, and m ∈ mT for every element m of
M.
(3) T = T 2 if and only if r(M ) ⊆ T 2 .
(4) For every prime ideal P of A with r(M ) ⊆ P , we have either A = T + P or T ⊆ P .
 any nonempty collection {Bi }i∈I of right ideals of A with Bi ⊇ r(M ) (i ∈ I),we have that
(5) For
the sum i∈I h(Bi ) is a direct sum of right ideals of the ring h(A) if and only if the sum i∈I (M Bi ) is
a direct sum of submodules of the module M .
(6) A proper submodule N of M is a minimal prime submodule of M if and only if there exists a
prime ideal P of A such that P ⊇ r(M ), N = M P = M , and P/r(M ) is a minimal prime ideal of the
factor ring A/r(M ).
(7) Every prime submodule of M contains a minimal prime submodule of M .
(8) If N is a proper submodule of M , then N -rad(M ) is the intersection of all submodules K of M
such that N ⊆ K and K/N is a minimal prime submodule of M/N .
(9) If N is a proper submodule of M and B is the intersection of all prime ideals P of A with
P ⊇ r(M/N ), then N -rad(M ) = M B.
Proof. (1) Since A is a right invariant ring, all right primitive factor rings of A are division rings;
therefore, we can use Proposition 2.10. By Proposition 2.10(2), J(M ) = M JM . Let QM be the set of all
maximal ideals Q of A with r(M ) ⊆ Q and let J ∗ (M ) be the intersection of all modules M Q (Q ∈ QM ).
If P is a maximal ideal of A with M = M P , then A = P + r(M ), r(M ) ⊆ P , and P ∈ QM . Therefore,
∗ ∗
JM ⊆ JM and M JM ⊆ M JM = J(M ).
Let Q ∈ QM . If M = M Q, then J(M ) ⊆ M Q. If M = M Q, then it follows from Proposition 2.10(1)
that M Q is a maximal submodule of M , and therefore J(M ) ⊆ M Q. Thus, J(M ) ⊆ M Q in any case.
Since r(M ) ⊆ Q, it follows from Theorem 4.3 that
   
  
J ∗ (M ) = MQ = M  (Q + r(M )) M  (Q) = M JM ∗
.
Q∈QM Q∈QM Q∈QM

Therefore, J(M ) ⊆ J ∗ (M ) = M JM
∗ .

(2) By Theorem 4.3,

 
MT = M( E) = M E = M.
E∈E E∈E
Since M = M T , we have r(T ) ⊆ r(M ). Since M = M T , it follows from Note 1.12(1) that

m ∈ mT for
every element m of M . Let C be a right ideal of the right invariant ring A such that T C = 0. Then

TC ⊆ T C = 0 and C ⊆ r(T ) ⊆ r(M ) ⊆ T.


Therefore, C = T C = 0 and T is an essential right ideal of A.
(3) If T = T 2 , then T 2 = T ⊇ r(M ).

3873
 

Assume that r(M ) ⊆ T 2 . By (2), M = M T . Therefore, T 2 ∈ E. Since T = E∈E E , we have
T ⊆ T 2 ⊆ T and T = T 2 .
(4) Let P be a prime ideal of A with r(M ) ⊆ P . If M = M P , then T ⊆ P . Assume that M = M P .
Then there exists a cyclic submodule X of M such that X ⊆ M P . There exists an ideal B of A such that
X = M B. Since X ⊆ M P , we have B ⊆ P . Let X = xA. By Note 1.12(2), there exists an element t of
T such that x(1 − t) = 0. Then X(1 − t) = 0, since A is a right invariant ring. Therefore, M B(1 − t) = 0
and B(A(1 − t)A) ⊆ r(M ) ⊆ P . Since B ⊆ P and P is a prime ideal containing B(A(1 − t)A), we have
A(1 − t)A ⊆ P . Therefore, 1 − t ∈ P and A = T + P .
(5) The proof follows from the fact that by Theorem 4.3,
    
   
M h(Bi )  M h(Bj ) = M h(Bi )  h(Bj )
j=i j=i

for every i ∈ I.
(6) The proof follows from Theorem 4.11.
(7) The proof follows from (6) and Theorem 4.11.
(8) The proof follows from (7).
is a faithful A-module. Let P denote
(9) Without loss of generality, we can assume that M

the set of
all prime ideals P of A such that r(M/N ) ⊆ P . Then B = P ∈P P . By Theorem 4.3, M B = P ∈P (M P ).
Let P ∈ P. If M = M P then N -rad(M ) ⊆ M P . If M = M P then N = M r(M/N ) ⊆ M P implies
N -rad(M ) ⊆ M P by Proposition 2.28(1). It follows that N -rad(M ) ⊆ M P .
Conversely, assume that K is a prime submodule of M containing N . By Proposition 2.28(1) there
exists a prime ideal Q of A such that r(M/N ) ⊆ Q and K = M Q. Since M r(M/N ) = N ⊆ K = M Q =
M , it follows that r(M/N ) ⊆ Q by Theorem 4.11. Therefore, B ⊆ Q. Thus, M B ⊆ K. It follows that
M B ⊆ N -rad(M ). Therefore, N -rad(M ) = M B.

4.13. Theorem. Let A be a right invariant ring with commutative multiplication of ideals, M be a
nonzero multiplication right A-module, and T be the intersection of all ideals E of A such that M = M E
and r(M ) ⊆ E. Then the following conditions are equivalent.
(1) M is a finitely generated module.
(2) T = A.
(3) r(M ) ⊆ T 2 and T is a finitely generated ideal of A.
(4) M = M P for every prime ideal P of A.
(5) M = M P for every prime ideal P of A such that r(M ) ⊆ P and P/r(M ) is a minimal prime
ideal of the ring A/r(M ).
Proof. The equivalence of conditions (1) and (2) follows from Theorem 3.11.
The implications (2)=⇒(3) and (4)=⇒(5) are obvious.
(3)=⇒(2) By Proposition 4.12(3), T = T 2 . In addition, T is a finitely generated ideal of A. By
Proposition 3.10(2), there exists a central idempotent e of A such that T = eA. Then M = M T = M e
and 1 − e ∈ r(M ). By the definition of the ideal T , this ideal contains r(M ). Therefore, 1 − e ∈ T = eA.
Then e = 1 and T = A.
The implication (1)=⇒(4) follows from Theorem 3.11.
(5)=⇒(2) Assume that T = A. Let Q be a maximal ideal of A such that r(M ) ⊆ T ⊆ Q. Then
Q/r(M ) is a prime ideal of the factor ring A/r(M ). There exists an ideal P of A such that r(M ) ⊆ P ⊆ Q
and P/r(M ) is a minimal prime ideal of the ring A/r(M ). Then P is a prime ideal of A and T + P ⊆ Q.
Since A = T + P , it follows from Proposition 4.12(4) that T ⊆ P . Therefore, M = M P .

4.14. Corollary. Let A be a right invariant ring with commutative multiplication of ideals, M be a
faithful nonzero multiplication right A-module, and T be the intersection of all ideals C of A such that
M = M C. Then the following conditions are equivalent.
(1) M is a finitely generated module.

3874
 (2) T is a finitely generated ideal of A.
(3) M = M P for every minimal prime ideal P of A.
Proof. Since r(M ) = 0, the assertion follows from Theorem 4.13.

A module

M is called finitely cogenerated if for every nonempty
collection of submodules Ni (i ∈ I)
of M with i∈I Ni = 0 there exists a finite subset J of I such that j∈J Nj = 0.
4.15. Proposition. Let A be a right invariant ring with commutative multiplication of ideals, M be a
multiplication right A-module, and h : A → A/r(M ) be the natural ring epimorphism. Then the following
assertions hold.
(1) Soc(M ) = M Soc(h(A))h(A) .
(2) For every multiplication right ideal B of A, the right A-module M B is a multiplication module.
(3) If N is a submodule of M , then N is an essential submodule of the A-module M ⇐⇒
there exists a right ideal B of A such that N = M B, B ⊇ r(M ), and h(B) is an essential right ideal
of h(A).
(4) M is a uniform module if and only if A/r(M ) is a right uniform ring.
(5) The A-module M is finitely cogenerated if and only if the right h(A)-module h(A) is finitely
cogenerated.
(6) If h(A) is a right Noetherian ring, then M is a Noetherian module.
(7) If M = 0 and r(M ) contains a prime ideal P of A, then A/P is a domain and M is a finitely
generated, nonzero, uniform, torsion-free multiplication module over the domain A/P .
(8) Let P1 , . . . , Pn be prime ideals of A such that Pi ⊆ Pj (i = j) and M = M Pi (1 ≤ i ≤ n). Then
there exists an element m of M such that m ∈ / M Pi (1 ≤ i ≤ n).
(9) J(M ) is a superfluous submodule in M .
(10) If M is an Artinian module, then M is a cyclic module.
Proof. (1) We denote by E(M ) the set of all essential submodules of the module M . For any module, its
socle coincides with the intersection of all essential submodules of the module. Therefore, it follows from
(2) and Theorem 4.3 that
  
Soc(M ) = N= (M B) = M ( B) = M (Soc(h(A)h(A) ))).
N ∈E(M ) B∈E(h(A)h(A) ) B∈E(h(A)h(A) ))

(2) Let P be a maximal ideal of A. Assume that the following condition holds: (∗) either M = M P
and P does not contain the annihilator of any cyclic submodule of M , or B = BP and P does not contain
the annihilator of any cyclic submodule of BA . Since BP = P B, we have M BP = M P B = M B in both
cases. Therefore, P does not contain the annihilator of any cyclic submodule of M B by Note 1.12(2).
Now assume that the above condition (∗) does not hold. By Theorem 4.1, there exist elements p1 , p2 ∈
P , b ∈ B, and m ∈ M such that B(1 − p1 ) ⊆ bA and M (1 − p2 ) ⊆ mA. We set p = p1 + p2 − p1 p2 ∈ P .
Then
M B(1 − p) = M B(1 − p1 )(1 − p2 ) ⊆ M bA(1 − p2 ) ⊆ M (AbA)(A(1 − p2 )A)
= M (A(1 − p2 )A)(AbA) = M (1 − p2 )AbA ⊆ mA(AbA) = mbA;
i.e., mbA is a cyclic submodule of M B such that P does not contain r(M/X). It follows from Theorem
4.1 that M B is a multiplication module.
(3) The implication ⇐= follows from Proposition 2.3(1).

=⇒ There exists an ideal B of A such that N ⊇ r(M ) and N = M B. Assume that h(B) h(C) = 0
for some ideal C of A with C ⊇ r(M ). By Theorem 4.3, we have
  
N (M C) = (M B) (M C) = M (B C) = 0.
Since N is an essential submodule of M , we have M C = 0. Therefore, C ⊆ r(M ) and h(C) = 0. Thus
h(B) is an essential right ideal of A.
(4) The proof follows from Proposition 4.12(5).

3875
 (5) Assume

that M is finitely cogenerated. Let {Bi }i∈I be a nonempty collection of right ideals of A
such that i∈I h(Bi ) = 0 and Bi ⊇ r(M ) (i ∈ I). By Theorem 4.3,
 
 
(M Bi ) = M Bi = M r(M ) = 0.
i∈I i∈I


Since M is finitely cogenerated, there exists a finite subset J of I such that j∈J (M Bj ) = 0. Therefore,
   
M( Bj ) ⊆ (M Bj ) = 0, Bj = r(M ), and h(Bj ) = 0.
j∈J j∈J j∈J j∈J

It follows that the right h(A)-module h(A) is finitely cogenerated.

Conversely, assume that the right h(A)-module
h(A) is finitely cogenerated. Let Ni (i ∈ I) be a
nonempty collection of submodules of M
such that
i∈I Ni = 0. For
each i ∈ I there exists an ideal Ci
of A such that Ni = M Ci . Clearly M
( i∈I Ci ) ⊆ i∈I Ni . Thus i∈I Ci = 0 and by hypothesis there
exists a finite subset J of I such that j∈J Ci = 0. By Theorem 4.3,
  
Ni = (M Ci ) = M ( Ci ) = 0.
i∈J i∈J i∈J
Therefore, M is finitely cogenerated.
(6) The proof follows from Theorem 4.3.
(7) By Theorem 4.11, 0 is a completely prime submodule of M , A/P is a domain, and M/0 ∼ = M is
a nonzero torsion-free multiplication module over the domain A/P . Let A = A/P . Then M is a nonzero
multiplication right module over the domain A, 0 is a minimal prime ideal of A, and rA (M ) = 0. By
Theorem 4.13 and Proposition 2.3(4), M is a finitely generated uniform A-module. Therefore, M is a
finitely generated uniform A-module.

(8) For each i ∈ {1, . . . , n}, we have j=i Pj ⊆ Pi . By (7), M/M Pi is a nonzero torsion-free module
over the domain A/Pi (i = 1, . . . , n). Therefore,

M ( Pj ) ⊆ M Pi .
j=i
Therefore, we can choose an element mi so that

mi ∈ M ( Pj ) \ M Pi .
j=i

We set m = m1 + · · · + mn ∈ M . It is directly verified that m ∈

/ M Pi (1 ≤ i ≤ n).
(9) Since every factor module of M is a multiplication module, it is sufficient to prove that either
M = 0 or M = J(M ). Let M = 0 and let X be a nonzero cyclic submodule M . There exists a maximal
ideal P of A with X = XP . By Note 1.13, M = M P . By Proposition 2.10(1), M P is a maximal
submodule in M . Therefore, M = J(M ).
(10) By (9), J(M ) is a superfluous submodule in M . By Theorem 2.14(2), M is a cyclic module.

4.16. Proposition. Let A be a right invariant ring with commutative multiplication of ideals and let M
be a multiplication right A-module. Then the following conditions are equivalent.
(1) M is an Artinian module.
(2) A/r(M ) is a right Artinian ring.
(3) M is a cyclic, Artinian, and Noetherian module.
Proof. (1)=⇒(2) By Proposition 4.15(10), M is a cyclic module. Then M is a faithful cyclic Artinian
right module over the right invariant ring A/r(M ). Therefore, A/r(M ) is a right Artinian ring.
(2)=⇒(3) It follows from Theorem 4.3 that M is an Artinian module. By Proposition 4.15(10), M
is a cyclic module. Since the right Artinian ring A/r(M ) is right Noetherian, M is a Noetherian module
by Proposition 4.15(6).

