4.

4 Eulerian graphs

Here is one of the oldest problems concerning graph drawing.

[Problem.] Draw a graph with a single closed line

without lifting the pencil from the paper. (한붓그리기)

Mathematically, this can be formalized as follows.

A walk where all the edges are distinct is called a tour.
A tour (v0, e1, v1, …, em−1, vm−1, em, vm) in a graph is closed if v0 = vm.
A tour is Eulerian if it uses all the edges of the graph.

We want to determine whether a graph contains an Eulerian tour or not.

First consider ‘closed tour’
4.4 Eulerian graphs

What are necessary conditions for having a closed Eulerian tour?

4.4 Eulerian graphs

What are necessary conditions for having a closed Eulerian tour?

[Theorem 1.8.1 (Euler 1736)]

A graph G has a closed Eulerian tour if and only if
G is connected and every vertex has even degree.
[Theorem 1.8.1 (Euler 1736)]
A graph G has a closed Eulerian tour if and only if
G is connected and every vertex has even degree.
PF) (=>) This is easy. If a vertex appears k times in an Eulerian tour (counting first=end),
then its degree must be 2k.

(<=) Suppose every vertex has even degree.

We prove by induction on | E(G) | . If | E(G) | = 0, then it is trivial.
Assume | E(G) | > 0.

Note that G has a cycle C. Let P be a longest path in G.

Then the last vertex v should have a neighbor in V(P − v). This creates a cycle.
If E(G) = E(C), then we are done. Suppose G − E(C) has an edge.
PF) (<=)

We choose a maximal closed walk W in G that traverses every edge of G at most once.
Let F be the set of edges used in W.
If F = E(G), then we are done. We may assume that E(G)∖F ≠ ∅.

Observe every vertex of G − F has even degree.

Let U be a component of G − F containing an edge.
By induction U admits an Euler tour W′.
By concatenating W with W′, we get a longer closed walk,
that traverses every edge of G at most once, a contradiction.


[Theorem 1.8.1 (Euler 1736)]
A graph G has an Eulerian tour that is not closed if and only if
G is connected and has exactly two vertices of odd degree.

PF) How to reduce to the original theorem?

Directed graphs

We introduce directed graphs (or digraphs), where every edge has a direction.

[Definition] A directed graph G is a pair (V, E), where E ⊆ V × V.

The ordered pair (x, y) ∈ E is called a directed edge.

We say that a directed edge (x, y) has head y and tail x.

It is natural to define a directed path, walk, tour, closed/ Eulerian tour in a similar way,
respecting directions.
Directed graphs

Now, we need a neighborhood concept for directed graphs.

+
For a vertex v, the number of edges whose tail is v is denoted by degG (v), and
−
the number of edges whose head is v is denoted by degG (v).

[Proposition 4.5.2]
A directed graph G is contains a closed Eulerian tour if and only if
it is connected (when we forget the directions) and $
+ −
degG (v) = degG (v) for each vertex v of G.

We can prove in a similar way. We leave it as an exercise.

An application.

[Problem] Find a cyclic sequence of digits 0 and 1, as long as possible

such that no two k-tuples of consecutive digits are the same.

Let ℓ(k) denote the maximum possible number of digits in such a sequence.

Can you determine ℓ(k)?

k
Clearly, ℓ(k) ≤ 2 , which is the all possible digits in k-tuples.
 k
[Proposition] For each k ≥ 1, we have ℓ(k) = 2 .

k k
PF) As ℓ(k) ≤ 2 , it is enough to construct a cyclic sequence of length 2 ,
k
which shows ℓ(k) ≥ 2 .
If k = 1, then it is trivial. Assume k ≥ 2.

Define a graph G such that

- V(G) is the set of all (0,1)-sequences of length k − 1,
- Directed edges of G are all pairs of (k − 1)-digit sequences of the form
((a1, a2, …, ak−1), (a2, …, ak−1, ak))

k
Directed edges are in bijection with (a1, …, ak), and so | E(G) | = 2 .
PF) We want to see that this directed graph has a closed Eulerian tour.

− +
It is straightforward to verify that degG (v) = degG (v) = 2 for all vertices v.
Also, it is connected when you forget directions
(any vertex can be made to 0-sequence 1-by-1).

So, it has a closed Eulerian tour.

Now, let (e1, …, eK ) be the sequence of edges in a closed Eulerian tour.

i i i i i
Each edge ei has the form ((a1, …, ak−1), (a2, …, ak−1, ak)).

1 2 K
We obtain a sequence (a1 , a1 , …, a1 ).
Each subsequence of k consecutive digits corresponds to one directed edge,
and since no directed edge is repeated, no two k-digits are the same.
4.6 2-connectivity

Sometimes, you want to measure the 'connectivity' of a graph.

For instance, you made a network so that when you lost some of nodes,
you still want to make all connected.

This property is captured by 'vertex connectivity' or 'edge connectivity'.

[Definition 4.6.1] A graph G is 2-connected if it has at least 3 vertices, and

by deleting any single vertex, we obtain a connected graph.

[General definition] A graph G is k-connected G is if it has at least k + 1 vertices, and

by deleting any set of at most k vertices, we obtain a connected graph.

We can similarly define the 'edge version' of the connectivity, but we skip the definition.
[Question] What can we expect for 2-connected graphs?
Some graph operations

We defined G − e and G + f for edges e, f, and G − v for a vertex v.

(Edge subdivision)
Let e = xy be an edge of G.
G % e is the graph obtained from G by removing e and
adding a new vertex v and new edges vx, vy.

A graph G′ is a subdivision of a graph G if

G′ is obtained from G by repeatedly subdividing edges.

For a path P and two vertices v, w, we write vPw be the subpath from v to w.


