388
Views
0
CrossRef citations to date
0
Altmetric
Research Articles

On characterizing proper max-point-tolerance graphs

Pages 313-325 | Received 08 May 2020, Accepted 28 Jul 2023, Published online: 16 Aug 2023

Abstract

Max-point-tolerance graphs (MPTG) were introduced by Catanzaro et al. in 2017 as a generalization of interval graphs. This graph class has many practical applications in the study of the human genome as well as in signal processing for networks. The same class of graphs was also studied by Soto and Caro in 2015 with a different name, p -BOX(1) graphs. In our article, we consider a natural subclass of max-point-tolerance graphs, namely, proper max-point-tolerance graphs (proper MPTG), where intervals associated with the vertices are not contained in each other properly. We present the first characterization theorem of this graph class by defining certain linear ordering on the vertex set. In the course of this study, we prove proper max-point-tolerance graphs are asteroidal triple-free, and perfect. We also find that proper max-point-tolerance graphs are equivalent to unit max-point-tolerance graphs. Further, we show that MPTG (proper MPTG) and max-tolerance graphs (proper max-tolerance graphs) are incomparable. In conclusion, we demonstrate relations between proper MPTG with other variants of MPTG and max-tolerance graphs.

2010 MSC:

1 Introduction

A graph G=(V,E) is an interval graph if each vertex is mapped to an interval on real line in such a way so that any two of the vertices are adjacent if and only if their corresponding intervals intersect. Many people [Citation5, Citation6] have done extensive research work on interval graphs for their worldwide practical applications in Computer Science and Discrete Mathematics. In 1962, Lekkerkerker and Boland [Citation12] first characterized this graph class by proving it chordal and asteroidal triple-free. After that, many combinatorial problems, like the maximal clique problem, the independent set problem, the coloring problem, the domination problem, and most importantly, the recognition problem, have been solved in linear time for interval graphs.

Due to their significant applications in many theoretical and practical situations, interval graphs were generalized to several variations [Citation1, Citation10, Citation14]. One of the most relevant variations of this graph class related to this article is tolerance graph which was first introduced by Golumbic and Monma [Citation8] in 1982. A simple undirected graph G=(V,E) is a min-tolerance graph (typically known as tolerance graph) if each vertex uV corresponds to a real interval Iu and a positive real number tu, called tolerance, such that uv is an edge of G if and only if |IuIv|min{tu,tv}. Golumbic and Trenk [Citation9] introduced max-tolerance graphs where each vertex uV corresponds to a real interval Iu and a positive real number tu (known as tolerance) such that uv is an edge of G if and only if |IuIv|max{tu,tv}. For max-tolerance graphs, we may assume tu|Iu| for each uV since otherwise u becomes isolated. Max-tolerance graphs having a representation in which no interval properly contains another are called proper max-tolerance graphs. Some combinatorial problems, such as finding the maximal cliques, were obtained in polynomial time whereas the recognition problem was proved to be NP-hard [Citation11] for max-tolerance graphs in 2006. Also, a geometrical connection of max-tolerance graphs to semi-squares was found by Kaufmann et al. [Citation11]. The research work by Golumbic and Trenk [Citation9] can be referred to for more information on tolerance graphs.

In 2015, Soto and Caro [Citation15] introduced a new graph class, namely p -BOX graphs, where each vertex corresponds to a box and a point inside it in the Euclidean d-dimensional space. Any two vertices are adjacent if and only if each of the corresponding points is at the intersection of their respective boxes. When the dimension is one, this graph class is denoted by p-BOX(1). In 2017, Catanzaro et al. [Citation3] studied these dimension one graphs independently by giving them a different name, max-point-tolerance graphs (MPTG) where each vertex uV corresponds to a pair of an interval and a point (Iu,pu), where Iu is an interval on the real line and puIu, such that uv is an edge of G if and only if {pu,pv}IuIv. More precisely, each pair of intervals can “tolerate” a non-empty intersection without creating an edge as long as at least one distinguished point does not fall in the intersection. Then G is said to be represented by {(Iv,pv)|vV}. They characterize MPTG by defining a special linear ordering to its vertex set. The graphs MPTG have a number of practical applications in human genome studies for DNA scheduling and in modeling telecommunication networks for sending and receiving messages [Citation3]. Recently in [Citation13] central-max-point-tolerance graphs (central MPTG) have been studied considering pu as the center point of Iu for each uV. In the course of this research, the class of central MPTG is proven to be equivalent to unit max-tolerance graphs, where each interval possesses a unit length.

A natural and well-studied subclass of interval graphs is the class of proper interval graphs where no interval contains another properly. This graph class has various characterizations in terms of the linear ordering of its vertices, the consecutive 1’s property of its associated augmented adjacency matrix, the forbidden graph structure (K1,3) etc [Citation7]. It is known [Citation2] that proper interval graphs are the same as unit interval graphs where each interval possesses the same length. So the natural question arises as to what will be the characterization of MPTG when the intervals are proper.

In this article, we introduce proper-max-point-tolerance graphs (proper MPTG). It is a MPTG having a representation where no interval is properly contained in another. We find this graph class to be asteroidal triple-free and perfect. We obtain the first characterization theorem of this graph class by introducing certain linear ordering on its vertex set, which can be an independent interest of study for the class of proper MPTG graphs. Interestingly, we note that interval graphs form a strict subclass of proper MPTG. Analogous to unit interval graphs, we define unit-max-point-tolerance graphs (unit MPTG) as a MPTG where all the intervals have equal length. Next, we show that proper MPTG graphs are the same as unit MPTG. We investigate the connection between max-tolerance graphs (proper max-tolerance graphs) and MPTG (proper MPTG) and be able to prove that these graph classes are incomparable. Incidentally, we settle a query raised in the book [Citation9] by Golumbic and Trenk about whether unit max-tolerance graphs are the same as proper max-tolerance graphs or not. In the Conclusion section, we show the relations between the subclasses of MPTG and max-tolerance graphs related to proper MPTG.

2 Preliminaries

A matrix whose entries are only 0, 1 is a binary matrix. The augmented adjacency matrix of a simple undirected graph G=(V,E) is obtained from the adjacency matrix of G by replacing 0’s by 1’s along the principal diagonal [Citation7] and is denoted by A(G) throughout the article.

The following characterization of MPTG is known:

Theorem 2.1.

[Citation3, Citation13] Let G=(V,E) be a simple undirected graph. Then the following are equivalent.

  1. G is a MPTG.

  2. There is an ordering of vertices such that for every quadruplet x, u, v, y the following condition holds: Ifx<u<v<y and xv,uyE,thenuvE.(Tex translation failed) (2.1)

  3. There is an ordering of vertices of G such that for any u<v, u,vV, (2.2) uvEuwE for all w>v or, wvE for all w<u.(2.2)

  4. There exists an ordering of vertices such that every 0 above the principal diagonal of the augmented adjacency matrix A(G) has either all entries right to it are 0 (right open) or, all entries above it are 0 (up open).

If the vertices of G satisfy any of the aforementioned conditions with respect to a vertex ordering, then we call the ordering as MPTG ordering of G. One can verify that MPTG orderings need not be unique for G.

From [Citation13] one can note that MPTG and max-tolerance graphs are not the same. In Theorem 2.4 we prove that these graph classes are incomparable. To obtain the proof, we need certain properties of proper max-tolerance graphs, as described in Corollary 2.3.

Observation 2.2.

Let G=(V,E) be a proper max-tolerance graph. Then there exists a vertex ordering () of V such that for every quadruplet v1,v2,v3,v4 the following holds:  If v1v2v3v4,v1v3,v2v4Ev2v3E and (v1v2E or v3v4E or v1v2,v3v4E).(2.3)

Proof.

Let G be a proper max-tolerance graph with interval Ii=[ai,bi] and tolerance ti for each vertex viV. Then arranging the intervals according to the increasing order of their left endpoints, one can show b2a3=b2a4+a4a3>t2 as a4>a3 and v2v4E. Again b2a3=b2b1+b1a3>t3 as b2>b1 and v1v3E. Therefore we get |I2I3|=b2a3>max{t2,t3} which imply v2v3E. Now if v1v2E then b1a2<t1 or t2. If b1a2<t1 then t1>b1a2>b1a3t1 introduces contradiction. Hence b1a2<t2. Thus we get t3b1a3<b1a2<t2. Again if v3v4E one can find t2<t3 which is again a contradiction. Hence the result follows. ▪

Corollary 2.3.

K2,3 is not a proper max-tolerance graph.

Proof.

On the contrary, if K2,3 is a proper max-tolerance graph, then one can find a vertex ordering ( say) from Observation 2.2 with respect to which the vertices of K2,3 satisfies (2.3). Let {v1,v5},{v2,v3,v4} are two partite sets of it. Now if v1v5. Then adjacency of vertices helps us to conclude that any two vertices from {v2,v3,v4} can not occur simultaneously within v1,v5 or left to v1 or right to v5 in . Hence exactly one among them can occur within v1,v5 and between two other vertices, one (say v4) occurs left to v1 and the other (say v3) occurs right to v5 respectively. But again (2.3) gets contradicted for vertices {v4,v1,v5,v3} as v1,v5 are nonadjacent. A similar contradiction will arise when v5v1. ▪

Theorem 2.4.

Max-tolerance graphs and the class of max-point-tolerance graphs are not comparable.

Proof.

