Development of Calculus in India

K. Ramasubramanian and M. D. Srinivas

In this article we shall present an overview of the development of calculus in Indian mathematical
tradition. The article is divided naturally into two parts. In the first part we shall discuss the de-
velopments during what may be called the classical period, starting with the work of Āryabhat.a
(c. 499 CE) and extending up to the work Nārāyan.a Pan.d.ita (c. 1350). The work of the Kerala
School starting with Mādhava of Saṅgamagrāma (c. 1350), which has a more direct bearing
on calculus, will be dealt with in the second part. Here we shall discuss some of the contributions
of the Kerala School during the period 1350–1500 as outlined in the seminal Malayalam work
Yuktibhās.ā of Jyes.t.hadeva (c. 1530).

PART I : T HE C LASSICAL P ERIOD

Āryabhat.a to Nārāyan.a Pan.d.ita (c. 500–1350 CE)

1. Introduction

In his pioneering history of calculus written sixty years ago, Carl Boyer was to-
tally dismissive of the Indian contributions to the conceptual development of the
subject.1 Boyer’s historical overview was written around the same time when (i)
Ramavarma Maru Thampuran and Akhileswarayyar brought out the first edition
of the Mathematics part of the seminal text Gan.ita-yukti-bhās.ā, and (ii) C.T. Ra-
jagopal and his collaborators, in a series of pioneering studies, drew attention to
the significance of the results and techniques outlined in Yuktibhās.ā (and the work
of the Kerala School of Mathematics in general), which seem to have been forgot-
ten after the initial notice by Charles Whish in early nineteenth century. These and
the subsequent studies have led to a somewhat different perception of the Indian
contribution to the development of calculus as may be gleaned from the following
quotation from a recent work on the history of mathematics:2

1C. B. Boyer, The History of the Calculus and its Conceptual Development, Dover, New York
1949, pp. 61–62.
2L. H. Hodgekin, A History of Mathematics: From Mesopotamia to Modernity, Oxford 2005,
p. 168.
202 K. Ramasubramanian and M. D. Srinivas

We have here a prime example of two traditions whose aims were completely
different. The Euclidean ideology of proof which was so influential in the Is-
lamic world had no apparent influence in India (as al-Biruni had complained long
before), even if there is a possibility that the Greek tables of ‘trigonometric func-
tions’ had been transmitted and refined. To suppose that some version of ‘calcu-
lus’ underlay the derivation of the series must be a matter of conjecture.
The single exception to this generalization is a long work, much admired in
Kerala, which was known as Yuktibhās.ā, by Jyes.t.hadeva; this contains some-
thing more like proofs—but again, given the different paradigm, we should be
cautious about assuming that they are meant to serve the same functions. Both
the authorship and date of this work are hard to establish exactly (the date usu-
ally claimed is the sixteenth century), but it does give explanations of how the
formulae are arrived at which could be taken as a version of the calculus.

The Malayalam work Gan.ita-yukti-bhās.ā (c. 1530) of Jyes.t.hadeva indeed

presents an overview of the work of Kerala School of mathematicians dur-
ing the period 1350–1500 CE. The Kerala School was founded by Mādhava
(c. 1340–1420), who was followed by the illustrious mathematician-astronomers
Parameśvara (c. 1380–1460), his son Dāmodara and the latter’s student
Nı̄lakan.t.ha Somayājı̄ (c. 1444–1550). While the achievements of the Kerala
School are indeed spectacular, it has now been generally recognised that these are
in fact very much in continuation with the earlier work of Indian mathematicians,
especially of the Āryabhat.an school, during the period 500–1350 CE.
In the first part of this article, we shall consider some of the ideas and meth-
ods developed in Indian mathematics, during the period 500–1350, which have a
bearing on the later work of the Kerala School. In particular, we shall focus on the
following topics: Mathematics of zero and infinity; iterative approximations for
irrational numbers; summation (and repeated summations) of powers of natural
numbers; use of second-order differences and interpolation in the calculation of
jyā or Rsines; the emergence of the notion of instantaneous velocity of a planet in
astronomy; and the calculation of the surface area and volume of a sphere.

2. Zero and Infinity

2.1. Background

The śānti-mantra of Īśāvāsyopanis.ad (of Śukla-yajurveda), a text of Brah-

mavidyā, refers to the ultimate absolute reality, the Brahman, as pūrn.a, the per-
fect, complete or full. Talking of how the universe emanates from the Brahman,
it states:

:pUa;NRa;ma;dH :pUa;NRa;Æa;ma;dM :pUa;Na;Ra;tpUa;NRa;mua;d;. ya;tea Á

:pUa;NRa;~ya :pUa;NRa;ma;a;d;a;ya :pUa;NRa;mea;va;a;va;a;Za;Sya;tea Á Á

That (Brahman) is pūrn
. a; this (the universe) is pūrn
. a; [this] pūrn
. a emanates
Development of Calculus in India 203

from [that] pūrn

. a; even when pūrn
. a is drawn out of pūrn
. a, what remains is
also pūrn
. a.

Pān.ini’s As..tādhyāyı̄ (c. 500 BCE) has the notion of lopa which functions as a
null-morpheme. Lopa appears in seven sūtras of Chapters 1, 3, 7, starting with
A;d;ZRa;nMa l+.ea;paH Á (1.1.60).

Śūnya appears as a symbol in Piṅgala’s Chandah. -sūtra (c. 300 BCE). In Chapter
VIII, while enunciating an algorithm for evaluating any positive integral power
of 2 in terms of an optional number of squaring and multiplication (duplication)
operations, śūnya is used as a marker:
.+.pea ZUa;nya;m,a Á ;a;d
õH ZUa;nyea Á (8.29-30).

Different schools of Indian philosophy have related notions such as the notion of
abhāva in Nyāya School, and the śūnyavāda of the Bauddhas.

2.2. Mathematics of zero in Brāhmasphut.a-siddhānta (c. 628 CE) of

Brahmagupta

The Brāhmasphut.a-siddhānta (c. 628 CE) of Brahmagupta seems to be the first

available text that discusses the mathematics of zero. Śūnya-parikarma or the
six operations with zero are discussed in the chapter XVIII on algebra (kut..takā-
dhyāya), in the same six verses in which the six operations with positives and
negatives (dhanarn.a-s.ad.vidha) are also discussed. Zero divided by zero is stated
to be zero. Any other quantity divided by zero is said to be taccheda (that with
zero-denominator):3
;Da;na;ya;ea;DRa;na;mxa;Na;mxa;Na;ya;ea;DRa;na;NRa;ya;ea:=+nta:=M .sa;mEa;k”+.Ma Ka;m,a Á

+.Na;mEa;k”+.aM . a ;Da;na;mxa;Na;Da;na;ZUa;nya;ya;eaH ZUa;nya;m,a Á Á

õ ;Za;ea;DyMa ;Da;nMa ;Da;na;a;dx;Na;mxa;Na;a;d;a;Da;k+:mUa;na;a;t,a Á

+:na;ma;a;Da;k+:a;a;d

v.ya;~tMa ta;d;nta:=M .~ya;a;dx;NMa ;Da;nMa ;Da;na;mxa;NMa Ba;va;a;ta Á Á

ZUa;nya;a;va;h;a;na;mxa;Na;mxa;NMa ;Da;nMa ;Da;nMa Ba;va;a;ta ZUa;nya;ma;a;k+:a;Za;m,a Á

Za;ea;DyMa ya;d;a ;Da;na;mxa;Na;a;dx;NMa ;Da;na;a;d

õ ;a ta;d;a [ea;pya;m,a Á Á

+.Na;mxa;Na;Da;na;ya;ea;Ga;Ra;ta;ea ;Da;na;mxa;Na;ya;ea;DRa;na;va;Da;ea ;Da;nMa Ba;va;a;ta Á

ZUa;nya;NRa;ya;eaH Ka;Da;na;ya;eaH Ka;ZUa;nya;ya;ea;va;Ra va;DaH ZUa;nya;m,a Á Á

;Da;na;Ba;€M ;Da;na;mxa;Na;&+.ta;mxa;NMa ;Da;nMa Ba;va;a;ta KMa Ka;Ba;€M Ka;m,a Á

Ba;€+:mxa;Nea;na ;Da;na;mxa;NMa ;Da;nea;na &+.ta;mxa;Na;mxa;NMa Ba;va;a;ta Á Á

3Brāhmasphut asiddhānta of Brahmagupta, Ed. with his own commentary by Sudhakara

.
Dvivedi, Benaras 1902, verses 18.30–35, pp. 309–310.
204 K. Ramasubramanian and M. D. Srinivas

Ka;ea:;d
Äâx ;ta;mxa;NMa ;Da;nMa va;a ta;. Ce+.dM Ka;mxa;Na;Da;na;a;va;Ba;€M va;a Á

+.Na;Da;na;ya;ea;vRa;gRaH .~vMa KMa Ka;~ya :pa;dM kx +:a;ta;yRa;t,a ta;t,a Á Á

. . . [The sum of] positive (dhana) and negative (r.n . a), if they are equal, is zero
(kham). The sum of a negative and zero is negative, of a positive and zero is
positive and of two zeros, zero (śūnya).
. . . Negative subtracted from zero is positive, and positive from zero is negative.
Zero subtracted from negative is negative, from positive is positive, and from zero
is zero (ākāśa). . . . The product of zero and a negative, of zero and a positive, or
of two zeroes is zero.
. . . A zero divided by zero is zero. . . . A positive or a negative divided by zero is
that with zero-denominator.

2.3. Bhāskarācārya on Khahara

Bhāskarācārya II (c. 1150), while discussing the mathematics of zero in Bı̄ja-

gan.ita, explains that infinity (ananta-rāśi) which results when some number is
divided by zero is called khahara. He also mentions the characteristic property of
infinity that it is unaltered even if ‘many’ are added to or taken away from it, in
terms similar to the invocatory verse of Īśāvāsyopanis. ad mentioned above:4
ãÉa .=+a;a;ZaH Á Á
Ka;h:=+ea Ba;vea;t,a Kea;na Ba;€+:(

È
îåMÁ ;a:ˆa;&+.t,a KMa Ka;&+.tMa ˆa;yMa . a ZUa;nya;~ya va;ga va;d mea :pa;dM . a Á Á
;a;d
õ .*+

. . . A;ya;ma;na;nta;ea 3/0 .=+a;a;ZaH Ka;h:=H I+.tyua;. ya;tea Á

A;a;sma;
/ / / / / a;nva;k+:a:=H
/ / Ka;h:=e na .=+a;Za;a;va;a;pa :pra;a;va;e;Sva;a;pa ;a;naH sxa;tea;Sua Á

ba;hu;Sva;a;pa .~ya;a;Œ+:ya;sxa;a;;k+:a;le Y;na;ntea Y;. yua;tea BUa;ta;ga;Nea;Sua ya;d

õ ;t,a Á Á

A quantity divided by zero will be (called) khahara (an entity with zero as di-
visor). Tell me . . . three divided by zero . . . This infinite (ananta or that without
end) quantity 03 is called khahara.
In this quantity, khahara, there is no alteration even if many are added or taken
out, just as there is no alteration in the Infinite (ananta), Infallible (acyuta)
[Brahman] even though many groups of beings enter in or emanate from [It] at
times of dissolution and creation.

2.4. Bhāskarācārya on multiplication and division by zero

Bhāskarācārya while discussing the mathematics of zero in Lı̄lāvatı̄, notes that

when further operations are contemplated, the quantity being multiplied by zero
should not be changed to zero, but kept as is. Further he states that when the quan-
tity which is multiplied by zero is also divided by zero, then it remains unchanged.

4Bı̄jaganita of Bhāskarācārya, Ed. by Muralidhara Jha, Benaras 1927, Vāsanā on

.
Khas.ad.vidham 3, p. 6.
Development of Calculus in India 205

He follows this up with an example and declares that this kind of calculation has
great relevance in astronomy:5

ya;ea;gea KMa [ea;pa;sa;mMa va;ga;Ra;d;Ea KMa Ka;Ba;a;Æa:ja;ta;ea .=+a;a;ZaH Á

Ka;h:=H .~ya;a;t,a Ka;gua;NaH ãÉa;ntya;(

KMa Ka;gua;Na;a;( ãÉa Zea;Sa;a;va;Da;Ea Á Á

ZUa;nyea gua;Na;k ãÉea;tpua;na;~ta;d;a .=+a;a;ZaH Á

e .ja;a;tea KMa h;a:=+(

ú
x +:ta O;;va :℄ea;ya;~ta;TEa;va Kea;na;ea;a;na;ta:ãÁ
A;a;va;k *.a yua;taH Á Á

ú
*.a;yua;gBa;va;a;ta ;a;k
KMa :pa:ãÁ M va;d Ka;~ya va;ga
ú
ãÉa :pa:ãÁ
mUa;lM ;Ga;nMa ;Ga;na;pa;dM Ka;gua;Na;a;( *.a Á

Kea;na;ea:;d
Äâx ;ta;a d;Za . a k H Ka;gua;Na;ea ;a;na:ja;a;DRa -
yua;€+:a;~:a;Æ ãÉa;gua;a;Na;taH
/ / a;Ba;( Ka;&+.ta;a;~:a;Sa;
/ / a;H Á Á

. . . A::℄a;a;ta;ea .=+a;a;ZaH ta;~ya gua;NaH 0 Á .sa;a;Da [ea;paH /

1 2 Á gua;NaH 3 Á h:=H

0 Á dx;ZyMa 63 Á

ta;ta;ea va;[ya;ma;a;Nea;na ;a;va;l+.ea;ma;a;va;a;Da;na;a I+.;k+:mRa;Na;a va;a l+.b.Da;ea .=+a;a;ZaH 14 Á

A;~ya ga;a;Na;ta;~ya g{a;h;ga;a;Na;tea ma;h;a;nua;pa;ya;ea;gaH Á

. . . A quantity multiplied by zero is zero. But it must be retained as such when
further operations [involving zero] are contemplated. When zero is the multiplier
of a quantity, if zero also happens to be a divisor, then that quantity remains
unaltered . . .
. . . What is the number which when multiplied by zero, being added to half of
itself multiplied by 3 and divided by zero, amounts to sixty-three?
. . . Either following the inverse process or by choosing a desired number for the
unknown (‘rule of false position’), the quantity is obtained to be 14. This kind of
calculation is of great use in mathematical astronomy.

Bhāskara works out his example as follows:

 
x 3
0 x+ × = 63
2 0
3x
×3 = 63.
2
Therefore, x = 14. (1)

Bhāskara, it seems, had not fully mastered this kind of “calculation with infinites-
imals” as is clear from the following example that he presents in Bı̄jagan.ita while
solving quadratic equations by eliminating the middle term:6

5Lı̄lāvatı̄ of Bhāskarācārya, Ed. by H. C. Bannerjee, Calcutta 1927, Vāsanā on verses 45–46,

pp. 14–15.
6Bı̄jaganita, cited above, Vāsanā on avyaktavargādi-samı̄karan am 5, pp. 63–64.
. .
206 K. Ramasubramanian and M. D. Srinivas

kH .~va;a;DRa;sa;a;h;ta;ea .=+a;a;ZaH Ka;gua;Na;ea va;a;gRa;ta;ea yua;taH Á

ú
ãÉa .ja;a;ta;aH
.~va;pa;d;a;Bya;Ma Ka;Ba;€+:( *.a;d;Za;ea;. ya;ta;a;m,a Á Á
:pa:ãÁ
Say what is the number which when added to half of itself, multiplied by zero,
squared and the square being augmented by twice its root and divided by zero,
becomes fifteen?

Clearly the problem as stated is

2
0(x + 2x ) + 2 × 0(x + x2 )
  
= 15. (2)
0

Bhāskara in his Vāsanā seems to just cancel out the zeros without paying any
heed to the different powers of zero involved. He converts the problem into the
equation
h x i2 h xi
x+ +2× x + = 15. (3)
2 2

Solving this, by the method of elimination of the middle term, Bhāskara obtains
the solution x = 2. The other solution (− 10
3 ) is not noted.

3. Irrationals and iterative approximations

3.1. Background
√ 7
Baudhāyana-śulva-sūtra gives the following approximation for 2:

:pra;ma;a;NMa txa;ta;a;yea;na va;DRa;yea:†a;ƒa . a;tua;TeRa;na;a;tma;. a;tua;a;~:M

/ /a;Za;ea;nea;na Á .sa;a;va;Zea;SaH Á

The measure [of the side] is to be increased by its third and this [third] again by
its own fourth less the thirty-fourth part [of the fourth]. That is the approximate
diagonal (saviśes.a).

√ 1 1 1
2 ≈ 1+ + −
3 3.4 3.4.34
577
=
408
= 1.4142156. (4)

7Baudhāyanaśulvasūtram (1.61-2), in The Śulvasūtras, Ed. by S. N. Sen and A. K. Bag, New

Delhi 1983, p. 19.
Development of Calculus in India 207

The above approximation is accurate to 5 decimal places. Baudhāyana-śulva-

sūtra—in the context of discussing the problem of circling a square—also gives
an approximation for π: 8

. a;tua:=+(rMa ma;Nq+.lM ;a;. a;k+:a;SRa;Ša;[Na;ya;a;Da ma;Dya;a;tpra;a;. a;a;ma;Bya;a;pa;a;ta;yea;t,a Á

ya;d;a;ta;a;Za;Sya;tea ta;~ya .sa;h;txa;ta;a;yea;na ma;Nq+.lM :pa;a:=+a;l+.Kea;t,a Á

If it is desired to transform a square into a circle, [a cord of length] half the

diagonal of the square is stretched from the centre to the east; with one-third [of
the part lying outside] added to the remainder [of the half-diagonal] the [required]
circle is drawn.

If a is half-the side of the square, then the radius r of the circle is given by
a √
r≈ (2 + 2). (5)
3
This corresponds to π ≈ 3.0883.

3.2. Algorithm for square-roots in Āryabhat.ı̄ya

The Āryabhat.ı̄ya of Āryabhat.a (c. 499 CE) gives a general algorithm for com-
puting the successive digits of the square root of a number. The procedure, given
in the following verse, is elucidated by us via an example:9
Ba;a;gMa h:=e+d;va;ga;Ra;a;Ša;tyMa ;a;d;gua;Nea;na va;gRa;mUa;le+.na Á
7 5
õ

va;ga;Ra;d
õ ;geRa Zua:;d
Äâe l+.b.DMa .~Ta;a;na;a;nta:=e mUa;l+.m,a Á Á
5 6 2 5
4 9
Always divide the non-square (even) place by 14) 7 2 (5
twice the square-root [already found]. Having sub- 7 0
tracted the square [of the quotient] from the square 2 5
(odd) place, the quotient gives the [digit in the] 2 5
next place in the square-root. 0 0

3.3. Approximating the square-root of a non-square number

The method for obtaining approximate square-root (āsanna-mūla) of a non-square

number (amūlada-rāśi) is stated explicitly in Triśatikā of Śrı̄dhara (c. 750):10

8Baudhāyanaśulvasūtram (1.58), ibid., p. 19.

9Āryabhatı̄ya of Āryabhata, Ed. by K. S. Shukla and K. V. Sarma, New Delhi 1976,
. .
Gan
. itapāda 4, p. 36.
10Triśatikā of Śrı̄dhara, Ed. by Sudhakara Dvivedi, Varanasi 1899, verse 46, p. 34.
208 K. Ramasubramanian and M. D. Srinivas

å
.=+a;Zea:=+mUa;l+.d;~ya;a;h;ta;~ya va;geRa;Na :ke +:na;a;. a;n}.a;h;ta;a Á

mUa;lM Zea;Sea;Na ;a;va;na;a ;a;va;Ba:jea;‘

ç Åu +Na;va;gRa;mUa;le+.na Á Á

Multiply the non-square number by some large square number, take the square-
root [of the product] neglecting the remainder, and divide by the square-root of
the multiplier.

Nārāyan.a Pan.d.ita (c. 1356) has noted that the solutions of varga-prakr.ti (the
so called Pell’s equation) can be used to compute successive approximations to the
square-root of a non-square number:11
mUa;lM g{a;a;hùÅ:aM ya;~ya . a ta;dÒU+pa;[ea;pa:jea :pa;de ta:ˆa Á

.$yea;ŽM &+.~va;pa;de;na . a .sa;mua:;d

Äâ :=e+t,a mUa;l+.ma;a;sa;Ša;m,a Á Á

[With the number] whose square-root is to be found as the prakr.ti and unity as
the ks.epa, [obtain the greater and smaller] roots. The greater root divided by the
lesser root is an approximate value of the square-root.

Nārāyan.a considers the example

10x 2 + 1 = y 2 , (6)

and gives the approximate values:

√ 19 721 27379
10 ≈ , , , (7)
6 228 8658
which are obtained by successive compositions (bhāvanā) of the solution x =
6, y = 9:12

228 = (2)(6)(19), 721 = (10)(6)2 + (19)2 , and so on.

3.4. Approximate value of π in Āryabhat.ı̄ya

Āryabhat.a (c. 499) gives the following approximate value for π:13

åò
. a;tua:=+a;Da;kM Za;ta;ma;;gua;NMa d
õ ;a;Sa;a;;~ta;Ta;a .sa;h;~:a;a;Na;a;m, a Á

õ ;ya;a;va;Sk+:}Ba;~ya;a;sa;Ša;ea vxa:†a;pa;a:=+Na;a;hH Á Á
A;yua;ta;d

11Ganitakaumudı̄ of Nārāyana Pandita, Ed. by Padmakara Dvivedi, Part II, Benaras 1942,
. . ..
verse 10.17, p. 244.
12Bhāvanā or the rule of composition enunciated by Brahmagupta is the transformation (X, Y )→
(X 2 + DY 2 , 2XY ) which transforms a solution x = X, y = Y of the equation x 2 − Dy 2 = 1, into
another solution with √
larger values for x, y, which correspond to higher convergents in the continued
fraction expansion of D and thus give better approximations to it.
13Āryabhatı̄ya, cited above, Ganitapāda 10, p. 45.
. .
Development of Calculus in India 209

One hundred plus four multiplied by eight and added to sixty-two thousand: This
is the approximate measure of the circumference of a circle whose diameter is
twenty-thousand.
62832
Thus as per the above verse π ≈ 20000 = 3.1416.

3.5. Successive doubling of the sides of the circumscribing polygon

It appears that Indian mathematicians (at least in the Āryabhat.an tradition)

employed the method of successive doubling of the sides of a circumscribing
polygon—starting from the circumscribing square leading to an octagon, etc.—
to find successive approximations to the circumference of a circle. This method
has been described in the later Kerala texts Yuktibhās.ā (c. 1530) of Jyes..thadeva
and Kriyākramakarı̄ commentary (c. 1535) of Śaṅkara Vāriyar on Lı̄lāvatı̄, of
Bhāskara II. The latter cites the verses of Mādhava (c. 1340–1420) in this con-
nection and notes at the end that:14
O;;vMa ya;a;va;d;Ba;a;M .sUa;[ma;ta;a;ma;a;pa;a;d;a;ya;tMua Za;k”+.a;m,a Á
Thus, one can obtain [an approximation to the circumference of the circle] to any
desired level of accuracy.

E A3 A2 A1

C2
B3
D2

C1

D1
B2

S
O

F IGURE 1. Finding the circumference of a square from circum-

scribing polygons.

14Lı̄lāvatı̄ of Bhāskara II, Ed. with commentary Kriyākramakarı̄ of Śaṅkara Vāriyar by

K. V. Sarma, Hoshiarpur 1975, comm. on verse 199, p. 379.
210 K. Ramasubramanian and M. D. Srinivas

We now outline this method as described in Yuktibhās.ā.15 In Figure 1, E O S A1

is the first quadrant of the square circumscribing the given circle. E A1 is half the
side of the circumscribing square. Let O A1 meet the circle at C1 . Draw A2 C1 B2
parallel to E S. E A2 is half the side of the circumscribing octagon.
Similarly, let O A2 meet the circle at C2 . Draw A3 C2 B3 parallel to EC1 . E A3
is now half the side of a circumscribing regular polygon of 16 sides. And so on.
Let half the sides of the circumscribing square, octagon etc., be denoted

l1 = E A1 , l2 = E A2 , l3 = E A3 , . . . (8)

The corresponding karn.as (diagonals) are

k1 = O A1 , k2 = O A2 , k3 = O A3 , . . . (9)

And the ābhādhas (intercepts) are

a 1 = D1 A 1 , a 2 = D2 A 2 , a 3 = D3 A 3 , . . . (10)

Now
√ r
l1 = r, k1 = 2r and a1 = √ . (11)
2
Using the bhujā-kot.i-karn.a-nyāya (Pythagoras theorem) and trairāśika-nyāya
(rule of three for similar triangles), it can be shown that

l1
l2 = l1 − (k1 − r ) (12)
a1
k22 = r 2 + l22 (13)
[k22 − (r 2 − l22 )]
and a2 = . (14)
2k2
In the same way ln+1 , kn+1 and an+1 are to be obtained in terms of ln , kn and
an . These can be shown to be equivalent to the recursion relation:16

r
q
ln+1 = [ (r 2 + ln2 ) − r ]. (15)
ln

15Ganita-yukti-bhāsā of Jyesthadeva, Ed. and Tr. by K. V. Sarma, with Exp. Notes by K. Ra-
. . ..
masubramanian, M. D. Srinivas and M. S. Sriram, 2 Vols, Hindustan Book Agency, New Delhi 2008.
Reprint Springer 2009, Vol. I Section 6.2, pp. 46–49, 180–83, 366–69.  
16If we set r = 1 and l = tan θ , then equation (15) gives l
n n n+1 = tan θn . Actually, θ = π
n
2 2n+1
π ≈ 2n π = π .
and the above method is based on the fact that for large n, 2n tan n+2
2 2n+2 4
Development of Calculus in India 211

4. Summation (and repeated summations) of powers of natural numbers

(saṅkalita)

4.1. Sum of squares and cubes of natural numbers in Āryabhat.ı̄ya

The ancient text Br.haddevatā (c. 5t h century BCE) has the result

2 + 3 + . . . + 1000 = 500, 499. (16)

Āryabhat.a (c. 499 CE), in the Gan.itapāda of Āryabhat.ı̄ya, deals with a general
arithmetic progression in verses 19–20. He gives the sum of the squares and cubes
of natural numbers in verse 22:17
.sEa;k+:sa;ga;. C+.pa;d;a;na;Ma kÒ +:ma;a;t,a ;a:ˆa;sMa;va;a;gRa;ta;~ya :Sa;Ž+eMa Y;ZaH Á

va;gRa;a;. a;a;ta;Ga;naH ãÉa Á Á

.sa Ba;vea;t,a ;a;. a;a;ta;va;ga;eRa ;Ga;na;a;. a;a;ta;Ga;na;(
The product of the three quantities, the number of terms plus one, the same in-
creased by the number of terms, and the number of terms, when divided by six,
gives the sum of squares of natural numbers (varga-citi-ghana). The square
of the sum of natural numbers gives the sum of the cubes of natural numbers
(ghana-citi-ghana).

