# 7

The First 50 Million Prime Numbers*
Don Zagier
for or

sin I -

~F(s) s sin -s

;

the a l g e b r a i s t

it is of a f i n i t e field"

"the c h a r a c t e r i s t i c

"a p o i n t or

in Spec

~"

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 n r~1 ISil<'~'<irSn = 1 [I - log2( ~ + (-1)r 2 pi1"''pir - I S x);

)]

([x] = b i g g e s t

integer

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) = [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) ].

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.

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.

0 o 0 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. prime 2 3 5 7 11 13 17 19 23 29 31 37 41 43 47 53 59 61 67 71 73 79 83 89 97 compo s i te 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 More interesting.I is prime.907 9.with n u m b e r s which grow regularly.000.999.OOO.~ 1876 1951 1951 Lucas Ferrier Miller.I (which I w o u l d n o t care to write down).8 To support the first of these claims.I 21279 . however. despite small oscillations. 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.901 9.1 = 170141183460469231731687303715884105727. you can see the data on the successive title-holders (3). Not u n t i l 1951.I 219937 . but for prime numbers people have gone to a great deal of trouble to do just that.+ Wheeler + EDSAC I 2521 .000 9.I 22281 . upon looking at these numbers one has the feeling of being in the presence of one of the inexplicable secrets of creation. 3.999.943 9. In the a c c o m p a n y i n g table.I 24423 . 5 etc. To the contrary.931 9. 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 .0OO.999.999.0OO.999.971 9.OOO and 10. in 1876 Lucas proved that the number 2 127 .~ ~'~ ~O 2127I 39 44 41 ~ 79 ~ 157 183 I 386 664 687 969 1281 ~ 1332 # 2917 2993 ~ 3376 6002 1971 Tuckerman + IBM 360 1963 Gillies + ILIAC 2 1957 1961 Riesel + BESK Hurwitz + Selfridge + IBM 7090 Lehmer + Robinson + SWAC ~o ~ . you can look it up in the G u i n n e s s book of records.d i g i t number 219937 .999. 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.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. A l r e a d y in this picture we can see that.100 ~ ( 2 1 4 8 + I) 114(2127180(2127I) + I I)2+ I ~.937 9.which is perhaps not surprising when one sees this number: 211213 . like squares or powers of two.1 24253 . For example.I 23217 . the lucky fellow is the 6 0 0 2 .973 9.999.999.0OO.I 1952 10. The l a r g e s t known prime number ~ ~-.I 29941 .991 The prime numbers between 10. if you d o n ' t believe me.900 and 10. I have already shown you a list of the prime numbers up to 1OO.999.999. 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). ~(x) by and large grows quite regularly.I 2607 .019 I0. thus ~(x) begins with O and jumps by I at every prime number 2.I 22203 .I 29689 . .929 9. Here is the same information p r e s e n t e d g r a p h i c a l ly. At t h e m o m e n t . were larger prime numbers discovered. and for 75 years this remained u n s u r p a s sed . with the appearance of electronic computers. is the question about the laws g o v e r n i n g prime numbers.

0 8. 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 . 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.OOO. one million. etc. Gauss.7 22.3 when we go from a power of 10 to the next.5 4. If we extend this list.OOO 10.. of course).OOO ~(x) 4 25 168 1.OOO 1. 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. 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.0OO. 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.e. 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.512 x/~(x) 2.9 no e x c e p t i o n to this rule. then we But when I extend the domain of x .3 as the logarithm of 10 (to the base e.534 455.OOO.592 78. 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. The prime number theorem s~ates that ~(x) is a s y m p t o t i c a l l y .0OO 100. b e l o w 10OO there are 168. for the graph now looks like this: 6000 5000 4000 3000 2000 1000 / 0 J 10000 I 20000 t 30000 l 40000 i 50000 x For me.7 15.equal to x/log x.. ~(x) x 10 1OO 1OOO 10. i. an interval of 1.e.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. with a relative error of 0% . 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 . there are sure to crop up scientists looking for the explanation.t a k i n g l y clear. Now.052.O 17.455 50.0 6.498 664. up to 10.OO0 1OO.761. o n e . that is. M a t h e m a t i c i a n s immediately recognize 2. 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..v a l u e s from a hundred to fifty thousand. 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.0OO. The r e g u l a r i t y observed in the primes forms .e. then we find the following table (in which the v a l u e s of ~(x). or about one-sixth..0OO. r e p r e s e n t thousands of hours of dreary calculation).1 10.i.OOO 10.. But if we compare the g r a p h of the function x/log x with that of ~(x).OOO 1.229 9. B e l o w 1OO there are 25 primes. w h e r e v e r nature reveals a pattern.OO0 there are 1229 prime numbers.f o u r t h of the numbers.4 12.847. 100 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.OOO.O00. listed so n o n c h a l a n t l y here.579 5.0 0 50 X I ..4 19. a b o u t one-eighth.

