http://hdl.handle.net/2027/mdp.39015017386932
^1
'1H°D'uoiipois ^^
rr
Mama Aaaadi
SCHOOL
MATHEMATICS
STUDY GROUP
v. \
Financial support for the School Mathematics Study Group has been provided by the
* * »
Page
1. Prime Numbers 1
2. Congruence 9
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
1. PRIME NUMBERS
size, we get
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
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 )
Euclid's Elements.
have a contradiction since A is not one of the primes p.,, p2, .**,
pn (Why?), and these were supposed to be all the primes.
our proof shows that there is a prime p which is not either 2,3
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
(Prove this in the general case by assuming that all prime divisors
(*) A =
MPiPg... Pn)1.
Notice that A is of the form 4k  1.
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.
posite integers.
For example, there are 50 consecutive integers all of which
are composite. We can actually exhibit such a consequence.
number, 51! +
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
2
1
there any other prime triples? See if you can prove it one way or
the other.
2. CONGRUENCES
We write
a a b (mod m)
a = a (mod m)
a + c = b + d (mod m)
(l) a  c = b  d (mod m)
ac = bd (mod m)
For (a+c)  (b+d) = (ab) + (cd) and each term in the right member
Namely, we have:
as promised.
N =
aD
+
10a1
+
102a2 +... +
10kak.
(The digits of N are therefore a0, a^ ak. )
(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.
a(bc) = 0 (mod m) .
It does not follow that ma or m(bc). For some of the prime
factors of m might divide a and the remaining prime factors
(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.
system.
plete residue system and so one of its members ax, say, is con
10 (mod
7
5
The preceding discussion can be applied to the solution in in
tegers of equations like
change for a dollar using only nickels and dimes? The equation is
4x  35 = 13y and
4x = 35 (mod 13).
Since is as we saw in Example
"4"
t
1
x = 13t  1.
(la) a = qb + r,
where
(lb) 0 ^ r < b.
and remainder and that the remainder is less than the divisor. We
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
(1) d  a, d  b,
(2) if dx
a, d^ b, then 61  d.
in the above definition.) Suppose there were two gcd's d^ and d2.
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
shall assume a, b are both positive; the remaining cases are easily
(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 36. Then from (3.2): 3 13 and 3 5.6 imply
3 1 33 (see Exercise 1). Prom (3.1): 36 and 3 2. 33 imply 372. So
divisor. Then from (3.1): d72 and d2*33; hence, d6, since
3 = 33  5 (72  2*33)
or
3   5.72 + 11.33.
(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, . . .
(4.(i»l)) rn_2
=
qnrn^1 + rn, 0 £ rn < r^
(4.(n+2)) and so
rnrn2
by (4.(n+l)).
Working upwards we conclude that r n a, rnJ^* ^ d is a common
(4.(n+l)) we have rn
=
rn_2
 qnrn„i. Both rn_1 and
rn_2 can
ax + by = d.
ax + by = 1.
factorized into primes, but remember that we have not yet discussed
have, by Theorem 1,
ax + by  1
acx + bey = c
As a corollary we get
THEOREM 3. If the prime pbc and pfb, then pc. (pfb means
"p does not divide b" . ) For if pfb, p must be prime to b, since,
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
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
(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 . )
P2Pg... Pr  q2q3*.. qs »
and needs no proof. However, there are many number systems besides
ANSWERS TO QUESTIONS
PRIME NUMBERS
A  PjPg
. . .
Pn
+ 1 shows that A > p1p2 ... Pn > PA»
since
each prime > 1.
(4kx + l)(4kg + 1) = 16
k^ +
4kx +
4k2 + 1 .
4(4^1^ +
kx + kg) + 1.
form hk  1.
a contradition.
CONGRUENCES
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)
2$  2X 25  2(l+ 3t) _
* m m m
3 3
dj, d2
are integers. Therefore
b + c =
adjL + adp ■ a(d, + dp), so a(b c).
+

r.j61
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
QA
.638
V.2
Financial support for the School Mathematics Study Group has been provided by the
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, McGrawHill, New York, 1948.
ii
1.
ARITHMETIC FUNCTIONS
I
THE NUMBER OF DIVISORS OF AN INTEGER
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, , b1 .
3.122 + 1.12 + 2
or 446 in the ten.
base
The symbol 4et21 would represent
3
Divisors
1, 2
1, 3
Number
2
of
Divisors
1
4 1, 2, 4 3
5 1, 5 2
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?
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
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 .
What are the divisors of p ? Can you find the sum of 0"{p )?
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
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,
m,
If nip m~ m
n =
p^ p2 pg .. prr , we found that f(n) =
(m1+l)
.
(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
(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
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
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
® ^
If]

