MATCH MATCH Commun. Math. Comput. Chem.

77 (2017) 95-104
Communications in Mathematical
and in Computer Chemistry ISSN 0340 - 6253

Resolvent Energy of Unicyclic, Bicyclic

and Tricyclic Graphs

Luiz Emilio Allem1 , Juliane Capaverde1 , Vilmar Trevisan1 ,

Ivan Gutman2,3 , Emir Zogic3 , Edin Glogic3

1
Instituto de Matematica, UFRGS, Porto Alegre, RS, 91509900, Brazil
emilio.allem@ufrgs.br, juliane.capaverde@ufrgs.br,
trevisan@mat.ufrgs.br
2
Faculty of Science, University of Kragujevac,
Kragujevac, Serbia
gutman@kg.ac.rs
3
State University of Novi Pazar, Novi Pazar, Serbia
ezogic@np.ac.rs , edinglogic@np.ac.rs

(Received December 26, 2015)

Abstract
The resolvent energy of a graph G of order n is defined as ER = ni=1 (n i )1 , where
P

1 , 2 , . . . , n are the eigenvalues of G. In a recent work [Gutman et al., MATCH Commun.

Math. Comput. Chem. 75 (2016) 279290] the structure of the graphs extremal w.r.t.
ER were conjectured, based on an extensive computeraided search. We now confirm the
validity of some of these conjectures.

1 Introduction
Let G be a graph on n vertices, and let 1 2 n be its eigenvalues, that
is, the eigenvalues of the adjacency matrix of G. The resolvent energy of G is defined
in [3, 4] as
n
X 1
ER(G) = . (1)
i=1
n i
 -96-

It was shown in [3] that

1 X Mk (G)
ER(G) = (2)
n k=0 nk
Pn
where Mk (G) = i=1 ki is the k-th spectral moment of G.
In what follows, we present results found in the literature that confirm some
of the conjectures made in [3, 4] on the resolvent energy of unicyclic, bicyclic and
tricyclic graphs. These results were originally stated in [2,5,6] in terms of the Estrada
index, another spectrumbased graph invariant related to the spectral moments by
the formula

X Mk (G)
EE(G) = .
k=0
k!
Most of the proofs in [2, 5, 6] are based on the spectral moments and work for the
resolvent energy without any change. The proofs that involve direct calculations with
Estrada indices can be easily modified to give the equivalent results concerning the
resolvent energy, as we show below. Thus, the following are determined:

the unicyclic graph with maximum resolvent energy (Theorem 1);

the unicyclic graphs with minimum resolvent energy (Theorems 2 and 3);

the bicyclic graph with maximum resolvent energy (Theorem 4);

the tricyclic graph with maximum resolvent energy (Theorem 5).

2 Unicyclic graphs with maximum and minimum

resolvent energy
Let Xn denote the unicyclic graph obtained from the cycle C3 by attaching n 3
pendent vertices to one of its vertices, and Xn the unicyclic graph obtained from C4
by attaching n 4 pendent vertices to one of its vertices, as in Figure 1.
 -97-

Figure 1: Unicyclic graphs with maximum resolvent energy.

Lemma 1. Let G be a unicyclic graph on n 4 vertices, G

6= Xn , Xn .

1. If G is bipartite, then Mk (G) Mk (Xn ), for all k 0, and Mk0 (G) < Mk0 (Xn )
for some k0 .

2. If G is not bipartite (that is, G contains an odd cycle), then Mk (G) Mk (Xn )
for all k 0, and Mk0 (G) < Mk0 (Xn ) for some k0 .

Proof. Part (i) follows from Lemmas 3.2, 3.5 and 3.8 in [2], and (ii) follows from
Lemmas 3.2, 3.5 and 3.7 in [2].

We denote the characteristic polynomial of a graph G by (G, ). For a proper

subset V1 of V (G), G V1 denotes the graph obtained from G by deleting the vertices
in V1 (and the edges incident on them). Let G v = G {v}, for v V (G). We
make use of the following lemma.

