Chapter 1

Basic Chemical Graph Theory

Graph Theory applied in Chemistry is called Chemical Graph Theory. This inter-
disciplinary science takes problems (like isomer enumeration, structure elucidation,
etc.) from Chemistry and solve them by Mathematics (using tools from Graph
Theory, Set Theory or Combinatorics), thus influencing both Chemistry and Math-
ematics. Partitioning of a molecular property and reconstructing it from fragmental
contributions is one of the main tasks of this theory. For further discussion, some
basic definitions in Graph Theory are needed.

1.1 Basic Definitions in Graphs

A graph G(V, E) is a pair of two sets, V and E, V ¼ V(G) being a finite nonempty set
and E ¼ E(G) a binary relation defined on V (Harary 1969). A graph can be
visualized by representing the elements of V by points/vertices and joining pairs
of vertices (i, j) by an edge/bond if and only if (i, j) 2 E(G). The number of vertices
in G equals the cardinality n ¼ jV(G)j of this set. The term graph was introduced by
Sylvester (1874). There is a variety of graphs, some of them being mentioned
below.
A path graph is a non-branched chain. A tree is a branched structure. A star is a
set of vertices joined in a common vertex. A cycle is a chain which starts and ends in
one and the same vertex (Fig. 1.1).
A complete graph Kn is the graph of with any two vertices are adjacent. The
number of edges in such a graph is n(n  1)/2. Fig. 1.2 illustrates the complete
graphs with n ¼ 1–5.
In a bipartite graph, the vertex set V can be partitioned in two disjoint subsets:
V1 [ V2 ¼ V(G); V1 \ V2 ¼ ∅ such that any edge (i, j) 2 E(G) joins V1 with V2
(Harary 1969; Trinajstić 1983; Diudea 2010). A graph is bipartite if and only if
all its cycles are even (Ionescu 1973). If any vertex i 2 V1 is adjacent to any vertex
j 2 V2 then G is a complete bipartite graph, Km,n, with m ¼ |V1| and n ¼ |V2| and the

© Springer International Publishing AG 2018 1

M.V. Diudea, Multi-shell Polyhedral Clusters, Carbon Materials: Chemistry
and Physics 10, DOI 10.1007/978-3-319-64123-2_1
2 1 Basic Chemical Graph Theory

Path Tree Star Cycle

Fig. 1.1 A variety of graphs

K1 K2 K3 K4 K5

Fig. 1.2 Complete graphs Kn; n ¼ 1–5

K1,3 K2,3 K3,3

Fig. 1.3 Complete bipartite graphs

f2

f1 f3 f4

Fig. 1.4 A planar graph and its faces

number of edges being m  n. A star is a complete bipartite graph K1,n. Fig. 1.3
illustrates some complete bipartite graphs.
A planar graph is a graph that can be drawn in the plane with no edge
intersections (Harary 1969). The regions defined by a plane graph are called faces
f, the infinite region being the exterior face (see the face f4 in Fig. 1.4). For any
spherical polyhedron with v ¼ |V| vertices, e ¼ |E| edges and f ¼ |F| faces, the Euler
(1752–1753) formula is true: ve + f ¼ 2. A graph is planar if and only if it has no
subgraphs homeomorphic to K5 or K3,3 (Kuratowski 1930).
A subgraph of a graph G is a graph G1 ¼ (V1, E1) having V1  V and E1  E
(Fig. 1.5).
A homeomorph of a graph G is a graph resulted by inserting vertices of degree
2 (Fig. 1.6).
A graph G is labeled, G(Lb), when its points are distinguished (e.g. by their
numbers) from those of the corresponding abstract graph. There are n! possibilities
of numbering a graph of order n.
1.1 Basic Definitions in Graphs 3

Fig. 1.5 A graph and one of its subgraphs

Fig. 1.6 Homeomorphs of the tetrahedron

Two graphs G ¼ (V, E) and G1 ¼ (V1, E1) are isomorphic (written G ffi G1) if
there exists a function f : V ! V1 which is a bijection (one-to-one and onto) and for
any vertices i , j 2 V(G) (i, j) 2 E(G) $ ( f(i), f( j)) 2 E1(G1). The function f is called
an isomorphism. If f is the permutation operation, then there exists a permutation
for which G(Lb) and G1(Lb) coincide and then f is an automorphism.
A walk wn is an alternating sequence of vertices and edges v1, e1,..,vn1, en1, vn in
which any two subsequent vertices are adjacent: (vi, vi þ 1) 2 E(G). Revisiting of
vertices and edges of G is allowed. Set, in a walk wn, the vertices V(wn) and the
traversed edges E(wn); the length of the walk is defined as the cardinality of its edge
set: l(wn) ¼ |E(wn)|. If the walk starts and ends in one and the same vertex (i.e., v1 ¼ vn)
the walk is a closed (or a self-returning) walk, otherwise it is an open walk. If no other
conditions are imposed, the walk is called a random walk (Harary 1969; Diudea 2010).
A path pn (or a self-avoiding walk) is the walk whose vertices are visited once. Its
vertices v1, v2, . . ., vn are all-distinct and no branching is allowed. The length of the
path is l( pn) ¼ |E( pn)|. If the path joining i and j is minimal, it is called a topological
distance or a geodesic; if the path is maximal, it is called a detour.
A circuit/cycle is a closed path, thus being both self-returning and self-
avoiding walk.
A trail (i.e., Eulerian walk) is a walk with all its edges distinct. Revisiting of
vertices is allowed.
A path is Hamiltonian if it visits once all the vertices in G. If such a path is a
closed one, then it is a Hamiltonian circuit. Fig. 1.7 illustrates the above discussed
walks.
When l( pij) is expressed in number of edges, the distance is called topological
distance; when it is measured in meters or submultiples (nm, pm) it is a metric
distance. Vertex degree d(i) is the number of edges incident in that vertex. If all the
4 1 Basic Chemical Graph Theory

