You are on page 1of 8

What Is the Most Surprising Result in Mathematics?

Author(s): TOM M. APOSTOL


Source: Math Horizons, Vol. 4, No. 2 (November 1996), pp. 8-14
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/25678086 .
Accessed: 14/01/2015 21:33

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

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

Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access to
Math Horizons.

http://www.jstor.org

This content downloaded from 138.100.4.44 on Wed, 14 Jan 2015 21:33:11 PM


All use subject to JSTOR Terms and Conditions
TOMM. APOSTOL

What Is the Most


Surprising

I have posed this question tomany mathematicians, and


it's no that I have heard many different
perhaps surprise
are a few of the more intriguing ones.
replies. Here

of t
Example1. The ubiquity
Because tt is defined as the ratio of the circumference
to the diameter of any circle, it's not surprising that tt
occurs in many formulas for areas and volumes of fig
Figure 1
ures that are generated by circles. But it is surprising A One-to-one between in a short
correspondence points
that 7r shows up in many situations that have nothing interval and a
long interval.
to do with circles. For example, the sum of the squares
of the reciprocals of the positive integers is 7r2/6.The
sum of their fourth powers is 7T4/90, and the sum of intermingling of decimal expansions, and can be shown
their sixth powers is 7r6/945. In 1736 Euler showed that as follows.
for any even exponent 2k the sum of the 2?;th powers To map a point (x, y) of the unit square to a point
z in the unit interval, write each of x and y in their
of the reciprocals of the positive integers is 7r2kmulti
decimal expansions. For example, is
plied by a rational number. The rational factor is, in nonterminating |
as as .Now
fact, (-l)*"1^ times 22k~1 divided by (2k)\, where the written 0.24999... , not 0.25000... suppose
numbers B2k are the Bernoulli numbers. we have
There are many other occurences of tt. For exam
x = 0.3 06 007 8 9... and y
= 0.001 2 3 004 6....
curve
ple, the2 area of the region under the Gaussian
? e~x is the tt. tt
y square root of The number also ap
Separate each decimal expansion into groups of inte
pears in many probability problems. For example, the gers, where the last integer in each group is nonzero.
probability that two integers chosen at random will have If you run out of zeros, put one digit in each group.
no common is 6/tt2. There are many
prime factor in Now interleave the two expansions by taking alternate
more examples, including equations in quantum me groups to form the nonterminating decimal
chanics where tt is often lumped together with Planck's
constant. z = 0.3 001 06 2 007 3 8 004 9 6....

It's an easy exercise for the reader to show that this


Example 2. gives
a one-to-one
mapping of the unit square onto the
unit interval. And you can use a similar argument to
(0,l]~(0,l]x(0,l]~(0,l]x(0,l]x(0,l]
show that the unit cube can be mapped in a one-to-one
This says there is a one-to-one correspondence between fashion onto the unit interval.
in a unit interval and points in a unit square, or in
points
a unit cube. As a young student I was surprised to learn
that a short interval has the same number of points as
a saw the correspondence in
long one, until I displayed
1. But it seems counter intuitive that an inter
Figure
val has the same number of points as a square. Here the
correspondence is less obvious (Figure 2). It involves

Figure 2
TOM M. APOSTOL, emeritus professor of mathematics at Cal There is a one-to-one
correspondence
between points
in a

tech, is director of Mathematics!. unit interval and in a unit or in a unit cube.


Project points square,

8 MathHorizonsNovember1996

This content downloaded from 138.100.4.44 on Wed, 14 Jan 2015 21:33:11 PM


All use subject to JSTOR Terms and Conditions
Result in Mathematics?

The next surprising result is somewhat related to


function cp(t)
the previous one.

