Research article

Number of maximal 2-component independent sets in forests

  • Received: 03 January 2022 Revised: 03 April 2022 Accepted: 11 April 2022 Published: 20 May 2022
  • MSC : 05C30, 05C69

  • Let G=(V(G),E(G)) be a graph. For a positive integer k, we call SV(G) a k-component independent set of G if each component of G[S] has order at most k. Moreover, S is maximal if there does not exist a k-component independent set S of G such that SS and |S|<|S|. A maximal k-component independent set of a graph G is denoted briefly by Mk-CIS. We use tk(G) to denote the number of Mk-CISs of a graph G. In this paper, we show that for a forest G of order n,

    t2(G){3n3,ifn0 (mod 3)andn3,43n43,ifn1 (mod 3)andn4,5,ifn=5,423n83,ifn2 (mod 3)andn8,

    with equality if and only if GFn, where

    Fn{n3P3,ifn0 (mod 3)andn3,n43P3K1,3,ifn1 (mod 3)andn4,K1,4,ifn=5,n83P32K1,3,ifn2 (mod 3)andn8.

    Citation: Shuting Cheng, Baoyindureng Wu. Number of maximal 2-component independent sets in forests[J]. AIMS Mathematics, 2022, 7(7): 13537-13562. doi: 10.3934/math.2022748

    Related Papers:

    [1] Chang Wan, Zehui Shao, Nasrin Dehgardi, Rana Khoeilar, Marzieh Soroudi, Asfand Fahad . Mixed domination and 2-independence in trees. AIMS Mathematics, 2020, 5(6): 5564-5571. doi: 10.3934/math.2020357
    [2] Xiaoling Zhou, Chao Yang, Weihua He . The linear $ k $-arboricity of symmetric directed trees. AIMS Mathematics, 2022, 7(2): 1603-1614. doi: 10.3934/math.2022093
    [3] 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
    [4] 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
    [5] Xiaoling Zhou, Chao Yang, Weihua He . The linear $ k $-arboricity of digraphs. AIMS Mathematics, 2022, 7(3): 4137-4152. doi: 10.3934/math.2022229
    [6] S. Masih Ayat, S. Majid Ayat, Meysam Ghahramani . The independence number of circulant triangle-free graphs. AIMS Mathematics, 2020, 5(4): 3741-3750. doi: 10.3934/math.2020242
    [7] Yassine Sabbar, Aeshah A. Raezah . Threshold analysis of an algae-zooplankton model incorporating general interaction rates and nonlinear independent stochastic components. AIMS Mathematics, 2024, 9(7): 18211-18235. doi: 10.3934/math.2024889
    [8] Shuangyan Zhao, Jiachao Wu . An efficient algorithm of fuzzy reinstatement labelling. AIMS Mathematics, 2022, 7(6): 11165-11187. doi: 10.3934/math.2022625
    [9] Chunyan Qin, Xiaoxia Zhang, Liangchen Li . $ Z_{3} $-connectivity of graphs with independence number at most 3. AIMS Mathematics, 2023, 8(2): 3204-3209. doi: 10.3934/math.2023164
    [10] Erdal Bayram, Gülşah Çelik, Mustafa Gezek . An advanced encryption system based on soft sets. AIMS Mathematics, 2024, 9(11): 32232-32256. doi: 10.3934/math.20241547
  • Let G=(V(G),E(G)) be a graph. For a positive integer k, we call SV(G) a k-component independent set of G if each component of G[S] has order at most k. Moreover, S is maximal if there does not exist a k-component independent set S of G such that SS and |S|<|S|. A maximal k-component independent set of a graph G is denoted briefly by Mk-CIS. We use tk(G) to denote the number of Mk-CISs of a graph G. In this paper, we show that for a forest G of order n,

    t2(G){3n3,ifn0 (mod 3)andn3,43n43,ifn1 (mod 3)andn4,5,ifn=5,423n83,ifn2 (mod 3)andn8,

    with equality if and only if GFn, where

    Fn{n3P3,ifn0 (mod 3)andn3,n43P3K1,3,ifn1 (mod 3)andn4,K1,4,ifn=5,n83P32K1,3,ifn2 (mod 3)andn8.



    Let G=(V(G),E(G)) be a graph. A set SV(G) is called an independent set of G if no two vertices of S are adjacent in G. A maximal independent set is an independent set that is not a proper subset of any other independent set. Let k be a positive integer. We call S a k-component independent set of G if each component of G[S] has order at most k. Clearly, the 1-component independent sets are the usual independent sets. A k-component independent set is maximal (maximum) if the set cannot be extended to a larger k-component independent set (if no k-component independent set of G has larger cardinality). A maximal k-component independent set of a graph G is denoted briefly by Mk-CIS. We use tk(G) to denote the number of Mk-CISs of G.

    In 1986, Wilf [12] proved that the maximum number of maximal independent sets for a tree of order n is 2n12 if n is odd and 2n21+1 if n2 is even. In 1988, Sagan [9] gave a simple graph-theoretical proof and characterized all extremal trees. In 1991, Zito [15] determined that the maximum number of maximum independent sets for a tree of order n is 2n32 if n>1 is odd and 2n22+1 if n is even, and also characterized all extremal trees. In 1993, Hujter and Tuza [4] proved that the maximal number of maximal independent sets in triangle-free graphs is at most 2n2 if n4 is even and 52n52 if n5 is odd, and characterized the extremal graphs. The number of the maximal independent sets on some classes of graphs were also studied in [5,6,10,13].

    In 2021, Tu, Zhang and Shi [11] showed that the maximum number of maximum 2-component independent sets in a tree of order n is 3n31+n3+1 if n0 (mod3), 3n131+1 if n1 (mod3), and 3n231 if n2 (mod3), and also characterized the extremal graphs.

    In 1981, Yannakakis [14] proved that the problem of computing the number of maximum 2-component independent sets for bipartite graphs is NP-complete. The complexity of the problem on some special families of graphs were studied in [1,2,7,8].

    In this paper, we establish a sharp upper bound for t2(G) of a forest G of order n and characterize all forests achieving the upper bound.

    Let G be a graph and v a vertex in G. The neighborhood NG(v) is the set of vertices adjacent to v and the closed neighborhood NG[v] is NG(v){v}. In the sequel, we use t(G) to present t2(G) for simplicity and S denotes an M2-CIS of a tree T under consideration.

    Theorem 2.1. For any forest F of order n3, t(F)f(n), where

    f(n)={3n3,ifn0 (mod 3)andn3,43n43,ifn1 (mod 3)andn4,5,ifn=5,423n83,ifn2 (mod 3)andn8,

    with equality if and only if

    Fn{n3P3,ifn0 (mod 3)andn3,n43P3K1,3,ifn1 (mod 3)andn4,K1,4,ifn=5,n83P32K1,3,ifn2 (mod 3)andn8.

    Lemma 2.2. (Cheng, Wu [3])Let n and k be two integers with nk+12.For any tree T of order n, there exists a vertex v such thatTv has d(v)1 components, each of which has order at most k, but the sum of theirorder is at least k.In particular, every nontrivial tree T has a vertex v such that all its neighbors but one are leaves.

    Lemma 2.3. For any positive integer n1,

    t(K1,n1)={1,if1n2,n,ifn3.

    We define five special trees, denoted by Ti for each i{1,,5}:

    T1 is a tree of order n obtained from K1,3 by subdividing an edge of K1,3n4 times, where 5n9.

    T2 is obtained from 2P4P3 by adding edges connecting a leaf of each copy of P4 to a leaf x of P3.

    T3 is obtained from (K1,3P4)P3 by adding edges connecting a leaf of K1,3 and P4 to a leaf x of P3.

    T4 is obtained from aK1,3P3 by adding edges connecting a leaf of each copy of K1,3 to a leaf x of P3 for an integer a2.

    T5 is obtained from bK1,3 by adding edges connecting a leaf of each copy of K1,3 to a fixed vertex x for an integer b2, as shown in Figure 1.

    Figure 1.  Ti, i{2,,5}.

    Lemma 2.4. t(Ti)t(Fn) for each i{1,,5}.

    Proof. By a straightforward calculation,

    t(T1)={4<5=t(K1,4)=t(F5),ifn=5,6<32=t(2P3)=t(F6),ifn=6,10<43=t(P3K1,3)=t(F7),ifn=7,13<42=t(2K1,3)=t(F8),ifn=8,17<33=t(3P3)=t(F9),ifn=9.

    Obviously, |V(T2)|=|V(T3)|=11, t(T2)=28<423=t(2K1,3P3)=t(F11), and t(T3)=31<423=t(2K1,3P3)=t(F11).

    Note that |V(T4)|=4a+3. Observe that for an M2-CIS S of T4, either xS or xS with dT[S](x)1. Let us define t0x=|{S:dT[S](x)=0}|=2a, t1x=|{S:dT[S](x)=1}|=(a+3)3a1, tˉx=|{S:xS}|=4a. Thus, t(T4)=t0x+t1x+tˉx=4a+(a+3)3a1+2a. We consider three cases in terms of the modularity of a (mod3).

