Planar graphs

Dr.Gangaboraiah, PhD Department of Community Medicine Kempegowda Institute of Medical Sciences Banashankari 2nd Stage, Bangalore-70 Mobile: 98451 28875 E-mail: gbphd@kimsbangalore.edu.in

Introduction Earlier it was indicated that a graph can be represented by more than one figure (geometrical drawing). In some figures which are representing graphs the edges intersect (cross over) at points, which are not vertices of the graph and in some others the edges meet only at the vertices.

A graph which can be represented by at least one plane drawing, in which the edges meet only at the vertices is called a planar graph. A graph which cannot be represented by a plane drawing, in which edges meet only at the vertices is called a nonplanar graph.

Thus, in every plane drawing of a nonplanar graph at least two edges of the graph intersect at a point which is not a vertex of the graph.

Planar graph
A graph G is said to be a planar graph if it can be represented by a figure drawn on a plane (of the paper, black board etc.,) in such a way that no two of its edges intersect (cross over except at a vertex on which both are incident).

Nonplanar graph
A graph G is said to be a nonplanar graph that cannot be represented by a figure drawn on a plane without a cross over between its edges.

Embedding
A figure representing a graph drawn on any surface in such a way that no two edges intersect is called embedding. An embedding of a planar graph is called plane representation.

For example, consider the following three drawings representing the complete graph of four vertices, viz., K4.

Observe that in the first of the drawings, two edges intersect and in the second and third no two edges intersect.

10

Therefore, the first drawing is not an embedding of K4 while the other two are embeddings.
Further, the second and third drawings illustrate that K4 can be drawn on a plane in such a way that no two of its edges intersect. Therefore, K4 is a planar graph.

11

The second and third drawings are its plane representations.

Two standard examples of nonplanar graphs are the complete graph K5 and the complete bipartite graph K3, 3.

12

Theorem: The complete graph of five vertices K5 (Kuratowskis first graph) is nonplanar.

Proof: Recall that in the graph K5 there are five vertices and every vertex is joined to all other four vertices by an edge so that the graph contains ten edges.

13

Let us name the vertices as v1, v2, v3, v4, and v5 and the edges as e1, e2, , e10 as specified below.

14

e1=(v1, v2)

e6=(v2, v4)

e2=(v1, v3)
e3=(v1, v4) e4=(v1, v5)

e7=(v2, v5)
e8=(v3, v4) e9=(v3, v5)

e5=(v2, v3)

e10=(v4, v5)

Note that the edges e1, e5, e8, e10 and e4 form a pentagonal circuit.

15

Consider the edge e7=(v2,v5). This edge can be drawn either inside or outside the pentagon shown in figure 2 (slide no. 8). Consider the edge e2=(v1, v3) and e3=(v1, v4). If we draw these edges inside the pentagon, they will intersect e7; therefore, let us draw them outside. Therefore, we have to draw inside the pentagon.

16

If we draw the edge e9 outside the pentagon, it intersects the edge e3. Therefore, we have to draw it inside the pentagon. Thus, both of the edges e6 and e9 are to be drawn inside the pentagon, but then they themselves intersect.

17

This demonstrates that in every possible plane drawing of K5 at least two edges of K5 intersect at a point which is not a vertex of K5. Hence, K5 is nonplanar.

18

Theorem 2 Kuratowskis second graph is nonplanar. Proof: Be definition, K3, 3 is the graph in which the vertex set is partitioned into two sets V1 and V2 each containing three vertices such that every vertex in

19

V1 is joined to every vertex in V2 by an edge and vice-versa. Let us name the vertices in V1 as v1, v2, v3 and the vertices in V2 as v4, v5, v6.

20

e1=(v1, v4) e4=(v2, v4) e7=(v3, v4)

e2=(v1, v5) e5=(v2, v5) e8=(v3, v5)

e3=(v1, v6) e6=(v2, v6) e9=(v3, v6)

Observe that the edges e1=(v1, v4), e4=(v4, v2), e5=(v2, v5), e8=(v5, v3), e9=(v3, v6), e3=(v6, v1) form a hexagonal circuit.