From [Citation13] one can verify that the graph G2 in Example 2.6 (see ) is a max-tolerance graph, which is not a MPTG. Now in Lemma 2.5 we prove Km,n, m,n13 is not a max-tolerance graph. Next, in Lemma 2.6 we show that Km,n for any two positive integers m, n is a max-point-tolerance graph. Thus, Km,n for m,n13 and G2 separate MPTG from max-tolerance graphs. Hence, these graph classes become incomparable. ▪

Figure 1: G2 in Example 2.6 of [Citation13].

Figure 1: G2 in Example 2.6 of [Citation13].

We verify the proof of the aforementioned theorem with the help of the following lemmas:

Lemma 2.5.

Complete bipartite graph Km,n when m,n13 is not a max-tolerance graph.

Proof.

We assume on the contrary that Km,n for m,n13 is a max-tolerance graph with interval representation {Iv=[av,bv]|vV} and tolerance tv for each vertex vV where the vertex set V is partitioned into two partite sets V1={x1,,xm}, V2={y1,,yn} and E be the edge set of Km,n.

Claim 1: No two intervals from same partite set can containFootnote1 an interval from other partite set.

proof:

On the contrary, we assume that Ix1 and Ix2 contain Iy1. As x1y1,x2y1E, |Iy1|=|Ix1Iy1|=|Ix2Iy1|tx1,tx2 from definition. Now if |Ix1Ix2|<tx1 then tx1|Iy1||Ix1Ix2|<tx1 as Ix1,Ix2Iy1. Similar contradiction will arise if |Ix1Ix2|<tx2. Hence, the claim is justified.

Claim 2: There always exist three vertices from each partite set whose corresponding intervals are not contained in each other. (we only need m,n7)

proof:

Note that V1,V2 are posets with respect to the interval containment relation. Hence, if our claim is not true, as m,n7 each poset must contain a totally ordered subset having cardinality at least 3. Let the totally ordered subset of V1 be {Ix1,Ix2,Ix3} satisfying Ix3Ix2Ix1.

We will show there can not be two intervals from V2 that intersect Ix1 at the same end (left or right). On the contrary, let Iy1 and Iy2 intersect Ix1 in its left end and by1by2. Now as Ix2 contained in Ix1, |Iy1Iy2||Iy1Ix2|ty1. Again ty2|Iy2Ix2||Ix2|=|Ix1Ix2|<tx1Footnote2 |Iy1Ix1||Iy1Iy2|. Hence it follows that |Iy1Iy2|ty1,ty2 which is a contradiction as y1,y2 are nonadjacent.

Similarly, one can show that there can be at most one interval from V2 that intersects Ix2 at its same end. Now, as n7 there must exist atleast n2 intervals corresponding to vertices of V2 properly contained in Ix2 which is not possible due to Claim 1. So the claim is true.

Since |V1|13(7) we can always choose three vertices A={x1,x2,x3} from V1 satisfying the conditions of Claim 2. Similarly, we choose vertices B={x4,x5,x6} from V1A and C={x7,x8,x9} from (V1A)B by observing |V1A|10 and |(V1A)B|7 respectively. Now as each Ixi for 1i9 can be contained in at most one Iyj from Claim 1, we can discard at most nine such yj’s from V2 where 1jn. Again, as n13, if we discard these yj’s from V2 we can choose two vertices y10,y11 (say) for which Iy10Iy11 and Iy11Iy10. Otherwise, V2 becomes a totally ordered set having at least four yj’s in it as n13. Now as m13, following proof of Claim 2 one can find a contradiction. Clearly IxiIy10,Iy11 for 1i9. Again, as Iy10 and Iy11 can be contained in at most one interval of V1 from Claim 1, between A, B, C we can always get one set C (say) for which Iy10,Iy11 are not contained in any of the intervals of that set. Now it is easy to see {y10,y11}C form K2,3 for which no interval is contained in another. Therefore, they form an induced proper max-tolerance graph. But it can not be true from Corollary 2.3. ▪

Lemma 2.6.

Km,n is max-point-tolerance graph for all positive integers m, n.

Proof.

Let V1={x1,,xm},V2={y1,,yn} be two partite sets of Km,n. We assign intervals Ixi=[i,m+n+iϵm] and points pi=i for 1im for vertices of V1 and Iyj=[jϵn,j+m] and points pj=j+m for 1jn for vertices of V2 where 0<ϵ<1. It is easy to check that the above assignment actually gives a max-point-tolerance representation of Km,n. ▪

3 Proper max-point-tolerance graphs (proper MPTG)

Proposition 3.1.

If a graph G has a proper MPTG representation S, then G has a proper MPTG representation T in which all the points associated with each interval are distinct.

Proof.

Let G=(V,E) be a proper MPTG and S={(Iv,pv)|vV} be a proper MPTG representation of G where Iv=[av,bv] be an interval, and pv is a point within it. We arrange the intervals according to the increasing order of pv’s, p1 through pn. We iterate the following rule at each shared point, moving from left to right until all pv-values become distinct. Suppose pi is the smallest pv-value shared by multiple points in S. Then pi=pi+1==pm. It is easy to see that there can be atmost two intervals having their endpoints as pi as the intervals are proper we can consider the interval endpoints to be distinct. If they are two in number, then one of them must be the left endpoint (say ak1), and another must be the right endpoint (say bk2). Take a positive value of ϵ smaller than min{d(pm,pm+1),d(pm,ak),d(pm,bk)} where pm+1, ak, bk are the associated points, left endpoint and right endpoint, respectively, occurred just after pm along the real line. To obtain a new proper MPTG representation T of G we assign ak1=pi, pl=pi+ϵml+2 for all ilm, bk2=pi+ϵ. It is easy to check now that the adjacencies remain intact and the intervals remain proper in T. ▪

Proposition 3.2.

Complete graphs Kn for any positive integer n are proper MPTG.

Proof.

We give a proper MPTG representation of Kn by assigning intervals Ivk=[ak,bk] and points pk for each vertex vkV where 1kn. Iv1=[1+1n,n+1], p1=2, Iv2=[1+2n,n+2], p2=2.5 and Ivi=[1+in,n+i], pi=i for 3in. ▪

Proposition 3.3.

Complete bipartite graphs Km,n, for any positive integer m and n are proper MPTG.

Proof.

Follows from the same MPTG representation of Km,n given in Lemma 2.6. ▪

Proposition 3.4.

Caterpillars are proper MPTG.

Proof.

Let G=(V,E) be a caterpillarFootnote3 with spine vertices s1,s2,,sk and additional vertices a1,a2,,al. Let aki1+1,,aki be the vertices adjacent to the spine vertex si where 1ik, k0=0 and 1kil. We set intervals Isi=[2i3,2i+2] and points psi=2i for spine vertices where 1ik and intervals Iaj=[2i3nj+1n+1,2i+2j2n+1], points paj=2i+2j12n+1 for additional vertices adjacent to the spine vertex si where n=kiki1= number of vertices adjacent to si and 1jn. ▪

Let NG(x) (NG(x)) be the set of all vertices that are adjacent (nonadjacent) to a vertex x of G. They are called neighbors (non neighbors) of x in G. The subgraph induced by neighbors of x in G is called neighborhood of x in G. An independent set of three vertices where each pair is joined by a path that avoids the neighborhood of the third is called an asteroidal triple. A graph is said to be asteroidal triple-free (AT-free) if it does not contain any asteroidal triple. For vertices u, v of a graph G, let D(u,v) denote the set of vertices that intercepts (i.e., either the vertex is in the path or adjacent to some vertex of the path) all u, v paths. For vertices u, v, x of G, u and v are said to be unrelated to x if uD(v,x) and vD(u,x) [Citation4].

Theorem 3.5.

[Citation4] A graph G is asteroidal triple-free if and only if, for every vertex x of G, no component F of NG(x) contains unrelated vertices.

Lekkerkerker and Boland exhibited the importance of asteroidal triples in [Citation12] by proving interval graphs to be chordal and AT-free. It was proved in [Citation3] that interval graphs form a strict subclass of MPTG. But there are certain MPTG graphs that are not AT-free (see ). But interestingly, in the following lemma, we find the class of proper MPTG graphs to be asteroidal triple-free.

Figure 2: MPTG G1 having asteroidal triple {v1,v2,v3}.

Figure 2: MPTG G1 having asteroidal triple {v1,v2,v3}.

Lemma 3.6.

If G is a proper MPTG then it is asteroidal triple-free.

Proof.

Let G=(V,E) be a proper MPTG with representation {(Iv,pv)|vV} where Iv=[av,bv] be an interval and pv be a point within it. On contrary to Theorem 3.5 let’s assume there exists a vertex xV and a component F of NG(x) containing a pair of unrelated vertices u, v (say). Let P1,P2 be two arbitrary paths of D(x,v) and D(x,u) respectively such that u(v) is not adjacent to any vertex of P1(P2) where P1=(x,v1,,vn=v),P2=(x,u1,,um=u).

Let pu,pv occur on the same side of px on the real line (say pu<pv<px). Now, if pu1<pv then pu1<pv<px clearly. Now as u1,x are adjacent (3.1) axpu1<pv<pxbu1(3.1)

which imply pvIu1Ix. But as v is not adjacent to u1,x, pu1<av<bv<px imply IvIu1,Ix from (3.1), which is a contradiction as the intervals are proper. Again, if pum1>pv one can similarly find contradiction as um1uE and vu,vum1E. Hence pum1(pu1) must occur left (right) to pv, i.e., pum1<pv<pu1. Now, replacing u by um1 and x by u1 one can similarly find pum2<pv<pu2. Hence, using the above technique for a finite number of times (as the length of P2 is finite), one can get pumk<pv<puk for some 1<k<m such that uk,umk are adjacent in P2. But this again introduces a similar contradiction.

