b232 PDF

The Discovery of the Series Formula for π by Leibniz, Gregory and Nilakantha
Author(s): Ranjan Roy

Source: Mathematics Magazine, Vol. 63, No. 5 (Dec., 1990), pp. 291-306
Published by: Mathematical Association of America
Stable URL: http://www.jstor.org/stable/2690896
Accessed: 27-02-2017 22:02 UTC
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 [email protected].
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
Mathematical Association of America is collaborating with JSTOR to digitize, preserve and extend access
to Mathematics Magazine
This content downloaded from 195.251.161.31 on Mon, 27 Feb 2017 22:02:42 UTC
All use subject to http://about.jstor.org/terms
ARTICLES
The Discovery of the Series Formula for 7r by

Leibniz, Gregory and Nilakantha
RANJAN ROY
Beloit College
Beloit, WI 53511
1. Introduction
The formula for -r mentioned in the title of this article is
4 3 57 . (1)
One simple and well-known moderm proof goes as follows:
x I
arctan x = | 1 +2 dt
x3 +5 - +2n + 1 x t2n+2
+ -w3 - +(-I)rl2n+1 +(-I)l?lf dt.
The last integral tends to zero if Ix < 1, for
'o t+2dt < jt dt - iX2n+3 20 as n oo.
Thus, arctan x has an infinite series representation for Ix I < 1:
3 X5
arctan x = x- - + - - . (2)
The series for -r/4 is obtained by setting x = 1 in (2). The series (2) was obtained
independently by Gottfried Wilhelm Leibniz (1646-1716), James Gregory
(1638-1675) and an Indian mathematician of the fourteenth century or probably
the fifteenth century whose identity is not definitely known. Usually ascribed to
Nilakantha, the Indian proof of (2) appears to date from the mid-fifteenth century and
was a consequence of an effort to rectify the circle. The details of the circumstances
and ideas leading to the discovery of the series by Leibniz and Gregory are known. It
is interesting to go into these details for several reasons. The infinite series began to
play a role in mathematics only in the second half of the seventeenth century. Prior to
that, particular cases of the infinite geometric series were the only ones to be used.
The arctan series was obtained by Leibniz and Gregory early in their study of infinite
series and, in fact, before the methods and algorithms of calculus were fully
developed. The history of the arctan series is, therefore, important because it reveals
early ideas on series and their relationship with quadrature or the process of findin
the area under a curve. In the case of Leibniz, it is possible to see how he used and
291
292 MATHFMATICS MAGAZINE
transformed older ideas on quadrature to develop his methods. Leibniz's work, in fact,
was primarily concerned with quadrature; the -r/4 series resulted (in 1673) when he
applied his method to the circle. Gregory, by comparison, was interested in finding an
infinite series representation of any given function and discovered the relationship
between this and the successive derivatives of the given function. Gregoiy's discov-
ery, made in 1671, is none other than the Taylor series; note that Taylor was not born
until 1685. The ideas of calculus, such as integration by parts, change of variables, and
higher derivatives, were not completely understood in the early 1670s. Some particu-
lar cases were known, usually garbed in geometric language. For example, the
fundamental theorem of calculus was stated as a geometric theorem in a work of
Gregory's written in 1668. Similar examples can also be seen in a book by Isaac
Barrow, Newton's mentor, published in 1670. Of course, very soon after this transi-
tional period, Leibniz began to create the techniques, algorithms and notations of
calculus as they are now known. He had been preceded by Newton, at least as far as
the techniques go, but Newton did not publish anything until considerably later. It is,
therefore, possible to see how the work on arctan was at once dependent on earlier
concepts and a transitional step toward later ideas.
Finally, although the proofs of (2) by Leibniz, Gregory and Nilakantha are very
different in approach and motivation, they all bear a relation to the modern proof
given above.
2. Gottfried Wilhelm Leibniz (1 646-1716)
Leibniz's mathematical background' at the time he found the -r/4 formula can be
quickly described. He had earned a doctor's degree in law in February 1667, but had
studied mathematics on his own. In 1672, he was a mere amateur in mathematics.
That year, he visited Paris and met Christiaan Huygens (1629-1695), the foremost
physicist and mathematician in continental Europe. Leibniz told the story of this
meeting in a 1679 letter to the mathematician Tschirnhaus, "at that time ... I did not
know the correct definition of the center of gravity. For, when by chance I spoke of it
to Huygens, I let him know that I thought that a straight line drawn through the
center of gravity always cut a figure into two equal parts,... Huygens laughed when
he heard this, and told me that nothing was further from the truth. So I, excited by
this stimulus, began to apply myself to the study of the more intricate geometry,
although as a matter of fact I had not at that time really studied the Elements
[Euclid] ... Huygens, who thought me a better geometer than I was, gave me to read
the letters of Pascal, published under the name of Dettonville; and from these I
gathered the method of indivisibles and centers of gravity, that is to say the
well-known methods of Cavalieri and Guldinus. "2
'For fuither information about Leibniz's mathematical development, the reader may consult: J. E.
Hofmann, Leibniz in Paris 1672-1676 (Cambridge: The Cambridge University Press, 1974) and its review
by A. Weil, Collected Papers Vol. 3 (New York: Spiinger-Verlag, 1979). An English translation of Leibniz's
own account, Historia et origo calculi differentialis, can be found in J. M. Child, The Early Mathematical
Manuscripts of Leibniz (Chicago: Open Court, 1920). An easily available synopsis of Leibniz's work in
calculus is given in C. H. Edwards, Jr., The Historical Development of the Calculus (New York:
Springer-Verlag, 1979).
2The Early Mathematical Manuscripts, p. 215.
Bonaventura Cavalieri (1598-1647) published his Geometria Indivisibilibus in 1635. This book was very
influential in the development of calculus. Cavalieri's work indicated that
VOL. 63, NO. 5, DECEMBER 1990 293
The study of Pascal played an important role in Leibniz's development as a

