- The Law of Quadratic Reci-procity An Honors Thesis ( ID 499) By Robert C. Matney Thesis Director Dr. Padmini Joshi (j) £ ·f fJ· __~J_c~L~~_{_I~ __ ~_~_~~(~4:/~·~~_~~________ (advisor's Ball State University Muncie, Indiana May, 1977 signature ) One of the most oriGinal, and nerhaps one of the rr~ost consenuential, discoveries in the eichteenth century in the field of number theory is the Law of (o~uadratic Heci proci ty. In the lanGUage of the period, if there exists an x such r) that x~-p is divisible by a, then n is said to be a Quadratic residue of a; if no such x exists, p is said to be a ouao.ratic non-residue of q. The Law of Quadratic decinrocity states that, if n and a are odd nrimes, then: In (i) each is a residue or non-residue of the other if at least one of them is of t1:1e form 4N+l, and (ii) is a residue or non-residue of q according as n is a non-residue or residue of p when both are of the form 4N+3 180~\, D Lef!'endre develo;)ed a symbol to describe the rela- tionship between nand c. The symbol is defined as follows: J1'or any number p and any odd nrirr.e q, (E) q ={ 6 -1 if "D is a ("1uadrrltic residue of q if a divides p if 1) is a ouadratic non-residue of q Stated in terms of the Legendre syrrbol, the Law of Quadratic Reciprocity becomes: For any distinct odd priITles n 8.nd q, 0-1 a-I (2)(~) = (-l)~ ~ CI P In 1783, .:":uler pave (C"ryroof of this theorem, but he made an 8.ssUIDlJtion which, al tho'J.,f"h true, w[-:s not 1)roved to be valid until rruch later. LeFendre also fave a "Droof, but he too made certain 8.ss1.lIT'Dtions which had not been nroved. 2 In 1949, however, the assumptions made by Euler were proved by elementary means, and his proof can now be considered to be complete. The first complete proof of the law was given by Gauss. He accomplished this in 1796, when he was only eighteen years old. Of this proof, Gauss wro:te: "For a whole year this theorem tormented me and absorbed my great.est effoTts until, at last, I obtained a proof • • • He was so fascinated by the result, which he called the "gem of higher arithmetic," that he went on to give six more proofs. Since Gauss's time,. the theorem has been proved in at least fifty different ways. A large number o·f these are given by P. Bachmann in his work Niedere Zahlentheorie. In this paper, we shall consider seven of the numerous proofs. These are: Gauss's First Proof Gauss's Third Proof Gauss's Fifth Proo·f Lange's Second Proof Frobenius' Proof Eisenstein's Proaf Lehmer's Proof We shall also consider some interesting resul.ts which follow from the discussion of the theorem. Some of the proofs have been altered slightly to provide for consistent notation throughout the paper. Before we can examine the proofs, we will need certain preliminary results. Many of the proo;fs rely on Gauss's Lemma, so we shall first consider this very important lemma. G. Archibald. 2 The proof given here was presented by Ralph 3 Lemma 1 (Gauss t s Lemma): Let p be an odd prime and let (p,q) = 1. If u denotes the number of least positive residues ~reater than ¥ modulo p, of the P;l integers in the set p-l) q, 2q, 3q, • • . , ( 2 q (A) then PROOF: Since (p,q) = 1, none of the intef'ers in (A) is congruent to 0 and no two belong to the same residue class, modulo p. Let be the least positive residues greater than ¥' and let b l , b 2 , by • • • , b v be the least nositive residues less than ¥. Then p-l u + v =2 Each of (B) is positive and less than ~ , and each is in a different residue class, b. J modulo p. =p-a. 1 for some j such that Suppose that (mod p) 1 ~ j ~ v 4 Let a. and b. be the least residues, respectively, of Aq J ~ and Bq, modulo p. o :: a. ~ -Then + bj ~ Aq + Bq Since A and B are positive and A + B. (e) ~ (A + B)q '0-1 S ~ , p does not divide Consequently, the P'21 integers p-al , p-a 2 , • • • , p-au ' b l , b 2 , • • • , b v '0-1 and are, therefore, an arrangeare distinct numbers ~ ~ 2 ment of the integers 1, 2, 3, ..., p-l 2 The product of the integers in (e) is • • • Also, .. (a) ( 2q) • (D) modulo p. From (e) and (D), we f;et '0-1 (P;l)! (q) 2 Ii (-1) u(P;l)!. (mod p) and so p-l (q) 2 _ (_l)u (mod p) Since p-l (9.) _ (q) 2 p We get (_l)u (mod p) 5 and Gauss's Lemma is nroved. Using this result, we can develop the followinp lemma, which is necessary for Gauss's Third Proof and Eisenstein's Proof. Lemma 2: If P is an odd prime and q an odd integer such that (p,q) then (9.) P = (_l)M where PROOF: If k is :in integer {Teater than 0 then kq = nQ + r where o and Q is an ~ r < p inte~er. Thus So rJ [-kqpJ = [ Q + P since = . Q =1 6 Then = p[~] q = p[~] 2q (I) + r1 ..