ISSN 2148-838Xhttp://dx.doi.org/10.13069/jacodesmath.423751 J. Algebra Comb. Discrete Appl. 5(2) • 85–99 Received: 19 August 2017 Accepted: 26 Ferbruary 2018 Journal of Algebra Combinatorics Discrete Structures and Applications Some results on the comaximal ideal graph of a commutative ring Research Article Subramanian Visweswaran, Jaydeep Parejiya Abstract: The rings considered in this article are commutative with identity which admit at least two maximal ideals. Let R be a ring such that R admits at least two maximal ideals. Recall from Ye and Wu (J. Algebra Appl. 11(6): 1250114, 2012) that the comaximal ideal graph of R, denoted by C (R) is an undirected simple graph whose vertex set is the set of all proper ideals I of R such that I 6⊆ J(R), where J(R) is the Jacobson radical of R and distinct vertices I1, I2 are joined by an edge in C (R) if and only if I1 + I2 = R. In Section 2 of this article, we classify rings R such that C (R) is planar. In Section 3 of this article, we classify rings R such that C (R) is a split graph. In Section 4 of this article, we classify rings R such that C (R) is complemented and moreover, we determine the S-vertices of C (R). 2010 MSC: 13A15, 05C25 Keywords: Comaximal ideal graph, Special principal ideal ring, Planar graph, Split graph, Complement of a vertex in a graph 1. Introduction The rings considered in this article are commutative with identity which admit at least two maximal ideals. Let R be a ring. We denote the set of all maximal ideals of R by Max(R). We denote the Jacobson radical of R by J(R). We denote the cardinality of a set A using the notation |A|. Motivated by the research work done on comaximal graphs of rings in [9, 12, 13, 15, 16] and on the annihilating-ideal graphs of rings in [5, 6], M. Ye and T. Wu in [18] introduced a graph structure on a ring R, whose vertex set is the set of all proper ideals I of R such that I 6⊆ J(R) and distinct vertices I1 and I2 are joined by an edge if and only if I1 + I2 = R. M. Ye and T. Wu called the graph introduced by them in [18] as the comaximal ideal graph of R and denoted it using the notation C (R) and investigated the influence of certain graph parameters of C (R) on the ring structure of R. Let R be a ring such that |Max(R)| ≥ 2. Subramanian Visweswaran (Corresponding Author), Jaydeep Parejiya; Department of Mathematics, Saurashtra Univesity, Rajkot, Gujarat, India 360 005 (email: s_visweswaran2006@yahoo.co.in, parejiyajay@gmail.com). 85 https://orcid.org/0000-0002-4905-809X https://orcid.org/0000-0002-2072-2719 S. Visweswaran, J. Parejiya / J. Algebra Comb. Discrete Appl. 5(2) (2018) 85–99 The aim of this article is to classify rings R such that C (R) is planar; C (R) is a split graph; and C (R) is complemented. It is useful to recall the following definitions from commutative ring theory. A ring R is said to be quasilocal (respectively, semiquasilocal) if R has a unique maximal ideal (respectively, R has only a finite number of maximal ideals). A Noetherian quasilocal (respectively, semiquasilocal) ring is referred to as a local ring (respectively, a semilocal ring). Recall that a principal ideal ring is said to be a special principal ideal ring (SPIR) if R has a unique prime ideal m. It is clear that m is nilpotent. If R is a SPIR with m as its only prime ideal, then we denote it by saying that (R, m) is a SPIR. Suppose that a ring T is quasilocal with m as its unique maximal ideal such that m 6= (0) and nilpotent. Let n ≥ 2 be least with the property that mn = (0). If m is principal, then it follows from (iii) ⇒ (i) of [3, Proposition 8.8] that {mi|i ∈{1, . . . , n− 1}} is the set of all nonzero proper ideals of T . Hence, (T, m) is a SPIR. We next recall the following definitions and results from graph theory. The graphs considered in this article are undirected and simple. Let G = (V, E) be a graph. Recall from [4, Definition 1.2.2] that a clique of G is a complete subgraph of G. The clique number of G, denoted by ω(G) is defined as the largest integer n ≥ 1 such that G contains a clique on n vertices [4, Definition, page 185]. We set ω(G) = ∞ if G contains a clique on n vertices for all n ≥ 1. Let n ∈ N. A complete graph on n vertices is denoted by Kn. A graph G = (V, E) is said to be bipartite if V can be partitioned into two nonempty subsets V1 and V2 such that each edge of G has one end in V1 and the other in V2. A bipartite graph with vertex partition V1 and V2 is said to be complete if each element of V1 is adjacent to every element of V2. Let m, n ∈ N. Let G = (V, E) be a complete bipartite graph with vertex partition V1 and V2. If |V1| = m and |V2| = n, then G is denoted by Km,n [4, Definition 1.1.12]. Let G = (V, E) be a graph. Recall from [4, Definition 8.1.1] that G is said to be planar if G can be drawn in a plane in such a way that no two edges of G intersect in a point other than a vertex of G. Recall that two adjacent edges are said to be in series if their common end vertex is of degree two [7, page 9]. Two graphs are said to be homeomorphic if one graph can be obtained from the other by insertion of vertices of degree two or by the merger of edges in series [7, page 100]. It is useful to note from [7, page 93] that the graph K5 is referred to as Kuratowski’s first graph and the graph K3,3 is referred to as Kuratowski’s second graph. The celebrated theorem of Kuratowski states that a graph G is planar if and only if G does not contain either of Kuratowski’s two graphs or any graph homeomorphic to either of them [7, Theorem 5.9]. Let G = (V, E) be a graph. It is convenient to name the following conditions satisfied by G so that it can be used throughout Section 2 of this article. (i) We say that G satisfies (Ku1) if G does not contain K5 as a subgraph (that is, equivalently, if ω(G) ≤ 4). (ii) We say that G satisfies (Ku∗1) if G satisfies (Ku1) and moreover, G does not contain any subgraph homeomorphic to K5. (iii) We say that G satisfies (Ku2) if G does not contain K3,3 as a subgraph. (iv) We say that G satisfies (Ku∗2) if G satisfies (Ku2) and moreover, G does not contain any subgraph homeomorphic to K3,3. Suppose that a graph G = (V, E) is planar. It follows from Kuratowski’s theorem [7, Theorem 5.9] that G satisfies both (Ku∗1) and (Ku ∗ 2). Hence, G satisfies both (Ku1) and (Ku2). It is interesting to note that a graph G can be nonplanar even if it satisfies both (Ku1) and (Ku2). For an example of this type, refer [7, Figure 5.9(a), page 101] and the graph G given in this example does not satisfy (Ku∗2). It is not hard to construct an example of a graph G such that G satisfies (Ku1) but G does not satisfy (Ku∗1). Let R be a ring such that |Max(R)| ≥ 2. In Section 2 of this article, we try to classify rings R such that C (R) is planar. It is proved in [18, Theorem 3.1] that ω(C (R)) = |Max(R)|. Hence, C (R) satisfies (Ku1) if and only if |Max(R)| ≤ 4. In Section 2 of this article, we first focus on classifying rings R such that C (R) satisfies (Ku2). It is shown in Lemma 2.1 that if C (R) satisfies (Ku2) , then |Max(R)| ≤ 3. 