mathematician. It was from Pascal that he learned the ideas of the "characteristic
triangle" and "transmutation." In order to understand the concept of transmutation,
suppose A and B are two areas (or volumes) which have been divided up into
indivisibles usually taken to be infinitesimal rectangles (or prisms). If there is a
one-to-one correspondence between the indivisibles of A and B and if these
indivisibles have equal areas (or volumes), then B is said to be obtained from A by
transmutation and it follows that A and B have equal areas (or volumes). Pascal had
also considered infinitesimal triangles and shown their use in finding, among other
things, the area of the surface of a sphere. Leibniz was struck by the idea of an
infinitesimal triangle and its possibilities. He was able to derive an interesting
transmutation formula, which he then applied to the quadrature of a circle and
thereby discovered the series for nr. To obtain the transmutation formula, consider
two neighboring points P(x, y), and Q(x + dx, y + dy) on a curve y =f(x). First
Leibniz shows that area (AOPQ) = (1/2) area (rectangle (ABCD)). See FIGURE 1.
Here PT is tangent to y =f(x) at P and OS is perpendictular to PT. Let p denote
the length of OS and z that of AC = BD = ordinate of T.
0 a C D b
FIGURE 1
a a' ?
J x" dx = n+1'
0
when n is a positive integer.

Blaise Pascal (1623-1662) made important and fundamental contributions to projective geometry,
probability theory and the development of calculus. The work to which Leibniz refers was ptublished in
1658 and contains the first statement and proof of
f0sin4do =cosO0-cos0.
00
This proof is presented in D. J. Struik's A Source Book in Mathematics 1200-1800 (Cambridge: Harvard
University Press, 1969), p. 239.
Paul Guldin (1577-1643), a Swiss mathematician of considerable note, contributed to the development
of calculus, and his methods were generally more rigorous than those of Cavalieri.
294 MATHEMATICS MAGAZINE
Since AOST is similar to the characteristic A PQR,
dx ds
p z'
where ds is the length of PQ. Thus,
area (OPQ) = 2jpds = 2 zdx. (3)
Now, observe that for each point P on y =f(x) there is a corresponding point A.
Thus, as P moves from L to M, the points A form a curve, say Z = g(x). If sector
OLM denotes the closed region formed by y =f(x) and the straight lines OL and
OM, then (3) implies that
area (sector OLM)

2
= 2bfbg(x) dx. (4)
This is the transmutation formula of Leibniz. From (4), it follows that the area under
y=f(x) is
fydx = bf(b)- f (a) + area (sector OLM)

I f(b) - +b re
2 a I..,zdx). (5)
This is none other than a particular
iseasily seen from FIGURE 1 that
y x dy (6)
Substituting this value of z in (5), it follows
fbydx = [xy]b ff(b x)dy,

(a f(a)
which is what one g

Now consider a cir
In this case, (6) imp
_ x(1 x) x
z = V2x-x2 = (7)
2x-x2 ~2 -x y
so that
2z2
X +z2 (8)
In FIGURE 2, let AOB =
0 = area ( A AOB) + area (region between arc AB and line AB). (9)
By the transmutation formula (4), the second area is xf0z dt where z is given by
Now, from FIGURE 3 below it is seen that
VOL. 63, NO. 5, DECEMBER 1990 295
B(x, y) z = g(x)
A C 0
FIGURE 2 x
FIGURE 3
zdtf 2(xz xdu). (10)
Using (8) and (10), it is now possible to rewrite (9) as
0= y + lxz 1+t2dt
=[z(2-x) + xz]- +2 dt (since y =z(2-x))
z t2
=zf 2it.
1+ t2
At this point, Leibniz was able to use a technique employed by Nicolaus Mercator
(1620-1687). The latter had considered the problem of the quadrature of the
hyperbola y(l + x) = 1. Since it was already known that
|xdx n + I
he solved the problem by expanding 1/(1 + x) as an infinite series and integrating

