Content uploaded by Hwankoo Kim
Author content
All content in this area was uploaded by Hwankoo Kim on Dec 24, 2021
Content may be subject to copyright.
Bull. Korean Math. Soc. 00 (20XX), No. 0, pp. 1–0
https://doi.org/10.4134/BKMS.b210155
pISSN: 1015-8634 / eISSN: 2234-3016
THE CLASS OF WEAK w-PROJECTIVE MODULES
IS A PRECOVER
Hwankoo Kim, Lei Qiao, and Fanggui Wang
Abstract. Let Rbe a commutative ring with identity. Denote by wPw
the class of weak w-projective R-modules and by wPw
⊥the right orthog-
onal complement of wPw. It is shown that (wPw,wPw
⊥) is a hereditary
and complete cotorsion theory, and so every R-module has a special weak
w-projective precover. We also give some necessary and sufficient con-
ditions for weak w-projective modules to be w-projective. Finally it is
shown that when we discuss the existence of a weak w-projective cover
of a module, it is enough to consider the w-envelope of the module.
1. Introduction
Throughout this paper Ris always a commutative ring with identity. We
first review some related concepts of w-modules. A finitely generated ideal J
of Ris called a GV-ideal if the homomorphism R→HomR(J, R) induced by
the inclusion map J →Ris an isomorphism. Denote by GV(R) the set of
GV-ideals of R. For any R-module N, set
torGV(R)(N) = {x∈N|there exists J∈GV(R) such that Jx = 0},
which is a submodule of N, called the total GV-torsion submodule of N. If
torGV(R)(N) = N, then Nis called a GV-torsion module; if torGV(R)(N) = 0,
then Nis called a GV-torsion-free module. A GV-torsion-free module Nis
called a w-module if Ext1
R(R/J, N ) = 0 for any J∈GV(R). Denote by Wthe
class of w-modules. The set of maximal w-ideals of Ris denoted by w-Max(R).
By [10, Theorem 6.2.15], an R-module Tis a GV-torsion module if and only if
Tm= 0 for any m∈w-Max(R).
We also need the concept of strong w-modules. An R-module Nis called
astrong w-module if Extk
R(T, N ) = 0 for any GV-torsion module Tand any
k>1. For a discussion of strong w-modules, please refer to [12]. Denote by
W∞the class of strong w-modules.
Received February 21, 2021; Accepted October 14, 2021.
2010 Mathematics Subject Classification. 13C10, 13D05, 13D07, 13D30.
Key words and phrases. Weak w-projective precover, w-operation (theory), cotorsion
theory.
c
20XX Korean Mathematical Society
1
$KHDGRI3ULQW
2 H. KIM, L. QIAO, AND F. WANG
Since the w-operation on an integral domain can establish the concept of w-
modules, which allows the w-operation to work in the category of modules, in
1997 the concepts of w-projective modules and w-flat modules over an integral
domain were first introduced [8]. In [4] the definition of w-flat modules was
extended to any commutative ring as follows. A module Mis called a w-flat
module if the functor M⊗ − preserves a w-exact sequence into a w-exact se-
quence. In [11] the concepts of the w-flat dimension (w-fd) of a module and the
w-weak global dimension (w-w.gl.dim) of a ring have been successively intro-
duced. Using the w-weak global dimension of a ring, a Pr¨ufer v-multiplication
domain (PVMD for short) can be characterized homologically as an integral
domain of w-w.gl.dim(R)61.
In 2015, the concept of w-projective modules was also extended to any com-
mutative ring [9]. Let Mbe an R-module. Set L(M) := (M/torGV(R)(M))w.
Then Mis called a w-projective module if Ext1
R(L(M), N ) is a GV-torsion mod-
ule for any w-module N. Denoted by Pwthe class of w-projective modules.
One can use the w-projective modules to introduce the w-projective ideals. One
hopes that some rings that used to be described by ideals can be described by
the w-projective modules. For example, in [11] it is proved that an integral
domain Ris a PVMD if and only if every finitely generated submodule of a
projective module is w-projective. As we all know, an integral domain Ris a
Dedekind domain if and only if each nonzero ideal is invertible; Ris a Krull
domain if and only if each nonzero ideal is w-invertible. Therefore, in the above
sense, Krull domains can actually be considered as w-Dedekind domains. But a
Dedekind domain is exactly an integral domain with global dimension at most
1, in other words, every submodule of a projective module is projective. In
[14], the authors can only prove that an integral domain Ris a Krull domain
if and only if every submodule of a finitely generated projective module is w-
projective. That is to say, the concept of w-projective modules cannot be used
to obtain a complete characterization of the Krull domains corresponding to
the Dedekind domains.
In order to give a complete homological characterization of Krull domains,
the concept of weak w-projective modules is introduced in [12] with the aid of
w-projective modules. Denote by RMthe category of all R-modules. Set
Pw†∞=N∈M
Nis GV-torsion-free and
Extk
R(M, N ) = 0 for any M∈ Pwand any k>1.
