Sequences and

Combinatorial Problems
S. I. Gelfand M. L. Gerver A. A. Kirillov
N. N. Konstantinov A.G. Kushnirenko

1
SE Q U EN C ES AND C O M B IN A T O R IA L PROBLEM S
T H E P O C K E T M A T H E M A T IC A L L IB R A R Y
J acob T. S c h w a r t z , Editor

P R IM E R S :

1. T H E C O O R D IN A T E M E T H O D
by I. M. Gclfand et al.
2. F U N C T IO N S A N D G R A P H S
by I. M. Gelfand et al.

W O R K B O O K S:

1. SEQUENCES
AN D COM BIN A TO RIA L PROBLEMS:
by S. I. Gclfand et al.
2. L E A R N L IM IT S T H R O U G H P R O B L E M S !
by S. I. Gelfand et al.
3. M A TH EM ATICAL PROBLEM S:
AN ANTHOLOGY
by E. B. Dynkin et al.

CO U RSES:

1. L IM IT S A N D C O N T IN U IT Y
by P. P. Korovkin
2. D IF F E R E N T IA T IO N
by P. P. Korovkin
 SEQUENCES AND
COMBINATORIAL PROBLEMS
BY

S. I. G elfand , M. L. G erver, A. A. K irillov ,

N. N. K onstantinov , A.G. K u s h n i r e n k o

Revised English Edition

Translated and Freely Adapted by
R ichard A. S ilverman

G O R D O N AN D BREACH
SCIENCE PUBLISHERS
NEW YORK • LONDON • PARIS
 Copyright © 1969 by Gordon and breach science publishers in c .
150 Fifth Avenue, New York, N .Y . 10011

Library of Congress Catalog Card Number: 68-55332

Editorial office for the United Kingdom:

Gordon and Breach Science Publishers Ltd.
12 Bloomsbury Way
London W. C. 1

Editorial office for France:

Gordon & Breach
7-9 rue Emile Dubois
Paris 14'

Distributed in Canada by:

The Ryerson Press
299 Queen Street West
Toronto 2B , Ontario

All rights reserved. No part of this book may be reproduced or utilized

in any form or by any means, electronic or mechanical, including photo­
copying, recording, or by any information storage and retrieval system,
without permission in writing from the publishers. Printed in east Germany.
 Contents

Preface vii
Preliminaries 1
Sample Problems 8
Detailed Solutions 25
Test Problems 81

V
 Preface

This book has a very simple structure. It begins with a brief

section called “Preliminaries” presenting the modicum of back­
ground information needed to solve the 89 problems stated in
the next section, called “Sample Problems.” These problems
are, for the most part, equipped with hints or answers or both.
But thenub of thebook is the section called “Detailed Solutions,”
where you will find all 89 sample problems worked out in full
detail. In our opinion, just studying these solutions (after first
spending a decent amount of time trying to solve the problems
on your own!) is a perfectly plausible way of learning about
sequences and combinatorial problems. Finally, to make sure
you have mastered the subject matter of the book, you should
attack all 37 problems in the section called “Test Problems.”
In fact, think of this section as a (rather tough) final examination
on which you must get at least a passing grade. Good luck!
 Preliminaries

1. Sequences

Suppose a real number u„ is associated with every positive

integer zz. Then the numbers

u i ,u 2, ...

are said to form an infinite sequence or simply a sequence, and

the numbers themselves are called the terms of the sequence. The
following are all examples of sequences:
1. The sequence all of whose terms are ones:

1, 1 ,...,1 , ...

Here u„ = 1, i.e., the number 1 is associated with every positive

integer.
2. The sequence of consecutive odd numbers:

1,3, 5 ,...
Here u„ = 2n — 1.
3. Consider the sequence specified by the formula

n (#z + 1)
i—

Write out the first seven terms of the sequence, and find z/100,
u„-3 and un+1.

1
2 Sequences and Combinatorial Problems

Answer. 1, 3, 6, 10, 15, 21, 28, ...

Wioo = 5050,

_ (n — 3) (n — 3 + 1) (n — 3)(n —2)
Un —3 ' 7 ^

(n + l)(n + 2)
"n + 1 =

The //th term of a sequence is called the general term of the

sequence. A sequence is often specified by giving an explicit
formula for its general term. Thus the sequence such that u„ = n2
starts off like
1,4,9, 16, 25, 36, 49, ...

The sequence with general term w„ is often denoted by {w„}, i.e.,

by writing the general term inside curly brackets.
Example 1. Find a formula for the general term of the se­
quence
2, 5, 8, 11, 14, 17, 20, ...

Answer. One possibility is

j/„ = 3n — 1.
Another is
w„ = (3/i — 1) x (the number of digits in n).

There are an infinite number of other possibilities!

Example 2. As a less obvious problem, find a formula for the
general term of the sequence

0, 13, f , 62, 171, . . .

Answer. One possibility is

n3 - 1
2
 Preliminaries 3

Example 3. One could hardly write a reasonable formula for

the general term of the sequence

3, 1,4, 1,5, 9, 2, 6, 5, 3, 5 ,... (1)

Nevertheless, (1) is just the sequence whose nth term is the nth
digit in the decimal representation of the number n = 3.1415926...
Hence there is actually a definite rule associating a term w„ of the
sequence (1) with every positive integer n. For example, ux = 3,
u2 = 1, «7 = 2, etc. Thus, despite the absence of an explicit
formula for the general term of this sequence, it is possible, at
least in principle, to find the number in any given position, be it
the first, seventh or 1007th position. In particular, it can be shown
that the sequence (1) is not periodic, i.e., that no block of terms
repeats itself over and over again like the underlined digits in the
decimal expansion of

\ = 0.142857142857...

As you recall from elementary geometry, n is the ratio of the

circumference of a circle to its diameter. The number n was
first estimated by Archimedes, who knew that it lies between
3y and 3^f. For ordinary purposes, the value n = 3.14 is often
good enough, but more accurate values like n = 3.14159 are
sometimes needed. In principle, n can be computed to any desired
accuracy. However, it is hard to imagine a problem requiring
knowledge of n to more than ten decimal places or so. Believe it
or not, somebody called Shanks once wasted a large part of his
life calculating n to 707 places. Then, more recently, several
thousand digits of n were calculated on a high-speed electronic
computer, and it turned out that Shanks’ calculations were in­
correct starting from the 202nd digit!
Example 4. Calculate the first 10 digits of the sequence

1, 1,2, 3, 5 ,... ( 2)
4 Sequences and Combinatorial Problems

formed by the following rule: The first two terms equal 1

(iq = 1,«2 = 1), while starting with the third term, every term is
the sum of the preceding two terms, i.e.,

wn = Un-l + »n-2 0 ^ 3).

There is an explicit formula for un in this case, but it is not too

simple (see Prob. 43, p. 15). The terms of the sequence (2) are
called the Fibonacci numbers, and the sequence itself is called the
Fibonacci sequence.
Example 5. Find the first few terms of the sequence

«1,«2, •••

whose uth term equals the sum of all the positive integers from
1 to n inclusive.
Answer. 1, 3, 6, 10, 15, 21, ...
Solution. uv = I, u2 = 1 + 2 = 3, u3 = 1 + 2 + 3 = 6,
2/4= 1 + 2 + 3 + 4 = 10, ... , u„ = 1 + 2 + ••• + n, ...

2. Mathematical Induction

If the first person in a line is a woman and if there is another

woman standing behind every woman (except the last), then
every person in the line is a woman. The reasoning behind this
somewhat facetious example occurs again and again in mathe­
matics and is called the principle o f mathematical induction. We
now give a more serious formulation of this principle: Given a
sequence o f assertions, i f the first assertion is true and if every
true assertion is followed by another true assertion, then every
assertion in the sequence is true.
Example 1. Prove that
n (n + 1)
1 + 2 + 3 + •••+ /! (3)
2
 Preliminaries 5

for every positive integer n. This formula comprises a whole

sequence of assertions:
1 •2
1) 1=
2
2 -3
2) 1+ 2 =
2
3 -4
3) 1+ 2+ 3 =
2
4 •5
4) 1+ 2 + 3 + 4 =

The first assertion of obviously true. We now verify that every

true assertion is followed by another true assertion. Suppose
assertion k is true, i.e., suppose (3) is valid for n = k so that
k (k + 1)
1 + 2 + 3 + - + * = (4)
2

Adding k + 1 to both sides of (4), we obtain

i „ - , „ k (k + 1) „
1 + 2 + 3 + ••• + * + (* + 1) = ----------- + (* + 1)
2
= (k + ! ) ( * + 2)
2
But this is just assertion * + 1, which comes right after asser­
tion *. Thus we have shown that every true assertion is followed
by another true assertion. Hence, according to the principle of
mathematical induction, every assertion in the sequence is true,
i.e., formula (3) holds for every positive integer n.
The same problem can be solved without recourse to mathe­
matical induction. Writing
u„ = 1 + 2 + 3 + ••• + n,
 6 Sequences and Combinatorial Problems

we have
u„ — 1 + 2 + 3 + ••• + (w — 2) + (// — 1) + //, (5)
and
u„ = n + (ji — 1) + (// — 2) + ••• + 3 + 2 + 1 (6)
(the second sum is the first sum written backwards). Adding
equations (5) and (6), we find that
2un = [1 + /j] + [2 + (n — 1)] + [3 + (// —2)] + •••
+ [(w —2) + 3] + [(w — 1) + 2] + [/i + 1].
Each term in brackets equals n + 1, and there are exactly n such
terms. In other words,
2u„ = (n + 1) + (n + 1) + ••• + (n + 1) + (// + 1) = n (n + 1),
n times
and hence
n (n + 1)
«n = -----------,
2
which is just another way of writing (3).
Another somewhat different form of the principle of mathe­
matical induction goes as follows: Given any assertion involving
an arbitrary positive integer n, suppose that
a) The assertion is true for n — 1;
b) Validity o f the assertion for n = k implies its validity for
n = k + 1.
Then the assertion is true for every positive integer n.
Warning. Note that hypothesis b) does not state that the
assertion is true for n = k + 1 but only that if it is true for
n = k, then it is true for it = k + 1.
Example 2. Prove than n5 —n is divisible by 5 for every po­
sitive integer n. The proof involves two steps:
a) If n = 1, n5 —n equals 0 and hence is trivially divisible
by 5.
b) Let n = k be an arbitrary positive integer k, and suppose
k 5 — k is divisible by 5. Then (k + l ) 5 — (k + 1) is also
 Preliminaries 1

divisible by 5. In fact, it follows from

(k + l)5 = k 5 + 5k4- + 10/c3 + 10A:2 + 5k + 1
that
(k + l)5 — (k + 1) = (k 5 + 5k4- + 10/c3 + 10k2 + 5k + 1)
= (k5 - k) + 5 (k4- + 2k 3 + 2k2 + k).
But each of the terms on the right is divisible by 5, the first by
hypothesis, the second since it is obviously a multiple of 5.
Since the sum of two numbers divisible by 5 is itself divisible by 5,
it follows that (k + l)5 — (k + 1) is divisible by 5. Thus hypo­
theses a) and b) figuring in the second formulation of the principle
of mathematical induction are satisfied. Therefore n5 — n is
divisible by 5 for every positive integer n.
There are other versions of the principle of mathematical
induction equivalent to those just given. For example, given any
assertion involving an arbitrary positive integer n, suppose that
a) The assertion is true for n = 1;
b) Validity o f the assertion for n ^ k implies its validity for
n = k + 1.
Then the assertion is true for every positive integer n.
The difference between this formulation of the principle of
mathematical induction and the preceding one is that hypothe­
sis b) now states that the assertion is true for all n ^ k and not
just for n = k. However, it is easy to see that the two formula­
tions arc equivalent in the sense that every theorem which can
be proved by applying the principle of mathematical induction
in one form can also be proved by applying the principle in the
other form (think this through).
It should also be noted that hypothesis a) is only needed to
“get started,” so to speak. Suppose that hypothesis a) is replaced
by a new hypothesis
a') The assertion is true for n = 8, say,
while hypothesis b) is left the same. Then it follows from a') and
b) that our assertion is true for all n starting from 8. This is
precisely the situation encountered in Problem 9, p. 9.
 Sample Problems

We begin with a group of problems (1-14) involving the

principle of mathematical induction.
1. Prove that
n2(n + l)2
l 3 + 23 + 33 + ••• + n3
4
for every positive integer n.
2. Prove that
1 1 J ___ n
+ + +
1 -2 2 -3 n(n + 1) n + 1
for every positive integer n.
Hint. The problem can be solved in two ways, either by mathe
matical induction or by using the formula
1 _ J_ _ 1
n(n + 1 ) n n + 1
3. Calculate the sum
1 1 _______1_______
1 -4 4 -7 + (3n - 2) (3n + 1)’

Answer.-----------.
3/i + 1
4. Prove that
1 1 1
+ ••• +
a {a + 1) (q + 1) (o + 2) (a + n - 1) (<7 + «)
n
a (a + n)

8
 Sample Problems 9

holds for every positive integer n and every a not equal to zero or
a negative integer.
5. Prove that
1 • 1! + 2 • 2! + ••• + n • n\ = (n + 1)! - 1,
where n is any positive integer and n ! (read “/i factorial”) denotes
the product of all positive integers from 1 to n inclusive.
6. Prove that

— )= n + 1
n2 ) 2n
for every integer n ^ 2.
Hint. The problem can be solved in two ways, either by
mathematical induction or by using the formula

j ___L = ( » - 1)(« + o
n2 n2
In using mathematical induction, start from n = 2.
7. Verify that
1 - 22 + 32 - ••• - ( - I ) " - 1 n2 - ( - 1 ) " - 1 n{n + 1}.
2
• •
\8 , Into how many parts is the plane divided by n straight lines
no two of which arc parallel and no three of which intersect in
a single point?
Hint. Suppose n lines have already been drawn, and show that
drawing the (n + l)st line increases the number of parts of the
plane by n + 1.
n2 + n + 2
Answer. ----------------.
2
9. Prove that any integer greater than 7 can be written as a
sum consisting of the integers 3 and 5 exclusively. (For example,
8 = 3 + 5, 9 = 3 + 3 + 3 , 10 = 5 + 5, 1 1 = 3 + 3 + 5, etc.)
Hint. Given that A:(A: > 7) can be written as a sum of threes
and fives, show that k + 1 can also be written as a sum of threes
2 Silverman V
 10 Sequences and Combinatorial Problems

and fives. Then apply mathematical induction, starting from

n = 8.
10. Suppose the plane is derived into parts by n straight lines
as in Problem 8. Prove that the plane can be colored black and
white in such a way that any two parts with a common side have
different colors, as shown in Figure 1 for the case n = 6. (Such
a coloring is said to be regular.)
11. Prove that the sum of the cubes of three consecutive posi­
tive integers is divisible by 9.
Fig. 1
12. Verify that
11 n + 2 + 12 2 n + 1

is divisible by 133 for any integer n ^ 0.

