You are on page 1of 16

JP Journal of Algebra, Number Theory and Applications

© 2022 Pushpa Publishing House, Prayagraj, India


http://www.pphmj.com
http://dx.doi.org/10.17654/0972555522035
Volume 59, 2022, Pages 1-16 P-ISSN: 0972-5555

1-SEPARABLE TORSION FREE MODULES


OVER INTEGRAL DOMAINS

Abstract

A torsion free module over an integral domain is called separable


(resp. 1-separable) if every finite subset of elements (resp. each
element) is contained in a completely decomposable direct summand.
In this work, it is shown that over an integral domain, a 1-separable
module M is separable in case each homogeneous direct summand of
finite rank in M has the property that its pure rank 1 submodules are
direct summands.

I. When 1-separable implies Separable

The author extends his work on separable torsion free abelian groups
[3] to torsion free modules over integral domains. In this article, all rings
Received: July 27, 2022; Revised: October 14, 2022; Accepted: October 18, 2022
2020 Mathematics Subject Classification: 13C05, 13G05 (Primary); 20K15, 20K20.
Keywords and phrases: torsion free module, integral domain, separable module, quasi-
isomorphism, torsion free abelian group.
How to cite this article: E. F. Cornelius, Jr., 1-separable torsion free modules over integral
domains, JP Journal of Algebra, Number Theory and Applications 59 (2022), 1-16.
http://dx.doi.org/10.17654/0972555522035
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Published Online: November 2, 2022
2 E. F. Cornelius, Jr.
are integral domains (commutative rings with 1, without zero divisors), all
modules are unitary torsion free over integral domains, and all sums are
direct, unless otherwise indicated.
A torsion free module over an integral domain R is separable if
each finite subset of elements is contained in a completely decomposable
summand, completely decomposable meaning a direct sum of rank 1
submodules. An R-module has rank 1 if it is isomorphic to an R-submodule
of the quotient field K of R. A torsion free module over an integral domain
is 1-separable if each element is contained in a completely decomposable
summand.
Two torsion free modules A and B, over an integral domain R, are said to
be quasi-isomorphic, A ≈ B, if there exist submodules A′ ⊆ A and B′ ⊆ B
− B, and B′ ~
such that A′ is isomorphic to B, A′ ~ − A. Quasi-isomorphism is
an equivalence relation on torsion free R-modules [10, pp. 224-225].
Quasi-isomorphism frequently is supplemented with conditions on rings
or modules, to obtain stronger structural results for modules.
A submodule N of the torsion free module M over an integral domain is
pure if, whenever an element rx ∈ N for x ∈ M and r ∈ R, r ≠ 0, then
x ∈ N . Equivalently, N = N ∩ rM for all nonzero r ∈ R, or also M N is
torsion free. Purity is transitive in the sense that if L is a pure submodule of
N and N is a pure submodule of M, then L is pure in M. Purity sometimes is
called relative divisibility in literature dealing with modules, e.g., [7, p. 2].
The equivalence classes of rank 1 modules under quasi-isomorphism are
called types. Isomorphic rank 1 modules have the same type. The type of a
rank 1 module M is denoted T ( M ). The type of a nonzero element x, T ( x ) ,
is the type of the pure rank 1 submodule P generated by x, T( P ). The
nonzero elements of a rank 1 module have the same type [10, pp. 224-225].

The set of types admits a natural partial order. For types σ and τ, σ ≤ τ
if and only if there exist rank 1 R-modules A and B with T ( A) = σ and
T( B ) = τ such that A is a submodule of B. σ < τ if A is a submodule of B,
1-separable Torsion Free Modules Over Integral Domains 3
but there is no isomorphism from B into A. The smallest type is the common
type of all fractional ideals of R, while the largest type is the type of K, the
quotient field of R. Types form a lattice with inf (σ, τ) = T ( A ∩ B ) and
sup(σ, τ ) = T( A + B ) [10, p. 225].

If h is a nonzero module homomorphism from a rank 1 module A to a


rank 1 module B, then h is injective, and T ( A) ≤ T ( B ).

