Escolar Documentos
Profissional Documentos
Cultura Documentos
http://hdl.handle.net/2027/mdp.39015028495656
U&mris<
1 8' 7
33
■- library
Financial support for the School Mathematics Study Group has been provided by the
MAY 2 3 1977
CONTENTS
* * *
Page
1. Prime Numbers 1
2. Congruence
Answers to Questions 29
Preface
Not all of mathematics can (or should) be taught in formal
size, we get
large do the primes get? You can probably find a prime larger than
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 -n/iOI? See if you can prove that if an integer n has a proper
divisor, it has a proper divisor which is not more than^ru )
P]P2P3... Pn + 1,
A =
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
or 5 (and so is greater than 5) and p^2«3«5+l = 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
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
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.
( 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 odd divisors.)
(*) A =
Mp-,^... Pn)-L
Notice that A is of the form 4 k - 1.
no p* divides A.
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.
51 J + 2, 51 J + 3, 5U + *, ..., 5U + 51.
There are 5° consecutive integers in this sequence. The first,
51 JI + 2, is divisible by 2, since 5U is and 2 is, and the sum of
two integers divisible by 2 is again divisible by 2. The second
easily see how to modify this proof if you want a block of length
there any other prime triples? See if you can prove it one way or
the other.
2. CONGRUENCES
multiple of some other number. For example, odd numbers are those
which differ from 1 by a multiple of 2, squares of odd numbers
We write
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)
Namely, we have:
N =
aQ
+ 10a
x
+
102a2 +... +
10^^.
(The digits of N are therefore a0, a^, ..., ak.)
Since 10 m 1 (mod 9), we have 102 s l2 = 1 (mod 9), and, in
general, 10r = 1 (mod 9). Hence, using the first equation of (l),
we get
ao + al + • • • + ak — ° (m°d 9).
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
= m + m, 3, m
Since a *s P1**111© "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
unique solution in the range 1 <j x < m. (Why can't x = 0?) The
15
a",
"a; aa 1 (mod m) . To we need
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 is prime, then
p
ap = a (mod p) .
(4) a?"1 =
(mod p), (a,p) =
1
1
(mod
a«2a...(p-l)a 1.2...(p-l)
p)
(5) = (mod .
But the left member has - factors and equals ap~ *1«2. . . (p-l) .
p
a =
aa = aap"2 ap_1 =
p)
= (mod .
1
(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
x0, yo is one solution, then certainly x0 + 13t, y0 - 4t is also a
solution if t is any integer. (Check this by substitution.)
Equations of this type are called Dlophantine 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
4x s 35 (mod 13).
Since ¥ mod 13 is 10, as we saw in Example 1, we have
x = 13t - 1.
(3a) ax + by = c,
we assume a prime to b. Then we solve the congruence ax = c
(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
(1) d I a, d | b,
in the above definition.) Suppose there were two gcdfs 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
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
(3.3) 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|3 and 3| 5-6 imply
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.
3 = 33 - 5 (72 - 2-33)
or
3 ■ - 5*72 + 11-33.
(4.1) a = qb + r, 0 _£ r < b
(4.2) b - qxr + 0 £ r
xj_
r1$
<
(4.3) r =
q2r1 + rg, rg
0
^
i^
<
(4.4) =
q3r2 + r3,
°
r2
r3 r2*
<
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.
(4.(1*1)) =
qnrn-1 rn,
+
rn-2 rn
0
rn-1
4
<
(4.(m-2)) rn-1
=
qnfl rft.
(4.(n+2)) and so
rn|rn_2
by (4.(n+l)).
Working upwards we conclude that rn a, rnl^' ^ *s a common
^
(4.(n+l)) we have
rn
=
rn-2
- Qnrn-1* Botn rn-2 can
rn-l and
ax + by - d.
ax + by = 1.
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
acx + bey = c
As a corollary we get
THEOREM 3. If the prime p|bc and p1b, then p|c. (plb means
" p does not divide b" . ) For if plb, p must be prime to b, since,
as a prime, p has no divisors other than 1 and itself. We can
leave as an exercise.
