1. Introduction
Chemical compounds (like hydrocarbons) can be represented by means of graphs. A topological index
T is just a number that encapsulates some property of graphs and such that
T correlates with a certain molecular characteristic; therefore, it can be employed to grasp chemical properties and physical properties of chemical substances. They play an important role in mathematical chemistry, in particular, in the QSPR/QSAR (quantitative structure-property relationship/quantitative structure-activity relationship) investigations. Computational and mathematical properties of topological indices have been studied in depth for more than 50 years (see, e.g., [
1,
2,
3,
4,
5,
6,
7,
8] and the references therein). In particular, a main subject on this field is to obtain sharp bounds of topological indices.
See [
9] for a review in a dialog manner discussing relevance of topological descriptors to chemical/physical properties.
One of the main topological indices is the following index, called
Randić index, defined in [
1] by
where
G is a graph and
is the degree of the vertex
.
Along the paper, will denote a (non-oriented) simple (without loops and multiple edges loops) finite graph without isolated vertices. Hence, every vertex has degree at least 1.
Two of the most popular alternatives to the Randić index are the
first Zagreb and second Zagreb indices, denoted by
and
, respectively, and defined by
These two indexes are very useful in mathematical chemistry and so, they have been extensively studied, see [
10,
11,
12,
13] and the references therein. Further development of Zagreb-type indices deals with the applications to more complex chemical objects, e.g., large carbon-based species with regular structures such as polycyclic aromatic hydrocarbons [
14] and carbon nanostructures [
15].
Please note that there are topological indices of different types. They may treat only vertices, only edges, or both edges and vertices of the graph to calculate an index. Thus, the first Zagreb index belongs to the first class (every index in the first class, as the first Zagreb index, also belongs to the second class).
Along this paper we obtain results for topological indices in this first class.
The so-called
harmonic index, is defined in [
16] by
For more information about the properties of that index we refer to [
17,
18,
19,
20,
21,
22], and the book [
23].
The
inverse degree index is defined as
This index first attracted attention through many conjectures obtained by the computer programme called Graffiti [
16]. Since then, several authors have studied its connections with other parameters of graphs: diameter, matching number, edge-connectivity, Wiener index, etc.; also, its chemical applications have been studied by many researchers (see [
24,
25,
26,
27,
28,
29]).
In [
30,
31,
32] the
general first Zagreb and general second Zagreb indices are defined by
Please note that the first Zagreb index is , the inverse degree index is , the forgotten index F is , …; moreover, the Randić index R is , the second Zagreb index is , the modified Zagreb index is , ….
The variable topological indices were introduced as a new way of characterizing heteroatoms (see [
33,
34]), and to assess the structural differences (see [
35]). The idea behind the variable topological indices is that the parameter is determined during the regression in such a way that the error of estimate for a fixed property is minimized.
The
sum lordeg index was introduced in [
36]. It is defined as
This index is interesting from an applied viewpoint since it correlates very well with the octanol-water partition coefficient for octane isomers [
36], and so, it appears in numerical packages for the computation of topological indices [
37]. For these reasons, in [
38] is stated the open problem of finding appropriate bounds for this index.
In [
39] the
harmonic polynomial is introduced as
In [
40,
41,
42] several properties of the harmonic polynomial are obtained. Please note that
.
In [
41] the
inverse degree polynomial is introduced as
It should be noticed that . Thus, the inverse degree polynomial can be used to obtain information about the inverse degree index of a graph G.
Given any function
, let us define the
f-
index
and the
f-
polynomial of
G by
if
. Also, let us define
. In particular,
if
. It is clear that
.
Please note that many important indices can be obtained from by choosing appropriate functions f: if , then is the first Zagreb index; if , then is the inverse degree index ; if , then is the forgotten index F; in general, if , then is the general first Zagreb index ; if , then is the sum lordeg index . Thus, each theorem in this paper about is a result for each one of these indices.
The
f-polynomial of other graph operations (e.g., join, corona product, Mycielskian and line) is studied in [
43].
Operations on graphs play an important role in Mathematical Chemistry, see e.g., [
44,
45], since many chemical structures appear as operations of graphs: The crystal structure of sodium chloride is the Cartesian product of two path graphs. The kernel of the iron crystal structure is the join of the cube graph
and an isolated vertex. The cyclobutane is the Cartesian product of two path graphs
. The alkane
can be represented as the corona product of the path graph
and the null graph
. The cyclohexane
can be represented as the corona product of the cycle graph
and the null graph
. The carbon nanotube
can be seen as the Cartesian product of the path graph
and the cycle graph
. The fence (respectively, the closed fence) is the lexicographic product of
and
(respectively,
and
). The zigzag polyhex nanotube
is the generalized hierarchical product of the path graph
and the cycle graph
. See [
46,
47].