A module is homogeneous if each nonzero element has the same type. A


completely decomposable module is homogeneous if it is the direct sum of
rank 1 submodules of the same type, because of the lattice structure of types.

A module is homogeneously decomposable if it is the direct sum of


homogeneous submodules [10, pp. 225-226]. For a detailed discussion of
types and their properties, see [11, Ch. 2].

Baer [1] introduced a class of abelian groups (Z- modules) that is more
general than completely decomposable groups (a torsion free abelian group
G is called completely decomposable if it is a direct sum of groups of rank
1), but behave “locally” as if they were completely decomposable and were
termed “separable groups”. Recall that a torsion free group G is called
separable if every finite subset of elements of G is contained in a completely
decomposable direct summand of G. However, the present author in [3]
proved that a torsion free abelian group G is separable if every element of G
is contained in a completely decomposable direct summand of G. In the
present paper, the author extends his work on separable torsion free abelian
groups to torsion free modules over integral domains. However, for such
modules, an additional assumption seems reasonably necessary to obtain
satisfactory decompositions, one that posits that some pure rank 1
submodules are summands.

Proposition 1. Let M be a torsion free module over an integral domain,


with the property that every pure submodule of finite rank is a completely
decomposable summand. Then every pure submodule inherits that summand
property; in particular, every summand of M does.
4 E. F. Cornelius, Jr.
Proof. Let N be a pure submodule of M and let L be a pure finite rank
submodule of N. By hypothesis, L, being pure, is a completely decomposable
finite rank summand of M. By the modular law, L also is a completely
decomposable summand of N. □

This result for torsion free abelian groups has been known for some time
[6, p. 503, Lemma 4.5]. Proposition 1 also holds if every pure rank 1
submodule is a summand.

Proposition 2. Let M be a torsion free module over an integral domain.


Then the following are equivalent:

(1) Every pure rank 1 submodule is a summand.

(2) Every pure submodule of finite rank is a completely decomposable


summand.

Proof. Clearly (2) implies (1), so assume that (1) holds, and let N be a
pure submodule of finite rank n. The proof proceeds by induction on n.
When n = 1, (2) holds, so let n > 1 and assume that (2) holds for all
submodules of rank < n. Let x1, ..., xn −1, xn be independent elements of N
and let L be the pure submodule generated by x1, ..., xn −1. By the induction
hypothesis, L is a completely decomposable summand of finite rank in M,
M = L ⊕ C for some complement C. Then xn = l + c for some l ∈ L and
c ∈ C , c ≠ 0. Since C inherits the summand property, the pure rank 1
submodule P generated by c is a summand of C, C = P ⊕ D. Then
N = L ⊕ P and M = ( L ⊕ P ) ⊕ D. Because L is completely decomposable
of finite rank and P has rank 1, N is a completely decomposable finite rank
summand of M. □

Definition. A torsion free module over an integral domain has the pure
summand property if it satisfies (1) or (2) and hence both of the properties
described in Proposition 2.

The author generally prefers the minimalist pure rank 1 characterization


in Proposition 2(1).
1-separable Torsion Free Modules Over Integral Domains 5
Theorem 1. A torsion free module of finite rank over an integral domain
is homogeneous and completely decomposable if every pure submodule of
rank 1 is a summand.

Proof. If M possesses the stated property, by Proposition 2, it is


completely decomposable, M = P1 ⊕ ⋯ ⊕ Pm with each Pi of rank 1. To
see that the Pi ’ s are of the same type, it suffices to consider two of them, Pj

and Pk . Let x be a nonzero element of Pj and y a nonzero element of Pk ,

and consider the pure submodule P generated by x + y in Pj ⊕ Pk . If π j is

the natural projection of Pj ⊕ Pk onto Pj , then π j ( x + y ) = x, so that

T ( P ) = T ( x + y ) ≤ T ( x ) = T ( Pj ). Now P is a summand of Pj ⊕ Pk , since

Pj ⊕ Pk inherits the summand property, so Pj ⊕ Pk = P ⊕ C . Thus, T ( Pj )

≤ T ( P ) and so T ( Pj ) = T ( P ). The argument is symmetrical with Pk ,

which also has the same type as P. M is, therefore, homogeneous. □

The pure summand property possessed by a torsion free module over an


integral domain has a profound effect on its type and structure. The property
is the most fundamental kind of separability.