In other words,

n(n + 1)(2n + 1)
12 + 22 + 32 + . . . + n 2 = (17)
6
13 + 23 + 33 + . . . + n 3 = [1 + 2 + 3 + . . . + n]2
n(n + 1) 2
 
= . (18)
2

4.2. Repeated sum of natural numbers in Āryabhat.ı̄ya

Āryabhat.a also gives the repeated sum of the sum of the natural numbers (saṅka-
lita-saṅkalita or vāra-saṅkalita):18
O;;k+:ea:†a:=+a;dùÅ;au;pa;a;. a;tea;gRa;. C+.a;dùÅ;ae;k+:ea:†a:=+a:ˆa;sMa;va;gRaH Á

ñ
ÂÅå +Å :€ H
:Sa:*. .sa ;a;. a;a;ta;Ga;naH .sEa;k+:pa;d;Ga;na;ea ;a;va;mUa;l+.ea va;a Á Á
Of the series (upaciti) 1, 2, . . . , n, take three terms in continuation of which the
first is the given number of terms (gaccha), and find their product; that [product],
or the number of terms plus one subtracted from its own cube divided by six, gives
the repeated sum (citi-ghana).

17Āryabhatı̄ya, cited above, Ganitapāda 22, p. 65.

. .
18Āryabhatı̄ya, cited above, Ganitapāda 21, p. 64.
. .
212 K. Ramasubramanian and M. D. Srinivas

We have
n(n + 1)
1 +2 +3 +...+n = . (19)
2
Āryabhat.a’s result expresses the sum of these triangular numbers in two forms:

(1 + 1) (2 + 1) (n + 1) [n(n + 1)(n + 2)]

1 +2 +...+n =
2 2 2 6
[(n + 1)3 − (n + 1)]
= . (20)
6

4.3. Nārāyan.a Pan.d.ita’s general formula for Vārasaṅkalita

In his Gan.ita-kaumudı̄, Nārāyan.a Pan.d.ita (c. 1356) gives the formula for the
r t h -order repeated sum of the sequence of numbers 1, 2, 3, . . . , n:19
O;;k+:a;a;Da;k+:va;a:=+Æa;ma;ta;aH :pa;d;a;a;d:+.pa;ea:†a:=+a :pxa;Ta;k, .teMa Y;Za;aH Á
ñÍ
O;;k+:a;dùÅ;ae;k+:. a;ya;h:=+a;~ta:.ÈÅ
‘+a;ta;ea va;a:=+sa;*:+.
öÅ a;l+.ta;m,a Á Á
The pada (number of terms in the sequence) is the first term [of an arithmetic
progression] and 1 is the common difference. Take as numerators [the terms
in the AP] numbering one more than vāra (the number of times the repeated
summation is to be made). The denominators are [terms of an AP of the same
length] starting with one and with common difference one. The resultant product
is vāra-saṅkalita.

Let
n(n + 1)
1 +2 +3 + ...+n = = Vn(1). (21)
2
Then, Nārāyan.a’s result is
(r−1) (r−1)
Vn(r) = V1 + V2 + . . . + Vn(r−1) (22)
[n(n + 1) . . . (n + r )]
= . (23)
[1.2 . . . (r + 1)]
Nārāyan.a’s result can also be expressed in the form of a sum of polygonal num-
bers:
n
X [m(m + 1) . . . (m + r − 1)] [n(n + 1) . . . (n + r )]
= . (24)
[1.2 . . . r ] [1.2 . . . (r + 1)]
m=1

19Ganitakaumudı̄ of Nārāyana Pandita, Ed. by Padmakara Dvivedi, Part I, Benaras 1936, verse
. . ..
3.19–20, p. 123.
Development of Calculus in India 213

This result can be used to evaluate the sums nk=1 k 2 , nk=1 k 3 , . . . by induction.
P P
It can also be used to estimate the behaviour of these sums for large n.

4.4. Summation of geometric series

The geometric series 1 + 2 + 22 + . . . 2n is summed in Chapter VIII of Piṅgala’s

Chandah.-sūtra (c. 300 BCE). As we mentioned earlier, Piṅgala also gives an
algorithm for evaluating any positive integral power of a number (2 in this context)
in terms of an optimal number of squaring and multiplication operations.
Mahāvı̄rācārya (c. 850), in his Gan.ita-sāra-saṅgraha gives the sum of a geo-
metric series and also explains the Piṅgala algorithm for finding the required power
of the common ratio between the terms of the series:20
:pa;d;Æa;ma;ta;gua;Na;h;a;ta;gua;a;Na;ta;pra;Ba;vaH ç Åu +Na;Da;nMa ta;d;a;dùÅ;aU;na;m,a Á
.~ya;a;‘

ñÍ
O;;k+:ea;na;gua;Na;a;va;Ba;€M gua;Na;sa;*:+.
öÅ a;l+.tMa ;a;va:ja;a;na;a;ya;a;t,a Á Á

.sa;ma;d;l+.a;va;Sa;ma;~va:+.pa;ea gua;Na;gua;a;Na;ta;ea va;gRa;ta;a;a;q+.ta;ea ga;. CH Á

È
îå Á a v.yea;k+:ea:†a:=+Ba;a;Æa:ja;taH
.+.pa;ea;naH :pra;Ba;va.*+;e .sa;a:=+m,a Á Á
The first term when multiplied by the product of the common ratio (gun . a) taken
as many times as the number of terms (pada), gives rise to the gun . adhana.
This gun . adhana, when diminished by the first term and divided by the common
ratio less one, is to be understood as the sum of the geometrical series (gun . a-
saṅkalita).

That is
a(r n − 1)
a + ar + ar 2 + . . . + ar n−1 = . (25)
(r − 1)
Vı̄rasena (c. 816), in his commentary Dhavalā on the S.at.khan.d.āgama, has made
use of the sum of the following infinite geometric series in his evaluation of the
volume of the frustum of a right circular cone:21
 2  n
1 1 1 4
1+ + + ...+ +... = . (26)
4 4 4 3

The proof of the above result is discussed in the Āryabhat.ı̄ya-bhās.ya (c. 1502)
of Nı̄lakan.t.ha Somayājı̄. As we shall see later (section 10.1), Nı̄lakan.t.ha makes
use of this series for deriving an approximate expression for a small arc in terms
of the corresponding chord in a circle.

20Ganitasārasaṅgraha of Mahāvı̄rācārya, Ed. by Lakshmi Chanda Jain, Sholapur 1963,

.
verses 2.93–94, pp. 28–29.
21See, for instance, T. A. Sarasvati Amma, Geometry in Ancient and Medieval India, Motilal
Banarsidass, Delhi 1979, Rep. 2007, pp. 203–05.
214 K. Ramasubramanian and M. D. Srinivas

5. Use of Second-order differences and interpolation in computation of

Rsines (Jyānayana)

Jyā, Kot.i and Śara

The jyā or bhujā-jyā of an arc of a circle is actually the half-chord (ardha-jyā

or jyārdha) of double the arc. In the Figure 2, if R is the radius of the circle, jyā
(Rsine), kot.i or kot.i-jyā (Rcosine) and śara (Rversine) of the cāpa (arc) EC are
given by:

jyā (ar c EC) = C D = R sin(6 C O E) (27)

kot.i (ar c EC) = O D = R cos(6 C O E) (28)
śara (ar c EC) = E D = Rvers(6 C O E)
= R − R cos(6 C O E). (29)

For computing standard Rsine-tables (pat.hita-jyā), the circumference of a circle

E
S
E

C D

F IGURE 2. Jyā, Kot.i and Śara.

is divided into 21600′ and usually the Rsines are tabulated for every multiple of
225′, thus giving 24 tabulated Rsines in a quadrant. Using the value of π ≈
62832
20000 = 3.1416, given by Āryabhat.a, the value of the radius then turns out to be
3437′ 44′′ 19′′′ . This is accurate up to the seconds, but is usually approximated
to 3438′. Using a more accurate value of π, Mādhava (c. 1340–1420) gave the
value of the radius correct to the thirds as 3437′ 44′′ 48′′′ which is also known by
the Kat.apayādi formula devo-viśvasthalı̄-bhr.guh. .
Development of Calculus in India 215

5.1. Computation of Rsines

Once the value of the radius R is fixed (in units of minutes, seconds etc.) the 24
Rsines can be computed (in the same units) using standard relations of jyotpatti
(trigonometry). For instance, Varāhamihira has given the following Rsine values
and relations in his Pañcasiddhāntikā (c. 505):22

R
R sin(30◦ ) = (30a)
2
R
R sin(45◦ ) = √ (30b)
2
√
3
R sin(60◦ ) = R (30c)
2
R sin(90◦ ) = R (30d)

R sin(A) = R cos(90 − A) (31)

R sin2 (A) + R cos2 (A) = R2 (32)
   
A 1 1
R sin = [R sin2 (A) + R vers2 (A)] 2
2 2
 1
R 2 1
= [R − R cos A] 2 . (33)
2

The above Rsine values (30) and relations (31)–(33) can be derived using the
bhujā-kot.i-karn.a-nyāya (Pythagoras theorem) and trairāśika (rule of three for
similar triangles), as is done for instance in the Vāsanā-bhās.ya of Pr.thūdaka-
svāmin (c. 860) on Brāhmasphut.asiddhānta (c. 628) of Brahmagupta. Equa-
tions (30)–(33) can be used to compute all 24 tabular Rsine values.

5.2. Āryabhat.a’s computation of Rsine-differences

The computation of tabular Rsine values was made much simpler by Āryabhat.a
who gave an ingenious method of computing the Rsine-differences, making use of
the important property that the second-order differences of Rsines are proportional
to the Rsines themselves:23

22Pañcasiddhāntikā of Varāhamihira, Ed. by T. S. Kuppanna Sastry and K. V. Sarma, Madras

1993, verses 4.1–5, pp. 76–80.
23Āryabhatı̄ya, cited above, Ganitapāda 12, p. 51.
. .
216 K. Ramasubramanian and M. D. Srinivas

:pra;Ta;ma;a;ƒa;a;pa:$ya;a;Da;Ra;dùÅ;aE:+.nMa Ka;a;Nq+.
/ /  tMa ;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 Á Á
The first Rsine divided by itself and then diminished by the quotient will give
the second Rsine-difference. The same first Rsine, diminished by the quotients
obtained by dividing each of the preceding Rsines by the first Rsine, gives the
remaining Rsine-differences.

Let B1 = R sin (225′ ), B2 = R sin (450′ ), ..., B24 = R sin (90◦ ), be the
twenty-four Rsines, and let 11 = B1 , 12 = B2 − B1 , ..., 1k = Bk − Bk−1 , ...
be the Rsine-differences. Then, the above rule may be expressed as24

B1
12 = B1 − (34)
B1
(B1 + B2 + . . . + Bk )
1k+1 = B1 − (k = 1, 2, . . . , 23). (35)
B1

This second relation is also sometimes expressed in the equivalent form

(11 + 12 + . . . + 1k )
1k+1 = 1k − (k = 1, 2, . . . , 23). (36)
B1

From the above it follows that

−Bk
1k+1 − 1k = (k = 1, 2, . . . , 23). (37)
B1

Since Āryabhat.a also takes 11 = B1 = R sin(225′ ) ≈ 225′, the above relations

reduce to

11 = 225′ (38)
−Bk
1k+1 − 1k = (k = 1, 2, . . . , 23). (39)
225′

5.3. Derivation of the Āryabhat. a-relation for the second-order

Rsine-differences

Āryabhat.a’s relation for the second-order Rsine-differences is derived and made

more exact in the Āryabhat.ı̄ya-bhās.ya (c. 1502) of Nı̄lakan.t.ha Somayājı̄ and
Yuktibhās.ā (c. 1530) of Jyes.t.hadeva. We shall present a detailed account of

24Āryabhata is using the approximation 1 − 1 ≈ 1′ and the second terms in the RHS of
. 2 1
(34)–(36) and the RHS of (37) and (39) have an implicit factor of (12 − 11 ). See (45) below which is
exact.
Development of Calculus in India 217

the first and second-order Rsine-differences as given in Yuktibhās.ā 25 later in

Section 16. Here we shall only summarize the argument.
In Figure 3, the arcs EC j and EC j +1 are successive multiples of 225′ . The
Rsine and Rcosine of the arcs EC j and EC j +1 are given by

Bj = C j P j , B j +1 = C j +1 P j +1 (40)
and Kj = C j T j , K j +1 = C j +1 T j +1, (41)

respectively. Let M j +1 and M j be the mid-points of the arcs C j C j +1 , C j −1 C j

and the Rsine and Rcosine of the arcs E M j and E M j +1 be denoted respectively
by B j − 1 , B j + 1 , K j − 1 , K j + 1 .
2 2 2 2

Cj E

E S
M j+1
α Mj Qj
Cj
Pj
C j+1 M j+1 G
Q j+1
F C j+1 Pj+1
F

N
Tj+1 Uj+1 Tj Uj O

F IGURE 3. Derivation of Āryabhat.a relation.

25Ganita-yukti-bhāsā, cited above, Section 7.5.1, pp. 94–96, 221–24, 417–20.

. .
218 K. Ramasubramanian and M. D. Srinivas

Let the chord of the arc C j C j +1 , be denoted by α and let R be the radius. Then
a simple argument based on trairāśika (similar triangles) leads to the relations:26
α
B j +1 − B j = K j+ 1 (42)
R
α
2

K j− 1 − K j+ 1 = Bj. (43)
2 2 R
Thus we get

1 j +1 − 1 j = (B j +1 − B j ) − (B j − B j −1 )
 α 2
= − Bj. (44)
R
We can also express this relation in the form

−B j (11 − 12 )
1 j +1 − 1 j = . (45)
B1

The above relations are exact. Āryabhat.a’s relation (39) corresponds to the ap-
proximations, B1 ≈ 225′ and 11 − 12 ≈ 1′ so that
 
 α 2 (11 − 12 ) 1
= ≈ . (46)
R B1 225′

In Tantrasaṅgraha, Nı̄lakan.t.ha Somayājı̄ has given the finer approximation:27

!
 α 2 (11 − 12 ) 1
= ≈ ′ . (47)
R B1 233 12

26Equations (42) and (43) are essentially the relations:

α  
h
R sin(x + h) − R sin x = R cos x +
R 2
    α
h h
R cos x − − R cos x + = R sin x,
2 2 R

with α = 2R sin h2 . These lead to (44) in the form:

 α 2
(R sin(x + h) − R sin x) − (R sin x − R sin(x − h)) = − R sin x.
R

27Tantrasaṅgraha of Nı̄lakantha Somayājı̄, Ed. with Laghu-vivrti of Śaṅkara Vāriyar by

.. .
S. K. Pillai, Trivandrum 1958, verse 2.4, p. 17.
Development of Calculus in India 219

This is further refined by Śaṅkara Vāriyar in his commentary Laghu-vivr.ti in the

form:28
 
 α 2 (11 − 12 ) 1
= ≈ . (48)
R B1 233′ 32′′

Since α = 2R sin 112′ 30′′ , we find that the above relation is correct up to seconds.
Commenting on Āryabhat.a’s method of computing Rsines, Delambre had re-
marked:29
The method is curious: it indicates a method of calculating the table of sines by
means of their second-differences... This differential process has not up to now
been employed except by Briggs, who himself did not know that the constant
factor was the square of the chord 1A (= 3◦ 45′ ) or of the interval, and who could
not obtain it except by comparing the second differences obtained in a different
manner. The Indians also have probably done the same; they obtained the method
of differences only from a table calculated previously by a geometric process.
Here then is a method which the Indians possessed and which is found neither
amongst the Greeks, nor amongst the Arabs.

5.4. The Rsine-table of Āryabhat. a

In the Gı̄tikā-pāda of Āryabhat.ı̄ya, Āryabhat.a has given a table of Rsine-

differences:30
ma;a;Ka
 Ba;a;Ka
 :P+.a;Ka
 ;Da;a;Ka
 :Na;a;Ka
 Va;a;Ka

å å
.z+.a;Ka
 h;~.Ja .~k+:a;k ;a;k+:Sga .~.Ga;a;k ;a;k+:Gva Á

;Gl+.a;k ;a;k+:g{a h;k”+.a ;Da;a;k ;a;k+:. a

ëÅÉ
.~ga ZJa z
õÅÉ *
ì: ‹a :P C k+:l+.a;DRa:$ya;aH Á Á

225, 224, 222, 219, 215, 210, 205, 199, 191, 183, 174, 164, 154, 143, 131, 119,
106, 93, 79, 65, 51, 37, 22, and 7—these are the Rsine-differences [at intervals of
225’ of arc] in terms of the minutes of arc.

The above values follow directly from Āryabhat.a’s relation (39) for the second
order Rsine-differences. To start with, 11 = B1 = R sin(225′ ) ≈ 225′ . Then we
get, 12 = B1 − BB11 = 224′ and so on.
The Rsine-table of Āryabhat.a31 (see Table 1), obtained this way, is accurate up
to minutes. In this table, we also give the Rsine values given by Govindasvāmin
(c. 825) in his commentary on Mahābhāskarı̄ya of Bhāskara I, and by Mādhava

28Ibid., comm. on verse 2.4.

29Delambre, Historie de l’ Astronomie Ancienne, t 1, Paris 1817, pp. 457, 459f, cited from
B. B. Datta and A. N. Singh, ‘Hindu Trigonometry’, Ind. Jour. Hist. Sc. 18, 39–108, 1983, p. 77.
30Āryabhatı̄ya, cited above, Gı̄tikāpāda 12, p. 29.
.
31See, for instance, A. K. Bag, Mathematics in Ancient and Medieval India, Varanasi 1979,
pp. 247–48.
220 K. Ramasubramanian and M. D. Srinivas

(c. 1340–1420) as recorded in the Āryabhat.ı̄ya-bhās.ya (c. 1502) of Nı̄lakan.t.ha

Somayājı̄. Though Govindasvāmin gives the Rsine values up to the thirds, his
values are accurate only up to the seconds; those of Mādhava are accurate up to
the thirds.

Arc-length Āryabhat.a Govindasvāmin Mādhava

(c. 499) (c. 825) (c. 1375)
3◦ 45′ 225′ 224′ 50′′ 23′′′ 224′ 50′′ 22′′′
7◦ 30′ 449′ 448′ 42′′ 53′′′ 448′ 42′′ 58′′′
◦
11 15 ′ 671 ′ ′ ′′
670 40 11 ′′′ 670′ 40′′ 16′′′
◦
15 00 ′ 890 ′ ′ ′′
889 45 08 ′′′ 889′ 45′′ 15′′′
◦
18 45 ′ 1105 ′ ′
1105 01 30′′ ′′′ 1105′ 01′′ 39′′′
◦
22 30 ′ 1315 ′ ′
1315 33 56 ′′ ′′ 1315′ 34′′ 07′′′
◦
26 15 ′ 1520 ′ ′
1520 28 22′′ ′′′ 1520′ 28′′ 35′′′
◦
30 00 ′ 1719 ′ ′
1718 52 10′′ ′′′ 1718′ 52′′ 24′′′
◦
33 45 ′ 1910 ′ ′
1909 54 19′′ ′′′ 1909′ 54′′ 35′′′
◦
37 30 ′ 2093 ′ ′
2092 45 46′′ ′′′ 2092′ 46′′ 03′′′
◦
41 15 ′ 2267 ′ ′
2266 38 44′′ ′′′ 2266′ 39′′ 50′′′
◦
45 00 ′ 2431 ′ ′
2430 50 54′′ ′′′ 2430′ 51′′ 15′′′
◦
48 45 ′ 2585 ′ ′
2584 37 43′′ ′′′ 2584′ 38′′ 06′′′
◦
52 30 ′ 2728 ′ ′
2727 20 29′′ ′′′ 2727′ 20′′ 52′′′
◦
56 15 ′ 2859 ′ ′
2858 22 31′′ ′′′ 2858′ 22′′ 55′′′
◦
60 00 ′ 2978 ′ ′
2977 10 09′′ ′′′ 2977′ 10′′ 34′′′
◦
63 45 ′ 3084 ′ ′
3083 12 51′′ ′′′ 3083′ 13′′ 17′′′
◦
67 30 ′ 3177 ′ ′
3176 03 23′′ ′′′ 3176′ 03′′ 50′′′
◦
71 15 ′ 3256 ′ ′
3255 17 54′′ ′′′ 3255′ 18′′ 22′′′
◦
75 00 ′ 3321 ′ ′
3320 36 02′′ ′′′ 3320′ 36′′ 30′′′
◦
78 45 ′ 3372 ′ ′
3371 41 01′′ ′′′ 3371′ 41′′ 29′′′
◦
82 30 ′ 3409 ′ ′
3408 19 42′′ ′′′ 3408′ 20′′ 11′′′
◦
86 15 ′ 3431 ′ ′
3430 22 42′′ ′′′ 3430′ 23′′ 11′′′
◦
90 00 ′ 3438 ′ ′
3437 44 19′′ ′′′ 3437′ 44′′ 48′′′
TABLE 1. Rsine-table of Āryabhat.a, Govindasvāmin and Mādhava.

5.5. Brahmagupta’s second-order interpolation formula

The Rsine table of Āryabhat.a gives only the Rsine values for the twenty-four
multiples of 225′ . The Rsines for arbitrary arc-lengths have to be found by inter-
polation only. In his Khan.d.akhādyaka (c. 665), Brahmagupta gives a second-
order interpolation formula for the computation of Rsines for arbitrary arcs. In this
Development of Calculus in India 221

work, which is in the form of a manual (karan.a) for astronomical calculations,

Brahmagupta uses a simpler Rsine-table which gives Rsines only at intervals of
15◦ or 900′:32

ga;ta;Ba;ea;gya;Ka;Nq+.k+:a;nta:=+d;l+.a;va;k+:l+.va;Da;a;t,a Za;tEa;nRa;va;Æa;Ba:=+a;‹ya;a Á

ta;dùÅ;au;a;ta;d;lM yua;ta;ea;nMa Ba;ea;gya;a;d

U ;na;a;a;Da;kM Ba;ea;gya;m,a ÁÁ

Multiply the residual arc after division by 900′ by half the difference of the tabular
Rsine difference passed over (gata-khan . d.a) and to be passed over (bhogya-
khan ′
. d.a) and divide by 900 . The result is to be added to or subtracted from half
the sum of the same tabular sine differences according as this [half-sum] is less
than or equal to the Rsine tabular difference to be passed. What results is the true
Rsine-difference to be passed over.

Let h be the basic unit of arc in terms of which the Rsine-table is constructed,
which happens to be 225′ in the case of Āryabhat.ı̄ya, and 900′ in the case of
Khan.d.akhādyaka. Let the arc for which Rsine is to be found be given by

s = jh + ε for some j = 0, 1, . . . (49)

Now R sin( j h) = B j are the tabulated Rsines. Then, a simple interpolation

(trairāśika) would yield
ε
R sin( j h + ǫ) = B j + (B j +1 − B j )
h
ε
= R sin( j h) + 1 j +1 . (50)
h
Instead of the above simple interpolation, Brahmagupta prescribes
 ε   1   ε  (1 ∼ 1 ) 
j j +1
R sin( j h + ǫ) = B j + (1 j + 1 j +1) ± . (51)
h 2 h 2

Here, the sign is chosen to be positive if 1 j < 1 j +1 , and negative if 1 j > 1 j +1

(as in the case of Rsine). So Brahmagupta’s rule is actually the second-order
interpolation formula
 ε   1   ε  (1 − 1 ) 
j j +1
R sin( j h + ε) = R sin( j h) + (1 j + 1 j +1 ) −
h 2 h 2
j +1 + 1 j ) j +1 − 1 j )
 ε  (1  ε 2 (1
= R sin( j h) + +
h 2 h 2
j +1 − 1 j )
ε εhε i (1
= R sin( j h) + 1 j +1 + −1 . (52)
h h h 2

32Khandakhādyaka of Brahmagupta, Ed. by P. C. Sengupta, Calcutta 1941, p. 151.

..
222 K. Ramasubramanian and M. D. Srinivas

6. Instantaneous velocity of a planet (tātkālika-gati)

6.1. True daily motion of a planet

In Indian Astronomy, the motion of a planet is computed by making use of two

corrections: the manda-sam . skāra which essentially corresponds to the equation
of centre and the śı̄ghra-sam. skāra which corresponds to the conversion of the
heliocentric longitudes to geocentric longitudes. The manda correction for planets
is given in terms of an epicycle of variable radius r , which varies in such a way
that
r r0
= , (53)
K R

where K is the karn.a (hypotenuse) or the (variable) distance of the planet from
the centre of the concentric and r0 is the tabulated (or mean) radius of the epicycle
in the measure of the concentric circle of radius R.

r
Γ

P0
K
∆µ α
R
M

F IGURE 4. Manda correction.