    If a=3s, s1, then |V(T4)|=12s+3 and t(F12s+3)=34s+1. Moreover, since 43s34s and (s+1)33s+23s234s for any s1, it follows that for any s1,

    t(T4)=43s+(s+1)33s+23s34s+234s=34s+1=t(F12s+3).

    If a=3s+1, s1, then |V(T4)|=12s+7 and t(F12s+7)=434s+1. Moreover, since 43s+1434s and (3s+4)33s+23s+1834s for any s1, it follows that for any s1,

    t(T4)=43s+1+(3s+4)33s+23s+1434s+834s=434s+1=t(F12s+7).

    If a=3s+2, s0, then |V(T4)|=12s+11 and t(F12s+11)=4234s+1. Moreover, since 43s+24234s and (3s+5)33s+1+23s+23234s for any s0, it follows that for any s0,

    t(T4)=43s+2+(3s+5)33s+1+23s+24234s+3234s=4234s+1=t(F12s+11).

    Note that |V(T5)|=4b+1 and t(T5)=4b+b3b1b2b1. We consider three cases in terms of the modularity of b (mod3).

    If b=3s, s1, then |V(T5)|=12s+1 and t(F12s+1)=434s1. Moreover, since 43s34s and s33s34s1 for any s1, it follows that for any s1,

    t(T5)=43s+s33s3s23s134s+34s1=434s1=t(F12s+1).

    If b=3s+1, s1, then |V(T5)|=12s+5 and t(F12s+5)=4234s1. Moreover, since 43s+1434s and (3s+1)33s434s1 for any s1, it follows that for any s1,

    t(T5)=43s+1+(3s+1)33s(3s+1)23s434s+434s1=4234s1=t(F12s+5).

    If b=3s+2, s0, then |V(T5)|=12s+9 and t(F12s+9)=34s+3. Moreover, since 43s+24234s and (3s+2)33s+11134s for any s0, it follows that for any s0,

    t(T5)=43s+2+(3s+2)33s+1(3s+2)23s+14234s+1134s=34s+3=t(F12s+9).

    In this section, we give the proof of Theorem 2.1.

    Proof. Let F be a forest of order n. It is straightforward to check that the result is true if n5. We proceed with the induction on the order n of F. If FK1,n1, then by Lemma 2.3, the result trivially holds. Next we assume that F is not a star. By Lemma 2.2, for a tree T, there exists a vertex x with d(x)1 neighbors being leaves. Let N(x)={x1,,xd(x)1,y}, where y is the neighbor of x which is not a leaf of T, as shown in Figure 2.

    Figure 2.  T.

    Claim 1. If d(x)6, then t(T)t(Fn).

    Proof. Let Tx and Ty be two components of Txy containing x and y respectively. Then |V(Tx)|=d(x). Observe that for an M2-CIS S of T, either xS or xS with dT[S](x)=1. Let us define

    tˉx=|{S:xS}|,

    t1x=|{S:dT[S](x)=1}|

     =|{S:dT[S](x)=1,{x,y}S}|+|{S:dT[S](x)=1,{x,xi}S,

      i{1,,d(x)1}}|.

    Thus, t(T)=tˉx+t1x. Since tˉx=t(Ty) and t1xd(x)t(Ty), we have

    t(T)(d(x)+1)t(Ty). (3.1)

    Let V(Tx)=V(Tx). We consider three cases in terms of the modularity of d(x) (mod3).

    Case 1. d(x)=3s, s2

    Let Tx=sP3. Then t(Tx)=3s. By (3.1), it follows that for any s2,

    t(T)(3s+1)t(Ty)3st(Ty).

    By the induction hypothesis, t(Ty)t(Fnd(x)). Hence, t(T)t(Fn).

    Case 2. d(x)=3s+1, s2

    Let Tx=(s1)P3K1,3. Then t(Tx)=43s1. By (3.1), it follows that for any s2,

    t(T)(3s+2)t(Ty)43s1t(Ty).

    By the induction hypothesis, t(Ty)t(Fnd(x)). Hence, t(T)t(Fn).

    Case 3. d(x)=3s+2, s2

    Let Tx=(s2)P32K1,3. Then t(Tx)=423s2. By (3.1), it follows that for any s2,

    t(T)(3s+3)t(Ty)423s2t(Ty).

    By the induction hypothesis, t(Ty)t(Fnd(x)). Hence, t(T)t(Fn).

    Claim 2. If d(x)=4 or 5, then t(T)t(Fn).

    Proof. The meanings of notations here are same as those adopted in Claim 1. Let F1=T(N[x]{y}), F2=TN(x)N(y) and F3=TN[x]. Combining these observations with the definition of Fi, we get that tˉx=t(F1), t1x=t(F2)+(d(x)1)t(F3). Since t(T)=tˉx+t1x, we have

    t(T)=t(F1)+t(F2)+(d(x)1)t(F3). (3.2)

    Let ni be the order of Fi for each i{1,2,3}. Then n1=nd(x), n2=nd(x)d(y) and n3=nd(x)1. We consider three cases in terms of the modularity of n (mod3).

    Case 1. n=3s, s2

    Subcase 1.1. d(y)=3l, l1

    If d(x)=4, then n1=3(s2)+2, n2=3(sl2)+2 and n3=3(s2)+1. By the induction hypothesis, t(F1)423s4, t(F2)423sl4 and t(F3)43s3. Moreover, since 423l+5234 for any l1, by (3.2), it follows that for any s2 and l1,

    t(T)=(42+423l+432)3s4=(423l+52)3s43s=t(Fn).

    If d(x)=5, then n1=3(s2)+1, n2=3(sl2)+1 and n3=3(s2). By the induction hypothesis, t(F1)43s3, t(F2)43sl3 and t(F3)3s2. Moreover, since 43l+1633 for any l1, by (3.2), it follows that for any s2 and l1,

    t(T)=(4+43l+12)3s3=(43l+16)3s33s=t(Fn).

    Subcase 1.2. d(y)=3l+1, l1

    If d(x)=4, then n2=3(sl2)+1. By the induction hypothesis, t(F2)43sl3. Moreover, since 123l+5234 for any l1, by (3.2), it follows that for any s2 and l1,

    t(T)=(42+123l+432)3s4=(123l+52)3s43s=t(Fn).

    If d(x)=5, then n2=3(sl2). By the induction hypothesis, t(F2)3sl2. Moreover, since 33l+1633 for any l1, by (3.2), it follows that for any s2 and l1,

    t(T)=(4+33l+12)3s3=(33l+16)3s33s=t(Fn).

    Subcase 1.3. d(y)=3l+2, l0

    If d(x)=4, then n2=3(sl2). By the induction hypothesis, t(F2)3sl2. Moreover, since 93l+5234 for any l0, by (3.2), it follows that for any s2 and l0,

    t(T)=(42+93l+432)3s4=(93l+52)3s43s=t(Fn).

    If d(x)=5, then n2=3(sl3)+2. By the induction hypothesis, t(F2)423sl5. Moreover, since 423l+14435 for any l0, by (3.2), it follows that for any s2 and l0,

    t(T)=(432+423l+433)3s5=(423l+144)3s53s=t(Fn).

    Case 2. n=3s+1, s2

    Subcase 2.1. d(y)=3l,l1

    If d(x)=4, then n1=3(s1), n2=3(sl1) and n3=3(s2)+2. By the induction hypothesis, t(F1)3s1, t(F2)3sl1 and t(F3)423s4. Moreover, since 93l+25432 for any l1, by (3.2), it follows that for any s2 and l1,

    t(T)=(32+93l+42)3s3=(93l+25)3s343s1=t(Fn).

    If d(x)=5, then n1=3(s2)+2, n2=3(sl2)+2 and n3=3(s2)+1. By the induction hypothesis, t(F1)423s4, t(F2)423sl4 and t(F3)43s3. Moreover, since 423l+64433 for any l1, by (3.2), it follows that for any s2 and l1,

    t(T)=(42+423l+423)3s4=(423l+64)3s443s1=t(Fn).

    Subcase 2.2. d(y)=3l+1, l1

    If d(x)=4, then n2=3(sl2)+2. By the induction hypothesis, t(F2)423sl4. Moreover, since 423l+1+25432 for any l1, by (3.2), it follows that for any s2 and l1,

    t(T)=(32+423l+1+42)3s3=(423l+1+25)3s343s1=t(Fn).

    If d(x)=5, then n2=3(sl2)+1. By the induction hypothesis, t(F2)43sl3. Moreover, since 123l+64433 for any l1, by (3.2), it follows that for any s2 and l1,

    t(T)=(42+123l+423)3s4=(123l+64)3s443s1=t(Fn).

    Subcase 2.3. d(y)=3l+2, l0

    If d(x)=4, then n2=3(sl2)+1. By the induction hypothesis, t(F2)43sl3. Moreover, since 43l+25432 for any l0, by (3.2), it follows that for any s2 and l0,

    t(T)=(32+43l+42)3s3=(43l+25)3s343s1=t(Fn).