Exercise 6. Let p and p1, p2, ..., Pn
be primes. If
p|(p-jP2...p ), 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
reasoning, q^
- 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 p1 _£
p. - q, £ q. = p.; since the first and last
members of this chain of inequalities are the same, we have equality
throughout . )
q2q3... qs ,
=
P2P3... Pr
and proceed as before. We get p « q , p_ = q„, etc.
If s > r, we would have
and needs no proof. However, there are many number systems besides
ANSWERS TO QUESTIONS
PRIME NUMBERS
A =
P1P
. . . p +1 shows that A >
P-^2
'" Pn ^ pi* slnce
each prime > 1.
(4^ +
l)(4k2 + 1) - 16 kjkg +
4k2 +
4k2 + 1 .
4(4^1^ + kx + kg) + 1.
a contradition.
CONGRUENCES
(mod 0
(mod
0
aQ a^ ak
Exercise 3. Let the integer be = + 10a, ... (this
N
+ +102<Ja2,
a
(mod
1
if is odd. Hence, = - aj + - a3 +. . .+
N
k
a0 a2 a2-
and ■ if and only if
N
+ + ... + a2j s +
a^+ ac +.. .+ &pi-l
+
aQ &2 aij. a^
aQ lOa-j^ lO^aj^.
3k
Hence, N as
aQ + 10a! (mod K) and N ■ 0 (mod 4) if
and only if aQ + 10a !
■ 0 (mod 4).
Exercise 5* Suppose x^, Xg both satisfy ax, = b, ax s b (mod m)
xl " x2
= ®* The two x's are the same.
25 - 2x 25 - 2(-l 3t)
J m
3
m
3
+
m _
square.
SCHOOL
MATHEMATICS
STUDY GROUP
Lb
v.:.
Financial support for the School Mathematics Study Group has been provided by the
CONTENTS
* * *
Page
1 . Arithmetic Functions - I - The Number of Divisors of
Integer
an 1
8. A New Field 51
Answers to Questions 57
Preface
ii
1.
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 any number base, b , we would need b symbols for the num
bers 0, 1, 2, . . . , b-1 .
3-122 + 1-12 + 2
or 446 in the base ten.
The symbol 4et21 would represent
2
3
Divisors
1, 2
1, 3
Number
2
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?
V/hich numbers do not have an even number of divisors? 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 OP THE DIVISORS OF AN INTEGER
"In the beginning created the heavens and the earth."
God
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 sura 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 directly is not too
<r(n) easy.
However, we can use the approach of the experimental scientist and
collect some data. Suppose we make a table for <r(n) .
n Divisors of n ^"(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 .
If n = p q
k s
, can you guess what <T(n) is in this case?
Check your answer in a few cases and see if you can prove it.
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 m2 mr
n =
pl p2 * ' "
pr wnere the P's are distinct primes and the
2, 3, 5, 7, 13, 17, 19, 31, 61, 89, 107, 127, 521, 607, 1279, 2203,
and 228l. Accordingly seventeen even perfect numbers are known.
The last five of these were found in 1952 by SWAC, the digital com-
ppQ-i
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.
The Find an odd perfect number.
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 an odd per if
fect number exists, it must have at least six different prime fac-
14
tors and cannot be less than 1.4x10
There is one result about perfect numbers which is true whether
the perfect number or odd. is even 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 .)
If n =
m,
p1 p2 p3°
nip m«
.. pr
m.
, we found that f (n) = (m,+l) .
If
m
... (l + p +....+ p ) . 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
summand by a 1 we would get for the sum exactly T (n) , the
number of divisors of n . This is easily seen by replacing each
p in the formula for <7"(n) by 1 and then the formula reduces
to our formula for f(n) .
(HINT: (Tk(n)
=
(lk +
p£ + (p2)k + ... + (p^)k) ... (lk + pk +
„ , m .
ARITHMETIC FUNCTIONS
III
THE DISTRIBUTION OFir(n)
PRIMES AND THE FUNCTION
(HINT: d If
divides n , then § divides n also.)
Exercise 5- Determine whether 178l and 4079 are primes.
