Вы находитесь на странице: 1из 18

Singular and Nonsingular Modules Relative to a

Torsion Theory
Seil EKEN and Mustafa ALKAN

Abstract
In this paper we introduce and study torsion-theoretic generalizations
of singular and nonsingular modules by using the concept of -essential
submodule for a hereditary torsion theory . We introduce two new mod-
ule classes called -singular and non- -singular modules. We investigate
some properties of these module classes and present some examples to
show that these new module classes are dierent from singular and non-
singular modules. We give a characterization of -semisimple rings via
non- -singular modules. We prove that if M= (M ) is non- -singular for
a module M , then every submodule of M has a unique -closure. We
give some properties of the torsion theory generated by the class of all
-singular modules. We obtain a decomposition theorem for a strongly
-extending module by using non- -singular modules.
2010 Mathematics Subject Classication: 16S90, 16D60, 16D70.
Keywords and phrases: -essential submodule, -singular module,
non- -singular module, -semisimple module.

1 Introduction
In recent years, some authors have dened several relative notions of module the-
ory by using the concept of -essential submodule which was dened by Gomez
Pardo in 1985. In 2011, S. Asgari and A. Haghany introduced a dimension
for a module M , called -rank of M , by using -essential submodules (see [5]).
Note that in [5] a -essential submodule was called a pseudo -essential submod-
ule. In 2012, the present authors introduced a torsion-theoretic generalization
of extending modules by using -essential submodules (see [8]) and gave some
characterizations of modules M over a ring R such that every submodule has
a unique closure relative to a hereditary torsion theory on M od-R (see [9]). In
2014, T. Albu used -essential submodules to prove the Relative Osofsky-Smith
Theorem (see [1]).
Throughout this paper, R will denote an associative ring with identity, M od-
R will be the category of unitary right R-modules, and all modules and module
homomorphisms will belong to M od-R. Let be a hereditary torsion theory
on M od-R. In this paper we dene and study torsion-theoretic generalizations
of singular and nonsingular modules by using -essential submodules. After

1
this introductory section, this paper is divided into three sections. In section 2,
we recall some lattice-theoretic concepts and then apply these lattice-theoretic
results to the lattice (Sat (M ); ) of -saturated submodules of a module M to
obtain corresponding torsion-theoretic notions and results. In the third section
we dene the concept of -singular submodule which is a relative notion of
the singular submodule of a given module M . We introduce two new module
classes called -singular modules and non- -singular modules. We present some
examples to show that these new module classes are dierent from singular and
nonsingular modules. We show that if A is -essential in B, then B=A is -
singular (see Theorem 3.7) and if B= (B) is non- -singular then the converse
of this result is also true (see Proposition 3.12). We give a characterization
of -semisimple rings via non- -singular modules (see Theorem 3.14). We show
that if RR is -torsionfree and nonsingular, then a right R-module M is singular
if and only if it is -singular (see Proposition 3.10). It is well-known that if P
is a projective module, then P=X is singular if and only if X is large in P . We
prove a torsion-theoretic generalization of this result. We prove that if P is a
-torsionfree -projective right R-module and X is a submodule of P , then P=X
is -singular if and only if X is -essential in P (see Theorem 3.13). In Section
4, we present further torsion theoretic results by using the preradical Z which
is dened in the third section. We prove that if R= (R) is non- -singular, then
the class of all -singular right R-modules is closed under module extensions
and -essential extensions (see Theorem 4.8). Let M be a right R-module and
N M and let N denote the (unique) submodule of M containing N such
that N =N = Z(M=N ) where Z(M=N ) is the singular submodule of M=N .
This process can be repeated, so one can dene N , N , and so on (see [12]).
N is called the Goldie closure of N in M . In [12], Goldie proved that the
quotient M=N is nonsingular. We dene a torsion-theoretic generalization
of the Goldie closure of a submodule and generalize Goldies stated result to
the torsion-theoretic setting (see Denition 4.1 and Theorem 4.5). In [9], the
present authors dened the concept of -UC module which is a torsion-theoretic
generalization of the concept of UC module which was dened in [16]. We prove
that if M is a right R-module such that M= (M ) is non- -singular then M
is a -UC module (see Theorem 4.7). We give some properties of the torsion
theory generated by the class of all -singular modules (see Proposition 4.9 and
Theorem 4.10). We obtain a decomposition theorem for a strongly -extending
module by using non- -singular modules (see Theorem 4.14).
Now we recall some fundamental concepts of torsion theory from [6] and
[11]. Let := (T ; F) be a torsion theory on M od-R. Modules in T will be
called -torsion and modules in F will be called -torsionfree modules. Let
M 2 M od-R. Then the -torsion submodule of M , denoted by (M ), is dened
to be the sum of all -torsion submodules of M . The torsion class T is given
by T = fM 2 M od-R : (M ) = M g. F is referred to as torsionfree class
and it is given by F = fM 2 M od-R : (M ) = 0g. The torsion class T is
closed under homomorphic images, direct sums and module extensions. The
torsionfree class F is closed under isomorphisms, submodules, direct products
and module extensions. In this paper, all torsion theories are assumed to be

2
hereditary, that is, we assume that submodules of -torsion modules are -
torsion, unless stated otherwise. A torsion theory is called stable if the torsion
class is closed under injective hulls; equivalently, (M ) is a closed submodule of
M for every module M . If I is an idempotent ideal of R, then it is well-known
that I determines a hereditary torsion theory I with torsion class fM 2 M od-
R : M I = 0g. We refer to I as the torsion theory corresponding to I. The
torsion theory generated by the class of all singular modules is called Goldie
torsion theory. We will denote Goldie torsion theory by G . It is well-known
that G is hereditary and the G -torsion submodule G (M ) of a module M is
the second singular submodule of M . That is, G (M ) is the submodule Z2 (M )
of M such that Z2 (M )=Z(M ) = Z(M=Z(M )), where for a module M , Z(M )
denotes the singular submodule of M .
Let R be a ring and M be a right R-module. A submodule N of M is
called -dense ( -saturated) in M if M=N is -torsion ( -torsionfree). The set
of all -dense right ideals of R will be denoted by F (R). It is well known that
(M ) = fm 2 M : (0 : m) 2 F (R)g, where (0 : m) = fr 2 R : mr = 0g.
We let Sat (M ) denote the set of all -saturated submodules of M . Sat (M )
is closed under arbitrary intersections, (M ) = \fN : N M; N 2 Sat (M )g
and if N is a -saturated submodule of M , (N ) = (M ). The -saturation of
N in M , denoted by N c , is dened to be the intersection of all the -saturated
submodules of M that contain N . It is well-known that (M=N ) = N c =N .