    If d(x)=5, then n2=3(sl2). By the induction hypothesis, t(F2)3sl2. Moreover, since 323l+64432 for any l0, by (3.2), it follows that for any s2 and l0,

    t(T)=(42+323l+423)3s4=(323l+64)3s443s1=t(Fn).

    Case 3. n=3s+2, s2

    Subcase 3.1. d(y)=3l,l1

    If d(x)=4, then n1=3(s1)+1, n2=3(sl1)+1 and n3=3(s1). By the induction hypothesis, t(F1)43s2, t(F2)43sl2 and t(F3)3s1. Moreover, since 43l+1342 for any l1, by (3.2), it follows that for any s2 and l1,

    t(T)=(4+43l+32)3s2=(43l+13)3s2423s2=t(Fn).

    If d(x)=5, then n1=3(s1), n2=3(sl1) and n3=3(s2)+2. By the induction hypothesis, t(F1)3s1, t(F2)3sl1 and t(F3)423s4. Moreover, since 333l+914232 for any l1, by (3.2), it follows that for any s2 and l1,

    t(T)=(33+333l+43)3s4=(333l+91)3s4423s2=t(Fn).

    Subcase 3.2. d(y)=3l+1, l1

    If d(x)=4, then n2=3(sl1). By the induction hypothesis, t(F2)3sl1. Moreover, since 33l+1342 for any l1, by (3.2), it follows that for any s2 and l1,

    t(T)=(4+33l+32)3s2=(33l+13)3s2423s2=t(Fn).

    If d(x)=5, then n2=3(sl2)+2. By the induction hypothesis, t(F2)423sl4. Moreover, since 423l+914232 for any l1, by (3.2), it follows that for any s2 and l1,

    t(T)=(33+423l+43)3s4=(423l+91)3s4423s2=t(Fn).

    Subcase 3.3. d(y)=3l+2, l0

    If d(x)=4, then n2=3(sl2)+2. By the induction hypothesis, t(F2)423sl4. Moreover, since 423l+2+1342 for any l0, by (3.2), it follows that for any s2 and l0,

    t(T)=(4+423l+2+32)3s2=(423l+2+13)3s2423s2=t(Fn).

    If d(x)=5, then n2=3(sl2)+1. By the induction hypothesis, t(F2)43sl3. Moreover, since 123l+914232 for any l0, by (3.2), it follows that for any s2 and l0,

    t(T)=(33+123l+43)3s4=(123l+91)3s4423s2=t(Fn).

    Claim 3. t(T)t(Fn) if one of the following conditions holds:

    (1) n6, n0 or 1 (mod3), d(x)=3;

    (2) n8, n2(mod3), d(x)=3, d(y)3, where yN(x).

    Proof. The meanings of notations here are same as those adopted in Claims 1 and 2. By (3.2), we have

    t(T)=t(F1)+t(F2)+2t(F3). (3.3)

    Let ni be the order of Fi for each i{1,2,3}. We consider three cases in terms of the modularity of n (mod3).

    Case 1. n=3s, s2.

    n1=3(s1), n3=3(s2)+2. By the induction hypothesis, t(F1)3s1 and t(F3)423s4.

    Subcase 1.1. d(y)=3l, l1

    Since n2=3(sl1), by the induction hypothesis, t(F2)3sl1. Moreover, since 333l+5934 for any l1, by (3.3), it follows that for any s2 and l1,

    t(T)(33+333l+242)3s4=(333l+59)3s43s=t(Fn).

    Subcase 1.2. d(y)=3l+1, l1

    Since n2=3(sl2)+2, by the induction hypothesis, t(F2)423sl4. Moreover, since 423l+5934 for any l1, by (3.3), it follows that for any s2 and l1,

    t(T)(33+423l+242)3s4=(423l+59)3s43s=t(Fn).

    Subcase 1.3. d(y)=3l+2, l0

    Since n2=3(sl2)+1, by the induction hypothesis, t(F2)43sl3. Moreover, since 123l+5934 for any l0, by (3.3), it follows that for any s2 and l0,

    t(T)(33+123l+242)3s4=(123l+59)3s43s=t(Fn).

    Case 2. n=3s+1, s2

    By the definition of ni, n1=3(s1)+1, n3=3(s1). By the induction hypothesis, t(F1)43s2 and t(F3)3s1.

    Subcase 2.1. d(y)=3l, l1

    Since n2=3(sl1)+1, by the induction hypothesis, t(F2)43sl2. Moreover, since 43l+1043 for any l1, by (3.3), it follows that for any s2 and l1,

    t(T)(4+43l+23)3s2=(43l+10)3s243s1=t(Fn).

    Subcase 2.2. d(y)=3l+1, l1

    Since n2=3(sl1), by the induction hypothesis, t(F2)3sl1. Moreover, since 33l+1043 for any l1, by (3.3), it follows that for any s2 and l1,

    t(T)(4+33l+23)3s2=(33l+10)3s243s1=t(Fn).

    Subcase 2.3. d(y)=3l+2, l0

    Since n2=3(sl2)+2, by the induction hypothesis, t(F2)423sl4. Moreover, since 423l+90433 for any l0, by (3.3), it follows that for any s2 and l0,

    t(T)(432+423l+233)3s4=(423l+90)3s443s1=t(Fn).

    Case 3. n=3s+2, s2

    By the definition of ni, n1=3(s1)+2, n3=3(s1)+1. By the induction hypothesis, t(F1)423s3 and t(F3)43s2.

    Subcase 3.1. d(y)=3l, l1

    Since n2=3(sl1)+2, by the induction hypothesis, t(F2)423sl3. Moreover, since 423l+40423 for any l1, by (3.3), it follows that for any s2 and l1,

    t(T)(42+423l+83)3s3=(423l+40)3s3423s2=t(Fn).

    Subcase 3.2. d(y)=3l+1, l1

    Since n2=3(sl1)+1, by the induction hypothesis, t(F2)43sl2. Moreover, since 123l+40423 for any l1, by (3.3), it follows that for any s2 and l1,

    t(T)(42+123l+83)3s3=(123l+40)3s3423s2=t(Fn).

    Subcase 3.3. d(y)=3l+2, l1

    Since n2=3(sl1), by the induction hypothesis, t(F2)3sl1. Moreover, since 323l+40423 for any l1, by (3.3), it follows that for any s2 and l1,

    t(T)(42+323l+83)3s3=(323l+40)3s3423s2=t(Fn).

    In view of Claim 3, we consider the case that d(y)=2 and n=3s+2 where s2.

    Claim 4. Assume that d(y)=2 for the remaining neighbor y of x and d(z)1 for the neighbor of y other than x, as shown in Figure 3. If Tz has an isolated vertex or an isolated edge, then t(T)t(Fn).

    Figure 3.  T.

    Proof. Let Tx and Ty be two components of Txy containing x and y respectively.

    Case 1. Tz has exactly an isolated vertex.

    By Lemma 2.4, we distinguish two subcases in terms of d(z)3.

    Subcse 1.1. d(z)=3

    Observe that for an M2-CIS S of Ty, either yS or yS with dT[S](y)1. Let us define ˜t0y=|{S:dT[S](y)=0}|, ˜t1y=|{S:dT[S](y)=1}|, ˜tˉy=|{S:yS}|. Thus, t(Ty)=˜t0y+˜t1y+˜tˉy.

    Observe that for an M2-CIS S of T, either yS or yS with dT[S](y)1. Let

    t0y=|{S:dT[S](y)=0}|=˜t0y,

    t1y=|{S:dT[S](y)=1}|

     =|{S:dT[S](y)=1,{x,y}S}|+|{S:dT[S](y)=1,{y,z}S}|

     =˜t0y+˜t1y.

    tˉy=|{S:yS}|=|{S:yS,xS}|+|{S:yS,xS}|

     (˜t0y+2˜t1y+2˜tˉy)+˜tˉy=˜t0y+2˜t1y+3˜tˉy.

    Clearly, t(T)=t0y+t1y+tˉy3˜t0y+3˜t1y+3˜tˉy. Since t(Ty)=˜t0y+˜t1y+˜tˉy, t(T)3t(Ty)=t(Tx)t(Ty). By the induction hypothesis, t(Ty)t(Fn3). Hence, t(T)t(Tx)t(Fn3)t(Fn).

    Subcase 1.2. d(z)4

    Let Tyz=TyTz where zTz. Observe that for an M2-CIS S of Tz, either zS or zS with dT[S](z)=1. Let us define tˉz=|{S:zS}| and t1z=|{S:dT[S](z)=1}|. Thus, t(Tz)=tˉz+t1z.

    The meanings of notations here are same as those adopted in Subcase 1.1. Note that t0y+t1yt1z+2tˉz and tˉy=2t(Tz)+t1z. Thus, t(T)2t(Tz)+2(t1z+tˉz). Since t(Tz)=tˉz+t1z, t(T)4t(Tz)=t(Ty)t(Tz). By the induction hypothesis, t(Tz)t(Fn4). Hence, t(T)t(Ty)t(Fn4)t(Fn).

    Case 2. Tz has two isolated vertices.