An R-module Mis called a weak w-projective module if Ext1
R(M, N )=0
for any N∈ Pw†∞. Denote by wPwthe class of weak w-projective modules.
In [12] the authors pointed out: Every w-pro jective module must be weak w-
projective. Conversely, every weak w-projective module of finite type and any
weak w-projective ideal of an integral domain are all w-projective. At the same
time, in [12] it is also given an example of a weak w-projective module over
a UFD, which is not w-projective. In [12] it is also introduced the concept
of the w-projective dimension (w-pd) of a module and the global w-projective
$KHDGRI3ULQW
WEAK w-PROJECTIVE MODULES 3
dimension (w-gl.dim) of a ring. With the help of the concepts of weak w-
projective modules and the global w-projective dimension of a ring, in [12]
the authors give a homological characterization of Krull domains: An integral
domain Ris a Krull domain if and only if every submodule of a projective
module is weak w-projective, equivalently, w-gl.dim(R)≤1.
Let Abe a class of modules, Mbe an R-module, A∈ A, and ϕ:A→Mbe
a homomorphism. Then (A, ϕ) is called an A-precover of Mif for any A0∈ A
and any homomorphism f:A0→M, the following diagram
A0
h
xx
f
Aϕ//M
is commutative, equivalently, for any A0∈ A, HomR(A0, A)ϕ∗
→HomR(A0, M )→
0 is an exact sequence. Let (A, ϕ) be an A-precover of a module M. When
A0=A,f=ϕ, and the above diagram is commutative, it is said that (A, ϕ) is
an A-cover of Mif his an isomorphism. If any R-module Mhas A-precover
(resp., cover), then we say that Ais a precover (resp, cover) class.
Let Sbe a class of modules. Set
⊥S:= {A∈M|Ext1
R(A, C) = 0 for any C∈ S}
and
S⊥:= {B∈M|Ext1
R(C, B) = 0 for any C∈ S},
are called the left orthogonal complement and the right orthogonal complement
of S, respectively [3]. Then obviously one has wPw=⊥(P†∞
w). In [1], the
authors introduced and studied the right orthogonal complement of the class
of w-flat modules. Also set
⊥∞S:= {A∈M|Extk
R(A, C) = 0 for any C∈ S and any k>1},
and
S⊥∞:= {B∈M|Extk
R(C, B) = 0 for any C∈ S and any k>1}.
In recent years, the cotorsion theory has received great attention from re-
searchers. Let A,Bbe two classes of modules. Then (A,B) is called a cotorsion
theory if B=A⊥and A=⊥B. In addition, (A,B) is called a hereditary co-
torsion theory if whenever 0 →A1→A→A2→0 is exact with A, A2∈ A,
one has A1∈ A. And (A,B) is called a complete cotorsion theory if for any
R-module M, there is an exact sequence 0 →K→A→M→0 with A∈ A
and K∈ B. When a formulated pair (A,B) of modules becomes a cotorsion
pair, the classical homology method can be used very smoothly to characterize
rings and modules. For the projective modules, a well-known theorem of Ka-
plansky states that a projective module over an arbitrary ring is a direct sum of
countably generated projective modules. In 2020, Wang and Qiao established
the w-version of Kaplansky’s theorem [13]: If Mis a w-projective w-module,
then Mhas a w-projective w-ℵ0-continuous ascending chain (see the definition
$KHDGRI3ULQW
4 H. KIM, L. QIAO, AND F. WANG
later). Using this result, this article obtains the main result: (wPw,wPw⊥) is
a hereditary and complete cotorsion theory, and so every module has a special
weak w-projective precover.
2. Basic results
Denoted by FT the class of GV-torsion-free modules. Let Sbe a class of
modules. Define:
S†:= S⊥∩ FT
={N∈M|Nis GV-torsion-free and Ext1
R(M, N ) = 0 for any M∈ S}
and
S†∞:= S⊥∞∩ FT
=N∈M
Nis GV-torsion-free and
Extk
R(M, N ) = 0 for any M∈ S and any k>1.
Set
GV(R)∗:= {R/J |J∈GV(R)}.
Obviously GV(R)∗is a set of modules.
Proposition 2.1. Let S,S1be classes of modules. Then:
(1) S ⊆ ⊥(S†∞)⊆⊥(S†).
(2) If S ⊆ S1, then S†
1⊆ S†and S†∞
1⊆ S†∞.
(3) (S ∪ S1)†=S†∩ S †
1.
Proof. These are obvious.
For k>1, set
Wk:= {N∈ FT | Exti
R(R/J, N ) = 0 for any J∈GV(R) and any 1 6i6k}.
By convention, we set W0:= FT . A module Nis called a wk-module if
N∈ Wk. It is known that a GV-torsion-free module Nis a w-module if and
only if Ext1
R(C, N ) = 0 for any GV-torsion-module C([10, Theorem 6.2.7]).
