Emch - Sridharan - Srinivas - Contributions Ot The History of Indian Mathematics (2005)

Contributions
to the
History of Indian Mathematics
Contributions
to the
History of Indian Mathematics
Editors
Gérard G. Emch
R. Sridharan
M. D. Srinivas
Preface
The rst Joint India-AMS meeting in Mathematics was held in

Bangalore in December 2003. One of its themes was the \History of
Indian Mathematics". Two sessions on this theme were organised jointly
by Gerard G. Emch from the U.S., and R. Sridharan from India. These
sessions were held on the 18th and the 20th of December.
The speakers at these sessions covered a wide spectrum of topics
ranging from Vedic Prosody and ancient Buddhist logic at one end to
the contributions of Srinivasa Ramanujan and Indian contributions to
Quantum Statistics at the other.
The lectures were enthusiastically received and it was felt that a
volume based on these lectures in detail might be useful to the mathe-
matical community; we thus formed an editorial committee to compile
such a volume. Articles were invited from the speakers and were refer-
eed.
This volume, which is the outcome, begins with an overview1 of the
subject and is divided into three sections.
The rst section which deals with the ancient period has two articles,
one on Vedic Prosody and the work of Pingala
_ and the other on Buddhist
Logic.
The next section which discusses the mathematics of the classical
and medieval periods begins with two articles, one on the work of Brah-
magupta on Bhavana and its applications, another on the contributions
of Bhaskara II to the mathematics of Karan.i or surds. The next article
is on the use of power series techniques by the medieval Kerala School
of Mathematics. The next two articles focus on the nature of algorithms
in Indian Mathematics and Astronomy. The nal article of this section
is on the notion of proofs in Indian Mathematics and the tradition of
Upapattis in Mathematics and Astronomy of India.
1
The overview is based on a lecture delivered by one of the editors at the Nehru
Centre, Mumbai, during September 2002 and is to be published by the Nehru Centre
in \Science in India: Past and Present", 2005 (in Press). We are grateful to Mr
Sahani, the Executive Secretary of the Nehru Centre for allowing us to include it in
this volume.
viii
The third section is devoted to the modern period. The rst arti-
cle points to some surprising contributions of Srinivasa Ramanujan on
partial fractions while the second surveys the history of some of the
contributions of Indian mathematicians to Quantum Statistics.
The editors are grateful to the organisers of the AMS-India confer-
ence for their hospitality and to the contributors for their enthusiastic
response. We are very happy to acknowledge the generous help of C. S.
Seshadri and his colleagues at the Chennai Mathematical Institute (in
particular C. S. Aravinda and V. Balaji) in making this volume possible.
We are particularly thankful for the immense task accomplished by of
P. Vanchinathan who prepared the camera-ready copy for the publisher.
We thank J. K. Jain of Hindustan Book Agency whose active support
and interest has made our endeavour a pleasure.
June 17, 2005 Gerard G. Emch

R. Sridharan
M. D. Srinivas
Contents
Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .vii
I. Introductory Overview
Mathematics in Ancient and Medieval India
R. Sridharan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
II. Ancient Period

Sanskrit Prosody, Pingala
_ Sutras and Binary Arithmetic
R. Sridharan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .33
Shedding Some Localic and Linguistic Light on the

Tetralemma Conundrums
F. E. J. Linton . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
III. Classical and Medieval Period

Brahmagupta's Bhavana: Some Re ections
Amartya Kumar Dattta . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
The Karan.: How to Use Integers to Make Accurate

Calculations on Square Roots
Francois Patte . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
Relations between Approximations to the Sine in

Kerala Mathematics
Kim Plofker . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
Algorithms in Indian Mathematics

M. S. Sriram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
Algorithms in Indian Astronomy

K. Ramasubramanian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
Proofs in Indian Mathematics

M. D. Srinivas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .209
IV. Modern Period

Ramanujan and Partial Fractions
George E. Andrews . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
Contributions of Indian Mathematicians to Quantum Statistics

Gerard G. Emch . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
Mathematics in Ancient and Medieval
India
R. Sridharan
Chennai Mathematical Institute, Chennai 600 017
rsridhar@cmi.ac.in
Baruch Spinoza (1632-1677), that gentle philosopher of the seven-

teenth century laid down as his credo \Sedulo curavi, humanas actiones
non redere, non lugere, neque detestari, sed intelligere". He says that
when he sets out to interpret the thoughts and history of a bygone age,
he obeys the above injunctions set by him for himself { not to ridicule,
not to mourn, never to detest, but try to understand { an excellent piece
of advice to anyone who interprets ancient writings. I hope I abide by
Spinoza's admonitions.
Indian contributions to decimal systems and place value:
Some concepts are so fundamental that they get absorbed in the

general thinking of all the peoples of the world and it is indeed hard to
pinpoint a specic period of time for their birth. One such is the basic
idea of introducing the place value system and the number zero, the
credit for which goes to Ancient India. The notion of the place value
has become òbvious' and is now learnt even by young children at an
early age. But, to have a true appreciation of the depth of this notion,
one has merely to look at the clumsy Roman numerals and note that, not
too long ago, these were very much in vogue in the West, in spite of their
cumbersome notation. It has been held by a few of the Western scholars
that the Arabs were responsible for the decimal notation even though
the Arabs themselves clearly indicated that they were merely conveyers
of the wisdom of India to the West. But it is now generally accepted that
Lecture given at the Nehru Centre, Mumbai, on 21st September 2002. To be
published by the Nehru Centre in \Science in India: Past and Present", (2005, in
Press)
1
2 R. Sridharan
the decimal notation is indeed a contribution of India, though, xing an

exact date for this invention is extremely hard. A possible guess could be
100 B.C. Of course, large numbers have always been favourites of Indians
at all epochs and are mentioned in Vedic literature, in the epics like the
Ramayan.a and the Mahabharata. A true assessment of the immensity
of this great idea of the decimal system and the invention of zero can
be found in the following quotation from Professor G.B. Halstead (\On
the foundation of the technique of Arithmetic" Chicago, 1912). \The
importance of the creation of the zero-mark can never be exaggerated.
This gives to airy nothing not merely a local habitation and a name,
a picture, a symbol, but helpful power; it is the characteristic of the
Hindu race whence it sprang. It is like coining Nirvana into dynamos.
No single mathematical creation has been more potent for the general
on go of intelligence and power".
The mathematical sophistication of the Harappan culture:
As we know, the Indus Valley civilisation (now, more properly termed

the Harappan civilisation) dates back to 2500-3000 B.C. This was a city
civilisation and hence, obviously, for town planning, a knowledge of prac-
tical geometry must have been a must. Apart from this, we have clear
evidence of the geometric sophistication of this ancient civilisation. The
patterns on the pottery excavated at these sites show that the artisans
were very well acquainted with the concept of circles and well versed in
the use of dividers for scratching on the surface of pottery. The mo-
tif of triangles is also common on pottery. Apart from the use of such
elementary geometry, a sophisticated system of weights and measures
used by them shows that the civilisation was well acquainted with basic
arithmetic. One of the highlights of this civilisation was its knowledge
of brick building technology. The most ideal ratio of the length, breadth
and the thickness of a brick to have eÆcient binding strength was known
to the Harappan people.

The vedic period and the Sulva geometry:
There is a huge gap in our knowledge of the interim between the

end of the Harappan civilisation and the beginnings of the vedic civili-
sation. Whatever happened during this period, it is safe to assume that
the skills and the culture of the earlier period could not all have disap-
peared and it is indeed reasonable to postulate that there was a basic
continuity in many human endeavours. In particular, the technology of
Mathematics in India 3
brick building during the Harappan age was perhaps put to good use for
the construction of the sacricial altars during the vedic period. (Let
us remember that culture sometimes is dened as a state of the mind
produced by things forgotten.)
Mathematics of the vedic period consists of those geometric tech-
niques needed for the construction of the Vedis (altars) and Agnis (re
places) described by the Brahman.as for the obligatory (nitya) and op-

tional (kamya) rites. This knowledge is contained in Sulva Sutras which

are a part of the so called Kalpa Sutras (more particularly of the Srauta
Sutras) attached to the Vedas as Vedangas _ (\limbs" of the Vedas).

There are at least nine extant Sulva Sutras: Laugaks.i, Manava, Varaha,

Baudhayana, Vadhula, Apastamba, Hiran.yakesin, Katyayana and Mai-

trayan.a. The date of the Sulva Sutras is purely conjectural except that
from their language, it is clear that they are pre-Pan.inian, and that
800 B.C. would be a probable date for the codication of the Sutras
(one must remember that the vedic rituals must have preceded even the

Vedas so that the Sulva tradition must be much older than the age of
their codication).

The word Sulva means a measuring rope or a rod (though the Sutras
themselves do not use this; rajju is used for a measuring rod). We shall

give a very brief sketch of the contents of the Sulva Sutras. We begin
most importantly with the theorem of the square on the diagonal of a
right angled triangle, popularly attributed to Pythagoras. A point to re-

member is that the Sulva Sutras deal with right angled triangles not for
their own sake but as two halves of a square or a rectangle when cut by
a diagonal. While it is true that various particular integral values of the
sides of a right angled triangle like 3, 4, 5 were known to all the ancient
civilisations of the world, the real geometric signicance of the theorem
was perhaps rst realised by the altar building vedic priests. It is be-
lieved that the theorem was well known to the Indians at least by 800

B.C. which is the estimated date of the oldest Sulva Sutra, namely, that

of Baudhayana. Satapatha Brahman.a gives 36 units as the length of the
east-west line segment (prac) of the so called Maha vedi, and 30 units
as one of the north-south sides and the prac and half the side, making
a right angled triangle with sides 36, 15 and 39. The Sulva geometers

were well aware of the notion of similarity of gures; (the Satapatha
Brahman.a mentions the mode of increasing the size of the vedi, four-
4 R. Sridharan
teen fold, to accommodate the 101st performance of the sacrice). The
Sautr
paman.paltar, pfor example, uses a right angled triangle whose sides
are 5 3, 12p 3, 13p 3 and the p Asvamedha altar involved a triangle with
sides of 15 2, 36 2 and 39 2. Thus irrational numbers occurred very
naturally, the Vedic priests being interested in constructing altars whose
areas were given multiples of the area of another altar. Many rational
and integral solutions of the Pythagorean equation x2 + y2 = z 2 were

also quite well known to the Sulva geometers. For example, the triples
(3, 4, 5), (5, 12, 13), (7, 24, 25), (8, 15, 17), (12, 35, 37) are explicitly
used.

The question whether the Sulvak aras had a proof of the Pythagoras
theorem can be answered in the following way: Their approach to geom-
etry was purely functional and they did not have the deductive approach

to geometry as the Greeks had. The Sulvak aras looked upon geometry
as a necessity and a tool and their approach was therefore empirical.
Among the many simple geometrical results that the Sulva Sutras
mention, are the various properties of the diagonals of a rectangle, con-
structions of isosceles right angled triangles, parallelograms, rhombi,
construction of a trapezium similar to another and so on.
As one of the most remarkable statements found in p
the Sulva Sutras,
I would like to highlight a rational approximation to 2 which occurs

in the Baudhayana, Apastamba
and Katyayana Sulva Sutras. The ap-
proximation is
p 1 1 1
21+ + ;
3 (3)(4) (3)(4)(34)
which is correct up to ve decimal places. It is indeed remarkable,
whereas Greek geometry came to a grinding (though temporary) halt
when the Greeks met with the phenomenon of irrationality of numbers
for the rst time, Indian thought went ahead, quite happy to accommo-
date the irrationals. (Incidentally, a commentator by name Rama who
lived in the middle of the 15th century A.D., in a place called Naimis.a
near modern Lucknow, improved upon this approximation and obtained
p 1 1 1 1 1
21+ + + ;
3 (3)(4) (3)(4)(34) ((3)(4)(34)(33) (3)(4)(34)(34)
which gives a better approximation, correct up to seven decimal places).

The Sulvak aras sometimes had to deal with very sophisticated in-
determinate equations in two or even four unknowns for the construc-
tion of the altars. For instance, in the construction of Garud.a Cayana,
Baudhayana explains a procedure to have a Vedi constructed with ve
layers of bricks, each layer consisting of 200 bricks, covering an area of
7 12 sq. purshas and he indicates a construction of this vedi, using square
bricks of four dierent sizes. This leads to a fairly intricate indetermi-
nate equation.

Baudhayana and Apastamba give dierent solutions to this problem.
Another curiosity is a derivation of an arithmetical identity from
geometry and is found in Baudhayana Sulva Sutra (cf. [5]), where he
wishes to construct larger and larger squares starting with a small one,
by adding successively gnomons to it. The following geometrical gure
illustrates this process, and gives a geometric proof of the identity: 1 +
3 + 5 + + 2n 1 = n2 .
** ** **
** 1 *
** * *

We end our very short sketch of the mathematics of the Sulva period with

the following passage which is a numerical curiosity from the Satapatha
Brahman.a (cf. [8, p. 351]) about nding all the divisors of 720.1
Now this Prajapati, the year, there are seven hundred and twenty
days and nights, his lights (being) those bricks; three hundred and sixty
enclosing stones and three hundred and sixty bricks with (special) for-
mulas. This Prajapati, the year has created all existing things, both what
breathes and the breathless, both gods and men: Having created all ex-
isting things he felt like one emptied out and was afraid of death. He
bethought to himself \How can I get these beings back into my body? how
can I be again the body of all these beings?". He divided his body into
two. There were three hundred and sixty bricks in one and as many in
1 See also I. 164 of the Rgveda esp., I. 164.11 and I. 164.48.
.
6 R. Sridharan
the other: He did not succeed. He made himself three bodies{in each of
them there were three eighties of bricks: He did not succeed. He made
himself four bodies of a hundred and eighty bricks each: He did not suc-
ceed. He made himself ve bodies{in each of them there were a hundred
and forty-four bricks: He did not succeed. He made himself six bodies
of a hundred and twenty bricks each: He did not succeed. He did not
develop himself seven fold (na saptadha vyabhavat). He made himself
eight bodies of ninety bricks each: He did not succeed. He made himself
nine bodies of eighty bricks each: He did not succeed. He made him-
self ten bodies of seventy-two bricks each: He did not succeed. He did
not develop eleven fold. He made himself twelve bodies of sixty bricks
each: He did not succeed. He did not develop either thirteen fold or
fourteen fold. He made himself fteen bodies of forty-eight bricks each:
He did not succeed. He made himself sixteen bodies of forty-ve bricks
each: he did not succeed. He did not develop seventeen fold. He made
himself eighteen bodies of forty bricks each: He did not succeed. He did
not develop nineteen fold. He made himself twenty bodies of thirty-six
bricks each: he did not succeed. He did not develop either twenty-one or
twenty-two or twenty-three fold. He made himself twenty-four bodies of
thirty bricks each. Then he stopped at the fteenth attempt and because
he stopped at the fteenth arrangement, there are fteen forms of waxing
and fteen of the waning (of the moon).
(Let me add on my own, one more numerical curiosity based on
the number seven hundred and twenty. The set of permutations of six
symbols which is seven hundred and twenty constitutes a group, the
so called symmetric group of degree six. This group contains fteen
transpositions (those permutations which interchange only two symbols)
and fteen other elements of order two which disturb all the six symbols.)
The contribution of the Jainas:

The Sulva period of mathematical development in India represents
also simultaneously the mathematical development brought about by
the Jains. Fixing dates for Jaina canonical literature is as diÆcult as
(if not more diÆcult than) xing the date for Vedic literature. Once,
it was thought that Vardhamana Mahavra, a contemporary of Buddha
was the founder of Jainism but it has now been conclusively estab-
lished that he really was the last of the Trthankaras; some of the earlier
Trthankaras
_ like R.s.abha, Bharata being well known gures in Hindu
Puranic literature also. Thus, Jainism which arose as a revolt against

the vedic sacricial practices is perhaps as old as the Vedas themselves
or even belongs to an earlier period. The mathematics of the Jains is

also as old as the mathematics of the Sulvas (if not older). But most
unfortunately, our knowledge of the mathematical contributions of the
Jains is only through commentaries on earlier works, the original math-
ematical works of the Jains having not come to light yet. In fact, there
is a genuine need to search for original Jain manuscripts.
A name which stands out as a Jain mathematician of fame is Uma-
ambara Jains. He most probably
svati who belonged to the sect of Svet
lived around 150 B.C. and belongs to the Kusumapura school near mod-
ern Patna. We note in parenthesis that the later Astronomer - math-

ematician Aryabhat . a I (born 476 A.D.) also belonged to the Kusuma-
pura school. Umasvati was a great metaphysician. But he seems to
have contributed considerably to mathematics too. His famous work
is \Tatvarthadhigama
p Sutra Bhas.ya". Here are found the well known
approximation 10 to and substantial p contributions to the geometry
of the circles. The approximation 10 to was used by the Jains from
500 B.C. to 1500 A.D. The Jains came to study the circle in connection
with their theory of cosmology .
The Jains should have been well aware of the place value system
(their heavy calculations giving circumstantial evidence for this). They
had names for very large numbers too and eventually had a concept of
innity, however imperfect.
The Jains were aware of the law of indices: am an = am+n .
The Jain work Sthananga _ Sutra (c. 300 B.C.) classies the study of
mathematics under ten dierent heads: parikarma, vyavahara, rajju,
rasi, kalasavarn.a, yavat-tavat, varga, ghana, varga-varga and vikalpa.
Parikarma probably refers to the four fundamental operations of arith-
metic namely, addition, subtraction, multiplication and division; vyava-
hara probably means applied arithmetic, kalasavarn.a probably refers to

the study of fractions, rajju means geometry (as in the Sulva Sutras) rasi
means measurements in general, varga means squaring, ghana means
cubing and varga-varga may mean raising numbers to higher powers.
yavat-tavat is the \unknown" x in the ancient Indian mathematics.
vikalpa is the name given by the Jains to the theory of permutations
and combinations. Even though permutations were known from ancient
8 R. Sridharan
times (they occur in the prosody of Pingala,

_ as also in Susruta's work
\Susruta Sam.hita" around 6th century B.C. where mention is made
of 63 combinations which can be made out of 6 rasas (tastes)), the
Jains, however were the ones who treated the theory of permutations
and combinations systematically as a topic in mathematics and whose
general formulae are completed by the Jain mathematician Mahavra
(850 A.D.).
The Baks
.al
manuscript:
In 1881 A.D., in the course of excavations, a farmer came across a

mathematics manuscript in birch-bark at a village at Baks.al near Pe-
shawar. This manuscript was not complete; only about 70 leaves were
available, a greater portion having been apparently lost. The language
of the manuscript being Gatha (a modied form of prakr.ta) in S arada
script. This manuscript was later printed and published by the then
Government of India, edited by G.R. Kaye, with an introduction (Kaye
has now become well known for his bias and prejudice against Indians).
This manuscript is a collection of rules along with illustrative examples,
devoted to arithmetic and algebra, there being not much of geometry,
except for a few isolated problems. Perhaps there was a section on geom-
etry which had been lost. The topics on arithmetic deal with fractions,
square roots, problems on prot and loss and problems related to the rule
of three. The manuscript (whose date is controversial, Kaye having put
it around the 12th century A.D., though Datta believes that it belongs
to the 3rd or 4th century A.D.) must itself have been a commentary of
a much earlier work. We mention some of the mathematics contained
in the manuscript. For example, it nds the sum of a nite number of
fractions by reducing them to a common denominator; gives the sum of
n consecutive terms of an arithmetic or a geometric progression; much
more importantly gives a formula for the determination of approximate
values of quadratic surds. This formula is usually attributed to Heron,
a Greek mathematician of 200 A.D. An English translation of this Sutra
in question runs as follows:
If the number is a non square, subtract the nearest square number,
divide the remainder by twice (the nearest square), half square of this is
divided by the sum of the approximate root and this when subtracted
will give the corrected root.
Expressed algebraically, this reads that if A = a2 + r, then, approx-
imately,
p r ( 2ra )2
A=a+ :
2a 2(a + 2ra )
p
This
p formula
p is used p to give rational approximations to surds like 41,
105, 889 and 339009. Using such approximations, the following
kind of problem is for instance solved in the Baks.al manuscript: Given
the sum of the terms of an arithmetic progression whose rst term and
the common dierence are also given, to nd the approximate value of
the number of terms of this progression.
There are also problems involving indeterminate equations in the
Baks.ali manuscript. Similar problems are discussed by Bhaskara II
(about whom we shall discuss later) in Llavat. One is tempted to
conclude that such problems were commonly considered in all works
even before Bhaskara.
Chandas S
utra of Pingala
_ and Binary Arithmetic: ([11,12,18,19])
Prosody (Chandas) was one of the basic studies undertaken in India
right from ancient times. Indeed, prosody was an important Vedanga_
as is borne by the following sloka.
Chandah. padantu vedasya hastau kalpo'tha pat.hyate j

. srotramucyate jj
Jyotis.amayanam. chaks.urniruktam
About Pingala,
_ not much is known except that he lived most probably
three or four centuries before Christ and that he authored an impor-
tant work on Chandas written in Sutra style in eight chapters. (it is
not clear whether he lived during the time of the grammarian Pan.ini
or not). This work was commented upon by several people (Bharata,
Halayudha, Hemacandra, Jayadeva, Jayakrti, Kedara, Yadava Prakasa,
: : :) of which the commentary of Halayudha belonging to the 10th cen-
tury, called Mr.ta Sanjva~n, is very well known.
In the rst seven chapters of his work, Pingala
_ adopts the conven-
tional method of metrical analysis of the padas or verses. However, in
the last chapter of his work, he introduces a new method which leads
to binary arithmetic. Before explaining this method, let me begin with
some basic facts on metres. A metre consists of syllables contained in
a foot or pada (there being generally four feet in a stanza). A syllable
10 R. Sridharan
is of two kinds, either it is light (represented by ^ and called laghu)

consisting of a short vowel followed by at most one consonant or it is
heavy (represented by { and called guru) which consists either of a long
vowel, a diphthong or a vowel followed by two or more consonants. The
usual method of scanning verses is to divide any verse into units of three
syllables and assign names to metres on the basis of the combination of
the triplets of syllables. As we have remarked, in the rst seven chapters
of his work, Pingala
_ uses this method of classifying metres. However, in
the eighth chapter, Pingala
_ introduces a new method. Let me make a
remark on a standard convention in prosody. The prosodists ignore the
rst and the last syllable of a foot2 , so that, for example for the Gayatr
verse
tat saviturvaren.iyam
whose metrical representation is
{ ^^ { ^ { ^^,
the rst and the last syllables are ignored so that the Gayatr metre is
represented by the following sequence of six syllables.
^^ { ^ { ^.
Pingala
_ associates to a syllabic foot, a table called Prastara which
consists of rows of laghus and gurus, listed horizontally. The idea of
Pingala
_ is to attach a numerical value to each row in the Prastara by
giving the value 0 to a guru and the value 1 to a laghu. The rule for
constructing the Prastara (for the six syllabic verse) is as follows:
First start with all gurus, (six in number in our example). Next, a
new row of gurus is started but with the proviso that under the rst guru
in the line above a laghu is written instead. This procedure is continued
until we reach a row consisting entirely of six laghus
2 cf. e.g. A.A. MacDonnell \Vedic Grammar", Appendix II, p. 437 for general
rules of vedic prosody.
{ { { { { {
^ { { { { {
{ ^ { { { {
^ ^ { { { {
.. .. .. .. .. ..
. . . . . .
^ ^ ^ ^ ^ ^
Pingala
_ attaches a numerical value to each row as follows. Take for
example the syllabic pattern ^ { { { { {. The number attached to
it is just the binary expansion 1 1 + 0 2 + 0 22 + + 0 25 = 1,
increased by 1. i.e 1 + 1 = 2. The rst horizontal line would then
correspond to 0 1 + 0 2 + + 0 25 of 0 increased by 1, so that it
has the numerical value 1. In other words, Pingala _ makes correspond
to each row, the corresponding dyadic expansion and increases it by 1.
For instance the third row has numerical value 3 and so on and the last
row of the Prastara which is simply
^^^^^^
which corresponds to the binary expansion 1 1+1 2+1 22 + +1 25 =
63, has as its value 63 + 1 = 64. The syllabic pattern for the Gayatr
verse we started with, according to this numbering system has the value
44. The method of constructing a Prastara has been further explained
in the works of Kedara (8th century A.D.) and Trivikrama (12th century
A.D.) and others. To recapitulate, to each horizontal row of a Prastara
consisting of a syllabic pattern one associates the dyadic expansion by
assigning the values 0 and 1 to gurus and laghus respectively and by
multiplying them by powers of 2 depending on their position and add
1 to get the required number. The most remarkable contribution of
Pingala
_ are the Sutra (24 and 25 in Chapter VIII of Pingala's
_ work)
which show how to get back conversely, from any number, a syllabic
pattern in a unique manner so that the procedure just described gives a
one-to-one onto correspondence between numbers and syllabic patterns.
The Sutras in question are the cryptic
24. l ardhe,
25. saike g.
Translated, they read \a laghu if divisible by 2; if not add 1 and a guru".

Let us interpret this remarkably brief Sutra and write down for example
12 R. Sridharan
the syllabic pattern corresponding to 44. Note that 44 is divisible by 2

so that we divide 44 by 2 and we start with a ^, the quotient is 22 which
is still divisible by 2 so that we divide by 2 and add one more laghu and
^^. The quotient is 11 which is not divisible by 2; add 1 divide by
2 and and put a { so that we get ^^ {. Since the new number 6 is
divisible by 2 we add a ^ and get ^^ { ^. The quotient is 3 which is
not divisible by 2; we add 1 divide by 2 and put a { so that the pattern
is ^^ { ^ {. The number 2 being divisible by 2 we put a ^ and get
the pattern ^^ { ^ { ^ and end here. This is indeed the metrical
pattern for the foot of the Gayatr verse we started with.
To summarise, by means of a Prastara, Pingala
_ gave an equivalence
between the binary representations of numbers and their decimal rep-
resentations. I would like to remark that the study of Sanskrit prosody
initiated by Pingala
_ had a continuing tradition in India, the study being
taken up by the Jains too.
We note that in Europe, Leibniz, in the late seventeenth century,
discovered binary expansions of numbers. When he later came across
the Chinese hexagram depictions of Fu Hsi in the \Book of Changes",
he interpreted them as representing binary expansions and hence at-
tributed the discovery of binary representation to the Chinese. One has
to remember that the Indian contributions to mathematics had not yet
reached Europe at that time.
Indian mathematics during the classical age

Aryabhat
.a
Among the mathematicians of the classical age in India, the rst im-

portant name is that of Aryabhat
. a I (not to be confused with Aryabhat .a
II who lived nearly ve centuries later). Not much is known about

Aryabhat
. a (from now on when we talk of Aryabhat
. a we mean Aryabhat .a

I) except that in 499 A.D., he wrote his work Aryabhat . ya, in which he
mentions he was then 23 years old, so that one concludes that he was
born in 476 A.D. It has been, on the basis of tradition, assumed that he
hailed from Kerala, though this is not certain. But, it is certain that he
belongs to the Kusumapura school. It is there that he wrote his famous
work mentioned above. His work is very concise and hard to understand.
Most of it is devoted to Astronomy, though there are 33 slokas devoted
to mathematics. The various subjects dealt with in his work are: the
theorem of Pythagoras, methods of extracting square roots and cube

roots, progressions, geometrical problems, problems involving solutions
of quadratic equations and most importantly, solutions of indeterminate

equations of the rst degree. Aryabhat . a's method of solution of inde-
terminate linear equations was termed by the later mathematicians as
Kut.t.aka (method of pulverization). After the discovery of the Baks.al

manuscript, parts of Aryabhat . a's work, except those dealing with linear
indeterminate equations have lost much of their interest. He was most
probably summarising the state of mathematical knowledge of his time.

One should, however, make a special mention of Aryabhat . a's work on
his approximation for and his tabulation of the values of sine. We

shall rst mention very brie y Aryabhat . a's discussion of the sine.
A
r
O
C
We note that in the arc ACB of a circle with centre O shown in the
diagram, the chord AB represents 2r sin , r denoting the radius of the
. a took r = 360 60=2, where is a suitable approxi-
circle. Aryabhat
mation for , explained in the following sloka:
caturadhikam. satamas.t.agun.am. dvas.as.t.istatha sahasran.am j

ayutadvayavis.kambhasyasanno vr.ttaparin.ahah. jj
We note the crucial word asanna in the sloka which means \approxi-

mately". Aryabhat . a dened approximately to be that number which
makes the circumference of a circle of diameter 20000 units to be 62832
so that is approximately 3.1416. It is also very likely, (looking at the

above sloka), that Aryabhat . a was well aware that is not a rational

number. The sine table of Aryabhat . a consists of the values of r sin
for various angles with r as above. It is amusing to note in passing
the etymology of the word `sine' in this connection. In Sanskrit, the arc
ACB was called Capa or Dhanus. The chord AB was called jya or jva.
With the Arabs jva became jiba and then jaib. Because of the existence
14 R. Sridharan
of a word in Arabic, meaning heart and sounding similar, the Romans

began calling what was originally jya as sinus which means `heart' in
Latin.

We shall now brie y discuss Aryabhat . a's work on linear indetermi-
nate equations. i.e. those of the form ax by = c, where a; b; c are
integers, x and y being unknowns. The Kut.t.aka method of solution of
such an equation is essentially an algorithm to obtain the greatest com-
mon divisor of a and b. This is indeed the most original mathematical

contribution of Aryabhat . a. He was obviously led to consider such equa-
tions through problems in Astronomy - those dealing with planetary

motions and conjunctions. Aryabhat . a is very brief (indeed too brief)
in his work when he describes this method. There are indeed just two

slokas in Aryabhat.ya on Kut.t.aka. These are
adhikagrabhagaharam. chindyadunagra bhagaharen.a
ses.aparasparabhaktam. matigun.amagrantare ks.iptam
adha upari gun.itam antyayugunagra ccheda bhajiteses.am
adhikagra ccheda gun.am. dvi cchedagram adhikagrayutam
These slokas present many diÆculties in translating and have led to a

lot of dierent interpretations. B. Datta (1932) has given a translation
based on the interpretation of the above slokas by the commentator

Bhaskara I (625 A.D.) of Aryabhat . a.
The Kut.t.aka method has been very popular with the later mathe-

maticians of India like Bhaskara I, Mahavra, Aryabhat . a II, Bhaskara
II, and right up to the middle ages. There is in fact a text book from

South India called Kut.t.akara Siroman . i by Devaraja. One also has to
mention that in the solution of second degree indeterminate equations
considered rst by Brahmagupta, Kut.t.aka plays an important role.
The Kut.t.aka method solves the problem of determining an integer
which when divided by two integers a and b leaves remainders r1 and
r2 . A generalization of this problem for several integers a1 ; : : : ; an and

corresponding remainders r1 ; : : : ; rn was also considered by Aryabhat . a.
The Kut.t.aka method applies to this problem too. We note that this
latter is just the classical Chinese Remainder theorem.

Aryabhat . a I was followed by Varahamihira, an astronomer of great
repute who wrote in 505 A.D. his great work \Pa~nca Siddhantika", ex-
pounding the various prevalent astronomical doctrines named Paulisa,

Romaka, Vasis.t.ha, Saura and Brahma. Among these, he singles out
the Saura as the only correct one. Varahamihira also wrote the famous
treatises \Br.ihajjataka" and \Br.hat Sam.hita", dealing with various as-
pects of Astronomy. Incidentally he seems to be the rst astronomer in
India who knew the need for the correction due to the precession of the
equinoxes.
Brahmagupta
Brahmagupta is perhaps the most remarkable mathematician of the

classical age in India. His contributions to mathematics are still remem-
bered and with admiration. He was born in 598 A.D. in Sind (now in
Pakistan) and belongs to the Ujjain School, from where he wrote his
famous work \Brahma Sphut.a Siddhanta", in his 30th year, under the

reign of the Saka King Vyaghramukha. Brahmagupta's idea in writing
his work was to bring up to date an older astronomical work entitled
\Brahma Siddhanta". This great work of Brahmagupta on Astronomy
has more than four chapters devoted to pure mathematics. Among the
commentaries on Brahmagupta's work, one should mention especially,
that of Pr.thudaka Swami (860 A.D.).
It is interesting to note that Indian Astronomy became known to
the Arab world only through Brahma Sphut.a Siddhanta. Khalif Ab-
basid Al Mansoor (712{775 A.D) founded the city of Baghdad (on the
bank of Tigris) and made it a centre of learning. A scholar from Ujjain
was invited at his initiative in 770 A.D. to explain Indian Astronomy.
Through the Khalif's orders Brahmagupta's work was translated into
Arabic by Al Fazari and named \Sind Hind" or \Hind Sind". It is thus
that the Arab world became cognizant of Indian Astronomy.
We shall pick two brilliant contributions of Brahmagupta, one in
Algebra and the other in Geometry which clearly reveal the stature and
sophistication of this rst rate mathematician.
(a) Varga Prakr.ti (Solution of quadratic indeterminate equations)
It must have been well known to all ancient civilizations of the world
that the product of the sum of two squares is again a sum of two squares;
this comes from the identity
(x2 + y2 )(z 2 + t2 ) = (xz yt)2 + (xt yz )2 :

16 R. Sridharan
Brahmagupta stated for the rst time a more general version of the
above identity called Bhavana in his work cited above. It asserts that
for any arbitrary (integer) N , we have
(x2 Ny2 )(z 2 Nt2 ) = (xz Nyt)2 N (xt yz )2 : ()
Bhavana is an example of the notion of \composition" of quadratic forms

invented by Gauss several centuries later. (The above identity () can be
interpreted as saying that binary quadratic forms of the type x2 Ny2
admit \composition". The theory of composition of quadratic forms is
still an active area of research in mathematics).
Brahmagupta used Bhavana as a tool to study equations of the form
x2 Ny2 = k for integers x; y for any given positive integer N and
a given integer k. Such equations are called Vargaprakr.ti (square na-
tured). That Brahmagupta considered quadratic indeterminate equa-
tions is itself remarkable, in view of the fact that in Astronomy (as we

noticed when we were discussing Aryabhat . a's work) one comes across
only linear indeterminate equations and there is a priori no reason to
consider Vargaprakr.ti, unless out of sheer mathematical curiosity. This
clearly shows the working of the mind of a pure mathematician, who
has the penchant for picking up signicant problems. Brahmagupta
showed, using Bhavana, that if there is one solution (; ) of the equa-
tion x2 Ny2 = 1, then one has an innity of solutions; indeed, applying
() one sees that (2 + N 2 ; 2 ) is also a solution and one could iter-
ate this procedure to get an innity of solutions. But Brahmagupta did
more. He proved using () that if one of x2 Ny2 = 1, x2 Ny2 = 2
or x2 Ny2 = 4 has a solution in integers, then x2 Ny2 = 1 has a
solution also in integers. Brahmagupta, by this method solves the equa-
tion x2 92y2 = 1 (x = 1151, y = 120 is a solution) and the equation
x2 83y2 = 1 (x = 82, y = 9 is a solution). He however was unable to
solve the equation x2 Ny2 = 1 in general.
The credit for giving an algorithm to solve the above equation in
general goes to an unknown Indian mathematician who lived after Brah-
magupta but before Bhaskaracarya II (whose contributions we shall dis-
cuss subsequently). Bhaskaracarya II, in his work \Bjagan.ita" (circa
1150 A.D.) describes an algorithm to solve this equation and says \people
call this method Cakravala" (the \cyclic method"). Using the Cakravala
method, he solves the rather diÆcult equations x2 61y2 = 1 and
x2 67y2 = 1 (x = 1766319049; y = 226153980 is the least positive

integral solution of the rst equation and x = 48842, y = 5967 is the
least positive integral solution of the second).
In 1954, K.S. Shukla of Lucknow University discovered (cf. [15]) in
Maharaja's Palace library in Thiruvananthapuram a commentary called
\Sundar" by a certain Udaya Divakara on the work of Bhaskara I,
called \Laghu Bhaskarya", where the same Cakravala method is de-
scribed with reference to a certain Jayadeva. The commentary itself
dates back to 1073 A.D., so that the Cakravala method must be ante-
rior to Bhaskara II, at least by a hundred years. The Cakravala method
is a beautiful algorithm which reminds one of Fermat's methods. Curi-
ously enough, Fermat (1601-1665) set up as a challenge to some of his
contemporaries like Frenicle, Brouncker and Wallis the problem of nd-
ing of an integral solution of the equation x2 61y2 = 1 (They indeed
found a solution). Thus began, in the 17th century in Europe the study
of such equations which is the genesis of modern Arithmetic Geometry.
A. Weil (cf. [21]) has this to say in this connection: \What would have
been Fermat's astonishment if some missionary just back from India
had told him that his problem had been successfully tackled there by
the native Indians almost six centuries earlier?". One must add that
Euler who looked at this problem of Fermat, wrongly attributed it to a
certain British mathematician named Pell (who had nothing to do with
this equation) and called it Pell's equation. His name has since stuck
to the equation x2 Ny2 = 1 through the mistake of Euler. It should
really be called \Brahmagupta equation".
As an example of the prejudice of some of the mathematicians of the
West, who would insist on depriving India of the honour of discovering
mathematical ideas and try to prove that all mathematics originated
in Greece, I quote the reasons of (a very ne mathematician) Van der
Waerden to dismiss the claim that Brahmagupta was indeed the rst one
to study the so called Pell's equation and attributing it to the Greeks.
\1) In Brahmagupta's treatise, the problem of solving Pell's equation
is not motivated at all.
2) The main subject of Brahmagupta's treatise is Astronomy. His
Astronomical system is based upon the epicycle hypothesis which is
a Greek invention. As one of his tools, Brahmagupta uses a table of
sines. The Greeks have tables of chords which can be easily trans-
18 R. Sridharan
formed into a table of sines. Some eighty years before Brahmagupta,

Varahamihira presents excerpts from ve Siddhantas one of which, the
Romaka Siddhanta, was based on the solar and lunar theory of Hip-
parchus while another Siddhanta is ascribed to \Paulisa, the Greek".
Hence the Astronomy and the mathematical tools of the Hindus at
the time of Brahmagupta and before, were largely derived from Greek
sources. This makes it even more probable that the Hindu method of
solving Pell's equation also goes back to Greek sources.
Let us keep in mind that in history of science, independent inventions
are exceptions: the general rule is dependence" (cf. [20]).
(b) Brahmagupta's work on rational cyclic quadrilaterals
I now discuss a typical contribution of Brahmagupta to geometry. As

we have already remarked, Sulvak aras were familiar with the construc-
tion of right angled triangles with rational sides. It is indeed remarkable
that several centuries later, Brahmagupta was still interested in the
problem of construction of gures with rational magnitudes, more pre-
cisely, in the question of constructing cyclic quadrilaterals with rational
sides and diagonals.
We begin with the following remark: Suppose that we have two right
angled triangles ADB and A0 D0 C 0 with rational sides. We consider
another triangle ADC which is similar to A0 D0 C 0 . By juxtaposing the
triangles ADB and ADC along the side AD, we get a triangle ABC with
rational sides and rational area. Conversely, any triangle with rational
sides and rational area can be obtained by juxtaposing two rational right
angled triangles. Indeed, one has to drop the perpendicular from one
of the vertices to the opposite side and consider the two right angled
triangles thus formed.
A0
A A
B D B D C D0 C0
This result is attributed to Euler by (cf. [7]) (though no proof is found in

Euler's published papers). However this must certainly have been obvi-
ous to Brahmagupta, as it is indeed a particular case of Brahmagupta's
result on quadrilaterals by looking at a triangle as a degenerate quadri-
lateral. What Brahmagupta did was much more. He constructed cyclic
quadrilaterals with rational sides and diagonals by juxtaposing two right
angled triangles in either of the following two ways.
A A
B C
B D C
E E
Figure 1 Figure 2
The rst gure corresponds to juxtaposing the rational right angled

triangle ADB and ADC along AD and producing AD to meet the
circumcircle of ABC at E . The quadrilateral ABCE is seen to be one
with all sides, the intercepts on the diagonals and hence the diagonals all
rational. The other gure corresponds to juxtaposing two rational right
angled triangles along their hypotenuses. Various numerical examples
were considered by Brahmagupta.
The work of Brahmagupta reached Europe through the translation
by Colebrooke in the 19th century. The famous French geometer Chasles
was very much impressed by this work and he published an appreciation
of this. (cf. [2]). The German mathematician Kummer (cf. [10]) in 1846
considers the general problem for all quadrilaterals, giving a complete
solution for the cyclic case. Kummer modies Brahmagupta's method
of construction and shows that one can obtain all cyclic quadrilaterals
whose sides, diagonals and area are rational, by employing a juxtapo-
sition of three rational right angled triangles. (It is rather amusing to
20 R. Sridharan
note that he blames Brahmagupta for not having thought of this so-
lution!) He relates the solution in the possibly non-cyclic case to the
existence of rational points of certain elliptic curves.
The questions studied by Brahmagupta on quadrilaterals inspired
further study and elaborations by successive generations of Indian math-
ematicians like Mahavra, Bhaskara II and the interest in these problems
continued even in the middle ages. For instance, Gan.esa (c 1545 A.D.)
has results on rational quadrilaterals.
The next few names of mathematicians of note in the classical age
in India are those of Sr dhara (c 750 A.D.) who composed a work on
arithmetic called Pat.gan.ita and a work on algebra. The latter is no
longer extant. We know about the contributions of Srdhara from the
later work of Bhaskara II, who also mentions the name of another al-
gebraist Padmanabha. One of the most celebrated mathematicians af-
ter Brahmagupta was Mahavra (C 815 A.D.). He was in the court of
the Ras.t.rakut.a king Amoghavars.a Nr.patunga
_ (who ruled over a part
of the present state of Karnat.aka). Mahavra wrote a work entitled
\Gan.itasarasam.graha" (C 850 A.D.) which is really the rst text book
on arithmetic whose material is found even in the present text books in
South India. Mahavra mainly improved on the work of his predecessors
and probably did not make any profound contributions himself.
Bh
askarac
arya II
The next celebrated and essentially the last mathematician of the

classical period is Bhaskara II who was called Bhaskaracarya in view
of the veneration that people had for his learning. Bhaskara II be-
longs (according to his own statement in his work \Siddhantasiroman.i"
written around 1150 A.D.) to Bijjad.a Bd.a which has been identied
with modern Bijapur. His work Siddhantasiroman.i is divided into
four parts: Llavat, Bjagan.ita (Algebra), Grahagan.ita (Astronomy)
and Goladhaya (study of the celestial sphere). Bhaskara prided himself
as a poet and wrote all his mathematics in verse form. (To my mind,
verse form and mathematics do not go well with each other).
As we already said, Siddhantasiroman.i is essentially a text book, a
consolidation of the state of mathematics at the time of Bhaskara II,
but his work was systematic and lucid so that Bhaskara was very much
arya".
respected and termed an \Ac
The part of Siddhantasiroman.i, called Llavat deals with various

aspects of arithmetic, whereas Bjagan.ita deals with algebraic, theoret-
ical questions. Llavat has several problems in arithmetic, geometry,
quadratic equations, permutations and combinations and Kut.t.aka. As
a typical example, we mention a problem on permutations related to
Chandas from Llavat:
Prastare mitra gayatryah. syuh. pade vyaktayah. kati j

Ekadi guravascasu kati katyucyatam pr.thak? jj
[The metre Gayatr, (as noted before) has six syllables. The number
of possible times the long syllable occurs in any pada in a metre with
six syllables is six, the number of times it occurs twice is fteen and so
on, so that the sum total of all its occurrences is 64 (including the case
when it does not occur at all. For the Gayatr metre with four padas,
the number of all such possibilities is 644 = 16777216)].
There is a tradition that Llavat was the name of Bhaskara's daugh-
ter who became a widow at an early age and that Bhaskara wrote his
work Llavat to provide consolation for his daughter. It is not clear that
this fanciful story has any basis.
In Bhaskara's Bjagan.ita, there are problems which reduce to solving
quadratic equations. Occasionally, there is an example of a problem
which reduces to an easily solved cubic, or even a biquadratic.
Bhaskara was familiar with the notion of integration as a limit of
nite sums. For instance, he calculates the surface area of the sphere,
by dividing it into annuli and drawing a system of parallel circles, or
by dividing the surface into lunes by drawing meridian circles through a
pair antipodal points on the sphere and dissecting each lune into a large
number of small quadrilaterals by drawing circles parallel to the equator.
Bhaskara also had ideas about dierentiation in connection with the so
called instantaneous method (tatkalika) by dividing the day into a large
number of small intervals and comparing the positions of the planet
at the end of successive intervals. The tatkalika gati is essentially the
instantaneous velocity of the planet. Bhaskara also knew the addition
formula for the sine and special values of the sine function. Bhaskara's
innitesimal approach probably found its fruition in the mathematics of
Kerala region in Medieval India which we now proceed to discuss.
22 R. Sridharan
Mathematics in Medieval India (cf. [1,12,13,17]

It was believed till nearly a hundred and fty years ago that creative
contributions to mathematics form India came to a halt after Bhaskara
II. However, in 1835, Charles M. Whish, a civilian employee of the East
India Company sprang a surprise, when he published in the \Trans-
actions of the Royal Asiatic Society of Great Britain and Ireland", a
paper entitled \On the Hindu quadrature of the circle and the innite
series for the proportion of the circumference to the diameter exhibited
astras: Tantrasangraham,
in four S _ Yuktibhas.a, Karan.apaddhati and
Sadratnamala". Whish gave an account of the contents of these works
and assigned possible dates for these texts, some which have since been
conrmed. Since the middle of the 20th century, several Indian scholars
have worked on the contributions of the Kerala school during the middle
ages and it is now obvious that, at least in Kerala, Indian mathemat-
ics did not end with Bhaskara and in fact continued right through the
middle ages.
The great pioneer of the Kerala school was Madhava of Sangama- _
grama who lived possibly during 1340 A.D. - 1425 A.D. He was usually
referred to as a Golavid, an expert on the celestial sphere. Though he did
not publish much, (he has works like Ven.varoha and Sput.achandrapti,
both giving rules of thumb for xing lunar positions and some other
works like Agan.ita, Laghuprakaran.a), Madhava, however seems to have
been a versatile genius and a source of inspiration to his students. We
shall talk about Madhava in connection with a specic problem later.
Next in line comes Paramesvara who was a prolic writer on Astron-
omy and is especially remembered as the author of Dr.ggan.ita system of
Astronomy. Based on his substantial observations spread over fty-ve

years, he made some signicant corrections to Aryabhat . a's astronomical
calculations.
The greatest personality of the Kerala school was Nlakan.t.ha So-
mayaji, who was a student of Damodara (son of Paramesvara and like
his father an Astronomer). Nlakan.t.ha Somayaji lived sometime during
1440 and 1550 A.D., wrote a rst rate commentary on Aryabhat .ya,
appropriately called Bhas.ya (reminiscent of the great Mahabhas.ya of
Pata~njali on the As.t.adhyay of Pan.ini!). This work must have been
composed late in his life since he quotes there from his other works in-
cluding Tantrasangraham.
_ It may be interesting to note that among his
works there is one which is called Sundararaja Prasnottara which records

Nlakan.t.ha's answers to specic questions of a certain Sundararaja, a
contemporary scholar from the Tamil country. The respect Sundararaja
had for Nlakan.t.ha bordered on veneration.
Jyes.t.hadeva was another student of Damodara and he authored Yuk-
tibhas.a, a scientic text written in Malayalam modelled on Nlakan.t.ha's
Tantrasangraham.
_ Yuktibhas.a is indeed unique in the sense that it is
the rst text in mathematics in India to state precise mathematical
propositions along with their proofs. This was composed around 1520
A.D. Karan.apaddhati was written by Putumana Somayaji around 1730
A.D. This is a work of ten chapters and verses and summarises all the
mathematical contributions of the Kerala school in the sixth chapter,
often quoting from Madhava, Nlakan.t.ha and others.
Sadratnamala belongs to the period after 1800, its author being
Sankara
_ Varma, a ruler of the local principality Kad.athanad. in north
Kerala. This work is merely an exposition of the achievements of the
Kerala mathematicians.
Though most of the above texts deal with Astronomy or with some
special topics, they always contain sections on mathematics; for example
Yuktibhas.a contains usual mathematical material on arithmetic, frac-
tions, Kut.t.aka, a preliminary section with a proof of the Pythagorean
theorem, approximation of the circumference of the circle by the perime-
ters of inscribed regular polygons and nally and most importantly an
innite series for =4, which is customarily attributed to Gregory and
Leibniz. It also contains several results on convergence of series, series
for the sine and the cosine and a correct computation of the volume of
the sphere, by the method of integration. It has also problems on the
geometry of quadrilaterals.

During our discussion of the work of Aryabhat . a, we mentioned his
approximation for as 3.1416 and his \qualication" that the latter is in-

deed an approximation. Nlakan.t.ha raises in his Bhas.ya of Aryabhat
.ya
the question \why this qualication?", \why not the precise value?" and
he answers \because it is impossible to obtain a precise one", asserting
essentially that is irrational. He also refers to the great Madhava
and says that Madhava had even a better approximation to namely
3.1415926536 (this approximation is quoted in Karan.apaddhati IV, 7).
In other words, the mathematicians of Kerala already had a good no-
24 R. Sridharan
tion of irrational numbers. Madhava also seems to have had the idea to
look for an \innite series" for which would eventually lead to ratio-
nal approximations. Proceeding in this manner, the Kerala school did
produce power series expansions for tan 1 x, sin x, cos x etc. Of course,
they did not formally discuss the convergence of these series, but had
a clear idea about convergence and in fact constructed more and more
rapidly convergent series, and broke innite series at a nite stage to
estimate the error term. Let us remember, all these were achieved at
least two centuries before Europe ever came to consider such questions.
We conclude our account of the contributions of the Kerala school,
by detailing just one example, which shows the level of sophistication
and depth reached by them in analysis. Yuktibhas.a, discussed earlier
contains the following verse which is attributed to Madhava
Vyase varidhinihate rupahr.te vyasasagarabhihate j

Trisaradivis.amasam.khya bhaktamr.n.am. svam pr.thak kramat kuryat jj
which is merely a re-rendering of a verse in Karan.apaddhati. The con-
tent of the verse above is to relate the circumference c of a circle of
diameter d in terms of d by the series
4d 4d
c = 4d
3
+
5
;
or equivalently 4 = 1 13 + 51 + , which is the so called Gregory-
Leibniz series. We shall present the proof of the above equality due to
the Kerala school.
Consider a quarter of a circle of unit radius with centre O inscribed
in a unit square as in the following gure:
O P 0
P 1
A
Pr 1
B0 B
C Pr
Pn
Partition P0 Pn into n equal parts. Let OPr 1 ; OPr meet the circle at A
and B . We draw AB 0 and Pr 1 C as perpendiculars to OPr from A and
Pr 1 respectively.
From the similarity of the triangles OAB 0 and OPr 1 C , we get, in
particular
AB 0 Pr 1 C
= :
OA OPr 1
From the similarity of the triangles CPr 1 Pr and OP0 Pr , we obtain
Pr 1 C OP0
= ;
Pr 1 Pr OPr
so that we get
OA OP0 Pr 1 Pr Pr 1 Pr
AB 0 = = :
OPr 1 OPr OPr 1 OPr
We note that if n is suÆciently large, the segment AB 0 tends to arc
AB and OPr 1 and OPr are approximately equal, so that we have the
approximation
1=n 1=n
arc AB = :
OPr 1 1 + ( r n 1 )2
2
We therefore have, by using the geometric series 1

1+x
= 1 x + x2 ,
circumference P
= =4 = limn!1 nr=1 1+(1r=n1 )2
8 n
!
1 nX1 1 X
n 1
1 X
n 1
= lim
n!1 n r=1
1
n3 r=1
r +
2
n5 r=1
r 4
:
Remarkably, the Kerala
P mathematicians knew that for any integer p 0,
n 1 p
one has that np+1 r=1 r tends to p+1
1 1
as n ! 1. Substituting this in
the equation above and pushing the limit inside (which is correct, but
needs justication), we nd the innite series expansion for =4 namely
1 1
4
=1
+
3 5
:
In fact, the same method also exhibits the series
x3 x5
tan 1
x=x
3
+
5
:
26 R. Sridharan
P
That np1+1 nr=11 rp tends to p+1
1
as n ! 1 was proved in Europe by
Roberval in 1634, nearly a century later than the discoveries of the
Kerala mathematicians.
As we mentioned earlier, the story does not end here. The Kerala
school had several concrete innite series, they also could manipulate
to obtain rapidly convergent series and they even had vague notions of
integration and dierentiation.
The achievements of the Kerala school may in part be explained
by the leadership of that brilliant mind Madhava. The reason why
the discoveries of the Kerala school did not reach the West is perhaps
because the Portuguese dominated the West Coast of India and were
not interested in the contributions of the natives. It indeed needed the
British functionary Whish to bring to the knowledge of the rest of the
world these achievements and it further took nearly a century for the
Indian scholars to take cognition of this great and yet neglected chapter
in the history of mathematics.
Conclusion
To sum up, the contributions of India to mathematics during the

ancient, classical and medieval periods have been very noteworthy and
many times profound. However, it has taken quite a while for the world
to understand and appreciate their depth. There could be several rea-
sons for this. It can in part be attributed to the overwhelming awe
the rest of the world had for India's contributions to religion and tran-
scendental philosophy, vis-a-vis the other contributions of India! Yet
another reason could be the method, peculiar to ancient India, to hand
down knowledge to posterity. Due, perhaps, to the paucity of writ-
ing material, important ideas were preserved in the cryptic \Sutras"
which could not be deciphered unless through extensive commentaries
by certain distinguished teachers belonging to later traditional schools
of learning based on the gurukula system. This sometimes had the eect
of hiding the real meaning of the Sutras from the uninitiated and even
if the Sutras did make sense, their full meaning was not always appar-
ent. This had also the unfortunate eect of obscuring the original date
of discovery of a particular Indian contribution, there being a possible
ambiguity between the date of the actual discovery and the date of a
later commentary on it.
I would like to end this article with a couple of general remarks.

The Indian habit of using the verse form to state mathematical prob-
lems and propositions has not been of much help in the propagation of
mathematics; very often versication could be more of an impediment
than help. One might also add that unlike in Greece, where mathemat-
ics was held in high esteem for its own sake, in India, scientic thought
was generally held subservient to tradition and was not cultivated for
its own sake. This, in my opinion, is indeed a weakness in the approach
in ancient and medieval India to science, in general, and mathematics,
in particular.
Acknowledgements:
I thank the Nehru Centre, in particular, Mr. Satish Sahney, for

inviting me to give this lecture in the series \Science in India: Past and
Present". I stand in great intellectual debt to C. N. Srinivasaiengar,
from whose beautiful little book [17], I have borrowed liberally.
I take this opportunity to put on record my indebtedness to Sripathy
for all the help I have received from him. I thank Vijayalakshmi for
achieving the near impossible feat of deciphering my manuscript and
bringing it to the present form. I have pleasure in expressing my warm
thanks to V. Balaji, K.R. Nagarajan, Shiva Shankar and M.D. Srinivas
for all the support that they have given me on every occasion.
I dedicate this little piece of labour (whatever be its worth) to two
of my esteemed colleagues, for whom I have admiration: to Seshadri for
his undaunted eorts for the cause of mathematics, especially for the
Chennai Mathematical Institute and his ever youthful enthusiasm (in
spite of the fact of being coeval with me), to Parimala whose dedication
and devotion to mathematics are absolute and can be matched only by
her tenacity, forthrightness and fortitude.
References
[1] K. Balagangadharan: Mathematical Analysis in Medieval Kerala,
Science in the West and India, Some Historical Aspects, Edited by
B.V. Subbarayappa and N. Mukunda, Himalaya Publishing House,
1995, p.290{312.
28 R. Sridharan
[2] M. Chasles: Apercu historique sur l'origine et development des

methodes en geometrie, Paris 1875, p.436
[3] H.T. Colebrooke: Algebra, Arithmetic and Mensuration from the
Sanskrit of Brahmagupta and Bhaskara, London, 1817.
[4] B. Datta: Early literary evidence of the use of the zero in India,
Amer. Math. Monthly, Vol. 33, 1926, p.449-454, p.556-572.

[5] B. Datta: The Science of the Sulva, A Study in Early Hindu Ge-
ometry, University of Calcutta, 1932.
[6] B. Datta and A.N. Singh: History of Hindu Mathematics, A Source
book, I & II, Asia Publishing House, 1962.
[7] L.E. Dickson: Rational triangles and quadrilaterals, Amer. Math.
Monthly 28, 1901, p. 244{250.

[8] J. Eggeling: The Satapatha Brahman.a, Part IV, Vol 43, \Sacred
Books of the East".
[9] G.G. Joseph: The Crest of the Peacock, Non European Roots of
Mathematics, AÆliated East-West Press, 1995 (2nd edition, Pen-
guin 2000).
[10] E.E. Kummer: Uber die Vierecke, deren Seiten und Diagonalen
rational sind, Crelle, Vol. 37, 1847, p.1{20.
[11] N.N. Namboodri: Binary Number System: The Mathematical Base
of Indian Prosody, Journal of the Institute of the Asian Studies, 12,
No.2, 1986, p.77{100.
[12] M. Ojanguren: Dyadische Metrik und Kreisumfang: Zwei Beispiele
alter indischer Mathematik, Preprint, Lausanne, 2000.
[13] T.A. Saraswathi Amma: Geometry in ancient and Medieval India,
Motilal Banarsidass Publishers, 1979.
[14] C.O. Selenius: Rationale of the Cakravala process of Jayadeva and
Bhaskara II, Historia Mathematica, 2, 1975, p.168-183.
arya Jayadeva, the Mathematician, Ganita, 5, 1954,
[15] K.S. Shukla: Ac
p.1{20.
[16] R. Sridharan: Ancient Indian Contributions to Quadratic Algebra:

Sulva to Cakravala, Science in the West and India, Some Historical
Aspects, Edited by B.V. Subbarayappa and N.Mukunda, Himalaya
Publishing House, 1995, p.280{289.
[17] C.N. Srinivasiengar: The History of Ancient Indian Mathematics,
World Press, 1988
[18] A. Van Nooten: Binary Numbers in Indian Antiquity, Journal of
Indian Philosophy, 1993, p.31{50

[19] A. Weber: Uber die Metrik der Inder, Harrwitz and Gofsmann,
Berlin, 1863.
[20] B.L. Van der Waerden: Pell's Equation in Greek and Hindu Math-
ematics, Russian Math Surveys 31:5 (1976), 210{225 from Uspeki
Math. Nauk 31:5 (1976) 57{70).
[21] A. Weil: Number Theory, An approach through History, Birka-
hauser, 1982.
Sanskrit Prosody, Piṅgala Sūtras and
Binary Arithmetic
R. Sridharan
Chennai Mathematical Institute, Chennai 600 017
rsridhar@cmi.ac.in
Abstract
In India, the science of prosody, had its beginnings in the Vedic

period and has been held in high esteem, being regarded as a
vedāṅga or a limb of the Veda. The earliest work on prosody was
by Piṅgala which is generic and on which all the subsequent works
are based. It is an amazing fact that this early work already deals
with matters relating to problems of combinatorics. The main aim
of this paper is to give a brief description of this work of Piṅgala
and discuss in detail the mathematics arising out of it.
§1. Introduction
Classical Sanskrit composition is of two kinds: gadya (prose) and padya
(poetry). Sanskrit prosody, the study of the metrical scanning of San-
skrit poetry, called chandas śāstra, has its beginnings already in the
vedic times. In his classic work, Vedic Metre in its Historical develop-
ment, first published in 1905, E. V. Arnold ([3]) begins with high praise
for the R. gveda by remarking that The Rigveda is not a book, but a
library and a literature. At the end of the first chapter, he adds: To
whatever conclusions we may be further led in detail, it must be plain
that as works of mechanical art, the metres of Rigveda stand high above
those of modern Europe in variety of motive and in flexibility of form.
They seem indeed to bear the same relation to them as the rich har-
monies of classical music to the simple melodies of the peasant. And in
proportion as modern students come to appreciate the skill displayed by
the Vedic poets, they will be glad to abandon the easy but untenable
theory that the variety of form employed by them is due to chance, or
33
34 R. Sridharan
the purely personal bias of individuals; and to recognize instead that we

find all the signs of a genuine historical development, that is of united
efforts in which a whole society of men have taken part, creating an
inheritance which has passed through generations from father to son,
and holding up an ideal which has led in turn to seek rather to enrich
his successors than to grasp at his own immediate enjoyment. If this
was so, then the vedic bards are also to be counted amongst ‘great men
and.....such as sought out musical tunes and set forth verses in writing’.
To quote another great British Indologist on the Indian contributions
to metres, here is what H. D. Colebrooke has to say in his essay entitled
On Sanscrit and Prakrit poetry, published in “Asiatic Researches” in
1808:xxx (and reprinted in Essays on History, Literature and Religion
of Ancient India ([9])): The prosody of Sanskrit is found to be richer
than that of any other known language, in the variations of the metre,
regulated, either by quantity, or by the number of syllables both with
or without rhyme and subject to laws imposing in some instances rigid
restrictions, in others, allowing ample latitude.
A few words about the arrangement of the article: Since a general
reader, (for instance a mathematician) to whom this article is addressed,
may not be familiar with principles of prosody in general and Sanskrit
prosody in particular, this article begins with a rather discursive account
of some aspects of vedic prosody and also introduces some basic facts on
the rules of Sanskrit prosody in the first few preliminary sections. The
main theme of the paper which is the discussion of the classical work
of Piṅgala on prosody is taken up, beginning §5. The crucial section is
§8 , from the point of view of a mathematician, which deals with the
combinatorial aspects of Piṅgala’s work.
§2. The beginnings of prosody

As we said earlier, the beginnings of Sanskrit prosody go back to the
vedic times. The Brāhman.as speak eloquently of the origins of metres,
colouring them with mysticism. The study of prosody has been held,
right from the early times, with the greatest of esteem; At the end of
section 7 of prapāt. haka 1 of the Nidāna Sūtra ([16]), we have for instance
the following stanza:
chandasām
. vicayam
. jānan yah. śarı̄rādvimucyate
Piṅgala and Binary Arithmetic 35
chandasāmeti sālokyamānantyāyāśnute śriyam
CdsA\ Evcy\ jAnn^ y, frFrAEmQyt.

CdsAmEt sAloÈmAn(yAyAt Eym^;
Translated roughly into English the above stanza reads:

One who a has deep knowledge of Chandas, shares, after liberation
from his body, the same abode of the Chandas, acquiring eternity, glory
and beatitude
Prosody has been described as the feet of the Vedas; it is thus one
of the limbs of the Veda - a vedāṅga. In Pān.inı̄ya śiks.ā, we have the
following verse:
Chandah. pādantu vedasya hastau kalpo’tha pat.hyate

Jyotis.āmayanam
. caks.urniruktam
. śrotramucyate
Cd, pAdt vd-y h-tO kSpo_T pÕt.

>yoEtqAmyn\ c"En zÄ\ o/mQyt ;
Chandas are the feet of the Vedas, Kalpa the hands, Astronomy the
eyes and Nirukta the ears.
The importance of the knowledge of chandas for understanding the
Vedas is emphasised in Br.haddevata (a text dealing with the gods of
the R. gveda, supposedly written by that ancient venerable vedic seer
Śaunaka, verse 136, VIII, ([5]) as follows:
aviditvā r..sim
. chando daivatam . yogameva ca
yo’dhyāpayejjapedvāpi pāpı̄yāñjāyate tu sah.
aEvEd(vA Eq\ Cdo {dvt\ yogm

v c .
yo_@yApy>jpAEp pApFyAÒAyt t s, .
One who teaches or recites the Veda without having proper knowl-
edge of the applications, the seers, metres and Gods, becomes indeed a
sinner.
36 R. Sridharan
As one of the earliest references to prosody, we have a verse (I.1.5) in

the Mun.d.akopanis.ad ([10]), which lists chandas as essential for attaining
the “lower knowledge”, the higher one being that of the Brahman.
tatrāparā r.gvedo yajurvedo sāmavedo atharvavedah.

