This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

FEW REMARKS ON ESSENTIAL SUBMODULES

\nameSourav Koner and Biswajit Mitra CONTACT Email: [email protected] & [email protected] Department of Mathematics, The University of Burdwan, India 713104 & Department of Mathematics, The University of Burdwan, India 713104
Abstract

In the present paper, modules over integral domains and principal ideal domains that are proper essential extensions of some submodules are classified. We introduce a new class of modules that we call SM\mathrm{SM} modules and show that the class of Artinian modules, locally finite modules, and modules of finite lengths are all proper subclasses of SM modules. We also show that non-semisimple SM\mathrm{SM} modules possess essential socles. Further, we show that non-semieimple modules over integral domains with nonempty torsion-free parts do not possess essential socles.

keywords:
Proper essential submodule, Socle of a module, Direct sums of modules, Free modules, Modules over principal ideal domains and integral domains.

MSC (2020): Primary: 16D10; Secondary: 16D80.

1 Introduction

An essential submodule of a module is a submodule that intersects any other nonzero submodule nontrivially. It is named by Eckman and Schpof (see [ES (1953)]) though the idea behind essential submodule in a left RR-module MM is due to Johnson (see [RJ (1951)]). Essential submodules play a crucial role in noncommutative algebra. As an example, a necessary and sufficient condition that a left ideal in a ring RR is essential, is the existence of a nonzero-divisor (that is, a regular element), is again a necessary and sufficient condition for a ring RR to have a left classical Artin semisimple quotient ring, that is, RR is semiprime left Goldie ring. This shows that the self-generating property of essential submodules generate other essential objects. In much of the literature, the definition of an essential submodule includes the parent module. In this paper, we work with proper essential submodules whose definition is the following:

Definition 1.1.

A proper nontrivial submodule EE of a left RR-module MM is said to be a proper essential submodule of MM if EN{0M}E\cap N\neq\{0_{M}\} for every nontrivial submodule NN of MM.

Throughout the paper, unless otherwise stated, we assume RR is a ring with unity 101\neq 0 which is not a field. Observe that any non-trivial proper ideal II in an integral domain RR is a proper essential submodule of the RR-module RR. However, it is not necessary for an RR-module to contain a proper essential submodule. Because, if we consider a finite \mathbb{Z}-module MM of order nn, where nn is square-free, and if NpN_{p} denote the submodule of MM of order pp where pp is a prime dividing nn, then M=pnNpM=\bigoplus_{p\mid n}N_{p} shows that MM does not contain any proper essential submodule. The following proposition is our observation and is supposed to prevail in the literature, but we could not trace the result either in direct or indirect form anywhere. So we proved this result and incorporated it here.

Proposition 1.2.

If an RR-module MM is semisimple, then MM does not contain any proper essential submodule. Hence, any 𝕜\mathbbm{k}-vector space VV does not contain any proper essential subspace.

Proof.

Let M=αΛNαM=\bigoplus_{\alpha\in\Lambda}N_{\alpha}, where each NαN_{\alpha} is a simple submodule of MM. If a submodule EE of MM intersects every other non-trivial submodule of MM non-trivially, then NαEN_{\alpha}\subseteq E for all αΛ\alpha\in\Lambda. But this implies αΛNαE\bigoplus_{\alpha\in\Lambda}N_{\alpha}\subseteq E, that is, MEM\subseteq E.

Let {vα}αΛ\{v_{\alpha}\}_{\alpha\in\Lambda} be a basis for the 𝕜\mathbbm{k}-vector space VV. Then we have V=αΛVαV=\bigoplus_{\alpha\in\Lambda}V_{\alpha}, where VαV_{\alpha} is the subspace generated by vαv_{\alpha}. As each VαV_{\alpha} is a simple 𝕜\mathbbm{k}-subspace of VV, we can use the first part of this proposition to conclude that VV does not contain any proper essential subspace. ∎

A part of the following theorem (1.3) is very much well-known (see [TL (1999)]). Therefore, we include proof of the other part of the theorem which we again did not find in the literature.

Theorem 1.3.

Let MM be a left module over RR. Then one can show the equivalence

(a)(\mathrm{a}) Every submodule of MM is a direct summand.

(b)(\mathrm{b}) MM has no proper essential submodule.

(c)(\mathrm{c}) MM is semisimple.

(d)(\mathrm{d}) Soc(M)=M\mathrm{Soc}(M)=M.

Proof.

Equivalency of (a), (c) and (d) are found in [TL (1999)]. Here, we show the equivalence of (a) and (b), that is, (a)(b)\mathrm{(a)}\Leftrightarrow\mathrm{(b)}.

(a)(b):\mathrm{(a)}\Rightarrow\mathrm{(b)}: Suppose MM has a proper essential submodule. Then from the definition, it trivially follows that the proper essential submodule can not be the direct summand.

(b)(a):\mathrm{(b)}\Rightarrow\mathrm{(a)}: Conversely let NN be a proper submodule of MM. Then NN is not proper essential. There exists a submodule TT such that NT={0}N\cap T=\{0\}. Choose a maximal such TT, then NTN\bigoplus T is either proper essential in MM or NT=MN\bigoplus T=M. But as per hypothesis, NT=MN\bigoplus T=M, so the result follows. ∎

Theorem (1.4) can be found in [JL (1966)]), so, we skip its proof.

Theorem 1.4.

The intersection of all essential submodules of a left RR-module MM is equal to its socle Soc(M)\mathrm{Soc}(M).

2 Proper essential submodules over integral and principal ideal domains

Theorem 2.1.

Let {Mω}ωΛ\{M_{\omega}\}_{\omega\in\Lambda} be a family of left RR-modules indexed by a nonempty set Λ\Lambda. Then ωΛMω\bigoplus_{\omega\in\Lambda}M_{\omega} has a proper essential submodule if and only if some MωM_{\omega} has a proper essential submodule.

Proof.

Let M=ωΛMωM=\bigoplus_{\omega\in\Lambda}M_{\omega} and ω0Λ\omega_{0}\in\Lambda be such that Mω0M_{\omega_{0}} has a proper essential submodule Eω0E_{\omega_{0}}. Let X=ωΛXωX=\bigoplus_{\omega\in\Lambda}X_{\omega}, where Xω0=Eω0X_{\omega_{0}}=E_{\omega_{0}} and Xω=MωX_{\omega}=M_{\omega} if ωω0\omega\neq\omega_{0}. We claim that XX is a proper essential submodule of MM. Let NN be any nontrivial submodule of MM and let (nω)ωΛ(n_{\omega})_{\omega\in\Lambda} be a nonzero element of NN. Clearly, (nω)ωΛX(n_{\omega})_{\omega\in\Lambda}\in X if nω0=0Mω0n_{\omega_{0}}=0_{M_{\omega_{0}}}. If nω00Mω0n_{\omega_{0}}\neq 0_{M_{\omega_{0}}}, consider the cyclic module nω0\langle n_{\omega_{0}}\rangle. As Eω0nω0{0Mω0}E_{\omega_{0}}\cap\langle n_{\omega_{0}}\rangle\neq\{0_{M_{\omega_{0}}}\}, we conclude that XN{0M}X\cap N\neq\{0_{M}\}.

For the converse, assume that MM has a proper essential submodule EE but MωM_{\omega} does not have a proper essential submodule for all ωΛ\omega\in\Lambda. Observe that, for all ωΛ\omega\in\Lambda, we have either EMωE\cap M_{\omega} is proper essential in MωM_{\omega} or EMω=MωE\cap M_{\omega}=M_{\omega}. As MωM_{\omega} does not have any proper essential submodule, it must be that EMω=MωE\cap M_{\omega}=M_{\omega} for all ωΛ\omega\in\Lambda, that is, MEM\subseteq E, a contradiction. ∎

Free modules play a central role in algebra, since any module is the homomorphic image of some free module. The following corollary is now immediate from theorem (2.1).

Theorem 2.2.

Let RR be an integral domain and let AA be a nonempty set. Then the free RR-module R(A)R(A) on the nonempty subset AA has a proper essential submodule.

Proof.