3876
 The implication (3)=⇒(1) is obvious.

4.17. Lemma. Let A be a right invariant ring, P be a maximal ideal of A such that the right localization
AP exists, and M be a right A-module. Then the following assertions hold.
(1) If m is an element of M and N is a submodule of M , then (mA)P ⊆ NP if and only if mA(1 −
p)A ⊆ N for some element p of P . In particular, (mA)P = 0 if and only if mA(1 − p)A = 0 for some
element p of P .
(2) If m is an element of M and N is a submodule of M , then (N : mA)P = (NP : (mA)P ).
(3) If AP is a right invariant ring, then (X : Y )P is contained in the ideal (XP : YP ) of AP .
(4) If MP is a cyclic AP -module, then M contains an element x such that for every element m of
M , there exists an element p of P with m(1 − p) ∈ xA.
(5) If M is a finitely generated module and MP is a cyclic AP -module, then there exist elements
x ∈ M and p ∈ P such that M (1 − p) ∈ xA.
Proof. Let S = A \ P , f : A → AP be the canonical ring homomorphism, g : M → MP be the canonical
A-module homomorphism, h : A → A/P be the natural epimorphism, and m be an element of M .
(1) Since A is a right invariant ring, (N : m) = (N : mA). Therefore, the assertion follows from
Lemma 3.17(3).
(2) Since A is a right invariant ring, (N : m) = (N : mA). Therefore, the assertion follows from
Lemma 3.19(3).
(3) The proof follows from Lemma 3.19(2).
(4) By assumption, MP is a cyclic AP -module. There exist elements x ∈ M and s ∈ S such that
MP = g(x)f (s)−1 AP = g(xA)AP . Since g(m) ∈ g(xA)AP , it follows from Lemma 3.16(6) that there
exists an element s of S such that ms ∈ xA. Since A/P is a simple right invariant ring, A/P is a division
ring. Therefore, the nonzero element h(s) of h(A) is invertible in A/P . Therefore, there exist elements
t ∈ S and p ∈ P such nthat st = 1 − p. Then m(1 − p) = mst ∈ xA.
(5) Let M = i=1 mi A. By (4), there exist elements p1 , . . . , pn ∈ P such that mi (1 − pi ) ∈ xA

by p the element 1 − (1 − p1 ) · · · (1 − pn ) of the ideal P . Since A is a right
for i = 1, . . . , n. We denote
invariant ring, A(1 − p) ⊆ ni=1 ((1 − pi )A). Therefore,

n 
n
M (1 − p) = mi A(1 − p) ⊆ mi (1 − pi )A ⊆ xA.

i=1 i=1

4.18. Theorem. Let A be a right invariant ring with commutative multiplication of ideals. Then for
every maximal right ideal P of A, the right localization AP exists and AP is a right invariant ring with
commutative multiplication of ideals. Further, if M is a right A-module then the following conditions are
equivalent.
(1) M is a multiplication module. 
(2) MP = 0P for every maximal right ideal P of A such that P ⊇ m∈M (mA : M ).
(3) For each maximal ideal P of A with MP = 0P , there exist elements s ∈ A \ P and x ∈ M with
M s ⊆ xA.
(4) For every right A-module extension M ⊆ K and each maximal ideal P of A with MP = 0P , the
AP -module MP is cyclic and (NP : MP ) = (N : M )P for each submodule N of K.
(5) For each maximal ideal P of A with MP = 0P , the AP -module MP is cyclic and (NP : MP ) ⊆
(N : M )P for each submodule N of M .
Proof. For every maximal right ideal P of A, it follows from Lemma 3.20(6) that the right localization
AP exists and AP is a right invariant ring with commutative multiplication of ideals.
The implication (1)=⇒(2) follows from Lemma 3.18(5). 
(2)=⇒(3) Let P be a maximal ideal of A with MP = 0P . It follows from (2) that P ⊇ m∈M (mA :
M ). Therefore, there exists an element x of M such that P ⊇ (xA : M ). Therefore, there exists an
element s ∈ A \ P with M s ⊆ xA.

3877
 (3)=⇒(4) Let K be a right A-module such that M is a submodule of K, P be a maximal ideal with
MP = 0P , and f : A → AP be the canonical ring homomorphism. Since MP = 0P , it follows from (3)
that there exist elements s ∈ A \ P and x ∈ M with M s ⊆ xA. Therefore, MP coincides with the cyclic
AP -module (xA)P . By Lemma 4.17(3), (N : M )P ⊆ (NP : MP ). If y ∈ (N : M s), then
AyA ⊆ (N : M s), M (AsA)(AyA) ⊆ N, M (AyA)(AsA) ⊆ N,
−1
ys ∈ (N : M ), andf (y) = f (ys)f (s) ∈ (N : M )P .
In addition, (N : xA)P ⊇ (NP : (xA)P ) by Lemma 4.17(2). Thus,
(NP : MP ) = (NP : (xA)P ) ⊆ (N : xA)P ⊆ (N : M s)P ⊆ (N : M )P .
The implication (4)=⇒(5) is obvious. 
(5)=⇒(2) Let P be a maximal right ideal of A such that P ⊇ m∈M (mA : M ). Assume that
MP = 0P . It follows from (5) that the AP -module MP is cyclic and (NP : MP ) ⊆ (N : M )P for each
submodule N of M . There exists an element x of M such that MP = (xA)P . Then A P = ((xA)P : MP ) =
(xA : M )P , since
 (NP : M P ) ⊆ (N : M )P for each submodule N of M . Since P ⊇ m∈M (mA : M ), we
have PP ⊇ ( m∈M (mA : M ))P ⊇ (xA : M )P = AP . This is a contradiction, since PP is the Jacobson
radical of AP by Lemma 4.17(1).
(2)=⇒(1) Let P be a maximal right ideal of A. By Theorem 4.1, it is sufficient to prove that either
P does not contain the annihilator of any element of M or there exist elements p ∈ P and m ∈ M such
that M (1 − p) ⊆ mA. 
Assume that P ⊇ m∈M (mA : M ). By assumption MP = 0P . It follows from Lemma 3.17(3) that
for every element m of M , there exists an element p of P such that m(1 − p) = 0. Therefore, P does not
contain the annihilator of any element of M .
Assume that P ⊇ m∈M (mA : M ). Then there exists an element m of M such that (mA : M ) ⊆ P .
Therefore, A = P + (mA : M ) and there exists an element p of P such that 1 − p ∈ (mA : M ). Then
M (1 − p) ⊆ M (mA : M ) ⊆ mA.

4.19. Corollary. Let A be a right invariant ring with commutative multiplication of ideals and let M be
a projective right ideal of A. Then M is a multiplication right ideal.
Proof. It follows from Theorem 4.18 that for every maximal right ideal P of A, the right localization
AP exists and AP is a right invariant ring. Let N be a submodule of MA and let P be a maximal
ideal of A with MP = 0P . Since A is a right invariant ring, we have M N ⊆ N , N ⊆ (N : M ), and
NP ⊆ (N : M )P . By Theorem 4.18, it is sufficient to prove that MP is a cyclic right AP -module and
(NP : MP ) ⊆ (N : M )P . By Lemma 3.17(1), AP is a local ring. By Lemma 3.17(4), the nonzero right
ideal MP of AP is free.
Assume that the free right AP -module MP is cyclic. Then (NP : MP ) = NP ⊆ (N : M )P .
Assume that MP is a noncyclic free right AP -module with basis {mi }i∈I consisting

of at

least two
elements. Let mi and mj be two distinct elements of the basis. Then mi mj ∈ mi A Amj ⊆ mi A mj A =
0. Then mj ∈ r(mi ) = 0; this is a contradiction.

4.20. Theorem. Let A be a right invariant ring with commutative multiplication of ideals, K be an
A-module, and M be a submodule of K. Then the following conditions are equivalent.
(1) M is
a multiplication module.


(2) M B N = M ((N : M ) B) for every ideal B of A and each submodule N of K.
(3) If B is an ideal of A and N is a submodule of M with N ⊆ M B, then N = M C for some ideal
C ⊆ B.
Proof. (1)=⇒(2) It follows from Theorem 4.18 that for each maximal ideal P of A with MP = 0P , the
ring AP is a right invariant ring with commutative multiplication of ideals, the AP -module

MP
is cyclic,
and (NP : MP ) = (N : M )P for each submodule N of K. By Lemma 3.16(8), XP YP = (X Y )P for
any two submodules X and Y of K. By Lemma 3.19(1), (XB)P = XP BP for any submodule X of K.

3878



First, assume that M =
mA is cyclic. We
show that N mAB = mA((N
: mA) B). It is
sufficient to prove that mAB N ⊆ mA((N : mA) B). Let x
be an element of N mAB.
There exists
an element b of B such that x = mb ∈ N . Then b ∈ (N : mA) B,
so x ∈ mA((N : mA) B).

Now we consider the general case. If MP = 0P , then (M B N )P = 0P = (M ((N : M ) B))P .
Therefore, we can assume that MP = 0P .


Then MP is a cyclic AP -module and, therefore, MP BP NP = MP ((NP : MP ) BP ). Therefore,
   
(M B N )P = (M B)P NP = MP BP NP = MP ((NP : MP ) BP )
 
= MP ((N : M )P ) BP ) = (M ((N : M ) B))P .




Thus, (M B
N )P = (M ((N : M ) B))P for every maximal ideal P of A. By Lemma 3.18(3), M B N =
M ((N : M ) B).



(2)=⇒(3) We have N = M B N = M ((N : M ) B), where (N : M ) B ⊆ B.
(3)=⇒(1) We can take B = A.

4.21. Theorem. Let A be a right invariant ring with commutative multiplication of ideals and let M be
a finitely generated right A-module. Then the following conditions are equivalent.
(1) M is a multiplication module.
(2) For every maximal right ideal P of A, the right AP -module MP is cyclic.
Proof. For every maximal right ideal P of A, it follows from Lemma 3.20(6) that the right localization
AP exists and AP is a right invariant ring with commutative multiplication of ideals. The implication
(1)=⇒(2) follows from by Proposition 3.21(2).
(2)=⇒(1) It follows from Lemma 4.17(4) that for every maximal right ideal P of A, there exist
elements x ∈ M and p ∈ P such that M (1 − p) ∈ xA. This implies that for every maximal ideal P of A,
there exists a cyclic submodule X of M such that P does not contain r(M/X). By Theorem 4.9, M is a
multiplication module.

4.22. Theorem. Let A be a right invariant ring with commutative multiplication of ideals and let M be
a right A-module. Then the following conditions on M are equivalent.
(1) M is a multiplication
 module.
(2) X = X(m∈M (mA : M )) for every cyclic submodule X of M .
(3) X = X( m∈M (mA : M )) for every submodule X of M .
Proof. (1)=⇒(2) Let X be a cyclic submodule ofM . By Proposition 3.1, every cyclic submodule
mA of M
 is a multiplication module. Since M = m∈M mA, it follows from Theorem 4.5 that A =
r(X) + m∈M (mA : M ). Therefore,
 
X = XA = X(r(X) + (mA : M )) = X( (mA : M )).
m∈M m∈M

(2)=⇒(3) Let X be a submodule of M . It follows from (2) that xA = xA( m∈M (mA : M )) for
every element x of X. Therefore,
     
X= xA = (xA( (mA : M ))) = ( xA)( (mA : M )) = X( (mA : M )).
x∈X x∈X m∈M x∈X m∈M m∈M

(3)=⇒(1) Let P be a maximal ideal of A. By Theorem 4.1, it is sufficient to prove that either
P does not contain the annihilator of any element of M or there exist elements p ∈ P and x ∈ M
such that M(1 − p) ⊆ xA. If P ⊇ m∈M (mA : M ) and m is an element of M , then, by assumption,
mA = mA( m∈M (mA : M )) ⊆ mAP , so r(m) ⊆ P .
Assume that P ⊇ m∈M (mA : M ). Then there exists an element x of M such that (xA : M ) ⊂ P .
Therefore, (xA : M ) + P = A and there exist elements b ∈ (xA : M ) and p ∈ P such that 1 = b + p. Then
M (1 − p) ⊆ xA.

3879
4.23. Proposition. Let A be an invariant ring with commutative multiplication of ideals and let M be
an ideal of A that contains a regular element of A. Then the following conditions are equivalent.
(1) M is a multiplication right A-module.
(2) M is a multiplication left A-module.
(3) M is an invertible ideal.
(4) MA and A M are finitely generated multiplication modules.
Proof. It is sufficient to prove the equivalence of conditions (1), (3), and (4).
The equivalence of conditions (3) and (4) follows from Proposition 3.14.
The implication (4)=⇒(1) is obvious.
(1)=⇒(4) By Proposition 3.14, it is sufficient to prove that M is a multiplication left A-module.
Let N be a submodule of the left A-module M . Since A is an invariant ring, N is a submodule of the
right A-module M . Since M is a multiplication right A-module, there exists an ideal B of A such that
N = M B. Since A is a ring with commutative multiplication of ideals, BM = M B = N . Therefore, M
is a multiplication left A-module.

4.24. Theorem. Let A be an invariant ring with commutative multiplication of ideals and let M be a
nonzero right A-module such that r(M ) contains a prime ideal P of A. Then the following conditions are
equivalent.
(1) M is a multiplication module.
(2) M is a finitely generated multiplication module.
(3) A/P is a domain and MA ∼ = BA for some invertible ideal B of the domain A/P .
Proof. The implication (1)=⇒(2) follows from Proposition 4.15(7).
(2)=⇒(3) By Proposition 4.15(7), M is a finitely generated, nonzero, uniform, torsion-free multi-
plication module over the domain A/P . It is directly verified that every finitely generated, uniform,
torsion-free module over an invariant domain is isomorphic to an ideal of the domain. Therefore, there
exists a nonzero ideal B of the domain A/P such that MA ∼= BA . By Proposition 4.23, B is an invertible
ideal.
The implication (3)=⇒(1) follows from Proposition 3.14.

4.25. Theorem. Let A be a right invariant ring with commutative multiplication of ideals and let M be
a multiplication right A-module. Then the following assertions hold.
(1) If P is a prime ideal of A, then A/P is a domain and M/M P is a finitely generated, nonzero,
uniform, torsion-free multiplication module over the domain A/P .
(2) For any finite set {P1 , . . . , Pn } of prime ideals of A, the factor modules
   
M/(M P1 . . . M Pn ), M/M (P1 . . . Pn ), and M/M (P1 · · · · · Pn )





are finitely generated; in addition, if at least one of the modules M P1 . . . M Pn , M (P1 . . . Pn ),
and M (P1 · · · · · Pn ) is a superfluous submodule of M , then M is a finitely generated module.


(3) For any finite set {M1 , . . . , Mn } of prime submodules


of M , the factor module M/(M1 . . . Mn )
is finitely generated; in addition, if the module M1 . . . Mn is a superfluous submodule of M , then M
is a finitely generated module.
(4) If M contains only finitely many minimal prime submodules, then M is a finitely generated
module.
(5) If every semiprime ideal of A is a finite intersection of prime ideals, then M is a finitely generated
module.
(6) If either A satisfies the ascending chain condition on semiprime ideals or every semiprime factor
ring of A does not contain an infinite direct sum of nonzero ideals, then M is a finitely generated module.
(7) M is a finite-dimensional A-module ⇐⇒
the ring A/r(M ) does not contain an ideal that is a direct sum of infinitely many nonzero ideals of
A/r(M ).

3880
Proof. (1) The proof follows from Proposition 4.15(7).
(2) By (1), all modules M/M Pi are finitely generated. By Lemma 2.4(3), the factor modules
   
M/(M P1 . . . M Pn ), M/M (P1 . . . Pn ), and M/M (P1 · · · · · Pn )
are finitely generated. Therefore, there exist finitely generated submodules X, Y , and Z of M such that
   
M = X + M P1 . . . M Pn ) = Y + M (P1 . . . Pn ) = Z + M (P1 · · · · · Pn ).





