Some remarks on a paper of Chetwynd
and Hilton on critical star multigraphs
David Cariolaro
Institute of Mathematics
Academia Sinica
Nankang, Taipei
11529 Taiwan
E-mail: cariolaro@math.sinica.edu.tw
June 30, 2008
Abstract
In [A.G. Chetwynd and A.J.W. Hilton, Critical star multigraphs, Graphs
and Combinatorics 2 (1986), 209-221] Chetwynd and Hilton started the
investigations of the edge-chromatic properties of a particular class of
multigraphs, which they called star multigraphs. A star multigraph is a
multigraph such that there exists a vertex v ∗ that is incident with each
multiple edge. Star multigraphs turn out to be useful tools in the study
of the chromatic index of simple graphs. The main goal of this paper is
to provide shorter and simpler proofs of all the main theorems contained
in the above mentioned paper. We shall also give an extension of one of
the Chetwynd-Hilton results to list-colouring and prove a general result
of independent interest.
Keywords: star multigraph, chromatic index, edge-colouring, fans
MSC 2000: 05C15
1
Introduction
All graphs considered in this paper are loopless, undirected and finite, but may
contain multiple edges. The term “multigraph” will be used here as a synonym
of “graph”. Let G be a graph. The vertex set and edge set of G are denoted by
V (G) and E(G), respectively. The degree of a vertex v in G, denoted by dG (v),
is the number of edges incident with v in G. ∆(G) denotes the maximum degree
of G. If u, v are vertices of G, we denote by uv the set of edges joining u and v.
The cardinality of uv is denoted by µ(uv) and called the multiplicity of uv. If
µ(uv) = 1, we say that uv is a simple edge and, if µ(uv) > 1, we say that uv is a
1
multiple edge. A graph is simple if all its edges are simple. If e is an edge joining
the vertices u and v, we denote this by e ∈ uv, or e = uv when uv is a simple
edge. Two edges are adjacent if they are distinct and have at least one common
endpoint, and parallel if they are distinct and have two common endpoints. If
S is a set of vertices or edges of G, we denote by G − S the graph obtained
from G by deleting all the elements of S, together with the edges incident to
any vertex in S.
An edge-colouring of G is a map ϕ : E(G) → C, where C is a set, called the
colour-set, whose elements are called colours, and ϕ has the property that adjacent edges are assigned distinct colours. If C is chosen so that |C| is minimum,
then ϕ is called an optimal colouring and the integer |C| is called the chromatic
index of G, denoted by χ′ (G). G is said to be k-edge-colourable if k ≥ χ′ (G).
Clearly χ′ (G) ≥ ∆(G), since all the edges incident with a vertex of maximum
degree must receive a distinct colour. If χ′ (G) = ∆(G), we say that G is Class
1 and, otherwise, we say that G is Class 2. An edge e of G is called critical
if χ′ (G − e) < χ′ (G). G itself is called critical if it is Class 2, has no isolated
vertices, and all its edges are critical. It is well known and easy to see that
every Class 2 graph G contains a critical subgraph with the same chromatic
index. For an introduction to edge-colouring, and for graph-theoretic notation
and terminology, not explicitly introduced here, we refer the reader to Fiorini
and Wilson [7].
A celebrated theorem of Vizing [11] is equivalent to the statement that every
Class 2 simple graph G satisfies χ′ (G) = ∆(G)+1. An extension of this theorem
(Theorem 2 below) was first proved by Chetwynd and Hilton in [5], where the
study of a new class of graphs, known as star multigraphs, was begun. A star
multigraph is a graph G such that there exists a vertex v ∗ (called a star centre) to
which all multiple edges of G are incident. Equivalently, G is a star multigraph
if there exists a vertex v ∗ such that G − v ∗ is a simple graph. Thus a star centre
may not be unique, but is unique unless G is a simple graph or there exists
only one multiple edge in G. Star multigraphs, as stated in [5], are “vital tools
in the investigations of the chromatic index of certain kinds of simple graphs”.