HI •
<
multiple of . If
we are counting the numbers taken out by our
3
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
■K]n][y +
fefe]+[*,]
+
+[sfe] .
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
k
.
tr(n) = + 
k
M
1
.
piHF])+ [M +[M)
*
«*>».(B*
' "
(hm
[2.3.5J
.1
3
17
1R25)
= 25  (12 + 8 + 5) + (4 + 2 + 1)  0 + 3  1
= 9 
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.
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
(1) 2x + 3y =
5 (4) 4x + 6y = 9
(2) 2x + 4y = 5 (5) 4x + 6y = 8
(3) 3x + 3y = 5 (6) 2x  4y = 4 .
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 nonzero 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 + .
,
.
,
.
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 nonzero 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 .
(Suppose a > b . )
a = b,q, +
r^
b = +
rl'Q2 r2
rl =
V2% +
r3
r2
=
r3.q4 + r.4
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 .
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 .
(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 .
5.
a =
£ . Suppose that £ has been reduced to that p and q are
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
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 nonnegative.
If b = integer a is a . These
0, the nor.n of the rational
o
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
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 .
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
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
(17071783).
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 .
Accordingly we will
consider only solutions which have no
common factors. In this situation not all three integers x , y ,
(3) 4u2 + y2 = z2 or
u
4u
2 = z
2  y2
(*) ^2 =
(z + y)(z  y) .
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
>
(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 .
(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 ) ',
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
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 .
7.
APPROXIMATIONS OP IRRATIONALS BY RATIONALS
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 .
1 H
% 3
m +
q' m +
q '
   ' m +
^q^
 ^e P°^nt wnich represents <X
result .
Theorem : If is irrational, then to every integer q > 1
333/106 , 355/113 .
47
.
<*  p/q .
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.
< / + 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
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
<
q nq
with 1 <, q 1 n
I
«**
P/Q I < 1/nq < l/q2. or
(5) q
<
*9
(7)
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 .
_ 2
1/2 q
=
since q > 1 .
Case III. V2"< p/q < 2 . Then p2 > 2q2 and p 2q2 =
or
1 1^1 4q
2 + V2~
8.
A NEW FIELD
. .
c\
0
>
,
>
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
.
I
What are the units in
J
?
and, in particular,
divides
X
J
.
1
c\
K
,
= a + \f5 we have =
(a+b (ab y/^S) or
pi
N
°{
b
\/~=b)
,
a2
= + 5b2 .
N<^
N
J
a
,
(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 .
N X = a2 + 5b2 = 1 .
are a =  1 , b = o .
(?)
.3 : 3
J
e
o( <^ .
by (l), Nc*0 N0 = N3 = = .
4
N
.
9
3
,
or 1. If =9, nQ = and is a unit.
<^
6
2 N
1
,
= a + 5b = we have
4
N
a
3
,
^
1
3
,
.
7
If you have read the Supplement Prime Numbers you will know
J
.
,
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 nonunit 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
it 'T ' TT s
SI 'T ' 's 'e '+7 '9 ST 9
ei 'I ' CT s
trT 'T ' 's '2 irT +7
6T 'T* 6T S
OS T OS
9
'
TS
'
S2
'
es rT es
S
'
rT 'E
8
+rS
'
S3 'T 'S S3
E
'
93
'
'e '6 IZ
^7
22 rT
'
82 fT 'S 'Ml
%
85
9
L
'
'
6s rT 6S
S
'
oe fT 'OT 'ST OG
8
'
58
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 .
(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> p
2
, >i>
k
> q> pq> p
2
q>..>P k2222
q> q > pq > p q
pkq2 
3(k + 1) .
.
ql
Jc+1 „3
p_i
n ., ^k+1 _
+ <1 + <l ;  *
p 1
Pk+1
p 

1
1 . qS+1
ql
 1
<<(n) =
m1+l
£l
Px
 1
. £2
m^+l
P2
li ...
 1
P
mr+l
Pr


1
1
 1 3"  1
6 == 23 ; <f{6) = .
 34 = 12 .
22~  1 2 1
23  32 
12 = 223 ; (f (12) =
2 
1
1
*
3 
= 7.4 = 28 .
 
18 = 232 ; d (18) =
22
2 
1
1
. 33
3  ■j
= 313 = 39 
2*  1 32 
24 = 233 ; <^(24) =
  y = 154 = 60 .
f \
2 1 3
23  1 .

28 = 227 ;
(J (28) =
 = 78 = 56 .
2 1 7 
 
30 = 2.3.5 ; d (30) =
22
2 
1
1
*
32
3  72

144 = 2432 ; (f[144) =
25
2 
1
1
. 33
3 53
= 3113 = 403 .
62
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 =
a  b = da'  db« =
d(a»  b') (Distributive Law).
Therefore d divides a  b .
(365, 146) = 73 .
and ax + by = c .
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 .
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 .
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 +
No + not a Gaussian
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 .
Exercise 9*
Yes. For suppose 3 = $(y .
Then N(3) =
N(4 )N( & ) = 9 .
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
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 .
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,
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
V/m§q' cq"
V/m +
£ 2 (2r+l)q'
l/ra
 _ 1_
]/m
>7
>
q
l/m + r + r + 1 (2r + 1) q£
> 1 >
(2r + 1) q£
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
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^
P P P
factors (except l). We have 20b = 54 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>
N N <3 .
2 2
8. Since a, b are rational integers, we have a + 5b >
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{ ,
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 .
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.
m, 2t« 2t3 2t
Then n has the form n = 2 p2 p3 . . . p and we
may write
m, 1
if
t3
*
"2"
..Pr
2
.
pg P3 n^
)
n = m,l
tg
tg
if
t
75
g
m, is even.
(2
. .
p2 p3
.
q. e.d.
If n =
2m"1(2ml) and 2ml is a prime, then n is a
perfect number.
Proof:
We need only show that (n) = 2n The prime divisors of
U
.
2
 
(T(n)=fH (^J2
1
Then
21
(2m_l) _
x
((2ml) +1) .
((2ml)  1).
ra .
,
1)
(2
{(2m _1} _ 1}
74
If m  7 , 2ml is a prime.
If m = 13 , 2ml is a prime.
_ mnnu
1
m, iru m, m, nu1
^
Then 2ml = 2 ^1 =
(2 X) ^1 =
(2 1l) {(2 1)
m, nu2 m,
+ (2 X)
d
+ ... + 2
.1
+ 1) .
Since 2
.1
> 4 , 2 1 ^ 3 and 21 is not a prime, q.e.d,
21 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) >
q + q = 0.
,2m = CT(q) . Hence these are the only
n =
2m(2m+1l) q.e.d.
n
... + d»
i£
= 2
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
J4~
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