1

The Life of Pi: From Archimedes to Eniac and Beyond

Jonathan M. Borwein, FRSC
Prepared for Berggren Festschrift Draft VIII. 19/06/2012
Laureate Professor & Director CARMA
Research currently supported by the Australian Research Council.
Email: [email protected]

1 Preamble: π and Popular Culture

The desire to understand π, the challenge, and originally the need, to calculate ever more accurate values of π,
the ratio of the circumference of a circle to its diameter, has challenged mathematicians–great and less great—for
many centuries. It has also, especially recently, provided compelling examples of computational mathematics.
π, uniquely in mathematics, is pervasive in popular culture and the popular imagination.

I shall intersperse this largely chronological account of π’s mathematical status with examples of its ubiquity.
More details will be found in the selected references at the end of the chapter—especially in Pi: a Source Book
[9]. In [9] all material not otherwise referenced may be followed up upon, as may much other material, both
serious and fanciful. Other interesting material is to be found in [21], which includes attractive discussions of
topics such as continued fractions and elliptic integrals.

Fascination with π is evidenced by the many recent popular books, television shows, and movies—even
perfume—that have mentioned π. In the 1967 Star Trek episode “Wolf in the Fold,” Kirk asks “Aren’t there
some mathematical problems that simply can’t be solved?” And Spock ‘fries the brains’ of a rogue computer by
telling it: “Compute to the last digit the value of Pi.” (Figure 1 illustrates how much more is now known, see
also http://carma.newcastle.edu.au/piwalk.shtml.) The May 6, 1993 episode of The Simpsons has the
character Apu boast “I can recite pi to 40,000 places. The last digit is one.”

In November 1996, MSNBC aired a Thanksgiving Day segment about π, including that scene from Star
Trek and interviews with the present author and several other mathematicians at Simon Fraser University. The
1997 movie Contact, starring Jodie Foster, was based on the 1986 novel by noted astronomer Carl Sagan. In
the book, the lead character searched for patterns in the digits of π, and after her mysterious experience found
confirmation—that the universe had meaning— in the base-11 expansion of π. The 1997 book The Joy of Pi [11]
has sold many thousands of copies and continues to sell well. The 1998 movie entitled Pi began with decimal
digits of π displayed on the screen. And in the 2003 movie Matrix Reloaded, the Key Maker warns that a door
will be accessible for exactly 314 seconds, a number that Time speculated was a reference to π.

As a striking example, imagine the following excerpt from Yann Mandel’s 2002 Booker Prize winning novel
Life of Pi being written about any other transcendental number:

“My name is
Piscine Molitor Patel
known to all as Pi Patel.

For good measure I added

π = 3.14
and I then drew a large circle which I sliced in two with a diameter, to evoke that basic lesson of
geometry.”

Equally, National Public Radio reported on April 12, 2003 that novelty automatic teller machine withdrawal
slips, showing a balance of $314, 159.26, were hot in New York City. One could jot a note on the back and,
1 This paper is an updated and revised version of [13] and is made with permission of the editor.

1
apparently innocently, let the intended target be impressed by one’s healthy saving account. Scott Simon, the
host, noted the close resemblance to π. Correspondingly, according to the New York Times of August 18 2005,
Google offered exactly “14, 159, 265 New Slices of Rich Technology” as the number of shares in its then new
stock offering. Likewise, March 14 in North America has become π Day, since in the USA the month is written
before the day (‘314’). In schools throughout North America, it has become a reason for mathematics projects,
especially focussing on π.

In another sign of true legitimacy, on March 14, 2007 the New York Times published a crossword in which
to solve the puzzle, one had first to note that the clue for 28 down was “March 14, to Mathematicians,” to
which the answer is piday. Moreover, roughly a dozen other characters in the puzzle are pi—for example, the
clue for 5 down was “More pleased” with the six character answer hapπer. The puzzle is reproduced in [14].

Other and more recent examples—including the US Congressional House Resolution 224 designating of
National Pi Day in 2009—may be found at http://www.carma.newcastle.edu.au/jon/piday.pdf which is
annually updated.

It is hard to imagine e, γ or log 2 playing the same role. A corresponding scientific example [4, p. 11] is

“A coded message, for example, might represent gibberish to one person and valuable information to
another. Consider the number 14159265... Depending on your prior knowledge, or lack thereof, it is
either a meaningless random sequence of digits, or else the fractional part of pi, an important piece
of scientific information.”

Again, a scientist can use π confident that it is part of shared societal knowledge—even if when pressed his
definition might not be as good as Pi’s.

For those who know The Hitchhiker’s Guide to the Galaxy, it is amusing that 042 occurs at the digits ending
at the fifty-billionth decimal place in each of π and 1/π—thereby providing an excellent answer to the ultimate
question of ‘life, the universe and everything‘, which was “What is forty two? ” A more intellectual offering is
“The Deconstruction of Pi” given by Umberto Eco on page three of his 1988 book Foucault’s Pendulum, [9, p.
658]. The title says it all.

Pi. Our central character

π = 3.14159265358979323 . . .
is traditionally defined in terms of the area or perimeter of a unit circle, but see Figure 4 where the subtlety
of showing the two are the same is illustrated. The notation of π itself was introduced by William Jones in
1737, replacing ‘p’ and the like, and was popularized by Leonhard Euler who is responsible for much modern
nomenclature. A more formal modern definition of π uses the first positive zero of the sine function defined as
a power series. The first thousand decimal digits of π are recorded in Figure 2.

Despite continuing rumours to the contrary, π is not equal to 22/7 (see End Note 1). Of course 22/7 is one
of the early continued fraction approximations to π. The first six convergents are
22 333 355 103993 104348
3, , , , , .
7 106 113 33102 33215
The convergents are necessarily good rational approximations to π. The sixth differs from π by only 3.31 10−10 .
The corresponding simple continued fraction starts

π = [3, 7, 15, 1, 292, 1, 1, 1, 2, 1, 3, 1, 14, 2, 1, 1, 2, 2, 2, 2, 1, 84, 2, 1, 1, . . .],

using the standard concise notation. This continued fraction is still very poorly understood. Compare that for
e which starts
e = [2, 1, 2, 1, 1, 4, 1, 1, 6, 1, 1, 8, 1, 1, 10, 1, 1, 12, 1, 1, 14, 1, 1, 16, 1, 1, 18, . . .].

2
Figure 1: A 100 billion step planar walk on the binary digits of π. Colors change from red and orange to violet
and indigo as we proceed (snapshot taken from image in [1])

3 . 1415926535897932384626433832795028841971693993751058209749445923078164062862089986280348253421170679
8214808651328230664709384460955058223172535940812848111745028410270193852110555964462294895493038196
4428810975665933446128475648233786783165271201909145648566923460348610454326648213393607260249141273
7245870066063155881748815209209628292540917153643678925903600113305305488204665213841469519415116094
3305727036575959195309218611738193261179310511854807446237996274956735188575272489122793818301194912
9833673362440656643086021394946395224737190702179860943702770539217176293176752384674818467669405132
0005681271452635608277857713427577896091736371787214684409012249534301465495853710507922796892589235
4201995611212902196086403441815981362977477130996051870721134999999837297804995105973173281609631859
5024459455346908302642522308253344685035261931188171010003137838752886587533208381420617177669147303
59825349042875546873115956286388235378759375195778185778053217122680661300192787661119590921642019893

Figure 2: 1,001 Decimal digits of π with the ‘3’ included

Figure 3: A pictorial proof of Archimedes’ inequalities

3
 Figure 4: Construction showing uniqueness of π, taken from Archimedes’ Measurement of a Circle

A proof of this observation shows that e is not a quadratic irrational since such numbers have eventually periodic
continued fractions.

Archimedes’ famous computation discussed below is:

10 10
(1) 3 <π<3 .
71 70
Figure 3 shows this estimate graphically, with the digits shaded modulo ten; one sees structure in 22/7, less
obviously in 223/71, and presumptively not in π.

2 The Childhood of π

Four thousand years ago, the Babylonians used, among other values, the approximation 3 18 = 3.125. Then,
or earlier, according to ancient papyri, Egyptians assumed a circle with diameter nine has the same area as a
square of side eight, which implies π = 256/81 = 3.1604 . . . . Some have argued that the ancient Hebrews were
satisfied with π = 3:

“Also, he made a molten sea of ten cubits from brim to brim, round in compass, and five cubits
the height thereof; and a line of thirty cubits did compass it round about.” (I Kings 7:23; see also 2
Chronicles 4:2)

One should know that the cubit was a personal not universal measurement. In Judaism’s further defense,
several millennia later, the great Rabbi Moses ben Maimon Maimonedes (1135–1204), also known as the RaM-
BaM, is translated by Langermann, in “The ‘true perplexity’ [9, p. 753] as fairly clearly asserting the irrational
nature of π:

4
 “You ought to know that the ratio of the diameter of the circle to its circumference is unknown, nor
will it ever be possible to express it precisely. This is not due to any shortcoming of knowledge on
our part, as the ignorant think. Rather, this matter is unknown due to its nature, and its discovery
will never be attained.” (Maimonedes)

In each of these three cases the interest of the civilization in π was primarily in the practical needs of
engineering, astronomy, water management and the like. With the Greeks, as with the Hindus, interest was
centrally metaphysical and geometric.

Archimedes and π. Around 250 BCE, Archimedes of Syracuse (287–212 BCE) is thought to be the first to
show that the “two possible Pi’s” are the same. Clearly for a circle of radius r and diameter d, Area= π1 r2
while P erimeter = π2 d, but that π1 = π2 is not obvious, and is often overlooked. Figure 4. reproduces his
proof (construction) showing the coincidence of the two constants.

