
Citation: S. Masih Ayat, S. Majid Ayat, Meysam Ghahramani. The independence number of circulant triangle-free graphs[J]. AIMS Mathematics, 2020, 5(4): 3741-3750. doi: 10.3934/math.2020242
[1] | Chunyan Qin, Xiaoxia Zhang, Liangchen Li . Z3-connectivity of graphs with independence number at most 3. AIMS Mathematics, 2023, 8(2): 3204-3209. doi: 10.3934/math.2023164 |
[2] | Igal Sason . Observations on graph invariants with the Lovász ϑ-function. AIMS Mathematics, 2024, 9(6): 15385-15468. doi: 10.3934/math.2024747 |
[3] | Wenke Zhou, Guo Chen, Hongzhi Deng, Jianhua Tu . Enumeration of dissociation sets in grid graphs. AIMS Mathematics, 2024, 9(6): 14899-14912. doi: 10.3934/math.2024721 |
[4] | Syafrizal Sy, Rinovia Simanjuntak, Tamaro Nadeak, Kiki Ariyanti Sugeng, Tulus Tulus . Distance antimagic labeling of circulant graphs. AIMS Mathematics, 2024, 9(8): 21177-21188. doi: 10.3934/math.20241028 |
[5] | Sufyan Asif, Muhammad Khalid Mahmood, Amal S. Alali, Abdullah A. Zaagan . Structures and applications of graphs arising from congruences over moduli. AIMS Mathematics, 2024, 9(8): 21786-21798. doi: 10.3934/math.20241059 |
[6] | Xia Hong, Wei Feng . Completely independent spanning trees in some Cartesian product graphs. AIMS Mathematics, 2023, 8(7): 16127-16136. doi: 10.3934/math.2023823 |
[7] | Zongrong Qin, Dingjun Lou . The k-subconnectedness of planar graphs. AIMS Mathematics, 2021, 6(6): 5762-5771. doi: 10.3934/math.2021340 |
[8] | Mohamed Basher . On the reflexive edge strength of the circulant graphs. AIMS Mathematics, 2021, 6(9): 9342-9365. doi: 10.3934/math.2021543 |
[9] | Yunfeng Tang, Huixin Yin, Miaomiao Han . Star edge coloring of K2,t-free planar graphs. AIMS Mathematics, 2023, 8(6): 13154-13161. doi: 10.3934/math.2023664 |
[10] | Miao Fu, Yuqin Zhang . Results on monochromatic vertex disconnection of graphs. AIMS Mathematics, 2023, 8(6): 13219-13240. doi: 10.3934/math.2023668 |
Erdös proposed the tenth problem in the Third Conference on discrete mathematics in Clemson University as follows: Whether there exists a regular triangle-free graph Gn on n vertices where the largest independent sets have a size equal to the degree of Gn [5].
Bauer determined all n,r,l such that there exists a r-regular kl-free graph on n vertices [1]. Bauer proposed the following structures: Suppose n is even. Let the vertices in the two parts be labelled as:
1,2,…,n/2 and 1′,2′,…,(n/2)′. |
Now connect each vertex i to (i+k)′ mod n/2 for 0⩽k⩽r−1. In this case, α(G)=n/2 and the independence ratio is 1/2. If n is odd, then r must be even and 5r/2⩽n⩽3r−1. Form the graph G as in Figure 1. Each of parts A,B,C,D,E are independent sets. The size of them are: |B|=|C|=r/2,|A|=3r−n,|D|=|E|=n−2r. Each vertex of B and C is adjacent to each vertex of A. Then connect each vertex of B to every vertex of D and each vertex of C to every vertex of E. Since |D|=|E|=n−2r≥r/2, join any vertex of D to some vertices of E to make a r/2-regular bipartite graph. Note that B∪C is an independent set of size r and B∪E is an independent set of size (n−3r/2). If n=5r/2, then α(G)=r. Otherwise, α(G)>r and the independence ratio of this graph is at least r/(3r−1) [1].
Sidorenko set some 2k-regular triangle-free graphs such that their independence number are equal to their degree. It is obvious that the maximum degree of vertices is a lower bound for the independence number in triangle-free graphs. Sidorenko graphs are circulant graphs [6]. Brandt investigated the class S of triangle-free graphs where the neighbourhood of vertices are maximal independent sets [2]. The Sidorenko graphs are in S [2].
