Solutions Manual

for

GRAPHS & DIGRAPHS

6th Edition

Gary Chartrand
Western Michigan University

Linda Lesniak
Western Michigan University

Ping Zhang
Western Michigan University
 Table of Contents
Preface to the Sixth Edition xi

1. Introduction 1

1.1 Graphs 1
1.2 The Degree of a Vertex 5
1.3 Isomorphic Graphs 7
1.4 Regular Graphs 12
1.5 Bipartite Graphs 13
1.6 Operations on Graphs 16
1.7 Degree Sequences 18
1.8 Multigraphs 25
• Exercises for Chapter 1 28

2. Connected Graphs and Distance 37

2.1 Connected Graphs 37

2.2 Distance in Graphs 44
• Exercises for Chapter 2 51

3. Trees 57

3.1 Nonseparable Graphs 57

3.2 Introduction to Trees 62
3.3 Spanning Trees 69
3.4 The Minimum Spanning Tree Problem 81
• Exercises for Chapter 3 86

4. Connectivity 95

4.1 Connectivity and Edge-Connectivity 95

4.2 Theorems of Menger and Whitney 102
• Exercises for Chapter 4 110

5. Eulerian Graphs 115

5.1 The Königsberg Bridge Problem 115

5.2 Eulerian Circuits and Trails 117
• Exercises for Chapter 5 123

6. Hamiltonian Graphs 125

i
 6.1 Hamilton’s Icosian Game 125
6.2 Sufficient Conditions for Hamiltonian Graphs 128
6.3 Toughness of Graphs 134
6.4 Highly Hamiltonian Graphs 140
6.5 Powers of Graphs and Line Graphs 145
• Exercises for Chapter 6 154

7. Digraphs 161

7.1 Introduction to Digraphs 161

7.2 Strong Digraphs 166
7.3 Eulerian and Hamiltonian Digraphs 167
7.4 Tournaments 169
7.5 Kings in Tournaments 179
7.6 Hamiltonian Tournaments 180
• Exercises for Chapter 7 184

8. Flows in Networks 191

8.1 Networks 191

8.2 The Max-Flow Min-Cut Theorem 199
8.3 Menger Theorems for Digraphs 207
• Exercises for Chapter 8 212

9. Automorphisms and Reconstruction 217

9.1 The Automorphism Group of a Graph 217

9.2 Cayley Color Graphs 223
9.3 The Reconstruction Problem 228
• Exercises for Chapter 9 235

10. Planar Graphs 239

10.1 The Euler Identity 239

10.2 Maximal Planar Graphs 248
10.3 Characterizations of Planar Graphs 252
10.4 Hamiltonian Planar Graphs 264
• Exercises for Chapter 10 268

11. Nonplanar Graphs 275

11.1 The Crossing Number of a Graph 275

ii
 11.2 The Genus of a Graph 286
11.3 The Graph Minor Theorem 300
• Exercises for Chapter 11 302

12. Matchings, Independence and Domination 305

12.1 Matchings 305

12.2 1-Factors 310
12.3 Independence and Covers 317
12.4 Domination 322
• Exercises for Chapter 12 329

13. Factorization and Decomposition 335

13.1 Factorization 335

13.2 Decomposition 343
13.3 Cycle Decomposition 345
13.4 Graceful Graphs 351
• Exercises for Chapter 13 358

14. Vertex Colorings 363

14.1 The Chromatic Number of a Graph 363

14.2 Color-Critical Graphs 371
14.3 Bounds for the Chromatic Number 374
• Exercises for Chapter 14 385

15. Perfect Graphs and List Colorings 393

15.1 Perfect Graphs 393

15.2 The Perfect and Strong Perfect Graph Theorems 402
15.3 List Colorings 405
• Exercises for Chapter 15 410

16. Map Colorings 415

16.1 The Four Color Problem 415

16.2 Colorings of Planar Graphs 426
16.3 List Colorings of Planar Graphs 428
16.4 The Conjectures of Hajós and Hadwiger 434
16.5 Chromatic Polynomials 438
16.6 The Heawood Map-Coloring Problem 444

iii
iv

• Exercises for Chapter 16 448

17. Edge Colorings 453
17.1 The Chromatic Index of a Graph 453
17.2 Class One and Class Two Graphs 460
17.3 Tait Colorings 467
• Exercises for Chapter 17 476
18. Nowhere-Zero Flows and List Edge Colorings 481

18.1 Nowhere-Zero Flows 481

18.2 List Edge Colorings 491
18.3 Total Colorings 495
• Exercises for Chapter 18 500
19. Extremal Graph Theory 503

19.1 Turán’s Theorem 503

19.2 Extremal Subgraphs 505
19.3 Cages 509
• Exercises for Chapter 19 520

20. Ramsey Theory 523

20.1 Classical Ramsey Numbers 523
20.2 More General Ramsey Numbers 532
• Exercises for Chapter 20 538

21. The Probabilistic Method 543

21.1 The Probabilistic Method 543
21.2 Random Graphs 554
• Exercises for Chapter 21 561

Hints and Solutions to Odd-Numbered Exercises 563

Bibliography 589
Supplemental References 606
Index of Names 607
Index of Mathematical Terms 613
List of Symbols 625
Chapter 1

Introduction

Section 1.1. Graphs

1. This information can be modeled (or represented) by the graph shown in
Figure 1.1, where each small circle indicates a box and a line segment between
two boxes indicates that these two boxes contain at least one wire segment of
the same color.
B
.... 8 ..
B1
.... .... ...... .......
......................... ........... .....
.
............ ............ .................. ..
...... .... .........
.... . ... ..... ......... ..... ... .....
..... ... .... ......... ............
.... ...................
. .....
..... ... ... ....
..... ........
.....
..... . ... .......... .....
....
. . .. ..... .... .. ..... .....
. .
... .........
. .
. ......... .........
. ... . ... ........
. ... ..... ...
...... .. . ... .
..... ..................................................................................................................................................................................
B 7 ......................................... .......
.....
.....
... ........ B2
... ..... ........ ..... ... .....
..... ....
..
...
... ......... .......... ......... .... ..... ... ...
... ....... ........... ... . ..
... .. ...... . ..
.. ..... .. . ...... . .
... ... ...... ........ .................... ........ ...
.... ........ .........
.............
... ....... ...
.... ......... ... ......
..... ....
..
... .... ........ ......... ... ..........
... ........ .. .
. ..... ..
.....
... .......... ..... ... .........
... ......... ............
... ......... ..... ... ... ............ .......
... ...... ..... ..... ... .. .
............... ... ..... ... ... ... ............. ......
.. .........................................................................................................................................................................
.
. .
B3
B 6 .................................... ...... .....
......
...
....
...
... .. ... ...
.....
..... .......... .....
..... ..... .... .... .....
..... .............. .....
.....
..... .... ................. ..... .. ... ..
. . . .........
..... ... ......... ... . . ..
..... ... ......... ........ ..... ..... .... .........
..... .. ......... ..... .. .. .. ....
............. ......... ............... .....
... .... ........... ....
.... . . .. .
..

B5 B4

Figure 1.1: The graph in Exercise 1

2. The graph G is shown in Figure 1.2. The degree of ∅ is 7, the degree of each
of {1}, {2} and {3} is 4, the degree of each of {1, 2}, {1, 3} and {2, 3} is 2 and
the degree of S is 1. The size of G is 13.

Section 1.2. The Degree of a vertex

3. Denote the degree of the remaining vertices by x. Since there are 8 vertices
of degree x, it follows that 5 · 4 + 6 · 5 + 7 · 6 + 8x = 2 · 58. Thus, x = 3 and
so the degree of each remaining vertex is 3.

1
2 CHAPTER 1. INTRODUCTION

∅ ....
... .....
...................................
.......... ..... ... ........ ........
... . . ... ............. ......... ..... ............. .................
.
.... ...... .... . ... ....
... ...... ......................
... ....
S .... ..... ..
..... ... ... ............. .. ... {1}
.... .... . ...... ... .. ..
...
......
..
. ...... .
. ............. ........ .... ....
. ..... .. ...
..... .. .........
...
..... ... ..
..... ................. ...
.. ...
.
.... .. .... ..
.. ........ .....
..... ........... ... ......... ...
..... ........ .... ... ... ...... ...
...... ........
...... ........... .. ... ... .............
{2, 3} ....
...
...
. ...
... ...............................
. . ... .
{2}
... ................. .... ....
.
. .. ... .. ..
... ..........
... . . .. .... .......... ... ... ..
.... ... ...
... ........
..... .......... ... ... ...
.......... ............
{1, 3} ..... .............. .. .
....
.........
. . .. ................ {3}
.....
. .
.. . ..
...
.....................
.....
{1, 2}

Figure 1.2: The graph G in Exercise 2

4. Proof. Assume, to the contrary, that G has at most k + 2 vertices of de-

gree k + 1, at most k vertices of degree k + 2 and at most k + 1 vertices of
degree k + 3. Since the order of G is n = 3k + 3, it follows that G has exactly
k + 2 vertices of degree k + 1, exactly k vertices of degree k + 2 and exactly
k + 1 vertices of degree k + 3. In each case, G has an odd number of odd
vertices, which is impossible.

5. Proof. Let G be a graph with r vertices of degree r, r + 1 vertices of degree

r +1 and r +2 vertices of degree r +2. Thus, the order of G is 3r +3. First, we
show that r is odd. Assume, to the contrary, that r is even. Then G contains
an odd number r + 1 of odd vertices, which is impossible by Corollary 1.5.
Thus, r is odd and G contains 2r + 2 vertices of odd degree.

6. Let G be the graph of order 2k with V (G) = {u1 , u2 , . . . , uk , v1 , v2 , . . . , vk }.

For each i with 1 ≤ i ≤ k, join each vertex ui to the i vertices v1 , v2 , . . . , vi .
Then deg ui = i and deg vi = k + 1 − i for 1 ≤ i ≤ k.

Section 1.3. Isomorphic Graphs

7. (a) G1 ∼
= G2 .
(b) H1 ∼6 H2 . For example, H1 has two vertices of degree 4, while H2 has
=
three vertices of degree 4.

8. (a) There are 34 such graphs, each of which has size m for some m with
0 ≤ m ≤ 10. By using complementary graphs the number of graphs of
order 5 and size m equals the number of graphs of order 5 and size 10−m
(see Figure 1.3).
(b) The minimum size of a graph G of order 5 such that every graph of order
5 and size 5 is isomorphic to some subgraph of G is 7. First, observe
that the graph G = P4 ∨ K1 has the desired property. If the minimum
size were 6, then G must consist of a 5-cycle C and an edge joining two
 3

q .
.....
q q
......
..... ....
q
q q q .....
....
..... q q ..... ........
..... .. q q q
q q q q q q q q

q
......... ...... q q..
......... ....... q
..... ........ ....... ... ... ........
q
......
.....
..... .........
.... q q....
....
.....
. q q ..... ... .....
..... ... . q ...
.. q .....
. q
... ...
...
q q q q q
...
.. ... q q q
q q q q
q q ... .
.........
....
.... .....
.........
..... .........
............. ..... .........
q .....
q
. ...
..... .... ..... .... ... ........
q q
.....
..........
q
......
.....
..... ........
q
..... q
......
..... ........
..... ... q q
..... .. .. ...
.... .. .. ...
.... .... ..... ... q ... q
... ..... q .......
...
.... .....
.. ..........
... ......
..

... .....
... ... .. .. ... ... ... .......
...
... .........
q
... ....
.....
.
.
q
...
q...
. q q
.. ...
...
... q
...
.. q q q ...
. q
.... q ...
q
.....
....

q
......... q
........
q .....
.... .....
q
.
.........
q
.........
q ..
..........
..... ......... ..... ........ ..... .........
q
......
..... ........
.... .... q q
.......
.....
q
.....
... q
......
.....
..... ........
.....
q q
......
..... ........
.... ....
.. q q
.......
..... .....
q
. q
..........
..... ... q
... ... ... ...... ... .....
... ... ... ... ........
... .......
... ... ... .......... ... ......... ... ... ... ... .......
q...
. ...q
...
q
... .......
......................q q... ....
...
..
q
...
q
...
. q q
...
... q..
.
... q ... .
q
.....
.

Figure 1.3: Graphs for Exercise 8

nonconsecutive vertices of C. This graph, however, does not have the

desired property.

9. (a) Proof. Let φ : V (G) → V (H) be an isomorphism from G to H. Then

|S| = |T | and the mapping φ0 : S → T defined φ0 (v) = φ(v) for all v ∈ S
is a bijection from S to T . To show that φ0 is an isomorphism from
G[S] to H[T ], it remains to show that φ0 maps adjacent vertices in G[S]
to adjacent vertices in H[T ] and maps nonadjacent vertices in G[S] to
nonadjacent vertices in H[T ]. Let u, v ∈ V (G[S]) = S. Since G[S] is an
induced subgraph of G, it follows that u and v are adjacent in G[S] if and
only if u and v are adjacent in G. Since φ is an isomorphism from G to
H, it follows that u and v are adjacent in G if and only if φ(u) = φ0 (u)
and φ(v) = φ0 (v) are adjacent in H. Since (1) φ0 (u), φ0 (v) ∈ T and
(2) H[T ] is an induced subgraph of H, it follows that φ0 (u) and φ0 (v)
are adjacent in H if and only if φ0 (u) and φ0 (v) are adjacent in H[T ].
Therefore, u and v are adjacent in G[S] if and only if φ0 (u) and φ0 (v) are
adjacent in H[T ]. Therefore, φ0 is an isomorphism from G[S] to H[T ]
and so G[S] ∼= H[T ].
(b) Let r = 3. Then G[S] has no edges and G[T ] has one edge. Thus
G[S] ∼
6 G[T ] and so G ∼
= 6 H.
=

10. (a) The three graphs with this property are shown in Figure 1.4(a).
(b) The graphs G, H, F1 and F2 in Figure 1.4(b) have this property.

Section 1.4. Regular Graphs

11. Proof. Denote the size of G by m. Thus, m = rn/2. The average degree of
4 CHAPTER 1. INTRODUCTION

................. ........ ............... .

..............
.
........... ..... .........
..... .....
.......... ..... ... ........
.
..... ........ .... ..... .... .
.
.....
.
. .....
.....
...... ........... ..............
..... ........
................. ..........
............ ............... ...............
...

u (a) u u
......
.... ....
.....
.... .... .........
...... ... .......... ............ .............. ...... .... .......... .....................
...... ..... ...... ......
......
..... ...... .... ......
..... ...... ......
...............
..........
... ... ... ... . .. ..... .
...... .........
y ....
. .
............
...... ....
v .... ..... ... .... y .. .
.............
. ........
....... ...
...... ... v y ..
...
.......... .......... v
... .
....
..
.
. ... ... .. ....
. .
......
. ...
... .. . .. ... ... ... ..
. ...
G: ...
... ..........
......
.. .
... H: ... ... ... ...
... F1 : ......
......
....
.
F2 : ...
. ... ... ... ...
....... ...
................ ................ ...............
..
...........
.
.............. .............. ....
...........
.
..............
.
..............
. .... .
x w x w x w
(b)

Figure 1.4: The graphs Exercise 10

G is  
2m rn/2
=2 = r.
n n
Since G is not r-regular, G contains a vertex v such that deg v 6= r. We
consider two cases.

Case 1. deg v > r. Not all vertices of G have degree r or more, for otherwise,
 rn  X
2m = 2 = rn = deg v > rn,
n
v∈V (G)

which is impossible. Hence G contains a vertex u with deg u < r. Therefore,

∆(G) ≥ deg v ≥ r + 1 and δ(G) ≤ deg u ≤ r − 1 and so ∆(G) − δ(G) ≥ 2.

Case 2. deg v < r. Not all vertices of G have degree r or less, for otherwise,
 rn  X
2m = 2 = rn = deg v < rn,
n
v∈V (G)

which is impossible. Hence, G contains a vertex w with deg w > r. Therefore,

∆(G) ≥ deg w ≥ r + 1 and δ(G) ≤ deg v ≤ r − 1 and so ∆(G) − δ(G) ≥ 2.
12. Let k ≥ 2. For 0 ≤ i ≤ k − 1, let Gi = C3(k−i) + iC3 . The graphs
G0 , G1 , . . . , Gk−1 are pairwise non-isomorphic.
13. Let V (G) = {u, v, w, x} and E(G) = {uv, vw, wx, xu, vx}. Let e = vx. Then
G − e = C4 and G − u = C3 .
14. (a) Let G1 = 2C3 and G2 = C6 .
(b) Let H1 = 3C3 and H2 = C9 .
15. If G itself is r-regular, then there is nothing to prove. So we may assume
that G is not r-regular. Let G0 be another copy of G and join corresponding
 5

vertices whose degrees are less than r, calling the resulting graph G1 . If G1
is r-regular, then G1 has the desired properties. If not, then we continue this
procedure until arriving at an r-regular graph Gk , where k = r − δ(G).

16. For each i with 1 ≤ i ≤ j, let vi be the vertex of G with deg vi = r − i and let
G0 be another copy of G where the vertex vi0 in G0 corresponds to the vertex
vi in G. Let H be the graph of order 2n obtained from G and G0 by joining vi
0 0
to vj−i+1 , vj−i+2 , . . . , vj0 for 1 ≤ i ≤ j. Then H is an r-regular graph of order
2n containing G as an induced subgraph.

17. The Petersen graph.

18. (a) G6,1 = K6 and G5,2 is the Petersen graph.

(b) The graph Gn,k is an n−k n



k -regular graph of order k .

Section 1.5. Bipartite Graphs

19. Let x be the number of vertices of degree 8 in W . Then n = 10 + 4 + 3 + x
and m = 6 · 10 = 60. Since m = 4 · 2 + 3 · 4 + 8x = 60, it follows that x = 5
and so n = 22.

20. Proof. We have seen that the size of the complete bipartite graph Kb n c,d n e
2 2
is bn2 /4c. For every bipartite graph with partite sets U and W with s = |U | ≤
|W | = t and s + t = n, clearly Ks,t has the maximum size. If 0 ≤ t − s ≤ 1,
then Ks,t = Kbn/2c,dn/2e and the size of Ks,t is bn2 /4c. Suppose that t−s ≥ 2.
Then t = dn/2e + p and s = bn/2c − p for some p ≥ 1. Then the size of Ks,t
is

(dn/2e + p) (bn/2c − p) = bn2 /4c + p (bn/2c − dn/2e) − p2 < bn2 /4c.

Hence, Kb n c,d n e is the only bipartite graph having size bn2 /4c.
2 2

21. Proof. Suppose that the partite sets of a 3-partite graph G of order n =
3k and size m are A, B and C, where |A| = a, |B| = b and |C| = c and
a + b + c = 3k. We may assume that a ≥ b ≥ c. Then a ≥ k and c ≤ k. Hence,
a = k + x, c = k − y and b = 3k − a − c = k − x + y, where x, y ≥ 0. Then

m ≤ ab + ac + bc = 3k 2 − x2 + xy − y 2
y 2 3y 2
 
2
= 3k − x − + ≤ 3k 2 ,
2 4

where m = 3k 2 if and only if x = y = 0 and so a = b = c = k, in which case,

G = Kk,k,k .

22. Proof. Define a relation R on V (G) by u R v if uv ∈ / E(G). The relation R

is clearly reflexive and symmetric. For vertices u, v and w of G, if uv ∈
/ E(G)
and vw ∈ / E(G), then uw ∈ / E(G). Thus, R is transitive and so R is an
6 CHAPTER 1. INTRODUCTION

equivalence relation on V (G). Hence, V (G) is partitioned into equivalence

classes V1 , V2 , . . . , Vk where k ≥ 2. Thus, two vertices are related if and
only if they belong to the same equivalence class. Also, vertices belonging to
different equivalence classes are not related and are therefore adjacent. Thus,
G is a complete k-partite graph with partite sets V1 , V2 , . . . , Vk .

Section 1.6. Operations on Graphs

23. (Every bipartite graph having no partite set with three or more vertices.) If
G has a partite set with three or more vertices, then G has a triangle. Thus,
we may assume that every partite set has at most two vertices and so the
order of G is at most 4. Then G is a subgraph of K2,2 = C4 and is therefore
bipartite.
24. Proof. Suppose that G contains a vertices of degree less than k, b vertices of
degree greater than k and c vertices of degree k. Thus, G contains b vertices
of degree less than k, a vertices of degree greater than k and c vertices of
degree k. Since the vertices of G and G have exactly the same degrees, a = b.
Hence, the order of G is 2a + c. Since G has odd order, c is odd.
25. (a) The graph G is C7 or C3 + C4 .
(b) The graph G is one of C9 , C6 + C3 , C5 + C4 or 3C3 . Thus, there are four
such graphs.
26. Proof. Assume, to the contrary, that there exists an r-regular self-complementary
graph G of even order n ≥ 2. Then G is (n−1−r)-regular and so r = n−1−r.
Hence, 2r = n − 1, which is odd. This is impossible.
27. Proof. We proceed by induction on n. The graph C5 shows that the result
holds for n = 1. Assume that there is a regular self-complementary graph H
of order 5k , where k ≥ 1. Let Hi (i = 1, 2, . . . , 5) be five copies of H. Let G
be the graph consisting of the graphs H1 , H2 , . . . , H5 together with all edges
joining the vertices of Hi and the vertices of Hi+1 for i = 1, 2, . . . , 5, where
H6 = H1 . Then G is a regular self-complementary graph of order 5k+1 .
28. Proof. Suppose that the order n of G2 is n = 2k and that there are exactly a
vertices of degree less than k in G2 . Since G2 is self-complementary, there are
exactly a vertices of degree greater than k in G2 . Furthermore, if deg v < k,
then degG v = n − deg v > n − k = k. Observe that G is obtained from G1
and G2 by joining each vertex in G2 whose degree is greater than k to every
vertex in G1 (that is, joining each vertex in G2 whose degree is less than k to
every vertex in G1 ). Since G1 ∼ = G1 and G2 ∼ = G2 , the graph G is obtained
from G1 and G2 by joining each vertex in G2 whose degree is less than k to
every vertex in G1 , which is isomorphic to G.
29. The following proof essentially makes use of Exercise 28.
Proof. We use induction on n when (1) n = 4k, k ≥ 1 and (2) n = 4k + 1,
k ≥ 0.
 7

(1) Suppose that n = 4k. Since P4 is self-complementary, the statement is

true for k = 1. Assume that there exists a self-complementary graph G of
order 4k, where k ≥ 1. We show that there exists a self-complementary
graph H of order 4(k + 1). Let H be the graph obtained from G and
P4 = (v1 , v2 , v3 , v4 ) by joining each of v1 and v4 to every vertex of G.
Then H is self-complementary and the order of H is 4(k + 1).
(2) Suppose that n = 4k + 1. Since K1 is self-complementary, the statement
is true for k = 0. Now assume that there exists a self-complementary
graph G of order 4k + 1, where k ≥ 0. We show that there exists a self-
complementary graph H of order 4(k +1)+1. Again, let H obtained from
G and P4 = (v1 , v2 , v3 , v4 ) by joining each of v1 and v4 to every vertex of
G. Then H is self-complementary and the order of H is 4(k + 1) + 1.

30. (a) and (b) see Figures 1.5(a) and (b), respectively.

v
.
v
....... .......
..........
...................
...... ...... ... ...
...
....... ...... .. ....
.
.
...... ......
... ...
...... ...... .
.............. ...
... .
.......
.. .... ...
. ......
...... .. z ...... ..
... ........
.....
w
z .....
..........
...
..
............. ............
w ......
...... ..
..
.
. u ...
..
......... .... ..........
......
... ... ...... . . . ..
... u ... .......
.. ............... .............
... ... ... ............ ........ ..
G: ...
...
... ...
... H: .
... ... ...... ...... .... ...
.. .... .................. ........ ....
.. .. ... .... ...... .... ..
.
.............. ... ...... ....... ...... .........
......... ...
. ............. ..................... .......
y x y x
(a)
v v
......... .
.......
.................... ............
..... ......
...... ... ....
........ u ......
...... .. ... ...
. . ......
......
..
.
... ...................
......
..... z ...
..................
....... ....
.. ... ...... ............
... ........ w
.. . . ....... .... ................... .............
.
. ... . .......... ........
................................ ............ .. . ....... .....
z .................
... ............... ........
.............. .. w ... ...............
... ..................... .....
... ...
.
... ....... ........... s
..... ....... ....... ................ .......... ....
. ..
...... ...... ..
.
... ...... ....... ..
...
... ..
...... .... .
........ ....
.... .. . .............
.
..............
s .
..............
..........
........
G: ..... ...... ... H: y .........
...
...
...
.....
.....
...................
...... ...
..
..........
... ....... ...... ..
...... .. x
... ........ .
... ... ....... .......... ....
.. ..........
...
..
... ...
...
..............
t ...
..............
... ...........
. ..... .......... .......
................... ..... ..
..
... .....
y x (b) t u

Figure 1.5: The graphs in Exercise 30(a) and (b)

31. Proof. Since there is a self-complementary graph of order n for every integer
n with n ≡ 0 (mod 4) or n ≡ 1 (mod 4) by Exercise 29, we may assume
that n ≡ 2 (mod 4) or n ≡ 3 (mod 4). We verify this for n ≡ 2 (mod 4)
by induction. That the result holds for n = 6 follows from Exercise 30(a).
Assume, for some integer k j ≥ 6 with k ≡ 2 (mod 4), that there exists a
k

k
graph G1 of order k and size 2 /2 that is isomorphic to a graph H1 ⊆ G1 .
Add to G1 the path P = (p1 , p2 , p3 , p4 ) of order 4 and join p1 and p4 to
everyj vertex kof G1 and denote this graph by G. Then G has order k + 4,
size k+4


2 /2 and is isomorphic to a graph H of G, where H is constructed
from H1 by adding (p2 , p4 , p1 , p3 ) and joining p2 and p3 to every vertex of H1 .
Hence, the result holds if n ≡ 2 (mod 4).
8 CHAPTER 1. INTRODUCTION

The proof for the case when n ≡ 3 (mod 4) is similar.

32. The subgraph H in Figure 1.6 is isomorphic to P .
.......
...................... .......
.............. ....................
..... ...... .......
...... ...... ..... ....... ... ......
.. . ...... ...... ................. .........................................................
......
. . ...... ............
. .. .......
..
......
. ..
......
...... .
.
.
.... ....
. ...
. .........
... ......
.... ... .... .. ..
... . ... .
........
......
...... ... ... ... ........... ... ....
. .... . . ...... . .. ... .
. ... ....
..... . .... .. ... . ...............................
....... .
. ...... . ..
.. ......... ...... .......... .... .....
.................... .. .... ........
.... .......... . .. .... ............. ...
.
.. ....... .. .
..........
....... .
P : ...... ..........................
... .........
...... .. .
... ...
...
... .
.... .
.. ............ ...... .......
..
..
.
..
.
.
.
... ...
. H: ... ............
.... ... ...... .........
.
. .... .... .
........... ........ ...
... ...
... . . .
.. . .. ... ... ..... . .. .. ... ..
... ....... .. .
.......
...........
.
. .
... ... ... ..... .... ... ....
. ... ...
... . ........ ............... .. ..... ..... ... . ...
. ... ...
... . ....... .. . .
..... .. . .
. .... ... ..
... .. ... ..... . . .
... .. ........ ........... .... ...
...... ...... ... .
. ...
. ....
.
... ..... ............. ....... . ....
.... ... ......
......... ... ....
...
... .................. ... .... ..
.. .
... ..... .
. ..
.. ......... .... ... ..
......
.
.
... ... .
.. .
. ... ....... .
..... . ... .
... ... ....
... .......... ..................
.. .. ...... ..
... ..... .... ... ........... ..
... ....... ....
.... ... ... ...........
.... ..... ........ .....
... ...... .... .... .... .
.............
........
. .
..... .....
...... ..... . .
.. .... .........
.......... ....... ........ ... ...
..... ............. ............................................
......... ...............

Figure 1.6: The graphs Exercise 32

33. (a) The degree of u1 in G1 + G2 is degG1 u1 .

(b) The degree of u1 in G1 ∨ G2 is degG1 u1 + n2
(c) The degree of (u1 , u2 ) in G1  G2 is degG1 u1 + degG2 u2 .
34. The graph P3 ∨ 2P3 has order 9 and size 24, the graph P3  2P3 has order 18
and size 24, and the graph Q1 + Q2 + Q3 has order 14 and size 17.

Section 1.7. Degree Sequences

35. See Figure 1.7. Delete the edges uw and xy of G and add the edges ux and
wy. This produces a graph isomorphic to H.

v v ....
........ .... ..
................... ...................
..... ..... .... .....
..
......
. ..... ..
....... .....
.. ..... .. .....
..... .. ..... ........
.............. .............. u ...............
u w ......
...... ......
........
.
w
...... ...........
..........
.
..
G: H: ...... ............
...... ......
.............. .............. ...... ...........
................ .......
y x y x

Figure 1.7: The graphs G and H in Exercise 35

36. (a) The minimum number of 2-switches is 1. Label the vertices of G1 as

shown in Figure 1.8(a). Deleting the edges v1 v2 and v5 v6 and adding the
edges v1 v6 and v2 v5 produces a graph G01 that is isomorphic to H1 .
(b) The minimum number of 2-switches is 2. Label the vertices of G2 as
shown in Figure 1.8(b). Since G2 contains three pairwise disjoint trian-
gles and H2 is bipartite, at least two 2-switches are required to transform
G2 into H2 . Consider the 2-switch in which the edges v2 w2 and u3 w3
are deleted from G2 and the edges u3 w2 and v2 w3 are added giving the
graph G02 .
 9

v1 ..........
v 1
....... ........... ......................
.............. ....... ....
..... ......... ..... ....
..... ..... ..... ...
...... ..... ...
. .
.....
.

v8 ..............
.
....
. .......
.......... ... ....
........... v
.......................2
v2 .....
... v 8 .... .......
......
.....
... ...
... ...
G1 : v7 .............. ..............
v3 G01 : ...
...
v
... 7 .......
....... ..............
v 3
...
...
...
..... ...
..... .
....... ..... ..
v6 .............. .............. ..... .. . . .
.......... . ..
.....
..... .....
. v4 .....
...... v 4 ....
.....
..... .
..................
.
..
...... v6 ....
..
.....
.
.
....
.
...
.... ....
........... ...................... ......
................................
v5 (a) v5
v1 u 2 v1 u 2 v 1 u2
........ ................. .......... ................. ........ ................
.......... ... .... ......... ..... ......... .....
.
..
...... ... ...... ..
.....
. .....
..... .
..
...... .....
.....
.. ... ..... .. ...
..... ..... ..... ..... ...
. .....
..... .. . .... w w .. ..
............. w w ......
u .........
1 ......
...............
..... 1w ..... .......
.............. ..............
v2 u1
.............................................
. .....
.....
1 2
. . ...............
... ....
.............
.
..
v2 u1 ....
...
.......... 1
..
.
.
.. .. . .....
. .. .
2 ..
.. ....
. . .
.......
.
.
.
. . ...........
... v2
. . . . ..
.....
..... .........
.. .
w .....
..... .........
.. ...
...
.
..... .
..... .... .......... .........
. ...
. .....
........... 2 .......... ... ..
... .. . ....... ..
............................................
... .....
........
....... .......
G : x.
... G : 0 x . ...
... .....
...
... .......
..
2 ....
.
2
. . . .. .. ..........
...... G002 : .....
..... x .
........... ........... ...
.. . ....... ....
...........
w w ...... ....... . ... ... ...
............ 3 3................. ..... w 3................. ...
..... ........ ..... ...
.
...
..... .......... ..... .....
.
................ ...... .............. ............... ............... ..............
v3 u3 v3 u3 v3 u3

(b)

Figure 1.8: The graphs in Exercise 36

Next consider the 2-switch in which the edges u1 w1 and xw3 are deleted
from G02 and the edges w1 w3 and u1 x are added. This produces the graph
G002 . Since G002 ∼
= H2 , a minimum number of two 2-switches is required to
transform G2 to H2 .
37. Gs = {C9 , C4 + C5 , C3 + C6 , 3C3 } and G = P4 = (3C3 , C3 + C6 , C9 , C4 + C5 ).
38. The three graphs G1 , G2 , G3 in Figure 1.9 all have the degree sequence

s : 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3

and form a triangle in G.

39. Proof. Suppose that this statement is false. Thus, for each graph H with
V (H) = {v1 , v2 , . . . , vn } where deg vi = di for 1 ≤ i ≤ n, there is a vertex vk
(1 ≤ k ≤ n) such that both (1) and (2) fail. Among all such graphs H and
all such vertices vk , let G be one with a vertex vk for which NG (vk ) has a
maximum number of vertices in common with either W1 = {v1 , v2 , . . . , vdk } if
k > dk or with W2 = {v1 , v2 , . . . , vdk +1 } − {vk } if 1 ≤ k ≤ dk . We consider
two cases.
Case 1. k > dk . Thus, NG (vk ) 6= W1 and the vertex vk is adjacent to a vertex
vj ∈
/ W1 and is not adjacent to a vertex vi with 1 ≤ i ≤ dk and so di ≥ dj .
Consequently, there is a vertex v` such that vi v` ∈ E(G) and vj v` 6∈ E(G).
Replacing the edges vk vj and vi v` by vk vi and vj v` is a 2-switch in G that
produces a graph G1 with V (G1 ) = {v1 , v2 , . . . , vn } where degG1 vi = di for
10 CHAPTER 1. INTRODUCTION

..........
G1 ...... G2 G2
s s s s s s

s s st s s ....
st s s st
....
...
...
s us sv s s us ..
.... sv..
.....
s s us sz
.. ......
..............

w s sx w s ..
.
.... .
..... .....
.....
sx w s sx
..... ...
.....
..... ...
s sv
.
s sz sz s
. ...
s ..
..... .....
y y .... . y

...........
G1 ...... G3 G3
s s s s s s
s s st s s..... . st u s ss
... ...
... ...
... ...
s u s sv s s u s ...
... .....
... ...
...
sv s w s sz
......
......
w s sx w s .....
... ...
.. .....
.
...
sx y s sx
.
.. ...
... ...
s t s sv s
... ...
y s sz y s
... ...
sz

...........
G3 ...... G2 G2
s s s s s s

u s ss u s .....
ss s s sv
...
...
...
w s sz w s ..
.
.. .... sz s us sx
.. ...... ...
............ ....
........
sx
...
y s sx y s .....
..
..
.. ..
..... .... ....
... ...
..
w s sz
.....
. .
.
....
.
. .... ....
s.............t....... s sv s s t s .....
.....
..... ..
... sv s y s st s

Figure 1.9: The graphs Exercise 38

1 ≤ i ≤ n in which NG1 (vk ) has a greater number of vertices in common with

W1 . This is a contradiction.
Case 2. 1 ≤ k ≤ dk . Thus, NG (vk ) 6= W2 and so the vertex vk is adjacent
to a vertex vj ∈/ W2 and is not adjacent to a vertex vi with 1 ≤ i ≤ dk + 1
and so di ≥ dj . Consequently, there is a vertex v` such that vi v` ∈ E(G)
and vj v` 6∈ E(G). Replacing the edges vk vj and vi v` by vk vi and vj v` is a
2-switch in G that produces a graph G2 with V (G2 ) = {v1 , v2 , . . . , vn } where
degG2 vi = di for 1 ≤ i ≤ n in which NG2 (vk ) has a greater number of vertices
in common with W2 . This is a contradiction.

40. See Figure 1.10.

(a) G2 = G1 − xz + xy (see Figure 1.10).

(b) The degree sequence of G1 is 5, 4, 3, 3, 2, 2, 2, 1 and the degree sequence
of G3 is 4, 3, 3, 3, 3, 2, 2, 2. When an edge rotation of G1 takes place, one
 11

u ........
v ........ .............. ..............
.............. ........
.......... ................. ..... ................
..... ..... ..... ........ ....
......... .
..
...... ..... .....
..... ............... ........
.... ..... ..... ..... ............ ...
............... ..... ...... ....
t .
......
..
... .
.............. w .............. ........
............
.............
. .
.........
.................
....
... .
....... . ..
...
. ............ .... ....
G1 : .....
.......
..........
.........
G2 : .
.. .
....... ....
....... ......... G3 : .....
.... ....
....
..... .......... .. .
....... ......... ....
....
................. ................. .............. ......................
..
..............
.................. ....... .....
s .....
..... ................
......... ..... z .....
.....
... ........
... ........
........ ...
..
..... ......... ......... .
.. ... . ..
..... ................. ..
.. ..
.....
. ...
........
.........
................
.
............. . ................
....... ................ .............. ........ ..........

x y

Figure 1.10: The graphs in Exercise 40

vertex of G1 has its degree increased by 1 and another has its degree
decreased by 1. Hence, the degree sequence of G3 cannot be obtained.
(c) Proof. Let G and H be two nonempty, noncomplete graphs of the
same order n and same size m. We
may assume
that V (G) = V (H) =
{v1 , v2 , . . . , vn }. Furthermore, k2 ≤ m < k+1 2 for some integer k with
2 ≤ k ≤ n − 1. Let S = {v1 , v2 , . . . , vk } and let F be the graph of order n
and size m with V (F ) = {v1 , v2 , . . . , vn } such
that F [S] = Kk and vk+1
is adjacent to v1 , v2 , . . . , vj , where j = m − k2 < k. We claim that G can
be transformed into F by a sequence of edge rotations. Suppose that this
is not the case. Among all graphs into which G can be transformed by a
sequence of edge rotations, let F 0 be one for which F 0 [S] has maximum
size. We consider two cases.
Case 1. The size of F 0 [S] is less than k2 . Then F 0 contains two

nonadjacent vertices vi and vj with 1 ≤ i < j ≤ k. If either vi or vj

is adjacent to a vertex v` with ` > k, then a graph F 00 can be obtained
from F 0 by a single edge rotation so that vi vj ∈ E(F 00 ), producing a
contradiction. Otherwise, F 0 contains an edge vp vq where p, q > k. Since
vi vp ∈/ E(F 0 ), a graph F 00 containing vi vj can be obtained from F 0 by a
single edge rotation so that vi vj ∈ E(F 00 ), again a contradiction.

Case 2. The size of F 0 [S] is k2 . Among all graphs F 0 with F 0 [S] ∼

= Kk
into which G can be transformed by a sequence of edge rotations, let G0
be one for which vk+1 is adjacent
to a maximum number of the vertices
v1 , v2 , . . . , vj , where j = m − k2 . Thus vk+1 v` ∈ / E(G0 ) for some `
with 1 ≤ ` ≤ j. If vk+1 vp ∈ E(G0 ) where p > j, then a graph G00
containing vk+1 vp can be obtained from G0 by a single edge rotation, a
contradiction. Otherwise, vk+1 is isolated in G0 and G0 contains an edge
vp vq where p, q > k+1. Thus, vk+1 vp ∈ / E(G0 ) and G0 can be transformed
into a graph containing vk+1 vp by a single edge rotation which in turn
can be transformed into a graph containing vk+1 v` , again producing a
contradiction.
Thus, as claimed, G can be transformed into the graph F with the de-
sired properties by a sequence of edge rotations, as can H. Consequently,
F can be transformed into the graph H by a sequence of edge rotations,
implying that G can be transformed into the graph H by a sequence of
12 CHAPTER 1. INTRODUCTION

edge rotations.

41. (b), (c) and (e) are graphical.

42. For a vertex v of a graph G of order n, we have degG v + degG v = n − 1.
Thus, d1 , d2 , · · · , dn is a degree sequence of a graph G of order n if and only
if (n − 1) − d1 , (n − 1) − d2 , · · · , (n − 1) − dn is a degree sequence of G.
43. If there exists a graph G with degree sequence x, 1, 2, 3, 5, 5, then the order
of G is 6. Since there are at least two vertices of degree 5, it follows that
δ(G) ≥ 2 and so no vertex of G has degree 1.
44. Since every graph has an even number of odd vertices, x must be even. So
x = 0, 2, 4, 6 are the only possibilities. We cannot have x = 0 since not all
the degrees can be different. We now consider x = 2, 4, 6, and apply the
Havel-Hakimi theorem (Theorem 1.10).

x=2
s : 7, 6, 5, 4, 3, 2, 2, 1
s1 : 5, 4, 3, 2, 1, 1, 0
s2 : 3, 2, 1, 0, 0, 0
s3 : 1, 0, −1, 0, 0. Since s3 is not graphical, s is not graphical and so
x 6= 2.
x=4
s : 7, 6, 5, 4, 4, 3, 2, 1
s1 : 5, 4, 3, 3, 2, 1, 0
s2 : 3, 2, 2, 1, 0, 0
s3 : 1, 1, 0, 0, 0. Note that s3 is a degree sequence of G = K2 + 3K1 and
so s3 is graphical. Therefore, s is graphical and x = 4 is possible.
x=6
s : 7, 6, 6, 5, 4, 3, 2, 1
s1 : 5, 5, 4, 3, 2, 1, 0
s2 : 4, 3, 2, 1, 0, 0.
Ignoring isolated vertices, not all degrees can be different and so s2 is
not graphical. Therefore, s is not graphical and x 6= 6.

Therefore, the sequence is graphical for x = 4 only.

P3 P7
45. Since i=1 di = 17 and 3(3 − 1) + i=1 min{3, di } = 16, it follows from
Theorem 1.13 that the sequence s : 6, 6, 5, 4, 3, 2, 2 is not graphical.
46. First, there is no such graphical sequence where one term occurs 3 times since
no graph contains an odd number of odd vertices. Thus, the only possible
graphical sequences are those in which two terms occur twice each. Again,
since no graph contains an odd number of odd vertices, the sequence 4, 4, 3, 2, 2
is not graphical. On the other hand, both sequences 4, 4, 3, 3, 2 and 4, 3, 3, 2, 2
are graphical.
 13

47. Let S = {a1 , a2 , · · · , an } be a set of positive integers and let k = lcm {a1 +
1, a2 +1, · · · , an +1}. (Note that we could also let k = (a1 +1)(a2 +1) · · · (an +
1).) Let G be the union of the graphs aik+1 Kai +1 for 1 ≤ i ≤ n. Then G has
the desired property.
Let S = {2, 6, 7}. If k = 1, 3, then the graph with degree set S has an odd
number of odd vertices, which is impossible. If k = 2, then a degree sequence
of the graph G is 7, 7, 6, 6, 2, 2. This implies that G has order 6 but maximum
degree 7, which is impossible. The sequence

s : 7, 7, 7, 7, 6, 6, 6, 6, 2, 2, 2, 2

is graphical by Theorem 1.12.

48. The graph

  Fn constructed in Theorem 1.15 contains exactly two  vertices
  of de-
gree n2 . Since the degrees of the vertices of Fn are 1, 2, . . . , n2 , n2 , n2 +
1, . . . , n−1, the degrees of the vertices of F n are n−2, n−3, . . . , n2 −1, n2 −
   

1, . . . , 1, 0.

49. Proof. Assume first that s01 and s02 are bigraphical sequences. Then there
exists a bipartite graph G0 with partite sets V10 and V20 such that V10 =
{u2 , u3 , . . . , ur } and V20 = {w1 , w2 , . . . , wt }, where degG0 ui = ai for 2 ≤ i ≤ r
and 
bj − 1 if 1 ≤ j ≤ a1
degG0 wj =
bj if a1 + 1 ≤ j ≤ t.
A new bipartite graph G can be constructed from G0 by adding a new vertex
u1 and joining u1 to wi for 1 ≤ i ≤ a1 . Thus the sequences s1 and s2 are
bigraphical sequences.
For the converse, suppose that s1 and s2 are bigraphical sequences. Hence,
there exist bipartite graphs with partite sets V1 = {u1 , u2 , . . . , ur } and V2 =
{w1 , w2 , . . . , wt } such that deg ui = ai and deg wj = bj for 1 ≤ i ≤ r and
1 ≤ j ≤ t. If there exists such a bipartite graph H and a vertex u of H of
degree a1 in V1 adjacent to vertices of degrees b1 , b2 , . . . , ba1 , then s01 and s02
are bigraphical sequences for H − u. Hence, we may assume that no bipartite
graph with bigraphical sequences s1 and s2 has this property. Among all such
graphs, let G be one having a vertex u1 of degree a1 in V1 such that the sum
of the degrees of the neighbors of u1 is maximum. Hence there exist integers
i and j with 1 ≤ i < j ≤ t such that bi > bj and u1 is not adjacent to wi but
is adjacent to wj . Because deg wi > deg wj , there is a vertex uk ∈ V1 with
deg uk = ak such that wi is adjacent to uk but wj is not adjacent to uk . Let

F = (G − {u1 wj , uk wi }) + {u1 wi , uk wj }.

Thus, the bipartite graph F also has bigraphical sequences s1 and s2 but the
sum of the degrees of the neighbors of u1 in F is greater than the sum of
degrees of the neighbors of u1 in G. This is a contradiction.
14 CHAPTER 1. INTRODUCTION

50. These are the two nearly irregular graphs of order 10 and each is the comple-
ment of the other.

51. (a) Let n = r + s, where s ≥ 0 is even. Then G = rK1 + (s/2)K2 has the
desired properties.
(b) Since every even vertex has degree at least 0 and every odd vertex has
degree at least 1, the minimum size of such a graph is at least s/2. Since
the graph rK1 + (s/2)K2 has this size, the minimum size is s/2.
(c) Suppose first that n is even. Then n − 1 is odd. So, every odd vertex
has degree at most n − 1 and every even vertex

has degree at most n − 2.
Hence the maximum size of G is at most n2 − r/2. Since r and s are
both even, the graph (r/2)K2 + sK1 has size n2 − r/2, which is the

maximum size. Suppose next that n is odd. Then n − 1 is even. Every

even vertex has degree at most n − 1 and every odd vertex
has degree at
most n − 2. Hence, the maximum size of G is at most n2 − s/2. Since
(s/2)K2 + rK1 has size n2 − s/2, this is the maximum size.

52. (a) Proof. Let uv ∈ E(G). Let W00 be the set of vertices in V (G) − {u, v}
adjacent to neither u nor v, let W10 be the set of vertices in V (G)−{u, v}
adjacent to u but not v, let W01 be the set of vertices in V (G) − {u, v}
adjacent to v but not u and let W11 be the set of vertices in V (G)−{u, v}
adjacent to both u and v. For i, j ∈ {0, 1}, let |Wij | = aij . Then
a00 + a10 + a01 + a11 = n − 2, a10 + a11 ≥ (2n + 1)/3 and a01 + a11 ≥
(2n + 1)/3. Therefore,

4n + 2
a01 + a10 + 2a11 ≥
3
and so
4n + 2
(a01 + a10 + a11 ) + a11 = (n − 2 − a00 ) + a11 ≥ .
3
Thus,
4n + 2 n+8 n+8
a11 ≥ − (n − 2) + a00 = + a00 ≥ .
3 3 3
Since n ≥ 4, W11 contains at least four vertices. If any two vertices in
W11 are adjacent, then u and v belong to a complete subgraph of order 4.
Suppose, on the other hand, that no two vertices of W11 are adjacent.
Then for each x ∈ W11 ,
n+8 2n − 8
deg x ≤ n − = ,
3 3
contradicting the assumption that deg x ≥ (2n + 1)/3 for every vertex v
of G.
 15

(b) Since every vertex in Kr,r,r has degree 2r = 2n/3 and Kr,r,r contains no
complete subgraph of order 4, the result in (a) is best possible.

Section 1.8. Multigraphs

53. Only the sequence in (c) is a degree sequence of an irregular multigraph (see
Figure 1.11). There is no multigraph containing an odd number of odd ver-
tices. So, (a) and (b) are not the degree sequences of any multigraph.
.........................
....... .
........................... ....................................... ................................................... ................................ ..................................
......... ........... ........... ............... ............ ...........
.
.............. ................. ........... .. ..
. ........ ......... .........
.......... .
....... .
............................ ........................................
1 3 5 7 6 4 2

Figure 1.11: The graphs Exercise 53

54. The graphs of order 3 and 4 are shown in Figure 1.7 in the text. See Fig-
ure 1.12.
3
..
.......
s ..
............
s
...... ... ...
s s s ...........
...... s s ........
...........
.....................
...
............................
s ... ...
... ..... ...........
...... .
..... ... ....
... ..... ...
...
..
. ... ... .... ...
1 3 .. ... ... ....
2 s s
... ... ...
... .... ..... ............................... ....
s s
.. ...........
................... ....
........ .
..........................................
............................ ............................... 4
s s s s ...........
...... s s......
......... ..
. s .
..............................................................
s 5
1 4 5 2 4
3
s s
s 1
s s s .....
.
.......
s
..... s ... ...
... ...
... ....
... ....
s
.
.............. s
.....
..... .....
.....
3 ........... .... .... ..... ...
.....
..... .....
.....
2 .................
... ... ... .
........ ........ ...... ... ... ... ..
..... ..
.. ........... ..... ........... ...... ....... .. ... .. ... ....
..... ..... ...... ..... ..... ........ ............ ......... ... ... . .
..... ..........
s s s ... ... ................................... ... ... . .
s
...... .
........ ........ ...................... .............
............................... s .
................................................................
......................
....... ......................................
... s
............................ 7
6 6
s
..... ....
s 5 s ..........
.....................................................
.....................
..
.................
s6
....... ..... ....
3 ......s....................................................................................................s... 9
.....
..... ... ...... .........
.....
..... ...... ...... ... ..
.....
..... .........
s
... ....... ......
.... ........ ...... ................... ........ ..... s s
.........
. ...........
......
s
...... ....... .. ........ ...... ...
......... .......................... ......
.................................. .....
.....
.
..... ... ....
... ...
.. .. .... .
....... ...... ............ ..
...... ...... ....... .... ...
..... ..... ........ .... ..
..... ... .. .... ..... ..... ... .... ..
8 ..
..
. ........... .... .... ....... ......... ................ ..... .... .....
..... ...... . . . .. .. . . . ..
s s s
.....
.....
.... .... ..... ........ .................. ..... ....
... ...
.
..... s ... .. .. .... ...... ...... .....
... ... ... .... ....... ...... .......
... ...... ... ...... ....... ......
... ...
... ..
... .... ......... ....... ........ ... ...
1 .......................... ......... ... ..
s ..........................
s... ...
............................................................................................
11 ...............
...... ..................................... ............... . . ..
7
s9
.
7 .........s.........................................................
....................................................................................... ........ ...
s
........ s.
........
......................... .........
...................
.......... .........................
....... .............................
...... ........ ....... ..... .........
.....................................
..........
... ....... ....... ... ......... ............................... ........... ..... .................................. .................
... ....... .............. ..... 15
s ... ... ....... .......... .......... ....... ...
.. . ..
... ......... .. ... ... ....... ......... ....... ..... ............................ ............ ........ ....... ....
..
s
... ... ... ....... .......... ......... .............. ......... .. .. . .
... ... .......... ... ... ......... .......................................... ....... .... ... ....
... ... ...... .
. .. . . .. ...
....................................................... . ....
. .
... ... ... ... ... ... . .. .. ..
... ..... ... ... ....................... .. .. ..
s
.....
. ... ...
... ...
.
..
. .
.... ... ....... ....
. .. . . .
..... .... ....
. .
... ... ... .... ... ... ... ... ... ... ...
... ... ... ... ... ... .... .... ... ... ..
... ... .
. . .
. . . . .... .... .....
... ..... ... ... .. .. .. .. . . . .
. . . .
..... .... ... ... ... .. .. .. ..... .... ....
..... ..... ..... ..... ....
..... ..... ........ .... .... .... .... ............................
...... ..... ..... ... .. .. ... ......... ......
....... ...... ..... ........ ............. .......
s
.......................................................
...................... .......................
.........
11

Figure 1.12: Graphs and multigraphs in Exercise 54

55. Each of the sequences in (a), (c), (e), (g) and (h) is the degree sequence of a
multigraph. See Figure 1.13.
16 CHAPTER 1. INTRODUCTION

s s . .. s .......
..........
............. ....................
s s
... .... ...... .
... ...
... .... ..... ..... ... ..
..... ... .....
... ...
... ...
..
s s ......
... ....
...
...
...
... .
..
.... ..
...
s
.. ....
s ... ...
..............
..... . .. .....
... ...
... ....
...
...
...
... .....
s ... ..
.... .
.
..
.. s
................ ......
...... ....... ...... .....
... ... ... ... .... ..... ..... .... ... ... .... ... .... ...... ... ...... ......
... .... ... ... .. .. ... .. ... .... .... ... ... ...... .... .... .... .....
.... . .... .... ... ... ... .. ... ... .. ..... .... .... ..... .... ..... ....
... .... ... ... ... ... ... ... ... .. ... ... ... .. ... ... ... ...
... ... ... .... ... ... .. ....
s
... ... . ...... ...... .......
s
... ...
. s
...... .. . ....
.............. s
..............
...
.......................................
.......................................... s . s
....... ......
....... ........
.. s......................
......
.
...........................................

(a) (c) (e) (g) (h)

Figure 1.13: Multigraphs in Exercise 55

Since the sum of the degrees of the vertices of a multigraph G is twice the
number of edges of G, neither s4 nor s6 is the degree sequence of a multi-
graph. The sequence s2 is also not the degree sequence of any multigraph G;
otherwise, G has (1 + 2 + 5)/2 = 4 edges and 5 edges incident with a vertex
of G, which is impossible.

56. Proof. First, let G be a multigraph of order n and size m havingPdegree

n
sequence s : d1 , d2 , . . . , dn , where d1 ≥ d2 ≥ · ·P
· ≥ dn . Then 2m = i=1 di
1 n
is even. Suppose, to the contrary, that d1 > 2 i=1 di = m. Then there is a
vertex of G incident with more edges than those in G. This is impossible.
We verify the converse by induction on n. Suppose that s : d1 is a sequence
of one nonnegative integer such that d1 is even and d1 ≤ 21 d1 . Then d1 = 0,
so s is the degree sequence of the only multigraph K1 of order n = 1. For
n = 2, supposeP2that s : d1 , d2 is a sequence
P2 of a nonnegative integers such
that d1 ≥ d2 , i=1 di is even and d1 ≤ 12 i=1 di . This implies that d1 = d2 .
The multigraph of order 2 whose two vertices are joined by d1 parallel edges
has degree sequence s. Hence, the statement is true for n = 2.
Assume for an integer k ≥ 2 that if s1 : e1 , e2 , . . . , eP k is an sequence of
k
k nonnegative integers such that e1 ≥ e2 ≥ · · · ≥ ek , i=1 ei is even and
1
P k
e1 ≤ 2 i=1 ei , then s1 is a degree sequence of a multigraph of order k. Let
s : d1 , d2 , . . . , dk+1 be a sequence of k + 1 nonnegative integers such that
Pk+1 Pk+1
d1 ≥ d2 ≥ · · · ≥ dk+1 , i=1 di is even and d1 ≤ 21 i=1 di . Since d1 ≥ d2
Pk+1
and d1 ≤ i=2 di , the terms d2 , d3 , . . . , dk+1 can be decreased by a total of
d1 resulting in a sequence d02 , d03 , . . . , d0k+1 of nonnegative integers such that
Pk+1 0 Pk+1
di = i=2 di −d1 ≥ 0, d02 ≥ d03 ≥ · · · ≥ d0k+1 where d02 −d03 ≤ 1. Because
Pi=2
k+1 Pk+1 Pk+1 0 0
Pk+1 0
i=1 di is even, i=2 di −d1 = i=2 di is even. Consequently, d2 ≤ i=3 di
k+1
and so d02 ≤ 21 i=2 d0i . By the induction hypothesis, s0 : d02 , d03 , . . . , d0k+1 is
P
the degree sequence of a multigraph G0 of order n. We may assume that
V (G0 ) = {v2 , v3 , . . . , vk+1 } and degG0 vi = d0i for 2 ≤ i ≤ k + 1. Let G be
the multigraph obtained from G0 by adding a new vertex v1 to G0 and joining
v1 to the vertices of G0 so that the degrees of V (G0 ) become d2 , d3 , . . . , dk+1 .
Then degG v1 = d1 and so s : d1 , d2 , . . . , dk+1 is the degree sequence of G.

57. (a) m = 4 · 1 + 3 · 2 + 2 · 3 + 1 · 4 = 20.

 17

(b) At most two edges can be replaced by one edge each and at most two
edges can be replaced by two parallel edges. Thus, the size of H is at
least 2 · 1 + 2 · 2 + 3 = 9. This is possible (see the graph H in Figure 1.14).
At least two edges can be replaced by one edge each and at most one
edge can be replaced by four parallel edges and, if so, one edge can be
replaced by three parallel edges. Thus, the size is at most 2·1+4+3+2 =
11. The graph F in Figure 1.14 shows that this can occur.

2
......
4
... . ........
....................... ....................
...... ...... ......
. . .
...... ......
...... .. ......... ......
...... ...... ..... ......
..... 2 2 ...... ...... 3 4 ......
. .
.. ......... .............. ...........
4
.
.............
.
.... ...
5 3 ...... ........
. 5
H: .....
...
... ..
.
..... F : ...
...
... ..
..
..
.. .
...
... 1 3 ...
. ...
...2 1 ...
..
...
... ..
.. ... ...
.
............. 1 ...
...............
..
.............
...
1 ...
.............
...
.... .
1 3 2 1

m=9 m = 11

Figure 1.14: Graphs H and F in Exercise 57(b)

(c) Let U and W be the partite sets of G, where U = {u1 , u2 , . . . , us } and

W = {w1 , w2 , . . . , wt } such that deg u1 ≥ deg u2 ≥ · · · ≥ deg us and
deg w1 ≥ deg w2 ≥ · · · ≥ deg wt . Then the minimum value of m is
( s t
)
X X
min i deg ui , i deg wi .
i=1 i=1
18 CHAPTER 1. INTRODUCTION
Chapter 2

Connected Graphs and

Distance

Section 2.1. Connected Graphs

1. Let G = K2,3 and let u and v be the two vertices of G with deg u = deg v = 3.

2. Consider the path P3 = (v, u, w) of order 3.

3. (a) The statement is true. Proof. Suppose that A1 = A2 . Then A1 and A2

are both n × n matrices for some positive integer n. Hence the orders
of G1 and G2 are n. Let u1 , u2 , . . . , un be a labeling of the vertices
of G1 that gives the adjacency matrix A1 and let v1 , v2 , . . . , vn be a
labeling of the vertices of G2 that gives the adjacency matrix A2 . Define
f : V (G1 ) → V (G2 ) by f (ui ) = vi for 1 ≤ i ≤ n. Since A1 = A2 , two
vertices ui and uj are adjacent in G1 if and only if vi and vj are adjacent
in G2 . Hence G1 ∼= G2 .
(b) The statement is false. Let G2 = (u1 , u2 , u3 ) be a path of order 3
and G3 = (v1 , v3 , v2 ) be ∼
  a path of
 order 3. Then G2 = G3 but A2 =
0 1 0 0 0 1
1 0 1 and A3 = 0 0 1 and so A2 6= A3 .
0 1 0 1 1 0

4. (a) The matrix A is the adjacency matrix of the graph K2 with V (K2 ) =
{v1 , v2 }. Since there is one v1 − v1 walk of length 4, one v2 − v2 walk of
length4, and no v1 − v2 walks or v2 − v1 walks of length 4, the matrix
1 0
A4 is .
0 1
(b) Since deg v1 = deg v2 = 1 and there exist
 no v1 − v2 paths or v2 − v1
2 1 0
paths of length 2, it follows that A = . Since A2 is the identity
0 1
matrix I, it follows that A4 = A2 · A2 = I · I = I.

19
20 CHAPTER 2. CONNECTED GRAPHS AND DISTANCE
 
0 1 1 1
 1 0 1 1  2
5. The adjacency matrix of G1 is A =  1 1 0 1 . To compute A , observe


0 1 1 0
that deg v1 = deg v4 = 2 and deg v2 = deg v3 = 3. For i 6= j, the (i, j)-
entry of A2 is the

2
 number of different vi − vj paths of length 2. Thus A =
2 1 1 2
 1 3 2 1  3
 1 2 3 1 . To compute A , observe that v1 and v4 belong to one
 

2 1 1 2
triangle and v2 and v3 belong to two triangles each. For i 6= j, the (i, j)-
entry of A3 is the

3
 number of different vi − vj walks of length 3. Thus A =
2 5 5 2
 5 4 5 5 
 5 5 4 5 .
 

2 5 5 2
 
0 1 0 0 0
 1 0 1 0 0 
 
6. The adjacency matrix of the graph G2 is A =   0 1 0 1 0 . The

 0 0 1 0 1 
0 0 0 1 0
k
(i, j)-entry of A is the number of different vi − vj walks of length k in G.
Thus
   
1 0 1 0 0 0 2 0 1 0
 0 2 0 1 0 
 3  2 0 3 0 1 
 
2

A =  1 0 2 0 1 , A =  0 3 0 3 0 
  
 and
 0 1 0 2 0   1 0 3 0 2 
0 0 1 0 1 0 1 0 2 0
 
2 0 3 0 1
 0 5 0 4 0 
 
A4 =   3 0 6 0 3 .

 0 4 0 5 0 
1 0 3 0 2

7. V (G) = {v1 , v2 , v3 , v4 , v5 } and E(G) = {v1 v2 , v1 v3 , v2 v3 , v3 v4 , v4 v5 }.

8. Since there is one vi − vi walk of length 0 and no vi − vj walks of length 0
for i 6= j, the (i, j)-entry of A0 gives the number of vi − vj walks of length 0.
Therefore, Theorem 2.2 also holds for k = 0.
9. Let Pk+1 = (u1 , u2 , . . . , uk+1 ) be a path of order k + 1 and let G be the graph
obtained from Pk+1 by adding k new vertices v1 , v2 , . . . , vk and joining vi to
ui and ui+1 for 1 ≤ i ≤ k. Then G has the desired property.
[Note: Also, consider G = K1 ∧ kK2 .]
 21

10. (a) See Figure 2.1.

s ..........
............. ....................
s .....
.......
s
......
.......................................................................
.....
. ...
... ...
s
.
s
...
... ..
............................................................................................
... ...
.... ..
... ...
s ...
...
... .s
..
.
..
... ...
... ...
s .....
.....
......
......
s.
...
..
..
.....
........
s
.......................................

Figure 2.1: The graph in Exercise 10

(b) Let G = K3,3 . Then g(G) = 4 and c(G) = 6 but G contains no 5-cycle.
11. (a), (b). Let P = (u1 , u2 , . . . , u` ) be a longest path in G. Thus, u1 is only
adjacent to vertices of P , for otherwise, there is a path longer than P . Let
k be the greatest integer, 2 ≤ k ≤ `, such that u1 uk ∈ E(G). Therefore,
C = (u1 , u2 , . . . , uk , u1 ) is a k-cycle, where k ≥ δ(G) + 1.
12. Note that g(G) = 3 and c(G) = 6k.
13. Proof. Let G be a graph. Since (v) is a trivial path for every vertex v
of G, the vertex v is connected to itself. Suppose that u is connected to
v. Then G contains a u − v path P , say P = (u = u1 , u2 , . . . , uk = v).
Then (v = uk , uk−1 , . . . , u2 , u1 = u) is a v − u path and v is connected to
u. Next, suppose that u is connected to v and v is connected to w. Then
G contains a u − v path P = (u = u1 , u2 , . . . , uk = v) and a v − w path
Q = (v = v1 , v2 , . . . , u` = w). Then

(u = u1 , u2 , . . . , uk = v = v1 , v2 , . . . , u` = w)

is a u − w walk. By Theorem 2.1, G contains a u − w path.

14. Proof. Suppose that there is a partition {V1 , V2 } of V (G) such that no edge
joins a vertex of V1 and a vertex of V2 . Let x ∈ V1 and y ∈ V2 . Then x and y
are not connected in G and so G is not connected.
For the converse, suppose that for every partition {V1 , V2 } of V (G), there
exists an edge of G joining a vertex of V1 and a vertex of V2 . We claim that
G is connected, for otherwise, there exist two distinct vertices x and y in G
that are not connected. Let U1 be the set of vertices that are connected to x
and let U2 = V (G) − U1 . Then there is no edge joining a vertex of U1 and a
vertex of U2 .
15. Proof. Assume, to the contrary, that the statement is not true. Then there
exists a connected graph G of order n and an integer k with 2 ≤ k ≤ n − 1
such that deg u + deg v ≥ k for every pair u, v of nonadjacent vertices of G
but G contains no path of length k. Let P be a longest path in G. Suppose
that P is a u − v path of length `. Then 2 ≤ ` < k. We consider two cases.
22 CHAPTER 2. CONNECTED GRAPHS AND DISTANCE

Case 1. uv ∈ E(G). Then G has a cycle C of length ` + 1 < n. Since G is

connected, there is a vertex w ∈ V (G) − V (C) such that w is adjacent to some
vertex x on C. Then G has a path of length ` + 1, which is a contradiction.

Case 2. uv ∈
/ E(G). By hypothesis, deg u + deg v ≥ k. Suppose that

P = (u = u0 , u1 , · · · , u` = v).

Since P is a longest path in G, every vertex adjacent to u (and to v) belongs

to P . If u is adjacent to ui (1 ≤ i ≤ `), then v is not adjacent to ui−1 , for
otherwise, G contains a cycle of length ` + 1 and we can proceed as in Case 1.
Thus deg v ≤ ` − deg u. However then

deg u + deg v ≤ ` < k,

which is a contradiction.
16. Proof. Let G be a disconnected graph of order n ≥ 6 having three compo-
nents. We show that there exists some vertex v ∈ V (G) such that degG v ≤
(n − 3)/3.
Let G1 , G2 and G3 be the three components of G with |V (Gi )| = ni ≥ 1 for
1 ≤ i ≤ 3. Suppose that n1 ≤ n2 ≤ n3 . Then n1 ≤ n/3. If v ∈ V (G1 ), then
n n−3
degG v = degG1 v ≤ n1 − 1 ≤ −1= .
3 3
Since δ(G) ≤ (n − 3)/3, it follows that ∆(G) ≥ (n − 1) − (n − 3)/3 = 2n/3.
17. Let V (G) = {u = v1 , v2 , · · · , vn = v}. Since G is connected, it follows that
G contains a vi − vi+1 path Pi for i = 1, 2, · · · , n − 1. Proceeding along the
paths P1 , P2 , . . . , Pn−1 in the given order produces a u − v walk containing all
vertices of G.
18. A graph G of order n has the property that every induced subgraph of G is
connected if and only if G = Kn .
Proof. First, if G = Kn , then every induced subgraph of G is complete
and therefore connected. For the converse, suppose that G is not complete.
Then G contains two nonadjacent vertices u and v. Let S = {u, v}. Then
G[S] = K 2 is disconnected.
19. (a) Proof. Let the distinct degrees be a, b and c. Let deg x = a and deg y =
b. Then there exists an x − y path P 0 . If there is a vertex of degree c on
P 0 , then the proof is complete. Otherwise, there are adjacent vertices u
and v with deg u = a and deg v = b. Let P = (u, v) be this subpath of
P 0 . Now, let z be a vertex with deg z = c and let Q be a shortest path
from z to P . This produces a path containing vertices of degrees a, b and
c.
(b) No. Consider K1 ∨ (K1 + K2 + K3 ).
 23

20. Proof. If every two vertices of G are of the same parity, then the result
follows immediately. If this is not the case, then let V1 = {v ∈ V (G) :
v is odd} and V2 = {v ∈ V (G) : v is even}. Thus {V1 , V2 } is a partition of
V (G). Since G is not bipartite, V1 and V2 cannot be partite sets for G. So
either two vertices of V1 are adjacent or two vertices of V2 are adjacent. In
either case, G contains two adjacent vertices whose degree sum is even.

21. Proof. Assume, to the contrary, that G has k (or more) components. Then
G has a component G1 of order at most n/k. Each vertex of G1 has degree
at most (n/k) − 1 = (n − k)/k, which is a contradiction.

22. Proof. Let M = (k − 1)(n − k − 1) + k+1

2 . Assume, to the contrary, that
there is a graph G of order n ≥ k + 1 and size m ≥ M , no subgraph of which
has minimum degree k. Thus δ(F ) ≤ k − 1 for every subgraph F of G. In
particular, δ(G) ≤ k − 1. Let v1 ∈ V (G) such that degG v1 ≤ k − 1. Consider
G1 = G−v1 . The size of G1 is at least M −(k−1). Since δ(G1 ) ≤ k−1, there is
a vertex v2 ∈ G1 such that degG1 (v2 ) ≤ k−1. Let G2 = G1 −v2 = G−{v1 , v2 }.
The size of G2 is at least M −2·(k−1). In general, let Gi = G−{v1 , v2 , . . . , vi }
such that degGi−1 (vi ) ≤ k − 1 for 1 ≤ i ≤ n − k − 1, where G0 = G. Thus the
size of of Gi is at least M − i(k −
1). In particular, the size of Gn−k−1 is at
least M − (k − 1)(n − k − 1) = k+1 2 . Since Gn−k−1 has order k + 1, it follows
Gn−k−1 = Kk+1 , which has minimum degree k, producing a contradiction.

23. Proof. Select some vertex v1 in G. Since G is connected, there is a vertex

v2 in G that is adjacent to v1 . If n ≥ 3, then the subgraph induced by
S2 = {v1 , v2 } is not a component of G. Thus there is a vertex v3 ∈ / S2 that is
adjacent to some vertex in S2 . Continuing in this way, we obtain a sequence
v1 , v2 , . . . , vn where each vertex vi (2 ≤ i ≤ n) is adjacent to some vertex in
the set {v1 , v2 , . . . , vi−1 }.

24. Proof. We show, by induction, for each integer p with 2 ≤ p ≤ n that the
subgraph G[{v1 , v2 , . . . , vp }] is connected. Since v1 v2 ∈ E(G), the subgraph
G[{v1 , v2 }] is connected. Assume for an integer k with 2 ≤ k < n that
H = G[{v1 , v2 , . . . , vk }] is connected. Suppose that vk+1 is adjacent to v` ∈
{v1 , v2 , . . . , vk }. Since there is a v` − vi path in H for every i with 1 ≤ i ≤ k,
there is a vk+1 − vi path in H 0 = G[{v1 , v2 , . . . , vk+1 }] for every i with 1 ≤
i ≤ k. Furthermore, since there is a vi − vj path in H for every two vertices
vi , vj ∈ V (H), there is a vi − vj path in H 0 . Thus, H 0 is connected. By
the Principle of Mathematical Induction, the subgraph G[{v1 , v2 , . . . , vp }] is
connected for every integer p with 2 ≤ p ≤ n. In particular, G is connected.

Section 2.2. Distance in Graphs

25. Proof. Observe that
d(u, v) + d(u, w) + d(v, w) = [d(u, v) + d(v, w)] + d(u, w) ≥ 2d(u, w)
by the triangle inequality.
24 CHAPTER 2. CONNECTED GRAPHS AND DISTANCE

26. Proof. If G is a bipartite graph, then G contains no odd cycle by Theo-

rem 1.18 and so G contains no induced odd cycle. For the converse, assume
that G contains no induced odd cycle. We show that G does not contain any
odd cycle (which implies that G is bipartite by Theorem 1.18). If this is not
the case, among all odd cycles of G, let C be one of minimum size. If C is not
induced, then there is a chord in C. This, however, produces an odd cycle of
smaller size in G, which is impossible.

27. Since m is an integer, n is even and so n = 2k for some integer k ≥ 3. We

claim that G = Kk−1,k+1 , which has order n and size m = (n−2)(n+2)
4 = k 2 −1.

If G is a graph of order n ≥ 6 and size m = (n−2)(n+2)

4 containing no odd cycle,
then G is a bipartite graph with partite sets U and W where say |U | ≤ |W |. It
is known (by Theorem 1.8) that the size of G is at most k 2 . If |U | = |W | = k,
then G = Kk,k − e has size k 2 − 1, but then only two vertices of G have degree
less than k. Thus, |U | < |W |. Then |U | = k − a and |W | = k + a for some
positive integer a. If a ≥ 2, then the size of G is at most k 2 − a2 < k 2 − 1,
which does not occur and so a = 1. Thus, G is a subgraph of Kk−1,k+1 . Since
the size of Kk−1,k+1 is k 2 − 1, it follows that G = Kk−1,k+1 .

28. (a) Proof. Assume, to the contrary, that there exists a connected graph
G containing two paths P 0 and P 00 of maximum length ` that have no
vertex in common. Suppose that P 0 is u0 − v 0 path and P 00 is a u00 − v 00
path. Let

s = min{d(w0 , w00 ) : w0 ∈ V (P 0 ), w00 ∈ V (P 00 )}.

Thus s ≥ 1. Let x ∈ V (P 0 ) and y ∈ V (P 00 ) such that d(x, y) = s and let

P be an x − y path of length s. Let Q0 be the u0 − x subpath or x − v 0
subpath of P 0 , whichever has length at least d`/2e. Similarly, let Q00 be
either the u00 − y subpath or the y − v 00 subpath of P 00 , whichever has
length at least d`/2e. Then Q0 , P, Q00 is a path of length more than `,
which is a contradiction.
(b) Proof. Suppose that ` is odd, say ` = 2k + 1 for some positive integer
k. Suppose that P is a u − v path and Q an x − y path such that
V (P ) ∩ V (Q) = {w}. Then either the u − w subpath or the w − v
subpath of P has length at least k + 1, say the w − v path P 0 has length
at least k + 1. Similarly, we may assume that the x − w subpath Q0 of Q
has length at least k + 1. Then P 0 and Q0 form a path of length at least
2k + 2 > 2k + 1 = `, a contradiction.

29. (a) d1 (v) = deg v.

(b) For 1 ≤ k ≤ n − 1, let mk denote
P the number of pairs u, v of vertices of
G such that d(u, v) = k. Then v∈V (G) dk (v) = 2mk .
P  P
P n−1 2
(c) v∈V (G) k=1 dk (v) = v∈V (G) (n − 1) = n(n − 1) = n − n.
 25

30. Proof. Let u and v be two vertices of G. We show that u and v are connected
and dG (u, v) ≤ 2. If u and v are in different components of G, then uv is
an edge in G and so dG (u, v) = 1. Suppose that u and v are in the same
component G1 of G. Let w be a vertex that is in another component G2 of G.
Then (u, w, v) is a path in G. Thus dG (u, v) ≤ 2. Therefore diam(G) ≤ 2.

31. First, let a = 1. For b = 1, let G = P2 ; while for b = 2, let G = P3 . For a ≥ 2,

let G = C2a if b = a; while for b > a, let G be obtained from C2a and Pb−a+1
by identifying an end-vertex of Pb−a+1 with a vertex of C2a .

32. (a) Assume, to the contrary, that 2, 3, 3, 3 is the eccentricity sequence of

some graph G (necessarily of order 4). Let u be a vertex with eccentricity 3
and suppose that v is a vertex of G such that e(u) = d(u, v) = 3. Let
P = (u, w, x, v) be a u − v geodesic. Then neither w nor x has eccentricity 3,
a contradiction.
(b) Since a graph with eccentricity sequence a, b, b, b has order 4, it follows
that either (i) a = 1 and b = 2 or (ii) a = 2 and b = 3. The eccentricity
sequence of the graph K1,3 is 1, 2, 2, 2; while 2, 3, 3, 3 is not the eccentricity
sequence of any graph by (a). Thus a = 1 and b = 2.

33. For n ≥ 4, let e1 , e2 and e3 be the edges of a triangle in Kn . Let F =

Kn − e1 − e2 and G = Kn − e1 − e2 − e3 . Then F 6∼= G but the eccentricity
sequence of both is 1, 1, . . . , 1, 2, 2, 2.

34. (a) K2 . (b) P3 .

35. Let H = K2,k where the partite sets of H are U = {u1 , u2 } and W =
{w1 , w2 , . . . , wk } and let G be the graph obtained from H by adding two
vertices v1 and v2 and the two edges u1 v1 and u2 v2 . Then deg wi = e(wi ) = 2
for i = 1, 2, . . . , k.

36. Proof. Let w be a vertex with e(w) = k and let u be a vertex with d(w, u) =
e(w) = k. For a central vertex v of G, let P be a u − v path of length
d(u, v). Thus d(u, v) ≤ e(v) = rad(G). Since d(u, w) = k, it follows that
e(v) < k ≤ e(u). By Theorem 2.7, there is a vertex x on P such that e(x) = k.
Because d(u, x) ≤ d(u, v) < k and d(w, u) = k, it follows that x 6= w.
Ps−1
37. Proof. Let n = 2 + (r − 1) i=1 (r − 2)i and suppose that G is a connected
graph of order n such that ∆(G) < r and diam(G) < s. Let x ∈ V (G) such
that e(x) = diam(G) < s. Let Ai (x) denote the set of vertices at distance i
from x where 0 ≤ i ≤ s − 1. Then |A0 (v)| = 1 and |Ai (x)| ≤ (r − 1)(r − 2)i−1
for 1 ≤ i ≤ s − 1. Since
s−1
[ s−1
X
n = |V (G)| = | Ai (v)| ≤ 1 + (r − 1) (r − 2)i ,
i=0 i=1

this is a contradiction.
26 CHAPTER 2. CONNECTED GRAPHS AND DISTANCE

38. No. Suppose that G is a connected graph with diam(G) ≥ 2 such that H =
Per(G) is complete. Let u and v be two antipodal vertices of G. Then 1 =
dG (u, v) = diam(G) ≥ 2, which is a contradiction.
39. For the graph G in Figure 2.2 and the four distinct vertices v1 , v2 , v3 , v4 in G,
vi+1 is an eccentric vertex of vi for i = 1, 2, 3. Also, e(v1 ) = 4, e(v2 ) = 6 and
e(v3 ) = 8.

v.....4. v1
... ... ............ ............. ............ .............. .............. ............
... ... ... ... .. .. ...

. ............. ............ ............. ............. .............

v2 ............... ... .... ... ... ...

............ ............ ............ ............ ..............

... ... ... ... ..
v3

Figure 2.2: The graph in Exercise 39

40. In the graph G of Figure 2.3, each vertex is labeled with its total distance.
The median is also shown.
38
....
..... .....
....
37 29
.........
....
..... .......... ........ ........
...... .... ...........
.... ......
.......
....... .. 28 26 26................... ......
......
........... 34 42 52
.. ..... ... ....
G: ........
........ ....
. .
.............
...... ......
....
..... .....
..........
...... .......
.. .
...........
..
...... .......
....
.. ..
..... ......
.... Med(G) : .............. ..... .....
....
... ...... ..
...
..... ...........
....
..
.
u v ...... ...... .........
........ .......
.........
.
.
..
u v
37 ....
..... ...... 29
...
38

Figure 2.3: The median of a graph

41. Let 2k + 1 ≥ 3 and let G be the graph obtained by identifying the center v of
K1,2k+1 with the end-vertex v2k+1 of the path P = (v1 , v2 , . . . , v2k+1 ). Then
vk+1 is the only central vertex of G. Observe that
   
2k + 1 2k + 2
td(v) = + (2k + 1) = ,
2 2
   
2k 2k + 2
td(v2k ) = + 1 + 2(2k + 1) = + 2,
2 2
and for an end-vertex x adjacent to v,
 
2k + 3
td(x) = + 2(2k).
2
Since all other vertices of G have a total distance greater than that of v, the
vertex v is the only median vertex of G. Thus
d(Cen(G), Med(G)) = k.
 27

42. (a) Proof. Let u and v be vertices at distance 3 in G. Select any two
vertices x and y. We consider three cases. Recall that the closed neigh-
borhood N [u] of u consists of u and all the neighbors of u.
Case 1. x ∈ N [u] and y ∈ N [u]. Then (x, v, y) is a path in G and
d(x, y) ≤ 2.
Case 2. x ∈ N [u] and y ∈
/ N [u]. Then (x, v, u, y) is a path in G and
d(x, y) ≤ 3.
Case 3. x ∈ / N [u] and y ∈
/ N [u]. Then (x, u, y) is a path in G and
d(x, y) ≤ 2.
Since d(x, y) ≤ 3 in each case, the diameter is at most 3.
(b) If G has diameter at least 4, then G has diameter at most 3; so G cannot
be self-complementary. If diam(G) = 1, then G is complete and is not
self-complementary. So the diameter must be 2 or 3 for G to be self-
complementary.
Figure 2.4(a) shows a self-complementary graph of order 4k and diam-
eter 3 and a self-complementary graph of order 4k + 1 and diameter 2,
where the bold lines between two graphs indicate that all possible edges
join these two graphs.
(c) If the diameter of G is 2, the diameter of G can be any integer k for
2 ≤ k ≤ n − 1. For k = 2, see Figure 2.4(b). For 3 ≤ k ≤ n − 1, let G be
the graph obtained by identifying a vertex of Kn−k+1 with an end-vertex
of the path Pk . Then diam(G) = 2 and diam(G) = k.
43. Proof. Certainly, d(G, H) = 0 if and only if G ∼ = H. Since each 2-switch
from F to F 0 results in an inverse 2-switch from F 0 to F , the distance d is
symmetric. Let F, G and H be three graphs in Gs such that d(F, G) = a
and d(G, H) = b. Then there are a 2-switches that transforms F into G and
b 2-switches that transforms G into H. Hence there is a sequence of a + b
switches that transforms F into H. Therefore,

d(F, H) ≤ a + b = d(F, G) + d(G, H)

and so the triangle inequality holds.

44. Proof. Let G be a connected graph. Since (1) D(u, v) ≥ 0, (2) D(u, v) = 0 if
and only if u = v and (3) D(u, v) = D(v, u) for every pair u, v of vertices of G,
it remains only to show that detour distance satisfies the triangle inequality.
Let u, v and w be any three vertices of G. Since the inequality D(u, w) ≤
D(u, v) + D(v, w) holds if any two of these three vertices are the same vertex,
we assume that u, v and w are distinct. Let P be a u − w detour in G of
length k = D(u, w). We consider two cases.

Case 1. v lies on P . Let P1 be the u − v subpath of P and let P2 be the v − w

subpath of P . Suppose that the length of P1 is s and the length of P2 is t. So
28 CHAPTER 2. CONNECTED GRAPHS AND DISTANCE

.......
.................... ............................ .......
.................... ...........................
...... ......
qqqqqqqqqqqq qqqqqqqqqqqq ........
..... ..... .....
.....
qqqqqqqqqqqq qqqqqqqqqq .........
... .... ... .... .
... ... ... ... ... ...
....
qqqqq .. Kk
... ....
qqqqqqq
...

qqqqqqqqqq qqqqqqqqqqq ......... qq qq

... ...
....
... K k ..
.
...
..
...
.. ....
... K k ......
..
qqq qqqqq qqqqqqqq .....
.
.
... Kk
..
.
...
... .. .. ... ...
.... .. ..... .. .... .. .... .
...... ...
. .
..... ...... . .
.. ...... ....
.......................... .......
...................... .......................... ...........................
qqq q qqq qq qqq q qqq q
qqqqq qqqqq qqqqq qqq qq
q
qq qqqqqqq qq q qqqq qqqqqqq qqqqqq
q qq q qq qq qq qq qqq
....................... ....................... ........................ ....................
...... ..... ...... ..... ...... ..... ....... .....
.... ... .... ... .... ..... ...
... ... ... ... ... ...
... ... ...
.... ... .... ... .... ... .... ...
.... K .... ....
. K .
K . ....
K .
..
... k
..
... ... k ..
.. ... k ..... qqq qq
... k ......
qqq
... ... . ... ...
q
.... . . .... .. .... .. .
q
.. ..
. ... .... ..
qqqqqqqqq qqqq
...... .. ...... . ...... .. .
..........................
qq
.......................... ......................... ...... .
..........................
q
qqqqqqqqq qqqq qqq qqqqqqqqqqqqqqq
qqqqqqqqqqqq qq qqqqqqqqq
qqqq..........q..qqqq
....
(a)

......... ...............
......... ... .....
...... ......
...... ......
.
..
....... ......
.. ......
...... ......
qqqqqqqqqq
......
qqq q
..........
...............
qqqq qq
... ...
...
qqq qqqqqq
..

qqq
qqq qqqqq q q q
q
qqq qqqqqq q q qqqqq qq
qqq qqq qq q q
qqq ..............q..q.q...........................................q.q..q.......... qqq
q..q....... .. q
.....
...
....
....
Kn−4 ............
........
.............. ........
..........................................

(b)

Figure 2.4: Graphs in Exercise 42

s + t = k. Therefore,

D(u, w) = k = s + t ≤ D(u, v) + D(v, w).

Case 2. v does not lie on P . Since G is connected, there is a shortest path

Q from v to a vertex of P . Suppose that Q is a v − x path. Thus x lies on
P but no other vertex of Q lies on P . Let r be the length of Q. Thus r > 0.
Let the u − x subpath P 0 of P have length a and the x − w subpath P 00 of
P have length b. Then a ≥ 0 and b ≥ 0. Therefore, D(u, v) ≥ a + r and
D(v, w) ≥ b + r. So

D(u, w) = k = a + b < (a + r) + (b + r) ≤ D(u, v) + D(v, w),

and so the triangle inequality holds.

Chapter 3

Trees

Section 3.1. Nonseparable Graphs

1. Proof. Assume, to the contrary, that v is a cut-vertex of G. Then there
is a vertex w belonging to a component of G − v that does not contain u.
Since every u − w path contains v, it follows that d(u, w) > d(u, v), which is
a contradiction.
2. Proof. Since G is nonseparable, G − u is connected. Since u is a cut-vertex
of G − v, it follows that (G − v) − u = G − {u, v} is disconnected and so
(G − u) − v is disconnected. Thus v is a cut-vertex of G − u.
3. Proof. Suppose first that u and v are two vertices in a nonseparable graph
G of order 3 or more. By Theorem 3.3, there is a cycle C of order 3 or more
containing u and v. Therefore, there are two internally disjoint u − v paths
on C.
For the converse, suppose that for every two vertices x and y of G, there
are two internally x − y paths. Thus G is connected. Assume, to the contrary,
that G contains a cut-vertex v. By Theorem 3.2, G contains two vertices u
and w such that v lies on every u − w path of G. Then G does not contain
two internally disjoint u − w paths, which is a contradiction.
4. Proof. Since H is clearly connected, it remains to show that H contains no
cut-vertices. Let x ∈ V (H). If x = v, then H − x = G, which is connected.
If x ∈ V (G), then G − x is connected since G is nonseparable. Since v is
adjacent to at least one vertex of G − x, it follows that H − x is connected.
Thus H has no cut-vertices.
5. Proof. Let H be the graph obtained by adding two vertices u and w to
G and joining u to the two vertices of U and joining w to the two vertices
of W . By Corollary 3.5, H is nonseparable. By Corollary 3.4, H contains
two internally disjoint u − w paths. Deleting u and w from these two paths
produces two disjoint paths connecting the vertices of U and the vertices of
W.

29
30 CHAPTER 3. TREES

6. (a) Proof. Suppose first that G contains disjoint k-cycles C and C 0 . Let
u1 , u2 ∈ V (C) and w1 , w2 ∈ V (C 0 ). Let U = {u1 , u2 } and W = {w1 , w2 }.
By Corollary 3.6, G contains two disjoint paths P1 and P2 connecting
the vertices of U and the vertices of W . Let xi be the last vertex of Pi
(i = 1, 2) belonging to C and let yi be the first vertex of W following xi
on Pi . For i = 1, 2, let Qi be the xi − yi subpath of Pi . Suppose that the
length of Qi is `i ≥ 1. Since there is an x1 − x2 path on  C of length at
least k2 and a y1 − y2 path on C 0 of length at least k2 , it follows that
G contains a cycle of length at least 2 k2 + `1 + `2 ≥ k + 2, which is a
 

contradiction.
Next assume that G contains k-cycles C and C 0 having exactly one
vertex v in common. Suppose that u ∈ V (C) and w ∈ V (C 0 ), where
u, w 6= v. Since v is not a cut-vertex of G, there is a u − w path P in G
not containing v. Let x be the last vertex of P belonging to C and let y
be the first vertex of C 0 following x on P . Let Q be the x − y subpath
of P, where Q has length `. Then G contains a cycle of length at least
2 k2 + ` ≥ k + 1, again, producing a contradiction.
(b) Every cycle in K2,t , t ≥ 3, has length 4 and there are two cycles in the
graph K2,t that have only two vertices in common.

7. (a) Proof. Suppose that there is a graph G containing a cut-vertex v of

degree 1. By Theorem 3.2, there are vertices u and w distinct from v such
that v lies on every u − w path of G. Thus deg v ≥ 2, a contradiction.
(b) Proof. Let v be a cut-vertex of a graph G such that deg v = 2. By
Theorem 3.2, there are vertices u and w distinct from v such that v lies
on every u − w path of G. Let P be a u − w path in G. Then there are
vertices x and y on P such that (x, v, y) is a subpath of P and xy ∈
/ E(G).
Thus x and y belong to distinct components Gx and Gy of G − v. Since
deg y ≥ 2, it follows that y is adjacent to a vertex z in Gy . Since v is
adjacent only to x in G − y, it follows that y is on every v − z path in G
and so y is a cut-vertex of G. Similar, x is a cut-vertex of G.
(c) The unique cut-vertex in K1,k has this property.

