Independence and Matching Numbers of Unicyclic Graphs From Null Space
Luiz Emilio Allem, Daniel Alejandro Jaume, Gonzalo Molina, Maikon, Machado Toledo, and Vilmar Trevisan

TL;DR
This paper characterizes singular unicyclic graphs using null space support and derives formulas for their independence and matching numbers, enabling linear algebra-based computations.
Contribution
It introduces new formulas for independence and matching numbers of unicyclic graphs based on null space support, advancing graph theoretical analysis.
Findings
Characterization of singular unicyclic graphs
Closed-form formulas for independence and matching numbers
Linear algebra methods for graph property computation
Abstract
We characterize unicyclic graphs that are singular using the support of the null space of their pendant trees. From this, we obtain closed formulas for the independence and matching numbers of a unicyclic graph, based on the support of its subtrees. These formulas allows one to compute independence and matching numbers of unicyclic graphs using linear algebra methods.
| Support | -vertices | |
|---|---|---|
| Support | Core | -vertices |
|---|---|---|
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Independence and Matching Numbers of Unicyclic Graphs from Null Space
L. Emilio Allem
UFRGS - Universidade Federal do Rio Grande do Sul, Instituto de Matemática, Porto Alegre, Brazil
,
Daniel A. Jaume
Universidad Nacional de San Luis, Departamento de Matemáticas, San Luis, Argentina
,
Gonzalo Molina
Universidad Nacional de San Luis, Departamento de Matemáticas, San Luis, Argentina
,
Maikon M. Toledo
UFRGS - Universidade Federal do Rio Grande do Sul, Instituto de Matemática, Porto Alegre, Brazil
and
Vilmar Trevisan
UFRGS - Universidade Federal do Rio Grande do Sul, Instituto de Matemática, Porto Alegre, Brazil
Abstract.
We characterize unicyclic graphs that are singular using the support of the null space of their pendant trees. From this, we obtain closed formulas for the independence and matching numbers of a unicyclic graph, based on the support of its subtrees. These formulas allows one to compute independence and matching numbers of unicyclic graphs using linear algebra methods.
Key words and phrases:
Unicyclic, null space, nullity, support, independence number, matching number.
1991 Mathematics Subject Classification:
05C50, 15A18
1. Introduction
Recently, in [8], the authors studied the null space of the adjacency matrix of trees and they presented a null decomposition of trees. In general, this null decomposition divides a tree into two forests (one of the forests can be empty), one composed by singular trees and the other composed by non-singular trees. The technique used was the analysis of the support of the tree, where the support is defined as the subset of vertices for which at least one of its corresponding coordinates of the eigenvectors of the null space of the adjacency matrix is nonzero.
As an application, in [8], the null decomposition was used to obtain closed formulas for two classical parameters. The first one is the independence number of a graph , denoted by . Notice that the problem of computing is -hard [9] and several mathematicians have studied ( for example [2, 6, 12]). The second one is the matching number of a graph , denoted by [5, 11]. Historically the matching theory started with bipartite graphs and one of the earliest works was published in 1916 [10].
In this paper, we extend the results of [8] to unicyclic graphs. In a more general sense, we obtain structural information of the unicyclic graphs using the support of their subtrees. In particular, we obtain closed formulas for the independence and matching numbers of unicyclic graphs that depend on the support and the core of their subtrees. Next we give an outline of this paper. It is worth pointing out that, in practice, this means that these classical parameters can be computed using linear algebra.
In section 2, we present some basic notations and definitions of support of a graph. In section 3, we characterize singular unicyclic graphs using the support of their pendant trees. In section 4, we obtain a closed formula for the independence number of unicyclic graphs using the support of the subtrees of these unicyclic graphs. In section 5, we obtain closed formulas for the matching number of unicyclic graphs based on the support of subtrees of these unicyclic graphs.
2. Basic definitions and notation
In this section we present some notation and basic definitions. In particular, we explain the notion of support of a graph. We use the graphs of Figure 1 to illustrate the concepts used here.
Let be a simple graph of order , with vertex set and edge set , the adjacency matrix of is defined as
[TABLE]
Denote by the -eigenspace of , thus . The [math]-eigenspace () is the focus of our work and will be denote by . The nullity of a graph , denoted by , is the multiplicity of the eigenvalue zero in the spectrum of or, equivalently, the dimension of the [math]-eigenspace of . The graph is called singular if is a singular matrix or Otherwise, the graph is called non-singular.
