J. Comp. & Math. Sci. Vol. 1(2), 155-162 (2010).

On Minimum Metric Dimension of

Circulant Networks
1

BHARATI RAJAN, 1INDRA RAJASINGH, 1CHRIS MONICA. M*

and 2PAUL MANUEL

Department of Mathematics, Loyola College, Chennai 600 034 (India)

Department of Information Science, Kuwait University, Kuwait 13060
Email: chrismonicam@yahoo.com

ABSTRACT
Let M = {v1 , v 2 ,..., vn } be an ordered set of vertices in
a graph G. Then

(d (u, v1 ), d (u, v 2 ),..., d (u, vn )) is called the

M-coordinates of a vertex u of G. The set M is called a metric

basis if the vertices of G have distinct M-coordinates. A minimum
metric basis is a set M with minimum cardinality. The cardinality
of a minimum metric basis of G is called minimum metric
dimension and is denoted by (G). This concept has wide
applications in motion planning and in the field of robotics. In this
paper we determine the minimum metric dimension of certain
classes of circulant networks. We prove that for circulant graphs
G(n; {1, 2}), (G (n; {1, 2}) = 3, when n = 4 l, 4 l + 2,
4 l + 3, l 1 and 2 < (G (n; {1, 2}) 4, for 4 l+ 1, l 1. We
have similar results for circulant digraphs G(n; {1, 2, 3}) and
certain subclasses of circulant graphs.
Key words: metric basis, metric dimension, circulant
graphs, robotics, Cayley graph.

1. INTRODUCTION
Major issues involved in the design
of interconnection networks are quick
*This research is supported by The
Major Project-No.F.8-5-2004(SR) of University
Grants Commission, India.

communication among vertices, high

robustness, rich structure in the sense
of embeddable properties, fault tolerance
and VLSI. The circulant graphs are an
important class of topological structures
of interconnection networks. They are
symmetric with simple structures and
easy extendability. Circulant graphs have

Journal of Computer and Mathematical Sciences Vol. 1 Issue 2, 31 March, 2010 Pages (103-273)

156

Bharati Rajan et al., J.Comp.&Math.Sci. Vol.1(2), 155-162 (2010).

been used for decades in the design

of computer and telecommunication
networks due to their optimal faulttolerance and routing capabilities4. The
term circulant comes from the nature
of its adjacency matrix; a matrix is
circulant if all its rows are periodic
rotations of the first one. Circulant
matrices have also been employed for
designing binary codes. Circulant graphs
also constitute the basis for designing
certain data alignment networks for
complex memory systems18. Most of
the earlier research concentrated on
using the circulant graphs to build interconnection networks for distributed
and parallel systems2,4. For example,
undirected circulant networks arise in
the context of Mesh Connected Computer suited for parallel processing of
data, such as the well-known ILLIAC
type computers1. Generally, the ILLIAC
network with n2 processors can be
represented as a circulant graph G(n2;
{1, n}).
By using circulant graph we can
adapt the performance of the network
to user needs. For a given number of
processors and a given crossbar switch
technology, we can choose the performance of the network. If subsequently,
the user needs to increase this
performance we can increase the
degree of the circulant graph without
changing the number of processors.
The opposite modification is also
possible, we can increase the number
of processors without changing the
degree of the circulant graph. This

flexibility which is not possible with

other topologies, allows us to optimize
the ratio price/performance according
the user. The family of circulant graphs
includes the complete graph and the
cycle among its members.
Circulant graphs are intensively
researched in computer science, graph
theory and discrete mathematics.
2. An Overview of the Paper
Let M = {v1 , v2 ,..., v n } be an
ordered set of vertices in a graph G.
Then (d (u , v1 ),d (u , v2 ),..., d ( u, vn )) is
called the M-coordinates of a vertex u
of G. The set M is called a metric basis
if the vertices of G have distinct Mcoordinates. A minimum metric basis
is a set M with minimum cardinality.
If M is a metric basis then it is clear
that for each pair of vertices u and v
of V \ M, there is a vertex w M such
that d(u, w) d(v, w). The cardinality
of a minimum metric basis of G is
called minimum metric dimension and
is denoted by (G); the members of a
metric basis are called landmarks. A
minimum metric dimension (MMD)
problem is to find a minimum metric
basis.
This problem has application in
the field of robotics. A robot is a
mechanical device which is made to
move in space with obstructions around.
It has neither the concept of direction
nor that of visibility. But it is assumed
that it can sense the distances to a set