    The meanings of notations here are same as those adopted in Subcases 1.1 and 1.2. Note that t(Tz)=t1z+tˉz, t0y=tˉz, t1ytˉz+t1z, tˉy=2t(Tz)+t1z. Thus, t(T)2t(Tz)+2(t1z+tˉz)=4t(Tz)=t(Ty)t(Tz). By the induction hypothesis, t(Tz)t(Fn4). Hence, t(T)t(Ty)t(Fn4)t(Fn).

    Case 3. Tz has an isolated edge.

    Note that t1z+tˉzt(Tz). By a similar argument as in the proof of Case 2, we show that t(T)t(Fn).

    In view of Claim 4, we consider the case that d(z)=2.

    Claim 5. Assume that there exists a path P:=xyzw in T with d(x)=3,d(y)=d(z)=2, as shown in Figure 4. We have t(T)t(Fn) if one of the following conditions holds:

    Figure 4.  T.

    (1) Tw has an isolated vertex or an isolated edge;

    (2) Tw has no isolated vertex or isolated edge, where d(w)2.

    Proof. Let N(w)={w1,,wd(w)1,z}. We consider two cases in the following.

    Case 1. Tw has an isolated vertex or an isolated edge.

    Subcase 1.1. Tw has an isolated vertex.

    Let Tyz=TyTz where zTz. Observe that for an M2-CIS S of Tz, either zS or zS with dT[S](z)1. Let us define ˜t0z=|{S:dT[S](z)=0}|, ˜tˉz=|{S:zS}|, ˜t1z=|{S:dT[S](z)=1}|. Thus, t(Tz)=˜t0z+˜t1z+˜tˉz.

    Observe that for an M2-CIS S of T, either zS or zS with dT[S](z)1. Let

    t0z=|{S:dT[S](z)=0}|=2˜t0z,

    t1z=|{S:dT[S](z)=1}|

     =|{S:dT[S](z)=1,{y,z}S}|+|{S:dT[S](z)=1,{z,w}S}|

     =˜t0z+3˜t1z,

    tˉz=|{S:zS}|

     =|{S:zS,yS,dT[S](y)=0}|+|{S:zS,yS,dT[S](y)=1}|

      +|{S:zS,yS}|

     =˜tˉz+(|{S:zS,yS,dT[S](y)=1,wS}|

      +|{S:zS,yS,dT[S](y)=1,wS}|)+2˜tˉz

     3˜tˉz+(˜t0z+˜tˉz)=˜t0z+4˜tˉz.

    Obviously, t(T)=t0z+t1z+tˉz4˜t0z+3˜t1z+4˜tˉz. Since t(Tz)=˜t0z+˜t1z+˜tˉz, t(T)4t(Tz)=t(Ty)t(Tz). By the induction hypothesis, t(Tz)t(Fn4). Hence, t(T)t(Ty)t(Fn4)t(Fn).

    Subcase 1.2. Tw has an isolated edge.

    By Lemma 2.4, d(w)3. Let Tzw=TzTw where wTw and Tz=K1,4 where V(Tz)=V(Tz). Then t(Tz)=5. Observe that for an M2-CIS S of Tw, either wS or wS with dT[S](w)1. Let us define t0w=|{S:dT[S](w)=0}|, tˉw=|{S:wS}|, t1w=|{S:dT[S](w)=1}|. Thus, t(Tw)=t0w+t1w+tˉw.

    The meanings of notations here are same as those adopted in Subcase 1.1. Note that t0z=2tˉw, t1z2t0w+t1w+2tˉw and tˉz=t(Tw)+t0w+3t1w.

    We obtain that t(T)=t0z+t1z+tˉzt(Tw)+3t0w+4t1w+4tˉw. Since t(Tw)=t0w+t1w+tˉw, t(T)5t(Tw)=t(Tz)t(Tw). By the induction hypothesis, t(Tw)t(Fn5). Hence, t(T)t(Tz)t(Fn5)t(Fn).

    Case 2. Tw has no isolated vertex or isolated edge, where d(w)2.

    The meanings of notations here are same as those adopted in Subcase 1.1. Let F1=T(N[x]{z}), F2=TN[x]V(P), F3=TN[x]V(P)N(w) and F4=TN[x]V(P)N(w)N(wi). Combining these observations with the definition of Fi, we get that t0z=2t(F2), t1z=t(F2)+3t(F3), tˉz=t(F1)+t(F3)+3(d(w)1)t(F4). Since t(T)=t0z+t1z+tˉz, we have

    t(T)=t(F1)+3t(F2)+4t(F3)+3(d(w)1)t(F4). (3.4)

    Let ni be the order of Fi for each i{1,2,3,4}. Then n1=3(s1), n2=3(s2)+2. By the induction hypothesis, t(F1)3s1 and t(F2)423s4. We consider three cases in terms of the modularity of d(w) (mod3).

    Subcase 2.1. d(w)=3l,l1

    Since n3=3(sl1), n43(sl2)+2, by the induction hypothesis, t(F3)3sl1 and t(F4)423sl4. Moreover, since 48l+203l+25423 for any l1, by (3.4), it follows that for any s2 and l1,

    t(T)(32+42+4323l+42(3l1)3l)3s3=(48l+203l+25)3s3423s2=t(Fn).

    Subcase 2.2. d(w)=3l+1,l1

    Since n3=3(sl2)+2, n43(sl2)+1, by the induction hypothesis, t(F3)423sl4 and t(F4)43sl3. Moreover, since 108l+643l+754232 for any l1, by (3.4), it follows that for any s2 and l1,

    t(T)(33+423+433l+4l333l)3s4=(108l+643l+75)3s4423s2=t(Fn).

    Subcase 2.3. d(w)=3l+2,l1

    Since n3=3(sl2)+1, n43(sl2), by the induction hypothesis, t(F3)43sl3 and t(F4)3sl2. Moreover, since 27l+253l+25423 for any l1, by (3.4), it follows that for any s2 and l1,

    t(T)(32+42+423l+32(3l+1)3l)3s3=(27l+253l+25)3s3423s2=t(Fn).

    In view of Claim 5, we proceed to consider the case that d(w)=2.

    Claim 6. Assume that there exists a path P:=xyzwu in T with d(x)=3,d(y)=d(z)=d(w)=2, as shown in Figure 5. We have t(T)t(Fn) if one of the following conditions holds:

    Figure 5.  T.

    (1) Tu has an isolated vertex or an isolated edge;

    (2) Tu has no isolated vertex or isolated edge, where d(u)2.

    Proof. Let Twu=TwTu where uTu and N(u)={u1,,ud(u)1,w}.

    Case 1. Tu has an isolated vertex or an isolated edge.

    Subcase 1.1. Tu has an isolated vertex.

    By Lemma 2.4, d(u)3. Let Tw=2P3 where V(Tw)=V(Tw). Then t(Tw)=9. Observe that for an M2-CIS S of Tu, either uS or uS with dT[S](u)=1. Let us define t1u=|{S:dT[S](u)=1}|, tˉu=|{S:uS}|. Thus, t(Tu)=t1u+tˉu.

    Observe that for an M2-CIS S of T, either wS or wS with dT[S](w)1. Let

    t0w=|{S:dT[S](w)=0}|4tˉu,

    t1w=|{S:dT[S](w)=1}|

     =|{S:dT[S](w)=1,{w,u}S}|+|{S:dT[S](w)=1,{w,z}S}|

     4t1u+(t1u+4tˉu)=5t1u+4tˉu.

    tˉw=|{S:wS}|

     =|{S:wS,zS,dT[S](z)=0}|+|{S:wS,zS,dT[S](z)=1}|

      +|{S:wS,zS}|

     =2t1u+t(Tu)+t1u=3t1u+t(Tu).

    Clearly, t(T)=t0w+t1w+tˉwt(Tu)+8(t1u+tˉu). Since t(Tu)=t1u+tˉu, t(T)9t(Tu)=t(Tw)t(Tu). By the induction hypothesis, t(Tu)t(Fn6). Hence, t(T)t(Tw)t(Fn6)t(Fn).

    Subcase 1.2. Tu has an isolated edge.

    The meanings of notations here are same as those adopted in Subcase 1.1, with exception that adding the definition of t0u. More precisely, let t0u=|{S:dT[S](u)=0}|. Then t(Tu)=t0u+t1u+tˉu.

    Note that t0w=2tˉu, t1w4t1u+3tˉu and tˉw2t0u+3t1u+t(Tu). Thus, t(T)=t0w+t1w+tˉwt(Tu)+2t0u+7t1u+5tˉu. Moreover, t(Tu)=t0u+t1u+tˉu, t(T)8t(Tu)<9t(Tu)=t(Tw)t(Tu). By the induction hypothesis, t(Tu)t(Fn6). Hence, t(T)t(Tw)t(Fn6)t(Fn).

    Case 2. Tu has no isolated vertex or isolated edge, where d(u)2.