Punnim showed that the values of independence number of regular graphs cover a line segment of integer numbers, i.e., for all n,r there exist a⩽b∈N such that the independence number of any r-regular graph on n vertices is in [a,b] and for every integer c∈[a,b], there exists a r-regular graph on n vertices with independence number equals to c. This interval is called the independence interval [4].
In this current work, the independence interval of 2-regular graphs are determined in the second section. The maximum independent set for circulant 3-regular triangle-free graphs will be determined in Section 2. Moreover, a subinterval of independence interval for 3-regular graphs is determined in this section. Finally, it will be proved that the independence ratio of 3-regular triangle-free graphs is at least 3/8. Note that Staton proved that the independence ratio for 3-regular triangle-free graphs is at least 5/14 [7]. Heckman and Thomas give a new proof for this theorem and they design a linear-time algorithm to find the maximum independent set for cubic graphs [3].
On the last section, some class of regular triangle-free graphs with independence number equals to degree for odd degree will be determined. In this case the lower and upper bounds are close to the even case which determined by Sidorenko [6].
Let G be a graph. The number of vertices of G is denoted by n(G). The independence number of G is the maximum size of an independent set of G and it is denoted by α(G). The independence ratio is defined by i(G)=α(G)n(G).
Definition 2.1. Suppose S⊂{1,…,n−1} is such that
x∈S⟺n−x∈S. |
The circulant graph produced by S is defined as follows. The vertex set is {0,1,…,n−1} and x,y are adjacent if and only if x−y∈S. Trivially, if x<y, it means n−(x−y)∈S. The circulant graph produced by S is denoted by G[S]. The circulant graph G[S] is triangle-free if and only if
∀x,y∈S,x−y∉S. |
If r=2, we could determine the minimum independence number, as follows.
Theorem 2.2. Let G be a 2-regular triangle-free circulant graph. If the set
X={d;d≠1,d|n,d≠n3,nd is odd} | (2.1) |
is non-empty, then minα(G)=n−max(x)2, else minα(G)=n2 and the independence ratio is at least 2/5.
Proof. Let G=G[S] be a triangle-free circulant graph and S={t,n−t}. Since G is a 2-regular graph, G is a union of disjoint circles. If gcd(n,t)=1, then G[S] is a Hamiltonian circle and α(G[S])=⌊n/2⌋.
If gcd(n,t)=d≠1, then G[S]≃G[d,n−d] and we have:
n−d−d≠d⟹n≠3d,d|n. |
But the graph G[S] is disjoint union of d circles of size n/d and α(G[S])=d⌊n/2d⌋. Since n=dk, we have two cases:
(1). k=2s is an even integer. Then α(G[S])=d⌊n/2d⌋=ds=n/2.
(2). k=2s+1 is an odd integer. Then α(G[S])=d⌊n/2d⌋=ds=(n−d)/2.
Therefore, if X={d;d≠1,d|n,d≠n/3,n/d is odd } is a non-empty set, then
min(α(G[S]))=(n−max(X))/2. |
Otherwise, min(α(G[S]))=n/2.
Since nd is an odd number and d≠n/3, then nd⩾5 and consequently d⩽n5. Therefore, the independence ratio is at least
n−n52n=25. |
Note that every 2-regular graph G is a disjoint union of circles, i.e., G=C1∪⋯∪Ck. Thus α(G)=∑α(Ci). Suppose the size of Ci is ni. If ni is odd, α(Ci)=(ni−1)/2, otherwise α(Ci)=ni/2. Suppose G=C1∪⋯∪Ct∪Ck and C1,…,Ct are odd circles and Ct+1,˙,Ck are even circles. Thus:
α(G)=(n1−1)/2+⋯+(nt−1)/2+nt+1/2+⋯+nk/2=n/2−t/2. |
The independence ratio in this case is 1/2−t/2n. Since t⩽n3, then i(G)⩾13. In fact, the max/min of independence number is determined by the size of t as follows.
Proposition 2.3. The independence interval for 2-regular graphs is [⌈n3⌉,⌊n2⌋]. In fact, if n>3:
minα(G)=⌈n3⌉={k;n=3k; G is union of k triangles;k+1;n=3k+1;G is union of k−1 triangles and a rectangle;k+1;n=3k+2;G is union of k−1 triangles and a pentagon. |
The max/min of independence number of 2-regular triangle-free graphs is determined as follows.