Journal of Computer and Mathematical Sciences Vol. 1 Issue 2, 31 March, 2010 Pages (103-273)

Bharati Rajan et al., J.Comp.&Math.Sci. Vol.1(2), 155-162 (2010).

of landmarks. Evidently, if the robot
knows its distances to a sufficiently large
set of landmarks its position in space
is uniquely determined.
The concept of metric basis and
minimum metric basis has appeared
in the literature under a different name
as early as 1975. Slater in16 and later
in17 had called these sets as locating
sets and reference sets respectively. Slater
called the cardinality of a reference
set as the location number of G. He
described the usefulness of these ideas
when working with sonar and loran
stations. Independently, Harary and
Melter8 discovered these concepts as
well but used the terms metric basis and
metric dimension, rather than reference
set and location number. Chartrand et
al.6 have called a metric basis and a
minimum metric basis as a resolving
set and minimum resolving set. We
adopt the terminology of Harary and
Melter.

157

Garey and Johnson7 noted that

this problem is NP-complete for general
graphs by a reduction from 3-dimensional matching. Recently Manuel et
al. 15 have proved that the MMD
problem is NP-complete for bipartite
graphs by a reduction from 3-SAT
narrowing down the gap between the
polynomial classes and NP-complete
classes for the MMD problem.
In this paper, we derive a minimum
metric basis for certain classes of
circulant networks.
3. Topological Properties of Circulant
Networks
The circulant network is a
natural generalization of double loop
network, which was first considered
by Wong and Coppersmith18. A circulant
undirected graph, denoted by G(n;
{1, 2 j}), 1<j n/2, n 3 is defined
as an undirected graph consisting of
the vertex set V = {0, 1 n 1} and
the edge set E = {(i, j): j i s (mod n),

If G has p vertices then it is

clear that 1(G) p 1. Also for the
complete graph Kp, the cycle Cp and
the complete bipartite graph Km,n, the
minimum metric dimensions are (Kp)
= p 1, (Cp) = 2 and (Km, n) = m
+ n 2 8. This problem has been
studied for grids11, trees, multi-dimensional grids10, Petersen graphs3, Torus
Networks 13 , Benes and Butterfly
networks15 and Honeycomb networks14.

s{1, 2 j}}.
The graph in Figure 1 is the
circulant graph G(8; {1, 2}). It is also
clear that G(n;1) is an undirected
cycle C n and G(n; {1, 2 n/2}) is
a complete graph K n . We observe
e
that C n = G(n, 1) is a subgraph of
G(n; {1, 2 j}) for every j, 1 < j
n/2.

Journal of Computer and Mathematical Sciences Vol. 1 Issue 2, 31 March, 2010 Pages (103-273)

158

Bharati Rajan et al., J.Comp.&Math.Sci. Vol.1(2), 155-162 (2010).

0

is hamiltonian, and is a Cayley graph.

In this paper, we make use of an

interesting property of circulant graphs,
that it is diametrically uniform12.