    The meanings of notations here are same as those adopted in Subcase 1.1. Let F1=TN(x)V(P), F2=TN(x)(V(P){u}), F3=TN(x)V(P)N(u) and F4=TN(x)V(P)N(u)N(ui). Combining these observations with the definition of Fi, we get that t0w=2t(F1), t1w=3t(F1)+4t(F3), tˉw=t(F2)+2t(F3)+3(d(u)1)t(F4). Since t(T)=t0w+t1w+tˉw, we have

    t(T)=5t(F1)+t(F2)+6t(F3)+3(d(u)1)t(F4). (3.5)

    Let ni be the order of Fi for i{1,2,3,4}. Then n1=3(s2)+1 and n2=3(s2)+2. By the induction hypothesis, t(F1)43s3 and t(F2)423s4. Now we consider three subcases in terms of d(u) (mod3).

    Subcase 2.1. d(u)=3l, l1

    By the definition of ni, n3=3(sl2)+2 and n43(sl2)+1. By the induction hypothesis, t(F3)423sl4 and t(F4)43sl3. Moreover, since 108l+603l+764232 for any l1, by (3.5), it follows that for any s2 and l1,

    t(T)=(543+42+6423l+432(3l1)3l)3s4=(108l+603l+76)3s4423s2=t(Fn).

    Subcase 2.2. d(u)=3l+1,l1

    By the definition of ni, n3=3(sl2)+1 and n4=3(sl2). By the induction hypothesis, t(F3)43sl3 and t(F4)3sl2. Moreover, since 81l+723l+764232 for any l1, by (3.5), it follows that for any s2 and l1,

    t(T)=(543+42+6123l+34l3l)3s4=(81l+723l+76)3s4423s2=t(Fn).

    Subcase 2.3. d(u)=3l+2,l1

    By the definition of ni, n3=3(sl2) and n4=3(sl3)+2. By the induction hypothesis, t(F3)3sl2 and t(F4)423sl5. Moreover, since 48l+703l+764232 for any l1, by (3.5), it follows that for any s2 and l1,

    t(T)=(543+42+6323l+42(3l+1)3l)3s4(48l+703l+76)3s4423s2=t(Fn).

    In view of Claim 6, we proceed to consider the case that d(u)=2.

    Claim 7. Assume that there exists a path P:=xyzwuq in T with d(x)=3,d(y)=d(z)=d(w)=d(u)=2, as shown in Figure 6. We have t(T)t(Fn).

    Figure 6.  T.

    Proof. Let Tuq=TuTq where qTq and N(q)={q1,,qd(q)1,u}.

    Case 1. Tq has an isolated vertex.

    By Lemma 2.4, d(q)3. Let Tu=P3K1,3 where V(Tu)=V(Tu). Then t(Tu)=12. Observe that for an M2-CIS S of Tq, either qS or qS with dT[S](q)=1. Let us define t1q=|{S:dT[S](q)=1}|, tˉq=|{S:qS}|. Thus, t(Tq)=t1q+tˉq.

    Observe that for an M2-CIS S of T, either uS or uS with dT[S](u)1. Let

    t0u=|{S:dT[S](u)=0}|

     =|{S:dT[S](u)=0,zS,dT[S](z)=1}|+|{S:dT[S](u)=0,zS,

      dT[S](z)=0}|

     2tˉq+(t1q+tˉq)=t1q+3tˉq.

    t1u=|{S:dT[S](u)=1}|

     =|{S:dT[S](u)=1,{u,q}S}|+|{S:dT[S](u)=1,{w,u}S}|

     4t1q+(6tˉq+t1q)=5t1q+6tˉq,

    tˉu=|{S:uS}|

     =|{S:uS,wS,dT[S](w)=0}|+|{S:uS,wS,dT[S](w)=1}|

      +|{S:uS,wS}|

     =2t1q+3t(Tq)+t1q=3t1q+3t(Tq).

    Clearly, t(T)=t0u+t1u+tˉu3t(Tq)+9(t1q+tˉq). Since t(Tq)=t1q+tˉq, t(T)12t(Tq)=t(Tu)t(Tq). By the induction hypothesis, t(Tq)t(Fn7). Hence, t(T)t(Tu)t(Fn7)t(Fn).

    Case 2. Tq has no isolated vertex.

    The meanings of notations here are same as those adopted in Case 1. Let F1=TN(x)V(P)N(q), F2=TN(x)V(P)N(q)N(qi), F3=TN(x)V(P) and F4=TN(x)(V(P){q}). Combining these observations with the definition of Fi, we get that t0u=3t(F3), t1u=4t(F1)+4t(F3), tˉu=2t(F1)+3(d(q)1)t(F2)+3t(F4). Since t(T)=t0u+t1u+tˉu, we have

    t(T)=6t(F1)+3(d(q)1)t(F2)+7t(F3)+3t(F4). (3.6)

    Let ni be the order of Fi for i{1,2,3,4}. Then n3=3(s2) and n4=3(s2)+1. By the induction hypothesis, t(F1)3s2 and t(F4)43s3. We consider three subcases in terms of d(q) (mod3).

    Subcase 2.1. d(q)=3l,l1

    By the definition of ni, n1=3(sl2)+1 and n23(sl2). By the induction hypothesis, t(F1)43sl3 and t(F2)3sl2. Moreover, since 9l+53l+1142 for any l1, by (3.6), it follows that for any s2 and l1,

    t(T)(83l+3(3l1)3l+11)3s2(9l+53l+11)3s2423s2=t(Fn).

    Subcase 2.2. d(q)=3l+1,l1

    By the definition of ni, n1=3(sl2) and n23(sl3)+2. By the induction hypothesis, t(F1)3sl2 and t(F2)423sl5. Moreover, since 16l+183l+33423 for any l1, by (3.6), it follows that for any s2 and l1,

    t(T)(183l+42l3l+33)3s3(16l+183l+33)3s3423s2=t(Fn).

    Subcase 2.3. d(q)=3l+2,l0

    By the definition of ni, n1=3(sl3)+2 and n23(sl3)+1. By the induction hypothesis, t(F1)423sl5 and t(F2)43sl4. Moreover, since 36l+443l+994232 for any l0, by (3.6), it follows that for any s2 and l0,

    t(T)(2423l+12(3l+1)3l+99)3s4(36l+443l+99)3s4423s2=t(Fn).

    By Lemma 2.2, we consider the case that there exists a vertex with one neighbor being leaf.

    Claim 8. Assume that there exists a path P:=xyz in T with d(x)=1,d(y)=2, as shown in Figure 7. We have t(T)t(Fn) if one of the following conditions holds:

    Figure 7.  T.

    (1) Tz has an isolated vertex or an isolated edge other than the component xy;

    (2) Tz has no isolated vertex or isolated edge, where d(z)3.

    Proof. Let Tyz=TyTz where zTz and N(z)={z1,,zd(z)1,y}.

    Case 1. Tz has an isolated vertex.

    Let Ty=P3 where V(Ty)={x,y,z1}. Then t(Ty)=3.

    Subcase 1.1. Tz has exactly an isolated vertex, say z1.

    Observe that for an M2-CIS S of Tzz1, either zS or zS with dT[S](z)1. Let us define t0z=|{S:dT[S](z)=0}|, tˉz=|{S:zS}|, t1z=|{S:dT[S](z)=1}|. Thus, t(Tzz1)=t0z+t1z+tˉz.

    Observe that for an M2-CIS S of T, either yS or yS with dT[S](y)=1. Let

    tˉy=|{S:yS}|t(Tzz1)+t1z,

    t1y=|{S:dT[S](y)=1}|

     =|{S:dT[S](y)=1,{x,y}S}|+|{S:dT[S](y)=1,{y,z}S}|

     t(Tzz1)+t0z+tˉz,

    Clearly, t(T)=tˉy+t1y2t(Tzz1)+t0z+t1z+tˉz. Since t(Tzz1)=t0z+t1z+tˉz, t(T)3t(Tzz1)=t(Ty)t(Tzz1). By the induction hypothesis, t(Tzz1)t(Fn3). Hence, t(T)t(Ty)t(Fn3)t(Fn).

    Subcase 1.2. Tz has two isolated vertices.

    Observe that for an M2-CIS S of Tzz1, either zS or zS with dT[S](z)=1. Let tˉz=|{S:zS}|, t1z=|{S:dT[S](z)=1}|. Then t(Tzz1)=tˉz+t1z.

    The meanings of notations here are same as those adopted in Subcase 1.1. Note that tˉy2t1z and t1y2tˉz+t(Tzz1). We have t(T)=tˉy+t1yt(Tzz1)+2(tˉz+t1z). Moreover, t(Tzz1)=tˉz+t1z, t(T)3t(Tzz1)=t(Ty)t(Tzz1). By the induction hypothesis, t(Tzz1)t(Fn3). Hence, t(T)t(Ty)t(Fn3)t(Fn).

    Case 2. Tz has an isolated edge, say z1z11.

    Let Ty=K1,3 where V(Ty)={x,y,z1,z11}. Then t(Ty)=4. The meanings of nations here are same as those adopted in Subcase 1.1. It is sufficient to note that t0z+t1zt(Tzz1z11), tˉyt(Tzz1z11)+t0z+t1z and t1y2t(Tzz1z11). Thus, t(T)=tˉy+t1y4t(Tzz1z11)=t(Ty)t(Tzz1z11). By the induction hypothesis, t(Tzz1z11)t(Fn4). Hence, t(T)t(Ty)t(Fn4)t(Fn).