Example 3. Existence of space-filling curves


-2-101234
The element of surprise here may be age-related. As a
student many years ago I was stunned to learn that there Figure 4
of the saw-tooth function ip.
are continuous curves that pass through every point of Graph

the unit square. But today many schoolchildren are able


to generate fractals and space-filling curves with their
Define x(t) and y{t) in terms of ipby the following
computers so this probably doesn't seem as surprising
to them as it did to me. There are explicit recipes for equations:

curves. Here is one of my


constructing such space-filling
favorites; itwas concocted by I. J. Schoenberg in 1938.
n=l n=l
Let <pbe a saw-tooth function of period 2 that is
?
symmetric about the y axis, has the constant value (p(t) It can be shown that these are continuous functions of t,
0 on the interval [0, the constant value (f(t) = 1 on
|], and, as t varies from 0 to 1, the point (x(t),y(t)) passes
= ? 1 on
[?, 1], and is linear, ip(t) 3t [|, |]. Its graph is
x
through every point of the unit square [0,1] [0,1]. In
shown in Figure 4. other words, x(t) and y(t) provide a parametric repre
sentation of a continuous curve that fills the unit square.
However, such a continuous curve cannot be one-to
one.

Example 4. The Banach-Tarski paradox

This theorem says that a solid sphere can be divided


into a finite number of pieces (5 will do) that can then
be rearranged to form two new spheres, each of which
has the same volume as the original sphere. Needless
to say, the pieces are nonmeasurable sets and are hard
to visualize. Of course, each new sphere can be divided
a total of four new
again, giving us spheres, each having
the same volume as the original sphere. And the process
can be indefinitely. The proof of the Banach
repeated
Tarski paradox shows that such a dissection exists in the
abstract world of mathematics, but it does not explain

Figure 3 Figure 5
An to a curve. The Banach-Tarski
approximation space-filling paradox

MathHorizonsNovember1996 9

This content downloaded from 138.100.4.44 on Wed, 14 Jan 2015 21:33:11 PM


All use subject to JSTOR Terms and Conditions
how one could carry it out in practice on a real sphere that there are infinitely many primes. There are many
in the real world. If this dissection could actually be car proofs of this theorem, but Euclid's original proof is the
ried out in practice itwould imply that you could start most elegant. It is a proof by contradiction that goes
with a single spherical pea and, after a finite number of as follows. Suppose there were only a finite number of
dissections, produce enough new peas to fill the sun. primes, and let P denote their product. Look at the
number Q = P + 1. Since Q is greater than 1 itmust
be divisible by some prime occuring in the product P
Example5. Stirling'sestimateforfactorials P is
(because the product of all the primes). But this is
because such a would also divide their
impossible prime
^nn+\j2e-n < nj < + ? P = 1.
difference Q
y/^nn+l/2e-n^i
A more sophisticated proof of Euclid's theorem was
we see the appearance of our old friend
Again given by Euler who showed that the series of the recipro
7r together with another ubiquitous constant, Euler's cals of the primes diverges. Every freshman calculus stu
number e. It's truly surprising that the product of all dent learns that the harmonic series
Y^=i n diverges.
the positive integers up to n is very nearly equal to If you discard all the composite numbers in this series
? as n ?> oo the
V/27rnn+1//2e_n.Because 0 approxi there are still enough terms left over tomake the series
^
mation becomes more exact as n ?> oo.
diverge.
The can surely be regarded as A question that presents itself at the very threshold
foregoing examples
surprising results inmathematics. Clearly, the answer to of mathematics is this: How are the primes distributed
the question in the title is a matter of opinion. In the rest the integers? Detailed examination of a table of
among
of this article I would like to explain why my answer is reveals great irregularities in distribution. Some
primes
a result about prime numbers called the prime number are very close
primes together, like 3 and 5; 11 and 13;
theorem. To lay the groundwork, let me begin with a 17 and 19; or 59 and 61; these are examples of pairs of
few remarks about prime numbers. twin primes, primes that differ by 2. Pairs of twin primes
keep occuring as far as the tables go. Table 1 contains
aboutprimenumbers
Backgroundinformation some data about twin
primes.
To date, no one knows whether twin primes even
are
Physicists study the fundamental particles because they tually cease to exist or if there infinitely many twin
are the building blocks of the physical world. The build prime pairs. In 1919, Viggo Brun proved that the sum
world are the prime of the reciprocals of the twin primes converges with sum
ing blocks of the mathematical
numbers. A prime is an integer greater than 1 whose 1.90216054_This expresses the scarcity of the twin
even if there are infinitely many of them.
only divisors are 1 and the number itself.The first few primes
. At the other extreme, there are huge gaps between
primes are listed here: 2,3,5,7,11,13,17,19,23,29,...
Integers greater than 1 that are not prime
are called consecutive primes. It is easy to show that arbitrarily
Each composite number is a product of large gaps exist between primes. Choose any integer n >
composite.
The number 1 is neither nor 1 and look at the n? 1 consecutive numbers n! + 2, n! + 3,
primes. prime composite.
Prime numbers derive their importance from a the n! + 4,...,n! + n. These are all
composite (the first is
orem of Gauss called the fundamental theorem of arith divisible by 2, the second by 3, the third by 4, etc.), and
since n is arbitrary thismeans there must be arbitrarily
metic, which states that every integer n > 1 can be fac
we see that
tored as a product of primes in one and only one way large gaps between consecutive primes. So
tables reveal primes close together and also primes very
(apart from the order of the factors). The factorization
of integers into prime factors was known tomany math far apart. On the average the primes appear to become
more widely spaced in the higher parts of the table.
ematicians before Gauss, including the ancient Greeks,
but the uniqueness of the factorization was first ex Euclid's theorem on the infinitude of primes can be
stated another way. Arrange the primes in increasing,
plicitly stated by Gauss in his Disquisitiones Arithmeticae,
Sec. 16. Gauss deduced the fundamental theorem from order and let pn denote the nth prime, so that pi = 2,
= = can regard pn as a function of n.
30 in Book 7 of Euclid's Elements, which p2 3? p3 55_We
Proposition
states as n?>oo. How fast
states that if a prime divides a product of two integers, Euclid's theorem that pn-+oc