Archimedes’ Method. The first rigorous mathematical calculation of π was also due to Archimedes, who
used a brilliant scheme based on d oubling inscribed and circumscribed polygons
6 7→ 12 7→ 24 7→ 48 7→ 96
10
and computing the perimeters to obtain the bounds 3 71 < π < 3 17 , that we have recapitulated above. The
case of 6-gons and 12-gons is shown in Figure 5; for n = 48 one already ‘sees’ near-circles. Arguably no
computational mathematics approached this level of rigour again until the 19th century. Phillips [9, pp. 15-19]
calls Archimedes the ‘first numerical analyst’.

Archimedes’ scheme constitutes the first true algorithm for π, in that it is capable of producing an arbitrarily
accurate value for π. It also represents the birth of numerical and error analysis—all without positional notation
or modern trigonometry. As discovered severally in the 19th century, this scheme can be stated as a simple,
numerically stable, recursion, as follows [15].

√
Archimedean Mean Iteration (Pfaff-Borchardt-Schwab). Set a0 = 2 3 and b0 = 3—the values for
circumscribed and inscribed 6-gons. Then define
2an bn √
(2) an+1 = (H) bn+1 = an+1 bn (G).
an + bn
This converges to π, with the error decreasing by a factor of four with each iteration. In this case the error is
easy to estimate, the limit somewhat less accessible but still reasonably easy [14, 15].

Variations of Archimedes’ geometrical scheme were the basis for all high-accuracy calculations of π for the
next 1800 years—well beyond its ‘best before’ date. For example, in fifth century CE China, Tsu Chung-Chih
used a variation of this method to get π correct to seven digits. A millennium later, al-Kāshī in Samarkand
“who could calculate as eagles can fly” obtained 2π in sexagecimal:
16 59 28 01 34 51 46 14 50
2π ≈ 6+ + 2 + 3 + 4 + 5 + 6 + 7 + 8 + 9,
601 60 60 60 60 60 60 60 60
good to 16 decimal places (using 3·228 -gons). This is a personal favourite; reentering it in my computer centuries
later and getting the predicted answer gave me horripilation (‘goose-bumps’).

3 Pre-calculus Era π Calculations

In Figures 6, 8, and 11 we chronicle the main computational records during the indicated period, only com-
menting on signal entries.

5
 Figure 5: Archimedes’ method of computing π with 6- and 12-gons

Little progress was made in Europe during the ‘dark ages’, but a significant advance arose in India (450 CE):
modern positional, zero-based decimal arithmetic—the “Indo-Arabic” system. This greatly enhanced arithmetic
in general, and computing π in particular. The Indo-Arabic system arrived with the Moors in Europe around
1000 CE. Resistance ranged from accountants who feared losing their livelihood to clerics who saw the system
as ‘diabolical’—they incorrectly assumed its origin was Islamic. European commerce resisted into the 18th
century, and even in scientific circles usage was limited until the 17th century. This is, of course a greatly
simplified version of extraordinary events, for a recent article on the matter see [6].

The prior difficulty of doing arithmetic is indicated by college placement advice given a wealthy German
merchant in the 16th century:

“A wealthy (15th Century) German merchant, seeking to provide his son with a good business edu-
cation, consulted a learned man as to which European institution offered the best training. ‘If you
only want him to be able to cope with addition and subtraction,’ the expert replied, ’then any French
or German university will do. But if you are intent on your son going on to multiplication and
division—assuming that he has sufficient gifts—then you will have to send him to Italy.’ ” (George
Ifrah, [14])

Discussions about Roman arithmetic continue. Claude Shannon (1916–2001) had a mechanical calculator
wryly called Throback 1 built to compute in Roman, at Bell Labs in 1953 to show that it was practicable, if a
tad messy, to compute using Roman numerals!

Ludolph van Ceulen (1540–1610). The last great Archimedean calculation, performed by van Ceulen using
262 -gons—to 39 places with 35 correct—was published posthumously. The number is still called Ludolph’s
number in parts of Europe and was inscribed on his head-stone. This head-stone disappeared centuries ago but
was rebuilt, in part from surviving descriptions, recently as shown in Figure 7. It was reconsecrated on July
5th 2000 with Dutch royalty in attendance. Ludolph van Ceulen, a very serious mathematician, was also the
discoverer of the double angle formula for the cosine.

6
 Name Year Digits
Babylonians 2000? BCE 1
Egyptians 2000? BCE 1
Hebrews (1 Kings 7:23) 550? BCE 1
Archimedes 250? BCE 3
Ptolemy 150 3
Liu Hui 263 5
Tsu Ch’ung Chi 480? 7
Al-Kashi 1429 14
Romanus 1593 15
van Ceulen (Ludolph’s number∗ ) 1615 35

Figure 6: Pre-calculus π Calculations

4 Pi’s Adolescence

François Viéte (1540–1603). The dawn of modern mathematics appears in Viéte’s or Viéta’s product (1579)
√ √
√ √ √ √
2 2 2+ 2 2+ 2+ 2
= ···
π 2 2 2
considered to be the first truly infinite product; and in the first infinite continued fraction for 2/π given by Lord
Brouncker (1620–1684), first President of the Royal Society of London:
2 1
= .
π 9
1+
25
2+
49
2+
2 + ···
This was based on the following brilliantly ‘interpolated’ product of John Wallis2 (1616–1703):
∞
∏ 4k 2 − 1 2
(3) = ,
4k 2 π
k=1

which led to the discovery of the Gamma function (see End Note 2) and a great deal more. Variations on these
formulas of Vı́ete and Wallis continue to be published.

A flavour of Viéte’s writings can be gleaned in this quotation from his work, first given in English in [9, p.
759]. What we now take for granted was reason for much passionate argument.

“Arithmetic is absolutely as much science as geometry [is]. Rational magnitudes are conveniently
designated by rational numbers, and irrational [magnitudes] by irrational [numbers]. If someone
measures magnitudes with numbers and by his calculation get them different from what they really
are, it is not the reckoning’s fault but the reckoner’s.
Rather, says Proclus, arithmetic is more exact than geometry.3 To an accurate calculator,
if the diameter is set to one unit, the circumference
√ of the inscribed dodecagon will be the side of the
binomial [i.e. square root of the difference] 72 − 3888. Whosoever declares any other result, will be
mistaken, either the geometer in his measurements or the calculator in his numbers.” (Viéte)
2 One of the few mathematicians whom Newton admitted respecting, and also a calculating prodigy!
3 The capitalized phrase was written in Greek.

7
 Figure 7: Ludolph’s rebuilt tombstone in Leiden

This fluent rendition is due to Marinus Taisbak, and the full text is worth reading. It certainly underlines how
influential an algebraist and geometer Viéte was—as an early proponent of methods we now take for granted.
Viéte, who was the first to introduce literals (‘x’ and ‘y’) into algebra, nonetheless rejected the use of negative
numbers.

Leonard Euler (1707–1783). Not surprisingly the great Euler ‘master of us all’ [20] made many contributions
to the literature on π. Equation (3) may be derived from Euler’s product formula for π, given below in
∫ π/2
(4), with x = 1/2, or by repeatedly integrating 0 sin2n (t) dt by parts. One may divine (4) as Euler did
by considering sin(πx) as an ‘infinite’ polynomial and obtaining a product in terms of the roots—which are
0, {1/n2 : n = ±1, ±2, · · · }. It is thus plausible that
∏∞ ( )
sin(π x) x2
(4) = c 1− 2 .
x n=1
n

Euler in 1735, full well knowing that the whole argument was heuristic, argued that, as with a polynomial,
c was the value at zero, 1, and the coefficient of x2 in the Taylor series must be the sum of the roots. Hence, he
was able to pick off coefficients to evaluate the zeta-function at two:
∑ 1 π2
ζ(m) := and marvellously ζ(2) = .
nm 6
n≥1

The explicit formula for ζ(2) solved the so called Basel problem posed in 1644. This method also leads to the
evaluation of ζ(2n) as a rational multiple of π 2n :
π2 π4 π6 π8
ζ(2) = , ζ(4) = , ζ(6) = , ζ(8) = ,...
6 90 945 9450
Indeed, it produces:
2m
(2π)
ζ(2m) = (−1)m−1 B2m ,
2 (2m)!
∑
in terms of the Bernoulli numbers, Bn , where t/(exp(t) − 1) = n≥0 Bn tn /n!, gives a generating function for
the Bn which are perforce rational; see also [28].

8
 Much less is known about odd integer values of ζ, though they are almost certainly not rational multiples of
powers of π. More than two centuries later, in 1976 Roger Apéry, [9, p. 439], [15], showed ζ(3) to be irrational,
and we now also can prove that at least one of ζ(5), ζ(7), ζ(9) or ζ(11) is irrational, but we cannot guarantee
which one. All positive integer values of ζ are strongly believed to be irrational. Though it is not relevant to
our story, Euler’s work on the zeta-function also lead ultimately to the celebrated Riemann hypothesis [14].

5 Pi’s Adult Life with Calculus

In the later 17th century, Newton and Leibniz founded the calculus, and this powerful tool was quickly exploited
to find new formulae for π. One early calculus-based formula comes from the integral:
∫ x ∫ x
dt
tan−1 x = 2
= (1 − t2 + t4 − t6 + · · · ) dt
0 1 + t 0
x3 x5 x7 x9
= x− + − + − ···
3 5 7 9
Substituting x = 1 formally proves the well-known Gregory-Leibniz formula (1671–74)
π 1 1 1 1 1
(5) = 1− + − + − + ···
4 3 5 7 9 11
James Gregory (1638–75) was the greatest of a large Scottish mathematical family. The point, x = 1, however,
is on the boundary of the interval of convergence of the series. Justifying substitution requires a careful error
estimate for the remainder or Lebesgue’s monotone convergence theorem, but most introductory texts ignore
the issue, as they do the issue of Figure 4. The arctan integral and series was known centuries earlier to the
Kerala school: identified with Madhava (c. 1350 – c. 1425) of Sangamagrama near Kerala, India, who may well
have computed 13 digits of π by methods similar to those described in the next section.

