Sei sulla pagina 1di 3

We use induction on k. The graphs K1 and K2 show that the result is true for k=1 and 2 respectively.

Now suppose that k>=2 and that Mk is a k-chromatic graph containing no triangles. We use Mk to
construct a (k+1)-chromatic graph Mk+1 containing no triangles.

First suppose that v1, v2,....,vn are the vertices of Mk. The vertices of the new graph Mk+1 are
defined to be those of Mk together with an extra n+1 vertices denoted by u1,u2,...,un,v. The edge
set of Mk+1 is defined to consist of all the edges of Mk together with an edge from v to each of the
ui’s and an edge from each ui to each of the neighbours of vi.

Starting with M2 = K2, Figure 6.17 shows the construction of M3 and M4.

We now show that Mk+1 has no triangles. Assume this is false, i.e., that there is a triangle in Mk+1.
Since Mk, by assumption, contains no triangles, such a triangle must have either v or at least one of
the ui’s as a vertex. Since no two of the ui’s are adjacent and v is only adjacent to the ui’s this forces
the triangle to be of the form vjvkuivj. However,since ui is adjacent to vj and vk, these latter vertices
must be neighbours of vi and we get the triangle vjvkvivj, impossible since this lies in Mk. Thus Mk+1
has no triangles, as required.

Now we show that X(Mk+1)=k+1. Given a k-colouring of Mk we can extend this to a(k+1)-colouring of
Mk+1 by colouring each ui by the colour assigned to vi and then colouring v with a(k+1)st colour. (It
is easily checked that this is a (k+1)-colouring of Mk+1, because of the way the edge set of Mk+1 is
defined). Thus X(Mk+1)<=k+1 and so we must now show that Mk+1 can not be k-coloured.

Suppose, to the contrary, that Mk+1 has a k-colouring using the colours 1,2,...,k. Suppose that the
vertex v is coloured k. Then each vertex ui can not be coloured k, since v and ui, are adjacent. Also,
since X(Mk)=k, the colour k must be used to colour some vertices in Mk. Recolour those vertices vi
coloured k in Mk by the colour assigned to their corresponding ui. Then, since ui and vi have exactly
the same neighbours in Mk, this has produced a(k-1)-colouring of Mk, which is a contradiction since
X(Mk)=k. It now follows that X(Mk+1)=k+1, finishing the proof.

We note in passing that the graph M4 shown in Figure 6.17 is know as the Grotzsch graph.
Although Mycielski’s construction shows that in general the clique number may be quite different
from the chromatic index there is a large class of graphs where the numbers are equal. To describe
this class we first need a result on the complement G of a graph G.

Suppose that there is a subset U of G such that the induced subgraph G[U] of G is isomorphic to P4.
Since P4 is its own complement, the induced subgraph G[U] of G is also isomorphic to P4. In
particular, G[U] and G[U] are both connected. This argument shows that (a) implies (b).

Now suppose that G does not containt a vertex-induced subgraph isomorphic to P4. We adopt a
proof by contradiction to show that (a) holds. Thus we assume that G does contain a nonempty
subset U such that both G[U] and G[U] are connected. Clearly we can choose such a subset U to have
as small a number of vertices as possible. Thus if T is any proper subset of U then either G[T] or G[T]
is connected. Also a moment’s thought shows that our minimum subset U has at least 4 vertices.

Let u1 be a vertex in U and set U1=U-{u1}. Then, by the minimality of U, either G[U1] or G[U1] is
disconnected. Suppose that F[U1] is disconnected. (It suffices to do this since, once our argument is
finished, we may replace G by G because P4 is the complement of itself.) Then, since G[U] is
connected, there is a vertex u2 in U1 such that u1u2 ∉ 𝐸(𝐺), the edge set of G.

Denote the vertex set of the component of G[U1] containing u2 by U2. Then any vertex u3 in U2 has
no neighbours in the set U1-U2. Moreover, since G[U] is connected, there is such a vertex u3
adjacent to u1 in G and also a vertex u4 ∈ U1 – U2 also adjacent to u1 in G. So thus far we have
u1u2∈ 𝐸(𝐺) and u1u3,u1u4∈ 𝐸(𝐺).