• • • • • • • (P;l)q + r2 • .• = p[(P;l)q] + r p-1 2 where r.< 1. O~ p and 1 ~i :s; n-1 2 Let P =1 A = + 2 + 3 + • • • + P;l 2 =~ = and let • • • + au B = • • • with the a. 's and b.'s the least J 1. in Gauss's ~ositive residues as given Le~ma. If M = [~] + [~] + • • • + [(P;l)~] then =A+ B from Gauss's Lemma. (1) Pq Since = By adding the eouations in (I) we e-et pM + A + B 7 is an arrangement of the integers 1, 2, 3, • p-l . · , 2 we see that (2) P = up - A + B Subtracting (2) from (1), we = p(M p(q - 1) p(q - 1) s p(K - u) ~et - u) + 2A Thus (mod 2) Since p and q are both odd, q - 1 - 0 (mod 2) p -1 (mod 2) - u (mod 2) and (3) becomes 0 :: 1,~ or M == u (mod 2) and thus By Gauss's Lemma, Therefore which corepletes the lemma. The final lemrra which we will prove is necessary for the discussion of Gauss's First Proof. The proof given here follows a presentation of Dirichlet and Dedekina. 3 8 Lemma 3: If 0 = 4M + 1, there exists an odd nrime pt < q for which c}) = -1 PROOF: He consider two cases; either q = 8N + 5 or q = 8N + 1. Case 1: If q = 8N + 5 then 0+1 == 3 (mod 4) ~ can be Consequently, since not all Tlrime factors of q+1 2 congruent to 1, mod 4, q+l must have a prime factor 2 pt - 3 (Irod 4) That is, q + 1 _ 0 (mod p') or q _ -1 (mod u') So "0'-1 (~) = (~) = C-l)~ = -1 p P since p' =3 (mod 4). Therefore (~) = -1 p Case 2: Let a = 8N + 1. AssuJYe q is a quadratic residue of every prime which does not exceed 2m + 1 < q. Then q is also a quadratic residue of every positive interer which is a product of numbers, each ~ 2m + 1. 9 If • • 2m(2m + 1) then x 2 - q solution. (mod M) 'Then (k,M) is solvable. =1 Let =k x be one a.nd • • (A) -- k(k 2 _ 12 )(k 2 - 0 (k 2 .' .' 22) • m2 ) (k + m) (k + m - l)(k + m - 2) • • • (k - m) (mod M) since the product in (A) is divisible by M. =1 since (k,M) M = Therefore, and (m+l) (m+l)2 - 12)(m+l)2 _ 22) • • the product (B) a == g _ 12 -1 . (m+l)2 _ 12 m+l q - 22 • • (m+l)2 _ 22 • . q - m2 (m+l)2 _ m2 is an inte{':'er. On the other hand, a is not an integer if if m <J<f<m + 1 m2 then < q m = [JCf] since < (m + 1)2 and all of the fre.ctions in (B) are proper fractions. Since c ~ 17, we see that 4q < q2 - 2q + 1 8 < (q - 3)2 = and (q _ 1)2 Thus 2 ro: < 2m + 1 q - 1, < 2 vq + 1 < q , and (when q = = [Vcr]), q Thus, for m, our orip'inal assumption is invalid. if m 8N + 1, there exists a prime p' < 2¥iQ + 1 Consequently for which 10 ,From these two cases, if prime p ,< q (~) p q = 4M + 1, there exists an odd for which = -1 which completes the ~roof. With these results, we are now ready to consider the reciprocity law. We shall first exarrine Gauss's First Proof, which he presented in his Disquisitiones Arithmeticae. 4 This proof uses mathematical induction to achieve the desired results, and is quite comnlex in nature. Gauss considered eight cases, but these can be combined to f'i ve rise to two mutually exclusive cases. The proof given here follows the presentcdion {'i ven by Dirichlet and Dedekind. 5 GAUSS'S FIRST PROOF Sirce we will be using mathematical induction, we will uroceed as follows. If the theorerr. is assumed true for all pairs of odd primes less than a prime q then it will be shown that the theorerr Vlill be true for a cOJTlbined with any prime less than q. Since (~) = (~) = -1 we see that the theorem is true for the tvvo smallest odd primes. In the course of the proof we use the fact that, on the assurr.ption that the theorem is true for every pair of odd primes less than a, the statement - P-l Q-l (1) (~)(~) = (_1)-2-"""2 11 will be true for all odd inteFers P and Q such that (p,Q) and all prime factors of P and Q are less than q. =1 The Law of quadratic Reciprocity will be called the Theorem, and statement (1) will be referred to as the Generalized Theorem. The p-enera1ized theorem is proved by using the truth of the original theorem for all nairs of odd pri~es one of which is a factor of P and the other of Q. We begin by assuming the theorem is true for every pair of distinct odd primes less than q. we see that, for q = 4M Since + 3, 9-1 (=2) q = (-1) 2 = (E) q _(E) q and either (E.) = 1 (A) q For q = (.=E.) or q =1 4M + 1, q-1 (::E) q = (-1) 2 (E) q = (E. ) q and either (E) (::E) 1 q = q = (B) or (c) These can be combined into two cases: At least one of (E) and (::E) is 1; q q (A and B) 12 (ii) = q Consider case (i) • (E) q and (E) 4M + 1 q = -1 •, (0) Either =1 Or (.