A Curious Anomaly in the Gregory Series. In 1988, it was observed that Gregory’s series for π,
∞
∑ ( )
(−1)k+1 1 1 1 1 1
(6) π = 4 = 4 1− + − + − + ···
2k − 1 3 5 7 9 11
k=1

when truncated to 5,000,000 terms, differs strangely from the true value of π:

3.14159245358979323846464338327950278419716939938730582097494182
230781640...
3.14159265358979323846264338327950288419716939937510582097494459
230781640...
2 -2 10 -122 2770

Values differ as expected from truncating an alternating series, in the seventh place—a “4” which should be a
“6.” But the next 13 digits are correct, and after another blip, for 12 digits. Of the first 46 digits, only four differ
from the corresponding digits of π. Further, the “error” digits seemingly occur with a period of 14, as shown
above. Such anomalous behavior begs for explanation. A great place to start is by using Neil Sloane’s Internet-
based integer sequence recognition tool, available at www.research.att.com/~njas/sequences. This tool has
no difficulty
∑∞ recognizing the sequence of errors as twice Euler numbers. Even Euler numbers are generated by
sec x = k=0 (−1)k E2k x2k /(2k)!. The first few are 1, −1, 5, −61, 1385, −50521, 2702765. This discovery led to
the following asymptotic expansion:

∑ (−1)k+1
N/2 ∞
∑
π E2m
(7) −2 ≈ .
2 2k − 1 m=0
N 2m+1
k=1

9
 Name Year Correct Digits
Sharp (and Halley) 1699 71
Machin 1706 100
Strassnitzky and Dase 1844 200
Rutherford 1853 440
Shanks 1874 (707) 527
Ferguson (Calculator) 1947 808
Reitwiesner et al. (ENIAC) 1949 2,037
Genuys 1958 10,000
Shanks and Wrench 1961 100,265
Guilloud and Bouyer 1973 1,001,250

Figure 8: Calculus π Calculations

Now the genesis of the anomaly is clear: by chance the series had been truncated at 5,000,000 terms—exactly
one-half of a fairly large power of ten. Indeed, setting N = 10, 000, 000 in Equation (7) shows that the first
hundred or so digits of the truncated series value are small perturbations of the correct decimal expansion for
π. And the asymptotic expansions show up on the computer screen, as we observed above.

On a hexadecimal computer with N = 167 the corresponding strings and hex-errors are:

3.243F6A8885A308D313198A2E03707344A4093822299F31D0082EFA98EC4E6C8
9452821E...
3.243F6A6885A308D31319AA2E03707344A3693822299F31D7A82EFA98EC4DBF6
9452821E...
2 -2 A -7A 2AD2

with the first being the correct value of π. (In hexadecimal or hex one uses ‘A,B, . . ., F’ to write 10 through 15
as single ‘hex-digits’.) Similar phenomena occur for other constants. (See [9].) Also, knowing the errors means
we can correct them and use (7) to make Gregory’s formula computationally tractable, notwithstanding the
following discussion of complexity!

6 Calculus Era π Calculations

Used naively, the beautiful formula (5) is computationally useless—so slow that hundreds of terms are√ needed
to compute two digits. Sharp, under the direction of Halley4 , (see Figure 8) actually used tan−1 (1/ 3) which
is geometrically convergent. Moreover, Euler’s (1738) trigonometric identity
( ) ( )
−1 −1 1 −1 1
(8) tan (1) = tan + tan
2 3
produces the geometrically convergent rational series
π 1 1 1 1 1 1 1 1
(9) = − + − + ··· + − + − + ···
4 2 3 · 23 5 · 25 7 · 27 3 3 · 33 5 · 35 7 · 37

An even faster formula, found earlier by John Machin, lies similarly in the identity
( ) ( )
π 1 1
(10) = 4 tan−1 − tan−1 .
4 5 239
4 The astronomer and mathematician who largely built the Greenwich Observatory and after whom the comet is named.

10
Figure 9: Newton’s method of computing Pi: “I am ashamed to tell you to how many figures I carried these
computations, having no other business at the time.” Issac Newton, 1666.

This was used in numerous computations of π, given in Figure 8, starting in 1706 and culminating with
Shanks’ famous computation of π to 707 decimal digits accuracy in 1873 (although it was found in 1945 to
be wrong after the 527-th decimal place, by Ferguson, during the last adding machine-assisted pre-computer
computations5 ).

Newton’s arcsin computation. Newton discovered a different more effective—actually a disguised arcsin—
formula. He considering the area A of the left-most region shown in Figure 9. Now, A is the integral
∫ 1/4 √
(11) A = x − x2 dx.
0

√
Also, A is the area of the circular sector, π/24, less the area of the triangle, 3/32. Newton used his newly
developed binomial theorem in (11):
∫ 1 ∫ 1 ( )
4 4 x x2 x3 5x4
A = x 1/2
(1 − x) 1/2
dx = x 1− −
1/2
− − − ··· dx
0 0 2 8 16 128
∫ 1 ( )
4 x3/2 x5/2 x7/2 5x9/2
= x1/2 − − − − · · · dx
0 2 8 16 128
Integrate term-by-term and combining the above produces
√ ( )
3 3 1 1 1 1
π= + 24 − − − ··· .
4 3 · 8 5 · 32 7 · 128 9 · 512

Newton used this formula to compute 15 digits of π. As noted, he later ‘apologized’ for “having no other
business at the time.” (1665-1666 was the year of the great plague which closed Cambridge, and of the great
fire of London of September 1666.). It was also directly after Newton’s Annus mirabilis that led ultimately to
his Principia.) A standard chronology ([26] and [9, p. 294]) says “Newton significantly never gave a value for
π.” Caveat emptor, all users of secondary sources.

The Viennese computer. Until quite recently—around 1950—a computer was a person. As a teenager, this
computer, one Johan Zacharias Dase (1824–1861), would demonstrate his extraordinary computational skill by,
for example, multiplying
79532853 × 93758479 = 7456879327810587
5 This must be some sort a record for the length of time needed to detect a mathematical error.

11
in 54 seconds; two 20-digit numbers in six minutes; two 40-digit numbers in 40 minutes; two 100-digit numbers
in 8 hours and 45 minutes. In 1844, after being shown
( ) ( ) ( )
π −1 1 −1 1 −1 1
= tan + tan + tan
4 2 5 8

he calculated π to 200 places in his head in two months, completing correctly—to my mind—the greatest
mental computation ever. Dase later calculated a seven-digit logarithm table, and extended a table of integer
factorizations to 10,000,000. On Gauss’s recommendation Dase was hired to assist this project, but Dase died
not long afterwards in 1861 by which time Gauss himself already was dead.

An amusing Machin-type identity—meaning an equation that expresses π as a linear combination of

arctan’s—due to the Oxford logician Charles Dodgson is
( ) ( ) ( )
1 1 1
tan−1 = tan−1 + tan−1 ,
p p+q p+r

valid whenever 1 + p2 factors as qr. Dodgson is much better known as Lewis Carroll, the author of Alice in
Wonderland.

7 The Irrationality and Transcendence of π

One motivation for computations of π was very much in the spirit of modern experimental mathematics: to
see if the decimal expansion of π repeats, which would mean that π is the ratio of two integers (i.e., rational),
or to recognize π as algebraic—the root of a polynomial with integer coefficients—and later to look at digit
distribution. The question of the rationality of π was settled in the late 1700s, when Lambert and Legendre
proved (using continued fractions) that the constant is irrational.

The question of whether π was algebraic was settled in 1882, when Lindemann proved that π is transcenden-
tal. Lindemann’s proof also settled, once and for all, the ancient Greek question of whether the circle could be
squared with straight-edge and compass. It cannot be, because numbers that are the lengths of lines that can be
constructed using ruler and compasses (often called constructible numbers) are necessarily algebraic, and squar-
ing the circle is equivalent to constructing the value π. The classical Athenian playwright Aristophanes already
‘knew’ this and perhaps derided ‘circle-squarers’ (τ ετ ραγωσιειν) in his play The Birds of 414 BCE. Likewise,
the French Academy had stopped accepting proofs of the three great constructions of antiquity—squaring the
circle, doubling the cube and trisecting the angle—centuries before it was proven impossible.

We next give, in extenso, Ivan Niven’s 1947 short proof of the irrationality of π. It well illustrates the
ingredients of more difficult later proofs of irrationality of other constants, and indeed of Lindemann’s proof of
the transcendence of π building on Hermite’s 1873 proof of the transcendence of e.

8 A Proof that π is Irrational

Proof. Let π = a/b, the quotient of positive integers. We define the polynomials

xn (a − bx)n
f (x) =
n!

F (x) = f (x) − f (2) (x) + f (4) (x) − · · · + (−1)n f (2n) (x)

the positive integer being specified later. Since n!f (x) has integral coefficients and terms in x of degree not
less than n, f (x) and its derivatives f (j) (x) have integral values for x = 0; also for x = π = a/b, since

12
f (x) = f (a/b − x). By elementary calculus we have

d
{F ′ (x) sin x − F (x) cos x} = F ′′ (x) sin x + F (x) sin x = f (x) sin x
dx
and
∫ π
f (x) sin x dx = [F ′ (x) sin x − F (x) cos x]π0
0

(12) = F (π) + F (0).

Now F (π) + F (0) is an integer, since f (j) (0) and f (j) (π) are integers. But for 0 < x < π,

π n an
0 < f (x) sin x < ,
n!
so that the integral in (12) is positive but arbitrarily small for n sufficiently large. Thus (12) is false, and so is
our assumption that π is rational. QED

Irrationality measures. We end this section by touching on the matter of measures of irrationality. The
infimum µ(α) of those µ > 0 for which

p 1
α− ≥ µ
q q

for all integers p, q with sufficiently large q, is called the Liouville-Roth constant for α and we say that we have
an irrationality measure for α if µ(α) < ∞.

