7 sin The First 50 Million Prime Numbers* Don Zagier I - for ~F(s) s sin -s ; the a l g e b r a i s t it is "the c h a r a c t e r i s t i c of a f i n i t e field" or "a p o i n t in Spec ~" or To my parents "a n o n - a r c h i m e d e a n valuation"; the c o m b i n a t o r i s t d e f i n e s t h e p r i m e n u m b e r s i n d u c t i v e l y b y the r e c u r s i o n (I) Pn+1 = 1 [I - log2( ~ + n r~1 (-1)r ISil<'~'<irSn 2 pi1"''pir ([x] = b i g g e s t )] - I integer S x); and, f i n a l l y , the l o g i c i a n s h a v e r e c e n t iv b e e n d e f i n i n q the p r i m e s as t h e p o s i t i v e v a l u e s o f the p o l y n o m i a l (2) F(a,b,c,d,e,f,g,h,i,j,k,l,m,n,o,p,q, r,s,t,u,v,w,x,y,z) = I w o u l d l i k e to t e l l y o u t o d a y a b o u t a subject which, although I have not worked in it m y s e l f , has a l w a y s e x t r a o r d i n a r i l y c a p t i v a t e d me, a n d w h i c h has f a s c i n a t e d mathematicians from the earliest times until the present - namely, the question o f t h e d i s t r i b u t i o n of p r i m e n u m b e r s . You certainly all know what a prime n u m b e r is: it is a n a t u r a l n u m b e r b i g g e r t h a n 1 w h i c h is d i v i s i b l e b y no o t h e r n a t u r a l n u m b e r e x c e p t for I. T h a t a t l e a s t is the n u m b e r t h e o r i s t ' s d e f i n i tion; o t h e r m a t h e m a t i c i a n s sometimes have other definitions. For the function-theo r i s t , for i n s t a n c e , a p r i m e n u m b e r is an i n t e g r a l r o o t o f t h e a n a l y t i c f u n c tion ::The following article is a revised version of the author's inaugural lecture ~Antrittsvorlesung) held on May 5, 1975 at Bonn University. Additional remarks and references to the literature may be found at the end. Translated from the German by R. Perlis..The original German version will also be published in Beihefte zu Elemente der Mathematik No. 15, Birkh~user Verlag, Basel. [k + 2] [1 - (wz+h+j-q) 2 (2n+p+q+z-e) 2 - (a2y2-y2+1-x2)2 ({e4+2e3}{a+1}2+1-o2) 2 (16{k+1}3{k+2}{n+1}2+1-f2)2 ({(a+u~-u2a) 2-1}{n+4dy}2+1-{x+cu}2) 2 - (ai+k+1-l-i) 2 ({gk+2g+k+1}{h+j}+h-z) 2 (16r2y~{a2-1}+1-u2) 2 (p-m+l{a-n-1}+b{2an+2a-n2-2n-2}) 2 (z-pm+pla-p21+t{2ap-p2-1}) 2 (qTx+y{a-p-1}+s{2ap+2a-p2-2~-2})2 ( a ~ 1 2 - 1 2 + 1 - m 2 ) 2 - (n+l+v-y) ]. But I hope that you are satisfied with the f i r s t d e f i n i t i o n t h a t I gave. T h e r e a r e two f a c t s a b o u t t h e d i s t r i b u t i o n o f p r i m e n u m b e r s of w h i c h I h o p e to c o n v i n c e y o u so o v e r w h e l m i n g l y t h a t t h e y w i l l b e p e r m a n e n t l y e n g r a v e d in y o u r h e a r t s . T h e f i r s t is that, d e s p i t e t h e i r s i m p l e d e f i n i t i o n a n d r o l e as the b u i l d ing b l o c k s of t h e n a t u r a l n u m b e r s , the p r i m e n u m b e r s b e l o n g to t h e m o s t a r b i trary and ornery objects studied by mathe m a t i c i a n s : t h e y g r o w like w e e d s a m o n g the n a t u r a l n u m b e r s , s e e m i n g to o b e y no o t h e r l a w t h a n t h a t o f c h a n c e , a n d nob o d y c a n p r e d i c t w h e r e the n e x t o n e w i l l sprout. T h e s e c o n d f a c t is e v e n m o r e a s t o n i s h i n g , for it s t a t e s j u s t t h e o p posite: that the prime numbers exhibit s t u n n i n g r e g u l a r i t y , t h a t t h e r e a r e laws governing their behaviour, and that they obey these laws with almost military precision. 8 To support the first of these claims, let me begin by showing you a list of the prime and c o m p o s i t e numbers up to 100 (where apart from 2 I have listed only the odd numbers). prime 2 3 5 7 11 13 17 19 23 29 31 37 41 compo s i te 43 47 53 59 61 67 71 73 79 83 89 97 9 15 21 25 27 33 35 39 45 49 51 55 57 63 65 69 75 77 81 85 87 91 93 95 99 2127 - 1 = 170141183460469231731687303715884105727. Not u n t i l 1951, with the appearance of electronic computers, were larger prime numbers discovered. In the a c c o m p a n y i n g table, you can see the data on the successive title-holders (3). At t h e m o m e n t , the lucky fellow is the 6 0 0 2 - d i g i t number 219937 - I (which I w o u l d n o t care to write down); if you d o n ' t believe me, you can look it up in the G u i n n e s s book of records. The l a r g e s t known prime number and lists of the primes among the 1OO numbers i m m e d i a t e l y preceding and following 10 million: The prime numbers between 9,999,900 and 10,000,000 9,999,901 9,999,907 9,999,929 9,999,931 9,999,937 9,999,943 9,999,971 9,999,973 9,999,991 The prime numbers between 10,0OO,OOO