You are on page 1of 153

Essays on number theory.

School Mathematics Study Group.


[New Haven? 1960-

http://hdl.handle.net/2027/mdp.39015017386932

Public Domain, Google-digitized


http://www.hathitrust.org/access_use#pd-google

We have determined this work to be in the public domain,


meaning that it is not subject to copyright. Users are
free to copy, use, and redistribute the work in part or
in whole. It is possible that current copyright holders,
heirs or the estate of the authors of individual portions
of the work, such as illustrations or photographs, assert
copyrights over these portions. Depending on the nature
of subsequent use that is made, additional rights may
need to be obtained independently of anything we can
address. The digital images and OCR of this work were
produced by Google, Inc. (indicated by a watermark
on each page in the PageTurner). Google requests that
the images and OCR not be re-hosted, redistributed
or used commercially. The images are provided for
educational, scholarly, non-commercial purposes.
n

^1
'1H°D'uoiipois ^^

rr
Mama Aaaadi-
SCHOOL
MATHEMATICS
STUDY GROUP

ESSAYS ON NUMBER THEORY I

UNIVERSITY OF MICHIGAN LIBRARIES


SCHOOL
MATHEMATICS
STUDY GROUP

ESSAYS ON NUMBER THEORY I

Written for the SCHOOL MATHEMATICS STUDY GROUP


Under a grant from the NATIONAL SCIENCE FOUNDATION
541

v. \

Financial support for the School Mathematics Study Group has been provided by the

National Science Foundation.

Copyright i960 by Yale University.

PHOTOLITHOPRINTED BY CUSHING- MALLOY. INC.


ANN ARBOR, MICHIGAN.UNITEDSTATES OFAMERICA
CONTENTS

* * »

Page
1. Prime Numbers 1

2. Congruence 9

3. The Fundamental Theorem of Arithmetic 19

Answers to Questions 29
Preface
Not all of mathematics can (or should) be taught in formal
textbooks. Just as an English course is enlivened by selections
from literature, a mathematics course can gain depth and interest

from special readings.

The present volume might be read in conjunction with the

SMSG First Course in Algebra or Intermediate Mathematics. It intro


duces the subject of number theory, a branch of mathematics highly

esteemed for its naturalness, conceptual clarity, and elegance. We

hope that these essays will prove enjoyable and stimulating.


1

1. PRIME NUMBERS

Among the natural numbers are the prime numbers: a positive


integer is called a prime if it has no divisors except 1 and itself.
Thus 6 is not a prime, for it has the divisor 2, and 2 is different
from both 1 and 6. We call 6 a composite number. But 7 is a prime,

because its only divisors are 1 and 7. By agreement we shall not


count the integer 1 among either the primes or the composite integers.
Thus every integer greater than 1 is either prime or composite.

If a number is composite, it has a prime divisor (that is, a


divisor which is prime). For you have seen in this course or in an
earlier course that every integer greater than 1 can be written as
a product of prime factors. Obviously each prime in the product is
a divisor of the integer in question.
The smallest prime is 2; it is also the only prime which is an

even integer. No larger even number can be prime for it has 2 as

a divisor. If we write down the primes in the order of increasing

size, we get

2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31,

The dots at the end mean that there are still larger primes. How

large do the primes get? You can probably find a prime larger than
100, but could you find one larger than 1,000,000?
(How would you go about finding a prime larger than 100? If
you test a given number, 101 say, by dividing it in succession by

various numbers, would you have to try all integers less than 101

as possible divisors? Or would it be sufficient to try all primes

less than 101? If so, wouldn't it be enough to try all primes less

than \/l01? See if you can prove that if an integer n has a proper
divisor, it has a proper divisor which is not more than^/nl )

To put the question in another way, are there a finite number

of primes or are there infinitely many? If you think this over a

bit, you will probably find it impossible to think of any way to


attack the problem. Yet the answer was known to the Greeks.' In
fact, we find the proof that there are infinitely many primes in

Euclid's Elements.

This proof is a proof by contradiction: suppose there were

only n primes, say Pi, P2, ..., Pn, where n is a certain in


teger. Now let
A = . .
Pn + .1,
P1P2P3*
that is, A is the integer obtained by multiplying together all
the primes and adding 1 to the result.
Now A is either prime or composite. If A is prime, we

have a contradiction since A is not one of the primes p.,, p2, .**,
pn (Why?), and these were supposed to be all the primes.

The other possibility is that A is composite. In that case


A is divisible by a prime; call it p. Could p be equal to pi?
Certainly not, for if we divide A by p^ we get a remainder of 1,

since p, certainly divides the product P.,p?... p . Since p

divides A, p cannot be p.^ in the same way, p could not be


P2, P3, or any of the primes in the set p.^, p2, ..., Pn. Again we

have a contradiction, namely, we assumed that p^, p2, ..., pn were

all the primes, but here is a number p, undoubtedly a prime,

which is not one of pi, p2, ..., Pn.


So whether A is prime or composite (and these are the only
possibilities) we arrive at a contradiction to our original assump

tion that there are only a finite number of primes. Therefore,


this assumption must be false, and we have proved the theorem.

THEOREM. There are infinitely many primes.


Although this proof is an example of an "indirect proof", it
can be turned around to make a direct proof. What we have actually
shown is that if we have any_ set of primes {p^, p2, ..., Pn), there
is a prime p which is not in the set and which is not more than

A. For example, suppose we know that 2,3,5, are primes. Then

our proof shows that there is a prime p which is not either 2,3

or 5 (and so is greater than 5) and p _£ 2*3*5 + 1 = 31. So there


is a prime larger than 5 but not larger than 31. Actually, of
course, there are several primes between 5 and 31. Our proof, then,

does this: given any set of primes, it gives a limit below which
there must be a new prime. In this way we can produce in succession

an indefinite number of primes.


There are a very large number of fascinating questions having

to do with primes. Although most of these are quite easy to state,


the answers to many are not known. For example, it was conjectured
that every even number greater than 2 is the sum of two primes

(examples: 10 = 3 + 7, 46 = 23 + 23, 100 = kj + 53). This is


called the Goldbach conjecture, after the name of the man who first
4

proposed the problem in 1742. It has never been proved or disproved.

However, there are some questions about primes that we can

answer. Suppose we write down the sequence of positive integers

for the form 4k - 1:

3, 7, 11, 15, 19, ....


Are there an infinite number of primes in this sequence?
(By an integer of the form 4k - 1, we mean an integer which is
equal to 4k - 1 if we chose the right integer k. Thus 3 = 4.1 - 1,

27 ■ 4*7 - 1, and so on. Of course, the value of k is different


in each case. )

Before we discuss this question, let us notice a few things

about odd numbers. Every odd number is either of the form 4k + 1

or 4k - 1. (Can you prove this?) Furthermore, if you multiply two

numbers of the form 4k + 1, the product is also of the form 4k + 1.

(Check this.) Naturally, if you multiply any finite number of in


tegers of the form 4k + 1, the product is still of this form, be
cause you could multiply the first two integers, then multiply the
result by the third integer, than multiply this result by the

fourth integer, and so on.

Now suppose A is an integer of the form 4k - 1, then we can

conclude that A has at least one prime divisor of the form 4k - 1.

For example, 19 is already prime, while 27 has the prime divisor 3.

(Prove this in the general case by assuming that all prime divisors

are of the form 4k + 1 and arriving at a contradiction. Notice


that A has only divisors.)
odd

Now we can return to the original question: are there infi


nitely many primes of the form 4k - 1? Suppose there are only a
finite number of such primes; call them p^, p2, . .., pn. Let

(*) A =
MPiPg... Pn)-1.
Notice that A is of the form 4k - 1.

We follow the proof of the previous theorem. A is either


prime or composite. If A is prime, we have a contradiction, for
A is then a prime of the form 4k - 1 but A is not one of the

primes p,, p2, ..., p . The only other possibility is that A is


composite. Since A is of the form 4k - 1, A must have a prime

divisor p of the form 4k - 1, as we have Just seen. But p is


not one of the primes pi, p2, ..., p , for p divides A whereas

no p1 divides A.
So whether A is prime or composite, there is a prime of the

form 4k - 1 which is not one of the primes p^, p2, ..., pn. This
is the contradiction we were looking for, and we have proved the

following result.
THEOREM. There are infinitely many primes of the form 4k - 1.

See if you can construct a similar proof that there are infi
nitely many primes of the form 6k - 1.

Actually, it is true that there are infinitely many primes of

the form ak + b, where a and b are any integers which have no


common divisor (except 1). (For instance, there are infinitely
many primes of the form 5k + 3.) The proof of this, however, is
very difficult. The first proof was given by P. G. L. Dirichlet

(1805 - 1859), a famous German mathematician.

Notice that we did not try to prove the above results by


looking for a formula for the nth prime. Such a formula, if it
exists, would be highly complicated, because the distribution of
primes among the integers is so irregular. As an example of this
irregularity, we prove:

THEOREM. There are arbitrarily long sequences of consecutive com

posite integers.
For example, there are 50 consecutive integers all of which
are composite. We can actually exhibit such a consequence.

First, we introduce a new notation: 1J = 1, 2! = 1*2,

3 J = 1*2*3, etc. In general, n.' (read "n factorial") is the


product of all the integers from 1 to n, inclusive. Now we con

sider the sequence

51! + 2, 51! + 3, 51! +


**,
..., 51! + 51.
There are 50 consecutive integers in this sequence. The first,
51 + 2, is divisible by 2, since 5U is and is, and the sum of
2
!

two integers divisible by is again divisible by 2. The second


2

is divisible by 3, for the same reason. In


3,

number, 51! +

general, 51! + is divisible by as long as is not more


k

than 51. So every integer in the sequence is composite. You can

easily see how to modify this proof if you want a block of length
n instead of length 50.
We have seen that consecutive primes can be far apart; can

they be close together? and are consecutive primes that differ


2

by 1, but obviously there are no other such pairs, for if is an


p

odd prime, is even and greater than and so is not a prime.


p
+

2
1

The next possibility is that consecutive primes differ by 2, and

there are many examples of such "prime twins": (3,5), (5,7),


(17,19), (29,31). But are there infinitely many such pairs? No

one knows, although many famous mathematicians have exerted


themselves trying to find out. What is your guess?

Finally, we might consider "prime triples", like 3,5,7. Are

there any other prime triples? See if you can prove it one way or
the other.
2. CONGRUENCES

In many situations in mathematics, what is important is not

the particular value of an integer but the fact that it differs


from another number by a multiple of 2 or a multiple of 5 or a
multiple of some other number.
'
For example, odd numbers are those

which differ from 1 by a multiple of 2, squares of odd numbers

differ from 1 by a multiple of 8, etc. The same thing happens in


ordinary life. What can be said about two times which read the
same on the clock? About two dates which have the same month and
day but are in consecutive years?

Two numbers which differ by a multiple of 2 are said to be

congruent modulo 2. We make the formal definition:


DEFINITION. Two integers a,b are said to be congruent modulo m

(where m is a positive integer) if and only if a - b is divisible


by m.

We write
a a b (mod m)

to indicate that a and b are congruent modulo m. In particu


lar, a = 0 (mod m) means that m divides a and conversely, m

is called the modulus of the congruence.


10

Congruence is a relation between the integers a and b

(with respect to the modulus m) . Many of the properties of the


more familiar relation of equality carry over to congruences.

Exercise 1. Prove that

a = a (mod m)

if a = b (mod m), then b = a (mod m)

if a = b (mod m) and b = c (mod m), then a = c (mod m) .

Congruences can be added, subtracted, and multiplied like


ordinary equations. Thus,

if a s b-(mod m) and c = d (mod m), then

a + c = b + d (mod m)

(l) a - c = b - d (mod m)

ac = bd (mod m)

For (a+c) - (b+d) = (a-b) + (c-d) and each term in the right member

is divisible by m, hence the right member is also. Similarly for


the second congruence. Finally, ac - bd =
(a-b)c + (c-d)b. Since
m divides a-b, it divides (a-b)c; for the same reason, m divides
(c-d)b. This establishes the third congruence above.
Using (l) we can answer questions such as: is 2 + 1 a

prime? (A prime is an integer > 1 which has no factors except 1

and itself.) This number is of some historical interest, for


Fermat stated that all numbers of the form 2^+1 are primes,

whereas Euler showed that 2 + 1 (i.e. n = 5) is divisible by 641.

We can easily prove Euler 's result by means of congruences without


expanding this very large number and dividing it by 64 1.
11

Namely, we have:

22 = 4 (mod 641), 2k = 16 (mod 641 ),


28 ■ 256 (mod 641), 216 = (2562) = 65536 = 154 (mod 64l),
232 = (154)2 = 23716 = 640 (mod 641).
Each congruence is obtained from the preceding one by multiplying

it by itself. The largest number we had to calculate was (256)2.


Prom the last congruence we get

232 + 1 = 641 = 0 (mod 641),

as promised.

Another thing we can do easily by means of congruence is to


prove the familiar rule for "casting out nines". Let N be a

positive integer and write it in the base 10:

N =
aD
+
10a1
+
102a2 +... +
10kak.
(The digits of N are therefore a0, a^ ak. )

Since 10 s 1 (mod 9), we have 102 = l2 = 1 (mod 9), and, in


general, 10r = 1 (mod 9). Hence, using the first equation of (l),
we get

(2) N = aQ +
&i +
&2 + - . - + ak (mod 9).
Now N is divisible by 9 if and only if N s 0 (mod 9), which,
according to (2), occurs if and only if the sum of the digits

ao + al + . . . + ak s 0 (mod 9).
Exercise 2. Prove that an integer is divisible by 3 if and only if
the sum of its digits is divisible by 3.

Exercise 3. Prove that an integer is divisible by 11 if and only


if the sum of the digits in the "odd places" is congruent modulo

11 to the sum of the digits in the "even places". By "odd places"

we mean the units place, the hundreds place, etc.


12

Exercise 4. Prove that a number is divisible by 4 if and only if


the part of the number occupying the units and tens places is
divisible by 4.

In arithmetic we have the cancellation law: if ab = ac and

a 4 0 then b = c. What is the analogue for congruences?

Suppose we have ab = ac (mod m), which is the same thing as

a(b-c) = 0 (mod m) .

It does not follow that m|a or m|(b-c). For some of the prime
factors of m might divide a and the remaining prime factors

divide b - c. However, if m is prime to a (i.e., m and a

have no common divisors other than l), then m|(b-c). This is a

consequence of the theorem: _if m| xy and m


_is prime to x, then

m] y. For a proof see the Supplement The Fundamental Theorem of

Arithmetic. So we have the result:

Theorem 1. Lf ab = ac (mod m) and m


_is prime to a, then

b = c (mod m) . In other words, we can cancel a common factor from


a congruence provided the common factor is prime to the modulus.

The equation ax = b, where a,b,x are integers, cannot be solved

for x unless a happens to divide b. By contrast, the congruence


ax s b (mod m) can always be solved for x provided only that
(a,m) = 1. We shall now show how this comes about.
The numbers 0, 7, 35, -14 are mutually congruent modulo 7

(that is, any pair of the numbers is congruent). Likewise:


2, -5, 16, -47 are congruent modulo 7. Consider the set of all
integers congruent a fixed integer modulo 7; this set is called a

residue class modulo 7. For example, the numbers congruent to 0

modulo 7 form one residue class RQ; the numbers congruent to 1


13

modulo 7 form another residue class R, . R0 and R.^


have no common

elements, for if an integer a = 0 and a ■ 1, then 0 = 1 (mod 7),


by Exercise 1. Consider the residue classes RQ, R.^ R2, ..., Rg.

(RA
is the set of numbers m i (mod 7).) Every integer n is in
one of these classes. For we can write n = 7q + i where

0 ^ i <£
6. Then n € R.^. (This explains the word "residue", which
means remainder. ) In this way, the set of integers is partitioned
into 7 sets, the residue classes modulo J, no two of which contain
common elements.

Of course, there was nothing special about the choice of the


modulus 7 in the above discussion. We have, in fact, the general
result: Let m be an integer ^ 1, and Rj_ (1 = 0, _1, . . ., m-l)
the set of integers which are congruent to 1 modulo m. Every in
teger n jLs an element of Rj_ for some _i. Moreover, Rj and R.

have no common elements if i 4 J. Finally, n€ R^ if and only if


n = qm + 1 for some _gt.

The set (0, 1, 2, ..., m-l) is called a complete residue

system.

DEFINITION. A complete residue system modulo m (m > 1) is a set


which contains one and only one element from each residue class R^

(i = 0, 1, 2, ..., m-l). Thus (m, m + 2, 2 - m, 3, 4, ..., m - 2,


-1) is another complete residue system. No complete residue system

can have fewer than m elements, for it must contain an element


congruent to 0, 1, 2, ..., m-l. Nor can it have more than m

elements. For if we distribute the r > m elements among the m

residue classes, one class is bound to have at least two elements.


We see, therefore, that a complete residue system modulo m
_is a
14

set of m mutually incongruent integers, and conversely.


Certain complete residue systems are of particular interest.
Theorem 2. _If a _is prime to m, then S = (a, 2a, . . ., ma)

is a complete residue system modulo m.

Proof . S certainly contains m numbers. Suppose two of


them were congruent, ra = sa (mod m) . Since (a,m) = 1, we have,

by Theorem 1, r = s (mod m) . But 1 £ r ^ m, 1


£ s ^ m, so that
0 £ \r - s| < m. Hence, m|(r - s) only if r - s = 0. This shows
that no two elements of S are congruent modulo m. Therefore, S

is a complete residue system.


Theorem 2 is what we need to discuss the equation.

(3) ax = b (mod m), a prime to m.

We look for a solution x.

Since a is Prime to m, the set (a, 2a, ..., ma) is a com

plete residue system and so one of its members ax, say, is con

gruent to b. This proves that (3) always has a solution in inte


gers.
Theorem _3. The congruence

ax = b (mod m), a prime to m in which a,b,m


are given, always has an integral solution x.
Of course there are many solutions of (3), for x + m, x + 2m,

x - m are solutions if x is a solution. But there is only one

solution in any given complete residue system.


Exercise 5. Prove: if (a,m) = 1, there is only one x satisfying
ax = b (mod m) and 0 ^ x < m.