Corollary 1. A torsion free module over an integral domain is


homogeneous and separable if it has the pure summand property.

Proof. Let M be such a module with the pure summand property,


x1, ..., xn be a finite set of elements of M, and P be the pure submodule
generated by the xi . Then P is a summand of M and inherits the pure
summand property. By Theorem 1, P is a homogeneous completely
decomposable summand containing the xi . □

Corollary 2. A torsion free module of countable rank over an integral


domain is homogeneous and completely decomposable if it has the pure
summand property.
6 E. F. Cornelius, Jr.
Proof. Suppose M is such a module with the pure summand property
over the integral domain R. Let {xi : i ∈ N + } be a maximal independent
subset of M and let Pn be the pure submodule generated by x1, ..., xn . By
Theorem 1, Pn is a homogeneous completely decomposable summand of M
and so of Pn +1, Pn +1 = Pn ⊕ Cn +1 for some complement Cn +1. If C1 = P1,

then Pn = C1 ⊕ ⋯ ⊕ Cn and M = i =1 ⊕ Ci . To see the latter point, let

x ∈ M ; because the xi are a maximal independent set, rx = r1x1 + ⋯ +


rm xm for some positive integer m and r, r1, ..., rm ∈ R, r ≠ 0. Thus
∞ ∞
x ∈ Pm ⊆ i =1 ⊕ Ci because Pm is pure, so M = i =1 ⊕ Ci . □

Corollary 3. Let M be a homogeneously decomposable torsion free


module over an integral domain such that each homogeneous summand in
the decomposition of M has countable rank and possesses the pure summand
property. Then M is completely decomposable.

Proof. By Corollary 2, each summand of countable rank is homogeneous


and completely decomposable, so that M is the direct sum of homogeneous
completely decomposable summands, making it, too, completely
decomposable. □

Torsion free abelian groups with the pure summand property possess not
only the further properties proved above, but also satisfy converses. For
example, if G is a torsion free abelian group, then it is homogeneous and
separable if and only if it has the pure summand property [6, p. 503, Lemma
4.5], [3, Corollary]. For general torsion free modules over integral domains,
the converse is not so clear, i.e., if a torsion free module over an integral
domain is the direct sum of rank 1 submodules of the same type, it may not
be the case that every pure rank 1 submodule is a summand. For that reason,
the pure summand property may be a helpful assumption.
1-separable Torsion Free Modules Over Integral Domains 7
The proof of the main Theorem 2 requires that completely decomposable
summands of finite rank be homogeneously decomposed for simplification.
If M is a 1-separable torsion free module over an integral domain and x is a
nonzero element of M contained in a completely decomposable summand
C = C1 ⊕ ⋯ ⊕ Cn with the Ci of rank 1, then x = x1 + ⋯ + xn , xi ∈ Ci ,
i = 1, ..., n. C can be simplified, first by requiring that each component xi
of x be ≠ 0, then by collecting the Ci according to type into disjoint direct
sums S j of quasi-isomorphic rank 1 submodules, S j = C jl ⊕ ⋯ ⊕ C jl ,
1 k j

j = 1, ..., m, so that all of the Ci with the same type are in a single S j .
By design, S j is a completely decomposable homogeneous submodule,
consisting of a finite number of rank 1 submodules of the same type,
and C = S1 ⊕ ⋯ ⊕ Sm . The collection of the rank 1 summands into
homogeneous submodules is the desired homogeneous decomposition.

