AKCE International Journal of Graphs and Combinatorics

ISSN: 0972-8600 (Print) 2543-3474 (Online) Journal homepage: https://www.tandfonline.com/loi/uakc20

Unsolved Algorithmic Problems on Trees

Stephen T. Hedetniemi & T.W. Haynes (Communicated by)

To cite this article: Stephen T. Hedetniemi & T.W. Haynes (Communicated by) (2006) Unsolved
Algorithmic Problems on Trees, AKCE International Journal of Graphs and Combinatorics, 3:1, 1-37

To link to this article: https://doi.org/10.1080/09728600.2006.12088803

© 2006 Taylor & Francis, LLC, trading as

Taylor & Francis Group

Published online: 10 Mar 2020.

Submit your article to this journal

Article views: 14

View related articles

Citing articles: 1 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=uakc20
AKCE J. Graphs. Combin., 3, No. 1 (2006), pp. 1-37

Unsolved Algorithmic Problems on Trees

Stephen T. Hedetniemi
Clemson University
Clemson, SC 29634
e-mail: [email protected]

Communicated by: T.W. Haynes

Received 13 July 2005; accepted 25 April 2006

Abstract
The literature on algorithms and complexity results for domination and domination-
related problems is extensive, and deals with a somewhat bewildering variety of concepts
and definitions. At the same time, what one observes is that no matter what the definition
or concept is, an algorithm problem related to that concept is almost always solvable in linear
time when the inputs are restricted to trees. The fact that this is so follows almost directly
from the now well-developed theory of algorithms on partial k -trees, or graphs of bounded
treewidth. In light of this, it is somewhat surprising that quite a few algorithmic problems on
trees remain unsolved. In this paper we offer a list of more than 60 algorithm problems that
have yet to be solved for trees, quite a few of which are newly defined here. For about 50 of
these problems the associated NP-completeness questions have not yet been settled either.

Keywords:
2000 Mathematics Subject Classification: 05C

1. Introduction

In order to understand the problems to be discussed in this paper, we will need some
definitions and terminology.
Let G = (V, E) be a graph of order n = |V | .
Let v be a vertex in V . The open neighborhood N (v) of v is the set of all vertices
adjacent to v . The closed neighborhood N [v] of v is the set N (v) ∪ {v} . For a set S
of vertices, the open neighborhood N (S) of S is defined by N (S) = ∪v∈S N (v) and the
closed neighborhood N [S] of S is the set N (S) ∪ S . Given a set S ⊆ V , by G[S] we
mean the subgraph of G induced by S , where G[S] = (S, E ∩ (S × S)) .
A set S ⊆ V is called independent if no two vertices in S are adjacent. The vertex
independence number β0 (G) equals the maximum size of an independent set in G , while
the lower independence number(also called the independent domination number) i(G) ,
equals the minimum size of a maximal independent set in G .
2 Unsolved Algorithmic Problems on Trees

A set S ⊆ V is a dominating set if every vertex in V − S is adjacent to a vertex in S ,

that is, N [S] = V . The domination number γ(G) is the minimum size of a dominating
set S in G ; such a set is called a γ -set of G . The upper domination number Γ(G)
equals the maximum size of a minimal dominating set in G .
A set S is a packing if for every vertex v ∈ V , |N [v] ∩ S| ≤ 1 . The packing number
P (G) equals the maximum size of a packing in G , while the it lower packing number
p(G) equals the minimum size of a maximal packing in G . The following well-known
inequality chain relates these invariants:
p(G) ≤ P (G) ≤ γ(G) ≤ i(G) ≤ β0 (G) ≤ Γ(G).

Given a set S ⊆ V and a vertex u ∈ S , an external private neighbor of u (with respect

to S ) is a vertex v ∈ V − S which is adjacent to u but no other vertex in S . A vertex
u ∈ S is its own private neighbor if it is not adjacent to any vertex in S . Finally, an
internal private neighbor of u (with respect to S ) is a vertex v ∈ S that is adjacent to
u but is not adjacent to any other vertex in S .
A set S ⊆ V is called irredundant if every vertex in S is either its own private
neighbor or has an external private neighbor, with respect to S . This can also be defined
as follows. Define the private neighbor set of a vertex v ∈ S , to equal the set pn[v, S] =
N [v] − N [S − {v}] . A set S is irredundant if for every vertex v ∈ S , pn[v, S] 6= ∅ . The
irredundance number ir(G) equals the minimum size of a maximal irredundant set in G ,
while the upper irredundance number IR(G) equals the maximum size of an irredundant
set in G .
The concepts of independence, domination and irredundance are related in a natural
way, as first observed by Cockayne, Hedetniemi and Miller in 1978 [25]:
Theorem 1. For any graph G ,
ir(G) ≤ γ(G) ≤ i(G) ≤ β0 (G) ≤ Γ(G) ≤ IR(G).

This inequality chain has become the focus of a substantial amount of research in
domination theory, and will be a reference for many of the open problems to follow.
Nearly all of the algorithm problems in the following sections start by considering the
class of all sets of vertices (or edges) having some property P in a graph G , each set in
which is called a P -set. Particular focus is on the class of all minimal or maximal P -sets,
and one attempts to determine the minimum and maximum sizes of a minimal P -set, or
the minimum and maximum sizes of a maximal P -set.
In this paper we will occasionally depart from traditional notation. In an effort to
provide a more uniform system of notation, we will use subscripts or lower case symbols
to refer to minimum size parameters (e.g. β1 (G) or γt (G) ) and superscripts or upper
case symbols to refer to maximum size parameters (e.g. β 1 (G) or Γt (G) ).
A property P is called hereditary if whenever a set S has property P , so does every
subset of S . For example, the property of being an independent set is hereditary. For
 Stephen T. Hedetniemi 3

hereditary properties one considers the class of all maximal P -sets, and one seeks first
an algorithm to determine the maximum size of a P -set, and second, an algorithm to
determine the minimum size of a maximal P -set; we refer to this second problem as the
mini-maximal problem. One reason for having an interest in hereditary properties is that
a hereditary P -set is maximal if and only if it is 1 -maximal, meaning that in order to
decide if a set S is maximal, all one need do is check every set S ∪ x , for every x not in
S [21].
Similarly, a property P is called super-hereditary if whenever a set S has property P ,
so does every superset of S . For example, the property of being a vertex cover is super-
hereditary. For super-hereditary properties P , one considers the class of all minimal
P -sets. One then seeks to determine the minimum size of a P -set and the maximum
size of a minimal P -set. We refer to this second problem as the maxi-minimal problem.
Similar to hereditary properties, the property of being a minimal super-hereditary P -set
can be determined by considering every set S − {x} , for every element x ∈ S , that is, a
super-hereditary set S is minimal if and only if for every x ∈ S , S − {x} is not a P -set
[21].
Thus, most of the algorithmic problems that follow come in pairs, either those of finding
the maximum and mini-maximal sizes, or of finding the minimum and maxi-minimal sizes
of a P -set. In many instances, only the problems of finding either a minimum or a
maximum size P -set have been considered, while the maxi-minimal or mini-maximal
problems have not been considered. In these cases, we will indicate that the problem
invariant is newly mentioned here by saying (NEW HERE). In these cases it would be
interesting to undertake a graph theoretical study of the properties and values of these
new invariants and their relationships with other related invariants.
The following sections contain a listing of more than 60 algorithm problems for trees
that do not appear to have been solved, and an almost equally long listing of associated
NP-completeness questions that do not appear to have been settled. For sure, many of
these problems are easy to solve and are simple exercises; the fact that they have not been
solved is simply because they may have been overlooked. But at the same time, many of
these problems are truly difficult, and some may never be solved. Hopefully the reader
will find at least some of these problems to be fascinating to ponder.
If the reader knows that any of these problems have been solved, please send an e-mail
with this information to: [email protected].
The interested reader is encouraged to refer to two chapters in [64]: the first by Kratsch
[78] contains a survey of algorithmic results on domination and related invariants, while
the second, by Hedetniemi, McRae and Parks [74] contains a survey of NP-completeness
results for many domination and domination-related invariants.
In addition, anyone interested in developing algorithms for solving these types of prob-
lems is encouraged to understand the different approaches that can be used to construct
polynomial time algorithms for problems that are restricted to partial k -trees for some
fixed k , or graphs of bounded treewidth [2, 3, 4, 8, 96].
4 Unsolved Algorithmic Problems on Trees

2. Irredundance Numbers

We have previously defined the irredundance number ir(G) and the upper irredundance
number IR(G) . There are three other types of irredundance that have received attention
in the literature.
We say that a set S ⊂ V is:

1. open irredundant if every vertex in S has an external private neighbor. The open ir-
redundance number oir(G) equals the minimum size of a maximal open irredundant
set in G , and the upper open irredundance number OIR(G) equals the maximum
size of an open irredundant set in G .

2. open-open irredundant if every vertex in S has either an internal or an external pri-

vate neighbor. The open-open irredundance number ooir(G) equals the minimum
size of a maximal open-open irredundant set in G , while the upper open-open irre-
dundance number OOIR(G) equals the maximum size of an open-open irredundant
set in G .

3. closed-open irredundant if every vertex in S has a private neighbor of some kind, ei-
ther itself, an internal or an external private neighbor. The closed-open irredundance
number coir(G) equals the minimum size of a maximal closed-open irredundant set
in G , while the upper closed-open irredundance number COIR(G) equals the max-
imum size of a closed-open irredundant set in G .

First introduced by Cockayne, Hedetniemi and Miller in 1979 [25], there are now more
than 100 papers on irredundance in graphs; the reader is referred to a discussion of
this literature in [9, 15, 18, 20, 21, 43, 63, 73]. The complexity of various irredundance
numbers is discussed in the paper by Fellows, Fricke, Hedetniemi and Jacobs [42], who
first demonstrated the NP-completeness of the decision problems associated with IR(G)
and COIR(G) , and by Golumbic and Laskar [58]. The decision problem associated
with OIR(G) was first settled by Even, Goldreich, Moran and Tong [38], while that for
OOIR(G) was settled by Cameron [11].
The complexity of the (lower) irredundance number ir(G) was first settled by Pfaff,
Laskar and S.T. Hedetniemi [89] who showed the NP-completeness for bipartite graphs,
and by Laskar, Pfaff, Hedetniemi and Hedetniemi [79] who demonstrated the NP-
completeness for chordal graphs. The NP-completeness of the decision problems associated
with oir(G) , ooir(G) and coir(G) for bipartite graphs, chordal graphs, line graphs and
line graphs of bipartite graphs was shown by McRae [85].
The values of the lower irredundance numbers of trees have proved to be very difficult to
compute. An impressive linear algorithm for computing the irredundance number ir(T )
of any tree T was constructed in 1987 by Bern, Lawler, and Wong [6]. This is a dynamic
programming algorithm which consists of a 20 by 20 matrix of possible combinations of
 Stephen T. Hedetniemi 5

subtrees that can combine to form an optimal solution. Although the authors provided
little or no proof of the correctness of this algorithm, there is little doubt that it is correct.
It seems reasonable, therefore, that a similar complexity will be required in order to
compute the value of the other three lower irredundance numbers.
Algorithm 1. Design an algorithm for computing the value of oir(T ) for any tree T .
Algorithm 2. Design an algorithm for computing the value of ooir(T ) for any tree T .
Algorithm 3. Design an algorithm for computing the value of coir(T ) for any tree T .

In 1990 [77], Jacobson and Peters showed that for any tree T , β0 (T ) = Γ(T ) = IR(T ) .
And since a simple greedy algorithm, first designed in 1966 by Daykin and Ng [29], will
suffice to compute the value of β0 (T ) , it is easy to compute the value of IR(T ) for any
tree T .
It is also well known that for any tree T , OIR(T ) = β 1 (T ) , for example, as noted by
Golumbic and Laskar [58], where β 1 (T ) denotes the well known matching number of a
tree T (cf. Section ). And since a simple greedy algorithm will suffice to compute the
value of β 1 (T ) [88], it is easy to compute the value of IR(T ) for any tree T .
In the same paper [58], the authors point out that Fricke, S.M. Hedetniemi and Laskar
[47] had shown that for any tree T , β ∗ (T ) = OOIR(T )/2 , where β ∗ (T ) is the induced
matching number, to be defined later in Section . They then show that the same equality
holds for all bipartite graphs. Since it was shown by Cameron [11] that the induced
matching number β ∗ (G) can be computed in polynomial time for all chordal graphs, it
follows that the value of OOIR(T ) can be computed in polynomial time for all trees.
Similarly, in [58] it is shown that for all bipartite and circular arc graphs, COIR(G) =
β [1] (G) , where β [1] (T ) is the 1 -dependence number, to be defined later in Section . Since
it is easy to see that the [1] -dependence number of a tree can be computed in linear time,
using a simple, Wimer-style algorithm, [96] and [30], it follows that COIR(T ) can be
computed in linear time for any tree T .

2.1 Independent Open Irredundance

In 1999, Cockayne [12] introduced the study of a large class of generalized irredundant
sets in graphs. Each type of a generalized irredundant set S ⊂ V is defined by the types of
private neighbors (i.e. self, internal or external) that each vertex in the set must have. For
example, if every vertex in S is its own private neighbor then the set S is an independent
set, and if every vertex in S either is its own private neighbor or has an external private
neighbor, then S is an irredundant set. Among the many possible Boolean functions on
three variables that describe a possible type of generalized irredundance, one of the more
interesting is the following.
A set S is called an independent open irredundant set (or ioir-set if S is an independent
set and every vertex in S has an external private neighbor. In [12], Cockayne identifies
6 Unsolved Algorithmic Problems on Trees

12 types of generalized irredundant sets the properties of which are hereditary. Perhaps
the most interesting of these not previously mentioned are the ioir-sets. One can therefore
define ioir(G) to equal the minimum size of a maximal ioir-set and IOIR(G) to equal the
maximum size of an ioir-set.
These generalized irredundant sets are also studied by Finbow in [43] and Cockayne
and Finbow in [18].

NP-problem 1. Does G have a maximal ioir-set of size at most k ?

Algorithm 4. Design an algorithm to compute the value of ioir(T ) for any tree T .

NP-problem 2. Does G have an ioir-set of size at least k ?

Algorithm 5. Design an algorithm to compute the value of IOIR(T ) for any tree T .

2.2 Total Irredundance

In 1993, S.M. Hedetniemi, S.T. Hedetniemi and Jacobs [70] generalized the concept
of irredundance in a very natural way. A set S ⊂ V is said to be a total irredundant
set if every vertex v ∈ V has a private neighbor with respect to S . This means that
every vertex in S either is its own private neighbor, or has an external private neighbor
with respect to S , and every vertex in V − S either is not adjacent to any vertex in
S , or is adjacent to at least one vertex in V − S which is not adjacent to any vertex
in S . This definition gives rise to the two invariants irt (G) and IRt (G) , which equal
the minimum and maximum sizes of a maximal total irredundant set in G . In [70] it is
shown that the value of IRt (T ) can be computed in linear time for any tree T , while
the decision problem corresponding to IRt (G) is NP-complete. No results, algorithm or
NP-completeness, were given for irt (G) .
Two other papers on total irredundance have appeared since then, [62] and [39].

NP-problem 3. Does G have a maximal total irredundant set of size at most k ?

Algorithm 6. Design an algorithm to compute the value of irt (T ) for any tree T .

Although it has not been done in the previous papers on total irredundance, one can
easily define the other types of total irredundance: open total irredundance ( oirt (G) and
OIRt (G) ), open-open total irredundance ( ooirt (G) and OOITt (G) ) and closed-open total
irredundance ( coirt (G) and COIRt (G) )(NEW HERE).

NP-problem 4. Does G have a maximal total open irredundant set of size at most k ?

Algorithm 7. Design an algorithm to compute the value of oirt (T ) for any tree T .

NP-problem 5. Does G have a maximal total open-open irredundant set of size at most
k?
 Stephen T. Hedetniemi 7

Algorithm 8. Design an algorithm to compute the value of ooirt (T ) for any tree T .
NP-problem 6. Does G have a maximal total closed-open irredundant set of size at most
k?
Algorithm 9. Design an algorithm to compute the value of coirt (T ) for any tree T .
NP-problem 7. Does G have a total open irredundant set of size at least k ?
Algorithm 10. Design an algorithm to compute the value of OIRt (T ) for any tree T .
NP-problem 8. Does G have a total open-open irredundant set of size at least k ?
Algorithm 11. Design an algorithm to compute the value of OOIRt (T ) for any tree T .
NP-problem 9. Does G have a total closed-open irredundant set of size at least k ?
Algorithm 12. Design an algorithm to compute the value of COIRt (T ) for any tree T .

2.3 External Redundance

In 1994, McRae [84] introduced the concept of external redundance, which was designed
to extend the irredundance, domination, independence inequality chain, and was based on
an analysis of the conditions under which an irredundant set is a maximal irredundant set.
This was further discussed by Cockayne, Hattingh, Hedetniemi, Hedetniemi and McRae
in 1997 [21].
We say that a set S ⊂ V is external redundant if for every vertex v ∈ V −S , there exists
a vertex w ∈ S ∪ {v} , such that pn[w, S ∪ {v}] = ∅ , and if w ∈ S then pn[w, S] 6= ∅ .
Stated in other words, an external redundant set has the property that any attempt to
add a vertex v ∈ V − S to S results in the creation of a new vertex which does not have
a private neighbor; either the vertex v will not have a private neighbor in the set S ∪ {v} ,
or a vertex in S which has a private neighbor with respect to S no longer has a private
neighbor with respect to the larger set S ∪ {v} .
An equivalent definition might be helpful. Given a set S ⊆ V , we define the private
neighbor count of S to equal the value pnc(S) = |{v ∈ S : pn[v, S] 6= ∅}| . Thus,
the private neighbor count of a set S equals the number of vertices in S which have
private neighbors. A set S is called external redundant if for every vertex v ∈ V − S ,
pnc(S ∪ {v}) ≤ pnc(S) . Thus, you cannot increase the private neighbor count of S by
adding a vertex to S .
The external redundance number er(G) equals the minimum size of a maximal external
redundant set in G , while the upper external redundance number ER(G) equals the
maximum size of an external redundant set in G . In [21] no discussion is given of the
complexity of the decision problem associated with either er(G) or ER(G) , and no
algorithms are mentioned.
It is pointed out however that:
8 Unsolved Algorithmic Problems on Trees

Lemma 1. If S ⊆ V is a maximal irredundant set, then S is a minimal external redun-

dant set.

Thus, one can include the external redundance numbers in the previous inequality chain,
as follows:

Theorem 2. For any graph G ,

er(G) ≤ ir(G) ≤ γ(G) ≤ i(G) ≤ β0 (G) ≤ Γ(G) ≤ IR(G) ≤ ER(G).

NP-problem 10. Does G have an external redundant set of size at most k ?

Algorithm 13. Design an algorithm to compute the value of er(T ) for any tree T .

NP-problem 11. Does G have a minimal external redundant set of size at least k ?

Algorithm 14. Design an algorithm to compute the value of ER(T ) for any tree T .

2.4 Perfect Neighborhood Sets

Let G = (V, E) be a graph and let S ⊂ V be a an arbitrary set of vertices. A vertex

v ∈ V is said to be perfect with respect to S if |N [v] ∩ S| = 1 . Thus, if a vertex v ∈ V is
perfect then it is either an element of S that is an isolate in the subgraph G[S] induced
by S , or is not in S but is adjacent to exactly one vertex in S . A vertex is said to
be ap if it is not perfect but is adjacent to a perfect vertex (with respect to S ). A set
S ⊂ V is called a perfect neighborhood set if every vertex v ∈ V is either perfect or ap
with respect to S . The perfect neighborhood number θ(G) equals the minimum size of a
perfect neighborhood set in G . The upper perfect neighborhood number Θ(G) equals the
maximum size of a perfect neighborhood set in G .
Perfect neighborhood sets were introduced by Hedetniemi, Hedetniemi and Henning in
1997 [68], and by Fricke, Haynes, Hedetniemi, Hedetniemi and Henning in 1999 [46]. The
problem of deciding if a graph G has a perfect neighborhood set of size at most k was
shown in [68] to be NP-complete for bipartite and chordal graphs. Also in [68] a linear
algorithm is constructed for computing the value of θ(T ) for any tree T .
The upper perfect neighborhood number was defined and studied in [46], where it was
shown, somewhat surprisingly, that for all graphs G , Θ(G) = Γ(G) . Thus, the value of
Θ(T ) can be computed for all trees T , since for all trees T , β0 (T ) = Γ(T ) .
A similar definition can be given for open perfect neighborhood sets, in which we say
that a vertex v ∈ V is called open perfect if |N (v) ∩ S| = 1 . A set S ⊂ V is an open
perfect neighborhood set if every vertex v ∈ V is either open perfect or is adjacent to an
open perfect vertex. The open perfect neighborhood number θo (G) equals the minimum
size of an open perfect set in G , and the upper open perfect neighborhood number Θo (G)
equals the maximum size of an open perfect neighborhood set.
 Stephen T. Hedetniemi 9

The problem of deciding if a graph has an open perfect neighborhood set of size at
most k was shown to be NP-complete for bipartite and chordal graphs by Hedetniemi,
Jacobs, Laskar and Pillone in 1996 [72], however no algorithms for trees were constructed
for computing the values of θo (G) or Θo (G) .

Algorithm 15. Design an algorithm for computing the value of θo (T ) for any tree T .

NP-problem 12. Does G have an open perfect neighborhood set of size at least k ?

Algorithm 16. Design an algorithm for computing the value of Θo (T ) for any tree T .

2.5 Annihilation Number

In a series of papers, Cockayne, Favaron, Mynhardt and Puech [26, 16, 17, 24] introduced
several new irredundance-related invariants and established an interesting sequence of
inequalities among them, especially for trees. But noticeably no NP-completeness results
are given nor are algorithms for trees constructed.
The concept of a private neighborhood pn[u, S] of a vertex u in a set S enables one
to define a partition V (G) = {I, A, EI, EA, M, R} , where:

• I = {u ∈ S|u is an Isolate in G[S]},

• A = S − I , the vertices in S that are Adjacent to another vertex in S ,

• EI = {v ∈ V − S|v is an External private neighbor of some vertex in I},

• EA = {v ∈ V − S|v is an External private neighbor of some vertex in A},

• M = {v ∈ V −S||N (v)∩S| ≥ 2} , the vertices in V −S that are Multiply dominated

by S ,

• R = V − N [S] , the Remaining vertices.

With these sets defined, we can say that:

• S is a dominating set if N [S] = V ,

• S is irredundant if every vertex in A has a private neighbor in EA ,

• S is a packing if M ∪ A = ∅ ,

• S is a perfect neighborhood set (PN-set) if the set I ∪ EI ∪ EA is a dominating

set.
10 Unsolved Algorithmic Problems on Trees

A vertex v ∈ V − S is said to annihilate a vertex u ∈ S if ∅ =

6 pn[u, S] ⊆ N [v] . That
is, vertex u ∈ S has a private neighbor with respect to S , but does not have a private
neighbor with respect to the set S ∪ {v} . Let Ann(S) = {v ∈ V − S|v annihilates some
vertex u ∈ S} . We say that a set S is U -annihilated by every subset U ⊆ Ann(S) . In
[20] Cockayne, Grobler, Hedetniemi and McRae show:

Theorem 3. A set S is maximal irredundant if and only if S is irredundant and N[R]-

annihilated.

One can observe that the class of N [R] -annihilated sets is the same as the class of
external redundant sets and is contained in the class of R -annihilated sets (called it Ra -
sets). Sets that are both R -annihilated and irredundant, hereafter called Rai -sets, were
first introduced by Favaron and Puech [41], who called them semi-maximal irredundant
sets.
Define ra(G) to equal the smallest size of an Ra-set (R-annihilated set) in G , and
rai(G) to equal the smallest size of an Rai-set (R-annihilated and irredundant set) in G .
The following impressive inequality chain was proved in [17], where (i) γ2 (G) equals
the distance- 2 domination number of G , that is the size of a smallest set S having the
property that every vertex in V is distance at most 2 from some vertex in S , (ii) θi (G)
equals the smallest cardinality of an independent perfect neighborhood set in G , and (iii)
p(G) and P (G) are the two packing numbers.

Theorem 4. For any tree T ,

rai(T )
γ2 (T ) ≤ θ(T ) ≤ θi (T ) ≤ p(T ) ≤ ra(T ) ≤ ≤ ir(T ) ≤ P (T ) = γ(T ).
er(T )

The complexity and algorithm questions for these invariants have not been addressed.

NP-problem 13. Does G have an independent perfect neighborhood set of size at most
k?

Algorithm 17. Design an algorithm for computing the value of θi (T ) for any tree T .

NP-problem 14. Does G have an Ra -set set of size at most k ?

Algorithm 18. Design an algorithm for computing the value of ra(T ) for any tree T .

NP-problem 15. Does G have an Rai -set set of size at most k ?

Algorithm 19. Design an algorithm for computing the value of rai(T ) for any tree T .
 Stephen T. Hedetniemi 11

3. k -dependence, k -dependent domination and k -dependent irredundance

A set S ⊂ V is called a k -dependent set if the maximum degree of a vertex in G[S] is at

most k . The k -dependence number β[k] (G) equals the maximum size of a k -dependent
set in G , while the lower k -dependence number i[k] (G) equals the minimum size of a
maximal k -dependent set in G . The concept of k -dependent sets was first introduced
by Fink and Jacobson in 1984 [44].
In 1989, Hare, Hedetniemi and Hedetniemi [60] showed that the problem of deciding
if a graph G has a 1 -dependent set of size at least k is NP-complete for bipartite and
chordal graphs, or any class of graphs, for which DOMINATING SET is NP-complete,
that is closed under the corona GoK1 operation. They also constructed an algorithm for
computing the value of β[1] (T ) for any tree T and the value of i[1] (T ) for any tree T .
NP-problem 16. Does G have a maximal 1 -dependent set of size at most k ?

The concept of k -dependence extends quite naturally to both domination and irre-
dundance. In 2002, Favaron, Hedetniemi, Hedetniemi and Rall introduced the concept
of k -dependent domination [40]. A k -dependent dominating set is a set which is both
k -dependent and dominating. The k -dependent domination number γ[k] (G) equals the
minimum size of a k -dependent dominating set and the upper k -dependent domination
number Γ[k] (G) equals the maximum size of a minimal k -dependent dominating set in
G.
In the same paper [40], the authors introduce the concept of k -dependent irredundance.
The k -dependent irredundance number ir[k] (G) equals the minimum order of a maximal
k -dependent irredundant set in G , while the upper k -dependent irredundance number
IR[k] (G) equals the maximum size of a k -dependent irredundant set. The natural rela-
tionship between k -dependent dominating sets and k -dependent irredundant sets follows
from the following lemma in [40]:
Proposition 1. Every minimal k -dependent dominating set in a graph G is also a
maximal k -dependent irredundant set in G .
Corollary 1. For any graph G ,

ir[k] (G) ≤ γ[k] (G) ≤ Γ[k] (G) ≤ IR[k] (G).

In [40], the authors establish the NP-completeness of the decision problems associated
with each of the invariants ir[k] (G) , γ[k] (G) , Γ[k] (G) and IR[k] (G) . But no algorithms
for computing any of these values were presented. It is a simple matter to construct an
algorithm for computing the value of ir[1] (T ) and γ[1] (T ) for any tree T , given algorithms
for computing ir(T ) and γ(T ) , which already exist. However, it is more of a challenge
to compute the upper values.
Algorithm 20. Design an algorithm for computing the value of Γ[1] (T ) for any tree T .
12 Unsolved Algorithmic Problems on Trees

Algorithm 21. Design an algorithm for computing the value of IR[1] (T ) for any tree T .

It is worth noting that in [80], Lawler and Slater present a linear algorithm for finding
the minimum number of K1 and K2 components which dominate the vertices of a tree.

4. Perfect and Nearly Perfect Sets

A number of graph theoretical invariants are defined in terms of a set S ⊂ V of vertices

and conditions under which a given vertex satisfies the perfect equality |N [v] ∩ S| = 1 , or
the nearly perfect inequality |N [v] ∩ S| ≤ 1 .
The concept of nearly perfect sets was first defined in a paper by Cockayne, Hartnell,
Hedetniemi and Laskar in 1993 [22], but was first studied by Dunbar, Harris, Hedetniemi,
Hedetniemi, McRae and Laskar in 1995 [33].
A set S ⊂ V is called nearly perfect if for every vertex v ∈ V − S , |N (v) ∩ S| ≤ 1.
A nearly perfect set S is 1 -minimal if for every vertex u ∈ S , the set S − {u} is not a
nearly perfect set, and is 1 -maximal if for every vertex u ∈ V − S , the set S ∪ {u} is not
a nearly perfect set. The 1 -maximal nearly perfect number of a graph G , np (G) , equals
the minimum size of a 1 -maximal nearly perfect set in G . The 1 -minimal nearly perfect
number Np (G) equals the maximum size of a 1 -minimal nearly perfect set in G .
In [33] a linear algorithm is presented for computing the value of np (T ) for any tree T .
It is also shown that the corresponding decision problem is NP-complete when restricted
to either bipartite or chordal graphs. Somewhat surprisingly, it is shown in [33] that the
value of Np (G) can be computed in polynomial time for any graph G .
However, the related concept of a total nearly perfect set was not discussed. As defined
in [22], a set S is a total nearly perfect set if for every vertex v ∈ V , |N (u) ∩ S| ≤ 1 . In
the same way that np (G) and Np (G) are defined, one can define ntp (G) and Ntp (G) .

NP-problem 17. Does G have a 1 -maximal nearly total perfect set of size at most k ?

Algorithm 22. Design an algorithm to compute the value of ntp (T ) for any tree T .

NP-problem 18. Does G have a 1 -minimal nearly total perfect set of size at least k ?

Algorithm 23. Design an algorithm to compute the value of Ntp (T ) for any tree T .

5. Types of Domination

The two books on domination in graphs by Haynes, Hedetniemi and Slater [63] and
[64] present many different types of dominating and related sets. Since the publication of
these two books, perhaps another 20-30 different types of dominating sets have appeared
in the literature. Not meaning to be totally comprehensive, we present in the next several
subsections a variety of open domination problems for trees.
 Stephen T. Hedetniemi 13

The most studied types of domination are (i) dominating sets, (ii) independent domi-
nating sets and (iii) total dominating sets.
The first algorithm for computing the value of γ(T ) for any tree T was presented
in 1975 by Cockayne, Goodman and Hedetniemi [19], while the NP-completeness of the
associated dominating set problem was first given by Garey and Johnson in [49] (see also
[74]).
The first algorithm for computing the value of β0 (T ) for any tree T was presented in
1966 by Daykin and Ng [29], while the first algorithm for computing the value of i(T )
was given by Beyer, Proskurowski, Hedetniemi and Mitchell in 1977 [7].
It was pointed out by Cockayne, Favaron, Payan and Thomason in 1981 [15], that
for any bipartite graph, β0 (G) = Γ(G) , and thus the β0 -algorithm for trees suffices to
compute the upper domination number of a tree.
A set S ⊂ V is called a total dominating set if for every vertex v ∈ V , N [v] ∩ S 6= ∅ .
Thus, every vertex in V −S is adjacent to at least one vertex in S and every vertex in S is
adjacent to at least one vertex in S other than itself. The total domination number γt (G)
equals the minimum size of a total dominating set in G , while the upper total domination
number Γt (G) equals the maximum size of a minimal total dominating set in G . First
introduced by Cockayne, Dawes and Hedetniemi in 1980 [13], the algorithmic complexity
of total domination was first considered by Laskar, Pfaff, Hedetniemi and Hedetniemi in
1984 [79]. They showed that the decision problem associated with the total domination
number γt (G) is NP-complete for undirected path graphs, and constructed the first linear
algorithm for computing the value of γt (T ) for any tree T . A discussion of the NP-
completeness results associated with γt (G) can be found in [63]. The NP-completeness
of the decision problem associated with the upper total domination number was settled
by Jacobson and Peters in 1990 [77] for bipartite graphs, and later by McRae [85] for line
graphs and line graphs of bipartite graphs. The first algorithm for computing the upper
total domination number of a tree was given by Fricke, Hare, Jacobs and Majumdar in
1990 [45].

5.1 Irredundant Domination

In 1990, Hedetniemi, Hedetniemi and Jacobs [69] introduced the concept of private
domination, as follows. A set S is called a private dominating set if S is a dominating
set and every vertex in S has an external private neighbor. The private domination
number equals the maximum size of a (minimal) private dominating set, and is denoted
Γp (G) .
It is well-known that every graph G has a γ -set that is also a private dominating set
(cf. Bollobas and Cockayne [9]). Thus, the minimum size of a private dominating set
always equals the domination number.
In [69] the authors show that the following problem is NP-complete.
14 Unsolved Algorithmic Problems on Trees

PRIVATE DOMINATING SET

INSTANCE: A graph G and a positive integer k .
QUESTION: Does G have a minimal private dominating set of size at least k ?
The authors also construct a linear algorithm for computing the value of Γp (T ) for any
tree T .
The concept of private domination, however, can be placed in the broader context of
irredundant domination, as follows.
As defined, a private dominating set is a dominating set which is also open irredundant.
Since there are three other kinds of irredundance, it follows that there are three other
kinds of irredundant domination.
First, notice that any minimal dominating set is automatically an irredundant set.
Thus, if we define γir (G) to equal the minimum size of an irredundant dominating set,
then it would follow that γ(G) = γir (G) . Similarly, if Γir (G) equals the maximum size
of a minimal irredundant dominating set, then Γ(G) = Γir (G) .
In the interests of uniformity of notation, we could say that Γp (G) , the private domi-
nation number, is also the open irredundant domination number Γoir (G) , and the lower
open irredundant domination number, γoir (G) , equals the minimum size of an open irre-
dundant dominating set. But by the theorem of Bollobas and Cockayne [9], it follows that
γoir (G) = γ(G) .
Next, we can define the open-open irredundant domination number Γooir (G) (NEW
HERE) to equal the maximum size of a minimal open-open irredundant dominating set,
and γooir (G) to equal the minimum size of an open-open irredundant dominating set.
Again, since every open irredundant set is also an open-open irredundant set, it follows
that
γ(G) = γoir (G) = γooir (G).

Finally, we can define the closed-open irredundant domination number, Γcoir (G) (NEW
HERE), to equal the maximum size of a minimal closed-open irredundant dominating
set, and the lower closed-open irredundant domination number, γcoir (G) , to equal the
minimum size of a closed-open irredundant dominating set. By much the same reasoning,
however, we can conclude that since every γ -set is automatically a closed-open irredundant
dominating set, γ(G) = γcoir (G) .

NP-problem 19. Does G have a minimal open-open irredundant dominating set of size
at least k ?

Algorithm 24. Design an algorithm for computing the value of Γooir (T ) , for any tree
T.

NP-problem 20. Does G have a minimal closed-open irredundant dominating set of size
at least k ?

Algorithm 25. Design an algorithm for computing the value of Γcoir (T ) for any tree T .
 Stephen T. Hedetniemi 15

5.2 Paired Domination

A dominating set S ⊂ V is said to be a paired dominating set if the subgraph G[S]

induced by S has a perfect matching. The paired domination number γpr (G) equals
the minimum size of a paired dominating set, while the upper paired domination number
Γpr (G) (NEW HERE) equals the maximum size of a minimal paired dominating set.
Paired domination was introduced by Haynes and Slater [66], and studied further in
Haynes, Henning and Slater [65] and Profitt, Haynes, and Slater [90], who settled the
corresponding NP-completeness question [66]. Goddard [51] has recently constructed a
simple algorithm for computing γpr (T ) for any tree T .
However, none of these papers discusses the upper paired domination number Γpr (G)
defined here.
NP-problem 21. Does G have a minimal paired dominating set of size at least k ?
Algorithm 26. Design an algorithm for computing the value of Γpr (T ) for any tree T .

5.3 Efficient and Perfect Domination

A dominating set S is said to be efficient if for every vertex v ∈ V , |N [v] ∩ S| = 1.

Two basic facts about efficient dominating sets (see [63]) are well known in domination
theory: (i) not every graph has an efficient dominating set, for example, the five-cycle does
not have one, and (ii) if a graph G has an efficient dominating set, then every efficient
dominating set in G has the same size, which equals the domination number γ(G) .
However, every graph does have a perfect dominating set. A dominating set S is said
to be perfect if for every vertex v ∈ V − S , |N [v] ∩ S| = 1. Perfect dominating sets were
introduced by Weichsel [95] and studied by Yen and Lee [97] and [98] and by Cockayne,
Hartnell, Hedetniemi and Laskar in 1993 [22].
In [97], Yen and Lee show that the problem of deciding if a graph G has a perfect dom-
inating set of size at most k is NP-complete for bipartite and chordal graphs. They also
construct a linear algorithm for computing the minimum weight of a perfect dominating
set in an arbitrary, weighted tree T .
Define the perfect domination number, γp (G) , to equal the minimum size of a perfect
dominating set in G , and Γp (G) (NEW HERE) to equal the maximum size of a minimal
perfect dominating set in G .
NP-problem 22. Does G have a minimal perfect dominating set of size at least k ?
Algorithm 27. Design an algorithm for computing the value of Γp (T ) for any tree T .
16 Unsolved Algorithmic Problems on Trees

5.4 Restrained Domination

In 1994, Telle [94] first defined the concept of a restrained dominating set. This was
later studied by Domke, Hattingh, Hedetniemi, Laskar and Markus in 1994 [30]. A set
S ⊂ V is a restrained dominating set if every vertex in V − S is adjacent to a vertex in S
and a vertex in V − S other than itself. The restrained domination number γr (G) equals
the minimum size of a restrained dominating set in G , while the upper restrained dom-
ination number Γr (G) (NEW HERE) equals the maximum size of a minimal restrained
dominating set in G . In [30] the authors showed that the decision problem associated
with the restrained domination number γr (G) is NP-complete for bipartite and chordal
graphs. They also proposed a linear algorithm for computing the value of γr (T ) for any
tree T . However, the algorithm given in [30] is incorrect. A refined and corrected version
of this algorithm has recently been constructed by Goddard [51].

NP-problem 23. Does G have a minimal restrained dominating set of size at least k ?

Algorithm 28. Design an algorithm for computing the value of Γr (T ) for any tree T .

5.5 Signed Domination

In 1995, Dunbar, Hedetniemi, Henning and Slater [35] introduced the concept of signed
domination in graphs. A signed dominating function is a function of the form f : V →
{−1, 1} such that for all vertices v ∈ V , f [v] ≥ 1 . By f [v] we mean the sum of the values
P all vertices u ∈ N [v] . The weight of a signed dominating function is simply
of f (u) over
the sum v∈V f (v) . The signed domination number γ−11 (G) equals the minimum weight
of a signed dominating function in G . The upper signed domination number Γ−11 (G)
equals the maximum weight of a minimal signed dominating function in G .
The algorithmic complexity of signed domination in graphs is discussed by Hattingh,
Henning and Slater in 1995 [61]. They showed that the problem of deciding if a graph
G has a signed dominating function of weight at most k is NP-complete for bipartite
and chordal graphs, and the similar problem of deciding if a graph has a minimal signed
dominating function of weight at least k is NP-complete for bipartite and chordal graphs.
They also constructed a linear algorithm for computing the value of γ−11 (T ) for any tree
T.

Algorithm 29. Design an algorithm for computing the value of Γ−11 (T ) for any tree T .

5.6 Minus Domination

In 1996, Dunbar, Hedetniemi, Henning and McRae [34] introduced the concept of mi-
nus domination in graphs (this work was delayed in publication until 1999). A minus
dominating function is a function of the form f : V → {−1, 0, 1} such that for all
vertices v ∈ V , f [v] ≥ 1 . The weight of a minus dominating function is simply the
 Stephen T. Hedetniemi 17

P
sum v∈V f (v) . The minus domination number γ−101 (G) equals the minimum weight
of a minus dominating function in G . The upper minus domination number Γ−101 (G)
equals the maximum weight of a minimal minus dominating function in G .
In [32], Dunbar, Goddard, Hedetniemi, McRae and Henning showed that decision prob-
lems associated with γ−101 (G) and Γ−101 (G) are NP-complete for bipartite and chordal
graphs. They also presented a linear algorithm for computing the value of γ−101 (T ) for
any tree T . However, they did not present an algorithm for the upper minus domination
number.

Algorithm 30. Design an algorithm for computing the value of Γ−101 (T ) for any tree
T.

5.7 Secure Domination and Secure Total Domination

In 2003, Cockayne, Favaron and Mynhardt [14] defined a new type of domination, as
follows. A set S ⊂ V is a secure dominating set if for every vertex v ∈ V − S , there
exists a vertex u ∈ N (v) ∩ S such that S − {u} ∪ {v} is a dominating set. A secure
dominating set has the property that for every vertex v ∈ V − S , there is a neighbor of
v in S , say u , such that the new set obtained by deleting u from S and adding v is
another dominating set. This change of sets corresponds to moving a guard stationed at
vertex u ∈ S to the vertex v ∈ V − S , so that the resulting placement of guards is still
a dominating set. The secure domination number γs (G) equals the minimum size of a
secure dominating set in G .

NP-problem 24. Does G have a secure dominating set of size at most k ?

Algorithm 31. Design an algorithm for computing the value of γs (T ) for any tree T .

In a very recent paper, Benecke, Cockayne and Mynhardt [5] defined a secure total
dominating set to be a set S ⊂ V which is (i) a total dominating set, and (ii) for every
vertex u ∈ V − S , there exists a vertex v ∈ S ∩ N (u) such that S − {v} ∪ {u} is a total
dominating set. The secure total domination number γst (G) equals the minimum size of
a secure total dominating set in G .

NP-problem 25. Does G have a secure total dominating set of size at most k ?

Algorithm 32. Design an algorithm for computing the value of γst (T ) for any tree T .

5.8 Eternal Security

Recent papers by Burger, Cockayne, Grundlingh, Mynhardt, Winterbach and van Vu-
uren [10] and Goddard, Hedetniemi and Hedetniemi [52] define a new kind of domina-
tion, as follows. Consider placing a guard on each vertex of a dominating set S0 of
a graph G . If for every vertex v ∈ V − S0 there exists an adjacent vertex u ∈ S0 ,
18 Unsolved Algorithmic Problems on Trees

for which the resulting set S1 = S0 − {u} ∪ {v} is a dominating set, then we say that the
set S0 is 1 -secure. A set S0 is eternally 1 -secure if for any sequence v1 , v2 , . . . , vk of
vertices, there exists a corresponding sequence of guards at vertices u1 , u2 , . . . , uk , and a
sequence of dominating sets S1 , S2 , . . . , Sk , so that ui ∈ Si−1 and Si = Si−1 − {ui } ∪ {vi }
and vi ∈ N [ui ] .
The eternal 1 -security number σ1 (G) , (denoted γ∞ (G) in [10]) equals the minimum
size of an eternally 1 -secure dominating set. It is pointed out in [52] that for any bipartite
graph (hence, for any tree), σ1 (G) = β0 (G) .
A more general type of security is defined in [52] in which one permits the movement
of any number of guards from vertices in a current dominating set to adjacent vertices in
order to form a new dominating set. The eternal m -security number, σm (G) , is defined
to equal the minimum size of a dominating set which can handle an arbitrary sequence of
single attacks using multiple guard shifts.
In [52], the complexities of the decision problems associated with the invariants σ1 (G)
and σm (G) are left as open problems. Indeed, the problems of deciding if a given dominat-
ing set is either an eternally 1 -secure dominating set or an eternally m -secure dominating
set are left as open problems.

Algorithm 33. Design an algorithm for deciding if a given set S ⊂ V is eternally 1 -

secure.

Algorithm 34. Design an algorithm for deciding if a given set S ⊂ V is eternally m -

secure.

NP-problem 26. Does G have an eternally 1 -secure dominating set of size at most k ?

NP-problem 27. Does G have an eternally m -secure dominating set of size at most
k?

Algorithm 35. Design an algorithm for computing the value of σm (T ) for any tree T .

5.9 Capacitated Domination

In a recent 2006 paper [55], Goddard, Hedetniemi, Huff and McRae consider capac-
itated domination. An r -capacitated dominating set ( r CDS) of a graph G is a set
S = {v1 , v2 , . . . , vk } ⊆ V for which there exists a partition Π = {V1 , V2 , . . . , Vk } of V
satisfying the following conditions for every i , 1 ≤ i ≤ k :

• vi ∈ Vi ,

• |Vi | ≤ r + 1 ,

• vi is adjacent to all other vertices in Vi .

 Stephen T. Hedetniemi 19

The minimum size of an r CDS in G is called the r -capacitated domination number

and is denoted κr (G) (Note: the authors actually use the Hebrew letter ”daleth”). The
authors study the basic properties of capacitated dominating sets, determine exact values
for various classes of graphs, and show that the problem of deciding if a graph G has
an r -capacitated dominating set of size at most k is NP-hard. They also construct
a linear algorithm for computing the value of κr (T ) for any tree T . They mention,
but do not discuss, the maxi-minimal version of this problem. One can define the upper
r -capacitated domination number κr (G) to equal the maximum size of a minimal r -
capacitated dominating set.

NP-problem 28. Does G have a minimal r -capacitated dominating set of size at least
k?

Algorithm 36. Design an algorithm for computing the value of κr (T ) for any tree T .

6. Broadcast Numbers

Recently Erwin [36] and [37], and Dunbar, Erwin, Haynes, Hedetniemi and Hedetniemi
[31] introduced the concepts of broadcasts in graphs.
We say that a function f : V → {0, 1, . . . , diam(G)} is a broadcast if for every vertex
v ∈ V , f (v) ≤ e(v) , where diam(G) denotes the diameter of G and e(v) denotes the
eccentricity of vertex v . Also, let rad(G) denote the radius of G .
Given a broadcast f , we define Nf [u] = {v | d(u, v) ≤ f (u)} to be the broadcast
neighborhood of u . Further, we define V 0 = {v | f (v) = 0} and V + = V − V 0 =
{u | f (u) > 0} (if there is some potential for ambiguity, then we shall let V + = Vf+ ,
V 0 = Vf0 , and so on). We say that every vertex in V + is a broadcast vertex, and the
broadcast neighborhood Nf [V + ] = v∈V + Nf [v] . If u ∈ V + is a broadcast vertex, v ∈ V
S
and d(u, v) ≤ f (u) , then vertex v can hear a broadcast from vertex u . The set of vertices
that a vertex v ∈ V can hear is defined as H(v) = {u ∈ V + | d(u, v) ≤ f (u)} . For a
vertex v ∈ V + , the private f -neighborhood pnf [v] is the set {u ∈ V | H(u) = {v}} . If
v ∈ pnf [v] , then we say that v is its own private f -neighbor. We define the cost of a
broadcast to be f (V ) = Σv∈V f (v) = Σv∈V + f (v) .
A broadcast f of some type is said to be minimal (respectively, maximal) if there does
not exist a broadcast g 6= f of the same type such that for every u ∈ V , g(u) ≤ f (u)
(respectively, g(u) ≥ f (u) ). Given two distinct broadcasts f and g , we say that f ≤ g
(respectively, f ≥ g ) if and only if for every vertex u ∈ V , f (u) ≤ g(u) (respectively,
f (u) ≥ g(u) ), for all u ∈ V .
Let fS : V → {0, 1} be the characteristic function of a set S ⊆ V of a graph G ,
that is, fS (u) = 1 if u ∈ S , and fS (u) = 0 otherwise. We will be interested in the
characteristic functions of a variety of sets of vertices in a graph.
With this terminology we can define a number of different kinds of broadcasts.
20 Unsolved Algorithmic Problems on Trees

6.1 Dominating broadcasts

As defined in [37], we say that a broadcast f is dominating if Nf [V + ] = V (G) , or

equivalently, if for every v ∈ V , |H(v)| ≥ 1 . The minimum cost f (V ) of a dominating
broadcast f of a graph G is the broadcast domination number γb (G) . We say that a
dominating broadcast f of cost equal to γb (G) is a γb -broadcast. The upper broadcast
domination number equals the maximum cost of a minimal dominating broadcast, and is
denoted Γb (G) .
In a very recent tour de force, Heggernes and Lokshtanov [75] designed a polynomial
algorithm for computing the value of γb (G) for any graph G . At present no algorithm
as been designed for the upper broadcast domination number.
NP-problem 29. Does G have a minimal dominating broadcast of size at least k ?
Algorithm 37. Design an algorithm for computing the value of Γb (T ) for any tree T .

6.2 Independent broadcasts

A broadcast f is called independent if for every vertex v ∈ V + , Nf [v] ∩ V + = {v} , or

equivalently, |H(v)| = 1 . If f is an independent broadcast, then no broadcast vertex can
hear a broadcast from any other broadcast vertex. We note that an independent broadcast
need not be a dominating broadcast. The maximum cost of an independent broadcast of
G is called the broadcast independence number and is denoted βb (G) . The lower broadcast
independence number ib (G) equals the minimum cost of a maximal independent broadcast
of G .
The following inequalities illustrate the relationships between all four broadcasting in-
variants.

γ(G)
W ≤ i(G) ≤ β0 (G)
V ≤ Γ(G)
V
|  | |
γb (G) ≤ ib (G) ≤ βb (G)  Γb (G)

NP-problem 30. Does G have a maximal independent broadcast of size at most k ?

Algorithm 38. Design an algorithm for computing the value of ib (T ) for any tree T .
NP-problem 31. Does G have an independent broadcast of size at least k ?
Algorithm 39. Design an algorithm for computing the value of βb (T ) for any tree T .

6.3 Independent dominating broadcasts

A broadcast f is called independent dominating if it is both independent and dom-

inating. The maximum cost of a minimal independent dominating broadcast of G
 Stephen T. Hedetniemi 21

is called the upper broadcast independent domination number and is denoted Γib (G) . Sim-
ilarly, the broadcast independent domination number γib (G) equals the minimum cost of
an independent dominating broadcast of G . Clearly, γib (G) ≤ i(G) and Γib (G) ≥ β0 (G) ,
since the characteristic function fS of any maximal independent set S is a minimal, in-
dependent dominating broadcast. Note that if f is a minimal independent dominating
broadcast, then for any broadcast g 6= f for which g ≤ f , we know that g is independent
but is not a dominating broadcast.
In domination theory it is well known that a strict inequality γ(G) < i(G) often holds.
However, for broadcasts we get a different result.
Theorem 5. [37] If f is a broadcast on a graph G that is dominating but not independent,
then there is a broadcast g on G that is dominating, independent, with g(V ) ≤ f (V ) ,
and Vg+ ⊂ Vf+ .
Corollary 2. [37] Every graph G has a γb -broadcast which is independent, that is, for
any graph G ,
γb (G) = γib (G).
Proposition 2. For any graph G ,
γib (G) ≤ ib (G) ≤ rad(G) ≤ diam(G) ≤ Γib (G).
NP-problem 32. Does G have a minimal independent dominating broadcast set of size
at least k ?
Algorithm 40. Design an algorithm for computing the value of Γib (T ) for any tree T .

6.4 Efficient broadcasts

A broadcast f is called efficient if every vertex hears exactly one broadcast, that is, for
every vertex v ∈ V , |H(v)| = 1 . The maximum cost of an efficient broadcast is the upper
broadcast efficiency number Γeb (G) , and the broadcast efficiency number γeb (G) equals
the minimum cost of an efficient broadcast.
For example, Figure 1 illustrates twelve distinct efficient broadcasts in the path P7 ,
where γeb (P7 ) = 3 and Γeb (P7 ) = 6.
Theorem 6. Every graph G has a γb (G) -broadcast which is efficient.

By definition, γb (G) ≤ γeb (G) , so Theorem 6 implies that γb (G) = γeb (G) . We know
that Γeb (G) ≤ min{βb (G), Γb (G), Γib (G)} for any graph G , because every efficient domi-
nating broadcast is independent and minimal dominating. Also, observe that any diameter
broadcast is an efficient dominating broadcast. Therefore we have the following corollary.
22 Unsolved Algorithmic Problems on Trees

0 0 2 0 0 0 1 1 0 0 1 0 0 1
s s s s s s s s s s s s s s

2 0 0 0 0 2 0 0 2 0 0 0 1 0
s s s s s s s s s s s s s s

0 3 0 0 0 0 1 0 0 0 3 0 0 0
s s s s s s s s s s s s s s

3 0 0 0 0 1 0 3 0 0 0 0 0 2
s s s s s s s s s s s s s s

4 0 0 0 0 0 1 0 0 4 0 0 0 0
s s s s s s s s s s s s s s

6 0 0 0 0 0 0 0 5 0 0 0 0 0
s s s s s s s s s s s s s s

Figure 1: Efficient broadcasts in P7 .

Corollary 3. For every graph G ,

γb (G) = γib (G) = γeb (G) ≤ ib (G) ≤ rad(G) ≤ diam(G) ≤ Γeb (G) ≤ min{βb (G), Γb (G), Γib (G)}.

NP-problem 33. Does G have a minimal efficient dominating broadcast of size at least
k?

Algorithm 41. Design an algorithm for computing the value of Γeb (T ) for any tree T .

6.5 Packing broadcasts

A broadcast f is called a packing if every vertex hears at most one broadcast, that is,
for every vertex v ∈ V , |H(v)| ≤ 1. The maximum cost of a packing broadcast of G is
called the broadcast packing number and is denoted Pb (G) . Similarly, the lower broadcast
packing number equals the minimum cost of a maximal packing broadcast, and is denoted
pb (G) .
Note first that the characteristic function fS of a maximal packing need not be a
maximal packing broadcast. This can be seen by considering the path P5 with values
assigned as indicated in Figure 2(a). This center vertex in Figure 2(a) is a maximal
packing, but it is not a maximal packing broadcast, since the value of 1 can be increased
to 2 , which results in the maximal packing broadcast (see Figure 2(b)).
 Stephen T. Hedetniemi 23

0 0 1 0 0 0 0 2 0 0
s s s s s s s s s s
(a) (b)

Figure 2: A maximal packing which is not a maximal packing broadcast.

Note that one may have a maximal packing broadcast which is not a dominating broad-
cast. For example, consider the values 1,0,0,1,0,0 in the graph of P6 . We know that for
any graph G ,
p(G) ≤ P (G) ≤ γ(G).

Proposition 3. Every efficient broadcast is

(i) a maximal packing broadcast,

(ii) a minimal dominating broadcast, and
(iii) a minimal independent dominating broadcast.

Corollary 4. For any graph G ,

pb (G) ≤ γeb (G) = γib (G) = γb (G) ≤ ib (G) ≤ Γeb (G) ≤

min{Pb (G), Γb (G), Γib (G), βb (G)}.

NP-problem 34. Does G have a maximal packing broadcast of size at most k ?

Algorithm 42. Design an algorithm for computing the value of pb (T ) for any tree T .

NP-problem 35. Does G have a packing broadcast of size at least k ?

Algorithm 43. Design an algorithm for computing the value of P b (T ) for any tree T .

7. Matchings

A set M ⊆ E of edges is called independent if no two edges in M have a vertex in

common. A matching in a graph G is any independent set M of edges. The matching
number β 1 (G) equals the maximum size of a matching in G . The minimum maximal
matching number β1 (G) equals the minimum size of a maximal matching in G . We will
also refer to β1 (G) as the lower matching number.
Perhaps the first algorithm for computing the value of β 1 (T ) for a tree was that of
Mitchell [86], which appeared earlier in a 1975 paper by Mitchell, Hedetniemi and Good-
man [88]. A β 1 (T ) -algorithm was also constructed independently by Savage in 1980 [91].
The first algorithm for computing the value of β1 (T ) was by Mitchell and Hedetniemi in
1977 [87]. It this paper it was also pointed out for the first time that for any tree T ,
γ 0 (T ) = β1 (T ) , that is, the edge domination number of any tree can always be achieved
with a maximal matching of minimum cardinality.
24 Unsolved Algorithmic Problems on Trees

Given a matching M we say that every vertex that is incident to an edge in M is

saturated by M , and every edge in E − M both of whose vertices are saturated by M is
called a saturated edge. Let V (M ) denoted the set of vertices saturated by M , and let
E(M ) be the set of edges saturated by M . The subgraph G[M ] = (V (M ), E(M ) ∪ M )
is the subgraph induced by M .
We say that a graph G has a perfect matching if it has a matching M for which
V [M ] = V (G) .
A matching M is called induced (or sometimes strong) if no two edges in M are joined
by an edge. Equivalently, a matching is induced if it does not saturate any edges. Thus,
in this case, G[M ] = (V (M ), M ) . Induced matchings were introduced by Cameron [11].
The induced matching number β ∗ (G) of a graph G equals the maximum size of an
induced matching in G , while the lower induced matching number β∗ (G) equals the
minimum size of a maximal induced matching in G . Algorithms for computing the value
of β ∗ (T ) for any tree have been constructed by Cameron [11](chordal graphs), Fricke and
Laskar [48] and Zito [99] and [100]. An algorithm for computing the value of β∗ (T ) was
recently constructed by Goddard, Hedetniemi, Hedetniemi and Laskar [54].
The next several sections will discuss open problems related to a variety of different
kinds of matchings.

7.1 Uniquely Restricted Matchings

A matching M is said to be uniquely restricted if there is no other matching of the

same size on the vertices saturated by M . Equivalently, a matching is uniquely restricted
if and only if the subgraph G[M ] contains no (alternating) cycles. Thus, in a tree, every
matching is automatically uniquely restricted.
The uniquely restricted matching number β ur (G) equals the maximum size of a uniquely
restricted matching in G , while the lower uniquely restricted matching number βur (G)
(NEW HERE) equals the minimum size of a maximal uniquely restricted matching in G .
Uniquely restricted matchings were introduced by Golumbic, Hirst and Lewenstein [57],
and have been studied by Levit and Mandrescu [81], [82] and [83].
Notice that every induced matching is a uniquely restricted matching. Therefore:

βur (G) ≤ β∗ (G) ≤ β ∗ (G) ≤ β ur (G).

UNIQUELY RESTRICTED MATCHING

INSTANCE: graph G and a positive integer k .
QUESTION: Does G have a uniquely restricted matching of size at least k ?
Golumbic, Hirst and Lewenstein [57] show that UNIQUELY RESTRICTED MATCH-
ING is NP-complete for bipartite and split graphs, but can be solved in polynomial time
for proper interval graphs, trees, cacti and threshold graphs. They leave as unsolved the
complexity of UNIQUELY RESTRICTED MATCHING for interval graphs.
 Stephen T. Hedetniemi 25

It is interesting to note that one may have to spend some time, given an arbitrary
matching M in a given class of graphs, deciding whether M is a uniquely restricted
matching. The authors show that this question can usually be answered in time O(|V | +
|E|) .
Apparently the authors did not consider the mini-maximal version of uniquely restricted
matchings. Trivially, the lower uniquely restricted matching number of any tree equals
the lower matching matching number of a tree, and therefore an algorithm exists for this
problem [87]. However the NP-completeness question has not been settled.

NP-problem 36. Does G have a maximal uniquely restricted matching of size at most
k?

7.2 Connected, Isolatefree, Acyclic and Disconnected Matchings

A recent paper by Goddard, Hedetniemi, Hedetniemi and Laskar [54] introduces four
kinds of matchings
A matching M is said to be connected if the induced subgraph G[M ] is connected.
The connected matching number β c (G) equals the maximum size of a connected matching
in G . The lower connected matching number is denoted βc (G) .
A matching is said to be isolate free if either |M | = 1 or G[M ] has no K2 component.
The corresponding invariants are denoted β if (G) and βif (G) .
A matching M is said to be acyclic if the induced subgraph G[M ] contains no cycles.
The acyclic matching number β a (G) equals the maximum size of an acyclic matching in
G , while the lower acyclic matching number is denoted βa (G)
Finally a matching is called disconnected if G[M ] is a disconnected graph. The corre-
sponding two invariants are denoted β dc (G) and βdc (G) .
The following results are given in [54].

1. β ∗ (G) ≤ β a (G) ≤ β ur (G) ≤ β 1 (G).

2. β ∗ (G) ≤ β dc (G) ≤ β 1 (G).

3. β c (G) ≤ β if (G) ≤ β 1 (G).

4. For every connected graph G , β c (G) = β if (G) = β 1 (G).

5. For any tree T , β 1 (T ) − 1 ≤ β dc (T ) ≤ β 1 (T ) .

6. For any connected graph G , β1 (G) ≤ βif (G) ≤ βc (G).

It is also pointed out in [54] that the connected and isolate free matching numbers can be
computed in polynomial time, while the ACYCLIC MATCHING Problem is NP-hard. In
26 Unsolved Algorithmic Problems on Trees

addition it is mentioned that LOWER ACYCLIC MATCHING, LOWER CONNECTED

MATCHING AND LOWER DISCONNECTED MATCHING are all NP-hard.
An interesting open problem is that of determining the complexity of the following
problem.
NP-problem 37. DISCONNECTED MATCHING
INSTANCE: A graph G and a positive integer k .
QUESTION: Does G have a disconnected matching of size at least k ?
Algorithm 44. Design an algorithm to compute the value of βif (T ) for any tree T .
Algorithm 45. Design an algorithm to compute the value of βc (T ) for any tree T .
Algorithm 46. Design an algorithm to compute the value of βdc (T ) for any tree T .
Algorithm 47. Design an algorithm to compute the value of β dc (T ) for any tree T .

7.3 b -Matchings

For each vertex v ∈ V specify a value b(v) , where 1 ≤ b(v) ≤ d(v) , where d(v) is
the degree of v . In this way we define a vector called a b -vector. A b -matching is a
set of edges M having the property that no vertex v is incident with more than b(v)
edges of M . The b -matching number β (b) (G) equals the maximum number of edges in
a b -matching in G . The lower b -matching number β(b) (G) (NEW HERE) equals the
minimum number of edges in a maximal b -matching in G .
One of the earliest papers on b -matchings in graphs was that of Goodman, Hedetniemi
and Tarjan in 1976 [59]. This paper contains a linear time, greedy algorithm for computing
the value of β (b) (T ) for any tree T .
However, the authors did not define or consider the mini-maximal version of this prob-
lem.
NP-problem 38. Does G have a maximal b -matching of size at most k ?
Algorithm 48. Design an algorithm for computing the value of β(b) (T ) for any tree T .

8. Influence Numbers

In [93], Stege and Van Rooij introduced the concept of the influence number of a set of
vertices and the influence number of a graph (see also [1]). Let S ⊂ V be a set of vertices.
For any vertex u ∈ V the value d(u, S) equals the minimum distance d(u, v) between
vertex u and a vertex v ∈ S .
The influence of S is defined to equal
X 1
η(S) = .
u∈S
2d(u,S)
 Stephen T. Hedetniemi 27

The influence number η(G) of a graph G equals the maximum influence of a set S
in G . In [27] Daugherty settled the NP-completeness of the decision problem associated
with η(G) , while McRae [27] has constructed a linear algorithm for computing the value
of η(T ) for any tree T .
Similarly, Hartnell, Hedetniemi, Hedetniemi and Rall (cf. [27]) defined the total influ-
ence of a set S to equal
X X 1
d(u,v)
.
u∈S v∈V −S
2

The total influence number ηt (G) equals the maximum total influence of a set S ⊂ V .
Daugherty, Lyle and Laskar have recently published a paper on the total influence number
of a graph [28]. However, the complexity of the decision problem associated with the total
influence number has not been settled, nor has an algorithm been constructed for trees.

NP-problem 39. Does G have a total influence set of size at least k ?

Algorithm 49. Design an algorithm for computing the value of ηt (T ) for any tree T .

The mini-maximal problems for the influence and total influence numbers have not yet
been defined and studied. We say that a set S is a maximal (total) influence set if no
proper superset of S has greater (total) influence. While it is difficult to define a symbol
for these two invariants, we will use a subscript of mm, for mini-maximal, to denote these
two values. Let ηmm (G) (NEW HERE) equal the mini-maximal influence number of G ,
and ηmmt (G) (NEW HERE) equal the mini-maximal total influence number of G .

NP-problem 40. Does G have a maximal influence set of size at most k ?

Algorithm 50. Design an algorithm for computing the value of ηmm (T ) for any tree T .

NP-problem 41. Does G have a maximal total influence set of size at most k ?

Algorithm 51. Design an algorithm for computing the value of ηmmt (T ) for any tree
T.

9. Vertex Partition and Coloring Problems

Several problems involved with partitioning the vertex set of a graph into sets, each of
which has a given property P , have recently appeared, whose solutions for trees have not
been found. We present several examples.

9.1 Broadcast Chromatic Number

As introduced by Goddard, Hedetniemi, Hedetniemi, Harris and Rall [53], a function

π : V → {0, 1, . . . , k} is called a broadcast coloring of order k if π(u) = π(v) implies that
28 Unsolved Algorithmic Problems on Trees

d(u, v) > π(u) . The minimum order of a broadcast coloring of a graph G is called the
broadcast chromatic number, and is denoted by χb (G) . Equivalently, a broadcast coloring
is a partition Π = {V1 , V2 , . . . , Vk } of V such that each color class Vi is an i -packing,
that is, the distance between any two vertices in Vi is greater than i .
In [53] the authors show that one can determine in polynomial time if an arbitrary graph
has a broadcast coloring of order 1 , 2 , or 3 . However, they show that the problem of
deciding if a graph has a broadcast coloring of order 4 is NP-hard. The computation of
the value of χb (T ) for any tree has proved to be quite difficult.

Algorithm 52. Design an algorithm for computing the value of χb (T ) for any tree T .

We note that a paper similar to [53] has been written by Sloper [92], who uses a
slightly different definition of a broadcast coloring. We will give these colorings a different
name and call them broadcast e-colorings (”e” for eccentricity). A broadcast e-coloring
of a graph G is a function color : V → N such that (i) color(u) = color(v) implies
that d(u, v) > color(u) and (ii) for every v ∈ V , color(v) ≤ e(v) , where e(v) is the
eccentricity of v . According to this definition, many graphs will not have a broadcast
e-coloring. In [92] Sloper presents several classes of trees which have broadcast e-colorings
and several classes of trees which are not broadcast e-colorable. This raises an interesting
problem:

NP-problem 42. Given an arbitrary tree T , is it broadcast e-colorable?

9.2 Dominator Partitions

In a very recent paper, Cockayne, Hedetniemi, Hedetniemi, Laskar, McRae and Wallis
[23] introduced the concept of a dominator partition. We say that a vertex v ∈ V is a
dominator of a set S ⊂ V if v dominates (is adjacent to) every vertex in S . A partition
π = {V1 , V2 , . . . , Vk } is called a dominator partition if every vertex v ∈ V is a dominator
of at least one block Vi of π . A dominator partition is said to be minimal if any partition
π 0 obtained from π by forming the union of any two classes Vi ∪ Vj , i 6= j , is no longer
a dominator partition. The dominator partition number πd (G) equals the minimum k
such that G has a dominator partition of order k . The upper dominator partition number
Πd (G) equals the maximum order of a minimal dominator partition of G . In something
of a surprise, the authors prove in [23]:

Theorem 7. For any graph G , γ(G) ≤ πd (G) ≤ γ(G) + 1 .

They then show that the problem of deciding if a graph G has a dominator partition of
order at most k is NP-complete, even for bipartite, planar or chordal graphs. In addition
they show that the value of πd (T ) can be computed in linear time for any tree T . Also
in polynomial time, one can decide if πd (G) ≤ 2 . However, they were not able to settle
the complexity of the following interesting question:
 Stephen T. Hedetniemi 29

NP-problem 43. Does G have a dominator partition of order at most 3 ?

Very recently, Goddard, Jacob and Laskar [56] presented several results on the upper
dominator partition number. However, they have not solved the following two problems.

NP-problem 44. Does G have a minimal dominator partition of order at least k ?

Algorithm 53. Design an algorithm for computing the value of Πd (T ) for any tree T .

9.3 Dominator Colorings

In a very recent paper Hedetniemi, Hedetniemi, McRae and Blair [71] introduce the
concept of dominator colorings of graphs. A subsequent paper on dominator colorings
by Gera, Rasmussen and Horton [50] was recently presented at the 37th Southeastern
International Conference on Graph Theory, Combinatorics and Computing. A dominator
coloring of a graph G is a dominator partition π = {V1 , V2 , . . . , Vk } in which each class
Vi is an independent set. Thus, dominator partitions are proper colorings, in which each
vertex v ∈ V is a dominator of at least one color class. As with dominator colorings,
one can say that a dominator coloring π is minimal if any partition π 0 obtained from
π by forming the union of any two classes Vi ∪ Vj , i 6= j , is no longer a dominator
coloring. We define the dominator chromatic number χd (G) to equal the minimum order
of a dominator coloring of G , while the dominator achromatic number ψd (G) equals the
maximum order of a minimal dominator coloring of G .
Let χ(G) denote the well known chromatic number of a graph G .

Theorem 8. [71] For any graph G , max{γ(G), χ(G)} ≤ χd (G) ≤ γ(G) + χ(G) .

It is also shown that the problem of deciding if a graph G has a dominator coloring of
order at most k is NP-complete. The following is also proved:

Proposition 4. [71] For any tree T , γ(T ) ≤ χd (T ) ≤ γ(T ) + 1.

However the authors have not been able to construct an algorithm for trees, and have
not settled the decision problem associated with ψd (G) .

Algorithm 54. Design an algorithm for computing the value of χd (T ) for any tree T .

NP-problem 45. Does G have a minimal dominator coloring of order at least k ?

Algorithm 55. Design an algorithm for computing the value of ψd (T ) for any tree T .

10. Problems Involving Independent Sets

This concluding section is something of a postscript to most of the preceding sections.

Most of the these problems ask you to find the minimum or maximum size of a set having
30 Unsolved Algorithmic Problems on Trees

one or more specified properties. What if, in addition, you add the requirement (if it
isn’t already stipulated) that the set be independent? Immediately a number of problems
take on added interest. We conclude by listing several of these that we think are of some
interest.

Algorithm 56. Design an algorithm for computing the minimum size of a 1 -maximal
independent nearly perfect set in a tree T .

Algorithm 57. Design algorithms for computing the minimum and maximum size of a
maximal independent total irredundant set in a tree T .

Algorithm 58. Design algorithms for computing the minimum and maximum size of an
independent open perfect neighborhood set in a tree T .

Algorithm 59. Design an algorithm for computing the minimum size of an independent
k -capacitated dominating set in a tree T .

Algorithm 60. Design an algorithm for computing the maximum influence of an inde-
pendent set in a tree T .

Algorithm 61. Design an algorithm for computing the maximum total influence of an
independent set in a tree T .

11. Summary

The tables below provide a quick summary of these many open problems.

NP T ree Lower N umber U pper T ree NP

Algorithm Algorithm
[89] oir open irredundance OIR [77] [38]
[85] ooir open-open irredundance OOIR [47] [11]
[85] coir closed-open irredundance COIR [58] [42]
ioir independent open irredundance IOIR
irt total irredundance IRt [70] [70]
oirt total open irredundance OIRt
ooirt total open-open irredundance OOIRt
coirt total closed-open irredundance COIRt
er external redundant ER
[72] θo open perfect neighborhood Θo
θi independent perfect neighborhood Θi
ra R-annihilated Ra
rai R-annihilated irredundance Rai
[60] i[1] [1]-dependence β[1] [60] [60]
[40] Simple γ[1] [1]-dependent domination Γ[1] [40]
[40] Simple ir[1] [1]-dependent irredundance IR[1] [40]
ntp total nearly perfect Ntp
 Stephen T. Hedetniemi 31

NP T ree Lower N umber U pper T ree NP

Algorithm Algorithm
[49] [19] γooir open open irredundant domination Γooir
[49] [19] γcoir closed-open irredundant domination Γcoir
[66] [51] γpr paired domination Γpr
[97] [97] γp perfect domination Γp
[30] [30] γr restrained domination Γr
[61] [61] γ−11 signed domination Γ−11 [61]
[32] [32] γ−101 minus domination Γ−101 [32]
γs secure domination Γs
γst secure total domination Γst
σ1 eternal 1-security Σ1
σm eternal m-security Σm
[55] [55] κr r-capacitated domination κr
[75] [75] γb broadcast domination Γb
ib independent broadcast βb
[75] [75] γib independent dominating broadcast Γib
[75] [75] γeb efficient dominating broadcast Γeb
pb packing broadcast Pb
[87] βur uniquely restricted matching β ur [57] [57]
[54] βif isolate-free matching β if [54] [54]
[54] βc connected matching βc [54] [54]
[54] βdc disconnected matching β dc
β(b) (b)-matching β (b) [59]
ηmm influence η [27] [27]
ηmmt total influence ηt
[53] χb broadcast chromatic
[23] [23] πd dominator partition Πd
[71] χd dominator chromatic ψd
nip independent nearly perfect Nip
κik independent k-capacitated domination κik
ηi independent influence ηi
ηit independent total influence η it

Acknowledgments

The author gratefully acknowledges Professors Sandra Hedetniemi, Wayne Goddard, Renu
Laskar, Brian Dean and David Jacobs at Clemson University, Professor Alice McRae at Ap-
palachian State University, Professor Doug Rall at Furman University, Professor Jean Dunbar at
Converse College, Professor E. J. Cockayne at the University of Victoria, Professor Johan Hattingh
at Georgia State University, Professor Peter J. Slater at The University of Alabama in Huntsville,
and Professor Teresa Haynes at East Tennessee State University, for their generous help in listing
these open problems and providing appropriate bibliographic entries. Any omissions or incorrect
citations, should they be present, are entirely the fault of the author, for which apologies are
offered.
32 Unsolved Algorithmic Problems on Trees

References

[1] P. Agah, A. Hertel, P. Hertel, A. Scott, U. Stege and I. van Rooij, Profit problems in graphs:
Classical graph problems reconsidered, Manuscript, 2005.
[2] S. Arnborg, Efficient algorithms for combinatorial problems on graphs with bounded decom-
posability - a survey, BIT, 25 (1985),2-23.
[3] S. Arnborg, J. Lagergren and D. Seese, Easy problems for tree-decomposable graphs, J. Al-
gorithms, 12(1991), 308-340.
[4] S. Arnborg and A. Proskurowski, Linear time algorithms for NP-hard problems restricted to
partial k -trees, Discrete Appl. Math., 23 (1989),11-24.
[5] S. Benecke, E.J. Cockayne and C.M. Mynhardt, Secure total domination in graphs, submitted
for publication.
[6] M.W. Bern, E.L. Lawler and A.L. Wong, Linear-time computation of optimal subgraphs of
decomposable graphs, J. Algorithms, 8(1987), 216-235.
[7] T. Beyer, A. Proskurowski, S. Hedetniemi and S. Mitchell, Independent domination in trees,
Proc. Eighth Southeastern Conf. on Combinatorics, Graph Theory and Computing, Util.
Math., Winnipeg, 321-328, 1977.
[8] H.L. Bodlaender, A tourist guide through treewidth, Acta Cybernet., 11(1993),1-23.
[9] B. Bollobas and E.J. Cockayne, Graph theoretic parameters concerning domination, indepen-
dence and irredundance, J. Graph Theory, 3(1979),241-249.
[10] A.P. Burger, E.J. Cockayne, W.R. Grundlingh, C.M. Mynhardt, W. Winterbach and J.H.
van Vuuren, Infinite order domination in graphs, J. Combin. Math. Combin. Comput., 50
(2004),179-194.
[11] K. Cameron. Induced matchings, Discrete Appl. Math., 24(1989),97-102.
[12] E.J. Cockayne, Generalized irredundance in graphs: hereditary properties and Ramsey num-
bers, J. Combin. Math. Combin. Comput., 31(1999), 15-31.
[13] E.J. Cockayne, R.M. Dawes and S.T. Hedetniemi, Total domination in graphs, Networks,
10(1980), 211-219.
[14] E.J. Cockayne, O. Favaron and C.M. Mynhardt, Secure domination, weak Roman domination
and forbidden subgraphs, Bull. Inst. Combin. Appl., 39 (2003), 87-100.
[15] E.J. Cockayne, O. Favaron, C. Payan and A.G. Thomason, Contributions to the theory of
domination, independence and irredundance in graphs, Discrete Math., 33(1981), 249-258.
[16] E.J. Cockayne, O. Favaron, J. Puech and C.M. Mynhardt, Packing, perfect neighbourhood,
irredundant and R -annihilated sets in graphs, Australas. J. Combin., 18(1998), 253-262.
[17] E.J. Cockayne, O. Favaron, J. Puech and C.M. Mynhardt, An inequality chain of domination
parameters for trees, Discuss. Math. Graph Theory, 18(1)(1998), 127-142.
[18] E.J. Cockayne and S. Finbow, Generalised irredundance in graphs: Nordhaus-Gaddum
bounds, Discuss. Math. Graph Theory, (To appear).
 Stephen T. Hedetniemi 33

[19] E.J. Cockayne, S.E. Goodman and S.T. Hedetniemi, A linear algorithm for the domination
number of a tree, Inform. Process. Lett., 4(1975),41-44.
[20] E.J. Cockayne, P.J.P. Grobler, S.T. Hedetniemi and A.A. McRae, What makes an irredundant
set maximal? J. Combin. Math. Combin. Comput., 25(1997),213-223.
[21] E.J. Cockayne, J.H. Hattingh, S.M. Hedetniemi, S.T. Hedetniemi and A. A. McRae, Us-
ing maximality and minimality conditions to construct inequality chains, Discrete Math.,
176(1997),43-61.
[22] E.J. Cockayne, B.L. Hartnell, S.T. Hedetniemi and R. Laskar, Perfect domination in graphs,
J. Combin. Inform. Sys. Sci., 18(1-2)(1993), 136-148.
[23] E.J. Cockayne, S.M. Hedetniemi, S.T. Hedetniemi, R. Laskar and C.K. Wallis, Dominator
partitions of graphs, Submitted for publication, 2006.
[24] E.J. Cockayne, S.M. Hedetniemi, S.T. Hedetniemi and C.M. Mynhardt, Irredundant and
perfect neighbourhood sets in trees, Discrete Math., 188(1-3)(1998), 253-260.
[25] E.J. Cockayne, S.T. Hedetniemi and D.J. Miller, Properties of hereditary hypergraphs and
middle graphs, Canad. Math. Bull., 21(1978),461-468 .
[26] E.J. Cockayne and C.M. Mynhardt, On a conjecture concerning irredundant and perfect
neighbourhood sets in graphs, Papers in honour of Stephen T. Hedetniemi, J. Combin. Math.
Combin. Comput., 31(1999),241-253.
[27] S.M. Daugherty, The influence and total influence numbers of a graph, MS paper, Dept.
Mathematical Sciences, Clemson University, May 2005.
[28] S. Daugherty, J. Lyle and R. Laskar, On the total influence number of a graph, Congr. Numer.,
174(2005),107-121.
[29] D.E. Daykin and C.P. Ng, Algorithms for generalized stability number of tree graphs, J.
Austral. Math. Soc., 6(1966),89-100.
[30] G.S. Domke, J.H. Hattingh, S.T. Hedetniemi, R.C. Laskar and L.R. Markus, Restrained
domination in graphs, Discrete Math., 203(1999),61-69.
[31] J.E. Dunbar, D.J. Erwin, T.W. Haynes, S.M. Hedetniemi and S.T. Hedetniemi, Broadcasts
in graphs, Discrete Appl. Math., 154(1)(2006),59-75.
[32] J. Dunbar, W. Goddard, S. Hedetniemi, A. McRae and M.A. Henning, The algorithmic com-
plexity of minus domination in graphs, Discrete Appl. Math., 68(1996), 73-84.
[33] J.E. Dunbar, F.C. Harris, S.M. Hedetniemi, S.T. Hedetniemi, A.A. McRae and R.C. Laskar,
Nearly perfect sets in graphs, Discrete Math., 138(1995),229-246.
[34] J. Dunbar, S. Hedetniemi, M.A. Henning and A. McRae, Minus domination in graphs, Discrete
Math., 199(1999), 35-47.
[35] J. Dunbar, S.T. Hedetniemi, M.A. Henning and P.J. Slater, Signed domination in graphs,
In Proc. Seventh Quadrennial Internat. Conf. on the Theory and Applications of Graphs, Y.
Alavi and A. Schwenk, Eds., Graph Theory, Combinatorics and Algorithms 1(1995),311-322.
[36] D. Erwin, Cost domination in graphs, PhD Thesis, Western Michigan University, 2001.
[37] D. Erwin, Dominating broadcasts in graphs, Bull. Inst. Combin. Appl., 42(2004),89-105.
34 Unsolved Algorithmic Problems on Trees

[38] S. Even, O. Goldreich, S. Moran and P. Tong, On the NP-completeness of certain network
testing problems, Networks, 14(1984),1-24.
[39] O. Favaron, T. Haynes, S.T. Hedetniemi, M.A. Henning and D. Knisley, Total irredundance
in graphs, Discrete Math., 256(2002),115-127.
[40] O. Favaron, S.M. Hedetniemi, S.T. Hedetniemi and D.F. Rall, On k -dependent domination,
Discrete Math., 249(2002),83-94.
[41] O. Favaron and J. Puech, Irredundant and perfect neighborhood sets in graphs and
claw-free graphs, 16th British Combinatorial Conference (London, 1997), Discrete Math.,
197/198(1999),269-284.
[42] M. Fellows, G. Fricke, S. Hedetniemi and D. Jacobs, The private neighbor cube, SIAM J.
Discrete Math., 7(1994), 41-47.
[43] S. Finbow, Generalisations of irredundance in graphs, PhD Thesis, University of Victoria,
2003.
[44] J.F. Fink and M.S. Jacobson, n -domination in graphs, In Graph Theory with Applications to
Algorithms and Computer Science, Y. Alavi and A.J. Schwenk, Eds., (Kalamazoo, MI 1984),
Wiley, 283-300, 1985.
[45] G.H. Fricke, E.O. Hare, D.P. Jacobs and A. Majumdar. On integral and fractional total
domination, Congr. Numer., 77(1990), 87-95.
[46] G.H. Fricke, T.W. Haynes, S.M. Hedetniemi, S.T. Hedetniemi and M.A. Henning, On perfect
neighborhood sets in graphs, Discrete Math., 199 (1999), 221-225.
[47] G. Fricke, S.M. Hedetniemi and R. Laskar, Open-irredundance in trees, Unpublished
manuscript, 1989.
[48] G. Fricke and R. Laskar, Strong matchings in trees, Congr. Numer., 89 (1992), 239-243.
[49] M.R. Garey and J.S. Johnson, Computers and Intractability, A Guide to the Theory of NP-
Completeness, Freeman, 1979.
[50] R. Gera, C. Rasmussen and S. Horton, Dominator colorings and safe clique partitions, Pre-
sented at 37th Southeastern Internat. Conf. on Graph Theory, Combinatorics and Computing,
(Florida Atlantic University, March 6-10, 2006).
[51] W. Goddard, Private communication, April 2006.
[52] W. Goddard, S.M. Hedetniemi and S.T. Hedetniemi, Eternal security in graphs, J. Combin.
Math. Combin. Comput., 52(2005), 169-180.
[53] W. Goddard, S.M. Hedetniemi, S.T. Hedetniemi, J.M. Harris and D.F. Rall, Broadcast chro-
matic numbers of graphs, Ars Combin., (To appear).
[54] W. Goddard, S.M. Hedetniemi, S.T. Hedetniemi and R. Laskar, Generalized subgraph-
restricted matchings in graphs, Discrete Math., 293 (2005), 129-138.
[55] W. Goddard, S.T. Hedetniemi, J.L. Huff and A.A. McRae, Capacitated domination, Submit-
ted for publication, January 2006.
[56] W. Goddard, J. Jacob and R. Laskar, On the upper dominator partition problem of cer-
tain graphs. Presented at the Thirty Seventh Southeastern Internat. Conf. on Combinatorics,
Graph Theory and Computing, (Florida Atlantic University, March 6-10, 2006).
 Stephen T. Hedetniemi 35

[57] M.C. Golumbic, T. Hirst and M. Lewenstein, Uniquely restricted matchings, Algorithmica,
31 (2001), 139-154.
[58] M.C. Golumbic and R.C. Laskar, Irredundancy in circular arc graphs, Discrete Appl. Math.,
44 (1993), 79-89.
[59] S. Goodman, S. Hedetniemi and R.E. Tarjan, b -matchings in trees, SIAM J. Comput., 5(1)
(1976), 104-108.
[60] E.O. Hare, S.M. Hedetniemi and S.T. Hedetniemi, 2 -dependence numbers of trees, Unpub-
lished manuscript, dated September 11, 1989.
[61] J.H. Hattingh, M.A. Henning and P.J. Slater, The algorithmic complexity of signed domina-
tion in graphs, Australas. J. Combin., 12(1995), 101-112.
[62] T.W. Haynes, S.T. Hedetniemi, M.A. Henning and D.J. Knisley, Stable and unstable graphs
with total irredundance number zero, Ars Combin., 61 (2001), 33-46.
[63] T.W. Haynes, S.T. Hedetniemi and P.J. Slater, Fundamentals of Domination in Graphs, Mar-
cel Dekker, 1998, 446 pp.
[64] T.W. Haynes, S.T. Hedetniemi and P.J. Slater, Eds. Domination in Graphs: Advanced Topics,
Marcel Dekker, 1998, 497 pp.
[65] T.W. Haynes, M.A. Henning and P.J. Slater, Trees with equal domination and paired-
domination numbers, Ars Combin.,76 (2005),169-175.
[66] T.W. Haynes and P.J. Slater, Paired-domination in graphs, Networks, 32(3)(1998), 199-206.
[67] T.W. Haynes and P.J. Slater, Paired-domination and the paired domatic number, Congr.
Numer., 109(1995), 65-72.
[68] S.M. Hedetniemi, S.T. Hedetniemi and M.A. Henning, The algorithmic complexity of perfect
neighborhoods in graphs, J. Combin. Math. Combin. Comput., 25 (1997), 183-192.
[69] S.M. Hedetniemi, S.T. Hedetniemi and D.P. Jacobs, Private domination: theory and algo-
rithms, Congr. Numer., 79 (1990), 147-157.
[70] S.M. Hedetniemi, S.T. Hedetniemi and D.P. Jacobs, Total irredundance in graphs: theory
and algorithms, Ars Combin., 35A (1993), 271-284.
[71] S.M. Hedetniemi, S.T. Hedetniemi, A.A. McRae and J.R.S. Blair, Dominator colorings of
graphs, Preprint, 2006.
[72] S.T. Hedetniemi, D.P. Jacobs, R. Laskar and D. Pillone, Open perfect neighborhood sets in
graphs, Unpublished manuscript, dated August 23, 1996.
[73] S. Hedetniemi, R. Laskar and J. Pfaff, Irredundance in graphs: a survey, Congr. Numer.,
48(1985), 183-194.
[74] S.T. Hedetniemi, A.A. McRae and D.A. Parks, Complexity Results, Chapter 9 in Domination
in Graphs: Advanced Topics, T.W. Haynes, S.T. Hedetniemi and P.J. Slater, Eds., Marcel
Dekker, 233-269,1998.
[75] P. Heggernes and D. Lokshtanov, Optimal broadcast domination of arbitrary graphs in poly-
nomial time, Rept. No. 290, Dept. Informatics, University of Bergen, February 2005.
[76] D. Hershkowitz and H. Schneider, Ranks of zero patterns and sign patterns, Linear and
Multilinear Algebra, 34(1993),3-19.
36 Unsolved Algorithmic Problems on Trees

[77] M.S. Jacobson and K. Peters, Chordal graphs and upper irredundance, upper domination and
independence, Discrete Math., 86(1990), 59-69.
[78] D. Kratsch, Algorithms, Chapter 8 in Domination in Graphs: Advanced Topics, T.W. Haynes,
S.T. Hedetniemi and P.J. Slater, Eds., Marcel Dekker, 191-231,1998.
[79] R. Laskar, J. Pfaff, S.M. Hedetniemi and S.T. Hedetniemi, On the algorithmic complexity of
total domination, SIAM J. Alg. Disc. Meth., 5 (1984),420-425.
[80] E.L. Lawler and P.J. Slater, A linear time algorithm for finding an optimal dominating sub-
forest of a tree, In Graph Theory with Applications to Algorithms and Computer Science,
(Kalamazoo, MI, 1984), Wiley, 501-506, 1985.
[81] V.E. Levit and E. Mandrescu, Unicycle bipartite graphs with only uniquely restricted max-
imum matchings, Combinatorics, computability and logic (Constanţa, 2001), Springer Ser.
Discrete Math. Theor. Comput. Sci., Springer, 151-157, 2001.
[82] V.E. Levit and E. Mandrescu, Bipartite graphs with uniquely restricted maximum matchings
and their corresponding greedoids.
[83] V.E. Levit and E. Mandrescu, Local maximum stable sets in bipartite graphs with uniquely
restricted maximum matchings, Stability in graphs and related topics. Discrete Appl. Math.,
132(2003), 163-174.
[84] A.A. McRae. External redundant set in graphs, Presented at 25 th Southeastern. Conf. on
Combinatorics, Graph Theory and Computing, (Boca Raton, FL, 1994).
[85] A.A. McRae. Generalizing NP-completeness proofs for bipartite graphs and chordal graphs,
PhD Thesis, Clemson University, 1994.
[86] S.L. Mitchell, Linear algorithms on trees and maximal outerplanar graphs: design, complexity
analysis and data structures study, PhD Thesis, University of Virginia, 1977.
[87] S.L. Mitchell and S.T. Hedetniemi, Edge domination in trees, Congr. Numer., 19 (1977),
489-509.
[88] S. Mitchell, S. Hedetniemi and S. Goodman, Some linear algorithms on trees, Proc. Sixth
Southeastern Conf. on Combinatorics, Graph Theory and Computing, Util. Math., Winnipeg,
467-483, 1975.
[89] J. Pfaff, R. Laskar and S.T. Hedetniemi, NP-completeness of total and connected domination
and irredundance for bipartite graphs, Tech. Rept. 428, Dept. Mathematical Sciences, Clemson
University, July 1983.
[90] K.E. Proffitt, T.W. Haynes and P.J. Slater, Paired-domination in grid graphs, Congr. Numer.,
150(2001),161-172.
[91] C. Savage, Maximum matchings and trees, Inform. Process. Lett., 10(4-5)(1980), 202-205.
[92] C. Sloper, Broadcast-coloring of trees, Rept. No., 233 September 2002, Dept. of Informatics,
University of Bergen.
[93] U. Stege and I. van Rooij, Profit domination in graphs, Preprint, 2005.
[94] J.A. Telle, Vertex partitioning problems: characterization, complexity and algorithms on
partial k -trees, PhD Thesis, University of Oregon, 1994.
[95] P.M. Weichsel, Large subgraphs of hypercubes, In Steiner Systems and Codes, preprint.
 Stephen T. Hedetniemi 37

[96] T.V. Wimer, S.T. Hedetniemi and R. Laskar, A methodology for constructing linear graph
algorithms, Proc. Sundance Conf., (Sundance, UT, 1985), Congr. Numer. 50(1985), 43-60.
[97] C.C. Yen and R.C.T. Lee, The weighted perfect domination problem, Inform. Process. Lett.,
35(1990),295-299.
[98] C.C. Yen and R.C.T. Lee, A linear algorithm to solve the weighted perfect domination problem
in series-parallel graphs, undated manuscript.
[99] M. Zito, Induced matchings in regular graphs and trees, WG 1999:89-100. Graph-theoretic
concepts in computer science (Ascona, 1999), Lecture Notes in Comput. Sci., 1665:89-100,
Springer, Berlin, 1999.
[100] M. Zito, Linear time maximum induced matchings algorithm for trees, Nordic J. Comput.,
7(1)(2000),58.