Proposition 2.4. The independence interval for 2-regular triangle-free graphs is [⌈2n5⌉,⌊n2⌋]. In fact, the minimum independence number for 2-regular triangle-free graphs is as follows (n⩾5):
minα(G)=⌈2n5⌉={2k;n=5k;G is union of k pentagons;2k+1;n=5k+1;G is union of k−1 pentagons and a hexagon;2k+1;n=5k+2;G is union of k−1 pentagons and a heptagon;2k+2;n=5k+3;G is union of k−1 pentagons and 2 rectangles;2k+2;n=5k+4;G is union of k pentagons and a rectangle. |
If r=3, n must be an even number and n/2∈S. Note that 3-regular circulant graph has a 2-regular circulant as an induced subetaaph. In the rest of this section, the minimum independence number of 3-regular triangle-free circulant graphs is determined.
Lemma 2.5. If S={d,n/2,n−d} and G[S] is a triangle-free circulant graph and gcd(n,d)=1, then
α(G[S])={n/2;n/2 is odd n/2−1;n/2 is even. | (2.2) |
Proof. Since G[S] is a triangle free circulant graph, we have:
n/2−d≠d⟹d≠n/4; | (2.3) |
n−d−d≠d⟹d≠n/3. | (2.4) |
Suppose gcd(n,d)=1, then G[S] has a Hamiltonian circle. If kdn≡n2 and k⩽n, then k=n2 and therefore G[S]≅G[{1,n/2,n−1}]. Note that α(G[S])⩽n/2.
Suppose n/2 is an odd integer. Since the numbers {1,n/2,n−1} are odd and any odd vertex is adjacent to 3 even vertices, then the set of all odd vertices is an independent set in G[S] and α(G[S])=n/2.
Otherwise, suppose n/2=2k is an even integer. Then for every odd number 2l+1<n/2, the number n/2−(2l+1)=2(k−l)−1 is an odd number less than n/2. Thus the set of odd integers is not an independent set in G[S]. Nevertheless, the set {1,…,2k−1,2k+2,…,4k−2} is an independent set of size 2k−1. Now suppose there exists an independent set I of size 2k. Since {a,n/2+a} is an edge in G[S] for arbitrary integer 0⩽a⩽n/2−1, then I have at most exactly one element from each of these pairs. Without less of generality, suppose I has 0. Then, I must have n/2+1, and so on. Therefore, one could see
I={0,n/2+1,2,n/2+3,…,n−1} |
and it is a contradiction, because n−1 is adjacent to 0.
Theorem 2.6. Suppose n is an even number and S={d,n/2,n−d;d|n}. If G[S] is a triangle-free circulant graph, then
α(G[S])={n/2−d;n/2d is even n/2;n/2d is odd (n−d)/2;n/d is odd |
Proof. Note that d≠n/2,n/3,n/4. Then we have two cases:
ⅰ) If n/d is even, then n=2td and G0={0,d,…,(2t−1)d}. Thus n/2=td is one of the vertices of G0. Note that G0 is a n/d-cycle with addition edges {kd,(k+t)d} for k=0,1,…t−1. If n/2d is even, α(G0)=n2d−1=t−1. But G[S] is d copies of G0. Therefore, α(G[s])=d(t−1)=n/2−d. Otherwise, if n/2d is odd, α(G0)=n2d and α(G[s])=n/2.
ⅱ) If n/d is odd, then n=(2t+1)d. Thus n/2=td+(d/2). Note that d is an even divisor of n. Let Gi={i,i+d,…,i+2td}, for i=0,…,d−1. The graph G0 is a n/d-cycle and every vertex of Gi is adjacent to one and only one vertex of Gi+d2 for i=0,…,d/2−1. But α(Gi)=⌊n2d⌋=t and therefore α(G[s])=d2(2t)=(n−d)/2.