2 Preliminaries
In this section we recall some lattice-theoretic concepts from [2], [3], [4] and [17]
and then apply these lattice-theoretic results to the lattice (Sat (M ); ) of all -
saturated submodules of a module M to obtain corresponding torsion-theoretic
notions and results.
It is known that Sat (M ), the set of all -saturated submodules of a module
M , is an upper continuous modular lattice with respect to inclusion having
least element (M ) and greatest element M (see [17, Chapter 9, Proposition
4.1]). Following [1], if P is any property on lattices, we say that a module M
is/has -P, if the lattice (Sat (M ); ) is/has P. Also, a submodule N of a
module M is/has -P if its -saturation N c , which is an element of the lattice
(Sat (M ); ) is/has P.
Let L be a bounded lattice with the least element 0 and the greatest element
1. For a; b 2 L, we write

b=a := [a; b] = fx 2 L : a x bg:

An element e of L is said to be essential (in L) if e ^ x 6= 0 for every x 6= 0


in L. We denote the set of all essential elements of L by E(L) [17, Chapter III].
A submodule N of a module M is called -essential in M if N c is an essential
element of the lattice Sat (M ) [13]. Let a; b be elements of L. If a b and
a is essential in [0; b], then b is said to be an essential extension of a. If a has
no proper essential extension in L, then a is said to be essentially closed in L

3
[17, Chapter III]. A submodule K of a module M is called -closed in M if K
is an essentially closed element in Sat (M ) [13]. For an element a 2 L, we say
that c 2 L is a closure of a in L if a 2 E(c=0) and c is essentially closed in L.
A lattice L is called essentially closed if for each element a 2 L there exists a
closure of a in L. Every upper continuous modular lattice is essentially closed
[4]. So Sat (M ) is an essentially closed lattice for a module M . Let K be a
submodule of a module M . A closure of K c in Sat (M ) is said to be a -closure
of K in M [13]. An element b 2 L is a pseudo-complement in L if there exists
an element a 2 L such that a ^ b = 0 and b is maximal in the set of all elements
c in L with a ^ c = 0; we say in this case that b is a pseudo-complement of a. L
is called pseudo-complemented if every element of L has a pseudo-complement,
and strongly pseudo-complemented if for all a; b 2 L with a ^ b = 0, there exists
a pseudo-complement p of a in L such that b p. Every upper continuous
modular lattice is strongly pseudo-complemented [4]. So Sat (M ) is a strongly
pseudo-complemented lattice for a module M . Let L be a submodule of a
module M . A submodule K of M is said to be a -complement of L in M
if K is a pseudo-complement of Lc in Sat (M ) [13]. An element c of L is a
complement (in L) if there exists an element a of L such that a ^ c = 0 and
a _ c = 1; we say in this case that c is a complement of a (in L). The lattice
L is said to be complemented if every element of L has a complement in L [17,
Chapter III]. A submodule N of a module M is said to be a -direct summand
of M if its -saturation N c has a complement in the lattice Sat (M ) [1].
The lattice L is said to be simple if L = f0; 1g and 0 6= 1 [2], [3]. A module
M is called a -simple module if Sat (M ) is a simple lattice [11]. An element
a 2 L is said to be an atom if a 6= 0 and a=0 = f0; ag, i.e., a=0 is a simple lattice.
We denote by A(L) the set, possibly empty, of all atoms of L. The socle Soc(L)
of a complete lattice L is the join of all atoms of L, i.e., Soc(L) := _A(L). L
is said to be semi-atomic if 1 is a join of atoms of L [2], [3]. The -socle of a
module M is the socle Soc(Sat (M )) of the complete lattice Sat (M ) and M
is called a -semisimple module if Sat (M ) is a semi-atomic lattice [11]. An
element c of L is called compact in L if whenever c _x2A x for a subset A of
L, there is a nite subset F of A such that c _x2F x. The lattice L is said
to be compact if 1 is a compact element in L, and compactly generated if it is
complete and every element of L is a join of compact elements [17, Chapter III].
A module M is said to be -compactly generated if Sat (M ) is a compactly
generated lattice [1].

Proposition 2.1 Let M be a right R-module.


(1) If M is -semisimple then every submodule of M is a -direct summand
of M .
(2) Suppose that M is a -compactly generated module. Then M is -
semisimple if and only if every submodule of M is a -direct summand of M .

Proof (1) follows from [17, Chapter III, Proposition 5.4] and (2) follows from
[17, Chapter III, Proposition 5.5].

4
3 -singular and Non- -singular Modules
In this section we dene the relative notion of the singular submodule of a
given right R-module M . We dene and study two new module classes called
-singular and non- -singular modules. We present some examples to show that
these new module classes are dierent from singular and nonsingular modules.
Let M be a right R-module. We dene the set

Z (M ) = fx 2 M : xI (M ) for some I E RR g:

Note that in [13] Gomez Pardo dened a left exact preradical which is
denoted by ZF assigning each right R-module M to ZF (M ) = fm 2 M :
ann(m) E RR g. In [13] Gomez Pardo gave necessary and su cient conditions
for ZF to be a radical and he established the existing relationship between this
functor and spectral Gabriel topologies. Note that ZF (M ) Z (M ) for every
right R-module M .