Since R(A)R(A) is free on the subset AA, we have R(A)ARR(A)\simeq\bigoplus_{A}R. The proof is now completed using theorem (2.1). ∎

An immediate consequence of theorem (2.1) is the classification of finitely generated modules over principal ideal domains possessing proper essential submodules. Before that, we look the following lemma.

Lemma 2.3.

The non-simple left cyclic module m\langle m\rangle over an integral domain RR has a proper essential submodule if and only if there exists a maximal ideal \mathcal{M} of RR containing AnnR(m)\mathrm{Ann}_{R}(m) such that for any ideal II of RR containing AnnR(m)\mathrm{Ann}_{R}(m) we have αβAnnR(m)\alpha\beta\notin\mathrm{Ann}_{R}(m) for some αAnnR(m)\alpha\in\mathcal{M}\setminus\mathrm{Ann}_{R}(m) and for some βIAnnR(m)\beta\in I\setminus\mathrm{Ann}_{R}(m). Hence, the non-simple cyclic module m\langle m\rangle over a principal ideal domain RR has a proper essential submodule if and only if the generator of the AnnR(m)\mathrm{Ann}_{R}(m) is not square-free.

Proof.

Let \mathcal{M} be a maximal ideal of RR that contains AnnR(m)\mathrm{Ann}_{R}(m) and II be any ideal of RR containing AnnR(m)\mathrm{Ann}_{R}(m). If for some αAnnR(m)\alpha\in\mathcal{M}\setminus\mathrm{Ann}_{R}(m) and for some βIAnnR(m)\beta\in I\setminus\mathrm{Ann}_{R}(m) we have αβAnnR(m)\alpha\beta\notin\mathrm{Ann}_{R}(m), then we see that /AnnR(m)\mathcal{M}/\mathrm{Ann}_{R}(m) is proper essential in R/AnnR(m)R/\mathrm{Ann}_{R}(m). As R/AnnR(m)mR/\mathrm{Ann}_{R}(m)\simeq\langle m\rangle, we conclude that m\langle m\rangle has a proper essential submodule.

Conversely, suppose m\langle m\rangle has a proper essential submodule EE. Since we have mR/AnnR(m)\langle m\rangle\simeq R/\mathrm{Ann}_{R}(m), there exists some ideal JJ of RR containing AnnR(m)\mathrm{Ann}_{R}(m) such that J/AnnR(m)EJ/\mathrm{Ann}_{R}(m)\simeq E. Put JJ in some maximal ideal \mathcal{M}. As J/AnnR(m)J/\mathrm{Ann}_{R}(m) is proper essential in R/AnnR(m)R/\mathrm{Ann}_{R}(m), for any ideal II of RR containing AnnR(m)\mathrm{Ann}_{R}(m), we have J/AnnR(m)I/AnnR(m){0m}J/\mathrm{Ann}_{R}(m)\cap I/\mathrm{Ann}_{R}(m)\neq\{0_{\langle m\rangle}\}. As JJ is contained in \mathcal{M}, we get that /AnnR(m)I/AnnR(m){0m}\mathcal{M}/\mathrm{Ann}_{R}(m)\cap I/\mathrm{Ann}_{R}(m)\neq\{0_{\langle m\rangle}\}.

Let AnnR(m)=a\mathrm{Ann}_{R}(m)=\langle a\rangle and let a=up1p2pna=up_{1}p_{2}\cdots p_{n} where pip_{i}’s are distinct primes and uu is a unit. The only maximal ideals that contain a\langle a\rangle are the ideals pi\langle p_{i}\rangle where 1in1\leq i\leq n. If αpia\alpha\in\langle p_{i}\rangle\setminus\langle a\rangle and βqia\beta\in\langle q_{i}\rangle\setminus\langle a\rangle where qi=p1p2pi1pi+1pnq_{i}=p_{1}p_{2}\cdots p_{i-1}p_{i+1}\cdots p_{n}, we see that αβa\alpha\beta\in\langle a\rangle. Now, the first part of this lemma shows that m\langle m\rangle does not have a proper essential submodule.

Assume now that a=up1α1p2α2pnαna=up_{1}^{\alpha_{1}}p_{2}^{\alpha_{2}}\cdots p_{n}^{\alpha_{n}}, where pip_{i}’s are distinct primes, uu is a unit and αi>1\alpha_{i}>1 for some ii. It is now evident that the maximal ideal pi\langle p_{i}\rangle of RR satisfies the property of the first of this lemma. ∎

Theorem 2.4.

A finitely generated left module MM over a principal ideal domain RR has a proper essential submodule if and only if the betti number of MM is at least 11 or some invariant factor is not square free.

Proof.

The fundamental theorem of finitely-generated modules over principal ideal domain yields the invariant-factor decomposition M=RnR/a1R/amM=R^{n}\bigoplus R/\langle a_{{1}}\rangle\bigoplus\cdots\bigoplus R/\langle a_{{m}}\rangle. The proof is now completed using theorems (2.1), (2.2), and lemma (2.3) ∎

Lemma 2.5.

A left RR-module MM has a proper essential submodule if and only if the left R/AnnR(M)R/\mathrm{Ann}_{R}(M)-module MM has a proper essential submodule. Hence, if an RR-module MM has a proper essential submodule, then AnnR(M)\mathrm{Ann}_{R}(M) can not be a maximal ideal.

Proof.

At first, assume that RR-module MM has a proper essential submodule EE. Now, let NN be any submodule of R/AnnR(M)R/\mathrm{Ann}_{R}(M)-module MM. Consider the cyclic R/AnnR(M)R/\mathrm{Ann}_{R}(M)-module n\langle n\rangle, where nNn\in N be any element. As, nE{0M}\langle n\rangle\cap E\neq\{0_{M}\}, there exists rRr\in R such that rn0rn\neq 0 and rnnErn\in\langle n\rangle\cap E. Then (r+AnnR(M))nnE(r+\mathrm{Ann}_{R}(M))n\in\langle n\rangle\cap E (as R/AnnR(M)R/\mathrm{Ann}_{R}(M)-module). Hence, this shows that R/AnnR(M)R/\mathrm{Ann}_{R}(M)-module MM has a proper essential submodule EE.

Conversely, assume that R/AnnR(M)R/\mathrm{Ann}_{R}(M)-module MM has a proper essential submodule EE. Let NN be any submodule of MM. Then, nE{0M}\langle n\rangle\cap E\neq\{0_{M}\} for nNn\in N with n0Mn\neq 0_{M}. This implies, for some (r+AnnR(M))R/AnnR(M)(r+\mathrm{Ann}_{R}(M))\in R/\mathrm{Ann}_{R}(M) with r+AnnR(M)AnnR(M)r+\mathrm{Ann}_{R}(M)\neq\mathrm{Ann}_{R}(M), (r+AnnR(M))nE(r+\mathrm{Ann}_{R}(M))n\in E, that is, rnErn\in E. Thus, EE is proper essential in MM.

Now, applying proposition (1.2) and the first part of this lemma, we get that AnnR(M)\mathrm{Ann}_{R}(M) can not be a maximal ideal. ∎

Let MM be a module over an integral domain RR. If pRp\in R is a prime, recall that the pp-primary component AnnM(p)\mathrm{Ann}^{\ast}_{M}(p) of MM is the set of all elements of MM that are annihilated by some positive power of pp and its sub-primary component AnnM(pm)\mathrm{Ann}_{M}(p^{m}) is the set of all elements of MM that are annihilated by some fixed power of pp (i.e., annihilated by pmp^{m}).

Lemma 2.6.

Let MM be a non-faithful left module over a principal ideal domain RR and let p1,p2,,pnp_{1},p_{2},\ldots,p_{n} be the distinct primes that divide the generator of AnnR(M)\mathrm{Ann}_{R}(M). Then MM has a proper essential submodule if and only if some sub-primary component of MM has a proper essential submodule for some prime pip_{i}.

Proof.

