A Note on Gutmans Conjecture

1
Dragan Stevanovic dragance@pmf.ni.ac.yu

Faculty of Science, Visegradska 33, 18000 Nis, Serbia and Montenegro

Let G = (V, E) be a simple graph, with the degree of a vertex i V denoted

by di . In [1] G is called SQR if ( di )iV is an eigenvector of (an adjacency
matrix of) G. Trivially, connected regular and semiregular graphs are SQR
and Gutman conjectured in [1] that a connected SQR graph is either regular
or semiregular.
A symmetric matrix A of order n all of whose entries are nonnegative, and
all of whose row sums d1 , d2 , . . . , dn are positive, is called almost regular if
there is a positive number r such that, if aij > 0, then di dj = r2 . Clearly, if
G is connected and either regular or semiregular, then its adjacency matrix is
almost regular.
Actually, Gutmans conjecture may be generalized to say that if an irre-
ducible nonnegative symmetric matrix A with positive row sums d1 , d2 , . . . , dn
has an eigenvector ( di )ni=1 then A is almost regular.
The following theorem is proved in [2].
Theorem 1 ([2]) Let A be a P nonnegative symmetric matrix with positive row
n
sums d1 , . . . , dn and set m = i=1 di . Then
P p s
i,j aij di dj m
(A) P aij , (1)
m i,j di dj

where (A) denotes the spectral radius of A, and A is almost regular if and only
if any two of these functions in A in (1) are equal, in which case all three are
equal.

Proof of this theorem also contains proof of generalized Gutmans conjecture.

For the sake of completeness, we reproduce here just the corresponding part of
the proof.
First, notice that ( di )ni=1 is a positive eigenvector of a nonnegative ma-
trix A, and thus, its corresponding eigenvalue is the spectral radius (A). Now,
if all the di are equal, we are done. Otherwise set = min{d1 , . . . , dn } and
= max{d1 , . . . , dn }. Choose u and v such that du = and dv = . Now
assume that there exists w with auw > 0 and dw < . Then we have
n r n r
X dj X
(A) = auj < auj = .
j=1
j=1

1 The author gratefuly acknowledges support from the Grant 1389 of the Serbian Ministry

of Science, Technology and Development.

1
On the other hand,
n r n r
X dj X
(A) = avj avj = ,
j=1
j=1

which is a contradiction.
Assuming existence of w with avw > 0 and dw > leads to an analogous
contradiction. We conclude that whenever aij > 0 then di = and dj = or
vice versa, and that A is almost regular matrix with (A) = .

References
[1] I. Gutman, A Spectral Property of Graphs, Graph Theory Notes of New York
XLIV (2003), 1013.
[2] A.J. Hoffman, P. Wolfe, M. Hofmeister, A Note on Almost Regular Matrices,
Linear Algebra Appl. 226228 (1995), 105108.