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

On the existence of minimally tough graphs having large minimum degrees

Morteza Hasanvand
Department of Mathematical Sciences, Sharif University of Technology, Tehran, Iran
๐—†๐—ˆ๐—‹๐—๐–พ๐—“๐–บ.๐—๐–บ๐—Œ๐–บ๐—‡๐—๐–บ๐—‡๐–ฝโ€‹@โ€‹๐–บ๐—…๐—Ž๐—†.๐—Œ๐—๐–บ๐—‹๐—‚๐–ฟ.๐–พ๐–ฝ๐—Ž\mathsf{morteza.hasanvand@alum.sharif.edu}
Abstract

Kriesel conjectured that every minimally 11-tough graph has a vertex with degree precisely 22. Katona and Varga (2018) proposed a generalized version of this conjecture which says that every minimally tt-tough graph has a vertex with degree precisely โŒˆ2โ€‹tโŒ‰\lceil 2t\rceil, where tt is a positive real number. This conjecture has been recently verified for several families of graphs. For example, Ma, Hu, and Yang (2023) confirmed it for claw-free minimally 3/23/2-tough graphs. Recently, Zheng and Sun (2024) disproved this conjecture by constructing a family of 44-regular graphs with toughness approaching to 11.

In this paper, we disprove this conjecture for planar graphs and their line graphs. In particular, we construct an infinite family of minimally tt-tough non-regular claw-free graphs with minimum degree close to thrice their toughness. This construction not only disproves a renewed version of Generalized Krieselโ€™s Conjecture on non-regular graphs proposed by Zheng and Sun (2024), it also gives a supplement to a result due to Ma, Hu, and Yang (2023) who proved that every minimally tt-tough claw-free graph with tโ‰ฅ2t\geq 2 has a vertex of degree at most 3โ€‹t+โŒˆ(tโˆ’5)/3โŒ‰3t+\lceil(t-5)/3\rceil. Moreover, we conjecture that there is not a fixed constant cc such that every minimally tt-tough graph has minimum degree at most โŒˆcโ€‹tโŒ‰\lceil ct\rceil.

Keywords: Toughness; minimum degree; minimally tough graph; claw-free; planar; regular.

1 Introduction

In this article, all graphs are considered simple. Let GG be a graph. The vertex set, the edge set, the number of components, and the independence number of GG are denoted by Vโ€‹(G)V(G), Eโ€‹(G)E(G), ฯ‰โ€‹(G)\omega(G), ฮฑโ€‹(G)\alpha(G), respectively. For a positive real number tt, a graph GG is said to be tt-tough if ฯ‰โ€‹(G)โ‰คmaxโก{1,1tโ€‹|S|}\omega(G)\leq\max\{1,\frac{1}{t}|S|\} for all SโІ(G)S\subseteq(G). The maximum positive real number tt such that GG is tt-tough is called the toughness of GG that is denoted by tโ€‹(G)t(G). In another word, for a non-complete graph GG, tโ€‹(G)=minโก{|S|ฯ‰โ€‹(Gโˆ–S):ฯ‰โ€‹(Gโˆ–S)โ‰ฅ2}t(G)=\min\{\frac{|S|}{\omega(G\setminus S)}:\omega(G\setminus S)\geq 2\}. Likewise, GG is said to be minimally tt-tough, if Gโˆ’eG-e is not tt-tough, for every edge eโˆˆEโ€‹(G)e\in E(G). A graph GG is called claw-free, if it does not an induced star of size three. A circulant graph ๐’žโ€‹(n,A)\mathcal{C}(n,A) is a graph GG with vertices v1,โ€ฆ,vnv_{1},\ldots,v_{n} and two vertices viv_{i} and vjv_{j} are adjacent if and only if iโˆ’jโˆˆAโІโ„คni-j\in A\subseteq\mathbb{Z}_{n} (mod nn), where โ„คn\mathbb{Z}_{n} denotes the cyclic group of order nn and โ„คn={1,โ€ฆ,n}\mathbb{Z}_{n}=\{1,\ldots,n\}. A graph HH is called ss-solid, if it is obtained from a graph GG by replacing every vertex vv with sโ€‹(v)s(v) copies v1,โ€ฆ,vsโ€‹(v)v_{1},\ldots,v_{s(v)}, and two vertices viv_{i} and wjw_{j} are adjacent if vv and ww are adjacent in the original graph GG, where ss is a positive integer-valued function on Vโ€‹(G)V(G); two 22-solid graphs are drawn in Figureย 6. We denote this graph by ๐’ฎโ€‹(G,s)\mathcal{S}(G,s). We denote by Sโ€‹(G)S(G) the subdivision graph of GG which can be obtained from it by inserting a new vertex on each edge. For a graph GG, the line graph Lโ€‹(G)L(G) is a graph whose vertex set is Eโ€‹(G)E(G) and also two e1,e2โˆˆEโ€‹(G)e_{1},e_{2}\in E(G) are adjacent in Lโ€‹(G)L(G) if they have a common end in GG. The square of a graph GG is a graph with the same vertex set and two vertices are adjacent if their distance in GG is at most 22. For two graphs GG and HH, the Cartesian product Gโ€‹โ–กโ€‹HG\square H refers to the graph with the vertex set {(v,w):vโˆˆVโ€‹(G),wโˆˆVโ€‹(H)}\{(v,w):v\in V(G),w\in V(H)\} such that (v,w)(v,w) and (vโ€ฒ,wโ€ฒ)(v^{\prime},w^{\prime}) are adjacent if vโ€‹vโ€ฒโˆˆEโ€‹(G)vv^{\prime}\in E(G) or wโ€‹wโ€ฒโˆˆEโ€‹(H)ww^{\prime}\in E(H). The complete graph and the path graph of order nn are denoted by KnK_{n} and PnP_{n}, respectively. A graph is called rr-regular, if all vertex-degrees are the same integer number rr. A graph GG of order at least k+1k+1 is called kk-connected, if it remains connected after removing any set of vertices of size at most kโˆ’1k-1.

In 1973 Chvรกtalย [4] introduced the concept of toughness inspired by a property of Hamiltonian graphs. In fact, every Hamiltonian graph must be 11-tough. He also conjectured that tough enough graphs admit a Hamiltonian cycle. This concept is stronger than connectivity, as he noted that every tt-tough graph is 2โ€‹t2t-connected as well.

Theorem 1.1

.([4, Proposition 1.3]) Every tt-tough graph is also 2โ€‹t2t-connected.

For higly connected graphs, Mader (1971) [10] proved that every minimally kk-connected graph has a vertex of degree kk. Broersma, Engsberg, Trommel (1999) [2] defined the concept of minimally tough graphs. Motivated by Maderโ€™s result, one may ask whether minimally tough graphs have small minimum degrees (even in special families of graphs). Kriesell conjectured that every minimally 11-tough graph has minimum degree precisely two (similar to Hamiltonian cycles).

Conjecture 1.2

.([6]) Every minimally 11-tough graph has a vertex of degree 22.

A generalization of this conjecture is also proposed by Katona and Varga (2018)ย [8] which says that minimally tough graph have minimum degree close to twice their toughnesses.