000.455 LS(X) log 2 + log---~ "'" log-----~ or.000.215 50.252.719 41.185 26.355. 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 _ I I + + I There is one more a p p r o x i m a t i o n which I would like to mention.. 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). = Li(x) or.000 1.179 46.622 36. From this.O00 200.703 36. 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.847. in honour of Riemann. .146.000.146. 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 1.248 46.000.89 ~i(3/~) _ . rapidly c o n v e r g i n g power se- ~ I n=1 n~ (n+1) (log x) n n! ' zeta func- where ~(n+1) tion (8). the cube of a prime as a third.I.355.090 16.326 26.325 21.336. 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 3000 2000 1000 I i I I I X v ~ I ~ (x t~ X log x 10 000 20 000 30 000 40 000 501000 .000 800.517 31.867 31.355 21.000.0OO. is the R i e m a n n .937 16.455 11. w h i c h was first given by Gauss. i.89 ~ i ( ~ ) .000 For those in tle function that R(x) is given by the ries R(x) = I + ~ (X) 5.931 41. W i t h a more careful calcul a t i o n and with m o r e detailed data on ~(x).252. the audience who know a littheory.761.761. 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. equivalently. (x) = Li(x) . If we now compare the graph of Li(x) with that of ~(x). 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). as the following values show: X I00.000. counting the square of a prime as half a prime. 10 000 20 000 30 000 40 000 50 000 Therefore it is natural to ask for better approximations. by the logari~mic integral x I Li(x) = 2/ lo--~-~t.O79. .847. If we take another look at our table of the ratios of x to ~(x). 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.1.078.534 R(X) 5.000 600.000 400.000 300.552 11.000 500.324.009.08366 Another good a p p r o x i m a t i o n to ~(x).000 900. we find that this ratio is almost exactly log x .000 700.000.10 see that. what is e s s e n t i a l l y the same (6).324.. 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). x ~(x) % log x . 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.08366 from log x. This leads to the a p p r o x i m a t i o n (x) + 89 ~(/~) + ~ ~(3~s + .OO0.009.252. (7) The f u n c t i o n on the right side of this formula is denoted by R(x).e.252.000. etc.949 50.000.336. perhaps I m i g h t add an entire function of log x.

