Teorija grafov – Barvanja grafov (Lekcija 3)

Teorija grafov – Barvanja grafov (Lekcija 3)

Avtor: Alen Orbanić, Primož Potočnik, Tomaž Pisanski

Vertex-colorings of graphs

  • Let = be a graph and a set of “colors”
 
Definition
A vertex-coloring (barvanje točk) of is a function : . The coloring is proper (pravilno) if () . is -vertex-colorable (točkovno -obarvljiv) if there exists a proper vertex-coloring with = .


  • is -colorable is -colorable for any .
  • Chromatic number (kormatično število) of :

    = mink : is -vertex-colorable

Vertex-colorings – examples

(cube_vertex_coloring.png)
2-coloring of the cube
(petersen_vertex_coloring.png)
3-coloring of the Petersen

... more examples

  • .
  • .
  • If , then .
 
Corollary
If contains a cycle of odd length, then .

Graphs with

  • Clearly, if and only if .
 
Lemma
if and only if is bipartite.

PROOF: ... think of color classes as bipartition sets ...

  • We know that graphs with cannot have cycles of odd length We will now show that the converse holds as well:
 
Lemma
If contains no cycles of odd length, then .

...

  • PROOF. WLOG: is connected.
  • Choose . For let:

    • = "blue" if is even;
    • = "red" if is odd.
  • If this is not a proper coloring, then there are two adjacent vertices that are both at even or both at odd distance from .
  • Find shortest paths , from to and to . Then is a closed walk of odd length.
(dvepoti.png)
  • To complete the proof, we need to show the following:
  • Exercise.: If a graph contains a closed walk of odd length, then it also contains a cycle of odd length.

Characterization of bipartite graphs

This proves the following characterization of bipartite graphs.

 
Theorem

If is a graph, then the following statements are equivalent:

  • is bipartite.
  • .
  • contains no cycles of odd length.

Cliques

 
Definition
A subset is called a clique (klika), if the induced subgraph is a complete graph.
(klike.png)

Maximal Clique

 

Definition
A maximal clique (maksimalna klika) is a clique that is not contained in any other clique. A largest clique (največja klika) is a clique with the largest number of vertices among all cliques.

= "the size of the largest clique in ".

  • Since = , it follows that .

The Brooks theorem

 

Theorem
(Brooks) Let be a graph. Then

.


Moreover, unless is a complete graph or a cycle of odd length.

Proof of the Brooks theorem

  • WLOG: is connected.
  • We already know that . So we need to show two things:

    • .
    • If or , then .
  • Finding a -coloring is easy:

    • Let be the set of colors. Order the vertices of in some linear order. Color the first vertex with color .
    • Suppose that we have already colored the first vertices. Let be the next vertex, and let be the smallest integer that does not appear as a color of some neighbor of . Color v with the color .
    • Repeat this procedure until all vertices are colored.

(pozresna_01.png)

(pozresna_02.png)

(pozresna_03.png)

(pozresna_04.png)

(pozresna_05.png)

(pozresna_06.png)

(pozresna_07.png)

Proof of the Brooks theorem – killing unessential greens

  • In the rest of the proof, we may assume that or . It remains to show, that we may change the -coloring in such a way that one of the colors “disappears”.
  • For the rest of the proof: = “green”.
  • Let = “the set of green vertices”.
  • If there is such that one of the non-green colors does not appear among its neighbors, then we may use this color for . Apply this throughout . This procedure is called “killing unessential greens”.
(nonessential.png)


  • After unessential greens are “killed”, we get a -coloring in which each green vertex has valence , and no two neighbors of are of the same color.

Proof of the Brooks theorem – pushing the green color

The second procedure allow us to “push” the green color from any to any other along any path from to .

  • Kill unessential s. If or , then stop.
  • Let be the first vertex on the path from to . Let be the color of . Since we killed unessential greens, no green neighbours of have any other neighbours of color .
(shift_01.png)

Proof of the Brooks theorem – pushing the green color

  • Change the color of green neighbors of to , and change the color of to green.
(shift_02.png)
  • Go to step 1 with in place of , and with being the part of old from to
(shift_03.png)

This procedure changed the color of some old green vertices, and cyclically rotated the colors along .

The proof of Brook’s theorem – reduction to regular graphs

  • Let be a vertex of smallest valence. For each green , choose a shortest path from to , and push the green color to . Now is the only green vertex.
  • If , then there is a non-green color which does not appear among neighbors of . Hence we can kill the green color at , and finish.
  • It follows: We may thus assume that is regular (all vertices have valence ).
  • The proof now splits into two cases:

    • is -connected;
    • is not -connected.

The proof of Brook’s theorem – the 3-connected case

  • Suppose is -connected.
  • Since is not complete, there exist . Push the green color from all green vertices to .
  • Since there in no green color in , there exist of the same color.
(kill_green_01.png)

The proof of Brook’s theorem – the 3-connected case II

  • Consider the graph . Since is -connected, is connected. Choose a shortest path from to in and push the green color from to along this path.
(kill_green_02.png)

The proof of Brook’s theorem – the 3-connected case III

  • This results in a proper coloring of where the only green vertex is , where and (two neighbors of ) have the same color. Therefore, the green color of can be “killed”, giving a -coloring of .
(kill_green_03.png)

The proof of Brook’s theorem

  • Suppose now that is not -connected.
  • The rest of the proof of is by induction on . By inspection, we see that the theorem holds for . Assume now that and that theorem holds for all graphs with less than vertices.
  • If , then , and so .
  • If , then or , and the theorem holds.
  • Assume henceforth that .