Now let

𝑋 = {𝑥 ∈ 𝑈2 : 𝑥 𝑖𝑠 𝑎 𝑛𝑒𝑖𝑔ℎ𝑏𝑜𝑢𝑟 𝑜𝑓 𝑢1 𝑖𝑛 𝐺} and

𝑌 = {𝑦 ∈ 𝑈2 : 𝑦 𝑖𝑠 𝑎 𝑛𝑒𝑖𝑔ℎ𝑏𝑜𝑢𝑟 𝑜𝑓 𝑢1 𝑖𝑛 𝐺̅ }.

Then X and Y are both nonempty since 𝑢3 ∈ 𝑋 and 𝑢2 ∈ 𝑌. Also clearly 𝑈2 = 𝑋⋃𝑌 and 𝑋⋂𝑌 = 𝑣.
Moreover, since G[𝑈2 ] is connected (by the definition of 𝑈2 0 there are vertices 𝑥 ∈ 𝑋 and 𝑦 ∈ 𝑌
such that 𝑥𝑦 ∈ 𝐸(𝐺).

Finally let 𝑍 = {𝑢4 , 𝑢1 , 𝑥, 𝑦}. Then we have 𝑢4 𝑢1 , 𝑢1 𝑥, 𝑥𝑦 ∈ 𝐸(𝐺) but 𝑢1 𝑦 ∉ 𝐸(𝐺), by the definition
of Y, and 𝑢4 𝑥, 𝑢4 𝑦 ∉ 𝐸(𝐺), since 𝑢4 is not in the component G[𝑈2 ] of G[𝑈1 ]. It follows that G[Z] is
isomorphic to P4,in contradiction to our initial assumption. Thus (b) does imply (a), as required.

We now use Theorem 6.10 to prove our final result of this section. Due to Seinsche [56], it displays a
class of graphs where the clique number does equal the chromatic index.

We use induction on n, the number of vertices of G. For n = 1 the result is true since X(K1)=cl(K1)=1.

Now suppose that the result is true for all graphs satisfying the hypothesis and with at most n
vertices where 𝑛 ≥ 1 is fixed. Let G be a graph with 𝑛 + 1 vertices in which no subset of four vertices
induces 𝑃4 as a subgraph. Then, by Theorem 6.10, for any nonempty subset U of V(G) either G[U] or
𝐺̅ [𝑈] is disconnected. In particular, taking U=V(G), we see that either G or its complement 𝐺̅ is
disconnected.
First suppose that G is disconnected and denote its components by 𝐺1 , 𝐺2 , … , 𝐺𝑖 , so that 𝑡 ≥ 2. Then
clearly no 𝐺𝑖 contains 𝑃4 as an induced subgraph. Thus, since each 𝐺𝑖 has less vertices than G, our
induction assumption gives 𝑋(𝐺𝑖 ) = 𝑐𝑙(𝐺𝑖 ) for each 𝑖 = 1, … , 𝑡. Since 𝑋(𝐺) = max{𝑥(𝐺𝑖 ): 1 ≤ 𝑖 ≤
𝑡}, by Theorem 6.1(e), and 𝑐𝑙(𝐺) = max{𝑐𝑙(𝐺𝑖 ): 1 ≤ 𝑖 ≤ 𝑡} (see Exercise 6.4.1), it follows that
𝑋(𝐺) = 𝑐𝑙(𝐺), as required.

Now suppose that 𝐺̅ is disconnected and denote the complements of its components by
𝐻1 , 𝐻2 , … , 𝐻𝑠 , so that 𝑠 ≥ 2. Since G does not contain 𝑃4 as an induced subgraph neither do each of
the 𝐻𝑖 ′𝑠. Thus, again by our induction assumption, 𝑋(𝐻𝑖 ) = 𝑐𝑙(𝐻𝑖 ) for 𝑖 = 1, … , 𝑠. Moreover, by
Exercise 1.5.5 (d), G is the join

𝐺 = 𝐻1 + 𝐻2 + ⋯ + 𝐻𝑠
Hence, by the Exercise 6.3.4 (a), 𝑋(𝐺) = ∑𝑠𝑖=1 𝑋(𝐻𝑖 ). Since 𝑐𝑙(𝐺) = ∑𝑠𝑖=1 𝑐𝑙(𝐻𝑖 ) (by Exercise 6.4.3) it
follows that 𝑋(𝐺) = 𝑐𝑙(𝐺), as required. The result now follows by induction.

It is quite easy to see that the converse of Theorem 6.11 is false. Just take G to be 𝑃4 itself.

Potrebbero piacerti anche