In particular, . the equation ax = 1 (mod m) (a,m) = 1, has a

unique solution in the range 1


^ x < m. (Why can't x = 0?) The
15

number x is called the reciprocal of a (mod m) and often written


a; thus, aa = 1 (mod m) . To solve (3), we need only to know a,
for x = ab evidently satisfies (3).
The reciprocal of an integer mod m can be found by trial. A

way which is sometimes quicker is provided, for m a prime, by the

following theorem, which is important in its own right.


FERMAT'S LITTLE THEOREM. If p is prime, then
ap = a (mod p) .

To prove this, notice first that if pla, the result is trivial,


since ap - a = a(ap~1 - 1) is evidently divisible by p. But if
pf a, we can cancel the factor a and get

(k) aP"1 ■ 1 (mod p), (a,p) = 1

as the congruence we have to prove. Again we use Theorem 2. Since


S = (a, 2a, pa) is a complete residue system as well as

T = (1, 2, 3, p), each element of S must be congruent to


some element of T and conversely. Now pa = p (mod p). Hence,

the product of all the elements of S other than pa is congruent

to the product of all the elements of T other than p:

(5) a.2a. . .(p-l)a = 1-2. ..(p-l) (mod p).


But the left member has p - 1 factors and equals a?'1 -1*2. . . (p-l) .
Cancelling the factor 1*2... (p-l), which is prime to p, from both

members of (5), we get (4). This completes the proof of Fermat's


theorem.

We can use this theorem to calculate the reciprocal of a num

ber to a prime modulus. Clearly a = ap~2 if (a,p) = 1, for


aa = aap~2 = aP"1 = 1 (mod p) .

Example 1. Solve 4x = 7 (mod 13). First calculate k. Since 13


16

is a prime and (4, 13) = 1, we have ¥ s 4'1 . Now r- 16 s 3,


= 9 = .4, 48 = 16 = 3, 43 = 3.4 = -1.
Hence, = 48.43 ■ 3--1 ■ 13). Finally,,
-4"

10 (mod

x = TT*7 ■ 10-7 ■ (mod 13). Check: 4.5 = 20 = (mod 13).

7
5
The preceding discussion can be applied to the solution in in
tegers of equations like

(6) 4x + 13y = 35.

Since this is a single equation in two variables, we see that there


are infinitely many solutions if there is one solution. For if
Xo» yo is one solution, then certainly xD + 13t, yQ - 4t is also a
solution if is any integer. (Check this by substitution.)
t

Equations of this type are called Diophantine equations after the


Greek mathematician Diophantus who studied them. Such an equation
arises, e.g., from the problem: in how many ways can you make

change for a dollar using only nickels and dimes? The equation is

5x + lOy = 100, with the restriction x 0, 0.


^>
y
^

We can solve (6) as follows. If x and are integers


y

which satisfy (6), then

4x - 35 = 13y and

4x = 35 (mod 13).
Since is as we saw in Example
"4"

mod 13 10, 1, we have

x = 35*10 = - 4.-3 b -1 (mod 13). Hence x is of the form

13 - where is an integer; write


t

t
1

x = 13t - 1.

Substituting this in (6) gives


17

Thus (x = -1 + 13t, y = 3 - 4t) is a set solutions of (6) for


each integer t. For instance, (-1, 3), (12, -l), (168, - 49) are
solutions obtained by taking t = 0, 1, 13 respectively. Notice
that there are no positive solutions (i.e., solutions in which x

and y are both positive).


Exercise 6. In how many ways can a total weight of 25 pounds be

built up out of 2 pound and 3 pound weights?

In the general case


(3a) ax + by - c,

we assume a prime to b. Then we solve the congruence ax = c

(mod b) which is possible by Theorem 3. Let x= x (mod b), then


x =
x0 + b t. Substituting in (3a), we find y = (c - ax0)/b - at.
Note that (c - ax0)/b is an integer since ax0= c (mod b). For
reasons of space, we do not complete the discussion by considering

the generalization of Fermat's theorem to the case in which the


modulus is not a prime, nor do we treat congruences ax = b (mod m)

where a is not prime to m, or the corresponding Diophantine


equation ax + by = c where a and b are not relatively prime.
For these matters and many other fascinating topics, the reader is
referred to books on Number Theory such as:

Uspensky and Heaslet, Elementary Number Theory (McGraw-Hill).


19

3. THE FUNDAMENTAL THEOREM OP ARITHMETIC

One of the first steps in your study of algebra was to extend


the system of integers to the larger system of rational numbers.
The purpose of this step was to make division possible in all cases:
the equation a = bx, where a and b are rational numbers and
b 4 0, always has a solution x which is rational. By the same

token, if we stay within the set of integers, division is not always


possible: given integers a and b 4 0, sometimes there is an in
teger c such that a = be, sometimes not. In the first case we

say that b divides a, or b is a divisor of a, or a is divi


sible by b. We write b | a for "b divides a" (notice that the bar,
| , is vertical, not slanting).

The study of the properties of integers is called Arithmetic.


In this Supplement we shall study the important property of divisi
bility. Throughout, the letters of the alphabet shall stand for
integers.
Certain simple facts are observed at once and easily proved.
We leave these as exercises.

Exercise 1. Prove: if a | b, a | c, then a | (b+c), a | (b-c),


and a | bf, where f is any integer.
Exercise 2. Prove: if a | b, then a | (-b) and (-a) | b.
20

An important tool in the study of divisibility is the so-called


division algorithm.

DIVISION ALGORITHM. If a, b are positive integers then

(la) a = qb + r,
where

(lb) 0 ^ r < b.

The integers q and r are uniquely determined.


The division algorithm merely states the familiar fact that
when two integers are divided in the usual way we get a quotient

and remainder and that the remainder is less than the divisor. We

have stated the algorithm only for a, b positive but it actually


holds for all a, b provided b / 0.
To give a formal proof of the Division Algorithm as stated

above, consider the multiples Ob, b, 2b, ... of b. Since Ob = 0 < a

but kb > a for some positive integer k, there must be a largest


integer q such that qb ^ a. Set a - qb = r. (q is the "quotient,"

r the "remainder.") First, r ^ 0. Next, if r ^ b we would have

a - (q+l)b =a-qb-b = r-b^0, i.e., (q+l)b £ a, so qb was

not the largest multiple of b which is £ a. Therefore, r < b,

and we have proved equation (l).


Notice that if b > a, we have q = 0, r = a < b : a = 0*b + a.

We still have to show that q and r are unique.


Suppose we could have

(2) a = qib + r1 = q2b + r2, 0 ^ r2 < b, 0 ^ r2 < b.

If q.j^ > q2 , we get q1 ^ qg + 1 and

a ■ q^b + r ^ (q2+l)b + r^ > q2b +


r2 + rj > q2b +
r2
= a,
21

so that a > a, a contradiction. So q^ £ q2. By interchanging q^

and q2, we conclude q2 £ q^. Therefore, q^ = q2. Subtracting

qlb =
q2b from (2) gives r^
=
r2. Hence, q and r are unique.
Interest next centers on the common divisors of two integers.
Since -d is a divisor whenever d is, and only then, we may as well

consider only positive divisors. For example, 8 and 12 have the

common divisors 1, 2, 4; 4 and 9, however, have no common divisor


other than 1. We say that 4 and 9 are relatively prime or that
k Is prime to We call 4 the greatest common divisor (gcd) of 8

and 12 (written (8,12) ■ 4) because every common divisor of 8 and

12 divides 4. Clearly, (4, 9) - 1.

DEFINITION. Let a and b be integers not both 0. By the gcd of


a and b, written (a,b), we mean the positive integer d having

the following properties:

(1) d | a, d | b,

(2) if dx|
a, d^ b, then 61 | d.

We can see without much difficulty that there cannot be more


than one gcd. (Hence, the use of "the" in "the positive integer d"

in the above definition.) Suppose there were two gcd's d^ and d2.

Since d1 is a gcd, dg|d1; since dg is a gcd, d^ldg. Both


d^
and d~

are positive; therefore d^ ■ d2.


Exercise 3. Why did we assume in the definition of gcd that a and

b were not both 0?

Factorizing integers in order to find their gcd is troublesome

when the numbers are large. And we can never be sure that every
pair of integers has a gcd, no matter how many special cases we

work out. We shall now give a practical method of finding the gcd
22

which involves only the division algorithm. This method, which is


called the Euclidean Algorithm, also proves the existence of the gcd.

Let a and b be the two integers whose gcd is desired. We

shall assume a, b are both positive; the remaining cases are easily

handled once this case is settled. Before treating the general


case, let us consider the particular integers 72 and 33.

Write, by the division algorithm,

(3.1) 72 - 2.33 + 6

(3.2) 33 r 5.6 +©

(3.3) 6 = 2.3
Prom this chain of equations we deduce that (72, 33) = 3. For
from (3.3), we have that 3|6. Then from (3.2): 3 13 and 3| 5.6 imply
3 1 33 (see Exercise 1). Prom (3.1): 3|6 and 3| 2. 33 imply 3|72. So

3 is a common divisor of 72 and 33. Suppose d is another common

divisor. Then from (3.1): d|72 and d|2*33; hence, d|6, since

6 - 72 - 2.33. Prom (3.2): d| 33 and d | 5.6 imply d|(33 - 5.6), i.e.


d|3. This shows that any common divisor d of 72 and 33 divides 3.

Therefore 3 is the gcd: (72, 33) = 3.

We can obtain an additional result by writing the equations

(3) in reverse order:


6 » 2.3
33 = 5.6+3 ► 3 - 33 - 5.6
72 - 2.33 + 6 —+6 = 72 - 2*33

3 = 33 - 5 (72 - 2*33)
or
3 - - 5.72 + 11.33.

The gcd 3 is a linear function of 72 and 33 with coefficients


which are integers .
23

In the general case we would have the chain of equations:

(4.1) a = qb+r, 0 ^ r < b

(4.2) b - qxr +
rlt 0 £ i^ <
r

(4.3) r =
q2r1 + r2, ° ^
r2 < ^

(4.4) ^
=
q3r2 + r3, ° £
r3 <
V
This process must come to an end. For the sequence b, r, r.,, r2, . . .

is a decreasing sequence of non-negative integers and so must even


tually reach 0.

The last two steps are:

(4.(i»l)) rn_2
=
qnrn^1 + rn, 0 £ rn < r^

We claim that rn is the gcd of a and b. The proof is


exactly the same as in the special case. Certainly rn r , by

(4.(n+2)) and so
rn|rn-2
by (4.(n+l)).
Working upwards we conclude that r n a, rnJ^* ^ d is a common

divisor of a and b, then d|r by (4.1). Working downwards, we

arrive at the fact that d r by (4.(n+l)).


Moreover, we can express rn in terms of a and b. From

(4.(n+l)) we have rn
=
rn_2
- qnrn„i. Both rn_1 and
rn_2 can

be expressed in terms of earlier r's by means of equations in the


set (4). After a finite number of steps we have r expressed as

a linear function of a and b with integral coefficients.


Notice that in the above proof we used the properties proved
2k

in Exercise 1 in an essential way (if an integer divides two other

integers, it divides their sum, difference, and any multiple of


either integer) .

Summarizing our results, we have:

THEOREM 1. Every pair of integers a, b other than the pair


0, 0, possesses a unique gcd. If d = (a,b) is the gcd of a

and b, there exist integers x and y such that

ax + by = d.

In particular, if a and b are relatively prime, there exist


integers x and y such that

ax + by = 1.

Exercise 4. Prove the assertions of Theorem 1 in the cases in


which a and b are not both positive.

Consider the cases: a > 0, b < 0; a < 0, b > 0; a < 0, b < 0.

Theorem 1 enables us to prove the result which will lead


directly to the Fundamental Theorem of Arithmetic.

THEOREM 2. If a|bc and (a,b) = 1, then a|c.

This theorem does not seem so remarkable if we imagine a, b, c

factorized into primes, but remember that we have not yet discussed

factorization into primes.


The proof of the theorem is very simple. Since (a,b) = 1, we

have, by Theorem 1,

ax + by - 1

for certain integers x, y. Multiply by c:

acx + bey = c

Now a certainly divides acx, and a|bcy since a|bc by hypothesis.


Therefore, by Exercise 1, a|c.
25

As a corollary we get

THEOREM 3. If the prime p|bc and pfb, then p|c. (pfb means

"p does not divide b" . ) For if pfb, p must be prime to b, since,

as a prime, p has no divisors other than 1 and itself. We can

then apply Theorem 2.


Exercise 5. Let p be a prime. Then (p, a) = 1 if and only if
pfa.
A slight extension of Theorem 3 is the following, which we

leave as an exercise.
Exercise 6. Let p and p1, p2, ..., Pn
be primes. If
pj (PjPo. . »Pn), then p is equal to one of the

primes pi#
We are now in a position to prove the

FUNDAMENTAL THEOREM OF ARITHMETIC: Every integer > 1 can be

written as a product of primes. If the primes are written in the

order of increasing magnitude, the factorization is unique.


We regard a prime as its own "product of primes". Thus 2 = 2

is a factorization into primes.


The proof is in two parts; first, we have to show that n > 1

is a product of prime factors. If n is a prime, we are through.

If not, n =
a^g, where 1 < a.^ < n, 1 < a2 < n. If a^ a2 are
both primes, we have our factorization; otherwise, we repeat the

same process on a, and a2, obtaining n =


aoaj,ac.ag, with
1 <
a3 < al, 1 <
a4 < al» 1 <
a5 < a2' X < a6 < a2*
In the
successive steps of the process, the factors get smaller and smaller,
but since they are positive integers they must eventually reach 2

if they are not primes at some intermediate stage. Thus n has a


26

factorization Into primes.


Suppose there were two factorizations of n:

(5) P^..* Pr
=
q^*** %
where the p's and qfs are primes and Pt ^ P2 ^ . . . ^ Pr,
^l £ ^2 .£
* .. .^ qs* Since Pi divides p.jP2 ... pr, it divides
q q ... q . By Exercise 6, p. ■ q. for some i. By the same

reasoning, q1 - pj for some J. The fact that the p*s and q's
are arranged in increasing order means that i = j = 1, p = p,,
q ■ q_. (For p <^
p. = q, £ q. = p.; since the first and last
members of this chain of inequalities are the same, we have equality
throughout . )

Now divide both members of (5) by p, ■


q^, getting

P2Pg... Pr - q2q3*.. qs »

and proceed as before. We get p = q , p. ■ q,,, etc.

If s > r, we would have

which is impossible. Hence £ r; by symmetry,


s r ^ s, and we

conclude r = s. This is the end of the proof.


Here is an application of the Fundamental Theorem.

Exercise 7» If the product of two relatively prime Integers Is a

perfect square, each of the integers is a perfect


square.

You may feel that the Fundamental Theorem is completely obvious

and needs no proof. However, there are many number systems besides

the rationals. In these systems we can define integers, divisibili


ty, and primes: we can do arithmetic. But in most of these number

systems, while we can factor an integer into primes, the factoriza-


27

tion is not unique! Certain integers in these fields have two or


more factorizations into entirely different primes. (See the
Supplement: A New Field. ) This shows that we cannot regard the

uniqueness of factorization into primes in the rational field as


something which is obvious. It needs to be proved.
29

ANSWERS TO QUESTIONS

PRIME NUMBERS

1. If n has a proper divisor d (i.e., 1 < d < n), we can write

dd' = n. Now either d or d* is <£$n, because if both

d,d* >y^» dd' > n.

2. A / pi (i = 1, 2, . . . , n) because the equation

A - P-jPg
. . .
Pn
+ 1 shows that A > p1p2 ... Pn > PA»
since
each prime > 1.

3. Since an integer when divided by 4 must have one of the re


mainders 0, 1, 2, 3, we can write any integer in one of the
forms 4k, 4k + 1, 4k + 2, 4k + 3. Now 4k = 2 . 2k and

4k + 2 - 2(2k + 1), so these integers are even. On the other

hand, 4k + 1 and 4k + 3 have remainders of 1 when divided by


2; they are odd. But 4k + 3 - 4(k + l) - 1.

4. Let the two numbers be 4^ + 1 and 4kg + 1.

(4kx + l)(4kg + 1) = 16
k-^ +
4kx +
4k2 + 1 .

4(4^1^ +
kx + kg) + 1.

5. A is the product of its prime divisors. (Some primes may

occur more than once in the product, e.g., 60 2 . 2 . 3 .


5.)
30

Since A is odd, all prime divisors of A are odd. If all


prime divisors were of the form kk + 1, the product would
also be of this form, contradicting the fact that A is of the

form hk - 1.

6. Suppose that there are only a finite number of primes of the


form Gk - 1, namely, p, , p~, ... , p . Consider
A 6(p1Pp...p ) - 1. If A
= is prime, we have a contradic
tion, for A is not p, or p2 or . . . or p .

Suppose A is composite. A is of the form 6k - 1.

Since A is odd, A has only odd divisors, and therefore


only odd prime divisors. Now, every odd prime is of the form

6k - 1, or else 6k + 1. (Proof: 6k, 6k + 2, and 6k + h

are all even, and 6k + 3 is not prime.)


Since (6kx
+ l) (6k2
+ l) - 6(6^^ +
kx
+ kg) + 1, the

product of integers of the form 6k + 1 is again an integer

of this form. Therefore, A has a prime factor of the form

6k - 1. But this cannot be p,, p0, ..., p ; again we have

a contradition.

7. Consider the sequence

(n + 1)! + 2, (n + l)« + 3, ..., (n + l)» + n + 1

and follow the reasoning in the text.


8. There is only one "prime triple". For suppose n - 2, n, n + 2

are all prime, n must be of the form 3k +1, 3k - 1, or 3k.

In the first case, n + 2 is of the form 3k + 3, i.e., n + 2 is


divisible by 3. But then n + 2 = 3 since it is prime. Then

n = 1, which is not prime. Next, suppose n is of the form


31

3k - 1, then n - 2 is of the form 3k - 3, i.e., n - 2 is divi


sible by 3. Hence n - 2 = 3, so n » 5, n + 2 = 7. This gives

the prime triple 3, 5, 7. Finally, if n = 3k, then n ■ 3, and

n - 2 = 1, not a prime. Hence, the only prime triple is (3,5»7).


33

CONGRUENCES

Exercise 1. a - a = 0 and 0 is divisible by m. If


a = b (mod m), a - b is divisible; hence, so is
b - a. Therefore b = a (mod m) . If a = b,
b = c (mod m), then a - b and b - c are divisible by
m. Therefore, so is their sum (a-b)+ (b- c) = a - c,

It follows that a = c (mod m) .