Lemma 2.2. (1) If 16i6k, then Wk⊆ Wi.
(2) Wkis closed under extensions.
(3) Let N∈ Wk. Then N∈ Wk+1 if and only if Extk+1
R(M, N )=0for
any GV-torsion module M.
Proof. (1) and (2) are trivial. We will prove only (3). It is enough to show the
necessity. Assume that N∈ Wk+1 . If k= 0, then N∈ W1=W. Thus by
[10, Theorem 6.2.7], Ext1
R(M, N ) = 0 for any GV-torsion module M. Consider
the case k= 1. Let Mbe a GV-torsion module. Then for any x∈M, there
exists Ix∈GV(R) such that Ixx= 0. Set F:= L
x∈M
R/Ix. Then Fis a GV-
torsion module. Let exdenote the element in Fthat takes the value 1+Ixat the
component x, and the other components take the value 0. Define h:F→M
$KHDGRI3ULQW
WEAK w-PROJECTIVE MODULES 5
by h(ex) = x. Then his an epimorphism. Set A:= Ker(h). Then it follows
from the exact sequence 0 = Ext1
R(A, N )→Ext2
R(M, N )→Ext2
R(F, N )=0
that Ext2
R(M, N ) = 0. Now the assertion follows by induction.
Proposition 2.3. The following are equivalent for a GV-torsion-free module
N.
(1) N∈ W∞.
(2) Exti
R(R/J, N )=0for any J∈GV(R)and any i>1.
Proof. (1) ⇒(2) This is trivial.
(2) ⇒(1) Let k>1 and set
W0
k:= N∈ FT
Exti
R(M, N ) = 0 for any GV-torsion module M
and any 1 6i6k.
By Lemma 2.2, W0
k=Wk. Thus N∈T∞
k=1 W0
k=W∞.
Let Mand Nbe R-modules. A homomorphism f:M→Nis called a w-
monomorphism (resp., a w-epimorphism, a w-isomorphism) if fm:Mm→Nm
is a monomorphism (resp., an epimorphism, an isomorphism) for any maximal
w-ideal mof R. And Mis said to be w-isomorphic to Nprovided that there
exist an R-module Land two w-isomorphisms f:L→Mand g:L→N.
Theorem 2.4. Let Sbe a class of modules such that S ⊆ F T . Set A:= ⊥S.
Then the following are equivalent.
(1) Ais closed under w-isomorphisms.
(2) GV(R)∪GV(R)∗⊆ A.
(3) S ⊆ W2.
Proof. (1) ⇒(2) Let J∈GV(R). Since R∈ A and Jand Rare w-isomorphic,
it follows that J∈ A. Also since R/J and 0 are w-isomorphic, it follows that
R/J ∈ A.
(2) ⇒(3) Let N∈ S. Then Ext1
R(R/J, N ) = 0 and Ext1
R(J, N ) = 0 for any
J∈GV(R). Thus Nis a w2-module. Therefore S ⊆ W2.
(3) ⇒(1) Let f:M→M0be a w-isomorphism. By [10, Proposition 6.3.4],
there exist a module Band exact sequences 0 →A→M→B→0 and
0→B→M0→C→0, where Aand Care GV-torsion modules. If M∈ A,
then for any N∈ S it follows from the exact sequence 0 = HomR(A, N )→
Ext1
R(B, N )→Ext1
R(M, N ) = 0 that Ext1
R(B, N ) = 0. Again by the exact
sequence 0 = Ext1
R(C, N )→Ext1
R(M0, N )→Ext1
R(B, N ) = 0 it follows that
Ext1
R(M0, N ) = 0, that is, M0∈ A.
On the other hand, assume that M0∈ A. By Lemma 2.2, Ext2
R(C, N ) = 0.
By the exact sequence 0 = Ext1
R(M0, N )→Ext1
R(B, N )→Ext2
R(C, N ) = 0 it
follows that Ext1
R(B, N ) = 0. Also by the exact sequence 0 = Ext1
R(B, N )→
Ext1
R(M, N )→Ext1
R(A, N ) = 0, it follows that Ext1
R(M, N ) = 0, i.e., M∈ A.
Therefore Ais closed under w-isomorphisms.
$KHDGRI3ULQW
6 H. KIM, L. QIAO, AND F. WANG
Corollary 2.5. Let Sbe a class of modules. Set A:= ⊥S. If S ⊆ W∞, then
Ais closed under w-isomorphisms.
Proof. This follows directly from Theorem 2.4 and the fact that W∞⊆ W2.
Example 2.6. (1) It is easy to see that (GV(R)∗)†=W.
(2) By Proposition 2.3, (GV(R)∗)†∞=W∞.
(3) By Theorem 2.4, (GV(R)∗∪GV(R))†=W2.
Proposition 2.7. Let Sbe a class of modules satisfying GV(R)∗⊆ S. Then:
(1) S†⊆ W and S†∞⊆ W∞.