21

22

Consider the edge e6=(v2, v6). This edge can be drawn either wholly inside the hexagon or wholly outside it. Let us draw it inside- the other case is similar. Consider the edge e2=(v1, v5). If we draw this edge inside the hexagon, it intersects the edge e6.

23

Therefore, let us draw it outside the hexagon.

Now, consider the edge e7=(v3, v4). If this edge is drawn inside the hexagon, it crosses the edge e6 and if it is drawn outside the hexagon, it crosses the edge e2.

24

This demonstrates that in every possible plane drawing of K3, 3 at least two edges of K3, 3 intersect at a point which is not a vertex of K3, 3.

Hence K3, 3 is a nonplanar graph.

25

Common properties of K5 and K3, 3 By the virtue of the definitions and Theorems 1 and 2 the Kuratowskis graphs K5 and K3, 3 share some common properties. 1. Both are regular graphs, where as K5 is 5-regular, K3, 3 is 3-regular.

26

2. Both are nonplanar. 3. Removal of one edge makes each a planar graph . (In theorem 1, the drawing would have contained only non intersecting edges if e9 was not there. In theorem 2, the drawing would have contained only nonintersecting edges if e7 was not there.

27

4. Removal of one vertex makes each a planar graph (Note that K4 and K2, 3 are planar.

5. K5 is the nonplanar graph with the smallest number of vertices, and K3,3 is the nonplanar graph with the smallest number of edges.

28

Homeomorphic graphs Two graphs G1 and G2 are said to be homeomorphic if one of these can be obtained from the other by insertion of new vertices of degree two into its edges or by the merger of edges in series.

29

Observe that the second graph is obtained from the first graph by inserting the vertex v which of degree 2 into an edge. Hence, the two graphs are homeomorphic.

The following theorem known as the Kuratowskis theorem is fundamental in the study of planar graphs.

30

31

32

Theorem 3

A necessary and sufficient condition for a graph G to be a planar is that G does not contain either K5 or K3, 3 as subgraphs or any subgraph homeomorphic to either of these.
Note: If a graph G contains K5 or K3, 3 as a subgraph, then in view of Kuratowskis theorem we may infer that G is nonplanar.

33

Example: The Petersen graph may be represented pictorially as:

Since the Petersen graph is edge contractible to K5, it is not a planar graph

34

Example 1
Show that (i) The graph of order 5 and size 8 (ii) The graph of order 6 and size 12 are planar

35

Solution To show that a graph is planar, it is enough if we draw one plane diagram representing the graph in which no two edges cross each other. The figures in slide 34 shows that they are planar.

36

Example 2: If n4, show that the complete graph Kn is planar. Solution: The diagrams represent graphs K1, K2, K3, and K4.

37

Solution: The diagrams represent graphs K1, K2, K3, and K4. In none of these diagrams, no two edges cross each other. Therefore, K1, K2, K3, and K4 are all planar graphs. Since K1, K2, K3, and K4 are planar graphs and K5 is a nonplanar graph, it may be infer that K5 is the complete nonplanar graph of the smallest order.

38

Example 3 Show that the graphs K2, 2 and K2, 3 are planar graphs. Solution In K2, 2 the vertices set is partitioned into two sets of vertices say {v1, v2} and {v3, v4}.

39

Solution In K2, 2 the vertices set is partitioned into two sets of vertices say {v1, v2} and {v3, v4}.

40

respectively and there is an edge joining every vertex in V1 with vertex V2 and vice-versa. The first figure in slide no. 36 represents this graph. Evidently, in this graph, no two edges cross each other. Therefore, K2, 2 is planar.

41

In K2,3, the vertex set is partitioned into two sets V1, and V2, with V2, containing two vertices v1, v2, and V2 containing three vertices, say, v3, v4, v5 and there is an edge joining every vertex in V with every vertex in V2 and vice-versa. The figure 2 in slide no. 39 represents this graph. Observe that in this figure, no two edges cross each other. Therefore, K2,3 is planar.