As an example, we observe that the set is a basis for the null space of the tree of Figure 1, hence . And we notice that 0 is not an eigenvalue of the tree of Figure 1, so .
Definition 2.1**.**
A set of vertices of a graph is an independent set in if no two vertices in are adjacent. A maximum independent set is an independent set of maximum cardinality. The cardinality of any maximum independent set in , denoted by , is called the independence number of . denotes the set of all maximum independent sets of .
For example, in Figure 1 the vertex subsets and of the tree are the only independent sets of maximum cardinality. Therefore,
[TABLE]
and .
Definition 2.2**.**
A matching in is a set of pairwise non-adjacent edges, that is, no two edges in share a common vertex. A maximum matching is a matching of largest cardinality in . The matching number of , denoted by , is the size of a set of any maximum matching. denotes the set of all maximum matching of . A vertex is saturated by , if it is an endpoint of one of the edges in the matching . Otherwise the vertex is said non-saturated. Moreover, a matching is said to be perfect if it saturates all vertices of .
In the figures, we use zig zag edges to represent the edges of a matching.
In Figure 1, the tree has matching and the tree has perfect matching . Therefore, and .
The Edmond-Gallai vertices of , denoted by , is the set of all vertices of that are non-saturated by some maximum matching in .
For example, the maximum matchings of are
[TABLE]
Thus and
Definition 2.3**.**
Let be a graph with vertices and let be a vector of . The support of in is
[TABLE]
Let be a subset of . Then the support of in is
[TABLE]
As a convention, we use rectangular vertices in figures to represent vertices of the support. Consider the tree (Figure 1) and the set of vectors
Then
The following result shows that in order to compute the support of an eigenspace of , it is enough to analyse the coordinates of the vectors of a basis of this eigenspace.
Lemma 2.4**.**
[8]** Let be a graph, and an eigenvalue of . Let be a basis of , then
We are interested in the support of the null space of , that is, our focus is , which, for purposes of notation, is denoted by . In practice to compute we will use Lemma 2.4 and not the definition of support. That is, we compute a basis of the null space and consider the entries of the vectors in the basis to obtain the support.
For example, notice that is a basis of (Figure 1), thus
[TABLE]
Moreover, note that (Figure 1) is non-singular, that is, , therefore, .
Theorem 2.5**.**
[8]** Let be a tree, then is an independent set of .
The following well known result characterizes singular trees in terms of their matchings.
Lemma 2.6**.**
[3]** is a nonsingular tree if and only if has a perfect matching.
As we can see in Figure 1, the tree is a singular tree, because it does not have perfect matching. The tree is a non-singular tree, because it has perfect matching.
Our first goal is to characterize singular unicyclic graphs in terms of the support of their pendant trees, which is the subject of next section.
Using Lemma 2.6 above, and Lemma 3.12, Corollary 4.14 of [8] we obtain the next result.
Lemma 2.7**.**
Let be a tree, then .
It means that only the vertices of the support of a tree are not saturated by some maximum matching in this tree.
3. Singular Unicyclic Graphs
In this section we characterize singular unicyclic graphs using the support of their pendant trees, which is the statement of Theorem 3.6.
For cycles, the problem of characterizing singular graphs is solved.
Lemma 3.1**.**
[13]** A cycle of vertices is singular if and only if is divisible by .
Hence, for the remaining of this section, we will consider a unicyclic graph . Let be a unicyclic graph and let be the unique cycle of . For each vertex , we denote by the induced connected subgraph of with maximum possible number of vertices, which contains the vertex and no other vertex of . is called the pendant tree of at . Notice that is obtained by identifying the vertex of with the vertex on for all vertices . In Figure 2 we have two unicyclic graphs and with their pendant trees , , , , and , respectively.
Definition 3.2**.**
[7]** For a tree with at least two vertices, the vertex is called mismatched in if there exists a maximum matching of that does not saturate ; otherwise, is called matched in . If a tree consists of only one vertex it is considered mismatched.
A unicyclic graph is said be of Type if there exists a vertex on the cycle of such that is matched in , otherwise, is said to be of Type .
To emphasize, a unicyclic graph is of Type , if there exists a vertex of its cycle that is saturated by all maximum matchings of the pendant tree . is of Type if any vertex of its cycle is not saturated by some maximum matching of .
As an example, consider the unicyclic graph in Figure 2. We notice that is of Type , because the vertex is matched in . Indeed, the maximum matchings of are and and both saturate . On the other hand, the unicyclic graph of Figure 2 is of Type , because the pendant trees , and have maximum matchings that do not saturate , and , respectively. For example, , and are maximum matchings in , and , respectively, that do not saturate , and , respectively.
