7 Planarity

Embeddings & Euler’s Formula

Planar Embedding: If G is a graph, an embedding of G in the plane is a function φ which
assigns each vertex of G a distinct point in the plane and assigns to each edge e with ends
u, v a simple rectifiable curve with ends φ(u) and φ(v) so that this curve minus its ends is
disjoint from the image of V (G) ∪ (E(G) \ {e}).

Planar and Plane: A plane graph is a graph G together with an embedding of G in the
plane. A graph is planar if there exists an embedding of it in the plane.

Faces: If G is a plane graph, then the space obtained from the plane by removing all points
in the image of G consists of finitely many connected components, each of which is called a
face of G. We let F (G) denote the set of all faces of G. Every face a ∈ F (G) is bounded
by a closed walk (not necessarily a cycle) called a boundary walk. The size of a, denoted
size(a), is the length of this walk, and an edge or vertex is incident with a if it appears in
this walk. Note that every edge appears exactly twice in the boundary walks of the faces.

Theorem 7.1 (Euler’s Formula) If G is a connected plane graph, then

|V (G)| − |E(G)| + |F (G)| = 2.

Proof: We proceed by induction on |E(G)|. If |E(G)| = 0, then |V (G)| = 1 = |F (G)| so the

formula holds. For the inductive step, let G be a plane graph with |E(G)| > 0 and choose
e ∈ E(G). If e is a non loop edge, then contracting e results in a plane graph (namely
G · e) with the same number of faces but one fewer edge and one fewer vertex. So, the result
follows by applying induction to G · e. If e is a loop edge, then e separates the plane1 , so
there are two distinct faces incident with e, and deleting e results in a plane graph (namely
G − e) with the same number of vertices but one fewer edge and one fewer face. So, the
result follows by applying induction to G − e. ¤
1
The seemingly obvious statement that every simple closed curve in the plane separates it into two regions
is the Jordan Curve Theorem and is surprisingly difficult to prove. We shall take this as an assumption.
 2
Duality
Duality: Let G be a plane graph, and construct a new plane graph G∗ as follows. For each
face a ∈ F (G), add a vertex a∗ in a. For each edge e ∈ E(G) which lies in the boundary walk
of the faces a, b, add an edge from a∗ to b∗ which crosses e but is otherwise disjoint from the
image of G (if e appears twice in the boundary walk of a, then e∗ is a loop at a∗ ). This can
be done so that G∗ is a plane graph, and we call any plane graph constructed in this manner
a dual of G. As suggested by the name, if G∗ is a dual of G, then G is a dual of G∗ . This
gives correspondences V (G) ∼ F (G∗ ), E(G) ∼ E(G∗ ), and F (G) ∼ V (G∗ ), and for any set
X of vertices, edges, or faces of G we let X ∗ denote the corresponding set of elements in G∗ .

Observation 7.2 Let G, G∗ be dual plane graphs.

(i) If a ∈ F (G), then size(a) = d eg(a∗ ).

P
(ii) a∈F (G) size(a) = 2|E(G)|.

Proof: (i) follows immediately from the definitions. For (ii), note that
X X
2|E(G)| = 2|E(G∗ )| = d eg(a∗ ) = size(a). ¤
a∗ ∈V (G∗ ) a∈F (G)

Proposition 7.3 If G, G∗ are connected dual plane graphs, then G is bipartite if and only
if G∗ is Eulerian.

Proof: By possibly adding parallel edges to G, we may assume that every edge of G appears
in the boundary walks of two distinct faces. For the ”only if” direction, note that if G is
bipartite, then every boundary walk has even length, so every vertex of G∗ has even degree.
For the ”if” direction, assume G∗ is Eulerian, let C be a cycle in G, and let A ⊆ F (G) be
the set of faces which lie inside C. Now, E(C) is the symmetric difference of the edge sets in
the boundary walks of faces in A. Since each such boundary walk has even length, |E(C)|
is even as well. ¤

Bond: A bond is a nonempty edge cut which is minimal (with respect to inclusion). In a
connected graph, an edge cut δ(X) is a bond if and only if setting Y = V (G) \ X, both
G − X and G − Y are connected.
 3
Proposition 7.4 If G = (V, E) and G∗ = (V ∗ , E ∗ ) are dual plane graphs, then

(i) C ⊆ E is the edge set of a cycle (in G) if and only if C ∗ is a bond (in G∗ ).

