437
Views
1
CrossRef citations to date
0
Altmetric
Original Article

On n-connected splitting matroidsFootnote

&
Pages 50-56 | Received 31 May 2017, Accepted 11 Dec 2017, Published online: 10 Jun 2020

Abstract

In general, the splitting operation on a binary matroid M does not preserve the connectivity of M. In this paper, we provide sufficient conditions to preserve n-connectedness of a binary matroid under splitting operation. As a consequence, for an (n+1)-connected binary matroid M, we give a precise characterization of when the splitting matroid MT is n-connected.

1 Introduction

We refer to Oxley [Citation1] for standard terminology in graphs and matroids. The matroids under consideration are loopless. For a matroid M, let E(M) and r(M) denote the ground set and the rank function of M, respectively. For sets A and B, let AΔB denote the set (AB)(BA). GT.

Fleischner [Citation2] defined the splitting operation for graphs with respect to a pair of adjacent edges. As an extension of this operation to binary matroids, Raghunathan, Shikare and Waphare [Citation3] defined the splitting operation for binary matroids with respect to a pair of elements. The effects of the splitting operation on various parameters of a matroid like graphicness, cographicness, and connectedness have been well studied in the literature, for example,see [4–6,3,7,8Citation[4]Citation[5]Citation[6]Citation[3]Citation[7]Citation[8]].

The splitting operation for binary matroids with respect to any set TE(M), which is a generalization of the splitting operation with respect to a pair of elements, is defined by Shikare, Azadi and Waphare [Citation9] as follows.

Definition 1.1

[Citation9]

Let M be a binary matroid with standard matrix representation A over the field GF(2) and let TE(M). Let AT be the matrix obtained by adjoining one extra row to the matrix A whose entries are 1 in the columns labeled by the elements of the set T and zero otherwise. The vector matroid of the matrix AT, denoted by MT, is called as the splitting matroid of M with respect to T, and the transition from M to MT is called as the splitting operation with respect to T. (See Remark 4.3.)

The splitting operation on a matroid M is related to a lift of M. A matroid M is a coextension of M if MAM for some nonempty subset A of E(M). If A =1, then M is a lift of M. Let M be a binary matroid and TE(M) and M be a lift of M by a with aE(M) such that aT is a cocircuit in M. Then M a MT.

In general, the splitting operation does not preserve the connectivity of a given matroid. Borse and Dhotre [Citation5] obtained the following result to get a connected splitting matroid MT when T =2.

Theorem 1.2

[Citation5]

Let M be a connected and vertically 3-connected binary matroid and TE(M) with T =2. Suppose every cocircuit Q of M with TQ has size at least 4 and further, Q does not contain a 2-circuit of M. Then MT is connected.

Malvadkar, Shikare and Dhotre [Citation10] obtained the following generalization of Theorem 1.2 by assuming an additional condition on the girth of M.

Theorem 1.3

[Citation10]

Let n2 be an integer and M be an n-connected, vertically (n+1)-connected binary matroid with E(M) 2n+1 and girth at least n+1. Suppose TE(M) with T n. Then the matroid MT is n-connected if and only if for AE(M) with A =n1, there is a circuit C in M such that CT is odd and CA=.

However, in Lemma 3.7 we prove that an n-connected, vertically (n+1)-connected binary matroid M with E(M) 2n+1 and girth at least n+1 is (n+1)-connected. Therefore the matroid M considered in Theorem 1.3 is actually (n+1)-connected, which is a stronger condition on M to preserve n-connectedness under the splitting operation. Without assuming this stronger condition, we generalize Theorem 1.2 as follows.

Theorem 1.4

Let n2 be an integer and M be an n-connected and vertically (n+1)-connected binary matroid with E(M) 2n+1 and TE(M) with T =n. Suppose T is not an n-circuit in M and for every cocircuit Q of M intersecting T, Q max 2 QT ,n+1 and further, QΔT is not an n-circuit of M. Then MT is n-connected.

We also obtain, in Theorem 4.1, some necessary conditions for MT to be n-connected. On combiningTheorems 1.4, 4.1, Lemma 3.7 and Theorem 1.3, we prove the following characterization for the matroid MT to be n-connected.

Theorem 1.5

Let n2 and M be an (n+1)-connected binary matroid with E(M) 2n+1 and TE(M) with T =n. Then the following statements are equivalent.

(1)

The matroid MT is n-connected.

(2)

For every cocircuit Q in M intersecting T, Q max 2 QT ,n+1 .

