Math 3322: Graph Theory1 Mikhail Lavrov

Lecture 6: The degree of a vertex

August 29, 2024 Kennesaw State University

1 Degrees and the handshake lemma

First, some definitions. The degree of a vertex v in a graph G is the number of edges that have v
as an endpoint. (We also say such edges are incident on v.) We write the degree of v as deg(v). If
we have multiple graphs that share the vertex v for any reason, we might write degG (v) to specify
that we mean the degree in G.
Here is a graph with vertices labeled according to their degree:
1 4 4 3

1 0 3 2

There’s a lot of associated terminology. A vertex with degree 0 is called an isolated vertex, and a
vertex with degree 1 is sometimes called a leaf. A graph G has a maximum degree (the largest
degree of any vertex) and a minimum degree (the smallest degree of any vertex).
We write ∆(G) for the maximum degree and δ(G) for the minimum degree. Graph theory uses a
lot of Greek letters for properties of graphs; this is only the beginning.
The first tool we’ll need to make use of degrees is the Handshake Lemma (also known as the degree
sum formula).
Lemma 1.1. In any graph G, the vertex degrees add up to twice the number of edges:
X
degG (v) = 2|E(G)|.
v∈V (G)

Proof. Many proofs exist; for the sake of practice, let’s do a proof by induction. We will prove that
for any graph with m edges, the sum of degrees is 2m, by induction on m.
The base case is m = 0. Here, we have a graph with no edges. No matter how many vertices we
have, their degrees are all 0, the sum of the degrees is 0, and 2m is also 0.
Assume that the degree sum formula holds for all (m − 1)-edge graphs. Let G be a graph with
m ≥ 1 edges, and let xy be any edge of G. We can apply the inductive hypothesis to G − xy (the
graph we get by deleting edge xy from G), a graph with m − 1 edges.
1
This document comes from the Math 3322 course webpage: http://facultyweb.kennesaw.edu/mlavrov/
courses/3322-fall-2024.php

1
What is the relationship between the degrees of G and the degrees of G − xy? Both x and y have
one extra incident edge in G they don’t have in G − xy: edge xy itself. So
degG (x) = 1 + degG−xy (x) and degG (y) = 1 + degG−xy (y).
For any other vertex v, G − xy and G have the same number of edges, so we have
degG (v) = degG−xy (v).
Also, G − xy and G have the same set of vertices. So if we add up the vertex degrees in G − xy
and G, the result is that
X X
degG (v) = 2 + degG−xy (v).
v∈V (G) v∈V (G−xy)

Applying the inductive hypothesis, we get that the degree sum in G − xy is 2(m − 1), so the degree
sum in G is 2(m − 1) + 2 = 2m.
By induction, the degree sum formula holds for all graphs.

Let’s answer some quick questions using the handshake lemma!

Q1. The hypercube graph Qn has 2n vertices; how many edges does it have?
A1. Every vertex has n neighbors: we can modify a bit sequence x1 x2 . . . xn in n places. So the
sum of degrees is n · 2n , and therefore the number of edges is n · 2n−1 .
Q2. In particular, Q3 has 8 vertices of degree 3. Can we have a 7-vertex graph where all the
vertices have degree 3?
A1. No: then the degree sum would be 7 · 3 = 21, so there would be 10.5 edges, which is
impossible.
Q3. A soccer ball has 12 black pentagonal panels (and some white hexagonal panels I’m too lazy
to count). Panels are stitched along their edges, and meet at corners; at each corner, a pentagon
and two hexagons meet. How many edges are there where two panels meet?
A3. Each black pentagon has 5 corners, which will be the 12 · 5 = 60 vertices of our graph; the
edges will be the edges where panels meet. Here, each vertex has degree 3, so the sum of degrees
is 60 · 3 = 180, and there are 90 edges.
(Some slightly fancier logic can convince us that there are 20 hexagons; see the practice problems
at the end of the lecture notes.)
Q4. Suppose you have a graph G with 9 vertices and 20 edges. What can the minimum degree
δ(G) of this graph be?
A4. Since the sum of the degrees is 2 · 20 = 40, the average of the degrees is 40 9 ≈ 4.44. So the
minimum degree can be at most 4: if every vertex had degree 5 or more, then the sum of degrees
would be at least 5 · 9 = 45. We can convince ourselves that it’s possible to have a minimum degree
of 0, 1, 2, 3, or 4.
We can generalize the answer to the first question. Let’s say that a vertex v is even if deg(v) is an
even number, and odd if deg(v) is an odd number. Then:

2
Corollary 1.2. Every graph G must have an even number of odd vertices (possibly 0).

Proof. By the handshake lemma, the sum of degrees is always an even number (twice the number
of edges.) If you subtract of all the even degrees, you still have an even number. So the sum of all
odd degrees is even.
The sum of an odd number of odd values is odd. The sum of an even number of odd values is even.
So there must be an even number of odd degrees to sum, if we want an even total.

2 Degrees and cycles

The following theorem will be very useful to us in a couple of weeks:
Theorem 2.1. Every graph G whose minimum degree δ(G) is at least 2 contains a cycle.
The intuition for this theorem is as follows. Just start at any vertex of G and walk around, taking
care not to leave a vertex the way you entered it. Eventually you will run into a vertex you’ve seen
before. The first time that happens, your trajectory from that vertex and back forms a cycle.
Whenever we have an intuition of the form “keep doing this thing until it does what we want”, this
suggests a proof with the extremal principle. Just take (by the extremal principle) the situation in
“this thing” has been done for as long as it can without “doing what we want”. Then it is forced
to “do what we want” in the very next step.
In this particular case, if we walk around for as long as we possibly can without revisiting a vertex,
what we’re getting is a very long path. This suggests the following proof:

Proof. Let (v1 , v2 , . . . , vk ) be a longest path in G. (We know a longest path exists because there’s
an upper limit to the length of a path: the number of vertices in G.)
We know δ(G) ≥ 2 and therefore in particular deg(v1 ) ≥ 2. Is it possible that v1 has a neighbor w
which is not one of v2 , . . . , vk ? It’s not! In that case, (w, v1 , v2 , . . . , vk ) would be a longer path.
So v1 has at least two neighbors, which are all in the set {v2 , v3 , . . . , vk }. One of v1 ’s neighbors is
v2 : the next vertex on the path. This doesn’t help us. But there must be another vertex vi with
i > 2 which is adjacent to v1 .
Then (v1 , v2 , . . . , vi , v1 ) is the cycle we wanted.

3 Average and minimum degree

Even if you have lots and lots and lots of edges, your minimum degree can be very small. For
example, a 100-vertex graph might consist of a 99-vertex complete graph and a single isolated
vertex. This has 4851 edges, which is close to the maximum number of edges a 100-vertex graph
can have: 4950. And yet the minimum degree is 0, and we can’t apply any nice results like
Theorem 2.1 to this graph.
Here’s a way to partially solve this problem:

3
Theorem 3.1. Let G be a graph with average degree at least d. Then G contains a subgraph H
with δ(H) > d2 .
This could be done using another application of the extremal principle: if you pick a subgraph
H with the highest average degree, then it turns out to work. Rather than do this, we will use
induction, so that we get to practice induction. (Take note of how we structure this proof to avoid
the “induction trap”!)

Proof. We induct on n, the number of vertices in G.

When n ≤ d, the theorem “holds trivially”: that is, when n ≤ d, the highest possible degree in G
is d − 1, so G cannot have average degree d or more to begin with! We could use that as our base
case, but it’s a bit more satisfying to use n = d + 1.
In this case, the highest possible degree in the graph is d. The average degree can only be this high
if every vertex has degree d: if G = Kd+1 . In this case, G itself is the subgraph H we’re looking
for. This base case also holds.
Either way, suppose that the theorem holds for all (n − 1)-vertex graphs with average degree at
least d. Let G be an n-vertex graph with average degree at least d.
At this point, let’s say something about average degree. If G has vertices v1 , v2 , . . . , vn , then the
average of the degrees is

deg(v1 ) + deg(v2 ) + · · · + deg(vn ) 2|E(G)|

= .
n n
So the statement “G has average degree at least d” is equivalent to the statement “G has at least
1
2 nd edges.