000.OOO10. at least if the interval is long enough to make statistics meaningful.000.000100.000. 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 As you can see.32''')(18.OOO = 18.a/(log x) 2 as the expected number of twin primes in the interval [x. As already mentioned.OOO.427 .000. Actually. let alone that they are distributed according to the conjectured law.000.e.000.OOO.OOO 100. For example.000.000. . which don't contain any primes at all. between 113 and 127.OOO1OO.OOO (1OO. 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) 2.150.000.427)2= 150~000 584 pairs of prime twins. If one looks at tables of primes.OO0.000. % 8142 Correspondingly.0001.000. but a heuristic argument suggests the asymptotic formula (II) g(x) ~ (log x) 2 (1. Let g(x) be the length of the largest prime-free interval or "gap" up to x.11 At this point I should emphasize that Gauss's and Legendre's approximations to ~(x) were obtained only empirically.000 1. Naturally.150. we should expect a bit more. although he was led to his function R(x) by theoretical considerations.O00.. e.000. I would like to give a few m o r e numerical examples. This is especially surprising in the case of the prime pairs. as well as in several equally long intervals around larger powers of 10: In the following picture. the number of primes in an interval of length a about x should be approximately a/log x.O00. the probability for a number of the order of magnitude x to be prime is roughly equal to I/log x. namely the problem of the gaps between primes.OOO 1. and that even Riemann. An easy heuristic argument (9) gives C. the probability that two random numbers near x are both prime is approximately I/(log x) 2.150. pairs of primes differing by 2. x+a] where C is a . so g(2OO) = 14.OO0) 150.000.000.000. For example.).3 (more exactly: C = 1.OOO100.000 8142 8154 584 601 7238 7242 461 466 6514 6511 5922 374 389 5974 309 276 5429 5433 259 276 5011 5065 221 208 186 4653 4643 191 4343 4251 166 161 log 150.000. one sometimes finds unusually large intervals.000. That was first accomplished in 1896 by Hadamard and (independently) de la Vall~e Poussin.000.000. the number g(x) grows very erratically. the largest gap below 200 is the interval from 113 to 127 just mentioned. their proofs were based on Riemann's work.000 10.150. Lal and Blundon (io) giv-ing the exact number of primes and prime twins in this interval.0001. While still on the theme of the pred i c t a b i l i t y of the prime numbers. 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.OOO.000. Thus if one asks how many prime twins (i.000 because Interval Prime numbers I Prime twins I O ~ ~ 1OO.constant with value about 1.OOO.OOO1.150. that is.O00.150. the agreement with the theory is extremely good.150.g. since it has not yet even been proved that there are infinitely many such pairs.OOO 10.OOO. you can see how well even the wildly irregular function g(x) holds to the expected behaviour. Here are data computed by Jones.000..000 100.000. 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).3203236316.150. but small in comparison to x.00010.000 1. I want to give one last illustration of the predictability of primes.O00.000. never proved the prime number theorem.

and it can be shown that Li(x) stays better as x grows. I think.000. and Riemann up to 10 m i l l i o n (12) .000 . the g r a p h of R(x) . fL AA .000 < The o s c i l l a t i o n of the function K ( x ) . Also.1. .as we already saw in the picture up to 50. 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). 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. X(X)+3OO- z(x}+200i ~(x)+l~176 1 3~'(x]1 ~(x)-1ool Legendre V~ Riemann / i. 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 .o O~ x ~< 1...fV Vm " R. 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. 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 ). That remains so until a billion.emann . Gauss. but after 5 million Li(x) is better. In this range. As you can see. to s h o w you the promised 50 m i l l i o n primes.. 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 .O8366) is c o n s i d e r a b l y better than Gauss's Li(x).I have plotted only the differences b e t w e e n them: R (x)-X(x) -- /~ . M -. But u p to 10 m i l l i o n there are only some 600 thousand prime numbers. 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. shows w h a t the person who decides to study number theory has let himself in for.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.~ ( x ) become larger and larger.. G a u s s ' s approximation was always bigger than ~(x).12 700 600 500 &O0 300 200 100 0 10 102 10 3 10 4 10 5 106 107 108 109 1010 1011 1012 This picture. as you can see in the picture on the following page (in w h i c h the above data are plotted logarithmically).000.A..Rlemann " V ~ / ~ ~ ' " - ~N x (m rmlilons) .~(x) looks like this(13) : 600~ .A -300| Up to now. In the picture up to 10 million. but have only shown you a few thousand.