86 S. Visweswaran, J. Parejiya / J. Algebra Comb. Discrete Appl. 5(2) (2018) 85–99 Let R be a ring such that |Max(R)| = 2. It is proved in Proposition 2.7 that C (R) is planar if and only if C (R) satisfies (Ku2) if and only if R ∼= R1 × R2 as rings, where (Ri, mi) is a quasilocal ring for each i ∈ {1, 2} and for at least one i ∈ {1, 2}, (Ri, mi) is a SPIR with m2i = (0). Let R be a ring with |Max(R)| = 3. It is shown in Proposition 2.13 that C (R) satisfies (Ku2) if and only if R ∼= R1×R2×R3 as rings, where (Ri, mi) is a SPIR with m2i = (0) for each i ∈{1, 2, 3}. It is proved in Theorem 2.18 that C (R) is planar if and only if R ∼= R1 × R2 × R3 as rings, where Ri is a field for at least two values of i ∈{1, 2, 3} and if i ∈{1, 2, 3} is such that Ri is not a field, then (Ri, mi) is a SPIR with m2i = (0). Let G = (V, E) be a graph. Recall that G is a split graph if V is the disjoint union of two nonempty subsets K and S such that the subgraph of G induced on K is complete and S is an independent set of G. Let R be a commutative ring with identity. In [16] P. K. Sharma and S.M. Bhatwadekar introduced and investigated a graph associated with R, whose vertex set is the set of all elements of R and distinct vertices x, y are joined by an edge if and only if Rx + Ry = R. The graph studied in [16] is named as the comaximal graph of R in [12]. In [9], M.I. Jinnah and S.C. Mathew classified rings R such that the comaximal graph of R is a split graph. Let R be a ring such that |Max(R)| ≥ 2. In Section 3 of this article, we try to classify rings R such that C (R) is a split graph. It is proved in Lemma 3.2 that if C (R) is a split graph, then |Max(R)| ≤ 3. Let R be a ring such that |Max(R)| = 3. It is shown in Theorem 3.3 that C (R) is a split graph if and only if R ∼= F1 × F2 × F3 as rings, where Fi is a field for each i ∈ {1, 2, 3}. Let R be a ring such that |Max(R)| = 2. It is proved in Theorem 3.5 that C (R) is a split graph if and only if R ∼= F ×S as rings, where F is a field and S is a quasilocal ring. Let G = (V, E) be a graph. Recall from [2, 11] that two distinct vertices u, v of G are said to be orthogonal, written u ⊥ v if u and v are adjacent in G and there is no vertex of G which is adjacent to both u and v in G; that is, the edge u − v is not an edge of any triangle in G. Let u ∈ V . A vertex v of G is said to be a complement of u if u ⊥ v [2]. Moreover, we recall from [2] that G is complemented if each vertex of G admits a complement in G. Furthermore, G is said to be uniquely complemented if G is complemented and whenever the vertices u, v, w of G such that u ⊥ v and u ⊥ w, then a vertex x of G is adjacent to v in G if and only if x is adjacent to w in G. Let R be a ring such that R is not an integral domain. Recall from [1] that the zero-divisor graph of R denoted by Γ(R) is an undirected graph whose vertex set is Z(R)\{0} (here Z(R) denotes the set of all zero-divisors of R) and distinct vertices x, y are joined by an edge if and only if xy = 0. The authors of [2] determined in Section 3 of [2] rings R such that Γ(R) is complemented or uniquely complemented. For a ring R, we denote the set of all units of R by U(R) and we denote the set of all nonunits of R by NU(R). The Krull dimension of a ring R is simply denoted by dimR. In [15, Proposition 3.11] it is proved that the subgraph of the comaximal graph of R induced on NU(R)\J(R) is complemented if and only if dim( R J(R) ) = 0. Section 4 of this article is devoted to find a classification of rings R such that C (R) is complemented. Let R be a ring such that |Max(R)| ≥ 2. It is verified in Remark 4.1(ii) that if C (R) is complemented, then it is uniquely complemented. It is shown in Theorem 4.7 that C (R) is complemented if and only if R is semiquasilocal. Moreover, in Section 4, a discussion on the S-vertices of C (R) is included. Let G = (V, E) be a graph. Recall from [14, Definition 2.9] a vertex a of G is said to be a Smarandache vertex or simply a S-vertex if there exist distinct vertices x, y, and b of G such that a−x, a−b, and b−y are edges of G but there is no edge joining x and y in G. In [14], A.M. Rahimi investigated the S-vertices of the zero-divisor graph of a commutative ring and the zero-divisor graph of a ring with respect to an ideal. For a ring R with |Max(R)| ≥ 2, it is noted in Remark 4.8 that if |MaxR)| = 2, then no vertex of C (R) is a S-vertex. Let R be a ring such that |Max(R)| ≥ 3. It is shown in Proposition 4.9 that a vertex I of C (R) is a S-vertex if and only if I is not contained in at least two distinct maximal ideals of R. Let A, B be sets. If A is a subset of B and A 6= B, then we denote it symbolically using the notation A ⊂ B. Let G be a graph. We denote the vertex set of G by V (G). Let R be a ring. For a proper ideal I of R, as in [15], we denote {m ∈ Max(R)|m ⊇ I} by M(I). 87 S. Visweswaran, J. Parejiya / J. Algebra Comb. Discrete Appl. 5(2) (2018) 85–99 2. Some preliminary results and on the planarity of C (R) As is already mentioned in the introduction, the rings considered in this article are commutative with identity which admit at least two maximal ideals. Lemma 2.1. Let R be a ring. If C (R) satisfies (Ku2), then |Max(R)| ≤ 3. Proof. Suppose that |Max(R)| ≥ 4. Let {mi|i ∈ {1, 2, 3, 4}} ⊆ Max(R). Let V1 = {m1, m2, m1m2} and let V2 = {m3, m4, m3m4}. Observe that V1 ∪ V2 ⊆ V (C (R)), V1 ∩ V2 = ∅, and the subgraph of C (R) induced on V1 ∪ V2 contains K3,3 as a subgraph. Hence, C (R) does not satisfy (Ku2). This is in contradiction to the hypothesis that C (R) satisfies (Ku2). Therefore, |Max(R)| ≤ 3. Lemma 2.2. Let R be a ring such that |Max(R)| ≥ 2. If C (R) satisfies (Ku2), then there exist nonzero rings R1 and R2 such that R ∼= R1 ×R2 as rings. Proof. Assume that C (R) satisfies (Ku2). We assert that R admits a nontrivial idempotent. Suppose that R does not have any nontrivial idempotent. By hypothesis, |Max(R)| ≥ 2. Let m1, m2 be distinct maximal ideals of R. Observe that m1 + m2 = R. Hence, there exist a ∈ m1 and b ∈ m2 such that a + b = 1. Therefore, Ra + Rb = R. It is clear that for all i, j ∈ N, Rai + Rbj = R. Since we are assuming that R has no nontrivial idempotent, we obtain that Rai 6= Raj and Rbi 6= Rbj for all distinct i, j ∈ N. Let V1 = {Ra, Ra2, Ra3} and let V2 = {Rb, Rb2, Rb3}. Note that V1∪V2 ⊆ V (C (R)) and V1∩V2 = ∅. For all i, j ∈ N, Rai + Raj ⊆ m1 and Rbi + Rbj ⊆ m2. Hence, no two members of Vi are adjacent in C (R) for each i ∈{1, 2}. It is clear from the above discussion that the subgraph of C (R) induced on V1∪V2 is K3,3. This is a contradiction. Therefore, R admits at least one nontrivial idempotent. Let e be a nontrivial idempotent of R. Observe that the mapping f : R → Re×R(1−e) defined by f(r) = (re, r(1−e)) is an isomorphism of rings. Let us denote the ring Re by R1 and R(1 − e) by R2. It is clear that R1 and R2 are nonzero rings and R ∼= R1 ×R2 as rings. Remark 2.3. Let R be a ring such that |Max(R)| = 2. If C (R) satisfies (Ku2), then we know from Lemma 2.2 that there exist nonzero rings R1, R2 such that R ∼= R1 × R2 as rings. As |Max(R)| = 2, it follows that Ri is quasilocal for each i ∈ {1, 2}. We assume that