1. Introduction
A single number representing a chemical structure, by means of the corresponding molecular graph, is known as topological descriptor. Topological descriptors play a prominent role in mathematical chemistry, particularly in studies of quantitative structure–property and quantitative structure–activity relationships. Moreover, a topological descriptor is called a topological index if it has a mutual relationship with a molecular property. Thus, since topological indices encode some characteristics of a molecule in a single number, they can be used to study physicochemical properties of chemical compounds.
After the seminal work of Wiener [
1], many topological indices have been defined and analysed. Among all topological indices, probably the most studied is the Randić connectivity index (
R) [
2]. Several hundred papers and, at least, two books report studies of
R (see, for example, [
3,
4,
5,
6,
7] and references therein). Moreover, with the aim of improving the predictive power of
R, many additional topological descriptors (similar to
R) have been proposed. In fact, the first and second Zagreb indices,
and
, respectively, can be considered as the main successors of
R. They are defined as
where
is the edge of
G between vertices
u and
v, and
is the degree of vertex
u. Both
and
have recently attracted much interest (see, e.g., [
8,
9,
10,
11]) (in particular, they are included in algorithms used to compute topological indices).
Another remarkable topological descriptor is the
harmonic index, defined in [
12] as
This index has attracted a great interest in the lasts years (see, e.g., [
13,
14,
15,
16,
17,
18]). In particular, in [
16] appear relations for the harmonic index of some operations of graphs.
In [
19], the
harmonic polynomial of a graph
G is defined as
and the harmonic polynomials of some graphs are computed. For more information on the study of polynomials associated with topological indices and their practical applications, see, e.g., [
20,
21,
22,
23].
This polynomial owes its name to the fact that .
The characterization of any graph by a polynomial is one of the open important problems in graph theory. In recent years, there have been many works on graph polynomials (see, e.g., [
21,
24] and the references therein). The research in this area has been largely driven by the advantages offered by the use of computers: it is simpler to represent a graph by a polynomial (a vector with dimension
) than by the adjacency matrix (an
matrix). Some parameters of a graph allow to define polynomials related to a graph. Although several polynomials are interesting since they compress information about the graphs structure; unfortunately, the well-known polynomials do not solve the problem of the characterization of any graph, since there are often non-isomorphic graphs with the same polynomial.
Polynomials have proved to be useful in the study of several topological indices. There are many papers studying topological indices on graph operations (see, e.g., [
25,
26,
27]).
Along this work,
indicates a finite, undirected and simple (i.e., without multiple edges and loops) graph with
. The main aim of this paper is to obtain several computational properties of the harmonic polynomial. In
Section 2, we obtain closed formulas to compute the harmonic polynomial of many classical symmetric operations of graphs: Cartesian product, corona product, join, Cartesian sum and lexicographic product. These formulas are interesting by themselves and, furthermore, allow to obtain new inequalities for the harmonic index of these operations of graphs. Besides, we provide in the last section an algorithm that computes this polynomial with complexity
.
We would like to stress that the symmetry property present in the operations on graphs studied here (Cartesian product, corona product, join, Cartesian sum and lexicographic product) was an essential tool in the study of the topological indexes, because it allowed us to obtain closed formulas for the harmonic polynomial and to deduce the optimal bounds for that index.
2. Definitions and Background
The following result appears in Proposition 1 of [
19].
Proposition 1. If G is a k-regular graph with m edges, then
Propositions 2, 4, 5, 7 in [
19] have the following consequences on the graphs:
(the complete graph with
n vertices),
(the cycle with
vertices),
(the
n-dimensional hypercube),
(the complete bipartite graph with
vertices),
(the path graph with
n vertices), and
(the wheel graph with
vertices).
In Propositions 2.3 and 2.6 in [
28] appear the following result.
Proposition 3. If G is a graph with m edges, then:
for every and ;
on and is strictly increasing on ;
is strictly convex on if and only if G is not isomorphic to a union of path graphs ; and
for every .
Considering the Zagreb indices, Fath-Tabar [
29] defined the first Zagreb polynomial as
The harmonic and the first Zagreb indices are related by several inequalities (see [
30], Theorem 2.5 [
31] and [
32], p. 234). Moreover, the harmonic and the first Zagreb polynomials are related by the equality
,
In [
33], Shuxian defined the following polynomial related to the first Zagreb index as
Given a graph
G, let us denote by
its
subdivision graph.
is constructed from
G by inserting an additional vertex into each of its edges. Concerning
, in Theorem 2.1 of [
25], the following result appears.
Theorem 1. For the subdivision graph of G, the first Zagreb polynomial is Since the harmonic and the first Zagreb polynomials are related by the equality , we have the following result for the harmonic polynomial of the subdivision graph.
Proposition 4. Given a graph G, the harmonic polynomial of its subdivision graph is Similarly, we can obtain the harmonic polynomial for the other operations on graphs appearing in [
25].
Next, we obtain the harmonic polynomial for other classical operations: Cartesian product, corona product, join, Cartesian sum and lexicographic product. It is important to stress that, since large graphs are composed by smaller ones by the use of products of graphs (and, as a consequence, their properties are strongly related), the study of products of graphs is a relevant and timely research subject.
Let us recall the definitions of these classical products in graph theory.
The Cartesian product of the graphs and has the vertex set and is an edge of if and , or and .
Given two graphs and , we define the corona product as the graph obtained by adding to , copies of and joining each vertex of the i-th copy with the vertex .
The join is defined as the graph obtained by taking one copy of and one copy of , and joining by an edge each vertex of with each vertex of .
The Cartesian sum of the graphs and has the vertex set and is an edge of if or .
The lexicographic product of the graphs and has as vertex set, so that two distinct vertices of are adjacent if either , or and .