With this theorem, we have considerably reduce! the work of
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 \/r\ . It goes like this.
Write down all the integers ^ n . For example, take n = 25 .
iMMMHiO^MlSWlSlfilT 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
where the
[|]
n
[§]
[SJ pk
. .
,
.
ing process. It
taken out as a multiple of
was and also as a
2
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 = p,p2 where p, and p2 are primes, it will
,
be taken out when we sieve with p.. and again when we sieve with
p2 . So that in all such cases in order to have the integer taken
out only once, we must add it back in once. better estimate of
A
16
ir(n) would then be an expression like
n-[i]-[§]---[U+fe]+[^3l+-+fe]
•
Even quite
this expression won't
consider do. We must numbers of
the form p-^PoPo • These numbers will be sieved out 3 times; when
--(iy+K+---ft])+(te]+ fe3]
+
-+[^A])
(• * •' etc.
This expression seems to go on indefinitely. However, as soon as
— < 1 , =
' anc* ^^e comPllcated expression actually has only
(m) ^
„«, . 25
'.([f|+
p|]+[f])+
"
(fcfl
♦
[£] +\g\)
3 1
[-2^5-]+
17
IK 25) = 25 - (12 + 8 + 5) + (4 + 2 + 1) -0 + 3-1
= 9 •
(HINT: log n =
log10n , where M =
0.4342945...)
4. W
(1) 2x + 3y = 5 (4) 4x + 6y = 9
(2) 2x + 4y =
5 (5) 4x + 6y = 8
(3) 3x + 3y = 5 (6) 2x - 4y = 4 .
integral solutions?
Which of these have
Look at the coefficients of x and y and the constant term
in each of the equations for which you found a solution. Is it
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 _d 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 very clumsy notation.
a 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
r
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 + 1) 4 + 17-2 = 17 {(2-5 + 1)- 4 + 2)
3315 = 17 {(2-5 + 1)
-
4 + 2) 4+17 (2-5 + 1)
l)j
=
({(2-5 +1) (2-5 +
•
17 4 + 2}-4 +
•
Thus we have shown that 17 is a divisor of both 3315 and 782 .
,
.
,
it divides 782 187-4 = 34
- But then d* divides 187 and if
.
253 = 122-2 + 9
122 = 9-13 + 5
9=5-1+4
5 = 4-1 -+{7)
4 = 1-4 + 0
The greatest common divisor is 1 , the last non-zero remainder.
Such numbers which have 1 for their greatest common divisor are
called relatively prime . Check that (253, 122) = 1 by factoring
the two numbers .
rn-2= rn-rqn +
rn • where
rn is the
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 - -r t where t is an integer t .
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.
our method we get the solution By
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 + k6 t , y- 195t . To find
= -89
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
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 79 persons. What is the share
obtained by each? (Mahaviracarya)
5. l4x - 45y = 11 .
5.
and m = 2a
2
n = a + IT2 .
29
Then
(1) a =
^§ and
2
-
(2) b =
/ 4n
-±V*H
+-1 m2
Since b is rational
- ? 2
(3) 4n m = c , where c is some rational integer.
Substituting (3) in (2) we have
(4) b -±c .
2 2
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 .
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 .
A p > v p '
Exercise 6.
Is 3 + 111 divisible by 2 + 3i ? by -i ?
Lemma. N(<*^3)
=
N(<* ) N(/9 ) , where N(<* ) denotes the norm
of °* '
„
Proof: If we let c* be the conjugate of <* and ($
be the
conjugate of /3 ,
N( <*
)
- «* ot
N(/3 ) =
P/3_ __ _
N(<*/3) =
<*/S */3 Since <*
/? =
oc-/3
=(««)( (3 ft)
=
N(o< )N(/9)
The lemma also can 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
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 .
Exercise 8.
Is 5 a Gaussian prime?
Exercise 9-
Is 3 a Gaussian prime?