=.E) q =1 Or ~rove In order to the theorew, we need to show that p-1 q-l 2 (E) (.9.) = (-1) 2 (2) p q Let w be either 'P or -p, so tha.t (~) = 1. There are two distinct solutions x of the congruence x 2 :: w (mod q) o< x < q such that If Xo is one solution, the other is q - xO. that one of the two solutions that is even. o< e < Let e be Then q and e2 - w (3 ) = fq then w > 0 ,i.e. w = p. Then p - e 2 q (p - e ) , which is impossible. So f > O. If f < 0 and I Consider the two 3.lternatives: First, suppose p~f. From (3) we see that 2 >0 13 and since fq is a Quadratic residue, mod p, we have (~) - (-f-)(~) Iwl Iwl Iwl 1 - and so (';1) = (Twi-) Now 0 <f < q This is obvious if fq o < Iwl < q = > p e2 ~ (q_1)2 + p = q 2 - 2q + 1 + < = q2 - 2q - w When - -p Sinc e (f, w) = 1 , 0 < f , we have the f:enera1ized theorem =P then a(q-1) Iwl-1 f-l , we have 1 w-l f-1 f ) = (-l)~ ~ ( -;w;- = = p and w w q (l!l) = (~) = f IWI If If w (/~' )(lfL) = (-1) (5) 0 • = f < q - 1. 'rhus w -p , from (5) we have 2 ~ < q , a.nd 14 = -w-l f-l (-1)~ -2-(=!) = -\,\,+1 f-l (-1)~ -2- = f -(w+l) f-l (-1) 2 -2- w-l f-l = (_1)2 2 In both instances w-l f-l f) ( 'TWT = (_1)--2- --2- Since e is even, from (3) we have -w - 1 ~ fa - 1 (mod 4) Therefore =f _ W~1 q ;1 (wod 2) Letting f' = -y f-l we and at a-I = 2"" ha.VE! fa - 1 = (2f t + 1)(20' + 1) = 4f'qt + 2(f' + q') and so fa-l ~ = 2f'q' + f' + q' rrherefore w+l - --2f-l + ~ q-l - --2- (mod 2) 15 f' ul ti plying by vv-l -2- , w-l w+l - ""'2 -y - we get w-l f-l w-l <1-1 -2- -y + -y 2'" (mod 2) ~;nce w-l d w+l are consecu t~" ........ ve"~n t ep'ers, 2- an . - 2 w-l w+l is -2-~ even, and thus "'1-1 f-l VI-I a-I -2-~--Y2 (mod 2) Therefore w-l 0.-1 (-1-) = (_1)-2-- ~ (6) w whether w = p or w = -n. The theorem '.Nill be established for p and q if we show that (2) is valid. If w = from (4) and (6) we get p , "0-1 \1-1 (~) = (L) = (-l)~ -y p n Since statement (2) is established for nand q when If w = -p , then since we see that (2) = (-1) q q Then, from (4) and (6), n f (.:.) = (-) = p p -p-l q-l 2 2 (-1) w = p • 16 -(11+1) q-l = (-1) 2 2 2±1 !l::!. = (-1) 2 2 0-1 0-1 0-1 = (-1) 2 2 (-1) 2 """"-- lJ-l a-I ..lo...- ~ = (-1) 2 2 (-1) q lJ-l a-I = (-1)~ ~(£) q = -lJ This gives us (2) when w for case (i) when (f,p) = 1. and corT'l)letes the proof Consider now the instance where f p/f. Then = wf 1 where f1 is odd and f1 < Frorr (3) , we see that q w/e i.e. e = weI , where e is 1 even, and we get -I e 1 2w (7) ConseouHnt1y 1 (8) w{f1 = = flq ." (-f,q) Iw Also (f ,e ) = 1 and we have 1 1 (~) and /.1 r Now tie may, therefore, assume the truth of (1), the e:enera1ized theorem for Iwl and If~; that is, 17 If p w - then fl > and 0 f = (-1) l -1 p-l 2 2 From (8) we pet and (¥:-) 1 =1 Thus = (-1) f l -1 p-l 2 2 Conseauently, f1-l = (-1) 2 :£::! p-l fl+l .E::1. 2 (-1) 2 = (-1) 2 2 From (7), we see that Since fl and are both odd, both cannot be of the form q 4N + 1 nor can both be of the form 4N + 3. =q;1 fl~l Since (E) q = 1, Conseouently, (mod 2) (2) is valid; that is, p-l q-l (~) p If w = = (-1) 2 -p ,then fl 2 (E) q < o. Since Ifll and we can assume that (1) holds for Ifll < q and and p. p 'rhus <q , 18 -f1 -l n-l = (-1) 2 2 fl+1 p-1 = (-1) 2 2 From (8), and 1 = (lf~l) so that = (-fp 1 ) = ( .9.P ) (lfi-) p fl+1 p-l 2 2 (-E--) If , 1 = (-1) = fl+1 p-1 (-1) 2 2 ( -1 ) = fl+1 (-1) 2 = (-1) TtlT E=! -f1 -1 2 (-1) 2 fl+1 p-1 + fl+1 2 2 2 Since = 1 (!! ) q (E) = (-1) q q we see that (E) q = (-1) q = q-1 (-1) 2 19 As we showed above, fl+l _ 0-1 (mod 2) 2 2 Thus (-4-) = p p-l q-l ~ (.9.) == (-1) 2 2 (-1) 2 p and so p-1 9-1 2 (~) q (.9.) == (-1) 2 p This cOIDDletes the proof of (2), and establishes the theorem for case (i). Consider now the more difficult case (ii). Let q = 4M + 1 and there exists a prime ( q) P'" If (~) q =1 p <q , with (~) p' <q such that = -1. From Lemma 3, = -1 then from case (i) the theorem could be used, and This, however, contradicts our assumption that Thus, and the theorem is valid for pt and 0; that is, P.' = -1. (q) 20 - .E.:=1 (~)(~) = (-1) P q 2 q-1 2 In order to complete the proof, we need to show that, if p < f. q , p p' (~) , and = <%) = -1 -1 , then ; that is, Since ppt is quadratic residue of q, and there exists a positive '3. even integer e such that e 2 _ pp' (10) Here, < q. f First, suppose = fq We now consider three possibilities. (f,pp') = 1. Then, from (10), <ljff) = 1 Likewise, Since (1) is applicable to If I and pp', when f-l "01)'-1 (p§') When f < = (p~,) = (_1)--2-- I 2 0 , Q) <PPT = < f) pp' = (-1) = = -1)( If