The underlying philosophy is that, if the chromatic index of a certain simple
graph H is unknown, it may be helpful to embed H into a star multigraph
G (typically by the addition of a vertex v ∗ and suitably chosen multiple edges
joining v ∗ to the vertices of H) in the attempt to draw information about H
from the knowledge of the edge-colouring properties of G. It was this line of
investigation that led Chetwynd and Hilton in [6] to formulate the Overfull
Conjecture, which is now considered one of the most interesting and difficult
conjectures in edge-colouring. Thus, it appears that star multigraphs offer a
very fruitful line of investigation.
In [5] Chetwynd and Hilton completely classified star multigraphs with at
most two vertices of maximum degree. In [6] they classified certain star multigraphs with three vertices of maximum degree. (To the best of our knowledge
the problem of the classification of all the star multigraphs with three vertices
of maximum degree remains open to this date.) The proofs of the results contained in [6] are considerably more involved than those in [5], and we shall not
2
be concerned with them in this paper. Instead, we shall provide simplifications
to the proofs of all the main results of [5], and sometimes our simplifications
will be considerable. We remark that all the proofs in [5] and [6] are (in the
style of the authors) very precise and, at times, ingenious. However the reading of [5] and [6] can prove to be difficult for those readers not accustomed
with edge colouring, in particular for those parts of the proofs where fans or
the so-called fan argument is used. One of our goals will be to provide some
conceptual simplifications to the Chetwynd-Hilton proofs. Indeed, having put
on a sound and formal basis the theory of fans in [2], we shall use some of our
results to express, very succinctly, facts concerning fans in a rigorous way, in
the hope to give to the reader a better understanding of the material presented.
In particular, we shall frequently use an expression for the chromatic index of a
Class 2 multigraph which we recently obtained [3, 4] (Lemma 1 below). We call
this expression the Fan Formula. Quite unexpectedly, the Fan Formula gives an
exact expression for the chromatic index of a Class 2 graph under very general
conditions. Thus the paper may also be viewed as an attempt to show the power
and wide applicability of this formula. Before we state it, however, we need to
expose the reader to a certain amount of jargon. This will be compensated later
by considerably shortening the proofs of the theorems.
An e-tense colouring φ of a graph G is a partial edge-colouring of G which
assigns no colour to e and whose restriction to E(G − e) is an optimal colouring
of G − e. The colour set of φ is defined to be the colour set of its restriction to
G − e. The edge e is called the uncoloured edge.
Given an e-tense colouring φ of G with colour set C, and a vertex w ∈ V (G),
we say that a colour α ∈ C is missing at w (or that w is missing the colour α) if
there is no edge, having w as an endpoint, which is assigned the colour α by φ.
Let e be an edge of G and let u be an endpoint of e. Let φ be an e-tense
colouring of G. A fan at u with respect to φ is a sequence of edges of the form
F = [e0 , e1 , e2 , . . . , ek−1 , ek ],
where e0 = e, ei ∈ uvi , and where the vertex vi is missing the colour of the edge
ei+1 , for every i = 0, 1, . . . , k − 1. The vertex u is called the pivot of the fan.
The fan F is said to terminate at the edge ek . A fan is maximal if it cannot be
extended to a larger fan. An edge f is called a fan edge at u if it appears in at
least one fan at u. A vertex w is called a fan vertex at u if it is joined to u by
at least one fan edge. The set of fan vertices1 is denoted by V (F). If w is a fan
vertex at u, we denote by µ∗ (uw) the number of fan edges joining u and w, and
call µ∗ (uw) the fan multiplicity of the edge uw.
We are now ready to state the Fan Formula [3, 4].
Lemma 1 Let G be a Class 2 multigraph and let e ∈ uv be a critical edge.
Let φ be a tense colouring with respect to the edge e, and let V (F) be the set
1 As the notation suggests, the set of fan vertices is the vertex set of a graph, which is called
the Fan Digraph and was introduced and studied in [2]. However, this concept will not be
necessary in the present context. We refer the reader to [2] for further details.