śiks.ā kalpo vyākaran.am
. niruktam
. chando jyotis.amiti
t/AprA `vdo yjv

do sAmvdo aTv v
d, .
Ef"A kSpo &yAkrZ\ EnzÄ\ Cdo >yoEtqEmEt ;
The six vedāṅgas which were considered essential for the understand-
ing of the Vedas are as mentioned above: śiks.ā, phonetics; kalpa, the
knowledge of the sacrificial rites; vyākaran.a, grammar; nirukta, etymol-
ogy; chandas, prosody; and jyotis.a, astronomy.
These vedāṅgas, whose beginnings can be traced already to the
Brāhman.as and the Āran.yakas, did not originally refer to independent
branches of knowledge but were only indicated as fields of study, essen-
tial for the understanding of the Vedas. As time went by, it was realised
that there was a real need to develop them as auxiliary subjects associ-
ated with the study of the Vedas. Hence, independent texts were written
in the (mnemonic) sūtra style to expound these subjects. ([28]) . It is
perhaps worthwhile to mention, by the way, that the sūtra style of writ-
ing is something unique to Indian literature. A very succinct definition
of a sūtra is found ([14]) in the Vis.n.udharmottara purān.a and runs thus:
alpāks.aram asandigdham
. sāravat viśvatomukham
astobham anavadyam . ca sūtram
. sūtravido viduh.
aSpA"r\ asEd`D\ sArvt^ EvtomK\.

a-toBm^ anv\ c s/\ s/Evdo Evd, ;
A sūtra should have the least number of syllables, should contain
no doubtful words, no redundancy of words, should have unrestricted
validity, should contain no meaningless words and should be faultless!
The study of prosody which began in the vedic period evolved to
apply to classical Sanskrit and to the Prakrit poetry as well and had its
continued impact on the poetry of the later ages too.
To quote an example, the ślokas of the epics like the Rāmāyan. a and
the Mahābhārata are derived by and large from the vedic metre anus..tup.
Indeed the vedic metre anus..tup came to be monopolised by the poets
of the classical age. On the other hand, the vedic metres tris..tup and
jagatı̄ led to metres used by poets and bards at the courts of various
kings.
In this connection, it is amusing to note, parenthetically, that the
great poet Kālidāsa (who himself is probably the author of a work enti-
tled Śrutabodha [21] on classical metres - though this work is attributed
by some to Vararuci) employs a vedic metre in a very appropriate con-
text in his great play Abhij ñāna Śākuntalam.
When Śakuntalā is about to leave the hermitage of the sage Kan.va
to go to meet her husband Dus.yanta, Kan.va offers a benediction, which
is set in the following beautiful stanza in the vedic metre tris..tup, with
11 syllables in each of its four pādas.
amı̄ vedı̄m paritah. klr.ptadhis.n.yāh.

samidvantah. prānta sam . stı̄rn.a darbhāh.
apaghnanto duritam . havyagandhaih.
vaitānāstvām
. vahnayah . pāvayantu
Translated in to English, the stanza reads:

May these sacrificial fires, fixed in their places around the altar,
nourished by holy wood, with the darbhā grass strewn around their
boundaries, removing sin by the fragrance of the oblations, purify thee!
amFv
Ed\ pErt, kØED yA,
sEmt, þAt sM-tFZ dBA , .
apÍto dErt\ h&ygD{,
v{tAnA-vA\ v¡y, pAvyt ;
38 R. Sridharan
§3. Units of prosody, the syllables

The etymology of the Sanskrit word for prosody chandas traces it to
various roots, for instance, it can de derived from the root chad, which
means “to cover”; incidentally, this is not the only possible etymological
derivation; there are several other possibilities too! Whatever be the
etymology of chandas and the consequent derived meanings, it denotes
the science of syllables in verses.
A syllable (aks.ara in Sanskrit) is a vowel with or without one or
more consonants. A syllable is called a laghu (short), (denoted by l), if
it consists of a short vowel followed by at most one consonant. A syllable
which is not a laghu is called a guru (long), (denoted by g). But there
is a proviso by which even a short syllable will be treated as long while
scanning, when it is followed by a conjunct consonant, an anusvāra (a
nasal) or a visarga (an aspirant). Unlike in classical Sanskrit prosody,
where the nature of the syllables is also an important aspect of prosody,
vedic metre is governed solely by the number of syllables in a verse,
called the length of the metre. (a verse is called a pāda in Sanskrit),
which forms the basic unit of Sanskrit poetry. Verses combine to form
a r.k, or a stanza, which is a unit of a vedic hymn.
A stanza consists, generally, of not less than three and not more
than fifteen verses. A stanza may consist of metrically identical (sama)
or metrically different (vis.ama) verses. Two or three stanzas combine
to form a strophe.
The following is an example of a .rk in gāyatrı̄ metre (a stanza with
three verses each of which has 8 syllables):
agnim ı̄le | puróhitam |

yajñásya de | vam .rtvijam |
hótāram ra | tnadhā
´tamam |
which has the following arrangement of 8 syllables in each of its verses:

g l g g l g l l
g g l g l g l l
g g g g l g l l
(According to one of the rules of prosody, the first and the last syllables
of a verse are ignored for scanning purposes.)

§4. Some works on prosody other than Piṅgala’s
As we shall notice presently, Piṅgala wrote a definitive work (in sūtra
style) on prosody, probably around the middle of the third century B.C.
As is the case with such definitive works, (for example the As..tādhyāyı̄
of Pān.ini), Piṅgala’s work systematises and improves upon the work
of many earlier authors on the subject. The names of Yāska and the
otherwise unknown prosodists like Saitava, Rāta, Mān.d.avya, Tān.d.ı̄,
Kraus.t.iki and Kāśyapa are mentioned as some of those who preceded
him. Like Pān.ini once again, who dealt with classical Sanskrit gram-
mar rather than vedic grammar, Piṅgala, though he begins his work
with vedic metres, deals for a substantial part with classical metres.
It should be remarked that works like the Chando viciti (called Tatva
subodhinı̄), which is a part of the Nidāna Sūtra (which is a śrauta
sūtra of the Sāmaveda, and is supposed, according to some, to have
been written by the great Patañjali, who wrote the Mahābhās.ya - the
“great commentary” - on Pān.ini’s As..tādhyāyı̄), R . kprātiśākhya (written
by the venerable Śaunaka), Sāṅkhyāyana Brāhman.a, associated to the
R
. gveda and R . ksarvānukraman. ı̄, also deal with various aspects of vedic
metres. The Agni purān.a, Nārada purān.a, Garud.a purān.a, and the
Vis.n.udharmottara purān.a, Nāt.ya Śāstra by Bharata, and Varāhamihira’s
Br.hat sam . hitā are some of the fairly old texts which have separate sec-
tions dealing with Sanskrit prosody. Subsequent to the classic work of
Piṅgala, apart from commentaries on it, like that of Halāyudha (called
Mr.ta sañjı̄vanı̄ ([11]), that of Yādava Prakāśa ([19]), there have been
many authors like Kedāra Bhat.t.a ([25]), Svayambhū (847 A.D.),
Ks.emendra (1100 A.D.) and Gaṅgādāsa (1500 A.D.)([6]) and others
([26]), who have written texts on prosody. As we mentioned earlier,
even Kālidāsa is said to be the author of the text Śrutabodha, dealing
with classical Sanskrit prosody. All of these are heavily influenced by
the monumental work of Piṅgala. There have also been many Jain au-
thors who have written on prosody, like the author of Jānāśrayı̄ (6th to
the 7th century A.D)([12]), Jayakı̄rti (1000 A.D.) Jayadeva (1000 A.D.)
and that polymath from Gujarat, Hemacandra (1088–1172 A.D.)([7]).
§5. Piṅgala, the author of Chandas Sūtra

As is the case with many of the ancient personages in India, very little
40 R. Sridharan
is known about Piṅgala himself except that he was highly venerated

and referred to as Piṅgalācārya or Piṅgala Nāga. (Nāga in Sanskrit
means a serpent and serpents are supposed to be endowed with great
wisdom). Some think that he was identical with Patañjali the author
of the Mahābhās.ya. S.adguruśis.ya in his commentary (1187 A.D.) on
R. ganukraman.ı̄ refers to Piṅgala as pān.inı̄yānuja which can be inter-
preted to mean that Piṅgala was a younger contemporary of Pān.ini
or even that he was the younger brother of Pān.ini. Though, conjec-
turally, it is thought that Piṅgala lived in the middle of the third century
B.C.,the precise period of Piṅgala is hard to determine. Most probably,
Piṅgala was a younger contemporary of Pān.ini and belonged to the third
century B.C. With reference to his place of birth, we are equally igno-
rant, though it is surmised that he might have been born somewhere on
the west coast of India. That he lived near a coast is perhaps obliquely
corroborated by the statement in the Pañcatantra (2,36) (cf. [18], p.
255) about the manner in which Piṅgala met his death. Stressing the
theme that even the meritorious ones can not take it for granted that
they are safe from assault, it is mentioned there
chandojñānanidhim
. jaghāna makaro velātat.e piṅgalam
Translated into English, it reads Piṅgala, the repository of the knowledge

of metres was killed by a crocodile on the sea shore. The full verse in
fact says that Pān.ini was killed by a lion, Jaimini by an elephant and
Piṅgala by a crocodile. Albrecht Weber in his book, “Uber die Metrik
der Inder” ([27]), guesses that this enumeration is perhaps in the order
of time and therefore Piṅgala probably was later in time than Pān.ini
and Jaimini.
§6. Piṅgala’s Chandas Sūtra

However uncertain one is about Piṅgala as a man and his life history, his
work on chandas (in eight chapters, containing 315 sūtras) is very much
extant and has been commented upon, as we said earlier, by several dis-
tinguished authors including Halāyudha (11th century), Yādava Prakāśa
(11th century), the latter being the well known teacher of Ramānuja.
As we also mentioned, there are several later texts on Sanskrit prosody
based on Piṅgala’s work, one of the most important one being by Kedāra
Bhat.t.a (12-13th Century). We note also that in the Agni purān.a ([1]),
chapters 327-334 give a summary of the Chandas Śāstra as expounded
by Piṅgala, beginning with a description of prosody thus:
chando vaks.ye mūlajaistaih. piṅgaloktam

. yathākramam
Cdo v#y mlj{-t{, Ep½loÄ\ yTAm\ .

In Varāhamihira’s Br.hatsam. hitā ([23]) in section 104, which deals
with grahagocarādhyāya (movements of planets), verse 58, emphasising
the rule of prosody (already found in the first chapter of Piṅgala’s Chan-
das Sūtra), reads:
prakr.tyāpi laghuryaśca vr.ttabāhye vyavasthitah.

sa yāti gurutām. loke yadā syuh. susthitā grahāh.
þk(yAEp lGy vbA &yvE-Tt, .

s yAEt gztA\ lok
ydA -y, sE-TtA g}hA, ;
Very much like the final syllable in a verse which is deemed long by
the rules of prosody even if it is short, a person though of mean birth,
and reprehensible in character, becomes respectable in this world, if the
planets are favourable.
§7. A brief discussion of Piṅgala’s Chandas Sūtra

As we said earlier, Piṅgala’s Chandas Sūtra contains 315 sūtras dis-
tributed over eight chapters. Among these, the sūtras of the first three
chapters and the first seven sūtras of the fourth are devoted to vedic
metres. As mentioned before, the two basic building blocks of Sanskrit
prosody are the guru (g) and the laghu (l). These correspond to the
Greek syllables: thesis and arsis. From these, the following groups of
disyllables can be built:
g g − which in Greek is the disyllable spondee

l g − iambic
g l − trochaeus
l l − pyrrhic
42 R. Sridharan
Obviously, the number of trisyllables is eight and are as written

below:
Trisyllable Greek name Sanskrit name
ggg molossus magan.a
lgg bacchius yagan.a
glg amphimacer ragan.a
llg anapaestus sagan.a
ggl antibacchius tagan.a
lgl amphibrachys jagan.a
gll dactylus bhagan.a
lll tribrachys nagan.a
In the first chapter of his work, Piṅgala gives the mnemonics ma, ya,
ra, sa, ta, ja, bha, na to the set of trisyllables written above. Any trisyl-
lable is called a gan.a, so that the trisyllables are denoted, respectively
by magan.a, yagan.a, ragan.a, sagan.a, tagan.a, jagan.a, bhagan.a, nagan.a.
These are referred to in the Piṅgala’s Chandas Sūtra as as..tau vasava iti.
Piṅgala remarks that these gan.as along with the guru and laghu form
the basis of all prosody.
Many works on prosody, like Vr.tta Ratnākara of Kedāra Bhat.t.a
(1150 A.D.), the commentary of Yādava Prakāśa (circa 1050 A.D.) and
many other commentators of Piṅgala’s work have the following couplet
which expresses poetically the pre-eminence of the above ten units of
prosody:
myarastajabhnagairlāntaih. ebhirdaśabhiraks. araih.

samastam. vāṅmayam
. vyāptam . trailokyamiva vis.n.unā
Myr-tjÞg {lA t{, eEBd fEBr"r{, .

sm-t\ vAÁy\ &yAØ\ /{loÈEmv Ev ZnA ;
The world of speech is enveloped by the ten units ma, ya, ra, sa, ta,
ja, bha, na, g and l, like Lord Vis.n.u permeating the three worlds.
(The same statement is also made by S.adguruśis.ya in his commentary
of R
. ksarvānukraman. ı̄.) We quote another śloka given by Kedāra Bhat.t.a
in his Vr.tta Ratnākara, which gives a mnemonic for the eight gan.as:
ādimadhyāvasānes. u bhajasā yānti gauravam

yaratā lāghavam
. yānti manau tu guru lāghavam
aAEdm@yAvsAnq BjsA yAEt gOrv\ .

yrtA lAGv\ yAEt mnO t gzlAGv\ ;
A rough translation in to English of the above śloka reads:
The guru moves into the first, middle and the last position in bha,
ja and sa. The laghu moves into the first, middle and the last positions
in ya, ra and ta. ma and na represent all gurus and all laghus.
There are many features of Sanskrit prosody which distinguish it
from the Greek. Greek prosody had its origin in music and dance,
whereas in India, prosody began with the vedic chants. Also, whereas
in Greek prosody, scanning is achieved though the analysis of the posi-
tion and nature of disyllables, in Sanskrit, it is through the analysis of
trisyllables and the two single syllables g and l.
We include at this point a few facts on vedic as well as classical
prosody. In general, metrical music deals with three factors: the sound
value of a syllable, syllabic quantity and the time taken for the utterence
of a syllable. In vedic metres, the music depends only on the modulation
of the voice in the pronunciation of the syllables; the essential features of
the syllables, namely whether they are short or long do not matter. On
the other hand, the music of classical metres depends on the essential
features of the syllables, their variations and their order of succession.
Hence, in classical prosody, a single letter could not be the unit of a
metrical line as in vedic metres. A mere mention of the number of syl-
lables which are all independent units sufficed to give an idea of the
metrical line in the vedic metre and there was no need to give the es-
sential features of the letters nor was it necessary to say how they were
related to each other. But both these points required to be stated in the
case of classical metres. Hence a method had to be found for scanning
classical metres. Units of two syllables and their fourfold combinations
are a choice and such a choice was indeed tried out by a Jain prosodist,
as has been pointed out by H. D. Velankar in his book Jayadāman([13]).
But these were found to be inadequate to express the basic constituents
of the music of a metre, especially in the case of longer verses. So a
new unit had to be found by the classical prosodists, which was neither
too long nor too short. In ancient India, the number 3 was the least
44 R. Sridharan
number which denoted multiplicity; the number 2 did not really signify
plurality and indeed enjoyed too special a status. This is perhaps one
of the reasons why, as H. D. Velankar suggests in his Jayadāman([13]),
groups of three syllables were chosen by the ancient prosodists of India
for scanning classical metres.
The second chapter of Piṅgala’s Chandas Sūtra introduces and dis-
cusses various aspects of the seven basic vedic metres: gāyatrı̄, us.n.ik,
anus..tup, br.hatı̄, paṅkti, tris..tup and jagatı̄. Gāyatrı̄ consists generally
of three pādas of eight syllables each and hence has 24 syllables in all,
and from then on, the number of syllables in these metres increases by 4
at a time, so that us.n.ik has 28 syllables, anus..tup has 32, br.hatı̄ has 36,
paṅkti has 40, tris..tup has 44 and jagatı̄ has 48. Eight different varieties
of these metres, are also discussed. Thus the seven basic vedic metres
are divided into eight forms each, and totally there are 56 different kinds
of metres.
In the third chapter, the notion of a pāda (foot) in Sanskrit prosody
(which is very different from the notion of a ‘foot’ in Greek prosody)
is discussed. Rules regarding filling of a pāda are also discussed. For
example, in the gāyatrı̄ when the number of syllables falls short of the
required number of eight syllables, as in the following:
tatsaviturvaren. yam
t(sEvtv ry\
where there are only seven syllables, one should scan it as:
tatsaviturvaren. iyam
t(sEvtv rEZy\
changing y to iy.
In this chapter, nine forms of the gāyatrı̄ metre in terms of the
number of pādas are described. To give an example, one could have
a gāyatrı̄ stanza containing four pādas with six syllables each, which is
called catus.pāda gāyatrı̄. Halāyudha, in his commentary gives such an
example from the Atharvaveda, Kān.d.a 6, Sūkta 1.1 ([4]).
It is interesting to note that a non vedic catus.pāda gāyatrı̄ stanza

(attributing it to the Pāñcālās) is also quoted in the Nidāna Sūtra
(prapāt. haka 1, Chando viciti)([16]), whose meaning is unclear.
It is also interesting to note that in §16 of the R
. kprātiśākhya ([20]),
there is an example of such a stanza (stanza 7), which is given by
Śaunaka. The stanza runs as follows:
indrah. śacı̄patir balena vı̄l.itah.

duścyavano vr.s.ā samatsu sāsahih.
idý, fcFpEtr^ bl
n vFEt, .
dôyvno vqA sm(s sAsEh, ;
This stanza is also found in the Nidāna Sūtra (prapāt. haka 1, Chando
viciti)([16]). A small part of this stanza occurs in the R . gveda (Eighth
Man.d.ala, 19;20) namely:
yenā samatsu sāsahah.
y
nA sm(s sAsh,
(Yādava Prakāśa in his commentary of Piṅgala Chandas Sūtra notices
this fact).
Similar forms of other metres are also discussed in this chapter. Men-
tion is made of a class of those metres whose first and last verses have
correct number of syllables, but whose middle verses have smaller num-
ber of syllables. Such metres are called pipı̄lika madhya that is, with
a middle like that of an ant! For example, there are gāyatrı̄ stanzas in
which the first and the last pādas have eight syllables but whose middle
pāda has only three syllables. A general rule states that the number of
syllables in the first pāda determines the metre.
To the seven basic metres are sometimes associated the seven svaras
of music : namely .sad.ja, .rs.abha, gāndhāra, madhyama, pañcama, dhai-
vata and nis.āda (respectively); also the following colours: sita (silvery),
sāraṅga (variegated), piśaṅga (brown), kr..sn.a (black), nı̄la (blue), lohita
(red) and gaura (white); and to the seven rishis: Agniveśya, Kāśyapa,
46 R. Sridharan
Gautama, Āngı̄rasa, Bhāradvāja,Kauśika and Vāsis..tha. These identifi-

cations are intended as alternate methods to identify these metres, in
case there is a confusion!
In chapter 4, after discussing fifteen kinds of vedic metres from utkr.ti
to jagatı̄, Piṅgala introduces the cryptic statement ‘from now on classical
metres’ and from then on, he deals only with classical metres till the end
of the book. He in fact deals in the rest of this chapter with the so called
mātrā vr.ttas, that is those metres of classical Sanskrit based on the
syllabic instants (a syllabic instant being the time taken to pronounce a
short syllable: a long syllable takes twice as much time and is therefore
said to constitute two syllabic instants). He discusses, in particular,
the Āryā and the Vaitālı̄ya metres. (We note, incidentally, that the
Āryabhat. ı̄ya of Āryabhat.a is written in the Āryā metre.)
In the fifth chapter, Piṅgala discusses the so called vr.tta chandas.
He classifies stanzas with four pādas into three types: sama, ardhasama
and vis.ama. Samavr.ttas are those which consist of the same number of
syllables in each pāda, while ardhasamavr. ttas have the same number of
syllables in the first and the third pādas, as well as in the second and
the fourth pādas. vis.ama vr.ttas are those in which all the pādas have
unequal number syllables.
The aim of the sixth chapter of Piṅgala’s Chandas Sūtra is principally
to define the notion of yati (caesura). The sūtra which describes yati
is yati vicchedah. The word vicchedah. signifies ‘resting place’. It is the
mechanical pause introduced in the middle of the verse. As against the
irregular pauses in the vedic metres like tris..tup and jagatı̄, it is regularly
admitted in classical metres. While the origin of yati can be traced to the
need for the ease of recitation, it evolved into an art and ornamentation
in classical poetry. The concept of yati has been discussed at length
by all the later prosodists and has become a regular feature of classical
vr.ttas. The effectiveness of yati in classical Sanskrit poetry, is best
illustrated in the beautiful verses of the exquisite Meghadūta of Kālidāsa
(in the slow-moving, majestic metre of mandākrāntā, a classical metre,
with seventeen syllables, with pauses at the end of the fourth and tenth
syllables in each pāda).
In the seventh chapter, Piṅgala describes and discusses metres ati-
jagatı̄, śakvarı̄, atiśakvarı̄, as..ti, atyas..ti, dhr.ti, atidhr.ti, kr.ti, prakr.ti,
ākr.ti, vikr.ti, sam
. kr.ti, abhikr.ti and utkr.ti which are the so called atichan-
das (hyper metres) containing 52, 56, 60, 64, 68, 72, 76, 80, 84, 88, 92,
96, 100 and 104 syllables respectively. At the end of the chapter, he also
explains the metre dan.d.aka.
The eighth chapter which is the concluding chapter of Piṅgala’s book
begins with the sūtra, atrānuktam
. gāthā; Piṅgala’s idea is to include in
this chapter those metres which had not been mentioned in the earlier
chapters.
The last fifteen sūtras of this chapter (sūtras 20 till 35) are the most
interesting ones from the point of view of mathematics and deal with
binary arithmetic and combinatorial questions arising out of the study
of prosody. We shall discuss these in the next section.
§8. Piṅgala’s sūtras and binary arithmetic

Since prosody deals with two symbols l and g and their repetitions, it is
rather an easy matter for us (who live in this computer age) to guess1
that the study of prosody should naturally lead to questions on binary
arithmetic. Indeed, the study of prosody did lead the ancient Indian
mathematicians to binary arithmetic and combinatorics, as is evidenced
by the sūtras 20-35 in the eighth chapter. As is usual with Piṅgala,
these sūtras are cryptic to the point of being obscure. However, as
is customary with the ancient Indian system of preserving knowledge,
the later commentators of Piṅgala’s Chandas sūtra have provided ample
explanations of the sūtras ([17], [22]).
The sūtras 20-23 deal with the construction of the so called prastāra
of a metre, which can be translated roughly into English as a matrix or
an array of syllables. The laghus and gurus in a metre of a given length
are listed horizontally as rows (or lines) in a prastāra. This device of
a prastāra can be thought of as a table written either on the ground
or on a board. The rules for the construction of a prastāra, for metres
of length one, two or three are given in these sūtras. For example, the
prastāra for a metre of length 1 is obtained by first writing the symbol
g (for guru) and beneath it the symbol l (for laghu). The prastāra for
1
One remembers the words of Schiaparelli, the Italian historian of Early Greek
Astronomy who wrote in the introductory section of his paper on the work of Eudoxus
on Astronomy: “Tutto il nostro merito sta nell’esser venuti al mondo piu tardi”: ‘Our
sole merit consists in having come to the world a little later’.
48 R. Sridharan
a metre of length 2 starts with a horizontal row with two gurus: g g.

We begin the next row , by writing l (for a laghu) below the first entry
g of the first row and write g below the second entry g of the first row,
so that this row reads l g. In the third row, we begin with a g and
write a l beneath the next entry g of the second row so that, the third
row reads g l. We begin the fourth row with an l and write a l below
the the next entry l of the third row. The prastāra for a metre with two
syllables is now complete and is the array of four horizontal rows
g g 1
l g 2
g l 3
l l 4
consisting of two syllables each. The general rules for constructing the
prastāras of metres of a given length n are similar and explained by
the sūtras. Namely, we start with a horizontal row consisting entirely
of n gurus. The rest of the rows of the prastāra are constructed by
using the following rule: Start any row and continue filling the row with
gurus until we see for the first time a guru in the previous row. Then
write a laghu as the entry for this row below this guru and from then
on, copy the rest of the syllables from the previous row. We continue
filling rows this way until we reach a row consisting of all laghus, where
we stop. 2 This method applied to two syllables gives obviously the
prastāra of two syllables we have written above. The prastāra for a
metre of three syllables, using the rule described above gives the table
for the eight gan.as (trisyllables) we wrote down in the beginning of the
previous section.
The sūtra 23 reads vasavastrikah., which simply enumerates the num-
ber of trisyllables as eight! (there are eight vasus according to the vedic
lore!)
2
As has been kindly pointed out by Professor M.G. Nadkarni, this rule applied
to infinite sequences of zeros and ones (g = 0, l = 1) gives rise to a transformation
on the space of sequences of zeros and ones. It is a very basic object in ergodic
theory called dyadic coding machine or odometer transformation, and when viewed
as a transformation of the unit interval, it is called von Neumann transformation, a
name given by Kakutani. This transformation plays a very important role in orbit-
equivalance theory and related areas.
Let us also add one more fact regarding the construction of the
prastāra. We number the rows of a prastāra serially with the first row
of the prastāra consisting of all gurus being numbered as 1.
Before discussing the rest of the sūtras, it is perhaps convenient to in-
troduce a stanza which lists the various techniques, termed as pratyayas,
by which some arithmetic questions related to metres can be analysed.
This stanza is found in text books on prosody subsequent to Piṅgala’s
work. For instance, it is found in Kedāra Bhat.t.a’s vr.tta ratnākara, ([25])
Yādava Prakāśa’s commentary of Piṅgala’s Chandas Sūtra, Hemacan-
dra’s chandonuśāsana, and in many other works on prosody. The stanza
in question ([25], p. 187) reads as follows:
prastāro nas..tamuddis..tam ekadvayādi lagakriyā

saṅkhyā caivādhvayogaśca .sad.ete pratyayāh. smr.tāh.
þ-tAro n£mE£\ ekyAEd lgEyA .

sºA c{vA@vyog qXt þ(yyA, -mtA, ;
As we said, the above stanza enumerates the various components of
some of the arithmetic aspects of prosody, namely: (i) prastāra (whose
meaning we just now explained), (ii) nas..tam, (iii) uddis..tam, (iv) ekad-
vayādilagakriyā, (v) saṅkhyā, (vi) adhva yoga.
The following stanza ([25], p. 188) summarises what we said already
about the way a prastāra is constructed:
pāde sarvagurāvādyāllaghum . nyasya guroradhah.

yathopari tathāśes. am
. bhūyah. kuryādamum
. vidhim
ūne dadyād gurūneva yāvat sarvalaghurbhavet
prastāro’yam
. samākhyātah. chandovicitivedibhih.
pAd sv grAvAA¥G\ y-y grorD, .

yTopEr tTAf q\ By, kyA dm\ EvEDm^ ;
Un dAd^ g!nv yAv(sv lGB vt^ .
þ-tAro_y\ smAHyAt, CdoEvEcEtv EdEB,;
We shall describe now each of the other aspects listed above re-
lated to the arithmetic of prosody. Sūtras 24 and 25 of chapter eight of
Piṅgala’s Chandas Sūtra, which read:
50 R. Sridharan
(24) l’ardhe
and
(25) saike g
refer to the process of nas..tam and this word means ‘vanishing’ or ‘dis-
appearance’. Suppose that the prastāra of the metre (which is usually
written on the sand) has been erased by mistake. The process described
shows how to recover the metre only through the knowledge of the num-
ber of the row in which the particular metre had appeared. This process
is illustrated by the following example: Suppose that we know that a
certain metre with a fixed number of syllables say 6, occurs as the 44th
row in the prastāra, how does one write down the corresponding me-
tre? The answer is given by the two sūtras above as elaborated further
by the following stanza in Vr.tta ratnākara ([25], p.192), which explains
the process nas..tam. (There are similar explanations of these sūtras in
Halāyudha’s and Yādava Prakāśa’s commentaries.)
nas..tasya yo bhavedaṅkastasyārdhe’rdhe same ca lah.

vis.ame caikamādāya tasyārdhe’rdhe gururbhavet
n£-y yo Bvd¬-t-yAD
_D sm c l, .
Evqm c
{kmADAy t-yAD
_D
gzB vt^ ;
The procedure indicated is best explained by applying it to the ex-
ample mentioned above: the number 44 being divisible by 2, we write
an l (for laghu) and divide 44 by 2 to get 22. The number 22 being
still divisible by 2, we append an l to the earlier laghu and divide 22 by
2, to get 11. Up to this point, the procedure is indicated by sūtra 24
which says if it is possible to halve, then an l. When we, however, hit
the number 11 which is odd, sūtra 25 takes over and it says otherwise
add 1 and a g. Now add 1 to 11 and write a g (a guru). The number
now is 12, which is divisible by 2; and we divide by 2 to get 6. Now 6
being divisible by 2, sūtra 24 applies, we write an l and divide 6 by 2.
We get 3 which is odd. Sūtra 25 tells us that we should add 1 and write
a g. We now get 4, which we divide by 2 to get 2. We write an l and
divide 2 by 2 to get 1 as quotient and we stop here (since the metre has
6 syllables) and the metre we are looking for is
l l g l g l
This is also the general rule given in the sūtras 24 and 25 (and
explained in the stanza) for writing down a metre, given the number of
its row in the prastāra.
The process uddis..tam is indicated by two sūtras of the Chapter 8 of
Piṅgala’s Chandas Sūtra which read:
(26) pratilomagan
. am dvirlādyam
and
(27) tatogyekam
. jahyāt
and expanded upon in the following couplet of Kedāra Bhat.t.a ([25],

p.194):
uddis..tam
. dvigun.ādyadyuparyaṅkānsamālikhet
laghusthā ye tu tatrāṅkāstaih. saikairmiśritairbhavet
uE£\ EgZAApy ¬AsmAElKt^ .

lG-TA y t t/A¬A-t{, s
{k{Em Et
{B v
t^ .
These sūtras, as interpreted by the couplet above, answer the follow-
ing question: Suppose that one is given a metre with a certain number
of syllables what is the number of the row representing this metre in the
prastāra?
The process uddis..tam is thus the converse of nas..tam and can be
translated as ‘determination’; it gives a method of determining the num-
ber of the row representing a metre with a certain number of syllables.
The answer, as given by the couplet is the following: We make the
number 1 correspond to the first syllable from the left and from then
on, make powers of 2, namely 2, 4, 8, . . . correspond to each succeeding
syllable. Ignoring the powers of 2 corresponding to the gurus of the
metre and adding the powers of 2 corresponding only to the laghus of
the metre and increasing this sum by 1 gives the requisite number of
the line in the prastāra. (Put in the mathematical language, one thinks
of the metre as a mnemonic for a dyadic expansion by thinking of the
laghus as representing 1 and the gurus as representing 0!) Let us consider
as an example, the metre
l l g l g l,
52 R. Sridharan
as above. Then the number in question is 1 + 2 + 8 + 32 = 43 increased

by 1, that is 44.
It should be remarked at this point that the sūtras 26 and 27 as stated
by Piṅgala do not suggest the above procedure outlined by the couplet.
The sūtras themselves have been interpreted by Halāyudha in a different
way and this interpretation is also found in the commentary of Piṅgala’s
Chandas sūtra by Yādava Prakāśa. We shall discuss this presently. But
before doing this, let us note that the processes of uddis..tam and nas..tam
described above, together give a one to one correspondence between
non-negative integers and their dyadic expansions, via, metres. In fact,
given any metre we get an integer by the process described above, (by
assigning the value 1 to a laghu and 0 to a guru and assigning the
value 2i−1 to the syllable which occurs at the ith position from the left;
summing these numbers and adding 1 to it we get the number of the row
corresponding to this metre in the prastāra). Conversely, sūtras 24 and
25 (explained further by the process of nas..tam) assign to every integer
a metre. These two processes are obviously inverses of each other. We
further note that the metre which consists only of gurus corresponds to
the dyadic expansion of 0 and since this is the first row of the prastāra,
the number of the row corresponding to any metre is one more than the
number given by the corresponding dyadic expansion.
We shall now give the interpretations of Halāyudha and Yādava
Prakāśa of the sūtras 26 and 27 of Piṅgala, which give a very interesting
method of computing the number represented by a dyadic expansion.
We shall explain this principle now, mainly through examples, and
then state the general principle without proof (the proof is easy to es-
tablish).
Consider for example the string of syllables:
l g l
The rule given by the sūtras (and explained by Halāyudha and Yādava
Prakāśa in their commentaries), applied to the above metre says that we
first look at the syllable on the extreme right. Noting that this syllable
is a laghu, we attach the value 2 to it. We next look at the syllable to
its immediate left. Noting that this is a guru, we attach to it the value
3 which is one less than twice the number 2, attached to the previous
syllable. We then look at the next syllable to the left which is a laghu.
To this we attach twice the value attached to the previous syllable and
this is 6. The number of the row representing this metre in the prastāra
is 6!
We note that according to our earlier computation, the above row of
syllables represents the number 1 + 4 increased by 1 which is 6 again!
As we shall remark presently, the above process applies, in general,
to all metres of a given length n and the number associated to the
first syllable is indeed the number of the row of the given metre in the
prastāra of metres of length n.
We look at the gāyatrı̄ metre, considered earlier, as another example.
l l g l g l
We assign the value 2 for the laghu on the extreme right, the value
4 − 1 = 3, for the next syllable on its left which is a guru and then 6 for
the next which is a laghu, then 11 for the next syllable which is a guru
and 22 for the next syllable which is a laghu and finally 44 for the first
syllable on the extreme left which is a laghu. This is the number for the
gāyatrı̄ row in the prastāra for a metre of six syllables!
The general rule can now be formulated: If we take a metre of any
length, and wish to find out what its number is as a row in the prastāra
of metres of this length, we start by giving the value 2 or 2−1 = 1 to the
syllable on the extreme right, according as it is a laghu or a guru. We
multiply this number by 2 and attach this number to the next syllable
on its left, if it happens to be a laghu or attach this number decreased by
1 if this syllable happens to be a guru. Keep on repeating this procedure
till we reach the beginning syllable of the metre. The number attached
to this syllable is the number of the row in the prastāra.
It is easily verified that the number obtained by the procedure indi-
cated above coincides with the number given by the dyadic expansion
(by assigning the value 1 to a laghu, 0 to a guru, increased by 1).
Thus, the above is another method of finding a number through its
dyadic expansion and this does not use addition of terms (as the earlier
one did) and is more algorithmic, suited to the computer. In this sense,
this ancient method is as modern as that of the computer! Actually, the
54 R. Sridharan
sūtra 26 says that we first reverse the metre and carry the process from
left to right.
We now turn to the sūtras 28 to 32 and 34, 35 of Piṅgala, which deal
with the combinatorics given rise to by the study of metres. The sūtras
in question are:
(28) dvirardhe;
(29) rūpe śūnyam;
(30) dvi śūnye;
(31) tāvadardhe tadgun
. itam;
(32) dvirdyūnam. tadantānām;
(34) pare pūrn.am;
and
(35) parepūrn.amiti.
The questions asked and answered are: How many metres with a
given length have gurus ocurring once, twice etc? How many metres
are there with a given length? These questions which naturally arise in
the study of prosody, obviously deal with the theory of permutations
and combinations. We shall see that in this connection, the so called
Pascal triangle, from which one can read off the binomial coefficients
was already constructed by the ancient prosodists of India.
These topics are covered under the headings ekadvayādilagakriyā and
saṅkhyā by the later prosodists like Kedāra Bhat.t.a and others. (As a
matter of fact, Piṅgala’s Chandas Sūtra deals with these topics in the
reverse order.) Piṅgala’s sūtras 28-32 treat saṅkhyā and 34 and 35 with
the computation of number of metres of a given length with prescribed
number of gurus and laghus in it, through the combinatorics of what is
now known as the Pascal triangle.) The two verses in Kedāra Bhat.t.a’s
work ([25], p.196) which describe the first process is the following:
varn.ān vr.ttabhavān saikān auttarādharya tah. sthitān

ekādikramataścaitānuparyupari niks.ipet
upāntyato nivarteta tyajennekaikamūrdhvatah.
uparyādyāt gurorevamekadvayādilagakriyā
vZA n^ vBvAn^ s{kAn^ aOrADy t, E-TtAn^ .

ekAEdmt {tAnpy pEr EnE"pt^ ;
upA(yto Envt t (yjàk{km@v t, .

upyA At^ grorvmkyAEdlgEyA ;
The method to find the number of metres of length n in which gurus

and laghus occur once, twice etc, as suggested in the above verse, is the
following:
We start with a row of length n + 1 consisting of the number 1. (In
what follows we assume for simplicity that n = 6 and the next figure
illustrates the procedure for n = 6.)
1 1 1 1 1 1 1
1 2 3 4 5 6
1 3 6 10 15
1 4 10 20
1 5 15
1 6
1
We start the second row with a 1. For the next position we take
the sum of the number which precedes it in the row (which is 1 in our
example) and the number of the previous row in the position above it
(which is 1 again in our example), and the sum here is 2. We choose the
next number of the row to be once again the sum of the two numbers,
one which is in the preceding position in the row and the number in the
position above it in the previous row.
Hence, in this case, we take 2+1=3 as the next number in the second
row. The third number in the second row is chosen similarly and we
continue this procedure, and end the second row with the number of
entries one less than that of the first row. In our example, the second
row has therefore 6 entries, the last entry being 5 + 1 = 6. We start
the third row once again with a 1; choose the number for the second
position of the row the sum of the number in the row in the position
preceding it which is 1 in our case and the number in the position above
56 R. Sridharan
it in the second row which is 2 so that we take 1+2=3 as the second

number of the third row. We stop this row once again with the number
of its entries one less than the second row, which is five in our example,
the last entry being 10+5=15. We continue this process until we stop
with the (n + 1)th row which has just one entry namely 1.
The number of metres with n syllables in which guru appears only
once is given by the last number of the second row, which is n. This
number is obviously also the number of metres of length n, in which the
laghus appear n − 1 times. The number of metres of length n in which
guru appears exactly twice is given by the last number of the third row
which is seen to be n(n − 1)/2. More generally, the number of metres of
length n in which the guru appears
i-times is given by the last term of
n
the (i + 1)th row, and which is .
i
Thus, the array constructed with the specifications of the two verses
above gives a computation for the binomial coefficients and is the so
called Pascal triangle, (with its base tilted by 45 degrees) which was
constructed by Pascal in 1654. This device had however been used
by the Indian prosodists, under the name meru prastāra, at least two
thousand years earlier, in connection with the study of metres.
It is interesting to note that Bhāskarācārya II, the mathematician,
who lived in the 12th century A.D, in his famous book of problems
called Lı̄lāvatı̄, has the following verse ([8]) which asks for the number of
metres with a prescribed length and with a specified number of gurus or
laghus (and the commentary provides a very simple algorithm for finding
these.) The verse in question (for the gāyatrı̄ metre), for example, is the
following:
prastāre mitra gayatryāh. syuh. pāde vyaktayah. kati

ekādi guravaścāśu kathyatām
. tatpr.thak pr.thak?
þ-tAr Em/ ! gAyìyA, -y, pAd &yÄy, kEt .

ekAEdgrvAf kLytA\ t(pTk^ pTk^ ?
The figure below gives the solution: We begin a row with the length
of the metre as its first entry. The succeeding entries of the row are those
gotten by decreasing this number successively by one at a time, the last
entry of the row being 1. Below this row, we start a new row beginning
with 1, the succeeding numbers in this row being those obtained by
increasing the numbers successively by one, the last entry of this row
being the length of the metre. We fill in a new row above these two rows
by the following numbers. The first entry in the new row shall be the
number obtained by multiplying the first entries of the two rows below,
so that we get as the first entry of the row above as 1 × 6 = 6. The next
entry of the new row is obtained by multiplying the first two entries of
the first row and dividing it by the product of the first two entries of the
second row, so that we get in our example, the number in the new row
to be 6×5
1×2 = 15. The third entry in the new row shall be the product of
the first three entries of the first row divided by the product of the first
three entries of the second row, which for our example is the number
6×5×4
1×2×3 = 20.
We continue the process, till we get the last entry of the new row
which is 1.
6 15 20 15 6 1
6 5 4 3 2 1
1 2 3 4 5 6
More generally, for any metre with length n, we get, as the first
n
entry of the new row, the number n = n · 1 = , the second entry
1
n
to be n·(n−1)
1·2 = and, more generally, for the ith entry the number
2
n·(n−1)···(n−i+1) n
1·2···i = .
i
These, as we know, give the number of metres of length n, in which
the guru (and similarly the laghu) occurs exactly once, twice, . . . , i-
times. In particular,in the example of the gāyatrı̄ metre, the numbers
are 6, 15, 20, 15, 15, 6.
Lı̄lāvatı̄ ([8], Appendix p. 48) has another problem on the deter-
mination of the number of sama, ardhasama and vis.ama vr.ttas in the
metre anus..tbh, which is preceded by a general rule valid for any metre,
given below.
pādāks.aramitagacche gun.avargaphalañjaye dvigun.e

58 R. Sridharan
samavr.ttānām
. saṅkhyā tadvargo vargavargaśca
svasvapadonau syātāmardhasamānāñca vis.amān.ām
pAdA"rEmtgQC gZvg PlÒy EgZ .

smvAnA\ sºA tgo vg vg .
-v-vpdonO -yAtAmD smAnAÑ EvqmAZAm^ ;
The number of syllables in the four verses of the vr.tta in anus.t.up,
being 32 (the anus.t.up, has 8 syllables), the number of arrangements
of the long and short syllables of all the pādas is 232 = 4294967296.
Evidently, the number of arrangements where the pādas are all alike is
the number of arrangements of the syllables in a single pāda and is hence
28 = 256. The number of arrangements occuring as ardhasama vr.ttas
is 216 − 256 = 65280. The number of vis.ama vr.ttas which is the total
number of all the arrangements ‘minus’ the number of possibilities where
two pādas are alike is 232 − 216 = 4294967296 − 65536 = 4294901760.
We discuss finally saṅkhyā. As we remarked earlier, this has been
discussed in the Piṅgala’s sūtras 28-32. Kedāra Bhat.t.a ([25], p.201), on
the other hand, has the following verse describing saṅkhyā.
lagakriyāṅkasandohe bhavetsaṅkhyā vimiśrite

uddis..tānkasamāhārah. saiko vā janayedimām
lgEyA¬sdoh Bv(s\HyA EvEmEt

.
uE£A¬smAhAr, s
{ko vA jny EdmAm^ ;
As the verse says, there are two methods of computing the number
of metres of length n. One can either sum up the numbers (obtained
by the process of lagakriyā) which count the number of metres in which
the gurus occur once, twice, etc. (These numbers are the entries on the
extreme right in the meru prastāra we constructed earlier.) In other
words, one is heresumming
up all the
binomial
coefficients of n and
n n n n
therefore one gets + ···+ = 2 , which is obviously the
0 1 n
number of metres of length n.
Otherwise, one can use the process of uddis..tam. We note that the
required number is the number of metres in the prastāra of the metre.
The length of the metre being n, as we have remarked earlier, the last
row of the prastāra consists of n laghus, and then we know that it cor-
responds to the dyadic expansion 1 + 21 + 22 + · · · + 2n−1 = 2n − 1, and
the number of rows of the prastāra is gotten by adding 1 to it, so that
one obtains 2n .
Piṅgala’s sūtras, mentioned above, give a somewhat elaborate
method of arriving at this number, which we shall not discuss, since,
in any case, it is simple combinatorics.
The sūtra 33 of Piṅgala reads ekona adhvā, which deals with the
space required on the sacrificial ground for writing the prastāra of a
metre of a given length. Since there is no mathematics involved in it,
we shall pass over this sūtra and its explanation given by Kedāra Bhat.t.a
in his Vr.tta Ratnākara.
§9. Concluding remarks

To summarise,our aim in this article has been to highlight the con-
tributions of Piṅgala to Sanskrit prosody with a special emphasis on
the combinatorial aspects. As we mentioned earlier, the influence of
Piṅgala on the later prosodists has been profound. Particularly inter-
esting is the development of Prakrit prosody with emphasis on mātra
vr.ttas. One of the greatest of the later prosodists is Hemacandra whose
name we already have mentioned. The construction of the prastāra and
the other devices mentioned in the earlier section can be extended to
mātra vr.ttas too, as has been explained by Kedāra Bhat.t.a [25]. The
work of Hemacandra ([8], [2]) has a complete chapter on the combina-
torics of prosody with special reference to mātra vr.ttas (in particular, to
the Ārya metre). We do not discuss these here.
Acknowledgements: This article is an expanded version of a half an
hour talk at the joint AMS - India meeting held at Bangalore in Dec
2003. I thank the organisers of the meeting, in particular Professor K.B.
Sinha and Professor Gadadhar Misra, the first for inviting me to organise
the session on “History of Indian Mathematics” in collaboration with
Professor Emch during the meeting and the second for the hospitality
provided during the meeting. It is my pleasure to acknowledge the
immense support and cooperation of Professor G. Emch. Many of my
friends and staff at the CMI have given me unstinted help during the
preparation of this article. I would like to thank particularly C. S.
60 R. Sridharan
Aravinda, K. Paramasamy, Matthias Meulien and Sripathy who helped

me in texing. (Aravinda also went through the tex version several times
to correct many mistakes and helped me in very many ways) I admire
the remarkable patience they all showed while explaining the arcanas of
texing to a novice like me. I am very grateful to V. Balaji for his immense
kindness and ever willing help. He is, among many other things, solely
responsible for the appearance in this article of the beautiful devanagari
script. I am indebted to M. D. Srinivas who has been, on all occasions,
ready to offer his help in every possible way, and to K. R. Nagarajan,
to whom I turn for help and advice on all matters.
References
[1] Agni Mahāpurān.am (English translation according to M.N. Dutt,

edited and revised by K.L.Joshi, Parimal Publishers Delhi, 2001.
[2] Ludwig Alsdorf, The Pratyayās: Indian Contribution To Combina-

torics (Translated from German by Sreeramula Rajeswar Sarma),
Indian Journal of History of Science, 26(1), 1991.
[3] E. V. Arnold, Vedic Metre in in its Historic Developement, Motilal

Bananrasi Dass, Delhi, 1905.
[4] Atharvaveda Sam . hitā (Sanskrit Text, English translation, Notes

and index of verses According to the translation of W. D. Whitney
and Bhās.ya of Sāyan.ācārya, Edited and revised by K. L. Joshi),
INDICA, reprinted by Parimal Publications, Delhi, 2000.
[5] Br.haddevatā (attributed to Śaunaka), A. A. Macdonell, Motilal Ba-

narsidass publishers private limited Delhi, 1904.
[6] Chandomañjarı̄ of Gangādāsa, commentary by Pt. Paramesh-

waradin Pandey, Edited by Pt. Satyanarayana Shastri Khanduri,
Krishnadas Academy, Varanasi, 1987.
[7] Chando’nuśāsana by Hemacandrasūri, edited by H. D. Velankar,

Adhisthata Singhi Jain Sastran Siksapitha, Bharatiya Vidya Bha-
van, Bombay, 1961.
[8] H. D. Colebrooke: Translation of Lı̄lāvatı̄ (with notes by H. C.

Banarjee), Asian Educational Sciences, New Delhi, Madras, 1993.
[9] H. D. Colebrooke, Essays on History, Literature and Religions of

Ancient India (Miscellaneous Essays, Vol 2), Cosmo Publications,
New Delhi 1977.
[10] Daśopanis. ads (with the commentary of of Sri Upanishad Brahma

yogi), Edited by C. K. Kunhan Raja, revised by Prof. A. R. Ra-
manathan, The Adyar Library Research Centre, Chennai, 1984.
[11] Halāyudha’s commentary on Piṅgala’s sūtras, edited by Pt.

Kedaranatha, Vasudeva Laxmana Sastri Panasikara, Chaukhamba
publishers, Varanasi, 2002.
[12] Jānāśrayı̄ Chandoviciti, published by the curator, the university

manuscripts library, Trivandrum (with an introduction by P. K.
Narayana Pillai), Trivandrum Government press, 1949.
[13] Jayadāman edited by H. D. Velankar, Haritos.amālā No.1, Bombay,

1949.
[14] Arati Mishra, Origin and Developemnent of Sanskrit metres, The

Asiatic Society Calcutta 16, 1989.
[15] Nātyaśāstra by Bharatamuni (with the commentary “Abhinava

Bhārati” of Abhinava Gupta, 4 volumes, edited by Parimal Pub-
lishers Delhi, 2001.
[16] Nidāna Sūtra of Patañjali: Edited with an introduction, a frage-

mentary commentary and indices by K. N. Bhatanagar, Meherc-
hand Lachhman Das, Oriental and Foreign booksellers, 1971.
[17] M. Ojanguren, Dyadische Metrik und Kreisumfang: Zwei Beispiele

alter indischer Mathematik, Preprint, Lausanne, 2000.
[18] Pañcatantram of Viṡṅuśarma (edited with Hindi commentary by

Shyamacharanapandeya), Motilal Banarsidass, Delhi, 2002.
[19] Piṅgala chandas with Yādavācārya’s commentary (edited by

Srīdhara Vāsis.t.ha), Nag Publishers, Delhi 2004.
62 R. Sridharan
[20] R
. kprātiśākhyam of Śaunaka (with the commentary of Uvvat.a),
Virendra Kumar Sharma, Chowkamba Sanskrit Pratisthan, 1970.
[21] Śrutabodha of Kālidāsa, (edited with Ānandavardhinı̄ commentary

in Sanskrit, by Sri Pt. Kanakalal Sharma), Chowkhmba Sanskrit
Samsthan, second edition, Varanasi, 1984.
[22] A. Van Nooten, Binary Numbers in Indian Antiquity, Journal of

Indian Philosophy, 1993, p.31-50
[23] Varāhamihira’s Br.hatsam

. hitā (with English translation, exhaustive
notes and literary comments by M. Ramakrishna Bhatt), Motilal
Banarsidass, Delhi, 1997.
[24] Virhāṅkā’s Vr.ttajāti samuccaya, edited by H. D. Velankar, 1962.
[25] Vr.tta Ratnākara by Kedarabhatta (with commentary by Sridha-

rananada Shastri), Motilal Banarsidass, Delhi, 1993.
[26] Vr.tta Ratnāvalı̄ (by Veṅkat.eśa kavi), edited by A. Veeraraghavan,

Sarasvati Mahal Library, Tanjore, 1999.
[27] A. Weber, Über die Metrik der Inder, Harrwitz und Gofsmann,
1863.
[28] H. Winternitz, A History of Indian literature, (English translation),

reprinted, 3 volumes, Motilal Banarsidass publishers, Delhi, 1999.
Shedding Some Localic and Linguistic
Light on the Tetralemma Conundrums
F.E.J. Linton
Mathematics Dept., Wesleyan University
Middletown, CT 06459 USA
inton@wesleyan.edu
Abstract
Numerous authors over the centuries have puzzled over what
has been called the Buddhist paradigm of catus.kot.i. A classic
example: the four statements, considered both mutually exclusive
and jointly exhaustive,
(i) the Tathagata exists after death;
(ii) the Tathagata does not exist after death;
(iii) the Tathagata both does and does not exist after death;
(iv) the Tathagata neither does nor does not exist after death.
We oer some linguistic gedanken-experiments illustrating ev-
eryday situations in which appropriate analogues to the above four
statement-forms are entirely plausible as mutually exclusive or
jointly exhaustive alternatives; and we oer a framework, based
on the logical paradigms of locale or topos theory, illustrating how
forms (iii) and (iv), in particular, need be neither contradictory,
nor paradoxical, nor even mutually equivalent.
1 Foreword
As an exemplary model { what in German might be called a Vorbild or

Musterexemplar { of catus.kot.i or tetralemma, one would be hard-pressed
to nd a more quintessentially perfect instance than the following, taken
from verse XVIII.8 of the Mulamadhyamakakarika by Nagarjuna, as
kindly rendered into English by the generous referee:
63
64 F. E. J. Linton
Anything is either true,

Or not true,
Or both true and not true,
Or neither . . . .
This fragment provides what a mathematician of a certain bent might

call a universal example of catus.kot.i in all regards { form and content,
structure and message { without a single super uous word or restriction.
Of course, to the reader steeped in the bivalent logical tradition
prevalent in the West from the time of Aristotle, or even before, through
the time of Boole, and beyond, already the rst two lines of this frag-
ment would seem to cover all the bases, with the last two being quite
super uous, little more than incomprehensible, contradictory, mystify-
ing mumbo-jumbo, if not downright misleading mystical nonsense.
The very modest aim of the presentation that follows will be to
tease out of the catus.kot.i any lingering ber of paradox, conundrum, or
mysticality, so as to allow such a reader to recognize, in each of the four
alternatives that the catus.kot.i sets forth, a familiar, viable, and relevant
state of aairs. The means by which to accomplish this aim will number
but two: suitable models of (nonstandard) logical systems (cf. [R] for
a thorough if technical introduction to such ideas), and examples from
everyday language.
The reader expecting anything more will, alas, come away disap-
pointed. Neither the history of the catus.kot.i, nor the role of catus.kot.i
in Indian philosophy, nor any critical analysis of their many commenta-
tors, nor any sensitive comparisons or contrasts of those commentators,
or their comments, one with another { no such scholarly discourse { will
be found here. For such material, the reader is better advised to visit
the pages of [B], [C], [G], or [Si], or, even better, the many works cited
therein.
Nor will the reader nd any attempt to provide information as to the
nature of Truth, or Reality, or what it means to be Valid, or to Exist {
here again, for etymological as well as for epistemological enlightenment,
our advice would be to consult articles like [K] and [So], or to browse
through the pages of [JIPR] and of its subsequent sister volumes, where
similar articles have appeared.
Tetralemma Conundrums 65
Finally, beyond oering all due gratitude to the reader who can for-
give these omissions (and abject apologies to the reader who cannot),
I must express profound thanks to Professors Emch and Sridharan for
having encouraged the preparation of the present material in the rst
place; to the referee (alas, anonymous), whose numerous valuable sug-
gestions upon an earlier draft, I hope, I have adequately incorporated
into the present revision; to Professor P. Vanchinathan for a master-
ful translation of my HTML submission into the requisite LaTeX; to
the airlines UAL and Lufthansa, whose frequent yer program and air-
craft, respectively, graciously facilitated my participation in the Banga-
lore conference by providing complimentary air transport between the
North American continent and India; and to the Faculty Research Grant
program of Wesleyan University, for its generosity in underwriting se-
lected additional travel expenses connected with the presentation of this
material at Bangalore.
2 Taming the Terrible Catus

. kot
.i
What the extract from Nagarjuna cited in the foreword suggests, as

regards the catus.kot.i quoted in the abstract (to be found as item /1/ on
page 28 of [B]), is that, writing P for the proposition that the Tathagata
exists after death, the four propositions
(i0 ) P
(ii0) : P (not-P )
(iii0) P &:P (P and not-P )
(iv0 ) : P &::P (neither P nor not-P )
(corresponding to (i){(iv)) are mutually exclusive and cover all possibil-

ities. What sort of logic can be at work here?
Classically, of course, at least in the western tradition, where P
and :P are complementary and ::P = P , the last two formulations
are identically trivial and the rst two already cover all possibilities
(principle of excluded middle, or tertium non datur).
66 F. E. J. Linton
In the intuitionistic logic of a topos, on the other hand { and, unlike

in an earlier lecture [L], we shall refrain here from attempting to oer
any introduction to the notion of topos, or to the sort of logic prevailing
there, preferring to send the interested reader to such standard exposi-
tions of those matters as [J] (especially Chapter 5, sections 1 and 2), or
[L&S] (especially the marvellously informal overview of pp. 123{128), or
[M&M] { the rst two formulations are no longer complementary. They
do remain mutually exclusive, however, and the last two, consequently,
are still identically trivial. It is just that the rst two need no longer
cover all possibilities, that is, the principle of excluded middle need no
longer hold (concrete illustrations of such state of aairs appears below).
If, instead, we envision a logic dual to that of a topos, more like the
logic of the lattice of closed subsets of a topological space, we nally
reach a situation where both P &:P and :P &::P may be non-trivial.
But now P; :P; P &:P; and :P &::P may well no longer be mutually
exclusive. Indeed, at least for closed P , we have the order-inclusions
: &::
P P P &:P P and
: &::
P P P &:P : P;
so that if P is \regular-closed", say, that is, if P = ::P , the last

two formulations coincide and fall within both P and :P (indeed, they
constitute the boundary of P ).
Somehow, :P must not be getting treated purely as the negation of
P . Let us write Q temporarily for this negation of P , and see what we
can make of statements (i) through (iv) in such a setting. They become
(i00 ) P ,
(ii00 ) Q ,
(iii00) P &Q, and
(iv00 ) : P &:Q( = :(P _ Q )),
where the last formulation is logically tantamount to the negation of \P

or Q (or both)", i.e., to the negation of what the rst two alone jointly
cover. Certainly the last item here excludes each of the earlier ones,
and all are, in general, non-trivial. But if all four are to be mutually
exclusive, what (i00 ) and (ii00) are implicitly intending to represent must
surely be rather
(i000 ) P &:Q and

(ii000) : P &Q,
respectively. Then, at least classically, we obtain the four mutually ex-

clusive, jointly exhaustive, atomic generators of the free Boolean algebra
on the two free generators P and Q, viz.:
(a) P &:Q = P Q;
(b) : P &Q = Q P:
(c) P &Q = P &Q, and

(d) : P &:Q(= :((P Q) _ (Q P ) _ (P &Q)) = :(P _ Q )):
But how is one now to make any sense of the hope that Q may stand
for :P ? That is, how shall we maintain the mutual exclusivity and
individual non-triviality of the four items
(a0 ) P &::P;
(b0 ) : P &:P;
(c0 ) P &:P , and
(d0 ) : P &::P ,
obtained from (a){(d) by putting :P in place of Q ?