3
4

Figure 1. Circulant undirected

graph G(8; {1,2})
A circulant digraph was proposed
by Elspas and Turner 5. A circulant
digraph, denoted by G(n; {1, 2 j}),
1<j n1, n2 is defined as a digraph
consisting of the vertex set V = {0, 1
n 1} and the edge set E = {(i, j):

j i s (mod n) , s {1, 2 j}}. See

For each vertex u of a graph G,

the maximum distance d(u, v) to any
other vertex v of G is called its
eccentricity and is denoted by ecc(u).
In a graph G, the maximum value of
eccentricity of vertices of G is called
the diameter of G and is denoted by
. Let G be a graph with diameter . A
vertex v of G is said to be diametrically
opposite to a vertex u of G, if dG(u, v)=.
A graph G is said to be a diametrically
uniform graph if every vertex of G has
at least one diametrically opposite vertex.
The set of diametrically opposite
vertices of a vertex x in G is denoted
by D(x).

Figure 2. It is clear that G(n; 1) is a directed

cycle C n and G(n; {1,2n1 }) is a

4. Minimum Metric Dimension of

Circulant Networks

complete digraph K n .
Theoretical properties of circulant
graphs have been studied extensively
and are surveyed in2. A circulant graph

Theorem 110: Let G be a graph

with minimum metric dimension 2 and
let {u,v}V be a metric basis in G.
Then the following are true:
a) There is a unique shortest path
between u and v.
b) The degree of each u and v is
at most 3.

Figure 2. Circulant digraph

G(8; {1, 2})

Since G(n;{1, 2 j}), 1<j

n/2 is 2j-regular, we have the
following Lemma as an application of
Theorem 1.

Journal of Computer and Mathematical Sciences Vol. 1 Issue 2, 31 March, 2010 Pages (103-273)

Bharati Rajan et al., J.Comp.&Math.Sci. Vol.1(2), 155-162 (2010).

Lemma 1: (G(n; {1, 2 j}))
> 2, 1 < j n/2.
Proposition 1: The undirected
circulant graph G(n;{1, 2}) is
diametrically uniform. Moreover, for
any vertex x in G(n; {1, 2}), the set
D(x) of diametrically opposite vertices
of x satisfies
1.

D( x) 1 if n = 4l + 2

2.

D( x) 2 if n = 4l + 3

3.

D( x) 3 if n = 4l

4.

D( x) 4 if n = 4l + 1.

159

Fix a V . Then the mirror through the

vertex a divides the vertex set of G(n;
{1, 2}) into two sets S1 and S2. Let
S1= { a1 , a2 ,...,a t }, S2 = { b1 , b2 ,...,bt }
for t= ( n 1) / 2 . Then ai and bi are
images of each other with respect to
the mirror through a. See Figure 3.

The proof of Proposition 1

follows by considering the breadth first
search tree rooted at the vertex x of
the graph.
Definition 1: For any vertex x
of G(n; {1, 2}), the geometric diameter
of C n through x is called a mirror.
For any x V , if D(x) is odd,,
then the mirror through x passes
through one of the members of D(x).
This vertex is denoted by x*.

(n 1) / 2 .

Proposition 2: Let t =
Then

{a *, at , bt }
{a , a , b , b }

D( a) t 1 t t 1 t
{a *}
{a t , bt }

if
if
if
if

n 4l
n 4l 1
n 4l 2
n 4l 3

for any vertex a of G(n; {1, 2}).

Figure 3. G(12; {1, 2})

Lemma 2: Let G(n; {1, 2}) be
an undirected circulant graph where
n=4l, 4l+ 2 or 4l + 3, l 1. Let M =
{a, a1}. Then any two vertices in S1 ( S2 )
have distinct M-coordinates.
Proof: Let u, v S1 be such that
d(a, u)=d(a, v). Then u and v are consecutive vertices on C n . Let u a i ,

v ai 1 . Without loss of generality,,

let i be odd. The shortest paths

aa1a 3 a5 ...ai and aa 2 a 4 ...a i 1 are of

Journal of Computer and Mathematical Sciences Vol. 1 Issue 2, 31 March, 2010 Pages (103-273)

160

Bharati Rajan et al., J.Comp.&Math.Sci. Vol.1(2), 155-162 (2010).