Let AnnR(M)=a\mathrm{Ann}_{R}(M)=\langle a\rangle and a=up1α1p2α2pnαna=up_{1}^{\alpha_{1}}p_{2}^{\alpha_{2}}\cdots p_{n}^{\alpha_{n}}, where pip_{i}’s are distinct primes and uu is a unit. For 1in1\leq i\leq n, if we set qi=p1α1p2α2pi1αi1pi+1αi+1pnαnq_{i}=p_{1}^{\alpha_{1}}p_{2}^{\alpha_{2}}\cdots p_{i-1}^{\alpha_{i-1}}p_{i+1}^{\alpha_{i+1}}\cdots p_{n}^{\alpha_{n}}, then the homomorphism φi:MqiM\varphi_{i}:M\rightarrow q_{i}M shows that M/Ker(φi)qiMM/\mathrm{Ker}(\varphi_{i})\simeq q_{i}M. Since we have Ker(φi)=piαiM\mathrm{Ker}(\varphi_{i})=\langle p_{i}^{\alpha_{i}}\rangle M and qiM=AnnM(piαi)q_{i}M=\mathrm{Ann}_{M}(p_{i}^{\alpha_{i}}), we conclude that M/piαiMAnnM(piαi)M/\langle p_{i}^{\alpha_{i}}\rangle M\simeq\mathrm{Ann}_{M}(p_{i}^{\alpha_{i}}). As the ideals p1α1\langle p_{1}^{\alpha_{1}}\rangle, p2α2\langle p_{2}^{\alpha_{2}}\rangle, …, pnαn\langle p_{n}^{\alpha_{n}}\rangle are pairwise comaximal, the chinese remainder theorem yields Mi=1nAnnM(piαi)M\simeq\bigoplus_{i=1}^{n}\mathrm{Ann}_{M}(p_{i}^{\alpha_{i}}). The result now follows from theorem (2.1). ∎

Theorem 2.7.

Let MM be a left torsion module over a principal ideal domain RR. Then MM has a proper essential submodule if and only if AnnM(pi)AnnM(pj)\mathrm{Ann}_{M}(p^{i})\subsetneq\mathrm{Ann}_{M}(p^{j}) for some some prime pp and for some positive integers i,ji,j with i<ji<j.

Proof.

Let mMm\in M be such that rm=0rm=0 for some rRr\in R. Let r=up1α1p2α2pnαnr=up_{1}^{\alpha_{1}}p_{2}^{\alpha_{2}}\cdots p_{n}^{\alpha_{n}} be the prime factorisation of rr into primes, where uu is some unit. Consider the cyclic submodule m\langle m\rangle. As the submodule m\langle m\rangle is annihilated by the ideal r\langle r\rangle, we can apply lemma (2.6) to conclude that m=i=1nAnnm(piαi)\langle m\rangle=\bigoplus_{i=1}^{n}\mathrm{Ann}_{\langle m\rangle}(p_{i}^{\alpha_{i}}). Thus, M=p-primeAnnM(p)M=\sum_{{p{\text{-}\mathrm{prime}}}}\mathrm{Ann}^{\ast}_{M}(p). Now choose finite number of primes q1,q2,,qlq_{1},q_{2},\ldots,q_{l} in RR and consider the submodules AnnM(q1),AnnM(q2),,AnnM(ql)\mathrm{Ann}^{\ast}_{M}(q_{1}),\mathrm{Ann}^{\ast}_{M}(q_{2}),\ldots,\mathrm{Ann}^{\ast}_{M}(q_{l}). If mAnnM(qj)ijAnnM(qi)m\in\mathrm{Ann}^{\ast}_{M}(q_{j})\bigcap\sum_{i\neq j}\mathrm{Ann}^{\ast}_{M}(q_{i}) for some fixed jj with 1jl1\leq j\leq l, then we have qjsjm=0q_{j}^{s_{j}}m=0 for some sjs_{j}\in\mathbb{N} and m=m1++mj1+mj+1++mlm=m_{1}+\cdots+m_{j-1}+m_{j+1}+\cdots+m_{l}. Also, for all 1il1\leq i\leq l with iji\neq j, we have qisimi=0q_{i}^{s_{i}}m_{i}=0 for some sis_{i}\in\mathbb{N}. As RR is a principal ideal domain, there exists x,yRx,y\in R such that qjsjx+ξy=1q_{j}^{s_{j}}x+\xi y=1, where ξ=q1s1q2s2qj1sj1qj+1sj+1qlsl\xi=q_{1}^{s_{1}}q_{2}^{s_{2}}\cdots q_{j-1}^{s_{j-1}}q_{j+1}^{s_{j+1}}\cdots q_{l}^{s_{l}}. But this implies m=qjsjxm+ξy(m1++mj1+mj+1++ml)m=q_{j}^{s_{j}}xm+\xi y(m_{1}+\cdots+m_{j-1}+m_{j+1}+\cdots+m_{l}), that is, m=0m=0. Hence, we have that M=p-primeAnnM(p)M=\bigoplus_{{p{\text{-}\mathrm{prime}}}}\mathrm{Ann}^{\ast}_{M}(p). Now, if MM has a proper essential submodule EE, then for some prime pp, EAnnM(p)E\cap\mathrm{Ann}^{\ast}_{M}(p) is proper essential in AnnM(p)\mathrm{Ann}^{\ast}_{M}(p) (see theorem (2.1)). If AnnM(pi)=AnnM(pj)\mathrm{Ann}_{M}(p^{i})=\mathrm{Ann}_{M}(p^{j}) for all positive integers i,ji,j, then observe that AnnR(AnnM(p))=p\mathrm{Ann}_{R}(\mathrm{Ann}_{M}(p))=\langle p\rangle. Thus, using lemma (2.5), we conclude R/pR/\langle p\rangle-vector space AnnM(p)\mathrm{Ann}_{M}(p) has a proper essential subspace EAnnM(p)E\cap\mathrm{Ann}_{M}(p). Again, applying the second part of lemma (2.5), we get a contradiction.

Conversely, assume that AnnM(pi)AnnM(pj)\mathrm{Ann}_{M}(p^{i})\subsetneq\mathrm{Ann}_{M}(p^{j}) for some prime pp and for some positive integers i,ji,j with i<ji<j. We claim that AnnM(pi)\mathrm{Ann}_{M}(p^{i}) is proper essential in AnnM(p)\mathrm{Ann}_{M}(p). Let NN be any submodule of AnnM(p)\mathrm{Ann}_{M}(p) and let nNn\in N. If nAnnM(pi)n\notin\mathrm{Ann}_{M}(p^{i}), then suppose ll be the smallest positive integer such that nAnnM(pl)n\in\mathrm{Ann}_{M}(p^{l}). Observe now that plinAnnM(pi)p^{l-i}n\in\mathrm{Ann}_{M}(p^{i}) and plin0p^{l-i}n\neq 0. This implies plinNAnnM(pi)p^{l-i}n\in N\cap\mathrm{Ann}_{M}(p^{i}). Now apply theorem (2.1) to conclude that MM has a proper essential submodule. ∎

If RR is an integral domain, then a torsion-free left RR-module MM contains a copy of RR as a left RR-module. Also, the RR-module D1RD^{-1}R has a proper essential submodule. Therefore, it is natural to consider a torsion-free RR-module MM that contains D1RD^{-1}R as its submodule. In this regard, the question that arises is whether MM contains a proper essential submodule. The following theorem answers this question.

Theorem 2.8.

Let MM be a left torsion-free module over an integral domain RR and let DD denotes the set of all nonzero elements in RR. If MM contains D1RD^{-1}R as its submodule, then the RR-module MM has a proper essential submodule.

Proof.