Closed walk Path Trail Cycle Hamiltonian Hamiltonian

path circuit

Fig. 1.7 A variety of walks

H
H
H C
H
C H

C H

H H
H
C C C
H H
H C H
H

Fig. 1.8 A molecular graph and its hydrogen depleted representation

vertices in G have the same degree, the graph is called a d-regular graph; otherwise
it is irregular.
An invariant of a graph is a graph theoretical property, which is preserved up to
isomorphism; it remains unchanged, irrespective of the numbering or pictorial
representation of G.
A chemical graph is a model of a chemical system, used to characterize the
interactions among its components: atoms, bonds, groups of atoms or molecules. It
is also called a reaction graph (Diudea et al. 2002, 2006).
A structural formula of a chemical compound can be represented by a molecular
graph, its vertices being atoms and edges corresponding to covalent bonds. Usually
hydrogen atoms are not depicted in which case we speak of hydrogen depleted
molecular graphs (Fig. 1.8).
A graph is said connected if any two vertices i and j are the endpoints of a path;
otherwise it is disconnected.
The distance dij between two vertices i and j is the length of a shortest path
joining them, if any: dij ¼ min l( pij); otherwise dij ¼ 1. A shortest path is often
called a geodesic (Harary 1969).
The eccentricity of a vertex i, ecci, is the maximum topological distance between
i and any vertex j of G : ecci ¼ max dij. The radius of a graph, r(G), is the minimum
eccentricity among all vertices i in G: r(G) ¼ min ecci ¼ min max dij. Conversely,
the diameter of a graph, d(G), is the maximum eccentricity in G: d(G) ¼ max
ecci ¼ max max dij.
The detour δij between two vertices i and j is the length of a longest path joining
these vertices, if any: δij ¼ max l( pij); otherwise δij ¼ 1.
1.2 Topological Matrices and Indices 5

In a connected, non-directed graph, the distance and the detour are metrics, that
is, for any vertices i, j and k, the following equivalence relations are true

mði; jÞ  0; mði; jÞ ¼ 0 $ i ¼ j
mði; jÞ ¼ mðj; iÞ
mði; jÞ þ mði; kÞ  mðj; kÞ

Within this book, only molecular graphs will be considered.

1.2 Topological Matrices and Indices

A molecular graph can be represented by a sequence of numbers, a polynomial, a

single number or a matrix. These representations are aimed to be unique, for a given
structure. Topological matrices can be accepted as a rational basis for designing
topological indices (Randić 1991).

1.2.1 Adjacency Matrix

Since early nineteenth century, a matrix A(G) has been associated to an organic
molecule to show its atomic adjacency/connectivity (Sylvester 1874). This is a
square table, of dimensions n  n, whose entries are


1 if i 6¼ j and ði; jÞ 2 EðGÞ
½AðGÞij ¼
0 if i ¼ j or ði; jÞ 2 = EðGÞ

A(G) characterizes a graph up to isomorphism. It allows the reconstruction of the

graph. A(G) is symmetric vs. its main diagonal, so that the transpose AT(G) leaves A
(G) unchanged: AT(G) ¼ A(G) (Trinajstić 1983). Figure 1.9 illustrates the adjacency
matrix. If instead 1, the bond order is used, the corresponding matrix is called the
connectivity matrix C(G), with information identical to a connection table. Raising at
a power e, of a square matrix, can be eluded by applying the algorithm of Diudea
et al. (1994). It evaluates a (topological) property of a vertex I by iterative summation
of the first neighbors contributions. The algorithm, called eWM, is defined as

M þ I ¼ 0 WM
eþ1  X 
WM ii ¼ ½Mij ½e WM jj
j
eþ1 
WM ij
¼ ½e WM ij ¼ ½Mij
6 1 Basic Chemical Graph Theory

2 3
G1.2.1

A 1
wi A2 2
wi A3 3
wi
1 0 0 0 1 1 1 1 1 0 3 0 1 1 3 5
2 0 0 1 1 2 1 2 1 1 5 1 2 3 4 10
3 0 1 0 1 2 1 1 2 1 5 1 3 2 4 10
4 1 1 1 0 3 0 1 1 3 5 3 4 4 2 13

Fig. 1.9 Adjacency matrices An; n ¼ 1–3, for the graph G1.2.1

where M is any square matrix; the diagonal elements [eWM]ii equal the row sum
RSi of Me and are walk degrees ewM , i (weighted by the property collected in M—
Diudea et al. 2002)
X
½e WM ii ¼ ½Me ij ¼ e wM, i
j

The half sum of the local invariants ewM,i in G is a global invariant, called the
walk number eWM.

1X e
e
W M ¼e W M ðGÞ ¼ i
W M, i
2

When M ¼ A, the quantity eWM (or simply eW ) represents the so called

molecular walk count.
The sum of diagonal elements in a square matrix is called the trace Tr(Me) ¼ ∑i[Me]ii
(Trinajstić 1983). The half sum of diagonal elements provides the global invariant e
SRWM (Self Returning Walk) eSRWM ¼ (1/2)∑i[Me]ii ¼ MOM(Me) which equals the
moment of order e of the matrix M, MOM(Me). When M ¼ A, the elements [Ae]ii count
both self-returning walks and circuits of length e. (Diudea et al. 2002) MOM(Ae) is
related to the spectral properties of molecular graphs (e.g., the energy of molecular
orbitals—Graovac and Babić 1990). Figure 1.10 shows the graphical evaluation of ewi
and eW numbers.

1.2.2 Distance Matrix

Distance matrix D(G) was introduced by Harary (1969). It is a square symmetric

table, of dimensions n  n, whose entries are defined as
1.2 Topological Matrices and Indices 7

e e
W
e
W
3 2 3 2
1 2 2 1 5 1 3 1 5

6 3 7 4
2 6 5 2 11 3 8 2 12

17 7 19 10
3 12 9 3 24 7 18 4 29
Fig. 1.10 Graphical evaluation of ewi and eW: e ¼ 1–3

1 2 3 4 5 6 7 8 RSi
1 0 1 2 3 4 5 2 3 20
7 2 1 0 1 2 3 4 1 2 14
3 2 1 0 1 2 3 2 1 12
2 4 6
1 3 5 4 3 2 1 0 1 2 3 2 14
5 4 3 2 1 0 1 4 3 18
8 6 5 4 3 2 1 0 5 4 24
G1.2.2 7 2 1 2 3 4 5 0 3 20
8 3 2 1 2 3 4 3 0 18

Fig. 1.11 Distance matrix D(G1.2.4)

 
min l pi, j , if i 6¼ j
½DðGÞij ¼
0 if i ¼ j

The non-diagonal entries of this matrix are just the topological distance between
i and j. Figure 1.11 illustrates the distance matrix.
The half sum of all entries in D(G) provides the well-known Wiener W topo-
logical index
XX
W ¼ W ðGÞ ¼ ð1=2Þ i j ½Dij

When one considers the genuine distances between atoms (i.e., the distances
measured through space), the 3D geometric matrix is obtained (Crippen 1977).