3
of fan vertices at u with respect to φ. Then
χ′ (G) =
1
|V (F )|
1
= ⌈ |V (F
)| ·
P
·
P
w∈V (F ) (degG (w)
w∈V (F ) (degG (w)
+ µ∗ (uw)) +
|V (F )|−2
|V (F )|
+ µ∗ (uw))⌉
We shall say that the Fan Formula is written at u, to indicate that the pivot
of the fans is the vertex u. Notice that |V (F)| ≥ 2 holds always under the
hypotheses of Lemma 1, from which the second equality above follows easily.
The Fan Formula may be deduced easily from the main theorem of [2], the Fan
Theorem.
We shall often use the following property, discovered independently by Andersen [1] and Goldberg [8, 9] and implicit in the work of Vizing (see [2, Lemma
2]).
Lemma 2 Let G be a Class 2 graph, e a critical edge and φ an e-tense colouring
of G. Let u be an endpoint of e. Let V (F) be the set of fan vertices at u. Then,
for any two distinct vertices x, y ∈ V (F) ∪ {u}, and for any colour α missing at
x and β missing at y, we have α 6= β.
The exposition and the organization of the results of this paper does not
follow the same order as in Chetwynd and Hilton [5]. Instead the paper is
organized as follows. Section 2 is dedicated to star multigraphs with one vertex
of maximum degree and to the proof that every Class 2 star multigraph G
satisfies χ′ (G) = ∆(G) + 1. For a special class of star multigraphs we also
determine the list-chromatic index.
In Section 3 we prove two lemmas. The first of these is a general edgecolouring result which will be only of partial use here, but which may prove to
be very helpful in the future. The second lemma specifies the edge-colouring
properties of the complete graph with one edge replaced by r parallel edges, and
will be very helpful in Section 4.
Finally, in Section 4 we completely classify star multigraphs with two vertices
of maximum degree.
There are some additional results in [5] which we will not prove. Their proof
is, in our view, sufficiently simple as it is in the original source, and most of
them are corollaries of the results proved by us, or can be proved by an easy
adaptation of the arguments adopted by us.
2
Star multigraphs with one vertex of maximum
degree
The first of our theorems will be stated in terms of list-colouring. We first
give some relevant definitions. If G is a graph, an edge-list-assignment L is a
function which assigns to each edge e of G a list L(e) of colours (i.e. a set). We
say that G is L-choosable if it is possible to select a colour from each list and
4
assign it to the corresponding edge in such a way that the resulting colouring
is a proper edge-colouring. If all the lists have cardinality k, we call L a klist-assignment. G is said to be k-edge-choosable if, for every k-list-assignment
L, G is L-choosable. The minimum k for which G is k-edge-choosable is called
list-chromatic index of G and denoted by χ′l (G). Clearly χ′l (G) ≥ χ′ (G), as may
be seen by taking all the lists to be coincident with a fixed colour-set C, and the
well known List-Colouring Conjecture asserts that χ′l (G) = χ′ (G) for any graph
G.
We are ready to prove our first theorem. Hilton and Chetwynd proved the
edge-colouring part of the same theorem as part of their proof of [5, Theorem
1]. We shall generalize this result to its list-colouring analogue. In doing so
we shall use the easy fact that the List-Colouring Conjecture holds for graphs
G with ∆(G) ≤ 2 and a result of Harris [10] that says that, if a simple graph
satisfies ∆(G) ≥ 3, then χ′l (G) ≤ 2∆(G) − 2.
Theorem 1 Let G be a star multigraph such that every vertex is incident
with a multiple edge. Then
∆(G) + 1 if |V (G)| = 3, |E(G)| = ∆(G) + 1;
χ′l (G) = χ′ (G) =
∆(G)
otherwise.