Lemma 3.1 Let M be a right R-module. Then the following statements are
satised.
(1) Z (M ) is a submodule of M .
(2) If f : M ! N is any R-homomorphism, then f (Z (M )) Z (N ).
(3) If N M , then Z (N ) = Z (M ) \ N .
(4) (M ) Z (M ).
(5) If RR is -torsionfree or is a stable torsion theory, then the singular
submodule Z(M ) is contained in Z (M ).
(6) If RR and M are -torsionfree, then Z(M ) = Z (M ).
(7) Z (M= (M )) = Z (M )= (M ).
(8) Z (M ) -soc(RR ) (M ).

Proof (1) Since RR E RR , we always have 0 2 Z (M ). Let x; y 2 Z (M ).


Then xI (M ) and yJ (M ) for some I E RR and J E RR . Since
I \ J E RR by [8, Proposition 2.2-(3)] and (x y)(I \ J) (M ), we get that
x y 2 Z (M ). Let r 2 R. Consider the map fr : RR ! RR , dened by
fr (s) = rs for all s 2 R. Clearly fr is an R-homomorphism. Since I E RR , we
have fr 1 (I) = fs 2 R : rs 2 Ig E RR by [13, Proposition 2.4-(b)]. Also we
have xr fr 1 (I) xI (M ). This shows that xr 2 Z (M ). Thus Z (M ) is
a submodule of M .
(2) Let x 2 f (Z (M )). Then x = f (m) for some m 2 Z (M ). Since mI
(M ) for some I E RR , we have f (mI) = f (m)I = xI f ( (M )) (N ).
This shows that x 2 Z (N ). Thus f (Z (M )) Z (N ).
(3) Clear from the denition and part (2).
(4) This statement follows from the denition of Z (M ) and the fact that
RR E RR .
(5) Z(M ) = fx 2 M : xI = 0 for some I E RR g. If RR is -torsionfree, then
the set of all large right ideals of R is equal to the set of all -essential right
ideals of R. This fact shows that Z(M ) Z (M ).

5
If is a stable torsion theory, then every large submodule of M is -essential
in M by [5, Corollary 2.4]. This implies that Z(M ) Z (M ).
(6) If an R-module is -torsionfree, then the concepts of large and -essential
submodules coincide. The result follows from this fact.
(7) This statement follows from the denitions and part (2).
(8) For any x 2 Z (M ), we have xI (M ) for some I E RR . By [13,
Proposition 2.2], I c E RR . By [2, Proposition 3.1], we have -soc(RR ) I c .
The exactness of the sequence 0 ! xI ! xI c ! xI c =xI ! 0 gives that
xI c (M ). Thus x -soc(RR ) (M ) and so Z (M ) -soc(RR ) (M ).

Lemma 3.1 shows that Z is a left exact preradical (e.g. see [17, Chapter
VI]).

Denition 3.2 Let M be a right R-module. The submodule Z (M ) is called


the -singular submodule of M . M is called a -singular module provided M =
Z (M ) and M is called a non- -singular module provided Z (M ) = 0.

By using the denitions and Lemma 3.1, we get the following corollary.

Corollary 3.3 (1) Every -torsion module is -singular and every non- -singular
module is -torsionfree.
(2) If is a stable torsion theory or RR is -torsionfree, then every singular
module is -singular and every non- -singular module is nonsingular.
(3) A right R-module M is -singular if and only if M= (M ) is -singular.

Let denote the stable torsion theory (M od-R; 0). Then a right R-module
M which is not singular is clearly -singular. The following example shows that
there is a stable torsion theory other than in which a -singular module
need not to be singular.

F [x] 0
Example 3.4 Let F be a eld and R = . Con-
F [x; y]=(x; y 2 ) F [y]=(y 2 )
sider the right R-module M := RR .
By [14, page 36, exercise 6], Z2 (RR ) 6= Z(RR ). This implies that Z2 (RR )
is not a singular module. But Z2 (RR ) is a G -singular module since it is G -
torsion.

The following example shows that a singular module need not to be -


singular for a torsion theory which is not stable. It also shows that a non- -
singular module need not to be nonsingular.
2 3
F F F
Example 3.5 Let R = 4 0 F 0 5 where F is a eld and let I denote the
0 0 F
2 3
F F F
torsion theory on M od-R corresponding to the idempotent ideal I = 4 0 0 0 5,
0 0 0
that is TI = fN 2 M od-R : N I = 0g. Consider the right R-module M := RR .

6
(1) M= I (M ) is a singular
2 right R-module:
3
0 F F
Note that I (M ) = 4 0 F 0 5. It is easy to see that I (M ) is a large
0 0 F
submodule of M . Hence M= I (M ) is a singular right R-module.
(2) M=
2 I (M ) is3a non- I -singular right R-module which is not nonsingular:
a b c
Let 4 0 d 0 5 + I (M ) 2 Z I (M= I (M )) where a; b; c; d; e 2 F . Then
0 0 e 2 3
a b c
there is a I -large right ideal J of R such that 4 0 d 0 5 J I (M ). Since
0 0 e
RR is not I -torsion, I (RR ) is not I -large in RR . It follows from [13, Propo-
sition 2.4-(a)-(i)] that J 6 2I (RR ). 3
x y z
So there exists an element 4 0 t 0 5 2 J where x; y; z; t; k 2 F and x 6= 0.
0 0 k
2 32 3 2 3
a b c x y z ax bt + ay ck + az
Since 4 0 d 0 5 4 0 t 0 5 = 4 0 dt 0 5 2 I (M ), we
0 0 e 0 0 k 0 0 ke
2 3
a b c
have ax = 0 and hence a = 0. This shows that 4 0 d 0 5 2 I (M ) and
0 0 e
hence Z I (M= I (M )) = 0. Considering part (1), we deduce that M= I (M ) is a
non- I -singular right R-module which is not nonsingular.
(3) By parts (1) and (2), we see that M= I (M ) is a singular right R-module
which is not I -singular.

The following example shows that a nonsingular module need not to be


non- -singular.

