Higher Engineering Mathematics

Prof. P.N. Agrawal

Department of Mathematics
Indian Institute of Technology Roorkee
Lecture-35
Kuratowski’s Theorem

Hello friends, welcome to my lecture on Kuratowski’s Theorem. We have seen that a graph or a
multi-graph is planar if it can be drawn in a plane so that its edges do not cross. Kuratowski gave
a Theorem which characterises non-planar graphs.

(Refer Slide Time: 0:48)

A graph is not planar if and only if it contains a subgraph homeomorphic to K 33 or K 5. We have

also seen that K 33 and K 5 are nonplanar graphs. Now, so in a way this theorem otherwise can
also be interpreted as a graph is planar if and only if it has no subgraph which is homeomorphic
to K 33 or K 5. So, he, Kuratowski gave a characterisation of planar graphs.
(Refer Slide Time: 1:18)

Now, let us see some examples on this Kuratowski’s Theorem. Let us look at this example, here
we have a graph, this one, this graph we have. We have 6 vertices here, A, B, C, D, E, F okay
and they are joined by these edges. Now this graph has a subgraph which is isomorphic to K 33.
Let us see that that means we want to show that this graph is nonplanar graph, okay. If we can
show that this graph has a subgroup which is homeomorphic to K 33 or K 5 then it will be a
nonplanar graph by the Theorem of Kuratowski. Okay. So, let us see how we can do that.
We remove the edge CF Okay, let us remove the edge CF. okay. Let us remove the edge CF
from this graph. By removing vertices or edges from a graph we can arrive at a sub graph, we
have earlier seen that. So when we remove the edge CF, okay, and arrange E and F as F and E,
okay redraw the graph then what do we notice? The graph becomes in this form okay. So let us
remove the edge CF This edge, this edge we remove and transpose E and F. okay. Transpose E
and F. Okay. So, E vertex now comes here and F vertex now is here. Okay.

And then we see that A which is joined to E okay, it is joined to E like this, okay so this is AE
and then A is joined to B, this is joined to B here. Then A is joined to D. This is A joined to D
and C is joined to B, C is joined to B here. C is joined to E, okay. This is CE. So we have CE
here okay and then C is joined to D. And then EF okay F is joined to D. So this is F is joined to
D here. And then F is joined to E, so this is F is joined to E and then F is joined to B, okay this
one F is joined to B okay. So F is joined to B.

So D, 6 vertices, A, B, C, D have been partitioned into 2 sets ACF and BDE okay. And each
vertex of one set, ACF is joined to every other vertex of the other set BDE okay. So this graph,
this graph is K 33 okay. This graph is isomorphic to, this graph is isomorphic to K 33. Okay. So the
given graph, okay this given graph has a subgroup, okay the given graph has a subgraph which is
isomorphic to K 33. Okay. And therefore the given graph has a subgraph which is homeomorphic
to K 33 because by the Theorem, by Kuratowski Theorem, it follows that since this graph is
isomorphic to K 33 okay, we have K 33 is a nonplanar graph okay K 33 is a nonplanar graph.

And so it contains a subgraph which is homeomorphic to K 33 or K 5. Okay so it is this graph is

nonplanar K 33 is nonplanar. So we have a subgraph which is homeomorphic to K 33 and therefore,
this graph okay this graph had a subgraph which is isomorphic to K 33 and K 33 has a subgraph
which is homeomorphic to K 33. So this graph has a subgraph which is homeomorphic to K 33.

And therefore this graph is nonplanar, the given graph. We can remove the edge CF and
transpose the vertices E and F and arrive at this graph okay which is isomorphic to K 33 okay.
And then use Kuratowski’s Theorem to obtain a subgraph of K 33 which is homeomorphic to K 33
and therefore the given graph has a subgraph which is homeomorphic to K 33.
(Refer Slide Time: 6:02)

Now let us look at this one, 2 nd one. The graph shown in this figure has a subgraph okay
homeomorphic to K 5. Identify the subgraph. So what we do here? Let us remove the vertex H,
the edges FH and HG. Okay. So let us remove the vertex H okay. When we remove the vertex H,
the edges FH and GH will also be removed okay. Then edges FH and GH are also removed.
Okay. And then remove this edge CA okay. Remove edge CA, also remove edge AC, okay this
AC. Then transpose A and D, okay transpose A and D and C and E okay.