R = R1 × R2 where (Ri, mi) is a quasilocal ring for each i ∈{1, 2} and try to classify such rings R in order that C (R) satisfies (Ku2). Lemma 2.4. Let R1, R2 be rings and let R = R1 ×R2. Suppose that Ri has at least two nonzero proper ideals for each i ∈{1, 2}. Then C (R) does not satisfy (Ku2). Proof. We are assuming that Ri has at least two nonzero proper ideals for each i ∈ {1, 2}. Let I1, I2 be distinct nonzero proper ideals of R1 and let J1, J2 be distinct nonzero proper ideals of R2. Let V1 = {I1×R2, I2×R2, (0)×R2} and let V2 = {R1×J1, R1×J2, R1×(0)}. Observe that V1∪V2 ⊆ V (C (R)) and V1 ∩ V2 = ∅. As (Ii × R2) + (R1 × Jk) = R1 × R2 for all i, k ∈ {1, 2, 3} (where we set I3 is the zero ideal of R1 and J3 = zero ideal of R2), it follows that the subgraph of C (R) induced on V1 ∪V2 contains K3,3 as a subgraph. Therefore, C (R) does not satisfy (Ku2). Let I be an ideal of a ring R. Then the annihilator of I in R, denoted by AnnRI is defined as AnnRI = {r ∈ R|Ir = (0)}. Lemma 2.5. Let (R, m) be a local ring which is not a field. The following statements are equivalent: (i) R has only one nonzero proper ideal. (ii) (R, m) is a SPIR with m2 = (0). Proof. (i) ⇒ (ii) We are assuming that R has only one nonzero proper ideal. Hence, m is the only nonzero proper ideal of R. Let x ∈ m, x 6= 0. Then m = Rx. Note that AnnRm is a nonzero proper ideal of R and so, AnnRm = m. Hence, m2 = (0). Therefore, (R, m) is a SPIR with m2 = (0). 88 S. Visweswaran, J. Parejiya / J. Algebra Comb. Discrete Appl. 5(2) (2018) 85–99 (ii) ⇒ (i) As R is not a field, m 6= (0). Thus if (R, m) is a SPIR with m2 = (0), then it is clear that m is the only nonzero proper ideal of R. Lemma 2.6. Let R = R1 × R2, where (Ri, mi) is a quasilocal ring for each i ∈ {1, 2}. The following statements are equivalent: (i) C (R) satisfies (Ku2). (ii) For at least one i ∈{1, 2}, (Ri, mi) is a SPIR with m2i = (0). Proof. (i) ⇒ (ii) Assume that C (R) satisfies (Ku2). We know from Lemma 2.4 that for at least one i ∈ {1, 2}, Ri has at most one nonzero proper ideal. Hence, for that i, either mi = (0) or in the case mi 6= (0), we obtain from Lemma 2.5 that (Ri, mi) is a SPIR with m2i = (0). Therefore, there exists at least one i ∈{1, 2} such that (Ri, mi) is a SPIR with m2i = (0). (ii) ⇒ (i) Without loss of generality, we can assume that (R1, m1) is a SPIR with m21 = (0). Observe that {M1 = m1×R2, M2 = R1×m2} is the set of all maximal ideals of R. Since |Max(R)| = 2, it follows from (3) ⇒ (1) of [18, Theorem 4.5] that C (R) is a complete bipartite graph with vertex partition V1 and V2, where Vi is the set of all proper ideals A of R such that M(A) = {Mi} for each i ∈{1, 2}. Note that if A ∈ V1, then A = I × R2 for some ideal I of R1 such that I ⊆ m1. Since there are at most two proper ideals of R1, we obtain that |V1| ≤ 2. It is now clear that C (R) satisfies (Ku2). Proposition 2.7. Let R be a ring such that |Max(R)| = 2. The following statements are equivalent: (i) C (R) is planar. (ii) C (R) satisfies both (Ku∗1) and (Ku ∗ 2). (iii) C (R) satisfies (Ku2). (iv) R ∼= R1 ×R2 as rings, where (Ri, mi) is a quasilocal ring for each i ∈{1, 2} and for at least one i ∈{1, 2}, (Ri, mi) is a SPIR with m2i = (0) Proof. (i) ⇒ (ii) This follows from Kuratowski’s theorem [7, Theorem 5.9]. (ii) ⇒ (iii) This is clear. (iii) ⇒ (iv) This follows from Remark 2.3 and Lemma 2.6. (iv) ⇒ (i) Let us denote the ring R1 ×R2 by T . Without loss of generality, we can assume that (R1, m1) is a SPIR with m21 = (0). Let V1, V2 be as in the proof of (ii) ⇒ (i) of Lemma 2.6 and it is already noted there that |V1| ≤ 2 and C (T ) is a complete bipartite graph with vertex partition V1 and V2. It is now clear that C (T ) is planar. Since R ∼= T as rings, we get that C (R) is planar. Let R be a ring such that |Max(R)| = 3. We next try to classify such rings R in order that C (R) satisfies (Ku2). Lemma 2.8. Let R1, R2 be rings and let R = R1 ×R2. If R1 admits at least two maximal ideals and if C (R1) does not satisfy (Ku2), then C (R) does not satisfy (Ku2). Proof. We are assuming that C (R1) does not satisfy (Ku2). Then there exist subsets A = {I1, I2, I3} and B = {J1, J2, J3} of V (C (R1)) such that A ∩ B = ∅ and Ii + Jk = R1 for all i, k ∈ {1, 2, 3}. Let V1 = {I1×R2, I2×R2, I3×R2} and let V2 = {J1×R2, J2×R2, J3×R2}. Observe that V1∪V2 ⊆ V (C (R)), V1 ∩ V2 = ∅, and as (Ii × R2) + (Jk × R2) = R1 × R2 = R for all i, k ∈ {1, 2, 3}, it follows that the subgraph of C (R) induced on V1 ∪ V2 contains K3,3 as a subgraph. This proves that C (R) does not satisfy (Ku2). Lemma 2.9. Let R be a ring such that |Max(R)| = 3. If C (R) satisfies (Ku2), then R ∼= R1 ×R2 ×R3 as rings, where Ri is a quasilocal ring for each i ∈{1, 2, 3}. 89 S. Visweswaran, J. Parejiya / J. Algebra Comb. Discrete Appl. 5(2) (2018) 85–99 Proof. Assume that C (R) satisfies (Ku2). As |Max(R)| = 3, it follows from Lemma 2.2 that there exist nonzero rings T1 and T2 such that R ∼= T1 ×T2 as rings. Since R has exactly three maximal ideals, it follows that either T1 or T2 is not quasilocal. Without loss of generality, we can assume that T1 is not quasilocal. Hence, the number of maximal ideals of T1 is exactly two. Let us denote the ring T1 × T2 by T. Since R ∼= T as rings, we obtain that C (T ) satisfies (Ku2). Now, it follows from Lemma 2.8 that C (T1) satisfies (Ku2). Hence, we obtain from Lemma 2.2 that there exist nonzero rings T11 and T12 such that T1 ∼= T11 × T12 as rings. Therefore, R ∼= T11 × T12 × T2 as rings. Hence, on renaming the rings T11, T12, and T2, we obtain that there exist rings R1, R2, and R3 such that R ∼= R1 × R2 × R3 as rings. Since |Max(R)| = 3, it is clear that Ri is quasilocal for each i ∈{1, 2, 3}. Lemma 2.10. Let R be a ring such that |Max(R)| = 3. If C (R) satisfies (Ku2), then R ∼= R1×R2×R3 as rings, where (Ri, mi) is a SPIR with m2i = (0) for each i ∈{1, 2, 3}. Proof. Assume that C (R) satisfies (Ku2). We know from Lemma 2.9 that R ∼= R1 ×R2 ×R3 as rings, where Ri is a quasilocal ring for each i ∈{1, 2, 3}. Let mi denote the unique maximal ideal of Ri for each i ∈ {1, 2, 3}. Let us denote the ring R1 × R2 × R3 by T. Since R ∼= T as rings, we obtain that C (T ) satisfies (Ku2). Let i ∈ {1, 2, 3}. It follows from Lemma 2.4 that Ri has at most one nonzero proper ideal. Hence, either mi = (0) in which case, Ri is a field or mi 6= (0) is the only nonzero proper ideal of Ri in which case, we obtain from Lemma 2.5 that (Ri, mi) is a SPIR with m2i = (0). This proves that R ∼= R1 ×R2 ×R3 as rings, where (Ri, mi) is a SPIR with m2i = (0) for each i ∈{1, 2, 3}. Lemma 2.11. Let R be a ring such that |Max(R)| = 3. Let {m1, m2, m3} denote the set of all maximal ideals of R. Let i ∈{1, 2, 3}. Let us denote the set of all proper ideals I of R such that M(I) = {mi} by Wi. If |Wi| ≤ 2 for each i ∈{1, 2, 3}, then C (R) satisfies (Ku2). Proof. Suppose that C (R) does not satisfy (Ku2). Then there exist subsets V1 = {I1, I2, I3} and V2 = {J1, J2, J3} of V (C (R)) such that V1∩V2 = ∅ and Ii + Jk = R for all i, k ∈{1, 2, 3}. After renaming the maximal ideals of R (if necessary), we can assume without loss of generality