Exercise 2. Let the integer be N =


aQ + 10a
^^
+ ... + 10
a^. Since
10 s 1 (mod 3), 10k a 1 (mod 3), and
N = a0 + &i + . . . + a^ (mod 3). Hence, N = 0 if
and only if aQ
+
&^
+ . . . +
ak
■ 0 (mod 3).
Exercise 3. Let the integer be N - a + 10a, + ... +102«Ja2, (this
integer has an odd number of places). Since
10 m
-). (mod 11), 10k = 1 if k is even, 10k m -1
if k is odd. Hence, N m a0
- ai + a2 - a3 +. . .+
a2.
and N ■ 0 if and only if
aQ
+
a2 + a^ + ... +
a2* = ai +
a^
+ ac +. . .+
&21-1
(mod 11). The proof when N has an even number of
places is practically the same.
Exercise h. N - aQ + 10a j + ... + lO^aj^. Since 10 s 2,
34

10k = 2k (mod 4). Thus 10k s 0 (mod 4) if k ^ 2.

Hence, N ■
aQ +
10a! (mod 4) and N = 0 (mod 4) if
and only If aQ + 10a j ■ 0 (mod 4).
Exercise 5. Suppose x,, Xp both satisfy ax, = b, ax = b (mod m)

and 0 <; x^ < m, 0 £ .x^, < m. Subtracting the two con

gruences, we get a(x, - Xp) = 0 (mod m) . Since


(a,m) = 1 we can cancel the factor a and then have

x, - Xp — 0 (mod m) . But 0 ^ x, - Xp < m, so

xl " x2 = ^. ^ne ^w° x's are ^ne same*

Exercise 6. We have to solve the equation 2x + 3y = 25 in integers


x, y such that x ^ 0, y ^ 0. We have 2x = 25 (mod 3),
x s -1 (mod 3) (since 2x m -x (mod 3)). Hence,

x = -1 + 3t, t an integer. Then

2$ - 2X 25 - 2(-l+ 3t) _
* m m m
3 3

Since x ^ 0, t ^ 1. Since y ^ 0, t ^ 4. Hence

there are four solutions (2,7), (5,5), (8,3),


(11,1).
35

THE FUNDAMENTAL THEOREM OF ARITHMETIC

Exercise 1. Since a|b, a|c, we have b = ad,, c ■


ad2, where

dj, d2
are integers. Therefore
b + c =
adjL + adp ■ a(d, + dp), so a|(b c).
+

Likewise, a|(b - c). Also bf ■ ad.f - a(d,f);


hence a|bf.
Exercise 2. For a certain integer g, we have b = ag. Hence,
-b = -ag = a(-g), and b = ag = (-a)(-g). Thus

a|(-b) and (-a)|b.


Exercise 3. Every integer / 0 divides 0, hence, all integers are
common divisors of the pair 0, 0. There is no

greatest common divisor of this pair of integers.


Exercise 4. Case I. a > 0, b < 0. Since b =|-b|, we have d|b
if and only if d|(|b|). Hence if d = (a,|b|), we

have d = (a, b). Since there are integers x, y such


that d = ax + |b|y, we have d = ax - by = ax + b(-y).
Case II. a < 0, b > 0. This is the same as Case I
with a and b interchanged.
Case III, a < 0, b < 0. We have

(a, b) = (-|a|, -|b|) = (|a|, |b|). If d = |a|x = |b|y,


then d = -ax -by = a(-x) + b(-y).
36

Exercise 5» If p|a» then p and a have the common factor p.

Hence, if (p, a) - 1, pi a. Suppose pi a. Then a

does not have p as a factor. Since the only factors


of p are 1 and p, it follows that (p, a) = 1.

Exercise 6. Write P.,P2*.. Pn =


pi*pop3* *
,pn* Now applylnS
Theorem 3 we get that pIpi or p|p2Pg...Pn. If
p|Pl, P = Pi and the result is proved. If not, write

P2p3,,,pn " p2*p3p4..*pn and Proceed in the same way.

If p does not divide any of p1, p2, ..., Pj^, then

p'pn-lpn ' therefore, p|pn so that p =


pn.
Note: If the student is familiar with Mathematical Induction, he

can construct an elegant proof by assuming the result to be true

when the product has k factors.


Exercise 7. We are given ab = c2 with (a, b) - 1. Write the
el ek
factorization of 0,0 =
?^^ p2
e2
. . . p.

where e,, ... ek are positive integers (e. is


simply the number of times the prime pA occurs in
the factorization) . Also, write down the factoriza-
tions of a and b: a = q..
f1 . . . q
f s,
b = ... rtSt. We then have

-
r.j61

'ifl ..* C3rigl- rt8t -pi2ei pi


2ek

Since the two members of this equation are both facto


rizations of the same number ab (or r), the same

primes must occur in both. Hence, p, is some q or


r; say,
P.^®1
- Q^1 .
It follows that t±
=
2er If
we do this for every prime q, we see that a is a
37

product of primes raised to even powers. Therefore,


a is a square. By the same reasoning, b is a

square.
.*.
,53*
V. I

DATE DUE
AUG 8 T9ST.

OCT 10 1B62
OCT 2i
w go
JIM »

JUL gg 1963"

?s W * A
SCHOOL
MATHEMATICS
STUDY GROUP

ESSAYS ON NUMBER THEORY II

UNIVERSITY OF MICHIGAN LIBRARIES


SCHOOL
MATHEMATICS
STUDY GROUP

ESSAYS ON NUMBER THEORY II

Written for the SCHOOL MATHEMATICS STUDY GROUP


Under a grant from the NATIONAL SCIENCE FOUNDATION
MatheTatici

QA

.638
V.2-

Financial support for the School Mathematics Study Group has been provided by the

National Science Foundation.

Copyright i960 by Yale University.

PHOTOLITHOPRINTED BY CUSHING. MAU.OY. IMC.


ANN ARBOR. MICHIGAN,UNITEDSTATES OFAMERICA
CONTENTS
-* * *
Page
1. Arithmetic Functions - I The Number of Divisors of
an Integer 1

2. Arithmetic Functions - II - The Sum of the Divisors


of an Integer 7
3. Arithmetic Functions - III - The Distribution of Primes
and the Function it (n) 13

4. The Euclidean Algorithm and Linear Diophantine Equations 19

5. The Gaussian Integers 27

6. Fermat's Method of Infinite Descent 37

7. Approximation of Irrationals by Rational s 45

8. A New Field 51

Answers to Questions 57
Preface
This volume contains the eight chapters :
(1) Arithmetic Functions - I - The Number of Divisors of an
Integer
(2) Arithmetic Functions - II - The Sum of the Divisors of an
Integer
(3) Arithmetic Functions - III - The Distribution of Primes and
the Function ir(n)
(4) The Euclidean Algorithm and Linear Diophantine Equations
(5) The Gaussian Integers
(6) Fermat's Method of Infinite Descent
(7) Approximation of Irrationals by Rationals
(8) A New Field
These supplements were written for students who are especially
good in mathematics and who have a lively interest in the subject.
The author's aim in (l) and (2) is to lead the reader to discover
for himself some interesting results and to experience the thrill
of mathematical discovery. The others are more expository in
nature, but they contain exercises to clarify the material and to
give the reader a chance to work with the concepts which are intro
duced. It is suggested that the supplements be read with pencil
and paper at hand. All questions should be pondered and answered,
if possible when they occur. A casual reading of these supplements
is, in most cases, unprofitable, and in some cases impossible.
Answers have been provided. However, it is suggested that
these answers should not be consulted until the reader has finished
working through the unit or until he reaches a point where he needs
an answer in order to proceed.
For the most part the units are independent of each other.
However, some have somewhat tenuous ties with certain chapters of
the 11th grade material of the SMSG, (Intermediate Mathematics) .
In particular, Sections (l) and (2) may be used at any time
after the student has completed Chapter 3 of Intermediate Mathe
matics. While they are independent, Section (2) is easier and more
meaningful if
Section (l) has been done previously.
i
Section (3) may be read also after Chapter 3 of Intermediate
Mathematics. However, on the last pagelogarithms are mentioned
and for this reason it may be more useful after Chapter 8 of Inter
mediate Mathemat i c s . (logarithms and exponents).
Section (4) may be used at any time after Chapter 2 of Inter
mediate Mathematics (in which linear equations are discussed) .
Section (5) is designed to follow Chapter 5 on complex numbers
and also to pave the way for the section entitled "A New Field".
Section (6) naturally follows Chapter 9 on induction.
Section (8) assumes familiarity with Chapters 5 and 15 of
Intermediate Mathematics.
Suggestions for further reading are:
The Enjoyment of Mathematics by Hans Rademacher and Otto Toeplitz,
Princeton University Press, Princeton, 1957.
What Is Mathematics? by Courant and Robbins, Oxford, New York, 194l.
Number Theory and Its History by £. Ore, McGraw-Hill, New York, 1948.

ii
1.

ARITHMETIC FUNCTIONS

Leopold Kronecker, one of the great mathematicians of the


nineteenth century is supposed to have said in an after dinner
speech "God made the integers; all the rest is the work of man."
The basic role of the integers in the development of the real
number system lends some weight to Kronecker 's statement. In
your work with functions the domain of definition of the func
tion has usually been the set of real numbers or some subset of
this set. There are many interesting functions, however, which
have for their domain of definition the set of positive integers.
Such functions are called arithmetic functions . In the units
which follow we will consider several arithmetic functions which
prove useful in stating and answering many questions about
integers.

I
THE NUMBER OF DIVISORS OF AN INTEGER

Some people from time to time advocate changing the base of


our number system from ten to twelve. To say that our numbers
are written in the base ten means that we interpret a symbol
like 312 to stand for
3.102 + 1.10 + 2.

If we were using the base six then 312 would stand for
3.62 + 1.6 + 2 ,
which would be 116 in the base ten.
In for
...
any number base, b , we would need b symbols the num
bers 0, 1, 2, , b-1 .

In particular if we used the base twelve we would need two new


symbols, say t and e for 10 and 11.
Then 312 in the base twelve would represent

3.122 + 1.12 + 2
or 446 in the ten.
base
The symbol 4et21 would represent

4.124 + 11.123 + 10.122 + 2.12 + 1 ,

which would be 10347 in the base ten.


The claim is that the base twelve would make arithmetic easier.
made
The fractions 1/3, 1/4, 1/6, and 1/12 instead of having repre
sentations .333 ... , .25, .166. . . , and .083 . . .
would have the simple form .4, .3, .2, and .1 .
Whatever the merits of this proposal, it seems unlikely to be
adopted. However, it does suggest an interesting mathematical
problem. Suppose to find a number with a large number
we wanted
of divisors, but which was not too large to serve as a base for
system of numbers. The advantage would be that the more divisors
the number has, the more fractions would have convenient finite
representations. As a start we might make a table for the first
few integers .
11
Integer

3
Divisors

1, 2
1, 3
Number

2
of
Divisors
1

4 1, 2, 4 3

5 1, 5 2

Extend this table for all the integers up through 30.


Which number in the table has the smallest number of divisors?
If we extend our table will we ever encounter another integer with
this number of divisors? Why not?
Make a list of the numbers in the table with two divisors.
The numbers in this list are given a special name; they are called
prime numbers .

Now list the numbers with three divisors. Do you notice any
property which they have in common besides that of having the same
number of divisors? Are there other numbers in the table with this
property? Try to state a theorem about all the numbers with three
divisors.
How many numbers in the list have an even number of divisors?
Which numbers not have an even number of divisors?
do Check this
list with your theorem. Can you guess how many numbers less than
fifty have an even number of divisors? Less than 101?
Which numbers in your table have a prime number of divisors?
Do you notice any other property that these numbers have in common?
Could you make a guess about the form of a number with a prime
number of divisors. How many divisors does 8 have? 32? 27?

64? 2 ? 3 ? See if you can devise a theorem which states exactly


when the number of divisors is a prime.
Make table showing the number of times each integer
another
appears in the number of divisors column of your first table. That
is, how many integers up to thirty have one divisor, two divisors,
three divisors, etc. We can see from this new table that most of
the numbers up through thirty seem to have an even number of
4
divisors. One of the distinguishing traits of a mathematician is
his tendency to generalize his results. This means that once he
has solved a particular problem, he begins to think of a large
class of 'similar problems. This tendency to try to see the origi
nal problem as a special case of a much larger problem is one dif
ference between a mathematician and a person who likes to solve
problems. In the light of the information we now have about the
divisors of numbers, see if you can generalize your theorem about
the numbers which have three divisors.
starting point for our discussion was the problem of find
The
ing a number base which was not too large, but which had many
divisors. From this point of view ten has as many divisors as any
other number up to ten. However, the restriction that the number
not be too large was designed to keep the arithmetic simple. The
smaller the base the easier the addition and multiplication tables
are to learn. Taking into account both of these things, six would
seem to be a better choice than ten. It would then be unnecessary
to learn such troublesome parts of the multiplication table as
7x9, 9x6, etc. Unfortunately, for this base there are also dis
advantages. Large numbers would require many more digits in their
representation than they require in the base ten. So we are forced
to conclude that ten isn't really such a bad number base after all.
Suppose we pursue our aim of finding a number with a large
number of divisors, even if it isn't the most practical number base.
Which number up through thirty has the largest number of divisors?
Up to fifty
are there any numbers with nine divisors? Ten divisors?
More than ten divisors?
Of the numbers less than 100, which one has the greatest
number of divisors?
If you have an answer to the last question, you are probably in
a good position to devise a formula for the number of divisors of
any particular integer n . (If not, try to consider some special
cases. For example, we know how many divisors any prime has. How

many divisors does p k , a power of a prime, have?) The usual


notation for the number of divisors of n is t(n) , where is t
the Greek letter tau. Try to write out an explicit expression for
r(n).
5

If you are having trouble actually writing down the expression,


you are probably being handicapped by a lack of a suitable notation.
While this has nothing to do with the idea which enables you to
determine the number of divisors particular integer, devis
for any
ing a suitable notation turns out to be of great importance in many
parts of mathematics. Lack of a suitable notation for numbers is
thought by many to explain the Greek preference for geometry and
the relatively small amount of arithmetic and algebra they were
able to develop. Perhaps if you write n in the form
m nip m

n = p. p„ ...pr >where the p's are the distinct prime divisors of


n and the m's tell' you how many times the prime is a factor of
n, you will find this notation helpful in writing our your expres
sion for T (n) .

Find all the numbers less than 100 which have six divisors.
Find the smallest positive integer with fifteen divisors.
Find all primes that are one less than a perfect square. One

less than a perfect cube. One less than a fourth power. How many

primes are one less than a k power? Why?


2.

ARITHMETIC FUNCTIONS

II
THE SUM OF THE DIVISORS OF AN INTEGER
"In the beginning God created the heavens and the earth."
The Genesis account of creation goes on to tell how God labored
for six days, and on the seventh day He rested. As early as the
sixth century B. C. the Pythagorean brotherhood classified integers
into deficient, abundant, and perfect numbers according to whether
the sum of the proper divisors of the integer was less than, great
er than, or equal to the integer itself. Proper here means that
the integer itself is not counted as one of its divisors. Thus the
fact that 6 and 28 were perfect numbers, that is, 6=1+2+3
and 28 =1+2+4+7+14 , special significance.
gave them a
The ancients saw in the number six a symbol of the perfection of
the creation. The discovery that the phases of the moon repeat
every 28 days may also have had a part in the designation of these
as perfect numbers.
Can you find any other perfect numbers?
Euclid includes in his ELEMENTS a rule for obtaining even per
fect numbers. Before we consider Euclid's rule, let us take a
detour and consider the problem of finding the sum of the divisors
of a number. The sum of the divisors of an integer is an arithmetic
function, that is a function defined over the positive intergers.
We first note that the sum of the divisors is equal to the sum of
the proper divisors and the number itself. The usual notation for
the sum of the divisors of n is <T(n) where (T is the Greek letter

11
sigma. To try to find a formula for <r(n) directly is not too easy.
However, we can use the approach of the experimental scientist and
collect some data. Suppose we make a table for 0"(n) .
n Divisors of n 0"(n)
1

2 1, 2 3

3 1, 3 4

4 1, 2, 4 7

5 1, 5 6
8
Extend the table for all n less than 31 .

In our notation perfect <T(P) = 2P .


a number P is if
Mark the integers which are deficient with a D , those which are

abundant with A , and those which are perfect with P .


How many of each kind are there in your table?
You probably have already noticed that the easiest numbers for
which to compute (T(n) were the primes. (A prime is a number which
has exactly two divisors.)
Complete the following theorem: If p is a prime,
<T(p) =

The easiest case after that of the prime is probably that of


an integer which is a power of a prime.

What are the divisors of p ? Can you find the sum of 0"{p )?

(HINT: xr+1 - 1 = (x - 1) (xr + xr_1 + . . . + x + l) .

To prove this simply multiply out the right hand side.)


Vr
Now suppose that n = p q where both p and q are primes.
What are the divisors of n ? How many are there? What is their
sum?
k 2
Now suppose n = p q . What are the divisors of n ? How
many are there? What is their sum?

If n =
k s
p q , can you guess what <T(n) is in this case?
Check your answer in a few cases and see you can prove it. if
Now it shouldn't be too hard to devise a formula for <T(n)
for any n , provided we write n in the form
ml "^ mr
n =
pl p2 " " '
pr wnere tne P's are distinct primes and the

m's tell us how many times the prime is a factor of n .

Use your formula to compute <T(6), <T(l2), <T(l8), (T(24),


<r(28), <r(3o), <r(m) .

your table of <r(n) list all n for which <7"(n) is odd.


From
Do you notice any property these integers have in common? Complete
the following theorem and try to prove it:
If 0"(n) is odd, then n is
9

Let us now return to our original problem of finding perfect


numbers . We remember that in order for n to be perfect

<T(n) = 2n . Euclid arrived at the following rule: n =


2m"1(2m - l)
is perfect number if 2m - 1 is prime. Use Euclid's result to
a
find other perfect numbers. Try to prove Euclid's theorem:
If n =
2m_1(2m - 1) and 2m - 1 is a prime, then n is a perfect
number.
your computations with Euclid's theorem,
As you can see from
one good mathematics problem often leads to another. Euclid's
theorem tells us that 2m_1(2m - l) is perfect if 2m - 1 is a
prime. So that we can find as many perfect numbers as we can find
primes of the form 2m - 1 .