Polynomials, in general, have lately proved to be useful in graph theory and, in particular, in Mathematical Chemistry (see [
41,
43,
45,
48,
49,
50,
51,
52,
53,
54,
55]).
The main goal in this paper is to obtain information of many topological indices (each case of for some particular choice of f) of several graph products, from the information on topological indices of these factors, which are much easier to calculate than the products. Our approach is to obtain information about the corresponding f-polynomials, which are easy to calculate (see for instance Theorems 1, 11 and 17); then, we can deduce information on the index by using the formula (see for instance Theorems 4, 13 and 18). This is a good approach since the bounds of the f-polynomial of a product of two graphs allow use of analytic tools to bound the index of such a graph product, simplifying the proofs.
2. Background
The study of the effect of graph operations on topological indices is an active topic of research (see, e.g., [
41,
54,
55,
56]). We study in this section the
f-polynomial of several graph products: lexicographic product, Cartesian sum and Cartesian product.
The Cartesian product of the graphs and has the vertex set and is an edge of if and , or and .
The lexicographic product of the graphs and has as vertex set, so that two distinct vertices of are adjacent if either , or and .
The Cartesian sum of the graphs and has the vertex set and is an edge of if or .
In [
57] is defined the following
Zagreb polynomialPlease note that .
In ([
43], Propositions 1, 2 and 3) appear the following useful results.
Proposition 1. If G is a graph of order n and , then:
is a polynomial if and only if for every ,
is a positive function on ,
is a continuous function on if and only if ,
is a continuous function on if and only if for every ,
is an integrable function on for every , and ,
is increasing on if and only if for every ,
is strictly increasing on if and only if for every , and for some ,
is convex on if for every ,
is strictly convex on if for every , and for some ,
is concave on if for every ,
is strictly concave on if for every , and for some ,
.
If and , then Proposition 1 gives .
Proposition 2. If G is a k-regular graph with n vertices and , then
Next, we show the polynomial for some important graphs: (complete graph), (cycle graph), (hypercube graph), (complete bipartite graph), (star graph), (path graph), and (wheel graph).
Choose
and
.
f satisfies the
-
additive property 1 (
) if
for every
with
.
f satisfies the
-
additive property 2 (
) if
for every
with
.
f satisfies the
-
additive property 3 (
) if
for every
with
.
3. Inequalities for Cartesian Products
First, we prove pointwise inequalities of involving and .
Theorem 1. Let , and let and be two graphs with minimum degree at least δ. For , the f-polynomial of the Cartesian product satisfies.
If , then If , then If and and have and vertices, respectively, then Proof. If
, we have
and its corresponding monomial of the
f-polynomial is
Assume that
. Since
for every
,
and
,
If , then by a quite similar argument the result is obtained.
If
, then
The inequality involving is obtained in a similar way. □
Remark 1. If , then the condition and have minimum degree at least δ, is satisfied for every graph .
Theorem 1 has the following consequence when we consider .
Theorem 2. Let and be two graphs of order and , respectively, and . For , the f-polynomial of the Cartesian product satisfies.
If , then and If , then and If , then and Proof. The inequalities are consequences of the following facts and Theorem 1.
If
, then
and
for every
, and so,
.
If
, then
and
for every
, and so,
.
If
, then
is a decreasing function on
and
for every
, and so,
. □
Theorem 2 yields for the inverse degree polynomial:
Corollary 1. Let and be two graphs, the ID polynomial of the Cartesian product is Since is an increasing function on , and Theorem 1 has the following consequence. Recall that a vertex with degree 1 is called pendant vertex. Please note that is a positive function on , and so, is well-defined for every graph G without pendant vertices.
Please note that a graph has minimum degree at least 2 if and only if it does not have pendant vertices.
Theorem 3. Let and be two graphs with minimum degree at least 2 and of order and , respectively. If , then and the f-polynomial of the Cartesian product satisfies for Next, we obtain bounds for by using the previous inequalities for . This is a good approach since the pointwise bounds of allow use of analytic tools to bound . We start with the case .
Theorem 4. Let , and be two graphs of order and , respectively, and minimum degree at least δ. If , then Proof. Theorem 1 gives
for every
. Thus, Proposition 1 leads to
The same argumentation gives . □
To deal with the cases and , we need some technical results.