Let us now look more closely at rational primes of the form
4n + 3 . Suppose a rational prime p = 4n + °t/*
3 =
. Then
N(p) =
N(<x)N(f3 ) =
P
o
• If P is not a Gaussian prime, then
there must exist <* and (i such that N( oC ) 1 and N( (3) ^ 1. /
In that case, N( <* ) = p and N( ft ) = p . But * = x + yi , if
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 fi cannot be p , the norm of one of
them must be 1 and that one is unit,
a and the other is an asso
ciate of p . Since p has no divisors except units and associ
ates of p , we have proved the following theorem.
Theorem 2. Every rational prime of the form 4n + 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 many rational primes of the form 4n + 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
From 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 .
3h
By hypothesis N( « ) =
N( (3 )n( f ) - P , where p is a
rational prime.
Since N( (% ) and N( X ) 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 , (c) 17 , (d) 29 , (e) 101 , (f) 1721 .
(b) 13 ,
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
2
primes of the form 4n + 1 . Suppose p « 4n + 1 = x +y 2 . We
p = x2 + y2 =
(x + yi)(x - yi) .
Gaussian prime.
to question three, then is: The only rational
The answer
integers which are Gaussian primes are the rational primes of the
form 4n + 3 .
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 and on the other hand there are only finitely many
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 = ^2 + l2 , 29 = 22 + 52 .
Exercise 1.
Write 37, ^1, 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 that there was no prime of the form 4n + 1 which
means
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 Dlophantus , Fermat made his
famous note regarding the problem of finding rational solutions of
the equation
(1) x2 + y2 = z2 .
'?'?'?
We
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 that the only possibility for a
show
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) ifu2 + y2 = z2 or
n
4u
2 = z
2 - y2
(*) 4u2 =
(z + y)(z - y) •
2
and ab = c , then a is a square and b is a square.
Proof: By the Fundamental Theorem of Arithmetic (see the
supplement entitled The Fundamental Theorem of Arithmetic) , we may
write c as tne pro duct ol prime factors p. , p« , . . . p .
Then
2 2
c =
(PjPq" -Pn)
= ab .
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 Accordingly, if we let
b
.
,
xl x2
p. be the first prime that divides b_, p. be the second,
xk+l 1k+2
etc . we must have
;
P
2
2
a = p. p. . .p. = (p, p. . .p.
.
xl
)
x2 xk xl x2 xk
2
2 ...p, ...p.
2
= p. p. = (p., p,
b
)
Tc+l 1k+2 xn xk+l xk+2 xn
q.e.d.
We now return to our problem of finding the solutions to the
divisor of
2
equation
2
+ .
z
+
y
2
z
= v and = w and
y
Z
(5) + = 2v
y
z
(6) - = 2w2
y
z
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 .
(8) x = 2vw
,
(9) y = v2 - w2 ,
(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) ,
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
a = d'
2 2 2
From (ll) we have that a = b + y , and again by Theorem 2
b = 2rs ,
2
(13) y = r2 - s ,
■2 2
a = r + s ; where r and s have no
common factor. But from (12) and (13) we have
2b = or = 4rs
By Theorem 1, then r = x.
kl '
s = x, + y_ and since by (12) a = d ,
we have
xl +
yl = d where 1 < d < a < z .
7.
O.a^ap... an
, i.e., a-j/lO +
a2/l02 + . . .a /±0n . )
H
<••»+/
•■#.-£.
%
is of length —
M.
; the points of subdivision are then
m +- , m
m 4
+ -
2
, m + 3zi . The point which represents <*.
(1) |
« _ R
q <
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
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 - p , s<* - p , which lie in the same box.
" /+ 1
Pr <
n
s <* " i + l
n < Ps< n
The second inequality may be written
+
*s <x
P - -n
Adding the first and last inequalities, we get
Hence,
(3) _ £ ^
q nq
<
q nq
with 1 1 q < n .
2
n > q , nq > q , so that
I
<*■
-p/q I < i/nq £ i/q2, or
(5) _ R
q
49
>*.n' ^ n '
^
■* ' El
q1"
s
* -1
n~*qp" s^
1 / 1
This shows that p/q is not the same as p'/q' , since the first
is further from o< than l/n , whereas the second is nearer to ©<
than l/n . That is, p'/q' ^s nearer to o( than p/q is.