In Figure 4, C is the centre of concentric on which the mean planet P0 is lo-

cated. CU is the direction of the ucca (aphelion or apogee as the case may be).
P is the true planet which lies on the epicycle of (variable) radius r centered at
P0 , such that P0 P is parallel to CU . If M is the mean longitude of a planet, α the
Development of Calculus in India 223

longitude of the ucca, then the correction (manda-phala) 1µ is given by

r 
R sin(1µ) = R sin(M − α)
 rK 
0
= R sin(M − α). (54)
R
For small r , the left hand side is usually approximated by the arc itself. The
manda-correction is to be applied to the mean longitude M, to obtain the true or
manda-corrected longitude µ given by
r   1 
0
µ=M− R sin(M − α). (55)
R R

If n m and n u are the mean daily motions of the planet and the ucca, then the true
longitude on the next day is given by
r   1 
0
µ + n = (M + n m ) − R sin (M + n m − α − n u ). (56)
R R

The true daily motion is thus given by

r   1 
0
n = nm − [R sin{(M − α) + (n m − n u )} − R sin (M − α)]. (57)
R R

The second term in the above is the correction to mean daily motion (gati-phala).
An expression for this was given by Bhāskara I (c. 629) in Mahābhāskarı̄ya,
where he makes use of the approximation:33
 ( (n − n ) ×
R sin{(M − α) + (n m − n u )} m  u
≈ 1
− R sin(M − α) 225 Rsine-difference at (M − α).
(58)
In the above approximation, (n m − n u ) is multiplied by tabular Rsine-difference
at the 225′ arc-bit in which (the tip of the arc) (M − α) is located. Therefore,
under this approximation, as long as the anomaly (kendra), (M − α), is in the
same multiple of 225′, there will be no change in the gati-phala or the correction
to the mean velocity. This defect was noticed by Bhāskara also in his later work
Laghubhāskarı̄ya:34
A;Æa;Ba;Ša:+.pa;ta;a Bua;€ ãÉa;a;pa;Ba;a;ga;a;va;. a;a;a:=+NaH Á
e +:(

ãÉa .ja;a;va;a;na;a;mUa;na;Ba;a;va;a;dùÅ;a;sa;}Ba;va;a;t,a Á Á
.=+vea;a:=+nd;ea;(

33Mahābhāskarı̄ya of Bhāskara I, Ed. by K. S. Shukla, Lucknow 1960, verse 4.14, p. 120.

34Laghubhāskarı̄ya of Bhāskara I, Ed. by K. S. Shukla, Lucknow 1963, verses 2.14-5, p. 6.
224 K. Ramasubramanian and M. D. Srinivas

O;;va;ma;a;l+.ea;. ya;ma;a;nea;yMa .ja;a;va;a;Bua;a;€+:a;vRa;Za;a;ya;teRa Á

Whilst the Sun or the Moon moves in the [same] element of arc, there is no change
in the rate of motion (bhukti), because the Rsine-difference does not increase or
decrease; viewed thus, the rate of motion [as given above] is defective.

The correct formula for the true daily motion of a planet, employing the Rcosine
as the ‘rate of change’ of Rsine, seems to have been first given by Muñjāla (c. 932)
in his short manual Laghumānasa 35 and also by Āryabhat.a II (c. 950) in his
Mahā-siddhānta:36
È
îå Á ;a Bua;a;€+:gRa:$ya;a;Ba;€+:a k+:l+.a;a;d;P+.l+.m,a Á Á
k+:ea;a;f;P+.l;*+
The kot.iphala multiplied by the [mean] daily motion and divided by the radius
gives the minutes of the correction [to the rate of the motion].

This gives the true daily motion in the form

r   1 
0
n = n m − (n m − n u ) R cos(M − α). (59)
R R

6.2. The notion of instantaneous velocity (tātkālikagati) according to

Bhāskarācārya II

Bhāskarācārya II (c. 1150) in his Siddhāntaśiroman.i clearly distinguishes the

true daily motion from the instantaneous rate of motion. And he gives the Rcosine
correction to the mean rate of motion as the instantaneous rate of motion. He
further emphasizes the fact that the velocity is changing every instant and this is
particularly important in the case of Moon because of its rapid motion.37
å å
;a;d;na;a;nta:=+~.pa;;Ka;ga;a;nta:=M .~ya;a;d
, ga;a;taH .~.Pu +.f;a ta;tsa;ma;ya;a;nta:=+a;le ÁÁ
È
îå Á ;a mxa;d
k+:ea;f;a;P+.l;*+ u ;ke +:ndÒ+Bua;a;€+:a;~:a:$ya;e
/ / a:;d
Äâx ;ta;a k+:a;kR +:mxa;ga;a;a;d;ke +:ndÒe Á

å
ta;ya;a yua;ta;ea;na;a g{a;h;ma;Dya;Bua;a;€+:~ta;a;tk+:a;a;l+.k+:a ma;nd;pa;a:=+~.Pu+.f;a .~ya;a;t,a Á Á

.sa;ma;a;pa;a;ta;Tya;nta;sa;ma;a;pa;. a;a;l+.nMa ;a;va;Da;ea;~tua ta;tk+:a;l+ja;yEa;va yua:$ya;tea Á

ú
.sua;d *.a;a;l+.na;ma;a;dùÅ;a;ya;a ya;taH
U :=+sa:ãÁ :pra;a;ta;[a;NMa .sa;a na .sa;ma;a ma;h;tya;taH Á Á
The true daily motion of a planet is the difference between the true planets on
successive days. And it is accurate (sphut.a) over that period. The kot.iphala
(Rcosine of anomaly) is multiplied by the rate of motion of the manda-anomaly
(mr.du-kendra-bhukti) and divided by the radius. The result added or subtracted
from the mean rate of motion of the planet, depending on whether the anomaly is
in Karkyādi or Mr.gādi, gives the true instantaneous rate of motion (tātkālikı̄
manda-sphut. agati) of the planet.

35Laghumānasa of Muñjāla, Ed. by K. S. Shukla, New Delhi 1990, verse 3.4, p. 125.
36Mahāsiddhānta of Āryabhata II, Ed. by Sudhakara Dvivedi, Varanasi 1910, verse 3.15, p. 58.
.
37Siddhāntaśiroman i of Bhāskarācārya, Ed. by Muralidhara Chaturvedi, Varanasi 1981,
.
verses 2.36–8, p. 119.
Development of Calculus in India 225

In the case of the Moon, the ending moment of a tithi38 which is about to end or
the beginning time of a tithi which is about to begin, are to be computed with the
instantaneous rate of motion at the given instant of time. The beginning moment
of a tithi which is far away can be calculated with the earlier [daily] rate of
motion. This is because Moon’s rate of motion is large and varies from moment
to moment.

Here, Bhāskara explains the distinction between the true daily rate of motion
and the true instantaneous rate of motion. The former is the difference between
the true longitudes on successive days and it is accurate as the rate of motion, on
the average, for the entire period. The true instantaneous rate of motion is to be
calculated from the Rcosine of the anomaly (kot.iphala) for each relevant moment.
Thus if ωm and ωu are the rates of the motion of the mean planet and the ucca,
then ωm − ωu is the rate of motion of the anomaly, and the true instantaneous rate
of motion of the planet at any instant is given by Bhāskara to be
r   1 
0
ω = ωm + (ωm − ωu ) R cos(M − α), (60)
R R
where (M − α) is the anomaly of the planet at that instant.
Bhāskara explains the idea of the instantaneous velocity even more clearly in
his Vāsanā:39
å
A;dùÅ;a;ta;na:(õ;a;~ta;na;~.Pu+.f;g{a;h;ya;eaH A;Ea;d;a;ya;k+:ya;ea;a;dR;na;a;DRa:ja;ya;ea;va;Ra A;~ta;k+:a -
å
;a;l+.k+:ya;ea;va;Ra ya;d;nta:=M k+:l+.a;a;d;kM .sa;a .~.Pu +.f;a ga;a;taH Á A;dùÅ;a;ta;na;a;. C
õÅÅ +.~ta;nea

nyUa;nea va;kÒ +:a;ga;a;ta::℄eRa;ya;a Á ta;tsa;ma;ya;a;nta:=+a;l I+.a;ta Á ta;~ya k+:a;l+.~ya

ma;Dyea A;na;ya;a ga;tya;a ãÉa;a;l+.a;ya;tMua

g{a;h;( yua:$ya;ta I+.a;ta Á I+.yMa ;a;k+:l
ñÍ
.~TUa;l+.a ga;a;taH Á A;Ta .sUa;[ma;a ta;a;tk+:a;a;l+.k+:a k+:Tya;tea Á tua;*:+.
ç Å ga;tyUa;na;a

. a;ndÒ+ga;a;taH :ke +:ndÒ+ga;a;taH Á A;nyea;Sa;Ma g{a;h;a;Na;Ma g{a;h;ga;a;ta;=e+va :ke +:ndÒ+ga;a;taH Á

mxa;d
u ;ke +:ndÒ+k+:ea;a;f;P+.lM kx +:tva;a .tea;na :ke +:ndÒ+ga;a;ta;gRua;Nya;a Á ;a:ˆa:$ya;ya;a

Ba;a:$ya;a Á l+.b.Dea;na k+:k”+.a;Ra;a;d;ke +:ndÒe -

g{a;h ga;a;ta;yRua;€+:a k+:a;ya;Ra Á mxa;ga;a;d;Ea

å
tua .=+a;h;ta;a k+:a;ya;Ra Á O;;vMa ta;a;tk+:a;a;l+.k+:a ma;nd;pa;a:=+~.Pu+.f;a .~ya;a;t,a Á

ta;a;tk+:a;a;l+.k”+.a;a Bua;€”+.a;a . a;ndÒ+~ya ;a;va;a;Za;M -

:pra;ya;ea .ja;na;m,a Á ta;d;a;h

‘.sa;ma;a;pa;a;ta;Tya;nta;sa;ma;a;pa;. a;a;l+.na;m,a ’ -
ãÉa;ndÒ H
I+.a;ta Á ya;tk+:a;a;l k+:( ta;sma;a;t,a

ç Å +ta;ea
k+:a;l+.a;‘ va;a ga;}ya;ea va;a ya;d;a;sa;Ša;a;~ta;tTya;nta;~ta;d;a
/ / / ta;a;tk+:a;a;l+.k”+.a;a

ga;tya;a ;a;ta;a;Ta;sa;a;Da;nMa k+:tua yua:$ya;tea Á ta;Ta;a .sa;ma;a;pa;. a;a;l+.nMa . a Á ya;d;a tua

d
U :=+ta:=+a;~ta;Tya;nta;e
/ / / a d
U :=+. a;a;l+.nMa va;a . a;ndÒ+~ya ta;d;a;dùÅ;a;ya;a .~TUa;l+.ya;a k+:tua

ãÉa;ndÒ+ga;a;ta;mRa;h;tva;a;t,a :pra;a;ta;[a;NMa .sa;ma;a na

yua:$ya;tea Á .~TUa;l+.k+:a;l+.tva;a;t,a Á ya;ta;(

Ba;va;a;ta A;ta;~ta;d;TRa;ma;yMa ;a;va;Zea;Sa;ea Y;Æa;Ba;a;h;taH Á

38Tithi is the time taken by the Moon to lead the Sun exactly by 12◦ in longitude.
39Siddhāntaśiroman i, cited above, Vāsanā on 2.36–38, p. 119–20.
.
226 K. Ramasubramanian and M. D. Srinivas

A;Ta ga;a;ta;P+.l+.va;a;sa;na;a Á A;dùÅ;a;ta;na:(õ;a;~ta;na;g{a;h;ya;ea:=+nta:=M ga;a;taH Á A;ta

O;;va g{a;h;P+.l+.ya;ea:=+nta:=M ga;a;ta;P+.lM Ba;a;va;tua;ma;hR;a;ta Á A;Ta ta;tsa;a;Da;na;m,a Á

A;dùÅ;a;ta;na:(õ;a;~ta;na;ke +:ndÒ+ya;ea:=+nta:=M :ke +:ndÒ+ga;a;taH Á Bua:ja:$ya;a;k+.=+Nea ya;;‘

匁 +ea;gya -
å
Ka;NqM .tea;na .sa;a gua;Nya;a õ ;d;~:Eaò H
Za:=+a;d (225) Ba;a:$ya;a Á ta:ˆa ta;a;va;t,a

ta;a;tk+:a;a;l+.k+:Ba;ea;gya;Ka;Nq+.k+.=+Na;a;ya;a;nua;pa;a;taH Á ya;a;d ;a:ˆa:$ya;a;tua;ya;ya;a

åò
-
k+:ea;a;f .$ya;ya;a;dùÅ;aM Ba;ea;gya;Ka;NqM Za:=+a;d
õ ;d;~:a;tu a;yMa l+.Bya;tea ta;de;;ya;a

åò
;a;k+:Æa;ma;tya:ˆa k+:ea;a;f:$ya;a;ya;aH Za:=+a;d
õ ;d;~:a;a gua;Na;a;~:a:$ya;a
/ / h:=H Á

å
:P+.lM ta;a;tk+:a;a;l+.kM -
.~.Pu+.f Ba;ea;gya;Ka;NqM .tea;na :ke +:ndÒ+ga;a;ta;gRua;Na;na;a;ya;a

å
õ ;d;~:Eaò ;Ba;Ra:$ya;a Á
Za:=+a;d
åò
A:ˆa Za:=+a;d
õ ;d;~:a;Æ a;ma;ta;ya;ea;gRua;Na;k+:Ba;a:ja;k+:ya;ea;~tua;ya;tva;a;Ša;a;Zea kx +:tea :ke +:ndÒ-
ga;teaH k+:ea;a;f:$ya;a;gua;Na;a;~:a:$ya;a;h:=H
/ / .~ya;a;t,a Á :P+.l+.ma;dùÅ;a;ta;na:(õ;a;~ta;na;ke +:ndÒ -
d;ea:$yRa;ya;ea:=+nta:=M Ba;va;a;ta Á ta;tP+.l+.k+.=+Na;a;Ta .~va;pa;a:=+a;Da;na;a gua;NyMa Ba;Ma;ZEaH

(360) Ba;a:$ya;m,a Á :pUa;va ;a;k+:l gua;Na;k H k+:ea;a;f:$ya;a .sa;a ya;a;va;t,a :pa;a:=+a;Da;na;a

gua;Nya;tea Ba;Ma;ZEaH ;a;h;ya;tea ta;a;va;tk+:ea;a;f;P+.lM

ò
.ja;a;ya;ta I+.tyua;pa;pa;ŠMa k+:ea;f;a ‘ -
È
îå Á ;a mxa;d
:P+.l;*+ u ;k ’
e +:ndÒ+Bua;a;€ :a:=+tya;a;a;d Á O;;va;ma;dùÅ;a;ta;na:(õ;a;~ta;na;g{a;h;P+.l+.ya;ea:=+nta:=M

ç Å +teaH
ta;‘ :P+.lM k+:k”+.a;Ra;a;d;ke +:ndÒe g{a;h;NRa;P+.l+.~ya;a;pa;. a;a;ya;ma;a;na;tva;a;t,a tua;l+.a;d;Ea

;Da;na;P+.l+.~ya;a;pa;. a;a;ya;ma;a;na;tva;a;t,a ;Da;na;m,a Á ma;k+.=+a;d;Ea tua ;Da;na;P+.l+.~ya;a;pa;. a;a -

ya;ma;a;na;tva;a;t,a mea;Sa;a;d;a;vxa;Na;P+.l+.~ya;ea;pa;. a;a;ya;ma;a;na;tva;a;dx ;Na;m,a I+.tyua;pa;pa;Ša;m,a Á

The true daily velocity is the difference in minutes etc., between the true planets
of today and tomorrow, either at the time of sunrise, or mid-day or sunset. If to-
morrow’s longitude is smaller than that of today, then we should understand the
motion to be retrograde. It is said “over that period”. This only means that, dur-
ing that intervening period, the planet is to move with this rate [on the average].
This is only a rough or approximate rate of motion. Now we shall discuss the
instantaneous rate of motion... In this way, the manda-corrected true instanta-
neous rate of motion (tātkālikı̄ manda-parisphut. agati) is calculated. In the
case of Moon, this instantaneous rate of motion is especially useful...Because of
its largeness, the rate of motion of Moon is not the same every instant. Hence, in
the case of Moon, the special [instantaneous] rate of motion is stated.

Then, the justification for the correction to the rate of motion (gati-phala-
vāsanā). . .The rate of motion of the anomaly is the difference in the anomalies of
today and tomorrow. That should be multiplied by the [current] Rsine-difference
used in the computation of Rsines and divided by 225. Now, the following rule
of three to obtain the instantaneous Rsine-difference: If the first Rsine-difference
225 results when the Rcosine is equal to the radius, then how much is it for the
given Rcosine. In this way, the Rcosine is to be multiplied by 225 and divided
by the radius. The result is the instantaneous Rsine-difference and that should be
multiplied by the rate of motion in the anomaly and divided by 225 . . .

Thus, Bhāskara is here conceiving also of an instantaneous Rsine-difference,

though his derivation of the instantaneous velocity is somewhat obscure. These
Development of Calculus in India 227

ideas are more clearly set forth in the Āryabhat.ı̄ya-bhās.ya (c. 1502) of Nı̄lakan.t.ha
Somayājı̄ and other works of the Kerala School.

6.3. The śı̄ghra correction to the velocity and the condition for retrograde
motion

Bhāskara then goes on to derive the correct expression for the true rate of motion
as corrected by the śı̄ghra-correction. In the language of modern astronomy, the
śı̄ghra-correction converts the heliocentric longitude of the planets to the geocen-
tric longitudes. Here also, the Indian astronomers employ an epicycle, but with a
fixed radius, unlike in the case of the manda-correction.
If µ is the manda-corrected (manda-sphut.a) longitude of the planet, ζ is the
longitude of the śı̄ghrocca, and rs , the radius of the śı̄ghra-epicycle, then the cor-
rection (śı̄ghra-phala) 1σ is given by
r 
s
R sin(1σ ) = R sin (µ − ζ ), (61)
K

where (µ − ζ ) is the śı̄ghrakendra and K is the hypotenuse (śı̄ghrakarn.a) given

by

K 2 = R 2 + rs2 − 2Rrs cos (µ − ζ ). (62)

The calculation of the śı̄ghra-correction to the velocity is indeed much more dif-
ficult as the denominator in (61), which is the hypotenuse which depends on
the anomaly, also varies with time in a complex way. This has been noted by
Bhāskara who was able to obtain the correct form of the śı̄ghra-correction to the
velocity (śı̄ghra-gati-phala) in an ingenious way.40
ñÍ ú È Å
:P+.l+.Ma;Za;Ka;a;*:+.
öÅ a;nta:=+a;Za;a:*+;n îåÁ ;a dÒ+a;ë
êÁÁ a;a.*+ ëeÁ +:ndÒ+Bua;a;€ H
*: (rua;a;ta;&+.a;d
õ ;Za;ea;Dya;a Á

å
.~va;Za;a;Gra;Bua;€e H .~.Pu+.f;Kea;f;Bua;a;€ H Zea;SMa . a va;kÒ +:a ;a;va;pa:=+a;ta;Zua:;d
Äâ ;Ea ÁÁ
The Rsine of ninety degrees, less the degrees of śı̄ghra-correction for the longi-
tude (śı̄ghra-phala), should be multiplied by the rate of motion of the śı̄ghra-
anomaly (drāk-kendra-bhukti) and divided by the hypotenuse (śı̄ghra-karn . a).
This, subtracted from the rate of motion of the śı̄ghrocca, gives the true velocity
of the planet. If this is negative, the planet’s motion is retrograde.

If ω is the rate of motion of the manda-corrected planet and ωs is the rate of

motion of the śı̄ghrocca, then the rate of motion of the śı̄ghra-anomaly is (ω −ωs ),

40Siddhāntaśiroman i, cited above, verse 2.39, p. 121.

.
228 K. Ramasubramanian and M. D. Srinivas

and the true velocity of the planet ωt is given by

 
(ωs − ω)R cos(1σ )
ωt = ωs − . (63)
K

The details of the ingenious argument given by Bhāskara for deriving the correct
form (63) of the śı̄ghra-correction to the velocity has been outlined by D. Arkaso-
mayaji in his translation of Sı̄ddhāntaśiroman.i.41
Since Bhāskara’s derivation is somewhat long-winded, here we shall present
a modern derivation of the result just to demonstrate that the expression given by
Bhāskara is indeed exact.

P’
∆σ

R
∆σ
K

r θ

vs
E θ

E’

F IGURE 5a. Velocity of a planet as seen from the Earth.

In Figure 5a, S, E and P represent the positions of the Sun, Earth and an
exterior planet respectively. Let v and v s be the linear velocities of the planet and
the Earth with respect to the Sun. P P ′ and E E ′ are lines perpendicular to the line
E P joining the Earth to the planet. Let R, r represent the radii of the orbits of
the planet and the Earth (assumed to be cicular) around the Sun respectively and
K , the distance of the planet from the Earth. For an exterior planet, the śı̄ghra-
correction 1σ is given by the angle S P̂ E.

41D. Arkasomayaji, Siddhāntaśiroman i of Bhāskarācārya, Tirupati 1980, pp. 157–161.

.
Development of Calculus in India 229

If v t is the linear velocity of the planet as seen from the Earth, then the angular
velocity is given by

dθ vt
ωt = = . (64)
dt K
The magnitude of v t in terms of v and v s (for the situation depicted in Figure 5a)
is

v t = v cos 1σ + v s cos θ. (65)

Also, from the triangle S E P, the distance of the planet from the Earth—known as
karn.a, and denoted K in the figure—may be expressed as

K = R cos 1σ + r cos θ,
K − R cos 1σ
or cos θ = . (66)
r
Using (66) in (65) we have
vs
vt = v cos 1σ + (K − R cos 1σ )
r  
vs K R
= + cos 1σ v − v s
r r
R


vt vs cos 1σ v − v s r
or = + . (67)
K r K
Making use of (64) and the fact that v = Rω and v s = r ωs , the above equation
reduces to
 
(ωs − ω)R cos 1σ
ωt = ωs − ,
K

which is same as the expression given by Bhāskara (63).

Bhāskara in his Vāsanā42 explains as to why in the śı̄ghra process a different
procedure for finding the rate of motion of the planet has to be employed than the
one used in the manda process:

A:ˆa;ea;pa;pa;a:†aH Á A;dùÅ;a;ta;na:(õ;a;~ta;na;Za;a;Gra;P+.l+.ya;ea:=+nta:=M ga;teaH Za;a;Gra;P+.lM .~ya;a;t,a Á

ta;ƒa ya;Ta;a ma;a;ndM ga;a;ta;P+.lM g{a;h;P+.l+.va;d;a;na;a;tMa ta;Ta;a ya;dùÅ;a;a;na;a;ya;tea kx +:tea Y;a;pa

k+:Na;Ra;nua;pa;a;tea .sa;a;nta:=+mea;va .~ya;a;t,a Á ya;Ta;a ;Da;a;vxa;a:;d

Äâ ;de Á na ;a;h :ke +:ndÒ+ga;a;ta:ja;mea;va

42Ibid., Vāsanā on 2.39.

230 K. Ramasubramanian and M. D. Srinivas

:P+.l+.ya;ea:=+nta:=M .~ya;a;t,a ;a;k+:ntva;nya;d;a;pa A;dùÅ;a;ta;na;Bua:ja;P+.l+(õ;a;~ta;na;Bua:ja -

:P+.l+.a;nta:=e ;a:ˆa:$ya;a;gua;Nea Y;dùÅ;a;ta;na;k+:NRa;&+.tea ya;a;dx;ZMa :P+.lM na ta;a;dx;ZMa (õ;a;~ta;na -
k+:NRa;&+.tea Á .~va;pa;a;nta:=e Y;a;pa k+:NeRa Ba;a:$ya;~ya ba;hu;tva;a;d, ë+:nta:=M
ba;– .~ya;a;a;d-

tyea;ta;d;a;na;ya;nMa ;a;h;tva;a;nya;t,a ma;h;a;ma;a;ta;ma;a;;‘

匁 H
k+:a;pa;ta;m,
/// a Á ta;dùÅ;a;Ta;a...
Here is the justification. The śı̄ghra-correction to the rate of motion is the dif-
ference between the śı̄ghra-phalas of today and tomorrow. If that is derived in
the same way as the manda-correction to the rate of motion, the result will be
incorrect even if it were to be divided by the hypotenuse (śı̄ghra-karn . a)... The
difference is not just due to the change in the anomaly [which is the argument
of the Rsine] but also otherwise... The result of dividing by today’s hypotenuse
is different from that of dividing by that of tomorrow. Even if the hypotenuses
turn out to differ by small amount, the quantities they divide are large and thus
a large difference could result. Hence, this way of approach [which was adopted
in the case of manda-correction to the rate of the motion] has been forsaken and
another has been devised by the great intellects. That is as follows...

6.4. The equation of centre is extremum when the velocity correction van-
ishes

Later, in the Golādhyāya of Siddhāntaśiroman.i, Bhāskara considers the situa-

tion when the correction to the velocity (gati-phala) vanishes:43
k+:[ya;a;ma;Dya;ga;a;ta;yRa;g{ea;Ka;a;pra;a;ta;vxa:†a;sMa;pa;a;tea Á

å
ma;DyEa;va ga;a;taH .~.pa;;a :pa:=M :P+.lM ta:ˆa Kea;f;~ya Á Á
Where the [North-South] line perpendicular to the [East-West] line of apsides
through the centre of the concentric meets the eccentric, there the mean velocity
itself is true and the equation of centre is extremum.