Let rdD1R\frac{r}{d}\in D^{-1}R be such that rd0M\frac{r}{d}\neq 0_{M}. Now, consider the collection 𝒞={NMRN,rdN}\mathcal{C}=\{N\leq M\mid R\subseteq N,\frac{r}{d}\notin N\} of submodules of MM. Clearly 𝒞\mathcal{C} is a partially ordered set under inclusion. By Zorn’s Lemma, it has a maximal element 𝔼\mathbbm{E}. We claim that 𝔼\mathbbm{E} is a proper essential submodule of MM. If not, there exists a non-trivial submodule KK that intersects 𝔼\mathbbm{E} trivially, that is, 𝔼K={0M}\mathbbm{E}\cap K=\{0_{M}\}. Observe that KK must not contain any nonzero element in D1RD^{-1}R, because, if rdD1R\frac{r^{\prime}}{d^{\prime}}\in D^{-1}R is such that rd0M\frac{r^{\prime}}{d^{\prime}}\neq 0_{M} and rdK\frac{r^{\prime}}{d^{\prime}}\in K, then rKr^{\prime}\in K, contrary to our assumption that 𝔼K={0M}\mathbbm{E}\cap K=\{0_{M}\}. Let αK\alpha\in K be such that α0M\alpha\neq 0_{M}. Clearly, we have 𝔼α={0M}\mathbbm{E}\cap\langle\alpha\rangle=\{0_{M}\}. Consider the submodule 𝔼α\mathbbm{E}\oplus\langle\alpha\rangle. By maximality of 𝔼\mathbb{E}, it must be that rd𝔼α\frac{r}{d}\in\mathbbm{E}\oplus\langle\alpha\rangle. This implies r/d+α𝔼αr/d+\alpha\in\mathbbm{E}\oplus\langle\alpha\rangle. If rd+α𝔼\frac{r}{d}+\alpha\in\mathbbm{E}, then r+dα𝔼r+d\alpha\in\mathbbm{E}. But this gives, dα𝔼d\alpha\in\mathbbm{E}, a contradiction. Also, if rd+αα\frac{r}{d}+\alpha\in\langle\alpha\rangle, then rαr\in\langle\alpha\rangle, a contradiction. Thus, we can write rd+α=e+rα\frac{r}{d}+\alpha=e+r^{\ast}\alpha, where e𝔼{0M}e\in\mathbbm{E}\setminus\{0_{M}\} and rR{0R}r^{\ast}\in R\setminus\{0_{R}\}. But this implies r+dα=de+drαr+d\alpha=de+dr^{\ast}\alpha. Hence, we get that rde=d(r1)αr-de=d(r^{\ast}-1)\alpha. Observe that rde𝔼r-de\in\mathbbm{E} whereas d(r1)ααd(r^{\ast}-1)\alpha\in\langle\alpha\rangle. Hence, it must be that rde=0Mr-de=0_{M} and d(r1)=0Rd(r^{\ast}-1)=0_{R}. As d0Rd\neq 0_{R}, it must be that r=1r^{\ast}=1. Plugging the value of rr^{\ast} in the equation rd+α=e+rα\frac{r}{d}+\alpha=e+r^{\ast}\alpha yields rd=e\frac{r}{d}=e. But this is absurd because 𝔼\mathbbm{E} being a member of 𝒞\mathcal{C} cannot contain rd\frac{r}{d}. ∎

Another question that arises naturally from the definition of proper essential submodule is: if MM has a proper essential submodule, does the quotient module M/NM/N also have a proper essential submodule where NN is any proper non-trivial submodule of MM? In general, the answer is no (example: \mathbb{Z}-module p2\mathbb{Z}_{p^{2}}). But under certain constraints, we can give an affirmative answer. The converse, however, always holds.

Lemma 2.9.

Let MM be a left RR-module and NN be a submodule of MM such that M/NM/N has a proper essential submodule E/NE/N. Then, the submodule EE is a proper essential submodule of MM. Moreover, if a module MM has a proper essential submodule EE and if NN is a submodule of MM such that M/NM/N is torsion-free and NEN\neq E, then the submodule (E+N)/N(E+N)/N is proper essential in M/NM/N.

Proof.

If N={0M}N=\{0_{M}\}, there is nothing to show. Assume N{0M}N\neq\{0_{M}\}. Let LL be a nontrivial proper submodule of MM and u(L+N)/NE/Nu\in(L+N)/N\cap E/N. Clearly, u=l+N=e+Nu=l+N=e+N for some lLNl\in L\setminus N and eENe\in E\setminus N. But this shows that lEl\in E. Hence, we get that EE is proper essential in MM.

If N={0M}N=\{0_{M}\}, there is nothing to show. Therefore, assume N{0M}N\neq\{0_{M}\}. Let L/NL/N be a non-trivial proper submodule of M/NM/N. Let lLNl\in L\setminus N be such rlErl\in E, for some nonzero rRr\in R and rl0Mrl\neq 0_{M}. As rlNrl\notin N, we conclude that rl+N(E+N)/NL/Nrl+N\in(E+N)/N\cap L/N, that is, (E+N)/N(E+N)/N is proper essential in M/NM/N. ∎

Corollary 2.10.

Let MM be a left module over an integral domain RR that contains D1RD^{-1}R as its submodule and MTor(M)M\neq\mathrm{Tor}(M). Then, the RR-module MM has a proper essential submodule.

Proof.

Consider the quotient module M/Tor(M)M/\mathrm{Tor}(M). As M/Tor(M)M/\mathrm{Tor}(M) is torsion-free, so applying theorem (2.8) we get that M/Tor(M)M/\mathrm{Tor}(M) has a proper essential submodule. Again, applying lemma (2.9), we see that MM has a proper essential submodule. ∎

The correspondence between elements in HomR(M,N)\mathrm{Hom}_{R}(M,N) with matrices over RR plays a fundamental role in the theory of modules. Keeping this in mind, the most natural question that we investigate is whether the set of matrices (of some dimension) over RR (as an RR-module) has a proper essential submodule. We obtain an affirmative answer in the following theorem. Let m×n(R)\mathcal{M}_{m\times n}(R) denote the set of all m×nm\times n matrices over RR.

Theorem 2.11.

Let RR be an integral domian. Then the left RR-module m×n(R)\mathcal{M}_{m\times n}(R) possess a proper essential submodule.

Proof.

Let MM and NN are the free RR modules on the sets of nn and mm elements respectively. Clearly, we have MRnM\simeq R^{n} and NRmN\simeq R^{m}. As HomR(Rn,Rm)HomR(M,N)\mathrm{Hom}_{R}(R^{n},R^{m})\simeq\mathrm{Hom}_{R}(M,N) and HomR(M,N)m×n(R)\mathrm{Hom}_{R}(M,N)\simeq\mathcal{M}_{m\times n}(R), we get that HomR(Rn,Rm)m×n(R)\mathrm{Hom}_{R}(R^{n},R^{m})\simeq\mathcal{M}_{m\times n}(R). Since we have HomR(Rn,Rm)i=1nj=1mHomR(R,R)\mathrm{Hom}_{R}(R^{n},R^{m})\simeq\bigoplus_{i=1}^{n}\bigoplus_{j=1}^{m}\mathrm{Hom}_{R}(R,R), we obtain m×n(R)Rmn\mathcal{M}_{m\times n}(R)\simeq R^{mn}. Now, we can apply theorem (2.2) to obtain the desired result. ∎

3 Essential socles in a new class of modules

Definition 3.1.

A left RR-module MM is said to be an SM\mathrm{SM} module if every nontrivial submodule of MM contains a simple submodule.

Proposition 3.2.

Let MM be a left module over RR such that MM is either Artinian, or, locally finite, or, is of finite length. Then MM is an SM\mathrm{SM} module.

Proof.

Let MM be an Artinian module. If MM is simple there is nothing to show. Therefore assume MM is not simple. Let NN be any submodule of MM. If we consider any strictly descending chain CN:N0=NN1N2NsC_{N}:N_{0}=N\supsetneq N_{1}\supseteq N_{2}\supsetneq\cdots\supsetneq N_{s} of submodules of NN that terminates with the nontrivial proper submodule NsN_{s}, then NsN_{s} is a simple module contained in NN. Because, otherwise one can enlarge the chain CNC_{N}.

Now assume MM is a locally finite module. Again, if MM is simple then MM must be finite and there is nothing to show. So, assume MM is not simple. Let NN be any submodule of MM and let nNn\in N. Let 𝒢\mathcal{G} be the collection of all submodules of n\langle n\rangle. As n\langle n\rangle is finite, we can choose S𝒢S\in\mathcal{G} such that |S||S| is least. Clearly, SS can not contain any nontrivial proper submodule LL, because then LL being a member of 𝒢\mathcal{G} contradicts the minimality of |S||S|. This shows that SS is a simple submodule contained in NN.

Finally, assume that the length of MM is finite. If l(M)=1l(M)=1, then MM is simple and there is nothing to show. So, assume l(M)>1l(M)>1. Let NN be any submodule of MM. Let \mathcal{H} be the collection of all nontrivial proper submodule of NN. Choose KK\in\mathcal{H} such that l(K)l(K) is minimal. If KK would not be simple and PP is a nontrivial proper submodule of KK, then l(P)<l(K)l(P)<l(K) contradicts the minimality of l(K)l(K). Thus, KK is a simple submodule contained in NN. ∎

Theorem 3.3.