When we transpose A and D, are transposed A and D okay, you see D goes to the position of A
and A goes to the position of D. You can see here. Okay. So transpose A and D and C and E.
okay. We arrive at this subgraph okay. This is a sub graph okay because it has been obtained by
deleting, removing the vertex H. Thereby removing the edges FH and GH and also the edge, AC.
Okay. So this is a subgraph of this given graph, of (b) is a subgraph of (a). okay. Moreover, it is
homeomorphic to K 5 because in K 5, we have this graph.

K 5 is this graph. Okay. Just K 5. So we are adding two vertices, A here and C here. Okay. When
we add the vertices A and C on the edge BF and BG okay the graph that we get, that graph is
homeomorphic to this graph. So this, so in K 5 okay we add the vertices A and C to arrive at this
graph okay and we know that K 5 is a nonplanar graph, okay. So this graph is a subgroup of this
graph. This graph is a subgraph of this graph. Okay. And moreover, this is homeomorphic to K 5.
And therefore, the given graph A has a subgraph which is homeomorphic to K 5 and so the given
graph is nonplanar. The graph in (a) is, this graph is nonplanar graph.

(Refer Slide Time: 9:41)

Okay. The graph shown in the figure below has a graph homeomorphic to K 33. Let us identify
that. Okay. So what we do in this? We delete, we remove the edges BD. We remove edges BD,
DF, okay and CE and EG okay. And then redraw this graph okay. okay. After removing the
edges BD, this one, BD and DF, CE and EG okay. Let us redraw this graph. When we redraw
this graph, it becomes like this. You see BC can be drawn as BC and then the vertex A here
okay. Then this FHG can be drawn like this, this FH and then HG. Over this edge FG with put
this vertex H, so FHG, we can draw like this and then once we remove BD, DF and CE and EG
okay, the graph takes this form.

Now, this graph is homeomorphic to K 33 okay, so because why it is homeomorphic to K 33?

Because in the K 33 graph this A vertex and H vertex are not there okay. The remaining graph is
isomorphic to K 33. So this graph has been obtained from the graph which is isomorphic to K 33 by
adding vertices A and H and therefore this is homeomorphic to K 33. This graph is homeomorphic
to K 33. So the given graph has a subgraph which is homeomorphic to K 33 and therefore this
graph is nonplanar. Okay.
(Refer Slide Time: 11:52)

Now, let us look at this graph. The graph shown in the figure below has a subgraph isomorphic
to K 5. Okay. So what we do here, let us remove the vertex A and then the edges AB, AC, and
AD will be removed okay. So let us remove the vertex A okay. Then the edges AB, AD and AC
will also be removed. Okay. And then let us redraw the graph. We bring this vertex D above this
edge BC okay. So we bring the vertex D above the edge BC okay this edge above this edge BC
and redraw the graph.

We see that we arrive at, we redraw the graph, we get the graph (b). And we know that because
here every vertex is connected to every other vertex okay then therefore this graph is isomorphic
to K 5. This is isomorphic to K 5. But K 5 is a nonplanar graph. So the given graph is also
nonplanar. This is also nonplanar.
(Refer Slide Time: 14:17)

Now, let us use Kuratowski’s Theorem to show that the graph below given below is nonplanar.
Okay, what we do is let us remove the edges joining v1 to v6. Okay. Let us remove the edges
joining v1 to v 6 and v2 to v5. Okay. So we remove this edge okay and also the edge joining v2 to
v 5 okay and we arrive at this subgraph. Okay. We have removed v1 the edge joining v1 to v 6 okay
this edge and we have also removed the edge joining v2 to v 5 okay. So we have v1 to v3, v 3 to v4 ,
we have v1 to v 6. v1 to v 6 we have removed. Okay. And we have also removed v2 to v5.

