On distances in Sierpiński graphs: almost-extreme vertices

and metric dimension

Sandi Klavžar
Faculty of Mathematics and Physics, University of Ljubljana
Jadranska 19, 1000 Ljubljana, Slovenia
and
Faculty of Natural Sciences and Mathematics, University of Maribor
Koroška 160, 2000 Maribor, Slovenia
sandi.klavzar@fmf.uni-lj.si
Sara Sabrina Zemljič
Institute of Mathematics, Physics and Mechanics
Jadranska 19, 1000 Ljubljana, Slovenia
sara.zemljic@gmail.com

July 20, 2012

Abstract
Sierpiński graphs Spn form an extensively studied family of graphs of fractal na-
ture applicable in topology, mathematics of the Tower of Hanoi, computer science,
and elsewhere. An almost-extreme vertex of Spn is introduced as a vertex that is
either adjacent to an extreme vertex of Spn or is incident to an edge between two
subgraphs of Spn isomorphic to Spn−1 . Explicit formulas are given for the distance
in Spn between an arbitrary vertex and an almost-extreme vertex. The formulas
are applied to compute the distance of almost-extreme vertices and to obtain the
metric dimension of Sierpiński graphs.

Keywords: Sierpiński graph; almost-extreme vertex; distance of vertex; metric di-

mension

AMS Subj. Class. (2010): 05C12

1 Introduction
Sierpiński graphs Spn were introduced from at least three reasons. In [17], then were
motivated by topological studies of universal spaces (cf. [16]) and the fact that the
base-3 Sierpiński graphs S3n are isomorphic to the Tower of Hanoi graphs on 3 pegs.

1
Independently, a class of graphs called WK-recursive networks was introduced in com-
puter science in [3], see also [4]. WK-recursive networks are very similar to Sierpiński
graphs, they can be obtained from Sierpiński graphs by adding a link (an open edge)
to each of its extreme vertices.
Graphs Spn were studied by now from numerous points of view, the reader in invited
to the recent paper [11] about colorings of these graphs and references therein; see
also [5] for more coloring results. From among other investigations we explicitly mention
only a few of them. An appealing application of Sierpiński graphs is due to Romik [22]
who designed, based on Sierpiński labelings, a finite automaton utmost useful for the
Tower of Hanoi problem. In [18] the structure of Sierpiński graphs was the key to
determine the exact genus for a first infinite family of fractal graphs. Recently, the hub
number of Sierpiński-like graphs was determined in [14].
Metric issues received a special attention on Sierpiński graphs. This is in particular
due to the fact that shortest paths in the base-3 Sierpiński graphs correspond to optimal
solutions in the Tower of Hanoi puzzle. In the seminal paper [17] a formula for the
distance between vertices in Spn was proved, we state it as Theorem 2.2. Then, in [10],
additional metric properties of these graphs were investigated, in particular establishing
a connection with the Stern’s diatomic sequence via regular states in the Tower of Hanoi
that admits two different optimal solutions. Parisse [19] followed with a paper in which
he studied, among other matters, the diameter, the eccentricity, the radius, and the
center of these graphs. Wiesenberger [24] obtained a formula for the average distance
in Spn . The formula is far from being trivial, it extends over several lines! Very recently,
Hinz and Parisse [12] succeeded to determine the average eccentricity and standard
deviation for all Sierpiński graphs.
The paper is organized as follows. In the next section definitions, concepts, and
results needed in this paper are given. Then, in Section 3, we obtain distances between
almost-extreme vertices and other vertices. The advantage of the obtained formulas
compared to Theorem 2.2 is that we need not to compute the minima of related expres-
sions. In the subsequent section we use the obtained formulas to compute the distance
of almost-extreme vertices. In the final section we make another application by com-
puting the metric dimension of Sierpiński graph. We point out here that in general it
is very difficult to determine the exact metric dimension, see [9] and references therein
for the complexity issues on metric dimension.