Suppose we consider this problem a bit. If m is 2 ,

2 -1=3, which is a prime. This gives the perfect number 6 .

If m is 3 , 2m - 1 = 7 , which is also prime This gives the


perfect number 28 . If m is 4 , 2m - 1 = 15 , which is not

prime. For m = 5 , 2 -1=31, which is prime and you can see


that the perfect number which corresponds to m = 5 is already quite
large. Test values of m up to 13 to see how many more perfect
numbers you can find.
The primes of the form 2m - 1 are called Mersenne primes
after a French monk, Father Marin Mersenne (1588-1648), who listed
eleven values of m less than or equal to 257 for which he

claimed 2m - 1 was prime . Modern digital computers have been em

ployed to check and extend Mersenne 's results and it has been found
that two values 67 and 257 which Mersenne stated gave primes, do
not, and that there are three others less than 257 which do give
primes and which Mersenne missed. Your own calculations have prob
ably convinced you that for large values of m it may be hard to
tell whether 2 - 1 is prime or not. However, we could decrease
the number of trials by noticing that if m itself is not prime,

then 2 - 1 cannot be . Therefore we have to test only 2P - 1


10
where p is prime. Think this over and see if you can prove the

statement: If m is not prime, 2m - 1 is not prime.


The Mersenne primes with m less than 2300 are now complete
ly determined. The values of m which give Mersenne primes are
2, 3, 5, 7, 13, 17, 19, 31, 61, 89, 107, 127, 521, 607, 1279, 2203,
and 2281. Accordingly seventeen even perfect numbers are known.
The last five of these were found in 1952 by SWAC, the digital com.

puter at U.C.L.A. The Mersenne prime 2 - 1 is also the largest


prime known. It has at least 686 digits and gives a perfect number

with at least 1372 digits.


There are still two unsolved problems concerning perfect num
"
bers We have shown a number in Euclid's form 2 (2 - l) is
perfect whenever 2m - 1 is prime. It can also be proved that any
even perfect number must have this form Try to prove this for
yourself. (It is not very easy.) However, it is still unknown
whether there are a finite number of even perfect numbers or infin
itely many That is we do not know whether or not there are infin
itely many Mersenne primes .

other problem sounds easier. Find an odd perfect number.


The
At the present time no odd perfect numbers are known and many mathe
maticians think it likely that none exist. However, no one has been
able to prove this The best that is known is that if an odd per
fect number exists, it must have at least six different prime fac-
tors and cannot be less than 1.4x10 14
There is one result about perfect numbers which is true whether
the perfect number is even or odd. Prove that the sum of the
reciprocals of all the divisors of a perfect number is 2.
(HINT: Call the divisors d, , dp , ..., d, and notice that for
every divisor d, , .? = d, , is also a divisor of n . )

We able to restate our original problem of determin


have been
ing perfect numbers in terms of the function <r (n) . But this
arithmetic function is useful in other problems besides that of
finding perfect numbers. If you have read part I of this unit, you
11
may remember that we found an expression for the number of divisors
of an integer n , T (n) .

m,
If nip m~ m
n =
p^ p2 pg .. prr , we found that f(n) =
(m1+l)
.

(nig + 1) ... (mr


+ l) . In this part we found that
m,
P p nip
mc
Plx) (l
=
<r(n) (1 +
?1
+
p^
+ . . . + +
p2 + p| + . . . +
v2)
m
... (1 +
Pr +....+ p_ ) . If this expression for <T(n) is
multiplied out we get a sum which contains as summands all the
divisors of n and each exactly once. Hence if we replaced each
suramand by a 1 we would get for the sum exactly f(n) , the
number of divisors of n . This is easily seen by replacing each
p in the formula for <T(n) by 1 and then the formula reduces
to our formula for f(n) .
Thus we can look at ?*(n) as a generalization of t(n) .
This is sometimes indicated by writing 0£(n) = T (n) , the sub-
script zero indicates that we are taking the sum of zero powers
of the divisors of n . <T%(n) =
(T(n) is the sum of the first
powers of the divisors of n . Similarly mathematicians found it
natural to ask for the sum of the k powers of the divisors of
n . Try to devise a formula for the sum of the k powers of the
divisors of n .

(HINT: (Tk(n)
=
(lk + pk + (p2)k + ... + (p^)k) ... (lk + pk +

(p^)k
- ... + (p/)k) - Simplify.)
13
3.

ARITHMETIC FUNCTIONS

III
DISTRIBUTION OP PRIMES AND THE FUNCTION ir(n)
THE
One of the most interesting problems in the study of the
integers has to do with the distribution of primes. A prime is an
integer which has exactly two divisors, 1 and the integer itself.
The first few primes are 2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31.
In the supplement entitled Prime Numbers several interesting facts
about primes are discussed One of these is that there are infin
itely many primes It is also shown in that supplement that there
are arbitrarily large gaps in the sequence of primes. On the other

hand, primes can be as close together as 2 and 3 or 3 and 5.


It isn't hard to see that no two consecutive integers can be prime
after the pair 2 and 3 . Why not? However, as far as the table
of primes has been extended, we still find pairs of primes whose
difference is 2 . Such primes are called "twin primes". The
first few twin primes are 3 and 5 > 5 and 7 , 11 and 13 .
Exercise 1. Make a table of all primes less than 100.
Exercise 2. Find all pairs of twin primes less than 100.
One of the famous unsolved problems of number theory (the
study of properties of the positive integers) is the question:
"Are there infinitely many pairs of twin primes?"
Another unsolved problem is that of finding an expression for

the n prime number. your table of primes that


You can see from
the distribution of primes seems to be very irregular. Since mathe
maticians have not succeeded in finding a formula for the next prime
after any given prime, a related question could be asked: "How many
primes are there less than or equal to a given integer n ?" We
might give a name to this function which gives the number of primes
< n . It is usually called ir(n) .

Exercise 3. Compute 7r(n) from your table of primes for


n = 10 , 20, 30, hO, 50, 75, 100 .

You can see that finding ir(n) for large values of n is


quite job. aIn fact extending the table of primes gets to be a
formidable job. To decide that a given integer n is prime, we
14

need to be sure that no integer less than n divides n, except 1

of course. After a few trials we notice that it isn't necessary to


try as divisors all integers less than n . 2 doesn't divide If
n then no multiple of 2 will either. If 3 doesn't divide n
then no multiple of 3 will. We could continue in this way and it
quickly becomes evident that we only need to try as divisors prime
numbers less than n , and not all of these. If we don't find a
prime < \/n which divides n , then n must be prime. We can

restate this fact as a theorem.


Theorem. If no prime < \/n divides n , then n is a prime.
Exercise 4. Prove this theorem.

(HINT: If d divides n ,
-| then divides n also.)
Exercise 5- Determine whether 178l and 4079 are primes.
With this theorem, considerably reduce! the work of
we have
deciding whether a given integer is a prime -- we need only try as
divisors, primes which are < \/n . For large n this is a great
help. However, it only tells us about a particular integer n .
Eratosthenes (c. 230 b.c.) devised a method, which we now call the
sieve of Ertosthenes, for sieving out all primes less than a given
integer if
we know the primes up to \/n . It goes like this.
Write down all the integers ^ n . For example, take n = 25 .
123 4. £678£10 1112 13 141£l6l7 18 19

20 21 22 23 24 25

In this case V25 = 5 The primes < %/25 are 2, 3, and 5 .

Underline all multiples of underline all multiples of 3


2. Then .

Then all multiples of 5 - (Note that some numbers will be under


lined more than once.) Now all the integers which are not under
lined are prime. These numbers are precisely the primes greater
than y/25 and < 25 .

Exercise 6. Use this sieve method to extend your table of


primes up to 225-
If
we return to our problem of finding ir(n) we may use the
idea of the sieve of Erathosthenes to devise a formula for ir(n) .

What we actually did was to take out the integers which were not
15

prime. There were n integers in our list. First, we took out the

multiples of 2 . There would have been ^


of these if n had

been even. Since n was odd, £ was not an integer, and in that
case we took out a number equal to the greatest integer less than

7j , that is . Similarly when we took out the multiples of 3,

we took out a number equal to the greatest integer less than ^


,

in this case ^l^L


. We see then that it would be convenient to
have an expression for this number of numbers which we sieve out
each time. Let us define, then, the function [x] to be the
greatest integer < x . Some examples of this new function are :

[3] = 3 , [2.61] = 2 , [-5-1] = -6 , = 1 > etc- I* ls clear


that [x] takes on only integral values, although its domain of
definition is the set of real numbers. Strictly speaking, then it
is not an arithmetic function, but an integral -valued function.
The number of integers sieved out each time is now represented by

® ^-
If]

-
HI •
<

What we have in mind is to devise an expression for ir(n)


-
-[§]-[§]-... - is
[|]

like where the


[§J
n pk
,

largest prime There are-\/n difficulties


with this . two
<

method. In the first place, was underlined twice in our siev


6

ing process. It was taken out as multiple of and also as


2
a

multiple of . If
we are counting the numbers taken out by our
3

sieve, then we only want to count once. We have taken it out


6

twice. The same thing happened to all multiples of 6. We can

remedy this situation in


by adding back =
[5] tne number
>

of numbers n which are -multiples of


in insures
£2^3"]
. Adding it
6
<

that is taken out only once. However, the same sort of thing
6

happens with other numbers like 10, 14, 15, 21, etc. In general
if an integer m = where p^ and Pg are primes, it will
P-^2
,

be taken out when we sieve with p^ and again when we sieve with
So that in all such cases in order to have the integer taken
Pg

out only once, we must add it back in once. better estimate of


A
16
ir(n) would then be an expression like

-■K]-n]---[y +
fefe]+[*,]
+
-+[sfe] .

Even this expression won't quite do. We consider numbers of


must
the form P.1P0P0 . These numbers will be sieved out 3 times; when

we sieve by p, , by p2 , and by p3 . Then they will be added

back 3 times when we add back the muptiples of P.iPp >


P.1P3 > anc*

p0Po So these numbers haven't actually been taken out at all.


n
Consequently we remedy this situation by subtracting T 1

Lplp2p3J
If continue in this manner we can take out all multiples of
we

every prime once and only once and the number of numbers remaining
will be given by the expression

"
( LP1P2P3 J
+
lplp2pJ
+ .'. +
lpn_2Pk_iPk]J
+ (...)
(. . .) etc.
This expression seems to go on indefinitely. However, as soon as

< 1 ,
[—]
= 0 , and the complicated expression actually has only

finitely many terms .

We said that there were two difficulties. We have fixed up


the one of these caused by sieving out numbers more than one time.
The other is that we have taken out all multiples of the primes,
including the primes p, , p2 , . . . , p, , themselves We can

correct this mistake by writing ir(n) = M + ir( V**) - 1 .


Of course Vn) = So that the above formula becomes
ir{

k
.

tr(n) = + -
k
M

1
.

The - comes from the fact that is not prime.


a
1

Let us try the formula for n = 25 . The primes <V25 are


2, 3, and
5
.

piHF])+ [M +[M)
*

«*>-».(B|*
' "
(hm
[2.3.5J
.1
3
17

1R25)
= 25 - (12 + 8 + 5) + (4 + 2 + 1) - 0 + 3 - 1

= 9 -

Exercise 7. Compute ir(l50) using the formula above. ir(225).


Exercise 8. Find the number of primes between 100 and 200 .

The formula we have obtained is an improvement over the origi


nal method of actually sieving, but it is still very time consuming
for large values of n Mathematicians have succeeded in showing

that for very large values of n, ir(n) is asyptotically equal to

lQgnn
; that is - approaches 1 as n gets very large
log n
(log n is the natural logarithm of n ). This theorem is known as
"the prime number theorem" . Until 19^8 the only proofs of this
theorem which were known involved of the deepest
some and most dif
ficult mathematics An elementary proof was found in 1948 by Atle
Selberg. However, this proof is very long and complicated and
elementary only in a technical sense.

Exercise 9- ir(lO,000,000) = 664,580 . Compute


log n
for n = 10,000,000.
(HINT: log n = » loginn , where M =
0.4342945...)
4. 19

THE EUCLIDEAN ALGORITHM AND LINEAR DIOPHATINE EQUATIONS

point in your mathematical experience, you have un


At some
doubtedly encountered word or story problems . Here is one taken
from the Ganita-Sara-Sangraha of Mahaviracarya, a Hindu writer of
the ninth century. "Into the bright and refreshing outskirts of a
forest which were full of numerous trees with their branches bent
down with the weight of flowers and fruits, trees such as jambu

trees, date palms, hintala trees, palmyras, punnaga trees and mango
trees -- filled with the many sounds of crowds of parrots and
cuckoos found near springs containing lotuses with bees roaming
around them --a number of travelers entered with joy. There were
63 equal heaps of plantain fruits put together and seven single

fruits. These were divided evenly amoung 23 travelers. Tell me


now the number of fruits in each heap." If we translate the prob
lem into ordinary algebraic language (it is a shame to do such a
thing to so beautiful a problem, but it does help to simplify the
process of finding a solution), it looks something like this:
63x + 7 = 23y ,

where x is the number of fruit in each heap and y is the number


each traveler receives From the nature of the problem it is clear
that only solutions in positive integers are acceptable.
The question now is, how do we find solutions in integers to
such equations .

One way might be to draw a graph of the straight line


ax + by = c and see if it
points with positive
passes through any
integral coordinates- This particular equation does. Draw a graph
of the equation. Can you find a solution from your graph?
Suppose our flowery Hindu problem had translated into the
equation 3x + 6y = 13 . Does this equation have a solution in
positive integers? Why?
Solve 3x + 6y = 24 for x and y positive integers. Is
there more than one solution? How many are there? For what posi
tive solution is x smallest? For which positive solution is y
smallest?
Consider the equation 2x - y = 6 . Find a solution with x
and y positive integers Is there more than one such solution?
20

How many positive solutions are there?


After the last three examples, it would seem that an equation
ax + by = c with a, b, and c integers may have no solutions, a
finite number of positive solutions, or an infinite number of posi
tive solutions. Can you tell which one of these cases you have by
looking at the graph of the equation? From your consideration of
the graph of the equation try to write down conditions on the line
which will cover all possibilities for the number of positive solu
tions.
If the numbers involved are quite large finding solutions from
a graph might be very difficult. Fortunately
completely we can
solve this problem of finding integral solutions without using
graphical methods at all. To do this we need to be able to tell
when a solution exists; and if
solution exists, we would like to
a
have a method (besides guessing or trial and error) which will
always lead us to a solution. Finally it would be nice if we could
devise an expression which would tell us all possible solutions in
integers for the equation. All (these) things are possible for
those who like mathematics.
First consider the following equations :

(1) 2x + 3y =
5 (4) 4x + 6y = 9

(2) 2x + 4y = 5 (5) 4x + 6y = 8

(3) 3x + 3y = 5 (6) 2x - 4y = 4 .

of these have integral solutions?


Which
Look at the coefficients of x and y and the constant term
in each of the equations for which you found solution. Is it a
true that any number which divides both the coefficient of x and
the coefficient of y divides the constant term? Do you think
this must be true of any equation which has a solution? State this
result as a theorem and write out an informal proof for the theorem.
(HINT: Call ci the greatest common divisor of a and b . If we
used the notation gcd(a,b) = d , then gcd (2,4) = 2; gcd (9,12)=
3 ; gcd (2a, 3a) = a ; gcd (abc, abe) = ab, etc You can see that
this is a very clumsy notation. We might abbreviate, when it is
clear that we mean the greatest common divisor of two integers, by-
omitting the letters gcd. Then (40,24) = 8 means gcd (40,24)= 8.
21

Unfortunately, this notation (a,b) is used in several different


ways in various parts of mathematics. However, as we have noted
above, if a and b are integers and we write (a,b) = d for the
greatest common divisor it isn't easily confused with the other
uses of the symbol . )
This theorem which you have arrived at states what is called a
NECESSARY condition that the equation ax + by = c have a solution
in integers. This is a reasonable use of the word necessary since
the equation cannot have a solution unless (a,b) divides c . The
condition is truly necessary for a solution of the equation.
Mathematicians love to find a neat condition which is necessary
and which also insures that a given problem has a solution. That
is, it would be nice if two things were true -- (l) that
ax + by = c has no solution unless (a,b) divides c and (2)
that if(a,b) divides c , the equation always does have a solu
tion in integers. You have met this idea before in Chapter I where
the phrase "if and only if" was used. We could restate our hopeful
statement above as : The equation ax + by = c has a solution in
integers if
and only (a,b) divides c .if
Look again at our six equations above. Does it seem to be
true that if
(a,b) divides c, there is a solution? We shall
now try to devise a way to prove that this is always true.

How do you find the greatest common divisor of two integers?


In all the cases we have considered, it has been easy to do just by
looking at the two integers . How did you do it in the seventh and
eighth grades when adding fractions with different denominators?
One way of course is to write out the factors of each integer and

pick out those which are common. For instance, to find (248, 312),
we write 248 - 23 .
31 and 312 = 23 . 3 . 13. Then clearly
(248, 312) = 8 . However, suppose the numbers are large and it
isn't easy to find the factors of either number. For example,
suppose we are asked to find (782, 3315) . The usual method works
of course, but is not as easy as in the cases we have previously
encountered. Another method which solves this problem is attributed
to Euclid (who lived about 300 B.C.).
22
It goes like this:
3315 = 782.4 + 187
782 = 187.4 + 34

187 = 34.5 +@
34 = 17.2 + 0
Euclid's (or algorithm) gives 17 , the last non-zero re
method
mainder, as the greatest common divisor of 3315 and 782. That
17 is the greatest common divisor can be proved as follows. First,
proceeding from the bottom to the top, we can see that 17 divides
each number on the left hand side as follows :
34 = 17.2
187 = 17 (2.5 + 1)
782 = 17 (2.5 + l) 4 + 17.2 = 17 {(2.5 + 1). 4 + 2]
3315 = 17 {(2.5 + 1)
-
4 + 2} 4+17 (2.5 + l)

l)j
=
( {(2.5
17 + 1) .
4 + 2} .4 + (2.5 + .

Thus we have shown that is a divisor of both 3315 and 782


17 .