    In view of Claim 8, we consider the case that d(z)=2.

    Claim 9. Assume that there exists a path P:=xyzw in T with d(x)=1,d(y)=d(z)=2, as shown in Figure 8. We have t(T)t(Fn) if one of the following conditions holds:

    Figure 8.  T.

    (1) n6, n0 or 1(mod3);

    (2) n8, n2(mod3), d(w)3.

    Proof. Let F1=TN[y], F2=TV(P), and F3=TV(P)N(w). Observe that for an M2-CIS S of T, either zS or zS with dT[S](z)1. Let us define

    tˉz=|{S:zS}|=t(F1), t0z=|{S:dT[S](z)=0}|=t(F2),

    t1z=|{S:dT[S](z)=1}|

     =|{S:dT[S](z)=1,{y,z}S}|+|{S:dT[S](z)=1,{z,w}S}|

     =t(F2)+t(F3).

    Since t(T)=tˉz+t0z+t1z, we have

    t(T)=t(F1)+2t(F2)+t(F3) (3.7)

    Let ni be the order of Fi for each i{1,2,3}. We consider three cases in terms of the modularity of n(mod3).

    Case 1. n=3s, s2.

    By the definition of ni, n1=3(s1), n2=3(s2)+2. By the induction hypothesis, t(F1)3s1 and t(F2)423s4. We distinguish three subcases according to d(w)(mod3).

    Subcase 1.1. d(w)=3l, l1,

    Since n3=3(sl1), by the induction hypothesis, t(F3)3sl1. Moreover, since 333l+5934 for any l1, by (3.7), it follows that for any s2 and l1,

    t(T)(33+242+333l)3s4=(333l+59)3s43s=t(Fn).

    Subcase 1.2. d(w)=3l+1, l1

    Since n3=3(sl2)+2, by the induction hypothesis, t(F3)423sl4. Moreover, since 423l+5934 for any l1, by (3.7), it follows that for any s2 and l1,

    t(T)(33+242+423l)3s4=(423l+59)3s43s=t(Fn).

    Subcase 1.3. d(w)=3l+2, l0

    Since n3=3(sl2)+1, by the induction hypothesis, t(F3)43sl3. Moreover, since 123l+5934 for any l0, by (3.7), it follows that for any s2 and l0,

    t(T)(33+242+123l)3s4=(123l+59)3s43s=t(Fn).

    Case 2. n=3s+1, s2

    By the definition of ni, n1=3(s1)+1, n2=3(s1), by the induction hypothesis, t(F1)43s2 and t(F2)3s1.

    Subcase 2.1. d(w)=3l, l1

    Since n3=3(sl1)+1, by the induction hypothesis, t(F3)43sl2. Moreover, since 43l+1043 for any l1, by (3.7), it follows that for any s2 and l1,

    t(T)(4+23+43l)3s2=(43l+10)3s243s1=t(Fn).

    Subcase 2.2. d(w)=3l+1, l1

    Since n3=3(sl1), by the induction hypothesis, t(F3)3sl1. Moreover, since 33l+1043 for any l1, by (3.7), it follows that for any s2 and l1,

    t(T)(4+23+33l)3s2=(33l+10)3s243s1=t(Fn).

    Subcase 2.3. d(w)=3l+2, l0

    Since n3=3(sl2)+2, by the induction hypothesis, t(F3)423sl4. Moreover, since 423l+90433 for any l0, by (3.7), it follows that for any s2 and l0,

    t(T)(432+233+423l)3s4=(423l+90)3s443s1=t(Fn).

    Case 3. n=3s+2, s2

    By the definition of ni, n1=3(s1)+2, n2=3(s1)+1. By the induction hypothesis, we have t(F1)423s3 and t(F2)43s2.

    Subcase 3.1. d(w)=3l, l1

    Since n3=3(sl1)+2, by the induction hypothesis, t(F3)423sl3. Moreover, since 423l+40423 for any l1, by (3.7), it follows that for any s2 and l1,

    t(T)(42+83+423l)3s3=(423l+40)3s3423s2=t(Fn).

    Subcase 3.2. d(w)=3l+1, l1

    Since n3=3(sl1)+1, by the induction hypothesis, t(F3)43sl2. Moreover, since 123l+40423 for any l1, by (3.7), it follows that for any s2 and l1,

    t(T)(42+83+123l)3s3=(123l+40)3s3423s2=t(Fn).

    Subcase 3.3. d(w)=3l+2, l1

    Since n3=3(sl1), by the induction hypothesis, t(F3)3sl1. Moreover, since 323l+40423 for any l1, by (3.7), it follows that for any s2 and l1,

    t(T)(42+83+323l)3s3=(323l+40)3s3423s2=t(Fn).

    In view of Claim 9, we proceed to consider the case that d(w)=2 and n=3s+2 where s2.

    Claim 10. Assume that there exists a path P:=xyzwu in T with d(x)=1,d(y)=d(z)=d(w)=2, as shown in Figure 9. We have t(T)t(Fn) if one of the following conditions holds:

    Figure 9.  T.

    (1) Tu has an isolated vertex or an isolated edge;

    (2) Tu has no isolated vertex or isolated edge, where d(u)4.

    Proof. Let N(u)={u1,,ud(u)1,w}.

    Case 1. Tu has an isolated vertex or an isolated edge.

    Subcase 1.1. Tu has an isolated vertex.

    Let Tzw=TzTw where wTw. Observe that for an M2-CIS S of Tw, either wS or wS with dT[S](w)1. Let us define ˜t0w=|{S:dT[S](w)=0}|, ˜t1w=|{S:dT[S](w)=1}|, ˜tˉw=|{S:wS}|. Then t(Tw)=˜t0w+˜t1w+˜tˉw.

    Observe that for an M2-CIS S of T, either wS or wS with dT[S](w)1. Let

    t0w=|{S:dT[S](w)=0}|=˜t0w,

    t1w=|{S:dT[S](w)=1}|

     =|{S:dT[S](w)=1,{z,w}S}|+|{S:dT[S](w)=1,{w,u}S}|

     =˜t0w+˜t1w

    tˉw=|{S:wS}|

     =|{S:wS,uS,dT[S](u)=1}|+|{S:wS,uS}|

     3˜tˉw+˜t0w.

    It is easy to see that t(T)=t0w+t1w+tˉw3˜t0w+˜t1w+3˜tˉw. Since t(Tw)=˜t0w+˜t1w+˜tˉw, t(T)3t(Tw)=t(Tz)t(Tw). By the induction hypothesis, t(Tw)t(Fn3). Hence, t(T)t(Tz)t(Fn3)t(Fn).

    Subcase 1.2. Tu has an isolated edge.

    Let Twu=TwTu where uTu and Tw=K1,3 where V(Tw)=V(Tw). Then t(Tw)=4. Observe that for an M2-CIS S of Tu, either uS or uSdT[S](u)1. Let us define t0u=|{S:dT[S](u)=0}|, tˉu=|{S:uS}|, t1u=|{S:dT[S](u)=1}|. Then t(Tu)=t0u+t1u+tˉu.

    The meanings of notations here are same as those adopted in Subcase 1.1. Note that t0w=tˉu, t1wt1u+tˉu and tˉw=t(Tu)+t0u+2t1u.

    Thus, t(T)=t0w+t1w+tˉwt(Tu)+t0u+3t1u+2tˉu. Moreover, t(Tu)=t0u+t1u+tˉu, t(T)4t(Tu)=t(Tw)t(Tu). By the induction hypothesis, t(Tu)t(Fn4). Hence, t(T)t(Tw)t(Fn4)t(Fn).

    Case 2. Tu has no isolated vertex or isolated edge, where d(u)4.

    The meanings of notations here are same as those adopted in Subcase 1.1. Let F1=T(V(P){u}), F2=TV(P), F3=TV(P)N(u) and F4=TV(P)N(u)N(ui). Combining these observations with the definition of Fi, we get that t0w=t(F2), t1w=t(F2)+t(F3) and tˉw=t(F1)+t(F3)+2(d(u)1)t(F4). Since t(T)=t0w+t1w+tˉw, we have

    t(T)=t(F1)+2t(F2)+2t(F3)+2(d(u)1)t(F4). (3.8)

    Let ni be the order of Fi for each i{1,2,3,4}. Then n1=3(s1)+1, n2=3(s1). By the induction hypothesis, t(F1)43s2 and t(F2)3s1. We want to prove that t(T)423s2=t(Fn) for any s2. By (3.8), we complete the proof by showing that for any s2,

    t(F3)+(d(u)1)t(F4)3s1. (3.9)

    Subcase 2.1. d(u)=3l, l2

    Since n3=3(sl1)+1, n43(sl1), by the induction hypothesis, t(F3)43sl2 and t(F4)3sl1. Moreover, since 9l+13l3 for any l2, it follows that for any s2 and l2,

    43sl2+(3l1)3sl1=9l+13l3s23s1.

    i.e., (3.9) holds.