Conjecture 1.3

.([8]) Every minimally tt-tough graph has a vertex of degree โŒˆ2โ€‹tโŒ‰\lceil 2t\rceil, where t>0t>0.

Recently, Zheng and Sun (2024) [15] showed that Conjectureย 1.3 fails for real numbers tt close enough to 11. As a consequence, they also remarked that Conjectureย 1.2 is sharp according to their construction (the necessary toughness cannot be increased even a little).

Theorem 1.4

.([15]) For every inetegr kk with kโ‰ฅ2k\geq 2, there is a 44-regular minimally (1+12โ€‹kโˆ’1)(1+\frac{1}{2k-1})-tough graph.

They also proposed the following revised version of Conjectureย 1.3 for non-regular graphs and posed one open problem for further investigation in regular graphs. Their problem is recently settled by Cheng, Li, and Liu (2024) [3] who disproved Conjectureย 1.3 by a new family of 44-regular graphs and 66-regular graphs.

Conjecture 1.5

.([15]) Every non-regular minimally tt-tough graph has a vertex of degree โŒˆ2โ€‹tโŒ‰\lceil 2t\rceil, where t>0t>0.

In this paper, we first present all counterexamples of Conjectureย 1.3 having at most eleven vertices. Among all of them, one graph was already discovered in [15], and one of them is not regular which shows that Conjectureย 1.5 fails for small graphs. Next, we show that Generalized Kriesellโ€™s Conjecture fails even in 44-regular planar graphs and the square of planar graphs (by calling a special counterexample in [5]). Note that Generalized Krieselโ€™s Conjecture seems true for a large ratio of graphs and it has been recently verified for several families of graphs, see [7, 9, 11]). In particular, Ma, Hu, and Yang (2023) confirmed it for claw-free minimally 3/23/2-tough graphs.

Theorem 1.6

.([12]) Every claw-free minimally 3/23/2-tough graph has minimum degree 33.

In general, Ma, Hu, and Yang (2023) [13] confirmed a weaker version of it on claw-free graphs by giving the explicit linear bound 3โ€‹t+โŒˆtโˆ’53โŒ‰3t+\lceil\frac{t-5}{3}\rceil on the minimum degrees. In Sectionย 3, disprove Generalized Krieselโ€™s Conjecture for claw-free graphs by constructing a family of (non-regular) counterexamples whose minimum degrees are close to thrice their highnesses. It remains to decide how much the number โŒˆtโˆ’53โŒ‰\lceil\frac{t-5}{3}\rceil in the following theorem can be reduced.

Theorem 1.7

.([12]) Every minimally tt-tough claw-free graph with tโ‰ฅ2t\geq 2 has a vertex of degree at most 3โ€‹t+โŒˆtโˆ’53โŒ‰3t+\lceil\frac{t-5}{3}\rceil.

Motivated by these new graph constructions, one may ask whether every minimally tt-tough graph has minimum degree at most โŒˆcโ€‹tโŒ‰\lceil ct\rceil for a constant number cc. We conjecture that the answer is false and put forward a conjecture in Sectionย 4 for this purpose.

2 Small graphs and planar graphs

By a computer search, we observed that among all connected graphs of order at most 1010, there is a unique counterexample of Conjectureย 1.3 which was already found in [15, Theoremย 9] (see the left graph is Figureย 1) and there are only three counterexamples of order 1111 (the middle-right graph was also found in [3, Theoremย 1.6]). One of these graphs is not regular (the right graph) and consequently Conjectureย 1.5 fails. (We will construct an infinite family of counterexamples in Theoremsย 3.1 andย  3.2). Note that there is another counterexample which is shown in the right graph of Figureย 6.

Refer to caption
Figure 1: All minimally tโ€‹(G)t(G)-tough graphs GG of order at most 1111 satisfying ฮดโ€‹(G)>โŒˆ2โ€‹tโ€‹(G)โŒ‰\delta(G)>\lceil 2t(G)\rceil, where tโ€‹(G)โˆˆ{4/3,3/2,6/4,5/2}t(G)\in\{4/3,3/2,6/4,5/2\}, respectively (from left to right).

In the following theorem, we are going to introduce planar 44-regular counterexamples using a simple structure based on the copy of the complete graph K5K_{5} minus an edge. We originally found the smallest graph by applying a computer search on the restricted outputs of the program plantri due to Brinkmann and McKay (2007) [1].

Theorem 2.1

. There are infinitely many planar minimally 3/23/2-tough 44-regular graphs.

Proof.

Let mm be an even number with mโ‰ฅ4m\geq 4. For each ii with 1โ‰คiโ‰คm1\leq i\leq m, let vi,1,โ€ฆ,vi,6v_{i,1},\ldots,v_{i,6} be six numbers. First, we add five edges vi,1โ€‹vi,2,โ€ฆ,vi,4โ€‹vi,5,vi,5โ€‹vi,1v_{i,1}v_{i,2},\ldots,v_{i,4}v_{i,5},v_{i,5}v_{i,1} (which forms a cycle). Next, we add four edges vi,6โ€‹vi,jv_{i,6}v_{i,j}, where jโˆˆ{1,โ€ฆ,4}j\in\{1,\ldots,4\}. Finally, if ii is even, we add three edges vi,1โ€‹vi+1,1,vi,2โ€‹vi+1,5,vi,5โ€‹vi+1,2v_{i,1}v_{i+1,1},v_{i,2}v_{i+1,5},v_{i,5}v_{i+1,2}, and if ii is odd, we add three edges vi,4โ€‹vi+1,4,vi,3โ€‹vi+1,5,vi,5โ€‹vi+1,3v_{i,4}v_{i+1,4},v_{i,3}v_{i+1,5},v_{i,5}v_{i+1,3} (if i=mi=m, we compute i+1i+1 modulo mm that means i+1=1i+1=1). Call the resulting graph GG. Obviously, this graph is planar and 44-regular. For the smallest case mm=4, the graph GG is illustrated in Figureย 2.

Refer to caption
Figure 2: A minimally 3/23/2-tough planar 44-regular graph of order 2424.

We claim that tโ€‹(G)=3/2t(G)=3/2. We first show that tโ€‹(G)โ‰ฅ3/2t(G)\geq 3/2. Let SS be an arbitrary cut-set of GG. Define GiG_{i} to be the induce subgraph of GG with the vertex set Vi={vi,j:1โ‰คjโ‰ค6}V_{i}=\{v_{i,j}:1\leq j\leq 6\} and let Si=SโˆฉViS_{i}=S\cap V_{i}. Since GiG_{i} has independence number at most two, we must have ฯ‰โ€‹(Giโˆ–Si)โ‰ค2\omega(G_{i}\setminus S_{i})\leq 2. Let ฯ‰i\omega_{i} be the number of components of Giโˆ–SiG_{i}\setminus S_{i} which is not connected to a component of Gi+1โˆ–Si+1G_{i+1}\setminus S_{i+1} in Gโˆ–SG\setminus S. Define ฮธi=|Si|โˆ’32โ€‹ฯ‰i\theta_{i}=|S_{i}|-\frac{3}{2}\omega_{i}. Let start with the following simple but useful subclaims.

