<

<

Method for Construction of

Characteristic Polynomials via
Graph Linearization
KAKALI DATTA, ASOK K. MUKHERJEE
Department of Chemistry, University of Burdwan, Burdwan-713104, India
Received 12 May 1997; accepted 22 May 1997

ABSTRACT: A new method for construction of characteristic polynomials CP. of

complicated graphs having arbitrary edge and vertex weights has been developed. The
method first converts the graph into isospectral linear chains with weighted vertices and
edges and then builds up the CP coefficients recursively. Two types of graphs have been
used for illustration, viz., i. graphs that can be linearized by symmetry factorization and
ii. graphs without symmetry which are to be linearized by an algorithm involving walks
of unit length. Both types have been illustrated, of which type i. includes the Schlegel of
fullerene fragment C 20 and another large graph with many fused rings. Q 1997 John
Wiley & Sons, Inc. Int J Quant Chem 65: 199]204, 1997

Introduction

here exist quite a large number of methods

for construction of characteristic polynomials
CP. of graphs. They may be classified into two
major types, viz. i. methods w 1]7x that depend on
counting of elementary subgraphs such as K 2 s
and rings and ii. methods w 8]10x based on CaleyHamilton theorem and Newtons identities w 11x
which require calculation of traces of the various
powers of the adjacency matrix A. of the graph.
While methods of type i. bring out the combinatorial dependence of the CP coefficients on the graph

Correspondence to: A. K. Mukherjee.

International Journal of Quantum Chemistry, Vol. 65, 199]204 (1997)
Q 1997 John Wiley & Sons, Inc.

structure, they are computationally difficult, particularly for large graphs with many fused rings.
On the other hand, methods of type ii. take into
account such dependence only indirectly traces of
An may be correlated to self-returning walks of
length n w 12x. , but are computationally facile.
Randic
w 10x has shown how a method of the second type developed by Barakat w 9x can be executed
combinatorially through the use of selected Young
diagrams after computing the traces of various
powers of A. The object of the present study is to
develop a computationally facile recursive method
for evaluation of CP coefficients of undirected
graphs having arbitrary vertex and edge weights
after linearizing the graph through symmetry factorization where possible or through a graph linearization algorithm recently developed w 13x which
uses walks of unit length.

CCC 0020-7608 / 97 / 030199-06

DATTA AND MUKHERJEE

Method
The algorithm for linearizing a graph using
walks of unit length was described in Ref. w 13x .
Sometimes this is not required and the graph may
be converted into an isospectral linear chain with
proper edge and vertex weights by symmetry factorization. The present algorithm for building up
the CP coefficients of a weighted linear chain is
based on the recurrence relation
iy1.
2
iy2.
Cr i. s Cr iy1. q h i Cry1
y k iy1,
i C ry2

C0 i. s 1,

y1.

Cr i. x iyr ,

4. Add columnwise to get the coefficients Cr i..

Insert the sign of y1. ry1 before each Cr i..

Illustrations

1.

where Cr i. means the r th coefficient of the CP,

P L i ; x ., of a linear chain with i vertices:
P Li ; x . s

2. Multiply each coefficient by h i and write

them below the above row displacing one
place to the right.
3. Multiply the coefficients of the i y 2.th row
2
by yk iy1,
i and write them displaced two
places to the right.

2.

rs0

where h i is the weight of the ith vertex and k iy1, i

is the weight of the edge connecting the vertices
labeled i y 1 and i. Equation 1. can be derived
from the Heilbronner recurrence relation w 14x applied to a weighted linear chain from one end:
2

.
P L i ; x . s x y h i . P L iy1 ; x . y k iy1,
i P L iy2 ; x .
3.

An easy procedure for executing the recursive

work required by Eq. 1. follows:
1. Write down the coefficients without sign. of
the i y 1.th chain in a row.

CASE I: GRAPHS LINEARIZED BY

SYMMETRY FACTORIZATION
a. The graph G 1 Fig. 1. can be converted
into mirror-plane fragments Gq
and Gy
by
1
1
McClellands method w 15]17x as shown in Figure
1. We have chosen this graph because it contains
as a subgraph the carbon atom skeleton of
biphenylene, which was used by Hosoya w 4x to
illustrate his method of construction of CP coefficients. The building up of the CP coefficients of
Gq
1 by the present method is shown in Table I.
These yield, according to Eq. 2.,
8
7
6
5
4
.
P Gq
1 ; x s x y 5 x q 3 x q 17x y 19 x

y 17x 3 q 21 x 2 q 5 x y 5.