then itmust divide at least one of the factors. does pn go to infinity? Does it grow like a constant
Another important result about primes found in Eu times n, a constant times y/n, a constant times log n, or
clid's Elements is Proposition 20 in Book 9, which states what?

x I 103 | 104 | 105 | 106 1 107 1 108 1 109 I 1010 I 1011 "
# I 35 | 205 1 1,224 [ 8,169 [ 58,980 [ 440,312| 3,424,506
| 27,412,679
[ 224,376,048
Table 1
In this table, # = number of twin prime pairs < x

10 MathHorizonsNovember1996

This content downloaded from 138.100.4.44 on Wed, 14 Jan 2015 21:33:11 PM


All use subject to JSTOR Terms and Conditions
ObservationsofGauss and Legendreon the
N0MBll? y nombhb/
ofprimes
distribution Limitbx. Limite x- ? -?,

Par la furtuule.Par les tables. Par la formulc.Par les tables.

A lot of people have studied this question, starting with


ioooo i23o i23o 200000
Gauss in 1791 when he was only 14 years old. Actu 1798a 17984
20000 2268 2263 250000 22035 22045
ally, the young Gauss considered a different function,
3oooo 325a 3ooooo 26023
the number of primes less than or equal to x, which is 3246 25988

now denoted by tt(x). (Here's 7r as a 40000 4^?5 4204 35oooo 29961 29977
again, but symbol
a 5oooo 5136 400000 33854 33861
for a function, not number.) Gauss examined a table 5134
of primes compiled by Lambert. By the way, Lambert 60000 6049 6o58 5ooooo 4i533 4i538 II
was the firstmathematician to prove that tt is irrational. 70000 6949 6y?)6 600000 49096 49?9^
Gauss counted the number of primes in blocks of 100, 80000 700000 56565 56535
7838 7837
1000, and 10,000 consecutive integers and made a note 90000 8717 8713 800000 63955 ^937
in his diary that the function 1/ log x was a good approx 100000 9588 9592 900000 71279 71268
imation to the average density of distribution (number 15oooo 13844 13849 1000000 78543 78493
of primes per unit interval). A few years later in 1796,
when Vega published an extended table of primes up 6
Figure
to 400,031, Gauss substantiated his hypothesis further A table from Legendre's book.