1.2.3 Detour Matrix

In cycle-containing graphs, when the shortest path (i.e., geodesic) is replaced by the
longest path between two vertices i and j, the Detour matrix Δ(G) can be
8 1 Basic Chemical Graph Theory

Fig. 1.12 Detour matrix 0 1 2 3 4 5 6 RSi

Δ(G1.2.3) 1 0 1 3 3 4 5 16
1 2 1 0 2 2 3 4 12
3 3 2 0 2 3 4 14
2
4 3 2 2 0 1 2 10
6 5 4 3 3 1 0 1 12
3 4 5
G1.2.3 6 5 4 4 2 1 0 16

constructed (Harary 1969; Diudea et al. 2002; Lukovits 1996; Amić and Trinajstić
1995)

 
maxl pi, j , if i 6¼ j
½ΔðGÞij ¼
0 if i ¼ j

Figure 1.12 shows a Detour matrix.

1.2.4 Combinatorial Matrices

Two path-calculated matrices have been proposed (Diudea 1996; Diudea et al.
1998): the distance-path Dp (Fig. 1.13) and the detour-path Δp (Fig. 1.14), whose
elements are combinatorially calculated from the classical matrices, distance D
(or distance-edge) and detour Δ (or detour-edge)


  npði; jÞ, ði; jÞ 2 DðGÞ , if i 6¼ j
Dp ¼ ð1:1Þ
ij 0 if i¼j


  npði; jÞ, ði; jÞ 2 ΔðGÞ , if i 6¼ j
Δp ij ¼ ð1:2Þ
0 if i¼j

 2
½Mij þ 1
npði; jÞ ¼ ¼ ð1=2Þ ½Mi, j þ ½Mi, j ; M ¼ D; Δ
2

In the above, np(i, j) is the number of internal paths (Klein et al. 1995) of length
1  | p|  |(i, j )| included in the path p(i, j).
Half-sum of entries in these matrices provide the Hyper-Wiener and Hyper-
Detour indices (Lukovits 1996; Diudea et al. 1998).

1.2.5 Wiener Matrices

Randić proposed the so-called Wiener matrix, W, and exploited it as a source of

structural invariants, useful in QSPR/QSAR (Randić et al. 1993, 1994).
1.2 Topological Matrices and Indices 9

1 2 3 4 5 6 7 8 RSi
1 0 1 3 6 10 15 3 6 44
7 2 1 0 1 3 6 10 1 3 25
3 3 1 0 1 3 6 3 1 18
2 4 6
1 3 5 4 6 3 1 0 1 3 6 3 23
5 10 6 3 1 0 1 10 6 37
8 6 15 10 6 3 1 0 15 10 60
G1.2.2 7 3 1 3 6 10 15 0 6 44
8 6 3 1 3 6 10 6 0 35

Fig. 1.13 Distance-path matrix Dp(G1.2.2)

1 2 3 4 5 6 7 8 9 10
10 1 0 15 21 6 10 15 21 28 28 1
2 15 0 15 10 6 10 15 21 21 21
1 3 21 15 0 15 21 15 21 28 28 28

6 2 4 6 10 15 0 15 10 15 21 21 10
8
5 10 6 21 15 0 15 21 28 28 15
7
9 6 15 10 15 10 15 0 15 21 21 21
5 3 7 21 15 21 15 21 15 0 1 1 28

4 8 28 21 28 21 28 21 1 0 3 36
9 28 21 28 21 28 21 1 3 0 36
G1.2.4
10 1 21 28 10 15 21 28 36 36 0

Fig. 1.14 Detour-path matrix Δp(G1.2.4)

In trees, the non-diagonal entries in such a matrix are defined as:

 
We=p i, j
¼ ni, e=p  nj, e=p

where ni and nj denote the number of vertices lying on the two sides of the edge/
path, e/p (having i and j as endpoints); when defined on edge, the half-sum of entries
give the Wiener index while defined on path, the hyper-Wiener index is calculated.
We is calculated from Wp by relation We ¼ Wp • A, where the symbol • indicates the
Hadamard (Horn and Johnson 1985) pair-wise matrix product:
[Ma • Mb]ij ¼ [Ma]ij [Mb]ij. In fact, We is an adjacency matrix weighted by the number
of external paths which include a given edge e. Fig. 1.15 illustrates this matrix.

1.2.6 Cluj Matrices

Cluj matrices CJ(G) have been proposed by Diudea (Diudea 1997a; Diudea et al.
1997a, b; Janežič et al. 2007); they are defined on Cluj fragments CJi , j , p which
10 1 Basic Chemical Graph Theory

1 2 3 4 5 6 7 8 RSi
1 0 7 5 3 2 1 1 1 20
7
2 7 0 15 9 6 3 7 3 50
2 4 6 3 5 15 0 15 10 5 5 7 62
1 3 5
4 3 9 15 0 12 6 3 3 51
8 5 2 6 10 12 0 7 2 2 41
G1.2.2 6 1 3 5 6 7 0 1 1 24
7 1 7 5 3 2 1 0 1 20
8 1 3 7 3 2 1 1 0 18

Fig. 1.15 Wiener matrix We/p(G1.2.2): the italicized entries represent We matrix

collect vertices v lying closer to i than to j, the endpoints of a path p(i, j). These
fragments represent the vertex proximities (see also Gutman 1994) of i vs. any
vertex j, joined by the path p, with the distances measured in the subgraph G \ p:
   