The class of left SM\mathrm{SM} modules properly contains the class of Artinian modules, locally finite modules, and modules of finite lengths.

Proof.

Let 𝒞1,𝒞2,𝒞3\mathcal{C}_{1},\mathcal{C}_{2},\mathcal{C}_{3}, and 𝒞4\mathcal{C}_{4} denote the class of all SM\mathrm{SM} modules, modules of finite lengths, locally finite modules, and Artinian modules respectively. Consider the free pn\mathbb{Z}_{p^{n}}-module M=pnM=\bigoplus_{\mathbb{N}}\mathbb{Z}_{p^{n}} on the set \mathbb{N}, where pp is some fixed prime and nn is a natural number such that n>1n>1. Let NN be any nontrivial proper submodule of MM and let nNn\in N. Observe that |n|pn|\langle n\rangle|\leq p^{n}. If we choose a submodule LL of n\langle n\rangle such that |L|=p|L|=p, then clearly LL is simple and contained in NN. This shows that MM is an SM\mathrm{SM} module.

As l(pn)=nl(\mathbb{Z}_{p^{n}})=n, therefore the divergence of the series i=1n\sum_{i=1}^{\infty}n shows that the length of the free pn\mathbb{Z}_{p^{n}}-module pn\bigoplus_{\mathbb{N}}\mathbb{Z}_{p^{n}} is not finite. Hence, we have that 𝒞2𝒞1\mathcal{C}_{2}\subsetneq\mathcal{C}_{1}.

For each natural number ii, we define the submodules Li=FjL_{i}=\bigoplus_{\mathbb{N}}F_{j}, where Fj=pnF_{j}=\mathbb{Z}_{p^{n}} if j>ij>i, otherwise, Fj={0pn}F_{j}=\{0_{\mathbb{Z}_{p^{n}}}\}. Now, observe that the strictly descending chain of submodules L0=ML1L2LsL_{0}=M\supsetneq L_{1}\supsetneq L_{2}\supsetneq\cdots L_{s}\supsetneq\cdots do not terminate. Thus, 𝒞4𝒞1\mathcal{C}_{4}\subsetneq\mathcal{C}_{1}.

Now, consider an infinite RR-module WW of finite length nn (for example finite dimensional vector space). If W0=WW1W2Wn={0W}W_{0}=W\supsetneq W_{1}\supsetneq W_{2}\supseteq\cdots\supsetneq W_{n}=\{0_{W}\}, then observe that W=R(A)W=R(A), where A={wiwiWiWi+1,0in2}A=\{w_{i}\mid w_{i}\in W_{i}\setminus W_{i+1},0\leq i\leq n-2\} and R(A)R(A) is the module generated by the subset AA. As W𝒞3W\notin\mathcal{C}_{3}, therefore, we get that 𝒞3𝒞1\mathcal{C}_{3}\subsetneq\mathcal{C}_{1}. ∎

Theorem 3.4.

Let MM be a left SM\mathrm{SM} module over RR. Then Soc(M)\mathrm{Soc}(M) is proper essential in MM if and only if MM is not semisimple.

Proof.

Assume first that Soc(M)\mathrm{Soc}(M) is proper essential in MM. Then, clearly we have Soc(M)M\mathrm{Soc}(M)\subsetneq M. Hence, applying theorem (1.4), we conclude that MM is not semisimple.

Conversely, assume that MM is not semisimple. Again, applying theorem (1.4), we see that MM contains a proper essential submodule. Now, let NN be any submodule of MM. As NN contains a simple submodule, we get that Soc(M)N{0M}\mathrm{Soc}(M)\cap N\neq\{0_{M}\}. Further, as Soc(M)\mathrm{Soc}(M) is contained in every proper essential submodule of MM, we conclude that Soc(M)\mathrm{Soc}(M) is proper essential in MM. ∎

4 Torsion-free modules over integral domains

In general, it is not clear from theorem (2.8) that under what condition a left torsion-free module MM over an integral domain RR contains an essential submodule. For this purpose, we define the closure of a submodule NN of MM that gives some sort of measure in the sense that it is that maximal subspace of MM in which NN is essential.

Definition 4.1.

Let MM be a left torsion-free module over an integral domain RR. For any submodule NN of MM, we denote ClM(N)\mathrm{Cl}_{M}(N) to mean the closure of NN in MM defined in the following way: ClM(N)={mMrmN,for some nonzero rR}\mathrm{Cl}_{M}(N)=\{m\in M\mid rm\in N,\text{for some nonzero $r\in R$}\}.

Throughout this section, unless otherwise stated, we assume MM is a left torsion-free module over an integral domain RR and DD is the set of nonzero elements in RR. The following proposition is quite straightforward and follows directly from definition (4.1).

Proposition 4.2.

Let N,N1,N2,TN,N_{1},N_{2},T are submodules of MM and let f:MSf:M\rightarrow S be any homomorphism of modules. Then, we have the following:

(a)\mathrm{(a)} ClM({0M})={0M}\mathrm{Cl}_{M}(\{0_{M}\})=\{0_{M}\} and ClM(N)\mathrm{Cl}_{M}(N) is a submodule of MM.

(b)\mathrm{(b)} NClM(N)N\subseteq\mathrm{Cl}_{M}(N) and ClM(ClM(N))=ClM(N)\mathrm{Cl}_{M}(\mathrm{Cl}_{M}(N))=\mathrm{Cl}_{M}(N).

(c)\mathrm{(c)} If N1N2N_{1}\subseteq N_{2}, then ClM(N1)ClM(N2)\mathrm{Cl}_{M}(N_{1})\subseteq\mathrm{Cl}_{M}(N_{2}).

(d)\mathrm{(d)} f(ClM(T))ClS(f(T))f(\mathrm{Cl}_{M}(T))\leq\mathrm{Cl}_{S}(f(T)).

Proof.

(a):\mathrm{(a)}: ClM({0M})\mathrm{Cl}_{M}(\{0_{M}\}) is the set of all those elements mm in MM such that rm=0rm=0 for some nonzero rRr\in R. As MM is torsion-free, we see that ClM({0M})={0M}\mathrm{Cl}_{M}(\{0_{M}\})=\{0_{M}\}. If m1,m2m_{1},m_{2} are in ClM(N)\mathrm{Cl}_{M}(N), then r1m1Nr_{1}m_{1}\in N and r2m2Nr_{2}m_{2}\in N for some nonzero elements r1,r2r_{1},r_{2} in RR. Since r1r2(m1m2)Nr_{1}r_{2}(m_{1}-m_{2})\in N and r1r20r_{1}r_{2}\neq 0, we conclude that (m1m2)ClM(N)(m_{1}-m_{2})\in\mathrm{Cl}_{M}(N). Now, let mClM(N)m\in\mathrm{Cl}_{M}(N) and rRr\in R be such that r0r\neq 0 and rmNrm\in N. For any sRs\in R, we have r(sm)Nr(sm)\in N. Thus, smClM(N)sm\in\mathrm{Cl}_{M}(N). Hence, ClM(N)\mathrm{Cl}_{M}(N) is a submodule of MM.

(b):\mathrm{(b)}: For any element nNn\in N, we have 1nN1n\in N, so NClM(N)N\subseteq\mathrm{Cl}_{M}(N). It is evident that ClM(N)ClM(ClM(N))\mathrm{Cl}_{M}(N)\subseteq\mathrm{Cl}_{M}(\mathrm{Cl}_{M}(N)). Now, suppose mClM(ClM(N))m\in\mathrm{Cl}_{M}(\mathrm{Cl}_{M}(N)). This implies, for some nonzero rRr\in R, rmrm lie in ClM(N)\mathrm{Cl}_{M}(N). Again, for some nonzero sRs\in R, (sr)mN(sr)m\in N. As, sr0sr\neq 0, we conclude that mClM(N)m\in\mathrm{Cl}_{M}(N). Hence, ClM(ClM(N))ClM(N)\mathrm{Cl}_{M}(\mathrm{Cl}_{M}(N))\subseteq\mathrm{Cl}_{M}(N).