that I1 ⊆ m1. Since I1 + Jk = R for each k ∈{1, 2, 3}, it follows that Jk 6⊆ m1 for each k ∈{1, 2, 3}. By hypothesis, |W2| ≤ 2 and |W3| ≤ 2. Therefore, we obtain that W2∩V2 6= ∅ and W3∩V2 6= ∅. This implies that J1J2J3 ⊆ m2m3. It follows from Ii + Jk = R for all i, k ∈ {1, 2, 3} that Ii + J1J2J3 = R for each i ∈ {1, 2, 3} and so, Ii + m2m3 = R. Hence, we get that Ii ∈ W1 for each i ∈{1, 2, 3}. This implies that |W1| ≥ 3. This is in contradiction to the assumption that |W1| ≤ 2. Therefore, C (R) satisfies (Ku2). Let R be a ring such that |Max(R)| = 3. Let i ∈ {1, 2, 3} and let mi, Wi be as in the statement of Lemma 2.11. In Proposition 2.12, we classify such rings R in order that |Wi| ≤ 2 for each i ∈{1, 2, 3}. Proposition 2.12. Let R be a ring such that |Max(R)| = 3. Let {mi|i ∈ {1, 2, 3}} denote the set of all maximal ideals of R. Let i ∈ {1, 2, 3}. Let us denote the set of all proper ideals I of R such that M(I) = {mi} by Wi. Then the following statements are equivalent: (i) |Wi| ≤ 2 for each i ∈{1, 2, 3}. (ii) R ∼= R1 ×R2 ×R3 as rings, where (Ri, ni) is a SPIR with n2i = (0) for each i ∈{1, 2, 3}. Proof. (i) ⇒ (ii) Let i ∈{1, 2, 3}. We claim that mi is principal. First, we verify that m1 is principal. Suppose that m1 is not principal. Observe that m1 6⊆ m2 ∪ m3. Let a1 ∈ m1\(m2 ∪ m3). As m1 6= Ra1 by assumption, it follows from [10, Theorem 81] that there exists a2 ∈ m1\(Ra1 ∪ m2 ∪ m3). Note that m1 6= Ra2 and it is clear from the choice of the elements a1, a2 that Ra1 6= Ra2 and {Ra1, Ra2, m1}⊆ W1. This is in contradiction to the assumption that |W1| ≤ 2. Therefore, m1 is principal. Similarly, it can be shown that m2 and m3 are principal. Let i ∈ {1, 2, 3}. Observe that m2i = m 3 i . Suppose that m 2 i 6= m 3 i . Then {mi, m2i , m 3 i} ⊆ Wi. This is impossible, since |Wi| ≤ 2. Therefore, m 2 i = m 3 i . Since mi + mj = R for all distinct i, j ∈ {1, 2, 3}, it follows from [3, Proposition 1.10(i)] that J(R) = ∩3i=1mi = ∏3 i=1 mi. Hence, (J(R))2 = ∏3 i=1 m 2 i = ∏3 i=1 m 3 i = (J(R)) 3. Now, as J(R) = ∏3 i=1 mi is principal, there exists a ∈ J(R) such that J(R) = Ra. From (J(R))2 = (J(R))3, we obtain that Ra2 = Ra3. Hence, a2 = ra3 for some r ∈ R. Since 1 − ra is a unit in R, we obtain that a2 = 0 and so, (J(R))2 = (0). Since 90 S. Visweswaran, J. Parejiya / J. Algebra Comb. Discrete Appl. 5(2) (2018) 85–99 m2i + m 2 j = R for all distinct i, j ∈ {1, 2, 3} and ∩ 3 i=1m 2 i = ∏3 i=1 m 2 i = (0), we obtain from the Chinese remainder theorem [3, Proposition 1.10(ii) and (iii)] that the mapping f : R → R m21 × R m22 × R m23 given by f(r) = (r + m21, r + m 2 2, r + m 2 3) is an isomorphism of rings. Let i ∈ {1, 2, 3}. Let us denote the ring R m2 i by Ri. Let us denote mim2 i by ni. Since mi is a principal ideal of R, we obtain that ni is a principal ideal of Ri and it is clear that n2i = (0 + m 2 i ). This shows that (Ri, ni) is a SPIR with n 2 i is the zero ideal of Ri for each i ∈{1, 2, 3} and R ∼= R1 ×R2 ×R3 as rings. (ii) ⇒ (i) Assume that R ∼= R1 × R2 × R3 as rings, where (Ri, ni) is a SPIR with n2i = (0) for each i ∈ {1, 2, 3}. Let us denote the ring R1 × R2 × R3 by T . Observe that T is semilocal with {N1 = n1 ×R2 ×R3, N2 = R1 ×n2 ×R3, N3 = R1 ×R2 ×n3} as its set of all maximal ideals. Let us denote the set of all proper ideals A of T such that M(A) = {Ni} by Ui for each i ∈{1, 2, 3}. Since Ri has at most one nonzero proper ideal for each i ∈ {1, 2, 3}, it follows that |Ui| ≤ 2. From R ∼= T as rings, we obtain that |Wi| ≤ 2 for each i ∈{1, 2, 3}. Proposition 2.13. Let R be a ring such that |Max(R)| = 3. The following statements are equivalent: (i) C (R) satisfies (Ku2). (ii) R ∼= R1 ×R2 ×R3 as rings, where (Ri, mi) is a SPIR with m2i = (0) for each i ∈{1, 2, 3}. Proof. (i) ⇒ (ii) Assume that C (R) satisfies (Ku2). We know from Lemma 2.10 that R ∼= R1×R2×R3 as rings, where (Ri, mi) is a SPIR with m2i = (0) for each i ∈{1, 2, 3}. (ii) ⇒ (i) Assume that R ∼= R1 × R2 × R3 as rings, where (Ri, mi) is a SPIR with m2i = (0) for each i ∈{1, 2, 3}. Let {Mi|i ∈{1, 2, 3}} denote the set of all maximal ideals of R. Let i ∈{1, 2, 3} and let us denote the set of all proper ideals I of R such that M(I) = {Mi} by Wi. We know from (ii) ⇒ (i) of Proposition 2.12 that |Wi| ≤ 2. Hence, we obtain from Lemma 2.11 that C (R) satisfies (Ku2). Let R be a ring such that |Max(R)| = 3. We try to classify such rings R in order that C (R) is planar. If C (R) is planar, then we know from Kuratowski’s theorem [7, Theorem 5.9] that C (R) satisfies (Ku2). Hence, we obtain from (i) ⇒ (ii) of Proposition 2.13 that R ∼= R1 × R2 × R3 as rings, where (Ri, mi) is a SPIR with m2i = (0) for each i ∈{1, 2, 3}. Lemma 2.14. Let R = R1 × R2 × R3, where (Ri, mi) is a SPIR with mi 6= (0) but m2i = (0) for each i ∈{1, 2, 3}. Then C (R) does not satisfy (Ku∗2). Proof. The proof of this lemma closely follows the proof given in [17, Lemma 3.13]. Note that |Max(R)| = 3 and {M1 = m1 × R2 × R3, M2 = R1 × m2 × R3, M3 = R1 × R2 × m3} is the set of all maximal ideals of R. Let us denote the subgraph of C (R) induced on W = {v1 = M1, v2 = M21, v3 = M3, v4 = M2, v5 = M 2 2, v6 = M 2 3, v7 = M1∩M2} by H. Observe that in H, the edges M3−M1∩M2 and M1∩M2−M23 are in series and moreover, in H, vi is adjacent to v4 and v5 for each i ∈{1, 2, 3}. Further- more in H, v1 and v2 are adjacent to v6. Therefore, on merging the edges v3 −v7 and v7 −v6, we obtain a graph H1 which contains K3,3 as a subgraph. Hence, H contains a subgraph which is homeomorphic to K3,3. This shows that C (R) does not satisfy (Ku∗2). Lemma 2.15. Let R = F × R2 × R3, where F is a field and (Ri, mi) is a SPIR with mi 6= (0) but m2i = (0) for each i ∈{2, 3}. Then C (R) does not satisfy (Ku ∗ 1). Proof. The proof of this lemma closely follows the proof given in [17, Lemma 3.14]. Observe that |Max(R)| = 3 and {M1 = (0)×R2×R3, M2 = F ×m2×R3, M3 = F ×R2×m3} is the set of all maximal ideals of R. Let us denote the subgraph of C (R) induced on W = {v1 = M2, v2 = M1M3, v3 = M22, v4 = M3, v5 = M1M2, v6 = M 2 3, v7 = M1} by H. Note that in H, the edges e1 : v1 −v2, e2 : v2 −v3 are edges in series and the edges e3 : v4 − v5, e4 : v5 − v6 are edges in series. Observe that in H, v1 is adjacent to all the elements of W except v3 and v5; v3 is adjacent to all the elements of W except v1 and v5; v4 is adjacent to all the elements of W except v2 and v6; v6 is adjacent to all the elements of W except v2 91 S. Visweswaran, J. Parejiya / J. Algebra Comb. Discrete Appl. 5(2) (2018) 85–99 and v4; v7 is adjacent to all the elements of W except v2 and v5. Let H1 be the graph obtained from H on merging the edges e1 and e2 and on merging the edges e3 and e4. It is clear that H1 is a complete graph on five vertices. This proves that C (R) contains a subgraph H such that H is homeomorphic to K5. Therefore, C (R) does not satisfy (Ku∗1). Lemma 2.16. Let R = F1 ×F2 ×R3, where F1 and F2 are fields and (R3, m3) is a SPIR with m3 6= (0) but m23 = (0). Then C (R) is planar. Proof. Observe that |Max(R)| = 3 and {M1 = (0)×R2×R3, M2 = F1×(0)×R3, M3 = F1×F2×m3} is the set of all maximal ideals of R. Observe that V (C (R)) equals {v1 = M1, v2 = M2, v3 = M3, v4 = M1M2, v5 = M2M3, v6 = M1M3, v7 = M 2 3, v8 = M1M 2 3, v9 = M2M 2 3}. It is not hard to verify that C (R) is the union of the cycle Γ : v1 −v3 −v2 −v7 −v1, the edges e1 : v3 −v4, e2 : v4 −v7, e3 : v1 −v2, and the pendant edges e4 : v1 − v5, e5 : v1 − v9, e6 : v2 − v6, and e7 : v2 − v8. Note that Γ can be represented by means of a rectangle. The edges e1, e2 are edges in series and their common end vertex v4 can be plotted inside the rectangle representing Γ and the edges e1, e2 can be drawn inside this rectangle. The edges ei(i ∈{3, 4, 5, 6, 7}) are such that one of their end vertices ∈{v1, v2} and they can be drawn outside the rectangle representing Γ in such a way that there are no crossing over of the edges. This proves that C (R) is planar. Lemma 2.17. Let R = F1 ×F2 ×F3, where Fi is a field for each i ∈{1, 2, 3}. Then C (R) is planar. Proof. Note that |Max(R)| = 3 and {M1 = (0) × F2 × F3, M2 = F1 × (0) × F3, M3 = F1 × F2 × (0)} is the set of all maximal ideals of R. Observe that V (C (R)) equals {v1 = M1, v2 = M2, v3 = M3, v4 = M1M2, v5 = M2M3, v6 = M1M3}. It is clear that C (R) is the union of the cycle Γ : v1 − v2 − v3 − v1 and the pendant edges e1 : v1−v5, e2 : v2−v6, and e3 : v3−v4. The cycle Γ can be represented by means of a triangle and the one of the end vertex of ei is vi for each i ∈ {1, 2, 3} and the edges e1, e2, e3 can be drawn outside the triangle representing Γ in such a way that there are no crossing over of the edges. This proves that C (R) is planar. Theorem 2.18. Let R be a ring such that |Max(R)| = 3. The following statements are equivalent: (i) C (R) is planar. (ii) C (R) satisfies both (Ku∗1) and (Ku ∗ 2). (iii) R ∼= R1 × R2 × R3 as rings, where Ri is a field for at least two values of i ∈ {1, 2, 3} and if i ∈{1, 2, 3} is such that Ri is not a field, then (Ri, mi) is a SPIR with m2i = (0). Proof. (i) ⇒ (ii) This follows from Kuratowski’s theorem [7, Theorem 5.9]. (ii) ⇒ (iii) Since C (R) satisfies (Ku∗2), we get that C (R) satisfies (Ku2). Therefore, we obtain from Proposition 2.13 that R ∼= R1 × R2 × R3 as rings, where (Ri, mi) is a SPIR with m2i = (0) for each i ∈ {1, 2, 3}. Let us denote the ring R1 × R2 × R3 by T. Since R ∼= T as rings, we obtain that C (T ) satisfies (Ku∗1) and (Ku ∗ 2). As C (T ) satisfies (Ku ∗ 2), it follows from Lemma 2.14 that Ri is a field for at least one value of i ∈ {1, 2, 3}. Suppose that Ri is a field for exactly one value of i ∈ {1, 2, 3}. Without loss of generality, we can assume that R1 is a field and R2, R3 are not fields. In such a case, we obtain from Lemma 2.15 that C (T ) does not satisfy (Ku∗1). This is a contradiction. Therefore, Ri is a field for at least two values of i ∈{1, 2, 3}. This proves that R ∼= R1 ×R2 ×R3 as rings, where Ri is a field for at least two values of i ∈ {1, 2, 3} and if i ∈ {1, 2, 3} is such that Ri is not a field, then (Ri, mi) is a SPIR with m2i = (0). (iii) ⇒ (i) Suppose that R ∼= R1 × R2 × R3 as rings, where Ri is a field for at least two values of i ∈ {1, 2, 3} and if i ∈ {1, 2, 3} is such that Ri is not a field, then (Ri, mi) is a SPIR with m2i = (0). Let us denote the ring R1 ×R2 ×R3 by T. Note that either Ri is a field for each i ∈{1, 2, 3}, in which case, we obtain from Lemma 2.17 that C (T ) is planar or there are exactly two values of i ∈{1, 2, 3} such that Ri is a field and in such a case, we obtain from Lemma 2.16 that C (T ) is planar. Since R ∼= T as rings, we get that C (R) is planar. 92 S. Visweswaran, J. Parejiya / J. Algebra Comb. Discrete Appl. 5(2) (2018) 85–99 3. When is C (R) a split graph? Let R be a ring such that |Max(R)| ≥ 2. The aim of this section is to classify rings R such that C (R) is a split graph. Throughout this section, we assume that K is a nonempty subset of V (C (R)) such that the subgraph of C (R) induced on K is complete and S is a nonempty subset of V (C (R)) such that S is an independent set of C (R). Lemma 3.1. Let R be a ring such that C (R) is a split graph with V (C (R)) = K ∪S and K ∩S = ∅. If |Max(R)| ≥ 3, then Max(R) = K. Proof. First, we claim that Max(R)) ⊆ K. Suppose that Max(R) 6⊆ K. Then there exists m ∈ Max(R) such that m /∈ K. Hence, m ∈ S. We are assuming that |Max(R)| ≥ 3. Therefore, there exist distinct maximal ideals m′, m′′ of R such that m′ 6= m and m′′ 6= m. Since m+m′ = m+m′′ = m+m′m′′ = R, we get that m′, m′′, and m′m′′ are adjacent to m in C (R). As m ∈ S, we obtain that m′, m′′, m′m′′ ∈ K. Hence, m′ and m′m′′ must be adjacent in C (R). This is impossible, since m′+m′m′′ = m′ 6= R. Therefore, Max(R) ⊆ K. We next verify that K ⊆ Max(R). Let I ∈ K. Then I is a proper ideal of R and so, there exists a maximal ideal m of R such that I ⊆ m. We assert that I = m. Suppose that I 6= m. Then I and m are adjacent in C (R). This is impossible, since I + m = m 6= R. Hence, I = m and this proves that K ⊆ Max(R) and so, K = Max(R). Lemma 3.2. Let R be a ring. If C (R) is a split graph, then |Max(R)| ≤ 3. Proof. Suppose that |Max(R)| ≥ 4. Now, V (C (R)) = K ∪ S with K ∩ S = ∅. Since C (R) is a split graph by assumption, we obtain from Lemma 3.1 that Max(R) = K. Let {mi|i ∈{1, 2, 3, 4}}⊆ Max(R). Note that for all distinct i, j ∈ {1, 2, 3, 4}, mimj /∈ Max(R) = K and so, mimj ∈ S. Hence, both m1m2 and m3m4 must be in S. Therefore, m1m2 cannot be adjacent to m3m4 in C (R). This is impossible, since m1m2 + m3m4 = R. Therefore, |Max(R)| ≤ 3. Let R be a ring such that |Max(R)| = 3. In Theorem 3.3, we classify such rings R in order that C (R) is a split graph. Theorem 3.3. Let R be a ring such that |Max(R)| = 3. The following statements are equivalent: (i) C (R) is a split graph. (ii) R ∼= F1 ×F2 ×F3 as rings, where Fi is a field for each i ∈{1, 2, 3}. Proof. (i) ⇒ (ii) We are assuming that C (R) is a split graph. Then V (C (R)) = K∪S with K∩S = ∅. As |Max(R)| = 3, we obtain from Lemma 3.1 that Max(R) = K. Let {m1, m2, m3} denote the set of all maximal ideals of R. Let i ∈ {1, 2, 3} and let us denote the set of all proper ideals I of R such that M(I) = {mi} by Wi. We assert that Wi = {mi} for each i ∈ {1, 2, 3}. First, we show that W1 = {m1}. It is clear that m1 ∈ W1. Let I ∈ W1 be such that I 6= m1. As K = Max(R), it follows that I must be in S. Note that m2m3 ∈ S. It is clear that I 6= m2m3. Now, I, m2m3 ∈ S and S is an independent set of C (R), we get that I and m2m3 cannot be adjacent in C (R). However, I + m2m3 = R. This is a contradiction and so, I = m1. This shows that W1 = {m1}. Similarly, it can be shown that W2 = {m2} and W3 = {m3}. We next show that mi is principal for each i ∈{1, 2, 3}. Note that m1 6⊆ m2∪m3. Hence, there exists x1 ∈ m1\(m2 ∪ m3). Observe that Rx1 ∈ W1 = {m1} and so, m1 = Rx1. Similarly, using the facts that W2 = {m2} and W3 = {m3}, it can be proved that m2 = Rx2 for any x2 ∈ m2\(m1 ∪ m3) and m3 = Rx3 for any x3 ∈ m3\(m1 ∪ m2). It is clear that J(R) = ∩3i=1mi = ∏3 i=1 mi = Rx1x2x3. Let i ∈{1, 2, 3}. Note that m2i ∈ Wi = {mi} and so, mi = m 2 i . This implies that ∏3 i=1 mi = ∏3 i=1 m 2 i . Hence, Rx1x2x3 = Rx 2 1x 2 2x 2 3. Therefore, x1x2x3 = rx 2 1x 2 2x 2 3 for some r ∈ R. As x1x2x3 ∈ J(R), we