CJ i, j, p ¼ v v 2 V ðGÞ; DðG\ pÞ i; v < DðG\ pÞ j; v

The entries in CJ matrices are taken, by definition, as the maximum cardinality

among all such fragments: ½UCJi, j ¼ max CJ i, j, p ; it is because, in graphs
p
containing rings, more than one path can join the pair (i, j), thus resulting more
than one fragment related to i (with respect to j and path p). Cluj matrices are
defined by using either distance or detour concepts: when path p belongs to the set
of distances DI(G), the suffix DI is added to the name of matrix; when path
p belongs to the set of detours DE(G), the suffix is DE; when the matrix symbol
is not followed by a suffix, it is by default DI.
Cluj matrices are defined in any graph and, except for some symmetric graphs,
are unsymmetric: UCJDI, UCJDE. Symmetric Cluj matrix SCJ can be obtained by
the Hadamard multiplication of UCJ with its transpose: SCJ ¼ UCJ • UCJT. The
edge-based symmetric matrix is calculable from the path-based one:
SCJe ¼ SCJp • A. Fig. 1.16 illustrates the Cluj matrices in a ring-containing graph.
In trees, the path joining any two vertices is unique, then CJi , j , p represents the
set of external paths passing through i to j. Fig. 1.17 illustrates the Cluj matrix in a
tree graph. In calculating CJ matrix, the path p(i, j) is characterized by a single
endpoint, however, the information is twice: the row-sum is identical to that
provided by We matrix while the column-sum equals that in the distance D matrix.
Indices derived from topological matrices (see above) are calculated as half sum
of matrix entries
XX X
I ðGÞ ¼ ð1=2Þ i j ½Mij ¼ i>j
½Mij ¼ ð1=2Þ uMuT
1.2 Topological Matrices and Indices 11

UCJDI UCJDE
1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8
1 0 4 2 2 2 2 3 5 0 1 1 1 1 2 1 2
2 4 2 3 0 2 2 2 3 2 4 1 0 1 1 1 1 2 1
7
5 3 5 4 0 4 4 4 3 6 2 2 0 3 4 2 2 2
6 4 3 5 3 0 3 3 4 5 2 2 2 0 4 2 2 3
1 3 8
5 5 5 2 2 0 4 4 5 1 1 1 1 0 1 1 1
6 3 4 3 3 3 0 4 7 3 2 2 2 2 0 2 7
7 3 3 2 3 3 3 0 6 1 3 1 1 1 1 0 1
8 1 1 1 1 1 1 1 0 1 1 1 1 1 1 1 0

Fig. 1.16 Cluj matrices in a ring-containing graph G1.2.5

1 2 3 4 5 6 7 8 RS(We)
1 0 1 1 1 1 1 1 1 7
7 2 7 0 3 3 3 3 7 3 29
3 5 5 0 5 5 5 5 7 37
2 4 6
1 3 5 4 3 3 3 0 6 6 3 3 27
5 2 2 2 2 0 7 2 2 19
8 6 1 1 1 1 1 0 1 1 7
7 1 1 1 1 1 1 0 1 7
8 1 1 1 1 1 1 1 0 7
G1.2.2
CS(D) 20 14 12 14 18 24 20 18

Fig. 1.17 Cluj matrix UCJe/p (G1.2.4); the italicized entries represent UCJe matrix

where u and uT are the unit vector (of order n) and its transpose (Estrada and
Rodriguez 1997; Estrada et al. 1997). In trees, the following equalities hold:
SCJp ¼ Wp; SCJe ¼ We. The half-sum in the symmetric Cluj matrices provides
the Wiener W and hyper-Wiener WW indices, from SCJe and SCJp, respectively.
Relation with Cluj polynomial CJS(x) (calculated as the first derivative in x ¼ 1, see
Chap. 11) is as follows: ∑i∑j[UCJe]ij ¼ CJS(1)

1.2.7 Distance-Extended Matrices

Diudea (1997a, b) has performed the Hadamard product on the unsymmetric Cluj
matrix: [D • UCJ]i , j ¼ [D]i , j • [UCJ]i , j to provide a new matrix, that shows in trees,
the equalities: CS(D • UCJ) ¼ CS(Dp) and RS(D • UCJ) ¼ RS(Wp). This matrix
(illustrated in Fig. 1.18) is a direct proof of the theorem of Klein et al. (1995): in
trees, the sum of all internal paths (given by Dp) equals the sum of all external paths
(given by Wp) with respect to all pairs (i, j). Thus, the half-sum of entries in
D • UCJ equals the hyper-Wiener WW index (Diudea and Gutman 1998).
12 1 Basic Chemical Graph Theory

1 2 3 4 5 6 7 8 RS(Wp)
7 1 0 1 2 3 4 5 2 3 20
2 7 0 3 6 9 12 7 6 50
2 4 6 3 10 5 0 5 10 15 10 7 62
1 3 5
4 9 6 3 0 6 12 9 6 51
8 5 8 6 4 2 0 7 8 6 41
6 5 4 3 2 1 0 5 4 24
7 2 1 2 3 4 5 0 3 20
G1.2.2
8 3 2 1 2 3 4 3 0 18
CS(Dp) 44 25 18 23 37 60 44 35

Fig. 1.18 Distance-extended Cluj matrix D • UCJ(G1.2.2)

1.2.8 Walk Matrices

Diudea (1996, 1999) has proposed the walk matrix, W(M1M2M3), defined by the
principle of a single endpoint characterization of a path
h  i
½WM1 M2 M3 ij ¼ ½M2 ij W M1 i ½M3 ij ¼ RS ðM1 Þ½M2 ij ½M3 ij
i

where WM1i is the walk degree of the vertex i, of extent [M2]ij, weighted by the
property collected in M1 and M3 (i.e., the ith row-sum of the matrix M1, raised to
power [M2]ij and multiplied by the entries of M3); the diagonal entries are zero. It is
a square, (in general) non-symmetric matrix that mixes three square matrices, in
fact being a true matrix operator (see below).
Let first (M1 M2 M3) be (M1 1 1), where 1 is the matrix with “1” off-diagonal
elements. The entries of matrix W(M1 1 1) will be
 
WðM1 11Þ ij ¼ ½RSðM1 Þi ¼ W M1 , i

Next, consider the combination (M1 1 M3); the corresponding walk matrix can
be written as the Hadamard product

WðM1 1M3 Þ ¼ WðM1 11Þ • M3

Examples in Fig. 1.19 are for the case: M1 ¼ A and M3 ¼ D.