equal length. Then the lengths of shortest

paths aa 2 a 4 ...a i 1 and a1 a 2 a 4 ...ai 1
are the same and
d(a1, ai) = d(a, ai) 1
= d(a,ai+1) 1
= d(a1, ai+1)1 d(a1, ai+1).
Thus any two vertices in S1 have
distinct M-coordinates. The proof is
similar for any two vertices in S2.
Theorem 2: Let G(n; {1, 2}) be
an undirected circulant graph. Then,
for l 1 ,
1. (G (n; {1, 2}) = 3,
when n = 4l, 4l + 2, 4l +3.
2. 2 < (G (n; {1, 2}) 4,
when n = 4l + 1.
Proof: Let n = 4l. Let a be a
vertex of G(4l;{1,2}). By the structure
of G(4l; {1, 2}), vertices a2 i 1 , a2 i ,

b2 i 1 , b 2i , 1 i t 2 are at distance
i from a. Note that i < . The vertices
at distance are a*, at , bt where a* is
a vertex on the mirror through a. Let
M={a, a1}. Then by Lemma 2, any two
vertices in S1(S2) have distinct M-coordinates. Now among all pairs (x, y), x
S1, y S2 only the pairs ( a2 i , b2 i 1 ),

Figure 4. G(9; {1, 2}) and

M = {a , p, q, r}
Similarly, for i > 2,
d( b2 i 1 , a2 )=d( b2 i 1 , b3 )+d( b3 , a2 )
= i 2 + d( b3 , a2 ).
Hence d( a2 i , a2 ) d( b2 i 1 , a2 ), for

1 i t 2 1. By Theorem 1, (G(n;
{1, 2}) > 2. Thus M = {a, a1 , a2 } is a
minimum metric basis. Proofs for G(n;
{1,2}) when n=4l+2, 4l+3 or 4l+1
are similar to that of G(n; {1, 2}).
It is important to note that the
mirror through the vertex a does not
pass through any of the diametrically
opposite vertices of a when n= 4l +1
or 4l+ 3. See Figure 4 and Figure 5.

1 i t 2 1 , where a2 (t 2 1) = a*,
are equidistant from both a and b. Now,
include a2 in M. It ready follows that
d( a4 , a2 )d( b3 , a2 ). Now, for
or i>2,
d( a2 i , a2 ) = d( a2 i , a4 ) + d( a4 , a2 )
= i 2 + d( a4 , a2 ).

Figure 5. G(11; {1, 2}) and

M = {a , p, q}

Journal of Computer and Mathematical Sciences Vol. 1 Issue 2, 31 March, 2010 Pages (103-273)

Bharati Rajan et al., J.Comp.&Math.Sci. Vol.1(2), 155-162 (2010).

The following theorems can be
proved proceeding along the same
lines as in Theorem 2.
Theorem 3: Let G(n; {1, 2,
3}), n 8 be a circulant undirected
graph. Then
1. 2 < (G(n, {1, 2, 3})) 5,
when n = 6l, l 2 .
2. 2 < (G(n, {1, 2, 3})) 4,
when n = 6l + 2, 6l+ 4, l 1 .
3. 2 < (G(n, {1, 2, 3})) 5,
when n = 6l + 1, l 2 .
4. 2 < (G(n, {1, 2, 3})) 4,
when n = 6l + 3, 6l + 5, l 1 .

Figure 6. G(9; {1, 2, 3}) and

M = {p, q, r}
A metric basis of a digraph G(V,
E) is a subset M V such that for
each pair of vertices u, v V \ M, there
exists a vertex w M such that d(w, u)
d(w, v). We have the following conjecture. See Figure 6.
Conjecture 1: Let G (n; {1, 2 j}
be a directed circulant graph. Then

(G (n; {1, 2 j}) = j, 1 j n 1 .