In his Vāsanā, Bhāskara explains this correlation between vanishing of the

velocity correction and the extrema of the correction to the planetary longitude:44
k+:[ya;a;vxa:†a;ma;Dyea ya;a ;a;ta;yRa;g{ea;Ka;a ta;~ya;aH :pra;a;ta;vxa:†a;~ya . a yaH .sMa;pa;a;ta;~ta:ˆa

å
ma;DyEa;va ga;a;taH .~.pa;;a Á ga;a;ta;P+.l+.a;Ba;a;va;a;t,a Á ;a;kM +:. a ta:ˆa g{a;h;~ya :pa:=+mMa :P+.lM

.~ya;a;t,a Á ya:ˆa g{a;h;~ya :pa:=+mMa :P+.lM ta:ˆEa;va ga;a;ta;P+.l+.a;Ba;a;vea;na Ba;a;va;ta;v.ya;m,a Á

ya;ta;ea Y;dùÅ;a;ta;na:(õ;a;~ta;na;g{a;h;ya;ea:=+nta:=M ga;a;taH Á :P+.l+.ya;ea:=+nta:=M ga;a;ta;P+.l+.m,a Á

g{a;h,:~ya ga;tea;va;Ra :P+.l+.a;Ba;a;va;~Ta;a;na;mea;va ;Da;na;NRa;sa;a;nDaH

// Á ya;t,a :pua;na;lR+.Œ+:ea;€M

å
‘ma;DyEa;va ga;a;taH .~.pa;;a vxa:†a;d
õ ;ya;ya;ea;ga;gea ’
dùÅ;au;. a:=e I+.a;ta ta;d;sa;t,a Á na ;a;h

vxa:†a;d
õ ;ya;ya;ea;gea g{a;h;~ya :pa:=+mMa :P+.l+.m,a Á
The mean rate of motion itself is exact at the points where the line perpendicular
[to the line of apsides], at the middle of the concentric circle, meets the eccentric

43Siddhāntaśiroman i, cited above, Golādhyāya 4.39, p. 393.

.
44Ibid., Vāsanā on Golādhyāya 4.39.
Development of Calculus in India 231

circle. Because, there is no correction to the rate of motion [at those points]. Also,
because there the equation of centre [or correction to the planetary longitudes] is
extreme. Wherever the equation of centre is maximum, there the correction to the
velocity should be absent. Because, the rate of motion is the difference between
the planetary longitudes of today and tomorrow. The correction to the velocity is
the difference between the equations of centre. The place where the correction to
the velocity vanishes, there is a change over from positive to the negative. And,
what has been stated by Lalla, “the mean rate of motion is itself true when the
planet is on the intersection of the two circles [concentric and eccentric]”, that is
incorrect. The planet does not have maximum equation of centre at the confluence
of the two circles.

C’

N S
C

F IGURE 5b. Equation of centre is extremum where the correc-

tion to velocity vanishes.

Bhāskara explains that when the anomaly is ninety degrees, or the mean planet
is at N along the line C N perpendicular to the line of apsides C E (see Figure
5b), the equation of centre is maximum. It is precisely then that the correction to
the velocity vanishes, as it changes sign from positive to negative. It is incorrect
to state (as Lalla did in his Śis.yadhı̄vr.ddhida-tantra) that the correction to the
velocity is zero at the point where the concentric and eccentric meet.

7. Surface area and volume of a sphere

In Āryabhat.ı̄ya (Golapāda 7), the volume of a sphere has been incorrectly es-
timated as the product of the area of a great circle by its square-root. Śrı̄dhara
232 K. Ramasubramanian and M. D. Srinivas

(c. 750) seems to have given the correct expression for the volume of a sphere
(Triśatikā 56), though his estimate of π is fairly off the mark. Bhāskarācārya
(c. 1150) has given the correct relation between the diameter, the surface area and
the volume of a sphere in his Lı̄lāvatı̄:45
vxa:†a;[ea:ˆea :pa;a:=+a;Da;gua;a;Na;ta;v.ya;a;sa;pa;a;dH :P+.lM ya;t,a

[ua;NNMa :vea;dE:+:pa;a:= :pa;a:=+taH k+:nd

u ;k+:~yea;va .ja;a;l+.m,a Á

È
ga;ea;l+.~yEa;vMa ta;d;a;pa . a :P+.lM :pxa;Ž;jMa v.ya;a;sa;a;na.*+
îåM Á

ñ
:Sa;a:*.
Âå+Å :BRa;€M Ba;va;a;ta ;a;na;ya;tMa ga;ea;l+.ga;BeRa ;Ga;na;a;K.ya;m,a Á Á
In a circle, the circumference multiplied by one-fourth the diameter is the area,
which, multiplied by four, is its surface area going around it like a net around
a ball. This [surface area] multiplied by the diameter and divided by six is the
volume of the sphere.

The surface area and volume of a sphere have been discussed in greater detail
in the Siddhāntaśiroman.i (Golādhyāya 2.53-61), where Bhāskara has also pre-
sented the upapatti or justification for the results in his commentary Vāsanā. As
regards the surface area of the sphere, Bhāskara argues as follows:46

A;Ta ba;a;l+.a;va;ba;ea;Da;a;Ta ga;ea;l+.~ya;ea;pa;a:= d;ZRa;yea;t,a Á BUa;ga;ea;lM mxa;Nma;yMa d;a:+:ma;yMa

va;a kx +:tva;a tMa . a;kÒ +:k+:l+.a;pa;a:=+a;DMa (21600) :pra;k+:pya ta;~ya ma;~ta;ke ;a;ba;ndu M

åò ñÍ
kx +:tva;a ta;sma;a;a;d
â õâ ;nd;ea;ga;eRa;l+.Sa;NNa;va;a;ta;Ba;a;gea;na Za:=+a;d É +;a
õ ;d;~:a;sa;*
Ë ùe ;na
Á
(225)
;Da;nUa:+.pea;NEa;va vxa:†a:=e+Ka;a;mua;tpa;a;d;yea;t,a Á :pua;na;~ta;sma;a;de;va ;a;ba;nd;eaH .tea;nEa;va ;a;d
õ ;gua;Na -
.sUa:ˆea;Na;a;nya;Ma ;a:ˆa;gua;Nea;na;a;nya;a;mea;vMa . a;tua;a;va;Za;a;ta;gua;NMa ya;a;va;ƒa;tua;a;va;Za;a;ta;vRxa:†a;a;a;na

Ba;va;a;nta
// Á O;;Sa;Ma vxa:†a;a;na;Ma Za:=+nea:ˆa;ba;a;h;vaH (225) I+.tya;a;d;a;a;na .$ya;a;Da;Ra;a;na

v.ya;a;sa;a;Da;Ra;a;na .~yuaH Á õx :†a;pra;ma;a;Na;a;a;na Á

.tea;Bya;ea Y;nua;pa;a;ta;a;d ta:ˆa ta;a;va;d;ntya -
vxa:†a;~ya ma;a;nMa . a;kÒ +:k+:l+.aH (21600) Á ta;~ya v.ya;a;sa;a;Da ;a:ˆa:$ya;a 3438 Á

.$ya;a;Da;Ra;a;na . a;kÒ +:k+:l+.a;gua;Na;a;a;na ;a:ˆa:$ya;a;Ba;€+:a;a;na vxa:†a;ma;a;na;a;a;na .ja;a;ya;ntea Á

d õR ;ya;ea;vRxa:†a;ya;ea;mRa;Dya
õ ;ya;ea;d O;;kE +:kM va;l+.ya;a;k+:a:=M [ea:ˆa;m,a Á ta;a;a;na . a;tua;a;va;Za;a;taH Á

ba;hu:$ya;a;pa;[ea ba;hU;a;na .~yuaH Á ta:ˆa ma;h;d;Da;ea;vxa:†Ma BUa;Æa;ma;mua;pa;a:=+ta;nMa l+.Gua;mua;KMa

åò
Za:=+a;d
õ ;d;~:a;Æ a;ma;tMa l+.}bMa :pra;k+:pya l+.}ba;gua;NMa ku +:mua;Ka;ya;ea;ga;a;DRa;Æa;ma;tyea;vMa :pxa;Ta;k,

:pxa;Ta;k, :P+.l+.a;a;na Á .tea;Sa;Ma :P+.l+.a;na;Ma ya;ea;ga;ea ga;ea;l+.a;DRa;pxa;Ž+P+.l+.m,a Á ta;a;‰

ë +gua;NMa

.sa;k+:l+.ga;ea;l+.pxa;Ž+P+.l+.m,a Á ta;d
õùÅ;a;a;sa;pa;a:=+a;Da;Ga;a;ta;tua;ya;mea;va .~ya;a;t,a Á
In order to make the point clear to a beginner, the teacher should demonstrate it
on the surface of a sphere. Make a model of the earth in clay or wood and let its
circumference be 21, 600 minutes. From the point at the top of the sphere at an
arc-distance of 1/96th of the circumference, i.e., 225′ , draw a circle. Similarly
draw circles with twice, thrice,... twenty-four times 225′ [as the arc-distances] so

45Lı̄lāvatı̄, cited above (fn. 5), verse 203, p. 79–80.

46Siddhāntaśiroman i, cited above, Vāsanā on Golādhyāya 2.57, p. 362.
.
Development of Calculus in India 233

that there will be twenty-four circles. These circles will have as there radii Rsines
starting from 225′ . The measure [circumference] of the circle will be in propor-
tion to these radii. Here, the last circle has a circumference 21, 600′ and its radius
is 3, 438′ . The Rsines multiplied by 21, 600 and divided by the radius [3, 438]
will give the [circumference] measure of the circles. Between any two circles,
there is an annular region and there are twenty-four of them. If more [than 24]
Rsines are used, then there will be as many regions. In each figure [if it is cut and
spread across as a trapezium] the larger lower circle may be taken as the base and
the smaller upper circle as the face and 225′ as the altitude and the area calculated
by the usual rule: [Area is] altitude multiplied by half the sum of the base and
face. The sum of all these areas is the area of half the sphere. Twice that will be
the surface area of the entire sphere. That will always be equal to the product of
the diameter and the circumference.

Here Bhāskara is taking the circumference to be C = 21, 600′, and the corre-
sponding radius is approximated as R ≈ 3, 438′. As shown in Figure 6, circles
are drawn parallel to the equator of the sphere, each separated in latitudes by 225′ .
This divides the northern hemisphere into 24 strips, each of which can be cut and
spread across as a trapezium. If we denote the 24 tabulated Rsines by B1 , B2 ,...
B24 , then the area A j of j -th trapezium will be
 
C (B j + B j +1)
Aj = 225.
R 2

Therefore, the surface area S of the sphere is estimated to be

   
C B24
S=2 B1 + B2 + ...B23 + (225). (68)
R 2

Now, Bhāskara states that the right hand side of the above equation reduces to
2C R. This can be checked by using Bhāskara’s Rsine-table. Bhāskara himself
has done the summation of the Rsines in his Vāsanā on the succeeding verses,47
where he gives another method of derivation of the area of the sphere, by cutting
the surface of the sphere into lunes. In that context, he computes the sum
   
B24 R
B1 + B2 .... + B23 + = B1 + B2 .... + B23 + B24 −
2 2
≈ 54233 − 1719 = 52514. (69)

47Siddhāntaśiroman i, cited above, Vāsanā on Golādhyāya 2.58–61, p. 364.

.
234 K. Ramasubramanian and M. D. Srinivas

E F
C D
225
A B
225

F IGURE 6. Surface area of a sphere.

Thus, according to Bhāskara’s Rsine table

  
B24
B1 + B2 + .... + B23 + (225) = 52514 × (225)
2
= 11815650
≈ (3437.39)2. (70)

Taking this as R 2 = (3438)2, we obtain the surface area of the sphere to be 48

 
C
S=2 R 2 = 2C R. (71)
R

Of course, the grossness of the result (70) is due to the fact that the quadrant of the
circumference was divided into only 24 bits. Bhāskara also mentions that we may
consider dividing the circumference into many more arc-bits, instead of the usual
24 divisions which are made for computing Rsine-tables. This is the approach
taken in Yuktibhās.ā, where the circumference of the circle is divided into a large

48As has been remarked by one of the reviewers, it is indeed intriguing the Bhāskara chose to
sum the tabular Rsines numerically, instead of making use of the relation between Rsines and Rcosine-
differences which was well known since the time of Āryabhat.a. In fact, the proof given in Yuk-
tibhās.ā (cited below in fn. 49) makes use of the relation between the Rsines and the second order
Rsine-differences to estimate this sum.
Development of Calculus in India 235

number, n, of equal arc-bits. If 1 is the Rsine of each arc-bit, the surface area is
estimated to be
 
C
S=2 (B1 + B2 + ....Bn )(1). (72)
R

Then it is shown that in the limit of large n,

(B1 + B2 + ....Bn )(1) ≈ R 2 , (73)

which leads to the result 2C R for the surface area.49

As regards the volume of a sphere, Bhāskara’s justification is much simpler:50
ga;ea;l+.pxa;Ž+P+.l+.~ya v.ya;a;sa;gua;a;Na;ta;~ya :Sa;qM+.Za;ea ;Ga;na;P+.lM .~ya;a;t,a Á
ñÍ
A:ˆa;ea;pa;pa;a:†aH Á :pxa;Ž+P+.l+.sa;*
É +;a;a;
Ë ù a;na .+.pa;ba;a;hU;a;na v.ya;a;sa;a;DRa;tua;ya;vea;Da;a;a;na
Á

.sUa;. a;a;Ka;a;ta;a;a;na ga;ea;l+.pxa;Že :pra;k+:pya;a;a;na Á .sUa;. ya;g{a;a;Na;Ma ga;ea;l+.ga;BeRa .sMa;pa;a;taH Á

O;;vMa .sUa;. a;a;P+.l+.a;na;Ma ya;ea;ga;ea ;Ga;na;P+.l+.Æa;ma;tyua;pa;pa;Ša;m,a Á ya;t,a :pua;naH

[ea:ˆa;P+.l+.mUa;le+.na [ea:ˆa;P+.lM gua;a;Na;tMa ;Ga;na;P+.lM .~ya;a;a;d;a;ta ta;t,a :pra;a;yaH

. a;tua;veRa;d;a;. a;a;yRaH :pa:=+ma;ta;mua;pa;nya;~ta;va;a;n,a Á

The surface area of a sphere multiplied by its diameter and divided by six is its
volume. Here is the justification. As many pyramids as there are units in the sur-
face area with bases of unit side and altitude equal to the semi-diameter should
be imagined on the surface of the sphere. The apices of the pyramids meet at the
centre of the sphere. Then the volume of the sphere is the sum of the volumes
of the pyramids and thus our result is justified. The view that the volume is the
product of the area times its own root, is perhaps an alien view (paramata) that
has been presented by Caturavedācārya [Pr.thūdakasvāmin].

We may note that it is the Āryabhat.ı̄ya rule which is referred to as paramata

in the above passage. Bhāskara’s derivation of the volume of a sphere is similar
to that of the area of a circle by approximating it as the sum of the areas of a
large numbers of triangles with their vertices at the centre, which is actually the
proof given in Yuktibhās.ā. In the case of the volume of a sphere, Yuktibhās.ā,
however, gives the more “standard” derivation, where the sphere is divided into
a large number of slices and the volume is found as the sum of the volumes of
the slices—which ultimately involves estimating the sum of squares of natural
numbers (varga-saṅkalita), 12 + 22 + 32 + ... + n 2 , for large n.51

49Ganita-yukti-bhāsā, cited above, Section 7.18, pp. 140–42, 261–63, 465–67. In modern ter-
. .
Rπ
minology, this amounts to the evaluation of the integral 02 R sin θ Rdθ = R 2 .
50Siddhāntaśiroman i, cited above, Vāsanā on Golādhyāya 2.61, p. 364.
.
51Ganita-yukti-bhāsā, cited above, Section 7.19, pp. 142–45, 263–66, 468–70.
. .
236 K. Ramasubramanian and M. D. Srinivas

PART II : WORK OF THE K ERALA S CHOOL

Mādhava to Śaṅkara Vāriyar (c. 1350–1550 CE)

8. Kerala School of Astronomy

The Kerala School of Astronomy in the medieval period, pioneered by Mādhava

(c. 1340–1420) of Saṅgamagrāma, extended well into the 19th century as exem-
plified in the work of Śaṅkaravarman (c. 1830), Rājā of Kad.attanad.u. Only
a couple of astronomical works of Mādhava (Ven.vāroha and Sphut.acandrāpti)
seem to be extant now. Most of his celebrated mathematical discoveries—such as
the infinite series for π and the sine and cosine functions—are available only in
the form of citations in later works.
Mādhava’s disciple Parameśvara (c. 1380–1460) of Vat.asseri, is reputed to
have carried out detailed observations for over 50 years. A large number of orig-
inal works and commentaries written by him have been published. However, his
most important work on mathematics, the commentary Vivaran.a on Lı̄lāvatı̄ of
Bhāskara II, is yet to be published.
Nı̄lakan.t.ha Somayājı̄ (c. 1444–1550) of Kun.d.agrāma, disciple of
Parameśvara’s son Dāmodara (c. 1410–1520), is the most celebrated mem-
ber of Kerala School after Mādhava. Nı̄lakan.t.ha has cited several important
results of Mādhava in his various works, the most prominent of them being
Tantrasaṅgraha (c. 1500) and Āryabhat.ı̄ya-bhās. ya. In the latter work, while
commenting on Gan.itapāda of Āryabhat.ı̄ya, Nı̄lakan.t.ha has also dealt exten-
sively with many important mathematical issues.
However, the most detailed exposition of the work of the Kerala School, start-
ing from Mādhava, and including the seminal contributions of Parameśvara,
Dāmodara and Nı̄lakan.t.ha, is to be found in the famous Malayalam work Yuk-
tibhās.ā (c. 1530) of Jyes.t.hadeva (c. 1500–1610). Jyes.t.hadeva was also a dis-
ciple of Dāmodara but junior to Nı̄lakan.t.ha. The direct lineage from Mādhava
continued at least till Acyuta Piśārat.i (c. 1550–1621), a disciple of Jyes.t.hadeva,
who wrote many important works and a couple of commentaries in Malayalam
also.
At the very beginning of Yuktibhās.ā, Jyes.t.hadeva states that he intends to
present the rationale of the mathematical and astronomical results and procedures
which are to be found in Tantrasaṅgraha of Nı̄lakan.t.ha. Yuktibhās.ā, compris-
ing 15 chapters, is naturally divided into two parts, Mathematics and Astronomy.
Topics in astronomy proper, so to say, are taken up for consideration only from the
eighth chapter onwards, starting with a discussion on mean and true planets.
Development of Calculus in India 237

The first seven chapters of Yuktibhās.ā are in fact in the nature of an inde-
pendent treatise on mathematics and deal with various topics which are of rele-
vance to astronomy. It is here that one finds detailed demonstrations of the re-
sults of Mādhava such as the infinite series for π, the arc-tangent, sine and the
cosine functions, the estimation of correction terms and their use in the genera-
tion of faster convergent series. Demonstrations are also provided for the classi-
cal results of Āryabhat.a (c. 499) on kut..tākāra (linear indeterminate equations),
of Brahmagupta (c. 628) on the diagonals and the area of a cyclic quadrilat-
eral, and of Bhāskara II (c. 1150) on the surface area and volume of a sphere.
Many of these rationales have also been presented mostly in the form of San-
skrit verses by Śaṅkara Vāriyar (c. 1500–1560) of Tr.kkut.aveli in his commen-
taries Kriyākramakarı̄ (c. 1535) on Lı̄lāvatı̄ of Bhāskara II and Yukti-dı̄pikā on
Tantrasaṅgraha of Nı̄lakan.t.ha. In fact, Śaṅkara Vāriyar ends his commentary
on the first chapter of Tantrasaṅgraha with the acknowledgement:52

I+.tyea;Sa;a :pa:=+kÒ +:ea;q+.a;va;a;sa;a;d

õ :ja;va:=+sa;ma;a;a:=+ta;ea ya;ea Y;TRaH Á
ñÍ
.sa tua ta:n:a;sa;*:
çÅÅ" +.h;~ya :pra;Ta;mea Y;Dya;a;yea ma;ya;a k+:a;Ta;taH Á Á

Whatever has been the meaning as expounded by the noble dvija of Parakrod. a
[Jyes.t.hadeva] the same has now been stated by me for the first chapter of
Tantrasaṅgraha.

In the following sections we shall present an overview of the contribution of the

Kerala School to the development of calculus (during the period 1350–1500), fol-
lowing essentially the exposition given in Yuktibhās.ā. In order to indicate some
of the concepts and methods developed by the Kerala astronomers, we first take
up the issue of irrationality of π and the summation of infinite geometric series
as discussed by Nı̄lakan.t.ha Somayājı̄ in his Āryabhat.ı̄ya-bhās.ya. We then con-
sider the derivation of binomial series expansion and the estimation of the sum of
integral powers of integers, 1k + 2k + . . . + n k for large n, as presented in Yuk-
tibhās.ā. These results constitute the basis for the derivation of the infinite series
for π4 due to Mādhava. We shall outline this as also the very interesting work
of Mādhava on the estimation of the end-correction terms and the transforma-
tion of the π-series to achieve faster convergence. Finally we shall summarize the
derivation of the infinite series for Rsine and Rcosine due to Mādhava.
In the final section, we shall deal with another topic which has a bearing on
calculus, but is not dealt with in Yuktibhās.ā, namely the evaluation of the in-
stantaneous velocity of a planet. Here, we shall present the result of Dāmodara,
as cited by Nı̄lakan.t.ha, on the instantaneous velocity of a planet which involves

52Tantrasaṅgraha of Nı̄lakantha Somayājı̄, Ed. with Yukti-dı̄pikā of Śaṅkara Vāriyar by

..
K. V. Sarma, Hoshiarpur 1977, p. 77. The same acknowledgement appears at the end of the subsequent
chapters also.
238 K. Ramasubramanian and M. D. Srinivas

the derivative of the arc-sine function. There are indeed many works and com-
mentaries by later astronomers of the Kerala School, whose mathematical contri-
butions are yet to be studied in detail. We shall here cite only one result due to
Acyuta Pis.ārat.i (c. 1550–1621), a disciple of Jyes.t.hadeva, on the instantaneous
velocity of a planet, which involves the evaluation of the derivative of the ratio of
two functions.

9. Nı̄lakan.t.ha’s discussion of irrationality of π

In the context of discussing the procedure for finding the approximate square root
of a non-square number, by multiplying it by a large square number (the method
given in Triśatikā of Śrı̄dhara referred to earlier in Section 3.3), Nı̄lakan.t.ha
observes in his Āryabhat.ı̄ya-bhās. ya:53

O;;vMa kx +:ta;ea Y;pya;a;sa;Ša;mea;va mUa;lM .~ya;a;t,a Á na :pua;naH k+.=+Na;a;mUa;l+.~ya ta:†va;taH

:pa;a:=+. Ce+.dH k+:tua Za;k”+.a I+.tya;Æa;Ba;pra;a;yaH Á ta;ta;ea ya;a;va;d;pea;[a;m,a AM;Za;a;na;Ma

.sUa;[ma;tva;a;ya ma;h;ta;a va;geRa;Na h;na;na;mua;€+:m,a Á

Even if we were to proceed this way, the square root obtained will only be ap-
proximate. The idea [that is being conveyed] is, that it is actually not possible
to exactly de-limit (paricchedah. ) the square root of a non-square number. Pre-
cisely for this reason, multiplication by a large square was stated (recommended)
in order to get as much accuracy as desired.

Regarding the choice of the large number that must be made, it is mentioned
that one may choose any number—as large a number as possible—that gives the
desired accuracy.54

ta:ˆa ya;a;va;ta;a ma;h;ta;a gua;Na;nea bua:;d

Äâ ;a;va;lM+.Ba;a;vaH .~ya;a;t,a ta;a;va;ta;a h;nya;a;t,a Á
ëÅ
ma;h:†va;~ya A;a;pea;Æa;[a;k+:tva;a;t,a ë
*:+:
Á a;. a;d;a;pa na :pa;a:=+sa;ma;a;a;‹a;a:=+a;ta Ba;a;vaH Á
You can multiply by whichever large number you want up to your satisfaction
(buddhāvalam . bhāvah.). Since largeness is a relative notion, it may be under-
stood that the process is an unending one.

In this context, Nı̄lakan.t.ha cites the verse given by Āryabhat.a specifying the
ratio of the circumference to the diameter of a circle (value of π), particularly
drawing our attention to the fact that Āryabhat.a refers to this value as “approxi-
mate”.55

53Āryabhatı̄ya of Āryabhata, Ed. with Āryabhatı̄ya-bhāsya of Nı̄lakantha Somayājı̄ by

. . . . ..
K. Sām-baśiva Śāstrı̄, Trivandrum Sanskrit Series 101, Trivandrum 1930, comm. on Gan
. itapāda
4, p. 14.
54Ibid.
55Ibid.
Development of Calculus in India 239

va;[ya;a;ta . a – ‘A;yua;ta;dõ;ya;a;va;Sk+:}Ba;~ya A;a;sa;Ša;ea vxa:†a;pa;a:=+Na;a;hH ’ I+.a;ta Á

å
ta:ˆa v.ya;a;sea;na :pa;a:=+a;Da::℄a;a;nea A;nua;ma;a;na;pa:=+}.pa:=+a .~ya;a;t,a Á ta;tk+:mRa;Nya;a;pa

mUa;l+.a;k+.=+Na;~ya A;nta;Ba;Ra;va;a;de;va ta;~ya A;a;sa;Ša;tva;m,a Á ta;tsa;va ta;d;va;sa:=e

O;;va :pra;a;ta;pa;a;d;a;ya;Sya;a;maH Á
As will be stated [by the author himself] – ‘this is [only] an approximate measure
of the circumference of the circle whose diameter is twenty-thousand.’ In finding
the circumference from the diameter, a series of inferences are involved. The
approximate nature of this also stems from the fact that it involves finding square
roots. All this will be explained later at the appropriate context.