Irrationality measures are difficult. Roth’s theorem, [15], implies that µ(α) = 2 for all algebraic irrationals,
as is the case for almost all∑reals. Clearly, µ(α) = 1 for rational α and µ(α) = ∞ if and only if α is a so-called
Liouville number such as 1/10n! . It is known that µ(e) = 2 while in 1993 Hata showed that µ(π) ≤ 8.02.
Similarly, it is known that µ(ζ(2)) ≤ 5.45, µ(ζ(3)) ≤ 4.8 and µ(log 2) ≤ 3.9.

A consequence of the existence of an irrationality measure µ for π, is the ability to estimate quantities such
as lim sup | sin(n)|1/n = 1 for integer n, since for large integer m and n with m/n → π, we have eventually
1 1
| sin(n)| = | sin(mπ) − sin(n)| ≥ |mπ − n| ≥ .
2 2 mµ−1

Related matters are discussed at more length in [2].

9 π in the Digital Age

With the substantial development of computer technology in the 1950s, π was computed to thousands and
then millions of digits. These computations were greatly facilitated by the discovery soon after of advanced
algorithms for the underlying high-precision arithmetic operations. For example, in 1965 it was found that the
newly-discovered fast Fourier transform (FFT) [15, 14] could be used to perform
√ high-precision multiplications
much more rapidly than conventional schemes. Such methods (e.g., for ÷, x see [15, 16, 14]) dramatically
lowered the time required for computing π and other constants to high precision. We are now able to compute
algebraic values of algebraic functions essentially as fast as we can multiply, OB (M (N )), where M (N ) is the
cost of multiplication and OB counts ‘bits’ or ‘flops’. To convert this into practice: a state-of-the-art processor
in 2010, such as the latest AMD Opteron, which runs at 2.4 GHz and has four floating-point cores, each of

13
which can do two 64-bit floating-point operations per second, can produce a total of 9.6 billion floating-point
operations per second.

In spite of these advances, into the 1970s all computer evaluations of π still employed classical formulae,
usually of Machin-type, see Figure 8. We will see below methods that can compute N digits of π with time
complexity OB (M (N )) log OB (M (N )). Proving that the log term is unavoidable, as seems likely, would yield
an algorithmic proof—quite different from current proofs—that π is not algebraic. Such a proof would be a
significant contribution to number theory and complexity theory.

Electronic Numerical Integrator and Calculator. The first computer calculation of π was performed
on ENIAC—a behemoth with a tiny brain from today’s vantage point. The ENIAC was built in Aberdeen,
Maryland by the US Army:

Size/weight. ENIAC had 18,000 vacuum tubes, 6,000 switches, 10,000 capacitors, 70,000 resistors,
1,500 relays, was 10 feet tall, occupied 1,800 square feet and weighed 30 tons.
Speed/memory. A, now aged, 1.5GHz Pentium does 3 million adds/sec. ENIAC did 5,000, three
orders faster than earlier machines. The first stored-memory computer, ENIAC stored 200 digits.
Input/output. Data flowed from one accumulator to the next, and after each accumulator finished
a calculation, it communicated its results to the next in line. The accumulators were connected to
each other manually. The 1949 computation of π to 2,037 places on ENIAC took 70 hours in which
output had to be constantly reintroduced as input.

A fascinating description of the ENIAC’s technological and commercial travails is to be found in [25]. Note that
ENIAC as a child of the 1940’s was called a ‘calculator’ not a computer.

Ballantine’s (1939) Series for π. Another formula of Euler for arccot is

∞
∑ 2 ( )
(n!) 4n 1
x n+1 = arctan .
2
n=0 (2 n + 1)! (x + 1)
x
This, intriguingly and usefully, allowed Guilloud and Boyer to reexpress the formula, used by them in 1973 to
compute a million digits of π, viz, π/4 = 12 arctan (1/18) + 8 arctan (1/57) − 5 arctan (1/239) in the efficient
form
∑∞ 2 ∑∞ 2 ( )
(n!) 4n (n!) 4n 1
π = 864 + 1824 − 20 arctan ,
n=0
(2 n + 1)! 325n+1 n=0
(2 n + 1)! 3250n+1 239
where the terms of the second series are now just decimal shifts of the first.

Ramanujan-type elliptic series. Truly new types of infinite series formulae, based on elliptic integral
approximations, were discovered around 1910 by Srinivasa Ramanujan (1887–1920) (shown in Figure 10) but
were not well known (nor fully proven) until quite recently when his writings were fully decoded and widely
published. They are based on elliptic functions and are described at length in [9, 15, 14].

G.N. Watson (see [14]) elegantly describes his feelings on viewing formulae of Ramanujan, such as (13) below:

“... a thrill which is indistinguishable from the thrill which I feel when I enter the Sagrestia
Nuova of the Cappella Medici and see before me the austere beauty of the four statues
representing ‘Day’, ‘Night’, ‘Evening’, and ‘Dawn’ which Michelangelo has set over the
tomb of Giuliano de’Medici and Lorenzo de’Medici”

One of these series is the remarkable:

√ ∞
1 2 2 ∑ (4k)! (1103 + 26390k)
(13) = .
π 9801 (k!)4 3964k
k=0

14
 Figure 10: Ramanujan’s seventy-fifth birthday stamp (courtesy Naomi Borwein-Yao

Each term of this series produces an additional eight correct digits in the result. When Gosper used this formula
to compute 17 million digits of π in 1985, and it agreed to many millions of places with the prior estimates,
this concluded the first proof of (13). As described in [17], this computation can be shown to be exact enough
to constitute a bona fide proof ! Actually, Gosper first computed the simple continued fraction for π, hoping
to discover some new things in its expansion, but found none. At the time of writing 500 million terms of the
continued fraction for π have been computed by Neil Bickford (a teenager) without shedding light on whether
the sequence is unbounded (see [1]).

At about the same time, David and Gregory Chudnovsky found

√ the following rational variation of Ramanu-
jan’s formula. Amazingly one can show that it exists because −163 corresponds to an imaginary quadratic
field with class number one:
∞
∑
1 (−1)k (6k)! (13591409 + 545140134k)
(14) = 12
π
k=0
(3k)! (k!)3 6403203k+3/2

Each term of this series produces an additional 14 correct digits. The Chudnovskys, shown in Figure 14,
implemented this formula using a clever scheme that enabled them to use the results of an initial level of
precision to extend the calculation to even higher precision. They used this in several large calculations of π,
culminating with a then record computation of over four billion decimal digits in 1994. Their remarkable
story was told compellingly by Richard Preston in a prizewinning New Yorker article “The Mountains of Pi”
(March 2, 1992).

While the Ramanujan and Chudnovsky series are in practice considerably more efficient than classical for-
mulae, they share the property that the number of terms needed increases linearly with the number of digits
desired: if you want to compute twice as many digits of π, you must evaluate twice as many terms of the series.

Relatedly, the Ramanujan-type series

∞
( ( ) )3
1 ∑ 2n
42 n + 5
n
(15) =
π n=0
16n 16

allows one to compute the billionth binary digit of 1/π, or the like, without computing the first half of the series,
and is a foretaste of our later discussion of Borwein-Bailey-Plouffe (or BBP) formulae.

15
 Name Year Correct Digits
Miyoshi and Kanada 1981 2,000,036
Kanada-Yoshino-Tamura 1982 16,777,206
Gosper 1985 17,526,200
Bailey Jan. 1986 29,360,111
Kanada and Tamura Sep. 1986 33,554,414
Kanada and Tamura Oct. 1986 67,108,839
Kanada et. al Jan. 1987 134,217,700
Kanada and Tamura Jan. 1988 201,326,551
Chudnovskys May 1989 480,000,000
Kanada and Tamura Jul. 1989 536,870,898
Kanada and Tamura Nov. 1989 1,073,741,799
Chudnovskys Aug. 1991 2,260,000,000
Chudnovskys May 1994 4,044,000,000
Kanada and Takahashi Oct. 1995 6,442,450,938
Kanada and Takahashi Jul. 1997 51,539,600,000
Kanada and Takahashi Sep. 1999 206,158,430,000
Kanada-Ushiro-Kuroda Dec. 2002 1,241,100,000,000
Takahashi Jan. 2009 1,649,000,000,000
Takahashi April. 2009 2,576,980,377,524
Bellard Dec. 2009 2,699,999,990,000

Figure 11: Post-calculus π Calculations

10 Reduced Operational Complexity Algorithms

In 1976, Eugene Salamin and Richard Brent independently discovered a reduced complexity algorithm for π. It is
based on the arithmetic-geometric mean iteration (AGM) and some other ideas due to Gauss and Legendre
around 1800, although neither Gauss, nor many after him, ever directly saw the connection to effectively
computing π.

√
Quadratic Algorithm (Salamin-Brent). Set a0 = 1, b0 = 1/ 2 and s0 = 1/2. Calculate

ak−1 + bk−1 √
(16) ak = (Arithmetic) bk = ak−1 bk−1 (Geometric)
2
2a2k
(17) ck = a2k − b2k , sk = sk−1 − 2k ck and compute pk = .
sk

Then pk converges quadratically to π. Note the similarity between the arithmetic-geometric mean iteration (16),
(which for general initial values converges quickly to a non-elementary limit) and the out-of-kilter harmonic-
geometric mean iteration (2) (which in general converges slowly to an elementary limit), and which is an
arithmetic-geometric iteration in the reciprocals (see [15]).

