Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Quadratic Reciprocity: Its Conjecture and Application

Author(s): David A. Cox


Source: The American Mathematical Monthly, Vol. 95, No. 5 (May, 1988), pp. 442-448
Published by: Taylor & Francis, Ltd. on behalf of the Mathematical Association of America
Stable URL: https://www.jstor.org/stable/2322482
Accessed: 31-07-2019 21:18 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

Mathematical Association of America, Taylor & Francis, Ltd. are collaborating with
JSTOR to digitize, preserve and extend access to The American Mathematical Monthly

This content downloaded from 143.107.231.30 on Wed, 31 Jul 2019 21:18:18 UTC
All use subject to https://about.jstor.org/terms
THE TEACHING OF MATHEMATICS

EDITED BY JOAN P. HUTCHINSON AND STAN WAGON

Quadratic Reciprocity: Its Conjecture and Application

DAVID A. Cox
Department of Mathematics, Amherst College, Amherst, MA 01002

If p and q are distinct odd primes, then the law of quadratic reciprocity states
that

(_(_ q (_1)(p-1)(q 1)/4 (0.1)

where (a/p) is the Legendre symbol, defined to be +1 when a is a quadratic


residue modulo p and to be -1 otherwise. Quadratic reciprocity is one of the gems
of elementary number theory, but at the same time, it's never been easy to see where
the theorem comes from. Other formulations are not a great help. For example,
statement (0.1) is equivalent to the following:

If p or q is congruent to 1 modulo 4, then p is a quadratic residue


modulo q if and only if q is a quadratic residue modulo p. If p and q
are congruent to 3 modulo 4, then p is a quadratic residue modulo q if
and only if q is a quadratic nonresidue modulo p.

(Again, p and q are odd primes.) This version of the theorem is easier to read, but it
still sheds no light on how to conjecture such an amazing result.
Of course, once we know to look for some relation between quadratic residues
modulo different primes, it's fairly easy to find examples that lead to a statement
like (0.1) or (0.2). But how does one begin to suspect that such a relation should
exist?
This paper has two goals. The first is to answer the question just raised, i.e., to
present a series of questions and examples that lead naturally to quadratic reciproc-
ity. The second goal is to use quadratic reciprocity to prove the following theorems
of Fermat for odd primes p:

p x2 + 2 x, y E Z p 1mod4
p _ X2?+ 2y2, X y El Z p 1,3mod8 (0.3)
p = x2+3y2, x, y E Z p= 3orp 1mod3.
Our two goals are closely related, for we will use Fermat's results to motivate the
questions that lead to reciprocity. This might seem circular, but it's how things
happened historically. Euler knew of Fermat's theorems, but since Fermat never
published the proofs, Euler had to find his own. This took many years, and along
the way Euler discovered quadratic residues and quadratic reciprocity. This paper
will follow Euler closely, both in the examples leading to reciprocity and in the
proofs of (0.3). For an excellent account of Euler's work on number theory, the
reader should consult Weil's book [4].
There is one other aspect of our second goal which deserves mention. Many
books on number theory present quadratic reciprocity but fail to give interesting

442

This content downloaded from 143.107.231.30 on Wed, 31 Jul 2019 21:18:18 UTC
All use subject to https://about.jstor.org/terms
1988] THE TEACHING OF MATHEMATICS 443

applications. It is important to show that quadratic reciprocity is not an isolated


result, but rather is a theorem that leads immediately to other theorems of
independent interest. (I should point out that several books do this very nicely, in
particular Ireland-Rosen [2] and Nagell [3]. I am grateful to the referees for pointing
out these references and for suggesting improvements in the exposition.)

1. We will first discuss Euler's strategy for proving Fermat's results (0.3). One
half of the theorem is easy: if p = x2 + ny2, n = 1, 2 or 3, then the desired
congruence for p follows by considering x and y modulo 4, 8 and 3 respectively.
The converse is more difficult, and this is where Euler's strategy comes in. For the
case of X2 + y2, it consists of the following two steps:

Step 1. If p 1 mod 4 is prime, then show that p divides a sum of the form
x2 + y2, where x and y are relatively prime integers.

Step 2. If an odd prime p divides a sum of the form X2 + y2, where x and y are
relatively prime, then show that p is a sum of two squares.

The cases x2 + 2y2 and X2 + 3y2 can be formulated similarly, and it is clear that
these steps suffice to prove Fermat's results (0.3).
Notice that both steps involve prime divisors of x2 + ny2, where n = 1, 2 or 3. It
thus makes sense to look more closely at this situation, and as happens often in
mathematics, we can better understand our problem by posing it more generally.
Hence, in considering prime divisors of x2 + ny2, we should let n be an arbitrary
integer, positive or negative. We still want x and y to be relatively prime, for
otherwise any prime could divide x2 + ny2 just take x and y to be multiples of p.
We can restate our problem in terms of the Legendre symbol as follows:

LEMMA 1.1. Let p be a prime not dividing n. Then there are relatively prime
integers x andy such that pIx2 + ny2 if and only if (-n/p) = 1.