    Subcase 2.2. d(u)=3l+1, l0

    Since n3=3(sl1), n43(sl2)+2, by the induction hypothesis, t(F3)3sl1 and t(F4)423sl4. Moreover, since 16l+93l32 for any l0, it follows that for any s2 and l0,

    3sl1+42l3sl3=16l+93l3s33s1.

    i.e., (3.9) holds.

    Subcase 2.3. d(u)=3l+2, l1

    Since n3=3(sl2)+2, n43(sl2)+1, by the induction hypothesis, t(F3)423sl4 and t(F4)43sl3. Moreover, since 36l+283l33 for any l1, it follows that for any s2 and l1,

    423sl4+4(3l+1)3sl3=36l+283l3s43s1.

    i.e., (3.9) holds.

    In view of Claim 10, we proceed to consider the case that d(u)=2 and d(u)=3 respectively.

    Claim 11. Assume that there exists a path P:=xyzwuq in T with d(x)=1,d(y)=d(z)=d(w)=d(u)=2, as shown in Figure 10. We have t(T)t(F).

    Figure 10.  T.

    Proof. Let N(q)={q1,,qd(q)1,u} and Twu=TwTu where uTu.

    Case 1. Tq has an isolated vertex.

    Let Tw=K1,3 where V(Tw)=V(Tw). Then t(Tw)=4. Observe that for an M2-CIS S of Tu, either uS or uS with dT[S](u)1. Let us define ˜t0u=|{S:dT[S](u)=0}|, ˜tˉu=|{S:uS}|, ˜t1u=|{S:dT[S](u)=1}|. Thus, t(Tu)=˜t0u+˜t1u+˜tˉu.

    Observe that for an M2-CIS S of T, either uS or uS with dT[S](u)1. Let

    t0u=|{S:dT[S](u)=0}|=2˜t0u,

    t1u=|{S:dT[S](u)=1}|

     =|{S:dT[S](u)=1,{w,u}S}|+|{S:dT[S](u)=1,{u,q}S}|

     =˜t0u+3˜t1u.

    tˉu=|{S:uS}|

     =|{S:uS,wS}|+|{S:uS,wS,dT[S](w)=1}|+|{S:uS,

      wS,dT[S](w)=0}|

     =˜tˉu+(˜t0u+˜tˉu)+˜tˉu=˜t0u+3˜tˉu.

    Then t(T)=t0u+t1u+tˉu=4˜t0u+3˜t1u+3˜tˉu. Moreover, t(Tu)=˜t0u+˜t1u+˜tˉu, t(T)4t(Tu)=t(Tw)t(Tu). By the induction hypothesis, t(Tu)t(Fn4). Hence, t(T)t(Tw)t(Tu)t(Fn4).

    Case 2. Tq has no isolated vertex.

    The meanings of notations here are same as those adopted in Case 1. Let F1=TV(P)N(q), F2=TV(P)N(q)N(qi), F3=TV(P) and F4=T(V(P){q}). Combining these observations with the definition of Fi, we get that t0u=2t(F3), t1u=3t(F1)+t(F3) and tˉu=t(F1)+2(d(q)1)t(F2)+t(F4). Since t(T)=t0u+t1u+tˉu, we have

    t(T)=4t(F1)+2(d(u)1)t(F2)+3t(F3)+t(F4). (3.10)

    Let ni be the order of Fi for i{1,2,3,4}. Then n3=3(s2)+2, n4=3(s1). By the induction hypothesis, t(F3)423s4 and t(F4)3s1. We want to prove that t(T)423s2=t(Fn) for any s2. By (3.10), we complete the {proof} by showing that for any s2,

    4t(F1)+2(d(q)1)t(F2)233s3. (3.11)

    Subcase 2.1. d(q)=3l, l1

    n1=3(sl1), n23(sl2)+2, by the induction hypothesis, t(F1)3sl1 and t(F2)423sl4. Moreover, since 96l+763l233 for any l1, it follows that for any s2 and l1,

    43sl1+242(3l1)3sl4=96l+763l3s4233s3.

    i.e., (3.11) holds.

    Subcase 2.2. d(q)=3l+1, l0

    n1=3(sl2)+2, n23(sl2)+1, by the induction hypothesis, t(F1)423sl4 and t(F2)43sl3. Moreover, since 72l+643l233 for any l0, it follows that for any s2 and l0,

    433sl4+8l3sl2=72l+643l3s4233s3.

    i.e., (3.11) holds.

    Subcase 2.3. d(q)=3l+2, l0

    n1=3(sl2)+1, n23(sl2), by the induction hypothesis, t(F1)43sl3 and t(F2)3sl2. Moreover, since 18l+223l23 for any l0, it follows that for any s2 and l0,

    423sl3+2(3l+1)3sl2=18l+223l3s3233s3.

    i.e., (3.11) holds.

    Claim 12. Assume that there exists a vertex x with d(x)3 such all components of Tx are isomorphic to K1,3, but one, denoted by Ty where y is the neighbor of x lying in Ty, is not isomorphic to K1,3, as shown in Figure 11. We have t(T)t(Fn).

    Figure 11.  T.

    Proof. For an integer a2, assume that N(x)={x1,,xa,y}. Let Txy=TxTy where yTy. Then |V(Tx)|=4a+1.

    Case 1. Ty has no isolated vertex.

    Observe that for an M2-CIS S of Ty, either yS or yS with dT[S](y)1. Let us define t0y=|{S:dT[S](y)=0}|, tˉy=|{S:yS}|, t1y=|{S:dT[S](y)=1}|. Thus, t(Ty)=t0y+t1y+tˉy.

    Observe that for an M2-CIS S of T, either xS or xS with dT[S](x)1. Let

    t0x=|{S:dT[S](x)=0}|2a(t1y+tˉy),

    t1x=|{S:dT[S](x)=1}|

     =|{S:dT[S](x)=1,yS}|+|{S:dT[S](x)=1,yS}|

     3a(t0y+tˉy)+a3a1t(Ty),

    tˉxy=|{S:xS,dT[S](xi)=1ordT[S](xi)=dT[S](xj)=0,i,j{1,,a}}|

     =(4a2aa2a1)t(Ty),

    t1ˉxy=|{S:xS,dT[S](y)=1,dT[S](xi)1,exactlyonexi,dT[S](xi)=

      0orxiS,i{1,,a}}|

     =(2a+a2a1)t1y,

    t0ˉxy=|{S:xS,dT[S](y)=0,dT[S](xi)1,xiS,i{1,,a}}|

     =a2a1t0y,

    tˉx=|{S:xS}|=tˉxy+t1ˉxy+t0ˉxy

     =(4a2aa2a1)t(Ty)+(2a+a2a1)t1y+a2a1t0y.

    Since t(Ty)=t0y+t1y+tˉy, (2a+1+a2a1)t1y+(3a+a2a1)t0y+(3a+2a)tˉy(3a+a2a1)t(Ty). Moreover, t(T)=t0x+t1x+tˉx. We get that

    t(T)[4a+(a+3)3a12a]t(Ty).

    Let V(Tx)=V(Tx). Then t(TxTy)=t(Tx)t(Ty). By the induction hypothesis, t(Ty)t(Fn(4a+1)). We want to prove t(T)t(Tx)t(Fn(4a+1))t(Fn) for any a2. Therefore, we need to show that for any a2,

    4a+(a+3)3a12at(Tx). (3.12)

    We distinguish into three cases based on the modularity of a(mod3).

    Subcase 1.1. a=3s, s1

    Let Tx=(4s1)P3K1,3 where |V(Tx)|=12s+1. Then t(Tx)=434s1. Since 43s34s and (s+1)33s34s1 for any s2, by (3.12), it follows that for any s2,

    43s+(s+1)33s23s34s+34s1=434s1=t(Tx).

    If s=1, then a=3. Note that t0x8(t1y+tˉy), t1x23t0y+25tˉy+29t(Ty) and tˉx=44t(Ty)+20t1y+12t0y. Meanwhile, t(Ty)=t0y+t1y+tˉy, t(T)108t(Ty)=t(Tx)t(Ty). By the induction hypothesis, t(Ty)t(Fn(4a+1)). Hence, t(T)t(Tx)t(Fn(4a+1))t(Fn).

    Subcase 1.2. a=3s+1, s1

    Let Tx=(4s1)P32K1,3 where |V(Tx)|=3(4s+1)+2. Then t(Tx)=4234s1. Since 43s+1434s and (3s+4)33s434s1 for any s2, by (3.12), it follows that for any s2,

    43s+1+(3s+4)33s23s+1434s+434s1=4234s1=t(Tx).

    The result is true for s=1.

    Subcase 1.3. a=3s+2, s0

    Let Tx=(4s+3)P3 where |V(Tx)|=3(4s+3). Then t(Tx)=34s+3. Since 43s+24234s and (3s+5)33s+11134s for any s1, by (3.12), it follows that for any s1,

    43s+2+(3s+5)33s+123s+24234s+1134s=34s+3=t(Tx).

    The result is true for s=0.

    Case 2. Ty has an isolated vertex.

