For graphs and , the proper Ramsey Number is the smallest integer so that any -edge-coloring on contains either a monochrome or a properly colored . We determine the proper Ramsey number of against and .
Keywords: Graph theory, Edge coloring, Ramsey theory, Proper Ramsey number
1. Introduction
In this paper, we give values for the proper Ramsey number of against and . Let and be graphs and the chromatic index of . The proper Ramsey number,
, of and is defined to be the smallest positive
integer so that every -edge coloring of contains either a monochrome or a properly colored . Necessarily, if , then there exists at least
one coloring of which
contains neither. Such a coloring is called a critical coloring
with respect to .
English et al. [1] determined for a variety of graph pairs in
which , including
for against . Extending that work,
Olejniczak [2]
found the remaining values of where is even. The problem we address here
concerns a pair of graphs for which , namely to determine
and .
Note that the proper Ramsey number is one of several variations on
the ordinary Ramsey number. It is closely related to the
edge-chromatic Ramsey number introduced by Eroh [3]. The edge-chromatic
Ramsey number is defined to
be the the smallest positive integer so that if the edges of are colored with any number of
colors, then the colored graph contains either a monochrome or a properly colored . Eroh [3] shows that this number exists only for
certain classes of graph pairs. The proper Ramsey number, on the other
hand, is a restriction of edge-chromatic Ramsey number, in the sense
that we color with a fixed
number of colors. Olejniczak [2] shows that the proper Ramsey number
exists for all pairs of graphs
and . Further variations on the
Ramsey number are discussed in Chartrand and Zhang [4].
In the following, graphs are assumed to be finite, simple, and
undirected. We refer to Chartrand and Zhang [5] for our graph-theoretic
notation.
2. Essential Preliminaries
At several points in this paper, we shall need to select a vertex of
a graph so that it is incident to the greatest number of edges of the
same color. We therefore adopt the following definition.
Definition 1. Let be a graph, and let be the indexed family of colors
of edges in . If is a vertex of , then the color degree of with respect to the color is the number of edges incident
to that have the color . We denote the color degree by .
Note that there must be some
and some for which for all and for every vertex in . That is, there is some vertex incident
to the greatest number of like-colored edges. The proof is trivial. Let
be
the set of all color degrees of all vertices in . Then is a set of integers bounded above by
. It therefore has a
greatest element. We denote this greatest element and call it the maximal
color degree.
We shall also require the following facts about and
Lemma 1. Every -edge-coloring of that does not contain a monochrome
contains a properly colored
beginning with each color.
Proof. Let be a -edge-colored copy of containing no monochrome . Consider an arbitrary vertex of . Without loss of generality, suppose
that is incident to more red
edges than blue edges. Then is
incident to either three or two red edges.
Suppose that has three red
edges. Then must contain either a
red or a blue . Let , , and be the three red edges. Then and must be blue. But then produces a monochrome regardless of whether it is red or
blue. So must not have three red
edges.
Suppose that has two red
edges. Let and be red and blue. Then must be blue to avoid a red . So is a properly colored beginning with a blue edge. On the
other hand, and cannot both be blue, since a blue
would result. So one is red.
Thus one of or is a properly colored beginning with a red edge.
Lemma 2. Any -edge-coloring of not containing a monochrome decomposes into two Hamiltonian cycles of
different colors.
Proof. Let be a -edge-colored copy of . We first show that for every color and vertex
in . Suppose otherwise. Then for some and . Without loss of generality, suppose
that this vertex is labeled and
the color is red. Then there are at least three vertices, label them
, , and , adjacent to through a red edge. If any of the edges
, , or is red, a monochrome results. So none of them are red. But
then they must all be blue, and again a monochrome results. Thus each monochrome
subgraph of is -regular.
Consider then, the red subgraph of . This subgraph must be , since there is no other -regular graph on vertices. The same argument applies to
the blue subgraph. Since these subgraphs have disjoint edge sets and
identical vertex sets, they decompose into two Hamiltonian cycles of
different colors.
3. Results
Theorem 1. .
Proof. We first show that . Consider the
following coloring of .