Proof. Suppose that p divides such a number x2 + ny2. Then x2 - ny2mod p.


Since x and y are relatively prime, it follows that p + y. The integers modulo p
form a field, so that yb I mod p for some b. Multiplying our congruence by b2,
we see that (xb)2 - n mod p, which implies that (-n/p) = 1. The converse is
trivial, and the lemma is proved.

We are now ready to turn to Step 1. Using Lemma 1.1, we see that is suffices to
prove the following for primes p:

I p- lmod4
P

-2
I 1 p 1,3mod8 (1.2)

=I 1 p I p mod3.

As we did before, let's pose this more generally. Thus we want congruence
conditions on p that imply (- n/p) = 1. Looking at (1.2), we see that the way to
unify the congruence conditions is to work modulo 4n.

This content downloaded from 143.107.231.30 on Wed, 31 Jul 2019 21:18:18 UTC
All use subject to https://about.jstor.org/terms
444 DAVID A. COX [May

It's now time to do some examples. We will work out the case n - - 5 in detail.
This means studying the remainders modulo 20 of those primes p for which
(5/p) = 1. One way to generate such primes is to factor the numbers x2 - 5 for
variable x. If we let x range from 1 to 30, we get the following prime factors
(excluding 2 and 5):
11, 19, 29, 31, 41, 59, 61, 71, 79, 89
109, 131, 139, 179, 181, 191, 251, 471, 571.
Working modulo 20, all of these primes are congruent to +1 or + 11, and the first
row lists all primes under 100 which satisfy this congruence condition. If we let x
range from 1 to 80, we get all primes under 200 that satisfy the congruence. It
appears that we will eventually pick up all such primes, which leads to the
conjecture that for a prime p i 2,5,

p +1, +llmod20.

=1 p-1,7modI2
P

Other examples can be worked out similarly. Here are some sample cases:

=I1 p 1,3,7,9 mod 20


P

_~ =1p- +mod12
= 1 p 1,9,11,15,23,25mod28 (1.3)

_ - 1 p ?1,+3, +9mod28.
5
p ?1, ?11mod20

Here we always assume that p is an odd prime not dividing n.


In looking for something to unify these conjectures, the bottom three results of
(1.3) offer the most hope because of the +'s. We will thus concentrate on the case
(N/p) for N> 0.

_ =1 p- +Ilmod12
The key problem is to discern a pattern in the numbers that follow
course, numbers in congruences can be written many ways. For example, 11
- 9 mod 20 and 3 -25 mod 28. Using these two facts we can rewrite the bottom
part of (1.3) as follows:
3

S
(_)=1 p- +?1,?9mod20 (1.4)

-7+
l " } 1 <, p ?I+, ?25, +9mod28.

This content downloaded from 143.107.231.30 on Wed, 31 Jul 2019 21:18:18 UTC
All use subject to https://about.jstor.org/terms
1988] THE TEACHING OF MATHEMATICS 445

All of a sudden there is a pattern: all of the numbers that appear are odd squares!
Before we get too excited, let's try another case:

p +1, +5mod24.

Unfortunately, ? 5 is not congruent to a square modulo 24. The same phenomenon


happens for (10/p) and (14/p). So why does the pattern work for 3, 5 and 7 but not
for 6, 10 and 14? The obvious difference is that the former numbers are all prime!
This says that there is something very special about the prime case, i.e., the case
of (q/p) = 1, where p and q are distinct odd primes. From (1.4), we get the
following conjecture.

CONJECTURE 1.5. If p and q are distinct odd primes, then

( I) = 1 #p- +?2mod 4q for some odd /P.

Notice that we have now given an answer to the basic question posed in the
introduction-the above examples single out the prime case very nicely. It remains
to show that Conjecture 1.5 is equivalent to the usual formulation of quadratic
reciprocity. We do this as follows.

THEOREM 1.6. Conjecture 1.5 is equivalent to the law of quadratic reciprocity for
distinct odd primes (see (0.1)).