Therefore, if at least one of the modules M P1 . . . M Pn , M (P1 . . . Pn ), and M (P1 · · · · · Pn ) is a
superfluous submodule of M , then M is equal to one of the finitely generated modules X, Y , and Z.
(3) By Theorem 4.11, there exist prime ideals P1 , . . . , Pn of A such that Mi = M Pi for every i.
Therefore, the assertion follows from (2).
(4) Without loss of generality, M is faithful. By Proposition 4.12(6), there exist a positive integer
n and minimal prime ideals Pi (1 ≤ i ≤ n) of A such that M Pi are the minimal prime submodules of
M . It follows that if P is a minimal prime ideal of A and P = Pi (1 ≤ i ≤ n), then M = M P . If
S is the intersection of all minimal prime ideals
of A
the exception of the Pi (1 ≤ i ≤ n), then

with
the prime radical rad(A) of A coincides with S P1 · · · Pn . By Theorem 4.3, M = M S. For each
1 ≤ i ≤ n, the module M/M Pi is finitely generated by (1). We set B = P1 · · · · · Pn . By Lemma 2.4(3),
the factor
module M/M B is finitely generated. Let M = m1 A + · · · + mk A + M B. Since M = M S and
BS ⊆ B S = rad(A) ⊆ J(A), we have
M = m1 A + · · · + mk A + M BS = m1 A + · · · + mk A + M J(A);
in addition M J(A) is a superfluous submodule of M by Proposition 2.15(2). Therefore, M = m1 A + · · · +
mk A.
(5) We have that the intersection of all prime ideals of A that contain r(M ) is a finite intersection
of prime ideals. It follows from Proposition 4.12(6) and Proposition 4.12(8) that M contains only finitely
many minimal prime submodules. By (4), M is a finitely generated module.
(6) It is directly verified that every semiprime ideal of A is a finite intersection of prime ideals.
Therefore, the assertion follows from (5).
(7) The implication ⇐=(1) follows from Proposition 2.3(2).
=⇒ Let h : A → A/r(M ) be the natural ring epimorphism ∞
 and let {Bi }i=1 be a nonempty set of
ideals of A such that all ideals Bi contain r(M
 ) and the sum i∈I h(Bi ) is a direct sum of ideals of the
ring h(A). By Proposition 4.12(5), the sum i∈I (M Bi ) is a direct sum of submodules of the module M .
In addition, every module M Bi is a fully invariant submodule of M by Note 1.5(1). By assumption, there
exists a positive integer n such that M Bi = 0 for all i ≥ n. It follows that h(Bi ) = 0 for all i ≥ n.

4.26. Proposition. Let A be an invariant ring with commutative multiplication of ideals and let M be
a nonzero faithful multiplication right A-module. Then the following assertions hold.
(1) If there exists a cyclic submodule X of M such that M s ⊆ X for some regular element s of A,
then M is a finitely generated, torsion-free module and there exists an invertible ideal B of the ring A
such that BA ∼ = M.
(2) If M is a uniform module, then M is a finitely generated, torsion-free module, and there exists
an invertible ideal B of the ring A such that BA ∼
= M.
Proof. (1) Let X = xA and let B be the ideal (M aA : x) of A. Since A is an invariant ring and xA
is a cyclic A-module, M s = M As = M sA is a submodule of xA and M s = xB. Since xs ⊆ M a, the
ideal B contains the regular element s. The invariant ring A has the classical ring of fractions Q. By
Proposition 3.13(3), M is a torsion-free module. Therefore, we can assume that M is a submodule of the
right A-module M Q. Then M = xBs−1 . Since A is an invariant ring with commutative multiplication of
ideals, Bs−1 = s−1 B and M = xs−1 B. Since M is a torsion-free module, it is directly verified that there
exists an isomorphism ϕ : MA → BA such that ϕ(xs−1 b) = b (b ∈ B). Therefore, BA is a multiplication
right A-module. By Proposition 4.23, B is an invertible ideal of A and M is a finitely generated module.

3881
 (2) Since M is a faithful uniform multiplication right A-module, A is a right uniform ring by Propo-
sition 4.15(4). The right uniform, left invariant ring A is a left uniform ring. We denote by P the set of
all right zero-divisors of A. Since A is a left uniform ring, all nonzero left ideals are essential left ideals.
Therefore, the set P is an ideal of A, since P coincides with the left singular ideal Sing(A A) of A. Since
A is an invariant ring with commutative multiplication of ideals, the ideal P coincides with the set of all
left zero-divisors of A, P = Sing(AA ), and A \ P coincides with the set of all regular elements of A.
Assume that M = M P . Let m be a nonzero element of M . By Note 1.12(2), there exists an element
p of the ideal P such that m(1 − p) = 0. We set z(M ) = {a ∈ A | ma = 0 for some 0 = m ∈ M }. By
Proposition 3.13(2), P = z(M ). Since m(1 − p) = 0, 1 − p ∈ z(M ) = P and 1 = p + (1 − p) ∈ P ; this is a
contradiction.
We have proved that M = M P . There exists a cyclic submodule X of M such that X
M P . Since
M is a multiplication module, there exists an ideal S of A such that M S = X. Since X ⊆ M P , we have
S ⊆ P . Therefore, there exists an element s ∈ S \ P such that M s ⊆ X. Since s is a regular element, it
follows from (1) that M is a finitely generated, torsion-free module and there exists an invertible ideal B
of the ring A such that BA ∼ = M.

4.27. Theorem. Let A be an invariant ring with commutative multiplication of ideals and let M be a
multiplication right A-module. Then the following assertions hold.
(1) If M is a finite-dimensional module, then M is a finitely generated module.
(2) If there exists a submodule N of J(M ) such that the factor module M/N is a finite-dimensional
module, then M is a finitely generated module.

Proof.

(1)
Since M is a finite-dimensional module, there exist submodules M1 , . . . , Mn of M such that
M1 · · · Mn = 0 and all factor modules M/Mi are uniform. By Proposition 4.26(2), every faithful
uniform multiplication A/r(Mi )-module M/Mi is a finitely generated A/r(Mi )-module. Therefore, every
A-module M/Mi is finitely generated. By Proposition 2.11(4), M is a finitely generated module.
(2) By (1) applied to the finite-dimensional multiplication module M/N , the module M/N is finitely
generated. Therefore, there exists a finitely generated submodule X of M such that M = X + N ⊆
X + J(M ) ⊆ M . Then M = X + J(M ). By Proposition 2.15(2), J(M ) is a superfluous submodule of M .
Therefore, M = X.

Let L be a partially ordered set. If L1 , L2 ∈ L and L2 ≤ L1 , then L1 /L2 denotes the partially ordered
subset in L such that N ∈ L1 /L2 if and only if L2 ≤ N ≤ L1 . The deviation dev(L) of L can be defined
by transfinite induction as follows:
dev(L) = −1 if and only if L contains exactly one element.
We consider an ordinal α ≥ 0; also, assume that we have already defined which partially ordered sets
have deviation β for ordinals β < α.
Then dev(L) = α if and only if
(i) we have not already defined dev(L) = β for some β < α, and
(ii) for every (countable) descending chain L1 > L2 > . . . of elements of L, we have dev(Li /Li+1 ) < α
for all but finitely many subscripts i (this means that, for all but finitely many i, the deviation of Li /Li+1
has previously been defined, and, therefore, is an ordinal less than α).
We accept the convention that an assertion like dev(L) = α is an abbreviation for “the deviation of
L exists and equals to α.” We note that an arbitrary partially ordered set does not necessarily have a
deviation.
If the lattice Lat(M ) of submodules of a module M has deviation, then this deviation is called the
Krull dimension of M ; it is denoted by Kdim(M ).

4.28. Lemma ([24]). For a right module M over a ring A, the following assertions hold.
(1) If N is a submodule of M , then M has the Krull dimension if and only if N and M/N have the
Krull dimension. In this case, Kdim(M ) = sup{Kdim(N ), Kdim(M/N )}.

3882
 (2) If M is a finitely generated module and the module AA has the Krull dimension, then M has the
Krull dimension.
(3) If M has the Krull dimension, then M is a finite-dimensional module.
Proof. (1) The assertion is directly verified.
(2) The proof follows from (1) and the fact that the finitely generated module M is isomorphic to a
factor module of a finite direct sum of copies of the module AA with the Krull dimension.
(3) Assume the contrary. Without loss of generality, we can assume that Kdim(M) > −1 and M
has the Krull dimension α with α the minimal among the Krull dimensions of modules that are not
finite-dimensional. Let ⊕∞ ∞
i=1 Ni be a submodule of M such that Ni = 0 for all i and let Mn = ⊕j=1 N(2n j) .
We consider the infinite descending chain M1 ⊃ M2 ⊃ M3 ⊃ . . . . Every factor module Mi /Mi+1 is
an infinite direct sum of nonzero submodules, and Kdim(Mi /Mi+1 ) ≤ α. The minimality of α forces
Kdim(Mi /Mi+1 ) = α for all i. By definition, Kdim(M ) > α; this is a contradiction.

4.29. Proposition. Let A be an invariant ring with commutative multiplication of ideals and let M be a
multiplication right A-module. Then M has the Krull dimension if and only if the right A/r(M )-module
A/r(M ) has the Krull dimension. In this case, Kdim(M ) = Kdim((A/r(M ))A/r(M ) ).
Proof. Assume that A/r(M ) has the Krull dimension. By Lemma 4.28(3), the factor ring A/r(M ) is
right finite-dimensional. By Proposition 2.3(2), M is a finite-dimensional module. By Theorem 4.27(1),
M is a finitely generated A-module. Since M is a finitely generated A/r(M )-module, MA/r(M ) has the
Krull dimension by Lemma 4.28(2). Therefore, MA has the Krull dimension.
Conversely, assume that M has the Krull dimension. By Lemma 4.28(3), M is a finite-dimensional
module. By Theorem 4.27(1), M is a finitely generated module. Assume that M = m1 A + · · · + mk A
for some positive integer k and elements mi ∈ M (1 ≤ i ≤ k). We define ϕ : AA → M (k) by ϕ(a) =
(m1 a, . . . , mk a) for all a ∈ A. Then ϕ is an A-monomorphism and hence A has the Krull dimension and
Kdim(M ) = Kdim(A) by Lemma 4.28(1) and Lemma 4.28(2).
It follows from Lemma 4.28(1) that Kdim(M ) = Kdim((A/r(M ))A/r(M ) ).

4.30. Example ([17]). There exist a commutative domain A and a multiplication A-module M such
that M is not a finitely generated module.
Proof. Let K be a field, A be the polynomial ring K[x1 , x2 , x3 , . . . ] in a countably infinite set of inde-
terminates x1 , x2 , x3 . . . , B = x1 A + x2 A + x3 A + . . . , C = (x1 − x21 )A + (x2 − x22 )A + (x3 − x23 )A + . . . ,
and M = B/C. By Corollary 4.6(2), M is a multiplication ideal of the ring A/C. Therefore, M is a
multiplication A-module. The module M is not finitely generated.

The commutative versions of 4.1, 4.3, 4.6(2), 4.9, and 4.19 are proved in [17]. The commutative
versions of 4.2, 4.5, 4.6(1), 4.7, and 4.8 are proved in [78]. In the presented form, the above assertions are
proved in [88] and [89].

5. Multiplication Modules over Regular Rings

5.1. Proposition ([33]). (1) If {Mi }i∈I is a set of countably generated modules and P is a direct
summand of ⊕i∈I Mi , then P is a direct sum of countably generated modules.
(2) Every direct summand of a direct sum of finitely generated modules is a direct sum of countably
generated modules.
(3) Every projective module is a direct sum of countably generated projective modules.
Proof. (1) Assume that M = ⊕i∈I Mi = P ⊕ Q. We proceed by transfinite induction to obtain a well-
ordered properly ascending chain of submodules Sα of M with α being an ordinal satisfying the following
conditions: 
(i) M = Sα ;
(ii) S0 = 0;

3883
 (iii) if α is a nonzero nonlimit ordinal,then the module Sα /Sα−1 is countably generated;
(iv) if α is a limit ordinal, then Sα = β<α Sβ ;
(v) Sα is a direct sum of some
of the modules M
i for all α;
(vi) Sα = Pα ⊕ Qα , Pα = P Sα , and Qα = Q Sα . 
Let S0 = 0. If α is a limit ordinal, then we define Sα = β<α Sβ . Assume that α is a nonlimit ordinal
and Sα−1 = M . We choose Mi such that Mi ⊆ Sα−1 . Let {x11 , x12 , x13 , . . . } be a countable generator
system of the module Mi . Then x11 is a sum of the P -component and the Q-component. These two
components are contained in a direct sum of finitely many modules Mi which has a countable generator
system {x21 , x22 , x23 , . . . }. Decomposing the element x12 as x11 before, we obtain a countable sequence
{x31 , x32 , x33 , . . . }.
Continuing by induction on m, we obtain a new countable row for each xmn contained in the sequence
{x11 , x12 , x21 , x13 , x22 , x31 . . . }. By Sα , we denote the submodule generated by Sα−1 and all the entries
xmn of the infinite matrix obtained above.
It is directly verified that the obtained chain of the submodules Sα satisfies conditions (i)–(vi). For
each α, the module Pα is a direct summand of M . Therefore, Pα is a direct summand of Pα+1 . Since

Sα+1 /Sα ∼
= Pα+1 /Pα ⊕ Qα+1 /Qα ,

the module Pα+1 /Pα is countably generated. If α is a limit ordinal, then Pα = β<α Pβ . Therefore, P is
isomorphic to the direct sum of the countably generated modules Pα+1 /Pα .
(2) The proof follows from (1).
(3) The proof follows from (1) and the fact that every projective module is a direct summand of a
direct sum of free cyclic modules.

5.2. Lemma. Let M be a regular right module over a ring A. Then the following assertions hold.
(1) Every submodule of M is a regular module.
(2) Every finitely generated submodule of M is a direct summand of M .
(3) For every countable infinite set {Xi }∞
i=1 of
cyclic submodules of M
, there exists a countable infinite
∞
set {Yi }∞
i=1 of cyclic submodules of M such that i=1 X i = ⊕ ∞ Y and
i=1 i
n
i=1 Xi = ⊕i=1 Yi for every n.
n

(4) Every countably generated submodule of M is a direct sum of a countable set of cyclic modules.
(5) Every projective submodule of M is a direct sum of cyclic modules that are isomorphic to direct
summands of AA .
(6) If n is a positive integer and X is an n-generated submodule of M , then X is a direct sum of n
cyclic modules.

Proof. (1) The assertion is directly verified.

(2) We prove by induction on n that all n-generated submodules of M are direct summands of M .
For n = 1, the assertion follows from the definition of a regular module. Assume that n > 1 and the
assertion holds for all positive integers that are less than n. Let N = X + Y be an n-generated submodule
of M , where the modules X and Y have less than n generators.

By the induction

hypothesis, there exists
a direct decomposition

M = X ⊕ Z. We set T ≡ N Z. Then N = N (X ⊕ Z) = X ⊕ T . Since
T ∼= N/X ∼
= Y /(X Y ), the module T is cyclic.

There exists a direct decomposition M = T ⊕ U . Since
T ⊆ Z, we have Z = T ⊕ W , where W = Z U . Therefore,

M = X ⊕ Z = X ⊕ T ⊕ W = N ⊕ W.

(3) Using induction on n, we construct a countable infinite set {Yi }∞

i=1 of cyclic submodules of M
such that
 n 
n
Yi = ⊕ni=1 Yi = Xi
i=1 i=1

3884
for n = 1, 2, . . . . For n = 1, we set Y1 = X1 . Assume that n > 1 and we have constructed the required
cyclic modules Y1 , . . . , Yn−1 . By (2), there exist direct decompositions
 n 

M = (⊕ni=1 Yi ) ⊕ Zn−1 = Xi ⊕ Zn .
i=1

Since

n−1 
n
⊕n−1
i=1 Yi = Xi ⊆ Xi
i=1 i=1
n
i=1 Yi ) ⊕ Zn−1 , there exists a direct decomposition
and M = (⊕n−1 i=1 Xi = (⊕i=1 Yi ) ⊕ Yn . In addition,
n−1

 n   n  n−1    
  n−1     n−1
∼
Yn = Xi / ⊕ Yi = Xi / ∼
Xi = Xn / Xn Xi .
i=1
i=1 i=1 i=1 i=1

Therefore, Yn is the required cyclic module.