and 10,0OO,100 10,0OO,019 I0,OOO,079 I hope you will agree that there is no apparent r e a s o n why one number is prime and another not. To the contrary, upon looking at these numbers one has the feeling of being in the presence of one of the inexplicable secrets of creation. That even m a t h e m a t i c i a n s have not penetrated this secret is perhaps m o s t c o n v i n c i n g l y shown by the ardour w i t h w h i c h they search for bigger and bigger primes - with n u m b e r s which grow regularly, like squares or powers of two, nobody would ever bother writing d o w n examples larger than the p r e v i o u s l y known ones, but for prime numbers people have gone to a great deal of trouble to do just that. For example, in 1876 Lucas proved that the number 2 127 - I is prime, and for 75 years this remained u n s u r p a s sed - which is perhaps not surprising when one sees this number: 2127- I ~ ( 2 1 4 8 + I) 114(2127180(2127- I) + I I)2+ I ~ ~ , 0 0 ~-,~ ~'~ ~o ~O ~,~ o 39 1876 Lucas 44 1951 Ferrier 41 ~ 79 ~ 1951 Miller.+ Wheeler + EDSAC I 1952 Lehmer + Robinson + SWAC 2521 - I 157 2607 - I 21279 - I 22203 - I 183 I 386 664 22281 - I 687 23217 - 1 969 24253 - I 24423 - I 1281 ~ 1332 # 29689 - I 2917 29941 - I 2993 ~ 211213 - I 3376 219937 - 1 6002 1957 1961 Riesel + BESK Hurwitz + Selfridge + IBM 7090 1963 Gillies + ILIAC 2 1971 Tuckerman + IBM 360 More interesting, however, is the question about the laws g o v e r n i n g prime numbers. I have already shown you a list of the prime numbers up to 1OO. Here is the same information p r e s e n t e d g r a p h i c a l ly. The function d e s i g n a t e d ~(x) (about which I will be c o n t i n u a l l y speaking from now on) is the number of prime numbers not e x c e e d i n g x; thus ~(x) begins with O and jumps by I at every prime number 2, 3, 5 etc. A l r e a d y in this picture we can see that, despite small oscillations, ~(x) by and large grows quite regularly. 9 no e x c e p t i o n to this rule. It is not difficult to flnd an empirical formula w h i c h gives a good d e s c r i p t i o n of the growth of the prime numbers. B e l o w 1OO there are 25 primes, that is, o n e - f o u r t h of the numbers; b e l o w 10OO there are 168, or about one-sixth; up to 10,OO0 there are 1229 prime numbers, i.e. a b o u t one-eighth. If we extend this list, computing the p r o p o r t i o n of prime numbers to natural n u m b e r s up to one hundred thousand, one million, etc., then we find the following table (in which the v a l u e s of ~(x), listed so n o n c h a l a n t l y here, r e p r e s e n t thousands of hours of dreary calculation). ~(x) 0 X I ,,,. 100 50 But when I extend the domain of x - v a l u e s from a hundred to fifty thousand, then this r e g u l a r i t y b e c o m e s b r e a t h - t a k i n g l y clear, for the graph now looks like this: 6000 5000 4000 3000 2000 1000 / 0 J I t l i 10000 20000 30000 40000 50000 x For me, the smoothness with which this curve climbs is one of the m o s t a s t o n i s h ing facts in m a t h e m a t i c s . Now, w h e r e v e r nature reveals a pattern, there are sure to crop up scientists looking for the explanation. The r e g u l a r i t y observed in the primes forms x ~(x) x/~(x) 10 1OO 1OOO 10,OO0 1OO,OOO 1,OOO,OOO 10,0OO,0OO 100,0OO,OOO 1,O00,0OO,OOO 10,OOO,OOO,OOO 4 25 168 1,229 9,592 78,498 664,579 5,761,455 50,847,534 455,052,512 2.5 4.0 6.0 8.1 10.4 12.7 15.O 17.4 19.7 22.0 Here we see that the ratio of x to ~(x) always jumps by a p p r o x i m a t e l y 2.3 when we go from a power of 10 to the next. M a t h e m a t i c i a n s immediately recognize 2.3 as the logarithm of 10 (to the base e, of course). Thus we are led to c o n j e c t u r e that x ~(x) ~ log x where the sign ~ m e a n s that the ratio ~(x)/(x/log x) tends to I as x goes to infinity. This r e l a t i o n s h i p (which was not proved until 1896) is k n o w n as the prime number ~eorem. Gauss, the g r e a t e s t m a t h e m a t i c i a n of them all, d i s c o v e r e d it at the age of fifteen by studying prime number tables contained in a book of logarithms that had been given to him as a present the previous year. T h r o u g h o u t his life Gauss w a s keenly interested in the d i s t r i b u t i o n of the prime n u m b e r s and he m a d e e x t e n s i v e calculations. In a letter to Enke (4) he d e s c r i b e s how he "very often used an idle quarter of an hour to count through another chiliad [i.e., an interval of 1,OO0 numbers] here and there" ~ntil finally he had listed all the prime numbers up to 3 m i l l i o n (!) and compared their d i s t r i b u t i o n w i t h the formula w h i c h he had conjectured. The prime number theorem s~ates that ~(x) is a s y m p t o t i c a l l y - i.e., with a relative error of 0% - equal to x/log x. But if we compare the g r a p h of the function x/log x with that of ~(x), then we 10 see that, a l t h o u g h the function x / l o g x q u a l i t a t i v e l y m i r r o r s the b e h a v i o u r of ~(x), it c e r t a i n l y does not agree with ~(x) s u f f i c i e n t l y well to explain the smoothness of the latter: 6000 5000 4000 3000 6000 2000 5000 1000 &O00 ~ ~t~(x I 10 000 log x 3000 X 20 000 30 000 40 000 501000 . 