13. Prove that the inequality
(1 + a)n > 1 + na
holds for any integer n ^ 2 provided that a > —1 and <7^0.
14. Prove that
, 1 1 1 /-
1+ + —= + - + - = • > V n
V2 v 3 v n
for any integer n ^ 2.
Hint. Use the inequality

- > \!k + 1 - V k .
sjk + 1
The next group of problems (15-43) deals with sequences.
15. Consider the sequence uu u2, ... such that

«i 2, u2 = 3, u„ = 3«„_! - 2u, (n > 2).

Prove that
un = 2"~1 + 1.
16. Consider the sequence Wj, u2, ... such that
ui = 1, un+1 = un + 8n (/i 1).
Prove that un = (2n — l)2.
 Sample Problems 11

17. Given a sequence ulf u2, ..., let Au,, Au2, ... be the se­
quence with general term

Au„ = u„+l - u„
(the capital Greek letter A is read “delta”). This new sequence is
called the sequence of first differences of the original sequence
Mj, m2, __Given that u„ = n2, what iszlM„?
Answer. Au„ = 2n + 1.
18. Suppose two sequences uY, u 2, ... and vu v2, ... have the
same first differences, i.e., suppose z1m„ = Av„ for all n.
a) Can it be asserted that u„ — v„l
b) Can it be asserted that u„ = v„ provided it is known that
«i =
Answer, a) No; b) Yes. Use the formula uY + Aut + z!m2
+ ••• + Au„-t = U„.
19. Suppose every term of as equence ivj, iv2, ... equals the sum
of the corresponding terms of the sequences ux, u2, ... and
v u v2, ..., i.e., suppose h’„ = u„ + v„ for all n. Prove that
Aw„ — Au„ + Av„ for all it.
20. Given a sequence with general term u„ = nk, prove that
the general term z1m„ of the sequence of first differences is a
polynomial of degree k — 1 in n. Find the leading coefficient of
this polynomial.
Answer. The leading coefficient is k.
21. Given a sequence whose general term un is a polynomial
of degree k in n with leading coefficient a0, prove thatzlM„ is a
polynomial of degree k — 1 in it. Find the leading coefficient
of this polynomial.
Hint. Use the results of Problems 19 and 20.
22. Starting from a sequence

^1 >^2 >••• > >•••>

from the sequence of first differences

Au i = M2 - M j , z 1m 2 = m3 - u2, . . . , z 1m„ = Mn+ i - m„ , ...,

 12 Sequences and Combinatorial Problems

then the sequence of second differences

A 2ux — A u2 — A ul} A 2u2 — A u3 —A u2, ...,

A 2un = Aun+1 — Aun, ...

and so on, k times up to the sequence of kth differences

A kut = A k~1u2 —A k~ 1u1, A ku2 = A k 1x/3 —A k~1u2, ...,

O n ■n 0
A ku„ = A k~lun+l —A k~iu„, ...
Cki

n i n3
■O c
O t-
For example, if un = n3, then the first, second and third differ­
in
WO
■a ■a ences are as shown in Figure 2. Suppose un = nk. Show that all
i
i ! i the terms of the sequence of fcth differences equal the same num­
) 1
2 ! 8
3 27
)13
)12
)is
)6 ber. What is this number?
;37
it
5 125
64
)61
124 ) 6
J30 J ?
Answer. k\
J SI
6 | 216
23. Given a sequence {u„} = {nk}, prove that there is another
sequence {«„} such that
a) Au„ = v„;
Fig. 2
b) un is a polynomial of degree k + 1 in n.
What is the leading coefficient of this polynomial?
Hint. Use induction in &and the results of Problems 19 and 20.
24. Suppose a sequence {w„} is such that every term of the
sequence of fourth differences {zl4j/„} vanishes. Prove that un
is a polynomial of degree 3 in n.
25. Given a sequence {i/„} such thatzlwn is a polynomial of
degree k in n, prove that un is a polynomial of degree k + 1
in n.
Hint. Use the results of Problems 19 and 23.
26. Calculate the sum
l 2 + 22 + ••• + n2.

Hint. Use Problem 25 to show that the sum is a polynomial

of degree 3 in n.
n(n + 1) (2/i + 1)
Answer. --------------------- .
 Sample Problems 13

27. Calculate the sum

1 ■2 + 2 - 3 + ••• + zz (/z + 1).

n (n + 1) (zz + 2)
Answer. ---------- —--------- .
3
28. By an arithmetic progression is meant a sequence {w„}
such that
Mn+l d
for all n, where the number cl is called the difference of the
progression. By a geometric progression is meant a sequence {z/„}
such that
Un+1 <7wn>
where the number q is called the ratio of the progression.
a) Given an arithmetic progression {z/,,}, write a formula
for u„ in terms of Wj and the difTerer.ee d.
b) Given a geometric progression {z/„}, write a formula for un
in terms of ut and the ratio q.
Answer, a) w„ = zq + (n — 1) d; b) z/„ = qn~1u1.
29. a) Find a formula for the sum

S„ — Wi + u2 + + u„

of the first n terms of an arithmetic progression,

b) Find a formula for the product

P„ = z/,z/2 1/„

of the first n terms of a geometric progression.

Answer, a) Sn = /zz/j + — d; b) Pn — u1qnin~l)l2.
2
30. a) What is the sum of the first 15 terms of the arithmetic
progression with first term 0 and difference
b) What is the product of the first 15 terms of the geometric
progression with first term 1 and ratio \ / 10?
Answer, a) 35; b) 1035.
14 Sequences and Combinatorial Problems

31. a) The third term of an arithmetic progression equals 0.

Find the sum of the first five terms.
b) The third term of a geometric progression equals 4. Find
the product of the first five terms.
Answer, a) 0; b) 45 = 1024.
32. Let Sn be the sum of the first n terms of a geometric pro­
gression with first term ux and ratio q. Prove that

q - 1

Hint. Consider the quantity qS„ — S„.

33. A messenger arrives in a city of three million inhabitants,
bearing interesting news which he tells to two other people
10 minutes later. Then 10 minuter later each of these people
tells the news to two more people (who haven’t heard the news
yet) and so on, as long as there are people who don’t know
the news. Flow long does it take the whole city to learn the
news?
Answer. 3 hours and 30 minutes.
34. A bicyclist and a horseback rider compete in a five-lap
race. Both take the same time to complete the first lap, but the
bicyclist traverses each remaining lap 1.1 times more slowly
than the preceding lap. The horseback rider also slows down,
but by the same amount on each remaining lap. Both contestants
arrive at the finish line at exactly the same time. Who takes
longer to complete the fifth lap, and how much longer does he
take?
Answer. It takes the horseback rider 0.985 less time than the
bicyclist to complete the fifth lap.
35. Find the sum S of all odd numbers less than 1000.
Answer. S = 250,000.
36. Find the sum S of all three-digit numbers not divisible
by 2 or by 3.
Hint. Let denote the sum of all three-digit numbers, S (2>
 Sample Problems 15

the sum of all three-digit numbers divisible by 2, S (3) the sum

of all three-digit numbers divisible by 3 and S {6) the sum of all
three-digit numbers divisible by 6. Then S = S (1) — S {2)
- S (3) + S(6).
Answer. S = 164,700.
37. Suppose the sum of the first n terms of a sequence equals
Sn = 3n2. Prove that the sequence is an arithmetic progression.
Find the first term ux and the difference d of the progression.
Answer. ut = 3, d = 6.
38. Does there exist a geometric progression with the num­
bers 27, 8 and 12 as terms (these numbers need not appear
consecutively or in the order given here)? At what positions in
the sequence can these numbers appear?
Answer. Yes, such a progression exists. If 27 appears in the
mth place, 8 in the nth place and 12 in the pth place, then m, n
and p satisfy the equation m — 3p + 2n = 0.
39. Answer the same questions for the numbers 1, 2 and 5.
Answer. No such progression exists.
40. The squares of the twelfth, thirteenth and fifteenth terms
of an arithmetic progression form a geometric progression. Find
all possible values of the ratio of the geometric progression.
Answer. The ratio can take the values 1, 4, -y- + 1 Vl7,
i - W n .
41. Find all geometric progressions {w„} such that

= m„-i + »/„_ 2
for all n ^ 3.
Answer. Any progression with ratio \ + i^ /5 or —i^/5.
42. The terms of a certain sequence are sums of the corre­
sponding terms of two geometric progressions. What is the
third term of the sequence?
/'^ n ju W v T h c third term also equals zero.
43. Represent the terms of the Fibonacci sequence

1, »’2 = 1. = w„-i + "’„-2 (» ^ 3)

 16 Sequences and Combinatorial Problems

(see Example 4, p. 3) as sums of corresponding terms of two

geometric progressions {w„} and {r„}.
„ - ,1 + yf 5\ " /1 - y / 5 '
Answer. ——
V5
The remaining problems (44-94) are all of a combinatorial
Here is one way of
lighting the bulbs: character, i.e., they typically ask “how many ways are there of
Qt 999
Here is another
doing such and such?” Problems of this kind, involving the
calculation of numbers of distinct combinations, are of great
(the bulbs arealloff): importance in probability theory, computational mathematics,

ft f i t the theory of automata and mathematical economics.

^ 44. In how many ways can five light bulbs be lit?
Fig. 3 Hint. Each bulb can be on or olT. Two ways of lighting the
bulbs are distinct if they differ in the state of at least one bulb
(see Figures 3 and 4, where white and black indicate on and off
These ways are different: respectively). There are various approaches to the problem.

9f f f t One is to simply enumerate all possible ways of lighting the

bulbs, namely
fSo are9these:
fff 0) The bulbs are all off:
(1 way)
1) One bulb is on:
f t 9 1 9 o • • •
• o • •
919ft • • o •
Fig. 4 • • • o
• • • • o (5 ways)
2) Two bulbs are on (draw a figure and count the ways).
(? ways)
3) Three bulbs are on. (Don’t hurry. There is no need
to draw a separate figure or count. Think things over and
you will find that you already know the answer.) (? ways)
4) Four bulbs arc on (verify that the number of ways
is the same as in case 1). (5 ways)
5) The bulbs are all on:
o o o o o (1 way)
 Sample Problems 17

Another approach goes as follows: Suppose there is one bulb

instead of five. Then how many ways can the bulb be lit? Next
suppose there are two bulbs, three bulbs, etc. How many times
does the number of ways of lighting the bulbs increase each time
an extra bulb is included?
Answer. There are 32 ways of lighting five bulbs.
45. Given n light bulbs, let Cl denote the number of ways of
turning on k bulbs (k = 1, ..., n). Thus, in the preceding prob­
lem we found that Co = 1 while Cf = C l = 5. Prove that
Cn0 + C\ + ■■■ + Cnk + - + Cnn = 2".
Hint. Using the symbol Cl allows us to express the mathe­
matical content of the sentence “There are 10 ways of lighting
5 bulbs such that 3 bulbs are on” by the formula
Cl = 10.
A comparison of the two methods of solving Problem 44 leads
to the formula
Co5 + c? + c | + c | + C | + Cl = 25.
46. A city has n traffic lights, each of which can be in any of
three states (green, yellow or red). In how many ways can the
traffic lights be lit?
Answer. 3".
47. How many ways can n traffic lights be lit if k lights can
be in any of three states (green, yellow or red), while n — k
lights can be in either of two states (green or red)?
Answer. 3k2n~k.
48. What is the largest number of distinct license plates
consisting of 3 letters followed by 4 digits?
49. How many six-digit numbers are there containing no
zeros or eights?
Hint. Suppose you already know the number of five-digit
numbers containing no zeros or eights. Now can you calculate
the number of six-digit numbers with the same property?
Answer. 86 = 262,144.
18 Sequences and Combinatorial Problems

50. In a certain country, no two people have the same set of

teeth, i.e., one person may have a full set of teeth, one may be
missing the lower left molar, another the upper right molar, and
so on. What is the maximum population of the country?
Hint. There are 32 teeth in a full set.
51. Imagine that the expression
(x - \ ) ( x - 2) ••• (x - 100)
is multiplied out and all terms involving x to the same power
are combined. What is the coefficient of x:99?
Answer. —5050.
52. How many ways are there of writing a positive integer
n ^ 2 as a sum of two positive integers? There are two ways of
interpreting this problem, i.e., with or without the assumption
that the order of terms is important. In other words, it can be
assumed that 8 = 3 + 5 and 8 = 5 + 3 are two ways of
writing 8 as a sum of two terms or just one way. The answers
differ in the two cases. Solve both problems.
Answer. There are n — 1 ways if the order is important.
Otherwise there are \n ways if n is even and \ {n — 1) ways if n
is odd.
53. How many ways arc there of writing a positive integer
« ^ 3 as a sum of three positive integers if order is important?
A,m -er. (" ~ * > ~ 2) .

54. By Pascal's triangle is meant the array

1
1 1
1 2 1
\ /
1 3 3 1
V 7
1 4 6 4
V ' \ /
1 5 10 10 5 1
 Sample Problems 19

where the sides of the triangle are made up of ones and each of
the other numbers is the sum of the two numbers appearing
above it (to the left and right), as indicated by the dashed lines.
Show that the sum of the numbers in the (n + l)st row of Pas­
cal’s triangle equals 2n.
Hint. Show that the sum of the numbers in the (« + l)st row
is twice as large as the sum of those in the nth row.
55. What is the largest number of bishops that can be put on
a chessboard without any two bishops threatening each other?
Prove that this number of bishops can be put on in iV different
ways, where N is a perfect square.
Hint. To understand the problem, recall that in chess a bishop
moves along a diagonal. For example, a bishop on the square d3
abcdef gh
can go to any of the squares shown in Figure 5 (bl, c2, etc.) in
a single move. Thus two bishops threaten each other if they Fig. 5
occupy the same diagonal.
Warning. The problem is sometimes “solved” by putting ■ lE JiD lll
14 bishops on a chessboard as shown in Figure 6. It is then
claimed that if more than 14 bishops are put on the board, then
two will necessarily threaten each other. This seems plausible, but
not completely convincing. In any event, this approach does not
lead to a solution of the second (and basic) part of the problem. ID lD lD lU
56. A mother has two apples and three pears. Every day for Fig. 6

five consecutive days she gives out one piece of fruit. How many
ways can this be done?
Answer. Cf = 10.
57. Answer the same question for k apples and n pears (given
out one at a time over the course of n + k days).
Answer. C * " . This answer will do for now, if you are unable
to write an explicit formula for C*+n in terms of n and k. Later
on, we shall find such a formula (see Probs.60, 66 and 67).
58. Answer the same question for 2 apples, 3 pears and 4 or­
anges (given out one at a time over the course of 9 days).
59. In how many ways can 8 rooks be put on a chessboard
without any two rooks threatening each other?
 20 Sequences and Combinatorial Problems

Hint. A rook can move horizontally or vertically. For ex­

ample, the rook on square d3 in Figure 7 threatens all squares
in file d and row 3. Two rooks cannot be put in the same file
(they would threaten each other), i.e., only one rook can be put
in each file. But there are 8 files and 8 rooks. Hence there is
abcd'efgh
exactly one rook in any given file. On how many different
squares can this rook be placed?
Fig. 7
60. How many distinct ways are there of choosing 2 objects
from among n objects?
Hint. This quantity (also equal to the number of ways of
turning on 2 bulbs out of n) has previously been denoted by Cl.
Therefore the problem is to calculate Cl (for example, it has al­
ready been shown in Prob.56 that C\ = 10).
. ~n n (n - 1)
Answer. C2 = ----------- .