200 years ago by Euclid.89 log x < ~(x) < 1. x2xl) " Every prime p smaller than 2n appears in the numerator. He showed that for s u f f i c i e n t l y large x x 0. e v e n t u a l l y exceeds any prev i o u s l y given number. since R(x) is always smaller than Li(x)). who proved the existence of infinitely m a n y primes.~(x) grows steadily to infinity. namely that the sum of the r e c i p r o c a l s of the prime numbers di- i. His proof. that is.~(n) each bigger than n. but c e r t a i n l y no p bigger than n can appear in the denominator...~(x) are so big that ~(x) a c t u a l ly becomes larger than Li(x). 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).e. used the function ~(s) Lt(x)-R(x) "~/// / t = I + i_ + I__ + .7 ~ x .. and that is impossible..11 x log x 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) . we will prove (x) < 1. 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). This i n e q u a l i t y is valid for x < 12OO. which is also very simple. 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. Euler proved more. In the one direction. It is amusing to rem a r k that. This was refuted already 2. Up to now no such numbers have been found and perhaps none ever will be found.x2xl (n ) = ~ = (nx(n-1)X. On the this c o e f f i c i e n t other hand 2n (2n) ! (2n) x(2n-1) x. 2s 500 3s 200 10C 10 I 20 I 30 i 50 I 70 I 1 100 200 x (in millions) J 500 I 1000 whose importance for the study of ~(x) was fully r e c o g n i z e d only later. In the 18th century.. i. In any case. A n o n . 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. so we get . then by m u l t i p l y i n g them together and adding I. His argument can be formulated in one sentence: If there were only finitely m a n y primes. a l t h o u g h the sum of the reciprocals of all primes is divergent. n<p<2n P But the p r o d u c t has ~(2n) . the prime number t h e o r e m is correct with a relative error of at most 11%. one would get a number which is not div i s i b l e by any prime at all.13 200C 100C verges. 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) . but Littlewood proved that they exist and Skewes(14) proved that there is one that is smaller than 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).e. 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). 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 Since 22n= n (I+1)2n= (o2n)+(~n)+'''e(~n)+'''+(2n2n) is at m o s t 22n. . is d i v i s i b l e by every prime b e t w e e n n and 2n: 2n (n) 9 factors.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. with the work of Riemann.

This forhas the f o r m logarithms < 2n log 2 < 1 39 n log n " ~ . -4. 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 .09 ~ < 1.. i. . 4 2 4 8 7 8 I P5 = ~ + 3 2 . 0 2 2 0 4 0 1 P3 = ~ . 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 - Rk(X) = R(x) + TI(X) + T2(x) + . 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 . then (~).7 (n/log n). . i.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.2 5 . 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 R ( x p) and t a k i n g ~(n) logarithms ~(x) = R(x) - Z > n log 2 'log n - log log (n+1) n P (n > 200). i.14. 4 2 4 8 7 8 55 = 89 . T h e s e r o o t s (apart f r o m the s o . ways ered mula 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 . i. and a d d i n g relations gives ~(2n) for n.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. 9 3 5 0 5 7 i. 0 1 0 8 5 6 I P4 = ~ + 3 0 .c a l l e d "trivial" r o o t s 0 = -2. i.. 0 2 2 0 4 0 I P3 = ~ + 2 5 . The f i r s t t e n of them are as follows(19): I Pl = ~ + 1 4 . i.14 (2n) -~ (n) n or.134725 I P2 = ~ . 0 1 0 8 5 6 54 = 89 . Pl ~ 89 .3 2 .he d i s c o v a n exact f o r m u l a for ~(x). which would have f a r . 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 this > 2 n 3 log n In closing. 1 3 4 7 2 5 I P2 = ~ + 2 1 . n + I < 3..7 l o g it is a l s o v a l i d for 2n+I. H e n c e the t h e o r e m Since ~(2n+I) < ~(2n) 2n+I (2n+i") is v a l i d also for 2n. 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. c o m p l e t i n g the induction. 9 3 5 0 5 7 i. 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.. If pVp is the largest power of p pVP dividing < n. i.7 log n log (2n) (n > 1200). taking ~(2n) < ~ p< n<p<2n 2n (n) 22n < ram.3 0 .~(n) " = + 89 Li(x) Z + 89 P Li (x p) + B y induction. + Tk(X) . these n 2n < 3. It has b e e n v e r i f i e d for 7 m i l l i o n roots.09 ~ o g n + I (n > 1200) < 1. 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. the t h e o r e m is v a l i d so ~(n) < I . Corollary: satisfies Evergbinemial coefficient"(~) (~) = ~ p~n pVp _< n~(n) of the c o r o l l a r y (~) 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 coefficients w i t h g i v e n n. i. -6.2 1 . 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 n w e find 2n = (I+I) n = n Z k=O (~) S gives (n+l). .r e a d h i n g c o n s e q u e n c e s for number theory(20).

