7

sin ~F(s)
The First 50 Million I - s ;

Prime Numbers* sin --

s

Don Zagier for 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
"a n o n - a r c h i m e d e a n valuation";
To my parents
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)

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

([x] = b i g g e s t integer S x);

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-the-
o 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 ~Antrittsvor-
lesung) held on May 5, 1975 at Bonn University.
Additional remarks and references to the lit-
erature 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.
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) ].
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 math-
e 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 no-
b 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 pre-
cision.
8
To support the first of these claims, 2127 - 1 =
let me begin by showing you a list of the 170141183460469231731687303715884105727.
prime and c o m p o s i t e numbers up to 100
Not u n t i l 1951, with the appearance of
(where apart from 2 I have listed only
electronic computers, were larger prime
the odd numbers).
numbers discovered. In the a c c o m p a n y i n g
prime compo s i te table, you can see the data on the suc-
cessive title-holders (3). At t h e m o m e n t ,
2 43 9 63 the lucky fellow is the 6 0 0 2 - d i g i t num-
3 47 15 65 ber 219937 - I (which I w o u l d n o t care to
5 53 21 69 write down); if you d o n ' t believe me, you
7 59 25 75 can look it up in the G u i n n e s s book of
11 61 27 77 records.
13 67 33 81
17 71 35 85
19 73 39 87
The l a r g e s t known prime number
23 79 45 91
29 83 49 93
31 89 51 95 ~ ~ ,
37 97 55 99
41 57 ~-,~ 0 0
~'~ ~o o

and lists of the primes among the 1OO

numbers i m m e d i a t e l y preceding and fol- ~O ~,~
lowing 10 million:
2127- I 39 1876 Lucas
The prime numbers between 9,999,900 and
10,000,000 ~ ( 2 1 4 8 + I) 44 1951 Ferrier
114(2127- I) + I 41 ~ Miller.+
9,999,901
9,999,907 180(2127- I)2+ I 79 ~ 1951 Wheeler +
EDSAC I
9,999,929
2521 - I 157
9,999,931
9,999,937
2607 - I 183 I Lehmer +
9,999,943
21279 - I 386 1952 Robinson
9,999,971
9,999,973 22203 - I 664 + SWAC
9,999,991 22281 - I 687
23217 - 1 969 1957 Riesel +
The prime numbers between 10,0OO,OOO and
BESK
10,0OO,100
24253 - I 1281 ~ Hurwitz +
10,0OO,019 24423 - I 1332 # 1961 Selfridge
+ IBM 7090
I0,OOO,079 29689 - I 2917
29941 - I 2993 ~ 1963 Gillies +
I hope you will agree that there is ILIAC 2
no apparent r e a s o n why one number is 211213 - I 3376
prime and another not. To the contrary, 219937 - 1 6002 1971 Tuckerman
upon looking at these numbers one has + IBM 360
the feeling of being in the presence of
one of the inexplicable secrets of crea-
tion. 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 More interesting, however, is the
w h i c h they search for bigger and bigger question about the laws g o v e r n i n g prime
primes - with n u m b e r s which grow regu- numbers. I have already shown you a list
larly, like squares or powers of two, no- of the prime numbers up to 1OO. Here is
body would ever bother writing d o w n ex- the same information p r e s e n t e d g r a p h i c a l -
amples larger than the p r e v i o u s l y known ly. The function d e s i g n a t e d ~(x) (about
ones, but for prime numbers people have which I will be c o n t i n u a l l y speaking from
gone to a great deal of trouble to do now on) is the number of prime numbers
just that. For example, in 1876 Lucas not e x c e e d i n g x; thus ~(x) begins with O
proved that the number 2 127 - I is prime, and jumps by I at every prime number 2,
and for 75 years this remained u n s u r p a s - 3, 5 etc. A l r e a d y in this picture we can
sed - which is perhaps not surprising see that, despite small oscillations,
when one sees this number: ~(x) by and large grows quite regularly.
 9
no e x c e p t i o n to this rule. It is not dif-
ficult 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
~(x) 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 follow-
ing 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 ~(x) x/~(x)

10 4 2.5
1OO 25 4.0
1OOO 168 6.0
10,OO0 1,229 8.1
1OO,OOO 9,592 10.4
1,OOO,OOO 78,498 12.7
10,0OO,0OO 664,579 15.O
100,0OO,OOO 5,761,455 17.4
1,O00,0OO,OOO 50,847,534 19.7
10,OOO,OOO,OOO 455,052,512 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
X
next. M a t h e m a t i c i a n s immediately recog-
I ,,,. nize 2.3 as the logarithm of 10 (to the
0 50 100 base e, of course). Thus we are led to
c o n j e c t u r e that
But when I extend the domain of x - v a l u e s x
from a hundred to fifty thousand, then ~(x) ~ log x
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 where the sign ~ m e a n s that the ratio
clear, for the graph now looks like this: ~(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
6000 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
5000 at the age of fifteen by studying prime
number tables contained in a book of log-
4000 arithms that had been given to him as a
present the previous year. T h r o u g h o u t his
3000 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
2000 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 of-
1000 ten used an idle quarter of an hour to
count through another chiliad [i.e., an
/ J I t l i x interval of 1,OO0 numbers] here and there"
0 10000 20000 30000 40000 50000
~ntil finally he had listed all the prime
numbers up to 3 m i l l i o n (!) and compared
For me, the smoothness with which this their d i s t r i b u t i o n w i t h the formula w h i c h
curve climbs is one of the m o s t a s t o n i s h - he had conjectured.
ing facts in m a t h e m a t i c s . The prime number theorem s~ates that
Now, w h e r e v e r nature reveals a pat- ~(x) is a s y m p t o t i c a l l y - i.e., with a re-
tern, there are sure to crop up scien- lative error of 0% - equal to x/log x.
tists looking for the explanation. The But if we compare the g r a p h of the func-
r e g u l a r i t y observed in the primes forms tion 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 6000
~(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 5000

~
smoothness of the latter:
4000
~t~(x
3000
6000
2000
5000
1000
&O00
I X
10 000 20 000 30 000 40 000 501000 .
3000 log x

2000
There is one more a p p r o x i m a t i o n which
1000 I would like to mention. Riemann's re-
search on prime numbers suggests that
I i I I I X
v
the p r o b a b i l i t y for a large number x to
10 000 20 000 30 000 40 000 50 000 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
Therefore it is natural to ask for better the square of a prime as half a prime,
approximations. If we take another look the cube of a prime as a third, etc. This
at our table of the ratios of x to ~(x), leads to the a p p r o x i m a t i o n
we find that this ratio is almost exact-
ly log x - I. W i t h a more careful calcu- (x) + 89 ~(/~) + ~ ~(3~s + ... = Li(x)
l a t i o n and with m o r e detailed data on
~(x), Legendre (5) found in 1808 that a or, equivalently,
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 ob-
tained if in place of I we subtract (x) = Li(x) - 89 ~ i ( ~ ) - 89 ~i(3/~) _ . . .

1.08366 from log x, i.e. (7)

x The f u n c t i o n on the right side of this

~(x) % log x - 1.08366 formula is denoted by R(x), in honour of
Riemann. It represents an a m a z i n g l y good
Another good a p p r o x i m a t i o n to ~(x), a p p r o x i m a t i o n to ~(x), as the following
w h i c h was first given by Gauss, is ob- values show:
tained by taking as starting point the
empirical fact that the frequency of X ~ (X) R(X)
prime numbers near a very large number x
is almost exactly I/log x. From this, I00,0OO,O00 5,761,455 5,761,552
the number of prime numbers up to x 200,000,000 11,078,937 11,O79,090
should be a p p r o x i m a t e l y given by the 300,000,000 16,252,325 16,252,355
logari~mic sum 400,000,000 21,336,326 21,336,185
500,000,000 26,355,867 26,355,517
_ I I + + I 600,000,000 31,324,703 31,324,622
LS(X) log 2 + log---~ "'" log-----~
700,000,000 36,252,931 36,252,719
800,000,000 41,146,179 41,146,248
or, what is e s s e n t i a l l y the same (6), by 900,000,000 46,009,215 46,009,949
the logari~mic integral 50,847,455
1,000,OO0,000 50,847,534
x I For those in the audience who know a lit-
Li(x) = 2/ lo--~-~t. tle function theory, perhaps I m i g h t add
that R(x) is an entire function of log x,
If we now compare the graph of Li(x) with given by the rapidly c o n v e r g i n g power se-
that of ~(x), then we see that w i t h i n the ries
accuracy of our picture the two c o i n c i d e
iexactly. R(x) = I + ~ I (log x) n
T h e r e is no point in showing you the pic- n=1 n~ (n+1) n! '
ture 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 where ~(n+1) is the R i e m a n n zeta func-
even better a p p r o x i m a t i o n to ~(x). tion (8).
 11
At this point I should emphasize that
Gauss's and Legendre's approximations to Interval Prime Prime
~(x) were obtained only empirically, and numbers twins
that even Riemann, although he was led I I
to his function R(x) by theoretical con- O ~ ~
siderations, never proved the prime num-
ber theorem. That was first accomplished
in 1896 by Hadamard and (independently)
de la Vall~e Poussin; their proofs were
based on Riemann's work. 1OO,OO0,OOO- 8142 8154 584 601
1OO,150,OOO
While still on the theme of the pre-
1,OOO,O00,OOO- 7238 7242 461 466
d i c t a b i l i t y of the prime numbers, I would 1,OOO,150,OOO
like to give a few m o r e numerical exam-
ples. As already mentioned, the probabil- 10,OOO,O00,OOO- 6514 6511 374 389
ity for a number of the order of magni- 10,OOO,150,OOO
tude x to be prime is roughly equal to 100,OOO,O00,OOO- 5922 5974 309 276
I/log x; that is, the number of primes 100,000,150,000
in an interval of length a about x should
be approximately a/log x, at least if 1,000,000,000,000- 5429 5433 259 276
the interval is long enough to make sta- 1,000,000,150,000
tistics meaningful, but small in compar- 10,000,000,000,000- 5011 5065 221 208
ison to x. For example, we would expect 10,000,000,150,000
to find around 8142 primes in the inter-
val between 100 m i l l i o n and 100 m i l l i o n 100,000,000,000,000-
4653 4643 191 186
plus 150,000 because 100,000,000,150,000
1,000,000,000,000,000-
4343 4251 166 161
1,000,000,000,150,000
150,OOO 150,OOO
log (1OO,O00,OO0) = 18.427 . % 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 approxi-
m 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 heu-
ristic 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 .con-
stant 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
150~000 584
(1.32''')(18.427)2=
As you can see, the agreement with the
theory is extremely good. This is espe-
cially 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 inter-
vals, 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 inter-
val 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 formu-
la (II)

g(x) ~ (log x) 2
pairs of prime twins. Here are data com-
puted by Jones, Lal and Blundon (io) giv--
ing the exact number of primes and prime In the following picture, you can see
twins in this interval, as well as in how well even the wildly irregular func-
several equally long intervals around tion g(x) holds to the expected behav-
larger powers of 10: iour.
12
700 This picture, I think, shows w h a t the
person who decides to study number the-
600 ory 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
500 x/(log x - 1.O8366) is c o n s i d e r a b l y bet-
ter than Gauss's Li(x), but after 5 mil-
&O0 lion Li(x) is better, and it can be shown
that Li(x) stays better as x grows.
300
But u p to 10 m i l l i o n there are only
200 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
100 way out to a b i l l i o n (American style:
10 ). In this range, the g r a p h of
0 R(x) - ~(x) looks like this(13) :
10 102 10 3 10 4 10 5 106 107 108 109 1010 1011 1012

600~ R(x)-X(x) /~ AA
.A,.,, ,A , fL
.;o
-300| -- O~< x ~< 1,000,000,000
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 The o s c i l l a t i o n of the function K ( x ) - ~ ( x )
their disorderliness. Also, I have not become larger and larger,;but even for
yet fulfilled the promise of m y title to these a l m o s t i n c o n c e i v a b l y large values
s h o w you the first 50 m i l l i o n primes, but of x they never go beyond a few hundred.
have only shown you a few thousand. So In c o n n e c t i o n with these data I can
here is a graph of ~(x) compared w i t h the m e n t i o n yet another fact about the num-
a p p r o x i m a t i o n s of Legendre, Gauss, and ber of prime numbers ~(x). In the pic-
Riemann up to 10 m i l l i o n (12) . Since these ture up to 10 million, G a u s s ' s approxima-
four functions lie so close together that tion was always bigger than ~(x). That
their graphs are i n d i s t i n g u i s h a b l e to the remains so until a billion, as you can
naked eye - as we already saw in the pic- see in the picture on the following page
ture up to 50,000 - I have plotted only (in w h i c h the above data are plotted log-
the differences b e t w e e n them: arithmically).

X(X)+3OO-

z(x}+200-

~(x)+l1~176 Legendre
V~ Riemann /

i,fV Vm
3~'(x]1 " R,emann , M -- Rlemann " V ~ / ~ ~ ' " - ~N

~(x)-1ool
x (m rmlilons)
 13
200C verges, i.e. e v e n t u a l l y exceeds any pre-
v i o u s l y given number. His proof, which
is also very simple, used the function
100C
~(s) = I + i_ + I__ + ... ,
2s 3s
Lt(x)-R(x)
500 "~/// whose importance for the study of ~(x)
/ t
was fully r e c o g n i z e d only later, with
the work of Riemann. It is amusing to re-
m a r k that, a l t h o u g h the sum of the recip-
200 rocals of all primes is divergent, this
sum over all the known primes (let's say
the first 50 million) is smaller than
four(15).
10C I I i I I 1 J I
10 20 30 50 70 100 200 500 1000 The first major result in the direc-
x (in millions) tion 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
S u r e l y this graph gives us the impres-
sion that with i n c r e a s i n g x the differ- x
x x < ~(x)
0.89 log < 1.11 log x
ence Li(x) - ~(x) grows steadily to in-
finity, that is, that the logarithmic in-
tegral 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 i.e., the prime number t h e o r e m is cor-
the number of primes up to x (this w o u l d rect with a relative error of at most
agree with the o b s e r v a t i o n that R(x) is 11%. His proof uses binomial c o e f f i c i e n t s
a better a p p r o x i m a t i o n than Li(x), since and is so p r e t t y that I c a n n o t r e s i s t at
R(x) is always smaller than Li(x)). But least sketching a simplified v e r s i o n of
this is false: it can be shown that there it (with somewhat worse constants).
are points where the o s c i l l a t i o n s of In the one direction, we will prove
R(x) - ~(x) are so big that ~(x) a c t u a l -
ly becomes larger than Li(x). Up to now (x) < 1.7 ~
x .
no such numbers have been found and per-
haps none ever will be found, but Little-
wood proved that they exist and Skewes(14) This i n e q u a l i t y is valid for x < 12OO.
proved that there is one that is smaller Assume i n d u c t i v e l y that it has been
than 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 Since n

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 - 22n= (I+1)2n= (o2n)+(~n)+'''e(~n)+'''+(2n2n)
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. this c o e f f i c i e n t is at m o s t 22n. On the
In the last part of my lecture I would other hand
like to talk about some theoretical re-
sults about ~(x) so that you don't go 2n (2n) ! (2n) x(2n-1) x...x2xl
away w i t h the feeling of having seen only (n ) = ~ = (nx(n-1)X. x2xl) "
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 Every prime p smaller than 2n appears in
being prime is m u c h too random for us to the numerator, but c e r t a i n l y no p bigger
be able to prove a n y t h i n g about it. This than n can appear in the denominator.
was refuted already 2,200 years ago by
Euclid, who proved the existence of in-
finitely m a n y primes. His argument can is d i v i s i b l e by every prime b e t w e e n n and
be formulated in one sentence: If there
2n:
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 di- 2n
v i s i b l e by any prime at all, and that is P (n) 9
n<p<2n
impossible. In the 18th century, Euler
proved more, namely that the sum of the But the p r o d u c t has ~(2n) - ~(n) factors,
r e c i p r o c a l s of the prime numbers di- each bigger than n, so we get
14
(2n) -~ (n) 2n 22n ram, he d i d s o m e t h i n g w h i c h is in m a n y
n < ~ p< (n) < ways 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 -
n<p<2n ered a n exact f o r m u l a for ~(x). This for-
mula has the f o r m
or, taking logarithms

~(2n) - ~(n) < 2n log 2 < 1 39 n + 89 + 89 +

log n " ~ "
= Li(x) - Z Li (x p)
B y induction, the t h e o r e m is v a l i d for n, P
so ~(n) < I .7 (n/log n), and a d d i n g these
relations gives
w h e r e the sum r u n s o v e r the r o o t s of the
n 2n zeta f u n c t i o n ~(s)(18) . T h e s e r o o t s
~(2n) < 3.09 ~ < 1.7
log n log (2n) (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 -
(n > 1200). 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 be-
H e n c e the t h e o r e m is v a l i d also for 2n. tween 0 a n d 1. The f i r s t t e n of them are
Since as follows(19):

n I
~(2n+I) < ~(2n) + I < 3.09 ~ o g n + I Pl = ~ + 1 4 . 1 3 4 7 2 5 i,
I
2n+I (n > 1200) P2 = ~ + 2 1 . 0 2 2 0 4 0 i,
< 1.7 l o g (2n+i")
I
P3 = ~ + 2 5 . 0 1 0 8 5 6 i,
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. I
P4 = ~ + 3 0 . 4 2 4 8 7 8 i,
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 I
P5 = ~ + 3 2 . 9 3 5 0 5 7 i,
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 for-
m 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) :
Pl ~ 89 - 14.134725 i,
Lemma: Let p be a prime. If pVp is the lar-
I
P2 = ~ - 2 1 . 0 2 2 0 4 0 i,
gest power of p dividing (~), 1
then P3 = ~ - 2 5 . 0 1 0 8 5 6 i,
pVP < n.
54 = 89 - 3 0 . 4 2 4 8 7 8 i,
Corollary: Evergbinemial coefficient"(~)
55 = 89 - 3 2 . 9 3 5 0 5 7 i.
satisfies

(~) = ~ pVp _< 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

p~n 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 ex-
If w e add t h e i n e q u a l i t y of the c o r o l l a r y a c t l y I/2 is still u n p r o v e d : this is the
famous Riemann hypothesis, which would
for all b i n o m i a l coefficients (~) 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 num-
ber theory(20). It has b e e n v e r i f i e d for
w i t h g i v e n n, t h e n w e find 7 m i l l i o n roots.
n 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
2n = (I+I) n = Z (~) S (n+l).n ~(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 -
k=O m a n n ' s f o r m u l a in the f o r m

and t a k i n g logarithms gives

~(x) = R(x) - Z R ( x p)
~(n) > n log 2 log (n+1) P
- 'log n - log n

> 2 n (n > 200). The k th a p p r o x i m a t i o n to ~ (x) w h i c h this

3 log n f o r m u l a y i e l d s is the f u n c t i o n
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 - Rk(X) = R(x) + TI(X) + T2(x) + . . . + 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 func-
tion Tn(x) is a smooth, oscillating func-
tion of x. The first few look like this
(21):
~
0 ~ AA A A

0.3
0.2 i
o~ N
It follows that Rk(x) is also a smooth
function for each k. As k grows, these
= T1 (x)
~ 0 functions approach ~(x). Here, for exam-
-0.1 ple, are the graphs of the 10th and 29th
approximations,
-0.2
-0.3

0.2

N
i/ ~ = T2 (X)
0.1

~
15--
O.
N
T3 (x)

-0"II
-0.2

0.2
0.1
0
-0.1
-0.2
~/ ~~o
N
Tz. (x)

I
50
I
100
X
..-
16

Xtx)
.............. R10(x)
R29(x)

R29(x)

I I
50 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. 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, 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.