(2) If GV(R)⊆ S, then S†⊆ W2.
Proof. This follows immediately from Example 2.6.
3. The class of weak w-projective modules is a precover
Let Abe a class of modules and Mbe an R-module. If there is a continuous
ascending chain of submodules of M:
(3.1) 0 = M0⊆M1⊆ · · · ⊆ Mα⊆Mα+1 ⊆ · · · ⊆ Mλ=M
such that Mα+1/Mα∈ A for any α < λ, then Mis called an A-filtered module.
A continuous ascending chain (3.1) is called an A-filtration of M.
In order to determine when (S,S⊥) is a complete cotorsion theory, the fol-
lowing lemma is very effective and will be used later.
Lemma 3.1 (Eklof–Trlifaj).Let Sbe a set of modules. Then:
(1) Let Nbe an R-module. Then there exists a short exact sequence 0→
N→Q→A→0, where Q∈ S⊥and Ais an S-filtered module, and
thus A∈⊥(S⊥).
(2) (⊥(S⊥),S⊥)is a complete cotorsion theory.
Proof. See [2] or [7, Theorem 2.2].
In order to make Lemma 3.1 apply to the context of a class of related mod-
ules, we make corresponding modifications to it, but note that the idea belongs
to Eklof–Trlifaj essentially.
Lemma 3.2. Let S= GV(R)∗∪ S1be a set of modules, where S1⊆ FT .
(1) Let Nbe a GV-torsion-free module. Then there exists an exact sequence
(3.2) 0 →N→Q→A→0,
where Q∈ S†and Ais an S-filtered module such that A∈⊥(S†).
(2) Let Mbe an R-module. Then there exists an exact sequence
(3.3) 0 →B→P→M→0,
where P∈⊥(S†)and B∈ S†.
$KHDGRI3ULQW
WEAK w-PROJECTIVE MODULES 7
Proof. (1) Set X:= L
S∈S1
Sand Y:= L
J∈GV(R)
R/J. Then Xis a GV-torsion-
free module and Yis a GV-torsion module. Set S=X⊕Y. Then S⊥={S}⊥.
Thus we may assume that Sis the class of modules composed of the fixed
module Sand its direct sums. Let 0 →K1
µ1
−→ F1→X→0 and 0 →K2
µ2
−→
F2→Y→0 be exact sequences, where F1and F2are free modules. Set
F:= F1⊕F2and K:= K1⊕K2. Then 0 →Kµ
−→ F→S→0 is an exact
sequence, where µ:= µ1⊕µ2. Since Xis GV-torsion-free, K1is a w-module.
Since Yis GV-torsion, we have (K2)w=F2
Take a regular cardinal λso that Khas a generating system Zwith |Z|< λ.
Set Q0:= N. Then Q0is GV-torsion-free. For α < λ, if Qαhas been
constructed, select a free module F0
αand an epimorphism δα:F0
α→Qα. Set
Iα:= HomR(K, Qα) to be a new index set and define µα:K(Iα)→F(Iα)
as the homomorphism of direct sums, which is induced by µ. Then µαis a
monomorphism and Coker(µα) = S(Iα).
Define ϕα:K(Iα)⊕F0
α= ( L
f∈Iα
Kf)⊕F0
α→Qα, where Kf=K, by
ϕα([uf], z) = P
f∈Iα
f(uf) + δα(z), where uf∈Kf, z ∈F0
α. Since δαis an
epimorphism, so is ϕα. In addition, for any f∈Iα, let if:K→K(Iα)and
jf:F→F(Iα)be the natural imbeddings. Then one has
(3.4) f=ϕαifand jfµ=µαif.
Now assume that if β6α, then Qβhas been constructed (if αis a limit
ordinal, set Qα=S
β<α
Qβ), in particular, Qαhas been constructed. Construct
the following pushout diagram:
0//K(Iα)⊕F0
α
µα⊕1//
ϕα
F(Iα)⊕F0
α//
ψα
S(Iα)//
∼
=
0
0//Qα
hα//Qα+1 //Qα+1/Qα//0
One gets Qα+1. At this time ψαis an epimorphism. As you can see from the
above diagram, if Qαis a GV-torsion-free module, then Ker(ψα) = Ker(ϕα)
is a w-module, and thus Qα+1 is also a GV-torsion-free module. Hence by a
transfinite induction, we see that each Qαis a GV-torsion-free module.
Set Q:= S
α<λ
Qα= lim
→
α<λ
Qα. Then Qis a GV-torsion-free module. Set
A:= Q/N and Aα:= Qα/N . Then Aα+1/Aα∼
=Qα+1/Qα∼
=S(Iα). Since
Q=S
α<λ
Qα, one gets that A=S
α<λ
Aα. Thus Ais an S-filtered module, and
thus one has A∈⊥(S⊥). Since S†⊆ S⊥, one has A∈⊥(S†).