Now consider the case when pu,pv occur opposite to px on the real line (say pu<px<pv). As F is a connected component of NG(x) there must exist a path P=(u,,v) between u, v in F. It is easy to observe that px must fall within an interval Ik (say) for some vertex k of the path P. Again as k is not adjacent to x either pk<ax or pk>bx. Now if pk<ax then there must exist some vertices k,k+1 in P such that ak+1pk<ax<bx<pk+1bk where kk<v in P, which imply IxIk,Ik+1, which is not true as the intervals are proper. One can find a similar contradiction when pk>bx. Hence we are done by Theorem 3.5. ▪

It is easy to observe following the above lemma that all trees are not proper MPTG although they are all MPTG [Citation15]. Note that, it is not necessary that all AT-free graphs have to be perfect (C5 is AT-free). However, in Theorem 3.10 we prove proper MPTG graphs to be perfect.

Lemma 3.7.

Let Cn be a cycle of length n. Then Cn,n5 does not belong to the class of proper MPTG.

Proof.

Since proper MPTG graphs are AT-free following Lemma 3.6, Cn,n6 does not belong to the class of proper MPTG as {v1,v3,vn1} form an asteroidal triple.

On the contrary, let C5=(V,E) has a proper MPTG representation {(Iv,pv)|vV} where Iv=[av,bv] be an interval and pv be a point within it. Let {vi|1i5} be the vertices that occurred in a circularly consecutive way in clockwise (anticlockwise) order in C5. Now if p1 occurs between p2,p5 on the real line (say p2<p1<p5) then as v2v5E either p2<a5 or p5>b2.

If p2<a5 then a1<a5 (as v1v2E) and hence b1<b5 as the intervals are proper. Thus we get I1I5=[a5,b1]. Now as v4v5E, p4I5. Hence p2<a5p4. If p4 occurs left to p1 then p4[a5,p1]. But as v1v4E, b4<p1 which imply b4<p5 which contradicts v4v5E. Hence p4>p1. Now, as v3 is adjacent to both v2,v4, [p2,p4]I3. Therefore we get p1I3 as p2<p1<p4. But as v1v3E, either p3>b1 or p3<a1.

If p3>b1 then from the previous relations we get (3.2) a3p2<a5<p5b1<p3b2.(3.2)

Hence p5I3 from (3.2). But as v3v5E, b5<p3, which implies I5I3 from (3.2), which is a contradiction as the intervals are proper. Again, if p3<a1 then a4p3<a1 and therefore b4<b1. Combining, we get (3.3) a4<a1p2<a5p1<p5b4<b1.(3.3)

Hence p1I4 from (3.3). But as v1v4E, p4>b1. Thus we get I1I4 from (3.3), which is not true as the intervals are proper. One can similarly find contradiction when p5>b2.

Now if p2,p5 occurs on the same side of p1 (say p2<p5<p1) then p5[p2,p1]I1,I2. But as v2v5E, a5>p2 and hence a5>a1 and therefore b5>b1 as the intervals are proper.Hence (3.4) I1I5=[a5,b1].(3.4)

Now v4v5E imply p4I5. Now if p4<b1 then p4I1I5. But as v1v4E, b4<p1<b1 which imply a4<a1. Hence, by combining, we get a4<a1p2<a5p5,p4b4<p1b2. Thus p2I4 and p4I2, which is not true as v2v4E.

Next, if p4>b1, then b4p4>b1 imply a4>a1. Hence, we get (3.5) a1p2<a5p5<p1b1<p4b4,b5.(3.5)

Now, as v3 is adjacent to v2,v4, [p2,p4]I3. This imply p5,p1I3 from (3.5). But as v3v5,v1v5E, p3I1,I5 and therefore does not belong to I1I5. Hence p3<a5 or p3>b1 from (3.4).

Now if p3<a5 then p3<a1 from (3.5) as p3I1. Hence, combining, we get a4p3<a1<b1<p4b4. This implies I1I4 which is a contradiction. Again, if p3>b1 then p3>b5 as p3I5 from (3.5). Now as a3p2 from (3.5) we get I5I3 which is again not true as the intervals are proper. ▪

We note that the 4-wheel graphFootnote4 W4=K4 is a proper MPTG from Proposition 3.2. Again the 5-wheel graph W5 is also a proper MPTG with representation ([30,130],80) for the center vertex u and ([20,120],50),([10,100],70), ([60,150],90), ([40,140],110) for other vertices.

Corollary 3.8.

The n-wheel graphs Wn for n6 are not proper MPTG.

Proof.

If Wn,n6 be a proper MPTG then Wn{u} would also be a proper MPTG where u is the central vertex of Wn. But this is not true from Lemma 3.7 as Wn{u}=Cn1. ▪

Lemma 3.9.

Cn¯,n7 does not belong to the class of MPTG.

Proof.

One can note that {v1,v4,vn,v3} form an induced 4-cycle (C say) in Cn¯,n7. From (2.1) one can find that in any MPTG ordering of vertices of C neither {v1,vn} nor {v4,v3} can sit together within the other.

Now if we consider v1v3vnv4 in some MPTG ordering of C then it is easy to verify from (2.1) that when n8 neither vn3 nor vn2 can occur prior to v1 or after v4 as vn3,vn2 are adjacent to v3,vn respectively. Hence they have to sit within v1,v4. Now one can find contradiction from (2.1) applying it to the vertex set {v1,vn3,vn2,v4} as vn3,vn2 are nonadjacent. Similar contradictions will arise for other MPTG orderings of C. Again, for n=7 repeatedly applying (2.1) one can show there is no MPTG ordering of C7¯. ▪

From the aforementioned lemma, it is obvious that proper MPTG graphs can not contain Cn¯,n7 as an induced subgraph. Combining this fact with Lemma 3.7 one can conclude the following:

Theorem 3.10.

The class of proper MPTG graphs are perfect.

Definition 3.11.

Let G=(V,E) be an undirected graph with a linear ordering (say ``<) on its vertex set V. We call such an ordering as proper MPTG ordering of V if it satisfies the following conditions.

  1. For any v1,v2,v3V, if v1<v2<v3 and v1v3E, then v1v2E or v2v3E. (3-point condition)

  2. For any v1,v2,v3,v4V, if v1<v2<v3<v4 and v1v3,v2v4E, then v2v3E. (4-point   condition)

  3. For any v1,v2,v3,v4,v5V, if v1<v2<v3<v4<v5 and v1v4,v2v5E, then v1v3E or v3v5E when v1v2 or v4v5E. (5-point condition-1)

  4. For any v1,v2,v3,v4,v5V, if v1<v2<v3<v4<v5 and v1v3,v3v5E, then v1v2E or v2v4E or v4v5E. (5-point condition-2)

  5. For any v1,v2,vj,vk,v5,v6V, if v1<v2<vj,vk<v5<v6 and v1vj,vjv5,v2vk,vkv6E for some j, k such that 2<j,k<5, then v1v2E or v2v5E or v5v6E. (6-point condition)

One can observe that any proper MPTG ordering is a MPTG ordering as it satisfies condition 2, which is the same as (2.1). Therefore, any graph satisfying a proper MPTG ordering must be a MPTG by Theorem 2.1. We now prove that proper MPTG graphs are characterized as graphs with proper MPTG orderings. In the following lemma, we present a necessary condition for proper MPTG graphs.

Lemma 3.12.

Let G=(V,E) be a proper MPTG. Then G has a proper MPTG ordering of V.

Proof.

Let G be a proper MPTG with representation {(Ii,pi)|iV} where Ii=[ai,bi] be an interval on the real line, and pi is a point within it. We arrange the vertices of V according to the increasing order of pi’s. Let’s denote the ordering of V by “<”. We will show that with respect to this ordering all the conditions of Definition 3.11 hold true.

proof of condition (1): Let v1<v2<v3 for some v1,v2,v3V and v1v3E. Since v1v3E, a3p1<p2<p3b1 which imply p2[p1,p3]I1,I3. Now if v1v2,v2v3E, p1,p3I2 which imply p1<a2 and b2<p3 as p1<p2<p3. Hence a3p1<a2<b2<p3 which imply [a2,b2][a3,p3]I3 which contradicts that the intervals are proper. Hence the result follows. proof of condition (2): Since G is a MPTG, this verification follows from the proof of Theorem 5.4 of [Citation3]. proof of condition (3): Let v1<v2<v3<v4<v5 for some v1,v2,v3,v4,v5V and v1v4,v2v5E. Since v1v4E, a4p1<p2<p3<p4b1, which imply p2,p3I1,I4. Again as v2v5E, a5p2<p3<p4<p5<b2 which imply p3,p4I2,I5. Now if v1v3,v3v5E, p1,p5I3 which imply p1<a3 and b3<p5 as p1<p3<p5. If v1v2E then a2p1<a3<b3<p5<b2 which imply [a3,b3][a2,b2] which is a contradiction as the intervals are proper. Now if v4v5E then a4p1<a3<b3<p5b4 imply [a3,b3][a4,b4] which is again a contradiction as the intervals are proper. Thus, the condition holds true.