Example 3.6 Let R, I, I and M be as in Example 3.5. Then M is a nonsin-


gular right R-module by [15, (7.14d) Example]. But M is not a non- I -singular
right R-module. Because 0 6= I (M ) Z I (M ) by Lemma 3.1-(4) and hence
Z I (M ) 6= 0.

Theorem 3.7 (1) A module C is non- -singular if and only if HomR (A; C) = 0
for all -singular modules A.
f g
(2) If there exists a short exact sequence of modules 0 ! A ! B !
C ! 0 such that Im(f ) E B, then C is a -singular module.
(3) If C is a -torsionfree -singular module, then there exists a short exact
f g
sequence 0 ! A ! B ! C ! 0 such that Im(f ) E B.

Proof (1) If A is -singular and C is non- -singular, and f : A ! C is an


R-homomorphism, then f (A) = f (Z (A)) Z (C) = 0, by Lemma 3.1. Thus
HomR (A; C) = 0 whenever A is -singular and C is non- -singular.

7
Conversely, if HomR (A; C) = 0 for all -singular modules A, then in partic-
ular HomR (Z (C); C) = 0. Now the inclusion map Z (C) ! C is zero, hence
Z (C) = 0.
(2) Assume that we have such an exact sequence. Let c 2 C. Then c = g(b)
for some b 2 B. Consider the map k : RR ! B, dened by k(r) = br for all
r 2 R. Clearly k is an R-homomorphism. Since f (A) E B, by [13, Proposition
2.4-(b)], we have k 1 (f (A)) E RR , that is, the right ideal I = fr 2 R : br 2
f (A)g E RR . Now bI f (A) = ker g; hence g(bI) = g(b)I = cI = 0 (C)
and so c 2 Z (C). Thus C = Z (C), i.e., C is a -singular module.
(3) Assume that C is a -torsionfree -singular module. Choose a short exact
g
sequence 0 ! A ! B ! C ! 0 such that B is free. If fb g 2 is a basis
for B, then for each we have g(b )I = 0 for some I E RR ; hence b I A.
Since I E RR for all 2 , we get that b I E b R. By [8, Proposition
2.2-(4)], we have 2 b I E 2 b R = B. Since 2 b I A, we have
A E B by [13, Proposition 2.4-(a)-(i)].

Theorem 3.8 (1) The class of all non- -singular right R-modules is closed
under submodules, direct products, large extensions and module extensions.
(2) The class of all -singular right R-modules is closed under submodules,
factor modules and direct sums.
(3) If N is a non- -singular and -essential submodule of a module M , then
M= (M ) is non- -singular.

Proof (1) Since Z is a preradical, the closure under submodules and direct
products follows by [17, Chapter VI, Proposition 1.2].
Let A be a non- -singular module and A E B. Then A\Z (B) = Z (A) = 0.
This implies that Z (B) = 0 and hence B is non- -singular.
Suppose that 0 ! C ! B ! A ! 0 is an exact sequence of modules
with C; A non- -singular. According to Thoerem 3.7-(1), we have HomR (M; C) =
0 and HomR (M; A) = 0 for any -singular module M . By the exactness of the
sequence

0 ! HomR (M; C) ! HomR (M; B) ! HomR (M; A)

we obtain HomR (M; B) = 0. By Theorem 3.7-(1), B is non- -singular.


(2) Since Z is a left exact preradical, the closure under submodules,factor
modules and direct sums follows by [17, Chapter VI, Proposition 1.2 and Propo-
sition 1.7].
(3) Let N be non- -singular and N E M . Then N \ Z (M ) = Z (N ) = 0.
This implies that Z (M ) (M ). Since we always have the reverse inclusion,
we get the equality Z (M ) = (M ). Lemma 3.1-(7) implies that M= (M ) is
non- -singular.

Corollary 3.9 Let M be a -torsionfree -singular right R-module and f 2


HomR (R; M ). Then kerf E RR .

8
Proof f (R) is a -torsionfree -singular right R-module by Theorem 3.8-
(2). According to Theorem 3.7-(3), there exist a module L and a -essential
submodule K of L such that f (R) ' L=K. Since RR is projective, there exists
an R-homomorphism g : R ! L such that g = f where is the natural
epimorphism L ! L=K. Then kerf = g 1 (K) E RR by [13, Proposition
2.4-(b)].

Proposition 3.10 Let RR be -torsionfree and nonsingular. Then a right R-


module M is singular if and only if it is -singular.
Proof Suppose that M is a -singular right R-module. By the denitions
RR is non- -singular. The injective hull E(RR ) of RR , is also non- -singular
by Theorem 3.8-(1). Hence HomR (M; E(RR )) = 0 by Theorem 3.7-(1). [10,
4.1-(2)] implies that M is singular.
Conversely, if M is a singular right R-module, then M is -singular by
Lemma 3.1-(5).

Corollary 3.11 Let RR be -torsionfree and nonsingular. Then Z(M ) = Z (M ) =


ZF (M ) for a right R-module M .
Proof Since RR is -torsionfree, we have Z(M ) = fm 2 M : ann(m) E
RR g = ZF (M ). The result follows from Proposition 3.10.

Theorem 3.7-(2) shows that if A E B, then B=A is -singular. The next


proposition is a partial converse of this result.
Proposition 3.12 Let M be a right R-module such that M= (M ) is non- -
singular and N be a submodule of M . Then M=N is -singular if and only if
N E M.
Proof If N E M , then M=N is -singular by Theorem 3.7-(2). Suppose
that M=N is -singular. If M = (M ), then the result is clear. Assume that
M 6= (M ). Let x 2 M n (M ). Then (xI +N )=N (M=N ) for some I E RR .
Assume that xR \ N (M ). Then xI \ N (M ). The exactness of the
sequence
0 ! xI \ N ! xI ! (xI + N )=N ! 0
gives that xI (M ). This means that x 2 Z (M ). Since M= (M ) is non- -
singular, we have Z (M ) = (M ) by Lemma 3.1-(7). This means that x 2 (M ),
a contradiction. Thus xR \ N 6 (M ). Hence xr 2 N n (M ) for some r 2 R.
By [5, Proposition 2.3], N E M .