Let us simplify, for the moment, by assuming that ::P = P , so that
the four conjunctions above become
(a00 ) P &P ,
(b00 ) : P &:P ,
(c00 ) P &:P , and
(d00 ) : P &P .
68 F. E. J. Linton
Next, let us imagine that the second occurence of P in each of these

four conjunctions is merely a homonym for the P that occurs rst.
Mostly, in living languages, homonyms are words that sound alike, but
are spelled dierently and have dierent meanings, like red, the color,
and read, the past participle, or pear, the fruit, pair, the duo or couple,
and pare, the verb meaning to peel (and perhaps also cut up) a fruit
(perhaps even a pear) or vegetable. But there are homonyms also with
both sound and spelling identical, like sucker, which can at once signify
a person easily duped or taken advantage of, or a tendril on a vine.
How may we realize the two occurrences of P in (a00 ){(d00 ) as mere
homonyms of each other? It would be enough, for example, were our
lattice of propositions somehow spatial, that is, representable as some
sort of subsets of some particular space X , to place ourselves in the
Cartesian product X X of the space X with itself. For now, corre-
sponding to P , there arise two clearly distinguishable homonyms of P
in X X : one, the \vertical cylinder" P X over the P in the rst
spatial factor X ; the other, the \horizontal cylinder" X P alongside
the P in the second factor X (cf. Figures 1 & 2).
If we now simply treat each rst occurrence of P in the forms (a00 ){
(d00 ) as instances of the vertical cylinder P X , and each second occur-
rence as the horizontal one X P , then our four conjunctions correspond
to the four rectangles in Figure 3 in the following page ( P &P being in-
terpreted, for example, as the intersection, P P , of P X with X P ,
etc.).
For what it is worth, we exhibit a topos whose internal logic has
system of truth values inherently of this form. Indeed, where S is any
of the very classical topoi of absolutely standard sets { say, made up of
the sets in Godel's constructive hierarchy { the topos S S of ordered
pairs of such sets is such a topos. Its truth value object is the ordered
pair (2; 2) consisting of two copies of the usual two-element Boolean
algebra formed from the ordinal number 2 = f0; 1g, and this has exactly
four global elements: (1, 1) and (0, 0), playing the roles of True and
NotTrue, and serving as counterparts of P &P and :P &:P , respectively;
and (1, 0) and (0, 1), playing the roles of BothTrueAndNotTrue and
NeitherTrueNorNotTrue, counterparts in turn of P &:P and :P &P .
Not every topos whose truth value object has exactly four global ele-
ments has them arranged quite in this way, however. For example, if we
topologize the ordinal number 3 (whose points are the smaller ordinals
0, 1 and 2) by declaring open exactly those subsets of 3 that happen
themselves to be ordinal numbers (viz., ; (the empty subset), f0g; f0; 1g
and all of 3), then the topos of sheaves on this space 3 has truth value
object whose global elements likewise number four, but correspond ex-
actly, even as to their ordering, to the four open subsets of 3 that make
up the topology just described. Here, between True and NotTrue (or 3
and the empty set) we have two intermediate truth values, each neither
True nor NotTrue, but one \more true", as it were, and \less not true",
than the other:
NotTrue = ; < f0g < f0; 1g < 3 = True:
To within isomorphism, this topos may also be depicted as the topos

of double-transitions among sets, that is, as congurations of the form
A !
f
B !
g
C
70 F. E. J. Linton
made up of three sets and two functions, as depicted. The truth value
object for this topos is the conguration above, where 4, 3, and 2 are
the ordinal numbers 4 = f0; 1; 2; 3g; 3 = f0; 1; 2g; and 2 = f0; 1g, and
the functions f+ and f both carry 0 to 0 and 1 to 1, but f (2) = 1,
while f+(2) = f+ (3) = 2, as depicted above.
The four global elements here are simply the four length-two paths,
or orbits, seen to emanate from the various members of 4, the upper-
most and lowermost of which it seems plausible to accept as playing the
roles of True and NotTrue, respectively, while the remaining two paths,
clearly neither True nor NotTrue, somehow represent the values \more
True than NotTrue" and \more NotTrue than True". Or perhaps the
catus.kot.ian expressions \both true and yet not true" and \neither true
nor yet not true" better convey the sense of these intermediate truth
values, though we suspect this is not an illustration of the classical
paradigm the catus.kot.i had in mind.
But in fact, the logic of this topos does realize the way apparent
contradictions are commonly used in everyday speech. A daiquiri made
with far too much lime juice, for example, and a little too much sugar,
may well be called both sweet and not sweet; a coee prepared with just
barely not enough sugar for the taste of a particular coee-drinker may
be disparaged as neither sweet nor not sweet. If the best student to pass
through your department in the past ten years has an uncanny knack
for getting arrested at student political demonstrations, you will be apt
to wonder whether your department should once again post bail for this
student who is both really very bright and yet not really very bright.
Or, of another student, not quite so bright { generally dealing very well
with the more diÆcult problems and readings, but sometimes inexplica-
bly failing utterly when faced with far simpler ones { and yet having an
investment acumen that is simply uncanny, you may well think, some-
what perplexedly, this student is neither really all that bright, nor not
really all that bright.
There are, of course, also everyday linguistic settings in which the
last two catus.kot.i options (iii) and (iv), far from being mutually exclu-
sive, coincide completely. This state of aairs corresponds, perhaps, to
the Trairasika viewpoint (cf. [B], p. 35). A grape-fruit, for example,
sour, to some extent, like all its kin, but remarkably less so than most,
you might be equally happy to describe as both sour and not sour, or
as neither sour nor not sour. Would you like a topos whose truth value
object epitomizes just this situation, not envisioned in the catus.kot.i, of
the last two options (iii) and (iv) coinciding? The Sierpinski topos, as
it is known, is a case in point.
The objects of the Sierpinski topos are shortened versions of the con-
gurations shown above: only two sets, B and C , rather than three,
and only one function g. The truth value object is the right-hand frag-
ment of the truth value object shown above, and has only three global
elements, namely the three one-step paths emanating from the various
elements of 3, which have reason to be thought of as True, Neither-
WhollyTrueNorNotWhollyTrue, and NotTrue (taken from top to bottom),
respectively, though the middle value may equally well be thought of as
TrueInTheLongRunEvenIfNotTrueAtTheOutset. This middle truth value,
in other words, is at once BothTrueAndNotTrue and NeitherTrueNorNot-
True, and is the only alternative to the extreme values True and NotTrue.
3 Afterword
As a nal topic, perhaps not worthy of even this passing mention, let
us take up one objection on the part of some commentators to the
tetralemma paradigm, namely, that there should by rights be yet a fth
alternative, something like NoneOfTheAbove, to the classical four. The
Buddha himself, after all, is reported in one instance to have rejected,
each in its turn, all four alternatives of one particular quadrilemma,
indicating that the truth lay somehow elsewhere.
There are indeed topoi, readily described, whose global truth values
72 F. E. J. Linton
Figure 4:
easily realize the ideal of being ve in number. For that matter, that
ideal can be realized in three wholly dierent ways. In all cases, how-
ever, the lattice of global truth values must, for purely topos-theoretic
reasons (that is, by virtue of what has been called generalized abstract
nonsense), be what is known, to those in the lattice trade, as distributive.
That requirement rules out the last two lattices depicted in Figure 4
above. The remaining ve-element lattices number three: they too ap-
pear in Figure 4, as the rst three on the left: they are all distributive,
but none is Boolean.
And just which of their intermediate members (between True at the
top and NotTrue at the bottom) should be interpreted as BothTrueAnd-
NotTrue, as NeitherTrueNorNotTrue, or as NoneOfTheAbove, I leave as
my parting conundrum to you.
Bibliography
[S&RS ] Sanskrit and Related Studies (Proceedings of the International
Conference, September 23{26, 1993), Cracow Indological Studies
1 (1995). Institute of Oriental Philology, Jagiellonian University,
Cracow (Poland), 1995.
[JIPR ] Journal of Indian Philosophy and Religion 1 (1996). Society
for Indian Philosophy and Religion, Calcutta, 1996.
[B ] Balcerowicz, Piotr: Formal Analysis of Catus.kot.i|A Buddhist An-
ticipation of Multiple-Valued Logic?, in [S&RS], pp. 27{43.
[C ] Chakrabarti, Chandana: The Dialectic of Negation in the Vedantic

and the Platonic Traditions, in [JIPR], pp. 135{147.
[G ] Gareld, Jay L.: Emptiness and Positionlessness: Do the

adhyamika Relinquish all Views?, in [JIPR], pp. 1{34.
M
[J ] Johnstone, Peter T.: Topos Theory, London Mathematical Society

Monograph No. 10, Academic Press, London, New York, San
Francisco, 1977.
[K ] Kudelska, Marta: Etymology of the Word `satyam' in the Light of
the Concept of Brahman, in [S&RS], pp. 179{188.
[L ] Linton, F.E.J.: Time, Truth, and Topoi. SIPR conference oering,

Calcutta, 1997.
[L&S ] Lambek, Joachim, and Phil J. Scott: Introduction to Higher-
Order Categorical Logic, Cambridge Studies in Advanced Mathe-
matics 7, Cambridge University Press, Cambridge, London, New
York, Sydney, 1986.
[M&M ] Mac Lane, Saunders, and Eike Moerdijk: Sheaves in Geometry
and Logic: A First Introduction to Topos Theory, Universitext
series, Springer-Verlag, New York, Berlin, Heidelberg, 1992.
[R ] Rasiowa, Helena: An Algebraic Approach to Non-Classical Logics,
PWN, Warsaw (Poland), and North-Holland, Amsterdam, 1974.
[Si ] Siderits, Mark: Do Persons Supervene on Skandhas?, in [JIPR],
pp. 55{76.
[So ] Sohnen-Thieme, Renate: On the Concept and Function of satya
(`truth') in Ancient Indian Literature, in [S&RS], pp. 235{244.
Brahmagupta's Bhavana: Some
Re ections
Amartya Kumar Dutta
Stat-Math Unit; Indian Statistical Institute;
203, B.T. Road; Kolkata 700108
amartya@isical.ac.in
Abstract
We shall present Brahmagupta's treatment of the indetermi-
nate equation Dx2 + 1 = y2 highlighting some ideas of modern
algebra that are implicit in this ancient work of 628 CE and dis-
cuss the consequent pedagogic potential of Brahmagupta's results.
1 The Bhavana | An Introduction

Mathematics in India attained one of its highest peaks during the 7th
century CE with the arrival of the versatile astronomer-mathematician
Brahmagupta (born 598 CE). His major work Brahma Sphut.a Siddhanta
([B]), comprising over 1000 verses in 24 chapters, was composed in 628
CE. Two of these chapters (12 and 18), dealing exclusively with math-
ematics, were translated into English by Colebrooke in 1817 ([C]).
Brahma Sphut.a Siddhanta is probably the rst ancient Indian text
having a separate chapter (18) on algebra. A substantial portion of this
chapter is devoted to solutions of indeterminate equations of the rst and
second degree. This includes partial solutions of the celebrated Pell's
equation. Brahmagupta is the earliest known mathematician to have
systematically investigated integer solutions of Dx2 + 1 = y2 . In the
process he discovered signicant results on the more general equation
Dx2 + m = y2 called varga-prakr.ti (square-natured) in ancient India.
In the Preface of his treatise on history of number theory, L.E. Dickson
made a special mention of this work ([Di], p. xi):
77
78 Amartya Kumar Dutta
It is a remarkable fact that the Hindu Brahmegupta in the

seventh century gave a tentative method of solving ax2 + c =
y2 in integers, which is a far more diÆcult problem than its
solution in rational numbers.
All the results of Brahmagupta on this topic are clever applications of
a certain law of composition called \bhavana ". This principle can be
formulated, in modern language and notations, as follows:
Theorem 1 (Brahmagupta's Bhavana)

The solution space of the equation Dx2 + m = y2 admits the binary
operations
(x1 ; y1 ; m1 ) (x2 ; y2 ; m2 ) = (x1 y2 x2 y1 ; Dx1 x2 y1 y2 ; m1 m2 ):
In other words, if (x1 ; y1 ; m1 ) and (x2 ; y2 ; m2 ) are solutions of Dx2 +
m = y2 , then so are (x1 y2 + x2 y1 ; Dx1 x2 + y1 y2 ; m1 m2 ) and (x1 y2
x2 y1 ; Dx1 x2 y1 y2 ; m1 m2 ).
The consequent identities

(y1 2 Dx1 2 )(y2 2 Dx2 2 ) = (Dx1 x2 y1 y2 )2 D(x1 y2 x2 y1 )2
are now called Brahmagupta's identities. The result was rediscovered
by Euler during the middle of the 18th century. Euler highlighted the
result in his writings as \theorema eximium " (a theorem of capital im-
portance), \theorema elegantissimum " (a most elegant theorem), etc.1
For brevity, we shall adopt a notation suggested by Weil ([W2], p 21).
For a given positive integer D, (p; q; m) will denote a triple of numbers
satisfying Dp2 + m = q2 . Thus Theorem 1 states:
(p; q; m) (r; s; n) = (ps qr; Dpr qs; mn):
Ancient Indian algebraists had realised the importance of the two laws
and used the special technical term bhavana (composition) | the for-
mula obtained by taking the positive sign was called the samasabhavana2
([W2], p 284{285; [Se], p 168).
1
From the Vedic era, addition has been called samasa (\putting together") and
2
the sum obtained samasta (\whole", \total", etc) | see [D3], p 226. Note that the
prex sam (together) is used like the Latin con expressing \conjunction", \union",
etc.
Brahmagupta's Bh
avana 79
or yoga-bhavana (additive composition) and the one obtained by tak-

ing the negative sign was called the vises.a-bhavana or antara-bhavana
(subtractive composition). In the special case of equal roots and inter-
polators, the rule was called tulya-bhavana (composition of equals); the
general case was called atulya-bhavana (composition of unequals).3
The sheer beauty apart, Theorem 1 has a technical power, a glimpse
of which can be felt from the way it can be used to solve a diÆcult
indeterminate equation like 92x2 + 1 = y2 in a few steps. This example
is mentioned in ([C], p 364) immediately after the verses describing
Theorem 1.
Example 1 (Brahmagupta) Solve the equation 92x2 + 1 = y2 , in

integers.
Solution. One readily observes that 92 12 + 8 = 102 . Composing

the triple (1; 10; 8) with itself (by samasa-bhavana), and dividing the
resulting triple (20; 192; 82 ) by 82 , one obtains the triple ( 25 ; 24; 1) which,
when composed with itself, gives the integer triple (120; 1151; 1). Thus
(120; 1151) is a solution of 92x2 + 1 = y2 .
It is interesting to note that after stating this problem, Brahmagupta
had used the phrase kurvannavatsarad gan.akah. | \One who can solve
it within a year (is truly a) mathematician."
The Pedagogic Issue

What strikes a modern algebraist reviewing Brahmagupta's inge-
nious treatment of the equation Dx2 + m = y2 is an incredible sophisti-
cation in the very attitude towards an algebraic problem. Fundamental
concepts and principles of modern algebra are implicit in the bhavana
and its applications. But though the ideas are modern, the presentation
does not involve the elaborate language of modern abstract algebra. This
aspect of Brahmagupta's work makes it particularly relevant for the al-
gebra training of \would-be creative mathematicians"4 among students
(at higher-secondary or rst-year college level) who have already demon-
strated their skill in high-school algebra.
3
See ([Sh] p 5{8; [C] p 170{172; [DS] p 146{148).
4
Modication of a phrase from Weil ([1], p 228).
It is generally agreed that for a fuller realisation of one's creative

potential, \One should learn from the (works of the) Masters". But
does a higher-secondary student have an immediate, realistic and rapid
access to any of the Masters embodying, even partially, the spirit of mod-
ern algebra? We suggest that Brahmagupta's results and applications
can be utilised to promote the mathematical maturity and creativity
of a bright student who has completed high-school algebra and is on
the verge of making the sharp (and rather abrupt) transition from the
\classical algebra" of school mathematics to the \modern algebra" or
\abstract algebra" of college mathematics.
Arrangement
Some readers would be well-informed about ancient Indian results
on indeterminate equations, some might not be. Again, while the pro-
fessional algebraist knows the subtleties involved in the bhavana, the
richness and depth of the result might not be apparent to all histori-
ans. Keeping in mind the possible diverse background of the readers,
the paper has been arranged as follows.
In Section 2 we highlight certain features of modern algebra implicit
in the bhavana and its applications. Section 3 proposes the central theme
of the paper | the pedagogic potential of aspects mentioned in Section
2 in the context of \abstract algebra".
The applications of Theorem 1 are revisited in Sections 4 and 5
chie y from the pedagogic perspective: Section 4 describes the immedi-
ate applications of the bhavana, mostly by Brahmagupta himself, while
Section 5 traces its in uence on the cakravala. Section 6 touches an-
other thought-provoking topic for the students: the possible genesis of
the bhavana.
Section 7 discusses the possible motivation for investigation of the
equation Dx2 +1 = y2 . Section 8 contains a few miscellaneous historical
remarks on the bhavana.
To convey to the readers some avour of the original \composi-
tion" (pun intended!), Appendix presents the transliteration of Brah-
magupta's verses describing Theorem 1. For readers uncomfortable with
Sanskrit, these verses are followed by a translation. As Brahmagupta's
language is somewhat cryptic, a more lucid version of Theorem 1, due
Brahmagupta's Bh
avana 81
to Bhaskara II, has also been included in Appendix.
2 Bhavana and Modern Algebra
Andre Weil asserted ([W1], p 231{232):
An understanding in depth of the mathematics of any given

period is hardly ever to be achieved without knowledge ex-
tending far beyond its ostensible subject-matter. More often
than not, what makes it interesting is precisely the early oc-
currence of concepts and methods destined to emerge only
later into the conscious mind of mathematicians; the histo-
rian's task is to disengage them and trace their in uence or
lack of in uence on subsequent developments.
The entire section of Brahma Sphut.a Siddhanta on the Varga-Prakr.ti

demonstrates a phenomenal wizardry in classical algebraic manipula-
tion at an early stage in the history of symbolic algebra. But more
exciting is the implicit occurrence of some of the simple but powerful
ideas which characterise modern algebra. We mention below three re-
lated concepts that are embedded in the bhavana: binary composition,
the multiplicativity of the norm function, the composition and prop-
erty of a binary quadratic form. If we leave aside the set-theoretic
language, the principle of binary composition is quite explicit | in
fact, in ancient Indian mathematics, bhavana means \composition"! We
shall also analyse Brahmagupta's methods and highlight in his approach
the dynamic anticipation of a certain mathematics culture that even-
tually evolved much later, albeit on a more broad and rm founda-
tion.
Binary Composition
Theorem 1 denes an intricate binary operation (samasa-bhavana)
(x1 ; y1 ; m1 ) (x2 ; y2 ; m2 ) = (x1 y2 + x2 y1 ; Dx1 x2 + y1 y2 ; m1 m2 )
on S = f(x; y; m) 2 Z Z Z j Dx2 + m = y2 g where Z denotes
the set of integers. This sophisticated idea of constructing a binary
composition on an abstractly dened unknown set is the quintessence
of modern \abstract algebra". The idea occurs in Brahmagupta's work
in a uid amorphous form | it had not been crystallised in a precise
set-theoretic framework.
Brahmagupta did not present the collection S as a single entity. But
he had envisaged the key ingredient of a modern abstract structure:
binary composition. If we exclude the four elementary arithmetic op-
erations on usual numbers, the bhavana is perhaps the rst conscious
construction of a binary composition. Further, the binary operation
is quite a complicated one: it involves two integral triples of unknown
roots. Recall that even basic symbolic computations with unknown roots
| treating them as if they were known quantities | is a fairly modern
approach that emerged during the investigations on the general polyno-
mial in one unknown.
Norm Function
The bhavana laws have an elegant interpretation in terms of the
norm function
p | a verypimportant concept in modern mathematics. Let
A = Z[ D] (= fb + a Dja; b
2 Zg). The norm function on A is the
map N : A ! Z dened by
p p p
N (y + x D) = (y + x D)(y x D) = y2 x2 D:
Brahmagupta's identity can be reformulated as the statement:
The norm function N is multiplicative, i.e.,

N ( ) = N ()N ( ) 8; 2 A.
Consider
p f; g : Z Z ! A dened by f (x; y) =
the bijective maps p
y + x D and g(x; y) = y x D. Then the set S = f(x; y; m) 2
Z Z Z j Dx2 + m = y2 g is simply the graph of each of the functions
Brahmagupta's Bh
avana 83
N Æ f and N Æ g. Let p : Z Z Z ! Z Z denote the map dened

by p(x; y; z ) = (x; y). Then pjS is a bijection from S to Z Z and
hence = f Æ pjS and = g Æ pjS are bijections from S to A. The
multiplicative structure bhavana on S is essentially the multiplication in
the ring A: the samasa-bhavana on S is obtained by transferring the ring
multiplication on A via and the antara-bhavana is obtained through
the conjugate .
It is possible that Brahmagupta discovered the identity through an
algebraic manipulation which was, in essence, the verication of the
structure preserving property of the norm function | the natural mul-
tiplication in A providing the precise formula (see the last paragraph of
Section 6).
Binary Quadratic Form

y2 Dx2 is a binary quadratic form with discriminant D. In the
language of quadratic forms, Brahmagupta's identity says that two such
forms (say y2 Dx2 ; v2 Du2 ) can be composed to yield another such
form with discriminant D in a new pair of variables (xv yu; yv Dxu).
Again recall that the term bhavana means \composition"! The theory
of composition of quadratic forms, an important and rich topic initiated
by Gauss and Dirichlet, is still an active area of research.
Here one may also recall that Fermat's researches in number theory
led him to a deep study of the binary quadratic form y2 2x2 which
must have resulted in his realisation of the far-reaching importance of the
equation y2 Dx2 = 1. As is well-known, Fermat not only investigated
Pell's equation but also inspired others to take interest in the problem.
Before mentioning another aspect of Theorem 1 in the context of
quadratic forms, we recall a few denitions. Two quadratic forms f (x; y)
and g(x; y) over an integral domain K are dened to be equivalent if
there exists a homogeneous linear change of variables which takes the
form f to the form g; more precisely, if there exists an invertible matrix
A with entries in K such that g(x) = f (Ax) where x denotes the vector
(x; y)T . Thus f and g are equivalent if if there exist a; b; c; d in K such
that g(x; y)=f (ax + by; cx + dy) and ad bc is a unit in K .
An element c in K is said to be represented by a binary form f (x; y)
if the value c is attained by f , i.e., if there exist a; b 2 K such that
c = f (a; b). A quadratic form f is said to be strongly multiplicative if f

is equivalent to cf for every unit c represented by f .
Theorem 1 can be viewed as the result:
The binary form y2 Dx2 is strongly multiplicative (over

rationals).
For, if f denotes the form y2 Dx2 , c a non-zero number represented

by f , say c = b2 Da2 , and g = cf , then g(x; y) = (b2 Da2 )(y2
Dx2 ). Now Brahmagupta's identity prescribes the substitution that
has to be made to obtain g(x; y) = f (u; v): namely, the homogeneous
linear transformation given by u = bx + ay; v = Dax + by which is
invertible (since the determinant of Brahmagupta's transformation is
c 6= 0). This version of Theorem 1 was generalised in 1965 by A. Pster
using, what are now called, \Pster forms". Pster's discovery opened
up new directions in the theory of quadratic forms.
In his recent research monograph ([O]) in this area, Manuel Ojan-
guren begins Chapter 5 by quoting Brahmagupta's original Sanskrit
verses describing Theorem 1; the chapter itself is titled Also sprach
Brahmagupta5 . The rst result in the chapter (Lemma 5.1) states:
If K is a eld of characteristic dierent from 2, and c is a

non-zero element of K represented by the quadratic space
< 1; a >, then < 1; a > is isometric to < c >< 1; a >.
Readers familiar with the relevant terminology can see that Lemma 5.1
is a version of the fact that y2 Dx2 is strongly multiplicative. M.
Ojanguren referred to Lemma 5.1 as \Brahmagupta's lemma" ([O], p
55).6 Theorem 5.2 in ([O], p 55) states the generalisation of Lemma 5.1
by Pster: Pster forms are multiplicative.
In English: \Thus Spake Brahmagupta"!

5
Just before the statement of Lemma 5.1, Ojanguren writes ([O], p 54): \In [5,
6
Ch.18] Brahmagupta proves, in the slightly dierent formulation quoted above, the
following lemma.". With a quiet humour characteristic of Ojanguren, Item 5 is
listed in \References" as: 5. Brahmagupta: Brahmasphut.asiddhanta, Bhnmala,
628 (preprint).
Brahmagupta's Bh
avana 85
Quest for General Algebraic Principles

One profound message of modern algebraic research is that an iso-
lated problem can often be handled more eectively by viewing it as
part of a larger set-up. A deep understanding of the general picture
not only generates additional techniques for the special situation, it
also leads to new creations and discoveries whose overall impact could
be of much greater value than a mere solution to the original prob-
lem. This realisation has heralded a new attitude in algebra: the search
for general principles. Lagrange (1770 CE) pioneered the trend. The
culture of generalisation has been diligently pursued in 20th century
algebra.
Now let us consider certain aspects of Brahmagupta's ancient work
in this light. The example D = 92, mentioned in Section 1, serves as an
illustrative model.
(i) While trying to solve a specic hard problem Dx2 + 1 = y2 (in two
variables), Brahmagupta undertook a bold and farsighted exploration of
the general picture: the solution space of Dx2 + m = y2 (in three vari-
ables). In the process he discovered and extracted an important general
and abstract principle (Theorem 1), and made a clear enunciation of
this principle for posterity. It is amazing that in an attempt to solve an
indeterminate equation in two variables, a seventh-century mathemati-
cian thought of constructing, what amounts to, an intricate abstract
structure on the solution space of an equation in three variables. This
is an original attitude in mathematics the like of which was not to be
seen for the next 1000 years.
(ii) The power of the general principle can be seen from the deceptive
ease with which it immediately provides the solution of a diÆcult equa-
tion like 92x2 + 1 = y2 (Recall Example 1).
(iii) Apart from providing partial solutions to the original problem (see
Section 4), the samasa-bhavana principle contained the key to the the
subsequent complete solution (see Section 5).
(iv) In retrospect, one realises that the samasa-bhavana had also opened
up new possibilities discussed in Section 2. However, it was far too
ahead of the times to have immediate fruition beyond the solution of
Pell's equation. Its true potential began to be harnessed only after its
rediscovery in the 18th century. This general principle turned out to be
\a theorem of capital importance" and even a starting point of an area

in algebra.
A Caveat
Sometimes the brilliance of an algebraic research lies in its opening
up of new and unexpected horizons with immense possibilities through
surprisingly simple innovations. Ironically, the very simplicity of the
work tends to hinder or distract later generations from a deep percep-
tion of its true worth. A casual observer could very well fail to fully
appreciate the richness and magnicence of the ideas encapsuled in the
construction of a mere 2-step solution of 92x2 + 1 = y2 . Further, once
the mathematical community gets accustomed to an original idea like
the bhavana, it becomes all the more diÆcult to fathom the greatness
of the discovery. Historians need to be aware of this intrinsic risk of
missing the real depth and signicance of Brahmagupta's work on the
Varga-Prakr.ti.
3 Pedagogic Potential of Bhavana

We rst clarify the aspect of mathematics education which we shall
be addressing. Our target is the (potentially) research-oriented higher-
secondary or rst-year college student who is about to learn modern
algebra. We focus on two related aspects: (1) promotion of a culture of
creativity in the introduction to abstract structures and (2) promotion
of \the art of discovery".
Training in Abstract Algebra

During his rst encounter with \abstract algebra", the student of
mathematics is suddenly confronted with a formidable edice of ax-
iomatic structures. He learns systematically, but more or less passively,
a large number of denitions and basic properties. For the student,
this \algebra" has not much apparent connection with the manipulative
high-school algebra that he hitherto enjoyed. In fact, with passage of
time, he even tends to lose his classical manoeuvring skill as his train-
ing usually focusses, almost exclusively, on understanding of abstract
Brahmagupta's Bh
avana 87
structures. During this transitional period, the student has hardly any
chance to mentally participate in the process of discovery of the basics
of modern algebra | he has to patiently learn a radically new approach
to mathematics suspending his creative impulse. The situation is all the
more grave for the large number of students who do not have scope for
interaction with creative algebraists.
The current group-ring-eld approach of abstract algebra is undoubt-
edly neat, elegant and has enormously simplied mathematics. But sim-
plicity tends to hide mathematical subtleties. Abrupt introduction to
abstract structures, completely divorced from their original contexts,
could sti e the natural growth of thought-process and promote a sort of
mechanical pursuit of forms missing the substance. The achievement of
extreme elegance and simplicity, therefore, has a potential risk for new-
generation learners. They tend to approach algebra with a mechanised
mindset for too long.7
Is it possible to present before a high-school student a non-trivial but
accessible algebraic work of a mathematical genius which can creatively
orient him to the principles of modern algebra through the language of
high-school algebra?
We aÆrm that, before the student gets lost in the elaborate maze
of groups, rings, elds, vector spaces, etc, Brahmagupta's results can
be used to informally introduce him to the essence of abstract algebraic
ideas. An imaginative and eective use of Theorem 1 and Example 1
can convey to the fresh student the power of \binary composition" much
more deeply and rapidly than any of the standard pedagogic approaches.
It could be inspiring for him to realise that the simplicity of the solution
to Example 1 has its secret in the monoid structure of (S; ) of Sec-
tion 2 which in turn has its root in the natural ring structure of the set
A in Section 2. He could be excited by the observation that, through
samasa-bhavana, the set f(x; y; m) 2 Q Q QjDx2 + m = y2 g forms a
group or the later realisation that f(x; y) 2 N NjDx2 + 1 = y2 g forms
7
As the inevitable perils are gradually emerging, there is now an increasing trend to
supplement the organised and rapid presentation of abstract algebra in its generality
with some discussion on the genesis of the subject. It is a welcome sign that attempts
are being made in recent texts to impart some familiarity with the original approach
of Galois supplementing the standard 20th century presentation. But considerable
preparation is needed before one can get introduced to fragments of the thought-
process of a Lagrange or a Galois.
a cyclic monoid. What is more, the unexpected invocation of abstract

algebraic technique for a concrete problem by an ancient mathematician
could infuse in him a dynamic and creative vigour in his formal study of
abstract algebra. Thus, through this deep but easily accessible work of
a Master, the perceptive student can get an early exposure to modern
sophistication, see a natural application of abstract algebraic principles,
develop a air for exploring worthwhile generalisation, and truly imbibe
the spirit of the abstract algebra culture without getting swept away by
its formalism.
The Art of Discovery

In his plenary lecture at the ICM (1978), Andre Weil quoted the
following statement of Leibniz on history of mathematics:
Its use is not just that History may give everyone his due
and that others may look forward to similar praise, but also
that the art of discovery be promoted and its methods known
through illustrious examples.8
Weil added ([W1], p 229):
Deviating only slightly from Leibniz, we may say that its rst
use for us is to put or to keep before our eyes \illustrious
examples" of rst-rate mathematical work.
For our target students, the developments on Diophantine equations

in ancient India | the kut..taka of Aryabhat . a and his successors, the
bhavana of Brahmagupta and the cakravala of Jayadeva-Bhaskara |
could be utilised to promote the \art of discovery". Especially, Brah-
magupta's bhavana would serve not only as an \illustrious example of
rst-rate mathematical work" but also as an early example of modern
methods and techniques and of the harmonious blending of the classical
and modern algebra styles. It would help students realise how the seeds
of various abstract structures lie in earlier classical algebra.
Brahmagupta's bhavana is the rst known instance of an involved
abstract algebraic thinking. Its pedagogic value stems from the fact
8
Both the original and the above translated version occur in ([W1], p 227).
Brahmagupta's Bh
avana 89
that it is a self-contained gem. When the same ideas resurfaced in

Europe after more than a thousand years, they came as part of big
theories developed by a large number of mathematical giants. The raw
student needs a long time to assimilate even the subsequent simplied
versions of these theories. By contrast, the applications and in uence of
bhavana (discussed in next two sections), especially the cakravala, are
ready-made materials for promoting the art of discovery. The student
need not have prior preparation and the volume of work he has to dwell
on is well within a manageable limit.
Note that the solution of Pell's equation is taught quite neatly in
several texts on classical algebra and elementary number theory. But
those presentations oer a nished product to the students. Continued
fraction is dened, its properties are systematically established and, in
due course, the solution of Pell's equation is described. The approach
has its usefulness but does not serve the purpose alluded to by Leibniz
and Weil.
To look at the pedagogic aspect from another angle, a research scholar
gets trained from two sources: text books, research papers. They serve
complementary purposes. By and large, school and college students have
access only to the former. Can one give them some material which fulls
one aspect of the role played by research papers at a later stage?
We aÆrm that robust expositions on the kut..taka, bhavana, cakravala,
etc, in the style of modern expository research papers, can play that
role. A preliminary attempt was made in that direction by the present
author in ([Du1], [Du2]) for a journal on science education.9 Each of
these articles contained an introductory part meant to excite or inspire
the student on the topic; but the major part of the articles tried to give
some avour of research culture.
Besides, to use a phrase of Weil ([W1], p 231) made in a slightly
dierent context, the students could also gain by the encouragement to
recognize mathematical ideas in obscure or inchoate form,

and to trace them under the many disguises which they are
apt to assume before coming out in full daylight.
9
The pedagogic usefulness of bhavana was also perceived by M. Ojanguren. After
receiving a draft of ([Du2]), Ojanguren mentioned in an e-mail to the author that he
thought of using the results in an elementary course for school teachers.
4 Immediate Applications of Bhavana

by Brahmagupta
Brahmagupta's presentation of the results on Varga-Prakr.ti resembles a
typical modern arrangement where one rst develops the fundamental
theory or principle that emerges out of the investigations into a problem,
and then records the solution to the original problem(s) as one applica-
tion of the basic theory. Thus, after announcing Atha Varga Prakr.tih.,
Brahmagupta immediately states Theorem 1 | the cornerstone of the
section on Varga-Prakr.ti | in Verses 64{65 of ([1], Ch. 18). This is
followed by a few general results on the varga-prakr.ti which are useful
oshoots of Theorem 1: construction of innitely many rational solu-
tions of Dx2 + 1 = y2 (Verse 65), generation of innitely many integer
solutions of Dx2 + m = y2 from a given one (Verse 66), derivation of
solutions of Dx2 +1 = y2 from solutions of Dx2 4 = y2 (Verses 67{68),
etc. The results are then illustrated by several concrete numerical ex-
amples. The example D = 92 (Example 1) is mentioned in Verses 71{72.
Application 1: Rational Solutions of Dx2 + 1 = y2 .

The verses 64{65 in ([1], Ch. 18) contain, apart from Theorem 1,
the following two corollaries (see Appendix).
Corollary 1 (Brahmagupta)
If Dp2 + m = q2 , then D(2pq)2 + m2 = (Dp2 + q2 )2 .
Corollary 2 (Brahmagupta)
If (p; q) is a root of Dx2 c = y2 , then ( 2pq ; Dp c+q ) is a root of Dx2 +1 =
2 2
c
y2 .
In fact, immediately after stating their respective verses describing

Theorem 1, Brahmagupta, Jayadeva ([Sh], p 8) and Bhaskara II ([Ba],
p 23) describe non-trivial rational solutions of Dx2 + 1 = y2 .
This is actually an easy application | the full strength of Theorem
1 is not really needed. One can always choose a positive integer p (for
instance, p = 1) and a positive integer q such that q2 > Dp2 and put
m = q2 Dp2 to get Dp2 + m = q2 and then apply Corollary 2. The
Brahmagupta's Bh
avana 91
latter needs Corollary 1 which could be discovered directly. In fact,

nding a non-trivial rational solution of the Pell's equation could itself
be set as a reasonable challenge to the student inciting him to directly
discover Corollaries 1 and 2.
Brahmagupta simply made a terse statement amounting to Corollary
2 (see Appendix) but did not bother to give the trivial explanation as
to how to obtain some integer-triple (p; q; m). As in modern times, one
cannot always expect the Masters, who have to oer volumes of original
materials, to spell out all obvious details. (See Remark 4 in Appendix.)
But as the history of algebra from the 1870s will testify, clarity of
exposition can do wonders for the rapid progress of a subject. In an-
cient Indian algebra, the terseness of Brahmagupta was balanced by
pati (1039)
the clearer accounts of later expositors. For instance, Sr
described the method for obtaining rational solutions as follows:
\Unity is the lesser root. Its two squares [set at two places] are [each]
multiplied by the prakr.ti [and the product is] decreased [by the prakr.ti
and] increased by a [suitable] interpolator whose square-root will be the
greater root. From these two, two roots are obtained by bhavana. There
will be an innite [set of two roots]."10
In other words, applying Brahmagupta's bhavana on two copies of
the identity
D:12 + (q2 D) = q2 ;
one gets a rational solution
2q q2 + D
x= 2 ; y= 2
q D q D
of Dx2 + 1 = y2 . Innitely many roots can be obtained by varying q
(also by repeated use of bhavana).
The above solution was also described by Bhaskara II (1150),
Narayan.a (1350) and later writers like J~nanaraja (1503), Kr.s.n.a (1600)
and Kamalakara (1658). To quote Bhaskara II:
\Divide twice an optional number by the dierence between the
square of that optional number and the prakr.ti. This [quotient] will
10
Verse 33 of Chapter 14 (on algebra) of the astronomy treatise Siddhantasekhara;
quoted (in translation) in ([Si], p 40; [DS], p 153, 150).
be the lesser root [of a varga prakr.ti] when unity is the additive. From
that [follows] the greater root."11
Now the application of bhavana on the identity D:12 + q2 D = q2 to
obtain rational solutions might appear to be a relatively minor feat in the
history of Indian algebra. But the clear formulation of this identity, the
display of the action of bhavana on it, and explicit mention of division
by q2 D might have facilitated the invention of the cakravala (see
next section). We can also expect that, during actual computation,
there would be a tendency to choose q so as to avoid division by large
numbers. This leads to a q for which jq2 Dj is the minimum | another
component of the complex cakravala. In this connection, one is reminded
of Weil's caveat ([W1], p 235):
It is also necessary not to yield to the temptation (a nat-

ural one to the mathematician) of concentrating upon the
greatest among past mathematicians and neglecting work of
only subsidiary value. ... historically it can be fatal, since
genius seldom thrives in the absence of a suitable environ-
ment, and some familiarity with the latter is an essential
prerequisite for a proper understanding and appreciation of
the former. Even the textbooks in use at every stage of math-
ematical development should be carefully examined in order
to nd out, whenever possible, what was and what was not
common knowledge at a given time.
However not enough of historical materials have been found to ascertain
the actual sequences in emergence and ow of ideas. (See Section 8.)
Application 2: Generation of Innitely Many Roots

Verse 66 in ([B], Ch 18) can be stated as:
Theorem 2 (Brahmagupta) If the equation Dx2 + m = y2 has one

positive integral solution, it has innitely many.
The general equation Dx2 + m = y2 need not have any integral

solution (students could be encouraged to nd simple examples like
11
Verse 73 of Bjagan. ita ([Ba], p 23); quoted (in translation) in ([DS], p 154).
Brahmagupta's Bh
avana 93
D = 3; m = 2). However, when it does have an integral solution, Brah-

magupta's verse 66 prescribes the samasabhavana to use that solution
and a non-trivial integral solution of Dx2 + 1 = y2 to generate innitely
many integral solutions of Dx2 +m = y2 . In particular, from one positive
integral solution of Dx2 + 1 = y2 , one gets innitely many. The verses
65{66 tacitly describe innitely many rational solutions of Dx2 +1 = y2 .
pati (quoted earlier) and
This consequence is mentioned explicitly in Sr
Bhaskara II.
It would be interesting for students to observe that the method for
generating arbitrarily large solutions of Dxp2 + 1 = y2 was useful for
determination of rational approximation to D. If Da2 + 1 = b2 , then
b p b2 Da p
2 1
p :
D= =
a a(b + a D) a(b + a D)
Thus, for apsuÆciently large solution (a; b), ab will be a good approxi-
mation for D. This application was explicitly stated by the algebraist
Narayan.a around 1350 ([D2], p 187{188). By that time, the cakravala
method for determination of one (in fact, the minimum) positive integral
solution had already been invented |- Narayan.a himself was an expos-
itor of both the bhavana and the cakravala. Narayan.a illustrated the
p q 1 approximations
method for successive
p of surds by two numerical exam-
ples: 10 and 5 . For 10, he mentioned the rational approximations
19 , 721 and 27379 . For D = 10, since 9 is the nearest square, one has the
6 228 8658
initial triple (1; 3; 1). Now,
(1; 3; 1) (1; 3; 1) = (6; 19; 1);
(6; 19; 1) (6; 19; 1) = (721; 228; 1);
(6; 19; 1) (721; 228; 1) = (27379; 8658; 1)
q
and hence the three successive fractions of Narayan.a. The case of 15 is
similar. This method of getting successively closer approximations was
restated by Euler in 1732 ([D2], p 188).
Application 3: Integer Solution of Dx2 + 1 = y2 (Partial)

Verses 67{68 in ([B], Ch 18) give the following consequences of the
samasabhavana.
Theorem 3 (Brahmagupta)
(i) If Dp2 + 4 = q2 , then ( 12 p(q2 1); 21 q(q2 3)) is a solution of
Dx2 + 1 = y2 .
(ii) If Dp2 4 = q2 , and r = 21 (q2 +3)(q2 +1), then (pqr; (q2 +2)(r 1))
is a solution of Dx2 + 1 = y2 .
Since (1)2 = 1 and (2)2 = 4, it follows that from any positive

integer solution of Dx2 + m = y2 , where m 2 f 1; 2; 4g, one can
derive a positive integer solution of Dx2 + 1 = y2 by repeated use of
pati
the samasabhavana. This consequence was explicitly recorded by Sr
([Si], p 40 ;[DS], p 157). Brahmagupta applied it to numerical examples.
We quote two of them ([B], Ch 18, Verses 71{72, 75).
Example 2 (Brahmagupta) Solve, in integers, 13x2 + 1 = y2 .
Solution. 1312 4 = 32 . Now Theorem 3 yields the solution (180; 649).
Brahmagupta mentioned the equations 13x2 + 300 = y2 and 13x2

27 = y2 . By inspection, one has solutions (10; 40) and (6; 21) respec-
tively. Composing with (180; 649) one gets larger solutions for the two
equations.
Example 3 (Brahmagupta) Solve, in integers, 83x2 + 1 = y2 .
Solution. 83 12 2 = 92 . (1; 9; 2) (1; 9; 2) = (18; 164; 4). As the

quantities are all even, dividing by 22 , one gets the triple (9; 82; 1), i.e.,
the solution (9; 82).
Brahmagupta stated the formulae for D = 4 in a generality re-
quired for the harder case: p odd. Students could be encouraged to
deduce Theorem 3 from Theorem 1 and to nd simpler formulae when
p is even as also the formulae for D = 1 and D = 2. They should be
able to discover results like (p; q; 1) ) (2pq; 2q2 + 1; 1), (p; q; 2) )
(pq; q2 1; 1); and, for p even, (p; q; 4) ) (p=2; q=2; 1), (p; q; 4) )
(pq=2; q2 =2 + 1; 1). Brahmagupta used these simple formulae in his ex-
amples though he did not state them explicitly.
Brahmagupta's Bh
avana 95
Theorem 3 already enabled Brahmagupta to solve various diÆcult

cases like D = 83 or D = 92. In the next section, we shall see, with the
example of the famous case D = 61, how it accelerated the subsequent
cakravala algorithm.
5 Role of Bhavana in Cakravala

Brahmagupta attained partial success in his attempt to solve Dx2 + 1 =
y2 . His results can directly be applied only to those specic values of D
for which one gets some integer triple (p; q; m), where m 2 f1; 2; 4g,
through inspection or clever manipulations. Even when such a triple
becomes available for a special D, the methods lead to some positive
integral solution of Dx2 +1 = y2 | but not necessarily the minimum and
hence although the samasabhavana would fetch innitely many integral
solutions, it might not fetch all integral solutions. For instance, for
D = 3, an application of Brahmagupta's formula (for m = 4) on the
simple identity 3 22 + 4 = 42 yields the triple (15; 26; 1); but the
minimum triple (for D = 3) is (1; 2; 1).
But Brahmagupta's partial solution, apart from being a remarkable
landmark by itself, was also a signicant step towards the celebrated
cakravala algorithm | a perfect method (free from trial-and-error) for
obtaining, for any D, all positive integral solutions of Dx2 + 1 = y2 .
Actually the set of positive integral solutions of Dx2 + 1 = y2 forms
a cyclic monoid under the operation samasabhavana and the cakravala
method fetches the minimum solution, i.e., the generator of the above
monoid. Thus, after the application of cakravala, the samasabhavana
generates all integral solutions from the minimum one.
Historians have paid glowing tributes to the cakravala. H. Hankel
exclaimed12 :
It is beyond all praise; it is certainly the nest thing achieved
in the theory of numbers before Lagrange.
More recently, Selenius13 opined ([Se], p 180):
12
Zur Geschichte der Mathematik in Alterthum und Mittelalter (Leipzig, 1874), p
202; quoted in ([Bg], p 28; [Se], p 170).
13
C.O. Selenius gave a formal justication of the cakravala in terms of his gener-
The old Indian cakravala method for solving the mathemati-

cally fundamental indeterminate varga-prakr.ti equation [Dx2
+ 1 = y2 ] was a very natural, eective and labour-saving
method with deep-seated mathematical properties. ... More
than ever are the words of Hankel valid, that the cakravala
method was the absolute climax (\ohne Zweifel der Glanz-
punkt") of old Indian science, and so of all Oriental math-
ematics.... no European performances in the whole eld of
algebra at a time much later than Bhaskara's, nay nearly up
to our times, equalled the marvellous complexity and inge-
nuity of cakravala.
The earliest known author on the cakravala is Ac arya Jayadeva.14

His verses on cakravala have been quoted in ([Sh]). Bhaskara II 15 (1150)
too described the algorithm in ([Ba], Verse 75).16
The underlying idea of this brilliant algorithm can be put as follows:
From a triple (pn ; qn ; mn ) such that jmn j is \small", one has to construct
a triple (pn+1 ; qn+1 ; mn+1 ) with jmn+1 j \small" eventually arriving at
(p; q; 1).
As we remarked in the preceding section, triples of the type (1; y; y2
D) seem to have been \in the air" among Indian algebraists who ex-
pounded on Brahmagupta's work. The authors of cakravala begin with
an initial triple (p0 ; q0 ; m0 ) of this type: p0 = 1, q0 a number for which
jq02 Dj is minimum, and m0 = q02 D.
Now having arrived at an integer triple (pn ; qn ; mn ), it would be nat-
ural to compose it with the (variable) triple (1; y; y2 D) and explore the
outcome. The samasabhavana yields an integer triple (pn y + qn; Dpn +
alised continued fraction expansion. He gave a talk on it in a Short Communications
session of the \Algebra and Theory of Numbers" section of the 1962 ICM at Stock-
holm.
14
Jayadeva's verses on the complete solution of the indeterminate equation Dx2 +
1 = y 2 have been quoted in the text Sundar of Udayadivakara composed in 1073
CE. This text itself was discovered only in 1954 by K.S. Shukla. Nothing is known
so far about Jayadeva except for the obvious fact that he lived before 1073 CE. The
algebraist Jayadeva is not to be confused with the Vais.n.ava poet of the 12th century
who composed Gta-Govinda.
15
Also known as Bhaskaracarya.
16
Bhaskara's version has been quoted in the paper \Algorithms in Indian Mathe-
matics" by M.S. Sriram in this volume.
Brahmagupta's Bh
avana 97
qny; mn (y2 D)). Dividing by mn 2 (another natural idea given the ear-
lier emphasis on methods for obtaining rational solutions) one obtains
a rational triple (pn+1 ; qn+1 ; mn+1 ) = ( pnm y+qn ; Dpn +qny ; y2 D ). Now
n mn mn
the above triple would be an integer triple if y can be so chosen such
that pn+1 (and hence qn+1 ; mn+1 ) becomes an integer. This amounts
to nding integer solutions of the equation mn x pn y = qn . The lin-
ear indeterminate equation had been extensively discussed by Indian

algebraists from the time of Aryabhat . a (499 CE) and they knew the
complete solution to the problem. Now among the innite solutions,
one would obviously choose the solution for which jy2 Dj would be
minimum so that jmn+1 j is minimised (as desired). And that is pre-
cisely the prescription of the cakravala! Andre Weil's remark can hardly
be bettered ([W2], p 22):
As is the case with many brilliant discoveries, this one [cakra-

vala] can be seen in retrospect as deriving quite naturally
from the earlier work [samasabhavana].
Weil had written in a dierent context ([W1], p 231):
...in large part the art of discovery consists in getting a rm

grasp on the vague ideas which are \in the air", some of them
ying all around us, some (to quote Plato) oating around
in our own minds.
It would be instructive for the students to see that Brahmagupta's novel

ideas contained the key to the discovery of the cakravala and that the
astonishing algorithm was possibly the result of a marvellous interplay
between the two great preceding works | the kut..taka and the bhavana
| perhaps catalysed by the clarity brought about by expositions on
lesser results like rational solutions of varga-prakr.ti.
Selenius expressed doubt ([Se], 176{177) whether the original in-
ventors of cakravala actually thought in terms of the bhavana. In the
absence of historical details, one can never be sure regarding the chain
of thoughts involved in the discovery. But this reconstruction is, in
any case, useful for our pedagogic purpose. It is historically plausible
given that the verses on cakravala, in both Jayadeva and Bhaskara, are
preceded by verses on the bhavana; and that the bhavana was highly
regarded by Indian algebraists including Jayadeva and Bhaskara |

Jayadeva remarking that the bhavana pervades numerous algorithms.
The alternative interpretation of Selenius, in the framework of contin-
ued fractions, too has its pedagogic merits in promoting the culture of
creativity.
In retrospect, the original cakravala can be simplied by avoiding the
kut..taka.17 Denote by (xn ; yn ) the solution of mn x pn y = qn for which
jy2 Dj is minimal. Then yn+1 +yn 0 (mod mn). Thus the sequence yn
can be inductively constructed as follows: having constructed yn , yn+1
is dened to be an integer y satisfying the simple congruence y yn
(mod mn ) for which jy2 Dj is minimal.
Simplications are observed after initial breakthroughs. For the pur-
pose of promoting the art of discovery, it is desirable to highlight the
initial thought-process which actually achieved the breakthrough and
which tends to get camou aged in later simplications. There is a cer-
tain richness of thought in the cute approach through the blending of
kut..taka and bhavana which would be good for the students to imbibe.
Of course, they should also see the subsequent simplication which too
serves a lesser, but still important, pedagogic purpose.
The bhavana also aids the cakravala in rapidly arriving at the minimum
solution. Let us recall the famous example of Bhaskara ([Ba], Verse 76):
Example 4 (Bhaskara II) Solve, in positive integers, 61x2 + 1 = y2 .
Solution. As 64 is the perfect square nearest to 61, we have the initial

triple (1; 8; 3). Now one nds positive integer y for which y+8
3 is an
integer and jy2 61j is minimised. Clearly y = 7. Now
7+8 61 + 8 7 72 61
= 5; = 39; = 4:
3 3 3
Thus we have the second triple (5; 39; 4). Now rather than continuing
the cakravala, one can apply Brahmagupta's formula (Theorem 3) on
the triple (5; 39; 4) to immediately get the minimum positive solution
(226153980; 1766319049) of 61x2 + 1 = y2 .
17
This observation is mentioned in the pioneering work of A.A. Krishnaswamy
Ayyangar ([K], p 234).
Brahmagupta's Bh
avana 99
As displayed in the two tables in Sriram's paper in this volume, the

Brouncker-Euler algorithm needs 22 steps while the cakravala algorithm
(without the bhavana) needs 14 steps. Thus, although the \cyclic" char-
acter of cakravala gets destroyed (making theoretical justication more
complicated), the bhavana provides a very eective short-cut for the
purpose of practical numerical computations.
In 1657, Fermat, in an eort to arouse the interest of contemporary
mathematicians to number theory, issued a few challenge problems with
special emphasis upon the problem of nding integer solutions to the
equation Dx2 +1 = y2 , unaware of the works of Brahmagupta-Jayadeva-
Bhaskara. Regarding the challenge, Andre Weil remarked ([W2], p 81{
82):
What would have been Fermat's astonishment if some mis-

sionary, just back from India, had told him that his problem
had been successfully tackled there by native mathematicians
almost six centuries earlier!
It is interesting to note that Fermat specically mentioned the case

D = 61 which was used by Bhaskara to illustrate the cakravala.
6 Possible Genesis of the Bhavana

It would also be useful for a budding researcher to try to imagine, in
retrospect, how a great mathematician could possibly have arrived at
his discovery. Even if the actual thought-process had been completely
dierent, meaningful speculations can serve an important purpose by
providing fresh mathematical insights to students. We brie y summarise
the discussion on the genesis of bhavana made in this light in ([Du2], p
20{21).
It is well-known that the Vedic re-altar construction problems often
involved the so-called Pythagorean triples | integer triples (x; y; z ) sat-
isfying x2 + y2 = z 2 . The Sulba
Sutras mention several such triples. It is
possible that the Sulba authors were aware of the formula (2mn)2 +(m2

n2 )2 = (m2 +n2 )2 .18 It is likely that, while handling Pythagorean triples,
18
This identity can be obtained from the algebraic formulae pq = ( p+2 q )2 ( p 2 q )2
ancient Indians investigated the more general indeterminate equation

z = x2 + y2 and discovered the algebraic formulae19
(x1 2 + y1 2 )(x2 2 + y2 2 ) = (x1 x2 y1 y2 )2 + (x1 y2 x2 y1 )2 : ()
Now Brahmagupta explicitly gave the integer solution (2mn; m2

n ; m2 + n2 ) of the equation x2 + y2 = z 2 ([B], Chap 12, Verse 33;
2
[C], p 306). It is thus possible that Brahmagupta had also examined
x2 + y2 = c, arrived at the result () and, then, while considering the
more diÆcult Dx2 + m = y2 , he could have been on the lookout for
analogous formulae!
p readily follows from () by replacing x1 by
pBrahmagupta's identity
x1 D and x2 by x2 D. Weil remarks ([W2], p 14) that \this could
not have been fully realized until the eighteenth century". Complex
numbers appeared in formal mathematics only in the 16th century CE
and its vigorous use can be seen only from the 18th century.
True, that validity of complex numbers had not been recognised in
ancient Indian mathematics. But it is still possible that Brahmagupta
might have heuristically used them as a secret trick | an inspired \rough
work" | for guessing the magic identity which could then be veried
by recognised algebraic manipulations.20
Brahmagupta's identity (y2 Dx2)(t2 Dz 2) = (yt + Dxz )2 D(yz +
tx)2 may also be obtained by p splittingp the terms y2 Dx2, t2 Dz p 2,
observing the identity (y + x D)(t + z D) = (yt + Dxz )+(yz +pxt) D,
and multiplying this identity by the conjugate identity (y x D)(t
as well as ka2 = ( k+1
2
)2 a2 ( k 2 1 )2 a2 (putting p = m2 ; q = n2 ; a = n2 ; k = m2 =n2
and clearing denominators); and these two formulae are involved in the Sulba trans-
formation of a rectangle into a square (of equal area) and Katyayana's ingenious rule
for combining k squares into a single square. (See [D3], p 133{136.)
19
This surmise nds some support from parallel developments in Greek mathemat-
ics. The Greeks too were fascinated by the Pythagorean triples | both Euclid (300
BCE) and Diophantus (c. 250 CE) gave the general solution to the Pythagorean
equation | and there are indications that Diophantus was familiar with the identity
(*). For, Diophantus mentioned that 65 is expressible as sum of two dierent squares
in two dierent ways since 65 is the product of 13 and 5, each of which is a sum of
two squares ([W2], p 10{11).
20
To give an analogy, recall the non-commutative ring-theoretic result: \If 1 ab is
invertible, then so is 1 ba." and consider the eective trick of guessing the inverse
of 1 ba through power-series expansion of (1 ba) 1 even when the latter does not
make formal mathematical sense.
Brahmagupta's Bh
avana 101
p p
z D) = (yt + Dxz ) (yz + xt) D. This was the approach of Euler
([W2], p 15). One wonders if Brahmagupta's thought-process too had
taken a similar route! That would amount to an implicit handling of
the norm function mentioned in Section 2.
7 Possible Motivation for Varga-Prakr.ti

Pell's equation is the most famous Diophantine equation after x2 + y2 =
z 2 . Its importance in the study of binary quadratic forms and real
quadratic elds is well-known. But what prompted Brahmagupta's vig-
orous pursuit of Dx2 + 1 = y2 in the seventh century?
Dierent explanations can be oered on the basis of the scarce his-
torical materials available | each instructive in its own way.
(i) Astronomy
In Laghu-Bhaskarya (Chap 8; Verse 18) of Bhaskara I, there is a
problem from astronomy involving the simultaneous equations
8x + 1 = y2 ; 7y2 + 1 = z 2 :
By inspection, one readily gets an integral solution y = 3; z = 8 for
the second equation. Did such problems provide the original motivation
for exploring a systematic method of determination of integer solutions
to the general varga prakr.ti? Possible. In his commentary (Sundar)
on the Laghu-Bhaskarya, Udaydivakara (1073) prescribes Jayadeva's
general method for solving 7y2 + 1 = z 2 ([Sh], p 4).
But the central problems in the vast canvas of ancient Indian astron-
omy do not seem to seriously involve the Pell's equation. It is true that
the examples of Brahmagupta on varga prakr.ti (as also on other topics
in algebra) are put in the language of astronomy. Given the exalted
status of astronomy, it was natural to put the illustrative examples in
a glamorous astronomy framework for greater impact. But consider-
ing the prevailing interests in astronomy, it looks more likely that these
examples were invented to illustrate the rules; rather than the rules hav-
ing emerged from an attempt to solve those examples. Astronomy alone
might not have provided suÆcient impetus for the sustained eort that
would have gone into the partial (and later the general) solution to such
a diÆcult problem.
p
(ii) Rational Approximation to D

p
Aryabhat . a's method for computing D, when D is a perfect square,
is well-known ([DS], p 170{175) | a modied version of his algorithm
is taught in school arithmetic. But right from the Vedic times, there
have also been eorts p to obtain convenient, reasonably accurate, rational
approximations to D when D is not a perfect square. In the Sulba
Sutras (c. 800 BCE), the most ancient mathematics treatises available,
7 , 17 , 577 have been used as approximations for p2 ([D3], p 202; [Di],
5 12 408
p 341; [Sa], p 18). These three fractions may be interpreted as arising
out of solutions of 2x2 1 = y2 ; for 2:52 1 = 72 , 2:122 + 1 = 172 and
2:4082 + 1 = 5772 . 21 In fact, they are respectively the third, fourth
p and
eighth convergents of the simple continued fraction expansion of 2. 22
It is possible that, at some stage, Indians noticed that a suÆciently
large integer solution (a; b) of Dx2 + 1 = y2 (or even Dx2 c = y2 where
c is pa relatively small positive integer) will give a good approximation ab
for D and that this realisation triggered a quest not only for a method
of solving the varga-prakr.ti but also a device for generating arbitrarily
large solutions from a given one. The hypothesis may have a pedagogic
usefulness, but there does not seem to be any concrete evidence so far to
indicate that this was indeed the original motivation for Brahmagupta. p
Narayan.a, who explicitly used the varga-prakr.ti for approximating D
21
One may recall p here that Archimedes (287-212 BCE) gave the approximations
265
and 1351 for 3 and that, in his commentary on Archimedes, Eutocius (700 CE)
mentioned the relations 2652 3 1532 = 2 and 13512 3 7802 = 1 as a verication
153 780
of the validity of the approximations ([W2], p 16).

22
To cite two other examples of rational approximations: Sr dhara (750 CE),

Aryabhat a
.p II (950),
Sr pati (1039), N
ar
a yan
. a (1350) and Munsvara (1603) gave
p
[ DN 2 ]
the rule D N for a conveniently chosen large integer N ([Bg], p 98{99).
J~nanaraja (1503), his son Suryadasa (1541), and Kamalakara (1658) gave a method
p
of successive approximation ([D2], p 193{194) which may p be expressed as follows:
Let a0 = a = [ D] and ai+1 = 21 (ai + aDi ); then ai D and each ai is a better
p
approximation than the previous ai 1 . The case a2 D was mentioned earlier
p
in the ancient Bakhshal Manuscript in the form D a + 2ra 2((a2+a )r ) , where
r 2
2a
r = D a ([D1], p 11{12). The Sulba
2 fractions 12
17
and 408
577
are respectively the a2
and a3 for D = 2.
Brahmagupta's Bh
avana 103
(see Section 4), lived more than seven centuries after Brahmagupta. It
is however possible that Indians did have an early realisation of this
application of large solutions of varga-prakr.ti but did not mention it in
any major text till the discovery of a general algorithm for nding a
non-trivial solution of Dx2 + 1 = y2 for any D.
(iii) Pursuit as Pure Mathematics

The mathematical chapters of Brahmagupta's treatise ([B]) reveal
the mind and spirit of a great pure mathematician. Many of Brah-
magupta's discoveries show his enthusiastic pursuit of results from a
feel for their intrinsic mathematical worth rather than the requirements
of immediate applications. One may cite, for instance, his theorems (Ch
12, Verses 21, 28) on the area and diagonals of a cyclic quadrilateral
(rediscovered by W. Snell in 1619) and his ingenious construction (Ch
12, Verse 38) of a cyclic quadrilateral whose sides, diagonals, circum-
radius and area are all rational and whose diagonals are perpendicular
to each other (which drew the admiring attention of M. Chasles and
E.E. Kummer in 19th century).

Now Aryabhat . a had already presented a cryptic solution to the lin-
ear indeterminate equation ax by = c. This problem probably arose
from, and certainly had applications in, Indian astronomy. Subsequent
algebraists, including Brahmagupta, brought clarity and made rene-

ments on Aryabhat . a's pioneering work. Having successfully dealt with
the linear case, the pure mathematician's impulse might have driven
Brahmagupta to take up the harder problem of the quadratic indeter-
minate equation. The peculiarity of the varga-prakr.ti would have come
out from his investigations. He would also have recognised the funda-
mental nature and importance of the equation Dx2 + 1 = y2 on which
he clearly focussed considerable attention.
The achievements on Pell's equation tend to overshadow the fact
that, at least from the time of Brahmagupta, Indian algebraists had
produced a large bulk of work involving ingenious solutions of various
types of indeterminate equations.23 Most of these equations were clearly
investigated for their own sake. This zest for exploration of higher degree
indeterminate equations (after the application-oriented linear case) ts
23
See [DS] for some examples.
into the general spirit of the era | the \Classical Age" of Indian history
| when any department of knowledge or sphere of activity, once taken
up, was pursued to its extreme.
Brahmagupta's penchant for pursuing problems in algebra for the
sheer joy (and glory) is explicitly recorded towards the end of his chapter
on algebra, i.e., Kut.t.aka ([B], Chapter 18, Verses 99{100; [C], p 377):
sukhamatramam prasnah. prasnanyatsahasrasah. kuryat

anyairdattatprasnanuktairva sadhayetkaran.aih.
jana samsadi
_ daivavidam_ tejo nasayati bhanuriva bhanam
kut.t.akaraprasnaih. pat.hitairapi kim_ punarj~nataih.
\These questions are stated simply for delight. One may devise a
thousand others; or may resolve the problems proposed by others, by
the rules given here.
\As the sun eclipses the stars by his brilliance, so will an expert
eclipse the glory of other astronomers24 in assemblies of people, by the
recital of algebraic problems, and still more by their solution."
While stating concrete examples, especially those involving the varga
prakr.ti, he often used the phrase kurvannavatsarad gan.akah. | \One
who can solve it within a year (is truly a) mathematician." Clearly he
revelled in the challenge posed by the varga prakr.ti, the determination of
a solution of which was described by Jayadeva ([Sh], p 14{15) as being
\as diÆcult as setting a y against the wind":
prakat.itamatigahanamidam_ marutimukhe maks.ikakaran.am
daivavid means \destiny-knowing" or \astrologer" ([Mw], p 497). In this con-

24
text, \astronomer" seems more appropriate. As some of the great mathematical

astronomers (like Varahamihira) were also famous for their knowledge of astrology, a
sharp distinction was not always made. In the quoted line, the word daivavid might
have been preferred to a standard term for astronomer for greater impact. Apart from
the nuance of \godly" in daiva, the word seems to create a powerful sound-eect in
that line.
Brahmagupta's Bh
avana 105
8 Historical Context
In the history of indeterminate equations in India, the three greatest

achievements are the kut..taka, the bhavana and the cakravala. The

kut..taka algorithm occurs in Aryabhat .ya (499 CE). In the absence25

of evidence to the contrary, one may regard Aryabhat . a as the pioneer
of kut..taka.
Brahmagupta (628 CE) ourished a century after Aryabhat . a. A
truly original idea, the bhavana has the unmistakable stamp of Brah-
magupta's genius. One does not see any seed of the bhavana in the trea-

tises of his predecessors Aryabhat . a and Bhaskara I or in the Bakhs.al
manuscript. None of the results in these remarkable texts come close to
the algebraic sophistication shown by Brahmagupta's research on Pell's
equation. In fact, in several results of Brahmagupta, one discerns a
similar abrupt jump in the mathematical maturity level.
Brahmagupta was born in an era when symbolic algebra was at its
infancy. He himself established some of its seminal features | he intro-
duced the concept of zero as an integer in algebra, formulated the rules of
arithmetic operations involving negative numbers and zero, contributed
to the evolution of convenient notations and terminology (like the use
of distinct letters for several unknowns) and the formation and handling
of equations; and so on. It is astonishing that an ancient scientist, who
had to develop such basic principles in algebra, could have taken the
leap into investigation of a topic like Pell's equation | and that too
with the mind and approach of a modern algebraist.26
The bhavana in uenced, directly and indirectly, subsequent research
on indeterminate equations by Indian algebraists. For tracing the pos-
sible ow and development of ideas, it is desirable to have some clar-
ity regarding the knowledge and achievements at various time-points.
However, due to the absence of certain crucial historical documents, the
25
Hardly any complete post-Vedic mathematics treatise, prior to Aryabhat.ya, has
survived.
26
The achievement appears all the more startling when contrasted with the gen-
eral evolution of algebraic thought: till the 16th century CE, Arab and European
mathematicians struggled with problems involving equations of the type ax + b = c
(a; b; c positive numbers) as shown by the prevalence of the cumbrous \rule of false
position".
modern historian faces some diÆculties.