Each iteration of the algorithm doubles the correct digits. Successive iterations produce 1, 4, 9, 20, 42, 85, 173, 347
and 697 good decimal digits of π, and takes log N operations for N digits. Twenty-five iterations computes π
to over 45 million decimal digit accuracy. A disadvantage is that each of these iterations must be performed to
the precision of the final result. In 1985, my brother Peter and I discovered families of algorithms of this type.
For example, here is a genuinely third-order iteration6 :
6A fourth-order iteration might be a compound of two second-order ones; this third order one can not be so decomposed.

16
 √
Cubic Algorithm. Set a0 = 1/3 and s0 = ( 3 − 1)/2. Iterate
3 rk+1 − 1
rk+1 = , sk+1 = and 2
ak+1 = rk+1 ak − 3k (rk+1
2
− 1).
1 + 2(1 − s3k )1/3 2

Then 1/ak converges cubically to π. Each iteration triples the number of correct digits.

√ √
Quartic Algorithm. Set a0 = 6 − 4 2 and y0 = 2 − 1. Iterate

1 − (1 − yk4 )1/4
yk+1 = and ak+1 = ak (1 + yk+1 )4 − 22k+3 yk+1 (1 + yk+1 + yk+1
2
).
1 + (1 − yk4 )1/4

Then 1/ak converges quartically to π. Note that only the power of 2 or 3 used in ak depends on k.

Let us take an interlude and discuss:

‘Piems’ or pi-mnemonics. Piems are mnemonics in which the length of each word is the corresponding digit
of π. Punctuation is ignored. A better piem is both longer and better poetry.

Mnemonics for π

“Now I, even I, would celebrate

In rhyme inapt, the great
Immortal Syracusan, rivaled nevermore,
Who in his wondrous lore,
Passed on before
Left men for guidance
How to circles mensurate.” (30)

“How I want a drink, alcoholic of course, after the heavy lectures involving
quantum mechanics.” (15)

“See I have a rhyme assisting my feeble brain its tasks ofttimes resisting.” (13)

There are many more and much longer mnemonics than the famous examples given in the inset box—see
[9, p. 405, p.560, p. 659] for a fine selection. Indeed, there is whole cottage industry around the matter,
http://www.huffingtonpost.com/jonathan-m-borwein/pi-day_b_1341569.html?ref=science.

Philosophy of mathematics. In 1997 the first occurrence of the sequence 0123456789 was found (later than
expected heuristically) in the decimal expansion of π starting at the 17, 387, 594, 880-th digit after the decimal
point. In consequence the status of several famous intuitionistic examples due to Brouwer and Heyting has
changed. These challenge the principle of the excluded middle—either a predicate holds or it does not— and
involve classically well-defined objects that for an intuitionist are ill-founded until one can determine when or
if the sequence occurred (see [12]).

For example, consider the sequence which is ‘0’ except for a ‘1’ in the first place where 0123456789 first
begins to appear in order if it ever occurs. Did it converge when first used by Brouwer as an example? Does
it now? Was it then and is it now well defined? Classically it always was and converged to ‘0’. But, until the
computation was done any argument about its convergence relied on the principle of the excluded middle which
intuitionists reject. Intuitionistically, it converges now. Of course, if we redefine the sequence to have its ‘1’ in
the first place that 0123456789101112 first begins, then we are still completely in the dark. But is a sign of how
far computation of π has come recently that this was never thought necessary.

17
 Figure 12: Yasumasa Kanada in his Tokyo office

11 Back to the Future

In December 2002, Kanada computed π to over 1.24 trillion decimal digits. His team first computed π in
hexadecimal (base 16) to 1,030,700,000,000 places, using the following two arctangent relations:
1 1 1 1
π = 48 tan−1 + 128 tan−1 − 20 tan−1 + 48 tan−1
49 57 239 110443
1 1 1 1
π = 176 tan−1 + 28 tan−1 − 48 tan−1 + 96 tan−1 .
57 239 682 12943
The first formula was found in 1982 by K. Takano, a high school teacher and song writer. The second formula
was found by F. C. W. Störmer in 1896. Kanada verified the results of these two computations agreed, and
then converted the hex digit sequence to decimal. The resulting decimal expansion was checked by converting
it back to hex. These conversions are themselves non-trivial, requiring massive computation.

This process is quite different from those of the previous quarter century. One reason is that reduced
operational complexity algorithms require full-scale multiply, divide and square root operations. These in turn
require large-scale fast Fourier transform (FFT) operations, which demand huge amounts of memory, and
massive all-to-all communication between nodes of a large parallel system. For this latest computation, even
the very large system available in Tokyo did not have sufficient memory and network bandwidth to perform
these operations at reasonable efficiency levels—at least not for trillion-digit computations. Utilizing arctans
again meant using many more arithmetic operations, but no system-scale FFTs, and so the method can be
implemented using ×, ÷ by smallish integer values—additionally, hex is somewhat more efficient!

Kanada and his team evaluated these two formulae using a scheme analogous to that employed by Gosper
and by the Chudnovskys in their series computations, in that they were able to avoid explicitly storing the
multiprecision numbers involved. This resulted in a scheme that is roughly competitive in numerical efficiency
with the Salamin-Brent and Borwein quartic algorithms they had previously used, but with a significantly lower
total memory requirement. Kanada used a 1 Tbyte main memory system, as with the previous computation,
yet got six times as many digits. Hex and decimal evaluations included, it ran 600 hours on a 64-node Hitachi,
with the main segment of the program running at a sustained rate of nearly 1 Tflop/sec.

18
 Hex Digit Occurrences

0 62499881108
Decimal Digit Occurrences 1 62500212206
2 62499924780
0 99999485134 3 62500188844
1 99999945664 4 62499807368
2 100000480057 5 62500007205
3 99999787805 6 62499925426
4 100000357857 7 62499878794
5 99999671008 8 62500216752
6 99999807503 9 62500120671
7 99999818723 A 62500266095
8 100000791469 B 62499955595
9 99999854780 C 62500188610
D 62499613666
Total 1000000000000 E 62499875079
F 62499937801

Total 1000000000000

Figure 13: Seemingly random behaviour of single digits of π in base 10 and 16

12 Why π?

What possible motivation lies behind modern computations of π, given that questions such as the irrationality
and transcendence of π were settled more than 100 years ago? One motivation is the raw challenge of harnessing
the stupendous power of modern computer systems. Programming such calculations are definitely not trivial,
especially on large, distributed memory computer systems.

There have been substantial practical spin-offs. For example, some new techniques for performing the fast
Fourier transform (FFT), heavily used in modern science and engineering computing, had their roots in attempts
to accelerate computations of π. And always the computations help in road-testing computers—often uncovering
subtle hardware and software errors.

Beyond practical considerations lies the abiding interest in the fundamental question of the normality (digit
randomness) of π. Kanada, for example, has performed detailed statistical analysis of his results to see if there
are any statistical abnormalities that suggest π is not normal, so far the answer is “no” (see Figures 1 and
13). (Kanada reports that the 10 decimal digits ending in position one trillion are 6680122702, while the 10
hexadecimal digits ending in position one trillion are 3F89341CD5.) Indeed, the first computer computation of π
and e on ENIAC, discussed above, was so motivated by John von Neumann. The digits of π have been studied
more than any other single constant, in part because of the widespread fascination with and recognition of π.
Very recent work, suggesting π may well be normal, can be found in and traced from [1].

Changing world views. In retrospect, we may wonder why in antiquity π was not measured to an accuracy
in excess of 22/7? Perhaps it reflects not an inability to do so but a very different mindset to a modern more
experimental one. One can certainly find Roman ampitheatres where more accurate measurement than 22/7
would have been helpful. Bailey and I discuss this issue in more detail in [6].

In the same vein, one reason that Gauss and Ramanujan did not further develop the ideas in their identities
for π is that an iterative algorithm, as opposed to explicit results, was not as satisfactory for them (especially

19
 Figure 14: The remarkable Chudnovsky brothers π (courtesy D. and G. Chudnovsky)

Ramanujan). Ramanujan much preferred formulae like

3 3
π ≈ √ log (5280) , √ log (640320) ≈ π
67 163
correct to 9 and 15 decimal places; both of which rely on deep number theory. Contrastingly, Ramanujan in his
famous 1914 paper Modular Equations and Approximations to Pi [9, p.253] found
( )1/4
192
92 + = 3.14159265258 · · ·
22

“empirically, and it has no connection with the preceding theory.” Only the marked digit is wrong.

Discovering the π Iterations. The genesis of the π algorithms and related material is an illustrative example
of experimental mathematics. My brother and I in the early 1980’s had a family of quadratic algorithms for
π, [15], call them PN , of the kind we saw above. For N = 1, 2, 3, 4 we could prove they were correct but
were only conjectured for N = 5, 7. In each case the algorithm appeared to converge quadratically to π. On
closer inspection while the provable cases were correct to 5, 000 digits, the empirical versions agreed with π to
roughly 100 places only. Now in many ways to have discovered a “natural” number that agreed with π to that
level—and no more—would have been more interesting than the alternative. That seemed unlikely but recoding
and rerunning the iterations kept producing identical results.

Two decades ago even moderately high precision calculation was less accessible, and the code was being run
remotely over a phone-line in a Berkeley Unix integer package. After about six weeks, it transpired that the
package’s square root algorithm was badly flawed, but only if run with an odd precision of more than sixty digits!
And for idiosyncratic reasons that had only been the case in the two unproven cases. Needless to say, tracing
the bug was a salutary and somewhat chastening experience. And it highlights why one checks computations
using different sub-routines and methods.

13 How to Compute the N -th Digits of π

One might be forgiven for thinking that essentially everything of interest with regards to π has been dealt with.
This is suggested in the closing chapters of Beckmann’s 1971 book A History of π. Ironically, the Salamin–
Brent quadratically convergent iteration was discovered only five years later, and the higher-order convergent

20
algorithms followed in the 1980s. Then in 1990, Rabinowitz and Wagon discovered a “spigot” algorithm for
π—the digits ‘drip out’ one by one. This permits successive digits of π (in any desired base) to be computed
by a relatively simple recursive algorithm based on all previously generated digits.