2000 1000 I i I I I 10 000 20 000 30 000 40 000 50 000 X v Therefore it is natural to ask for better approximations. If we take another look at our table of the ratios of x to ~(x), we find that this ratio is almost exactly log x - I. W i t h a more careful calcul a t i o n and with m o r e detailed data on ~(x), Legendre (5) found in 1808 that a p a r t i c u l a r l y good a p p r o x i m a t i o n is obtained if in place of I we subtract 1.08366 from log x, i.e. x ~(x) % log x - 1.08366 Another good a p p r o x i m a t i o n to ~(x), w h i c h was first given by Gauss, is obtained by taking as starting point the empirical fact that the frequency of prime numbers near a very large number x is almost exactly I/log x. From this, the number of prime numbers up to x should be a p p r o x i m a t e l y given by the logari~mic sum _ LS(X) I I log 2 + log---~ + + "'" I log-----~ or, what is e s s e n t i a l l y the same (6), by the logari~mic integral x Li(x) I = 2/ lo--~-~t. If we now compare the graph of Li(x) with that of ~(x), then we see that w i t h i n the accuracy of our picture the two c o i n c i d e iexactly. T h e r e is no point in showing you the picture of L e g e n d r e ' s a p p r o x i m a t i o n as well, for in the range of the graph it is an even better a p p r o x i m a t i o n to ~(x). There is one more a p p r o x i m a t i o n which I would like to mention. Riemann's research on prime numbers suggests that the p r o b a b i l i t y for a large number x to be prime should be even closer to 1/log x if one c o u n t e d not only the prime numbers but also the powers of primes, counting the square of a prime as half a prime, the cube of a prime as a third, etc. This leads to the a p p r o x i m a t i o n (x) + 89 ~(/~) + ~ ~(3~s + ... = Li(x) or, equivalently, (x) = Li(x) - 89 ~ i ( ~ ) - 89 ~i(3/~) _ . . . (7) The f u n c t i o n on the right side of this formula is denoted by R(x), in honour of Riemann. It represents an a m a z i n g l y good a p p r o x i m a t i o n to ~(x), as the following values show: X ~ (X) R(X) I00,0OO,O00 200,000,000 300,000,000 400,000,000 500,000,000 600,000,000 700,000,000 800,000,000 900,000,000 1,000,OO0,000 5,761,455 11,078,937 16,252,325 21,336,326 26,355,867 31,324,703 36,252,931 41,146,179 46,009,215 50,847,534 5,761,552 11,O79,090 16,252,355 21,336,185 26,355,517 31,324,622 36,252,719 41,146,248 46,009,949 50,847,455 For those in tle function that R(x) is given by the ries R(x) = I + where ~(n+1) tion (8). the audience who know a littheory, perhaps I m i g h t add an entire function of log x, rapidly c o n v e r g i n g power se- ~ I n=1 n~ (n+1) (log x) n n! ' is the R i e m a n n zeta func- 11 At this point I should emphasize that Gauss's and Legendre's approximations to ~(x) were obtained only empirically, and that even Riemann, although he was led to his function R(x) by theoretical considerations, never proved the prime number theorem. That was first accomplished in 1896 by Hadamard and (independently) de la Vall~e Poussin; their proofs were based on Riemann's work. While still on the theme of the pred i c t a b i l i t y of the prime numbers, I would like to give a few m o r e numerical examples. As already mentioned, the probability for a number of the order of magnitude x to be prime is roughly equal to I/log x; that is, the number of primes in an interval of length a about x should be approximately a/log x, at least if the interval is long enough to make statistics meaningful, but small in comparison to x. For example, we would expect to find around 8142 primes in the interval between 100 m i l l i o n and 100 m i l l i o n plus 150,000 because Interval Prime numbers I O 1OO,OO0,OOO1OO,150,OOO 1,OOO,O00,OOO1,OOO,150,OOO 10,OOO,O00,OOO10,OOO,150,OOO 100,OOO,O00,OOO100,000,150,000 1,000,000,000,0001,000,000,150,000 10,000,000,000,00010,000,000,150,000 100,000,000,000,000100,000,000,150,000 1,000,000,000,000,0001,000,000,000,150,000 log 150,OOO (1OO,O00,OO0) 150,OOO = 18.427 150~000 8142 I ~ 8154 ~ 584 601 7238 7242 461 466 