...

  • Suppose that has a cut-vertex , and let be the components of .
(cut_vertex.png)
  • By induction, each is -colorable. By renaming colors in each if necessary, we may assume that in all , the vertex has the same color. This gives a -coloring of .

...

  • Suppose now that has a vertex-cut of size two:.
(two_cut.png)
  • In a similar way as in case we may use induction to show that is -colorable.
  • Homework H2: Finish the proof of the theorem in this case.

The Mycielski construction

  • Let be a graph on vertices with at least one edge. Construct a new graph on vertices in the following way:
  • (a disjoint union).
  • .
  • Homework H3: Show that .
  • Example: The graph, obtained in this way from is called the .
(mycielski.png)

Edge-colorings

  • Let be a graph and a set of “colors”. We define edge-colorings in a similar way as vertex-colorings:
 
Definition
An edge-coloring (barvanje povezav) of is a function . The coloring is proper if incident edges receive different collors. The graph is -edge-colorable (povezavno -obarvljiv) if there exists a proper edge-coloring with .
  • The minimal integer for which is -edge-colorable is called the chromatic index (kormatični indeks) of .
= mink : is −edge–colorable


  • Note that .

Vizing’s theorem

  • There is an obvious natural lower bound: .
  • The upper bound is given by Vizing’s theorem.
 
Theorem
(Vizing)
  • We skip the proof.

Class 1 vs. Class 2, theorem

  • Graphs with are graphs of class 1, the others are of class 2.
  • is of class , is of class .
  • Hypercubes are of class
  • The Petersen graph is of class .
  • In general, determining is difficult.
  • For some graphs, this task is easier. For example, bipartite graphs.
 
Theorem
If is bipartite, then .

Proof of theorem

  • By contradiction: Let be a positive integer. Among all graphs with choose a counterexample with the least number of edges.
  • Choose an edge , such that (What if such and edge does not exist?).
  • By hypothesis, . Color the edges with colors.

Proof of theorem II

  • There is a color , which does not appear at , and a color , which does not appear at .
(kempe_01.png)
  • If , color with .

Proof of theorem III

  • Assume now that .
  • Consider the subgraph induced by the edges of colors and . Clearly , so the connected components of are paths or cycles.
(kempe_02.png)

Proof of theorem IV

  • Note that swappings colors and in any component of gives a different proper coloring.
  • Since all paths from to are of odd length ( is bipartite!), and are in different components of . Swap the colors and in a component containing .
(kempe_03.png)

Proof of theorem V

  • Finally, color with .
(kempe_04.png)

Snarks

  • A regular graph of valence is called cubic graph (kubičen graf).
  • Homework H3: Show that every connected cubic graph with is -edge-connected.
  • On the other hand, it is not easy to find -edge-connected cubic graphs with .
  • Such a graph is called a snark. (The name comes from a poem “The hunting of the Snark” by Lewis Carol.)
  • The smallest such graph is the Petersen graph.
  • Constructing new families of snarks is still a difficult task.

Matchings

  • Consider a proper edge-coloring of . Consider the set of edges colored with a fixed color. No vertex of is incident with more than one edge from .
 
Definition
A matching (prirejanje) in a graph is a set such that each is incident with at most one .
  • Vertices, that are incident with some are saturated (nasičen).
  • If every vertex of is saturated, then the matching is perfect (popolno prirejanje).
  • A matching is maximal if it is the largest among all matching.

Stable sets and covers

  • Maximal matchings are related to “stable sets”, “vertex covers” and “edge covers”
 
Definition
A stable set in is a set such that no two vertices in are adjacent in . A vertex cover in is a set such that every edge of is incident with at least one vertex in . An edge cover in is a set such that every vertex of is incident with at least edge vertex in .
  • := “the size of a maximal matching ”;
  • := “the size of a largest stable set ”;
  • := “the size of a smallest vertex cover of ”;
  • := “the size of a smallest edge cover of ”.

Gallai’s theorem and the theorem

 
Theorem
(Gallai, 1959) If has no isolated vertices, then .


 
Theorem
() If is bipartite, then .

We skip the proofs.

Hall’s theorem

  • It is often difficult to decide, what is the size of a largest matching. For bipartite graphs, we have the following nice result:
 
Theorem
(Hall) Let be a bipartite graph with bipartition . Then has a matching in which every vertex of is saturated if and only if for every set .


  • Here is the set of vertices that are adjacent to some vertex in .

Proof of Hall’s theorem

  • PROOF. One direction is obvious.
  • For the other direction, we need the theorem. Suppose that there is no matching in which every is saturated. Then .
  • By the theorem, . Therefore, there is a vertex cover with .
  • Let . Then , and so

Homework

H1 Finish the proof of the Brooks theorem in the case where the vertex-connectivity of the graph is .

H2 Show that .

H3 Show that every connected cubic graph with is -edge-connected.

0%
0%