Lemma 2. [1] Let v V (G), and let C(v) be the set of cycles containing v. Then
X X
(G, ) = (G v, ) (G v w, ) 2 (G V (Z), )
vwE(G) ZC(v)

where (G v w, ) 1 if G is a single edge, and (G V Z, ) 1 if G is a cycle.

Theorem 1. Let G be a unicyclic graph on n 4 vertices. Then ER(G) ER(Xn ),

with equality if and only if G = Xn . Moreover, if G is bipartite, then ER(G)
ER(Xn ), with equality if and only if G
= Xn .
 -98-

Proof. Let G be a unicyclic graph. Lemma 1 and equation (2) imply that ER(G)
ER(Xn ) if G contains an odd cycle, and ER(G) ER(Xn ) if G contains an even
cycle (i.e., G is bipartite). Furthermore, equality occurs if and only if G
= Xn , in the
case of an odd cycle, or G
= Xn , in the bipartite case.
Now, an n-vertex unicyclic graph with maximum resolvent energy is either Xn or
Xn . We show that ER(Xn ) > ER(Xn ), for n 4. Let (Xn , ) and (Xn , ) denote
the characteristic polynomials of Xn and Xn , respectively. Then, by [3, Theorem 8],
we have
0 (Xn , n) 0 (Xn , n)
ER(Xn ) = and ER(Xn ) = (3)
(Xn , n) (Xn , n)
where 0 (G, ) = d
d
(G, ). By Lemma 2, it follows that

(Xn , ) = n4 (4 n2 2 + n 3)

(Xn , ) = n4 (4 n2 + 2n 8) .

Hence

0 (Xn , n) 0 (Xn , n)
ER(Xn ) ER(Xn ) =
(Xn , n) (Xn , n)

0 (Xn , n) (Xn , n) 0 (Xn , n) (Xn , n)

=
(Xn , n) (Xn , n)

10n4 24n3 + 10n2 4n + 16

= .
(n4 n3 n 3)(n4 n3 + 2n 8)
The polynomial p() = 104 243 + 102 4 + 16 does not have any real roots,
thus the numerator p(n) is positive for all n. The real roots of the polynomials
4 3 3 and 4 3 + 2 8 are less than 2, so the denominator is positive
for n 2. It follows that ER(Xn ) ER(Xn ) > 0.

Let Cn denote the unicyclic graph obtained by attaching a pendent vertex to a

vertex of Cn1 .

Lemma 3. Let G be a unicyclic graph on n 5 vertices. If G

6= Cn , Cn , then at
least one of the following holds:

1. Mk (G) Mk (Cn ) for all k 0, and Mk (G) > Mk (Cn ) for some k0 0.
 -99-

2. Mk (G) Mk (Cn ) for all k 0, and Mk (G) > Mk (Cn ) for some k0 0.

Proof. If G
6= Cn , Cn , it follows from Lemmas 5.3, 5.4, 5.5 in [2] that G can be
transformed into either Cn or Cn in a finite number of steps, in such a way that, at
each step, the k-th spectral moment does not increase for each k, and decreases for
some k0 .

Theorem 2. Let G be a unicyclic graph on n 5 vertices. If G 6

= Cn , Cn , then
ER(G) > min{ER(Cn ), ER(Cn )}.

Proof. Follows from Lemma 3.

Using arguments that are not based on spectral moments, we can strengthen
Theorem 2 as follows:

Theorem 3. Let G be a unicyclic graph on n 5 vertices. If G 6

= Cn , then ER(G) >
ER(Cn ).