2 Preliminaries
The graphs considered are simple and connected. The distance dG (u, v) between ver-
tices u and v in a graph G is the standard shortest path distance. We also write d(u, v)
if the graph G is clear from the context. The set {1, 2, . . . , n} is shortly denoted with
[n].
Let p, n ∈ N, then the Sierpiński graph Spn is defined on vertex set [p]n . Two vertices

2
s = s1 s2 . . . sn and t = t1 t2 . . . tn are adjacent if and only if there exists a δ ∈ [n] such
that (i) sd = td , for d ∈ [δ − 1]; (ii) sδ 6= tδ ; and (iii) sd = tδ and td = sδ for d ∈ [n] \ [δ].
For any n, S1n ∼ = K1 and S2n ∼ = P2n . Note also that Sp1 ∼ = Kp for any p. For S53 see
Fig. 1. Let n ≥ 2, then for i ∈ [p], let iSpn−1 be the subgraph of Spn induced by the
vertices of the form s = is2 . . . sn . The subgraph iSpn−1 is isomorphic to Spn−1 .
Vertices of the form i . . . i = in are called extreme vertices. Clearly, Spn contains p
extreme vertices and they are of degree p − 1; all the other vertices are of degree p. For
n, p ≥ 2 we introduce almost-extreme vertices of Spn as the vertices that are either of
the form i . . . ij = in−1 j or ij . . . j = ij n−1 , where i 6= j. In Fig. 1 the extreme vertices
of S53 are emphasized with filled circles and the almost-extreme vertices are emphasized
as triangles (vertices of the form ij 2 ) and as diamonds (vertices of the form i2 j).

111
115 112

114 113
155 122

211

144 133
555 222

444 333

Figure 1: S53 with its extreme and almost-extreme vertices emphasized

3
 Obviously, there are p(p − 1) vertices of the form in−1 j and also p(p − 1) vertices of
the form ij n−1 . The almost-extreme vertex in−1 j is adjacent to the extreme vertex in
and the almost-extreme vertex ij n−1 is incident to the edge between iSpn−1 and jSpn−1 .
Note also that in Sp2 , p ≥ 2, any vertex is either extreme or almost-extreme.
The distance between a vertex of Spn and an extreme vertex can be computed as
follows, where we use Iverson’s convention that (X) = 1, if statement X is true, and
(X) = 0, if X is false.

Lemma 2.1 [17] For any j ∈ [p] and any vertex s = s1 . . . sn of Spn ,
n
X
n
d(s, j ) = (sd 6= j) · 2n−d .
d=1

Moreover, there is exactly one shortest path between s and j n .

Note that it immediately follows from Lemma 2.1 that d(in , j n ) = 2n − 1 for any
i 6= j. More generally, the distance between arbitrary vertices of Spn can be computed
in the following way:

Theorem 2.2 [17] Let s = sis and t = sjt be vertices of Spn , where i 6= j and s, t ∈
[p]δ−1 , δ ∈ [n]. Then
n
d(s, t) = min d(s, j δ−1 ) + 1 + d(t, iδ−1 ),
o
d(s, k δ−1 ) + 1 + 2δ−1 + d(t, k δ−1 ) | k ∈ [p] \ {i, j} .

Note that in the above theorem, δ is the index of the first position in which differ
s and t.

3 Distances to almost-extreme vertices

In this section we simplify Theorem 2.2 for the case of almost-extreme vertices and
begin with the almost-extreme vertices that are adjacent to extreme vertices.

Proposition 3.1 Let n, p ≥ 2, let j n−1 k be an almost-extreme vertex of Spn , and let
is, s ∈ [p]n−1 , be an arbitrary vertex of Spn . Then

 d(s, j n−1 ) + 2n−1 ;


i 6= j, k ,
n−1
d(is, j k) = d(s, j n−1 ) + 2n−1 − 1; i = k ,
d(s, j n−2 k); i=j.


4
Proof. Set s = is. For i = j, d(js, j n−1 k) = d(s, j n−2 k).
If i = k, the distance d(ks, j n−1 k) equals d(s, j n−1 )+1+d(k n−1 , j n−2 k) = d(s, j n−1 )+
2 n−1 − 1, since all the other distances from the minimum in Theorem 2.2 are at least
2 > 2n − 1.
n

Let now i 6= j, k. Since i 6= j, we have s1 6= j and hence Lemma 2.1 implies that
d(s, j n ) = 2n−1 + d(s, j n−1 ). By Theorem 2.2,

d(s, j n−1 k) = min{d(s, j n−1 ) + 1 + d(in−1 , j n−2 k),

d(s, k n−1 ) + 1 + 2n−1 + d(k n−1 , j n−2 k), (1)
n−1 n−2 n−1 n−1
d(` ,j k) + 1 + 2 + d(s, k ), ` 6= i, j, k} .

For ` 6= i, j, k we have

d(`n−1 , j n−2 k) + 1 + 2n−1 + d(s, k n−1 ) = 2n + d(s, k n−1 ) > 2n − 1 .

Moreover,

d(k n−1 , j n−2 k) + 1 + 2n−1 + d(s, k n−1 ) = 2n − 1 + d(s, k n−1 ) .

The latter distance is equal to 2n −1 if and only if s = k n−1 . In that case, d(in−1 , j n−2 k)+
1 + d(s, j n−1 ) = 2n − 1. We conclude that the minimum in (1) is always achieved in
the first term, and is equal to 2n−1 + d(s, j n−1 ). 

Note that, by Proposition 3.1, for s ∈ iSpn−1 , where i 6= j, k, holds d(s, j n−1 k) =
d(s, j n ).
To state a similar formula for the other almost-extreme vertices, we will make use
of the following notation. Let s = s1 s2 . . . sn ∈ Spn , then set

n + 1; for j = 1, 2, . . . , n : sj 6= i,
firsti (s) =
min{j | sj = i}; otherwise .

Proposition 3.2 Let n, p ≥ 2, let jk n−1 be an almost-extreme vertex of Spn , and let
s = is, s ∈ [p]n−1 , be an arbitrary vertex of Spn . Then

d(s, k n−1 ) + 2n−1 + 1;


 i 6= j, k and firstj (s) > firstk (s) ,
 d(s, j n−1 ) + 2n−1 ;

i 6= j, k and firstj (s) < firstk (s) ,


d(s, jk n−1 ) = 2n − 1; i 6= j, k and firstj (s) = firstk (s) ,
d(s, j n−1 ) + 1; i = k,




d(s, k n−1 ); i=j.


Proof. For i = j, d(js, jk n−1 ) = d(s, k n−1 ).

If i = k, then d(ks, jk n−1 ) = d(s, j n−1 ) + 1, since all the other distances from the
minimum in Theorem 2.2 are at least 2n > 2n − 1.

5
 Let now i 6= j, k. By Theorem 2.2,

d(s, jk n−1 ) = min{d(s, j n−1 ) + 2n−1 , d(s, k n−1 ) + 2n−1 + 1,

d(s, `n−1 ) + 2n ; ` ∈ [p] \ {i, j, k}} .

Note that for every ` ∈ [p] \ {i, j, k},

d(s, `n−1 ) + 2n > 2n − 1 ,

and therefore this path cannot be a shortest path.

Furthermofirstj (s) > firstk (s). Thus s = s1 . . . sr ksr+1 . . . sn−r−2 and s1 , . . . , sr ∈
[p] \ {j, k}. Then by Lemma 2.1 and because s1 , . . . , sr 6= j it follows that

d(s1 . . . sr ksr+1 . . . sn−2 , j n−1 ) = 2n−1 + · · · + 2n−1−r + d(ksr+1 . . . sn−2 , j n−1−r )

= 2n−1 + · · · + 2n−2−r + d(sr+1 . . . sn−2 , j n−2−r ) .

Similarly, since s1 , . . . , sr 6= k,

d(s1 . . . sr ksr+1 . . . sn−2 , k n−1 ) = 2n−1 + · · · + 2n−1−r + d(sr+1 . . . sn−2 , k n−2−r ) .

Thus, for any 0 ≤ r ≤ n − 2,

d(sr+1 . . . sn−2 , k n−2−r ) + 1 ≤ 2n−2−r ≤ 2n−2−r + d(sr+1 . . . sn−2 , j n−2−r ) ,

hence the minimum is obtained with d(s, k n−1 ) + 2n−1 + 1.

In the cases firstj (s) < firstk (s) and firstj (s) = firstk (s) we argue along the same
lines, hence we skip the details. 

Proposition 3.2 is illustrated in Fig. 2 on S64 , where i = 1, j = 2, and k = 5. The

subgraph 1S63 is drawn explicitly, its vertices are emphasized as follows. The vertices
that satisfy the first line of the proposition’s equation are drawn black. Those vertices
that satisfy the second line in white and vertices that satisfy third line are gray.

4 Distance of almost-extreme vertices

Let u be a vertex of a connected graph G. Then the distance dG (u) of u is defined as
X
dG (u) = dG (u, v) .
v∈V (G)

Whenever G will be clear from the context we will write d(u) instead of dG (u). The
distance of a vertex in particular plays an important role in mathematical chemistry,
cf. [15], because it is a basic stone for the famous Wiener index of a graph. In this
section we determine the distance of almost-extreme vertices of Sierpiński graphs.
To make the paper self-contained we first reprove the following result that can be
found in [24] as well as in the proof of [19, Corollary 2.6].

6
 1111 1222

1S63

1666 1333

1555 1444

Figure 2: Illustration of Proposition 3.2 on S64

Lemma 4.1 For any n, p ≥ 1 and any i ∈ [p],

dSpn (in ) = pn−1 (p − 1) (2n − 1) .

Proof. Compute as follows:

n n
X X X p − 1 n X n−d
d(s, in ) = (sd 6= i) · 2n−d = p 2
p
s∈V (Spn ) s∈V (Spn ) d=1 d=1
n−1 n
=p (p − 1) (2 − 1) .

Using Lemma 4.1 and Proposition 3.1 we arrive at:

7
Theorem 4.2 Let n, p ≥ 2, and let j n−1 k be an almost-extreme vertex of Spn . Then

pn−1 2n (p − 1)2 − p(p − 1) − 1 + 1

n−1
dSpn (j k) = + 1.
p−1
Proof. By Proposition 3.1, in each subgraph iSpn−1 , i 6= j, k, we add for each vertex
s ∈ iSpn−1 the value d(s, j n−1 ) + 2n−1 . This gives the distance of the vertex j n−1 plus
pn−1 2n−1 in each such subgraph. Similarly (by Proposition 3.1), for kSpn−1 we add
dSpn−1 (j n−1 ) + (2p)n−1 − pn−1 . Hence
  X
d(j n−1 k) = (p − 1) dSpn−1 (j n−1 ) + (2p)n−1 − pn−1 + d(s, j n−2 k) .
js∈jSpn−1

For jSpn−1 we proceed analogously as for Spn , and similarly for each j ` Spn−` , ` < n − 1.
In each such subgraph we get
X
(p − 1)dSpn−`−1 (j n−`−1 + pn−`−1 2n−`−1 ) − pn−`−1 + d(s, j n−`−2 k) .
j `+1 s∈j `+1 Spn−`−1

At the end we add p − 1, since each vertex in j n−1 Sp1 is at distance 1 from j n−1 k.
Therefore,
X n−2
Xh   i
n−2
d(s, j k) = (p − 1) dSpn−`−1 (j n−`−1 ) + (2p)n−`−1 − pn−`−1 + p − 1 .
js∈jSpn−1 `=1

Consequently,
 
d(j n−1 k) = (p − 1) dSpn−1 (j n−1 ) + (2p)n−1 − pn−1 + p − 1
n−2
Xh   i
+ (p − 1) dSpn−`−1 (j n−`−1 ) + (2p)n−`−1 − pn−`−1
`=1
= (p − 1) pn−2 (p − 1) 2n−1 − 1 + (2p)n−1 − pn−1 + p − 1

n−2
Xh     i
+ (p − 1) pn−`−2 (p − 1) 2n−`−1 − 1 + (2p)n−`−1 − pn−`−1 ,
`=1

from which the claimed expression is obtained in a straightforward manner. 

Next, using Lemma 4.1 and Proposition 3.2 we can also give an expression for the
distance of the other almost-extreme vertices:

Theorem 4.3 Let n, p ≥ 2, and let jk n−1 be an almost-extreme vertex of Spn . Then

pn−1 2n+1 (p2 − 2) − p2 + 4 − p(p − 2)n

n−1
dSpn (jk )= .
2(p + 2)

8
Proof. Using Proposition 3.2 and the symmetry of Sierpinski graphs we have:
X X X
dSpn (jk n−1 ) = d(s, k n−1 ) + (d(s, j n−1 ) + 1) + d(s, jk n−1 )
js∈jSpn−1 ks∈kSpn−1 s∈iSpn−1
∀i6=j,k
X
= dSpn−1 (k n−1 ) + dSpn−1 (j n−1 ) + pn−1 + (p − 2) d(s, jk n−1 ) . (2)
s∈iSpn−1
i6=j,k

Note that in each subgraph i1 . . . i` Spn−` , where i1 , . . . , i` 6= j, k and ` < n − 1, all the
vertices of the subgraph i1 . . . i` kSpn−`−1 satisfy the first line in equation in Proposi-
tion 3.2, therefore (also by Lemma 2.1),

d(i1 i2 . . . i` ks, jk n−1 ) = d(s, k n−1 ) + 2n−1 + 2n−2 + · · · + 2n−` + 1

 
= d(s, k n−1 ) + 2n−` 2` − 1 + 1 .

Furthermore,
X X    
d(s, jk n−1 ) = dSpn−`−1 (k n−`−1 ) + 2n−` 2` − 1 + 1
s∈i1 ...i` kSpn−`−1 s∈i1 ...i` kSpn−`−1
   
= dSpn−`−1 (k n−`−1 ) + pn−`−1 2n−` 2` − 1 + 1 .

In each subgraph i1 . . . i` Spn−` , where i1 , . . . , i` 6= j, k and ` < n − 1, all the vertices

of the subgraph i1 . . . i` jSpn−`−1 satisfy the second line in equation in Proposition 3.2.
Similarly we get
X X   
d(s, jk n−1 ) = dSpn−`−1 (j n−`−1 ) + 2n−` 2` − 1
s∈i1 ...i` jSpn−`−1 s∈i1 ...i` jSpn−`−1
  
= dSpn−`−1 (j n−`−1 ) + pn−`−1 2n−` 2` − 1 .

Repeating this argument recursively for ` = 2, . . . , n − 2, we cover all the vertices of

iSpn−1 but those with firstj (s) = firstk (s). By Proposition 3.2, d(s, jk n−1 ) = 2n − 1
holds for vertices with firstj (s) = firstk (s) = n + 1 and there are exactly (p − 2)n−1
such vertices in iSpn−1 . In i1 . . . in−1 Sp1 we add to the sum 2n − 2 and 2n − 1 for
vertices i1 . . . in−1 j and i1 . . . in−1 k, respectively. There are (p − 2)n−1 pairs of such
vertices in Spn . Since there are (p − 2)` subgraphs i1 . . . i` S
pn−` , i1 , . . . i` 6= j, k in iSpn−1
and dSpn−1 (k n−1 ) = dSpn−1 (j n−1 ) = pn−2 (p − 1) 2n−1 − 1 , we can rewrite (2) in the

9
following way:

2pn−2 (p − 1) 2n−1 − 1 + pn−1

n−2
X h      i
+ (p − 2) (p − 2)`−1 pn−2−` (p − 1) 2n−1−` − 1 + pn−1−` 2n−` 2` − 1 + 1
`=1
n−2
X h    i
+ (p − 2) (p − 2)`−1 pn−2−` (p − 1) 2n−1−` − 1 + pn−1−` 2n−` 2` − 1
`=1
+ (p − 2)(p − 2)n−1 (2n − 1) + (p − 2)n−1 2n+1 − 3

= 2pn−2 (p − 1) 2n−1 − 1 + pn−1 + (p − 2)n (2n − 1) + (p − 2)n−1 2n+1 − 3

n−2
X h      i
+ (p − 2)` 2pn−2−` (p − 1) 2n−1−` − 1 + pn−1−` 2n−`+1 2` − 1 + 1 .
`=1

The rest of the computation is then straightforward. 

For a fixed p, the result of Theorem 4.3 simplifies. For instance, in the classical
(with respect to the Tower of Hanoi) case p = 3 we get:

Corollary 4.4 Let n ≥ 2, and let jk n−1 be an almost-extreme vertex of S3n . Then
 
n−1 n−1 7 n 1 3
dS3n (jk )=3 2 − − .
5 2 10

5 Metric dimension of Spn

The metric dimension of a graph turned out to be a natural concept while studying
several different problems and was consequently also reinvented in numerous disguises.
(An impressive list of its applications can be found in [9].) It is thus clear that this
dimension presents an intrinsic graph invariant. For the first time it was independently
introduced in 1974 and 1975 by Harary and Melter [8] and Slater [23], respectively.
We refer to the recent semi-survey paper of Bailey and Cameron [1] for a great source
on historical developments, connections to other invariants, non-standard terminology,
and a long list of references. Another survey source for the dimension is [6]. Here
we just recall that the metric dimension has been studied on Cartesian products of
graphs [2, 21], distance-regular graphs [7], and circulant graphs [13]. In this section we
determine the metric dimension of Sierpiński graphs.
Let G be a graph, then R ⊆ V (G) is a resolving set if each vertex of G is uniquely
determined by the distances to the vertices of R. More precisely, let R = {u1 , . . . , uk },
then R is resolving if (d(x, u1 ), . . . , d(x, uk )) 6= (d(y, u1 ), . . . , d(y, uk )) holds for any
x, y ∈ V (G). The metric dimension of G, denoted µ(G), is the size of a minimum
resolving set.

10
Theorem 5.1 For any p ≥ 3 and any n ≥ 1,

µ(Spn ) = p − 1 .

Moreover, if R is a minimum resolving set, then |R ∩ jSpn−1 | ≤ 1 holds for any j ∈ [n].

Proof. To show that µ(Spn ) ≤ p − 1 it suffices to demonstrate that the extreme vertices
1n , 2n , . . . , (p − 1)n form a resolving set. To prove this fact, we first show that all the
extreme vertices form a resolving set.
Let s = s1 s2 . . . sn and t = t1 t2 . . . tn be vertices of Spn , s 6= t. Since there exists an
index i such that si 6= ti , Lemma 2.1 implies that d(s, sni ) 6= d(t, sni ). Consequently, all
extreme vertices form a resolving set (cf. also an argument from [18]).
To see that {1n , 2n , . . . , (p − 1)n } form a resolving set, set

D(x) = (d(x, 1n ), d(x, 2n ), . . . , d(x, (p − 1)n ))

for a vertex x ∈ V (Spn ). We need to show that D(s) 6= D(t). Assume on the contrary
that D(s) = D(t). By Lemma 2.1, for any vertex w of V (Spn ) we have
p
X
d(w, in ) = (p − 1)(2n − 1) ,
i=1

(see also [19, Proposition 2.5]). It follows that d(s, pn ) = d(t, pn ) holds as well, but
then all extreme vertices would not resolve s and t, a contradiction. We conclude that
µ(Spn ) ≤ p − 1.
Suppose now that Spn contains a resolving set R with |R| ≤ p − 2. Then there exist
subgraphs jSpn−1 and kSpn−1 , j 6= k, such that R ∩ jSpn−1 = ∅ and R ∩ kSpn−1 = ∅. By
Proposition 3.1 it follows that for each r ∈ R, d(r, j n ) = d(r, j n−1 k), so R cannot be a
resolving set for Spn .
Let now R be a minimum resolving set. Then |R| = p − 1. If for some ` we have
|R ∩ `Spn−1 | ≥ 2, then by the pigeonhole principle there exist indices j 6= k such that
R ∩ jSpn−1 = ∅ and R ∩ kSpn−1 = ∅. But then by the same argument as above, R is not
resolving. 

While finalizing this paper we found out that Theorem 5.1 was independently (and
at the same time) obtained by Aline Parreau in her Ph.D. thesis [20].

Acknowledgements
This work has been financed by ARRS Slovenia under the grant P1-0297 and within
the EUROCORES Programme EUROGIGA/GReGAS of the European Science Foun-
dation. The first author is also with the Institute of Mathematics, Physics and Me-
chanics, Ljubljana.

11
References
[1] R. F. Bailey, P. J. Cameron, Base size, metric dimension and other invariants of
groups and graphs, Bull. Lond. Math. Soc. 43 (2011) 209–242.

[2] J. Cáceres, C. Hernando, M. Mora, I. M. Pelayo, M. L. Puertas, C. Seara,

D. R. Wood, On the metric dimension of Cartesian products of graphs, SIAM
J. Discrete Math. 21 (2007) 423–441.

[3] G. Della Vecchia, C. Sanges, A recursively scalable network VLSI implementation,

Future Generation Comput. Syst. 4 (1988) 235–243.

[4] J.-S. Fu, Hamiltonian connectivity of the WK-recursive network with faulty nodes,
Inform. Sci. 178 (2008) 2573–2584.

[5] H.-Y. Fu, {Pr }-free colorings of Sierpiński-like graphs, Ars Combin. 105 (2012)
513–524.

[6] W. Goddard, O. R. Oellermann, Distance in graphs, in: Structural Analysis of

Complex Networks Birkhäuser/Springer, New York, 2011, 49–72.

[7] J. Guo, K. Wang, F. Li, Metric dimension of some distance-regular graphs, J.

Comb. Optim., in press, DOI 10.1007/s10878-012-9459-x.

[8] F. Harary, R. A. Melter, On the metric dimension of a graph, Ars Combin. 2

(1976) 191–195.

[9] M. Hauptmann, R. Schmied, C. Viehmann, Approximation complexity of metric

dimension problem, J. Discrete Algorithms 14 (2012) 214–222.

[10] A. M. Hinz, S. Klavžar, U. Milutinović, D. Parisse, C. Petr, Metric properties of

the Tower of Hanoi graphs and Stern’s diatomic sequence, European J. Combin.
26 (2005) 693–708.

[11] A. M. Hinz, D. Parisse, Coloring Hanoi and Sierpiński graphs, Discrete Math. 312
(2012) 1521–1535.

[12] A. M. Hinz, D. Parisse, The average eccentricity of Sierpiński graphs, Graphs

Combin., in press: doi: 10.1007/s00373-011-1076-4.

[13] M. Imran, A. Q. Baig, S. A. U. H. Bokhary, I. Javaid, On the metric dimension of

circulant graphs, Appl. Math. Lett. 25 (2012) 320–325.

[14] C.-H. Lin, J.-J. Liu, Y.-L. Wang, W. C.-K. Yen, The hub number of Sierpiński-like
graphs, Theory Comput. Syst. 49 (2011) 588–600.

12
[15] X. Lin, On the Wiener index and circumference, MATCH Commun. Math. Com-
put. Chem. 67 (2012) 331–336.

[16] S. Lipscomb, Fractals and Universal Spaces in Dimension Theory, Springer, Berlin,
2009.

[17] S. Klavžar, U. Milutinović, Graphs S(n, k) and a variant of the Tower of Hanoi
problem, Czechoslovak Math. J. 47(122) (1997) 95–104.

[18] S. Klavžar, B. Mohar, Crossing numbers of Sierpiński-like graphs, J. Graph Theory

50 (2005) 186–198.

[19] D. Parisse, On some metric properties of the Sierpiński graphs Skn , Ars Combin.
90 (2009) 145–160.

[20] A. Parreau, Problèmes d’identification dans les graphes, Ph.D. Thesis, Université
de Grenoble, July 2012.

[21] J. Peters-Fransen, O. R. Oellermann, The metric dimension of Cartesian products

of graphs, Util. Math. 69 (2006) 33–41.

[22] D. Romik, Shortest paths in the Tower of Hanoi graph and finite automata, SIAM
J. Discrete Math. 20 (2006) 610–622.

[23] P. Slater, Leaves of trees, Congr. Numer. 14 (1975) 549–559.

[24] H. L. Wiesenberger, Stochastische Eigenschaften von Hanoi- und Sierpiński-

Graphen, Diplomarbeit, Ludwig-Maximilians-Universität, München, 2010.

13