6514 6511 5922 374 389 5974 309 276 5429 5433 259 276 5011 5065 221 208 4653 4643 191 4343 4251 186 166 161 . % 8142 Correspondingly, the probability that two random numbers near x are both prime is approximately I/(log x) 2. Thus if one asks how many prime twins (i.e. pairs of primes differing by 2, like 11 and 13 or 59 and 61) there are in the interval from x to x+a then we m i g h t expect approxim a t e l y a/(log x) 2. Actually, we should expect a bit more, since the fact that n is already prime slightly changes the chance that n+2 is prime (for example n+2 is then certainly odd). An easy heuristic argument (9) gives C.a/(log x) 2 as the expected number of twin primes in the interval [x, x+a] where C is a .constant with value about 1.3 (more exactly: C = 1.3203236316...). Thus between 1OO m i l l i o n and 1OO m i l l i o n plus 150 thousand there should be about (1.32''')(18.427)2= Prime twins 584 As you can see, the agreement with the theory is extremely good. This is especially surprising in the case of the prime pairs, since it has not yet even been proved that there are infinitely many such pairs, let alone that they are distributed according to the conjectured law. I want to give one last illustration of the predictability of primes, namely the problem of the gaps between primes. If one looks at tables of primes, one sometimes finds unusually large intervals, e.g. between 113 and 127, which don't contain any primes at all. Let g(x) be the length of the largest prime-free interval or "gap" up to x. For example, the largest gap below 200 is the interval from 113 to 127 just mentioned, so g(2OO) = 14. Naturally, the number g(x) grows very erratically, but a heuristic argument suggests the asymptotic formula (II) g(x) pairs of prime twins. Here are data computed by Jones, Lal and Blundon (io) giv-ing the exact number of primes and prime twins in this interval, as well as in several equally long intervals around larger powers of 10: ~ (log x) 2 In the following picture, you can see how well even the wildly irregular function g(x) holds to the expected behaviour. 12 700 This picture, I think, shows w h a t the person who decides to study number theory has let himself in for. As you can see, for small x (up to a p p r o x i m a t e l y I million) Legendre's a p p r o x i m a t i o n x/(log x - 1.O8366) is c o n s i d e r a b l y better than Gauss's Li(x), but after 5 million Li(x) is better, and it can be shown that Li(x) stays better as x grows. 600 500 &O0 300 200 100 0 10 102 10 3 10 4 10 5 106 107 108 109 1010 1011 1012 But u p to 10 m i l l i o n there are only some 600 thousand prime numbers; to s h o w you the promised 50 m i l l i o n primes, I have to go not to 10 m i l l i o n but all the way out to a b i l l i o n (American style: 10 ). In this range, the g r a p h of R(x) - ~(x) looks like this(13) : 600~ R(x)-X(x) /~ .A,.,, ,A -300| Up to now, I have s u b s t a n t i a t e d m y claim about the o r d e r l i n e s s of the primes m u c h m o r e t h o r o u g h l y than m y c l a i m about their disorderliness. Also, I have not yet fulfilled the promise of m y title to s h o w you the first 50 m i l l i o n primes, but have only shown you a few thousand. So here is a graph of ~(x) compared w i t h the a p p r o x i m a t i o n s of Legendre, Gauss, and Riemann up to 10 m i l l i o n (12) . Since these four functions lie so close together that their graphs are i n d i s t i n g u i s h a b l e to the naked eye - as we already saw in the picture up to 50,000 - I have plotted only the differences b e t w e e n them: AA , fL -- .;o O~< x ~< 1,000,000,000 The o s c i l l a t i o n of the function K ( x ) - ~ ( x ) become larger and larger,;but even for these a l m o s t i n c o n c e i v a b l y large values of x they never go beyond a few hundred. In c o n n e c t i o n with these data I can m e n t i o n yet another fact about the number of prime numbers ~(x). In the picture up to 10 million, G a u s s ' s approximation was always bigger than ~(x). That remains so until a billion, as you can see in the picture on the following page (in w h i c h the above data are plotted logarithmically). X(X)+3OO- z(x}+200i ~(x)+l1~176 3~'(x]1 ~(x)-1ool Legendre V~ i,fV Vm " R,emann , x (m rmlilons) M -- Rlemann " V ~ / ~ / Riemann ~ ' " - ~N 13 verges, i.e. e v e n t u a l l y exceeds any prev i o u s l y given number. His proof, which is also very simple, used the function 200C 100C ~(s) = I + i_ + 2s Lt(x)-R(x) "~/// 500 / t 200 10C 10 I 20 I 30 i 50 I 70 I 1 100 200 x (in millions) J 500 I 1000 S u r e l y this graph gives us the impression that with i n c r e