Lemma 1. [58] If are non-negative convex functions on , then Lemma 1 can be slightly improved as follows.
Proposition 4. If are convex non-negative functions on , then Proof. If
, then
are convex non-negative functions on
, and Lemma 1 gives
Since
are non-negative on
, the functions
defined by
are decreasing, and so, there exist their limits as
(although they can be
∞). By taking
in the above inequality, we obtain the result. □
Lemma 2. ([59] Corollary 5.2) If are convex non-negative functions on , then Since any non-negative concave function on
has finite lateral limits at
a and
b, ([
59], Corollary 4.3) can be stated as follows.
Lemma 3. If are non-negative concave functions on , then With these technical results, we can deal now with the cases and .
Theorem 5. Let , and be graphs of orders and , respectively, and minimum degree at least δ. If the f-polynomials of and are convex functions on , then
If , then If , then Proof. Assume first that
. Theorem 1 gives
for every
.
If or , then trivially holds.
If
and
, then
and
are continuous functions on
by Proposition 1, and so, they are convex on
. Proposition 1 and Lemma 2 give
Assume now that
. Theorem 1 gives
for every
. Since
and
are convex functions on
, Proposition 1 and Proposition 4 give
□
If the f-polynomials of and are concave functions on , we can obtain simpler bounds for .
Theorem 6. Let , and be two graphs with minimum degree at least δ. If and the f-polynomials of and are concave functions on , then Proof. Since
, Theorem 1 gives
for every
. Since
and
are concave functions on
, a simple combination of Proposition 1 and Lemma 3 yields
□
We can obtain more bounds for if the image of f does not intersect for some constant .
Theorem 7. Let , and be two graphs of orders and , respectively, and minimum degree at least δ, and . If , then
If , then If , then Proof. Let be the function . Hence, and Proposition 1 gives that and are convex functions on .
If
, then
and Theorem 5 gives
If
, then
and Theorem 5 gives
□
The first inequality in Theorem 7 can be improved as follows.
Theorem 8. Let , and be two graphs of orders and , respectively, and minimum degree at least δ, and . If , and , then Proof. Since on , we have and , and so, we obtain the result. □
Theorems 2, 8 (with ), 7 (with ) and 4 have the following consequence for the general first Zagreb index.
Theorem 9. Let and be two graphs of orders and , respectively, and .
If , then As a rather simple consequence of the above theorem, we have for the first Zagreb, forgotten and inverse degree indices:
Corollary 2. If and are two graphs with and vertices, respectively, then Since , Theorem 4 gives the following result for the index.
Theorem 10. If and are graphs without pendant vertices and with and vertices, respectively, then A particular consequence of Theorems 9 and 10 is the following.
Corollary 3. If and are the cycle graphs with and vertices, respectively, then 4. Inequalities for Lexicographic Products
We start this section by proving pointwise inequalities of involving the f-polynomials of and .
Theorem 11. Let , and be two graphs of orders and , respectively, and minimum degree at least δ. The f-polynomial of the lexicographic product satisfies the following inequalities for .
If , then If , then If , then Proof. If , then .
Assume that
. Since
for every
, one can prove by induction that
If , then same argument allows obtaining of the result.
If
, then
A similar argument gives . □
Theorems 2 and 11 have the following consequence when .
Proposition 5. Let and be two graphs with and vertices, respectively, and . The f-polynomial of the lexicographic product satisfies the following inequalities for .
If , then Theorem 5 gives the following equality for the inverse degree polynomial.
Corollary 4. Given two graphs and , of order and , respectively, the ID polynomial of the lexicographic product is Since , Theorem 11 has the following consequence.
Theorem 12. Let and be two graphs with minimum degree two and of order and , respectively. If , then the f-polynomial of the lexicographic product satisfies for Next, we obtain bounds for by using the previous inequalities for . We start with .
Theorem 13. Let , and be two graphs of order and , respectively, and minimum degree at least δ. If , then Proof. Theorem 11 gives
for every
. Thus, Proposition 1 leads to
. A similar argument gives the inequality
. □
We deal now with .
Theorem 14. Let , and be two graphs of orders and , respectively, and minimum degree at least δ, and . If , then the following inequalities hold.
If , then If , then Proof. Consider the function . We have and so, Proposition 1 implies that and are convex functions on the open interval and continuous on the closed interval ; hence, they are convex when .