term by term. He simultaneously had the expansion for log(l + x). Mercator pub-
lished this result in 1668, though he probably had obtained it a few years earlier. A
year later, John Wallis (1616-1703) determined the values of x for which the series is
valid. Thus, Leibniz found that
3 z5
0 =z-=
Z +Z_ 3 +5 - . (11)
In FIGURE 2, ABC = 0 and z = x/y = tan 0.
Of course, Leibniz did not invent the not
above until 1675, and his description of th
are the same.
The discovery of the infinite series for 7I was Leibniz's first great achievement. He
communicated his result to Huygens, who congratulated him, saying that this
remarkable property of the circle will be celebrated among mathematicians forever.
Even Isaac Newton (1642-1727) praised Leibniz's discovery. In a letter of October
24, 1676, to Henry Oldenburg, secretary of the Royal Society of London, he writes,
"Leibniz's method for obtaining convergent series is certainly very elegant, and it
would have sufficiently revealed the genius of its author, even if he had written
nothing else."3 Of course, for Leibniz this was only a first step to greater things as he
himself says in his "Historia et origo calculi differentialis."
3. James Gregory (1 638-1 675)
Leibniz had been anticipated in the discovery of the series for arctan by the Scottish
mathematician, James Gregory, though the latter did not note the particular case for
7-/4. Since Gregory did not publish most of his work on infinite series and also
because he died early and worked in isolation during the last seven years of his life,
his work did not have the influence it deserved. Gregory's early scientific interest was
in optics about which he wrote a masterly book at the age of twenty-four. His book,
the Optica Promota, contains the earliest description of a reflecting telescope. It was
in the hope, which ultimately remained unfulfilled, of constructing such an instru-
ment that he travelled to London in 1663 and made the acquaintance of John Collins
(1624-1683), an accountant and amateur mathematician. This friendship with Collins
was to prove very important for Gregory when the latter was working alone at St.
Andrews University in Scotland. Collins kept him abreast of the work of the English
mathematicians such as Isaac Newton, John Pell (1611-1685) and others with whom
Collins was in contact.5
Gregory spent the years 1664-1668 in Italy and came under the influence of the
Italian school of geometry founded by Cavalieri. It was from Stefano degli Angeli
(1623-1697), a student of Cavalieri, that Gregory learned about the work of Pierre de
Fermat (1601-1665), Cavalieri, Evangelista Torricelli (1608-1647) and others. While
in Italy, he wrote two books: Vera Circuli et Hyperbolae Quadratura in 1667, and
Geometriae Pars Universalis in 1668. The first book contains some highly original
ideas. Gregory attempted to show that the area of a general sector of an ellipse, circle
or hyperbola could not be expressed in terms of the areas of the inscribed and
circumscribed triangle and quadrilateral using arithmetical operations and root extrac-
tion. The attempt failed but Gregory introduced a number of important ideas such as
convergence and algebraic and transcendental functions. The second book contains
the first published statement and proof of the fundamental theorem of calculus in
geometrical form. It is known that Newton had discovered this result not later than
1666, although he did not make it public until later.
Gregory returned to London in the summer of 1668; Collins then informed him of
the latest discoveries of mathematicians working in England, including Mercator's
recently published proof of
2 X3
log( + x) =x- 2 + 3 -
3See H. W. Turnbull (ed.), The Correspondence of Isaac Newton (Cambridge: The University Press,
1960), Vol. 2, p. 130.
4Peter Beckmann has persuasively argued that Gregory must have known the series for 7/4 as well.
See Beckmann's A History of Pi (Boulder, Colorado: The Golem Press, 1977), p. 133.
5The reader might find it of interest to consult: H. W. Turnbull (ed.), James Gregory Tercentenary
Memorial Volume (London: G. Bell, 1939). This volume contains Gregory's scientific correspondence with
John Collins and a discussion of the former's life and work.
VOL. 63, NO. 5, DECEMBER 1990 297
Meditation on these discoveries led Gregory to publish his next book, Exercitationes
Geometricae, in the winter of 1668. This is a sequel to the Pars Universalis and is
mainly about the logarithmic function and its applications. It contains, for example,
the first evaluations of the indefinite integrals of sec x and tan x.6 The results are
stated in geometric form.
In the autumn of 1668, Gregory was appointed to the chair in St. Andrews and he
took up his duties in the winter of 1668/1669. He began regular correspondence with
Collins soon after this, communicating to him his latest mathematical discoveries and
requesting Collins to keep him informed of the latest activities of the English
mathematicians. Thus, in a letter of March 24, 1670, Collins writes, "Mr. Newtone of
Cambridge sent the following series for finding the Area of a Zone of a Circle to Mr.
Dary, to compare with the said Dary's approaches, putting R the radius and B the
parallel distance of a Chord from the Diameter the Area of the space or Zone between
them is =
B3 B5 B7 5B9 ,,7
2RB- 3R - 20I -R3 56RI5 576IR
This is all Collins writes about the series but it is