We’re assuming G has at least 21 nd edges. Also, if δ(G) > d2 , then we are already done: we were
looking for a subgraph with this minimum degree, and G itself can be that subgraph! So assume
that G has a vertex v with deg(v) ≤ d2 .
In that case, let G′ = G − v: the graph obtained from G by deleting v. We know that G′ has n − 1
vertices and at least 12 nd − d2 edges: we started with at least 12 nd edges, and we lost at most d2 of
them from deleting v. This simplifies to 12 (n − 1)d, so G′ has at least 12 (n − 1)d edges, which means
G′ has average degree at least d, too!
By applying the inductive hypothesis to G′ , we learn that G′ has a subgraph H with δ(H) > d2 .
Since H is a subgraph of G′ , and G′ is a subgraph of G, we have found the subgraph of G we
wanted.
By induction, the theorem holds for graphs with any number of vertices.

Corollary 3.2. If G has n vertices and at least n edges, then G contains a cycle.

Proof. If G has n vertices and at least n edges, it has average degree at least d = 2. By Theorem 3.1,
G has a subgraph H with δ(H) > d2 = 1. If δ(H) > 1, then δ(H) ≥ 2, so by Theorem 2.1, H
contains a cycle. This is also a cycle in G.

4
4 Practice problems
1. In an earlier lecture, we gave an abstract definition of the Petersen graph: the vertices
correspond to two-element subsets of {1, 2, 3, 4, 5}, and two vertices are adjacent if the subsets
are disjoint.
(a) Without going back to looking at a drawing of the Petersen graph, compute the number
of edges it has directly from the definition.
(b) Suppose that we generalize the graph: the vertices correspond to two-element subsets of
{1, 2, . . . , n}, and two vertices are adjacent if the two subsets are disjoint.
What is the degree of a vertex in this graph? How many edges does the graph have?
2. Suppose that G is a graph with 5 vertices and 7 edges. For which pairs (a, b) is it possible
that δ(G) = a and ∆(G) = b?
For the cases where it is possible, give an example. For the cases where it is not possible,
explain why not.
3. The five Platonic solids are:
• The tetrahedron, which has 4 vertices and 4 triangular faces, with 3 faces meeting at
every corner.
• The cube (or hexahedron), which has 8 vertices and 6 square faces, with 3 faces meeting
at every corner.
• The octahedron, which has 6 vertices and 8 triangular faces, with 4 faces meeting at
every corner.
• The dodecahedron, which has 20 vertices and 12 pentagonal faces, with 3 faces meeting
at every corner.
• The icosahedron, which has 12 vertices and 20 triangular faces, with 5 faces meeting at
every corner.
In each case, find the number of edges where two faces meet. (Later in the semester, we will
explore further constraints on the Platonic solids.)
4. Suppose G is a graph whose minimum degree δ(G) is at least 10. Then we can improve on
the conclusion of Theorem 2.1: not only does G contain a cycle, but G contains a cycle of
length at least 11.
Prove this! The structure of the proof is very similar, except that instead of saying “So v1
has at least two neighbors, . . . ” we can say “So v1 has at least 10 neighbors, . . . ” and go from
there. How can we use those neighbors to get a long cycle?
5. (a) Adapt the proof idea of Lemma 1.1 to prove the following result: if G is a bipartite
graph with bipartition (A, B), then
X X
deg(v) = deg(v).
v∈A v∈B

5
 (b) Consider the soccer ball we used as an example earlier in the lecture notes.
Define a bipartite graph whose vertices are the pentagons and hexagons on the soccer
ball, and use part (a) of this question to count the hexagons. (You’ll need to know that
each hexagon borders three pentagons.)
6. (a) Suppose that G is a 10-vertex graph with δ(G) = 5. Is it possible that G is not connected?
If so, give an example. If not, explain why not.
(b) Suppose that G is an 11-vertex graph with δ(G) = 6. Is it possible that G is not
connected?
If so, give an example. If not, explain why not.
(c) Generalize this argument: if G is an n-vertex graph, for what value of δ(G) can we be
certain that G is connected? Prove your answer.
7. An open problem called Conway’s 99-graph problem is to determine whether there is a 99-
vertex graph with the following properties:
• Every two adjacent vertices have exactly one common neighbor;
• Every two non-adjacent vertices have exactly two common neighbors.
If such a graph exists, then every vertex in it must have the same degree, d. What is d?