It is, then, a common divisor.


Now let us show that it is the greatest common divisor. We

do this by showing that any number d* which divides both 3315


and 782 must divide 17 integer divides
Then 17 it if an

,
.

cannot exceed 17 Hence 17 must be the greatest common divisor.


.

To prove this we simply reverse the process of the preceding para


graph. Suppose d* divides 3315 and 782 then it must divide
;

3315 - 782.4 = 187 Why? Next if d* divides 782 and 187

,
.

it divides 782 187 .4 34 . But then


- = d* divides 187 and if
34 it divides 187 - 34.5 = 17 . So we see that any number which
,

divides both 3315 and 782 must divide 17 Therefore 17


.

must be the greatest common divisor of these two numbers.


You may have noticed that we have used repeatedly a very
obvious fact, namely, that if an integer divides each of two
integers, it divides their sum and their difference. This is a
trivial but extremely useful theorem. Write out proof for this
a

theorem giving reasons for each step.


Suppose we try Euclid's method on 253 and 122.
23

253 = 122.2 + 9

122 = 9,13 + 5

9=5.1+4
5 = 4.1 +(T)
4=1.4+0
The greatest common divisor is 1 , the last non-zero remainder.
Such which have
numbers 1 for their greatest common divisor are
called relatively prime . Check that (253, 122) = 1 by factoring
the two numbers .

Find the g.c.d. of 1596 and 96 . Find (4l8, 1376);


(365, 146) .

To prove that Euclid's method always gives us the greatest


common divisor for any two integers a and b , we can proceed as
follows :

(Suppose a > b . )
a = b,q, +
r^
b = +
rl'Q2 r2

rl =
V2% +
r3

r2
=
r3.q4 + r.4

rn-2= 'n-i.On rn * where rn ls the


+

last non-zero remainder. (Is it clear that there will always be a


last non-zero remainder? Why?) To show that r is the greatest
common divisor, we must show that r is a common divisor; that is,
that it divides both a This is left to the reader. He
and b .

can same way that we did in the first example


argue in exactly the
with 17 , 3315 > and 782 . Then we must show that any common
divisor of a and b divides r . The argument again is the
same as in the example.
We now have a fool-proof method for obtaining the greatest
common divisor of any two integers . Actually we have done a good
bit more. Not only can we find d = (a,b), but we get as a bonus
24

a solution to the equation ax + by = d . In the case of 17,


3315, and 782 we are in a position to solve the equation
3315x + 782y = 17 . The solution is as follows :
Euclidean Algorithm Solution of 3315x + 782y = 17
3315 = 782.4 + 187 187 = 3315 - 782.4
782 = 187.4 + 34 34 = 782 - 187.4
= 782 - (3315 - 782.4)4
= 782 (1 + 4.4) - 3315 (4)
187 = 34.5 + 17 17 =187-34.5
=
(3315 - 782.4) - {782(1 + 4.4) -
3315(4)}.5
= 3315 (1 + 4.5) - 782 (4 + 4.4)5}
= 3315 (21) - 782 (89)
So that set x if we = 21 and y = -89 we have a solution to the
original equation.
You will
that in the beginning of this discussion we
remember
were trying to find solutions in integers to equations of the form
ax + by = c . We found that in order for the equation to have a

solution at all, (a,b) = d had to divide c . The claim was made


that if
this happened, there was always a solution in integers for
the equation. We are now in a position to show that this is true.

Suppose you stop reading at this point and try to find out how to
get a solution from what we have done so far.
Check your method with the following. From our discussion
above of Euclid's
algorithm, it is clear that we can always solve ax + by = d where
d = (a,b). To find a solution of the original equation let
c = d.c' . Now take the equation ax + by = d and multiply both
sides by c' We get
a(xc') + b(yc') = dc' = c .

It is xc' and yc ' are solutions to our problem.


clear then that
This is really a nice result. We have a method for finding a

solution to any equation which has a solution.


There is just one thing -- the solution we get may not be in
positive integers x and y . Of course there may not be any
solutions in positive integers, but in our "beautiful forest"
problem, clearly only positive solutions are acceptable. While it
25
is wonderful always to be able to get one a real mathe solution,
matician, at this point, would certainly wonder "Isn't there some
way to find all the solutions?". Try to find a way to find another
solution from the one obtained by Euclid's algorithm. Can you now

find all solutions?


(HINT: Suppose x and y satisfy the equation, i.e.,
ax + by = c = and suppose x and y are any other pair of

numbers which satisfy it, so that ax + by = c . Then subtract


the first
equation from the second, divide both sides of the
resulting equation by d , transpose, try to see what can be said
about (x - xQ) (y - yQ).)
and
When you have made as much as you can out of the "hint" ,

check your results with the reasoning in the answer sheet. You
will find there that the general solution may be given in the form
x =
x„o + §d t
y = y -
|t where t is an integer t .

It is easy to check that for any integer t the x and y given


above do satisfy the equation provided x and y do. Check

this for yourself. It is


clear from this check that this x and
y will satisfy the equation for any value of t . Is it also
clear that any solution must have this form for some integer t ?
Try to show that this is true.
We are now in a position to find all the positive solutions
for our original equation if any exist. Let us take the equation
3315x + 782y = 17 again. By our method we get the solution
x = 21 and y = -89 . Are there any positive solutions? Well if
we look at the general solution obtained above, for this equation

it assumes the form x = 21 + 46 t , y = -89 - 195t . To find


positive solutions we must have t which satisfies x = 21 + 46t
> 0 and y = -89 - 195t > 0 . However, if t satisfies both of
these inequalities it must be an integer > -21/46 and at the same
time < -89/195 . There is no integer satisfying both of these at
the same time.
26

(Plot these 2 numbers on the real line and look for the integers
to the right of -21/46 which are also to the left of -89/195.)
Consequently there are no positive solutions. Of course in this
particular problem this is clear from looking at the equation.
However, the method we have used will
lead you to the values of t
which give all positive solutions in any other problem.
Now you are in a position to find out the number of fruits in
each heap in our original problem. Go to it.
What is the smallest number of fruit there could have been in
each heap? Are there infinitely many positive solutions? Write
out the general formula for all solutions.
Here are a few more problems which you can solve using the
methods of this unit.
1. l6x + 7y = 601.
2. Find the positive solutions for the equation lOlx + 753y =

100,000.
3. Say quickly, mathematician, what is the smallest multiplier
by which 221 being multiplied and 65 added to the
product the sum divided by 195 becomes exhausted?
(From the Lilavati of Bhaskara (1150 A.D. ).)
4. In the forest 37 heaps of wood apples were seen by the
travelers. After 17 fruits were removed, the remainder
was divided evenly among persons.
79 What is the share
obtained by each? (Mahaviracarya)
5. l4x - 45y = 11 .

6. 40x - 63y = 135 .


27

5.

THE GAUSSIAN INTEGERS

In order to be able to find solutions to all quadratic equa-


p
tions ax + bx + c = 0 where a, b, and c are real numbers, we

found it to extend our number system to include numbers


necessary
whose squares are negative. In fact, we adjoin to the set of if
real numbers a number with the property that its square is -1 ,
define addition and multiplication for this extended set, we
achieve a new number system in which every quadratic equation
(or more generally every polynomial equation) with coefficients
in the new system has solutions.
This extension of our number system consists of the set of
all numbers of the form a + bi where a and b are real num-
bers and i is a number with the property that i p
= -1 . This
new system is called the field of
of our complex numbers . Most
work in mathematics in grades one through eight is arithmetic.
Let us now investigate arithmetic in the complex number system.
In ordinary arithmetic we were mainly concerned with the positive
integers. The question naturally arises "What are the integers of
our new extended number system?" To try to devise a reasonable
definition, we try to generalize some property of the ordinary
integers in the system of rational numbers so that these "rational
integers" will still be "integers" in the extended system, and so
that as many characteristics of the ordinary rational integers as
possible will retained. be
In keeping with our interest in solving equations, the property
of the rational integers that we choose to generalize is the prop
erty that they are solutions of linear equations, x + a = 0 ,
with rational integral coefficients. For our purpose it is con
venient to restrict our attention to a subset of the complex num
bers, namely the set of numbers {a + bi} where a and b are
rational. We then define Gaussian integers to be those complex
numbers a + bi , a and b rational, which satisfy an equation
2
of the form z + mz + n = 0 where m and n are ordinary
28
rational integers. These new integers are called Gaussian integers
in honor of Carl Frederich (1777-1855) the German mathemati Gauss
cian who is ranked with Archimedes and Newton as one of the three
greatest mathematicians of all time. Gauss was the first person
to systematically develop the properties of these new integers,
and, in particular to show that the Fundamental Theorem of Arith
metic (Every integer can be written as the product of primes and in
essentially only one way.) holds for these integers.
Suppose we now consider the form which the new Gaussian inte
gers must have. We remember that the definition requires that they
be numbers of the form a + bi , a and b rational, which
p
satisfy an equation z + mz + n = 0 , m and n ordinary ration
al integers. If b = 0 , the Gaussian integer is a rational number

a =
£ . Suppose that £ has been reduced to that p and q are

rational integers with no common factors. Thence since a = *


p
satisfies z + mz + n = 0 , we have
2
£2 -+ m£ + n = 0
q. q
2 2
p + mpq + nq =0
o
p = .q(mp :- nq) .

2
q then divides p . But since p and q have no common factors,
q must divide p and q must actually be 1 . (If not q is a

common factor of q and p . ) But if q = 1 , then a is actual


ly a rational integer.
There remains the case when b / 0 . In this case from the
quadratic formula we have that if a + bi is a root then a - bi
is also.
Accordingly
(z - (a + bi) (z - (a - bi)) = z + mz + n
2 - 2ac 2 2 2
z + a + b = z + mz + n

and m = 2a
2 2
n = a^ + b^ .
29
Then

(1) a =
3? and

+i An -
■*/
=
(2) b m2

Since b is rational
p P
(3) 4n - m = c , where c is some rational integer.
Substituting (3) in (2) we have

(4) b - ± c .

P P
The equation (3) can be written 4n = m + c
This means that m and c are either both even or both odd.
They cannot both be odd.
Exercise 1.
Prove that the sum of the square of two odd numbers is not a
multiple of 4 .
Therefore both m and c are even and a and b are rational
integers.
We have then in both cases that a and b must be rational
integers and we are now able to say that the Gaussian integers are
complex of the form
numbers a + bi where a and b are actually
rational integers.
It is easy to check that the sum, difference, and product of
two Gaussian integers is a Gaussian integer.
Exercise 2.
Show that the sum, difference, and product of two Gaussian
integers is a Gaussian integer.
We see then that our new integers behave at least in these
respects like ordinary rational integers. When we come to division
we must look a little more closely.

Exercise 3.
Is the quotient of two rational integers a rational integer?
Justify your answer.
Exercise 4.
Is the quotient of two Gaussian integers a Gaussian integer?
Justify your answer.
30
previous exercise shows us that division is not always
The
possible in the set of Gaussian integers. Let us then define
division for Gaussian integers precisely. We say that the Gaussian
integer <* is divisible by the Gaussian integer (3 if there is a
Gaussian integer If such that c* =
(5 }f

Example 1 .

Is 2 + 3i divisible by 1 '+ i ?

SOLUTION:
If3i is divisible by 1
2 + + i , then there must be a
Gaussian integer x + yi such that
(1 + i)(x + yi) = 2 + 3i .

Then (x - y) + (x + y) i = 2 + 3i and
x - y = 2 ,
x + y = 3 .

x = 5 y = I
Since these are the only possible values for x and y if
x + yi satisfies the original equation, and since these are not
rational integers, our answer is "No, 2 + 3i is not divisible
by 1 + i
."
Exercise 5-
Is 2 + 3i divisible by 2 - 3i ? by i ?

Exercise 6.
Is 3 + 11i divisible by 2 + 3i ? by -i ?
We have seen that the conjugate, a - bi , of the complex
number, a + bi , is useful in many questions concerning complex
numbers. We use the conjugate to define the norm of a complex
integer. The norm of a + bi is defined as (a + bi)(a - bi) =
2 2
a + b . We immediately notice several things about the norm of
a complex integer. In the first place, it is a rational integer
since a and b are. In the second place it is non-negative.
If b = integer a is a . These
0, the nor.n of the rational
o

properties prove very useful in trying to settle many questions


about Gaussian integers.
If we look into the divisibility properties of the Gaussian
integers, we are led to consider the integers which correspond to
31
1 and -1 integers; 1 and -1 are the only.
among the rational
rational integers which divide every rational integer. We call
these numbers the units of the system of rational integers .
Similarly we define units for the Gaussian integers to be those
Gaussian integers which divide every Gaussian integer. We can

determine the units for the Gaussian integers quite easily by


first using our new notion of the norm.
We first need the preliminary theorem or
Lemma. N(<*^3)
=
N(o< ) n(^ ) , where N(<* ) denotes the norm
of <* .
_
Proof: If we let ot be the conjugate of <* and fi be the
conjugate of (3
,

N( c*
) = ok oi

N(/3 ) . /3 /§ _ _
N(*/3) =
«fi <x/3 Since « fi = oc . /3

=
N(« )N(/3)
The lemma can also be proved directly from the definition of
the norm. Let c*. = a + bi ,^3=c + di and write out the details
of this proof.
It is now easy to show
Theorem 1. u is unit if and only if N(u) = 1 .
a
Proof: If u is a unit, it divides every integer and in
particular the integer 1

Then 1 = u*v for some Gaussian integer v .

By the lemma, N(u)N(v) . N(l) =

But N(l) = 1 = N(u)N(v) . Since the norm of any integer is a


positive rational integer N(u) = N(v) = 1 and the "if" part of
the theorem is proved.
Now suppose N(u) = 1 .

Let u = e + fi . Then e2 + f2 = 1 and either e = 0 and


f = +1 or
f + 0 and
e = + 1 .

Hence if N(u) =1, u = 1 , -1 , i , -i . But 1 and -1 clearly


32

divide any Gaussian integer a + bi . Also a + bi =


i(b - ai)
and - bi = i(-b + ai) . Hence these four integers divide every
a
Gaussian integer and are therefore units.
q.e.d.
have as a bonus from this theorem, the
1,-1,
We
Corollary: The units of the Gaussian integers are
i , and -i .

When finding the divisors of a rational integer n , it is

only necessary to consider positive divisors of positive integers


n , since for any divisor d of n , -d is always a divisor of
n . Similarly if n is negative whenever d divides n so does
-d . We could describe this situation by saying that n and -n
are associates; i.e., the associates of an integer n are integers
obtained by multiplying n by units. In the case of rational
integers n has only the associates n and -n . If we extend
the associates of <X to be the Gaussian integers obtained from
by multiplying <* by units. Thus the associates of any Gaussian
integer are ci , -o( , <K , i
and -i .
If we now consider the divisors of a Gaussian integer, °< ,
we need only concern ourselves with divisors which are not units

or associates of .

Exercise J.
Show that if ck and ft are associates their norms are equal.
We are now able to define a Gaussian prime as a Gaussian in

teger which is not a unit and which has no divisors except units
and its associates. Several interesting questions can now be asked.
1. Are rational primes Gaussian primes?
2. Are there infinitely many Gaussian primes?
3. rational integers are Gaussian primes?
Which
We can answer the first without much trouble. 2 is a ration
al prime. However 2 = (l + i)(l - i) . Since 1 + i and 1 - i
have norm 2 , they are not units. The associates of 2 are 2 ,
-2 , 2i , and -2i . Therefore since 1 + i and 1 - i are
neither units nor associates of 2 , the rational prime 2 is not
a Gaussian prime.
33

Exercise 8.
Is 5 a Gaussian prime?
Exercise 9.
Is 3 s a Gaussian prime?
Let us now look more closely at rational primes of the form
4n + 3 . Suppose a rational prime p = 4n + 3=<*K . Then
N(p) =
N(cx)N(^3 ) = p . If p is not a Gaussian prime, then
there must exist ck and ft such that N( oC ) ^ 1 and N( ft ) £ 1.
In that case, N( <*
) = p and N( ft ) = p . But if <* = x + yi ,
o p
N(oc)=x+y = p = 4n + 3. This is impossible for no integer
of the form 4n + 3 is the sum of two squares.
Exercise 10.
Prove that no rational integer of the form 4n + 3 is the
sum of two squares by considering all possible cases for x and
y (both even, both odd, one even and one odd) .
Since the norm of <* and ft cannot be p , the norm of one of
them must be 1 and that one is a unit, and the other is an asso
ciate of p . Since divisors except units
p has no and associ
ates of p , we have proved the following theorem.
Theorem 2. Every rational prime of the form hn + 3 is a
Gaussian prime.
This proves also that there are infinitely many Gaussian
primes, since in the supplement Prime Numbers it is proved that
there are infinitely rational primes of the form 4n
many + 3 .

.And we have thus answered question two in the affirmative.


Exercise 11.
Is 1 + i a Gaussian prime?
Exercise 12.
Is 1 - i
a Gaussian prime?
Exercise 13.
Is any composite rational integer a Gaussian Prime?
preceding discussion and exercises, we have the re
Prom the
sult that the only rational integers which are Gaussian primes are
rational primes of the form 4n + 3 and possibly some rational
primes of the form 4n + 1 .
34

To settle the question about the existence of rational primes


of the form 4n + 1 which might also be Gaussian primes we need
two results. The first is a theorem which is rather easy to prove.
Theorem 3. If N( <* ) is a rational prime, * is a Gaussian
prime .

Proof: Suppose <* = fit*


Then N(c* ) =
N( ft )N( Y ) .

By hypothesis N( <x ) =
N( (3 )N( f ) = p , where p is a
rational prime.
Since N( ft ) and N( ) are rational integers, one of these is
1 and the other is p . The one whose norm is 1 is a unit and
we have the result that <* can only be written as a unit times an
associate of * . Therefore <* is a Gaussian prime.
q.e.d.
The other result which we need is that any rational prime of
the form 4n + 1 is the sum of two squares.
Exercise 14.
Write the following rational primes as the sum of two squares,
(a) 5 (b) 13 , (c) 17 , (d) 29 , (e) 101 , (f) 1721 .
,
Since the proof of this result requires more machinery from
the theory of numbers than we have available, we will not give the
proof here. (A proof can be found in any elementary number theory
book. )
We are now in a position to settle the question about rational

primes of the form 4n + 1 . Suppose p - 4n + 1 = x


P
+ y
p
. We

can factor p as follows:


p = x2 + y2 =
(x + yi)(x - yi) .
Then the Gaussian integers x + yi and x -yi have norm p and
o
by Theorem 3 Since the norm of p is p
are Gaussian primes.
and the norm of x + and yi
is p , by Exercise 7 the x - yi
primes x + yi and x - yi are not associates of p . Then p
is the product of primes, which are not associates of p . We have
therefore proved
35

Theorem 5. No rational prime of the form 4n + 1 is a


Gaussian prime.
The answer to question three, then is: The only rational
integers which are Gaussian primes are the rational primes of the
form 4n + 3 .