Proof. Let p and q be distinct odd primes, and set p* = (-1)(p-l)/2p. Note
that p* 1 mod 4. We will assume the following two properties of the Legendre
symbol:

(-) = (1.7)

ab a b

(see [3, p. 135]). Using (1.7) it is an easy exercise to show that quadratic reciprocity
is equivalent to the statement

{qA {P*

Since each side equals ? 1, it follows that quadratic reciprocity can be written as the
equivalence

q P

and comparing this to Conjecture 1.5, it thus suffices to show

p ?f32mod4q (-) = 1. (1.8)

Note that /32 1 mod4 since P is odd. Thus the + sign in (1.8) must be

This content downloaded from 143.107.231.30 on Wed, 31 Jul 2019 21:18:18 UTC
All use subject to https://about.jstor.org/terms
446 DAVID A. COX [May

(-1)(p-1)/2, and we then have

p ? ,2 mod4q p (1)(P-1)/232 mod 4q

/ p*2mod 4q.

Now, to prove (1.8), suppose that p* = - 2 mod 4q. This implies p* -,2 mod q,
so that (p*/q) = 1 follows immediately. Conversely, if (p*/q) = 1, then p*-
a2mod q for some a. Letting a3 = a or a + q, depending on whether a is
even, we obtain p* = f2mod4q, and the theorem is proved.

With this theorem and the preceding examples, it becomes clearer how someone
could have conjectured quadratic reciprocity. In our case, however, the "someone"
is none other than Euler, for we have followed roughly the same path he took to
discover reciprocity. In fact, our examples (1.3) are taken from Euler's 1744 paper
"Theoremata circa divisores numerorum in hac forma pxx ? qyy contentorum" [1,
vol. 2, pp. 194-222], and in the same paper there are a series of Annotations that
include Conjecture 1.5 as a special case. For more details, see [4, pp. 186-187,
204-210].
We should also mention the other parts of reciprocity, the so-called complemen-
tary theorems for odd primes p:

( _1) - _1)(P 1)/2

2 (1.9)
p- 1 _)(p2-l/

Notice that we've already used the formula for (- I/p) in the proof of Theorem 1.6.
For our purposes, the key fact is that once we've done a full treatment of
reciprocity (statements (0.1) and (1.9)), Step 1 of the proof of Fermat's results
follows immediately-the equivalences (1.2) are now an easy exercise which we
leave to the reader. It remains to prove Step 2, which will be done in ?2.

2. Earlier, we stated Step 2 only for the case X2 + y2. We will state the general
case, for X2 + ny2, n = 1, 2, 3, as the following theorem.

THEOREM 2.1. Let p be an odd prime that divides a sum a2 + nb2, where n = 1, 2
or 3 and a and b are relatively prime. Then p can be written in the form x2 + ny2.

Proof. We will follow the arguments Euler used in 1752 for the case X2 + Y2 [1,
vol. 2, pp. 300-307]. We will generalize his arguments to X2 + ny2, n = l,2, 3.
The crucial step is the following lemma.

LEMMA 2.2. Let q = X2 + ny2, n a positive integer, and suppose that q divides a
number N = a2 + nb2, where a and b are relatively prime. If either q is prime, or
q = 4 and n = 3, then N/q = c2 + nd 2, where c and d are relatively prime.

Proof. Let us first consider the case where q is prime. Since q divides both
x 2N = X2(a2 + nb2 ) and a2q = a 2(X2 + ny 2), it divides their difference

x 2(a2 + nb2) - a2(x2 + ny2) = n(x2b2 - a2y2) = n(xb - ay)(xb + ay).

Since q is prime, it must divide one of these factors.

This content downloaded from 143.107.231.30 on Wed, 31 Jul 2019 21:18:18 UTC
All use subject to https://about.jstor.org/terms
1988] THE TEACHING OF MATHEMATICS 447

If qln, then q = n since q = x2 +


a = nd. Then N = n2d 2 + nb2, which implies N/q = b2 + nd 2, as desired.
If qlxb - ay or qlxb + ay, we can assume that the former holds by changing the
sign of y. Then xb - ay = dq = d(x2 + ny2). This implies that

xb - dx2 = ay + dny2 = y(a + ndy), (2.3)

from which we conclude that xly(a + ndy). Since x and


prime), we must have xla + ndy, i.e.,

a + ndy = cx (2.4)

so that a = cx - ndy. Substituting (2.4) into (2.3), we obtain

x(b - dx) = y(cx),

which implies that b = dx + cy.


However, we also have the famous identity

(c2 + nb2)(x2 + ny2) = (cx - ndy) + n(dx + cy)2.

Using the above formulas for a and b, this becomes

(c2 + nd2)q = a2 + nb2 = N,

so that N/q = c2 + nd2, as desired. Since a and b are relatively prime, the
formulas for a and b show that c and d are also relatively prime.
It remains to consider the case n = 3 and q = 4. Here, we have 41a2 + 3b2, so
that a and b have the same parity. Since a and b are relatively prime, they must be
odd. Since 4 = 12 + 3. 12, the argument for the prime case (with x = y = 1) would
work, provided that 41b - a or 41b + a. But the latter holds for any pair of odd
numbers, which proves the lemma in this case.