Let us prove that Q∈ S⊥. For this, it is sufficient to prove that µ∗:
HomR(F, Q)→HomR(K, Q) is an epimorphism. Let g:K→Qbe a ho-
momorphism. Since the generating system Zof Ksatisfies |Z|< λ and Q=
$KHDGRI3ULQW
8 H. KIM, L. QIAO, AND F. WANG
S
α<λ
Qα, there exists an ordinal α < λ such that Im(g)⊆Qα. Thus there exists
a homomorphism f:K→Qαsuch that g(x) = f(x) for any x∈K. By the
pushout diagram above and (3.4), one has ψαjfµ=ψαµαif=hαϕαif=hαf.
Define σ:F→Qby σ(z) = ψαjf(z)∈Qα+1 ⊆Q. Then one can ver-
ify directly that g=σµ =µ∗(σ). Thus µ∗is an epimorphism. Therefore
Q∈ S⊥∩ F T =S†.
(2) Take an exact sequence 0 →N→F→M→0, where Fis a projec-
tive module. Then Nis a GV-torsion-free module. By (1), there is an exact
sequence 0 →N→Q→A→0, where Q∈ S†and A∈⊥(S†). Consider the
following commutative diagram with two exact rows:
0
0
0//N//
F
//M//0
0//Q//
P//
M//0
A
A
0 0
where the square diagrams in the upper left and lower corners are pushout
diagrams. Since F , A ∈⊥(S†), one has P∈⊥(S†). Therefore one gets the
exact sequence (3.3) by taking B:= Q.
Let Abe a class of modules. Then an A-precover f:C→Mof Mis said
to be special if fis surjective and Ker(f)∈ A⊥. In other words, there is an
exact sequence 0 →K→C→M→0 with C∈ A and K∈ A⊥.
Theorem 3.3. Let S= GV(R)∗∪S1be a set of modules, where S1⊆ FT . Set
A:= ⊥(S†). If Ais closed under w-isomorphisms, then (A,A⊥)is a complete
cotorsion theory.
Proof. Note that (A,A⊥) is the cotorsion theory generated by S†. Let us prove
that any module Mhas a special A-precover.
By Lemma 3.2, there is an exact sequence (3.3), where P∈ A and B∈ S†⊆
⊥(S†)⊥=A⊥. Therefore Mhas a special A-precover.
Proposition 3.4. Let Sbe a class of modules such that GV(R)∗⊆ S. Set
B:= ⊥(S†∞). Then:
(1) S†∞is closed under direct products, direct summands, and cokernels of
monomorphisms.
(2) Bis closed under direct sums, direct summands, kernels of epimor-
phisms, and w-isomorphisms.
(3) B†=B†∞=S†∞.
$KHDGRI3ULQW
WEAK w-PROJECTIVE MODULES 9
Proof. (1) Obviously S†∞is closed under direct products and direct summands.
Obviously S⊥∞is closed under cokernels of monomorphisms. By [12, Propo-
sition 2.2(2)], W∞is also closed under cokernels of monomorphisms. Since
S†∞=S⊥∞∩ W∞,S†∞is closed under cokernels of monomorphisms.
(2) Obviously Bis closed under direct sums and direct summands. By (1),
Bis closed under kernels of epimorphisms. By Corollary 2.5, Bis closed under
w-isomorphisms.
(3) Obviously we have that S†∞⊆⊥(S†∞)⊥∩ FT =B†. Since Bis
closed under kernels of epimorphisms, we have B⊥∞=B⊥. Thus we have
B†=B⊥∞∩ FT =B†∞. Since S ⊆ B, it follows that B†=B†∞⊆ S†∞.
Therefore B†=S†∞.
Let Mbe an R-module. Then Mis said to be w-ℵ0-generated if there exist
a countably generated free module Fand a w-epimorphism φ:F→M.
Let Mbe a w-projective w-module. If there is a continuous ascending chain
of w-projective w-submodules of M:
0 = M0⊆M0
1⊆M0
2⊆ · · · ⊆ M0
α⊆ · · · ⊆ M0
λ=M
such that each factor M0
α+1/M0
αis a w-ℵ0-generated w-projective module, then
it is said that Mhas a w-projective w-ℵ0-continuous ascending chain. It follows
from [13, Theorem 3.5] that if Mis a w-projective w-module, then Mhas a
w-projective w-ℵ0-continuous ascending chain.
Proposition 3.5. (1) wPw†=P†∞
w.
(2) Let S= GV(R)∗∪ S1be a set of modules, where S1is the class of
w-projective w-ℵ0-generated w-modules. Then S†∞=P†∞
w.
(3) Let S= GV(R)∗∪ S1be a set of modules, where S1={R}. Then
S†∞=W∞.
Proof. (1) This follows immediately from Proposition 3.4 by setting S:= Pw.