Actually, it can be shown that the Gaussian primes are of


three kinds :

(1) rational primes of the form 4n + 3 and their


associates,
(2) 1 + i , 1 - i and their associates,
(3) integers of the form x + yi and x - yi where x and
2
y are positive, x is even and x + y
2
is a rational
prime, and their associates.*

*A linen manufacturing company: N. W. Linnenfabrieken, E. J. E.


van Dissel and Zonen, P.O. Box 272, Eindhoven, Holland, makes a
tablecloth 28" x 28" in which the Gaussian primes form the woven
design. It is available in red, green, blue, and yellow at $2.00
each.
37

FERMAT' S METHOD OF INFINITE DESCENT

The theory of numbers (the study of properties of the positive


integers) fascinating and difficult branch of mathematics.
is a
The French provincial government official and amateur mathemati
cian Pierre Fermat (l601?-l665) devoted much of his leisure time
to systematically cultivating this branch of mathematics.
One of the baffling aspects of number theory is the absence
of many general methods for attacking problems in this field.
Fermat devised an ingenious method which he called "the method of
infinite descent" to handle certain kinds of problems. It is some
what like mathematical induction in reverse. Instead of showing
that a certain proposition, P(n) , is true for n = 1 , and when
ever P(k) is true, P(k+l) is also, we begin at the other end.
We first suppose that P(n) is true for some integer. We then
show that if it is true for any particular integer, it is true for
a smaller one. Since on the one hand this argument can be repeated
indefinitely other hand there are only finitely many
and on the
positive integers less than a given positive integer, we have a
contradiction. This means that our assumption that the proposi
tion is true for some integer is wrong, and we have the result that
the proposition is not true for any integer. In this form it would
seem to be especially useful for disproving theorems.
The argument can be modified, however, to prove positive
statements. Fermat said that he used it to prove that any prime
of the form 4n + 1 can be written as the sum of two squares. For
instance 5 = 22 + l2 , 13 = 22 + 32 , 17 = 42 + l2 , 29 = 22 + 52 .

Exercise 1.
Write 37, 4l, 89, 101 as the sum of two squares. Can this be
done in more than one way?
38
Fermat's argument goes as follows. Suppose an arbitrarily chosen
prime, p = 4n + 1 , is not the sum of two squares. He then shows

that there is a smaller prime of this form which is not the sum of
two squares. Continuing in this way he arrives at the result that
5 is not the sum of two squares. But 5=2 2 +1 2 . This con
tradiction means that there was no prime of the form 4n + 1 which
was not the sum of two squares. We do not have Fermat's proof of

this theorem and in fact it was not until 17^9 that the first
rigorous proof was given by the Swiss mathematician Leonard Euler
(1707-1783).
Fermat discovered many deep and interesting properties of the
integers. Very few of his proofs have come down to us; however,
his method of infinite descent can be used to prove a special case
of one of the most famous theorems in mathematics, Fermat's Last
Theorem. In a margin of Bachet's Diophantus , Fermat made his
famous note regarding the problem of finding rational solutions of
the equation

(1) x2 + y2 = Z2 .

"On the contrary, it is impossible to separate a cube into two


cubes, a fourth into two fourth powers, or, generally,
power any
power above the second into two powers of the same degree: I have
discovered a truly marvellous demonstration which this margin is
too narrow to contain." Mathematicians are uncertain as to whether
Fermat actually had a proof; however, no proof for all powers
greater than 2 has yet been found.
2 2 2
The equation x + y = z , of course, does have solutions;
for instance 3+4=5.
2 2 2
In fact we now obtain all solutions
for this case as follows. We first note that we need only look for
solutions x , y , and z which have no common factors, since if
x + y = z , then certainly (kx)2 + (ky)2 =
(kz)2 , and converse
ly.
Exercise 2.
Show that if any two of the integers x , y , and z in (l)
have a common divisor, d , then d divides the third.
39

Accordingly we will
consider only solutions which have no
common factors. In this situation not all three integers x , y ,

and z can be even. Why not?


Exercise 3.
Show that not all three integers x , y , and z in (l) can
be odd.
Exercise 4.
Show that it is impossible for two of the integers x , y ,
and z in (l) to be even and one of them odd.
The preceding exercises show that the only possibility for a
solution to (l) is for one of the integers to be even and the
other two to be odd. Suppose x is even and y and z are odd.
Let
(2) x = 2u . Then (l) becomes

(3) 4u2 + y2 = z2 or
u
4u
2 = z
2 - y2

(*) ^2 =
(z + y)(z - y) .

Since z and y are odd, z + y and z - y are even. If we


consider any common divisor of z + y and z - y , it must divide
their sum, 2z , and their difference, 2y . We know that 2 is a
common divisor, but if there were any other besides 2 , it would
have to divide both z and y . However, we excluded this case
in the beginning.
At this point we must pause to prove
Theorem 1. If the greatest common divisor of a and b is 1
2
and ab = c , then a is a square and b is a square.
Proof: of Arithmetic (see the
By the Fundamental Theorem
supplement entitled The Fundamental Theorem of Arithmetic) , we may
write c as the product of prime factors p1 , pg , . . . pn .
Then
2 2
c =
(P^Pg- - = ab -
-Pn)
Clearly p1 divides ab . If p^^
divides a it does not divide
2
b since a and b have no common factors. In this case p^

must then divide a . If p, does not divide a , then it must


40
2
divide b and similarly in this case p.. will divide b_ . We

can make the same argument for each prime p. . Hence if any prime

divides a , so does its square; and this prime does not divide b_.
The same statement can be made for b_
. Accordingly, if we let
p. be the first prime that divides a , p. be the second, etc.;
xl x2
p. be the first prime that divides b_, p. be the second,

pop
1k+l xk+2
etc.; we must have
p
a =
p1 p1 ...p1 =
(p., p., ...P4 )
xl x2 xk 11 x2 xk
2
b = p. p. ...p, = (p., p1 ...p^ )
k+1 xk4.2 xn xk+l .Lk+2 """n

q. e.d.
We now return to our problem of finding the solutions to the
2 2 2
equation x + y = z . Since the greatest common divisor of
z + y and z - y is 2 , we can write (4) in the form
tu = *. 2 2
>

where the greate !St common divisor of Z and Y is


2
Then u = Z.Y and by theorem 1
2 2
Z = v and y = w and
p
(5) z + y = 2v

(6) z - y = 2w2 .

Exercise 5.
Show that v and w have no common factors.
Then substituting (5) and (6) in (4) we have

4u2 =
(2v2) (2w2) or
2 2 2
u = v w and
(7) u = vw .

Substituting (7) in (2) we have


(8) x = 2vw ,

\ {(5) - (6)) gives


4i
(9) y = v2 - w2 ,

\ ((5) + (6)) gives

(10) z = v2 + w2 .

Since y and z are both odd, one of v and w is even and the
other odd.
In the beginning we supposed that x, y , and z were any solu
tion without common factors and we have found the form which they
must assume.
We have then
2 2 2
Theorem 2. The solutions of x + y = z are given by
x = 2kvw ,

y =
k(v2 - w2) ,
o p
z =
k(v + w ) ',

where k is any integer and v and w are any integers chosen


so that they have no common factor and so that one is even and the
other odd.
Fermat ' s Last Theorem can be stated as follows : There are no
integers for which xn + yn = zn if n is
x, y , and z
greater than 2 . The proof for the special case n = 4 serves
as a good illustration of Fermat 's method of infinite descent.
Theorem 3. There is no solution in integers for

x
4
+ y
4
= z
4
. As above, if the equation has a solution x , y, z
and any pair of these integers
factor, that common has a common
factor then divides the third integer and both sides of the equa
tion can be divided by the fourth power of that common factor. So
if there is a solution, we can assume that the x, y, and z are
relatively prime in pairs; that is, every pair has greatest common
divisor 1 .
We also notice that if we can show that x
4
+ y
4
= z
2
is
impossible then so is x 4 + y 4 = z 4 , since the sum of two if
fourth powers isn't a square, it certainly can't be a fourth power.
therefore prove the simpler statement that x 4 + y 4 = z 2
We

has no solution in integers. The proof by infinite descent follows.


42
Proof: Suppose there is a solution x , y , and z relative-
4 4 2
ly prime in pairs, say x + y = z . This equation can be
rewritten
(x2)2 + (y2)2 =
(z)2 .

However from Theorem 2 we know that

x2 = 2ab ,

(11) y
2 = a
2 - .2
b ,

z
2
= a
2.2 + b ;

for integers
some a and b with greatest common divisor 1 and
one of these even and the other odd. Suppose that b is even.
Since x = 2ab =
a(2b) and a and b have no common factors,
by Theorem 1, 2b is a square and a is a square. Set

a = d2 .

■p p p
From (ll) we have that a = b + y , and again by Theorem 2
b = 2rs ,

2
(13) y = r2 - s ,

a = r2 + s
2
; where r and s have no
common factor. But from (12) and (13) we have
p
2b = c = 4rs .

2
By Theorem 1, then r =
x^
,
4 4 2
s =
x^
+
y^ and since by (12) a = d ,

x,4 4 2
we have +
y1
= d , where 1 < d < a < z .

But now we have a solution x, y , d to the equation


4 4 2
x + y = z in which d is less than z . What we have actually
shown is that if x
4
+ y
4
= z
2
has a solution we can always find
another solution with smaller z . But this is impossible since
there are only finitely many positive integers less than a given
2
integer z . Therefore there is no solution to x
4
+ y
4
= z and
43
4 4 4
consequently no solution to x + y = z
Fermat also used this of proof to show that if both of
method
the legs of a right triangle are integers, the area cannot be a
square . The proof of this statement is similar to the one given
above .
45

7.
APPROXIMATIONS OP IRRATIONALS BY RATIONALS

In Chapter 1 of your text you learned that a real number can


be represented by an infinite decimal. The main features of this
representation are the following:
Let the real number be g + <* , where g is an integer and
0 < <* Since the integral part g
< 1 . offers no difficulties,
we can consider only and write it as a decimal: =
O.a^aga^...

(1) Each decimal section is a rational number. (By the nth


decimal section of the decimal O.a^^g..., we mean the number

0.a1a2... an , i.e., + + . )
a1^/lO a^y/lO2 ...an^/lOn
(2) The difference | crt
- . . can be made as
0.a^a2. an|
small as we please if we choose n large enough.
(3) <*. - £ 10"n
0.a1a2...an .

(4) The denominator of each decimal section is a power of


10 .

You may not have noticed property (3) before. It is easily


proved, for «. - 0.a1a2...an = 0.0 . . an+1 an+2
..-
.^0
n
£ 0.0 . . . 0 99 ... = 0.0 ... 0 1 . (The numbers under the braces
n n-1

indicate the number of 0's .) Of course, (2) is a consequence of


(3).
Properties (2) and (4) seem to be rather special; they arise
from the fact that we are using decimal sections to approximate oc .
There seems to be no particular reason to do this, and in fact we
might get better approximations if we used general rational numbers
p/q approximations.
as
Before doing this, however, let us realize that there is no
particular point in approximating rational s by other rationals.
Prom now on we shall assume that ok , the number being approxi
mated, is irrational .
46
It is very easy to produce rational numbers p/q for which
I
- p/q | < l/q . Suppose «* lies between the consecutive
integers m and m+1 . Let q be any integer > 1 , and divide
the interval (m , m+l) on the number line into q parts or sub
intervals (see figure) . Each part

1 H

% 3

is of length — ; the points of subdivision are then

m +
q' m +
q '
- - - ' m +
^q^-
- ^e P°^-nt wnich represents <X

will fall inside of the subintervals; it cannot fall on an end-


one
point of a subinterval since is irrational whereas the end-
points are rational. If falls in the subinterval whose left
endpoint is m + i/q , then clearly
m + i/q < <* < m + ,

so that \ ok - (m + i/q) | < l/q , or

(i) I* -|| <


i .

where p = qm + i . Since this process can be carried out no matter


what integer q may be (as long as q > l) , we have proved this

result .
Theorem : If is irrational, then to every integer q > 1

there corresponds an integer p such that (l) is true.


Let us check this result with some famous approximations of
ir = 3.14159265... which were known to the Greeks, namely 22/7 >

333/106 , 355/113 .
47

.
<* - p/q .

22/7 .00126. . . .1429. . .

333/106 .0000832. . . .009434. . .

355/113 . 000000266 . . . .OO8850. . .

These approximations are considerably better than what would


be expected from the Theorem! Is a better result than (l) pos
sible?
Before trying for a better result, let us spend a few words
trying to make the whole concept of approximation more precise.
If <* is the number being approximated and p/q is the approxi
mation, then certainly we want to make |
<* - p/q| "small". But
small - compared to what? We can make | oc
- p/q| as small as
we please provided we can take q large enough , as Theorem 1
shows. If we want to have | oc - p/q| < 0.001 , we have only to
choose q > 1000 . In other words, we can make | <* - p/q| small
but we pay for it by having to use a large denominator q . This
suggests that we might try for a result in which is still
approximated by p/q but the denominator of the right member of
(l) is larger than q .

To get such a result, we shall introduce a completely obvious


but very important principle :

THE BOX PRINCIPLE. If


objects are placed in
n + 1 n boxes,
there is a box which contains at least two objects.

Even though this obvious, give a formal proof of it.


theorem is so
You probably feel that nothing of any importance could possi
bly come out of anything that sounds so trivial as the Box Princi
ple, but wait! Let n be any positive integer. Divide the
interval 0 ... 1 into n equal subintervals; these will be our
n "boxes" .

Now for each integer j in the range 1 < j n + 1 , let p .


3

be the greatest integer less than j oc , that is,


0 j<* - j 1, 2, , n
<
Pj < 1 , + . . . + 1

(Note that j <* - p. cannot be either 0 or 1 , for j <x - p

is Irrational . ) Consider the n + 1 numbers


48
<* - p. 2« - p, . . , (n + 1)<* -p.
n+1

they all lie between 0 and 1 and so are distributed among our n
boxes. Hence, according to the Box Principle, there must be two
of them, say r <x
Pr > s<* - p
^s
, which lie in the same box.

If this box is the subinterval ^"/n ( 1r + l)/n , we have

-< / + 1

.
r«x "Pr <
n

" 9 + 1
Ps< n
The second inequality may be written
* + 1
< -soc + P x
< - —
n ^s n
Adding the first and last inequalities, we get

(2) -1/n < (r - s) * -(pr«- ps) < 1/n .

Since r and s are both integers between 1 and n+1


r - is
but are not equal , that
s-r,
we see s between 1 and n ;
r-s
| |

1 ^ |r - s| <£
n . Set q = or whichever is posi-
tive; p =
pr - pr - ps if r - s > 0 , P =
Ps
-
Pr if
r - s < 0 . Then 1 <^
q <[
n , and (2) becomes -1/n < qc*
p 1/n
< , or |q <* - p| < 1/n .

Hence,
(3) _ £
q nq

This gives us the theorem:


Theorem. For each irrational number o< and each positive
integer n , there is a rational number p/q such that

<
q nq

with 1 <, q 1 n

Of course, (4) is considerably better than (l), simply


because 1/nq is considerably smaller than l/q when n is large,
We can obtain a more useful form of (4) by noting that, since
2
n > q , nq > q , so that

I
«**
-P/Q I < 1/nq < l/q2. or
(5) q
<
*9

So far we have existence of only one rational


shown the
approximation p/q with the property (5). Actually, there are
infinitely many such rational approximations, as we now show.
Choose an integer n' < n and find, by the last theorem, a
rational p'q' such that
I
<* - p'q' I < 1/n'q' , with 1 < q' £ n' .
Now <*> - p/q is not zero, so there must be an integer t for
which
(6) I
cC . p/q J > i/t
We shall increase t , if necessary, to make t > n (this only
strengthens the inequality (6)); then we can use t for the n'
above .

So now we have the following:


1 . 1 .1
q n ' n'q1 ^ rP" s n
This shows that p/q since the first
is not the same as p'/ci' ,
is further from o< than l/n , whereas the second is nearer to
than l/n . That is, p'/cl' is nearer to o< than p/q is.
Thus, p'/cl' is a new approximation to °< , and moreover,
1 1
n'q' S
<
q,2
so that (5) is satisfied. Now starting with p'/ci' we could
produce a still better approximation p''/q''
t° °^ which also
satisfies (5). We can continue this process indefinitely. This
proves the following theorem:
Theorem : For each irrational number there are infinitely
many different rational numbers P-jQ^ > i = 1> 2> 3, ... for which

(7)

Can we do better even than (7)? Is it true, for instance,


that there are infinitely many different p/q for which

I ok - p/q I < l/q ? There are infinitely many irrationals o< for
which the last inequality holds, but there are also infinitely
many for which it does not.
50
We shall give an example. Let <* =
\/2~". Suppose we wish
to approximate by a rational number p/q .

Case I. p/q < . We have y/2.q_ > p , and

2q2 _ p2 = (q V? - p) (qV?+p) >1,


2 - 2
because 2 q p is a positive integer and so is at least 1

Hence, remembering that p/q<\/2— and *\/2~ < 1.42 ,


we get

q V2~- P > s >


q \/2~ + p q \/2~ = q V2" 2 V2"Q
and since q > 0 ,

_ 2
1/2- q
=

Case II. p/q > 2 . Since < 1.42 ,


1
I
V2"- q| > °-58 >
¥^ ^q

since q > 1 .

Case III. V2"< p/q < 2 . Then p2 > 2q2 and p 2q2 =

(p-q V2~) (P + q V2~) > 1 >

or
1 1^1 4q
2 + V2~

There are no more cases, for V^- / p/q - (Why?)