Now some of the crucial unresolved questions are: Who invented the
cakravala and when? What was the progress in indeterminate equations
during the intermediary period (between Brahmagupta and the inven-
tion of cakravala) and how were the results presented? To what extent
had the awareness of this discovery spread among the mathematicians
in the course of time? One has to remember the caveat of Weil ([W1],
p 234) that \one should not invariably assume a mathematician to be
fully aware of the work of his predecessors".
Perhaps one has to await the discovery of the lost algebra treatises
dhara, Padmanabha, Jayadeva, Sr
of Sr pati and other algebraists who
ourished during the 5 centuries between Brahmagupta (628 CE) and
Bhaskara II (1150 CE), or at least come across relevant quotations from
these texts.
So far, the earliest author who refers to cakravala is Udayadivakara
(1073 CE) who quotes Jayadeva's verses on cakravala. But we have
no idea regarding the dates or other results of Jayadeva. Given the
brilliance of the work, it is surprising that Jayadeva is not mentioned
by Bhaskara II | he does not seem to have been referred to by most of
the mathematicians.
The earliest extant original expositor27 of the cakravala (so far) is
Bhaskara II (1150 CE) who attributes the name of the method (and
hence the method) to earlier writers without specifying any name. Be-
ing a gifted algebraist himself, Bhaskara II would have undoubtedly
realised the greatness of this work. Now, at the end of his algebra
text Bjagan.ita, he makes a general acknowledgement of Brahmagupta,
dhara and Padmanabha as his sources. There is then a strong possi-
Sr
dhara or Padmanabha invented, or at least wrote on, the
bility that Sr
cakravala. But none of Padmanabha's works has been found. There is
dhara's date is estimated to be around 750
no clue as to his dates. Sr
CE. But his algebra treatise is no longer extant. Some of its content,
like the method of \completing the square" in quadratic equations, is
known from references by Bhaskara. Another relevant missing text is
pati (dated 1039 CE).
the Bjagan.ita (algebra) of Sr
27
By original expositor we mean here one who is presenting the result in his own
language and not quoting his predecessor(s).
Brahmagupta's Bh
avana 107
If a book specialises on the most advanced topics of the time, or is

too diÆcult, it is likely to be published in fewer numbers with fewer
editions. Perhaps, for a similar reason, there was not enough circulation
of the exclusively algebra treatises of the 8th{11th centuries.
Appendix: Original Verses on Brahmagupta's

Bhavana
Theorem 1 (in Section 1) was stated by Brahmagupta in Verses 64{65

of Chapter 18 (titled Kut.t.akadhyayah.28 ) of Brahma Sphut.a Siddhanta
([B]) in the following form:
Atha Varga Prakr.tih.
mulam_ dvidhes..tavargad gun.akagun.adis..ta yutavihnacca

adyavadho gun.akagun.ah. sahantyaghatena kr.tamantyam
vajravadhaikyam_ prathama m_ praks.epah. ks.epavadhatulyah.
In symbols, the rst line states:

[If]
p
yi = Dxi 2 + mi ; i = 1; 2;
while the second and third line taken together mean:
. =Kut . .taka (algebra) + adhy . (chapter). The subject algebra
28
Kut..tak
adhy ayah ayah
was called kut..taka-gan.ita before the use of the current Sanskrit term bjagan.ita. The
term kut..taka (pulverisation) indicates that it was regarded as the science of solving
problems by a process of simplication | by breaking quantities (for instance, the
coeÆcients or solutions of a given equation) into smaller pieces. The term, reminis-
cent of Fermat's descent, was particularly appropriate for the solution of the linear

indeterminate equation pioneered by Aryabhat . a ([Du1], p 10{22).
[then] y = Dx1 x2 y1 y2 ; x = x1 y2 x2 y1 ; m = m1 m2
satisfy y2 = Dx2 + m:
(Mention of y = Dx1 x2 + y1 y2 and x = x1 y2 + x2 y1 is explicit; mention
of y = Dx1 x2 y1 y2 and x = x1 y2 x2 y1 also seem implicit.)
A literal translation of Line 1 would be : \[Given] the square roots [yi

for i = 1; 2] of [the following] two [quantities]: the [two] desired squares
[xi 2 ] multiplied by the gun.paka [D] and increased or decreased by the
desired [cip]." Thus yi = Dxi 2 ci , where ci are positive integers;
i.e., yi = Dxi 2 + mi , where mi are integers. Another literal transla-
tion could begin as follows: \[Consider following] two [integral/rational
quantities] which have square-roots [i.e., are perfect squares yi 2 ]: ",
i.e., yi 2 = Dxi 2 + mi (i = 1; 2).
A literal translation of Line 2: \The product of the [given] initial (i.e.,

rst) roots [x1 ; x2 ] multiplied by the gun.aka [D] together with the prod-
uct of the [given] nal (i.e., second) roots [y1 ; y2 ] yields a [new] nal (i.e.,
second) root [y] [of the equation y2 = Dx2 + m].", i.e., y = Dx1 x2 + y1 y2
or y = Dx1 x2 y1 y2 .
A literal translation of Line 3: \Cross-multiplication [yields new] rst

[root of the above equation]; [new] interpolator is equal to the product
of the previous interpolators."
Theorem 1 was also stated in verse form by Ac arya Jayadeva,

Bhaskara II (1150), Narayan.a (1350), J~nanaraja (1503) and Kamalakara
(1658).
Remark 1
As pointed out by Datta-Singh ([DS], p. 146), the word dvidha in
Line 1 has two-fold implication:
(i) Consider two triples satisfying Dx2 + m = y2 .
(ii) Consider two copies of a triple satisfying Dx2 + m = y2 .
Brahmagupta's Bh
avana 109
Thus (ii) is a special case of (i) and, together with the next two lines,
indicate Corollary 1: If Dp2 + m = q2 , then D(2pq)2 + m2 = (Dp2 + q2 )2 .
That the audience is also expected to read the implication (ii) in
Line 1, and therefore Corollary 1 in Lines 1{3, is conrmed by the very
next Line 4:
praks.epasodhakahr. te mule praks.epake rupe
\On dividing the roots [obtained] by the [original] additive or sub-
tractive, [roots for the] interpolator unity [will be found]." OR \On
dividing [the roots obtained] by the square root of the [new] additive or
subtractive, [roots for the] interpolator unity [will be found]."
Thus, we have, Corollary 2: If (p; q) is a root of Dx2 c = y2 , then
( ; Dp +q ) is a root of Dx2 + 1 = y2 .
2 pq 2 2
c c
Line 4 is in harmony with the previous lines if the implication (ii) is
also considered; otherwise the language used might appear abrupt.
Remark 2
Both the additive and subtractive versions of Theorem 1 seem to
nd simultaneous mention in Brahmagupta's verses through the expres-
sions saha and aikyam_ with the ambiguous sense of \combining" Dx1 x2
with y1 y2 and x1 y2 with x2 y1 . The word \combining" certainly includes
the implication \adding". Thus the reference to the more important
additive principle (producing new roots Dx1 x2 + y1 y2 and x1 y2 + x2 y1 )
is obvious. Presumably, \combining" also includes \taking dierence";
the subtractive principle (generating Dx1 x2 y1 y2 and x1 y2 x2 y1 ) is
unlikely to have been overlooked by Brahmagupta. (Also see Remark
4 below.) In any case, Jayadeva, Bhaskara II and subsequent writers
clearly describe the two principles successively in separate verses.
Remark 3
Theorem 1 in the Introduction was stated in three equivalent forms.
Brahmagupta's verses (quoted above) seem to resemble best the second
form. But the perception of Theorem 1 in the rst form (i.e., in terms
of an operation ) comes out in the applications by Brahmagupta and
others, in the version of Theorem 1 due to Bhaskara (quoted below) and,
above all, in the choice and use of the term bhavana (composition).
Remark 4
During our literal translation, we had to insert several brackets `[ ]'
to include expressions which are implicit in Brahmagupta's verses. The
3 lines, conveying Theorem 1, could appear obscure for a person not
already familiar with the essence of the result. Cryptic brevity is a gen-

eral feature of the original treatises of the earlier greats like Aryabhat
.a
(499 CE) and Brahmagupta (especially the former). The verses were
meant to indicate broad hints | not complete details. There used to be
emphasis on the use of a learner's own intellect for lling up the details.
Here one may point out that most modern mathematicians feel that a
researcher eventually gains more insight into his area from a terse text
than from a clearly spelt-out text, from an obscure important paper
than from a lucid one.
Besides, conciseness was also a practical necessity. Palm-leaf manu-
scripts do not last long. Thus a treatise could be preserved only through
memorisation or through repeated copying. In either case, the more con-
cise the work, the better its chance of survival through faithful memo-
risation or copying. Brevity would have been particularly indispensable
for an ancient author with enormous output. One can appreciate Brah-
magupta's compulsions when one considers that, in spite of its oppressive
terseness, the number of verses in Brahma Sphut.a Siddhanta exceeds a
thousand.
With passage of time (and improvement of writing materials), the
later treatises, even by stalwarts, give more and more elaborate versions
of the results of the predecessors. Jayadeva's verses on the bhavana
([Sh], p. 4{8) have greater clarity than Brahmagupta's, Bhaskaracarya's
version (quoted below) is still more lucid.
Detailed expositions on the works of the Masters were meant to be
orally transmitted through the guru-sis.ya (mentor-disciple) link. The
commentaries too try to clarify the implications of the passages of the
original treatises. Brahmagupta's verses are explained by Pr.thudaka
Svami (860 CE); Bhaskara's verses by Kr.s.n.a (1600) and Suryadasa
(1541).29
Brahmagupta's result, once stated, is not diÆcult to prove. Proofs
29
The footnotes in [C], p 170{172 and 363{364 contain excerpts from these com-
mentaries on Brahmagupta's and Bhaskara's verses describing the bhavana.
Brahmagupta's Bh
avana 111
of such results are not to be expected in the original treatises (where

even the statements of main results are terse). They were communicated
orally and sometimes recorded in the later works. A proof of Theorem
1 is described by Kr.s.n.a in his Bjapallava (1600) | see ([DS], p 148{149).
The version of Bhaskara II

Bhaskaracarya's version ([Ba], p 22) of Brahmagupta's bhavana oc-
curs in the section \Varga Prakr.ti" of his text \Bjagan.ita" (Verses
70{71). We present below the transliteration followed by a translation.
is.tam_ hrasvam_ tasya vargah. prakr.tya ks.un.n.au yukto varjita va sa yena
mulam_ dadyat ks.epakam_ tam_ dhanarn.am_ mulam_ tacca jyes..thamulam_
vadanti
hrasvajyes..tha ks.epakan nyasya tes.am_ tananyanva'dho nivesya kramen.a
sadhyanyebhyo bhavanabhirbahuni mulanyes.am_ bhavana procyate'tah.
vajrabhyasau jyes..thalaghvostadaikya m_ hrasvam_ laghvorahatisca prakr.tya
ks.un.n.a jyes..thabhyasa yug jyes..thamulam_ tatrabhyasah. ks.epayoh. ks.epakah.
syat
hrasvam_ vajrabhyasayorantara m_ va laghvorghato yah. prakr.tya vinighnah.
ghato yasca jyes..thayostadviyogo jyes..tham_ ks.epo'trapi ca ks.epaghatah.
\[Consider] the desired lesser root [x].30 Its square, multiplied by
prakr.ti [D], is added or subtracted by some quantity [c] such that it [the
sum or dierence Dx2 c] gives a [integer] square root. The interpolator
[m = c], positive or negative, is called the ks.epa; the square root [y] is
called the greater root.
\Set down successively the lesser root [x1 ], the greater root [y1 ] and
the interpolator [m1 ]. Place under them, the same or another [triple
x2 ; y2 ; m2 ], in the same order. From them, by repeated applications of
the bhavana, numerous roots can be sought. Therefore, the bhavana is
being expounded.31
30
and y in Dx2 + m = y 2 , which were called adya (initial or rst) and antya
x
(nal) respectively by Brahmagupta, were named hrasva (lesser) and jyes..tha (greater)
respectively by Bhaskara II; the gun.aka D of Brahmagupta was called prakr.ti by
Bhaskara II.
31
The last sentence comes from the phrase bhavana procyate'tah. which can also
\[Consider] the two cross-products of the two greater and the two
lesser roots. The sum [x1 y2 + x2 y1 ] of the two cross-products is a lesser
root. Add the product of the two [original] lesser roots multiplied
by the prakr.ti [D] to the product of the two greater roots. The sum
[Dx1 x2 + y1 y2 ] will be a greater root. The product of the two [previous]
interpolators will be the [new] interpolator.
\Again the dierence between the two cross-products is a lesser root.
The dierence between the product of the two [original] lesser roots
multiplied by the prakr.ti and the product of the two greater roots will
be a greater root. Here also, the interpolator is the product of the two
[previous] interpolators."
We mention here that Bhaskaracarya referred to Brahmagupta as
gan.aka cakra cud.aman.i32 (jewel among mathematicians).
Acknowledgement. The author sincerely thanks Prof. Sitanath Goswami
for verifying the accuracy of the Sanskrit-English translations and ac-
companying notes, and Professors Amit Roy, Kalyan Mukherjea and
Raja Sridharan for stimulating mathematical discussions and valuable
suggestions.
References
[B] Brahmagupta, Brahma Sphut.a Siddhanta (628). The Sanskrit

text, edited by R.S. Sharma, has been published by the Indian
Institute of Astronomical Research (1966).
[Ba] Bhaskaracarya, Bjagan.ita (1150). The Sanskrit text, edited

and translated by S.K. Abhyankar, has been published by the
Bhaskaracharya Pratishthana, Pune (1980).
[Bg] A.K. Bag, Mathematics in Ancient and Medieval India,

Chaukhambha Orientalia, Varanasi (1979).
be translated as \Therefore, it is called bhavana." Both meanings might have been

intended. After all, Bhaskara II began the text with a deliberate pun on samkhy
_ .,
ah
satpurus. a, b
ja, vyakta, etc!
. aka= mathematician, cakra c . i= crest-jewel; round jewel in a coronet.

32
gan ud.
aman
Brahmagupta's Bh
avana 113
[C] H.T. Colebrooke, Algebra, with Arithmetic and Mensuration,

from the Sanscrit of Brahmegupta and Bhascara, John Murray,
London (1817).
[D1] B. Datta, The Bakhshal Mathematics, Bulletin of the Calcutta
Mathematical Society Vol 21 No 1 (1929), p. 1{60.
[D2] B. Datta, Narayan.a's Method for Finding Approximate Value
of a Surd, Bulletin of the Calcutta Mathematical Society Vol
23 No 4 (1931), p. 187{194.
[D3] B. Datta, The Science of the Sulba, University of Calcutta
(1932); reprint (1991).
[DS] B. Datta and A.N. Singh, History of Hindu Mathematics: A
Source Book Part II (Algebra), Motilal Banarsidass, Lahore
(1935); Asia Publishing House, Bombay (1962).
[Di] L.E. Dickson, History of the Theory of Numbers Vol II, Chelsea
Publishing Co. NY (1952).
[Du1] A.K. Dutta, Diophantine Equations: The Kuttaka, Resonance
7(10) (October 2002), p. 6{22.
[Du2] A.K. Dutta, Diophantine Equations: The Samasabhavana, Res-
onance 8(11) (November 2003), p. 10{24 and 9(1) (January
2004), p. 61{63.
[K] A.A. Krishnaswamy Ayyangar, New Light on Bhaskara's
Chakravala or Cyclic Method of solving Indeterminate Equa-
tions of the Second Degree in two Variables, J. Indian Math.
Soc. 18 (1929{30), p. 225{248.
[Mw] M. Monier-Williams, Sanskr.it-English Dictionary, Clarendon
Press, Oxford (1899); reprinted Munshiram Manoharlal, New
Delhi (2002).
[O] Manuel Ojanguren, The Witt Group and The Problem of Luroth,
Dottorato di ricerca in Mathematica, ETS EDITRICE PISA,
University of Pisa (1990).
[Sa] T.A. Sarasvati Amma, Geometry in Ancient and Medieval In-
dia, Motilal Banarsidass, Delhi (1979; 1999).
[Se] C.O. Selenius, Rationale of the Chakravala Process of Jayadeva

and Bhaskara II, Historia Mathematica 2 (1975), p. 167{184.
[Sh] arya Jayadeva, The Mathematician, Gan.ita 5
K.S. Shukla, Ac
(1954), p. 1{20.
[Si] pati: An Eleventh-Century Indian Mathemati-
K.N. Sinha, Sr
cian, Historia Mathematica 12 (1985), p. 25{44.
[Sr] R. Sridharan, Ancient Indian contributions to Quadratic Alge-
bra, Science in the West and India (ed B.V. Subbarayappa and
N. Mukunda), Himalaya Publishing House, Bombay (1995).
[W1] Andre Weil, History of Mathematics: Why and How, Proceed-
ings of the International Congress of Mathematicians, Helsinki
1978 Vol 1 (ed. O. Lehto), Helsinki (1980), p. 227{236.
[W2] Andre Weil, Number Theory: An approach through history from
Hammurapi to Legendre, Birkhauser (1984).
The Karan
. : How to Use Integers to
Make Accurate Calculations on Square
Roots
Francois Patte
UFR de Mathematiques et Informatique
Universite de Paris 5 Rene Descartes
75270 Paris Cedex 06
francois.patte@math-info.univ-paris5.fr
Abstract
This paper presents the karan., a mathematical construction
to use integers to make calculations with square roots. Indian
p p
mathematicians p for this purpose (e.g.
invented pnew operations
( 2 + 8)2 = 2 + 8 + 2 2 8 = ( 18)2 for the sum of what
they call the karan. 2 and 8, the sum of which is the karan. 18).
This construction seems to be sophisticated, even useless, but we
can nd an explanation in a commentary (17th century): if all
the calculations on square roots are made with karan., and the
approximate value is taken only at the end, the result is more ac-
curate than if approximate values are taken at the beginning of
the calculation.
This paper intends to present a mathematical construction made

by Indian mathematicians in order to use integers in the calculations
involving square roots: these mathematical objects are called karan..
To begin with, it may be useful to introduce the authors and the
texts on which this paper is based. According to the Indian tradition,
knowledge develops and is transmitted through a fundamental text and
the commentaries written on it. Here, the main text is the Bjagan.ita
of Bhaskaracarya, and two commentaries are used: the Suryaprakasa by
Suryadasa ([1]) and the Bjapallava by Kr.s.n.adaivaj~na ([2]).
115
116 Fran
cois Patte
There are two Indian mathematicians by the name of Bhaskara: one

who lived in the 7th century and is often referred to as Bhaskara I
and one who lived in the 12th century and is known as Bhaskaracarya
(Bhaskara-acarya: Bhaskara the master) or Bhaskara II; in this paper,
his name will be shortened to Bhaskara because there is no ambiguity
here: we shall never use the works of the rst Bhaskara. In the same
way, Kr.s.n.adaivaj~na, is abbreviated as Kr.s.n.a.
1 The author: Bhaskara

Bhaskaracarya was born in 1036 saka, which is 1114 A.D., as he puts
it himself at the end of the third part of his main work, the Siddhanta-
siroman.i([3]): \I was born in the year 1036 of saka kings, during my
thirty-sixth year I composed the Siddhantasiroman.i".1
Then, he gives the name of a town and its location, the name of
the gotra he belongs to and the name of his father: \There was at
Vijjad.avid.a, a town located in the Sahya mountains. . . a twice-born from
an.d.ilya lineage. . . the virtuous Mahesvara. . . , born from the latter,
the S
the clever poet Bhaskara. . . " 2
What remains a mystery up to now is the exact name and location of
the town: Vijjad.avid.a. The Sahya mountains are located in the northern
part of the Maharashtra state and hold the well-known sites of Ajanta
and Ellora; in these mountains, an inscription was discovered about
1850 in the basement of a temple at Patnadev, a small place near the
modern Chalisgaon; this inscription gives Bhaskara's genealogy from the
10th to the 13th century, approximately, and was engraved to celebrate
the foundation by his grandson of a school dedicated to the studies of
Bhaskara's works in this very place. The name of Vijjad.avid.a is not
quoted in this inscription which explains that king Jaitrapala, from the
Yadava dynasty, made Bhaskara's son, Laks.mdasa, his astrologer and
took him from \this town" to his capital; at that time, the capital of
the Yadavas was Devagiri which has been identied as Daulatabad near
1
rasagun.apurn.amahsamasakanr.pasamaye 'bhavan mamotpattih. rasagun.avars.e-
n.a maya siddhantasiroman.ih. racitah.
2
ast sahyakulacalasritapure. . . vijjad.avid.e san.d.ilyagotro dvijah.. . . mahesvara-
kr.t. . . tajjas. . . suddhh. kavir bhaskarah.
The Karan
.
117
Aurangabad and is not that far from Patnadev.
1.1 Bhaskara's works

The main work of Bhaskara is the Siddhantasiroman. i ; it is composed
of four parts: the rst two are mathematical ones, the last two are as-
tronomical ones. The two mathematical parts are respectively entitled:
Llavat and Bjagan.ita.
Let us brie y explain the title Siddhantasiroman. i : \The Diadem on
the Siddhanta". Siddhanta means \settled opinion". In India this name
was given to the fundamental astronomical works; there were ve major
Siddhantas but there remains only one of them: the Suryasiddhanta,
\The Siddhanta of the Sun", thus called because it is assumed to have
been revealed by the god Sun. The work of Bhaskara is based on it.
The Llavat is a treatise of elementary calculus; numeration, oper-
ations, rules of proportions, calculation of areas, volumes and so on; it
belongs to the class of pa.t-gan.ita or vyakta-gan.ita. pa.t means method,
gan.ita means calculus and vyakta means manifested; this is a \method
of calculus" or a \calculus on manifested numbers".
We can nd the name of the title at the end of the rst stanza of the
Llavat which describes the method (pa.t): \(. . . ) I proclaim with soft
and correct words, using short syllables, a method of good computation,
which causes great satisfaction, which is clear and possesses the grace
of the play." 3
The Moghul Abu al Fayd. Fayd. reports a legend about this title,
which is also a name given to a girl in India, in his translation of this work
into Persian, in 1587. He says that Llavat is the name of Bhaskara's
daughter to whom this book was dedicated. We did not nd anything
about this in Sanskrit commentaries; commentators merely explain the
formation of the word lla-vat : \that which possesses the play, that
which is like a game".
Bjagan.ita : this is the generic name for algebra. Bja means seed,
so Bjagan.ita is calculus on seeds, the seeds which potentially contain
3
(. . . ) pat.m
. sadgan.itasya vacmi caturaprtipradam prasphut.am sam.ks.ipta-
ks.arakomalamalapadair lalityall avat
m
118 Fran
cois Patte
calculus on manifested numbers; another name for this is avyakta-gan.ita :

calculus on non-manifested numbers. We can nd in the commentaries
some parallel with the vyakta and avyakta worlds, the manifested and
non-manifested worlds of the Sam. khya philosophy, the non-manifested
world containing the manifested world before the creation.
The Bjagan.ita expounds calculus on negative and positive num-
bers, calculus on unknown quantities | these are the avyakta, the
non-manifested numbers which contain the possibility of making \real
computations" if you replace them by numbers (manifested-numbers);
another name for unknown quantities is varn.a : color, letter | and it ex-
plains resolution methods of equations: linear, Diophantine, algebraic...
The fourth chapter, following the chapter on unknown quantities, is
devoted to the karan..
The two astronomical parts of the Siddhantasiroman. i are entitled:
Grahagan.itadhyaya, \Lesson on the Computation of Planets", and Go-
ladhyaya, \Lesson on Spheres". This part contains some trigonometry:
calculation of additions of sines etc.
The Siddhantasiroman. i was written in verse, as is usually done for
this kind of treatises, and it is often diÆcult to understand without
the help of commentaries. Here, we shall use two commentaries on the
Bjagan.ita : one is the Suryaprakasa by Suryadasa, composed around
1530; the other is the Bjapallava by Kr.s.n.adaivaj~na, composed around
1604.
The other Bhaskara's known works are: a commentary on his Sid-
dhantasiroman. i, the Mitaks.ara \Having Measured Syllables" or Vasa-
nabhas.ya ; these are the names given for small commentaries to explain
works very brie y. The Karan.akutuhala : \The Wonder of Astronomical
Calculations" is a practical treatise on astronomy. It is dated 1183 by
Bhaskara himself; this is the last date that is known about him and,
for this reason, some people think that he died about 1185. We do not
know exactly when and where. Some think that he was in charge of the
astronomical observatory in Ujjain and that he died there, but there are
no texts and no inscriptions to support these facts. The astronomical
instruments that we can see nowadays in Ujjain were erected in the
18th century by the maharajah of Jaipur, Jai Singh II. The last work
is a commentary on a mathematical work by Lalla, an astronomer from
The Karan
.
119
. yadhvr.ddhidatantra : \A Treatise to Increase

the 8th century: the Sis
the Understanding of Students".
2 The commentators: S
uryadasa and
Kr.s.n.adaivaj~
na
2.1 S
uryadasa
We do not know much about Suryadasa. As is usually the case, he

belonged to a family of astronomers and, very likely, lived in the west-
ern part of the Godavar valley during the rst part of the 16th cen-
tury. He is known for two commentaries: one on the Llavat, entitled
Gan.itamr.takupika, \The Well of Nectar which is The Calculus", and one
on the Bjagan.ita : the Suryaprakasa, \The Brightness of the Sun"; this
title is a play on words: his own name, Suryadasa, but also the name of
Bhaskara, which also means the Sun, as does the word surya. The name
of his father is also known, J~nanaraja. He wrote an astronomical trea-
tise too: the Siddhantasundaraprakr. ti or \The Charming Foundation of
the Siddhanta", and a mathematical work, in imitation of Bhaskara's
Bjagan.ita : the Siddhantasundarabja.
2.2 Kr.s.n.adaivaj~
na
Kr.s.n.adaivaj~na was born in a family of astronomers who had settled in

Varanasi at the end of the 16th century. He was a protege of the Moghul
emperor Jahangir (1605-1627). His commentary on the Bjagan.ita, the
Bjapallava or \The Sprout of the Seed", is dated Saturday, the fourth
tithi of the dark fortnight of the Caitra month, in the year 1523 of
the saka era, namely: Tuesday, March 12th, 1602. He also wrote some
examples (udaharan.a) using the horoscopes he made about members of
the Moghul dynasty and he may have composed a commentary on the
Llavat. Unlike the others, Bhaskara and Suryadasa, he seems to have
had as a guru, not his own father, but a nephew of Gan.esa, another
great Indian mathematician.
120 Fran
cois Patte
3 The karan.i
3.1 The \concept" of karan
.

The word karan. means producer. It is in the feminine for, in Sulbas utra
texts, \The Aphorisms on the Rope", it was originally in relation to
the word rajju, or rope, which is feminine too. This rajju karan. has
been used to produce geometric squares, that is a right angle; there are

some aphorisms in the Sulbas utra explaining how to place some marks
on a rope, at the distance of 3, 4 and 5, in order to produce a right-
angled triangle and hence a square. So, the word karan. was used to
designate the side of a square and because of that, it can mean the
square root of a number. From the 5th century before the common era
to the 17th century, the word karan. seems to have had many meanings
related to squares and square roots and consequently it is not easy to
translate. Moreover, the mathematical concept expressed by this word
in this papaer, has no equivalent in modern mathematics. For this reason
we will not translate it. One common translation into English is surd.
The word surd is ambiguous, for it means either irrational number or
square root; there is no idea of irrational numbers in karan., no idea of a
number which cannot be expressed by a ratio, by a fraction of integers.
As for the meaning \square root", this is not a correct translation for
karan., as karan. are not merely numbers, but numbers with a set of
operations.
Bhaskara does not give any denition of karan.; the rst rule, \sutra"
as this kind of stanza is called in Sanskrit, deals straight with the rules
of addition and multiplication, but we can nd some information in the
commentaries. Here is what Suryadasa says about this rst sutra 4 :
\Now, [the author] who examines the nature of these karan.s, under
the pretext of explaining the rule of multiplication, says: \vargen.a".
One will multiply a square vargen.a by a square number; likewise,
one will divide a square by a square only; on the contrary, one will not
multiply or divide a square by a number. It is pointed out by this, that
what is of the nature of karan. by name is of the nature of a number
4
We will see the sutra itself later on. The transliteration of all quoted texts is
given at the end of the article, see page 133.
The Karan
.
121
accepted with the quality of square; this is said by Narayan.a:
\The name of karan. will be for that [number] the square root
of which must be taken.5 "
We shall examine the rst part of this text later, and use now only
the underlined part. What is important here is that if you consider
the number two as a karan., you must accept that this number is a
square; this is reinforced by the above quotation from Narayan.a who is
a mathematician from the 14th century. A karan. is a number for which
you have to compute the square root, so you must think that it is a
square, you must think of a mental squaring operation.
Here is now what Kr.s.n.a says as an introduction to his commentary
on this chapter:
\Now are commented the six operations on the karan.. In this mat-
ter, one should understand that the six operations on karan., are six
operations done through the squares of two square root quantities. Be-
cause the origin of these six operations is preceded by a state of square,
therefore, [there will be] also the use of the expression \state of karan.",
in relation with these six operations, for a quantity producing a square
root; this usage will not be possible if the calculation proceeds with the
state of square root as the rst step. The technical word \six operations
on karan." must be understood because of the necessity of such calcula-
tions about square root numbers; in this [case], that quantity for which
a square root without a remainder is not possible, when its square root
is required, is a karan., but this is not merely a quantity which does not
produce a square root; if it were so, there would always be the use of the
expression \state of karan." for two, three, ve, six, and so on.
| Let it be so!
No! If the operations were done on that basis, for instance, eight
added to two would be eighteen, etc."
This commentary requires some explanations. Note that the \six
operations" are the operations which are studied in the Bjagan.ita,
whatever the objects (negative numbers, unknown quantities, karan.),
5
Quoted by Suryadasa from Narayan.a's B
jagan
. it
avatam
. sa.
122 Fran
cois Patte
namely: addition, subtraction, multiplication, division, squaring and

square root. First, we recognise the same idea as in the commentary by
Suryadasa: a karan. is a number which is the square of its square root.
Of course, this is a truism. For a modern mathematician, every number
is the square of its square root; but at the time of Bhaskara, and his
commentators, it was not so obvious that a number which was not a
square could have a square root; we can see that in this text: Kr.s.n.a
speaks of a quantity producing a square root, the Sanskrit compound
mula-da (which gives a root) is used by Indian mathematicians to des-
ignate a perfect square. Here, we are told that a square can be called
a karan. but | and it is the second fact to be noticed | \in relation
with these six operations", that is to say that we cannot separate, for a
number, the state of being a karan. from the class of operations dened
for them.
This becomes clearer with the end of the text: Kr.s.n.a explains that
a \quantity for which a square root without a remainder is not possible,
when its square root is required, is a karan. but this is not merely a
quantity which does not produce a square root". Indeed! For if we keep
in mind that every number which is a square could be called a karan.,
every integer should be a karan.. The explanation comes right after,
taking the form of a small dialogue, as is often the case in Sanskrit
commentaries: an opponent argues \Let it be so!" : and what if we take
this denition for granted? Kr.s.n.a answers: \If the operations were to be
done on that basis, for instance, eight added to two would be eighteen."
So we cannot call a number a karan. without thinking of how to make
operations with it.
3.2 Rules of computation
We can now analyse the text of Bhaskara and see why \eight added to
two would be eighteen" ; here are the two rst rules of this chapter:
Let us x the mahat as the sum of the two karan.s and the
laghu as twice the square root of their product, the sum and
the dierence of these two are as for the integers. One will
multiply and divide a square by a square. {1{ But the square
root of the larger [number] divided by the smaller [number],
The Karan
.
123
plus one or minus one, multiplied by itself and by the smaller

[number], will also be respectively, the sum and the dierence
of these two [karan.s]. One will leave them apart if there is
no square root. {2{6
These sutras give two ways of calculation for the sum and the dif-
ference of two karan.s : in the rst one, two new objects are dened:
the mahat, the greater, which is the sum of the two karan.s, that is
to say the sum of the two integers which measure the two karan.s ; the
second object is the laghu, the smaller, which is twice the square root of
the product of the same integers. The names chosen for these technical
words are obvious, for the sum of two integers is always greater than
twice the square root of their product. Once the mahat and the laghu
have been calculated, the two integers produced are added and the re-
sult is the karan. sum of the two karan.s, that is to say: the square of
the sum of the square roots of the two integers measuring the karan..
If the dierence of the mahat and the laghu is computed, the result is
the dierence of the two karan.s. There is a problem with this denition:
the dierence between two karan.s is the same whether you compute the
dierence between the greater one and the smaller one or the smaller
minus the greater.
Before going deeper in the explanations of these rules, let us see the
example given by Bhaskara, and its solution by Suryadasa.
Say the sum and the dierence of the two karan.s measured
by two and eight and by three and twenty-seven and after
a long while, if you know the six operations on karan., say,
dear, [the sum and the dierence] of these two measured by
three and seven.7
Now the commentary by Suryadasa:

6
yogam . karan.yor mahatm . prakalpya ghatasya mulam . dvigun.am
. laghum.
ca yogantare rupavad etayoh. sto vargen.a vargam . gun.ayed bhajec ca laghvya hr.tayas
tu padam . mahatyah. saikam. nirekam . svahatam. laghughnam yogantare stah. kramasas
tayor va pr.thaksthitih. syad yadi nasti mulam
7
dvikas.t.amityos tribhasam.khyayos ca yogantare bruhi pr.thak karan.yoh. trisap-
tamityos ca ciram . vicintya cet s.ad.vidham. vetsi sakhe karan.yah.
124 Fran
cois Patte
In this case, putting down ka 2 ka 8, by: \Let us x the mahat as

the sum of the two karan.s ", the mahat is 10.
Now, the product of these two karan.s is 16, its square root 4, when
multiplied by two, the laghu is produced: 8.
\The sum and the dierence of these two are as for the integers": 18
and 2; these two karan.: ka 18 and ka 2 are the sum and the dierence.
The solution of this exercise put forward by Bhaskara is clear. We
can observe how the mathematicians proceeded to dierentiate a simple
number and a number which must be considered as a karan.: they wrote
the rst syllable of the word karan. before the number: ka 2 and this
means that this \2" is the square of the square root of 2, so, in this
addition we have to nd out the square of the square root which is the
sum of the square root of 2 and the square root of 8; with our modern
notations, we have:
p p p p
( 2 + 8)2 = 2 + 8 + 2 2 8 = ( 18)2
If we could mix our modern formalism with the ancient Indian one,
we would write that the set of karan. is a set of numbers with this
particular addition:
p
ka a ka b = ka (a + b + 2 ab)
It is obvious that this addition (and subtraction) is only dened
if the product of the two integers measuring the karan. is a square.
What happens if this is not the case is expressed by the last sentence of
Bhaskara's rule: \One will leave them apart if there is no square root."
This situation is illustrated by the third example given: the sum of the
karan. measured by three and seven is impossible because 21 is not a
square, so the solution of this exercise is merely: ka 3 ka 7.
This creates an extension of the meaning of karan.. In Sanskrit texts
there are some statements like: \Let the karan. be measured by 2, 3 and
5" as if a karan. was also a composition of several karan.s the sum of
which is not possible but which can be used for other calculations.
The second way for adding, or subtracting, the karan.s given by
Bhaskara (see page 122) is very simply derived from the rst one and
there is no need to comment upon it:
The Karan
.
125
p p r
8 18
( 8 2)2 = 2(
2
1)2 = 2
Let us focus on the sentence of the rule: \One will multiply and divide
a square by a square." We have seen how one of the commentators,
Suryadasa, had explained it (page 120) in order to put forward the
nature of square of the karan..
Here is now what Kr.s.n.a says about it: \vargen.a vargam. gun.ayed
bhajec ca": here is what is said: when you want to multiply karan.s,
if there is the state of multiplicand or the state of multiplier for some
integer | or if you want to divide karan.s, if there is the state of divi-
dend or the state of divisor for some integer | then, having squared the
integer, the multiplication and the division can be performed, because a
karan. has the nature of square."
Another idea springs from this commentary: how to \embed" the
integers in the karan. set; if we want to make operations mixing integers
and karan., we have to square the integers in order to give them a state
of karan.s and this will change the general rules given for the operations.
For instance, the rule for squaring the integers given in the Llavat uses
the identity (a + b)2 = a2 + b2 + 2 ab. So Kr.s.n.a explains afterwards: \for
the square also, the fullment is in like manner because it (the square)
is a kind of multiplication according to its nature of product of two equal
[numbers]. Or, according to the method stated for the manifested num-
bers: \The square of the last [digit] must be placed and [the other digit
must be] multiplied by the last one increased two times..."8 , there will
be a fullment [of this method] for the squaring of the karan. also but,
as has been said: \One will multiply and divide a square by a square",
when it is said: \multiplied by the last increased two times", we must
understand: \multiplied by the last increased four times."
Let us now make a digression. Notice the extreme degree of con-
ciseness of Sanskrit works such as the Bjagan.ita : in one single rule,
Bhaskara describes four of the six operations: addition, subtraction,
multiplication and squaring. Although he describes the addition and
8
sthapyo 'ntyavargo dvigun.antyanighnah. svasvoparis.t.ac ca tathapare 'nk
_ as
tyaktvantyam utsarya punas ca rasim | Llavat
126 Fran
cois Patte
the subtraction, by giving two methods, because these operations dier

from the same operations applied to numbers or unknown quantities, a
simple sentence is enough for him to say how to handle the multiplication
and consequently, the squaring.
This method of exposition is usual in the fundamental works of San-
skrit literature and the works of Bhaskara are of this kind; they were
used for centuries as a basis for mathematical teaching: students learned
them by heart, then their masters composed commentaries which be-
came original lectures. In order that they may be easily remembered,
they were composed in verse, \using short syllables", as seen earlier
(page 117), and any unnecessary rule was avoided. So, in this chapter
on the karan., there will not be any proper rule for multiplication: once
it is understood | with the help of the commentator | that for any
\simple" karan. 9 the multiplication is merely given by:
p p p
( 2)2 ( 8)2 = ( 2 8)2
the general rule for karan. measured by more than one integer has al-
ready been given in a preceding chapter, the chapter on unknown quan-
tities:
One must think here, in like manner, of the rule of multipli-

cation by parts stated for manifested numbers, in the case of
the square of non-manifested numbers and in the case of the
multiplication of karan.s.10
And even here, the rule refers to another rule given in the Llavat
which describes the property of distributivity of the multiplication with
respect to the addition | the \multiplication by parts":
(. . . ) Or the multiplicand, equal in number to the number of

parts of the multiplier, being placed under each of them, is
multiplied by these parts and added up.11
9
Let us call \simple" a karan. measured by a single integer
10
avyaktavargakaran.gun.anasu cintyo vyaktoktakhan.d.agun.anavidhir evam atra
11
gun.yas tv adho 'dho gun.akhan.d.atulyas taih. khan.d.akaih. sam.gun.ito yuto va
The Karan
.
127
In fact, it is clear from the text of the Llavat and its commentaries
that the \parts" in question are either the result of splitting an integer
into two (or more) parts in order to perform a mental calculation or the
digits of the number with their decimal place value; so, this last rule can
be used to perform calculations on polynomial-like quantities: a number
considered in such a way being merely a polynomial in the powers of ten.
Let us see now the example given by Bhaskara for the multiplication
of the karan.:
Set the multiplier as the karan. counted by two, three and

eight and the multiplicand as counted by [the karan.] three
with the integer ve; say the product quickly. Or the multi-
plier is the two karan.s measured by three and twelve less the
integer ve.12
and the solution given by Suryadasa:

Here, the multiplier is: ka 2 ka 3 ka 8.
In like manner, the multiplicand is counted by three with ve units;
in this multiplicand, there are: ka 3 ru 5.
One notices an integer: after taking its square, the state of karan.
must be brought about because it has been said: \One will multiply and
divide a square by a square". By so doing, ka 3 ka 25 are produced.
Now, according to the method of the rule: \One must think here, in
like manner, of the rule of multiplication by parts stated for manifested
numbers, in the case of the square of non-manifested numbers and in the
case of the multiplication of karan.", after multiplication, ka 54 ka 450
ka 9 ka 75 are produced.
The rule \One will leave them apart if there is no square root" is used
in this example and we discover a new formalism to denote integers: r u,
rst syllable of the Sanskrit word rupa, the meaning of which is \unity",
\integer". The commentator squares this integer to transform it into a
karan. before he performs the multiplication.
12
dvitryas.t.asam.khya gun.akah. karan.yo gun.yas trisam . khya ca sapa~ncarupa va-
dham . pracaks . v
a
s u vipa~
n car
u pe gun
. o 'thav
a tryarkamite karan.yau
128 Fran
cois Patte
According to Bhaskara's rule for the multiplication \by parts" (see

page 126), we could represent the way to do this multiplication with the
following table:
ka 2 ka 3 ka 8
ka 3 ka 25 ka 3 ka 25 ka 3 ka 25
ka 6 ka 50 ka 9 ka 75 ka 24 ka 200
y z y z
The multiplier is split in to three parts, as stated in Bhaskara's rule,
and placed in the rst row; then the multiplicand is put under each part
of the multiplier and the multiplication is performed in each cell of the
table, the results being written in the third row.
We have put an identical symbol under the karan. for which the
addition is possible.
p
y: 6 + 24 + 2 6p 24 = 54
z: 50 + 200 + 2 50 200 = 450
Once the addition is done, the result is the one given by Suryadasa.
The second example raises a problem because there is no mathemat-
ical notation for addition or subtraction; the Indians have developed for-
malisms in some branches of knowledge like grammar and mathematics
but there are no signs to denote the operations. In this chapter on the
karan., putting two karan.s side by side indicates that it is the sum of
these two karan. which is considered. This is the meaning of: \One
will leave them apart if there is no square root". For the subtraction,
the notation is almost the same because it has been explained at the
beginning of the Bjagan.ita that a subtracted positive number becomes
a negative number, therefore, subtracting a number is only adding its
opposite. There is a sign to denote negative numbers: a dot is placed
over them; applying this notation to the karan. leads to:

ka 8 ka 2
meaning that the karan. 2 is subtracted from the karan. 8.
As long as only karan.s are considered, no problems occur: the sub-
traction rule applies as it is formulated: the dierence is the karan.
The Karan
.
129
p
measured by the number 8 + 2 2 2 8 = 2. But in his example,
Bhaskara says: \(. . . ) Or the multiplier is the two karan. measured
by three and twelve less the integer ve" and according to the Indian
notation system, we have to write:

ka 3 ka 12 ru5
Because karan. and integers are mixed in this multiplier, we have to
square the integer in order to transform it into a karan. and when doing
this, we will loose the \negative sign" showing that the last component
must be subtracted.
To solve this problem, Bhaskara introduced a restriction to the gen-
eral rule which says that the square of a negative quantity is positive:
The square of negative integers will also be negative if it is

calculated for the reason of a state of karan.. Likewise, the
square root of a karan. the nature of which is negative will
be negative for the reason of creation of a state of integer.13

With this rule, the multiplier becomes: ka 3 ka 12 ka 25, which

can be simplied
p as ka 27 ka 25, by the addition of the karan.s 3 and
12 (3 + 12 + 2 3 12 = 27). Now, the multiplication can be performed
in the same way as in the rst example; let us summarize this with a
table:

ka 25 ka 27
ka 3 ka 25 ka 3 ka 25

ka 75 ka 625 ka 81 ka 675
Depending on the commentator, the result can be simplied in more

than one way: noticing that 81 and 625 are squares, we can give them
back their state of integers and subtract one from the other, because the

square root of the \negative square", 625, remains negative according

u 16. The two remaining karan.s, ka 75
to the last given rule; we get r
13
ks.ayo bhavec ca ks.ayarupavargas cet sadhyate 'sau karan.tvahetoh. r.n.atmikayas
ca tatha karan.ya mulam. ks.ayo rupavidhanahetoh.
130 Fran
cois Patte
and ka 675 can be subtracted, for 675 75 = 50625 is a square and we

get ka 300.
It may seem strange for a modern mathematician to state a general

rule, such as: \the square of a positive or a negative number is a positive
number", then to restrict its range of application by another rule which
may even contradict the general one. Nevertheless this is found very
often in mathematical Sanskrit texts because the paradigm of logic in
Sanskrit scientic knowledge is grammar rather than mathematics. We
can see an example of this here and this procedure repeatedly occurs
in Pan.ini's grammar which is the fundamental text of Indian scientic
tradition.
There are two more operations to complete the six operations de-
scribed in the Bjagan.ita : division and square root.
Division is easy to perform | and Bhaskara does not give any rule for
this, only examples | because the algorithm given for the multiplication
is the same as the one given for the unknown quantities and thus, as
is shown by the two preceding tables, the division is very similar to
today's Euclidean division of polynomials: it is suÆcient to read these
tables in the reverse order, making the third row the dividend and the
rst one the divisor to nd out that the middle row is the quotient of the
division. All the examples given for the multiplication are used in this
way to explain the division in both chapters: the one about unknown
quantities and the present one on the karan..
The square root is rather diÆcult and we shall not discuss it in this
paper. Let us just say that its algorithm is based on the identity:
(a + b + c)2 = a2 + b2 + c2 + 2 ab + 2 ac + 2 bc
which the mathematicians inverted in order to nd out the quantities

a, b, c from the left member of the identity. We have just written down
three quantities, but there are some examples given by Bhaskara with
more than three karan.s.
The Karan
.
131
3.3 The use of the karan

.
What was the purpose of Indian mathematicians when they constructed

these karan.s ? They knew perfectly
p well how to calculate the approxi-
mate values of a square root A, using the rst two or three values of
the sequence:
an+1 = 12 (an + aA )
n
Let us consider what Kr.s.n.a says about it (this text is the paragraph
that follows the text quoted on page 121):
\But it may be argued that these are only words! Why, then, trouble
yourself to study these operations on karan.s for, in common practice,
there is no use of karan.s but only of the approximate values of their
square roots and, with the use of six operations on numbers, these six
operations [on karan.s] are meaningless. Moreover, even if the calcula-
tion with karan.s is done, in common use, [calculation] with approximate
square roots from the beginning is better than this and is preferable to it.
| This is not correct. If a rough square root is taken from the
beginning, there will be a big roughness in its multiplication and so on;
but if the calculation of karan.s, which is minute, is performed, later,
when the approximate square root is taken, there will be some dierence
but not very much; for this great distinction, the six operations with
karan.s must necessarily be undertaken."
As already seen this justication for the construction of karan. takes
the form of a dialogue; an opponent develops the idea that this con-
struction is useless and that only approximate calculations of square
roots is enough for everyday transactions. He is told in return that if
many operations are done with approximate values of square roots the
nal error is much bigger than if the calculations were done through the
sophisticated construction of karan. and if the approximate value of the
result is taken at the very end of the calculation.
The next paragraph justies the location of this chapter in the com-
plete book of Bhaskara's mathematical works; in Sanskrit commentaries,
it is mandatory to give the reasons why a particular subject is studied
132 Fran
cois Patte
and to justify its place in the succession of the topics developed by an

author.
\Although it is suitable that these [six operations on karan.] should
be undertaken before the six operations on varn.a14 , because they are
closer to the operations on manifested numbers, according to the maxim
of the needle and the kettle15 , it is however suitable to undertake them
immediately after the operations on varn.a for the reason that a great
eort is required by their examination and understanding."
References
[1] See, for instance, Bjagan.ita of Bhaskara with the commentary
Suryaprakasa of Suryadasa, vol. I, ed. Pushpa Kumari Jain, Gaek-
wad's Oriental Series, No.182, Vadodara, 2001.
[2] See, for instance, Bjagan.ita of Bhaskara with the commentary
Bjapallavam of Kr.s.n.a Daivaj~na, ed. T.V. Radhakrishna Sastri, Tan-
jore Saraswathi Mahal Series No.78, Tanjore, 1958.
[3] See, for instance, Siddhantasiroman. i of Bhaskara with Vasanabhas.ya
and Vasanavarttika, ed. Muralidhara Chaturveda, Varanasi, 1981.
14
Unknown quantities.
15
This maxim is used to show that when two things | an easy one and a diÆcult
one | must be done, the easier one should be rst attended to just like when you
have to prepare a needle and a kettle, you should rst take up the preparation of the
needle as it is an easier work compared to the preparation of a kettle. See Apte's
Practical Sanskrit-English Dictionary.
The Karan
.
133
Appendix: Sanskrit texts
Suryadasa, page 120:

athaitasyah. karan.ya gun.anavidhikathanavyajena svarupam. nirupa-
yann aha vargen.e ti|
vargen.a vargankena
_ vargam. gun.ayet tatha vargen.aiva vargam. bha-
jen na param. tu rupen.a vargam. gun.ayed bhajed vety arthah. anena
karan.tvam. nama vargatvenabhimatankatvam
_ . sucitam. bhavati tad uk-
tam. narayan.ena
mulam. grahyam. raser yasya tu karan.nama tasya syat

iti
Kr.s.n.a, page 121:
atha karan.s.ad.vidham. vyakhyayate atredam avagantavyam. mula-
rasyor vargadvara yat s.ad.vidham. tat karan.s.ad.vidham iti asya s.ad.-
vidhasya vargatvapuraskaren.aiva pravr.tter ata evasmin s.ad.vidhe mula-
darasav api karan.tvavyavaharah. karan.tvapuraskaren.a gan.itapravr.ttav
ayam. na syat karan.s.ad.vidham iti sam.j~na tu karan.rasav etasya gan.ita-
syavasyakatvad dras.t.avya tatra yasya raser mule 'peks.ite niragram.
mulam. na sam.bhavati sa karan. na tv amuladarasimatram tatha sati
dvitripa~ncas.ad.adis.u sarvada karan.tvavyavaharah. syat
| astu sa iti cet
na tatha sati tatprayuktam. karyam. syat yathas.t.au dvisam.yuta
as.t.adasaiva syur ity adi
Kr.s.n.a page 125:
\vargen.a vargam. gun.ayed bhajec ca" iti etad uktam. bhavati|
karan.gun.ane kartavye yadi rupan.am. gun.yatvam. gun.akatvam. va syat
karan.bhajane va kartavye yadi rupan.am. bhajyatvam. bhajakatvam. va
syat tada rupan.am. vargam. kr.tva gun.anabhajane karye karan.ya var-
garupatvad iti
Kr.s.n.a page 125:
vargasyapi samadvighata taya gun.anavises.atvad uktavat siddhih.
\sthapyo 'ntyavargo dvigun.antyanighna" ityadina vyaktoktaprakaren.a
134 Fran
cois Patte
va karan.vargasyapi siddhih. syat kim. tu \vargen.a vargam. gun.ayed"

ityuktatvad dvigun.antyanighna ity atra caturgun.antyanighna iti dras.t.a-
vyam
Kr.s.n.a page 131:
nanv astu paribhas.amatram idam. tathapi kim anena karan.s.ad.vidha-
nirupan.asramen.a na hy asti loke karan.bhir vyavaharah. kintu tadasanna-
mulair eva tats.ad.vidham. ca rupas.ad.vidhenaiva gatartham kim. ca kr.te
'pi karan.gan.ite tatas tadasannamulenaiva vyavaharah. tadvaram. prag
eva tadadara iti cet
maivam prag eva sthulamulagrahan.e tadgun.anadav atisthulata
syat kr.te 'pi suks.me karan.gan.ite pascat tadasannamulagrahan.e kim.
cid evantaram. syan na mahad ity asti mahan vises.a iti karan.s.ad.vidham
avasyam arambhanyam
tad yady api vyaktas.ad.vidhantarangatv
_ ad varn.as.ad.vidhat prag eva-
rabdhum. yuktam. tathapy etasya nirupan.avagamayoh. prayasagauravat
suckat.ahanyayena varn.as.ad.vidhanantaram arambho yukta eva
Relations between Approximations to
the Sine in Kerala Mathematics
Kim Plofker
Department of Mathematics, University of Utrecht,
and
International Institute for Asian Studies, Leiden
kim plofker@yahoo.com
Abstract
The mathematicians of the school of M adhava in late me-
dieval Kerala (South India) described various series expressions for
trigonometric quantities. This paper examines the reasoning un-
derlying two such formulas and the possible connections between
them.
1 Introduction
The school of Madhava in fteenth- and sixteenth-century Kerala is fa-
mous for its brilliant mathematical discoveries in a number of areas,
including astronomical modeling, trigonometric series, innitesimals, it-
erative functions, and solution of equations. Some of their best-known
work involves expressions (attributed to Madhava himself) for the Sine
and Cosine that are equivalent to what we now call Taylor series ex-
pansions. The modern name can be misleading, however, since formulas
that in modern mathematics represent essentially the same concept can
dier widely in their original Sanskrit context. Exploring the apparent
motivations of these rules, as suggested by their accompanying \yuktis"
or rationales, provides a better understanding of their relationships and
their underlying concepts of trigonometric series.
135
136 Kim Plofker
Most of the results discussed here date back to the time of Madhava
himself or his student Paramesvara, but the yuktis explaining them in
detail did not appear, as far as the existing texts indicate, until about a
century later. It is not known how they originated: they may have been
written along with the original results in texts now lost, or passed down
orally from teacher to pupil, or devised by later members of the school
to make sense of the enigmatic formulas, or some combination of all of
these.
2 The Sine and Cosine series in the sources

We begin our examination of the Kerala power series by quoting their
formulations as they appear in Sanskrit sources, translating them as
literally as possible into modern mathematical notation, and showing
their relation to the modern forms of these series.
2.1 The \Madhava-Newton" power series

The following statements about the Sine (here capitalized to indicate
that it is scaled to a non-unity trigonometric radius R, instead of being
scaled to 1 like the modern trigonometric functions) and Versine (the
\versed sine", equal to the Radius minus the Cosine) are part of a yukti
laid out by one Sa_nkara in the rst half of the sixteenth century. They
form part of his commentary Yuktid
pik
a (\Lamp of Rationales") on
an astronomical text called Tantrasangraha
_ (\Epitome of Astronomical
Treatises") written in 1500 by his teacher Nlakan.t.ha, who studied under
the son of Madhava's student Paramesvara. (See [12], 184{190, in which
similar statements are attributed to a manuscript of the Tantrasangraha
_
itself, [7], 169{173, and [9].)

The yukti is derived from a Malayalam exposition by Jyes.t.hadeva,
another of Sa_nkara's teachers, entitled or \Vernacular [Ex-

Yuktibh
as.a
position] of Rationales". This part of Sa_nkara's commentary applies
to the beginning of the second chapter, where the , like
other Sanskrit astronomical treatises, introduces trigonometry. Sa_nkara
Tantrasangraha
_
is explaining ([13], 118) a way to nd accurate Sine and Versine values

for a given arc (measured in arcminutes):
Approximations in Kerala Mathematics 137
[
Yuktid
pik
a 2, 440{441:] Having multiplied the arc and the
results of each [multiplication] by the square of the arc, di-
vide by the squares of the even [numbers] together with
[their] roots, multiplied by the square of the Radius, in order.
Having put down the arc and the results one below another,
subtract going upwards. At the end is the Sine: : : ([3], 57)
[
Yuktid
pik
a 2, 442{444:] Having multiplied unity and the
results of each [multiplication] by the square of the arc, di-
vide by the squares of the even [numbers] minus their roots,
multiplied by the square of the Radius, in order. But divide
the rst [instead] by twice the Radius. Having put down the
results one below another, subtract going upwards. At the
end is the Versine: : :
In the computation for the Sine discussed in verses 440{441, the initial
\result" is , the second , the third , and so
2 2 2 2 2 2
forth. For n = 1; 2; 3; : : :, the nth result is divided by R times a term 2
depending on the corresponding even number 2n: namely, (2n) + 2n. 2
Thus the sequence of results will be

3 3
=
R (2 + 2) R 62 2
; 2

= ;
5 5
R (2 + 2)(4 + 4) R 120
4 2 2 4
7
7
6 2 2
=
R (2 + 2)(4 + 4)(6 + 6) R 5040 2
; (1)
6
and so forth. Since, for the nth such result, the denominator term
Y
n
Y
n
((2i) + 2i) = (2i(2i + 1)) = (2n + 1)!;

2
(2)
i=1 i=1
it follows that Sa_nkara's sequence of recursive subtractions of terms from

the next higher term will give

3 5 7
Sin = R2 3! R4 5! R6 7!
::: : (3)
138 Kim Plofker
The Versine computation in the subsequent verses similarly gives

2 4 6
Vers = R2 R3 2(42 4) R5 2(42 4)(6 6)
2
:::

2 4 6
= R 2! R3 4! R5 6!
::: : (4)
Since Kerala trigonometry uses the standard value for R of about 3438
360 60=2, the division of 0 by R produces what we would call in
( )
radians. This means that the above expressions are equivalent to the
following modern series normalized to R = 1:
sin(x) = x x3! + x5! x7! + : : :
3 5 7
1 cos(x) = x2! x4! + x6! : : :

2 4 6
(5)
Other verses quoted by Sa_nkara in this yukti ( 2, 437{439:
Yuktid
pik
a
[13], 117{118), and elsewhere ascribed to Madhava, contain numerical
values for the coeÆcients (up to the fth-order term) of a slightly dier-
ent form of the same series for Sine and Versine. Hence it is clear that
these power series were known (and probably originally developed) by
Madhava himself.
2.2 Madhava and the \Taylor-series-like approximation"
Near the beginning of its second chapter, the states a
Tantrasangraha
_
rule for determining the Sine of some arbitrary arc by means of Sines
and Cosines from a Sine-table. More detailed analyses of this rule, and
of an elaboration of it recorded by Madhava's student Paramesvara,
are given in [4], [5], and [10]. Brie y, choosing from the Sine-table the
tabulated arc closest to , and knowing the tabulated Sine and Cosine
of , we manipulate them as follows:
[
Tantrasangraha
_ 2, 10cd{12ab:] Having set down the two
composite Sines [i.e., the tabulated Sine and Cosine of the
tabulated arc] closest to the [arbitrary] arc whose Sine and
Cosine are sought, one should compute the arc [of] deciency
or excess [depending on whether the desired Sine or Cosine

is less or greater than the tabulated one].
And set down as a divisor 13751 divided by twice the arcmin-
utes of that [dierence arc], for the purpose of the mutual
correction of those [quantities]. Having rst divided [either]
one [of the tabulated Sine or Cosine by that divisor], add or
subtract [the result] with respect to the other, according as
the [dierence] arc is excessive or decient.
Now in the same way, apply that [corrected quantity] times
two to the other: this is the correction. ([13], 112)
The dierence ( ) is found in arcminutes, and used to compute
a \divisor" D:
D=
13751 = 2 3437;45 2R : (6)
2( ) ( ) ( )
The Sine and Cosine of are \mutually corrected" after being divided
by D, to give a rst approximation to the Sine and Cosine of :
Sin Sin + CosD ; Cos Cos SinD : (7)
The sign of the correction term depends on whether the function of
is greater or less than that of . These approximate results are then
doubled, divided by D, and applied in another \mutual correction":

Sin
Sin Sin + Cos D D ; 2

Cos Cos Sin + Cos 2
: (8)
D D
Recalling that D 2R=( ), and again normalizing to R = 1 as for
modern trigonometric functions, we get equivalent expressions with x
and a in the place of and :
sin(x) sin(a) + cos(a) (x a) sin(2 a) (x a) ;
2
cos(x) cos(a) sin(a) (x a) cos(2 a) (x a) :2

(9)
Nlakan.t.ha describes this method too as \spoken by Madhava" ([13],
120).
140 Kim Plofker
2.3 The common equivalent in the modern Taylor series

Both sets of Madhava's Sine and Cosine rules, as represented in equa-
tions (5) and (9), can be derived from the general Taylor polynomial
f 00 (a) f 000(a)
f (x) = f (a) + f 0 (a)(x a) + ( x a) +
3! (x a) + : : : (10)
2 3
2!
where f (x) is the sine or cosine function whose value at x is sought.
Equation (9) gives the rst three terms of the Taylor series, while equa-
tion (5) is equivalent to the corresponding Maclaurin series where a = 0.
The Taylor and Maclaurin expansions per se , of course, are theo-
retically and historically very closely intertwined. They emerged in the
seventeenth and eighteenth centuries, in forms very similar to their mod-
ern representation, as part of the early calculus toolkit for dealing with
arbitrary functions in the form of polynomials. The Maclaurin series
was from the beginning considered as a special case of the Taylor series.
3 The series as explained in their yuktis

It is clear that the context in which Madhava developed his \mutual
corrections" of the Sine and Cosine or Sine and Versine was very dierent
from the generalized techniques of successive dierentiation employed
by Taylor, Maclaurin, and their predecessors. What was the relation, if
any, between Madhava's two methods as perceived by the members of
his school?
3.1 Correcting Sines by Versines
Sa_nkara's exposition culminating in the \Madhava-Newton" series is de-
tailed and intricate, comprising nearly a hundred Sanskrit verses ([13],
109{118). We can only touch upon its highlights here: more thorough
summaries (though mostly without translations) of Jyes.t.hadeva's ratio-
nale, from which Sa_nkara derived this explanation, may be found in [11],
[7], [9], [12], and [14].
The foundation of the yukti is its division of a given arc into n unit
arcs . Each unit arc is bisected by a radius. For each successive arc
∆θ
∆θ
R
∆θ
Figure 1: The quadrant with radius R, divided into unit arcs
= i, the Sine and Cosine line segments Sin and Cos are drawn,
as are the \medial" Sines and Cosines at the bisection points, Sin and
i i i
i:5
Cos .i:5
As Figure 1 illustrates, the similar right triangles thus produced

imply that the dierences Sin and Cos between pairs of successive
Sines and Cosines can be found from the medial Cosines and Sines.
Each triangle with hypotenuse R and vertical leg Sin is similar to the
smaller triangle with hypotenuse Crd and horizontal leg Cos , as
i
in the pair of shaded triangles in the gure. As Sa_nkara puts it,

i:5
[
Yuktid
pik
a 2, 349{352:] : : : The change in the dierence be-
tween Sines produced from the junctions of the correspond-
ing arcs should be [proportional] to the change in the Cosine
produced from the center of the arc. [Likewise,] the dier-
ence in Sine produced from the center of the arc is known as
the change in Cosine-dierence at the junctions of the corre-
sponding arc. Thus the changes in dierence are mutual[ly
dependent].
Then divide the Sine produced to the junction of the rst arc,
142 Kim Plofker
multiplied by the Chord, by the Radius: [it is] the dierence

of the Cosines in the centers of those arcs.
In our notation,
Cos = Sin Crd
i+1
R
; Sin = Cos ;
i:5
Crd
Ri+1 i:5
i:5
Crd
Cos = Sin R ; Sin = Cos R :
i
Crdi:5 (11)
i
For any consecutive pair of, e.g., Sine-dierences, their \second dier-
ence" or \dierence of the dierences" Sin also depends on these
quantities:
[Yuktid
pik
a 2, 352{354:] : : : Having multiplied that [Cosine-
dierence, Cos ] by the Chord, divide again by the Ra-
1:5
dius. [That] should be the dierence of the dierences of the
Sines produced to the junctions of the rst and second [arcs].
Therefore, divide the rst Sine, multiplied by the square of
the Chord, by the square of the Radius. [The] quotient is the
dierence of the dierences of the rst and second [Sines].
[Yuktid
pik
a 2, 358{359:] : : : In the same way, the division
by the square of the Radius of every Sine multiplied by the
square of the Chord should be similarly the dierence of [its
own dierence and] the dierence after that.
That is to say, in terms of the relations from equation 11,
Sin = Sin Sin = (Cos
i i i+1 Cos ) Crd = Cos Crd
i:5 1 i:5
R i:5
R
= Sin Crd
2
i
R2
: (12)
Consequently, any desired Sine-dierence Sin (for i > 0) can i+1
be expressed in terms of the previous Sine-dierence and the second-

dierence between them. But the previous Sine-dierence itself can be
expressed in terms of another second-dierence, and so on, until we are
left with a sum of second-dierences. And each second-dierence, we
recall from equation (12), is dependent upon the corresponding Sine, as
Sa_nkara goes on to state:
[Yuktid
pik
a 2, 361{363:] However many dierences of [Sine-
]dierences there are, when the sum of that many, beginning
with the rst, is subtracted from the rst dierence, the
desired dierence is obtained. When it is desired to obtain
the total of the dierences of [Sine-]dierences, [the quotient]
from the sum of Sines multiplied by the square of the Chord,
divided by the square of the Radius, should be the sum of
the dierences of the dierences.
Sin = Sin
i+1 i Sin i
= Sin i 1 Sin i 1 Sin i
... ... ... ...