161

4. CONCLUSION
We have obtained the minimum
metric dimension of undirected circulant
graphs G(n; {1, 2}), G(n; {1, 2,
3}). The problem for G(n; {1, 2 j}),
j > 2 is under investigation.
R E FE R E N C ES
1. G.H. Barnes, R.M. Brown, M. Kato,
D. J. Kuck, D. L. Slotnick, R. A.
Stokes, The ILLIAC IV computers,
IEEE Transactions on computers,
17, 746-757(1968).
2. J.C. Bermond, F. Comellas, and D.F.
Hsu, Distributed loop computer
networks: A survey, ournal of
Parallel and Distributed Computing,
24, 2-10 (1995).
3. Bharati Rajan, Indra Rajasingh, J.
A. Cynthia and Paul Manuel, On
Minimum Metric Dimension, The
Indonesia-Japan Conference on
Combinatorial Geometry and Graph
Theory, September 13-16, 2003,
Bandung, Indonesia.
4. F.T. Boesch and J. Wang, Reliable
Circulant Networks with Minimum
Transmission Delay, IEEE Transactions on Circuit and Systems, 32,
1286-1291 (1985).
5. B. Elspas and J. Turner, Graphs
with circulant adjacency matrices,
Journal of Combinatorial Theory,
9, 297-307 (1970).
6. Gary Chartrand, Linda Eroh, Mark
A. Johnson, Ortrud R. Oellermann,
Resolvability in graphs and the
metric dimension of a graph,
Discrete Applied Mathematics,

Journal of Computer and Mathematical Sciences Vol. 1 Issue 2, 31 March, 2010 Pages (103-273)

162

7.

8.

9.

10.

11.

12.

13

Bharati Rajan et al., J.Comp.&Math.Sci. Vol.1(2), 155-162 (2010).

105, 99-113 (2000).
M. R. Garey and D. S. Johnson,
Computers and Intractability: A
Guide to the Theory of NP Completeness, W. H. Freeman and
Company (1979).
F. Harary and R. A. Melter, The
metric dimension of a graph, Ars
Combinatorica, 2, 191-195 (1976).
S. Khuller, E. Rivlin and A. Rosenfeld,
Graphbots: Mobility in discrete
spaces, IEEE Transactions on
Systems, Man and Cybernetics,
28(1), 29-38 (1998).
S. Khuller, B. Ragavachari, and A.
Rosenfeld, Landmarks in Graphs,
Discrete Applied Mathematics, 70,
217-229 (1996).
R.A. Melter and I. Tomcscu, Metric
bases in digital geometry, Computer
Vision, Graphics, and Image processing, 25, 113-121 (1984).
Paul Manuel, Bharati Rajan, Indra
Rajasingh and Amutha Alaguvel,
Tree Spanners, Cayley Graphs
and Diametrically Uniform Graphs,
LNCS, 2880, 334-345 (2003).
Paul Manuel, Bharati Rajan, Indra
Rajasingh, Chris Monica. M, Landmarks in Torus Networks, Journal

14.

15.

16.
17.

18.

19.

of Discrete Mathematical Sciences

& Cryptography, 9(2), 263-271
(2006).
Paul Manuel, Bharati Rajan, Indra
Rajasingh, Chris Monica. M, On
Minimum Metric Dimension of
Honeycomb Networks, Journal of
Discrete Algorithm, 6(1), 20-27
(2008).
Paul D. Manuel, Mostafa I. AbdEl-Barr, Indra Rajasingh and Bharati
Rajan, An Efficient Representation
of Benes Networks and its Applications, Journal of Discrete Algorithm,
6(1), 11-19 (2008).
P.J. Slater, Leaves of trees, Congr.
Number., 14, 549-559 (1975).
P.J. Slater, Dominating and reference sets in a graph, J. Math.
Phys. Sci., 22, 445-455(1988).
G.K. Wong and D.A. Coppersmith,
A combinatorial problem related
to multimodule memory organization, Journal of Association for
Computing Machinery, 21, 392401 (1974).
J. Xu, Topological Structures and
Analysis of Interconnection Networks, Kluwer Academic Publishers
(2001).

Journal of Computer and Mathematical Sciences Vol. 1 Issue 2, 31 March, 2010 Pages (103-273)