A further step can be taken - one crucial to this discussion - provided


that each pure rank 1 submodule of S j is a summand of S j . Each such pure
rank 1 submodule would have the same type as the rank 1 submodules
constituting S j , as S j is homogeneous. Let Pj denote the pure submodule
of S j generated by the component of x contained in S j , x jl + ⋯ + x jl ,
1 k j

j = 1, ..., m.

Under this pure summand hypothesis, Pj is a summand of S j , and


hence of C, so that x ∈ C ′ = P1 ⊕ ⋯ ⊕ Pm with the Pj ’ s rank 1 submodules
having distinct types. C ′ is a summand of C and so of M.

With these refinements, the proof of the main result can proceed through
a sequence of three lemmas. Also needed is some new notation. For a type τ
and a torsion free module N over an integral domain, let N (τ ) denote the
submodule of N, consisting of the elements with type ≥ τ [10, p. 225].
8 E. F. Cornelius, Jr.
Lemma 1. Let M be a homogeneously decomposable torsion free
module of finite rank over an integral domain, in which each homogeneous
summand in the decomposition of M has the property that its pure rank
1 submodules are summands. Then every summand of M is completely
decomposable.

Proof. By Theorem 1, each homogeneous summand of M is completely


decomposable, so let M = M1 ⊕ ⋯ ⊕ M n be a homogeneous decomposition
in which each M i is completely decomposable and homogeneous of distinct
type σi . If n = 1, then Corollary 1 supplies the answer. So suppose n > 1,
that the result holds for fewer summands, and that M = H ⊕ C . Suppose
further that σn is maximal among the distinct σi .

Then, by [10, p. 225], M n = M (σn ) = H (σn ) ⊕ C (σn ) , so that, by the


modular law, H = H (σn ) ⊕ H ′ and C = C (σn ) ⊕ C ′ for some H ′ ⊆ H
and C ′ ⊆ C . Thus, M = H (σn ) ⊕ H ′ ⊕ C (σn ) ⊕ C ′ = H ′ ⊕ C ′ ⊕ M n and
M Mn ~ − M1 ⊕ ⋯ ⊕ M n −1 − ~ H ′ ⊕ C ′. H (σn ) is a summand of M n and so
is completely decomposable by Proposition 2. The induction hypothesis
implies that H ′ is completely decomposable, because it is isomorphic to a
summand of M1 ⊕ ⋯ ⊕ M n −1, which is completely decomposable. Hence
H = H (σn ) ⊕ H ′ is completely decomposable because both summands
are. □

Lemma 2. Let M be a 1-separable torsion free module over an


integral domain, in which each homogeneous summand of finite rank has
the property that its pure rank 1 submodules are summands. Suppose
M = A ⊕ B with A the direct sum of rank 1 submodules Ai . Let b be a
nonzero element of B, which is contained in a completely decomposable
summand of M and satisfies T (b ) </ T ( Ai ) for all Ai in the decomposition
of A. Then b is contained in a completely decomposable summand of B.
1-separable Torsion Free Modules Over Integral Domains 9
Proof. By Theorem 1 and the discussion on homogeneous
decomposition supra, b may be assumed to be contained in a summand of M,
P1 ⊕ ⋯ ⊕ Pm , with the Pj rank 1 submodules having distinct types, such
that the component of b in Pj is nonzero for all j. Observe that T (b ) ≤
T ( Pj ) for all j, and that T (b ) = T ( Pk ) for at most one Pj because the Pj
have distinct types; for j ≠ k , T (b ) < T ( Pj ). For each j ≠ k , Pj ⊆ B
because T (b ) < T ( Pj ) and T (b ) </ T ( Ai ) for all Ai . This is so, because if
Pj , j ≠ k , had a nonzero component in some Ai , then the projection πi of
M onto Ai would send Pj into Ai , necessitating T (b ) < T ( Pj ) ≤ T ( Ai ) ,
contradicting T (b ) </ T ( Ai ) for all Ai . Thus, Pj ⊆ B for each j ≠ k . If P
denotes the pure submodule generate by b, then because P and all Pj , j ≠ k ,

are contained in B, so is Pk , meaning that b ∈ P1 ⊕ ⋯ ⊕ Pm ⊆ B. □

Lemma 3. Let M be a 1-separable torsion free module over an


integral domain, in which each homogeneous summand of finite rank has
the property that its pure rank 1 submodules are summands. Suppose
M = A ⊕ B with A completely decomposable of finite rank. Then B is
1-separable.

Proof. Let b be a nonzero element of B, which is contained in a


completely decomposable summand of M, P1 ⊕ ⋯ ⊕ Pm , with the Pj of
rank 1, having distinct types and the components of b ≠ 0. Let C denote a
complement to this summand in M, so that M = P1 ⊕ ⋯ ⊕ Pm ⊕ C = A ⊕ B.
The proof proceeds by induction on the rank of A.
Suppose that A has rank 1. Segregate the Pj ’ s as follows. Let L be the
sum of those Pj ’ s for which T ( Pj ) ≥/ T ( A) , and let S be the sum of those
Pj ’ s that satisfy T ( Pj ) ≥ T ( A). Without loss of generality, assume that L
and S are nontrivial. If L = Pj1 ⊕ ⋯ ⊕ Pjk , note that for a nonzero element
l ∈ L, T (l ) ≥/ T ( A) because if it were ≥, then T ( Pji ) ≥ T (l ) ≥ T ( A) , in
10 E. F. Cornelius, Jr.
contradiction to the definition of L. With this notation, M = L ⊕ S ⊕ C =
A ⊕ B and b ∈ ( L ⊕ S ) ∩ B.

A consideration of types reveals that A ⊆ S ⊕ C , for if a were a


nonzero element of A with a = l ′ + s + c, l ′ ∈ L, l ′ ≠ 0, s ∈ S , and
c ∈ C , the projection of M onto L would map a to l ′ and require T (l ′) ≥
T (a ) = T ( A). Thus, l ′ = 0 and A ⊆ S ⊕ C. Application of the modular
law to S ⊕ C and M = A ⊕ B yields S ⊕ C = A ⊕ D with D = ( S ⊕ C )
∩ B, so that M = L ⊕ S ⊕ C = A ⊕ L ⊕ D. Observe that A ⊕ L has finite
rank.

Express b ∈ ( L ⊕ S ) ∩ B as the sum of its L and S components,


b = l ′′ + s′; note l ′′, s′ ≠ 0, assuming that L and S are nontrivial. Now
write s′ as the sum of its A and D components, s′ = a ′ + d , respectively.
By design, T ( s′) ≥ T ( A) and consequently T (d ) ≥ T ( A) , because also
T (a′) ≥ T ( A). With reference to the composition of L as Pj1 ⊕ ⋯ ⊕ Pjk , if

T (d ) < T ( Pji ) for some ji , then T ( A) ≤ T (d ) < T ( Pji ) would contradict

the definition of L as the sum of those Pj ’ s for which T ( Pj ) ≥/ T ( A). Thus


T (d ) </ T ( Pji ) , i = 1, ..., k , and T (d ) </ T ( A) , so Lemma 2, applied to
M = ( A ⊕ L ) ⊕ D and d, yields a decomposition D = E ⊕ F , in which E
is a finite rank completely decomposable submodule of D, containing d.

When results are combined, M = ( A ⊕ L ⊕ E ) ⊕ F with A ⊕ L ⊕ E


completely decomposable of finite rank containing l ′′ + a ′ + d = l ′′ + s′ = b.
Another application of the modular law to M = A ⊕ B and A ⊕ L ⊕ E
decomposes A ⊕ L ⊕ E into A ⊕ H , where H = ( A ⊕ L ⊕ E ) ∩ B and
b ∈ H ⊆ B. Now H, as a finite rank summand of the completely
decomposable submodule A ⊕ L ⊕ E , also of finite rank, is homogeneously
decomposable by [10, Corollary 1]. H, as a summand of A ⊕ L ⊕ E , is also
a summand of M so by assumption, homogeneous summands of H have
the pure summand property. H is, therefore, completely decomposable
1-separable Torsion Free Modules Over Integral Domains 11
by Corollary 3. Thus, M = A ⊕ H ⊕ F with b ∈ H ⊆ B and B = H ⊕
[B ∩ ( A ⊕ F )]. Inasmuch as each element b ∈ B is contained in a
completely decomposable summand of B, B by definition is 1-separable.
The induction hypothesis is that if M is a torsion free 1-separable
module over an integral domain, in which each homogeneous summand
has the property that its pure rank 1 submodules are summands, and if
M = A ⊕ B with A completely decomposable of finite rank ≤ n, then B is
1-separable.

Now suppose that M is a torsion free 1-separable module over an


integral domain, in which each homogeneous summand has the property
that its pure rank 1 submodules are summands, and M = A ⊕ B with A
completely decomposable of rank n + 1. Express A = I ⊕ J with I of
rank n, so that M = I ⊕ ( J ⊕ B ). Under the induction hypothesis, J ⊕ B
is 1-separable, and by Proposition 1, it inherits the pure summand property.
Therefore, the fact that B is 1-separable follows from the proof of the first
case applied to J ⊕ B, because J has rank 1. □

Note that if L or S is trivial or d = 0 in the proof of Lemma 3, then the


proof is significantly simplified. The principal result can now be established
by an elementary induction argument.

Theorem 2. Let M be a 1-separable module over an integral domain, in


which each homogeneous summand of finite rank has the property that its
pure rank 1 submodules are summands. Then M is separable.

Proof. Assume that each subset of M having at most n elements is


contained in a completely decomposable summand. Let x1, ..., xn , xn +1 be
any elements of M. By hypothesis, x1, ..., xn are contained in a completely
decomposable summand A of finite rank, M = A ⊕ B. By Lemma 3, B is
1-separable, so that the B component of xn +1 is contained in a completely
decomposable summand C of B. A ⊕ C is then a completely decomposable
summand of M that contains all of the xi ’ s. □
12 E. F. Cornelius, Jr.
II. Whither go 1-separable Modules?

The extension of the 1-separable result for torsion free abelian groups to
various torsion free modules over integral domains may not be easy, because
the proof for abelian groups was rather difficult. László Fuchs described the
work as “an unexpected result” [9, p. 507] that was “not easy to prove”
[8, p. 15]. It was called “quite an astonishing result” by Reid and Wickless
[16, p. 4], quoting Göbel [9, p. 16]. Extending the result without additional
assumptions might prove to be even more difficult.
Metelli arrived at the result using a more complicated approach, coining
the description “1-separable” in the process [12, p. 266], whereas the
author’s proof used only established methods and theorems. Fuchs and
Viljoen extended the result to the classes of groups, expanding upon the
author’s ideas, but did not use 1-separable terminology [8]. Rangaswamy
extended and simplified the work of Fuchs and Viljoen [15]. However, these
extensions were limited to torsion free abelian groups; few, if any, ventured
into the realm of more general rings and modules.
Quasi-isomorphism of torsion free abelian groups does provide some
generalization, as noted in the author’s Remark [3]. Such work usually is
done in the context of a divisible group, a vector space V over the rational
numbers Q [4, 5]. Torsion free abelian groups A and B are said to be quasi-
equal, A =ɺ B, if there are positive integers m and n such that mA ⊆ B and
nB ⊆ A. A and B are quasi-isomorphic if there exist isomorphic groups
A′ ⊆ A and B′ ⊆ B such that mA ⊆ A′ and nB ⊆ B′ [2, Section 3]. This
definition of quasi-isomorphic comports with that given for modules because
if f is an isomorphism between A′ and B′, then A → mA → f (mA) ⊆ B
yields an isomorphism from A to a subgroup of B and similarly for n and
f −1. A′ ⊆ A is a quasi-summand of A if A =ɺ A′ ⊕ C for some complement
C ⊆ A. There exist torsion free abelian groups of rank 2, that are quasi-
isomorphic but not isomorphic [2, Lemma 9.4].
Quasi-isomorphic torsion free abelian groups of rank 1 are isomorphic,
as are all quasi-isomorphic torsion free rank 1 modules over principal ideal
1-separable Torsion Free Modules Over Integral Domains 13
domains [10, Theorem 1]. Thus, it may be unnecessary to delve into types as
used in torsion free abelian groups [6, Ch. 12, Section 1]. Quasi-isomorphic
rank 1 modules need not be isomorphic over general integral domains; for
the case of Dedekind domains, consult [10, pp. 228-231]. Recall that an
integral domain is Dedekind if every proper ideal factor into primes, or
equivalently, every fractional ideal is invertible.
Quasi-decomposition may imply decomposition in some cases. If each
pure subgroup of a reduced torsion free abelian group G is a quasi-summand,
then G is the direct sum of a finite number of rank 1 subgroups with
comparable types [4, Theorem 1.6]. Hence, if each pure rank 1 subgroup of a
reduced torsion free abelian group is a quasi-summand, then the group is
separable and its elements have comparable types.
As previously noted, to obtain useful structure under quasi-isomorphism
of modules, it may be necessary or desirable to impose additional conditions
on rings or modules. In [10], although the definition of equivalence for
rank 1 modules is stated for integral domains, the principal results are for
Dedekind domains.
Olberding put conditions on integral domains and their modules to
obtain interesting structural results. His paper “Prüfer domains and pure
submodules of direct sums of ideals” [13] begins as follows:

If A = I1 ⊕ I 2 ⊕ ⋯ ⊕ I n is a direct sum of finitely many ideals of a


Dedekind domain, then every pure submodule of A is a direct summand of
A. It is well-known that valuation domains also possess this property. This
suggests that the class of integral domains for which pure submodules of
direct sums of finitely many ideals are summands is significantly larger than
the class of Dedekind domains. In this note, we prove that in fact the integral
domains which inherit this property are precisely the h-local Prüfer domains.
In Olberding’s paper, he employs a definition of quasi-equal analogous
to that used in the study of torsion free abelian groups: If X and Y are
submodules of the quotient field K of an integral domain R, then X is quasi-
equal to Y if there exist nonzero elements r, s ∈ R such that rX ⊆ Y and
14 E. F. Cornelius, Jr.
sY ⊆ X . This, of course, is a much stronger condition than mere quasi-
isomorphism, although multiplications r : X → rX ⊆ Y and s : Y → sY
⊆ X do satisfy the definition of quasi-isomorphism. Olberding also puts
conditions on rings to obtain structural result for modules.

Recall that a domain is h-local if every element belongs only to finitely


many maximal ideals, and every prime ideal is contained only in a single
maximal ideal. A Prüfer domain is a semihereditary integral domain;
equivalently, a Prüfer domain may be defined as a commutative ring without
zero divisors in which every finitely generated ideal is invertible. A
valuation domain R is an integral domain such that for every element x in its
field of fractions K, at least one of x or x −1 belongs to R. Additional
information may be found in [14]. With these definitions, the principal
results from [13] germane to this article are set forth below.

[A] Let R be an integral domain with quotient field K. Then for any
proper submodule X of K, the following statements are equivalent:

(i) Every pure submodule of a finite direct sum of submodules of K,


each quasi-equal to X, is a summand.

(ii) Every pure rank 1 submodule of a finite direct sum of submodules of


K, each quasi-equal to X, is a summand [13, Lem. 3.1(1) & (3)].

[B] The following are equivalent for an integral domain R with quotient
field K:

(i) R is an h-local Prüfer domain.

(ii) Every pure submodule of a finite direct sum of quasi-equal


submodules of K is a summand [13, Th. 3.2(1) & (2)].