It is well-known that if P is a projective module then P=X is singular if and


only if X is large in P . In the following theorem we give a torsion-theoretic
generalization of this result.
Recall that a right R-module M is called -projective if it is projective with
respect to every R-epimorphism having a -torsionfree kernel (see [6]). Note
that in [11] a -projective module was called a -codivisible module.

9
Theorem 3.13 Let P be a -torsionfree -projective right R-module and X be
a submodule of P . Then P=X is -singular if and only if X E P .

Proof ((=) By Theorem 3.7-(2).


(=)) There exists a free right R-module F and an exact sequence
0 ! K ! F ! P ! 0 with K F . Since P is -torsionfree, K is
-saturated in F . This implies that (K) = (F ). Consider the exact sequence

0 ! K= (K) ! F= (K) ! P ! 0:

Since P is -projective and K= (K) is -torsionfree, this sequence splits, i.e.


P is isomorphic to a direct summand of F= (K) = F= (F ). Let F= (F ) =
A= (F ) B= (F ), where A= (F ) ' P . Let Y = (F ) be a submodule of A= (F )
such that A= (F )=Y = (F ) ' A=Y is -singular. If Y = (F ) is not -essential in
A= (F ) then Y = (F ) B= (F ) is not -essential in F= (F ) by [8, Proposition
2.2-(4)]. Since F= (F )= (Y = (F ) B= (F )) ' A=Y , it su ces to prove the
theorem for the right R-module F= (F ).
Let L= (F ) be a submodule of F= (F ) such that F= (F )=L= (F ) ' F=L
is -singular. F = i2I Ri where each Ri is a copy of R and F= (F ) '
i2I (Ri = (Ri )). Let i 2 I. Then Ri = (Ri )=(Ri \ L)= (Ri ) ' Ri = (Ri \ L) '
(Ri + L) =L is -singular by Theorem 3.8-(2). Ri =(Ri \ L)c is also -singular
again by 3.8-(2). Corollary 3.9 implies that (Ri \L)c E Ri . By [13, Proposition
2.2], we have Ri \ L E Ri for each i 2 I. By using [8, Proposition 2.1-(3) and
Proposition 2.2-(4)] we get that ( i2I Ri ) \ L E i2I Ri = F and so L E F .
By [5, Proposition 2.3], we have L= (F ) E F= (F ). This completes the proof.

Theorem 3.14 (1) If RR is -semisimple then M= (M ) is non- -singular for


every right R-module M .
(2) Suppose that RR is a -compactly generated module. Then RR is -
semisimple if and only if M= (M ) is non- -singular for every right R-module
M.

Proof (1) Let RR be -semisimple and M be a right R-module. (M )


Z (M ) by Lemma 3.1-(4). Let x 2 Z (M ). Then xI (M ) for some I E RR .
Since RR is -semisimple, there exists a right ideal J of R such that R= (I + J)
and I \ J are both -torsion by Proposition 2.1-(1). xR=(xI + xJ) is also -
torsion since it is a homomorphic image of R=(I + J). I E RR implies that
J (RR ) and hence xJ (M ). The exactness of the sequence

0 ! xI + xJ ! xR ! xR=(xI + xJ) ! 0

gives that xR (M ) and hence x 2 (M ). Thus Z (M ) = (M ). By Lemma


3.1-(7), Z (M= (M )) = Z (M )= (M ) = 0. Thus M= (M ) is non- -singular.
(2) Suppose that M= (M ) is non- -singular for every right R-module M .
Lemma 3.1-(7) implies that Z (M ) = (M ) for every right R-module M . Let

10
A RR . By [13, Proposition 2.9], there exists a right ideal B of R such that
A\B (RR ) and A + B E RR . By Theorem 3.7-(2), R=(A + B) is -
singular. By the assumption R= (A + B) = Z (R= (A + B)) = (R= (A + B)),
i.e. R=(A + B) is -torsion. Thus every submodule of RR is a -direct summand
of RR . By Proposition 2.1-(2), RR is -semisimple. Conversely, if RR is -
semisimple then the result follows from part (1).

4 Further Results
Recall that for any submodule N of a right R-module M , N denotes the
(unique) submodule of M containing N such that N =N = Z(M=N ). This
process can be repeated, so one can dene N , N , and so on (see [12]). In
the following denition we give a torsion-theoretic analogue of this idea.

Denition 4.1 Let N be a submodule of a right R-module M . We dene N ^


to be the submodule of M containing N such that N ^=N = Z (M=N ). This
process can be repeated, so we can dene N ^; N ^^ and so on.

It is clear that 0^ is just Z (M ).

Lemma 4.2 Let M be a right R-module and L N M . Then L^ N ^. In


particular Z (M ) = 0^ N ^.

Proof Let x 2 L^. Then x+L 2 Z (M=L) and so (x+L)I (M=L) for some
I E RR . This implies that (L : xr) 2 F (R) for every r 2 I. Since L N; we
have (N : xr) 2 F (R) for every r 2 I. This implies that (x + N )I (M=N )
and hence x 2 N ^. Thus L^ N ^.

In [12], Goldie dened an equivalence relation on the submodules of a module


M as follows. Two submodules K, L of M are related (written K L), if for any
submodule X of M , we have X \ K 6= 0 if and only if X \ L 6= 0. Generalizing
this idea to the torsion-theoretic setting we give the following denition.

Denition 4.3 Let M be a right R-module. We say that two submodules S, T


of M are -related (written S T ) if for any submodule X of M we have that
X \S 6 (M ) if and only if X \ T 6 (M ).

Clearly, is an equivalence relation on the submodules of M . It is easily


checked that if S T , then S T if and only if S E T .

Proposition 4.4 Let M be a right R-module and L; N M . Then:


(1) N + 0^ N ^.
(2) N ^ N ^^ and N ^ E N ^^
(3) If S N then S N ^.