2fB (R2 - x2)1/2 dx after expanding by the bi
integration. Newton had discovered the binomial expansion for fractional exponents
in the winter of 1664/1665, but it was first made public in the aforementioned letter
of 1676 to Oldenburg.
There is evidence that Gregory had rediscovered the binomial theorem by 1668.8
However, it should be noted that the expansion for (1 - x)'/2 does not necessarily
6
A proof of the formula
fOsec d =logtan(- +2)
was of considerable significance and interest to mathematicians in the 1660's due to its connection with a
problem in navigation. Gerhard Mercator (1512-1594) published his engraved "Great World Map" in
1569. The construction of the map employed the famous Mercator projection. Edward Wright, a Cam-
bridge professor of mathematics, noted that the ordinate on the Mercator map corresponding to a latitude
of 00 on the globe is given by cJ sec 0 do, where c is suitably chosen according to the size of the map. In
1599, Wright published this result in his Certaine Errors in Navigation Corrected, which also contained a
table of latitudes computed by the continued addition of the secants of 1', 2', 3', etc. This approximation to
f0 sec ? do was sufficiently exact for the mariner's use. In the early 1640's, Henry Bond observed that the
values in Wright's table could be obtained by taking the logarithm of tan(wr/4 + 0/2). This observation
was published in 1645 in Richard Norwood's Epitome of Navigation. A theoretical pioof of this observation
was very desirable and Nicolaus Mercator had offered a sum of money for its demonstration in 1666. John
Collins, who had himself written a book on navigation, drew Gregory's attention to this problem and, as we
noted, Gregory supplied a proof. For more details, one may consult the following two articles by F. Cajori:
"On an Integration ante-dating the Integral Calculus," Bibliotheca Mathematica Vol. 14 (1913/14), pp.
312-19, and "Algebra In Napier's Day and Alleged Prior Invention of Logarithms," in C. G. Knott (ed.),
Napier Memorial Volume (London: Longmans, Green & Co., 1915), pp. 93-106. More recently, J. Lohne
has established that Thomas Harriot (1560-1621) had evaluated the integral J0 sec 0 do in 1594 by a
stereographic projection of a spherical loxodrome from the south pole into a logarithmic spiral. This work
was unpublished and remained unknown until Lohne brought it to light. See J. A. Lohne, "Thomas Harriot
als Mathematiker," Centaurus, Vol. 11, 1965-66, pp. 19-45. Thus it happened that, although J sec 0 dO is a
relatively difficult trigonometric integral, it was the first one to be discovered.
James Gregory, p. 89.
8See The Correspondence of Isaac Newton, Vol. 1, p. 52, note 1.
imply a knowledge of the binomial theorem. Newton himself had proved the
expansion by applying the well-known method for finding square roots of numbers to
the algebraic expression 1 - x. Moreover, it appears that the expansion of (1 - x)1/2
was discovered by Henry Briggs (1556-1630) in the 1620's, while he was construct-
ing the log tables.9 But there is no indication that Gregory or Newton knew of this. In
any case, for reasons unknown, Gregory was unable to make anything of the
series-as evidenced by his reply of April 20, "I cannot understand the series you
sent me of the circle, if this be the original, I take it to be no series." l0 However, by
September 5, 1670, he had discovered the general interpolation formula, now called
the Gregory-Newton interpolation formula, and had made from it a number of
remarkable deductions. He now knew how "to find the sinus having the arc and to
find the number having the logarithm." The latter result is precisely the binomial
expansion for arbitrary exponents. For, if we take x as the logarithm of y to the base
1 + d, then y = (1 + d)x and Gregory gives the solution as
x ~x(X -1) x_x_-_1__x_-_2

(1 + d) = 1 + xd + d2 + X(x-1)(x-2)d3 +..
1 -2 1-2 -3
It is possible that Newton's series in Collins' letter had set Gregory off on the course
of these discoveries, but he did not even at this point see that he could deduce
Newton's result. Soon after, he did observe this and wrote on December 19, 1670, "I
admire much my own dullness, that in such a considerable time I had not taken
notice of this."'12 All this time, he was very eager to learn about Newton's results on
series and particularly the methods he had used. Finally on December 24, 1670,
Collins sent him Newton's series for sin x, cos x, sin- lx and x cot x, adding that
Newton had a universal method which could be applied to any function. Gregory then
made a concentrated effort to discover a general method for himself. He succeeded.
In a famous letter of February 15, 1671 to Collins he writes, "As for Mr. Newton's
universal method, I imagine I have some knowledge of it, both as to geometrick and
mechanick curves, however I thank you for the series ye sent me and send you these
following in requital."'13 Gregory then gives the series for arctan x, tan x, sec x,
log secx, log tan( + -), arcsec(2eX), and 2arctantanh x/2. However, what he
had found was not Newton's method but rather the Taylor expansion more than forty
years before Brook Taylor (1685-1731). Newton's method consisted of reversion of
series, expansion by the binomial theorem, long division by series and term by term
integration.14 Thinking that he had rediscovered Newton's method, Gregory did not
9See D. T. Whiteside, "Henry Briggs: The Binomial theorem Anticipated," The Mathematical Gazette,
Vol. 15, (1962), p. 9. Whiteside shows how the expansion of (1 + x)l'2 arose out of Brigg's work on
logarithms.
'0James Gregory, p. 92.
"1In their review of the Gregory Memorial Volume, M. Dehn and E. Hellinger explain how th
binomial expansion comes out of the interpolation formula. See The American Mathematical Monthly, Vol.
50, (1943), p. 149.
12James Gregory, p. 148.
13Ibid., p. 170.
14It should be mentioned that Newton himself discovered the Taylor series around 1691. See D. T.
Whiteside (ed.), The Mathematical Papers of Isaac Newton, Vol. VII (Cambridge: The Cambridge
University Press, 1976), p. 19. In fact, Taylor was anticipated by at least five mathematicians. However, th
VOL. 63, NO. 5, DECEMBER 1990 299
publish his results. It is only from notes that he made on the back of a letter from
Gedeon Shaw, an Edinburgh stationer, dated January 29, 1671, that it is possible to
conclude that Gregory had the idea of the Taylor series. These notes contain the
successive derivatives of tan x, sec x, and the other functions whose expansions he
sent to Collins. The following extract from the notes gives the successive derivatives
of tan 0; here m is successively y, dy d'y, etc., and q = r tan 0. Gregory writes15
1st 2nd 3rd 4th
q2 ~2q 3 ___ 2 ___