Proof. If |V (G)| ≤ 3 the statement of the theorem is immediate, so we shall
assume |V (G)| ≥ 4. Let v ∗ be the star centre (which is necessarily unique in
this case) and let V (G) = {v ∗ , v1 , v2 , . . . , vs }, where s ≥ 3. Let µ(v ∗ vi ) = ki
for all i = 1, 2, . . . , s. By assumption, ki ≥ 2, so v ∗ has degree ∆(G). Let
L = {L(e) : e ∈ E(G)} be a ∆(G)-list-assignment. Choose a distinct colour
from each of the lists assigned to the edges incident with v ∗ , thus obtaining a
partial colouring ψ. We now aim to extend ψ to the edges of the simple graph
H = G − v ∗ . Consider the list-assignment L1 = {L1 (e) : e ∈ E(H)}, where,
if e = vi vj , L1 (e) is obtained from L(e) by suppressing the colours assigned
by ψ to the edges of the form v ∗ vi and v ∗ vj . Clearly G is L-choosable if H is
L1 -choosable. By construction,
X
|L1 (e)| ≥ |L(e)|−ki −kj = ∆(G)−ki −kj =
kℓ ≥ 2(s−2) ≥ 2∆(H)−2, (1)
ℓ6=i,j
Ps
where we have used the fact that dG (v ∗ ) = ℓ=1 kℓ = ∆(G) and s = |V (H)| ≥
∆(H) + 1. By Harris’ theorem and (1), H is L1 -choosable as long as ∆(H) ≥ 3.
We may then assume ∆(H) ≤ 2. If ∆(H) = 0, there is clearly nothing to prove.
If ∆(H) = 1, then H is L1 -choosable unless there is an edge e ∈ E(H) such that
L1 (e) = ∅. But, in view of (1), this may occur only if s = 2, which is contrary
to assumption. If ∆(H) = 2 and H contains no odd cycle, then χ′l (H) = 2,
and hence, from (1), H is L1 -choosable. If ∆(H) = 2 and H contains an odd
cycle, then χ′l (H) = 3; in this case it suffices to show that at least one of the
inequalities in (1) is strict. If s ≥ 4, the last inequality is strict. Hence we may
5
assume that s = 3. It is easy to see that, by a different initial choice for the
partial list-colouring ψ, the first inequality in (1) may be assumed to be strict for
all the edges of H. Hence, in any case, H may be assumed to be L1 -choosable
and thus G is L-choosable, concluding the proof.
✷
Using Theorem 1, we can now prove that Vizing’s theorem for simple graphs
extends to star multigraphs. Our proof is very short.
Theorem 2
Let G be a star multigraph. Then χ′ (G) ≤ ∆(G) + 1.
Proof. Without loss of generality, we may assume that G is Class 2. Furthermore, replacing G with a critical subgraph with the same chromatic index, we
may assume that G is critical. If every vertex is incident with a multiple edge,
then, by Theorem 1, G is (∆(G) + 1)-edge-choosable, and hence (∆(G) + 1)edge-colourable. Therefore we may assume the existence of a vertex u incident
only with simple edges. Choosing an arbitrary edge e incident with u and any
e-tense colouring φ, and writing the Fan Formula at u, we have
P
1
∗
χ′ (G) = ⌈ |V (F
w∈V (F ) (degG (w) + µ (uw))⌉
)| ·
1
≤ ⌈ |V (F
)| ·
P
w∈V (F ) (degG (w)
+ 1)⌉ ≤ ∆(G) + 1,
where we have used the fact that µ∗ (uw) = µ(uw) = 1 for any w ∈ V (F). This
proves the theorem.
✷
By Theorem 2, any Class 2 star multigraph with maximum degree ∆ has
necessarily chromatic index ∆ + 1. We proceed by considering star multigraphs
with only one vertex of maximum degree.
Theorem 3 Let G be a star multigraph with only one vertex of maximum
degree. Then G is Class 2 if and only if G contains a subgraph on 3 vertices
with ∆(G) + 1 ≥ 3 edges.