11
Proof (1) It su ces to show that N + 0^ E N ^. Let X be a submodule of N ^
such that X \ (N + 0^) (N ^). For any x 2 X there is a right ideal I E RR
such that (xI + N )=N (M=N ). This implies that (xI + N + 0^)=(N + 0^)
(M=(N + 0^)). By the exactness of the sequence

0 ! xI \ (N + 0^) ! xI ! (xI + N + 0^)=(N + 0^) ! 0

we obtain xI (M ). This shows that x 2 X \ 0^ (N ^) and hence


X (N ^). Thus N + 0^ E N ^.
(2) Replacing N by N ^ in (1), we get N ^^ N ^ + 0^ = N ^.
(3) Let S N and x 2 S. We must show that (xI + N )=N (M=N ) for
some I E RR . Put I := (N : x). We claim that (N : x) is -essential in RR .
By [13, Proposition 2.2], it su ces to show that (N : x)c is -essential in RR . It
is easy to see that (N : x)c = (N c : x). So we may assume that N is -saturated
in M . Let r 2 Rn (RR ). We may assume that r 62 (N : x). Thus xr 62 N
and so xrR \ S = xrR 6 (M ). Otherwise xr 2 (M ) N , a contradiction.
Since S N , we have xrR \ N 6 (M ). Thus there exists an s 2 R such that
xrs 2 N n (M ). This implies that rs 2 (N : x)n (RR ). By [5, Proposition 2.1],
I = (N : x) E RR .

In [12], Goldie proved that for any submodule N of a right R-module M ,


the quotient M=N is nonsingular. In the following theorem we give a torsion-
theoretic generalization of this result.

Theorem 4.5 For any submodule N of a right R-module M , we have N ^^^ =


N ^^. In other words M=N ^^ is non- -singular.

Proof Let N M . Replacing N by N ^ in Proposition 4.4-(2), we deduce


that N ^^^ N ^^ N ^. Applying Proposition 4.4-(3) with N replaced N ^
and S with N ^^^, we get N ^^^ N ^^ and hence N ^^^ = N ^^.

Corollary 4.6 For any submodule N of a right R-module M , N ^^ is -closed


in M .

Proof Consider any submodule Y of M such that N ^^ E Y M . Then,


by Proposition 4.4-(3) and Theorem 4.5, Y N ^^ implies Y N ^^ and so
Y = N ^^.

Let L be a bounded lattice. An element a 2 L has a unique closure in L if a


has exactly one closure in L [4]. We say that an essentially closed lattice L is a
UC lattice if every element of L has a unique closure in L. A module M is said
to be a -UC module if Sat (M ) is a UC lattice, i.e., a module M is a -UC
module if and only if every submodule of M has a unique -closure in M [9].
The concept of -UC module is a torsion-theoretic generalization of the concept
of UC module which was dened in [16].

12
Theorem 4.7 Let M be a right R-module such that M= (M ) is non- -singular
and N M . Then:
(1) N E N ^.
(2) N ^ is the largest submodule of M which is -related N .
(3) N ^^ = N ^.
(4) N = N ^ if and only if N is -closed in M .
(5) N ^ is the smallest -closed submodule of M containing N . In particular,
N ^ is the unique -closure of N in M . Consequently, M is a -UC module.
(6) If Ni c M (i 2 I) then \i2I Ni c M.

Proof (1) Since M= (M ) is non- -singular, we have 0^ = Z (M ) = (M )


by Lemma 3.1-(7). By Lemma 4.2, (M ) N ^ and hence (M ) = (N ^).
Proposition 4.4-(1) gives that N + 0^ = N + (N ^) E N ^. By [5, Proposition
2.3], (N + (N ^)) = (N ^) E N ^= (N ^). [5, Proposition 2.3] implies that N E
N ^.
(2) If S N , then Proposition 4.4-(3) gives S N ^.
(3) Part (1) and Proposition 4.4-(2), imply that N ^^ N^ N . So part
(2) shows that N ^ = N ^^.
(4) First assume that N = N ^. Then N = N ^^ by part (3). N ^^ is -closed
in M by Corollary 4.6. So N is -closed in M . Conversely, if N is -closed in
M , then N = N ^ by part (1).
(5) Note that N ^ = N ^^ c M by part (3) and Corollary 4.6. Moreover,
if N X c M , then N ^ X^ = X by part (4) and Lemma 4.2. Thus N ^
is the smallest -closed submodule of M containing N . In particular, if X is a
-closure of N , then we have N ^ = X.
(6) M is a -UC module by part (5). Now the result follows from [9, Propo-
sition 3.11].

In general the class of -singular modules is not closed under module ex-
tensions or -essential extensions. For example, consider the torsion theory
= (0; M od-R). Let R = Z=4Z. Consider R as a right R-module on itself.
Then Z (R) = 2R and R=2R ' 2R, i.e. R=2R is -singular. Thus R is an
extension of the -singular module 2R by the -singular module R=2R. But R
is not -singular. We also note that R is a -essential extension of the -singular
module 2R (see [14, page 33]). The following theorem shows that if R= (R) is
non- -singular the class of all -singular modules is closed under module exten-
sions and -essential extensions.
From now on, we denote 0^^ by Z 2 (M ) for a right R-module M .

Theorem 4.8 Let R= (R) be non- -singular. Then:


(1) Z (M=Z (M )) = 0 for every right R-module M . Hence Z is a radical.
(2) A right R-module M is -singular if and only if HomR (M; C) = 0 for
every non- -singular module C.
(3) The class of all -singular modules is closed under module extensions and
-essential extensions. Hence the class of all -singular modules is a torsion
class.