(4) The proof follows from (3).
(5) By Proposition 5.1, every projective module is a direct sum of countably generated modules.
Therefore, the assertion follows from (4) and the fact that every projective cyclic right A-module is
 summand of AA .
isomorphic to a direct
(6) Let X = ni=1 Xi , where all the modules Xi are cyclic. For every i > n, we set Xi = 0. By (3),
there exists a set {Yi }∞
i=1 of cyclic submodules of M such that X = ⊕i=1 Yi .
n

5.3. Proposition. For a ring A, the following conditions are equivalent.
(1) Every right ideal of A is an idempotent multiplication right ideal.
(2) Every left ideal of A is an idempotent multiplication left ideal.
(3) A is a strongly regular ring.
(4) a ∈ Aa2 for every element a of A.
(5) A is a reduced regular ring.
(6) A is a normal regular ring.
(7) A is an invariant regular ring.
Proof. It is sufficient to prove the equivalence of conditions (1), (3), (5), (6), and (7).
(1)=⇒(3) By Proposition 3.1, A is a right invariant ring. If a is an element of A, then a ∈ aA =
aAaA = a2 A.
(3)=⇒(5) If a is an element of the strongly regular ring A such that a2 = 0, then a ∈ a2 A = 0.
(5)=⇒(6) Let e be an idempotent of A. Then 0 = (eA(1 − e))2 = (1 − e)Ae)2 . Since A is a reduced
ring, eA(1 − e) = (1 − e)Ae = 0. Therefore, e is a central idempotent.
(6)=⇒(7) Since A is a normal regular ring, every principal right or left ideal of A is generated by a
central idempotent. Therefore, every principal right or left ideal of A is an ideal of A. Therefore, A is an
invariant ring.
(7)=⇒(1) Since A is an invariant regular ring, every right ideal of A is generated by central idempo-
tents. By Proposition 3.4, every right ideal B of A is a multiplication right ideal. Since B is generated
by idempotents, B 2 = B.

5.4. Proposition. Let A be a ring, M be a right A-module such that every 2-generated submodule of M
is a distributive regular module, and R = End(M ). Then the following assertions hold.
(1) M is a distributive regular Bezout module and every countably generated submodule of M is a
direct sum of a countable set of cyclic modules.
(2) If X is a submodule of M , then f (X) ⊆ X for every homomorphism f : X → M ; in particular,
M is an invariant module and X coincides with every submodule Y of M such that X ∼ =Y.
2
(3) Ker(f ) = Ker(f ) for every endomorphism f of M .

3885


(4) For every endomorphism f of M and each submodule X of M with X Ker(f ) = 0, the endo-
morphism f induces the automorphism f |X of X.
(5) M = f (M ) ⊕ Ker(f ) for every endomorphism f of M .
(6) R is a strongly regular ring.
(7) If A is a right invariant ring, then M is a semihereditary, faithful, distributive, regular Bezout
multiplication left module over the strongly regular ring R, the lattice Lat(R M ) coincides with the lattice
Lat(MA ), and for every finitely generated submodule X of the left R-module M , there exists a central
idempotent e of R such that R X is isomorphic to the projective principal left ideal Re of the ring R.
Proof. (1) By Lemma 2.17, M is a distributive module.
Let X be a cyclic submodule of M and let E be the set of all 2-generated submodules of M containing
X. For every E ∈ E, there exists a submodule Y (E) of E such that E = X ⊕ Y (E). Since M is a
distributive module,    
X ( Y (E)) = (X Y (E)) = 0.
E∈E E∈E
Let m be an element of M . Since X + mA ∈ E, we have

m ∈ X + mA = X ⊕ Y (X + mA) ⊆ X ⊕ Y (E).
E∈E

Therefore, M = X ⊕ E∈E Y (E) and M is a regular module.
Let N be a finitely generated submodule of M . Since M is a regular module, it follows from Lemma
5.2(6) that N is a finite direct sum of cyclic modules. By Lemma 2.18(2), N is a cyclic module. Therefore,
M is a Bezout module.
By Lemma 5.2(4), every countably generated submodule of M is a direct sum of a countable set of
cyclic modules.
(2) Let x be an element of X. By (1), there exists a direct decomposition M = xA ⊕ Y . By Lemma
2.18(4), f (xA) ⊆ xA. Therefore, f (X) ⊆ X.
(3) Let X be a cyclic submodule of Ker(f 2 ). By (1), the cyclic module f (X) is contained in the
regular module X. Therefore, there exists a direct decomposition X = f (X) ⊕ Y . By (2), f (Y ) ⊆ Y .
Thus,
f (X) = f (f (X) ⊕ Y ) ⊆ f 2 (X) + f (Y ) ⊆ 0 + Y = Y.


Therefore, f (X) = Y f (X) = 0, f (Ker(f 2 )) = 0, and Ker(f 2 ) ⊆
Ker(f ) ⊆ Ker(f 2 ).
(4) By (2), f induces the endomorphism f |X of X. Since X Ker(f ) = 0, the endomorphism f |X
of X is a monomorphism. Let x be an element of X. It follows from (1) and (2) that xA = f (xA) ⊕ Y
for some submodule Y of xA. By (2), f (Y ) ⊆ Y . Therefore,
 
f (Y ) ⊆ Y f (xA) = 0, Y ⊆ X Ker(f ) = 0,
xA = f (xA) ⊕ Y = f (xA), X = f (X),
and f |X is an epimorphism.

(5) Let X be a cyclic submodule of M such that f (X) ⊆ f (M ) Ker(f ). It follows from (1) and (2)
that X = f (X) ⊕ Y for some submodule Y of X. By (2), f (Y ) ⊆ Y . Therefore,

f (Y ) ⊆ Y f (X) = 0 and f (X) = f (f (X) ⊕ Y ) ⊆ f 2(X) + f (Y ) = 0.


Therefore, f (M ) Ker(f ) = 0.
Let m be an element of M . It follows from (1) and (2) that mA = f (mA) ⊕ Z for some submodule
Z of X. By (2), f (Z) ⊆ Z. Therefore,

f (Z) ⊆ Z f (mA) = 0, Z ⊆ Ker(f ), and mA = f (mA) ⊕ Z ⊆ f (M ) ⊕ Ker(f ).
Therefore, M = f (M ) ⊕ Ker(f ).

3886
 (6) Let f be an endomorphism of M . By (5) and (4), M = f (M ) ⊕ Ker(f ) and f induces the
automorphism α of f (M ). Since M = f (M ) ⊕ Ker(f ), there exists an automorphism β of M such that
β(x + y) = α(x) + y for all elements x ∈ f (M ) and y ∈ Ker(f ). Let π be the projection of M on f (M )
with kernel Ker(f ). We denote by g the endomorphism πβ −1 of M . For any element m of M , we have
gf 2 (m) = g(α(f (m))) = gβ(f (m)) = πβ −1 β(f (m)) = π(f (m)) = f (m).
Therefore, f = gf 2 . By Proposition 5.3, A is a strongly regular ring.
(7) By (6), R is a strongly regular ring. It is obvious that M is a faithful left R-module. Let m be an
element of M and let a be an element of A. Since A is a right invariant ring, r(m) ⊆ r(ma). Therefore,
there exists an epimorphism f : mA → maA such that f (m) = ma. Since mA is a direct summand of
the regular A-module M , there exists an endomorphism f of M such that f |mA = f . Then f (m) = ma.
Therefore, mA ⊆ Rm. By (2), Rm ⊆ mA. Therefore, mA = Rm, the lattice Lat(R M ) coincides with
the lattice Lat(MA ), and the set of cyclic submodules of MA (resp. the set of direct summands of MA )
coincides with the set of cyclic submodules of R M (resp. the set of direct summands of R M ). Therefore,
R M is a distributive regular Bezout module.
We prove that M is a multiplication left R-module. By the left-side version of Proposition 2.1, it is
sufficient to prove that for every cyclic submodule X of the left R-module M , there exists an ideal Q of
R such that X = QM . By the above argument, X is a cyclic submodule of the regular right A-module.
Therefore, there exists an idempotent e of the strongly regular ring R such that X = eM ; in addition,
X = RX by (2). Therefore, X = (Re)M and we can set Q = Re.
Let X be a finitely generated submodule of the left R-module M . Since R M is a Bezout module, X
is a cyclic left R-module. There exists an element x of X such that X = Rx = xA. Since xA is a direct
summand of MA , there exists a central idempotent e of the ring R such that e(M ) = X = e(x)A = Re(x).
It is sufficient to prove that Re(x) is a free left module over the ring Re. Let f be an element of the ring R
such that (f e)(e(x)) = (f e)(x) = 0. Then f e(M ) = (f e)(e(M )) = f e(X) = 0 and Re (Re(x)) ∼ = Re. Since
every finitely generated submodule of R M is isomorphic to a direct summand of R R, the left R-module
M is semihereditary.

5.5. Theorem. Let A be a ring, M be a regular multiplication right A-module, and R = End(M ). Then
the following assertions hold.
(1) M is a distributive regular invariant Bezout module, R is a strongly regular ring, every countably
generated submodule of M is a direct sum of a countable set of cyclic modules, and every submodule X of
M coincides with each submodule Y of M such that X ∼ =Y.
(2) If A is a right invariant ring, then M is a semihereditary right A/r(M )-module and for ev-
ery finitely generated submodule X of MA/r(M ) , there exists a central idempotent e of the ring A/r(M )
such that the right A/r(M )-module X is isomorphic to the projective direct summand eA/r(M ) of
A/r(M )A/r(M ) .
(3) If A is a right invariant ring, then M is a semihereditary, faithful, distributive, regular Bezout
multiplication left module over the strongly regular ring R, the lattice Lat(R M ) coincides with the lattice
Lat(MA ), and for every finitely generated submodule X of the left R-module M , there exists a central
idempotent f of R such that R X is isomorphic to the projective principal left ideal Rf of the ring R.
Proof. Let X be a 2-generated submodule of M . By Lemma 5.2(2), X is a direct summand of M . By
Note 1.7, the direct summand X of the multiplication module M is a multiplication module. Since all
2-generated submodules of M are multiplication modules, it follows from Proposition 2.25 that M is a
distributive module. By Lemma 5.2(1), every submodule of M is a regular module. Now (1) and (3)
follow from Proposition 5.4.
It remains to prove (2). Let X be a finitely generated submodule of MA/r(M ) . By (1), M is a Bezout
A/r(M )-module. Therefore, X is a cyclic A/r(M )-module. By Proposition 3.23(2), there exists a central
idempotent e of the ring A/r(M ) such that the right A/r(M )-module X is isomorphic to the projective
direct summand eA/r(M ) of A/r(M )A/r(M ) . Therefore, M is a semihereditary right A/r(M )-module.

3887
5.6. Lemma. Let A be a regular ring, M be a right A-module, and N be a submodule of M . Then the
following assertions

hold.
(1) N M B = N B for every left ideal B of A.
(2) For any ideal B of A, there exists a monomorphism f : (N/N B)A → (M/M B)A .



Proof. (1) Since N B ⊆ N M B, it is sufficient to prove that any element x ofN M B is contained in
mn ∈ M and b1 , . . . , bn ∈ B such that x = ni=1 mi bi . We denote by
N B. There exist elements m1 , . . . ,
C the finitely generated left ideal ni=1 Abi of A. Since A is a regular ring, there exists an idempotent
e ∈ C ⊆ B such that C = Ae. Therefore, there exist elements a1 , . . . , an ∈ A such that bi = ai e
set m = ni=1 mi ai . Then x = me ∈ N and x = xe
∈ N C ⊆ N B.
(i = 1, . . . , n). We


(2) By (1), N M B = N B. Since N/N B coincides with N/(N M B), the module N/(N M B)
is naturally isomorphic to the submodule (N + M B)/M B of M/M B.

5.7. Proposition. For a right module M over a strongly regular ring A, the following conditions are
equivalent.
(1) M is a distributive module.
(2) For any two elements x and y of M , there exists an element a of A such that xa ∈ yA and
y(1 − a) ∈ xA.
(3) For any two elements x and y of M , there exists a central idempotent e of A such that
xe ∈ yeA, y(1 − e) ∈ x(1 − e)A, and xA + yA = (x(1 − e) + ye)A.
(4) M is a Bezout module.
(5) M is a regular module and End(M ) is a normal ring.
(6) For any maximal ideal B of A, the dimension of the vector space M/M B over the division ring
A/B does not exceed 1.
Proof. The equivalence of conditions (1) and (2) follows from Lemma 2.17.

(2)=⇒(3) Let x, y ∈ M . By assumption, there exists an element a ∈ A with xa, y(1 − a) ∈ xA yA.
Since A is a strongly regular ring, there exists a central idempotent e of A such that aA = eA. Then
xe ∈ xaAe ⊆ yeA and there exists an element b of A such that a = eb. Since 1 − e is a central idempotent
of A, we have
1 − e = (1 − e)(1 − eb) = (1 − e)(1 − a) = (1 − a)(1 − e) ∈ (1 − a)A.
Therefore, y(1 − e) ∈ y(1 − a)A(1 − e) ⊆ x(1 − e)A. Thus,
x(1 − e) = (x(1 − e) + ye)(1 − e) ∈ (x(1 − e) + ye)A, ye = (x(1 − e) + ye)e ∈ (x(1 − e) + ye)A,
xe ∈ yeA ⊆ (x(1 − e) + ye)A, and y(1 − e) ∈ x(1 − e)A ⊆ (x(1 − e) + ye)A.
Therefore,
xA + yA = xeA + x(1 − e)A + yeA + y(1 − e)A ⊆ (x(1 − e) + ye)A ⊆ xA + yA.
(3)=⇒(4) It follows from (3) that every 2-generated submodule of M is a cyclic module. Therefore,
M is a Bezout module.
(4)=⇒(5) Let X be a 2-generated submodule of M . Since M is a Bezout module, X is a cyclic
module over the strongly regular ring A. By Proposition 5.3, A is an invariant regular ring. Therefore,
the lattice of all submodules of the cyclic A-module X can be identified with the lattice Lat(A/B) of all
ideals of some factor ring A/B of the invariant regular ring A. Then A/B is an invariant regular ring
and every finitely generated right or left ideal of A is generated by a central idempotent. Therefore, X is
a distributive regular module. Since every 2-generated submodule of M is a distributive regular module,
it follows from Proposition 5.4(1) and Proposition 5.4(6) that M is a regular module and End(M ) is a
normal ring.

3888
 (5)=⇒(1) Let x and y be two elements of the regular module M . There exist idempotents e, f , and
g of the ring End(M ) such that
xA = e(M ), yA = f (M ), and (x + y)A = g(M ).
Since End(M ) is a normal ring, the idempotents e, f , and g are central. Therefore,
  
(x + y)A = g(M ) (e(M ) + f (M )) = ge(M ) + gf (M ) ⊆ xA (x + y)A + yA (x + y)A ⊆ (x + y)A).

By Lemma 2.17, M is a distributive module.

(1)=⇒(6) Assume the contrary. Then M has a distributive factor module M/N that is a direct sum
of two isomorphic nonzero modules. This contradicts Lemma 2.18(3).
(6)=⇒(1) Let N be a 2-generated submodule of M and let X be a simple right A-module. By
Lemma 2.17, it is sufficient to prove that the module X ⊕ X is not a homomorphic image of N . Assume
the contrary. Since A is an invariant ring and X is a simple module, there exists a maximal ideal B
of A such that (X ⊕ X)B = 0. Therefore, dim(N/N B) > 1, where dim(N/N B) is the dimension of
the vector space N/N B over the division ring A/B. By Lemma 5.6(2), there exists a monomorphism
f : (N/N B)A → (M/M B)A . Therefore, dim(M/M B) > 1; this is a contradiction.