Consider as . Decompose each subgraph into two Hamiltonian cycles;
in each subgraph, color one
cycle red and the other blue. Color the remaining edges (those between
the sugraphs) green. This
coloring contains no monochrome . The red and blue subgraphs are each
copies of , which do not contain
. The green sugraph is , which contains no odd cycles by
virtue of being bipartite. Thus, the green subgraph contains no
Furthermore, this coloring contains no proper . Consider any 3-cycle in this graph.
Either all of the vertices lie in one subgraph, or two lie in one and the third lies in the other. In
the first case, all of the edges of the cycle are either red or blue, so
it cannot be properly colored. In the other, two adjacent edges are
green, so it cannot be properly colored. This coloring is illustrated in
Figure 1.
Figure 1. This coloring on shows that .
We now show that . The proof is by contradiction. Let be a -edge-colored copy of containing no monochrome and no proper . We show that this assumption leads
to a contradiction. Let be a
vertex of maximal color degree, and , , and the subgraphs of so that each is joined to entirely edges of a single color, say
by red edges, by blue, and by green. Since contains no monochrome , the edges within must each be of a color different
than that of the edges joining
to . Since contains no proper , the edges between and must each be one of two colors,
specifically those of the edges joining and to . For instance, the edges between and must all be either red or blue, since
coloring any such edge green yields a proper .
We will suppose that , which can always be accomplished by a relabeling of
colors. Note that . It follows that the possible orders of these
subgraphs correspond to the three-part integer partitions of 10. We
consider these possibilities in order of decreasing .
Case 1. Suppose that . Then contains a 2-edge-colored . It follows that must contain a monochrome , since . But we assumed otherwise,
so this is a contradiction.
Case 2. Suppose that . Then either or . Without loss of generality,
suppose that . Since
is a -edge-colored copy of containing no monochrome , it can only be colored in one way.
By Lemma 2, has one blue Hamiltonian cycle and
one green Hamiltonian cycle. Similarly, must contain a proper , and so it has at least one red edge
adjacent to a green edge. From the discussion above, all the edges
between and are either red or blue.
Let
so that is the blue
Hamiltonian cycle and
is the green Hamiltonian cycle, and let , , and belong to so that is red and green. Consider the set of edges from
to . If any three of them are blue, then a
blue results, for in any set of
three vertices in , two of
them are adjacent through a blue edge. Thus, at least three of the edges
from to are red. For the same reason, at least
three of the edges from to
are red. But this means that
there exists at least one vertex of with red edges to both and , yielding a red , so this is impossible.
Case 3. Suppose that . There are two
possibilities.
Subcase 3.1.. Suppose . By Lemma 1, contains both a blue-green-blue and a
green-blue-green . We will first
examine the subgraph . Let
so that and are green and is blue. Let so that is red and is green. Recall that all the edges
between and are either red or blue. Consider the
set of edges . At least
one of these edges is red, otherwise is a blue . (Recall that is the subgraph of induced by the set of vertices .) Without loss of generality, suppose
that is red. Then we must have
the following:
must be red,
otherwise is a proper
.
must be red,
otherwise is a proper
.
must be red,
otherwise is a proper
.
must be blue,
otherwise is a red
.
must be blue,
otherwise is a proper
.
must be red,
otherwise is a blue
.
must be red,
otherwise is a proper
.
must be blue,
otherwise is a a red
.
must be blue,
otherwise is a proper
.
must be blue,
otherwise is a proper
.
must be blue,
otherwise is a proper
.
It then follows that the remaining edges of must all be blue to avoid proper
-cycles in , , and . The resulting graph is shown
in Figure 1. Thus, has four blue edges and two green
edges.
Figure 2. Coloring of when , and edge is red. Solid edges are red, dashed
edges blue, and bold edges green. Note the number of edges of each color
in .
Next consider the subgraph . Relabel the vertices of so that , and are blue, and is green. (We are guaranteed the
existence of such a path by Lemma 1.) Let
so that is red and is blue. Note that the starting
conditions here are isomorphic to those in the case of through the permutation of the
colors blue and green. It follows that must be colored in the same way as with the colors blue and green
swapped. Hence, must have four
green edges and two blue edges.
But then must have
both four green edges and four blue edges, a contradiction since has only six edges.