obtain that 1 − rx1x2x3 is a unit in R and so, x1x2x3 = 0. Since mi + mj = R for all distinct i, j ∈ {1, 2, 3} and J(R) = (0), we obtain from the Chinese remainder theorem [3, Proposition 1.10(ii) and (iii)] that the mapping f : R → R m1 × R m2 × R m3 given by f(r) = (r + m1, r + m2, r + m3) is an isomorphism of rings. Let 93 S. Visweswaran, J. Parejiya / J. Algebra Comb. Discrete Appl. 5(2) (2018) 85–99 us denote the field R mi by Fi for each i ∈ {1, 2, 3}. Therefore, R ∼= F1 × F2 × F3 as rings, where Fi is a field for each i ∈{1, 2, 3}. (ii) ⇒ (i) We are assuming that R ∼= F1×F2×F3 as rings, where Fi is a field for each i ∈{1, 2, 3}. Let us denote the ring F1 ×F2 ×F3 by T. Note that V (C (T )) = {m1 = (0)×F2 ×F3, m2 = F1 ×(0)×F3, m3 = F1×F2×(0), m1m2 = (0)×(0)×F3, m2m3 = F1×(0)×(0), m1m3 = (0)×F2×(0)}. Let K = {m1, m2, m3} and let S = {m1m2, m2m3, m1m3}. It is clear that V (C (T )) = K ∪ S, K ∩ S = ∅, the subgraph of C (T ) induced on K is complete and S is an independent set of C (T ). Therefore, C (T ) is a split graph. As R ∼= T as rings, we obtain that C (R) is a split graph. Let R be a ring such that |Max(R)| = 2. We next try to classify such rings in order that C (R) is a split graph. Lemma 3.4 is well-known. We include a proof of Lemma 3.4 for the sake of completeness. Lemma 3.4. Let G = (V, E) be a complete bipartite graph. The following statements are equivalent: (i) G is a split graph. (ii) G is a star graph. Proof. (i) ⇒ (ii) Assume that G is a split graph. Hence, there exist nonempty subsets K, S of V such that V = K ∪ S, K ∩ S = ∅, the subgraph of G induced on K is complete, and S is an independent set of G. By hypothesis, G is a complete bipartite graph. Let G be complete bipartite with vertex partition V1 and V2. We claim that S ∩ Vi = ∅ for some i ∈ {1, 2}. Suppose that S ∩ Vi 6= ∅ for each i ∈ {1, 2}. Let si ∈ S ∩ Vi for each i ∈ {1, 2}. Then s1 and s2 are adjacent in G. This is impossible, since S is an independent set of G. Therefore, either S ∩ V1 = ∅ or S ∩ V2 = ∅. Without loss of generality, we can assume that S∩V2 = ∅. Hence, S = S∩V = S∩(V1 ∪V2) = (S∩V1)∪(S∩V2) = S∩V1 and so, S ⊆ V1. It follows from V = V1 ∪ V2 = S ∪ K and S ⊆ V1 that V2 ⊆ K. Since no two distinct elements of V2 are adjacent in G, whereas any two distinct vertices of K are adjacent in G, it follows that |V2| = 1. This shows that G is a star graph. (ii) ⇒ (i) Suppose that G is a star graph. Hence, G is a complete bipartite graph with vertex partition V1 and V2 such that |Vi| = 1 for at least one i ∈ {1, 2}. Without loss of generality, we can assume that |V1| = 1. With K = V1 and S = V2, it is clear that V = K ∪ S, K ∩ S = ∅, the subgraph of G induced on K is complete, and S is an independent of G. Therefore, G is a split graph. Theorem 3.5. Let R be a ring such that |Max(R)| = 2. The following statements are equivalent: (i) C (R) is a split graph. (ii) R ∼= F ×S as rings, where F is a field and S is a quasilocal ring. Proof. Let {m1, m2} denote the set of all maximal ideals of R. (i) ⇒ (ii) Assume that C (R) is a split graph. As |Max(R)| = 2, we know from (3) ⇒ (1) of [18, Theorem 4.5] that C (R) is a complete bipartite graph with vertex partition V1 and V2, where for each i ∈ {1, 2}, Vi is the set of all proper ideals I of R such that M(I) = {mi}. As we are assuming that C (R) is a split graph, we obtain from Lemma 3.4 that C (R) is a star graph. Hence, there exists a vertex I of C (R) such that I is adjacent to each vertex J of C (R) with J 6= I. We can assume without loss of generality that I ∈ V1. In such a case, we obtain that V1 = {I}. It is clear that m1 ∈ V1 and so, I = m1. Let a ∈ m1\m2. As Ra ∈ V1, we get that m1 = Ra. Observe that a2 ∈ m1\m2. Hence, Ra2 ∈ V1 and so, Ra = Ra2 = m1. Now, there exists r ∈ R such that a = ra2. This implies that e = ra is a nontrivial idempotent element of R and moreover, m1 = Ra = Re. Note that the mapping f : R → R(1 − e) × Re defined by f(x) = (x(1 −e), xe) is an isomorphism of rings. Hence, f(m1) = (0) ×Re is a maximal ideal of R(1 − e) × Re. Therefore, R(1 − e) is a field. Since |Max(R)| = 2, it follows that the ring Re is quasilocal. Thus with F = R(1 − e) and S = Re, we obtain that F is a field and S is a quasilocal ring and R ∼= F ×S as rings. (ii) ⇒ (i) Assume that R ∼= F × S as rings, where F is a field and S is a quasilocal ring. Let us denote the ring F ×S by T. Let V1 = {(0)×S} and V2 = {F ×I|I is a proper ideal of S}. Note that C (T ) is a 94 S. Visweswaran, J. Parejiya / J. Algebra Comb. Discrete Appl. 5(2) (2018) 85–99 complete bipartite graph with vertex partition V1 and V2 and as |V1| = 1, it follows that C (T ) is a star graph. Hence, we obtain from (ii) ⇒ (i) of Lemma 3.4 that C (T ) is a split graph. Since R ∼= T as rings, we get that C (R) is a split graph. 4. Some more results on C (R) Let R be a ring such that |Max(R)| ≥ 2. The aim of this section is to classify rings R such that C (R) is complemented and to determine the S-vertices of C (R). Let R be a ring. Let X be the set of all prime ideals of R. Recall from [3, Exercise 15, page 12] that for a subset E of R, the set of all prime ideals p of R such that p ⊇ E is denoted by V (E). We know from [3, Exercise 15, page 12] that the collection {V (E)|E ⊆ R} satisfies the axioms for closed sets in a topological space. The resulting topology is called the Zariski topology. The topological space X is called the prime spectrum of R and is denoted by Spec(R). Let E ⊆ R. We know from [3, Exercise 15(i), page 12] that V (E) = V (I), where I is the ideal of R generated by E. The subspace of Spec(R) consisting of all the maximal ideals of R with the induced topology is called the maximal spectrum of R and is denoted by Max(R). The collection {V (I)∩Max(R)|I varies over all ideals of R} is the collection of all closed sets of Max(R). As in [15], we denote V (I)∩Max(R) by M(I). Thus M(I), as mentioned in the introduction, is the set of all maximal ideals m of R such that m ⊇ I. As in [15], for an element a ∈ R, we denote M(Ra) simply by M(a) and Max(R)\M(a) by D(a). Let G = (V, E) be a simple graph. Let v ∈ V . Then the set of all u ∈ V such that v is adjacent to u in G is called the set of neighbours of v in G and we use the notation NG(v) to denote the set of all neighbours of v in G. Remark 4.1. Let R be a ring such that |Max(R)| ≥ 2. The following statements hold. (i) Let I, J ∈ V (C (R)) be such that I ⊥ J in C (R). Then IJ ⊆ J(R). (ii) If C (R) is complemented, then it is uniquely complemented. Proof. (i) Suppose that IJ 6⊆ J(R). Then there exists a maximal ideal m of R such that IJ 6⊆ m. Hence, I 6⊆ m and J 6⊆ m and so, I + m = J + m = R. Now, m ∈ V (C (R)) is such that m is adjacent to both I and J in C (R). This is impossible, since I ⊥ J in C (R). Therefore, IJ ⊆ J(R). (ii) Let I ∈ V (C (R)). We are assuming that C (R) is complemented. Hence, there exists at least one J ∈ V (C (R)) such that I ⊥ J in C (R). Let J1, J2 ∈ V (C (R)) be such that I ⊥ J1 and I ⊥ J2 in C (R). We know from (i) that IJi ⊆ J(R) for each i ∈ {1, 2}. Let A ∈ V (C (R)) be such that J1 is adjacent to A in C (R). Hence, J1 + A = R. We claim that J2 + A = R. Suppose that J2 + A 6= R. Then there exists m ∈ Max(R) such that J2 + A ⊆ m. It follows from I + J2 = R that I 6⊆ m. As IJ1 ⊆ J(R) ⊆ m, we get that J1 ⊆ m. Therefore, J1 + A ⊆ m. This is impossible, since J1 + A = R. Hence, J2 + A = R. This shows that NC (R)(J1) ⊆ NC (R)(J2). Similarly, it can be shown that NC (R)(J2) ⊆ NC (R)(J1). Therefore, NC (R)(J1) = NC (R)(J2). This proves that C (R) is uniquely complemented. Lemma 4.2. Let R be a ring such that |Max(R)| ≥ 2. The following statements are equivalent: (i) C (R) is complemented. (ii) M(I) is a closed and open subset of Max(R) for each I ∈ V (C (R)). Proof. We adapt an argument found in the proof of [15, Proposition 3.10]. (i) ⇒ (ii) Assume that C (R) is complemented. Let I ∈ V (C (R)). It is clear that M(I) is a closed subset of Max(R). Since C (R) is complemented, there exists J ∈ V (C (R)) such that I ⊥ J in C (R). Hence, I and J are adjacent in C (R) and there is no A ∈ V (C (R)) such that A is adjacent to both I and J in C (R). That is, I + J = R and there is no proper ideal A of R with A + I = A + J = R. Note that there exist a ∈ I and b ∈ J such that a + b = 1. We claim that M(I) = D(b). Let m ∈ M(I). As m ⊇ I, a ∈ I , and a + b = 1, it follows that b /∈ m. Hence, m ∈ D(b). This shows that M(I) ⊆ D(b). Let m ∈ D(b). 95 S. Visweswaran, J. Parejiya / J. Algebra Comb. Discrete Appl. 5(2) (2018) 85–99 Hence, m + J = R. Since, I ⊥ J in C (R), it follows that I + m 6= R and so, I ⊆ m. That is, m ∈ M(I). This proves that D(b) ⊆ M(I) and so, M(I) = D(b) is a closed and open subset of Max(R). (ii) ⇒ (i) Let I ∈ V (C (R)). By assumption, M(I) is a closed and open subset of Max(R). Hence, there exists an ideal J of R such that M(I) = Max(R)\M(J). This implies that I + J = R and IJ ⊆ J(R). If A ∈ V (C (R)) is such that A + I = A + J = R, then A + IJ = R. This is impossible, since IJ ⊆ J(R). Therefore, there is no A ∈ V (C (R)) such that A is adjacent to both I and J in C (R). This proves that I ⊥ J in C (R). Therefore, C (R) is complemented. Proposition 4.3. Let R be a ring such that |Max(R)| ≥ 2. If R is semiquasilocal, then C (R) is complemented. Proof. Let {mi|i ∈ {1, 2, . . . , n}} denote the set of all maximal ideals of R. Note that V (C (R)) = {I|I is a proper ideal of R with I 6⊆ J(R)}. Let I ∈ V (C (R)). Let i1, . . . , it ∈ {1, 2, . . . , n} be such that M(I) = {mi1, . . . , mit}. It is clear that 1 ≤ t < n. Let us denote the set {1, 2, . . . , n}\{i1, . . . , it} by {it+1, . . . , in}. Consider the ideal J = ∩nj=t+1mij . It is clear that M(J) = {mij|j ∈ {t + 1, . . . , n}}. It follows from I + J = R and IJ ⊆ J(R) that M(I) = Max(R)\M(J) is a closed and open subset of Max(R). Therefore, we obtain from (ii) ⇒ (i) of Lemma 4.2 that C (R) is complemented. Corollary 4.4. Let R be a ring such that |Max(R)| ≥ 2. If R is semiquasilocal, then C (R) is uniquely complemented. Proof. This follows from Proposition 4.3 and Remark 4.1(ii). Let R be a ring. As in [12], we call the graph studied by P.K. Sharma and S.M. Bhatwadekar in [16] as the comaximal graph of R and as in [12], we denote it using the notation Γ(R). It is useful to recall here that the vertex set of Γ(R) is the set of all elements of R and distinct vertices a and b are joined by an edge in Γ(R) if and only if Ra + Rb = R. Moreover, as in [12], we use the notation Γ1(R) to denote the subgraph of Γ(R) induced on U(R); we use Γ2(R) to denote the subgraph of Γ(R) induced on NU(R); for a ring R with |Max(R)| ≥ 2, we use Γ2(R)\J(R) to denote the subgraph of Γ(R) induced on NU(R)\J(R). It is shown in [15, Proposition 3.11] that Γ2(R)\J(R) is complemented if and only if dim( R J(R) ) = 0. We prove in Theorem 4.7 that for a ring R with |Max(R)| ≥ 2, C (R) is complemented if and only if R J(R) ∼= F1 ×F2 ×···×Fn as rings, where Fi is a field for each i ∈{1, 2, . . . , n}. Lemma 4.5. Let R be a ring such that |Max(R)| ≥ 2. The following statements are equivalent: (i) C (R) is complemented. (ii) C ( R J(R) ) is complemented. Proof. (i) ⇒ (ii) We are assuming that C (R) is complemented. Observe that J( R J(R) ) is the zero ideal of R J(R) . Let I J(R) ∈ V (C ( R J(R) )). Then it is clear that I ∈ V (C (R)). As C (R) is complemented, there exists J ∈ V (C (R)) such that I ⊥ J in C (R). We know from the proof of (i) ⇒ (ii) of Lemma 4.2 that there exists b ∈ J such that M(I) = D(b). It is not hard to verify that M( I J(R) ) = D(b + J(R)). Hence, for any I J(R) ∈ V (C ( R J(R) )), M( I J(R) ) is a closed and open subset of Max( R J(R) ). Therefore, we obtain from (ii) ⇒ (i) of Lemma 4.2 that C ( R J(R) ) is complemented. (ii) ⇒ (i) We are assuming that C ( R J(R) ) is complemented. Let I ∈ V (C (R)). Let us denote the ideal I + J(R) by A. Then A J(R) ∈ V (C ( R J(R) )). Since C ( R J(R) ) is complemented, it follows from the proof of (i) ⇒ (ii) of Lemma 4.2 that there exists b ∈ R\J(R) such that M( A J(R) ) = D(b + J(R)). Now, it is easy to verify that M(I) = D(b). Thus for any I ∈ V (C (R)), M(I) is a closed and open subset of Max(R). Therefore, we obtain from (ii) ⇒ (i) of Lemma 4.2 that C (R) is complemented. 96 S. Visweswaran, J. Parejiya / J. Algebra Comb. Discrete Appl. 5(2) (2018) 85–99 Let R be a ring. Recall from [8, Exercise 16, page 111] that R is said to be von Neumann regular if for each element a ∈ R, there exists b ∈ R such that a = a2b. We know from (a) ⇔ (d) of [8, Exercise 16, page 111] that R is von Neumann regular if and only if dimR = 0 and R is reduced. Hence, if R is von Neumann regular, then J(R) = nil(R) = (0). Let R be a von Neumann regular ring. Let a ∈ R. We know from (1) ⇒ (3) of [8, Exercise 24, page 113] that a = ue, where u is a unit of R and e is an idempotent element of R. Hence, any ideal of R is a radical ideal of R. Let I be any ideal R. Since the set of all prime ideals of R equals the set of all maximal ideals of R, it follows from [3, Proposition 1.14] that I = r(I) is the intersection of all the maximal ideals m of R such that m ∈ M(I). Lemma 4.6. Let R be a von Neumann regular ring such that |Max(R)| ≥ 2. The following statements are equivalent: (i) C (R) is complemented. (ii) R ∼= F1 ×F2 ×···×Fn as rings for some n ≥ 2, where Fi is a field for each i ∈{1, 2, . . . , n}. Proof. (i) ⇒ (ii) We are assuming that C (R) is complemented. Since J(R) = (0), it is clear that V (C (R)) equals the set of all nonzero proper ideals of R. Let I be a nonzero proper ideal of R. As C (R) is complemented, we know from the proof of (i) ⇒ (ii) of Lemma 4.2 that M(I) = D(b) for some nonzero nonunit b of R. Note that b = ue, where u is a unit of R and e is an idempotent element of R. Therefore, M(I) = D(b) = D(e) = M(1 − e). Hence, I = ∩m∈M(I)m = ∩m∈M(1−e)m = R(1 − e). This proves that each ideal of R is finitely generated and so, R is Noetherian. Therefore, we obtain from [8, Exercise 21, page 112] that there exist n ∈ N and fields F1 . . . , Fn such that R ∼= F1 ×···× Fn as rings. Since |Max(R)| ≥ 2, it follows that n ≥ 2. (ii) ⇒ (i) Assume that R ∼= F1 × F2 × ···× Fn as rings for some n ≥ 2, where Fi is a field for each i ∈ {1, 2, . . . , n}. Let