42

Note: The graph K2, 3 can be obtained by removing one vertex from K3, 3. Thus, the nonplanar graph K3, 3 becomes planar if one vertex is removed from it.

43

Example 4 (Three utility problem) Suppose there are three houses and three utility points (electricity, water, sewage, say) which are such that each utility joint is joined to each house. Can the lines of joining be such that no two lines cross each other?

44

Since each house is joined to each utility point, the graph has to be K3, 3. This graph is nonplanar and therefore at least two of its edges cross each other. As such, it is not possible to have the lines joining the houses and the utility points such that no two line cross each other. Example 5 Verify that the following graphs are homeomorphic but not isomorphic.

45

Solution Each graph can be obtained from the other by adding or removing appropriate vertices. Therefore, they are homeomorphic. They are not isomorphic is evident if we observe the incident relationship which is not identical.

46

Eulers formula If G is a planar graph, then G can be represented by a diagram in a plane in which no two edges intersect (cross over). Such a diagram divides the plane into a number of parts called regions (or meshes, or windows or faces) of which exactly one (part) is unbounded. The number of edges on the boundary of a region (face) is called the degree of that region*.

47

*If a pendant edge is a part of the boundary of a region, by convention this edge is counted twice while determining the degree of the region. For example, in the diagram of a planar graph shown in the above figure, the diagram divides the plane into 6 regions R1, R2, R3, R4, R5, and R6.

48

Observe that whereas each of the regions R1 to R5 is bounded, the region R6 is unbounded. That is, R1 to R5 are in the interior of the graph while R6 is in the exterior.

49

Further observe that, in the above figure, the boundary of the region R1 is made up of two edges. Therefore, the degree of R1 is 2 and we write d(R1)=2.

The boundary of each of the regions R2 and R4 is made up of 3 edges, therefore, d(R2)=d(R4)=3.

50

The boundary of the regions R3 consists of 4 edges of which one is a pendant edge. Therefore, d(R3)=5.
The region R5 is bounded by a single edge (self-loop), therefore, d(R5)=1. The boundary of the exterior region R6 consists of six edges; therefore, d(R6)=6.

51

Note that,
d(R1)+d(R2)+d(R3)+d(R4)+d(R5)+d(R6)=20

which is twice the number of edges in the graph. This property is analogous to the handshaking property and is true for all planar graphs.

52

In the figure, two trivial interesting cases are depicted. The first of these figures corresponds to the null graph, N1, with one vertex and no edges.

53

Evidently, the diagram determines only one region R- the unbounded region. The second figure corresponds to a tree. In this case also, the diagram determines only one region R-the unbounded region.
It should be pointed out that the regions are determined by a diagram of a planar graph and not by the graph itself.

54

This means that if we change the diagram of the graph, the regions determined by the old one in the sense that the unbounded region in the old. However, the interesting fact is that the total number of regions in the two diagrams remains the same.

Theorem 1: A connected planar graph G with n vertices and m edges has exactly m-n+2 regions in every one of its diagrams.

55

Proof Let r denote the number of regions in a diagram of G. The theorem states that r = m n + 2 or n m + r = 2 (1) The proof is by induction. If m = 0, then n must be equal to 1. Because, if n>1, then G will have at least two vertices and there must be an edge connecting them (because G is connected) so that m 0.

56

If n = 1, a diagram of G will have only one region in it- the entire plane region (as shown in figure one of slide no. 39). Thus, if m = 0, we have n = 1 and r = 1 so that n m + r = 2. This verifies the theorem for m = 0.

Now, assume that the theorem holds for all graphs with m = k number of edges, where k is a specified non-negative integer.

57

Consider a graph Gk+1 with k+1 edges and n vertices. If Gk+1 is a tree, then Gk+1 will have n-1 edges and a diagram of Gk+1 will determine only one region-the entire plane region (as shown in figure 2 in slide no. 39). Thus, for Gk+1, we have k + 1 = n 1 and r = 1, so that n (k+1) + r = 2. This means that the result (1) is true when m = k + 1 as well if the graph is a tree.