(2) Since S ⊆ Pw, we have P†∞
w⊆ S†∞. Let N∈ S †∞. For any w-
projective w-module P, by [13, Theorem 3.5] Pis an S1-filtered module. Thus
Exti
R(P, N ) = 0 for any i>1. By Proposition 2.7, Nis a strong w-module. Let
Pbe a w-projective module. Then one has the following two exact sequences:
0→torGV(R)(P)→P→P/torGV(R)(P)→0
and
0→Q→Qw→Qw/Q →0,
where Q:= P/torGV(R)(P) is GV-torsion-free. Considering two long exact se-
quences induced by the above two exact sequences, it follows that Exti
R(P, N ) =
0 for any w-projective module Pand any i>1 since Pwis closed under w-
isomorphisms. Thus N∈ P⊥∞
w∩ FT =P†∞
w. Therefore S†∞=P†∞
w.
(3) This is trivial.
$KHDGRI3ULQW
10 H. KIM, L. QIAO, AND F. WANG
Theorem 3.6. Let S= GV(R)∗∪ S1be a set of modules, where S1⊆ FT .
Set B:= ⊥(S†∞). Then (B,B⊥)is a hereditary and complete cotorsion theory.
Proof. For any M∈ S, fix a projective resolution of M. Let LMbe the set
of all syzygies of this projective resolution of M(including Mitself as −1
syzygy). Set L:= S
M∈S
LM. Then Lis again a set. Note that Lcan be split
into L= GV(R)∗∪ L1, where L1is the set of all syzygies of M∈ S1and all
non-negative syzygies of R/J ∈GV(R)∗. Then L1⊆ F T .
Let N∈ S⊥∞. For any X∈ L, there exists an exact sequence
(3.5) 0 →X→Pk→ · · · → P1→P0→M→0,
where each Piis a projective module and M∈ S. Thus one has Ext1
R(X, N )∼
=
Extk+2
R(M, N ) = 0. Therefore N∈ L⊥.
On the other hand, let N∈ L⊥. For any Y∈ S and any k>−1, by
considering the exact sequence (3.5), one has Extk+2
R(Y, N )∼
=Ext1
R(X, N ) = 0.
Thus N∈ S⊥∞. Therefore L⊥=S⊥∞. By Theorem 3.3, (B,B⊥) is a complete
cotorsion theory. It follows by Proposition 3.4 that (B,B⊥) is a hereditary
cotorsion theory.
Now we are ready to state the main theorem.
Theorem 3.7. (wPw,wPw⊥)is a hereditary and complete cotorsion theory,
and so every module has a special weak w-projective precover.
Proof. Let S1be the collection of all w-ℵ0-generated w-projective w-modules
and set S:= GV(R)∗∪ S1. Since the collection of all ℵ0-generated modules
is a set, Sis also a set. By Proposition 3.5(2), S†∞=W∞. By Theorem 3.6,
(wPw,wPw⊥) is a hereditary and complete cotorsion theory.
According to [5, 6], we say that a module Mis a w∞-projective module if
Ext1
R(M, N ) = 0 for any strong w-module N. Denote by Pw∞the class of
w∞-projective modules. Then Pw∞=⊥W∞.
Theorem 3.8. (Pw∞,P⊥
w∞)is a hereditary and complete cotorsion theory.
Proof. Set S1:= {R}and S:= GV(R)∗∪ S1. Then Sis a set of modules. By
Proposition 3.5, S†∞=W∞. Thus Pw∞=⊥(S†∞). Now the assertion follows
by Theorem 3.6.
Proposition 3.9. Let Mbe a w-module. Then there is a special weak w-
projective precover of M,ϕ:P→Msuch that Pis a w-module and Ker(ϕ)∈
P†∞
w.
Proof. We use the notation Las in the proof of Theorem 3.6 and the notation
Sas in Proposition 3.5(2). Then L†=S†∞=P†∞
w. Now the assertion follows
by Theorem 3.6.
$KHDGRI3ULQW
WEAK w-PROJECTIVE MODULES 11
Recall that a class of modules is said to be hereditary if it is closed under
isomorphic copies and submodules.
Lemma 3.10. If Pwis a hereditary class of modules, then wPw†=Pw†.
Proof. If Pwis a hereditary class of modules, then P⊥
w=P⊥∞
w, and thus
P†
w=P†∞
w. Now the assertion immediately follows by applying Proposition
3.5(1).
In the following result, we give some necessary and sufficient conditions for
weak w-projective modules to be w-projective.
Theorem 3.11. The following conditions are equivalent for a ring R:
(1) Every weak w-projective module is w-projective.
(2) Every weak w-projective w-module is w-projective.
(3) (Pw,Pw⊥)is a hereditary cotorsion theory and every w-module has a
special Pw-precover of a w-module.
Proof. (1)⇒(3) This follows by Theorem 3.7 and Proposition 3.9.