I) (pp' pPT pp'-l 2 -f-1 "Op'-l (_1)---2- -f+l pp'-l 2 --y (-1) 2 f >0 we have 21 = -(f-l) pp"-l (-1) 2 2 f-l pp'-1 = (-1)~ 2 But, from (10), -pp _ fq Since q =1 (mod 4) 4), we see that "'Op'+1 2 np'+l 2 (mod f _ -pp' (mod 4) and so f-l -2Also, (mod 2) pp' 1 pp'+l 2- and 2 are consecutive integers, so their product is even. Consequently, q-l P12'-l 2 2 =0 == 12E'+l 12E'-l _ f-l 1212'-1 =2" 2 2 2 (mod 2) Therefore (p§t) = q-l 1212'-1 2 (-1) 2 In the first instance, then, Second, suppose that where (fl,p) = 1. p'l f p'e 1 2 _ p . Consequently, since -. (11) gives us (9) is true. and Also, we have (10) gives us (11) =1 = (f,p) e = = 1. Then f = p'f 1 p'e 1 ' and equation 22 p'e 2 ( If 11 ) = I that is, ~ (If .. 1 ) 1: ., ., =1 (~) p Now, (p§ i ) Since (1) holds for (~) If11 and pp', we have f = (-1) -1 ppt_l 12 2 (~) Ifll -1 pp'-l = (-1) Thus If f1 > 0 ...,'/e have fl-l ppt_1 (~) pp = (-1) 2 f = 2 (~)(~) p P -1 pp'-l (-1) ~ 2 (E':)(g.) p p Since the theorem is valid for p and pI, f -1 ppt_1 pt_1 p-1 l (~) = (-1) 2 2 (-1) 2 2 pp = If fl < 0 fl-l pp'-l + pt-1 p-1 (-1) 2 2 2 2 we h8.ve 2 2 23 -f:±-l E"P'-l 2 2 = (-1) (-~) (P%r) (~) ( T ) (f~1) 1')-1 -f -l pp'-l 1 + -2 + ~ 2 2 CE.:) (~) 2 = (-1) p P p-l -f1 -l pp'-l + -2+ lcl~ 2 2 2 2 = (-1) -f1 -l EE'-l E-1 E'+l + 2 2 2 2 = (-1) From (11) , we know that fl = -~ (mod 4) and fl;l _ "0;1. - -lJ f q (mod 4) 1 ., that is, (mod 2) Therefore, f -1 pp'-l p'-l = p+l E2:!. (PE'-l + 2 ~2+22-2 = P;l.p ,.E;l _ 1 = P;l P;l =0 E.2) 2 (mod 2) Since fl+1 -~ 2 := 2 (mod 2), fl+l 1')1)'-1 ;. + p-l p'+l 2 2 2 2 ~ = p-1 2 - PE'+P' 2 In both cases, - P fl> 0 (p§') = = p-l 2 EP'-l + p-l p'+l 2 2 2 - ,.p-l p+l = p-1 p+l 2 2 2 2 - 0 or fl < 0 (mod 2) , 1 and (9) is true. Since we did not use any slJecial "OroDerty of p', we can interchange p and p' to nroduce the same result when --------.---- - ..- . - ----- prf 24 and ( f , 1) ,) = 1. Finally, let np'lf; that is, f = p~'f2. Then and nn'e ~ 22 - 1 = (12) c From this, and -f 0) 1= ( ~ Thus Since (1) holds forlf 2/and 1)'0', we have If 2 1-1 npt-l = (-1) 2 2 \vhen f:2 > 0, (p§') = (-1) pp'-1 + f -1 YJp'-l 2 ~ 2 f2+1 P12'-l 2 = (-1) 2 V'Ihen f2 < 0 , (p?) = (-1) 'Op'-l + np'-l + -f -1 P12'-1 22 2 2 2 f2+1 P12'-1 2 2 = (-1) From (12), since f 2 a s f2 ~ -1 (mod 4), ._--_.__ ...--- .... _------.... e = PP'e 2 25 ,- f2~1 _ 0 (mod 2) and therefore f2~1 Pp;-l _ 0 (mod 2) Consequently, (p§d = 1 and (9) is valid. Irhis comuletes the proof of (9), and establishes the theorem for case (ii). Since in both cases (i) and (ii), the first assu.IT:11tion is true, and since the theorem is true for the two smallest odd nrimes, by com11lete mathematical induction, the theOreI11 is tru,e for all odd 1)rimes. 'rhis comnletes (}auss' s First Proof. Gauss recopnized ths,t this first proof was extremely corr.nlicated, and he continued to work on the theorem. develoned six more proofs during his lifetime. will be discussed here. Gauss's 'rhird Proof. by Harriet Griffin. 6 Two of these The next proof we shall discuss is The exnosi tion followed here is p,;i ven (}AUSS'S 'rIURD PROOF Let 1') and q be distinct odd primes, with q less than 11. Let [%J + N= [*J + lVi ,- = He [~J + • • • [~J + [( + • .' • + P;l )%J [(q;l)~J 26 Then 2::l q-l 2 (2) (-9.) = (-1) q p 2 PROOF: If k then = is [SJ p 1, 2, 3, .- • • p-l , 2 Also, not an integer. [%J = 0 since , q <p and [~J ~ 1. p If [~J 80 = s that with kq = sp + r o< r < p then (k+l)q = < 1 sp + r + q But r+q p Therefore s < - (k+l)q p < 8 + 1 Further, (n-l)q 2p -~----------,--- = = pq - p + p - 2p q 27 = = since TJ a - < 2p [~J = • n(q 2p 1) + 1) - Ct 21) q = -2 1 + p - q 2p q-1 2 Assume that, for s [(k+~)gJ and k <p-1 2 = , s + 1 'rhen !f9. p < s + 1 < (k+1 )..9p 'rhus k < (s+l)p <k q + 1 'rherefore, is the last term of IV; such that [k~J = s '1:'hen the 12st terrn of Rnd I~" that has the value s-l 30, for the number of terms having the value 0-1 """'2 'rheref ore, 11'1 = [% ] + [~J + • • • + [ ( P;l ) ~ ] is is [S~J · 28 = O-([*J - [~J) +l-([~J - [*J)+ + + == s.( [(S+~)PJ - [~J )+ q-l (n-l [(¥)*J) 2 ••• 2 • -[~J - [~J • • • - [C"!21 )~J + CJ 21) (q;l) == _ N + (0-1 ) ( q -1 ) 2 2 Therefore, M + N By Lemma 2, and 'rhus p-l q-l = (-1) 2 2 and so which nroves the theorem. Eisenstein's Proof, which will be nresented later, uses an annroach which is similar to this proof. His proof, however, makes use of a geometrical interpretation of the numbers Mana. N. - The next nroof 'Nhich we shall consider is Gauss's Fifth Proof. The nresentation piven here is an adaptation 29 - of the proof y:hich W2.8 r'i ven by USDensky and Heaslett. 