(ii) if {S, T } is a partition of E, then (V, T ) is a spanning tree of G if and only if (V ∗ , S ∗ )

is a spanning tree of G∗ .

Proof: For (i), we begin by proving the ”only if” direction. Let C be the edge set of a cycle
in G. Then the image of C separates the plane into two connected components, and this
gives a partition of the faces of G. Since any two faces in the same component may be joined
by a curve in the plane disjoint from V (G) ∪ C, it follows that C ∗ is a bond of G∗ . For the
”if” direction, suppose that C ∗ is a bond of G∗ . Then C ∗ gives us a partition of the vertices
of G∗ (or equivalently a partition of the faces of G) into two connected components. The
edges of G which lie on the boundary of these two components are precisely those in C, so
C is the edge set of a cycle, as desired.
For (ii), we have:

(V, T ) is a spanning tree of G.

⇔ T does not contain the edge set of a cycle and S does not contain a bond.
⇔ T ∗ does not contain a bond and S ∗ does not contain the edge set of a cycle.
⇔ (V ∗ , S ∗ ) is a spanning tree of G∗ . ¤

Another proof of Euler’s Formula: Let G be a connected plane graph with dual G∗ .
Choose a partition {S, T } of E(G) so that T is the edge set of a spanning tree in G. Then
S ∗ is the edge set of a spanning tree in G∗ so we have

2 = (|V (G)| − |T |) + (|V (G∗ )| − |S|) = |V (G)| − |E(G)| + |F (G)|. ¤

Applications of Euler’s Formula

Triangulation: A triangulation of the plane is a plane graph in which every face has size
three.

Lemma 7.5 If G is a simple planar graph, then

 4
(i) 6 ≤ 3|V (G)| − |E(G)| with equality for triangulations.

(ii) 4 ≤ 2|V (G)| − |E(G)| if G has no triangle.

Proof: By possibly adding edges, we may assume that G is connected. For (i), note that if
every face has size ≥ 3, part (ii) of Observation 7.2 implies 2|E(G)| ≥ 3|F (G)|. Plugging
this into Euler’s Formula gives 6 = 3|V (G)| − 3|E(G)| + 3|F (G)| ≤ 3|V (G)| − |E(G)|. This
formula holds with equality if every face is a triangle, giving us (i). The proof of (ii) is similar
to that of (i) except that every face has size ≥ 4 so we get 2|E(G)| ≥ 4|F (G)| by Observation
7.2 and then Euler’s Formula yields 4 = 2|V (G)| − 2|E(G)| + 2|F (G)| ≤ 2|V (G)| − |E(G)|.
¤

Observation 7.6 The graphs K5 and K3,3 are not planar.

Proof: Since K5 has 5 vertices and 10 edges, it cannot be planar by (i) of the previous lemma.
Since K3,3 has 6 vertices and 9 edges, it cannot be planar by (ii) of this lemma. ¤

Theorem 7.7 Let G be a connected d-regular plane graph and assume that G∗ is k-regular.
If d, k ≥ 3, then G is one of Tetrahedron, Cube, Octahedron, Dodecahedron, Icosahedron.

Proof: Let v = |V (G)|, e = |E(G)|, and f = |F (G)|. It follows from Observation 7.2 that
2e = kf and from the degree sum formula that 2e = dv. Substituting into Euler’s formula
we get
2e 2e
2=v−e+f = d
−e+ k

so by elementary manipulations we have:

1 1 1 1
d
+ k
= e
+ 2

1 1 1
If k, d ≥ 4 then d
+ k
≤ 2
so the above equation cannot be satisfied. Similarly, if one
1 1 1
of d or k is equal to 3 and the other is ≥ 6, then d
+ k
≤ 2
so this equation cannot be
satisfied. Thus, the only possible values for (d, k) are (3, 3), (3, 4), (4, 3), (3, 5), and (5, 3).
Further, our equation implies that in these cases, G must (respectively) have 6, 12, 12, 30,
and 30 edges. It then follows from an easy case analysis that Tetrahedron, Octahedron,
Cube, Dodecahedron, and Icosahedron are the only possibilities. ¤
 5
The Structure of Planar Graphs
Minor: Let G be a graph. Any graph which can be formed from G by a sequence of vertex
and edge deletions and edge contractions is called a minor of G. Note that if H is a graph
of maximum degree 3, then G has an H minor if and only if G contains a subdivision of H.

Series-Parallel: A graph G is series-parallel if G can be constructed from the null graph

by applying the following operations (repeatedly):

• adding a vertex of degree ≤ 1.

• adding a loop or parallel edge.

• subdividing an edge.

Theorem 7.8 A graph is series-parallel if and only if it has no K4 minor.

Proof: The ”only if” direction follows by an easy induction argument, since none of the
above operations can introduce a K4 minor. We prove the ”if” direction by induction on
|V (G)| + |E(G)|. As a base, note that this is trivial when G is null. For the inductive step,
let G be a non null graph without a K4 minor. It follows from Theorem 6.8 that G must
have either a parallel edge or a vertex of degree ≤ 2. If G has a parallel edge or a vertex
of degree ≤ 1, then by deleting this element and applying induction we deduce that G is
series parallel. If v ∈ V (G) has degree two, then the result follows by applying the reverse
operation of subdivision to v (i.e. delete v and then add a new edge between its neighbors)
and then applying induction. ¤

Lemma 7.9 If G is a plane graph, then G is 2-connected if and only if every face of G is
bounded by a cycle.

Proof: We prove the ”only if” direction by way of the contrapositive. Assume that there is
a face a which is not bounded by a cycle. Choose a vertex v so that the boundary walk of
a passes through the vertex v twice. Then we may draw a closed curve starting and ending
at v with interior contained in a. This curve separates the plane into two components each
of which must contain a vertex of G, so we find that v is a cut vertex. Thus, G is not
2-connected.
 6
For the ”if” direction we also prove the contrapositive. Let G be a plane graph which is
not 2-connected. Choose a proper 1-separation (H1 , H2 ) of G with V (H1 ) ∩ V (H2 ) = {v}.
Now, H1 and H2 are plane graphs meeting only at the vertex v with H2 embedded in a face
of H1 . It follows that there is a face a of G with boundary walk passing through v twice, so
a is not bounded by a cycle. ¤

Lemma 7.10 Let C be a cycle and let X, Y ⊆ V (C). Then one of the following holds:

(i) |X| ≤ 1 or |Y | ≤ 1.

(ii) X =Y.

(iii) There exist x1 , x2 ∈ X and y1 , y2 ∈ Y so that x1 , y1 , x2 , y2 are distinct and occur on C

in this order.

(iv) There are vertices u, v ∈ V (C) so that if P, Q are the two paths of C between u and v,
then X ⊆ V (P ) and Y ⊆ V (Q).

Proof: We shall assume that |X|, |Y | ≥ 2 and X 6= Y as otherwise one of (i) or (ii) holds. By
possibly switching X and Y , we may assume that X \ Y 6= ∅ and choose x1 ∈ X \ Y . Let y1
be the first vertex in Y clockwise from x1 and let y2 be the first vertex in Y counterclockwise
from x1 . Since |Y | ≥ 2 we have that y1 6= y2 . Let P, Q be the two paths of C between y1 and
y2 and assume that x1 lies on P . If X ⊆ V (P ), then (iv) holds, otherwise (iii) holds. ¤

Theorem 7.11 (Kuratowski-Wagner) A graph is planar if and only if it has no K3,3 or

K5 minor.