Subclaim 1: If Giโˆ–SiG_{i}\setminus S_{i} is disconnected, then |Si|โ‰ฅ3|S_{i}|\geq 3 or Si={vi,1,vi,4}S_{i}=\{v_{i,1},v_{i,4}\}

Subclaim 2: If {vi,1,vi,5}โІSi\{v_{i,1},v_{i,5}\}\subseteq S_{i} or {vi,4,vi,5}โІSi\{v_{i,4},v_{i,5}\}\subseteq S_{i}, then ฮธiโ‰ฅ1/2\theta_{i}\geq 1/2. In particular, in both cases, we have ฮธiโ‰ฅ1\theta_{i}\geq 1, if |Si|โ‰ฅ3|S_{i}|\geq 3 or there is one component of Giโˆ–SiG_{i}\setminus S_{i} connecting to a component of Gi+1โˆ–Si+1G_{i+1}\setminus S_{i+1}.

If there is a component of GG having one vertex from every subgraph GiG_{i}, then ฯ‰โ€‹(Gโˆ–S)โ‰คโˆ‘1โ‰คiโ‰คm(ฯ‰โ€‹(Giโˆ–Si)โˆ’1)+1\omega(G\setminus S)\leq\sum_{1\leq i\leq m}(\omega(G_{i}\setminus S_{i})-1)+1. Moreover, ฯ‰iโ‰ค1\omega_{i}\leq 1 and ฮธiโ‰ฅ0\theta_{i}\geq 0 for all indices ii. Since SS is a cut-set, there is a subgraph GiG_{i} in which Giโˆ–SiG_{i}\setminus S_{i} is disconnected. If |Si|โ‰ฅ3|S_{i}|\geq 3, then ฮธiโ‰ฅ3/2\theta_{i}\geq 3/2. Otherwise, by Subclaim 1, Si={vi,1,vi,4}S_{i}=\{v_{i,1},v_{i,4}\}. If there is a subgraph GjG_{j} in which Gjโˆ–SG_{j}\setminus S is connected and |Sj|>0|S_{j}|>0, then ฮธi+ฮธjโ‰ฅ1/2+1=3/2\theta_{i}+\theta_{j}\geq 1/2+1=3/2. Otherwise, there must be at least two other subgraphs Gj1G_{j_{1}} and Gj2G_{j_{2}} in which Gjtโˆ–SjtG_{j_{t}}\setminus S_{j_{t}} is connected. Thus ฮธi+ฮธj1+ฮธj2โ‰ฅ1/2+1/2+1/2=3/2\theta_{i}+\theta_{j_{1}}+\theta_{j_{2}}\geq 1/2+1/2+1/2=3/2. Therefore, โˆ‘1โ‰คiโ‰คmฮธiโ‰ฅ3/2\sum_{1\leq i\leq m}\theta_{i}\geq 3/2 which implies that |S|โ‰ฅ32โ€‹ฯ‰โ€‹(Gโˆ–S)|S|\geq\frac{3}{2}\omega(G\setminus S). Now, assume that there is no component of Gโˆ–SG\setminus S having one vertex from every subgraph GiG_{i}. Therefore,

โˆ‘1โ‰คiโ‰คmฮธi=โˆ‘1โ‰คiโ‰คm(|Si|โˆ’32โ€‹ฯ‰i)=|S|โˆ’32โ€‹ฯ‰โ€‹(Gโˆ–S).\sum_{1\leq i\leq m}\theta_{i}=\sum_{1\leq i\leq m}(|S_{i}|-\frac{3}{2}\omega_{i})=|S|-\frac{3}{2}\omega(G\setminus S).

We say an index ii is good if ฮธiโ‰ฅ0\theta_{i}\geq 0, and bad otherwise. If all indices are good, then |S|ฯ‰โ€‹(Gโˆ–S)โ‰ฅ3/2\frac{|S|}{\omega(G\setminus S)}\geq 3/2. Otherwise, by applying the following assertion, one can conclude that there is a partition PP of indices {1,โ€ฆ,m}\{1,\ldots,m\} such that for every IโˆˆPI\in P, โˆ‘iโˆˆIฮธiโ‰ฅ0\sum_{i\in I}\theta_{i}\geq 0 and so โˆ‘1โ‰คiโ‰คmฮธiโ‰ฅ0\sum_{1\leq i\leq m}\theta_{i}\geq 0 and |S|ฯ‰โ€‹(Gโˆ–S)โ‰ฅ3/2\frac{|S|}{\omega(G\setminus S)}\geq 3/2.

Subclaim 3: If bb is a bad index, then there exists an index jโˆˆ{1,2,3}j\in\{1,2,3\} such that โˆ‘bโ‰คiโ‰คb+jฮธiโ‰ฅ0\sum_{b\leq i\leq b+j}\theta_{i}\geq 0 and all indices b+1,โ€ฆ,b+jb+1,\ldots,b+j are good.

Proof of Subclaim 3. By the isomorphic property, we may assume that bb is odd. Suppose the assertion false. This implies that ฮธb+maxโก{0,ฮธb+1,ฮธb+1+ฮธb+2,ฮธb+1+ฮธb+2+ฮธb+3}โ‰คโˆ’1/2\theta_{b}+\max\{0,\theta_{b+1},\theta_{b+1}+\theta_{b+2},\theta_{b+1}+\theta_{b+2}+\theta_{b+3}\}\leq-1/2. In addition, any component of Gbโˆ–SbG_{b}\setminus S_{b} is not connected to a component of Gb+1โˆ–Sb+1G_{b+1}\setminus S_{b+1}. Obviously, ฯ‰bโˆˆ{1,2}\omega_{b}\in\{1,2\}. To derive a contradiction, we shall consider the following two cases:

Case A: ฯ‰b=1\omega_{b}=1.