4.

The CP of Gy
1 can be similarly constructed. It is
y
noteworthy here that the portions of Gq
1 and G 1
up to the sixth vertex are the mirror plane frag-

FIGURE 1. Linearization of a graph with many fused rings by using plane of symmetry.

200

VOL. 65, NO. 3

CONSTRUCTION OF CHARACTERISTIC POLYNOMIALS

TABLE I

Building up of the coefficients of G1+.

L 0:

h1L 0 :
L 1:

1
1

h 2 L 1:

yk1,2 2 L 0 :
L 2:

y1
1

h3 L 2:

y1

y1

y1

y1

y1

y2

y1

y2

y1

y1

y4

y1

y1

y2

y1

y6

y1

y6

y1

y3

y10

y6

y1

y3

y2

y13

y5

12

y2

y13

y5

12

y2

y1

y4

y1

10

y6

y3

y17

y19

17

21

y5

y5

2
yk 2,
3 L 1:

L 3:

h4 L 3 :
2
yk 3,
4 L 2:

L 4:

h5 L 4 :
2
yk 4,
5 L 3:

L 5:

h6 L 5:
2
yk 5,
6 L 4:

L 6:

h7 L 6:
2
yk 6,
7 L 5:

L 7:

h8 L 7 :
2
yk 7,
8 L 6:

L 8:
(y1) r:

5
x
(y)

3
x
(y)

x
(y)

INTERNATIONAL JOURNAL OF QUANTUM CHEMISTRY

x
(y)

201

DATTA AND MUKHERJEE

ments of G 2 and so

we obtain the CP of C 20 in the factorized form,

. y .
P G 2 ; x . s P Lq
6 ; x P L6 ; x

P G 3 ; x . s w P G4q ; x .x hs2 w P G4q ; x .x hs0.618

s x 6 y 4 x 5 q x 4 q 10 x 3

= w P G4q ; x .x hsy1.618 x q 1 .

y6 x y 6 x q 3 .
2

s x 3 y 3 x 2 y x q 5.

= x 6 q 4 x 5 y x 4 y 10 x 3

= x 3 y 1.618 x 2 y 2.382 x q 2.236.

y6 x 2 q 6 x q 3 .
s x 12 y 14 x 10 q 69 x 8 y 154 x 6
q 162 x 4 y 72 x 2 q 9,

= x 3 q 0.618 x 2 y 4.618 x y 2.236.

5.

which is the same as that obtained by Hosoya w 4x

but now obtained far more easily.
b. The graph G 3 Fig. 2. is the Schlegel of
fullerene fragment, C 20 , drawn by the circumscribing principle given by Dias w 18x . Following the
method of Davidson w 19x and Shen w 20x and utilizing a fivefold symmetry axis it can be shown w 21x
that the adjacency matrix of this Schlegel can be
factorized into five blocks each of which can be
represented by the graph G4 Fig. 2., with h s w 0
q w 5 s 2 for the first block, w q w 4 s 2 cos2pr5.
for the second and fifth blocks, and w 2 q w 3 s
2 cos4pr5. for the third and fourth blocks w w s
exp2p ir5.x . Mirror plane fragmentation of G4
gives the three-chain G4q and an isolated vertex
G4y. with weight y1. Since such fragmentation is
possible for all the five blocks, it immediately
follows that y1 is a fivefold degenerate eigenvalue of G 3 . The present procedure gives as shown
in Table II..
P G4q ; x . s x 3 y h q1 . x 2 q h y3 . x q 2 h q 1 . ,
P G4y ; x . s x q 1 . .

= x q 1.

s x 3 y 3 x 2 y x q 5.
= x 6 y x 5 y 8 x 4 q 6 x 3
2

5
q16 x 2 y 5 x y 5 . x q 1 . .

7.

TABLE II
Building up of the coefficients G4+.
L 0:

h1L 0 :
L 1:

1
1

h 2 L 1:

yk1,2 2 L 0 :
L 2:

y2

y2 h

y1

y1

2
yk 2,
3 L 1:

( h + 1)

1
(y1) r:

0
y2

h3 L 2:

6.