61. In how many ways can a class be seated if there are 26 stu­
dents and 28 seats?
62. A father has five apples, all of different sizes, which he
gives to his eight sons in such a way that each son either gets a
whole apple or nothing. In how many ways can this be done?
8 8!
Answer. C 1 • 5! = — .
3!
63. Let C2 denote the number of ways of lighting n bulbs
such that k bulbs are on. Prove that Cnk is the (k + l)st number
in the (n + l)st row of Pascal’s triangle (defined in Prob. 53).
64. In how many ways can four rooks be placed on a chess­
board without any two threatening each other?
65. There are 3 officers and 40 men in a company. In how
many ways can a patrol consisting of 1 officer and 3 men be
chosen?
66. In how many ways can 3 objects be chosen from n ob­
jects (without regard for order)?
~ 0 (n ~ 2)
A
Answer. C3 = w(«
r"
---------- .
1-2-3
 Sample Problems 21

67. In how many ways can k objects be chosen from n objects

(without regard for order)?
Hint. Use the method of the preceding problem or mathe­
matical induction.
, ^n n{n - 1) ••• (n - k + 1)
Answer. Ck = ------------------------------- .
1 -2 - Ik
68. In how many ways can six different consecutive notes be
played on a piano? How many six-note chords are there?
Hint. A piano has 88 keys. In a chord all notes are played
simultaneously.
<59. How many terms are obtained after multiplying out the
expression
(1 + o ) ( l + b )(l + c )(l + d ) ( l + e ) ( l + / ) ( 1 + g )?
Answer. 27.
70. How many of the terms obtained in the preceding problem
are products of three letters?
„
Answer. ^ 3 = ---------
C 7 ' 6 ' 5 = 35.
^
1 -2 -3
71. How many terms are obtained if we multiply on the ex­
pression
(1 + x + y) 2 0

without combining similar terms?

Answer. 3i0.
72. Suppose the expression
(\ + x 5 + x 1)20

is multiplied out and similar terms are combined. What are the
coefficients of x 11 and x 18?
73. Suppose the expression
(1 + x)56

is multiplied out and similar terms are combined. What are the
coefficients of x 8 and x-48?
22 Sequences and Combinatorial Problems

Answer. The coefficient of x8 is Cg6, and that of x48 is

✓ -*56 _ 56
^48 — <^8 *
74. Prove that

(1 + x)n = Cn0 + Clx + C \x2 + ••• + CnV ,

a special case of the binomial theorem (see Prob.77). Corre­

spondingly, the numbers Cl (k = 0, 1, are called the
binomial coefficients.
75. Let a0, a , , a2, ..., an be the numbers in the (n + l)st
row of Pascal’s triangle. Then it follows from Problems 63 and 74
that
(1 + x)n = a0 + arx + a2x 2 + ••• + a ^ x ' ’" 1 + a„xn.

Give another proof of this fact, starting directly from the defi­
nition of Pascal’s triangle.
76. Suppose k zeros are inserted between every pair of con­
secutive digits of the number 14641. Prove that the resulting
number is a perfect square.
77. Suppose the expression

(o + b)n

is multiplied out and similar terms are combined. Find the term
involving ak. Prove the binomial theorem

(ia + b)n = Cla" + Clabn~l + Cla2bn~2

+ ••• + Clakbn~k + ••• + O " .

Hint, (a + b)n = a" ^1 + —^ .

78. In the expression (x + y + z)n find the term involving

x ky l .
Hint. First find all terms involving x*.
79. Calculate the sums

Co + Cl + Cl + •••
 Sample Problems 23
and
CI + C\ + Cns + •••.

Hint. Recall from Problem 45 that

2" = c nQ + c l + c n2 + ••• + c ;,

but this time prove the formula by setting x = 1 in the binomial

theorem
(1 + .Y)» = Cn0 + Clx + C2V + ■•• + CnV

(see Prob.74).
80. Find the sum of the coefficients of the polynomial ob­
tained when the expression

(1 + x - 3x2)1965

is multiplied out and similar terms are combined.

Answer. —1.
81. How many diagonals can be drawn in a convex polygon
with n sides?
Answer. Cl — n.
82. How many distinct even five-digit numbers can be formed
from the digits 0, 1, 3, 4 and 5 if no repetitions are allowed?
Answer. 42.
83. How many distinct ways are there of putting a penny, a
nickel, a dime, a quarter, a half-dollar and a silver dollar in two
pockets?
84. How many distinct divisors does the number 2■ 3 •5 -7 -11
have?
85. Prove that
C n + 1 + c n + 2 + . . . + C n +k = Q " + * + 1 _ ] .

86. How many ways are there of distributing 3 / j objects among

three people in such a way that each person receives n objects?
87. A room contains several people, of whom 6 know English,
7 know French and 6 know German. Two know both English
24 Sequences and Combinatorial Problems

and French, 4 know both English and German, 3 know both

French and German and one person knows all three languages.
How many people are there in the room? How many know only
English?
Answer. There are 11 people in the room, one of whom knows
only English.
88. A library has a certain number of subscribers, each of
whom has borrowed at least one book. Suppose that for every
set of k books, it is known how many people have borrowed
them at one time or another (here k ranges from 1 to n, the
total number of books in the library). How many subscribers
are there?
89. How many six-digit telephone numbers are there contain­
ing the combination 12?
Answer. 49401.
 Detailed Solutions

/. The problem is solved by using mathematical induction:

a) The formula
n2 (n + l)2
l 3 + 23 + 33 + ••• + n3 ( 1)
4
obviously holds for n = 1.
b) Suppose (1) holds for n = k, where k is an arbitrary posi­
tive integer, so that
13 -.3 -13 73 k 2(k + l)2 ...
I3 + 23 + 33 + ••• + k 3 = — --------— . (2)
4
Then we want to show that (1) also holds for n = k + 1, i.e., that
Ck + l)2 (A: + 2)2
l 3 + 23 + 33 + - + k 3 + (k + l)3
4
But
l 3 + 23 + 33 + ••• + k 3 + ( k + l)3
k 2(k + l)2 'k 2
+ (k + l)3 = (k + 1)2( — + k + 1
4
(k + l)2 (k + l)2 (k + 2f
(k2 + 4k + 4) =

Therefore if (1) holds for n = k, it also holds for n = k + 1,

and hence, by mathematical induction, (1) holds for all n.
2. First solution
a) The formula
1 1 I n
------ H-------- + ••• + --------------— (1)
1 -2 2 -3 n (n + 1) n + 1
obviously holds for n = 1.
3 Silverman V
25
26 Sequences and Combinatorial Problems

b) Suppose (1) holds for n = k, where k is an arbitrary posi­

tive integer, i.e., let

1 1 I k
------ + ------- + ••• + ----------- = ----------.
1 -2 2 -3 k (k + 1) k + 1

Then (1) also holds for n = k + 1. In fact,

1 1 1 , 1
------ +------- + ••• + ------------ H---------------------
1 -2 2 -3 k (k + 1) (k + \)(k + 2)

k 1 k { k + 2) + \
~ k + 1 (k + 1)(k + 2) “ (fc+ 1)(A: + 2)

k 2 + 2k + 1 (k + l)2 k + 1
~ (k + 1)(k + 2) ~ (k + 1)(k + 2 ) ~ k + 2 '

Therefore, by mathematical induction, (1) holds for all n.

Second solution. By inspection,

1 _ _1_ _ 1
n (n + 1) n n + 1

for any positive integer n. Therefore the left-hand side of (1)

can be written as

--------h ------ H-------- + + + -----------

1 -2 2 -3 3 -4 (/i - 1) /; /;(/;+ !)
 Detailed Solutions 27

Since all the terms except the first and the last cancel, this sum
reduces to
1 1 1 1
+ + ••• + = 1-
1 -2 2 -3 n (n + 1) n + 1 n + 1
as in the first solution.

3. First solution
The formula
1 1 1
+ + + ( 1)
1 -4 4 -7 (3n — 2) (3n + 1) 3n + 1
is easily verified by mathematical induction:
a) The formula holds for n = 1.
b) Suppose (1) holds for an arbitrary positive integer n = k,
i.e., let
1 , 1 1 k
-------- 1--------- + ••• H------------------------- = ------------ .
1 -4 4 -7 (3k — 2) (3k + 1) 3k + 1
Then (1) also holds for n = k + 1. In fact
1 1 1 1
+ + ••• + +
1 ■4 4 •7 (3k - 2) (3k + 1) (3k + 1) (3k + 4)
k 1 k (3k + 4) + 1
+
3k + 1 (3k + 1) (3k + 4) (3k + 1) (3A: + 4)
_ 3k2 + 4k + 1 _ (3k + 1) (k + 1) _ k + 1
~ (3k + 1) (3k + 4) ~ (3k + 1) (3fc + 4) “ 3k + 4 '
Therefore (1) holds for all n.

Second solution
Starting from the identity
_______ 1_______ __ 1___________ 1
(3/z - 2) (3/z + 1) ~ 3(3/2 - 2) 3(3/2 + 1)’
 28 Sequences and Combinatorial Problems

we write the left-hand side of (1) as

1 1 1 1
+ ------+ --------- + +
1 -4 4 -7 7-10 (3n - 5) (3n - 2)
1 1 1 \
(3/2 - 2) (3n + 1 ) \ 3 ■1 3- 4,
1 1 A f 1 1
+ +
3-4 3-7 3-7 3- 10
1 1 1 1
+ +
,3 (3 /2 -5 ) 3(3/i - 2 ) , ^3 (3// — 2) 3(3n+1),
Since all the terms except the first and the last cancel, this sum
reduces to
1 , 1 1
-------- h ------ + ••• + ----------------------
1 -4 4 -7 (3// - 5) (3// - 2)
- 1 _ 1 _ n
~ 3 3(3« + 1) ~ 3/2 + 1 ’
as in the first solution.
4. One way of solving the problem involves the use of mathe­
matical induction:
a) The formula
1 1 1
------------- + ------------------------ + ••• + --------------------------------
a (a + 1) (a + l)(a + 2) (a + n - 1) (a + n)
n
0)
a {a + n)
obviously holds for n = 1.
b) Suppose (1) holds for an arbitrary positive integer n k.
so that
1 1
------------- H -------------------
a (a + 1) (a + l)(a + 2) (a + k — \)(a + k)
_ ^
a (a + k)
 Detailed Solutions 29

Then (1) also holds for n = k + 1. In fact,

1 1 ________ 1________
a (a + 1) {a + l)(fl + 2) (a + k - l)(o + k)
1
+ --------------------------
(a + k) (a + k + 1)
k 1
--------------- 1---------------------------
a{a + k) (a -{■ k) (a + k + \)
_ k(a + k + 1) + a _ k 1 + ak + a + k
a(a + k)(a + k + 1) a (a + k) (a + k + 1)
_ {a + k) (k + 1) _ k + 1
a(a + k)(a + k + 1) a(a + k + 1)
Therefore (1) holds for all n.
5. We again use mathematical induction:
a) The formula

1 • 1! + 2 * 2 ! + ••• + / ; • « ! = (n + 1)! — 1
!
holds for n = 1. 1
b) Suppose (1) holds for n = k, i.e., 1

1 • 1! + 2 • 2! + ••• + k • kl = (k + 1)1 - 1.

Then (1) also holds for n = k + 1. To see this, we merely note

that 1 • 1! + 2 • 2! + ••• + k • k\ + (k + 1) (k + 1)!
= (k + 1)! - 1 + (k + 1) (k + 1)! = (k + 2) (k + 1)! - 1.

BUt (k + 2)(k + l)\ = (k + 2)\

by the definition of «!, and hence

1 • 1! + 2 • 2! + ••• + (k + 1) (k + 1)! = (k + 2)! - 1.

It follows that (1) holds for all n.

30 Sequences and Combinatorial Problems

6. First solution
a) The formula

' i - r i - i = ^ ( 1)
h : n2 J In
obviously holds for n = 2.
b) Suppose (1) holds for any integer k ^ 2, i.e., let

K )('-i) ('-4 -4 4
Then (1) also holds for n = k + 1. In fact
1 1
.-I ' i-T 1- 1
4j k 2j \ (k + i y ,
k + 1 1 \ ( H 1 ) [ ( H 1 ) 2 -1]
1-
2fc \ (fc + l)2/ 2k (k + l)2
(A: + 1) k (k + 2) k + 2
2k (k + l)2 2 (k + 1)
Applying mathematical induction, we find that (1) holds for
all n ^ 2.
Second solution
Starting from the identity
, 1 (n - 1) (n + 1)

we write the left-hand side of (1) as

1
1-1 ------
' 4 ) ■4 ) 0 - 4 (» - I)2
(2 - 1) (2 + 1) (3 - 1) (3 + 1) (4 - 1) (4 + 1)
22 42
(n — 2) n (n - 1) (n + 1)
(n - l)2 n2

J s-rr - I
 Detailed Solutions 31

Since all the factors except 2 — 1 = 1 and n + 1 in the numer­

ator and 2 and n in the denominator cancel, this product redu­
ces to

as in the first solution.

7. Use mathematical induction:
a) The formula
1 - 22 + 32 -------+ ( - I ) " - 1 /12 = ( - 1 )" -111- +- (1)
2
clearly holds for n = 1.
b) Suppose (1) holds for any positive integer k , so that

1 - 22 + 32 - ••• + ( - I ) * '1 k 2 =
2
Then (1) also holds for n = k + 1, since

1 - 22 + 32 -------+ ( —I)*"1 k 2 + ( - 1 )k (k + l)2

= (_ !)* -! i f c L i l + (_1)*(/C+ 1)2

2

= ( - 1 ) — (* + - * - 1

= ( - ! ) » - ■ ( * + 1) h _ . (k + 1) (fc + 2)
2 2

Therefore (1) holds for all n.