8. (a) The statement is false. For each integer n ≥ 2, let Gn = 2Kn ∨ K1 . Then
G has exactly one cut-vertex v but the connected subgraph Kn ∨ K1 =
Kn+1 contains v and is nonseparable.
(b) The statement is false. The graph G = Cn has this property but G is
nonseparable.

9. Proof. Since v is a cut-vertex of G, it follows that G − v is disconnected.

Let u, w ∈ V (G)−{v}. We show that there is a u−w path in G not containing
v, which implies that v is not a cut-vertex of G. We consider two cases.

Case 1. u and w belong to distinct components of G − v. Then u and w are

nonadjacent in G and so (u, w) is a path in G − v.
 31

Case 2. u and w belong to the same component G1 of G − v. Let G2 be

another component of G and let x ∈ V (G2 ). Then neither u nor w is adjacent
to x in G. This implies that ux and wx are edges in G. Thus (u, x, w) is a
u − w path in G − v.
10. No. In the nonseparable graph G = K4 − e, where V (G) = {u, v, w, x} with
uv ∈
/ E(G), let P = (u, w, x, v). Then there is no u − v path Q such that P
and Q are internally disjoint.
11. (a) Proof. By Theorem 3.3, every two vertices of a nonseparable graph lie
on a common cycle. For the converse, suppose that G is a graph of order
3 or more such that every two vertices u and v of G lie on a common
cycle. Then there exist two internally disjoint u − v paths in G. By
Corollary 2.4, G is nonseparable.
Next, let G be a nonseparable graph, where e = uv and f = xy are
two edges of G. The edges e and f are subdivided by introducing two
new vertices s and t, resulting in the four new edges us, sv, xt and ty,
producing a nonseparable graph H. Thus s and t lie on a common cycle
C 0 . Since degH s = degH t = 2, the edges us, sv, xt and ty lie on the
cycle C 0 . Replacing us and sv by e and replacing xt and ty by f produces
a cycle C in G containing e and f . For the converse, suppose that G is
a connected graph of order 3 or more with the property that every two
edges of G lie on a common cycle. Let u and v be two vertices of G.
Let e = ux and f = vy be distinct edges of G incident with u and v,
respectively. Then e and f lie on a common cycle C of G. So u and v lie
on C. Thus G is nonseparable.
A similar proof can be given of the statement: A graph G of order 3
or more is nonseparable if and only if every vertex and edge of G lie on
a common cycle of G.
(b) The graph H is complete if and only if G is nonseparable.
Proof. First, if G is nonseparable, then every two vertices of G lie on
a common cycle of G. Thus every two vertices of H are adjacent and
so H is complete. For the converse, suppose that H is complete. Then
every two vertices of H are adjacent. Thus every two vertices of G lie on
a common cycle of G and so G is nonseparable.
12. Proof. Since G is connected (by Theorem 10.3), it remains to show that G
contains no cut-vertex. Assume, to the contrary, that G contains a cut-vertex
v. Then G contains (necessarily nonadjacent) vertices u and w such that v
lies on every u − w path in G. Thus deg u + deg w ≥ n. Let U be the set of
vertices adjacent to u and let W be the set of vertices adjacent to w. Since G
has order n, the set U ∩ W contains at least two vertices. Let v1 , v2 ∈ U ∩ W .
At least one of these vertices, say v1 , is different from v. Then (u, v1 , w) is a
u − w path in G not containing v, contrary to the hypothesis.
13. The statement is false. The graph G = K2,3 is itself a block of order 5 in G
and there is no cycle in G containing all the vertices of G.
32 CHAPTER 3. TREES

14. k ≥ ` + 1. Proof. We proceed by induction on `. If ` = 0, then k = 1; while

if ` = 1, then k ≥ 2. Assume, for an integer ` ≥ 2, that every connected graph
with b blocks and c cut-vertices, where 0 ≤ c < `, satisfies b ≥ c + 1. Let G
be a connected graph with k blocks and ` cut-vertices. By Theorem 3.8, G
contains a cut-vertex v with the property that, with at most one exception,
all blocks containing v are end-blocks. Since G has at least two cut-vertices,
exactly one block B containing v is not an end-block. For each end-block
B 0 containing v, delete V (B 0 ) − {v} from G. In the resulting graph H, the
vertex v is not a cut-vertex. Thus H has j blocks and ` − 1 cut-vertices,
where j ≤ k − 1. By the induction hypothesis, j ≥ (` − 1) + 1 = ` and so
k ≥ j + 1 ≥ ` + 1.
15. The statement is false. See Figure 3.1.
...............
.

............. ......... ................. ............. .........

........ ........ . ..... ..........
..... ........ . .................. .................. ......... ........
....
..
............ ..
.............. ................. ...............
.
.. .............. ............. ............ ............
u ............... ... ... .... ... ........... .......... v
.... .... ....

Figure 3.1: The graph in Exercise 15

16. The statement is false. The vertices of the graph G in Figure 3.2 are labeled
with their eccentricities. Thus Cen(G) lies in an end-block.

4 3 4 5
........ .
.............. ....... .......
........ ............ ..............
..... .....
..... ..... ......... ..... .........
..
...... ..... .
......
. ..... .
......
. .....
.. ..... ...... ..... ...... .....
..... ............ ............ .....
.... .
.............
G: 5 ...............
.....
..... ..
.
...
........... .........
..... ... ..
.....
.......... ..................
.
....
...
3 4 5
.......

4 3

Figure 3.2: The graph in Exercise 16

17. (a) Proof. Assume, to the contrary, that G has a cut-vertex u that is
a peripheral vertex. Let v be a vertex that is farthest from u. Then
d(u, v) = e(v) and so e(v) = d(u, v). Since u is farthest from v, it follows
that u is not a cut-vertex of G by Exercise 1. This is a contradiction.
(b) The statement is false. The vertices of the graph G in Figure 3.3 are
labeled with their eccentricities. Then P er(G) lies in a block that is not
an end-block of G.
18. (a) Proof. Consider the wheel W2k+2 = C2k+2 ∨ K1 , where v is the cen-
tral vertex of W2k+2 . Then rad(W2k+2 ) = 1 and rad(W2k+2 − v) =
rad(C2k+2 ) = k + 1. Thus rad(W2k+2 − v) = rad(W2k+2 ) + k.
(b) Proof. First observe that there are graphs G for which rad(G − v) =
rad(G) − 1 for some vertex v of G that is not a cut-vertex. For exam-
ple, for an end-vertex v of P2k , rad(P2k − v) = rad(P2k−1 ) = k − 1 =
rad(P2k ) − 1.
 33

3 .
..............

4 .
3 .......
2 3 .
4
............... .............. .......... .......... .
............
..... ..... .......... ..... ......... .................. .......
.....
..... ........ ..... ..
....... ..... ......... ..... ..
.......
..... ..... . ..... ..... . ..... .... . .
..... ..... .
.......... ......... ..... .... .........
G: .......
..... .........
.
.........
..... ......... .....
........
..... .........
.........
..... ..........
.....
..
....... ..... .
..
...... .....
..... .... .
..
. . .....
...... ..... .
..
.. .....
.....
.... ..... .. ..... . .. ........
. .. . ... ... . . . . .. . .. .
. .
. . . .
.... .
............ ......... ......... ......... ........

4 3 2 3 4
....
...........

Figure 3.3: The graph in Exercise 17(b)

Now, let G be a nontrivial connected graph and let v be a vertex of

G that is not a cut-vertex. Let u ∈ V (G) and let w be a vertex of
G that is distinct from u. Suppose that dG (u, w) = k. If there exists
a u − w geodesic in G that does not contain v, then dG−v (u, w) = k.
If, on the other hand, every u − w geodesic in G contains v, then
dG−v (u, w) > k. This implies that if u is a vertex of G with eG (u) = k
and dG (u, v) 6= k, then eG−v (u) ≥ k. Suppose that dG (u, v) = k
and P = (u = u0 , u1 , . . . , uk = v) is a u − v geodesic in G. Then
dG (u, uk−1 ) = dG−v (u, uk−1 ) = k − 1. Thus eG−v (u) ≥ k − 1. Therefore,
rad(G − v) ≥ rad(G) − 1.
(c) Proof. By assumption, H is a connected induced subgraph of G of
minimum order having radius r. For a non-cut-vertex v of H, rad(H −
v) 6= r. If rad(H −v) < r, then by (b), rad(H −v) = r−1. If rad(H −v) ≥
r + 1, then there are induced subgraphs of H − v whose radius is 1. Since
the removal of any non-cut-vertex of a connected graph can reduce the
radius by at most 1 (by (b)), it follows that repeated deletion of non-
cut-vertices from H − v will result in an induced subgraph of H − v (and
therefore of G) with radius r. This is a contradiction.

Section 3.2. Introduction to Trees

19. Proof. Since G is a connected graph of order 3 or more, either deg u ≥ 2
or deg v ≥ 2, say deg v ≥ 2. Then there is a vertex w different from u that
is adjacent to v. Assume, to the contrary, that v is not a cut-vertex. Thus
G − v is connected, and so there is a u − w path P in G − v. However then,
P together with v and the two edges uv and vw form a cycle containing the
bridge uv. This contradicts Theorem 3.10.

20. Proof. Assume, to the contrary, that there exists a connected graph G whose
vertices have even degrees but G contains a bridge e = uv. Then G−e contains
exactly two components, one of which contains u and the other contains v.
Then each of the components of G − e has exactly one odd vertex, which is
impossible.
34 CHAPTER 3. TREES

21. (a) The only example is a double star where each central vertex has degree
4.
(b) Let T be a tree of order n where 75% of the vertices have degree 1 and the
remaining 25% vertices have degree x ≥ 2. Then (3n/4) · 1 + (n/4) · x =
2(n−1). Solving for n, we have n = 8/(5−x). The only possible solutions
for n are when x = 3 or x = 4. If x = 3, then T = K1,3 ; while if x = 4,
then T is the tree in (a).
P
22. The average degree of T is ( v∈V (T ) deg v)/n = 2(n − 1)/n = 2 − 2/n.

23. There are 20 forests of order 6 (see Figure 3.4).

s s s s s s s s s s
.......................................................................................... s s s s s s
..................................................................... s s

s s s s
..................................................................... s s s s s s s s s s s s s s

s s s
............................................ s s s s s s s s
........................ s s s s s s s
s
s s s s s s s s s s s s s s s s s
s ... s s s s
...
...
s s s
.................................................................... s s s s
...
. s s s s
.......................................... s s s s s s s....................s
s s s . s...
...
s
... ...
... .....
...
s s
... ...
s s s s s s s s s s ....
..
...
s s s ........
..................... ...
...
. s
............
...
....
s .
.. s s
...

Figure 3.4: The 20 forests of order 6 in Exercise 23

24. Proof. Assume first that G is a forest. Then every induced subgraph H of
G is a forest. If H contains a nontrivial component (a tree), then H contains
at least two end-vertices. Otherwise, H contains at least one isolated vertex.
In either case, G contains a vertex of degree at most 1. For the converse,
assume that G is not a forest. Then G contains a cycle C. Thus G[V (C)] is
an induced subgraph of G that contains the spanning cycle C. This implies
that every vertex of G[V (C)] has degree at least 2.

25. Claim: A graph G is a forest if and only if every connected subgraph of G

is an induced subgraph of G.

Proof. We first show that every connected subgraph of a forest F is an

induced subgraph of F . Assume, to the contrary, that this is not the case.
 35

Then there is a connected subgraph H of F that is not an induced subgraph

of F . This implies that H contains two nonadjacent vertices u and v that are
adjacent in F . Since H is connected, H and therefore F contains a u − v path
P . However then P together with uv is a cycle in F , which is impossible.
For the converse, suppose that G is a graph that is not a forest. Then G has
a cycle C. Let e be an edge of C. Then C −e is a connected subgraph of G with
vertex set V (C). Since G[V (C)] contains e, it follows that G[V (C)] 6= C − e.
Thus C − e is not an induced subgraph of G.
26. The statement is false. For example, let G be the graph consisting of two
components, one of which is P2 and the other is obtained from K3 by adding
a pendant edge at each vertex of K3 .
27. Proof. Assume first that at least one of e1 and e2 is not a bridge, say e1 is
not a bridge. Then G − e1 is connected. Thus G − e1 − e2 = (G − e1 ) − e2
has at most two components.
For the converse, suppose that e1 and e2 are bridges of G. Since e1 is a
bridge, G − e1 has exactly two components, say H and F . We may assume
that e2 ∈ E(H). Then e2 is a bridge of H, implying that H − e2 consists of
two components. Therefore, G − e1 − e2 has three components.
28. Proof. Let G be a 3-regular graph. If G has a bridge e, then each vertex
incident with e is a cut-vertex. (See the proof of Exercise 19.) For the converse,
suppose that G has a cut-vertex u. We may assume that G is connected,
for otherwise, consider the component of G containing u. Then G − u is
disconnected and so G − u has at least two components. Since deg u = 3,
it follows that G − u has at most three components. If G − u has three
components, say G − u = G1 + G2 + G3 , then u is adjacent to exactly one
vertex vi in Gi and each edge uvi (1 ≤ i ≤ 3) is a bridge of G. If G−u has two
components, say G − u = G1 + G2 , then u is adjacent to exactly one vertex
in G1 or to exactly one vertex in G2 , say u is adjacent to exactly one vertex
v1 in G1 . Then uv1 is a bridge of G.
29. Proof. By Theorem 3.9, the center of every graph lies in a block. Since the
only blocks of T are K2 , the only possible centers are K1 and K2 .
30. (a) Proof. Let P be a longest path in T , say P is a u − v path. Thus the
length of P is diam(T ). Then u and v are end-vertices of T and P − u − v
is a path of length diam(T ) − 2 in T 0 . Thus diam(T 0 ) ≥ diam(T ) − 2.
Assume, to the contrary, that T 0 contains a longest path Q of length j,
where j ≥ diam(T )−1. Suppose that Q is an x−y path. Each of x and y
is adjacent to an end-vertex of T , for otherwise, x and y are end-vertices
of T and so do not belong to T 0 . This implies that T contains a path of
length diam(T ) + 1, which is impossible.
Certainly, no central vertex of T is an end-vertex of T . Thus each
central vertex of T belongs to T 0 . Let u ∈ V (T 0 ) and suppose that
eT (u) = k. For each vertex v of T such that dT (u, v) = k, the vertex v
36 CHAPTER 3. TREES

is an end-vertex of T . Hence eT 0 (u) ≤ k − 1. Let w be the neighbor of v

on the u − v path in T . Then dT 0 (u, w) = k − 1. Thus eT 0 (u) = k − 1.
Therefore, for every vertex u in T such that u is not an end-vertex of T ,
it follows that eT 0 (u) = eT (u) − 1. In particular, if u is a central vertex
of T , then eT (u) = rad(T ) = rad(T 0 ) + 1.
Let u be a central vertex of T . Then u is not an end-vertex of T . As
we saw, eT 0 (u) = eT (u) − 1. Then any central vertex of T is a central
vertex of T 0 . Thus Cen(T ) = Cen(T 0 ).
(b) Proof. Let T1 be the tree obtained from T by deleting the end-vertices
of T , let T2 be the tree obtained by deleting the end-vertices of T1 and so
on until we arrive at a tree T` which is either K1 or K2 . Then diam(T ) =
diam(T` ) + 2` and rad(T ) = rad(T` ) + `. If T` = K1 , then diam(T` ) = 0
and rad(T` ) = 0. So diam(T ) = 2` and rad(T ) = `. Hence diam(T ) =
2 rad(T ) if Cen(T ) = K1 . If T` = K2 , then diam(T` ) = 1 and rad(T` ) =
1. So diam(T ) = 2` + 1 and rad(T ) = ` + 1. Thus diam(T ) = 2(rad(T ) −
1) + 1 = 2 rad(T ) − 1 if Cen(T ) = K2 .
31. (a) Proof. Since b(vi ) ≤ deg vi for 1 ≤ i ≤ n, it then follows by the First
Theorem of Graph Theory (Theorem 1.4) that
n
X n
X
b(vi ) ≤ deg vi = 2m,
i=1 i=1

as desired.
Pn
(b) Proof. First, assume that i=1 b(vi ) = 2m. By (a), b(vi ) = deg vi for
1 ≤ i ≤ n. This implies that every edge of G is a bridge and so G is a
tree.
For the converse, assume that G is a tree. Then every edge of G is a
bridge. This implies that b(vi ) = deg vi for 1 ≤ i ≤ n. ItP
then follows
n
by the First Theorem of Graph Theory (Theorem 1.4) that i=1 b(vi ) =
2m.
32. Claim: T = K1 or T = P4 .
Proof. Let T be a tree of order n. Since T and T are both trees of order n,
it follows that the sizes of T and T are n − 1. Thus n − 1 + n − 1 = 2(n − 1) =
n

n(n−1)
2 = 2 . Hence 4(n − 1) = n(n − 1) and so (n − 1)(n − 4) = 0, implying
that n = 1 or n = 4. If n = 1, then T ∼ = K1 . If n = 4, then T ∼ = P4 or
∼ ∼
T = K1,3 . Since P4 = P4 and K 1,3 is not a tree, it follows that T = P4 .
33. Proof. Since each end-block of a tree consists of a pendant edge and thus a
leaf, the result follows from Theorem 3.8.
34. Proof. Assume, to the contrary, that dk = deg vk > n−1
 
k for some integer
k with 1 ≤ k ≤ n and so
 
n−1 n−1 n+k−1
deg vk ≥ +1≥ +1= .
k k k
 37

Thus
n
X k
X n
X
2n − 2 = deg vi = deg vi + deg vi
i=1 i=1 i=k+1
 
n+k−1
≥ k + (n − k) = 2n − 1,
k

which is a contradiction.

35. Proof. Suppose that the order of G is n and the size of G is m. Since
G − u and G − v are both trees of order n − 1, the size of each of G − u and
G − v is n − 2. Therefore, m − deg u = n − 2 and m − deg v = n − 2. Thus
deg u = deg v = m − (n − 2).

36. Proof. Assume, to the contrary, that there exists a tree T containing two
distinct edges e1 and e2 such that the two components of T −e1 are isomorphic
and the two components of T − e2 are isomorphic. Let T1 and T2 be the two
components of T − e1 . Since T1 ∼ = T2 , these two trees have the same size,
say k. Thus the size of T is 2k + 1. The edge e2 belongs either to T1 or
to T2 . Without loss of generality, assume that e2 belongs to T2 . Since one
component of T − e2 contains T1 and e1 , the size of one component of T − e2
is at least k + 1 and the size of the other component of T − e2 is therefore at
most k − 1. Thus the two components of T − e2 are not isomorphic, which is
a contradiction.

37. Proof. The graph C n+2 is (n − 1)-regular. Since δ(C n+2 ) = n − 1, it follows
by Theorem 3.20 that T is isomorphic to a subgraph of C n+2 .

38. Proof. Since T is not a star, the order n of T is at least 4. We proceed by

induction on n ≥ 4. If n = 4, then T = P4 . Since P 4 = P4 , the result is true
for n = 4. Assume that for every tree T of order n − 1 ≥ 4 that is not a star,
T ⊆ T.
Let T be a tree of order n that is not a star. We first show that there is an
end-vertex v ∈ V (T ) such that T − v is not a star. If T = Pn , then let v be an
end-vertex of T . So we may assume that T 6= Pn and let w ∈ V (T ) such that
deg w ≥ 3. If every vertex adjacent to w is not an end-vertex of T , let v be
any end-vertex of T ; otherwise, let v be an end-vertex of T that is adjacent
to w. Thus T − v is not a star.
By the induction hypothesis, T − v ⊆ T − v. Hence T − v is isomorphic to
a subgraph F of T − v. Let φ be an isomorphism from T − v to F . Let u be
the vertex in T that is adjacent to v. We consider two cases.

Case 1. φ(u) 6= u. Since v is adjacent to φ(u) in T , we can extend the

isomorphism φ from T − v to F to an isomorphism from T to a subgraph F 0
of T by defining φ(v) = v.
38 CHAPTER 3. TREES

Case 2. φ(u) = u. Since T is not a star, there exists an end-vertex x in T

such that x is not adjacent to u in T . Define an isomorphism φ∗ from T to a
subgraph F 0 of T by defining φ∗ (v) = φ(x), φ∗ (x) = v and φ∗ (w) = φ(w) for
all w ∈ V (T − v − x). Observe that if x is a adjacent to a vertex y in T , then
v is adjacent to φ(y) in T . Moreover, φ(x) is adjacent u in T .

39. Claim. The 4-cycle C4 is the only graph with this property.
Proof. First, observe that if G = C4 and S is a set of three vertices of G,
then G[S] = P3 . Suppose then that G 6= C4 . If G is a tree, then G cannot
have three or more end-vertices. Thus G = Pn = (v1 , v2 , . . . , vn ) for some
n ≥ 4. However, G[{v1 , v2 , v4 }] = K2 + K1 , which is not a tree. If G is not a
tree, then G contains cycles. Let Ck = (v1 , v2 , . . . , vk = v1 ) be a smallest cycle
in G. If k = 3, then G[{v1 , v2 , v3 }] ∼
= K3 , which is not a tree. If k ≥ 5, then
G[{v1 , v2 , v4 }] ∼
= K2 + K1 , which is not a tree. Thus k = 4. Since G 6= C4 ,
there is a vertex v ∈ V (G) − V (Ck ) such that v is adjacent a vertex in Ck ,
say v is adjacent to v1 . Then v is adjacent to neither v2 nor vk as G contains
no triangles. So G[{v, v2 , vk }] = K 3 , which is not a tree.

40. Proof. Let T be a tree with maximum degree ∆ = ∆(T ) = k. By The-

orem 3.14, the numbers ni of vertices of degree i (1 ≤ i ≤ ∆) satisfy the
identity n1 = 2 + n3 + 2n4 + · · · + (∆ − 2)n∆ ≥ 2 + (k − 2) = k.

41. Proof. If T is a path, the result is immediate. So we may assume that

T is not a path. Therefore, T contains vertices that are not on P . For
w ∈ V (T ) − V (P ), let Pw denote the unique shortest path from w to P and
denote the length of Pw by `(Pw ). For a vertex u of T , we now show that
e(u) = max{d(u, u0 ), d(u, ut )}. Assume, without loss of generality, that

max{d(u, u0 ), d(u, ut )} = d(u, u0 ).

We consider two cases, depending on whether u belongs to P .

Case 1. u ∈ V (P ). Thus u = uk for some integer k with t/2 ≤ k ≤ t.

Obviously, d(u, ui ) ≤ d(u, u0 ) for 0 ≤ i ≤ t. Let z ∈ V (T ) − V (P ) and
let Pz be a z − uj path. If j ≤ k, then since P is a longest path in T , it
follows that `(Pz ) ≤ j and so d(u, z) ≤ d(u, u0 ). On the other hand, if j > k,
then since P is a longest path in T , it follows that `(Pz ) ≤ t − j and so
d(u, z) ≤ d(u, ut ) ≤ d(u, u0 ). Thus, e(u) = d(u, u0 ).

Case 2. u ∈ V (T ) − V (P ). Let Pu be a u − uk path, where 0 ≤ k ≤ t. Letting

max{d(uk , u0 ), d(uk , ut )} = d(uk , u0 ), we then proceed as in Case 1 to show
that e(u) = max{d(u, u0 ), d(u, ut )} = d(u, uk ) + d(uk , u0 ) = d(u, u0 ).

42. (a) Proof. Let v be a central vertex in T and let P = (u0 , u1 , . . . , ut ) be a

longest path in a tree T . By Exercise 41, e(v) = rad(T ) = max{d(v, u0 ), d(v, ut )},
say e(v) = d(v, u0 ). Assume, to the contrary, that v is not on P . Let
 39

P 0 = (v = v0 , v1 , . . . , vr = w) be a shortest path from v to a ver-

tex w of P . Then e(v) = d(v, w) + d(w, u0 ). Again, by Exercise 41,
e(v1 ) = max{d(v1 , u0 ), d(v1 , ut )}. Since d(v1 , u0 ) = d(v, u0 ) − 1 and
d(v1 , ut ) = d(v, ut ) − 1, it follows that e(v1 ) < e(v), contradicting the
fact that v is a central vertex in T .
(b) let P = (u0 , u1 , . . . , ut ) be a longest path in a tree T . By (a), every central
vertex v of T lies on P and e(v) = max{d(v, u0 ), d(v, ut )}. If t = 2k + 1
is odd, then v = vk or v = vk+1 and e(vk ) = e(vk+1 ) = rad(T ); while if
d = 2k is even, then v = vk and e(vk ) = rad(T ). Since |e(vi )−e(vi+1 | ≤ 1
for 1 ≤ i ≤ d − 1 (Theorem 2.7), it follows for every integer k with
rad(T ) < k ≤ diam(T ) that P contains two vertices having eccentricity k
in T .

Section 3.3. Spanning Trees

43. See Figure 3.5.

2 2 2
........ ....... .............
......... ............ ... ....
....
.... ..... ..... ....
..
.. ..
..... ...
... ....
... ... ....
.... .... ... ....
..... ..... ...
. ....
.
1 ............. ............. 3 1 ............. ............ 3
..
1 .............. ............. 3
... ... ... ... ...

(2) (1) (3)

..
1 ..................... ............
.... 2 ...
1 .............. ...........
.......... 2 1 ........................ .............
... 2
....
1 ............. .....
... ...
........... 2
..... ...... ..... ......
.....
..... ..
......
. .....
..... ..
......
.
..... . .
..... ...... ..... ......
..... ..... ..... .....
............ ...... ..... ......
3 ............. ... ... 4 3 ................. ............. 4 3 ...............
............
....... 4 3 ................ ............. 4
... ... . ... . . ...

(1,1) (2,2) (4,4) (3,3)

.... ..
1................ .............
... 2 1 ................ .............
... 2 1 ............. .............
.. 2 1 ............... .............
... 2

.. .....
3 ............. .............
... 4 3 .............
... .............
... 4 3 ........... .............
... 4 ....
3 ............. .............
... 4
(3,4) (1,3) (1,2) (2, 4)
............. .............. ... ....
1 .................... ............... ..... .............
1 ...
........
2 ...
1 ............. ... ....
........... 2 . 2 1 .................... ... 2
...... ...... ..... .....
..
......
. .
..
...... .....
..... .....
.....
. ..
...... ......
.....
..... .....
..... ..... ..... .....
..... ..... ......... .....
3 .............. ............. 4 3 ............. ............ 3 ............ .... .... 3 ............. ..........
4
... ... .... .... 4 .... ... 4 ... .......

(2,3) (3,2) (1,4) (4,1)

1 ....................... ........ ....

1 ...................... .......... .... ........... ....
1 ..................... ...........
......
......... 2
..... ..........
......... 2 1 ..................... .......... 2
..... ..........
......... 2
.....
..... ......... .......... ..... .......... ..... .....
....... .. .......... ..
... ...........
..... .......... ........... ..... .....
...... ..... ...... ......... ...... .......... ...... ..........
...... ............ ...... ............ ...... ............ ...... ............
.............. ... ... 3 ............... ... ... 4 3 .............. ... .. ............... ... ... 4
3 . ... 4 ... ... . ..... 4 3 ... ...

(4,3) (3,1) (2,1) (4,2)

Figure 3.5: The trees in Exercise 43

44. (a) Let T be a tree with vertex set V (T ) = {1, 2, . . . , n}, n ≥ 3, that has
constant Prüfer code (a1 , a2 , . . . , an ) where ai = ` for some ` ∈ V (T )
40 CHAPTER 3. TREES

for 1 ≤ i ≤ n. Let V1 = V (T ) − {`} = {b1 , b2 , . . . , bn−1 }, where b1 <

b2 < . . . < bn−1 . Then b1 , b2 , . . . , bn−2 are end-vertices adjacent to the
vertex `. Removing b1 , b2 , . . . , bn−2 from T , we obtain K2 containing the
edge bn−1 `. Thus every vertex in V1 is adjacent to the vertex ` and so
T = K1,n−1 whose central vertex is `.
(b) Let T be a tree with vertex set V (T ) = {1, 2, . . . , n}, n ≥ 3, that has
Prüfer code (a1 , a2 , . . . , an ) where ai ∈ {`1 , `2 } for some `1 , `2 ∈ V (T )
for 1 ≤ i ≤ n, `1 6= `2 and each of `1 and `2 appears at least once
in (a1 , a2 , . . . , an ). In this case, n ≥ 4. Let V1 = V (T ) − {`1 , `2 } =
{b1 , b2 , . . . , bn−2 }, where b1 < b2 < . . . < bn−2 . We know that `1 and `2
are not end-vertices, while b1 , b2 , . . . , bn−3 in V1 are end-vertices, each of
which is adjacent to `1 or to `2 . Removing b1 , b2 , . . . , bn−3 from T , we
obtain P3 whose central vertex is bn−2 ∈ {`1 , `2 }. Thus every vertex in
V2 is adjacent to `1 or to `2 , which implies that T is a double star whose
central vertices are `1 and `2 .
(c) Let T be a tree of order n ≥ 3. Observe that if T contains a vertex `
whose degree is 3 or more, then ` appears twice in the Prüfer code of T .
This implies that ∆(T ) ≤ 2. On the other hand, T is connected and so
T = Pn .
45. The labeled tree in Figure 3.6 has Prüfer code (4, 5, 7, 2, 1, 1, 6, 6, 7).
11 10
....
9
............. ..... .....
... .............
... ...

............. ............... ............. ............. ............... ............. ............. .............

... . ... ... . ... ... ...
3 4 5 7 6 1 2 8

Figure 3.6: The tree in Exercise 45

46. Since seven numbers appear twice each and one number appears three times,
the length of the Prüfer code is 17. However, the length of the Prüfer code of
a tree of order n is n − 2, so n − 2 = 17 and n = 19. Thus the number of leaves
is 19−7−1 = 11. Also, by Theorem 3.14, n1 = 2+n3 +2n4 = 2+7+2·1 = 11.
47. Since every vertex v of a tree T appears deg v − 1 times in its Prüfer code, it
follows that the degree sequence of T is 4, 3, 3, 2, 2, 1, 1, 1, 1, 1, 1.
48. (a) If G contains three or more u − v paths for two distinct vertices u and
v, then G contains at least two cycles, which contradicts the fact that G
has cycle rank 1.
(b) The statement in (a) is false if G has cycle rank 2. Let the partite sets
of G = K2,3 be U = {u1 , u2 } and W = {w1 , w2 , w3 }. Then the graph G
has order n = 5 and size m = 6. Thus m − n + 1 = 2 and G has cycle
rank 2. There are three u1 − u2 paths in G.
49. For each even integer n ≥ 2, rad(Pn ) = n/2.
 41

50. Let T be the tree obtained from the path P2r by adding n − 2r new vertices
and joining n each of these vertices to one of the central vertices of P2r . The
tree T has the desired property.

Section 3.3. Spanning Trees

51. Proof. If G is a unicyclic graph of order n, then G has size n and at least
three vertices of G lie on a cycle of G. Hence if sn : d1 , d2 , . . . , dn is a degree
sequence ofPG, then 1 ≤ di ≤ n − 1 for 1 ≤ i ≤ n, at most n − 3 terms of sn
n
are 1 and i=1 di = 2n.
We verify the converse by induction. Suppose that P3d1 , d2 , d3 is a sequence
of integers with 1 ≤ di ≤ 2 for 1 ≤ i ≤ 3 such that i=1 di = 2 · 3 = 6. Then
the sequence is necessarily 2, 2, 2, which is the degree sequence of the unicyclic
graph K3 . The statement is therefore true for n = 3. Assume for an integer
k ≥ 3 that if sk : d1 , d2 , . . . , dk is a sequence of integers
Pk with 1 ≤ di ≤ k − 1,
at most k − 3 terms of which are 1 which satisfies i=1 di = 2k, then sk is a
degree sequence of a unicyclic graph of order k. Let

sk+1 : d1 , d2 , . . . , dk+1

be a sequence of
Pintegers with 1 ≤ di ≤ k, at most k − 2 terms of which are 1
k+1
which satisfies i=1 di = 2k + 2. If every term di = 2, then sk+1 is a degree
sequence of Ck+1 , which is a unicyclic graph. Suppose that not all integers di
are 2. Then some term is 1, say dk+1 = 1 and some integer dj is 3 or more.
Let dk be the maximum term in sk+1 . Then dk ≤ k. If dk = k, then no other
Pk+1
term has this value since i=1 di = 2k + 2. Let

s0 : d01 , d02 , . . . , d0k

be the sequence where d0i = di for 1 ≤ i ≤ k − 1 and d0k = dk − 1. Then

Pk
1 ≤ d0i ≤ k − 1 for 1 ≤ i ≤ k, i=1 d0i = 2k and at most k − 3 terms of s0
are 1. By the induction hypothesis, s0 is the degree sequence of some unicyclic
graph H with V (H) = {v1 , v2 , . . . , vk } where deg vi = d0i . Let G be the graph
obtained from H by adding a new vertex vk+1 which is adjacent to vk . Then
sk+1 is the degree sequence of the unicyclic graph G.
52. (a) Let T be any spanning tree of G that is a distance-preserving tree, say
from the vertex v. Let x, y ∈ V (T ) such that dT (x, y) = diam(T ). Now
dT (x, v) = dG (x, v) ≤ eG (v) and dT (y, v) = dG (y, v) ≤ eG (v). Thus

diam(T ) = dT (x, y) ≤ dT (x, v) + dT (v, y)

≤ 2eG (v) ≤ 2 diam(G).

(b) The statement is true. Let k be a positive integer. Consider G = Kk+2 .

Then T = Pk+2 is a spanning tree of G and diam(T ) = k + 1 > k · 1 =
k diam(G).
42 CHAPTER 3. TREES

v v v
....... ......... .......
... ... ...........
........... ......... .................... ...... . ...........
..... ..... .....
......... ..... .
..
......
......
...... ....
....... ......
..... ......
....... ... ..... ........... .....
............. ......... ............... ... .... ............... ........
..... ... ... ...........
...
...... .. ...
........
..
.....
. ...
.....
. .....
...... ......
...... ......
..... ......
...... .....
.. .
...... ...
.. .
... ....
...
..
. .............. ............. ............ ....
.... ..... .
. ..............
.. ... ... ... ..

G T1 T2

Figure 3.7: The graphs in Exercise 53

53. See Figure 3.7.

54. (a) Suppose that T is a spanning tree of G that is distance-preserving from
some vertex of G. The fact that diam(G) ≤ diam(T ) is obvious. For
diam(T ) ≤ 2 diam(G), see the proof in Exercise 52(a).
(b) Let G be the graph of order 2a + 1, where
V (G) = {v0 , v1 , . . . , va } ∪ {u1 , u2 , . . . , ua }
such that

E(G) = {vi vi+1 : 0 ≤ i ≤ a − 1} ∪ {ui ui+1 : 1 ≤ i ≤ a − 1}

∪{ui vi : 1 ≤ i ≤ a} ∪ {vi ui+1 : 0 ≤ i ≤ a − 1}.

(See Figure 3.8 for the graph G when a = 4.) Let E denote the edge set
of the path
P = (v0 , v1 , · · · , va ).
For a < b ≤ 2a, we define a spanning tree Tb of G such that Tb is
distance-preserving from v0 . For b = a + j with 1 ≤ j ≤ a, let
Qj = (v0 , u1 , · · · , uj )
and
E(Tb ) = E ∪ E(Qj ) ∪ {ui vi+1 : j ≤ i ≤ a − 1}.
See Figure 3.8 for the trees Tb for a < b ≤ 2a.
55. See Figure 3.9.
56. (a) The matrices D and D − A of G are
   
2 0 0 0 0 2 −1 0 0 −1
 0 3 0 0 0   −1 3 −1 0 −1 
   
D=  0 0 2 0 0   D−A= 0
 −1 2 −1 0 
.
 0 0 0 2 0   0 0 −1 2 −1 
0 0 0 0 3 −1 −1 0 −1 3
To calculate a cofactor of D − A, we delete the entries in row 5 and
column 5 and obtain
 43

u 1 u 2 3 u 4 u
.......... ........... ......... ........ ............. .......... .......... ...
... ....
............. ............. ............ ...........
.... ......
....... ........ .......... .........
G: ..... ........
. .... ........
. .
..... ........ .
. T4 : ..... .... ..... ....
.
.....
. ..... ..
. ..... .
.. ..... ..
. ... ..... .... ..... ....
..... ..... ..... .... ..... ..... ... .....
..... .......
.
..
. .
...
.. ..
.
..... ..... ........ ..... ........ ..... ........ ... ..... ..
....
.... .. ........... .......... ........ ........... ...... ....... ...
............. .............. ........ ......... ... .... ............. ... ...
.... ............. ............. .............
... . ... ...
v0 v1 v2 v3 v4

T5 : .......
......... ............. ......... .......
..... ... T6 : .......
.......... ............. .............. ..............
..... .. ......... ..... .. .....
... ..... .. .....
..
..... .... .... .....
.
.....
. ..
...... ..
..... ..
.. ..
..... ..
.. ..
.....
. .... ... .. ... .... ....
..... ..... .... ..... .... .... .....
..... .
.. .... ....
............. ..... ..... . . .
............. ... .
..... ..... .............. ... ............. ............ ............. ...........
...
..... ..... .............
... ... ... ... ... ... ... ... ...

.......... ............. ............. ........... T8 : ............ ............ ............ .............

T7 : ..........
.....
... ... .........
.... .........
.....
.... .... ...
.
..... ..
.....
. ......
. .... .
..... .....
...
....
..... ... ... ............................................................. ............................................................... ... ............ ............
............. ............. ............. .... ..... .... .... ............. .... ....
...

Figure 3.8: The graphs in Exercise 54(b)

............. ........... ............ ..........
...... ......... ........ ........
.....
..... .....
.....
T : .....
..... .....
....
G: .....
..... .
....
................
.
. .....
..... .......
.... ....
.
..
.
... .............
............. .... ..... ............. ................................................................. ...
... ... ... .
u v u v

Figure 3.9: The graphs in Exercise 55

 
 2
 −1 0 0 
 −1 3 −1 0 
(−1)5+5  = 11.
 0 −1 2 −1 
 0 0 −1 2 
Consequently, there are 11 distinct spanning trees of the graph G.
(b) The distinct labeled spanning trees of G are shown in Figure 3.10.

v1
........ ........ ....... ............... .................
........... ........... .......... ... .
.... ...... ... ..... ... ... ... ......
..
..... ... ...
. ... ..... ... ..... ...
.. .... .. .. .. .... ... ....
v2 ............. ............ .............. .............. ............... ............... ............... .............. ...... .............
.. v5 ........

v3 .............
.. ............... v4 ............... ............... ............... ............
... ............. ............. ............
...
...............
.. ..

........... ............... ....... ................ ............... .................

....... ..... ... ...
....
.... .... .. ... ... ...
... .... .... ... ... ...
.... .... .. ....
. ... ...
. ...
.......
. . .
............... ............... ............... ............... .............. ............... ............... .............. ............... ............... ............... ..........

............... ............... ............... .............. .............. .............. ............... ............... ............... ............... ............... ..............

Figure 3.10: The distinct spanning trees of a graph

57. (a) See Figure 3.11.

(b) Proof. A spanning tree T of G having v as an end-vertex is obtained
from a spanning tree T 0 of G − v by joining a vertex u of T 0 to v.
44 CHAPTER 3. TREES

v1 v2 ...............
............... ............... ............... ................ ............... ............... ............... ............... ....... .................. ........
..... ..... .... .. ..........
..... ..... ..... ..... ... .....
..... .........
..... ......... ..... .....
..... ........
G: ........
... ...
..... ........
.....
.....
.....
.
...
.....
.
.. ...
... ...
...
..... .........
..... ........... ........ ..... ..... .........
................ ........ .............. ............... ............... ......... ............... .............. ............... ............... ............... .......

v3 v4
...............
...
... ... ............. ......
... ..
...
.............. ....... ...
............ ...
... ...
.......... .. ........ ..... ........... ..........
..... ..... ..... ........ .....
.
..
..... ..
...... ......... .
......
. . . .
..... ..... ..... ..... .....
.... ..... ..... ......... .....
.... .... ..... ......... .....
................ ............. ............... ............... ............... ........ .............. ...............
.. .. ..

Figure 3.11: The graphs in Exercise 57

Since G − v has order n − 1, the number of spanning trees of T − v is

(n − 1)(n−1)−2 = (n − 1)n−3 . Since there are n − 1 vertices in T − v to
join to v, the number of spanning trees of G having v as an end-vertex is
(n − 1)(n − 1)n−3 = (n − 1)n−2 . Note: For n = 4, (n − 1)n−2 = 32 = 9.

58. We use the Matrix-Tree theorem to prove that there are nn−2 distinct labeled
trees of order n.
Proof. Let tn be the number of distinct labeled trees of order n. For n ≥ 2,
tn is the number of spanning trees of Kn and so tn is any cofactor of the
n × n matrix D − A = [cij ], where cii = n − 1 and cij = −1 for all i, j with
1 ≤ i, j ≤ n and i 6= j. Let B be the (n − 1) × (n − 1) matrix obtained from
D − A by deleting row n and column n. Then

tn = (−1)n+n det(B) = det(B).

For 2 ≤ i ≤ n−1, replace the ith row Ri in B by Ri −R1 , obtaining the matrix
B 0 . Observe that det(B 0 ) = det(B). Next replace the first column C1 in B 0 by
Pn−1 00
i=1 Ci where Ci is the ith column for 1 ≤ i ≤ n − 1, obtaining B . Observe
00 0 00
that det(B ) = det(B ) = det(B) and B = [bij ] is an upper triangular
(n − 1) × (n − 1) matrix such that b11 = 1 and bii = n for 2 ≤ j ≤ n − 1.
Since det(B 00 ) is the product of the entries on the main diagonal, it follows
that det(B) = nn−2 and so tn = nn−2 .

59. Proof. Assume first that e is an edge of a connected graph G that is not
a bridge. Then G − e is connected. Then every spanning tree of G − e is a
spanning tree of G that does not contain e.
For the converse, suppose that e is an edge that does not belong to every
spanning tree of G. Let T be a spanning tree of G not containing e. Then
T + e has a cycle containing e. Therefore, e is not a bridge.

60. By Exercise 59, an edge of F belongs to every spanning tree of G if and only
if e is a bridge of G. Thus F is a forest and so F contains no cycles.

61. Since the size of a complete graph of order n is n2 = n(n−1)

2 and each spanning
tree has n − 1 edges, the maximum possible number of spanning trees in a
 45

connected graph of order n ≥ 4, no two of which have an edge in common,

is n(n−1) n
2(n−1) = 2 . (We will see in Chapter 10 that this is attainable when n is
even.)

62. (a) Proof. Let v be an end-vertex of a spanning tree T of G. Then T − v

is a tree that is a spanning tree of G − v and so G − v is connected. Thus
v is not a cut-vertex of G.
(b) Proof. Since every tree contains at least two end-vertices, it then fol-
lows by (a) that G contains at least two vertices that are not cut-
vertices.
(c) Proof. The edges of G incident with v do not produce a cycle. Let E
be the set of edges incident with v. If there is a spanning tree of G with
edge set E, then the proof is complete; otherwise, there is an edge e1
such that E ∪ {e1 } does not produce a cycle. If there is a spanning tree
of G with edge set E ∪ {e1 }, then the proof is complete. Continuing in
this manner, we obtain a spanning tree T with E ⊆ E(T ).
(d) Proof. We show that ∆(G) ≤ 2. Assume, to the contrary, that ∆(G) =
k ≥ 3. Let v be a vertex of G with deg v = k. It then follows by (c)
that G contains a spanning tree T that contains all edges of G that are
incident with v. Thus T contains a vertex whose degree exceeds 2 and
so T has more than two end-vertices. By (a), G has more than two
vertices that are not cut-vertices, a contradiction. Therefore, as claimed
∆(G) ≤ 2. Since G is connected, G is a path or G is a cycle. Since a
cycle has no cut-vertices, G is a path.

63. For each positive integer k different from 2, we show that there is a connected
graph with exactly k spanning trees. First, a connected graph G has exactly
one spanning tree if and only if G is a tree. For k ≥ 3, the graph G = Ck has
exactly k spanning trees, as there are k possible edges that can be removed
from G to obtain a spanning tree of G. In order for G to have more than one
spanning tree, G must contain a cycle and therefore at least three spanning
trees.

64. Applying the proof of Theorem 3.25, observe that the tree T can be trans-
formed into T 0 by a sequence T = T0 , T1 , . . . , Tk = T 0 of spanning trees of G
such that Ti is transformed into Ti+1 (0 ≤ i ≤ k − 1) by an edge exchange. In
an edge exchange from Ti to Ti+1 , these two spanning trees have n − 2 edges
in common for each i.

65. If r = t, let G = K1,r . Suppose then that 2 ≤ r < t. Let G be the graph
obtained by attaching r − 1 pendant edges to a vertex v of Kt−r+2 . Then at
least one vertex of Kt−r+2 different from v is an end-vertex in a spanning tree
of G. At the other extreme, all other vertices of Kt−r+2 different from v can
be an end-vertex in a spanning tree of G, for a total of (t − r + 1) + (r − 1) = t
end-vertices. This is illustrated in Figure 3.12 for r = 4 and t = 7.
46 CHAPTER 3. TREES

............... ......... ............. ............. ............. .......

........... .......... ... ... ..... .... ..
... ....... ....... .. ... .. ...
... ........ ............. .... ... ...
.
... ....... . ...
... ............. ............. .... ... ...
.........
. ......... ... ...
.................. .... . ...
.. ................... ............. ........ ............... .... ... ......
.... ... ..
. ... ........... ...... ... ... .............
.. ..... .. ...
..... ..
..... ... ... ....... .....
..... ...
..... .. .... ........
..... ... .... ........ ..
. ..... ... .. ......
..... .. .. ..... .... ..... .. . .....
..... ... ... .... .... ................
.....
................ ..... ............. ............. ............ ............... .............. .............
...................................................................... ... ... .... . .... ...
...
...
....
.
.
............. ............. ............
... ... ...

G T1 T2

Figure 3.12: A graph G and two spanning trees of G in Exercise 65

66. (a) The adjacency matrix A of G and the degree matrix D of G are
   
0 1 1 0 2 0 0 0
 1 0 1 0   0 2 0 0 
A=  1 1 0 1  and D =  0 0 3 0 .
  

0 0 1 0 0 0 0 1
Thus  
2 −1 −1 0
 −1 2 −1 0 
D−A=  −1 −1
.
3 −1 
0 0 −1 1
(b) A cofactor of the matrix D − A is
 
 2 −1 0 

(−1)1+1  −1 3 −1  .
 0 −1 1 
(c) The matrix C described in the proof of Theorem 3.26 is
 
1 −1 0 0
 −1 0 1 0 
C= .
 0 1 −1 1 
0 0 0 −1
(d) The matrix C3 described in the proof of Theorem 2.25 is
 
1 −1 0 0
C3 =  −1 0 1 0 .
0 0 0 −1
(e) The matrix C3 · C3t is
 
  1 −1 0  
1 −1 0 0  −1 2 −1 0
 −1 0 0 
 =  −1
0 1 0 · 0 2 1 .
1 0 
0 0 0 −1 0 0 1
0 0 −1
The determinant of C3 · C3t is 3.
 47

(f) Note that

   
 1 −1 0   1 −1 0 
t
   
det(C3 · C3 ) =  −1 0 1  ·  −1
  0 0 
 0 0 0   0 1 0 
   
 1 −1 0   1 −1 0 
 
+  −1 0 0  ·  −1 0 0 

 0 0 −1   0 0 −1 
   
 1 0 0   1 −1 0 
  
+  −1 1 0 · 0
  1 0 
 0 0 −1   0 0 −1 
   
 −1 0 0   −1 0 0 
  
+  0 1 0 · 0 1 0 
 
 0 0 −1   0 0 −1 
= 0 + 12 + (−1)2 + 12 = 3.

Section 3.4. The Minimum Spanning Tree Problem

67. See Figure 3.13.

u
.........
......................
..... ......
....
....... ......
3 .
.....
......
.
......
......
......
6
.....
. ......
..
.
. .
. .... ... 4 ......
......
..
............. ........
..
v .......................
... 4
............. x ... .
5
. .
......
.. ...
...................... w
... ............. .
.
. .. .. .. .. ... ............. ..
... .
.
.
.
.....
.
... ....... ..
.
...
... .. ... ...
... ... .....
... ... ... ...
... ... ... ..
... 5 .... ...
6 ..
. 3
6 ...
...
.
.. ...
...
.
..
...
... ..... ... ..
... ....
... ... 5 ... ..
.......
. ........
... ... ... ...
... ...
y z
......... ... ..........
........ ......... ... ...
3 ........... ........... ........... .......... ..........
.
..
.
......
.....
. 3 .
..
.
......
.....
3
.
..
......
...... 3................. 3.................
.
...... ..
......
..
...... ..... ..
......
............... ............... ............... ............. ..................... ............... ..... ............... ..............
4 ....................... 4 .............
... . ... ...
..
... ........
........ ... . ..........
................ .....
... ........ 4........ ..
.
............... ............. ... .............
......... ... .. . ... ..............
.. . ....
. ... .... .... ........ ..
. ....... ..
. . 3 ... .... 3 5 ... ..
.
... 3 .
... 5 ..... .. ....
.
... 3
... .. .
... ... .... ..
... .... ............ .. ............. ..... ... 5 ...
............. ............. ............. ............. .... ............... .... ............. ............. .............
... ... . ... ...

............. ............. ............. .......

..... ... .........
...........
... ... ...
3................ ... 3................
4 ..... 4 ...
....... ............. ....... .
.........
.
..
. ....... ...... 4
............. .............. ............. 5 ....... .. 5 ..... ... ... ... 5 ...... .. .............. 5 .......
....... ..
... .... ... ......... ....... ... ........ ...... ... .......... ....... ...
.......... .......
......... ......... .... ......... .........
............. ... ................ ... ............... ................ ... ................. .
..
... ... ... ... ...... ... ... ... . . . ..
..
. 3 ... 3 ... 3 ... 3 5 ....
.... ... 3
... ... ... ... .... ...
.. .. .. .. .... ..
.. .. . . .... .....
............ .............. ............... ............. ............... ............. ............ .
.............. .
....
...
... .... . ...
.........
.... ... . ... . ... .... . ...

Figure 3.13: The graphs in Exercise 67

48 CHAPTER 3. TREES

68. See Figure 3.14.

2 ................
u .............................. .......
. v
7 .......
x........................5........
.......
.......
....... w
..............
5 ............
... .......
... ...... 11
.........
. 6 .......
.........
. .......
.......
..
...
..... 3 11 .......
.
.............. ..........
z
y ... .......
9

2 2.......... 2 .............
2 .............
2
............. ............. ............. ............. .............. ............... .........
... ... ... ....... ... . ...
....
.... . .........
....
5............. 5.........
......
. ....
............. ...
............. ...
.... ..... ............ ..... ...
............. ........... ............. 5 ...
............. ..............
...
.......
.... 5 3 ................ 5
3................. . ...
3............ .
.... 3............
.... .
.... .
.... .... .... .... 9
.............. ............. ............... ............. .............. ............... ............... .............. .............. .............
. ... . ... ... . ... . . ...

..............
2 ............. ...............
2 ............. .............
2 ............. ............. 2 ............. .............
2 .............
. ... .
....
... ... ...... ...
....
.... ...
.....
...
.... .
..... .... .....
.
..... 5
.
.... 5 ..... 5 ......
.
.. .... .
.. ....
5
............ ...
.............
...
............. ............
5 ............. ............ 5 ............. ............ ............. ............
.... ...... ..... ..... 5 .....
... ..
3............. 3................
... ..
....
............. .............. ............. ............. ............. ............. ....
............... .............
... ............... 9 .............
... . ... ... ... ... ... ... ...

Figure 3.14: The graphs in Exercise 68

69. Proof. Let T be a minimum spanning tree of G. First, suppose that T is

the unique minimum spanning tree of G. Assume, to the contrary, that there
exist edges e and f of G such that (1) e ∈ E(G) − E(T ), (2) f is on the cycle
of T + e and (3) w(f ) ≥ w(e). Then T 0 = T + e − f is a spanning tree of G
with
w(T 0 ) = w(T ) + w(e) − w(f ) ≤ w(T ),
which implies that either T 0 is another minimum spanning tree of G or T is
not a minimum spanning tree of G, a contradiction.
For the converse, assume that the weight of each edge e ∈ E(G) − E(T )
exceeds the weight of every edge on the cycle of T + e. We claim that T
is a unique minimum spanning tree, for otherwise, let T 0 6= T be another
minimum spanning tree of G. So w(T 0 ) = w(T ). Let e0 ∈ E(T 0 ) − E(T ) and
let C 0 be the cycle in T + e0 . By our assumption, w(e0 ) > w(f 0 ) for every edge
f 0 in C 0 . Thus let f 0 ∈ E(C) and so

w(T 0 ) = w(T ) + w(e0 ) − w(f 0 ) > w(T ) = w(T 0 ),

which is a contradiction.

70. Proof. Suppose that G has order n and assume, to the contrary, that there
is a minimum spanning tree T of G that cannot be obtained from Kruskal’s
algorithm. Let E(T ) = {e1 , e2 , . . . , en−1 }, where w(e1 ) ≤ w(e2 ) ≤ . . . ≤
w(en−1 ). Among all minimum spanning trees of T obtained by Kruskal’s
algorithm, let T 0 be one whose edges are selected in the order f1 , f2 , . . . , fn−1
 49

(and so w(f1 ) ≤ w(f2 ) ≤ . . . ≤ w(fn−1 )) and such that k (< n − 1) is the

maximum positive integer for which {e1 , e2 , . . . , ek } = {f1 , f2 , . . . , fk }. Thus
fk+1 is not in T .
Let e be an edge of the cycle C in T + fk+1 that does not belong to T 0 and
let T0 = T + fk+1 − e. Hence T0 is a spanning tree of G and so w(T ) ≤ w(T0 ).
Thus w(e) ≤ w(fk+1 ). Furthermore,

w(ek+1 ) ≤ w(e) ≤ w(fk+1 ) ≤ w(ek+1 ). (3.1)

Since there is equality throughout (3.1), it follows that w(ek+1 ) = w(fk+1 ).

Because e1 , e2 , . . . , ek+1 do not produce a cycle in G and ek+1 is an edge of
minimum weight in E(G) − {e1 , e2 , . . . , ek }, there is a minimum spanning tree
obtained by Kruskal’s algorithm that contains e1 , e2 , . . . , ek+1 . This produces
a contradiction.
71. By Kruskal’s algorithm, there is only one choice at each step for the edge
selected. That the minimum spanning tree obtained is unique follows from
Exercise 70
72. Let G be a connected weighted graph, where each edge e of G is assigned a
weight w(e). Let a = max{w(e) : e ∈ E(G)}. For each e ∈ E(G), define a
new weight of e by w0 (e) = a + 1 − w(e). This produces a new connected
weighted graph G0 . We then use Kruskal’s algorithm to obtain a minimum
spanning tree T 0 of G0 . Replacing w0 (e) by w(e) for each edge e of T 0 results
in a maximum spanning tree T of G.
50 CHAPTER 3. TREES
Chapter 4

Connectivity

Section 4.1. Connectivity and Edge-Connectivity

Pk
1. Let G = Kn1 ,n2 ,···,nk with n1 ≤ n2 ≤ · · · ≤ nk and n = i=1 ni . Let Vi be a
partite set of G with |Vi | = ni (1 ≤ i ≤ k).
Claim: κ(G) = λ(G) = δ(G) = n − nk .
Proof. By Theorem 4.4, κ(G) ≤ λ(G) ≤ δ(G) = n − nk . Let S be a set of
vertices with |S| < δ(G). Then there exist u, v ∈ V (G) − S such that u ∈ Vi
and v ∈ Vj , where i 6= j. Furthermore, uv ∈ E(G − S) and each vertex of
G − S is either adjacent to u or adjacent to v. Thus, G − S is connected. So
κ(G) = λ(G) = δ(G) = n − nk .

2. Proof. Observe that {v1 , v2 , . . . , vk } is a vertex-cut of H. Thus κ(H) ≤ k.

Let W be a set of k − 1 vertices of H. We show that H − W is connected.
We consider two cases.

Case 1. w ∈ W . Then H − W = G − (W − {w}). Since G is k-connected and

|W − {w}| = k − 2, it follows that G − (W − {w}) is connected and so H − W
is connected.

Case 2. w ∈ / W . Since G is k-connected, G − W is connected. Since |W | =

k − 1, it follows that H − W is obtained from G − W by adding a new vertex
w and joining w to a vertex of G − W . Therefore, H − W is connected.

3. (a) Proof. Let H = G ∨ K1 , where V (H) − V (G) = {v}. We show that

κ(H) ≥ k + 1. Let S be a set of vertices of H with |S| = k. There are
two cases.
Case 1. v ∈ / S. Since every vertex in G is adjacent to v in H, every
vertex in G − S is adjacent to v in H − S and so H − S is connected.
Case 2. v ∈ S. Then H − S = G − (S − {v}). Since κ(G) ≥ k and
|S − {v}| = k − 1, it follows that G − (S − {v}) is connected.

51
52 CHAPTER 4. CONNECTIVITY

In either case, S is not a vertex-cut of H. Thus, the removal of k or fewer

vertices from H does not disconnect H and so κ(H) ≥ k + 1. Therefore,
H is (k + 1)-connected.
(b) Proof. Let H = G ∨ K1 , where V (H) − V (G) = {v}. We show that
λ(H) ≥ k + 1. Let S be a set of edges of H with |S| = k. Suppose
that ` edges of S belong to G, where 0 ≤ ` ≤ k and the remaining k − `
edges are incident with v. Let S 0 be the set of ` edges belonging to G.
If ` = k, then v is adjacent to every vertex of G in H and so H − S is
connected. If ` < k, then G − S 0 is connected. Since v is adjacent to at
least k + 1 − ` > 0 vertices of G, it follows that H − S is connected.
4. Let H = G ∨ K1 . Observe that H has diameter at most 2. By Theorem 4.7,

λ(H) = δ(H) = 1 + dn .

5. Let G be a k-connected graph and let T be a tree of order k + 1. Then

k ≤ κ(G) ≤ δ(G). By Theorem 3.20, G contains T as a subgraph.
6. (a) Proof. Since S is a minimum vertex-cut, |S| = κ(G) = k. Assume, to
the contrary, that G − S has ` components, where ` ≥ 3. Let G1 be a
component of smallest order n1 in G − S. Then
n−k n−k
n1 ≤ ≤ .
` 3
If u ∈ V (G1 ), then
degG u ≤ degG1 u + k ≤ n1 − 1 + k
n−k n + 2k − 3
≤ −1+k = ,
3 3
contrary to hypothesis.
(b) Proof. Assume, to the contrary, that G − S has ` components, where
` ≥ t + 1. Let G1 be a component of smallest order n1 in G − S. Then
n−k n−k
n1 ≤ ≤ .
` t+1
If u ∈ V (G1 ), then
degG u ≤ degG1 u + k ≤ n1 − 1 + k
n−k n + kt − t − 1
≤ −1+k = ,
t+1 t+1
contrary to hypothesis.
7. Proof. Assume, to contrary, that there exists a graph G of order n contain-
ing at least k pairwise nonadjacent vertices such that
n + (k − 1)(` − 2)
degG v ≥
k
 53

for all v ∈ V (G) but that G is not (`, k)-connected. Thus, κk (G) = t ≤ ` − 1.
Since G contains at least k pairwise nonadjacent vertices, there is a set S
of t vertices such that G − S contains at least k components. Let G1 be a
component of smallest order n1 in G − S. Thus n1 ≤ n−t k . For v ∈ V (G1 ), it
follows that
 
n−t n + kt − t − k
degG v ≤ −1 +t=
k k
n + (k − 1)(t − 1) − 1 n + (k − 1)(` − 2) − 1
= ≤ ,
k k
producing a contradiction.

8. Choose k and n such that n ≥ k + 1 and let N = (n − k + 1)/2 (so k and n are
of opposite parity). Let G = 2KN ∨ Kk−1 . Then U = V (Kk−1 ) is a vertex-cut
and so κ(G) ≤ k − 1. Let v ∈ V (2KN ). Then deg v = (N − 1) + (k − 1) =
(n + k − 3)/2. In fact, δ(G) = (n + k − 3)/2.

9. Proof. Let F1 and F2 be two copies of Kc+1 with V (F1 ) = {u1 , u2 , . . . , uc+1 }
and V (F2 ) = {v1 , v2 , . . . , vc+1 }. Let G be the graph obtained from F1 and F2
by adding the a edges ui vi (1 ≤ i ≤ a) and b − a edges joining u1 and b − a
vertices of F2 different from v1 . Then κ(G) = a, where {u1 , u2 , . . . , ua } is a
minimum vertex-cut, λ(G) = b, where the b edges between F1 and F2 form a
minimum edge-cut and δ(G) = c.

10. Claim: If G is a connected graph of order n ≥ 2, then

κ(G) + con(G) = n + 1.

Proof. Let S be a minimum vertex-cut of G. Then |S| = κ(G). Since G − S

is an induced disconnected subgraph of order n − κ(G) in G, it follows that
con(G) ≥ n − κ(G) + 1 or κ(G) + con(G) ≥ n + 1.
Let H be an induced disconnected subgraph of order con(G) − 1 and let
S = V (G) − V (H). Since H = G − S is disconnected, S is a vertex-cut of G
and so

κ(G) ≤ |S| = n − (con(G) − 1) = n − con(G) + 1,

or κ(G) + con(G) ≤ n + 1. Combining the two inequalities gives us the desired

result.

11. Proof. Let k = 2`, where ` ≥ 1. Any k-connected graph of order n has
minimum degree at least k and so has size at least kn/2. We show that
there exists a k-connected graph G of order n and size kn/2. Let V (G) =
{v1 , v2 , . . . , vn } where vi is adjacent to the vertices vi±1 , vi±2 , . . . , vi±` . Thus,
G is a k-regular graph of order n and size kn/2. It remains to show that
κ(G) = k. Assume, to the contrary, that κ(G) < k. Then there exists a
54 CHAPTER 4. CONNECTIVITY

set S of k − 1 vertices of G such that G − S is disconnected. Let v1 and vr

be vertices in two components of G − S. Let S = {vi1 , vi2 , . . . , vik−1 }, where
1 < i1 < i2 < · · · < ik−1 ≤ n. Since S contains k − 1 = 2` − 1 vertices,
at most ` − 1 of the integers i1 , i2 , . . . , ik−1 are less than r or at most ` − 1
of these integers exceed r, say the former. Let j1 = 1. There is a vertex
vj2 ∈/ S such that 2 ≤ j2 ≤ ` + 1. Also, there is a vertex vj3 ∈ / S such that
j2 +1 ≤ j3 ≤ j2 +`. Continuing in this manner, there are integers j1 , j2 , . . . , js
with 1 = j1 < j2 < . . . < js = r such that ji+1 − ji ≤ ` and vji ∈ V (G − S) for
1 ≤ i ≤ s − 1. Since vji vji+1 ∈ E(G − S), the path (v1 = vj1 , vj2 , . . . , vjs = vr )
is a v1 − vr path in G − S, producing a contradiction.
12. The statement is false. For example, let G = K1 ∨ (K1 + K2 ), where v ∈ V (G)
with deg v = 1. Then κ(G) = 1 and κ(G − v) = 2.
13. (a) Proof. Let W = {w1 , w2 , . . . , wk−2 } be a set of k − 2 vertices of G − e.
We show that (G − e) − W is connected. First, assume that e is incident
with some vertex of W . Then (G − e) − W = G − W . Since G is k-
connected and |W | = k − 2, it follows that G − W is connected. Thus,
we may assume that e is not incident with any vertex of W . Let x, y ∈
V (G) − W . We show that x and y are connected in (G − e) − W . There
are two cases.
Case 1. e = xy. Since G contains at least k + 1 vertices, there exists
z ∈ V (G) − (W ∪ {x, y}). Since G is k-connected, G − (W ∪ {y}) and
G − (W ∪ {x}) are both connected. Thus there is an x − z path P in
G − (W ∪ {y}) and a z − y path Q in G − (W ∪ {x}). Note that P and Q
do not contain e and so are paths in G − e − W . Then the walk obtained
by following P by Q is an x − y walk in G − e − W and so G − e − W
contains an x − y path by Theorem 2.1.
Case 2. e 6= xy. So e is either not incident with x or not incident with
y, say e is not incident with x. Suppose that e is incident with a vertex
u 6= x. Since G is k-connected, G − (W ∪ {u}) is connected and so there
is an x − y path P in G − (W ∪ {u}). This implies that P is also an x − y
path in G − W . Since P does not contain u, it follows that P does not
contain e. Therefore, P is an x − y path in G − e − W .
(b) Proof. Let X = {e1 , e2 , . . . , ek−2 } be a set of k − 2 edges of G − e.
Since |X ∪ {e}| = k − 1, the graph G − e − X is connected.
14. (a) If G is a 0-regular graph, then G = K n for some positive integer n. Then
κ(G) = λ(G) = 0.
(b) Let G be a 1-regular graph. If G is disconnected, then κ(G) = λ(G) = 0.
If G is connected, then G = K2 and so κ(G) = λ(G) = 1.
(c) If G is disconnected, then κ(G) = λ(G) = 0. If G is connected, then
G = Cn , where n ≥ 3. Thus, κ(G) = λ(G) = 2.
(d) The graph G shown in Figure 4.1(a) is 4-regular with κ(G) = 1 and
λ(G) = 2. Thus r = 4.
 55

(e) The graph G shown in Figure 4.1(b) is 4-regular with κ(G) = 2 and
λ(G) = 4. Thus r = 4.

........ ... ................ ...........

.................................... ..... ... ................. ........................................................................................... ................................ ...
........ ........... ......... .... .. .
..... ..... .......... .. ........ .... ........ ........................................ ......... ........ ....
. ...... ...
.... ................................. ...................................... . . ... .... ............. ........ .. ................ ............... .......
... . .. .
....... .......... . .......................
...
.... ..... .... ................................ ....... ...........................
...
... ............. ........................ ..............
................................ . . ... ..... . . .
.......... ........................................
...
...............................
.
..........
..... ........................................ ................................ .................. .................................... .
.. .
..... ..
..... ..... . .. ........ .................... . . ..... ........ .. ............. ........
..... ......... ............................ ............. ..... ......... ................................... .............. .............. ... ........... ..
....................... .......... ............
...........
... ........ ...
........ .......
. ..
......... ................................... ... ................
.........
. ...... ......................................... ..... ........................................
...... ......

(a) (b)

Figure 4.1: The graphs in Exercises 14(d) and 14(e)

15. Claim: (1) κ(G) ≥ κ(G), (2) λ(G) ≥ λ(G), and (3) λ(G) ≥ κ(G).
Proof. Observe that (1) and (2) are immediate consequences of the defi-
nitions. For (3), let H be a subgraph of G such that κ(H) = κ(G). Then
λ(G) ≥ λ(H) ≥ κ(H) = κ(G).

16. Claim: max{κ(G) : G ∈ G} = k.

Proof. Let G ∈ G and let E = [V (G1 ), V (G2 )], so |E| = k. Let W be the
set of vertices of G1 incident with at least one edge of E. Then G − W is
disconnected and |W | ≤ k. Thus, max{κ(G) : G ∈ G} ≤ k.
Since G1 and G2 are k-connected, each has order at least k+1. Let V (G1 ) =
{u1 , u2 , . . . , un } and V (G2 ) = {v1 , v2 , . . . , vn }. Suppose that E = {u1 v1 , u2 v2 ,
· · ·, uk vk } is the set of k edges added to G1 and G2 , producing the graph
H. Let U = {u1 , u2 , · · ·, uk } and V = {v1 , v2 , · · · , vk }. Since H − U is
disconnected, it follows that κ(H) ≤ k. We show that κ(H) = k in this case.
Let S be a subset of V (H) of cardinality k − 1 and let Si = S ∩ V (Gi ) for
i = 1, 2. Since Gi (i = 1, 2) is k-connected and |Si | < k for i = 1, 2, it follows
that Gi − Si is connected. Furthermore, there exists j (1 ≤ j ≤ k) such
that uj , vj ∈ / S. Thus, uj , vj and uj vj are present in H − S and so H − S is
connected. Hence, S is not a vertex-cut and so κ(H) > k − 1, implying that
κ(H) = k.

Section 4.2. Theorems of Menger and Whitney

17. Proof. Let W be a set of k − 1 vertices of H. We show that H − W is
connected. We consider two cases.
Case 1. w ∈ W . Then H − W = G − (W − {w}). Since G is k-connected and
|W − {w}| = k − 2, it follows that G − (W − {w}) is connected and so H − W
is connected.
Case 2. w ∈ / W . Since G is k-connected, G − W is connected. Since |W | =
k − 1, it follows that S − W 6= ∅. Thus H − W is obtained from G − W by
adding a new vertex w and joining w to the vertices of S − W . Therefore,
H − W is connected.
56 CHAPTER 4. CONNECTIVITY

18. Proof. Construct a graph H from G by adding a new vertex v and joining
v to each of v1 , v2 , . . . , vt . It is a consequence of Corollary 4.12 that H is
t-connected. By Theorem 4.11, H contains t internally disjoint u − v paths
Q1 , Q2 , . . . , Qt . Then Qi − v is a u − vi path for each i (1 ≤ i ≤ t), every two
paths of which have only u in common.

19. Proof. Let G be a k-connected graph of order n ≥ 2k and let the graph H
be obtained from G by adding two vertices v1 and v2 and joining vi to the
vertices of Vi for i = 1, 2. By applying Corollary 4.12 twice, it follows that
H is k-connected. Thus H contains k internally disjoint v1 − v2 paths. This
produces k disjoint paths connecting V1 and V2 .
For the converse, let G be a graph of order n ≥ 2k that is not k-connected.
Then G contains a vertex-cut S with |S| = k − 1. Let G1 be a component of
G − S of minimum order n1 . If n1 ≥ k, let V1 be a set of k vertices of G1
and let V2 be a set of k vertices of G − S, none of which belong to G1 . Then
there do not exist k disjoint paths connecting V1 and V2 since all such paths
must pass through S. If n1 < k, then define V1 to be the union of V (G1 ) and
k − n1 vertices of S. Let V2 be the union of the remaining n1 − 1 vertices of
S and k − n1 + 1 vertices of G − S not belonging to G1 . Here too, there do
not exist k disjoint paths connecting V1 and V2 .

20. Proof. Since G is k-connected, δ(G) ≥ k. By t applications of Corol-

lary 4.12, it follows that Gt is k-connected.

21. Proof. Let u ∈ S1 and v ∈ S2 . Then there exist k internally disjoint u − v

paths P10 , P20 , . . . , Pk0 in G. Let ui be the last vertex of Pi0 belonging to S1 and
let vi be the first vertex following ui on Pi0 that belongs to S2 . For 1 ≤ i ≤ k,
let Pi be the ui − vi subpath of Pi0 . Then the paths P1 , P2 , . . . , Pk have the
desired properties.

22. Proof. Since G is k-connected, G − v is (k − 1)-connected, where k − 1 ≥ 2.

Therefore, by Theorem 4.15 the vertices v1 , v2 , · · · , vk−1 lie on a common cycle
C of G − v. Since G is k-connected, G contains k − 1 internally disjoint v − vi
paths Qi (1 ≤ i ≤ k − 1). Let ui be the first vertex of Qi on C and denote
the v − ui subpath of Qi by Pi .

23. Proof. Assume first that G is k-edge-connected. Thus λ(G) ≥ k and so the
removal of any fewer than k edges of G results in a connected graph. Hence, for
every two vertices u and v, the minimum number of edges that separate u and
v is at least k. By Theorem 4.17, G contains at least k pairwise edge-disjoint
u − v paths.
For the converse, suppose that G contains k pairwise edge-disjoint u − v
paths for each pair u, v of distinct vertices of G. Hence, for every pair u, v of
distinct vertices of G, the maximum number of pairwise edge-disjoint u − v
paths in G is at least k. By Theorem 4.17, the minimum number of edges that
separate u and v is at least k. Thus λ(G) ≥ k and so G is k-edge-connected.
 57

24. The statement is false. Let G = K1 ∨ 2K2 . Then G is 2-edge-connected. Let

v1 be the cut-vertex of G, and let v and v2 be two nonadjacent vertices of G.

25. The statement is true. Proof. Let u ∈ S1 and v ∈ S2 . Then there exist k
edge-disjoint u − v paths Q1 , Q2 , · · · , Qk . For each i = 1, 2, · · · , k, let ui be
the last vertex of Qi that belongs to S1 and let vi be the first vertex of S2
that belongs to Qi after ui is encountered. Let Pi be the ui − vi subpath of
Qi .

26. Proof. Since Qn is n-regular, δ(Qn ) = n and so κ(Qn ) ≤ λ(Qn ) ≤ n. Thus,

it suffices to show that κ(Qn ) ≥ n. We show that for every two vertices u
and v of Qn , there are n internally disjoint u − v paths. We verify this by
induction. This is clearly true for n = 1 and n = 2.
Assume that every two vertices of Qn−1 are connected by n − 1 internally
disjoint paths, where n − 1 ≥ 2. Suppose that Qn is constructed from two
copies H and H 0 of Qn−1 , where V (H) = {v1 , v2 , . . . , v2n−1 } and V (H 0 ) =
{v10 , v20 , . . . , v20 n−1 }, such that vi vi0 ∈ E(Qn ). Let u and v be two vertices of
Qn . We consider two cases.

Case 1. u, v ∈ V (H) or u, v ∈ V (H 0 ), say the former. By the induction

hypothesis, H contains n − 1 internally disjoint u − v paths P1 , P2 , . . . , Pn−1 .
Let P 0 be a u0 − v 0 path in H 0 . Define Pn = (u, P 0 , v). Then {P1 , P2 , . . . , Pn }
is a collection of n internally disjoint u − v paths in Qn .

Case 2. u ∈ V (H) and v ∈ V (H 0 ). Then v = w0 for some w ∈ V (H). We

consider two subcases.

Subcase 2.1. w 6= u. By the induction hypothesis, H contains n − 1 internally

disjoint u−w paths P10 , P20 , . . . , Pn−1
0
. Each of these paths necessarily contains
at least one neighbor of w in H. Suppose that the next-to-last vertex of Pi0
(1 ≤ i ≤ n − 1) is xi . For 1 ≤ i ≤ n − 2, let Pi = (Pi0 − w, x0i , w0 ). Let
0
Pn−1 = (Pn−1 , w0 = v). Let Pn−1 00
be the u0 − w0 path in H 0 corresponding
0 00
to Pn−1 and let Pn = (u, Pn−1 ). Then {P1 , P2 , . . . , Pn } is a collection of n
internally disjoint u − v paths in Qn .

Subcase 2.2. w = u. Then we show that Qn contains n internally disjoint

u − u0 paths. Let x ∈ V (H) such that x 6= u. Then H contains n − 1
internally disjoint u − x paths P10 , P20 , . . . , Pn−1
0
. For each i with 1 ≤ i ≤ n − 1,
let xi ∈ V (H) such that xi is adjacent to x on Pi0 and let Qi be the u − xi
subpath of Pi0 . Now for each i with 1 ≤ i ≤ n − 1, let Q∗i be the x0i − u path
in H 0 corresponding to Qi (by reversing the order of the vertices on Qi ). For
1 ≤ i ≤ n − 1, define Pi = (Qi , Q∗i ) and Pn = (u, v). Then {P1 , P2 , . . . , Pn } is
a collection of n internally disjoint u − u0 paths in Qn .

27. In the proof of Theorem 4.17, there is a step where each edge of S is sub-
divided, introducing k new vertices w1 , w2 , . . . , wk , which are then identified
producing a new vertex w. The resulting structure may be a multigraph (not
58 CHAPTER 4. CONNECTIVITY

a graph) and it need not occur that there are k pairwise edge-disjoint u − w
paths in the underlying graph of H and k pairwise edge-disjoint w − v paths
in the underlying graph of H. See Figure 4.2.
u1 =
.....
u2 u1 =
.......
u2 u1 =
......
u2
.... .... . . .... ....
..... ........... ..
... ... ... ....
... .... ... ...
...
. ... .... ...
... ... ........... .. ... ...........
. ... ...... .. ...
e 1 ... .
.
..
...
...
...
... 2 e w t 1.....
.
...
.
. ...
... tw
... 2
......
... ..tw
..... .........
... ... ... ..... .....
..
.
. ...
... .
.
...
...
... ...
...... .....
.....
... ... .
. . ..... .....
.
.
.. ... .
...
.
...
..
......
. .....
...........
...... ... ...... ...
. ........
...... ............ ............. ............. ....
.. .
.
........ ............
... . .
v1 v2 v1 v2 v1 v2
in G in H

Figure 4.2: The graphs in Exercise 27

28. Proof. Suppose that κ(G) = k. Let u and v be two vertices of G such that
d(u, v) = diam(G). Since G has connectivity k, it follows that G contains k
internally disjoint u − v paths P1 , P2 , . . . , Pk . Since d(u, v) = diam(G), each
path Pi (1 ≤ i ≤ k) contains at least diam(G) − 1 vertices different from u
and v. Thus n ≥ k(diam(G) − 1) + 2.
29. Proof. Let u and v be two adjacent vertices of a 3-connected graph G. Then
there are three internally disjoint u − v paths P1 , P2 , P3 , where say the lengths
of P1 and P2 are at least 2. Then P1 and P2 together with the edge uv form
a chorded cycle in G. This need not be the case for a 2-connected graph. For
example, the graph G = Cn for n ≥ 4 contains no chorded cycle.
30. (a) Proof. Suppose that the statement is false. Then there is a 3-connected
graph G and two vertices u and v such that all paths in every set of three
internally disjoint u−v paths have the same length. Let P1 , P2 and P3 be
three internally disjoint u − v paths of length a, where then a ≥ 2. Since
G − {u, v} is connected, G − {u, v} contains an x − y path P connecting
two of the three paths P1 , P2 and P3 such that x, say, belongs to P1 , y
belongs to P2 and no other vertex of P belongs to P1 , P2 or P3 . Suppose
that the length of P is b ≥ 1, the length of the u − x subpath of P1 is
k and the length of the u − y subpath of P2 is `. Then there are u − v
paths of lengths k + b + a − ` and ` + b + a − k internally disjoint with
P3 . Hence, k + b + a − ` = ` + b + a − k = a, which implies that b = 0, a
contradiction.
(b) Let u and v be any two antipodal vertices of Cn , where n ≥ 4 is even.
31. (a) Proof. By Whitney’s theorem (Theorem 4.11), κ(G) = k. Let S be a
vertex-cut of G with |S| = k. Since n ≥ k +3, it follows that G−S has at
least three vertices. Since G − S is disconnected, there are vertices u and
v belonging to different components of G−S. Let w ∈ V (G)−(S∪{u, v}).
Then u and v belong to different components of G − (S ∪ {w}) and so
S ∪ {w} is a vertex-cut with |S ∪ {w}| = k + 1.
 59

(b) Proof. Let u and v be two vertices of G with d(u, v) = diam G = k.

Since G is k-connected, there exist k internally disjoint v − u paths Qi
(i = 1, 2, · · · , k). Since d(u, v) = k, each path Qi has length k or more.
For each i with 1 ≤ i ≤ k, let vi be a vertex on Qi such that the v − vi
subpath Pi has length i. This gives the desired result.

32. Proof. Assume, to the contrary, that there exists a graph G of order n and
an integer k with 1 ≤ k < n such that deg u + deg v ≥ n + k − 2 for every
two nonadjacent vertices u and v of G but G is not k-connected. Thus, there
is a set S of k − 1 vertices of G such that G − S is disconnected. Let G1
and G2 be two components of G − S such that v1 ∈ V (G1 ) and v2 ∈ V (G2 ),
where |V (Gi )| = ni for i = 1, 2. Then v1 and v2 are not adjacent in G and
deg vi ≤ (ni − 1) + (k − 1) = ni + k − 2 for i = 1, 2. Since n1 + n2 ≤ n − k + 1,
it follows that deg v1 + deg v2 ≤ (n1 − 1) + (n2 − 1) + 2(k − 1) ≤ n + k − 3,
which is a contradiction.
60 CHAPTER 4. CONNECTIVITY
Chapter 5

Eulerian Graphs

Section 5.1. The Königsberg Bridge Problem

1. Yes. Since the resulting multigraph has exactly two odd vertices A and B, it
has an Eulerian trail, starting at A or B and terminating at the other vertex.

2. As observed, at least two of the letters A, B, C, D are neither the first nor
the last term of a sequence of eight letters that represents a route that crosses
each bridge exactly once. Suppose, without loss of generality, that B is such
a letter. Then B occurs only as an interior term in the sequence. Since each
such occurrence of B represents two bridges that are crossed and there are
three bridges that lead into and out of B, such a route is impossible.

3. In this case, 14 bridges would need to be crossed (each bridge crossed twice)
and, if this is possible, such a route could be represented by a sequence of 15
letters. This is possible and one such route is: A, B, A, B, A, C, A, C, A, D,
B, D, C D, A.

4. Six. There is a boat ride that can go under all bridges except bridge e. If a
boat ride exists that goes under f and g, then it must start by going under f
or under g and terminate under the other. If, in addition, it travels under e,
then it can travel under e, f and g only.

Section 5.2. Eulerian Circuits and Trails

5. Proof. Suppose that F is an r1 -regular graph of order n1 and H is an r2 -
regular graph of order n2 . Then r1 and r2 are both odd and so n1 and n2 are
both even. In the graph G, degG x = r1 +1 for each x ∈ V (F ), degG x = r2 +1
for each x ∈ V (H) and degG v = n1 + n2 . Thus, each vertex of G is even.
Since G is connected, G is Eulerian.

6. The statement is true.

61
62 CHAPTER 5. EULERIAN GRAPHS

Proof. Since G has neither an Eulerian circuit nor an Eulerian trail, G con-
tains 2k odd vertices, where k ≥ 2. Observe that degH v = 2k and all other
vertices of H are also even. Since H is connected, H is Eulerian.
7. The graph G  H can be constructed by replacing each vertex v of G by a
copy Hv of H. Let x be a vertex of G  H. Then x belongs to Hv for some
vertex v of G. The vertex x is adjacent to its neighbors in Hv as well as to
one vertex in Hu for every neighbor u of v in G. Thus

degG  H x = degHv x + degG v.

Hence, degG  H x is even if and only if degHv x and degG v are both even or
both odd (that is, they are of the same parity).
If degHv x is even, then degG v is even for every vertex v of G; while if degHv x
is odd, then degG v is odd for every vertex v of G. Since G  H is connected,
it follows that G  H is Eulerian if and only if both G and H are Eulerian
or every vertex of G and H is odd.
8. Since deg v ≥ 1, it follows that deg u − deg v ≤ n − 2. Thus deg u − deg v =
n − 2, which implies that deg u = n − 1 and deg v = 1. Since deg u is odd,
n = deg u + 1 is even.
9. No. Suppose that G is not Eulerian. Then r is odd and so n is even. Thus G
is (n − 1 − r)-regular. Since (n − 1) − r is even, G is Eulerian.
10. Proof. Suppose that the order of T is n ≥ 3. Since every tree has at least
two leaves, there is a vertex v of T with deg v = 1. Furthermore, there is a
vertex u of T with deg u = 2. Thus,

degG v = n − 2 and degG u = n − 3.

One of the two integers n − 2 and n − 3 is even and the other is odd. Hence,
T contains a vertex of odd degree and so T is not Eulerian.
11. Proof. For each vertex v in G, either deg v ≡ 0 (mod 4) or deg v ≡
2 (mod 4). Let V = V1 ∪ V2 such that each vertex v in V1 has deg v ≡
0 (mod 4) while each vertex v in V2 has deg v ≡ 2 (mod 4). Suppose that
V1 = {u1 , u2 , · · · , ua } and V2 = {v1 , v2 , · · · , vb }. Then deg ui = 2ki where ki
is even for 1 ≤ i ≤ a, while deg vj = 2`j where `j is odd for 1 ≤ j ≤ b. Let m
be the size of G. By the First Theorem of Graph Theory (Theorem 1.4),

X X X a
X b
X
2m = deg v = deg v + deg v = 2 ki + 2 `j .
v∈V v∈V1 v∈V2 i=1 j=1

This implies that

a
X b
X
ki + `j = m. (5.1)
i=1 j=1
 63

Pa Pb
(Note that if V1 = ∅, then let i=1 ki = 0; while if V2 = ∅, then let j=1 `j =
0.)
By (5.1),
b
X a
X
`j = m − ki .
j=1 i=1

Thus, b is even if and only if m is even.

12. The statement is true. Proof. Let G be a PEulerian bipartite graph of size m
and let U be a partite set of G. Since m = u∈U deg u and each vertex of G
is even, m is even.

13. Both statements are false.

(a) Let G be a graph obtained from two copies F and H of C3 by joining a

vertex u in F to a vertex v in H.
(b) Any u − v trail must contain an odd number of edges incident with u
and an odd number of edges incident with v. So, if G contains two edge-
disjoint u − v trails, then there is at least one edge incident with u on
neither trail and at least one edge incident with v on neither trail.

14. Proof. Let P be a u − v geodesic in G. Thus, P contains d(u, v) edges and

G − E(P ) contains m − d(u, v) edges. Since u and v have degree 1 on P and
every other vertex of P has degree 2 on P , every vertex of G − E(P ) has
even degree in G − E(P ). Thus, every nontrivial component of G − E(P )
is Eulerian. By Theorem 5.4 and the fact that m − d(u, v) > n, the graph
G − E(P ) contains two or more cycles such that every edge of this graph lies
on exactly one of these cycles. Therefore, P and these cycles have the desired
property.

15. (a) Since every vertex has odd degree, n is even and at least one edge inci-
dent with each vertex is not in any Eulerian subgraph of G. Thus, the
maximum size of an Eulerian subgraph of G is m − n/2.
(b) The Petersen graph P has order n = 10 and size m = 15. Since the
circumference of P is 9, it follows that P does not contain an Eulerian
subgraph (a cycle in this case) of size m − n/2 = 10.

16. (a) Proof. Let G be an Eulerian graph of odd order n = 2k + 1 ≥ 3. If

G contains at most two vertices of the same degree, then G contains at
most two vertices of each of the degrees 2, 4, 6, . . . , 2k. However then, G
contains at most n − 1 vertices, which is impossible.
(b) Proof. First, observe that G1 = K3 is the only Eulerian graph of order
3 and it contains three vertices of the same degree. Assume, to the
contrary, that the statement is false. Then there exists a smallest odd
integer n ≥ 5 for which the statement is false. Then n = 2k + 1 for
some integer k ≥ 2. Consequently, there is a unique Eulerian graph
64 CHAPTER 5. EULERIAN GRAPHS

Gk−1 of order 2k − 1 containing three vertices of one of the degrees

2, 4, . . . , 2k−2 and two vertices of every other degree. Then Gk = Gk−1 ∨
K2 is an Eulerian graph of order 2k + 1 containing three vertices of one
of the degrees 2, 4, . . . , 2k and two vertices of every other degree. By
assumption, there is an Eulerian graph Hk with this property such that
Gk 6∼= Hk . First, observe that Gk has two vertices x and y of degree
2k, for if Gk had three vertices of degree 2k, then Gk has no vertices
of degree 2. Thus, x and y are adjacent to each other and to all other
vertices of Gk . Therefore, Gk = H ∨ K2 for some graph H of order
2k − 1. Since the degree of every vertex of H is 2 less than its degree
in Gk , H is an Eulerian graph of order 2k − 1 containing exactly three
vertices of the same degree and at most two vertices of any other degree.
By assumption, H is unique and so H is unique. Since Gk = H ∨ K2 , it
follows that Gk ∼ = Hk .
17. (a) In G, every edge belongs to two 3-cycles and two 4-cycles or four cycles
in all. In H, every edge belongs to three 3-cycles, six 4-cycles and six
5-cycles, or 15 cycles in all.
(b) By Theorem 5.5, the graph G is not Eulerian and H is Eulerian.
Chapter 6

Hamiltonian Graphs

Section 6.2. Sufficient Conditions for Hamiltonian Graphs

1. (a) A Hamiltonian cycle of this graph is shown (with solid edges) in Fig-
ure 6.1. (Also, one could follow the consonants in Figure 6.2 (in Sec-
tion 6.1 in the text) in alphabetical order to construct a different Hamil-
tonian cycle.)

R
qq...q.............
q q qqqqqqqq ..... .....................
qqqqqq ..........qqqqqq
qqqqqq Z...........................Q
......
......
qqqqqqqq... P....................
..

qq qqq q ..q.....qq
q ....q...........
qq q q q
qqqq .........qqqqqq q...q..q...qq........ .............q...qq ..................... .......................
. . . ......
...