The sum of all entries in WðM1 1M3 Þ can be calculated by
X  
uWðM1 1M3 Þ uT ¼ i
RS WðM1 1M3 Þ i ¼ uðM1 M3 ÞuT

where u and uT are the unit vector (of order n) and its transpose. The above
equation represents a joint of Cramer and Hadamard matrix algebra, by means of
1.2 Topological Matrices and Indices 13

6
15
2 4 10 12
1 3 5 15 9 17

7 14
G1.2.6
A (G) D (G;RS(Di)

1 2 3 4 5 6 7 RSi 1 2 3 4 5 6 7 RSi
1 0 1 0 0 0 0 0 1 1 0 1 2 3 4 2 3 15
2 1 0 1 0 0 1 0 3 2 1 0 1 2 3 1 2 10
3 0 1 0 1 0 0 1 3 3 2 1 0 1 2 2 1 9
4 0 0 1 0 1 0 0 2 4 3 2 1 0 1 3 2 12
5 0 0 0 1 0 0 0 1 5 4 3 2 1 0 4 3 17
6 0 1 0 0 0 0 0 1 6 2 1 2 3 4 0 3 15
7 0 0 1 0 0 0 0 1 7 3 2 1 2 3 3 0 14
W(A11) (G) W(D11) (G)

1 2 3 4 5 6 7 RSi 1 2 3 4 5 6 7 RSi
1 0 1 1 1 1 1 1 6 1 0 15 15 15 15 15 15 90
2 3 0 3 3 3 3 3 18 2 10 0 10 10 10 10 10 60
3 3 3 0 3 3 3 3 18 3 9 9 0 9 9 9 9 54
4 2 2 2 0 2 2 2 12 4 12 12 12 0 12 12 12 72
5 1 1 1 1 0 1 1 6 5 17 17 17 17 0 17 17 102
6 1 1 1 1 1 0 1 6 6 15 15 15 15 15 0 15 90
7 1 1 1 1 1 1 0 6 7 14 14 14 14 14 14 0 84

(A1D) = W(A11) D W(D1A) = W(D11) A

1 2 3 4 5 6 7 diSDi 1 2 3 4 5 6 7 diSDi
1 0 1 2 3 4 2 3 15 1 0 15 0 0 0 0 0 15
2 3 0 3 6 9 3 6 30 2 10 0 10 0 0 10 0 30
3 6 3 0 3 6 6 3 27 3 0 9 0 9 0 0 9 27
4 6 4 2 0 2 6 4 24 4 0 0 12 0 12 0 0 24
5 4 3 2 1 0 4 3 17 5 0 0 0 17 0 0 0 17
6 2 1 2 3 4 0 3 15 6 0 15 0 0 0 0 0 15
7 3 2 1 2 3 3 0 14 7 0 0 14 0 0 0 0 14
ADj 24 14 12 18 28 24 22 142 ADi 10 39 36 26 12 10 9 142

Fig. 1.19 W(M1M2M3) algebra; diSDi are local contributions to the degree-distance index
(Dobrynin and Kochetova 1994)

W(M11M3). The row-sum vector in W(M11M3) can be achieved by the pair-wise

product of the row-sums in M1 and M3, respectively
  
RS WðM1 1M3 Þ i ¼ ½RSðM1 Þi • ½RSðM3 Þi
14 1 Basic Chemical Graph Theory

This vector represents a collection of pair-wise products of local (topological)

properties (encoded as row-sums in M1 and M3). When M1 ¼ M3 ¼ M, then the
sum of entries can be written as
 
uWðM n MÞ uT ¼ uððMÞn MÞuT ¼ u Mnþ1 uT

thus proving that W(M1M2M3) is a true matrix operator.

A particular case of the walk matrix, RW(AD1) (i.e., the reciprocal of W(AD1)) is
identical to the restricted random walk matrix proposed by Randić (1995), (see also
Diudea and Randić 1997).

1.2.9 Reciprocal Matrices

In Chemical Graph Theory, the distance matrix accounts for through bond interac-
tions of atoms in molecules. However, these interactions decrease as the distance
between atoms increases. This reason led to the introduction, by the QSAR Group
of Timisoara (Ciubotariu 1987; Ciubotariu et al. 2004) and next by the groups of
Balaban (Ivanciuc et al. 1993) and Trinajstić (Plavšić et al. 1993), of the reciprocal
distance matrix, RD. The entries in this matrix are defined by

½RDij ¼1=½Dij

RD matrix allows the calculation of a Wiener analogue number, called the

Harary index (Plavšić et al. 1993), in the honor of Frank Harary. Several matrices
having entries as reciprocal (topological) property: [RM]i , j ¼ 1/[M]i , j ; M ¼ We/p ,
Dp, and UCJ have been proposed by Diudea (1997c).

1.2.10 Layer and Shell Matrices

1.2.10.1 Layer Matrices

Layer matrices (Skorobogatov and Dobrynin 1988) have been proposed in connec-
tion with sequences of walks (Halberstam and Quintas 1982; Bonchev et al. 1989;
Dobrynin 1993); they are built up on the layer partitions in a graph.
Let G(v)k be the kth layer of vertices v lying at distance k, in the partition G(i)

GðvÞk ¼ fvjdi, v ¼ kg
 
GðiÞ ¼ GðvÞk ; k 2 0; 1; . . . ; ecci
1.2 Topological Matrices and Indices 15

6 i\k 0 1 2 3 4 RS
1 1 1 2 2 1 7
2 1 3 2 1 0 7
2 4 3 1 3 3 0 0 7
4 1 2 2 2 0 7
1 3 5
5 1 1 1 2 2 7
6 1 1 2 2 1 7
7 1 1 2 3 0 7
7 CS 7 12 14 12 4
G1.2.6
kCS 12 28 36 16 92a

Fig. 1.20 Layer matrix of counting LC(G1.2.6); a2  Wiener index

with ecci being the eccentricity of i. The entries in the layer matrix (of vertex
property) LM (Diudea 1994, 2010; Diudea et al. 1994; Diudea and Ursu 2003) are
defined as
X
½LMi, k ¼ pv
vjd i, v ¼k

Layer matrix is a collection of the above defined entries:

n   o
LMðGÞ ¼ ½LMi, k ; i 2 V ðGÞ; k 2 0; 1; ::; dðGÞ

with d(G) being the diameter of the graph (i.e., the largest distance in G). The
zero column is just the column of vertex properties, [LM]i , 0 ¼ pi. When pi ¼ 1 (i.e.,
the counting property), LM matrix is named LC (i.e., Layer matrix of Counting—
Fig. 1.20). Any atomic/vertex property can be considered as pi. Moreover, any
square matrix M can be taken as info matrix, i.e., the matrix supplying local/vertex
properties, as row-sum RS, column-sum CS or diagonal entries, as implemented in
TopoCluj software (Ursu and Diudea 2005).

