1.

1 Planar Graphs
notes by Nabil H. Mustafa and János Pach

I want to express a radical alternative that I learned from Sir Michael Atiyah. His
view was that the most significant aspects of a new idea are often not contained in the
deepest or most general theorem which they lead to. Instead, they are often embodied
in the simplest examples, the simplest definition and their first consequences.

David Mumford

One major way that graph theory interacts with geometry is through the study of graphs
that can be drawn, or embedded, in Euclidean spaces with certain constraints. For example,
given a graph G = (V, E), one can ask if it can be drawn in the plane in such a way that
no two edges of G cross each other.
We define the notion of a drawing more precisely. v4
An embedding of a graph G = (V, E) in the plane
consists of functions that map its vertices and edges v5
to subsets of the plane. First, the function φV : V → v
3
R2 maps each vertex v ∈ V to a point φV (v) ∈ R2 .
Then for each edge e = (u, v) ∈ E, the continuous v2
function φe : [0, 1] → R2 maps e to a continuous
curve in R2 between the mappings of u and v, i.e.,
φe (0) = φV (u) and φe (1) = φV (v). We will assume v1
that φV is injective, and that no curve for any edge e
passes through any of the points of the vertices, unless the vertex is an endpoint of e.
An embedding of G is called planar if the curves of every two edges of E are disjoint, except
possibly at a common vertex. A graph G = (V, E) is called planar iff there exists a planar
embedding of G. Any such embedding of G partitions R2 into connected components,
called the faces of the embedding. Each face is bounded by elements of V and E. The
unbounded face is called the outer face of the embedding, and all other faces are the inner
faces. The size of a face is its number of bounding vertices.

B ASIC P ROPERTIES OF P LANAR G RAPHS

One of the most basic facts about planar graphs is the following:

Lemma 1.1 (Euler’s formula). Let G = (V, E) be a connected planar graph, and consider a
planar embedding with the set of faces F . Then

|V | − |E| + |F | = 2.

8
Proof. The proof is by induction on the number of cycles in G. The base case is when G
is a tree, in which case |F | = 1 (the outer face), |E| = |V | − 1 and the relation holds.
Otherwise, let G0 = (V 0 , E 0 ) be the graph obtained by removing any edge e ∈ E belonging
to a cycle in G, and let F 0 be the resulting set of faces. Then |V 0 | = |V |, |E 0 | = |E| − 1
and |F 0 | = |F | − 1. By induction, |V 0 | − |E 0 | + |F 0 | = 2, and we get the desired relation for
G.
Euler’s formula implies that the number of faces is the same for any embedding of a planar
graph. It also implies an upper-bound on the number of edges in a planar graph:
Lemma 1.2. Let G = (V, E) be a planar graph on n vertices and m edges, and where the size
k
of each face is at least k ≥ 3 for an integer k. Then m ≤ k−2 (n − 2). In particular, any planar
graph has at most 3n − 6 edges.

Proof. Consider any planar embedding of G, and let F be the set of faces of this embedding.
The proof is by double-counting the sum of the sizes of the faces of G:
X

k · |F | ≤ size of f ≤ 2m.
f ∈F

From Euler’s relation, |F | = 2 − n + m,

k · (2 − n + m) ≤ 2m
k
m ≤ (n − 2).
k−2

P LANARITY T ESTING C RITERION

Clearly planar graphs are a small subset of all possible graphs. Two important examples
of non-planar graphs are K5 and K3,3 . From Lemma 1.2, it follows that K5 is not planar,
as it has 5 vertices yet with greater than 3 · 5 − 6 = 9 edges; similarly K3,3 is not planar,
as it has 6 vertices, no triangular face (so k = 4), and yet greater than 2 · 6 − 4 = 8 edges.
Surprisingly, any graph not containing a subdivision† of these two subgraphs is planar:
Theorem 1.3 (Kuratowski’s theorem). A graph is planar iff it does not contain a subdivision
of K5 and K3,3 .

Kuratowski’s theorem can be used to prove another important theorem, the Hanani-Tutte
theorem, which states that if a graph G can be embedded in the plane such that every edge
has an even number of crossings, then G is planar. Here we prove a slightly weaker variant
directly by induction.
†
Recall that given G = (V, E) and an edge e = {u, v} ∈ E, the subdivision of e yields a graph G0 = (V 0 , E 0 ),
where V 0 = V ∪ {w}, and E 0 = E \ {{u, v}} ∪ {{u, w}, {w, v}}. A graph H is a subdivision of G if it can be
derived from G by a sequence of subdivisions.

9
Theorem 1.4 (Weak Hanani-Tutte theorem). Let G = (V, E), |V | = n, |E| = m, be a graph
for which there exists an embedding such that every pair of edges cross an even number of
times. Then G is a planar graph.