Proof: We first prove the ”only if” direction. If we take a plane graph and either delete
an edge or vertex, or contract an edge, the resulting graph may still be embedded in the
plane. Thus, every minor of a planar graph is planar. Since K5 and K3,3 are not planar
(Observation 7.6), it follows that no graph with a K5 or K3,3 minor is planar.
For the ”if” direction, we let G be a graph with no K5 or K3,3 minor and we proceed
by induction on |V (G)| + |E(G)|. If G is not connected, then by applying the induction
hypothesis to each component, we obtain a plane embedding of each component, and by
combining these, we get a plane embedding of G. Thus, we may assume G is connected. We
will proceed with a sequence of similar (but more complicated) steps.
 7
Suppose that G is not 2-connected.
In this case, we may choose a nontrivial 1-separation (H1 , H2 ) of G (unless |V (G)| ≤ 2 in
which case the theorem is trivial). Since H1 and H2 have no K5 or K3,3 minor, by induction,
we may embed them in the plane. Combining these embeddings gives a plane embedding of
G. Thus, we may assume G is 2-connected.

Suppose that G is not 3-connected.

In this case, choose a nontrivial 2-separation (H1 , H2 ) of G with {u, v} = V (H1 ) ∩ V (H2 )
(unless |V (G)| ≤ 3 in which case the result is trivial). Add a new edge uv to H1 and H2 to
form H1+ and H2+ . Choose vertices z1 ∈ V (H1 ) \ V (H2 ) and z2 ∈ V (H2 ) \ V (H1 ) and apply
Menger’s theorem to choose two internally disjoint paths from z1 to z2 . It follows from the
existence of these paths that H1+ and H2+ are minors of G (for instance, to see that H1+ is
a minor, delete all vertices and edges of H2 not on the two paths chosen above and then
contract all but one of the edges in H2 which remain). It follows from this that H1+ and
H2+ have no K5 or K3,3 minor, so by induction, we may choose planar embeddings of them.
By combining these embeddings on the edge uv and then removing it, we obtain a plane
embedding of G. Thus, we may assume that G is 3-connected.

Suppose that G has an edge uv so that G − {u, v} is not 2-connected.

If such an edge exists, then we may choose a nontrivial 3-separation (H1 , H2 ) of G − uv
where V (H1 ) ∩ V (H2 ) = {u, v, w}. For i = 1, 2 let Hi+ be the graph obtained from Hi by
adding a new vertex adjacent to u, v, and w. Then choose vertices z1 ∈ V (H1 ) \ V (H2 ) and
z2 ∈ V (H2 ) \ V (H1 ) and apply Menger’s theorem to choose three internally disjoint paths
from z1 to z2 . It follows from the existence of these paths that H1+ and H2+ are minors of
G (for instance, to see that H1+ is a minor, delete all vertices and edges of H2 not on the
three paths chosen, and then contract all edges in H2 except for one on each of the paths).
It follows that H1+ and H2+ have no K5 or K3,3 minor, so by induction, we may choose plane
embeddings of them. By combining these embeddings, we may obtain a plane embedding of
G. Thus, we may assume that G has no such edge uv.

We now have sufficient connectivity for our inductive procedure. Choose an edge xy of G,
let G0 = G·xy, let z be the vertex formed by contracting xy, and let G00 = G−{x, y} = G0 −z.
Now, G0 has no K5 or K3,3 minor, so by induction, we may choose a planar embedding of
 8
it. Furthermore, it follows from our assumptions that G00 = G − {x, y} is 2-connected, so
(by Lemma 7.9) the face of G00 which contains the vertex z is bounded by a cycle C. Thus,
all neighbors of x and y in G lie on the cycle C. We now start with our embedding of
G00 = G − {x, y} and try to extend this to a embedding of G. Let X be the set of neighbors
of x in G and Y be the set of neighbors of y (so X and Y are subsets of V (C)). We now
apply Lemma 7.10 to C for X and Y . If either (i) holds or (ii) holds and |X| = |Y | = 2,
then we may extend our embedding of G00 to a plane embedding of G. If (ii) holds with
|X| = |Y | ≥ 3, then G contains a K5 minor, contradicting our assumption. If (iii) holds,
then G contains a K3,3 minor, contradicting our assumption. The only remaining possibility
is (iv), but in this case we may once again extend our embedding of G00 to an embedding of
G. ¤