In this case, |Sb|โ‰ค1|S_{b}|\leq 1 and โˆ’3/2โ‰คฮธbโ‰คโˆ’1/2-3/2\leq\theta_{b}\leq-1/2. If Sbโˆฉ{vb,3,vb,4,vb,5}=โˆ…S_{b}\cap\{v_{b,3},v_{b,4},v_{b,5}\}=\emptyset, then {vb+1,3,vb+1,4,vb+1,5}โІSb+1\{v_{b+1,3},v_{b+1,4},v_{b+1,5}\}\subseteq S_{b+1} and ฮธb+1โ‰ฅ3/2\theta_{b+1}\geq 3/2. Since ฮธb+ฮธb+1โ‰ฅ0\theta_{b}+\theta_{b+1}\geq 0, we derive a contradiction. Assume that |Sbโˆฉ{vb,3,vb,4,vb,5}|=1|S_{b}\cap\{v_{b,3},v_{b,4},v_{b,5}\}|=1. In this case, ฮธb=โˆ’1/2\theta_{b}=-1/2. If Sb={vb,5}S_{b}=\{v_{b,5}\}, then since the unique component of Gbโˆ–SbG_{b}\setminus S_{b} is not connected to a component of Gb+1โˆ–Sb+1G_{b+1}\setminus S_{b+1}, we must have {vb+1,4,vb+1,5}โІSb+1\{v_{b+1,4},v_{b+1,5}\}\subseteq S_{b+1} and ฮธb+1โ‰ฅ1/2\theta_{b+1}\geq 1/2 using Subclaim 2. Since ฮธb+ฮธb+1โ‰ฅ0\theta_{b}+\theta_{b+1}\geq 0, we derive a contradiction. If Sb={vb,4}S_{b}=\{v_{b,4}\}, then it is not difficult to check that Sb+1={vb+1,3,vb+1,5,vb+1,6}S_{b+1}=\{v_{b+1,3},v_{b+1,5},v_{b+1,6}\} and ฮธb+1=0\theta_{b+1}=0. Similarly, since any component of Gb+1โˆ–Sb+1G_{b+1}\setminus S_{b+1} is not connected to a component of Gb+2โˆ–Sb+2G_{b+2}\setminus S_{b+2}, we must have {vb+2,1,vb+2,5}โІSb+2\{v_{b+2,1},v_{b+2,5}\}\subseteq S_{b+2} and ฮธb+2โ‰ฅ1/2\theta_{b+2}\geq 1/2 using Subclaim 2. Since ฮธb+ฮธb+1+ฮธb+2โ‰ฅ0\theta_{b}+\theta_{b+1}+\theta_{b+2}\geq 0, we derive a contradiction. If Sb={vb,3}S_{b}=\{v_{b,3}\}, then it is not difficult to check that Sb+1={vb+1,1,vb+1,3,vb+1,4}S_{b+1}=\{v_{b+1,1},v_{b+1,3},v_{b+1,4}\} and ฮธb+1=0\theta_{b+1}=0. Since any component of Gb+1โˆ–Sb+1G_{b+1}\setminus S_{b+1} is not connected to a component of Gb+2โˆ–Sb+2G_{b+2}\setminus S_{b+2}, we must have Sb+2={vb+2,2,vb+2,5,vb+2,6}S_{b+2}=\{v_{b+2,2},v_{b+2,5},v_{b+2,6}\} and ฮธb+2=0\theta_{b+2}=0. Therefore, we must have Sb+3={vb+3,4,vb+3,5}S_{b+3}=\{v_{b+3,4},v_{b+3,5}\} and ฮธb+2โ‰ฅ1/2\theta_{b+2}\geq 1/2 using Subclaim 2. Since ฮธb+ฮธb+1+ฮธb+2+ฮธb+3โ‰ฅ0\theta_{b}+\theta_{b+1}+\theta_{b+2}+\theta_{b+3}\geq 0, we again derive a contradiction.

Case B: ฯ‰b=2\omega_{b}=2.

In this case, by Subclaim 1, we must have Sb={vb,1,vb,4}S_{b}=\{v_{b,1},v_{b,4}\} and so ฮธb=โˆ’1\theta_{b}=-1. Since there is not a component of Gbโˆ–SbG_{b}\setminus S_{b} connecting to a component of Gb+1โˆ–Sb+1G_{b+1}\setminus S_{b+1}, it is not difficult to check that {vb+1,3,vb+1,5}โІSb+1\{v_{b+1,3},v_{b+1,5}\}\subseteq S_{b+1}, {vb+1,1,vb+1,2}โˆฉSb+1=โˆ…\{v_{b+1,1},v_{b+1,2}\}\cap S_{b+1}=\emptyset, and ฮธb+1โ‰ฅ0\theta_{b+1}\geq 0. Consequently, Sb+2={vb+2,1,vb+2,2}S_{b+2}=\{v_{b+2,1},v_{b+2,2}\} and ฮธb+2=1/2\theta_{b+2}=1/2 using Subclaim 2. This implies that Sb+3={vb+3,4,vb+3,5}S_{b+3}=\{v_{b+3,4},v_{b+3,5}\} and ฮธb+3=1/2\theta_{b+3}=1/2 using Subclaim 2 again. Since ฮธb+ฮธb+1+ฮธb+2+ฮธb+3โ‰ฅ0\theta_{b}+\theta_{b+1}+\theta_{b+2}+\theta_{b+3}\geq 0, we derive contradiction, as desired.

Therefore, tโ€‹(G)โ‰ฅ3/2t(G)\geq 3/2. From now on, for notational simplicity, for each odd index ii, we set Si={vi,1,vi,4}S_{i}=\{v_{i,1},v_{i,4}\}, and for even index ii, we set Si=Viโˆ–{vi,1,vi,4}S_{i}=V_{i}\setminus\{v_{i,1},v_{i,4}\}; recall that Vi={vi,j:1โ‰คjโ‰ค6}V_{i}=\{v_{i,j}:1\leq j\leq 6\}. If we set S=โˆช1โ‰คiโ‰คmSiS=\cup_{1\leq i\leq m}S_{i}, then it is easy to check that |S|=3โ€‹m|S|=3m and ฯ‰โ€‹(Gโˆ–S)=2โ€‹m\omega(G\setminus S)=2m. This implies that tโ€‹(G)โ‰ค|S|ฯ‰โ€‹(Gโˆ–S)=3/2t(G)\leq\frac{|S|}{\omega(G\setminus S)}=3/2. Hence tโ€‹(G)=3/2t(G)=3/2 and the claim is proved. It remains to show that GG is minimally tโ€‹(G)t(G)-tough. Let ee be an arbitrary edge of GG. We shall consider the following cases:

Case 1: Both ends of ee lie in the set {vi,2,vi,3,vi,6}\{v_{i,2},v_{i,3},v_{i,6}\}.

By the isomorphic property, we may assume that ii is odd. Let vv be the unique vertex in {vi,2,vi,3,vi,6}\{v_{i,2},v_{i,3},v_{i,6}\} not incident with ee. If we set S=(โˆช1โ‰คjโ‰คmSj)โˆช{v}S=(\cup_{1\leq j\leq m}S_{j})\cup\{v\}, then it is easy to check that |S|=3โ€‹m+1|S|=3m+1 and ฯ‰โ€‹((Gโˆ’e)โˆ–S)=2โ€‹m+1\omega((G-e)\setminus S)=2m+1. This implies that tโ€‹(Gโˆ’e)โ‰ค|S|ฯ‰โ€‹(Gโˆ–S)=3โ€‹m+12โ€‹m+1<3/2t(G-e)\leq\frac{|S|}{\omega(G\setminus S)}=\frac{3m+1}{2m+1}<3/2.

Case 2: eโˆˆ{vi,1โ€‹vi,6,vi,2โ€‹vi,1,vi,2โ€‹vi+1,5}e\in\{v_{i,1}v_{i,6},v_{i,2}v_{i,1},v_{i,2}v_{i+1,5}\}.

By the isomorphic property, we may assume that ii is even. If e=vi,1โ€‹vi,6e=v_{i,1}v_{i,6}, then we set S=(โˆช1โ‰คjโ‰คmSj)โˆ–{vi,6}S=(\cup_{1\leq j\leq m}S_{j})\setminus\{v_{i,6}\}. Otherwise, we set S=(โˆช1โ‰คjโ‰คmSj)โˆ–{vi,2}S=(\cup_{1\leq j\leq m}S_{j})\setminus\{v_{i,2}\}. In both cases, it is easy to check that |S|=3โ€‹mโˆ’1|S|=3m-1 and ฯ‰โ€‹((Gโˆ’e)โˆ–S)=2โ€‹m\omega((G-e)\setminus S)=2m. This implies that tโ€‹(Gโˆ’e)โ‰ค|S|ฯ‰โ€‹(Gโˆ–S)=3โ€‹mโˆ’12โ€‹m<3/2t(G-e)\leq\frac{|S|}{\omega(G\setminus S)}=\frac{3m-1}{2m}<3/2.