7 The notatijn has been altered for consistency. G.~USS' S :PIFTH YiWOF Let p and q be distinct odd primes. 'rhen -0-1 0-1 (~)(~) = (-l)~ ~ q p PROOF: By Gauss's Lemma, and where .~ is the nlJ.Ir."ber of elements in the set (A) q, 2q, 30, • • whose least Dositive residues modulo ~ 2 , are greater than and v is the number of elements in the set p, 2p, 3p, • • • , (a;l)p whose least positive residues moclulo %. p C1 are greater than Then The least uositive residue of any nurrber modulo p is either o (f) (f' ) or one of the numbers 1, 2, 3, n-l ...,~ . . . , p-l 30 -. Similarly, the least T)osi ti ve residue of any number modulo a is either 0 or one of the nrunbers (F) 1, 2, 3, (Ft) q+l 0+3 ~ 2 ' 2 . . . , q-l . . . , 0-1 ' 2 Correspondingly, the numbers (c) 1, 2, 3, • .., none of which is (livisible by p and be distributed in eight classes Class 1: 9.S Class 2: follows: Let their nUJl;ber be a. Those numbers whose least nositive residues mod p arEl in (f) and mod q in (F'). Class 3:: Let their nl.unber be b. Those numbers whose le8st ';osi ti ve residues mod 1J on. are in (f') En<1 mod q in 4: Those nlmbers whose least Let their number be g. ~ositive c:cre in (ft) ,nd mod a in (F'). Class 5: mod simul tsneously, C2n '1'hose numbers whose least Dosi ti ve residu.es wod p are in (f) <ind mod a in (F). Class (l Those multi 1)le8 of p are in (ft). (1 residues ITod D Let their number be d. wllose lecwt -positive residues '1'hese 8,re the elements of (A). Thus their nurr:ber is u. Class 6: Those multinles of mod. p are in (f). Class 7: whose le8.st posi ti ve residues . n-l Their number lS ~ - u. Those multi1Jles of mod a are in (F'). r 1J whose least positive residues rI'hese Cl.re the elements of (B). Thus their number is v. Class 3: Those multi ')les of p whose least nosi ti ve residues mod q are in (F). Their nurrber is 9-1 - v. --------------_ -._---------.. 2 31 4, and 7 comprise all of the numbers in (0) whose Classes 2, (F'). least Dositive residues mod q are in (F'), residue in If r is a given these numbers are r, q + r, 2q + r, • • • (P;3)q + r and the ineQuality or tq is true for < (p-3)q - t 2· n-3 =~ + (q + q-l _ r) 2 but not for p-l t = -r · Since r can n-l values, and each r gives p-l take on ~ 2 numbers, the number of members in Classes 2, 4, and 7 is number is also ( JJ) b + d + v. So ~ and q in the discussion above, we get the numoer of members in Glasses g + d + u 3, 4, and 5, which is = p-l .9..::! 2 2 To each number x in (C) which is also in Class (D) But this p-l ..-.a-I b + d + v = .-....2 2 By interchanging (2) p-l.q-l 2 2 • l?9.:!:1 E9.:!:l 2 ' 2 ' ..., 3 in the set pq - 1 there corresponds the number pq - x whose least positive residues mod p are in (f) and mod q in (F'), and vice versa. Therefore in Class 3 there are exactly as many numbers as there are elements of (D) whose least ~ositive residues 32 (F'). mod p are in (f) and mod a in -. Sets (C) and (D) together form the set 1, 2, 3, • • • , pq - 1 (E) Consequently, the number of members in Classes 2 and 3 is the same as the nurr:ber of elements of (E) whose least positive residues mod p are in (f) and mod q in O~,). But to any pair of such residues there corresponds a unic:ue number in (E), and since the nUIr:ber of pairs is P;l q;l , there are the same number of elements in Classes 2 and p-1 q-1 b+g=T 2 Adding (1) and (2) we get b + g + u + v + 2d = 2(P;1 So p-l 0-1 + u + v + 2d 2~ or u + v + 2d = p-1 q-1 2 2 Thus _ p-l q-1 u + v = ~ ~ ( mod 2 and so (_l)u+v By Gauss's Le~~a, = lJ-l a-I (-1) 2 2 ~~ ) ~) 3. That is, 33 Therefore and the theorem is proved. These three proofs by Gauss show the fascination which the theorem held for that ewinent ma.thematician. Gauss was not the only scholar to devote considerable effort to proving the theorem. of three The next proof we shall consider is the second ~iven by Lange. due to Barnard and Child. The diagrammatic illustration is 8 LANGE'S SECOND PROOF If p and, unless 0 are unequal oda, primes then p ~ q ~ 3 (mod 4) , in which case (~) = _(So) q p PROOF: Without loss of eenerality, we can aSSUJIle that p < q. Consider the sets . .' A = {q, 2q, 3q, • , (P;l)q} B = {p, 2p, 3p, • • • , (q;l)p} represented by the uoints of division of two scales set edge to edge. 34 - q I f 1? sp - -- I p (¥)q rq 2p - - I (Q;3)p (s-l)p t 1 (Q;l)p Let u be the number of elements in A with negative absolute < q , between two consecutive members of B. If least residues, modulo D. (s-l)p < rq Since p each rq in A lies < sp with and If we let rq, 1 .s: r ~ p-l 2 (s-l)p, and sp be the points A, B, and C, respectively, then IBAI and ICAI are the least positive and numerically least negative residues of rq, modulo p. Thus the absolute least residue of ra, modulo p is neggtive if ICAI < IBAI ; sp - rq that is, if 1 < '2 P So u is the number of pairs of positive integers (r,s) such that I s < ~ o< Let v bE~ sp - rq < I 2P the number of eleTI'ents in B 'ltvi th negB-tive absolute least residues, modulo q. Let (P;l)q and (q;l)p be the points X and Y respectively. 