S. Assuming that n lines have already been drawn, we draw
the (ti + l)st line. This line intersects the other n lines in just
n points, since by hypothesis it intersects each of the other lines
but does not intersect any two of them in the same point. It is
clear that these n points of intersection divide the (n + l)st line
into n + 1 segments (two of which are infinite). Each of these
segments divides an “old” part of the plane into two “new”
32 Sequences and Combinatorial Problems

parts. Therefore the n + 1 segments together create 2n + 2 new

parts out of 72 + 1 old parts of the plane, i.e., drawing the
(n + l)st line increases the number of parts of the plane by
72+1.
It is now an easy matter to calculate the number of parts into
which 72 lines divide the plane. One line divides the plane into
2 parts, drawing a second line increases the number of parts
by 2, drawing a third line increases the number of parts by 3,
and so on. Therefore the total number of parts equals

2 + 2 + 3 + 4 + ---+72 = l + (l + 2 + 3 + ••• + 72)

_ 1 + 11 +
2

(recall Example 1, p. 4). Therefore n lines divide the plane

into
j 72 (72 + 1) _ 722 + 72 + 2
2 ~ 2
parts.
9. The proof is by mathematical induction:
a) For 72 = 8 we have 8 = 3 + 5.
b) Suppose k is a sum of threes and fives. Then this sum
either contains a five (possibly several) or it does not. In the
first case, replace a five by 2 threes. Then the new numbers are
again all threes and fives and add up to k + 1. In the second
case, there are at least 3 threes, and we can replace 3 threes by
2 fives. The new numbers are again all threes and fives and
again add up to k + 1. In either case, if k can be written as a
sum of threes and fives exclusively, so can k + 1. It follows by
mathematical induction that any integer greater than 7 equals
a sum of threes and fives.
10. The proof is again by induction:
a) If 72 = 1, so that the plane is divided by just one line, then
there is an obvious regular coloring. In fact, we need only color
one half-plane white and the other half-plane black.
 Detailed Solutions 33
b) Suppose there is a regular coloring of the plane divided
by n lines, and draw the (n + l)st line. This line divides the
plane into two half-planes. In one of these half-planes, change
the color of every part (i.e., change white to black and black to
white), and in the other half-plane leave the color of every part
the same. Then the new coloring is regular. To see this, consider
any two adjacent parts of the plane divided by n + 1 lines. The
line separating the two parts is either one of the old lines or
else it is the new (n + l)st line. In the first case, the two parts
had different colors before drawing the new line and changing
colors in one of the resulting half-planes. Therefore they either
change c^lor or they do not, but in any event they still have
different colors. In the second case, the pieces had the same
color before drawing the new line, but after drawing the new
line one of the pieces has its color changed with the result that
the two pieces now have different colors.
Thus if the plane divided by n lines can be regularly colored,
so can the plane divided by n + 1 lines. It follows by mathe­
matical induction that given any positive integer n, there is a
regular coloring of the plane divided by n lines.
11. We must verify the assertion that
n3 + (n + l)3 + (n + 2)3
is divisible by 9 for every positive integer n. As usual, the proof
is in two steps:
a) The assertion is true for n = 1, since l 3 + 23 + 33 = 36
is divisible by 9.
b) Suppose the assertion is true for a positive integer n = k,
so that A:3 + (k + l)3 + (k + 2)3

is divisible by 9. Then the assertion is true for n = k + 1. In fact

(k + l)3 + (k + 2)3 + (k + 3)3
= (k + l)3 + (k + 2)3 + (A:3 + 3 • 3k2 + 3 • 32A: + 33)
= [A:3 + (A: + l)3 + (A: + 2)3] + [9 (k2 + 3k + 3)],
34 Sequences and Combinatorial Problems

where both terms in brackets are divisible by 9, the first by

hypothesis, the second since it is obviously a multiple of 9.
Therefore the sum of the terms is also divisible by 9, i.e., the
assertion is true for n = k + 1. It follows by mathematical
induction that the assertion is true for all n.
12. This time the induction goes as follows:
a) For n = 0 the number
i r + 2 + 122»+ i = i p + 12 = 121 + 12 = 133

is trivially divisible by 133.

b) If the number l l n+2 + 122n+1 is divisible by 133 for some
n — k, then it is divisible by 133 for n — k + 1. In fact
1J(fc+ 1) +2 J22(fc+1) + 1 _ 11fc+3 _j_ J22fc +3

= 11 • 11* +2 + 144 • 122fc+1

= 11 • 11* +2 + 11 • 122k+1 + 133 • 122fc+1
= 11 (11A+2 + 122ft+1) + 133 • I22k+1,
where the last expression on the right is obviously divisible by
133, being the sum of two terms divisible by 133 (the first by
hypothesis, the second by inspection). Therefore 11"+2 + 122n+1
is divisible by 133 for n = k + 1. It follows by mathematical
induction that l l n+2 + 122n+1 is divisible by 133 for any posi­
tive integer n.
13. Use mathematical induction, starting from n = 2:
a) The inequality
(1 + a)n > 1 + na (1)

obviously holds for n = 2. In fact

(1 -f- a)2 = 1 -f- 2a -J- a2 > 1 + 2a,

since a2 > 0 if a ^ 0.
b) Suppose (1) holds for some positive integer k ^ 2, i.e., let
(1 + a)k > 1 + ka. (2)
 Detailed Solutions 35

Then (1) also holds for n = k + 1. To see this, note that 1 + a > 0
by hypothesis and multiply both sides of (2) by 1 + a, obtaining
(1 + a)k (1 + a) > (1 + ka) (1 + a)
°r (1 + a)k+1 > [1 + (k + I) a] + ka.
But ka2 > 0 if a ^ 0, and hence
(1 + a)k+l > 1 + (k + I) a.
It now follows by mathematical induction that (1) holds for
all n ^ 2 if a > —1, a ^ 0.
14. The proof (again by induction) goes as follows:
a) The inequality
1 1
1+ + + + Jr > ^ ( 1)
V2 73 V”
holds for n = 2, since then it becomes

1+
- > y/2
>/2
which is equivalent to the obvious inequality
y/2 + 1 > 2.
b) Suppose (1) holds for n = k, so that
1
. + -^ + - U ( 2)
n/2 ^3 + 7 i > '/ k '
Then (1) also holds for n = k + 1. This is an immediate conse­
quence of the inequality
1
"> \ J k 1 — \jky (3 )
Jk+ 1
as we see by adding (2) and (3). To verify (3), we merely multiply
both sides by s/k + 1 4- yjk, obtaining the inequality
36 Sequences and Combinatorial Problems

which is obviously true. The validity of (1) for all n ^ 2 now

follows by mathematical induction.
15. The result follows by mathematical induction, once we
note that
a) u„ — 2"-1 + 1 for n = 1 and n = 2.
b) If ti„ = 2n~ 1 + 1 holds for n ^ k (k ^ 2), then it also
holds for n = k + 1. In fact
wk+1 = 3uk - 2uk- t = 3 (2k-1 + l ) - 2 ( 2 ‘- J + 1)
= 3 •2k- 1 + 3 - 2*"1 - 2
= 2 • 2k“ 1 + 1 = 2k + 1.
16. Again the result follows by mathematical induction, after
noting that
a) un — (2n — l)2 for 72=1.
b) If un = (2/7 — l)2 holds for n = k, then it also holds for
n = k + 1. In fact
i/k+1 = uk + 8fc = (2k — l)2 + 8fc = 4k 2 — 4k + 1 + 8fc
= 4k2 + 4k + \ = (2k + l)2 = [2 (k + 1) - l]2.
17. Au„ = u„+1 - u„ = (n + l)2 - n2 = 2n + 1.
18. Since obviously
Mi + (m2 - Mt) + (m3 - z/2) + ••• + (u„ - un- 1) = u„
(all the terms cancel except the last), the formula
ux + Au i + z1m2 + ••• + Au„-i = u„ (1)
is an obvious consequence of the definition of the first differ­
ences. Using (1) and the analogous formula for v„, we have
u„ — v„ = (mj + Auy + A u2 + ••• + ^lw„-i)
- (t?! + A vi + A v2 + ••• + A v„ -X)
= Mj — vk + zIm j — Av i + z1m2 — z l t ; 2 + •••

+ Au„-1 — Av„-i.
 Detailed Solutions 37
By hypothesis
Au j = Avy,A u2 = A v2, ...,Au„_ ! = A v„ -1,
and hence
lln ~ Vn = III - Wi­
lt follows that if ux — vy ^ 0, then w„ — v„ # 0, i.e., un differs
from t’„ for every /i. On the other hand, if u2 = v2, then un = v„
for all n.
19. It need only be observed that
Aw„ = (h'b+1 - u-n) = (wn+1 + vn+1) - (w„ + v„)
= (wn+ 1 - Mn) + (vn+l - V„) = Au„ + Avn.

20. Noting that

Aun = (n + 1)* - nk,
we first show that
(n + l)k = nk + knn~1 + (1)

where the dots denote terms in n of degree less than k — 1. This

is done by mathematical induction in k\
a) Formula (1) holds for k = 1.
b) Suppose (1) holds for k = k 0, i.e., let
(n + l)ko = n ° + k 0n'£°’ 1 + ...,
where the dots denote terms in n of degree less than k 0 — 1.
Then (1) also holds for k = k Q + 1. In fact
(n + l)k0+ 1 = (n + l)k° (n + 1)
= (nk0 + k 0n ° ~ l + ■••)(«+ 1)
= nko+l + k 0nko + ... + nko + k Qn ° ~ l + •••
= Hko+1 + (k0 + 1) «k° + ...,
where the last set of dots denotes terms in « of degree less than k 0.
Therefore (1) holds for all k.
38 Sequences and Combinatorial Problems

It now follows from (1) that

Au„ = (n + l)k —nk — nk + knk~l + ••• —nk

— knk~l + ...,
where the last set of dots denotes terms in n of degree less than
k — 1. In other words, Au„ is a polynomial in n of degree k — 1,
with leading coefficient equal to k.
21. Let
u„ = a0nk + a{nk~l + ••• + ak- x n + ak,

and regard the sequence {u„} as the sum of k + 1 sequences

{fliM*"1}, ...,{<!*-!«}, {fl*},

respectively. Then, according to Problem 19, {Au„} is the sum

of the first differences of these k + 1 sequences. Moreover, it
follows from Problem 20 that the general term of the first se­
quence is a polynomial of degree k — 1 in n with leading coeffi­
cient a0k, while the general term of the remaining sequences are
all polynomials of degree less than k — 1. Therefore Au„ is a
polynomial of degree k — 1 in n, with leading coefficient a0k.
22. If u„ = nk, then the fcth difference A ku„ = k \ This can be
seen by using induction in k :
a) If Ac = 1, then u„ = n and hence

Aun = u„+ i - u„ = (n + 1) - n = 1 = 1!

b) Suppose A ku„ = k\ for all k ^ k 0. Then A ku„ = kl for

k = k 0 + 1. To see this, consider the sequence {w„} = {nko+ 1}.
Then, as we know from Problem 20, Au„ is a polynomial in n
of degree k 0 with leading coefficient k 0 + 1, say

Au„ = (k0 + l)« k° + fli/i*0-1 + a2n °~ 2 + ••• + n + ak<3.

But the sequence {Au„} is the sum of the k 0 + 1 sequences

{(k0 + l ) ^ 0}, {fliW*0" 1}, {a2nk°~2}, ...,{ a to_ 1n},{ako},

 Detailed Solutions 39

+ ••• + z lkotf*0_ 1rt + A k°ako

= (k0 + 1) Zl*°ai*° + a! AAk°~1nk°~1 + a2A 2Ak°~ 2n °~ 2
+ ••• + ak0- i A k° 1An + Ak°ako 0)
(justify this equation!). Moreover, by hypothesis,

A knk = k\ for all k ^ k 0,

and hence (1) reduces

Ako+lu„ = (k0 + 1) k 0\ + a ,A ( k 0 - 1)! + a2A 2 (k0 - 2)!

+ - + a k0- l Ak° - 1 1 + Ak°a,
= (k0 + 1)A:0! = (k0 + 1)!

since all the other terms vanish (if the terms of a sequence all
equal the same constant, then its first differences, second
differences, third differences,etc. all vanish). Therefore A ku„ = k\
for k = k 0 + 1, and the proof by induction is complete.
23. The proof is by induction:
a) Let k = 0 so that v„ = 1 for all n. Then the assertion is
clearly true, the required sequence being just the sequence with
general term u„ = n. Let this sequence be denoted by {wj;0)}.
b) Suppose the assertion is true for all k ^ k 0, so that there

( 1)

To find uiko+1>, we start from the sequence with general term

n a+2. According to Problem 21, Anko+2 is a polynomial of
40 Sequences and Combinatorial Problems

degree k 0 + 1 in n with leading coefficient k 0 + 2. Therefore

*o + 2

\ko + 2 /
is a polynomial of degree k 0 + 1 in n with leading coefficient 1.
Let this polynomial be

nAc0n+ 1 +■ a0nkn0 +■ a^nkQ—1 + ••• +, ako.

Then for {i/*0+1)} we choose the sequence with general term

*o + 2
„ „<*0> „ „<*0-l) „ „<0)
*£°+1 - "■ *2jk0^n •
kn + 2

To verify that this sequence has the required properties, we use

Problem 19, obtaining
*0+2
Au(„k°+1) = a (-?-----A (a0u(ko)) - A ia tu ? * - " )
\ko + 2

--------- a oaky no))

= nko+l + a0nk° + a j/i*° 1 + ••• + ako

—a0n ° - aj/7*0" 1 - ••• - ako = n 0+\

which agrees with (1). The induction is now complete, i.e., for
every k there exists a sequence {w(t)} such that Auik) = nk.
Moreover, it is clear from the above construction that u<k) is a
polynomial of degree + 1 in n with leading coefficient 1j(k + 1).
24. Since A Aun = 0 by hypothesis, we have A 3u„ = c (for
all n) where c is a constant. Therefore A 2un is a polynomial of
degree 1 in n. Let this polynomial be a0n + aY. Then, by the
solution of Problem 18,

Aun = A ux + A 2ux + A 2u2 + ••• + A 2un- 1

= A u1 + Oj (n - 1) + a0 [1 + 2 + + (n - 1)].
 Detailed Solutions 41

But
n (n — 1)
1 + 2 + ■•• + (ti — 1) =
2
(see Example 1, p. 4), and hence
n 0n - 1)
Aun = Aui + <7j (n - 1) + a0

\x.,A u n is a polynomial of degree 2 in n. Let this polynomial be

b0n2 + b^i + b2. Then

un = uk + A u v + Au 2 + ••• + A u„ -l

= u i + b2 {ii —1) + bi —-------- + b0\ \ 2 + 22 + ••■+(/! — l)2] .