Addressing the issue—later in his commentary, as promised earlier—while dis-

cussing the value of π Nı̄lakan.t.ha observes:56
ñÍ
:pa;a:=+a;Da;v.ya;a;sa;ya;eaH .sa;* ËÉ +;a;a;sa;}ba;nDaH
ù
Á
,
:pra;d;a;ZRa;taH Á . . . A;a;sa;ŠaH A;a;sa;Ša;ta;yEa;va

ñÍ ñÍ
A;yua;ta d - õ ;ya;sa;*ÉË+;a;
ù
Á
a;va;Sk+:}Ba;~ya I+. y a
M É +;a;a
:pa;a:=+a;Da;sa;*
Ë ù
Á
o+.€+:a Á k u +:taH :pua;naH

ñÍ
va;a;~ta;va;Ma .sa;* ËÉ +;a;a;m,
ù
Á
a o+.tsxa:$ya A;a;sa;ŠEa;va I+.h;ea;€+:a ?
o+.. ya;tea Á ta;~ya;a

va;€u+:ma;Za;k”+.a;tva;a;t,a Á ku +:taH ?
The relation between the circumference and the diameter has been presented.
. . . Approximate: This value (62,832) has been stated as only an approximation to
the circumference of a circle having a diameter of 20,000. “Why then has an ap-
proximate value been mentioned here instead of the actual value?” It is explained
[as follows]. Because it (the exact value) cannot be expressed. Why?

Explaining as to why the exact value cannot be presented, Nı̄lakan.t.ha contin-

ues:57
yea;na ma;a;nea;na ma;a;ya;ma;a;na;ea v.ya;a;saH ;a;na:=+va;ya;vaH .~ya;a;t,a , .tea;nEa;va ma;a;ya;ma;a;naH

:pa;a:=+a;DaH :pua;naH .sa;a;va;ya;va O;;va .~ya;a;t,a Á yea;na . a ma;a;ya;ma;a;naH :pa;a:=+a;DaH

;a;na:=+va;ya;vaH .tea;nEa;va ma;a;ya;ma;a;na;ea v.ya;a;sa;ea Y;a;pa .sa;a;va;ya;va O;;va; I+.a;ta

ëÅ
O;;ke +:nEa;va ma;a;nea;na ma;a;ya;ma;a;na;ya;eaH o+.Ba;ya;eaH ëÁ
*:+:a; a;pa na ;a;na:=+va;ya;va;tvMa

.~ya;a;t,a Á ma;h;a;nta;m,a A;Dva;a;nMa ga;tva;a;a;pa A;pa;a;va;ya;va;tva;m,a O;;va l+.Bya;m,a Á

ëÅ
;a;na:=+va;ya;va;tvMa tua *:+:a;
ëÁ a;pa na l+.Bya;m,a I+.a;ta Ba;a;vaH Á

Given a certain unit of measurement (māna) in terms of which the diameter

(vyāsa) specified [is just an integer and] has no [fractional] part (niravayava),
the same measure when employed to specify the circumference (paridhi) will
certainly have a [fractional] part (sāvayava) [and cannot be just an integer].
Again if in terms of certain [other] measure the circumference has no [fractional]
part, then employing the same measure the diameter will certainly have a [frac-
tional] part [and cannot be an integer]. Thus when both [the diameter and the
circumference] are measured by the same unit, they cannot both be specified [as

56Ibid., comm. on Ganitapāda 10, p. 41.

.
57Ibid., pp. 41–42.
240 K. Ramasubramanian and M. D. Srinivas

integers] without [fractional] parts. Even if you go a long way (i.e., keep on re-
ducing the measure of the unit employed), the fractional part [in specifying one
of them] will only become very small. A situation in which there will be no [frac-
tional] part (i.e, both the diameter and circumference can be specified in terms of
integers) is impossible, and this is the import [of the expression āsanna].

Evidently, what Nı̄lakan.t.ha is trying to explain here is the incommensurability

of the circumference and the diameter of a circle. Particularly, the last line of
the above quote—where Nı̄lakan.t.ha clearly mentions that, however small you
may choose your unit of measurement to be, the two quantities will never become
commensurate—is noteworthy.

10. Nı̄lakan.t.ha’s discussion of the sum of an infinite geometric series

In his Āryabhat.ı̄ya-bhās. ya, while deriving an interesting approximation for the

arc of a circle in terms of the jyā (Rsine) and the śara (Rversine), Nı̄lakan.t.ha
presents a detailed demonstration of how to sum an infinite geometric series. The
specific geometric series that arises in this context is:

 2  n
1 1 1 1
+ + ...+ +... = .
4 4 4 3

We shall now present an outline of Nı̄lakan.t.ha’s argument that gives an idea

of how the notion of limit was understood in the Indian mathematical tradition.

10.1. Nı̄lakan.t.ha’s approximate formula for the arc in terms of jyā and śara

In Figure 7, AB is the arc whose length (assumed to be small) is to be determined

in terms of the chord lengths AD and B D. In the Indian mathematical literature,
the arc AB, the semi-chord AD and the segment B D are referred to as the cāpa,
jyārdha and śara respectively. As can be easily seen from the figure, this termi-
nology arises from the fact that these geometrical objects look like a bow, a string
and an arrow respectively. Denoting them by c, j , and s, the expression for the arc
given by Nı̄lakan.t.ha may be written as:

s 
1
c≈ 1+ s 2 + j 2. (74)
3
Development of Calculus in India 241

B
E
F
A D C

F IGURE 7. Arc-length in terms of jyā and śara.

Nı̄lakan.t.ha’s proof of the above equation has been discussed in detail by Saras-
vati Amma.58 It may also be mentioned that the above approximation actually
does not form a part of the text Āryabhat.ı̄ya; but nevertheless it is introduced by
Nı̄lakan.t.ha while commenting upon a verse in Āryabhat.ı̄ya that gives the arc
in terms of the chords in a circle.59 The verse that succinctly presents the above
equation (74) goes as follows:60
.sa:ˆyMa;Za;a;a;d;Sua;va;ga;Ra;t,a .$ya;a;va;ga;Ra;Q.•a;a;t,a :pa;dM ;Da;nuaH :pra;a;yaH Á
The arc is nearly (prāyah.) equal to the square root of the sum of the square of
the śara added to one-thirds of it, and the square of the jyā.

The proof of (74) given by Nı̄lakan.t.ha involves:

(1) Repeated halving of the arc-bit, cāpa c to get c1 . . . ci . . . .
(2) Finding the corresponding semi-chords, jyā ( ji ) and the Rversines,
śara (si ).
(3) Estimating the difference between the cāpa and jyā at each step.
If δi denotes the difference between the cāpa and jyā at the i t h step, that is,

δi = ci − ji ,

58T. A. Sarasvati Amma, cited above (fn. 21), pp. 179–182.

59vxa:†ea (Āryabhat.ı̄ya, Gan
Za:=+sMa;va;gRaH A;DRa:$ya;a;va;gRaH .sa Ka;lu ;Da;nua;Sa;eaH Á . itapāda, verse 17).
60Āryabhatı̄ya-bhāsya on Āryabhatı̄ya, cited above (fn. 50), comm. on Ganitapāda 12 and
. . . .
17, p. 63 and p. 110. That the verse cited is from another work of his, namely Golasāra, has been
alluded to by Nı̄lakan.t.ha in both the instances of citation.
242 K. Ramasubramanian and M. D. Srinivas

then it is seen that this difference decreases as the size of the cāpa decreases.
Having made this observation, Nı̄lakan.t.ha proceeds with the argument that
• Generating successive values of the ji -s and si -s is an ‘unending’ process
(na kvacidapi paryavasyati) as one can keep on dividing the cāpa into half
ad infinitum (ānantyāt vibhāgasya).
• It would therefore be appropriate to proceed up to a stage where the differ-
ence δi becomes negligible (śūnyaprāya) and make an ‘intelligent approxi-
mation’, to obtain the value of the difference between c and j approximately.
The original passage in Āryabhat.ı̄ya-bhās.ya which presents the above argument
reads as follows:61
ta:ˆa .$ya;a;. a;a;pa;ya;ea:=+nta:=+~ya :pua;naH :pua;naH nyUa;na;tvMa . a;a;pa;pa;a:=+ma;a;Na;a;pa;tva -
å å
kÒ +:mea;Nea;a;ta ta:†a;d;DRa;. a;a;pa;a;na;a;m,a A;DRa:$ya;a;pa:=+}.pa:=+a Za:=+pa:=+}.pa:=+a . a
ëÅ
A;a;na;a;ya;ma;a;na;a na ëÁ a;. a;d;a;pa
*:+: :pa;yRa;va;~ya;a;ta A;a;na;ntya;a;d, ;a;va;Ba;a;ga;~ya Á
ú
ta;taH ;a;k+:ya;nta;a:*.
ãÁa;t,a :pra;de;ZMa ga;tva;a . a;a;pa;~ya ãÉa
.ja;a;va;a;ya;a;(

A;pa;a;ya;~tva;m,a A;a;pa;a;dùÅ;a . a;a;pa:$ya;a;nta:=M . a ZUa;nya;pra;a;yMa l+.b.Dva;a :pua;na:=+a;pa

k+:pya;ma;a;na;ma;nta:=+m,a A;tya;pa;ma;a;pa k+:Ea;Za;l+.a;t,a :℄ea;ya;m,a Á

10.2. Nı̄lakan.t.ha’s summation of the infinite geometric series

The question that Nı̄lakan.t.ha poses as he commences his detailed discussion on

the sum of geometric series is very important and arises quite naturally whenever
one encounters the sum of an infinite series:62
k+:TMa :pua;naH ta;a;va;de;va va;DRa;tea ta;a;va;d
õ ;DRa;tea . a ?
How do you know that [the sum of the series] increases only up to that [limiting
value] and that it certainly increases up to that [limiting value]?

Proceeding to answer the above question, Nı̄lakan.t.ha first states the general
result
"    #
1 2
 3
1 1 a
a + + +... = .
r r r r −1

Here, the left hand side is an infinite geometric series with the successive terms
being obtained by dividing by a common divisor, r , known as cheda, whose value

61Ibid., comm. on Ganitapāda 17, pp. 104–05.

.
62Ibid., p. 106.
Development of Calculus in India 243

is assumed to be greater than 1. He further notes that this result is best demon-
strated by considering a particular case, say r = 4. In his own words:63
å
o+.. ya;tea Á O;;vMa yaH tua;ya;. Ce+.d;pa:=+Ba;a;ga;pa:=+}.pa:=+a;ya;aH A;na;nta;a;ya;aH A;a;pa

.sMa;ya;ea;gaH ta;~ya A;na;nta;a;na;a;ma;a;pa k+:pya;ma;a;na;~ya ya;ea;ga;~ya A;a;dùÅ;a;a;va;ya;a;va;naH

å
:pa:=+}.pa:=+Ma;Za;. Ce+.d;a;t,a O;;k+:ea;na;. Ce+.d;Ma;Za;sa;a;}yMa .sa;vRa:ˆa .sa;ma;a;na;mea;va Á ta;dùÅ;a;Ta;a

å
– . a;tua:=M+Za;pa:=+}.pa:=+a;ya;a;mea;va ta;a;va;t,a :pra;Ta;mMa :pra;a;ta;pa;a;dùÅ;a;tea Á
It is being explained. Thus, in an infinite (ananta) geometrical series (tulya-
ccheda-parabhāga-paramparā) the sum of all the infinite number of terms con-
sidered will always be equal to the value obtained by dividing by a factor which is
one less than the common factor of the series. That this is so will be demonstrated
by first considering the series obtained with one-fourth (caturam. śa-paramparā).

What is intended to be demonstrated is

"    #
1 2
 3
1 1 a
a + + +... = . (75)
4 4 4 3

Besides the multiplying factor a, it is noted that, one-fourth and one-third are
the only terms appearing in the above equation. Nı̄lakan.t.ha first defines these
numbers in terms of one-twelfth of the multiplier a referred to by the word rāśi.
For the sake of simplicity we take the rāśi to be unity.

1 1 1 1
3× = ; 4× = .
12 4 12 3

Having defined them, Nı̄lakan.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, which leads to the general result,

"  2  n #  n  
1 1 1 1 1 1
− + +...+ = . (76)
3 4 4 4 4 3

63Ibid., pp. 106–07.

244 K. Ramasubramanian and M. D. Srinivas

Nı̄lakan.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 be-
tween 13 and sum of powers of 14 (as given by the right hand side of (76)), 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
+ + ...+ +... = . (77)
4 4 4 3

A brief extract from the text presenting the above argument is given below:64

yea .=+a;Zea;d
õ ;Ra;d;Za;Ma;Za;aH .tea;Sa;Ma ;a:ˆa;kM ;a;h . a;tua:=M+ZaH Á . a;tua;SkM . a ˆyMa;ZaH Á

ta;ƒa;tua;;yea ˆyMa;Za;a;tma;ke Ba;a;ga:ˆa;yMa . a;tua:=M+Zea;na;a;pUa;NRa;m,a Á yaH :pua;naH ta;~ya

å
. a;tua;Ta;eRa Y;}.ZaH ta;~ya;a;a;pa :pa;a;d:ˆa;yMa . a;tua:=M+Za;~ya . a;tua:=M+Zea;na;a;pUa;NRa;m,a Á

d
õ ;a;d;Za;Ma;Za;a;na;Ma ˆa;ya;a;Na;Ma ...

ñÍ
ta;~ya :pua;naH :pua;na:=+a;ta;sUa;[ma;tva;a;de;va na :k *:Å +.a;k+:a:=H
e +:va;lM ˆyMa;Za;tvea;na A;çÅ ,
;a;na:+.pya;ma;a;Na;~ya va;a ;a;kÒ +:ya;ma;a;Na;~ya va;a A;a;na;ntya;a;t,a Á A;a;na;ntya;a;de;va

;a;Za;;tva;a;de;va k+:mRa;Na;~ta;~ya A;pa;a:=+pUa;a;tRa;Ba;Ra;a;ta Á O;;vMa .sa;vRa;d;a;a;pa

å ïîéå x +:Sya;a:ˆa
.sa;a;va;Zea;Sa;a;Na;Ma k+:mRa;Na;Ma :pa:=+}.pa:=+a;ya;Ma k+:a;t=+: yeRa;na;a;k .sa;a;Ša;a;h;ta;a;ya;Ma

ãÉa;a;ya;tea . a;tua;gRua;Na;ea:†a:=e gua;Na;ea:†a:=+a;K.yea ga;a;Na;tea Y;a;pa Á

:pa;a:=+pUa;a;tRaH .~ya;a;de;vea;a;ta ;a;na;(

Three times one-twelfth of a rāśi is one-fourth (caturam . śa) [of that rāśi]. Four
times that is one-third (tryam . śa). [Considering] four times that [one-twelfth of
the rāśi] which is one-third, three by fourth of that falls short by one-fourth [of
one-third of the rāśi]. Three-fourths of that [i.e., of 4.3 1 of the rāśi] which is
one-fourth of that (tryam . śa), again falls short [of the same] by one-fourth of
one-fourth [of one-third of the rāśi] . . .

Since the result to be demonstrated or the process to be carried out is never

ending (ānantyāt) and the difference though very small (atisūks.matvāt) [still
exists and the sum of the series] cannot be simply taken to be one-third. It seems
that the process is incomplete since always something remains because of its never
ending nature. In fact, since in all the problems involving [infinite] series, by
bringing in all the terms and placing them together, the process would [in princi-
ple] become complete, here, in the mathematics involving repeated multiplication
of one-fourth, a similar conclusion may be drawn.

64Ibid., p. 107.
Development of Calculus in India 245

11. Derivation of binomial series expansion

Yuktibhās.ā presents a very interesting derivation of the binomial series for (1 +

x)−1 by making iterative substitutions in an algebraic identity. The method given
65
in the text may be summarized
as follows.
c
Consider the product a b , where some quantity a is multiplied by the multi-
plier c, and divided by the divisor b. Here, a is called gun.ya, c the gun.aka and
b the hāra, which are all assumed to be positive. Now the above product can be
rewritten as:
c (b − c)
a =a−a . (78)
b b

In the expression a (b−c)

b in (78) above, if we want to replace the division by b
(the divisor) by division by c (the multiplier), then we have to make a subtractive
correction (called śodhya-phala) which amounts to the following equation.
 
(b − c) (b − c) (b − c) (b − c)
a =a − a × . (79)
b c c b

Now, in the second term (inside parenthesis) in (79)—which is what we referred

to as śodhya-phala, which literally means a quantity to be subtracted—if we again
replace the division by b by division by c, then we have to employ the relation (79)
once again to get another subtractive term
 
c (b − c) (b − c) (b − c)
a = a− a −a ×
b c c b
 
(b − c) (b − c) (b − c) c
= a− a −a × ×
c c c b
" " !##
(b − c) (b − c)2 (b − c)2 (b − c)
= a− a − a − a × . (80)
c c2 c2 b

2
Here, the quantity a (b−c)c2
is called dvitı̄ya-phala or simply dvitı̄ya and the
one subtracted from that is dvitı̄ya-śodhya-phala. If we carry out the same set of
operations, the m t h śodhya-phala subtracted from the m t h term will be of the form

 m  m
(b − c) (b − c) (b − c)
a −a × .
c c b

65Ganita-yukti-bhāsā, cited above, Vol. I, Sections 6.3.3–4, pp. 54–58, 188–191, 375–378.
. .
246 K. Ramasubramanian and M. D. Srinivas

Since the successive śodhya-phalas are subtracted from their immediately pre-
ceding term, we will end up with a series in which all the odd terms (leaving
out the gun.ya, a) are negative and the even ones positive. Thus, after taking m
śodhya-phalas we get

(b − c) 2 (b − c) m
   
c (b − c)
a = a−a +a − . . . + (−1)m a
b c c c
 m
(b − c) (b − c)
+(−1)m+1 a . (81)
c b

Regarding the question of termination of the process, both the texts Yuktibhās.ā
and Kriyākramakarı̄ clearly mention that logically there is no end to the process
of generating śodhya-phalas. We may thus write our result as:66

(b − c) 2 (b − c) m−1
   
c (b − c)
a = a−a +a − . . . + (−1)m−1 a
b c c c
 m
(b − c)
+(−1)m a + .... (82)
c

It is also noted that the process may be terminated after having obtained the
desired accuracy by neglecting the subsequent phalās as their magnitudes become
smaller and smaller. In fact, Kriyākramakarı̄ explicitly mentions the condition
under which the succeeding phalas will become smaller and smaller:67
x +:tea Y;a;pa yua;a;€+:taH ëÅ
O;;vMa mua;huH :P+.l+.a;na;ya;nea k ëÁ
*:+:a; a;pa na .sa;ma;a;a;‹aH Á ta;Ta;a;a;pa

ya;a;va;d;pea;[Ma .sUa;[ma;ta;a;ma;a;pa;a;dùÅ;a ãÉa;a;tya;a;nyua;pea;[ya

:pa;a;( :P+.l+.a;na;ya;nMa -
.sa;ma;a;pa

na;a;ya;m,a Á I+.h;ea:†a:=+ea:†a:=+P+.l+.a;na;Ma nyUa;na;tvMa tua gua;Na;h;a:=+a;nta:=e gua;Na;k+:a:=+a;ŠyUa;na

O;;va .~ya;a;t,a Á
Thus, even if we keep finding the phalas repeatedly, logically there is no end
to the process. Even then, having carried on the process to the desired accu-
racy (yāvadapeks.am . sūks.matāmāpādya), one should terminate computing
the phalas by [simply] neglecting the terms that may be obtained further (pāścā-
tyānyupeks.ya). Here, the succeeding phalas will become smaller and smaller
only when the difference between the gun . aka and hāra is smaller than gun. aka
[that is (b ∼ c) < c].

66It may be noted that if we set (b−c) = x, then c = 1 . Hence, the series (82) is none other
c b (1+x)
than the well known binomial series
a
= a − ax + ax 2 − . . . + (−1)m ax m + . . . ,
1+x
which is convergent for −1 < x < 1.
67Kriyākramakarı̄ on Lı̄lāvatı̄, cited above (fn. 14), comm. on verse 199, p. 385.
Development of Calculus in India 247

12. Estimation of sums of 1k + 2k + . . . n k for large n

As mentioned in section 4.1, Āryabhat.a has given the explicit formulae for the
summation of squares and cubes of integers. The word employed in the In-
dian mathematical literature for summation is saṅkalita. The formulae given by
Āryabhat.a for the saṅkalitas are as follows:

n(n + 1)
Sn(1) = 1 + 2 + ··· + n =
2
n(n + 1)(2n + 1)
Sn(2) = 2 2
1 + 2 + ···+ n = 2
6
n(n + 1) 2
 
Sn(3) = 13 + 23 + · · · + n 3 = . (83)
2

From these, it is easy to estimate these sums when n is large. Yuktibhās.ā gives a
general method of estimating the sama-ghāta-saṅkalita

Sn(k) = 1k + 2k + · · · + n k , (84)

when n is large. The text presents a general method of estimation, which does not
make use of the actual value of the sum. In fact, the same argument is repeated
even for k = 1, 2, 3, although the result of summation is well known in these
cases.

12.1. The sum of natural numbers (Mūla-saṅkalita)

Yuktibhās.ā takes up the discussion on saṅkalitas in the context of evaluating the

circumference of a circle which is conceived to be inscribed in a square. It is
half the side of this square that is being referred to by the word bhujā in both
the citations as well as explanations offered below. Half of the side of the square
(equal to the radius) is divided
n equal bits, known as bhujā-khan.d.as. It is
into
these bhujā-khan.d.as nr , 2 nr · · · whose powers are summed.


To start with, Yuktibhās.ā discusses just the basic summation of bhujā-khan.d.as
called Mūla-saṅkalita. We now cite the following from the translation of Yuk-
tibhās.ā:68
Now is described the methods of making the summations (referred to in the ear-
lier sections). At first, the simple arithmetical progression (kevala-saṅkalita) is
described. This is followed by the summation of the products of equal numbers
(squares). . . .
Here, in this mūla-saṅkalita (basic arithmetical progression), the final bhujā is

68Ganita-yukti-bhāsā, cited above, Section 6.4, pp. 61–67, 192–97, 382–88.

. .
248 K. Ramasubramanian and M. D. Srinivas

equal to the radius. The term before that will be one segment (khan .d. a) less. The
next one will be two segments less. Here, if all the terms (bhujās) had been equal
to the radius, the result of the summation would be obtained by multiplying the
radius by the number of bhujās. However, here, only one bhujā is equal to the
radius. And, from that bhujā, those associated with the smaller hypotenuses are
less by one segment each, in order. Now, suppose the radius to be of the same
number of units as the number of segments to which it has been divided, in order
to facilitate remembering (their number). Then, the number associated with the
penultimate bhujā will be less by one (from the number of units in the radius);
the number of the next one, will be less by two from the number of units in the
radius. This reduction (in the number of segments) will increase by one (at each
step). The last reduction will practically be equal to the measure of the radius, for
it will be less only by one segment. In other words, when the reductions are all
added, the sum thereof will practically (prāyen . a) be equal to the summation of
the series from 1 to the number of units in the radius; it will be less only by one
radius length. Hence, the summation will be equal to the product of the number
of units in the radius with the number of segments plus one, and divided by 2.
The summation of all the bhujās of the different hypotenuses is called bhujā-
saṅkalita.
Now, the smaller the segments, the more accurate (sūks.ma) will be the result.
Hence, do the summation also by taking each segment as small as an atom (an . u).
Here, if it (namely, the bhujā or the radius) is divided into parārdha (a very
large number) parts, to the bhujā obtained by multiplying by parārdha add one
part in parārdha and multiply by the radius and divide by 2, and then divide by
parārdha. For, the result will practically be the square of the radius divided by
two. . . .

The first summation, the bhujā-saṅkalita, may be written in the reverse order
from the final bhujā to the first bhujā as
 
 nr  (n − 1)r r 
Sn(1) = + + .... + . (85)
n n n

Now, conceive of the bhujā-khan.d.a nr as being infinitesimal (an.u) and at the same
time as of unit-measure (rūpa), so that the radius will be the measure of n, the
pada, or the number of terms. Then

Sn(1) = n + (n − 1) + .... + 1. (86)

If each of the terms were of the measure of radius (n) then the sum would be
nothing but n 2 , the square of the radius. But only the first term is of the measure
of radius, the next is deficient by one segment (khan.d.a), the next by two segments
and so on till the last term which is deficient by an amount equal to radius-minus-
one segment. In other words,

Sn(1) = n + [n − 1] + [n − 2].... + [n − (n − 2)] + [n − (n − 1)]

= n.n − [1 + 2 + ... + (n − 1)]. (87)
Development of Calculus in India 249

When n is very large, the quantity to be subtracted from n 2 is practically (prāyen.a)

the same as Sn(1) , thus leading to the estimate

Sn(1) ≈ n 2 − Sn(1) , (88)

n2
or Sn(1) ≈ . (89)
2
It is stated that the result is more accurate, when the size of the segments are small
(or equivalently, the value of n is large).69
If instead of making the approximation as in (88), we proceed with (87) as it is,
(1) (1)
we get Sn = n 2 − (Sn − n), which leads to the well-known exact value of the
sum of the first n natural numbers
n(n + 1)
Sn(1) = , (90)
2
With the convention that the nr is of unit-measure, the above estimate (89) is stated
in the form that the bhujā-saṅkalita is half the square of the radius.