58

Suppose Gk+1 is not tree and let r be the number of regions which a diagram of Gk+1 has. Since Gk+1 is not a tree, it contains at least one circuit. Consider an edge e in a circuit and remove it from Gk+1. The resulting graph, Gk+1- e, will have n vertices and k edges and a diagram of the graph will have r-1 regions. Since Gk+1- e has k edges, the theorem holds for this graph (by induction assumption made).

59

That is, we have r-1 = k n + 2, or n- (k + 1 ) + r = 2. This means that the result (1) is true when m = k + 1 as well it is true for m = k and the graph is not a tree. Thus, the result (1) if it is true for m = k + 1, it is true for m = k 0, whether the graph is a tree or not. Hence, by induction, it follows that the result (1) is true for all non-negative integers m.

60

Remarks The formula (1) which gives the number of regions (which a diagram of a connected planar graph has) in terms of the number of vertices and the number of edges is known as the Eulers formula. Corollary 1 If G is a connected simple planar graph with n ( 3) vertices, m (>2) edges

61

and r regions, then (i) m (3/2) r and (ii) m 3n-6 Further, if G is a triangle-free, then (iii) m 2n-4. Proof Since the graph G is simple, it has no parallel edges and no self-loops. As such, every region must be bounded

62

by three or more edges. Therefore, the total number of edges that bound all regions is greater than or equal to 3r. On the other hand, an edge is in the boundary of at most two regions. Therefore, the total number of edges that bound all regions is less than or equal to 2m. Thus, 3r 2m or m (3/2) r. This is required result (i).

63

Now, substituting for r from the Eulers formula in the result just proved, we get m (3/2) (m n + 2) Which simplifies to m 3n 6. This is the required result (ii). If G is triangle-free, every region must be bounded by four or more edges, so that the total number of edges that bound all is greater than or equal to 4r. Consequently, we have 4r 2m.

64

Substituting for r from the Eulers formula in this, we get m 2 (m n + 2) or m 2n-4. This is the required result (iii). Corollary 2 Kuratowskis first graph, K5, is nonplanar. Proof The graph K5 is simple, connected and has n=5 vertices and m=10 edges.

65

If this graph is planar, then from the result (ii) of Corollary 1, we should have m 3n-6, i.e., 10 15-6, which is not true. Therefore, K5 is nonplanar. Corollary 3 Kuratowskis second graph K3,3 is nonplanar Proof: Note that K3,3 is simple, connected and has n=6 vertices and m=9 edges.

66

Suppose K3, 3 is planar. By examining the K3, 3 figure we note that K3, 3 has no triangles. (In fact no bipartite graph has a circuit of odd length). Therefore, by result (iii) of Corollary 1 we should have m 2n- 4, i.e., 9 12- 4, which is not true. Hence K3, 3 is nonplanar.

67

Corollary 4 Every connected simple planar graph G contains a vertex of degree at most 5. Proof Suppose each vertex of G is of degree greater than or equal to 6. Then, if d1, d2, , dn are the degrees of the n vertices of G, we have d1 6, d2 6, d3 6,, dn 6.

68

Adding these, we get

d1+d2++dn 6n. By handshaking property, the left hand side of this inequality is equal to 2m where m is the number of edges in G. Thus, 2m 6n or 3n m.

69

By result (ii) of Corollary 1, we should have m 3n 6. Thus, 3n 3n 6. This cannot be true. Therefore, G must have a vertex of degree at most 5.

70

Example Verify Eulers formula for the graph shown in slide no. 45

Solution: Here n=6, m=10 and r=6 Therefore, n m + r = 6 10 + 6 = 2 Here the Eulers formula is verified.

71

Example What is the maximum number of edges possible in a simple connected planar graph with eight vertices? Solution: m 3n-6 = 18 ( Here n=8). Thus the maximum number of edges possible is 18.