Thus, p'/cl' is a new approximation to °< , and moreover,
C<
-4 <
n'q' <
,.2
1
(7) ot _ <
2q2 - p2 = (q V? - p) (qV?+p)>l,
2 - if
because 2 q p is a positive integer and so is at least 1 .
1/2"
_ P.
q
=
v*"-f >
4 q
or
v^-i >
f+V5-
>
2 + V2"
>
^q£
8.
A NEW FIELD
The field you are most familiar with is the rational 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 \/-5 (=a + ib \/5~) , where
a and b are rational numbers. We define two elements of K ,
a + b to be equal
\/~5 > if
they are equal as complex numbers,
i.e., if
and only a = c and b = d . if
It is easy to check
that K is closed under addition, subtraction and multiplication.
(Do this.) K contains 0 = 0 + 0 • V^B" and 1=1+ 0- \pS .
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 .
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
M = a2 + 5b2 .
notice that N
We \ =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 0of
J , not a unit, whose only divisors are - 1 and - ){ . This, of
course, agrees with the definition of prime in I .
1. and a
6
N
1
,
= = we have a
d\
+ 5b
N
3
,
(^
1
3
,
.
7
If you have read the Supplement Prime Numbers you will know
J
.
,
2 + 3 \A5 , 6 +
\f^ .
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 = where neither nor is a unit or is
-
9)
a
(?)
0
o(
,
so
^
=
<*
• we
N
.
<3
6
N
Not,
,
,
have N§ No( and also N^ N«* .
<
<
1
1
<
<
This
that shows for N$ a Hence, has a
6
.
^
S
<
Q
,
(2) 21 = 3 •
7
21 =
/T5) (4 - /C5)
(4 +
ofq.auit[q.T«iV suoT^ouriji
I
JcaqutnM jo
saosfATQ saoefAfa
T T
T
rT ' s 5
s
rT e 3
e
rT '3 ^7 E
rT s s
rT ' 'S 'e 9 +7
9
rT ' Z s
Z
'
'T. 'e 6 e
6
TT L
T '
'< TT S
ST 'T' 'S 'e >M '9 ST 9
CT rT ' CT 5
^T 'T' 's 'Z irT *r
ST rT ''e 'S ST tr
6T 'T' 6T s
03 '3 '5 'ot os
'fr
fT '
9
S3 fI 's 'TT 33
+7
'
es 'I es
s
'
+rS
'
S3 'T 's SS
e
'
ZS fT 'e '6 zs
tr
'
9
'
6s rT 6S
3
'
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.
m- nip m
For n =
px p2 ... pr , r(n) =
(m][
+ l) .
12, 18, 20, 28, 32, 44, 45, 52, 63, 68, 75, 76, 92, 99 •
144 .
59
3. 7- None. At most one. Since x
k - 1 - (x - 1)
/ \
(xk"± +1) ,
if x > 2 the number is not a prime.
6o
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 K LviS(?rs of n <D[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, *, 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, *, 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, *, 5, 10. , 20 42 A
21 1., 3, 7, 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, 1^. , 28 56 P
29 1.. 29 30 D
p - i
1, P, P 2 ,...,P k , Q, pq> P
2
q,...,Pk q . 2(k + 1) .
pkq2 •
3(k + 1) .
uqq;
(1 + q +
o
q2) £
~^+l
^-
p-1
-i
=
„3
q "
q-1
n
1 •
„k+l
£
p - =-i
1
n
*
- -
p-1
qS+1
Pk+1 1 .
q-1 1
....
m, +1 nu+l
V
m +1
- - -
^(n)
"
I -
1
. P2
-
I1 Pr
-
1
.
Pi P2 Pr 1
6 = 2-3; d"(6) =
%^-A
•
il^l = 3.4 = 12 .
,2 n „3
18 = 2-32 ; </(l8) =
22~_l
'
33 I i = 3-13 = 39 .
4
" 2 "
*
24 = 23-3 ; (^(24) = • = 15.4 60
2g | 33
= .