(c):\mathrm{(c)}: Let mClM(N1)m\in\mathrm{Cl}_{M}(N_{1}). Then, rmN1rm\in N_{1} for some nonzero rRr\in R. As rmN2rm\in N_{2} as well, we conclude that mClM(N2)m\in\mathrm{Cl}_{M}(N_{2}), that is, ClM(N1)ClM(N2)\mathrm{Cl}_{M}(N_{1})\subseteq\mathrm{Cl}_{M}(N_{2}).

(d):\mathrm{(d)}: Let sf(ClM(T))s\in f(\mathrm{Cl}_{M}(T)). Then, there exists mClM(T)m\in\mathrm{Cl}_{M}(T) such that f(m)=sf(m)=s. Also, there exists rRr\in R such that r0r\neq 0 and rmTrm\in T. This shows that rf(m)f(T)rf(m)\in f(T), that is, sClS(f(T))s\in\mathrm{Cl}_{S}(f(T)). Hence, we get that f(ClM(T))ClS(f(T))f(\mathrm{Cl}_{M}(T))\leq\mathrm{Cl}_{S}(f(T)). ∎

The question that we investigate is whether the closure operator commutes with arbitrary direct sums of submodules. We see that it is not true in general, but in bigger space, the closure operator commutes with arbitrary direct sum of submodules.

Lemma 4.3.

Let N1,N2,,NnN_{1},N_{2},\ldots,N_{n} are the submodules of MM. If DD denotes the set of all nonzero elements in RR, then ClD1M(i=1nNi)=i=1nClD1M(Ni)\mathrm{Cl}_{D^{-1}M}\big{(}\bigoplus_{i=1}^{n}N_{i}\big{)}=\bigoplus_{i=1}^{n}\mathrm{Cl}_{D^{-1}M}(N_{i}).

Proof.

It is enough to show that ClD1M(LS)=ClD1M(L)ClD1M(S)\mathrm{Cl}_{D^{-1}M}(L\bigoplus S)=\mathrm{Cl}_{D^{-1}M}(L)\bigoplus\mathrm{Cl}_{D^{-1}M}(S) for any two submodules L,SL,S of MM. Observe that if mdClD1M(L)ClD1M(S)\frac{m}{d}\in\mathrm{Cl}_{D^{-1}M}(L)\cap\mathrm{Cl}_{D^{-1}M}(S), then r1mdL\frac{r_{1}m}{d}\in L and r2mdS\frac{r_{2}m}{d}\in S for some nonzero r1,r2Rr_{1},r_{2}\in R. As (r1r2)mdLS(r_{1}r_{2})\frac{m}{d}\in L\cap S, we see that md=0M\frac{m}{d}=0_{M}. Now, suppose (m1d1+m2d2)ClD1M(L)ClD1M(S)(\frac{m_{1}}{d_{1}}+\frac{m_{2}}{d_{2}})\in\mathrm{Cl}_{D^{-1}M}(L)\bigoplus\mathrm{Cl}_{D^{-1}M}(S). We have then r1m1d1L\frac{r_{1}m_{1}}{d_{1}}\in L and r2m2d2S\frac{r_{2}m_{2}}{d_{2}}\in S for some r1,r2R{0}r_{1},r_{2}\in R\setminus\{0\}. As (r1r2)(m1d1+m2d2)LS(r_{1}r_{2})(\frac{m_{1}}{d_{1}}+\frac{m_{2}}{d_{2}})\in L\bigoplus S, we see that (m1d1+m2d2)ClD1M(LS)(\frac{m_{1}}{d_{1}}+\frac{m_{2}}{d_{2}})\in\mathrm{Cl}_{D^{-1}M}(L\bigoplus S). Conversely, let mdClD1M(LS)\frac{m}{d}\in\mathrm{Cl}_{D^{-1}M}(L\bigoplus S). Therefore, for some nonzero rRr\in R, we have rmdLS\frac{rm}{d}\in L\bigoplus S. Let rmd=l+s\frac{rm}{d}=l+s. Then, r(mdlr)=sr(\frac{m}{d}-\frac{l}{r})=s, that is, (mdlr)ClD1M(S)(\frac{m}{d}-\frac{l}{r})\in\mathrm{Cl}_{D^{-1}M}(S). Also, observe that lrClD1M(L)\frac{l}{r}\in\mathrm{Cl}_{D^{-1}M}(L). Hence, we get that md=lr+(mdlr)\frac{m}{d}=\frac{l}{r}+(\frac{m}{d}-\frac{l}{r}), that is, mdClD1M(L)ClD1M(S)\frac{m}{d}\in\mathrm{Cl}_{D^{-1}M}(L)\bigoplus\mathrm{Cl}_{D^{-1}M}(S). ∎

Theorem 4.4.

If {Nα}αΛ\{N_{\alpha}\}_{\alpha\in\Lambda} be a family of submodules indexed by a nonempty set Λ\Lambda, then ClD1M(αΛNα)=αΛClD1M(Nα)\mathrm{Cl}_{D^{-1}M}\big{(}\bigoplus_{\alpha\in\Lambda}N_{\alpha}\big{)}=\bigoplus_{\alpha\in\Lambda}\mathrm{Cl}_{D^{-1}M}(N_{\alpha}). In particular, we have the following, ClM(αΛNα)=M(D1(αΛClM(Nα)))\mathrm{Cl}_{M}\big{(}\bigoplus_{\alpha\in\Lambda}N_{\alpha}\big{)}=M\cap\big{(}D^{-1}\big{(}\bigoplus_{\alpha\in\Lambda}\mathrm{Cl}_{M}(N_{\alpha})\big{)}\big{)}.

Proof.

Let mdClD1M(αΛNα)\frac{m}{d}\in\mathrm{Cl}_{D^{-1}M}\big{(}\bigoplus_{\alpha\in\Lambda}N_{\alpha}\big{)}. Then, for some r0r\neq 0, we have rmdαΛNα\frac{rm}{d}\in\bigoplus_{\alpha\in\Lambda}N_{\alpha}. Let rmd=i=1nnαi\frac{rm}{d}=\sum_{i=1}^{n}n_{\alpha_{i}}. This implies, mdClD1M(i=1nNαi)\frac{m}{d}\in\mathrm{Cl}_{D^{-1}M}(\bigoplus_{i=1}^{n}N_{\alpha_{i}}). Now, applying lemma (4.3), we get that ClD1M(i=1nNαi)αΛClD1M(Nα)\mathrm{Cl}_{D^{-1}M}(\bigoplus_{i=1}^{n}N_{\alpha_{i}})\leq\bigoplus_{\alpha\in\Lambda}\mathrm{Cl}_{D^{-1}M}(N_{\alpha}). Hence, we obtain that mdαΛClD1M(Nα)\frac{m}{d}\in\bigoplus_{\alpha\in\Lambda}\mathrm{Cl}_{D^{-1}M}(N_{\alpha}). Conversely, suppose i=1nmαidαiαΛClD1M(Nα)\sum_{i=1}^{n}\frac{m_{\alpha_{i}}}{d_{\alpha_{i}}}\in\bigoplus_{\alpha\in\Lambda}\mathrm{Cl}_{D^{-1}M}(N_{\alpha}), where mαidαiClD1M(Nαi)\frac{m_{\alpha_{i}}}{d_{\alpha_{i}}}\in\mathrm{Cl}_{D^{-1}M}(N_{\alpha_{i}}). Then, for each ii with 1in1\leq i\leq n, there exists some nonzero rαir_{\alpha_{i}} such that rαimαidαiNαi\frac{r_{\alpha_{i}}m_{\alpha_{i}}}{d_{\alpha_{i}}}\in N_{\alpha_{i}}. But then (rα1rα2rαn)(i=1nmαidαi)i=1nNαi(r_{\alpha_{1}}r_{\alpha_{2}}\cdots r_{\alpha_{n}})\big{(}\sum_{i=1}^{n}\frac{m_{\alpha_{i}}}{d_{\alpha_{i}}}\big{)}\in\bigoplus_{i=1}^{n}N_{\alpha_{i}} shows that (rα1rα2rαn)(i=1nmαidαi)αΛNα(r_{\alpha_{1}}r_{\alpha_{2}}\cdots r_{\alpha_{n}})\big{(}\sum_{i=1}^{n}\frac{m_{\alpha_{i}}}{d_{\alpha_{i}}}\big{)}\in\bigoplus_{\alpha\in\Lambda}N_{\alpha}, that is, i=1nmαidαiClD1M(αΛNα)\sum_{i=1}^{n}\frac{m_{\alpha_{i}}}{d_{\alpha_{i}}}\in\mathrm{Cl}_{D^{-1}M}\big{(}\bigoplus_{\alpha\in\Lambda}N_{\alpha}\big{)}.