Theorem 2.7. Suppose n is an even number and 1<d<n. If gcd(n,d)=t≠1, then
G[d,n/2,n−d]≅G[t,n/2,n−t]. |
Proof. Since gcd(n,d)=t, then G[d,n−d]≅G[t,n−t] and this graph is t disjoint cycles of order n/t. We have two cases:
Case 1. t|n/2. Then n=2kt and d=d′t. Therefore, d′ is an odd integer. Suppose V(G0)={0,d,…,(n/t−1)d}={0,t,…,(n/t−1)t}. Thus n/2∈V(G0). G0 is a 3-regular triangle-free circulant graph on n/t vertices and S={1,n/2t,n/t−1} and
α(G0)={n/2t;n/2t is odd n/2t−1;n/2t is even | (2.5) |
Thus we have α(G[S])=tα(G0) and
α(G[S])={n/2;n/2 is odd n/2−t;n/2 is even | (2.6) |
Case 2. t∤n/2. Thus n/2∉V(G0) and there exists i<t such that n/2∈V(Gi)={i,i+d,…,i+(n/t−1)d}. Note that n=kt and t∤n/2. Therefore, k is an odd number and t is an even. But 0∈V(G0) is adjacent to n/2∈V(Gi) and for every j=0,1,…,n/t−1, jd is adjacent to n/2+jd∈V(Gi). In fact, we have t/2 copies of graph such as the graph Figure 2 on 2n/t vertices.
The maximum independent set in these graphs has order n/t−1 as such as: {0,n/2+d,2d,n/2+3d,…,(n/t−3)d,n/2+(n/t−2)d}. This independent set is shown by black vertices in Figure 2. Then α(G[S]))=(n−t)/2.
We summarize Theorems 2.6,2.7 and Lemma 2.5 as follows:
Theorem 2.8. Suppose n is an even number and 1⩽d<n. If gcd(n,d)=t and S={d,n/2,n−d}, then
α(G[S])={n/2−t;n/2t is even n/2;n/2t is odd (n−t)/2;n/t is odd |
By Theorem 2.7, one could determine the independence number for every 3-regular circulant triangle-free graphs. The minimum independence number for these graphs are determined for n⩽44 as follows:
\boldsymbol n | \boldsymbol d | \boldsymbol \alpha | \boldsymbol n | \boldsymbol d | \boldsymbol \alpha | \boldsymbol n | \boldsymbol d | \boldsymbol \alpha | \boldsymbol n | \boldsymbol d | \boldsymbol \alpha | |||
6 | 1 | 3 | 16 | 2 | 6 | 26 | 2 | 12 | 36 | 3 | 15 | |||
8 | 1 | 3 | 18 | 2 | 8 | 28 | 4 | 12 | 38 | 2 | 18 | |||
10 | 2 | 4 | 20 | 4 | 8 | 30 | 6 | 12 | 40 | 5 | 15 | |||
12 | 1 | 5 | 22 | 2 | 10 | 32 | 4 | 12 | 42 | 6 | 18 | |||
14 | 2 | 6 | 24 | 3 | 9 | 34 | 2 | 16 | 44 | 4 | 20 |
Corollary 2.9. Suppose n = 2p = 2(2k+1) and p\geqslant 5 is a prime number and G[S] is a 3 -regular triangle-free circulant graph. Then \min\alpha(G[S]) = p-1 and
i(G[S]) = \frac{1}{2}-\frac{1}{2p}\geqslant \frac{2}{5}. |
Corollary 2.10. The independence ratio of circulant 3 -regular triangle-free graphs is at least \frac{3}{8} and this lower bound is sharp, i.e. there exists a 3 -regular triangle-free circulant graph on 8k vertices such that its independence number equals to 3k , for all k .
Proof. Note that the minimum independence number is \frac{n}{2}-d and d must divide n and \frac{n}{2d} is even. Since d\neq n/2,n/3,n/4 , then d\leqslant\frac{n}{8} . Thus \alpha(G)\geqslant\frac{n}{2}-\frac{n}{8} = \frac{3n}{8} and therefore the independence ratio is at least \frac{3}{8} .
Suppose d\vert 8k and n/d is an even integer. Then the maximum value of d is k . Therefore S = \{k,4k,7k\} and \alpha(G[S]) = 4k-k = 3k and consequently the independence ratio is 3/8 .
In this section, the circulant triangle-free graphs with independence number equal to their degree are investigated. These graphs are denoted by \mathcal{A} -graph in this work. The first integer n such that there exists a r -regular \mathcal{A} -graph with n vertices is obviouse:
Proposition 3.1. The least number n such that there exists a r -regular \mathcal{A} -graph on n vertices is 2r .
Proof. Suppose G is a triangle-free r -regular graph on n vertices, then n is at least 2r .