...qq .....q..q q
.............
qqq qqq G....q.q B .q.q.q.. q q C q
........
qq
......

qqq .q..q.........................qqqq qqq....q.....................qq qq

qqq ......q...q qqq.q..q......qqqq
qqq qqq ......
q q... qqq
.... .....

q
qqq .....q...... .......... ..........q q q
qqq .......... ...... ............ qqq
qq...........
.
...... q
... ....q
...
.... ....
.. . ......

Figure 6.1: A Hamiltonian cycle in the graph of the dodecahedron

(b) The path (R, Q, P, C, B, G) does not lie on a Hamiltonian cycle since
there is no way to enter and exit vertex Z.

2. Proof. Suppose that G is a graph containing a vertex v such that v is

adjacent to three vertices v1 , v2 , v3 , all of degree 2. Then any Hamiltonian
cycle C must contain the edges vv1 , vv2 and vv3 . However then, degC v ≥ 3,
which is impossible.

3. (a) By Theorem 2.6, a bipartite graph contains no odd cycles. Since G

has odd order, G does not have a Hamiltonian cycle and so G is not
Hamiltonian.
(b) The Herschel graph is a bipartite graph of odd order.

65
66 CHAPTER 6. HAMILTONIAN GRAPHS

4. (a) Proof. Suppose that C 0 = (u1 , u2 , . . . , un , u1 ) is a Hamiltonian cycle of

G and C 00 = (v1 , v2 , . . . , vp , v1 ) is a Hamiltonian cycle of H. The graph
G  H can be constructed by replacing each vertex ui (1 ≤ i ≤ n) of G
by a copy Hi of H, where Ci00 = (vi,1 , vi,2 , . . . , vi,p , vi,1 ) is a Hamiltonian
cycle of Hi such that vj,t and vk,t are adjacent in G  H if uj uk ∈ E(G).
Let Pi be the vi,1 − vi,3 path on Ci00 not containing vi,2 and let Pi0 be the
vi,3 − vi,1 path on Ci00 not containing vi,2 . Then
P1 , P20 , P3 , P40 , . . . , Pn−1 , Pn0 , vn,2 , vn−1,2 , . . . , v1,2 , v1,1
is a Hamiltonian cycle of G  H if n is even and
P1 , P20 , P3 , P40 , . . . , Pn−1
0
, Pn , vn,2 , vn−1,2 , . . . , v1,2 , v1,1
is a Hamiltonian cycle of G  H if n is odd.
(b) Proof. We proceed by induction on n ≥ 2. First, it is clear that Q2
and Q3 are Hamiltonian. For an integer n ≥ 4, assume that Qk is
Hamiltonian for every integer k with 2 ≤ k < n. Since Qn−2 and Q2
are Hamiltonian and Qn = Qn−2  Q2 , it then follows by (a) that Qn is
Hamiltonian.

5. For n = 2k + 1, the graph G = K1 ∨ 2Kk has order n. Then σ2 (G) = n − 1

but G is not Hamiltonian since G has a cut-vertex.

6. (a) Proof. By Corollary 6.2, G is Hamiltonian. Let C be a Hamiltonian

cycle of G. Then v1 lies on C. As we proceed around C, we either
encounter the vertices in the order v1 , v2 , v3 or v1 , v3 , v2 . In the latter
case, proceed around C in the reverse direction.
(b) The graph G in Figure 6.2 of order 8 with δ(G) = 4 contains no Hamil-
tonian cycle encountering the vertices v1 , v2 , v3 , v4 in this order.
...... ......
.... ....... ........ ......
..... .......... ........ ........ ...... .........
..... .. .. ........ ........ ... .....
..
....... ..... ...... .............. v 1 .
. . ........... ..... .....
........ . .....
..... ... ... ......................... ... .....
..
....... ... ...
.. ... ........ ..
. . . .
......
. ............... ..... .....
......
. ...
........... ..
............... ... . ...... ................
.. ........
.....
.....
... ... ............ ... ... ..... .. ...............
...
..
....... . .
.. . ... ....
... .... ... . .
...... ... .. ... .. .
. .. ............... .........
.............................. ...................
G: v 3 ....................................
....................
...
. ..
. .
......
... ......
.
.
.
. ..
.
......
.
. ..... ..
............................
v4
.....
............................. ... ........ .................
..... .. ... .. ..................... .....
.....
..... ... .
............................................................................. ...
......
... .. ..... .. ..... .
.....
... ..... ....... .................. .....
.....
........
..... .....
.....
..... .... ..... ..............
.......
....... ........ .....
..... .. ... ........ v ....... .... .....
................. 2 ....... ........ ........
....... ....... .....
......

Figure 6.2: The graph in Exercise 6(b)

(c) Proof. Assume, to the contrary, that κ(G) ≤ 2. Then G has a vertex-
cut S = {v1 , v2 }. Let v3 and v4 be in distinct components of G − S. For
any Hamiltonian cycle C of G, either v3 or v4 must be encountered after
v1 but before v2 . Thus, there is no Hamiltonian cycle encountering the
vertices v1 , v2 , v3 , v4 in this order.

7. Since G contains a Hamiltonian path, G is connected. Let v ∈ V (G). By

hypothesis, G contains a Hamiltonian path P with initial vertex v. Since
 67

P − v is a Hamiltonian path of G − v, the vertex v is not a cut-vertex of G.

Hence the graph G − v is connected and so G is 2-connected.
By Theorem 6.9, the Petersen graph is not Hamiltonian. On the other
hand, for each vertex v of the Petersen graph, there is a Hamiltonian path
with initial vertex v.

8. (a) Proof. Let G be a graph of order n ≥ 2, where the degrees of the

vertices of G satisfy d1 ≤ d2 ≤ · · · ≤ dn . Suppose that there is no integer
k < (n + 1)/2 for which dk ≤ k − 1 and dn−k+1 ≤ n − k − 1. Thus for
every integer k < (n + 1)/2, either dk > k − 1 or dn+1−k > n − k − 1.
Let H = G ∨ K1 . Then the order of H is n0 = n + 1. Let the degrees
of the vertices of H be d01 ≤ d02 ≤ · · · ≤ d0n ≤ d0n+1 , where d0i = di + 1
(1 ≤ i ≤ n) and d0n+1 = n. Thus, for each integer k < n0 /2, either d0k > k
or d0n0 −k > n − k = n0 − k − 1. By Theorem 6.8, H is Hamiltonian. Let
C be a Hamiltonian cycle of H. Then removing the vertex v of H that
does not belong to G results in a Hamiltonian path of G.
(b) Proof. Let G be a self-complementary graph of order n, the degrees of
whose vertices satisfy d1 ≤ d2 ≤ · · · ≤ dn . Since G is self-complementary,
dk + dn+1−k = n − 1 for each k (1 ≤ k ≤ n). Therefore, for every integer
k with k < (n + 1)/2, either dk > k − 1 or dn+1−k > n − k − 1. So by
(a), G has a Hamiltonian path.

9. Proof. Assume, to the contrary, that the result is false. Then among all
counterexamples of some order 2k, let G be one of maximum size. Thus, G
is a non-Hamiltonian bipartite graph with partite sets U and W such that
|U | = |W | = k and deg v > k/2 for every vertex v of G. Since G is not
Hamiltonian, G 6= Kk,k and because of the defining property of G, if u ∈ U
and w ∈ W , where uw ∈ / E(G), then G + uw is Hamiltonian. This implies
that G has a Hamiltonian u − w path

P = (u = u1 , w1 , u2 , w2 , · · · , wk−1 , uk , wk = w),

where ui ∈ U and wi ∈ W for 1 ≤ i ≤ k. If uwi ∈ E(G), then wui ∈

/ E(G),
for otherwise, G has a Hamiltonian cycle. Then

k
|{ui ∈ U : ui w ∈
/ E(G)}| ≥ degG u > .
2
This implies that
k k
deg w < k − = ,
2 2
which is a contradiction.
Observe that the bound is sharp since for an even integer k, the graph H =
2K k , k has the property that deg v = k/2 for every vertex v of H but H is
2 2
disconnected and is therefore not Hamiltonian.
68 CHAPTER 6. HAMILTONIAN GRAPHS

10. (a) Proof. Let
G be a graph of order n and size m such that n ≥ 3 and
m ≥ n−1 2 + 2. If G = Kn , then G is Hamiltonian. So we may assume
that G 6= Kn and thus contains pairs of nonadjacent vertices. Let u and
v be two nonadjacent vertices of G. Thus, the size of the graph G − u − v
is m − deg
u − deg v. Since the order of G − u − v is n − 2, its size is at
most n−22 . Hence,
 
n−2
m − deg u − deg v ≤ .
2

Since m ≥ n−1


2 + 2, it follows that
   
n−1 n−2
+2≤m≤ + deg u + deg v.
2 2

Thus, deg u + deg v ≥ n and so σ2 (G) ≥ n. By Ore’s theorem (Theo-

rem 6.1), G is Hamiltonian.

G be a graph of order n and size m such that n ≥ 3 and
(b) Proof. Let
m ≥ n−12 +1. Let H = G∨K1 , where v is the vertex of H not belonging
to G, and let H have order n0 and size m0 . Then n0 = n + 1 and
   
0 n−1 n−1
m = m+n≥ +1+n= + (n − 1) + 2
2 2
   0 
n n −1
= +2= + 2.
2 2

By (a), H is Hamiltonian and so G = H − v has a Hamiltonian path.

11. The graph G = Kk,k+1 is not Hamiltonian, has order n = 2k + 1, κ(G) = k

and α(G) = k + 1, where κ(G) = α(G) − 1.

12. (a) Proof. Let T be an independent set with |T | = k + 1 and let S =

V (G) − T . Thus |S| = k. If G contains a Hamiltonian cycle C, then
since |S| > |T |, two consecutive vertices on C must belong to S. This is
impossible.
(b) Let G be the graph obtained from C2k = (v1 , v2 , · · · , v2k , v1 ) by joining
all pairs vi , vj of vertices by an edge if i and j are odd.

13. Proof. Let H = G ∨ K1 . Then κ(H) = κ(G) + 1 and α(H) = α(G).

Therefore, κ(H) ≥ α(H). By Theorem 6.10, H is Hamiltonian. Let C be
a Hamiltonian cycle of H. Then removing the vertex v of H that does not
belong to G results in a Hamiltonian path of G.

14. Proof. We proceed by induction on the order n ≥ 4 of a tree that is not

a star. Since P 4 = P4 , the result is true for n = 4. Assume that the result
holds for all trees of order k, where k ≥ 4. Let T be a tree of order k + 1 such
that T is not a star. Then T contains an end-vertex v such that T − v is not
 69

a star. Suppose that v is adjacent to u in T . Then T − v has a Hamiltonian

path P . Since v is adjacent to every vertex of P in T except u, it follows that
T has a Hamiltonian path.

Section 6.3. Toughness of Graphs

15. (a) Proof. Let G = Kr,2r,3r for some positive integer r. Then G has order
n = 6r and δ(G) = 3r = n/2. By Dirac’s theorem (Corollary 6.2), G is
Hamiltonian.
(b) Suppose that the partite sets of G are V1 , V2 , V3 , where |V1 | = r, |V2 | = 2r
and |V3 | = 3r + 1. Let S = V1 ∪ V2 . Then |S| = 3r and k(G − S) =
3r + 1. Since k(G − S) > |S|, it follows by Theorem 6.11 that G is not
Hamiltonian.
Pk−1
(c) Claim: The graph G is Hamiltonian if and only if nk ≤ i=1 ni .
Proof. Let V1 , V2 , . . . , Vk be the partite sets of G where |Vi | = ni for
Pk−1
1 ≤ i ≤ k. First suppose that nk > i=1 ni and let S = V1 ∪ V2 ∪ · · · ∪
Vk−1 . Since
k−1
X
k(G − S) = nk > ni = |S|,
i=1
it follows by Theorem 6.11 that G is not Hamiltonian.
Pk−1 Pk
For the converse, assume that nk ≤ i=1 ni . Then n = i=1 ni ≥
nk + nk and so nk ≤ n2 . If v ∈ Vk , then
k−1
X n n
deg v = δ(G) = ni = n − nk ≥ n − = .
i=1
2 2

By Dirac’s theorem (Corollary 6.2), G is Hamiltonian.

16. Proof. Let G be a 1-tough graph. Then |S| ≥ k(G − S) for every vertex-cut
S of G. Assume, to the contrary, that G contains a cut-vertex v. Let T = {v}.
Then k(G − T ) ≥ 2, contradicting the assumption that |T | ≥ k(G − T ).
17. (a) Let G = P5 = (v1 , v2 , v3 , v4 , v5 ) and S = {v2 , v4 }. Then k(G − S) = 3 >
|S| = 2.
(b) Theorem If G is a graph with a Hamiltonian path, then

k(G − S) ≤ |S| + 1

for every nonempty proper subset S of V (G).

Proof. Let G be a graph with a Hamiltonian path and let S be a
nonempty proper subset of V (G). Let H = G ∨ K1 and let v be the
vertex of H not belonging to G. Since G has a Hamiltonian path, H is
Hamiltonian. Let S = T ∪ {v}. By Theorem 6.11, k(H − T ) ≤ |T |. Since
H −T = H −(S ∪{v}) = (H −v)−S = G−S and |T | = |S ∪{v}| = |S|+1,
it follows that k(G − S) ≤ |S| + 1.
70 CHAPTER 6. HAMILTONIAN GRAPHS

18. Since G contains a vertex that is adjacent to three vertices of degree 2, it

follows that G is not Hamiltonian (by Exercise 2). It remains to show that
t(G) ≥ 1, that is, |S|/k(G − S) ≥ 1 for every vertex-cut S of G. Since G has
no cut-vertex, κ(G) ≥ 2. Because G has vertices of degree 2, it follows that
κ(G) ≤ 2 and so κ(G) = 2. Hence we show that

k(G − S) ≤ |S| (6.1)

for every vertex-cut S of G with |S| ≥ 2. Label the vertices of G as shown in

Figure 6.3(a). Let S be a vertex-cut of G. Obviously, (6.1) holds if |S| ≥ 4.
Thus, we may assume that |S| = 2 or |S| = 3. If u4 ∈ / S, then k(G − S) = 1
and so (6.1) holds. Next, suppose that u4 ∈ S. The graph G − u4 is shown in
Figure 6.3(b). If |S| = 2, then k(G − S) = 1 or k(G − S) = 2; while if |S| = 3,
then k(G − S) = 1, 2 or 3. Therefore, (6.1) holds.

u1 u5
............... ...............
..... ......
..... ..........
..
..... ... ..
. ... ....
..... ... ...
.... ... ....
.
...... .. ... ....... u
.
.
... .............. ..... .......... 1
.
...
. ... ...
.
...
.
.
...
. ... ... .....
.
. . u .. ... ...
.
..... ..
. 5 ...... ...
. ...
.
..... 7 .............................
u ... .. ...
..
.
.
... ..................... .......................... ...
... ........... ...............
..... ..........
. .......
. ... .......... ... ..
........ .....
.... .......... ........ ... ....... ... .
... .....
..
.
..
.
... ........ ............ . ..
.. u 4 ..... . ...........
...
.
...
... ..
.
.
...
.
..
.
u 3 u .
2 ..........
........... ........... .......... ....... . .
.
..
..................... .
...... ............... .
......... ..... .............. .....
..........
... ..
.... ...............................
. . .... .......
. .. ...
....................... u 6 ....................
.... ....... u7 u6
u3 u2
(a) (b)

Figure 6.3: Graphs in Exercise 18

19. (a) Proof. Since G has order n = 101 and δ(G) ≥ n/2, it follows by
Corollary 6.2 that G is Hamiltonian. Let C = (v1 , v2 , . . . , v101 , v1 ) be
a Hamiltonian cycle of G. It suffices to show that v1 lies on a cy-
cle of length 27. Partition the set V (C) − {v1 } into the 25 sets Si =
{vi , vi+25 , vi+50 , vi+75 }, where 2 ≤ i ≤ 26. If v1 is adjacent to three or
more vertices in any set Si , then v1 lies on a cycle of length 27; other-
wise vi is adjacent to at most two vertices in every set Si . This implies,
however, that deg v1 ≤ 50, which is a contradiction.
(b) Theorem Let k be a positive integer. If G is a graph of order 4k + 1
such that δ(G) ≥ 2k + 1, then every vertex of G lies on a cycle of length
k + 2.
Proof. Since G has order 4k + 1 and δ(G) ≥ n/2, it follows that G is
Hamiltonian. Let C = (v1 , v2 , . . . , v4k+1 , v1 ) be a Hamiltonian cycle of
G. We show that v1 lies on a cycle of length k + 2. The set V (C) − {v1 }
can be partitioned into the k sets Si = {vi , vi+k , vi+2k , vi+3k }, where
2 ≤ i ≤ k + 1. If v1 is adjacent to three or more vertices of any set Si ,
then v1 lies on a cycle of length k + 2; otherwise vi is adjacent to at most
two vertices in every set Si . This implies that deg v1 ≤ 2k, which is a
contradiction.
 71

20. Proof. To produce a disconnected graph, the vertices of two partite sets
must be removed. If we let S be such a vertex-cut, then

|S| 2r
= =2
κ(G − S) r

and so t(Kr,r,r ) ≤ 2. If U is any other vertex-cut of Kr,r,r , then |U | > 2r and

|U |
> 2.
k(G − U )

Thus, t(Kr,r,r ) = 2.
|S0 |
21. Proof. Let S0 be a vertex-cut of G such that k(G−S 0)
= t(G). Since H
is a spanning subgraph of G, it follows that S0 is also a vertex-cut of H.
Furthermore, k(H − S0 ) ≥ k(G − S0 ). Therefore,
 
|S| |S0 | |S0 |
t(H) = min ≤ ≤ = t(G),
k(H − S) k(H − S0 ) k(G − S0 )

as desired.

22. (a) Proof. Let S be an independent set of vertices of G with |S| = α(G).
Then U = V (G) − S is a vertex-cut. So

|U | n − α(G)
t(G) ≤ = .
k(G − U ) α(G)

(b) By (a), n ≥ t(G)α(G) + α(G) = α(G)(t(G) + 1).

23. Proof. Let G = Kr,s with partite sets U and W such that |U | = r and
|W | = s with r ≤ s. Since U is a vertex-cut,

|U | r
t(G) ≤ = .
k(G − U ) s
|S|
Next, let S be a vertex-cut of G such that t(G) = k(G−S) . Then |S| ≥ r and
k(G − S) ≤ s. Hence,
|S| r
t(G) = ≥ .
k(G − S) s
Thus t(G) = r/s.

24. The statement is false. For every positive integer k, the complete bipartite
graph Kk,k+1 of order 2k + 1 is k-connected graph but not Hamiltonian and
so no such constant k0 exists.
72 CHAPTER 6. HAMILTONIAN GRAPHS

Pk
25. (a) Proof. Let n = i=1 ni and let V1 , V2 , . . . , Vk be the partite sets of
G = Kn1 ,n2 ,...,nk , where |Vi | = ni for i = 1, 2, . . . , k. Since V (G) − Vk is
a vertex-cut of G,

|V (G) − Vk | n − nk
t(G) ≤ = .
|Vk | nk
|S|
Next, let S be a vertex-cut of G such that t(G) = k(G−S) . Then |S| ≥
|S| n−nk
n − nk and k(G − S) ≤ nk . Hence, t(G) = k(G−S) ≥ nk and so
t(G) = n−n
nk .
k

(b) Proof. If r = 0, then every disconnected graph has toughness 0. Hence

we may assume that r = a/b, where a and b are positive integers. Let
G = K1,1,···,1,b be a complete (a + 1)-partite graph. Then by (a), t(G) =
a/b.
26. Claim: For every tree T of order 3 or more, t(T ) = 1/∆(T ).
Proof. Let v be a vertex such that deg v = ∆(T ). Let S 0 = {v}. Then
|S 0 |/k(T − S 0 ) = 1/∆(T ). So, t(T ) ≤ 1/∆(T ). Let S be a vertex-cut of T
such that
|S|
t(T ) = .
k(T − S)
Let S = {v1 , v2 , · · · , vk }. Whenever vi is removed from a subtree of T , there
is an increase in the number of components of T − S of at most deg vi − 1. So,

|S| k
t(T ) = = Pk
k(T − S) 1 + i=1 (deg vi − 1)
k 1
≥ ≥ .
1 + k∆(T ) − k ∆(T )

Therefore, t(T ) = 1/∆(T ).

27. Proof. Assume, to the contrary, that G is Hamiltonian. Then G contains
a Hamiltonian cycle C. At least one of the five pairs {si , ti }, 1 ≤ i ≤ 5,
of vertices contains no vertex that is adjacent to either p or q on C, say
{s1 , t1 } has this property (see Figure 6.8). Then s1 u1 , s1 w1 , t1 v1 , t1 x1 ∈
E(C). Necessarily, u1 w1 , u1 x1 , v1 x1 , v1 w1 ∈ / E(C). This, however, implies
that v1 y1 , w1 y1 , u1 z1 , x1 z1 ∈ E(C), which is impossible. Thus, G is not Hamil-
tonian.
28. By Theorem 6.14, if G is a noncomplete claw-free graph, then t(G) = 12 κ(G).
Thus, the Matthews and Sumner Conjecture can be restated as follows:
Every 2-tough noncomplete claw-free graph is Hamiltonian.
 73

Section 6.4. Highly Hamiltonian Graphs

29. Proof. Let G be a graph of order n ≥ 4 and size m ≥ n−1


2 + 3. If G is
complete, then G is Hamiltonian-connected. Otherwise, let u
and v be two
nonadjacent vertices of G. Then G−u−v contains at least n−1
2 +3−deg u−
deg v edges but at most n−2


2 edges. Thus,
   
n−1 n−2
+ 3 − deg u − deg v ≤
2 2

and so deg u + deg v ≥ n + 1. Therefore, σ2 (G) ≥ n + 1, which implies by

Theorem 6.15 that G is Hamiltonian-connected.

30. Proof. Let S = {u, v} ⊆ V (G). Since G is Hamiltonian-connected, there is

a u − v Hamiltonian path P = (u = u1 , u2 , · · · , un = v) in G. This implies
that G − u − v is connected. Thus, S is not a vertex-cut. Therefore, κ(G) ≥ 3
and G is 3-connected.

31. Proof. Suppose that the statement is false. Then there exist two vertices
u and v for which there is no Hamiltonian u − v path. Consider the graph
G − v of order n − 1. Since degG−v x + degG−v y ≥ n − 1 for every pair x, y of
nonadjacent vertices of G−v, it follows that G−v contains a Hamiltonian cycle
C = (u = u1 , u2 , . . . , un−1 = w, u). Thus, P = (u = u1 , u2 , . . . , un−1 = w) is
a Hamiltonian u − w path in G − v. Since G contains no Hamiltonian u − v
path, vw ∈/ E(G).
Suppose that degG w = k. Thus w is adjacent to un−2 and k − 1 vertices
in the set {u = u1 , u2 , . . . , un−3 }. If w is adjacent to uj where 1 ≤ j ≤ n − 3,
then v is not adjacent to uj+1 ; for otherwise

Q = (u = u1 , u2 , . . . , uj , w, un−2 , un−3 , . . . uj+1 , v)

is a Hamiltonian u − v path in G, which is impossible. Therefore,

degG v ≤ (n − 2) − (k − 1),

and so degG v + degG w ≤ n − 1, a contradiction.

32. Proof. Since the theorem holds for k = 1, 2, we may assume that k ≥ 3. Let
P be u − v path of order ` where 3 ≤ ` ≤ k, say P = (u = u1 , u2 , . . . , u` = v).
Let S = {u2 , u3 , . . . , u`−1 } and let H = G − S. Then the order of H is
n0 = n − ` + 2. Since σ2 (G) ≥ n + k − 1, it follows that

σ2 (H) ≥ n + k − 1 − 2(` − 2) = n + k − 2` + 3 ≥ n − ` + 3 = n0 + 1.

Thus, H is Hamiltonian-connected and so H contains a Hamiltonian u − v

path P 0 . The paths P and P 0 form a Hamiltonian cycle of G containing P
and so he(G) ≥ k.
74 CHAPTER 6. HAMILTONIAN GRAPHS

33. (a) Proof. Let G be a graph
of order n ≥ 3 such that δ(G) ≥ rn for some
rational number r ∈ 21 , 1 and so δ(G) ≥ drne. First, suppose that


(n − 1)/n ≤ r < 1. Then δ(G) ≥ rn ≥ [(n − 1)/n]n = n − 1 and so

G = Kn . Since (2r − 1)n + 1 ≤ n, it follows that he(G) ≥ (2r − 1)n + 1
Next, suppose that 1/2 ≤ r < (n − 1)/n. Let k = d2rne − n + 1. Observe
that
1 ≤ k = 2drne − n + 1 ≤ 2(n − 1) − n + 1 = n − 1.
Since
2drne d2rne
δ(G) ≥ drne = ≥
2 2
n + (d2rne − n + 1) − 1 n+k−1
= = .
2 2
it follows by Corollary 6.17 that
he(G) ≥ k = d2rne − n + 1 ≥ 2rn − n + 1 = (2r − 1)n + 1.
(b) Proof. Let G be a graph of order n ≥ 3 for which δ(G) ≥ (n + k − 1)/2
for some integer k with 1 ≤ k < n. Let r = n+k−1
2n and so 1/2 ≤ n+k−1
2n <
1. Because δ(G) ≥ rn, it follows by Corollary 6.19 that
   
n+k−1
he(G) ≥ (2r − 1)n + 1 = 2 − 1 n + 1 = k.
2n

34. (a) Proof. Let G = Kn − e, where e = uv. It is straightforward to see that

every path of order n − 3 lies on a Hamiltonian cycle of G. On the other
hand, a path of order n − 2 in G containing neither u nor v lies on no
Hamiltonian cycle of G. Thus, he(G) = n − 3.
(b) Proof. First, let P = (u1 , u2 , . . . , un−1 ) be a path of order n − 1 in
G and let un be the vertex of G not belonging to P . Since P 0 =
(u1 , u2 , . . . , un−2 ) and P 0 = (u2 , u3 , . . . , un−1 ) are paths of order n − 2 in
G and he(G) ≥ n − 2, it follows that un−1 un ∈ E(G) and u1 un ∈ E(G),
respectively. Hence, C = (u1 , u2 , . . . , un , u1 ) is a Hamiltonian cycle of G
containing P . Therefore, he(G) ≥ n − 1. Next, let Q = (v1 , v2 , . . . , vn )
be a path of order n in G. Since Q0 = (v1 , v2 , . . . , vn−1 ) is a path of
order n − 1 in G, it follows that vn−1 vn , v1 vn ∈ E(G) and so C 0 =
(v1 , v2 , . . . , vn , v1 ) is a Hamiltonian cycle of G containing Q. Thus,
he(G) = n.

35. The graph G of Figure 6.4 is pancyclic, but there is no u − u0 path of length
2. So G is not panconnected.

36. Proof. If there is a bipartite graph G of order 3 or more that is Hamiltonian-

connected, then G is Hamiltonian. Suppose that U and V are the partite sets
of G. Then |U | = |V | ≥ 2 and any two vertices in the same partite set are not
connected by a Hamiltonian path.
 75
..............................
............ ........
....... ......
...... ......
...... .....
u u
..... ...........
.
..... ............. .... ...........
....... ..... .......
.......
.......
.......
..........
... ...
............
..
.........
...

u0 u
...
...
....... ...... ......
..... ............. ..... .........
. ......
...... .....
......
........ ...
........
........... ......
.......................................

Figure 6.4: The graph in Exercise 35

37. Claim: If G is a graph of order n ≥ 3 such that σ2 (G) ≥ n + 1, then G is

pancyclic.
This follows from Theorem 6.20 since, with f (n) = n + 1, G is not isomorphic
to K n2 , n2 .
38. By Theorem 6.9, the Petersen graph P is not Hamiltonian. We have men-
tioned that for every vertex v of P , the graph P −v is Hamiltonian. Therefore,
P is hypo-Hamiltonian.

Section 6.5. Powers of Graphs and Line Graphs

39. Claim: The subdivision graph S(G) of a graph G is Eulerian if and only if
G is Eulerian.
Proof. First observe that G is connected if and only if S(G) is connected.
Suppose that G is Eulerian. Thus, degG v is even for every vertex v of G. Since
degS(G) v = degG v if v ∈ V (G) and degS(G) v = 2 if v ∈ V (S(G)) − V (G),
it follows that S(G) is Eulerian. For the converse, suppose that S(G) is
Eulerian. Since every vertex of S(G) is even, every vertex of G is even and so
G is Eulerian.
40. Claim: The subdivision graph S(G) of a graph G of order n ≥ 3 is Hamil-
tonian if and only if G = Cn .
Proof. If G = Cn , then S(G) = C2n , which is Hamiltonian. For the con-
verse, we show that if G 6= Cn , then S(G) is not Hamiltonian. If G is discon-
nected, then S(G) is disconnected. So we may assume that G is connected. If
δ(G) = 1, then δ(S(G)) = 1 and S(G) is not Hamiltonian. Thus, δ(G) ≥ 2. If
G has a vertex v of degree 3 or more, then v is adjacent to at least three ver-
tices of degree 2 in S(G). By Exercise 2, S(G) is not Hamiltonian. Therefore,
G is 2-regular and so G = Cn
41. If G is a connected graph of order n and size m, then V (G) and V (S(G)) −
V (G) are two independent sets of S(G). If m = n − 1, then G is a tree and
α(S(G)) = |V (G)| = n; otherwise m ≥ n and α(S(G)) = |V (S(G)) − V (G)| =
m.
42. Proof. Since G1 = G and G is connected, the result is true for k = 1.
Hence we may assume that G is a connected graph of order n ≥ 3 and k is
76 CHAPTER 6. HAMILTONIAN GRAPHS

an integer with 2 ≤ k ≤ n − 1. Let S ⊆ V (Gk ) = V (G) with |S| ≤ k − 1.

We show that Gk − S is connected. Let u and v be two vertices of Gk − S.
We show that Gk − S contains a u − v path. Suppose that dG (u, v) = `. If
` ≤ k, then uv ∈ E(Gk ) and so uv ∈ E(Gk − S). Hence, we may assume
that ` > k. Suppose that P = (u = v0 , v1 , . . . , v` = v) is a u − v geodesic
in G. Let u = v0 = u0 , u1 , u2 , . . . , ut = v` = v be the subsequence of u =
v0 , v1 , . . . , v` = v consisting of those vertices of P that do not belong to S.
For i = 0, 1, . . . , t − 1, at most k − 1 vertices of S lie between ui and ui+1
on P . Thus, dG (ui , ui+1 ) ≤ k, implying that ui ui+1 ∈ E(Gk − S) and so
P 0 = (u = v0 = u0 , u1 , u2 , . . . , ut = v` = v) is a u − v path in Gk − S. Thus,
Gk − S is connected and so Gk is k-connected.
43. Proof. Let u, v ∈ V (G) such that dG (u, v) = diam(G) = `. Suppose that
` = qk + r for some integers q, r, where q ≥ 1 and 0 ≤ r < k and that

P = (u = v0 , v1 , . . . , v` = v)

is a u − v geodesic in G. Then

0 (u = v0 , vk , v2k , . . . , vqk , v` = v0 ) if r > 0
P =
(u = v0 , vk , v2k , . . . , vqk = v0 ) if r = 0

is a u − v geodesic in Gk and so

q+1 if r > 0
dGk (u, v) =
q if r = 0.

Thus, dGk (u, v) = d`/ke and so diam(Gk ) ≥ d`/ke. On the other hand, if
x, y ∈ V (Gk ) = V (G), then dG (x, y) = `0 ≤ `. This implies that dGk (x, y) =
d`0 /ke ≤ d`/ke. Therefore, diam(Gk ) = d`/ke.
2
44. (a) For n ≥ 3, let G = Kn . Then Kn2 = K1,n−1 = Kn . Note that if G is a
connected graph of diameter 2, then G 6= Kn and G2 = Kn .
(b) If G = K2 , then H 2 = G only when H = K2 .
45. Since G2 contains a vertex adjacent to more than two vertices of degree 2, it
follows by Exercise 2 that G2 is not Hamiltonian.
46. Proof. It suffices to establish the result for trees since if T is a spanning tree
of a graph G and T 3 − v is Hamiltonian for a vertex v of T , then G3 − v is also
Hamiltonian since T 3 − v is a spanning tree of G3 − v. Let T be a tree of order
4 or more and let v be a vertex v of T . Let T1 , T2 , · · ·, Tk be the components
of T − v, where then k = degT v. For each i with 1 ≤ i ≤ k, let vi be a vertex
of Ti adjacent to v. Also, if Ti is nontrivial, then let ui be a vertex of Ti that
is adjacent to vi ; otherwise, let ui = vi . If v is an end-vertex of T , then T − v
is a tree and T 3 − v = T13 . Since the cube of a connected graph of order 3 or
more is Hamiltonian, T13 is Hamiltonian and so T 3 − v is Hamiltonian. Hence
we assume that k ≥ 2. Thus, Ti3 is Hamiltonian-connected for 1 ≤ i ≤ k. Let
 77

Pi be a Hamiltonian ui − vi path for 1 ≤ i ≤ k, where Pi is trivial if Ti is

trivial. Since d(vi , uj ) ≤ 3 for 1 ≤ i 6= j ≤ k, it follows that vi uj ∈ E(T 3 − v).
Thus (P1 , P2 , · · ·, Pk , v1 ) is a Hamiltonian cycle in T 3 − v.
47. Proof. By Exercise 8(b), every self-complementary graph is connected. Also,
every self-complementary graph of order at least 5 has diameter 2 or 3.
If diam(G) = 2, then G2 is complete and is thus Hamiltonian-connected.
Hence, we may assume that diam(G) = 3. If G is 2-connected, then G2 is
Hamiltonian-connected by Theorem 6.23. Thus, we may further assume that
G has one or more cut-vertices. Let u1 be a vertex of G with e(u1 ) = 3 and for
0 ≤ i ≤ 3, let Ai (u1 ) be the set of vertices whose distance from u1 is i. Thus,
A0 (u1 ) = {u1 }. Necessarily, no vertex of A0 (u1 ) or A3 (u1 ) is a cut-vertex of
G. Let u2 be a vertex of G such that d(u1 , u2 ) = 3 and let N (ui ) = Vi for
i = 1, 2.
Assume first that V (G) = V1 ∪V2 ∪{u1 , u2 }. If every vertex of V1 is adjacent
to every vertex of V2 , then G2 = Kn − e, which is Hamiltonian-connected.
Hence we may assume that there exist vertices v1 ∈ V1 and v2 ∈ V2 such that
v1 v2 ∈
/ E(G). Observe that

X = {u1 u2 , v1 v2 } ∪ {u1 w2 : w2 ∈ V2 } ∪ {u2 w1 : w1 ∈ V1 } ⊆ E(G).

Since the spanning subgraph F of G with edge set X has the property that
2
F 2 is Hamiltonian-connected, so too is G = G2 .
Next, suppose that V (G) 6= V1 ∪ V2 ∪ {u1 , u2 }. Let

X = V (G) − (V1 ∪ V2 ∪ {u1 , u2 }) .

Let H be the spanning subgraph of G with edge set

E(H) = {xu1 , xu2 : x ∈ X} ∪ {u1 u2 } ∪ {u2 w1 : w1 ∈ V1 } ∪ {u1 w2 : w2 ∈ V2 }.

2
Since H 2 is Hamiltonian-connected, so too is G = G2 .
48. Proof. We proceed by induction on j. First, we show that t(G2 ) ≥ κ(G).
For this result, it is useful to define t(Kn ) = +∞ for every positive integer n.
If G2 is complete, then t(G2 ) = +∞ > κ(G). Thus, we may therefore assume
that G2 is not complete. Let S be a vertex-cut of G2 . We show that
|S|
≥ κ(G)
k(G2 − S)

which will implies that t(G2 ) ≥ κ(G).

Let V1 , V2 , . . . , V` be he vertex sets of the ` components of G2 − S. Since there
are no edges between Vi and Vj (1 ≤ i, j ≤ `, i 6= j) in G2 , the same is true
for G.
Next, we define ` subsets S1 , S2 , . . . , S` of S as follows. For each i = 1, 2, . . . , `,
a vertex u in S belongs to Si if an only if there is a vertex v ∈ Vi that is adjacent
78 CHAPTER 6. HAMILTONIAN GRAPHS

to u in G. Each vertex u in S belongs to at most one set Si , for otherwise,

there are vertices vi ∈ Vi and vj ∈ Vj , i 6= j, such that u is adjacent to both
vi and vj , implying that vi vj ∈ E(G2 ), a contradiction. Thus

Si ∩ Sj = ∅ for 1 ≤ i, j ≤ `, i 6= j. (6.2)

Furthermore, each set Si is a vertex-cut of G since the only paths in G from

Vi to V (G) − Vi contain a vertex of Si . Hence,

|Si | ≥ κ(G) for i = 1, 2, . . . , `. (6.3)

Combining (6.2) and (6.3), we have

`
X
|S| ≥ |Si | ≥ ` · κ(G) = k(G2 − S) · κ(G) (6.4)
i=1

It follows from (6.4) that

|S|
≥ κ(G).
k(G2 − S)

Thus, the result holds for j = 1. Suppose that if k = 2j for some positive
integer j, then t(Gk ) ≥ k2 κ(G). Now let k = 2j+1 and we show that
2k
t(G2k ) ≥ 2 κ(G) = k · κ(G).

By Theorem 6.13 and the induction hypothesis that

t(G2k ) = t((Gk )2 ) ≥ κ(Gk ) ≥ 2t(G2 ) = 2 k2 κ(G) = k · κ(G).

49. No. Let G be the graph obtained by the paths P2 = (u, w) of order 2 and
P4 = (v1 , v2 , v3 , v4 ) of order 4 by adding each vertex of u and w to every
vertex of P4 . Then G is claw-free but G is not a line graph by Theorem 6.25.
50. Proof. Let G be a nontrivial connected graph and e, f ∈ V (L(G)), where
say e = uv and f = xy for some u, v, x, y ∈ V (G). We show that e and f are
connected in L(G). This is obvious if e = f and so we may assume that e 6= f .
Suppose that k ≥ 1 is the minimum distance between a vertex of {u, v} and
a vertex of {x, y}, say dG (v, x) = k and

P 00 = (v = w0 , w1 , . . . , wk = x)

is a v − x geodesic in G. Therefore, neither u nor y lies on P 00 . Thus

P 0 = (u, v = w0 , w1 , . . . , wk = x, y)

is a u − y geodesic in G. Let ei = wi−1 wi for i = 1, 2, . . . , k. Then

P = (e, e1 , . . . , ek , f )

is an e − f path in L(G). Therefore, L(G) is connected.

 79

51. Suppose that G is a graph of order n and size at least 3. Let d1 , d2 , . . . , dn be

a degree sequence of G. Observe that (a, b, c, a) is a triangle in L(G) if and
only if the subgraph induced by the three edges a, b, c is either K3 or K1,3 .
Let τ (H) be the number of triangles in the graph H. Then
n  
X di
τ (L(G)) = τ (G) + .
i=1
3

52. Proof. Assume that G is Eulerian. Since G is connected, it follows by

Exercise 50 that L(G) is connected. Let e be a vertex in L(G). Then e = uv
is an edge in G. Moreover, degL(G) e = degG u + degG v − 2. Since u and v
are even vertices, degL(G) e is even. Thus L(G) is Eulerian.
53. (a) Claim: For a graph G of order n, G ∼ = L(G) if and only if G = Cn .
Proof. Without loss of generality, we assume that G is connected.
First, if G = Cn , then L(G) = Cn . For the converse, assume that G
is a connected graph of order n and size m such that G ∼ = L(G). Since
the order of L(G) is m, it follows that n = m and so G has exactly one
cycle. If G 6= Cn , then G contains a vertex of degree 3 or more. However
then, L(G) contains at least two cycles, which is a contradiction.
(b) By (a) L(G) ∼ = L2 (G) if and only if L(G) = Cn for some integer n ≥ 3.
If n ≥ 4, then G = Cn ; while if n = 3, then G = K3 or G = K1,3 .
54. (a) The statement is true. Proof. Since G is Hamiltonian, G is 2-connected.
Since the square of every 2-connected graph is Hamiltonian-connected,
G2 is Hamiltonian-connected.
(b) The statement is false. If G = K1,3 or G = K4 , then L(G) is Eulerian.
(c) The statement is false. If G = C4 , then L(C4 ) = C4 is not Hamiltonian-
connected.
(d) The statement is true. Proof. Since G has a dominating circuit, L(G)
is Hamiltonian. So L(G) has a Hamiltonian cycle C, which is a domi-
nating circuit of L(G).
55. Proof. Since G is connected, L(G) is connected and so L2 (G) is also con-
nected (by Exercise 50). Thus to show that L2 (G) is Eulerian, it suffices to
show that every vertex of L2 (G) has even degree.
Assume that L3 (G) is an Eulerian graph. Then every vertex of L3 (G) has
even degree. If G = Cn for some n ≥ 3, then Lk (G) = Cn for all k ≥ 1. If
G = K1,3 , then Lk (G) = C3 for all k ≥ 1. Hence, we may assume that G
is neither a cycle nor K1,3 . Since L3 (G) is Eulerian, G is not a path. Thus,
G has order 4 or more and contains at least one vertex of degree 3 or more.
Hence, Lk (G) contains a triangle for all k ≥ 1.
Let z be a vertex of L3 (G). Thus z has even degree in L3 (G). Suppose that
z corresponds to the edge xy in L2 (G). Thus
degL3 (G) z = degL2 (G) x + degL2 (G) y − 2.
80 CHAPTER 6. HAMILTONIAN GRAPHS

Hence, the degrees of x and y are of the same parity and so every two adjacent
vertices in L2 (G) are of the same parity. Since L2 (G) is connected, every
vertex of L2 (G) is even or every vertex of L2 (G) is odd. If every vertex of
L2 (G) is even, then L2 (G) is Eulerian and the proof is complete. Assume
then that every vertex of L2 (G) is odd. Let x be a vertex of L2 (G). Thus,
x is an odd vertex of L2 (G). Suppose that x corresponds to the edge uv in
L(G). Then
degL2 (G) x = degL(G) u + degL(G) v − 2.
Since x is an odd vertex, the degrees of u and v are of opposite parity and
so the degrees of every two adjacent vertices of L(G) are of opposite parity.
Since L(G) contains a triangle, this is impossible.

56. See Figure 6.5.

s .
................ ......s
........ ............
........ ............... ........ ... .....
s .
...... ............ ....
.... ...
.... ..
.... .... s
........ .............
........
.
.....
... ......
... .....
... ......
...... ...... ... .
.
...... ...... .... ... .. ... ... .....
........... ......
...... ...
... .
.
.
.. . .
.
. ... ...
...
.....
.....
......
. ...... ...... .. .. .... ... .....
...... . . .
s
... .....
s ...
...... .
s
.. . . . .
s
. .
s s s
.
s s
.. ..... . . . ... . .....
.. ....... .. ..... . . .... .
.. . .. .
...
. ....
.... .... .... ..... .... ... ...... .. ..... ..
. ...... ........
.
. . .... ..... . .
s
.
s s
. . . .
s s s s s s
.... ... .
... ... .. ..... ..... ..... ..... ..... .....
... ... .. . .
. .. .......... ..........
.... .....
G: ...... .. ....
..
. ... L(G) : ....... ...
.. .
... .......
.........
.. .......
...
s s
.... ..... .. ......
s
...
... .......
s s s
..... . . .
..... ........ ... ... ...
s s s
. ..... ..... ... .....
....... .... ...
...... ... ..... ........ .. ...... ... ...
...... . .
.
..... ..... ... ... ... ... .....
...... ...... ..... ... ... .. .. .....
...... ...... ..... ... ... .. .. .....
...... ...... ..... ... ... .. .
. ..
.... . .. .
...... ...... .....
..... ..... ...
.
.....
......
...... ....
....... ..... ...
... ..
... ... ... .......
s
...... ..........
..........
.
..... ..
..... ...
..... ... . s.
.....
. ......
.... ............ ... ... .........
.. ..
.
........ ........ ........ ... .....
s ....... ..............
........ s
........ ............
.....

Figure 6.5: The graphs in Exercise 56

57. (a) Proof. Assume, to the contrary, that there is a k-edge-connected graph
G, k ≥ 2, such that L(G) is not k-connected. Then there is a vertex-cut
S in L(G) with |S| = ` ≤ k − 1. Then S is a set of ` edges of G. Let e
and f be two vertices in L(G) − S that belong to different components
of L(G) − S. Suppose that e = uv and f = xy, where u, v, x, y ∈ V (G).
Since |S| < k and G is k-edge-connected, G − S is connected. Hence,
G − S contains a u − x path
P = (u = v0 , v1 , . . . , v` = x).
Let ei = vi vi+1 for 0 ≤ i ≤ ` − 1. Then ei ∈
/ S for 0 ≤ i ≤ ` − 1. Thus,
P 0 = (e0 , e1 , . . . , e`−1 )
is a path in L(G) − S. This, however, implies that e and f are connected
in L(G) − S, which is impossible.
(b) Proof. Let {U, W } be a partition of V (L(G)). By Theorem 4.2, it
suffices to show that |[U, W ]| ≥ 2k − 2. The sets U and W in L(G)
correspond to subsets X and Y , respectively, of E(G). Thus, {X, Y } is
a partition of E(G). For each vertex v of G, let δX (v) and δY (v) denote
 81

the number of edges incident with v that belong to X and belong to Y ,

respectively. Thus,
degG v = δX (v) + δY (v) ≥ k.
Hence, X
|[U, W ]| = δX (v) · δY (v).
v∈V (G)

Since G is connected, there exist vertices v in G such that δX (v) > 0 and
δY (v) > 0. Let V 0 be the set of all such vertices. We consider two cases.
Case 1. |V 0 | = 1. Let V 0 = {v}. Since the set of edges in X incident
with v is an edge-cut of G, as is the set of edges of Y incident with v, it
follows that
δX (v) ≥ λ(G) ≥ k and δY (v) ≥ λ(G) ≥ k.
Hence, |[U, W ]| ≥ k 2 ≥ 2k − 2 and so L(G) is (2k − 2)-edge connected.
Case 2. |V 0 | ≥ 2. Let v1 , v2 ∈ V 0 . Since δX (vi ) + δY (vi ) ≥ k for
i = 1, 2, it follows that δY (vi ) ≥ k − δX (vi ) > 0. Thus δX (vi ) ≥ 1 and
k − δX (vi ) ≥ 1. This implies that
δX (vi ) − 1 ≥ 0 and k − δX (vi ) − 1 ≥ 0.
Hence,
(δX (vi ) − 1)(k − δX (vi ) − 1) ≥ 0
or
kδX (v1 ) − k − (δX (vi ))2 + 1 ≥ 0.
Thus,
kδX (vi ) − (δX (vi ))2 ≥ k − 1
and so δX (vi ) · δY (vi ) ≥ k − 1 for each i. Therefore, |[U, W ]| ≥ 2(k − 1)
and so L(G) is (2k − 2)-edge connected.
58. We claim that there is no graph G such that T (G) = K1,3 . Assume, to the
contrary, that there is a graph F such that T (F ) = K1,3 . Since T (F ) =
(S(F ))2 , it follows that (S(F ))2 = K1,3 . Thus, S(F ) is a connected graph of
order 4. If diam(S(F )) ≤ 2, then (S(F ))2 = K4 . Hence, diam(S(F )) ≥ 3.
Since the order of S(F ) is 4, it follows that diam(S(F )) = 3 and S(F ) = P4 .
However then, (S(F ))2 6= K1,3 .
59. (a) Proof. Suppose that G is a nontrivial connected graph. Then G2 is
2-connected and so S(G2 ) is also 2-connected. It then follows by Theo-
rem 6.23 that T (G2 ) ∼
= (S(G2 ))2 is Hamiltonian.
(b) Proof. Suppose that G is a nontrivial connected graph. Then S(G) is
also a nontrivial connected graph. Thus, H = (S(G))2 is 2-connected.
Since T (G) ∼
= H, it follows that (T (G))2 ∼
= H 2 is Hamiltonian by Theo-
rem 6.23.
82 CHAPTER 6. HAMILTONIAN GRAPHS
Chapter 7

Digraphs

Section 7.1. Introduction to Digraphs

1. (a) Let D be the digraph with V (D) = {v1 , v2 , v3 , v4 , v5 } where (vi , vj ) ∈
E(D) if and only if 1 ≤ i < j ≤ 5.
(b) Yes. Let D be the digraph with V (D) = {u, v, w, x, y} and
E(D) = {(u, v), (v, u), (u, w), (w, u), (v, w), (w, v), (u, x), (u, y), (w, y), (y, x)}.
2. There are 27 digraphs of order 4 and size 4 (see Figure 7.1).
3. Let D be the digraph of order 2k such that V (D) is partitioned into two sets
U and W where |U | = |W | = k, where every vertex u ∈ U is adjacent to every
vertex of W .
4. Let V (D0 ) = {v1 , v2 , . . . , v5 } where (vi , vj ) ∈ E(D0 ) if 1 ≤ i < j ≤ 5. Then
od vi = 5−i. Let Di = D0 +(vi , v1 ), where 2 ≤ i ≤ 5. Then Di has exactly two
vertices of outdegree 6−i. Let D0 be the digraph of order 4 constructed in the
same manner as D0 . Thus, D0 has exactly one vertex of outdegree 0, 1, 2, 3.
Adding an isolated vertex to D0 produces a digraph of order 5 having two
vertices of degree 0. So, a can be 0, 1, 2, 3 or 4.
5. The statement is false. Let D be a directed triangle. Then all three vertices
have outdegree 1 and indegree 1.
6. The statement is false. Let D1 consist of the directed 6-cycle (u1 , u2 , . . . , u6 , u1 )
and let D2 consist of of the two directed 3-cycles (v1 , v2 , v3 , v1 ) and (v4 , v5 , v6 , v4 ).
Then idD1 ui = idD2 vi = 1 and odD1 ui = odD2 vi = 1 for i = 1, 2, . . . , n, but
D1 6∼
= D2 .
7. Proof. Observe that if v is a vertex of a regular tournament of order n, then
od v = id v = (n − 1)/2. Thus n is odd. Let V (D) = {v1 , v2 , . . . , vn }, where
n = 2r + 1 such that

(vi , vj ) ∈ E(D) if and only if j = i + 1, i + 2, . . . , i + r (mod n).

83
84 CHAPTER 7. DIGRAPHS

............... .................... ............ ............. ................. ............. ..............

..
....................... .............. ............. ......................... ............
. ... ... ... . ... ...
.. .
...... ...... ... ...... ..... .......... .
........ ....... ...... ... ... ... ... ...... .........
......
..........
. ...... ..... .
............. ..... ......
............... .......... ............
.. ...........
...... ...........
. . .............. ..................... ............... ............ ......................
. . .............
... . ... . . . ...

............ ................. ............ ............. ..................... ........... ............. ....................... ............ ............. ................. .........
.... ..... .... ... ..........
... ....... ... .........
...... ...... ..... .....
.......... ............
.
..... ...... ...... .....
......... . ........ . .......... ... ... ....................
....... .. .
....
.. . .
........ ...........
. . . . ... ... .............
. ............. .. ..... .
. ....... ..
..... ..... .....
.... .....
................ ....... ... ................ . ....... ..... .............
. .
........... ......... .............. ............... ............. ... .......... ......... ............... ................. ...
... .. ... ...

... ........... ......... ............. ........... .........

............. ...... .. .. ............. ..................... ............
........... ........
...... ............. ...... ..........
........ ... . ....
......
... ......... ... ...
..... ..... ...... ......
....... ...... ..... ... .. ..... ....... ......
........ ................... ... ..
...... .................... ...... .................... ........ ....................
. ........... . ........ . . . ....... .
. .........
..... ....
.
...... .....
.............. ...................... ........... ............... ....................... ............ ............ ................. ............. ............. ........... .........
...... .......
... ..... . ... ... ... ...

... ...
.......................... .................... ............ ................................... ................
....... .... ................ ......... ..... ................. .......................................... ....... .......... ...........................
................ . ............ ............. ............. .......... ................... .. . .
. ............... ................... .. .
............................. ................ ................ ...... .................. .............. ....... .................. .................................................
.................... ........................... ......................... ...... ....
.... ........ .................. ... .. ....... ..... ... ... ............ .
.
.
........ .... .. .
..
.. ...... .......... ........ .............. ..... . .
..........
........ ... ............... .. . .......... . .....
........................ . ...... . ......
........ ...... ................. ......... ...... ............. .............
............... . . ... .... ............. .............. .............. ............. ........ ............... ... ...
. ......................................... ... ... . ....... ... . . ...
...........

.
........................ ........................... ........................ ......................... ..................
......... ..... ................. ........ ..... ............... ........ ...... ................. ......... ...... ................ .......... ..... .................
............. ............. ............ ............. .............
................... .................... .......... .... ................. .......... ..... .......
. ............ ....... ...................... ........... ...... ....................
. .
................... ...........................
. ................................... ................... ..... ....................... .....
... ......... ...... ....... ..... ....
.......
........
.......
........ .... .......
.......
........ .......... ........ ...... .
........ . .........
. . .. .. . ................... .
..... ..............
. . . .. .. .
.
....... .. .....
. . .
.... ....... ............... ........... .......... . .
..... ..... . ..........
. .
............... ..
..
.............. ............... ..............
.. . ..... ...... ... ... ... . ... ..

..................... ................ ....................... ....................... ...........................

......... ...... ................ .......... ......................... ......... ...... ................ ......... ...................... ........ ..... ...............
............ . . .............. ............ ... ... .............. ... .... ............ . .
........... ...... .................... ........... .... ................. ........... ... ......... ... .. ............ ....... ................
......................... ..... .......................... ...................................... .................................. . .......................
... .. ...... .... . ... .. ...
.
....... . ...
.. ... ... .. ... .. ..........
......... .. .. ...... ...... ...
. ............. . .
......
.................. ............. ...................................................................... ..................................................................... ... ......
................................................................... ................................................................
... ... ... ... ... . ... ... .....

Figure 7.1: The 27 digraphs in Exercise 2

Then the resulting digraph D is an r-regular tournament of order n.

8. (a) Proof. Observe that
n
X n
X 2
(od vi )2 = ((n − 1) − id vi )
i=1 i=1
n
X
(n − 1)2 − 2(n − 1) id vi + (id vi )2
 
=
i=1
n
X n
X n
X
= (n − 1)2 − 2(n − 1) id vi + (id vi )2
i=1 i=1 i=1
  X n
n
= n(n − 1)2 − 2(n − 1) + (id vi )2
2 i=1
n
X n
X
2 2 2
= n(n − 1) − n(n − 1) + (id vi ) = (id vi )2 .
i=1 i=1
 85

P4 3
(b) The statement is false. For the tournament T in Figure 7.2, i=1 (od vi ) =
P4
30 and i=1 (id vi )3 = 24.

v1 ......................... ...........
......
.....
..........
......... v2
.....
..... ..
.......
..... ...
..... .....
..... ..... ...
..... ........ ....
.........
.. .. ...
... ..
. ...
............
..... .. .....
. ........ ...................
............. ......
..... .....
.
. ...... .....
. .....
.....
.............. ........... ............
v4 ... .. .... ... v3

Figure 7.2: The tournament T in Exercise 8(b)

9. (a) The sum of the entries in row i is od vi and the sum of the entries in
column i is id vi .
(b) An n × n matrix A is the adjacency matrix of a digraph of order n if and
only if A is a zero-one matrix in which every entry on the main diagonal
is 0.
10. (a) Proof. Let P be a longest path in D, where say P = (u = v0 , v1 , . . . , v` =
v). Since od v ≥ k ≥ 1, there are k vertices x1 , x2 , . . . , xk such that
(v, xj ) ∈ E(D) (1 ≤ j ≤ k). If xj ∈ / V (P ) for some j (1 ≤ j ≤ k),
then P 0 = (u = v0 , v1 , . . . , v` = v, xj ) is a longer path in D, which is
impossible. Thus, xj ∈ V (P ) for 1 ≤ j ≤ k. Let vi be the first vertex on
P such that (v, vi ) ∈ E(D). Then (vi , vi+1 , . . . , vk = v, vi ) is a cycle of
length at least k + 1 in D.
(b) Proof. Let P be a longest path in D, where say P = (u = v0 , v1 , . . . , v` =
v). Since id u ≥ k ≥ 1, there are k vertices x1 , x2 , . . . , xk such that
(xj , u) ∈ E(D) (1 ≤ j ≤ k). If xj ∈ / V (P ) for some j (1 ≤ j ≤ k),
then P 0 = (xj , u = v0 , v1 , . . . , v` = v) is a longer path in D, which is
impossible. Thus, xj ∈ V (P ) for 1 ≤ j ≤ k. Let vi be the last vertex
on P such that (vi , u) ∈ E(D). Then (vi , v0 = u, v1 , . . . , vi ) is a cycle of
length at least k + 1 in D.
11. Proof. Let G be a bipartite graph with partite sets U and W . Let D be the
orientation of G such that (u, w) ∈ E(D) if and only if u ∈ U and w ∈ W .
Then D has no directed path of length 2.
For the converse, if G is not bipartite, then G contains an odd cycle Ck ,
where k ≥ 3. Any orientation of G must direct two consecutive edges of Ck
in the same direction, producing a directed path of length 2.
12. Proof. For b ≥ 3, consider the directed (b + 1)-cycle (v0 , v1 , v2 , · · · , vb , v0 ).
Then join vb to v1 , v2 , v3 , · · · , vb−a+1 , producing a strong digraph with the
desired properties.
13. See the strong digraph D of Figure 7.3, where u1 and v1 are adjacent from
every vertex D1 and u5 and v2 are adjacent to every vertex D1 . Then
Cen(D) = D1 .
86 CHAPTER 7. DIGRAPHS

..............................
....... .....
..... ....
. ................... ........................
v3 v .
2.......................................................................
..
.. . ... ............................. u u4
. ......
t..... ................. t .. . .
..........
..............
.
...
...
...
.........................
............... 5
t
........ ................. t....................
.......... ..................... ..
...
. .... .
. ........................... ...............
............... . .
............................... .
. .... .....
...............
. .. ... .. ........
v4 t
.
t u3
.
...........
.....
.
.
........ .
.
.. .................. .............. ..... ........
.......
......
...... ........ .... ............. .... ........ .......
......
...... .
...........
...........
.
.
.
..................... ...
...
.
... ..
...
.... ........ D 1
.............
.. ..................
. .
..........................
.
... . .......................
.
t t u2
.
.......
t
...... ..
t ...
... . .
. .............
...... ........... ....................... ..... ...
. .
.. ... .. . .......... .......
...... ................................ ........ .. ......... .
. ......
............................. ..... ... ...............................
............................
v5 v 1
. ...........................
.... .
.....
.....................
..
u 1
...... .....
......... ......
.....................

Figure 7.3: The strong digraph D in Exercise 13

14. Proof. We proceed by induction on the order n of D. If n = 1, then the

result follows vacuously. Assume that the statement is true for all digraphs
with fewer than n vertices where n ≥ 2. Let D be a digraph of order n and
let v be a vertex of D. Let D0 = D − ({v} ∪ N + (v)), where N + (v) is the set
of all vertices adjacent from v. By the induction hypothesis, there is a set S 0
in D0 such that no two vertices in S 0 are adjacent in D0 and each vertex not
in S 0 is reachable from a vertex in S 0 by a directed path of length at most
2. If v is adjacent from some vertex u of S 0 , then every vertex of N + (v) is
reachable from u by a directed path of length 2. Hence, in this case, S = S 0
satisfies the required property. If, on the other hand, v is not adjacent from
any vertex of S 0 , then v is joined to no vertex of S 0 and S = S 0 ∪ {v} has the
desired property.

Section 7.2. Strong Digraphs

15. Proof. Among all u − v walks in D, let P = (u = u0 , u1 , . . . , uk = v) be a
u − v walk of minimum length for which every arc of P belongs to W . Thus,
the length of P is k. We claim that P is a u − v path. Assume, to the
contrary, that this is not the case. Then some vertex must be repeated in P ,
say ui = uj for some i and j with 0 ≤ i < j ≤ k. If we then delete the vertices
ui+1 , ui+2 , . . . , uj from P , we arrive at the u − v walk

W 0 = (u = u0 , u1 , . . . , ui−1 , ui = uj , uj+1 , . . . , uk = v),

whose length is less than k and such that every arc of W 0 belongs to W , which
is impossible.
16. Proof. Since P = (u = v0 , v1 , v2 , . . . , vr = v) is a u − v path in D if and
only if

Q = (v = vr , vr−1 , . . ., v1 , v0 = u)

~ it follows that D is strong if and only if D

is a v − u path in D, ~ is strong.

17. (a) Proof. Let e = uv and suppose that e is oriented as (u, v). Since G
is connected and has no bridges, it follows by Theorem 7.5 that G has
 87

a strong orientation D. If (u, v) ∈ E(D), then the result holds; while if

(u, v) ∈
/ E(D), then the converse D~ of D is a strong orientation of G and
~
(u, v) ∈ E(D), in which case the result follows from Exercise 16.
(b) Let Cn = (v1 , v2 , . . . , vn , v1 ) be an n-cycle, n ≥ 3. Orient v1 v2 and v1 vn
as (v1 , v2 ) and (v1 , vn ), respectively. Then there is no strong orientation
of Cn containing (v1 , v2 ) and (v1 , vn ).

18. Proof. First, assume that D is a strong orientation of G. Let B be a block

of G, and let D0 be the subdigraph of D induced by V (B). Let u and v be
two vertices of D0 . Since D is strong, there is a directed u − v path P in D.
Since u, v ∈ V (B), every vertex of P belongs to B. Thus, P is a directed u − v
path in D0 .
For the converse, let D be an orientation of G such that the subdigraph
of D induced by the vertex set of each block of G is strong. Let u and v be
two vertices of D. We show that D contains a directed u − v path. If u and
v belong to a common block of G, then it is immediate that D contains a
directed u − v path. Hence, we may assume that u and v belong to distinct
blocks B and B 0 of G. Let P be a u − v path in G. Thus, there exist k ≥ 1
cut-vertices w1 , w2 , · · · , wk on P such that wi belongs to blocks Bi and Bi+1
for i = 1, 2, · · · , k, where B = B1 and B 0 = Bk+1 . By hypothesis, there exists
a directed u−w1 path P0 in the subdigraph of D induced by V (B1 ), a directed
wi − wi+1 path Pi (1 ≤ i ≤ k − 1) in the subdigraph of D induced by V (Bi+1 )
and a directed wk − v path Pk in the subdigraph of D induced by V (Bk+1 ).
The directed paths P0 , P1 , · · · , Pk produce a directed u − v path in D.

19. (a) Proof. The result is true if G contains no bridges. Thus, we may
assume that G contains at least one bridge. We consider two cases.
Case 1. G contains exactly one bridge, say e = xy. Then G − e
has exactly two components G1 and G2 , each of which has a strong
orientation. Let Di be a strong orientation of Gi (i = 1, 2). Then the
orientation D of G obtained from D1 and D2 by adding the arc (x, y)
has the desired property.
Case 2. G contains exactly two bridges, say e = xy and f = wz. We
may assume that
d(y, w) = min{d(p, p0 ), p ∈ {x, y}, p0 ∈ {w, z}}.
Then G − e − f has exactly three components G1 , G2 and G3 , each
of which has a strong orientation. Let Di be a strong orientation of
Gi (i = 1, 2, 3). Furthermore, y and w belong to the same component.
Suppose that x ∈ V (G1 ), y, w ∈ V (G2 ), and z ∈ V (G3 ). Then the
orientation D of G obtained from D1 , D2 and D3 by adding the arcs
(x, y) and (w, z) has the desired property.
(b) The statement (a) is false if G contains three bridges. For example, no
orientation of G = K1,3 has the desired property.
88 CHAPTER 7. DIGRAPHS

20. Proof. Assume, to the contrary, that D is not strong. Then D contains
two vertices u and v for which there is no u − v path. Then (u, v) is not
an arc of D. If N + (u) ∩ N − (v) 6= ∅, then there exists a vertex w such
(u, w), (w, v) ∈ E(D) and so (u, w, v) is a u − v path, a contradiction. Thus,
N + (u) ∩ N − (v) = ∅. However then, {u}, N + (u), N − (v), {v} are pairwise
disjoint sets with |{u}∪N + (u)∪N − (v)∪{v}| ≥ n+1, again a contradiction.

Section 7.3. Eulerian and Hamiltonian Digraphs

21. Proof. First, let D be an Eulerian digraph. Then D contains an Eulerian
circuit C. Let v be a vertex of C. Assume first that v is not the initial vertex
of C (and so v is not the terminal vertex either). Whenever v is encountered
on C, an arc is used to enter v and another is used to exit v. This contributes
1 to both the indegree and outdegree of v. If v is encountered k times on C,
then od v = id v = k. If v is the initial vertex of C, then an arc is used to exit
v. The final arc of C enters v. Any other occurrences of v on C contribute
1 to both the indegree and outdegree of v, and so od v = id v in this case as
well.
For the converse, let D be a nontrivial connected digraph for which od v =
id v for every vertex v of D. For a vertex u of D, let T be a trail of maximum
length with initial vertex u. Suppose that T is a u − v trail. Assume first that
v 6= u and that v is encountered k times on T , where k ≥ 1. Then T contains
k arcs directed towards v and k − 1 arcs directed away from v. However,
since od v = id v, there is an arc directed away from v that does not belong
to T . This means, however, that T can be extended to a longer trail with
initial vertex u. Since this is impossible, u = v and T is a circuit C in D.
Consequently, D contains circuits and C is a u − u circuit of maximum length
in D.
We claim that C contains all of the arcs of D and that C is therefore an
Eulerian circuit. Assume, to the contrary, that C does not contain all of the
arcs of D. Since D is connected, there is a vertex w on C that is incident with
arcs not on C. Let D0 = D − E(C) be the spanning subdigraph of D whose
arcs are those not belonging to C. Since odD v = idD v and odC v = idC v for
every vertex v on C, it follows that odD0 v = idD0 v for every vertex of D0 .
Let T 0 be a trail of maximum length in D0 with initial vertex w. As before,
T 0 is a w − w circuit C 0 in D00 . If we insert C 0 at w in C, then we produce a
circuit C 00 in D containing more arcs than C, which is impossible. Hence, C
is an Eulerian circuit.

22. Proof. First, assume that G is Eulerian. Then G contains an Eulerian

circuit C = (v1 , v2 , · · · , vk , v1 ). Let D be the orientation of G obtained by
replacing each edge vi vi+1 (1 ≤ i ≤ k) of G by the arc (vi , vi+1 ) and replacing
vk v1 by (vk , v1 ). Then C is an Eulerian circuit of D and so D is Eulerian.
For the converse, assume that G is an Eulerian orientation of D. Thus,
D is connected and od v = id v for every vertex v of D. Consequently, G is
 89

connected and
degG v = odD v + idD v = 2idD v.
So, G is Eulerian.
23. Proof. First, suppose that D contains an Eulerian trail, say a u − v trail.
Add a new vertex w to D along with the arcs (v, w) and (w, u), denoting
the new digraph by D0 . Then D0 contains an Eulerian circuit and so D0 is
Eulerian. By Theorem 7.6, odD0 x = idD0 x for every vertex x of D0 , which
implies that od u = id u + 1 and id v = od v + 1 in D, while od w = id w for
all other vertices w of D.
For the converse, suppose that D contains two vertices u and v such that
od u = id u + 1 and id v = od v + 1, while od z = id z for all other vertices z
of D. Let D0 be the digraph obtained from D by adding a new vertex w and
two new arcs (v, w) and (w, u). Then od x = id x for all vertices x of D0 and
so by Theorem 7.6, D0 is Eulerian. Let C be an Eulerian circuit of D0 . The
arcs (v, w) and (w, u) are consecutive on C, so C − w is an Eulerian trail of
D that begins at u and ends at v.
24. Proof. Let D0 be the digraph constructed from D by adding k new vertices
w1 , w2 , . . ., wk and the 2k arcs (v, wi ) and (wi , u) for 1 ≤ i ≤ k. Since
od x = id x for every vertex x of D0 , the digraph D0 is Eulerian and therefore
contains an Eulerian circuit C. Suppose that C begins (and ends) at u. Since
od u − id u = k and id v − od v = k, it follows that C contains k arc-disjoint
u − v trails. By Theorem 7.3, for each such u − v trail T , there is a u − v path
each of whose arcs belongs to T . Since none of these paths contains any of
the vertices wi (1 ≤ i ≤ k), D contains k arc-disjoint u − v paths.
25. (a) Yes. If x 6= v, then od x ≥ id x. If the vertex x is encountered k times
in T 0 , then idT 0 x = k and odT 0 x = k − 1 and there is at least one arc
incident with x and directed away from x that is not on T 0 . Thus, T 0
can be extended.
(b) No. Let D be the digraph with V (D) = {u, v, w, x, y} and
E(D) = {(u, w), (w, v), (w, x), (x, y), (y, w)}.
The u − v trail T = (u, w, v) cannot be extended in the digraph D but
T is not an Eulerian trail in D.
26. Proof. First, suppose that D is a nontrivial connected digraph such that
E(D) can be partitioned into subsets Ei , 1 ≤ i ≤ k, where the subdigraph
D[Ei ] induced by the set Ei is a cycle for each i. Then each vertex v of D
belongs to j of these cycles for some j with 1 ≤ j ≤ k. Thus, od v = id v = j
and so D is Eulerian by Theorem 7.6.
For the converse, we proceed by induction on the size m of D. If m = 2,
then D is the directed 2-cycle and so D has the desired property. Assume then
the result holds for every Eulerian digraph of size less than m, where m ≥ 3.
Let D be an Eulerian digraph of size m. Since D is Eulerian, od v = id v ≥ 1
90 CHAPTER 7. DIGRAPHS

for every vertex v of D by Theorem 7.6. Thus, by Theorem 7.2, D contains

at least one cycle C. If D is the directed m-cycle, then the proof is complete.
Otherwise, there are arcs of D not in C. Removing the arcs of C from D
produces a digraph D0 , each nontrivial component of which is Eulerian. Since
D0 has fewer than m arcs, it follows by the induction hypothesis that the arc
set of each nontrivial component of D0 can be partitioned into subsets Ei0 ,
1 ≤ i ≤ ki0 , where the subdigraph D[Ei0 ] induced by the set Ei0 is a cycle for
each i. Now all of these cycles of the components of D0 , together with C, give
the desired result.
27. Proof. Let V (D) = {v P1n, v2 , . . . , vn }.PnBy the First Theorem of Digraph
Theory (Theorem 7.1), i=1 od vi = i=1 id vi = m, where m is the size
of D. Assume that od vi ≥ id vi for 1 ≤ i ≤ k and that od vi < id vi for
k + 1 ≤ i ≤ n. Then
Pk Pn
i=1 (od vi − id vi ) = t and i=k+1 (id vi − od vi ) = t.

Let D0 be a digraph obtained from D by adding t new vertices x1 , x2 , . . . , xt

in such a way that
(1) each vertex xj (1 ≤ j ≤ t) is adjacent to some vertex in {v1 , v2 , . . . , vk }
and adjacent from some vertex {vk+1 , vk+2 , . . . , vn } and
(2) idD0 v = odD0 v for every vertex v ∈ V (D0 ).
Then D0 is Eulerian by Theorem 7.6. Since D0 is Eulerian, D0 has a Eulerian
circuit C. Since deg xj = 2 for 1 ≤ j ≤ t, the circuit C contains every
vertex xj and the two arcs incident with xj as consecutive arcs in C. Then
C − {x1 , x2 , . . . , xt } consists of t open trails in D whose arc sets produce a
partition of E(D).
28. Proof. Let m be the size of D. By the First Theorem of Digraph Theory
(Theorem 7.1),
Xn n
X
od vi = id vi = m.
i=1 i=1

Thus,
n
X
(od vi − id vi ) = 0. (7.1)
i=1

Since od vi ≥ id vi for 1 ≤ i ≤ n, it follows that od vi − id vi ≥ 0 for 1 ≤ i ≤ n.

By (7.1), od vi − id vi = 0 and so od vi = id vi for 1 ≤ i ≤ n, implying by
Theorem 7.6 that D is Eulerian.
29. Proof. Suppose that the order of D is n. Then the vertices of D can be labeled
as v1 , v2 , . . . , vn so that r(vi ) = i for 1 ≤ i ≤ n. Necessarily, D contains no cy-
cle, for otherwise, if u and v belong to a cycle, then a vertex is reachable from v
if and only if it is reachable from u, implying that r(u) = r(v), which is impos-
sible. We show that D contains the Hamiltonian path (vn , vn−1 , . . . , v2 , v1 ).
 91

Suppose that it does not. First, we claim that (vn , vn−1 ) ∈ E(D). Suppose
that this is not the case. Since r(vn ) = n, there is a vn − vn−1 path P in
D such that P has an internal vertex x. Since x can reach every vertex that
vn−1 can reach, r(x) ≥ r(vn−1 ), which is impossible. Thus, as claimed, vn
is adjacent to vn−1 in D. Let k be the largest positive integer such that
(vk , vk−1 ) ∈
/ E(D). Then k ≤ n − 1. Since D contains no cycle, D contains
no path from vk to any of the vertices vn , vn−1 , . . . , vk+1 . Thus, vk cannot
reach any of vn , vn−1 , . . . , vk+1 and must reach all of v1 , v2 , . . . , vk−1 . Hence,
D contains a vk − vk−1 path with an internal vertex vj such that (vk , vj ) is an
arc and 1 ≤ j ≤ k − 2. Since D contains a vj − vk−1 path and r(vk−1 ) = k − 1,
it follows that r(vj ) ≥ k − 1, which is a contradiction. Thus, D contains the
Hamiltonian path (vn , vn−1 , . . . , v2 , v1 ).
30. Proof. First, we show that that D is strong. We claim for every two vertices
u and v that D contains a u − v path. If (u, v) ∈ E(D), then (u, v) is a u − v
/ E(D), then od u + id v ≥ n − 1 and N + (u) and N − (v) are
path. If (u, v) ∈
disjoint sets with |N + (u) ∪ N − (v)| ≥ n − 1. Then D contains at least n + 1
vertices, which is impossible.
Suppose that n = 2k + 1 for some positive integer k. Let D be the digraph
∗
obtained by identifying a vertex in two copies of Kk+1 . Then for every two
vertices u and v with (u, v) ∈
/ E(D), od u + id v = 2k = n − 1 but D is not
Hamiltonian.
31. For k ≥ 3, let D1 and D2 be copies of the digraph Kk∗ obtained by replacing
each edge uv of Kk by the symmetric pair (u, v) and (v, u) of arcs. Let D be
obtained by identifying a vertex of D1 and a vertex of D2 , denoting this vertex
by x. Then the order of D is n = 2k −1. Observe that od v ≥ k −1 = (n−1)/2
and id v ≥ k − 1 = (n − 1)/2 for every vertex v of D but D is not Hamiltonian
as D has a cut-vertex x.
32. Consider the digraph D of order 4 in Figure 7.4, which is not strong. Although
deg v = 4 for every vertex v of D, the digraph D is not Hamiltonian.
......... ......
........... ........
............... .................
... ............... ........ .....
....
. ... ...... ...
....... .... ...
... ... ....... .
............. ........ ...
.......... ... .....
..... .......... ...
. ... ..... ..... .......... ...
..... ... ..... .......... .. ...
......... .
D: ...
...
... ........
..
.
.
..........
... ......
.
..... .
.. .
.
...
...
.
. ..
..........
.
....... ..... . .
................. ... ..
... . .... . ..
... ... ...... ......
..... ..... ..
...
.... .......... ..... .. .
...
... .. ...... ..... .. ..
............... ........ ..
.... ..... ........... ..............
... ...... .......

Figure 7.4: The digraph Exercise 32

Section 7.4. Tournaments

33. The tournaments T 0 and T 00 in Figure 7.5 are strong tournaments of order 5.
Since T 00 has vertices with outdegree 3 and every vertex in T 0 has outdegree
2, T 0 ∼
6 T 00 .
=
92 CHAPTER 7. DIGRAPHS

r .
...........
r ...
..........
..... ..... .....
..... ... .... ..... .. .. .....
..... .... ..... ..... .... .... .......... .
. .
. ..... .... ..... .......... . ....
.
............. ... ... .................. ............... ... ...
. ..................
........ ... ..... ...... ...... .... ....... ......
...... ..... ..... ........ ........ .....
..... .. .....
r r r
..... ..... .....
r
..... ..... ..... ....... . ..... .....
.. . .. .. .
....................................................................................................................
. . . .
.........
. . ...................... ... ..
..
T0 : .........
... ....... .
..
. .
.... ........
. T 00 : .........
. ..
.
.
. ...
..... ....
.........
... ....... .............. ..
.
........
. .... .... ....
. .
... .....
... ....... ..
.
...
. ....
. .
... ....... ... ... ...... .. ... ....... ... ... ....... ..
......... ........ ........ . ......... ........ . .
. ....... ... ... . ....
......... ...
... ................ ............... ... ....... ...... ........ . ........... ............
...
... .... ......................... .... ......... ............ .. ............. ................. ....
... .... .................... ... .......
... ... ...... ....... ... ... ... . . ..
.. . . . .
. ..
... .. ....... ...... ... .. ... ... ......... ........... .... ....
r
...... .. ..
r r
.............
r
... ... ......
........... ...... .. ..
...... ....................... .......... ..........
.... . ..................... .

Figure 7.5: The tournaments T 0 and T 00 in Exercise 33

34. There is one tournament of order 7 having seven strong components. This is
the transitive tournament of order 7. There is no tournament of order 7 having
six strong components. There are five tournaments of order 7 having five
strong components. Every tournament of order 7 with four strong components
has exactly one strong component of order 4. Since there is only one strong
tournament of order 4, there are four tournaments of order 7 having four
strong components. A tournament of order 7 with three strong components
has either (1) two strong components of order 3 or (2) one strong component of
order 5. There are three tournaments of order 7 in the case of (1). Since there
are six strong tournaments of order 5, there are 18 tournaments of the type (2).
Finally, we determine the number of tournaments of order 7 with two strong
components. If one of the strong components has order 3 and the other has
order 4, then there are two such tournaments. All other tournaments of order
7 that are not strong contain a strong component of order 6. Suppose that
there are x strong tournaments of order 6. Then there are 2x tournaments
of order 7 containing a strong component of order 6. Therefore, the total
number of tournaments of order 7 that are not strong is 33 + 2x.
We now compute x. There is one tournament of order 6 with 6 strong
components (the transitive tournament), four tournaments of order 6 with 4
strong components, three tournaments of order 6 with 3 strong components,
one tournament of order 6 with 2 strong components, both of which have
order 3, and twelve tournaments of order 6 with 2 strong components, one of
order 5 and one of order 1. We have seen that there are 56 tournaments of
order 6. Thus x = 56 − 21 = 35. Hence there are 21 tournaments of order 6
that are not strong. Therefore, the total number of tournaments of order 7
that are not strong is 33 + 2x = 33 + 2(35) = 33 + 70 = 103.

35. If D is an r-regular tournament of order n and size m, then

 
n n(n − 1)
nr = m = = ,
2 2

implying that n = 2r + 1, which is odd. For r = (n − 1)/2 ≥ 1, an r-regular

tournament T of order n with V (T ) = {v1 , v2 , . . . , vn } can be constructed by
joining vi to vi+1 , vi+2 , . . . , vi+r for each i (1 ≤ i ≤ n), where the subscripts
are expressed modulo n.
 93

36. Let T1 be a tournament of order 9 for which V (T1 ) = {v1 , v2 , · · · , v9 } and vi

is adjacent to vi+1 , vi+2 , vi+3 , vi+4 for 1 ≤ i ≤ 9, where the subscripts are
expressed modulo 9. To construct a tournament T2 of order 9, we let C1 ,
C2 , C3 be three pairwise disjoint directed 3-cycles, orient all edges between
C1 and C2 from C1 to C2 , all edges between C2 and C3 from C2 to C3 , and
all edges between C3 and C1 from C3 to C1 . Both T1 and T2 are 4-regular.
The subdigraph of T1 induced by those vertices adjacent from a vertex of T1 is
transitive; while the subdigraph of T2 induced by those vertices adjacent from
a vertex of T2 contains a directed 3-cycle. Thus, T1 and T2 are nonisomorphic.

37. Proof. If Te were not transitive, then Te would contain a triangle (S1 , S2 , S3 , S1 ).
Let vi ∈ V (Si ) for i = 1, 2, 3. Then (v1 , v2 , v3 , v1 ) is a directed 3-cycle in T
and so v1 , v2 , v3 must belong to the same strong component of T , which is a
contradiction.

38. (a) Proof. Let S1 , S2 , · · · , Sk be the k ≥ 1 strong components of T . If

k = 1, then the result follows trivially. Thus, we may assume that k ≥ 2.
Suppose that u and v are two vertices in T such that u and v belong
to distinct strong components of T . Without loss of generality, we may
assume that (u, v) ∈ E(T ). Let

W = {x ∈ V (T ) : (v, x) ∈ E(T )}.

Then |W | = od v. Since (u, v), (v, x) ∈ E(T ), it follows that (u, x) ∈

E(T ). Thus, od u ≥ 1 + |W | = 1 + od v > od v and so od u 6= od v.
(b) Proof. By (a), two vertices with the same outdegree belong to the
same component. Since all vertices of a regular tournament have the
same outdegree, there is only one strong component.

39. (a) The sequence π : 0, 1, 1, 4, 4 is not a score sequence of a tournament; for

if it were, then by Theorem 4.17 the sequence π1 : 0, 1, 1, 4 would also be
the score sequence of a tournament. However, no tournament of order 4
contains a vertex of outdegree 4.
(b) Notice that

π: 1, 1, 1, 4, 4, 4
π10 : 1, 1, 1, 4, 3
π1 : 1, 1, 1, 3, 4
π2 : 1, 1, 1, 3.

Since π2 is a score sequence of a tournament, so are π1 and π. See

Figure 7.6, where the tournament Ti has score sequence πi for i = 1, 2
and the tournament T has score sequence π.
(c) Notice that

π: 1, 3, 3, 3, 3, 3, 5
94 CHAPTER 7. DIGRAPHS

t
..... .....
..... .... .....
r
. .....................................
...... ...... ......
..... .. .. ......... . ......... .......... .. .......
.. ..... .. ... ...... .. ............. ............
................. .... ............ ................. ..
.
.................. .... ......... .......... .....
r
...
r
..
r
.
r r r
... .. .... . ........ ...... ...
........................... ..... ... .. .... ...... ..
.. .
. ... . ...
............ ..... .............. .. ... ..................... ............
.
.
.................... ...
. . .. . . ...
... .......... ........... ... ................... ... ... ..... . ... .... ...................... . .
.......... .... ...
................. ... ... ............... .. ...................... ..... . . . . .. . . ..
r
. .
......... ..................... .... ....
... ..............
r r
...
... ......... .................... ..... ....... ... . ... ..
...
. ... .... ... ............
.... .. . ........... ................ ... .... ... . ... ... ...
............ ...
. ..... ...... . . ... ... .... .......... .... .........
.
..
.
....... .. . .... ... ........ ...... . .. ... ... ....... ........
. ...... ..
. . . .
... ........... ............... ... .... ...... ....
... .. .... ..........
. .... .... ....... ........ ... ...........
. ..
. .
. .
. .
...
. ....
... . ....
. ... ... .... .......... ... ... ..... ...
. ................ ... ..
... ..... ...... ...... . .
...... ... . ...... .. . .
... .....
r
... ...
.. r
..... ...
.. .
...
....
.....
r
..... ...
.....
.... .. ..
..... ................................................................ .... t
... .
...... .....
....... ....
. .
......... .......
....................................
T2 T1
T

Figure 7.6: The tournaments in Exercise 39(b)

π10 : 1, 3, 3, 3, 3, 2
π1 : 1, 2, 3, 3, 3, 3
π20 : 1, 2, 3, 2, 2
π2 : 1, 2, 2, 2, 3
π30 : 1, 2, 2, 1
π3 : 1, 1, 2, 2.

Since π3 is a score sequence of a tournament, so is π. The tournament

T of Figure 7.7 has score sequence π.
.................................
............. ... .............................
........ ....................... .........
...... ........ .......
........ ...... ......
................. ...... .....
.....
......
..............
... .. . ........ ........................ ... .......
. . .. .. ... .............. ...... ............. ..... ...
... ............. ...... .. ...
. . . . .
...................................... .. ...
... .
...
...
...
. .
.....
... ...... ....... . ........ . ..
.. ....
....... ...... ..... ... ...... .. .
... ... ................. ...
. .. .................. .........
. . .
................... ......
..
.
. . ....
. .. .
. ... . .... .
.............. . .
... .. . . ... ...... .....
.
............. .. . .. .. . ...
....... ......... ............. ............. . . .. . .. . . . .......... ...... . . ..
..... ...

T : u..... ....................................
.............
......... ........................................ ............. ........ ......... ........
.
. ..... .....
.......
.. .
..
. . . .
...........
r
.... ............... .........
.......... .. .......
....
t
... ....... ................ ..... .
.. ..
. . ................. ....
. ... ..... .......
... ............. .............. ..... .
............. .... .... .......
.
.............. .......... ............... ..............
. .....................
... . ..................................... .................... ....................... .
........................... ......
... . . .
......... ... . . . . .... ......
.
. ... ...
... ... ........ ......
..... ... ..
... . . ....
. .
. .
..
..
.. . ....... ... .... ....
... .
..... ...................... . ..
... ..... .............. ....
..............
...... ........ ...
...... ...... ....
.......
...
........
............ . . ........ ....... ..
..... ...................... . ... .
..... .....
..... .....
...... .....
......
....... ... ........
........ ......
........... ........
..................................................

Figure 7.7: A tournament having score sequence π in Exercise 39(c)

(d) The sequence 2,
3, 3, 4, 4, 4, 4, 5 has eight terms. The sum of the terms
is 29 6= 28 = 82 , so it cannot be the score sequence of a tournament of
order 8.
40. If π is a score sequence of a tournament T , then π1 is the score sequence of
the converse tournament of T .
41. Suppose that T is a tournament Pk with score
sequence π : s1 , s2 , . . . , sn such
that s1 ≤ s2 ≤ · · · ≤ sn and i=1 si = k2 for all k with 1 ≤ k ≤ n. Let
V (T ) = {v1 , v2 , . . . , vn } where od vi = si for 1 ≤ i ≤ n. Let 1 ≤ k < n
and let T 0 be the subtournament of T induced by {v1 , v2 , . . . , vk }. Also, let
 95

Pk
odT 0 vi = s0i . Thus, i=1 s0i = k2 . Hence, s0i = si for 1 ≤ i ≤ n, implying

that vk is adjacent to v1 , v2 , . . . , vk−1 for k = 2, 3, . . . , n and so sk = k − 1.

Thus, T is a transitive tournament.

42. Proof. Suppose first that π : s1 , s2 , . . . , sn is a nondecreasing sequence that

is a score sequence of a strong tournament T of order n. We may assume that
V (T ) = {v1 , v2 , . . . , vn } such that od vi = si for i = 1, 2, . . . , n. Since T is a
tournament of order n,
n  
X n
si = .
i=1
2

Next, let k be an integer with 1 ≤ k ≤ n − 1 and S = {v1 , v2 , . . . , vk }. Then

the induced subdigraph T1 = T [S] is a tournament of order k and so

k  
X k
odT1 vi = .
i=1
2

Since T is a strong tournament, some vertex vj (1 ≤ j ≤ k) in T1 is adjacent

in T to a vertex not in T1 so that odT vj > odT1 vj . Since odT vi ≥ odT1 vi for
all i with 1 ≤ i ≤ k, we have
k k k  
X X X k
si = odT vi > odT1 vi = .
i=1 i=1 i=1
2

Next, we verify the converse. Let π : s1 , s2 , . . . , sn be a nondecreasing

sequence of nonnegative integers such that
k  
X k
si >
i=1
2

for 1 ≤ k ≤ n − 1 and
n  
X n
si = .
i=1
2

By Theorem 7.19, π is a score sequence of a tournament T of order n. Let

V (T ) = {v1 , v2 , . . . , vn } such that od vi = si for i = 1, 2, . . . , n. It remains to
show that T is strong.
Assume, to the contrary, that T is not strong. Then T contains t ≥ 2 strong
components S1 , S2 , . . . , St such that if u ∈ V (Si ) and v ∈ V (Sj ) where 1 ≤
i < j ≤ t, then (u, v) ∈ E(T ). Suppose that W = V (St ) and U = V (T ) − W .
Then the outdegree of each vertex of U exceeds the outdegree of each vertex
of W . Let |W | = k. Thus, 1 ≤ k ≤ n − 1. Therefore,

{odT w : w ∈ W } = {s1 , s2 , . . . , sk }.
96 CHAPTER 7. DIGRAPHS

Hence, odT w = odSt w for each w ∈ W and so

k  
X X X k
si = odT w = odSt w = ,
i=1
2
w∈W w∈W

producing a contradiction.

43. (a) There is exactly one circular tournament of order 3, namely the unique
strong tournament of order 3.
(b) Proof. Suppose that there is a tournament T that contains three ver-
tices not having both positive outdegree and positive indegree. Then two
of these vertices, say u and v, both have outdegree 0 or indegree 0, say
the former. However, either (u, v) or (v, u) is an arc of T , producing a
contradiction.
(c) None. Assume, to the contrary, that there exists a circular tournament
of order 4 or more. By (b), T contains a vertex v having both positive
outdegree and positive indegree. Since T contains at least three vertices
other than v, either od v ≥ 2 or id v ≥ 2, say the former. Then there
exist three vertices u, w and x such that (u, v), (v, w) and (v, x) are arcs
of T . Either (w, x) or (x, w) is an arc of T , say the former. Since (v, w)
and (w, x) are arcs of T and T is circular, (x, v) is an arc of T . This is
impossible.

44. Proof. Let U be the set of k vertices of a tournament T of order 2k for

which each vertex of U has outdegree r and let W be the set of k vertices of
T having outdegree s. Let T 0 be the subtournament induced by W . Then
 
k X X
= odT 0 v ≤ odT v = sk.
2
v∈W v∈W

Thus, s ≥ (k − 1)/2. Hence, the minimum possible value of s is (k −
1)/2 and

so k must be odd for s to attain this value. Let T1 and T2 be k−12 -regular
tournaments of order k and join every vertex of T1 to T2 . Then in T , every
vertex of T1 has outdegree (3k − 1)/2 and every vertex of T2 has outdegree
(k − 1)/2.

45. Proof. Let P = (u = v0 , v1 , v2 , . . . , vk = v) be a shortest u − v path. Then

every vertex vi (2 ≤ i ≤ k) is adjacent to u, that is, (vi , u) ∈ E(T ) for
2 ≤ i ≤ k, for otherwise, there exists a vertex vj for some j ∈ {2, 3, . . . , k}
such that (u, vj ) ∈ E(T ). Then the u − v path (u, vj , vj+1 , . . . , vk = v) is
shorter than P , a contradiction. Hence, id u ≥ k − 1.

46. Proof. Assume that T is a tournament of order n with ∆ − δ < n/2 that
is not strong. Then T contains k ≥ 2 strong components S1 , S2 , . . . , Sk such
that every vertex of Si is adjacent to every vertex of Sj for 1 ≤ i < j ≤ k.
 97

Let ni be the order of Si (1 ≤ i ≤ k). Let v and w be vertices of T such that

od v = ∆ and od w = δ. Then v belongs to S1 and w belongs to Sk . Thus,
n1 − 1 n1 + 1
∆≥ + (n − n1 ) = n −
2 2
and
nk − 1
δ≤ .
2
Thus,
n1 + 1 nk − 1
∆−δ ≥ n− −
2 2
n1 + nk
= n− .
2
Since n1 + nk ≤ n, it follows that
n n
∆−δ ≥n− = ,
2 2
which is a contradiction.
47. Proof. Let A be the set of vertices of T to which v is adjacent. Since od u <
od v, there is x ∈ A such that (x, u) ∈ E(T ) and so (u, v, x, u) is a triangle.
48. In the strong tournament of order 3, every vertex has outdegree 1. Next, let
T 0 be the subtournament induced by any four vertices of a tournament T of
order at least 4. Then the size of T 0 is 6. Since v∈V (T 0 ) od v = 6, we cannot
P
have four vertices of outdegree 1.
49. Proof. Let X be the set of vertices to which u is adjacent. Then v is
also adjacent to every vertex in X; for otherwise, there is a directed u − v
path (u, x, v) for some x ∈ X. Let Y be the set of vertices from which u is
adjacent. Then v is also adjacent from every vertex in Y , for otherwise, there
is a directed v − u path (v, y, u) for some y ∈ Y . Thus, idD u = idD v and
odD u = odD v.
50. Proof. Let (u = v0 , v1 , . . . , vk = v) be a u − v path of length k ≥ 2 in
~ v) = k, it follows that (u, vi ) ∈
T . Since d(u, / E(T ) for 2 ≤ i ≤ k and so
(vi , u) ∈ E(T ) for 2 ≤ i ≤ k. Therefore, for each i with 2 ≤ i ≤ k,

(u = v0 , v1 , v2 , . . . , vi−1 , v = vi , u = v0 )

is a cycle of length i + 1 in T .
51. Proof. Suppose that u and v do not lie on a common cycle. Furthermore,
assume, without loss of generality, that (u, v) ∈ E(T ). Let W be the set of
vertices to which v is adjacent in T ; that is,

W = {w : (v, w) ∈ E(T )}.

98 CHAPTER 7. DIGRAPHS

Since u and v do not lie on a common cycle, it follows that (u, w) ∈ E(T )
for all w ∈ W ; for otherwise, u and v lie on a triangle. Therefore, od u ≥
od v + 1 > od v.

52. Proof. We show that u and v belong to the same strong component of T ,
which then implies the result. Assume, to the contrary, that u and v belong to
distinct strong components of T , say S1 and S2 , respectively. We may assume
that (u, v) ∈ E(T ). Let

W = {x ∈ V (T ) : (v, x) ∈ E(T )}.

Then (u, w) ∈ E(T ) for every w ∈ W ; for otherwise, there exists some vertex
w0 ∈ W with (w0 , u) ∈ E(T ), producing a directed 3-cycle (u, v, w0 , u), contra-
dicting the fact that u and v belong to distinct strong components of T . Since
u and v belong to directed 3-cycles, there is u0 ∈ W such that (u, u0 ) ∈ E(T ).
However then, od u − od v ≥ 2, which is a contradiction. Hence, as claimed,
u and v belong to the same strong component of T .

Section 7.5. Kings in Tournaments

53. Proof. By the proof of Theorem 7.20, every vertex of maximum outdegree in
a tournament is a king. Since every vertex in a nontrivial regular tournament
has the same outdegree, every vertex is a king.
54. The statement is false. Let T be a tournament of order n = 2k ≥ 6 with
V (T ) = {v1 , v2 , . . . , v2k } such that vi is adjacent to vi+1 , vi+2 , . . . , vi+k−1
(addition modulo k) and where (vi+k , vi ) are arcs of T for 1 ≤ i ≤ k. Thus,
od v ≥ k −1 = (n−2)/2 for every vertex v of T and od vi = k −1 for 1 ≤ i ≤ k.
Then

d(v1 , v) ≤ 2 if and only if v ∈ {v2 , v2 , . . . , v2k−1 },

that is, v1 is not adjacent to vn and v1 is not adjacent to a vertex that is

adjacent to vn . Thus, v1 is not a king of T .
55. Let T be the tournament with V (T ) = {u, v, w, x} and
E(T ) = {(v, u), (u, w), (u, x), (w, v), (x, v), (w, x)}.
Then u, v, w are kings of T but x is not a king.
56. Let T be the tournament with V (T ) = {u, v, w, x, y} and
E(T ) = {(v, u), (u, w), (x, u), (u, y), (v, y), (v, x), (w, v), (w, y), (x, w), (y, x)}.
Then u, v, w, x are kings of T but y is not a king.
57. Solution. Assume that n ≥ 6 is even. Let Tn−3 is a regular tournament
of order n − 2. Then the tournament T of order n has V (T ) = {v1 , v2 , v3 } ∪
V (Tn−3 ) and
 99

E(T ) = {(v1 , v2 ), (v2 , v3 ), (v3 , v1 )} ∪ E(Tn−3 ) ∪ E1 ∪ E2 ∪ E3 ,

where E1 = {(v, v1 ) : v ∈ E(Tn−3 )}, E2 = {(v, v2 ) : v ∈ E(Tn−3 )} and
E3 = {(v3 , v) : v ∈ E(Tn−3 )}. Then v1 is the only vertex of T that is not
a king. [Note: For another possible construction of T , see the solution of
Exercise 54.]

58. Proof. Let T be a nontrivial tournament and let T~ be the converse of T .

Thus, T~ contains a king v. In T~ , for every vertex u distinct from v, either v
is adjacent to u or v is adjacent a vertex that is adjacent to u. Therefore, v
is a serf in T .

Section 7.6. Hamiltonian Tournaments

59. Proof. Since T is not a transitive tournament, either T is strong or T con-
tains k ≥ 2 strong components S1 , S2 , . . . , Sk such that for every two vertices
u, v ∈ V (T ) with u ∈ V (Si ) and v ∈ V (Sj ) with 1 ≤ i < j ≤ k, (u, v) is an
arc of T and S` is nontrivial for some ` with 1 ≤ ` ≤ k.
If T is strong, then T contains a Hamiltonian cycle C = (v1 , v2 , . . . , vn , v1 )
and so T contains a Hamiltonian path with initial vertex vj for each j ∈
{1, 2, . . . , n}.
Suppose that T is not strong. Thus, T has a strong component S` of order
at least 3, where 1 ≤ ` ≤ k. Let P = (v1 , v2 , . . . , vn ) be a Hamiltonian path
in T . Let r be the minimum positive integer and t the maximum positive
integer such that vr and vt belong to S` . Then there are at least t − r + 1
Hamiltonian paths in S` as above. Replacing vr , vr+1 , . . . , vt by these t − r + 1
paths produces t − r + 1 ≥ 3 Hamiltonian paths in T .
60. (a) If a tournament T of order 5 is not strong, then its largest strong com-
ponent S has order 4. The strong component S has five Hamiltonian
paths; so T itself has 1 · 5 = 5 Hamiltonian paths. Any other structure
for T has either one or three Hamiltonian paths. Thus, the maximum
number of Hamiltonian paths in T is five.
(b) If all strong components of T have order 1, then T is transitive and
there is only one Hamiltonian path. There are three Hamiltonian paths
in a strong tournament of order 3 and five in a strong tournament of
order 4. If T has x strong components of order 3 (x = 1, 2, 3) and all
other strong components have order 1, then T has 3x Hamiltonian paths.
If T has y strong components of order 4 (y = 1, 2) and all other strong
components have order 1, then T has 5y Hamiltonian paths. If T has
one strong component of order 3, one of order 4 and one of order 1, then
T has 1 · 3 · 5 = 15 Hamiltonian paths. Thus, the possible values of k
are 1, 3, 5, 9, 15, 25, 27.
61. There are three Hamiltonian paths in S1 with initial vertex v1 . Because of
the symmetry of S1 , it follows that S1 has 15 Hamiltonian paths. Thus, S2
has k/15 Hamiltonian paths.
100 CHAPTER 7. DIGRAPHS

62. The statement is false. Let D be the tournament of order 6 obtained from two
directed cycles C = (u, v, w, u) and C 0 = (x, y, z, x) by joining each vertex of
C to every vertex of C 0 . Then every vertex belongs to a cycle. Since there is
no x − u directed path in D, for example, it follows that D is not strong.

63. (a) The statement is false. The arc (u, v) in the strong tournament T of
order 4 shown in Figure 7.8 does not lie on a Hamiltonian cycle of T .

u ......
......................
........ .......
. .. .
...................... ...............
....... . . ...........
........ ....
........... ................
............. ...... .
........... ..........
...........
. .
... ..
. ...
.........
.................. .. ....
........ .... ...........................
...................... .
........

Figure 7.8: The tournament of Exercise 63

(b) The statement is false. The strong tournament T of order 4 shown in

Figure 7.8 is vertex-pancyclic by Theorem 7.27, but does not contain a
Hamiltonian u − v path.

64. Proof. Let C = (v1 , v2 , · · · , v` , v1 ) be an `-cycle of T . Then the digraph

D induced by the vertices of C is a strong tournament of order ` since D
contains a Hamiltonian cycle. By Corollary 7.29, D is pancyclic and so D,
and therefore T , has an s-cycle for every integer s with 3 ≤ s ≤ `.

65. There must be a strong component of order k, another strong component of

order k−1 or k and k trivial strong components. So, n ≥ k+(k−1)+k = 3k−1.
Chapter 8

Flows in Networks

Section 8.1. Networks

1. No, f is not a flow since f (s, t) = 3 > 2 = c(s, t).
2. (a) Let X = V (D) − {v} and let K = [X, X]. Since cap(K) = 9, it follows
that val(f ) ≤ 9.
(b) The value of the flow f defined on the network N in Figure 8.1 is the
net flow out of u, which is 9.
4,3
....... ........... ................
............. ...... ... ......
...... ........... ......
......
3,3 .... .
.....................
......
......
...... .....
......
......
...... 4,4
.....
.
........ .
..
... ...2,0
...........
... .. 2,0
............. ..
.
. ...
4,1
......
.........
.... . . ..
........ .................
.... ..... ......
. ..
..
......
...
6,4 .........
3,3 . . . ......
.
..
............
......
..........
............. .......... .............. ... ...
v
N: u ........
......
.... .
. .
. ..... . .......... . .
..
.
.................
..
.
..........
. .
......
.
..........
...... . ...
... . ...
......
...... 5,1 .....
..
...
2,2 .
............
..........
. . . .
...... . ... .. ...... ... ..
..........
............. ...... ...... ...... ......
...... ...... 3,0 . ...... 5,5
......
.
......
. .......
2,2 ...... ...
.......................
...
...........
..
......
................
...
........ ...... ...

4,3

Figure 8.1: The network N in Exercise 2

3. (a) Since the net flow out of the vertex y is 0, it follows that 4 = b + 2 and
so b = 2. The net flow out of the vertex w is also 0, so b + 2 = a + 3 or
4 = a + 3. Thus, a = 1. Since the net flow out of x is 0, 3 + 3 + c = 7
and so 6 + c = 7. Hence, c = 1.
(b) The value of f is the net flow out of u, which is val(f ) = 1 + 4 − a = 4.
Observe also that val(f ) is the net flow into v, which is 6 − 1 − c = 4.
4. Figure 8.2 shows a flow where the maximum flow value is 3. Since
f (t, y) = f (y, x) + f (y, w) + f (y, z) ≥ 3,
there is no flow where the maximum flow value along any arc is less than 3.

101
102 CHAPTER 8. FLOWS IN NETWORKS

r 2 s 1
....... ................. ......... ........... .........
......
...........
......
......
......
.......... ...... ..........
...... ..
...... ...
v
...... ...... ...... .....
1 ......
......
.
...
.
2 .............
.........
...
. .
....... 2 .
...
.
...... ....
................
...........
.............
1 ... ..
...... ......
......
1 .........
......
.......... ...
.
...
....... . ...... ...... ...
..........
..
.
2 .... .
......
..
.
......
.
.
...
.
. ....

N: u
.
..............
. ...........
......
.........
.........
.
............
...........
w
.
...
.
.. 1
...
...... .
...... t ......
......
...
...
.
...........
..
......... 1 ...
..
....... ..
....
. ..........
.
...
2 ........... ...................... .. ... .......
1 ... ..
...... .
... ...
........ ...
......
.
...
.. ......
...... 3 .....
.
.
...
.......
...... 2 ...
...
...
..... ......
..
...
.
...... ............... ..............
.............. ........................ ...
..........
...... ...
...
x 1 y 1 z

Figure 8.2: The network N in Exercise 4

5. Since f is a flow in D, the net flow out of t in D is 0, that is,

X X
f (t, x) − f (x, t) = 0.
x∈N + (t) x∈N − (t)

Since id t = 0, it follows that

X
f (x, t) = 0.
x∈N − (t)

Thus, x∈N + (t) f (t, x) = 0. Since f (t, x) ≥ 0 for each x ∈ N + (t), we have
P

f (t, x) = 0 for each x ∈ N + (t). This implies that for the flow f 0 , the net flow
out of each vertex in N 0 equals the net flow out of this vertex for the flow f .
Hence, f 0 is a flow on N 0 and val(f ) = val(f 0 ). A similar argument holds for
od t = 0.
6. (a) See Figure 8.3(a). The net flow out of u is 3 and the net flow out of v is
−3. The net flow out of every other vertex is 0.
(b) The maximum value of a flow in N is 4. It can be obtained from the
flow in (a) by augmenting along the semipath
(u, (x, u), x, (w, x), w, (w, t), t, (v, t), v)
(see Figure 8.3(b)).
(c) For X = {u}, the cut [X, X] is minimum and f (X, X) − f (X, X) = 4.
7. (a) Let U be the set of all maximal paths P in D whose initial vertex is u
and such that no arc of P belongs to A. Define the set X by
[
X= V (P ).
P ∈U

Clearly, u ∈ X. Since every u − v path in D contains at least one arc of

A, it follows that v ∈ X = V (D) − X and [X, X] is a u − v separating
set of arcs in D. It remains to show that [X, X] ⊆ A. Assume, to the
contrary, that there is an arc (x, y) ∈ [X, X] but (x, y) ∈
/ A. Since x ∈ X,
 103

s 3,2 t
.......
......... ................. .................
. .
...... ...... ... . ..........
..
..
..
...... ... .... ......
1,1 ......... .
. .
. ..
...........
...
....
... .. ..
............
.
... . ........
...............
. ...... 3,1
... ....
.
... ......
5,3 ..........
. .
.
.....
. .....
.
.. . 2,1 ......
......
......
... . ... .
.
..
... .
...........
3,3 . . ...........
. ... ..
.. . . ..
.
.
..
w .. ...........
2,1 ......
......
.........
.... .... ... ......
v
u .. ....
.....
. . . .
.....
... ..
... ......... .
......
.. ..
....
..........
....
..... . . ... ..
..... ....
.......... ..... ............. ..
......
........... 8,7 ..... ......... .............
..... . ...... .
..... . ..... 3,2
........... ..... .
2,1 .....
.....
..... .
...
.............. ... ...
3,1
......
......
...... .
..
..... 4,3
..... .......... ...... ........
. ............
.............. ......
........... .............. ......
........... ... ....
... .. ...
x 6,6 y 5,5 z
(a)
s....... 3,2 t
. . ................. ............
........... ...................
...... ..... ...
...... ..
......
1,1 ................... ... . .......
......
...... .
.....
.
...............
....
..............
.. .................
........ 3,0
..
.....
. .... 5,3 ......
..
... 2,2 ......
......
......
...... ... .. . ......
.......... .. .
3,3 .... ..
. . ........ ..
.
w 2,1 ......
..........
............. ........... . ...........
v
u .......
.....
......
.....
.
.
. . .
. .
................
......
......
....
... ..
.........
.....
........ .... .... .. .......
. . ......
.
........... .
........... 8,6 ..... ............ ...
...........
....... . ..... ..
3,2 .......
.......... ..... . .....
2,0 .....
.....
..... ....
.
............
.....
.. ..
3,1
......
......
..
.. .
..
.....
.... 4,3
..... ....... ...... .
........... ...................
... ..
...
...........
...... ............. ...........
...... .... ....
... ...
x 6,6 y 5,5 z
(b)

Figure 8.3: The network N in Exercise 6

there exists a u − x path Q in D such that no arc of Q belongs to A.

Then the u−y path Q0 obtained from Q by adding the arc (x, y) contains
no arc belonging to A. This implies that Q0 is a subpath of some path
in U and so y ∈ X, which is a contradiction.
(b) Let D be a u − v path of length 2, say (u, x, v). Let X = {u} and
A = E(D). Then [X, X] = {(u, x)}, which is a proper subset of A.
8. Proof. If (x, y) is an arc of D, where x, y ∈ X, then f (x, y) is a term in
the sum f + (x) and f (x, y) is a term is a term in the sum f − (y). Thus,
f (x, y) − f (x, y) = 0 occurs in f + (X) − f − (X). Therefore,
f + (X) − f − (X) = f (X, X) − f (X, X)
as desired.
9. (a) Proof. By Theorem 8.2, val(f ) = f + (X) − f − (X) and by Exercise 8,
f + (X) − f − (X) = f (X, X) − f (X, X). Thus,
val(f ) = f (X, X) − f (X, X) = c(X, X) − 0 = cap(K).
By Corollary 8.3, f is a maximum flow and K is a minimum cut.
(b) Proof. By the the Max-Flow Min-Cut Theorem (Theorem 8.6), the
value of a maximum flow f in a network N equals the capacity of a min-
imum cut K = [X, X]. Let c be the capacity function of N . Therefore,
val(f ) = f (X, X) − f (X, X) = c(X, X) = cap(K).
104 CHAPTER 8. FLOWS IN NETWORKS

Since f (X, X) ≥ 0, it follows (as in the proof of Theorem 8.10) that

val(f ) = f (X, X) − f (X, X)
≤ f (X, X) ≤ c(X, X) = cap(K) = val(f ).
Thus f (X, X) = 0 and c(X, X) = f (X, X). Therefore, f (a) = c(a) for
all a ∈ [X, X] and f (a) = 0 for all a ∈ [X, X].
10. (a) Assign a flow of 3 to each of the arcs (w, t), (t, s) and (s, w) and assign
a flow of 0 to every other arc.
(b) If the underlying digraph of a network contains a directed cycle, then
such a flow exists.

Section 8.2. The Max-Flow Min-Cut Theorem

11. After applying Algorithm 8.7 to the flow f , we obtain a new flow f 0 (see
Figure 8.4(a)) with 5 = val(f 0 ) > val(f ) = 4. After applying Algorithm 8.7
to the flow f 0 , we then obtain another flow f 00 (see Figure 8.4(b)) is obtained
with val(f 00 ) = 6.

s 2,1 t
........ ........... ............
........... ...... ............ ..........
...... ......
...... ......
......
3,1 ..... .
....................
1,1 ..
............... .
. .. . ......
...... 3,2
.
........ .... .......... ......
......
.......
. .
........ ..........
.............
..... ....
.
..
.........
4,4 .
.... .
......
..
3,3 1,1 ......
......
..........
(a) ..
............. ........... ............ ........... ....... ........... .. ..
u .............
........
......
. ...... ......
.... .
.
........ ......
..........
.
......... v
........
.........
................
x ..
.
....
y ............
.. ..
... . ...... ...
........... ... .............
........
2,2 ...... ........ ..
........ .......... ........
........ ..... ........
3,0 ........
........ . . .
........ ......... ..............
......... .
5,2
...... .....
......
w
A flow f 0 after augmenting f along
the semipath (u, (w, u), w, (v, w), v)

s 2,2 t
......... ...........
...... .............
........... ........... ...........
..... ......
..... ...... ......
3,2 ... .
....................
1,1 ............. . . .
. .. ......
...... 3,3
.......... ... ........... ......
...... .
.......
. ......... ..........
.
.... ..... ............
.
...
..
.......
.
4,4 .
...........
...........
.
3,3 1,1 ......
......
...........
(b) u ............. ........... .. . . . ... ..
... . . . ..... .. .
.............
........
...... ........ ...........
....
.... ...
... ...........
..........
.
......... v
........
.........
................
x .......
y ..... .
.... . ........
............ ... ..................
........
2,2 .... ............
........ ......... ........
........ ....... ........
3,0 ........
........ . . .
........ ........ ...............
......... . .
.
5,2
....... .......
......
w
A flow f after augmenting f 0 along 00

the semipath (u, (u, s), s, (s, t), t, (t, v), v)

Figure 8.4: The network N in Exercise 11

After applying Algorithm 8.7 to the flow f 00 , we learn that there is no f 00 -

augmenting semipath and so f 00 is a maximum flow. Since u and s are the only
 105

vertices labeled by the algorithm, K = [X, X] = {(u, x), (s, t)} is a minimum
cut, where X = {u, s}.
12. A maximum flow has value 7 (see Figure 8.5) and a minimum cut K = [X, X],
where X = {u}, X = V (D) − {u}, and D is the underlying digraph of the
network N .
s 4,3 t
........ ...........
...... ................
........... .... ..
..... ... ...........
3,3 ...... ...
.................... .. .. ......
...... 5,5
. ...
. . ... ......
...... ...... ......
..
.... . . . . ......
..
.............
. 2,2 .........
................
..........
.
..
4,4 r . .... . ......
......
...........
............. . . . .... .. ....
.
u ..........
.....
................ ...........
.
..... .....
........
.. v
.....
..... ... ... .........
........
.
..... .
.......
.............. 1,1 .....
. ....... 1,1
..... .....
. . .. ....
....... ....... .................
1,0
.....
.....
.....
.....
3,0 ...
.......
....... ...... 2,1
...................... ... ................................ .............
.........
.......
..... 6,2
.......................... ... ..
...
. .......... ........ ..... ...
.... ...
...
y
.................
.......... ....
.......

w 2,0 x

Figure 8.5: The network N in Exercise 12

13. (a) The statement is false. Let N be a network such that c(a) > 0 for every
arc a in the underlying digraph D and f is a flow in N with val(f ) > 0.
If each of f1 and f2 is the zero flow, then f1 and f2 agree on [X, X] and
[X, X] but f1 and f2 are not maximum flows.
(b) The statement is true. Proof. Since f1 and f2 are maximum flows and
K = [X, X] is a minimum cut in N , it follows by the Max-Flow Min-Cut
Theorem (Theorem 8.10) that val(f1 ) = val(f2 ) = cap(K) = c(X, X).
For i = 1, 2,
val(fi ) = fi (X, X) − fi (X, X)
≤ fi (X, X) ≤ c(X, X) = cap(K).
Since val(fi ) = cap(K) for i = 1, 2, it follows that fi (a) = c(a) for each
arc a ∈ [X, X] and fi (a) = 0 for each arc a ∈ [X, X].
14. See Figure 8.6, where c1 (u, w) = 1, c2 (u, w) = 2, c1 (w, v) = 3, c2 (w, v) = 4.
1, 2 3, 4
............. ...........
...... ............. .......... .............
... ... ...... ...
u w v

Figure 8.6: The network N in Exercise 14

Section 8.3. Menger Theorems for Digraphs

15. The digraph D0 constructed from D is shown in Figure 8.7. The maximum
number of pairwise arc-disjoint u − v paths in D0 is 1 and one such path is
P 0 = (u, w0 , w00 , x0 , x00 , v). This gives rise to the u − v path P = (u, w, x, v)
in D.
106 CHAPTER 8. FLOWS IN NETWORKS

w00 x0
............. ................... .............
......... ......
... ..
... .. ... ....
. ......... ....
. ... ......
......... ...
... ........
. .. ... ........ ......
0.......... . ... ....... ......... 00
w ...
.
... ....
... ...
.
.
...
...
... . ... x
... ..................
...... ... ... .......... ......
............ ... ..
........
......
.. . . .......... .......
.
. . . . . . ..........
...............
...
..... ......
. ...
. . .. ... .... ......
..... ... ... ...... . ... ......
... ... ... ..... . ... ......
...
..............
D0 :
.........
u ...........
......
...
...
. . ..
..
... ....
.
. .
.
...... .
...
...
...
.... ..
..... ....
.
v
...... .
... ............ .. . . ... ......
...... .. ... . . . ...
.
...................
... ....
. ..
... ... ...............
..... ... ... .........
......
....... .. ... ......
......
.................. ..... . ....... .... . ......
... ... .. ......... ... ..................
0 ........ ..... . 00
..
z .......
y ... .. .........
....... ..
..
.............. ..... ... ............
...... .. ... .....
... ... ... ...
... .. .. ..
............. ................
... .
y 00 z0

Figure 8.7: The digraph D0 in Exercise 15

16. The maximum number of internally disjoint u − v paths in D is 3. Three such

paths are P1 = (u, w1 , x1 , x2 , v), P2 = (u, w2 , x3 , v) and P3 = (u, w4 , w3 , x4 , v).

17. Suppose that S = {s1 , s2 , . . . , sr } and T = {t1 , t2 , . . . , tq }. Add a new vertex

s (the new source) and the arcs (s, si ), 1 ≤ i ≤ r, and P add a new vertex t (the
new sink) and the arcs (ti , t), 1 ≤ i ≤ q. Let C = e∈E(D) c(e) and assign C
as the capacity for each of the r + q new arcs. Call this new network N̂ with
underlying network D̂. Then there is a one-to-one correspondence between
the flows in N and the flows in N̂ . Furthermore, corresponding flows have the
same value.
Chapter 9

Automorphisms &
Reconstruction

Section 9.1. The Automorphism Group of a Graph

1. Observe that

Z2 × Z2 ∼
= Aut(G1 ) = {, (t u), (y z), (t u)(y z)}.
Aut(G2 ) = {, (1 2), (6 7), (1 2)(6 7),
(3 5)(1 6)(2 7), (3 5)(2 6)(1 7), (3 5)(1 7 2 6), (3 5)(1 6 2 7)}.

2. (a) See the table below.

 α1 α3 α3 φ1 φ2 φ3 φ4
  α1 α2 α3 φ1 φ2 φ3 φ4
α1 α1 α2 α3  φ4 φ3 φ1 φ2
α2 α2 α3  α1 φ2 φ1 φ4 φ3
α3 α3  α1 α2 φ3 φ4 φ2 φ1
φ1 φ1 φ3 φ2 φ4  α2 α1 α3
φ2 φ2 φ4 φ1 φ3 α2  α3 α1
φ3 φ3 φ2 φ4 φ1 α3 α1  α2
φ4 φ4 φ1 φ3 φ2 α1 α3 α2 

(b) and (c) See Figure 9.1.

3. Yes. K4 − e.

4. Let C5 = (1, 2, 3, 4, 5, 1). Then

Aut(C5 ) = {, (1 2 3 4 5), (1 3 5 2 4), (1 4 2 5 3), (1 5 4 3 2),

(2 5)(3 4), (1 3)(4 5), (2 4)(1 5), (3 5)(1 2), (1 4)(2 3)}.

107
108 CHAPTER 9. AUTOMORPHISMS & RECONSTRUCTION

.. α2
..
α..1 α3 .............. ..
...... .....
....
..... ...... .......
.... ..
..........
... ....... ...... .....
... ....... ...... ..... φ2
... .......
... ......  ....
.
...... .....
. α1 ................................. .....
... ..
...................
...... . .......... . ... ..... ........... .......... ..... ..
...
... ................. ... ... ....... .............. φ .
... .......... ......... ....
... ....................... ... ... ....... .......... 4 .... .
... .
.....
.... ... .............................. ..... ...
G: .
... ..... ..... ..... ...... ..
....... ....
...... .........
..... .......
..............
...
...
.....
.....
.....
.
................ .....
.....
...
..
...... .... ....... ......
. ... ..... .... ... ..... ..
..... ....... .
..... ... ....... ..... .....
.
.. ... ... ..... .......
... ..... ...........
G: ...
... ............. .
... .......................................
.... .... ..
. .
...
.

.... .. .
... . .
... ...... ... ... ...... ...
... ...... .... ...
. ..
.....
......
.
... ..
.
.. . .
. . .. ..
.
... ............. ....... ..............................
............... .... .....
..
...............
......
......
......
...
... φ 1
...
... α ..
..
. 3 .
..
...
... ....
....... . . . .... . .
. .
. . ... ... ... .................. ............. ...
...
... ... ...
............................. .. . ... ... ...
φ1 ..........
..... .
.
..
...
. .
.
.. α 2 .....
.. ...
.
...
....
........ ..
. .
....... φ4 ...
..... ...
... ... ....
..
..... ... ... ... ..... ..... ... . .
.
. .
...
..
.. . ... .... ..... . .. .....
.....
..... .
.
. .. . ... .
... .
....
. .....
..... .... ... .....
..... .. ... . .. .... ..
... ... .........
.. ..... ... ... ......
..... ... ... ... .. .... ...... .. . ...........
...... ... ............. ...................
... . ... ....
.............. .... .....
....
....
...
φ2 φ3 φ3

Figure 9.1: The graphs G and G

............... .......
...........
.....
.....
.
..
.....
..
............
........ ................ ............
....... ............
...
..... .....
..... .....
..... .....
. ... ..
.......... .. . . ..
..
..........
.......... ....... .. ..
......... ......
.....
.....
..
......
..
................. ...............

Figure 9.2: The graph in Exercise 5

5. For the graph G in Figure 9.2, Aut(G) ∼

= Z4 .
6. Consider G = 2Kr where the two copies of Kr in G have vertex sets U and
V , respectively. Let u ∈ U . An automorphism φ maps u to a vertex in U or
to a vertex in V . We consider these two cases.

Case 1. φ(u) ∈ U . Then each vertex in U is mapped by φ into a vertex in

U and each vertex in V is mapped into a vertex in V . Thus in this case, for
each automorphism of Kr with vertex set U , there are r! automorphisms of
Kr with vertex set V . Since there are r! automorphisms of Kr with vertex
set U , the number of such automorphisms of G is (r!)2 .

Case 2. φ(u) ∈ V . Then each vertex in U is mapped by φ into a vertex in

V and each vertex in V is mapped into a vertex in U . Again, the number of
such automorphisms is (r!)2 .

Thus | Aut(G)| = (r!)2 + (r!)2 = 2(r!)2 . Since the order of G is n = 2r, it

(2r)!
follows by Theorem 9.3 that the number of distinct labelings of G is 2(r!) 2.

(2r)!
Thus the number of distinct labelings of Kr,r = G is 2(r!) 2 . In general, for the

graph G = kKr , | Aut(G)| = k!(r!)k . Thus the number of distinct labelings of

(kr)!
Kr,r,...,r = G is k!(r!)k.

7. We have already noted that 6 is the smallest order of a nontrivial graph whose
automorphism group consists only of the identity automorphism; so (k, n) is
 109

not realizable when k = n and 2 ≤ n ≤ 5. These are the only pairs that are not
realizable. First, notice that all pairs (k, n) when k = 1 are realizable because
Kn has this property. Also, all pairs (n, n) where n ≥ 6 are realizable because
the graph G of order n and size n with V (G) = {v1 , v2 , . . . , vn } consisting of
the path (v1 , v2 , . . . , vn−1 ) and the edges vn vn−3 and vn vn−2 has this property.
For n ≥ 3, the star K1,n−1 has two orbits and so (2, n) is realizable. For
3 ≤ k < n, let G be constructed from the path P = (v1 , v2 , . . . , vk−1 ) and
the complete graph H = Kn−k+2 where V (H) = {vk−1 , vk−2 , . . . , vn } by
identifying the two vertices in P and H labeled by vk−1 . Then G is a graph
of order n with k orbits.
8. All pairs (k, n) have this property except (1, 2). The graph Kn shows that
(k, n) is realizable for k = n ≥ 1. For 2 ≤ k < n, the graph Kk + K n−k shows
that (k, n) is realizable. The graph K1,n−1 shows that (k, n) is realizable for
k = 1 and n ≥ 3.
9. Let n = 2k for some positive integer k and let G be the bipartite graph
with partite sets U = {u1 , u2 , . . . , uk } and W = {w1 , w2 , . . . , wk } where ui
(1 ≤ i ≤ k) is adjacent to wj if i + j > k. Then deg ui = deg wi = i for
1 ≤ i ≤ k and the only vertex similar to ui is wi .
10. The graph H is a complete k-partite graph where k is the number of orbits
of G. Since the orbits of G produce a partition of V (G) into k subsets, the
graph H is a union of k complete graphs and so H is a complete k-partite
graph.
11. Since the image of a cut-vertex in an automorphism of a graph is a cut-
vertex, the graph G3 is not vertex-transitive. For some vertex v in the graph
G4 , G4 [N (v)] = 2K2 , while for other vertices v, G4 [N (v)] = K1,3 . Therefore,
G4 is not vertex-transitive.
12. (1) Z1 ∼
= Aut(D1 ),
(2) Z2 ∼
= Aut(D2 ) = {, (u w)(v x)},
(3) Aut(D3 ) ∼
= S3 since u and w are fixed, but v, x, y can be permuted among
themselves.

Section 9.2. Cayley Color Graphs

13. Let Γ ∼ = Z4 be generated by the element a and let ∆1 = {a} and ∆2 =
{a, a2 , a3 }. Then D∆1 (Γ) and D∆2 (Γ) are shown in Figure 9.3.

(a) Γ = {g1 = e, g2 = a, g3 = a2 , g4 = a3 } and ∆1 = {a}.

(b) Γ = {g1 = e, g2 = a, g3 = a2 , g4 = a3 } and ∆2 = {a, a2 , a3 }.

14. Proof. Let E be the arc set of D∆ (Γ). First, suppose that α is a color-
preserving automorphism of D∆ (Γ) and let g ∈ Γ and h ∈ ∆. Then (g, gh) ∈
E and its color is h. Since α is an automorphism, (α(g), α(gh)) ∈ E and
110 CHAPTER 9. AUTOMORPHISMS & RECONSTRUCTION

a3
.........
.. ............... .....................
................... ......
...... .....
........ .....
....
.... a ...
.....
e .
...
.
.... . .......
. .
.......... .
.
...........
....
..
............. a
......... ...........
.... ....
.
...... .....
.
. ....... .....
.....
.. ..... ...
a........... ... ..... ..... ........
..........
... ..... .....
e ................ ............. a .. ..... .
...... ....
...
...... ... . 2 ........ . .
... 2
.... ...
....
....
a .....
..... ........
.
...
..
a ...
...
a ....... ... ..
...... a a3 ...
... a
.
.
.. ..
. .
..
.
.
.
.
.
......
.
... .......
.
.
.
.
... ...
...
... a3
. ...
... ........... .....
..
.....
. .
..... a ...
.
..
.... .... ..
.......... .......... ............. .......... .... ... ..
..... ..
3 ...... ....... ...
a2 ......
.....
..... .....
..... ...
a a
....
....
..... ......
..... .....
.....
. ......
...
...
..... .
.. ... ..
. .
...... .......... .....
....................
..... .. .................
3 ........... ......
.. . 2
D∆1 (Γ) a ...
..... a ...... a
..... .....
...... ......
....... ..............
........... ...........
.............................

a3
D∆2 (Γ)

Figure 9.3: Graphs for Exercise 9

furthermore, its color is also h since α is color-preserving. Therefore, α(gh) =

(α(g))h.
For the converse, suppose that α is a permutation of V (D∆ (Γ)) such that
α(gh) = (α(g))h for every g ∈ Γ and h ∈ ∆. Let (g1 , g2 ) ∈ E. Then
g2 = g1 h for some h ∈ ∆ and so α(g2 ) = α(g1 h) = (α(g1 ))h by assumption.
This then implies that (α(g1 ), α(g2 )) ∈ E. We can similarly verify that if
(α(g10 ), α(g20 )) ∈ E, then (g10 , g20 ) ∈ E. Therefore, α is an automorphism.
Furthermore, both (g1 , g2 ) = (g1 , g1 h) and (α(g1 ), α(g2 )) = (α(g1 ), (α(g1 ))h)
are colored h and so α is color-preserving.

15. The group of color-preserving automorphisms of the Cayley color graph D∆ (Γ)
consists of

, α = (g1 g2 g3 g4 )(g5 g6 g7 g8 ),
(g1 g3 )(g2 g4 )(g5 g7 )(g6 g8 ), (g1 g4 g3 g2 )(g5 g8 g7 g6 ),
β = (g1 g5 )(g2 g6 )(g3 g7 )(g4 g8 ), (g1 g6 g3 g8 )(g2 g7 g4 g5 ),
(g1 g7 )(g2 g8 )(g3 g5 )(g4 g6 ), (g1 g8 g3 g6 )(g2 g5 g4 g7 ).

This group is isomorphic to Z4 × Z2 and so by Theorem 9.5, Γ ∼ = Z4 × Z2

which is generated by a and b. In this case, D∆ (Γ) is color-preserving and
Aut(D∆ (Γ)) ∼
= Z4 × Z2 , which is generated by α and β.

16. Let G be the graph constructed in the proof of Theorem 9.6 from the Cayley
color graph D∆ (Γ). If, instead of using paths of length 2k − 1 and 2k, we use
paths of length rk − 1 and rk for r ≥ 3, then the resulting graph Gr has the
desired property.
 111

17. Let Γ ∼
= Zn = {e, a, a2 , · · · , an−1 } and ∆ = {a}. Then D∆ (Γ) is the directed
cycle of order n all of whose arcs are colored a. The underlying graph of
D∆ (Γ) is Cn .
18. Consider Exercise 15. The underlying graph of the Cayley color graph D∆ (Γ),
where Γ ∼
= Z4 × Z2 and ∆ = {a, b}, is Q3 and so Q3 is a Cayley graph.

Section 9.3. The Reconstruction Problem

19. Note that the order of G is n = 6, the size of G is m = 9 and the degree
sequence of G is s : 4, 4, 3, 3, 2, 2. Now consider G5 . Let v ∈ V (G) such that
G5 = G − v (see Figure 9.4(a)). We know that deg v = m − m5 = 4. Since
G contains two vertices of degree 4 and no vertex of degree 1, it follows that
vx, vz ∈ E(G). Necessarily, v is adjacent to two of the vertices u, w and y.
If vy ∈
/ E(G), then v is adjacent to both u and w in G. However then, the
resulting graph G produces G − u ∼ = G − w, as shown in Figure 9.4(b), which
is not isomorphic to any of the subgraphs Gi (1 ≤ i ≤ 6). Thus, v is adjacent
to y and one of u and w. Since u and w are similar, these two situations
are equivalent. Thus, the only possible graph for G is the graph shown in
Figure 9.4(c). Observe that G − vi ∼ = Gi for 1 ≤ i ≤ 6.

...
v5 v3
. ... ...
u ................................. ...................
...... .....
.......
............
...... .........
................
.. .....
.....
.......
....... x y z .....
..
...... .....
........ .. ...
....... .....
.....
.....
.....
........... ... ...
..................................................... ............ .............. .............. ..... .........
......
...........
....... .
... v1 .........
......
......
.....
.....
..... ......
.
... ...
.... v2
...
.......
. ...... ..... ........
... .................
................ ............
w .... ............. ............. ........ ........
... ...
v4 v6
(a) (b) (c)

Figure 9.4: The graphs Exercise 19

20. Note that the order of G is n = 5, the size of G is m = 7 and the degree
sequence of G is s : 3, 3, 3, 3, 2. Let v ∈ V (G) such that G − v = G2 . Then
deg v = m − m2 = 3 and so there is only one vertex in G that is not adjacent
to v in G. Since all the vertices of G2 are similar, the only possible graph for
G is the graph shown in Figure 9.5.

v1................ .....
.............
.... v5
.....
.........
...
....
.
.....
v 2....................
...
..
...... .........
.
.... .....
..... .....
..... .....
..... .....
..
......
. .....
.....
..
v4 ............. ..........
....... v3
....

Figure 9.5: The graph in Exercise 20

21. The order of G is n = 7, the size of G is m = 12 and the degree sequence of

G is s : 4, 4, 4, 3, 3, 3, 3. Now consider G1 = G − v1 . We know that deg v1 =
112 CHAPTER 9. AUTOMORPHISMS & RECONSTRUCTION

m − m1 = 4. Since there are three vertices of degree 4 and four vertices of

degree 3 in G, it follows that v1 is adjacent to the four vertices of degree 2 in
G1 . Thus G = K3,4 with partite sets U = {v1 , v2 , v3 } and W = {v4 , v5 , v6 , v7 }.

22. The outdegrees of the tournaments D1 and D2 of Figure 5.17 are the same,
namely 1, 2, 2, 2, 3. If D1 and D2 were isomorphic, then every isomorphism
would have to map the vertex of outdegree 1 in D1 to the vertex of outdegree
1 in D2 . Furthermore, the vertex in D1 adjacent from the vertex of outdegree
1 must map into the vertex in D2 adjacent from the vertex of outdegree
1. However, this vertex in D1 has outdegree 2, while this vertex in D2 has
outdegree 3. Hence no isomorphism exists.

23. (a) Proof. Let G be a reconstructible graph. Then G can be reconstructed

from its vertex-deleted subgraphs G − v, v ∈ V (G). Consider the com-
plement G of G. Since G − v = G − v, it follows that by taking the
complement of each vertex-deleted subgraph G − v of G, the graph G
can be reconstructed. Thus G is reconstructible.
(b) Proof. Let G be a graph whose complement G is disconnected. By
Theorem 9.10, G is reconstructible. By (a), G is reconstructible.

24. Proof. If G = Cn for some odd integer n ≥ 3, then G is reconstructible

since G is regular. If G 6= Cn but G is not bipartite, then G contains an
odd cycle of length less than n. Thus G contains a vertex v such that G − v
has an odd cycle. Since the property that a graph contains an odd cycle is
recognizable, the property of being bipartite is also recognizable.

25. Since none of the graphs Gi (1 ≤ i ≤ 11) are connected, it follows that G
is disconnected. The graph G has order n = 11, size m = 10 and degree
sequence

2, 2, 2, 2, 2, 2, 2, 2, 2, 1, 1.

Following the proof of Theorem 9.10, we observe that k(G) = 3 since min{k(Gi ) :
1 ≤ i ≤ 11} = 3. Since F = C4 is a component of maximum order among the
graphs Gi , C4 is a component of maximum order in G. For a vertex v of C4 ,
let F 0 = F − v = P3 . Let ` be the minimum number of components in any
graph Gi that are isomorphic to F . Then G has ` + 1 components isomorphic
to F ; that is, G has two components that are C4 . Among the graphs Gi
(1 ≤ i ≤ 11) with k(G) = 3 components having minimum number of compo-
nents isomorphic to C4 , the graphs Gi (1 ≤ i ≤ 8) have the maximum number
of components isomorphic to F 0 = P3 , namely, each such graph Gi has two
components isomorphic to P3 . Replacing a copy of F 0 by F in one of the
graphs Gi (1 ≤ i ≤ 8) gives us the graph G, that is, G = 2C4 + P3 . (Observe
also that once it has been determined that two of the three components of
G are C4 , the remaining component has degree sequence 2, 1, 1 and so this
component must be P3 .)
 113

26. If only the two subgraphs G − x and G − y are given, then it is impossible to
determine whether G contains the edge xy.
27. No. The graph obtained by connecting two vertices of K4 by a new path of
length 3 has order 6, size 9 and degree sequence 4, 4, 3, 3, 2, 2.

28. The only solution is the graph obtained from P5 = (v1 , v2 , v3 , v4 , v5 ) by adding
two new vertices x, y and the two edges xv3 and yv4 .
29. The only solution is the graph obtained from C6 = (v1 , v2 , v3 , v4 , v5 , v6 , v1 ) by
adding the two edges v2 v5 and v3 v5 .

30. If some solution G is Hamiltonian, then G − v has a Hamiltonian path for

each v ∈ V (G). The graphs G2 and G3 do not contain Hamiltonian paths.
114 CHAPTER 9. AUTOMORPHISMS & RECONSTRUCTION
Chapter 10

Planar Graphs

Section 10.1. The Euler Identity

1. Proof. Certainly, a graph G is planar if and only if each of its components
is planar, so we may assume that G is connected. Since every subgraph of a
planar graph is planar, it follows that every block of a planar graph is planar.
For the converse, we employ induction on the number of blocks of G. If G has
only one block and this block is planar, then, of course, G is planar. Assume
that every graph with fewer than k ≥ 2 blocks, each of which is planar, is a
planar graph and suppose that G has k blocks, all of which are planar. Let
B be an end-block of G and denote by v the cut-vertex of G belonging to B.
Delete from G all vertices of B different from v, calling the resulting graph
G0 . By the inductive hypothesis, G0 is a planar graph. Since the block B is
planar, it may be embedded in the plane so that v lies on the exterior region.
In any region of a plane embedding of G0 containing v, the plane block B may
now be suitably placed so that the two vertices of G0 and B labeled v are
identified. The result is a planar embedding of G; hence, G is planar.
2. Proof. For a fixed positive integer n, we proceed by induction on m. For
m = 0, G = Kn and so k(G) = n and r = 1. Thus n − 0 + 1 = 1 + n.
Assume that the result is true for all planar graphs of order n and size less
than m. Let G be a planar graph of order n and size m with k components
and let e ∈ E(G). Then the planar graph G − e has order n0 = n and size
m0 = m − 1. Suppose that G − e has k 0 components and r0 regions. By the
induction hypothesis, n0 − m0 + r0 = 1 + k 0 . We consider two cases.

Case 1. e is a bridge. Then k = k 0 − 1 and r0 = r. Then

n − m + r = n0 − (m0 + 1) + r0 = (n0 − m0 + r0 ) − 1 = 1 + k 0 − 1 = k 0 = 1 + k.

Case 2. e is not a bridge. Then k 0 = k and r = r0 + 1. Then

n − m + r = n0 − (m0 + 1) + (r0 + 1) = n0 − m0 + r0 = 1 + k 0 = 1 + k.

115
116 CHAPTER 10. PLANAR GRAPHS

This completes the proof.

3. Let G be the graph consisting of two paths P = (u, v, w, x) and Q = (y1 , y2 , z2 , z1 )
where the end-vertices y1 and z1 are joined to every vertex of P (see Fig-
ure 10.1). The complement G of G consists of P and Q, both of which are
paths of order 4, where the end-vertices y2 and z2 of Q are joined to every
vertex of P . That is, G and G are isomorphic and so G is self-complementary.
Hence, G is also planar.
u.......
... .....
.......... ........
..... .....
.
.......
. .....
. .....
..... .....
....
.....
.......
..
. ........v
...................
.....
.....
.....
. ... ..... ...... .....
... ....
G: ...
..... .
..
........ .
..
.. .........
...
.....
.....
.....
.... ..... ...... .....
..
.
.
....
..
... ......... .
. .
. .... ....
. . . ..
.
........
. w
..
..... ..... .............
.........
.. ......
.....
....
. ...
.
... ........ .............. ........ ......... .........
.
.. ...
. .... ...... .....
... ......... ...... ........ .
....
... ... ...... ... ...
...... x ... ......... ......... ........
.
..... ...... ........ ..... ... ........ ..... .....
..... ...... ........ ................. ...... ................................... ........ ...... ....
.
.................................. ................................ .................. ..................................
....................................
.
..............................................
.
. .
.... .. .. .. .......................
.... ............ .... ..... ........
. .......
... ..
y1 y2 z2 z1

Figure 10.1: The graph in Exercise 3

4. Proof. Suppose that there exists a planar graph G of order n ≥ 11 such

that G is also planar. Since the size of a planar graph of order n ≥ 11 is at

3n − 6 and the sum of the sizes G and2 G is the size of Kn , it follows that
most
n
2 is at most 2(3n − 6) = 6n − 12. Since n − 13n + 24 = (n − 2)(n − 11) + 2, it
follows that n2 − 13n + 24 > 0 when n ≥ 11. However, since 2

n − 13n + 24 > 0,
2 n
it follows that n − n > 12n − 24 = 4(3n − 6) and so 2 > 2(3n − 6), which
is a contradiction.
5. (a) Proof. Let G be a 3-regular planar graph of order n and size m con-
taining no triangles or 4-cycles. Since G is 3-regular, 2m = 3n. Let there
be a planar embedding of G resulting in r regions. Denote by N the
sum of the edges on the boundary of a region over all r regions. Since G
contains no triangles or 4-cycles, the boundary of each region contains
at least 5 edges and so N ≥ 5r. Furthermore, N counts each edge at
most twice and so N ≤ 2m. Thus, 5r ≤ N ≤ 2m. Since n − m + r = 2,
it follows that n = 2 + m − r and so

10n = 20 + 10m − 10r ≥ 20 + 10m − 4m = 20 + 6m = 20 + 9n.

Therefore, n ≥ 20.
(b) Since the Petersen graph is a 3-regular graph of order 10 containing no
triangles or 4-cycles, it follows by (a) that it is not planar.
6. (a) Assume, to the contrary, that G contains k vertices of degree 5, where
k ≤ 11. So G contains n − k vertices of degree 6 or more. Thus, 2m ≥
5k + 6(n − k) = 6n − k and so m ≥ 3n − k2 ≥ 3n − 11 2 > 3n − 6. This
implies that G is not planar, which is a contradiction.
 117

(b) The graph of the icosahedron is a 5-regular graph of order 12.

7. Since the size of G is 3(n − 1) = 3n − 3 > 3n − 6, it follows that G is not
planar.
8. Suppose that there are r regions. If we sum the number of edges on the
boundaries of the regions, we obtain 2m = 3(r − 1) + k and so 3r = 2m +
3 − k. Since n − m + r = 2, it follows that 3n − 3m + 3r = 6. Therefore,
3n − 3m + (2m + 3 − k) = 6. Solving for m, we obtain m = 3n − k − 3.
9. Let G be a connected k-regular plane graph of order n and size m resulting in
r regions where r = n. By the Euler Identity (Theorem 10.1), n − m + r = 2.
Since m = nk/2 and n = r, it follows that n − m + r = 2n − nk/2 = 2.
Therefore, k = 4 − 4/n. Since k is an integer, n | 4 and so n ∈ {1, 2, 4}. If
n = 1, then G = K1 and r = 1. If n = 2, then k = 2 and r = 2. There is no
such graph. If n = 4, then k = 3 and r = 4. Thus, G = K4 . Therefore, either
G = K1 or G = K4 .
10. Claim: F5 = 12 + F7 .
Proof. In this case, V − E + F = 2 and 2E = 3V3 = 5V5 + 6V6 + 7F7 .
Therefore,

12 = 6V − 6E + 6F = 6V3 − 6E + 6(F5 + F6 + F7 )
= (10F5 + 12F6 + 14F7 ) − (15F5 + 18F6 + 21F7 ) + (6F5 + 6F6 + 6F7 )
= F5 − F7 .

Hence, F5 = 12 + F7 .

Section 10.2. Maximal Planar Graphs

11. The graphs G1 and G2 of Figure 10.2 are maximal planar graphs of order 6.
The graph G1 = K2,2,2 is 4-regular, while G2 is not.
......
.............
... .....
.......... .
.... ...
.
... ........ ...
............ .
... .. ... ... ... .................. .....
.
.... .... .... ..... .. ... ........ ..... ...
. . . .
... .. ..... ..... .... ....... ...... ....... ....
.. ... .. .. .... ..
... ... ... ... ... ..... ..................... ........ .....
G1 : ... ...... .. .
............. ..... G2 : ... ..... ...... ...... ..... ...
.... ............... . . . ... ...... ........ ...... ..... ...
.
. .
... ........ ................ ........ ..... ........................ ..... ...... .... ..
...... ........
. ..... ........... .. . .. ... ....
............... .... ...... .
.
. .. . .... ............ ..............
. . .............. ..... ....................................................
... ..... .............. .................... ................................. .......................
......................... ..... ..
... ....................... ......

Figure 10.2: The graphs in Exercise 11

12. Proof. Let G be a 4-regular maximal planar graph of order n and size m.
Since G is 4-regular, 2m = 4n or m = 2n. Since G is a maximal planar
graph, m = 3n − 6 and so 2n = 3n − 6. Thus, n = 6 and so G is a 4-regular
graph of order 6. There is only one such graph, namely, K2,2,2 (the graph of
the octahedron). Note that if G is a 4-regular graph of order 6, then G is a
1-regular graph of order 6 and so G = 3K2 . Thus, such a graph is unique.
118 CHAPTER 10. PLANAR GRAPHS

13. Proof. We have seen that if G is a maximal planar graph of order n ≥ 3

and size m, then m = 3n − 6. It remains to verify the converse. Assume,
to the contrary, that there exists a planar graph of order n ≥ 3 and size
m = 3n − 6 that is not maximal planar. Then edges can be added to G to
produce a maximal planar graph G0 of order n and size m0 > m = 3n − 6.
Since m0 > 3n − 6, this is a contradiction.

14. Proof. Certainly every maximal planar graph of order 4 or more is 2-

connected. Assume, to the contrary, that there exists a maximal planar
graph G of order 4 or more such that κ(G) = 2. Then there is a vertex-
cut S = {u, v}. Since G is maximal planar, the neighbors of u lie on a cycle
C of length degG u in G. Furthermore, v is a cut-vertex of the graph G − u
and so there are two or more branches of G − u at v. Since C belongs only
to one of these branches, say B, the vertex u is adjacent only to vertices of B
and to no vertex of any other branch, except possibly v. However then, v is a
cut-vertex of G as well, which is impossible.

15. The statement is false. Let G be a maximal planar graph of order n = 6 and
size m = 3n − 6 = 12. Since n − m + r = 2, it follows that r = 2 − n + m =
2 − 6 + 12 = 8; so there are 7 interior regions of G. Let H be the graph
obtained from G by placing a vertex in each interior region and joining it to
the three vertices on its boundary. Thus, no two added vertices are adjacent
in H. Let S = V (G). Then k(H − S) = 7. Since k(H − S) > |S|, it follows
by Theorem 6.11 that H is not Hamiltonian.

16. Let G be a maximal planar graph of order n ≥ 3 and size m, a planar embed-
ding of which results in r regions where r = n. Since n − m + r = 2, it follows
that n − (3n − 6) + n = 2 and so n = 4. Thus, G = K4 .

17. The graph G is nearly maximal planar as there is a planar embedding (see
Figure 10.3) where the boundary of every region is a cycle and only one
boundary is not a triangle.
.........
...........................
....... ........
....... ........
............. ........
....... ............
............... ... .
.......... ...........
....... .......
....... .......
G: .......
.......
............................
...........
... ..
..............................
................ .........
.. .........
......... ................
................. ........
...

Figure 10.3: The graph G in Exercise 17

18. (a) Observe that (n − 1) + (3n − 6) = n2 and so (n − 2)(n − 7) = 0. Since

K 2 is not maximal planar, n 6= 2. Thus, n = 7.

(b) The tree T of order 7 obtained by subdividing each edge of K1,3 exactly
once has the property that T is maximal planar.
 119

19. Proof. Assume, to the contrary, that the result is false. Then there exists
a maximal planar graph G of order n ≥ 4 and size m that satisfies none of
(1)-(3). Thus, δ(G) = 5 and m = 3n − 6; so
X
(6 − deg v) = 6n − (6n − 12) = 12, (10.1)
v∈V (G)

which is a consequence of Theorem 10.13. Assign to each vertex v of G a

charge of 6 − deg v. By (10.1), the sum of the charges of the vertices of G
is 12. Therefore, each vertex of degree 5 is assigned a charge of +1, each
vertex of degree 6 is assigned a charge of 0 and each vertex of degree 7 or
more is assigned a negative charge. By assumption, at most one neighbor
of each vertex of degree 5 has degree 5 or 6. For each vertex v of G having
degree 5, we redistribute its charge of +1 equally to four neighbors of v not
having degree 5 or 6. Upon doing this, the sum of the charges of the vertices
of G remains 12. The charge of every vertex of degree 5 or 6 is now 0. Let
u be a vertex of degree k where k ≥ 7. Initially, the charge of u is 6 − k.
After the redistribution of charges, at most 2k
 
3 neighbors of u will transfer a
charge of +1/4 to u, resulting in a charge of at most (6 − k) + 2k
 
 2k  3 /4. Since
(6 − k) + 3 /4 ≤ 0, the sum of the charges of the vertices of G is at most
0, which is impossible.
20. Let T be a tree of order n ≥ 2 and size m = n − 1. Hence, the sum of the
charges of the vertices of T is
X X X
(2 − deg v) = 2− deg v = 2n − 2m = 2n − 2(n − 1) = 2.
v∈V (G) v∈V (G) v∈V (G)

Section 10.3. Characterizations of Planar Graphs

21. (a) The statement is true. Proof. Assume that G is planar. Since H is a
subgraph of G, it follows that H is planar. Since H 0 is a subdivision of
H, it follows that H 0 is planar.
(b) Since no two vertices of U are adjacent in G0 , it follows that G[U ] = K k .
22. Proof. Assume, to the contrary, that G = T + e1 + e2 + e3 is nonplanar.
We consider two cases.

Case 1. G contains a subdivision H of K5 . Assume that H contains k vertices

of degree 2. Then the order of H is 5 + k and the size of H is 10 + k. On the
other hand, H contains at most 4 + k edges of T . This implies that the size
of H is at most (4 + k) + 3 = 7 + k < 10 + k, which is a contradiction.

Case 2. G contains a subdivision H of K3,3 . Assume that H contains k

vertices of degree 2. Then the order of H is 6 + k and the size of H is 9 + k.
On the other hand, H contains at most 5 + k edges of T . This implies that
the size of H is at most (5 + k) + 3 = 8 + k < 9 + k, which is a contradiction.
120 CHAPTER 10. PLANAR GRAPHS

23. Proof. Assume, to the contrary, that the graph G of order n ≥ 5 contains no
subdivision of K3,3 but is nonplanar. Necessarily then, G contains a subgraph
H of order k ≥ 5 that is a subdivision of K5 . Since H contains at most k − 1
edges belonging to T , the subgraph H contains at most (k−1)+5 = k+4 edges.
Since H is a subdivision of K5 , it follows that H contains 10 + (k − 5) = k + 5
edges. This is a contradiction.
24. (a) n = 8, m = 16 and 3n − 6 = 18.
(b) Since m < 3n − 6, this provides no information about the planarity of
K4  K2 .
(c) The graph K4  K2 is nonplanar as it contains a subdivision of K5 . See
Figure 10.4.
.....
............
s s.......................
...... .... ... .....
.... . ... ....
... ... ... ...

...
...
.. ...
.
..... .....
... ....
s .............
...
... ...
.
...
...
...
..... ..... ... ..
... ....... .. ..
s ..
.
..
... ... .... ... . .. ..
... ... ..... .
.... .....
...
.
.
..
... ... .
... ... ...
..... .. ....
....
s
...... .....
............. .....
.... ....................
...

Figure 10.4: A subdivision of K5 in K4  K2 in Exercise 24

25. Claim: The only such graph is G = K5 .

Proof. Since G is nonplanar, G contains either a subdivision of K5 or of
K3,3 . Since G − e is planar for each edge e of G, G itself must be a subdivision
of K5 or of K3,3 . If G is a subdivision of K3,3 , then G contains 6 vertices of
degree 3 and k vertices of degree 2 for some k ≥ 0. Then n = 6 + k and
m = 9 + k. Since m = 3n − 5, it follows that 9 + k = 3(6 + k) − 5 = 13 + 3k. So
2k = −4, which is impossible. If G is a subdivision of K5 , then G contains 5
vertices of degree 4 and k vertices of degree 2 for some k ≥ 0. Then n = 5 + k
and m = 10+k. Since m = 3n−5, it follows that 10+k = 3(5+k)−5 = 10+3k.
So, 2k = 0. Thus, k = 0 and G = K5 .
26. First, notice that S 2,2 = P4 = P4 and that S 2,3 is a proper subgraph of K5 ;
so S 2,2 and S 2,3 are planar. The planar graphs S 2,4 and S 3,3 are shown in
Figure 10.5. If a + b ≥ 7, then S a,b contains Ka+b−2 as a subgraph and so
S a,b is nonplanar. Thus, S a,b is planar if and only if a + b ≤ 6.
27. (a) Proof. Assume, to the contrary, that G contains H = K3,3 as a sub-
graph. Let v be the vertex of G that is not a vertex of H. Then G − v
contains H as a subgraph and so G − v is nonplanar, which is a contra-
diction.
(b) The graph G of order 7 obtained by subdividing one edge of K3,3 has
this property.
28. The graphs C 3 , C 4 and C 5 are clearly planar. The graph C 6 = K3  K2 is
also planar. The graph C 7 contains a subdivision of K3,3 and is nonplanar.
 121

............. ........
... ...........
... .....
...
. ...
... ...
......... ... ...
............... .... ...
...
... ... ... .. ..
... ...... ...
... . ..
.. ........ ..... ... ...... .....
.
... ... ..... . ........ ....
S 2,4 : .
..
. .. ..
... .................... .....
.
.. S 3,3 : ..
.
.. ........
. ...
... ......
.
....... ...
.........
... ..... ..... ... .......... ............
.... .........
.
..... ...
...... ... ............... ............ ..
..........
.. ... ..... ... ...... ... .....
... ..... ......... ..... ... ..... .....
... ...... ..................... ... .. ..... .....
.
............ ............. ......... ..............
.
.....
..... .....
....... ....... ....... ....... ..... ....
.
.................... ....... ....... ..... .....
..
... .. ............... ....................
................ ..... .... ... ...
...
.... ....

Figure 10.5: The graphs in Exercise 26

See Figure 10.6, where C7 = (v, u1 , u2 , u3 , w1 , w2 , w3 , v), U = {u1 , u2 , u3 } and

W = {w1 , w2 , w3 }.

v
............
....
.. ..
.. .....
s u1
.
w .......
..... ..... s .
.... ....
3 ............................ .... .... ...........................
.

..... ........... . . .
..... ......... ................... ........
..... ... ................... ... ......
..... ......... .......... ...
... . . . .......
..............
..... ....
........ ........... ... ..........
w .. .. s
....................
.. .....
......
.... ...
2 ........................... ..... ...................... ...... ...............
........
... s u2
....... .. .. .
......... ........ ........ .................
... ............... .............
.. .
.. ... ...... ............ ..
.. ..... ...............................
s
............. .........
..... ........
......... .
s..............
..

w1 u3

Figure 10.6: A subdivision of K3,3 in C 7 in Exercise 28

n

n(n−3)
For n ≥ 8, the graph C n has size m = 2 −n = 2 . Since n ≥ 8, it
follows that

n2 − 9n + 12 = (n2 − 9n + 8) + 4 = (n − 1)(n − 8) + 4 ≥ 4 > 0.

Therefore, n2 − 3n > 6n − 12 and so

n(n − 3)
m= > 3n − 6.
2

Hence, C n is nonplanar.

29. The graph Cn2 is nonplanar if and only if n ≥ 5 and n is odd. First observe that
C32 = C3 , which is planar. For n ≥ 4 and n even, Cn2 is planar as there exists
a planar embedding of Cn2 . See Figure 10.7, where Cn = (v1 , v2 , . . . , vn , v1 )
and n is even.
Since C52 = K5 is nonplanar, it remains to show that if n ≥ 7 and n is odd,
then Cn2 is nonplanar. Let Cn = (v1 , v2 , . . . , vn , v1 ). We show that Cn2 contains
a spanning subgraph H that is a subdivision of K3,3 . Let U = {v1 , v2 , vn−2 }
and W = {v3 , vn , vn−1 }. Then

{v1 v3 , v1 vn , v1 vn−1 , v2 v3 , v2 vn , vn−2 vn−1 , vn−2 vn } ⊆ E(H).

122 CHAPTER 10. PLANAR GRAPHS
.........................
............ .......
....... ......
...... .....
..
...... vs .....
.... 1 ...
..
.. ....................................................
. ... ..
. .
.. ............. .......... .......... ........ .....
v s .. v
s
.
. ...... . ...... ...... ... 2
....
n .
.
................. .
.....
. ..
. ...... ..... .. ....
......... .............
. .. . .
.. ..
...... ..... .......
...... ... ....
..... .... ...... ...... ... .....
.... ... ...... ......
...... .....
. .....
...
. .... .......... ....... ... .
...
...
.... .. ........ . ...
sv s
. . ....... ...
... ........... ........ ...
....
...
.......
... ... n−1 v
3 .... ..... ... ...
..
... .... .... . .. ..
... . .. ..
.
. ... .
...
... ... ...
... .
.
. .
..
.... ... ... .
.. ... .....
s
..... .. .
v n−2 ...... .......
.........
.. ... ... ......
..........
.......
s
.... .
.
.
.. .... 4 v
... ...
...
r r ..
..
..
r r r ...
r

Figure 10.7: A planar embedding of Cn2 for even n in Exercise 29

Also, the paths (v2 , v4 , v6 , . . . , vn−1 ) and (v3 , v5 , v7 , . . . , vn−2 ) belong to H.

Thus, H is a subdivision of K3,3 and Cn2 is nonplanar. See Figure 10.8 for H
when n = 9.
v 1
s ........
... ...
........... ..
.......... ..........................
v s
...
sv
. .... .
.......... ...... ...... ...........
9 ............................................................................................................................. 2
.. ....... ...... .. ...
...... ... ...
............... ...... ... ..
...... ...... ...
........ ........ ...
...... .. ... ........ ...
v8 s
...... ..
......
...... ....
.
.
... ...........
...
... ...
s v3
....
... ... ... ... ...
... ... ... ... ...
... .... ... .
. .
.
... ..... ..
... ...... ....
... ... ... ... ....
... .. .... ... ..
... ... ... . .. .....

v7 s
.......... ... ... ......
... ..
.... ...
.............................
.. ... .................
..........
v4
...........
................ .. . . .................
........... ........ ...
...
... .......... ................. ..
... ............................. ..
. ........... .................
.................. ........
.
....
v6 v5

Figure 10.8: A subdivision H of K3,3 in C92 in Exercise 29

30. Although Kn is a contraction of Kn+1 , the graph Kn+1 is not a subdivision

of Kn .
31. The statement is false. Although K4 is a minor of K5 and K4 is planar, K5
is not planar.
32. (a) The graph G contains K3,3 (see Figure 10.9).

...
t ...............
... ....
.... .....
..
..... .....
...
. .....
.
t
..
. . .....
.
.. . . ..
..............
. ........
........ .... .
. .........
........ .
... ........
........
....................
... ...
..................
...
........ ..............
t

Figure 10.9: The subgraph K3,3 in the graph G of Exercise 32(a)

 123

(b) The graph G does not contain a subdivision of K5 since only four vertices
of G have degree 4 or more.
(c) Contracting the edge joining the two vertices of degree 3 in G produces
the graph K5 .

33. (a) In this case, m = (4 · 10)/2 = 20 and 3n − 6 = 24. Since m ≤ 3n − 6, no

conclusion can be made.
(b) The statement is true (see the graph in Figure 10.10 ).
..........
............................
...... ... .......
..
........
. ... .......
...... ... ......
. .. ........ ... ......
......
......... . .
... ......
.
. ....... ..
..... . .. .. . ............. ...... .... ......
......
... .......... .... ..... ......... ....
....................................... .... .. . ...... ...
.......
... ..
...... .............
... ........................ .
... . ............. ...................................
.
. .
... ........
.
.
............ .
...
..... ....
... . .. . .. .
... ... ... ..... ...
...
... .... ... ..... .
... ..... ....
... ... .... ... ..... .
... ..
. ... ... ......... ...
... .. .
............. .. ... ..
... ...
. . ... . . .................. ....
... .. . ... ......
... .. ........ ............. . ..
..
... ... ..... ....... .....
... .. ..... ....... ... ....
............ ....... ... ..
... . ....
..... .. ..
... .... .......
...

Figure 10.10: A planar 2-connected 4-regular graph of order 10 in Exercise 33

(c) Since H is vertex-transitive, if H contains K5 as a subgraph, then there

is a subgraph K5 containing u1 , for example. Since the neighbors u2 and
u5 of u1 are nonadjacent, for example, H contains no K5 . For each i with
1 ≤ i ≤ 5, let Vi = {ui , vi } (see Figure 10.21 in Chapter 10). Since the
subgraph of H induced by each set Vi is connected and since for every
two sets Vi and Vj (i 6= j), there is a vertex in Vi is adjacent to a vertex
in Vj , it follows that K5 is a minor of H. Therefore, H is nonplanar.
(d) Proof. Assume, to the contrary, that there is a graph G of order n ≥ 5
and size m ≥ 3n − 5 such that no subgraph of G has minimum degree 4.
Thus δ(F ) ≤ 3 for every subgraph F of G. In particular, δ(G) ≤ 3. Let
v1 ∈ V (G) such that degG (v1 ) ≤ 3. Consider G1 = G−v1 . The size m1 of
G1 is at least (3n−5)−3. Since δ(G1 ) ≤ 3, there is a vertex v2 ∈ G1 such
that degG1 (v2 ) ≤ 3. Now consider G2 = G1 − v2 = G − {v1 , v2 }. The size
m2 of G2 is at least (3n−5)−2·3. In general, let Gi = G−{v1 , v2 , . . . , vi }
such that degGi−1 (vi ) ≤ 3 and the size of mi of Gi is at least (3n − 5) − 3i
for 1 ≤ i ≤ n − 5, where G0 = G. In particular, the size of Gn−5 is at
least (3n − 5) − 3(n − 5) = 3n − 5 − 3n + 15 = 10. Since Gn−5 has order
5, it follows that Gn−5 = K5 , which produces a contradiction.

34. (a) Since no graph can contain an odd number of odd vertices, the minimum
possible order of such a graph G is at least 12. Since the graph G of order
12 in Figure 10.11(a) has a subdivision of K5 , the minimum number is
12.
(b) The graph H contains a subdivision of K3,3 (see Figure 10.11(b)).
124 CHAPTER 10. PLANAR GRAPHS

....
... ... u1 u2
.................

. . .... .....
...... .. .. ......
...... .... ..... ..........
. . .........
s ..
... .....
..............
... .... ............
...
.............
.
.... ...... . ..
. .. .. ... .....
.... ...... .... ..... ........... ... ..... ...
....
... ...... . ... ........... .... ..
v . ...
v1
... ..........
... .. ....... .......... ..... .
.. . . ...
. ...
...
......... 1 ......
.
... ...
...
.s
vs2
.
... .. ........... ... ... .. ... ... .. ............. ... ............. ...
.............. ............ ... ......
... . v ... ... ...... ...........
.
.... ..... .... ... ... ... ..
.. 2 ... ... .............
....... .......
....... .... .. .
. ...
..... .... . .... . ... ..
...... ...
......
..... ... ... v .... .. ..
. ... v 6................. ......
......
v
.
.
........
.
.
.................... u6
.
.....
..
............. ...
6
.............
.
. .
.
......
. ....
.. ................................
.
.
...
.... . s u3
...
.............
...
s
...... ...
... s
...... 3
......
......
.. .............
...
..... ...... ... .
.
.. .... ..
......
...... ... ...........
.....
.....
..... ......
......
...
... ...
..
.
. ...
...
v 3 ....
.
. ...... .
...... ...............
.......... ...
..
.............. ......
.... ...
... .
. .. .. ....... .
s
. .......
... ...
... ...
...
v ........ ... .. ... ...... ................
...
... .. . ...
... 5 ............. ....... ... ................ ....
... . .. ... .. ...
...
...
. ....
...
. . ...
...
... .....
..... v 4 ...
.... ..
..
.
. v5 v4....
... ... ..
................................................................................ .......
s
... ... ....... ... ... .... ....
....... . . ..... ..... .....
..... .. ... ... . ... ....
..... ..... ..... ..... ...
.......... .
.............. u5
.......
. ... u4
(b) (c)
(a)

Figure 10.11: The graphs in Exercise 34

The graph H does not contain a subdivision of K5 ; for suppose that F is

a subdivision of K5 in H. We may assume that the five solid vertices, as
shown in Figure 10.11(c), are the vertices of degree 4 in F . Some vertex
of degree 4 in F , say v2 , is not adjacent to the vertex v5 . Then there do
not exist a v2 − v1 path and a v2 − v3 path containing neither v4 nor v6
that are internally disjoint.
(c) Since H contains a subdivision of K3,3 , it follows that H contains K3,3
as a minor. Contracting the edges ui vi (1 ≤ i ≤ 6) produces K6 , so H
has K5 as a minor.
(d) Either (b) or (c) shows that H is nonplanar.
35. Claim. If G is a connected graph of order n ≥ 4 such that G ∨ K1 is
outerplanar, then G = Pn .
Proof. Suppose that G 6= Pn . Assume first that G is not a tree. If G
contains a triangle, then G ∨ K1 contains K4 , which is impossible. If G
contains a k-cycle, k ≥ 4, then G ∨ K1 contains a subdivision of K2,3 , again,
an impossibility. Hence, G is a tree. If G contains a vertex of degree 3 or
more, then G ∨ K1 contains K2,3 , a contradiction. Hence, in all cases, G ∨ K1
is not outerplanar.
36. (a) Proof. Let v ∈ V (G) and consider the induced subgraph G − v of G.
Then the order of G − v is p = n − 1 and its size is q = m − degG v. Since
G has property πk , it follows that
 
k+1
m − degG v ≤ k(n − 1) − .
2

Since m = kn − k+1


2 , we have
   
k+1 k+1
kn − − degG v ≤ k(n − 1) −
2 2
and so degG v ≥ k.
 125

(b) Proof. Assume, to the cobtrary, that ω(G) ≥ k + 2. Then G contains

an induced subgraph H = Kk+2 of order p = k + 2 and size is q = k+2


2 .
Since G has property πk ,

k 2 + 3k
   
k+1 k+1
q ≤ kp − = k(k + 2) − =
2 2 2
2
 
k + 3k + 2 k+2
< = = q,
2 2

which is a contradiction.
(c) Every maximal outerplanar graph of order n ≥ 3 has property π2 . The
graph K2 ∨ K 3 has property π2 but is not outerplanar since it contains
K2,3 as a subgraph. For n ≥ 5, the graph G = K2 ∨ K n−2 of order n and
size m = 2n − 3 has property π2 but is not outerplanar since it contains
K2,3 as a subgraph.
(d) Every maximal planar graph of order n ≥ 4 has property π3 . The graph
K3 ∨ K 3 has property π3 but is not planar since it contains K3,3 as
a subgraph. For n ≥ 6, the graph G = K3 ∨ K n−3 of order n and
size m = 2n − 6 has property π3 but is not planar since it contains K3,3
as a subgraph.

Section 10.4. Hamiltonian Planar Graphs

37. Proof. Assume, to the contrary, that G = K2,3 is Hamiltonian. Then G
contains a Hamiltonian cycle C. Consider the planar embedding of K2,3 in
Figure 10.12.
.........
...................................
.......... ..........
..........
......
............. ..........
..........
......... ..........
.......... ..........
............... .........
. ...... ............
... ........ ....
.........
......... .......
.........
......... .
.............
.........
......... ......
.......
......... ... ...............
........... ............
.......

Figure 10.12: A planar embedding of K2,3 in Exercise 37

Thus for i 6= 4, ri = 0 and ri0 = 0; while r4 + r40 = 3 and r4 ≥ 1 and r40 ≥ 1.

Hence, r4 − r40 = 1 or r4 − r40 = −1. In either case, it follows by Theorem 10.27
that
Xn
0= (i − 2)(ri − ri0 ) = 2(r4 − r40 ) = ±2,
i=3

producing a contradiction.
38. Consider the graphs G1 and G2 in Figure 10.13.

(a) Proof. Assume, to the contrary, that G1 is Hamiltonian. Then G1

contains a Hamiltonian cycle C. Thus, ri = 0 and ri0 = 0 for i 6= 4 and
126 CHAPTER 10. PLANAR GRAPHS

............... ............. ............

..................
. .... ......... ...... .. .. ......
.....
..... ..... ...... .... ..... ..........
.
..... ....... .
.
..... .... ...
... ..........
..
......... .......
.. .....
.... .... ... ... ...... ... ... ......
... ......... ...... .. ... ......
..... . .
....... ... ....... ......
..
.....
. .
. .
.... .... . . . .
.
.
...
. ...
. . .... .... ..........
........
.... . ........
..
..... ... ............. .. ..
........... ... .. ......
. .......... ....
. ...
......... ... .. .... ......... .. ..
..
......
. ... ..
.
. ............
.. .
. .
... ..
.
.
..... .................. .
....
... ... ............ ..... ...
G1 : .........
.... ...
............. G2 : ...
...
....
.... .......
...........
.. . ..
..
.
..... ...
.
.......
..... ... ......... ............................................................... ....
. ..... ... .. . ....... . ... .. .
..
..... ..... ... ... ... ..
..
... ..... . ...
.
..............
. .........
... ... ............. ............
... ... ... ...

Figure 10.13: The graphs G1 and G2 in Exercise 38

r4 + r40 = 7. Thus, r4 − r40 6= 0. By Theorem 10.27,

n
X
0= (i − 2)(ri − ri0 ) = 2(r4 − r40 ) 6= 0,
i=3

producing a contradiction.
(b) Proof. Assume, to the contrary, that G2 is Hamiltonian. Then G2
contains a Hamiltonian cycle C. Thus, ri = 0 and ri0 = 0 for i 6= 4, 5;
while r4 + r40 = 1 and r5 + r50 = 6. By Theorem 10.27,
n
X
0= (i − 2)(ri − ri0 ) = 2(r4 − r40 ) + 3(r5 − r50 ).
i=3

Since r4 − r40 = ±1 and 3(r5 − r50 ) = −2(r4 − r40 ), it follows that 3 | 2, a

contradiction.

39. Proof. Assume, to the contrary, that the Grinberg graph G is Hamiltonian.
We apply Theorem 10.27 to a Hamiltonian cycle of G. If i 6= 5, 8, 9, then
ri = ri0 = 0. Thus,

3(r5 − r50 ) + 6(r8 − r80 ) + 7(r9 − r90 ) = 0

and so

3(r5 − r50 ) + 6(r8 − r80 ) = 3[(r5 − r50 ) + 2(r8 − r80 )]

= −7(r9 − r90 ).

Thus 3 | 7(r9 − r90 ). Since 3 and 7 are relatively prime, 3 | (r9 − r90 ), which is
a contradiction since r9 = 0 and r90 = 1.

40. Proof. Assume, to the contrary, that the Herschel graph H is Hamiltonian.
We apply Theorem 10.27 to a Hamiltonian cycle of H. If i 6= 4, then ri =
ri0 = 0. Thus, 2(r4 − r40 ) = 0, implying that r4 = r40 , but r4 + r40 = 9, which is
impossible.
 127

41. Proof. Assume, to the contrary, that G has a Hamiltonian cycle C that
contains both e and f . See Figure 10.14.
.....
qqq ... ..
..... .....
...... .... ...........
.......
. qqq ......
qqq
... ......
....
...... ......
......
qqq e
......
........ ......
qqq
..
. ......
.. . .. .... ......
.. . ...... ......
......
....
............
... .. ..
...................
.
R1 .
.
.
.. . .......
.
... .....
.... ..... .........
R2 ......
.............
. ..
..... ..
... ........... .......
.
.. .....
.. . .
. .
. . ........... ..........
........ ... ........ ..... .....
... ......... .... .. . ....
.... ...... ... ... .
.. ...
...
... ... .. .......... ...
....... . ..
... ...
...
. .
...
... ...
... ... ... ...
... ...
...
...
... R3 ...
...
... .
...
.. ..
.
...
... ..... .. ...
......
qqq
... ..
..... ...... ..... .... ...

f qqqqqqq
... ..... ........ ..
...
... ... .
...
...
... ... ...
..
qq
... ...
..
...
R4 ...

qqq
... ... ....
. .... .. .
...... .......
..... ...... .... ......
..... .....

Figure 10.14: The graph in Exercise 41

Then one of the regions R1 and R2 is in the interior of C and the other is in
the exterior of C. Similarly, one of R3 and R4 is in the interior of C and the
other is in the exterior of C. Thus, r4 ≥ 2 and r40 ≥ 2. Since r4 + r40 = 5, it
follows that

r4 − r40 = 1 or r4 − r40 = −1.

If i 6= 4, 5, then ri = ri0 = 0. Thus, by Theorem 10.27,

2(r4 − r40 ) + 3(r5 − r50 ) = 0,

implying that 3 divides 2(r4 − r40 ) and so 3 | 2, which is impossible.

128 CHAPTER 10. PLANAR GRAPHS
Chapter 11

Nonplanar Graphs

Section 11.1. The Crossing Number of a Graph

1. See Figure 11.1.

...... ...........
.. ........
.......... ...
.................. .......... ......... ......... .....
.... .. ........ ................ ......... .... .. ......
.... ..... .............. ......... . . . . .. . . ........ .............. ...
.. ... ... ....... ..
........ ....... . . . .
... . ..... ... ...
...
... ... ..... ............... ..... ... ...
.. ... ... ......
. . ..
. ..... ......... ......... ..... ...
...
. ... ... .
. . .
... ......... . .
.... ... ... ............. . ... . ...
... ... ... ... ...... ... ........... ..... .... ...
... ... ...... ..... .. ..... ...
... ... .
.... . . .
.............. ... ......... . ...
.... ... .. . ....
... ... .... .... ....... ....... ... ..
. . . ..
... ... ... .......... .
.
. ............ . ... .
...
... .
.. ....... .... .
..... ....... .... . .
... . . . . ..... .. ..
... ... .. ... ... .........
...
... ... ..... ......... ..
....... ........
..... ...
... .
... ... ......... .......... ...... .................. .............. ...
... . .. ... ............ .......... ..... ... ..
... ..................................... ..................... ....
... ... ....................................................................................................... ... ....
.... ....... ......... ...
..... .... ..... .....
..... ...... ..... .....
...... ..... ..... ......
...... .......
....... ....... .... ....... ..........
.................. .... ....
................................................................................

Figure 11.1: A drawing of K7 with nine crossings in Exercise 33

2. Since K3,3 is nonplanar, cr(K3,3 ) ≥ 1. Figure 11.2(a) shows that cr(K3,3 ) = 1.

.........
.............. ......................
....... ......
...... ....
..... ..... ... ... .........

....
....
. ...
.....
. u
.......................
.....
....
...
... ....
...
..................
.....
..... ..
.......
.... ...
..... .... .....
...
... .. ..
..... ... ... ......... ... ... ................. ...
.... .
. .... ....... .
.... . .
.. .. ....... ..... ... .... ... ...
........
...........................
... ......... ..
. ... ..
.. ....... ............... ...
u ...... .. . .. . .. ............ . .. ...
...... ..... ... ........... .. ...
... ... .... ..... ..... ...
..... ......... .
. . ... .
... .... ..............
... ........ ...................... .... ............ ... ... ..... ... ............... . ...
... ................... ........... ... ... ... ... ... ... ..... ......... ...
.
.... ............. ... ...
.... ... .. ..... ... ... ... ...
. . .... .. ... .. ..
........ . .... . ... .. .... .
..... ........ .. .. .... .. .
. ... ...
.. ... ... ... .. ... ...
.......... .
..... ....... .... ....... .......
u ... ... ... . .
u
.
......... .. .
... ... . .... ... .. ... ........... ...
... ..... .....
u
... .......
... ..
... ............ ... .................. u ...
...
.......
.......
....... .... .... ............
.
... ...
...
...
.....
..... .....
.....................
..
..
...
.... ............ ........... .. ... .......
....... ... ... ... ... .. ..
..... ......... . ... ....... ..... .. ... .................... ...
.
...... ..... ... . .. .
....... ........ . .
...
. . ... ..... .....
..... .....
.......... ...... ... ......... ... .........
....................... ..... ..... ............ ...........
..... ..... ... .. ... ...
......
....... ...
. ....... .... ....
.......... ...
.................................
(a) (b) (c)

Figure 11.2: The graphs in Exercises 2, 5 and 7

129
130 CHAPTER 11. NONPLANAR GRAPHS

3. Figure 11.3 shows a drawing of K5,5 with 16 crossings. Therefore, cr(K5,5 ) ≤

16.
t .
...........
............... ........
..... .. .. ..... .......
..... .. .. ..... .......
. . .. ..... .... ..... .......... .............
.. ... ..... .......
..... ... ..... .......
..... ...
... ..... .......
.....
t
... ..... .......
..... .. . ..... .......
.. ......
. .... ......................................... ..... .......
.......
.. .
.... . . . .. .................. ........................................................... .......
. ..
..... .... ..... ...... .... ....... ..... .. ........... .................
.. . ............. ..... ................ .............
.. .. .. .
..
... ................ . ....... ........
t t t t
.
t
. .
.
.. .. .. .
. .
....... ... ..
. ..
............... ................ .......
................... ......... . .. ..... .......................
. ........... ......... . .
.................. ...........
.................. .... ......... ...........................
............ .............
........... .............. ..................
.... .... .... .......... .......... ................. ....... ......
..... ....... ............. .. ..... .... ..... ........................... ......................................... ........................
............. .... ....................................................... ...... .
t
... ...
..... ...... ... .. ..................................... .. ........ .....
..... ....... ... .... . .... ... ... ..... .... ........ ..........
..... ...... ..... ........ .............. ....
..... ....... . . .. .. ..... ......
..... ...... ..... ..... ... ... ......... .................. .
...
..... ...... .. .. . . .
........ ... .... .... .......... ....................
. ....
..... ......
..... ....... .. .. ...... ........ ..... ......
............ .. ....... ........ ..... ......
t
..... .....
..... ... .............................. .
..
..
.....
..... . .... .. ... ....
..... .... . .... ..........
..... ... ... ..... ..
..... ... ... ....... ..........
..... .. ... ..... ......
..... ... .. ..... ......
........ ...... .......
t .... ..............
..........

Figure 11.3: A drawing of K5,5 with 16 crossings

4. Since K2,2,2 is planar, cr(K2,2,2 ) = 0.

5. Since K1,2,3 contains K3,3 as a subgraph, K1,2,3 is nonplanar and cr(K1,2,3 ) ≥

1. Figure 11.2(b) shows that cr(K1,2,3 ) = 1.

6. Since there is a drawing of C3  C3 with three crossings (see Figure 11.4(a)),

it follows that cr(C3  C3 ) ≤ 3. Since C3  C3 is nonplanar, cr(C3  C3 ) ≥ 1.
Assume, to the contrary, that cr(C3  C3 ) = 1. Let there be given a drawing
of C3  C3 with one crossing. Suppose that e is an edge that is crossed in this
drawing. Then C3  C3 − e is planar. The edges of C3  C3 are equivalent.
Because (C3  C3 ) − e contains a subdivision of K3,3 (see Figure 11.4(b)), it
follows that (C3  C3 ) − e is nonplanar, producing a contradiction.

qq
........ ................
................ ... ...
...
qqq
...
..... ... ...
...
...
... ... .......... .....
.
...
q .....................
e qqqq
.. ........ ... ...
.... ............. ...
... .. .
. .. .
. .. ...
... ... ... ... .. ........
. ...
qqq
... ... .... ... ... ...
.. .. . .. ...
s
... ... ... ... ... ...
... . . ... ... .
.. ...
. ....... .... ... .... . .
...
...
.... .... .
... ....
. ... . ... ... .
.... .......
. ...
. ... .... .
..... ..... ...
..
...
... ...
. .....
. ....
.
.....
...
.... .... .
....
. ...
.
...
. ... . ... .....
. ..
.............
...
...
.
... .
... .......... ...
.
...
.
...
... .... ........... .. .
. . . . . ...
.
... ... ..... ...
. ............... ...
. . .
. .. . ... .... .......... ....... ................ ...
.. .. ................ ........... ........... ................... ........ ..
.. .
. ...
. ..
...... ........... .....
.
. . .... ... ...... ...... .....
.
.
.... ... ......... .... ... . .
...... ... ....... ....
. ... .... .
.. .
..... ...... ....
..
s
.. .
. ....... ...... .. ..
.. .. ... . .
.... ... ....... ....
. .. .. .. .
......................................................................................................... ......... .....
..
... ......... .......
. ..
... .
. . ..
... .............................................
.. . ..... ................. . .
.. .... ...... .......
. .
. .
... ...... ....... ....
... ..
.. .
.. . .
..... ..... ...
... ...... ...... ..... ... ....... ..... ..... ...
... ........ ........
.. ....... ............
s
...... .... .. .......
................................... ............... ................ ... .......
... ................................................................................................................................................................................
.
.. ......................................... .......................... .
... ..........................................................................................................................................

(a) (b)

Figure 11.4: Graphs in Exercise 6

7. (a) See Figure 11.2(c).

(b) The statement is false and (a) provides a counterexample.
 131

8. Since cr(C3  Ct ) ≥ cr(C3  Ct ) = t, it follows that cr(C3  Ct ) ≥ t. As

there is a straight-line embedding of C3  Ct with t crossings (see Figure 11.5
for t = 5), cr(C3  Ct ) ≤ t. Thus, cr(C3  Ct ) = t.
.....
.... ....
..... ....... ..........
..... .. .....
.
........ .... .....
. .....
..... .... .....
.
. . ...... . .
. .... .....
.....
.
. ...... .
.
.... .........
.... .. .....
..
. ..... ...... ... .........
... .
.....
.....
....
. .
... ....... .....
. .. . . .
.......... . .........
. ....... ...... .
. .. ........
.....
.....
. ....... .. .... ................ .......... .
.....
.....
..... .... .... .....
. . ..
..................... .
......
..
. .... .......
..
.. ... ...................................
.... . .
. . . . . .. ....
........... .......... ........ ....
. .....
. .. ... ....... ..
... ...... .............. .. ....
.
. ......
.
. . ... ... .......... . ....
. ..
... ..... ...... ................... ..... .... ........
... ............ ............ .. ... . .
......... .... . ...
... ................ ....... .. .. ...
... .. . ..
...
... ...
...
. ...
. ..
.
.
..
.
.
...
... ...
... ... ... ... ...
... ... ... ... ... ...
... ... ... .. .. ..
... . . .
. ...
. ..
. ..
.
.
... ... ... . .
... ... . ...... ... ...
... ... ....................................................................................... ....
... ... .. ..... .. .
...
... ..... ... ............................ ....
. ... . .. . . . .
... ... ... ... ............... ......... ..... ....
... ..... ......................................... ..... .. ..
... ... ...... ..... ......
....... .............. ....... ...
.
.... .......................................................................................................
. .. ...
....

Figure 11.5: A straight-line embedding of C3  C5 with five crossings

9. See Figure 11.6.

3 ........
..................
..... .....
..... .....
... ..... .....
..... .....
..
...... .....
.....
.... .
...... .
...
...
.. ..
.. .
.
6
.
. .....
. .
.....
.....
.....
.
.....
. ........ .................. .....
......
. ....... ... ..... .........
. .....
.....
......
. ........ .... ... ........ .....
...
. ..... .. ... ...... .....
... ..
. . . .
....... .....
....
. ..... .. ... . .....
.
.....
.
...
.
.
.....
... 3
.....
.
.
.. ..
......
..............
........
. . 3
..............
. .......
.
. .......
.....
...
..
..
.... ........ ...
..
............ . ...... .... ...... ...
. . ........ ....... .........
.
.. . . ..
. ... .. . ..
. .. ......... ........ ......
.... ........ .....
. .... ......
. . ..
... ........ ........
... .
... ........ ..... .... .
.. .
...........
.
.. ......... ......... ........
. .. ..
.. .. .. .. . ..
. ............ . .
. ........ ......... .......
.. .. ... ...... 6
..... ...... ........ ..................... ........... .
........... ............. ........... ........
.
..... ...... ........ ...
. . . . .................................................. . ... .. .
. ..
. ...........
. .... .
........... . . .
........... .................................
. . . . .......................................
. ...
..... .
...... .
..... .................. ..... ......................................... ........... .................
.............................
....................... ........................ 3 .......................
...................................................................................................................................................................................................................................................................................................................................................
6 . ... 6

Figure 11.6: The graph in Exercise 9

10. The statement is true. Suppose that G is not outerplanar. Then G contains a
subdivision of K4 or a subdivision of K2,3 . If G contains a subdivision of K2,3 ,
then G  K2 contains the structure shown in Figure 11.7(a), which contains a
subdivision of K3,3 shown in Figure 11.7(b). If G contains a subdivision of K4 ,
the situation is similar. Thus, G  K2 is nonplanar and so cr(G  K2 ) 6= 0.
132 CHAPTER 11. NONPLANAR GRAPHS
..............................
............... .........
......... ....... ...................................
.............
....... ...... ........
.........
. . . ...... ...... .......
.......
......
...
.... . .....
..... .. . . ...... .....
.
.....
. ..... .
. ..... .....
.....
...
. ...
... .
....
. ....
.... ... .... ...
...
.. .
t
. ... .. ...
... ........... .........
. ....... .....
. .... ... .... . . ......
..........
...
.... .
.... . .
. . . ...
... ....
. ..
. . . . ... ...... ... ...... ... . ...
.. ... ..... .... .... ...... .... . ... .... .
... ...... .. .. .... .. .. ... .... .... .. ...
. .. .
.. ... ....
.
... .. .. ........ . ... ... .. ........ ... . . .
. ... .. .
..
.
........ ...
... .. .. ........ ....
.. ... .. .. ... .. .. ... .... ... .. .... ...
..... ...... .. ... ... ... .. ...... ... ..
r
...
r
. .
r t ..........
. ... ... .. .... .... .. ..... ....... ...
......... . . .. . . .. . .... .....
............. ............. .
. .
...
...
......................... .........
.
. .........
.
. .
........... .
.... ...
. .... ... .
.... ... ...............
.
..
......... . ...... ........ ............ .. ..
.....
.... .. ..... ....... .... ................. ..
.........
.... ... ...... ......
.....
... ............
.
... ... ....... .. ......... ... .
. ... . ..
.. ..
....... .. .... ..
. .......... ...... . .. .. ....... .
..... .. ... ........................................... ..... . .. ..... . .. . .
. . ..
.... ......................................... .... ....
.. .
.... .. .. .... ... ...
.. .... ... .... ...
..... .. ... ..... .. ..... ..... ....... ...... ....
..... ... ......... ...... ... ........
.............. . .
..................................................................................................................
. ..... . ....
..... .... t ...... ...
.... ...
.......

(a) (b)

Figure 11.7: Graphs in Exercise 10

Section 11.2. The Genus of a Graph

11. Since K3,3 ⊆ K4,4 , it follows that γ(K4,4 ) ≥ 1. The toroidal embedding of
K4,4 in Figure 11.8 shows that γ(K4,4 ) = 1. Since K4,4 has order n = 8 and
size m = 16 and n − m + r = 0, it follows that r = 8.
t...... t
.......
..... ........
.......................... t
.....
.
..... ..... ..... .....
..... ..... .....
......
... ...
.....
.....
.....
R 7......... . ... .....
.....
.....
R 8............ ..........
... ..... ... ..
.....
..... .......... ..... .........
.. .
........
R ....... R
R 5 ............................ 1 ...........................
.....
5
.. ..... ..
..... ..... ..... .....
..... ..... ..... .....
..... ..... ..... .....
...... ..... ......... .....
t
.
t
. .....
t .. ..... .....
..... ........ ....
.....
..... R 2 .... ....
... .......
.
..
.
.....
R 3 ............
...
.....
..... . ...... .. ....
..... ....... .....
.. .... ...
..... .. ..... .
..... ..... ..... .....
..... ........ ..... ........
.......... ............
.. ... ........
R ..
..
6......... ......... ... .
..
R 4 ........... ............. 6
...
. . ... R
..... ..... .. .....
..... ..... .... .....
..... ..... ..... .....
.......
R .. ..
..... .......
R .....
t .
.....
. 7 ..
t
..... ..... ..
.. 8
........................ t
.....

Figure 11.8: Embedding of K4,4 on the torus

12. (a) Proof. Let G be drawn on the sphere with cr(G) crossings. For each
crossing of two edges e and f , place a handle over e and draw f on the
handle. Then we have an embedding of G on the surface of genus cr(G).
Therefore, γ(G) ≤ cr(G).
(b) Proof. For k = 1, let G = K5 (or G = K3,3 ). Then γ(G) = cr(G) = 1.
For k = 2, let G = K3,4 . Then γ(K3,4 ) = 1 and cr(K3,4 ) = 2 by
Theorem 11.4. For k ≥ 3, let G = C3  Ck . By Theorem 11.6, cr(G) = k.
Thus the graph C3  Ck is not planar. Since C3  Ck can be embedded
on the torus (see Figure 11.9), γ(C3  Ck ) = 1.
13. Proof. Since the boundary of every region contains at least k edges and
every edge is on the boundary of at most two regions, kr ≤ 2m. By Corol-
lary 11.12,
2m
2 − 2γ(G) = n − m + r ≤ n − m + .
k
 133
...........
......
... .
...... .....
...... u1 s sv1 s w1 ... ...

u2 s sv2 s w2
s s s
p p p
p p p
p p p
s s s
s s s
uk vk wk
...........
......

Figure 11.9: A toroidal embedding of C3  Ck

The result follows by solving this inequality for γ(G).

14. Proof. For each component Gi of G, let Bi1 , Bi2 , · · · , Bi,ji be the blocks of
Gi . By Theorem 11.16,
ji
X
γ(Gi ) = γ(Bip ).
p=1

Then G has blocks {Bi1 , Bi2 , · · · , Bi,ji : 1 ≤ i ≤ k}. By Theorem 11.16,

X ji
k X k
X
γ(G) = γ(Bip ) = γ(Gi ),
i=1 p=1 i=1

giving the desired result.

m n
15. By Corollary 11.15, γ(Ks,t ) ≥ 4 − 2 + 1. The result then follows by letting
n = s + t and m = st.
16. (a) Let G = K3,3 ∨ K n . Since the order of G is n + 6 and the size of G is
6n + 9, it follows by Theorem 11.13 that
6n + 9 n + 6 n−1
γ(G) ≥ − +1= .
6 2 2

(b) Since K3,3 ∨ K n (n ≥ 1) contains K3,3 as a subgraph, each graph K3,3 ∨

K n is nonplanar. For n = 3, K3,3 ∨ K 3 = K3,3,3 . Since K3,3,3 can be
embedded on the torus as shown in Figure 11.10, γ(K3,3,3 ) = 1. Because
K3,3 ∨ K n ⊆ K3,3,3 for n = 1, 2, 3, it follows that γ(K3,3 ∨ K n ) = 1 for
n = 1, 2, 3.
134 CHAPTER 11. NONPLANAR GRAPHS

.........................................................................................................................................................................................................................................................................................................................
... ... ... ..... ...
... ... ... .....
..... ...
.... ...
...
...
... .....
..... ...
... ... ... ..... ...
... ...
...
...
...
.....
..... ....
.
... ... ... .
..... ...
.................... ... .... ..
.....
... ............ .... . ...
............ ...
.. ............ ... ... ...
... ............
v
y
...
v
...
u
...... ...
.................................................................................................................................................................................................................................................................................................................................
... ..... ..... ......... ...
..... ..... .........
... ..... ...... .......... ...
.. ..... ..... ......... ...
..... .........
.... ..... .
........ ......... ...
...
... ..... ..... .........
..... ..... ......... .
... ..... ..... ......... ....
.............. ..... ..... ......
.............. ..... ..... ...
... .............. ..... ..... ...
... ..............
..............
.....
.....
.....
..... ...
..
v u v
y
... .. ... .
..........................................................................................................................................................................................................................................................................................................................................
.. ..... ..... ............. ...
..... ..... ...
... ..... ...... ............. ...
... ..... .....
.....
.........
......... ...
... ..... ..... .........
..... ...
... ..... .....
.....
.........
......... ..
.... ..... ..... ......... ....
.................. ..... . .
.......... ..... ...... ........
... .......... .
....... .. ..... ...
... .......... ..... ..... ...
.. .......... ..... ..... ...
.......... ..... .....
....
v
y v
...
u .......... ..... ..... .
................................................................................................................................................................................................................................................................................................................................
... ...
...
... ...
...
...
... ... .
......
... ...
...
. .. ......
... ... ... . .....
...... .... ... .... . ....
..... . . . ...
... ... ...
.... ... ... ... ...
... ... ... ... ...
... ... ... ... ...
... ... ... ...
... ... ... .... ...
... ... ... ... .
........................................................................................................................................................................................................................................................................................................................
.......

Figure 11.10: Embedding K3,3,3 on the torus in Exercise 16(b)

17. Since the order of the bipartite graph G = C4  C6  K2 is 48 and its size is
120, it follows by Corollary 11.15 that
120 48
γ(G) ≥ − + 1 = 7.
4 2
To show that γ(G) = 7, it suffices to produce an embedding of G on a surface
of genus 7. First, note that C4  C6 can be embedded on the torus as shown
in Figure 11.11.
... ... ... ... .................
... ... ... ... .
....... ... ... ... ... ......
........ ... ... ... ... ... ...
s ...
... ... ... ....
...................... ...
..
s s ...
... . .. ....
... ............... ...
s ..

................... . ... ... ................ ...

.......... ....... ... . . ... ........ ..... ...
s ... .... ...
... ...
...
s s ...... ... ...
...
...
s ...
...
...
... ... ... ...
... ... ...
s
...
...
... .. ..... ....
................ ..... ...
..
s s ...
... ... ... ....
... ...... ...... ...
s ..

... ..... ....... ... ... ... ...... ...... ... . . .

... ...... ...... ... . . ... .... ..... ... ... . .
s ... .... ...
...
...
...
...
s s ... ... ...... ...
...
...
s ...
...
... ... ... ...
... ... ...
s
...
s ...
... ... .. ....
........ ........ ... ...
..
s ...
... . . .... s
... ...... ....... ...
..

.......... ....... ... .

... ....... ....... ... .
. .
... ...... .......... .
.... ...... ........
. ..
s ... .. ..... ...
...
...
...
...
s s ... s
... .... ....... ...
...
...
...
... ... ... ...
... ... ... ...
... ... ... ... .
.....
...........

Figure 11.11: Embedding C4  C6 on the torus in Exercise 17

Observe that there are six regions with pairwise disjoint boundaries as
shown in Figure 11.11. We then consider an embedding of C4  C6 on another
copy of a torus that is a “mirror image” of the given embedding. Applying
the method in the proof of Theorem 11.20, we insert a handle between each
 135

pair of regions (one in the first torus and the other in the second torus) and
create a surface of genus 1 + 1 + (6 − 1) = 7. Then we embed (C4  C6 )  K2
in the same fashion described in the proof for γ(Qn ).
18. Since γ(K5 ) = 1, the result holds for γ = 1. For γ ≥ 2, let G be the connected
graph consisting of γ blocks, each of which is a copy of K5 . The result then
follows by Theorem 11.16.
19. Proof. For n ≥ 6k + 6, let G be a maximal planar graph of order n. Then
m = 3n − 6. The order of G  K2 is 2n and the size of G  K2 is

2m + n = 7n − 12.

By Theorem 11.13,
7n − 12 2n n−6
γ(G  K2 ) ≥ − +1= ≥ k,
6 2 6
giving the desired result.
20. See Figure 11.12.
..
.....
.. ..
a ............. .........
b
........ .........
...... .....
...... .....
...... .
........
...... .. ......
..... ......
...........
..... ...........
......... ......
... ......
..... .............
............... ... ...
... ...
... ..
......
.
b a

Figure 11.12: Embedding K5 on the Möbius strip in Exercise 20

21. Let there be given a toroidal embedding of K7 . Insert holes in the regions R1
and R2 and place a handle H between these two holes. Place a new vertex v
on H and draw edges from v to the four vertices on the boundaries of R1 and
R2 . This gives an embedding of G on S2 . Thus, γ(G) ≤ 2. Since G has order
n = 8 and size 21 + 4 = 25, it follows by Theorem 11.13 that
m n 25 8 7
γ(G) ≥ − +1= − +1=
6 2 6 2 6
and so γ(G) ≥ 2. Thus, γ(G) = 2.
22. Proof. Since G can be embedded on S3 , it follows that γ(G) ≤ 3. Because
H is a 6-regular graph of order 12, the size of H is 36. The graph G = H  K2
has order 24 and size 2 · 36 + 12 = 84. Therefore, by Theorem 11.13,

γ(G) ≥ m/6 − n/2 + 1 = 84/6 − 24/2 + 1 = 3.

Hence, γ(G) = 3.
136 CHAPTER 11. NONPLANAR GRAPHS
..............................................
.............
................................................................. .................. ...........
.......... ...........
.......... ........ ......... .........
........
........ ........
. .. . w ....... .... .......
r t
.
. ..
t
...... ....... .......
......... ............ .
............ ........... ...........
. .. ...
.
... ... . ...
.............
...... ..
..............................
....... . ........ .................... ......... ....... ..... ......... ....
.....
..... .
.....
.
..
.
...... .
........ .....
. ........
. ...
....
.
.
...... .
.......
.... ..... .. ..
..... ..... .
...
.. ... ..... ........
..... .
. .. . ...... .. ..... ...............
u
................ ................
... .........
.............
..................
.. ..
........... v
..
..
r..
.
.
.
...
..... .................
..... ......... ..... ......... .........
... . .. .....
........ .....
.. ..
...... .....
..... .........
. ..... .. ...
..... ..... ..... .
t r
........ ..
.
....... ..... ..... ...... .........
.....
.. ....... ...
.............
............. ... ............. . .
........ ..... .....
.......
........ ... ... ........ ......
.........
........... ....... ......
.
................. ........
......................................................
(b)
(a) ...........
................
..................................................
...........
......... .........
....... ........
.
.............
.......
.
..........
w ..
. .. .
.......
......
........
...... ... .
.. ......... .... ... ....
..... .
..... ... .
..... ... . ..... .......
.....
..... ...... ..
.....
........... ..
.....
. ........
.
..... . .
... ............. . ..... . ...
..
. .. .. ....
............. ........... ............. ..... ...
.. .. ...... ...... .................. ...............
u ............ ..............
..... ........ .........................
........ ..............
.
....
v
.....
.
......
. ..... ... ..... .....
.....
...
. ..... ....
. .
...
. .. ......
.. . . ....... .....
........
............... ..........
.
....... ..........
...... .... ...
.... .........
........
......... ..
. ...........
..........
............. .
.............
.....................................................................

(c)

Figure 11.13: Graphs in Exercise 23

23. Yes. Consider the graph G in Figure 11.13(a).

The graph G in Figure 11.13(a) is nonplanar since it contains a subdivision
of K3,3 shown in Figure 11.13(b). By inserting a handle H between the holes
shown in the drawing of G in Figure 11.13(c), the edge uw can be drawn on
H and so γ(G) = 1. The edge e = uv can be added to G and placed on H. So
γ(G + e) = 1. Since cr(G) = 1 and G + e contains W4  K1 (see Exercise 7),
it follows that cr(G + e) ≥ 2.
24. Proof. The size of H is m = rn/2. Since r ≡ 5 (mod 6), it follows that r =
5 + 6q for some nonnegative integer q. Let the order and size of G = H  K2
be denoted by n0 and m0 , respectively. Thus, n0 = 2n and m0 = 2m + n. By
Theorem 11.13,

m0 n0 2m + n 2n 2m − 5n
γ(G) ≥ − +1= − +1= + 1.
6 2 6 2 6
rn (5+6q)n
Since m = 2 = 2 , it follows that

(5 + 6q)n − 5n (r − 5)n
γ(G) ≥ + 1 = qn + 1 = + 1.
6 6
(r−5)n
Since G can be embedded on Sk where k = 6 + 1, we have γ(G) = k.
 137

Section 11.3. The Graph Minor Theorem

25. (a) Let F be the set of all forests. Let F ∈ F. Since the deletion of a vertex,
the deletion of an edge of F or the contraction of an edge of F results in
a forest, F is a minor-closed family of graphs.
(b) There is only one forbidden minor of F, namely K3 . That is, a graph F
belongs to F if and only if K3 is not a minor of F .
26. No. The contraction of an edge in the bipartite graph C4 produces C3 , which
is not bipartite.
27. Since the deletion of a vertex, the deletion of an edge or the contraction of
an edge of an outerplanar graph is outerplanar, the set O is a minor-closed
family of graphs. The set of the forbidden minors of O is {K4 , K2,3 } by
Theorem 10.24.
28. By Theorem 11.22, there exist positive integers i1 and j1 with i1 6= j1 such
that one of Gi1 and Gj1 is a minor of the other, say j1 > i1 . Now consider the
infinite subset {Gj1 +1 , Gj1 +2 , . . .}. By Theorem 11.22, there exist integers
i2 and j2 and i2 6= j2 such that one of Gi2 and Gj2 is a minor of the other.
Continuing in this manner gives the desired result.
138 CHAPTER 11. NONPLANAR GRAPHS
Chapter 12

Matchings, Independence
and Domination

Section 12.1. Matchings

1. Proof. Assume, to the contrary, that there exists a tree T with two distinct
perfect matchings M1 and M2 . Let H be a spanning subgraph of T with
E(H) = (M1 − M2 ) ∪ (M2 − M1 ). By Theorem 12.1, each component of H is
(i) an isolated vertex, (ii) an even cycle whose edges are alternately in M1 and
in M2 or (iii) a nontrivial path whose edges are alternately in M1 and in M2
such that each end-vertex of the path is either an M1 -unmatched vertex or an
M2 -unmatched vertex but not both. Since T has no cycles, (ii) is impossible.
Since every vertex of T is incident with both an edge in M1 and an edge in
M2 , (i) is impossible. Since no vertex is either an M1 -unmatched vertex or
an M2 -unmatched vertex, (iii) is impossible.

2. (a) In order for a graph G of order n = 2k to have a maximum matching of

k edges and maximum size m, the graph G = Kn and so m = n2 .

(b) For n = 2k + 2, the graph Kn−1 + K1 has order n, size n−1

2 and a
maximum matching of size k. If G is a graph whose size is at least n−1


2 +
1, then G has a Hamiltonian path, implying that G has a maximum
matching of size k + 1. Thus, m = k−1


2 .

3. Proof. Assume that Hall’s condition is satisfied. Let H be the graph ob-
tained from G by adding two new vertices u and w where u is joined to every
vertex of U and w is joined to every vertex of W . We show that U can be
matched to a subset of W . We claim that the minimum number of vertices
in a u − w separating set in H is |U |. Assume, to the contrary, that there is
a u − w separating set S in H with |S| < |U |. Let U1 = S ∩ U , U2 = U − S,
W1 = S ∩ W and W2 = W − S (see Figure 12.1). Since |S| < |U |, it follows
that U2 6= ∅. No vertex in U2 is adjacent to any vertex in W2 , for otherwise,

139
140 CHAPTER 12. MATCHINGS, INDEPENDENCE AND DOMINATION

u and w are connected in H − S, which is impossible. Thus, NG (U2 ) ⊆ W1

and so
|NG (U2 )| ≤ |W1 |. (12.1)

Since |S| < |U |, it follows that

|S| = |U1 | + |W1 | < |U | = |U1 | + |U2 |

and so |W1 | < |U2 |. By (12.1),

|NG (U2 )| ⊆ |W1 | < |U2 |,

which contradicts the assumption that G satisfies Hall’s condition. Therefore,

the minimum number of vertices in a u − w separating set in H is |U |, as
claimed. By Menger’s theorem (Theorem 4.10), the maximum number of
internally disjoint u − w paths in G is |U |, which implies that U can be
matched to a subset of W .

U W
................... ...........................
....... .....
..... .... ..... .....
.... ... ..... ...
... ... ... ...
.... ... ...
. ...
... ... ..
. ...
... ... ... ...
... ...
... ....
...
....... ...... U 1
...
...
... ..... ..
..... ...
... W 1
.
.........
.....
... ...
. ..... ..... ..........
...
.... ...
... ..... ..... ... ... .......
.... ... ..... ..
. ... ... .......
... . ..... ..
.
..... .. ... ..... ........ .. ... .....
............. .... ... ........ ..... ... ...........
.. ..
H: u ..........
......
...
... ...
..
.
..
......
..... ........
...
... ....
.
.....
........... v
...... .. ..... . .........
...... .. .... ..... ..... ...
......... .... ..... ... .
.. .......
...... .. .... ..... .
.......... ... ..... ..
... .....
......
U ... ..... ...
........
...
...
...
2
..
...
...
...
...
W 2 ..
.
. .
...
... ... ... ...
... ... ...
... ...
.. ..
...
..... .....
. ...
..... .
....
......
..........................
... ..... ....
........ ............
........

Figure 12.1: The sets U1 , U2 , W1 , W2 in H in the proof of Exercise 3

4. Proof. Let S be a set of k vertices of U where 1 ≤ k ≤ r, say

S = {ui1 , ui2 , . . . , uik },

where i1 < i2 < · · · < ik . Therefore, ik ≥ k. By hypothesis, uik is adjacent

to all vertices wj such that ik + j ≥ r + 1. Therefore, uik is adjacent to all
vertices wj such that j ∈ T = {r + 1 − ik , r + 2 − ik , . . . , r}. Now, |T | =
r − (r + 1 − ik ) + 1 = ik ≥ k. Thus, |N (uik )| ≥ k and so |N (S)| ≥ k. Hence,
G satisfies Hall’s condition. By Hall’s theorem (Theorem 12.3), there is a
matching from U to a subset of W . Since |U | = |W |, it follows that G has a
perfect matching.
[Note: For 1 ≤ i ≤ r, ui and wr+1−i are adjacent. Therefore, M = {ui wr+1−i :
1 ≤ i ≤ r} is a perfect matching.]
 141

5. (a) Proof. Assume, to the contrary, that there is a bipartite graph G of size
m with partite sets U and W with |U | = |W | = k such that m ≥ k 2 −k+1
but G has no perfect matching. By Hall’s theorem (Theorem 12.3),
there is a nonempty subset S of U such that |N (S)| = j < ` = |S|.
This implies that m ≤ k 2 − `(k − j). Since m ≥ k 2 − k + 1, it follows
that −k + 1 ≤ −`(k − j). Since j ≤ ` − 1 and k ≥ `, it follows that
`(` − 1) ≥ `j ≥ `k − k + 1 = k(` − 1) + 1 ≥ `(` − 1) + 1, which is a
contradiction.
(b) Let G = Kk,k−1 + K1 .
6. Claim: The subdivision graph S(G) of a connected graph G has a perfect
matching if and only if G is unicyclic.
Proof. Suppose that G has order n and size m. First, assume that S(G)
has a perfect matching. Since S(G) is a bipartite graph with partite sets
V (S(G)) − V (G) and V (G) and S(G) has a perfect matching, it follows that
|V (S(G)) − V (G)| = |V (G)| and so m = n. Hence, G is unicyclic.
For the converse, assume that G is a unicyclic graph. Then G contains a
single cycle C. In S(G), the cycle S(C) contains a perfect matching. There-
fore, if G is a cycle, then S(G) has a perfect matching. We may therefore
assume that G is not a cycle. So, if we remove the edges of S(C) from S(G),
then we have one or more nontrivial trees S(T ). Let v be the vertex of S(T )
that belongs to S(C). For each vertex u in S(T ) for which dS(T ) (v, u) is even,
let eu be the edge on the v − u path incident with u. The set of all such edges
eu in all such trees S(T ) together with a perfect matching in S(C) produce a
perfect matching in S(G).
7. Proof. Observe first that α0 (G) ≤ |U |. We consider two cases, depending
on whether def(U ) ≤ 0 or def(U ) > 0.

Case 1. def(U ) = max{|S| − |N (S)|} ≤ 0, where the maximum is taken over

all nonempty subsets S of U . Then Hall’s condition is satisfied. Hence, U can
be matched to a subset of W and so α0 (G) = |U |.

Case 2. def(U ) = max{|S| − |N (S)|} > 0, where the maximum is taken over
all nonempty subsets S of U . Let S be a subset of U such that def(U ) =
|S| − |N (S)|. We show that N (S) can be matched to a subset of S and U − S
can be matched to a subset of W − N (S). Remove all the edges joining a
vertex in N (S) and a vertex in S. Assume, to the contrary, that there is
T ⊆ N (S) such that |N (T )| < |T |. Thus,

|S − N (T )| = |S| − |N (T )| > |S| − |T |.

Then S 0 = S − N (T ) is a subset of U with N (S 0 ) = N (S) − T . Hence,

|S 0 | − |N (S 0 )| = |S − N (T )| − |N (S) − T |
> (|S| − |T |) − (|N (S)| − |T |) = def(U ),
142 CHAPTER 12. MATCHINGS, INDEPENDENCE AND DOMINATION

which is a contradiction. Also, assume, to the contrary, that there is T ⊆ S

such that |N (T )| < |T |. Then S 0 = S ∪ T is a subset of U with N (S 0 ) =
N (S) ∪ N (T ). Thus,

|S 0 | − |N (S 0 )| = (|S| + |T |) − (|N (S)| + |N (T )|)

> (|S| + |T |) − (|N (S)| + |T |) = def(U ),

which is a contradiction. Hence, N (S) and S satisfy Hall’s condition and so

α0 (G) = |N (S)| + |S| = |N (S)| + (|U | − |S|)

= |U | − def(U ),

as desired.

8. (a) The statement is false. If T is a tree of order n, then S(T ) has order
2n − 1.
(b) Let T1 = P2 , T2 = P3 and T3 = K1,3 .

9. M (C3 ) = C3 , M (2P3 ) = C4 , M (C5 ) = C5 , M (K2,3 ) = C6 .

10. (a) Let G be an r-regular maximal planar graph of order n, where 4 < n <
12. Then the size m of G satisfies m = 3n − 6 and m = rn 2 . Thus,
rn
2 = 3n − 6 and so rn = 6n − 12. Hence, (6 − r)n = 12. Since n | 12
and 4 < n < 12, it follows that n = 6 and so r = 4. There is only one
4-regular graph G of order 6, namely K2,2,2 , which is maximal planar.
(b) Since G = 3K2 , it follows that E(G) is a perfect matching M . Then
G + M = K6 and γ(K6 ) = 1.
(c) Proof. Let G be a maximal planar graph of order n and size m = 3n−6.
Let M be a perfect matching in G. Thus, n is even and |M | = n/2. The
size of G + M is m0 = m + n2 = 7n
2 − 6. Therefore,

7n
m0 n −6 n
γ(G + M ) ≥ − +1= 2 − +1
6 2 6 2
7n n n
= −1− +1= ,
12 2 12
as desired.

11. (a) The statement is true. Proof. Let M be a matching of maximum size
that contains both e and f . Then M is a maximal matching of G.
(b) The statement is false. Let G be the graph obtained from the path
(u, v, w, x, y) by adding a new vertex z and the edge wz. Let e = vw and
f = wx. Then there is no maximum matching of G containing either e
or f .

12. (a) The complete graphs Kn of even order n ≥ 2.

 143

(b) Proof. Let M = {f1 , f2 , . . . , fj } be a matching in G such that |M | <

1 0 0
2 α (G) and let M = {e1 , e2 , . . . , ek } be a maximum matching in G,
where say fi = xi yi for 1 ≤ i ≤ j and ei = ui vi for 1 ≤ i ≤ k. Let
S = {xi , yi : 1 ≤ i ≤ j} and S 0 = {ui , vi : 1 ≤ i ≤ k}. Thus, 2k = |S 0 |,
2j = |S| and k > 2j. For some integer ` where 1 ≤ ` ≤ k, there exists a
pair {u` , v` } of vertices of S 0 such that u` , v` ∈
/ S, for otherwise, k ≤ 2j.
Therefore, M ∪ {u` v` } is a matching in G.

Section 12.2. 1-Factors

13. (a) The statement is false. Replace the vertex v by a triangle in Figure 12.6
in Chapter 12.
(b) The converse of Errera’s theorem is false.
14. Proof. Let G be a cubic graph with at most two bridges. If G contains no
bridges, then by Petersen’s theorem (Theorem 12.8), G has a 1-factor. Hence,
we may assume that G has exactly one bridge or exactly two bridges. Let S
be a subset of V (G) of cardinality k ≥ 1. We show that the number ko (G − S)
of odd components of G − S is at most |S|. Since this is true if G − S has
no odd components, we may assume that G − S has ` ≥ 1 odd components
G1 , G2 , . . . , G` . Let Xi (1 ≤ i ≤ `) denote the set of edges joining the vertices
of S and the vertices of Gi . For each i with 1 ≤ i ≤ `, observe that
X X
degG v = degGi v + |Xi |.
v∈V (Gi ) v∈V (Gi )
P P
Since v∈V (Gi ) degG v = 3|V (Gi )| is odd and v∈V (Gi ) degGi v = 2|E(Gi )|
is even, it follows that |Xi | is odd for 1 ≤ i ≤ `. We now consider two cases.

Case 1. G has exactly one bridge. We may assume that |X1 | ≥ 1 and |Xi | ≥ 3
for 2 ≤ i ≤ `. Since |S| = k and every vertex of S has degree 3 in G, at most
3k edges join the vertices of S and the vertices of G − S. Therefore,
`
X
1 + 3(` − 1) ≤ |Xi | ≤ 3k.
i=1

Then ` = ko (G − S) ≤ k + 2/3. Since ko (G − S) is an integer, ko (G − S) ≤

k = |S|.

Case 2. G has two bridges. We may assume that |Xi | ≥ 1 for i = 1, 2 and
|Xi | ≥ 3 for 3 ≤ i ≤ `. Then
`
X
2 + 3(` − 2) ≤ |Xi | ≤ 3k.
i=1

Thus, ko (G − S) ≤ |S| + 4/3 and so ko (G − S) ≤ |S| + 1. Since G has even

order, ko (G − S) and |S| are of the same parity and so ko (G − S) 6= |S| + 1.
Therefore, ko (G − S) ≤ |S|.
144 CHAPTER 12. MATCHINGS, INDEPENDENCE AND DOMINATION

By Tutte’s theorem (Theorem 12.7), G has a 1-factor.

15. (a) Proof. Let S = {u, v}. Then ko (G − S) = 4 > |S| = 2. By Tutte’s
theorem (Theorem 12.7), G has no 1-factor.
(b) Proof. Let G be a bridgeless graph having exactly one vertex of degree
7 and all others of degree 3. If G is disconnected, then it follows by
Petersen’s theorem (Theorem 12.8) that every cubic component of G
contains a 1-factor. Hence, it remains to show that the one component
of G having the vertex of degree 7 also has a 1-factor. Thus, we may
assume that G is connected.
Assume, to the contrary, that G does not have a 1-factor. By Tutte’s
theorem (Theorem 12.7), there exists a set S ⊆ V (G) such that ko (G −
S) > |S|. Let G1 , G2 , · · · , Gk denote the odd components of G − S. Let
k = ko (G − S) and j = |S|. Thus, k > j. Since every vertex of G
has odd degree, G has even order and so ko (G − S) and |S| are of the
same parity. Hence, k ≥ j + 2. Since no component Gi of G − S is a
component of G, at least one edge must join Gi and S. However, since
G is bridgeless, there cannot be exactly one edge between any Gi and
S. There cannot be exactly two edges between any Gi and S either,
for otherwise Gi has an odd number of odd vertices. Therefore, there
must be at least three edges between each Gi and S. Hence, there are
at least 3k edges joining the vertices of V (G) − S and the vertices of S.
The set S, however, contains at most one vertex of degree 7; so at most
3(j − 1) + 7 = 3j + 4 edges can exist between V (G) − S and S. Therefore,
3k ≤ 3j + 4 ≤ 3(k − 2) + 4 = 3k − 2, which is impossible.
(c) Proof. The graph G in (a) is such an example.

16. (a) Proof. Assume, to the contrary, that there is a connected cubic graph
G having a 1-factor F and a bridge e such that F does not contain
e. Then F is a 1-factor of G − e. Suppose that G1 and G2 are the
components of G − e and the order of Gi is ni for i = 1, 2. Since Gi has
ni − 1 odd vertices for i = 1, 2, it follows that n1 − 1 and n2 − 1 are both
even. Thus, n1 and n2 are odd. This implies that G1 and G2 have no
1-factor, which is impossible.
(b) It is possible that a 1-factor contains neither e nor f . It is possible that
a 1-factor contains both e and f . See Figure 12.2.
........ .......
............................... ..................
....... ... ......... e ......... ..... .............
............. ..... .............. . . .. .
................ ... ........
...... ... ......... ............ ... ....... ..
................. .... ...
...
..... ...
... ..
...
... .....
. ... ..... ..
... ... ... ...
.... .....
G: ......
... .....
...
.. ...
... ..... ....
............... ... ... ... ..
.......
....... ......
. .....
.... .
.
................ ... ......
....... .. ........ ..... ........ .... ...........
....... ... ........ .
.. .........
....... ....... ................ f ........
........ .. ..
..... ...
... .............................
.......

Figure 12.2: The graph in Exercise 16(b)

 145

It is impossible that a 1-factor contains exactly one of e and f . Assume,

to the contrary, that G contains a 1-factor F such that F contains e but
not f . Then F is also a 1-factor of G − f . Suppose that G1 and G2 are
the components of G − e − f , where Gi has order ni for i = 1, 2. By
the same argument used in (a), n1 and n2 are even. Let e = uv where
u ∈ V (G1 ) and v ∈ V (G2 ). This implies that each of G1 − u and G2 − v
has a 1-factor, which is impossible since each has odd order.
17. (a) Proof. Assume that G has order n and size m. Let V1 = {v1 , v2 , · · · , vk },
V2 = {vk+1 , vk+2 , · · · , vn } and E 0 = [V1 , V2 ]. Let H = G[V2 ]. Then
n
X k
X n
X
2m = degG vi = degG vi + degG vi
i=1 i=1 i=k+1
k
X n
X
= degG vi + |E 0 | + degH vi .
i=1 i=k+1
Pn
Since 2m and i=k+1 degH vi are even, it follows that

k
X
degG vi + |E 0 |
i=1

Pk
is even. Since degG vi is odd for all i (1 ≤ i ≤ k), i=1 degG vi is even
if and only if |V1 | is even. Thus, |V1 | + |E 0 | is even, giving the desired
result.
(b) Proof. For k = 1, this is Petersen’s theorem (Theorem 12.8). So we
may assume that k ≥ 2. Assume that there is a proper subset S of V (G)
so that the number k of odd components of G − S exceeds j = |S|. Let
G1 , G2 , · · · , Gk be the odd components of G − S. Since G is 2k-edge-
connected, there are at least 2k edges between Gi and S for each i with
1 ≤ i ≤ k. Let Ei be the set of edges between S and the component Gi
for 1 ≤ i ≤ k. Thus, |Ei | ≥ 2k for 1 ≤ i ≤ k. However, there cannot
be a set Ei of exactly 2k edges between S and Gi (1 ≤ i ≤ k) since |Ei |
and |V (Gi )| are of the same parity by (a) and |V (Gi )| is odd. Thus,
|Ei | ≥ 2k + 1 for 1 ≤ i ≤ k. Therefore, there are at least k(2k + 1) edges
between S and G − S. On the other hand, there are at most j(2k + 1)
edges between S and G − S. Hence, k(2k + 1) ≤ j(2k + 1), implying that
k ≤ j, which is a contradiction.
18. Proof. Assume, to the contrary, that V (G) contains a proper subset S such
that the number ko (G − S) = k of odd components of G − S exceeds |S| = j.
Let G1 , G2 , · · · , Gk be the odd components of G − S. Since n is even, k and
j are of the same parity. So, k ≥ j + 2.
For each integer i with 1 ≤ i ≤ k, let Ei be the set of edges in G joining
Gi and S. Then there are at most r − 1 sets Ei with |Ei | = r − 2, say
146 CHAPTER 12. MATCHINGS, INDEPENDENCE AND DOMINATION

E1 , E2 , · · · , Er−1 . Thus, |Ei | ≥ r − 2 for 1 ≤ i ≤ r − 1. Next, we show that

|Ei | and r are of the same parity. Let H be the subgraph induced by V (Gi ),
where r ≤ i ≤ k. Then
X X X
2m = degG vi = degG v + degG v
v∈V (G) v∈V (G)−V (Gi ) v∈V (Gi )
X X
= degG vi + |Ei | + degH vi
v∈V (G)−V (Gi ) v∈V (Gi )
X
= r(n − |V (Gi )|) + |Ei | + degH vi .
v∈V (Gi )
P
Now, v∈V (Gi ) degH vi is even as it is the size of H and so r(n−|V (Gi )|)+|Ei |
is even. We consider two cases.

Case 1. r is odd. Since r and n − |V (Gi )| are both odd, it follows that |Ei | is
odd.

Case 2. r is even. In this case, r(n − |V (Gi )|) + |Ei | is even and r is even,
implying that |Ei | is even.
Since |Ei | ≥ r − 1 for each i with r ≤ i ≤ k and |Ei | and r are of the same
parity, it follows that |Ei | ≥ r for all r ≤ i ≤ k. On the other hand, there are
at most jr edges joining G − S and S. Therefore, (r − 1)(r − 2) + (k − r + 1)r ≤
jr ≤ (k − 2)r, which implies that 2 ≤ 0, a contradiction.
19. (a) 3. (b) Yes.
20. Proof. Suppose first that G is factor-critical. Then n is odd. Let S be a
nonempty proper subset of V (G) and let v ∈ S. Hence, G − v contains a
1-factor. Let S 0 = S − {v}. Then S 0 is a proper subset of V (G − v). By
Tutte’s theorem (Theorem 12.7), ko (G0 − S 0 ) ≤ |S 0 | and so

ko (G − S) = ko (G − v − S 0 ) ≤ |S 0 | < |S|.

For the converse, let G be a graph of odd order n such that ko (G − S) ≤ |S|
for every nonempty proper subset S of V (G). We show that G is factor-
critical. Let v ∈ V (G) and let S 0 be a proper subset of V (G − v). Now, let
S = S 0 ∪ {v}. Thus, S is a nonempty proper subset of V (G). By assumption,
ko (G − S) ≤ |S|. Since n is odd, ko (G − S) 6= |S| and so ko (G − S) < |S|.
Thus,
ko (G − v − S 0 ) = ko (G − S) < |S| = |S 0 | + 1
and so ko (G − v − S 0 ) ≤ |S 0 |. By Tutte’s theorem, G − v has a 1-factor and
so G is factor-critical.
 147

Section 12.3. Independence and Covers

21. Proof. Let G be a bipartite graph and let H be a subgraph of G. Then
H is bipartite. Let U and W be the partite sets of H with |U | ≤ |W |. Then
α(H) = |W | ≥ |V (H)|/2. For the converse, let G be a graph that is not
bipartite. Let F be a subgraph of G consisting of an odd cycle C of G.
Suppose that the order of C is 2k + 1. Then α(F ) = k < |V (F )|/2.
22. Proof. It suffices to verify these bounds for connected graphs. For a graph
G of order n, the upper bound α0 (G) ≤ bn/2c is immediate. Since α0 (Kn ) =
bn/2c, for example, this bound is sharp.
Next, we show by induction on the order n ≥ 2 of a tree that
n
α0 (T ) ≥ .
1 + ∆(T )
Equality holds for n = 2 and n = 3. Assume, for every tree T 0 of order n0
where 2 ≤ n0 < n with n ≥ 4, that
n0
α0 (T 0 ) ≥ .
1 + ∆(T 0 )
Let T be a tree of order n. If T = K1,n−1 , then α0 (T ) = 1+∆(Tn
) = 1, which
verifies that sharpness of the lower bound. Hence, we may assume that T is
not a star. Then T contains an edge e such that T − e contains two nontrivial
trees T1 and T2 where Ti has order ni for i = 1, 2. By the induction hypothesis,
ni
α0 (Ti ) ≥ for i = 1, 2.
1 + ∆(Ti )
Let Xi be an independent set of α0 (Ti ) edges in Ti for i = 1, 2. Then X1 ∪ X2
is an independent set of T . Thus,
n1 n2
α0 (T ) ≥ α0 (T1 ) + α0 (T2 ) ≥ +
1 + ∆(T1 ) 1 + ∆(T2 )
n1 n2 n1 + n2 n
≥ + = = .
1 + ∆(T ) 1 + ∆(T ) 1 + ∆(T ) 1 + ∆(T )

Now let G be a connected graph of order n ≥ 2 and let T be a spanning tree

of G. Then
n n
α0 (G) ≥ α0 (T ) ≥ ≥ ,
1 + ∆(T ) 1 + ∆(G)
as desired.

23. Proof. By Exercise 22,

n jnk
≤ α0 (G) ≤ .
1 + ∆(G) 2
By Gallai’s theorem (Theorem 12.11),
148 CHAPTER 12. MATCHINGS, INDEPENDENCE AND DOMINATION

α0 (G) + β 0 (G) = n.

Thus,
n n · ∆(G)
β 0 (G) = n − α0 (G) ≤ n − =
1 + ∆(G) 1 + ∆(G)
and jnk lnm
β 0 (G) = n − α0 (G) ≥ n − = ,
2 2
as desired.

If n = 2k and G = Pk  K2 , then β 0 (G) = n

n
2 = 2 . If G = K1,n−1 , then
n·∆(G)
β 0 (G) = n − 1 = 1+∆(G) .

24. A nonempty graph G has the property that every two distinct maximal inde-
pendent sets of vertices are disjoint if and only if G is a complete multipartite
graph.
Proof. First, observe that every complete multipartite graph has this prop-
erty. For the converse, assume that G is a graph that is not a complete
multipartite graph. Suppose that the minimum number of independent sets
into which V (G) can be partitioned is k. For each partition {V1 , V2 , · · · , Vk }
of V (G) into independent sets, there is a vertex u in some set Vi that is not
adjacent to a vertex v of some set Vj where i 6= j. Among all such partitions
of V (G) into k independent sets, let {U1 , U2 , · · · , Uk } be one where U1 con-
tains such a vertex u and |U1 | is maximum. We may assume that v ∈ U2 .
Because of the maximality of U1 , v is adjacent to some vertex of U1 . Thus,
any maximal independent set containing U1 does not contain v and any max-
imal independent set containing {u, v} obviously contains v. Thus, these two
maximal independent sets are distinct. Since both contain u, they are not
disjoint.
25. α(G) + β(G) = n and α0 (G) + β 0 (G) = n − k.
26. The statement is true. Proof. Assume first that G has a perfect matching.
Then α0 (G) = n/2. By Gallai’s theorem (Theorem 12.11), β 0 (G) = n/2 and
so α0 (G) = β 0 (G). For the converse, assume that α0 (G) = β 0 (G). Then by
Theorem 12.11, α0 (G) = n/2, implying that G has a perfect matching.

Section 12.3. Domination

27. Let Q3 = C4  K2 , where C4 = (v1 , v2 , v3 , v4 , v1 ) and the corresponding copy
of C4 is (v10 , v20 , v30 , v40 , v10 ). Then {v1 , v30 } is a minimum dominating set in Q3
and so γ(Q3 ) = 2.
Every vertex in Q4 dominates 5 vertices. Since Q4 contains 16 vertices,
γ(Q4 ) ≥ 4. Observe that Q4 = Q3  K2 . A dominating set of cardinality
4 can be obtained from minimum dominating sets in each copy of Q3 . Then
γ(Q4 ) = 4.
 149

28. (a) For n ≥ 3, γ(Cn ) = dn/3e.

First, we write n = 3q + r, where 0 ≤ r ≤ 2. Since Cn is 2-regular,
every vertex of Cn dominates exactly three vertices. Therefore, any q
vertices of Cn dominate at most 3q distinct vertices of Cn . If r = 0, then
this says that γ(Cn ) ≥ q. If r = 1 or r = 2, then γ(Cn ) ≥ q + 1.
Suppose first that r = 0. Let S be the set consisting of any vertex v of
Cn and every third vertex of Cn beginning with v as we proceed cyclically
about Cn . Then every vertex of Cn is dominated by exactly one vertex
of S. Since S contains exactly q vertices, γ(Cn ) ≤ q. Next, suppose that
r = 1 or r = 2. Now let S be the set consisting of any vertex v of Cn
and every third vertex of Cn beginning with v as we proceed cyclically
about Cn until we have a total of q + 1 vertices. Then every vertex of
Cn is dominated by at least one vertex of S. So, S is a dominating set
of Cn and γ(Cn ) ≤ q + 1. In both cases, γ(Cn ) = dn/3e.
(b) For n ≥ 2, γ(Pn ) = dn/3e. First, observe that γ(Pn ) ≥ γ(Cn ) = dn/3e.
For each cycle Cn and minimum dominating set S of Cn , let e be an
edge of Cn that joins two vertices not belonging to S. Then S is also a
dominating set for Cn − e = Pn and so γ(Pn ) ≤ dn/3e.
29. Let G be a graph of order n. Then γ(G) = 1 if and only if ∆(G) = n − 1.
Proof. Assume that γ(G) = 1. Then G contains a dominating set consisting
of a single vertex v. Since v dominates every vertex of G, it follows that v
is adjacent to all other vertices of G and ∆(G) = n − 1. Next suppose that
∆(G) = n − 1. Then G contains a vertex v of degree n − 1. Since v is
adjacent to all other vertices of G, it follows that {v} is a dominating set and
γ(G) = 1.
30. Proof. We begin with the lower bound. Let S be a minimum dominating
set of G. Then [
V (G) − S ⊆ N (v),
v∈S

implying that |V (G) − S| ≤ |S| · ∆(G). Therefore, n − γ(G) ≤ γ(G) · ∆(G)

and so γ(G) ≥ n/(1 + ∆(G)).
Next we establish the upper bound. Let v be a vertex of G with deg v =
∆(G). Then V (G) − N (v) is a dominating set of cardinality n − ∆(G); so,
γ(G) ≤ n − ∆(G).
31. Proof. Let d = diam(G) ≥ 3 and let v ∈ V (G) such that e(v) = d. Define
V0 = {v}, V1 = {x : d(v, x) = 1}, V2 = {x : d(v, x) = 2} and V3 = {x :
d(v, x) ≥ 3}. Then

V (G) = V (G) = V0 ∪ V1 ∪ V2 ∪ V3 .

Since d ≥ 3, it follows that V3 6= ∅. Let w ∈ V3 and let S = {v, w}. Observe

that v is adjacent in G to every vertex in V2 and V3 and w is adjacent in G to
v and every vertex in V1 . Thus, S is a dominating set of G and so γ(G) ≤ 2.
150 CHAPTER 12. MATCHINGS, INDEPENDENCE AND DOMINATION

Since G is connected, G has no isolated vertices and so ∆(G) ≤ n − 2, where

n is the order of G. Thus, γ(G) = 2 by Exercise 29.

32. (a) No. Assume, to the contrary, that there exists a graph G without isolated
vertices such that γ(G) = β(G) but γ(G) is strictly less than each of the
numbers α(G), α0 (G) and β 0 (G). Thus γ(G) < α0 (G). Since α0 (G) ≤
β(G) by Theorem 12.12, it follows that

α0 (G) ≤ β(G) = γ(G) < α0 (G),

which is impossible.
(b) (i) Does there exist a graph G without isolated vertices such that γ(G) =
α(G) but γ(G) is strictly less than each of the numbers β(G), α0 (G)
and β 0 (G)?
Yes. Let G = 2K4 . Then γ(G) = α(G) = 2, β(G) = 6, α0 (G) = 4
and β 0 (G) = 4.
(ii) Does there exist a graph G such that γ(G) = α0 (G) but γ(G) is
strictly less than each of the numbers α(G), β(G) and β 0 (G)?
Yes. Let G the graph shown in Figure 12.3. Then γ(G) = α0 (G) = 4,
α(G) = 7, β(G) = 5 and β 0 (G) = 8.
............. ........... ............ .........
..... ....... ....... .........
... ... ..... ...
...
...
..... ...
... .
. ..... ....
... ... ............
........... ... ....
... ... ...
... ..
G: ...
...
..
.
............ .......
... ........ .... ....
... ..... .........
.
. ... ..... ...
.. ... ..... ...
... ... ... ...
... ......... ..... .........
............ ... ...
... ............ ........
... ...

Figure 12.3: The graph G in Exercise 32(b)(ii)

(iii) Does there exist a graph G such that γ(G) = β 0 (G) but γ(G) is
strictly less than each of the numbers α(G), β(G) and α0 (G)?
No. Assume, to the contrary, that there exists a graph G such
that γ(G) = β 0 (G) but γ(G) is strictly less than each of the numbers
α(G), β(G) and α0 (G). Thus, γ(G) < α(G). Since α(G) ≤ β 0 (G) by
Theorem 12.12, it follows that

α(G) ≤ β 0 (G) = γ(G) < α(G),

which is impossible.

33. Proof. Let G be a graph of order n without isolated vertices. Since α(G) +
β(G) = n and α0 (G) + β 0 (G) = n, it follows that

γ(G) ≤ min{α(G), α0 (G), β(G), β 0 (G)} ≤ n/2,

as desired.
 151

34. Proof. Let G be a graph of order n ≥ 2. The lower bound follows imme-
diately from the observation that if γ(G) = 1, then γ(G) ≥ 2. It remains
to verify the upper bound. If G has an isolated vertex, then γ(G) ≤ n and
γ(G) = 1; while if G has an isolated vertex, then γ(G) ≤ n and γ(G) = 1. So
in these cases, γ(G) + γ(G) ≤ n + 1. If neither G nor G has an isolated vertex,
then γ(G) ≤ n/2 and γ(G) ≤ n/2 by Corollary 12.17 and so γ(G)+γ(G) ≤ n.

35. (a) The statement is true.

Proof. Let S be an independent minimum dominating set. Since S
is a minimum dominating set, |S| = γ(G) ≤ i(G). On the other hand,
since S is an independent dominating set, i(G) ≤ |S| = γ(G). Thus,
γ(G) = i(G).
(b) Let G be a double star with V (G) = {u, u1 , u2 , v, v1 , v2 } where deg u =
deg v = 3, N (u) = {u1 , u2 , v} and N (v) = {v1 , v2 , u}. Then γ(G) =
2 and {u, v} is the only minimum dominating set in G, which is not
independent.

36. (a) Let U and W be the partite sets of G, where |U | ≤ |W |. Since U is an

independent dominating set and |U | ≤ n/2, it follows that i(G) ≤ n/2.
(b) Let G = Kr,r . Then n = 2r and i(G) = r.

37. If G = P7 , then γ(G) = i(G) = 3.

For k ≥ 4, let G be the graph obtained from P3 = (u, v, w) by adding 3(k − 2)
new vertices u1 , u2 , . . . , uk−2 , v1 , v2 , . . . , vk−2 and w1 , w2 , . . . , wk−2 and joining
(1) all vertices ui (1 ≤ i ≤ k − 2) to u, (2) all vertices vi (1 ≤ i ≤ k − 2) to
v and (3) all vertices wi (1 ≤ i ≤ k − 2) to w. Then {u, v, w} is a minimum
dominating set of G and so γ(G) = 3. We show that i(G) = k. Since
S = {u, v1 , v2 , . . . , vk−2 , w} is an independent dominating set, it follows that
i(G) ≤ k. Let T be any independent dominating set of G with |T | = i(G).
Observe that if x ∈ {u, v, w} and x ∈ / T , then xi ∈ T for all i with 1 ≤ i ≤ k−2.
If v ∈ T , then u, w ∈ / T and so ui , wi ∈ T for 1 ≤ i ≤ k − 2. Thus,

|T | ≥ 1 + 2(k − 2) = 2k − 3 > k,

which is impossible. Thus, v ∈ / T and so vi ∈ T for 1 ≤ i ≤ k − 2. If exactly

one of u and w is not in T , say u ∈
/ T , then ui ∈ T for 1 ≤ i ≤ k − 2. Thus,

|T | ≥ 1 + 2(k − 2) = 2k − 3 > k,

which is impossible. If both u and w are not in T , then

|T | = 3(k − 2) = 3k − 6 = k + (2k − 6) > k,

which again is impossible. Thus, u, w ∈ T and T = S. Hence, i(G) = k.

38. (a) Proof. By Theorem 12.19, γ(G) ≤ β(G) and, by definition, i(G) ≤
α(G). Therefore, γ(G) + i(G) ≤ α(G) + β(G) = n by Theorem 12.11.
152 CHAPTER 12. MATCHINGS, INDEPENDENCE AND DOMINATION

(b) Let G be the corona of Kk . Then the order of G is 2k and γ(G) = i(G) =
k.
39. The statement is true.
Proof. Let G be a nontrivial connected graph and let S be a minimal domi-
nating set of G. By Theorem 12.16, the set V (G) − S is a dominating set. Let
T be a minimal dominating subset of V (G) − S. Thus, S and T are disjoint.
40. The statement is true. Since L(G) is claw-free, the result follows by Theo-
rem 12.22.

41. (a) Proof. Since every total dominating set in G is also a dominating set,
γ(G) ≤ γt (G). It remains to show that γt (G) ≤ 2γ(G). Let S =
{v1 , v2 , . . . , vk } be a minimum dominating set of G. The vertices in
V (G) − S are therefore dominated by the vertices of S. Since G contains
no isolated vertices, each open neighborhood N (vi ) is nonempty. Now,
let ui ∈ N (vi ) (1 ≤ i ≤ k) and let S 0 = {u1 , u2 , . . . , uk }. Thus, the ver-
tices of S are dominated by the vertices of S 0 . Therefore, the set S ∪ S 0
is a total dominating set of G. Hence γt (G) ≤ |S ∪ S 0 | ≤ 2|S| = 2γ(G).
(b) For each integer n ≥ 3, let Hn be the corona of the n-cycle Cn . Then
γ(Hn ) = γt (Hn ) = n. For each integer n ≥ 3, let Fn be the graph
obtained by adding a pendant edge to each end-vertex of Hn . Then
γ(Fn ) = n and γt (Fn ) = 2n.

42. (a) G1 = Kn for n ≥ 2

(b) G2 = Pn for n ≥ 3.
(c) Claim: A graph G has an irregular dominating set if and only if N [u] 6=
N [v] for every two distinct vertices u and v of G.
Proof. If N [u] 6= N [v] for every two distinct vertices u and v, then
the set N [u] of vertices that dominate u is different from the set N [v] of
vertices that dominate v. Hence, V (G) is an irregular dominating set of
G.
For the converse, suppose that G has an irregular dominating set S =
{v1 , v2 , . . . , vk }. Let u and v be any two distinct vertices of G. Then
dom(u) 6= dom(v). We may assume that dom(u) = (a1 , a2 , . . . , ak ) and
dom(v) = (b1 , b2 , . . . , bk ), where, say, a1 = 1 and b1 = 0. Then v1 ∈ N [u]
and v1 ∈ / N [v], that is, N [u] 6= N [v].
43. No. If G is the claw K1,3 , then γ(G) = i(G) = 1.
Chapter 13

Factorization and
Decomposition

Section 13.1. Factorization

1. Suppose that the Petersen graph P is 1-factorable. Since P is cubic, there

is a 1-factorization {F1 , F2 , F3 } of P into three 1-factors. Then the spanning
subgraph F with edge set E(F1 ) ∪ E(F2 ) is a 1-factorable 2-factor. Because
P contains no triangles or 4-cycles, each 2-factor in P is either a Hamiltonian
cycle or two 5-cycles. By Theorem 6.9, the Petersen graph is not Hamiltonian.
Thus, the 2-factor consists of two 5-cycles. Since such a 2-factor is not 1-
factorable, P is not 1-factorable.

2. The 4-cycle C4 has this property.

3. (a) Since Qn is bipartite and n-regular, the result follows by Theorem 13.1.
(b) If Qn is k-factorable, then k | n. On the other hand, if k | n, then
n = kr for some integer r. By (a), Qn is 1-factorable and so Qn has n
1-factors F1 , F2 , · · · , Frk . Grouping every k 1-factors together produces
a k-factorization of G with r k-factors.

4. Let n = 2k, where k ≥ 5 is an odd integer, and let Hi = Kk for i = 1, 2. Now

let ui vi ∈ E(Hi ) for i = 1, 2 and let

G = H1 + H2 − u1 v1 − u2 v2 + u1 u2 + v1 v2 .

Then G is a (k − 1)-regular graph of order 2k. However, G is not 1-factorable,

for suppose that it is. Then G has a 1-factorization F consisting of k − 1
(≥ 4) 1-factors. Necessarily, there exists a 1-factor F ∈ F containing neither
u1 u2 nor v1 v2 . Since H1 and H2 have odd orders, some edge of F must join
a vertex of H1 and a vertex of H2 . This is impossible.

153
 154 CHAPTER 13. FACTORIZATION AND DECOMPOSITION

5. Since n ≡ 0 (mod 4) and n ≥ 8, it follows that n = 4k for some integer k ≥ 2.

Let H1 = K2k−1 and H20 = K2k+1 , let C be a Hamiltonian cycle in H20 and
let H2 = H20 − E(C). Next, let G = H1 + H2 . Thus, G is a (2k − 2)-regular
graph of order n, where 2k − 2 = n2 − 2. Since H1 and H2 have odd order, G
does not contain 1-factor and is therefore not 1-factorable.

6. Let C = (v1 , v3 , v5 , v7 , v1 , v2 , v4 , v6 , v8 , v2 , v3 , v4 , v5 , v6 , v7 , v8 , v1 ) be an Eule-

rian circuit of G = C82 . The graphs G, H, F1 and F2 described in the proof
of Theorem 13.4 are shown in Figure 13.1. Corresponding to the 1-factor
F1 of H is the 2-factor G1 = C 0 + C 00 where C 0 = (v1 , v3 , v5 , v7 , v1 ) and
C 00 = (v2 , v4 , v6 , v8 , v2 ). Corresponding to the 1-factor F2 of H is the 2-factor
G2 = C 000 = (v1 , v2 , v3 , , v4 , v6 , v7 , v8 , v1 ).

....
v1
... ....
................ .......
........ .... .... ........
.
...
.......... ......... ......... ................
.
... . ..... .............
v 8 ................... .....
.....
.....
..... ........ 2 v
. . ..... .....
..... .... .....
.. ..
u1 u2 u3 u4 u5 u6 u7 u8
... ... .....
... ... ......... ..... ... .. ................. ................. ................ ................. .................... ................. .....
..... ..
.......
..... .
... .........
.
..... .. ...
........ ... ........
... ......
......... ........ ......... ......... ........ ....................... ....................................................
... ...... ... ...... ... ...... ... ...... ... ...... ...... ...... .....
.. .......
........ ...
... ...... ...
... ...... ... ...... ... ..... ... ..... .. .....................................................................................
.. ...... .. ... .............
... .... . ... ...... .
... ..... ... ..... . .. ....... ..... . .....
... ..........................................................................................
. ...
G: v7 ...................... .... ... ............ v3 ... ......
... .....
... .....
... ......
... .....
... ...... . .......... ..........
...
. . ........ ... ..... ...
... ..... .. ... ..... ... ... ..... ... ..... ... .. ............. ............... ............ ......... ... ...... ...
... ........ ..... ..... .......................................................... ...............................
(a) ... .......
... ......... .......... ....
....... .. H: ...
...
...
. . .....
... . .. .. . . .. .. .
..... ..... ..................................................... ................................ ........ ....
. .
. . . . .
.....
.....
..... ....
... .....
. ......
. ..... .....
.
...
.
... ... ...... ..
..... .... ... ... .... .............. .............. ..... ... .............. ... ..... ... ..... ... ....... ...
... .. ...... ... ... ...
. ... ............................................. ..... ............ ..... ... ..... .. ..... ...
... .
..
.
........
......
..... ...
.....
... ... ..
.. (b) ................................. ...... ... ............... ..........
........................... ..................
. .
...
........
...... ...
..
..... ...
..
..
.
..... ...
.
........
..
..... ..
..
.
.
..
..... ...
. .
...... ....
........
v ... ..... ........ ...................... ........ ........
6 .................... ........... ..... ........... .. .... .
.......... .................
.......
......... ...........
.......
........
.......
.............
.
...............
....... ..... ....... v ........................ ......... ........
....... ......... ........ .............. 4 ........ ........ ...
....... .......... ......
....... .........
......
w1 w2 w3 w4 w5 w6 w7 w8
v5
u1 u2 u3 u4 u5 u6 u7 u8 u1 u2 u3 u4 u5 u6 u7 u8
............... ................. .................. .................. ................. ............... ......... ......... ................ ................ ............... ................ ................ ................ ................ ........
..... ..... ..... ..... ..... . .............. ........... ... ... . ................
..... ..... ..... ..... ..... ................................. ......................... ...
...
...
... ... ...
...
...
... . ...................
..... .....
.....
.....
.....
.....
. ....... .... ....... ..
.
.... .. ..
.... ... ... ... ... ... ..................................... .....
.....
..... ..... ..... .................................. ................................. .......... ... ... ...
... ... ............ .... ..
..
...
. ..
. .
.... .
...
F1 : .....
.....
.....
....................................
. ..
. ........................ .................................
................
. ..
.....
.....
.....
.....
.....
.....
..... F2 : ...
...
...
...
...
..
... ...................................
.................... ...
...
...
...
...
...
...
..........
. . .. .... . .
....... .................. ..
. .. .
. .. . . . . . .
..... ........ .....
..... .....
..... ... ................................... ... ...
.
...
.
...
...
...
...
.
..... .
... .. . . . . . . .. .
...... ..
. .. ..... .
.... . .. . . .
. .
. .. .. .. . ..
............... ......... ..... ..... ..... ......... ........ ....... ....... ........ ........ .......
...................... ................ .......
........
............
........
.. .
.... .... . . ............ ............
........ ................... ... .... ... .. ... .. ........ . ....... . .
....... . ............
...
.. .. ...... ...... .... .. ..... .....
w1 w2 w3 w4 w5 w6 w7 w8 w1 w2 w3 w4 w5 w6 w7 w8

v v1
........
............
1 (c) ......
...... .....
..... ......... ........ .... ..............
.... ........ ........
..
.. ..... .
............. ........
v ...... ..... .....
v ............. ..............
8 .......................................................................................................................... v 8 .......... ........ 2
... v
.... ...
.. ..... 2 ... ...
.. .....
... ........ ..... ..
. ...
G1 : .. ......
............
.....
.....
..... G2 : ...
..
. ...
...
...
.... ..... ... ...
....... ..... ..
..............
........... ... ........... .
.............. v3
v 7 ................ .... . .
........... v3 v7 ... ...
.
.....
....... ......
. ... ..
... ....... ....... ... .
...
.... ...
.. ... ..... ... ...
.. ......... ..... ... ...
.. ... ..... . ..
v ............................................................................................................................. v v ................ ..........
.
6 .....
..... ..
.....
... 4 6 ..........
........ .......
..
... ... .
4 v
..... ..... ........ ......
..... ..... ........ .......
.............. ............... ............
.... .... ... ....
.... .....
v5 v5
(d)

Figure 13.1: The graphs in Exercise 6

7. See Figure 13.2.

8. Proof. It suffices to show that there exists an r-regular graph of order n if

at least one of r and n is even (and 0 ≤ r ≤ n − 1). Suppose first that n is
 155

v1 v1 v1 v1
................. .............. ............ ......
.
......
...... v8 ........
... ...... ...... v2
v8 ...
v8..............................
. .
v8 ........................ ........ .............. ............ .. ............ v2 ............... ........ ......
..... v2 ... ................. . ............ v2
........ ... .......... .... .
. .. ... ...... .............
......... ..... ...................... .........
r
... ...... . .
......... ... ..... v0 ............ .............. ... ... v0 ....
r ... .
v7......... v.7......................... ......... .........
r
v ......... v7..... ..
... ...... ...... ... .......... .
r
... 0 .
. ............. . .
v7 ........................ ..... ....... ..........
........
... .......... ...
. .. .
.......
. ....... ... ... ............ .............
......... v0 ...................
.... ...... v3
......... .. ... ..... ............ ......... ..... ... ... ... ..... ..... ......... .....
.......... v3 v3 .. .. .. ..... ............. .....
... .........
.............
... ...... .........
............... ...
..... ...
....... ..
.
. ...
. .... v3
. . . . . .......................... ................
..
...
v6 ................... ......
.
...... v6........ ................. ..........
..... v6........
... ..
.. ............ v6 ......... ..... .......
.... ......... v4
............ ......... ... ............
...... v4 ........ v4 ............ v4 ...................

v5 v5 v5 v5

Figure 13.2: A Hamiltonian factorization of K9 in Exercise 7

even. Then n = 2k for some positive integer k. By Theorem 13.2, K2k can
be factored into 1-factors F1 , F2 , . . . , F2k−1 . The union of r of these 1-factors
produces an r-regular graph of order n.
Next, suppose that n is odd. Then r is necessarily even. The graph K1 is
0-regular of order 1, so we may assume that n = 2k +1 ≥ 3. By Theorem 13.5,
K2k+1 can be factored into Hamiltonian cycles F1 , F2 , . . . , Fk . The union of
r/2 of these Hamiltonian cycles gives an r-regular graph of order n.
9. Proof. Since n is even, there exists a k-regular graph H of order n − 2 with

V (H) = {v1 , v2 ,· · ·, vn−2 }.

Let G be obtained from H by adding two new vertices u and v and the edges
uvi , vvi for all i with 1 ≤ i ≤ k. Then deg vi = k + 2 for 1 ≤ i ≤ k while the
remaining vertices of G have degree k.
10. Proof. By Theorem 13.5, the graph G = K2k+1 has a Hamiltonian factor-
ization. Let v ∈ V (G). Deleting v from G and from each factor produces the
desired factorization of K2k .
11. Assume, to the contrary, that K2k+1 can be factored into

Hamiltonian paths.
Since every Hamiltonian path has 2k edges, 2k | 2k+1
2 and so 2k | k(2k + 1).
Thus, 2 | (2k + 1), which is impossible.
12. Proof. Since the result is true for 1 ≤ k ≤ 3, we may assume that k ≥ 4.
Let H = K2k−1 , where V (H) = {v1 , v2 , . . . , v2k−1 }. Let the vertices of H
be placed cyclically about a regular (2k − 1)-gon and let the edges of H be
straight line segments. Let

F1 = (v1 , v2 , v2k−1 , v3 , v2k−2 , . . . , vk+1 )

be the Hamiltonian v1 − vk+1 path of H whose edges are all those parallel
to v1 v2 or to v2 v2k−1 . By rotating the path F1 through a clockwise angle of
2π(` − 1)/(2k − 1) radians for ` = 2, 3, . . . , k − 1, we arrive at Hamiltonian
v` − vk+` paths F` that are pairwise edge-disjoint. Furthermore,
k−1
!
[ [
E(H) = E(F` ) {vi v2k−i : 1 ≤ i ≤ k − 1}.
`=1
156 CHAPTER 13. FACTORIZATION AND DECOMPOSITION

We now construct a graph G = K2k by adding a new vertex v2k and joining
it to every vertex of H. For ` = 1, 2, . . . , k − 1, define G` as the Hamiltonian
cycle of G obtained by joining v2k to v` and to vk+` in F` . Let G0 be the
1-factor of G with

E(G0 ) = {vi v2k−i : 1 ≤ i ≤ k − 1} ∪ {vk v2k }.

Then {G0 , G1 , . . . , Gk−1 } is a factorization of K2k into k − 1 Hamiltonian

cycles and a 1-factor.
13. Let b = 2k + 1, where k ≥ 1. Then 2a + b + 1 = 2a + 2k + 2. By Theorem 13.7,
the graph K2a+b+1 can be factored into a + k Hamiltonian cycles and a 1-
factor. Since each Hamiltonian cycle has even length, each of the Hamiltonian
cycles can be factored into two 1-factors. Thus, K2a+b+1 can be factored into
a Hamiltonian cycles and 2k + 1 = b 1-factors.
14. By Theorem 13.7, the graph K6k+4 can be factored into 3k + 1 Hamiltonian
cycles H1 , H2 , . . . , H3k+1 and a 1-factor, which we denote by F2k+1 . For
j = 1, 2, . . . , k, the Hamiltonian cycle H2k+1+j can be factored into two 1-
factors F2j−1 and F2j . For i = 1, 2, . . . , 2k + 1, the 3-factor Gi with E(Gi ) =
E(Hi ) ∪ E(Fi ) is Hamiltonian.
15. Let C 0 = (v1 , v2 , v4 , v6 , v8 , v7 , v5 , v3 , v1 ) and C 00 = (v1 , v7 , v5 , v6 , v4 , v3 , v2 , v8 , v1 ).
Then C 0 and C 00 form a Hamiltonian factorization of C82 .
16. Proof. Let H = G − E(C) = C3 + C5 , where C3 = (u1 , u2 , u3 , u1 ) and
C5 = (w1 , w2 , w3 , w4 , w5 , w1 ). Since every vertex of C3 is incident with two
edges of C, it follows that C5 has two chords that belong to C. There are two
possibilities, depending on whether these two chords are adjacent.
Case 1. The two chords are adjacent. We may assume that w1 w3 and w1 w4
are the two chords of C5 that belong to C. Necessarily, no vertex of C3 is
joined to both w3 and w4 . Hence we may assume that

u1 w2 , u1 w5 , u2 w3 , u2 w5 , u3 w2 , u3 u4 ∈ E(G)

and so C = (u1 , w5 , u2 , w3 , w1 , w4 , u3 , w2 , u1 ). Thus,

C 0 = (u1 , u2 , w3 , w4 , u3 , w2 , w1 , w5 , u1 ) and C 00 = (u1 , u3 , u2 , w5 , w4 , w1 , w3 , w2 , u1 )
constitute a Hamiltonian factorization.
Case 2. The two chords are not adjacent. We may assume that w1 w3 and
w2 w4 are the two chords of C5 that belong to C. Necessarily, no vertex of C3
is joined to both w1 and w3 or to both w2 and w4 . Also, no two vertices of C3
are adjacent only to the vertices w1 , w2 , w3 , w4 . Hence, we may assume that
u1 w2 ∈ E(C). There are three possibilities.
Subcase 2.1. u1 w1 ∈ E(G). Then we may assume that

u2 w3 , u2 w5 , u3 w4 , u3 w5 ∈ E(G).
 157

In this situation,

C 0 = (u1 , u2 , w5 , u3 , w4 , w3 , w2 , w1 , u1 ) and
C 00 = (u1 , u3 , u2 , w3 , w1 , w5 , w4 , w2 , u1 )

give a Hamiltonian factorization of G.

Subcase 2.2. u1 w5 ∈ E(G). Then we may assume that

u2 w1 , u2 w5 , u3 w3 , u3 w4 ∈ E(G).

Here,

C 0 = (u1 , u2 , w1 , w2 , w3 , u3 , w4 , w5 , u1 ) and
C 00 = (u1 , u3 , u2 , w5 , w1 , w3 , w4 , w2 , u1 )

produce a Hamiltonian factorization of G.

Subcase 2.3. u1 w3 ∈ E(G). Then we may assume that

u2 w1 , u2 w5 , u3 w4 , u3 w5 ∈ E(G).

Here,

C 0 = (u1 , w3 , w2 , w4 , w5 , w1 , u2 , u3 , u1 ) and
C 00 = (u1 , u2 , w5 , u3 , w4 , w1 , w2 , u1 )

constitute a Hamiltonian factorization of G.

17. For n ≥ 3, let G be the graph obtained by adding a new vertex v to Kn−1 and
joining v to two distinct vertices u and w of Kn−1 . Then any Hamiltonian
cycle of G contains both edge uw and vw.

18. (a) Since δ(G) = 4k + 1 = n/2, it follows that G is Hamiltonian by Dirac’s

theorem.
(b) Since every Hamiltonian cycle of G must contain exactly two edges in
G[B], it follows that G contains at most k = (n − 2)/8 edge-disjoint
Hamiltonian cycles.

Section 13.2. Decomposition

19. Let P = (u, x, z, w, v). Rotating P twice clockwise through an angle of
2π/3 radians, we obtain P 0 = (v, z, y, u, w) and P 000 = (w, y, x, v, u). Thus,
{P, P 0 , P 00 } is a cyclic P5 -decomposition of the graph of the octahedron shown
in Figure 13.7 in Chapter 13.
158 CHAPTER 13. FACTORIZATION AND DECOMPOSITION

20. Proof. Assume, to the contrary, that there exists a K1,4 -decomposition of
the graph of the octahedron. Since the size of G is 12, there are three graphs
in this decomposition. Each vertex in G is either (1) the vertex of degree 4 in
a subgraph K1,4 , (2) does not appear in a subgraph K1,4 , or (3) a vertex of
degree 1 in a subgraph K1,4 . However, any vertex of degree 1 in a subgraph
K1,4 must belong to four such subgraphs, which is impossible.
21. Proof. Suppose that there exists a Steiner triple system of order n ≥ 3.
Then the complete graph Kn is K3 -decomposable. In order for Kn to be K3 -
decomposable, the size of Kn must be divisible by 3, that is, 3 must divide
n


2 and so n(n − 1)/6 must be an integer. Hence either 3 | n or 3 | (n − 1).
Furthermore, since each vertex of Kn belongs to a certain number of copies
of K3 , its degree, namely n − 1, must be even and so n is odd. This says that
either n = 3p for some odd integer p or n = 3q + 1 for some even integer q.
Therefore, either n = 6k + 1 or n = 6k + 3 for some integer k and so either
n ≡ 1 (mod 6) or n ≡ 3 (mod 6).
22. Since the graph K9 is 3K3 -factorable as shown in Figure 13.3, nine schoolgirls
can take four daily walks in three rows of three girls each so that no two girls
walk in the same row twice.
v 1 v1
..................... ........
... ..........1 ...........
... ............ ..... ..
.......... ... ... ....
v ....... ..... ...
v
8 .......... v 0......
..v 2 8 .................................... .... ..............v
v ... ... 2
... ...
... ...
... ... ....0 .....
s .. ..
.......... .....................0 .. .
..
..... .....
......... ................... 0 ... ....
..... ............
..
s ... . .. ...
... .. ...
................. ............. ..... ... .......... ... ..
v .............. .. 4 ..... ........................ .......... ... ..... .
.
...........
F1 : 7 ........ ..... ...... ... v3 F2 : v 7 ........... ... ..... ........
..... .. ... .........
. .. v3
.. ... .. ... .
. .. .
.
2 ...... ...
... 3...... 2 ... ..
... .. .
. .
.
.
..
.
...... ... ....
.. .. . .
.
.
... ... 3 ... ... .. ... .................
........ ... .
..............
v v ....... ... ...... ................ v
v6 ..................... ...... 4 6 ........ . ... ........ 4
........ ...... .. ..
.
...... ... ................
1 ... v5
v5
v1 v1
.............. .....
...... ....
... ........ .... .........
... ........
v ........ ... ........ v .............. .....
..... .............
8........ ... v
............ 2 8 ..................... ... .... 2
........... ....
v
. . .......... .........
.................... ................
.
.. ............
v s
. . . . . . . .. .. . ... .. ..
sv
.. .. . .
.. ......... ... ........ ........................ ........
.. ......... 0... .
... ......... .......
v ....... ......... .. 0 .....
..... .. ......... ..... ........ ..
F3 : 7 .............. ... ......... ...... v3 F4 : v .
7 ....................................... ........... ..... v3
.... ................ .... ..... .. ..............
. .
.. . . .. .. ..... .. .... .........
....... ... ....... ..... .........
........ .... .. .. .........
......... .. ... ............ ...............
v ............... .. ....... ............. ......... .. ..
6.......................................................................... 4 v v6 ...... ..... ......... 4 v
.. ..... .......
.............. .....................
......
v5 v5

Figure 13.3: A cyclic 3K3 -factorization of K9 in Exercise 22

23. (a) Let V (Kn ) = {v1 , v2 , . . . , vn }. For 1 ≤ i ≤ n − 1, let Gi be the star of

size i with central vertex vi+1 containing the edges vj vi+1 for 1 ≤ j ≤ i.
(b) Let n = 2k + 1 ≥ 5. By Theorem 13.5, there exists a Hamiltonian
factorization F = {F1 , F2 , . . . , Fk } of Kn . For 1 ≤ i ≤ k, Fi can be
decomposed into two paths Gi and G0i , where Gi has size i and G0i has
size n − i. Then G1 , G2 , . . . , Gk , G0k , G0k−1 , . . . , G01 is an ascending sub-
graph decomposition of Kn with the desired property.
 159

24. Denote the four paths by

F1 = (u1 , u2 , u3 , u4 , u5 , u6 , u7 , u8 ),
F2 = (v1 , v2 , v3 , v4 , v5 , v6 , v7 , v8 ),
F3 = (w1 , w2 , w3 , w4 , w5 , w6 , w7 , w8 ),
F4 = (x1 , x2 , x3 , x4 , x5 , x6 , x7 , x8 ).
Let {G1 , G2 , . . . , G7 } be the edge-induced subgraphs of K8 with
E(G1 ) = {u4 u5 }, E(G2 ) = {v3 v4 , v5 v6 }, E(G3 ) = {w2 w3 , w4 w5 , w6 w7 },
E(G4 ) = {x1 x2 , x3 x4 , x5 x6 , x7 x8 }, E(G5 ) = {u3 u4 , u5 u6 , v2 v3 , v4 v5 , v6 v7 },
E(G6 ) = {u2 u3 , u6 u7 , v1 v2 , v7 v8 , w3 w4 , w5 w6 },
E(G7 ) = {u1 u2 , u7 u8 , w1 w2 , w7 w8 , x2 x3 , x4 x5 , x6 x7 }.
Then {G1 , G2 , . . . , G7 } is an ascending subgraph decomposition of 4P8 in
which each subgraph Gi is 1-regular.
25. Let U = {u1 , u2 , u3 } and W = {w1 , w2 , · · · , w7 } be the partite sets of K3,7 .
Then {G1 , G2 , . . . , G6 } (shown in Figure 13.4) is an ascending subgraph de-
composition of K3,7 in which each subgraph Gi is connected.

u 1 u.....2 u3 u1 u2 u3
.......
......... .......... ......... ................ ............... ...............
............ ......
........
........ ... ...
.
...
..........
. ... ....
G1 : ........
........
........
.
G2 : .... ....
.. ...
........ .. ...
.......... .. ...
.
..................
..
...
............... ... ..
...
............ ............... ............... ............... ............... ............. ............
.. ............. ............... ............... ............. ...............
.. .. ..
w1 w2 w3 w4 w5 w6 w7 w1 w2 w3 w4 w5 w6 w7

u1 u2 u.....3 u1 u2 u3
............... .............. .... ..... ............... ................ ................
..... .. ..... .. ..
..... ...
...
.....
..... ... .....
..... ... ..... .
.... ...
G3 : .....
.....
.....
.....
...
...
...
G4 : ..... ..
..... ...
......
...
...
...
..... ... ............ ...
.....
............ ... .. .. ....... ...
. ........ .
............... ............... ............... ... .... ............. ............... ............... ............... ............... ............... ............... ............... ... .... ............
.. .. .. ...
w1 w2 w3 w4 w5 w6 w7 w1 w2 w3 w4 w5 w6 w7

u1 u 2 u3 u 1 u 2 u 3
............... ........ ....... ............................... ......... ........
................ ...........
........... ........................ ............................ ........
.
..................... ...... ..... ............................. ....... ..................
... ............... ..................... .......... ..
.. .. . .
.
.. . ..... ....... ... .. . . ...... ...... ....... .......... .............
G5 : ..
.
....... ......... ....
..
.
...... ..... G6 : ...... ....... ..................... .
...... ...... ..... ...... ............................. ................................. ............
.
....... ......... .... ........... ........
. ...... . ....... ........ .......... .......
....... . . . ......... .................. .
.. ....... ............... .............................
.......
. .....
. ............. ......... .
.
..
................... .
.
.....
. ....
. ..
. ... . .
..
... .... ............... ............... .......................
.
............... .................. ............... ................. ............... ............... ............... ..................... ............... ............. ............. ... .... ... .... ........
... ... .. .. ... ..
w1 w2 w3 w4 w5 w6 w7 w1 w2 w3 w4 w5 w6 w7

Figure 13.4: An ascending subgraph decomposition of K3,7 in Exercise 25

160 CHAPTER 13. FACTORIZATION AND DECOMPOSITION

26. See Figure 13.5.

u1 u2 u1 u2
s s v.....1.. v2 s s v.....1. v....2..
......... ............. .......... ... ....
G1 : ..
G2 :
...

u6 s ..
......
su3 u6 s s u3
......
......
.
.......
.....
.
.. ............... .............. ............... ..............
s ......
...... s v4 v3
s s v4 v3
u5 u4 u5 u4
u1 u2 u1
s ......
s v1 v2 s
.............
u2
s v1 v2
...... ...
............ ........ ... ....... ................. ...............
...... ......... ... ........ .....
G3 : .
...
....... ...
... G4 : ... ...... .....
su3 s su3
..
s
... ...... .....
...... .. ......
...... .
...
... ... .....
u6 ... u6 ...
...
.....
.....
... ... .....
.
.... ... .....
. ... .. .
..................................................
..............
s s ...............
s ...
s
..
v4 v3
..
v4 v3
u5 u4 u5 u4

u1
s u2
s .. v1 v......2.
u1
s us2
... v1 v2
... ............... ... .... ... ... .............. ...............
... .. ... ...
G5 : ..
. G6 : ... ...
s s u3
. .. ...
s
.
u6
...
................................................................
....... u6
.
.....
.
su3
...
..
....... ..... ...
... ...
....... ... ...
. ...
... ............ ...
.............. .............. ...
s ...
...
s
.......
..
v3
s
...
s
...
... ............... ...............
v4 v4 v3
u5 u4 u5 u4

Figure 13.5: An ascending subgraph decomposition of K6 + K4 in Exercise 26

Section 13.3. Cycle Decomposition

27. In addition, 3 | 92 , 4 | 92 and 9 | 92 . Hence, there is a C3 -decomposition,

a C4 -decomposition and a C9 -decomposition. For a C3 -decomposition, see

Figure 13.3 for the solution of Exercise 22. For a C9 -decomposition, see Fig-
ure 13.2 for the solution of Exercise 7. For a C4 -decomposition, consider
a regular 9-gon whose nine vertices are equally spaced and arranged in the
order v1 , v2 , . . . , v9 . Now consider the 4-cycle C = (v1 , v2 , v4 , v7 , v1 ). A C4 -
decomposition of K9 is obtained by cyclically rotating C about the regular
9-gon.

28. The minimum number of cycles of distinct lengths in a decomposition of K7

is 4, which occurs with the decomposition {C3 , C5 , C6 , C7 }, where V (K7 ) =
{v1 , v2 , . . . , v7 }, C3 = (v1 , v3 , v6 , v1 ), C5 = (v1 , v4 , v2 , v6 , v5 , v1 ),
C6 = (v2 , v5 , v3 , v7 , v4 , v6 , v2 ) and C7 = (v1 , v2 , . . . , v7 , v1 ).

29. The decomposition in Figure 13.6(a) consists of one 12-cycle, one 6-cycle
and three 4-cycles and so there is no odd cycle. The decomposition in Fig-
ure 13.6(b) consists of one 12-cycle, one 9-cycle, one 6-cycle and one 3-cycle
and so there are two odd cycles. The decomposition in Figure 13.6(c) consists
of one 12-cycle, one 9-cycle and three 3-cycles and there are four odd cycles.
The decomposition in Figure 13.6(d) consists of six 5-cycles and there are six
 161

odd cycles. The decomposition in Figure 13.6(e) consists of one 9-cycle and
seven 3-cycles and there are eight odd cycles.
s s s
qq qqqq qq qqqq qqqqqq
. . ... .. .... ..
..... ......... ..... .......... ..... ........
..... ..... .....
qq q qqqq qq q qqqq qq q qqqq
..... ..... .....
..... ..... ..... ..... ..... .....
.
....... ..... .
......
. .....
..... . ....... .....
s qq qqq s
. .
s qq qqqq s
.
s q qqq s ... .... .. ... .....
.. .
...... ..... ..... ..... ..... .........
. ..... .... . .. . ...
........... ..... ..... ..... ..... ..... ............. .
................ ..... ..... ..... ..... ..............
qq qqq
. .
qqq
..
..... . .
..... ..
qqq
... ...... ...... ..
qq qq
. . .. ...... .... ..
... ... .... ... . ....... ...... ... ... ......... ..... ..
qq q qqq qqq
... ..... . ... ........ .... ...
qqq ..... . ...... .. . ..
..... .. ..... ... ... ...... ..........
qq q qq q
...... .
... . .................
qq ..
qqq
... ... ........ .
qqsq
... ...
... ..... ..... . .. ...
...... ............ ..
. . .. .. ...
q
. ... .......... ........... ..
.... ....
s ... ............
qq sqqqq qs qqqqq qqsqqqq
. ..... ...... ..
s qqqqq q s
......... ......... ..... ..... ........... .......................
qqqqq s qqqqq q ... .. ...... .
qqqqq s
..... ..... ...
s qq sqq . .......
s
...... .

qqq qqqq qqqqq qqqqqqqqq

q
.. ........
................ .... ...... ..... ..... .. ............ ..... .. ..... ... ... ..... ........ ... . ..... ..... ...........
............. . .. ...... ... . .............. ..
... . .....
qqqq . ..
.. ... qqqqq ... ....
..
..... ... ...
qq q
.........
..... ..........
q
. .. .
... ........ qqqqq ...... ..
....... .... .............. ... ....
.......
....... ....
qqq
... .. ..........
.
...
... .. .
........ ..
qq
.... .... ..... .... .
q qqqqqq
..... .. ....
qq
.... .... ............. .... ..........
q qqqqq .
qq
.... ... ............. ....
q qqqq qq qq qqqqqqq
... ....... .. ..
... .............. ... ....
qqq qqq qqqq
... .. ... ..... . ... ........ .
...... ... ....... . ...
... . . . ... . . . . ... ........
qq
. . ...

qqqq q
. . ... ..........
qqqqqqq sqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq q
... .... . . . . . . . ... ..
qqqqqqqqqq qsqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq
.
q
...
qqqq
.. ... .. .. .. ... ..............
qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq
. . ..... .... .... ..... . .. ... ... ............. ................... ..................
q
.. . .. ... . . ... . ...
q qq
.
q
...
q
. .
q
.
qs qq ....... ........... ...
...
q qq q
qsqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq q qq qq qq qsq qqqqqqqqqqqqqqqqqqqqqqqqqqqqqq
...
.. ...
q qqqqq ...
qqq .. ...
...
q q
sqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq
... .. .................... .. ...
.. . .... ...... .. ...
...
...
.. . ... ... ...
....
. .. ....... ...
qqqqq q s qqq qs
. .. .
. . . ... ... . . ...
s
. .
.
.. ... ..
.
.
.
....... ......
...
.
... ..
.
...
... ...
...
. .......... . .
. ...
. .. ....
...
. .
.
.. . .
.
..
.
... .
.
. ...
............ . .
. .
..
.... ..... . .
..
. . .
...
.. . . ......... . ..
. . .... .... ... .
. .
... . . ... ........ .. .
.. .... ..
..... ... .
... ..
. .... . ....
s
............ ... . ...... ..... .. ... ............ . ..... ... .. ......
s .....
......... . ........ .
... ............ ......... ... .... ..... ...... ........ ... .... .... .....
.
s
.......... ..... . .
s ........ ... ............ . .
........ ... ............ .
s s
. .. ............... ........ .... .
........ .......... .......... ............
. .
(a) (b) (c)
..........s .......s
. ...
. ....
...... . .. ..... ......... .. ......
..... . . ...... ... . .. ...
...... ... ... ........... ...... ... .. .......
s .....
.............. ..... ........ .............. .............
. s .
. s ......
...... ... ..... ...........
. .
................ .............. ..... ......... ........ ... s.
..... . ..
... .... .. ... ..... ..... ... . ....... .. .. ..
... ....... ....... .... ... .... ... ...... ..
... .. ..... ..... ... .. ... .. .. .. ..... ..... ...
... . ... .. .. ... ... ...... . ..
... .. ..... .. ... ..... .... .... ...... ..... . ..... . ..... ....
s ......
s
..........
sq ... ...
s .... ....
................... ............ ..
qq
.. .....
s
. ..................
s s
.
qqq s
qqq qqqqq
............ .. ..............
..... .. ....................................
.............. .... ... . .. .. .................... .......... ...... . ......
.. .. .. ... . .. . ..... .. ... ...... ..... ..... ....
... ..
... ..... ..... ... .. .... ... .. ... ... ...... .... ..... ..... ........ ..
... .... .......... .. .. ... ... .. . ........ .. ... .... ... ... ........... .. . qqq qq . ......... ... ...
.... .. ... . . . .
. .....
...
... ... .
...
.......
...... ........ ...
. ... ........ ......
qqq qq
.. .... ... .......
.
. . .... ...... . ...
.. .
... ........
.. .
... ....
...
. ...
.
. .... .. .... ..
s . .
.... ....
. ..
. ......... ....... ..... .
... . . ...
.
... ... .. ....
.
.. ........
... ...... ...
..........
.. s .. .. .. ..... ..
.... ... .
.......................................... .............................................. ..
........ ...
. ........ ....
s
. .
s ..........
. .. ...... .... .... ......... ......... ..... ........ . .. .......
............... ..... ..... ..... .. .........
.......... .... ... .........
.
.....
..... .. ....
..... .. . ..................
.
... .........
.. s ........
.
................................... ...
. . ......
....
... .....
.....
..... ... ...
.
. .
.
.................................
. ...
........
s
..... . . ................
s s
........
........ ... ... ........
s
................
s
...... ................ ............... ...... . .........
.... ..... ...........

(d) (e)

Figure 13.6: Graphs in Exercise 29

The only way to have 10 odd cycles is to have ten 3-cycles. Since there do
not exist three edge-disjoint 3-cycles in K5 , this is impossible.

30. The only possible decomposition of K7 of this type consists of three 3-cycles
and three 4-cycles. Let V (K7 ) = {v1 , v2 , . . . , v7 }. Three 3-cycles are
(v1 , v2 , v3 , v1 ), (v3 , v4 , v5 , v3 ) and (v5 , v6 , v7 , v5 )
and three 4-cycles are
(v2 , v6 , v3 , v7 , v2 ), (v1 , v6 , v4 , v7 , v1 ) and (v1 , v4 , v2 , v5 , v1 ).

31. (a) Consider K1,2k .

(b) Let G be the graph obtained by subdividing k edges of K1,2k once each.

32. Assume that the size of G is even. (If the size of G is odd, then the proof is
similar.) If t = 0, then the result is true vacuously. Thus, we may assume that
t ≥ 2. It suffices to show that there exists a circuit decomposition of G, exactly
t − 2 of which have odd length. Among all circuit decompositions of G, con-
sider those circuit decompositions containing exactly t circuits of odd length;
and, among those, consider one, say D = {C1 , C2 , . . . , Ck } for some positive
integer k, where the distance between some pair Ci , Cj of circuits of odd length
is minimum. We claim that this minimum distance is 0. Assume that this is
not the case. Suppose that P is a path of minimum length connecting a vertex
162 CHAPTER 13. FACTORIZATION AND DECOMPOSITION

wi in Ci and a vertex wj in Cj , and let wi x be the edge of P incident with wi

(where it is possible that x = wj ). Then wi x belongs to a circuit Cp among
C1 , C2 , . . . , Ck . Necessarily, Cp has even length, for otherwise, the distance
between Ci and Cp is 0, producing a contradiction. Since Ci and Cp have the
vertex wi in common, Ci and Cp may be replaced by the circuit C 0 consisting
of Ci and Cp (that is, E(C 0 ) = E(Ci ) ∪ E(Cp )) and C 0 has odd length. How-
ever then, the circuit decomposition D0 = ({C1 , C2 , . . . , Ck }−{Ci , Cp })∪{C 0 }
has exactly t circuits of odd length and the distance between Cj and C 0 in D0
is smaller than the distance between Ci and Cj in D, which contradicts the
defining property of D. Thus, as claimed, the distance between Ci and Cj is 0
and so Ci and Cj have a vertex in common. Hence the circuit C ∗ consisting
of Ci and Cj has even length. Then ({C1 , C2 , . . . , Ck } − {Ci , Cj }) ∪ {C ∗ } is a
circuit decomposition of G, exactly t − 2 of which have odd length.

Section 13.4. Graceful Graphs

33. If this graph G is graceful, then it is possible to label the five vertices of G with
five of the seven labels 0, 1, . . . , 6 in such a way that the resulting edge labels
are 1, 2, . . . , 6. If the three vertices of a triangle are labeled with three integers
of the same parity, then all edge labels of the triangle are even numbers. If
the three vertices of a triangle are labeled with exactly one or exactly two
even labels, then two edges have an odd label. So both triangles must have
an even number of odd labels. Thus, it is not possible for three edges to be
labeled 1, 3, 5.
34. It is graceful. See Figure 13.7.
............
... .....
..
..
... 2 .......
...
............ ...... ...........
..... ......
.
. ...... ......
......
.
....
.................
..
5 1 ............
...
. .....
.... ... .... ...
... 7
..................
.
. 3
..... ......
. . ..
..... ..... .....
.
..... 7
.....
..... ............... ...........
3 .... .. ..
..... .. ........
.....
0
...
.
.... .... ....
6 ..... ........ ......... 4
..... .....
.... .....
..... ..................
.
.................... .. ...
.....6 ... ..... ..
..... ......
........
4
..... ......
........
.

Figure 13.7: A graceful labeling of the graph in Exercise 34

 163

35. (a) In Figure 13.8, label the seven vertices of K7 by 0, 1, 2, . . . , 6. In particu-

lar, the vertex labels 0, 1 and 3 produce a graceful labeling of K3 . Since
the size of K3 is 3, it follows by Theorem 13.18 that there exists a cyclic
K3 -decomposition of K7 .
(b) Proceed as in (a) with a regular 9-gon rather than a regular 7-gon to
obtain a K3 -decomposition of a 6-regular graph of order 9 into nine
copies of K3 .
0 .............. ..............
................... 1
... ......... ........ ..............
... ............. .............. ......... ..............
6 .............. ... ... ... 1 .... ..
... ..
... .... .....
... ... ..
..... ...
... .. ... .......
3 .. 2 .............. .... ......
.........................................................................
.
. .
H2 : .... ..........
H1 : .............. ... .. .
.............. 2 .... ............. H3 : ......... ..
5 ... ..
... ...
..
.
.... ......
.........
......... ....
.... ...... ..... .
................
... ..
... ... ...................... .........
.
............. .......
............. ..
. ..............
4 3
....... .............. .......
.............. ........ ................
... ... .............. ....... ...
.............. ..... ... ............................................................... ....... ... ..............
.............. . ... ................
................. .
.. ...
. ... ........
........ ........ ...
..
H4 : ... ..... H5 : .. H6 : .. ........ H7 : ........
........ .....
... ...... ... ... ... ........
... ...... ... ........ ........ ...
.............. ... ............. ......................... ..............
........ ..
............
............. .....
.....
..... .............. ..... ... . . .............. ......
. .. ..... ..... ...
... ..... ..... ...
.....
... . . ............... .............. ............. .............. .
............. ............
............... .............. . ..

Figure 13.8: A K3 -decomposition of K7 in Exercise 35(a)

36. Figure 13.9 shows a cyclic F -decomposition of K12 .

....
...... ......
..
...
...... ...... 0 ....
............ ...
...... ......
2 ........
.. . .. ...... ...
............. ... ......
..
. ....
...
.............. .....
..... ...
3 ............. ...
............
....
..... ......
...
1 .....
..............
....
..
............
...... .................. 0 ....
...... .....
..
...
...
...
.... .
...... ..... ..
.. ...
.... 2 ... 3 ... 5 ...
...... .................................................................................................................................... ...... ..
.
...
...... ..... ..... ...... ....
...... ...... . ...
.. ... ..
.............
.... .. ..........
..
.
................. .........
4 .
.............
.............
..
...
..
.............
....
..... ......
4 ....
...... .....
5 ...
..... ......
...
...... ...................
..
...
..... .......
...
... .. ...
..... ....
............. ...... ......
..
1

Figure 13.9: The graphs for Exercise 36

164 CHAPTER 13. FACTORIZATION AND DECOMPOSITION

37. Figure 13.10 shows a cyclic P6 -decomposition of K10 .

....
...... ........
.. ........
0
........ 1
........
...
...... ......
..
...
....... ......
.
2 .....
............................................................................................................
..... . ............... ...... ......
....... ........... .....
1 .....
.
.....
. ...........
........... 0 ...
...
.....
....
.. 3 ...........
...........
...........
...........
...
...
..... .................. ...
..... ....
...... ...... ..... ... .
..... ..
. ............ 2 ........
.. .. ..
..
... ...
... ...... ...
..... .... 4 .............. ..
.
.
.
..
...... .... ...
.
..... ............
..... ........
3 .....
..... .... ....
......... ...... ...... ...... ......
.... .. ..
.....
.......
...... ......
..
4 ...
...... ......
.. .... ...
...... ...... ...... ......
.. ..

Figure 13.10: The graphs for Exercise 37

38. (a) Since K1,k is graceful (by Theorem 13.23) and has size k, it follows by
Theorem 13.18 that K2k+1 is K1,k -decomposable.
(b) By the same argument as in (a), the graph K2k−1 is K1,k−1 -decomposable.
In fact, K2k−1 can be decomposed into 2k − 1 copies of K1,k−1 . Each
vertex of K2k−1 is therefore the center of a star K1,k−1 . Add a vertex
v to K2k−1 and join v to each vertex of K2k−1 , resulting in K2k . The
graph K2k is therefore decomposed into 2k − 1 copies of K1,k .

39. Proof. Let m = 2k. Then K3m+1 = K6k+1 . Arrange the vertices v1 , v2 ,
· · ·, v6k+1 cyclically about a regular (6k + 1)-gon and join every two vertices
by a straight line segment, producing the complete graph K6k+1 . Let C =
(v1 , v2 , · · · , v6k+1 , v1 ).
For every two distinct vertices x and y of K6k+1 , dC (x, y) ≤ 3k. Assign
each edge xy of K6k+1 the value dC (x, y). Since G is graceful, its vertices can
be labeled 0, 1, · · · , 2k so that each edge is labeled with one of the integers
1, 2, · · · , 2k. Embed G in K6k+1 so that the labelings of the edges coincide.
Rotating this copy of G about K6k+1 6k times gives a decomposition of K6k+1
into 6k + 1 copies of G. Every vertex of K6k+1 is incident with two edges
labeled 2k + 1, 2k + 2, · · · , 3k that are not used in this decomposition. Hence,
the subgraph H remaining after the decomposition is m-regular.
 165

40. (a) Consider the drawings of F1 , F2 and F3 in Figure 13.11(a), each of which
is the first subgraph in an Fi -decomposition. A cyclic Fi -decomposition
can then be obtained.
(b) Yes. A decomposition {F1 , F2 , F3 } is shown in Figure 13.11(b).
..
...........
..
...... .....
..
..... ..... ... ...
.... ............
..................... ................ .......
... ........ ........ ... .
.........
...
. ........ .............. ... ..
.......
.
..
...
. ....
....... ............... ... .......
... ........ ... .. ....... ..
..
................... ........
.............. ... ...... .............. ...... .....
.... .. .. . . ... .... ......
.......
....
.
... .............
.. ................. ... ..............
... ......... ..... . ... ..
...
. .... ....
..
..... ..... ....... .
. . ..
..... .......
.... ... ...... .....
..... ....... ......
.
.
.
. .......
.... ... ...
.... ....
..... .........
... ........ .... ..........................
... ......... ...........
. ..
... .........
...........
.. ..
... .........
............................... ... ... ..
...... .....
... ... ....
... ..... ... .. ..
... .. .. .....
.
.
... .. ...
... .. ... .... ....
.......... ... ... .............. ..............
... ..
.... ...............
....

F1

.. ....... ..
..
...... ...... .... ... ... ... ...... ....
........... .................. .. ................... ....
....... ....... ... ...
....... ............. ... .. ........
. ........
.......... ... .. ........
....... .............. ... ... ........
....... ....... ... ... ...... ...............
................ ............... ... .. .... ... .... .
.
... ... .... ... ... ...
. .... ....... ........
...
..... .
.... ... ... .
.
... ...
..........
.. ..
. ....
. ...
.. .
....... ..............
. .............
..... ...... ..
...................
. ..... ...... .
... ..
.... ...... .....
.... ..... ....

.. .............
...... ..... .....
....
..
.. .. ...... ..... ....
..............
..... ...... ...... .... ....
.... ....

F2 F3
(a)

.......
.................
..... ... ..... .
..... ....
..... .... .....
..... ........... .....
..... .... ...
. . .
.
.....
..... . ..
.....
..... .. ..... .....

qqqqqqqqqqqqq q
qq qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq qqqqqqqqqqqqqqq
.. ..... .. .
. ........
..... ..... . ...
.
... .... ......
..... .. . .. . .....
... . .
q
... ..... ...... ... .
... ....
qqqqqqqq qqq ..
qqqqqqqq qqqqqqqqqqqq
... .... . . .... ..
... ... ...
... ...
...
qqq qqq
.
. . ..

qqqqqqqqqqqqq
... .. ...
... ...
... . ...
qqqqqqqq qqqqqqqqq
... ... ...
... ..
... ... ..
... ...
. .... ..
... .... .... ..... ...... ...
.... .......
...
... ... ..... ...
... ... ..... ..
..
. .
...
... . ..... .
..... .............
..... .... ... ....
.... ....

(b)

Figure 13.11: The graphs for Exercise 40

166 CHAPTER 13. FACTORIZATION AND DECOMPOSITION

41. The Petersen graph P is not K1,3 -decomposable, for otherwise P can be de-
composed into five subgraphs K1,3 . The five central vertices in these five
subgraphs K1,3 necessarily form an independent set. Since α(P ) = 4, this is
is impossible. On the other hand, P is cyclically F -decomposable when F is
any of P4 , 3K2 , P3 + K2 . The first subgraph in such a cyclic F -decomposition
is shown in Figure 13.12.
......
......... ..... ..
......... ...... ......
........ .... ............... ......
...... .. ....... . .......
...... ........
. .....
..
...... ......
.
.. ......
.
.....
..
......
....
............. ...... ....
..............
.................
.
............. .. ................ .. ....... .
.............
...... ...... . ...... ...... ...... ..... ....
..... ..... ... ..... ... ...
............. .... .. ..
...... .
.......
..... .. ...... ..... ... ..
.... ......... ... ...
.... ........
... .... ....... ... ... ..
..... ... ..........
.. ...... .....
.. ..... .....
.
... ....... .
.....
.
.
............. ..
...... ..... ....
.... .......
... .. . ..... ..
..... .... .... ...... ..... .............. ...... .....
..... .... ....
...
.. ...
...... ..... ....
.............. .
...
......
.... ............. ..... .....
..
...... ..... .............
.... .... .... .....

P4 3P2 P 3 + P2

Figure 13.12: The graphs for Exercise 41

42. See Figure 13.13.

............ ..........
..... .....
......... .............
0
........... .............. ............... .............................. ......... .............. ...... ..... ..............................
0
... ...... ..... .... ...
..... ..... ...................
8 16 16
8..
. .. .
..... .......
. 8 15 15 ... ........
.......... .....
.....
.
.....
.... 8 ..................
......
..... ..... ...... ..... ................
............ ........... ..... ...... 1 15 1 .... ....
..... ..... 1 14 .. .
.... ..... 7 ... ... ............
9 ............
.. .
1
............. .... . ......... ...
. ... ... ...
... ... ........... 2 14 ..........
.
..
.. 13 ... ..... .....
9 ..... ...
.... . . . ...2. ............ ............ 15
............
..... .... .... ...
14 ... ...
7
........... ............ ... ...
...
...
... .. 3 13 ..
... .............
3 12 ... ........... ..... ...
C .. .... ..... C 2
. ..
..... ....
........ 15 .........
.
.
..... ....
6
..........
... 16 ............
10 ........... 2
. . ... 12
.
...
... .
. ........ ..
... .
. 4 ...
..
...4 11 ... ... .
..........
.
... .... ............ ..... ....
............ . 10 .... .... 14
. . ............ ........... ...........
..... ...
6 ..... ....
13 ...
5
... 11 ....
.
........... ............. .
5
.... . ... ....... ...
.... 10 ...
. ... .
..... ..... ...............
.... ... 5 ..... ....
.... ........
... .
.... ....
.......... ...
..... .....
..............
6 ..... 10 3 ..
. ............
11 6 ................. 9 .. 3 . ..... ...... ............ ......
..
7 .. .
.... ....... .. ..... ... .... 9 13 ... .....
....... ............
......... ..
5 ... ...
........... .......
..... ..........
12 11 7
....... ..
...................
.
............. ..... ...
.................. ....................... .......
4
......... .......... ..........

Figure 13.13: Graceful labelings of C15 and C16 in Exercise 42

 167

43. A cyclic C6 -decomposition of K13 can be obtained by rotating the 6-cycle

shown in Figure 13.14.

v0......... v
............
1
......... ...
...............
....................
..............
. ............... .......
............. ... v
v12 ... ...
...
...... ..
......
...
2
..
. ...
... ..
.
... ...
v11 ............. ... ... .....
.... ..
...
...
... ...
.. .......... ...
.
..... ....... v3
... .. ........ ...
. .
... ....... ...
... ...... ...
..... ..
... ..... .... .. ..........
v10 ................ ... .
.....
.
....... .
. .
... ...... v4
... ..... ... ..
..
... ..... ..
.. .......
. .. .....
.
.
. ... .
... ..... ... ...
... ..... ... ...
............. ..... .............
v9 ...
.
...
.. ..
. ......
.....
.
.
.. ..
.. .... ... v5
... .......... .. ...
.
... ........ .. ...
.
... ... ....
................ ..............
v .............
v 6
8 ...
v7

Figure 13.14: A cyclic C6 -decomposition of K13 in Exercise 43

44. (a) See Figure 13.15.

..........
.... ...
........... .......................
0
.. ..
............. ........... ..........
..................
.
..................... 8 . .
12 .... .... 12 .... .... 8
............ ..............
. ...
... ...
... ...
.... ...
...
6
..
... ... 10
. ...
.. . .. . . ..
.
.. ..
..... ..... ............
6
............ .... .... 18
... ............
.
...
... ...
..
...
...
C 9 ..
.
.
5
... ... 3
... ...
... ...
.. ... ..
.
..... .... ............
1.... .... .... ...
............ ....... .... 15
.....
..... ..... .....
.
15.....
.....
.
.....
... 1
..... .....
..... .....
..... ........... 2 .....
.... ..
...................
.
16 ... ...
.........
... ..
........... 14

Figure 13.15: A labeling of C9 in Exercise 44(a)

(b) For every integer i ∈ {1, 2, . . . , 9}, there are two adjacent vertices in G
whose distance is i on the 19-cycle (v0 , v1 , v2 , . . . , v18 , v0 ). This implies
that K19 can be decomposed into copies of C9 .
(c) What (a) and (b) say is that if there is a labeling of the vertices of a
graph G of size m with distinct elements of the set {0, 1, . . . , 2m} such
that the set of induced edge labels has a nonempty intersection with
each of the sets Si = {i, 2m + 1 − i} for i = 1, 2, . . . , m, then K2m+1 is
G-decomposable.
168 CHAPTER 13. FACTORIZATION AND DECOMPOSITION

45. See Figure 13.16.

............... 5 ............... 4 ...............
.....
3 .............. 2 ...............
.....
1 ..............
P6 : .... 0 . ....
5 . 1 . ..... . 2 . ..... 3 .
.............. ............. .............. 4
.............. .............. ...............

....
..... ...... 6 ............... 5 ............... 4 ..
...... ........ 3 ....
..... ...... 2 ..
...... ........ 1 ....
..... ......
..... .....
P7 : .... 0 .
.............. 6 .
............... 1
..............
. .... 5
...............
.... 2 .
..............
.... 4
...............
.... 3 .
..............

............... 8 ............... 7 ............... 6 ................ 5 ..

...... ........ 4 ....
..... ....... 3 ..
...... ........ 2 ............... 1 ..
...... ........
..... ..... ....
P9 : 0
..............
. 8
..............
. 1
..............
. ... 7 ...
.............
.... 2
...............
.... 6
...............
.... 3
...............
.... 5
..............
. .... 4
...............

..
...... ........ 9 ...............
.....
8 ....
..... ...... 7 ....
..... ....... 6 ...............
..... 5 ....
..... ....... 4 ...............
....
3 ...............
.....
2 ...............
.....
1 ....
..... .......
P10 : .... 0
............... 9 .
...............
....
1 .
..............
.... 8
.............. 2
..............
. .... 7
............... 3
..............
. 6 .
............... 4
..............
. .... 5
...............

Figure 13.16: Graceful labelings of P6 , P7 , P9 and P10 in Exercise 45

46. (a) Let G be a double star with

V (G) = {u, u1 , u2 , . . . , ua } ∪ {v, v1 , v2 , . . . , vb }
where u and v are not end-vertices, u is adjacent to ui (1 ≤ i ≤ a) and
v is adjacent to vj (1 ≤ j ≤ b) and u. Thus, n = a + b + 2, m = a + b + 1
and m = n − 1. Then the labeling f of G defined by f (u) = n − 1,
f (ui ) = i − 1 for 1 ≤ i ≤ a, f (v) = a and f (vj ) = a + j for 1 ≤ j ≤ b is
a graceful labeling.
(b) Let T be a caterpillar of order n and let P = (v0 , v1 , . . . , vk ) be the
spine of T . We define a graceful labeling f : V (T ) → {0, 1, . . . , n − 1}.
First, define f (v0 ) = 0. Suppose that deg v0 = d0 ≥ 2. Label the
d0 − 1 neighbors of v0 not on P by n − 1, n − 2, . . . , n − (d0 − 1) and
then define f (v1 ) = n − d0 . The edges incident with v0 are then labeled
n − 1, n − 2, . . . , n − d0 . Suppose that deg v1 = d1 ≥ 2. Label the d1 − 2
neighbors of v1 not on P by 1, 2, . . . , d1 − 2 and define f (v2 ) = d1 − 1.
The edges incident with v1 are then labeled n − d0 , n − d0 − 1, . . . , n −
d0 − d1 + 1. Continuing in this manner produces a graceful labeling of
T . (See Figure 13.17 for a caterpillar of order n = 14.)

13 1 2 11 4 5 9 8 7
............... ............. ............... ...
.......... ............... ............... ............... ............... ...............
.... .. .. ... .. ... ..
... .. ..
... ... ...
... ..
...
... ..
... ... ... .
. ... 2 ..
... ... ... ..
. ... ...
13 11 ...
... ..
..
10 8 6 ... ...
.. 5 3 ...
... ...
.. 1
... .... ...
... .... ... ..
.
... .. ... ... ... . .
......... .. ............
............... .......... ............... ............... ........

0 12 12 9 3 7 10 4 6
v0 v1 v2 v3 v4

Figure 13.17: A graceful labeling of a caterpillar of order n = 14 in Exercise 46(b)

 169

47. (a) If F is a forest of order n with k ≥ 2 components, then the size of F

is m = n − k ≤ n − 2. Thus, any mapping from V (F ) to {0, 1, . . . , m}
cannot be one-to-one. Hence, F does not have a graceful labeling.
(b) We show that the disconnected graph G = P3 +C3 , where P3 = (v1 , v2 , v3 )
and C3 = (v4 , v5 , v6 , v4 ), is not graceful. Assume, to the contrary, that
there exists a graceful labeling f : V (G) → {0, 1, . . . , 5} of G. For each
of the numbers 0, 1, 2, . . . , 5, some vertex is assigned that number. In
particular, some vertex is labeled 0. We consider three cases.
Case 1. f (v1 ) = 0. Then f (v2 ) = 5. Then either f (v3 ) = 1 or f (v3 ) = 4,
say the former. The vertices v4 , v5 and v6 are labeled 2, 3 and 4, which
implies that two edges of G are labeled by 1, which is impossible.
Case 2. f (v2 ) = 0. Then f (v1 ) = 5. Thus f (v3 ) = 1 or f (v3 ) = 4, say
the former. Then v4 , v5 and v6 are labeled 2, 3 and 4, which again is
impossible.
Case 3. f (v4 ) = 0. Then f (v5 ) = 5. Thus f (v6 ) = 1 or f (v6 ) = 4. In
either case, two edges are labeled 1, producing a contradiction.
48. Proof. Since Kk is graceful for 2 ≤ k ≤ 4, the result is true for 2 ≤ k ≤ 4.
Hence, we may assume that k ≥ 5. Let the vertices of Kk be labeled with
distinct elements of the set {0, 1, 2, . . . , r} for some positive integer r in such
a way that distinct edges receive distinct labels. We may assume that some
vertex v of Kk is labeled
0 and some vertex of Kk is labeled r. Hence the edges
of Kk are labeled with k2 elements of the set S = {1, 2, . . . , r} . Consequently,
there are s = r − k2 elements of S for which no edge of Kk has this label.

Suppose that A = {a1 , a2 , . . . , as } is the subset of S such that no edge of Kk

is labeled with an element of A. Therefore, no vertex of Kk is labeled with
any element of A either.
Next, let G be that graph obtained by adding s vertices v1 , v2 , . . . , vs to
Kk and joining each of these vertices to v. Assign the label ai to vi for
i = 1, 2, . . . , s. Then G is a graceful graph with χ(G) = k.
49. Let G be a graceful graph of order n and size m where f is a graceful labeling of
G that assigns the vertices of G the labels a1 , a2 , . . . , an . The complementary
labeling f replaces the vertex label ai by m − ai . Since ai ∈ {0, 1, 2, . . . , m}
(1 ≤ i ≤ n), it follows that m − ai ∈ {0, 1, 2, . . . , m}. Furthermore,

|f (ai ) − f (aj )| = |(m − f (ai )) − (m − f (aj ))| = |f (ai ) − f (aj )|.

So f is also a graceful labeling of G.

170 CHAPTER 13. FACTORIZATION AND DECOMPOSITION
Chapter 14

Vertex Colorings

Section 14.1. The Chromatic Number of a Graph

1. Proof. Suppose that the vertices of G are colored 1, 2, 3. Let v be a vertex
colored 3. Thus, the neighbors of v are either all colored 1 or are all colored
2. If all neighbors of v are colored 1, then recolor v with the color 2; while
if all neighbors of v are colored 2, then recolor v with the color 1. Let w be
another vertex colored 3. Since v and w are not adjacent, no neighbor of w
has been recolored. Thus, each neighbor of w is colored 2 or 1 and w can be
recolored 1 or 2, respectively. We continue to recolor each vertex colored 3,
resulting in a proper coloring of G with the colors 1 and 2. Thus, χ(G) ≤ 2
and so χ(G) = 2.

2. Proof. Let c be a k-coloring of G using colors 1, 2, . . . , k and let V1 , V2 , . . . , Vk

be the color classes. We claim that for each pair i, j of integers with 1 ≤ i 6=
j ≤ k, there is an edge joining a vertex in Vi and a vertex in Vj ; for otherwise,
Vi ∪ Vj is an independent set and so all vertices in Vi ∪ Vj can be colored the
same. This implies that G can be colored by k − 1 colors, which is a contra-
diction. Thus, as claimed, for each pair i, j of integers with 1 ≤ i 6= j ≤ k,

there is an edge joining a vertex in Vi and a vertex in Vj . Since there are k2
pairs of color classes, m ≥ k2 .

3. Claim: χ(H) = k. Proof. Let c be a k-coloring of G with color classes

V1 , V2 , . . . , Vk . Two vertices of H are adjacent if and only if the vertices
belong to different sets among V1 , V2 , . . . , Vk . Thus, H is a complete k-partite
graph with partite sets V1 , V2 , . . . , Vk and so χ(H) = k.

4. Proof. First, observe that χ(G − S) ≤ k − 1. If χ(G − S) ≤ k − 2, then

χ(G) ≤ k −1, which is impossible. Thus, χ(G−S) = k −1. By Corollary 14.5,
there is a component H of G − S such that χ(H) = k − 1.

5. Proof. Let G be a connected graph of order n and diameter d and let

P = (v0 , v1 , v2 , . . ., vd ) be a v0 − vd geodesic of length d in G. Assign

171
172 CHAPTER 14. VERTEX COLORINGS

the color 1 to the vertex vi if i is odd, the color 2 to the vertex vi if i is even,
and the colors 3, 4, . . . , n − d + 1 to the remaining n − d − 1 vertices of G. This
produces an (n − d + 1)-coloring of G and so χ(G) ≤ n − d + 1.
6. The complete k-partite graph G = K2,2,...,2 is (2k − 2)-regular. Since Kk is
a subgraph of G, it follows that χ(G) ≥ k. Let the k partite sets of G be
V1 , V2 , . . . , Vk . By assigning the color i (1 ≤ i ≤ k) to each vertex of Vi , a
k-coloring of G is produced. Thus, χ(G) ≤ k and so χ(G) = k.
7. Proof. If G is r-regular, let H = G. Otherwise, let G0 be another copy of
G and join corresponding vertices whose degrees are less than r, calling the
resulting graph G1 . Let Si be the color class whose vertices are colored with
color i, where 1 ≤ i ≤ k. Let Si0 be the set corresponding to Si in G0 . Color
the vertices in Si0 with i + 1 for 1 ≤ i ≤ k − 1 and color the vertices in Sk0
with color 1 (see Figure 14.1). Then the resulting graph G1 has χ(G1 ) = k
and the degree of each vertex in G1 is 1 more than the corresponding vertex
in G if the degree of each such vertex in G is less than r. If G1 is r-regular,
let H = G1 . If not, we continue this procedure until arriving at an r-regular
graph Gp , where p = r − δ(G) and χ(Gp ) = k, as desired.

G G0
.......................
...... ... ........
...................
....
1 ....... ..
.....................
.....
.......
...
.
...
.................... 2
..................... ....................
....... ... .......
..... ..... ...
2 ....... ...
..................... ......
.......................
...
3
q q
q q
q q
........................ ........................
...... .. ......
k−1 ....
.......
.....................
... ....
.......
.....................
..
... k
...................
........ ... .....................
.......
.....
k .......
.....................
.
... .....
.......
...
...
.....................
1

Figure 14.1: The graph in Exercise 7

8. The graph H is a bipartite graph, where one partite set is the set of vertices
added to G to form H and the other partite set is V (G). Thus, χ(H) = 2,
which implies that k = 2.
9. (a) Proof. Let G be a nonempty graph and let H be a graph obtained
from G by subdividing a single edge xy of G. Suppose that the edge
xy is replaced by a new vertex z and the two new edges xz and zy in
H. Assume that χ(G) = k and let there be a k-coloring of G with the
colors 1, 2, . . . , k. By assigning z the color k + 1, a (k + 1)-coloring of
H is obtained, which implies that χ(H) ≤ k + 1 = χ(G) + 1. Hence,
χ(H) − χ(G) ≤ 1.
Next, suppose that χ(H) = k and let c be a k-coloring of H with
the colors 1, 2, . . . , k where z is colored k, say. Thus, neither x nor y is
 173

colored k, say c(x) = i and c(y) = j, where 1 ≤ i, j ≤ k − 1 then. If

i 6= j, then c is a coloring of G and χ(G) ≤ k. If i = j, then assign
y the new color k + 1 in G and so χ(G) ≤ k + 1 = χ(H) + 1. Hence,
χ(G) − χ(H) ≤ 1. Therefore, |χ(G) − χ(H)| ≤ 1.
(b) (i) If G = P2 , then H = P3 . Then χ(H) = χ(G) = 2.
(ii) If G = C3 , then H = C4 . Then χ(H) = 2 and χ(G) = 3.
(iii) If G = C4 , then H = C5 . Then χ(H) = 3 and χ(G) = 2.

10. (a) Proof. Since each coloring that assigns distinct colors to distinct ver-
tices is a balanced coloring, every graph has a balanced coloring.
(b) Figure 14.2 shows a balanced coloring using three colors and so χb (G) ≤
3. The graph G is a (bipartite) tree of order 8 with five vertices in one
partite set and three in the other partite set; so there is no balanced
2-coloring of G. This says that χb (G) = 3.

1 x
...............
s ................
.....
2
.....
.....
1
.....
.....2
........ 3 2 3
G: ............... ............... ............... ..............
t .. ..
.........
.....
.
... .
.........
v w y z
.....
u .................
1

Figure 14.2: A balanced coloring of the graph G in Exercise 10(b)

11. (a) The statement is false. The graph of Figure 14.3 has an edge e not in
any cycle.
............... ...................... ......
....... ..
......
.... ..
.....
..... ........ ........ ............. ..... ...
.....
..... ........ ........ ..... .....
.....
..... ......... ........
........ e .....
.......... ..... .........
.......... ........... ... ..........
..
........... . .................... .......
..
. .............. ........ ..... ..........
....... ..... ........... ........ ..... .....
.....
.
.
. ..
.....
.. ..........
. ... ........
........ ..
......
. .....
... ......... ....... . .
............... ..........
............... .......... ... ..
.. .......
.

Figure 14.3: The graph in Exercise 11(a)

(b) The statement is true.

Proof. Assume, to the contrary, that there exists some edge e = uv of G
such that e is on every odd cycle of G. Then G − e contains no odd cycle
and so χ(G − e) = 2. Hence, G − e is bipartite. Let U and W be the
partite sets of G − e. Necessarily, u, v ∈ U or u, v ∈ W , say the former.
Suppose that |U | = k and |W | = n − k. Summing the degrees of the
vertices in U and W , we obtain kr − 2 = (n − k)r and so (2k − n)r = 2.
This implies that r | 2. Since r ≥ 3, this is impossible.

12. (a) Let k = max{χ(H) : H is a subgraph of G}. Since G is a subgraph of

G, it follows that χ(G) ≤ k. On the other hand, χ(H) ≤ χ(G) for every
subgraph of G and so χ(G) ≥ k. Thus, χ(G) = k.
174 CHAPTER 14. VERTEX COLORINGS

(b) What is max{χ(H) : H is a proper subgraph of G}?

This number is either χ(G) or χ(G) − 1 as C4 and C3 illustrate.
13. Proof. We may assume, without loss of generality, that

χ(G1 ) = max{χ(Gi ) : 1 ≤ i ≤ k}.

Since G1 is a subgraph of G, it follows that χ(G1 ) ≤ χ(G). For each i with

1 ≤ i ≤ k, let ci be a proper coloring of Gi using the colors 1, 2, . . . , χ(G1 ).
Then the coloring c of G defined by c(v) = ci (v) if v ∈ V (Gi ) for 1 ≤ i ≤ k
is a proper coloring using χ(G1 ) colors and so χ(G) ≤ χ(G1 ). Therefore,
χ(G) = χ(G1 ).
14. Proof. We proceed by induction on k. If k = 1, then G itself is the only block
of G and so the result holds. Assume that if H is a nontrivial connected graph
with k blocks H1 , H2 , . . . , Hk , where k ≥ 1, then χ(H) = max{χ(Hi ) : 1 ≤ i ≤
k}. Let G be a nontrivial connected graph with k + 1 blocks G1 , G2 , . . . , Gk+1 .
We may assume, without loss of generality, that Gk+1 is an end-block of G and
let v be the cut-vertex of G belonging to Gk+1 . Then G0 = G − (V (Gk+1 ) −
{v}) is a nontrivial connected graph with the k blocks G1 , G2 , . . . , Gk . By
the induction hypothesis, χ(G0 ) = max{χ(Gi ) : 1 ≤ i ≤ k}. Suppose that
max{χ(Gi ) : 1 ≤ i ≤ k} = χ(G1 ) = a and χ(Gk+1 ) = b. We show that
χ(G) = max{a, b}. Since G0 and Gk+1 are subgraphs of G, it follows that
χ(G) ≥ χ(G0 ) = a and χ(G) ≥ χ(Gk+1 ) = b. Thus, χ(G) ≥ max{a, b}.
Let c0 be a coloring of G0 using the colors 1, 2, . . . , a and let c00 be a coloring
of Gk+1 using the colors 1, 2, . . . , b. Permute the colors if necessary so that
c0 (v) = c00 (v) = 1. Define a coloring c of G by c(v) = 1, c(x) = c0 (x) if
x ∈ V (G0 ) − {v}, and c(x) = c00 (x) if x ∈ V (Gk+1 ) − {v}. Since c is a proper
coloring of G, it follows that χ(G) ≤ max{a, b}. Because a = max{χ(Gi ) :
1 ≤ i ≤ k} and b = χ(Gk+1 ), we have

χ(G) = max{a, b} = max{χ(Gi ) : 1 ≤ i ≤ k + 1},

as desired.
15. Proof. We proceed by induction on k ≥ 2. Let G = G1 ∨ G2 . For i = 1, 2,
suppose that χ(Gi ) = ki and let ci be a proper coloring of Gi using the
colors 1, 2, . . . , ki . Define a coloring c of G by c(v) = c1 (v) if v ∈ V (G1 ) and
c(v) = k1 + c2 (v) if v ∈ V (G2 ). Since c is a proper coloring of G using k1 + k2
colors, it follows that χ(G) ≤ k1 +k2 = χ(G1 )+χ(G2 ). On the other hand, let
c be a proper coloring of G. Then c uses at least χ(Gi ) colors for the vertices
of Gi for each i with i = 1, 2. Furthermore, since each vertex in G1 is adjacent
to every vertex in G2 , no vertex in G1 can be assigned the same color as a
vertex of G1 by c. This implies that c uses at least χ(G1 ) + χ(G2 ) distinct
colors and so χ(G) ≥ χ(G1 ) + χ(G2 ). Thus, the result holds for k = 2.
Assume for graphs F1 , F2 , . . . , Fk and F = F1 ∨ F2 ∨ · · · ∨ Fk that χ(F ) =
Pk 0
i=1 χ(Fi ). Let H = H1 ∨ H2 ∨ · · · ∨ Hk+1 and let H = H1 ∨ H2 ∨ · · · ∨ Hk .
 175

Pk
Then H = H 0 ∨ Hk+1 . By the induction hypothesis, χ(H 0 ) = i=1 χ(Hi ) and
since the result is true for the join of two graphs, χ(H) = χ(H 0 ) + χ(Hk+1 ).
Therefore,
k
X k+1
X
χ(H) = χ(H 0 ) + χ(Hk+1 ) = χ(Hi ) + χ(Hk+1 ) = χ(Hi ),
i=1 i=1

giving the desired result.

16. (a) Proof. Let G be a disconnected graph with α(G) = 2. Since every two
vertices from different components are not adjacent, G contains exactly
two components G1 and G2 . If G1 , say, is not complete, then G1 contains
two nonadjacent vertices u and v. Let w ∈ V (G2 ). Then S = {u, v, w} is
an independent set, contradicting the assumption that α(G) = 2. Hence,
G1 and G2 are complete, one of which has order at least dn/2e.
(b) Proof. Let G be a connected graph with α(G) = 2. Since G contains
pairs of nonadjacent vertices, G contains two vertices u and w such that
dG (u, w) = 2. Hence, uw ∈/ E(G). Let P = (u, v, w) be a u − w geodesic
in G. Suppose, to the contrary, that there is a vertex x in G that is
adjacent to neither u nor w, then {u, w, x} is an independent set, which
is impossible.

17. The statement is false. Let G = K1,1,1,3 . Then χ(G) = 4, n = 6 and α(G) = 3.
n
Thus, α(G) = 2 and n − α(G) + 1 = 4.

18. Let G be a graph of order n for which χ(G) = n/α(G) = n − α(G) + 1. Then

(α(G))2 − (n + 1)α(G) + n = 0

and so (α(G) − 1)(α(G) − n) = 0. Hence, α(G) = 1 or α(G) = n, implying

that G = Kn or G = K n .

19. Since n = 2k 2 + k, α(G) = k and ω(G) = 2k, it follows that

n 2k 2 + k
= = 2k + 1 and n + 1 − α(G) = 2k 2 + 1.
α(G) k

Therefore, 2k + 1 ≤ χ(G) ≤ 2k 2 + 1. For 1 ≤ j ≤ 2k + 1, let

V (Gj ) = {v(j−1)k+i : 1 ≤ i ≤ k}.

The coloring c : V (G) → {1, 2, . . . 2k + 1} of G that assigns v` the color

a (1 ≤ a ≤ 2k + 1) if ` ≡ a (mod(2k + 1)) is a proper coloring and so
χ(G) = 2k + 1.

20. Let G = W5 = C5 ∨ K1 . Then χ(G) = 4. In this case, n = 6, α(G) = 2 and

ω(G) = 3. Thus, n/α(G) = 3 and n − α(G) + 1 = 5.
176 CHAPTER 14. VERTEX COLORINGS

21. For the graph G in Figure 14.4, χ(G) = α(G) = 3. The set S = {u, w, y}
is the unique independent set of three vertices in G. If all vertices of S are
assigned the same color, then each of the remaining vertices of G must receive
a different color. Hence, there is no 3-coloring of G in which some color class
contains three vertices. In fact, any 3-coloring of G using the colors 1, 2, 3 is
either that shown in Figure 14.4 or a permutation of these colors.

u
.............
.................
.... .....
........
. 1 .....
..... .....
.
....... .....
.....
G: .
....
.
.... 2 3 .....
.
.............
z ..............
.
. ... . ........ v
..
.
.
.... ........ .
..
..... .........
..
...
. .....
......
. .....
.....
...
. ..... ....
..
. ....... .
.. .....
..... ..... .... .....
..
......
. ..... 1 .
..
...... .....
.....
.. 3 ...... .....
2
........... . . ................
y ......
.
... ....
...
. .
. w
x

Figure 14.4: A graph G in Exercise 21

22. Yes. Let G be the graph of order 10 with V (G) = U ∪ W , where U =

{u1 , u2 , u3 , u4 , u5 } and W = {w1 , w2 , w3 , w4 , w5 } such that G[U ] = K5 , G[W ] =
K 5 , and such that wi is adjacent to every vertex of U except ui for i =
1, 2, 3, 4, 5. We may assume that ui is colored i. Any 5-coloring of G must
also assign wi the color i. Then each color class has exactly two vertices. Since
W is independent in G and α(G) = |W | = 5, every 5-coloring of G results in
each color class containing α(G) − 3 vertices.

23. Proof. Since G is a k-colorable graph, χ(G) ≤ k. Assume, to the contrary,

that χ(G) ≤ k − 1. Let c be a coloring of G using ` ≤ k − 1 colors and
let V1 , V2 , . . . , V` be the resulting color classes. We may assume, without loss
generality, that |V1 | ≤ |V2 | ≤ · · · ≤ |V` |. Thus, |V` | ≥ n/` and |V1 ∪ V2 ∪
V`−1 | = n − |V` | ≤ n − n` = n(`−1) . If x ∈ V` , then deg x ≤ `−1


·· · ∪  ` ` n≤
k−2
k−1 n < δ(G), which is a contradiction.
If G = K`,`,...,` is a complete (k − 1)-partite graph, then the
 order
 of G is
k−2
n = `(k − 1) and every vertex of G has degree `(k − 2) = k−1 n. Since
χ(G) = k − 1, this bound is sharp.

24. Proof. Let v ∈ V (G) and suppose that u and w are adjacent vertices such
that d(u, v) = d(v, w). Let P be a u − v geodesic and Q a w − v geodesic,
and let x be the first vertex that P and Q have in common. Then d(u, x) =
d(x, w) = d for some positive integer d. Let P 0 be the u − x subpath of P and
Q0 the w − x subpath of Q. Then P 0 , Q0 and uw produce a cycle C of length
2d + 1. Since C is an odd cycle, χ(G) ≥ 3.
For the converse, suppose that χ(G) ≥ 3. Assume, to the contrary, that
there is v ∈ V (G) such that d(u, v) 6= d(v, w) for every two adjacent vertices
u and w of G. Let Vi = {x ∈ V (G) : d(v, x) = i} for 1 ≤ i ≤ k where k is
 177

the greatest distance from v to a vertex of G, that is, e(v) = k. Observe that
each set Vi is independent in G for 1 ≤ i ≤ k. Furthermore, if x ∈ Vi , y ∈ Vj
and |i − j| ≥ 2, then xy ∈
/ E(G). Now, let U be the union of sets Vi where i is
odd and let W be the union of sets Vi where i is even. Then G is a bipartite
graph with partite sets U and W . Thus, χ(G) = 2, which is a contradiction.
25. (a) The number χ(C5 ) is not defined. For G = C5 , any partition of V (G)
into independent sets requires at least one set to consist of a single vertex,
which is not an maximal independent set.
(b) Suppose that χ(G) = k and let {V1 , V2 , . . . , Vk } be a partition of V (G)
into k maximal independent sets. Assigning the color i to each vertex in
Vi for 1 ≤ i ≤ k, we obtain a proper k-coloring of G and so χ(G) ≤ k =
χ(G).
(c) For n, s, t ≥ 1, χ(Kn ) = n and χ(Ks,t ) = 2.
26. (a) Let c be the vertex coloring where c(v) = i if v ∈ Vi for 1 ≤ i ≤ k.
Assume, to the contrary, that there exists some vertex u with c(u) = a
for which the coloring c0 defined by c0 (x) = c(x) if x 6= u and c0 (u) = j
for some j ∈ {1, 2, . . . , k} − {c(u)} is a proper coloring of G. This implies
that Vj ∪ {u} is an independent set of vertices of G, contradicting the
fact that Vj is a maximal independent set of vertices of G.
(b) Suppose that χ(G) = k and let {V1 , V2 , . . . , Vk } be a partition of V (G)
into k maximal independent sets. Therefore, if v ∈ / Vi (1 ≤ i ≤ k), then
v is adjacent to some vertex of Vi . Thus, Vi is P a dominating set of G and
k
so γ(G) ≤ |Vi | for 1 ≤ i ≤ k. Hence, kγ(G) ≤ i=1 |Vi | = |V (G)| and so
γ(G) · χ(G) ≤ n.
(c) Let Gi = Kk for 1 ≤ i ≤ 6 and let G be the graph for which V (G) =
∪6i=1 V (Gi ) and two vertices u and v of G are adjacent if u ∈ V (Ga )
and v ∈ V (Gb ) where |a − b| ≤ 1 and G7 = G1 . Then χ(G) = 2k and
χ(G) = 3k. Thus, χ(G) − χ(G) = k.
27. (a) Proof. Let {V1 , V2 , . . . , Vk } be a partition of V (G) into k independent
sets. Then v belongs to one of these sets, say v ∈ V1 . If V1 = Ui for
some i (1 ≤ i ≤ `), then the proof is complete. Hence, we may assume
that V1 is a proper subset of some set Ui (1 ≤ i ≤ `), say V1 ⊂ U1 and
so U1 − V1 6= ∅. Remove any vertices of U1 − V1 from V2 , V3 , . . . , Vk ,
producing V20 , V30 , . . . , Vk0 where then Vi0 ⊆ Vi for i = 2, 3, . . . , k. Thus
{U1 , V20 , V30 , . . . , Vk0 } is a partition of V (G) into k independent sets, giving
the desired result.
(b) Proof. First suppose that S is a color class in some k-coloring of G.
Then χ(G[V (G) − S]) ≤ k − 1. If χ(G[V (G) − S]) ≤ k − 2, then χ(G) ≤
k − 1, which is a contradiction.
For the converse, suppose that χ(G[V (G) − S]) = k − 1 for some
independent set S of vertices of G. Let c be a (k−1)-coloring of G[V (G)−
S] using the colors 1, 2, . . . , k − 1 and assign each vertex of S the color k.
Then S is a color class in a k-coloring c of G.
178 CHAPTER 14. VERTEX COLORINGS

28. (a) Proof. Suppose that χ(G) = k. Among all k-colorings of G, let c be
one such that the color class U containing v has the maximum number
of vertices. Thus, U is a maximal independent set of G containing v and
so U ∈ {U1 , U2 , . . . , U` }. By Exercise 27(b),
χ(G[V (G) − U ]) = k − 1.
So
min χ(G[V (G) − Ui ]) ≤ χ(G[V (G) − U ]) = k − 1.
1≤i≤`

Therefore,
k = χ(G) ≥ 1 + min χ(G[V (G) − Ui ]).
1≤i≤`

Next, we claim that χ(G[V (G) − Ui ]) ≥ k − 1 for each i with 1 ≤ i ≤ `,

for if χ(G[V (G) − Ui ]) ≤ k − 2, then χ(G) ≤ k − 1, which is impossible.
Therefore,
k = χ(G) = 1 + min χ(G[V (G) − Ui ])
1≤i≤`

as desired.
(b) Since G[V (G) − Ui ] ⊆ G for each i (1 ≤ i ≤ `), it follows that
χ(G[V (G) − Ui ]) ≤ χ(G).
Thus,
χ(G) ≤ 1 + max χ(G[V (G) − Ui ] ≤ 1 + χ(G).
1≤i≤`

The upper bound can be attained, for if G = P4 = (u, v, w, x) and

U = {u, x}, then χ(G[V (G) − U ]) = 2 = χ(G).
29. Proof. Assume, to the contrary, that there is a Hamiltonian-connected
graph G of order n ≥ 3 with χ(G) = 2. Then G is a bipartite graph with par-
tite sets U and W . Since G is not a star, we can assume that 2 ≤ |U | ≤ |W |.
Let u1 , u2 ∈ U . Since every u1 − u2 path P contains fewer than |W | vertices
that belong to W , the path P is not Hamiltonian and so G is not Hamiltonian-
connected.
30. A connected graph G has a 2-colorable line graph if and only if G is a nontrivial
path or an even cycle.
Proof. Assume that G has a 2-colorable line graph. If there exists a vertex
v of G with deg v ≥ 3, then L(G) has a triangle and so χ(L(G)) ≥ 3. Hence,
∆(G) ≤ 2, implying that G is a path or cycle. If G is an odd cycle, then
L(G) = G and χ(L(G)) = 3. Therefore, G is a nontrivial path or an even
cycle. The converse is immediate.
31. A graph G is constructed with vertex set V (G) = {F1 , F2 , . . . , F10 }, where
Fi is adjacent to Fj (i 6= j) if Fi and Fj cannot be placed in the same tank
(see Figure 14.5). Then the minimum number of tanks required to house all
fish is χ(G). Since χ(G) = 3, this number is 3. A 3-coloring of G is given in
Figure 14.5.
 179

F2 F6
...... ...
.......................
.... ..
........... ...... ..........
........... ..........
3 .......... 2 3 ..........
...........
.
.......
...... . ..........
.......... . 1
............. .........
F3 ................................... ...........
..... ..
.......... ..... F7
........... ...
...........
........... ..... F1 F10 .........................
..... ..
. .............................
............. .... ..........
.................... ..........
..........
..... .
..................
............... .
1 2 ..........
..........
..............
.. .....
F4 ............................. ..........
..........
.......... F8
.......... ..
2 ..........
.......... 3
.......... ....... 1
... .........
.......................
3
..... ...
... ...................

F5 F9

Figure 14.5: A graph G in Exercise 31

32. The minimum number of phases is the chromatic number of the graph in
Figure 14.6, which is 3.

L1 ...
... ...
..................
..... .. .. ....
1
. .... .... ..... ........
.
L 7 ..........
.
.... .. .. ....
.... .... .... ........ L
.............. 2
............. ....... .. ... .....
1 ...........
............. ....
.....
3 ......
..
. .........
...
... ..... ... .....
..
.... . . ..
.... .
.
..... .. ..
. .
... ....... ..
.... ... .................... ... ..
..... ... ............. ........ ..
.... .. ... ........... ... ........ .....
... ....
. .
... .
. .......... .... ..
...... .......
L6 .................. ...
...
... ... ............... L3
... .......
........ ... ... ...
........ ... ... ..
3 ..........
... ... ...
... ...
.
2
... .............. .. .
... ........
........ .... ....
.... ........ ......
.
... . . ...........
L .... .....
5 .. .......
L4
3 2

Figure 14.6: The graph in Exercise 32

Section 14.2. Color-Critical Graphs

33. (a) Proof. Since G − v is a subgraph of G, it follows that χ(G − v) ≤ k.
Assume, to the contrary, that χ(G − v) = ` ≤ k − 2. Then there is
an `-coloring c of G − v using the colors 1, 2, . . . , `. However then, the
(` + 1)-coloring c0 of G defined by c0 (x) = c(x) for all x ∈ V (G) − {v}
and c0 (v) = ` + 1 is a proper coloring of G. Thus, χ(G) ≤ ` + 1 ≤ k − 1,
which is a contradiction. Therefore, χ(G − v) = k or χ(G − v) = k − 1.
(b) Proof. Let e = uv. Since G − e is a subgraph of G, it follows that
χ(G − e) ≤ k. Assume, to the contrary, that χ(G − e) = ` ≤ k − 2. Then
there is an `-coloring c of G − e using the colors 1, 2, . . . , `. If c(u) 6= c(v),
then c is also a coloring of G, which is impossible. Thus, we may assume
that c(u) = c(v). However then, the (` + 1)-coloring c0 of G defined by
c0 (x) = c(x) for all x ∈ V (G) − {v} and c0 (v) = ` + 1 is a proper coloring
of G, which is a contradiction.
34. Proof. It suffices to show that χ(G−u) = k −1. By Exercise 33, χ(G−u) =
k − 1 or χ(G − u) = k. Since G − u is a subgraph of G − e, it follows that
180 CHAPTER 14. VERTEX COLORINGS

k − 1 ≤ χ(G − u) ≤ χ(G − e) = k − 1. Thus, χ(G − u) = k − 1.

35. Let G be a 3-critical graph of order n ≥ 3. Then G contains an odd cycle C.
If V (G) − V (C) 6= ∅, then χ(G − v) = 3 for each v ∈ V (G) − V (C); while
if E(G) − E(C) 6= ∅, then χ(G − e) = 3 for each e ∈ E(G) − E(C). Thus,
G = Cn .
36. Claim: The only k-critical graph G such that G − v is (k − 1)-critical for
every vertex v of G is the complete graph Kk .
Proof. First, observe that Kk has the desired property. Assume, to the
contrary, that there is a k-critical graph G other than Kk such that G − v
is (k − 1)-critical for every vertex v of G. Since χ(G) = k and G 6= Kk , the
order of G is at least k + 1. Let there be given a k-coloring of G. Then some
color class S of G contains at least two vertices. Let x, y ∈ S, where then
xy ∈/ E(G). Since G is k-critical, χ(G − x) = k − 1. Since there is a (k − 1)-
coloring of (G − x) − y, the graph G − x is not (k − 1)-critical, producing a
contradiction.
37. The statement is true.
Proof. Suppose that G is a k-critical graph and H is an `-critical graph. By
Theorem 14.7, χ(G ∨ H) = k + `. We show that G ∨ H is (k + `)-critical. If F
is a proper subgraph of G ∨ H, then F is a subgraph of (G ∨ H) − e for some
edge e of G ∨ H. Thus, it suffices to show that χ((G ∨ H) − e) < χ(G ∨ H)
for every edge e of G. We consider two cases.

Case 1. e ∈ E(G) or e ∈ E(H), say the former. Since (G∨H)−e = (G−e)∨H

and χ(G−e) = k−1, it follows that χ((G∨H)−e) = χ(G−e)+χ(H) = k+`−1.

Case 2. e = uv, where u ∈ V (G) and v ∈ V (H). Since χ(G − u) = k − 1 and

χ(H−v) = `−1, there is a (k−1)-coloring of G−u using the colors 1, 2, . . . , k−1
and an (` − 1)-coloring of H − v using the colors k, k + 1, . . . , k + ` − 2. We
then can obtain a (k + ` − 1)-coloring of (G ∨ H) − e by assigning u and v the
color k + ` − 1. Thus, χ((G ∨ H) − e) ≤ k + ` − 1.
38. For an integer k ≥ 3, let G be the 2-connected graph obtained from two copies
F and H of Kk−2 by adding three new vertices u, v, w and joining (1) u to
every vertex in F and H, (2) v to every vertex in F and w and (3) w to every
vertex in H. See Figure 14.7.
Note that the order of G is n = 2k − 1 and α(G) = 2. Thus
n 1
χ(G) ≥ =k−
α(G) 2
and so χ(G) ≥ k. Since there is a k-coloring of G, it follows that χ(G) = k.
Since there is a (k − 1)-coloring of G − x for every vertex x of G, it follows
that G is k-critical.
39. (a) The statement is false. The graph G = (K2 + K1 ) ∨ K1 has this property.
 181

u
............
..................................
........... ...... ... .......... ............
............ ........... ..... ..... ............. .....................
... .. .
. ............... .
...
. . ... ... ....... ...........
..... .. ....... ....
......................................... ............ ... ...
....... ...............................................
............. ............ ... ...
... ....... ..........
...
......... ....... ..... ... .............. .......
......
. . .
.
....... ..... ..
...
...
.........
.. .....
... ... ... ...
. . ...
..... ... ...
. ...
...
... K k−2 ..
....
.
...
...
...
K k−2 ..
...
.
... ... ..... ...
..... ..... .....
...... .. .......... ..
........ ..... ...
.......
............... ........ ............
... .......................................................
..... ....................................... ....
..... ... . ... .. ....
..... ... ..
.
. ... ... .
.....
.
..... . .
..... ... .. ... .. .....
..... .. .. ... .. .....
..... .... .... ... ... .........
..... ... .. ... ... ......
..... ... .. .
... .. ..... .
. ...
..... .. ..
..... ......... . .. ...
...... ..
..... ................
...

v w

Figure 14.7: The graph G in Exercise 38

(b) The statement is false. The graph G = P3 + C9 has this property.

40. (a) Proof. Let G be a maximal k-degenerate graph of order n, where n ≥

k + 1. Thus, δ(G) ≤ k. Suppose that δ(G) 6= k. Then δ(G) = ` < k.
Let u ∈ V (G) such that degG u = `. Since ` < k ≤ n − 1, it follows that
there are vertices v of G that are not adjacent to u. Since G is maximal
k-degenerate, G+uv is not k-degenerate and so δ(G+uv) > k. However,
degG+uv u = ` + 1 ≤ k, producing a contradiction.
l m
n
(b) Let G = Kn . We claim that χPk (G) = k+1 . Divide n by k + 1,
obtaining
n = (k + 1)q + r where 0 ≤ r ≤ k.
n
If n/(k + 1) is an integer, then q = k+1 . So, V (G) can be partitioned
into q subsets V1 , V2 , . . . , Vq such that |Vi | = k + 1 for 1 ≤ i ≤ q. Then
G[Vi ] = Kk+1 , which is k-degenerate. If n/(k l +m 1) is not an integer, then
n
V (G) can be partitioned into q + 1 = k+1 subsets V1 , V2 , . . . , Vq+1
such that |Vi | = k + 1 for 1 ≤ i ≤ q and |Vq+1 | = r. Thus, G[V
l i ] mis
n
k-degenerate for each i (1 ≤ i ≤ q + 1). Therefore, χPk (Kn ) ≤ k+1 .
l m
n
Suppose that χPk (Kn ) = ` < k+1 . Then any partition of V (G) into `
subsets means that some subset V 0 of V (G) has more than k + 1 vertices.
However G[V 0 ] is not k-degenerate, which is a contradiction. Therefore,
 
n
χPk (Kn ) = .
k+1

(c) Proof. Let G be a graph that is `-critical with respect to χPk . Then
χPk (G) = ` and χPk (G−v) = `−1 for every vertex v of G. Let v ∈ V (G)
such that degG v = δ(G). Let H = G − v. Since χPk (H) = ` − 1,
there is a partition of V (H) into ` − 1 subsets V1 , V2 , . . . , V`−1 such that
H[Vi ] is k-degenerate for 1 ≤ i ≤ ` − 1. Assume, to the contrary, that
182 CHAPTER 14. VERTEX COLORINGS

degG v < (k + 1)(` − 1). Then v is adjacent to less than k + 1 vertices

in some set Vi (1 ≤ i ≤ ` − 1), say Vj . However then, G[Vj ∪ {v}] is
k-degenerate, implying that χPk (G) ≤ ` − 1, which is a contradiction.

Section 14.3. Bounds for the Chromatic Number

41. (a) Proof. Let G be a k-chromatic graph. Among the k-colorings of G
using the colors 1, 2, . . . , k, let C1 be the set of those colorings that color
a maximum number of vertices 1. Among the colorings in C1 , let C2 be
the set of those colorings that color a maximum number of vertices 2.
Continuing in this manner, let c be one of the resulting colorings with
color classes S1 , S2 , . . . , Sk , where Si is the set of vertices colored i for
i = 1, 2, . . . , k. Therefore, |S1 | ≥ |S2 | ≥ · · · ≥ |Sk |. Furthermore, each
vertex colored j for some j with 2 ≤ j ≤ k is adjacent to a vertex colored
i for each i with 1 ≤ i ≤ j − 1. Now let s be an ordering of the vertices of
G in which the vertices of S1 are listed first, followed by the vertices of
S2 , then the vertices of S3 , and so on. The greedy coloring determined
by s is precisely the k-coloring c.
(b) Let p be a given positive integer and let H = Kp+2,p+2 with partite sets
U and W , where U = {u1 , u2 , . . . , up+2 } and W = {w1 , w2 , . . . , wp+2 }.
Thus, χ(H) = 2. Let G = H − {ui wi : 1 ≤ i ≤ p + 2}. Consider the
sequence s : u1 , w1 , u2 , w2 , u3 , w3 , . . . , up+2 , wp+2 . The greedy coloring
determined by s results in c(ui ) = c(wi ) = i for i = 1, 2, . . . , p + 2. This
produces a (p + 2)-coloring using p more colors than χ(G).
42. A tree T of order 2k has this property (see the tree T in Figure 14.8 for
k = 4). List the eight end-vertices first and then followed by the four vertices
of degree 2, followed by the two vertices of degree 3 and then the two vertices
of degree 4.
......
..... ..
....... ......
.......
............... .
...........
.
. ........... .
................ ...........
......
... ..... ......
......
...... ..........
...... ....
........... ......
. .. ............. ................ .............
........... ... ..... ...
...........
. ......
.....
. ......
.
...
. ......
........... ......
............ ...... ....
....... ...... ... .............. .............
....... .......
....... .............
.. ...
............... ...
...
...
...... ...
...
...
...
...
.
.............
............ ...
...

Figure 14.8: The tree T in Exercise 42 for k = 4

43. (a) See Figure 14.9.

(b) Yes, because the chromatic number of T is 2. For example, the greedy
coloring with the sequence v1 , v8 , v4 , v6 , v2 , v3 , v5 , v7 uses two colors.
(c) No. Any greedy coloring of T uses at most 1 + ∆(T ) = 1 + 3 = 4 colors.
 183

v3
............
.... 1
v1 v5 v8 3 2
T : ............ ............. ............ ............ ............. ............. 1
.... ... .... .... ... ...

1 2 4 v7 v6 v4
.
1 ...............
v2

Figure 14.9: The tree T in Exercise 43(a)

44. (a) Since G has two vertices of degree 6 and six vertices of degree 5, the
upper bound for χ(G) given in Theorem 14.18 is 1 + 5 = 6.
(b) χ(G) = 5. A 5-coloring of G is shown in Figure 14.10. No 4-coloring of
G exists.
5
........
................
...... .. .. ......
. ...
...... .... ..... ..........
.. . ......
...... ... .... ......
..... ... ......
..... ... ......
...... .. ... ......
... ....... .
... ... ......
.. ...... .. .... ...
. ......
....
.. . .. .
..............................................................................................................................................................................
2 ......... .........
........... ........... .
...
....... . ... .
.
...... ....
.. ......
......................
1
... ....... ........... ........
... ...... .. .
...... ......................
.
... 3 . .
........ ........... ...... ..
.................
..
..
.....
.. ..
... ......... .. ......... .................... ..... ......... ..
... . . .. . . . ...
... .................... ............................... ... ............. ..
... .. ..... .....
. ............... ..
. ....
... .. ....... ....... ... ..
... ... ..... ............. ............. ..... ..... ..
... .. ...
... ... ...
... ... ...
... .. ...
3 ..
... .
............
..
....... ............ 4 ... ... ...
.. .
. .
.
. ....... .
... ... ... ............ ....... ..... .... ....
...... ... .......... ....... .. .....
..... .......
........
..........................................................................................................

1 2

Figure 14.10: The graph G in Exercise 44

45. Let T be the double star containing two vertices of degree 4. Then the upper
bounds for χ(T ) given by Theorems 14.18, 14.19 and 14.20 are 5, 2 and 3,
respectively.
46. (a) Since δ(T ) = 1 for every nontrivial tree T , the upper bound for χ(T ) is
1 + 1 = 2.
(b) By Theorem 10.25, the minimum degree of every outerplanar graph is at
most 2. Therefore, if G is outerplanar, then χ(G) ≤ 3.
47. Since the chromatic number of every 5-regular graph is at least 2 and χ(K5,5 ) =
2, it follows that s = 2. Since χ(G) ≤ 1 + ∆(G) = 6 and χ(K6 ) = 6, it follows
that t = 6. See Figure 14.11 for G3 , G4 and G5 .
48. Claim. If G is a disconnected graph, no component of which is complete or
an odd cycle, then χ(G) ≤ ∆(G).
Proof. Let G1 , G2 , · · · , Gk be the components of G, none of which are com-
plete graphs or odd cycles. Then χ(G) = max{χ(Gi )} ≤ max{∆(Gi )} =
∆(G).
184 CHAPTER 14. VERTEX COLORINGS

2 r
........... ................................. ..........
r 3
................. ............... ......... ...........
........ ........ ......... ....... ...... .. ..
... ...... ........ ............. ...... ............ ...... ....
... ..... ............. ......... . . 1 2 1 2
r3
... ...... ...... .......
... ........ ..... ................................................................ ............. ....
... .....................................................
.... ... ... ..
1r
.............................. ...... ...
r r
....................................................................... r
.......
... ............... ... .. ... ...... .............. ..... .............................
.......r
1 r r r
.

r r2
. .. .
. ...... .................... .... ........... ................. ... ...... ......... .. ... .......... .. . .. .. . .
......... ........... .................................... ... ........ .......... ....... ........ ................................. ....
G3 : r ...
... .............................................
....... .... 1 2
...... .................................
...... ...... ..
............ ... ....
. ... ........ ................ .. ........... ......... ..
...... ............. .................. ........... ......
...
...
.. ......... .............................. . . . G4 : .... ........... ................ ..... ..... .. .. .
.
... ..... ...... ........................................................ ........... ..... ....
. . . .
3 ... ............. .......................... .......................... ............. ...... ....
2
.. ........ ............
...... ... ...
...... .. .. .
. .
.. ...................... ... ... .............. ... ...................... .... ....
r
.
r
.
r
...................... .... ................ ............. .... .................
r
.......... ......... ...... ... ..
r
. . .. . .
............
.
r
............ . . .......... .......
................ ...........
.... .................
3 4 3 4
3 1

1 2
r
..
............
...... ......
r ...............
......... .........
..... ..... ... .... .......
....... .........
. ... ... ......
.. . .
r2 3r .. .. ...
... .... .... .. ............. .... ...
... . . .... ........ ... ...
... ... ............ .......... .. ...
...
.... ..... ............ ..... ........ ...
.. .... .....
G5 :
...
....
...
...... ... ...
.... ... ....
..... .. ...
r3 4 r . .
.
. .. ......
... .. .. ..
... .. .. ...
...
.
...
... ..... ...... ..... ... .. ..
... ..... ............. .. ...... ... ..
.
.

r4 r
. ... .. . ..
...
... ..... ................. 5 .........
........... ... ... ....
... ...
... .. .. .
.... ... ... ... ... ...
..... ... ..
..... ...... ... .... ........
.. .... .....
...... .......
r5
.............
..
1r ... ..... ......
........................

Figure 14.11: The graphs in Exercise 47

49. By Brooks’ theorem (Theorem 14.20), χ(G) ≤ ∆(G) = 3 unless G = K4 .

Since n > 4, G 6= K4 . So it follows that χ(G) ≤ 3. Because G has girth 3, we
have χ(G) = 3.
50. Proof. By the Greedy Coloring Algorithm, the greedy coloring c with re-
spect to the sequence v1 , v2 , . . . , vn has c(v1 ) = 1. Next, c(v2 ) = 1 if v1
and v2 are not adjacent and c(v2 ) = 2 if v1 and v2 are adjacent. Suppose
that c has assigned colors to the vertices in the set S = {v1 , v2 , . . . , vj } where
2 ≤ j ≤ n−2. The vertex vj+1 is assigned the smallest color k that is assigned
to no neightbor of vj+1 in S. Since vj+1 is adjacent to at least one vertex
in {vj+2 , vj+3 , . . . , vn }, it follows that k ≤ ∆(G). Since deg vn < ∆(G), the
color assigned to vn by the greedy coloring c is also at most ∆(G). Hence,
χ(G) ≤ ∆(G).
kn2
51. Proof. Since the size of G is m = 12 (kn)n = 2 , it follows by Theorem 14.22
2
n2
that χ(G) ≥ n2 n−2m = n2 −kn 1
2 = 1−k .

52. By Theorem 14.22, χ(G) ≥ 36/(36 − 26) = 3.6 and so χ(G) ≥ 4.

53. (a) If G is a graph of order n and size m = n2 /4, then by Theorem 14.22,
n2 n2
χ(G) ≥ = 2 = 2.
n2 − 2m n − 2(n2 /4)
This bound is sharp. For example, let G = Kr,r for some positive integer
r. Then n = 2r, m = n2 /4 and χ(G) = 2.
(b) If G is a graph of order n and size m where m > n2 /4, then by The-
orem 14.22, χ(G) > 2 and so χ(G) ≥ 3. This bound is sharp. Let
G = Kr,r + e. Then n = 2r, m = n2 /4 + 1 and χ(G) = 3.
 185

54. If G = Kn1 ,n2 ,...,nk , then χ(G) = k. For n1 = n2 = · · · = nk = r, n = rk and

m = r2 k2 . By Theorem 14.22, χ(G) ≥ k.

55. Let G be a nonempty bipartite graph with partite sets U and W and let D be
the orientation of G obtained by directing each edge uw of G, where u ∈ U
and w ∈ W , from u to w. Then `(D) = 1 and so χ(G) = 2 by Corollary 14.27.
56. (a) Proof. Suppose that χ2 (G) = k and let a 2-tone k-coloring c of G be
given. Let v be a vertex of G with deg v = ∆. Then for each neighbor
x of v, it follows that c(u) ∩ c(v) = ∅. For every two neighbors x and
y of v, we have |c(x) ∩ c(y)| ≤ 1. Therefore, for the k − 2 colors used
in the 2-tone k-coloring c of G that are distinct from the colors in c(v),
it follows thatl k−2

2
2 ≥ m and so k − 5k + (6 − 2∆) ≥ 0. Therefore,
∆
√
5+ 8∆+1
χ2 (G) = k ≥ 2 .
(b) Since Cn is 2-regular for each n ≥ 3, it follows by (a) that χ2 (Cn ) ≥ 5.
The 2-tone 6-coloring of C7 and the 2-tone 5-coloring of C8 shown in
Figure 14.12 give 5 ≤ χ2 (C7 ) ≤ 6 and χ2 (C8 ) = 5. (We write ab for
{a, b}.)

12
12 t
t .......
..........
...................................
.......
.........................................
34 t
......
.....
.......
...... ......
..... ........
...... .....
..... t 45
56 t ...
.
...
... .....
...
... t 34
... ..
....
.
... ...
...
...
...
. .
... ... .... ...
...
.... ... ...
t 13
...
15 t
... ... ...
... ..
.... ... .
... ... ... ...
34 t t15
... ..
. ... ...
... ... ... ...
... ...
... ... ... ...
... ... ... ...
... ...
.....
.....
t
....
. t
...
..... . t
...
t ...... .
.... ..... .
..
...... .....
........
....................................... 23 ......
...... 25
t
.......
.......... .......
.................................
25 36
14

Figure 14.12: A 2-tone 6-coloring of C7 and a 2-tone 5-coloring of C8

We claim that χ2 (C7 ) = 6. Assume, to the contrary, that there is a

2-tone 5-coloring c of C7 . Since no color can be used for more than three
vertices, four of the five colors 1, 2, 3, 4, 5 must be used for three vertices
each, say 1, 2, 3 and 4. Then the color 5 is used for exactly two vertices.
Hence, the colors assigned to five vertices of C7 are 2-element subsets of
{1, 2, 3, 4}. This implies that there is a path P = (v1 , v2 , v3 ) on C7 for
which 5 ∈ / c(v1 ) ∪ c(v2 ) ∪ c(v3 ). We may assume that c(v1 ) = 12 and
c(v2 ) = 34. Since 3, 4 ∈ / c(v3 ) and c(v3 ) 6= 12, we have a contradiction.
Hence, as claimed, χ2 (C7 ) = 6.
186 CHAPTER 14. VERTEX COLORINGS
Chapter 15

Perfect Graphs and List

Colorings

Section 15.1. Perfect Graphs

1. (a) G = C5 . (b) G is the Grötzsch graph.
2. Let G1 be a triangle-free k-chromatic graph (whose existence is guaranteed
by Theorem 15.1). Then G = G1 + K` has the desired properties.
3. The statement is true.
Proof. By Theorem 15.1, for each integer k ≥ 3, there is a triangle-free k-
chromatic graph G. Let H be a k-critical subgraph of G. Since G is triangle-
free, H has the desired property.
4. By Theorem 15.1, there is a triangle-free 4-chromatic graph G. If G is regular,
then this is a desired example. Otherwise, let s = δ(G) < ∆(G) = r. Let
G0 = G. Let G1 be constructed from two copies of G0 where the vertex v
in one copy is joined to v in the other copy if degG v < r. The graph G1
is also 4-chromatic, triangle-free and δ(G1 ) = s + 1. If G1 is r-regular, then
this is a desired example; if not, let G2 be constructed from two copies of G1 ,
where the vertex v in one copy is joined to v in the other copy if degG1 v < r.
Continuing in this manner, we arrive at Gr−s , which is r-regular, triangle-free
and 4-chromatic.
5. Consider the graph G of the octahedron shown in Figure 15.1(a). We need to
show that χ(H) ≤ ω(H) for every induced subgraph H of G (since χ(H) ≥
ω(H) for every graph H). Clearly, ω(G) = 3 and a 3-coloring of G is indicated
in Figure 15.1(a). It remains to show that χ(H) ≤ ω(H) for every proper
induced subgraph H of G. Without loss of generality, we may assume that
H is connected and H is not complete. Then H is one of the three graphs
H1 , H2 and H3 in Figure 15.1(b). We see that χ(H1 ) = ω(H1 ) = 2, χ(H2 ) =
ω(H2 ) = 3 and χ(H3 ) = ω(H3 ) = 3.

187
188 CHAPTER 15. PERFECT GRAPHS AND LIST COLORINGS

1
...
... ....
..................
..... .. ... ....
.
..
...... .... .... .........
.... ... .... ........
..... ... .....
..... ... .....
..... ...
... .....
.
......
. .. ...
... .....
G: ......
.
.... .
... .
....
.....
.....
..
.. .
...
.
.. 3 .
..
..
......
... ...
...
... ..
.........
. .
2 .....
.....
.....
..
....
. . ..
... .... .... . ... ........... .....
..
....
. ..
.......
. ...
... . .
. .......
.........
.....
.....
..
....
. .
........ . .. . .... .....
... ..... ... .. .......
.
.
..
..
... .......... .
...
. .
.
..
......... .......... .........
.. . ....... .....
.
...... .......... . .. ..... . .
... .. . ....................................
. ....... ....
. .. ....... ................. ...... .....
..... ....... .................. .................
........................................
2 ...
........... ....................
............................... 1 ............ .
...... 3
(a)
1
........
1 2...
............. .. ... ............. ... ...
... 1 .....
..................
.....
......
.....
.........
.....
2 .... ..... .....
.....
.... .....
..... 3
........
.....3
..................
....

H1 : ............. .............. ... .. .... ..

... 2 H2 : .....
..... ........
.....
.
H1 : ..................
.....
. .....
..... .. ..
..... ...... ...... .....
.....
..... ...... .... ..........
............. ............. ..............
... 1 .... ....
....
. .......

1 2 1
(b)

Figure 15.1: Graphs in Exercise 5

[Also, observe that G is chordal and, consequently, is perfect.]

6. Proof. Since χ(G) = n − 1, it follows that ω(G) ≤ n − 1. Suppose that

ω(G) ≤ n − 2. Let H be a maximum clique of G and let u, v ∈ V (G) − V (H).
Thus, each of u and v is not adjacent to at least one vertex in H, say u is not
adjacent to u0 in H and v is not adjacent to v 0 in H. If u0 6= v 0 , then assign
the color 1 to u and u0 , the color 2 to v and v 0 , and the colors 3, 4, . . . , n − 2
to the remaining n − 4 vertices of G, producing an (n − 2)-coloring of G,
which is impossible. Hence, we may assume that u0 = v 0 , where u0 is the only
vertex of H not adjacent to u and v 0 is the only vertex of H not adjacent
to v. If u and v are not adjacent, then assign the color 1 to u, v and u0
and the colors 2, 3, . . . , n − 2 to the remaining n − 3 vertices of G, producing
an (n − 2)-coloring of G, which is impossible. If u and v are adjacent, then
each of u and v is adjacent to each vertex in V (H) − {u0 }. This implies that
(V (H) − {u0 }) ∪ {u, v} induces a clique of order n − 1 in G, a contradiction.

7. (a) The statement is false. The graph G = C5 has order n = 5, χ(G) = 3 =

n − 2 and ω(G) = 2 = n − 3.
(b) The statement is true. Proof. Let G be a graph of order n ≥ 6 with
χ(G) = n − 2. We claim that ω(G) = n − 2. Assume, to the contrary,
that ω(G) = k ≤ n − 3. Then G contains a maximum clique H = Kk .
Let u, v, w ∈ V (G) − V (H). Then u is not adjacent to some vertex u0
of H, v is not adjacent to some vertex v 0 of H and w is not adjacent to
some vertex w0 of H. If the vertices u0 , v 0 and w0 are distinct, then we
can assign the color 1 to u and u0 , the color 2 to v and v 0 , the color 3 to
w and w0 and the colors 4, 5, . . . , n − 3 to the remaining n − 6 vertices
 189

of G. Then χ(G) ≤ n − 3, producing a contradiction. Hence, we may

assume that no such distinct vertices u0 , v 0 and w0 exist.
Next, suppose that u and v are not adjacent to u0 and w is not adjacent
to w0 , where u0 6= w0 . If u and v are not adjacent, then the color 1 can
be assigned to u, v and u0 , the color 2 can be assigned to w and w0 ,
and the colors 3, 4, . . . , n − 3 can be assigned to the remaining n − 5
vertices of G. Then χ(G) ≤ n − 3, producing a contradiction. Hence,
we may assume that uv ∈ E(G). If u and v are adjacent to w0 , then
G[V (H) ∪ {u, v} − {u0 }] = Kk+1 , producing a contradiction. Therefore,
at least one of u and v is not adjacent to w0 . If vw0 ∈ / E(G), then we
can proceed as above unless vw ∈ E(G), which we assume to be the
case. If uw ∈ E(G), then G[V (H) ∪ {u, v, w} − {u0 , w0 }] = Kk+1 , which
is impossible. Hence, we may assume that u0 u, u0 v, w0 v, w0 w, uw ∈ / E(G)
and u0 w, w0 u, uv, vw ∈ E(G). If k < n − 3, then there is another vertex
x ∈ V (G) − V (H) and we may assume that u, v and x play the same
role as u, v and w. Thus, either G[V (H) ∪ {u, v, x} − {u0 , w0 }] = Kk+1
or G[V (H) ∪ {v, w, x} − {u0 , w0 }] = Kk+1 , which is impossible. Hence,
we may assume that k = n − 3. However then, assigning the color 1
to u and u0 , the color 2 to w and w0 , the color 3 to v and some other
vertex of H and the colors 4, 5, . . . , n − 3 to the remaining n − 6 vertices
of G, produces an (n − 3)-coloring of G. Then χ(G) ≤ n − 3, producing
a contradiction.
Finally, we may assume that u, v and w are all not adjacent to exactly
one vertex of H, say u0 . If none of u, v and w are adjacent, then we
assign the color 1 to u, v, w and u’ and the colors 2, 3, 4, . . . , n − 3 to the
remaining n − 4 vertices of G and so χ(G) ≤ n − 3, which is impossible.
Hence, we may assume that uv ∈ E(G). However then, G[V (H)∪{u, v}−
{u0 }] = Kk+1 , producing a contradiction.
8. Proof. Let G be a k-critical graph for some integer k ≥ 2 having the prop-
erty that every proper induced subgraph of G is perfect. If G is complete, then
G itself is perfect. It remains to show that if G is perfect, then G is complete.
Suppose that G is perfect but is not complete. Since G is not complete, G
contains two nonadjacent vertices u and v. Since G is critically k-chromatic,
χ(G − v) = k − 1. Because G − v is perfect, ω(G − v) = k − 1. Since G is
perfect and χ(G) = k, it follows that ω(G) = k. It therefore follows that v
belongs to every k-clique of G. Similarly, u belongs to every k-clique of G.
Since u and v are not adjacent, this is a contradiction.
9. (a) Let V (G) = {v1 , v2 , . . . , vn } and E(G) = {e1 , e2 , . . . , em }. Let A de-
note the power set of {1, 2, . . . , m}. Define an intersection graph H by
V (H) = {S1 , S2 , . . ., Sn } such that Si consists of the subscripts of the
edges of G that are incident with vi in G. Then H ∼ = G.
(b) Let V (G) = {v1 , v2 , . . . , vn } and let A = {1, 2, . . . , n}. Let F be the set
of 2-element subsets of A such that {i, j} ∈ F if and only if vi vj ∈ E(G).
Then L(G) is the intersection graph of F.
190 CHAPTER 15. PERFECT GRAPHS AND LIST COLORINGS

10. Let A = {1, 2, 3} and let S1 = {1, 2}, S2 = {1, 3}, S3 = {2, 3} and S4 = {3}.
Then the intersection graph of S = {S1 , S2 , S3 , S4 } is isomorphic to K4 − e.
Thus, k ≤ 3. Since there are only three nonempty subsets of {1, 2} and the
order of K4 − e is 4, it follows that k = 3.

11. Proof. Suppose that G is an interval graph with V (G) = {v1 , v2 , . . . , vn }.

Then there exists a collection S of n closed intervals of real numbers, say
S = {Ii : 1 ≤ i ≤ n}, such that vi and vj are adjacent if Ii ∩ Ij 6= ∅. Let H
be an induced subgraph of G with V (H) = {vi1 , vi2 , . . . , vik } ⊆ V (G), where
1 ≤ k ≤ n, and let
S 0 = {Iij : 1 ≤ j ≤ k} ⊆ S.

Let vir , vis ∈ V (H), where 1 ≤ r, s ≤ k. Since H is an induced subgraph of

G, it follows that vir and vis are adjacent in H if and only if vir and vis are
adjacent in G. Thus, vir and vis are adjacent in H if and only if Iir ∩ Iis 6= ∅.
Therefore, H is also an interval graph.

12. There are five connected graphs of order 4 that are interval graphs. Four of
these are shown in Figure 15.2.
............. v1
... v2 v3 v2 v3
.............. ......... ............. .......... v1 v2
... .... ......... .........
.............. ..... ..... ......
..... ..... ................ .............
v2 ..... ..... ..... ....
. ..... ....... ..... ..
.. . .....
.....
...
..............
....... .................... .....
v1 ........
v1 .....
.....
............ .....
.... v3 .....
.....
.....
...
............. ............. ........
... ............. ...........
............. v4 ... .
... v4
v4 v4 v3

Figure 15.2: Graphs in Exercise 12

For P4 , let I1 = [1, 2], I2 = [2, 3], I3 = [3, 4], I4 = [4, 5].
For K1,3 , let I1 = [2, 5], I2 = [1, 2], I3 = [3, 4], I4 = [5, 6].
For (K2 + K1 ) ∨ K1 , let I1 = [2, 4], I2 = [1, 2], I3 = [1, 3], I4 = [4, 5].
For K4 − e, let I1 = [1, 3], I2 = [1, 2], I3 = [2, 4], I4 = [3, 4].
For K4 , let Ii = [1, 2] for i = 1, 2, 3, 4. For distinct intervals, let I1 = [1, 8],
I2 = [2, 7], I3 = [3, 6], I4 = [4, 5].
The remaining connected graph of order 4 is C4 = (v1 , v2 , v3 , v4 , v1 ). We
show that C4 is not an interval graph. Assume, to the contrary, that C4 is
an interval graph and let I1 , I2 , I3 , I4 be the corresponding intervals. Since
I1 ∩ I3 = ∅, we may assume that I1 = [a1 , b1 ] and I3 = [a3 , b3 ], where b1 < a3 .
Then I2 = [a2 , b2 ] and I4 = [a4 , b4 ], where (1) a2 ≤ b1 and a3 ≤ b2 and (2)
a4 ≤ b1 and a3 ≤ b4 . However then, I2 ∩ I4 6= ∅, a contradiction.

13. (a) The interval graph G is the path (J1 , I1 , J2 , I2 , J3 , I3 , J4 , I4 , J5 ) of order

9. Thus, the size of G is 8.
 191

(b) The interval graph G is a tree obtained from the path (J1 , I1 , J2 , I2 , J4 ,
I3 , J5 ) of order 7 by adding the vertex J3 and the edge I2 J3 . The size
of G is 7.
(c) The interval graph G is the forest 2P5 , one of whose components is
(J1 , I1 , J2 , I2 , J3 ) and the other is (J4 , I3 , J5 , I4 , J6 ). The size of G is 8.
14. (a) Since the length of each interval Jj is less than the length of each interval
Ii , each interval Jj has a nonempty intersection with at least one but at
most two intervals Ii . Suppose that the graph G is not acyclic. Then
G contains one or more cycles. Let C be a cycle in G and let r be the
smallest integer such that Jr is a vertex in C and s the smallest integer
such that Is is a vertex in C. Necessarily, degG (Jr ) = 2. Suppose that
Is and Is+1 are adjacent to Jr . However then, no vertex Jt where t > r
can be adjacent to Is , which is a contradiction.
(b) By (a), the graph G is a forest. Thus, the size of G is a + b − k(G). It
remains only to determine the number k(G) of components in G. Let
gcd(a, b) = d. Thus, a = a0 d and b = b0 d, where a0 and b0 are relatively
prime. Thus, for each j with 1 ≤ j ≤ b,
 
Jj = (j−1)a
b0
0 ja0
, b0 .

The number of components in G is then equal to the number of integers

j (1 ≤ j ≤ b) such that b0 | ja0 . Since a0 and b0 are relatively prime, the
number of components in G equals the number of integers j (1 ≤ j ≤ b)
such that b0 | j. Since b = b0 d, it follows that k(G) = d, that is, the size
of G is a + b − gcd(a, b).
15. (a) Let ui correspond to Ii for 1 ≤ i ≤ 5. Then G is the graph with V (G) =
{u1 , u2 , . . . , u5 } and E(G) = {u1 u2 , u1 u3 , u2 u3 , u2 u4 , u3 u4 , u4 u5 }.
(b) Let J1 = (0.9, 7.1), J2 = (5.9, 8.1), J3 = (5.9, 9.1), J4 = (6.9, 20.1) and
J5 = (19.9, 22.1). By letting ui correspond to Ji for (1 ≤ i ≤ 5), we
arrive at the same graph G in (a).
16. The graph Kk−2 ∨ C4 has chromatic number k. Since G contains a 4-cycle
with no chords, G is not chordal.
17. Proof. Assume, to the contrary, that there is an interval graph G that is
not a chordal graph. Then G contains a cycle C = (v1 , v2 , . . . , vk , v1 ), k ≥ 4,
with no chord. Since G is an interval graph, corresponding to each vertex vi
(1 ≤ i ≤ k) of C is an interval Ii = [ai , bi ]. Since v2 is adjacent to both v1
and v3 but v1 v3 ∈
/ E(G) and v1 is adjacent to vk but v2 vk ∈ / E(G), we may
assume that a1 < a2 ≤ b1 and b1 < a3 ≤ b2 . Continuing in this manner, we
have a1 < a2 < · · · < ak < a1 , which is impossible.
18. The graph G in Figure 15.3 is chordal since its only cycle is a triangle.
Assume, to the contrary, that G is an interval graph. Corresponding to
each vertex vi (1 ≤ i ≤ 6) of G is an interval Ii = [ai , bi ]. The intervals I4 , I5
192 CHAPTER 15. PERFECT GRAPHS AND LIST COLORINGS

v4 ......................... ............
......... v5
..... .....
.....
..... ..
......
.
.
......... .....
.... ... .............
v 1 ............ ..
.
........
.... v 2
... ..
... .....
... .......
............
.......
v3

............. v6
...

Figure 15.3: The graph in Exercise 18

and I6 are mutually disjoint, so we may assume that a4 < b4 < a5 < b5 <
a6 < b6 . Since v3 v6 ∈ E(G) but v3 v5 ∈ / E(G), it follows that b5 < a3 . Since
v1 v4 ∈ E(G) but v1 v5 ∈/ E(G), it follows that b1 < a5 . Hence, I1 ∩ I3 = ∅ and
so v1 v3 ∈
/ E(G), which is a contradiction.

Section 15.2. The Perfect and Strong Perfect Graph The-

orems
19. For n = 7, let G7 = C 7 . Note that ω(G7 ) = 3 and χ(G7 ) = 4. For n ≥ 8, let
Gn = C 7 + (n − 7)K1 .
20. Let G = C2k+1 = (v1 , v2 , . . . , v2k+1 , v1 ). Observe that ω(G) = k and the
subgraph of G induced by the set S = {v2i : 1 ≤ i ≤ k} is a complete graph
of order k.
Next, we show that χ(G) = k + 1. Since χ(G) ≥ ω(G) = k, we first show
that χ(G) 6= k. Assume, to the contrary, that χ(G) = k. Then there is a k-
coloring c of G. Since G[S] = Kk , we may assume that c(v2i ) = i for 1 ≤ i ≤ k.
Since v1 is adjacent to v2i for 2 ≤ i ≤ k, it follows that c(v1 ) = 1. Similarly,
c(v2k+1 ) = k. This, however, implies that c(v2j−1 ) = j for 2 ≤ j ≤ k. Since
v2k−1 and v2k+1 are adjacent in G and c(v2k−1 ) = c(v2k+1 ) = k, this is a
contradiction. Thus, χ(G) ≥ k + 1. On the other hand, the coloring c0 defined
by c0 (v2i−1 ) = c0 (v2i ) = i for 1 ≤ i ≤ k and c0 (v2k+1 ) = k + 1 is a proper
(k + 1)-coloring of G. Thus, χ(G) ≤ k + 1 and so χ(G) = k + 1.
21. Since C8 is a bipartite graph, C8 is perfect. By the Perfect Graph Theorem
(Theorem 15.10), C 8 is also perfect.
22. (a) The statement is true. Since the order of each component of G is 25 and
χ(G) = max{χ(Gi ) : 1 ≤ i ≤ k}, it follows that χ(G) ≤ 25. Since one
component of G is K25 , χ(G) = 25. Also, ω(G) = max{ω(Gi ) : 1 ≤ i ≤
k} = ω(K25 ) = 25.
(b) The statement is false. One component of G is C25 . Since this is an
induced subgraph of G and χ(C25 ) = 3 and ω(C25 ) = 2 , it follows that
G is not perfect.
23. The converse is true, that is, if a replication graph Rv (G) of G is perfect, then
G is perfect.
 193

Proof. Suppose that G is not perfect. Then by the Strong Perfect Graph
Theorem (Theorem 15.12), G contains an induced odd cycle C of length 5 or
more. However, C is also an induced cycle of Rv (G) and so by the Strong
Perfect Graph Theorem, Rv (G) is not perfect.

24. (a) Let G be a bipartite graph with partite sets U and W . Orient each edge
uw of G, where u ∈ U and w ∈ W , from u to w, producing a transitive
orientation of G and so G is a comparability graph.
(b) There is a transitive tournament of order n for each n ≥ 2.
(c) Since every induced subgraph of a comparability graph is a comparability
graph, it suffices to show that if G is a comparability graph, then χ(G) =
ω(G). Furthermore, since χ(G) ≥ ω(G), we need only to show that
χ(G) ≤ ω(G) for every comparability graph G. By Theorem 14.26,
χ(G) ≤ 1 + `(D). However, since D is transitive, every path in D of
length k corresponds to a clique with k + 1 vertices in G. Therefore,
χ(G) ≤ 1 + `(D) ≤ ω(G).
(d) By (a) and (b), every bipartite graph and every complete graph is a
comparability graph. By (c), each such graph is perfect.
(e) Let G be the graph obtained from the 3-cycle (u, v, w, u) by adding a
pendant edge xw and letting D be the orientation of G whose arcs are
(u, v), (v, w), (u, w) and (x, w).

25. We claim that χ(Shad(G)) = χ(G) and ω(Shad(G)) = ω(G). First, we show
that χ(Shad(G)) = χ(G). Let χ(G) = k. Since G ⊆ Shad(G), it follows that
χ(Shad(G)) ≥ k. Let c be a k-coloring of G. The coloring c0 of Shad(G)
defined by c0 (vi ) = c(vi ) and c0 (ui ) = c(vi ) for 1 ≤ i ≤ n is a k-coloring of
Shad(G). Thus, χ(Shad(G)) ≤ k. Hence, χ(Shad(G)) = χ(G).
Next we show that ω(Shad(G)) = ω(G). Suppose that ω(G) = `. Since
K` ⊆ Shad(G), it follows that ω(Shad(G)) ≥ `. Assume, to the contrary, that
ω(Shad(G)) > `. Since the subgraph of Shad(G) induced by {u1 , u2 , . . . , un }
is isomorphic to G, there is a maximum clique H = Kt with t > ` such
that V (H) = V1 ∪ U1 , where V1 ⊆ {v1 , v2 , . . . , vn } and U1 ⊆ {u1 , u2 , . . . , un },
and |V1 | is maximum. Suppose that V1 = {vi1 , vi2 , . . . , via } and uib ∈ U1 .
Then uib is adjacent to vij for 1 ≤ j ≤ a. In particular, vib ∈ / V1 . However,
vib is adjacent to the vertices of V1 ∪ U1 − {uib }. Thus, the subgraph of
Shad(G) induced by (V1 ∪ {vib }) ∪ (U1 − {uib }) is also a maximum clique,
which contradicts the defining property of H.

26. (a) Assume that G is perfect. Let H be an induced subgraph of G. Then

ω(H) = χ(H). Consequently, V (H) can be partitioned into at most
χ(H) = ω(H) color classes, each with at most α(H) vertices. Thus,
|V (H)| ≤ α(H)χ(H) = α(H)ω(H).
(b) This follows directly from the observation that α(H) = ω(H) and α(H) =
ω(H) for all graphs H.
194 CHAPTER 15. PERFECT GRAPHS AND LIST COLORINGS

Section 15.3. List Colorings

27. Proof. Let there be given any collection L = {L(v) : v ∈ V (G)} of color lists
of size 1 + ∆(G) for the vertices of G. We may assume all colors are positive
integers. Suppose that the vertices of G are listed in the order v1 , v2 , . . . , vn
and the greedy coloring algorithm is applied. Then v1 is assigned the smallest
color in L(v1 ) and for 2 ≤ i < n, the vertex vi is assigned the smallest
available color in L(vi ) that is not used to color the neighbors of vi in the
set S = {v1 , v2 , . . . , vi−1 }. Since at most deg vi neighbors of vi belong to S,
at most deg vi colors in L(vi ) are used to color the neighbors of vi . Because
|L(vi )| = 1 + ∆(G) ≥ 1 + deg vi , there is at least one color available in L(vi )
for vi . Thus, G is L-choosable and so χ` (G) ≤ 1 + ∆(G).

28. Proof. Let C = (v1 , v2 , . . . , vn , v1 ) be an odd cycle and suppose that L(vi )
is a list of three colors for each i = 1, 2, . . . , n. Consider the path P =
(v1 , v2 , . . . , vn−1 ). By Theorem 15.14, every tree is 2-choosable and so P
is certainly 3-choosable. Hence, there exists a 3-list-coloring c of P where
c(vi ) ∈ L(vi ) for i = 1, 2, . . . , n − 1. Since |L(vn )| = 3, there exists a color
a ∈ L(vn ) such that c(v1 ) 6= a and c(vn−1 ) 6= a. Defining c(vn ) = a gives a
3-list-coloring c of C in which c(vi ) ∈ L(vi ) for i = 1, 2, . . . , n. Thus, C is
3-choosable.

29. Proof. Clearly, χ` (K2,3 ) ≥ 2. Let G = K2,3 with partite sets {u1 , u2 }
and {v1 , v2 , v3 } and let there be given an arbitrary collection L = {L(v) :
v ∈ V (G)} of color lists of size 2 for the vertices of G. First assume that
L(u1 ) ∩ L(u2 ) 6= ∅. Let a ∈ L(u1 ) ∩ L(u2 ). Assign a to u1 and u2 and assign
a color in L(vi ) − {a} to vi (1 ≤ i ≤ 3), producing a list coloring of G.
Next assume that L(u1 ) ∩ L(u2 ) = ∅, say L(u1 ) = {a, b} and L(u1 ) = {c, d}.
Among the four sets {a, c}, {a, d}, {b, c}, {b, d}, there is at least one set that
is not L(vi ) for a vertex vi (1 ≤ i ≤ 3), say this set is {a, c}. Assign a to u1 ,
c to u2 , and assign a color in L(vi ) − {a, c} to vi , producing a list coloring of
G. Thus, G is L-choosable and so χ` (G) = 2.

30. Proof. Let G = K2,4 with partite sets {u1 , u2 } and {v1 , v2 , v3 , v4 }. First
consider the collection L0 = {L(v) : v ∈ V (G)} of color lists of size 2 for
the vertices of G, where L(u1 ) = {a, b}, L(u1 ) = {c, d}, L(v1 ) = {a, c},
L(v2 ) = {a, d}, L(v3 ) = {b, c} and L(v4 ) = {b, d}. Since G is not L0 -choosable,
it follows that χ` (G) ≥ 3.
To show that χ` (G) ≤ 3, let there be given an arbitrary collection L =
{L(v) : v ∈ V (G)} of color lists of size 3 for the vertices of G. First assume
that L(u1 )∩L(u2 ) 6= ∅, say a ∈ L(u1 )∩L(u2 ). Assign a to u1 and u2 and assign
a color in L(vi ) − {a} to vi (1 ≤ i ≤ 4), producing a list coloring of G. Next
assume that L(u1 ) ∩ L(u2 ) = ∅, say L(u1 ) = {a, b, c} and L(u1 ) = {d, e, f }.
Assign the color a to u1 , the color d to u2 , and assign a color in L(vi ) − {a, d}
to vi (1 ≤ i ≤ 4), producing a list coloring of G. Thus, G is L-choosable and
so χ` (G) = 3.
 195

31. Proof. We show that the graph K3,27 is not 3-choosable. Let U and W be
the partite sets of K3,27 , where U = {u1 , u2 , u3 } and |W | = 27. Let L(u1 ) =
{1, 2, 3}, L(u2 ) = {4, 5, 6}, and L(u3 ) = {7, 8, 9}. Assign to the 27 vertices of
W distinct 3-element sets S such that |S ∩ L(ui )| = 1 for i = 1, 2, 3. From this
set of lists for K3,27 and for each function c : U ∪ W → {1, 2, . . . , 9}, there is a
vertex w ∈ W such that c(w) ∈ {c(u1 ), c(u2 ), c(u3 )} and c(w) ∈ L(w). Since
this is not a list coloring, K3,27 is not 3-choosable.

32. Proof. Let the partite sets of K10,10 be U and W . Suppose
that the 10
vertices of U and the 10 vertices of W are assigned the 53 = 10 distinct
3-element subsets of {1, 2, 3, 4, 5}. Suppose that some vertex of U is assigned
the color 1. Then at most 6 vertices of U are colored 1. Then no vertex of W
can be colored 1. Suppose that some vertex of W is assigned the color 2. Then
we may assume that some vertex of U is colored 3. At most three vertices not
colored 1 can be colored 3. The vertices w0 and w00 with L(w0 ) = {1, 3, 4} and
L(w00 ) = {1, 3, 5} must then be colored 4 and 5, respectively. However then,
the vertex u0 ∈ U with L(u0 ) = {2, 4, 5} has no choice for a color. Therefore,
χ` (K10,10 ) > 3.

33. Proof. First we show that Pn  K2 is not 2-choosable for n ≥ 4. For n ≥ 4,

let Pn  K2 be the graph obtained from the paths P = (v1 , v2 , . . . , vn ) and
P 0 = (v10 , v20 , . . . , vn0 ) together with the edges vi vi0 for i = 1, 2, . . . , n. Define

L(v1 ) = L(v4 ) = {2, 3}, L(v10 ) = L(v40 ) = {1, 3},

L(v2 ) = L(v3 ) = L(v20 ) = L(v30 ) = {1, 2}

and if n > 4, let L(vi ) and L(vi0 ) be any 2-element sets for 5 ≤ i ≤ n. We may
assume that c(v2 ) = 1 and c(v3 ) = 2 in the resulting list coloring c. Then
c(v20 ) = 2 and c(v30 ) = 1. Thus, c(v4 ) = 3. However then, c(v40 ) ∈/ L(v40 ) and
so Pn  K2 is not 2-choosable.
We show by induction that Pn  K2 is 3-choosable for n ≥ 1. This is
obvious if n = 1. Assume that Pk  K2 is 3-choosable for a positive integer
k and consider G = Pk+1  K2 . Let G be constructed from the two paths
(v1 , v2 , . . . , vk+1 ) and (v10 , v20 , . . . , vk+1
0
) where vi vi0 ∈ E(G) for 1 ≤ i ≤ k + 1.
Let there be given 3-element lists L(vi ) and L(vi0 ) for 1 ≤ i ≤ k + 1. Let
0
H = G−{vk+1 , vk+1 }. By the induction hypothesis, there is a proper coloring
c of H where c(vi ) ∈ L(vi ) and c(vi0 ) ∈ L(vi0 ) for 1 ≤ i ≤ k. Define c(vk+1 )
0
to be any element of L(vk+1 ) for which c(vk+1 ) 6= c(vk ) and define c(vk+1 ) to
0 0 0
be any element of L(vk+1 ) for which c(vk+1 ) ∈ / {c(vk+1 ), c(vk )}. Thus, G is
3-choosable.

34. Yes. Let U and W be partite sets of K3,3 , where U = {u1 , u2 , u3 } and W =
{w1 , w2 , w3 }. We may assume that L(u1 ) = {1, 2, 3}, |L(ui )| = 2 for i = 2, 3,
and |L(wi )| = 2 for i = 1, 2, 3. If ∩3i=1 L(ui ) 6= ∅ or ∩3i=1 L(wi ) 6= ∅, then there
is a list coloring for these lists. Then ∩3i=1 L(ui ) = ∅ and ∩3i=1 L(wi ) = ∅. We
consider cases.
196 CHAPTER 15. PERFECT GRAPHS AND LIST COLORINGS

Case 1. Two of the sets L(wi ), i = 1, 2, 3, are equal, say L(w1 ) = L(w2 ) =
{a, b}. Then L(w3 ) = {c, d}, where {a, b} ∩ {c, d} = ∅. Then we may define
c(w1 ) = c(w2 ) so that a list coloring results.

Case 2. No two of the sets L(wi ), i = 1, 2, 3, are equal. We consider four

subcases.

Subcase 2.1. Every two of the sets L(wi ) have one element in common, say
L(w1 ) = {a, b}, L(w2 ) = {a, c}, L(w3 ) = {b, c}. Define c(w1 ) = c(w2 ) = a
unless {L(u2 ), L(u3 )} = {{a, b}, {a, c}}, in which case, we define c(w1 ) =
c(w3 ) = b, c(u2 ) = c(u3 ) = a and c(w2 ) = c, resulting in a list coloring.

Subcase 2.2. Two pairs of the sets L(wi ) have one element in common, say
L(w1 ) = {a, b}, L(w2 ) = {a, c}, L(w3 ) = {b, d}. We may define c(w1 ) =
c(w2 ) = a unless {L(u2 ), L(u3 )} = {{a, b}, {a, d}}, in which case, we may
define c(w1 ) = c(w3 ) = b, resulting in a list coloring.

Subcase 2.3. Exactly two of the sets L(wi ) have one element in common, say
L(w1 ) = {a, b}, L(w2 ) = {a, c}, L(w3 ) = {d, e}. We may define c(w1 ) =
c(w2 ) = a unless {L(u2 ), L(u3 )} = {{a, d}, {a, e}}, in which case we may
define c(w1 ) = b and c(w2 ) = c. If L(u1 ) = {b, c, d}, define L(w3 ) = e;
otherwise, define L(w3 ) = d.

Subcase 2.4. The sets L(wi ) are pairwise disjoint, say L(w1 ) = {a, b}, L(w2 ) =
{c, d}, L(w3 ) = {e, f }. Let {c(w1 ), c(w2 ), c(w3 )} = {a, c, e}, unless L(u2 ) ⊆
{a, c, e} or L(u3 ) ⊆ {a, c, e}. We can redefine at most two of c(w1 ), c(w2 ), c(w3 )
so that L(u2 ) 6⊆ {c(w1 ), c(w2 ), c(w3 )} and L(u3 ) 6⊆ {c(w1 ), c(w2 ), c(w3 )}. We
may redefine the other color among c(w1 ), c(w2 ), c(w3 ), if necessary, so that
L(w1 ) 6⊆ {c(w1 ), c(w2 ), c(w3 )}. Hence, a list coloring can be defined from the
given lists.
Chapter 16

Map Colorings

Section 16.2. Colorings of Planar Graphs

1. Every two of Brazil, Argentina, Bolivia and Paraguay are neighboring coun-
tries and so four colors are required to color only these four countries.

2. One such map M is shown in Figure 16.1.

..
..........
..... ..........
....
.
.....
..... .....
.....
.....
1
.....
......... 2 ...........
. 3
................... ........ ........ ....................
.....
.....
... ............ ..........
M: ...
...
...
.
...1 ... . ...
...
..... ..
...
...4 .... .
...
2 ....
... ..... ... ....
... ...
....... 3 ... ..
......

Figure 16.1: A map in Exercise 2

3. Proof. If G is a maximal outerplanar graph of order 3 or more, then there is

a planar embedding of G such that the boundary of every interior region of G
is a triangle. Thus, χ(G) ≥ 3. By Theorem 10.23, H = G ∨ K1 is planar. By
the Four Color Theorem, χ(H) ≤ 4 and so χ(G) ≤ 3. Therefore, χ(G) = 3.
4. Proof. Triangulating the n-gon results in a maximal outerplanar graph G of
order n. By Exercise 3, χ(G) = 3. Color the vertices of G with the colors 1, 2
and 3. Assume, without loss of generality, that the color 1 is used to color the
minimum number of vertices of G. Then the number of vertices of G assigned
the color 1 is at most bn/3c. By appropriately stationing security guards near
the corners of the art gallery that correspond to the vertices of G colored 1,
every location in the gallery has a straight line view by some guard.

197
198 CHAPTER 16. MAP COLORINGS

Section 16.3. List Colorings of Planar Graphs

5. Proof. We proceed by induction on the order of an outerplanar graph. Since
every graph of order 3 or less is 3-choosable, the result holds for outerplanar
graphs of order 3 or less. Assume that every outerplanar graph of order k is
3-choosable, where k ≥ 3. Let G be an outerplanar graph of order k + 1. For
each vertex v of G, let L(v) be a set with |L(v)| = 3. Let x be a vertex of
degree 2 or less. Then G − x is an outerplanar graph of order k. Hence there
exists a list coloring c of G − x such that c(u) ∈ L(u) for each vertex u in
G − x. Since deg x ≤ 2, the vertex x is adjacent to at most two vertices of
G − x and so there exists an element a ∈ L(x) such that no neighbor of x is
colored a. Let c(x) = a. Thus, G is 3-choosable.
6. Proof. We proceed by induction on the order of a planar graph. Certainly
every planar graph of order 6 or less is 6-choosable. Assume that every planar
graph of order k is 6-choosable, where k ≥ 6 and let G be a planar graph of
order k + 1. For each vertex v of G, let L(v) be a set with |L(v)| = 6. Let x
be a vertex of G such that degG x ≤ 5. Then G − x is a planar graph of order
k. Since G − x is 6-choosable, there exists a list coloring c of G − x such that
c(u) ∈ L(u) for each vertex u in G − x. Hence, there exists a color a ∈ L(x)
such that a is not the color of any neighbor of x. Let c(x) = a. Since this
gives a list coloring of G, it follows that G is 6-choosable.

Section 16.4. The Conjectures of Hajós and Hadwiger

7. Let G = K4,4 with partite sets U = {u1 , u2 , u3 , u4 } and W = {w1 , w2 , w3 , w4 }.
Assume, to the contrary, that G contains a subdivision F of K5 . Suppose
that V (F ) = X ∪ Y such that |X| = 5, degF x = 4 for each x ∈ X and
degF y = 2 for each y ∈ Y . We may assume, without loss of generality, that
|X ∩ U | = k ≥ 3 and u1 , u2 , u3 ∈ X. If k = 3, then we may further assume
that X = {u1 , u2 , u3 , w1 , w2 }. Since u1 u2 , u1 u3 , u2 u3 ∈
/ E(F ), it follows that
|Y ∩ (W − {w1 , w2 })| ≥ 3, which is impossible. If k = 4, then we may assume
that X = {u1 , u2 , u3 , u4 , w1 }. However then, |Y ∩ (W − {w1 })| ≥ 6, which is
impossible.

8. Let G = K4,5 with partite sets

U = {u1 , u2 , u3 , u4 } and W = {w1 , w2 , w3 , w4 , w5 }.
Then G contains the subdivision of K5 as shown in Figure 16.2.
 199

u2
........
.................
...... .. .. ......
...... .. ... ......
........... .... .... ...........
.. ... ......
w ...... ... ............ w
3................. ...
...
... ... ..... 4
... ...
...... ... ..
.. ... .....
.
. ........ .. .
. .....
.....
. .
...... ..
. ... .....
..... . .
... .....
..
. . .
. ..........
u ................ .. ......... ...
1 ..................
... ....... .
.
.
.
..
. ...... ....
... ..............
. .
............ u3
. ...
...
...
...
.......
.........
....... ....
... w 5
.
...
...
... .............
......
... ....
...
.
..
... ........ .... ...
... ......... .......... ..
... .. ............
. . ... .
....... .... ...
... . ....... ....... ... .
... ... ........... ... ...
... ... ....... ............ ... ...
... .. ....... ....... ... ..
... .... ............ ....... .. ....
... .. ........ ....... ... ..
.. .. .... ....... .. ..
............... .............
............
........
.... ...
w1 u4 w2

Figure 16.2: The graph in Exercise 8

9. Consider the graph G of order n = 13 consisting of three copies G1 , G2 , G3

of K3 and two copies G4 , G5 of K2 , where each vertex of Gi is adjacent to
each vertex of Gi+1 for i = 1, 2, . . . , 5, where G6 = G1 . See Figure 16.3. Then
α(G) = 2 and so χ(G) ≥ n/α(G) = 6.5. Thus, χ(G) ≥ 7. The 7-coloring
shown in Figure 16.3 verifies that χ(G) = 7.

1
...............
...
... .....
.. ...
. ...
. ...
............. .............
2 ... ... 3
qqqq qqqqq
qqqq G2 qqqqq
qqq qqqqq
qqqq qqqqq
4
.........
...........
qqqqq qqq 4
........
..........
... ..... ... .....
.
... ... ..... ...
... ... . ...
...
5 .............. ............ 6 5 ............ ..............
G: . ... ... . 6
G1 G3
qqq qq
qqq qqq
qqq qqq qq
qqq qqq
qq
1 .................... qqq
........
2 .....
......... .....
..... .....
.
.....
. G4 .........
... ...
...... ...
............. G5 7
3

Figure 16.3: The graph in Exercise 9

We claim that G does not contain a subdivision of K7 , for suppose that G

does contain such a subgraph H. Thus, H contains seven vertices of degree 6
and every two of these vertices are connected by six internally disjoint paths
in H.
First, we show that G4 cannot contain exactly one vertex v of degree 6 of
H. If this were the case, then v is connected to each of the remaining six
vertices of degree 6 by six internally disjoint paths. Since these paths must
pass through the vertices of G3 and G5 , which have a total of five vertices, this
200 CHAPTER 16. MAP COLORINGS

is impossible. Similarly, G5 also cannot contain exactly one vertex of degree

6 of H.
Next we show that not both vertices of G4 are vertices of degree 6 of H.
Suppose that u and v are vertices of G4 and both have degree 6 in H. Since
u and v are connected to five other vertices of degree 6 of H by mutually
disjoint paths, these paths must contain all vertices of G3 and G5 . Thus, the
vertices of G3 and G5 must also be vertices of degree 6 of H, which implies
that the three vertices of G1 are vertices of degree 6 of H. This is impossible.
Hence, no vertex of G4 or of G5 is a vertex of degree 6 of H.
Therefore, all three vertices of G1 , G2 or G3 are vertices of degree 6 of H.
This implies that at least three internally disjoint paths must pass through
G4 and G5 , which is impossible.

10. Proof. Let G be a graph of order n and size m containing Kk as a minor.

Then it is possible to partition V (G) into k subsets V1 , V2 , . . . , Vk such that
Gi = G[Vi ] is connected for i = 1, 2, . . . , k and such that there is a vertex of Gi
adjacent to a vertex of Gj for all pairs i, j of distinct integers with 1 ≤ i, j ≤ k.
Suppose that Gi has order ni and size mi . Then mi ≥ ni − 1. Hence
k   X k    
X k k k
m≥ mi + ≥ (ni − 1) + =n+ − k,
i=1
2 i=1
2 2

as desired.

Section 16.5. Chromatic Polynomials

11. Proof. The chromatic polynomial of a graph G of order n can be expressed
as
P (G, λ) = λn − a1 λn−1 + a2 λn−2 + · · · + (−1)n+1 an−1 λ,
where ai ≥ 0 for 1 ≤ i ≤ n − 1. Thus,
a1
X a2
X
P (G, λ) = P (K n , λ) − P (K n−1 , λ) + P (K n−2 , λ) −
i=1 i=1
an−1
X
n+1
· · · + (−1) P (K1 , λ).
i=1

12. (a) See Figure 16.8 (at the end of this chapter). Thus,

P (C6 , λ) = λ(6) + 9λ(5) + 20λ(4) + 10λ(3) + λ(2)

= λ6 − 6λ5 + 15λ4 − 20λ3 + 15λ2 − 5λ
= (λ − 1)6 + (λ − 1).
 201

(b) P (C6 , 2) = 2. This is not surprising since χ(C6 ) = 2 and there are two
colorings of C6 using the colors 1 and 2.
13. (a) See Figure 16.4. Thus, we obtain

P (K2,2,2 , λ) = λ(6) + 3λ(5) + 3λ(4) + λ(3)

= λ6 − 12λ5 + 58λ4 − 137λ3 + 154λ2 − 64λ.

u r rv
......... ............ .....r ........... ...........r
r
.............
....... ....................... ....... ..... ..... .... .....
........ ............................... ..... .. .. .....
r .............. ....... ........ .....
r = u ...r........................................................................................................................................................r +
...........................................................................................
.................. ..................... u r ..... .. .. .....
........... .... ..... ...........
........ ........
r
r .. .. ......... ... .
r v .r..
...................................................................... ............... ...................r
... ...... .. .
r
.... .. .
.. ............. ...
......... ............. r
v

...
....
..
. r ..
................................
r r r .... ....
......................... ...........
.....
...
...
..
.... r
.
...........
r
.......
... ........
.....
...
...
.... .................... ....... .... . .......
r +
...

=
...
r
............. ....... ......... .....
..... ............... .... ............. ..................... r + .... ............. ...................r
... ......... ...................... ...
...
+
r
... ....... ... .... ........
..... .......................................................... .
.
.. r
.. ................. ....
............. ...
.. ................. ...
... ................... ... ...................
r r
... ....... ... .... .................... .. ...
r r
.
... .... ..... ...................................
r r .. .
................... ......... ............. ............ ........... .. ...
r r
...
... ................ ................ .................................. ... ... ...................... ..
.... .. ...
...
.. .... ...

r r
........r
.... ......
................. .......... ......
= ..... .. ................. .. ... ....
r ..... ........ ............ .. ..
... ............ ..... .... ..........
........ .. ......
r u + 2 ....r...................................................................................ru +
........................................................................................ r......... .....
ru
.
... ... ......................... .. ...... .........
... ... ........... ... ......
r
.......................................
........... .... rv ..r
............ ...........
. ..r
v r .
.....
.
...... ...........
.
rv
...

= (K6 + K5 ) + 2(K5 + K4 ) + (K4 + K3 )

= K6 + 3K5 + 3K4 + K3

Figure 16.4: Graphs for Exercise 13

(b) P (K2,2,2 , 3) = 6. This is not surprising because χ(K2,2,2 ) = 3 and there

are exactly 3! = 6 ways to color the vertices of K2,2,2 using the colors 1,
2 and 3.
14. Proof. We proceed by induction on n ≥ 3. Since

P (C3 , λ) = λ(3) = λ(λ − 1)(λ − 2) = (λ − 1)3 − (λ − 1)

= (λ − 1)3 + (−1)3 (λ − 1),

the result holds for n = 3. Assume that

P (Cn−1 , λ) = (λ − 1)n−1 + (−1)n−1 (λ − 1)

for an integer n ≥ 4. We show that

P (Cn , λ) = (λ − 1)n + (−1)n (λ − 1).

Since Pn is a tree of order n, it follows that P (Pn , λ) = λ(λ − 1)n−1 by

Corollary 16.15. Let the end-vertices of Pn be u and v and let H be the graph
obtained by identifying u and v. Thus, H = Cn−1 . Then

P (Pn , λ) = P (Pn + uv, λ) + P (H, λ)

= P (Cn , λ) + P (Cn−1 , λ).
202 CHAPTER 16. MAP COLORINGS

Therefore,
P (Cn , λ) = P (Pn , λ) − P (Cn−1 , λ)
= λ(λ − 1)n−1 − [(λ − 1)n−1 + (−1)n−1 (λ − 1)]
= λ(λ − 1)n−1 − (λ − 1)n−1 + (−1)n (λ − 1)
= (λ − 1)n−1 (λ − 1) + (−1)n (λ − 1)
= (λ − 1)n + (−1)n (λ − 1),
giving the desired result.
15. (a) If G is a unicyclic graph of order n containing an n-cycle, then G = Cn
and by Exercise 14, P (G, λ) = (λ−1)n +(−1)n (λ−1). Suppose then that
G is a unicyclic graph of order n containing a k-cycle where 3 ≤ k < n.
Then G contains an end-vertex v. By Theorem 16.14, P (G, λ) = (λ −
1)P (G − v, λ). Continuing in this manner, we obtain P (G, λ) = (λ −
1)n−k P (Ck , λ). Therefore, every two unicyclic graphs of the same order
and containing a cycle of the same length are chromatically equivalent.
(b) By (a), the number of distinct chromatic polynomials for unicyclic graphs
of order n is n − 2 (one for each possible cycle length).
16. Let G be a graph with components G1 , G2 , . . . , Gk . For 1 ≤ i ≤ k, P (Gi , λ) is
the number of proper λ-colorings of Gi . Since a λ-coloring of Gi is independent
of a λ-coloring of Gj for 1 ≤ i 6= j ≤ n, it follows that
k
Y
P (G, λ) = P (Gi , λ).
i=1

17. (a) Proof. Suppose that the statement is false. Then among all counterex-
amples, let S be the set of those of minimum order n. Let G be a graph
in S of minimum size m. Thus, P (G, λ) = λg(λ) where g(0) = 0. By
Corollary 16.15, G is not a tree. Let e = uv be an edge of G that is not a
bridge and let H be the graph obtained by identifying u and v in G − e.
Then P (G, λ) = P (G − e, λ) − P (H, λ). Let P (G − e, λ) = λf (λ) and
P (H, λ) = λh(λ). By assumption, f (0) 6= 0 and h(0) 6= 0. Hence, both
the coefficient a of λ in P (G − e, λ) and the coefficient b of λ in P (H, λ)
are nonzero. Since the orders of G − e and H are of opposite parity and
the coefficients in every chromatic polynomial alternate in sign, one of a
and b is positive and the other is negative and so the coefficient a − b of
λ in P (G, λ) is nonzero. Since g(0) = 0, it follows that the coefficient of
λ in P (G, λ) is 0, a contradiction.
(b) Proof. Assume first that G has exactly k components G1 , G2 , . . . , Gk .
If any component Gi is trivial, then P (Gi , λ) = λ. If a component
Gi is nontrivial, then by (a), P (Gi , λ) = λfi (λ), where fi (0) 6= 0. By
Exercise 16,
Yk
P (G, λ) = P (Gi , λ) = λk f (λ),
i=1
 203

where f (0) 6= 0.
For the converse, suppose that G is a graph for which P (G, λ) = λk f (λ)
for some positive integer k where f (0) 6= 0. Suppose that G has exactly
` components. Then by (a) and Exercise 16,
P (G, λ) = λ` g(λ) and g(0) 6= 0.
Since λk f (λ) = λ` g(λ), and g(0) 6= 0, and f (0) 6= 0, it follows that
k = `.
18. Proof. Let F be a forest of order n with k components T1 , T2 , . . . , Tk , where
Pi khas order ni for 1 ≤ i ≤ k. Thus, each component Ti is a tree and
T
i=1 ni = n. By Corollary 16.15 and Exercise 16,

k
Y
P (F, λ) = λ(λ − 1)ni −1 = λk (λ − 1)n−k ,
i=1

as desired.
19. Proof. We proceed by induction on the number of blocks in G. The result is
obvious if G has only one block. Assume that the result is true for connected
graphs with r − 1 blocks, where r ≥ 2. Let G be a graph with r blocks, say
B1 , B2 , . . . , Br . We may assume that Br is an end-block with cut-vertex v.
Let H be the graph obtained from G by deleting the vertices of Br except v.
Then H has r − 1 blocks B1 , B2 , . . . , Br−1 . By the induction hypothesis,
Qr−1
P (Bi , λ)
P (H, λ) = i=1 r−2 .
λ
Let F be the disconnected graph with the two components H and Br . By
Exercise 16, P (F, λ) = P (H, λ)P (Br , λ). Since the number of λ-colorings of
F in which v is assigned the same color in Br as it is in H equals
P (H, λ) · P (Br , λ)
,
λ
it follows that
r
P (H, λ) · P (Br , λ) Y P (Bi , λ)
P (G, λ) = = .
λ i=1
λr−1

This completes the proof.

20. Let G = C5 and let H be the graph obtained from C3 by adding two pendant
edges. Then G and H have order 5, size 5 and chromatic number 3. By
Exercises 14 and 15,

P (G, λ) = (λ − 1)5 − (λ − 1) = λ5 − 5λ4 + 10λ3 − 10λ2 + 4λ

and
204 CHAPTER 16. MAP COLORINGS

P (H, λ) = (λ − 1)2 P (C3 , λ) = (λ − 1)2 λ(3) = λ5 − 5λ4 + 9λ3 − 7λ2 + 2λ.

21. Proof. The chromatic polynomial P (Kr,r , λ) is a polynomial of degree 2r

such that the coefficient of λ2r−1 is −r2 and P (Kr,r , 2) > 0. Since Kr,r is the
only bipartite graph of order 2r and size r2 , it is chromatically unique.

22. Proof. Let G be a graph for which P (G, λ) = λ(λ−1)n−1 . Since λ | P (G, λ)
but λ2 - P (G, λ), the graph G is connected by Exercise 17(b). Because the
degree of P (G, λ) is n and the coefficient of λn−1 is −(n − 1), it follows that
G has order n and size n − 1. Hence, G is a tree.

23. Proof. Let G be a connected graph of order n. The vertices of G can be

listed as v1 , v2 , . . . , vn such that vi (2 ≤ i ≤ n) is adjacent to at least one
vertex of {v1 , v2 , . . . , vi−1 }. We now consider the number of ways that the
vertices of G can be colored with 1, 2, . . . , λ. The vertex v1 can be assigned
any one of these λ colors. For each i with 2 ≤ i ≤ n, the vertex vi can be
colored with at most λ − 1 colors. Thus, P (G, λ) ≤ λ(λ − 1)n−1 .

24. The statement is false. Suppose that there is a graph G with P (G, λ) =
λ4 − 3λ3 + 3λ2 . Thus, G has order 4 and size 3 and so χ(G) ≥ 2. However,
P (G, 1) = 1, which implies that χ(G) = 1. This is impossible.

25. The statement is true. First, we determine the chromatic polynomial of G =

K2,4 . See Figure 16.5.

u ....... ............... ......... y

..... ......
...... ......... ......
...... .. ..... ...........
x ....................... ..................
............. ........... ...... .................................
.......
.... ...
...... .... ....... ...... ...... ..... ..... ...... ... ..........
.
... ... ...... ... .. .. ...
......
..
...
... ......
...... ... ... ......
.
...... ...... .. ..... ... ... . .
.. . .. ... .... ..... .....
.
...... ...
... ...... ... ... ................... ..... ...
.
... .. .... ......
...... ...... .......... ... ....
G: .............. ...... . ... ... .... ... .. ... ... ... ..
.......
......
. .
........
.... .
.
.
.......
.....
.. .
........... = ... ... ..... ... ..
... ... ... ... . + ...
... ..
..
... ....
..... .....
......
...... .... .
.
..
........ ... .......... .. ............ .............. ............. .............
...... .. .. ..
... ... ... ... .. ...... ...
...... .... .... ........... ... ..... ...
...... ... .. .....
...... .............. G 1 ...
............
..
..... ....
.... ........... G2
v

Figure 16.5: The graphs G, G1 , and G2 in Exercise 25

We compute P (G1 , λ). There are λ choices for a color for x and once a color
has been selected for x, there are λ − 1 choices for a color for y. The color
for the four remaining vertices can be any of λ − 2 colors that are not used to
color x and y. Thus, P (G1 , λ) = λ(λ − 1)(λ − 2)4 . Since G2 is a tree of order
5, it follows that P (G2 , λ) = λ(λ − 1)4 . Therefore,

P (G, λ) = λ(λ − 1)(λ − 2)4 + λ(λ − 1)4

= λ6 − 8λ5 + 28λ4 − 50λ3 + 44λ2 − 15λ.

If G is not chromatically unique, then there must be a graph H of order 6,

size 8 and chromatic number 2 for which P (G, λ) = P (H, λ). However, the
only graph H with these properties that is not isomorphic to K2,4 is the graph
 205

obtained by deleting an edge from K3,3 . See Figure 16.6. From this, we obtain

P (H, λ) = λ(λ − 1)(λ − 2)(λ − 3)3 + 3λ(λ − 1)(λ − 2)3

+λ(λ − 1)(λ − 2)(λ − 3)2 + λ(λ − 1)(λ − 2)2 + λ(λ − 1)3
= λ6 − 8λ5 + 28λ4 − 51λ3 + 47λ2 − 17λ.

Since P (G, λ) 6= P (H, λ), the graph G is chromatically unique.

r ur r u.r.................r...............v..r
......... ...... ........
............ .............
u
r vr r
....... ........ .........
............ ... ...........
........ ...... .......
.......... ... ... ...... ..
..... ................ ...................... ... ......... ........ ...
H: ............ ............... ... ......... .... ... .. ........... ... ..
r r r
........... ............
...... .... ..........
= ............ .. ..............
r r r
......... .......... ........
+ r
............. .............
... ..... r
v ..... .....

...
...
...
r ur vr
......... ........ ........
............ .................
ur v
r
............. ...............
...
...
...
...
... ......
.. .... r
......................
. r r .....r......
.. ............ ........ ..............
....
r
... ....
...
...
.... .
..... ................. . ..
. ................ ... ... ... ... ......... ....... .. ..
.
... ..... ..... . ..... .
= ... . . .. .. . ..
. +
.............. ......................
.
. .
. ........ ........ ... . . +. .
. ... .................. .. ... .......... ... ..
r r r
...
r r r r
.
r r r
... .. .. . . . .. . .. .
.. .
. . .. .......... ...... .. ...
... . ... . .. ........ ....... ..... ......... .... ... .......... ...........
... ........ ........ ... . . . ... ..
.
.... .. .... ...
... .

r r r r................... r
..... ... .....
... ... ...........
... ...
r r r r r
...
...... ......
r r r
.
.... ... ..
.. ...... ........
...
.. ....... .... .......
= ... ................................. . . ... . . . ........ .. .. .......
....................... .. ... . .. . .
....................... ... . .
. .. ................... .... .
. .
.... .... + ..r.... .........r
..... ... .................. ...
r r r+ r r r +
. . .
r r
.
r r r+ r r r +
.
............. ................... . . .
........... ............... .... . . . .
... ............... .......... . ... .... ... .....................
... . . .
. .. . . .
. . .. . .....
... .... .. .. ..... .... ... .... ... ...
.
.
..
... ... ... ..
.... ... .... ...
u
=
r r vr
............. ........... ..........
............... ................. ..
r r r +2(K2 + K 3 ) + (K3 + K 2 ) + (K2 + K 2 ) + K1,3
. .. . .
................................ ..
............ ..... .... ..............

= (K3 + K 3 ) + 3(K2 + K 3 ) + (K3 + K 2 ) + (K2 + K 2 ) + K1,3

Figure 16.6: The graphs for Exercise 25

26. Since G is a maximal planar graph of order n ≥ 3, its size is m = 3n − 6.

Since n − m + r = 2, it follows that r = 2 − n + m = 2 − n + (3n − 6) = 2n − 4.
Thus, 2n − 4 vertices are added to G. Since each of these vertices is adjacent
to three mutually adjacent vertices of G, there are λ − 3 choices for its color.
Thus, P (H, λ) = (λ − 3)2n−4 P (G, λ).

Section 16.6. The Heawood Map-Coloring Problem

27. (a) See Figure 18.2.
... ... ...........
... ... ......
... ...
...
...
... a 5 .
....
. c
... .......... ...
..
...... ... ............................. ..... 4 .
........ 4 ... ........ ........ .. ........
. ....... ........... ..........
. ................. ... ...
... .... .
. ..
..... .....
. .
. .... .....
b ...... .....
....... b
.
.............
.......
1 2 ..... ....
.....
... ........
. ..
...
.......................... ....
... ...... ....... .
.
... .. ... .... ....... .
... ............. ....... ....
................
3
.............
.. .. 5
........ ..........
5 ..... .....
.....
..... . ......
.
.....
.....
............
........
..... ....
.
................
.
a
.....
.....
..... a .....
.....
..... c
4 .....
.....
.....
.....
.....
..... ...........
4
......

Figure 16.7: An embedding of the Petersen graph P on the torus

206 CHAPTER 16. MAP COLORINGS

(b) By the Generalized Euler Identity (Theorem 11.10), n − m + r = 0 and

so 10 − 15 + r = 0. Thus, r = 5.
(c) Since every two regions are adjacent, 5 colors are needed.
28. The statement is false. For example, by the Four Color Theorem (Theo-
rem 16.2), χ(S0 ) = 4. Now 2 = χ(K3,3 ) ≤ χ(S0 ) but K3,3 cannot be embed-
ded on S0 since K3,3 is nonplanar.
29. By Theorem 11.18, γ(K8 ) = 2. By Theorem 16.18, χ(S2 ) = 8. Thus, χ(S2 ) =
χ(K8 ) = 8.
 207

r r r r
u r ..........
.....
r r
...
... ...... ....
... ...
...
...
....
r u ....r............ r ru
..... ..........
ru
...
u r
...........
..... .......... rv r
........
..... ......... r
... . r ... .....
.......
... ...
r ...
r
.. .
..... ........... ..... .... ............
r + r ........
..... .
r
...
u ..... . .....
......... r .... ... ....
..... ............ r +r r
......
......
rv
..
+
... ......... ..........
.r..... r rv + u r v = r
.
r ..
= ...... ......... ... ...... ... ..... ......
r
..... ..... . . .........
.. ............ ... ...... .. ... ...
v r ...
. ..... .
.......r.......
. ..... ......... .
r
... . ... . ... ..
....... ... ... ...
v v
ur
.... .....r.
u ......
r ..... ........... r ur
.....
r u..r ....... ..... ...
......r
... .... ...
..... . ...
.. ...
r
.
. ...r ..ru r
. . . .
. ...... ... .
....r ..
. . . u ...
........
r
... ...
r
.. . . . . .
r + ......... ... + ... ........
. .
...r
. ..... v .
..r .. ...r +
. .. .. .. . . . .
. .. ... ... .... .
. .... ... .. .. ... ....... . ...
.... . . .. . . . . ..
. + ... . + K
r
.
r
. . . . . . . . . . .. . .
= .... . ..
v .......... 3 ...
............. + .. ......r ..
r
.... . .. . .
... ........r r
.
.r .
. . .
... u ...r
+ . . . . . . . ..
....r .. ... r ...r
.... . ... ..... .. ... . ...
r ........r
... ..... ... ... ..... .. . . . .
. ..
.. .
... .... . ... . .. . . .
... .
. .
.......r
. ....... .... . ... .. .... . .. .... v ..
... ... .... v ...
. .... v .
... . ..
.... v .. v
r .. .. .r
...r ... ........r
...... .
.. . .... .....
. ..... ......... ...... .... ...... ..........
.. ...
....r .....r .. ..r
.... ...
.. .r .r r .........
.... .
.. ..r
r ......... ... ...... .........r
..r
... ..... .......... ..... ... .......
..r
.......... .............r
..... ..... ...
... . ..... ........
.
... ..
.....
..... ..... .. ......... .........
.....
.....
= ..... ... ............... + .. . .
.. + . . ..... .. ...... . + . . . . + r ..........r ......
.
........ ... ...
...r
...
.............. ..............r ..... +
...
... ..........
...r.. r .r ....
..... .....
..... ..........
.r ..... ... ..r ..... .........r
... ... ...
... r .. r
.. ... ..... ...
.........r ..
...
... ........ . .... .. .
... ...
....
. .. .. .. .. ...
....
r
. ..... ..
. .
.. .
. ..
.
r .....r ......
... . ... . ...
... ... ...
r
...
.. .r
..
...r
....... ......r .. .r
... .......... ...
... ... ... ...
+(K4 + K3 )+ ....... ............................. + K4 ...... + ...... ................ + K3 ...... + ....... ................ + ......r ...... +K3
.. . .. .. ..
... r .....r ... r ..r
... ..r
... ... .... r ..
.
... ..
. ...r ....
... ... ...
... ... ....
... .... ...
. . ...
......r........ u r r
...r.......
... .. . . ......
..r u
....r
. ..
... ....
.. .......r
.
...r
...
...... ......r
... ...
.
. .
..r...
ur ru
= ............ .............
..r................ .............r +
.. ........ ...... ....... ......
...... u ..... r
3
.r ..r r ....r
.........
r ....r
+ + .....
.
. .. ...... 4 . +2K + 3K +
.....r rv
......... ....... .. ... .. . ..
. 4 3
..
v .. v v
... v
...
r r ....
....
..
r ..... ...
.....

...r
.. ... . . . ...
..r
..
....r......... ............r .... r ........
. ..... ..
. ... .. . .
...... . . .. .
..... ..... ...
..... .r .......r
...
... ....r.. ... ......r
... .......... ......... ... ......... ....... ... ... ....
.. .......... .........
..... ...r
... ............. ................r + r ..........r ....
..... ......
= ... .................................. + .... .................. .... ... + 3 ..... ..................................... +K4 .... +
.. .. . . ......
... ..r
... ....
........ ...........r ...r .. ......r ... ... .....r..... ..........r
.
... ..........r
... .r ..r
. . . ..
. ..... . ... .. ... ....................... .
.... ........ ...
... .... .
.... . .
....
. . ... .

+4(K4 + K3 ) + 2K4 + 3K3 + (K3 + K2 )

...r
....r...
u
r ....r ... ....
..r
..
r r .......
....r
..... ..........
r ....... ..... ... ....r
....r r
3 ......r..............................................r +
...... ...... ........ .... ... .........
................ ............. .....
. .
. . . . . ...... ........ ......
... ................... ...................... ... + ......
= v r r u+
.....
..r ...r .......... + r .........r +9K4 + 8K3 + K2
.
.....r
....... ........r ..... ....r
r
. ...
.... .......... .................. ....
......... .... .. ......
r v .
.................
... u .
. .
v
...
...
r ...r ..
..... ....
r .....
... .r....... ........r
. ...
.. .. . . .. ...
.. .... ...
..... .......................... .... .
.. ....... ........
.
. . r
= ..... ...r..............................r + .r................................r ...... +2 ...... ...r..........................r+ ............. ....... +3(K + K ) + 9K + 8K + K
.... ...... ..... r ...
..
... ......... ....... ..r
... .....r .. ... r
... .... .....r
... .... r ....r ..... 5 4 4 3 2
... .....r ..
... ................
. ... .... ..
... . ...
....r
u
...r
u
...r
...... ........... ...... ......... .r ru
................. ...............r ......r ...... ............r +2
...... . ...... .. .....
.....
=
...r ......... .... ......r +3 r . r +3K5 + 12K4 + 8K3 + K2
........................ . ... .............. .. .....
v ............................... ....r
..... .......r
. . . ..
..
....r
... v v

...
... ur .....
... ...
... .
ur .....
...
r
... ................... ... ............... ...
= ..... .................... .............
... .............................
r ...
...
... r
... ...... ... ....... r +K
..... ................... .............
...
...
r
... .................. ...........
... ........... ........... r +K5 .. +3
.. r r
... ... ............... ...
... ..... ....
4 ....
..
+2(K4 + K3 ) + 3K5 + 12K4 + 8K3 + K2
... .....................
.... r .. .
...
...
...
.... v .....
.

.. v
ru
.....
.... ...
...
= r .
... ...............................
..... ......................................... r ...
...
.

r rv+K5 ....... +3(K5 + K4 ) + 4K5 + 17K4 + 10K3 + K2

... ....................................................
... .......... .. ...........
..
... ......................
.... r
..... ..
...
= K6 + 9K5 + 20K4 + 10K3 + K2

Figure 16.8: Graphs for Exercise 12

208 CHAPTER 16. MAP COLORINGS
Chapter 17

Edge Colorings

Section 17.1. Chromatic Index and Vizing’s Theorem

1. Since ∆(G) = 3, it follows that χ0 (G) = 3 or χ0 (G) = 4. Suppose that
χ0 (G) = 3. Then there exists a 3-edge coloring of G using the colors 1, 2,
3. We may assume that vx is colored 1, wx is colored 2 and xy is colored
3 (see Figure 17.1). Then wy must be colored 1 and vy must be colored 2.
However, uv and uw must then both be colored 3, which is impossible. Thus,
χ0 (G) = 4.

u ......
.... ...
....... ..................
....... .......
.
............ .......
...... .......
...... .......
.............. ..........
v .............
........
........ 2 .......
..... ..
........ .... w
........ .
.... ......
.
........ ..........
.............
1 ........
... ...
........ ................
........
........ ........
..
...
. ......... ..............
..
............. ......... y
x . .

Figure 17.1: The graph in Exercise 1

2. Let H be the graph obtained by adding a pendant edge to a vertex v of

maximum degree in a graph G. Then ∆(H) = ∆(G) + 1. Let ∆(G) = k and
let there be given a χ0 (G)-edge coloring of G, using the colors 1, 2, . . . , k and
possibly k + 1. At least one of the colors 1, 2, . . . , k + 1 is not used for an edge
incident with v. This color may be used for the pendant edge added to G to
obtain H. Thus, χ0 (H) = k + 1 = ∆(H).
3. Proof. Since ∆(G  K2 ) = 1 + ∆(G), it suffices to show that there exists a
(1 + ∆(G))-edge coloring of G  K2 . Because χ0 (G) ≤ 1 + ∆(G), there exists
a (1 + ∆(G))-edge coloring of G. Let G0 be another copy of G and let the
identical (1 + ∆(G))-edge colorings be assigned to G and G0 . Let v ∈ V (G)
and let v 0 be the vertex of G0 that corresponds to v. There is at least one
color α of the 1 + ∆(G) colors that is not used to color the edges of G incident

209
210 CHAPTER 17. EDGE COLORINGS

with v. Also, α is not used to color the edges of G0 incident with v 0 . Assign
the edge vv 0 of G  K2 the color α. Proceeding in this manner for each such
edge of G  K2 , a (1 + ∆(G))-edge coloring of G  K2 is produced.

4. (a) Proof. Suppose that ∆(G) = k ≥ 1. Then χ0 (G) = k or χ0 (G) = k + 1.

Now ∆(G ∨ K1 ) > ∆(G) and so ∆(G ∨ K1 ) ≥ k + 1. Hence, χ0 (G ∨ K1 ) ≥
k + 1 ≥ χ0 (G).
(b) Let G = Kn , where n is an odd integer. Then χ0 (G) = ∆(G) + 1 = n.
Since G ∨ K1 = Kn+1 and n + 1 is even, χ0 (G ∨ K1 ) = χ0 (Kn+1 ) =
∆(Kn+1 ) = n. Thus, χ0 (G ∨ K1 ) = χ0 (G) for every complete graph G of
odd order.

5. (a) Proof. Let Vi = V (Gi ) for i = 1, 2 where |Vi | = ni . Then G1 ∨

G2 contains the complete bipartite graph Kn1 ,n2 as a subgraph. Let
n = max{n1 , n2 }. Then χ0 (Kn1 ,n2 ) = n. Let there be given an n-edge
coloring of Kn1 ,n2 in G1 ∨ G2 using the colors 1, 2, . . . , n. Suppose that
k = max{k1 , k2 }. Then χ0 (Gi ) ≤ k for i = 1, 2. Let there be given a
k-edge coloring of both G1 and G2 with the colors n + 1, n + 2, . . . , n + k.
Since there is then an (n + k)-edge coloring of G1 ∨ G2 , it follows that

χ(G1 ∨ G2 ) ≤ n + k = max{n1 , n2 } + max{k1 , k2 },

as desired.
(b) For G1 = K2 and G2 = C5 , n1 = 2, n2 = 5, k1 = 1 and k2 = 3. Hence,
B = max{n1 , n2 } + max{k1 , k2 } = 5 + 3 = 8. Since ∆(G1 ∨ G2 ) = 6,
either χ0 (G1 ∨ G2 ) = 6 or χ0 (G1 ∨ G2 ) = 7. In fact, χ0 (G1 ∨ G2 ) = 6. A
6-edge coloring of G1 ∨ G2 is shown in Figure 17.2.
6
.....
... ..... .......
.......................... ......... ........
..... .. ... ..... ........ ....... ........... ....
....... ..... ... ... .....
..
....... .... ..... .......... .............. . ...
. ........ .......... .... .... ........
.... ... ....... ......... .......... ....... .. ... .....
..... ... ..... . . . . .....
..... ...
...
... ..... ....................... ......... ...
...
... .....
1 ..... .. ... ..................... ......... .. ... .....
...
...... ...
. .. . .
. .
... ..... .......... .................. .
.
.....
.....
.
.... .. . ........
. ........ ........ ... ..... 1
.. . .... .. . .
.. .. ...
..
..... . ...........
.
.
...
..
. .... ........
. ... ............ .... .....
.....
.... ............
. ... ..... ..... .. ....... .. .....
.
.. . . ... ..
. . . .
. . .
..
..... . . . ..
....... ..
.
......
.
.......
. ..
........
. . ..
.....
.....
....
. .
....... ... ..... ....
. ... ...... ... ......... .....
.. .. . .. . ...... ... .........
..... .
....... .. . ....
. ... .
. .. . . 2 .....
.
........ 5 .. .
...........
.
.
.
. .......... ...
. ... . ......
........
.
...
...
.
........
. ..........
.....
....
...
.... .
. .
........ 5 ..
. ....
.... ...
. .
. . ..
...... . ........... .........
.... ............ . ........
4 ... . ... 2
. . ..
..
.
... .......... ..
..
. .......
. 4 .... ..
.......... .
..
3 3 . ....... ..
.
....... ........
.
....... ....
....... ....
................
.. .. ..
.................. .................. ... ... ..... .. .... ..
........ . .... ...... ........
........
2 ..
. ........
.........
.......... 1 5 4 ...... ........ ....
..........
............ ..
. ..... .......
.
............. .........
................ .............
........................ ................
.....................................................................................
3

Figure 17.2: The graph in Exercise 5(b)

6. (a) Since ∆(G) = 7 and µ(G) = 4, it follows that χ0 (G) ≤ 7 + 4 = 11 by

Theorem 17.3. By Theorem 17.4, χ0 (G) ≤ (3 · 7)/2 and so χ0 (G) ≤ 10.
(b) Since every two edges of G are adjacent and the size of G is 10, it follows
that χ0 (G) = 10.
 211

Section 17.2. Class One and Class Two Graphs

7. Proof. We show that G is overfull. Since H is r-regular of odd order, r is
even, implying that G is obtained from H by removing 12 (r −2) or fewer edges.
Since at most r − 2 vertices of H have their degrees decreased in forming G,
it follows that ∆(G) = r. Suppose that the order of H is n and so n ≥ r + 1.
Hence, the size m of G satisfies the following:
 
nr r − 2 n−1
m ≥ − = r+1
2 2 2
jnk jnk
= ∆(G) + 1 > ∆(G).
2 2
Hence, G is overfull and so by Corollary 17.9, G is of Class two.
8. The statement is false. Let G1 = K2 , H = K3 and G2 = K4 . Then G1 and
G2 are of Class one and H is of Class two.
9. Proof. Since G is 3-regular, it follows that G has even order. Let C be
a Hamiltonian cycle of G. Color the edges of C alternately 1 and 2. The
remaining edges of G form a perfect matching in G. Coloring these edges the
color 3 produces a 3-edge coloring of G and so χ0 (G) = 3 = ∆(G).
10. Let Gi (i = 1, 2) have order ni and size mi . Then n1 = 5, m1 = 9 and
∆(G1 ) = 4. Since m1 > bn1 /2c∆(G1 ), it follows that G1 is overfull. Also,
n2 = 9, m2 = 21 and ∆(G2 ) = 5. Therefore, m2 > bn2 /2c∆(G2 ) and G2 is
overfull. Since G1 and G2 are both overfull, it follows by Corollary 17.9 that
G1 and G2 are of Class two.
11. The graphs of the five regular polyhedra are of Class one (see Figure 17.3).
Since the tetrahedron is a complete graph of even order, it is of Class one.
Since the cube Q3 is a bipartite graph, χ0 (Q3 ) = ∆(Q3 ) = 3 by König’s
theorem (Theorem 17.7) and so Q3 is of Class one. The dodecahedron is a
Hamiltonian cubic graph and is of Class one by Exercise 9. Since there is a
4-edge coloring of the graph of the octahedron and a 5-edge coloring of the
graph of the icosahedron, these graphs are also of Class one.
12. Proof. Let G be a cubic graph with connectivity 1. Then G is a connected
graph containing at least one cut-vertex. Let u be a cut-vertex of G. Then
u is incident with a bridge e of G. Assume, to the contrary, that G is of
Class one. Then there exists a 3-edge coloring of G, using the colors 1, 2,
3. Suppose that c(e) = 1. Since each color class is a perfect matching, the
edges of G colored 1 and 2 form a 2-regular subgraph of G, which is a union
of cycles. Thus, every edge colored 1 lies on a cycle of G, which produces a
contradiction since e is a bridge.
13. Proof. If G is regular, say r-regular, then G is 1-factorable by Theo-
rem 10.15 and so χ0 (G) = r = ∆(G). Hence, we may assume that G is
not regular. Thus, G contains vertices with degree less than ∆(G) = r. Let
212 CHAPTER 17. EDGE COLORINGS

....... .........
............ .............. ................ .............
... ..... ... .. .. ..
...
......
..... .... . ... .... .... .....
....
...
... ..... ..
......
. ..
... ... ... ...
.. ..
. ... ..... .. .... ... ... .... .....
... ... ..... .... ............... ... ..
... ... ....... ... 2
.. ....
... ...
3
.... ... .. ... ...
..
.
.. ... 1 .
... ... 1 ... ...
... 4
.. ......... ... .... ......... .
................ ...
...
. ....
..................
.......
...
.. ... ..........
. .. .. ..
. ..... ........... ............. ... ......... .....
4 .. ........ ....
. ....... .....
4 2
.. .
... ... . . . ..... ...
.... ............ ........ ... ......... .... .... .. . .....
3 .............
... .. ...... ...
.................. ............. ..... ..... ... ......... ... .
. .. .. ... ..
....... ..... ..... .
. .. ........ . ...........
.......... ..............
...... .
..............
.. ..... ... ..... ........... .......... ..........
......... ........ .......... 2 1 . .
... ... ...
... ..........................................................................................................................
...
tetrahedron 3
.........
cube ....... octahedron
..................... ..... ...
...... ...... ..............
...... ...... ...... ..........
..
. .
....... ...... .. ....... .......
......
. ........ ......
...... 1 . .
... ...
... ...
.. ........ ........................... ...... ... .. 3 ... ....
... ... 5
. ... . .... .
..
.. ..... . .... . ....... ...
.... ...
.
......
...... .... .... .... .. . ..
. . . .. .. .. ...
...
.... ...
. . ..
.
.
. ... . .
. ...... ..
.
.. .... . .. ................ ..... .....
.
..... ......... . .... .... .. ......... ........ .. ... ....... ... ... ....... .... .... 2
.... ....... ......... . . . ...... .. ...... .... .. ..
2 4
.. .... ........ ... .... ....... .... ....
... ....... ... .... .. ... .. .. ...... ...... . 4
....... .......
. ..
..... ..... ............. ....
. .
.. . ... ... ... ....... ... ....
... ..... ... ... .. ..... 1 ... ..... ... ...
... ... .
. . . ... ... ... ...... ... ... 5 ..... ... ...
... ... ...
...
... ... ... ............. ... ... ........ ...
... ... .. ... .. .. ... ...... .. ....... .
... ....... ..... .. .. ..
.. .................. 3 .. ... ........ .....
...
... . ..... . . .
. .
. .. . .
. .... 3
.. ............... ............................................................................................ .....
. . .. .
... ...................... .............. .....
................................. ... .
. ... .... ...
..... ..... .
...... .... .. ......... ......... .. .......
... ... .................... .. .
..
.. ........ .... .. ....
... .... 2 . .
.. ........ ........... .....
1
...
... ...
...
... ...
.. .. .
...
...
.
.
. .. 5
... ..... ... 5 ...
......
.
..
4... ...
...........
. .
1 ..... 4
.....
. .
... ....... ....
.. .... ..
... ... ... ... ... ...... .. ...... .... ........ .....
..... ..... ........ .....
...
........ ... ..
.. .. 1
... ......... ..... .......... 2 . .... . .. .. . ... . . . . . . . 3 ..
. 5
..... .... ....... .....
...
... ............ ............. ........... .. ...... . ....... ............ ............
...... .... ..
. .. .
. ... ...... ......
. ... ... ............ ...... .... .. ... .
............ ..... .... ...............
... ..... ...
... .. . ... ...... ......................................... 5
. ... .......... ... ..... . .
4 1 ............ ... .... ..
............. ........ . . ............ ...................................................................... ....... ................................................................. .............
3 2
... . ...... ....... .......................................................................................... ........ ........
....... ........ 2
....................................................................
.................................... 3
.........................................................................
. .......
4
dodecahedron
icosahedron

Figure 17.3: The graphs of the five regular polyhedra in Exercise 11

the partite sets of G be U0 and W0 . Now let G0 be another copy of G whose

partite sets U0 and W0 of G are respectively U00 and W00 in G0 and where the
vertex v 0 in G0 corresponds to v in G. Let U1 = U0 ∪ W00 and W1 = W0 ∪ U00
so that |U1 | = |W1 |. Let G1 be the bipartite graph with partite sets U1 and
W1 such that

E(G1 ) = E(G) ∪ E(G0 ) ∪ {uu0 : degG u < r}.

If G1 is not r-regular, we construct a new bipartite graph G2 from G1 and

another copy G01 of G1 as above. Continuing in this manner, we arrive at an
r-regular bipartite graph Gk , where k = r − δ(G), containing G as an induced
subgraph. Since Gk is an r-regular bipartite graph, Gk is 1-factorable. Thus,
χ0 (Gk ) = r and so χ0 (G) = r = ∆(G) as well.
14. Proof. Let G be an r-regular graph of odd order n and size
m. Then m=
rn/2, α0 (G) ≤ (n − 1)/2 and ∆(G) = r. Since rn/2 > n−1 2 r ≥ α 0
(G)∆(G),
it follows by Theorem 17.8 that H is of Class two.
15. Proof. Let G be an overfull graph of order n and size m. Since α0 (G) ≤ n2 ,
 

it follows that jnk

m> · ∆(G) ≥ α0 (G) · ∆(G).
2
By Theorem 17.8, G is of Class two.
16. Proof. Let H be an overfull subgraph of G. Then clearly, ∆(H) ≤ ∆(G).
Assume, to the contrary, that ∆(H) < ∆(G). Suppose that H has order n0
 213

and size m0 . Since H is an overfull subgraph of G, it follows that

 0 
n −1
m0 > ∆(G). (17.1)
2
Also, X
2m0 = degH v ≤ n0 ∆(H) < n0 ∆(G).
v∈V (H)

Thus,
n0
 
0
m < ∆(G)
2
and so
n0 − 1
 
0
m ≤ ∆(G),
2
contradicting (17.1).
17. Let G = Cn  K2 where n ≥ 3 is odd. Then G is a 3-regular graph of order
2n. Let G1 and G2 be the two copies of Cn in G. Note that G1 (or G2 ) is not
an overfull subgraph of G. Hence, the conclusion of Theorem 17.13 cannot
necessarily be made if G1 is not an overfull subgraph of G.
The Petersen graph is another example.
18. This statement is false. Consider the graph G shown in Figure 17.4(a). The
graph L(G) is shown in Figure 17.4(b) together with a 6-edge coloring of
L(G). Then ω(L(G)) = 4. Since ∆(L(G)) = 6, it follows that χ0 (L(G)) = 6
or χ0 (L(G)) = 7. Since there is a 6-edge coloring of L(G), it follows that
χ0 (L(G)) = 6. Therefore, χ0 (L(G)) 6= ω(L(G)) and χ0 (L(G)) 6= 1 + ω(L(G)).
Thus, G is neither of Type one nor of Type two.
............. .......
...... .......
...
3..................... .... ..................1...
......
.... .............................................2
. . . .. .. ....... ...
... ............ ... .................................
............. .... ............. ... .....
..... .........
..... 2
..... .
.........

G: 1 ........................................ 3
. .
..................
L(G) : 4.................... ...................6.
............. ............. ............. .....
.... 5 .............
... ... ... .....................................................................................
...... ..
...... 5
...... .....
.....
...... .....
...... ... ..........
.............
... 6 ...................... 4
.

(a) (b)

Figure 17.4: The graphs G and L(G) in Exercise 18.

19. Proof. Let G be a graph of odd order n and size m for which n·∆(G)−2m <
∆(G). Then
(n − 1)∆(G) j n k
m> = ∆(G) ≥ α0 (G)∆(G).
2 2
By Theorem 17.8, G is of Class two.
214 CHAPTER 17. EDGE COLORINGS

20. The star K1,k is a planar graph of Class one with maximum degree k. For each
integer k with 2 ≤ k ≤ 5, the graph Gk of Figure 17.5 has order nk = 2k − 1,
maximum degree ∆(Gk ) = k and size mk = k 2 − k + 1. Since
jn k
k
∆(Gk ) = (k − 1)k = k 2 − k,
2
it follows that mk > n2k ∆(Gk ) and so Gk is overfull. By Corollary 17.9, Gk
 

is of Class two for each k with 2 ≤ k ≤ 5.

...
... ...
.............
... ..... ...
.......... ..... .... .... .....
............... .. . .. ...
..... ... .. ... ...
....
... ... .... .... .....
.. ... ... .. ... ...
... .
.....
. ... ... ... ... ...
... .... ..............
... .
...
. .
........... ..........
...
... .......
... ............... ................ .....
.... ..... . .... . . ..
. ..
...... ....
..
..... ... ..
.
.. ......................... ..
....... ...... .
.
.
.. ....... .
...... ....
..
.... ...
.... ............... ....... .. ... .......
. .... .... ..... ...
...... .......
................... .............. . ..
... .....
. ..
..... ... .......... . .
............. .... ..... ... ...... .......... ... . ........ .......... .....
... ............................................................................... ........ .... ..
. . .......................... ........ ........
........ ........
. . ...
....................... ...............
........
.....
G2 G3
...
.... ......
........................... G4
....... ... ..... ...........
... ....... ..... ... ......
... .
... ....... .....
.....
..... ...
.....
..... . . . .....
..
..... ..... ............................. ..... ....
...
.... .
. ............ .........
.
.....
. ... ...
. ... ............
. ..
.
.. ... ...
... ........................ ....... ............. ...... ..... ...
G5 : . ..
.
....
..
.
... ...
. ...
....... .....
. ..
.... .... ... ....
. ...
...
...
... .... ..
.. .
. . ...
... ........ ... ....
. .
.............
... ..
........... .
...
.... ....
.
.... ....... . .
.... .... . .
.............. .... ...
... ...... .... . . ... . .. . .......... .. ............. .
. ..
....... ............................................... ........... ..... ....
...................
........................................................ ..... .
. ......................
.......................................................................... .
. ....... .
. ... . .....

Figure 17.5: The graphs in Exercise 20

21. Proof. The graph H has order n = 4k + 1 and size m = k(4k + 1) = 4k 2 + k.

Thus, G has order n and size m0 = k(4k + 1) − (k − 1) = 4k 2 + 1. Since G is
obtained from H by removing k−1 edges, there are vertices in G having degree
2k and so ∆(G) = 2k. Hence b n2 c∆(G) = (2k)(2k) = 4k 2 < 4k 2 + 1 = m0 . So
G is overfull, implying that G is of Class two by Corollary 17.9.

22. (a) Proof. Let n = 4k + 1 be the order of G. Since δ(G) > n/2, G is
Hamiltonian.
(b) Proof. The size of H is m = (k + 1)(4k + 1) − (2k + 1) = 4k 2 + 3k. The
edge independence number of H is α0 (H) = 2k and the maximum degree
of H is ∆(H) = 2k +1. Since m > α0 (H)·∆(H) = 2k(2k +1) = 4k 2 +2k,
the graph H is overfull and so is of Class two by Corollary 17.9.

23. (a) The schedule given in the solution of Example 17.10 gives France two
consecutive days off. Only France has such a schedule. If we perform
the cyclic permutation (2 5 1) of colors (that is, color 2 is replaced by 5,
color 5 is replaced by 1 and color 1 is replaced by 2), then this is rectified.
Day 1: AS, BF, CI
Day 2: AG, BC, FS
Day 3: BS, GI
 215

Day 4: AF, CG
Day 5: AI, BG, CS.
Day 6: BI, CF, GS.
(b) The schedule given in the solution of Example 17.11 has player B only
playing in one match on Day 4. All other players play at least two
matches every day. Interchanging the colors of AC-BD and AC-DE rec-
tifies this.
24. (a) Let G = 2Kn/2 .
(b) Let C be a Hamiltonian cycle in K(n+2)/2 and then let
G = K(n−2)/2 + (K(n+2)/2 − E(C)).
25. Proof. Let G be a nontrivial self-complementary r-regular graph of order
n ≥ 2 and size m. Then, of course, G is also an r-regular graph of order n and
2 2
size m. Thus, r = (n − 1)/2 and m = 12 n2 = n 4−n . Since n2 ∆(G) = (n−1)

 
4
n2 −n (n−1)2
and 4 > 4 , it follows that G is overfull and therefore of Class two.
26. Proof. Suppose that G is r-regular. Since n is even, either n ≡ 2 (mod 4)
or n ≡ 0 (mod 4). We consider these two cases.
Case 1. n ≡ 2 (mod 4). If r ≥ n/2, then G is of Class one by Theorem 17.14.
Suppose that r < n/2. Thus r ≤ (n/2) − 1 = (n − 2)/2. Then G is (n − 1 − r)-
regular. Since n − 1 − r ≥ (n − 1) − (n − 2)/2 = n/2, it follows that G is of
Class one.
Case 2. n ≡ 0 (mod 4). If r ≥ (n − 2)/2, then G is of Class one by
Theorem 17.14; otherwise, r ≤ (n−2)/2−1 = (n−4)/2. Then G is (n−1−r)-
regular, where n − 1 − r ≥ (n − 1) − (n − 4)/2 = (n + 2)/2. Hence, G is of
Class one.

Section 17.3. Tait Colorings

27. (a) Let S = {s, t}. Since k(G − S) > |S|, it follows by Theorem 6.11 that G
is not Hamiltonian.
(b) The graph G has a Tait coloring since by Corollary 17.17, every cubic
bridgeless planar graph is of Class one.
28. (a) Proof. Assume first that G has a Tait coloring, using the colors 1, 2,
3. We may assume that the edge uv is colored 1. Consider the edge
coloring of H in which every edge of H that belongs to G is colored the
same as in G and where the structure replacing uv is colored as shown
in Figure 17.6. This produces a Tait coloring of H.
Next assume that H has a Tait coloring, using the colors 1, 2, 3. We
may assume that the edge xy is colored 1 in H. Then either (1) wx and
yz are both colored 2 and wy and xz are both colored 3 or (2) wx and
yz are both colored 3 and wy and xz are both colored 2, say (1) holds.
216 CHAPTER 17. EDGE COLORINGS

u
.............
..
1
.....
.... ... w
...... ...............
2................ ......
......3
.. ......
.....
............. 1 ...........
x ........
...... .
.......
. .
.......... y
......
...... ........
.
...... ... ......... 2
3 ....... .....
...... z
1
.............
...
v

Figure 17.6: The graph in Exercise 28(a)

Then uw and vz are colored 1. Replacing this structure by the edge

uv and coloring this edge the color 1 produces the graph G with a Tait
coloring.
(b) By applying the result in (a) three times, the graph F is transformed
into the graph C3  K2 . Since C3  K2 is a bridgeless cubic planar
graph, it follows by Corollary 17.17 that C3  K2 is of Class one and so
has a Tait coloring. By the result in (a), F has a Tait coloring. (Note:
Since F itself is a bridgeless cubic planar graph, F is of Class one and so
has a Tait coloring.)

29. Let G be the cubic graph shown in Figure 17.7.

s
qqqqq..............
qqqqqqqqqqqqqq ....... ............................................3......
qqqqqqqqqqqqqqqqq
1
qqqqqqqqqqqqqqq
qqqq 2...... ..............
......q.......qqqq
..............
..q.. .....
qq q
2qqqqqq qqqqq 3 q qqqq ...........
.. ..
qq qqq .........3
1 ... .....
. .. 1..................................2
qqqqqqqq qqq...... qq
.
......... . ..... .....
.............
qqqq1qq 1qqqqqq..q........ ................ ... .. ............... .........
... ... ..... 2 2 ............... ... ...... 3 3......................
qqqqqqqqqq
......
...... ........... ......
. .
G: qq qq
. .
3 ................................
. ...
........... ....

qqqqqq qqqqqqqq..q.......
3 2 1 2 .......... 1
...... ...........
.............qq
..... ...... .....
.....qqq
............
.............. ...............
q
.............
... qq
qqqqqq qqqq q qq....... .
.
q qqqq ........ .
.
........ .
. .......... ..
... .
.........

1 qq.q..................... 3
... . ..

2 q................qqqqqqqqqqq3
. .
...... .
. ......
1 ...................................... 2
qqqqqqqqqqqqqqqqqqqq .............
qqqqqqqqqqqqqqqqqqqq 2 .........
...
...
...
..
qqqqqqqq......................................................3
1 ....
t

Figure 17.7: The graph in Exercise 29

(a) The graph G is not planar as it contains a subdivision of K3,3 (shown

with bold edges in Figure 17.7).
(b) Let S = {s, t}. Since k(G − S) > |S|, it follows by Theorem 6.11 that G
is not Hamiltonian.
(c) The graph G has a Tait coloring. Such an edge coloring is shown in
Figure 17.7.

30. The statement is true.

 217

Proof. Assume first that G has a Tait coloring, using the colors 1, 2, 3.
Consider some edge uv of G and suppose that uv is colored 1. Now consider
the edge coloring of the structure replacing uv of G in the formation of H
shown in Figure 17.8(a). This produces a Tait coloring of H.

u u
.............. .............
.. ...

1 1
......... s
.........................
.......
..... .....
s
2 ...... 3 2 ...... ... ........... 3
........... ......
...... ............. ......
......
... ...........
.......
.
.............. x .................
............
w ........... 1
.......
.......
1
...
.. ........ ....
. .
............. w ... .......
........ . ...
..
.... ..
....... ..... x
....... ............. ... ........ ...
...
.....
.............
3 .......
.
.................. ...
..
2 .... ...
.... ............
.......
.......
....... .... ........ ........
.
... ........ ...............
.....
.... ..... .. .............
y ...................
y ........
.......
.. .
........... z . .......
....... .... ..
...... ..... z
....... ...... ....... .......
.............. ........... ....... ..... ...... ....
2 ... .....
..... 3 ...... ....
......
t t
1
...
............. ...............
.

v v
(a) (b)

Figure 17.8: The graphs in Exercise 30

Next, assume that H has a Tait coloring c, using the colors 1, 2, 3. Consider
the structure (Figure 17.8(b)) that replaces some edge uv of G. We may
assume that us has been colored 1 and that sw is colored 2 and sx is colored
3. Then one of wy and wz is colored 1 and the other is colored 3. Similarly,
one of xy and xz is colored 1 and the other is colored 2. If c(wz) = 1, then
c(xy) = 1; while if c(wy) = 1, then c(xz) = 1. In either case, c(tv) = 1 and
so us and tz are both colored 1. Replacing this structure by uv for each edge
uv of G and coloring the edge uv with the same color as us and tv produces
the graph G with a Tait coloring.

31. (a) Since the graph G is a cubic graph, either χ0 (G) = 3 or χ0 (G) = 4. If G
has a Tait coloring, then the triangle in G can be contracted to a single
vertex (see Figure 17.9), producing a Tait coloring of the Petersen graph
P . Since χ0 (P ) = 4, this is impossible. Thus, χ0 (G) = 4.
..
.....
.....
2............... .....
.... 2 .....
..... .....
............... .....
.......
.... .......
. .....
. ..
..
...
... ........... ................
1 ...
...
... 3 ...... .... ......
... ....... 1
... ... ... .......
.
. ... ... ... .......
. ... ...
........... ................
.......
... ... ...... 1
.......
3 ...
...
...
... 2 .......
.....
...
...
... .
...
3 ...
...
...
.

Figure 17.9: The graphs in Exercise 31

218 CHAPTER 17. EDGE COLORINGS

(b) The graph G is not a snark. Even though G is a cubic graph of Class
two, its girth is not at least 5.
Chapter 18

Nowhere-Zero Flows and List

Edge Colorings

Section 18.1. Nowhere-Zero Flows

1. See Figure 18.1.

−1 1 1
............. ...........
...... ............. ............. ................... ............. ............ ..........
...... .........
....
...
... ......
...
... ...... ........
...
...
.
.............. ... ....... ...
....
........... ....... .......... .......
......... . . ...... .. ... ..... .............
−1 ........
...
... ....
.......
−1 1 ...
...
... ....
.......
. 1 1 ...
... ....
.........
.
−1
........... ..........
.. ... ........ ............
........... ...
... ...
.
. .
... .....
.
... .....
. ... ...... ...
−1 .
..
.............
......
........... −1 1 . .
...........
.
.
..
.
.
.
. ............
...... 1 −1 ..
.....
....
...........
......−1
..... .... ... . ..
..
.
..........
.
. ...
... ... .. ... ...
. ..... .. ............
............ ........... ..... ...... ............ ........... ...
.............. ........... ...
..............
..... ...... ... ..... ...... ...... ......
−1 −1 1
(a) D1 (b) D2 (c) D3

Figure 18.1: The orientations D1 , D2 and D3 in Exercise 1

2. Proof. Let G be a graph having a nowhere-zero k-flow for some integer

k ≥ 2. By Theorem 18.3, there is an orientation D of G and a nowhere-zero
k-flow φ on D all of whose flow values are positive. Suppose that f = (x, y) is
an arc of D that belongs to E2 . Proceeding as in the proof of Theorem 18.2, we
let D0 be the orientation of G obtained by reversing the direction of f , resulting
in the arc f 0 = (y, x). Then the function φ0 : E(D0 ) → {±1, ±2, . . . , ±(k − 1)}
defined by

if e 6= f 0

0 φ(e)
φ (e) =
−φ(f ) if e = f 0

is a nowhere-zero k-flow on D0 . Furthermore, φ0 (e) > 0 if e 6= f 0 and φ0 (f 0 ) <

0. Continuing this procedure for all other arcs in E2 gives the desired result.

219
220CHAPTER 18. NOWHERE-ZERO FLOWS AND LIST EDGE COLORINGS

3. Proof. Suppose that φ1 and φ2 are flows on an orientation D of a graph G.

For integers a and b, consider the function aφ1 + bφ2 : E(D) → Z. Since φ1
and φ2 are flows on D,

σ1+ (v; φ1 ) = σ1− (v; φ1 ) and σ2+ (v; φ2 ) = σ2− (v; φ2 )

for each vertex v of D. We show that

σ + (v; aφ1 + bφ2 ) = σ − (v; aφ1 + bφ2 ).

Thus, for v ∈ V (D),

X
σ + (v; aφ1 + bφ2 ) = (aφ1 + bφ2 )(v, w)
(v,w)∈E(D)
X
= (aφ1 (v, w) + bφ2 (v, w))
(v,w)∈E(D)
X X
= a φ1 (v, w) + b φ2 (v, w)
(v,w)∈E(D) (v,w)∈E(D)

= aσ1+ (v; φ1 ) + bσ2+ (v; φ2 )

= aσ1− (v; φ1 ) + bσ2− (v; φ2 )

X X
= a φ1 (w, v) + b φ2 (w, v)
(w,v)∈E(D) (w,v)∈E(D)
X
= (aφ1 (w, v) + bφ2 (w, v))
(w,v)∈E(D)
X
= (aφ1 + bφ2 )(w, v)
(w,v)∈E(D)

= σ − (v; aφ1 + bφ2 )

and so aφ1 + bφ2 is a flow on D.

4. Since the genus of the Petersen graph P is 1, it follows that P can be embedded
on the torus. Since the order of P is 10 and the size of P is 15, it follows
from the Generalized Euler Identity (Theorem 11.11) that n − m + r = 0
and so 10 − 15 + r = 0. Thus, r = 5. An embedding of P on the torus is
shown in Figure 18.2. Observe that each of the five regions is adjacent to the
other four. Let c be a 5-region coloring that assigns the colors 1, 2, 3, 4, 5
to these five regions (see Figure 18.2). Let D be an orientation of P . If e is
an arc of D and the region R lies to the right of e and region R0 to the left
of e, then define φ(e) = c(R) − c(R0 ). Then φ(e) ∈ {±1, ±2, ±3, ±4}. Let
v be a vertex of P . Suppose that we encounter the edges vv1 , vv2 , vv3 as we
proceed cyclically about v. For i = 1, 2, 3, let Ri be the region having vi and
vi+1 on the boundary, where v4 = v1 , and suppose that Ri is colored ci . If
 221

all arcs incident with v are directed away from v or directed towards v, then
σ + (v; φ) = σ − (v; φ) = 0. If exactly one arc, say (v, v1 ), is directed away from
v, then σ + (v; φ) = c1 −c3 = σ − (v; φ); while if exactly two arcs, say (v, v1 ) and
(v, v2 ), are directed away from v, then σ + (v; φ) = c2 − c3 = σ − (v, φ). Thus,
φ is a nowhere-zero 5-flow.
... ... ...........
... ... ......
... ...
...
...
... a 5 ....
. ... c
. ... .......... .
........ ... ............................. ..... 4 ...
......... 4 ... ........ ........ .. ......
. ....... ........... ........
. ................. ... ...
.......
.... ... .....
.......
. .....
... .....
b .
.. ...
.
.... ....
..
. . .
.
1 2
......
..............
b
..... .
... .......
. ..
... .... .. .
................ ....
... ......
. ....... .
.
... . ... ....... .
.
... ..... ..
.. ........... .......
....... ........
................
....... 3 .... ..
. . 5
..... ..........
5 .....
..... .....
.. ..
..

.....
.....
............
... ..
.......
... ....
................ a
.....
.....
..... a
.....
.....
..... c
4 .....
.....
.....
.....
.....
..... ......
4
. ...........

Figure 18.2: An embedding of the Petersen graph on the torus

5. Proof. Let G be a bridgeless graph containing an Eulerian trail and let u

and v be the two odd vertices of G. Since G is bridgeless and G is connected,
G is 2-edge-connected. By Theorem 4.18, G contains two edge-disjoint u − v
paths P and P 0 . Direct the edges of P from u to v and assign each of these
arcs the flow value 1. Direct the edges of P 0 from v to u and assign each arc
of P 0 the flow value 2. Remove the edges of P and P 0 from G, resulting in a
graph H, each vertex of which has even degree except for u and v. Since u and
v are the only odd vertices of H, they belong to the same component H1 of H.
Thus, H1 has an Eulerian trail T . Direct the edges of T from u to v, assigning
each resulting arc the flow value 1. Every other nontrivial component H 0 of
H is Eulerian. Direct the edges of H in the direction of an Eulerian circuit of
H 0 and assign each resulting arc the flow value 1. The resulting function just
defined on the arcs of the orientation D of G is a nowhere-zero 3-flow.

6. Yes. The graph K3  K2 has a cycle double cover with three Hamiltonian
cycles. See Figure 18.3.

7. Let V (K6 ) = {1, 2, 3, 4, 5, 6}. A cycle double cover of K6 consisting of five

Hamiltonian cycles is
(1,2,3,4,5,6,1), (1,3,6,2,5,4,1), (1,4,2,6,5,3,1), (1,5,2,3,4,6,1), (1,2,4,6,3,5,1).

8. (a) Yes. Suppose that V (K4 ) = {1, 2, 3, 4}. Then

(1, 2, 3, 1), (1, 2, 4, 1), (1, 3, 4, 1), (2, 3, 4, 2)
is such a cycle double cover of K4 .
(b) No. Since the size of K5 is 10 and 2 · 10 = 20 is not a multiple of 3, no
such cycle decomposition is possible.
222CHAPTER 18. NOWHERE-ZERO FLOWS AND LIST EDGE COLORINGS

.................
.......... . ..
.............. ...
...
... ...
....
. ... ...
...
... ... ...
...
..... ... ...
...
.. .. ...
... ...
. ... ............. ... ............ ...
...
K3  K : 2 ........
..
.. .......
.. ....
...
...
...
...
......
...
...
...
...
.... .... . ... ... ...
...
... ... ... ... ... ...
.
.. .
.. .
... ... ... ...
.. .. .... ... ... ...
... ..
. . . ... ........ ...
.
.. .
.
......... . .
.. ....
. ... ........ ... .... ...
.. .
..........
. .
.... ......... ... ........ .. ...
... ..
.. .. ......... ..... ........ .. ...
..
.
.. ............. ..
. ..
.......... ... .
...
..........
. ...
...
. ...
................... ............... ....... ..
...... ... ...
... ................. .............
... ..

........ .......
........... ............
... ... .....
... .. ... ...
..... ...
. ...
...
...
.. ... ...
... ..... ...
...
... ............. ... .......... ...
..... ...... .
. .. .
.
.......... ...
... ... .
... .. .....
. ...
...
.
.... ..... .... ..... ...
... ...
..
... ... .... ... . ...
... ... .. .
.. ... ...
... ... .
..
. .
..
. .... ...
... ... .
..
. .. .
......... ...
.. ..........
. ................. .. .........
. . .
...........
...
...
... ...... ... ........ ... ...
.......... .. ...
... ........ ...
........ .
........
........ .....
.
... ........ ..
.. ........... ...
. .
...
... ........ .. . . .. .
......... ........
............. ....... ...
....... ................... .... ...
.....
... ...

Figure 18.3: A cycle double cover of K3  K2 consisting of three Hamiltonian cycles

(c) Yes. Suppose that V (K5 ) = {1, 2, 3, 4, 5}. Then

(1, 2, 3, 4, 1), (1, 2, 4, 5, 1), (1, 3, 2, 5, 1), (1, 3, 5, 4, 1), (2, 3, 4, 5, 2)
is such a cycle double cover of K4 .

Section 18.2. List Edge Colorings

9. Proof. Let ∆(G) = ∆ and for each edge e of G, let L(e) be a list of 2∆ − 1
colors (positive integers). Let e1 , e2 , . . . , em be an ordering of the edges of
G. Assign the smallest color in L(e1 ) to e1 . Suppose that colors have been
assigned to the edges e1 , e2 , . . . , ei from their respective color lists for 1 ≤ i <
m. We now assign to ei+1 the smallest color in L(ei+1 ) not assigned to any
of the edges e1 , e2 , . . . , ei that are adjacent to ei+1 . Since ei+1 is adjacent to
at most 2(∆ − 1) edges among e1 , e2 , . . . , ei and |L(ei+1 )| = 2∆ − 1, at least
one color from L(ei+1 ) is available for ei+1 . Thus, χ0` (G) ≤ 2∆(G) − 1.
10. k = 21. Since e1 , e2 and e3 must be assigned distinct colors, we must have
(1) |L(er )| ≥ 3 for some r (1 ≤ r ≤ 3),
(2) |L(es )| ≥ 2 for some s (1 ≤ s ≤ 3 and s 6= r) and
(3) |L(et )| ≥ 1 for some t (1 ≤ t ≤ 3 and t 6= r, s).
The same is true for e7 , e8 and e9 . Since it can occur that
" 3 # " 9 #
[ \ [
L(ei ) L(ei ) = ∅,
i=1 i=7

it follows that we must have |L(ei )| ≥ 3 for i = 4, 5, 6. Thus, k ≥ 21. On the

other hand, any lists L(ei ) with |L(e1 )| = |L(e9 )| = 1, |L(e2 )| = |L(e8 )| = 2
and |L(ei )| = 3 for 3 ≤ i ≤ 7 have the desired property and so k = 21.
 223

Section 18.3. Total Colorings

11. Since G is a complete 3-partite graph, χ(G) = 3. Since ∆(G) = 4, it follows
that χ0 (G) ≥ 4. Since there is a 4-edge coloring of G shown in Figure 18.4(a),
it follows that χ0 (G) = 4. The 5-total coloring in Figure 18.4(b) and the fact
that χ00 (G) ≥ 1 + ∆(G) = 5 implies that χ00 (G) = 5.

...
1 ...... ........... 1 ............
................... ... . .... ...... ... .....
................. ................. ..... 2 .... 3 .
... ...... ......... ........ ..... ... .
.... .............. ..
... ...... ......... ......... .......... ..... . ............................
... ...... ... 2 3 ........... ..
.... ..
........... .... ........
... ...... .......... ........
.... .
.
.... ...
... ...... ............... ........ ..... .. ........ 5 ........ ..... .....
... ..
... ..........
........
........ ........
........
.....
..... .....
... .....
... .... . .. ......... 4 ........
.... .... ...
.
.... ..
.......................... 4 ..... ... ... ..... ........
........ ........ ..... .....
... .....
..... . . ...
. ... ... ..
... ......... ........ ................. ............... ..... .
... ..... .. ..
............. ..... ... ......... ......... 4 ..... .....
... 3 ..... 4 . 1 ...... .. ... ..... 5 ...... .
...
... .....
..... .....
. . .
...
... ........
....
. ..
.
. ...
...
.....
.....
... 1
..... ......
.
.
.....
..
....
... ......... ... ... ..... . .......... ... .....
.
.. .. .
.........
.
............... ... ..... 2 ... ... .... ...
...
... ... ... ..
... ...
... .....
..... ...........
... ...3 ..... ...
...
... ... ... .. ... ... ... ..... ...
4 ... 2 ... .. ... ........ ...... ...... ................ .
... ...
... ... ...
... 2 ... .
.....
... 1 .
.....
. ... .
...
.
.
...
... ... .. ..
...
... 3
.... .....
. 2
.... .....
.
...
... ... ..
. ... ... ........... .......... ... 2
...
1 ... .
...
3 ..
.
. 3 ... ... ..
. ....
... ... ... ... ... . .
..
.
... . ... ... ... ... ... ...
... .... ... .. ... ... ... ...
... .. ... ..... ... ... 4 5 ... ...
... ... .. ...
... ... ... ...
...
... ... ..
. .
....
... .. .. ... ... .. ... ...
... ... .. ... ... ..... ... ...
......... ... ... .. ...
.......... ... ... ... ...
....... ... ... .... .....
... ................. ...
...... ... ...
... .
... 1 .....
....... .........
(a) ..

(b)

Figure 18.4: A 5-total coloring of K2,2,2,

12. Since α(G) = 2, it follows that χ(G) ≥ n/α(G) = 3.5 and so χ(G) ≥ 4.
Since there is a 4-coloring of G, it follows that χ(G) = 4. Since 14 = m >
bn/2c · ∆(G) = 4 · 3 = 12, G is an overfull graph and so χ0 (G) = 1 + ∆(G) = 5.
First, observe that χ00 (G) ≥ 1 + ∆(G) = 5 and |V (G) ∪ E(G)| = 21. If
χ00 (G) = 5, then there exists a subset S ⊆ V (G) ∪ E(G) with |S| ≥ 21/5
consisting of an independent set of vertices and an independent set of edges
such that no vertex of S is incident with an edge of S. Thus, |S| ≥ 5. Since
α(G) = 2 and α0 (G) = 3, such a set S must consist of two nonadjacent vertices
and an independent set of three edges. However, since n = 7, one of these
two vertices must be incident with one of the three edges, which is impossible.
Hence, χ00 (G) ≥ 6. Since there is a 6-total coloring of G shown in Figure 18.5,
it follows that χ00 (G) = 6. Thus χ(G) = 4, χ0 (G) = 5 and χ00 (G) = 6.
224CHAPTER 18. NOWHERE-ZERO FLOWS AND LIST EDGE COLORINGS

1
.......
...... ......
2....................................................................5..
..... ... ..... .......
........ ..... ..... .......
5 ....................... .....
.... .....
.....
...........
.....
. 2
. .. .
..
..
. ...... ... ....
.... .... ......... 6 ..... ... ..
. .. .... ..... ... ....
.
.... ............. 3 4 ..... .. ...
4 .. .......
....... .
.....
...1
..
... .... ... ...... ..
.. ...... .... .. .... ..
............. ... .. ........ .....
G: 2
.
..
.... .........
..... ....... ..
... 3 4 .
.
.
.
.
.
..........
.. .
.... ....... 5
... .......... ... ......... ....
... ......... ........ ...
...
... ... ........... 5 . 2 ... .............. ...
........ .
... ....
... ... ........ .............. ..
.. .....
...
1 ... ..
... ...
..
.
..
..
..
.
..
... ...
....... .............. .... .....
.
. . 3
... . ............. ............ ..
................. ..... ......
. ...
4 6 1

Figure 18.5: A 4-total coloring of the graph in Exercise 12

13. Since ∆(G) = 3, it follows that χ00 (G) ≥ 4. Since there is a 4-total coloring
of G, shown in Figure 18.6, it follows that χ00 (G) = 4.
2
.......... ........ .......
.............
3 .........
........................ 1
......
4.......................................2 3.................. .............1......
.. .... .. ......
2 ........ ......
. ... ......
....... . . .. ... .
.......... ..
.
.................
. .............. 1 ............
... ......
4 ..........................2 4 .... ..
3................ 4.. ...... 3 .. .
..........
..........
....... 4 3 ....................... 4
... ....... ...... ....... ....... ... ....
..... .......
..... .....................
.. 3 ...
.......
.......................
.
... 3 4 .......
... . .
..
... ....
..... ....... 1 ........ 4 2 ...................... ......
... ..
..... .....
.....
..... 1 .....
...... .....
...... .... .....
...... ......
...... ......
....... .......
........
........ ..
...
.........
.......... .
1 ............ .......... 2
.................... ......................................................
...............

Figure 18.6: The graph in Exercise 13

14. Proof. Let G = Kn . Since ∆(G) = n − 1, it follows that χ00 (G) ≥ (n − 1) +

1 = n. Suppose first that n is odd. Then n = 2k + 1 for some positive integer
k. Let V (G) = {v1 , v2 , . . . , vn }. For j = 1, 2, . . . , n, let

Xj = {vj } ∪ {vj+i vj−i : 1 ≤ i ≤ k},

where the subscripts j + i and j − i are expressed as integers modulo n. By

assigning the color j to each element of Xj , we have an n-total coloring of G
and so χ00 (G) ≤ n. Thus χ00 (G) = n if n is odd.
Next suppose that n is even. Now |V (G) ∪ E(G)| = n + n2 = (n2 +

n)/2. Suppose that χ00 (G) = `. Then V (G) ∪ E(G) can be partitioned as
{X1 , X2 , . . . , X` } where each set Xi (1 ≤ i ≤ `) consists of an independent
set of vertices and an independent set of edges such that no vertex of Xi is
incident with an edge of Xi . Thus, |Xi | ≤ n/2. Hence

(n2 + n)/2
χ00 (G) ≥ = n + 1.
n/2
Let H be the graph obtained by adding a new vertex v to G and joining v
to every vertex of G. Thus, H = Kn+1 . Since n + 1 is odd, χ00 (H) = n + 1,
 225

as we showed above. Hence, χ00 (G) ≤ χ00 (H) = n + 1 and so χ00 (G) = n + 1
when n is even.
15. Proof. Let G = Ks,t , where 1 ≤ s < t. Since ∆(G) = t, it follows
that χ00 (G) ≥ t + 1. Let the partite sets of G be U and W , where U =
{u1 , u2 , . . . , us } and W = {w1 , w2 , . . . , wt }. For 1 ≤ j ≤ t, let

Ej = {ui wi+j−1 : i = 1, 2, . . . , s},

where i + j − 1 is expressed modulo t. For 1 ≤ j ≤ t, let Xj = Ej ∪ {ws+j },

where s + j is expressed modulo t, and let Xt+1 = U . By assigning each
element of Xi (1 ≤ i ≤ t + 1) the color i, we have a (t + 1)-total coloring of
G. Thus, χ00 (G) ≤ t + 1 and so χ00 (Ks,t ) = t + 1.
16. We may assume that G is a nontrivial connected graph with ∆(G) ≤ 2. We
show that χ00 (G) ≤ 2 + ∆(G). Since ∆(G) ≤ 2, it follows that G = Pn (n ≥ 2)
or G = Cn (n ≥ 3). First, suppose that G = Pn = (v1 , v2 , . . . , vn ). Define
the 3-total coloring c of G by c(v1 ) = 1, c(v1 v2 ) = 2, c(v2 ) = 3, c(v2 v3 ) = 1,
c(v3 ) = 2, c(v3 v4 ) = 3 and so on (see Figure 18.7 for n = 7). Thus, χ00 (G) ≤ 3.
In fact, χ00 (Pn ) = 3 for n ≥ 2.
1 2 3 1 2 3 1 2 3 1 2 3 1
.............. ............. ............. ............. ............. ............. .............
. ... ... ... ... ... ...

Figure 18.7: A 3-total coloring of P7 in Exercise 16

Next, suppose that G = Cn = (v1 , v2 , . . . , vn , v1 ). Then ∆(G) = 2 and

G − v1 vn = Pn . Define the 4-total coloring c0 of G such that c0 (v1 vn ) = 4 and
the restriction of c0 to Pn is c. Thus, χ00 (G) ≤ 4.
226CHAPTER 18. NOWHERE-ZERO FLOWS AND LIST EDGE COLORINGS
Chapter 19

Extremal Graph Theory

Section 19.1. Turán’s Theorem

1. T7,2 = K3,4 , t7,2 = 12
T7,3 = K2,2,3 , t7,3 = 16
T7,4 = K1,2,2,2 , t7,4 = 18
T8,2 = K4,4 , t8,2 = 16
T8,3 = K2,3,3 , t8,3 = 21
T8,4 = K2,2,2,2 , t8,4 = 24

2. (a) The minimum size m of such a graph is (4k + 1)2 .

The graph K4k,4k+2 has order n = 8k + 2, independence number 4k + 2,
size 4k(4k +2) = 162 +8k and is bipartite. Therefore, m ≥ 162 +8k +1 =
(4k + 1)2 . Let G be a graph of order n = 8k + 2 ≥ 10, size (4k + 1)2 and
independence number α(G) = 4k+2. Assume, to the contrary, that G has
no triangle. Let U be a maximum independent set and W = V (G) − U .
So, |U | = 4k + 2 and |W | = 4k. Let U 0 ⊆ U such that |U 0 | = 4k. Let
G0 = G[U 0 ∪ W ]. Thus, G0 has order 8k. By Theorem 1.9, if G0 has more
than 16k 2 edges, then G0 (and so G) contains a triangle. So we may
assume that G0 contains at most 16k 2 edges. The remaining edges of G
belong to [U − U 0 , W ]. Since there are at most 8k edges in [U − U 0 , W ],
it follows that G has at most 16k 2 + 8k edges, a contradiction.
(b) The minimum size m of such a graph is (4k + 1)2 − a2 + 1.
The complete bipartite graph K4k+1−a,4k+1+a has order n = 8k + 2,
independence number 4k+1+a, size (4k+1−a)(4k+1+a) = (4k+1)2 −a2
and is bipartite. Therefore, m ≥ (4k + 1)2 − a2 + 1. Let G be a graph
of order n = 8k + 2 ≥ 10, size (4k + 1)2 − a2 + 1 and independence
number α(G) = 4k + 1 + a. Assume, to the contrary, that G has no
triangle. Let U be a maximum independent set and W = V (G) − U .
So |U | = 4k + 1 + a and |W | = 4k + 1 − a. Let U 0 ⊆ U such that

227
228 CHAPTER 19. EXTREMAL GRAPH THEORY

|U 0 | = 4k + 1 − a. Let G0 = G[U 0 ∪ W ]. Thus, G0 has order 8k + 2 − 2a.

If G0 has more than (4k + 1 − a)2 edges, then G0 (and so G) contains a
triangle. So we may assume that G0 contains at most (4k + 1 − a)2 edges.
The remaining edges of G belong to [U − U 0 , W ]. Since there are at most
(2a)(4k + 1 − a) edges in [U − U 0 , W ], it follows that G has at most
(4k + 1 − a)2 + (2a)(4k + 1 − a) = (4k + 1 − a)(4k + 1 + a)
= (4k + 1)2 − a2
edges, which is a contradiction.

Section 19.2. Extremal Subgraphs

3. Solution. m = 13. First, observe that the graph H of order 6 and size 12
obtained by removing the edges of a triangle from K6 has three vertices of
degree 5 and three vertices of degree 3. Since H has less than five vertices of
degree at least 4, it follows that H does not contain a subdivision of K5 and
so m ≥ 13. There are two graphs of order 6 and size 13, both of which are
shown in Figure 19.1 (with the missing edges denoted by dotted lines) and
each of which contains a subdivision of K5 . So m = 13.
......
........ . . . . . . . . ..............
.............. ........... . . . . . . . . ................
... ...... ..................... ........ .... . .
..... ... ......... ......... .... ...... ... ........ ............................ ....... .... ..
.... ... ............... ....
.... ..................... ....................... .....
. ... .
...... ........ . .
...
... ............. ... .... ......... ... ... .......... ......... ........ ................... ..
... .
.
..... .........
.... ... .
. .............. ............... .. ... . ...... ....
........... .... ... ... ........................................................................................................................
.... ......... ....... ............ ..... .........
..... ........ .... ..... .........................
.. ..
.... ......... ....... ..... .............. ..... ..... .............
. .
.. ... . .
.... ........... ........... ..
.... . . . ... ..... ....... ............ .. ...
. .. .
.
.... ..... ....................... .... ....
........... ........
.
...................
. ..... .............................................................
... ...... . .... .... ...... ...
.... . . . . . . . . . ....... ..... .......

Figure 19.1: Graphs in Exercise 3

4. For k = 3 and n = 4, it follows that (k − 1)n − k2 = 2 · 4 − 32 = 5. Every

graph of order 4 and size 5 (namely the graph K4 − e) contains both P4 and
K1,3 as subgraphs, the only two trees of size 3.
5. Proof. Let G be a graph of order n ≥ 3 and size bn/2c + 1. We need
only show that G contains a vertex of degree 2 or more. Let V (G) =
{v1 , v2 , · · · , vn }. Then
n
X
deg vi = 2(bn/2c + 1) ≥ 2[(n − 1)/2] + 2 = n + 1.
i=1

Assume, to the contrary, that deg vi ≤ 1 for all i (1 ≤ i ≤ n). Then

n
X
deg vi ≤ n · 1 = n.
i=1

This is a contradiction.
Graphs of order n and size bn/2c not containing P3 as a subgraph are
(n/2)K2 if n is even and [(n − 1)/2]K2 + K1 if n is odd, both of which have
size bn/2c.
 229

6. Proof. Let G be a graph of order n ≥ 4 and size n. We show that G contains

2K2 as a subgraph. Since the size of G is n, it follows that G contains a cycle.
If G contains a cycle of length 4 or more, then G contains 2K2 as a subgraph.
Hence, we may assume that the only cycle that G contains is a triangle.
However, since n ≥ 4, it follows that G contains at least one more edge e,
which is necessarily not adjacent to at least one edge of the triangle and so G
contain 2K2 as a subgraph.
Graphs of order n and size n − 1 not containing 2K2 as a subgraph are K1,3
and K3 + K1 if n = 4; while for n ≥ 5, the star of order n is the only graph.
7. Proof. For n ≥ 4, the smallest integer m such that every graph of order n
and size m contains K1,3 as a subgraph is n + 1. Let G be a graph of order
n ≥ 4 and size n + 1. We show that G has a vertex of degree 3 or more. Let
V (G) = {v1 , v2 , · · · , vn }. Then
n
X
deg vi = 2(n + 1) = 2n + 2.
i=1
Pn
If deg vi ≤ 2 for all i (1 ≤ i ≤ n), then i=1 deg vi ≤ 2n, which is impossible.
Graphs of order n and size n not containing K1,3 as a subgraph are 2-regular
graphs, namely cycles or a union of cycles.
8. Proof. We show that every graph of order n and size 2n − 2 contains a
subdivision of K4 . We proceed by induction on n. For n = 4, 2n − 2 = 6 and
G = K4 . Hence, the statement is true for n = 4. Assume that every graph of
order n − 1 (≥ 4) and size 2(n − 1) − 2 = 2n − 4 contains a subdivision of K4 .
Let G be a graph of order n and size 2n − 2. Since
X
deg v = 4n − 4,
v∈V (G)

there is a vertex v in G such that deg v ≤ 3.

Assume first that deg v ≤ 2. Let H = G − v. Then H has order n − 1 and
size at least 2n − 2 − 2 = 2(n − 1) − 2 and so H contains a subdivision of K4 ,
as does G.
Hence, we may assume that deg v = 3. Let N (v) = {x, y, z}. If G[N [v]]
contains 6 edges, then G[N [v]] = K4 and so G contains K4 . So we may assume
that G[N [v]] contains at most five edges. Hence, G[N (v)] contains at most
two edges. We may assume that xy ∈ / E(G). Define G0 = G + xy − v. Then
G has order n − 1 and size at least 2n − 4 = 2(n − 1) − 2. By induction, G0
0

contains a subdivision of K4 and so G contains a subdivision of K4 .

9. (a) Proof. Let e = uv be an edge of G, let A = N (u) − {v} and let
B = N (v) − {u}. Then
n−2 ≥ |A ∪ B| = |A| + |B| − |A ∩ B|
 
2n − 2 4n − 4
≥ 2 − |A ∩ B| = − |A ∩ B|
3 3
230 CHAPTER 19. EXTREMAL GRAPH THEORY

and so |A∩B| ≥ (n+2)/3. Note that A∩B is not an independent set; for
otherwise, let w ∈ A ∩ B. Then deg w ≤ n − |A ∩ B| ≤ n − n+2
3 =
2n−2
3 ,
which is impossible. Let x, y ∈ A ∩ B such that xy ∈ E(G). Then
G[{u, v, x, y}] = K4 .
2n
(b) For an integer r ≥ 3, let G = Kr,r,r . Then deg v ≥ 3 = 2r but G does
not contain a complete subgraph of order 4.

10. Proof.
Assume, to the contrary, that there is a graph G of order n ≥ 3 and
size n−1
2 + 2 such that G contains a vertex v that does not lie on any triangle
in G. First, we claim that δ(G) ≥ 2; for otherwise, there is a vertex u of G
such
that degG u ≤ 1. Then the graph G − u has order n − 1 and size at least
n−1
2 + 1, which is impossible. Now, suppose that degG v = k ≥ 2. Since
v does not lie on a triangle in G, it follows that N (v) is an independent set
of vertices
in G.
The graph G0 = G − v has order n − 1 and size m0 . Then
m0 ≤ n−1 2 − k2 . Since
     
n−1 0 n−1 k
+2=m=m +k ≤ − + k,
2 2 2

2
it follows that k2 − k + 2 ≤ 0. However then, k 2 − 3k + 4 = k − 23 + 74 ≤ 0,

which is impossible.
On the other hand, let H be the graph obtained from the complete graph
Kn−1 by adding a new vertex v and

joining v to exactly one vertex of Kn−1 .
Then H has order n and size n−1
2 + 1 but the end-vertex v of H does not lie
on any triangle in H.

11. The minimum positive integer m such that every graph of order n and size m
contains two edge-disjoint triangles is bn2 /4c + 2.
Proof. Since the graph Kb n c,d n e + e does not contain two edge-disjoint
2 2
triangles, it follows that m ≥ bn2 /4c + 2. Every graph of order n ∈ {5, 6}
and size bn2 /4c + 2 contains two edge-disjoint triangles. Suppose that the
statement is false. Then there is a graph G of minimum order n and size
bn2 /4c + 2 that does not contain two edge-disjoint triangles. Thus n ≥ 7. We
consider two cases.
Case 1. n is even. So, n = 2k ≥ 8. Since
X
deg v = 2m = 2bn2 /4c + 4 = 2k 2 + 4,
v∈V (G)

it follows that G has size m = k 2 +2 and contains a vertex v of degree k or less.

Let H = G − v. Then H has order n − 1 = 2k − 1 and size at least k 2 − k + 2.
Since
b(n − 1)2 /4c + 2 = b(4k 2 − 4k + 1)/4c + 2 = k 2 − k + 2,
it follows that H contains two edge-disjoint triangles and so does G.
 231

Case 2. n is odd. So, n = 2k + 1 ≥ 7. Since

X
deg v = 2m = 2bn2 /4c + 4 = 2(k 2 + k) + 4 = 2k 2 + 2k + 4,
v∈V (G)

it follows that G has size m = k 2 + k + 2 and contains a vertex v of degree k

or less. Let F = G − v. Then F has order n − 1 = 2k and size at least (k 2 +
k + 2) − k = k 2 + 2. Since

b(n − 1)2 /4c + 2 = k 2 + 2,

it follows that F and therefore G as well contains two edge-disjoint triangles.

12. Proof. Let G be a graph of size m and let H be a 3-partite subgraph of

G of maximum size. We may assume, without loss of generality, that H is a
spanning subgraph of G with partite sets U , W and X. If degH v ≥ 23 degG v
for every vertex v of G, then
X 2 X 4m
2|E(H)| = degH v ≥ degG v =
3 3
v∈V (H) v∈V (G)

and so H contains at least 2m

 
3 edges. Otherwise, there exists some vertex
v ∈ V (H) = V (G) for which degH v < 32 degG v. Suppose that v ∈ U . Thus,
v is adjacent in G to more vertices in U than to vertices in at least one of W
and X, say v is adjacent to more vertices in U than to vertices in W . Consider
the spanning bipartite graph H 0 of G with partite sets U 0 , W 0 and X 0 , where
U 0 = U − {v}, W 0 = W ∪ {v} and X 0 = X, where E(H 0 ) consists of all edges
of G joining any pair of the sets U 0 , V 0 and X 0 . Then since degH v < degH 0 v,
it follows that |E(H 0 )| > |E(H)|, contradicting the choice of H.

Section 19.3. Cages

13. Proof. Let g = g(G). Suppose first that g is even, say g = 2k ≥ 4. Let
C = (v1 , v2 , . . . , v2k , v1 ) be a cycle of length g in G. Then d(v1 , vk+1 ) = k ≤
diam(G). Hence
g(G) = g = 2k ≤ 2 diam(G).
Next, suppose that g is odd, say g = 2k+1 ≥ 3. Let C = (v1 , v2 , . . . , v2k+1 , v1 )
be a cycle of length g in G. Then d(v1 , vk+1 ) = k ≤ diam(G). Hence

g(G) = g = 2k + 1 ≤ 2 diam(G) + 1,

completing the proof.

14. (a) Proof. First, n(3, 6) ≥ M (3, 6) = 14. Since the Heawood graph has
order 14, is 3-regular and has girth 6, it follows that n(3, 6) = 14.
232 CHAPTER 19. EXTREMAL GRAPH THEORY

(b) Proof. Let G be a 3-regular graph of order 14 and girth 6. Let v1 ∈

V (G). Suppose that N (v1 ) = {v2 , v3 , v4 }. Let N (v2 ) = {v1 , v5 , v6 },
N (v3 ) = {v1 , v7 , v8 } and N (v4 ) = {v1 , v9 , v10 }. Let U = {v1 , v2 , . . . , v10 }.
Then G[U ] is the graph shown in Figure 19.2(a). Without loss of gener-
ality, assume that v5 is adjacent to x1 and x2 . Then v6 is forced to be
adjacent to x3 and x4 . Then v7 can be adjacent to one of x1 and x2 and
one of x3 and x4 , say v7 is adjacent to x1 and x4 . Then v8 is forced to
be adjacent to x2 and x3 . Similarly, if v9 is adjacent to x1 and x3 , then
v10 is forced to be adjacent to x2 and x4 . This graph is the Heawood
graph shown in Figure 19.2(b).

v1
.........
....... ......
...... .. .............
....... .......
.
. ...... .......
v1 ....... . ........
.......
...
.. . .......
.......
.......
..... ..... .... .. .......
........ ..... .............. ..... ... .. . .......
........ ........ . . ... ... .
.. ............
.......... ........ . .
.......... .. .
v
......... .
......... v
. .
..
........
.......
.
... ........
........
........ .
... .... 2
. ...
v .
... .... 3
. ...
. .
.
.. ... 4
.. ....
.... .
. . ........ . ... . . ...
...
. .
. . .......... .... ..
. ...
..
. ...
.............. ............. ...
...
...
...
...
......... 2
. v ......... v 3
... ...
...
... ... 4
v ..
..
...
...
..
..
...
...
..
..
...
...
.. ... .. .... .. .. . ... . . . ...
. . . .....
.
...
.
v . .... .
..
.
..
. ...
...
... .
..... ...
...
... ..
.
.
. ...
... v . .....
.
5............
. .
...
........ 6
.
............... ...
. . 7...
..
..
.
v v
.
...... 8
.
.......... . ..
.
..
..... 9
........
v ...
.
........ v10
... ... ... ... ..... ....... ................. . .. ..
. . ............
...
...
... ..
...
... .. ...
........
... ...... ..... ....... ... ..... .
.
.. .... .... .
.... ... ........ ....
.
..
. ... ...
. .
. .
... ... ... ...... ..... .......
. . .
.....
.. ..
.
. ... ......... .
.
.. ..........
.
..
. ... ....... ....... ..... .. ....... ..
... ...... .....
......... ..
..
.. ....
... ... .....
...... ..... ...... ...... .. ... .......... ...
............ .... ..... ........... .
. . .
....... ............ ... ......
..... ... ......... ............ ........................ ..... ... ...... ...
... ... . ..... ... ... ... ..... .. .. ...... ..
... .....
..... ... ..... ........................ ............. ............. ..
v5 v6 v7 v8 v9 v10 ...
... ..... ...
.......
.. .
...... .......
.
......... ...
. . .
...... .. .......... ..
.................. ..
.. .
. .
...
..
... ...... ...... ............ ........... ...... ...
... ... ..... ....... . ...... .......... ..... ......................... ..
...
..... .................. . ... .......... .
. .
.. ....... ...... ...
.
... ... ...... ..... . .. . . ...... ......
... ... ........ ..... .... ........... ......... .... .... ........... ....
........... ..
..... .. ...... ..... ... ..
(a) G[U ] . .. ....
...............
...
...............
.......
..... ....
.......
...........
... ...
...
x1 x2 x3 x4
(b)

Figure 19.2: The graph in Exercise 14(b)

15. Proof. Let G be an (r, g)-cage and let H = G  K2 . Suppose that H is

an (s, g)-graph and assume, to the contrary, that H is an (s, g)-cage. Since
H contains 4-cycles, either g = g(H) = 3 or g = g(H) = 4. It is known that
n(s, 3) = s + 1 and n(s, 4) = 2s.
If g = 3, then the order of G is r + 1 and so the order of H is 2r + 2. If H
is an (r + 1, 3)-cage, then the order of H must be r + 2. Now r + 2 = 2r + 2,
implying that r = 0, which is impossible.
If g = 4, then the order of G is 2r and so the order of H is 4r. If H is
an (r + 1, 4)-cage, then the order of H must be 2(r + 1). Then 4r = 2r + 2,
implying that r = 1, which is impossible.

16. We claim that the maximum girth of a cubic graph of order 12 is 5. The graph
of Figure 19.3(a) is a (3, 5)-graph of order 12. Thus this maximum girth is
at least 5. Assume, to the contrary, that there is a cubic graph of order 12
having girth g ≥ 6. Thus G contains a subgraph shown in Figure 19.3(b). Let
U = {v5 , v6 , . . . , v10 } and W = {v11 , v12 }. Since the size of G is 18, the size
of the subgraph induced by U ∪ W is 9. Since the number of edges in [U, W ]
 233

is either 4 or 6, it follows that G[U ] contains at least three edges, producing

a cycle of length 5 or less, which is a contradiction.

.......
............ ..............
........... ........ ........................
v1
.................... ... . . .........
. . ................ .......................
..
.. ....
. ...
... ......... ...... .......
..... ..... ..... ....... .......
....... .
.
... .....
. .
.....
..... .
...
........ .......
... . ..
....
... ....... ..... .......
.......
.......
............
...
. ....... .....
. .
......... .. . .
....... .......
............................................................................................................................................. . ...
. ....
. .......
.......
.... ... . .
. .
. .. ... ....
...
....... ........ ..........
... ... ....... ......... . .
v
...
.. ...
... ......
... ......
.... ....
...
...
...
... .
..
... .... 2 v .
.
.. ..... 3 v ..........
... ... 4
... .
...... ..... ... ... .... .. ...
. ... ....
.. .. . . . . ..
. ... ... ... ..
. ...
. ........
.............. .
. .... ... ... .
...
... ...
...
...
...
..... ....... .
... .... ... ............... ... ..
... .. ...
... ........ . . .
. . .... ...... .. .... ... ... .. ... .. ...
... ................ ...
... .... ...
... . .. .
... .
. ...
..
........... ....... ..... ... . . ... .
. .. ...
...
..... ............. ....
.. ....... ... ... .
. .. .
............
. .
.............
.
....... ............
...
... ........ ....... ............ ............ ..
....... .. ....... ... .. .... .... ..... ... ........ ....
..............
.........
. ..
.. ......... ........
.. .
.
.........
.....
..................
...
......... ............. .....
.. ..
........ v5 v6 v7 v8 v9 v10
..... ....... .. ....... ....
..... ....... .... ....... ....
..... ............ .
.... ............... .... ....
............ .. ... ...
..............
...........
...
................. ............. .............
......................................... ...
........
v11 v12
(a)
(b)

Figure 19.3: The graphs in Exercise 16

17. (a) The number n(4, 3, 4) is the minimum order of a 4-regular graph contain-
ing an induced connected 3-regular subgraph of order 4. There is only
one 3-regular graph of order 4, namely K4 . Since K5 is a 4-regular graph
containing K4 as an induced subgraph, it follows that n(4, 3, 4) = 5.
(b) The number n(4, 3, 6) is the minimum order of a 4-regular graph contain-
ing an induced connected 3-regular subgraph of order 6 (but no smaller).
There are two 3-regular graphs of order 6, namely K3,3 and K3  K2 .
There is no 3-regular graph of order 5 and one 3-regular graph of order 4,
namely K4 . Hence, we seek a 4-regular graph of minimum order contain-
ing either K3,3 or K3  K2 as an induced subgraph but not containing
K4 . The order of such a graph must be at least 7. If n(4, 3, 6) = 7,
then there exists a 4-regular graph G of order 7 containing an induced
3-regular subgraph H of order 6. Hence, G contains a single vertex v
not in H. Since v cannot be joined to the vertices of H to produce
such a graph G, it follows that n(4, 3, 6) ≥ 8. Since the graph shown in
Figure 19.4 has the desired properties, n(4, 3, 6) = 8.
234 CHAPTER 19. EXTREMAL GRAPH THEORY
.......... .....
................. ...............................
......... . ......
....... ...... ......
.. ........ .. .
......... ......
. ...... ......
..... ...... ......
.
...... ..... ......
.... ..... .... ......
......
... ......... ......
.... ......
...
. ..... ......
..
. . ............ ........ .........
. .
.... ... ............
........ .......... . . ... ....
. .
.... ... .......... ..
.... ..
........................
... ...... ........... .... .... ...
...... .......... ......... ...... .......... ........
... ...... . ... . .... . .
.. ...... .............................. ............... ......
.
.... ............ ......................... ......................
. .. .. .
... ................. ................... .............................. ...........
... ..
.. ....... .. ........... . . .. . ..
. ..... ....
.......... ........
... ..... ........ ...
... ...... ........... ...... ...... .......... ......
... .. ...... .......... ........................ ...................
... ...... ................ ... ..
.
..... ..
... ...
... .
....... ...
...... ........
...
... ...... ..
. .......
.
... ...... ......
.. ........ ....
...
..... ...... ......
...... ......
..... ...... ......
..... ...... ......
......
....... ......
. ...........
........ ...... .. ......
............ .. .. .. ..
...............................................

Figure 19.4: The graph in Exercise 17(b)

18. (a) For a graph to have a vertex of degree 3, it must have order at least 4. The
graph K4 − e in Figure 19.5(a) is the unique graph having degree sequence 2,
2, 3, 3. Since it has girth 3, it is the unique (S, 3)-cage.
(b) Since K4 − e is the unique graph of order 4 having degree sequence 2, 2,
3, 3 and it has girth 3, any (S, 4)-cage must have order 6 or more. Since no
graph has degree sequence 2, 2, 2, 3, 3, 3 (as no graph has an odd number of
odd vertices), the order of an (S, 4)-cage is at least 8. Therefore, the graph in
Figure 19.5(b) is an (S, 4)-cage.
(c) By the same reasoning as in (b), the order of an (S, 5)-cage is at least 8
and the graph in Figure 19.5(c) is an (S, 5)-cage.
(d) The order of an (S, 6)-cage is at least 12 and the graph in Figure 19.5(d)
is an (S, 6)-cage.
.........
...................
..... .....
.
..
..... .....
.
..... .....
.
.....
. .. ..... .....
.....
................. ............ ............ ... ............. ... . .
... .
........ .....
..... ......... ... ... ............ .. ............ ........
.
.
. .
.. ..... .....
..
...... ..... ..
.. .. ... .
.
. ... ... .....
........
.....
. .....
..... ........... .
... ...
. ........
..... . ............... ............ ... . .....
.
... ...
............... ................ ... .......... .
.
..
.
......... .
.
.........
.
...
...
..... .. .......
..... ...... ................... ..
. .
. . ...
..... .
..... ..... ..... .. ..
. ... ...
..... ... .
............... ............... ...... ..... .
.. ... . ... ...
..... ........ .
.........
.. ........... .
.. .
.
..
. .... ...
............. ........... .......... . . ........
........ ..
..
. ...... ..
. ................ .......... . . ...
... ...... .. . ..... .......... .
............... ..
..
.. ............. .
.. ..........
..
.. ...... .....
............... ................ ... .
..... . ..... ..
..... ..
.... ......
.
. ............
............ ............... .......
(a) (b) (c)
(d)

Figure 19.5: The graphs in Exercise 18

Chapter 20

Ramsey Theory

Section 20.1. Classical Ramsey Numbers

1. Let G be a graph of order R(s, t) − 1.
(a) Let H = G ∨ K1 . Hence, H is a graph of order R(s, t). Therefore, either
Ks ⊆ H or Kt ⊆ H = G + K1 . If Ks ⊆ H, then Ks−1 ⊆ G; while if
Kt ⊆ H, then Kt ⊆ G.
(b) Let F = G + K1 . Hence, F is a graph of order R(s, t). Therefore, either
Ks ⊆ F or Kt ⊆ F = G ∨ K1 . If Ks ⊆ F , then Ks ⊆ G; while if
Kt ⊆ F = G ∨ K1 , then Kt−1 ⊆ G.
2. It is trivial that R(1, t) = 1. Suppose that t ≥ 2. If all edges of Kt−1 are
colored blue, then there is neither a red K2 nor a blue Kt . Thus, R(2, t) ≥ t.
On the other hand, let there be given a red-blue coloring of Kt . If there is
an edge of Kt that is colored red, then Kt contains a red K2 ; otherwise all
edges of Kt are colored blue, resulting in a blue Kt . Thus, R(2, t) ≤ t and so
R(2, t) = t.
3. Proof. Let G be a graph of order R(s, t). Then either (1) Ks ⊆ G, implying
that Ks0 ⊆ G, or (2) Kt ⊆ G, implying that Kt0 ⊆ G. Hence, R(s0 , t0 ) ≤
R(s, t).
Next, we show that equality holds if and only if s0 = s and t0 = t. Certainly
if s = s0 and t0 = t, then R(s0 , t0 ) = R(s, t). For the converse, assume, to the
contrary, that R(s0 , t0 ) = R(s, t) but s 6= s0 or t0 6= t. Hence, s < s0 or t0 < t,
say s0 < s. Let G0 be a graph of order R(s, t) − 1 and let H = G0 ∨ K1 . Hence,
H is a graph of order R(s, t). Therefore, either Ks ⊆ H or Kt ⊆ H = G0 +K1 .
If Ks ⊆ H, then Ks−1 ⊆ G0 and so Ks0 ⊆ G; while if Kt ⊆ H, then Kt ⊆ G0
and so Kt0 ⊆ G0 . Thus, R(s0 , t0 ) ≤ R(s, t) − 1 and so R(s0 , t0 ) 6= R(s, t),
producing a contradiction.
4. The 3-regular graph G of order 8 shown in Figure 20.1 contains no K3
and G contains no K4 . Hence, R(3, 4) ≥ 9. By Theorem 20.6, R(3, 4) ≤
b(42 + 3)/2c = 9. Hence, R(3, 4) = 9.

235
236 CHAPTER 20. RAMSEY THEORY

s ... ...
s .
......................
........ .... .... .........
s
..................... ....
.. ..
.. ..... s
... ........................ s
.......
........ .... ..... ........
... .....
s
...........................................................................................
...
......... ............. . ... .. . . .
... .....
... ...... ... .......................... ......... .... ... ... ....... ..... ....
... .................. .................. .. ... .. ..... ..... ...
.................... ........... .. ... .......... ..... ..
.
........ . ... ..... ... .......... ... .......
. ..... ...
s ........
.........
........ ........ ......... ......... ......... ................
....
. ...........
......... .. ... ...... ...... ... .... ................
. .
.
.... s
..
... s.. ..... ..
............. ..... s
........
......
.....
. . ..
..... .. .........
... ..... ..
... ....... ..........
.
.......... .. ....
.. .............. ..
....... . . ... ......... ....... ....
.
.... . ... ...
... ....... ....................................... .... ... ... ......
... .. ......
.
.. .
.... ...
... ....... ............... .................... ........ ..... .....
s
........................... .....
. ... .... s
.. ......................
. s
......
.....
.....
......... .........
...
.....
......... ..... ......... ..............
..... s
...
...

s s
... .. ......... ..............
..... .........

G G

Figure 20.1: The graph in Exercise 4

5. Proof. Let there be given a red-blue coloring of K18 . There is a vertex v

of K18 that is incident with 9 edges of the same color, say vvi is a red edge
for 1 ≤ i ≤ 9. Then the subgraph H induced by {v1 , v2 , · · · , v9 } is the graph
K9 each of whose edges is colored red or blue. Since R(3, 4) = 9, either H
contains a red K3 , implying that K18 has a red K4 , or H contains a blue K4 ,
implying that K18 has a blue K4 .
√
6. By Theorem 20.7, R(5, 5) > b25/2 c = b4 2c = 5 and so R(5, 5) ≥ 6. On the
other hand, R(5, 5) > R(4, 5) by Exercise 3. Since R(4, 5) = 25, it follows that
R(5, 5) ≥ 26. Next, let there be given a red-blue coloring of K50 . Then there
a vertex v of K50 that is incident with 25 edges of the same color, say vvi is a
red edge for 1 ≤ i ≤ 25. Then the subgraph H induced by {v1 , v2 , · · · , v25 } is
the graph K25 each of whose edges is colored red or blue. Since R(4, 5) = 25,
either H contains a red K4 implying that K50 has a red K5 or H contains a
blue K5 , implying that K50 has a blue K5 . Hence, 26 ≤ R(5, 5) ≤ 50.
7. Claim: min{tc } = 2.
Proof. The coloring that assigns each edge of 2K3 red and all other edges of
K6 blue produces two monochromatic triangles. Thus, min{tc } ≤ 2. Suppose
that min{tc } < 2. Then there exists a red-blue coloring c such that tc =
1. We may assume that each of the edges v1 v2 , v1 v3 , v2 v3 is colored red.
Necessarily, at least one of the edges v4 v5 , v4 v6 , v5 v6 is colored blue, say v4 v5
(see Figure 20.2). Since K6 contains only one monochromatic triangle, at least
two of the edges v4 v1 , v4 v2 , v4 v3 are blue and at least two of the edges v5 v1 ,
v5 v2 , v5 v3 are blue. Hence, v4 and v5 are joined to at least one of v1 , v2 , v3 by
blue edges, say v4 and v5 are joined to v3 by blue edges. Then all of the edges
v3 v4 , v3 v5 , v4 v5 are blue, implying that tc ≥ 2, producing a contradiction.
8. R(2, 3, 4) = 9. Since R(3, 4) = 9, there exists a blue-green coloring of K8 that
avoids both a blue K3 and a green K4 . Thus, R(2, 3, 4) ≥ 9. Let there be
given an red-blue-green coloring of K9 . If any edge of K9 is colored red, then
a red K2 results. If no edge of K9 is colored red, then we have a blue-green
coloring of K9 . Since R(3, 4) = 9, there is either a blue K3 or a green K4 .
Thus, R(2, 3, 4) = 9. [By the same argument, R(2, s, t) = R(s, t) for every
two positive integers s and t.]
 237

v1 R v2
.............. ......
... ...
... ...... ...........
...... ......
......
...... .
..
.......
.
...... ..
R ...... ......
....................... R
........

v3
v6
.......
...... .....
...... ... ...........
.
...
....... ......
.. ......
...... ......
...... ......
...... .........
............... ... ...
...
...

v4 B v5

Figure 20.2: A coloring of K6 in Exercise 7

9. No. Consider the red subgraph F and the blue subgraph H of K6 − e in

Figure 20.3, where e = uv. [Note that F ∼
= H.]

u ....
u
... ... ........
................ .........................................
........ ......
... . ... ...... . .....
... .. .....
. .......
. .
.
.....
. .. ....
... . ...
... ... . ...
... .
. ... ..
. .
. ...
... ... ... ...
...
. .
........... . ....
..
...... .....v ...
... ..
. .......... ......... .... ... ...........
...
F : ..
. ... ..........
. ......
.
v
..... ...
...... ...
..
...........
H: ....
...
z .. ..
.
. ..
. ... .w
. .
.
...
.
...
. .. .. . ...
............. ..... .... ... .............. .. ... .
. ..
z ....... ..... w ... ......
...... .. .
.
.
. ...
.
... ....... ..
........... .
..
.
.
... ... ... .. . . .
..
...
.... ... .......
.....
...
....... ...
... ..
...
..... .. ...... ...... ... ...
... . ..... .. ......... .. ....
... ... ...... ... .......... ................. ..........
..
...... ..... ... ................. .... .........
.... ..... ...... ........ ........
....
y x y x

Figure 20.3: The graph in Exercise 9

10. (a) Suppose that the edges vvi are colored the same for 1 ≤ i ≤ 4, say these
are all red edges. If any two of the six edges v1 v2 , v1 v3 , v1 v4 , v2 v3 , v2 v4 , v3 v4
are colored red, there are two red triangles. If at most one edge, say v1 v2 ,
is colored red, then {v1 , v3 , v4 } and {v2 , v3 , v4 } are the vertices of two blue
triangles.
(b) Suppose that GR , say, is 3-regular. If GR = K3  K2 , then GR has
two triangles; while if GR = K3,3 , then GB = 2K3 and contains two
triangles.
(c) If neither GR nor GB contain a vertex of degree 4 or more or is 3-regular,
then each of these graphs must contain only vertices of degree 2 or 3,
necessarily, some of each. Since every graph contains an even number of
odd vertices, one of these two graphs contains four vertices of degree 3
and two vertices of degree 2 while the other contains two vertices of
degree 3 and four vertices of degree 2. Suppose that GR contains four
vertices of degree 3 and two vertices of degree 2. The two vertices u
and v of degree 2 must be nonadjacent; for otherwise, e = uv ∈ E(GR )
and GR − e is a disconnected graph containing four vertices of degree 3
and two vertices of degree 1, belonging to different components. This is
238 CHAPTER 20. RAMSEY THEORY

impossible. Therefore, uv ∈ / E(GR ) and so G = GR + e is 3-regular. If

G = K3,3 , then K6 contains two blue triangles, otherwise, G = K3  K2 .
If e belongs to no triangle of G, then K6 contains two red triangles. If
e belongs to a triangle of G, then K6 contains a red triangle and a blue
triangle.

Section 20.2. More General Ramsey Numbers

11. Claim: R(F, H) = 7.
Proof. First, we show that R(F, H) ≥ 7. Since the red-blue coloring of K6
in which the red graph is 2K3 and the blue graph is K3,3 has neither a red F
nor a blue H, it follows that R(F, H) ≥ 7. Now, let there be given a red-blue
coloring of K7 . Since R(K3 , K3 ) = 6, K7 contains a monochromatic K3 . Let
U be the vertex set of a monochromatic K3 and let W be the set consisting
of the remaining four vertices of K7 . We consider two cases.

Case 1. The monochromatic K3 with vertex set U is blue. If any edge joining
two vertices of W is blue, then there is a blue H; otherwise, there is a red F .

Case 2. The monochromatic K3 with vertex set U is red. If any edge joining a
vertex of U and a vertex of W is red, then there is a red F . Otherwise, every
edge joining a vertex of U and a vertex of W is blue. If any edge joining two
vertices of W is blue, then there is a blue H; otherwise, there is a red F .

Thus, R(F, H) ≤ 7 and so R(F, H) = 7.

12. Claim: R(C3 , C4 ) = 7.
Proof. If the edges of K3,3 in K6 are colored red and the edges of K 3,3 = 2K3
are colored blue, then there is no red C3 and no blue C4 . Thus, R(C3 , C4 ) ≥ 7.
We show that R(C3 , C4 ) = 7. Suppose that there is a red-blue coloring of K7
for which there is no red C3 and no blue C4 . Since R(3, 3) = 6, there is a
monochromatic triangle in K7 . By assumption, this triangle is blue. Deleting
a vertex of this triangle produces a red-blue coloring of K6 , which necessarily
has a blue triangle. Since there is no blue C4 , these two blue triangles do
not have a common edge. Therefore, either K7 has two blue triangles with a
common vertex or two disjoint blue triangles. We consider these two cases.

Case 1. K7 has two blue triangles with a common vertex. See Figure 20.4(a).
Since K7 has no blue C4 , each edge ui vj (i = 1, 2; j = 1, 2) is red. The vertex
w is adjacent to at most one of the vertices v, u1 , u2 , v1 , v2 by a blue edge, for
otherwise K7 has a blue C4 . Hence, either w is adjacent to both u1 and v1 ,
say, by red edges, or to u2 and v2 by red edges. This, however, produces a
red C3 , which is impossible.

Case 2. K7 has two disjoint blue triangles. See Figure 20.4(b). Since K7 has
no blue C4 , at most one edge joining U = {u1 , u2 , u3 } and W = {w1 , w2 , w3 }
 239

is blue. Hence, we may assume that all edges joining U and W are red, except
possibly u1 w1 . Since K7 contains no red C3 , at least one of xu2 and xw2 is
blue, say xu2 is blue. Thus, xu1 and xu3 are red. Hence, xw1 and xw2 are
blue, producing a blue C4 and a contradiction.

u1 B u2 u3
............ .......... .........
......... .......... ........................
..... ..... B ....... ....... B
.....
..... .
........ ..
........... .......
. .......
..... .....
. .......
B .....
.....
.....
..... B . .... .
......... B .......
...........
.....
..... .....
..... ......... u .... ....
1 ...
... ...
... u2
.... .............
w ............ .......
... ... v
....................
... x
..... .....
..... ......... .......
B .....
...
. .....
.....
..... B w 1 .................... .......
.......... w2
...
...... ..... ....... B ...
.........
.......
..... ..... ....... .......
.
.... ..... ....... .....................
. .
.
..
..........
. .. .........
... ....
...
B ...... ....
..
B
v1 B v2 w3

(a) (b)

Figure 20.4: Blue triangles in K7 in Exercise 12

13. Claim: R(K3 , C5 ) = 9.

Proof. First, observe that if the edges of K4,4 are colored red in K8 and
the remaining edges of K8 are colored blue (resulting in a blue 2K4 ), then
there is no red K3 and no blue C5 . Thus, R(K3 , C5 ) ≥ 9. Suppose that there
is a red-blue coloring of the edges of K9 in which there is no red K3 . Since
R(K3 , K4 ) = 9, there must be a blue F = K4 in K9 . Each of the five vertices
in K9 not in F must be joined to at most one vertex of F by a blue edge, for
otherwise K9 contains a blue C5 . If these five vertices form a blue K5 , then
K9 contains a blue C5 . Hence, there are two vertices u and v not in F joined
by a red edge. Since u and v are joined to F by at least three red edges, K9
contains a red K3 . Thus, R(K3 , C5 ) ≤ 9 and so R(K3 , C5 ) = 9.

14. Proof. Let n = R(F 0 , H) + R(F, H 0 ) and let there be given a red-blue
coloring of Kn . Let v be a vertex of Kn . We consider two cases.

Case 1. v is incident with at least R(F 0 , H) red edges. Let A be the set
of vertices in Kn that are joined to v by a red edge. Thus, the order of
the (complete) subgraph of Kn induced by A is p ≥ R(F 0 , H). Hence, this
complete subgraph Kp contains either a red F 0 or a blue H. If Kp contains a
blue H, then Kn contains a blue H as well. On the other hand, if Kp contains
a red F 0 , then Kn contains a red F since v is joined to every vertex of A by
a red edge.

Case 2. v is incident with at most R(F 0 , H) − 1 red edges. Then v is incident

with at least R(F, H 0 ) blue edges. Let B be the set of vertices in Kn that are
joined to v by a blue edge. Thus, the order of the (complete) subgraph of Kn
induced by B is q ≥ R(F, H 0 ). Hence, this complete subgraph Kq contains
either a red F or a blue H 0 . If Kq contains a red F , then Kn contains a red
240 CHAPTER 20. RAMSEY THEORY

F as well. On the other hand, if Kq contains a blue H 0 , then Kn contains a

blue H since v is joined to every vertex of B by a blue edge.

15. Proof. By Theorem 13.5, K4k+1 is Hamiltonian-factorable. Thus, there

is a factorization of K4k+1 into 2k Hamiltonian cycles H1 , H2 , . . ., H2k .
If the edges of H1 , H2 , . . ., Hk are colored red and Hk+1 , Hk+2 , . . ., H2k
are colored blue, then there is no red K1,2k+1 and no blue K1,2k+1 . Hence,
R(K1,2k+1 , K1,2k+1 ) > 4k + 1 and so R(K1,2k+1 , K1,2k+1 ) = 4k + 2 = 2n.
Next, we assume that n is even. Then n = 2k for some integer k. We
show that R(K1,n , K1,n ) = R(K1,2k , K1,2k ) = 4k − 1 = 2n − 1. Let there be
given a red-blue coloring of K4k−1 . Now deg v = 4k − 2 for each vertex v
of K4k−1 . It is impossible to have 2k − 1 red edges and 2k − 1 blue edges
incident with every vertex of K4k−1 , for otherwise, the red and blue subgraphs
have an odd number of odd vertices. Therefore, there is either a red K1,2k
or a blue K1,2k . Thus, R(K1,2k , K1,2k ) ≤ 4k − 1. The graph K4k−2 can be
factored into 2k − 2 Hamiltonian cycles H1 , H2 , . . ., H2k−2 and a 1-factor F
by Theorem 13.2. If the edges of H1 , H2 , . . ., Hk−1 are colored red and all
other edges are colored blue, then there is no red K1,2k and no blue K1,2k .
Hence, R(K1,2k , K1,2k ) > 4k − 2 and so R(K1,2k , K1,2k ) = 4k − 1 = 2n − 1.

16. Proof. Let a red-blue coloring of Ks+t−1 be given resulting a red subgraph
F and a blue subgraph H. If H contains a vertex of degree t or more, then
K1,t ⊆ H. Thus, we may assume that every vertex of H has degree t − 1 or
less. Hence, in this case, δ(F ) ≥ s − 1, which implies that F contains every
tree of order s as a subgraph. Therefore, Ts ⊆ F .

17. (a) Since Rχ (s, t) ≤ R(s, t), it follows that Rχ (s, t) exists for all integers s
and t with 1 ≤ s ≤ t.
(b) Proof. Let n = Rχ (s − 1, t) + Rχ (s, t − 1). Let there be given a red-
blue coloring of the complete graph Kn . We show that Kn contains a
red subgraph with chromatic number at least s or a blue subgraph with
chromatic number at least t. We consider two cases.

Case 1. Some vertex v of Kn is incident with Rχ (s − 1, t) red edges.

Let A be the set of vertices that are joined to v by red edges. Then
|A| ≥ Rχ (s − 1, t) and so the order of the (complete) subgraph G of
Kn induced by A is at least Rχ (s − 1, t). Hence, either G contains a
red subgraph F with χ(F ) ≥ s − 1 or G contains a blue subgraph H
with χ(H) ≥ t. In the latter case, we have the desired result. On the
other hand, if G contains a red subgraph F with χ(F ) ≥ s − 1, then
Kn contains a red subgraph with chromatic number at least s since v is
joined to every vertex of F by red edges.

Case 2. No vertex of Kn is incident with Rχ (s − 1, t) red edges. Then for

each vertex v of Kn , v is incident with at least Rχ (s, t − 1) blue edges.
Let u be a vertex of Kn and let B be the set of vertices that are adjacent
 241

to u by blue edges. Then |B| ≥ Rχ (s, t−1) and so the complete subgraph
G induced by B contains either a red subgraph F with χ(F ) ≥ s or a
blue subgraph H with χ(H) ≥ t − 1. In the former case, we have the
desired result. So, we may assume that G contains a blue subgraph H
with χ(H) ≥ t − 1. Since u is joined to every vertex of H by blue edges,
Kn contains a blue subgraph whose chromatic number is at least t.
(c) Claim: Rχ (3, 3) = 5.
Proof. Let there be given a red-blue coloring of K4 such that both
the red subgraph and the blue subgraph is P4 . Since both subgraphs
are bipartite, Rχ (3, 3) > 4. For n = 5, if either the red or the blue
subgraph is bipartite, then the other subgraph contains a triangle and
so has chromatic number at least 3. Thus, Rχ (3, 3) = 5.
(d) Claim: For 3 ≤ s ≤ t, Rχ (s, t) = (s − 1)(t − 1) + 1.
Proof. The red-blue coloring of K(s−1)(t−1) in which the red subgraph
is the complete (t−1)-partite graph Ks−1,s−1,...,s−1 (and so has chromatic
number t − 1) and the blue subgraph is the graph (t − 1)Ks−1 (and so
has chromatic number s − 1) shows that Rχ (s, t) ≥ (s − 1)(t − 1) + 1.
Let n = (s−1)(t−1)+1 and let there be given a red-blue coloring of of
Kn . If the independence number of the blue subgraph G of Kn satisfies
α(G) ≥ t, then the red subgraph F contains Kt and so χ(F ) ≥ t. If
α(G) ≤ t − 1, then no color class of G has more than t − 1 vertices, and
so
n 1 + (s − 1)(t − 1)
χ(G) ≥ = > s − 1.
t−1 t−1
Thus, χ(G) ≥ s and Rχ (s, t) ≤ (s − 1)(t − 1) + 1.
Therefore, Rχ (s, t) = (s − 1)(t − 1) + 1.
18. Proof. Suppose that R(G1 , G2 , . . . , Gk ) = n and consider

R(G1 , G2 , . . . , Gk , Gk+1 ),

where Gk+1 = K2 . We show that R(G1 , G2 , . . . , Gk , Gk+1 ) = n as well.

Since R(G1 , G2 , . . . , Gk ) = n, there exists an edge coloring of Kn−1 with the
colors 1, 2, . . . , k such that there is no copy of Gi whose edges are colored
i for every integer i with 1 ≤ i ≤ k. Using this same coloring of Kn−1
with the colors 1, 2, . . . , k + 1 where no edge is colored k + 1, we see that
there is no copy of Gi whose edges are colored i for every integer i with
1 ≤ i ≤ k + 1. Therefore, R(G1 , G2 , . . . , Gk , K2 ) ≥ n. Now, let there be given
an edge coloring of Kn with the colors 1, 2, . . . , k + 1. If any edge of Kn is
colored k + 1, then there is a copy of Gk+1 whose edge is colored k + 1. If
no edge of Kn is colored k + 1, then we have an edge coloring of Kn using
the colors 1, 2, . . . , k. Since R(G1 , G2 , . . . , Gk ) = n, there is an integer i with
1 ≤ i ≤ k such that Kn contains a copy of Gi every edge of which is colored i.
Therefore, R(G1 , G2 , . . . , Gk , K2 ) ≤ n. Hence, R(G1 , G2 , . . . , Gk , K2 ) = n =
R(G1 , G2 , . . . , Gk ).
242 CHAPTER 20. RAMSEY THEORY

19. Proof. Let n = R(Kn1 , Kn2 , . . . , Knk , Ts ). First, we show that n ≤ 1 +

(r − 1)(s − 1). Let there be given an edge coloring of K1+(r−1)(s−1) using
the colors 1, 2, . . . , k, k + 1. We show that for some integer i with 1 ≤ i ≤ k
there is either a copy of Kni all of whose edges are colored i or a copy of Ts
all of whose edges are colored k + 1. Replace each of the colors 1, 2, . . . , k
by the color 0. Then each edge of K1+(r−1)(s−1) is colored either 0 or k + 1.
Since R(Kr , Ts ) = 1 + (r − 1)(s − 1) by Theorem 20.11, there is either a
copy of Kr each edge of which is colored 0 or a copy of Ts each edge of
which is colored k + 1. In the latter case, we have the desired result. Thus,
we may assume that there is a copy of Kr every edge of which is colored 0.
Replace the color 0 of each edge of Kr by the original color assigned to that
edge. Hence, we have an edge coloring of Kr colored with 1, 2, . . . , k. Since
R(Kn1 , Kn2 , . . . , Knk ) = r, it follows that for some integer i with 1 ≤ i ≤ k
that Kr and Hence, Kn contains a copy of Kni each edge of which is colored
i. Therefore, n ≤ 1 + (r − 1)(s − 1).
It remains to show that n ≥ 1 + (r − 1)(s − 1). Consider G = K(r−1)(s−1) .
We show that there is an edge coloring of G with the colors 1, 2, . . . , k + 1
such that for each i with 1 ≤ i ≤ k there is no copy of Kni all of whose
edges are colored i and there is no copy of Ts all of whose edges are colored
k + 1. Partition the vertex set of K(r−1)(s−1) as {V1 , V2 , . . . , Vr−1 } so that
|Vi | = s − 1 for 1 ≤ i ≤ r − 1. First, we color each edge of G[Vi ] with the
color k + 1 for 1 ≤ i ≤ r − 1. Since the order of Ts is s, there is no copy
of Ts every edge of which is colored k + 1. Since R(Kn1 , Kn2 , . . . , Knk ) = r,
there is an edge coloring of H = Kr−1 with the colors 1, 2, . . . , k such that
there is no copy of Kni all of whose edges are colored i for every i with
1 ≤ i ≤ k. Let V (H) = {v1 , v2 , . . . , vr−1 }. We now color each edge joining a
vertex of Va and a vertex of Vb (1 ≤ a, b ≤ r − 1, a 6= b) in G with the color
t (1 ≤ t ≤ k) if and only if va vb is colored t in H. Since there is no copy of
Kni all edges of which are colored i (1 ≤ i ≤ k) in H, there is no copy of Kni
all edges of which are colored i in G. Therefore, n ≥ 1 + (r − 1)(s − 1) and so
R(Kn1 , Kn2 , . . . , Kni , Ts ) = 1 + (r − 1)(s − 1).
20. (a) M R(F, H) ≤ R(F, H) since if for every red-blue coloring of Kn , there is
either a red F or a blue H, then there is either a monochromatic F or a
monochromatic H.
(b) Suppose that F is isomorphic to a subgraph of H and M R(F, H) =
n. Then n is the minimum order of a complete graph such that for
every red-blue coloring of Kn , there is either a monochromatic F or
a monochromatic H. If there is a monochromatic H, then there is a
monochromatic F and so M R(F, H) = R(F, F ).
21. (a) By Theorem 20.11, R(K1,3 , K3 ) = (4 − 1)(3 − 1) + 1 = 7.
(b) By Exercise 20 (a) and (a) above, M R(K1,3 , K3 ) ≤ 7. Since the red-blue
coloring of K5 having C5 as both red subgraph and blue subgraph con-
tains neither a monochromatic K1,3 nor a monochromatic K3 , it follows
that M R(K1,3 , K3 ) ≥ 6. Let there be given a red-blue coloring of K6 .
 243

By Theorem 20.1, R(3, 3) = 6 and so there is a monochromatic K3 and

so M R(K1,3 , K3 ) = 6.

22. (a) By Theorem 20.11, R(P4 , K3 ) = (4 − 1)(3 − 1) + 1 = 7.

(b) By Exercise 20 (a) and (a) above, M R(P4 , K3 ) ≤ 7. Let there be given
a red-blue coloring of K4 . Suppose that some vertex of K4 is incident
with three edges of the same color, say vv1 , vv2 and vv3 are all red. If
any of the edges v1 v2 , v1 v3 , v2 v3 is red, then there is a red K3 ; otherwise
all three edges are blue and there is a blue K3 . Hence, we may assume
that every vertex of K4 is incident with two edges of one color and one
edge of the other color. Suppose that vv1 and vv2 are red edges and vv3
is a blue edge. If v1 v2 is red, then there is a red K3 . Hence, we may
assume that v1 v2 is blue. If v2 v3 , say, is blue, then (v1 , v2 , v3 , v) is blue;
while if v2 v3 is red, then (v1 , v, v2 , v3 ) is red. Thus, M R(P4 , K3 ) ≤ 4.
Since the red-blue coloring of K3 in which not all edges are colored the
same avoids a monochromatic P4 and a monochromatic K3 , it follows
that M R(P4 , K3 ) ≥ 4. Hence, M R(P4 , K3 ) = 4.

23. We claim that M R(F, H) = 6.

Since the red-blue coloring of K5 that gives a red C5 and a blue C5 pro-
duces neither a monochromatic F nor a monochromatic H, it follows that
M R(F, H) ≥ 6.
Next, let a red-blue coloring of K6 be given. Since R(3, 3) = 6, either a red
K3 or a blue K3 is produced, say, without loss of generality, the former. Let
S1 = {u1 , u2 , u3 } be the vertex set of a red K3 and let S2 = {v1 , v2 , v3 } be
the set of the remaining vertices of K6 . If any edge joining a vertex of S1
and a vertex of S2 is red, a red H is produced; so we may assume that all
edges joining S1 and S2 are blue. If all edges joining vertices of S2 are red, a
red F is produced. Thus, we may assume that some pair of vertices of S2 are
joined by a blue edge. This produced a blue H. Hence, M R(F, H) ≤ 6 and
so M R(F, H) = 6.

24. Claim: RR(K1,k , K1,k ) = (k − 1)2 + 2.

Proof. We first show that RR(K1,k , K1,k ) > (k − 1)2 + 1. Let N = (k −
1)2 + 1. We consider two cases.

Case 1. k is odd. Then N is odd. By Theorem 13.5, KN can be factored into

2
N −1
2 = (k−1)
2 Hamiltonian cycles, say

Ci,1 , Ci,2 , . . . , Ci, k−1 ,

2

for 1 ≤ i ≤ k − 1. For each fixed integer i with 1 ≤ i ≤ k − 1, color the edges

of Ci,j with the color i for 1 ≤ j ≤ k−1
2 . There is neither a monochromatic
K1,k nor a rainbow K1,k . Thus, RR(K1,k , K1,k ) ≥ N + 1 in this case.
244 CHAPTER 20. RAMSEY THEORY

Case 2. k is even. Then N is even. By Theorem 13.2, KN can be factored

N − 1 1-factors, say

Fi,1 , Fi,2 , . . . , Fi,k−1

for 1 ≤ i ≤ k − 1. For each fixed integer i with 1 ≤ i ≤ k − 1, color the edges

of Fi,j (1 ≤ j ≤ k − 1) with the color i. There is neither a monochromatic
K1,k nor a rainbow K1,k . Thus, RR(K1,k , K1,k ) ≥ N + 1 in this case as well.
Therefore, RR(K1,k , K1,k ) ≥ (k − 1)2 + 2.
We next show that RR(K1,k , K1,k ) ≤ (k − 1)2 + 2. Let n = (k − 1)2 + 2.
Let there be given a k-edge coloring of the edges of Kn such that there is
no monochromatic K1,k . Let v ∈ V (Kn ). Since deg v = n − 1 and there is
no monochromatic K1,k , at most k − 1 edges incident with v can be colored
n
the same. Thus, d k−1 e = k edges incident with v are colored differently,
producing a rainbow K1,k .
25. The red-blue-green coloring of K4 in which the red subgraph is K1,3 , the blue
subgraph is K3 and no edge is colored green avoids both a rainbow K1,3 and
a monochromatic P4 . Thus, RR(K1,3 , P4 ) ≥ 5. Let there be given a red-
blue-green coloring of K5 that results in no rainbow K1,3 . Thus, every vertex
of K5 is incident with edges colored with one of two colors. Let V (K5 ) =
{v1 , v2 , v3 , v4 , v5 }.
First, assume that there is a vertex of K5 that is incident with at least three
edges of the same color, say v1 v2 , v1 v3 , v1 v4 are red. If any of the edges in
S = {v2 v3 , v2 v4 , v3 v4 } is red, then there is a red P4 . Thus, each edge in S is
colored blue or green. Since there is no rainbow K1,3 , either all edges in S are
colored blue or all are colored green, say the former. Similarly, since there is
no rainbow K1,3 , each edge joining a vertex in {v2 , v3 , v4 } and v5 is either red
or blue, producing a monochromatic P4 .
Next, assume that each vertex of K5 that is incident two edges of one color
and two edges of another color, say v1 v2 and v1 v3 are red and v1 v4 and v1 v5
are blue. If there is an edge joining a vertex in {v2 , v3 } and a vertex in {v4 , v5 }
is red or blue, then there is a monochromatic P4 . Thus, each edge joining a
vertex in {v2 , v3 } and a vertex in {v4 , v5 } is green and so produces a green P4 .
Thus, RR(K1,3 , P4 ) = 5.
26. (a) Since the red-blue coloring of K5 in which the red and blue graphs
are both C5 contains neither a red K1,3 nor a blue C4 , it follows that
R(K1,3 , C4 ) ≥ 6. Let there be given a red-blue coloring of K6 that has
no red K1,3 . Thus, each component of the red subgraph of K6 is ei-
ther a path (possibly trivial) or a cycle. If the red subgraph is C6 , then
the blue subgraph contains C4 . If the red subgraph is K1 + C5 , where
V (K1 ) = {v} and C5 =(v1 , v2 , v3 , v4 , v5 , v1 ), then (v, v1 , v3 , v5 , v) is a
blue C4 . The same is true if the red subgraph is K1 + P4 . Otherwise, the
red subgraph has one or more components whose vertex set has two or
three vertices and the blue subgraph contains C4 . Thus, R(K1,3 , C4 ) = 6.
 245

(b) Since there is a factorization of K4,4 into two Hamiltonian cycles,

BR(K1,3 , C4 ) ≥ 5.
Let there be given a red-blue coloring of K5,5 where the red subgraph
F contains no K1,3 . Let the partite sets of K5,5 be denoted by U =
{u1 , u2 , . . . , u5 } and W = {w1 , w2 , . . . , w5 }. We show that the blue sub-
graph H contains C4 . Edges can be added to F so that F is either C10 ,
C8 + K2 , 2C4 + K2 or C6 + C4 . We consider these four cases.
Case 1. F = C10 . We may assume that
C10 = (u1 , w1 , u2 , w2 , . . . , u5 , w5 , u1 ).
Then C 0 = (u1 , w2 , u5 , w3 , u1 ) is a 4-cycle in H.
Case 2. F = C8 + K2 . We may assume that
C8 = (u1 , w1 , u2 , w2 , . . . , u4 , w4 , u1 )
and V (K2 ) = {u5 , w5 }. The cycle C 0 in Case 1 is a 4-cycle in H in this
case as well.
Case 3. F = 2C4 + K2 . We may assume that the two 4-cycles are
(u1 , w1 , u2 , w2 , u1 ) and (u3 , u3 , u4 , w4 , u3 ) and V (K2 ) = {u5 , w5 }. In
this case, (u1 , w3 , u5 , w4 , u1 ) is a 4-cycle in H.
Case 4. F = C6 + C4 . We may assume that
C6 = (u1 , w1 , u2 , w2 , u3 , w3 , u1 ) and C4 = (u4 , u4 , u5 , w5 , u4 ).
Here, (u1 , w4 , u2 , w5 , u1 ) is a 4-cycle in H.
Thus, BR(K1,3 , C4 ) = 5.

27. The red-blue coloring of K2,2 in which the red and blue graphs are both P3
contains neither a red P4 nor a blue P4 . Thus, BR(P4 , P4 ) ≥ 3. Let there
be given a red-blue coloring of K3,3 with no red P4 . Let the partite sets of
K3,3 be denoted by U = {u1 , u2 , u3 } and W = {w1 , w2 , w3 }. We show that
the blue subgraph H contains P4 . Edges can be added to the red subgraph
F , if necessary, so that F is either 2P3 or K1,3 + 2K1 . We consider these two
cases.
Case 1. F = 2P3 . We may assume that the two paths of order 3 are
(w1 , u2 , w2 ) and (u2 , w3 , u3 ). Then (w1 , u3 , w2 , u2 ) is a path of order 4 in H.
Case 2. F = K1,3 + 2K1 . We may assume that the edges of K1,3 are
u2 w1 , u2 w2 and u2 w3 and that u1 and u3 are isolated vertices in F . Thus,
(u1 , w2 , u3 , w1 ) is a path of order 4 in H.
Hence, BR(P4 , P4 ) = 3.

28. Claim: n = 2k + 1.
Proof. We proceed by induction on k ≥ 2. For k = 2, we show that n = 5.
First, the red-blue coloring of K4 resulting in a red C4 and a blue 2K2 contains
246 CHAPTER 20. RAMSEY THEORY

no monochromatic subgraph with chromatic number at least 3. Let there be

given a red-blue coloring of K5 . Suppose that the red subgraph is bipartite
with partite sets U and W , where say |U | ≥ 3. Then there is a blue triangle
with chromatic number 3.
Assume for an integer k ≥ 2 that the minimum positive integer n such that
every k-edge coloring of Kn results in a monochromatic subgraph F of Kn
with χ(F ) ≥ 3 is 2k + 1.
We show that the minimum positive integer n such that every (k + 1)-edge
coloring of Kn results in a monochromatic subgraph with chromatic number
at least 3 is 2k+1 + 1.
First, we show that there is a (k + 1)-edge coloring of G = K2k+1 that avoids
every monochromatic subgraph H with χ(H) ≥ 3. Let V (G) = U ∪ W where
|U | = |W | = 2k . Let G1 = G[U ] = K2k and G2 = G[W ] = K2k . By
the induction hypothesis, there is a k-edge coloring of K2k , using the colors
1, 2, . . . , k, that contains no monochromatic subgraph with chromatic number
at least 3. Let there be given such a k-edge coloring of both G1 and G2 . By
coloring each edge in [U, W ] the color k + 1, we obtain a (k + 1)-edge coloring
of K2k+1 with the desired property.
Next, we show that for every (k + 1)-edge coloring of K2k+1 +1 , there is a
monochromatic subgraph F with χ(F ) ≥ 3. Let there be given a (k + 1)-edge
coloring of K2k+1 +1 using the colors 1, 2, . . . , k + 1. Let Gi be the subgraph
induced by edges colored i for 1 ≤ i ≤ k + 1. Suppose that χ(Gk+1 ) ≤ 2, that
is, Gk+1 is bipartite, having partite sets U and W , where say |U | ≥ 2k + 1.
Thus, K2k +1 is a subgraph of K2k+1 +1 , each of whose edges is colored with
some color i ∈ {1, 2, . . . , k}. It then follows by the induction hypothesis
that there is a monochromatic subgraph F in K2k +1 (and so in K2k+1 +1 )
with χ(F ) ≥ 3.
Chapter 21

The Probabilistic Method

Section 21.1. The Probabilistic Method

1. Follow the proof of Theorem 21.1 through the sentence

" #  
[ X n 1−(nt)
P AT ≤ P [AT ] = 2 .
t
T ∈T T ∈T

   
By assumption, P [∪AT ] < 1. Then P ∪AT > 0, that is, P ∩AT > 0.
Continue with the last sentence of the proof of Theorem 21.1.

2. Follow the proof of Theorem 21.2 through the sentence

“Because L0 ∩ L1 = ∅, this results in a proper coloring of Kr,r and so
χ` (Kr,r ) ≤ d2 log2 re.”
The proof will be completed once we have shown the existence of a partition
of L that satisfies (21.5) and (21.6).
Recall that the number of (ordered) partitions L = L0 ∪ L1 is the same as the
number of subsets L0 of L (since, then, L1 = L − L0 ), that is, 2n .
For a fixed i ∈ {0, 1} and v ∈ Vi , the number of subsets Li for which L1 ∩
L(v) = ∅ is 2n−|L(v)| . However,

2n 2n
2n−|L(v)| = < .
2|L(v)| 2r

So the number of partitions that fail to satisfy at least one of (21.5) and
n
(21.6) is less than (2r) 22r = 2n . So at least one partition of L satisfies (21.5)
and (21.6).

3. According to Theorem 21.2, χ` (K10,10 ) ≤ d2 log2 10e. Since log2 10 ≤ 3.33, it

follows that d2 log2 10e ≤ 7.

247
248 CHAPTER 21. THE PROBABILISTIC METHOD

s t
4. Assume that ns 2−(2) + nt 2−(2) < 1. We show that there exists a red-blue

coloring of Kn that contains neither a red Ks nor a blue Kt . Consider the

probability space S whose elements are the red-blue colorings of Kn , where
V (Kn ) = {v1 , v2 , . . . , vn }. The probabilities are defined by
1
P [vi vj is red] = P [vi vj is blue] =
2
for each pair vi , vj of distinct vertices of Kn and letting these events be
n
mutually independent. Then each of the 2( 2 ) red-blue colorings of Kn is
n
equally likely with probability 2−( 2 ) . For a fixed s-element subset U ⊆
{v1 , v2 , . . . , vn }, let AU be the event that the subgraph induced by U in Kn
is a red Ks . Similarly, for a fixed t-element subset W ⊆ {v1 , v2 , . . . , vn }, let
AW be the event that the subgraph induced by W in Kn is a blue Kt . Then

(s) s
( t ) t
P [A ] = 1 s = 2−(2) and P [A ] = 1 2 = 2−(2) . Consequently,
U 2 W 2
   


n −(2s) n −(2t)
P [ ∪U ⊆V (G),|U |=s AU ∪ ∪W ⊆V (G),|W |=t AW ]≤ 2 + 2 <1
s t
by assumption. But then
h

i
P ∪U ⊆V (G),|U |=s AU ∪ ∪W ⊆V (G),|W |=t AW > 0,

that is, 


P ∩U ⊆V (G),|U |=s AU ∩ ∩W ⊆V (G),|W |=t AW > 0.
So, there is a red-blue coloring of Kn for which there is neither a red Ks nor
a blue Kt . Thus, R(s, t) > n.
5. For fixed positive integers n and k satisfying nk (1−2−k )n−k < 1, consider the


n
probability space S whose elements are the 2( 2 ) different labeled tournaments
T with vertex set V (T ) = {v1 , v2 , . . . , vn }. The probabilities are defined by
1
P [(vi , vj ) ∈ E(T )] = P [(vj , v1 ) ∈ E(T )] =
2
for each pair vi , vj of distinct vertices of T and then letting these events
n
mutually independent. Thus, each of the 2( 2 ) different labeled tournaments
n
is equally likely with probability ( 12 )( 2 ) .
For a k-element subset U ⊆ {v1 , v2 , . . . , vn }, let AU denote the event that
there is no w ∈ V (G) − U that is adjacent to every vertex of U . Each vertex
w ∈ V (G) − U has probability ( 12 )k of being adjacent to every vertex of U and
there are n − k such vertices w. Thus, P [AU ] = (1 − 2−k )n−k since for each
vertex w ∈ V (G) − U , the probability that w is not adjacent to every vertex
of U is 1 − 2−k and these n − k events corresponding to the n − k possible
choices of w are mutually independent. Consequently,
 
X n
P [∪U ⊆V (T ),|U |=k AU ] ≤ P [AU ] = (1 − 2−k )n−k .
k
U ⊆V (T ),|U |=k
 249

By assumption, nk (1 − 2−k )n−k < 1. Consequently, P [∪U ⊆V (T ),|U |=k AU ] < 1

and so P [∪U ⊆V (T ),|U |=k AU ] > 0, that is, P [∩U ⊆V (T ),|U |=k AU ] > 0. Thus,
there is a tournament T of order n with property Uk .

Section 21.2. Random Graphs

6. (a) Since P [A ∩ B] = P [A ∪ B]) ≤ P [A] + P [B], it follows that
P [A ∩ B] = 1 − P [A ∩ B] ≥ 1 − (P [A] + P [B]).

(b) Proof. Suppose that almost all graphs have property Q1 . Then
lim P [G ∈ G(n, p) has property Q1 ] = 1.
n→∞

Similarly,
lim P [G ∈ G(n, p) has property Q2 ] = 1.
n→∞
Now, by (a),
P [G ∈ G(n, p) has property Q1 and Q2 ]
≥ 1 − (P [G ∈ G(n, p) does not have property Q1 ]
+P [G ∈ G(n, p) does not have property Q2 ]).
Since
lim P [G ∈ G(n, p) does not have property Qi ] = 0
n→∞
for i = 1, 2, the result follows.
(c) Proof. We proceed by induction on k. The case where k = 2 holds by
Theorem 21.8. Assume that for an integer k−1 ≥ 2, the result is true and
let Q1 , Q2 , . . . , Qk be graphical properties, each of which is possessed by
almost all graphs. By the induction hypothesis, all almost all graphs have
all of the properties Q1 , Q2 , . . . , Qk−1 ; call this property Q. Then almost
all graphs have property Q and almost all graphs have property Qk . By
Theorem 21.8 again, almost all graphs have properties Q and Qk , that
is, almost all graphs have all of the properties Qi for i = 1, 2, . . . , k.
7. Proof. By Theorem 21.7, almost all graphs have property Q2,0 . Suppose
that G has property Q2,0 . Then, for each pair {u, v} of vertices of G, there
is a vertex z ∈
/ {u, v} that is adjacent to both u and v. Consequently, G is
connected with diameter 2.
8. Proof. By Theorem 21.9, almost all graphs contain Kk+1 as a subgraph.
Thus almost all graphs have chromatic number at least k + 1, that is, almost
no graphs are k-colorable.
9. Proof. Let n be a positive integer such that (n − 3)(n − 4)/12 > k. By The-
orem 21.9, almost all graphs contain Kn as a subgraph. By Theorem 11.18,
γ(Kn ) > k. Therefore, almost all graphs have genus exceeding k and conse-
quently, almost no graphs have genus at most k.
250 CHAPTER 21. THE PROBABILISTIC METHOD

10. Let p(n) be a function of n and, for each graph G in G(n, p(n)), let the
random variable X(G) denote the number of copies of Kk in G. By (21.10),
if lim E[X] = 0, then lim E[X ≥ 1] = 0.
n→∞ n→∞
For a graph G ∈ G(n, p(n)) and each k-element
subset S of V (G), let AS be
the event that G[S] is complete. Since all k2 possible edges must be present
k
in G[S], we have P [AS ] = [p(n)](2) . Let XS be the indicator random variable
P
with XS = 1 if G[S] is complete and XS = 0 otherwise. Then X = XS ,
where the sum is taken over all k-element subsets S of V (G). Furthermore,
since XS is an indicator random variable,
k
E[XS ] = P [XS = 1] = P [AS ] = [p(n)](2) .

By the linearity of expectation,

 
n k k
[p(n)](2) < nk [p(n)](2) .
X
E[X] = E[XS ] =
k
k
Consequently, if lim nk [p(n)](2) = 0, then lim E[X] = 0 and so lim E[X ≥ 1] = 0,
n→∞ n→∞ n→∞
that is, almost no graph in G(n, p(n)) contains a copy of Kk . By assumption,

p(n)
lim = 0.
n→∞ n−2/(k−1)
Therefore,
 (k2)
p(n)
lim = 0.
n→∞ n−2/(k−1)
However,
 (k2)
p(n) k k(k−1)
[p(n)](2) · n k−1 · 2
2
=
n−2/(k−1)
k
= [p(n)](2) · nk .

11. Claim: If r(n) is a threshold function for some property Q, then so, too, is
2r(n). (Thus, threshold functions are not unique.)
Proof. Suppose that r(n) is a threshold function for some property Q. By
definition, this means that
p(n)
(i) lim = 0 implies that lim P [G ∈ G(n, p(n)) has Q] = 0
n→∞ r(n) n→∞
and
p(n)
(ii) lim = ∞ implies that lim P [G ∈ G(n, p(n)) has Q] = 1.
n→∞ r(n) n→∞

Consider the function r0 (n) = 2r(n). We must show that

 251

p(n)
(i) lim = 0 implies that lim P [G ∈ G(n, p(n)) has Q] = 0
n→∞ r0 (n) n→∞
and
p(n)
(ii) lim = ∞ implies that lim P [G ∈ G(n, p(n)) has Q] = 1.
n→∞ r0 (n) n→∞

p(n) p(n) p(n)

Since lim = 0 if and only if lim = 0 and lim = ∞ if and
n→∞ r(n) n→∞ c · r(n) n→∞ r(n)
p(n)
only if lim = ∞, the result follows.
n→∞ c · r(n)

12. Define a k-element subset U of vertices of a tournaments T to be bad if there is

no vertex w ∈ / U that is adjacent to every vertex of U . For each tournaments
T in T (n, 21 ), let X(T ) be the number of bad sets in T . We wish to show that
almost all tournaments have property Uk , that is,

lim P [T ∈ T (n, 21 ) has property Uk ] = 1.

n→∞

Certainly, if X(T ) = 0, then T has property Uk . Thus, by (21.10), it suffices

to show that lim E[X] = 0.
n→∞
For a tournament T in T (n, 12 ) and each k-element subset U of V (T ), let
XU be the indicator random variable defined
P by XU = 1 if the set U is bad
and XU = 0 otherwise. Then X = XU , where the sum is taken over
all k-element subsets U of V (T ). For a fixed set U and a vertex w ∈ / U,
the probability that w is adjacent to every vertex of U is ( 21 )k . Thus, the
probability that U is bad, that is, that no such vertex w exists, is (1−( 12 )k )n−k .
In other words,
 k !n−k
1
P [XU = 1] ≤ 1 − .
2
Since the number of k-element subsets U of V (T ) is nk , it follows that

   k !n−k
X n 1
E[X] = E[XU ] ≤ 1−
k 2
U ⊆V (T ),|U |=k
 k !n−k
1
≤ nk 1− .
2

Thus, by applying (21.2), we have lim E[X] = 0.

n→∞

13. If a graph G is connected with diameter 2, then G has at least two ver-
tices u and v for which e(u) = e(v) = 2. However, for every other vertex
w of G, e(w) = 1 or e(w) = 2. Consequently, G has eccentricity sequence
2, 2, . . . , 2, 1, 1, . . . , 1, where i terms equal 2 and n − i terms equal 1, for some
integer i with 2 ≤ i ≤ n. Since almost all graphs are connected with diame-
ter 2, the result follows.