Even insiders are sometimes surprised by a new discovery. Prior to 1996, most researchers thought if you
want to determine the d-th digit of π, you had to generate the (order of) the entire first d digits. This is not
true, at least for hex (base 16) or binary (base 2) digits of π. In 1996, Peter Borwein, Plouffe, and Bailey found
an algorithm for computing individual hex digits of π. It (1) yields a modest-length hex or binary digit string
for π, from an arbitrary position, using no prior bits; (2) is implementable on any modern computer; (3) requires
no multiple precision software; (4) requires very little memory; and (5) has a computational cost growing only
slightly faster than the digit position. For example, the millionth hexadecimal digit (four millionth binary digit)
of π could be found in four seconds on a 2005 Apple computer.

This new algorithm is not fundamentally faster than the best known schemes if used for computing all digits
of π up to some position, but its storage requirements, elegance and simplicity are of considerable interest, and
it is easy to parallelize. It is based on the following at-the-time new formula for π:
∑∞ ( )
1 4 2 1 1
(18) π= − − −
i=0
16i 8i + 1 8i + 4 8i + 5 8i + 6

which was discovered using integer relation methods (see [14]), with a computer search that ran over several
months and ultimately produced the (equivalent) relation:
( ) ( )
1, 41 1 1
(19) π = 4 · 2 F1 5 −4 + 2 arctan − log 5,
4
2

where the first term is a generalized Gaussian hypergeometric function evaluation. Maple and Mathematica can
both now prove (18). They could not at the time of its discovery. A human proof may be found in [14].

The algorithm in action. In 1997, Fabrice Bellard at INRIA—whom we shall meet again in Section 15—
computed 152 binary digits of π starting at the trillionth position. The computation took 12 days on 20
workstations working in parallel over the Internet. Bellard’s scheme is based on the following variant of (18):
∞
∑ ∞ ( )
(−1)k 1 ∑ (−1)k 32 8 1
π = 4 k
− + + ,
4 (2k + 1) 64 1024k 4k + 1 4k + 2 4k + 3
k=0 k=0

which permits hex or binary digits of π to be calculated somewhat faster than (18) depending on the imple-
mentation. (Most claims of improved speed of algorithms are subject to many caveats.)

In 1998 Colin Percival, then a 17-year-old student at Simon Fraser University (see Figure 15), accessed 25
machines to calculate first the five trillionth hexadecimal digit of π, and then the ten trillionth hex digit. In
September 2000, he found the quadrillionth binary digit is 0, a computation that required 250 CPU-years, using
1734 machines in 56 countries. We record some of Percival’s computational results in Figure 15.

Nor have matters stopped there. As described in [5, 8] in the most recent computation of π using the
BBP formula, Tse-Wo Zse of Yahoo! Cloud Computing calculated 256 binary digits of π starting at the two
quadrillionth bit. He then checked his result using Bellard’s variant. In this case, both computations verified
that the 24 hex digits beginning immediately after the 500 trillionth hex digit (i.e., after the two quadrillionth
binary bit) are: E6C1294A ED40403F 56D2D764.

Kanada was able to confirm his 2002 computation in only 21 hours by computing a 20 hex digit string
starting at the trillionth digit, and comparing this string to the hex string he had initially obtained in over 600
hours. Their agreement provided enormously strong confirmation. We shall see this use of BBP for verification
again when we discuss the most recent record computations of π.

21
 Colin Percival (1998)

Hex strings starting

Position at this Position

106 26C65E52CB4593
107 17AF5863EFED8D
108 ECB840E21926EC
109 85895585A0428B
1010 921C73C6838FB2
1011 9C381872D27596
1.25 × 1012 07E45733CC790B
2.5 × 1014 E6216B069CB6C1

Figure 15: Percival’s hexadecimal findings

14 Further BBP Digit Formulae

Motivated as above, constants α of the form

∞
∑ p(k)
(20) α = ,
q(k)2k
k=0

where p(k) and q(k) are integer polynomials, are said to be in the class of binary (Borwein-Bailey-Plouffe) BBP
numbers. I illustrate for log 2 why this permits one to calculate isolated digits in the binary expansion:
∞
∑ 1
(21) log 2 = .
k2k
k=0

We wish to compute a few binary digits beginning at position d + 1. This is equivalent to calculating
{2d log 2}, where {·} denotes fractional part. We can write
{{ d } { ∞ }}
∑ 2d−k ∑ 2d−k
(22) {2 log 2} =
d
+
k k
k=0 k=d+1
{{ d } { ∞ }}
∑ 2d−k mod k ∑ 2d−k
(23) = + .
k k
k=0 k=d+1

The key observation is that the numerator of the first sum in (23), 2d−k mod k, can be calculated rapidly by
binary exponentiation, performed modulo k. That is, it is economically performed by a factorization based on
the binary expansion of the exponent. For example,

317 = ((((32 )2 )2 )2 ) · 3

uses only five multiplications, not the usual 16. It is important to reduce each product modulo k. Thus, 317
mod 10 is done as
32 = 9; 92 = 1; 12 = 1; 12 = 1; 1 × 3 = 3.

A natural question in light of (18) is whether there is a formula of this type and an associated computational
strategy to compute individual decimal digits of π. Searches conducted by numerous researchers have been
unfruitful and recently D. Borwein (my father), Gallway and I have shown that there are no BBP formulae of
the Machin-type (as defined in [14]) of (18) for π unless the base is a power of two [14].

22
Figure 16: Ferguson’s “Eight-Fold Way” and his BBP acrylic circles. These three ‘subtractive’ acrylic circles
(white) and the black circle represent the weights [4, −2, −2, −1] in Equation (18)

Ternary BBP formulae. Yet, BBP formulae exist in other bases for some constants. For example, for π 2
we have both binary and ternary formulae (discovered by Broadhurst):
( )
9∑ 1
∞
16 24 8 6 1
π2 = − − − + .
8 64k (6k + 1)2 (6k + 2)2 (6k + 3)2 (6k + 4)2 (6k + 5)2
k=0
(24)
(
2 ∑ 1
∞
243 405 81 27
π2 = − − −
27 729k (12k + 1)2 (12k + 2)2 (12k + 4)2 (12k + 5)2
k=0
)
72 9 9 5 1
(25) − − − − + .
(12k + 6)2 (12k + 7)2 (12k + 8)2 (12k + 10)2 (12k + 11)2

These two formulae have recently been used for record digit computations performed on an IBM Blue Gene
system in conjunction with IBM Australia [8].

Remarkably the volume V8 in hyperbolic space of the figure-eight knot complement is well known to be
∞
√ ∑ 1 ∑ 1
2n−1
V8 = 2 3 (2n) = 2.029883212819307250042405108549 . . .
n=1
n n k=n k

Surprisingly, it is also expressible as

√ ∞ { }
3 ∑ (−1)n 18 18 24 6 2
V8 = − − − + ,
9 n=0 27n (6n + 1)2 (6n + 2)2 (6n + 3)2 (6n + 4)2 (6n + 5)2

again discovered numerically by Broadhurst, and proved in [14]. A beautiful representation by Helaman Ferguson
the mathematical sculptor is given in Figure 16. Ferguson produces art inspired by deep mathematics, but not
by a formulaic approach. For instance, his knowledge of hyperbolic geometry allows him to exploit surfaces of
negative curvature as shown in his“Eight-Fold Way”.

Normality and dynamics. Finally, Bailey and Crandall in 2001 made exciting connections between the
existence of a b-ary BBP formula for α and its normality base b (uniform distribution of base-b digits)7 . They
7 See www.sciencenews.org/20010901/bob9.asp .

23
make a reasonable, hence very hard, conjecture about the uniform distribution of a related chaotic dynamical
system. This conjecture implies: Existence of a ‘BBP’ formula base b for α ensures the normality base b of α.
Illustratively, or log 28 , the dynamical system, base 2, is to set x0 = 0 and compute
( )
1
xn+1 ←- 2 xn + mod 1.
n

15 Pi in the Third Millennium

15.1 Reciprocal series

A few years ago Jesús Guillera found various Ramanujan-like identities for π, using integer relation methods.
The three most basic are:
∑∞ ( )2n+1
4 n 5 2 1
(26) = (−1) r(n) (13 + 180n + 820n )
π2 n=0
32
∑∞ ( )2n+1
2 n 5 2 1
(27) 2
= (−1) r(n) (1 + 8n + 20n )
π n=0
2
∑∞ ( )2n+1
4 ? 7 2 3 1
(28) 3
= r(n) (1 + 14n + 76n + 168n ) ,
π n=0
8

where r(n) := (1/2 · 3/2 · · · · · (2n − 1)/(2n))/n!.

Guillera proved (26) and (27) in tandem, using most ingeniously the Wilf–Zeilberger algorithm for formally
proving hypergeometric-like identities [14, 7, 29]. No other proof is known and there seem to be no like formulae
for 1/π d with d ≥ 4. The third (28) is certainly true,9 but has no proof; nor does anyone have an inkling of
how to prove it; especially as experiment suggests that it has no ‘mate’ unlike (26) and (27) [7]. My intuition is
that if a proof exists it is more a verification than an explication and so I stopped looking. I am happy just to
know the beautiful identity is true. A very nice account of the current state of knowledge for Ramanujan-type
series for 1/π is to be found in [10].