and kept returning to this work as new tables of primes


appeared. Many years later, in 1849, he communicated
his observations in a letter to the astronomer Encke, and
the results were published posthumously in 1862. Based I'd like tomake a small detour here to tell you an inter
on tables listing to 3 million, Gauss observed esting story about that book. The copy in the archives
primes up
came from the collection of Eric
that 7r(x) is closely approximated by the integral of the Temple Bell, who was
function: Professor of Mathematics at Caltech from 1926 until his
density
retirement in 1953. Inside the book you'll find the fol
fx 1 own
7r(x)
~
/ :-dn. lowing inscription in Bell's handwriting. It says:
"This book survived the San Francisco Earthquake and
and he compared the value of the integral with the value Fire of 18 April, 1906. It was buried with about 600
of 7r(x) obtained by counting primes in the tables. He others, in a vacant lot, before the fire reached the spot.
that this integral is a good approximation The house next door to the lot fell upon the cache; the
conjectured
tar from the roof baked the 4 feet of dirt,
for all large values of x. Some excerpts from Gauss's covering the
tables can be found in Figure 7. books, to brick, and incinerated all but 4 books, of which
also considered this question in the 2nd this is one. Signed: E. T. Bell. Book buried just below
Legendre
edition of his number theory text, published in 1808, Grace Church, at California and Stockton Streets. House
and on the basis of his observations of tables of primes he number 729 California Street."
asserted that n(x) is closely approximated a Neither Gauss nor Legendre revealed how he ar
by slightly
different formula: rived at the appearance of the natural logarithm in
their formulas. Nor did theymake any explicit statement
~ about how good these approximations were thought to
^
logx- 1.08366'
be outside the range of the tables. It is generally un
Figure 6 shows a table from Legendre's book displaying derstood that both intended to imply that the ratio of
this approximation for various x up to a million. ?r(x) to each approximating formula on the right tends
Table 2 shows how ir(x) compares with what you get to the limit 1 as x tends to infinity. Integration by parts
from the formulas of Gauss and of Legendre. shows that the value of Gauss's integral is approximately
There is a copy of Legendre's number theory book so another of
x/logx, way stating Gauss's conjecture is
in the rare book collection of the Caltech archives, and to say that tt(x) is about x/logx. Both Gauss's and
Leg
endre's relations are equivalent to the simpler relation

x 7r(x) Gauss error error


Legendre 1 as x
7r(x)/r~-> 00,
10s 78,498 + 45
+130 logx
2 x 106 148,933 122
43~ which later came to be known as the prime number the
3 x 106 216,816 155
97~ orem. In my opinion this is one of the most
astonishing
5 x 106 129
125 results in all of mathematics. It describes a simple inti
348,513
mate relation between prime numbers and the natural
107 664,580 339
561
logarithm function which, at first glance, has nothing to
Table 2 do with prime numbers.