Proof. In view of Theorem 2, it is sufficient to prove that ER(Cn ) < ER(Cn ). In [3],
the validity of this latter inequality was checked for n 15. Therefore, in what
follows we may asume that n > 15, i.e., that n is sufficiently large.
Bearing in mind the relations (3), we get

d ln (G, ) 
ER(G) =
d 
=n

and therefore

d ln (Cn , ) d(ln Cn , ) 
 
ER(Cn ) ER(Cn ) =
d d 
=n

d (Cn , ) 
= ln . (4)
d (Cn , ) =n

The greatest eigenvalue of Cn is 2, and the greatest eigenvalue of Cn is certainly

less than 3. Therefore, bearing in mind that Cn and Cn contain no triangles and no
pentagons, for = n,

(Cn , ) = n n n2 + b2 (Cn ) n4 +

(Cn , ) = n n n2 + b2 (Cn ) n4 +
 -100-

and thus
b2 (Cn ) b2 (Cn )
 
(Cn , ) 1

=1+ +O
(Cn , ) 4 6
and
b2 (Cn ) b2 (Cn )
 
(Cn , ) 1
ln
= +O .
(Cn , ) 4 6
Then because of (4),

b2 (Cn ) b2 (Cn )
 
1
ER(Cn ) ER(Cn ) = 4 +O . (5)
n5 n7

Using the Sachs coefficient theorem [1], one can easily show that

1 1
b2 (Cn ) = n(n 3) and b2 (Cn ) = (n 3)(n 4) + 2n 7
2 2

from which one immediately gets that for sufficiently large values of n,

4
ER(Cn ) ER(Cn )
n5

i.e., ER(Cn ) < ER(Cn ).

3 Bicyclic graphs with maximum resolvent energy

Let (p, q, `) be the union of three internally disjoint paths Pp+1 , Pq+1 , P`+1 with com-
mon end vertices. Let Yn denote the bicyclic graph obtained from (2, 2, 1) by at-
taching n 4 pendent vertices to one of its vertices of degree 3, and let Yn denote the
bicyclic graph obtained from (2, 2, 2) by attaching n 5 pendent vertices to one of
its vertices od degree 3, as in Figure 2.

Figure 2: Bicyclic graphs with maximum resolvent energy.

 -101-

Lemma 4. Let G be a bicyclic graph on n 5 vertices, G

6= Yn , Yn . Then one of the
following holds:

1. Mk (G) Mk (Yn ) for all k 0, and Mk (G) < Mk (Yn ) for some k0 0.

2. Mk (G) Mk (Yn ) for all k 0, and Mk (G) < Mk (Yn ) for some k0 0.

Proof. Follows from Lemma 3.1, Theorems 3.2 and 3.3, and Lemma 3.4 in [5].

Theorem 4. Let G be a bicyclic graph on n 5 vertices. Then ER(G) ER(Yn ),

with equality if and only if G
= Yn .

Proof. By Lemma 4, a graph with maximum resolvent energy among n-vertex bicyclic
graphs is either Yn or Yn . Thus, it is sufficient to show that ER(Yn ) > ER(Yn ), for
n 5. Let (Yn , ) and (Yn , ) denote the characteristic polynomials of Yn and Yn ,
respectively. Then, by [3, Theorem 8], we have
0 (Yn , n) 0 (Yn , n)
ER(Yn ) = and ER(Yn ) = .
(Yn , n) (Yn , n)
By Lemma 2, it follows that

(Yn , ) = n4 4 (n + 1)2 4 + 2(n 4)

 

(Yn , ) = n4 4 (n + 1)2 + 3(n 5) .

 

Hence
0 (Yn , n) 0 (Yn , n)
ER(Yn ) ER(Yn ) =
(Yn , n) (Yn , n)

0 (Yn , n) (Yn , n) 0 (Yn , n) (Yn , n)

=
(Yn , n) (Yn , n)

16n4 34n3 + 8n2 + 2n + 60

= .
(n4 n3 n2 2n 8)(n4 n3 n2 + 3n 15)
The polynomial p(x) = 16x4 34x3 + 8x2 + 2x + 60 does not have any real roots,
thus the numerator p(n) is positive for all n. The real roots of the polynomials
x4 x3 x2 2x 8 and x4 x3 x2 + 3x 15 are less than 3, so the denominator
is positive for n 3. It follows that ER(Yn ) ER(Yn ) > 0.
 -102-

4 Tricyclic graphs with maximum resolvent energy

Let Zni , 1 i 6, be the graphs given in Figure 3.

Figure 3: Tricyclic graphs with maximum resolvent energy.