1.2.10.2 Shell Matrices

Entries in the shell matrix ShM (of vertex pair property, Diudea and Ursu 2003) are
defined as
X
½ShMi, k ¼ ½Mi, v
vjdi, v ¼k

The shell matrix is a collection of the above defined entries:

n   o
ShMðGÞ ¼ ½ShMi, k ; i 2 VðGÞ; k 2 0; 1; ::; dðGÞ
16 1 Basic Chemical Graph Theory

A shell matrix, ShM(G), will partition the entries of a square matrix according to
the vertex-distance partitions in the graph. The zero column entries [ShM]i , 0 can be
the diagonal entries of the info matrix. Tables 1.1, 1.2 and 1.3 exemplify the Shell
matrix of Cluj matrices for G1.2.6
Let now consider the behavior of the edge-calculated matrices: adjacency A, We
and UCJe matrices. These are sparse matrices, with non-zero entries only at k ¼ 1,
the other columns in the shell matrix being null (see Table 1.4). In case of UCJe
matrix, the column-sum equals the first derivative (in x ¼ 1) of Cluj polynomial
CJ-Sum(x) (see Sect. 11.6.2).

Table 1.1 Shell matrix of UCJ

ShUCJ(G1.2.6) UCJ(G1.2.6)
i\k 0 1 2 3 4 RS 1 2 3 4 5 6 7 RS
1 0 1 2 2 1 6 0 1 1 1 1 1 1 6
2 0 15 6 3 0 24 6 0 3 3 3 6 3 24
3 0 15 13 0 0 28 4 4 0 5 5 4 6 28
4 0 8 4 4 0 16 2 2 2 0 6 2 2 16
5 0 1 1 2 2 6 1 1 1 1 0 1 1 6
6 0 1 2 2 1 6 1 1 1 1 1 0 1 6
7 0 1 2 3 0 6 1 1 1 1 1 1 0 6
CS 0 42 30 16 4 92a CS 15 10 9 12 17 15 14 92
CS k 42 60 48 16 166b
a
2  Wiener index W
b
2  hyper-Wiener index WW (Wiener 1947; Randić 1993)
Shell matrix of UCJ (the bold values indicate entries in UCJe matrix)

Table 1.2 Shell matrix of UCJT

ShUCJT( G1.2.6) UCJT (G1.2.6)
i\k 0 1 2 3 4 RS 1 2 3 4 5 6 7 RS
1 0 6 5 3 1 15 0 6 4 2 1 1 1 15
2 0 6 3 1 0 10 1 0 4 2 1 1 1 10
3 0 6 3 0 0 9 1 3 0 2 1 1 1 9
4 0 6 4 2 0 12 1 3 5 0 1 1 1 12
5 0 6 5 4 2 17 1 3 5 6 0 1 1 17
6 0 6 5 3 1 15 1 6 4 2 1 0 1 15
7 0 6 5 3 0 14 1 3 6 2 1 1 0 14
CS 0 42 30 16 4 92a CS 6 24 28 16 6 6 6 92a
CS k 42 60 48 16 166b
2  Wiener index W
a

2  hyper-Wiener index WW
b
1.2 Topological Matrices and Indices 17

Table 1.3 Shell matrix of the symmetric Cluj matrix SCJ

ShSCJ(G1.2.6) SCJ(G1.2.6)
i\k 0 1 2 3 4 RS 1 2 3 4 5 6 7 RS
1 0 6 5 3 1 15 0 6 4 2 1 1 1 15
2 0 24 9 3 0 36 6 0 12 6 3 6 3 36
3 0 28 13 0 0 41 4 12 0 10 5 4 6 41
4 0 16 8 4 0 28 2 6 10 0 6 2 2 28
5 0 6 5 4 2 17 1 3 5 6 0 1 1 17
6 0 6 5 3 1 15 1 6 4 2 1 0 1 15
7 0 6 5 3 0 14 1 3 6 2 1 1 0 14
CS 0 92 50 20 4 166a CS 15 36 41 28 17 15 14 166
CS k 92 100 60 16 268b
2  hyper Wiener index
a

2  Tratch index (Tratch et al. 1990)

b

Table 1.4 Shell matrices of edge-defined info matrices

ShA(G1.2.6) ShWe(G1.2.6) ShUCJe(G1.2.6)
i\k 0 1 2 3 4 RS 0 1 2 3 4 RS 0 1 2 3 4 RS
1 0 1 0 0 0 1 0 6 0 0 0 6 0 1 0 0 0 1
2 0 3 0 0 0 3 0 24 0 0 0 24 0 15 0 0 0 15
3 0 3 0 0 0 3 0 28 0 0 0 28 0 15 0 0 0 15
4 0 2 0 0 0 2 0 16 0 0 0 16 0 8 0 0 0 8
5 0 1 0 0 0 1 0 6 0 0 0 6 0 1 0 0 0 1
6 0 1 0 0 0 1 0 6 0 0 0 6 0 1 0 0 0 1
7 0 1 0 0 0 1 0 6 0 0 0 6 0 1 0 0 0 1
CS 0 12 12a 0 92 92b 0 42 42c
a
2e
b
2W
c
CJS(x ¼ 1) ¼ v  e

1.2.10.3 Centrality Index

On the above layer/shell matrices, a centrality index (Diudea and Ursu 2003; Ursu
and Diudea 2005) is calculated as
" #1
ecci 
X 1=ðecci Þ2
CðLM\ ShMÞi ¼ ½LM\ ShM2k
ik
k¼1
X
CðLM\ ShMÞ ¼ CðLM\ ShMÞi
i

This index allows the finding of the graph center (e.g. the vertex having the
largest Ci value) and provides an ordering of graph vertices according to their
centrality.
18 1 Basic Chemical Graph Theory

1.3 Topological Symmetry

Molecular geometry and their crystals is completely described by the 230 symmetry
groups of the space (Hargittai and Hargittai 2010; Petitjean 2007). Symmetry of
molecules is reflected in several properties, such as dipole moments, IR vibrations,
13
C-NMR signals etc., that are dependent on the spatial structure of molecules.
Molecular topology reveals a different type of symmetry: the topological symmetry
(i.e., constitutional symmetry). It is defined in terms of connectivity, as a constitu-
tive principle of molecules and expresses equivalence relationships among ele-
ments of graph: vertices, bonds, faces or larger subgraphs. It makes use of Group
Theory formalism while the geometrical aspects are disregarded.
Let G1 ¼ (V1, E1) and G2 ¼ (V2, E2) be two graphs and a function f, mapping the
vertices of V1 onto the vertices belonging to the set V2, f: V1 ! V2. That is, the
function f makes a one-to-one correspondence between the vertices of the two sets.
The two graphs are called isomorphic, G1 G2, if there exists a mapping f that
preserves adjacency (i.e., if (i, j) 2 E1, then ( f(i), f( j)) 2 E2).
Using notions of the Group action (Hungerford 1974; Mirman 1999), in which
every element of the group acts like as a one-to-one mapping, the group G is said to
act on a set X if there is a function ϕ such that ϕ:G  ! X and for any element
x 2 X, there exists the relation ϕ(g,ϕ(h, x)) ¼ ϕ(gh, x), for all g, h 2 G, with ϕ(e,
x) ¼ x, e being the identity element of G. The mapping ϕ is called a group action
while the set {ϕ(gx)|g 2 G} is called the orbit of x. For a permutation σ on n objects,
the permutation matrix is an n  n matrix Pσ , with elements xij ¼ 1 if i ¼ σ( j) and
0 otherwise. For any permutation σ and τ on n objects, Pσ Pτ ¼ Pστ, while the set of
all permutation matrices is a group isomorphic to the symmetry group Sn on
n symbols. A permutation σ of the vertices of a graph H(V, E) (V being the set of
vertices and E the set of edges in H ) belongs to an automorphism group G if one
satisfies the relation Pσ TAPσ ¼ A, where A is the adjacency matrix of the graph H.
Given Aut(H ) ¼ {σ 1, . . . , σ m}, the matrix SG ¼ [sij], where sij ¼ σi( j), is called a
solution matrix for H and its calculation will provide the automorphism group of H.
Given a graph H(V, E) and the automorphism group Aut(H ), two vertices, i, j 2 V
are called equivalent if {ϕ(ij)|i, j 2 Aut(H )}, in other words, they belong to the same
orbit of automorphisms. Suppose v1, v2,...,vm are m disjoint automorphic partitions
of the set of vertices V(H ), then: V ¼ Vv1 [ Vv2 [ . . . [ Vvm and V vi \ V vj ¼ ∅. All
the automorphisms of a graph form a group (Gutman and Polanski 1986; Razinger
et al. 1993; Balasubramanian 1994, 1995a, b, c).
Let now consider a vertex invariant (i.e., a topological index), In ¼ In1, In2, . . .,
Inm, which assigns a value Ini to the vertex i. Two vertices i and j of a molecular
graph (with vertices meaning the atoms and edges the bonds in the molecule)
belong to the same invariant class if Ini ¼ Inj. The partitioning in classes of
vertices/atoms leads to m classes, with v1, v2, . . . vm atoms in each class; such a
partitioning may differ from the orbits of automorphism i.e. classes of equivalence,
since no vertex invariant is known so far to always discriminate two non-equivalent
vertices in any graph. The classes of vertices are eventually ordered according to
References 19

some rules. Note that topological symmetry equals the maximum geometrical
symmetry a molecular graph can have. Also note that topological symmetry, arising
from graph automorphisms, can be defined the same way as the Euclidean symme-
try (Petitjean 2007).
In Chemistry, such a study could answer to the question if two molecular graphs
represent or not one and the same chemical compound.
Within this book, topological symmetry/equivalence classes will be evaluated
by means of Centrality index (see Sect. 1.2.10.3) and checked by adjacency matrix
permutations.

References

Amić D, Trinajstić N (1995) On the detour matrix. Croat Chem Acta 68:53–62
Balasubramanian K (1994) Computer generation of automorphism graphs of weighted graphs. J
Chem Inf Comput Sci 34:1146–1150
Balasubramanian K (1995a) Computer generation of nuclear equivalence classes based on the
three-dimensional molecular structure. J Chem Inf Comput Sci 35:243–250
Balasubramanian K (1995b) Computational strategies for the generation of equivalence classes of
Hadamard matrices. J Chem Inf Comput Sci 35:581–589
Balasubramanian K (1995c) Computer perception of molecular symmetry. J Chem Inf Comput Sci
35:761–770
Bonchev D, Mekenyan O, Balaban AT (1989) Iterative procedure for the generalized graph center
in polycyclic graphs. J Chem Inf Comput Sci 29:91–97
Ciubotariu D (1987) Structură-reactivitate ı̂n clasa derivaţilor acidului carbonic. PhD Thesis,
Timisoara, Romania
Ciubotariu D, Medeleanu M, Vlaia V, Olariu T, Ciubotariu C, Dragos D, Seiman C (2004)
Molecular van der Waals space and topological indices from the distance matrix. Molecules
9:1053–1078
Crippen GM (1977) A novel approach to calculation of conformation: distance geometry. J
Comput Phys 24:96–107
Diudea MV (1994) Layer matrices in molecular graphs. J Chem Inf Comput Sci 34:1064–1071
Diudea MV (1996) Walk numbers eWM: Wiener-type numbers of higher rank. J Chem Inf Comput
Sci 36:535–540
Diudea MV (1997a) Cluj matrix invariants. J Chem Inf Comput Sci 37:300–305
Diudea MV (1997b) Cluj matrix, CJu: source of various graph descriptors. MATCH Commun
Math Comput Chem 35:169–183
Diudea MV (1997c) Indices of reciprocal property or Harary indices. J Chem Inf Comput Sci
37:292–299
Diudea MV (1999) Valencies of property. Croat Chem Acta 72:835–851
Diudea MV (2010) Nanomolecules and nanostructures—polynomials and indices, MCM, No. 10.
University Kragujevac, Serbia
Diudea MV, Gutman I (1998) Wiener-type topological indices. Croat Chem Acta 71:21–51
Diudea MV, Randić M (1997) Matrix operator, W(M1,M2,M3) and Schultz-type numbers. J Chem
Inf Comput Sci 37:1095–1100
Diudea MV, Ursu O (2003) Layer matrices and distance property descriptors. Indian J Chem
42A:1283–1294
Diudea MV, Topan MI, Graovac A (1994) Layer matrices of walk degrees. J Chem Inf Comput Sci
34:1071–1078
20 1 Basic Chemical Graph Theory