Suppose n = 2r, V = \{0,1,\dots,2r-1\} and S = \{2t-1;0\leqslant t\leqslant r\} . Then G[S] = K_{r,r} is a triangle-free circulant graph. The set of odd integers is an independent set of size r . If I is a set of size r+1 , then there exist two integers in I with different parity like a = 2k,b = 2l+1 . Since a-b\in S , then a,b are adjacent vertices in G[S] and it is a contradiction.
The above lower bound is trivial. But there does not exist r -regular \mathcal{A} -graphs for all n\geqslant 2r , as follows.
Proposition 3.2. There does not exist any 2r -regular circulant triangle-free graph on 4r+1 vertices.
Proof. Suppose S = \{x_1<\dots<x_r\leqslant 2r<x_{r+1}<\dots<x_{2r}\} . Since the graph G[S] is triangle-free, then the set \{x_r-x_{r-1},x_r-x_{r-2},\dots,x_r-x_1\} is disjoint from \{x_1,\dots,x_r\} . But \{x_r-x_{r-1},x_r-x_{r-2},\dots,x_r-x_1\}\cup \{x_1,\dots,x_r\}\subset\{1,\dots,2r\} . Therefore, we have two cases:
(1). x_r = 2r . Then 4r+1-x_r = 2r+1\in S . Therefore, 1 = 2r+1-2r\not\in S . Thus
1 \lt x_1 \lt x_2 \lt \dots \lt x_r = 2r, |
and
2r+1-x_r,2r+1-x_{r-1},\dots ,2r+1-x_1\not\in S. |
Moreover, the set \{2r-x_{r-1},2r-x_{r-2},\dots,2r-x_1\}\not\in S , too. Thus
\{2r-x_{r-1},2r-x_{r-2},\dots,2r-x_1\}\subseteq \{1,2r+1-x_r,2r+1-x_{r-1},\dots ,2r+1-x_1\}. |
Thus for every i = 1,2,\dots,r-1 , there exists j = 1,\dots,r such that 2r-x_i = 2r+1-x_j . Therefore, x_j = x_i+1 . Thus 2r-x_i = 2r+1-x_{i+1} , for all i and S = \{r+1,\dots ,2r-1,2r,2r+1,\dots ,3r\} . But 3r-(r+1) = 2r-1\in S and it is a contradiction.
(2). x_r = 2r-1 . Thus 4r+1-x_r = 2r+2\in S and 2r+2-(2r-1) = 3\not\in S . Therefore x_1<x_2<\dots <x_r = 2r-1\in S and 2r+2-(2r-1) = 3<2r+2-x_{r-1}<\dots<2r+2-x_1\leqslant 2r+1\not\in S . If x_1 = 1 , then 2\not\in S and \{2\}\not\in\{1 = x_1,x_2,\dots,x_r = 2r-1\}\dot{\cup}\{3,2r+2-x_{r-1},\dots,2r\} . Therefore, there exists x_i\in S such that 2r = 2r+2-x_i . Thus x_i = 2 and it is a contradiction. If x_1 = 2 , then \{2 = x_1<x_2<\dots<x_r = 2r\}\dot{\cup}\{3<2r+2-x_{r-1}<\dots<2r\} is a set of size 2r and this set is a subset of \{1,2,\dots,2r\} . It is a contradiction. Thus x_1\neq 1,2,3 . Then x_1\geqslant 4 and 2r+2-x_1\leqslant 2r-2 . Therefore the set
\{x_1 \lt x_2 \lt \dots \lt x_r = 2r-1\}\cup\{3 \lt \dots \lt 2r+2-x_1\leqslant 2r-2\} |
is a set of size 2r and it is a subset of \{3,4,\dots,2r-1\} . It is a contradiction too.
There exists similar result for circulant graphs with odd regularity:
Proposition 3.3. Let r>1 . There does not exist any (2r+1) -regular circulant triangle-free graph on 4r+4 vertices.
Proof. Suppose S = \{x_1<x_2<\dots<x_r<x_{r+1} = 2r+2<x_{r+2}<\dots<x_{2r+1}\} and G[S] is a (2r+1) -regular circulant triangle-free graph. Therefore 2r+2-x_1,2r+2-x_2,\dots,2r+2-x_r\not\in S .