a s i n g x the difference Li(x) - ~(x) grows steadily to infinity, that is, that the logarithmic integral Li(x) c o n s i s t e n t l y o v e r e s t i m a t e s the number of primes up to x (this w o u l d agree with the o b s e r v a t i o n that R(x) is a better a p p r o x i m a t i o n than Li(x), since R(x) is always smaller than Li(x)). But this is false: it can be shown that there are points where the o s c i l l a t i o n s of R(x) - ~(x) are so big that ~(x) a c t u a l ly becomes larger than Li(x). Up to now no such numbers have been found and perhaps none ever will be found, but Littlewood proved that they exist and Skewes(14) proved that there is one that is smaller than I__ + ... , 3s whose importance for the study of ~(x) was fully r e c o g n i z e d only later, with the work of Riemann. It is amusing to rem a r k that, a l t h o u g h the sum of the reciprocals of all primes is divergent, this sum over all the known primes (let's say the first 50 million) is smaller than four(15). The first major result in the direction of the prime number t h e o r e m was proved by C h e b y s h e v in 1850(16). He showed that for s u f f i c i e n t l y large x x x < ~(x) 0.89 log < 1.11 x log x i.e., the prime number t h e o r e m is correct with a relative error of at most 11%. His proof uses binomial c o e f f i c i e n t s and is so p r e t t y that I c a n n o t r e s i s t at least sketching a simplified v e r s i o n of it (with somewhat worse constants). In the one direction, we will prove x (x) < 1.7 ~ . This i n e q u a l i t y is valid for x < 12OO. Assume i n d u c t i v e l y that it has been proved for x < n and consider the m i d d l e binomial c o e f f i c i e n t 10101034 (a number of which H a r d y once said that it was surely the b i g g e s t that had ever served any definite purpose in m a t h e m a t ics). In any case, this example shows h o w u n w i s e it can be to base c o n c l u s i o n s about primes solely on numerical data. In the last part of my lecture I would like to talk about some theoretical results about ~(x) so that you don't go away w i t h the feeling of having seen only e x p e r i m e n t a l math. A n o n - i n i t i a t e would c e r t a i n l y think that the property of being prime is m u c h too random for us to be able to prove a n y t h i n g about it. This was refuted already 2,200 years ago by Euclid, who proved the existence of infinitely m a n y primes. His argument can be formulated in one sentence: If there were only finitely m a n y primes, then by m u l t i p l y i n g them together and adding I, one would get a number which is not div i s i b l e by any prime at all, and that is impossible. In the 18th century, Euler proved more, namely that the sum of the r e c i p r o c a l s of the prime numbers di- Since 22n= n (I+1)2n= (o2n)+(~n)+'''e(~n)+'''+(2n2n) this c o e f f i c i e n t other hand is at m o s t 22n. On the 2n (2n) ! (2n) x(2n-1) x...x2xl (n ) = ~ = (nx(n-1)X. x2xl) " Every prime p smaller than 2n appears in the numerator, but c e r t a i n l y no p bigger than n can appear in the denominator. is d i v i s i b l e by every prime b e t w e e n n and 2n: n<p<2n P 2n (n) 9 But the p r o d u c t has ~(2n) - ~(n) each bigger than n, so we get factors, 14 (2n) -~ (n) n or, < taking ~(2n) 2n (n) ~ p< n<p<2n ram, ways ered mula 22n < logarithms - ~(n) < 2n log 2 < 1 39 n log n " ~ B y induction, the t h e o r e m is v a l i d so ~(n) < I .7 (n/log n), and a d d i n g relations gives for n, these (n > 1200). H e n c e the t h e o r e m Since ~(2n+I) is v a l i d also for < 1.7 l o g 2n. n + I < 3.09 ~ o g n + I < ~(2n) 2n+I (2n+i") (n > 1200) it is a l s o v a l i d for 2n+I, c o m p l e t i n g the induction. For t h e b o u n d in the o t h e r d i r e c t i o n , w e need a s i m p l e l e m m a w h i c h c a n b e p r o v e d e a s i l y u s i n g the w e l l - k n o w n form u l a for the p o w e r of p w h i c h d i v i d e s n! (17) : Lemma: Let p be a prime. If pVp is the largest power of p then pVP Corollary: (~), dividing + 89 " n 2n < 3.09 ~ < 1.7 log n log (2n) ~(2n) he d i d s o m e t h i n g w h i c h is in m a n y m u c h m o r e a s t o n i s h i n g - he d i s c o v a n exact f o r m u l a for ~(x). This forhas the f o r m = Li(x) - + 89 Z P Li + (x p) w h e r e the sum r u n s o v e r the r o o t s of the zeta f u n c t i o n ~(s)(18) . T h e s e r o o t s (apart f r o m the s o - c a l l e d "trivial" r o o t s 0 = -2, -4, -6,..., w h i c h y i e l d a n e g l i g i b l e c o n t r i b u t i o n to the formula) are c o m p l e x n u m b e r s w h o s e r e a l p a r t s lie between 0 a n d 1. The f i r s t t e n of them are as follows(19): I Pl = ~ + 1 4 . 