= Sin 1 Sin 1 ::: Sin i
Xi
= Sin
1 Sin k
k =1
Sin Crd
X
i 2
= Sin
1 k
R 2
: (13)
k =1
From equation (13) we can see that if we were to add up i consecutive

Sine-dierences, the result would involve a sum of sums of Sines (or, al-
ternatively, of Cosine-dierences, which are equal to Versine-dierences).
But the sum of i consecutive Sine-dierences is also just the ith Sine it-
self:
[Yuktid
pik
a 2, 367{371:] Therefore, many sums of Sines one
below another, ending with the sum of the rst and sec-
ond, are to be made here successively. [The quotient] from
the sum of all those [sums], multiplied by the square of the
Chord, divided by the square of the Radius|which is the
result produced from the sums of the Sines|is equal to the
quotient [from division] by the Radius of the sum of [the
successive sums of] the Versine-dierences, times the Chord.
However many dierences are considered in the case of [any]
desired arc, subtract that [sum of sums] from the rst dif-
ference times that number of dierences. That should be
144 Kim Plofker
the Sine of the desired arc in the form of a sum of Sines of

[arc-]portions.
Sin Crd
X
i 1
X
k 2
Sin +Sin + : : : + Sin = iSin
1 2 i 1 j
R 2
k =1 j =1
Vers Crd
X
i 1
X
k
= iSin 1
R
j:5
k =1 j =1
= Sin : i (14)
Sa_nkara now simplies the above expression by taking Sin to be 1
approximately equal to the unit arc :
[Yuktid
pik
a 2, 375{376:] Because of the uniform smallness of
the arc-portions, the rst Sine-dierence is assumed [equal
to the unit arc]. That [Sin ] multiplied by the number
1
of dierences is the amount of the desired arc. Therefore,
subtract from the desired arc the result produced as stated.
The desired Sine remains there; it is [made] accurate by the
rule to be stated [below].
So we can rewrite equation (14) as
Sin Crd Vers Crd
X
i 1
X
k 2 X
i 1
X
k
Sin
i i j
2R
= i
R
j:5 : (15)
k =1 j =1 k =1 j =1
The process of \making the desired Sine accurate" commences with

several more simplifying assumptions:
[Yuktid
pik
a 2, 379{382:] : : :Whatever is the sum-of-sums of
the arcs|owing to ignorance of [the values of] the Sines|is
to be considered the sum [of the sums] of the Sines. But in
that case, the last Sine should be the desired arc: : : Because
of the smallness [of the unit], those arc-portions are consid-
ered [to be composed] with unity [i.e., as integers]. So the
integers in the desired arc are equal to that [arc]. Therefore
the sum of the Sines is assumed from the sum of the num-
bers having one as their rst term and common dierence.
That, multiplied by the Chord [between] the arc-junctures,
is divided by the Radius. The quotient should be the sum of
the dierences of the Cosines drawn to the centers of those
arcs.
[
Yuktid
pik
a 2, 384{385:] The other sum of Cosine-dierences,
[those] produced to the arc-junctures, [is] the Versine.[But]
the two are approximately equal, considering the minuteness
of the arc-division. On account of considering the arc-units
[equal] to unity because of [their] minuteness, the Chord too
is equal to that [unity], which makes no dierence in the
multiplier.
Sa_nkara points out, quite rightly, that we cannot use a sum of i 1 sums
of unknown Sines to compute an equally unknown ith Sine. He proposes
instead taking the sum of i sums of the successive arcs, which (if we take
Crd 1) is just a sum-of-sums of successive integers, whose
values are of course known. We will also assume that a sum of medial
Cosine- (or Versine-) dierences, which really equals a medial Versine,
is roughly equal to the adjacent non-medial Versine. This gives us the
following modied expression for Sin : i
Sin i
X
i
X
k
j
1 i X
i
X
k
Vers R1 i
X
i
Vers R1 : (16)
i
k =1 j =1
R2
k =1 j =1
j
k =1
k
A brief detour then provides a general expression for sums of successive

integers and their successive sums:
[
Yuktid
pik
a 2, 386{387:] Whatever is the product of however
many numbers beginning with the rst-term and increasing
[successively] by one, that [product] is divided by the product
of that many numbers beginning with one and increasing by
one. The results one after another are the sums-of-sums of
those [numbers].
That is, for a sequence of \however many" (say, q) consecutive integers
146 Kim Plofker
beginning with any \rst-term" p, Sa_nkara notes that

p (p + 1) (p + q 1) (p + q 1)! X p
X X j3 j2
1 2q = (p 1)!q! = ::: j : (17) 1

jq 1 =1 j2 =1 j1 =1
Yet another simplifying assumption follows:

[
Yuktid
pik
a 2, 389{393:] Half that square of the arc should
be the sum of the [consecutive] arcs. Because half the prod-
uct of the rst-term and the rst-term plus one is a sum:
therefore, from the cube and squared-square [etc.] of the de-
sired arc, divided by the product of numbers beginning with
one and increasing by one, there are many resulting sums
one after another.
So in this case, the rst sum of the arc-portions: : : should
be considered the sum of the Sines. But that sum of Sines,
multiplied by the Chord and divided by the Radius, is the
desired Versine.
We know from the familiar rule for the sum of an arithmetic series that
half the product p(p + 1) equals the sum of the rst p integers. Sa_nkara
now assumes that, since the desired arc i i + 1, the sum of the rst i
integers is approximately i =2, and generalizes that assumption (relying
2
on an earlier demonstration in his commentary) to conclude that

X X
i j2
i(i + 1)(i + 2) i X k
3 i 2
j =
j2 =1 j1 =1
1
123 6 2 ; k =1
X X X
i j3 j2
i(i + 1)(i + 2)(i + 3) i Xk 4 i 3
j =
j3 =1 j2 =1 j1 =1
1
1234 24 6 ; (18)k =1
etc. From this approximation and from equation (16), we can assume
that
X 1 i
k Vers :
i 2
(19)
R k =1
2R
i
We can also make a substitution for the double sum of integers in

equation (16):
[Yuktid
pik
a 2, 401{403:] : : :But the dierence of the desired
Sine and [its] arc is from the sum of the sum of the Sines.
The square of the desired arc should be divided by the Ra-
dius; thence is [found] the Versine. But thence whatever
[results] from the cube of the same desired arc [divided] by
the square of the Radius, from that the quotient with six is
approximately inferred [to be] the dierence of the arc and
Sine: : :
That is,
i Sin
X
i
X
k
j
1 i 1:
3
(20)
i
k =1 j =1
R2
6 R 2
Now Sa_nkara reminds us that this double sum of consecutive integers

or arcs should really have been a double sum of Sines, so every arc in
the sum needs to be corrected by the above expression for the dierence
between the arc and its Sine:
[Yuktid
pik
a 2, 406{408:] To remove the inaccuracy [result-
ing] from producing [the Sine and Versine expressions] from
a sum of arcs [instead of Sines], in just this way one should
determine the dierence of the [other] Sines and [their] arcs,
beginning with the next-to-last. And subtract that [dier-
ence each] from its arc: [those] are the Sines of each [arc]. Or
else therefore, one should subtract the sum of the dierences
of the Sines and arcs from the sum of the arcs. Thence should
be the sum of the Sines. From that, as before, determine the
sum of the Versine-dierences: : :
[Yuktid
pik
a 2, 417{419:] : : :[W]hatever sum-of-sums is in-
ferred from the determination of the Versine, the dierence
of the Sine and [its] arc is deduced from the sum-of-sums
after that [one]. Thence in this case, the quotient from the
product of the cube and square of the desired arc, [divided]
by whatever is the product of ve numbers beginning with
one and increasing by one, [is] also divided by the squared-
square of the Radius. The dierence of the Sine and [its] arc
determined in this way should be more accurate.
148 Kim Plofker
All the corrected consecutive terms must be re-summed to produce a

more correct Versine, using the approximation for successive sums-of-
sums in equation (18):
X i
k

1 =X k X k i
3 i
i
i 3 2 4
Vers
i k
6R R 2
R 6R 2R 24 R : (21) 3 3
k =1 k =1 k =1
And we can use the same sort of term-by-term correction to modify all
the Versines composing the Sine in equation (16):
X 1i
X k k
i
1 2 4
Sin i i Vers R i k
2R 4!R R 3
k =1 k =1
!
X
i
k2 X k4 i
=i 2R2 =1 4!R4
=1 k k
i3 i5
i +
3!R 5!R4 :
2
(22)
These recursive corrections of the Sine by the Versine and vice versa
are then applied indenitely to produce the general rules of equations (3)
and (4).
3.2 Correcting Sines by Cosines
We turn now to the 's \Taylor-series-like approxima-
Tantrasangraha
_
tions" of equation (8) to see what we can glean of the reasoning be-
hind them. Consider rst of all that if we omit their nal terms, we get
expressions equivalent to the following:
Sin Sin Cos D2 ; Cos Cos Sin D2 ; (23)
or equivalently
Sin Sin Cos Cos
2R 2Cos 2Sin : (24)
As discussed in ([10], 287{288), Sa_nkara analyzes in a dierent com-
mentary on the same text (
1
, or \Minor Commentary") an
Laghuvivr
. ti
1
Sa_nkara does not comment on this approximation in the . I'm in-
Yuktid
pik
a
debted to the reviewer of this paper for the important observation that such a yukti
is in fact provided in the of Jyes.t.hadeva.
Yuktibh
as.a
approximation very close to the rst of these, namely

Sin Sin
2R Cos + Cos ; (25)
as follows:
[
Laghuvivr
. ti onTantrasangraha
_ 2, 14cd{15ab:] Here, where
the divisor should be made from the Cosine of the medial
arc, it is said [to be made] with the sum of the Cosines of
both full [arcs], by assuming that that [sum] equals twice the
medial Cosine. But in reality, the sum of the Cosines of the
two full [arcs] is somewhat less than twice the medial Cosine.
Because of the deciency of that divisor, the result of that is
somewhat too big. But actually, that is what is desired: for
that result is really the Chord, which is a little less than its
arc: : :
In other words, if we considered the slightly smaller Crd( ) rather
than its arc, it would be more appropriate to write:
Crd( ) Sin Sin : (26)
2R 2Cos + 2
But this is essentially the same similar-triangle relation expressed in
equation (11): it states the dierence between two successive Sines in
terms of a linear proportion involving the Chord of the arc-dierence,
the Radius, and the medial Cosine. This exact linear proportion is what
validates the slightly adjusted one, using the arc instead of the Chord,
in equation (24). The \Madhava-Newton series" and the \Taylor-series-
like approximation" are therefore ultimately based on the same Rule of
Three derived from similar right triangles.
Sa_nkara does not explain here the reason for adding the second-order
correction term, but he proposes the option of including several more of
them (see [10], 292):
[
Laghuvivr
. ti onTantrasangraha
_ 2, 10cd{14ab:] Although
here, prior to that, the quotient from half the Cosine with
that same divisor [can] be applied to the Sine|and prior
150 Kim Plofker
to that, the quotient-result from a fourth part of the Sine

to the [half-]Cosine, and prior to that [the result] from an
eighth part of the Cosine to the [fractional] Sine and [simi-
larly, the result] from a sixteenth part of that [Sine] to the
[fractional] Cosine|yet because of the smallness of that, it
is to be considered negligible.
That is, the approximations in equation (8) might be expanded (if the
size of the correction made it worthwhile) to

Sin Sin + Cos Sin + 2 Cos Sin
+ Cos

4 8
Sin 1 1 1 1 1 2 ;
16 D D D D D D
Cos Cos Sin + Cos Sin
+ Cos Sin

2 4 8
+ Cos 1 1 1 1 1 2
16 D D D D D D : (27)
Recalling that D = 2R=( ), we may rewrite these as
Sin Sin + Cos ( ) Sin 2(R ) Cos 8(R )
2 3
2 3
+ Sin 32(R ) + Cos128

( ) Sin ( )
4 5 6
4 R 5 512R ; 6
Cos Cos Sin ( ) Cos 2(R ) + Sin 8(R )

2 3
2 3
+ Cos 32(R ) Sin 128

( ) Cos ( )
4 5 6
4 R 5 512R : 6
(28)
And if we set = 0 in order to compare these approximations with their
counterparts in equations (3) and (4), they reduce to
Sin 8R + 128 R ;
3 5
2 4
Cos R 2R + 32R 512 R

2 4 6
3 5
: (29)
3.3 Conclusion: the same or dierent series?

These commentaries of Sa_nkara contain almost the last known treat-
ments within Madhava's school of these two types of approximation
to the Sine and Cosine. (The anonymous source in [3] appears to
be somewhat later.) As understood in this mature form by Sa_nkara,
the \Madhava-Newton" series and the \Taylor-like" approximation are
clearly independent rules. The former is valued largely for its produc-
tion of Sine-values without the need for Sine-tables; the latter depends
on a known tabulated Sine and Cosine to determine those of a nearby
arc. Even if reduced to an equivalent form as in equation (29) (and there
is no evidence that the \Taylor-like" approximation was ever actually
handled this way in the Indian texts), they would be incompatible, since
the integer coeÆcients in the denominators of the former are factorials,
while those of the latter are powers of 2. 2
Yet the two series may be linked conceptually. They are both founded
on the Rule of Three arising from the same pair of similar right triangles
in the subdivision of the quadrant. And they are both developed, after
an initial approximation from this Rule of Three, via recursive mutual
corrections of Sine-terms by Cosine-terms and vice versa. The power
and fecundity of these basic concepts is illustrated in the variety of the
approximations derived from them. In particular, the origin of these
brilliant derivations in an elementary linear proportion recalls Bhaska-
ra's comment on mathematical foundations: \Just as this universe is
pervaded by Lord Narayan.a (who removes the suerings of those who
worship him and is the sole generator of this universe), with his many
forms|worlds and heavens and mountains and rivers and gods and men
and demons and so on|in the same way, this whole type of computation
is pervaded by the [rule of] three quantities."
References
[1] Bressoud, David, \Was Calculus Invented in India?", College Math-
ematics Journal 33, 2002, 2{13.

2
The reviewer has noted that the extension of the latter approximation suggested
by Jyes.t.hadeva in his demonstration also uses certain powers of 2 for the coeÆcients,
but they are not the same as Sa_nkara's.
152 Kim Plofker
[2] Feigenbaum, L., \Brook Taylor and the method of Increments",

Archive for History of Exact Sciences 34, 1985, 1{140.
[3] Gold, David, and Pingree, David, \A hitherto unknown Sanskrit
work concerning Madhava's derivation of the power series for sine
and cosine", Historia Scientiarum 42, 1991, 49{65.
[4] Gupta, R. C., \Second order interpolation in Indian mathematics up
to the fteenth century", Ind. J. Hist. Sc. 4, 1969, 86{98.
[5] Gupta, R. C., \An Indian form of third order Taylor series approxi-
mation of the sine", Historia Mathematica 1, 1974, 287{289.
[6] Ikeyama, Setsuro, (trans.), T. Hayashi, T. Kusuba, and M. Yano,
Studies in Indian Mathematics: Series, Pi and Trigonometry (in
Japanese; Tokyo, 1997), forthcoming.
[7] Katz, Victor J., \Ideas of calculus in Islam and India", Mathematics
Magazine 68 (3), 1995, 163{174.

[8] Pillai, S. K.,The Tantrasangraha,
_ a work on ganita by G
argyakerala
N
lakan
. .tha Somasutvan (Trivandrum Sanskrit Series 188), Quilon,
Sree Rama Vilasam Press, 1958.
[9] Pingree, David, \Power series in medieval Indian trigonometry", in
Peter Gaeke and David A. Utz, eds., Science and Technology in
India (Proceedings of the South Asia Seminar 1981/1982) , vol. II,

Philadelphia, University of Pennsylvania 1985, pp. 25{30.
[10] Plofker, Kim, \The èrror' in the Indian `Taylor series approxima-
tion' to the sine", Historia Mathematica 28, 2001, 283{295.
[11] Rajagopal, C. T., and Venkataraman, A., \Sine and cosine power
series in Hindu mathematics", Journal of the Royal Asiatic Society
of Bengal 15, 1949, 1{13.
[12] Sarasvati Amma, T. A., Geometry in Ancient and Medieval India ,
Delhi, Motilal Banarsidass, 1979.
[13] Sarma, K. V., Tantrasangraha
_ of N
lakan
. .tha Somay
aji , Hoshiarpur:
VVBIS& IS, Panjab University, 1977.
[14] Srinivas, M. D., \Proofs in Indian Mathematics", this volume.
Algorithms in Indian Mathematics
M.S.Sriram
Department of Theoretical Physics, University of Madras
Guindy Campus, Chennai - 600 025, India.
tphysmu@satyam.net.in
Abstract
Indian Mathematics is predominantly algorithmic. In fact, the

very word “Algorithm” is derived from the name of Al Khwarizmi
(c. 9th Century) whose works played a crucial role in the transmis-
sion of Indian algorithmic procedures to the Islamic and later to
the Western world. We shall discuss a few selected algorithms that
are representative of the Indian mathematical tradition from the
ancient Śulbasūtrās to the medieval texts of the Kerala School. In
particular, we shall outline some of the constructions described in
the Śulbasūtras, the algorithm for computing the cube-root given
by Āryabhat.a (c.499) and the kut..taka and cakravāla algorithms for
solving linear and quadratic indeterminate equations as discussed
by Āryabhat.a (c.499), Brahmagupta (c.628), Jayadeva (prior to
the 11th century) and Bhāskara (c.1150). We shall also discuss
the efficient algorithms for accurate computation of π and the sine
function due to Mādhava (c.14th century) as discussed in the texts
of the Kerala School of Mathematics and Astronomy.
1 Introduction
The very word algorithm is derived from the name of the famous ninth
century Islamic mathematician Al-Khwarizmi who was greatly influ-
enced by the Indian procedures in mathematics and wrote the famous
book Kitab al-hisab al-hindi on the art of Hindu reckoning [1]. Indian
Mathematics is predominantly algorithmic. It abounds in algorithms
for quick and efficient computations with numbers, and algebraic, geo-
metric and trigonometric quantities. However, it was understood that
153
154 M.S.Sriram
rationale and proofs have to be provided for all mathematical operations

and they are to be found in the commentaries [2].
In this article, we describe some representative algorithms in Indian
mathematics, as they have been developed over the ages. We begin with
the Śulbasūtras, which are adjuncts to Vedas and contain the procedures
for constructing various types of Vedic altars, of which we discuss some
interesting examples. Āryabhat.ı̄ya (c.499) is one of the earliest texts
which gives algorithms for finding the square, cube, square root, cube
root, etc. of a number. Detailed discussions of the same are found in
the commentary on Āryabhat.ı̄ya by Bhāskara I (c.629). We present the
algorithms for finding the square and the cube root given in Āryabhat.ı̄ya
and its commentary. We also describe the Kut..taka or the pulverizer
algorithm for solving linear indeterminate equations.
We next discuss the solution of quadratic indeterminate equations
or Vargaprakr.ti which was first considered by Brahmagupta in his
Brāhmasphut. a-siddhānta (c.628). An optimal algorithm for the same
called Cakravāla has been discussed by Bhāskara II in his Bı̄jagan.ita
(C.1150). The algorithm can at least be traced back to Jayadeva who
lived prior to 11th century. We shall compare the Cakravāla algorithm
with the one discovered by Wallis and Brouncker in the 17th century
and proved later by Lagrange. We then discuss the algorithms for con-
structing the sine tables in Indian astronomy both in their earlier ver-
sions and as given in Tantrasaṅgraha (c.1500) and Āryabhat.ı̄yabhās.ya
of Nīlakan.t.ha Somayājī, the famous Kerala astronomer-mathematician.
Finally, we describe the accurate calculation of π with a suitable re-
mainder term, due to Mādhava (14th cent.) as described in the Ker-
ala works Yuktidı̄pikā (c.1530), a commentary on Tantrasaṅgraha by
Śaṅkara Vāriyar and Yuktibhās.ā (c.1530) of Jyes.t.hadeva.
2 Geometrical construction in Śulbasūtras [3]
Śulbasūtras are part of Kalpasūtras, which are one of the six Vedāṅgas.
These are essentially manuals for geometrical constructions of Vedic
altars. Baudhāyana, Āpastamba, Kātyāyana etc. are some of the
Śulbasūtras. Some of the simpler procedures discussed here are for
dividing a line into equal parts, drawing perpendiculars, constructing
Algorithms in Indian Mathematics 155
squares and rectangles, etc. Pythagoras theorem or the ‘Theorem of the

square of the diagonal’ plays a crucial rule in most of these constructions.
This theorem was known in India, at least by the time of Baudhāyana
Śulbasūtra, where right triangles with sides (3,4,5), (15,3,39), (7,24,25),
etc., are mentioned. The earliest recorded explicit statement of the the-
orem is also to be found there.
ñÍ
d.a:GRa:.ca:tua.=;(ra:~ya.a:[Na:ya.a .=:ê
*.a õ
êÁu H :pa.a.(;Ra:ma.a:na.a ; a.ta:yRa.*+;.a:n
íé ÁÁ a.a ..ca ya:t,a :pxa:Ta:gBUa:tea k
u ++.taH
ta:mua:Ba:yMa k+=:ea: a.ta Á
(Baudhāyana Śulbasūtra 1.48)

The diagonal cord of a rectangle makes both (the squares)
that the vertical and horizontal sides make separately.
We give some illustrative examples of geometrical algorithms in the

Śulbasūtras in the following.
2.1 To draw a square equal to the difference of two

squares
..ca:tua.=;(ra.a:a:tua.=;(rMa ;
a.na:a$Ra:h.a:SRa:n,a ya.a:va:a.a:a$Ra:h.a:SeRa:t,a ta:~ya k+=;Nya.a vxa.;d
Äâ" :mua:a.+.Kea:t,a Á
Äâ" :~ya :pa.a.(õ;Ra:ma.a:na.Ma A:[Na:ya.a I+ta.=;t,a :pa.a.(õ;Ra:m,a o+pa:sMa:h:=e;t,a .sa.a ya.a ;

a.na:pa:tea:t,a
vxa.;d -
ta:d:pa:a..cC+nd
/// ù;a.a:t,
Å a ÁÁ
(Āpastamba Śulbasūtra 2.5)
Wishing to deduct a square from a square one should cut off

a segment by the side of the square to be removed. One of
the lateral sides of the segment is drawn diagonally across to
touch the other lateral side. The portion of the side beyond
this point should be cut off.
ABCD is the larger square. AE is the side of the smaller square. AD

is drawn diagonally across with point A fixed till D touches EF at P .
Considering the triangle AEP .
EP 2 = AP 2 − AE 2
= AD2 − AE 2
156 M.S.Sriram
D F C
A E B
Fig.1
Hence, the square with side EP is the required answer.
2.2 To convert a rectangle into a square

ñÍ
d.a:GRa:.ca:tua.=;(rMa .sa:ma:.ca:tua.=;(rMa ;
a..ca:k +a:SRa:n,a ; a.ta:yRa.*+;.a:nya.a:pa:
íé ÁÁ /// ù;aÅ Zea:SMa ;a.va:Ba.$ya.ea:Ba:ya:ta
a..cC+d
o+pa:d:Dya.a:t,a Á Ka:Nq+ma.a:ga:ntua:na.a .sMa:pUa.=;yea:t,a Á ta:~ya ;

a.na:
Ì+a:sa o+:H Á
(Āpastamba Śulbasūtra 2.7)

Wishing to turn a rectangle into a square, one should cut off
a part equal to the transverse side and the remainder should
be divided into two and juxtaposed at the two sides (of the
first segment). The bit (at the corner) should be filled by
an imported bit. The removal of this has been explained
already.
ABCD is the given rectangle (see Fig.2). Consider the square AB1 C1 D
with side AD. The remaining rectangle B1 BCC1 is divided into two
equal strips B1 B2 EC1 and B2 BCE. The strip B2 BCE is cut off and
applied to the side of the square DC1 . So the original rectangle has been
converted into a square with side AD1 with the small square C1 EC2 F
unfilled at one corner. Hence,
Area of the required square = Area of the square AB2 C2 D1
− Area of the square C1 EC2 F.
Using the previous procedure, this square is easily constructed.
F C2
D
1
D C
C1 E
A B B B
Fig.2
2.3 Combining equal squares
(a) Combining two equal squares

The rule in the Āpastamba Śulbasūtra (1.5) is
.sa:ma:~ya ;a.d
õ :k+=;Na.a Á
The diagonal of the square is the double-maker.
ABCD is the given square with side a and area a2 . Then the square
√
on the diagonal AC is the square with area 2a2 . That is, AC = 2a.
Incidentally, Fig.3 also indicates the proof.
Both Āpastamba (1.6) and Baudhāyana
√ Śulbasūtras (1.62) give a very
good approximation to the value of 2:
:pra:ma.a:NMa txa:ta.a:yea:na va:DRa:yea.a:a:tua:TeRa:na A.a:tma:.ca:tua:a.~:M

/a:Za.ea:nea:na .sa:a.va:Zea:SaH Á
The measure should be increased by one-third of itself, which

1
is again increased by its one-forth and diminished by 34 of
that (second) increment. This is the saviśes.a.
That is,
√ 1 1 1 577
2 1+ + − = 1.414216,
3 3.4 3.4.34 408
which is correct to 5 decimal places.
158 M.S.Sriram
D C
D C
a a
A a B
A a B
Fig.3
(b) An ingenious method for constructing a square with area

n times a given square is given in Kātyāyana-śulbasūtra (6.7)
ya.a:va:tpra:ma.a:Na.a:
a.na .sa:ma:.ca:tua.=;(ra.a:Nyea:k +a:k+.tua ;
a..ca:k +a:SeRa:t,a O;:k+ea:na.a:
a.na ta.a:
a.na Ba:va:a.nta
//
åò
; a.ta:yRa:k, ;a.d
õ :gua:Na.a:nyea:k+.ta O;:k+a: a.Da:k+a:
a.na Á ya: a.~:a:BR a:va: a.ta ta:~yea:Sua:~ta:tk+=:ea: a.ta ÁÁ
As many sides (of equal side) as you wish to combine into

one, the transverse line will be(equal to) one less than that;
twice a side will be(equal to) one more than that. It will be
a triangle. Its arrow(i.e., altitude) will do that.
Let the side of the original square be a. Construct an isosceles triangle

ABC (see Fig.4) with base
BC = (n − 1)a,
and sides
(n + 1)a
AB = AC = .
2
This is easily achieved by stretching the mid point A of a rope with
length (n + 1)a away from BC such that the rope is taut. Then the
altitude AD is the side of the square with area na2 , for,
AD 2 = AB 2 − BD2

(n + 1)a 2 (n − 1)a 2
= −
2 2
= na2 .
(n+1)a
2
B D C
(n−1)a
Fig.4
3 Algorithms for square and the cube root
3.1 Square
One of the earliest example of an algorithm in arithmetic is the one for

squaring a number. In his Āryabhat.ı̄ya-bhās.ya (c.629) Bhāskara I cites
the following as an ancient rule for squaring [4]:
x +.tva.a ;a.d
A:ntya:pa:d:~ya va:ga k õ :gua:NMa ta:de:va:.ca.a:ntya:pa:d:m,a Á
Zea:Sa:pa:dE.=:a:h:nya.a:t,a o+tsa.a:ya.eRa:tsa.a:yRa va:gRa:a.va:Da.Ea ÁÁ
To calculate the square of a number, the square of the last

digit is to be placed over it. And the rest of the digits,
doubled and multiplied by the last are to be placed over them
respectively. Then, (omitting the last digit and) moving the
rest by one place, the process is repeated.
Consider for example, the square of 125. The procedure is summarized

below.
1 5 6 2 5
25 52 = 25
4 20 22 = 4, 5.2.2 = 20
1 4 10 12 = 1, 2.2.1 = 4, 5.2.1 = 10
1 2 5
160 M.S.Sriram
It is noteworthy that this ’ancient rule’ for squaring is already opti-

mal. The mechanical method of squaring a n - digit number involves n2
multiplications, whereas the above method clearly involves only n(n+1)
2
multiplications.
3.2 Cube root
An algorithm for square root is given in Āryabhat.ı̄ya . Indeed, texts

prior to this work also deal with square roots. However, the first clear
enunciation of the cube root algorithm is found in Āryabhat.ı̄ya , which
we proceed to describe [5]
A:Ga:na.a:d, Ba.$ea:d, ;a.d

õ :ta.a:ya.a:t,a ;aa:gua:Nea:na ;Ga:na:~ya mUa:l+va:geRa:Na Á
va:gRa:a.~:a:pU ãÉa ;Ga:na.a:t,a

/ a:vRa:gua:a.Na:taH Za.ea:DyaH :pra:Ta:ma.a:d, ;Ga:na:( ÁÁ
(Having subtracted the greatest possible cube root from the

last cube place and then having written down the cube root
of the number subtracted in the line of the cube root), di-
vide the second non-cube place (standing on the right of
the last cube place) by thrice the square of the cube root
(already obtained); (then) subtract from the first non-cube
place (standing on the right of the second non-cube place)
the square of the quotient multiplied by thrice the previous
(cube root); and (then subtract) the cube (of quotient) from
the cube place (standing on the right of the first non-cube
place) (and write down the quotient on the right of the pre-
vious cube root in the line of the cube root, and treat this
as the new cube root. Repeat the process if there are still
digits on the right).
We consider the cube root of 17,71,561 as an example. Beginning from

the units place, the notational places are called cube place (c), first non-
cube place (n), second non-cube place(n ), cube place(c), first non-cube
place(n), second non-cube place(n ), and so on.
c n n c n n c 121
1 7 7 1 5 6 1 (line of cube root)
Subtract 13 1
Divide by 3.12 3) 0 7 (2
0 6
1 7
Subtract 3.1.22 1 2
5 1
Subtract 23 0 8
Divide by 3.122 432) 4 3 5 (1
4 3 2
3 6
Subtract 3.12.12 3 6
0 1
Subtract 13 1
0
The process ends and the cube root is 121. The algorithm is obviously
based on the algebraic identity: (a + b)3 = a3 + 3a2 b + 3ab2 + b3 .
4 Kut..taka or ‘pulveriser’ algorithm for linear

indeterminate equations
The subject of first order indeterminate equations was considered very

important by the ancient Indian mathematicians and most of them have
dealt with the Kut..taka or pulveriser method to solve them, beginning
with Āryabhat.ı̄ya . They are important in astronomy also, for instance,
in the calculation of Ahargan.a (the number of days elapsed from a given
epoch) from the mean longitudes of planets. This algorithm also plays
a crucial role in the solution of the much more difficult second order
indeterminate equations.
In one of its versions, the problem is to find an integer N which be-
ing divided by two given integers (a, b) will leave two given remainders
(r1 , r2 ). Thus we have
N = ax + r1 = by + r2 .
162 M.S.Sriram
Kut..taka method for this equation is summarized in two verses in

Āryabhat.ı̄ya (Gan.itapāda, 32, 33).
A: a.Da:k+a:g{a:Ba.a:ga:h.a.=M ;
a.C+ndù;a.a:d
Å U :na.a:g{a:Ba.a:ga:h.a:=e;Na Á
å
Zea:Sa:pa.=;~.pa.=;Ba:M ma: a.ta:gua:Na:ma:g{a.a:nta:=e ;a.[a:a:m,a ÁÁ
A:Da o+pa:a=;gua:a.Na:ta:ma:ntya:yua:gUa:na.a:g{a:.cCe+d:Ba.a:a$a:tea Zea:Sa:m,a Á
A: a.Da:k+a:g{a:.cCe+d:gua:NMa ;a.d
õ :.cCe+d.a:g{a:ma: a.Da:k+a:g{a:yua:ta:m,a ÁÁ
Below we give the translation of these verses by Datta and Singh fol-
lowing the interpretation of Bhāskara I [6]:
Divide the divisor corresponding to the greater remainder

by the divisor corresponding to the smaller remainder. The
remainder (and the divisor corresponding to the smaller re-
mainder) being mutually divided, the last residue should be
multiplied by such an optional integer that the product being
added (in case the number of quotients of the mutual divi-
sion is even) or subtracted (in case the number of quotients
is odd) by the difference of the remainders (will be exactly
divisible by the penultimate remainder. Place the quotients
of the mutual division successively one below the other in a
column, below them the optional multiplier and underneath
it the quotient just obtained). Any number below (i.e., the
penultimate) is multiplied by the one just above it and then
added to the one just below it. Divide the last number (ob-
tained by doing so repeatedly) by the divisor correspond-
ing to the smaller remainder; then multiply the remainder
by the divisor corresponding the smaller remainder and add
the greater remainder. (The result will be) the number cor-
responding to the two divisors.
In the following we take r1 > r2 , so that a is the divisor corresponding

to the greater remainder, and b the one corresponding to the smaller
remainder. Let
c = r1 − r2 .
We write down the procedure, when the number of quotients (ignoring

the first one q, as is usual with Āryabhat.a) is even.
b) a (q
bq
r1 ) b (q1
r1 q1
r2 ) r1 (q2
r2 q2
.
.
.
r2n ) r2n−1 (q2n
r2n q2n
r2n+1
Now a number t (mati) is chosen such that r2n+1 t + c is divisible by r2n

with quotient s. Then these are set down in the form of a vallı̄ (column)
and the successive columns are generated:
q1 q1 q1 ...... q1 β2n−1 + β2n−2
q2 q2 q2 ...... β2n−1
. . .
. . .
. . .
q2n−1 q2n−1 q2n−1 β1 + t = β2
q2n q2n t + s = β1
t t
s
Divide q1 β2n−1 + β2n−2 by b. The remainder is x and N = ax + r1 .

When we divide this by b, the quotient is y and the remainder is r2 .
When the number of quotients (omitting q) is an odd integer 2n − 1, the
number t is chosen such that r2n t − c is divisible by r2n−1 .
Example: To solve 45x + 7 = 29y.

Here
a = 45, b = 29, r1 = 7, r2 = 0.
164 M.S.Sriram
29) 45 (1
29
16) 29 (1
16
13) 16 (1
13
3) 13 (4
12
1
Here the number of quotients (omitting the first) is odd. t should be

chosen such that 1 × t − 7 is divisible by 3. Hence t is chosen to be 10.
Therefore we have,
1 1 1 92
1 1 51 51
4 41 41
10 10
1
Now, 92 = 29 × 3 + 5. Therefore,
N = 49 × 5 + 7 = 29 × 8.
Hence,
x = 5, y = 8.
5 Varga-prakr.ti: Quadratic indeterminate

equations
The quadratic indeterminate equation

x2 − Dy 2 = 1,
for a non square integer D is generally referred to as Pell’s equation,

though the 17th century English mathematician Pell had very little to
do with either posing the problem or solving it. The problem of finding
the integer solutions x, y to this equation was posed as a challenge to
the European mathematicians by Fermat in 1653, for specific values of
D = 61, 109, 149, etc. The English mathematicians Brouncker and
Wallis solved the equation. Fermat is credited with proving that the
equation has infinite number of solutions.
In fact, the quadratic indeterminate equations of the form
x2 − Dy 2 = K,
known as Varga-prakr.ti had been considered nearly thousand years ear-
lier by Brahmagupta in his Brāhmasphut. a-siddhānta (c.628). D, the
given non-square integer is called “prakr.ti”, K, a given integer, is called
Ks.epa and the integer solutions x and y are called Jyes..tha-mūla and
Kanis..tha-mūla respectively. The fact that the equation has infinite num-
ber of solutions is implied by Brahmagupta’s Bhāvanā Principle, to be
described below. The equation x2 − Dy 2 = 1, had been solved for all
D by Jayadeva (c.11th century or earlier) as cited by Udayadivākara
(c.11th century) in his commentary Sundarı̄ on Laghu-bhāskarı̄ya of
Bhāskara I [7]. Bhāskara II has also discussed this Cakravāla algorithm
in his Bı̄jagan.ita (c.1150) [8].
The motivation for solving such equations was probably to find rational
approximation for surds. From x2 − Dy 2 = 1, we find

√
D − x ≤ 1 ,
y 2xy
√
so that D ≈ xy , if x and y are large. For instance,
5772 − 2 × 4082 = 1,
so that
√ 577
2≈ .
408
√
It is noteworthy that this is the value of 2 given in Śulbasūtras. The
Cakravāla algorithm makes crucial use of the Bhāvanā Principle of Brah-
magupta which is summarised below.1
1
Further details regarding the Bhāvanā principle may be found in the contribution
166 M.S.Sriram
5.1 Brahmagupta’s Bhāvanā
If
x21 − Dy12 = K1 and x22 − Dy22 = K2 ,
then
(x1 x2 ± Dy1 y2 )2 − D(x1 y2 ± x2 y1 )2 = K1 K2 .
So, if the solution of the Varga-prakr.ti for the Ks.epas K1 and K1 are
known, the solution for the ks.epa K1 K2 can be found immediately [9].
In particular,
x2 − Dy 2 = 1 ⇒ (x2 + Dy 2 )2 − D(2xy)2 = 1
Thus, if one solution of the equation is known, an infinite number of

solutions can be found.
Brahmagupta’s Bhāvanā not only helps in finding any number of solu-
tions from just one solution, it also enables us to solve the K = 1 case,
provided we know a solution for K = −1, or ±2 or ±4. We give the
solutions (x, y) of x2 − Dy 2 = 1 in terms of the solutions (x1 , y1 ) of
x21 − Dy12 = K, when K assumes any of these five values below:
K = −1 : x = x21 + Dy12 , y = 2x1 y1 .

2
x1 + Dy12
K = ±2 : x = y = x1 y1 .
2
2 1 2 2
K = −4 : x = x1 + 2 x + 1 x1 + 3 − 1 ,
2 1

x1 y1 x21 + 1 x21 + 3
y = .
2 2
x1 − 2 x1 y1
K = +4 : x = , y= , if x1 is even ,
22 2
x1 − 3 y1 x21 − 1
x = x1 , y= , if x1 is odd.
2 2
of Amarthya Kumar Dutta in this volume.

5.2 Cakravāla algorithm for the Varga-prakr.ti x2 − Dy 2 = 1
In his Bı̄jagan.ita, Bhāskarācārya gives the Cakravāla algorithm for solv-

ing the equation x2 − Dy 2 = 1 in four verses (71 -75) [10].
:h
ò :~va.$yea:;pa:d:[ea:pa.a:n,a Ba.a.$ya:pra:[ea:pa:Ba.a.$a:k+a:n,a Á
kx +.tva.a k+.pya.ea gua:Na:~ta.a ta:Ta.a :pra:k ãÉyua:tea

x +. a.ta:ta:( ÁÁ
x +.tya.ea:neaY:Ta:va.a:pMa Zea:Sa:k
gua:Na:va:geRa :pra:k M ya:Ta.a Á
ta.ua [ea:pa:&+tMa [ea:paH v.ya:~taH :pra:k ãÉyua:tea

x +. a.ta:ta:( ÁÁ
gua:Na:l+a.b.DaH
/ :pa:dM :h x
ò :~vMa ta:ta.ea .$yea:;ma:Ta.eaY:sa:k+.t,a Á
tya:+a :pUa:vRa:pa:de:[ea:pa.a:n,a ..ca:k

Ò +.va.a:l+a.ma:dM .$a:guaH ÁÁ
õù;aeRÅ :k+.yua:ta.a:vea:vMa A:a.Ba:ea Ba:va:taH :pa:de

..ca:tua:d Á
..ca:tua:a.d
õR :[ea:pa:mUa:l;a:Bya.Ma .+pa:[ea:pa.a:TRa:Ba.a:va:na.a ÁÁ
Considering the lesser root, greater root and interpolator

(Ks.epa) as the dividend, addend and divisor respectively of
pulversier) the indeterminate multiplier of it should be so
taken as will make the residue of the prakr.ti diminished by
the square of that multiplier or the latter minus the prakr.ti
(as the case may be) to be the least. That residue divided by
the (original) interpolator is the (new) interpolator; it should
be reversed in sign in case of the subtraction from the prakr.ti,
The quotient corresponding to that value of the multiplier
is the (new) lesser root, like wise is obtained the greater
root. The same process should be followed putting aside
(each time) the previous roots and the interpolator. This
process is called Cakravāla (or the cyclic method). By this
method, there will appear two integral roots corresponding
to an equation with ±1, ±2 or ±4 as interpolator. In order to
derive integral roots corresponding to an equation with ad-
ditive unity from those of the equation with the interpolator
±2 or ±4, Bhāvanā (should be applied).
We describe algorithm in the following. To solve
x2 − Dy 2 = 1,
168 M.S.Sriram
we have to consider equations of the form
x2i − Dyi2 = Ki ,
in the intermediate stages, and also an indeterminate multiplier pi . Here

yi is the lesser root, xi is the greater root and Ki is the ks.epa. We start
with,
x20 − Dy02 = K0 , with x0 = 1, y0 = 0, K0 = 1,

and set p0 = 0. Given xi , yi , Ki and pi , yi+1 and pi+1 are obtained by
solving the kut..taka,
yi pi+1 + xi
yi+1 = ,
|Ki |
with the extra condition that |p2i+1 − D| is chosen to be minimum. Then
xi+1 and Ki+1 can be found from:
xi pi+1 + Dyi
xi+1 = ,
|Ki |
p2 − D
Ki+1 = i+1 .
Ki
They satisfy:
x2i+1 − Dyi+1
2
= Ki+1 .
Note that the above equation arises by doing Bhāvanā between
x2i − Dyi2 = Ki ,
p2i+1 − D.12 = p2i+1 − D,

and dividing the new x, y and K by |Ki |, |Ki | and |Ki2 | respectively.
This process is repeated till we get one of the values ±1, ±2, ±4 for the
Ks.epa. If we obtain the Ks.epa to be 1, we have solved the problem. If
the Ks.epa is −1 or ±2 or ±4, Bhāvanā can be used to obtain the solution
for x2 − Dy 2 = 1, as explained earlier.
Bhāskara II or his commentators have not outlined any proof that

Cakravāla algorithm always leads to a solution in a finite number of
steps. In 1929, A.A.Krishnaswami Ayyangar proved that the Cakravāla
algorithm always leads to a solution in a finite number of steps [11].
He also showed that the above procedure is equivalent to the one in

which the kut..taka equation for pi+1 is replaced by the condition that Ki
divides pi + pi+1 , with the other conditions being the same (including
minimization of |p2i+1 − D|).
Bhāskara himself considered the example of D = 61 in his Bı̄jagan.ita.

The successive iterated values of pi , Ki , xi , yi are given in Table 1.
The smallest solution x = 1766319049, y = 226153980 is reached after
the 14th step, if Cakravāla algorithm is applied mechanically. However,
just after two steps we find K = −4, so that the solution is found im-
mediately using the Bhāvanā.
Cakravāla algorithm for x2 − 61y 2 = 1
i pi Ki xi yi
0 0 1 1 0
1 8 3 8 1
2 7 -4 39 5
3 9 -5 164 21
4 6 5 453 5
5 9 4 1523 195
6 7 -3 5639 722
7 8 -1 29718 3805
8 8 -3 469849 60158
9 7 4 2319527 296985
10 9 5 9747957 1248098
11 6 -5 26924344 3447309
12 9 -4 90520989 11590025
13 7 3 335159612 42912791
14 8 1 1766319049 226153980
Table 1
After the first step, we have
82 − 61 × 12 = 3.
170 M.S.Sriram
In step 2, 8 + p2 is divisible by 3 and the minimum value of |p22 − 61| is

obtained when p2 = 7. Then x2 = 39, y2 = 5, K2 = −4 and we have
392 − 61 × 52 = −4.
Since K = −4, we can use Bhāvanā to obtain

2 1 2 2
x = (39 + 2) 39 + 1 39 + 3 − 1 = 1766319049,
2

1
y = (39 × 5) 392 + 1 392 + 3 = 226153980.
2
The Brouncker-Wallis-Euler-Lagrange algorithm for solving the equa-√

tion x2 − Dy 2 = 1, is based on the continued fraction expansion of D.
It can be shown to be identical to Cakravāla algorithm, except that the
condition
D − p2i+1 is minimum,
is replaced by
D − p2i+1 is minimum and positive.
The Cakravāla algorithm often skips some of the steps encountered in

the Brouncker-Wallis-Euler-Lagrange algorithm. Table 2 gives the suc-
cessive values of pi , Ki , xi , yi in the Brouncker-Wallis-Euler-Lagrange
algorithm for D = 61 where arrows indicate Cakravāla steps. The so-
lution is reached after 22 steps using this algorithm, in contrast to 14
steps in the case of Cakravāla algorithm. If the number of steps in the
Cakravāla algorithm and the Brouncker-Wallis algorithm are denoted by
nC and nB respectively, then it has been shown [12] empirically that for
large D,
nB
≈ 1.44.
nC
Thus, the Chakravāla algorithm is about 44% more efficient for large D.
Quadratic indeterminate equations in which ancient Indians did such
pioneering work is an active field of research even today and is at the
heart of computational mathematics [13].
Brouncker-Wallis-Euler-Lagrange algorithm for x2 − 61y 2 = 1
I pi Kj xi yi
→ 0 0 1 1 0
1 7 -12 7 1
→ 2 5 3 8 1
→ 3 7 -4 39 5
4 5 9 125 16
→ 5 4 -5 164 21
→ 6 6 5 453 58
7 4 -9 1070 137
→ 8 5 4 1523 195
9 7 -3 5639 722
10 5 12 24079 3083
→ 11 7 - 29718 3805
12 7 12 440131 56353
→ 13 5 -3 469849 60158
→ 14 7 4 2319527 296985
15 5 -9 7428430 951113
→ 16 4 5 9747967 1248098
→ 17 6 -5 26924344 3447309
18 4 9 63596645 8142716
→ 19 5 -4 90520989 11590025
→ 20 7 3 335159612 42912791
21 5 -12 1431159437 183241189
→ 22 7 1 1766319049 226153980
Table 2
6 Construction of the Sine-table
In the above figure AP = R sin θ. This is the Indian jyā , where R the
radius or trijyā is normally chosen such that one minute of arc on the
circle corresponds to one unit of distance. Hence R is the number of
minutes in a radian, and its value is very close to 3438, which is the
value mentioned in the earlier texts. More exact values are mentioned
in later texts, especially those of the medieval Kerala school.
172 M.S.Sriram
θ
O P
Fig : 5
Normally a quadrant is divided into 24 equal parts, so that each arc bit
α = 90 0
24 = 3 45 = 225 . Then the procedure for finding R sin iα, i =
1, 2, . . . 24 is explicitly given. The R sines of the intermediate angles are
to be determined by interpolation.
Āryabhat.ı̄ya gives the explicit algorithm for constructing the sine-table
(verse 12, Gan.itapāda):
:pra:Ta:ma.a:a.a:pa.$ya.a:Da.Ra:dù;aEÅ .+nMa Ka:a.Nq+tM

// a ;a.d
õ :ta.a:ya.a:DRa:m,a Á
ta:tpra:Ta:ma.$ya.a:Da.a:ZEa:~tEa:~tEa.+na.a:
a.na Zea:Sa.a:a.Na ÁÁ
The first Rsine divided by itself and then diminished by the

quotient gives the second Rsine difference. The same first
Rsine diminished by the quotients obtained by dividing each
of the preceding R-sines by the first Rsine gives the remain-
ing Rsine-differences.
This tells us that

R sin α
R sin 2α − R sin α = R sin α − (α = 225 ),
R sin α
R sin α + R sin 2α + . . . + R sin iα
R sin(i + 1)α − R sin iα = R sin α − .
R sin α
The second equation is equivalent to the relation:
R sin iα
R sin(i + 1)α − R sin iα = R sin iα − R sin(i − 1)α − .
R sin α
The values of the 24 Rsines are explicitly noted in another verse, and
the first Rsine is given by
R sin α ≈ 225
which obviously uses
sin α ≈ α,
when α is small. The exact recursion relation for the Rsine differences
is:
R sin(i + 1)α − R sin iα = R sin iα − R sin(i − 1)α − R sin iα 2(1 − cos α).
Now, 2(1 − cos α) = 0.0042822, which is approximated in Āryabhat.ı̄ya
to be
1 1
= = 0.00444444.
R sin α 225
The Āryabhat.ı̄ya ’s sine values are given in the Table 4. The same
procedure with the same values are to be found in many other works in-
cluding Sūryasiddhānta [14]. However the Rsine values given by Govin-
dasvāmi in his commentary on Mahābhāskarı̄ya of Bhāskara I which are
reproduced in Table 4 are more accurate (correct to five decimal places)
[15,16].
Far more accurate values for the sines were given by Mādhava of
Saṅgamagrāma (1340-1425), the father figure of the Kerala school of
astronomy and mathematics. These are cited in the two commen-
taries on Nīlakan.t.ha Somayāji’s Tantrasaṅgraha (c.1500)[17,18] namely
Laghuvivr.tti [17] and Yuktidı̄pikā [18], both composed by Śaṅkara
Vāriyar.
These results are based on the series expansion for sin θ, which we write
in the following form:
(Rθ)3 (Rθ)5 (Rθ)7 (Rθ)9 (Rθ)11

R sin θ = Rθ − + − + − + ...,
3!R2 5!R4 7!R6 9!R8 11!R10
where Rθ is the arc in minutes [19]. Explicit values of the magnitudes

of the terms starting in the reverse from sixth and up to the second in
(the RHS of) the above equation, given by Mādhava when the arc Rθ =
5400 = 90◦ , are mentioned in Yuktidı̄pikā in the kat.apayādi notation.
The verse giving these values is quoted below[18].
174 M.S.Sriram
;a.va:d
õ .Ma:~tua:a:ba:l.H k+.va.a:Za:
a.na:.ca:yaH .sa:va.Ra:TRa:Za.a:l+a.~Ta.=H
////
ñÍ ñÍ ú
;
a.na:a.vRa.;d
Äâ .a:*:+na:=
çÅ e ;ndÒ:+.a.îåÅ
*:+ga:a.d:tea:Svea:Sua kÒ +.ma.a:t,a :pa.*.a:su
ãÁ a Á
These values along with the modern values for comparison are listed in
Table 3. The last column in this table was computed using R = 3437.747
and Rθ = 5400.
Term Mādhava’s
no. in Sanskrit equivalent value according Modern
RHS in kat.apayādi to Yuktidı̄pikā value
VI ;a.va:d
õ .a:n,a 0 0 44 0 0 44.54
V tua:a:ba:l.H 0 33 6 0 33 5.6
IV k+.va.a:Za:
a.na:.ca:yaH 16 05 41 16 05 40.87
III .sa:va.Ra:TRa:Za.a:l+a.~Ta.=H
//// 273 57 47 273 57 47.11
II ;
a.na:a.vRa.;d
ñÍ
Äâ .a:*:+na:=
çÅ e ;ndÒ:+.k
, 2220 39 40 2220 39 40.10
Table 3
We find that the values given by Mādhava are indeed very accurate. For
an arbitrary arc Rθ (in minutes), the procedure is given in the later half
of the verse quoted above [18]:
A.a:Da:~tya.a:t,a gua:a.Na:ta.a:d:Ba.a::Da:nua:SaH kx +.tya.a ;a.va:&+tya.a:a.nta:ma:s,

// a -
ya.a:Ma Za.ea:Dya:mua:pa:yRua:pa:yRa:Ta ;Da:nea:nEa:vMa ;Da:nua:Sya:nta:taH ÁÁ
It is the following formula for R sin θ that is delineated by the above

verse:
R sin θ = Rθ − β 3 (2220 39 40 ) + β 5 (273 57 47 ) − β 7 (16 5 41 )
+β 9 (0 33 6 ) − β 11 (0 0 44 ),
Rθ
where β = 5400 . The 24 Rsines corresponding to Rθ = 225 , 450 , 675 . . .
were also explicitly stated by Mādhava in the kat.apayādi notation and
they have been given in Laghuvivr.tti in the set of verses beginning with
(rea:M na.a:ma va:a=;:a:na.Ma ;a.h:ma.a:a.dÒ;veRa:d:Ba.a:va:naH and ending with ta:tpa.=:a:a.d:k+.l;a:nta.a:~tua
ma:h.a.$ya.a ma.a:Da:va.ea:a.d:ta.aH Á They coincide with the modern values up to “thirds”
(corresponding to an accuracy of sines up to seventh or eighth decimal

places).
R sin θ according to
θ in min. Āryabhat.ı̄ya Govindasvāmi Mādhava(also Modern)
225 225 224 50 23 224 50 22
450 449 448 42 53 448 42 58
675 671 670 40 11 670 40 16
900 890 889 45 08 889 45 15
1125 1105 1105 01 30 1105 01 39
1350 1315 1315 33 56 1315 34 7
1575 1520 1520 28 22 1520 28 35
1800 1719 1718 52 10 1718 52 24
2025 1910 1909 54 19 1909 54 35
2250 2093 2092 45 46 2092 46 03
2475 2267 2266 38 44 2266 39 50
2700 2431 2430 50 54 2430 51 15
2925 2585 2584 37 43 2584 38 06
3150 2728 2727 20 29 2727 20 52
3375 2859 2858 22 31 2858 22 55
3600 2978 2977 10 09 2977 10 34
3825 3084 3083 12 51 3083 13 17
4050 3177 3175 03 23 3176 03 50
4275 3256 3255 17 54 3255 18 22
4500 3321 3320 36 02 3320 36 30
4725 3372 3371 41 01 3371 41 29
4950 3409 3408 19 42 3408 20 11
5175 3431 3430 22 42 3430 23 11
5400 3438 3437 44 19 3437 44 48
Table 4
In Table 4, we compare the values of the Rsines in Āryabhat.ı̄ya , Govin-
dasvāmin’s commentary of Mahābhāskarı̄ya, and Mādhava’s values as
176 M.S.Sriram
stated in Laghuvivr.tti [16] and the modern values.

It may be noted that the table gives the Rsine values corresponding to
arc lengths which are multiples of 225’. In the third and the fourth
columns, the values are in minutes, seconds and thirds. Note that the
modern values coincide with the Mādhava’s values in the last column.
7 Accurate computation of π using an

error-minimization algorithm
It is well known that an infinite series for π was first given by the Ker-
ala mathematicians, who invariably ascribe the result to Mādhava of
Saṅgamagrāma (1340 - 1425 AD):
Circumf erence 1 1 1
= π = 4(1 − + − . . . ).
Diameter 3 5 7
An ingenious geometrical proof of this is to be found in the celebrated

Malayalam text Yuktibhās.ā of Jyes.t.hadeva (c.1530) [20,21]. As is well
known, this series converges very slowly. An algorithm for accurate and
efficient computation of π using the technique of correction terms was
also given by Mādhava. This result is cited in Yuktidı̄pikā of Śaṅkara
Vāriyar [22].
v.ya.a:sea va.a:a=; a.Da:

a.na:h:tea .+pa:&+tea v.ya.a:sa:sa.a:ga.=:a:a.Ba:h:tea Á
ñÍ
-
;aa:Za.=:a:a.d ;a.va:Sa:ma:sa:*
ËÉ +;a.a:Ba:+.m
ù
Á
xa:NMa .~vMa :pxa:Ta:k, k Ò +.ma.a:t,a k
u +.ya.Ra:t,a ÁÁ
ñÍ
ya:tsa:* É
Ë+;a:ya.a.a
ù x x
h.=;Nea k+.tea ;
a.na:va.a.a &+ a.ta:~tua .$a.a:a.ma:ta:ya.a Á
Á
ñÍ
ta:~ya.a +.DvRa:ga:ta.a ya.a .sa:ma:sa:* Ë É +;a.a
ù ta:;lM gua:Na.eaY:ntea .~ya.a:t,a ÁÁ
Á
ta:d
õ :ga.eRa .+pa:yua:ta.ea h.a.=:ea v.ya.a:sa.a:a.b.Da:Ga.a:ta:taH
/ :pra.a:gva:t,a Á
ta.a:Bya.a:ma.a:Ma .~va:mxa:Nea kx +.tea ;Ga:nea [ea:pa O;:va k+=;Na.a:yaH ÁÁ
l+b.DaH :pa:a=; a.DaH .sUa:[maH ba:hu:kx +.tva.ea h.=;Na:ta.eaY: a.ta:sUa:[maH .~ya.a:t,a ÁÁ
The diameter multiplied by four and divided by unity (is

found and stored). Again the products of the diameter and
four are divided by the odd numbers like three, five, etc., and
the results are subtracted and added in order (to the earlier
stored result).
Take half of the succeeding even number as the multiplier
(gun.a term) at whichever (odd) number the division process
is stopped, because of boredom (by the slow converging pro-
cess). The square of that (even number) added to unity is
the divisor. The ratio has to be multiplied by the product
of the diameter and four as (stated) earlier.
The result obtained has to be added if the earlier term (in
the series) has been subtracted and subtracted if the earlier
term has been added. The resulting circumference is very
accurate; in fact more accurate than the one which may be
obtained by continuing the division process (with a large
number of terms in the series).
Essentially,

p+1
π 1 1 1 1 2
=1− + − ± ∓ ,
4 3 5 7 p (p + 1)2 + 1
to a fairly good approximation.
Describing different ways by which better approximations can be ob-
tained, finally Śaṅkara Vāriyar states a more accurate correction term
[22]:
O;:ByaH .sUa:[ma:ta.=:eaY:nya.ea ;a.va:

a.l+K.ya:tea k+.(É
ãa:na.a: a.pa .sMa:~k+a.=H Á
ñÍ
A:ntea .sa:ma:sa:* Ë É +;a.a:d:l+va:gR
ù aH .sEa:k+ea gua:NaH .sa O;:va :pua:naH Á Á
Á
ñÍ
yua:ga:gua:a.Na:ta.ea .+pa:yua:taH .sa:ma:sa:* ËÉ +;a.a:d:l+h:ta.e
ù a Ba:vea:d
, h.a.=H Á
Á
ñÍ
;aa:Za.=:a:a.d ;a.va:Sa:ma:sa:* É +;a.a:h.=;Na.a:t,
Ë ù a :pa.=;mea:ta:de:va va.a k+a:yRa:m,a ÁÁ
Á
Now a correction term better than the others will be pre-

sented. Here square of half of the even term increased by
unity, is the multiplier. The same (square) increased by
unity and multiplied by four, and further multiplied by half
the even term is the divisor. This correction term should be
applied after dividing by the odd numbers three, five, etc.
178 M.S.Sriram
In other words, a much better approximation for π4 is:

p+1 2
+1
π 1 1 1 1 2
=1− + − ± ∓ .
4 3 5 7 p 2
p+1
[(p + 1) + 4 + 1]
2
In Yuktibhās.ā , these correction terms are derived using an error-

minimization algorithm [21,23]. We give a summary of this procedure
as explained in the above references. If the series is truncated at odd
number p − 2 with a correction term Sl1 , π4 can be written as:
π 1 1 1 1 1
= 1 − + − ... − + .
4 3 5 7 p − 2 S1
On the other hand, if the series is truncated at p with a correction term
l
S2 , we will have
π 1 1 1 1 1 1
= 1 − + − ... − + − .
4 3 5 7 p − 2 p S2
If the correction terms are exact, then both should yield the same result.
That is,
1 1 1 1 1 1
= − or + = .
S1 p S2 S1 S2 p
S1 = S2 = 2p actually satisfies the equality. However both the correc-
tions should follow the same rule. If S1 = 2p − 2 and S2 = 2p + 2, that
is S is twice the even number above the last odd number in the series,
then the Sthaulya or inaccuracy is
1 1 1 1 1 1 4
∆= + − = + − = 3 ,
S1 S2 p 2p − 2 2p + 2 p 4p − 4p
and it can be shown that the error (corresponding to this choice of S)
is minimum. For other choices such as S1 = 2p − 3 and S2 = 2p + 1,
the inaccuracy ∆ will have a term proportional to p in the numerator,
whereas the denominator ≈ 4p3 . So the inaccuracy ∆ will be much
larger than its value for S1 = 2p − 2, when p is large. By choosing S2 (p)
of the form,
1
S2 (p) = A
,
2p + 2 + 2p+2
it can be shown that the inaccuracy is reduced further and inaccuracy

∆ is minimum at A = 4. So,

p+1
1 2
S2 (p) = = .
4 (p + 1)2 + 1
(2p + 2) +
2p + 2
This leads us to

p+1
π 1 1 1 1 2
= 1 − + − ... ± ∓ ,
4 3 5 7 p (p + 1)2 + 1
which is the first rule given in Yuktidı̄pikā . Proceeding in the same
manner, the error is further reduced and will be minimum when we take
1
S2 (p) = ,
4
(2p + 2) +
16
2p + 2 +
2p + 2
which leads to
2
p+1
+1
π 1 1 1 1 2
= 1 − + − ... ± ∓ .
4 3 5 7 p 2
p+1
[(p + 1) + 4 + 1]
2
This is the more accurate second rule given in Yuktidı̄pikā . When p = 99
(that is, considering 50 terms in the series), the resulting value of π is
found to be accurate to 11 decimal places. This is same as the value
π ≈ 3.14159265359222. . . given by the rule, ;a.va:bua:Da:nea.a:ga.$a.a:a.h:hu:ta.a:Za:na. . . , at-
tributed to Mādhava by Nīlakantha in his Āryabhatı̄yabhās. ya Gan.itapāda,
verse 2.
When the error-minimization algorithm is continued along the above
lines two steps further,
1
S2 (p) = .
4
(2p + 2) +
16
2p + 2 +
32
2p + 2 +
64
2p + 2 +
2p + 2
180 M.S.Sriram
When p = 99, this yields π = 3.141592653589793241, which is correct

to 17 decimal places. This is the same as the value for π stated by Rājā
Śaṅkara Varma in his Sadratnamālā (c.1823).
Acknowledgement
I thank M.D.Srinivas, Centre for Policy Studies, Chennai, and K.Rama-
subramanian, IIT Bombay, for many useful discussions on this topic.
References
1. See for instance, Jacques Sesiano, ‘Al-Khwarizmi’, Encyclopedia of

the History of Science, Technology, and Medicine in Non-Western
Cultures, Ed. by Helaine Selin, Kluwer Academic Publishers, Dor-
drecht, 1997.
2. M.D.Srinivas, ‘Proofs in Indian Mathematics’, Contribution to this

volume.
3. See for instance, T.A.Sarasvati Amma, Geometry in Ancient and

Medieval India, Motilal Banarasidass, 1979.
4. Āryabhat.ı̄ya of Āryabhat.a with the commentary of Bhāskara I and

Someśvara, K.S.Shukla, Indian National Science Academy, New
Delhi, 1976, p.49.
5. Āryabhat.ı̄ya Gan.itapāda, verse 5, Edited with English translation,

notes and comments by K.S.Shukla and K.V.Sarma, Indian Na-
tional Science Academy, New Delhi, 1976.
6. B.B.Datta and A.N.Singh, History of Hindu Mathematics, Part II,

Motilal Banarasidas, Lahore, 1938.
7. The manuscript of Udayadivākara’s commentary on Laghu-

bhāskarı̄ya is available in the Kerala University Oriental Research
Institute and Manuscripts Library, Trivandrum.
8. Bı̄jagan.ita of Bhāskarācārya edited by Achyutananda Jha,

Chowkhamba, Benaras, 1949.
9. Brāhmasphut. a-siddhānta (with Hindi translation) 4 vols., Ed. by

R.S.Sharma, Indian Institute of Astronomical and Sanskrit Re-
search, New Delhi, 1966.
10. Bı̄jagan.ita, cited earlier, Translation by Datta and Singh in Ref.6.
11. A.A.Krishnaswamy Ayyangar, Journal of Indian Mathematical So-
ciety, 18, 225, 1929.
12. Tomas Olivera Silva, ‘Record holder solutions of Pell’s equation’,
www.ieeta.pt-/ tos/pell.html, September 2003.
13. See for Example, S.Hallgren, ‘Polynomial-Time Quantum Algo-
rithms for Pell’s equation and the Principal ideal problem’, Cali-
fornia Institute of Technology preprint, November 2001.
14. Sūryasiddhānta , verses 15-16, Chapter 2, Ed. by K.C.Dvivedi,
Sampurnanand Sanskrit University, Varnasi, 1987.
15. Mahābhāskarı̄ya of Bhāskara I, Ed.by T.S.Kuppanna Sastri with
the Bhās.ya of Govindasvāmi, Govt. Oriental Manuscripts library,
Madras, 1957, p.200-1.
16. A.K.Bag, Mathematics in Ancient and Medieval India,
Chaukhambha Orientalia, Varanasi, 1979.
17. Tantrasaṅgraha of Nīlakan.t.ha with the commentary Laghuvivr.tti
of Śaṅkara Vāriyar, Ed. Surnad Kunjan Pillai, Trivandrum San-
skrit Series, No.188, Trivandrum, 1958, p.19.
18. Tantrasaṅgraha of Nīlakan.t.ha Somayājī with Yuktidı̄pikā and
Laghuvivr.tti of Śaṅkara Vāriyar, Ed.by K.V.Sarma, V.V.B Insti-
tute, Hoshiarpur, 1977, p.117.
19. A.K.Bag, ‘Mādhava’s sine and cosine series’, Indian Journal of
History of Science, 11, 54-57, 1976.
20. Gan.itādhyāya, of the text Yuktibhās.ā , Ed.by Ramavarma Tham-
puran and A.R. Akhileswara Aiyer, Trichur, 1947: See also
M.D.Srinivas in Ref.2.
21. Gan.ita-yuktibhās.ā, Vol. I, Ed. and Translated by K.V.Sarma,
with Explanatory Notes by K.Ramasubramanian, M.D.Srinivas
and M.S.Sriram (in press).
182 M.S.Sriram
22. Cited earlier; translation by K.Ramasubramanian.
23. Jolly John, ‘Derivation of the sam. skāras applied to the Mādhava
series in Yuktibhās.ā ’, in 500 years of Tantrasaṅgraha : A
Landmark in the History of Astronomy, Ed. by M.S.Sriram,
K.Ramasubramanian and M.D.Srinivas, Indian Institute of
Advanced Study, Shimla, 2002.
Algorithms in Indian Astronomy
K.Ramasubramanian
Cell for Indian Science and Technology in Sanskrit
Department of HSS, IIT Bombay
Powai, Mumbai 400 076, India
kramas@iitb.ac.in
Abstract
Indian Astronomy is rich in algorithms. The algorithms pre-

sented in the Indian astronomical texts have varying degrees of
complexities starting from the simple trairāśika rule, to the treat-
ment of parallax in a solar eclipse or the computation of the ele-
vation of lunar cusps. In the present article we will discuss a few
algorithms that are representative of the ingenuity and continuity
of the Indian astronomical tradition. We start with the interpo-
lation formula presented by Brahmagupta (c.665 AD) and then
proceed to describe a select few algorithms from Tantrasaṅgraha
of Nīlakan.t.ha composed in 1500 AD. Here we present the algo-
rithm for the calculation of time from shadow measurements and
the exact algorithm for the computation of lagna and the time
for the duration of an eclipse. We also comment on the iterative
process known as aviśes.akarma which aims at circumventing the
problem of interdependencies among several variables.
1 Introduction
It is not uncommon to find words which originate with a different conno-

tation and in due course pick up a completely different connotation. The
word algorithm forms a good example of this. Its origin can be traced
back to the Persian mathematician, al-Khwārazmī (800-847 AD). It is
quite interesting to note the observations made by D.E. Knuth in this
context [1]:
183
184 K.Ramasubramanian
In the middle ages, abacists computed on the abacus and

algorists computed by algorism. Following the middle ages,
the origin of this word was in doubt, and early linguists at-
tempted to guess at its derivation by making combinations
like algiros [painful] + arithmos [number]; others said no, the
word comes from “King Algor of Castile”. Finally, historians
of mathematics found the true origin of the word algorism: it
comes from the name of a famous Persian textbook author,
Abu Ja‘far Mohammed ibn Musa al-Khwārizmī (∼ 825 AD)
- literally, “Father of Jafar, Mohammed, son of Moses, na-
tive of Khwārizm.” Khwārizm is today the small Soviet city
of Khiva. al-Khwārizmī wrote the celebrated book Kitab al
jabr w’al-muqabala (“Rules of restoration and reduction”);
another word, “algebra”, stems from the title of his book,
although the book wasn’t really very algebraic.
Gradually the form and meaning of “algorism” got distorted; The change
from “algorism” to “algorithm” is any body’s guess. The remarks by
the well-known historian C.B.Boyer in this context are also noteworthy
[2].
. . . . . . when subsequently Latin translations of his (al

Khwārizmī’s) work appeared in Europe, careless readers be-
gan to attribute not only the book but also the numeration
to the author. The new notation came to be known as that of
al-Khwārizmī, or more carelessly, algorismi; ultimately the
scheme of numeration making use of Hindu numerals came
to be called simply algorism or algorithm.
The terms process, method, technique, procedure, routine, and so on

all essentially refer to a sequence of operations to be carried out to ac-
complish a given task. The word “algorithm” though similar, connotes
something more. For a procedure to be termed algorithm it must termi-
nate after n steps and upon termination it must yield a sensible result.
However, this is not true of all procedures. In this sense, all algorithms
are procedures; but all procedures are not algorithms.
Indian astronomy is essentially algorithmic in nature. The algorithms
presented are precise and fairly sophisticated [3]. Some of them are
Algorithms in Indian Astronomy 185
amazingly accurate [4]. We shall illustrate these points by considering

a few examples.
2 Brahmagupta’s interpolation formula
Interpolation is the art of reading between the lines in a table. The

rule of trairāśika [5,6] employed in Indian astronomy is close to what is
known as the first-order interpolation in modern parlance. This tech-
nique has been extensively applied to solve a variety of mathematical
and astronomical problems, beginning from the evaluation of sine func-
tion to the calculation of eclipses, at least from the time of Āryabhat.ı̄ya
(c.499 AD).
It is quite interesting to note that Brahmagupta introduced the second
order interpolation formula to determine more precise value of the sine
function, called jyā , for an arbitrary angle, from the set of tabulated
values of sine given at fixed intervals. The following verse from his
famous work Khan.d.akhādyaka (c.665 AD) explains the algorithm [7]:
ga:ta:Ba.ea:gya:Ka:Nq+k+a:nta.=;d:l+a.va:k+.l+Ga.a:ta:Za:tEa:nRa:va:a.Ba.=:a:a:ya.a Á
ta:dù;aÅu : a.ta:d:lM yua:ta.ea:nMa Ba.ea:gya.a:dU:na.a: a.Da:kM Ba.ea:gya:m,a ÁÁ
Multiply the residual arc left after division by 900 by half

the difference of the tabular difference passed over and that
to be passed over and divide by 900 ; by the result increase
or decrease, as the case may be, half the sum of the same two
tabular differences; the result which, whether less or greater
than the tabular difference to be passed, is the true tabular
difference to be passed over.
Generally in Indian astronomical texts, the interval chosen for tabulating

values of sines is 225 . However, Brahmagupta has chosen the interval to
be 15◦ = 900 for the sake of simplicity, as Khan.d.akhādyaka is a karan.a
text.1 This explains the number 900 figuring in the above verse.
1
Of the three kinds of texts, Siddhānta, Tantra and Karan.a, the last is meant to
serve as a simple manual used for quick computations.
Before we represent the content of the above verse in a mathematical

form, it would be useful to introduce the terminologies and the notation
employed. Let the equal intervals chosen be represented by α. That is,
xn − xn−1 = xn+1 − xn = α = 900, (1)
where the variable x denotes the angle and xi s are multiples of 900 .
The term Bhogya refers to the present interval between xn and xn+1 .
The term khand.aka refers to the I order tabular differences in the sine
values and we denote it by ∆n .
∆n = f (xn ) − f (xn−1). (2)
Let us suppose that the value of the function at three points

f (xn−1 ), f (xn ), and f (xn+1 ) are known, and it is required to find out
f (xn + βα). Now, the formula given by Brahmagupta may be written
as

βα (∆n + ∆n+1 ) (∆n − ∆n+1 )
f (xn + βα) = f (xn ) + −β , (3)
α 2 2
where 0 < β < 1. With a little algebraic manipulation, and suitably

rearranging the terms, the above equation may be rewritten as

βα βα βα (∆n − ∆n+1 )
f (xn + βα) = f (xn ) + × ∆n+1 + −1 × . (4)
α α α 2
Clearly, the formula given by Brahmagupta is identical with the stan-

dard quadratic interpolation formula [8].
3 Algorithm for finding the time from shadow
The technique of making a fairly good estimate of the time from the
shadow of an object cast by the Sun has been in vogue from time im-
memorial. Different cultures and traditions across the world have de-
vised simple instruments for this purpose. In the Indian astronomical
tradition the instrument used is called śaṅku .
Circle drawn around rays from

the Sanku the sun
E
X
Sanku
z
O
Y W
Figure 1: Zenith distance and the length of the shadow.
The śaṅku essentially consists of a rod of suitable thickness and height.