(3)⇒(2) Let Mbe a weak w-projective w-module. By assumption, there
is an exact sequence 0 →A→P→M→0 such that Pis a w-projective
w-module and A∈ Pw⊥. Since any GV-torsion module is w-projective, Ais a
w-module. By Lemma 3.10, A∈ Pw†=Pw†∞. Thus Ext1
R(M, A) = 0, and so
the above exact sequence is split. Therefore Mis a w-projective module.
(2)⇒(1) Let Mbe a weak w-projective module. It follows from [12, Corollary
2.7] that L(M) is a weak w-projective module. By assumption, L(M) is a w-
projective module. So Mis a w-projective module.
Proposition 3.12. Let Abe a class of modules which is closed under w-
isomorphisms. Let Mbe a GV-torsion-free module and ϕ:P→Mbe an
A-cover. Then:
(1) Pis a GV-torsion-free module.
(2) If ϕis a special A-cover and Mis a w-module, then Pis a w-module.
Proof. Set T:= torGV (P) and B:= P /T . Then Bis a GV-torsion-free module.
Let π:P→Bbe a natural homomorphism. Since Mis a GV-torsion-free
module, ϕinduces a homomorphism ψ:B→Msuch that ψ(x) = ϕ(x) for
any x∈F, that is ψπ =ϕ. Since Ais closed under w-isomorphisms, it follows
that B∈ A. Thus there is a homomorphism h:B→Psuch that ϕh =ψ. So
ϕhπ =ψπ =ϕ. Hence hπ is an isomorphism, and thus πis an isomorphism.
Therefore Pis a GV-torsion-free module.
(2) By (1), A:= Ker(ϕ) is also a GV-torsion-free module. Since Ais closed
under w-isomorphisms, Acontains all GV-torsion modules. So Ais a w-module.
It follows from the exact sequence 0 →A→P→M→0 that Pis a w-
module.
Theorem 3.13. Let Abe a class of modules closed under w-isomorphisms.
Let Mbe a GV-torsion-free module. Then Mhas a special A-cover if and only
$KHDGRI3ULQW
12 H. KIM, L. QIAO, AND F. WANG
if Mwhas a special A-cover. In addition, if Mis GV-torsion-free and Bis a
special A-cover of M, then Bwis a special A-cover of Mw.
Proof. Let ϕ:P→Mwbe an A-cover of Mw. Set T:= Mw/M. Then
Tis a GV-torsion module. Let π:Mw→Tbe a natural homomorphism.
Set g:= πϕ,A:= Ker(ϕ), and B:= Ker(g). Then one has the following
commutative diagram with exact rows and columns:
0
0
A
A
0//B//
ϕ0
Pg//
ϕ
T//0
0//M//
Mw
π//
T//0
0 0
where ϕ0=ϕ|B. It follows that ϕ0:B→Mis a special A-precover of M.
Let h:B→Bbe a homomorphism such that ϕ0h=ϕ0. By [10, Theorem
6.3.2], hcan be extended only to a homomorphism h0:P→P. So ϕh0is
an extension of ϕ0h. Again by [10, Theorem 6.3.2], ϕh0=ϕ. So h0is an
isomorphism. Thus his a monomorphism.
Let x∈B. Then there is y∈Psuch that h0(y) = x. So gh0(y) = πϕh0(y) =
πϕ(y) = g(y). Therefore b:= y−h0(y) = y−x∈Ker(g) = B. So y=b+x∈B,
which results in x=h(y). Thus his an epimorphism. So his an isomorphism,
and thus ϕ0:B→Mis an A-cover of M.
Conversely, let α:B→Mbe an A-cover of Mand P:= Bw. It follows
from Proposition 3.12(1) that Bis a GV-torsion-free module. By [10, Theorem
6.3.2], αinduces a unique homomorphism ϕ:P→Mw. Set T:= P /B and
T2:= Mw/M. Then Tand T2are GV-torsion modules. Thus one has the
following commutative diagram with two exact rows:
0//B//
α
Pπ//
ϕ
T//
β
0
0//M//Mw
π1//T2//0
Set A:= Ker(α), D:= Ker(ϕ), and T1:= Ker(β). It follows from the snake
lemma that one has the following exact sequence: 0 →A→D→T1→0.
Because A∈ A⊥, one has Ext1
R(T1, A) = 0. Thus D∼
=A⊕T1. Since Dis GV-
torsion-free, it follows that T1= 0, and so D=A. Since αis an epimorphism,
ϕis also an epimorphism, and thus βis an isomorphism. Hence ϕis a special
A-precover of Mw.
$KHDGRI3ULQW
WEAK w-PROJECTIVE MODULES 13
Now let h:P→Pbe a homomorphism such that ϕh =ϕ. Consider the
following diagram with exact two rows:
0//B//
h0
Pπ//
h
T//0
0//B//Pπ//T//0
Then πh =β−1π1ϕh =β−1π1ϕ=π, and so the square diagram on the right is a
commutative diagram. Thus h0:B→Bmakes the left square a commutative
diagram. Since αis the restriction of ϕon B, one has αh0=α. So h0is
an isomorphism, and thus his an isomorphism. Therefore ϕis an A-cover of
Mw.