MathHorizonsNovember1996 11

This content downloaded from 138.100.4.44 on Wed, 14 Jan 2015 21:33:11 PM


All use subject to JSTOR Terms and Conditions
.- .11
i 16S 51 89 j xot 81 151 85 aoi 77 I a$x 71 jot 85 351 74 4oz 70 451 9*
* 135 5* 97 3to?93 X5? 90 soa 87 451 88 30* 83 35a 80 40* 71 45* 76
3 "7 53% 103 87 153 88 003 78 453 7* 3<%7* 353** j 403*76 453 63
4 lao 54 9* *?* 80 *54 77 *?4 78 *54 8x 304.84 354 79 404 75 j454 7*
5 ?9 55 $0 ">S 91 15584 905 77 *5S ?6 305 _88iM^lppi^M^'
? ?4 56 93 106 ** 85 *?6 85 S?6 80 "i3t':|9?i
| 4|6 87 In^i l^^i;
7 117 53 99 *?7 9* 157 76 ??7 83 457 7* 307 8* 357 If mmftl sills:
8 107 58 91 i?8 76 15888 *o8 87 158 78 308 73 358 80 408 ?f 458 77
9 110 59 9?> *?9 9* *5987 109 85 45986 309 76 -359-tj iii#|#=B?^fiS
10 xx4 60 94 no 88 x6o85 wo 88 a6o 76 310 80 360p |ili|p ?0Gff0tm
ix xo? 6x 88 xxi 83 x6x85 *ix 84 *6x 77 3x1 79
14 xeg 87 xxa 84 16184 .41*86 464 .73 3x4 69 .^^WiM^^liMWi^
W^Sfe llifWt?'
13 X09 63 88 ?3 81 163 81 4x3 69 463 79 3*3 86 3%it.lllSll?.V^gim%
,14105 6493 ?4 88 16483 41481 46484 31486 36484 41469 4647^
<5 10a 65 80 115 8a 16577 ' 415 86 465 80 3x576 465 77 415,90 4*5 85
x6 xo8 66 98 ! xx693 x66So 416 74 466 78 316 77 36677 4x6 80 "46674
17 98 67 84 1*? 8x 167 8x .417 76 467 87 3x784 367 85 4*7 6? 467 69
18 X04 68 99 xx890 x6883 4x8 80 468 94 3*8 84 368 79 4?$ .** 468 83
19 94 69 80 xi9 79 16973 m 84 *% 75 3*9 8* 369m vmm*w^mM:
40x0* 7O8X 190 87 X7087 440 91 47078 S^M::;:^^g||^^^|^p^:
41 98 7* 98 *? 88 17x87 41x78 47184 3" 79 37* 70 4?**7S 1*7*87;
' 44x04 7*' 95 *4? 86 ,174,81 44480 47438 344 80 374 76. '444 73 .'
47478
43 zoo .73 90*V*i 88. 173W %i$$tXm'*$ -m^'i 37i--??!iliiiisiii!niiiiiiii
.\ 44 xc* 74 83 144 88. Ii74 79 ?4 80 474 7X ;
344 71-rj74^f3:':?m09$!W?HSm
*$Z9* 75 9*- "5.83 *75 83 ;?5 83 *75 80 3*5 tf.<
-p^M^^f^ ffgg^
46^98 7691' X46'84 17675 >46 84/356 83^ ii.'^M'^miu^^K.
-a?*?* 77 83 X4783 17795 aa? 76 477^83 |?7 7$ .m^mO^^mMm.
,48 94 7895 *a8 86 17873 448 8o#478 74 :.ilft^' ^MM^
: 49 98 79 84- "9 897 *?9 ?9"'m 89 *79 8*: 3*9 7imM/wm^al^i^m
30 94 80 91 13083. 18094, 43088 480 73 8i 38083 430 89 .480.
33f
3* '95 8x 88. 13x8$ lii 71 431 84 481 87, 33? 80 j8i -ml|#l?|?
34 94 8494 134 83 x8*,79' 434-78 48*85"* -33* 8*, |84^iiiiiiBli^l^^ i?!llil|
'33106 83 ?9 *33 8? X8391 43376 48377 7*-\m^MW^S^^t^l^K^
34 ioo 84 84 134 8s 18479 *34 7* ?84 7* 334 80 384 8x 434 80 484 ?s :
3S 94 8587- 13S 80- 18583 435 87
36 94 86 85 X3689 18691 '43673 486 77 336 77 386 80
:^MmM^M?.
37 99 87 88 X379? 79 j *37 76 :*&77* 337 84 387 78 437 73 "4*784
38 94 8893 138 80 x8887 &3873 *88 71 338 80 388%-4#m^^^^m
39 90 89 76 "*39 85 18980* 439 87 **9 85 339 77 389 75 439 75 489 83,
40 96 90 94 X4084 19088 I440 79 490 ?* .:a^b-*|--
IJWS^^P^I^^
4* 88 9x89 14*87 19175144x80 491-84
,44x01 9*85 X44 87 19481 44491.49*77 'p f; Mi-;
43 xo? 93 97 X438a 193 89 443 76 493 78 ;lllfs;l|! |?9i?lll iMillJif::
Wm^K^
44 85 94 86 14477 *9+ 84 444 77 *94 68 344 80 394 87 444 88 494 84
45 96 *9587 14579 *9574 445 78 *95 85
46 86 9695 X468| 19685 44680 49675 ^^j^c^m^W^^Wi^^
47 90 97 84 X47.84 X9776 H7 84 *97 8* 347 So 397 75 447 68 497 68
48 95 9884 14883 19887:448791 & n ti&^tov :j&- 'm
49 89 99 87 X4983 19996 449 88 499 73 349 77
50 98 too 87 150 91 400 77 450 80 350 82 1 400 8x 450 80
300 78 500 81

Figure 7
Excerpts from Gauss's tables. Primes counted by Gauss in intervals of length 1000.