Proof. Fix an embedding such that every pair of edges cross an even number of times.
Then we will inductively modify this embedding to G G0
construct a planar embedding of G. For the induc-
vj
tive argument to work, we will prove the statement
in a slightly stronger form: (i) we show it more
generally for multigraphs, and (ii) that the planar
embedding has the same sequence of edges around
each vertex as the given embedding.
vi vi,j
Let e = {vi , vj } be any edge in G. Con-
tract e by moving vj to the position of vi to
get a new graph G0 with the merged vertex vi,j
at the position of vi . The embedding for each
edge previously incident to vj is updated by ex-
tending it along the edge e to reach vi . There might exist self-intersections
of an edge in this modified embedding, but they can be removed (see figure).
Note also that even if G was a graph, G0 could be a multigraph (as a vertex with an edge to
both vi and vj in G now has two edges to vi in G0 ). Every pair of edges still cross an even
number of times in this embedding of G0 , and so by induction, G0 has a planar embedding.
This embedding can now be extended to an embedding of G by splitting vi,j in G0 to two
vertices vi , vj of G connected by a small-enough line segment. Here we need the stronger
inductive hypothesis that the embedding of G0 preserves the order of edges around each
vertex, as then all the edges incident to vj are contiguous around vi,j (same for the edges
incident to vi ) in the planar embedding of G0 and so can be assigned to vj after the split
without causing any additional intersections.
The base case is a graph G0 consisting of a single
vertex v with multiple self edges. Then, as every
pair of edges cross an even number of times, there
vj
must exist an edge e with an embedding that leaves
and enters v consecutively in the clockwise ordering vi vi,j
† 0
of the edges leaving/entering v . Then G \ {e} has, by induction, a planar embedding with
the same sequence of edges around v. Adding e back, by a small-enough curve that does
not intersect any other edge, gives the required embedding for G0 .

†
Take the edge e with the smallest number of edges in the clockwise ordering of its two endpoints around v.
If there is an edge that leaves between the two endpoints of e, it must enter also between the two endpoints,
but this contradicts the choice of e.

10
S OME O THER P ROPERTIES OF P LANAR G RAPHS

A proper coloring of a graph G = (V, E) is an assignment of colors to the vertices of V such

that the two vertices of every edge e ∈ E have different colors. The chromatic number of
a graph G, denoted χ(G), is the minimum number of colors required for a proper coloring
of G. It is easy to P see that any planar graph G = (V, E) has χ(G) ≤ 6: G has at most
3|V | − 6 edges, i.e., v deg(v) ≤ 6|V | − 12, and so there exists a vertex with degree at most
5. Inductively color G \ {v} with 6 colors, and then assign v the color missing among the
neighbors of v. This bound can be improved, and the following famous theorem gives a
precise answer for the chromatic number of planar graphs:

Theorem 1.5 (Four color theorem). Let G be a planar graph. Then χ(G) ≤ 4. Furthermore,
there exist planar graphs G for which χ(G) = 4† .

Since vertices of the same color class are independent, an immediate corollary of this
theorem is:

Corollary 1.6. Any planar graph on n vertices has an independent set of size at least n/4.

Let C be a convex polyhedron in R3 and let P be its set of vertices. Define the graph G =
(P, E) with the following set of edges: {pi , pj } ∈ E iff the segment pi pj lies on the bound-
ary, denoted ∂C, of C. G is called the 1-skeleton of C. For a point q ∈ R3 , the stereographic
v1

projection πq : C → R2 w.r.t. q projects ∂C onto R2 , v1

v2
where πq (p) is the intersection of the line qp with the v2

xy-plane. Note that the pre-image of any point p0 ∈ v4

R2 consists of at most two points in ∂C. In particular, v4

if one picks q to be close enough to a facet of C, then

the edges of E map to pairwise non-crossing set of v3 v3

segments in R2 , called the Schlegel diagram of C). Thus we can conclude that:

Theorem 1.7. The 1-skeleton of a convex polytope in R3 is a planar graph.

Q UESTIONS

(solutions)

1. Let L be a set of n lines in the plane, no two of which are parallel, and with no three
n


passing through a common point. Let P be the set of 2 intersection points of L.
Let G = (P, E) be the graph on P such that (pi , pj ) ∈ E iff pi and pj are consecutive
intersection points along some line l ∈ L.

(a) Using the existence of a low-degree vertices in G, prove that χ(G) ≤ 5.

†
For example, K4 is a planar graph and requires 4 colors.

11
 (b) Prove that χ(G) ≤ 3.

2. Let G = (V, E1 ∪ E2 ) be a graph, where (V, E1 ) and (V, E2 ) are planar graphs. Then
show that χ(G) ≤ 12.

3. Prove that a planar graph is bipartite if and only if in its planar embedding, all faces
have even length.

4. Let G = (V, E) be a planar graph with k connected components. Let F be the set of
faces in a fixed planar embedding of G. Then prove that |V | − |E| + |F | = k + 1.

5. Let P be a set of n points in R3 in convex position, and let G be the 1-skeleton of

P . Then prove that for any halfspace H, the subgraph of G induced by the points in
H ∩ P is connected.

6. Let C be a convex polyhedron with 12 vertices, and where each facet is a triangle.
Show that if each facet of C is labelled with a nonnegative integer such that the
sum of these integers over all faces is 39, then there must exist a vertex whose two
adjacent faces have the same label.

7. Let L be a set of n lines in the plane. The arrangement of L induces a set F of faces.
Let G = (F, E) be the planar graph on F where {fi , fj } ∈ E iff the two faces fi , fj are
adjacent in the arrangement, i.e., their boundaries share a common line in L. Prove
that χ(G) ≤ 2. Show the same is true for regions induced by circles instead of lines.

12