Subcase 3.2.. Suppose that and . We may assume that every
vertex has a similar distribution of edges. That is, we may assume that
the edges out of each vertex fall into color classes of size 4, 4, and
2, since otherwise we could apply a previous case by making a different
choice of .
Let
so that and are green and blue. Let so that and are green and red. We are guaranteed the existence
of these paths by Lemma 1.
Lastly, let . As in
the previous subcase, we will begin with one edge and show that the
choice of color for that edge fully determines the rest of the
coloring.
Consider the set of edges . At least one of these edges must be red, since
otherwise a blue results. This
scenario is entirely symmetrical with respect to these edges, so we are
free to suppose that the edge is
red. We must then have the following:
must be blue,
otherwise is a red
.
must be red,
otherwise is a proper
.
must be red,
otherwise is a proper
.
must be red,
otherwise is a proper
.
must be blue,
otherwise is a red
.
must be red,
otherwise is a blue
.
must be blue,
otherwise is a red
.
must be red,
otherwise is a proper
.
must be red,
otherwise is a proper
.
must be red,
otherwise is a proper
.
must be blue,
otherwise is a proper
.
must be blue,
otherwise is a proper
.
must be blue,
otherwise is a proper
.
must be blue,
otherwise is a proper
.
must be blue,
otherwise is a proper
.
It then follows that the remaining edges within are all blue to avoid proper -cycles in , , and . Similarly, the remaining edges
within are all red to avoid
proper -cycles in , , and . Thus, must be colored as shown in
Figure 3 in order to avoid both monochrome and proper .
Figure 3. Coloring of when and edge is red. Solid edges are red, dashed
edges blue, and bold edges green.
Since the remaining edges out of can only be red or green, each vertex
of must have one red edge and
one green edge into , otherwise
the color classes would not have the correct number of edges. Similarly,
each vertex of must have one
blue edge and one green edge into . (Recall, we may assume that every
vertex has two edges of one color, and four each of the two remaining
colors.) Specifically, exactly one of the edges in must be red, and the other
green. Similarly, exactly one of the edges in must be blue and the other
green.
Finally, consider the edge in . This edge cannot be blue, for then
would be a proper . Neither can it be red, for then
would be a proper . This is a
contradiction.
We have shown that cannot
take on any value between and
. But since it must take on one
of these values, we have arrived at a contradiction. Therefore .
We conclude that .
We now move on to the result for .
Theorem 2. .
Proof. We first exhibit a critical coloring on that contains neither a monochrome
nor a properly colored . To construct this coloring, consider
as . Decompose the subgraph into two Hamiltonian cycles,
and color one cycle red and the other blue. Color the remaining edges
(those between the and subgraphs) green. This coloring
clearly contains no monochrome :
the red and blue subgraphs are each copies of , while the green subgraph is the star
. On the other hand, any
5-cycle in the graph either contains the vertex or it does not. If it does,
then it must contain adjacent green edges. If it does not, then it
contains only red and blue edges. But cannot be properly colored with only
two colors, so the graph contains no proper . Thus .
Figure 4. This coloring on shows that .
We next show that . Let be a -edge-colored copy of containing neither a monochrome nor a properly colored . We show that this assumption
produces a contradiction. Let be
a vertex of with maximal color
degree, and let , , and be the subgraphs of so that each is joined to entirely by edges of the same color. We
will assume that and that , , and are joined to by red, blue, and green edges,
respectively. These assumumptions can can always be satisfied by a
relabeling of colors. Since , the
possible orders of these subgraphs must correspond to the three-part
integer partitions of 6. We consider the possible cases in order of
decreasing .
Case 1. Suppose that . Then contains a 2-edge-colored . It follows that must contain a monochrome , since . This is a
contradiction.
Case 2. Suppose that . Then, without loss of
generality, . By Lemma 2 the subgraph must be decomposed into Hamiltonian
cycles of different colors (blue and green) to avoid a monochrome . Let so that the blue
cycle is and the
green cycle . Let
.
Note that for each vertex in , there is a proper in beginning at and ending on that vertex, whose final
edge is blue. In alphabetical order of final vertex, , , , , and are proper such paths.