(3)

For AE(M) with A =n1, there is a circuit C in M such that CT is odd and CA=ϕ.

The circuits, cocircuits and rank function of the splitting matroid MT are given in the second section. In Section 3, we provide some basic concepts and results regarding connectedness and obtain few lemmas. The proofs of the main results are given in the last section. We also discuss the sharpness of Theorem 1.4 in the remark given at the last section.

2 Circuits and cocircuits of MT

It is clear from Definition 1.1 that the ground sets of the matroids M and MT are same. Further, M=MT if T is a cocircuit of M. Moreover, T is a cocircuit of MT if no proper subset of T is a cocircuit in M. Mills [Citation11] characterized the cocircuits of the splitting matroid MT in terms of the cocircuits of M when T =2.

Theorem 2.1

[Citation11]

Let M be a binary matroid with the collection of cocircuits C and let T= x,y E(M). Suppose T does not contain a cocircuit of M. Then each cocircuit of MT other than T belongs to one of the following collections.

(i)

C1= CT:CCandTisapropersubsetofC .

(ii)

C2= C:CCandCdoesnotcontainanymemberofC1 .

(iii)

C3= CΔT:CC, CT =1andCdoesnotcontainanymemberofC1 .

(iv)

C4= (C1C2)T:C1,C2C,C1C2=ϕ,xC1,yC2andCi does not contain any memberof C1 .

Azanchiler [Citation12] generalized Theorem 2.1 for arbitrary T as follows.

Theorem 2.2

[Citation12], pp. 35

Let M be a binary matroid with the collection of cocircuits C and TE(M) with T 2. Suppose T does not contain a cocircuit of M. Then each cocircuit of MT other than T belongs to one of the following collections.

(i)

C1= CT:CCandTisapropersubsetofC .

(ii)

C2= C:CCandCdoesnotcontainanymemberofC1 .

(iii)

C3= CΔT:CC,1 CT < T andCdoesnotcontainanymemberofC1 .

(iv)

C4= (C1C2Ck)T:CiC,CiTϕ,Ci are mutually disjoint and none of them contains any member of C1 .

The circuits and the rank function of the splitting matroid MT are given in [Citation9].

Lemma 2.3

[Citation9]

Let M be a binary matroid and TE(M) and C be the collection of circuits of M. Then the collection of circuits of MT is C1C2, where C1= CC: CT iseven and

C2= C1C2:C1,C2C,C1C2=ϕ, CiT isoddandC1C2containsnomemberofC1 .

Lemma 2.4

[Citation9]

Let M be a binary matroid, TE(M) such that T 2 and T does not contain a cocircuit of M. Suppose r and r are the rank functions of the matroid M and MT, respectively, and AE(M). Then

(i) r(A)=r(A), if A does not contain a circuit of M containing an odd number of elements of T;

(ii) r(A)=r(A)+1, otherwise;

(iii) r(MT)=r(M)+1.

3 Connected and vertically connected matroids

We provide here some basic definitions and results given in [Citation1]. Let n2 be an integer and M be a matroid with ground set E(M) and rank function r. An n-separation of M is a partition (X,Y) of E(M) such that min X , Y n and r(X)+r(Y)r(M)n1. If M is k-separated for some k, then connectivity λ(M) of M is the minimum j such that M has a j-separation; otherwise it is . An n-separation (X,Y) of M is a vertical n-separation if min r(X),r(Y) n. The vertical connectivity k(M) of M is the minimum j such that M has a vertical j-separation if M has two disjoint cocircuits, otherwise k(M)=r(M). The matroid M is n-connected if nλ(M) and it is vertically n-connected if nk(M). Obviously, every n-connected matroid is vertically n-connected. If M contains a circuit, then its girth, denoted by g(M), is the minimum circuit size. Similarly, if M contains a cocircuit, then its cogirth is the minimum cocircuit size.

We need the following known results.

Theorem 3.1

[Citation1], pp. 279

If a matroid M is not isomorphic to any uniform matroid Ur,m with m2r1, then λ(M)=min k(M),g(M) .

Theorem 3.2

[Citation1], pp. 273

If n2 and M is an n-connected matroid with E(M) 2(n1), then all circuits and all cocircuits of M have at least n elements.

Lemma 3.3

[Citation1], pp. 274

Let M be an n-connected matroid having at least 2n1 elements. Then M has no n-element subset that is both a circuit and a cocircuit.