1 3 4 7 2 5 i, I P2 = ~ + 2 1 . 0 2 2 0 4 0 i, I P3 = ~ + 2 5 . 0 1 0 8 5 6 i, I P4 = ~ + 3 0 . 4 2 4 8 7 8 i, I P5 = ~ + 3 2 . 9 3 5 0 5 7 i, Pl ~ 89 - 14.134725 i, I P2 = ~ - 2 1 . 0 2 2 0 4 0 i, 1 P3 = ~ - 2 5 . 0 1 0 8 5 6 i, 54 = 89 - 3 0 . 4 2 4 8 7 8 i, 55 = 89 - 3 2 . 9 3 5 0 5 7 i. < n. coefficient"(~) Evergbinemial satisfies (~) = ~ p~n pVp _< n~(n) If w e add t h e i n e q u a l i t y for all b i n o m i a l of the c o r o l l a r y coefficients (~) w i t h g i v e n n, t h e n w e find 2n = (I+I) n = and t a k i n g n Z k=O (~) S logarithms (n+l).n ~(n) It is e a s y to s h o w t h a t w i t h e a c h r o o t its c o m p l e x c o n j u g a t e a l s o appears. B u t that the real p a r t of e v e r y r o o t is exa c t l y I/2 is still u n p r o v e d : this is the famous Riemann hypothesis, which would have f a r - r e a d h i n g c o n s e q u e n c e s for number theory(20). It has b e e n v e r i f i e d for 7 m i l l i o n roots. W i t h t h e h e l p of the R i e m a n n f u n c t i o n R(X) i n t r o d u c e d a b o v e w e c a n w r i t e R i e m a n n ' s f o r m u l a in the f o r m gives ~(x) ~(n) > n log 2 - 'log n > 2 n 3 log n log - log = R(x) - Z R ( x p) P (n+1) n (n > 200). In closing, I w o u l d like to say a few w o r d s a b o u t R i e m a n n ' s work. A l t h o u g h R i e m a n n n e v e r p r o v e d the p r i m e n u m b e r t h e o - The k th a p p r o x i m a t i o n to ~ (x) w h i c h f o r m u l a y i e l d s is the f u n c t i o n Rk(X) = R(x) + TI(X) + T2(x) this + . . . + Tk(X) , 15 where Tn(X) = -R(xPn) - R(x pn) is the c o n t r i b u t i o n of the nth pair of roots of the zeta function. For each n the function Tn(x) is a smooth, oscillating function of x. The first few look like this (21): ~ 0 ~ AA A A 0.3 0.2 i o~ ~ 0 N = T1 (x) -0.1 It follows that Rk(x) is also a smooth function for each k. As k grows, these functions approach ~(x). Here, for example, are the graphs of the 10th and 29th approximations, -0.2 -0.3 0.2 i/ ~ N = T2 (X) 0.1 ~ 15-- O. N T3 (x) -0"II -0.2 0.2 0.1 0 -0.1 -0.2 ~/ N ~~o Tz. (x) X I 50 I 100 ..- 16 Xtx) .............. R10(x) R29(x) R29(x) I I 50 t00 a n d if w e c o m p a r e t h e s e c u r v e s w i t h the g r a p h of ~(x) u p to 100 (p. 9) we g e t the following picture: Remarks (I) J.M. G a n d h i , F o r m u l a e for the n t h prime, Proc. W a s h i n g t o n S t a t e Univ. Conf. o n N u m b e r Theory, W a s h i n g t o n S t a t e Univ., P u l l m a n , Wash., 1971 96 - 106 (2) J.P. Jones, D i o p h a n t i n e r e p r e s e n t a t i o n of the s e t o f p r i m e n u m b e r s , N o t i c e s of t h e A M S 22 (1975) A 326. I I 50 100 X I h o p e t h a t w i t h this a n d the o t h e r p i c t u r e s I h a v e shown, I h a v e c o m m u n i c a t ed a c e r t a i n i m p r e s s i o n of the i m m e n s e ibeauty o f the p r i m e n u m b e r s and of the iendless s u r p r i s e s w h i c h t h e y h a v e in istore for us. (3) T h e r e is a g o o d r e a s o n w h y so m a n y of t h e n u m b e r s in t h i s l i s t h a v e the f o r m M k = 2k - 1: A t h e o r e m of L u c a s s t a t e s that M k (k>2) is p r i m e if a n d o n l y if M k d i v i d e s Lk-1, w h e r e the n u m b e r s L n a r e d e f i n e d ind u c t i v e l y b y L I = 4 a n d L n + I = L~ - 2 (so L 2 = 14, L 3 = 194, L 4 = 37634, ...) a n d h e n c e it is m u c h e a s i e r to t e s t w h e t h e r M k is p r i m e t h a n it is to t e s t a n o t h e r n u m b e r of the same order of magnitude. T h e p r i m e n u m b e r s o f the f o r m 2K - I (for w h i c h k i t s e l f m u s t n e c - 17 e s s a r i l y be prime) are called Mersenne primes (after the French m a t h ematlcian M e r s e n n e who in 1644 gave 9 a list of such p r i m e s up to 10 79 , c o r r e c t up to 1018 ) and play a role in connection w i t h a c o m p l e t e l y different p r o b l e m of number theory. Euclid d i s c o v e r e d that when 2P - I is prime then the number 2P -I (2P-I) is "perfect", i.e. it equals the sum of its proper d i v i s o r s (e.g. 6 = I + 2 + 3 , 2 8 = I + 2 + 4 + 7 + %4, 496 = I + 2 + 4 § 8 + 16 + 31 (lOg X) "t dt R(x) = 7 t r(t+i)~(t+l) o (~(s) = the Riemann zeta function and F(s) = the gamma function) and R ( e 2 ~ X ) ' = ~2 (~ 2 x + 62 + 124 + 248) and Euler showed that every even p e r f e c t number has this form. It is u n k n o w n whether there are any odd perfect numbers at all; if they exist, they must be at least 10 IQ0 . There are exactly 24 values of p < 20,000 for w h i c h 2P - 1 is prime. (4) (8) R a m a n u j a n has given the following a l t e r n a t i v e forms for this function: Gauss, Werke II (1892) 444 For a d i s c u s s i o n of the history of the various a p p r o x i m a t i o n s to ~(x), in which an E n g l i s h translation of this letter also appears, see L.J. Goldstein: A history of the prime number theorem, Amer. Math. Monthly, 80 (1973) 599 - 615. C.F. 447. (5) A.M. Legendre, E s s a i sur la theorie de Nombres, 2nd edition, Paris, 1808, p. 394. _ ~2 ( 12x 4 3 + 3--~4x 5 + ~ x6+ . . 