Scribd
Upload
240 views22 pages

Hamiltonian Graphs

This document discusses Hamiltonian graphs, which are graphs that contain a Hamiltonian cycle (a cycle that passes through each vertex exactly once). Some key results discussed include: - Ore's theorem, which provides a sufficient condition for a graph to be Hamiltonian based on the sum of degrees of non-adjacent vertices. - Dirac's theorem, which states that if the minimum degree of a graph is at least half the number of vertices, the graph is Hamiltonian. - An algorithm for finding the Ore closure of a graph, which is the graph obtained by adding edges until the Ore condition is satisfied. - Results showing that a graph is Hamiltonian if and only if its Ore closure is Hamiltonian, or

Uploaded by

mmrmathsiubd
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
240 views22 pages

Hamiltonian Graphs

This document discusses Hamiltonian graphs, which are graphs that contain a Hamiltonian cycle (a cycle that passes through each vertex exactly once). Some key results discussed include: - Ore's theorem, which provides a sufficient condition for a graph to be Hamiltonian based on the sum of degrees of non-adjacent vertices. - Dirac's theorem, which states that if the minimum degree of a graph is at least half the number of vertices, the graph is Hamiltonian. - An algorithm for finding the Ore closure of a graph, which is the graph obtained by adding edges until the Ore condition is satisfied. - Results showing that a graph is Hamiltonian if and only if its Ore closure is Hamiltonian, or

Uploaded by

mmrmathsiubd
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 22

Hamiltonian graphs

Icosian game by sir William Rowan Hamilton, 1857

Hamiltonian cycle in graph G is a cycle that passes

Hamiltonian walk in graph G is a walk that passes

through each vertex exactly once.


through each vertex exactly once.

 If a graph has a Hamiltonian cycle, it is called a Hamiltonian graph .

 If a graph has a Hamiltonian walk, it is called a semiHamiltonian graph .

There are no known (non-trivial) conditions that would


be necessary and sucient for the existence of a Hamiltoinian cycle or a Hamiltoinian walk.

 In this lecture, only simple graphs are considered.

Theorem (Ore, 1960). Let G = (V; E ) be a simple graph,

where jV j =
implication

n  3. If for every two vertices u; w 2 V the

(u; w) 62 E ==> deg(u) + deg(w)  n


G is Hamiltonian.
Corollary (Dirac, 1952). If G = (V; E ) is a simple graph
having n vertices and for each v 2 V we have deg(v )  n2
then G is a Hamiltonian graph.
Proof of the Corollary. For every two verticesu u; w 2
V (whether they are neighbours or not) the inequality
deg(u) + deg(w)  n holds, thus Ore's theorem implies
that G is Hamiltonian.
holds, then the graph

Proof of the theorem. If n = 3 then the only graph satisfying the assumption is K3 . It is Hamiltonian.

Let n  4. Let the assumption of the theorem hold, but


let the conclusion be wrong.

If we add edges to the graph, the assumption will still hold.


Add edges to G until we reach the graph G0 such that it
is not Hamiltonian, but addition of any new vertex would
give a Hamiltonian graph.

Let e = (u; w) 2 V  V be an edge not present in


graph G0 [ feg has a Hamiltonian cycle

u = v0  v1  v2      vn
Grph

G0 . The

=wu :

G0 has a Hamiltonian walk


P : u = v0  v1  v2      vn

This walk has

1 edges.

=w :

Let




Eu be the set of edges (vi ; vi+1 ) where (u; vi+1 ) 2 E .


Ew be the set of edges (vi ; vi+1 ) where (vi ; w) 2 E .

Using the assumption of the theorem, we get jEu j + jEw j 


n. Thus, there is an edge (vi ; vi+1 ) in the intersection Eu \
Ew . Besides, i 6= 0 and i 6= n 2, since (u; w) 62 E .

vi+1
v1

v2

vn

vi

We have found a Hamiltonian cycle in

G0 .

Theorem (Bondy and Chvtal, 1976). Consider a sim-

ple graph G = (V; E ) and let u; v 2 V be non-neighbouring


vertices such that deg(u)+deg(v )  jV j. Then G is Hamiltonian i G [ f(u; v )g is Hamiltonian.

Proof. The direction  G Hamiltonian => G [ f(u; v )g


Hamiltonian is obvious. Proof of the other direction was
given in the proof of Ore's theorem.


Graph G = (V; E ) is called Ore-closed if for any two different vertices u; v 2 V the implication

deg(u) + deg(v )  jV j ==> (u; v ) 2 E


holds.

G0 = (V; E 0 ) is called Ore closure of graph G =


(V; E ) and denoted as O(G) if the following holds:
Graph





G0 is Ore-closed;
E  E 0;
E 0 is the least possible set with the above properties.

Lemma. Let G1 = (V; E1 ) and G2 = (V; E2 ) be Ore-closed


graphs. Then
Proof. Let
we have

G = (V; E1 \ E2 ) is Ore-closed.

u; v

2V

and

degG (u) + degG (v )  jV j. Then

degG1 (u)+ degG1 (v )  jV j and degG2 (u) + degG2 (v )  jV j;


since

degGi (u)  degG (u) and degGi (v )  degG (v ).

G1 and G2 are Ore-closed, we get (u; v )


(u; v ) 2 E2 , implying (u; v ) 2 E1 \ E2 .
As

E1 and

The Lemmma implies that all graphs have Ore closures.