Generally the height is taken to be 12 aṅgulas . For performing experi-
ments with śaṅku , it must be placed at the centre of a circle as shown
in Fig.1. Here, OX represents the śaṅku and OY is its shadow cast by
the sun. If the longitude of the sun, λ is known at a given instant, then
its declination, δ at that instant can be calculated using the relation (see
Fig.2)
sin δ = sin sin λ, (5)
where is the obliquity of the ecliptic. The procedure for obtaining
accurate values of the observer’s latitude, φ, are discussed in detail in
several Indian astronomical texts [9].
δ
λ
Celestial
Equator ε
N
Γ
Ecliptic (Vernal equinox)
Figure 2: Declination of the Sun.

The formula given by Nīlakan.t.ha in his celebrated work Tantrasaṅgraha

(c.1500 AD) for determining the time from shadow measurement turns
out to be essentially a function of φ, δ and z, where z is the zenith
distance of the sun. Since φ and δ are already found through shadow
measurements, only z needs to be known. For this, consider the triangle
OXY in Fig.1. It can be easily seen that
OY
sin z =
XY
OY
or z = sin−1 , (6)
XY
√
where XY is the hypotenuse given by 122 + OY 2 . Thus, the zenith
distance of the sun can also be obtained at any time by measuring the
shadow cast by the śaṅku .
The algorithm for finding the time from shadow measurements is pre-
sented by Nīlakan.t.ha as follows [10]:
È ñÍ
îå Á
v.ya.a:sa.a:DRa*+.a:t, a ta:taH Za:*:;e
öÐÅ aH l+}ba:k+a:Ma ;aa.$a.a:va:ya.a ÁÁ
h:tva.a dù;aÅu .$ya.a:a.va:Ba:e ta:t,a ..ca.=:$ya.a .~va:NRa:mea:va ..ca Á
ya.a:}ya.ea:d:gga.ea:l+ya.ea:~ta:~ya ..ca.a:pea v.ya:~tMa ..ca.=:a:sa:vaH ÁÁ
.sMa:~k+a:ya.Ra ga:ta:ga:}ya.a:~tea :pUa:va.Ra:pa.=;k+.pa.a:l+ya.eaH Á
The śaṅku is multiplied by radius (trijyā ) and divided by

lambaka . This is further multiplied by trijyā and divided
by dyujyā . To this quantity the carajyā is applied posi-
tively or negatively depending upon whether the Sun is in
the southern or the northern hemisphere.
To the arc of the result, the carāsavah. (ascensional differ-
ence) has to be applied in the reverse order. This gives the
time that has elapsed or yet to elapse in the eastern and the
western half of the hemisphere.
Before we express the above verse in mathematical notation we introduce

a few technical terms appearing in the above verse. The term trijyā
refers to R,2 lambaka is R cos φ and dyujyā is R cos δ. The word śaṅku
2
It is usually taken to be the measure of one radian in minutes. That is R ≈ 3438 .
in the first line of the quotation is used to denote R cos z.3 Carajyā is
R sin ∆α. With this, the the first two lines of the verse translate to the
relation
R cos z
R sin θ = ± R sin ∆α (7)
cos φ cos δ
In the latter half of the quotation given above it is mentioned that to
the arc of the above (θ), the ascensional difference, which is the angular
separation between the vernal equinox and the celestial object measured
along the celestial equator (see Fig.3), has to be applied in the reverse
order to obtain the required time t. That is,
t = θ ∓ ∆α. (8)
Substituting for θ we have

−1 R cos z
t = (R sin) ± R sin ∆α ∓ ∆α. (9)
cos φ cos δ
The above result can be easily understood using the tools of spherical
trigonometry.4 For this, consider Fig.3 where S is the Sun on its diurnal
path whose zenith distance is z, corresponding to the arc ZS. The
point where the sun sets in the western part of the horizon is denoted
by St . The segment P W is a part of the 6 0 clock circle, and the angle
Z P̂ W = 90o . Applying four-part formula to the triangle P W St , it can
be shown that, the ascensional difference ∆α is given by the relation
(well known in Indian Astronomy)
sin ∆α = tan φ tan δ. (10)
Further, it may be noted that S P̂ St = θ + ∆α is the angle to be covered

by the Sun from the given instant up to the sunset. This angle divided
by 6 and 15 give the time that is to elapse before sunset in hours and
nād.ikās (time unit of approximately 24 minutes) respectively.
3
Generally R cos z, which is the perpedicular distance of the sun from the horizon,
is called mahāśaṅku in order to distinguish it from the gnomon (dvādaśāṅgulaśaṅku)
used for shadow measurement. But sometimes mahāśaṅku would be simply referred
as śaṅku , as in the above verse, and what it refers to would be clear from the context.
4
Detailed demonstrations of the algorithms enunciated in Indian astronomical
and mathematical texts are given in the famous Malayalam work Gan.ita-Yuktibhās. ā
(c.1530 AD) of Jyes.t.hadeva [Ref.6, Vol.I & II].
φ
90 − Z
dirunal path
P 90−θ of the sun
θ A
90−δ
C
∆α S
r )
to
q ua vrtta
E ika
L t M
Horizo
n ha
W (G
St
B 6’
O
(T clo
iry ck
D ak cir
vr cle
tta
)
Q
Figure 3: Determination of time from shadow measurements.
Applying the cosine formula to the spherical triangle PZS, we have
cos z = sin φ sin δ + cos φ cos δ sin θ,
or,
cos z
sin θ = − tan φ tan δ
cos φ cos δ
cos z
= − sin ∆α.
cos φ cos δ
Hence, the time, t that is yet to elapse before sunset in angular measure
is
−1 cosz
t = θ + ∆α = sin − sin ∆α + ∆α, (11)
cos φ cos δ
which is the same as Eq.(9) given by Nīlakan.t.ha .
The method to determine the time elapsed after sunrise is exactly sim-
ilar. It is worth mentioning in this context that, in finding z, the cor-
rection due to the finite size of the Sun and its parallax have also been
taken into account by Nīlakan.t.ha [11]. If z were the apparent zenith
distance, then the true zenith distance z is given by
z = z + ∆θ with ∆θ = ds − p,
where ds is the angular semi-diameter of the Sun and p its parallax. As

∆θ is small,
R cos z = R cos(z + ∆θ)

R sin z
≈ R cos z − ∆θ . (12)
R
This is precisely what is stated to be the mahāśaṅku, R cos z, in an earlier

verse [12], which is to be used in the expression (Eq.9) for determining
the time from the shadow.
4 Algorithm for finding lagna
Lagna refers to the orient ecliptic point, that is, the point of the eclip-
tic which intersects with the eastern horizon at any desired instant.
Nīlakan.t.ha while discussing the procedure for finding lagna, in his
Tantrasaṅgraha first presents the conventional method, which could be
found in many of the earlier texts on Indian astronomy. Later, pointing
out that this procedure would give only approximate results, he proceeds
to give an exact algorithm for finding the lagna .
4.1 Conventional method
The following verses of Tantrasaṅgraha describe the standard procedure

used for the computation of lagna at any desired instant [13].
.sMa:~kx +.ta.a:ya:na:Ba.a:nUa:tTa.=:a:a.Za:ga:nta:v.ya:

a.l+ a.a:k+aH ÁÁ
ta:dÒ:a:a.Za:~va.ea:d:ya:pra.a:Na:h:ta.a .=:a:a.Za:k+.l;a:&+ta.aH Á
A:sa:va.ea .=:a:a.Za:Zea:Sa:~ya ga:ta.a:sua:Bya:~tya.$ea:a ta.a:n,a ÁÁ

ãÉa Zea:Sa:taH
o;a.=:ea.a.=:=:a:Za.a:na.Ma :pra.a:Na.aH Za.ea:Dya.a:( Á
ãÉa ta.a:va:taH
:pUa.=;
a.ya:tva.a .=;va.Ea .=:a:a.ZMa ;a.[a:pea:dÒ:a:Za.Ma:( ÁÁ
;a.va:Zua.;d
Äâ .a ya.a:va:ta.Ma :pra.a:Na.aH , Zea:Sa.a:a.~:M
/a:Za:
ç Åu ;Na.a:t,a :pua:naH Á
ta:dU:DvRa.=:a:a.Za:ma.a:na.a:a.a:n,a Ba.a:ga.a:n,a ;a.[a:va.a .=;va.Ea ta:Ta.a ÁÁ

È
îå Á å
:Sa:a.*+.a:a :pua:naH Zea:Sa.a:t,a ta:n}.a.a:na.a:a:k+.l;a A:a.pa Á
çÉ çÉ ñ
O;:vMa :pra.a:gl+îå
+;a:ma.a:nea:yMa A:~ta:l+îå Âå ÅÅ a:k
+;aM tua :Sa.*.+.yu , ÁÁ
v.ya:tya:yea:na.a:ya:nMa k+a:ya mea:Sa.a:a.d:tva:pra:a.sa.;d

Äâ :yea Á
From the longitude of the Sun corrected for ayana, the num-
ber of minutes to be elapsed in that rāśi (are calculated).
This is multiplied by the duration of rising of that rāśi and
is divided by the number of minutes in a rāśi . This gives
that prān.as for the remaining rāśi to rise and that has to be
subtracted from the duration elapsed (since the sunrise).
From the remainder (h ), the durations of rising of the rāśis
that follow have to be subtracted. Having added the degrees
remaining in that rāśi to the Sun, the other rāśis (degrees
corresponding to rāśis ), as many number of them, whose
rising times were subtracted are also added. The remaining
prān.as (r) are multiplied by 30 and divided by the duration
of rising of that rāśi . The result obtained is once again
added to the Sun.
The remainder when multiplied by 60 gives the result in min-
utes. Thus, prāglagna , the orient ecliptic point should be
obtained. The astalagna , setting ecliptic point is obtained
by adding six signs to that.
To know the longitude of the ecliptic points from the mes.adi,
the ayana correction has to be applied reversely.
The procedure described here may be understood with the help of Fig.4.
Here S represents the Sun in the eastern part of the hemisphere, Γ the
vernal equinox. R1 , R2 etc., are the ending points of the first rāśi (sign),
second rāśi and so on. h refers to the time elapsed after the sunrise and
H the hour angle of the the Sun.
Diurnal circle Z (zenith)
H P (pole)
tR
1
Γ
h
S R1
Eq
uat
R2 N
S
or
Horizo
n
Ecl
E ipti
c
Figure 4: Determination of prāglagna (orient ecliptic point) – by the

conventional method
Let λs be the sāyana longitude5 of the Sun. Suppose the Sun is in the ith
rāśi (in the Fig.4, it is shown to be in the first rāśi ), whose rising time
at the observer’s location is given by Ti . If θRi be the angle remaining to
be covered by the Sun in that rāśi (in minutes), then the time required
for that segment of the rāśi to come above the horizon is given by
θRi × Ti
tRi = , (13)
30 × 60
where Ti is in ghat.ikas . A factor of 60 in the denominator indicates
that the result tRi is expressed in prān.as.6 Subtracting this time tRi
from the time elapsed since sunrise h, we have
h = h − tRi .
From h the time required for the subsequent rāśis to rise, Ti+1 , Ti+2 ,
etc., are subtracted till the remainder r remains positive. That is,
r = h − Ti+1 − Ti+2 . . . − Ti+j−1 (r + ve)
5
This refers to the longitude measured from the vernal equinox. Here, it may be
noted that in Indian astronomy, nirayan.a longitudes are more commonly used. They
refer to the longitude measured from a fixed point, which is generally taken to be the
beginning point of star called Aśvini. It is also referred to as mes.ādi as in the last
line of the Sanskrit quotation given above.
6
Ghat.ikā is a measure which is close to 24 minutes and prān.a ≈ 4 seconds.
Suppose we are in the i + j th rāśi whose rising time is Ti+j . Then,

the portion of Ri+j , which would have come above the horizon in the
remaining time r is given by
r × 30
θEi+j = (in deg.). (14)
Ti+j
Now, the longitude of the prāglagna (L) is given by
L = λs + θRi + 30 + 30 + . . . + θEi+j . (15)
If so, then the astalagna is given by
astalagna = prāglagna + 180◦ , (16)
because the horizon divides the ecliptic exactly into two parts. The
lagnas obtained by the above procedure are sāyana lagnas. To obtain
the nirayan.a ones, one needs to subtract the ayanām
. śa.
After describing the above method, Nīlakan.t.ha remarks that it is only
approximate [14]:
O;:k+.a.sma:a:
/////// Ò +.ma.a:t,a k+a:l;ea ;a.h ;a.Ba:dù;a:te
a.pa .=:a:Za.Ea tua k Å a ÁÁ
.tea:na :Ea.=:a:a.Za:kM na.a.a k+.tua yua:M ya:ta:~ta:taH Á
çÉ
O;:va:ma.a:na.a:ta:l+îå
+;a:~ya .~TUa:l+tEa:va na .sUa:[ma:ta.a ÁÁ
The (rising) time gradually differs even in the same rāśi .

Hence, applying rule of three here, is not appropriate. There-
fore the lagna obtained by the above procedure will only be
approximate and not exact.
4.2 Exact algorithm
Subsequently, Nīlakan.t.ha gives an exact algorithm for finding lagna.

This is presented in several steps. First he finds the kālalagna (defined
in the following verse [15]), which is nothing but the time difference
between the desired instant and the time of rising of the vernal equinox
Γ.
.sa.a:ya:na:k x +.ta.aH
R +.Bua.$a.a:pra.a:Na.aH :pra.a:gva:t,a .~va:.ca.=;sMa:~k Á
. . . dù;aÅu :ga:ta:pra.a:Na:sMa:yua::H k+a:l;ea . . . Á
The right ascension of the sāyana Sun corrected by the as-

censional difference . . . this added to the prān.as elapsed gives
the kālalagna .
The time difference between the sunrise and the rising of the vernal
equinox Γ is α − ∆α (see Fig.5), where α is the right ascension of the
Sun and ∆α the ascensional difference. When this is added to the time
elapsed after sunrise, we obtain the kālalagna at the desired instant,
which we denote by L .
kālalagna = L = Time elapsed after sun rise + (α − ∆α). (17)
P
Di le
α
cir
urn
Ec
∆α
c
lip
al
tic
Horizo
n
X
Γ, Ε
Eq S
ua
tor
Figure 5: Determination of kālalagna .
Now, vitribhalagna is the point on the ecliptic which is 90◦ away from
the lagna. Consider Fig.6, where S is the Sun, Γ is the vernal equinox
and L is the orient ecliptic point, whose longitude is the lagna. V is
the vitribhalagna and K is the pole of the ecliptic. Dr.ks.epa is sine of
the zenith distance of the vitribhalagna (R sin ZV ). When Γ is on the

horizon, K is on the meridian. It can be easily seen that the kālalagna
is the hour angle of K, or L = Z P̂ K. Now,
ZK = KV − ZV = 90 − ZV,
as K is the pole of the ecliptic. Also, KP = and ZP = 90 − φ. Then,

using the cosine formula we obtain the following expression for dr.ks.epa
R sin ZV = R cos ZK = R cos sin φ + R cos φ sin cos L . (18)
Z (zenith)
K
H P (pole)
S Γ
Ecl
ipti
Eq
S c N
ua
Horizo
tor
n
L
E
Figure 6: Determination of prāglagna (orient ecliptic point) - exact

method.
This is essentially what is stated in the following verse [16].
A:ntya:dù;aÅu .$ya.a:h:ta.a:[a.a:dù;a:t, ãÉa l+}ba:k:H

Å a ;aa.$ya.a:Ma ya:( ÁÁ
çÉ È
k+a:l+l+îå îå Á
+;a.ea:tTa:k+ea: a.f*+H k+=:a:Ta.Ra:b.Dyua.=;gEa:&R+taH Á
dx:[ea:pa:~ta:a.; R +.mxa:ga.a:a.d:k

ÂåÅ ;dE:k+.aM ..ca k+a:le k+.a.k e ÁÁ
The aks.a multiplied by antyadyujyā and divided by trijyā,

and, lambaka multiplied by the kot.i of the kālalagna and di-
vided by 8452 (are kept separately). The dr.ks.epa is the dif-
ference or sum of the two depending upon whether kālalagna
is within the 6 signs beginning from karkat.aka or mr.ga Capri-

corn.
Here Nīlakan.t.ha defines two intermediate quantities x and y whose sum

or difference gives the expression for the sine of the zenith distance of
the the vitribhalagna . They are given by
antyadyujyā × aks.a R cos × R sin φ
x= = ,
trijyā R
and,
lambaka × kot.i of kālalagna R cos φ × R cos L × sin
y= = .
8452 R
It may be noted in the above equation that sin

sin 24
R = 3438 is taken to be
1
8452 . Now, dr
. ks.epa (= R sin ZV , refer Fig.6) is given to be
dr.ks.epa = x±y
Substituting for x and y we have,
R sin ZV = R cos sin φ ± R cos φ cos L sin , (19)
which is same as Eq.(18). From this, cos ZV is to be calculated. With

this, the lagna is to be found as follows [17]:
îå +a:d
ma:Dya.a: õ .a na:ta:pra.a:Na.aH ;
a.na:Za.a:Ta.a:d
õ .ea:a:ta.a:sa:vaH Á
O;:ta:d
â õâ .a:Na.ea:
a.na:ta.a ;aa.$ya.a ..ca.=:$ya.a:Q.a.a na:ta.a ya:a.d ÁÁ
ãÉea:a.=:$ya.ea:na.a ga.ea:le ya.a:}yea ;a.va:pa:yRa:ya.a:t,a

o+a:ta.a:( Á
È
îå Á
dù;auÅ .$ya.a l+}ba:k+.Ga.a:ta*+.a ;aa.$ya.a:a.a ..ca :pua:na:&R+ta.a ÁÁ
k+ea:fa.a dx:[ea:pa.$a.a:va.a:ya.a l+b.Da:.ca.a:pMa .=;va.Ea ;a.[a:pea:t,a Á

çÉ ëÐÅ
ta:+.îå
+;aM :pra.a:*:+.pa.a:l
ëÐÁ e .~ya.a:t,a ;
a.na:a.Za ..cea:t,a ta:a.d
õ :va:a$Ra:ta:m,a ÁÁ
The nata (hour angle) or unnata may be obtained depend-

ing upon whether the computation is done at midday or
midnight. If nata is obtained, then the bān.a (versed sine,
utkramajyā) of it is to be subtracted from trijyā , and the re-
sult is added to the carajyā. If unnata, (is obtained) then the
result deficient from carajyā is found. [This is the procedure

for the northern hemisphere]. For the southern hemisphere
the operation is reversed.
Here it is essentially stated that

R(cos φ cos δ[cos H + sin ∆α])
R sin(λl − λs ) =
cos ZV
R(cos φ cos δ cos H + sin φ sin δ)
= ,
cos ZV
where λl and λs are the lagna and the Sun’s longitude respectively. From
the above equation, taking the inverse sine we get λl − λs . To this, if we
add the longitude of the Sun we get the desired longitude that is lagna
(λl ). The rationale behind the above equation can be easily understood
with the help of Fig.6.
Applying the cosine formula to the spherical triangle PZS we have,
cos ZS = sin φ sin δ + cos φ cos δ cos H.
Similarly from the triangle ZVS we get,

cos ZS
cos V S = .
cos ZV
Now, V S + SL = V S + λl − λs = 90. Therefore, cos V S = sin(λl − λs ),
which results in the above expression for sin(λl − λs ).
5 Instantaneous velocity of the true planet
Both in the modern as well as ancient Indian astronomy, the true po-
sition of the planet7 is obtained from the mean position by applying
a correction to it. The correction term is known as mandaphala , in
Indian astronomy, while it is known as ‘equation of centre’ in modern
astronomy.
In Fig.7, A represents the direction of the mandocca (apside) and its
longitude ΓÔA = ,. P0 is the mean planet whose longitude called
7
Though the treatment is general, in this section planet refers to either the Sun
or the Moon.
A (direction of mandocca)
P (planet)
Q
θ0 − ϖ
P0
θ0 θ MS
ϖ Γ
O
Figure 7: The manda-samskāra or the equation of centre correction.
madhyamagraha (mean longitude) is given by θ0 = ΓÔP0 . The circle of

radius r with P0 as centre, is the epicycle on which the mandasphut. a
(true planet) is located. By construction, P P0 is parallel to OA. The
longitude of the mandasphut. a is given by θM S = ΓÔP . It can be easily
seen that
PQ r sin(θ0 − ,)
sin(θ0 − θM S ) = = ,
OP K
where
1
K = [(R + r cos(θ0 − ,))2 + r 2 sin2 (θ0 − ,)] 2 ,
is the mandakarn. a. As per the formulation in Tantrasaṅgraha , r also
varies such that
r r
= ,
K R
is a constant. Here r is a given parameter. Then,
r
sin(θM S − θ0 ) = − sin(θ0 − ,). (20)
R
The difference between the true (mandasphut. a ) and the mean position,
is called mandaphala, whereas the difference between the mandocca and
mean position is called mandakendra. Denoting them by ∆θ and M

respectively, the mandaphala is given by
r
∆θ = − sin−1 sin M . (21)
R
Thus the mandasphut. a (true longitude) of the planet is given by
θM S = θ0 + ∆θ. (22)
It may be noted that the calculation of true position of the planet in-
volves sine inverse function ( Eq.(22)). Hence, if one needs to find the
instantaneous velocity of the planet called tātkālikagati, one would have
to find the time derivative of this function. It is indeed remarkable that
an exact formula for the derivative of sine inverse function is given in
Tantrasaṅgraha as follows [18]:
x +. a.ta:pa:de:na .sMa:h:=e;t,a
..ca:ndÒ;ba.a:hu:P+l+va:gRa:Za.ea: a.Da:ta:aa.$ya:k+a:k Á
ta.a k+ea: a.f:P+l+

a.l+ a.a:k+a:h:ta.Ma :k
e +.ndÒ;Bua: a.+.a=;h ya:a l+Bya:tea Á Á 53 ÁÁ
õ :Za.ea:Dya mxa:ga.a:a.d:k
ta:a.d e ga:teaH ;a.[a:pya:ta.a:a.ma:h tua k+.kR +.f.a:a.d:ke Á
å
ta:;
ÂåÅ ;vea:t~.Pu +f:ta.=:a ga: a.ta:a.vRa:Da.eaH A:~ya ta:tsa:ma:ya.$a.a .=;vea.=;a.pa Á Á 54 ÁÁ
Let the product of the kot.iphala (in minutes)[r cos M ] and

the daily motion of the manda-kendra ( dM dt ) be divided by the
square root of the√square of the doh.phala subtracted from the
square of trijyā ( R2 − r 2 sin M ). The result thus obtained
has to be subtracted form the daily motion of the Moon if
the manda-kendra lies within six signs beginning from mr.ga
and added if it lies within six signs beginning from karkat.aka.
The result gives more accurate value of the Moon’s angular
velocity. In fact, the procedure for finding the instantaneous
velocity of the Sun is same as this.
If M be the manda-kendra , then the content of the above verse can be

expressed in mathematical form as
dM
d r cos M
−1 r dt
(sin ( sin M )) = . (23)
dt R R − r sin2 M
2 2
This verse appears in the context of finding the true rate of motion of
the Moon (instantaneous velocity) from its average rate of motion (mean
velocity). The term gati refers to the rate of change of the longitude of
the planet.
Recalling the expression for the true longitude of the planet, Eq.(22),
the rate of change of it is
d d d
θM S = θ0 − ∆θ. (24)
dt dt dt
Here, the first term in the RHS represents the mean velocity of the planet
and the second term the change in the mandaphala given by Eq.(23).
In the following section dealing with the computation of the duration of
an eclipse, the daily motion of the Moon denoted by dm is nothing but
the derivative of mandasphut. a given by Eq.(23),(24).
6 Algorithm for finding eclipse duration
In Indian astronomy, the total duration of an eclipse is found by deter-

mining the first and the second half durations separately. More often
than not, there will be significant difference between the two durations.
For obtaining accurate values, they are calculated using an iterative pro-
cedure, called aviśes.akarma, as will be explained in this section. The
sum of the two durations gives the total duration of the eclipse. For in-
stance, if T1 and T2 be the first and second half durations of the eclipse,
then the total duration of the eclipse is given by
T = T1 + T2 ,
where both T1 and T2 are found iteratively.
6.1 Half duration of the eclipse
The time interval between the Moon entering the umbral portion of the
shadow and the instant of opposition is the first half duration, T1 , and
that between the instant of opposition and the exit of Moon out of the
umbral region is the second half duration T2 . Naively, one may think
that these two durations would be equal. However, this is not true
because of the continuous change in the velocities of both the Sun and
the Moon.
In Fig.6(a), AX = β and OX = S represent the latitude of the Moon and

the sum of the semi-diameters of the shadow and the Moon respectively.
If dm − ds refers to the difference in the instantaneous daily motion of
the Sun and the Moon, then the first half duration (T1 ) is given by
√
OA OX 2 − AX 2 S2 − β2
T1 = = = (25)
Dif f. in daily motion dm − ds dm − ds
t
orbi
n’s
Moo
(a)
M .
X
O . Ecliptic
A
rbit
n ’s o
(b) Moo
M
O Y .
B
. Ecliptic
Figure 8: (a) The Earth’s shadow and the Moon just before the begin-
ning of the eclipse and just after the release. (b) The Earth’s shadow
and the Moon just after the beginning of total eclipse and just before
its release.
6.2 The need for iteration
In Eq.(25), β is the Moon’s latitude at the beginning of the eclipse. The

denominator represents the difference in the daily motion of the Sun and
the Moon. If λm and λs are the longitudes of the Sun and the Moon,
this difference in their daily motion called gatyantara or bhuktyantara is
given by
d
gatyantara = D(t) = (λm − λs).
dt
Initially, β and D(t) are calculated at the instant of opposition. If
Moon’s latitude and the rate of motion of the Sun and the Moon were
to be constant, then Eq.(25) would at once give the correct half dura-
tion of the eclipse and there would be no need for an iterative procedure.
However, they are continuously varying quantities. Hence, determining
T1 , using the latitude and gatyantara computed at the instant of opposi-
tion is only approximate and to get more accurate values aviśes.akarma,
a special kind of iterative procedure, is prescribed.
6.3 Concept of Aviśes.akarma
Aviśes.akarma refers to an iterative process that is to be carried out

whenever there is an interdependency (anyonyāśraya) between the quan-
tity to be calculated and the parameters which are involved in the cal-
culation. For instance, in eclipse calculation, without knowing the lati-
tude at the end of the eclipse, we will not be able to give the exact half
duration, and without knowing the half-duration it is not possible to
find the exact latitude of the Moon. To get over this tricky situation,
aviśes.akarma is recommended. The need for it is succinctly explained
Śaṅkara Vāriyar in his Yuktidı̄pikā as follows [19]:
;a.va:[ea:pa:k+ea: a.f:vxa.Ma ta:t,a na.a:na.a.+pMa :pra: a.ta:[a:Na:m,a Á
å
[ea:pa.ea ;a.Ba:a.ea Ba:vea:t,a .~.pa:ZRa:ma:Dya:ma.ea:[ea:Sua Za.a:ta:ga.eaH ÁÁ
;a.~Ta:tya:D
//// a ;a.d
õ :ta:yMa ;a.Ba:Ma [ea:pea ;a.Ba:ea Ba:vea:t,a ta:taH Á
ta:ta.eaY:a.va:Zea:Sa:a.va: a.Da:na.a ;a.~Ta:ya:DR

//// a:d
õ :ya:ma.a:na:yea:t,a ÁÁ
The declination circle (of the Moon) keeps changing every

second (continuously). [Hence], the latitudinal deflection of
the Moon is different at the beginning, middle and the end
of the eclipse.
Thus the two half durations (first and the second half) differ
because of the variation in the latitudinal deflection. There-
fore, the half-duration may be determined through the aviśes.a

process.
The word viśes.a means ‘distinction’; therefore aviśes.a is “without dis-

tinction”. Though the meanings of the words viśes.a and aviśes.a are
opposed to each other, the latter should not be taken to refer to tulya
or ‘completely identical’. In the context of mathematical calculations,
it only means – without distinction to a desired degree of accuracy. In
other words, in aviśes.akarma, the iterative process needs to be carried
out only up to a point wherein the two successive values of the results
are ‘without distinction’ for a desired degree of accuracy. Once this
accuracy is reached the process should be terminated.
6.4 The iterative process for half duration
Now, we illustrate the aviśes.a process by considering the example of

finding the half duration of an eclipse. Let tm be the time of opposition
or madhyakāla of a lunar eclipse, and ∆t0 be the zeroth order approx-
imation of the half duration of eclipse determined with the parameter
values obtained at tm . That is,

S 2 (tm ) − β 2 (tm )
∆t0 =
D(tm )
To get the first approximation, the sum of the semi diameters S, the
latitude of the Moon β and the difference in daily motion D are then
determined at tm − ∆t0 . With them the first approximation to the half
duration is obtained. It is given by

S 2 (tm − ∆t0 ) − β 2 (tm − ∆t0 )
∆t1 =
D(tm − ∆t0 )
To get the second approximation, S, β and D are determined at tm −∆t1 .
With these values the second approximation to the half duration is

S 2 (tm − ∆t1 ) − β 2 (tm − ∆t1 )
∆t2 =
D(tm − ∆t1 )
Similarly, determining S, β and D at tm − ∆t2 , the third approximation

is found.

S 2 (tm − ∆t2 ) − β 2 (tm − ∆t2 )
∆t3 =
D(tm − ∆t2 )
This process will be continued until,
∆tn − ∆tn−1 < , (26)
where is the desired degree of accuracy. At this stage, since ∆t has

converged to the desired accuracy, the iteration is terminated. In terms
of the notation used earlier, the first half duration T1 = ∆tn . Therefore,
the instant of the commencement of the eclipse, known as sparśakāla, is
given by
tb = tm − ∆tn . (27)
A similar procedure is to be adopted for the determination of moks.akāla

(te ), with the only difference that, instead of subtracting, the half du-
ration ∆t from tm , we need to add to it. If ∆ti be the second half
duration of the eclipse obtained after the ith iteration, then it is given
by

S 2 (tm + ∆ti−1 ) − β 2 (tm + ∆ti−1 )
∆ti =
D(tm + ∆ti−1 )
As in the case of sparśa, here again the process of iteration has to be

continued till ∆ti converges. That is,
∆tr − ∆tr−1 < . (28)
At this stage, the second half duration of the eclipse and the moks.akāla
are given by
T2 = ∆tr
te = tm + ∆tr (29)
Now, the total duration of the eclipse is te − tb = T1 + T2 .

7 Concluding Remarks
Many of the algorithms presented in the paper, barring some refine-

ments, can be found even in the celebrated text Āryabhat.ı̄ya , composed
by Āryabhat.a as early as 499 AD. The Kerala school of astronomy and
mathematics which is well known for its pioneering work in mathemat-
ical analysis and many innovations in the Indian astronomical tradition
has tried to perfect these algorithms, the culmination of which can be
seen in the works of Nīlakan.t.ha Somayājī. Particularly, the exact al-
gorithm for the computation of lagna and the formula for the instan-
taneous velocity of the planet presented by Nīlakan.t.ha , are indicative
of how in the Indian astronomical tradition there has been a continu-
ous endeavour to improvise and achieve better and better accuracy in
all computations. As regards the aviśes.akarma, it will be interesting to
study the convergence properties of this iterative process as employed
in different contexts in Indian Astronomy.
Acknowledgement
The author would like to express gratitude to Prof. M.D.Srinivas of
Centre for Policy Studies and Prof. M.S.Sriram of University of Madras
for many useful suggestions and discussions on this topic.
References:
1. Donald E.Knuth, The Art of Computer Programming, Vol. 1,

Addison Wesley, 1973, p 1-2.
2. C.B.Boyer, A History of Mathematics, John Wiley and Sons, 1989,

p 256.
3. See for instance, 500 years of Tantrasaṅgraha : A Landmark in the

History of Astronomy, Ed. by M.S.Sriram, K.Ramasubramanian
and M.D.Srinivas, Indian Institute of Advanced Study, Shimla,
2002.
4. (i) The verses attributed to Mādhava (c.14th century) by Śaṅkara

Vāriyar in his commentary to Tantrasaṅgraha (Chap 2, verses 437,
438) beginning with ;a.va:dõ.a:n,a tua:a:ba:l . . . and .~tea:naH .~:a.a:a.pa:Zua:naH . . . , for
obtaining the sine and cosine values for any desired angle, yield
results correct up to 7 decimal places. This is indeed a remarkable
result which may be considered far ahead of his times. (ii) For
more details and mathematical exposition of the above the reader
may refer to the article by M.S. Sriram published in the present
volume.
5. Lı̄lāvatı̄ of Bhāskarācārya with the commentary of Śaṅkara and

Nārāyan.a, Ed. by K.V.Sarma, VVRI, Hoshiarpur, 1975, ver.73, p
178.
6. Gan.ita-yuktibhās.ā, Ed. with English Translation by K.V.Sarma,

with Explanatory Notes by K.Ramasubramanian, M.D.Srinivas
and M.S.Sriram (in press), Vol. I, Chap 4.
7. Khan.d.akhādyaka of Brahmagupta, Ed. and Tr. by Bina Chat-

terjee, Motilal Banarsidass, 1970, Vol. 2, Uttarakhan. d.akhādyaka,
Chap 1, ver.4.
8. Erwin Kreyszig, Advanced Engineering Mathematics, John Wiley

& Sons, 1983, p 774.
9. See for instance, Siddhantaśiroman. i , Chap 3, verses 12-14.
10. For a detailed exposition on this, see K.Ramasubramanian

and M.S.Sriram, Corrections to the terrestrial latitude in
Tantrasaṅgraha , Indian Journal of History of Science, 38.2, 2003,
p 129-144.
11. Tantrasaṅgraha of Nīlakan.t.ha with the prose commentary

Laghuvivr.tti of Śaṅkara Vāriyar, Ed. by Surnad Kunjan Pillai,
Trivandrum Sanskrit Series no. 188, Trivandrum 1958, Chapter 3,
verses 23-25.
12. ibid., verses 19-21.
13. ibid., verses 95-100.
14. ibid., verses 100-101.
15. ibid., verse 102.
16. ibid., verses 104-105.

17. ibid., verses 107-109.
18. ibid., Chapter 2, verses 53-54.
19. Tantrasaṅgraha of Nīlakan.t.ha with Yuktidı̄pikā , commentary in

the form of verses by Śaṅkara Vāriyar, Ed. by K.V.Sarma, VVBIS,
Punjab University, 1977, Chap 4, verses 79 and 87, p 261-62.
Proofs in Indian Mathematics
M.D.Srinivas
Centre for Policy Studies
27, Rajasekharan Street
Mylapore, Chennai - 600 004, India.
policy@vsnl.com
Abstract
Contrary to the widespread belief that Indian mathematicians

did not present any proofs for their results, it is indeed the case
that there is a large body of source-works in the form of com-
mentaries which present detailed demonstrations (referred to as
upapatti-s or yukti-s) for the various results enunciated in the major
texts of Indian Mathematics and Astronomy. Amongst the pub-
lished works, the earliest exposition of upapatti-s are to be found
in the commentaries of Govindasvāmin (c.800) and Caturveda
Pr.thūdakasvāmin (c.860). Then we find very detailed exposition
of upapatti-s in the works of Bhāskarācārya II (c.1150). In the
medieval period we have the commentaries of Śaṅkara Vāriyar
(c.1535), Gan.eśa Daivajña (c.1545), Kr.s.n.a Daivajña (c.1600) and
the famous Malayalam work Yuktibhās.ā of Jyes.t.hadeva (c.1530),
which present detailed upapatti-s. By presenting a few selected
examples of upapatti-s, we shall highlight the logical rigour which
is characteristic of all the work in Indian Mathematics. We also
discuss how the notion of upapatti is perhaps best understood in
the larger epistemological perspective provided by Nyāyaśāstra,
the Indian School of Logic. This could be of help in explicating
some of the important differences between the notion of upapatti
and the notion of “proof” developed in the Greco-European tradi-
tion of Mathematics.
209
210 M.D.Srinivas
1 Alleged Absence of Proofs in Indian

Mathematics
Several books have been written on the history of Indian tradition in

mathematics.1 In addition, many books on history of mathematics de-
vote a section, sometimes even a chapter, to the discussion of Indian
mathematics. Many of the results and algorithms discovered by the
Indian mathematicians have been studied in some detail. But, little
attention has been paid to the methodology and foundations of Indian
mathematics. There is hardly any discussion of the processes by which
Indian mathematicians arrive at and justify their results and procedures.
And, almost no attention is paid to the philosophical foundations of In-
dian mathematics, and the Indian understanding of the nature of math-
ematical objects, and validation of mathematical results and procedures.
Many of the scholarly works on history of mathematics assert that Indian
Mathematics, whatever its achievements, does not have any sense of
logical rigour. Indeed, a major historian of mathematics presented the
following assessment of Indian mathematics over fifty years ago:
The Hindus apparently were attracted by the arithmetical

and computational aspects of mathematics rather than by
the geometrical and rational features of the subject which
had appealed so strongly to the Hellenistic mind. Their
name for mathematics, gan.ita, meaning literally the ‘science
of calculation’ well characterizes this preference. They de-
lighted more in the tricks that could be played with numbers
than in the thoughts the mind could produce, so that nei-
ther Euclidean geometry nor Aristotelian logic made a strong
impression upon them. The Pythagorean problem of the
incommensurables, which was of intense interest to Greek
geometers, was of little import to Hindu mathematicians,
who treated rational and irrational quantities, curvilinear
1
We may cite the following standard works: B.B.Datta and A.N.Singh, History of
Hindu Mathematics, 2 parts, Lahore 1935, 1938, Reprint, Delhi 1962; C.N.Srinivasa
Iyengar, History of Indian Mathematics, Calcutta 1967; A.K.Bag, Mathematics in
Ancient and Medieval India, Varanasi 1979; T.A.Saraswati Amma, Geometry in An-
cient and Medieval India, Varanasi 1979; G.C.Joseph, The Crest of the Peacock: The
Non-European Roots of Mathematics, 2nd Ed., Princeton 2000.
Proofs in Indian Mathematics 211
and rectilinear magnitudes indiscriminately. With respect to

the development of algebra, this attitude occasioned perhaps
an incremental advance, since by the Hindus the irrational
roots of the quadratics were no longer disregarded as they
had been by the Greeks, and since to the Hindus we owe also
the immensely convenient concept of the absolute negative.
These generalizations of the number system and the conse-
quent freedom of arithmetic from geometrical representation
were to be essential in the development of the concepts of
calculus, but the Hindus could hardly have appreciated the
theoretical significance of the change. . .
The strong Greek distinction between the discreteness of
number and the continuity of geometrical magnitude was
not recognized, for it was superfluous to men who were not
bothered by the paradoxes of Zeno or his dialectic. Ques-
tions concerning incommensurability, the infinitesimal, infin-
ity, the process of exhaustion, and the other inquiries lead-
ing toward the conceptions and methods of calculus were
neglected.2
Such views have found their way generally into more popular works
on history of mathematics. For instance, we may cite the following as
being typical of the kind of opinions commonly expressed about Indian
mathematics:
As our survey indicates, the Hindus were interested in and

contributed to the arithmetical and computational activities
of mathematics rather than to the deductive patterns. Their
2
C.B.Boyer, The History of Calculus and its Conceptual development, New York
1949, p.61-62. As we shall see in the course of this article, Boyer’s assessment –
that the Indian mathematicians did not reach anywhere near the development of
calculus or mathematical analysis, because they lacked the sophisticated methodology
developed by the Greeks – seems to be thoroughly misconceived. In fact, in marked
contrast to the development of mathematics in the Greco-European tradition, the
methodology of Indian mathematical tradition seems to have ensured continued and
significant progress in all branches of mathematics till barely two hundred year ago; it
also lead to major discoveries in calculus or mathematical analysis, without in anyway
abandoning or even diluting its standards of logical rigour, so that these results, and
the methods by which they were obtained, seem as much valid today as at the time
of their discovery.
212 M.D.Srinivas
name for mathematics was gan.ita, which means “the science

of calculation”. There is much good procedure and technical
facility, but no evidence that they considered proof at all.
They had rules, but apparently no logical scruples. More-
over, no general methods or new viewpoints were arrived at
in any area of mathematics.
It is fairly certain that the Hindus did not appreciate the
significance of their own contributions. The few good ideas
they had, such as separate symbols for the numbers from 1
to 9, the conversion to base 10, and negative numbers, were
introduced casually with no realization that they were valu-
able innovations. They were not sensitive to mathematical
values. Along with the ideas they themselves advanced, they
accepted and incorporated the crudest ideas of the Egyptians
and Babylonians.3
The burden of scholarly opinion is such that even eminent mathemati-

cians, many of whom have had fairly close interaction with contempo-
rary Indian mathematics, have ended up subscribing to similar views,
as may be seen from the following remarks of one of the towering figures
of twentieth century mathematics:
For the Indians, of course, the effectiveness of the cakravāla

could be no more than an experimental fact, based on their
treatment of great many specific cases, some of them of
considerable complexity and involving (to their delight, no
doubt) quite large numbers. As we shall see, Fermat was
the first one to perceive the need for a general proof, and
Lagrange was the first to publish one. Nevertheless, to have
developed the cakravāla and to have applied it successfully
to such difficult numerical cases as N = 61, or N = 67 had
been no mean achievements.4
3
Morris Kline, Mathematical Thought from Ancient to Modern Times, Oxford
1972, p.190.
4
Andre Weil, Number Theory: An Approach through History from Hammurapi to
Legendre, Boston 1984, p.24. It is indeed ironical that Prof. Weil has credited Fermat,
who is notorious for not presenting proofs for most of the claims he made, with the
realization that mathematical results need to be justified by proofs. While the rest of
this article is purported to show that the Indian mathematicians presented logically
Modern scholarship seems to be unanimous in holding the view that

Indian mathematics is bereft of any notion of proof. But even a cur-
sory study of the source-works that are available in print would reveal
that Indian mathematicians place much emphasis on providing what
they refer to as upapatti (proof, demonstration) for every one of their
results and procedures. Some of these upapatti-s were noted in the early
European studies on Indian mathematics in the first half of the nine-
teenth century. For instance, in 1817, H.T. Colebrooke notes the fol-
lowing in the preface to his widely circulated translation of portions of
Brāhmasphut. asiddhānta of Brahmagupta and Lı̄lāvatı̄ and Bı̄jagan.ita of
Bhāskarācārya:
On the subject of demonstrations, it is to be remarked that

the Hindu mathematicians proved propositions both alge-
braically and geometrically: as is particularly noticed by
Bhāskara himself, towards the close of his algebra, where
he gives both modes of proof of a remarkable method for the
solution of indeterminate problems, which involve a factum
of two unknown quantities.5
Another notice of the fact that detailed proofs are provided in the Indian
texts on mathematics is due to C.M.Whish who, in an article published
in 1835, pointed out that infinite series for π and for trigonometric func-
tions were derived in texts of Indian mathematics much before their
‘discovery’ in Europe. Whish concluded his paper with a sample proof
rigorous proofs for most of the results and processes that they discovered, it must be
admitted that the particular example that Prof. Weil is referring to, the effective-
ness of the cakravāla algorithm (known to the Indian mathematicians at least from
the time of Jayadeva, prior to the eleventh century) for solving quadratic indetermi-
nate equations of the form x2 − N y 2 = 1, does not seem to have been demonstrated
in the available source-works. In fact, the first proof of this result was given by
Krishnaswamy Ayyangar barely seventy-five years ago (A.A.Krishnaswamy Ayyan-
gar, “New Light on Bhāskara’s Cakravāla or Cyclic Method of solving Indeterminate
Equations of the Second Degree in Two Variables’, Jour. Ind. Math. Soc. 18, 228-
248, 1929-30). Krishnaswamy Ayyangar also showed that the cakravāla algorithm is
different and more optimal than the Brouncker-Wallis-Euler-Lagrange algorithm for
solving this so-called “Pell’s Equation.”
5
H.T. Colebrooke, Algebra with Arithmetic and Mensuration from the Sanskrit of
Brahmagupta and Bhāskara, London 1817, p.xvii. Colebrooke also presents some of
the upapatti-s given by the commentators Gan.eśa Daivajña and Kr.s.n.a Daivajña, as
footnotes in his work.
214 M.D.Srinivas
from the Malayalam text Yuktibhās.ā of the theorem on the square of

the hypotenuse of a right angled triangle and also promised that:
A further account of the Yuktibhās.ā, the demonstrations of

the rules for the quadrature of the circle of infinite series,
with the series for the sines, cosines, and their demonstra-
tions, will be given in a separate paper: I shall therefore
conclude this, by submitting a simple and curious proof of
the 47th proposition of Euclid [the so called Pythagoras the-
orem], extracted from the Yuktibhās.ā.6
It would indeed be interesting to find out how the currently prevalent

view, that Indian mathematics lacks the notion of proof, obtained cur-
rency in the last 100-150 years.
2 Upapatti-s in Indian Mathematics
2.1 The tradition of Upapatti-s in Mathematics and

Astronomy
A major reason for our lack of comprehension, not merely of the Indian
notion of proof, but also of the entire methodology of Indian mathemat-
ics, is the scant attention paid to the source-works so far. It is said that
there are over one hundred thousand manuscripts on Jyotih.śāstra, which
includes, apart from works on gan.ita (mathematics and mathematical
astronomy), also those on sam 7
. hitā (omens) and hora (astrology). Only
a small fraction of these texts have been published. A well known source
book, lists about 285 published works in mathematics and mathematical
astronomy. Of these, about 50 are from the period before 12th century
AD, about 75 from 12th − 15th centuries, and about 165 from 16th − 19th
6
C.M. Whish, ‘On the Hindu Quadrature of the Circle, and the Infinite Se-
ries of the Proportion of the Circumference to the Diameter Exhibited in the Four
Shastras, the Tantrasangraham, Yucti Bhasa, Carana Paddhati and Sadratnamala’,
Trans.Roy.As.Soc.(G.B.) 3, 509-523, 1835. However, Whish does not seem to have
published any further paper on this subject.
7
D. Pingree, Jyotih.śāstra: Astral and Mathematical Literature, Wiesbaden 1981,
p.118.
centuries.8
Much of the methodological discussion is usually contained in the de-
tailed commentaries; the original works rarely touch upon such issues.
Modern scholarship has concentrated on translating and analysing the
original works alone, without paying much heed to the commentaries.
Traditionally, the commentaries have played at least as great a role in
the exposition of the subject as the original texts. Great mathemati-
cians and astronomers, of the stature of Bhāskarācārya I, Bhāskarācārya
II, Parameśvara, Nīlakan.t.ha Somasutvan, Gan.eśa Daivajña, Munīśvara
and Kamalākara, who wrote major original treatises of their own, also
took great pains to write erudite commentaries on their own works and
on the works of earlier scholars. It is in these commentaries that one
finds detailed upapatti-s of the results and procedures discussed in the
original texts, as also a discussion of the various methodological and
philosophical issues. For instance, at the beginning of his commentary
Buddhivilāsinı̄, Gan.eśa Daivajña states:
å
(ra.a:Ba.a:~k+=:ea:+.va:.ca:sa.a:ma:a.pa .sMa:~.Pu +f.a:na.Ma
v.ya.a:K.ya.a:a.va:Zea:Sa:k+.Ta:nea:na na ..ca.a:a.~ta
/// ;
a..ca.a:m,a Á
A.a.ea:pa:pa:aa:k+.Ta:neaY: a.Ka:l+sa.a.=;BUa:tea
:pa:Zya:ntua .sua:¼a:ga:Na:k+a ma:ma bua:

a;d
Äâ :
a..ca.a:m,a ÁÁ
There is no purpose served in providing further explanations

for the already lucid statements of Śrī Bhāskara. The knowl-
edgeable mathematicians may therefore note the specialty of
my intellect in the statement of upapatti-s, which are after
all the essence of the whole thing.9
Amongst the published works on Indian mathematics and astronomy,

the earliest exposition of upapatti-s are to be found in the bhās.ya of
Govindasvāmin (c 800) on Mahābhāskarı̄ya of Bhāskarācārya I, and
the Vāsanābhās.ya of Caturveda Pr.thūdakasvāmin (c 860) on Brāhma-
8
K.V. Sarma and B.V. Subbarayappa, Indian Astronomy: A Source Book, Bombay
1985.
9
Buddhivilāsinı̄ of Gan.eśa Daivajña, V.G. Apte (ed.), Vol I, Pune 1937, p.3.
216 M.D.Srinivas
sphut.asiddhānta of Brahmagupta.10 Then we find very detailed ex-

position of upapatti-s in the works of Bhāskarācārya II (c.1150): his
Vivaran.a on Śis.yadhı̄vr.ddhidatantra of Lalla and Vāsanābhās.ya on his
own Siddhāntaśiroman. i.11 Apart from these, Bhāskarācārya provides an
idea of what is an upapatti in his Bı̄javāsanā on his own Bı̄jagan.ita in
two places. In the chapter on madhyamāharan. a (quadratic equations)
he poses the following problem:
Find the hypotenuse of a plane figure, in which the side

and upright are equal to fifteen and twenty. And show the
upapatti (demonstration) of the standard procedure of com-
putation.12
Bhāskarācārya provides two upapatti-s for the solution of this problem,

the so-called Pythagoras theorem; and we shall consider them later.
Again, towards the end of the Bı̄jagan.ita in the chapter on bhāvita
(equations involving products), while considering integral solutions of
equations of the form ax + by = cxy, Bhāskarācārya explains the nature
of upapatti with the help of an example:
The upapatti (demonstration) follows. It is twofold in each

case: One geometric and the other algebraic. The geometric
demonstration is here presented. . . The algebraic demonstra-
tion is next set forth. . . This procedure [of demonstration]
has been earlier presented in a concise instructional form
[sam. ks.iptapāt.ha] by ancient teachers. The algebraic demon-
strations are for those who do not comprehend the geomet-
ric one. Mathematicians have said that algebra is compu-
tation joined with demonstration; otherwise there would be
no difference between arithmetic and algebra. Therefore this
demonstration of bhāvita has been shown in two ways.13
10
The Āryabhat.ı̄ya-bhās.ya of Bhāskara I (c.629) does occasionally indicate the
derivation of some of the mathematical procedures, though his commentary does
not purport to present upapatti-s for the rules and procedures given in Āryabhat.ı̄ya.
11
This latter commentary of Bhāskara II is a classic source of upapatti-s and needs
to be studied in depth.
12
Bı̄jagan.ita of Bhāskarācārya, Muralidhara Jha (ed.), Varanasi 1927, p.69.
13
Bı̄jagan.ita, cited above, p.125-127.
Clearly the tradition of exposition of upapatti-s is much older and

Bhāskarācārya, and the later mathematicians and astronomers are
merely following the traditional practice of providing detailed upapatti-s
in their commentaries to earlier, or their own, works.14
In Appendix A we give a list of important commentaries, available in
print, which present detailed upapatti-s. It is unfortunate that none of
the published source-works that we have mentioned above has so far
been translated into any of the Indian languages, or into English; nor
have they been studied in depth with a view to analyze the nature of
mathematical arguments employed in the upapatti-s or to comprehend
the methodological and philosophical foundations of Indian mathematics
and astronomy.15
14
Ignoring all these classical works on upapatti-s, one scholar has recently claimed
that the tradition of upapatti in India “dates from the 16th and 17th centuries”
(J.Bronkhorst, ‘Pān.ini and Euclid’, Jour. Ind. Phil. 29, 43-80, 2001).
15
We may, however, mention the following works of C.T.Rajagopal and his col-
laborators which discuss some of the upapatti-s presented in the Malayalam work
Yuktibhās.ā of Jyes.t.hadeva (c.1530) for various results in geometry, trigonometry and
those concerning infinite series for π and the trigonometric functions: K. Mukunda
Marar, ‘Proof of Gregory’s Series’, Teacher’s Magazine 15, 28-34, 1940; K. Mukunda
Marar and C.T.Rajagopal, ‘On the Hindu Quadrature of the Circle’, J.B.B.R.A.S.
20, 65-82, 1944; K. Mukunda Marar and C.T.Rajagopal, ‘Gregory’s Series in the
Mathematical Literature of Kerala’, Math Student 13, 92-98, 1945; A. Venkatara-
man, ‘Some Interesting Proofs from Yuktibhās.ā ’, Math Student 16, 1-7, 1948;
C.T.Rajagopal ‘A Neglected Chapter of Hindu Mathematics’, Scr. Math. 15, 201-
209, 1949; C.T.Rajgopal and A. Venkataraman, ‘The Sine and Cosine Power Series
in Hindu Mathematics’, J.R.A.S.B. 15, 1-13, 1949; C.T. Rajagopal and T.V.V.Aiyar,
‘On the Hindu Proof of Gregory’s Series’, Scr. Math. 17, 65-74, 1951; C.T.Rajagopal
and T.V.V.Aiyar, ‘A Hindu Approximation to Pi’, Scr.Math. 18, 25-30, 1952.
C.T.Rajagopal and M.S.Rangachari, ‘On an Untapped Source of Medieval Keralese
Mathematics’, Arch. for Hist. of Ex. Sc. 18, 89-101, 1978; C.T.Rajagopal and
M.S.Rangachari, ‘On Medieval Kerala Mathematics’, Arch. for Hist. of Ex. Sc.
35(2), 91-99, 1986.
Following the work of Rajagopal and his collaborators, there are some recent studies
which discuss some of the proofs in Yuktibhās.ā . We may here cite the following:
T.Hayashi, T.Kusuba and M.Yano, ‘The Correction of the Mādhava Series for the
Circumference of a Circle’, Centauras, 33, 149-174, 1990; Ranjan Roy, ‘The Discovery
of the Series formula for π by Leibniz, Gregory and Nīlakan.t.ha’, Math. Mag. 63,
291-306, 1990; V.J.Katz, ‘Ideas of Calculus in Islam and India’, Math. Mag. 68,
163-174, 1995; C.K.Raju, ‘Computers, Mathematics Education, and the Alternative
Epistemology of the Calculus in the Yuktibhās.ā ’, Phil. East and West 51, 325-
362, 2001; D.F.Almeida, J.K.John and A.Zadorozhnyy, ‘Keralese Mathematics: Its
Possible Transmission to Europe and the Consequential Educational Implications’,
218 M.D.Srinivas
In this article, we shall present some examples of the kinds of upapatti-

s provided in Indian mathematics, from the commentaries of Gan.eśa
Daivajña (c.1545) and Kr.s.n.a Daivajña (c.1600) on the texts Lı̄lāvatı̄
and Bı̄jagan.ita respectively, of Bhāskarācārya II (c.1150), and from the
celebrated Malayalam work Yuktibhās.ā of Jyes.t.hadeva (c.1530). We
shall also discuss how the notion of upapatti is perhaps best understood
in the larger epistemological perspective provided by Nyāya-śāstra the
Indian School of Logic. This enables us to explicate some of the impor-
tant differences between the notion of upapatti and the notion of “proof”
developed in the Greco-European tradition of Mathematics.
2.2 Mathematical results should be supported by

Upapatti-s
Before discussing some of the upapatti-s presented in Indian mathemat-

ical tradition, it is perhaps necessary to put to rest the widely prevalent
myth that the Indian mathematicians did not pay any attention to, and
perhaps did not even recognize the need for justifying the mathemat-
ical results and procedures that they employed. The large corpus of
upapatti-s, even amongst the small sample of source-works published so
far, should convince anyone that there is no substance to this myth.
Still, we may cite the following passage from Kr.s.n.a Daivajña’s commen-
tary Bı̄japallava on Bı̄jagan.ita of Bhāskarācārya, which clearly brings
out the basic understanding of Indian mathematical tradition that cit-
ing any number of instances (even an infinite number of them) where a
particular result seems to hold, does not amount to establishing that as
a valid result in mathematics; only when the result is supported by a
upapatti or a demonstration, can the result be accepted as valid:
How can we state without proof (upapatti) that twice the

product of two quantities when added or subtracted from
J. Nat. Geo. 20, 77-104, 2001; D.Bressoud, ‘Was Calculus Invented in India?’,
College Math. J. 33, 2-13, 2002; J.K.John, ‘Derivation of the Sam . skāras applied
to the Mādhava Series in Yuktibhās.ā ’, in M.S.Sriram, K.Ramasubramanian and
M.D.Srinivas (eds.), 500 Years of Tantrasaṅgraha : A Landmark in the History of
Astronomy, Shimla 2002, p 169-182. An outline of the proofs given in Yuktibhās.ā can
also be found in T.A. Saraswati Amma, 1979, cited earlier, and in S.Parameswaran,
The Golden Age of Indian Mathematics, Kochi 1998.
the sum of their squares is equal to the square of the sum

or difference of those quantities? That it is seen to be so
in a few instances is indeed of no consequence. Otherwise,
even the statement that four times the product of two quan-
tities is equal to the square of their sum, would have to be
accepted as valid. For, that is also seen to be true in some
cases. For instance, take the numbers 2, 2. Their product
is 4, four times which will be 16, which is also the square
of their sum 4. Or take the numbers 3, 3. Four times their
product is 36, which is also the square of their sum 6. Or
take the numbers 4, 4. Their product is 16, which when mul-
tiplied by four gives 64, which is also the square of their sum
8. Hence, the fact that a result is seen to be true in some
cases is of no consequence, as it is possible that one would
come across contrary instances (vyabhicāra) also. Hence it
is necessary that one would have to provide a proof (yukti)
for the rule that twice the product of two quantities when
added or subtracted from the sum of their squares results in
the square of the sum or difference of those quantities. We
shall provide the proof (upapatti) in the end of the section
on ekavarn.a-madhyamāharan. a.16
2.3 Square of the hypotenuse of a right-angled triangle:

the so-called Pythagoras Theorem
Gan.eśa provides two upapatti-s for the rule concerning the square of the
hypotenuse (karn.a) of a right-angled triangle.17 These upapatti-s are
the same as the ones outlined by Bhāskarācārya II in his Bı̄javāsanā on
his own Bı̄jagan.ita, that we referred to earlier. The first involves the
avyakta method and proceeds as follows:
16
Bı̄japallava of Kr.s.n.a Daivajña, T.V. Radhakrishna Sastri (ed.), Tanjore, 1958,
p.54.
17
Buddhivilāsinı̄, cited earlier, p.128-129.
220 M.D.Srinivas
Take the hypotenuse (karn.a) as the

base and assume it to be yā. Let
the bhujā and kot.i (the two sides)
be 3 and 4 respectively. Take
the hypotenuse as the base and
4
3 draw the perpendicular to the hy-
potenuse from the opposite vertex
as in the figure. [This divides the
ya triangle into two triangles, which

are similar to the original] Now by
9 16 the rule of proportion ( anupāta),
yā = yā + yā
if yā is the hypotenuse the bhujā is
yā2 = 25
3, then when this bhujā 3 is the hy-
yā = 5
potenuse, the bhujā, which is now
the ābādhā (segment of the base)
on the side of the original bhujā will
9
be ( yā ).
Again if yā is the hypotenuse, the
kot.i is 4, then when this kot.i 4 is
the hypotenuse, the kot.i, which is
now the segment of base on the side
16
of the (original) kot.i will be ( yā ).
Adding the two segments (ābādhā-
s) of yā the hypotenuse and equat-
ing the sum to (the hypotenuse)
yā, cross-multiplying and taking
the square-roots, we get yā = 5,
the square root of the sum of the
squares of bhujā and kot.i .
The other upapatti of Gan.eśa is ks.etragata or geometrical, and proceeds

as follows:18
18
This method seems to be known to Bhāskarācārya I (c.629 AD) who gives a
very similar diagram in his Āryabhat.ı̄yabhās.ya, K.S. Shukla (ed.), Delhi 1976, p.48.
The Chinese mathematician Liu Hui (c 3rd century AD) seems to have proposed
similar geometrical proofs of this so-called Pythagoras Theorem. See for instance,
D.B.Wagner, ‘A Proof of the Pythagorean Theorem by Liu Hui’, Hist. Math.12,
71-3, 1985.
Take four triangles identical to

the given and taking the four hy-
potenuses to be the four sides,
form the square as shown. Now,
the interior square has for its side
the difference of bhujā and kot.i
. The area of each triangle is
half the product of bhujā and kot.i
and four times this added to the
area of the interior square is the
area of the total figure. This is
twice the product of bhujā and
kot.i added to the square of their
difference. This, by the earlier
c2 = (a − b)2 + 4( 12 ab)
cited rule, is nothing but the sum
= a2 + b2
of the squares of bhujā and kot.i
. The square root of that is the
side of the (big) square, which is
nothing but the hypotenuse.
2.4 The rule of signs in Algebra
One of the important aspects of Indian mathematics is that in many

upapatti-s the nature of the underlying mathematical objects plays an
important role. We can for instance, refer to the upapatti given by Kr.s.n.a
Daivajña for the well-known rule of signs in Algebra. While providing an
upapatti for the rule, “the number to be subtracted if positive (dhana) is
made negative (r.n.a) and if negative is made positive”, Kr.s.n.a Daivajña
states:
Negativity (r.n.atva) here is of three types – spatial, tempo-

ral and that pertaining to objects. In each case, it [neg-
ativity] is indeed the vaiparı̄tya or the oppositeness. . .For
instance, the other direction in a line is called the oppo-
site direction (viparı̄ta dik); just as west is the opposite of
east. . . Further, between two stations if one way of traversing
is considered positive then the other is negative. In the same
way, past and future time intervals will be mutually negative
222 M.D.Srinivas
of each other. . . Similarly, when one possesses said objects

they would be called his dhana (wealth). The opposite would
be the case when another owns the same objects. . . Amongst
these [different conceptions], we proceed to state the upa-
patti of the above rule, assuming positivity (dhanatva) for
locations in the eastern direction and negativity (r.n.atva) for
locations in the west, as follows. . . 19
Kr.s.n.a Daivajña goes on to explain how the distance between a pair of

stations can be computed knowing that between each of these stations
and some other station on the same line. Using this he demonstrates
the above rule that “the number to be subtracted if positive is made
negative. . . ”
2.5 The Kut..taka process for the solution of linear

indeterminate equations
To understand the nature of upapatti in Indian mathematics one will

have to analyse some of the lengthy demonstrations which are presented
for the more complicated results and procedures. One will also have to
analyse the sequence in which the results and the demonstrations are
arranged to understand the method of exposition and logical sequence
of arguments. For instance, we may refer to the demonstration given by
Kr.s.n.a Daivajña20 of the well-known kut..taka procedure, which has been
employed by Indian mathematicians at least since the time of Āryabhat.a
(c 499 AD), for solving first order indeterminate equations of the form
(ax + c)
= y,
b
where a, b, c are given integers and x, y are to be solved for in integers.