2525 + 40x 3 + --~--x +... (9) Namely: The p r o b a b i l i t y that for a r a n d o m l y chosen pair (m, n) of numbers both m and n are ~ 0 (mod p) is o b v i o u s l y [ (p-1)/p] 2 , while for a r a n d o m l y chosen number n, the. p r o b a b i l i t y that n and n+2 are both O (mod p) is I/2 for p = 2 and (p-2)/p for p ~ 2. Thus the probability for n and n+2 m o d u l o p to be prime twins differs by a factor - 1.5 < Li(x) < Ls(x), i.e. the d i f f e r e n c e b e t w e e n Li(x) and Ls(x) is bounded. We should also m e n t i o n that the l o g a r i t h m i c integral is often d e f i n e d as the C a u c h y principal value for p r 2 and by 2 for p = 2 from the c o r r e s p o n d i n g p r o b a b i l i t y for two i n d e p e n d e n t numbers m and n. Altogether, we have therefore improved our c h a n c e s by a factor p2 _ 2p C = 2 Xdt + s oTG~q I-5 2 p (6) M o r e precisely Li(x) = o§ log t ) (Bk = k th Bernoulli number; the symbol ~ m e a n s that the d i f f e r e n c e of the two sides tends to O as x grows to infinity). See G.H. Hardy, Ramanujan: Twelve Lectures on Subjects S u g g e s t e d by His Life and Work, Cambridge U n i v e r s i t y Press, 1940, Chapter 2. p-2 Ls(x) 1 = lim /i-~ dt ) but this d e f i n i t i o n d i f f e r s from that given in the text only by a constant. (7) The c o e f f i c i e n t s are formed as follows: the c o e f f i c i e n t of Li(n/x) is + 1/n if n is the p r o d u c t of an e v e n number of d i s t i n c t primes, -I/n if n is the p r o d u c t of an odd number of d i s t i n c t primes, and 0 if n contains m u l t i p l e prime factors. p>2 p prime p2 _ 2p + I = 1.32032. For a somewhat more careful p r e s e n t a t i o n of this argument, see Hardy and Wright, An Int r o d u c t i o n to the Theory of Numbers, C l a r e n d o n Press, Oxford, 1960, w 22.20 (p. 371 - 373) (10) M.F. Jones, M. Lal, and W.J. Blundon, S t a t i s t i c s on certain large primes, Math. Comp. 21 (1967) 103 - 107. (11) D. Shanks, On maximal gaps b e t w e e n successive primes, Math. Comp. 18 (1964) 646 - 651. The g r a p h of g(x) was m a d e from the tables found in 18 w h e r e ~(x) § 0 as x t e n d s to infinity and C ~ 0 . 2 6 1 4 9 7 is a constant. T h i s e x p r e s s i o n is s m a l l e r than 3.3 w h e n x = 109 , and e v e n w h e n x = 1018 it s t i l l lies b e l o w 4. the f o l l o w i n g papers: L.J. L a n d e r and T.R. P a r k i n , On f i r s t a p p e a r ance of p r i m e d i f f e r e n c e s , M a t h . Comp. 21 (1967) 483 - 488, R.P. Brent, The f i r s t o c c u r r e n c e o f large gaps b e t w e e n s u c c e s s i v e p r i m e s , Math. Comp. 27 (1973) 959 - 963. (16) (12) The d a t a for this g r a p h are t a k e n f r o m L e h m e r ' s table of p r i m e n u m b e r s (D.N. L e h m e r , L i s t of P r i m e N u m b e r s f r o m I to 1 0 , O O 6 , 7 2 1 , H a f n e r P u b l i s h i n g Co., N e w York, 1956). (13) T h i s and the f o l l o w i n g g r a p h w e r e m a d e u s i n g t h e v a l u e s of ~(x) found in D.C. M a p e s , F a s t m e t h o d for comp u t i n g the n u m b e r of p r i m e s l e s s than a g i v e n limit, Math. Comp. 17 (1963) 179 - 185. In c o n t r a s t to L e h m e r ' s d a t a u s e d in the p r e v i o u s graph, t h e s e v a l u e s w e r e c a l c u l a t e d from a f o r m u l a for ~(x) and n o t by c o u n t i n g the p r i m e s u p to x. P.L. C h e b y s h e v , R e c h e r c h e s n o u v e l l e s sur les n o m b r e s p r e m i e r s , Paris, 1851, CR P a r i s 29 (1849) 397 - 401, 738 - 739. For a m o d e r n p r e s e n t a t i o n (in German) of C h e b y s h e v ' s proof, see W. Schwarz, E i n f U h r u n g in M e t h o den und Ergebnisse der Primzahltheorie B I - H o c h s c h u l t a s c h e n b u c h 278/278a, M a n n h e i m 1969, Chapt. II.4, P. 42 48. (17) T h e l a r g e s t p o w e r of p d i v i d i n g p! is p [ n / P ] + [ n T p 2 ] + . . . , w h e r e [x] is the l a r g e s t i n t e g e r S x. Thus in the n o t a t i o n of the l e m m a w e h a v e Vp = r~17.