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.
(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 in-
d 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 -

e s s a r i l y be prime) are called Mer-
senne primes (after the French m a t h -
ematlcian M e r s e n n e who in 1644 gave
a list 9 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 dif-
ferent 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
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) C.F. Gauss, Werke II (1892) 444 -
447. For a d i s c u s s i o n of the histo-
ry 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 transla-
tion of this letter also appears,
see L.J. Goldstein: A history of
the prime number theorem, Amer.
Math. Monthly, 80 (1973) 599 - 615.
(5) A.M. Legendre, E s s a i sur la theorie
de Nombres, 2nd edition, Paris,
1808, p. 394.
(6) M o r e precisely
Ls(x) - 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 inte-
gral is often d e f i n e d as the C a u c h y
principal value
Xdt = lim /i-~
Li(x) = o§ log t
+
I-5
s oTG~q ) = 1.32032. For a somewhat more
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 fol-
lows: 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 con-
tains m u l t i p l e prime factors.
17
(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 ) ' = ~2 (~
2 x + 3--~4x
4 3 + ~ x6+ . .
5 1
+ 31 +
_ ~2 ( 12x 2525
+ 40x 3 + --~--x +... )
(Bk = k th Bernoulli number; the sym-
bol ~ 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, Rama-
nujan: Twelve Lectures on Subjects
S u g g e s t e d by His Life and Work, Cam-
bridge U n i v e r s i t y Press, 1940, Chap-
ter 2.
(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 num-
bers 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 proba-
bility for n and n+2 m o d u l o p to be
prime twins differs by a factor
p-2 2
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. Al-
together, we have therefore improved
our c h a n c e s by a factor
p2 _ 2p
C = 2 p>2 p2 _ 2p + I
dt
p prime
careful p r e s e n t a t i o n of this argu-
ment, see Hardy and Wright, An In-
t 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
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.
(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 com-
p 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.
(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 dis-
cuss 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). This b o u n d
has b e e n l o w e r e d to
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
w h e r e ~(x) § 0 as x t e n d s to infin-
ity 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.
(16) 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 Primzahltheo-
rie 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 ] - [~ - [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) = I + 1 + I_~ + ...
2s 3s
- 2 + 4-~ 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 "in-
t e r e s t i n g " r o o t s o f the zeta func-
tion, 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 equa- The 1977 Mathematics Calendar has found a
tions worthy successor:

n~ I (-I)n-I c o s ( y l o g n) = O,

Psychology
n8

n~ 1 (-I)n-I sin(ylog n) = O.
n8

The sum over the roots p in Rie-

mann's formula is not a b s o l u t e l y
Calendar '78
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 containing a selection of twelve themes
value of Im(p)). which p o r t r a y something of the exciting
Finally, I should m e n t i o n that, spirit w h i c h pervades p s y c h o l o g i c a l in-
a l t h o u g h Riemann stated the exact quiry:
formula for ~(x) already in 1859,
it w a s n ' t proved until 1895 (by von
Mangoldt). Dreaming
Cognitive Contours
(19) These roots were calculated a l r e a d y Social D o m i n a n c e in Monkeys
in 1903 by Gram (J.-P. Gram, Sur M c C o l l o u g h Effect
les zeros de la f o n c t i o n ~(s) de
Riemann, Acta Math. 27 (1903) 289 - Dynamic Effects of Repetitive Patterns
304). For a v e r y nice p r e s e n t i o n of Harlow and Piaget
the theory of R i e m a n n ' s zeta func-
tion, see H.M. Edwards, Riemann's Shadow and Light
Zeta Function, A c a d e m i c Press, N e w Memory Span Test
York, 1974.
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
(20) Namely the R i e m a n n hypothesis im-
plies (and in fact is e q u i v a l e n t to Split Brain
the statement) that the error in Jerome Bruner's Theory of Cognitive
Gauss's a p p r o x i m a t i o n Li(x) to ~(x) Learning
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.

(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 Rie-
m 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.