In 2008 Guillera [22] produced another lovely pair of third millennium identities—discovered with integer
relation methods and proved with creative telescoping—this time for π 2 rather than its reciprocal. They are
∞ ( ) ∞ ( 1 )2
∑ 1 x+ 2 n
1 3 ∑
2 n
(29) 2n (x + 1)3
(6(n + x) + 1) = 8x ,
n=0
2 n n=0
(x + 1)2n

and
∞
( ) ∞
( )2
∑ 1 x+ 2 n
1 3 ∑ x + 12 n
(30) (42(n + x) + 5) = 32x .
n=0
26n (x + 1)3n n=0
(2x + 1)2n

Here (a)n = a(a + 1) · ·(a + n − 1) is the rising factorial. Substituting x = 1/2 in (29) and (30), he obtained
respectively the formulae
∑∞ ∞
∑
1 (1)3n π2 1 (1)3n π2
2n ( 3 )3 (3n + 2) = and ( 3 )3 (21n + 13) = .
n=0
2 4 n=0
26n−2 3
2 n 2 n
8 Inthis case it is easy to use Weyl’s criterion for equidistribution to establish this equivalence without mention of BBP numbers.
9 Guillera ascribes (28) to Gourevich, who used integer relation methods. I ‘rediscovered’ (28) using integer relation methods with
30 digits. I then checked it to 500 places in 10 seconds, 1200 in 6.25 minutes, and 1500 in 25 minutes: with a naive command-line
instruction in Maple on a light laptop.

24
15.2 Computational records

The last decade has seen the record for computation of π broken in some very interesting ways. We have already
described Kanada’s 2002 computation in Section 11 and noted that he also took advantage of the BBP formula
of Section 13. This stood as a record until 2009 when it was broken three times—twice spectacularly.

Daisuke Takahashi. The record for computation of π of under 29.37 million decimal digits, by Bailey in
1986 had increased to over 2.649 trillion places by Takahashi in January 2009. Since the same algorithms were
used for each computation, it is interesting to review the performance in each case:

In 1986 it took 28 hours to compute 29.36 million digits on 1 cpu of the then new CRAY-2 at NASA Ames
using (18). Confirmation using the quadratic algorithm (16) took 40 hours. (The computation uncovered
hardware and software errors on the CRAY. Success required developing a speedup of the underlying FFT [14].)
In comparison, on 1024 cores of a 2592 core Appro Xtreme-X3 system 2.649 trillion digits via (16) took 64
hours 14 minutes with 6732 GB of main memory, and (18) took 73 hours 28 minutes with 6348 GB of main
memory. (The two computations differed only in the last 139 places.) In April Takahashi upped his record to
2,576,980,377,524 places.

Fabrice Bellard. Near the end of 2009, Bellard computed nearly 2.7 trillion decimal digits of π (first in
binary) using the Chudnovsky series (14). This took 131 days but he only used a single 4-core workstation with
a lot of storage and even more human intelligence! For full details of this feat and of Takahashi’s most recent
computation one can look at
http://en.wikipedia.org/wiki/Chronology_of_computation_of_pi .

Nor is that the current end of the matter:

Alexander Yee and Shigeru Kondo. In August 2010, they announced that they had used the Chudnovsky
formula to compute 5 trillion digits of π over a 90-day period, mostly on a two-core Intel Xeon system with
96 Gbyte of memory. They confirmed the result in two ways, using the BBP formula (as discussed above),
which required 66 hours, and a variant of the BBP formula due to Bellard, which required 64 hours. Changing
from binary to decimal required 8 days. This was upped to 10 trillion digits in October 2011. Full details are
available at http://www.numberworld.org/misc_runs/pi-5t/details.html.

16 . . . Life of π

Paul Churchland, writing about the sorry creationist battles of the Kansas school board, [19, Kindle ed, loc
1589] observes that:

“Even mathematics would not be entirely safe. (Apparently, in the early 1900’s, one legislator in a
southern state proposed a bill to redefine the value of pi as 3.3 exactly, just to tidy things up.)”

As we have seen, the life of π captures a great deal of mathematics—algebraic, geometric and analytic, both
pure and applied—along with some history and philosophy. It engages many of the greatest mathematicians
and some quite interesting characters along the way. Among the saddest and least-well understood episodes
was an abortive 1896 attempt in Indiana to legislate the value(s) of π. The bill, reproduced in [9, p. 231-235],
is is accurately described by David Singmaster, [27] and [9, p. 236-239].

At the end of the novel, Piscine (Pi) Molitor writes

25
 “I am a person who believes in form, in harmony of order. Where we can, we must give things a
meaningful shape. For example—I wonder—could you tell my jumbled story in exactly one hundred
chapters, not one more, not one less? I’ll tell you, that’s one thing I hate about my nickname, the
way that number runs on forever. It’s important in life to conclude things properly. Only then can
you let go.”

We may well not share the sentiment, but we should celebrate that Pi knows π to be irrational.

17 End Notes

1. Why π is not 22/7. Today, even the computer algebra systems Maple or Mathematica ‘know’ this since
∫ 1
(1 − x)4 x4 22
(31) 0 < 2
dx = − π,
0 1+x 7
though it would be prudent to ask ‘why’ each can perform the integral and ‘whether’ to trust it. Assuming we
do trust it, then the integrand is strictly positive on (0, 1), and the answer in (31) is an area and so strictly
positive, despite claims that π is 22/7 ranging over millennia.10 In this case, requesting the indefinite integral
provides immediate reassurance. We obtain
∫ 4
t
x4 (1 − x) 1 7 2 6 4
dx = t − t + t5 − t3 + 4 t − 4 arctan (t) ,
0 1 + x2 7 3 3

as differentiation easily confirms, and so the Newtonian fundamental theorem of calculus proves (31).

One can take the idea in (31) a bit further, as in [14]. Note that
∫ 1
4 1
(32) x4 (1 − x) dx = ,
0 630
and we observe that
∫ ∫ ∫
1 1
4
1
(1 − x)4 x4 1
4
(33) x4 (1 − x) dx < dx < x4 (1 − x) dx.
2 0 0 1 + x2 0

Combine this with (31) and (32) to derive: 223/71 < 22/7 − 1/630 < π < 22/7 − 1/1260 < 22/7, and so
re-obtain Archimedes’ famous computation
10 10
(34) 3 <π<3 .
71 70
The derivation above was first popularized in Eureka, a Cambridge student journal in 1971.11 A recent study
of related approximations is [24]. (See also [14].)

2. More about Gamma. One may define

∫ ∞
Γ(x) = tx−1 e−t dt
0
10 One may still find adverts in newspapers offering such proofs for sale. A recent and otherwise very nice children’s book “Sir
Cumference and the the Dragon of Pi (A Math Adventure)” published in (1999) repeats the error, and email often arrives in my
in-box offering to show why this and things like this are true.
11 (31) was on a Sydney University examination paper in the early sixties and the earliest source I know of dates from the 1940’s

[14].

26
for Re x > 0. The starting point is that

(35) x Γ(x) = Γ(x + 1), Γ(1) = 1.

In particular, for integer n, Γ(n + 1) = n!. Also for 0 < x < 1

π
Γ(x) Γ(1 − x) = ,
sin(πx)
since for x > 0 we have
n! nx
Γ(x) = lim ∏n .
k=0 (x + k)
n→∞

This is a nice consequence of the Bohr-Mollerup theorem [15, 14]√which shows that Γ is the unique log-convex
function on the positive half line satisfying (35). Hence, Γ(1/2) = π and equivalently we evaluate the Gaussian
integral ∫ ∞
√
e−x dx = π,
2

−∞
so central to probability theory. In the same vein, the improper sinc function integral evaluates as
∫ ∞
sin(x)
dx = π.
−∞ x

Considerable information about the relationship between Γ and π is to be found in [14, 21].

The Gamma function is as ubiquitous as π. For example, it is shown in [18] that the expected length, W3 , of
a three-step unit-length random walk in the plane is given by
( ) ( )
3 21/3 6 1 27 22/3 6 2
(36) W3 = Γ + Γ .
16 π 4 3 4 π4 3

We recall that Γ(1/2)2 = π and that similar algorithms exist for Γ(1/3), Γ(1/4), and Γ(1/6) [15, 14].

2. More about Complexity Reduction. To illustrate the stunning complexity reduction in the elliptic
algorithms for Pi, let us explicitly provide a complete set of algebraic equations approximating π to well over a
trillion digits.

The number π is transcendental and the number 1/a20 computed next is algebraic;
nonetheless they coincide for over 1.5 trillion places.
√ √
Set a0 = 6 − 4 2, y0 = 2 − 1 and then solve the system in Figure 17.

This quartic algorithm, with the Salamin–Brent scheme, was first used by Bailey in 1986 [17] and was
used repeatedly by Yasumasa Kanada (see Figure 12), in Tokyo in computations of π over 15 years or so,
culminating in a 200 billion decimal digit computation in 1999. As recorded in Figure 11, it has been used
twice very recently by Takahashi. Only thirty five years earlier in 1963, Dan Shanks—a very knowledgeable
participant—was confident that computing a billion digits was forever impossible. Today it is ‘reasonably easy’
on a modest laptop. A fine self-contained study of this quartic algorithm—along with its cubic confrere also
described in Section 10—can be read in [23]. The proofs are nicely refined specializations of those in [16].