us denote the ring F1 × F2 × ··· × Fn by T . Note that T is semilocal with {m1 = (0) × F2 ×···× Fn, m2 = F1 × (0) ×···× Fn, . . . , mn = F1 ×···× Fn−1 × (0)} as its set of all maximal ideals. We know from Proposition 4.3 that C (T ) is complemented. As R ∼= T as rings, we get that C (R) is complemented. Theorem 4.7. Let R be a ring such that |Max(R)| ≥ 2. The following statements are equivalent: (i) C (R) is complemented. (ii) R J(R) ∼= F1 ×F2 ×···×Fn as rings for some n ≥ 2, where Fi is field for each i ∈{1, 2, . . . , n}. (iii) R is semiquasilocal. Proof. (i) ⇒ (ii) We are assuming that C (R) is complemented. We know from (i) ⇒ (ii) of Lemma 4.5 that C ( R J(R) ) is complemented. Note that J( R J(R) ) equals the zero ideal of R J(R) . Let a ∈ R be such that a + J(R) is a nonzero nonunit of R J(R) . As C ( R J(R) ) is complemented, we obtain from (i) ⇒ (ii) of Lemma 4.2 that M(a + J(R)) is a closed and open subset of Max( R J(R) ). Therefore, it follows from [15, Lemma 1.2] that dim( R J(R) ) = 0. Thus R J(R) is reduced and zero-dimensional and so, R J(R) is von Neumann regular. It now follows from (i) ⇒ (ii) of Lemma 4.6 that R J(R) ∼= F1 ×F2 ×···×Fn for some n ≥ 2, where Fi is a field for each i ∈{1, 2, . . . , n}. (ii) ⇒ (iii) We are assuming that R J(R) ∼= F1 × F2 × ···× Fn as rings for some n ≥ 2, where Fi is a field for each i ∈ {1, 2, . . . , n}. Hence, R J(R) is semilocal. As |Max(R)| = |Max( R J(R) )|, we get that R is semiquasilocal. (iii) ⇒ (i) Since R is semiquasilocal, we obtain from Proposition 4.3 that C (R) is complemented. Let R be a ring with |Max(R)| ≥ 2. We next discuss some results regarding the S-vertices of C (R). 97 S. Visweswaran, J. Parejiya / J. Algebra Comb. Discrete Appl. 5(2) (2018) 85–99 Remark 4.8. Let R be a ring such that |Max(R)| = 2. Let {m1, m2} denote the set of all maximal ideals of R. We know from (3) ⇒ (1) of [18, Theorem 4.5] that C (R) is a complete bipartite graph with vertex partition V1 and V2, where Vi = {I|M(I) = {mi}} for each i ∈{1, 2}. Hence, it is clear that no vertex of C (R) is a S-vertex of C (R). Therefore, for a ring R with |Max(R)| ≥ 2, in determining the S-vertices of C (R), we assume that |Max(R)| ≥ 3. Proposition 4.9. Let R be a ring such that |Max(R)| ≥ 3. Let I ∈ V (C (R)). Then the following statements are equivalent: (i) I is a S-vertex of C (R). (ii) |Max(R)\M(I)| ≥ 2. Proof. (i) ⇒ (ii) Now, I ∈ V (C (R)) and we are assuming that I is a S-vertex of C (R). Hence, there exist distinct I1, I2, I3 ∈ V (C (R)) such that I −I1, I −I2, I2 −I3 are edges of C (R), but there is no edge joining I1 and I3 in C (R). Since I1 and I3 are not adjacent in C (R), there exists m ∈ Max(R) such that I1 + I3 ⊆ m. It follows from I + I1 = I2 + I3 = R that I 6⊆ m and I2 6⊆ m. As I2 is proper ideal of R, there exists m′ ∈ Max(R) such that I2 ⊆ m′. It is clear that m 6= m′ and it follows from I + I2 = R that I 6⊆ m′. Hence, {m, m′}⊆ Max(R)\M(I) and so, |Max(R)\M(I)| ≥ 2. (ii) ⇒ (i) Let I ∈ V (C (R)) be such that |Max(R)\M(I)| ≥ 2. Let {m1, m2} ⊂ Max(R) be such that I 6⊆ mi for each i ∈ {1, 2}. By hypothesis, |Max(R)| ≥ 3. Hence, there exists m3 ∈ Max(R)\{m1, m2}. Observe that I − m1, I − m2, m2 − m1 ∩ m3 are edges of C (R), but there is no edge of C (R) joining m1 and m1 ∩m3. This proves that I s a S-vertex of C (R). Acknowledgment: We are thankful to the referee for the support. References [1] D. F. Anderson, P. S. Livingston, The zero–divisor graph of a commutative ring, J. Algebra 217(2) (1999) 434–447. [2] D. F. Anderson, R. Levy, J. Shapiro, Zero–divisor graphs, von Neumann regular rings, and Boolean Algebras, J. Pure Appl. Algebra 180(3) (2003) 221–241. [3] M. F. Atiyah, I. G. Macdonald, Introduction to Commutative Algebra, Addison–Wesley, Reading, Massachusetts, 1969. [4] R. Balakrishnan, K. Ranganathan, A Textbook of Graph Theory, Universitext, Springer, 2000. [5] M. Behboodi, Z. Rakeei, The annihilating-ideal graph of commutative rings I, J. Algebra Appl. 10(4) (2011) 727–739. [6] M. Behboodi, Z. Rakeei, The annihilating-ideal graph of commutative rings II, J. Algebra Appl. 10(4) (2011) 741–753. [7] N. Deo, Graph Theory with Applications to Engineering and Computer Science, Prentice Hall of India Private Limited, New Delhi, 1994. [8] R. Gilmer, Multiplicative Ideal Theory, Marcel–Dekker, New York, 1972. [9] M. I. Jinnah, S. C. Mathew, When is the comaximal graph split?, Comm. Algebra 40(7) (2012) 2400–2404. [10] I. Kaplansky, Commutative Rings, The University of Chicago Press, Chicago, 1974. [11] R. Levy, J. Shapiro, The zero–divisor graph of von Neumann regular rings, Comm. Algebra 30(2) (2002) 745–750. [12] H. R. Maimani, M. Salimi, A. Sattari, S. Yassemi, Comaximal graph of commutative rings, J. Algebra 319(4) (2008) 1801–1808. [13] S. M. Moconja, Z. Z. Petrovic, On the structure of comaximal graphs of commutative rings with identity, Bull. Aust. Math. Soc. 83(1) (2011) 11–21. 98 https://doi.org/10.1006/jabr.1998.7840 https://doi.org/10.1006/jabr.1998.7840 https://doi.org/10.1016/S0022-4049(02)00250-5 https://doi.org/10.1016/S0022-4049(02)00250-5 https://mathscinet.ams.org/mathscinet-getitem?mr=3525784 https://mathscinet.ams.org/mathscinet-getitem?mr=3525784 https://doi.org/10.1007/978-1-4614-4529-6 https://doi.org/10.1142/S0219498811004896 https://doi.org/10.1142/S0219498811004896 https://doi.org/10.1142/S0219498811004902 https://doi.org/10.1142/S0219498811004902 https://mathscinet.ams.org/mathscinet-getitem?mr=360322 https://mathscinet.ams.org/mathscinet-getitem?mr=360322 https://doi.org/10.1080/00927872.2011.591861 https://doi.org/10.1080/00927872.2011.591861 https://doi.org/10.1081/AGB-120013178 https://doi.org/10.1081/AGB-120013178 https://doi.org/10.1016/j.jalgebra.2007.02.003 https://doi.org/10.1016/j.jalgebra.2007.02.003 https://doi.org/10.1017/S0004972710001875 https://doi.org/10.1017/S0004972710001875 S. Visweswaran, J. Parejiya / J. Algebra Comb. Discrete Appl. 5(2) (2018) 85–99 [14] A. M. Rahimi, Smarandache vertices of the graphs associated to the commutative rings, Comm. Algebra 41(5) (2013) 1989–2004. [15] K. Samei, On the comaximal graph of a commutative ring, Canad. Math. Bull. 57(2) (2014) 413–423. [16] P. K. Sharma, S. M. Bhatwadekar, A note on graphical representation of rings, J. Algebra 176(1) (1995) 124–127. [17] S. Visweswaran, P. Sarman, On the planarity of a graph associated to a commutative ring and on the planarity of its complement, São Paulo J. Math. Sci. 11(2) (2017) 405–429. [18] M. Ye, T. Wu, Co–maximal ideal graphs of commutative rings, J. Algebra Appl. 11(6) (2012) Article ID. 1250114 (14 pages). 99 https://doi.org/10.1080/00927872.2012.744025 https://doi.org/10.1080/00927872.2012.744025 http://dx.doi.org/10.4153/CMB-2013-033-7 https://doi.org/10.1006/jabr.1995.1236 https://doi.org/10.1006/jabr.1995.1236 https://doi.org/10.1007/s40863-017-0065-9 https://doi.org/10.1007/s40863-017-0065-9 https://doi.org/10.1142/S0219498812501149 https://doi.org/10.1142/S0219498812501149 Introduction Some preliminary results and on the planarity of C(R) When is C(R) a split graph? Some more results on C(R) References