13
Proof (1) Z (M ) E Z 2 (M ) by Proposition 4.4-(2). Suppose that
Z (M=Z (M )) = Z 2 (M )=Z (M ) 6= 0. Then there exists an element x 2
Z 2 (M )nZ (M ). Setting J = fr 2 R : xr 2 (M )g, we must have J 6E
RR , because x 2 = Z (M ). Thus there exists a right ideal K of R such that
J \K (RR ) but K 6 (RR ). Consider the map f : (K + (R))= (R) !
(xK + (M ))= (M ) dened by f (r + (R)) = xr + (M ) for all r 2 K. Then f is
an R-module isomorphism. Since R= (R) is non- -singular, (xK + (M ))= (M )
is also non- -singular. Thus Z ((xK + (M ))= (M )) = ((xK + (M ))= (M ))\
(Z (M )= (M )) = (Z (M ) \ (xK + (M )))= (M ) = 0 by Lemma 3.1. This
implies that Z (M ) \ (xK + (M )) (M ) and hence Z (M ) \ (xK + (M ))
(Z 2 (M )). Since Z (M ) E Z 2 (M ), we have xK + (M ) (M ) and hence
xK (M ). This implies that K J and hence K (RR ), a contradiction.
Thus Z (M=Z (M )) = 0.
(2) If M is -singular then, by Theorem 3.7-(1), HomR (M; C) = 0 for every
non- -singular right R-module C. Conversely assume that HomR (M; C) = 0
for every non- -singular right R-module C. According to part (1), M=Z (M )
is non- -singular. Hence the natural map M ! M=Z (M ) must be zero, and
so M = Z (M ).
(3) Let 0 ! C ! B ! A ! 0 be an exact sequence of right R-modules
with C, A -singular. According to part (2), we have HomR (C; M ) = 0 and
HomR (A; M ) = 0 for every non- -singular module M . By the exactness of the
sequence

0 ! HomR (A; M ) ! HomR (B; M ) ! HomR (C; M )

we obtain HomR (B; M ) = 0, and then part (2) shows that B is -singular.
Thus the class of all -singular modules is closed under module extensions.
Let N be a -singular right R-module and N E M . Then M=N is -singular
by Theorem 3.7-(2). Since M is an extension of N by M=N , the previous result
shows that M must be -singular.

Remark 1 Theorem 3.8-(1) shows that the class of all non- -singular right
R-modules is a torsionfree class for some torsion theory on M od-R. Denote
this torsion theory by . By Lemma 3.1-(4), is a generalization of , i.e.
. This process can be repeated for and we obtain the torsion theory
. In this way we obtain a chain of torsion theories 1 2 :::
Let be a stable torsion theory or RR be -torsionfree. Then every non- -
singular module is nonsingular by Lemma 3.1-(5). Thus is a generalization
of Goldie torsion theory, i.e. G .

Proposition 4.9 Let := (T ; F ) be the torsion theory in Remark 1.


Then:
(1) is a hereditary torsion theory on M od-R.
(2) If is a stable torsion theory or RR is -torsionfree then is a stable
torsion theory.

14
Proof (1) Theorem 3.7-(1) shows that is the torsion theory generated by
the class of all -singular modules. The class of all -singular modules is closed
under submodules and homomorphic images by Theorem 3.8-(2). Therefore
is a hereditary torsion theory on M od-R by [7, Corollary 1.2.12].
(2) Any generalization of Goldie torsion theory is stable by the proof of [18,
Proposition 1]. Now the result follows from Remark 1.

Theorem 4.10 Let := (T ; F ) be the torsion theory in Remark 1. Then


(M ) = Z 2 (M ) for a right R-module M .

Proof Let m 2 (M ). By Theorem 4.5, M=Z 2 (M ) 2 F . This implies


that HomR (mR; M=Z 2 (M )) = 0. Consider the map f : mR ! M=Z 2 (M ),
dened by f (mr) = mr + Z 2 (M ) for all r 2 R. Clearly f is an R-module
homomorphism. Since HomR (mR; M=Z 2 (M )) = 0, we have f = 0. This
implies that m 2 Z 2 (M ) and so (M ) Z 2 (M ).
Let N 2 F and h 2 HomR (Z 2 (M ); N ). Then hjZ (M ) = 0 by Theorem
3.7-(1). Since Z 2 (M )=Z (M ) is -singular, HomR (Z 2 (M )=Z (M ); N ) = 0.
Consider the map g : Z 2 (M )=Z (M ) ! N , dened by g(x + Z (M )) = h(x).
Since hjZ (M ) = 0, g is well-dened. g is clearly an R-module homomorphism
and so g must be zero. This shows that h(x) = 0 for every x 2 Z 2 (M ). Thus
HomR (Z 2 (M ); N ) = 0 for every N 2 F . This shows that Z 2 (M ) 2 T and
hence Z 2 (M ) (M ).

A module M is called an extending module if every closed submodule of


M is a direct summand of M . Various properties of extending modules can be
found in [10]. Following [1], we say that a module M is strongly -extending if
every -closed submodule of M is a direct summand of M . It is easily seen that
M is strongly -extending if and only if every submodule of M is -essential in
a direct summand of M . Note that in [8], a strongly -extending module was
called -extending module. But in this paper, we prefer the name "strongly
-extending" instead of " -extending".

Theorem 4.11 Let RR be strongly -extending and M be a right R-module.


Then the following are equivalent.
(1) M is non- -singular.
(2) M is -torsionfree and every cyclic submodule of M is projective and
extending.
(3) M is -torsionfree and every cyclic submodule of M is projective.

Proof (1) =) (2) Suppose that M is non- -singular. Then M is -torsionfree


by Lemma 3.1-(4). Let N be a cyclic submodule of M . Then there is a right
ideal I of R such that N ' R=I. Assume that I E J RR for some right ideal
J of R. Then J=I is -singular by Theorem 3.7-(2). Since N is non- -singular,
J=I is both -singular and non- -singular. This implies that I = J and hence
I is a -closed submodule of RR . Hence I is a direct summand of RR . Thus N
is isomorphic to a direct summand of RR . Therefore N is projective. N is also

15
strongly -extending by [8, Lemma 3.2-(2)]. This means that N is extending as
N is -torsionfree.
(2) =) (3) Clear.
(3) =) (1) For any m 2 Z (M ), mR is -torsionfree projective and isomor-
phic to R=ann(m). By Corollary 3.9, ann(m) E RR . Also ann(m) is a direct
summand of RR and so RR = ann(m) L for some L RR . This implies that
L (RR ) and hence mR (M ) = 0. Thus Z (M ) = 0.