Algorithm (for nding Ore closure). Consider a simple


graph

G = (V; E ).

u; v 2 V such that deg(u) + deg(v )  jV j and


(u; v ) 62 V . If there are no such vertices, output G and

1. Find

stop.

2. Add the edge

(u; v ) to E and return to step 1.

Proposition. The result of the algorithm does not depend


on the choice of vertices

u; v on step 1.

Proof. Assume we can get two dierent outcomes G1 =


(V; E [_ E1 ) and G2 = (V; E [_ E2 ) starting from graph G =
(V; E ) (so that E1 6= E2 ). W.l.o.g. assume E1 nE2 6= ;.

Elements of the set E1 nE2 are added to the graph G1 in


some order as the algorithm proceeds. Let (u; v ) be the
rst one in this order. Let E10  E1 be thet set of all edges
added before the edge (u; v ).

E10  E2 . Thus, in the graph G2 the condition


deg(u) + deg(v )  jV j holds. A contradiction with the
assumption (u; v ) 62 E2 .

We have

Theorem. The algorithm nds Ore closure of graph G.


Proof. The proof follows from these four claims:

1. Edge set of the output graph of the algorithm is a


superset of the edge set of the input graph.
2. The algorithm is monotone, i.e. if G1 = (V; E1 ) and
G2 = (V; E2 ), where E1  E2 , the algorithm turns
them into graphs G01 = (V; E10 ) and G02 = (V; E20 ),
where E10  E20 . The proof is similar to the proof
of the previous proposition.
3. The output graph of the algorithm is Ore-closed.
4. If the input of the algorith is an Ore-closed graph, the
algorithm will output it.

Corollary. A graph is Hamiltonian i its Ore closure is


Hmiltonian.
Proof. This is a consequence of the closure nding algorithm and Bondy-Chvtal theorem.


Corollary. Let G = (V; E ) be a simple graph with jV j =

n  3. If O(G) = Kn then G is Hamiltonian.


Proof.

Kn is Hamiltonian.

Theorem. Let G = (V; E ) be a non-Hamiltonian graph

on n vertices. Then there exists k < n2 such that G has


k vertices with degree at most k and n k vertices with
degree at most n k 1.

Proof. Let O(G) = (V; E 0 ). Since O(G) 6= Kn , there exist


vertices u and w such that (u; w) 62 E 0 . Take u and w so
that the sum degE 0 (u) + degE 0 (w) is maximal.

We have

(u; w)
Let

degE 0 (u) + degE 0 (w)  n 1, since otherwise


E 0 (according to the denition of Ore closure).
U = fu0 j u0 6= u; (u; u0 ) 62 E 0 g
W = fw0 j w0 6= w; (w; w0 ) 62 E 0 g :

W.l.o.g. assume

degE 0 (u)  degE 0 (w). Let k = degE 0 (u).

1.
2.
3.
4.
5.
6.

degE 0 (u) + degE 0 (w)  n 1.


degE 0 (u) + degE 0 (w) is the maximal possible.
k = degE 0 (u)  degE 0 (w).
1. and 3. give k  n 2 1 < n2 .
2. gives degE 0 (w0 )  degE 0 (u) for any w0 2 W . Besides,
degE 0 (u0 )  degE 0 (w) for any u0 2 U .
jU j = n 1 degE0 (u) and jW j = n 1 degE0 (w).
This is proven by a simple counting argument.

jW j  k .
8. 5. gives degE (w0 )  degE 0 (w0 )  degE 0 (u) = k for any
w0 2 W .
We have k vertices with degree  k.
7. 1. and 6. give

1.
4.
5.
6.

degE 0 (u) + degE 0 (w)  n 1.


k

n 1
2

< n2 .

degE 0 (u0 )  degE 0 (w) for any u0 2 U .

jU j = n

1 degE 0 (u).

jU j = n k 1. Thus jU [ fugj = n
For each u0 2 U we get from 5. and 1. that
degE (u0 )  degE 0 (u0 )  degE 0 (w)  n 1

9. 6. gives
10.

11. 4. gives
We have

degE (u)  degE 0 (u) = k 

n 1
2

k vertices with degree  n

n
k

1.

k.
k :

1 k.

Corollary. Consider a graph G = (V; E ) on n vertices


such that for each k < n2 the graph has less than k vertices
with degree at most k or less than n k vertices with degree
at most n k 1. Then G is Hamiltonian.
Proof. From the previous theorem: (A
: A). 

=> B) , (:B =>

The same claim for degree sequences:

Corollary. Consider a graph G = (V; E ) with degree se-

(a1 ; : : : ; an ). If for each k < n2 we have (ak  k) =


> (an k  n k) then G is Hamiltonian.
quence

Call the degree sequence (a1 ; : : : ; an ) Hamiltonian if each


graph G with degree sequence (b1 ; : : : ; bn ) where bi  ai
(1  i  n) is Hamiltonian.

Theorem. Degree sequence (a1 ; : : : ; an ) is Hamiltonian i


for each
Proof.

k<

n we have
2

(ak  k) => (an

( is proven in the previous slide

n

k ).

=> Assume that (a1 ; : : : ; an ) does not satisfy the required


condition. We will construct a graph with degree sequence
 (a1; : : : ; an) that is not Hamitlonian.
If the condition is not satised, we must have a k such that
ak  k and an k  n k 1.

For a given

k the largest such degree sequence is

(k;
: : : ; k; n k
| {z } |
k

1; :{z
::;n k
n 2k

1}; n
1; :{z
: : ; n 1}) :
|
k

A non-Hamiltonian graph with such a degree sequence:

Kk;k
Kn

You might also like