Diudea MV, Parv B, Gutman I (1997a) Detour-Cluj matrix and derived invariants. J Chem Inf
Comput Sci 37:1101–1108
Diudea M, Parv B, Topan MI (1997b) Derived Szeged and Cluj indices. J Serb Chem Soc
62:267–276
Diudea MV, Katona G, Lukovits I, Trinajstić N (1998) Detour and Cluj-detour indices. Croat
Chem Acta 71:459–471
Diudea MV, Gutman I, Jäntschi L (2002) Molecular topology. NOVA, New York
Diudea MV, Florescu MS, Khadikar PV (2006) Molecular topology and its applications. EFICON,
Bucharest
Dobrynin AA (1993) Degeneracy of some matrix invariant and derived topological indices. J Math
Chem 14:175–184
Dobrynin AA, Kochetova AA (1994) Degree distance of a graph: a degree analogue of the Wiener
index. J Chem Inf Comput Sci 34:1082–1086
Estrada E, Rodriguez L (1997) Matrix algebric manipulation of molecular graphs, Harary- and
MTI-like molecular descriptors. MATCH Commun Math Comput Chem 35:157–167
Estrada E, Rodriguez L, Gutierrez A (1997) Matrix algebric manipulation of molecular graphs,
distance and vertex-adjacency matrices. MATCH Commun Math Chem 35:145–156
Euler L (1752–1753) Elementa doctrinae solidorum. Novi Comm Acad Scient Imp Petrop
4:109–160
Graovac A, Babić D (1990) The evaluation of quantum chemical indices by the method of
moments. Int J Quantum Chem Quantum Chem Symp 24:251–262
Gutman I (1994) A formula for the Wiener number of trees and its extension to graphs containing
cycles. Graph Theory Notes NY 27:9–15
Gutman I, Polansky OE (1986) Mathematical concepts in organic chemistry. Springer, Berlin, pp
108–116
Halberstam FY, Quintas LV (1982) Distance and path degree sequences for cubic graphs. Pace
University, New York. A note on table of distance and path degree sequences for cubic graphs,
Pace University, New York
Harary F (1969) Graph theory. Addison-Wesley, Reading
Hargittai M, Hargittai I (2010) Symmetry through the eyes of a chemist. Springer, Dordrecht
Horn RA, Johnson CR (1985) Matrix analysis. Cambridge University Press, Cambridge
Hungerford TW (1974) Algebra. Graduate texts in mathematics, vol 73. Springer, New York
(Reprint of the original 1980)
Ionescu T (1973) Graphs-applications (in Romanian), Ed. Ped., Bucharest
Ivanciuc O, Balaban TS, Balaban AT (1993) Reciprocal distance matrix, related local vertex
invariants and topological indices. J Math Chem 12:309–318
Janežič D, Miličević A, Nikolić S, Trinajstić N (2007) Graph theoretical matrices in chemistry.
Mathematical chemistry monographs. University Kragujevac, Kragujevac
Klein DJ, Lukovits I, Gutman I (1995) On the definition of the hyper-Wiener index for cycle-
containing structures. J Chem Inf Comput Sci 35:50–52
Kuratowski K (1930) Sur le Problème des Courbes Gauches en Topologie. Fund Math 15:271–283
Lukovits I (1996) The detour index. Croat Chem Acta 69:873–882
Mirman R (1999) Point groups, space groups, crystals, molecules. World Scientific, River Edge
Petitjean M (2007) A definition of symmetry. Symmetry Cult Sci 18:99–119
Plavšić D, Nikolić S, Trinajstić N, Mihalić Z (1993) On the Harary index for the characterization
of chemical graphs. J Math Chem 12:235–250
Randić M (1991) Generalized molecular descriptors. J Math Chem 7:155–168
Randić M (1993) Novel molecular description for structure-property studies. Chem Phys Lett
211:478–483
Randić M (1995) Restricted random walks on graphs. Theor Chim Acta 92:97–106
Randić M, Guo X, Oxley T, Krishnapriyan H (1993) Wiener matrix: source of novel graph
invariants. J Chem Inf Comput Sci 33:700–716
References 21

Randić M, Guo X, Oxley T, Krishnapriyan H, Naylor L (1994) Wiener matrix invariants. J Chem
Inf Comput Sci 34:361–367
Razinger M, Balasubramanian K, Munk ME (1993) Graph automorphism perception algorithms in
computer-enhanced structure elucidation. J Chem Inf Comput Sci 33:197–201
Skorobogatov AV, Dobrynin AA (1988) Metric analysis of graphs. MATCH Commun Math
Comput Chem 23:105–151
Sylvester JJ (1874) On an application of the new atomic theory to the graphical representation of
the invariants and covariants of binary quantics—with three appendices. Am J Math 1:64–90
Tratch SS, Stankevich MI, Zefirov NS (1990) Combinatorial models and algorithms in chemistry.
The expanded Wiener number—a novel topological index. J Comput Chem 11:899–908
Trinajstić N (1983) Chemical graph theory. CRC Press, Boca Raton
Ursu O, Diudea MV (2005) TOPOCLUJ software program. Babes-Bolyai University, Cluj
Wiener H (1947) Structural determination of paraffin boiling points. J Am Chem Soc 69:17–20