m=q m = r + m=2q+ m =2r+ q + q
r r 2r r3
5th 6th
40q3 24q 5 136q2 240q4 120q6

m =16q + ~~+ -m = 16r + r + + ~
r2 r4 r r3 r
7th
m = 272q + 987-q + 1680 qO + 72Oq

r 2 r4 r6
8th
m = 272r + 3233 q + 11361 q3 + 13440 q5 + 5040 q,

r r 3 r5 r7'
It is clear from the form in whic

is formed by multiplying the derivative with respect to q of the preceding term by
r + e Now writing a =r0, Gregory gives the series in the letter to Collins as
r
follows:
a3 2a5 17a7 3233a9

3r2 15r4 315r6 181440r 8
The reasons for supposing that these notes were writ
Collins and were used to construct the series are (i) th
and (ii) Gregory's error in computing the coefficient

r2
be 1232 instead of 987 and which, in turn, leads to the error in the 8th m, where the
coefficient of g should be 3968 instead of 3233. This error is then repeated in the
r
series showing the origin of the series. Moreover, in the early parts of the notes,
Gregory is unsure about how he should write the successive derivatives. For example,
he attempts to write the derivative of sec 0 as a function of sec 0 but then abandons
the idea. He comes back to it later and sees that it is easier to work with m2 instead
of m since the m2's can be expressed as polynomials in tan 0. This is, of course,
sufficient to give him the series for sec 0. The series for log sec 0 and log tan(Or/4 + 0)
he then obtains by term by term integration of the series for tan 0 and sec 0
Taylor series is not unjustly named after Brook Taylor who published it in 1715. He saw the importance of
the result and derived several interesting consequences. For a discussion of these matters see: L.
Feigenbaum, "Brook Taylor and the Method of Increments," Archive for History of Exact Sciences, Vol. 34,
(1985), pp. 1-140.
15James Gregory, p. 352.
respectively. Naturally, the 3233 error is repeated. He must have obtained the series
for arc tan x from the 2nd m which can be written as
da r2 q2 q4
dq r2?q2
r +q ~ 2
r 4
r
The arctan series follows after term by term integration. Clearly, Gregory had made
great progress in the study of infinite series and the calculus and, had he lived longer
and published his work, he might have been classed with Newton and Leibniz as a
co-discoverer of the calculus. Unfortunately, he was struck by a sudden illness,
accompanied with blindness, as he was showing some students the satellites of
Jupiter. He did not recover and died soon after in October, 1675, at the age of
thirty-seven.
4. Kerala Gargya Nilakantha (c.1 450-c.1 550)
Another independent discovery of the series for arctan x and other trigonometric
functions was made by mathematicians in South India during the fifteenth century.
The series are given in Sanskrit verse in a book by Nilakantha called Tantrasangraha
and a commentary on this work called Tantrasangraha-vakhya of unknown author-
ship. The theorems are stated without proof but a proof of the arctan, cosine and sine
series can be found in a later work called Yuktibhasa. This was written in Malayalam,
the language spoken in Kerala, the southwest coast of India, by Jyesthadeva
(c.1500-c.1610) and is also a commentary on the Tantrasangraha. These works were
first brought to the notice of the western world by an Englishman named C. M.
Whish in 1835. Unfortunately, his paper on the subject had almost no impact and
went unnoticed for almost a century when C. Rajagopal16 and his associates began
publishing their findings from a study of these manuscripts. The contributions of
medieval Indian mathematicians are now beginning to be recognized and discussed
by authorities in the field of the history of mathematics.'7
It appears from the astronomical data contained in the Tantrasangraha that it was
composed around the year 1500. The Yuktibhasa was written about a century later. It
is not completely clear who the discoverer of these series was. In the Aryabhatiya-
bhasya, a work on astronomy, Nilakantha attributes the series for sine to Madhava.
This mathematician lived between the years 1340-1425. It is not known whether
16Rajagopal's work may be found in the following papers: (with M. S. Rangachari) "On an Untap
Source of Medieval Keralese Mathematics," Archive for History of Exact Sciences, Vol. 18, (1977), pp.
89-102; "On Medieval Kerala Mathematics," Archive for History of Exact Sciences, Vol. 35, (1986), pp.
91-99. These papers give the Sanskrit verses of the Tantrasangrahavakhya which describe the series for
the arctan, sine and cosine. An English translation and commentary is also provided. A commentary on the
proof of arctan series given in the Yuktibhasa is available in the two papers: "A Neglected Chapter of
Hindu Mathematics," Scripta Magnematica, Vol. 15, (1949), pp. 201-209; "On the Hindu Proof of
Gregory's Series," Ibid., Vol. 17, (1951), pp. 65-74. A commentary on the Yuktibhasa's proof of the sine
and cosine series is contained in C. Rajagopal and A. Venkataraman, "The sine and cosine power series in
Hindu mathematics," Journal of the Royal Asiatic Society of Bengal, Science, Vol. 15, (1949), pp. 1-13.
17See J. E. Hofmann, "Uber eine alt indische Berechnung von rr und ihre allgemeine Bedeutung,"
Mathematische-Physikalische Semester Berichte, Bd. 3, H. 3/4, Hamburg (1953). See also D. T. Whiteside,
"Patterns of Mathematical Thought in the later Seventeenth Century," Archive for History of Exact
Sciences, Vol. 1, (1960-1962), pp. 179-388. For a discussion of medieval Indian mathematicians and the
Tantrasangraha in particular, one might consult: A. P. Jushkevich, Geschichte der Mathematik in Mittelalter
(German translation Leipzig, 1964, of the Russian original, Moscow, 1961).
VOL. 63, NO. 5, DECEMBER 1990 301
Madhava found the other series as well or whether they are somewhat later
discoveries.
Little is known about these mathematicians. Madhava lived near Cochin in the very
southern part of India (Kerala) and some of his astronomical work still survives.
Nilakantha was a versatile genius who wrote not only on astronomy and mathematics
but also on philosophy and grammar. His erudite expositions on the latter subjects
were well known and studied until recently. He attracted several gifted students,
including Tuncath Ramanujan Ezuthassan, an early and important figure in Kerala
literature. About Jyesthadeva, nothing is known except that he was a Brahmin of the
house of Parakroda.
In the Tantrasangraha-vakhya, the series for arctan, sine and cosine are given in
verse which, when converted to mathematical symbols may be written as
y 1 ry 1 ry3 1 ry5 y
rarctanx= x- x 3 3 5 5 _ wherex<1,
s2 s2 s2
(22 + 2)r(2(2+ 2)r2 (4- (sine)