12 MathHorizonsNovember1996

This content downloaded from 138.100.4.44 on Wed, 14 Jan 2015 21:33:11 PM


All use subject to JSTOR Terms and Conditions
How toconjecturetheprimenumbertheorem
by had access to tables that only went up to 3 million. We

examininga tableofprimes have added the last four entries from more modern ta
bles. What can we learn by looking at these numbers?
I always wondered what led Gauss and Legendre to use Since we want to find how fast tt(x) grows with x, it is
the natural logarithm function in their formulas. They natural to look at the ratio x/tt(x) that compares the two
did not leave us any written clues. They just recorded quantities. Table 4 shows the corresponding ratios. No
their formulas and supporting data. I also wondered tice the gaps between successive entries in the last row
< x rather of numbers in Table 4.
why they studied n(x), the number of primes
than pn, which seems to be the more natural function Wesee that as each exponent of 10 increases
by 2,
to study. Actually it's easy to find a connection between the ratio x/tt(x) increases by an almost constant amount,

7r(x) and pn. 4.6, which is 2.3 times the change in the exponent of 10.
Take any large integer x. It lies between two con But ifx is expressed as a power of 10, then the exponent
secutive primes, say of x is the logarithm of x to the base 10. So the table
indicatesthat the change in the ratio x/tt(x) is approx
Pn < x < pn+U imately equal to 2.3 times the change in log10x. What
about this strange factor 2.3? A bright fourteen-year old
where, n, the index on pn, is just tt(x), the number of
such as Gauss, or even a bright Caltech freshman would
< x. Remember, our was this:
primes original question
How does pn grow as a function of n? We can ask a immediately realize that the factor 2.3 is very nearly the
more x on n? If logarithm of 10 to the base e, so
general question: How does depend
we can answer this for every x between pn and = =
question 2.31og10 x (logc 10)(log10 x) loge x.
pn+i then we can also answer our original question by
simply taking
x = pn At first
glance it seems that all
we This suggests that we compare the ratio x/tt(x) with
have done is ask a harder question. A table of primes logx (the natural logarithm of x). This comparison is
will give us an idea about how n depends on x, because shown in Table 5.
n is just 7r(x). If we can find out how n depends on x, Anyone who looks at this last row of numbers would
then maybe we can turn this around and find how x surely be tempted to conjecture that they approach 1
on n. as x ?> oc. Gauss, and other eminent
depends Legendre, many
So we are led to study the function tt(x), just as mathematicians of the early 19th century apparently
Gauss and Legendre did. The advantage of studying thought so, but they were not able to prove it.As far
as we know, neither Gauss nor
7r(x) is that we can tabulate this function very easily Legendre made any sig
a table of count how many nificant progress toward a proof.
by looking at primes?just
are < like Gauss and did. The first real step towards a proof was made in
primes there x, Legendre
x.
Table 3 shows what you get for various values of Gauss 1848 by the Russian mathematician, P. L. Chebyshev.

~
x 11021104 I 106 I 108 11010 11Q12 11014
tt(x) 25 1,229 78,498 5,761,455 455,052,512 37,607,912,018 3,204,941,750,802
Table 3