2
If Sn denotes the sum in brackets, then/lS1,, = n2 and hence by
Problem 23, Sn is a polynomial of degree 3 in n. Therefore u„
is also a polynomial of degree 3 in n.
25. Let
Aun = a0nk + aink~l + ••• + + ak.

and consider the sequence with general term

v„ = doun(«t) +i akun(Ik-1)> +, , ak-!Un(1) +i akuyn(0)\

+

where i/*10, wjj*” 15, ..., wj;15, i/£0) are the same as in the solution
of Problem 23. Then Avn = Au„ for all n. It follows from the
solution of Problem 18a that u„ — v„ = — vk for all n. There­
fore

un = (a0u™ + + ••• + ak- 1u<

'„1) + akw£0)) + («i - vt).

But, according to Problem 23, is a polynomial of degree

k + 1 in n, and hence so is un.
26. If
un = l 2 + 22 + ••• + n2,
then
= i/„+i - w„ = (n + l)2.
4 Silverman V
42 Sequences and Combinatorial Problems

is a polynomial of degree 2 in n. Therefore, by Problem 25, u„ is

a polynomial of degree 3 in n, say

un = a0n3 + axn2 + a2n + a3.

To find the coefficients a0, alf a2 and a3, we replace n by 0, 1,

2 and 3, obtaining
«o = a3 = 0,
«i = a0 + ax + a2 + a3 = 1,
u2 = 8a0 + 4a i + 2a2 + a3 = 5,
u3 = 21a0 + 9ax + 3a2 + a3 = 14.
Hence
(to + Wj + a2 = 1,
Soq + 4ax + 2a2 — 5,
21a0 + 9ax + 3a2 = 14,

and to solve this system we subtract three times the first equation
from the third equation and twice the first equation from the
second equation. This gives

24a0 6ax = 11,

6a0 + 2a i = 3,
which implies
6a0 = 2, a0 = ax — a2 =
It follows that
n (2n2 + 3n + 1)
u„ = + W + ^n =

n (n + 1) (2/i + 1)

i.e.,
n (n + 1) (2/i + 1)
I2 + 22 + ••• + n2 =
 Detailed Solutions 43

It should be noted that the answer can be verified by mathe­

matical induction if the answer is known in advance or if it has
been guessed somehow.
27. If
u„ = 1 ■2 + 2 • 3 + ••• + n (n + 1),
then
Au„ = (n + 1) (n + 2)

is a polynomial of degree 2 in n. It follows from Problem 25

that u„ is a polynomial of degree 3 in n, say

u„ = a0n 3 + a vn 2 + a 2n + a3.

To find the coefficients a 0 , a y , a 2 and a 3 , we replace n by 0, 1,

2 and 3, obtaining

uQ = a3 = 0,

Uy = a0 + ay + a2 + a 3 = 2,

u2 = 8a0 + 4ay + 2a2 + a3 = 8,

u3 = 27a0 + 9a y + 3a2 + a3 = 2 .
Hence
a O + Oy + 0 2 = 2 ,

8a0 4Oy + 2a2 — 8,

21a0 + 9ay + 3a2 = 20,

and to solve this system we subtract three times the first equation
from the third equation and twice the first equation from second.
This gives
24a0 + 6a! = 14,

6a0 + 2ay = 4,
which implies

6a0 = 2, a0 = i , ay = 1, a2 =
44 Sequences and Combinatorial Problems

It follows that
u„ = i /i3 + n2 + £ n
n (n2 + 3n + 2) n (n + 1) (/? + 2)
3 “ 3
i.e.,
n (n + 1) (n + 2)
1 -2 + 2 -3 + ••• + n ( n + 1) =
3
28. a) By the definition of an arithmetic progression,
u„ = un- 1 + d = un —2 + 2d = ••• = + {n - 1)d .
b) By the definition of a geometric progression,

Un = qun - 1 = q 2U„-2 = — = ^ n _ 1 » 1-

29. a) According to Problem 28a,

Sn — Uy + (mj + d) + ••• + [mj + (/i — 2) d]
+ [»i + (« - 1) d]. (1)
Writing this sum in reverse order, we have
Sn = [mi + (n - 1) d] + [?/i + (n - 2) d] + ■■■
+ (ui + d) + uY. (2)
We then add (1) and (2) term by term, obtaining

2-S'n = (mj + [«! + (/i — 1) d} + {(mj + d) + [Mj + {n — 2) d}

+ ••• + {[Mi + (n — 2) d] + (Mi + d)} + {[mx + (n — 1)*/] + Mj} .

There are n terms in curly brackets, each equal to 2uy + (n — 1)d.

Therefore
2Sn = n [2mi + (n — 1) d] = 2nuY + n{n — 1) d
or
 Detailed Solutions 45

b) According to Problem 28 b,
Pn = it 1 • qtty • q 2Ui ••• q"-2ul • q " - 1ux
_ u \ q l + 2 + - , + ( " - 2) + <"-!)_

But by Example 1, p. 4,
n (n — 1)
1 + 2 + ••• + (n - 2) + (n - 1) =

and hence
P„ = k V (" - |)/2.
30. a) Using the formula
„ n (n — 1) ,
S„ = nui H------------ d
2
found in Problem 29 a, we have
_ 15-14 1

5,5 “ 2 3 " 35-

b) Using the formula
p„ = «1y (" - 1)/2
found in Problem 29 b, we have
15-14 15-14
p 15 = (V io) 2 = io 2 * = io35

31. a) In terms of the difference d, we have

= u3 — 2d, u2 — Wj — d, 1/4. = u3 + d, u5 = u3 + 2d.

Therefore the sum of the first five terms equals

S 5 = («j - 2</) + (h3 - d) + u3 + («3 + d) + (m3 + 2d)

= 5»3 = 0.

b) In terms of the ratio q, we have

u 3 m 3 2
«i = — , «2 = — > uA. = qu3, us = q 2u3 .
<1 (1
46 Sequences and Combinatorial Problems

Therefore the product of the first five terms equals

Ps = — u3qu3q2u3 = i/f = 4s = 1024.

q q

32. It follows from

S„ = Uy + u2 + u3 + ••• + u„

= Uy + qUy + q 2Uy + ••• + q n~ l Uy

that

qS„ - Sn = q (Uy -f qUy + q2Uy + ••• + qn 2Uy + qn 1Uy)

- (Uy + qUy + q2Uy + q2Uy + ••• + qn~lUy)

= qUy + q2Uy + q2Uy + ••• + qn~1Uy + qnUy

- U y - qUy - q2Uy - q2Uy - ••• - qn~lUy

= qnui - Mi

(all but two terms cancel). Therefore

Sn (q - 1) = M, ( qn - 1) ,
i.e.,
q" - 1
Ml .
q - i

33. Ten minutes after the arrival of the messenger, 2 new

people learn the news, ten minutes later 4 more people learn the
news, and so on, with 2* new people learning the news 10k min­
utes after the messenger’s arrival. Therefore the number of
people who know the news after 10k minutes equals the number
who know the news when the messenger arrives (only the
messenger himself!) plus those who learn the news 10 minutes
later (2 people) plus those who learn the news 20 minutes later
(4 more people), and so on, up to and including the number
 Detailed Solutions 47

who learn the news 10k minutes later. This number is just
S k = 1 + 2 + 4 + ••• + 2 \
i.e., the sum of the first k + 1 terms of a geometric progression
with first term 1 and ratio 2, and according to the preceding
problem,

We must now find the smallest value of k for which S k (the

number of people who know the news after 10k minutes) ex­
ceeds 3,000,000 (the number of inhabitants of the city). This
value of k is 21, since

222 — 1 = 4,194,303.

Therefore the whole city knows the news after 210 minutes
= 3 hours and 30 minutes.
34. Suppose it takes the bicyclist a„ seconds to complete
the /ith lap (n = 1,2, 3, 4, 5), while it takes the horseback
rider b„ seconds. Then the numbers ait a2, a3, ax , a5 form a
geometric progression with ratio 1.1, while the numbers
bi, b2,b 3, f>4, 65 form an arithmetic progression with differ­
ence d. Moreover, by hypothesis, ak = and

a i + a2 + o3 + <74 + as ®= by + b2 + 63 + b^ + b$. (1)

Using the formulas found in Problems 29 and 32, we can write(1)

as
(1 .0 s - 1 at = 5at + 5 • 4
d.
1.1 - 1 2
which implies
Ol{10[(l.l)5 - 1] - 5} = 10d
or
d 10[(l.l)s - 1 ] - 5
(2)
a1 10
48 Sequences and Combinatorial Problems

We want to find the ratio

b$ ai 4d 1 4 d
Vs ~ ( i . i r ^ “ ( T I r + ( n r ^ 7 '

Substitution of (2) into (3) gives

h. 1 4 10[(1.1)5- 1] - 5
a5 (Hr (l.l)4 10
1
{1 + 4[(1.1)5 — 1] —2}
air
i 1 4420
[4(0.6105) - 1] = ------ 4- * 0.985 < 1 .
(Hr 1.4641

Therefore it takes the horseback rider 0.985 less time than the
bicyclist to complete the last lap.
35. Clearly 5 = l + 3 + 5 + --- + 997 + 999 is the sum of
the first 500 terms of an arithmetic progression with first term 1
and difference 2. Therefore

500 • 499
5 = 500 • 1 + ---------- 2 = 500 + 500 • 499 = 500 • 500
2
= 250,000.

36. Suppose that from S(1), the sum of all three-digit numbers
not divisible by 2 or by 3, we subtract S (2), the sum of all three-
digit numbers not divisible by 2, and S (3), the sum of all three-
digit numbers not divisible by 3. In so doing, the numbers divis­
ible by both 2 and 3, i.e., divisible by 6, have been subtracted
twice. Therefore to obtain the answer, we must add S(6), the
sum of all three-digit numbers divisible by 6, i.e.,

s = s (1) - s (2) - s (3) + s (6).

Each of the quantities S(1), S (2), S (3y and S (6) is the sum of a
certain number of terms of an arithmetic progression, as shown
 Detailed Solutions 49

in the following table, where i/j is the first term, n the number
of terms, and the difference:

n d

5 (1) 100 900 1

sw 100 450 2

5 (3) 102 300 3

s (6) 102 150 6

Therefore
900 ■899
5 (1) = 900 • 100 + ----------- 1 = 90,000 + 404,550 = 494,550,
2
450 • 449
S(2) = 450 • 100 + — ------ 2 = 45,000 + 202,050 = 247,050,
2

300 • 299
S(3) = 300 • 102 + — ----- - 3 = 30,600 + 134,550 = 165,150,
2
150 • 149
S(6) = 150 • 102 + —-------- 6 = 15,300 + 67,050 = 82,350,
2
and hence finally

S = - S i2) - S(3) + S(6) = 164,700.

37. Obviously the ;ith term of any sequence equals the sum
of the first n terms minus the sum of the first n — 1 terms, i.e.,

11n — 1.
In our case,

u„ = S„ — S„_j = 3n2 — 3 (/» — l)2 = 6/i — 3.

Moreover

1 = (6/» - 3) - [6(/» - 1) - 3] = 6
50 Sequences and Combinatorial Problems

for all n ^ 2. Therefore the difference between consecutive terms

of the sequence is constant and equals 6, i.e., the sequence is an
arithmetic progression with difference 6. Its first term is clearly
ul = S l = 3.
38. Suppose we have found a geometric progression with 27
as its mth term, 8 as its /ith term and 12 as its pth term. Let
denote the first term of the progression and q its ratio. Then

27 _
= q —1,*i > 8 = qn~1u1, 12 = qp- 1u1,

and dividing the first equation by the second and third gives
22 _ nm~n £ _ nm-p
8 V i 4 •
Raising the first of these equations to the power m — p and the
second to the power n — p, we equate their right-hand sides,
obtaining
(r r -p = d r -" (i)
which implies
2 3 m—3p —2m + 2/i _ —3p —2m + 2n

or
2>m —3p-\-2n _ 2 m—3p + 2 n
( 2)

Clearly (2) can hold only if m — 3p + 2n = 0 (why?).

Conversely, if m, n and p are distinct positive integers such
that m — 3p + 2n = 0, then (2) and (1) hold. Let
m—n / 2 7
q = V j •
Then
27 _ n m —n £ _ n m —p
= q » 4 = q (3 )

Let Ui be such that 12 = qp~lui (such a number always exists).

Then it follows from (3) that

8 = qn~lUl, 27 = qm~1Ui.

This means that the numbers 27, 8 and 12 are the mth, /;th and
pth terms of a geometric progression with first term ux and
 Detailed Solutions 51

ratio q. For example, if ux = 8, q = f, then in = 4 , n = 1,

P = 2.
39. Suppose we have found a geometric progression with 1
as its mth term, 2 as its nth term and 5 as its pih term. Let ux
denote the first term of the progression and q its ratio. Then

\ = q m- lult 2 = q"~lul , 5 = qp~luu

and dividing the second and third equations by the first gives

2 = qn- m, 5 = qp~m.

Raising the first of these equations to the power p — m and the

second to the power n —m, we equate their right-hand sides,
obtaining
2 p —m __ ^n —m

Clearly this relation can hold only if p —m = 0 and n — m = 0

(why?). But this is impossible, since m, n and p must obviously
be distinct. Hence there is no geometric progression with 1 as
its inth term, 2 as its nth term and 5 as its pth term.
40. Let Mi2, ui3 and u15 denote the twelfth, thirteenth and
fifteenth terms of the arithmetic progression. Then, by hypo­
thesis, the numbers u \2, wf3, u \5 form a geometric progression,
i.e.,
2 2
“ 13 _ »15
2 2 ‘
Ml2 U, 3

Taking the square root of both sides of this equation, we find

that either
U 13 __ U IS

U I2 U 13

or
»13 _ _ £ lS _
( 2)
»12 «1 3
52 Sequences and Combinatorial Problems

First wc consider (1). For any arithmetic progression with

difference d, we have

»i3 = ui 2 + "is = «i3 + 2 d = ul2 + 3d,

and hence
m12 + d m13 + 2d
*12 ‘ 13
I.C.,
i d 2d
1 + ----- = 1 + -----
*12 ‘ 13
or
2d
(3)
U 12 U 13

Assuming that d ^ 0, we find from (3) that

!h±. - o — - 4
u1 2 Ml 2

i.c., the ratio of the geometric progression is 4.