s2 s2 s2
r - = 2 -r 2 2 S + (cosine).
(22-2)r (22-2)r2 (42-4)r2
y s
x r-x
FIGURE 4
There are also some special features in the Tantrasangraha's treatment of the
series for 7i/4 which were not considered by Leibniz and Gregory. Nilakantha states
some rational approximations for the error incurred on taking only the first n terms of
the series. The expression for the approximation is then used to transform the series
for 7i/4 into one which converges more rapidly. The errors are given as follows:
+n-?f(n +i) i=1,2,3, (12)
where
1 ~~n/2 ad(n/2 )2?+1

fl(n) =-n f2(n) =2/ and f3(n) -(( 2/) 2
The transformed series are as follows:
Tr = 3 + 1 + 1 (13)
4 4 33- 3 53-5 73-7
and
IT 4 4 + 4
4 15+4d 1 35 +4 3 55 +4 5
Leibniz's proof of the formula for 7I/4 was found by the quadrat
proof in Jyesthadeva's book is by a direct rectification of an arc of a circle. In the
diagram given below, the arc AC is a quarter circle of radius one with center 0 and
OABC is a square. The side AB is divided into n equal parts of length 6 so that
n l = 1, P lPr = 8. E,F and Pr-iD are perpendicular to ?Pr. Now, the triangles
OEF and OP- 1D are similar, which gives
EF Pr_ 1 D- Pr-1D
OE =Opr , that is, EF= Opr;
0 A = P0
C B = Pn
FIGURE 5
The similarity of the A s Pr 1 Pr D and OAPr gives
Pr _ 1 r 1rD Pr 1'Pr
rOpr =- A
OA orp_D Op r
or Pr-1D
Thus,
Pr_- 1 Pr Pr - 1 Pr 8 8
EF =?Fr-6
?Fr-l OP 2 - 2- 15
1 + APr2 + r22
Since arc EG - EF - 5 8 arc of circle is

1 + r22 a oi
1iM ?r282 (14)

Of course, a clear idea of limits did not exist at that time so that the relation was
understood in an intuitive sense only. To evaluate the limit, Jyesthadeva uses two
lemmas. One is the geometric series
1 1- X + X2 - x3 + * .
VOL. 63, NO. 5, DECEMBER 1990 303
Jyesthadeva says that the expansion is obtained on iterating the following procedure:
1 =11
1 + x 1 + = 1x-x(1 ( 1 x
The other result is that
np?1
S(P) -=1P + 2P + .+nP ~ n+1 forlargen. (15)
A sketch of a proof is given by Jyesthadeva. He notes first that
nS(P- )=S(P) + S(P-) + S-')+ -- +Sr(P1) (16)

This is easy to verify. Relation (16) is also contained in the work of the tenth c
Arab mathematician Alhazen, who gives a geometrical proof in the Greek tradi
He uses it to evaluate S(3) and S'4' which occur in a problem about the volume of a
certain solid of revolution. Yuktibhasa shows that for p = 2,3
S(1))
S(P ) +S(') + ..+.S(L'- , (p17)
and then suggests that by induction the re

this is granted, it follows that if
s(l) - )~ n
it p
then by (16) and (17),
nS(lp 1) S(p) + _ or S() + n
and (15) is inductively proved.