proof of condition (4): Let v1<v2<v3<v4<v5 for some v1,v2,v3,v4,v5V and v1v3,v3v5E. Since v1v3E imply a3p1<p2<p3b1. Therefore p2I1I3. Now if v1v2E, then a2>p1 as p1<p2. Now as v3v5E, a5p3<p4<p5b3. This imply p4I3I5. Now if v4v5E, then b4<p5 as p4<p5. Now as p2,p4I3 if v2v4E imply either b2<p4 or p2<a4 as p2<p3<p4. If b2<p4 then a3p1<a2<b2<p4<p5b3 which imply [a2,b2][a3,b3] which is a contradiction as the intervals are proper. Again, if p2<a4 then a3p1<p2<a4<b4<p5b3 which imply [a4,b4][a3,b3] which is again a contradiction as the intervals are proper. Hence, the proof holds.

proof of condition (5): Let v1<v2<vj,vk<v5<v6 for some v1,v2,v3,v4,v5,v6V and v1vj,vjv5,v2vk,vkv6E for some j, k such that 2<j,k<5. Since v1vjE, ajp1<p2<pjb1, which imply p2I1,Ij. Now if v1v2E, then a2>p1 as p1<p2. Again as vkv6E, a6pk<p5<p6bk, which imply p5Ik,I6. Now if v5v6E, then b5<p6 as p5<p6. If v2v5E then either p2<a5 or b2<p5 as p2<p5. Now if p2<a5 then akp2<a5pj<p5<b5<p6bk as v2vk,vjv5E, which imply [a5,b5][ak,bk] which is a contradiction as the intervals are proper. Again, if b2<p5 then ajp1<a2<b2<p5bj as vjv5E, which imply [a2,b2][aj,bj] which again contradicts that the intervals are proper. Hence the result follows.

Therefore, < becomes a proper MPTG ordering of V. ▪

Definition 3.13.

Let G=(V,E) be an undirected graph possessing a proper MPTG ordering (say σ) of V. From Theorem 2.1 it is clear that G is a MPTG with respect to σ. Let {(Ii,pi)|iV} be a MPTG representation of G. Below we define an order relation σ among the right endpoints of each interval Ii=[ai,bi] associated to every vertex of G when vertices are arranged in σ order in A(G)=(ai,j) along both rows and columns of it. For convenience, we shall use the symbol instead of σ.

Let R1,R2 be two partial order relations defined on the set {bi|iV}. For i<j in A(G) we define,

bjR1bi if there exists some k2>j such that aj,k2=0Footnote5 and ai,k2=1 (3.6)

bjR2bi if there exists some k1<i such that ak1,i=0Footnote6 and ak1,j=1. (3.7)

Now we define an order “” between bi’s in the following way. For i<j, (3.8) bjbi if bjR1bi or bjR2bi or bjR1bk and bkR2bi for some kV satisfying i<k<j.(3.8) (3.9) bibj otherwise.(3.9)

We call the ordering “” as proper MPTG endpoint ordering corresponding to the proper MPTG ordering σ of V. It is important to note for i<j, bibj whenever ai,j=0. If not let bjbi, then from the above definition of , one of the following holds, bjR1bi, bjR2bi or bjR1bk and bkR2bi for some kV such that i<k<j. In the first two cases, condition 1 of Definition 3.11 and in the last case, condition 4 of Definition 3.11 gets contradicted.

In the following lemma, we will prove to be a total orderFootnote7.

Lemma 3.14.

Let G=(V,E) be an undirected graph that admits a proper MPTG ordering σ of V, then the associated proper MPTG endpoint ordering σ is a total order.

Proof.

Since σ is a proper MPTG ordering of V, the associated proper MPTG end point ordering σ= is obtained by defining a certain linear order between all the bi’s looking at the adjacency of the vertices in A(G)=(ai,j) (see Definition 3.13) when the rows and columns of it are arranged according to σ and {(Ii,pi)|iV} is the MPTG representation of G where Ii=[ai,bi] and piIi. Now to show is a total order, it is sufficient to prove “” is transitive. Let bi1bi2 and bi2bi3. Below we will show bi1bi3. Several cases may arise depending on the occurrence of i1,i2,i3 in σ.

Case (i) i1<i3<i2

We assume on contrary bi3bi1. We use the following observations repeatedly for the rest of the proof.

Observations:

  1. First we note ai3,i2=1 as i3<i2.

  2. Next bi3R1bj for any j<i3 imply existence of some k2>i3 such that ai3,k2=0,aj,k2=1. Moreover, k2>i2 otherwise (2.1) gets contradicted for vertices {j,i3,k2,i2}.

  3. Further, we see that bi2R1br for any r<i2 imply the existence of some k2>i2 such that ai2,k2=0,ar,k2=1. Note that if bi3R1bj for any j<i3 then if k2>k2, ai2,k2 becomes zero as ai2,k2=0 is right open, which helps us to conclude bi2R1bj from (3.6).

  4. Similarly, blR2bi1 for any l>i1 ensure the existence of some k1<i1 such that ak1,i1=0,ak1,l=1.

  5. Again bmR2bi3 for any m>i3 ensure there is some k1<i3 satisfying ak1,i3=0,ak1,m=1. We note that k1<i1 otherwise as ak1,i3=0 is up open, ai1,i3 becomes zero, which imply bi1bi3 which is a contradiction. Now if blR2bi1 holds for any l>i1 then if k1<k1<i1, ak1,i1 becomes zero as ak1,i1=0 is up open, which implies bmR2bi1 from (3.7).

Main proof: Let bi2R1bi3. Now if bi3R1bj for any j<i3 then we get k2 must be greater than k2 otherwise (2.1) gets contradicted for vertices {j,i3,k2,k2}. Hence, from observation 3 we get bi2R1bj.

Now if bi3R1bi1 we get bi2bi1 from last paragraph. Again if bi3R2bi1 then bi2bi1 follows from (3.8) as bi2R1bi3 and i1<i3<i2. Next, if bi3R1bk and bkR2bi1 for some kV such that i1<k<i3, we get bi2R1bk from last paragraph. Hence from (3.8) we get bi2R1bi1 as bkR2bi1 and i1<k<i2. Thus, all the above cases contradict our assumption.

Let bi2R2bi3. If bi3R1bi1 then k2>i2 from observation 2. Hence ai2,k2=1 otherwise bi2R1bi1 from (3.6) imply bi2bi1 which is not true. Hence condition 3 of Definition 3.11 gets contradicted for vertices {k1,i1,i3,i2,k2}. Again, if bi3R2bi1 then from observation 5 if k1<k1<i1 we get bi2bi1 which is not true. Again, k1k1 clearly. Hence k1<k1. But again (2.1) gets contradicted for vertices {k1,k1,i3,i2}. Now if bi3R1bk and bkR2bi1 for some kV such that i1<k<i3. Then ai2,k2=1 otherwise, we get bi2R1bk which imply bi2bi1 as bkR2bi1 and i1<k<i2. But we get a contradiction applying condition 3 of Definition 3.11 on vertices {k1,k,i3,i2,k2}.

Let bi2R1bk and bkR2bi3 for some kV such that i3<k<i2. Now if bi3R1bj for any j<i3 then if k2>k2, from observation 3 we get bi2R1bj.

Now if bi3R1bi1 then if k2>k2 from last paragraph we get bi2bi1 which is not true. Hence i2<k2<k2. Again ak,k2=1 from (2.1) applying to the vertices {i1,k,k2,k2}. Hence condition 3 of Definition 3.11 gets contradicted for vertices {k1,i1,i3,k,k2}. Next if bi3R2bi1 then if k1<k1<i1 then from observation 5 we get bkR2bi1 and hence from (3.8) we find bi2bi1 as bi2R1bk and i1<k<i2. Again when k1<k1, (2.1) gets contradicted for vertices {k1,k1,i3,k}. Now let bi3R1bk and bkR2bi1 for some kV satisfying i1<k<i3. As bi3R1bk then if k2>k2 then from the last paragraph we get bi2R1bk which helps us to conclude from (3.8) bi2bi1 as bkR2bi1. Hence i2<k2<k2. Thus, in this case, condition 3 of Definition 3.11 gets contradicted for vertices {k1,k,i3,k,k2} (note ak,k2=1 otherwise (2.1) gets contradicted for vertices {k,k,k2,k2}).

The verification of other cases is also long and rigorous. Therefore, we put them in Appendix. ▪

Definition 3.15.

Let G=(V,E) be a MPTG having representation {([ai,bi],pi)|iV} satisfying a proper MPTG ordering σ of V. Then from Definition 3.13 and Lemma 3.14 it follows that by observing the adjacencies of the vertices in A(G) one can find a total order between all bi’s in such a way that they can be put together in a single sequence P1 (say). Next, we arrange ai’s in a single sequence P2 (say) in the same order. Let |V|=n. Then

P1:bα1bα2bαn and P2:aα1aα2aαn where 1αin.

We now combine ai,bi,pi’s in a single sequence P by following the rule. Henceforth, we call P by canonical sequence corresponding to σ.

  1. First, we place all pi’s on the real line according to the occurrence of i’s in σ.

  2. Now, using induction, starting from the last element of P2 until the first element is reached, we place ai’s on the real line in the following way.

We first place aαn between p(αn)11 and p(αn)1. Now for aαk if (αk)1(αk+1)1 then we place aαk just before aαk+1, place aαk between p(αk)11 and p(αk)1 otherwise, where (αi)1 is the first column containing 1 in the αith row in A(G).