Next wc consider (2). In this case,

Mi 2 + d *<i3 + 2d
‘ 12 ‘ 13
I.C.,
i + = - 1— —
Ml 2 ‘ 13
or
„ d 2d
2 + -----+ ------ = 0.
111 2 111 3

It follows that
2mi2m13 + '/«i3 + 2r/M12 = 0
or
2u12 + Sdu12 + d^ = 0.
 Detailed Solutions 53

Therefore
d \2 d „
----] + 5 ------ + 2 = 0,
<12, «12
so that
- 5 ± V 17
»12 2

There are now two possibilities:

J _ 5 V 17
a)
Wi2 2 2

_W]3. _ 1 + _ 3 [ y fiT
>12 u 12

»13
<7 =
>«12 2 2

5^ V 17
b)
«12 2

«13 3_ V 17
= 1+
u 12 «12 2
\2
»13
q = = — + — V l7 .
‘ 12 2 2

In other words, (2) implies that q equals either y + ^ Vl7 or

T -W l7 .
There still remains the possibility that d = 0, but in this case
it is obvious that q = 1.
4 /. If {«„} is a geometric progression such that

w„ = w„-i + w„_2 (1)

for all ii ^ 3, then, by Problem 28b,

qn+1u1 = qnUi + qn 1ul

54 Sequences and Combinatorial Problems

for all n ^ 1. It follows that

q2 = q + I
and hence
1+
? = -----;

Conversely, it is easy to see that (2) implies (1).

42. Let the first progression have first term Uy and ratio p,
while the second progression has first term ty and ratio q.
Moreover, let the given sequence be {w„}. Then
Wy = w, + Vi = 0 , (1)
U’2 = puy + qv 1 = 0 , (2)
by hypothesis. But (1) implies

= -M i,
and then (2) in turn implies
puy - quy = (p - q )u i = 0
or p = q. It follows that
w3 = p 2Uy + q2Vy = p 2Uy - q2Uy = q2Uy - q2Uy = 0,
i.e., the third term of the given sequence also equals zero.
42. By hypothesis,
Uy + Vy = \, «2 + v2 = 1
and
Mn + V„ = U„-y + V„-y + m„_2 + v„-2. (1)
Let p and q be the ratios of the progressions {w„} and {r„}.
Clearly p ¥= q, since the sum of two geometric progressions
with the same ratio is itself a geometric progression and the
Fibonacci sequence is not a geometric progression. Equation
(1) can be written in the form
p"~1Uy + q"~1Vy = P"~2Uy + q"~2Vy + P"~3Uy +
 DeatUed Solutions 55

wiPn 3 {P2 - P - 1) = ~Viq" 2 (q2 - q - I). (2)

It follows that
p 2 - p - l = 0,
q2 — q — 1 = 0 .
In fact, suppose this is not true and let q2 — q — 1 / 0, say.
Then, dividing both sides of (2) by
vlpn~2(q2 - q - 1),
we obtain
H, P2 - P - l =
(3 )
vi q 2 - q - i \p)
for all n > 3. Since the left-hand side of (3) does not depend
on n, neither can the right-hand side. But this can be so only if
p = q, and, as just noted, p / q. This contradiction shows that
the assumption q2 — q — 1 / 0 is false. Similarly, it can be
shown that p 2 — p — 1 = 0. It follows that p and q are distinct
roots of the equation x 2 — x — 1 = 0. This equation has roots

1 ± y/5
2
and hence _
1 + %/5 1 - y/5
p = — ;— . q = — 7— >
2 2
say.
Finally, we find and vu using the pair of equations
i/i + t>i = 0,

1 + V5 1 - V5
+ = 1,
2
with solution
1 + y/5 _ - 1 + y/5
Ml
2V5 ’ ^ 2V5
56 Sequences and Combinatorial Problems

Therefore
i + \/5 / 1 + \/5 y - 1 = i / i + Vi y
u„ = pn~1ul
2 \ls \ 2 / V il 2 ) ’
-1 + V5 / 1 - s/sY ' 1
vn = qn~lVy
2^5 \ 2 )
and hence
n
1 1 + V5 i - V i VI
+ vn = —=
v5 2 2
44. First solution
The number of ways of lighting the bulbs such that one of
the five bulbs is on equals the number of ways such that one
bulb is off or equivalently such that four bulbs are on. In each
case, there are 5 ways in all (the symbol ~ means “corresponds
o • • • • • o o o 0
• o • • • o • 0 o o
• • o • • 0 o • 0 o
• • • o • o o o • o
• • • • o o o o o •
Similarly, the number of ways of lighting the bulbs such that
two bulbs are on equals the number of ways such that two bulbs
are off or equivalently such that three bulbs are on. In each
case, there are 10 ways in all:
o o• • • ~ • • o o o
o •o • • ~ • o • o o
o •• o • ~ • o o • o
o •• • o ~ • o o o •
• o o » » ~ o » » o o
• o » o » ~ o » o » o
• o » » o ~ o » o o »
• • o o » ~ o o » » o
• • o » o ~ o o » o *
• • • o o ~ o o o » *
 Detailed Solutions 51

There are two further ways of lighting the bulbs, namely all the
bulbs are on
o o o o o
or all the bulbs are off

This exhausts all the possibilities and shows that in all there are
5 + 5 + 1 0 + 1 0 + 1 + 1 = 32
ways of lighting the bulbs.
Second solution
1) Suppose there is only one bulb. Then the bulb can either
be off • or on o, i.e., there are two ways of lighting the bulb.
2) Suppose there are two bulbs. The first bulb can be in two
states • or o. Each of these two states can be combined with the
two states of illumination of the second bulb. If the second bulb
is off, there are two states

while if the second bulb is on, we have

giving a total of 4 ways of lighting the bulbs:

o •
• o
o o
3) Three bulbs. The first two bulbs can be in four states, and
each of these 4 states can be combined with either of the two
states of the third bulb:
• •
o • o •
o
• o • o
o o o o
5 Silverman V
58 Sequences and Combinatorial Problems

This gives a total of 8 ways of lighting the bulbs:

• • •
o • •
• o •
o o •
• • o
0 * 0
• o o
o o o

4) Four bulbs. The first three bulbs can be in 8 states, and

each of these states can be combined with either of two states
of the fourth bulb, i.e., there are 8 • 2 = 16 ways of lighting the
bulbs.
5) Five bulbs. There are now 16 • 2 = 32 ways of lighting the
bulbs.
The method just described can be modified slightly. For ex­
ample, another way of describing the transition from two to
three bulbs is

<:
o
<:
• o \/
o

o o<
xo

In any event, adding another bulb (which can be in either of two

states) doubles the number of ways of lighting the bulbs.
45. Let D„ denote the total number of ways of lighting n bulbs.
Then Dj = 2D j-1, since adding an extra bulb doubles the num­
ber of ways of lighting the bulbs (theyth bulb can be off or on).
 Detailed Solutions 59

But obviously = 2. Therefore

D„ = 2 D = 22D„-2 = - = V -'D y = 2" (1)

(this is essentially a proof by induction, anticipated in the second

solution to Prob.44). On the other hand, D„ can be calculated
by simple enumeration of the possibilities:
0) No bulbs are on. This can occur in Cq ways. But obviously
there is only one way in which no bulbs are on, regardless of
the value of n. Therefore Co = 1 for every n.
1) There are Cl ways in which one bulb can be on.
2) There are Cl ways in which two bulbs can be on.
k) More generally, there are C* ways in which k bulbs can
be on.
n) Finally, there are Cl ways in which all n bulbs can be on
(it is again obvious that Cnn = 1).
It follows that there are a total of

Dn = Cl + Cl + - + Cnk + - + Cnn (2)

ways of lighting n bulbs. Comparing (1) and (2), we find that

Co + c i + - + c i + ••• + c ; = 2n,

as required. (To make the formula “more elegant,” we write Co

and Cq instead of 1.)
46. Let Dn denote the total number of ways of lighting n
traffic lights. Then Dj = 3 D j-lt since adding an extra traffic
light triples the number of ways of lighting the lights (the jth
light can be green, yellow or red). But obviously D x = 3. There­
fore
Dn = 3D„-i = 32A ,-2 = - = V -'D y = 3".

47. There are 3* ways of lighting k lights, each of which can

be green, yellow or red (see Prob.46), and 2n~k ways of lighting
n — k lights, each of which can be green or red (see Prob. 45).
Thus there are a total of 3k2"~k ways of lighting all the lights,
60 Sequences and Combinatorial Problems

since for each of the 3fc ways of lighting the group of k lights,
there are 2n~k ways of lighting the remaining lights.
48. 263 • 104 = 175,760,000 license plates (there are 26 let­
ters in the alphabet).
49. Suppose there are a total of D„ u-digit numbers containing
no zeros or eights. Then Dj = 8A - i > since the extra digit can
be 1, 2, 3, 4, 5, 6, 7 or 9. But obviously Dy = 8. Therefore

A, = 8A - i = 82A -2 = - = 8- ‘A = 8".

In particular, there are

D6 = 86 = 262,144

six-digit numbers containing no zeros or eights.

50. Imagine that the positions of all the teeth in the mouth
have been numbered once and for all from 1 to 32. Then to each
inhabitant of the city assign a sequence of 32 zeros and ones
according to the following rule: The first term is a 1 if there is
a tooth in the first position and a 0 if not, the second term is a 1
if there is a tooth in the second position and a 0 if not, and so
on. Since different inhabitants are assigned different sequences,
the maximum number of inhabitants of the city is simply equal
to the total number of such sequences, namely 232 (recall the
solutions to Probs.44 and 45). Incidentally, 232 m 4,000,000,000
which exceeds the present population of the earth.
51. Consider the terms obtained after multiplying out the
expression
(jc - 1) (jc - 2) - (x - 100)

but before combining terms involving the same power of .v.

Each such term is the product of 100 factors, and each factor is
either the letter x o r a number. The first factor is either x or —1,
the second is cither x or —2, and so on. We arc only interested
in products in which x appears 99 times and a number appears
once. There are 100 such products, since the number can appear
 Detailed Solutions 61
in the first place, the second place, and so on, up to the hundredth
place. In other words, the products are

—1 • x • .y .v, x ■( —2) ■x ■x ••• x , ..., .y • ,\- ••• a' • ( —100).

^ ^ v v
99 times 99 times 99 times

Therefore the coefficient of x " equals

100 • 101
-(1 + 2 + ••• + 100) = -5050
2
(recall Example 1, p. 4).
52. Case 1 ( Order is important)
In this case each sum is determined by the first term, which
can be any positive integer less than //. Thus there are n — 1 ways
of writing the sum.
Case 2 (Order is unimportant)
If n is odd, there are only half as many ways as in Case 1,
since the sums

// = 1 + (// — 1 ) , // = (// — 1) + 1 ,

n = 2 + (n — 2), n = (n - 2) + 2,

etc., are regarded as the same. Hence the answer is now\{n — 1).
If n is even, the sum
n n
n ------ 1----
2 2

has no “mate,” but all the others do. Therefore the answer is
now
j + (h ~ 1) ~ 1 = £
2 2 '

53. First we find the number of sums where the first term
equals 1. In this case, the sum of the second and third terms
equals n — 1. Clearly there arc as many such sums as there are
ways of writing the number n — 1 as a sum of two terms, i.e.,
62 Sequences and Combinatorial Problems

n — 2 ways as wc know from the solution of the preceding prob­

lem. Similarly, there arc n — 3 sums such that the first term
equals 2, and so on, with just one sum such that the first term
equals n — 2. This gives a total of
(» - 2) (n - 1)
(n — 2) + (n — 3) + ••• + 2 + 1 =
2

ways of writing n as a sum of three positive integers (recall

Example 1, p. 4).
Now use the same method to show that the number of ways
of writing a positive integer n ^ 4 as a sum of four positive
integers equals
(n - 1) (n - 2) (« - 3)
6
Another solution of Problem 53 will be found in the remark on
p. 64.
54. If Sn denotes the sum of the numbers in the (n + l)st row
of Pascal’s triangle, then 5'n+1 = 2Sn. In fact, writing the nth
and (n + l)st rows as
1 a b f g 1
0+1 1+ a a + b ••• / + g g + 1 1+0

we find that
Sn = (0 + 1) + (1 + a) + (a + b) + •••

+ ( / + g) + (g + 1) + (1 + 0).
Combining the first terms in each set of parentheses and then
the second terms, we have
S„ = (0 + 1 + a + + / + g + 1)
+ (1 + a + b + + g + 1 + 0)
= 2(1 + a + b + ••• + g + 1) = 2iS'n_ i.
 Detailed Solutions 63
But obviously
Si = 1 + 1 = 2,
and hence
Sn = 2Sn- i = 22Sn- 2 = = 2n~1S 1 = 2",

as asserted.
55. The hint to the problem shows that it is indeed possible
to put 14 bishops on the board without any two threatening
each other. But we must still show that if more than 14 bishops

are put on the board, then two necessarily threaten each other.
First we ask how many bishops can be put on the black squares
without any two threatening each other. The answer is at least
7 (recall Figure 6), but no more as we know show. To this end,
we note that only one bishop can be put on each of the 7 verti­
cal “black diagonals” shown in Figure 8, since two bishops on
the same black diagonal will necessarily threaten each other.
64 Sequences and Combinatorial Problems

There are also 8 horizontal black diagonals (two consisting of a

single square) perpendicular to the vertical diagonals shown in
Figure 8, but if more than 7 bishops are put on these diagonals,
two will necessarily threaten each other. In fact, if a bishop is
put on each of the 8 horizontal black diagonals, then the two on
the diagonals consisting of a single square would be at opposite
corners of the board and hence would necessarily threaten each
other. However, there is no need to consider the horizontal
black diagonals in Figure 8, since the same answer is obtained
by considering just the vertical black diagonals.
A similar argument shows that 7 is the largest number of
bishops that can be put on the white diagonals without any two
threatening each other. Since a bishop on the white squares
cannot threaten a bishop on the black squares, 14 is the largest
number of bishops that can be put on the whole board without
threatening each other.
Now let W denote the number of ways of putting bishops on
white squares, B the number of ways of putting bishops on black
squares and A" the number of ways of putting bishops on squares
of either color, in each case without any two threatening each
other. Then clearly
W — B, N = B W = B 2,
i.c., N is a perfect square.
Remark. It is not hard to solve the more general problem of
an it x n chessboard (i.e., a chessboard n squares on a side)
where n is even. Prove that 2n — 2 bishops can be put on such a
board without any two threatening each other and that this num­
ber of bishops can be put on in 2" different ways.
56. One way of giving out the apples and pears is
o • o •
i.c., apples on the first and third days, pears on the second,
fourth and fifth days. Another way is
• • • o o
 Detailed Solutions 65