Lemma 3.4

[Citation1], pp. 70

Let M be a matroid and let YE(M). Then Y is a cocircuit if and only if E(M)Y is a hyperplane of M.

Corollary 3.5

[Citation4]

Let M be a matroid and let YE(M) such that r(MY)=r(M)1. Then Y contains a cocircuit of M.

Lemma 3.6

Let n2 be an integer and let M be an n-connected, vertically (n+1)-connected binary matroid with E(M) 2n+1. Then the cogirth of M is at least n+1.

Proof

By Theorem 3.2, the cogirth of M is at least n. Suppose it is n. Then there is a cocircuit Q of M with Q =n. Let Q=E(M)Q. By Lemma 3.4, Q is a hyperplane and so r(Q)=r(M)1. Again by Theorem 3.2, r(Q)n1 and r(Q)n1. Therefore r(Q)+r(Q)r(M)=r(Q)1. If r(Q)=n1, then (Q,Q) is a vertical (n1)-separation of M, a contradiction. Hence r(Q)=n. Therefore, if r(Q)n, then (Q,Q) is a vertical n-separation of M, again a contradiction. Consequently, r(Q)=n1. This implies that every n-element set in M Q=MQ is a circuit. It follows that M Q is isomorphic to the uniform matroid Un1,q, where q= Q n+1.

If n3, then qn+14 and so M contains the uniform matroid U2,4 as a minor, a contradiction to the fact that M is binary. Hence n=2. Therefore MQ is isomorphic to U1,q. This implies that MQ is the cycle matroid of the connected graph on two vertices and q edges. It follows that M=M(G), where G is the graph obtained from a triangle by adjoining q1 edges parallel to one of the edge. As G contains a vertex of degree two, it is not 3-connected and so M(G) is not vertically 3-connected. Therefore M is not vertically (n+1)-connected, a contradiction. This completes the proof. □

Lemma 3.7

Let n2 be an integer and let M be an n-connected, vertically (n+1)-connected binary matroid with E(M) 2n+1 and girth at least n+1. Then M is (n+1)-connected.

Proof

Since M is vertically (n+1)-connected, k(M)n+1. Also, the girth g(M) of M is at least n+1. By Lemma 3.6, the cogirth of M is also at least n+1. Let r be the rank of M. Then r+1g(M)n+1 and therefore rn2. Suppose M is isomorphic to the uniform matroid Ur,m for some m with m2r1. Then m>r. Suppose m=r+1. Then M is a circuit and hence the cogirth of M is 2, a contradiction. Thus mr+2. This implies that Ur,m contains U2,4 as a minor. Hence U2,4 is a minor of the binary matroid M, a contradiction. Therefore M is not isomorphic to the uniform matroid Ur,m with m2r1. Now, by Theorem 3.1, λ(M)=min k(M),g(M) n+1. Hence M is (n+1)-connected. □

4 Main results

In the following theorem, we provide some necessary conditions for the splitting matroid MT to be n-connected.

Theorem 4.1

Let n2 and M be an n-connected and vertically (n+1)-connected binary matroid with E(M) 2n+1 and TE(M) with T =n. Suppose the splitting matroid MT is n-connected and Q is a cocircuit of M intersecting T. Then

(i) Q max 2 QT ,n+1 ;

(ii) neither T nor QΔT is an n-circuit in M if n is even.

Proof

By Theorem 3.2, every circuit of M has size at least n and, by Lemma 3.6, every cocircuit of M has size at least n+1. Similarly, every cocircuit of MT contains at least n elements. As T =n, T does not contain any cocircuit of M. By Lemma 2.4 (iii), r(MT)=r(M)+1.

Suppose Q is a cocircuit of M containing k>0 elements of T. If T is a subset of Q, then, by Theorem 2.2 (i), QT is a cocircuit of the splitting matroid MT. Further, if T is not a subset of Q, then QΔT is a cocircuit of MT by Theorem 2.2 (iii). In both the cases, QΔT is a cocircuit of MT. Therefore, n QΔT = QT + TQ = Q + T 2k= Q +n2k giving 2k Q . Thus Q max 2k,n+1 . This proves (i).

We now prove (ii). Suppose n is even. Assume that QΔT=C is an n-circuit of M. In a binary matroid, intersection of a circuit and a cocircuit contains an even number of elements. Therefore CQ = QT is even. Consequently, CT = TQ =n QT is even as n is even. Hence, by Lemma 2.3, C is an n-circuit of MT. Already C is an n-cocircuit of MT. This is a contradiction by Lemma 3.3. Assume that T is an n-circuit of M. Then, by Lemma 2.3, T is also a circuit of MT. Further, by Definition 1.1, T is a cocircuit of MT. This is a contradiction by Lemma 3.3. □

We now prove Theorem 1.4.

Proof of Theorem 1.4

By Theorem 3.2, the girth of M is at least n and, by Lemma 3.6, the cogirth of M is at least n+1. Hence T does not a contain cocircuit of M as T =n. Therefore, by Lemma 2.4 (iii), r(MT)=r(M)+1. We prove the result by contradiction. Suppose MT is not n-connected. Then it has an (n1)-separation (A,B). Therefore min A , B n1 and r(A)+r(B)r(MT)n2.Suppose A =n1. Then A is independent in M and so, by Lemma 2.3, A is independent in MT also. Suppose A contains a cocircuit Q of MT. Then Q A =n1 and hence, by Theorem 3.2, Q is not a cocircuit of M. By Theorem 2.2, Q belongs to one of the three classes C1, C3 or C4. Suppose QC1. Then Q=CT, where C is a cocircuit of M containing T. Then, by hypothesis, C 2 QT =2 T =2n and so Q n, a contradiction. Suppose QC4. Then Q=(C1C2Ck)T, where k2 and Ci are mutually disjoint cocircuits of M and each of them contains at least one element of T. Since Ci n+1 and T =n, we have Q (C1C2)T C1 + C2 T 2n+2n=n+2>n1 Q , a contradiction. Hence QC3. Therefore Q=CΔT, where C is a cocircuit of M intersecting T but not containing T. Then C max 2k,n+1 , where k= CT . Hence Q = CΔT = C + T 2 CT 2k+n2k=n>n1 Q , a contradiction. Thus A does not contain a cocircuit of MT.

Hence r(B)=r(MT). This gives n1=r(A)=r(A)+r(B)r(MT)n2, which is a contradiction. Similarly, we get a contradiction if B =n1.

Therefore A n and B n. If r(A)n and r(B)n, then, r(A)+r(B)r(M)r(A)+r(B)(r(MT)1)n2+1=n1and so (A,B) is a vertical n-separation of M, a contradiction. Hence r(A)=n1 or r(B)=n1. Without loss of generality assume that r(A)=n1. Therefore every subset of A containing n elements is dependent in M A=MB and so in M. This shows that every subset of A of size n is an n-circuit of M A. If A n+1, then U2,4 is a minor of the binary matroid M A, which is a contradiction. Hence A =n. Thus A is an n-circuit of M. By Lemma 3.6, A does not contain a cocircuit of M. This gives r(B)=r(M).

Now, if r(B)=r(B)+1, then r(A)+r(B)r(M)r(A)+(r(B)1)(r(MT)1)=r(A)+r(B)r(MT)n2.This implies that (A,B) is an (n1)-separation of M, a contradiction. Hence r(B)=r(B). If A does not contain a cocircuit of MT, then r(MT)=r(B)=r(B)=r(M), a contradiction to Lemma 2.4 (iii).

Thus A contains a cocircuit Q of MT. Therefore Q A =n. As cogirth of M is at least n+1, Q is not a cocircuit of M. Further, AT as T is not an n-circuit of M. Therefore QT. By Theorem 2.2, QCi for some i 1,3,4 . Suppose QC1. Then Q=CT for some cocircuit C of M containing T. Therefore C 2 CT =2 T =2n. Hence Q = CT 2nn=n. This implies that Q=A. Since A is an n-circuit, CΔT is an n-circuit of M, a contradiction to the assumption that CΔT is not an n-circuit of M. Suppose QC3. Then Q=CΔT for some cocircuit C of M intersecting T but not containing T. Then C 2 CT and so Q = CΔT = C + T 2 CT T =n. This gives CΔT=Q=A. Hence CΔT is an n-circuit of M, a contradiction. Therefore QC4. Hence Q=(C1C2Ck)T, where k2 and Ci are mutually disjoint cocircuits of M and each of them contains at least one element of T. Since Ci n+1 and T =n, we have Q (C1C2)T C1 + C2 T 2n+2n=n+2>n Q , a contradiction. □

We prove Theorem 1.5, which is restated here for convenience. In view of Lemma 3.7, the equivalence of statements (1) and (3) follows directly from Theorem 1.3. However, we provide a shorter proof of the same.

Theorem 4.2

Let n2 and M be an (n+1)-connected binary matroid with E(M) 2n+1 and TE(M) with T =n. Then the following statements are equivalent.

(1)

The matroid MT is n-connected.

(2)

For every cocircuit Q in M intersecting T, Q max 2 QT ,n+1 .

(3)

For AE(M) with A =n1, there is a circuit C in M such that CT is odd and CA=ϕ.

Proof

Since M is an (n+1)-connected, it is n-connected as well as vertically (n+1)-connected. Further, by Theorem 3.2 every circuit and every circuit of M contains at least n+1 elements. Therefore T does not contain a cocircuit of M as T =n.

(1) (2) follows from Theorem 4.1.

(2) (1) follows from Theorems 4.1 and 3.2.

(1) (3). Suppose MT is n-connected but (3) does not hold. Then there is a set AE(M) with A =n1 such that CA ϕ for every circuit C in M with CT an odd integer. Let B=E(M)A. Then B n+1 and, by Lemma 2.4 (i), r(B)=r(B). Since A is not a circuit of M, by Lemma 2.3, A is independent in MT and so r(A)=r(A). By Theorem 3.2, A does not contain a cocircuit of M. Therefore r(B)=r(M) by Corollary 3.5. Hence r(A)+r(B)r(MT)=r(A)+r(B)r(M)1=r(A)1=n2.Thus (A,B) is an (n1)-separation of MT, a contradiction.

(3) (1). Assume that (3) holds but (1) does not hold. Then MT has an (n1)-separation (A,B). Therefore min A , B n1 and r(A)+r(B)r(MT)n2.

Suppose A n and B n. By Lemma 2.4, r(A)+r(B)r(M)r(A)+r(B)r(MT)+1n2+1n1.Hence (A,B) is an n-separation of M, a contradiction to the fact that M is (n+1)-connected.

Therefore we may assume that A =n1. By (3) and Lemma 2.4 (ii), r(B)=r(B)+1. Consequently, r(A)+r(B)r(M)r(A)+r(B)1r(MT)+1n2.This shows that (A,B) is an (n1)-separation of M, a contradiction. □

Remark 4.3

In the following example, we show that the condition on the matroid M given in Theorem 1.4 that M is vertically (n+1)-connected cannot be dropped. The graph G of (a) is 2-connected but not 3-connected. Hence its cycle matroid M(G) is 2-connected but not vertically 3-connected. Let T= e,f , where e and f are the edges of G as shown in (a). Then T satisfies the hypothesis of Theorem 1.4. The graph GT obtained by splitting away the edges e and f is shown in (b). Clearly, M(G)T=M(GT). However, M(GT) is not 2-connected and so M(G)T is not 2-connected. (a) The graph G

Fig. 1 Splitting of a graph.

Notes

Peer review under responsibility of Kalasalingam University.

References

  • Oxley J.G. Matroid Theory 1992 Oxford University Press Oxford
  • Fleischner H. Eulerian Graphs and Related Topics Part 1, Vol. 1 1990 North Holland Amsterdam
  • Raghunathan T.T. Shikare M.M. Waphare B.N. Splitting in a binary matroid Discrete Math. 184 1998 267 271
  • Borse Y.M. A note on n-connected splitting-off matroids Ars Combin. 128 2016 279 286
  • Borse Y.M. Dhotre S.B. On connected splitting matroids Southeast Asian Bull. Math. 36 1 Special Issue: Special Issue: International Conference on Discrete Mathematics 2012 17 21
  • Borse Y.M. Shikare M.M. Dalvi K.V. Excluded-Minors for the class of cographic splitting matroids Ars Combin. 115 2014 219 237
  • Shikare M.M. Dalavi K.V. Dhotre S.B. Splitting-off operation for binary matroids and its applications Graphs Combin. 27 6 2011 871 882
  • Shikare M.M. Waphare B.N. Excluded-Minors for the class of graphic splitting matroids Ars Combin. 97 2010 111 127
  • Shikare M.M. Azadi G. Waphare B.N. Generalized splitting operation for binary matroids and its applications J. Indian Math. Soc. New Ser. 78 1–4 2011 145 154
  • Malavadkar P.P. Shikare M.M. Dhotre S.B. A characterization of n-connected splitting matroids Asian-Eur. J. Combin. 7 4 2014 7 Article 14500600
  • Mills A. On the cocircuits of a splitting matroid Ars Combin. 89 2008 243 253
  • Azanchiler H. (Ph.D. Thesis) Some New Operation on Matroids and Related Results 2005 University of Pune