[C] An integral domain R is a valuation domain if and only if pure


submodules of finite direct sums of submodules of its quotient field K are
summands [13, Cor. 3.3].
Each case involves the interplay between conditions on rings and
modules sufficient to establish modular structure. The results are fascinating,
1-separable Torsion Free Modules Over Integral Domains 15
because they essentially say that the subject rings are determined by
the properties of their modules. Although Olberding does not use the
terminology 1-separable, separable, or pure summand property, those terms
are clearly applicable to his work.
As a final example of conditions placed on rings and modules to obtain
structural results from quasi-isomorphism, consider [7], where the definition
of a “separable” module M requires every finite set of elements of M to be
contained in a summand of finite rank, and for finite rank summands of M to
be completely decomposable. Olberding’s results on h-local Prüfer domains
are utilized.
Much remains to be explored about 1-separable modules over
integral domains under quasi-isomorphism. To firmly establish the class of
1-separable modules as a niche in the class of torsion free modules over
integral domains, examples are needed of 1-separable modules that are not
separable. The author has posted requests for examples on several websites
but has received no responses. Assuming that such examples exist, it would
be interesting to know what properties an integral domain must have for its
1-separable modules to be separable.

Acknowledgements

The author thanks Professor Phill Schultz of the University of Western


Australia for insightful suggestions that greatly improved the exposition of
this article.
The author also thanks the anonymous referees for their valuable
suggestions and constructive criticisms which improved the presentation of
the paper.

References

[1] R. Baer, Abelian groups without elements of finite order, Duke Math. J. 3 (1937),
68-122.
[2] R. A. Beaumont and R. S. Pierce, Torsion free groups of rank two, Mem. Amer.
Math. Soc. 38 (1961), 41 pp.
16 E. F. Cornelius, Jr.
[3] E. F. Cornelius, Jr., A sufficient condition for separability, J. Algebra
67(2) (1980), 476-478.
https://www.sciencedirect.com/science/article/pii/0021869380901714.
[4] E. F. Cornelius, Jr., Characterization of a class of torsion free groups in terms of
endomorphisms, Pacific J. Math. 79(2) (1978), 341-355. Submission date
corrected to February 5, 1974, Pacific J. Math. 85(2) (1979), 501.
[5] E. F. Cornelius, Jr., A generalization of separable groups, Pacific J. Math.
39(3) (1970), 603-613.
[6] László Fuchs, Abelian Groups, Springer, Switzerland, 2015.
[7] László Fuchs and Jorge E. Macías-Díaz, On completely decomposable and
separable modules over Prüfer domains, J. Comm. Alg. 2(2) (2010), 159-176.
[8] L. Fuchs and G. Viljoen, A generalization of separable torsion-free abelian
groups, Rend. Sem. Mat. Univ. Padova 73 (1985), 15-21.
[9] Rüdiger Göbel, László Fuchs - a personal evaluation of his contributions to
mathematics, Period. Math. Hungar. 32(1-2) (1996), 13-29.
[10] G. Kolettis, Jr., Homogeneously decomposable modules, Studies on Abelian
Groups, Springer-Verlag, 1968, pp. 223-238.
[11] E. Lee Lady, Finite Rank Torsion Free Modules over Dedekind Domains,
University of Hawaii, 1998. http://www.math.hawaii.edu/~lee/book/.
[12] C. Metelli, On type-related properties of torsionfree abelian groups, Abelian
Group Theory, R. Göbel, L. Lady and A. Mader, eds., Lecture Notes in
Mathematics, Springer-Verlag, Vol. 1006, 1983, pp. 253-267.
[13] Bruce Olberding, Prüfer domains and pure submodules of direct sums of ideals,
Mathematika 46 (1999), 425-432.
[14] Bruce Olberding, Characterizations and constructions of h-local domains,
Contributions to Module Theory, Walter de Gruyter, 2007, pp. 1-21.
[15] K. M. Rangaswamy, On C-separable Abelian groups, Comm. Algebra
13(6) (1985), 1219-1227.
[16] J. D. Reid and W. J. Wickless, Ross Allen Beaumont (in Memoriam), Abelian
Groups and Modules, Paul C. Eklof and Rüdiger Göbel, eds., International
Conference in Dublin, August 10-14, 1998, Birkhauser Boston, 1999.

You might also like