Now, what we do, we merge the two edges, these two edges. There are 2 edges at v 6 okay. Let us
merge the 2 edges. Series edges in series at v 6 okay. We have this, these 2 edges, this edge and
this edge, they are in series at v6 . So let us merge these 2 edges which are in series at v6. We
arrive at this graph. So v 5, we have v1, v2 here, v 4 , v 3 , v1 to v 3 and then and then v3 to v4 okay,
v7. Okay. And we have this. So we have this graph okay. When we merge these 2 edges, this one
and this one which are in series at v6 okay we arrive at this graph okay. v1, v2 , v 4 , v5 , v7, v1, v3
and yes, this one. Okay. And so, what we notice is that now this can be redrawn, let us redraw
this.

We have the following. We have v1, v 4 , v7, v2 , v3 , v 5. Okay. So we have v1 v2, and then we have
v1 v3, and then we have v1 v5. Okay. Like this. Then v4 v 2, v4 v 3,v4 v 5 . And this v2 v7. Okay v7 v3
and v7 v5. You can say here, we have v7. v7 is joined to v 5 okay. v7is joined to v3. v 3 we have
okay. And then v7is joined to v2 that is there. So v2 should have been there. v7 is there, yes v7is
joined to v2. So this one here. Okay. That we left out. So v7 is joined to v2. v7 is joined to v 5. v7 is
joined to v3.

Then v4 , v 4 is joined to v3 okay. v4 is joined to v5, v 4 is joined to v2 and then v1. v1 is joined to v5
okay, v1 is joined to v3, v1 is joined to v2 . So this graph can be redrawn and we see that this graph
is isomorphic to K 33 . This graph is isomorphic to K 33 and K 33 we have seen is A nonplanar
graph and therefore, the given graph is also nonplanar because it has a subgraph okay with this
subgraph which is isomorphic to K 33 and K 33 is nonplanar. So this graph is nonplanar.

(Refer Slide Time: 20:43)

Now, let us consider travelling salesman problem. A salesman is required to visit a number of
cities. What is the route he should take if he has to start from his home city, visit each city
exactly once and then return home travelling the minimum distance? Okay. So suppose we
represent cities by the vertices in a graph and roads by edges connecting the cities. The length of
the road may be represented as a weight associated with the corresponding edge. Okay. Now in
graph theoretic terminology the travelling salesman problem is equivalent to finding a short test
Hamiltonian circuit. A Hamiltonian circuit is, a Hamiltonian cycle is a which visits each vertex
exactly once.
(Refer Slide Time: 21:34)

( n−1 ) !
Now, if you take a complete weighted graph of n vertices, then there exists Hamiltonian
2
circuits. So if you take n vertices, okay we consider a complete undirected graph, complete
undirected graph where n are the number of vertices in a complete undirected graph with n

( n−1 ) !
vertices, there are, there exists different Hamiltonian cycles. If we consider a complete
2
directed graph, then there exists (n-1)! different Hamiltonian cycles because their direction is
associated. In the case of undirected, complete graph, there is no direction. So they become half

( n−1 ) !
and therefore we get Hamiltonian cycles in the case of complete undirected graph okay.
2

In the case of a complete directed graph with n vertices, there exists (n-1)! Hamiltonian cycles.
Okay. So in a complete weighted graph of n vertices, of course here we are talking about

( n−1 ) !
undirected graph, okay. There exists Hamiltonian circles. Now one possible approach to
2

( n−1 ) !
solve the travelling salesman problem is to consider calculate the weights of all the
2
Hamiltonian circles and then select the shortest one that is but in that case what will be this
problem will be very difficult to solve because the when n is large, the number of Hamiltonian
circuits will be too large.

So it is difficult to determine the shortest path to solve the travelling salesman problem and
therefore, because of this (n-1)! factorial, even for values of n which are not very large, it is
difficult to solve the consider all the Hamiltonian circuits and thereby determine the the circuit
Hamiltonian circuit along which travelling salesman problem has a solution. So it can be done
only manually by for a very small value of n okay. For arbitrary value of n, no efficient
algorithm to solve this problem is known till today.

(Refer Slide Time: 25:19)

Now the travelling salesman problem can be described as follows. This TSP means travelling
salesman problem. It can be written as (G,f,t), where G = (V, E) a complete graph. Okay V is the
set of vertices, E is the set of all edges in the complete graph, f is a function from V ×V to Z.
okay. V×V means all the pair of vertices of V. Okay. So f is a function from V×V to Z okay.
And t ∈Z okay. So G is a graph. This t∈ Z, G is a graph that contains a travelling salesman tour
with cost that does not exceed t. okay. So G is a graph which contains a travelling salesman tour
with cost not exceeding t.
(Refer Slide Time: 26:11)

Now let us consider this case, this example. We have , consider the following set of cities, find a
minimal path passing from all vertices once. So there are cities presented by A, B, C, D, E,
Okay. And the weights are given to each path, okay. To path AB, weight 2 is given, to path BC,
weight 4 is given. To path Could, the weight 3 is given. So let us find the path along which the
we get the minimum value of this minimum along which the travelling salesman problem have
the solution. So let us see, we have 2 paths okay.

Because we have to consider Hamiltonian cycle so we have to consider exact, each vertex
exactly once okay while going from along the this graph. Say suppose we start with A okay then
we must consider the Hamiltonian cycle ending at A. So let us see which path we can take. We
have the path A, B, C, D, A, B, C, D, E, and back to A. okay. This is one path. We can also
consider A, B, C, D A, B, C, E, D okay we can add another possibility could be, another
Hamiltonian cycle is A, B, C, then we go to E, then we go to D and then we go to A okay.

So if we follow this path, ABC, so what is the total of weights to this path 1 and this is path 2.
Both these paths when we follow, we visit each vertex exactly once. So along Path 1, some of
the weights is 2 plus 4 okay plus 3 plus 10 plus 5. And we get 2 +4 =6, 6 + 3 =9, 9 + 10 =19 + 5
=24 okay along path 1. And along Path 2 okay, we have the total as 2 plus 4 plus and then from
C to E we go. So 3 and then we go to D okay. From E to D. So we get 10. Okay. And then from
D, we go to A. So 1, 2, 3, no 1, 2, 3, 4 and then we go to this 5. Okay. So 2 plus 4 plus 3, then
10. I have missed the 3 here. 3 okay. AB, BC, CE okay. Then ED and then DA. So when we
follow this path, how much we get? 2 + 4 =6, 6 + 3= 9, 9 +10= 19 okay 9 + 10= 19 +12 this is 31
okay. So here we have 2 Hamiltonian cycles okay. One gives 24 value, another one gives 31
value. So 24 is the smallest okay. Therefore we should follow the path 1. Okay. That is A, B, C,
D, E and A okay.

So that is the solution of the, that is the minimal path here that gives the solution of the travelling
salesman problem in this case.

(Refer Slide Time: 30:35)

Now let us consider another one. Let us consider this problem here okay. So again we have let us
let us start with A okay. So then which path we have to take? Hamiltonian cycle. So path 1. Let
us see path 1. So path one could be A to B, B to C, C to D, D to E, E to F, F to H, H to G, okay
and G to A. So we can follow this path. Another path could be again A to B, B to C, C to D, D to
E, E to F, F to G, G to H, okay and H to A okay. Now let us see along Path 1, what total we are
getting. Along Path 1. Sum of weights is 2+1 + 3 + 2 +2 so after D we come to E, from E we go
to F, F to H. So we have 6 here. F to H, then H to G so we have 5 and then G to A, so we get 1.

So this 2 + 1 =3, 3 +3 =6 +2 =8 + 2= 10 + 6 =16 + 5 =21 +1= 22 okay. The sum is A to B, so we

have 2. Then B to C, we have 1. Then we have C to D, so 3 and then D to E, that is 2. Then E to
F, again 2 and then F to G, we get 7. And then we have G to H, so we get 5 and then H to A, so
we get 3. Okay. So this is 2 +1 =3+ 3= 6+ 2 =8 + 2= 10+7 =17 +5=22 and 3 =25 okay. So out of
these 2 Hamiltonian cycles, the one which gives the least value is the path 1 okay. So we should
follow path 1. So that is the solution of this travelling salesman problem.

(Refer Slide Time: 33:30)

So the total weights of Hamiltonian circuits described in these 2 paths are 22 and 25 okay. So in
the total weight Hamiltonian circuit in A is minimum, this one is minimum. The sales man
should travel along 2 circuit in A. So that is the end of my lecture. Thank you very much for your
attention.