Lemma 5. Let G be a bicyclic graph on n 4 vertices such that G

6= Zni , for
1 i 6. Then, some i {1, 2, 3, 4, 5, 6}, Mk (G) Mk (Zni ) for all k 0, and
Mk (G) < Mk (Zni ) for some k0 0.

Proof. Follows from Corollaries 3.5 and 3.9 and Lemma 3.10 in [6].

Theorem 5. Let G be a tricyclic graph on n 4 vertices. Then ER(G) ER(Zn1 ),

with equality if and only if G
= Zn1 .

Proof. By Lemma 5, a graph with maximum resolvent energy among n-vertex tricyclic
graphs is equal to Zni , for some 1 i 6.
By Lemma 2, it follows that

(Zn1 , ) = n5 (5 (n + 2)3 82 + 3(n 5) + 2(n 4)) = n5 f1 ()

(Zn2 , ) = n4 (4 (n + 2)2 6 + 3(n 5)) = n4 f2 ()

(Zn3 , ) = n4 (4 (n + 2)2 + 4(n 6)) = n4 f3 ()

(Zn4 , ) = n6 (6 (n + 2)4 63 + 3(n 4)2 + 2 (n 5)) = n6 f4 ()

 -103-

(Zn5 , ) = n5 (5 (n + 2)3 42 + 4(n 4) + 4) = n5 f5 ()

(Zn6 , ) = n6 (6 (n + 2)4 + 5(n 5)2 2(n 8)) = n6 f6 () .

For 2 i 6, we have

0 (Zn1 , n) 0 (Zni , n)
ER(Zn1 ) ER(Zni ) =
(Zn1 , n) (Zni , n)

0 (Zn1 , n) (Zni , n) 0 (Zni , n) (Zn1 , n)

= .
(Zn1 , n) (Zni , n)

Straightforward calculation yields

6n6 12n5 + 42n4 18n3 42n 120

ER(Zn1 ) ER(Zn2 ) =
n f1 (n) f2 (n)

28n6 56n5 + 44n4 8n3 + 136n2 + 80n 192

ER(Zn1 ) ER(Zn3 ) =
n f1 (n) f3 (n)

6n8 + 40n6 2n5 48n4 96n3 188n2 132n 40

ER(Zn1 ) ER(Zn4 ) =
n f1 (n) f4 (n)

16n7 20n6 + 56n5 40n4 + 4n3 52n2 188n

ER(Zn1 ) ER(Zn5 ) =
n f1 (n) f5 (n)

32n8 62n7 +30n6 +92n5 +94n4 2n3 432n2 432n128

ER(Zn1 ) ER(Zn6 ) = .
n f1 (n) f6 (n)

All the real roots of the polynomials that appear in the numerators are less than
2. Moreover, all the real roots of the polynomials fi , 1 i 6, are less than 3. It
follows that the numerator end denominator in the quotients above are positive for
n 3. Hence ER(Zn1 ) ER(Zni ) > 0 for 2 i 6.

References
[1] D. M. Cvetkovic, M. Doob, H. Sachs, Spectra of Graphs Theory and Applica-
tion, Academic Press, New York, 1982.

[2] Z. Du, B. Zhou, The Estrada index of unicyclic graphs, Lin. Algebra Appl. 436
(2012) 31493159.

[3] I. Gutman, B. Furtula, E. Zogic, E. Glogic, Resolvent energy of graphs, MATCH

Commun. Math. Comput. Chem. 75 (2016) 279290.
 -104-

[4] I. Gutman, B. Furtula, E. Zogic, E. Glogic, Resolvent energy, in: I. Gutman, X.

Li (Eds.), Graph Energies Theory and Applications, Univ. Kragujevac, Kragu-
jevac, 2016, pp. 277290.

[5] L. Wang, Y. Z. Fan, Y. Wang, Maximum Estrada index of bicyclic graphs, Discr.
Appl. Math. 180 (2015) 194199.

[6] Z. Zhu, L. Tan, Z. Qiu, Tricyclic graph with maximal Estrada index, Discr. Appl.
Math. 162 (2014) 364372.