    By Lemma 2.4, d(y)3. The meanings of notations here are same as those adopted in Case 1. Thus t(Ty)=t1y+tˉy,t0x2a+1tˉy,t1xa3a1t(Ty), tˉx(4a2aa2a1)t(Ty)+(2a+a2a1)t1y.

    Furthmore, 2a+1tˉy+(2a+a2a1)t1y(3a+a2a1)t(Ty), t(T)[4a+(a+3)3a12a]t(Ty). By a similar argument as in the proof of Case 1, we show that t(T)t(Fn).

    In view of Claim 11, it remains to consider the case that d(u)=3.

    Claim 13. Assume that there exists a path P:=xyzwu in T with d(x)=1,d(y)=d(z)=d(w)=2,d(u)=3. If Tu has no isolated vertex or isolated edge, then t(T)t(Fn).

    Proof. By Claims 1–12, it is not difficult to observe that there exists a vertex x such that Tx has two components, one is isomorphic to P4, the other one is isomorphic to P4 or K1,3, as shown in Figure 12. The meanings of notations adopted here are same in Case 1 of Claim 12. Let Txy=TxTy where yTy and Tx=3P3 where V(Tx)=V(Tx). Then t(Tx)=33.

    Figure 12.  Tx has two components, one is isomorphic to P4, the other one is isomorphic to P4 or K1,3.

    Case 1. Tx has two components which are isomorphic to P4.

    Subcase 1.1. Ty has no isolated vertex.

    Note that t0x4(t0y+tˉy), t1x6t(Ty)+9(t0y+t1y), tˉx6t(Ty)+3t1y+2t0y. Meanwhile, t(T)=t0x+t1x+tˉx. This implies that t(T)12t(Ty)+15t0y+12t1y+4tˉy. Since t(Ty)=t0y+t1y+tˉy, t(T)33t(Ty)=t(Tx)t(Ty). By the induction hypothesis, t(Ty)t(Fn9). Hence, t(T)t(Tx)t(Fn9)t(Fn).

    Subcase 1.2. Ty has an isolated vertex.

    By Lemma 2.4, d(y)3. Note that t0x4t(Ty), t1x9t1y+6tˉy and tˉx6t(Ty)+3t1y. Since t(T)=t0x+t1x+tˉx, t(T)10t(Ty)+12t1y+6tˉy. Moreover, t(Ty)=t1y+tˉy, t(T)22t(Ty)33t(Ty)=t(Tx)t(Ty). By the induction hypothesis, t(Ty)t(Fn9). Hence, t(T)t(Tx)t(Fn9)t(Fn).

    Case 2. Tx has two components which are isomorphic to P4 and K1,3, respectively.

    Subcase 2.1. Ty has no isolated vertex.

    Note that t1x6t(Ty)+9(t0y+tˉy), t0x4(t1y+tˉy), tˉx7t(Ty)+5t1y+3t0y. Since t(T)=t0x+t1x+tˉx, t(T)13t(Ty)+12t0y+9t1y+13tˉy. Moreover, t(Ty)=t0y+t1y+tˉy and t(T)26t(Ty)33t(Ty)=t(Tx)t(Ty). By the induction hypothesis, t(Ty)t(Fn9). Hence, t(T)t(Tx)t(Fn9)t(Fn).

    Subcase 2.2. Ty has an isolated vertex.

    By Lemma 2.4, d(y)3. Note that t1x9(t1y+tˉy), t0x4t(Ty) and tˉx7t(Ty)+5t1y. Since t(T)=t0x+t1x+tˉx, t(T)11t(Ty)+14t1y+9tˉy. Moreover, t(Ty)=t1y+tˉy and t(T)25t(Ty)33t(Ty)=t(Tx)t(Ty). By the induction hypothesis, t(Ty)t(Fn9). Hence, t(T)t(Tx)t(Fn9)t(Fn).

    From the above discussion, we proceed to consider the following.

    Claim 14. Assume there exists a path P:=xyz in T with d(x)=1 or d(x)=3 such that two neighbors of x distinct from y being leaves, d(y)=2 and d(z)3. If Tz has no isolated vertex or isolated edge, then t(T)t(Fn).

    Proof. By Claims 1–11 and 13, it remains to consider the case that there exists a vertex w such that Tw has at least two components which are isomorphic to K1,3. By Lemma 2.4 and Claim 12, we have t(T)t(Fn).

    This completes the proof of theorem.

    In this paper, we determine the maximum number of maximal 2-component independent sets of a forest of order n. It is an interesting problem to determine the maximum number of maximal 2-component independent sets of graphs of order n over some other families, such as trees, bipartite graphs, triangle-free graphs, all connected graphs.

    The work was supported by the national natural science foundation of China (No. 12061073) and XJEDU2019I001.

    The authors declare no conflict of interests.



    [1] V. E. Alekseev, R. Boliac, D. V. Korobitsyn, V. V. Lozin, NP-hard graph problems and boundary classes of graphs, Theor. Comput. Sci., 389 (2007), 219–236. https://doi.org/10.1016/j.tcs.2007.09.013 doi: 10.1016/j.tcs.2007.09.013
    [2] R. Boliac, K. Cameron, V. V. Lozin, On computing the dissociation number and the induced matching number of bipartite graphs, Ars Combin., 72 (2004), 241–253.
    [3] S. Cheng, B. Wu, On the k-component independence number of a tree, Discrete Dyn. Nat. Soc., 2021, 5540604.
    [4] M. Hujter, Z. Tuza, The number of maximal independent sets in triangle-free graphs, SIAM J. Discrete Math., 6 (1993), 284–288. https://doi.org/10.1137/0406022 doi: 10.1137/0406022
    [5] M. J. Jou, G. J. Chang, Maximal independent sets in graphs with at most one cycle, Discrete Appl. Math., 79 (1997), 67–73. https://doi.org/10.1016/S0166-218X(97)00033-4 doi: 10.1016/S0166-218X(97)00033-4
    [6] K. M. Koh, C. Y. Goh, F. M. Dong, The maximum number of maximal independent sets in unicyclic connected graphs, Discrete Math., 308 (2008), 3761–3769. doilinkhttps://doi.org/10.1016/j.disc.2007.07.079 doi: 10.1016/j.disc.2007.07.079
    [7] Y. Orlovich, A. Dolguib, G. Finkec, V. Gordond, F. Wernere, The complexity of dissociation set problems in graphs, Discrete Appl. Math., 159 (2011) 1352–1366. https://doi.org/10.1016/j.dam.2011.04.023
    [8] C. H. Papadimitriou, M. Yannakakis, The complexity of restricted spanning tree problems, J. Assoc. Comput. Mach., 29 (1982), 285–309. https://doi.org/10.1145/322307.322309 doi: 10.1145/322307.322309
    [9] B. E. Sagan, A note on independent sets in trees, SIAM J. Discrete Math., 1 (1988), 105–108. https://doi.org/10.1137/0401012 doi: 10.1137/0401012
    [10] B. E. Sagan, V. R. Vatter, Maximal and maximum independent sets in graphs with at most r cycles, J. Graph Theory, 53 (2006), 283–314. https://doi.org/10.1002/jgt.20186 doi: 10.1002/jgt.20186
    [11] J. Tu, Z. Zhang, Y. Shi, The maximum number of maximum dissociation sets in trees, J. Graph Theory, 96 (2021), 472–489. https://doi.org/10.1002/jgt.22627 doi: 10.1002/jgt.22627
    [12] H. S. Wilf, The number of maximal independent sets in a tree, SIAM J. Alg. Discrete Methods, 7 (1986), 125–130. https://doi.org/10.1137/0607015 doi: 10.1137/0607015
    [13] I. Wloch, Trees with extremal numbers of maximal independent sets including the set of leaves, Discrete Math., 308 (2008), 4768–4772. https://doi.org/10.1016/j.disc.2007.08.087 doi: 10.1016/j.disc.2007.08.087
    [14] M. Yannakakis, Node-deletion problems on bipartite graphs, SIAM J. Comput., 10 (1981), 310–327. https://doi.org/10.1137/0210022 doi: 10.1137/0210022
    [15] J. Zito, The structure and maximum number of maximum independent sets in trees, J. Graph Theory, 15 (1991), 207–221. https://doi.org/10.1002/jgt.3190150208 doi: 10.1002/jgt.3190150208
  • This article has been cited by:

    1. Wenke Zhou, Guo Chen, Hongzhi Deng, Jianhua Tu, Enumeration of dissociation sets in grid graphs, 2024, 9, 2473-6988, 14899, 10.3934/math.2024721
    2. Yuting Tian, Jianhua Tu, Enumerating maximal dissociation sets in three classes of grid graphs, 2024, 474, 00963003, 128711, 10.1016/j.amc.2024.128711
    3. Ziyuan Wang, Lei Zhang, Jianhua Tu, Liming Xiong, Upper bound for the number of maximal dissociation sets in trees, 2025, 348, 0012365X, 114545, 10.1016/j.disc.2025.114545
  • Reader Comments
  • © 2022 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(1948) PDF downloads(54) Cited by(3)

Figures and Tables

Figures(12)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog