Domination Null and Packing Null Vertices

Peter Johnson and Robert Rubalcaba,

Auburn University,
and Matthew Walsh,
Indiana University/Purdue University
at Fort Wayne
Abstract
A vertex v of a graph G is domination null in G if g(v) = 0 for ev-
ery minimum fractional dominating function g on G. Packing nullity
is dened analogously, with reference to fractional closed neighbor-
hood packings. We give classes of examples and examine graph oper-
ations that produce or preserve such vertices; several open problems
are posed.
Key words and phrases: fractional domination, fractional packing,
linear programming duality
The open neighborhood in a graph G of a vertex v V (G) will be
denoted, as usual, by N
G
(v), and its closed neighborhood, v N
G
(v), by
N
G
[v]. If only one graph is under discussion, the subscript indicating the
graph will be omitted.
If g : V (G) R is a function, and S V (G), we will let g(S) stand
for

uS
g(u). A fractional dominating function on G is a function g :
V (G) [0, 1] satisfying g(N[v]) 1 for each v V (G); a fractional closed
neighborhood packing on, or of, G is a function h : V (G) [0, 1] satisfying
h(N[v]) 1 for each v V (G). We will call these fractional packings,
for short. The minimum value of g(V (G)), among fractional dominating
functions g on G, is the fractional domination number of G, denoted
f
(G).
The maximum value of h(V (G)), among fractional packings h of G, is the
fractional (closed neighborhood) packing number of G, denoted
f
(G). A
fractional dominating function g of G such that g(V (G)) =
f
(G) is a
minimum fractional dominating function on G, and a fractional packing h
of G such that h(V (G)) =
f
(G) is a maximum fractional packing of G.
It is well understood (see [1] or [3]) that the problems of nding
f
(G)
and
f
(G) are dual linear programming problems, and therefore
f
(G) =

f
(G). There is an important corollary of the linear programming duality
theorem called the principle of complementary slackness (see [5]) that we
will make much use of. Some applications of this principle to fractional
graph theory are explored in [4], of which this paper is a spin-o. We
will not explain the principle here, but only its application to fractional
domination and packing: if v V (G) and (i) g(v) > 0 for some minimum
dominating function g on G, then h(N[v]) = 1 for every maximum fractional
packing h of G; and (ii) if h(v) > 0 for some maximum fractional packing
of G, then g(N[v]) = 1 for every minimum fractional dominating function
on G.
Thus, the existence of a minimum fractional dominating function, or
a maximum fractional packing, taking a positive value at a vertex puts
a severe constraint on the behavior of the dual extreme functions in the
closed neighbor set of the vertex, and this greatly facilitates the study of
the convex set of all maximum fractional packings, or minimum fractional
dominating functions, on G. These sets, and their relationship, are the
objects of study of [4]; it seems to us reasonable to focus on bad vertices
that impede that study.
Denitions A vertex v V (G) is domination null if and only if g(v) = 0
for every minimum fractional dominating function g on G; v is packing null
if and only if h(v) = 0 for every maximum fractional packing h of G.
Observations and examples
The study of (fractional) domination and packing is quite mature, and
we have no doubt that every one of the mathematical facts to be presented
in items 1-6 below either appears elsewhere in some form, or is easily in-
ferrable from results appearing elsewhere, or is folkloric. See [2] and [3] for
a concentrated account of this area.
Still, although we may be exploring well-explored territory in 1-6, we
think that there may be some value in this exploration arising from the new
focus of our study (a geologist and a botanist will make dierent discoveries
in the same territory) and from our use of the principle of complementary
slackness. This principle is well known in linear programming, but we have
found no exploitation of it in fractional graph theory, before now.
1. If G is regular, say of degree k, then the constant function 1/(k + 1) on
V (G) is both a fractional packing and a fractional dominating function on
G, and is therefore a maximum fractional packing and a minimum fractional
dominating function on G. Therefore G has neither packing nor domination
null vertices.
Similarly, if Gis the complete r-partite graph with parts of sizes n
1
, . . . , n
r
,
r 2 and each n
j
2, then, as shown in [1], there is a weighting of the ver-
tices of G with positive weights ((n
j
1)
1
[

r
i=1
(n
i
1)
1
+r 1]
1
goes
onto each vertex in the jth part) which is both dominating and packing;
2
therefore there are no null vertices of either type in G.
2. If G, on n vertices, has one or more vertices of degree n 1, then

f
(G) =
f
(G) = 1 and no vertex in G is packing null. No vertex of G
of degree n 1 is domination null, while every vertex of smaller degree
is domination null. Therefore, the path P
3
with 3 vertices is the smallest
graph to have domination null vertices.
3. The smallest graph to have packing null vertices is P
4
. To see this,
observe that the weighting
0
y
1 1 0
u x v
is both dominating and packing; therefore
f
(P
4
) =
f
(P
4
) = 2. It is now
easy to see that no vertex of P
4
is domination null: The constant weight-
ing 1/2 is a minimum fractional dominating function. Further, because
the weight on the closed neighborhood of x, and of y, with this constant
weighting is 3/2 > 1, it follows from the principle of complementary slack-
ness that both vertices must be packing null. It then follows easily that the
fractional packing/dominating function given above is the only maximum
packing of P
4
.
4. The weighting below on P
5
is both
1/2 1/2 1/2 0 1/2
a fractional packing and a fractional dominating function. Therefore

f
(P
5
) =
f
(P
5
) = 2 and each vertex other than the central vertex is
neither packing nor domination null. By the principle of complementary
slackness we have the equations F(V (P
5
)) = 2 and F(N[v]) = 1 for each of
those four vertices v, for any F either a maximum packing or a minimum
dominating function on P
5
. It follows easily that both maximum packings
of and minimum dominating functions on P
5
are of the form
1 s , 1 t t s 0
with s +t 1 for domination and s +t 1 for packing. Therefore the
central vertex is both packing and domination null.
5. We state without proof (but with occasional helpful remarks) which
vertices in P
n
, n 6, are packing and/or domination null. Let the vertices
3
of P
n
, from one end of the path to the other, be u
1
, u
2
, . . . , u
n
. The proofs
of the following results are easy exercises analogous to the arguments in 3
and 4, especially 4: nd fractional packings and domination functions with
the same total weight, and apply the principle of complementary slackness
to the vertices which are not packing or domination null.
n = 3k, k 2 :
f
(P
n
) =
f
(P
n
) = k and there are no packing null vertices.
(The constant function 1/3 is a maximum fractional packing.) Applying the
principle of complementary slackness one nds that (0, 1, 0, 0, 1, 0, . . . , 0, 1, 0)
is the unique minimum fractional dominating function, and thus the dom-
ination null vertices are the vertices u
j
, j , 2 mod 3.
n = 3k + 1, k 2 :
f
(P
n
) =
f
(P
n
) = k + 1; there are no domina-
tion null vertices and the vertices u
j
, j 0 or 2 mod 3, and only those,
are packing null. [In fact, there is a unique maximum fractional packing,
(1, 0, 0, 1, 0, 0, . . . , 1, 0, 0, 1).]
n = 3k + 2, k 2 :
f
(P
n
) =
f
(P
n
) = k + 1; the vertices u
j
, j 1 or
2 mod 3 are neither packing nor domination null, while the vertices u
3j
,
j = 1, . . . , k, are both packing and domination null. [In fact, both the max-
imum fractional packing and the minimum dominating functions are of the
form (t
1
, 1 t
1
, 0, t
2
, 1 t
2
, 0, . . . , 0, t
k+1
, 1 t
k+1
), with 1 t
i
+ t
i+1
1,
i = 1, . . . , k, for domination, and 1t
i
+t
i+1
1, i = 1, . . . , k, for packing.]
6. The smallest graph that has both domination and packing null vertices,
and none that are both, is C
5
with a chord:
1
2
x y
w v
0 0
1
2
1
2
1
2
1
2
1
4
1
4
0
u
The vertex weightings shown, one dominating, the other a packing, show
that
f
(G) =
f
(G) = 3/2, and that the fractional dominating function
shown is minimum, and the fractional packing maximum. Since the closed
neighborhood weight of v and w for the dominating function above is 5/4 >
1, it follows that the principle of complementary slackness that v and w are
packing null; from the given fractional packing, it is clear that they are the
only packing null vertices. Similarly, since 1/2 < 1, u is domination null,
and is the only such vertex in G.
4
The principle of complementary slackness gives g(N[u]) = g(N[x]) =
g(N[y]) = 1 for each minimum dominating function g on G. It is easy to
see from this and g(V (G)) = 3/2, and g(u) = 0, that all minimum fractional
dominating functions on G are of the following form:
1
2
t
1
2
1
2
t
0
A dual analysis shows that the fractional packing given previously is the
only maximum fractional packing of G.
There is something of general interest in this small example, other than
the presence of dierent kinds of null vertices, which relates to the possi-
bility of a converse, or strong converse, of the principle of complementary
slackness, as applied to fractional domination and packing. If v V (G)
is not packing null, then for every minimum fractional dominating func-
tion g on G, g(N[v]) = 1. If v V (G) is packing null in G, does there
necessarily exist a minimum fractional dominating function g on G such
that g(N[v]) > 1? Dually, if u is domination null in G, does there neces-
sarily exist a maximum fractional packing h of G such that h(N[u]) < 1?
If there is such a g for v, let us say that v is normally packing null, and
if g(N[v]) > 1 for every minimum fractional dominating function g on G,
we will say that v is strongly packing null. Normal and strong domination
nullity are dened dually.
All of the domination and/or packing null vertices in 1-5 are normal,
but not strong. In 6, with G = C
5
-with-a-chord, v and w are normally
(but not strongly) packing-null, while u is strongly domination-null. In 7,
8, and 9, below, we will see some more strongly packing null vertices, and
some more strongly domination null vertices. We do not yet have examples
of abnormally null vertices of either type. (An abnormally domination null
vertex v is domination null and h(N[v]) = 1 for every maximum fractional
packing h; abnormally packing null is dened dually.)
7. Suppose that m 3 and 1 < k < m. We form a graph G(k, m) = G
as follows: Let K be a clique on m vertices and let E = v
S
; S

V (K)
k

.
[Here,

V (K)
k

stands for the collection of k-subsets of V (K).] It is to be

5
understood that v
S1
,= v
S2
for S
1
,= S
2
, and that E is disjoint from V (K).
Form G, with vertex set V (K) E, with K as an induced subgraph, by
joining each v
S
to each vertex of S, for each S

V (K)
k

, and adding no
other edges.
If g(u) = 1/k for each u V (K) and g 0 on E, then g is a fractional
dominating function on G, with g(V (G)) = m/k; and if h(v) =

m1
k1

1
for each v E and h 0 on V (K), then h is a fractional packing of G with
h(V (G)) =

m1
k1

m
k

=
m
k
. Therefore
f
(G) =
f
(G) = m/k and g is
a minimum fractional dominating function, and h is a maximum fractional
packing.
Since g(N[u]) = m/k > 1 for each u V (k), the principle of comple-
mentary slackness implies that each u V (k) is packing null, and h shows
that no vertex of E is packing null. Similarly, g and h and the fact that
h(N[v]) =

m1
k1

1
< 1 for all v E shows that E is the set of domination
null vertices. We shall next show that g is the unique minimum fractional
dominating function on G, and thus that the vertices of V (K) are strongly
packing null.
Suppose that is some minimum fractional dominating function on G.
By the paragraph preceding, no vertex in E is packing null in G, so, by the
principle of complementary slackness, (v
S
) + (S) = 1 for each k-subset
S of V (K). But by remarks just above (v
S
) = 0 for each S. Therefore
(S) = 1 for each S

V (K)
k

; since k < m, it is easy to see that must be

constant on V (K). Since m/k = (V (G)) = (V (K)), the constant must
be 1/k.
The story of the set of domination null vertices, E, is a bit dierent.
We can prove: Each vertex of E is strongly domination null if and only if
k > m/2.
To see this, rst suppose that k > m/2. If, for some S

V (K)
k

, v
S
is not
strongly domination null, then for some maximum fractional packing of
G, 1 = (N[v
S
]) = (v
S
) + (S) = (v
S
) because vertices of S V (K)
are packing null. Now, because the vertices of V (K) are not domination-
null, (N[u]) = 1 for each u V (K). It then follows that for S
1

V (K)
k

,
S
1
,= S, S
1
S ,= , it must be that (v
S1
) = 0. But k > m/2 implies that
every S
1

V (K)
k

intersects S; thus (V (G)) = (E) = (v

S
) = 1, not
m/k, contradicting the assumption that is a maximum fractional packing
of G. Thus every v
S
E is strongly domination-null, after all.
On the other hand, suppose that k m/2. We will show that no v
S
E
is strongly domination-null (although they are clearly normal). Dene
by 0 on V (K) and on v
S1
E v
S
[S
1
S ,= , (v
S
) = 1, and,
for

S

V (K)
k

,

S S = , set (v

S
) =

mk1
k1

1
. It is straightforward to
see that is a fractional packing and that (N[v
S
]) = (v
S
) = 1. We have
6
(V (G)) = (v
S
)+

S|

SS=}
(v

S
) = 1+

mk
k

mk1
k1

1
= 1+
mk
k
=
m
k
, so is a maximum fractional packing, and we have shown that v
S
is
not strongly domination null, because (N[v
S
]) = 1.
We close this section (item 7) by showing that there is a unique maxi-
mum fractional packing of G if and only if k = m1. Let A be an m

m
k

zero-one matrix with rows indexed by the vertices of K and columns in-
dexed by the k-subsets of V (K), with entry A(v, S) = 1 if and only if v S.
It is easy to see that the rank of A is m, if 1 < k < m, but all we will need
here is that rk(A) m, with equality if k = m1.
Since the vertices of K are packing but not domination null, h(N[u]) =

{S|uS}
h(v
S
) = 1 for each u V (K), by the principle of complemen-
tary slackness, for every maximum fractional packing h of G. Further, it
is straightforward to see that for any function h : E [0, 1] satisfying

{S|uS}
h(v
S
) = 1 for all u V (K), h(E) =
m
k
, and so extending h by
setting h 0 on V (K) results in a maximum fractional packing of G.
Therefore, the maximum fractional packings of G correspond to vectors
(x
S
; S

V (K)
k

) = x of non-negative numbers such that A x =

1
.
.
.
1

=

1.
If k = m 1 then

m
k

= m, and A can be the m m matrix with zeroes

down the main diagonal and ones o the main diagonal, which has rank m,
so the equation A x =

1 has a unique solution,

(m1)
1
.
.
.
(m1)
1

. Otherwise,
if 1 < k < m1, then rkA m <

m
k

, and the equation A x =

1 does not
have a unique non-negative solution; just add a small vector in the kernel of
A to the constant vector (

m1
k1

1
), to get another non-negative solution.
8. The graph of item 7 makes sense with k = 1, but the story is quite
dierent in that case;
f
(G) =
f
(G) = m, but now there are no domination
null vertices. For each u V (K), assign some t(u) [0, 1] to u and 1t(u)
to v
{u}
; this gives a minimum fractional dominating function, and they are
clearly all like that.
Applying the principle of complementary slackness, which says here that
h(N[w]) = 1 for every w V (G) and every maximum fractional packing h
of G, one nds that there is a unique maximum fractional packing h of G:
h(v) = 1 for every v E and h(u) = 0 for every u V (K). [Recall that
m 3. If m = 2, G = P
4
and it is still true that the maximum fractional
packing is unique.] Thus the set of packing null vertices is V (K), and they
7
are all normal, but not strong.
9. Let H be a graph with a Hamiltonian cycle C
m
= C. Form G by adding
a new vertex of degree 2, v(e), for each edge e of C, with v(e) adjacent to
each end vertex of e.
Picture:
u
1
u
2
u
3
u
m1
u
m
v
m1
v
2
v
m
v
1
H
Assigning 1/2 to the v
j
and 0 to the u
i
gives a fractional packing of G
with total weight m/2. Assigning 1/2 to the u
i
and 0 to the v
j
gives
a fractional dominating function on G with total weight m/2. Therefore

f
(G) =
f
(G) = m/2, and the u
i
are not domination null, nor are the
v
j
packing null, in G. Further, because g(N[u
i
]) 3/2 > 1 for each
i, with g denoting the fractional dominating function just described, and
h(N[v
j
]) = 1/2 < 1 for each j, with h denoting the fractional packing just
described, it follows that the u
i
are packing null, and the v
j
are domination
null, in G.
The principle of complementary slackness applied to the v
i
now says that
if a minimum fractional dominating function assigns x
i
to u
i
, i = 1, . . . , m
(and 0 to each v
j
, necessarily) then we have x
i
+ x
i+1
= 1, i = 1, . . . , m.
If m is even, this says that the minimum fractional dominating functions
on G are of this form: around the cycle, assign t, 1 t, successively to
the vertices, and assign 0 to the v
j
. If H = C
m
, this implies that the u
i
are normally, but not strongly, packing null, if m is even (consider the two
weightings taking t = 0 and t = 1). This conclusion also holds if u
i
and u
j
,
[i j[ , = 1 mod m, are adjacent in H only if i j mod 2, with m even.
Otherwise, some of the u
i
will be strongly packing null in G.
If m is odd, the equations x
i
+ x
i+1
= 1, i = 1, . . . , m have a unique
solution, x
1
= = x
m
= 1/2. So there is a unique minimum fractional
dominating function in this case, and each u
j
is strongly packing null.
A similar application of the principle of complementary slackness shows
8
that the v
j
are normally but not strongly domination null if m is even, and
strongly domination null if m is odd.
10. See [2] for the denitions of dominating and (closed neighborhood)
packing sets of vertices in a graph, and the domination number and
the packing number . It is well known and elementary that, for every
graph G, (G)
f
(G) =
f
(G) (G). Let us say that a vertex is
integrally domination null in G if it belongs to no minimum dominating
set of vertices in G (a dominating set S V (G) such that [S[ = (G)),
and integrally packing null if it belongs to no maximum packing set in G
(a closed neighborhood packing S V (G) such that [S[ = (G)).
If (G) =
f
(G), then every domination null vertex is integrally domi-
nation null, because then the characteristic function of any minimum dom-
inating set in G is a minimum fractional dominating function, and so the
value of that function at the vertex must be zero i.e., the vertex does
not belong to the set. Similarly, if (G) =
f
(G), then every packing null
vertex in G is integrally packing null. But what if
f
(G) < (G), or if
(G) <
f
(G)? And what about the converse inclusions i.e., is every in-
tegrally domination (packing) null vertex necessarily domination (packing)
null?
We can answer the rst two questions by looking back to two of the
examples preceding. In 6, where G is C
5
with a chord, it is easy to see that
(G) = 1, and therefore no vertex is integrally packing null, whereas there
are two packing null vertices there. Also, (G) = 2, and it is easy to see
that every vertex of G is in some minimum dominating set, so there are no
integrally domination null vertices, while there certainly is one domination
null vertex.
In 9, with m odd, we leave it as an exercise to verify that (G) =
m1
2
,
(G) =
m+1
2
, and there are no integrally domination null vertices in G.
Further, if H = C
m
, then there are no packing null vertices in G.
The lesson of these examples seems to be that it is harder for a vertex to
be integrally null than just plain null, but we can show that this is just an
occasional phenomenon and simultaneously answer the other two questions
above by looking at the famed triple-yswatter, TF:
9
v
x
2
x
1
u
x
6 x
3
x
4 x
5
w
TF is regular of degree 3, so
f
(TF) =
f
(TF) = 16/4 = 4, and TF has
no domination nor packing null vertices (see Observation 1, above). But
TF has plenty of vertices that are both integrally domination and packing
null. We leave it as an exercise to verify that (G) = (G) = 4, and that
u, v, w, and x
1
, . . . , x
6
are integrally domination and packing null.
11. In the last of this string of observations and examples we restate in
simplied form two of the broad results that are given in greater detail
in [4]. Let D(G) denote the set of all minimum fractional dominating
functions on G, and let P(G) denote the set of all maximum fractional
packings of G. If G has no packing null vertices then by the principle of
complementary slackness, for each g D(G) and v V (G), g(N[v]) =
1; thus D(G) P(G). Similarly, if G has no domination null vertices,
P(G) D(G). Therefore, if G has neither packing nor domination null
vertices, as is the case with regular graphs, or complete r-partite graphs
with no part consisting of one vertex, then D(G) = P(G). [In [4] there is
a quick proof of what may be a folkloric result, that if G = K
n1,...,nr
, n
i

2, i = 1, . . . , r, r 2, then D(G) = P(G) = f, the function mentioned in
item 1.]
10
The remarks above can be ostensibly strengthened: recollect from the
discussion in item 6 that an abnormal packing-null vertex v would satisfy
g(N[v]) = 1 for every g D(G). Thus, if G has packing null vertices, but
only abnormal ones, then D(G) P(G), and, dually, if G has only abnor-
mal domination null vertices, then P(G) D(G). But these strengthenings
may be vacuous, as we have not yet discovered any abnormal null vertices
of either type.
Results
As usual, if S V (G), N(S) =

uS
N(u); and S) or S)
G
will denote
the subgraph of G induced by S.
If u V (G) and H is a graph, the explosion of u into H is the graph

G =

G(u, H) obtained by replacing u in G by a copy of H, every vertex of
which is adjacent to every vertex of N
G
(u). For example, if G is P
4
, u is
one of the vertices of G of degree 2, and H = K
2
, the empty graph on two
vertices, then

G(u, H) is C
4
with a pendant vertex:
u
H G

G(u, H)
Theorem 1 For any G, u V (G), and H, if

G =

G(u, H), then
f
(G)

f
(

G) with equality if either (a) H is a clique, or (b) u is domination null,
or (c) N
G
(u))
G
is a clique. Further, if u is not an isolated vertex of G,
then
f
(

G) <
f
(G) + 1.
Proof. Let h be a maximum fractional packing of G; then

h, dened by
distributing the weight h(u) over the vertices of the copy of H replacing u,
and setting

h = h on V (G) u, is a packing of G with the same total
weight as h. Thus
f
(G) =
f
(G) =

h(V (

G))
f
(

G) =
f
(

G).
Suppose that u is not an isolated vertex of G. Suppose that

h is a
maximum fractional packing of

G. Let v be a neighbor of u in G. Then

h(v) +

h(V (H))

h(N

G
[v]) 1, so

h(V (H)) 1

h(v). If

h(v) = 0 for
every v N
G
(u) then the function h : V (G) [0, ) dened by

h = h
on V (G) u and h(u) =

h(V (H)) is a fractional packing of G, since
then h(N
G
[v]) =

h(N

G
[v]) 1 for all v V (G) u and h(N
G
[u]) =

h(V (H)) 1. It would then follow that

f
(G) =
f
(G)
f
(

G), and thus
that
f
(G) =
f
(

G).
11
Otherwise, we have that

h(V (H)) < 1. The function h dened above, by
concentrating

h(V (H)) at u, may fail to be a fractional packing only because
h(N
G
[u]) =

h(V (H)) +

h(N
G
(u)) might be > 1. Assume so. For any
w V (H),

h(w)+

h(N
G
(u))

h(N

G
[w]) 1, and, from above,

h(V (H)) =
h(u) < 1. Therefore, h(N
G
[u]) = h(u) +

h(N
G
(u)) < 2. Consequently, the
function

h dened by

h(u) = 1

h(N
G
(u)) and

h = h =

h on V (G) u is
a fractional packing of G, because 0

h(u) = 1

h(N
G
(u)) <

h(V (H)) =
h(u), which implies that

h(N
G
[w]) h(N
G
[w]) 1 for all w V (G) u,
and

h(N
G
[u]) = 1. We have

h(V (

G))

h(V (G)) = h(u)(1h(N

G
(u))) =
h(N
G
[u])1 < 21 = 1. Thus
f
(

G) =
f
(

G) =

h(V (

G)) <

h(V (G))+1

f
(G) + 1 =
f
(G) + 1. If h(N
G
[u]) 1, by an argument given earlier,

f
(

G) =
f
(G). Thus, in any case, if u is not an isolated vertex of G then

f
(

G) <
f
(G) + 1.
Suppose that H is a clique. Suppose that

h and h are as above. For
any w V (H), h(N
G
[u]) =

h(N

G
[w]) 1, so h is a fractional packing of
G and it follows, as above, that
f
(G) =
f
(

G).
Suppose that u is domination null in G. Let g be a minimum fractional
dominating function on G, so g(u) = 0; dene g on

G by g 0 on V (H)
and g g on V (G) u. Then g is a fractional dominating function on

G,
so
f
(

G) g(V (

G)) = g(V (G)) =
f
(G). Since the inequality the other
way has already been established, the desired conclusion,
f
(

G) =
f
(G),
follows.
Suppose that N
G
(u))
G
is a clique. Let

h be a maximum fractional
packing of

G, and let h be dened on G by concentrating

h(V (H)) at u, as
above. For any v N
G
(u), h(N
G
[u]) =

h(N
G
(u))+

h(V (H))

h(N
G
[v])
1, so h is a fractional packing of G. As above, it follows that
f
(G) =
f
(

G).

Corollary 1 Suppose that S V (G) satises: for every v, w S, N

G
(v)
S = N
G
(w) S. Let G
o
be the graph obtained by collapsing S to a single
vertex u whose neighbors in G
o
are the vertices of N(S) S, with GS =
G
o
u. Then
f
(G
o
)
f
(G), with equality if either S)
G
is a clique or
u is domination null in G
o
or N(S) S)
G
is a clique. Further, if S)
G
is
not a union of connected components of G, then
f
(G) <
f
(G
o
) + 1.
Proof. The claims of the corollary follow from the theorem with G
o
playing
the role of G, G playing the role of

G =

G(u, H), and S)
G
playing the role
of H.
Theorem 2 (a) Suppose that u is domination null in G. Then for any H,
every vertex of V (H) is domination null in

G =

G(u, H).
12
(b) Suppose that S V (G) is a set of packing null vertices in G satisfy-
ing the hypothesis of Corollary 1, and let u and G
o
be as in that corollary.
Then u is packing null in G
o
.
Proof. (a): Supose that g is a minimum fractional dominating function
on

G, and let g be dened by g g on V (G) u, and g(u) = g(V (H)).
Then g is dominating on G, and g(V (G)) = g(V (

G)) =
f
(

G) =
f
(G),
applying Theorem 1. Thus g is a minimum fractional dominating function
on G. Therefore g(u) = 0 = g(V (H)); therefore g(w) = 0 for all w V (H);
g was arbitrary, so every w V (H) is domination null in

G.
(b): First we show that
f
(G) =
f
(G
o
). By Corollary 1,
f
(G
o
)

f
(G). On the other hand, if h is a maximum fractional packing of G, then
h(S) = 0; dene h
o
on G
o
by h
o
(u) = 0 and h
o
= h on V (G
o
) u =
V (G) S. Then h
o
is a fractional packing of G
o
, so
f
(G) = h(V (G)) =
h
o
(V (G
o
))
f
(G
o
). Thus
f
(G) =
f
(G
o
).
Now, if h
o
is some maximum fractional packing of G
o
, let h be dened
by distributing h
o
(u) any old way over S and setting h h
0
on V (G) S =
V (G
o
) u. Then h is a fractional packing of G with
f
(G) =
f
(G
o
) =
h
o
(V (G
o
)) = h(V (G)). Thus h is a maximum fractional packing of G.
Therefore h(S) = h
o
(u) = 0; h
o
was arbitrary, so u is packing null in G
o
.