Let us introduce another topological index that will be very useful in this work.
The
inverse degree of a graph
G is defined by
It is relevant to mention that the surmises inferred through the computer program Graffiti [
12] attracted the attention of researchers. Thus, since then, several studies (see, e.g., [
34,
35,
36,
37,
38]) focusing on relationships between
and other graph invariants (such as diameter, edge-connectivity, matching number and Wiener index) have appeared in the literature.
Let us define the
inverse degree polynomial of a graph
G as
Thus, we have . Note that .
The following result summarizes some interesting properties of the inverse degree polynomial. Recall that a vertex of a graph is said to be pendant if it has degree 1.
Proposition 5. If G is a graph with n vertices and k pendant vertices, then:
for every and ;
on ;
is strictly increasing on if and only if G is not isomorphic to a union of path graphs ;
is strictly convex on if and only if G is not isomorphic to a union of path graphs; and
for every .
Proof. Since every coefficient of the polynomial is non-negative, the first statement holds.
Since every coefficient of the polynomial is non-negative and is not identically zero, we have on .
Since every coefficient of the polynomial is non-negative, we have on if and only if there exists a vertex with , and this holds if and only if G is not isomorphic to a union of path graphs .
Similarly, is strictly convex on if and only if there exists a vertex with , and this holds if and only if G is not isomorphic to a union of path graphs.
Finally, if
, then
☐
Proposition 4 has the following consequence, which illustrates how these polynomials associated to topological indices provide information about the topological indices themselves.
Corollary 1. Given a graph G with maximum degree Δ
, the harmonic index of the subdivision graph satisfies 3. Computation of the Harmonic Index of Graph Operations
Let us start with the formula of the harmonic polynomial of the Cartesian product.
Theorem 2. Given two graphs and , the harmonic polynomial of the Cartesian product is Proof. Denote by and the cardinality of the vertices of and , respectively.
Note that if , then .
If
, then the corresponding monomial of the harmonic polynomial is
The same argument gives that the sum of the monomials corresponding to is , and the equality holds. ☐
Next, we present two useful improvements (for convex functions) of the well-known Chebyshev’s inequalities.
Lemma 1 ([
39]).
Let be non-negative convex functions defined on the interval . Then, Lemma 2 (Corollary 5.2 [
40]).
Let be non-negative convex functions defined on the interval . Then Theorem 3. Given two graphs and with and vertices, and and edges, respectively, the harmonic index of the Cartesian product satisfies Proof. Propositions 3 and 5 give that
are non-negative convex functions. Thus, Lemma 1 gives
These inequalities, Theorem 2 and give the lower bound.
Lemma 2 and Propositions 3 and 5 give
In addition, Lemma 2 and Propositions 3 and 5 give
These inequalities, Theorem 2 and give the upper bound. ☐
Theorem 4. Given two graphs and , with and vertices, respectively, the harmonic polynomial of the corona product is Proof. The degree of , considered as a vertex of , is . The degree of any copy of , considered as a vertex of , is .
If
, then the corresponding monomial of the harmonic polynomial of
is
If
, then each corresponding monomial of the harmonic polynomial of
is
If we add the corresponding polynomials of the copies of , then we obtain .
If
with
and
, then the corresponding monomial of the harmonic polynomial is
Thus, the equality holds. ☐
Theorem 5. Given two graphs and with and vertices, and edges, and and pendant vertices, respectively, the harmonic index of the corona product satisfies Proof. These inequalities, Theorem 4 and give the lower bound.
Lemma 2 and Proposition 3 give
In addition, Lemma 2 and Proposition 3 give
Lemma 2 and Proposition 5 give
These inequalities, Theorem 4 and give the upper bound. ☐
Theorem 6. Given two graphs and , with and vertices, respectively, the harmonic polynomial of the join is Proof. The degree of , considered as a vertex of , is . The degree of , considered as a vertex of , is .
If
, then the corresponding monomial of the harmonic polynomial of
is
If
, then the corresponding monomial of the harmonic polynomial of
is
If
with
and
, then the corresponding monomial of the harmonic polynomial is
Thus, the equality holds. ☐
Theorem 7. Given two graphs and with and vertices, and edges, and and pendant vertices, respectively, the harmonic index of the join satisfies Proof. We have seen in the proof of Theorem 5 that
Lemma 2 and Proposition 5 give
These inequalities, Theorem 6 and give the bounds. ☐
Theorem 8. Given two graphs and , with and vertices, respectively, the harmonic polynomial of the Cartesian sum is Proof. Note that if , then .
If
, then the corresponding monomial of the harmonic polynomial is
Hence, the sum of the corresponding monomials with
is
Similarly, the sum of the corresponding monomials with
is
If we add these two terms, then we take into account twice the corresponding monomials with
and
:
Hence, the equality holds. ☐
Theorem 9. Given two graphs and with and vertices, and and edges, respectively, the harmonic index of the Cartesian sum satisfies Proof. Lemma 2 and Propositions 3 and 5 give
In addition, Lemma 2 and Proposition 3 give
These inequalities, Theorem 8 and give the desired bounds. ☐
Theorem 10. Given two graphs and , with and vertices, respectively, the harmonic polynomial of the lexicographic product is Proof. Note that if , then .
If
, then the corresponding monomial of the harmonic polynomial is
If
with
, then the corresponding monomial of the harmonic polynomial is
Hence, the sum of their corresponding monomials is
We obtain the desired equality by adding these two terms. ☐
Theorem 11. Given two graphs and with and vertices, and edges, and and pendant vertices, respectively, the harmonic index of the lexicographic product satisfies Proof. Lemma 2 and Propositions 3 and 5 give
These inequalities, Theorem 10 and give the bounds. ☐