Case 3: e=vi,4โ€‹vi+1,4e=v_{i,4}v_{i+1,4}.

By the isomorphic property, we may assume that ii is odd. If we set S=((โˆช1โ‰คjโ‰คmSj)โˆ–{viโˆ’1,2,vi,4,vi+1,3})โˆช{vi,5}S=((\cup_{1\leq j\leq m}S_{j})\setminus\{v_{i-1,2},v_{i,4},v_{i+1,3}\})\cup\{v_{i,5}\}, then it is not difficult to check that |S|=3โ€‹mโˆ’2|S|=3m-2 and ฯ‰โ€‹((Gโˆ’e)โˆ–S)=2โ€‹mโˆ’1\omega((G-e)\setminus S)=2m-1. This implies that tโ€‹(Gโˆ’e)โ‰ค|S|ฯ‰โ€‹(Gโˆ–S)=3โ€‹mโˆ’22โ€‹mโˆ’1<3/2t(G-e)\leq\frac{|S|}{\omega(G\setminus S)}=\frac{3m-2}{2m-1}<3/2.

Case 4: e=vi,1โ€‹vi,5e=v_{i,1}v_{i,5}.

By the isomorphic property, we may assume that ii is even. If we set S=((โˆช1โ‰คjโ‰คmSj)โˆ–{viโˆ’2,2,viโˆ’1,4,vi,3,vi,5})โˆช{viโˆ’1,3,viโˆ’1,6,viโˆ’1,5,vi,4,vi+1,2},S=((\cup_{1\leq j\leq m}S_{j})\setminus\{v_{i-2,2},v_{i-1,4},v_{i,3},v_{i,5}\})\cup\{v_{i-1,3},v_{i-1,6},v_{i-1,5},v_{i,4},v_{i+1,2}\}, then it is not difficult to check that |S|=3โ€‹m+1|S|=3m+1 and ฯ‰โ€‹((Gโˆ’e)โˆ–S)=2โ€‹m+1\omega((G-e)\setminus S)=2m+1. This implies that tโ€‹(Gโˆ’e)โ‰ค|S|ฯ‰โ€‹(Gโˆ–S)=3โ€‹m+12โ€‹m+1<3/2t(G-e)\leq\frac{|S|}{\omega(G\setminus S)}=\frac{3m+1}{2m+1}<3/2.

The remaining edges are similar with respect to isomorphic property of the graph GG. Hence the proof is completed. โ–ก\Box

In 1999 Broersma, Engsberg, and Trommel [2] studied minimally tough square graphs. In particular, they showed that if the square of a graph is minimally 22-tough, then the original graph must be minimally 22-connected and triangle-free. We have recently used a small square graph in [5] to provide a solution for several conjectures and open problems related to list coloring. Surprisingly, we observed that this graph is another counterexample to Conjectureย 1.3. (Note that there is an infinite family of such examples, but this is enough for our purpose).

Theorem 2.2

. There is a minimally 33-tough 77-regular graph GG of order 1212. In particular, GG is the square of the planar line graph Lโ€‹(Sโ€‹(K4))L(S(K_{4})).

Proof.

Let HH be the cubic graph illustrated in Figureย 3 and let GG be the square of HH. It is not difficult to show that GG contains only four maximum independent sets of size 44 which are shown in Figureย 3 by numbers 1,โ€ฆ,41,\ldots,4.

Refer to caption
Figure 3: The square of the planar graph HH is a minimally 33-tough 77-regular graph.

Let II be a maximum independent set of GG. If we set S=Vโ€‹(G)โˆ–IS=V(G)\setminus I, then tโ€‹(G)โ‰ค|S|ฯ‰โ€‹(Gโˆ–S)=|S||I|=93=3t(G)\leq\frac{|S|}{\omega(G\setminus S)}=\frac{|S|}{|I|}=\frac{9}{3}=3. Let SS be a cut-set of Vโ€‹(G)V(G). Since GG is 77-connected, we must have |S|โ‰ฅ7|S|\geq 7. If ฯ‰โ€‹(Gโˆ–S)=2\omega(G\setminus S)=2, then |S|ฯ‰โ€‹(Gโˆ–S)โ‰ฅ72>3\frac{|S|}{\omega(G\setminus S)}\geq\frac{7}{2}>3. If ฯ‰โ€‹(Gโˆ–S)โ‰ฅ3\omega(G\setminus S)\geq 3, then according to the property of GG, the subgraph Gโˆ–SG\setminus S must consist of three isolated vertices and so S=|Vโ€‹(G)|โˆ’3=9S=|V(G)|-3=9 which implies that |S|ฯ‰โ€‹(Gโˆ–S)=3\frac{|S|}{\omega(G\setminus S)}=3. Consequently, tโ€‹(G)=3t(G)=3.

We are going to show that GG is minimally tโ€‹(G)t(G)-tough. Let e=uโ€‹ve=uv be an arbitrary edge of GG. If eโˆˆEโ€‹(H)e\in E(H), then it is not difficult to check that there is a unique edge eโ€ฒ=uโ€ฒโ€‹vโ€ฒe^{\prime}=u^{\prime}v^{\prime} of HH which is not connected to ee by an edge in GG. If we set S=Vโ€‹(G)โˆ–{u,v,uโ€ฒ,vโ€ฒ}S=V(G)\setminus\{u,v,u^{\prime},v^{\prime}\}, then tโ€‹(Gโˆ’e)โ‰ค|S|ฯ‰โ€‹((Gโˆ’e)โˆ–S)=83<3=tโ€‹(G)t(G-e)\leq\frac{|S|}{\omega((G-e)\setminus S)}=\frac{8}{3}<3=t(G). If eโˆ‰Eโ€‹(H)e\not\in E(H), then we may assume that uu and vv have a common neighbor ww in HH for which ww and vv lie in a triangle in HH. In this case, we again have tโ€‹(Gโˆ’e)โ‰ค|S|ฯ‰โ€‹((Gโˆ’e)โˆ–S)=83<3=tโ€‹(G)t(G-e)\leq\frac{|S|}{\omega((G-e)\setminus S)}=\frac{8}{3}<3=t(G), where S=Vโ€‹(G)โˆ–(Iโˆช{u})S=V(G)\setminus(I\cup\{u\}) and II is the maximum independent set of GG including vv. Hence the proof is completed. โ–ก\Box

3 Minimally tough claw-free graphs with minimum degree close to thrice their toughness

In the following theorem, we show that the upper bound in Theoremย 1.7 cannot be replaced by 3โ€‹tโˆ’23t-2 (by setting 2โ€‹t=m=โŒˆ2โ€‹n+13โŒ‰=2โ€‹n+132t=m=\lceil\frac{2n+1}{3}\rceil=\frac{2n+1}{3} provided that nโ‰ก31n\stackrel{{\scriptstyle 3}}{{\equiv}}1). Note that all graphs GG presented below are not regular, and alternatively disprove Conjectureย 1.5. In addition, GG is the line graph of the graph obtained from two copies of nn-stars by identifying mm pairs of pendant vertices belonging to different stars.