Proof. If G contains a subgraph on 3 vertices as stated by the theorem then
G is clearly Class 2. Assume now that G is Class 2. Replacing G with any of
its critical subgraphs with the same chromatic index, we may assume that G is
critical. If every vertex of G is incident with a multiple edge, then G, in view
of Theorem 1, has necessarily the form prescribed by the theorem. Hence we
may assume that G has a vertex u incident with simple edges only. Writing the
Fan Formula at u (with respect to any edge e incident with u and any e-tense
colouring φ), we have
χ′ (G) =
=
1
|V (F )|
1
|V (F )|
·
P
≤ ∆(G) +
·
P
w∈V (F ) (degG (w)
w∈V (F ) (degG (w)
1
|V (F )|
+
|V (F )|−2
|V (F )|
+ µ∗ (uw)) +
+ 1) +
|V (F )|−2
|V (F )|
= ∆(G) +
6
|V (F )|−2
|V (F )|
|V (F )|−1
|V (F )|
< ∆(G) + 1,
where we have used the fact that at most one of the neighbours of u has degree
∆(G). Hence G is Class 1, contradicting the assumption. This contradiction
proves the theorem.
✷
3
Some Lemmas
Before we continue our analysis of star multigraphs, we prove two lemmas in
this section. The first lemma expresses an edge-colouring property of arbitrary
critical multigraphs. A small application of this lemma will be given in the
proof of Lemma 2. The result itself, however, is of independent interest. It may
prove to be useful in the study of such conjectures as the Overfull Conjecture
or the Goldberg-Seymour Conjecture, as well as in the further study of star
multigraphs. We shall use the following notation: if G is a graph and x, y are
vertices of G, we denote by G + xy the graph obtained from G by adding an
edge joining x to y.
Lemma 3 Let G be a critical graph with more than three vertices and let x, y
be adjacent vertices. Then χ′ (G + xy) = χ′ (G).
Proof. Assume, on the contrary, that χ′ (G + xy) > χ′ (G). First observe that
every edge in G is incident with either x or y (or both). For, assume there was an
edge h neither incident with x nor with y. Since G is critical, χ′ (G−h) < χ′ (G),
so there is an optimal colouring ϕ of G and a colour α such that ϕ−1 (α) = {h},
i.e. h is the only edge receiving colour α. Such colouring is easily extendable
to a colouring of G + xy by colouring the extra edge joining x, y with colour α,
thus contradicting the assumption that χ′ (G + xy) > χ′ (G). Hence every edge
is incident with either x or y, i.e. the graph G − xy is bipartite with bipartition
({x, y}, V (G) \ {x, y}). Let φ be any optimal colouring of G. We may think
of φ as a tense colouring of G + xy, where the uncoloured edge is the extra
edge joining x and y. Let e be any edge joining x and y in G. Let ǫ = φ(e).
Since every edge of G is adjacent or coincident with e, there is no other edge
in G coloured ǫ, and hence the colour ǫ is missing at every vertex other than
x and y. Let α be a colour missing at x under the colouring φ and let β be
a colour missing at y. If α = β, then we can colour the uncoloured edge with
colour α, thus contradicting the assumption that χ′ (G + xy) > χ′ (G). It follows
that α 6= β. There must be a bicoloured α-β path joining x and y, otherwise
a colour exchange along a bicoloured α-β path starting at x would result in a
colouring φ′ such that x and y are missing the same colour under φ′ , and this
would contradict what was proved above. The bicoloured α-β path joining x
and y, in the present case, has necessarily length 2, and hence is of the form
xzy. If all the χ′ (G) colours used by φ appeared on one of the edges xy, yz, xz,
then G would have a subgraph on 3 vertices with the same chromatic index
as G, contradicting the fact that G is critical and |V (G)| ≥ 4. Hence there
exists a colour λ which is not present on the edges of the multitriangle xyz. By
symmetry, we may assume that λ is present at the vertex x, say on the edge
7
xw, where w 6= y, z. The vertex z is a fan vertex at x with respect to φ, and
is missing colour ǫ. Since w is also missing the colour ǫ under the colouring
φ, it cannot be a fan vertex at x, and hence the colour λ must be present at
y. It follows the existence of an edge of the form yt, where t 6= x, w, z, which
is coloured λ. Now, the colour β is present at at most one of w, t (since it is
present at z and there may be at most two edges coloured β). Suppose β is
missing at w. Then, by interchanging the colours of the bicoloured λ-β path
wxz, we may guarantee that β is missing at z. But now both z and w become
fan vertices at x under the current colouring φ′ , because [e0 , e1 , e2 ] is a fan at
x, where e0 is the uncoloured edge, e1 is the λ-edge joining x to z, and e2 is the
β-edge joining x to w. But z and w are both missing the same colour ǫ under
φ′ , which contradicts the fact that they are both fan vertices at x. Therefore we
may assume that β is missing at the vertex t under the colouring φ. However,
interchanging the colours of the edges of the α-β path joining x and y yields
a colouring φ′′ such that β is missing at x and α is missing at y, and hence
creates a situation symmetrical to the one of the other case, which also results
in a contradiction. This contradiction proves the lemma.