Then the edge must be
blue to avoid a proper ;
choosing red or green yields a proper in . But the same consideration
applies to every vertex of , so
every edge out of into must be blue. But this yields a
monochrome , such as , so this is impossible. This
case is illustrated in Figure 5.
Figure 5. The coloring that results if in the proof of Theorem 1. This clearly contains a monchrome
, contrary to our
assumption.
Case 3. Suppose that . Then there are two
possibilities.
Subcase 3.1. . Suppose that . By Lemma 1, contains a proper that begins and ends with blue edges.
Let , so that
is the proper , and . Note that is not blue, for then is a blue .
Consider the edges in . If any one of these
edges is not blue, then a proper results. Note that these are
symmetrically situated, so we may consider just one. Suppose that were not blue. Then would be a proper . So indeed, each of the four edges
indicated must be blue. But then
must be green to avoid a blue
in . But now a proper
cannot be avoided, since is a proper . So this subcase is
impossible.
Subcase 3.2.. Suppose that . Again, by
Lemma 1, contains a proper that begins and ends with blue edges.
Let , so that
is the proper , and let and . Consider the edges and . If either of these edges is not
green, then a proper results,
namely either of or
. But they also cannot
both be green, since then is a green . So this subcase is also
impossible.
Case 4. Suppose that . There are two possibilites.
Either or and .
Subcase 4.1.. Suppose . Let and . Note that and are -edge-colored copies of not containing a monochrome , so each contains a proper . We may assume that is green and blue. Then must be blue, otherwise is a proper . But note that is a proper with no blue edges. It follows that
is a proper . So cannot be blue. Hence, must be both blue and not blue, a
contradiction.
Subcase 4.2.. Suppose that and . Let and . Note that must contain a proper since it does not contain a
monochrome . We may assume that
is blue and green. But then,
must be green,
otherwise is a proper
.
is red or blue,
otherwise would be a
green .
must be blue,
otherwise is a proper
.
is red or green,
otherwise is a blue
.
Now consider . This
edge cannot be blue, for then is a proper . But neither can this edge be green,
for then is a proper
. Since this edge must be either
blue or green, this is a contradiction.
Case 5. Suppose that . Let , , and . Since we assumed that
has maximal color degree, it
follows that every vertex of has
the same number of edges of each color, namely 2. There are two
subcases.
Subcase 5.1.. Suppose that two of the edges in
share a color. Without
loss of generality, suppose that
and are both blue. In that case,
must be blue, since otherwise
is a proper . Since , , and are symmetrically situated, they must
all be blue as well. But then is a blue , a contradiction.
Subcase 5.2.. Suppose that each of , , and are different colors. Without loss of
generality, suppose is blue.
Then must be green and must be red. Note that each of
vertices and must be incident to two green edges, as
we established at the beginning of this case. The vertices of and each have one green edge already.
Since every vertex must be incident to two green edges, it follows that
there is a green edge out of to
each vertex in and . But this always produces a proper
. Since this configuration is
symmetric, it suffices to check one case. Indeed, if is green, then is a proper . This is a
contradiction.
We have shown that cannot
take on any of its possible values, and thus arrived at a contradiction.
It follows that a -edge-colored
must contain either a
monochrome or a properly
colored , so . Since we already
showed that , we
conclude that .
4. Concluding Remarks
These two theorems further the determination of the proper Ramsey
number of against . We add the small odd cycles and to the even cycles treated by
Olejniczak in [2]. There are only five remaining
cases: , , , , and . Since the three color Ramsey
number (see
[6]), any -edge-colored copy of necessarily contains a monochrome
, so any -edge-colored graph large enough to
contain a proper necessarily
contains a monochrome already.
We may therefore stop at .
Thus, finding these last five values would complete the determination of
.
Conflict of Interest
There was no conflict of interest in this study.
References:
English, S., Johnston, D., Olejniczak, D. and Zhang, P., 2017. Proper Ramsey numbers of graphs. Journal of Combinatorial Mathematics and Combinatorial Computing, 101, pp.281-299.[Google Scholor]
Olejniczak, D., 2019. Variations in ramsey theory. Ph.D. Thesis. Western Michigan University, Kalamazoo, MI.[Google Scholor]