•
28 = 22-7 ;
(f(28) =
^-=-J :L7^1
= 7-8 = 56 .
30 = 2.3.5 J </(30) =
22~_l
•
33 I I •
3b I 1
= 3.4.6 = 72
,5
144 = 24-32 ; (^(144) =
22~_\
'
33 I \ = 31-13 = 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-
1781 = 13-137 ; 4079 is prime.
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 623, 27b
log n 10,000,000 7 7
63
The Euclidean Algorithm and Linear Diophantine Equations
63x + 7 = 23y
63X + 7=Z*y
i < i
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)j (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.
See proof which follows.
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 .
(365, 146) = 73 .
and ax + by = c .
Divide by d
|(x - xQ)
=
-|(y - y0)
65
Since a/d and b/d have no common factors, x - x must be
divisible by £ ; let o d
Then substituting we hav«
H-t - -
J (7 -
x„) and y - y0=
-
z t Consequently
x - xQ
+
|t ,
y . y0
-
ft ; is a solution for every
integer t
CHECK:
a(x„ + O
"' + ^(y. - # t) « e .
"o d
7T ~wo d
ab - ab
^d t t
.
ax„ +" by
~" + = +
byQ
= c
"o o d axQ
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 = l4 - 63t .
Answers to Problems :
1. x = 3 + 7t , y = 79 - l6t .
2. x = 170 , y = 110 i x m
923 , y = 9 .
(x = 923 - lOlt).
+ 753t , y = 9
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) + (c + di) =
(a + c) + (b + d)i .
N(M(3)
=
(ac - bd)2 + (ad + be)2
.2 2 ,2,2 2,2
= ac
2 2
=bd +ad +bc , ,
=
(a2 + b2)(c2 + d2)
=
N(o{ )N(§>)
67
Exercise 7-
If JL and £ are associates, then q\ =A «u, where u is a
unit.
Then by the Lemma N( = N( ^ )N(u)
^ )
=
N( 3 )-l 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 - are not and i units. N(5) =25
they are not associates of 5 by Exercise 7- 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( <3 ) = 9 .
Case III: one even and one odd; say x = 2x' ,y=2y'+l
xc2 = y2 =
4(x«2 + y'2+ y')
2
+ 1
2
/ 4n + 3 for any
rational integer
;eger n . Therefor'
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 =
N(^ )N(0 )
But since are rational integers, of
<p{
and one
N(
$
N(
)
)
is and the other is =1 then
N(
them 2 . Suppose o( o^
;
)
is a unit. Therefore + is a prime since it can only be
i
1
written as a unit times an associate of + .
i
1
Exercise 12.
Yes. N(l - =2 and we can repeat the same argument given
i)
in Exercise 11.
Exercise 13.
No. Since every rational integer is a Gaussian integer, a
composite rational integer a = be has as divisors the Gaussian
integers and which are not units or associates of a
b
.
Exercise 14.
22
22
l2
52
(a) = + (d) 29 =
+
5
-
22
32
l2
(b) 13 + (e) 101 = 102 +
ll2
42
l2
Exercise 1.
37 = 62 + l2 ; ^1 = 52 + ^2 ; 89 = 52 + 82 ; 101 = 102 + l2 .
Exercise 2.
2 2 2
Given: d divides x , d divides z , and x + y = z
Show: d divides y .
d2x'2 + y2 = d2z'2 ,
,2/ ,2
y
2 = d (z« - x«,2x)
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 2 2 2 2
are all odd then x + y is even; but x + y = z and 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 z
2
is even, then so is z .
Exercise 5.
2
Given: z and y have no common factors; z + y = 2v ;
z - y = 2w .
p
Then adding z + y = 2v
z - y
2'z
=
=
2w
2(v
op + w ) , z = v
P
+ w
?