12.2. Summation of squares (Varga-saṅkalita)

We now cite the following from the translation of Yuktibhās.ā:70

Now is explained the summation of squares (varga-saṅkalita). Obviously, the
squares of the bhujās, which are summed up above, are the bhujās each multi-
plied by itself. Here, if the bhujās which are all multipliers, had all been equal to
the radius, their sum, (saṅkalita derived above), multiplied by the radius would
have been the summation of their squares. Here, however, only one multiplier
happens to be equal to the radius, and that is the last one. The one before that will
have the number of segments one less than in the radius. (Hence) if that, (i.e., the
second one), is multiplied by the radius, it would mean that one multiplied by the
penultimate bhujā would have been the increase in the summation of the squares.
Then (the segment) next below is the third. That will be less than the radius by
two segments. If that is multiplied by the radius, it will mean that, the summa-
tion of the squares will increase by the product of the bhujā by two (segments).
In this manner, the summation in which the multiplication is done by the radius
(instead of the bhujās) would be larger than the summation of squares by terms

69Śaṅkara Vāriyar also emphasizes the same idea, in his discussion of the estimation of
saṅkalitas in his commentary Kriyākramakarı̄ on Lı̄lāvatı̄ (cited above (fn. 14), comm. on verse
199, p. 382.):
Ka;Nq+.~ya;a;pa;tvea .sa;tyea;va l+.b.Da;~ya .sUa;[ma;ta;a . a .~ya;a;t,a Á

.d
Only when the segment is small (khan . asyālpatve) the result obtained would be
accurate.

70Ganita-yukti-bhāsā, cited above, Section 6.4, pp. 61–67, 192–97, 382–88.

. .
250 K. Ramasubramanian and M. D. Srinivas

which involve the successively smaller bhujās multiplied by successively higher

numbers. If (all these additions) are duly subtracted from the summation where
the radius is used as the multiplier, the summation of squares (varga-saṅkalita)
will result.
Now, the bhujā next to the east-west line is less than the radius by one (segment).
So if all the excesses are summed up and added, it would be the summation of the
basic summation (mūla-saṅkalita-saṅkalita). Because, the sums of the sum-
mations is verily the ‘summation of summations’ (saṅkalita-saṅkalita). There,
the last sum has (the summation of) all the bhujās. The penultimate sum is next
lower summation to the last. This penultimate sum is the summation of all the
bhujās except the last bhujā. Next to it is the third sum which is the sum of all
the bhujās except the last two. Thus, each sum of the bhujās commencing from
any bhujā which is taken to be the last one in the series, will be less by one bhujā
from the sum (of the bhujās) before that.
Thus, the longest bhujā is included only in one sum. But the bhujā next lower
than the last (bhujā) is included both in the last sum and also in the next lower
sum. The bhujās below that are included in the three, four etc. sums below it.
Hence, it would result that the successively smaller bhujās commencing from the
one next to the last, which have been multiplied by numbers commencing from 1
and added together, would be summation of summations (saṅkalita-saṅkalita).
Now, it has been stated earlier that the summation (saṅkalita) of (the segments
constituting) a bhujā which has been very minutely divided, will be equal to
half the square of the last bhujā. Hence, it follows that, in order to obtain the
summation (saṅkalita) of the bhujās ending in any particular bhujā, we will
have to square each of the bhujās and halve it. Thus, the summation of summa-
tions (saṅkalita-saṅkalita) would be half the summation of the squares of all the
bhujās. In other words, half the summation of the squares is the summation of the
basic summation. So, when the summation is multiplied by the radius, it would
be one and a half times the summation of the squares. This fact can be expressed
by stating that this contains half more of the summation of squares. Therefore,
when the square of the radius divided by two is multiplied by the radius and one-
third of it subtracted from it, the remainder will be one-third of the whole. Thus it
follows that one-third of the cube of the radius will be the summation of squares
(varga-saṅkalita).

With the same convention that nr is the measure of the unit, the bhujā-varga-
saṅkalita (the sum of the squares of the bhujās) will be

Sn(2) = n 2 + (n − 1)2 + .... + 12 . (91)

In above expression, each bhujā is multiplied by itself. If instead, we consider that

each bhujā is multiplied by the radius (n in our units), then that would give raise
to the sum

n [n + (n − 1) + ... + 1] = n Sn(1) . (92)

This sum is exceeds the bhujā-varga-saṅkalita by the amount

nSn(1) − Sn(2) = 1.(n − 1) + 2.(n − 2) + 3.(n − 3) + . . . + (n − 1).1.

Development of Calculus in India 251

This may be written in the form

nSn(1) − Sn(2) = (n − 1) + (n − 2) + (n − 3) + . . . +1
+(n − 2) + (n − 3) + . . . +1
+(n − 3) + . . . +1
+... . (93)

Thus,
(1) (1) (1)
nSn(1) − Sn(2) = Sn−1 + Sn−2 + Sn−3 + . . . . (94)

The right hand side of (94) is called the saṅkalita-saṅkalita (or saṅkalitaikya),
(1)
the repeated sum of the sums Si (here taken in the order i = n − 1, n − 2, . . . 1).
These are defined also by Śaṅkara Vāriyar in Kriyākramakarı̄ as follows:71
ñÍ ñÍ ñÍ
ta;Ta;a ;a;h .sa;*:+.
öÅ a;l+.ta;a;na;Ma ya;ea;ga;ea ;a;h öÅ a;l+.ta;sa;*:+.
.sa;*:+. öÅ a;l+.ta;mua;. ya;tea Á ta:ˆa
ñÍ ñÍ
A;ntya;sa;*:+.
öÅ a;l+.tMa .sa;va;Ra;sa;Ma Bua:ja;a;na;Ma ya;ea;gaH Á o+.pa;a;ntya;sa;*:+.
öÅ a;l+.tMa tua

A;ntya;Bua:ja;a;v.ya;a;ta;a:=+€+:a;na;a;Æa;ma;ta:=e+Sa;Ma ya;ea;gaH Á o+.pa;a;ntya;a;t,a :pUa;vRa;~ya

ñÍ
.sa;*:+.
öÅ a;l+.tMa :pua;na;~ta;d;va;a;Da;k+:a;na;a;mea;va Bua:ja;a;na;Ma ya;ea;gaH Á O;;vMa :pUa;vRa -
ñÍ ñÍ
.sa;*:+.
öÅ a;l+.ta;a;a;na .~va;ea:†a:=+a;t,a .sa;*:+.
öÅ a;l+.ta;a;t,a O;;kE +:ke +:na Bua:jea;na ;a;va:=+a;h;ta;a;a;na

Ba;va;a;nta
// Á

The sum of the summations is called as saṅkalita-saṅkalita. Of them the last

saṅkalita is the sum all the bhujās. The penultimate saṅkalita is the sum of
all the bhujās other than the last one. The saṅkalita of the one preceding the
penultimate is the sum of the bhujās ending with that. Thus, all the preceding
saṅkalitas will fall short by a bhujā from the succeeding saṅkalita.

n2
For large n, we have already estimated in (89) that Sn(1) ≈ 2 . Thus, for large n

(n − 1)2 (n − 2)2 (n − 3)2

nSn(1) − Sn(2) ≈ + + + .... (95)
2 2 2

Thus, the right hand side of (94) (the saṅkalita-saṅkalita or the excess of nSn(1)
(2)
(2) Sn
over Sn ) is essentially 2 for large n, so that we obtain

(2)
Sn
nSn(1) − Sn(2) ≈ . (96)
2

71Kriyākramakarı̄ on Lı̄lāvatı̄, cited above (fn. 14), comm. on verse 199, pp. 382–83.
252 K. Ramasubramanian and M. D. Srinivas

Again, using the earlier estimate (89) for Sn(1) , we obtain the result

n3
Sn(2) ≈ . (97)
3
Thus bhujā-varga-saṅkalita is one-third the cube of the radius.

12.3. Sama-ghāta-saṅkalita

We now cite the following from the translation of Yuktibhās.ā:72

Now, the square of the square (of a number) is multiplied by itself, it is called
sama-pañca-ghāta (number multiplied by itself five times). The successive
higher order summations are called sama-pañcādi-ghāta-saṅkalita (and will
be the summations of powers of five and above). Among them if the summation
(saṅkalita) of powers of some order is multiplied by the radius, then the product
is the summation of summations (saṅkalita-saṅkalita) of the (powers of
the) multiplicand (of the given order), together with the summation of powers
(sama-ghāta-saṅkalita) of the next order. Hence, to derive the summation of
the successive higher powers: Multiply each summation by the radius. Divide it
by the next higher number and subtract the result from the summation got before.
The result will be the required summation to the higher order.

Thus, divide by two the square of the radius. If it is the cube of the radius, divide
by three. If it is the radius raised to the power of four, divide by four. If it is
(the radius) raised to the power of five, divide by five. In this manner, for powers
rising one by one, divide by numbers increasing one by one. The results will be,
in order, the summations of powers of numbers (sama-ghāta-saṅkalita). Here,
the basic summation is obtained from the square, the summation of squares from
the cube, the summation of cubes from the square of the square. In this manner,
if the numbers are multiplied by themselves a certain number of times (i.e., raised
to a certain degree) and divided by the same number, that will be the summation
of the order one below that. Thus (has been stated) the method of deriving the
summations of (natural) numbers, (their) squares etc.

In the case of a general samaghāta-saṅkalita, (summation of equal powers) given

by

Sn(k) = n k + (n − 1)k + . . . + 1k , (98)

the procedure followed to estimate its behavior for large n is essentially the same
as that followed in the case of vargasaṅkalita. We first compute the excess of
nSn(k−1) over Sn(k) to be a saṅkalita-saṅkalita or repeated sum of the lower order

72Ganita-yukti-bhāsā, cited above, Section 6.4, pp. 61–67, 192–97, 382–88.

. .
Development of Calculus in India 253

saṅkalitas Sr(k−1) :

(k−1) (k−1) (k−1)

nSn(k−1) − Sn(k) = Sn−1 + Sn−2 + Sn−3 +... . (99)

(k−1)
If the lower order saṅkalita Sn has already been estimated to be, say,

nk
Sn(k−1) ≈ , (100)
k

then, the above relation (99) leads to73

(n − 1)k (n − 2)k (n − 3)k

nSn(k−1) − Sn(k) ≈ + + + ...
  k k k
1
≈ Sn(k) . (101)
k

Rewriting the above equation we have74

 
1
Sn(k) ≈ nSn(k−1) − Sn(k) . (102)
k

Using (100), we obtain the estimate

n k+1
Sn(k) ≈ . (103)
(k + 1)

73As one of the reviewers has pointed out, this argument leading to (101) is indeed similar to the
derivation of the following relation, which is based on the interchange of order in iterated integrals:
Z 1 Z 1 Z 1 Z 1 Z y Z 1 k
y
(1 − x)x k−1 d x = x k−1 dy d x = y x k−1 d x dy = dy.
0 0 x 0 0 0 k

74As Śaṅkara Vāriyar states in his Kriyākramakarı̄ on Lı̄lāvatı̄ (cited above (fn. 14), p. 383):

ñÍ ñÍ
öÅ a;l+.ta;a;na;ya;na;a;ya ta:†a;tsa;*:+.
A;ta o+†a:=+ea:†a:=+sa;*:+. öÅ a;l+.ta;~ya v.ya;a;sa;a;DRa;gua;Na;na;m,a
ñÍ
O;;k É +;a;a;‹a;~va;M
E +:k+:a;a;Da;k+:sa;*
Ë ù a;Za;Za;ea;Da;nMa . a k+:a;yRa;m,a I+.a;ta ;a;~Ta;ta;m,
/// a Á
Á
Therefore it is established that, for obtaining the sum of the next order, the previ-
ous sum, has to be multiplied by the radius and the present sum, divided by one
more than the previous [order], has to be diminished [from that product].
254 K. Ramasubramanian and M. D. Srinivas

12.4. Repeated summations (Saṅkalita-saṅkalita)

After having estimated the sum of powers of natural numbers samaghāta-saṅkalita

Yuktibhās.ā goes on to derive an estimate for the repeated summation (saṅkalita-
saṅkalita or saṅkalitaikya or vārasaṅkalita) of the natural number 1, 2, · · · , n.75
Now, are explained the first, second and further summations: The first summa-
tion (ādya-saṅkalita) is the basic summation (mūla-saṅkalita) itself. It has
already been stated (that this is) half the product of the square of the number of
terms (pada-vargārdha). The second (dvitı̄ya-saṅkalita) is the summation of
the basic summation (mūla-saṅkalitaikya). It has been stated earlier that it is
equal to half the summation of squares. And that will be one-sixth of the cube of
the number of terms.
Now, the third summation: For this, take the second summation as the last term
(antya); subtract one from the number of terms, and calculate the summation
of summations as before. Treat this as the penultimate. Then subtract two from
the number of terms and calculate the summation of summations. That will be the
next lower term. In order to calculate the summation of summations of numbers in
the descending order, the sums of one-sixths of the cubes of numbers in descend-
ing order would have to be calculated. That will be the summation of one-sixth
of the cubes. And that will be one-sixth of the summation of cubes. As has been
enunciated earlier, the summation of cubes is one-fourth the square of the square.
Hence, one-sixth of one-fourth the square of the square will be the summation
of one-sixth of the cubes. Hence, one-twenty-fourth of the square of the square
will be the summation of one-sixth of the cubes. Then, the fourth summation will
be, according to the above principle, the summation of one-twenty-fourths of the
square of squares. This will also be equal to one-twenty-fourth of one-fifth of
the fifth power. Hence, when the number of terms has been multiplied by itself
a certain number of times, (i.e., raised to a certain degree), and divided by the
product of one, two, three etc. up to that index number, the result will be the
summation up to that index number amongst the first, second etc. summations
(ādya-dvitı̄yādi-saṅkalita).

The first summation (ādya-saṅkalita) Vn(1) is just the mūla-saṅkalita or the

basic summation of natural numbers, which has already been estimated in (89)

Vn(1) = Sn(1) = n + (n − 1) + (n − 2) + . . . + 1
n2
≈ . (104)
2
The second summation (dvitı̄ya-saṅkalita or saṅkalita-saṅkalita or saṅkalitaikya)
is given by
(1) (1)
Vn(2) = Vn(1) + Vn−1 + Vn−2 +...
(1) (1)
= Sn(1) + Sn−1 + Sn−2 +... . (105)

75Ganita-yukti-bhāsā, cited above, Section 6.4, pp. 61–67, 192–97, 382–88.

. .
Development of Calculus in India 255

As was done earlier, this second summation can be estimated using the estimate
(89) for Sn(1)

n2 (n − 1)2 (n − 2)2
Vn(2) ≈ + + +... . (106)
2 2 2

Therefore
 
1
Vn(2) ≈ S (2) . (107)
2 n

Using the earlier estimate (97) for Sn(2) , we get an estimate for the dvitı̄ya-saṅkalita

n3
Vn(2) ≈ . (108)
6

Now the next repeated summation can be found in the same way

(2) (2)
Vn(3) = Vn(2) + Vn−1 + Vn−2 + . . .
n3 (n − 1)3 (n − 2)3
≈ + + + ...
6  6 6
1
≈ S (3)
6 n
n4
≈ . (109)
24

It is noted that proceeding this way we can estimate repeated summation Vn(k) of
order k, for large n, to be76

(k−1) (k−1)
Vn(k) = Vn(k−1) + Vn−1 + Vn−2 + . . .
n k+1
≈ . (110)
1.2.3. . . . (k + 1)

76These are again estimates for large n. As mentioned in Section 4, exact expressions for the first
(1) (2)
two summations, Vn and Vn , are given in Āryabhat.ı̄ya, Gan . itapāda 21; and the exact expression
(k)
for the k-th order repeated summation Vn has been given (under the name vāra-saṅkalita), by
(k)
Nārāyan.a Pan.d.ita (c. 1350) in his Gan
. itakaumudı̄, 3.19. This exact expression for Vn is also
noted in Section 7.5.3 of Yuktibhās.ā.
256 K. Ramasubramanian and M. D. Srinivas

13. Derivation of the Mādhava series for π

The following accurate value of π (correct to 11 decimal places), given by Mādhava,

has been cited by Nı̄lakan.t.ha in his Āryabhat.ı̄ya-bhās.ya and by Śaṅkara Vāriyar
in his Kriyākramakarı̄.77

;a;va;bua;Da;nea:ˆa;ga:ja;a;a;h;hu;ta;a;Za;na;a:ˆa;gua;Na;vea;d;Ba;va;a:=+Na;ba;a;h;vaH Á

na;va;a;na;Ka;vRa;Æa;ma;tea vxa;a;ta;a;va;~ta:=e :pa;a:=+a;Da;ma;a;na;Æa;ma;dM .ja;ga;d

u ;bRua;Da;aH Á Á

The π value given above is:

2827433388233
π≈ = 3.141592653592... (111)
9 × 1011

The 13 digit number appearing in the numerator has been specified using bhūta-
saṅkhya system, whereas the denominator is specified by word numerals.78

13.1. Infinite series for π

The infinite series for π attributed to Mādhava is cited by Śaṅkara Vāriyar in his
commentaries Kriyākramakarı̄ and Yukti-dı̄pikā. Mādhava’s verse quoted runs
as follows:79

v.ya;a;sea va;a;a:=+a;Da;a;na;h;tea .+.pa;&+.tea v.ya;a;sa;sa;a;ga:=+a;Æa;Ba;h;tea Á

ñÍ
É +;a;a;Ba;€+:mx
;a:ˆa;Za:=+a;a;d;a;va;Sa;ma;sa;*
Ë ù a;NMa .~vMa :pxa;Ta;k, kÒ +:ma;a;t,a ku +:ya;Ra;t,a Á Á
Á
The diameter multiplied by four and divided by unity [is found and saved]. 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 result
saved].

The series given by the verse may be represented as

 
1 1 1
Paridhi = 4 × Vyāsa × 1 − + − + . . . . . . . (112)
3 5 7

77Āryabhatı̄ya-bhāsya on Āryabhatı̄ya, cited above (fn. 53), comm. on Ganitapāda 10, p. 42;
. . . .
Kriyākramakarı̄ on Lı̄lāvatı̄, cited above (fn. 14), comm. on verse 199, p. 377.
78In the bhūta-saṅkhyā system, vibudha =33, netra =2, gaja =8, ahi =8, hutāśana =3,
trigun 11
. a =3, veda =4, bha =27, vāran . a =8, bāhu =2. In word numerals, nikharva represents 10 .
Hence, nava-nikharva =9 × 1011 .
79op. cit., p. 379.
Development of Calculus in India 257

The words paridhi and vyāsa80 in the above equation refer to the circumference
and diameter respectively. Hence the equation may be rewritten as
 
π 1 1 1
= 1 − + − +... . (113)
4 3 5 7

O S

Blown up version
of this quadrant

P0 Pi−1 Pi p
n

Ci
Ai−1
Ai
Bi

k i−1
ki

O S

F IGURE 8. Geometrical construction used in the proof of the

infinite series for π.

We shall now present the derivation of the above result as outlined in Yuk-
tibhās.ā of Jyes.t.hadeva and Kriyākramakarı̄ of Śaṅkara Vāriyar. For this pur-
pose, let us consider the quadrant O P0 Pn S of the square circumscribing the given
circle (see Figure 8). Divide the side P0 Pn into n equal parts (n very large). P0 Pi ’s

80Nı̄lakantha, in his Āryabhatı̄ya-bhāsya, presents the etymological derivation of the

.. . .
word vyāsa as ‘the one which splits the circle into two halves’: v.ya;a;sea;na ;a;h vxa:†Ma

v.ya;~ya;tea Á (Āryabhat.ı̄ya-bhās.ya, cited above (fn. 53), comm. on Gan

. itapāda 11, p. 43).
258 K. Ramasubramanian and M. D. Srinivas

are the bhujās and O Pi ’s are the karn.as denoted by ki . The points of intersection
of these karn.as and the circle are marked as Ai s.
The bhujās P0 Pi , the karn.as ki and the east-west line O P0 form right-angled
triangles whose hypotenuses are given by
 2
ir
ki2 = r 2 + , (114)
n

where r is the radius of the circle.

The feet of perpendiculars from the points Ai−1 and Pi−1 along the i t h karn.a
are denoted by Bi and Ci . The triangles O Pi−1 Ci and O Ai−1 Bi are similar.
Hence,

Ai−1 Bi Pi−1 Ci
= . (115)
O Ai−1 O Pi−1

Similarly triangles Pi−1 Ci Pi and P0 O Pi are similar. Hence,

Pi−1 Ci O P0
= . (116)
Pi−1 Pi O Pi

From these two relations we have,

O Ai−1 .O P0 .Pi−1 Pi
Ai−1 Bi =
O Pi−1 .O Pi
O Ai−1 O P0
= Pi−1 Pi × ×
O Pi−1 O Pi
r  r r
= × ×
n ki−1 ki
!
r  r 2
= . (117)
n ki−1 ki

It is then noted that when n is large, the Rsines Ai−1 Bi can be taken as the arc-bits
themselves.
→ :pa;a:=+DyMa;Za
:pa;a:=+a;Da;Ka;Nq+.~ya;a;DRa:$ya;a

i.e., Ai−1 Bi → A\
i−1 A i .

Thus, 18 th of the circumference of the circle can be written as sum of the con-
tributions given by (117). That is
" ! ! ! !#
C r  r2 r2 r2 r2
≈ + + + ···+ . (118)
8 n k0 k1 k1 k2 k2 k3 kn−1 kn
Development of Calculus in India 259

Though this is the expression that actually needs to be evaluated, the text men-
tions that there may not be much difference in approximating it by either of the
following expressions:
" ! ! ! !#
r2 r2 r2 r2
 
C r 
= + + + ··· + (119)
8 le f t n k02 k12 k22 2
kn−1

or
" ! ! ! !#
r2 r2 r2 r2
 
C r 
= + + + ··· + . (120)
8 right n k12 k22 k32 kn2

It can be easily seen that

   
C C C
< < . (121)
8 right 8 8 le f t

In other words, though the actual value of the circumference lies in between the
values given by (120) and (119) what is being said is that there will not be much
difference if we divide by the square of either of the karn.as rather than by the
product of two successive ones. Actually, the difference between (120) and (119)
is given by
" ! !#
r2 r2 r 
r   
1
− = 1− (since k02 , kn2 = r 2 , 2r 2 )
n k02 kn2 n 2
r  1
= (122)
n 2

Evidently this difference approaches zero as n becomes very large, as noted in

both the texts Yuktibhās.ā and Kriyākramakarı̄.
The terms in (120) are evaluated using the śodhya-phala technique (binomial
series, discussed earlier in Section 11) and each one of them may be re-written in
the form81
! ! !2
r r2 r r ki2 − r 2 r ki2 − r 2
= − + −... (123)
n ki2 n n r2 n r2

81It may be noted that this series is convergent since k 2 = r 2 + ir 2 and 0 ≤ (k 2 − r 2 ) < r 2
 
i n i
for i < n.
260 K. Ramasubramanian and M. D. Srinivas

Using (114) and (123) in (120), we obtain:

n
!
C X r r2
=
8 n ki2
i=1
 
n  
X r  r2
=
 (124)
n 2 ir 2
i=1 r + n
 
 
2 
n ir 2 ir 2
X r r n r n
=  −  2  +  2  − . . . (125)

n n r n r
i=1
r 
= [1 + 1 + . . . + 1]
n " #
 r   1   r 2  2r 2  nr 2
− + +...+
n r2 n n n
" #
2r 4
r  1   r 4    nr 4
+ + +...+
n r4 n n n
" #
r  1   r 6  6  nr 6
2r
− + +...+
n r6 n n n
+... . (126)

Each of the terms in (126) is a sum of results (phala-yoga) which we need to

estimate when n is very large, and we have a series of them (phala-paramparā)
which are alternatively positive and negative. Clearly the first term is just the sum
of the bhujā-khan.d.as.
The bhujās themselves are given by the integral multiples of bhujā-khan.d.a,
namely, nr , 2r nr
n , . . . n . In the series expression for the circumference given above,
we thus have the saṅkalitas or summations of even powers of the bhujās, such

2   2
2
as the bhujā-varga-saṅkalita, nr + 2r n + .... + nr
n , bhujā-varga-varga-

4   4
4
saṅkalita, nr + 2r n + ..... + nr
n , and so on.
If we take out the powers of bhujā-khan.d.a nr , the summations involved are
those of even powers of the natural numbers, namely edādyekottara-varga-
saṅkalita, 12 + 22 + ... + n 2 , edādyekottara-varga-varga-saṅkalita, 14 + 24 +
... + n 4 , and so on.
Development of Calculus in India 261

Now, recalling the estimates that were obtained earlier for these saṅkalitas,
when n is large,
n
X n k+1
ik ≈ , (127)
k+1
i=1

we arrive at the result82

 
C 1 1 1
= r 1 − + − + ··· , (128)
8 3 5 7

which is same as (112).

. skāra)
14. Derivation of end-correction terms (Antya-sam

It is well known that the series given by (112) for π4 is an extremely slowly con-
verging series. It is so slow that even for obtaining the value of π correct to 2
decimal places one has to find the sum of hundreds of terms and for getting it
correct to 4-5 decimal places we need to consider millions of terms. Mādhava
seems to have found an ingenious way to circumvent this problem. The technique
employed by Mādhava is known as antya-sam . skāra. The nomenclature stems
from the fact that a correction (sam. skara) is applied towards the end (anta) of the
series, when it is terminated after considering only a certain number of terms from
the beginning.

14.1. The criterion for antya-sam

. skāra to yield accurate result

The discussion on antya-sam

. skāra in both Yuktibhās.ā and Kriyākramakarı̄ com-
mences with the question:
How is it that one obtains the value of the circumference more accurately by doing
antya-sam 83
. skara, instead of repeatedly dividing by odd numbers?

82In modern terminology, the above derivation amounts to the evaluation of the following integral
 
n  
r2
Z 1
C X r  dx
= lim = r .