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. are the graphs of the 10th and 29th approximations.1 -0.2 0.15 where Tn(X) = -R(xPn) . oscillating function of x.2 ~/ N ~~o Tz. For each n the function Tn(x) is a smooth.1 -0.1 ~ O.2 N = T2 (X) i/ 0.. As k grows.2 0. 0.2 -0. (x) X 50 I 100 I . The first few look like this (21): ~ 0 ~ AA A A 0.3 o~ ~ -0.- . -0"II ~ 0 15-N T3 (x) -0.2 i N = T1 (x) It follows that Rk(x) is also a smooth function for each k. for example. these functions approach ~(x).1 -0. Here.3 0 0.

.. T h e p r i m e n u m b e r s o f the f o r m 2K .. P u l l m a n .106 (2) J.M.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. Wash..16 Xtx) . Jones. N o t i c e s of t h e A M S 22 (1975) A 326. 9) we g e t the following picture: 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. W a s h i n g t o n S t a t e Univ. 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~ .) 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. Conf..... L 4 = 37634. 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. . L 3 = 194. R10(x) R29(x) R29(x) I 50 I t00 I I X 50 100 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. o n N u m b e r Theory. Proc...I (for w h i c h k i t s e l f m u s t n e c - .P... G a n d h i . Remarks (I) J.2 (so L 2 = 14. W a s h i n g t o n S t a t e Univ.. F o r m u l a e for the n t h prime. (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 . 1971 96 .... 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 ..

Math. 1 62 + 124 + 248) and Euler showed that every even p e r f e c t number has this form. if they exist.J. see Hardy and Wright. An Int r o d u c t i o n to the Theory of Numbers. (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. See G. p 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. i. E s s a i sur la theorie de Nombres.F. Cambridge U n i v e r s i t y Press. 2 8 = I + 2 + 4 + 7 + %4. There are exactly 24 values of p < 20. 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 Xdt = lim /i-~ dt Li(x) = o§ log t + I-5 s oTG~q ) 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. 496 = I + 2 + 4 § 8 + 16 + 31 + _ (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: (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 ) ' = ~ (~ x 2 2 ~2 4 3 + 3--~4x 6 5 +~x+. Math.e.1 is prime. Altogether.. Legendre.. Oxford. Paris. Comp. we have therefore improved our c h a n c e s by a factor p2 _ 2p C = 2 p>2 p prime p2 _ 2p + I i. = 1. Ramanujan: Twelve Lectures on Subjects S u g g e s t e d by His Life and Work.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. w 22.651. and W.e. -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. Math. see L. C l a r e n d o n Press. 18 (1964) 646 .M. 1808. Chapter 2. Goldstein: A history of the prime number theorem.615. ( 12x + 40x 3 + --~--x +. p.g. it equals the sum of its proper d i v i s o r s (e. Amer. Jones. Shanks.107. It is u n k n o w n whether there are any odd perfect numbers at all.20 (p.000 for w h i c h 2P . 447.32032. 2nd edition. (6) M o r e precisely Ls(x) . 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. 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). 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. Comp. Hardy.H. (Bk = k th Bernoulli number.. The g r a p h of g(x) was m a d e from the tables found in . 21 (1967) 103 . (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. For a somewhat more careful p r e s e n t a t i o n of this argument. the. 6 = I + 2 + 3 . On maximal gaps b e t w e e n successive primes.F. 394. (11) D. Euclid d i s c o v e r e d that when 2P . S t a t i s t i c s on certain large primes.373) (10) M. Thus the probability for n and n+2 m o d u l o p to be prime twins differs by a factor p-2 2 (5) A. 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.J. 80 (1973) 599 . 1940. M. 1960. in which an E n g l i s h translation of this letter also appears. Blundon. 371 .I is prime then the number 2P -I (2P-I) is "perfect".1. and 0 if n contains m u l t i p l e prime factors. 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 . Lal. 2525 ) (4) 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). they must be at least 10 IQ0 . while for a r a n d o m l y chosen number n. Monthly.5 < Li(x) < Ls(x).