We now note that (14) can be rewritten, after expanding 1/(1 + r282) into a
geometric series, as
[ n n n
7 = lim 8E 1-_3 E r2 + 85 E r4-
4 n -r r=1 r=1 r=1
= iim [1- 7 r 2+ nSEr4_
=1-1+ 1-1+...
357
where we have used relation (15) and the fact that 8= 1/n. Now consider the
approximation (12) and its application to the transformation of series. Suppose that
1.1 1 1
0"n = - + 5-
3 5+ .~~~n
. .-n f(n+l1),
where f(n + 1) is a rational function of n which

of ir/4 than the nth partial sum Sn. Changin
'8See The Historical Development of the Calculus (mentioned in footnote 1), p. 84. Alhazen is the
latinized form of the name Ibn Al-Haytham (c. 965-1039).
on -2 = 1- + 5-* + * 2 ?f(n-1)
Subtracting the second relation from the first,
?U = -o2 - n+f(n + 1) +f(n- 1). (18)
Then
-n = 0;n-2- Un
= On4 + U,12 ? Un
... =o- 1-U3 +u5 -U7 + *+ Un

=1-f(2)-u3 + u5-U7 + *u*l.
It is clear that
lim on=y
nl -
and therefore
- = 1 -f(2) - u + u - U + * - - (19)
Thus, we have a new series for 7I/4 which depends on how the function f(n) is
chosen. Naturally, the aim is to choose f(n) in such a way that (19) is more rapidly
convergent than (1). This is the idea behind the series (13). Now equation (18)
implies that
f(n + 1) +fi(n -1) = -U. (20)
For (19) to be more rapidly convergent th

negligible compared to 1/n. It is reasonable to assume f(n + 1) -f(n - 1) -f(n).
These observations together with (20) imply that f(n) = 1/2n is a possible rational
approximation in equation (12). With this f(n), the value of un is given by (20) to be
1 1 1 1
Un n 2(n + 1) 2(n-1) n3 - n
Substituting this in (19) gives us (13), which is
7T 1 1 1 1
1 -- + + 3
4 333 53 5 73-7
The other series
IT 4 4 4
4 15+4 1 35+4 3 55+4 5
is obtained by taking f(n)= n2/7 in (19).

n2 + 1
It should be mentioned that Newton was aware of the correction fl(n)= 1/2n.
For in the letter to Oldenburg, referred to earlier, he says, "By the series of Leibniz
also if half the term in the last place be added and some other like devices be
employed, the computation can be carried to many figures." However, he says
nothing about transforming the series by means of this correction.
VOL. 63, NO. 5, DECEMBER 1990 305
It appears that Nilakantha was aware of the impossibility of finding a finite series of
rational numbers to represent 7-. In the Aryabhatiya-bhasya he writes, "If the
diameter, measured using some unit of measure, were commensurable with that unit,
then the circumference would not likewise allow itself to be measured by means of
the same unit; so likewise in the case where the circumference is measurable by some
unit, then the diameter cannot be measured using the same unit." 19
The Yuktibhasa contains a proof of the arctan series also and it is obtained in
exactly the same way except that one rectifies only a part of the 1/8 circle.
It can be shown that if 7-/4 = S71 +f(n), where Sn is the nth partial sum, then
f(n) has the continued fraction representation
1 [ 1 12 22 32 1
f(n) ? ? +(21)
Moreover, the first three convergents are
fl(n)= 2n f2()= n/2

f2(n) = +1 and (n/2
f3(n) )2?+1
which are the values quoted in (13). Clearly, Nilakantha was using some procedure
which gave the successive convergents of the continued fraction (21) but the text
contains no suggestion that (20) was actually known to him. This continued fraction
implies that
2 12 22 32
4 - v 2 2+2+ 2+'
which may be compared with the continued fraction of the seventeenth century
English mathematician, William Brouncker (1620-1684), who gave the result
4 12 32 52
X '2+ 2+ 2+
The third approximation
f3( n) = (n/2)2 + 1
is very effective in obtaining good numerical