We show next that in order to verify that a unicyclic graph is Type or , it suffices to check whether a vertex of the cycle is or is not in the support of the pendant tree .
Proposition 3.3**.**
A unicyclic graph is of Type if and only if there exists at least one pendant tree such that .
Proof.
Since is of Type we know that there exists a vertex in the cycle of such that is always saturated by any maximum matching in , that is, , by Lemma 2.7 we have . ∎
Immediately, we obtain the dual result.
Corollary 3.4**.**
A unicyclic graph is of Type if and only if every pendant tree is such that .
The following result computes the nullity of a unicyclic graph from the nullity of its pendant trees.
Lemma 3.5**.**
[7]** Let be a unicyclic graph and let be its cycle. If is of Type and be matched in , then
[TABLE]
If is of Type then
[TABLE]
We now obtain a characterization of singular unicyclic graphs using the support of their pendant trees.
Theorem 3.6**.**
Let be a unicyclic graph and let be the cycle of . is singular if and only if one of the following happens:
- (i)
There is a pendant tree , with and either does not have perfect matching or does not have perfect matching;
- (ii)
Every pendant tree , is such that and either one of the trees that compose the forest does not have perfect matching or the cycle has length equal to a multiple of .
Proof.
- (i)
As there is a pendant tree such that , we conclude, by Proposition 3.3 that is of Type . Moreover, by Lemma 3.5 we conclude that will be singular if and only if or has nonzero nullity, and by Lemma 2.6 we know that this only happens if does not have perfect matching or does not have perfect matching;
- (ii)
Since every pendant tree , is such that we conclude by Corollary 3.4 that is of Type . Then, by Lemma 3.5 we know that will be singular if and only if or has nonzero nullity, and by the Lemma 2.6 and Lemma 3.1 it happens if and only if at least one of the trees composing the forest does not have perfect matching or the cycle has length equal to a multiple of .
∎
4. Independence number of Unicyclic graphs
In this section we obtain closed formulas for the independence number of a unicyclic graph . This formula depends on the number of vertices of the support and also on the amount of -vertices of subtrees of this unicyclic graph . In order to understand our result, we start by presenting the null decomposition of trees, given in [8].
Definition 4.1**.**
Let be a tree. The -forest of , denoted by , is defined as the subgraph induced by the closed neighborhood of in :
[TABLE]
The -forest of , denoted by , is defined as the remaining graph:
[TABLE]
The Null Decomposition of is the pair .
is called the set of -vertices of .
We represent star vertices in the figures as the -vertices. As an example, the support of the tree in Figure 3 is
[TABLE]
The -forest of generated by the closed neighborhood of the support consists of
[TABLE]
The -forest of consists of:
[TABLE]
Figure 3 illustrates the null decomposition of the tree .
Definition 4.2**.**
The core of , denoted by , is defined to be the set of all the neighbours of the supported vertices of :
[TABLE]
For example, the core of tree (Figure 3) is:
[TABLE]
The next lemma gives closed formulas for the independence and matching numbers of trees and it is crucial to prove our main results.
Lemma 4.3**.**
[8]** Let be a tree. Then
[TABLE]
Our next result tells us that given any vertex in a non-singular tree, there will always be at least one maximum independent set that does not contain this vertex and another maximum independent set that contains this vertex.
Proposition 4.4**.**
Let be a non-singular tree and . Then there exist such that and .
Proof.
Since is a tree we have that is a bipartite graph. Then there are two disjoint subsets and of such that and for all we have and . As is a non-singular tree, it has perfect matching . As and for all we have and , then . That is, . Therefore, given a we have and or and . ∎
Lemma 4.5**.**
If is a tree and then there exist such that and .
Proof.
The null decomposition, in general, divides a tree into two forests (one them may be empty), a forest formed by singular trees, denoted by , and other formed by non-singular trees, denoted by (see Theorem 4.5 and Theorem 4.13 of [8]).
Moreover, we have and . Thus, if we have for some . As is non-singular using Proposition 4.4 we obtain such that and .
Let
[TABLE]
We observe that and are independent sets, because is an independent set of and , then . Notice that , then by Lemma 4.3. Moreover, we have and . ∎
Lemma 4.6**.**
Let be a tree and an independent set of . If and then .
Proof.