;
Subtracting y2 _ „2
0y = 2(y22 _
w2} > y = >
05- B <
1
cq
Then mq-p p
o
= (q \/m
- p) (q \/m + p) > 1 ,
v* -f .1 *-
q
P
1.1
+
' T)
q
P
2
1 >
(2r+l)q2
1
l/m £ \/m
]/m
-
q
q' l/m
v
1
+ £
q
.1
qr +
1
r + 1 (2r
1
+ l) q2
,
v=-| /
v
■■■ /
v 1
p j
(2r + 1) q^
A New Field
1. (a + b |/=5) + (c + d v/^5)
=
(a + c) + (b + d) j/^5
=
a2
+
h1
]p5
(a + b \/=5) - (c + d =
(a - c) + (b - d) /^5
1/^5)
=
a2
+
b2 /^5
(a + b \/^5)(c + d \/^5)
=
(ad - 5cd) + (be + ad) \f^>
=
a3 * b3 V^5 •
+
b
d |/^5
J^5 =
ac + 5bd
cT +
+
5d2
(be,- ad)
1/^5^ -1
+b -1 ^
when c + d y^F ^ 0 (i.e.,
not both c and d are 0),
we have a, , b, are rational since a, b, c, d are.
c>(
satisfies the monic equation of smallest possible degree
x - o( = 0 . This equation has coefficients in R since
*K is in R .
So 8 +
sf= o{
f[ +
<^(^2=^^^i +
^Q •
This gives a
2
+ 5b
2
= p . If a^b
are both even, the left
member is even. If a Jo 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 <K = p .
is a unit in J .
10. If 4 +
]/^S
= o( G> , we have 21 = N o( . n <3 . Now
N °( or for as we just saw, ^ = is impossible
-jf
p
3
7
,
N
1
<*
l
= Same
N
o(
,
for - I/C5
4
.
73
1, 2, 4, 8, 9, 16, 18, 25 . They are either a square or
twice square.
a If (j (n) is odd, then n is a square or twice
a square.
Proof:
m,+ 1
- m2+l - mr+l -
1 1 1
P-L
= -± — p2 Pr odd.
Given: <j (n) r^- - 1
' ' ' - 1
P2 Pr
Then each factor of the product must be odd.
If p.. = 2 , then m, may be any integer > 0 .
m,
2t2 2t3 2t
Then n has the form n = 2 • • •
Pr and we
P2 P3
may write
m, -1
2 (2
2
p22 p33...prr)2 if m1
is odd
n = m,-l
(2
d
z>2 P3
. . .
prr) if m1
is even.
q. e. d.
Proof:
We need only show that U(n) = 2n . The prime divisors of
n are 2 and 2m-l . We can make this statement only
because we are given that 2m-l is a prime.
Then
•(„)-££ d L
(stills-
-
(2-1) 1
(2m-l) (2m-l+l) =
2m(2m-l) = 2n .
If m ■ 7 , 2m-l is a prime.
dl, nu m, m, rau-1
m^nu
1 ^
Then 2m-l = 2 ^-1 =
(2 1) ^-1 =
(2 1-l) {(2 X)
m, m,~-2 m-,
+(2 x)
d
+ ... + 2
x
+ 1) .
,2A -1
m,
Since 2
-1
> 4 2 3 and 2^-1- is not a prime, q.e.d.
m-1 /_m
Every even perfect number has the form 2 (a -l) where
2m-l is a prime.
Proof:
U (m>n) = 0 (m) • Q (n) .
(Lemma: To prove this, one
only needs to write out the expressions for O (m-n) , ^(m) ,
and (f{r\) and verify that <T"(mn) =
(^(m) •
(T\n) .)
Let n = 2 q , where q is odd. Since n is perfect
(1) <T(n) = 2(n) = 2m+1q .
But by the lemma, ^T(n) =
<f{2m) 0i[q) . Substituting in
(1) we have
Now (^(2m)
= 2m+1-l . Substituting in (2) we have
n =
2m(2m+1-l) q.e.d.
1 1 1 d|
+
d«2
+ ... +
d'k
Then
*+% +'-'+\= »
cr; (n) m
<(">
n
k(m +1) -1
" k(m1+l)_1 k(m2+l)_1
<(n) -
fl k ,
£2
k - ,
Pr •
-
P1
1
P2
1
«*-*
1 V
£7-/^* o*otf-A-±o