0% found this document useful (0 votes)
599 views24 pagesChapter 1
Graphs are so named because they can be represented graphically. Each vertex is indicated by a point, and each edge by a line joining the points which represent its ends. A diagram of a graph merely depicts the incidence relation holding between its vertices and edges.
Uploaded by
Raj ReddyCopyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
599 views24 pagesChapter 1
Graphs are so named because they can be represented graphically. Each vertex is indicated by a point, and each edge by a line joining the points which represent its ends. A diagram of a graph merely depicts the incidence relation holding between its vertices and edges.
Uploaded by
Raj ReddyCopyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
/ 24
1 Graphs and Subgraphs
1.1 GRAPHS AND SIMPLE GRAPHS
Many real-world situations can conveniently be described by means of a
diagram consisting of a set of points together with lines joining certain pairs
of these points. For example, the points could represent people, with lines
joining pairs of friends; or the points might be communication centres, with
lines representing communication links. Notice that in such diagrams one is
mainly interested in whether or not two given points are joined by a line;
the manner in which they are joined is immaterial. A mathematical abstrac-
tion of situations of this type gives rise to the concept of a graph.
A graph G is an ordered triple (V(G),E(G), Wc) consisting of a
nonempty set V(G) of vertices, a set E(G), disjoint from V(G), of edges,
and an incidence function Yc that associates with each edge of G an
unordered pair of (not necessarily distinct) vertices of G. If e is an edge and
u and v are vertices such that Wo(e) = uv, then e is said to join u and v; the
vertices u and v are called the ends of e.
Two examples of graphs should serve to clarify the definition.
Example 1
G=(V(G), E(G), vo)
where
VG
E(G) = {e1, 2, €3, €4, €5, €6, €25 a}
and We is defined by
D1, Va, V3, V4, Us}
Wales) = v1v2, Yo(€2) = d203, Poles) = Vas, Wola) = VsUs
Wales) = V2d4, Wo(€s) = Vs0s, Poles) = V20s, Holes) = v2vs
Example 2
H =(V(H), E(H), x)
where
V(H) ={u, », w, x, y}
E(H) = {a, b,c, d,e, fg, h}
and yu is defined by
Yala) = uv, yu(b)=uu, Yulé)= vw, u(d) = wx
Wale) = vx, walf)=wx, dulg)=ux, wah) = xy 2 Graph Theory with Applications
Figure 1.1. Diagrams of graphs G and H
Graphs are so named because they can be represented graphically, and it
is this graphical representation which helps us understand many of their
properties. Each vertex is indicated by a point, and each edge by a line
joining the points which represent its ends.t Diagrams of G and H are
shown in figure 1.1. (For clarity, vertices are depicted here as small circles.)
There is no unique way of drawing a graph; the relative positions of points
representing vertices and lines representing edges have no significance.
Another diagram of G, for example, is given in figure 1.2. A diagram of a
graph merely depicts the incidence relation holding between its vertices and
edges. We shall, however, often draw a diagram of a graph and refer to it as
the graph itself; in the same spirit, we shall call its points ‘vertices’ and its
lines ‘edges’.
Note that two edges in a diagram of a graph may intersect at a point that
Figure 1.2. Another diagram of G
tn such a drawing it is understood that no line intersects itself or passes through a point
representing a vertex which is not an end of the corresponding edge—this is clearly always
possible. Graphs and Subgraphs 3
is not a vertex (for example e; and es of graph G in figure 1.1). Those graphs
that have a diagram whose edges intersect only at their ends are called
plartar, since such graphs can be represented in the plane in a simple
manner. The graph of figure 1.3a is planar, even though this is not
immediately clear from the particular representation shown (see exercise
1.1.2). The graph of figure 1.3b, on the other hand, is nonplanar. (This will
be proved in chapter 9.)
Most of the definitions and concepts in graph theory are suggested by the
graphical representation. The ends of an edge are said to be incident with
the edge, and vice versa. Two vertices which are incident with a common
edge are adjacent, as are two edges which are incident with a common
vertex. An edge with identical ends is called a loop, and an edge with
distinct ends a link. For example, the edge es of G (figure 1.2) is a loop; all
other edges of G are links.
Figure 1.3. Planar and nonplanar graphs
A graph is finite if both its vertex set and edge set are finite. In this book
we study only finite graphs, and so the term ‘graph’ always means ‘finite
graph’. We call a graph with just one vertex trivial and all other graphs
nontrivial.
A graph is simple if it has no loops and no two of its links join the same
pair of vertices. The graphs of figure 1.1 are not simple, whereas the graphs
of figure 1.3 are. Much of graph theory is concerned with the study of simple
graphs.
We use the symbols v(G) and e(G) to denote the numbers of vertices and
edges in graph G. Throughout the book the letter G denotes a graph.
Moreover, when just one graph is under discussion, we usually denote this
graph by G. We then omit the letter G from graph-theoretic symbols and
write, for instance, V, E, v and e instead of V(G), E(G), v(G) and e(G). 4 Graph Theory with Applications
Exercises
1.1.1 List five situations from everyday life in which graphs arise naturally.
1.1.2 Draw a different diagram of the graph of figure 1.3a to show that it
is indeed planar.
1.1.3. Show that if G is simple, then e =(5).
1.2. GRAPH ISOMORPHISM
Two graphs G and H are identical (written G=H) if V(G)= V(H),
E(G)= E(H), and tio = tu. If two graphs are identical then they can clearly
be represented by identical diagrams. However, it is also possible for graphs
that are not identical to have essentially the same diagram. For example, the
diagrams of G in figure 1.2 and H in figure 1.1 look exactly the same, with
the exception that their vertices and edges have different labels. The graphs
G and H are not identical, but isomorphic. In general, two graphs G and H
are said to be isomorphic (written G = H) if there are bijections 6: V(G) >
V(H) and ¢:E(G)— E(H) such that o(e) = uv if and only if Wu($(e)) =
6(u)6(v); such a pair (6, 6) of mappings is called an isomorphism between G
and H.
To show that two graphs are isomorphic, one must indicate an isomorph-
ism between them. The pair of mappings (6, 6) defined by
O(vi)=y, O(v2)=x, (vs) =u, (vs) =v, (Us) =w
and
o(e)=h, (e2)=g, —b(es)=b, ble) =a
oes\=e, b(e)=c, (e)=d, h(x) = f
is an isomorphism between the graphs G and H of examples 1 and 2; G and
H clearly have the same structure, and differ only in the names of vertices
and edges. Since it is in structural properties that we shall primarily be
interested, we shall often omit labels when drawing graphs; an unlabelled
graph can be thought of as a representative of an equivalence class of
isomorphic graphs. We assign labels to vertices and edges in a graph mainly
for the purpose of referring to them. For instance, when dealing with simple
graphs, it is often convenient to refer to the edge with ends u and v as ‘the
edge uv’. (This convention results in no ambiguity since, in a simple graph,
at most one edge joins any pair of vertices.)
We conclude this section by introducing some special classes of graphs. A
simple graph in which each pair of distinct vertices is joined by an edge is
called a complete graph. Up to isomorphism, there is just one complete
graph on n vertices; it is denoted by K,. A drawing of Ks is shown in figure
1.4a. An empty graph, on the other hand, is one with no edges. A bipartite Graphs and Subgraphs 5
(a) (b) (c)
Figure 1.4. (a) Ks; (b) the cube; (c) Kya
graph is one whose vertex set can be partitioned into two subsets X and Y,
so that each edge has one end in X and one end in Y; such a partition
(X, Y) is called a bipartition of the graph. A complete bipartite graph is a
simple bipartite graph with bipartition (X, Y) in which each vertex of X is
joined to each vertex of Y; if |X|=m and |Y|=n, such a graph is denoted
by Kan. The graph defined by the vertices and edges of a cube (figure 1.4b)
is bipartite; the graph in figure 1.4c is the complete bipartite graph Ka.
There are many other graphs whose structures are of special interest.
Appendix III includes a selection of such graphs.
Exercises
1.2.1 Find an isomorphism between the graphs G and H of examples 1
and 2 different from the one given.
1.2.2 (a) Show that if G=H, then »(G)=v(H) and e(G) = e(H).
(b) Give an example to show that the converse is false.
1.2.3 Show that the following graphs are not isomorphic:
1.2.4 Show that there are eleven nonisomorphic simple graphs on four
vertices.
1.2.5 Show that two simple graphs G and H are isomorphic if and only if
there is a bijection 6: V(G)— V(H) such that uve E(G) if and
only if 6(u)@(v) € E(H). 6
1.2.6
1.2.7
1.2.8
1.2.9
1.2.10
1.2.11
1.2.12
Graph Theory with Applications
Show that the following graphs are isomorphic:
v)
Let G be simple. Show that ¢ = (2) if and only if G is complete.
Show that
(a) €(Kmn) = mn;
(b) if G is simple and bipartite, then ¢ 0 has bipartition
(X, Y), then [X]=|¥].
Show that, in any group of two or more people, there are always two
with exactly the same number of friends inside the group.
If G has vertices v1, v2,..., v,, the sequence (d(v;), d(v2),..., d(v.))
is called a degree sequence of G. Show that a sequence
(di, da, ..., ds) of non-negative integers is a degree sequence of some
graph if and only if ) d, is even.
A sequence d= (di, da,..., da) is graphic if there is a simple graph
with degree sequence d. Show that
(a) the sequences (7, 6, 5, 4, 3, 3, 2) and (6, 6, 5, 4, 3, 3, 1) are not
graphic; “
(6) if d is graphic and d,=d.=...2d,, then ba is even and
Saskk—p+ E min{k,d} for 1sk=n
(Erdés and Gallai, 1960 have shown that this necessary condition is
also sufficient for d to be graphic.)
Let d= (dy, ds,..., d,) be a nonincreasing sequence of non-negative
integers, and denote the sequence (d:—1, ds—1,...,dau—1,
da,+2,...,dn) by d’.
(a)* Show that d is graphic if and only if d’ is graphic.
(b) Using (a), describe an algorithm for constructing a simple graph
with degree sequence d, if such a graph exists.
(V. Havel, S. Hakimi)
Show that a loopless graph G contains a bipartite spanning subgraph
H such that du(v) =4do(v) for all ve V.
Let S={x,, x2,..., xa} be a set of points in the plane such that the
distance between any two points is at least one. Show that there are
at most 3n pairs of points at distance exactly one.
1.5.10 The edge graph of a graph G is the graph with vertex set E(G) in
which two vertices are joined if and only if they are adjacent edges in 12 Graph Theory with Applications
G. Show that, if G is simple
(a) the edge graph of G has e(G) vertices and > (4() edges;
(b) the edge graph of K; is isomorphic to the complement of the
graph featured in exercise 1.2.6.
1.6 PATHS AND CONNECTION
A walk in G is a finite non-null sequence W = ve:v1e202... exd,, Whose
terms are alternately vertices and edges, such that, for 1=i(”5 1
(6) For v>1, find a disconnected simple graph G with e = ("3 ).
1.6.6 (a) Show that if G is simple and 5 >[v/2]—1, then G is connected.
(b) Find a disconnected ({v/2]— 1)-regular simple graph for v even
1.6.7 Show that if G is disconnected, then G* is connected.
1.6.8 (a) Show that if ee E, then w(G)<.w(G-—e)<(G)+1.
(b) Let ve V. Show that G—e cannot, in generat, be replaced by
G~—v in the above inequality.
1.6.9 Show that if G is connected and each degree in G is even, then, for
any ve V, w(G—v)<4d(v).
) then G is connected. 14 Graph Theory with Applications
1.6.10 Show that any two longest paths in a connected graph have a vertex
in common.
1.6.11 If vertices u and v are connected in G, the distance between u and
v in G, denoted by do(u, v), is the length of a shortest (u, v)-path in
G;; if there is no path connecting u and v we define do(u, v) to be
infinite. Show that, for any three vertices u, v and w, d(u, v)+
d(v, w)=d(u, w).
1.6.12 The diameter of G is the maximum distance between two vertices
of G. Show that if G has diameter greater than three, then G* has
diameter less than three.
1.6.13 Show that if G is simple with diameter two and A=v—2, then
e22v-4. .
1.6.14 Show that if G is simple and connected but not complete, then G
has three vertices u, v and w such that uv, owe E and uw¢E.
1.7 cycles
A walk is closed if it has positive length and its origin and terminus are the
same. A closed trail whose origin and internal vertices are distinct is a cycle.
Just as with paths we sometimes use the term ‘cycle’ to denote a graph
corresponding to a cycle. A cycle of length k is called a'k-cycle; a k-cycle is
odd or even according as k is odd or even. A 3-cycle is often called a
triangle. Examples of a closed trail and a cycle are given in figure 1.10.
Using the concept of a cycle, we can now present a characterisation of
bipartite graphs.
Theorem 1.2 A graph is bipartite if and only if it contains no odd cycle.
Proof Suppose that G is bipartite with bipartition (X, Y), and let C=
v00; . .. UxVo be a cycle of G. Without loss of generality we may assume that
vo X. Then, since vov; € E and G is bipartite, v,¢ Y. Similarly v,eX # 3,
in general, vie X and vase Y. Since vo€ X, %€ Y. Thus k =2i+1, for
some i, and it follows that C is even.
Closed trail: uevhxgwfwdvbu
Cycle: xaubvhx
Figure 1.10 Graphs and Subgraphs 15
It clearly suffices to prove the converse for connected graphs. Let G be a
connected graph that contains no odd cycles. We choose an arbitrary vertex
u and define a partition (X, Y) of V by setting
X={xeV|d(u,x) is even}
Y={yeV|d(u, y) is odd}
We shall show that (X, Y) is a bipartition of G. Suppose that v and w are
two vertices of X. Let P be a shortest (u,v)-path and Q be a shortest
(u, w)-path. Denote by u; the last vertex common to P and Q. Since P and
Q are shortest paths, the (u, u:)-sections of both P and Q are shortest
(u, us)-paths and, therefore, have the same length. Now, since the lengths of
both P and Q are even, the lengths of the (u:, v)-section P, of P and the
(us, w)-section Q, of Q must have the same parity. It follows that the
(v, w)-path P7'Q, is of even length. If v were joined to w, Pi'Q,wo would
be a cycle of odd length, contrary to the hypothesis. Therefore no two
vertices in X are adjacent; similarly, no two vertices in Y are adjacent O
Exercises
1.7.1 Show that if an edge e is in a closed trail of G, then e is in a cycle of
G.
1.7.2 Show that if 8=2, then G contains a cycle.
1.7.3* Show that if G is simple and 8 = 2, then G contains a cycle of length
at least 6+ 1.
1.7.4 The girth of G is the length of a shortest cycle in G; if G has no
cycles we define the girth of G to be infinite. Show that
(a) a k-regular graph of girth four has at least 2k vertices, and (up to
isomorphism) there exists exactly one such graph on 2k vertices;
(b) a k-regular graph of girth five has at least k?+1 vertices.
1.7.5 Show that a k-regular graph of girth five and diameter two has
exactly k*+1 vertices, and find such a graph for k =2, 3. (Hoffman
and Singleton, 1960 have shown that such a graph can exist only if
k =2, 3, 7 and, possibly, 57.)
1.7.6 Show that
(a) if ¢=v, G contains a cycle;
(b)* if e=v+4, G contains two edge-disjoint cycles. (L. Pésa)
APPLICATIONS
1.8 THE SHORTEST PATH PROBLEM
With each edge e of G let there be associated a real number w(e), called its
weight. Then G, together with these weights on its edges, is called a weighted 16 Graph Theory with Applications
1 2
Figure 1.11. A (uo, vo)-path of minimum weight
graph. Weighted graphs occur frequently in applications of graph theory. In
the friendship graph, for example, weights might indicate intensity of
friendship; in the communications graph, they could represent the construc-
tion or maintenance costs of the various communication links.
If H is a subgraph of a weighted graph, the weight w(H) of H is the sum
of the weights w(e) on its edges. Many optimisation problems amount
ents tn 8 ly opt
to finding, in a weighted graph, a subgraph of a certain type with minimum
(or maximum) weight. One such is the’ shortest path problem: given a railway
network connecting various towns, determine a shortest route between two
specified towns in the network.
Here one must find, in a weighted graph, a path of minimum weight
connecting two specified vertices uo and vo; the weights represent distances
by rail between directly-linked towns, and are therefore non-negative. The
path indicated in the graph of figure 1.11 is a (uo, vo)-path of minimum
weight (exercise 1.8.1).
We now present an algorithm for solving the shortest path problem. For
clarity of exposition, we shall refer to the weight of a path in a. weighted
graph as its length; similarly the minimum weight of a (u, v)-path will be
called the distance between u and v and denoted by d(u, v). These defini-
tions coincide with the usual notions of length and distance, as defined in
section 1.6, when all the weights are equal to one.
It clearly suffices to deal with the shortest path problem for simple graphs;
so we shall assume here that G is simple. We shall also assume that all the
weights are positive. This, again, is not a serious restriction because, if the
weight of an edge is zero, then its ends can be identified. We adopt the
convention that w(uv)=© if uvé E. Graphs and Subgraphs 17
The algorithm to be described was discovered by Dijkstra (1959) and,
independently, by Whiting and Hillier (1960). It finds not only a shortest
(uo, ¥o)-path, but shortest paths from up to all other vertices of G. The basic
idea is as follows.
Suppose that S is a proper subset of V such that woe S, and let § denote
V\S. If P= uo... ad is a shortest path from uo to S then clearly ie S and
the (uo, #)-section of P must be a shortest (uo, ii)-path. Therefore
d(uo, 5) = d(uo, t) + (ad)
and the distance from uo to S is given by the formula
d(Uo, S) = min{d(uo, u)+w(uv)} (1.1)
vs
This formula is the basis of Dijkstra’s algorithm. Starting with the set
So={uo}, an increasing sequence So, $;,..., S,-1 of subsets of V is con-
structed, in such a way that, at the end of stage i, shortest paths from uo to
all vertices in S, are known.
The first step is to determine a vertex nearest to uo. This is achieved by
computing d(uo, So) and selecting a vertex u,¢So such that d(uo, u:)=
(uo, So); by (1.1)
(uo, So) = min{d(uo, u) + w(ur)} = min{w(uov)}
1
and so d(uo, So) is easily computed. We now set S:={uo, us} and let Py
denote the path uous; this is clearly a shortest (uo, u:)-path. In general, if the
set S,={uo, ui, ..., us} and corresponding shortest paths P,, P2,..., Py have
already been determined, we compute d(uo, S,) using (1.1) ‘and select a
vertex ui€S, such that d(uo, ues) =d(uo, Si). By (1.1), d(Uo, tues) =
(Uo, uj) + w(ujtr1) for some jk; we get a shortest (Uo, Ux+1)-path by
adjoining the edge ujux., to the path P,.
We illustrate this procedure by considering the weighted graph depicted in
figure 1.12a. Shortest paths from uo to the remaining vertices are deter-
mined in seven stages. At each stage, the vertices to which shortest paths
have been found are indicated by solid dots, and each is labelled by its
distance from uo; initially uo is labelled 0. The actual shortest paths are
indicated by solid lines. Notice that, at each stage, these shortest paths
together form a connected graph without cycles; such a graph is called a tree,
and we can think of the algorithm as a ‘tree-growing’ procedure. The final
tree, in figure 1.12h, has the property that, for each vertex v, the path
connecting up and v is a shortest (uo, v)-path.
Dijkstra’s algorithm is a refinement of the above procedure. This refine-
ment is motivated by the consideration that, if the minimum in (1.1) were to
be computed from scratch at each stage, many comparisons would be ts dhs dhs AS
Pst: Test: Bey Wey
ENNIS
is hs bs Bs
DN Psp s ay pe 65 3 TABS 3
OSE SESE
Graph Theory with Applications
Figure 1.12. Shortest path algorithm Graphs and Subgraphs 19
repeated unnecessarily. To avoid such repetitions, and to retain computa-
tional information from one stage to the next, we adopt the following
labelling procedure. Throughout the algorithm, each vertex v carries a label
I(v) which is an upper bound on d(uo, v). Initially I(uo) =0 and I(v) = for
v# wo. (In actual computations © is replaced by any sufficiently large
number.) As the algorithm proceeds, these labels are modified so that, at the
end of stage i,
I(u)=d(uo,u) for ueS,
and
I(v) = min{d(uo, u)+ w(uv)} for ve,
Dijkstra’s Algorithm
1. Set I(uo)=0, I(v) == for v# uo, So={us} and i=0.
2. For each veS, replace I(v) by min{I(v); I(u)+w(uv)}. Compute
min{I(v)} and let wis: denote a vertex for which this minimum is attained.
Set Sir = S:U {uss}.
3. If i=v—1, stop. If i