( I V ] . 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. E i n f U h r u n g in M e t h o den und Ergebnisse der Primzahltheorie B I .. (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. N e w York. 738 . see W. W i t h this. w h e r e [x] is the l a r g e s t i n t e g e r S x. 8 (1933) 277 . J. 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) . 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. 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.410). . Soc. 27 (1973) 959 .. the "int e r e s t i n g " r o o t s o f the zeta function. P a r k i n . Paris. Comp.18 the f o l l o w i n g papers: L. 1956).. 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 . 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~17.7 by C o h e n a n d M a y h e w and to 1 .e. 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.L. 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. M a t h . P. M a p e s . (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 ] + . R. Brent..C.185..3 w h e n x = 109 . can be c h a r a c t e r i z e d in . i. Comp. (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) - 2 + 4~ n=l ns ' 7 1 = log log x + C + r p<x p w h i c h is v a l i d for Re(s) > I. 7 2 1 . . the r o o t s p = B + iy w i t h O < 8 < I. (3) 5 (1955) 48 .R. 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 . 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 . 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) = I + 1 + I_~ + . C h e b y s h e v . CR P a r i s 29 (1849) 397 . Chapt. O n the d i f f e r e n c e ~(x) li(x) (I). M a n n h e i m 1969. 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 . (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. L o n d o n Math. H a f n e r P u b l i s h i n g Co. 21 (1967) 483 . L e h m e r . Proc. Schwarz.283. 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 .H o c h s c h u l t a s c h e n b u c h 278/278a.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) . O O 6 .739. 2s 3s 6s = . and e v e n w h e n x = 1018 it s t i l l lies b e l o w 4. T h e r e f o r e 9p S (log n / l o g p).. Math. 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.401. f r o m w h i c h the c l ~ i m f o l l o w s .N. 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. II. Comp.963. 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. Soc. A c t a A r i t h m . Math.J. 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.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 w h e r e ~(x) § 0 as x t e n d s to infinity and C ~ 0 . T w e n t y . has b e e n l o w e r e d to This b o u n d 1010529. (18) T h e d e f i n i t i o n of ~(s) as I + I/2 s + I/3 s + . 17 (1963) 179 .. Skewes. 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 . 42 48. 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 . T h i s e x p r e s s i o n is s m a l l e r than 3.li(x). L a n d e r and T. 11 (1966) 397 .4.P.li(x) (II).[~ - [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). 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. Math.488. 6 5 x i O I165 w h e r e ~(x) is larger than Li(x). P. (14) S. 2 6 1 4 9 7 is a constant.53• and 1 . Lond. 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 . . 1851. (_I)n-1 101010964 for w h i c h ~(x) > Li(x).

Finally. Sur les zeros de la f o n c t i o n ~(s) de Riemann. N e w York. 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.983.-P. (19) These roots were calculated a l r e a d y in 1903 by Gram (J. Comp. 24 (1970) 969 .e. it w a s n ' t proved until 1895 (by von Mangoldt). %/ P. Math. 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. A c a d e m i c Press. 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. 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. 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 Social D o m i n a n c e 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 R e c o g n i t i o n of Faces Split Brain Jerome Bruner's Theory of Cognitive Learning in Monkeys (21) This and the f o l l o w i n g graphs are taken from H. see H. 1974. 27 (1903) 289 304). according to i n c r e a s i n g absolute value of Im(p)). G6hl.S. No cause for concern.. 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.19 an elementary way by the two equations n~ I (-I)n-I n8 n~ 1 (-I)n-I n8 c o s ( y l o g n) = O. Gram. The 1977 Mathematics Calendar has found a worthy successor: sin(ylog n) = O.M. Riesel and G. Acta Math. Edwards. Riemann's Zeta Function. The next Mathematics Calendar will appear in 1979. 1 o g x. (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 .