Since this upapatti is rather lengthy, we merely recount the essential
steps here.21
19
Bı̄japallava, cited above, p.13.
20
Bı̄japallava, cited above, p.85-99.
21
A translation of the upapatti may be found in M.D.Srinivas, ‘Methodology of
Indian Mathematics and its Contemporary Relevance’, PPST Bulletin, 12, 1-35,
1987.
Kr.s.n.a Daivajña first shows that the solutions for x, y do not vary if we
factor all the three numbers a, b, c by the same common factor. He then
shows that if a and b have a common factor, then the above equation
will not have a solution unless c is also divisible by the same common
factor. Then follows the upapatti of the process of finding the greatest
common factor of a and b by mutual division, the so-called Euclidean al-
gorithm. He then provides an upapatti for the kut..taka method of finding
the solution which involves carrying out a sequence of transformations
on the vallı̄ (line or column) of quotients obtained in the above mutual
division. This is based on a detailed analysis of the various operations
in reverse (vyasta-vidhi). The last two elements of the vallı̄, at each
stage, are shown to be the solutions of the kut..taka problem involving
the successive pair of remainders (taken in reverse order from the end)
which arise in the mutual division of a and b. Finally, it is shown how
the procedure differs depending upon whether there are odd or even
number of coefficients generated in the above mutual division.
2.6 Nīlakan.t.ha’s proof for the sum of an infinite

geometric series
In his Āryabhat.ı̄yabhās.ya while deriving an interesting approximation

for the arc of circle in terms of the jyā (Rsine) and the śara (Rversine),
the celebrated Kerala astronomer Nīlakan.t.ha Somasutvan presents a de-
tailed demonstration of how to sum an infinite geometric series. Though
it is quite elementary compared to the various other infinite series ex-
pansions derived in the works of the Kerala School, we shall present
an outline of Nīlakan.t.ha’s argument as it clearly shows how the no-
tion of limit was well understood in the Indian mathematical tradition.
Nīlakan.t.ha first states the general result22
3
1 1 2 1 a
a + + + ... = .
r r r r−1
where the left hand side is an infinite geometric series with the succes-
sive terms being obtained by dividing by a cheda (common divisor), r,
assumed to be greater than 1. Nīlakan.t.ha notes that this result is best
22
Āryabhat.ı̄yabhās.ya of Nīlakan.t.ha, Gan.itapāda, K.Sambasiva Sastri (ed.), Trivan-
drum 1931, p.142-143.
224 M.D.Srinivas
demonstrated by considering a particular case, say r = 4. Thus, what

is to be demonstrated is that
2 3
1 1 1 1
+ + + ... = .
4 4 4 3
Nīlakan.t.ha first obtains the sequence of results
1 1 1
= + ,
3 4 (4.3)
1 1 1
= + ,
(4.3) (4.4) (4.4.3)
1 1 1
= + ,
(4.4.3) (4.4.4) (4.4.4.3)
and so on, from which he derives the general result
2 n n
1 1 1 1 1 1
− + + ... + = .
3 4 4 4 4 3
Nīlakan.t.ha then goes on to present the following crucial argument to

derive the sum of the infinite geometric series: As we sum more terms,
the difference between 13 and sum of powers of 14 (as given by the right
hand side of the above equation), becomes extremely small, but never
zero. Only when we take all the terms of the infinite series together do
we obtain the equality
2 n
1 1 1 1
+ + ... + + ... = .
4 4 4 3
2.7 Yuktibhās.ā proofs of infinite series for π and the

trigonometric functions
One of the most celebrated works in Indian mathematics and astron-

omy, which is especially devoted to the exposition of yukti or proofs, is
the Malayalam work Yuktibhās.ā (c.1530) of Jyes.t.hadeva.23 Jyes.t.hadeva
23
Yuktibhās.ā of Jyes.t.hadeva, K. Chandrasekharan (ed.), Madras 1953.
Gan.itādhyāya alone was edited along with notes in Malayalam by Ramavarma
Thampuran and A.R.Akhileswara Aiyer, Trichur 1948. The entire work has been
edited, along with an ancient Sanskrit version, Gan.itayuktibhās.ā and English transla-
tion, by K.V.Sarma, with explanatory mathematical notes by K.Ramasubramanian,
M.D.Srinivas and M.S.Sriram (in press).
states that his work closely follows the renowned astronomical work
Tantrasaṅgraha (c.1500) of Nīlakan.t.ha Somasutvan and is intended to
give a detailed exposition of all the mathematics required thereof. The
first half of Yuktibhās.ā deals with various mathematical topics in seven
chapters and the second half deals with all aspects of mathematical as-
tronomy in eight chapters. The mathematical part includes a detailed
exposition of proofs for the infinite series and fast converging approx-
imations for π and the trigonometric functions, which were discovered
by Mādhava (c.1375). We present an outline of some of these proofs in
Appendix B.
3 Upapatti and “Proof”
3.1 Mathematics as a search for infallible eternal truths
The notion of upapatti is significantly different from the notion of ‘proof’

as understood in the Greek as well as the modern Western tradition
of mathematics. The ideal of mathematics in the Greek and modern
Western traditions is that of a formal axiomatic deductive system; it
is believed that mathematics is and ought to be presented as a set of
formal derivations from formally stated axioms. This ideal of mathe-
matics is intimately linked with another philosophical presupposition –
that mathematics constitutes a body of infallible eternal truths. Per-
haps it is only the ideal of a formal axiomatic deductive system that
could presumably measure up to this other ideal of mathematics being
a body of infallible eternal truths. It is this quest for securing certainty
of mathematical knowledge, which has motivated most of the founda-
tional and philosophical investigations into mathematics and shaped the
course of mathematics in the Western tradition, from the Greeks to the
contemporary times.
The Greek view of mathematical objects and the nature of mathematical
knowledge is clearly set forth in the following statement of Proclus (c.
5th century AD) in his famous commentary on the Elements of Euclid:
Mathematical being necessarily belongs neither among the

first nor among the last and least simple kinds of being, but
occupies the middle ground between the partless realities –
226 M.D.Srinivas
simple, incomposite and indivisible – and divisible things

characterized by every variety of composition and differenti-
ation. The unchangeable, stable and incontrovertible char-
acter of the propositions about it shows that it is superior
to the kind of things that move about in matter . . .
It is for this reason, I think, that Plato assigned different
types of knowing to the highest, the intermediate, and the
lowest grades of reality. To indivisible realities he assigned
intellect, which discerns what is intelligible with simplicity
and immediacy, and by its freedom from matter, its purity,
and its uniform mode of coming in contact with being is su-
perior to all other forms of knowledge. To divisible things
in the lowest level of nature, that is, to all objects of sense
perception, he assigned opinion, which lays hold of truth ob-
scurely, whereas to intermediates, such as the forms studied
by mathematics, which fall short of indivisible but are supe-
rior to divisible nature, he assigned understanding. . . .
Hence Socrates describes the knowledge of understandables
as being more obscure than the highest science but clearer
than the judgements of opinion. For, the mathematical sci-
ences are more explicative and discursive than intellectual
insight but are superior to opinion in the stability and ir-
refutability of their ideas. And their proceeding from hy-
pothesis makes them inferior to highest knowledge, while
their occupation with immaterial objects makes their knowl-
edge more perfect than sense perception.24
While the above statement of Proclus is from the Platonist school, the
Aristotelean tradition also held more or less similar views on the nature
of mathematical knowledge, as may be seen from the following extract
from the canonical text on Mathematical Astronomy, the Almagest of
Claudius Ptolemy (c.2nd century AD):
For Aristotle divides theoretical philosophy too,very fittingly,

into three primary categories, physics, mathematics and the-
ology. For everything that exists is composed of matter, form
24
Proclus: A Commentary on the First Book of Euclid’s Elements, Tr.G.R.Morrow,
Princeton, 1970, p.3,10.
and motion; none of these [three] can be observed in its sub-

stratum by itself, without the others: they can only be imag-
ined. Now the first cause of the first motion of the universe,
if one considers it simply, can be thought of as invisible and
motionless deity; the division [of theoretical philosophy] con-
cerned with investigating this [can be called] ‘theology’, since
this kind of activity, somewhere up in the highest reaches of
the universe, can only be imagined, and is completely sep-
arated from perceptible reality. The division [of theoretical
philosophy] which investigates material and ever-moving na-
ture, and which concerns itself with ‘white’, ‘hot’, ‘sweet’,
‘soft’ and suchlike qualities one may call ‘physics’; such an
order of being is situated (for the most part) amongst cor-
ruptible bodies and below the lunar sphere. That division [of
theoretical philosophy] which determines the nature involved
in forms and motion from place to place, and which serves
to investigate shape, number, size and place, time and such-
like, one may define as ‘mathematics’. Its subject-matter
falls as it were in the middle between the other two, since,
firstly, it can be conceived of both with and without the aid
of the senses, and, secondly, it is an attribute of all existing
things without exception, both mortal and immortal: for
those things which are perpetually changing in their insep-
arable form, it changes with them, while for eternal things
which have an aethereal nature, it keeps their unchanging
form unchanged.
From all this we concluded: that the first two divisions of the-
oretical philosophy should rather be called guesswork than
knowledge, theology because of its completely invisible and
ungraspable nature, physics because of the unstable and un-
clear nature of matter; hence there is no hope that philoso-
phers will ever be agreed about them; and that only math-
ematics can provide sure and unshakeable knowledge to its
devotees, provided one approaches it rigorously. For its kind
of proof proceeds by indisputable methods, namely arith-
metic and geometry. Hence we are drawn to the investigation
of that part of theoretical philosophy, as far as we were able
to the whole of it, but especially to the theory concerning
228 M.D.Srinivas
the divine and heavenly things. For that alone is devoted to

the investigation of the eternally unchanging. For that rea-
son it too can be eternal and unchanging (which is a proper
attribute of knowledge) in its own domain, which is neither
unclear nor disorderly.25
The view, that it is mathematics which can provide “sure and unshake-
able knowledge to its devotees” has persisted in the Greco-European
tradition down to the modern times. For instance, we may cite the
popular mathematician philosopher of our times, Bertrand Russel, who
declares, “I wanted certainty in the kind of way in which people want
religious faith. I thought that certainty is more likely to be found in
mathematics than elsewhere”. In a similar vein, David Hilbert, one of
the foremost mathematicians of our times declared, “The goal of my
theory is to establish once and for all the certitude of mathematical
methods”.26
3.2 The raison d’être of Upapatti
Indian epistemological position on the nature and validation of mathe-

matical knowledge is very different from that in the Western tradition.
This is brought out for instance by the Indian understanding of what
indeed is the purpose or raison d’être of an upapatti. In the beginning
of the golādhyāya of Siddhāntaśiroman. i, Bhāskarācārya says:
ma:Dya.a:dù;aMÅ dù;aÅu :sa:d.Ma ya:d.a ga:a.Na:tMa ta:~ya.ea:pa:pa:aMa ;a.va:na.a
:pra.Ea: a.QM :pra.Ea:Q;sa:Ba.a:sua .nEa: a.ta ga:Na:k+ea ;

a.naH;sMa:Za:ya.ea na .~va:ya:m,a Á
x :Zya:tea
ga.ea:le .sa.a ;a.va:ma:l;a k+=:a:ma:l+k+.va:t,a :pra:tya:[a:ta.ea d
ta:sma.a:d:s}yua:pa:pa:aa:ba.ea:Da:a.va:Da:yea ga.ea:l+pra:ba:nDa.ea:dù;a:taH

Å ÁÁ
27
Without the knowledge of upapatti-s, by merely mastering

the gan.ita (calculational procedures) described here, from
25
The Almagest of Ptolemy, Translated by G.J.Toomer, London 1984, p.36-7.
26
Both quotations cited in Ruben Hersh, ‘Some Proposals for Reviving the Philos-
ophy of Mathematics’, Adv. Math. 31, 31-50, 1979.
27
Siddhāntaśiroman.i of Bhāskarācārya with Vāsanābhās.ya and Vāsanāvārttika of
Nr.sim. ha Daivajña, Muralidhara Chaturveda (ed.), Varanasi 1981, p.326.
the madhyamādhikara (the first chapter of Siddhāntaśiro-

man.i) onwards, of the (motion of the) heavenly bodies, a
mathematician will not have any value in the scholarly as-
semblies; without the upapatti-s he himself will not be free
of doubt (nih.sam. śaya). Since upapatti is clearly perceivable
in the (armillary) sphere like a berry in the hand, I there-
fore begin the golādhyāya (section on spherics) to explain the
upapatti-s.
As the commentator Nr.sim . ha Daivajña explains, ‘the phala (object) of

upapatti is pān.d.itya (scholarship) and also removal of doubts (for oneself)
which would enable one to reject wrong interpretations made by others
due to bhrānti (confusion) or otherwise.’28
The same view is reiterated by Gan.eśa Daivajña in his preface to Bud-
dhivilāsinı̄:
v.ya:e va.a:v.ya:+.sMa:¼ea ya:du:a.d:ta:ma: a.Ka:lM na.ea:pa:pa:aMa ;a.va:na.a ta:t,a
;
a.na:Bra.Ra:nta.ea va.a +tea ta.Ma .sua:ga:Na:k+.sa:d:a.sa :pra.Ea:Q;ta.Ma .nEa: a.ta ..ca.a:ya:m,a Á
:pra:tya:[Ma dx:Zya:tea .sa.a k+=;ta:l+k+.

a.l+ta.a:d:ZRa:va:t,a .sua:pra:sa:a.a
Äâ :vxa.;d
ta:sma.a:d:g{ya.ea:pa:pa:aMa ;
a.na:ga:a.d:tua:ma: a.Ka:l+m,a o+tsa:he bua:
a;d Äù;aEÅ ÁÁ
29
Whatever is stated in the vyakta or avyakta branches of

mathematics, without upapatti, will not be rendered nirbhrā-
nta (free from confusion); will not have any value in an as-
sembly of mathematicians. The upapatti is directly perceiv-
able like a mirror in hand. It is therefore, as also for the
elevation of the intellect (buddhi-vr.ddhi), that I proceed to
enunciate upapatti-s in entirety.
Thus as per the Indian mathematical tradition, the purpose of upapatti

is mainly (i) To remove doubts and confusion regarding the validity
and interpretation of mathematical results and procedures; and, (ii) To
obtain assent in the community of mathematicians.
Further, in the Indian tradition, mathematical knowledge is not taken
to be different in any ‘fundamental sense’ from that in natural sci-
28
Siddhantaśiroman.i , cited above, p.326.
29
Buddhivilāsinı̄, cited above, p.3.
230 M.D.Srinivas
ences. The valid means for acquiring knowledge in mathematics are

the same as in other sciences: Pratyaks.a (perception), Anumāna (infer-
ence), Śabda or Āgama (authentic tradition). In his Vāsanābhās. ya on
Siddhāntaśiroman. i Bhāskarācārya refers to the sources of valid knowl-
edge (pramān. a) in mathematical astronomy, and declares that
ya:dù;aeÅ :va:mua:.cya:tea ga:a.Na:ta:~k+.nDea o+pa:pa:aa:ma.a:n,a O;:va.a:ga:maH :pra:ma.a:Na:m,a Á

30
For all that is discussed in Mathematical Astronomy, only an

authentic tradition or established text which is supported by
upapatti will be a pramān.a.
Upapatti here includes observation. Bhāskarācārya, for instance, says

that the upapatti for the mean periods of planets involves observations
over very long periods.
3.3 The limitations of Tarka or proof by contradiction
An important feature that distinguishes the upapatti-s of Indian math-

ematicians is that they do not generally employ the method of proof
by contradiction or reductio ad absurdum. Sometimes arguments, which
are somewhat similar to the proof by contradiction, are employed to
show the non-existence of an entity, as may be seen from the following
upapatti given by Kr.s.n.a Daivajña to show that “a negative number has
no square root”:
The square-root can be obtained only for a square. A neg-

ative number is not a square. Hence how can we consider
its square-root? It might however be argued: ‘Why will a
negative number not be a square? Surely it is not a royal
fiat’. . . Agreed. Let it be stated by you who claim that a
negative number is a square as to whose square it is: Surely
not of a positive number, for the square of a positive number
is always positive by the rule. . .Not also of a negative num-
ber. Because then also the square will be positive by the
rule. . .This being the case, we do not see any such number
whose square becomes negative. . . 31
30
Siddhāntaśiroman.i, cited above, p.30.
31
Bı̄japallava, cited earlier, p.19.
Such arguments, known as tarka in Indian logic, are employed only

to prove the non-existence of certain entities, but not for proving the
existence of an entity, which existence is not demonstrable (at least in
principle) by other direct means of verification.
In rejecting the method of indirect proof as a valid means for estab-
lishing existence of an entity which existence cannot even in principle
be established through any direct means of proof, the Indian mathe-
maticians may be seen as adopting what is nowadays referred to as the
‘constructivist’ approach to the issue of mathematical existence. But the
Indian philosophers, logicians, etc., do much more than merely disallow
certain existence proofs. The general Indian philosophical position is
one of eliminating from logical discourse all reference to such aprasiddha
entities, whose existence is not even in principle accessible to all means
of verification.32 This appears to be also the position adopted by the
Indian mathematicians. It is for this reason that many an “existence
theorem” (where all that is proved is that the non-existence of a hypo-
thetical entity is incompatible with the accepted set of postulates) of
Greek or modern Western mathematics would not be considered signif-
icant or even meaningful by Indian mathematicians.
3.4 Upapatti and “Proof”
We now summarize our discussion on the classical Indian understanding

of the nature and validation of mathematical knowledge:
1. The Indian mathematicians are clear that results in mathematics,

even those enunciated in authoritative texts, cannot be accepted
as valid unless they are supported by yukti or upapatti . It is not
enough that one has merely observed the validity of a result in a
large number of instances.
2. Several commentaries written on major texts of Indian mathemat-

ics and astronomy present upapatti-s for the results and procedures
enunciated in the text.
32
For the approach adopted by Indian philosophers to tarka or the method of
indirect proof see for instance, M.D.Srinivas, “The Indian Approach to Formal Logic
and the Methodology of Theory Construction: A Preliminary View”, PPST Bulletin
9, 32-59, 1986.
232 M.D.Srinivas
3. The upapatti-s are presented in a sequence proceeding systemat-

ically from known or established results to finally arrive at the
result to be established.
4. In the Indian mathematical tradition the upapatti-s mainly serve

to remove doubts and obtain consent for the result among the
community of mathematicians.
5. The upapatti-s may involve observation or experimentation. They

also depend on the prevailing understanding of the nature of the
mathematical objects involved.
6. The method of tarka or “proof by contradiction” is used occa-

sionally. But there are no upapatti-s which purport to establish
existence of any mathematical object merely on the basis of tarka
alone.
7. The Indian mathematical tradition did not subscribe to the ideal

that upapatti-s should seek to provide irrefutable demonstrations
establishing the absolute truth of mathematical results. There
was apparently no attempt to present the upapatti-s as a part of
a deductive axiomatic system. While Indian mathematics made
great strides in the invention and manipulation of symbols in rep-
resenting mathematical results and in facilitating mathematical
processes, there was no attempt at formalization of mathematics.
The classical Indian understanding of the nature and validation of math-

ematical knowledge seems to be rooted in the larger epistemological
perspective developed by the Nyāya school of Indian logic. Some of the
distinguishing features of Nyāya logic, which are particularly relevant in
this context, are: That it is a logic of cognitions (jñāna) and not “propo-
sitions”, that it has no concept of pure “formal validity” as distinguished
from “material truth”, that it does not distinguish necessary and con-
tingent truth or analytical and synthetic truth, that it does not admit,
in logical discourse, premises which are known to be false or terms that
are non-instantiated, that it does not accord tarka or “proof by contra-
diction” a status of independent pramān. a or means of knowledge, and
so on.33
33
For a discussion of some of these features, see J.N.Mohanty: Reason and Tradition
in Indian Thought, Oxford, 1992.
The close relation between the methodology of Indian mathematics and

Nyāya epistemology, has been commented upon by a leading scholar of
navya-nyāya:
The western concept of proof owes its origin to Plato’s dis-

tinction between knowledge and opinion or between reason
and sense. According to Plato, reason not merely knows ob-
jects having ontological reality, but also yields a knowledge
which is logically superior to opinion to which the senses
can aspire. On this distinction is based the distinction be-
tween contingent and necessary truths, between material
truth and formal truth, between rational knowledge which
can be proved and empirical knowledge which can only be
verified . . .
As a matter of fact, the very concept of reason is unknown
in Indian philosophy. In the systems which accept inference
as a source of true knowledge, the difference between percep-
tion and inference is not explained by referring the two to
two different faculties of the subject, sense and reason, but
by showing that inferential knowledge is caused in a special
way by another type of knowledge (vyāpti-jñāna [knowledge
of invariable concomitance]), whereas perception is not so
caused . . .
In Indian mathematics we never find a list of self-evident
propositions which are regarded as the basic premises from
which other truths of mathematics follow . . .
Euclid was guided in his axiomatization of geometry by the
Aristotelean concept of science as a systematic study with a
few axioms which are self-evident truths. The very concept
of a system thus involves a distinction between truths which
need not be proved (either because they are self-evident as
Aristotle thought, or because they have been just chosen as
the primitive propositions of a system as the modern logi-
cians think) and truths which require proof. But this is not
enough. What is important is to suppose that the number
of self-evident truths or primitive propositions is very small
and can be exhaustively enumerated.
234 M.D.Srinivas
Now there is no Indian philosophy which holds that some

truths do not require any proof while others do. The sys-
tems which accept svatah.prāmān. yavāda hold that all (true)
knowledge is self-evidently true, and those which accept para-
tah.prāmān.yavāda hold that all (true) knowledge requires
proof; there is no system which holds that some truths re-
quire proof while others do not . . . 34
3.5 Towards a new epistemology for Mathematics
Mathematics today, rooted as it is in the modern Western tradition,

suffers from serious limitations. Firstly, there is the problem of ‘foun-
dations’ posed by the ideal view of mathematical knowledge as a set
of infallible eternal truths. The efforts of mathematicians and philoso-
phers of the West to secure for mathematics the status of indubitable
knowledge has not succeeded; and there is a growing feeling that this
goal may turn out to be a mirage.
After surveying the changing status of mathematical truth from the
Platonic position of “truth in itself”, through the early twentieth cen-
tury position that “mathematical truth resides . . . uniquely in the logi-
cal deductions starting from premises arbitrarily set by axioms”, to the
twentieth century developments which question the infallibility of these
logical deductions themselves, Bourbaki are forced to conclude that:
To sum up, we believe that mathematics is destined to sur-

vive, and that the essential parts of this majestic edifice will
never collapse as a result of the sudden appearance of a con-
tradiction; but we cannot pretend that this opinion rests on
anything more than experience. Some will say that this is
small comfort; but already for two thousand five hundred
years mathematicians have been correcting their errors to
the consequent enrichment and not impoverishment of this
science; and this gives them the right to face the future with
serenity.35
34
Sibajiban Bhattacharya, ‘The Concept of Proof in Indian Mathematics and
Logic’, in Doubt, Belief and Knowledge, Delhi, 1987, p.193, 196.
35
N.Bourbaki, Elements of Mathematics: Theory of Sets, Springer 1968, p.13; see
also N.Bourbaki, Elements of History of Mathematics, Springer 1994, p.1-45.
Apart from the problems inherent in the goals set for mathematics,
there are also other serious inadequacies in the Western epistemology
and philosophy of mathematics. The ideal view of mathematics as a
formal deductive system gives rise to serious distortions. Some scholars
have argued that this view of mathematics has rendered philosophy of
mathematics barren and incapable of providing any understanding of
the actual history of mathematics, the logic of mathematical discovery
and, in fact, the whole of creative mathematical activity.36
There is also the inevitable chasm between the ideal notion of infalli-
ble mathematical proof and the actual proofs that one encounters in
standard mathematical practice, as portrayed in a recent book:
On the one side, we have real mathematics, with proofs,

which are established by the ‘consensus of the qualified’. A
real proof is not checkable by a machine, or even by any
mathematician not privy to the gestalt, the mode of thought
of the particular field of mathematics in which the proof is
located. Even to the ‘qualified reader’ there are normally
differences of opinion as to whether a real proof (i.e., one
that is actually spoken or written down) is complete or cor-
rect. These doubts are resolved by communication and ex-
planation, never by transcribing the proof into first order
predicate calculus. Once a proof is ‘accepted’, the results of
the proof are regarded as true (with very high probability).
It may take generations to detect an error in a proof. . . On
the other side, to be distinguished from real mathematics, we
have ‘meta-mathematics’. . . It portrays a structure of proofs,
which are indeed infallible ‘in principle’. . . [The philosophers
of mathematics seem to claim] that the problem of fallibility
in real proofs. . . has been conclusively settled by the presence
of a notion of infallible proof in meta-mathematics. . . One
wonders how they would justify such a claim.37
Apart from the fact that the modern Western epistemology of mathe-
matics fails to give an adequate account of the history of mathematics
36
I.Lakatos, Proofs and Refutations: The Logic of Mathematical Discovery, Cam-
bridge 1976.
37
Philip J.Davis and Reuben Hersh, The Mathematical Experience, Boston, 1981,
p.354-5.
236 M.D.Srinivas
and standard mathematical practice, there is also the growing awareness

that the ideal of mathematics as a formal deductive system has had seri-
ous consequences in the teaching of mathematics. The formal deductive
format adopted in mathematics books and articles greatly hampers un-
derstanding and leaves the student with no clear idea of what is being
talked about.
Notwithstanding all these critiques, it is not likely that, within the West-
ern philosophical tradition, any radically different epistemology of math-
ematics will emerge; and so the driving force for modern mathematics is
likely to continue to be a search for infallible eternal truths and modes of
establishing them, in one form or the other. This could lead to ‘progress’
in mathematics, but it would be progress of a rather limited kind.
If there is a major lesson to be learnt from the historical development of
mathematics, it is perhaps that the development of mathematics in the
Greco-European tradition was seriously impeded by its adherence to the
cannon of ideal mathematics as laid down by the Greeks. In fact, it is
now clearly recognized that the development of mathematical analysis
in the Western tradition became possible only when this ideal was given
up during the heydays of the development of “infinitesimal calculus”
during 16th – 18th centuries. As one historian of mathematics notes:
It is somewhat paradoxical that this principal shortcoming

of Greek mathematics stemmed directly from its principal
virtue–the insistence on absolute logical rigour. . .Although
the Greek bequest of deductive rigour is the distinguishing
feature of modern mathematics, it is arguable that, had all
the succeeding generations also refused to use real numbers
and limits until they fully understood them, the calculus
might never have been developed and mathematics might
now be a dead and forgotten science.38
It is of course true that the Greek ideal has gotten reinstated at the heart
of mathematics during the last two centuries, but it seems that most of
the foundational problems of mathematics can also be perhaps traced
to the same development. In this context, study of alternative episte-
mologies such as that developed in the Indian tradition of mathematics,
could prove to be of great significance for the future of mathematics.
38
C.H.Edwards, History of Calculus, New York 1979, p.79.
Appendices
A List of Works Containing Upapatti-s
The following are some of the important commentaries available in print,

which present upapatti-s of results and procedures in mathematics and
astronomy:
1. Bhās.ya of Bhāskara I (c.629) on Āryabhat.ı̄ya of Āryabhat.a (c.499),

K.S.Shukla (ed.), New Delhi 1975.
2. Bhās.ya of Govindasvāmin (c.800) on Mahābhāskarı̄ya of Bhāskara

I (c.629), T.S.Kuppanna Sastri (ed.), Madras 1957.
3. Vāsanābhās.ya of Caturveda Pr.thūdakasvāmin (c.860) on

Brāhmasphut. asiddhānta of Brahmagupta (c.628), Chs. I-III, XXI,
Ramaswarup Sharma (ed.), New Delhi 1966; Ch XXI, Edited and
Translated by Setsuro Ikeyama, Ind. Jour. Hist. Sc. Vol. 38,
2003.
4. Vivaran.a of Bhāskarācārya II (c.1150) on Śis.yadhı̄vr.ddhidatantra

of Lalla (c.748), Chandrabhanu Pandey (ed.), Varanasi 1981.
5. Vāsanā of Bhāskarācārya II (c.1150) on his own Bı̄jagan.ita, Ji-

vananda Vidyasagara (ed.), Calcutta 1878; Achyutananda Jha
(ed.) Varanasi 1949, Rep. 1994.
6. Mitāks.arā or Vāsanā of Bhāskarācārya II (c.1150) on his own

Siddhāntaśiroman. i, Bapudeva Sastrin (ed.), Varanasi 1866; Mu-
ralidhara Chaturveda (ed.), Varanasi 1981.
7. Vāsanābhās.ya of Āmarāja (c.1200) on Khand.akhādyaka of Brah-

magupta (c.665), Babuaji Misra (ed.), Calcutta 1925.
8. Gan.itabhūs.an.a of Makkībhat.t.a (c.1377) on Siddhāntaśekhara of

Śrīpati (c.1039), Chs. I-III, Babuaji Misra (ed.), Calcutta 1932.
9. Siddhāntadı̄pikā of Parameśvara (c.1431) on the Bhās.ya of Govin-

dasvāmin (c.800) on Mahābhāskarı̄ya of Bhāskara I (c.629), T.S.
Kuppanna Sastri (ed.), Madras 1957.
238 M.D.Srinivas
10. Āryabhat.ı̄yabhās.ya of Nīlakan.t.ha Somasutvan (c.1501) on

Āryabhat.ı̄ya of Āryabhat.a (c.499), K. Sambasiva Sastri (ed.), 3
Vols., Trivandrum 1931, 1932, 1957.
11. Yuktibhās.ā (in Malayalam) of Jyes.t.hadeva (c.1530); Gan.itādhyāya,

RamaVarma Thampuran and A.R. Akhileswara Aiyer (eds.),
Trichur 1948; K.Chandrasekharan (ed.), Madras 1953. Edited and
Translated by K.V. Sarma with Explanatory Notes by K. Rama-
subramanian, M.D. Srinivas and M.S. Sriram (in Press).
12. Yuktidı̄pikā of Śaṅkara Vāriyar (c.1530) on Tantrasaṅgraha of

Nīlakan.t.ha Somasutvan (c.1500), K.V.Sarma (ed.), Hoshiarpur
1977.
13. Kriyākramakarı̄ of Śaṅkara Vāriyar (c.1535) on Lı̄lāvatı̄ of

Bhāskarācārya II (c.1150), K.V.Sarma (ed.), Hoshiarpur 1975.
14. Sūryaprakāśa of Sūryadāsa (c.1538) on Bhāskarācārya’s Bı̄jagan.ita

(c.1150), Chs. I-V, Edited and translated by Pushpa Kumari Jain,
Vadodara 2001.
15. Buddhivilāsinı̄ of Gan.eśa Daivajña (c.1545) on Lı̄lāvatı̄ of

Bhāskarācārya II (c.1150), V.G.Apte (ed.), 2 Vols, Pune 1937.
16. T
. ı̄kā of Mallāri (c.1550) on Grahalāghava of Gan.eśa Daivajña
(c.1520), Balachandra (ed.), Varanasi 1865; Kedaradatta Joshi
(ed.), Varanasi 1981.
17. Bı̄janavāṅkurā or Bı̄japallavam of Kr.s.n.a Daivajña (c.1600) on

Bı̄jagan.ita of Bhāskarācārya II (c.1150), V.G.Apte (ed.), Pune
1930; T.V.Radha Krishna Sastri (ed.), Tanjore 1958; Biharilal Va-
sistha (ed.), Jammu 1982.
18. Śiroman. iprakāśa of Gan.eśa (c.1600) on Siddhāntaśiroman. i of

Bhāskarācārya II (c.150), Grahagan. itādhyāya, V.G.Apte (ed.), 2
Vols. Pune 1939, 1941.
19. Gūd.hārthaprakāśa of Raṅganātha (c.1603) on Sūryasiddhānta, Ji-

vananda Vidyasagara (ed.), Calcutta 1891; Reprint, Varanasi
1990.
20. Vāsanāvārttika, commentary of Nr.sim . ha Daivajña (c.1621) on

Vāsanābhās.ya of Bhāskarācārya II, on his own Siddhāntaśiroman. i
(c.1150), Muralidhara Chaturveda (ed.), Varanasi 1981.
21. Marı̄ci of Munīśvara (c.1630) on Siddhantaśiroman. i of

Bhāskarācārya (c.1150), Madhyamādhikāra, Muralidhara Jha
(ed.), Varanasi 1908; Grahagan. itādhyāya, Kedaradatta Joshi
(ed.), 2 vols. Varanasi 1964; Golādhyāya, Kedaradatta Joshi
(ed.), Delhi 1988.
22. Āśayaprakāśa of Munīśvara (c.1646) on his own

Siddhāntasārvabhauma, Gan.itādhyāya Chs. I-II, Muralid-
hara Thakura (ed.), 2 Vols, Varanasi 1932, 1935; Chs. III-IX,
Mithalal Ojha (ed.), Varanasi 1978.
23. Śes.avāsanā of Kamalākarabhat.t.a (c.1658) on his own

Siddhāntatattvaviveka, Sudhakara Dvivedi (ed.), Varanasi
1885; Reprint, Varanasi 1991.
24. Sauravāsanā of Kamalākarabhat.t.a (c.1658) on Sūryasiddhānta,

Chs. I-X, Sri Chandra Pandeya (ed.), Varanasi 1991.
25. Gan.itayuktayah., Tracts on Rationale in Mathematical Astronomy

by Various Kerala Astronomers (c.16th −19th century), K.V.Sarma
(ed.), Hoshiarpur 1979.
B Some Upapatti-s from Yuktibhās.ā (c.1530)
In this Appendix we shall present some of the proofs contained in the

Mathematics part of the celebrated Malayalam text Yuktibhās.ā 39 of
Jyes.t.hadeva (c.1530). This part is divided into seven chapters, of which
the last two, entitled Paridhi and Vyāsa (Circumference and Diame-
ter) and Jyānayana (Computation of Sines), contain many important
results concerning infinite series and fast convergent approximations
39
Yuktibhās.ā (in Malayalam) of Jyes.t.hadeva (c.1530); Gan.itādhyāya, Ra-
mavarma Thampuran and A.R. Akhileswara Aiyer (eds.), Trichur 1947; K. Chan-
drasekharan (ed.), Madras 1953; Edited, along with an ancient Sanskrit version
Gan.itayuktibhās.ā and English Translation, by K.V.Sarma, with Explanatory Notes
by K.Ramasubramanian, M.D.Srinivas and M.S.Sriram (in press).
240 M.D.Srinivas
for π and the trigonometric functions. In the preamble to his work,

Jyes.t.hadeva states that his work gives an exposition of the mathemat-
ics necessary for the computation of planetary motions as expounded
in Tantrasaṅgraha of Nīlakan.t.ha (c.1500). The proofs given in Yuk-
tibhās.ā have been reproduced (mostly in the form of Sanskrit verses
or kārikās) by Śaṅkara Vāriyar in his commentaries Yuktidı̄pikā 40 on
Tantrasaṅgraha and Kriyākramakarı̄ 41 on Lı̄lāvatı̄. Since the later work
is considered to be written around 1535 A.D., the time of composition
of Yuktibhās.ā may reasonably be placed around 1530 A.D.
In what follows, we shall present a brief outline of some of the mathe-
matical topics and proofs given in Chapters VI and VII of Yuktibhās.ā,
following closely the order which they appear in the text.
B.1 Chapter VI : Paridhi and Vyāsa (Circumferene and

Diameter)
The chapter starts with a proof of bhujā-kot.i-karn.a-nyāya (the so called

Pythagoras theorem), which has also been proved earlier in the first
chapter of the work.42 It is then followed by a discussion of how to
arrive at successive approximations to the circumference of a circle by
giving a systematic procedure for computing successively the perimeters
of circumscribing square, octagon, regular polygon of sides 16, 32, and
so on. The treatment of infinite series expansions is taken up thereafter.
B.1.1 To obtain the circumference without calculating

square-roots
Consider a quadrant of the circle, inscribed in a square and divide a side

of the square, which is tangent to the circle, into a large number of equal
40
Yuktidı̄pikā of Sankara Variyar (c.1530) on Tantrasaṅgraha of Nīlakan.t.ha Soma-
sutvan (c.1500), K.V.Sarma (ed.), Hoshiarpur 1977. At the end of each chapter of
this work, Śaṅkara states that he is only presenting the material which has been well
expounded by the great dvija of the Parakrod.ha house, Jyes.t.hadeva.
41
Kriyākramakarı̄ of Śaṅkara Vāriyar (c.1535) on Lı̄lāvatı̄ of Bhāskarācārya II
(c.1150), K.V.Sarma (ed.), Hoshiarpur 1975.
42
In fact, according to Yuktibhās.ā , almost all mathematical computations are
pervaded (vyāpta) by the trairāśika-nyāya (the rule of proportion as exemplified for
instance in the case of similar triangles)and the bhujā-kot.i-karn.a-nyāya.
parts. The more the number of divisions the better is the approximation
to the circumference.
0000
1111
0000
1111
0000
1111
0000
1111
di
0000
1111
11111
00000
00000
11111
00000
11111
00000
11111
b i
k i+1
k
i
R
C/8 (one eighth of the circumference) is approximated by the sum of the

jyārdhās (half-chords) bi of the arc-bits to which the circle is divided by
the karn.as (hypotenuses) which join the points which the divide tangent
to the centre of the circle. Let ki be the length of the ith karn.a. Then,

R R R R R R2
bi = di = =
ki ki n ki+1 n ki ki+1
Hence
n−1 n−1 2
π C 1 R2 1 R2
= = ≈
4 8R n ki ki+1 n ki
i=0 i=0
n−1
1 R2
= 2
n R
i=0
R2 + i2
n
Series expansion of each term in the RHS is obtained by iterating the
relation

a a a b − c
= − ,
b c b c
which leads to

a a a b − c a b−c 2
= − + + ....
b c b c c c
242 M.D.Srinivas
This (binomial) expansion is also justified later by showing how the

partial sums in the following series converge to the result.

100 100 100 10 − 8 100 10 − 8 2
= − + − ....
10 8 10 8 8 8
Thus
3
n 5
n
π 1 2 1
= 1− i + i4 − . . .
4 n n
i=1 i=1
When n becomes very large, this leads to the series given in the rule of
Mādhava vyāse vāridhinihate . . . 43
C π 1 1
= = 1 − + − ...
4D 4 3 5
B.1.2 Samaghāta-saṅkalita – Sums of powers of natural numbers
In the above derivation, the following estimate was employed for the
samaghāta-saṅkalita of order k, for large n :
nk+1
Sn(k) = 1k + 2k + 3k + . . . + nk ≈
(k + 1)
This is proved first for the case of mūla-saṅkalita
Sn(1) = 1 + 2 + 3 + . . . + n
= [n − (n − 1)] + [n − (n − 2)] + · · · + n
(1)
= n2 − Sn−1
Hence for large n,
n2
Sn(1) ≈
2
Then, for the varga-saṅkalita and the ghana-saṅkalita, the following es-
timates are proved for large n:
n3
Sn(2) = 12 + 22 + 32 + . . . + n2 ≈
3
n4
Sn(3) = 13 + 23 + 33 + . . . + n3 ≈
4
43
This result is attributed to Mādhava by Śaṅkara Vāriyar in Kriyākramakarı̄, cited
earlier, p.379; see also Yuktidı̄pikā , cited earlier, p.101.
It is then observed that, in each case, the derivation above is based on

the result
(k−1) (k−1) (k−1)
nSn(k−1) − Sn(k) = Sn−1 + Sn−2 + . . . + S1
It is observed that the right hand side of the above equation is a re-
peated sum of the lower order (k − 1) saṅkalita. Now if we have already
(k−1) k
estimated this lower order saṅkalita, Sn ≈ nk , then
(n − 1)k (n − 2)k (n − 3)k

nSn(k−1) − Sn(k) ≈ + + + ...
k k k
1
≈ Sn(k) .
k
Hence, for the general samaghāta-saṅkalita, we obtain the estimate
nk+1
Sn(k) ≈ .
(k + 1)
B.1.3 Vāra-saṅkalita – Repeated summations
The vāra-saṅkalita or saṅkalita-saṅkalita or repeated sums, are defined

as follows:
Vn(1) = Sn(1) = 1 + 2 + . . . + (n − 1) + n
(r−1) (r−1)
Vn(r) = V1 + V2 + . . . + Vn(r−1)
It is shown that, for large n

nr+1
Vn(r) ≈ .
(r + 1)!
B.1.4 Cāpı̄karan. a – Determination of the arc
This can be done by the series given by the rule44 is..tajyātrijyayorghātāt

. . . which is derived in the same way as the above series for C8 .

sin θ R sin θ 3 R sin θ 5
Rθ = R − + − ...
cos θ 3 cos θ 5 cos θ
44
See for instance, Kriyākramakarı̄, cited earlier, p.95-96.
244 M.D.Srinivas
sin θ
It is noted that | cos θ | ≤ 1, is a necessary condition for the terms in the
above series to progressively lead to the result. Using the above, for
θ = π6 , the following series is obtained:
√
1 1 1
C= 12D2 1 − + 2 − 3 + ... .
(3.3) (3 .5) (3 .7)
B.1.5 . skāra – Correction term to obtain accurate

Antya-sam
circumference
Let us set
C 1 1 1 1
= 1 − + − . . . + (−1)n−1 + (−1)n .
4D 3 5 (2n − 1) an
. skāra-hāraka (correction divisor), an will be accurate if

Then the sam
1 1 1
+ = .
an an+1 2n + 1
This leads to the successive approximations:45
π 1 1 1 1
≈ 1− + − . . . + (−1)n−1 + (−1)n ,
4 3 5 (2n − 1) 4n
π 1 1 1 1
≈ 1 − + − . . . + (−1)n−1 + (−1)n ,
4 3 5 (2n − 1) 4
4n +
4n
1 1 n−1 1 n n
= 1 − + − . . . + (−1) + (−1) .
3 5 (2n − 1) (4n2 + 1)
Later at the end of the chapter, the rule46 ante samasaṅkhyādalavargah.

. . . , is cited as the sūks.matara-sam
. skāra, a much more accurate correc-
tion:
π 1 1 (−1)n−1 (−1)n (n2 + 1)

≈ 1− + − ... + + ,
4 3 5 (2n − 1) (4n3 + 5n)
45
These are attributed to Mādhava in Kriyākramakarı̄, cited earlier, p.279; also
cited in Yuktidı̄pikā , cited earlier, p.101.
46
Kriyākramakarı̄, cited earlier, p.390; Yuktidı̄pikā , cited earlier, p.103.
B.1.6 Transformation of series
The above correction terms can be used to transform the series for the
circumference as follows:

C π 1 1 1 1 1 1 1
= = 1− − − − + − − ...
4D 4 a1 3 a1 a2 5 a2 a3
It is shown that, using the second order correction terms, we obtain the
following series given by the rule47 samapañcāhatayoh. . . .
C 1 1 1
= 5 − + − ...
16D (1 + 4.1) (35 + 4.3) (55 + 4.5)
It is also noted that by using merely the lowest order correction terms,
we obtain the following series given by the rule48 vyāsād vāridhinihatāt
...
C 3 1 1 1
= + 3 − + − ...
4D 4 (3 − 3) (53 − 5) (73 − 7)
B.1.7 Other series expansions
It is further noted, by using non-optimal correction divisors in the above

transformed series, one can also obtain the following results given in the
rules49 dvyādiyujām
. vā kr.tayo . . . and dvyādeścaturādervā . . .
C 1 1 1 1
= + 2 − 2 + 2 − ...
4D 2 (2 − 1) (4 − 1) (6 − 1)
C 1 1 1 1
= − − − − ...
8D 2 (42 − 1) (82 − 1) (122 − 1)
C 1 1 1
= 2
+ 2 + 2
− ...
8D (2 − 1) (6 − 1) (10 − 1)
47
48
49
246 M.D.Srinivas
B.2 Chapter VII : Jyānayanam – Computation of Sines
B.2.1 Jyā, koti and śara – R sin x, R cos x and R(1 − cos x)
First is discussed the construction of an inscribed regular hexagon with

side equal to the radius, which gives the value of R sin( π6 ). Then are
derived the relations:
π
R sin −x = R cos x = R(1 − versin x)
2
x 1 1
R sin = [(R sin x)2 + (R versin x)2 ] 2 .
2 2
Using the above relations several Rsines can be calculated starting from
the following:
π R
R sin =
6 2
π 2 12
R
R sin = .
2 2
This is one way of determining the pat.hita-jyā (enunciated or tabulated

sine values), when a quadrant of a circle is divided into 24 equal parts
of 3◦ 45 = 225 each. To find the Rsines of intermediate values, a first
approximation is
R sin(x + h) ≈ R sin x + h R cos x.
Then is derived the following better approximation as given in the rule50

ist.adoh.kot.idhanus.oh. . . . :

2 1
R sin(x + h) ≈ R sin x + R cos x − R sin x
∆ ∆

2 1
R cos(x + h) ≈ R cos x + R sin x − R cos x ,
∆ ∆
2R
where ∆ = h .
50
Tantrasaṅgraha , 2.10-14.
B.2.2 Accurate determination of sines
Given an arc s = Rx, divide it into n equal parts and let the pin.da-jyās
Bj , and śaras Sj− 1 , with j = 0, 1, . . . n, be given by
2

jx
Bj = R sin ,
n

(j − 12 )x
Sj− 1 = R vers
2 n
If α be the samasta-jyā (total chord) of the arc ns , then the second order
sine difference (jyā-khan.d.āntara) is shown to satisfy
α
(Bj+1 − Bj ) − (Bj − Bj−1 ) = (Sj− 1 − Sj+ 1 )
r 2 2
α 2
= Bj ,
r
for j = 1, 2, . . . n. From this are derived the relations
α
Sn− 1 − S 1 = (B1 + B2 + . . . + Bn−1 ),
2 2 r
α 2
Bn − nB1 = − [B1 + (B1 + B2 ) + . . . + (B1 + . . . + Bn−1 )]
αR

= − S 1 + S 3 + · · · + Sn− 1 − nS 1
r 2 2 2 2
If B and S are the jyā and śara of the arc s, then it is noted that, in
the limit of very large n, we have as a first approximation
js s
Bn ≈ B, Bj ≈ , Sn− 1 ≈ S, S 1 ≈ 0 and α ≈ .
n 2 2 n
Hence

1 s 2 s2
S ≈ [1 + 2 + . . . + n − 1] ≈ .
R n 2R
and
2
1 s 3
B ≈ s− [1 + (1 + 2) + . . . + (1 + 2 + . . . + n − 1))]
R n
s3
≈ s− .
6R2
248 M.D.Srinivas
Iterating these results we get successive approximations for the difference

between the Rsine and the arc (jyā-cāpāntara), leading to the following
series given by the rule51 nihatya cāpavargen.a . . . :
 s 3 s 5 
s s
 
R sin = B = R − R + R − . . .
R R 3! 5!
 s 4 s 6 
s s 2
 R R 
R − R cos = S = R − + − . . .
R R 4! 6!
While carrying successive approximations, the following result for vāra-

saṅkalitas (repeated summations) is used:
n
j(j + 1) . . . (j + k − 1) n(n + 1)(n + 2) . . . (n + k)
=
k! (k + 1)!
j=1
nk+1
≈ .
(k + 1)!
Then is obtained a series for the square of sine, as given by the rule52
nihatya cāpavargen. a . . .
x4 x6
sin2 x = x2 − + − ...
2 2 3
22 − 22 − 32 −
2 2 2
Chapter VII of Yuktibhās.ā goes on to discuss different ways of deriving
the jı̄ve-paraspara-nyāya53 , which is followed by a detailed discussion of
the cyclic quadrilateral. The chapter concludes with a derivation of the
surface area and volume of a sphere.
Acknowledgement
The author is deeply indebted to Gérard Emch, R.Sridharan and
M.S.Sriram for their valuable advice and encouragement. He is much
obliged to K.Ramasubramanian for his enthusiastic help in the prepa-
ration of this article.
51
Yuktidı̄pikā , cited earlier, p.118
52
Yuktidı̄pikā , cited earlier, p.119.
53
The relation between the sine and cosine of the sum or difference of two arcs with
the sines and cosines of the arcs.
Ramanujan and Partial Fractions
George E. Andrews
The Pennsylvania State University
University Park, PA 16802 USA
andrews@.psu.edu
Abstract
The subject of partial fractions is usually conned to the stan-
dard calculus course and is viewed as a useful albeit mundane
tool. This paper looks at partial fractions starting with Euler. We
then consider some of the very surprising and appealing discoveries
made by Ramanujan.
1 Introduction
The idea of partial fractions must surely have arisen after calculus and
basic analysis had a solid footing. We can certainly nd carefully pre-
pared examples in Euler's Introduction in Analysis Innitorum, Volume
1 [3, Chapter 2].
For example, [3, pp. 34{35] Euler proves that
1 1 1 2 1 7 1
= 3+ 2+ + + ; (1.1)
z (1
3 z ) (1 + z )
2 z z z 2(1 z ) 4(1 z ) 4(1 + z )
2
or [3, pp. 31{32]
z2 z 1 1 1 1
= + : (1.2)
(1 z ) (1 + z )
3 2
4(1 + z ) 4(z
2
1) 2(z 1)
2
2(z 1)3
Ramanujan was not the rst to observe that there are appealing and
sophisticated extensions of partial fractions. For example, there is the
251
252 George E. Andrews
classical partial fraction expansion for the reciprocal of Jacobi's theta

product [6, p.136]
1 ( 1)n qn(n+1)=2
X 1
Y (1 q n )2
= (1.3)
n= 1 1 zq n n=1
(1 zq n 1 )(1 q n =z )
P1 n n(n+1)=2
=0 ( 1) (2n + 1)q
nP
= 1 ( 1)n z n qn(n 1) ; (1.4)
n= 1
where jqj < 1 and z 6= q N for any integer N . Actually we nd (1.3) in
Ramanujan's Lost Notebook [5, p.1].
In the following pages, we shall examine some of Ramanujan's ideas.
Our object will be to choose formulas that illustrate how one can proceed
from rst principles to obtain some quite surprising formulas.
2 A Partial Fraction
In the following, we shall use the standard notation [4, p.3, eq. (1.2.15)
and p. 20, Ex. 1.2]. We shall always assume jqj < 1.
(A)n = (A; q)n = (1 A)(1 Aq ) (1 Aq n 1
);
1
Y
(A)1 = (A; q)1 = (1 Aq n );
n=0
(
N
( q ;q )
05M 5N
N
= M
(q ;q ) (q ;q ) N M ;
M 0 ; M < 0 or M > N:
Our rst result is the following:

Theorem 2.1.
N
X 2
N (q)n qn
n=0
n (z )n+1 (q=z )n

PN N n q 23 n2 + n2 (q)n
= N+ zqn qn : (2.1)
1 ( 1) 1 1
(1 z )(q) n=1 n (q )N +n 1 z
Proof. Both sides of (2.1) are rational functions of z with 2N + 1 sim-

ple poles at z = q N ; q N +1 ; : : : ; q 1 ; 1; q; : : : ; qN . Let us calculate the
residues of each side at these simple poles.
Ramanujan and Partial Fractions 253
If z = qm for 1 5 m 5 N , then the residue on the right-hand side is

3 2 m
N ( 1)m 1 q 2 m + 2 (q)m
:
m (q)N +m
On the left-hand side

N
X 2
N (q)n qn
lim (z qm )
z !qm n (z )n+1 (q=z )n
n=0
N
X 2
N (q)n qn +m
=
n=m
n (qm ; q)n+1 (q1 m ; q)m 1 (q)n m
XN 2
N (q)n qn +m(m+1)=2 ( 1)m 1
=
n=m
n (q)n+m (q)n m
N
X 2
( 1)m 1 qn +m(m+1)=2
= (q)N
n=m
(q)n+m (q)N n (q)n m
NXm 2
( 1)m 1 q(n+m) +m(m+1)=2
= (q)N
n=0
(q)n+2m (q)N n m (q)n
N m
(q)N ( 1)m 1 X (q N +m )n ( 1)n qn(n+1)=2+n(N +m)+m(3m+1)=2
=
(q)2m (q)N m n=0 (q2m+1 )n (q)n

(q)N ( 1)m 1 qm(3m+1)=2 q N +m ; 1 ; q; q N +m+1
= lim 2 1
(q)2m (q)N m !0 q 2m+1
(in the notation of [4, p.3, eq. (1.2.14)])
(q)N ( 1)m 1 qm(3m+1)=2
=
(q)2m (q)N m (q2m+1 )N m
(by the q-Chu-Vandermonde summation [4, p.11, eq. (1.5.2)])

N ( 1)m 1 q m(3m+1)=2 (q )m
=
m (q)N +m
Next we consider z = q m for 0 5 m 5 N . In this case, the residue

on the right-hand side is

N ( 1)m 1 qm(3m 1)=2 (q)m
:
m (q)N +m

N
X 2
N (q)n qn
limm(z q m)
z !q n (z )n+1 (q=z )n
n=0
N
X 2
N (q)n qn m
=
n=m
n (q m )m (q)n m (qm+1 )n
X
N
N
2
(q)n qn +m(m 1)=2 ( 1)m 1
=
n=m
n (q)m (q)n m (qm+1 )n
N
X 2
( 1)m 1 qn +m(m 1)=2
= (q)N
n=m
(q)n+m (q)N n (q)n m
NXm 2
( 1)m 1 q(n+m) +m(m 1)=2
= ( q )N
n=0
(q)n+2m (q)N n m (q)n
NXm
(q)N ( 1)m 1 qm(3m 1)=2
(q N +m )n ( 1)n qn(n+1)=2+n(N +m)
=
(q)2m (q)N m n=0
(q2m+1 )n (q)n

N ( 1)m 1
q m(3m 1)=2
(q)m
= ;
m (q)N +m
as before.
Hence by the classical theorem for the representation of a proper
rational function with simple poles by a partial fraction decomposition
[2, pp.56{57], we have established Theorem 2.1.
Note that the proof of Theorem 2.1 required nothing more sophisti-
cated than the very elementary q-Chu-Vandermonde summation.
Theorem 2.1 has some very appealing corollaries. While it does
not appear in Ramanujan's Lost Notebook, many instances of it do.
It should be noted that this result was rst explicitly stated by G.N.
Watson [7, p.64].
Corollary 2.1.
1
X qn
2
1 X ( 1)n qn(3n+1)=2
1
= : (2.2)
n=0
(z )n+1 (q=z )n (q)1 n= 1 1 zqn
Proof. Let N ! 1 in Theorem 2.1. The quadratic exponent on q easily

allows one to justify this limiting process, and a little algebra transforms
the result into the above expression.
From Corollary 2.1, G.N. Watson [7, p.64] and presumably Ra-
manujan deduced three important formulas for Ramanujan's third order
mock-theta functions:
1
X qn
2
f (q ) = ;
n=0
( q ; q ) 2
n
X1 qn
2
(q ) =
n=0
( q 2 ; q 2 )n
X1 qn2 ( q; q)
n
(q ) =
n=0
( q 3 ; q3 )
n
Corollary 2.2. [7, p. 64]
1 !
1 X ( 1)n qn(3n+1)=2
f (q ) = 1+4 ; (2.3)
(q) 1 n=1
1 + qn
1 !
1 X ( 1)n (1 + qnqn(3n+1)=2
(q ) = 1+2 ; (2.4)
(q)1 n=1
1 + q2n
1 !
1 X ( 1)n (1 + qn )qn(3n+1)=2
(q ) = 1+ (2.5)
(q) 1 n=1
1 q n + q 2n
Proof. To obtain (2.3) set z = 1 in (2.2) and simplify. To obtain (2.4)

set z = i in (2.2) and simplify. Finally to obtain (2.5) set z = ei=3 .
3 Another Partial Fraction

In this section we present a second result. This result is closely related
to the one in the previous section as Watson has shown [7, p. 63{66]
in that each is deducible from the same master formula. We treat this
theorem separately because we want to deduce these results from very
elementary identities.
Theorem 3.1.
N
X 2
N (q)n qn +n
n 1 1
n=0 (zq 2 )n+1 (q 2 =z )n+1
N 1 !
X N ( 1)n q3n(n+1)=2 (q)n 1 q n+ 2
= 1 + 1 (3.1)
:
n=0
n (q)n+N +1 1 zq n+ 2 z q n+ 2
Proof. We proceed exactly as we did with Theorem 2.1. Both sides

of (3.1) are rational functions of z with 2N + 2 simple poles at z =
1 1 1 1 1 1
q N 2 ; q N + 2 ; : : : ; q 2 , q 2 ; : : : ; q N 2 ; q N + 2 . We now calculate the resi-
dues of each side at these simple poles. Furthermore it is easy to check
that each side is symmetric in z and 1z ; hence we need only check the
residues at the positive power of q.
1
If z = qm+ 2 for 0 5 m 5 N , then the residue on the right-hand side
is clearly
1
N ( 1)m q 3m(m+1)=2+m+ 2 (q )m
:
m (q)m+N +1

N
X 2
1 N (q)n qn +n
lim 1 (z q m+ 2 ) 1 1
z !qm+ 2
n (zq 2 )n+1 (q 2 =z )n+1
n=0
X N 2 1
N (q)n qn +n+m+ 2
=
n=m
n (qm+1 )n+1 (q m )m (q)n m
2N
X
1 q n +n
= ( 1)m qm(m+1)=2+m+ 2 (q)N
n=m
(q)n+m+1 (q)N n (q)n m
NXm 2
1 q (n+m) +n+m
= ( 1)m qm(m+1)=2+m+ 2 (q)N
n=0
(q)n+2m+1 (q)N n m (q)n
1 2
( 1)m qm(m+1)=2+m+ 2 +m +m (q)N
=
(q)2m+1 (q)N m
NXm
(q N +m )n ( 1)n qn(n+1)=2+n(N +m+1)
(q)n (q2m+2 )n
n=0
3 2 5m 1
( 1)m q 2 m + 2 + 2 (q)N q N +m ; q ; q; q N +m+1
= lim 2 1
(q)2m+1 (q)N m !0 (q)2m+2
3 2 5m 1
( 1)m q 2 m + 2 + 2 (q)N
=
(q)2m+1 (q)N m (q2m+2 )N m
(by the q-Chu-Vandermonde summation [4, p. 11, eq. (1.5.2)])
1
N ( 1)m q 3m(m+1)=2+m+ 2 (q )m
=
m (q)m+N +1
As with Theorem 2.1, our theorem follows from the standard partial
representation of a proper rational function with simple poles [2, pp.
56{57].
Corollary 3.1. [7, p.66]
X 2
q n +n
1
1 X ( 1)n q3n(n+1)=2
1 1 = : (3.2)
n=0 (zq 2 )n+1 (q 2 =z )n+1
(q)1 n= 1 1 zqn+ 21
Proof. Let N ! 1 in Theorem 4.1. As in the proof of Corollary 2.2, the

quadratic exponents on q easily allow one to justify this limit. Algebraic
simplication then yields the assertion in Corollary 3.1.
From Corollary 3.1, we can easily deduce (as did Watson [7, p. 66]
and probably Ramanujan) three more important formulas for three more
of the third order mock-theta functions:
1 q2n(n+1)
X
! (q ) = ;
n=0
(q; q2 )2n+1
1
X q n(n+1)
(q ) = ;
n=0
( q; q2 )n+1
1 (q; q2 ) q2n(n+1)
X n
(q ) = :
n=0
(q3 ; q6 )n+1
Corollary 3.2. [7, p.66]
1
1
X 1 + q2n+1
! (q ) = ( 1)n q3n(n+1) ; (3.3)
(q2 ; q2 )1 n=0 1 q2n+1
1 X
1 (1 q2n+1 )
(q ) = ( 1)n q3n(n+1)=2 ; (3.4)
(q)1 n=0 (1 + q2n+1 )
1
1
X (1 q4n+2 )
(q ) = ( 1)n q3n(n+1) : (3.5)
q 2 ; q 2 )1 n=0 1 + q2n+1 + q4n+2
Proof. To obtain (3.3) replace q by q2 then set z = 1 in (3.2), and

simplify. To obtain (3.4), set z = i in (3.2) and simplify. To obtain (3.5)
replace q by q2 , then set z = e2i=3 in (3.2) and simplify.
4 The Simplest Partial Fraction

In the last two sections. we proved two partial fraction decompositions
relying only on the q-Chu-Vandermonde summation. In this section we
provide a proof of (1.3) that does not even require that much.
Theorem 4.1.
(q )N
(z )N +1 (q=z )N
X N
1 N (q)n ( 1)n qn(n+1)=2 1 1
= +
(1 z )(q )N n=1 n (q)n+N 1 zq n z qn
(4.1)
Proof. Both sides of (4.1) are rational functions of z with 2N + 1 simple

poles at z = q N ; q N +1 ; : : : ; q 1 ; 1; q; : : : ; qN .
First we calculate the residue at z = qm ; 1 5 m 5 N. The residue
on the right-hand side is clearly

N (q)m qm(m+1)=2
( 1)m 1
:
m (q)m+N
On the left-hand side, the residue is

(z qm )(q)N q m (q )N
limm =
z !q (z )N +1 (q=z )N (qm )N +1 (q1 m )m 1 (q)N m

( 1)m 1 qm(m+1)=2 (q)N m N (q )m q m(m+1)=2
= = ( 1) 1
:
(q)N +m (q)N m m (q)m+N
Next we consider z = q m , 0 5 m 5 N. The residue on the right-

hand side in this instance is

N (q)m ( 1)m qm(m+1)=2 N (q)m qm(m 1)=2
q m = ( 1)m 1
;
m (q)m+n m (q)m+N
and on the left-hand side, the residue is
(z q m )(q)N q m ( q )N
limm =
z !q (z )N +1 (q=z )N (q m )m (q)N m (qm+1 )N
m 1 m(m 1)=2

( 1) q (q)N N (q)m qm(m 1)=2
= = ( 1)m 1
:
(q)N +m (q)N m m (q)N +m
As before, our result now follows from the standard theorem on

partial fraction decompositions [2, pp.56{57]
Corollary 4.1. Identity (1.3) is valid.
Proof. The same argument used for Corollaries 2.1 and 3.1 holds here
as well.
5 Conclusion
There are many more results beyond those considered in this paper that
were included in Ramanujan's Lost Notebook. Bruce Berndt and I [1]
are publishing a full treatment of the results from the Lost Notebook,
and we present in Chapter 12 of that work a full account of Ramanu-
jan's partial fraction theorems. The proofs there are more succinct, but
rely on more sophisticated background. The object in this paper was to
illustrate the fact that one can deduce some of Ramanujan's most sur-
prising partial fraction formulas using nothing deeper than the q-analog
of the Chu-Vandermonde summation.
References
[1] G.E. Andrews and B. Berndt, Ramanujan's Lost Notebook, Part I,
Springer, Berlin (to appear).
[2] P. Dienes, The Taylor Series, Dover, New York, 1957.
[3] L. Euler, Introductio in Analysin Innitorum, Vol. 1, Lausanne,
1748.
[4] G. Gasper and M. Rahman, Basic Hypergeometric Series, Encycl.
Math and Its Appl., G.-C. Rota, ed., Vol. 35, Cambridge University
Press, Cambridge, 1990.
[5] S. Ramanujan, The Lost Notebook and Other Unpublished Papers,
Intro by G. E. Andrews, Narosa, New Delhi, 1987.
ements de la Theorie des fonctions Ellip-
[6] J. Tannery and J. Molk, El
tiques, Vol. III, Gauthier-Villars, Paris, 1896 [reprinted: Chelsea,
New York, 1972].
[7] G.N. Watson, The nal problem: an account of the mock theta
functions, J. London Math. Soc., 11(1936), 55{80.
Contributions of Indian
Mathematicians to Quantum Statistics
Gerard G. Emch
University of Florida, Gainesville FL 32611, USA
gge@math.u .edu
Abstract
While modern Algebra and Number Theory have well docu-
mented and established roots deep into India's ancient scholarly
history, the understanding of the springing up of statistics, specif-
ically quantum statistics, demands a closer inquiry. My pro ject is
two-fold. Firstly, I explore and delineate the cultural and educa-
tional circumstances that presided over the inception of the very
concept that quantum theory required its own dedicated statistical
analysis. My quest therefore is anchored in a brief review of the
pioneering contributions of personalities as diverse as those of Bose
and Chandrasekhar, or Raman and Krishnan, and Mahalanobis.
Secondly, I examine how the intellectual climate and some of the
local mathematical traditions have fostered the ongoing develop-
ment of quantum probability and stochastic processes theories in
India.
Science is not an impersonal stream of dis-

coveries. It is created by human beings and
its advances are very much products of highly
personal actions and reactions of some gifted
individuals. [70, p.31]
... science is a collective endeavour ... any
single life is but a fragment in a larger fabric.
[71, p.55]
261
262 Gerard G. Emch
1 Preliminary considerations
The main questions I want to explore in this essay are How and Why an
interest in quantum statistics developed among Indian mathematicians.
Indeed, the eld of quantum statistics might appear at rst sight to be
a singularity against the background of profound traditions, the deep
roots of which are illustrated in other communications reported in these
proceedings on the History of Indian Mathematics. Upon listening to
these enlightening contributions, we could not but be vividly reminded
of the continuity between the ancient schools and modern disciplines
such as algebra, geometry, combinatorics and number theory, in which
the accomplishments of Indian mathematicians are recognized the World
over as centrally in uential.
In contrast, quantum probability, and in particular the theory of
quantum stochastic processes, is a recent eld of enquiry in which con-
siderable pioneering work was achieved in twentieth{century India. It
would be futile to attempt and give here a detailed survey that would
do justice to all facets of this eort. Rather, I will focus on those which
relate most to areas of mathematical physics with which I am familiar,
and yet bring out the particular avours of the Indian contributions and
of the intellectual climate in which they originated.
Among foreign physicists and mathematicians, the very words of
\quantum" and \statistics" might evoke developments in Indian sci-
ence of cultures imported here from the outside in the early twentieth{
century. Such a naive cosmopolitan account would however miss sev-
eral Indian idiosyncrasies, two extremes among which are: a form of
governmental establishment that encourages the in uence of strong per-
sonalities, and yet a demographic diversity that allows the independent
pursuits of individual thinkers. To illustrate this dualism, suÆce it to
mention here only one of its many aspects: the owering in the nine-
teenth century of the \Bengali Renaissance" right in Calcutta, while the
port-city was the traditional seat of British presence.
I will thus sketch in Section 2 how strong personalities developed into
pioneers who paved the way towards the modern form of quantum statis-
tics: S.N. Bose, C.V. Raman and K.S. Krishnan, S. Chandrasekhar, and
P.C. Mahalanobis. In section 3, I turn to contemporary Indian gures
who have contributed to the developments of various aspects of the the-
Indian Contributions to Quantum Statistics 263
ory of quantum probability and of quantum stochastic processes: V.S.