Note that r+1\not \in S and it is not adjacent to 2r+2 , too. Moreover, x_r>r+1 . Otherwise, S = \{1,2,\dots,r\} and G[S] is not a triangle-free graph. If x_r<2r , then the set \{x_r+1,\dots,2r+1\} is disjoint from S . Thus these elements are in the form 2r+2-x_1,2r+2-x_2,\dots . Thus x_1 = 1,x_2 and G[S] is not triangle-free graph.
Now suppose x_r = 2r . Then 2r+1\not\in S and 2r+1 = 2r+2-x_1 . Thus x_1 = 1 . Since G[S] is triangle-free graph, 2 = (2r+2)-2r\not\in S . Moreover, (4r+4)-2r = 2r+4\in S . Similary, since G[S] is triangle-free graph, 4 = (2r+4)-2r\not\in S .
If 4 = r+1 , then r = 3 and 2,4,5\not\in S and 1,3,6\in S and G[S] is not triangle-free graph. If 4\neq r+1 , then 4 = (2r+2)-x_{r-1} and x_{r-1} = 2r-2 and x_{r+3} = (4r+4)-(2r-2) = 2r+6 . Since G[S] is triangle-free graph, then 6,8\not\in S . One could continue this process. If r = 2t-1 is an odd element, then 2r,2r-2,2r-4,\dots,r+3\in S . But 2r+6,r+3\in S , then G[S] is not a triangle-free graph.
If r is an even element, then 2,4,6,\dots\not\in S , but 2r\in S and it is a contradiction.
Now the main family of \mathcal{A} -graphs are presented as follows.
Definition 3.4. Suppose S = \{\pm k,\pm(k+1),\dots,\pm(2k-1)\} . The graph G[S] is denoted by G_{n,k} .
The graphs G_{n,k} are triangle-free graph if and only if n\geqslant 6k-2 .
Theorem 3.5. (Sidorenko's Theorem)- If 6k-2\leqslant n\leqslant 8k-3 , then G_{n,k} is a 2k -regular \mathcal{A} -graph, [6].
Note that if one add n/2 to S , then G[S] has triangle, because n/2\in S-S = [n-4k+2,n-2k] . Thus for odd degree, we must define new structures:
Theorem 3.6. Suppose n = 8k , let S = \{\pm 1,\pm 3,\cdots\pm (2k-1)\}\cup\{4k\} . Then G[S] is a (2k+1) -regular \mathcal{A} -graph on 8k vertices.
Proof. Since 4k-(2k-1) = 2k+1 , then (8k-(2k-1))-4k = 2k+1 and it is obviously triangle-free graph. Let V = \{0,1,\dots,8k-1\} .
Suppose 0<t_1<\dots<t_{2k+2}<8k is an independent set of size 2k+2 . If all t_i 's are odd integers. Suppose t_i<4k<t_{i+1} . Note that t_j-t_i are even, for all j>i . Thus t_j is adjacent to t_i if and only if t_j-t_i = 4k . Since there are 2k odd elements less than 4k , one must choose at most one element from each \{t_i,t_i+4k\} , for all t_i<4k . It is a contradiction.
If all t_i 's are even, there are at most 2k-1 even numbers less than 4k . if t_j>t_i , t_j is adjacent to t_i if and only if t_j-t_i = 4k . One must choose at most one element from each \{t_i,t_i+4k\} , for all t_i<4k . But there are 2k-1 set of \{t_i,t_i+4k\} and it is a contradiction.
Now suppose some t_i 's are odd and some of them are even. Let s_i = t_{i+1}-t_i , i = 1,\dots,2k+1 , s_{2k+2} = 8k-t_{2k+2}+t_1 . Then s_1+\dots+s_{2k+2} = 8k . There are some odd s_i and some even s_i . Since the sum of s_i 's are even, the number of odd s_i 's are even.
Let the number of odd s_i 's is r . If s_i is an odd element, then 2k+1\leqslant s_i<6k+1 . But s_1+\dots+s_{2k+2} = 8k , then r<4 . Thus we have two cases. If r = 0 is solved before.Now suppose r = 2 . Then s_i\geqslant 2, s_1+\dots+s_{2k+2}\geqslant (2k)\times 2+2\times (2k+1) = 4k+4k+2 = 8k+2 and it is a contradiction.
The above theorem is an upper bound for odd regularity. A non-trivial lower bound for these numbers is as follows:
Theorem 3.7. Suppose n = 6k+2 and S = \{ 3t+1; 0\leqslant t\leqslant 2k\} . Then G[S] is a (2k+1) -regular \mathcal{A} -graph on 6k+2 vertices.