8 n→∞ n
 
ir
2 
0 1 + x2
i=1 r2 + n

83k+:TMa ñÍ
:pua;na:=;ˆa É +;a;a;h:=+Ne
mua;hu;a;vRa;Sa;ma;sa;*
Ë ù
Á
a;na l+.Bya;~ya :pa;a:=+DeaH A;a;sa;Ša;tva;m,a A;ntya -
.sMa;~k+:a:=e+Na A;a;pa;a;dùÅ;a;tea Á o+.. ya;tea Á . . . (Kriyākramakarı̄ on Lı̄lāvatı̄, cited above (fn. 14),
comm. on verse 199, p. 386.)
262 K. Ramasubramanian and M. D. Srinivas

The argument adduced in favor of terminating the series at any desired term, still
ensuring the accuracy, is as follows. Let the series for π4 be written as

π 1 1 1 p−3 1 p−1 1
= 1 − + − . . . + (−1) 2 + (−1) 2 , (129)
4 3 5 7 p−2 a p−2
1
where a p−2 is the correction term applied after odd denominator p − 2. On the
l
other hand, if the correction term ap , is applied after the odd denominator p, then

π 1 1 1 p−3 1 p−1 1 p+1 1

= 1 − + − . . . + (−1) 2 + (−1) 2 + (−1) 2 . (130)
4 3 5 7 p−2 p ap

If the correction terms indeed lead to the exact result, then both the series (129)
and (130) should yield the same result. That is,

1 1 1 1 1 1
= − or + = , (131)
a p−2 p ap a p−2 ap p

. skāra) to
is the criterion that must be satisfied for the end-correction (antya-sam
lead to the exact result.

14.2. Successive approximations to get more accurate correction-terms

The criterion given by (131) is trivially satisfied when we choose a p−2 = a p =

2 p. However, this value 2 p cannot be assigned to both the correction-divisors84
a p−2 and a p because both the corrections should follow the same rule. That is,

a p−2 = 2 p, ⇒ a p = 2( p + 2)
or, a p = 2 p, ⇒ a p−2 = 2( p − 2).

We can, however, have both a p−2 and a p close to 2 p by taking a p−2 = 2 p − 2

and a p = 2 p + 2, as there will always persist this much difference between p − 2
and p when they are doubled. Hence, the first (order) estimate of the correction
divisor is given as, “double the even number above the last odd-number divisor
p”,

a p = 2( p + 1). (132)

But, it can be seen right away that, with this value of the correction divisor, the
condition for accuracy (131), stated above, is not exactly satisfied. Therefore a

84By the term correction-divisor (samskāra-hāraka) is meant the divisor of the correction term.
.
Development of Calculus in India 263

measure of inaccuracy (sthaulya) E( p) is introduced

 
1 1 1
E( p) = + − . (133)
a p−2 ap p

Now, since the error cannot be eliminated, the objective is to find the correction
denominators a p such that the inaccuracy E( p) is minimised. When we set a p =
2( p + 1), the inaccuracy will be
 
1 1 1
E( p) = + −
(2 p − 2) (2 p + 2) p
1
= . (134)
( p 3 − p)

This estimate of the inaccuracy, E p being positive, shows that the correction has
been over done and hence there has to be a reduction in the correction. This means
that the correction-divisor has to be increased. If we take a p = 2 p + 3, thereby
leading to a p−2 = 2 p − 1, we have
 
1 1 1
E( p) = + −
(2 p − 1) (2 p + 3) p
(−2 p + 3)
= . (135)
(4 p 3 + 4 p 2 − 3 p)

Now, the inaccuracy happens to be negative. But, more importantly, it has a term
proportional to p in the numerator. Hence, for large p, E( p) given by (135) varies
inversely as p 2 , while for the divisor given by (132), E( p) as given by (134) varied
inversely as p 3 .85
From (134) and (135) it is obvious that, if we want to reduce the inaccuracy and
thereby obtain a better correction, then a number less than 1 has to be added to the
correction-divisor (132) given above. If we try adding rūpa (unity) divided by the
1
correction divisor itself, i.e., if we set a p = 2 p+2+ (2 p+2) , the contributions from
 
the correction-divisors get multiplied essentially by 21p . Hence, to get rid of the
higher order contributions, we need an extra factor of 4, which will be achieved if
we take the correction divisor to be

4 (2 p + 2)2 + 4
a p = (2 p + 2) + = . (136)
(2 p + 2) (2 p + 2)

85It may be noted that among all possible correction divisors of the type a = 2 p + m, where m
p
is an integer, the choice of m = 2 is optimal, as in all other cases there will arise a term proportional to
p in the numerator of the inaccuracy E( p).
264 K. Ramasubramanian and M. D. Srinivas

Then, correspondingly, we have

4 (2 p − 2)2 + 4
a p−2 = (2 p − 2) + = . (137)
(2 p − 2) (2 p − 2)

We can then calculate the inaccuracy to be

 
 
 1 1  1
E( p) =  + −
 4 4  p
(2 p − 2) + (2 p + 2) +
2p − 2 2p + 2
" #
(4 p 3) (16 p 4 + 64)
= 4
−
(4 p + 16) 4 p(4 p 4 + 16)
−4
= . (138)
( p5 + 4 p)

Clearly, the sthaulya with this (second order) correction divisor has improved
considerably, in that it is now proportional to the inverse fifth power of the odd
number.86
At this stage, we may display the result obtained for the circumference with the
correction term as follows. If only the first order correction (132) is employed, we
have
 
1 ( p−1) 1 ( p+1) 1
C = 4d 1 − + . . . + (−1) 2 + (−1) 2 . (139)
3 p (2 p + 2)

If the second order correction (136) is taken into account, we have

 
 1 ( p−1) 1 ( p+1) 1 
C = 1 − 3 + . . . + (−1)
4d  2 + (−1) 2 
p 4 
(2 p + 2) +
(2 p + 2)
 ( p + 1) 
1 ( p−1) 1 ( p+1)
2
= 4d 1 − + . . . + (−1) 2 + (−1) 2 . (140)
 
3 p ( p + 1)2 + 1

86It may be noted that if we take any other correction-divisor a = 2 p + 2 + m , where m is

p (2 p+2)
an integer, we will end up having a contribution proportional to p2 in the numerator of the inaccuracy
E( p), unless m = 4. Thus the above form (136) is the optimal second order choice for the correction-
divisor.
Development of Calculus in India 265

The verse due to Mādhava that we cited earlier as defining the infinite series for
π
4 is, in fact, the first of a group of four verses that present the series along with
the above end-correction.87

v.ya;a;sea va;a;a:=+a;Da;a;na;h;tea .+.pa;&+.tea v.ya;a;sa;sa;a;ga:=+a;Æa;Ba;h;tea Á

ñÍ
É +;a;a;Ba;€+:mx
;a:ˆa;Za:=+a;a;d;a;va;Sa;ma;sa;*
Ë ù a;NMa .~vMa :pxa;Ta;k , k Ò +:ma;a;t,a k u +:ya;Ra;t,a Á Á
Á
ñÍ
ya;tsa;* ÉË+;a;ya;aY:ˆa
ù h:=+Ne a k x +:te a ; a;na;v x a:†a;a &+. a;ta;~tu a . j a;a;Æ a ;ma;ta;ya;a Á
Á
ñÍ
ta;~ya;a +:DvRa;ga;ta;a;ya;a .sa;ma;sa;* ËÉ +;a;a
ù ta;‰+lM gua;Na;ea Y;ntea .~ya;a;t,a Á Á
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 ;Da;nea [ea;pa O;;va k+.=+Na;a;yaH Á Á

l+.b.DaH :pa;a:=+a;DaH .sUa;[ma;ea ba;hu;kx +:tva;ea h:=+Na;ta;ea Y;a;ta;sUa;[maH .~ya;a;t,a Á Á

The diameter multiplied by four and divided by unity. 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.

Take half of the succeeding even number as the multipler at whichever [odd]
number the division process is stopped, because of boredom. The square of that
[even number] added to unity is the divisor. Their ratio has to be multiplied by
the product of the diameter and four as 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 cir-
cumference is very accurate; in fact more accurate than the one which may be
obtained by continuing the division process [with large number of terms in the
series].

Continuing this process further, Yuktibhās.ā presents the next order correction-
term which is said to be even more accurate:88

ñÍ
É +;a;a;d;l+.
A;ntea .sa;ma;sa;*
Ë ù va;gRaH .sEa;k+:ea gua;NaH .sa O;;va :pua;naH ÁÁ
Á
ñÍ
yua;ga;gua;a;Na;ta;ea .+.pa;yua;taH É +;a;a;d;l+.
.sa;ma;sa;*
Ë ù h;ta;ea Ba;vea;d, h;a:=H Á
Á

At the end, [i.e., after terminating the series at some point, apply the correction
term with] the multiplier being square of half of the [next] even number plus 1,
and the divisor being four times the same multiplier with 1 added and multiplied
by half the even number.

87Kriyākramakarı̄ on Lı̄lāvatı̄, cited above (fn. 14), comm. on verse 199, p. 379.
88Ganita-yukti-bhāsā, cited above, p. 82; Also cited in Yukti-dı̄pikā on Tantrasaṅgraha,
. .
cited above (fn. 49), comm. on verse 2.1, p. 103.
266 K. Ramasubramanian and M. D. Srinivas

In other words,89
 2
p+1
+1
1 2
=  
ap p+1
[( p + 1)2 + 4 + 1]
2
1
= . (141)
4
(2 p + 2) +
16
2p + 2 +
2p + 2

π
Hence, a much better approximation for 4 is:90

 2
p+1
+1
π 1 1 1 1 2
= 1 − + − + ···+ −  . (142)
4 3 5 7 p p+1
[( p + 1)2 + 4 + 1]
2

89The inaccuracy or sthaulya associated with this correction can be calculated to be

2304
E( p) = .
(64 p7 + 448 p5 + 1792 p3 − 2304 p)

The inaccuracy now is proportional to the inverse seventh power of the odd-number. Again it can be
shown that the number 16 in (141) is optimally chosen, in that any other choice would introduce a term
proportional to p2 in the numerator of E( p), given above.
In fact, it has been noted by C. T. Rajagopal and M. S. Rangachari that D. T. Whiteside has shown
(personal communication of D. T. Whiteside cited in C. T. Rajagopal and M. S. Rangachari, ‘On an
untapped source of medieval Kerala mathematics’, Arch. for Hist. Sc. 35(2), 89–102, 1978), that the
end correction-term can be exactly represented by the following continued fraction

1 1
= .
ap 22
(2 p + 2) +
42
(2 p + 2) +
62
(2 p + 2) +
(2 p + 2) + . . .

90It may be noted that this correction term leads to a value of π , which is accurate up to 11 decimal
places, when we merely evaluate terms up to n = 50 in the series (142). Incidentally the value of π ,
given in the rule vibudhanetra..., attributed to Mādhava that was cited in the beginning of Section 13,
is also accurate up to 11 decimal places.
Development of Calculus in India 267

15. Transforming the Mādhava series for better convergence

After the estimation of end-correction terms, Yuktibhās.ā goes on to outline a

method of transforming the Mādhava series (by making use of the above end-
correction terms) to obtain new series that have much better convergence prop-
erties. We now reproduce the following from the English translation of Yuk-
tibhās.ā:91
Therefore, the circumference (of a circle) can be derived in taking into consider-
ation what has been stated above. A method for that is stated in the verse
ú È
.sa;ma;pa:ãÁ îå RÁ ;mUa;l+.yua;ta;aH
*.a;a;h;ta;ya;ea ya;a .+.pa;a;dùÅ;a;yua:ja;Ma . a;tua.*+ ta;a;Æa;BaH Á

:Sa;ea;q+.Za;gua;a;Na;ta;a;t,a v.ya;a;sa;a;d, :pxa;Ta;ga;a;h;tea;Sua ;a;va;Sa;ma;yua;teaH Á

.sa;ma;P+.l+.yua;a;ta;ma;pa;h;a;ya .~ya;a;a;d;;v.ya;a;sa;sMa;Ba;vaH :pa;a:=+a;DaH Á Á (I)

The fifth powers of the odd numbers (1, 3, 5 etc.) are increased by
4 times themselves. The diameter is multiplied by 16 and it is suc-
cessively divided by the (series of) numbers obtained (as above). The
odd (first, third etc.) quotients obtained are added and are subtracted
from the sum of the even (the second, fourth etc.) quotients. The
result is the circumference corresponding to the given diameter.
Herein above is stated a method for deriving the circumference. If the cor-
rection term is applied to an approximate circumference and the amount of inac-
curacy (sthaulya) is found, and if it is additive, then the result is higher. Then
it will become more accurate when the correction term obtained from the next
higher odd number is subtracted. Since it happens that (an approximate circum-
ference) becomes more and more accurate by making corrections in succeeding
terms, if the corrections are applied right from the beginning itself, then the cir-
cumference will come out accurate. This is the rationale for this (above-stated
result).
When it is presumed that the correction-divisor is just double the odd number,
the following is a method to obtain the (accurate) circumference by a correction
for the corresponding inaccuracy (sthaulyām . śa-parihāra), which is given by
the verse:
, va;a;a:=+a;Da;a;na;h;ta;a;t,a :pxa;Ta;ga;a;‹Ma ˆya;a;dùÅ;a;yua;a;gva;mU
v.ya;a;sa;a;d / a;l+.Ga;nEaH Á
È
îå Á ya;a;sea .~va;mxa;NMa k x +:tva;a :pa;a:=+a;Da:=+a;nea;yaH Á Á (II)
;a:ˆa.*+;v. Ò +:ma;ZaH
k
The diameter is multiplied by 4 and is divided, successively, by the
cubes of the odd numbers beginning from 3, which are diminished by
these numbers themselves. The diameter is now multiplied by three,
and the quotients obtained above, are added to or subtracted from,
alternatively. The circumference is to be obtained thus.
If, however, it is taken that half the result (of dividing) by the last even number
is taken as the correction, there is a method to derive the circumference by that
way also, as given by the verse
x +:ta;ya;eaH È
îå Á a;va;Sk+:}Bea Á
õùÅ;a;a;a;d;yua:ja;Ma va;a k
d v.yea;k+:a h;a:=+a;d
, ;a;d
õ ;a;na.*+;

;Da;na;m,a +.Na;ma;ntea Y;ntya;ea;DvRa;ga;ta;Ea:ja;kx +:a;ta;a;d

õR ;sa;a;h;ta;a h:=+~ya;a;DRa;m,a Á Á (III)
The squares of even numbers commencing from 2, diminished by

91Ganita-yukti-bhāsā, cited above, Section 6.9, pp. 80–82, 205–07, 402–04.

. .
268 K. Ramasubramanian and M. D. Srinivas

one, are the divisors for four times the diameter. (Make the several
divisions). The quotients got by (the division) are alternately added
to or subtracted from twice the diameter. In the end, divide four times
the diameter by twice the result of squaring the odd number following
the last even number to which is added 2.

The method of sthaulya-parihāra, outlined above, essentially involves incor-

porating the correction terms into the series from the beginning itself. Let us recall
that inaccuracy or sthaulya at each stage is given by
 
1 1 1
E( p) = + − . (143)
a p−2 ap p

The series for the circumference (112) can be expressed in terms of these sthaulyas
as follows:
      
1 1 1 1 1 1 1
C = 4d 1 − + + − − + − −...
a1 a1 a3 3 a3 a5 5
  
1
= 4d 1 − + E(3) − E(5) + E(7) − . . . . (144)
a1

Now, by choosing different correction-divisors a p in (144), we get several

transformed series which have better convergence properties. If we consider the
correction-divisor (136), then using the expression (138) for the sthaulyas, we get

   
1 1 1 1
C = 4d 1 − − 16d − + − ...
5 (35 + 4.3) (55 + 4.5) (75 + 4.7)
 
1 1 1
= 16d − + −... . (145)
(15 + 4.1) (35 + 4.3) (55 + 4.5)

The above series is given in the verse samapañcāhatayoh. . . .(I). Note that each
term in the above series involves the fifth power of the odd number in the denom-
inator, unlike the original series which only involved the first power of the odd
number. Clearly, this transformed series gives more accurate results with fewer
terms.
If we had used only the lowest order correction (132) and the associated sthaulya
(134), instead of the correction employed above, then the transformed series is the
one given in the verse vyāsād vāridhinihatāt. . .(II)
 
3 1 1 1
C = 4d + 3 − 3 + 3 −... . (146)
4 (3 − 3) (5 − 5) (7 − 7)

Note that the denominators in the above transformed series are proportional to the
third power of the odd number.
Development of Calculus in India 269

Even if we take non-optimal correction-divisors, we often end-up obtaining

interesting series. For instance, if we take a non-optimal correction-divisor, say of
the form a p = 2 p, then the sthaulya is given by

1 1 1
E( p) = + −
(2 p − 4) 2 p p
1
=
( p 2 − 2 p)
1
= . (147)
( p − 1)2 − 1

Then, the transformed series will be the one given in the verse dvyādiyujām
.
vā kr.tayo. . .(III)92

 
1 1 1 1
C = 4d + − + +... . (148)
2 (22 − 1) (42 − 1) (62 − 1)

16. Derivation of the Mādhava series for Rsine and Rversine

16.1. First and second order differences of Rsines

We shall now outline the derivation of Mādhava series for Rsine (bhujā-jyā) and
Rversine (śara), as given in Yuktibhās.ā.93 Yuktibhās.ā begins with a discussion
of the first and second order Rsine-differences and derives an exact form of the
result of Āryabhat.a that the second-order Rsine-differences are proportional to
the Rsines themselves. We had briefly indicated this proof in Section 5.3.
Here we are interested in obtaining the Mādhava series for the jyā and śara of
an arc of length s indicated by EC in Figure 9. This arc is divided into n equal
arc bits, where n is large. If the arc length s = Rθ , then the j -th pin.d.a-jyā, B j is
given by94
   
js jθ
B j = jyā = R sin . (149)
n n

92The verse III in fact presents the series (148) along with an end correction-term of the form
(−1) p 4d .
2( p+1)2 +2
93Yuktibhāsā, cited earlier, Vol. I Section 16.5, pp. 94–103, 221–233, 417–427.
.
94Figure 9 is essentially the same as Figure 3 considered in section 5 except that the pindajyās B
.. j
are Rsines associated with multiples of the arc-bit ns into which the arc EC = s is divided. In Figure 3,
the B j ’s are the tabular Rsines associated with multiplies of 225′ .
270 K. Ramasubramanian and M. D. Srinivas

The corresponding kot.i-jyā K j , and the śara S j , are given by

   
js jθ
K j = kot.i = R cos , (150)
n n
    
js jθ
S j = śara = R 1 − cos . (151)
n n

Now, C j C j +1 represents the ( j + 1)-th arc bit. Then, for the arc EC j = jns ,
its pin.d.a-jyā is B j = C j P j , and the corresponding kot.i-jyā and śara are K j =
C j T j , S j = E P j . Similarly we have

B j +1 = C j +1 P j +1 , K j +1 = C j +1 T j +1 and S j +1 = E P j +1 . (152)

Cj E

E S
M j+1
α Mj Qj
Cj
Pj
C j+1 M j+1 G
Q j+1
F C j+1 Pj+1
F

N
Tj+1 Uj+1 Tj Uj O

F IGURE 9. Computation of Jyā and Śara by Saṅkalitas.

Let M j +1 be the mid-point of the arc-bit C j C j +1 and similarly M j the mid-

point of the previous ( j -th) arc-bit. We shall denote the pin.d.a-jyā of the arc
E M j +1 as B j + 1 and clearly
2

B j + 1 = M j +1 Q j +1 .
2

The corresponding kot.i-jyā and śara are

K j + 1 = M j +1 U j +1 and S j + 1 = E Q j +1 .
2 2
Development of Calculus in India 271

Similarly,

B j − 1 = M j Q j , K j − 1 = M j U j and S j − 1 = E Q j . (153)
2 2 2

Let α be the chord corresponding to the equal arc-bits ns as indicated in Fig-

ure 9. That is, C j C j +1 = M j M j +1 = α. Let F be the intersection of C j T j
and C j +1 P j +1 , and G of M j U j and M j +1 Q j +1 . The triangles C j +1 FC j and
O Q j +1 M j +1 are similar, as their sides are mutually perpendicular. Thus we have

C j +1 C j C j +1 F FC j
= = . (154)
O M j +1 O Q j +1 Q j +1 M j +1

Hence we obtain
α
B j +1 − B j = K j+ 1 , (155)
R 2
α
K j − K j +1 = S j +1 − S j = B j+ 1 . (156)
R 2

Similarly, the triangles M j +1 G M j and O P j C j are similar and we get

M j +1 M j M j +1 G GMj
= = . (157)
OC j O Pj Pj C j

Thus we obtain
α
B j+ 1 − B j− 1 = Kj, (158)
2 2 R α
K j− 1 − K j+ 1 = Sj+ 1 − Sj− 1 = Bj . (159)
2 2 2 2 R

We define the Rsine-differences (khan.d.a-jyā) 1 j by

1 j = B j − B j −1 , (160)

with the convention that 11 = B1 . From (155), we have

α
1j = K j− 1 . (161)
R 2
272 K. Ramasubramanian and M. D. Srinivas

From (159) and (161), we also get the second order Rsine-differences (the differ-
ences of the Rsine-differences called khan.d.a-jyāntara):

1 j − 1 j +1 = (B j − B j −1) − (B j +1 − B j )
α  
= K j− 1 − K j+ 1
αR 2 2

= Sj+ 1 − Sj− 1
R 2 2
 α 2
= Bj . (162)
R
Now, if the sum of the second-order Rsine-differences, is subtracted from the first
Rsine-difference, then we get any desired Rsine-difference. That is
 
11 − (11 − 12 ) + (12 − 13 ) + . . . + (1 j −1 − 1 j ) = 1 j . (163)

From (162) and (163) we conclude that

 α 2
11 − (B1 + B2 + · · · + B j −1) = 1 j . (164)
R

16.2. Rsines and Rversines from Jyā-saṅkalita

We can sum up the Rversine-differences (159), to obtain the śara, Rversine, at the
midpoint of the last arc-bit as follows:
   
Sn− 1 − S 1 = Sn− 1 − Sn− 3 + . . . . . . S 3 − S 1
2 2
α 2 2 2 2

= (Bn−1 + Bn−2 + . . . + B1 ) . (165)

R
Using (162), the right hand side of (165) can also be expressed as a summation of
the second order differences. From (164) and (165) it follows that the Rversine at
the midpoint of the last arc-bit is also given by
α  
Sn− 1 − S 1 = (11 − 1n ). (166)
R 2 2

Now, since the first Rsine-difference 11 = B1 , any desired Rsine can be obtained
by adding the Rsine-differences; these Rsine-differences have been obtained in
(164). Now, by making use of (164), the last pin.d.a-jyā can be expressed as fol-
lows:

Bn = 1n + 1n−1 + . . . + 11
Development of Calculus in India 273
 α 2  
= n11 − (B1 + B2 . . . + Bn−1 ) + (B1 + B2 . . . + Bn−2 ) + . . . + B1
R

 α 2  
= n B1 − Bn−1 + 2Bn−2 + · · · + (n − 1)B1 . (167)
R

The results (158) – (167), obtained so far, involve no approximations. It is now

shown how better and better approximations to the Rsine and Rversine can be
obtained by taking n to be very large or, equivalently, the arc-bit ns to be very
small. Then, we can approximate the full-chord and the Rsine of the arc-bit by the
arc-bit ns itself. Also, as a first approximation, we can approximate the pin.d.a-jyās
B j in the equations (164), (165) or (167) by the corresponding arcs themselves.
That is

js
Bj ≈ . (168)
n

The result for the Rsine obtained this way is again used to obtain a better approx-
imation for the pin.d.a-jyās B j which is again substituted back into the equations
(165) and (167) and thus by a process of iteration successive better approximations
are obtained for the Rsine and Rversine. Now, once we take B j ≈ jns , we will be
led to estimate the sums and repeated sums of natural numbers (ekādyekottara-
saṅkalita), when the number of terms is very large.