In all cases, then, we have:

If p/q is a rational number, then

(7) VST-? >


4q<

Equation (7) shows that the approximation (6) cannot be essential


ly improved for all irrationals <* . We can express this by say
ing that the approximation of a general irrational by a rational
is measured by the square of the denominator of the rational.
generalize (7) to other irrationals than
Can you ? Can
you give an infinite set of irrationals for which (7) is true?
51

8.

A NEW FIELD

The you are most familiar with is the rational field,


field
but you have also studied the real field and the field of complex
numbers. If
you read Section 5, Gaussian Integers, you learned
about the domain of integers in a certain subfield of the complex
field. Here we shall study a new field which shows some differ
ences from the fields you have studied before.
Consider the subset K of the set of complex numbers consist
ing of all numbers of the form a + b (=a + ib > where
a and b are rational numbers. We define two elements of K ,

a + b \/"5 > to be equal if


they are equal as complex numbers,
i.e., ifonly a = c
and and b = d . ifIt is easy to check
that K is closed under addition, subtraction and multiplication.
(Do this.) K contains 0 = 0 + 0 - \/-5 and 1 = 1+ 0- \f^> .
The set consisting of all non-zero elements of K is closed under
division. (Check.) Moreover, all the rules of calculation are
satisfied in K since they are satisfied for complex numbers.
In short, K is a field; it is a subfield of the field of complex
numbers .

Call the field of rationals R . In R we singled out cer


tain elements which we called integers . Denote the set of integers
in R by I . It is a little hard to see how we can define
integers in K , but experience has shown that the following defin
ition is satisfactory.
First, notice that every element = a + b
\/^5 of K
satisfies a polynomial equation of degree 2 whose coefficients
are rational numbers. Indeed, write

Certainly P(x) = 0 coefficients of P are


when x = and the
rational numbers. Notice that the coefficient of the leading term
is 1 : such polynomials are called monic. When b = 0 so that
o(
= a, the equation becomes
(x-a) = 0 ; hence is a root of
52
the equation of lower degree,
P(x) = x - a = 0

The number o( satisfies


polynomial equations, but many

P(x) = 0 has the smallest possible degree. This is obvious when


b = 0 since P(x) is then of the first degree. When b 0 , /
P(x) = 0 must have the root of = a - ib it has the root «( if
(Chapter 5, Section 6), so P(x) =0 is at least of degree 2 .
However, there might be more than one monic equation of lowest
degree satisfied by (X . Obviously this cannot be if b = 0 .
If b ^ 0 , any quadratic equation Q(x) = 0 satisfied by o(
must have the factors x - ek and x - o( and no others. Hence,
Q is of the form c(x-p< ) (x - o( ) , where c is a real or com
plex number. But since Q is monic we must have c = 1 and so
Q is identical with P .

So we see that each element c{ of K satisfies a unique


equation P(x) = 0 which is either linear or quadratic. Of
special interest are those elements cK of K whose unique monic
equations have not only rational coefficients but rational inte
gral coefficients. (We now have to say "rational integers" to
denote integers in R because we are going to define integers
in K . )
Definition. An element of K is an integer in K if and
only if the unique monic equation which it satisfies has rational
integral coefficients. We write J for the set of integers in K.
Is an integer in R (rational integer) also an integer in
K ? What monic equation does it satisfy? This shows that I is
a subset of J , or as we write it, I J . Algebraic structures
like I and J which are closed under addition and multiplica
tion, which possess an additive identity (0) and a multiplica
tive identity (l) , and which satisfy the associative, commuta
tive, and distributive laws and the cancellation law (ab = ac
and a ^ 0 imply b = c), are called integral domains .

Let us consider an element = a + b


\/~^> of J . The
equation which o( satisfies is, as we have seen,
2 - 2ax 2 2
x + a + 5b = 0 .

Since c\ € J , we have that 2a and a 2 + 5b


2
belong to I ;
53
hence, - (2a)2 + 4 (a2 * 5b2) = 20b2 £ I . From this you will
be able to deduce that 2b £ I , if you remember that b is
rational. (Do this.) So we have that 2a and 2b are rational
integers; write a =
a1/2
, b =
b.j/2 , where
a1
,
b^
£ I .

Let a2 + 5b2 = c , c £ I . We have 4a2 + 5. 4b2 =

a, + 5b, = 4c . Now 4 c is divisible by 4 ; hence so is


2 2
a, + 5b, . But the square of any odd integer has a remainder 1

when divided by 4 . By trying out the four possible cases


(a, even or odd, b, even or odd), we see that a, 2 + 5b,
2
is
divisible by 4 only if a, and b, are both even. Therefore,

a £ I and b € I . The integers in K are the numbers of the


form a + b ]/^3 , where a and b are rational integers .

We can now do arithmetic in J just as we did in I . We

shall use Greek letters to denote elements of


(2>

. .
c\
0

>
,

>

divides there is a such that if


Q

We say
£
o{
J

J
. o

If cK divides

£>
divides
6

then

/
y=(3 and
^

c^

4
.
^


(Even though this is obvious, give a proof of it.)
Q

and
y
.

In we had two special integers and -1 which divide


I

all integers. We call such an element a unit:


unit is an a
integer which divides all integers. There are two units in .

I
What are the units in
J
?

Let A be unit in Then A divides every element of


J
a

and, in particular,
divides
X
J

.
1

Before going further, we introduce the very convenient notion


of norm: if the norm of (written No( is mere
&
c\

c\
K
,

ly the product of cK by its complex conjugate Writing


«?

= a + \f-5 we have =
(a+b (a-b y/^S) or
pi

N
°{
b

\/~=b)
,
a2

= + 5b2 .
N<^

In particular, If £ we see that ^ isrational


o(

N
J

a
,

integer which, also, is positive. There is no difficulty in check


ing that
(l) No{6 = NcA .
N<3

(Do this.)
54
Let us return to the matter of the units of J . If )\ is a
unit we have ^ »1 for some integer ^<f J , since we have
seen that X must divide 1 . Using (l) we get
n > n y - 1 .

This shows that N /s 1 = , since any norm of an integer in K


is positive. But, putting
A = a + b /^5 > we have

N X = a2 + 5b2 = 1 .

The only solutions of this equation in rational integers a, b

are a = - 1 , b = o .

What we have proved is that if A is


unit of a J , then
X = - 1 . But obviously - 1 are units of J .
Hence, our result: The units of J are - 1 .

We notice that N \ = 1 if A is a unit. The converse is


also true: if
NX = 1 > A. is a unit. You will have no diffi
culty in proving this.
We can now define prime : a prime in J is an element 6 of
J , not a unit, whose only divisors are - 1 and - }f . This, of
course, agrees with the definition of prime in I .
For example, the integer 3 is a prime in J . Consider a
factorization of =
§ , where 6 J, Then,

(?)
.3 : 3

J
e
o( <^ .

by (l), Nc*0 N0 = N3 = = .
4
N

.
9

Since is a positive rational integer, we have N<K = 9,


<K
N

3
,
or 1. If =9, nQ = and is a unit.
<^

6
2 N

1
,

If contradiction, because this


2

= a + 5b = we have
4
N

a
3
,

equation cannot be solved in rational . integers. (Prove this.)


If Not = oi is a unit. Therefore, = implies that
o<

^
1

3
,

either o( or is a unit, which shows that is a prime in J.


(?)

In the same way, prove that is a prime in


J

.
7

(These statements can be generalized. See if you can prove


that any prime rational integer of the form 4n - is a prime in
1

If you have read the Supplement Prime Numbers you will know
J

.
,

that there are infinitely many rational primes of the form 4n -


1
.

Conclude, therefore, that there are infinitely many primes in J.)


55
Besides certain of the rational integers, there are other
integers in K which are primes. For example, 4 + \J~^b and
4 - are primes. (Prove.) Other examples are 1+2 ,
2+3 \f^> , 6 +
\f-$ .

of primes is simply that they are


Of course, the importance
the multiplicative building blocks: every rational integer not a
unit is a product of primes. (See Chapter 9, Section 3.) Is the
same result true in J ? It is.
To prove it, let S be the set of non-units in J which do
not have factorizations into primes. If S is empty our result
is established, so we assume S is not empty. Let N be the set
of positive integers which are the norms of elements of S . Then
N is a non-empty set of integers > 1 (because S contains no
units); as such it has a least element a (Chapter 9, Section 3).
Every element of J whose norm is less than a (and > l) does
not belong to S .
Let o( be an element of S such that N o(
= a . Then ^
has no factorization into primes. If o( is prime, we have the
trivial factorization o( = ; hence, o( is not prime. It
follows that o( = Q 2f , where neither 9) nor $ is a unit or is
- o{ . Since N«( = N <3
- N ^ , so that N Q divides No^ , we

have 1 < < No( and also 1 < < N<*

This shows that 6 S , for


f
N$ < a . Hence, Q has a
factorization into primes. By the same reasoning, # has a
factorization into primes. Multiplying these two factorizations
together, we see that ©< has a factorization into primes. This
contradiction was obtained on the assumption that S was not
empty; hence S is empty and our result is proved.
In the rational field, factorization into primes is unique :
no matter how we factorize an integer we always get the same
primes, each occurring the same number of times. E.g., 60 = 30*2
=
15-2-2 = 5-3-22 , 60 - 6-10 = 2-3*2-5 = 22-3-5 . Only the order
in which the factors occur is different. But this is not true in
every field.
56
Consider 21 as an integer in K . We have
21 = .
(2) 3 7

21 - (4 + /=5) (4 - /=£)
As we have seen, the integers in the right members are primes in
J. Furthermore, they differ by more than just units, i.e., 3 is
not equal to any other factor times a unit. Here, then, we have
two essentially different factorizations of 21 in J . Factoriza
tion into primes in J is not unique.
The central theorem used in the proof of unique factorization
in the rational field is the following: if a prime p divides a
product ab , then p divides either a or b (or both). This
theorem, however, is false in J . For from (2) we deduce that
4 +
\/-5 divides 3,7 (since it divides 21), but it does not
divide either 3 or 7 . If we assume, e.g., that
(4 +
V^X - 7 ,
we get, taking norms,
.N* = N7 = 49 , 21
so that No( is not a rational integer as it has to be.
Unique factorization can be restored to K by introducing
certain new elements called: ideals. Every non-unit integer in K
is a unique product of prime ideals. You will learn this beauti
ful theory if you continue your mathematical studies in college.
2S

shsmsnv oi SNOIisano

o.fQ.amttt!«*V suofq.oun,a

I
aaqiuriN jo
aaSaq.ui saosTAfa saosfAfa
i T I
z fT ' s z
e rT ' e s
+7 rT 's If e

S rT ' s s
9 rT ' 'S 'e 9 +7

I 'T ' 2 s
8 'T ' 's '* 8 +7

6 'T ' 'e 6 G

01 'T ' 'S '5 OT *7

it 'T ' TT s
SI 'T ' 's 'e '+7 '9 ST 9

ei 'I ' CT s
-trT 'T ' 's '2 irT +7

ST 'T ' 'e 'S ST +7

9T 'T ' 'S 'It '8 9T 5


Ll 'T * 2l 5

81 rT ' 'S 'e '9 '6 8T 9

6T 'T* 6T S

'S '5 'OT


'It

OS T OS
9

'

'T 'e '2 T2


^7

TS
'

'T 's 'TT 33


+7

S2
'

es rT es
S
'

*2 's '9 'ST


'9
'it

rT 'E
8

+rS
'

S3 'T 'S S3
E
'

'T 'S 'CT 93


+7

93
'

'e '6 IZ
^7

22 rT
'

82 fT 'S 'Ml
%

85
9
L
'
'

6s rT 6S
S
'

'S '£ '5


'9

oe fT 'OT 'ST OG
8
'
58

1. No. All other integers will at least be divisible by 1

and the integer itself. Hence the number of divisors will be


greater than or equal to 2 .
2, 3, 5, 7, 11, 13, 17, 19, 23, 29.
Yes. They are all perfect squares.
4, 9, 25 . Yes. 16.
Every number with exactly three divisors is a perfect square.
25 numbers have an even number of divisors. 1, 4, 9, 16, 25
do not.
42. 90.
The 14 numbers 2, 3, 4, 5, 7, 9, 11, 13, l6, 17, 19, 23,
25, 29. They are all primes or powers of a prime. Guess: It
must be a power of a prime. 4. 6. 4. 6. n+1. n+1. The
number of divisors of an integer n is a prime and only if if
n = qP~ where both p and q are primes.
Number of times n
appears as a number
n of divisors
1 1

2 10
3 3

4 9

5 1
6 4

7 0
8 2
The number of divisors of the integer n is odd if and only
if n is a square .

24 and 30 have 8 divisors. Yes. 36 has 9 divisors.


Yes. 48 has 10 divisors. No.
60, 72, and 96 all have 12 divisors.
nu m
... pr , r(n) =
riL,
For n =
Pl p2 (m1
+ l) .

(mg
+ 1) ... (mr + 1) .

12, 18, 20, 28, 32, 44, 45, 52, 63, 68, 75, 76, 92, 99 .

144 .
59
k
x - (x - 1)
/ \
3. 7. None. At most one. Since 1 =

(xk"i +1) ,
if x > 2 the number is not a prime.
60

Arithmetic Functions II
The first few perfect numbers are 6; 28; 496; 8128;
33,550,336. The first four were known by 100 A.D. Until only
1870
four more had been found. Between 1870 and 1950 four additional
ones were found.
n Divisors of n dT(n)
1 1 1 D

2 1, 2 3 D

3 1, 3 4 D

4 1, 2, 4 7 D

5 1, 5 6 D

6 1, 2, 3, 6 12 P

7 1, 7 8 D
'
8 1, 2, 4, 8 15 D

9 1, 3, 9 13 D

10 1, 2, 5, 10 18 D

11 1, 11 12 D

12 1, 2, 3, 4, 6, 12 28 A
13 1, 13 14 D

14 1, 2, 7, 14 24 D

15 1, 3, 5, 15 24 D

16 1, 2, 4, 8, 16 31 D

17 1, 17 18 D

18 1, 2, 3, 6, 9, 18 39 A

19 1, 19 20 D

20 1, 2, 4, 5, 10, 20 42 A
21 1, 3, 1, 21 32 D

22 1, 2, 11, 22 36 D

23 1, 23 24 D

24 1, 2, 3, 4, 6, 8, 12, 24 60 A
25 1, 5, 25 31 D

26 1, 2, 13, 26 42 D

27 1, 3, 9, 27 40 D
28 1, 2, 4, 7, 14, 28 56 P
29 1, 29 30 D

30 1, 2, 3, 5, 6, 10, 15, 30 72 A
6l
23 are deficient, 5 are abundant, and 2 are perfect.
If p is a prime ^(n) = p + 1 .

2 k
1* P* P > - . . >P

pk+1 - 1
P - 1

i, p, p2>--->pk> q> pq> p2q>--->pkq - 2(k + 1) .

+ p + p2 = ... + pk) + q(l + p + p2 + ... + pk)


=
(1 + q)(l + p + ... + pk)

i> p> p
2
,- ->i>
k
> q> pq> p
2
q>..->P k2222
q> q > pq > p q

pkq2 -
3(k + 1) .

.
q-l
Jc+1 „3
p_i
n ., ^k+1 _
+ <1 + <l ; - *
p 1

Pk+1
p -
-
1
1 . qS+1
q-l
- 1

<<(n) =
m1+l
£l
Px
- 1
. £2
m^+l

P2
li ...
- 1
P
mr+l

Pr
-
-
1
1

- 1 3" - 1
6 == 2-3 ; <f{6) = .
- 3-4 = 12 .
22~ - 1 2 1

23 - 32 -
12 = 22-3 ; (f (12) =
2 -
1
1
*
3 -
= 7.4 = 28 .

- -
18 = 2-32 ; d (18) =
22
2 -
1
1
. 33
3 - ■j
= 3-13 = 39 -

2* - 1 32 -
24 = 23-3 ; <^(24) =
- - -y = 15-4 = 60 .

f \
2 1 3

23 - 1 .
-
28 = 22-7 ;
(J (28) =
- = 7-8 = 56 .
2 1 7 -

- -
30 = 2.3.5 ; d (30) =
22
2 -
1
1
*
32
3 - 72

-
144 = 24-32 ; (f[144) =
25
2 -
1
1
. 33
3 -5-3
= 31-13 = 403 .
62

Arithmetic Functions III


One of consecutive integers must be even. Therefore
any two
aside from the pair 2 and 3 , any other pair of consecutive
integers must contain an even integer greater than 2, which is
composite.
Exercise 1.
The primes less than 100 are 2, 3, 5, 7, 11, 13, 17, 19,
23, 29, 31, 37, 4l, 43, 47, 53, 59, 6l, 67, 71, 73, 79, 83, 89, 97.
Exercise 2.
The twin primes less than 100 are 3, 5; 5, 7; 11, 13;
17, 19i 29, 31; 4l, 43; 59, 6l; 71, 73.
Exercise 3.
tt(10) = 4 , tt(20) = 8 , ir(30) = 10 , ir(4o) = 12 , tt(50) = 15 ,
tt(75) = 21 , ir(l00) = 25 .

Exercise 4.
Proof: Suppose n is not prime; then n = pq where p is a
prime 1 < p < n . By hypothesis p > \/n . But then q < j/n
(otherwise n pq > \/n j/n n) .= 1 , . = Therefore q must =

since if
q ^ 1 it has a prime divisor which is < j/n
Therefore q must = 1 , since q ^ 1 it has a prime divisor if
which is < \/n and which divides n , contrary to hypothesis.
If q must be 1 , then n is prime. q.e.d.
Exercise 5.
-
13.137 ; 4079 is prime.
1781
Exercise 6.
The primes greater than 100 and less than 225 are 101,
103, 107, 109, 113, 127, 131, 137, 139, 149, 151, 157, 163, 167,
173, 179, 181, 191, 193, 197, 199, 211, 223 .

Exercise 7.
=
tt(150) 35 ; tt(225) = 48 .

Exercise 8.
7r(200) = 46 and tt(100) = 25 . The answer is 21 .