We notice that, in general, , where , and (possibly we can have ). By Lemma 3.5 of [8] we have
[TABLE]
Note that is an independent set of . Indeed, and and are independent sets, because , and is an independent set by Theorem 2.5. Moreover, note that otherwise would not be an independent set, then , therefore, . ∎
Theorem 4.7 is one of the main results of this section. It gives a closed formula for the independence number of unicyclic graphs of Type . This formula depends on the support and -vertices of subtrees. It means that using this formula we can compute the independence number of unicyclic graphs of Type using linear algebra.
Theorem 4.7**.**
If is a unicyclic graph of Type and its pendant tree such that then
[TABLE]
Proof.
Note that there is an independent set such that . Indeed, if then by lemmas 4.5 and 4.6 there is a such that . Let . Let . We will prove that .
We notice that is an independent set in . To see that, we observe that the vertices of are not connected to each other, because is an independent set. Similarly, we conclude the same for . Moreover, the only adjacencies between and occur between vertices and and vertices and . Since , there is no possibility of adjacency between vertices of and vertices of . Suppose that , that is, there exists an independent set in such that . As there is a and such that . We have that
[TABLE]
In this case, we see that and are independent sets in and , respectively. Thus we have and , because and . Therefore, , which is a contradiction by (1). Hence, . By Lemma 4.3, we have
[TABLE]
Therefore, the independence number of is given by
[TABLE]
∎
The following example is an application of Theorem 4.7. Consider the unicyclic graph of Figure 4. We observe that is of Type . Indeed, , then by Proposition 3.3 we have that is a unicyclic graph of Type . Moreover, , and .
Therefore, by Theorem 4.7, we have that the independence number of is given by:
[TABLE]
We observe that is a maximum independent set of and .
Lemma 4.8**.**
Let be a unicyclic graph and its cycle. Let be a pendant tree such that . If then .
Proof.
Let , where is a connected component of . Let . As there is a such that does not saturate by Lemma 2.7. We observe that , then , because is an independent set of by Theorem 2.5. Then and . Suppose , that is, there is a such that , then by Lemma 2.7 in we obtain a does not saturate . Note that, and does not saturate . Then , because . Which is a contradiction, because does not saturate and all maximum matching in saturates because . Therefore, . Since the connected components of are connected components of we have . ∎
Theorem 4.9 is a similar result for unicyclic graphs of Type and gives a closed formula for the independence number of unicyclic graphs of Type .
Theorem 4.9**.**
Let be a unicyclic graph and its cycle. Let , where is a connected component of . If is a unicyclic graph of Type then
[TABLE]
Proof.
Let and such that . By Lemma 4.8, we have . As then, by lemmas 4.5 and 4.6, we have a such that . Consider and define . We will show that .
First, we notice that is an independent set in . Indeed, for all we observe that the vertices of are not connected to each other, because is an independent set. Similarly, we conclude the same for . Moreover, there is no chance that a vertex of is adjacent to a vertex , with , since the vertices of the trees and are not adjacent to each other. Now, we show that vertices of are not connected to vertices of . To see that, we observe that the only adjacency that exists between and is the adjacency between vertex and , but since , there is no possibility of adjacency between vertices of and .
Suppose now , that is, there is an independent set in such that . As , we see that there exist and such that . Thus we have
[TABLE]
As and are independent sets of and , respectively, we have and , because and . Thus, we have , which is a contradiction by (2). Therefore, . We observe that and, by Lemma 4.3, we have, for all ,
[TABLE]
Therefore, the independence number of is given by
[TABLE]
∎
As an example, consider the unicyclic graph of Figure 5. We first notice that is a unicyclic graph of Type , because , and . Then by Corollary 3.4 we obtain is a unicyclic graph of Type . Moreover, we have , where , , , and (see Figure 5).
Since , and have perfect matching, so they are non-singular and have empty support. and do not have perfect matching, so they are singular and computing their supports we obtain the Table 1.
Therefore, by Theorem 4.9 we have that the independence number of is given by:
[TABLE]
We observe that is a maximum independent set of and .
5. Matching number of Unicyclic graphs
In this section we obtain closed formulas for the matching number of unicyclic graphs.
Definition 5.1**.**
Let be a matching in the graph . An -alternating path is a path that alternates edges in and edges that are not in . An -augmenting path is an -alternating path, if it begins and ends at vertices non saturated by .
Consider the matching in graph of Figure 6. As the path is an -alternating path because its edges alternate outside and within the matching . Now the path is -augmenting, because it is an -alternating path and starts and ends at vertices non saturated by (vertices u and o).
The following is a classic result, it characterizes maximum matchings in a graph .
Lemma 5.2**.**
(Berge, 1957)[4] A matching is maximum in if and only if does not have an -augmenting path.