For any submodule NN of MM, if we show that ClD1M(N)=D1ClM(N)\mathrm{Cl}_{D^{-1}M}(N)=D^{-1}\mathrm{Cl}_{M}(N), then we have αΛClD1M(Nα)=D1(αΛClM(Nα))\bigoplus_{\alpha\in\Lambda}\mathrm{Cl}_{D^{-1}M}(N_{\alpha})=D^{-1}\big{(}\bigoplus_{\alpha\in\Lambda}\mathrm{Cl}_{M}(N_{\alpha})\big{)}, since D1D^{-1} distributes over arbitrary direct sum. So, suppose mdD1ClM(N)\frac{m}{d}\in D^{-1}\mathrm{Cl}_{M}(N). As mClM(N)m\in\mathrm{Cl}_{M}(N), we have rmNrm\in N for some nonzero rRr\in R. But then, we have (rd)(md)N(rd)(\frac{m}{d})\in N. Hence, mdClD1M(N)\frac{m}{d}\in\mathrm{Cl}_{D^{-1}M}(N). Conversely, suppose mdClD1M(N)\frac{m}{d}\in\mathrm{Cl}_{D^{-1}M}(N). Then, rmdN\frac{rm}{d}\in N for some nonzero rRr\in R. This implies, rmNrm\in N, that is, mClM(N)m\in\mathrm{Cl}_{M}(N). Hence, we get that mdD1ClM(N)\frac{m}{d}\in D^{-1}\mathrm{Cl}_{M}(N). The proof is now completed by taking intersection on both sides of the equation ClD1M(αΛNα)=D1(αΛClM(Nα))\mathrm{Cl}_{D^{-1}M}\big{(}\bigoplus_{\alpha\in\Lambda}N_{\alpha}\big{)}=D^{-1}\big{(}\bigoplus_{\alpha\in\Lambda}\mathrm{Cl}_{M}(N_{\alpha})\big{)} with MM. ∎

Theorem 4.5.

If NN is a submodule of MM, then Soc(ClM(N))=ClM(Soc(N))\mathrm{Soc}\big{(}\mathrm{Cl}_{M}(N)\big{)}=\mathrm{Cl}_{M}\big{(}\mathrm{Soc}(N)\big{)}. In particular, they are equal to Soc(N)\mathrm{Soc}(N).

Proof.

If SS is a simple submodule of NN then clearly it is also simle in ClM(N)\mathrm{Cl}_{M}(N). Conversely, if S=tS=\langle t\rangle is a simple submodule of ClM(N)\mathrm{Cl}_{M}(N), then rtNrt\in N for some nonzero rRr\in R. As rt=t\langle rt\rangle=\langle t\rangle, we conclude that SS is a simple submodule of NN. Hence, we conclude that Soc(ClM(N))=Soc(N)\mathrm{Soc}\big{(}\mathrm{Cl}_{M}(N)\big{)}=\mathrm{Soc}(N). Let Soc(N)=αΛSα\mathrm{Soc}(N)=\bigoplus_{\alpha\in\Lambda}S_{\alpha}, where SαS_{\alpha} is a simple submodule of NN for each αΛ\alpha\in\Lambda. Now, using theorem (4.4), we get that ClD1M(Soc(N))=αΛClD1M(Sα)\mathrm{Cl}_{D^{-1}M}\big{(}\mathrm{Soc}(N)\big{)}=\bigoplus_{\alpha\in\Lambda}\mathrm{Cl}_{D^{-1}M}(S_{\alpha}). As simple submodules of MM are simple submodules of D1MD^{-1}M and vice-versa, we see that SαS_{\alpha} is a simple submodule of D1MD^{-1}M for each αΛ\alpha\in\Lambda. In this position, we wish to use the fact that for any left torsion-free module ZZ (over some integral domain BB) and for any simple submodule LL of ZZ, we have ClZ(L)=L\mathrm{Cl}_{Z}(L)=L. To see this, let zClZ(L)z\in\mathrm{Cl}_{Z}(L) and bzLbz\in L for some nonzero bBb\in B. Let L=lL=\langle l\rangle. Clearly, we have l=bl=bz\langle l\rangle=\langle bl\rangle=\langle bz\rangle. If bz=blbz=bl, then b(zl)=0b(z-l)=0. As ZZ is torsion-free and b0b\neq 0, we conclude that z=lz=l, that is zLz\in L. If bzblbz\neq bl, then for some nonzero bBb^{\prime}\in B, we have bz=(bb)lbz=(b^{\prime}b)l, that is, b(zbl)=0b(z-b^{\prime}l)=0. Similar arguments as above shows that z=blz=b^{\prime}l, that is, zLz\in L. Hence, ClZ(L)L\mathrm{Cl}_{Z}(L)\subseteq L. Now, applying proposition (4.2), we conclude that ClZ(L)=L\mathrm{Cl}_{Z}(L)=L. Thus, we have that ClD1M(Soc(N))=αΛSα\mathrm{Cl}_{D^{-1}M}\big{(}\mathrm{Soc}(N)\big{)}=\bigoplus_{\alpha\in\Lambda}S_{\alpha}, that is, ClD1M(Soc(N))=Soc(N)\mathrm{Cl}_{D^{-1}M}\big{(}\mathrm{Soc}(N)\big{)}=\mathrm{Soc}(N). Now, MClD1M(Soc(N))=ClM(Soc(N))M\cap\mathrm{Cl}_{D^{-1}M}\big{(}\mathrm{Soc}(N)\big{)}=\mathrm{Cl}_{M}\big{(}\mathrm{Soc}(N)\big{)} shows that ClM(Soc(N))=Soc(N)\mathrm{Cl}_{M}\big{(}\mathrm{Soc}(N)\big{)}=\mathrm{Soc}(N). Hence, Soc(ClM(N))=ClM(Soc(N))=Soc(N)\mathrm{Soc}\big{(}\mathrm{Cl}_{M}(N)\big{)}=\mathrm{Cl}_{M}\big{(}\mathrm{Soc}(N)\big{)}=\mathrm{Soc}(N). ∎

Corollary 4.6.

Left torsion-free modules over integral domains are either semisimple or do not possess essential socles.

Proof.

If MM is not semisimple then Soc(M)M\mathrm{Soc}(M)\neq M (see theorem (1.3)). Now, if we set N=MN=M in theorem (4.5), we see that ClM(Soc(M))=Soc(M)\mathrm{Cl}_{M}\big{(}\mathrm{Soc}(M)\big{)}=\mathrm{Soc}(M). Thus, the socle Soc(M)\mathrm{Soc}(M) is not essential in MM. ∎

Theorem 4.7.

For submodules U,VU,V of MM, we have the following exact sequence: 0D1ClM(U)αD1ClM(V)βD1ClM/U(V/U)00\rightarrow D^{-1}\mathrm{Cl}_{M}(U)\xrightarrow{\alpha^{\prime}}D^{-1}\mathrm{Cl}_{M}(V)\xrightarrow{\beta^{\prime}}D^{-1}\mathrm{Cl}_{M/U}(V/U)\rightarrow 0. Moreover, it splits if the sequence 0U𝛼V𝛽V/U00\rightarrow U\xrightarrow{\alpha}V\xrightarrow{\beta}V/U\rightarrow 0 splits.

Proof.