Proof. It is obviouse that G[S] is a triangle-free circulant graph on 6k+2 vertices.
Now suppose we have an independent set I of size 2k+2 . Let 0\in I . Thus the other elements of I are elements of S_0 = \{3t; 1\leqslant t\leqslant 2k\} and S_2 = \{3t+2; 0\leqslant t\leqslant 2k-1\} . But |S_0| = |S_2| = 2k . Then I\cap S_0, I\cap S_2\neq \emptyset . Suppose a = 3t\in S_0, b = 3s+2\in S_2 .
We have two cases: If a>b , then a-b = 3t-3s-2\in S . Otherwise, every element of I\cap S_0 is less than every element of I\cap S_2 . Suppose 3s is the greatest element of I\cap S_0 , then |I\cap S_0|\leqslant s and for all element 3t+2\in I\cap S_2 , 3t+2>3s . Therefore t\geqslant s and I\cap S_2\subseteq\{3t+2; s\leqslant 2k-1\} . Thus |I\cap S_2|\leqslant 2k-s and |(I\cap S_0)\cup(I\cap S_2)|\leqslant 2k . It is a contradiction.
According to the above theorems, we have the following conjecture:
Conjecture 3.8. Suppose 6k+2\leqslant n\leqslant 8k and n is an even number. There exists a (2k+1) -regular \mathcal{A} -graph on n vertices.
Some of the well-known examples for (s,t,n) -graphs are circulant graphs. Thus t -regular triangle-free circulant graphs must be some good lower bounds for R(3,t+1) if their independence number is t . Thus it is better to find some \mathcal{A} -graphs with the maximum number of vertices. The algebraic properties of these graphs will be helpful to find their minimum independence number.
The authors declare no conflict of interest.
[1] |
D. Bauer, Regular Kn-free graphs, J. Combin. Theory Ser. B., 35 (1983), 193-200. doi: 10.1016/0095-8956(83)90071-0
![]() |
[2] |
S. Brandt, Triangle-free graphs whose independence number equals the degree, Discrete Math., 310 (2010), 662-669. doi: 10.1016/j.disc.2009.05.021
![]() |
[3] |
C. Heckman, C. Thomas, A new proof of the independence ratio of triangle-free cubic graphs, Discrete Math., 233 (2001), 233-237. doi: 10.1016/S0012-365X(00)00242-9
![]() |
[4] |
N. Punnim, The clique numbers of regular graphs, Graph. Combinator., 18 (2002), 781-785. doi: 10.1007/s003730200064
![]() |
[5] | R. D. Ringeisen, F. S. Roberts, Applications of Discrete Mathematics, In: Proceedings of the Third SIAM Conference on Discrete Mathematics held at Clemson University, SIAM, Philadelphia, PA, 1988. |
[6] |
A. F. Sidorenko, Triangle-free regular graphs, Discrete Math., 91 (1991), 215-217. doi: 10.1016/0012-365X(91)90114-H
![]() |
[7] |
W. Staton, Some Ramsey-type numbers and the independence ratio, T. Am. Math. Soc., 256 (1979), 353-370. doi: 10.1090/S0002-9947-1979-0546922-6
![]() |
1. | Ganesh Gandal, R Mary Jeya Jothi, Narayan Phadatare, On very strongly perfect Cartesian product graphs, 2022, 7, 2473-6988, 2634, 10.3934/math.2022148 |
\boldsymbol n | \boldsymbol d | \boldsymbol \alpha | \boldsymbol n | \boldsymbol d | \boldsymbol \alpha | \boldsymbol n | \boldsymbol d | \boldsymbol \alpha | \boldsymbol n | \boldsymbol d | \boldsymbol \alpha | |||
6 | 1 | 3 | 16 | 2 | 6 | 26 | 2 | 12 | 36 | 3 | 15 | |||
8 | 1 | 3 | 18 | 2 | 8 | 28 | 4 | 12 | 38 | 2 | 18 | |||
10 | 2 | 4 | 20 | 4 | 8 | 30 | 6 | 12 | 40 | 5 | 15 | |||
12 | 1 | 5 | 22 | 2 | 10 | 32 | 4 | 12 | 42 | 6 | 18 | |||
14 | 2 | 6 | 24 | 3 | 9 | 34 | 2 | 16 | 44 | 4 | 20 |