35 Then = ( q-p) 2 < £..2 So, <% IXYI IXY I Therefore, reuresents the absolute least resid_ue of (q;l)p , modulo q, and is positive. then IXyl Also, if is the absolute least residue of k-p, modulo q. Then for any sp in B with nerative absolute least residue, modulo q, there exists an r such that (r-l)q < sp < rq with 1 < r < p-l - 2 and - Therefore v is the number of pairs of positive integers (r,s) such that r 1 <2'P 1 s<~ o i.e. < rq - sp 1 <~ 36 - 21 q < sn - rq < 0 Since stl - rq i 0 u + v is the number of pairs of interers (r,s) such that r < Ip 2 1 (c) s <~ - 1 ~ < stl - rq < 12P Let r + rt and. s + st Since o 1 < r < 2'P 1 O<s<~ we see that o 1 < rt < 2P o < Sf < ~q Also, -----------,-- ---------,----- 37 -. r' s = (q+l) - s' 2 and the third inequality in (0) becomes - 12 < (Cl+l) - s' 2 (p+l) p - ~~ 2 - r' q <'21 P or - 1 -q 2~ Adding (q;P) _ t < 1-p 2 1 - s'P - '2 Q + r'q < ~p we get ~P < -s 'p < ~q + r' q Multiplying by -1, we have 1 - '2 q < s' p - r' q < 1'2 P Thus the pair of integers (r',s') satisfies the inequalities in (C). But (r,s) = (r' ,st) only when r = r' = p+l 4 q+l s = st = 4 That is, only when p - q - 3 (mod 4) Thus u + v is odd only when otherwise. p From Gauss's Lerr.ma, and - q - 3 (rr.od 4), ancl is even 38 Thus (D) Tiul t i plying by (~), we Fe t But (.9.) (.9.) = 1 • p p Thus And, since u + v is odd only when (l?) q unless = p _ q _ 3 (mod 4), (.9.) p p:: q _ 3 (mod 4), in which case This ,roves the theorem as it was originally stated. To show that this is equivalent to E.=! (E) (E)(.9.) = (-1) 2 q P q-l 2 we use the fact that, if p ~ q E 3 (mod 4) 1')-1 q-l product ...........=- is odd, but not otherwise. 2 2 then the Thus statements (D) and (E) are eouivalent, and the two methods of stating the theorem are the sarre. The next proof which we shall consider was given by Frobenius. The rredification nresented here was given by Daniel Shanks. He feels that this proof is possibly the simplest and most direct ureof of the Law of Quadratic Reciprocity.9 39 FROB1~~HIUS t If p = 2P + 1 and q = 2Q PROOF + 1 are unequal odd primes then PROOF: Assume that p < q. Let a be an odd integer, 0 <a < p such that qa = pat + r where and r is even. If u is the number of such a, then from Gauss's Lemma, We know that a, p, and q are odd and r is even. must a1:30 be odd. a, and 0 < a' < q • where v is the number of odd a' such that + r with <a <q 0< r < o<a <p o l Q and a oC,d. l Further, at is unioue1y determined by Then, b:r symrnetry, pal = qa Thus a , 40 Consider the function = qa R(a,a') pa' where a is an element of the set 1, 3, 5, • • • , p - 2 and a' :is 8.n element of the set 1, 3, 5, Ne l{now th8.t 0 .., • q < r <"p. o < pa' o < q 8. - 2 'rhus <p + r - pa' or - 1)a' <p and so O<R<lJ Since there are u values for r, there are also u values for -q <R < < -r < 0 -q < C1 a R such tha.t 0 Similarly, 1). 0 < r < q. So -q < R < O. 3ince or - Cia - r <0 which rreans that -q < qa - (q a + r) -0 < qa - pa t <0 -q <R < 0 <0 and so that Thus there v values for R such thgt (a,p) =1 , D and q are nrimes, and a <p, qa ~ pat so th8t 41 oa - pa' I. Thus ° R I. 0, rmel there are u + v values of R such that -(1 <R <p If then are two such R. Therefore, Rl + R 2 = = q (a l + (:12) - = q(p - = ap - + qa 2 a~ - pq So Rl + R2 = P - q If, however, a 1 = a 2 = p-l = p 2 and a-'1. ' -- a2' -.9.::12 - Q- then Pend Q are both odd, ana R = qP - - p(~t + - p(a - 1) 1) C1 ,) (pal' + na - 2 pQ = q(P;I) _ p(O;I) + p a ') 2 42 =E -2 9-2 = (E-l) = P - Q R = P - Q ~(1-1) So Thus u + v is even unless P c'nd Q are both odo.. is even unless P'J.nd Q a.re ooth odd. But PQ Therefore But 'rhus \F;hich -proves the theorem. (2::1) 2 = P Also, since and (2.::1) 2 . -- Q we see that 8 .. nd the~le t"vo foms of the theorem are erui valent. We shall now consider two geometric[:l Droofs of the theorem. The fir8t of these, by Eisenstein, provides us wi th a {,;eometrical inter1)retation of Gauss t s Third Proof. The variation ~iven Harriet Griffin. need the followin~ here is based on a presentation by Before we begin the proof, ho.,.;ever, . .ve definition: 43 Lattice Point: A tloint (x,y) is called. a lattice point if and only if x and yare both integers. Ne can now give the "9roof as stateo by Griffin. EISENSTEIN'S PROOF Let p and q be distinct odd primes. Then From Lemma 2, and where rf( = [~J + • • • + N = [~J • • • + + Then Consider the rectanr;le R with vertices 0(0,0) B(E 2 ' 9.) 