[Theorem 4.6.3] A graph G is 2-connected if and only if
for any two vertices v, w of G, there is a cycle containing v, w.

This theorem is a particular case of a very important result called ‘Menger’s theorem’.

Let G be a k-connected graph.

Then for any two vertices v, w of G,
there are k internally disjoint paths from v to w.
(internally disjoint = disjoint except endpoints)
[Theorem 4.6.3] A graph G is 2-connected if and only if
for any two vertices v, w of G, there is a cycle containing v, w.

(<=) This direction is clear.

Suppose G is not 2-connected.

Then there exists a vertex x such that G − x is disconnected.
So, there are two vertices v, w in G − x contained in distinct components.

Then there is no cycle containing v, w in G, a contradiction.

[Theorem 4.6.3] A graph G is 2-connected if and only if
for any two vertices v, w of G, there is a cycle containing v, w.

(=>) We prove by induction on distG(v, w).

Assume that distG(v, w) = 1. That is, vw is an edge.

We claim that G − vw is connected. Suppose that it is disconnected.
Then there are two components Cv, Cw, where
Cv contains v and Cw contains w.

As G has at least three vertices, one of Cv and Cw has at least two vertices.
Then vertices of v, w in the component disconnected the graph, a contradiction.
(=>) Assume that distG(v, w) ≥ 2.
Let P = v1v2⋯vk be a shortest path from v = v1 to w = vk.
Note that distG(v1, vk−1) = k − 1 (if you can shorten, you can shorten to vk).

By induction, there is a cycle C containing v1 and vk−1.

If C contains vk, then we are done.
Thus, we may assume that C does not contain vk.

Note that G is 2-connected. So, G − vk−1 is connected.

Let P be a path from vk to C − vk−1 in G − vk−1.

Let P1, P2 be the two paths from v to vk−1 in C, and

assume that the endvertex w of P other than vk lies in P2.
Then P1 ∪ G[{vk−1, vk}] ∪ P ∪ vP1w is a cycle containing v and vk.
The simplest 2-connected graphs are the cycles.
All the others can be constructed inductively from a cycle by adding paths:

Let G be a graph and H be a subgraph.

A path P in G is an H-path if
P has at least two vertices and it meets H exactly in its endvertices.

[Theorem] A graph is 2-connected if and only if

it can be constructed from a cycle by successively adding H-paths to
graphs H already constructed.

PF) (<=) Every graph constructed as described is 2-connected.

[Theorem] A graph is 2-connected if and only if
it can be constructed from a cycle by successively adding H-paths to
graphs H already constructed

PF) (=>) Suppose that G is 2-connected.

Note that G contains a cycle. Let H be a maximal subgraph constructible as above.

Any edge xy ∈ E(G)∖E(H) with x, y ∈ V(H) would be an H-path.

So, H has no such an edge. It means that H is an induced subgraph of G.

Assume H ≠ G. Then there is a vertex in V(G)∖V(H).

As G is connected, there is a vertex v ∈ V(G)∖V(H) and w ∈ V(H) that are adjacent.
As G is 2-connected there is a path P from v to H in G − w.
Then P with vw is an H-path. This contradicts the maximality of H.
4.7 Triangle-free graphs: an extremal problem

Let us consider a graph G with n vertices.

What is the maximum number of edges?

(2)
n
The answer is easy : .

QUESTION : What is the maximum number of edges of a graph with no triangles?

4.7 Triangle-free graphs: an extremal problem

Let T(n) be the maximum possible number of edges in a graph on n vertices with no
triangles.

T(1) = 0
T(2) = 1
T(3) = 2

It is also easy to check that T(4) = 4.

What is T(5)?
 ⌊4⌋
2
n
[Theorem 4.7.1 (Mantel 1907)] For every n ≥ 1, we have T(n) = .

PF) We prove by induction on n. It is clear when n ≤ 2. Suppose n ≥ 3.

Let G be an edge-maximal without triangle.
Clearly, G has an edge. Let uv be an edge.

⌊ 4 ⌋
2
(n − 2)
By the induction hypothesis, G − {u, v} has at most edges,

and each vertex of G − {u, v} has at most 1 neighbor in {u, v}.

⌊ 4 ⌋ ⌊4⌋ ⌊4⌋
2 2 2
(n − 2) n n
Hence | E(G) | ≤ + (n − 2) + 1 = . This shows T(n) ≤
 ⌊4⌋
2
n
[Theorem 4.7.1 (Mantel 1907)] For every n ≥ 1, we have T(n) = .

⌊4⌋
2
n
PF) To show T(n) ≥ , we need to give an example having this many edges.

2
If n = 2k + 1, then Kk,k+1 has k + k edges.
2
If n = 2k, then Kk,k has k edges.

In fact, we could show that this is the only possible example,

by modifying the proof. (See Theorem 4.7.2)
Can we extend this to general Kn?

Let H be a graph and n ≥ | V(H) | .

How many edges will suffice to force and H subgraph in any graph on n vertices?

A graph G ⊉ H on n vertices with the largest possible number of edges

is called extremal for n and H. Its number is denoted by ex(n, H).

As a case, we consider H = Kr with r > 1.

Ques) What would be the extremal graph for n and Kr?

 The unique complete (r − 1)-partite graphs on n ≥ r − 1 vertices
whose partition sets differ in size by at most 1 are called Turán graphs.
r−1
It will be denoted by T (n) and its number of edges by tr−1(n).

r−1
For n < r − 1, we define T (n) as Kn.

[Theorem 7.1.1 (Turán 1941)]

For all integers r, n with r > 1,
r−1
every graph G ⊉ Kr with n vertices and ex(n, Kr) edges is T (n).

The proof is exactly same. Try!