  1. Again using induction, starting from the first element of P1 until the last element is reached, we place bi’s on the real line by following the rule.

First we place bα1 between p(α1)2 and p(α1)2+1. Now for bαk if (αk)2(αk1)2 then we place bαk just after bαk1, place bαk between p(αk)2 and p(αk)2+1 otherwise where (αi)2 is the last column containing 1 in the αith row in A(G).

From the construction of P one can verify that if more than one bi(ai)’s occurs between two pi’s then they are arranged according to their occurrence in P1(P2) respectively. Again, it is easy to check that the order of the sequences P1,P2 get preservedin P.

Example 3.16.

Consider the MPTG G=(V,E) in whose augmented adjacency matrix A(G) is arranged according to a proper MPTG ordering σ=(v1<v2<<v7) of V. One can note that all the conditions of Definition 3.11 are satisfied by σ. A MPTG representation of G is provided by assigning an interval Ii=[ai,bi] and a point pi within Ii corresponding to each vertex vi where 1i7. Now looking at the adjacencies of the vertices in A(G) we construct P1,P2 sequences of endpoints bi,ai respectively, according to the construction described in Definition 3.15.

Figure 3: Augmented adjacency matrix A(G) of a proper MPTG G.

Figure 3: Augmented adjacency matrix A∗(G) of a proper MPTG G.

P1:b2b1b5b6b4b3b7.

P2:a2a1a5a6a4a3a7.

The combined sequence of ai,pi,bi is P:a2a1a5a6a4 p1a3p2p3b2a7p4b1p5b5p6b6p7b4b3b7.

We now present the main theorem of our article, which characterizes a proper max-point-tolerance graph.

3.1 Characterization of proper MPTG

Theorem 3.17.

Let G=(V,E) be an undirected graph. Then G is a proper MPTG if and only if it satisfies a proper MPTG ordering of V.

Proof.

Let G=(V,E) be a proper MPTG. Then it satisfies a proper MPTG ordering of V following Lemma 3.12.

Conversely, let G possess a proper MPTG ordering (say σ) of V. Then it satisfies all the conditions of Definition 3.11 with respect to σ. Let the augmented adjacency matrix of G, i.e., A(G)=(aij) is arranged according to σ. Then, from Theorem 2.1 G becomes a MPTG with respect to σ. Let [ai,bi] be the interval and pi be a point within it associated with each vertex iV in its MPTG representation. Now, from the adjacencies of the vertices in A(G) we define a proper MPTG endpoint ordering σ= among bi’s as described in Definition 3.13. Moreover, in Lemma 3.14 we prove as a total order. Next, we construct sequences P1(P2) consists of bi(ai) as described in Definition 3.15. Then we construct the canonical sequence P corresponding to σ, which is basically a combined sequence containing all ai,bi,pi’s. Now we associate the numbers 1 to 3n (|V|=n) to P starting from the first element of it, and define Ii=[ai¯,bi¯] and pi=pi¯ where ai¯,bi¯,pi¯ denote the numbers associated to ai,bi,pi in P. Below we prove {(Ii,pi)|iV} will actually give a proper MPTG representation of G.

Step 1. First we verify that the intervals Ii are proper.

First, we show ai<bi for each i in P. From the construction of P one can easily check that each ai is placed left to p(i)1 and each bi is placed right to p(i)2 where (i)1,(i)2 denote columns containing the first and last 1 in i th row of A(G). Now as p(i)1<pi<p(i)2 in P, ai<bi.

Since bi,ai’s are arranged in P1,P2 in the same order, and they individually keep their orderings intact in P, no interval can be contained in another. Hence the intervals {Ii=[ai¯,bi¯]|iV} are proper with respect to our given representation.

Step 2. Note that ai<pi<bi in P as (i)1i(i)2 from Definition 3.15.

Step 3. Next, we verify that ([ai¯,bi¯],pi¯) is the proper MPTG representation of G.

For this, it is sufficient to show that, with respect to the above representation, A(G) satisfies all of its adjacency relations.

Case (i): We show that ai,j=1 imply pi,pjIiIj.

Let for i<j, ai,j=1. If bibj in P1 (i.e., aiaj in P2) then as (j)1i, aj<pi in P. Again as (i)2>j, pj<bi in P. Further, if bjbi in P1 (i.e., ajai in P2) then as (j)2j, bj>pj in P. Again, (i)1i imply ai<pi. Now as the order of the elements of P1,P2 remains intact in P we get pi,pjIiIj.

Case (ii): Next, we show that ai,j=0 imply either piIj or pjIi.

The proof will follow from the following two claims.

Claim 1: For i<j when ai,j=0 is only up openFootnote8, pi<aj in P .

Let for i<j, ai,j=0 is only up open. Then there exists some k>j such that ai,k=1. From condition 1 of Definition 3.11 we get aj,k=1. Again, bibj as i<j in P1. Now as (i)2k>j, bi>pk>pj. Again, we note (j)1>i. Hence, to show aj>pi in P it is sufficient to prove that if there exists some lV such that ajal in P2 then (l)1>i.

On the contrary, let’s assume ajal for some l such that (l)1i. Several cases may arise depending on the occurrence of l in A(G). Note that l can not be equal to i as alaj in P2 then contradicts our assumption. Again lk otherwise blR2bj from (3.7) imply blbj in P1 and hence we get alaj in P2 which introduces contradiction. Next if l<i then al,j becomes zero as ai,j=0 is up open, which imply blbj in P1 and thus we get alaj in P2 which again contradicts ourassumption.

Now we consider the case when i<l<j. We need to use the following observations to find contradictions in this case.

Observations:

  1. Since (l)1i, ai,l=1 from (2.1) applying to the vertices {(l)1,i,l,k}.

  2. If bjR1br for any r<j then there exists some k2>j such that aj,k2=0,ar,k2=1. We note that k2>k as otherwise (2.1) gets contradicted for vertices {r,j,k2,k}.

  3. If bmR2bl for any m>l then there exists some k1<l such that ak1,l=0,ak1,m=1. If (l)1<k1<l then {(l)1,k1,l,m} contradict (2.1). Hence k1<(l)1i.

Main proof: Now if bjR1bl then from observation 2 we get k2>k. One can verify now that condition 3 of Definition 3.11 get contradicted for vertices {i,l,j,k,k2} using observation 1. Again, if bjR2bl then k1<(l)1i from observation 3. But in this case also (2.1) gets contradicted for vertices {k1,i,j,k}. Now if bjR1bk and bkR2bl for some kV such that l<k<j then k2 must be greater than k and k1<(l)1i from observations 2, 3. This implies ai,k=1 applying (2.1) on vertex set {k1,i,k,k} But in this case condition 3 of Definition 3.11 gets contradicted for vertices {i,k,j,k,k2}.

Now if l>j then if (l)1<i, a(l)1,j=0 as ai,j=0 is up open. This imply blR2bj from (3.7) i.e., blbj. Thus we get alaj, which is a contradiction. Again, for (l)1=i we get a similar contradiction.

Hence, for any l, ajal in P2 imply (l)1>i. Thus, from our above claim pi<aj in P is established, and therefore piIj.

Claim 2: For i<j when ai,j=0 is right open, then either pi<aj or bi<pj in P.

Let for i<j, ai,j=0 is right open. Then bibj in P1. As (i)2<j it is sufficient to prove that either for all rV satisfying brbi in P1, (r)2<j which imply bi<pj or for all lV such that ajal in P2, (l)1>i which imply pi<aj.

We assume, on the contrary, the existence of some r satisfying brbi and (r)2j. Note that r can not be greater than j as in that case ai,r becomes zero due to the fact ai,j=0 is right open, which implies bibr which contradicts our assumption. Again, when r<i, as (r)2j and ai,j=0 is right open, ai,(r)2 becomes zero, which implies biR1br from (3.6) and hence we get bibr which contradicts our assumption. A similar contradiction will occur when r=j. Hence i<r<j.

Case (i): If brR1bi, then there exists k2>r such that ar,k2=0,ai,k2=1. Note that k2 can not occur right to j in A(G) as ai,j=0 is right open. Hence r<k2<j. Now one can find ar,j=0 as ar,k2=0 is right open. Hence (r)2>j, which again contradicts (2.1) for vertices {i,r,k2,(r)2}.

Thus, for any rV if brR1bi holds, then we get (r)2<j. Hence, one can conclude bi<pj as (i)2<j in this case.

Case (ii): If brR2bi then there exist some k1<i such that ak1,i=0,ak1,r=1. From condition 1 of Definition 3.11, ai,r=1. We will show in this case pi<aj. For this we assume on contrary ajal for some lV such that (l)1i. Using Lemma 3.14 we get (3.10) brbibjbl(3.10)

In the following paragraph, we list some important observations, which we vastly use in the rest of the proof.

Observations:

  1. For l<j, al,j=1, for l<r, al,r=1 and for i<r, ai,r=1 otherwise (3.10) gets contradicted.

  2. For i<l, k1<(l)1i. If (l)1<k1 then a(l)1,i becomes zero as ak1,i=0 is up open, which implies blR2bi from (3.7), i.e., blbi which is not true from (3.10). Hence k1<(l)1i.