✷
We now prove a simple lemma which will be helpful in the sequel. For
positive integers n, r, we2 denote by Kn+r the complete graph Kn with one edge
replaced by an edge of multiplicity r. It is obvious that K3+r is Class 2 for every
+r
r, and easy to see that K2n
is Class 1 for every r and n. Indeed
+r
+r
χ′ (K2n
) ≤ χ′ (K2n ) + r − 1 = 2n + r − 2 = ∆(K2n
).
The following lemma deals with the remaining cases. We shall use the fact that
K2n+1 is Class 2 for every n and, in each optimal colouring of K2n+1 , every
vertex is missing a distinct colour.
Lemma 4 Let r, n be positive integers, n ≥ 2, r ≥ 2. Then the following are
equivalent properties:
(a) r = 2;
+r
(b) K2n+1
is Class 2;
+r
(c) K2n+1
is critical.
+2
Proof. Assume r = 2. We prove that K2n+1
is Class 2. This is immediate
+2
since, if K2n+1 was Class 1, it would have a (2n + 1) edge-colouring; but then
we would have a (2n + 1)-edge-colouring of K2n+1 with two vertices missing
+2
the same colour, which is impossible. Hence K2n+1
is Class 2. We now prove
+2
that it is critical. Let xy be the multiple edge of K2n+1
. Removing one of the
+2
two parallel edges yields the graph K2n+1 and, since K2n+1 is Class 2, we have
+2
+2
χ′ (K2n+1
) > 2n + 1 = χ′ (K2n+1 ). Remove now any other edge of K2n+1
, say
2 The
+2
+
corresponding notation for the graph Kn
in the Chetwynd-Hilton paper is Kn
.
8
e, and consider the graph K2n+1 − e. It is easy to construct a (2n + 1)-edgecolouring of K2n+1 − e with the two vertices x, y missing the same colour α, and
hence it is possible to colour with colour α an additional edge joining x and y,
+2
+2
+2
proving the inequality χ′ (K2n+1
− e) = 2n + 1 < χ′ (K2n+1
). Hence K2n+1
is
critical. Assume now r = 3. By Lemma 3 (using the assumption n ≥ 2) we have
+3
+2
+3
χ′ (K2n+1
) = χ′ (K2n+1
) = 2n + 2 = ∆(K2n+1
),
+3
and hence K2n+1
is Class 1. It follows by an easy induction argument that, for
+r
every r ≥ 4, K2n+1
is Class 1, since
+(r−1)
+r
+r
χ′ (K2n+1
) ≤ χ′ (K2n+1 ) + 1 = 2n + r = ∆(K2n+1
).
Thus the proof is completed.
4
✷
Star multigraphs with two vertices of maximum degree
We are now ready to prove the following theorem.
Theorem 4 Let G be a star multigraph with two vertices of maximum degree,
one of which a star centre. Then G is Class 2 if and only if G contains a subgraph
+2
on 3 vertices and ∆(G) + 1 edges or G = K2n+1
, for some n ≥ 2.
Proof. If G has the form prescribed by the theorem then clearly G is Class 2.