For example
1 1 20
1 - 3 + -Tg-+f3(20)
gives the value of 7- correct up to eight decimal places.20 Nilakantha himself gives
104348/33215 which is correct up to nine places. It is interesting that the Arab
mathematician Jamshid-al-Kasi, who also lived in the fifteenth century, had obtained
the same approximation by a different method.
19See Geschichte der Mathematik, p. 169.
20These observations concerning the continued fraction expansion of f(n) and its relation to the Indian
work and that of Brouncker, and concerning the decimal places in f (20), are due to D. T. Whiteside. See
"On Medieval Kerala Mathematics" of footnote 13.
5. Independence of these discoveries.
The question naturally arises of the possibility of mutual influence between or among
the discoverers of power series, in particular the series for the trigonometric func-
tions. Because of the lively trade relations between the Arabs and the west coast of
India over the centuries, it is generally accepted that mathematical ideas were also
exchanged. However, there is no evidence in any existing mathematical works of the
Arabs that they were aware of the concept of a power series. Therefore, we may grant
the Indians priority in the discovery of the series for sine, cosine and arctangent.
Moreover, historians of mathematics are in agreement that the European mathemati-
cians were unaware of the Indian discovery of infinite series.21 Thus, we may
conclude that Newton, Gregory and Leibniz made their discoveries independently of
the Indian work. In fact, it appears that yet another independent discovery of an
infinite series giving the value of 7r was made by the Japanese mathematician Takebe
Kenko (1664-1739) in 1722. His series is
= 4[1 2 + 22n1 (n !)2 ]22
This series was not obtained from the arctan series and its discussion is therefore not
included. However, the independent discovery of the infinite series by different
persons living in different environments and cultures gives us insight into the
character of mathematics as a universal discipline.
Acknowledgement. I owe thanks to Phil Straffin for encouraging me to write this paper and to the
referees for their suggestions.
21See "Patterns of Mathematical Thought in the later Seventeenth Century" of footnote 17. See also A.
Weil, "History of Mathematics: Why and How" in Collected Papers, Vol. 3 (New York: Springer-Verlag,
1979), p. 435.
22See D. E. Smith and Y. Mikami, A History of Japanese Mathematics (Chicago: Open Court, 1914).
This series was also obtained by the French missionary Pierre Jartoux (1670-1720) in 1720. He worked in
China and was in correspondence with Leibniz, but the present opinion is that Takebe's discovery was
independent. Leonhard Euler (1707-1783) rediscovered the same series in 1737. A simple evaluation of it
can be given using Clausen's formula for the square of a hypergeometric series.

b232 PDF

Uploaded by

Copyright:

Available Formats

b232 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

b232 PDF

Uploaded by

Copyright:

Available Formats

The Discovery of the Series Formula for π by Leibniz, Gregory and Nilakantha

Author(s): Ranjan Roy

The Discovery of the Series Formula for 7r by

The formula for -r mentioned in the title of this article is

One simple and well-known moderm proof goes as follows:

The last integral tends to zero if Ix < 1, for

'o t+2dt < jt dt - iX2n+3 20 as n oo.

Thus, arctan x has an infinite series representation for Ix I < 1:

2. Gottfried Wilhelm Leibniz (1 646-1716)

The study of Pascal played an important role in Leibniz's development as a

when n is a positive integer.

Since AOST is similar to the characteristic A PQR,

where ds is the length of PQ. Thus,

area (OPQ) = 2jpds = 2 zdx. (3)

area (sector OLM)

fydx = bf(b)- f (a) + area (sector OLM)

fbydx = [xy]b ff(b x)dy,

which is what one g

In FIGURE 2, let AOB =

zdtf 2(xz xdu). (10)

Using (8) and (10), it is now possible to rewrite (9) as

=[z(2-x) + xz]- +2 dt (since y =z(2-x))

he solved the problem by expanding 1/(1 + x) as an infinite series and integrating

3. James Gregory (1 638-1 675)

This is all Collins writes about the series but it is

fOsec d =logtan(- +2)

James Gregory, p. 89.

8See The Correspondence of Isaac Newton, Vol. 1, p. 52, note 1.

x ~x(X -1) x_x_-_1__x_-_2

'0James Gregory, p. 92.

12James Gregory, p. 148.

of tan 0; here m is successively y, dy d'y, etc., and q = r tan 0. Gregory writes15

1st 2nd 3rd 4th

q2 ~2q 3 ___ 2 ___

40q3 24q 5 136q2 240q4 120q6

m = 272q + 987-q + 1680 qO + 72Oq

m = 272r + 3233 q + 11361 q3 + 13440 q5 + 5040 q,

It is clear from the form in whic

a3 2a5 17a7 3233a9

and (ii) Gregory's error in computing the coefficient

15James Gregory, p. 352.

4. Kerala Gargya Nilakantha (c.1 450-c.1 550)

(22 + 2)r(2(2+ 2)r2 (4- (sine)

+n-?f(n +i) i=1,2,3, (12)

1 ~~n/2 ad(n/2 )2?+1

The transformed series are as follows:

The similarity of the A s Pr 1 Pr D and OAPr gives

Since arc EG - EF - 5 8 arc of circle is

1iM ?r282 (14)

The other result is that

A sketch of a proof is given by Jyesthadeva. He notes first that

nS(P- )=S(P) + S(P-) + S-')+ -- +Sr(P1) (16)

and then suggests that by induction the re

then by (16) and (17),

nS(lp 1) S(p) + _ or S() + n

and (15) is inductively proved.

= iim [1- 7 r 2+ nSEr4_

where f(n + 1) is a rational function of n which

Subtracting the second relation from the first,

q2 ~2q 3 _ 2 _