If
, then
and Theorem 11 implies
Notice that
implies
, and thus,
and
for
. Since
x is a convex function when
, Lemma 2 implies
If
, thus
and Theorem 11 implies
Now, we deduce several inequalities for many topological indices of lexicographic products.
Theorems 2, 14 (with ) and 13 have the following consequence for the variable first Zagreb index.
Theorem 15. Let and be two graphs of orders and , respectively, and .
If , then Theorem 15 has the following consequence for the first Zagreb, forgotten and inverse degree.
Corollary 5. If and are two graphs with and vertices, respectively, then Since , Theorem 13 allows deduction of the following result for the index.
Theorem 16. If and are graphs without pendant vertices and with and vertices, respectively, then 5. Inequalities for Cartesian Sums
We start this last section by proving pointwise inequalities of involving the f-polynomials of and .
Theorem 17. Let , and be two graphs of orders and , respectively, and minimum degree at least δ. For , the f-polynomial of the Cartesian sum satisfies.
If , then If , then If , then Proof. If , then .
Suppose that
. Since
for every
, one can prove by induction that
If , then a similar argument allows obtaining of the corresponding inequality.
Suppose that
. We deduce
A similar argument gives . □
Theorems 2 and 17 have the following consequence for .
Proposition 6. Let and be two graphs of orders and , respectively, and . For , the f-polynomial of the Cartesian sum satisfies the following inequalities for.
If , then Proposition 6 allows deduction of the following equality for the inverse degree polynomial.
Corollary 6. Given two graphs and , or order and , respectively, the ID polynomial of the Cartesian sum is Since , Theorem 17 has the following consequence.
Corollary 7. Let and be two graphs with minimum degree two and of order and , respectively. If , then the f-polynomial of the Cartesian sum satisfies for Next, we obtain bounds for by using the previous inequalities for . We start when .
Theorem 18. Let , and be two graphs of orders and , respectively, and minimum degree at least δ. If , then Proof. Theorem 17 gives
for every
. Hence,
by Proposition 1. A similar argument gives the inequality
. □
We consider now the case .
Theorem 19. Let , and be two graphs of orders and , respectively, and minimum degree at least δ, and . If , then the following inequalities hold.
If , then If , then Proof. Consider the function . Therefore, and Proposition 1 implies that and are convex functions on the open interval and continuous on the closed interval ; hence, they are convex when .
If
, then the function
belongs to
and Theorem 17 implies
Notice that
implies
, and thus,
and
for
. Since
x is a convex function on the interval
, Lemma 2 implies
If
, thus
belongs to
and Theorem 17 implies
Theorems 2, 19 (with ) and 18 have the following consequence for the general first Zagreb index.
Theorem 20. Let and be two graphs of orders and , respectively, and .
If , then Theorem 20 has the following consequence for the first Zagreb, forgotten and inverse degree indices.
Corollary 8. If and are two graphs of orders and , respectively, then Since , Theorem 18 implies the following result for the index.
Theorem 21. If and are graphs without pendant vertices and of order and , respectively, then We obtain the following result by using the previous ideas.
Lemma 4. Let G be a graph and Γ a subgraph of G, an increasing function, and . Then If f is a decreasing function and , then we obtain the converse inequalities.
Lemma 4 has the following consequence, relating the polynomials and indices of Cartesian products, lexicographic products and Cartesian sums.
Proposition 7. Let and be two graphs, an increasing function, and . Then If f is a decreasing function, then we obtain the converse inequalities.
6. Conclusions
There are several graph products which play an important role in graph theory. Three of these products are Cartesian product, lexicographic product and Cartesian sum. Many topological indices can be written as
for an appropriate choice of the function
f (e.g., first Zagreb, inverse degree, forgotten, general first Zagreb and sum lordeg indices). By using the
f-polynomial
introduced in [
43], we obtain in this paper several inequalities of every topological index which can be written as
for a function
f in these classical graph products, from the information on topological indices of their factors, which are much easier to calculate than the products. These results are interesting from the theoretical viewpoint, and also from the point of view of applications since many chemical compounds can be represented by graph products (see the introduction). Our approach is to obtain information about the corresponding
f-polynomials, which are easy to calculate (as in Theorems 1, 11 and 17); thus, we can deduce information on the
index by using the formula
(as in Theorems 4, 13 and 18). This is a good approach since the bounds of the
f-polynomial of a product of two graphs allow the use of analytic tools to bound the
index of such a product, simplifying the proofs.
In [
43] appear similar results for corona product and join. Consequently, two natural open questions are to study this problem for strong product and tensor product of graphs.