'
x 1102 I 104 I 106 ~I 108 1101Q I 1012 11014
rr(x) ~25 1,229" 78,498 5, 761,455 455,052,512 37,607,912,018 3,204,941,750,802
4.000 8.137 12.739 17.357 21.975 26.590 31.202
7T(x)

gaps: 4.137 4.602 4.618 4.618 4.615


4.612
Table 4

x 1 102 1 104 1 106 1 108 1 101Q I 1012 1 1014


logx 4.605 9.210 13.816 18.421 23.026 27.361 32.236

l?g* 1.151 1.132 1.085 1.061 1.048 1.039 1.033


x/7t{x)
Table 5

MathHorizonsNovember1996 13

This content downloaded from 138.100.4.44 on Wed, 14 Jan 2015 21:33:11 PM


All use subject to JSTOR Terms and Conditions
He proved that if the ratio n{x){logx)/x has a limit as ,
Western Michigan UNiwi^ny \-;
x ? oc, then this limitmust 1. But was un
equal Chebyshev
able to prove that this ratio actually tends to a limit.And
in 1850 he proved that this ratio lies between two nu
merical constants a and A for all sufficiently large x, so The Department ofMathematics and Statistics atWestern Michigan Uni
the ratio x/logx does, indeed, represent the true order versity consists of 35 faculty members with specialties inmost areas of

of magnitude of 7r(x). The problem mathematics, mathematics education, and statistics, with about 50 gradu
is difficult because
ate assistantships and doctoral associatesriips.Western Michigan Univer
there is no useful formula that generates the primes, so
sity is located in Southwestern Michigan, midway between Chicago and
it's hard to obtain information about their distribution. Detroit, and less than one hour's drive from Lake Michigan.
The next installment of this article will explain how
Degree Programs The Department has Ph.D. programs inMathematics,
G. F. B. Riemann attacked the problem with a new
Mathematics Education, and Statistics; and Master's degree programs in
method, based on a formula that Euler had discovered a Mathematics, Applied Mathematics, Computational Mathematics, Math
century earlier. Although Riemann came close to prov ematics Education, Statistics, and Biostatistics. Our graduate students
the number he never succeeded. receive a lot of individualized attention and encouragement.
ing prime theorem,
It took three more decades to develop the mathemat Financial Assistance In 1996-97, Graduate Assistantships and Doctoral
ics that was needed to carry out his methods. Part two Associateships carry a stipend of $10,000. An increase is expected for
of this article will tell part of the story of how this was 1997-98. Associateships carry tuition waivers. Assistantships pay in

in 1896, exactly one century ago. state tuition. Additional support may be awarded for Spring or Summer
finally accomplished sessions. Applications are due 15 February 1997. Late applications are
The next installment will appear in the February as long as openings remain. For additional information con
accepted
1997 Horizons. tact:

Chair, Graduate Committee phone: 616-387-4510


Dept. ofMathematics & Statistics fax: 616-387-4530
Western Michigan University email: grad@math-stat.wmich.edu
Kalamazoo, MI 49008-5152 url: www.wmich.edu/math-stat/

WMU is an Equal Opportunity/Affirmative Action Institution.

JL? $16,000 and $12,000 Fellowships


TQ1
*WvJ>jJLi
THEUNIVBRSITYOF SOUTHWESTERN Scholarships and Assistantships
LOUISIANA

The Department ofMathematics at theUniversity of Southwestern Louisiana offers


graduate
programs with emphases on applied mathematics, pure mathematics, and statistics. It has
fellowships, scholarships and assistantships for students seeking a Ph.D. degree. The stipends
per fiscal year are $16,000 for a Board of Regents Fellowship and $12,000 for a University
Fellowship. Both are renewable up to four years. In addition to assistantships, scholarships of
various amounts can be awarded. Such scholarships are renewable up to a maximum of five
years. Besides having waiver of tuition and fees (except theGraduate Student Organization fee
(currently $5.50 per semester)), low-cost housing is available to a fellowship or scholarship
holder.Applicationsmust be completedby February 15, 1997 forfellowshipsor scholarships,
and preferably byMarch 15 for assistantships. Although there is no deadline for
assistantships,
are GRE
early applications encouraged. High (verbal, quantitative, and analytical) scores and
outstanding academic achievement are required. Please contact:
Dr. C. Y. Chan, Head
DepartmentofMathematics Telephone: (318) 482-6702
University of Southwestern Louisiana E-mail: math@usl.edu
Lafayette,LA 70504-1010 Fax: (318) 482-5346

14 MathHorizonsNovember1996

This content downloaded from 138.100.4.44 on Wed, 14 Jan 2015 21:33:11 PM


All use subject to JSTOR Terms and Conditions

You might also like