Let φ:ClD1M/D1U(D1V/D1U)ClD1M(D1V)/ClD1M(D1U)\varphi:\mathrm{Cl}_{D^{-1}M/D^{-1}U}(D^{-1}V/D^{-1}U)\rightarrow\mathrm{Cl}_{D^{-1}M}(D^{-1}V)/\mathrm{Cl}_{D^{-1}M}(D^{-1}U) be a map defined by φ:md+D1Umd+ClD1M(D1U)\varphi:\frac{m}{d}+D^{-1}U\mapsto\frac{m}{d}+\mathrm{Cl}_{D^{-1}M}(D^{-1}U). Observe that we have md+ClD1M(D1U)ClD1M(D1V)/ClD1M(D1U)\frac{m}{d}+\mathrm{Cl}_{D^{-1}M}(D^{-1}U)\in\mathrm{Cl}_{D^{-1}M}(D^{-1}V)/\mathrm{Cl}_{D^{-1}M}(D^{-1}U). Because, for some nonzero rRr\in R, we have rmd+D1U=vd+D1U\frac{rm}{d}+D^{-1}U=\frac{v}{d^{\prime}}+D^{-1}U. But this implies r(mdvrd)D1Ur(\frac{m}{d}-\frac{v}{rd^{\prime}})\in D^{-1}U, that is, (mdvrd)ClD1M(D1U)(\frac{m}{d}-\frac{v}{rd^{\prime}})\in\mathrm{Cl}_{D^{-1}M}(D^{-1}U). As vrdClD1M(D1V)\frac{v}{rd^{\prime}}\in\mathrm{Cl}_{D^{-1}M}(D^{-1}V), md+ClD1M(D1U)\frac{m}{d}+\mathrm{Cl}_{D^{-1}M}(D^{-1}U) is an element of ClD1M(D1V)/ClD1M(D1U)\mathrm{Cl}_{D^{-1}M}(D^{-1}V)/\mathrm{Cl}_{D^{-1}M}(D^{-1}U). To see that φ\varphi is well-defined, observe that for elements md+D1U,md+D1U\frac{m}{d}+D^{-1}U,\frac{m^{\prime}}{d^{\prime}}+D^{-1}U in ClD1M/D1U(D1V/D1U)\mathrm{Cl}_{D^{-1}M/D^{-1}U}(D^{-1}V/D^{-1}U), if we have (mdmd)D1U(\frac{m}{d}-\frac{m^{\prime}}{d^{\prime}})\in D^{-1}U, then md+ClD1M(D1U)=md+ClD1M(D1U)\frac{m}{d}+\mathrm{Cl}_{D^{-1}M}(D^{-1}U)=\frac{m^{\prime}}{d^{\prime}}+\mathrm{Cl}_{D^{-1}M}(D^{-1}U), since proposition (4.2) yields D1UClD1M(D1U)D^{-1}U\subseteq\mathrm{Cl}_{D^{-1}M}(D^{-1}U). It is easy to see that φ\varphi is a module homomorphism. If md+D1UClD1M/D1U(D1V/D1U)\frac{m}{d}+D^{-1}U\in\mathrm{Cl}_{D^{-1}M/D^{-1}U}(D^{-1}V/D^{-1}U) be such that mdClD1M(D1U)\frac{m}{d}\in\mathrm{Cl}_{D^{-1}M}(D^{-1}U), then rmdD1U\frac{rm}{d}\in D^{-1}U for some nonzero rRr\in R, that is, mdD1U\frac{m}{d}\in D^{-1}U. This shows that φ\varphi is injective. Finally, for any md+ClD1M(D1U)ClD1M(D1V)/ClD1M(D1U)\frac{m}{d}+\mathrm{Cl}_{D^{-1}M}(D^{-1}U)\in\mathrm{Cl}_{D^{-1}M}(D^{-1}V)/\mathrm{Cl}_{D^{-1}M}(D^{-1}U), we have rmdD1V\frac{rm}{d}\in D^{-1}V for some nonzero rRr\in R. As r(md+D1U)D1V/D1Ur(\frac{m}{d}+D^{-1}U)\in D^{-1}V/D^{-1}U, we conclude that (md+D1U)ClD1M/D1U(D1V/D1U)(\frac{m}{d}+D^{-1}U)\in\mathrm{Cl}_{D^{-1}M/D^{-1}U}(D^{-1}V/D^{-1}U), that is, φ\varphi is surjective. Hence, φ\varphi is an isomorphism. It is now easy to see that ClD1M/D1U(D1V/D1U)=D1ClM/U(V/U)\mathrm{Cl}_{D^{-1}M/D^{-1}U}(D^{-1}V/D^{-1}U)=D^{-1}\mathrm{Cl}_{M/U}(V/U) and ClD1M(D1V)/ClD1M(D1U)=D1(ClM(V)/ClM(U))\mathrm{Cl}_{D^{-1}M}(D^{-1}V)/\mathrm{Cl}_{D^{-1}M}(D^{-1}U)=D^{-1}(\mathrm{Cl}_{M}(V)/\mathrm{Cl}_{M}(U)). This completes the first part of the theorem. Applying lemma (4.3), we can conclude the second part. ∎

We close this section with the following theorem that can also be viewed as an application of corollary (4.6).

Theorem 4.8.

Left modules over integral domains with nonempty torsion-free parts are either semisimple or do not possess essential socles.

Proof.

If MM is semisimple, there is nothing to show. So, assume that MM is not semisimple. This implies MM has a proper essential submodule EE (see theorem (1.3)). Now, applying lemma (2.9), we see that (E+Tor(M))/Tor(M)(E+\mathrm{Tor}(M))/\mathrm{Tor}(M) is proper essential in M/Tor(M)M/\mathrm{Tor}(M). Now, if 𝒫\mathcal{P} be the collection of all essential submodules of M/Tor(M)M/\mathrm{Tor}(M) and if we denote D¯=D/Tor(M)\overline{D}=D/\mathrm{Tor}(M) for D/Tor(M)𝒫D/\mathrm{Tor}(M)\in\mathcal{P}, then from corollary (4.6), we can conclude that the submodule D¯𝒫D¯\bigcap_{\overline{D}\in\mathcal{P}}\overline{D} is not proper essential in M/Tor(M)M/\mathrm{Tor}(M). Since we have (D¯𝒫D)/Tor(M)D¯𝒫D¯\big{(}\bigcap_{\overline{D}\in\mathcal{P}}D\big{)}/\mathrm{Tor}(M)\leq\bigcap_{\overline{D}\in\mathcal{P}}\overline{D}, therefore, we can conclude from lemma (2.9) that the submodule D¯𝒫D\bigcap_{\overline{D}\in\mathcal{P}}D can not be proper essential in MM. Further, since Soc(M)\mathrm{Soc}(M) is a submodule of D¯𝒫D\bigcap_{\overline{D}\in\mathcal{P}}D, we conclude that MM can not possess an essential socle. ∎

References

  • RJ (1951) R. E. Johnson, The extended centralizer of a module, Proc. Amer. Math. Soc., 19511951.
  • ES (1953) B. Eckman and A. Schpof, Uber injective moduln, Arch. Math., 19531953.
  • AWG (1964) A. W. Goldie, Torsion-free modules and rings, J. Algebra, 19641964.
  • JL (1966) J. Lambek, Lectures on Rings and Modules, First edition, Blaisdell, 19661966.
  • DHR (1989) D. V. Nguyen, H. V. Dinh, and R. Wisbauer, Quasi-injective modules with acc or dcc on essential submodules, Arch. Math., 19891989.
  • AF (1992) F.W. Anderson and K.R. Fuller, Rings and Categories of Modules, Graduate Texts in Mathematics, Second edition, Springer Verlag, 19921992.
  • TL (1999) T. Y. Lam, Lectures on modules and rings, First edition, Springer Verlag, 19991999.
  • SV (2004) P. F. Smith and M. R. Vedadi, Modules with chain conditions on non-essential submodules, Comm. Algebra, 20042004.
  • GK (2008) M. Ghirati and O. A. S. Karamzadeh, On strongly essential submodules, Comm. Algebra, 20082008.
  • UR (2014) A. Ulrich and G. Rüdiger, A note on essential extensions and submodules of generators, Houston J. Math., 20142014.
  • CS (2017) M. D. Cisse and D. Sow, On generalizations of essential and small submodules, Southeast Asian Bull. Math., 20172017.
  • MD (2019) M. Davoudian, Dimension on non-essential submodules, J. Algebra Appl., 20192019.
  • TY (2021) A. Tercan and R. Yaşar, Weak FI-extending modules with ACC or DCC on essential submodules, Kyungpook Math. J., 20212021.