Theorem 3.1

. Let nn and mm be two positive integers satisfying 23โ€‹n<m<n\frac{2}{3}n<m<n and nโ‰ฅ7n\geq 7. If GG is the graph obtained from the Cartesian product Knโ€‹โ–กโ€‹P2K_{n}\square P_{2} by deleting mm independent edges connecting two copies of KnK_{n}, then GG is a minimally m2\frac{m}{2}-tough claw-free graph satisfying

ฮดโ€‹(G)=ฮ”โ€‹(G)โˆ’1=nโˆ’1.\delta(G)=\Delta(G)-1=n-1.
Proof.

Let GiG_{i} be a copy of the complete graph KnK_{n} with vertices vi,1,โ€ฆ,vi,nv_{i,1},\ldots,v_{i,n}, where i=1,2i=1,2. For each jj with 1โ‰คjโ‰คm1\leq j\leq m, we add an edge v1,jโ€‹v2,jv_{1,j}v_{2,j} and call the resulting graph GG. The special case n=7n=7 and m=4m=4 is illustrated in Figureย 4 (the right graph).

Refer to caption
Figure 4: The line graph of the left graph is a minimally 5/25/2-tough graph GG satisfying ฮดโ€‹(G)=6\delta(G)=6 (which is shown in the right).

If we set S={v1,p:1โ‰คpโ‰คm}S=\{v_{1,p}:1\leq p\leq m\}, then ฯ‰โ€‹(Gโˆ–S)=2\omega(G\setminus S)=2 and |S|=m|S|=m which implies that tโ€‹(G)โ‰ค|S|ฯ‰โ€‹(Gโˆ–S)=m2t(G)\leq\frac{|S|}{\omega(G\setminus S)}=\frac{m}{2}. On the other hand, if SS is a cut-set of GG, then ฯ‰โ€‹(Gโˆ–S)=2\omega(G\setminus S)=2 and |S|โ‰ฅm|S|\geq m, because ฮฑโ€‹(G)=2\alpha(G)=2 and GG is mm-connected. This implies that |S|ฯ‰โ€‹(Gโˆ–S)โ‰ฅm2\frac{|S|}{\omega(G\setminus S)}\geq\frac{m}{2} and so tโ€‹(G)=m2t(G)=\frac{m}{2}. We are going to show that GG is minimally tโ€‹(G)t(G)-tough. Let ee be an arbitrary edge of GG. If e=v1,jโ€‹v2,je=v_{1,j}v_{2,j}, then tโ€‹(Gโˆ’e)โ‰ค|S|ฯ‰โ€‹((Gโˆ’e)โˆ–S)=mโˆ’12<m2=tโ€‹(G)t(G-e)\leq\frac{|S|}{\omega((G-e)\setminus S)}=\frac{m-1}{2}<\frac{m}{2}=t(G), where S={v1,p:1โ‰คpโ‰คmโ€‹ย andย โ€‹pโ‰ j}S=\{v_{1,p}:1\leq p\leq m\text{ and }p\neq j\}. If e=vi,j1โ€‹vi,j2e=v_{i,j_{1}}v_{i,j_{2}}, then tโ€‹(Gโˆ’e)โ‰ค|S|ฯ‰โ€‹((Gโˆ’e)โˆ–S)=n3<m2=tโ€‹(G)t(G-e)\leq\frac{|S|}{\omega((G-e)\setminus S)}=\frac{n}{3}<\frac{m}{2}=t(G), where S=(Vโ€‹(Gi)โˆ–{vi,j1,vi,j2})โˆช{v3โˆ’i,j1,v3โˆ’i,j2}S=(V(G_{i})\setminus\{v_{i,j_{1}},v_{i,j_{2}}\})\cup\{v_{3-i,j_{1}},v_{3-i,j_{2}}\}. This completes the proof. โ–ก\Box

In the following theorem, we construct two new families of non-regular and regular minimally tt-tough graphs with minimum degree at least โŒˆ3โ€‹tโŒ‰โˆ’1\lceil 3t\rceil-1.

Theorem 3.2

. If nn is an integer with nโ‰ฅ3n\geq 3, then the Cartesian product G=Knโ€‹โ–กโ€‹P3G=K_{n}\square P_{3} is a non-regular minimally n+13\frac{n+1}{3}-tough graph satisfying

ฮดโ€‹(G)=ฮ”โ€‹(G)โˆ’1=n.\delta(G)=\Delta(G)-1=n.

In addition, if G0G_{0} is the graph obtained from GG be deleting the middle vertex of an induced path v1โ€‹v2โ€‹v3v_{1}v_{2}v_{3} and replacing the new edge v1โ€‹v3v_{1}v_{3}, then G0G_{0} is nn-regular and minimally n+13\frac{n+1}{3}-tough.

Proof.

Let GiG_{i} be a copy of the complete graph KnK_{n} with vertices vi,1,โ€ฆ,vi,nv_{i,1},\ldots,v_{i,n}, where i=1,2,3i=1,2,3. For each pp with 1โ‰คpโ‰คn1\leq p\leq n, we insert two edges v1,pโ€‹v2,pv_{1,p}v_{2,p} and v2,pโ€‹v3,pv_{2,p}v_{3,p} into these graphs and call the resulting graph GG. This graph is isomorphic to the Cartesian product Knโ€‹โ–กโ€‹P3K_{n}\square P_{3}. The special case n=5n=5 is illustrated in Figureย 5 (the left graph).

Refer to caption
Figure 5: Two minimally 22-tough graphs with minimum degree 55 (regular and non-regular).