(pears the first three days, apples the next two days). Thus we
must enumerate all arrays made up of two light circles and three
dark circles. But this has already been done in the solution of
Problem 44, where the answer was found to be C2 = 10.
57. Every way of giving out the apples and pears corresponds
to an array of k light circles and n dark circles, with days on
which apples are given out indicated by light circles and days
on which pears are given out indicated by dark circles. There­
fore the problem is completely analogous to that of finding all
ways of lighting k + n bulbs such that k bulbs are on, a quantity
we denote by c£+n (recall Prob.45).
58. C \C \.
59. According to the hint, there is one and only one rook in
each file. Moreover, there is obviously one and only one rook
in each row. The rook in file a can be placed on any of 8 squares
if all 8 rows are previously free. But as a result of putting this
rook on the board, one row is no longer free (for example, if the
rook is on square a2, no further rooks can be put in row 2).
Hence there are now only 7 ways of putting a rook in file b
(in the case just considered, it can be placed on squares bl, b3,
b 4 ,..., b8). As a result of putting rooks in files a and b, two rows
are occupied, i.e., there are now only 6 free rows left and hence
only 6 ways of putting a rook in file c. Similarly, there are 5 ways
of putting a rook in file d, 4 ways in file e, 3 ways in file f, 2 ways
in file g and only one way in file h.
In other words, there are 8 ways of putting a rook in file a,
8 • 7 ways of putting rooks in files a and b, 8 - 7 - 6 ways of
putting rooks in files a, b and c, and so on. Hence there are

8 - 7 - 6 - 5 - 4 - 3 - 2 - 1 = 8!

ways of putting rooks in files a, b, c , ..., h (! denotes the factorial

as in Prob.5).
To check your understanding of this solution, ask yourself
in how many ways n rooks can be put on an n x n chessboard
66 Sequences and Combinatorial Problems

without any two rooks threatening each other. The answer is of

course n ! Make sure you see why.
We now solve the problem by another method, which gives a
different answer (?!). Hence this new solution must contain an
error, at least if you believe in the correctness of the first solution.
The second solution will be given for an n x n chessboard from
the outset. Suppose n rooks are put on an n x n board, with no
two rooks threatening each other. Then the first rook can be
placed on any one of n2 squares, for example on the square shown
in Figure 9a for the case n = 4 (the other rooks are not indi­
cated). As in the figure, we shade all squares threatened by this

(°)
Fig. 9

rook, including the square occupied by the rook itself. Then the
remaining n — 1 rooks must all occupy unshaded squares, since
otherwise one of them would threaten the first rook. Next we
delete the shaded squares, which together make up a row and a
file, leaving behind four pieces A, B, C and D (illustrated in
Fig. 9a for the case n = 4). By moving these pieces parallel to
themselves, they can be assembled into an (/j — 1) x (n — 1)
board, as shown in Figure 9b (where n = 4 again). No two
rooks on the (n — 1) x (n — 1) board threaten each other, since
otherwise they would have threatened each other on the ori­
ginal n x n board. Moreover, distinct arrangements of rooks on
the new board correspond to distinct arrangements of rooks
on the original board, and every arrangement of rooks on the
 Detailed Solutions 67

new board can be obtained by deleting a row and a file from the
original board. As already noted, there are n2 possible choices
of the square ossupied by the first rook. It follows that

L„ = n2L„. j , (1)
where L„ denotes the total number of arrangements of rooks on
an n x n board such that no two rooks threaten each other. But
obviously
L , = (1 !)2 = 1 ( 2)

(there is only one way of putting a rook on a board consisting

of a single square). Together (1) and (2) imply

L„ = n2L„-v = n2 (n - l )2 L„_2 = ••• = n2 (n - l)2 ••• 22Z-i

= n2 (n - l)2 ••• 22 • l 2 = (n!)2,

in seeming contradiction with the first solution. Where is the

mistake?
Pay close attention as we now reconcile the two solutions:
According to the second solution, there are (n!)2 distinct ways

of putting n rooks on an n x n board without any two rooks

threatening each other, while according to the first solution,
there are only n\ ways. The second solution would be correct if
the rooks were numbered, since then two arrangements like
those shown in Figure 10 would be distinct. However, if the
rooks are unnumbered, the two arrangements are the same, as
in Figure 11, and then the first solution is the correct one. Fig. 11
68 Sequences and Combinatorial Problems

Thus we have in fact proved two distinct results:

1) There are /„ = n \ arrangements of n indistinguishable rooks
on an n x n chessboard such that no two rooks threaten each
other.
2) There are L„ = (n!)2 arrangements of n distinguishable
rooks on an n x n board such that no two rooks threaten each
other.
Each of these propositions implies the other, as we now verify
for the case n = 4 (the proof generalizes at once to the case of
arbitrary n). On a 4 x 4 board, find 4 squares with the property
that if 4 rooks are placed on these squares, then no two rooks
threaten each other. Label these squares a, b, c and d. Clearly
there are /4 different ways of choosing the squares. Now label
the rooks themselves with the numbers 1, 2, 3, and 4, and let P4
be the number of distinct arrangements of the rooks 1, 2, 3 and 4
on the squares a, b, c and d. Obviously P4 does not depend on
the choice of the squares a, b, c and d. In fact, if A, B, C and D
are any other squares with the same property, there arc still P 4
distinct arrangements of the rooks 1, 2, 3 and 4 on the squares
A, B, C and D. It follows that

L4 = P4/4.

But obviously P4 = 4! (why?), and hence

P4 = 4! /4 = (4!)2.

More generally, if P„ denotes the number of distinct arrange­

ments of n numbered rooks on n fixed squares of a chessboard,
then
*T-71 = *P n*n>
I *P n = '*
»jim

and hence /„ = n\ implies L„ = («!)2 and conversely.

60. Label the n objects 1, 2 , Then every way of choosing
two of the n objects corresponds to a pair of numbers (the
numbers of the objects chosen), where obviously the pairs (k , /?;)
 Detailed Solutions 69

and (/», k) correspond to the same way. Therefore C2 equals

the number of pairs
(1,2), (1,3), , (n - 1,n).

The first number in the pair can be chosen in n ways and the
second number can be chosen in n — 1 ways, giving n(n — 1)
ways in all. But the pairs of numbers (k , m) and (m, k) are
counted twice, although they represent the same pair of objects.
It follows that
s,n _ n (/I - 1)
L-2 —------------- •
2

Remark. We can now give another solution of Problem 53.

Writing a number n as a sum of three terms can be thought of as
gathering n beads strung along a wire (as in an abacus) into three
groups. Since there are n — 1 spaces between the beads, we in
elTect must choose two of the spaces (where the beads will be
separated). But this can be done in

(« —2) (/I — 1)
c; = i

ways.
61. C28 , 26! = 281/2, where, as always, /?! denotes the product
1 • 2 ••• n.
62. Let P\ denote the number of ways of giving 5 apples, no
two of which are alike, to 8 sons. Then P5 can be calculated by
two methods:
1) The number of ways of choosing 5 of the 8 sons (namely
those to receive apples) equals Cf. But for each such group of
sons, there are 5! ways of giving out the apples (the apples arc
distinguishable). Therefore

P i = C l- 5!

2) The first apple can be given to any of the 8 sons, the

second apple can be given to any of the remaining 7 sons, and
70 Sequences and Combinatorial Problems

so on, up to the fifth apple which can be given to any of 4 sons

(the other 3 get no apples). It follows that

P® = 8 • 7 • 6 • 5 ■4 = —
3!
63. The proof is by induction:
a) The result is true for n = 1. In fact, the first number in the
second row of Pascal’s triangle is 1 = C q and the second number
is 1 = C[.
b) Suppose the result is true for n = j. Then it is also true for
n = j + 1. This is obvious for k = 0 or k = j + 1, since
CJ0+1 = 1, CjXi = 1, while the first and last numbers in any
row of Pascal’s triangle are both 1. Recalling that C[ is the num­
ber of ways of turning on k out of j bulbs, we now ask how many
ways there are of turning on k out of / + 1 bulbs. If the (J + l)st
bulb is on, there are C*_ 1 ways of illuminating the other bulbs
(any k — 1 of the remaining j bulbs can be turned on). On the
other hand, if the (j + l)st bulb is off, there are CJk ways of
illuminating the other bulbs (any k of the remaining./ bulbs can
be turned on). It follows that

C{+1 = C { -1 + Cl (1)

But Cl is the (k + l)st number in the (J + l)st row of Pascal’s

triangle, by hypothesis, while the next row consists of the num­
bers
1> CJ0 + c \ , C\ + CJ2 , ..., Cj~ 1 + CJj, 1,

by definition (recall Prob. 53). Using (1) we can write this row as
1 +1 +1 /"ij+1 1
W > 2 J •••> w y 1•

Therefore the result is true for n = j + 1, as asserted, and hence

for arbitrary n, by induction.
64. Cl • 8 • 7 • 6 • 5.
65. C \c t° .
 Detailed Solutions 71
66. We begin by finding the number of ways in which 3 ob­
jects can be chosen from n objects with regard for order. First
we choose any of the n objects, then any of the remaining n — 1
objects, and finally any of the n — 2 objects which still remain.
Thus the three objects can be chosen in n{n — 1) (n — 2) ways
if order is important. But this gives every group of objects 6
times if order is unimportant, since the objects A, B and C(say)
can be chosen in any of the 6 distinct orders ABC, ACB, BAC,
BCA, CAB, CBA. Hence there are

^n _ H (n - 1) (n - 2) n (n - 1) {n - 2)
^3 — ------------------------ —-----------------------
6 1*2-3

ways of choosing 3 objects out of n without regard for order

(for the meaning of C3, or more generally Cl, see Probs. 45
and 63).
67. By definition, Cl is the number of ways of choosing k
objects out of n (without regard for order). Generalizing the
calculation of Cl given in the preceding problem, we find at
once that

„ _ n(n — 1) •••(/! - k + 1) _ n(n - 1) ••• (n - k + 1)

*“ I • 2 ••• k ~ k\
0)
where k = 1,2, ..., n and k\ = 1 • 2 ••• k. Formula (1) can be
written even more concisely as
n\
C nk ( 2)
k\ (n — k)\

and holds for k = 0 if we define 0 ! = 1.

It is easy to prove (2) by induction in n. The formula clearly
holds for n = 1, since

1! 1!
C1 — = 1, C\ = = 1.
0!1! 1! 0 !
72 Sequences and Combinatorial Problems

Suppose (2) holds for // = j. Then it follows from the formula

r J+1 — cJ ± cJ

proved in the solution to Problem 63 that

Cfc+ 1 — J! +
(k — \)\ (j — k + 1)! k ! (J - k) !
j'-
— '
(k - 1)! (j - k ) \ \ j - k + 1 kj
J'- k + (j - k + 1)
(A; _ ] ) ! (_/_/;)! k ( j — k + 1)
J'- j+ 1
(k - 1 ) l ( j - k ) l k (j — k + 1)
where in the second step we use the obvious relations
k\ = k ( k — 1)!, (j - k + 1)! = (j - k + 1 ) ( ; - ^ ) !
Therefore
/'C*
-»./+ 1 — ( j + !)•'
k) (j + 1 - A:)!

i.e., (2) holds for n = j + 1. Hence, by induction, (2) holds for

arbitrary
The quantity C£ is called the number o f combinations o f n
objects taken k at a time, equal to the number of ways of choos­
ing k out of n objects without regard for order. There is also
a special notation for the number of ways of choosing k out
of n objects with regard for order, namely Pi. This notation has
already been used in the solution to Problem 62, where we proved
that
p \ = c l ■5! = —
3!
More generally, we have
ni
Pi = C lk\ =
(n - k)\
 Detailed Solutions 73
as you should verify. The quantity Pi is called the number o f
permutations o f n objects taken k at a time. If k = n, the quantity
Pi = Pi is simply called the number o f permutations o f n objects,
often denoted by P„.
68. There are
d88 _ 88!
■»6 — -----
6!
ways of playing six different consecutive notes. The number of
six-note chords equals
^88 88!
6 6! 82! '

69. Each of the terms obtained after multiplying out the

expression

(1 + o ) ( l + 6 ) ( 1 + c ) ( l + r f ) 0 + e ) ( \ + / ) ( 1 +g )

is the product of 7 factors, there being 7 sets of parentheses,

where each factor is either a letter or the number 1. Thus we must
find the number of products of 7 factors, each of which can be
“in two states.” In other words, our problem is equivalent to that
of finding all ways of lighting 7 bulbs, each of which can be
either off or on. The answer is clearly 27 (cf. Probs. 44 and 45).
70. We arc interested in products in which letters appear three
times and the number 1 appears four times (recall the solution
to Prob. 69). This is just the number of ways of choosing 3 out
of 7 letters, i.e.,

1-2-3

More generally, we might multiply out the expression

(1 + <?i) (1 + a2) ••• (1 + an)

and ask for the number of terms which arc products of k letters.
Clearly the answer is now Cl by exactly the same reasoning.
6 Silverman V
74 Sequences and Combinatorial Problems

71. Each of the terms obtained when the expression

( l+ x + ;v ) 20
is multiplied out without combining terms is a product of 20 fac­
tors, where each factor is 1, x or y. Thus we must find the num­
ber of products of 20 factors each of which can be “in three
states.” In other words, our problem is equivalent to that of
finding all ways of lighting 20 traffic lights, each of which can be
green, yellow or red. The answer is clearly 320 (cf. Prob. 46).
72. The coefficient of jc17 is

20- 19 • 18
9
2
The coefficient of x 18 is 0.
73. The expression
(1 + tfj) (1 + a2) ••• (1 + a56) (1)
reduces to
(1 + x)56
if we set
ax = a2 = — = a56 = x. (2)

But according to the solution to Problem 70, multiplying out

the expression (1) leads to C |6 terms which are products of
8 letters, and obviously each of these terms becomes x8 after
making the substitution (2). It follows that the coefficient of x8
in (1 + x)56, after multiplying out the expression and combining
similar terms, equals C f6. Similarly, the coefficient of x48 equals
Cjf. The numbers C |6 and C jf are equal, and hence so are the
coefficients of x8 and x48 in (1 + x)56. In just the same way, the
coefficients of x6 and x 50 are equal. More generally, the coeffi­
cients of x* and x n~k in the expression (1 + x)n are equal, since

k\{n — k)\ (n-k)\k\

74. The proof is almost literally the same as that of Problem 73.
 Detailed Solutions 75
75. The proof is by induction:
a) The formula
(1 + ,x)B = a0 + a^x + a2x 2 + ••• + an- lxn~l + anx" (1)
holds for n = 1, since
(1 + x )1 = 1 + 1 • x
and in this case the binomial coefficients obviously coincide with
the numbers in the second row of Pascal’s triangle.
b) Suppose (1) holds for n = j — 1, so that
(1 + a:)-'-1 = a0 + axx + ••• + akx k + ••• + aj
where a0, a l5 ak, ..., a j - k are the numbers in theyth row of
Pascal’s triangle. Then (1) also holds for n = j, i.e.,
(1 + x)J = b0 + bk + bkx + ••• + bkx k + ••• + bjXJ, (2)
where b0, bu ..., bk, ..., bj are the numbers in the (J + l)st row
of Pascal’s triangle. To see this, we note that
bo = o0 (= 1), by = a0 + ax, ..., bk = + ak, ...,
bj-1 = Oj-2 + Oj -1, bj = ffj-i (= 1),
by the definition of Pascal’s triangle, so that (2) becomes
(1 + x)J = 1 + (a0 + ak) x + ••• + (o*-i + ak) x k
+ + (aj -2 + aJ- l) x J~l + xJ. (3)
But (3) is obviously true, since
(1 + x)J = (1 + x)J- 1(l + x)
= (a0 + akx + ••• + akx k + ••• + aJ- lxJ~l) ( 1 +
= a0 + axX + ••• + akx k + ••• +
+ a0x + ••• + ak- i Xk + ••• + aJ- 2 xJ~i + Qj-iXJ
= fl0 + (°0 + Q\)x + •" + iflk- 1 + ak) X>1
+ ■■■ + (aj - 2 + aJ- 1) x J- 1 + xs .
It follows by induction that (1) holds for all n.
76 Sequences and Combinatorial Problems

76. First note that

C t = 1, C \ = 4, C l = 6, C t = 4, C t = 1.
Therefore
100 ••• 0400 0600 ••• 0400 ••• 01
H im e s k times H im es A; times

= C £ l 04*+4 + C 3 l 0 3* + 3 + C ^ I O 2**2 + c?10 *+1 + c

= (10k+1 + l)4

(recall Prob. 74), which is clearly a perfect square.