For a graph G on a n vertices, let

drat(G) = (number of domination null vertices of G) /n,
prat(G) = (number of packing null vertices of G) /n, and
brat(G) = (number of vertices both packing and domination null) /n.
For example, if G is the graph in item 6 above, C
5
with a chord, then
drat(G) = 1/5, prat(G) = 2/5, and brat(G) = 0.
Corollary 2 drat(G)[G is a connected simple graph contains numbers
arbitrarily close to 1, and positive numbers arbitrarily close to zero.
Proof. Take G
o
with a domination null vertex and explode it into an
arbitrarily large graph H, to get a graph G with drat(G) arbitrarily close
to 1. If G
o
was connected, G will also be connected.
On the other side, let G be K
n
minus an edge. Then
f
(G) =
f
(G) = 1
and, the set of domination null vertices consists of the two vertices of degree
n2. (This is a consequence of a more general observation in item 2 above.)
Thus, drat(G) = 2/n.
Problems
1. Do there exist graphs with abnormally domination null vertices? (Recol-
lect from 6, this would be a domination null vertex u such that h(N[u]) = 1
13
for every maximum fractional packing h of G.) We also ask the dual ques-
tion for packing null vertices and, should the answers to those two questions
be yes, we ask: are there any graphs with a vertex which is both abnor-
mally packing and abnormally domination null?
2. We vaguely wonder which graphs have the property that their vertex set
can be partitioned into domination and packing null vertices, as in examples
7 and 9 above. By Theorem 2 we can get loads of examples by exploding
domination null vertices in such graphs, so our question should really be
about such graphs not obtained from smaller graphs by such explosion.
In such a minimal graph, is the set of domination null vertices necessarily
independent? Minimal or not, in such a graph, is the set of packing null
vertices necessarily dominating (in the usual, integer sense see [2])? Is
there any such graph with a vertex which is both packing and domination
null?
3. To our amazement, we have not found graphs G and H, and a non-
isolated vertex u in G, such that
f
(G) <
f
(

G(u, H)). We are beginning
to think that something beyond our feeble comprehension is going on, and
that the main claim in Theorem 1 is just the easy part of a much stronger
statement. Here, we simply ask: do such G, H, and u exist? [If not,
then, in Corollary 1, we conclude
f
(G
o
) =
f
(G) if S)
G
is not a union of
connected components of G, i.e., if u is not isolated in G
o
.]
4. Regarding Theorem 2: can graphs G and H, and a packing null vertex
u in G, be found so that not all the vertices of V (H) are packing null in

G(u, H)? Dually, can a set S V (G) of domination null vertices, for some
G, be found, satisfying the hypothesis of Corollary 1, such that if u and G
o
are as in Theorem 2 (b), then u is not domination null in G
o
?
5. Describe the sets
Drat = drat (G)[G is a connected simple graph ,
Prat = prat (G)[G is a connected simple graph , and
Brat = brat (G)[G is a connected simple graph .
We are fairly condent that Drat will turn out to be the set of all rational
numbers in [0, 1), and that Brat will not, but dont know about Prat. (If
Prat turns out not to contain rationals arbitrarily close to 1, then the answer
to the rst question in problem 4 must be yes.) And we wonder what
Brat will look like; from observation 5, Brat has 1/3 as an accumulation
point. Is this Brats only non-zero accumulation point, and is 1/3 Brat?
6. The strong direct product of two graphs G and H, denoted here by GH,
14
is dened by: V (GH) = V (G) V (H) and (u
1
, v
1
), (u
2
, v
2
) are adjacent
in G H if and only if either u
1
= u
2
and v
1
, v
2
are adjacent in H, or
v
1
= v
2
and u
1
, u
2
are adjacent in G, or both u
1
and u
2
, and v
1
and v
2
, are
adjacent, in G and H, respectively.
It is proven in [5], in the proof of Theorem 7.4.9, that
f
(G H) =

f
(G)
f
(H). Suppose that u is domination (packing) null in G, and that v
is domination (packing) null in H. Is (u, v) necessarily domination (pack-
ing) null in GH?
References
[1] D. L. Grinstead and P. J. Slater, Fractional domination and fractional
packing in a graph, Congressus Numerantium 71 (1990), 153-172.
[2] T. W. Haynes, S. T. Hedetniemi, and P. J. Slater, Fundamentals of
Domination in Graphs, Marcel Dekker, Inc., New York, 1998.
[3] T. W. Haynes, S. T. Hedetniemi, and P. J. Slater, eds. Domination in
Graphs: Advanced Topics, Marcel Dekker, Inc., New York, 1998.
[4] Robert Rubalcaba and Matthew Walsh, Minimum dominating and
maximum packing functions on graphs, submitted.
[5] E. R. Scheinerman and D. H. Ullman, Fractional Graph Theory, Wiley-
Interscience, New York, 1997.
15