Varadarajan, K.R. Parthasarathy, K.B. Sinha, and M.D. Srinivas. In
the interest of conciseness, the bibliography at the end of this essay is
intended to be an integral part of my account, as the titles of the papers
I included provide in a nutshell the most essential information their au-
thors surely wanted to convey. I should also mention that in the course
of my essay I respect the contemporary usages when speaking of cities
like Calcutta [Kolkata] or Madras [Chennai].
2 The pioneers
2.1 Satyendra Nath Bose, 1894{1974
Wearing an illustrious Bengali patronym, Bose was born and died in

Calcutta, studied there, receiving his M.Sc. from Presidency College
in 1915 and continuing his studies in mathematics and physics at the
University College of the University of Calcutta, taking his Ph.D. in
1917; he rst taught there, until he went to Dacca in 1921, from which
he came back to Calcutta in 1945 to teach again there until 1956.
After he had translated and published a collection of Einstein papers
on relativity, he spent time rethinking (\in my own way" as he declared)
the published lectures of Boltzmann and of Planck on classical statistical
mechanics, as well as the the book of Gibbs, and several of the original
papers of Einstein, Bohr and Sommerfeld, i.e. the foundations of a well{
informed study of the current physical literature, namely what we call
today the \old" quantum mechanics.
Bose's own contributions to quantum statistics [10, 11] occurred dur-
ing a burst of creativity that lasted just a few years. This happened de-
spite the well{documented and enthusiastic support of Einstein who was
instrumental in the granting to Bose of a generous two{year travel fel-
lowship that brought him into contact with Langevin and Marie Curie
in Paris, and with Einstein and the Berlin establishment. After this,
Bose returned to teaching and some research, rst in Dacca and then
again in Calcutta.
The reason for Einstein's enthusiasm must have been the authentic
simplicity of Bose's derivation of the Planck black{body radiation law
264 Gerard G. Emch
which is best understood when written in the form:

(; T ) = n( ) u(; T )
with
( 2.1)
8 2 h
n( ) = and u(; T ) =
c3 exp(h=kT ) 1
where is the energy density distribution of the radiation, as a function
of the frequency and the temperature T at which the radiation is in
thermal equilibrium. Earlier, the rst factor, n( ) ; had been derived
from a classical notion, the number N ( ) = V n( ) of electromagnetic
modes of frequency ; V is the volume of the cavity in which the radia-
tion is trapped. This number obtains as a straightforward adaptation of
an argument by Rayleigh (1877{78) on standing sound waves, the only
two changes required here being to substitute the speed of light c for the
speed of sound; and to take into account that electromagnetic waves are
transversal waves, thus with two polarizations at the same frequency,
while sound waves in air are compression waves, thus with only one
mode for each frequency. The second term, u(; T ) ; had appeared as a
\lucky" manifestation of the genius of Planck who, in 1900, interpolated
between previously known formulas valid separately in low and high fre-
quency regimes. In the intervening years, Einstein had proposed that the
radiation consists of indivisible light quanta of energy h ; an interpre-
tation he supported by his concise and consistent explanation of all the
known features of the photo-electric eect, features which the Maxwell
theory of electromagnetism had thoroughly failed to account for. The
photon hypothesis enabled Debye and Einstein to account for u(; T )
and Bose was willing to accept their quantum counting argument. It
was on the rst term that he concentrated his critique; we read from
his famous June 4, 1924 letter to Einstein, that \I have tried to deduce
the coeÆcient 8 2 =c3 in Planck's law independent[ly] of the classical
electrodynamics, only assuming that the ultimate elementary region in
the phase space has the content h3 :" In postulating the irreducibility of
any such partition, Bose was striving to restore the consistency of the
premises; both terms in the Planck radiation formula were to have a
purely quantum origin: now, the Bose{Einstein statistics had found a
proper logical frame.
Einstein was suÆciently impressed by Bose's manuscripts to trans-
late them into German and arrange for their publication. Moreover, he
assimilated Bose's two papers thoroughly enough to realize that the new
statistics, if extended from radiation to material particles would imply
that a quantum ideal gas [23] would condensate in a super uid phase if
the temperature were lowered deeply enough: a phenomena that has no
classical analogue.
Although Einstein was not fully satised, Bose was not able to re-
spond to Einstein's objections, even after meeting personally with him in
Berlin. How Dirac was able to perform the necessary synthesis through
his symmetrization prescription has been described in [66] and [41,
pp.424{425]; incidentally, the word \Boson" to refer to particles that
obey the Bose{Einstein statistics was coined by Dirac.
As long as liquid Helium remained the only available laboratory
manifestation of the Bose{Einstein condensation, a hiatus remained
between theory and experiment, since the intermolecular interactions
surely present in super uid Helium were not accounted for in the orig-
inal model of an ideal quantum gas. The situation changed drastically
in the 1980s and 1990s with the observation of Bose{Einstein conden-
sation in dilute atomic gases at temperatures in the milli{ and then
micro{Kelvin ranges, thus projecting again Bose's name to the frontier
of contemporary physics.
2.2 Chandrasekhara Venkata Raman, 1888{1970 and

Kariamanikkam Srinivas Krishnan, 1898{1961
Raman's 1930 Nobel Prize serves here as a historical marker to witness

the development of modern physics in India at the time when the new
quantum statistics was emerging in the midst of signicant developments
in the classical theory of stochastic processes. The Nobel citation reads
for his work on the scattering of light and for the discovery of the eect
named after him. It was only in February 1928 [52] that the discovery
was made by Raman's research team at the Indian Association for the
Cultivation of Science; he was Professor of Physics at the University of
Calcutta since 1917 until 1933 when he joined the Indian Institute of
Science in Bangalore; there, in 1948, he became director of the Research
Institute that bears his name. His formal education had ended in 1907
with a M.A. from the University of Madras; he had obtained a B.A.
from Madras Presidency College in 1904, at the age of 16.
266 Gerard G. Emch
The Raman eect belongs to molecular spectroscopy, and it helps

determine the structure of molecules. Specically, when a spectro-
scope analyses the light emitted from a transparent medium on which
a monochromatic light beam of frequency o is directed, one observes
symmetrically on each side of the original spectral line a relatively faint
secondary spectrum
f = o m g ; ( 2.2)
now called the Raman spectrum. The observed frequency shifts m and
the relative intensities of the secondary lines { both of which are char-
acteristic of the material target rather than of the incident light beam {
allow us to interpret this spectrum as being due to the inelastic collisions
between the incident light quanta and the molecules, involving the dis-
crete energy levels hm associated by quantum theory to the vibrational
and/or rotational degrees of freedom of the molecules of the medium
{ primarily liquids, but also solids and even gases. The fact that the
so-called \Stokes" lines (those on the low-frequency side of the exciting
line) are more intense than the corresponding \anti{Stokes" lines (on
the high-frequency side) in the ratio
Ia =Is = exp( hm =kT ) ( 2.3)
is taken as a further conrmation of the quantum nature of the Raman
eect, so much so that to the physicist R.W. Wood is attributed the
opinion that \Raman's long and patient study of the phenomenon of light
scattering is one of the most convincing proofs of the quantum theory"
[74, p.208], thus suggesting it was on par with Einstein's explanation
of the photo-electric eect. As R.W. Wood may not be so well known
in the mathematical community, let it be noted that Niels Bohr, who
had recommended Raman for the 1930 Nobel Prize for Physics, not only
had also included Wood's name as a co-nominee, but had done so even
earlier for the 1929 Prize [54].
While the eect had been predicted before by Smekal (1923), its ac-
tual discovery demonstrated a degree of sophistication, both theoretical
and experimental, that was immediately recognized { in particular by
the Nobel committee { as a crowning achievement of what we call now
the \old" quantum physics epitomized by Bohr. All along, Raman con-
tinued to be very much involved in more general, albeit classical, physi-
cal aspects of oscillatory phenomena. Indeed two examples ought to be
mentioned here; the rst is that along with his better known activities
in optics, Raman maintained a creative interest in acoustics; see for in-
stance [51] where Raman describes the physics of musical instruments ...
including some Indian instruments such as the sitar and the tabla. The
second example of Raman's activity is closer to the subject of this es-
say: after having been struck by the deep blueness of the Mediterranean
sea, he questioned Rayleigh's theory on the blueness of the sky, and he
applied to the diraction of light the Einstein{Smolukowski stochastic
theory of uctuations and the attendant correlations [50], which was
extended by his collaborators; amongst them was K.S. Krishnan, who
belongs to this essay on several accounts. He performed some of the
crucial experiments leading to the unambiguous detection and thus the
very discovery of the Raman eect; and later, in the early 1940s, when as
professor of physics at Allahabad, Krishnan kindled Harish{Chandra's
rst scientic direction of research { theoretical physics { which Harish{
Chandra pursued in Bangalore with Bhabha [5], and then in Cambridge
under Dirac [71, p.56].
Krishnan's talents as an experimental physicist rst came to light in
Raman's IACS laboratory; recall [52] and see [54]. The cultivation of
these exceptionally acute abilities later made him famous as a pioneer
in the investigation of quite an array of diverse phenomena in condensed
matter physics { cf. e.g. [4] { so much so that the present address of
the National Physics Laboratory, of which he was the rst director, is
on K.S. Krishnan Road in Delhi. His early association with Raman's
research programme on \light scattered in a homogeneous medium" is
explicitly acknowledged in the opening paragraphs of both [35, 36], his
only overt contribution in and to mathematics. This belongs here for
three reasons: rst, it stems from an analysis of a physically important
stochastic process; second, it gives a mathematically elegant proof for a
special version of a theorem now known to all probability theorists; and
third, it provides my story with a link to Ramanujan, indispensable to
any paper in the history of modern Indian mathematics.
First, the theorem. Krishnan begins by citing [22] { in which Einstein
pursues the theory of critical opalescence proposed by Smolukowski [58]
lim
X+1 sin2 (n + )
=
Z
{ for its use of an approximation amounting to the formula:
1 sin2 x
dx 2 : ( 2.4)
2
n= 1 (n + ) x
!0 1
268 Gerard G. Emch
What Krishnan realizes is that this formula hides a deeper mathematical

truth, namely the following general result which he proceeds to prove.
f~(k) = (2) 2
1
1 Z
Let f be an even real function of a real variable x such that its Fourier
transform
f (x) exp( ikx)
1
exists and vanishes for jkj ko : Then
8 2 (0; 2=ko ] :
X+1 f (n + ) = p2f~(0) : ( 2.5)
n= 1
p
Note that: (i) sin2 x=x2 satises the assumptions, with ko = 2 ; (ii)
the validity of formula (2.5) extends over a range of values of rather
than as a limit, as (2.4) does; (iii) the assumptions are very permissive;
and (iv) (2.5) itself is a special case of the sampling theorem :
8 2 (0; 2=(ko + k)] :

X+1 f (n) exp( ink) = p2f~(k)
n= 1
( 2.6)
variously ascribed to Nyquist, Shannon or Whittaker; probabilists learn
this from [25, Thm.XIX.5.2]; and it is also recognized to be of cardinal
relevance in communication theory.
The second aspect of (2.5), which Krishnan recognizes already in [35],
is that the functions f for which the formula is valid cover a wide range;
from a wealth of examples, he is manifestly gratied in [36] to choose
and present, in the light of his own result (2.5), several of Ramanujan's
identities.
This spontaneous demonstration of such an interdisciplinary purview
appears to have been a major contributing factor in the systematic de-
velopment of the modern scientic establishment in India. Indeed sev-
eral other scientists of the time shared this broad vision. Consider thus
brie y the case of H.J. Bhabha (1909{1966), whom I mentioned already
in connection with Harish-Chandra [5]. Along with his long-lasting inter-
est in relativistic quantum theory of elementary particles [6], he belongs
to the present story also for his classical stochastic analysis of electron
cascades [7, 8], a theoretical aspect of cosmic rays physics; furthermore
{ as with Krishnan { experimental physicists know him perhaps even
better for his persistent and successful observational pursuits. Bhabha's

multidimensional drives are further evidenced by his being the founder
in 1945, and rst Director, of the Tata Institute for Fundamental Re-
search in Bombay; and by the eminent role he played on the international
political scene as president of the IAEC for the peaceful use of nuclear
power. Beyond the connes of this essay, I discern in [28] several ele-
ments that would help enlighten a discussion of the considerations (or
theses) I proposed in section 1 above; see also subsection 2.4 below.
2.3 Subrahmanyan Chandrasekhar, 1910{1995
Chandrasekhar is also an alumnus of Presidency College, Madras (B.Sc.

with honours in Physics, 1930), and then Trinity College, Cambridge
where he studied under R.H. Fowler, met Dirac, and was awarded the
Ph.D. in 1933. After Cambridge, he was associated with the University
of Chicago from 1937 to the end of his life, thus overlapping there with
the Marshall Stone era in mathematics. Although many of us know
Chandrasekhar as a most agile classical applied mathematician, he is
known primarily as an astronomer and was Editor of the Astrophysical
Journal from 1952 to 1971. He received the Nobel Prize in Physics in
1983, which brings us straight into his early involvement with quantum
statistics.
Chandrasekhar was prompted in this direction by no less a master
than Sommerfeld during an apparently dramatic private interview that
took place in the fall of 1928 in Madras, in the course of which the
\Professor of Professors" assured the eighteen year old student { so
proud of having studied on his own Maxwell{Boltzmann statistics and an
early edition of Sommerfeld's book Atomic Structure and Spectral Lines
{ that everything he knew had been superseded by the new quantum
mechanics of Schroedinger, Heisenberg, and Dirac. To drive home his
point, Sommerfeld mercifully passed to Chandrasekhar the galley{proofs
of a paper where he proposed to apply the Fermi{Dirac statistics to the
theory of electrons in metals. Chandrasekhar had thus inadvertently
landed on the hinge between the \old" quantum mechanics and the
\new", with a clear admonition that the passage from the former to the
latter was ineluctable. \Chandra would later characterize this encounter
as the `single most important event' in his scientic career." [75, p.1]
270 Gerard G. Emch
Chandrasekhar indeed understood the lesson: he soon wrote a paper

of his own [15]; and he read a 1926 paper by R.H. Fowler, which he de-
scribes later as \the fundamental discovery that the electron assembly
in the white dwarfs must be degenerate in the sense of the Fermi{Dirac
statistics" [17, fn.1,p.451]: indeed Fowler had identied the electron de-
generacy pressure as the pressure that keeps the stars from gravitational
collapse. Following up on this reading, Chandrasekhar published a string
of papers on relativistic quantum degenerate stellar congurations and
the theory of white dwarfs [16]; these early papers are part of his Nobel
prize citation.
It is in the course of this analysis that Chandrasekhar made the
discovery that white dwarfs have a maximum mass Mmax ' 1:4 M
where M is the mass of the sun:
hc
Mmax = ( )3=2 mB 2 ( 2.7)
G
where mB refers here to the mass of electrons (or, see below, neutrons).
Note that the coeÆcients in front of mB 2 depend only on universal
physical quantities: the Planck constant h ; the velocity of light c ; and
the gravitational constant G : The maximum mass Mmax is called the
Chandrasekhar limit.
Although Chandrasekhar's derivation is mathematically correct, it
seemed somewhat cumbersome. Already by the end of 1932, L.D. Lan-
dau not only presented a more elementary explanation of the Chan-
drasekhar limit but, upon hearing of the discovery of the neutron, he ap-
plied the above formula to then putative neutron stars. Chandrasekhar's
analysis implies that stars of mass larger than the critical mass Mmax
could not be self-sustained, but would collapse under their own gravity.
These theoretical predictions had to wait till the 1960s for observational
conrmation; in the meantime, these consequences of the Fermi{Dirac
statistics rst caused quite a surf, at the forefront of which stood Edding-
ton, an astronomer commanding an imposing authority, who is reported
to have spoken of a reductio ad absurdum calling for the interposition
of an as yet unknown fundamental theory. For the unfolding of the res-
olution of the con ict, cf. e.g. [53], the title of which already indicates
the complete extent to which Chandrasekhar was ultimately vindicated.
Chandrasekhar however did not pursue this line of research past the syn-
thesis he presented in [17]) (see in particular Chap. XI). Nevertheless,
he revisited in the mid{1960s the question of the general{relativistic

stability of stars.
The early period of Chandrasekhar's diverse activities { the most
pertinent to my purpose here { calls for three remarks relative to the
emergence of quantum statistics in India. The rst remark is that while
[15] { accepted upon Fowler's recommendation { is widely acknowledged
as Chandrasekhar's `rst' paper on the subject, it was preceded by an
earlier paper, really his rst publication, which had appeared in volume
3 (1928) of the Indian Journal of Physics; its very title indicates al-
ready the astronomical motivations that were to move Chandrasekhar
along much of his career: it announces unambiguously his interest in the
\Thermodynamics of the Compton Eect" (compare with [15]) \with
Reference to the Interior of the Stars."
The second remark is to draw attention to the mention, in the title
of [15] of the `New Statistics': in the matter of a few months since his
encounter with Sommerfeld recalled above, Chandrasekhar had caught
the gist of the Fermi{Dirac statistics. The Fowler{Chandrasekhar pa-
pers truly rank among the very rst and perhaps the most spectacular
applications of the Fermi{Dirac statistics, right with the semi{classical
theory of multi{electron atoms proposed independently in 1927-1928 by
Thomas and by Fermi (who had proposed in 1926 the statistics that
bears his name); and the 1928 Sommerfeld theory of electrons in met-
als. The impact of the discovery of the Chandrasekhar limit may be
compared with the prediction of the super uid condensate in perfect
quantum gases obeying the Bose{Einstein statistics (see Subsection 2.1.
above).
My third remark concerns the way Chandrasekhar later considered
quantum statistics: \It is somewhat misleading to use the word `statis-
tics' in Èinstein{Bose statistics' and `Fermi{Dirac statistics.' There is
only one statistics, namely the Gibbs statistics ... the symmetrical and
antisymmetrical cases simply correspond to two dierent assumptions ...
but nevertheless, we have the same statistical theory (Gibbs) underlying
both cases. It would be more logical to refer to Èinstein{Bose formulae'
and `Fermi{Dirac formulae'." [17, p.384, fn.5]. The reason for this foot-
note seems to have lost some of its bite today, perhaps partly in view
of the spin{statistics theorem of quantum eld theory; another reason
might be more idiosyncratic: classical mechanics and classical statistics
272 Gerard G. Emch
were to take an increasingly more important place in Chandrasekhar's

career; cf. for instance the masterful paper [18] that he reportedly had
rst prepared for his own use and published only at the insistence of von
Neumann. I nd it interesting to note here, from [65, p.i], that during
the summer of 1928 Chandrasekhar had visited his uncle C.V. Raman
in Calcutta, who was active on both fronts of physics, quantum and
classical: the Raman eect had just been discovered, and the Einstein{
Smolukowski theory of uctuations had recently been called upon { and
rened { to account for light scattering in uids; see Subsection 2.2
above. Perhaps in a belated re ection of this early encounter, Chan-
drasekhar selected in [19] \ ... a series of papers which Smolukowski
wrote during the last ve years of his life, papers in which the founda-
tions of the modern theory of stochastic processes were laid ..."
2.4 Prasanta Chandra Mahalanobis, 1893{1972
As the eminence of statistical studies in India is inseparable from Maha-

lanobis' name, it may be worth mentioning, in an essay on the contribu-
tions of Indian mathematicians to quantum statistics, that Mahalanobis
himself graduated in 1912 with honours in physics from Presidency Col-
lege in Calcutta, and that he was the only candidate to achieve First
Class in Physics when he took the tripos in Cambridge. Furthermore,
it was from within the Physics department of Presidency College in
Calcutta that, in the early 1920s, Mahalanobis planted a seed, the Sta-
tistical Laboratory, that under his tutelage became in 1931 the Indian
Statistical Institute, which in 1959 was oÆcially declared an \Institution
of National Importance" by an act of Parliament. Indeed, to ensure that
a ow of new ideas and problems would stimulate its team of statisti-
cians, the Institute included, from its beginning, practitioners of other
disciplines independently organized in a wide variety of departments;
for the immediate purpose of this essay, suÆce it to recall the strong
tradition maintained in their own elds by the ISI mathematicians and
physicists who personify the motto of the Institute \Unity in Diversity".
In the public-relations material distributed by ISI, one reads that
Mahalanobis was a science organizer who disliked all forms of bureau-
cracy in science. Be that as it may, the inheritance of this ideal may
begin to explain how our Indian colleagues manage to maintain their
personal scholarly production in diÆcult interdisciplinary elds, amidst

often demanding circumstances and extensive administrative responsi-
bilities. In this sense also, Mahalanobis ought therefore to be mentioned
here as one of the pioneers who created a home for the educational and
research initiatives discussed in the next section. As unmistakable signs
that Mahalanobis' successors have successfully pursued his lofty ideals I
ought to mention two of them who are directly related to my story: C.R.
Rao, who lists among his more than 50 Ph.D. students both Varadarajan
(1960) and Parthasarathy (1962), the contributions of whom I delineate
in subsections 3.1 and 3.2 below; and the current ISI Director, Kalyan
B. Sinha, cf. subsection 3.3.
3 Contemporary Figures
3.1 Veeravalli S. Varadarajan
In the 1950s and 1960s the frustrating inconsistencies of quantum eld

theory prompted fundamental re-examinations of the mathematical foun-
dations of quantum mechanics laid down in 1932 by von Neumann. The
henceforth universal Fock space framework { that imprudently extended
to innitely many degrees of freedom von Neumann's uniqueness the-
orem on the representation of the canonical commutation relations for
nitely many degrees of freedom { was the rst casualty of this on-
slaught, as the works by Friedrichs, and by Garding and Wightman,
unearthed the existence of uncountably many inequivalent irreducible
representations of the CCR for innitely many degrees of freedom.
This discovery was part of a renewed interest in earlier proposals to
revise the basic axioms of quantum mechanics, namely those by Jordan,
von Neumann and Wigner (1934) and by Birkho and von Neumann
(1936).
JvNW's \On an Algebraic Generalization of the Quantum Mechan-
ical Formalism" is an attempt to replace the algebra of all bounded
operators on a Hilbert space, which ends up with a complete classi-
cation of all nite{dimensional Jordan algebras; this approach matured
in the hands of Segal (1947), and of Haag and Kastler (1964) into the
C algebraic approach to quantum eld theory and statistical mechan-
274 Gerard G. Emch
ics; the conceptual problem of characterizing the Jordan algebras, the

elements of which are the self-adjoint elements of a C algebra, was
solved in the 1980s by Alfsen and his collaborators.
BvN's \The Logic of Quantum Theory" focuses on the structure of
the projectors on a Hilbert space, and is perhaps better characterized
as the germ of a non-Boolean proposition calculus or, even better, of an
innite{dimensional projective geometry.
This is where Varadarajan enters the story. A mathematician by
training { M.A. in mathematics and statistics, University of Madras,
1957; Ph.D. in mathematics, ISI{Calcutta, 1960 { he made a point \to
mention that it was through the lectures of Prof. G.W. Mackey at Seat-
tle in 1961 on the mathematical foundations of quantum mechanics that
I rst began to appreciate the beauty and depth of the subject" [69].
As a complement to this tribute to Mackey's in uence on the coming
generation of mathematicians and mathematical physicists throughout
the world, let me mention also that at the very same time Varadarajan
was imbibing the physical thinking underlying Mackey's project, some of
us were learning its mathematical underpinnings { especially systems of
imprimitivity and induced representations (on which Varadarajan was
already an expert) { starting from Mackey's mimeographic Chicago lec-
ture notes [38] that were circulating at CERN in Geneva and elsewhere
in Europe.
Moving forward in time, one discerns that the overarching theme
of Varadarajan's many-faceted research { much of which conducted at
UCLA { came to be the mutually stimulating roles geometry, and in
particular symmetry, plays in mathematics and in physics.
In mathematics, his eorts were directed rst to measure-theoretical
aspects of, and convergence problems in classical probability, a eld in
which he wrote an early burst of some fteen papers in a span of ve
years; his research turned then to the representation theory of Lie groups
and algebras, its rst published result being [47]. He later established
himself as one of the closest followers, the best interpreter and heir of
Harish{Chandra who, after a presentation by Varadarajan, indeed told
him that \he was very pleased ... and felt like a composer meeting a
conductor." [70, p.32]. Varadarajan's mathematical purview extends
as well to the theory of ber bundles and connections, cf. e.g. [73]
in which the physics reader is treated to a grand tour from Weyl and
Dirac, through Bohm-Aharanov, to Yang-Mills and Atiyah and Bott,

not to speak of Singer; cf. also [72].
Varadarajan's interest in the foundations of quantum mechanics is
already evidenced in [67] and continues to this day [3, 12]. With regard
to quantum statistics proper { the focus of this essay { the projective
geometry approach Varadarajan chooses in [68] allows him to oer what
I consider to be the best presentation of a central element in the inter-
pretation of von Neumann density matrices as quantum states, namely
Gleason's theorem [68, Thm.4.23] and its extension to probability mea-
sures on the lattice of projectors of a general von Neumann algebra
[68, pp.146-147]; the latter thus establish contact, at the representation
level, with the other branch of quantum axiomatics, the C algebraic
approach (see above).
3.2 Kalyanapuram Rangachari Parthasarathy
Shortly after Varadarajan (see subsection 3.1 above), Parthasarathy ob-

tained his Ph.D. from ISI{Calcutta in 1962, after having taken the BA
from the University of Madras in 1956. They actually wrote together
[47]. With certainly dierent emphases, they both had acquired a broad
background in functional analysis allowing them to master measure the-
ory and integration, Banach spaces and involutive algebras, and repre-
sentations of Lie groups and algebras. Pertaining more immediately to
Parthasarathy's training in probability, I wish to note his initial stay
abroad, at the Steklov Mathematical Institute in Moscow (1962-1963).
Early con uences, bearing Parthasarathy own stamp, obtained in his
classic/classical [42], and his sharp but perhaps somewhat less widely
known [43]. The collection [9] illustrates the diverse interests Partha-
sarathy developed during his career.
Several of these strands are later braided in his magnum opus [44],
the oÆcial birth certicate of the theory of quantum stochastic processes;
for a presentation oering additional perspectives, see also [45].
Three questions must be addressed about the birth of any mathemat-
ical theory, and this one is no exception: the rst is to account, however
succinctly, for the gestation period; the second concerns the actual time
and place of its birth; and the third has to do with the resonance of the
276 Gerard G. Emch
event.
Firstly, the gestation period. In this case, one may want to skip
the archaeological phase (1932{1947) in which the physical concepts
and mathematical tools were devised that allowed the passage from a
quantum probability theory for nitely many degrees of freedom, to
one better adapted to the consideration of systems with innitely many
degrees of freedom. Let it suÆce here to evoke the spirits of Fock,
Wiener, Itô and Segal. In the most elementary sense, the question was
to unify two heretofore separated domains of enquiry, the description of
the classical, stochastic process behind the evolution equation proposed
in 1908 by Langevin for the Brownian motion of a particle in a viscous
stochastic medium
dv=dt = v + (t) ( 3.1)
with v the velocity, the coeÆcient of viscosity, and a random function,
the time average of which vanishes; and the quantum, deterministic
evolution described by the Heisenberg equation
dB=dt = i [H; B ] ( 3.2)
where B and H are self{adjoint linear operators acting on a Hilbert
space, with B representing an observable, and H the Hamiltonian; equiv-
alently, upon viewing [H; ] as a derivation, the latter may be written
more intuitively in the Liouville form dB=dt = i L[B ]: In the modern
era, the problem is brought in context by mentioning, on one side, the
work of Itô, cf. Iteka and Watanabe (1981); and, on the other side,
the seminal contributions of Segal (1956, 1960), Streater (1969), Araki
(1970), Parthasarathy and Schmidt (1972), Hudson and Cockroft (1977),
Hudson (1980), or Barnet, Streater and Wilde (1982); Accardi, Frigerio
and Lewis (1982); cf. Hudson [30] who does speak with a particular
authority allowing for the expression of some freely admitted personal
bias; some of these may have been softened by a few additions in the
above list, still incomplete, even as it is limited to the years preceding
1984 for reasons to be presently explained.
Secondly, the actual birth place of the theory of quantum stochastic
processes is generally marked as the papers of Hudson and Parthasarathy
[32, 33, 34]; see already [31] for an early appearance of the quantum
stochastic dierential equation which they will use to describe quantum
diusion.
Thirdly, the resonance. While the above line travels mainly through
territories where the language of quantum physics is spoken, the waves
reached quite rapidly the French coast where classical probabilists viewed
the event as a mathematically genuine generalization of their language,
requiring new techniques; one of the most in uential advocates of the re-
quired reformulations is Paul-Andre Meyer in whose work [39] the change
in perspective is explicitly implemented. Parthasarathy also found it in-
structive to sketch an outline of quantum stochastic calculus that nev-
ertheless allows him to demonstrate how the familiar classical Markov
processes can be realized in the quantum framework [46]. Yet, as echoes
were also heard from further intellectual shores: \Quantum probability
has been parodied, and indeed bitterly criticized, as an industry consist-
ing in proving non{commutative analogues of well{known commutative
theorems. In my opinion it is, or should be, nothing of the kind ..."
[29]. Indeed, no one here would represent dierential geometry as just
generalizing Euclidean geometry to curved manifolds. Hudson contin-
ues, oering his own view of quantum stochastic theory: \It inhabits
a non{commutative, quantum world which is fundamentally dierent
and new, and which completely transcends conventional mathematical
experience." One of the talents of Parthasarathy is to combine un ap-
pable rmness, steady diligence, and a necessary dose of that Indian
ingredient, patience. As already mentioned, these were recognized by
the request that he present [45]. Another tribute to these talents is
the creation of the ourishing school of quantum stochastic theory in
India, going hand{in{hand with its perhaps more outspoken counter-
part in England. Indeed, this development is in large part a deliberate
consequence of Parthasarathy's dedication which is evidenced { perhaps
unwittingly { by his message to Sridharan, one of the organizers of \a
useful symposium which could inspire the young colleagues of Seshadri
to stay in India and bring up a new generation of mathematicians" [37,
p.xxv].
3.3 Kalyan Bidhan Sinha
Sinha joined Parthasarathy at ISI{Delhi in 1978. He had taken his

Ph.D. in Physics (1969) from the University of Rochester, where he had
joined us in 1965, fresh from a M.Sc. from the University of Delhi; his
B.Sc was from Presidency College, Calcutta, 1963. I am pleased to say
278 Gerard G. Emch
here again that the young faculty members of the physics department at
Rochester { which I joined in 1966 { had their pick from a long parade of
brilliant Indian Ph.D. students attracted there by the reputation of our
eminent senior colleagues, Robert Marshak, Emil Wolf, Leonard Man-
del, Bruce French, and Susumu Okubo. Some students, like Sudarshan
who had worked with Marshak, helped in no small way in establishing
our appeal; others, more adventurous perhaps, choose recent arrivals,
and I was fortunate indeed that Kalyan approached me in the very rst
weeks after I had arrived in Rochester. As part of his own research pro-
gramme, he attended my classes and seminars on the algebraic approach
to statistical mechanics and eld theory, just at the time when Marshak
had commissioned me to write a book on the subject [24]; the book it-
self appeared after Kalyan had left, but several footnotes bring explicit
witness to his early mastery, part of which, I am sure he acquired from
his interaction in Rochester with a then post{doctoral associate of mine,
Richard Herman.
Having nished his Ph.D. with me, Kalyan joined my own former
Ph.D. advisor, J.M. Jauch in Geneva, where the second phase of his
training was completed. Werner Amrein and Kalyan Sinha drove to
completion the book [2] on the rigorous mathematical treatment of par-
ticle scattering in quantum mechanics, an oeuvre which Jauch had left
in outline at the time of his unexpected death in 1974. After the book
had appeared, and was greeted with universal acclaim, Kalyan contin-
ued along this line of research, rst in various positions he occupied in
Europe, North America, and Brazil, and then upon his return in India,
where he started training his own graduate students.
The theory of quantum stochastic processes seemed to be pretty
much set on its path when Kalyan joined Parthasarathy at ISI{Delhi.
Their rst papers indeed conrmed its increasing scope [48]; see also [40,
1, 49]; and for reviews, see [55] and [57] respectively; these advances were
gained in part by taking further advantage of techniques in functional
analysis and operator algebra theory which Kalyan had acquired in his
earlier and diverse activities in mathematical physics.
As these techniques were opening new vistas in modern geometry,
Parthasarathy, Sinha, their collaborators and their students were ready
for the extension of quantum stochastic processes to general von Neu-
mann algebras [27, 26] and non{commutative manifolds [13, 14, 56].
At this point, the question must be asked as to whether these beau-

tiful mathematical constructs may have as immediate an impact on our
physical understanding of this world as had their classical counterparts.
From my own vantage point in non{equilibrium statistical mechan-
ics, I am particularly impressed by the fact that a bridge is now con-
structed from both ends of the problem. Traditionally, we had started
from the quantum mechanics of nite systems, then taking their innite
thermodynamical limit and various mixed limits { e.g. long{time/weak
coupling { to focus on the appropriate regimes. Beyond this mid{point,
we aimed at obtaining { and in particular cases, we did obtain { by a
projection on the proper subalgebra of relevant observables, some con-
tractive semigroups of completely positive maps, such as those encoun-
tered in the actual phenomenological world, e.g. those described by the
Bloch equation in quantum spin relaxation or electric conductivity.
Thus, those of us who are involved in trying to obtain quantitative
microscopic explanations of transport phenomena would be happy to get
from the quantum stochastic calculus some information on the mathe-
matical structures encountered at the mid-point of the putative bridge,
as this calculus starts from the semigroups and constructs their stochas-
tic dilations in ways that can be viewed as canonical. Enthusiasts of
the new canon may even speculate that travelling the road by way of
stochastic dilations is all there is: what was envisaged as the `mid{point'
may be what the microscopic world ultimately looks like, a brand-new
challenging world indeed.
3.4 Mandyam Doddamane Srinivas
The work of Srinivas oers original insights towards an approach to

quantum stochastic processes, although the name of some of the pre-
cursors { notably J.T. Lewis { appeared also in subsection 3.2 above.
The motivations here are quite dierent, re ecting the fact that, as a
physics graduate student in Rochester during the 1970s, Srinivas found
himself at the con uence of two streams: quantum axiomatics and quan-
tum optics. The rst was initiated upon my own arrival in Rochester
a few years before Srinivas joined us; the second was dominated on the
theoretical side by an eminent pilot, Emil Wolf, and on the other side
by the pioneering experiments conducted in the laboratory directed by
280 Gerard G. Emch
Leonard Mandel. Although Srinivas and I had several happy discus-

sions together, as explicitly acknowledged in [59], Wolf { who had also
directed recently the thesis of G.S. Agarwal, yet another of the brilliant
Indian students in Rochester { became Srinivas' Ph.D. advisor and they
wrote several papers together; cf. e.g. [63]. Srinivas completed his Ph.D.
thesis in 1976.
Srinivas' own motivation to study quantum stochastic processes
comes from a critique of von Neumann's description of the quantum
measurement process on two accounts: (1) the traditional presentation
is limited to the measurement of observables with purely discrete spec-
trum; and (2) no provision is made for repeated measurements, nor a
fortiori for measurements carried on continuously in time. His critique
[59] challenges the concept of so{called \experimentally veriable propo-
sitions", and examines means to replace it by notions that re ect more
closely experimental procedures such as \operations", \instruments",
and \eects".
Again, as narrated in the preceding subsection, proposals such as
these do not spring out of thin air; they cannot be entirely new, and they
follow steps taken earlier by others. Here, I should certainly refer to the
inspirations of G. Ludwig (1967), and especially to the contributions of
J.T. Lewis and E.B. Davies (1969); some of the latter are systematically
developed by Davies in [20], who does cite the above paper by Srinivas.
Srinivas later contributed several responses { among which [60] { to
the challenges just described. For applications to the original motivation
in quantum optics, cf. [62]. These, as well as other related papers by
Srinivas, are collected in [61].
4 Concluding remarks
I aimed at presenting some of the circumstantial evidences for the How
and Why of quantum statistics in India, as they do appear to an outsider;
and I hope that my account can prove useful, however biased it may be
by my being a mathematical physicist whose career had to unfold on
foreign soils. What I see from this position is that a vital Indian school
of quantum statistics has been created, enhanced by deliberate cross-
fertilizations between diverse scientic disciplines. I expect that any
revision of this essay in a not too distant future will be able to sight also
the rising stars of quantum communication theory, and to focus more
on ideas than on personalities, thus re ecting the staying power of the
eld rather than the genius of its early leaders.
Acknowledgements
First and foremost, I must record my deep appreciation for the persis-
tent prodding to which my co-editor R. Sridharan submitted me so that
I bind together in a paper the sketches I had drawn at the December
2003 Bangalore meeting. I also wish to acknowledge a referee's sugges-
tions that I nd a link with S. Ramanujan and that I mention, however
succinctly, such eminent personalities as H.J. Bhabha and C.R. Rao.
References
[1] L. Accardi and K.B. Sinha, Stop-time Algebras, in Quantum Prob-
ability and Applications IV, L. Accardi et al., eds., Lecture Notes
in Mathematics 1396 (1989) 68{72.
[2] W. Amrein, J.M. Jauch and K.B. Sinha, Scattering theory in quan-
tum mechanics, Benjamin, Reading, 1977.
[3] R. Arens and V.S. Varadarajan, On the concept of Einstein{
Podolsky{Rosen states and their structure, J. Math. Phys. 41 (2000)
638{651.
[4] G. Baskaran, ed., Special section: K.S. Krishnan's Birth Centenary,
Current Science 75 (1998) 1197{1275.
[5] H.J. Bhabha and Harish{Chandra, On the theory of point particles,
Proc. Roy. Soc. London A 183 (1944) 134{141; On the elds and
equations of motion of point particles, Proc. Roy. Soc. London A
185 (1946) 250{268.
[6] H.J. Bhabha, Relativistic wave equations for the elementary par-
ticles, Rev. Modern Phys. 17 (1945) 200{216; On a class of rela-
tivistic wave{equations of spin 3/2, Proc. Indian Acad. Sci. A 34
(1951) 335{354.
282 Gerard G. Emch
[7] H.J. Bhabha, On the stochastic theory of continuous parametric

systems and its implication to electron cascades, Proc. Roy. Soc.
London A 202 (1950) 301{322; Note on the complete stochastic
treatment of electron cascades, Proc. Indian Acad. Sci. A 32 (1950)
154{161.
[8] H.J. Bhabha and A. Ramakrishnan, The mean square deviation of
the number of electrons and quanta in the cascade theory, Proc.
Indian Acad. Sci. A 32 (1951) 141{153.
[9] R. Bhatia, ed., Analysis, geometry, and probability: Essays in hon-
our of K.R. Parthasarathy, Hindustan Book Agency, New Delhi,
1996.
[10] S.N. Bose, Planck's Gesetz und Lichtquantenhypothese,
Zeits.Physik 20 (1924) 178{181.
[11] S.N. Bose, Warmegleichgewicht in Strahlungsfeld bei Anwesenheit
von Materie, Zeits.Physik 24 (1924) 384{392.
[12] G. Cassinelli and V.S. Varadarajan, On Accardi's notion of com-
plementary observables, Innite Dimensional Analysis, Quantum
Probability and Related Topics, 5 (2002) 135{144.
[13] P.S. Chakraborty, D. Goswami and K.B. Sinha, Probability and
geometry on some non{commutative manifolds, J. Operator Theory
49 (2003) 185{201.
[14] P.S. Chakraborty and K.B. Sinha, Geometry on the quantum
Heisenberg manifold, J. Funct. Anal. 203 (2003) 425{452.
[15] S. Chandrasekhar, The Compton scattering and the new statistics,
Proceedings of the Royal Society A125 (1929) 231{237.
[16] S. Chandrasekhar, The maximum mass of ideal white dwarfs, The
Astrophysical Journal 74 (1931) 81{82; see also: The highly col-
lapsed congurations of a stellar mass, Mon. Not. Roy. Astron. Soc.,
91, 456-66 (1931); The density of white dwarf stars, Phil. Mag., 11,
592-96 (1931); Some remarks on the state of matter in the interior
of stars, Zeitschrift fur Astrophysik 5 (1932) 321{327.
[17] S. Chandrasekhar, An Introduction to the study of stellar structure,
(1st ed. Univ. Chicago Press, 1939), Dover Publication, 1958.
[18] S. Chandrasekhar, Stochastic problems in physics and astronomy,

Rev. Modern Phys. 15 (1943) 1-89.
[19] S. Chandrasekhar, Marian Smolukowski as the founder of the
physics of stochastic phenomena, in: Marian Smoluchowski, his life
and scientic work, R.S. Ingarden, ed., Polish Scientic Publishers
PWN, Warszawa, 2000, 21{28.
[20] E.B. Davies, Quantum theory of open systems, Academic Press,
London, 1976.
[21] R.S. Doran and V.S. Varadarajan, The mathematical legacy of
Harish{Chandra, Proceedings of Symposia in Pure Mathematics,
68, American Mathematical Society, Providence R.I., 2000, pp. 1{
35.
[22] A. Einstein, Theorie der Opaleszenz von homogenen Flussigkeiten
und Flussigkeitsgemischen in der Nahe des kritischen Zustandes,
Annalen der Physik 33 (1910) 1275{1298.
[23] A. Einstein, Quantentheorie des einatomigen idealen Gases, Sitz.
Ber. Akad. Wissens, Berlin, 1924, 261|267; 1925, 3{14.
[24] G.G. Emch, Algebraic methods in statistical mechanics and quan-
tum eld theory, Wiley{Interscience, New York, 1972.
[25] W. Feller, An Introduction to Probability Theory and its Applica-
tions, vol. II, John Wiley, New York, 1966.
[26] D. Goswami, J.M. Lindsay, K.B. Sinha and S. Wills, Dilation of
CP- ow on a von Neumann algebra, Pac. Journ. Math. 211 (2003)
221{247.
[27] D. Goswami and K.B. Sinha, Hilbert modules and stochastic di-
lation of a quantum dynamical semigroup on a von Neumann al-
gebra, Comm. Math.Phys. 205 (1999) 377{403; Minimal quantum
dynamical semigroup on a von Neumann algebra, Innite Dimen-
sional Analysis, Quantum Probability and Related Topics, 2 (1999)
221{229.
[28] G. Greenstein, A Gentleman of the Old School: Homi Bhabha and
the Development of Science in India, The American Scholar, Sum-
mer 1992, pp. 409{419.
284 Gerard G. Emch
[29] R.L. Hudson, Thinking non{commutatively, in: [9], pp.87{124.

[30] R.L. Hudson, Quantum stochastic analysis after four decades, in:
Probability towards 2000, Lecture Notes in Statistics 128, Springer,
New York, 1998, pp. 194{217.
[31] R.L. Hudson and K.R. Parthasarathy, Construction of quantum
diusions, in: Quantum probability and applications to the quantum
theory of irreversible processes, (Villa Mondragone, 1982), Lecture
Notes in Mathematics 1055, Springer, Berlin, 1984.
[32] R.L. Hudson and K.R. Parthasarathy, Quantum Itô's formula and
stochastic evolutions, Comm. Math. Phys. 93 (1984) 301-323.
[33] R.L. Hudson and K.R. Parthasarathy, Stochastic dilations of uni-
formly continuous completely positive semigroups, Acta Applican-
dae Mathematicae 2 (1984) 353{358.
[34] R.L. Hudson and K.R. Parthasarathy, Unication of Fermion and
Boson stochastic calculus, Comm. Math. Phys. 104 (1986) 457{470.
[35] K.S. Krishnan, A Simple Result in Quadrature, Nature 162 (1948)
215; see already: A.B. Bhatia and K.S. Krishnan, Light{scattering
in homogeneous media regarded as re ection from appropriate ther-
mal elastic waves, Proc. Roy. Soc. London A 192 (1948) 181-194.
[36] K.S. Krishnan, On the Equivalence of Certain Innite Series and
the Corresponding Integrals, Journal of the Indian Mathematical
Society 12 (1948) 79{88.
[37] V. Lakshmibai et al., eds., A tribute to C.S. Seshadri: Perspectives
in geometry and representation Theory, Hindustan Book Agency,
Chennai, 2003.
[38] G.W. Mackey, The theory of group representations, Mimeographic
notes written by M. Fell and D. Lowdenslager, University of
Chicago, 1955; later published with a major Appendix as his The
theory of unitary group representations, The University of Chicago
Press, Chicago, 1976. See also his: Mathematical foundations of
quantum mechanics, W.A. Benjamin, New York, 1963; Induced
representations of groups and quantum mechanics, W.A. Benjamin,
New York, 1968; Unitary group representations in physics, probabil-

ity, and number theory, Benjamin/Cummings, Reading MA, 1978.
[39] P.{A. Meyer, Quantum probability for probabilists, Springer, Berlin,
Heidelberg, 1993.
[40] A. Mohari and K.R. Parthasarathy, On a class of generalized
Evans{Hudson ows related to classical Markov processes, Quan-
tum Probability and Related Topics VII (1992) pp. 221{249.
[41] A. Pais, Subtle is the Lord: the science and the life of Albert Ein-
stein, Clarendon Press, Oxford, 1982.
[42] K.R. Parthasarathy, Probability measures on metric spaces, Aca-
demic Press, New York, 1967.
[43] K.R. Parthasarathy, Multipliers on locally compact groups,
Springer, Berlin, Heidelberg, New York, 1969.
[44] K.R. Parthasarathy, An introduction to quantum stochastic calcu-
lus, Birkhauser, Basel, 1992; reviewed by V.S. Varadarajan, Found.
Phys. 34 (1994) 427{430; Parthasarathy's ISI Lectures on quantum
stochastic calculus, 1988 were a forerunner of his book.
[45] K.R. Parthasarathy, Quantum Stochastic Calculus, in: Proceed-
ings of the International Congress of Mathematicians, Zurich 1994,
Birkhauser, Basel, 1995, pp. 1024{1035.
[46] K.R. Parthasarathy, Classical Markov processes from a quantum
probabilistic point of view, Math. Stud. 65 (1996) 143{154.
[47] K.R. Parthasarathy, R. Ranga Rao, and V.S. Varadarajan, Rep-
resentations of complex semi{simple Lie groups and Lie algebras,
Ann. Math. (2) 85 (1967) 383{429; main theorems announced in
Bull. Amer. Math. Soc. 72 (1966) 522{525.
[48] K.R. Parthasarathy and K.B. Sinha, Stochastic Integral represen-
tations of bounded quantum martingales in Fock space, J. Funct.
Anal. 67 (1986) 126{151; Unication of quantum noise processes
in Fock space, Quantum Probability and Related Topics 6 (1991)
371{384; Quantum ows with innite degrees of freedom and their
perturbations, Quantum Probability and Related Topics VII (1992)
109{123.
286 Gerard G. Emch
[49] K.R. Parthasarathy and K.B. Sinha, Stop{times in Fock space

stochastic calculus, Probability Theory and Related Fields 75
(1987) 317{349;
[50] C.V. Raman, On the molecular scattering of light in water and the
colour of the sea, Proceedings of the Royal Society 64 (1922).
[51] C.V. Raman, Musikinstrumente und ihre Klange, Handbuch der
Physik VIII (1927) 354{424.
[52] C.V. Raman, A new radiation, Indian Journal of Physics 2 (1928)
387{398; (with K.S. Krishnan) A new type of secondary radiation,
Nature 121 (1928) 501.
[53] S.L. Shapiro and S.A. Teukolsky, Black holes, white dwarfs, and
neutron stars: The physics of compact objects, Wiley, New York,
1983.
[54] R. Singh and F. Riess, Sir C.V. Raman and the story of the Nobel
prize, Current Science 75 (1998) 965{971.
[55] K.B. Sinha, Quantum stochastic calculus { a Review, RIMS Publi-
cations 923 (1995) 206{227.
[56] K.B. Sinha, Stochastic ows on UHF algebras, Quantum probabil-
ity and innite dimensional analysis, Greifwald, 2003; Probability
and geometry in non{commutative spaces, India-US meeting, Ban-
galore, 2003.
[57] K.B. Sinha, Quantum Stop Times, Lectures at a workshop on Quan-
tum Probability, Institut Fourier, Grenoble June 1998 in Quantum
Probability and Communications, World Scientic, Singapore (to
appear).
[58] M. Smolukowski, Molekular{kinetische Theorie der Opaleszenz von
Gasen im kritischen Zustande, so wie einiger verwandter Erschei-
nungen, Annalen der Physik 25 (1908) 205{226.
[59] M.D. Srinivas, Foundations of a quantum probability theory, J.
Math. Phys. 16 (1975) 1672{1685.
[60] M.D. Srinivas, Collapse postulate for observables with continuous

spectra, Comm. Math. Phys. 71 (1980) 131{158; A general law for
quantum joint probabilities: Generalization of the Wigner formula
and the collapse postulate for successive measurements of discrete
as well as continuous observables, Pramana - J. Phys. 30 (1988)
259{277.
[61] M.D. Srinivas, Measurements and quantum probabilities, University
Press (India), Hyderabad, 2001.
[62] M.D. Srinivas and E.B. Davies, Photon counting probabilities in
quantum optics, Optica Acta 28 (1981) 981{996; M.D. Srinivas,
Quantum theory of continuous measurements and its applications
to quantum optics, Pramana - J. Phys. 47 (1996) 1{23.
[63] M.D. Srinivas and E. Wolf, Some non{classical features of phase{
space representations of quantum mechanics, Phys. Rev. D3 11
(1975) 1477{1485.
[64] G. Srinivasan, ed., Chandrasekhar memorial issue, Journal of As-
trophysics 17 (1996), Numbers 3 & 4.
[65] G. Srinivasan, Subrahmanyan Chandrasekhar (19.10.1910 {
21.8.1995), in [64, i-vii].
[66] J. Stachel, Einstein and Bose, Max-Planck-Institut fur Wis-
senschaftsgeschichte, Berlin, Preprint 4, 1994.
[67] V.S. Varadarajan, Probability in physics and a theorem on simulta-
neous observability, Comm. Pure Appl. Math. 15 (1962) 189{217.
[68] V.S. Varadarajan, Geometry of quantum theory, 2nd ed., Springer,
New York, 1985; earlier editions: GQT, Vol.I, van Nostrand,
Princeton, 1968; GQT, Vol.II, van Nostrand, New York, 1970.
[69] V.S. Varadarajan, Quantum theory and geometry: Sixty years af-
ter von Neumann, International Journal of Theoretical physics, 12
(1993) 1816{1834.
[70] V.S. Varadarajan, Harish{Chandra, his work and its legacy, in [21,
pp.1{35]
[71] V.S. Varadarajan, Harish{Chandra memorial talk, in [21, pp.53{57]
288 Gerard G. Emch
[72] V.S. Varadarajan, Supersymmetry for mathematicians: an intro-

duction, American Mathematical Society, Providence R.I., 2004.
[73] V.S. Varadarajan, Vector bundles and connections in physics and
mathematics: Some historical remarks, in [37] pp. 502{541.
[74] G. Venkataraman, Journey into Light { Life and Science of C.V.
Raman, Indian Academy of Science, Bangalore (distributed by Ox-
ford University Press), 1988; Penguin Books, India, 1994.
[75] K.C. Wali, ed.; A quest for perspectives: Selected works of S. Chan-
drasekhar, with commentary, Vol.1, Imperial College Press, Lon-
don, 2001.

Emch - Sridharan - Srinivas - Contributions Ot The History of Indian Mathematics (2005)

Uploaded by

Copyright:

Available Formats

Emch - Sridharan - Srinivas - Contributions Ot The History of Indian Mathematics (2005)

Uploaded by

Document Information

Original Description:

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

Emch - Sridharan - Srinivas - Contributions Ot The History of Indian Mathematics (2005)

Uploaded by

Copyright:

Available Formats

Contributions

The rst Joint India-AMS meeting in Mathematics was held in

June 17, 2005 Gerard G. Emch

II. Ancient Period

Shedding Some Localic and Linguistic Light on the

III. Classical and Medieval Period

The Karan.: How to Use Integers to Make Accurate

Relations between Approximations to the Sine in

Algorithms in Indian Mathematics

Algorithms in Indian Astronomy

Proofs in Indian Mathematics

IV. Modern Period

Contributions of Indian Mathematicians to Quantum Statistics

Baruch Spinoza (1632-1677), that gentle philosopher of the seven-

Some concepts are so fundamental that they get absorbed in the

the decimal notation is indeed a contribution of India, though, xing an

As we know, the Indus Valley civilisation (now, more properly termed

There is a huge gap in our knowledge of the interim between the

Puranic literature also. Thus, Jainism which arose as a revolt against

times (they occur in the prosody of Pingala,

In 1881 A.D., in the course of excavations, a farmer came across a

Chandah. padantu vedasya hastau kalpo'tha pat.hyate j

is of two kinds, either it is light (represented by ^ and called laghu)

whose metrical representation is

Translated, they read \a laghu if divisible by 2; if not add 1 and a guru".

the syllabic pattern corresponding to 44. Note that 44 is divisible by 2

theorem of Pythagoras, methods of extracting square roots and cube

caturadhikam. satamas.t.agun.am. dvas.as.t.istatha sahasran.am j

of a word in Arabic, meaning heart and sounding similar, the Romans

These slokas present many diÆculties in translating and have led to a

pounding the various prevalent astronomical doctrines named Paulisa,

Brahmagupta is perhaps the most remarkable mathematician of the

(x2 + y2 )(z 2 + t2 ) = (xz  yt)2 + (xt  yz )2 :

(x2 Ny2 )(z 2 Nt2 ) = (xz  Nyt)2 N (xt  yz )2 : ()

Bhavana is an example of the notion of \composition" of quadratic forms

x2 67y2 = 1 (x = 1766319049; y = 226153980 is the least positive

formed into a table of sines. Some eighty years before Brahmagupta,

This result is attributed to Euler by (cf. [7]) (though no proof is found in

The rst gure corresponds to juxtaposing the rational right angled

The next celebrated and essentially the last mathematician of the

The part of Siddhantasiroman.i, called Llavat deals with various

Prastare mitra gayatryah. syuh. pade vyaktayah. kati j

Mathematics in Medieval India (cf. [1,12,13,17]

works there is one which is called Sundararaja Prasnottara which records

Vyase varidhinihate rupahr.te vyasasagarabhihate j

We therefore have, by using the geometric series 1

To sum up, the contributions of India to mathematics during the

I would like to end this article with a couple of general remarks.

I thank the Nehru Centre, in particular, Mr. Satish Sahney, for

[2] M. Chasles: Apercu historique sur l'origine et development des

[16] R. Sridharan: Ancient Indian Contributions to Quadratic Algebra:

In India, the science of prosody, had its beginnings in the Vedic

the purely personal bias of individuals; and to recognize instead that we

§2. The beginnings of prosody

chandasāmeti sālokyamānantyāyāśnute śriyam

CdsA\ Evcy\ jAnn^ y, frFrAEmQyt.

Translated roughly into English the above stanza reads:

Chandah. pādantu vedasya hastau kalpo’tha pat.hyate

June 17, 2005 Gerard G. Emch

The Karan.: How to Use Integers to Make Accurate

Chandah. padantu vedasya hastau kalpo'tha pat.hyate j

caturadhikam. satamas.t.agun.am. dvas.as.t.istatha sahasran.am j

These slokas present many diÆculties in translating and have led to a

pounding the various prevalent astronomical doctrines named Paulisa,

(x2 + y2 )(z 2 + t2 ) = (xz yt)2 + (xt yz )2 :

(x2 Ny2 )(z 2 Nt2 ) = (xz Nyt)2 N (xt yz )2 : ()

Bhavana is an example of the notion of \composition" of quadratic forms

The part of Siddhantasiroman.i, called Llavat deals with various

Prastare mitra gayatryah. syuh. pade vyaktayah. kati j

works there is one which is called Sundararaja Prasnottara which records

Vyase varidhinihate rupahr.te vyasasagarabhihate j

[2] M. Chasles: Apercu historique sur l'origine et development des

CdsA\ Evcy\ jAnn^ y, frFrAEmQyt.

Cd, pAdt vd-y h-tO kSpo_T pÕt.

aEvEd(vA Eq\ Cdo {dvt\ yogm

t/AprA `vdo yjv

aSpA"r\ asEd`D\ sArvt^ EvtomK\.

Cdo v#y mlj{-t{, Ep½loÄ\ yTAm\ .

þk(yAEp lGy vbA &yvE-Tt, .

Myr-tjÞg {lA t{, eEBd fEBr"r{, .

aAEdm@yAvsAnq BjsA yAEt gOrv\ .

þ-tAro n£mE£\ ekyAEd lgEyA .

pAd sv grAvAA¥G\ y-y grorD, .

uE£\ EgZAApy ¬AsmAElKt^ .

vZA n^ vBvAn^ s{kAn^ aOrADy t, E-TtAn^ .

upA(yto Envt t (yjàk{km@v t, .

pAdA"rEmtgQC gZvg PlÒy EgZ .

lgEyA¬sdoh Bv(s\HyA EvEmEt