Exercise 9.
ir(n) _" 664,580 ~
664,580 " 664,580 ~ 1.07" .
n M 4342945 b23,27«
log n 10,000,000 7 7
63
The Euclidean Algorithm and Linear Diophantine Equations

63x + 7 = 23y

63X + 7»Z*y

6 10 16 x

No.
No. 3x +3(x + 2y) = 13 . Three does not divide
6y =

thirteen, so that there are no integers x and y for which this


equation is satisfied.
3x + 6y = 24 . Solutions: (4, 2); (6, l) ; (2, 3) . Yes.
Three solutions. (2, 3). (6, l).
2x-y=6. (4, 2). Yes there are infinitely many,
x = 4 + t . y=2+2t is
solution for any integer t . Any a
non-negative t gives a positive solution.
Yes. If the slope, -a/b is positive and there is a solution
at all, then there are infinitely many positive solutions. If the
slope is negative and there is a positive solution, then there are
only finitely many. An equation may have solutions, and yet if
the slope is negative it may have no positive solutions. Then of
course there may be no solutions in integers at all .
(l)> (5)> and (6). Yes. Yes. If ax + by = c has a solution
in integers, then (a, b) = d divides c .
Proof: Let a = da' ; b = db' . Then da'x + db'y = c .

Hence if there is a solution d divides c .

Yes. For (l) . (2, 3) = 1, 1 divides 5 and (l, l) is a


solution.
For (5) . (4, 6) 2, 2 divides 8 and (2, 0) is a
solution .
64

For (6). (2, 4) = 2 , 2 divides 4 and (2, 0) is a


solution.
Factor each number. Take all prime factors common to both
numbers and raise each to the smallest exponent to which it appears
in either number.
proof which follows.
See
Theorem: If d divides a and d divides b , then d
divides a + b and a - b .
Proof: Let a = da' and b = db' . Then a + b = da' + db'
=
d(a« + b') (Distributive Law). Therefore d divides a + b .

a - b = da' - db« =
d(a» - b') (Distributive Law).
Therefore d divides a - b .

253 = 11.23 ; 122= 2»6l . (1596, 96) = 12 . (4l8, 1376) = 2 .

(365, 146) = 73 .

Yes. Given any positive integers a and b with say, a


greater than b , then there always exists integers q, and r,
such that a = bq, + r, with 0 £ r., < b . Similarly
b =
r.,qp + ro with 0 ^
r2 < r. . Continuing in this way we have

a decreasing sequence of positive integers. There are only b - 1


positive integers less than b . So that after at most b - 1
steps the remainder must be zero. If a and b aren't positive
integers, we can still find their greatest common divisor by using
the algorithm on |a| and |b| , which are positive.
If x and y is a solution then x + b and y - a is
also a solution.
General Solution: Suppose x and y satisfy the equation
and suppose x and y are any other solution. Then ax + by - c

and ax + by = c .

If we subtract we get a(x -x)+b(y-y)=0.


a(x - xQ) = -b(y - yQ) .

Divide by d

4(x
dv~
-
~
Ao'
xj =
"-" My
~d
- y)
65
Since a/d and b/d have no common factors, x - x must be

divisible by
* 4 ; let x - x =
^ t . Then substituting we hav«
d o d

t'T* = "
d
^y "
yo^ ' and y " yo=
"d* ' Consequently

X =
X^
o
+ -T
d
t ,

y -
|t ; is a solution for every
integer
CHECK:

a(x0
+
f t) +
b(yQ
-
| t) = c .

^o + byo + t ^S * ^o + byo c '


* " = =

Yes, it is clear that any solution must have this form since
x and y were assumed to be any solution of the equation and it
followed that they had this form for some t .
5. Yes. x = 5 - 23t, y = 14 - 63t .
Answers to Problems :

1. x = 3 + 7t , y = 79 - l6t .

2. x = 170 , y = 110 ; x = 923 , y = 9 .

(x = 923 + 753t , y = 9 - 101t).


3. 5 and 6 . (5 + 15t , 6 + 17t).

4. 4 . (9 - 795 , 4 - 37t).
5. x = 4 +45t , y = 1 + l4t .

6. x = 27 + 63t , y = 15 + 40t .
66

Gaussian Integers
Exercise 1.
Let the numbers be a = 2n + 1 and b = 2m + 1 .

Then a2 + b2 =
(2n + l)2 + (2m + l)2 = 4(n2 + m2 + n + m) + 2
2 2
Therefore for a and b odd, a + b leaves a remainder 2
when divided by 4 and consequently is never a multiple of 4 .

Exercise 2.
Let the Gaussian integers be a + bi and c + di ; a, b, c, d
rational integers, (l) (a + bi)
di) = (a + c) + (b + d)i . + (c +

Since a + c and b + d are rational integers the sum and


difference are Gaussian integers.
(2) (a + bi)(c + di) = (ac - bd) + (ad + bc)i .
Again since ac - bd and ad + be are rational integers the
product is a Gaussian integer.
Exercise 3.
p
No. is not a rational integer.
Exercise 4.
p
No. 2 and 3 are Gaussian integers and -j is not a Gaussian
integer.
Exercise 5-
The quotient is iri which is integer.
y||

No + not a Gaussian

Yes. + 3i = i(3 - 2i)


2

Exercise 6.
Yes. + Hi = + 3i)(3 +
(2

i)

.
3

Yes. + Hi =
-i(-ll + 3i) .
3

Proof of Lemma:
6
N(

N(

o{fi) = )n(
0(

bi,@=c
)

Let = a + di.
c{

Then,5((3 =
(ac - bd) + (ad + bc)i .

n(^8)
= (ac - bd)2 + (ad + bc)2
ac =bd .2 ,2 + ad .2
2.2 + bc
2
2
2

=
,

=
(a2 + b2)(c2 + d2)
=
N(o^ )N(§>)
.
67
Exercise J.
If A and & are associates, then cA =
ft *u, where u is a
unit.
Then by the Lemma N( ^ ) - N( ^ )N(U)
=
N( 3 ).1 since n(u) = 1 by
Theorem 1.
- N( <3 ) . Q.e.d.
Exercise 8.
No. 5 =
(2 + i)(2 - i) . N(2 + i) =
N(2 - i) - 5 .

Therefore 2 + i and 2 - i are not units. N(5) =25 and


they are not associates of 5 by Exercise J. Therefore 5 is
not a prime, since it has divisors which are neither units nor
associates of 5 .

Exercise 9*
Yes. For suppose 3 = $(y .

Then N(3) =
N(4 )N( & ) = 9 .

Then N(^ ) = 1, 3, or 9. If N^ ) =1, is a unit.


If N( # ) = 9 , then N( £ ) = 1 ^ is a unit.
and
Hence N( g>L ) must be 3 if 3 is not to be a prime.
But N( o( )
= a + b = 3 is impossible for rational
integers a and b . Therefore 3 has no divisors except units
and associates of 3 and is therefore a Gaussian prime.
Exercise 10.
We consider all possible cases for rational integers x and y.
Case I: x and y both even; let x = 2x' , y = 2y' .

x +y =4(x! +y')^4n+3 for any rational integer


n .

Case II: x and y both odd; let x=2x'+l, y = 2y' + 1 .

2 2
l)2 l)2
,
x + y =
(2x» + + (2y« +

=
4(x'2 + y«2 + x' + y») + 2 £ 4n + 3 for any
rational integer n .

Case III: one even and one odd; say x=2x' , y = 2y' +1
x2 = y2 =
4(x'2 + y'2+ y«) + 1 /S 4n + 3 for any
2 2
rational integer n . Therefore x + y ^ 4n + 3 for any inte
gers x and y . q.e.d.
68

Exercise 11.
Yes Suppose 1 + i = o{ V
N(l + i) =
N(<* M<3 )

2 =
we )
But since N( o{ ) and N( 65 ) are rational integers, one of
them is 2 other is 1 . Suppose
and the N( o( ) =1 ; then
is a unit. Therefore 1 + i is a prime since it can only be
written as a unit times an associate of 1 + i .
Exercise 12.
Yes. N(l - i) = 2 and we can repeat the same argument given
in Exercise 11.
Exercise 13.
No. Since every rational integer is a Gaussian integer, a
composite rational integer a = divisors the Gaussian
be has as
integers b and c which are not units or associates of a .
Exercise 14.

(a) 5 = 22 + l2 (d) 29 = 22 + 52

(b) 13 = 22 + 32 (e) 101 = 102 + l2


(c) 17 = 42 + l2 (f) 1721 + ll2 + 402
69

Fermat's Method of Infinite Descent

Exercise 1.
37 - 62 + l2 ; 41 52 + 42 ; 89 = 52 + 82 ; 101 - 102 + l2
Exercise 2:
2 2
Given: d divides x , d divides z , and x + y
Show: d divides y .

Let x = dx' , z = dz ' .

d2x'2 + y2 = d2z'2 ,

y2 =
d2(z'2 - x'2)
2 2
Therefore, d divides y and d divides y
Exercise 3.
If a number is odd its square is odd. If x , y , and z
2 2
are all odd then x + y is even; but x2 + y
2 = z
2
and
.2
z"~ is
odd. This contradiction shows that not all three numbers can be
odd.
Exercise 4.
If a number is even, its square is even; if a number is odd,
its square is odd. Consequently, the sum or difference of the
squares of two even numbers is an even number and it is impossible,

therefore for z*"


2
to be odd. But if 2
is even, then so is z
Exercise 5-
Given: z and y have no common factors; z + y = 2v ;
z - y = 2w2 .

Show that v and w have no common factors.


Suppose v and w have the common factor d ^ 1 .

2
Then adding z + y = 2v

z - y = 2w

2'z =
2 •
,

z = -2
v w2
2(vc: + w^) +

Subtracting 22 - -
2y =
2(v*"_ w2)
w ) ,. y = v2
v w

If d divides v and d divides w , then d divides


p
v2 + w = z and d divides v
2 - w
2
= y , contrary to the hypoth-
esis. Therefore v and w have no common factors .
70

Approximation of Irrationals by Rationals


1. Assume that no box contains more than one object. Then the
total number of objects is not more than n . This contra
dicts the fact that n + 1 objects were placed in the boxes.
2. Theorem: If m is a positive integer which is not a perfect
square, there is a constant c > 0 depending on m such that

V/m-§q' cq"

no matter what the rational number p/q may be.


Proof: Let r be the integer such that r - 1 < (/m < r .

Note that r > 2 .

Case I. p/q > /m


2 - 2 - p)(q \/m
Then m q p = (q \/m + p) > 1 ,

V/m +
£ 2 (2r+l)q'
l/ra

Case II. l/m < p/q < r + 1 .

- _ 1_
]/m
>7
>
q
l/m + r + r + 1 (2r + 1) q£

Case III. p/q > r + 1 .

> 1 >
(2r + 1) q£

since \/m < r while p/q > r + 1 .

For c , we can take c = 2r + 1 .


71

A New Field
1. (a + b \f^>) + (c + d \f^>)
=
(a + c) + (b + d) |/=5
= +
a2 b1

(a + b \f^>)
- (c + d =
(a - c) + (b - d) /=5
l/=5)
= +
a2 b2

(a + b y/I5)(c + d = (ad - 5cd) + (be + ad) |/^5


- a3 * b3 V=5 .

Since a, b, c, d are rational, so are a^, b^, a2, b2,


.
a3, b3
2. Since

a
c +
+ b
d
J^5
|/^5
=
ac + 5bd
c2 +
+

5d2
(bc«- ad) t/=5 = a
1
+ b
1 ^
when c + d y^5 / 0 (i.e.,
not both c and d are 0),
we have , are rational since a, b, c, d are.
a^ b^

3. An integer integer in K , since


o( in R is also an

o( satisfies the monic equation of smallest possible degree


x - o( = 0 . This equation has coefficients in R since
<^ is in R .
4. Write b = £ > where p, q are integers with no common

P P P
factors (except l). We have 20b = 5-4 p /q is a rational
p
integer. So q must divide 20 since it ha3 no factors
2 2
which divide p . q cannot divide 5 because 5 has no
P
factors which are squares. Hence, q must divide 4 ,
p
i.e., q = 4 . Then q = 2 , and 2b = 2p/q = p is an
integer, as claimed.
5. Let o( divide G and $ . Then 6= c\ $f ^
, <?
=
c{ $~2
.

So S ^= d
<TL

+ +
<t(f2=4(<f^ .
+

^2")
Hence, d divides Similarly,^ divides -
^
(3
+
(3

.
72
6. Use the theorem (Chapter Section 5) that the conjugate of 5>

a product of complex numbers is the product of the conjugates :

n(* )-(*/*)( 56 ) - )(<?<?) =


(*V)(<3£) =
-

N N <3 .

7. If nA = 1 , /\ = a + b v/^5 , we have a2 + 5b2 = 1 , the


only solutions of which in rational integers are a = ± 1 ,
b = 0 . Thus A = ± 1 , so A is a unit .

2 2
8. Since a, b are rational integers, we have a + 5b >

5b2 > 3 if £ 0 . b Therefore, b = 0 . Hence, a2 = 3 ,

which is impossible.
9. Let rational prime of the form
p be a 4n - 1 , and con
sider the factorization p =
^ <3 > where (^e.J". Taking o{ ,

norms we get Since No/ is a rational


N^.
,p,orl.
p2 = -
p
integer, we have either N o( «= p In the first
case we have N 0 = 1 , so 0 is a unit; in the last case,
q is a unit. Consider = p and let c>^
= a + b .

This gives a
2
+ 5b
2
= p . If a?b
are both even, the left
member is even. If a2b
are both odd, the left member is
even, since a
2
and 5b
2
are both odd. If a is even, b
odd, or a if is odd, b even, the left member is of the
form 4n + 1 . Hence, it is impossible that n <X = p .

Thus the factorization p = o{ ? is possible only if o< or


is a unit in J .

10. If 4 +
^5 = o( Q , we have 21 = N
o(
. n <3 . Now
N °< / 3 or 7 , for as we just saw, ^ Np = is impossible
when p is of the form 4n - 1 . Therefore N = 1 or
N 6 = 1 , so that either o( or fl are units . Same proof
for 4 - \f^> .
73
1, 2, 4, 8, 9, l6, 18, 25 . They are either a square or
twice a square. If () (n) is odd, then n is a square or twice
a square.
Proof:
111,+ 1 nu+1
p, - 1 - 1
mr+l - 1
— pg Pr
Given: (f (n) =
_ 1
.
Pg
. 1
. . .
pr - i odd.

Then each factor of the product must be odd.


If p. = 2 , then m, may be any integer > 0 .

If p. is odd, all the powers of p. will be odd. Since


2 mi
1 + p. + p? + . . . + p. must be odd, then m.+l must be odd;

i.e., m. is even. Let m. = 2t. .

m, 2t« 2t3 2t
Then n has the form n = 2 p2 p3 . . . p and we

may write
m, -1

if
t3

*
"2"

J"2. tr%2 is odd


(2

..Pr
2

.
pg P3 n^
)

n = m,-l
tg

tg

if
t

75
g

m, is even.
(2

. .
p2 p3
.

q. e.d.

In the second case n is clearly a square. In the first case n


is twice square.
a

If n =
2m"1(2m-l) and 2m-l is a prime, then n is a
perfect number.

Proof:
We need only show that (n) = 2n The prime divisors of
U

n are and -1 We can make this statement only


2

.
2

because we are given that 2m-l is a prime.

- -
(T(n)=fH (^J2
1

Then
2-1
(2m_l) _
x

((2m-l) +1) .
((2m-l) - 1).
ra .
,

-1)
(2

{(2m _1} _ 1}
74

(2m-l) (2m-l+l) = 2m(2m-l) = 2n .

If m = 6 , 2m-l is not a prime.

If m - 7 , 2m-l is a prime.

If m - 8 , 2m-l is not a prime.

If m = 9 , 2m-l is not a prime.

If m = 10 , 2m-l is not a prime.

If m = 11 , 2m-l is not a prime.

If m = 12 , 2m-l is not a prime.

If m = 13 , 2m-l is a prime.

If m is not prime, then 2m-l is not prime either.

Proof: Since m is not prime, let m = nu'iiu , where

m, and nu are greater than 1 .

_ mnnu
1
m, iru m, m, nu-1
^
Then 2m-l = 2 ^-1 =
(2 X) ^-1 =
(2 1-l) {(2 1)
m, nu-2 m,
+ (2 X)
d
+ ... + 2
.1
+ 1) .

Since 2
.1
> 4 , 2 -1 ^ 3 and 2-1 is not a prime, q.e.d,

Every even perfect number has the form 2 (a -l) where

2-1 is a prime.
Proof:
V (m.n) = £>(m) . O (n) . To prove this, one
(Lemma:
only needs to write out the expressions for D (m*n) , C^(m) >

and (f(n) and verify that 6^(mn) =


(X{m) .
(T\n).)
Let n = 2mq , where q is odd. Since n is perfect
(1) 0T(n) = 2(n) = 2m+1q .
But by the lemma, £T(n) =
(f{2m) 0"(q) . Substituting in
(1) we have

(2) ^(2m) (T(q) = 2m+1q


75

Now (f(2m) = 2m+1-l . Substituting in (2) we have

(3) (2m+1-l) (J-(q) = 2m+1q .

From (3) we see that 2 -1 divides q . Suppose we set


q = (2m+1-l)q." .

On the right hand side of (3) replace q by (2m+1-l)q'' ,

and dividing both sides by (2m+1-l) we have

(^) (j(o.) = 2m q . But q and q/ are divisors of q and

q + q = 0.
,2m = CT(q) . Hence these are the only

divisors of q and q must be a prime and q must be 1 .

Therefore q = 2m+ -1 and q is a prime. But then

n =
2m(2m+1-l) q.e.d.

Let the divisors of n be d, , cU , . . . , d..


Then i
1
d
+
1
±
d
+ ...+£d 1
=
d'
-^
+
6—
d«2
+

n
... + d»

= 2
n

Since n is perfect. (It should be shown that all d. are

distinct and actually include all divisors of n and each


only once . )

Another way of stating this result is:


G< (n) <(")

In fact a more general result holds: 0 (n) =


^
.» n

kCn^+l) x k(mg+l)
' 1 k(nyt.l) 1
- PX P2 Pr
<<»>
- - - 1
Pi 1
P2
1
Pr
V
S.M. 1>M

T/+& 0 ^Y jxr
J-4~
Mathematics

3A
X'H'i

,5 57
V,1

DATE DUE
AUG 8 9SJ
OCT 1 0 1 )62
OCT 2 4 1 )B2

m i j

Mn nf >rnu j
V