77. Since . by
(a + b)n = a" M H---- j ,

the term involving ak is

Cnkakbn~k.

The proof of the binomial theorem is now almost obvious.

78. We first find all terms involving xk, writing

(x + y + z f = [.v + (y + z)]n = ••• + C£x*(y + z)n~k + (1)

Then in the expression (y> + z)"~* we find the term involving^':

(y + z)n~k = ■■■ + C"~ky lzn~k~l + •••. (2)

It follows from (1) and (2) that the term involving x ky ‘ is

CnkCnr kx ky lzn- k- 1.

The explicit formula for CkC"~k is

c nc n -k _ (» - k)\ = n\___
* ' k\ (n — k)\ /! (n — k — /)! k \ /! (« - k - /)! '

79. Setting x = 1 and then x = —1 in the binomial theorem

(1 + x)n = Cn
0 + Cl x + C"x2 + ••• + O - ,
 Detailed Solutions 77

we obtain
2" = Cl + C" + Cl + ••• + c ; ( 1)
and
0 = Cl - Cl + Cl + ••• + (-1)" Cn
n ( 2)

Adding (1) and (2), and then subtracting (2) from (1), we find
that
CnQ + Cl + Cl + ••• = C\ + Cl + Cl + ••• = 2”- 1.

80. The result of multiplying out the expression

(1 + x —3a-2)1965

and combining similar terms is some polynomial

Uq + axx + a2x 2 + a3x 3 + •••

The sum of the coefficients of this polynomial equals the value

of the polynomial for x = 1, since

a0 + ax • 1 + a2 • l 2 + a3 • l 3 + •••
= Oq + a3 + a2 + a3 + •••.

Hence there is no need to actually carry out complicated alge­

braic operations. We need only set x = 1 in the original ex­
pression, obtaining

<?o + + a2 + a3 + ••• = (1 + 1 — 3 ■l 2)1965

= ( - 1 ) 1965 = - 1 .

81. There arc Cl distinct pairs of vertices of the polygon.

Joining all these points, we obtain Cl segments of which n are
sides and the rest diagonals. Hence there are Cl diagonals.
82. Being even, the number must end in 0 or 4. If it ends in 0,
any of the 4! = 24 arrangements of the first four digits is
allowed. If the number ends in 4, it cannot be allowed to begin
with 0, since then it would actually be a four-digit number.
This means that the 0 must be in the second, third or fourth
78 Sequences and Combinatorial Problems

position. In each of these three cases there are 3! = 6 arrange­

ments of the remaining digits 1, 3 and 5. Hence there are a total
°f 24 + 6 + 6 + 6 = 42
distinct even five-digit numbers made up of the digits 0, 1, 3, 4
and 5 with no repetitions.
83. 26 = 64.
84. 2s = 32.
85. In the solution to Problem 63, we found that
Ckn - - /ifl /nfl + 1
1 “T C* “ ^ k

[change j to n in formula (1), p. 70]. Therefore

/ /- t n + 1 f /- in + l\ i /-»n + 2 . /-in + 3 , , >nn + k
V.'-'O i (-'I J + C2 + ^3 +
/ /-in + 2 , /- in + 2 \ , /- .n + 3 , , /-iH + Jc
— i^Cx + C2 ) + C3 + ••• + C*
//-in + 3 . /-in + 3 \ , , /in + k
— (C2 + C3 ; + • * • + C*
Cn3 + 4 . “T .“T /-in
C*
+ fc
—
/-in + fc+1
••• — C * ,

and hence
C n1 + 1 ,
+
/-in + 1
C2
-
~T C 3
/-in + 2
+
, ,
+
/-in + k
v-'jfc
/-in + fc+1
— C*
|
— 1,

since C0 = 1.
86. C lnc l n = (3rt)
(n!)3
57. The condition of the problem can be written concisely in
the form e F G EF EG FG EFG
6 7 6 2 4 3 1
(E for English, F for French and G for German). Suppose the
person who knows all three languages leaves the room. Then
among the remaining people, nobody knows more than two
languages, and we have the following problem:
E F G EF EG FG EFG
5 6 5 1 3 2 0
 Detailed Solutions 79
Now suppose the three people who know both English and Ger­
man leave the room. Then the number of people who know other
pairs of languages does not change (since nobody in the room
knows all three languages):
E F G EF EG FG EFG
2 6 2 1 0 2 0
Finally suppose the two people who know both French and
German leave, together with the person who knows both
English and French. Then we have
E F G EF EG FG EFG
1 3 0 0 0 0 0
But this problem is trivial, i.e., there is one person in the room
who knows only English and three who know only French.
Moreover, 7 people in all left the room. Hence the room ori­
ginally contained 11 people, one of whom knows only English.
88. Suppose the number of subscribers who have borrowed k
given books is summed over all possible sets of k books, and let
S k denote this sum. Then

S = 5 , - 5 2 + 5 3 - S 4 + . . . + ( - l ) " - 1^ (1)
is the number of people subscribing to the library. To see this,
consider a subscriber who has taken out exactly k books, and
examine the contribution he makes to each term in the sum (1).
There is no loss of generality in assuming that our subscriber
has borrowed the first k books. Then he contributes Cy = k to
the term S ly since he is one of those who borrowed the first
book, one of those who borrowed the second book, etc., up to
and including the kth book. Our subscriber contributes C\ to S2,
since he is one of those who borrowed every pair of books from
among the first k books. Similarly, he contributes C„* to Sm if
m < k and 0 to Sm if m > k. In other words, he contributes
Cy - C k2 + C \ - C \ + ••• + ( - l ) k- l Ckk
80 Sequences and Combinatorial Problems

to the sum S. But by the solution to Problem 79,

-C o + C\ - C\ + Cl + ••• + ( - 1 )k~ 'C k = 0,

and moreover Cq = 1. Therefore the contribution of our sub­

scriber to the sum S equals 1. The same argument applies to
every subscriber, and hence S equals the number of subscribers
to the library, as asserted.
Remark. It is not hard to sec that Problem 87 is a special case
of Problem 88. On the other hand, the method used to solve
Problem 87 can also be used to solve Problem 88.
89. Consider the five sets of telephone numbers of the form

A B C
1 2 -------- - 1 2 ------ - - 12 --
D E
------1 2 - ----------12
where the spaces indicate missing digits (for example, D is the
set consisting of all numbers with 12 in the second and third
places and arbitrary digits elsewhere). Each of the sets A, R, C,
D and E contains 104 numbers. The sets A and D have no num­
bers in common, but the sets A and C share 102 numbers, i.c.,
all numbers of the form 12 1 2 __ It is easy to see that there
are 6 • 102 numbers which belong to two sets, but only one
number, namely 12 12 12, which belongs to three sets. There­
fore by the result of the preceding problem, there arc a total of

5 • 104 - 6 • 102 + 1 = 49401

six-digit telephone numbers containing the combination 12.

 Test Problems

Group I (Sequences and induction)

1. Calculate the sum

------------- + ---------------- + + -------------------------------------------------- .

1- 3 3-5 (2/i - 1) (2/i + 1)
2. Prove that
_P_ V_ n2 »(» + !)
1 -3 3 -5 (2n — 1) (2n + 1) “ 2(2n + 1)
for every positive integer n.
3. Calculate the product

where // ^ 3.
4. Prove that
(// + 1) (/i + 2) ••• (2/2 — 1) 2n
1 • 3 • 5 ••• (2// - 1)
5. Prove that
1 1 1 13
-----------1-----------+ ••• + — > ——
n + 1 n + 2 2« 24
for every integer n > 1.
6. For what positive integers does the inequality

2" > 2 / 2 + 1
hold?

81
 82 Sequences and Combinatorial Problems

7. Prove that
2n~l (an + bn) > (a + b)n,

where a + b > 0, a ^ b, n > 1.

8. Suppose the plane is divided into parts by n circles. Prove
that the plane can be colored black and white in such a way that
any two adjacent parts have different colors.
9. The sides of an arbitrary convex polygon are shaded from
the outside. Then some diagonals are drawn such that no three
diagonals intersect in a single point. Each of these diagonals is
also shaded on one side, i.e., a narrow shaded strip is drawn
along one side of each segment (see Fig. 12). Prove that among
the polygons into which the original polygon is divided by the
diagonal, at least one is entirely shaded from the outside.
Fig. 12 10. Prove that the number
10"+1 - 10(n + 1) + n
is divisible by 81 for every integer n ^ 0.
11. Calculate the sum
1*3 + 3*5 + ••• + w(/z + 2).
12. The sum of the first m terms of an arithmetic progression
equals the sum of the first n terms (m ^ it). Prove that the sum of
the first m + n terms of the progression equals zero.
13. Calculate the sum
6 + 66 + 666 + ••• + 666 ••• 6.
n times

14. Does there exist an arithmetic progression with the num­

bers 1, yj2 and 3 as terms (not necessarily in this order)?
15. Given a geometric progression whose ratio is an integer
other than 0 or —1, prove that the sum of any number of arbi­
trarily chosen terms cannot equal any term of the progression.
16. Each side of an equilateral triangle is divided into n equal
parts, and lines parallel to the sides are drawn through every
point of division. As a result, the original triangle is divided into
 Test Problems 83

congruent subtriangles. These subtriangles are then colored, some

black and others white, in such a way that every black triangle
shares sides with an even number of white triangles, while every
white triangle shares sides with an odd number of white tri­
angles. Prove that the subtriangles at the vertices of the original
triangle have the same color.
17. Find the fallacy in the following “proof” by induction
that all numbers are equal.
Theorem. All numbers are equal.
Proof. A single number equals itself, and hence the theorem
is true for n = 1. Suppose the theorem is true for n = k. Then
it is true for n = k + 1. In fact, given k + 1 numbers, arrange
them in some definite order. The first k numbers are equal, by
hypothesis, and hence equal the first number. Now eliminate
the second number. Then the remaining numbers, which include
the first number and the (k + l)st number, are all equal and in
particular equal the first number. Therefore all k + 1 numbers
equal the first number and hence are all equal. The theorem
now follows by induction.
Group 2 (Combinatorial problems)
18. In how many distinct ways can the faces of a cube be
numbered from 1 to 6?
19. Is 9 or 10 the more probable outcome of tossing a pair of
dice?
20. Prove that the numbers in the kth row of Pascal’s tri­
angle are odd if and only if k = 2" (n = 0, 1, 2, ...).
21. Which is larger 9950 + 10050 or 101s0?
Hint. Write 99 and 101 as 100 — 1 and 100 + 1.
22. Prove that the number 11100 — 1 ends in three zeros.
23. How many distinct divisors does the number 10! have?
Hint. 10! = 28345271.
24. By the integral part of x, written [x], is meant the largest
integer ^ x. Prove that
[(2 + V i ) ']
84 Sequences and Combinatorial Problems

is odd for any positive integer n.

25. a) In how many ways can 19 people be seated at a round
table?
b) In how many ways can 19 people be seated at a round table
if there are exactly r people between two given people?
26. Find the coefficient of a 50 in the expression

(1 + * )1000 + jc(1 + a)999 + + a 999 (1 + A') + A1000.

27. A child’s rattle consists of a circular wire on which arc

strung 3 white beads and 7 red beads. Some rattles look different
at first glance but can be made identical by suitably shifting the
beads (see Fig. 13). How many different rattles arc there?
28. In how many ways can a piece go from square al of a
chessboard to square h8 (see Figure 5, p. 19) if in each move the
piece can go either one square forward or one square to the
right?
29. In how many ways can two white rooks and two black
rooks be put on a chessboard without a white rook threatening
a black rook?
30. In how many ways can a checker at square a 1 of a checker­
board be promoted to a king (if there are no other checkers on
the board)?
31. In how many nonself-intersccting ways can one go from A
to B in Figure 14? (The motion is along circular and radial
routes with stops at the stations indicated by little circles.)
32. How many ways are there of putting 9 rooks on a 3 x 3 x 3
three-dimensional chessboard without any two rooks threatening
each other? (A rook threatens its row, file and vertical column.)
33. The cells of a 2 x « rectangular array arc filled in with
numbers from 1 to n in such a way that no row or column
contains the same number twice. How many such arrays are
there?
34. How many five-digit telephone numbers are there con­
taining the combinations 23 and 37?
 Test Problems 85

35. How many /i-digit telephone numbers are there containing

the combination 12?
36. A father had seven daughters. Every time a daughter got
married, each of her older unmarried sisters complained to the
father about his violating the custom that the eldest unmarried
daughter get married first. After the last daughter was finally
married, the father recalled that he had received seven com­
plaints. In how many ways could this have happened?
A more serious statement of this problem goes as follows:
By a permutation of the numbers from 1 to n we mean these
numbers arranged in any order. Any pair of numbers (r, s) in a
permutation is called an inversion if r > s and r appears to the
left of s. How many permutations of the numbers from 1 to n are
there containing k inversions? (In the problem of the daughters
n = k = 7.)
37. If a city has two tall buildings, their spires can be seen in
any order (from left to right) if the observer stands at the proper
point. The same is true for three tall buildings, provided they
do not all lie along some straight line (see Fig. 15). An architect
wants to locate seven tall buildings in such a way that their
spires can be seen in any order from suitable points. Will he
F ig . 15
succeed?