Using appropriate values of h as indicated earlier,

x
(y)

( h y 3)

y(1 + 2 h )
x
(y)

FIGURE 2. Linearization of the C 20 Schlegel using fivefold symmetry and plane of symmetry.

202

VOL. 65, NO. 3

CONSTRUCTION OF CHARACTERISTIC POLYNOMIALS

CASE II: GRAPHS THAT CANNOT BE
SYMMETRY FACTORIZED
Such a graph is G5 Fig. 3., which was used in
Ref. w 13x for illustration of the graph linearization
algorithm using walks of unit length. The isospectral linear chain that can be obtained from G5

through this algorithm is G5 L. Fig. 3.. Using the

present procedure the CP coefficients are built up
in Table III. It is found that
P G5 ; x . s P G5 L . ; x .
s x 5 y 2 x 4 y 5 x 3 q 5 x 2 q 5 x y 2. 8 .

FIGURE 3. Linearization of a graph without symmetry using walks of unit length.

TABLE III
Building up the CP coefficients of the nonsymmetric graph G.
1

h1L 0 :
1

h 2 L 1:

0.6667

yk1,2 2 L 0 :
L 2:

y3
1

h3 L 2:

1.6667

y2.3333

0.6190

1.0317

y1.4443

y1.5555

y1.5555

2.2857

y2.8571

y2.9998

y0.0683

y0.1561

+0.1951

+0.2049

y0.7041

y1.1735

+1.6428

2.2174

y3.7173

y3.9782

1.8477

y0.2174

y0.4821

+0.8081

+0.8649

y0.4017

y0.8005

y1.8297

+2.2871

+2.4013

y5

y5

2
yk 2,
3 L 1:

L 3:

h4 L 3 :
2
yk 3,
4 L 2:

L 4:

h5 L 4 :
2
yk 4,
5 L 3:

(y1) r:

2
6

L 5:

0.6667

(y)

L 1:

(y)

INTERNATIONAL JOURNAL OF QUANTUM CHEMISTRY

2
6

L 0:

(y)

203

DATTA AND MUKHERJEE

ACKNOWLEDGMENT
One of us K.D.. thanks the University of Burdwan for financial support. The authors also thank
the UGC, New Delhi, for financial assistance extended through DSA.

References
1.
2.
3.
4.
5.
6.
7.

H. Sachs, Publ. Math. Debrecen. 11, 119 1964..

J. Aihara, J. Am. Chem. Soc. 98, 6840 1976..
N. Trinajstic,
Croat. Chim. Acta 49, 593 1977..
H. Hosoya, Theor. Chim. Acta 25, 215 1972..
M. Randic,
J. Comput. Chem. 3, 421 1982..
K. Balasubramanian, Int. J. Quant. Chem. 21, 581 1982..
K. Balasubramanian and M. Randic,
Theor. Chim. Acta 61,
307 1982..

204

8. K. Balasubramanian, Theor. Chim. Acta 65, 49 1984..

9. R. Barakat, Theor. Chim. Acta 69, 35 1986..
10. M. Randic,
J. Math. Chem. 1, 145 1987..
11. W. Burnside and A. Panton, Theory of Equations, Vol. 1,
Dover, New York, 1956..
12. M. Randic,
SIAM J. Alg. Disc. Math. 6, 145 1985..
13. K. Datta and A. K. Mukherjee, J. Chem. Soc. Faraday Trans.,
91, 3095 1995..
14. E. Heilbronner, Helv. Chim. Acta 36, 170 1953..
15. B. J. McClelland, J. Chem. Soc. Faraday Trans. 2, 70, 1453
1974..
16. B. J. McClelland, Mol. Phys. 45, 189 1982..
17. B. J. McClelland, J. Chem. Soc. Faraday Trans. 2, 78, 911
1982..
18. J. R. Dias, J. Chem. Inf. Comput. Sci. 34, 248 1994..
19. R. A. Davidson, Theor. Chim. Acta 58, 193 1981..
20. M. Shen, Int. J. Quant. Chem. 38, 551 1990..
21. J. Sarkar and A. K. Mukherjee, Faraday Trans., to appear.

VOL. 65, NO. 3