Assume now that G is Class 2. Let H be a critical subgraph of G with the
same chromatic index (and hence with the same maximum degree). Arguing
by contradiction, we assume that H has more than 3 vertices. By Theorem 3
and the assumption of criticality, H must have two vertices of maximum degree,
and hence H and G have the same vertices of maximum degree. Let v ∗ be a
star centre. If there is a vertex u of H not adjacent to v ∗ , then, using the Fan
Formula at u and the fact that u is adjacent to at most one vertex of maximum
degree of H, we have
P
|V (F )|−2
1
∗
χ′ (G) = |V (F
w∈V (F ) (degG (w) + µ (uw)) + |V (F )|
)| ·
=
1
|V (F )|
·
P
≤ ∆(G) +
w∈V (F ) (degG (w)
1
|V (F )|
+
|V (F )|−2
|V (F )|
+ 1) +
|V (F )|−2
|V (F )|
= ∆(G) +
|V (F )|−1
|V (F )|
< ∆(G) + 1,
whence a contradiction with the fact that H is Class 2. It follows that every
vertex is either adjacent or coincident with v ∗ . Let V (H) = {v ∗ , v1 , v2 , . . . , vs },
where dH (v ∗ ) = dH (u) = ∆(H), and let ki = µ(v ∗ vi ) for every i = 1, 2, . . . , s.
Then
s
s
X
X
ki
(2)
ki = k1 +
∆(H) =
i=2
i=1
9
and
dH (v1 ) = k1 + t = ∆(H),
(3)
∗
where t is the number of neighbours of v1 in J = H − v . By comparison of (2)
and (3), we see that
t=s−1
(4)
and
ki = 1 for every i = 2, 3, . . . , s.
(5)
H ⊂ Kn+r
(6)
∆(H) = ∆(Kn+r )
(7)
It follows that
and
for some n ≥ 4 and r ≥ 2. Since H is Class 2, it follows from (6) and (7) that
Kn+r is Class 2 and, by Lemma 4, that n is odd, n ≥ 5 and r = 2. That is to
+2
+2
say, we have H ⊂ K2n+1
for some n ≥ 2 and ∆(H) = ∆(K2n+1
) = 2n + 1.
+2
+2
Since K2n+1 is critical and H is Class 2 and ∆(H) = ∆(K2n+1 ), we conclude
+2
that H = K2n+1
. Hence every vertex of H has degree 2n, except v ∗ and v1 ,
which have degree 2n + 1. The addition of any edge incident with any vertex
of H would therefore either increase the maximum degree of H or increase the
number of vertices of maximum degree in H. Since H has the same maximum
degree and the same number of vertices of maximum degree as G, it follows that
G = H, and the proof is completed.
✷
To complete the classification of star multigraphs with two vertices of maximum degree we now prove the following.
Theorem 5 Let G be a star multigraph with two vertices of maximum degree,
neither of which a star centre. Then G is Class 1.
Proof. We argue by contradiction, so let us assume that G is Class 2. Replacing
G by a critical subgraph with the same maximum degree (which must necessarily
satisfy the same conditions) we may assume that G is critical. Let v ∗ be a
star centre and let v1 , v2 be the vertices of maximum degree. By assumption,
v ∗ 6= v1 , v2 . If every vertex of G was adjacent or coincident to v ∗ , then v ∗ would
have maximum degree in G, contrary to assumption. Hence there exists a vertex
u which is not adjacent to v ∗ . Writing the Fan Formula at u with respect to
any edge e incident with u and any e-tense colouring, we see that
P
|V (F )|−2
1
∗
χ′ (G) = |V (F
w∈V (F ) (degG (w) + µ (uw)) + |V (F )|
)| ·
=
1
|V (F )|
·
P
≤ ∆(G) +
w∈V (F ) (degG (w)
2
|V (F )|
+
|V (F )|−2
|V (F )|
+ 1) +
|V (F )|−2
|V (F )|
= ∆(G) + 1,
10
where we have used the fact that there are at most two neighbours of u of
maximum degree. Since the sign of equality must hold in the last inequality
above, both v1 and v2 are fan vertices at u, and in particular u is adjacent
to both v1 and v2 and hence is distinct from v1 and v2 (which implies that
both v1 and v2 are adjacent to v ∗ , otherwise, replacing u with one of them and
repeating the argument, we reach a contradiction). Now3 , deleting the edge uv2
and adding an edge joining u to v ∗ , we obtain a star multigraph G′ with star
centre v ∗ and with ∆(G) = ∆(G′ ). Let ∆ = ∆(G) = ∆(G′ ). Notice that G′ has
at most two vertices of degree ∆, namely v1 and (possibly) v ∗ . Suppose G′ was
Class 2. Then, by Theorem 3 and Theorem 4, either G′ contains a subgraph
on 3 vertices and ∆ + 1 edges (which is impossible, since this would force v ∗
to have only two neighbours in G′ , whereas has the three neighbours u, v1 , v2 ),
+2
or G′ = K2n+1
for some n ≥ 2 (which is also impossible, since G′ contains two
non-adjacent vertices, namely u and v2 ). We conclude that G is Class 1. Thus
there exists a ∆-edge-colouring of G′ , and hence there exists a ∆-edge-colouring
φ of G′ − uv ∗ = G − uv2 such that u and v ∗ are missing the same colour α. We
can view φ as an e0 tense colouring of G, where e0 = uv2 is the uncoloured edge.