The proof of this lemma is similar in strategy to Lagrange's proof of the famous
four square theorem, which asserts that every positive integer is a sum of four or
fewer squares (see, for example, [3, Theorem 102]).
To complete the proof of Theorem 2.1, let p be an odd prime dividing a2 + nb2,
where a and b are relatively prime. Assume that p itself is not of this form. We will
show that there is an odd prime q < p with exactly the same properties. We would
then be done by Fermat's principle of infinite descent: applying the same argument
to q would give us q' < q, and continuing we would get an infinite decreasing
sequence p > q > q' > ... of positive integers, which contradicts the well-order-
ing property.
To produce q, we work with a2 + nb2. It is divisible by p, and remains so if we
replace a and b by a - kp and b - {p respectively. Furthermore, we may choose k
and 1' so that Ia - kpl < p72 and lb - epI < p72 because p is odd. Thus we may
assume that pIa2 + nb2 where IaI < p72 and IbI < p72. Since n < 3, it follows that
a2 + nb2 < (p72)2 + 3(p/2)2 = p2. Thus a2 + nb2 can be written as

a + nb2 = pql *. q, (2.5)

where the primes qi all satisfy qi < p.


We claim that one of these qi's is odd and n
Then all of the odd qi's can be written as

This content downloaded from 143.107.231.30 on Wed, 31 Jul 2019 21:18:18 UTC
All use subject to https://about.jstor.org/terms
448 GREGORY L. BULLOCK [May

Lemma 2.2 we can elimina


a2 + nb2 = 2ap

If n = 1 or 2, we can also apply Lemma 2.2 to 2 = 12 + 12 = 02 + 2- 12 to


eliminate factors of 2, showing that p = a2 + nb2, a contradiction. If n = 3, the
case q = 4 of Lemma 2.2 shows that we can reduce to either p = a2 + 3b2 or
2p = a2 + 3b2. It remains to show that the latter case cannot occur. But pla2
implies (- 3/p) = 1, which by quadratic reciprocity means p 1 mod 3, so that
2p--2mod3. Yet 2p = a2 + 3b2 implies 2p a2 lmod3, and thus we have a
contradiction.
This completes the proof of Theorem 2.1.

This finishes our proof of Fermat's three theorems (0.3), and the only gap is that
we haven't proved quadratic reciprocity. So once a course in number theory covers.
reciprocity, there really are nice applications waiting to be proved, and some of
them can even help motivate the statement of the theorem itself.

REFERENCES

1. L. Euler, Opera Omnia, series I, volume 2, Teubner, Leipzig and Berlin, 1915.
2. K. Ireland and M. Rosen, A Classical Introduction to Modern Number Theory, Springer-Verlag,
New York-Berlin-Heidelberg, 1982.
3. T. Nagell, Introduction to Number Theory, Chelsea, New York, 1981.
4. A. Weil, Number Theory: An Approach through History, Birkhauser, Boston-Basel-Stuttgart, 1984.

A Geometric Interpretation of the Riemann-Stieltjes Integral

GREGORY L. BULLOCK,
Mission Research Corporation, Santa Barbara, CA 93102

1. Introduction. For many students of mathematics, the moment of truth comes


in a rigorous study of real analysis. A promising student may become discouraged
during a real analysis course and subsequently abandon serious hopes of pursuing a
degree in mathematics and may even dissuade other students from doing the same.
A graphical explanation of some theorems of analysis can simplify the task of an
instructor and may save some students from unwarranted distress.
Herein are presented geometric interpretations of the Riemann-Stieltjes integral
(or, simply, the Stieltjes integral) and a few of the associated theorems as given in [1,
chapter 6]. For convenience, the definition of the integral is reproduced, and the
theorems that are interpreted here are stated without proof.

2. Definition of the Integral. Let [a, b] be--a given interval. Define a partition P
of [a, b] to be a set of points xo, xl x. , where a = xo < X1 < ... < xn =b.
Let a be a monotonically increasing function on [a, b]. For each partition P of
[a, bI write LAak = a(Xk)- a(Xk-l)-
For any real function f that is bounded on [a, b] let Mk = max{f(x), Xk-l <
x < Xk}, mk = min{f(x), Xkl <x <xk}, and set
n

U(P, f, a)= MkAak,


k=1

This content downloaded from 143.107.231.30 on Wed, 31 Jul 2019 21:18:18 UTC
All use subject to https://about.jstor.org/terms

You might also like