We now give a closed formula for the matching number of unicyclic graphs of Type . This formula depends on the core and -vertices of subtrees.
Theorem 5.3**.**
If is a unicyclic graph of Type and its pendant tree such that then
[TABLE]
Proof.
Let and . Let and . We will prove that . Suppose that . Hence, by Lemma 5.2, there is an -augmenting path, denoted by . Notice that is neither totally contained in nor totally contained in , otherwise would be -augmenting or -augmenting, which is a contradiction, because and , respectively. Moreover does not contain the edges and the edge simultaneous, because . Then starts at a vertex of and ends at a vertex of . Now we notice that the path contains the edge or the edge . Suppose that contains the edge , then we have (see Figure 7).
Let and . We observe that and . Let be a matching in given by . As is a matching in , we see that it does not saturate , because Hence (see Lemma 2.7). That is . We have
[TABLE]
By equation (3) we have , which is a contradiction.
The case where contains the edge is analogous. Therefore . Using Lemma 4.3 we have
[TABLE]
Therefore, the matching number of is given by
[TABLE]
∎
As an example of Theorem 5.3, consider the unicyclic graph of Figure 8. To see that it is of Type , we notice that , then by Proposition 3.3, we have is a unicyclic graph of Type . Moreover, , , , and .
Therefore, by Theorem 5.3, we have that the matching number of is given by
[TABLE]
We point out that is a maximum matching of and .
We now present a similar result for the matching number of unicyclic graphs of Type .
Theorem 5.4**.**
Let be a unicyclic graph and its cycle. Let , where is a connected component of . If is a unicyclic graph of Type then
[TABLE]
Proof.
For each pendant tree , with , choose an that does not saturate . Note that this maximum matching exists because is a unicyclic graph of Type , and this implies that for all . Hence, by Lemma 2.7, we have that exists. Choose an and let .
We will show that . Suppose by contradiction that . Then, by Lemma 5.2 there exists a -augmenting path denoted by in . Note that if then would be a -augmenting path in which is a contradiction because . Now, if then would be a -augmenting path in , which is a contradiction because .
Note that the only way to obtain an -augmenting path in is if we start the path at a vertex , with , and end at a vertex . If that happens, there would be an -alternating path starting in and ending in contained in . But since does not saturate , actually we would obtain a -augmenting path in , which is a contradiction, because . Therefore, .
By Lemma 4.3, we have
[TABLE]
Moreover, we have . Since is not saturated by in , we see that , that is, . Therefore, we have that the matching number is given by:
[TABLE]
∎
Consider the unicyclic graph of Figure 9. We see that is a unicyclic graph of Type , because , , , and . By Corollary 3.4, we have that is of Type . Notice that , where , , and (see Figure 9). We see that has perfect matching, then is non singular and so has empty support. Moreover, we have that , and do not have perfect matchings, then they are singular and their supports is given in Table 2.
Therefore, by Theorem 5.4, we have that the matching number of is given by:
[TABLE]
we point out that is a maximum matching of and .
Acknowledgments
Work supported by MATHAMSUD 18-MATH-01. Maikon Toledo thanks CAPES for their support. V. Trevisan acknowledges partial support of CNPq grants 409746/2016-9 and 303334/2016-9, and FAPERGS (Proj. PqG 17/2551-0001).
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] N. Alon and R. B. Boppana. The Monotone Circuit Complexity of Boolean Functions. Combinatorica , 7:1-22, 1987.
- 2[2] N. Alon and N. Kahale. Approximating the Independence Number via the J -function. Mathematical Programming , 80:253-264, 1998.
- 3[3] R. B. Bapat. Graph and Matrices. In Universitext. SBMAC- Springer, 2010. ISBN: 978-1-84882-980-0.
- 4[4] C. Berge. Two theorems in graph theory. Proceedings of the National Academy of Sciences 43(9): 842-844, 1957.
- 5[5] D. M. Cvetkovć, M. Doob and H. Sachs. Spectra of graphs: theory and application, vol. 87. Academic Pr, 1980.
- 6[6] A. M. Frieze. On the Independence Number of Random Graphs. Discrete Mathematics , 81:171-175, 1990.
- 7[7] S. Gong, Y. Fan and Z. Yin. On the nullity of graphs with pendant trees. Linear Algebra and its Applications , 433:1374-1380, 2010.
- 8[8] D. A. Jaume and G. Molina. Null Decomposition of Trees. Discrete Mathematics , 341:836-850, 2018.