5.8. Theorem. Let M be a multiplication right module over a strongly regular ring A. Then the following
assertions hold.
(1) M is a distributive regular invariant Bezout module, every submodule of M is a multiplication
module, every submodule X of M coincides with each submodule Y of M such that X ∼ = Y , and every
countably generated submodule of M is a direct sum of a countable set of cyclic modules.
(2) M is a semihereditary right A/r(M )-module, and for every finitely generated submodule X of
MA/r(M ) , there exists a central idempotent e of the ring A/r(M ) such that the right A/r(M )-module X
is isomorphic to the projective direct summand eA/r(M ) of A/r(M )A/r(M ) .
(3) End(M ) is a strongly regular ring, M is a semihereditary faithful, distributive, regular Bezout
multiplication left module over the strongly regular ring End(M ), and the lattice Lat(End(M ) M ) coincides
with the lattice Lat(MA ).
Proof. Let N be a submodule of M . There exists an ideal B of the invariant regular ring A such that
N = M B. There exists a set {ei }i∈I of central idempotents ei of A such that B = i∈I ei A. Let W
be a submodule of N . Since W ⊆ N = M B = i∈I M ei , we have W = W B. There exists an ideal C
of A such that M C = W . Since the ideal B is generated by central idempotents, we have BC = CB.
Therefore, N C = M BC = M CB = W B = W and N is a multiplication module. Since every submodule
of M is a multiplication module, M is a distributive module by Proposition 2.25(2). By Proposition 5.7,
M is a regular Bezout module. By Proposition 5.4(1), every countably generated submodule of M is a
direct sum of a countable set of cyclic modules. By Proposition 5.3, A is an invariant ring. In addition,
M is a multiplication regular A-module. Therefore, the remaining assertions follow from Theorem 5.5.

5.9. Proposition. Let M be a faithful multiplication right module over a ring A. Assume that either A
is a strongly regular ring or M is a regular module and A is a right invariant ring. Then the following
conditions are equivalent.
(1) M is a direct sum of countably generated modules.
(2) M = ⊕i∈I Mi and for every i ∈ I, the module Mi is isomorphic to a direct summand of the free
cyclic module AA .
(3) M is a projective module.
Proof. The implication (1)=⇒(2) follows from Theorem 5.5 and Theorem 5.8.
The implication (2)=⇒(3) follows from the fact that every direct sum of projective modules is a
projective module.
The implication (3)=⇒(1) follows from Proposition 5.1.

3889
 For a submodule X of a module M , we denote by cl(X) and op(X) the set of all maximal submodules
of M that contain X and the set of all maximal submodules of M that do not contain X, respectively.
Note 5.10 below is directly verified.
5.10. Note. Let M be a module. Then the following assertions hold.
(1) For any subset {Xi }i∈I of Lat(M ), we have
   
cl(Xi ) = cl( Xi ) and op(Xi ) = op( Xi ).
i∈I i∈I i∈I i∈I

(2) cl(0) = op(M ) = max(M ), cl(M ) = op(0) = ∅, and op(X) = max(M ) \ cl(X).
A module M is said to be spectral if the set max(M ) of all its maximal submodules is a topological
space such that the set of all closed sets of this space coincides with the set {cl(X) | X ∈ Lat(M )} and
 
cl(X) cl(Y ) = cl(X Y ) for all X, Y ∈ Lat(M ).
We denote the above topological space by max(M ). The set {cl(X) | X ∈ Lat(M )} of all closed subsets
in max(M ) is denoted by CL(M ). We denote the set {op(X) | X ∈ Lat(M )} by OP(M ); this set is the
set of all open subsets in max(M ).
5.11. Note. For a module M , the following conditions are equivalent.
(1) M is aspectral module.


(2) cl(X) cl(Y
) = cl(X Y ) for any two submodules X and Y of M .
(3) M = N + X Y for every maximal submodule N of M and any submodules X and Y of M such
that M = N + X = N + Y .
Proof. The implication (1)=⇒(2) follows
from the definition

of a spectral module.
(2)=⇒(3) Assume that
M = N + X Y . Then X Y ⊆ N , since N is a maximal submodule of
M . Therefore, N ∈ cl(X Y ) = cl(X) cl(Y ). Therefore, either X ⊆ N or Y ⊆ N . Therefore, either
N + X = M or N + Y = M ; this is a contradiction. 
(3)=⇒(2) Let
X and Y be two submodules  of M . If N ∈
cl(X) cl(Y ), then either X ⊆ N or

Y ⊆ N;
in both cases, X

Y ⊆ N . Therefore, cl(X)
cl(Y ) ⊆ cl(X Y ). Now assume that N ∈ cl(X Y ); this
means that X Y ⊆ N . Therefore, N + X Y = M . It follows from (3) that either M = N + X or
M = N + Y . Therefore,  either X ⊆ N or Y ⊆ N
, since N is 
a maximal submodule of M . This implies
the relation N ∈ cl(X) cl(Y ). Therefore, cl(X Y ) ⊆ cl(X) cl(Y ).
(2)=⇒(1) It follows from (2) that
 
cl(X) cl(Y ) = cl(X Y ) ∈ CL(M ) for all X, Y ∈ Lat(M ).
It follows from Note 5.10(1) that
 
cl(Xi ) = cl( Xi ) ∈ CL(M ).
i∈I i∈I

Therefore, the set CL(M ) satisfies the axioms of the set of all closed sets of a topological space. Therefore,
max(M ) is a topological space.

5.12. Proposition. Every multiplication module is a spectral module.
Proof. Let A be a ring, M be a multiplication right A-module, N be a maximal submodule of M ,
M = M/N , and X and Y be two submodules

of M such that M = N + X = N + Y . By Note 5.11, it
is sufficient to prove that M = N + X Y . There exist ideals B and C of A such that X = M B and
Y = M C. Then
 
M BC ⊆ M B M C = X Y, M = N + X = N + Y = N + M B = N + M C,

M = M B, and M = M C.

3890
Therefore, 
M = M BC and M = N + M BC ⊆ N + X Y.


Thus M = N + X Y.

5.13. Lemma. For a spectral right module M over a ring A, the following assertions hold.
(1) Every homomorphic image of M is a spectral module.
(2) Every direct summand of M is a spectral module.
(3) For any two isomorphic simple right A-modules X and Y , the module X ⊕Y is not a homomorphic
image of M .

(4) If {Xi }i∈I is a set of submodules of M such that i∈I cl(Xi ) = ∅, then the factor module M/X
does not have maximal submodules.
(5) If M is a finitely generated module, then the topological space max(M ) is compact.
(6) If cl(X) = cl(Y ) for any two distinct submodules X and Y of M , then M is a distributive module
and the lattice Lat(M ) is isomorphic to the distributive lattice of all open subsets of max(M ).
(7) If B is an ideal of A such that the factor ring A/B is a division ring, then the dimension of the
vector space M/M B over the division ring A/B does not exceed 1.
Proof. (1) It is sufficient to prove that every factor


module M/K is a spectral module. Let N , X, and
Y be three submodules of M such that K ⊆ N X Y , N/K is a maximal submodule of M/K,
M/K = N/K + X/K = N/K + Y /K, and M = N + X = N + Y.



Since M is a spectral module, M = N + X Y by Note 5.11. Then M/K = N/K + (X/K) (Y /K). By
Note 5.11, M/K is a spectral module.
(2) The proof follows from (1) and the fact that every direct summand of M is a homomorphic image
of M .
(3) Assume that there exist two isomorphic simple right A-modules X and Y such that the module
X ⊕ Y is a homomorphic image of M . Let X = xA, f : X → Y be an isomorphism, and N = (x + f (x))A.
Then Y = f (x)A, N is a simple module,
 
N X=N Y = 0, X ⊕ Y = N ⊕ X = N ⊕ Y,

M . By (1), the homomorphic image X ⊕ Y of M is a spectral module.

and N is a maximal submodule of
By Note 5.11, X ⊕ Y = N ⊕ (X Y ) = N . This is a contradiction, since N = X ⊕ Y = N ⊕ Y .
(4) We denote by X the submodule i∈I Xi of M . By Note 5.10(1),
 
cl(X) = cl( Xi ) = cl(Xi ) = ∅.
i∈I i∈I
This means that the module M/X does not have maximal submodules.

(5) Let {Ci }i∈I be a set of closed subsets of the space max(M ) such that i∈I Ci = ∅. For
 every i ∈ I,
there exists a submodule Xi of M such that Ci = cl(Xi ). We denote by X the submodule i∈I Xi of M .
By (4), the finitely generated module M/X does nothave a maximal submodule. Therefore, M/X = 0
and the finitely generated
 module M coincides with i∈I Xi . Therefore, there exists a finite subset J of
I such that M = j∈J Xj . Then
  
Cj = cl(Xj ) = cl( Xj ) = cl(M ) = ∅.
j∈J j∈J j∈J

Therefore, the space max(M ) is compact.

(6) It is directly verified that the mapping ϕ defined by ϕ(X) = op(X) is a surjective homomorphism
from the lattice Lat(M ) onto the distributive lattice of all open subsets of max(M ). If X and Y are
two distinct submodules of M , we have cl(X) = cl(Y ) by assumption; therefore op(X) = op(Y ) and
ϕ(X) = ϕ(Y ).
(7) The proof follows from (3).

3891
5.14. Proposition. Let M be a distributive module. Then M is a spectral module. Further, if there
exist nonzero submodules X and Y of M such that M = X ⊕ Y ⊕ Z for some Z ∈ Lat(M ), then
op(X) = cl(Y ⊕ Z), op(Y ) = cl(X ⊕ Z),
op(X) and op(Y ) are open-closed disjoint subspaces in max(M ), X and Y are spectral modules, and the
spaces max(X) and max(Y ) are homeomorphic to op(X) and op(Y ), respectively.
Proof. Let N be a maximal submodule of M
, and let X and Y

be two submodules of M such that
M = N + X = N + Y . Then M = (N + X) (N + Y ) = N + X Y . By Note 5.11, M is a spectral
module.
Since M is a distributive module,
   
N =N X ⊕ N (Y ⊕ Z and M/N ∼ = X/(N X) ⊕ (Y ⊕ Z)/(N (Y ⊕ Z)).



Since M/N is a simple module, either X = N X ⊆N or Y ⊕ Z = N (Y ⊕ Z) ⊆ N . We have that
either N ∈ cl(X) or N ∈ cl(Y ⊕ Z). Thus, N ∈ cl(X) cl(Y ⊕ Z). Therefore,
 
max(M ) = cl(X) cl(Y ⊕ Z) and cl(X) cl(Y ⊕ Z) = cl(X ⊕ Y ⊕ Z) = cl(M ) = ∅.
Therefore, op(X) = cl(Y ⊕ Z) and op(X) is an open-closed subspace in max(M ). Similarly, op(Y ) =
cl(X ⊕ Z) and op(Y ) is an open-closed subspace in max(M ). Since X and Y are direct summands of
the spectral module M , it follows from Lemma 5.13(2) that X and Y are spectral modules. Since M is
a distributive module, it is directly verified that the spaces max(X) and max(Y ) can be identified with
op(X) and op(Y ), respectively.

Assume that N ∈ op(X) op(Y ). Then
  
N ∈ cl(Y ⊕ Z) cl(X ⊕ Z) = cl((Y ⊕ Z) (X ⊕ Z)) = cl((Y X) ⊕ Z) = cl(Z).


Therefore, Z ⊆ N and the
simple module M/N
is isomorphic to the module X/(N
 X) ⊕ Y /(N ⊕ Y ).
Therefore, either X = N
X ⊆ N or Y = N Y ⊆
N . Therefore, N ∈ cl(X) cl(Y ); this contradicts
the relation N ∈ op(X) op(Y ). Therefore, op(X) op(Y ) = ∅.

5.15. Lemma. For a right module M over a ring A, the following conditions are equivalent.
(1) Every subfactor of M is a spectral module.
(2) Every 2-generated submodule of M is a spectral module.
(3) M is a distributive module.
Proof. The implication (1)=⇒(2) is obvious.
(2)=⇒(3) It follows from Lemma 5.13(3) that for any two isomorphic simple right A-modules X and
Y , the module X ⊕ Y is not a homomorphic image of any 2-generated submodule of M . By Lemma 2.17,
M is a distributive module.
The implication (3)=⇒(1) follows from Proposition 5.14 and the fact that every subfactor of a
distributive module is a distributive module.

5.16. Proposition. Let M be a right module over a ring A. Assume that either M is a regular module
or A is a strongly regular ring. Then the following conditions are equivalent.
(1) M is a spectral module.
(2) M is a distributive module.
(3) M is a spectral regular module, the lattice Lat(M ) is isomorphic to the distributive lattice of
all open subsets of the space max(M ), and for any two distinct points P, Q ∈ max(M ), there exist two
submodules X and Y of M such that op(X) and op(Y ) are open-closed compact disjoint neighborhoods of
points P ∈ op(X) and Q ∈ op(Y ) in the space max(M ).
Proof. (1)=⇒(2) Let X be a 2-generated submodule of M . By Lemma 5.15, it is sufficient to prove that
X is a spectral module.

3892
 Assume that M is a regular module. Then X is a direct summand of M by Lemma 5.2(2). By
Lemma 5.13(2), X is a spectral module.
Assume that A is a strongly regular ring. Let B be a maximal ideal of A. Then the invariant simple
ring A/B is a division ring. By Lemma 5.13(7), the dimension of the vector space M/M B over the
division ring A/B does not exceed 1. By Proposition 5.7, M is a distributive module.
(2)=⇒(3) By Lemma 5.15, M is a spectral module. By Proposition 5.7, we can assume that M is a
regular module. Since P and Q are two distinct maximal submodules of M , there exist elements x ∈ Q\P
and p ∈ P \ Q. Let W = xA + pA and let X = xA. Since M is a regular module, there exist direct
decompositions M = W ⊕ Z and W = X ⊕ Y . Therefore, M = X ⊕ Y ⊕ Z and X ⊆ P . Since W ⊆ Q and
X ⊆ Q, we have Y ⊆ Q. By Lemma 5.13(2) and Lemma 5.13(5), X and Y are spectral modules and the
spaces max(X) and max(Y ) are compact. By Proposition 5.14, op(X) and op(Y ) are open-closed disjoint
neighborhoods of the points P and Q of the space max(M ) and these neighborhoods are homeomorphic
to the spaces max(X) and max(Y ), respectively.
It remains to prove that the lattice Lat(M ) is isomorphic to the distributive lattice of all open subsets
of the space max(M ). By Lemma 5.13(6), it is sufficient to prove that cl(W1 ) = cl(W2 ) for any two distinct
submodules W1 and W2 of M . Assume that cl(W1 ) = cl(W2 ). Then op(W1 ) = op(W2 ). Let

T = W1 W2 , M = M/T, W 1 = W1 /T, and let W 2 = W2 /T.

Since W1 = W2 , at least one of the modules Wi is not equal to zero. For example, assume that W1 = 0.
Let L = L/T be a nonzero cyclic submodule of W1 , where T ⊂ L ⊆ W1 . By assumption, M is a
regular module. Therefore, there exists a direct decomposition M = L ⊕ Q. Therefore, there exists
a maximal submodule
Z/T of M that does not contain L. Then Z ∈ op(L) and Z ⊇ T . Therefore,
Z ∈ op(W1 ) = op(W1 ) op(W2 ) = op(T ); this is a contradiction.
The implication (3)=⇒(1) is obvious.

5.17. Theorem. Let M be a multiplication right module over a ring A. Assume that either M is
a regular module or A is a strongly regular ring. Then M is a spectral module, the lattice Lat(M ) is
isomorphic to the distributive lattice of all open subsets of the space max(M ), and for any two distinct
points P, Q ∈ max(M ), there exist two submodules X and Y of M such that op(X) and op(Y ) are
open-closed compact disjoint neighborhoods of points P ∈ op(X) and Q ∈ op(Y ) in the space max(M ).
Proof. It follows from Theorem 5.5(1) and Theorem 5.8(1) that M is a distributive regular module.
Therefore, the assertion follows from Proposition 5.16.

5.18. Theorem. If M is a multiplication right module over a regular ring A, then every submodule of
M is a distributive spectral multiplication module.


Proof. Let N be a submodule of M . By Lemma 5.6, N M B = N B for every ideal B of A. By Note
1.5(2), N is a multiplication module. Since every submodule of M is a multiplication module, M is a
distributive module by Proposition 2.25(2). Then N is a distributive module. By Proposition 5.12, N is
a spectral module.

6. Rings Whose Right Ideals Are Multiplication Modules

6.1. Proposition. For a right module M over a right invariant ring A, the following conditions are
equivalent.
(1) All finitely generated submodules of M are multiplication modules.
(2) All 2-generated submodules of M are multiplication modules.
(3) M is a distributive module.
Proof. The implication (1)=⇒(2) is obvious.
The implication (2)=⇒(3) follows from Proposition 2.25(2).

3893
 The implication (3)=⇒(1) follows from Theorem 3.8 and the fact that all submodules of distributive
modules are distributive modules.

6.2. Note. For a ring A, the following conditions are equivalent.

(1) Every ideal of A is a multiplication right ideal.
(2) Every right ideal of A is a multiplication right ideal.
(3) A is a right invariant, right distributive ring and every right ideal of A is a multiplication right
ideal.

Proof. (1)=⇒(3) By assumption, A is a multiplication right ideal. By Proposition 3.1, A is a right

invariant ring. If B is a right ideal of A, then B is an ideal and B is a multiplication right ideal by
assumption. By Proposition 6.1, A is a right distributive ring.
The implications (3)=⇒(2) and (2)=⇒(1) are obvious.

6.3. Theorem. For a ring A, the following conditions are equivalent.

(1) Every finitely generated ideal of A is a multiplication right ideal.
(2) Every finitely generated right ideal of A is a multiplication right ideal.
(3) Every 2-generated ideal of A is a multiplication right ideal.
(4) Every 2-generated right ideal of A is a multiplication right ideal.
(5) A is a right distributive, right invariant ring.

Proof. (1)=⇒(2) Let B be a finitely generated right ideal of A. By assumption, A is a multiplication

right ideal. By Proposition 3.1, A is a right invariant ring. Therefore, B is an ideal. By assumption, B
is a multiplication right ideal.
(2)=⇒(3) Let B be a 2-generated ideal of A. By assumption, A is a multiplication right ideal. By
Proposition 3.1, A is a right invariant ring. Therefore, B is a 2-generated right ideal. By assumption, B
is a multiplication right ideal.
(3)=⇒(4) Let B be a 2-generated right ideal of A. By assumption, A is a multiplication right
ideal. By Proposition 3.1, A is a right invariant ring. Therefore, B is an ideal. By assumption, B is a
multiplication right ideal.
(4)=⇒(5) By Proposition 3.1, A is a right invariant ring. By Proposition 2.25(2), A is a right
distributive ring.
The implication (5)=⇒(2) follows from Proposition 6.1.
(2)=⇒(1) Let B be a finitely generated ideal of A. By assumption, A is a multiplication right ideal.
By Proposition 3.1, A is a right-invariant ring. Therefore, B is a finitely generated right ideal. By
assumption, B is a multiplication right ideal.

6.4. Lemma
([83]). For a ring A and a distributive right A-module M , the following assertions hold.
(1) f (X) f −1 (X) ⊆ X for any submodule X of M and each endomorphism f of M .
(2) If X is a submodule of M and X + f (X) = M for some endomorphism f of M , then X = M .
(3) All maximal submodules of M are fully invariant submodules of M .
(4) If N is a submodule of M such that M/N is a semi-Noetherian module, then N is a fully invariant
submodule of M .
(5) If A is a local ring, then M is a uniserial module.


Proof. (1) Let y be an element of the module f (X) f −1 (X). There exists an element x of X such
that f (x) = y and f 2 (x) = f (y) ∈ X. By Lemma 2.17, there exists an element a of the ring A such that
xa ∈ yA and y(1 − a) ∈ xA ⊆ X. Then

y = ya + y(1 − a) ∈ f (xa)A + X ⊆ f (yA) + X = X.

Therefore, f (X) f −1(X) ⊆ X.

3894
 (2) Let m be an element of M . Since X + f (X) = M , there exist elements x, y ∈ X such that
f (m) = f (x) + y. Therefore,
f (m − x) = y ∈ X, m − x ∈ f −1(X),and m ∈ X + f −1(X).


Therefore, X + f −1(X) = M = X + f (X); in addition, f (X) f −1 (X) ⊆ X by (1). Then
 
M = (X + f (X)) (X + f −1 (X)) = X + f (X) f −1 (X) = X.
(3) Let X be a maximal submodule of M and let f be an endomorphism of M . Since X = M , it
follows from (2) that X + f (X) = M . Therefore, f (X) ⊆ X, since X is a maximal submodule of M .

(4) Let {Xi }i∈I be the set of all fully invariant submodules of M that contain N and let X =
i∈I Xi ⊆ M (the set {Xi }i∈I is not empty, since it contains the module M ). Then X is a fully invariant
submodule of M containing N . It is sufficient to prove that X/N = 0.
Assume that X/N = 0. By assumption, M/N is a semi-Noetherian module. Therefore, the nonzero
module X/N has a maximal submodule Y /N . Since Y is a maximal submodule of the distributive module
X, it follows from (3) that Y is a fully invariant submodule of X. In addition, X is a fully invariant
submodule of M . Therefore, Y is a fully invariant submodule of M . This implies Y ∈ {Xi }i∈I . Therefore,
X ⊆ Y ; we have a contradiction, since Y
X.
(5) Let x and y be two elements of M such that x ∈ / Y . By Lemma 2.17, there exists an element a
of A such that xa ∈ yA and y(1 − a) ∈ xA. Since x ∈ / Y and xa ∈ yA, the element a is not invertible.
Then 1 − a is an invertible element, since A is a local ring. Therefore, y = y(1 − a)(1 − a)−1 ∈ xA and M
is a uniserial module.

6.5. Lemma ([83]). For a right distributive ring A, the following assertions hold.
(1) If A is a right Noetherian ring, then A is a right invariant ring.
(2) If A is a right Artinian ring, then A is a finite direct product of right invariant, right uniserial,
right Artinian rings.
(3) If X and Y are two completely prime ideals of A that are not comparable with respect to inclusion,
then X + Y = A.
Proof. (1) The proof follows from Lemma 6.4(4).
(2) By (1), A is a right invariant ring. Without loss of generality, we can assume that A is an
indecomposable, right invariant, right Artinian ring. Then A is a local ring. The local right distributive
ring A is a right uniserial ring by Lemma 6.4(5).
(3) There exist elements x ∈ X \ Y and y ∈ Y \ X. By Lemma 2.17, there exists an element a of A
such that xa ∈ yA ⊆ Y and y(1 − a) ∈ xA ⊆ X. Since X and Y are two completely prime ideals, a ∈ Y
and 1 − a ∈ X. Therefore, X + Y = A.

6.6. Lemma. For a right distributive, right Noetherian ring A, the following assertions hold.
(1) If A is a semiprime ring, then A is a finite direct product of right Noetherian, right distributive
domains.
(2) If A is an indecomposable ring and A is not a domain, then A has a nonzero completely prime
nilpotent ideal X such that X = yX for every element y ∈ A \ X.
(3) Assume that the ring A is an indecomposable ring, A is not a domain, and A is not a right
Artinian right uniserial ring. Then A has a nonzero completely prime nilpotent ideal X and a maximal
ideal Y such that X = Y X
Y , XY is a maximal submodule of X, and XY = Y X.
Proof. By Lemma 6.5(1), A is a right invariant ring.
(1) Without loss of generality, we can assume that A is an indecomposable
ring. Since A is a right
Noetherian semiprime ring, there exist prime ideals X1 , . . . , Xn of A such that ni=1 Xi = 0 and for i = j,
the ideals Xi and Xj are not comparable with respect to inclusion. Every prime ring A/Xi is a prime
right invariant ring. Therefore, every ring A/Xi is a domain and every ideal Xi is a completely prime
ideal. If n = 1, then A is a domain. Assume that n ≥ 2. Without loss of generality, we can assume that

3895





X2 · · · Xn =
0. By
6.5(3), A = X1 +Xi for i > 1. By Lemma 2.4(1), A = X1 +X2 · · · Xn .

Lemma
In addition, X1 X2 · · · Xn = 0 and A is an indecomposable ring. This is a contradiction.
(2) The right Noetherian ring A has a nilpotent ideal X such that A/X is a semiprime ring. Let
h : A → A/X be the natural ring epimorphism. It follows from (1) that A/X is a finite direct product of
domains A1 , . . . , An and X = 0.
Assume that n = 1. Then X is a completely prime ideal. Let y ∈ A \ X and let x ∈ X. Since X is a
completely prime ideal and y ∈ A \ X, we have (X : y) = X. By Lemma 2.17, there exists an element a of
A such that xa ∈ yA and y(1 − a) ∈ xA ⊆ X. Since X is a completely prime ideal, 1 − a ∈ X. Therefore,
1 − a is a nilpotent element, Then a is an invertible element, since a = 1 − (1 − a). Therefore,
x = xaa−1 ∈ yA, X ⊆ yA, and X = y(X : y) = yX.
Assume that n ≥ 2. Then the factor ring A/X has a nonzero idempotent e = h(1). Since X is a nilpotent
ideal, there exists an idempotent e of A such that e = h(e). Then e = 0 and e = 1. In addition, eA and
(1 − e)A are ideals of A. This is a contradiction, since A is an indecomposable ring.
(3) By (2), A has a nonzero, completely prime, nilpotent ideal X such that X = yX for every element
y ∈ A \ X.
Assume that the ideal X is not maximal. The nonzero module XA has a maximal submodule K.
We denote by Y the ideal r(X/K) of A. Since A is a right invariant ring and the right A-module X/K
is simple, Y is a maximal ideal. Then XY ⊆ K = X. Since the ideal X is not maximal, Y ⊆ X. There
exists an element y ∈ Y \ X. Then X = yX ⊆ Y X ⊆ X. Therefore, X = Y X.
Assume that the ideal X is a maximal ideal of the right invariant ring A. Then A is a local,
semiprimary, right Noetherian ring. Therefore, A is a right Artinian ring. By Lemma 6.5(2), A is a right
uniserial ring. This is a contradiction.

6.7. Theorem. For a ring A, the following conditions are equivalent.
(1) Every ideal of A is a finitely generated multiplication right ideal.
(2) Every right ideal of A is a finitely generated multiplication right ideal.
(3) A is a right Noetherian ring and every 2-generated right ideal of A is a multiplication right ideal.
(4) A is a right Noetherian ring and every 2-generated ideal of A is a multiplication right ideal.
(5) A is a right distributive, right Noetherian ring.
Proof. The equivalence of conditions (1) and (2) follows from Note 6.2.
The implication (2)=⇒(3) is obvious.
The implications (3)=⇒(4) and (4)=⇒(5) follow from Theorem 6.3.
(5)=⇒(2) Let B be a right ideal of A. Since A is a right Noetherian, right distributive ring, A is a
right invariant ring by Lemma 6.5(1). The right ideal B of the right Noetherian ring A is finitely generated.
The finitely generated right ideal B of the right invariant, right distributive ring A is a multiplication
right ideal by Theorem 6.3.

6.8. Theorem. For a ring A, the following conditions are equivalent.
(1) Every ideal M of A is a multiplication right ideal such that M = M B for every proper ideal B
of A with r(M ) ⊆ B.
(2) Every right ideal M of A is a multiplication right ideal such that M = M B for every proper ideal
B of A with r(M ) ⊆ B.
(3) Every ideal M of A is a multiplication right ideal such that M = M P for every maximal ideal P
of A with r(M ) ⊆ P .
(4) Every right ideal M of A is a multiplication right ideal such that M = M P for every maximal
ideal P of A with r(M ) ⊆ P .
(5) A is a right distributive, right Noetherian ring.
Proof. The equivalences (1)⇐⇒(2) and (3)⇐⇒(4) follow from Note 6.2.
The implication (2)=⇒(4) is obvious.

3896
 (4)=⇒(5) By Theorem 6.3, A is a right distributive, right invariant ring. Let M be a right ideal of A.
Since M is a multiplication module over the right invariant ring A and M = M P for every maximal ideal
P of A with r(M ) ⊆ P , it follows from Theorem 3.11 that M is a finitely generated module. Therefore,
A is a right Noetherian ring.
(5)=⇒(4) By Lemma 6.5(1), A is a right invariant ring. Let M be a right ideal of A. By Theorem
6.7, M is a finitely generated multiplication right ideal. Since M is a finitely generated multiplication
module over the right invariant ring A, it follows from Theorem 3.11 that M = M P for every maximal
ideal P of A with r(M ) ⊆ P .

6.9. Lemma. Let A be a right invariant ring such that every factor ring of A is a finite direct product
of domains and right Artinian, right uniserial rings. Then the following assertions hold.
(1) If A is an indecomposable ring, then either A is a right Artinian, right uniserial ring or A is a
domain such that every proper factor ring of A is a finite direct product of right Artinian, right uniserial
rings.
(2) A is a finite direct product of right Artinian, right uniserial rings and right invariant domains
Di such that for every Di , each proper factor ring of Di is a finite direct product of right Artinian, right
uniserial rings.
Proof. (1) Assume that A is not a right Artinian, right uniserial ring. By assumption, A is a domain. Let
B be a proper nonzero ideal of the right invariant domain A, b be a nonzero element of B, h : A → A/b2 A
be the natural ring homomorphism, and f : A/(b2 A) → A/B be the surjective ring homomorphism such
that f (a + b2 A) = a + B for all a ∈ A. Since f (A) is a proper nonzero factor ring of A, it follows from
the assumption that there exist two proper nonzero ideals C and D of A such that

b2 A = C D, A = C + D, f (A) = f (C) × f (D),

either C = b2 A or C/b2 A is a finite direct product of right Artinian, right uniserial nonzero rings and
either D = b2 A or D/b2 A is a finite direct product

of nonzero domains. Since B = A, we have b2 A = A.
Assume that D = b A. Since b A = C D and A = C + D, we have C = A = b2 A and A/(b2 A) is
2 2

a finite direct product of right Artinian, right uniserial rings. Then A/B = f (A/(b2 A)) is a finite direct
product of right Artinian, right uniserial rings.
Assume that D = b2 A. There exists an element d ∈ D \ b2 A. Then
(f (bd))2 = f (bdbd) ∈ f (bAbA) = f (b2 A) = f (0)
and f (bd) is a nilpotent element of the reduced ring f (D) that is a finite direct product of nonzero
domains. Therefore, f (bd) = f (0) and there exists an element a of A such that bd = b2 a. Then d = ba,
since A is a domain. Thus,
(f (d))2 = (f (ba))2 = f (baba) ∈ f (bAbA) = f (b2 A) = f (0)
and f (d) is a nilpotent element of the reduced ring f (D). Therefore, f (d) = f (0) and d ∈ b2 A; this is a
contradiction.
(2) The proof follows from (1).

6.10. Theorem. For a ring A with commutative multiplication of ideals, the following conditions are
equivalent.
(1) Every right ideal M of A is a multiplication right ideal such that M = M P for every maximal
ideal P of A with r(M ) ⊆ P .
(2) A is a right distributive, right Noetherian ring.
(3) A is a finite direct product of right Artinian, right uniserial rings and right invariant domains
Di such that for every Di , each proper factor ring of Di is a finite direct product of right Artinian right
uniserial rings.
Proof. The equivalence of conditions (1) and (2) follows from Theorem 6.8.

3897
 (2))=⇒(3) By Lemma 6.5(1), A is a right invariant ring. Every factor ring of a right invariant, right
distributive, right Noetherian ring with commutative multiplication of ideals has the same properties. By
Lemma 6.9(2), it is sufficient to prove that A is a finite direct product of domains and right Artinian, right
uniserial rings. Without loss of generality, we can assume that A is an indecomposable right invariant,
right distributive, right Noetherian ring that is not a domain.
Assume that A is not a right Artinian, right uniserial ring. By Lemma 6.6(3), A has two ideals X
and Y such that XY = Y X. This is a contradiction.
The implication (3)=⇒(2) is directly verified.

6.11. Proposition. For an invariant ring A with commutative multiplication of ideals, the following
conditions are equivalent.
(1) Every right ideal of A is a multiplication right ideal and A does not contain an ideal that is an
infinite direct sum of nonzero ideals of A.
(2) Every left ideal of A is a multiplication left ideal and A does not contain an ideal that is an
infinite direct sum of nonzero ideals of A.
(3) A is a distributive Noetherian ring.
Proof. It is sufficient to prove the equivalence of conditions (1) and (3).
The implication (3)=⇒(1) follows from Theorem 6.7.
(1)=⇒(3) By Theorem 6.3, A is a right distributive ring. Since A is a right distributive, left invariant
ring, A is a left distributive ring. Let M be a right ideal of A. Since A is an invariant ring that does
not contain an infinite direct sum of nonzero ideals, A is a finite-dimensional ring. Therefore, M is a
finite-dimensional right A-module that is a multiplication module by assumption. By Theorem 4.27, M
is a finitely generated module. Therefore, A is a right Noetherian ring. Since A is a right Noetherian, left
invariant ring, A is a left distributive ring.

6.12. Proposition. For a ring A with commutative multiplication of ideals, the following conditions are
equivalent.
(1) Every right ideal of A is a multiplication right ideal, every left ideal of A is a multiplication left
ideal, and A does not contain an ideal that is an infinite direct sum of nonzero ideals of A.
(2) Every ideal of A is a multiplication right ideal, every ideal of A is a multiplication left ideal, and
A does not contain an ideal that is an infinite direct sum of nonzero ideals of A.
(3) A is a distributive Noetherian ring.
Proof. The equivalence of conditions (1) and (2) follows from Note 6.2.
The implication (3)=⇒(1) follows from Theorem 6.7.
(1)=⇒(3) By Note 6.2, A is an invariant ring. By Proposition 6.11, A is a distributive Noetherian
ring.

6.13. Proposition. There exists an invariant distributive uniform subdirectly indecomposable Bezout
ring A with the following properties.
(1) If N! (A) is the set of all left zero-divisors of A and Nr (A) is the set of all right zero-divisors
of A, then N! (A) and Nr (A) are distinct maximal right (left) ideals of A, the ring A does not have
other maximal right (left) ideals, A = N! (A) +
Nr (A), the factor rings A/N! (A) and A/Nr (A) are fields,
A/J(A) ∼ = A/N! (A) × A/Nr (A), the set N! (A) Nr (A) of all the two-sided divisors of zero of A coincides
with J(A), Nr (A) is the unique proper ideal of A whose right annihilator is equal to zero, and N! (A) is
the unique proper ideal of A whose left annihilator is equal to zero;
(2) A has an ideal M such that M 2 = 0, A/M is a commutative principal ideal domain that has
exactly two maximal ideals, M is a uniserial Artinian divisible A-module, M = aM = M a ⊂ aA = Aa
for each element a ∈ A \ M , and M is the unique right or left ideal of A that is not principal;
(3) A is not isomorphic to a subring of a local ring (in particular, A is not an order in a uniserial
ring);

3898
 (4) A is not a right or left localizable ring and there exists a left divisor of zero x in A such that 1 − x
is a right divisor of zero in A and J(A) contains neither x nor 1 − x.
Proof. Let Z be the ring of integers, Q be the field of rational numbers, and Q[i] be the field of fractions
of the ring of Gaussian integers Z[i] (i2 = −1). We denote by ϕ the automorphism q1 + q2 i → q1 − q2 i
of the field Q[i]. Let R1 and R2 be the localizations of the commutative principal ideal domain Z[i] with
respect to prime ideals generated by the prime elements 2 + i and 2
− i, respectively. Then R1 and R2
are commutative, uniserial, principal ideal domains. We set R ≡ R1 R2 . Let X and Y be the ideals of
the ring R generated by the elements 2 + i and 2 − i, respectively.
The commutative principal ideal domain R has exactly two maximal ideals X and Y ,

RX = R1 , RY = R2 , ϕ(RX ) = RY , X + Y = R, J(R) = X Y = (2 + i)(2 − i)R = 5R,
and the factor ring R/J(R) is isomorphic to the direct product of the fields A/X and A/Y . The ring R
coincides with the set of all uncancellable rational Gaussian fractions whose denominators are not divided
by 2 + i or 2 − i.
We denote by M the right R-module Q[i]/RX . Then all proper submodules of the module MR are
cyclic and form an infinite countable properly ascending chain 0 = s0 R ⊂ s1 R ⊂ s2 R ⊂ . . . such that
the simple module s1 R is isomorphic to the module R/Y = R/(2 − i)R and rR (sn ) = (2 − i)n R for any
positive integer n. We turn M into a left R-module such that rm = mϕ(r) for all elements r ∈ R and
m ∈ M . It is directly verified that M is an (R, R)-bimodule that is a uniserial, Artinian, divisible right
(resp. left) R-module.
We denote by A the trivial extension of the (R, R)-bimodule M by the ring R. We recall that A is
the external direct sum of the Abelian groups R and M with multiplication defined by
(r1 , m1 )(r2 , m2 ) = (r1 r2 , r1 m2 + m1 r2 ) = (r1 r2 , m2 ϕ(r1 ) + m1 r2 )
for all elements r1 , r2 ∈ R and m1 , m2 ∈ M . The pair (1, 0) is the identity element of the ring A. We
identify M , X, and Y with the ideals (0, M ), (X, 0), and (Y, 0) of A, respectively. We identify the ring R
with the subring (R, 0) of A.
The right (left) ideals of A contained in M coincide with R-submodules of M . We note that mA = Am
for each element m ∈ M , the factor ring A/J(A) is isomorphic to the direct product of the fields A/X
and A/Y , M 2 = 0, the factor ring A/M is isomorphic to the commutative principal ideal domain R,
M is a uniserial Artinian divisible right (resp. left) Bezout A-module, the simple right (left) A-module
s1 A = As1 is the least nonzero right (left) ideal of the ring A, rA (s1 ) = Y , and A (s1 ) = X.
Now we show that A satisfies the following properties.
(P1 )
X = (2 + i)A = A (s1 ) = N! (A) and Y = (2 − i)A = rA (s1 ) = Nr (A).
(P2 ) The element (2 + i) + (2 − i) = 4 = 5 − 1 is invertible in A, since 5 ∈ J(A).
(P3 ) Denoting by x the element (2 + i)/4 ∈ X, we have
1 − x = (2 − i)/4 ∈ Y, (2 + i)s1 = s1 ϕ(2 + i) = s1 (2 − i) = 0, and xs1 = s1 (1 − x) = 0.
(P4 )
x ∈ X \ J(A) = X \ Y ⊆ N! (A) and 1 − x ∈ Y \ X = Y \ J(A) ⊆ Nr (A).
(P5 ) The ideal M of the ring A and any right or left ideal of A are comparable with respect to
inclusion.
(P6 ) A is an invariant subdirectly indecomposable Bezout ring.
(P7 ) A is a distributive ring.
(P8 ) J(A) does not contain all the right (left) divisors of zero of A.
(P9 ) A is not isomorphic to a subring of a local ring.
(P10 ) The set A \ X is not left reversible and the set A \ Y is not right reversible.
(P11 ) A is not a right or left localizable ring.

3899
 Properties (P1 )–(P4 ) are directly verified.
Property (P5 ) follows from the fact that M is a divisible R-module.
Property (P6 ) follows from Property (P5 ) and the following facts: (i) A/M is a commutative Bezout
ring; (ii) M is a Bezout A-module; (iii) mA = Am for each element m ∈ M ; (i) s1 A is the least nonzero
right (left) ideal of the ring A.
Property (P7 ) follows from Property (P6 ) and the fact that every right Bezout module over a right
invariant ring is a distributive module by Note 3.3.
Property (P8 ) follows from Property (P4 ).
We prove Property (P9 ). Assume that there exists a local ring H such that A is a subring in H. By
Property (P4 ), the elements x and 1 − x are not regular. Therefore, the elements x and 1 − x are not
invertible in the local ring H; this is a contradiction.
We

prove Property
(P10 ). By Property (P4 ), x ∈ A \ Y . By Property (P3 ), xs1 = 0. Since
r(s1 ) (A \ Y ) = Y (A \ Y ) = ∅, the set A \ Y is not right reversible. Similarly, the set A \ X is not left
reversible.
Property (P11 ) follows from Property (P10 ) and the fact that every right (resp. left) denominator
set is a right (resp. left) reversible set.
It follows from the above argument and (P1 )–(P11 ) that A is the required ring.

7. Supplementary Results

Notes 7.1 and 7.2 below are verified by a direct calculation.

z m
7.1. Note. Let A be a ring, M be a right A-module, M # A be the set of all matrices
0 a
0 m
(z ∈ Z, m ∈ M, a ∈ A), and M # 0 be the set of matrices (m ∈ M ). Then the following
0 0
assertions hold.
(1) M # A is a ring with addition
z1 m1 z2 m2 z1 + z2 m1 + m2
+ =
0 a1 0 a2 0 a1 + a2
and multiplication
z1 m1 z2 m2 z1 z2 z1 m2 + m1 a2
= .
0 a1 0 a2 0 a1 a2
(2) M # 0 is a right ideal of the ring M # A.
(3) M is a multiplication right A-module if and only if M #0 is a multiplication right ideal of M #A.
7.2. Note ([4]). Let A be a commutative ring, M be an A-module, M $ A be the set of all rows (m, a)
(m ∈ M, a ∈ A), and M $ 0 be the set of rows (m, 0) (m ∈ M ). Then the following assertions hold.
(1) M $ A is a commutative ring with addition (m1 , a1 ) + (m2 , a2 ) = (m1 + m2 , a1 + a2 ) and multi-
plication (m1 , a1 )(m2 , a2 ) = (m2 a1 + m1 a2 , a1 a2 ).
(2) M $ 0 is an ideal of the ring M $ A.
(3) M is a multiplication right A-module if and only if M $ 0 is a multiplication ideal of M $ A.


A module M is said to be π-injective if for any two submodules X and Y with X Y = 0, there
exists a direct decomposition M = X ⊕ Y such that X ⊇ X and Y ⊇ Y .
7.3. Proposition. For a multiplication module M , the following assertions hold.

(1) If the module M is π-injective and X and Y are two submodules of M with X Y = 0, then
  
(X ⊕ Y ) Z = X Z ⊕Y Z
for any submodule Z of M .

3900
 (2) If all factor modules of M are π-injective modules, then M is a distributive module.

exists
a direct decomposition M = X ⊕ Y such that

Proof. (1) Since the module M is π-injective, there
X ⊇ X and Y ⊇ Y . By Lemma 2.12(1), Z = X Z + Y Z. Therefore,
     
(X ⊕ Y ) Z = (X ⊕ Y ) (X Z +Y Z) = X Z +Y Z.


(2) Let X and Y be two submodules
of
M , h : M → M/(X Y ) be the natural epimorphism, Z
be a submodule of M , and N ≡ X Z +
Y Z ⊆ M . Since M is a multiplication module, h(M ) is a
multiplication module; in addition, h(X) h(Y ) = h(0) and h(M ) is a π-injective module by assumption.
Applying (1) to the π-injective multiplication module h(M ), we obtain
  
(h(X) ⊕ h(Y )) h(Z) = h(X) h(Z) ⊕ h(Y ) h(Z).
Therefore,     
(X + Y ) Z +X Y =X Z +Y Z +X Y.
Using this relation and the modular law, we obtain
     
(X + Y ) Z = ((X + Y ) Z) (X Z+Y Z+X Y)
     
=X Z +Y Z +Z X Y =X Z+Y Z.

7.4. Corollary. Every injective multiplication right module over a right hereditary ring is a distributive
module.
Proof. The proof follows from Proposition 7.3(2) and the fact that every factor module of any injective
right module over a right hereditary ring is an injective module [82, p.33].

Lemma 7.5 below is a corollary of the dual basis lemma (e.g., see Theorem 5.4.2 of [34]).
7.5. Lemma. Let A be a ring and M be a right A-module generated by a finite set {mi }ni=1 of elements
of M . Then Mis a projective module if and only if there exist homomorphisms f1 , . . . , fn : M → AA
such that m = ni=1 mi fi (m) for every element m of M .
7.6. Theorem ([78]). For a finitely generated multiplication module M over a commutative ring A, the
following assertions hold.
(1) M is a projective A/r(M )-module.
(2) M is a quasiprojective A-module.
(3) If r(M ) = eA for some idempotent e of A, then M is a projective A-module.
Proof. (1) Without loss of generality, we can assume that r(M ) = 0. By Theorem 3.11, there  exist
elements a1 , . . . , an , b1 , . . . , bn of A and a finite set {m1 , . . . , mn } of elements of M such that 1 = ni=1 a2i bi
and M ai ⊆ mi A for i = 1, . . . , n. Let m be any element of M . For every i ∈ {1, . . . , n}, there exists
an element c(m, i) of A such that mai = mi c(m, i); we denote by ϕi (m) the element c(m, i)ai bi of A. If
mai = mi c(m, i) = mi c(m, i) for some element c(m, i) of A, then
M (c(m, i) − c(m, i))ai = M ai (c(m, i) − c(m, i)) ⊆ mi (c(m, i) − c(m, i))A = 0.
Therefore,
(c(m, i) − c(m, i))ai ∈ r(M ) = 0, c(m, i)ai = c(m, i)ai ,
and the element ϕi (m) = c(m, i)ai bi is well defined. It is directly verified that there exist homomorphisms
ϕi : M → AA such that ϕi (m) = c(m, i)ai bi (i = 1, . . . , n). We have
n n n
m=m·1= 2
mai bi = mi c(m, i)ai bi = mi ϕi (m).
i=1 i=1 i=1
By Lemma 7.5, M is a projective module.
(2) Since every homomorphic image of M is a right A/r(M )-module, the assertion follows from (1).

3901
 (3) We can assume that e = 1. Since A is a right invariant ring, e is a central idempotent of A.
Therefore, we can identify the factor ring A/r(M ) with the ring (1 − e)A. By (1), M is a projective
(1 − e)A-module. Therefore, M is a projective A-module.

7.7. Proposition. For a distributive module M over a commutative ring A, the following assertions
hold.
(1) Every finitely generated submodule of M is a quasiprojective A-module.
(2) Every finitely generated submodule N of M is a projective A/r(N )-module.
(3) If M is a prime module, then every finitely generated submodule of M is a projective A/r(M )-
module.
(4) If M is a faithful prime module, then every finitely generated submodule of M is a projective
A-module.
Proof. Let N be a finitely generated submodule of M . By Theorem 3.8, N is a multiplication module.
By Theorem 7.6(1) and Theorem 7.6(2), N is a quasiprojective A-module and N is a projective A/r(N )-
module.
Assume that N = 0 and M is a prime module. Then r(M ) = r(N ); therefore, N is a projective
A/r(M )-module. If r(M ) = 0, then N is a projective A-module.

Given a ring A, a right A-module X is said to be flat if for any monomorphism of left A-modules
L → L, the natural group homomorphism X ⊗ L → X ⊗ L is a monomorphism.
7.8. Proposition. Let M be a flat multiplication right module over a ring A. Then the following
assertions hold.
(1) If the lattice of ideals of A is distributive, then M is a distributive module.
(2) If every ideal of A is an idempotent ideal, then M is a distributive module.



Proof. (1) Since M is a flat right A-module, it is known that M B M C = M (B C) for any two left
ideals B and C of A (e.g., see Theorem 6.3 of [85] or Proposition 8.5 of [87]). By Note 1.11, M is a
distributive module.
(2) Let X, Y , and Z be three ideals of A. Then
    
(X + Y ) Z = ((X + Y ) Z)2 ⊆ (X + Y )Z = XZ + Y Z ⊆ X Z +Y Z ⊆ (X + Y ) Z.
Therefore, the lattice of ideals of A is distributive. By (1), M is a distributive module.

REFERENCES
1. M. M. Ali, “The Ohm type properties for multiplication ideals,” Beiträge Algebra Geom., 37, No. 2,
399–414 (1996).
2. D. D. Anderson, “Multiplication ideals, multiplication rings and the ring R(X),” Can. J. Math.,
28, 760–768 (1976).
3. D. D. Anderson, “Some remarks on multiplication ideals,” Math. Japonica, 4, No. 4, 463–469 (1980).
4. D. D. Anderson, “Some remarks on multiplication ideals II,” Comm. Algebra, 28, No. 5, 2577–2583
(2000).
5. A. Barnard, “Multiplication modules,” J. Algebra, 71, No. 1, 174–178 (1981).
6. J. W. Brewer and E. A. Rutter, “A note on finitely generated ideals which are locally principal,”
Proc. Amer. Math. Soc., 31, 429–432 (1972).
7. H. S. Butts and R. C. Phillips, “Almost multiplication rings,” Can. J. Math., 17, 267–277 (1965).
8. V. Camillo, “Distributive modules,” J. Algebra, 36, No. 1, 16–25 (1975).
9. V. P. Camillo, “Modules whose quotients have finite Goldie dimension,” Pacific J. Math., 69, No. 5,
337–338 (1977).
10. Y. H. Cho, “On multiplication modules, I,” Bull. Honam Math. Soc., 13, 75–80 (1996).
11. C. W. Choi, “Multiplication modules and endomorphisms,” Math. J. Toyama Univ., 18, 1–8 (1995).

3902
12. C. W. Choi and E. S. Kim, “On multiplication modules,” Kyungpook Math. J., 32, 97–102 (1992).
13. C. W. Choi and E. S. Kim, “Some remarks on multiplication modules,” Tamkang J. Math., 24,
No. 3, 309–316 (1993).
14. C. W. Choi and E. S. Kim, “A remark on multiplication modules,” Bull. Korean Math. Soc., 31,
No. 2, 163–165 (1994).
15. C. W. Choi and P. F. Smith, “On endomorphisms of multiplication modules,” J. Korean Math.
Soc., 31, No. 1, 89–95 (1994).
16. C. J. Cooper and J. A. Johnson, “The structure of fake multiplication modules and rings,” Tamkang
J. Math., 19, No. 2, 39–48 (1988).
17. Z. A. El-Bast and P. F. Smith, “Multiplication modules,” Comm. Algebra, 16, No. 4, 755–779
(1988).
18. Z. A. El-Bast and P. F. Smith, “Multiplication modules and theorems of Mori and Mott,” Comm.
Algebra, 16, No. 4, 781–796 (1988).
19. V. Erdoǧdou, “Multiplication modules which are distributive,” J. Pure Appl. Algebra, 54, Nos. 2–3,
209–213 (1988).
20. V. Erdoǧdou, “Regular multiplication rings,” J. Pure Appl. Algebra, 59, No. 1, 55–59 (1989).
21. J. Escoriza and B. Torrecillas, “Multiplication modules relative to torsion theories,” Comm. Algebra,
23, No. 11, 4315–4331 (1995).
22. S. Feigelstock and R. Raphael, “Distributive multiplication rings,” Period. Math. Hungar., 25, No. 2,
161–165 (1992).
23. R. W. Gilmer and J. L. Mott, “Multiplication rings as rings in which ideals with prime radical are
primary,” Trans. Amer. Math. Soc., 114, 40–52 (1965).
24. R. Gordon and J. C. Robson, “Krull dimension,” Memoirs Amer. Math. Soc., 133, 1–78 (1973).
25. M. Griffin, “Some results on v-multiplication rings,” Can. J. Math., 19, 710–722 (1967).
26. M. P. Griffin, “Valuations and Prufer rings,” Can. J. Math., 26, 412–429 (1974).
27. M. Griffin, “Multiplication rings via their total quotient rings,” Can. J. Math., 26, 430–449 (1974).
28. W. Heinzer and J. Ohm, “An essential ring which is not a v-multiplication ring,” Can. J. Math.,
25, 856–861 (1973).
29. P. A. Hummadi and A. G. Naoum, “Semi-multiplication modules,” Period. Math. Hungar., 27,
No. 3, 199–205 (1993).
30. R. K. Jain, “On multiplication rings,” Math. Student, 46, Nos. 2–4, 359–363 (1978).
31. R. K. Jain, “Generalized multiplication modules,” Riv. Mat. Univ. Parma (4), 7, 461–472 (1981).
32. R. K. Jain, “Polynomial rings and multiplication rings,” Indian J. Pure Appl. Math., 17, No. 11,
1317–1318 (1986).
33. I. Kaplansky, “Projective modules,” Ann. Math., 68, 372–377 (1958).
34. F. Kasch, Modules and Rings, Academic Press, London (1982).
35. E. S. Kim and C. W. Choi, “On multiplication modules,” Kyungpook Math. J., 32, No. 1, 97–102
(1992).
36. W. Krull, “Über Multiplikalionsringe,” Sitzber. Heidelberg. Akad. Wiss. Abhandl., 5, 13–18 (1925).
37. W. Krull, “Über allgemeine Multiplikationsringe,” Tohoku Math. J., 41, 320–326 (1936).
38. M. D. Larsen and P. J. McCarthy, “Overrings of almost multiplication rings,” J. Sci. Hiroshima
Univ. Ser. A-I Math., 31, 123–129 (1967).
39. M. D. Larsen and P. J. McCarthy, Multiplicative Theory of Ideals, Academic Press, New York–
London (1971).
40. C. B. Liu, “Some results on multiplication modules,” J. Luoyang Univ., 10, No. 4, 15–18 (1995).
41. G. M. Low and P. F. Smith, “Multiplication modules and ideals,” Comm. Algebra, 18, No. 12,
4353–4375 (1990).
42. F. Mehdi, “On multiplication modules,” Mathematics Student, 42, 149–153 (1974).

3903
 43. F. Mehdi, “On multiplication and weak multiplication modules,” Riv. Mat. Univ. Parma (4), 2,
107–112 (1976).
44. S. Mori, “Axiomatische Begründung des Multiplikationsringes,” J. Sci. Hiroshima Univ., A 3, 45–59
(1932).
45. S. Mori, “Über allgemeine Multiplikationsringe I,” J. Sci. Hiroshima Univ., A 4, 1–26 (1934).
46. S. Mori, “Über allgemeine Multiplikationsringe II,” J. Sci. Hiroshima Univ., A 4, 33–109 (1934).
47. S. Mori, “Allgemeine Z.P.I. Ringe,” J. Sci. Hiroshima Univ., A 10, 117–136 (1940).
48. S. Mori, “Struktur der Multiplikationsringe,” J. Sci. Hiroshima Univ., A 16, 1–11 (1952).
49. S. Mori, “Über Idealtheorie der Multiplikationsringe,” J. Sci. Hiroshima Univ., A 19, 429–437
(1956).
50. J. L. Mott, “Equivalent conditions for a ring to be a multiplication ring,” Can. J. Math., 16, No. 3,
429–434 (1964).
51. J. L. Mott, “Multiplication rings containing only finitely many minimal prime ideals,” J. Sci. Hi-
roshima Univ. Ser. A-I, 33, 73–83 (1969).
52. A. G. Naoum, “The Ohm type properties for finitely generated multiplication ideals,” Period. Math.
Hungar., 18, 287–293 (1987).
53. A. G. Naoum, “On the dual of a finitely generated multiplication module, II,” Beitrage Algebra
Geom., No. 27 83–88 (1988).
54. A. G. Naoum, “On the ring of endomorphisms of a finitely generated multiplication module,” Period.
Math. Hungar., 21, No. 3, 249–255 (1990).
55. A. G. Naoum, “Flat modules and multiplication modules,” Period. Math. Hungar., 21, No. 4,
309–317 (1990).
56. A. G. Naoum, “A note on projective modules and multiplication modules,” Beitrage Algebra Geom.,
No. 32 27–32 (1991).
57. A. G. Naoum, “On the ring of endomorphisms of a multiplication module,” Period. Math. Hungar.,
29, No. 3, 277–284 (1994).
58. A. G. Naoum and F. H. Al-Alwan, “Dedekind modules,” Comm. Algebra, 24, No. 2, 397–412 (1996).
59. A. G. Naoum and F. H. Al-Alwan, “Dense submodules of multiplication modules,” Comm. Algebra,
24, No. 2, 413–424 (1996).
60. A. G. Naoum and W. K. H. Al-Aubaidy, “A note on multiplication modules and their rings of
endomorphisms,” Kyungpook Math. J., 35, No. 2, 223–227 (1995).
61. A. G. Naoum, B. Al-Hashimi, and H. R. Hassan, “Maximal, prime and primary submodules of
finitely generated multiplication modules,” Dirasat Ser. B Pure Appl. Sci., 18, No. 1, 7–17 (1991).
62. A. G. Naoum, B. Al-Hashimi, and J. R. Kider, “On the module of homomorphisms on finitely
generated multiplication modules, I,” Period. Math. Hungar., 22, No. 2, 97–105 (1991).
63. A. G. Naoum, B. Al-Hashimi, and J. R. Kider, “On the module of homomorphisms on finitely
generated multiplication modules, II,” Period. Math. Hungar., 22, No. 3, 153–160 (1991).
64. A. G. Naoum, B. Al-Hashimi, and K. R. Sharaf, “The dual of a finitely generated multiplication
module,” Period. Math. Hungar., 20, No. 1, 65–74 (1989).
65. A. G. Naoum and M. M. Balboul, “On finitely generated multiplication ideals in commutative
rings,” Beiträge zur Algebra und Geom., 19, 75–82 (1985).
66. A. G. Naoum and M. A. K. Hasan, “On the sum and intersection of finitely generated projective
ideals,” Arab. J. Math. 4, Nos. 1–2, 61–75 (1983).
67. A. G. Naoum and M. A. K. Hasan, “On finitely generated projective ideals and the Ohm condition,”
Period. Math. Hungar., 16, 51–60 (1985).
68. A. G. Naoum and M. A. K. Hasan, “The residual of finitely generated multiplication modules,”
Arch. Math., 46, No. 3, 225–230 (1986).
69. A. G. Naoum, H. J. Mustafa, and M. S. Abdullah, “On finitely generated projective modules and
multiplication modules,” Period. Math. Hungar., 20, No. 1, 57–63 (1989).

3904
70. A. G. Naoum and K. R. Sharaf, “A note on the dual of a finitely generated multiplication module,”
Beitrage Algebra Geom., No. 27, 5–11 (1988).
71. Y. S. Park and C. W. Choi, “Multiplication modules and characteristic submodules,” Bull. Korean
Math. Soc., 32, No. 2, 321–328 (1995).
72. S. Singh, “Principal ideals and multiplication rings,” J. London Math. Soc. (2), 3, 311–320 (1971).
73. S. Singh, “Principal ideals and multiplication rings, II,” J. London Math. Soc. (2), 5, 613–623
(1972).
74. S. Singh and Y. Al-Shaniafi, “Quasi-injective multiplication modules,” Comm. Algebra, 28, No. 7,
3329–3334 (2000).
75. S. Singh and Y. Al-Shaniafi, “Multiplication modules,” Comm. Algebra, 29, No. 6, 2597–2609 (2001).
76. S. Singh and F.Mehdi, “Multiplication modules,” Can. Math. Bull., 22, No. 1, 93–98 (1979).
77. P. F. Smith, “On noncommutative AM-rings,” Houston J. Math., 11, No. 3, 405–422 (1985).
78. P. F. Smith, “Some remarks on multiplication modules,” Arch. Math., 50, No. 3, 223–235 (1988).
79. P. F. Smith, “Multiplication modules and projective modules,” Period. Math. Hungar., 29, No. 2,
163–168 (1994).
80. P. F. Smith, T. Ukegawa, and N. Umaya, “On noncommutative Noetherian multiplication rings,”
Math. Semin. Notes Kobe Univ., 11, No. 3, 259–276 (1983).
81. W. W. Smith, “Projective ideals of finite type,” Can. J. Math., 21, 1057–1061 (1969).
82. B. Stenström, Rings of Quotients: An Introduction to Methods of Ring Theory, Springer, Berlin
(1975).
83. W. Stephenson, “Modules whose lattice of submodules is distributive,” Proc. London Math. Soc.,
28, 291–310 (1974).
84. S. M. Tripathi, “On left orders and left multiplication rings,” Ganita, 28, 1–13 (1977).
85. A. A. Tuganbaev, Semidistributive Modules and Rings, Kluwer Academic, Dordrecht–Boston–
London (1998).
86. A. A. Tuganbaev, Distributive Modules and Related Topics, Gordon and Breach, Amsterdam (1999).
87. A. A. Tuganbaev, Rings Close to Regular, Kluwer Academic, Dordrecht–Boston–London (2003).
88. A. A. Tuganbaev, “Multiplication modules over noncommutative rings,” preprint.
89. A. A. Tuganbaev, “Distributive and multiplication modules and rings,” preprint.
90. H. Uda, “A characterization of Prufer v-multiplication domains in terms of polynomial grade,”
Hiroshima Math. J., 16, No. 1, 115–120 (1986).
91. T. Ukegawa, “Some properties of non-commutative multiplication rings,” Proc. Japan Acad., 54A,
No. 9, 279–284 (1978).
92. T. Ukegawa, “On the unique maximal idempotent ideals of nonidempotent multiplication rings,”
Proc. Japan Acad. Ser. A Math. Sci., 55, No. 4, 132–135 (1979).
93. T. Ukegawa, “Left Noetherian multiplication rings,” Osaka J. Math., 17, No. 2, 449–453 (1980).
94. T. Ukegawa, “On M -rings and general ZPI-rings,” Osaka J. Math., 19, No. 4, 923–930 (1982).
95. T. Ukegawa, “Some remarks on M -rings,” Math. Japonica, 28, 195–203 (1983).
96. C. A. Wood, “Commutative rings for which each proper homomorphic image is a multiplication
ring,” J. Sci. Hiroshima Univ. Ser. A-I Math., 33, 85–94 (1969).
97. C. A. Wood and D. E. Bertholf, “Commutative rings for which each proper homomorphic image is
a multiplication ring, II,” Hiroshima Math. J., 1, 1–4 (1971).

3905