72

Example What is the minimum number of vertices necessary for a simple connected graph with 11 edges to be a planar? Solution For a simple connected planar (n, m) graph, we should have by Corollary 1 m 3n 6 or n (m +6)/3 When m = 11, we get n 17/3. Thus, the required minimum number of vertices is 6.

73

Example Show that the condition m 3n 6 is not a sufficient condition for a connected simple graph with n vertices and m edges is not a planar.

Solution: Consider the graph K3,3 which is simple and connected and which has n=6 vertices and m=9 edges. We check that for this graph, m 3n-6. But the graph is nonplanar. This counter example proves the required result.

74

Dual graph: A dual graph of a given planar graph G is a graph which has a vertex for each plane region of G, and an edge for each edge in G joining two neighboring regions, for a certain embedding of G. The term "dual" is used because this property is symmetric, meaning that if H is a dual of G, then G is a dual of H (if G is connected).

75

76

Properties
The

dual of a planar graph is a planar multigraph - a graph that may have loops and multiple edges. If G is a connected graph and if G is a dual of G then G is a dual of G.

Example

77

78

Example: Prove that the Petersen graph is a non-planar graph.

79

Solution: Recall that the Petersen graph is a 3-regular graph of order 10 and size 15. The graph is repeated in figure 2 with vertices as A, B, C etc. Now consider the graph shown in figure 2. Verify that this graph is another representation of the Petersen graph.

80

Suppose we delete the edges {C, D} & {Q, T} along with their end vertices from this graph. The resulting graph is K3, 3. Thus the Petersen graph contains K3, 3 as a subgraph. Therefore, the Petersen graph is nonplanar (by Kurotowskis theorem).

81

Detection of planarity:
Given a graph G, the determination of its planarity or otherwise is an important problem. This problem can be tackled by employing what is known as elementary reduction. The steps involved in this reduction are as follows:

82

Step 1: Given a graph G, determine the set A = {G1, G2, G3, , Gk} Where G1, G2, G3, , Gk are subgraphs of G, every pair of which has exactly one vertex in common (such graphs are called blocks)

83

Step 2: Remove all loops from all of Gis. Step 3: Remove all but one edge between every pair of vertices joined by multiple edges (if any). Step 4: Eliminate all vertices of degree 2 by merging the edges incident on these vertices.

84

Step 5: Repeat step 3 and 4 repeatedly until each block Gi is reduced to a new graph Hi, which will be one of the following: (1) A graph with a single edge. (2) A complete graph of order four.

85

If Hi is the first or the second of the above possible forms, we conclude that Hi is a planar graph.
Consequently, each Gi with which we started is planar and therefore G is planar.

86

Example: Carry out the elementary reduction process for the following graph

87

Step 1: The given graph is a single block, hence, contains only G Step 2: Remove the loops e9 Step 3: Remove one of the two parallel edges e1,e8, say e8.

88

The resulting graph left out after the first three steps is

89

Step 4: Eliminate the vertices of degree 2 by merging the edges incident on these vertices. Thus, we merge (i) The edges e1 and e2 into an edge, say e10 (ii) The edges e6 and e7 into an edge, say e11

90

The resulting graph is

91

As per step 3, let us remove one of the parallel edges e5 and e10 and one of the parallel edges e3 and e11. The resulting graph is

92

As per step 4, let us merge the edges e3 and e4 into an edge e12 to get the following graph

93

As per step 3, let us remove one of two parallel edges e5 and e12, say the edge e12 and thus, we get the graph

94

This graph is the final graph obtained by the process of elementary reduction applied to the graph given in slide no. 86.

Example: Check the planarity (or otherwise) of the following graph

95

Step 1: Splitting G into blocks

96

Step 2: Removing loops and eliminating multiple edges

97

Step 3: Merging the edges incident on vertices of degree 2

98

Step 4: Eliminating the parallel edges

99

100

The reduction is now complete. The final reduced graph in slide no.88 has three block of which the first and the third (which are single edges) are obviously planar. The second one is evidently the complete graph K5, which is non-planar. Thus, the given graph contains K5 as a subgraph and is therefore non-planar.