Proposition 3.14. Let Abe a class of modules which is closed under w-
isomorphisms. Let Mbe an R-module and set T:= torGV(M). If ϕ:P→
M/T is a special A-cover which makes the pul lback diagram:
0//Tλ//P1
β//
α
P//
ϕ
0
0//T//Mπ//M/T //0,
then α:P1→Mis a special A-cover.
Proof. Because P1is w-isomorphic to P, one has P1∈ A. Set A:= Ker(ϕ).
Since Ker(α)∼
=A, it follows that α:P1→Mis a special A-precover. Let
h:P1→P1be a homomorphism such that αh =α. It follows from Proposition
3.12(1) that Pis a GV-torsion-free module. Thus hinduces a homomorphism
h:P→Psuch that ϕh =ϕ. So his an isomorphism. Thus one has the
following commutative diagram with two exact rows:
0//T//P1//
h
P//
h
0
0//T//P1//P//0
So his an isomorphism. Therefore αis a special A-cover.
Remark 3.15.Taking A:= wPw, by Theorem 3.13 and Proposition 3.14, in
order to discuss the existence of a weak w-projective cover of a module, just
consider whether the w-module has a weak w-projective cover.
Acknowledgements. The authors would like to express their sincere thanks
for the referee for his/her careful reading and helpful comments. This research
was supported by the Academic Research Fund of Hoseo University in 2019
(20190817).
$KHDGRI3ULQW
14 H. KIM, L. QIAO, AND F. WANG
References
[1] F. A. A. Almahdi, M. Tamekkante, and R. A. K. Assaad, On the right orthogonal
complement of the class of w-flat modules, J. Ramanujan Math. Soc. 33 (2018), no. 2,
159–175.
[2] P. C. Eklof and J. Trlifaj, How to make Ext vanish, Bull. London Math. Soc. 33 (2001),
no. 1, 41–51. https://doi.org/10.1112/blms/33.1.41
[3] E. E. Enochs and O. M. G. Jenda, Relative homological algebra, De Gruyter Expositions
in Mathematics, 30, Walter de Gruyter & Co., Berlin, 2000. https://doi.org/10.1515/
9783110803662
[4] H. Kim and F. Wang, On LCM-stable modules, J. Algebra Appl. 13 (2014), no. 4,
1350133, 18 pp. https://doi.org/10.1142/S0219498813501338
[5] Y. Y. Pu, W. Zhao, G. H. Tang, and F. G. Wang, w∞-projective modules and Krull
domains, Commun. Algebra, to appear.
[6] Y. Y. Pu, W. Zhao, G. H. Tang, and F. G. Wang, w∞-Warfield cotorsion modules and
Krull domains, Algebra Colloq., to appear.
[7] J. Trlifaj, Covers, envelopes, and cotorsion theories, Lecture notes for the workshop,
Homological Methods in Module Theory, 10–16 September, Cortona, 2000.
[8] F. Wang, On w-projective modules and w-flat modules, Algebra Colloq. 4(1997), no. 1,
111–120.
[9] F. Wang and H. Kim, Two generalizations of projective modules and their applications,
J. Pure Appl. Algebra 219 (2015), no. 6, 2099–2123. https://doi.org/10.1016/j.jpaa.
2014.07.025
[10] F. Wang and H. Kim, Foundations of commutative rings and their modules, Algebra
and Applications, 22, Springer, Singapore, 2016. https://doi.org/10.1007/978-981-
10-3337- 7
[11] F. Wang and L. Qiao, The w-weak global dimension of commutative rings, Bull. Korean
Math. Soc. 52 (2015), no. 4, 1327–1338. https://doi.org/10.4134/BKMS.2015.52.4.
1327
[12] F. Wang and L. Qiao, A homological characterization of Krul l domains II, Comm. Al-
gebra 47 (2019), no. 5, 1917–1929. https://doi.org/10.1080/00927872.2018.1524007
[13] F. Wang and L. Qiao, A new version of a theorem of Kaplansky, Comm. Algebra 48
(2020), no. 8, 3415–3428. https://doi.org/10.1080/00927872.2020.1739289
[14] F. G. Wang and D. C. Zhou, A homological characterization of Krull domains, Bull.
Korean Math. Soc. 55 (2018), no. 2, 649–657. https://doi.org/10.4134/BKMS.b170203
Hwankoo Kim
Division of Computer Engineering
Hoseo University
Asan 31499, Korea
Email address:hkkim@hoseo.edu
Lei Qiao
College of Mathematics and Software Science
Sichuan Normal University
Chengdu 610068, P. R. China
Email address:lqiao@sicnu.edu.cn
Fanggui Wang
College of Mathematics and Software Science
Sichuan Normal University
Chengdu 610068, P. R. China
Email address:wangfg2004@163.com
$KHDGRI3ULQW