We will also give a characterization of a strongly -extending module M via


Z 2 (M ). We need the following two results to give this characterization.
Lemma 4.12 [10, 7.5 Lemma] Let M1 and M2 be right R-modules and let
M = M1 M2 . Then M1 is M2 -injective if and only if for every submodule
N of M such that N \ M1 = 0 there exists a submodule M 0 of M such that
M = M1 M 0 and N M 0 .
Recall that any right R-modules fMi : i 2 Ig are called relatively injective
if Mi is Mj -injective for all distinct i; j 2 I.
Proposition 4.13 [8, Proposition 4.3] Let M1 and M2 be strongly -extending
modules. For each i 2 f1; 2g, let Pi be the submodule of Mi such that Mi =
(Mi ) Pi . If the right R-modules Pi (i 2 f1; 2g) are relatively injective then
M1 M2 is strongly -extending.
Theorem 4.14 A right R-module M is strongly -extending if and only if M =
(M ) K M 0 for some submodules K, M 0 of M such that Z 2 (M ) = (M ) K,
M 0 and Z 2 (M ) are both strongly -extending and K is M 0 -injective.
Proof ((=) M 0 is non- -singular by Theorem 4.5. Theorem 4.10 implies that
HomR (L; M 0 ) = 0 for every submodule L of K. Thus M 0 is K-injective. By
Proposition 4.13, M is strongly -extending.
(=)) Suppose that M is strongly -extending. By Corollary 4.6, Z 2 (M ) is
-closed in M . Hence Z 2 (M ) is a direct summand of M . By [8, Lemma 3.2],
Z 2 (M ) is strongly -extending. Since Z 2 (M ) is -saturated in M , (M ) =
(Z 2 (M )). By [8, Lemma 2.3], (M ) is -closed in Z 2 (M ). Hence (M ) is a
direct summand of Z 2 (M ). Thus M = (M ) K M 0 for some submodules
K; M 0 of M . M 0 and K M 0 are also strongly -extending by [8, Lemma 3.2].
Let N be a submodule of K M 0 such that N \ K = 0. There exist submodules
L1 , L2 of K M 0 such that K M 0 = L1 L2 and N E L1 . Since K M 0
is -torsionfree, we have N E L1 . Clearly L1 \ K = 0 and hence K L2 . It
follows that K M 0 = L1 L2 = K L1 (L2 \ M 0 ) and N L1 (L2 \ M 0 ).
By Lemma 4.12, K is M 0 -injective.

Acknowledgement
The second author was supported by the Scientic Research Project Admin-
istration of Akdeniz University.
The authors would like to thank the referee for his/her valuable comments
and suggestions which improved the presentation of this work.

16
References
[1] Albu, T. The Osofsky-Smith Theorem for Modular Lattices, and Applica-
tions (II), Communications in Algebra 42 (6), 2014, 2663-2683.

[2] Albu, T. and Iosif, M. On socle and radical of modular lattices, Ann. Univ.
Buchar. Math. Ser. 5 (LXIII) (2014), 187194.
[3] Albu, T. and Iosif, M. Lattice preradicals with applications to Grothendieck
categories and torsion theories, J. Algebra 444 (2015), 339366.

[4] Albu, T and Iosif, M. Modular C11 lattices and lattice preradicals,
http://imar.ro/~dbeltita/IMAR_preprints/2015/2015_2.pdf
[5] Asgari, Sh. and Haghany, A. Ranks of modules relative to a torsion theory,
Bulletin of the Iranian Mathematical Society, 37 (4), 2011, 19-33.
[6] Bland, P. E. Topics in Torsion Theory. Berlin. Wiley-VCH Verlag 1998.
[7] Crivei, S. Injective Modules Relative To Torsion Theories, Editura Fun-
datePentru StudEuropene, Cluj-Napoca (2004).
[8] eken, S. and Alkan, M. On -extending modules, Mediterr. J. Math. 9
(1), 2012, 129-142.
[9] eken, S. and Alkan, M. UC Modules With Respect to a Torsion Theory,
Turkish Journal of Mathematics, 36 (3), 2012, 376-385.
[10] Dung, N. V., Huynh, D. V., Smith, P. F., Wisbauer, R. (1994), Extending
Modules, Pitman Research Notes in Mathematics 313. Harlow: Longman.
[11] Golan, J. S. Torsion Theories. Pitmann Mon.and Surveys in Pure and
Appl.Math. 29, 1986.
[12] Goldie, A.W. Torsion-Free Modules and Rings, Journal of Algebra 1, 1964,
268-287.
[13] Gomez Pardo, J. L. Spectral Gabriel topologies and relative singular func-
tors. Communications in Algebra 13 (1), 1985, 2157.
[14] Goodearl, K.R. Ring Theory: Nonsingular Rings and Modules, Marcel-
Dekker, 1976.
[15] Lam, T.Y. Lectures on Modules and Rings, Springer-Verlag, New York,
1999.
[16] Smith, P.F. Modules for which every submodule has unique closure. Pro-
ceedings of the Biennial Ohio-Denison Conference, 1992, 302-313.
[17] Stenstrm, B. Rings of quotients, Springer-Verlag, 1975.

17
[18] Teply, Mark L. Torsionfree projective modules. Proc. Amer. Math. Soc. 27,
1971, 29-34.

Seil EKEN, Department of Mathematics, Istanbul Aydn University, De-


partment of Mathematics-Computer, Istanbul, Turkey, cekensecil@gmail.com.
Mustafa ALKAN, Akdeniz University Department of Mathematics, Antalya,
Turkey, alkan@akdeniz.edu.tr

18

Вам также может понравиться