2 C(o , ~) 10 44 0(0, ~) 9.) 2 ' 2 B(E (0, q;l) (P;l , 0) 0(0,0) The diagonal OB has eauation Y = -pO,x · ( q,p ) ,andSlnce = 1, there are no lattice 1')oints on the open line segment OB. Let Tl be the interior of R below OB, and let ~2 be the interior of R above OB. Consider a point (k, 0) on the x-axis, "'lith k an integer. For i.e. < k -< p-l 2 1 - the number of lattice points directly above the point in Tl is [~J. Thus the number of lattice uoints in Tl is Like'llJise, for j an integer, °< ,- j < 9.2 i.e. 1 < j - < 0,-1 - 2 the number of lattice noints in T2 directly to the right of the point (O,j) is [~J. Thus the number of lattice 'Points 45 [~J + [~J + • • • [(q;l)~J = N Hence the total number of lattice points in R is li: + N. '0-1) (a-I) But there are also exactly ( '2 2" lattice points in R. Thus, Nt + N = p-l q-l 2 2 Therefore, which ryroves the theorem. The final nroof of the theorem which we shall consider was -presented by D. H. Lehmer in 1957. It is, therefore, one of the most recent proofs available. The vRriation given here was presented by George Andrews. 11 LEJmmR'S PROOF If panel q are distinct odd primes, then (E) q unless p :: q (12 ) q ) = (£. p - ") ..) (mod 4) , in vthich case = -C Sp ) PROOF: Let u l clenot the number of inteeers in the set q, 2 q , 3 q, • • • , ( P;l ) q 46 with least 7)ositive residues greater than ~ , modulo p. "- Let u 2 denote the number of integers in the set p, 2p, 3p, • • ., wi th lea.st nosi tive residues greater than ~ , modulo q. Then, by Gauss's Lemma, and u (9.) == (-1) 1 p Thus, == (-1) ~+u 2 But if and only if So if and only if To prove the theorem, we need to show that u l + u 2 is odd if and only p:: a :: 3 (mod 4) • Consider the hexagon H with vertices OBCDEF as follows: c(E q(p-l») 0(0,0) 2' E(P(o.-l) .- 2q :1) '2 ----------_._-- 2p 47 E D c o B y = ~ P _L is the ecuation of 2p ox 1 y = .;.:.: p + -2 o is the equation of EF. lie~ If the point (x,y) BC~ <x <~ in H then o <y < ~ and (A) ~_L<y<9.!.+l p 2p . P 2 Let (a,b) be a lattice inside H. (P;l _ C~, 0,;1 _b) o Then the uoint is also a lattice noint, for if <a < ~ then 0> -a >- ~ and so tha.t --------------- 48 or o < 12±1: 2 - a But since a is an integer, p+l - n a <_p-l< .:. 222 Thus Similarly, if o <b <~ then A.lso, if we have -b > _~ _ 1p p and 0+1 ~ - b > 9+21 _ ~ _ 1 p 2 which Gives us or q+l _ b 2 > 9..E ~p _ S@:. + .9...... D ~D _ L 2n 49 in other V'lOrds .9.±! _ b (1) 2 > 9.P (p+l 2 - a) _ !L 2p Likevvise, if then -b aa +-"-(1 _ ........ < P 2p and oa - -p + .92p so 0+1 ~ - o Qa b< 1)(q+l) +-----2p 21) P and we have 0+1 _ b < pCHn+g q+l b < Q(p+l) 2p b < P9. ( p+21 2 -naa 2p or ~ - + ! - C18. 2 p- so that (2) S1.±! _ 2 - a) + ~ Co Thus, from (1) and (2) we have O<.2±.!-a<E 2 2 s. (p+21 _ a) _ 1) 8,nd %-< 9.±1:. _p 2 _ b < 9.P ( p~ 1 - 8, ) + ~ Thus for each lattice T)oint (8" b) in H, there corresnonds - " " t a 1 8 tt lce nOln (p+1 2 - a '0+1 2 - b) also in ~{. 50 Further, = (P~l (a,b) _ a,q;l - b) if and only if a -- p+l 2 - b a =(1+1 2 - b so ths.t = 2a -p+l 2 and 2b - = 0.+ 1 2 and 4b = a + 1 and 4a = p + 1 or, in other words, p ~ q ~ 3 (mod 4) Thus _ ( 8,b ) ~/ (p+l 2 lJ.nl e s s p _ a :: 3 3., 0+1 _ b) 2 ( m 0 c1 4). But if p ~ q - 3 ( mod 4), in addition to the nairs _ ~ q+l _ b) ) , (~ ( a, b 2 '--, 2 "there a -I P:l 8-nd b.j 9.:1 point, namely (P4l,CJ.~1). in H is octet if sr..cl or..ly if , Y/e p'et one more lattice Thus the number of lattice Doints n - Oc =3 (mod 4). Since (q,n) = 1, there are no lattice points on ODe (2", b) is a lattice noint below OD, then aa p and If 51 ~ C 8. - < bp < Cia Thus - % < bn AdelinE J, <0 we pet %< bp (B) qa - q (a - 1) <q Thus bl) has a le2st I)ositi.ve residue f;reater than %, mod Q. Conversely, if on has a 1e8.st nositive residue greater than %, modulo holds. 