Then u is a fan vertex at v2 missing colour α and, consequently, v ∗ cannot be a
fan vertex at v2 , since it is also missing colour α. Since dG (u) < ∆, there are at
least two missing colours at u, say α and β, and two maximal fans F1 , F2 , one
of which of the form F1 = [e0 , e1 , . . .], where e1 is coloured α, and the other of
the form [e0 , f1 , . . .], where f1 is coloured β. Since no fan vertex at v2 is joined
to v2 by a multiple edge, it follows that all the edges of F1 (and, similarly, F2 )
are incident with distinct fan vertices, and the fan vertices of F1 are distinct
from those of F2 , except for u, which is an endpoint of e0 . Moreover both F1
and F2 must terminate with a fan edge incident with a vertex of maximum
degree, which is necessarily v1 . Hence F1 and F2 both contain v1 as a fan
vertex, contradicting the fact that the only common fan vertex is u, and this
contradiction establishes the theorem.
✷
Final Remarks: In this paper we have obtained simplifications of the proofs
of the main results of [5] by means of some tools related to the study of fans,
in particular the Fan Formula. Such tools may be applied to classes of graphs
other than star multigraphs. Lemma 3 may have applications which go beyond the scope of this paper. In particular it may be useful in order to settle
the classification problem for star multigraphs with three vertices of maximum
degree.
References
[1] L.D. Andersen, On edge-colourings of graphs, Math. Scand., 40 (1977),
161-175.
3 The following ingenious argument is the same used by Chetwynd and Hilton in [5, Lemma
15]. However the present proof is considerably shorter because we do not prove, as a preliminary step, that ∆(G) ≤ |V (G)| − 1, which is unnecessary in our argument.
11
[2] D. Cariolaro, On fans in multigraphs, J. Graph Theory, 51 (2006), 301-318.
[3] D. Cariolaro, Some consequences of a theorem on fans, to appear in Taiwanese J. Mathematics.
[4] D. Cariolaro, The fire index, Graph Theory Notes of New York, LIV:1
(2008), 9-14.
[5] A.G. Chetwynd and A.J.W. Hilton, Critical star multigraphs, Graphs and
Combin., 2 (1986), 209-221.
[6] A.G. Chetwynd and A.J.W. Hilton, Star multigraphs with three vertices of
maximum degree, Math. Proc. Cambr. Phil. Soc., 100 (1986), 303-317.
[7] S. Fiorini and R.J. Wilson, Edge-Colourings of Graphs, Research notes in
Mathematics, Pitman, 1977.
[8] M.K. Goldberg, Remark on the chromatic class of a multigraph, Vycisl.
Mat. i Vycisl. Tech. (Kharkov), 5 (1975), 128-130 (in Russian).
[9] M.K. Goldberg, Edge-coloring of multigraphs: recoloring technique, J.
Graph Theory, 8 (1984), 123-137.
[10] A.J. Harris, Problems and conjectures in extremal graph theory, Ph.D. thesis, Cambridge University, 1984.
[11] V.G. Vizing, On an estimate of the chromatic class of a p-graph, Diskret.
Analiz., 3 (1964), 25-30 (in Russian).
12