( I V ] (14) S. Skewes, O n the d i f f e r e n c e ~(x) li(x) (I), J. L o n d o n Math. Soc, 8 (1933) 277 - 283. Skewes' p r o o f of this b o u n d a s s u m e s the v a l i d i t y of the R i e m a n n h y p o t h e s i s w h i c h w e discuss later. T w e n t y - t w o y e a r s l a t e r (On the d i f f e r e n c e ~(x) - li(x) (II), Proc. Lond. Math. Soc. (3) 5 (1955) 48 - 70) he p r o v e d w i t h o u t u s i n g the R i e m a n n h y p o t h e s i s that t h e r e e x i s t s an x less t h a n the (yet m u c h larger) bound 101010964 for w h i c h ~(x) > Li(x). has b e e n l o w e r e d to This b o u n d 1010529.7 by C o h e n a n d M a y h e w and to 1 . 6 5 x i O 1165 by L e h m a n (On the d i f f e r e n c e ~(x) - li(x), A c t a A r i t h m . 11 (1966) 397 - 410). L e h m a n e v e n s h o w e d t h a t t h e r e is a n i n t e r v a l of at l e a s t 10500 n u m b e r s b e t w e e n 1.53• and 1 . 6 5 x i O I165 w h e r e ~(x) is larger than Li(x). A s a c o n s e q u e n c e o f his i n v e s t i g a t i o n , it a p p e a r s l i k e l y t h a t t h e r e is a n u m b e r n e a r 6 . 6 6 3 x10370 w i t h ~(x) > Li(x) a n d t h a t there is no n u m b e r less t h a n 1020 w i t h this p r o p e r t y . (15) N a m e l y (as c o n j e c t u r e d by G a u s s in 1796 and p r o v e d by M e r t e n s in 1874) 7 1 = log log x + C + r p<x p - [~ - [i) E v e r y s u m m a n d in this sum is e i t h e r O o r I and is c e r t a i n l y O for r > (log n / l o g p) (since then [n/p r] = O). T h e r e f o r e 9p S (log n / l o g p), f r o m w h i c h the c l ~ i m f o l l o w s . (18) T h e d e f i n i t i o n of ~(s) as I + I/2 s + I/3 s + ... g i v e n a b o v e m a k e s s e n s e o n l y w h e n s is a c o m p l e x n u m b e r w h o s e r e a l p a r t is l a r g e r than I (since the series c o n v e r g e s for t h e s e v a l u e s of s only) and in this d o m a i n ~(s) has no zeroes. B u t the f u n c t i o n ~(s) can b e e x t e n d e d to a f u n c t i o n for all c o m p l e x n u m b e r s s, so t h a t it m a k e s s e n s e to s p e a k of its r o o t s in the w h o l e c o m p l e x plane. The s i m p l e s t w a y to e x t e n d the d o m a i n of d e f i n i t i o n of ~(s) at l e a s t to the h a l f , p l a n e Re(s) > O is to u s e the i d e n t i t y (1-21-s)~(s) - 2 + 4-~ = I + 1 + I_~ + ... 2s 3s 6s ... = (_I)n-1 n=l ns ' w h i c h is v a l i d for Re(s) > I, and to o b s e r v e that the s e r i e s o n the r i g h t c o n v e r g e s for a l l s w i t h p o s i t i v e r e a l part. W i t h this, the "int e r e s t i n g " r o o t s o f the zeta function, i.e. the r o o t s p = B + iy w i t h O < 8 < I, can be c h a r a c t e r i z e d in 19 an elementary way by the two equations n~ I (-I)n-I n8 c o s ( y l o g n) = O, n~ 1 (-I)n-I n8 sin(ylog n) = O. The sum over the roots p in Riemann's formula is not a b s o l u t e l y c o n v e r g e n t and t h e r e f o r e m u s t be summed in the proper order (i.e., according to i n c r e a s i n g absolute value of Im(p)). Finally, I should m e n t i o n that, a l t h o u g h Riemann stated the exact formula for ~(x) already in 1859, it w a s n ' t proved until 1895 (by von Mangoldt). The 1977 Mathematics Calendar has found a worthy successor: Psychology Calendar '78 containing a selection of twelve themes which p o r t r a y something of the exciting spirit w h i c h pervades p s y c h o l o g i c a l inquiry: Dreaming Cognitive Contours (19) These roots were calculated a l r e a d y in 1903 by Gram (J.-P. Gram, Sur les zeros de la f o n c t i o n ~(s) de Riemann, Acta Math. 27 (1903) 289 304). For a v e r y nice p r e s e n t i o n of the theory of R i e m a n n ' s zeta function, see H.M. Edwards, Riemann's Zeta Function, A c a d e m i c Press, N e w York, 1974. Social D o m i n a n c e in Monkeys M c C o l l o u g h Effect Dynamic Effects of Repetitive Patterns Harlow and Piaget Shadow and Light Memory Span Test W i l h e l m W u n d t Father of P s y c h o l o g y (20) Namely the R i e m a n n hypothesis implies (and in fact is e q u i v a l e n t to the statement) that the error in Gauss's a p p r o x i m a t i o n Li(x) to ~(x) is at most a c o n s t a n t times x l / 2 . 1 o g x. At the present it is even unknown w h e t h e r this error is smaller than xC for any c o n s t a n t c < I. R e c o g n i t i o n of Faces Split Brain Jerome Bruner's Theory of Cognitive Learning (21) This and the f o l l o w i n g graphs are taken from H. Riesel and G. G6hl, Some c a l c u l a t i o n s r e l a t e d to Riem a n n ' s prime number formula, Math. Comp. 24 (1970) 969 - 983. %/ P.S. No cause for concern. The next Mathematics Calendar will appear in 1979.