We claim that tโ€‹(G)=n+13t(G)=\frac{n+1}{3}. If we set S={v2,p:1โ‰คp<n}โˆช{v1,n,v3,n}S=\{v_{2,p}:1\leq p<n\}\cup\{v_{1,n},v_{3,n}\}, then |S|=n+1|S|=n+1 and ฯ‰โ€‹(Gโˆ–S)=3\omega(G\setminus S)=3. Hence tโ€‹(G)โ‰ค|S|ฯ‰โ€‹(Gโˆ–S)=n+13t(G)\leq\frac{|S|}{\omega(G\setminus S)}=\frac{n+1}{3}. It remains to show that tโ€‹(G)โ‰ฅn+13t(G)\geq\frac{n+1}{3}. Let SS be an arbitrary cut-set of Vโ€‹(G)V(G). If |S|โ‰ฅn+1|S|\geq n+1, then |S|ฯ‰โ€‹(Gโˆ–S)โ‰ฅ|S|ฮฑโ€‹(Gโˆ–S)โ‰ฅn+13\frac{|S|}{\omega(G\setminus S)}\geq\frac{|S|}{\alpha(G\setminus S)}\geq\frac{n+1}{3}. Assume that |S|โ‰คn|S|\leq n. According to the construction of GG, we must have ฯ‰โ€‹(Gโˆ–S)=2\omega(G\setminus S)=2. More precisely, |Sโˆฉ{v1,p,v2,p,v3,p}|=1|S\cap\{v_{1,p},v_{2,p},v_{3,p}\}|=1 for all integers pp with 1โ‰คpโ‰คn1\leq p\leq n. Therefore, |S|ฯ‰โ€‹(Gโˆ–S)=n2โ‰ฅn+13\frac{|S|}{\omega(G\setminus S)}=\frac{n}{2}\geq\frac{n+1}{3}. Hence the claim is proved. We are going to show that GG is minimally tโ€‹(G)t(G)-tough. Let ee be an arbitrary edge of GG. If e=v2,jโ€‹vi,je=v_{2,j}v_{i,j}, then tโ€‹(Gโˆ’e)โ‰ค|S|ฯ‰โ€‹((Gโˆ’e)โˆ–S)=n3<n+13=tโ€‹(G)t(G-e)\leq\frac{|S|}{\omega((G-e)\setminus S)}=\frac{n}{3}<\frac{n+1}{3}=t(G), where S={v2,p:1โ‰คpโ‰คnโ€‹ย andย โ€‹pโ‰ j}โˆช{v4โˆ’i,j}S=\{v_{2,p}:1\leq p\leq n\text{ and }p\neq j\}\cup\{v_{4-i,j}\}. If e=vi,j1โ€‹vi,j2e=v_{i,j_{1}}v_{i,j_{2}} and iโ‰ 2i\neq 2, then tโ€‹(Gโˆ’e)โ‰ค|S|ฯ‰โ€‹((Gโˆ’e)โˆ–S)=n3<n+13=tโ€‹(G)t(G-e)\leq\frac{|S|}{\omega((G-e)\setminus S)}=\frac{n}{3}<\frac{n+1}{3}=t(G), where S={vi,p:1โ‰คpโ‰คnโ€‹ย andย โ€‹pโ‰ j1,j2}โˆช{v2,j1,v2,j2}S=\{v_{i,p}:1\leq p\leq n\text{ and }p\neq j_{1},j_{2}\}\cup\{v_{2,j_{1}},v_{2,j_{2}}\}. If e=v2,j1โ€‹v2,j2e=v_{2,j_{1}}v_{2,j_{2}}, then tโ€‹(Gโˆ’e)โ‰ค|S|ฯ‰โ€‹((Gโˆ’e)โˆ–S)=n+24<n+13=tโ€‹(G)t(G-e)\leq\frac{|S|}{\omega((G-e)\setminus S)}=\frac{n+2}{4}<\frac{n+1}{3}=t(G), where S={v2,p:1โ‰คpโ‰คnโ€‹ย andย โ€‹pโ‰ j1,j2}โˆช{v1,j1,v1,j2,v3,j1,v3,j2}S=\{v_{2,p}:1\leq p\leq n\text{ and }p\neq j_{1},j_{2}\}\cup\{v_{1,j_{1}},v_{1,j_{2}},v_{3,j_{1}},v_{3,j_{2}}\}. This completes the proof of the first part.

Let G0G_{0} be the graph obtained from GG by deleting the vertex v2,nv_{2,n} and adding the new edge v1,nโ€‹v3,nv_{1,n}v_{3,n}. Obviously, G0G_{0} is nn-regular. The special case n=5n=5 is illustrated in Figureย 5 (the right graph). For notional simplicity, let us use HH instead of G0G_{0}. We claim that tโ€‹(H)=n+13t(H)=\frac{n+1}{3}. If we set S={v2,p:1โ‰คpโ‰คnโˆ’2}โˆช{v1,nโˆ’1,v3,nโˆ’1,v3,n}S=\{v_{2,p}:1\leq p\leq n-2\}\cup\{v_{1,n-1},v_{3,n-1},v_{3,n}\}, then |S|=n+1|S|=n+1 and ฯ‰โ€‹(Hโˆ–S)=3\omega(H\setminus S)=3. Hence tโ€‹(H)โ‰ค|S|ฯ‰โ€‹(Hโˆ–S)=n+13t(H)\leq\frac{|S|}{\omega(H\setminus S)}=\frac{n+1}{3}. It remains to show that tโ€‹(H)โ‰ฅn+13t(H)\geq\frac{n+1}{3}. Let SS be an arbitrary cut-set of Vโ€‹(H)V(H). If |S|โ‰ฅn+1|S|\geq n+1, then |S|ฯ‰โ€‹(Hโˆ–S)โ‰ฅ|S|ฮฑโ€‹(Hโˆ–S)โ‰ฅn+13\frac{|S|}{\omega(H\setminus S)}\geq\frac{|S|}{\alpha(H\setminus S)}\geq\frac{n+1}{3}. Assume that |S|โ‰คn|S|\leq n. According to the construction of HH, we must have ฯ‰โ€‹(Hโˆ–S)=2\omega(H\setminus S)=2. More precisely, |Sโˆฉ{v1,n,v3,n}|=1|S\cap\{v_{1,n},v_{3,n}\}|=1 and |Sโˆฉ{v1,p,v2,p,v3,p}|=1|S\cap\{v_{1,p},v_{2,p},v_{3,p}\}|=1 for all integers pp with 1โ‰คpโ‰คnโˆ’11\leq p\leq n-1. Therefore, |S|ฯ‰โ€‹(Hโˆ–S)=n2โ‰ฅn+13\frac{|S|}{\omega(H\setminus S)}=\frac{n}{2}\geq\frac{n+1}{3}. Hence the claim is proved. We are going to show that HH is minimally tโ€‹(H)t(H)-tough. Let ee be an arbitrary edge of HH. If e=v1,nโ€‹v3,ne=v_{1,n}v_{3,n}, then tโ€‹(Hโˆ’e)โ‰ค|S|ฯ‰โ€‹((Hโˆ’e)โˆ–S)=n3<n+13=tโ€‹(H)t(H-e)\leq\frac{|S|}{\omega((H-e)\setminus S)}=\frac{n}{3}<\frac{n+1}{3}=t(H), where S={v2,p:1โ‰คpโ‰คnโˆ’2}โˆช{v1,nโˆ’1,v3,nโˆ’1}S=\{v_{2,p}:1\leq p\leq n-2\}\cup\{v_{1,n-1},v_{3,n-1}\}. If e=v2,jโ€‹vi,je=v_{2,j}v_{i,j}, then tโ€‹(Hโˆ’e)โ‰ค|S|ฯ‰โ€‹((Hโˆ’e)โˆ–S)=n3<n+13=tโ€‹(H)t(H-e)\leq\frac{|S|}{\omega((H-e)\setminus S)}=\frac{n}{3}<\frac{n+1}{3}=t(H), where S={v2,p:1โ‰คpโ‰คnโˆ’1โ€‹ย andย โ€‹pโ‰ j}โˆช{v4โˆ’i,j,v1,n}S=\{v_{2,p}:1\leq p\leq n-1\text{ and }p\neq j\}\cup\{v_{4-i,j},v_{1,n}\}. If e=vi,j1โ€‹vi,j2e=v_{i,j_{1}}v_{i,j_{2}} and iโ‰ 2i\neq 2, then tโ€‹(Hโˆ’e)โ‰ค|S|ฯ‰โ€‹((Hโˆ’e)โˆ–S)=n3<n+13=tโ€‹(H)t(H-e)\leq\frac{|S|}{\omega((H-e)\setminus S)}=\frac{n}{3}<\frac{n+1}{3}=t(H), where S={vi,p:1โ‰คpโ‰คnโ€‹ย andย โ€‹pโ‰ j1,j2}โˆช{v2,j1,v}S=\{v_{i,p}:1\leq p\leq n\text{ and }p\neq j_{1},j_{2}\}\cup\{v_{2,j_{1}},v\} in which v=v2,j2v=v_{2,j_{2}} when nโˆ‰{j1,j2}n\not\in\{j_{1},j_{2}\}, and v=v4โˆ’i,nv=v_{4-i,n} when j1โ‰ j2=nj_{1}\neq j_{2}=n. If e=v2,j1โ€‹v2,j2e=v_{2,j_{1}}v_{2,j_{2}}, then tโ€‹(Hโˆ’e)โ‰ค|S|ฯ‰โ€‹((Hโˆ’e)โˆ–S)=n+24<n+13=tโ€‹(H)t(H-e)\leq\frac{|S|}{\omega((H-e)\setminus S)}=\frac{n+2}{4}<\frac{n+1}{3}=t(H), where S={v2,p:1โ‰คpโ‰คnโˆ’1โ€‹ย andย โ€‹pโ‰ j1,j2}โˆช{v1,j1,v1,j2,v3,j1,v3,j2,v1,n}S=\{v_{2,p}:1\leq p\leq n-1\text{ and }p\neq j_{1},j_{2}\}\cup\{v_{1,j_{1}},v_{1,j_{2}},v_{3,j_{1}},v_{3,j_{2}},v_{1,n}\}. Hence the proof is completed. โ–ก\Box