  3. Next we show ar,j=1. We assume on contrary that ar,j=0 then (r)2>j.

When l<r following observation 1, one can see that (2.1) gets contradicted for vertices {l,r,j,(r)2}. Next, when l>j, a(l)1,j=1 otherwise blR2bj (i.e., blbj) follows from (3.7) and the fact k1<(l)1i (observation 2) which contradicts (3.10). Hence, applying (2.1) on vertices {(l)1,r,j,(r)2} one can get a contradiction.

Again, when r<l<j the following cases lead us to contradiction.

If bjR1bl holds then aj,k2=0,al,k2=1 for some k2>j. Now if (r)2>k2 then aj,(r)2 becomes zero as aj,k2 is right open and hence from (3.6) we get bjR1br, i.e., bjbr which contradicts (3.10). Again if j<(r)2<k2, then ar,l=1 follows from (2.1) applying to the vertices {(l)1,r,l,(r)2} (observation 2). Hence, from observation 6, one can see that condition 3 of Definition 3.11 gets contradicted for {r,l,j,(r)2,k2}.

If bjR2bl then there exist k1<l such that ak1,l=0,ak1,j=1. Hence (2.1) gets contradicted for {k1,r,j,(r)2}.

Now if bjR1bk and bkR2bl for some kV such that l<k<j. Then there exist some k2>j,k1<l satisfying aj,k2=1,ak,k2=1,ak1,l=0,ak1,k=1. Similar contradiction will arise when (r)2>k2 as described in the last paragraph. Now when j<(r)2<k2, ar,k=1 applying (2.1) on vertices {k1,r,k,(r)2}. Hence, applying condition 3 of Definition 3.11 on vertex set {r,k,j,(r)2,k2} one can find contradiction.

  1. For i<l<j, ai,l=1. When i<l<r, using observations 1, 2 and applying (2.1) on vertices {(l)1,i,l,r} we get ai,l=1. Next if i<r<l, ar,l=1 (see proof of observation 3). Now if we assume on the contrary that ai,l=0 then (l)1<i. But then condition 3 of Definition 3.11 gets contradicted for the vertices {k1,(l)1,i,r,l}.

  2. Again for i<l<j, if bjR1bm and bmR2bl hold for some m such that l<m<j then k1<k1<i and ai,m=1 where ak1,l=0,ak1,m=1 for some k1<l.

If k1<k1 then ak1,i becomes zero as ak1,i=0 is up open which implies bmR2bi and hence we get bmbj from Lemma 3.14 (as bibj) which is not true from above. Again, k1i as ai,l=1 from observation 4. Now when i<k1<l, (2.1) gets contradicted for the vertices {i,k1,l,m}. Hence k1<k1<i.

Now if m<r, using observation 1 and applying (2.1) on the vertices {k1,i,m,r} we get ai,m=1. Again, when m>r, we obtain ar,m=1 by applying observation 3 and (2.1) to the vertices {k1,r,m,j}. Further applying condition 3 of Definition 3.11 on the vertices {k1,k1,i,r,m} one can get ai,m=1.

Main proof: If l<k1, then al,i=0 as ak1,i=0 is up open, which implies blbi as l<i. But this is not true from (3.10). Next if k1<l<i. Now using observations 1, 3 one can verify that condition 3 of Definition 3.11 gets contradicted for vertices {k1,l,i,r,j}. Again, when j<l, as ai,j=0 is right open, ai,l becomes zero. Hence, from observation 2, we get k1<(l)1<i. One can now show that condition 3 of Definition 3.11 gets contradicted for the vertices {k1,(l)1,i,r,l} using the fact ar,l=1 (see proof of observation 3).

Next, we consider the case i<l<j. Now if bjR1bl then there exists k2>j such that aj,k2=0,al,k2=1. From observations 3, 4, we get ar,j=1 and ai,l=1. Now applying condition 5 of Definition 3.11 on the vertex set {k1,i,l,r,j,k2} one can find contradiction.

If bjR2bl then there exists k1<l such that ak1,l=0,ak1,j=1. Proceeding similarly to observation 5 and replacing m by l one can prove k1<k1<i. Again, from observation 3, we get ar,j=1. Hence, by applying condition 3 of Definition 3.11 to the vertices {k1,k1,i,r,j}, one can find a contradiction.

Again if bjR1bk and bkR2bl for some kV such that l<k<j. Then there exist k2>j,k1<l such that aj,k2=0,ak,k2=1,ak1,l=0,ak1,k=1. From observation 5 one can find k1<k1<i. Also note ar,j=1 from observation 3 and ai,k=1 from observation 5. Hence, applying condition 5 of Definition 3.11 on the vertex set {k1,i,k,r,j,k2} one can findcontradiction.

Thus, in this case, we prove that if there exists some rV satisfying brR2bi such that (r)2j then for any lV, ajal in P1 imply (l)1>i. Hence pi<aj in P and therefore piIj.

Case (iii): If brR1bk and bkR2bi for some kV such that i<k<r. Then there exist k2>r,k1<i such that ar,k2=0,ak,k2=1,ak1,i=0,ak1,k=1. Note that k2>j otherwise (2.1) gets contradicted for vertex set {k,r,k2,(r)2} as (r)2j. Hence (k)2k2>j. Now as bkR2bi, i<k<j and (k)2>j, proceeding similarly as Case (ii), just replacing r by k one can find the same result. ▪

Remark 3.18.

By Theorem 3.17 we are able to provide a proper MPTG representation {([av,bv],pv)|vV} of a graph G (see ), which satisfies a proper MPTG ordering σ=(v1<v2<<v7). In Example 3.16 we found the canonical sequence P of the graph G from its augmented adjacency matrix A(G), arranged according to σ along both rows and columns.

P:a2a1a5a6a4p1a3p2p3b2

a7p4b1p5b5p6b6p7b4b3b7.

Therefore we obtain the proper MPTG representation of G from the following table according to the proof of the abovetheorem. a2a1a5a6a4p1a3p2p3b212345678910 a7p4b1p5b5p6b6p7b4b3b7 1112131415161718192021

Catanzaro et. al [Citation3] have shown that the interval graphs form a strict subclass of MPTG. We extend this result to the class of proper MPTG graphs.

Corollary 3.19.

Interval graphs form a strict subclass of proper MPTG.

Proof.

Let G=(V,E) be an interval graph having representation {Iv=[av,bv]|vV}. Now, arranging the intervals according to the increasing order of left endpoints (av) and ordering vertices of V with the same, it is easy to check that G satisfies all the conditions of Definition 3.11. Hence, G becomes a proper MPTG by Theorem 3.17. Also, one can easily verify that C4 is a proper MPTG (see ) but it is not an interval graph. Therefore, the result follows. ▪

Figure 4: C4 with its proper MPTG representation.

Figure 4: C4 with its proper MPTG representation.

In [Citation2] Bogart and West proved that proper interval graphs are the same as unit interval graphs. Here we can also conclude a similar result for the class of proper MPTG graphs.

Proposition 3.20.

Let G=(V,E) be an undirected graph. Then the following are equivalent:

  1. G is a proper MPTG.

  2. G is a unit MPTG.

Proof.

(1)(2) Let G=(V,E) be a proper MPTG with representation {(Iv,pv)|vV} where Iv=[av,bv] is an interval, and pv is a point within Iv. From Theorem 3.17 it follows that G must possess a proper MPTG ordering of V. Hence one can construct the sequences P1,P2,P as described in Definition 3.15. Let P1:bk1bkn and P2:ak1akn. We assign ak1,bk1 with real values α1,α1+l in P. At i th step, we assign aki,bki with real values αi,αi+l where max{αi1,mi}<αi<Mi where mi=max{αj+l|1ji1,bkjakiinP} and Mi=min{αj+l|1ji1,akibkjinP}. After the assignment is done for all aki,bki where 1in, assign values for pi in P in such a way so that P still remains an increasing sequence. Now it is easy to check that ([aki,bki],pi) gives us unit MPTG representation (unit length l) as the order of ai,bi,pi remains intact in P after assigning their values in the above way.

(2)(1) Converse is obvious. ▪

Remark 3.21.

The unit vs. proper question is addressed for many graph classes in reference [Citation9]. Golumbic and Trenk wrote, “For min-tolerance graphs, the classes are different. For max-tolerance graphs, the question is still open” [9, p. 215]. We settle this query below.

Proposition 3.22.

Unit max-tolerance graphs form a strict subclass of proper max-tolerance graphs.

Proof.

An unit max-tolerance graph is a proper max-tolerance graph with the same tolerance representation. Let {v1,,v6} are vertices of C6¯ occurred in circularly consecutive order in C6. Then I1=[2,8],t1=2.9,I2=[4,10],t2=4.5,I3=[1,7.1],t3=1,I4=[5,11],t4=3,I5=[3,9],t5=4.1,I6=[5.2,12],t6=1.5 is a proper max-tolerance representation of C6¯. But from [Citation13] one can verify that C6¯ is not a unit max-tolerance graph. Hence the result follows. ▪

In the following proposition, we will show that proper MPTG and proper max-tolerance graphs are not comparable, although both of them belong to the class of MPTG.

Proposition 3.23.

Proper MPTG and the class of proper max-tolerance graphs are not comparable.

Proof.

C5 is a proper max-tolerance graph with representation I1=[1,6],t1=0.25,I2=[1.2,8],t2=4.7,I3=[3,10],t3=4.8,I4=[5,12],t4=4,I5=[5.5,13],t5=0.35. But C5proper MPTG follows from Proposition 3.7.

Again, we note from Proposition 3.3 that K2,3 is a proper MPTG. But it is not a proper max-tolerance graph, as follows from Corollary 2.3. ▪

4 Conclusion

In this article, we introduce proper max-point-tolerance graphs and characterize this graph class in Theorem 3.17. We have proved in the previous section that interval graphs proper MPTG. Interestingly, one can note that both of these graph classes are AT-free and perfect (Section 3, [Citation12]). Therefore, any proper MPTG that is not an interval graph must contain C4 as an induced subgraph. Also interval graphs central MPTG = unit max-tolerance graph max-tolerance graph according to the reference [Citation13]. Again, Proposition 3.22 helps us to conclude that unit max-tolerance graphs proper max-tolerance graphs.

Next, we find that C6¯ proper MPTG central MPTG. It is a proper MPTG with the representation I1=[20,40],p1=39,I2=[15,38],p2=30,I3=[32,46],p3=33,I4=[25,42],p4=27,I5=[28,44],p5=37,I6=[10,36],p6=34. But it is not a central MPTG, follows from Theorem 3.9 of [Citation13]. Again, Cn,n5 central MPTG proper MPTG follows from [Citation15] and Lemma 3.7 of this article. Hence, these two classes become incomparable. Next, in Proposition 3.23 we see that proper max-tolerance graphs are not comparable to proper MPTG whereas both of these graph classes belong to the class of MPTG.

Further, we find that G1 in is a MPTG. It is also a max-tolerance graph with representation I1=[10,25],t1=5,I2=[45,53],t2=8,I3=[65,75],t3=10,I4=[20,40],t4=5,I5=[30,70],t5=10,I6=[45,60],t6=8,I7=[60,80],t7=10. But it is not a proper MPTG. Again, K2,3 is a max-tolerance graph with representation Ix=[20,0],tx=1,Iy=[0,20],ty=1 for one partite set and Iv1=[2,2],tv1=1,Iv2=[6,6],tv2=5,Iv3=[20,20],tv3=19 for the other partite set. But it is not a proper max-tolerance graph from Corollary 2.3. Lastly, from Theorem 2.4 we establish that MPTG and max-tolerance graphs are not comparable.

Combining these we obtain relations between some subclasses of max-tolerance graphs and MPTG related to proper MPTG in .

Figure 5: Hierarchy between subclasses of max-tolerance graph and MPTG.

Figure 5: Hierarchy between subclasses of max-tolerance graph and MPTG.

Acknowledgments

The author would like to thank Prof. Shamik Ghosh of Jadavpur University, India for his valuable suggestions to build up this paper.

Disclosure statement

No conflict of interest was reported by the author.

Additional information

Funding

This research is supported by UGC (University Grants Commission) NET fellowship (21/12/2014 (ii) EU-V) of the author.

Notes

1 Ix contains Iy means IxIy

2 If x, y are nonadjacent and IyIx then |IxIy|<tx

3 it is a tree with a path (s1,s2,,sk) (known as spine) such that every other vertex is of exactly one distance from the spine.

4 A n-wheel graph Wn is formed by connecting a single vertex to all the vertices of a cycle of order n1

5 This zero is right open from condition 4 of Theorem 2.1

6 This zero is up open from condition 4 of Theorem 2.1

7 is a partial order relation where any two elements are comparable

8 up open but not right open

References

  • Basu, A., Das, S., Ghosh, S., Sen, M. (2013). Circular-arc bigraphs and its subclasses. J. Graph Theory 73(4): 361–376.
  • Bogart, K. P., West, D. B. (1999). A short proof that “proper = unit”. Discrete Math 201: 21–23.
  • Catanzaro, D., Chaplick, S., Felsner, S., Halldórsson, B. V., Halldórsson, M. M., Hixon, T., Stacho, J. (2017). Max point-tolerance graphs. Discrete Appl. Math. 216: 84–97.
  • Corneil, D. G., Olariu, S., Stewart, L. (1997). Asteroidal triple-free graphs. Siam J. Discrete Math. 10: 399–430.
  • Fishburn, P. (1985). Interval Orders and Interval Graphs—A Study of Partially Ordered Sets, New York: Wiley.
  • Fulkerson, D. R., Gross, O. A. (1965). Incidence matrices and interval graphs. Pac. J. Math. 15: 835–855.
  • Golumbic, M. C. (2004). Algorithmic Graph Theory and Perfect Graphs, Vol. 57: Annals of Discrete Mathematics. New York: Elsevier Science.
  • Golumbic, M. C., Monma, C. L. (1982). A generalization of interval graphs with tolerances. Congress. Numer. 35: 321–331.
  • Golumbic, M. C., Trenk, A. (2004). Tolerance Graphs. Cambridge Studies in Advanced Mathematics 89. New York: Cambridge University Press.
  • Ghosh, S., Podder, M., Sen, M. K. (2010). Adjacency matrices of probe interval graphs. Discrete Appl. Math. 158: 2004–2013.
  • Kaufmann, M., Kratochvil, J., Lehmann, K. A., Subramanian, A. R. (2006). Max-tolerance graphs as intersection graphs: Cliques, cycles, and recognition. Paper presented at SODA ’06, January 22–26, Miami, FL.
  • Lekkerkerker, C. G., Boland, J. C. (1962). Representation of a finite graph by a set of intervals on the real line. Fund. Math. 51: 45–64.
  • Paul, S. (2020). On central-max-point tolerance graphs. AKCE Int. J. Graphs Comb. 17(3): 1069–1075.
  • Sen, M., Das, S., Roy, A. B., West, D. B. (1989). Interval digraphs: an analogue of interval graphs. J. Graph Theory 13(2): 189–202.
  • Soto, M., Caro, C. T. (2015). p− Box: A new graph model. J. Discrete Math. Theor. Comput. Sci. 17: 169–186.

Appendix

Here we provide a detailed verification of the other cases that occurred in Lemma 3.14.

Case (ii) i2<i3<i1

As bi1bi2, if bi1R1bi2 then there exists some k2>i1 such that ai1,k2=0,ai2,k2=1. We note that ai3,k2 must be one; otherwise, from (3.6) we get bi3R1bi2, i.e., bi3bi2 which contradicts our assumption. Hence bi1R1bi3 from (3.6), i.e., bi1bi3. Now if bi1R2bi2 then there exists some k1<i2 such that ak1,i2=0,ak1,i1=1. Now if ak1,i3=1 then bi3R2bi2 from (3.7) imply bi3bi2 which is again a contradiction. Hence ak1,i3=0, which imply bi1R2bi3 from (3.7), i.e., bi1bi3. Now if bi1R1bk and bkR2bi2 for some kV such that i2<k<i1. Then there exists k2>i1 and k1<i2 such that ai1,k2=0,ak,k2=1,ak1,i2=0,ak1,k=1. Now when i2<k<i3, if ai3,k2=0 then bi3R1bk imply bi3bi2 from (3.8) as bkR2bi2 which is a contradiction. Hence ai3,k2=1 and from (3.6) it follows bi1R1bi3, i.e., bi1bi3. Now when i3<k<i1, if ak1,i3=1 then bi3R2bi2 from (3.7) imply bi3bi2 which is a contradiction. Hence ak1,i3=0 which imply bkR2bi3 which helps us to conclude bi1bi3 from (3.8) as bi1R1bk.

Case (iii) i1<i2<i3

In this case, we assume on the contrary bi3bi1.

Let bi3R1bi1. Then there exists k2>i3 such that ai3,k2=0,ai1,k2=1. Now if ai2,k2=1 then bi3R1bi2 from (3.6) imply bi3bi2 which contradicts our assumption. Again, if ai2,k2=0 then bi2R1bi1 imply bi2bi1 which is again a contradiction. Now if bi3R2bi1 then there exists some k1<i1 such that ak1,i1=0,ak1,i3=1. Now if ak1,i2=0 then bi3R2bi2 from (3.7) imply bi3bi2 which is a contradiction. Again, if ak1,i2=1 then bi2R2bi1 imply bi2bi1 which is again a contradiction. Now let bi3R1bk and bkR2bi1 for some kV such that i1<k<i3. Then there exist k2>i3 and k1<i1 such that ai3,k2=0,ak,k2=1,ak1,i1=0,ak1,k=1. Now when i1<k<i2 then if ai2,k2=1 then bi3R1bi2 from (3.6) imply bi3bi2 which is not true. Again, if ai2,k2=0 then bi2R1bk from (3.6) which helps us to conclude bi2bi1 from (3.8) as bkR2bi1 which is a contradiction. Now when i2<k<i3 then if ak1,i2=1, bi2R2bi1, i.e., bi2bi1 from (3.7) which is not true. Further, if ak1,i2=0 then bkR2bi2 which helps us to conclude bi3bi2 from (3.8) as bi3R1bk, which is again a contradiction.

Case (iv) i3<i1<i2

Since bi2bi3, if bi2R1bi3 then there exist k2>i2 such that ai2,k2=0,ai3,k2=1. Now if ai1,k2=1 then bi2R1bi1 from (3.6), i.e., bi2bi1 which is a contradiction. Hence ai1,k2=0 which imply bi1R1bi3. Therefore bi1bi3. If bi2R2bi3 then there exist k1<i3 such that ak1,i3=0,ak1,i2=1. If ak1,i1=0 then bi2R2bi1 from (3.7) imply bi2bi1 which is a contradiction. Hence ak1,i1=1 which imply bi1R2bi3, i.e., bi1bi3. If bi2R1bk and bkR2bi3 for some kV such that i3<k<i2. Then there exist k2>i2,k1<i3 such that ai2,k2=0,ak,k2=1,ak1,i3=0,ak1,k=1. Now when i3<k<i1, if ai1,k2=1 then bi2R1bi1 from (3.6) which imply bi2bi1 which is a contradiction. Hence ai1,k2=0 which imply bi1R1bk which helps us to conclude bi1bi3 from (3.8) as bkR2bi3. Again when i1<k<i2, if ak1,i1=0 then bkR2bi1 imply bi2bi1 from (3.8) (as bi2R1bk) which contradicts our assumption. Hence ak1,i1=1 which imply bi1bi3 as earlier.

Case (v) i3<i2<i1

We need the following observations to prove this case.

Observations:

  1. First note ai3,i2=ai2,i1=1 as i3<i2 and i2<i1.

  2. Next, if bi1R1br for any r<i1 then there exists some k2>i1 such that ai1,k2=0,ar,k2=1.

  3. Further bmR2bi2 for any m>i2 imply existence of some k1<i2 such that ak1,i2=0,ak1,m=1. Now if i3<k1<i2 then ai3,i2 becomes zero as ak1,i2=0 is up open, which imply bi3bi2 which contradicts our assumption. Hence k1<i3.

  4. Again, bi2R1bj for any j<i2 ensure existence of some k2>i2 satisfying ai2,k2=0,aj,k2=1. Clearly k2>i1 otherwise (2.1) gets contradicted for vertices {j,i2,k2,i1}.

  5. Similarly, blR2bi3 for any l>i3 ensure existence of some k1<i3 satisfying ak1,i3=0,ak1,l=1.

  6. Let bi2R1bi3. Then k2>i1 from observation 4. Now if bpR2bi2 for any p>i2 then k1<i3 from observation 3. Next, if ak1,i3=1 then condition 3 of Definition 3.11 gets contradicted for vertices {k1,i3,i2,p,k2}. Hence ak1,i3=0, which imply bpR2bi3 from (3.7).

  7. Let bi2R2bi3. Then k1<i3 from observation 5. Now if bsR2bi2 for any s>i2 then k1<i3 from observation 3. Next if k1<k1 then (2.1) gets contradicted for vertices {k1,k1,i2,s}. Hence k1<k1<i3 which imply ak1,i3=0 as ak1,i3=0 is up open. Thus we get bsR2bi3 from (3.7).

Main proof: Let bi1R1bi2. Now if bi2R1bj for any j<i2 then using observations 2 and 4 we note k2>k2 otherwise (2.1) gets contradicted for the vertices {j,i2,k2,k2}. Now ai1,k2=0 clearly as ai1,k2=0 is right open. This imply bi1R1bj from (3.6) which imply bi1bj.

If bi2R1bi3, we get bi1bi3 from last paragraph. Again, if bi2R2bi3 then bi1bi3 from (3.8) as bi1R1bi2 and i3<i2<i1. When bi2R1bk and bkR2bi3 for some kV such that i3<k<i2, we get bi1R1bk from above paragraph. Hence we can conclude bi1bi3 from (3.8) as bkR2bi3.

Now let bi1R2bi2. If bi2R1bi3 then using observations 3, 4, 6 one can get bi1R2bi3, i.e.; bi1bi3. Again when bi2R2bi3 then bi1R2bi3 follows from observation 7, i.e., bi1bi3. Now if bi2R1bk and bkR2bi3 for some kV such that i3<k<i2 then if ak1,k=1, condition 3 of Definition 3.11 gets contradicted for vertices {k1,k,i2,i1,k2} (observations 3, 4). Hence ak1,k=0 which imply bi1R2bk which helps us to conclude bi1bi3 from (3.8) as bkR2bi3.

Let bi1R1bk and bkR2bi2 for some kV such that i2<k<i1. If bi2R1bi3 then from observations 3, 4, 6 we get bkR2bi3 which helps us to conclude bi1bi3 from (3.8) as bi1R1bk. Again if bi2R2bi3 then bkR2bi3 from observation 7 which helps us to conclude bi1bi3 from (3.8) as bi1R1bk.

Now if bi2R1bk and bkR2bi3 for some kV such that i3<k<i2. Then if k1<k1 (observations 3, 5) then ak1,k=1 applying (2.1) on vertices {k1,k1,k,k}. Therefore, condition 3 of Definition 3.11 gets contradicted for vertices {k1,k,i2,k,k2} (observations 3, 4). Hence we get k1<k1<i3. Thus ak1,i3 becomes zero as ak1,i3=0 is up open, which imply bkR2bi3 from (3.7). Hence from (3.8) we get bi1bi3 as bi1R1bk and i3<k<i1.

Case(vi) i2<i1<i3

Let’s assume bi3bi1 on the contrary. We make the following observations, which we will use throughout the proof.

Observations:

  1. First note that ai2,i1=1 as i2<i1.

  2. Next, if bi1R1br for any r<i1 then there exists k2>i1 such that ai1,k2=0,ar,k2=1. Now if i1<k2<i3 then ai1,i3 becomes zero as ai1,k2=0 is right open, which implies bi1bi3. But this contradicts our assumption. Hence k2>i3.

  3. Again, if bmR2bi2 for any m>i2 then there exists k1<i2 satisfying ak1,i2=0,ak1,m=1.

  4. If blR2bi1 for any l>i1 then there exists k1<i1 satisfying ak1,i1=0,ak1,l=1. Now if i2<k1<i1 then (2.1) gets contradicted for vertices {i2,k1,i1,l} using observation 1. Hence k1<i2.

  5. Again, if bi3R1bj for any j<i3 ensure existence of some k2>i3 such that ai3,k2=0,aj,k2=1.

  6. Let bi1R1bi2. Then k2>i3 from observation 2. Now if bjR2bi1 for any i1<j<k2 then from observation 4 we get k1<i2. Again, if ak1,i2=1 then condition 3 gets contradicted for vertices {k1,i2,i1,j,k2}. Hence ak1,i2=0.

  7. Let bi1R2bi2. Then k1<i2 from observation 3. Now if blR2bi1 holds for any l>i1 then k1<i2 from observation 4. Moreover k1<k1 otherwise (2.1) gets contradicted for vertices {k1,k1,i1,l}.

Main proof:

Let bi1R1bi2. Now if bi3R1bi1 holds then if k2>k2 then (2.1) gets contradicted for the vertices {i2,i1,k2,k2} (observations 2, 5). Again, if i3<k2<k2 then ai3,k2 becomes zero as ai3,k2=0 is right open, which implies bi3R1bi2 from (3.6), i.e., bi3bi2 which is a contradiction. Now if bi3R2bi1 then k1<i2 from observation 4. Again ak1,i2=0 from observation 6. But in this case, bi3R2bi2 from (3.7) imply bi3bi2 which is not true. Now if bi3R1bk and bkR2bi1 for some kV such that i1<k<i3. Note that k1<i2 from observation 4. Again ak1,i2=0 from observation 6. Hence we get bkR2bi2 which along with bi3R1bk helps us to conclude bi3bi2 from (3.8), which isnot true.

Now let bi1R2bi2. Now if bi3R1bi1 then bi3bi2 from (3.8) as bi1R2bi2, which is a contradiction. If bi3R2bi1 then k1<k1 from observations 3, 4, 7. Hence ak1,i2=0 as ak1,i2=0 is up open, which imply bi3R2bi2 from (3.7), i.e., bi3bi2 which is a contradiction. Again, if bi3R1bk and bkR2bi1 for some kV such that i1<k<i3 then k2>i3 and k1<i1 follows from observations 4, 5. Again, note k1<k1 from observation 7. Now ak1,i2 must be zero as ak1,i2=0 is up open, which implies bkR2bi2 from (3.7) which helps us to conclude bi3bi2 from (3.8) as bi3R1bk, which again contradicts ourassumption.

Now let bi1R1bk and bkR2bi2 for some kV such that i2<k<i1. Then k2>i3 from observation 2. Now if bi3R1bi1 then k2 (observation 5) can not be greater than k2 as in that case (2.1) gets contradicted for vertices {k,i1,k2,k2}. Again, if i3<k2<k2 then ai3,k2=0 as ai3,k2=0 is right open, which implies bi3R1bk from (3.6) which helps us to conclude bi3bi2 from (3.8) as bkR2bi2, which is again a contradiction.

Now if bi3R2bi1 then k1<i2 from observation 4. Now if k1<k1<i2 (observations 3, 4) then ak1,k=1 applying (2.1) on vertices {k1,k1,k,i3}. Now applying condition 3 of Definition 3.11 on vertices {k1,k,i1,i3,k2} one can find a contradiction. Now if k1<k1 then ak1,i2 becomes zero as ak1,i2=0 is up open, which implies bi3R2bi2 from (3.7), i.e., bi3bi2 which is a contradiction.

Again if bi3R1bk and bkR2bi1 for some kV such that i1<k<i3 then k1<i2 from observation 4. Now if k1<k1<i2 then ak1,k become one from (2.1) applying on vertices {k1,k1,k,k}. Hence applying condition 3 of Definition 3.11 on vertices {k1,k,i1,k,k2} one can get a contradiction. Now if k1<k1 then ak1,i2 becomes zero as ak1,i2=0 is up open, which imply bkR2bi2 from (3.7) which helps us to conclude bi3bi2 from (3.8) as bi3R1bk, which is again a contradiction.