0., then we can find 2-n integer a such that (B) But bp is<'n element of the set ..., (1-1) «-2n- and has least Dositive residue greater than There c;,re u %, modulo o. of these integers, and thus there are u')c.. lattice 2 points in H below OD. Similarly, we find u there ;:::,re u + u l if and only if 2 l l8.ttice points in H above OD. lattice "':Joints in H, and u l + u 2 is odd n:: q == 3 (mocl 4), Yvhich Droves the theorem. Thus (E.) == ( .9. ) q p unless p - (' - 3 Thus (mod 4) , in which cO.se (E.) == -(~) 1) q Wi th this nroof, we com''"llete our eJ.iscu8sion of the Law of Quaclratic Reci T)roci tv. There 2.re, however, three basic results which nrise from a disc1)ssion of the Legendre Symbol. The proofs of these corollaries are Quite si~ple, statements theJ'YIsel ves [-.re very important. The first -oroof we shall give deals Vii th the n8.ture of nuadr8.tic The second 2.nd third are concerned "'d th the Corollary 1: =1 then if and only if p-l r2 (mod p) _ 1 P;iOOF: If R is a primitive root of p, then p-l p-l 1)(R2 (R~ + 1) _ 0 (mod p) and thus p-l R 2 _ xl (mod p) If r - R2k (mod p) then -0-1 r~ = R(P-l)k _ 1 (mod p) ---------------- .. resid~les. 0u8.c1r~ltic of -1 and 2 resnectively. If p is an odd nrime and if (r,p) yet the _•..._._---- n8.tv.re 53 and we see that --- n-1 r 2 1 (mod n) 1 (mod p) SO (E) P 1 = if and only if p-1 r 2 - Corollary 2: If p is an odd prime then (-1) = 1 P if and only if (mod 4) p =1 PROOF: ;{e lcnow thet p-1 (-1) - 1 2 p if and only if - 1 (mod n) ( ffiOO_ 4) • Thus, by Corollary 1, (-1) = 1 p if ana. only if p-1 (-1) 2 - 1 (moel p) that is, if and only if - n :: 1 (mod 4) ,-------_._----_._-------- 54 Corollary 3: ..;....;;.~=--.~ for any odd nrime p. PROOF: Je ltnOVl that ¥- 2J [ p = 1 [~J = 0 Therefore, if k <p;l then [2~J = .,. 0 In Lernm9. 3, therefore, if then = 2 G M = 0, 20ncl in Dlace of congruence (3) we h8ve P(2 - 1) E 0 - u (mod 2) Thus p _ -u _ u (mod 2) cmd so Cl.nd These corollaries, together L8,W, fo:nn r:n imDortcmt neTt of ~ith t~le the )uadrRtic Recinrocity study of (lu8.dr:::tic residues. 55 The pro lJer 2.nplication of these DroY)erties cc::') 'lrovic1e mp.ny solutions to difficult nroblems. ized to Give the Fin;t Proof. ~}ener8.lized The theorem can he general- L8W, which \,,'e saw in G,auss's The properties can be used to determine the odd primes of which a given inteeer is a nuadratic residue or non-residue. limi tJ_ess. 'fhe ap"Cllic8.tions of the theorem are 21,lrrost This is why the Law of Juadratic ileci'':lroci t~T has occunied the attention of so many ereat mC'>the):laticians. The proofs -presented here e,re only 2, small the innumerable Y)roofs, botl'l elementary ano, theorem. s8JYl~llin~ cOTr~plex, of of the The best exY)osition on more of these proofs is the boolc by Bachmann which was noted e8,rlier. The other refer- ences given at the end of thts pS:lJer 2,re also of special interest to any study of this tyne. 56 .FoorNOTES lCarl Friedrich Gauss, DisC1uisitiones Arithmeticae (New Haven and London: Yale University Press, 1966) p. 87. 2Ralph G. Archibald, ~ Introduction to ~ Theoa of Numbers (Columbus, Ohio: Charles E. N;erril-PUblishing Co:), pp. 135-137. 3r-oid., -op. 277-278. 4 G~uss, pp. 89-98. 5Archibald, pp. 272-281. 6Harriet Griffin, Elementary Theory of Numbers (New York: McGraw-Hill Book Comnany, 1954), np:-146-l47. 7 J .. V. Us-pensky and !'Il.A. Heaslett, Elementa§if Number Theory (New York: KcGraw-Hill Book Company, 1939, pp. 284293 •. 8 S • Barnard and J .IVl e Child, Advanced Al~ebra (London: and Company, Limited, 1939), pp. 14 -149. Mc~l.lillan 9Daniel Shanks, Solved and Unsolved Problems in Number Theory (Washington: Spartan Books, 1962), pp. 40-47: lOGTr1Off"1n, pp~ 144- 145 • 11C~eorge E. Andrews, Number Theory (Philadelphia: VV.B. Saunders Company, 1971), P~. 120-123. 57 -BIBLIOGRAPHY Andrews, George E. Number Theory. Saunders Company, 1971. Philadelphia: W.B. Archibald, Ralph G. An Introduction to the Theory of Nwnbers. Columbus, Ohio, Charles E. ~erril Publishing Company, 1970. Bachmann, P. Niedere Zahlentheorie. Leinzig, B.G. Teubner, kathematische ~issenschaften, 1902. Barnard, S. and J.M. Child Advanced Al~ebra. I\~cN:illan and Compe,ny, Limited, 193 • London, Gauss, Carl Friedrich Disquisitiones Arithrneticae. (Translated by Arthur A. Clarke, S.J.), New Haven and London, Yale University Press, 1966. Griffin, Harriet Elementary Theory of Numbers. New York and London, W:cGraw-Hill Book Company, 1954. Shanks, Daniel Solved and Unsolved Problems in Number T1:.eory. Washington,-Spartan Books, 1962:Uspensky, J.V. and ~.A. Heaslett Elementary Number Theory. New York and London, McGraw-Hill Book Company, 1939. -