4 A revised conjecture

Motivated by Theoremย 3.2, one may ask whether every minimally tt-tough graph has minimum degree at most โŒˆcโ€‹tโŒ‰\lceil ct\rceil for a constant number cc. We conjecture that the answer is false (even possibly in 22-solid graphs) and put forward the following conjecture. To support this conjecture, by applying a computer search on vertex-transitive graphs on up to 3535 vertices [14], we could discover a minimally 44-tough 2424-regular (resp. 66-tough 3636-regular) 22-solid graph GG of order 7070 satisfying ฮดโ€‹(G)tโ€‹(G)=6\frac{\delta(G)}{t(G)}=6.

Conjecture 4.1

. For every positive real number cc, there is a minimally tโ€‹(G)t(G)-tough (possibly 2-solid) graph GG satisfying ฮดโ€‹(G)tโ€‹(G)โ‰ฅc\frac{\delta(G)}{t(G)}\geq c.

The structure of those graphs were not easy to draw, but to see a more clear graph example with smaller ratio, one can consider the 22-solid graph GG obtained from the circulant graph ๐’žโ€‹(39,{3,4})\mathcal{C}(39,\{3,4\}) for which ฮดโ€‹(G)tโ€‹(G)=5.2\frac{\delta(G)}{t(G)}=5.2 (more precisely, tโ€‹(G)=40/26t(G)=40/26 and ฮดโ€‹(G)=8\delta(G)=8). Note that this ratio for the left graph in Figureย 6 is 4.24.2. In addition, the main result in [15] says that this ratio for 22-solid graphs obtained from odd cycles must tend to 44. Fortunately, to compute efficiently the toughness of the ss-solid graph GG obtained from a graph HH, we only needed to consider cut-sets SS with the minimum |S|ฯ‰โ€‹(Gโˆ–S)\frac{|S|}{\omega(G\setminus S)} and |S||S| so that for every vertex vโˆˆVโ€‹(H)v\in V(H), SS either includes all copies of vv in GG or excludes all of them (together with using isomorphic properties of GG).

Refer to caption
Figure 6: Two minimally tโ€‹(G)t(G)-tough 22-solid (regular and non-regular) graphs GG with minimum degree 66 satisfying tโ€‹(G)โˆˆ{2014,2}t(G)\in\{\frac{20}{14},2\} and ฮดโ€‹(G)tโ€‹(G)โˆˆ{4.2,3}\frac{\delta(G)}{t(G)}\in\{4.2,3\}.

Finally, we pose the following question for further investigation. Note that Krieselโ€™s Conjecture says that fโ€‹(1)f(1) is finite and it must be 22 while Conjectureย 4.1 says that, regardless of fโ€‹(t)f(t) is finite or not, fโ€‹(t)t\frac{f(t)}{t} must tend to infinity when tt tends to infinity.

Problem 4.2

. Let tt be a positive real number. Is it true that there exists a positive integer fโ€‹(t)f(t) such that every minimally tt-tough graph GG satisfies ฮดโ€‹(G)โ‰คfโ€‹(t)\delta(G)\leq f(t)?

References

  • [1] G. Brinkmann and B.D McKay, Fast generation of planar graphs. MATCH Commun. Math. Comput. Chem. 58 (2007) 323โ€“357.
  • [2] H.J. Broersma, E. Engsberg, and H. Trommel, Various results on the toughness of graphs, Networks 33 (1999) 233โ€“238.
  • [3] K. Cheng, C. Li, and F. Liu, Constructions of minimally tt-tough regular graphs, (2024), arXiv:2412.12659.
  • [4] V. Chvรกtal, Tough graphs and Hamiltonian circuits, Discrete Math. 5 (1973) 215โ€“228.
  • [5] M. Hasanvand, The List Square Coloring Conjecture fails for bipartite planar graphs and their line graphs, 2211.00622v3.
  • [6] T. Kaiser, Problems from the workshop on dominating cycles, http://iti.zcu.cz /history/2003/Hajek/problems /hajek-problems.ps.
  • [7] G.Y. Katona, D. Soltรฉsz, and K. Varga, Properties of minimally tt-tough graphs, Discrete Math. 341 (2018) 221โ€“231.
  • [8] G.Y. Katona and K. Varga, Minimally toughness in special graph classes, arXiv:1802.00055, 2018.
  • [9] G.Y. Katona and H. Khan, Minimally tough chordal graphs with toughness at most 1/21/2, Discrete Math. 347 (2024) 113491.
  • [10] W. Mader, Eine Eigenschaft der Atome endlicher Graphen, Arch. Math. 22 (1971) 333โ€“336.
  • [11] H . Ma, X. Hu, and W. Yang, The structure of minimally tt-tough, 2โ€‹K22K_{2}-free graphs, Discrete Appl. Math. 346 (2024) 1โ€“9.
  • [12] H. Ma, X. Hu, and W. Yang, On the minimum degree of minimally 11-tough, triangle-free graphs and minimally 3/23/2-tough, claw-free graphs. Discrete Math. 346 (2023) 113352.
  • [13] H. Ma, X. Hu, and W. Yang, On the minimum degree of minimally tt-tough, claw-free graphs, arXiv:2311.08634.
  • [14] G. Royle and D Holt. Vertex-transitive graphs on fewer than 4848 vertices [Data set]. In Journal of Symbolic Computation 101 (2020) 51โ€“60. Zenodo.
  • [15] W. Zheng and L. Sun, Disproof of a conjecture on minimally t-tough graphs, Discrete Math. 347 (2024) 113982.