3. Following π on the Web. One can now follow Pi on the web through Wikipedia, MathWorld or elsewhere,
and indeed one may check the performance of π by looking up ‘Pi’ at http://www.google.com/trends. This
link shows very clear seasonal trends. with a large spike around Pi Day. The final spike (F) is for Tau Day
(6.28)—a joke that many seem not have realized is a joke.12
12 www.washingtonpost.com/blogs/blogpost/post/tau-day-replace-pi-make-music-with-tau/2011/06/28/AG6ub6oH_blog.html

27
 p
4
p
4
1− 1 − y0 4 1− 1 − y10 4
, a1 = a0 (1 + y1 )4 − 23 y1 1 + y1 + y1 2 , a11 = a10 (1 + y11 )4 − 223 y11 1 + y11 + y11 2



y1 = p
4
y11 = p
4
1+ 1 − y0 4 1+ 1 − y10 4
p
4
p
4
1− 1 − y1 4 1− 1 − y11 4
, a2 = a1 (1 + y2 )4 − 25 y2 1 + y2 + y2 2 , a12 = a11 (1 + y12 )4 − 225 y12 1 + y12 + y12 2



y2 = p
4
y12 = p
4
1+ 1 − y1 4 1+ 1 − y11 4
p
4
p
4
1− 1 − y2 4 1− 1 − y12 4
, a3 = a2 (1 + y3 )4 − 27 y3 1 + y3 + y3 2 , a13 = a12 (1 + y13 )4 − 227 y13 1 + y13 + y13 2



y3 = p
4
y13 = p
4
1+ 1 − y2 4 1+ 1 − y12 4
p
4
p
4
1− 1 − y3 4 4 9 2
1− 1 − y13 4
, a14 = a13 (1 + y14 )4 − 229 y14 1 + y14 + y14 2


y4 = p
4
, a4 = a3 (1 + y4 ) − 2 y4 1 + y4 + y4 y14 = p
4
1 + 1 − y3 4 1+ 1 − y13 4
p
4
p
4
1− 1 − y4 4 1− 1 − y14 4
, a5 = a4 (1 + y5 )4 − 211 y5 1 + y5 + y5 2 , a15 = a14 (1 + y15 )4 − 231 y15 1 + y15 + y15 2



y5 = p
4
y15 = p
4
1 + 1 − y4 4 1+ 1 − y14 4
p p
1 − 4 1 − y5 4 1− 4
1 − y15 4
, a6 = a5 (1 + y6 )4 − 213 y6 1 + y6 + y6 2 , a16 = a15 (1 + y16 )4 − 233 y16 1 + y16 + y16 2



y6 = p
4
y16 = p
4
1 + 1 − y5 4 1+ 1 − y15 4
p p
1 − 4 1 − y6 4 1− 4
1 − y16 4
, a7 = a6 (1 + y7 )4 − 215 y7 1 + y7 + y7 2 , a17 = a16 (1 + y17 )4 − 235 y17 1 + y17 + y17 2



y7 = p
4
y17 = p
4
1 + 1 − y6 4 1+ 1 − y16 4
p p
1 − 4 1 − y7 4 1− 4
1 − y17 4
, a8 = a7 (1 + y8 )4 − 217 y8 1 + y8 + y8 2 , a18 = a17 (1 + y18 )4 − 237 y18 1 + y18 + y18 2



y8 = p y18 = p
1 + 4 1 − y7 4 1+ 4
1 − y17 4
p p
1 − 4 1 − y8 4 1− 4
1 − y18 4
, a9 = a8 (1 + y9 )4 − 219 y9 1 + y9 + y9 2 , a19 = a18 (1 + y19 )4 − 239 y19 1 + y19 + y19 2



y9 = p y19 = p
1 + 4 1 − y8 4 4
1 + 1 − y18 4
p
4
p
1− 1 − y9 4 1 − 4 1 − y19 4
, a10 = a9 (1 + y10 )4 − 221 y10 1 + y10 + y10 2 , a20 = a19 (1 + y20 )4 − 241 y20 1 + y20 + y20 2 .



y10 = p
4
y20 = p
1+ 1 − y9 4 1 + 4 1 − y19 4

Figure 17: The system of equations used to compute π to 1.5 trillion places

4. The Difficulty of Popularizing Accurately. Let me finish on this theme. Even after many helpful
comments from readers, errors probably remain in my article. So I tell the story below with no particular
rancour.

Paul Churchland in [19] offers a fascinating set of essays full of interesting anecdotes—which I have no
particular reason to doubt. Nonetheless, the very brief quote at the start of Section 16, regarding the legislation
of values of π, contains four inaccuracies. As noted above: (i) the event took place in 1896/7 and (ii) in Indiana
(a northern state); (iii) the prospective bill, #246, offered a geometric construction with inconsistent conclusions
and certainly offers no one exact value. Finally, (iv) the intent seems to have been pecuniary, not hygienic [27].

As often, this makes me wonder whether mathematics popularization is especially prone to error or if the
other disciplines just seem better described because of my relative ignorance. On April 1, 2009, an article
entitled “The Changing Value of Pi” appeared in the New Scientist with an analysis of how the value of pi has
been increasing over time. I hope but am not confident that all readers noted that April 1st is “April Fool’s
day.” (See also entry seven of http://www.museumofhoaxes.com/hoax/aprilfool/.)

Acknowledgements. Thanks are due to many, especially my close collaborators P. Borwein and D. Bailey.

28
 Figure 18: Google’s trend line for ‘Pi’

References
[1] F. Aragon, D. H. Bailey, J.M. Borwein and P.B. Borwein, “Tools for visualizing real numbers: Planar
number walks.” Submitted Mathematical Intelligencer, June 2012.
[2] F. Amoroso and C. Viola, “Irrational and Transcendental numbers,” in volume 2 of Mathematics and
Culture, La matematica: Problemi e teoremi, Guilio Einaudi Editori, Turino. September 2008.
[3] J. Arndt and C. Haenel, Pi Unleashed, Springer-Verlag, New York, 2001.

[4] H.Christian von Baeyer, Information The New Language of Science, Weidenfeld and Nicolson, 2003.
[5] D.H. Bailey and J. M. Borwein, “Exploratory Experimentation and Computation.” Notices of the AMS,
58 (10) (2011), 1410–1419. See also http://www.ams.orolnotices.
[6] David H. Bailey and Jonathan M. Borwein, “Ancient Indian Square Roots: An Exercise in Forensic
Paleo-Mathematics.” American Mathematical Monthly. October (2012).

[7] D. Bailey, J. Borwein, N. Calkin, R. Girgensohn, R. Luke, and V. Moll, Experimental Mathematics in
Action, A K Peters Inc, 2007.

[8] D.H. Bailey, J.M. Borwein, A. Mattingly, and G. Wightwick, “The Computation of Previously Inaccessible
Digits of π 2 and Catalan’s Constant.” Notices of the AMS, Accepted, July 2011.

[9] L. Berggren, J.M. Borwein and P.B. Borwein, Pi: a Source Book, Springer-Verlag, (2004). Third Edition,
2004.

[10] N.D. Baruah, B.C. Berndt, and H.H. Chan, “Ramanujan’s series for 1/π: A survey,” Amer. Math. Monthly
116 (2009), 567–587.

[11] D. Blatner, The Joy of Pi, Walker and Co., New York, 1997.
[12] J.M. Borwein, “Brouwer-Heyting sequences converge,” Mathematical Intelligencer, 20 (1998), 14–15.

[13] J.M. Borwein, “La vita di pi greco: from Archimedes to ENIAC and beyond,” in volume 2 of Mathematics
and Culture, La matematica: Problemi e teoremi, Guilio Einaudi Editori, Turino. September 2008 (Italian).
November 2009 (French).
[14] J.M. Borwein and D.H. Bailey, Mathematics by Experiment: Plausible Reasoning in the 21st Century,
Second expanded edition, AK Peters Ltd, 2008.

29
[15] J.M. Borwein and P.B. Borwein, Pi and the AGM, John Wiley and Sons, 1987.

[16] J.M. Borwein and P.B. Borwein, “Ramanujan and Pi,” Scientific American, February 1988, 112–117.
Reprinted in pp. 187-199 of Ramanujan: Essays and Surveys, Bruce C. Berndt and Robert A. Rankin
Eds., AMS-LMS History of Mathematics, vol. 22, 2001. (Collected in [9].)
[17] J.M. Borwein, P.B. Borwein, and D.A. Bailey, “Ramanujan, modular equations and pi or how to compute
a billion digits of pi,” MAA Monthly, 96 (1989), 201–219. Reprinted in Organic Mathematics Proceedings,
www.cecm.sfu.ca/organics, 1996. (Collected in [9].)

[18] J. Borwein, D. Nuyens, A. Straub, and J. Wan. “Some Arithmetic Properties of Short Random Walk
Integrals.” Ramanujan Journal. 26 (2011), 109–132.

[19] P. Churchland, Neurophilosophy at work, Cambridge University Press, 2007.

[20] William Dunham, Euler: The Master of Us All Dolciani Mathematical Expositions, No 22, Mathematical
Association of America, 1999.

[21] P. Eymard and J.-P. Lafon, The Number π, American Mathematical Society, Providence, 2003.
[22] J. Guillera, “Hypergeometric identities for 10 extended Ramanujan-type series,” Ramanujan J., 15 (2008),
219–234.
[23] J. Guillera, “Easy proofs of some Borwein algorithms for π,” American Math. Monthly, 115 (2008),
850–854.
[24] S.K. Lucas, “Integral approximations to Pi with nonnegative integrands,” American Math. Monthly, 116
(2009), 166–172.
[25] S. McCartney, ENIAC: The Triumphs and Tragedies of the World’s First Computer, Walker and Co.,
New York, 1999.
[26] H. C. Schloper, “The Chronology of Pi,” Mathematics Magazine, Jan-Feb 1950, 165–170; Mar-Apr 1950,
216–288; and May-Jun 1950, 279–283. (Collected in [9].)
[27] D. Singmaster, “The legal values of Pi,” Mathematical Intelligencer, 7 (1985), 69–72. (Collected in [9].)

[28] H. Tsumura, “An elementary proof of Euler’s formula for ζ(2m).” American Math. Monthly, May (2004),
430–431.

[29] W. Zudilin, “Ramanujan-type formulae for 1/π: A second wind.” ArXiv:0712.1332v2, 19 May 2008.

There are many other Internet resources on π, a reliable selection is kept at www.experimentalmath.info.

30