16.3. Derivation of Mādhava series by iterative corrections to jyā and śara

As we noted earlier, the relations given by (165) and (167) are exact. But now we
shall show how better and better approximations to the Rsine and Rversine of any
desired arc can be obtained by taking n to be very large or, equivalently, taking the
arc-bit ns to be very small. Then both the full-chord α, and the first Rsine B1 (the
Rsine of the arc-bit), can be approximated by the arc-bit ns itself, and the Rversine
Sn− 1 can be taken as Sn and the Rversine S 1 may be treated as negligible. Thus
2 2
274 K. Ramasubramanian and M. D. Srinivas

the above relations (165), (167) become95

 s 
S = Sn ≈ (Bn−1 + Bn−2 + . . . + B1 ), (169)
nR
 s 2
B = Bn ≈ s − [(B1 + B2 + . . . + Bn−1 )
nR
+ (B1 + B2 . . . + Bn−2 ) + . . . + B1 ], (170)

where B and S are the Rsine and Rversine of the desired arc of length s and the
results will be more accurate, larger the value of n.
Now, as a first approximation, we take each pin.d.a-jyā B j in (169) and (170) to
be equal to the corresponding arc itself, that is

js
Bj ≈ . (171)
n
Then we obtain for the Rversine
 s h s  s  i
S ≈ (n − 1) + (n − 2) +...
n R   n n
1 s 2
= [(n − 1) + (n − 2) + . . .]. (172)
R n

For large n, we can use the estimate (89) for the sum of integers. Hence (172)
reduces to
  2
1 s
S≈ . (173)
R 2

Equation (173) is the first śara-sam . skāra, correction to the Rversine. We now
substitute our first approximation (171) to the pin.d.a-jyās B j in (170), which gives
the Rsine of the desired are as a second order repeated sum of the pin.d.a-jyās B j .
We then obtain
 2  
1 s 3
B ≈ s− [(1 +2 +. . .+(n −1))+(1 +2 +. . . (n −2))+. . . ]. (174)
R n

95As has been pointed out by one of the reviewers, in the following derivation instead of using
the relation (170), which involves repeated summation of pin. d.ajyās, one could use the much simpler
relation
s
B = Bn ≈ s − (S + Sn−2 + . . . + S1 ),
n R n−1
which essentially follows from (165) and (170). Then we can iterate between the above equation
and (169) which involve considering only sums of powers of integers. Yuktibhās.ā, however, em-
ployes successive iteration between (169) and (170), which involves consideration of repeated sums of
integers.
Development of Calculus in India 275

The second term in (174) is a dvitı̄ya-saṅkalita, the second order repeated sum,
and using the estimate (108), we obtain
 2 3
1 s
B≈s− . (175)
R 1.2.3

Thus we see that the first correction obtained in (175) to the Rsine-arc-difference
(jyā-cāpāntara-sam. skāra), is equal to the earlier correction to the Rversine (śara-
sam. skāra) given in (173) multiplied by the arc and divided by the radius and 3.
It is noted that the results (173) and (175) are only approximate (prāyika),
since, instead of the saṅkalita of the pin.d.a-jyās in (169) and (170), we have only
carried out saṅkalita of the arc-bits. Now that (175) gives a correction to the
difference between the Rsine and the arc (jyā-cāpāntara-sam . skāra), we can use
that to correct the values of the pin.d.a-jyās and thus obtain the next corrections to
the Rversine and Rsine.
Following (175), the pin.d.a-jyās may now be taken as
  3 
 2 js
js 1  n 
Bj ≈ − . (176)
n R 1.2.3


If we introduce (176) in (169), we obtain

  
1 s 2
S ≈ [(n − 1) + (n − 2) + . . .]
R n
 s   1 2  s 3  1 
− [(n − 1)3 + (n − 2)3 + . . .]. (177)
nR R n 1.2.3

The first term in (177) was already evaluated while deriving (173). The second
term in (177) can either be estimated as a summation of cubes (ghana-saṅkalita),
or as a tr.tı̄ya-saṅkalita, third order (repeated) summation, because each individ-
ual term there has been obtained by doing a second-order (repeated) summation.
Hence, recollecting our earlier estimate (110) for these saṅkalitas, we get
  2  3
1 s 1 s4
S≈ − . (178)
R 1.2 R 1.2.3.4

Equation (178) gives a correction (śara-sam . skāra) to the earlier value (173) of
the Rversine, which is nothing but the earlier correction to the Rsine-arc difference
. skāra) given in (175) multiplied by the arc and divided by the
(jyā-cāpāntara-sam
radius and 4.
276 K. Ramasubramanian and M. D. Srinivas

Again, if we use the corrected pin.d.a-jyās (176) in the expression (170) for the
Rsine, we obtain

1 2  s 3
 
B ≈ s− [(1 + 2 + .. + (n − 1)) + (1 + 2 + .. + (n − 2)) + ..]
R n
 4  
1 s 5
+
R n
 h
1 i
× (13 + 23 + ... + (n − 1)3 ) + (13 + 23 + ... + (n − 2)3 ) + ..
1.2.3
 2 3  4
1 s 1 s5
≈ s− + . (179)
R 1.2.3 R 1.2.3.4.5

The above process can be repeated to obtain successive higher order correc-
tions for the Rversine and Rsine: By first finding a correction (jyā-cāpāntara-
sam. skāra) for the difference between the Rsine and the arc, using this correction
to correct the pin.d.a-jyās B j , and using them in equations (169) and (170) get the
next correction (śara-sam . skāra) for the Rversines, and the next correction (jyā-
cāpāntara-sam . skāra) for the Rsine-arc-difference itself, which is then employed
to get further corrections iteratively. In this way we are led to the Mādhava series
for jyā and śara given by

 2  4
1 s3 1 s5
B = R sin(s) = s− +
R (1.2.3) R (1.2.3.4.5)
 6
1 s7
− + ...,
R (1.2.3.4.5.7)
  2  3
1 s 1 s4
S = R vers(s) = −
R 2 R (1.2.3.4)
 5
1 s6
+ − ... . (180)
R (1.2.3.4.6)

That is,

θ3 θ5 θ7
sin θ = θ− + − +... ,
(1.2.3) (1.2.3.4.5) (1.2.3.4.5.6.7)
θ2 θ4 θ6
vers θ = − + −... . (181)
(1.2) (1.2.3.4) (1.2..4.5.6)
Development of Calculus in India 277

17. Instantaneous velocity and derivatives

As we saw in Section 6.1, the mandaphala or the equation of centre for a planet
1µ is given by
r 
0
R sin(1µ) = R sin(M − α), (182)
R
where r0 is the mean epicycle radius, M is the mean longitude of the planet and α
the longitude of the apogee. Further as we noted earlier, Muñjāla, Āryabhat.a II
and Bhāskara II used the approximation

R sin(1µ) ≈ 1µ, (183)

in (182) and obtained the following expression as correction to the instantaneous

velocity of the planet:

d r 
0 d
(1µ) = R cos(M − α) (M − α). (184)
dt R dt
Actually the instantaneous velocity of the planet has to be evaluated from the more
accurate relation
h r  i
0
1µ = R sin−1 R sin(M − α) . (185)
R
The correct expression for the instantaneous velocity which involves the derivative
of arc-sine function has been given by Nı̄lakan.t.ha in his Tantrasaṅgraha.96

. a;ndÒ+ba;a;hu;P+.l+.va;gRa;Za;ea;a;Da;ta;a:ˆa:$ya;k+:a;kx +:a;ta;pa;de;na .sMa;h:=e+t,a Á

ta:ˆa k+:ea;a;f;P+.l+.a;l+.a;‹a;k+:a;h;ta;Ma :ke +:ndÒ+Bua;a;€+:a:=+h ya;ƒa l+.Bya;tea Á Á

ta;a;d
õ ;Za;ea;Dya mxa;ga;a;a;d;ke ga;teaH ;Æa;[a;pya;ta;a;Æa;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:ja;a .=+vea:=+a;pa Á Á

. iphala [r0 cos(M − α)] in minutes and the daily motion

Let the product of the kot
of the manda-kendra d(M−α) dt be divided by the square root of the square of
q 
the bāhuphala subtracted from the square of trijyā R 2 − r02 sin2 (M − α) .
The result thus obtained has to be subtracted from the daily motion of the Moon
if the manda-kendra lies within six signs beginning from Mr.ga and added if it

96Tantrasaṅgraha, cited above (fn. 52), verses 2.53–54, pp.169–170. Elsewhere, Nı̄lakantha
..
has ascribed these verses to his teacher Dāmodara (Jyotirmı̄mām
. sā, Ed. by K. V. Sarma, VVRI,
Hoshiarpur 1977, p. 40).
278 K. Ramasubramanian and M. D. Srinivas

lies within six signs beginning from Karkat.aka. The result gives a more accu-
rate value of the Moon’s angular velocity. In fact, the procedure for finding the
instantaneous velocity of the Sun is also the same.

If (M − α) be the manda-kendra, then the content of the above verse can be

expressed as

d(M − α)
d h −1  r0 i r0 cos(M − α)
sin sin(M − α) = q dt . (186)
dt R 2 2 2
R − r0 sin (M − α)

The instantaneous velocity of the planet is given by

d(M − α)
d d r0 cos(M − α)
µ = (M − α) − q dt . (187)
dt dt 2 2 2
R − r0 sin (M − α)

Here, the first term in the RHS represents the mean velocity of the planet and the
second term the rate of change in the mandaphala given by (186).
In his Āryabhat.ı̄ya-bhās.ya, Nı̄lakan.t.ha explains how his result is more correct
than the traditional result of Muñjāla and Bhāskarācārya:97

A;taH :P+.l+.sa;a;}yMa ku +:taH ? . . . :pua;na:=+a;pa ya;ea ;a;va;Zea;SaH ta:ˆa k+:ea;a;f:$ya;a;gua;a;Na -

ta;~ya ;a:ˆa:$ya;ya;a h:=+Na;mua;€+:m,a , I+.h k+:ea;a;f;P+.l+.gua;a;Na;ta;~ya :ke +:ndÒ+Ba;ea;ga;~ya

d;eaH P+.l+.k+:ea;f“a;a h:=+Na;mua;€+:m,a I+.a;ta Á .tea;na ta;tP+.lM . a;a;pa;a;kx +:tMa Bua:ja;a;P+.l -

ga;a;taH .~ya;a;t,a Á k+:Ta;m,a ?
. a;a;pa;ga;a;ta;sa;}ba;a;nDa:$ya;a;ga;tya;a;na;ya;ne
// a ya;t,a :ˆEa:=+a;a;Za;k+:mua;€M , .$ya;a;ga;tya;a

. a;a;pa;ga;tya;a;na;ya;nea ta;a;d
õ ;pa:=+a;tMa k+:mRa k+:a;yRa;m,a Á ta:ˆa :pUa;va;eRa;€e k+:mRa;a;Na

:ˆEa:=+a;a;Za;k+:d
õ ;yea;na ya;a d;eaH P+.l+.ga;a;taH A;a;na;a;ta;a ta;Ma v.ya;a;sa;a;DeRa;na h;tva;a

d;eaH P+.l+.k+:ea;f“a;a &+.tva;a ta;ƒa;a;pa;ga;a;ta;lR+.Bya;a Á ta:ˆea;dM :ˆEa:=+a;a;Za;k+:m,a ...

Hence, how can the results be equal? . . . Again the distinction being: there it was
prescribed that the multiplier kot.i-jyā was to be divided by trijyā, [but] here
it has been prescribed that the product of kot.iphala and the rate of change of
kendra be divided by kot.i of the doh.phala (doh.phalakot. yā).98 . . .

97Āryabhatı̄ya of Āryabhata, Ed. with Bhāsya of Nı̄lakantha Somayājı̄ by K. Sāmbaśiva

. . . ..
Śāstrı̄, Trivandrum Sanskrit Series 110, Trivandrum 1931, comm. on Kālakriyāpāda 22–25, pp. 62–
63.
98The terms dohphala and kotiphala refer to r0 sin(M − α) and r0 cos(M − α) respectively.
. . q R R
r
Hence, the term doh.phalakot. i refers to 1 − ( R0 sin(M − α))2 .
Development of Calculus in India 279

17.1. Acyuta’s expression for instantaneous velocity involving the derivative

of ratio of two functions

In the third chapter of his Sphut.anirn.ayatantra, Acyuta Pis.ārat.i (c. 1550–

1621), a disciple of Jyes.t.hadeva, discusses various results for the instantaneous
velocity of a planet depending on the form of equation of centre (manda-sam . skāra).
He first presents the formula involving the derivative of arc-sine function given by
Nı̄lakan.t.ha (in the name of (manda)-sphut.agati) as follows:99

k+:ea;a;f;P+.l+.a;h;ta;ke +:ndÒ+ga;tea;yRa;d, d;eaH P+.l+.k+:ea;a;f;k+:ya;a;‹a;ma;nea;na Á

å
h;a;na;yua;ta;a;mxa;ga;k+:kR +:f;k+:a;dùÅ;a;ea;mRa;Dya;ga;a;ta;BRa;va;a;ta .~.Pu +.f;Bua;a;€ H Á Á

Acyuta also gives the formula for the instantaneous velocity of a planet if one
were to follow a different model proposed by Munjāla for the equation of centre,
according to which mandaphala is given by

r0
sin(M − α)
1µ =  Rr , (188)
0
1 − cos(M − α)
R

instead of (182), where 1µ is small. If one were to use this formula for man-
daphala for finding the true longitude of the planet, then it may be noted that the
instantaneous velocity will involve the derivative of the ratio of two functions both
varying with time. Taking note of this, Acyuta observes:100
ïîéå
x +:t=+:
k a;~ya ma;a;nd;pa;a:=+Dea;a;nRa:ja;k+:NRa;tua;ya;Ea

vxa;a:;d
Äâ ;[a;ya;a;a;va;a;ta ma;tea k+:a;Ta;taH kÒ +:ma;ea Y;ya;m,a Á

A;DRa;~ya ma;a;nd;pa;a:=+DeaH [a;ya;vxa;a:;d

Äâ ;pa;[ea

yua;€M ;a;kÒ +:ya;a;kÒ +:ma;ma;Ta :pra;a;ta;pa;a;d;ya;a;maH Á Á

The procedure that was prescribed earlier is with reference to the School that
conceives of the increase and decrease in the circumference of the manda-vr.tta
in accordance with the karn . a. With reference to the School that conceives of
increase and decrease only according to the half [of it], now we prescribe the
appropriate procedure to be adopted.

Acyuta then proceeds to give the correct expression for the instantaneous velocity
of a planet in Munjāla’s model:101

99Sphutanirnayatantra of Acyuta Pisārati, Ed. by K. V. Sarma, VVRI, Hoshiarpur 1974,

. . . .
p. 19.
100Ibid., p. 20.
101Ibid., p. 21.
280 K. Ramasubramanian and M. D. Srinivas

kx +:ta;k+:ea;a;f;P+.lM ;a:ˆa:ja;a;va;ya;a ;a;va;&+.tMa d;eaH P+.l+.va;gRa;ta;~tua ya;t,a Á

mxa;ga;k+:kR +:f;k+:a;a;d;ke Y;mua;na;a yua;ta;h;a;nMa :P+.l+.ma:ˆa;k+:ea;a;f:ja;m,a Á Á

È
îå Á x +:ta;k+:ea;f;a;P+.l+.ya;a ;a:ˆa:ja;a;va;ya;a Á
;a;d;na;ke +:ndÒ+ga;a;ta.*+;mu a:;d
Äâ :=e+t,a k

:P+.l+.pUa;vRa;P+.lE+.k+:ta;ea d;lM ;a;d;na;Bua;€e +.=+a;pa .sMa;~kx +:a;ta;BRa;vea;t,a Á Á

Having applied the kot.iphala to trijyā [positively or negatively depending upon

the mandakendra], let the square of the doh.phala be divided by that. This may
be added to or subtracted from the kot.iphala depending on whether it is Mr.gādi
or Karkyādi. The product of this [result thus obtained] and the daily motion of
the manda-kendra divided by the kot.iphala and applied to trijyā will be the
correction to the daily motion.

Thus according to Acyuta, the correction to the mean velocity of a planet in

order to obtain its instantaneous velocity is given by
r 2
0
r
0
 sin(M − α)
cos(M − α) +  Rr 
R 0
cos(M − α)
1−
d(M − α)
R  , (189)
 r0 dt
1 − cos(M − α)
R
which is nothing but the derivative of the expression given in (188).

Acknowledgment
We are thankful to Prof. D. Mumford and Prof. C. S. Seshadri for organizing
the seminar on History of Mathematics at Chennai Mathematical Institute during
January–February 2008, which gave us an opportunity to put together some of the
material on development of calculus in India. We are grateful to our collaborator
Prof. M. S. Sriram and also to Prof. R. Sridharan and Prof. Raja Sridharan for their
valuable suggestions and continued encouragement. We would like to specially
thank the reviewers of the paper for their valuable comments and suggestions.

References

Major Source Works

(1) The Śulbasūtras of Baudhāyana, Āpastamba, Kātyāyana and Mānava:
Ed. and Tr. by S. N. Sen and A. K. Bag, Indian National Science Acad-
emy, New Delhi 1983.
(2) Āryabhat.ı̄ya of Āryabhat.a (c. 499): Ed. and Tr. by K. S. Shukla and
K. V. Sarma, Indian National Science Academy, New Delhi 1976.
(3) Āryabhat.ı̄yabhās.ya (c. 629) of Bhāskara I: Ed. by K. S. Shukla, Indian
National Science Academy, New Delhi 1976. Eng Tr. of Gan.itapāda by
Agathe Keller, 2 Vols. Birkhauser 2006.
Development of Calculus in India 281

(4) Brāhmasphut.asiddhānta (c. 628) of Brahmagupta: Mathematics Chap-

ters (XIII, XVIII) translated in H. T. Colebrooke, Algebra with Arithmetic
and Mensuration from the Sanskrit of Brahmagupta and Bhāskara, Lon-
don 1817. Ed. by Sudhakara Dvivedi, Varanasi 1902.
(5) Baks.āli Manuscript (c. 8th cent): Ed. and Tr. by T. Hayashi, in The
Baks.āli Manuscript: An Ancient Indian Mathematical Treatise, Egbert
Forsten, Groningen 2005.
(6) Triśatikā (c. 750) of Śrı̄dhara: Ed. by Sudhakara Dvivedi, Benaras 1899.
Tr. by N. Ramanujachari and G. R. Kaye, Bibliotheca Mathematica, 13,
203–217, 1912.
(7) Bhās.ya (c. 800) of Govindasvāmin on Mahābhāskarı̄ya of Bhāskara I:
Ed. by T. S. Kuppanna Sastri, Madras 1957.
(8) Gan.itasārasaṅgraha (c. 850) of Mahāvı̄ra: Ed. and Tr. by M. Ran-
gacharya, Govt. Press, Madras 1912. Ed. with Hindi Tr. by L C Jain, Jain
Samskrit Samrakshaka Sangha, Sholapur 1963.
(9) Vāsanā (c. 860) of Pr.thūdakasvāmin on Brāhmasphut.asiddhānta of
Brahmagupta: partially Ed. by Ramaswarup Sharma, 4 Vols, New Delhi
1966; Chapter 21, Ed. and Tr. by S. Ikeyama, INSA, New Delhi 2003.
(10) Lı̄lāvatı̄ (c. 1150) of Bhāskarācārya II: Tr. in H. T. Colebrooke, Alge-
bra with Arithmetic and Mensuration from the Sanskrit of Brahmagupta
and Bhāskara, London 1817. Ed. with Bhāskara’s Vāsanā and Bud-
dhivilāsinı̄ of Gan.eśa Daivajña (c. 1545) by V. G. Apte, 2 Vols, Pune
1937.
(11) Bı̄jagan.ita (c. 1150) of Bhāsakarācārya II: Tr. in H. T. Colebrooke, Al-
gebra with Arithmetic and Mensuration from the Sanskrit of Brahmagupta
and Bhāskara, London 1817. Ed. with Bhāskara’s Vāsanā by Muralid-
hara Jha, Benares 1927, Chaukhamba Rep. 1994.
(12) Siddhāntaśiroman.i (c. 1150) of Bhāskarācārya II: Ed. with Bhāskara’s
Vāsanā by Bapudeva Sastri, Varanasi 1866. Ed. with Bhāskara’s
Vāsanā and Nr.sim . ha Daivajña’s Vāsanāvārttika by Muralidhara
Chaturveda, Varanasi 1981. Grahagan.itādhyāya Tr. by D. Arkaso-
mayaji, Kendriya Sanskrit Vidya-peetha, Tirupati 1980, Rep. 2000.
(13) Gan.itakaumudı̄ (c. 1356) of Nārāyan.a Pan.d.ita: Ed. by Padmakara
Dvivedi, 2 Vols, Varanasi 1936, 1942. Chapters 13–14 Ed. and Tr.
by T Kusuba, Combinatorics and Magic-squares in India: A Study of
Nārāyan.a Pan.d.ita’s Gan.itakaumudı̄, Chapters 13–14, PhD Disserta-
tion, Brown University 1993. Tr. by Parmanand Singh, Gan.itabhāratı̄
1998–2002.
282 K. Ramasubramanian and M. D. Srinivas

(14) Tantrasaṅgraha (c. 1500) of Nı̄lakan.t.ha Somayājı̄: Ed. with

Laghuvivr.ti of Śaṅkara Vāriyar by Surnad Kunjan Pillai, Trivandrum
1958. Ed. with Yuktidı̄pikā of Śaṅkara Vāriyar by K. V. Sarma,
Hoshiarpur 1977.
(15) Āryabhat.ı̄yabhās.ya (c. 1502) of Nı̄lakan.t.ha Somayājı̄ on Āryabhat.ı̄ya
of Āryabhat.a: Vols. 1 and 2 on Gan.ita and Kālakriyāpāda, Ed. by
K. Sāmbaśiva Śastrı̄, Trivandrum 1930, 1931; Vol. 3 on Golapāda. Ed.
by Surnad Kunjan Pillai, Trivandrum 1957.
(16) Gan.ita-yukti-bhās.ā (c. 1530) of Jyes.t.hadeva (in Malayalam):
Gan.itādhyāya, Ed., with Notes in Malayalam, by Ramavarma Thampu-
ran and A. R. Akhileswarayyar, Trichur 1948. Ed. with Tr. by K. V. Sarma
with Explanatory Notes by K. Ramasubramanian, M. D. Srinivas and
M. S. Sriram, 2 Vols, Hindustan Book Agency, Delhi 2008, Rep. Springer
2009.
(17) Kriyākramakarı̄ (c. 1535) of Śaṅkara Vāriyar on Lı̄lāvatı̄ of
Bhāskarācārya II: Ed. by K. V. Sarma, Hoshiarpur 1975.
(18) Yuktidı̄pikā (c. 1535) of Śaṅkara Vāriyar on Tantrasaṅgraha of
Nı̄lakan.t.ha Somayājı̄: Ed. by K. V. Sarma, Hoshiarpur 1977.
(19) Buddhivilāsinı̄ (c. 1545) of Gan.eśa Daivajña on Lı̄lāvatı̄ of
Bhāskarācārya II: Ed. by V. G. Apte, 2 Vols, Pune 1937.
(20) Bı̄janavāṅkura (c. 1600) of Kr.s.n.a Daivajña on Bı̄jagan.ita of
Bhāskarācār-ya II: Ed. by V. G. Apte, Pune 1930. Ed. by T. V. Radha
Krishna Sastry, Tanjore 1958.
(21) Sphut.anirn.ayatantra (c. 1600) of Acyuta Piśārat.i: Ed. with Acyuta’s
commentary by K. V. Sarma, Hoshiarpur 1974.

Some Overviews of Indian Mathematics

(1) B. B. Datta and A. N. Singh, History of Hindu Mathematics, 2 parts, La-
hore 1935, 1938. Rep. Bharatiya Kala Prakasan, Delhi 2004.
(2) C. N. Srinivasa Iyengar, History of Indian Mathematics, Calcutta 1967.
(3) A. K. Bag, Mathematics in Ancient and Medieval India, Varanasi 1979.
(4) T. A. Sarasvati Amma, Geometry in Ancient and Medieval India, Delhi
1979. Rep. 2007.
(5) G. G. Joseph, The Crest of the Peacock: The Non-European Roots of
Mathematics, Peguin 1990, 2nd Ed., Princeton 2000.
Development of Calculus in India 283

(6) S. Parameswaran, The Golden Age of Indian Mathematics, Swadeshi Sci-

ence Movement, Kochi 1998.
(7) G. Emch, M. D. Srinivas and R. Sridharan Eds., Contributions to the His-
tory of Mathematics in India, Hindustan Book Agency, Delhi 2005.
(8) C. K. Raju, Cultural Foundations of Mathematics: The Nature of Mathe-
matical Proof and the Transmission of the Calculus from India to Europe
in the 16t h c.CE, Pearson Education, Delhi 2007.
(9) Kim Plofker, ‘Mathematics in India’, in V. J. Katz Ed., The Mathematics
of Egypt, Mesopotamia, China, India and Islam: A Source Book, Prince-
ton 2007, pp. 385–513.
(10) Kim Plofker, History of Mathematics in India: From 500 BCE to 1800
CE, Princeton 2008.

Some Articles
(1) C. M. Whish, ‘On the Hindu quadrature of the circle, and the infinite
series of the proportion of the circumference to the diameter exhibited in
the four Shastras, the Tantrasangraham, Yucti Bhasa, Carana Paddhati and
Sadratnamala’, Tran. Roy. As. Soc. (G.B.) 3, 509–523, 1834.
(2) Bapudeva Sastri, ‘Bhaskara’s knowledge of the differential calculus’, Jour.
As. Soc. Bengal 27, 213–216, 1858.
(3) William Spotiswoode, ‘Note on the supposed discovery of the principle
of the differential calculus by an Indian astronomer’, Jour. Royal As. Soc.
17, 221–222, 1860.
(4) P. C. Sen Gupta, ‘Infinitesimal calculus in Indian mathematics’, Jour.
Dep. Letters. Calcutta Univ., 22, 1–17, 1931.
(5) K. Mukunda Marar, ‘Proof of Gregory’s series’, Teacher’s Magazine 15,
28–34, 1940.
(6) K. Mukunda Marar and C. T. Rajagopal, ‘On the Hindu quadrature of the
circle’, J.B.B.R.A.S. 20, 65–82, 1944.
(7) K. Mukunda Marar and C. T. Rajagopal, ‘Gregory’s series in the mathe-
matical literature of Kerala’, Math. Student 13, 92–98, 1945.
(8) A. Venkataraman, ‘Some interesting proofs from Yuktibhās.ā’, Math. Stu-
dent 16, 1–7, 1948.
(9) C. T. Rajagopal, ‘A neglected chapter of Hindu mathematics’, Scr. Math.
15, 201–209, 1949.
284 K. Ramasubramanian and M. D. Srinivas

(10) C. T. Rajagopal and A. Venkataraman, ‘The sine and cosine power series
in Hindu mathematics’, J.R.A.S.B. 15, 1–13, 1949.
(11) C. T. Rajagopal and T. V. V. Aiyar, ‘On the Hindu proof of Gregory’s
series’, Scr. Math. 17, 65–74, 1951.
(12) C. T. Rajagopal and T. V. V. Aiyar, ‘A Hindu approximation to Pi’, Scr.
Math. 18. 25–30, 1952.
(13) R. C. Gupta, ‘Second order interpolation in Indian mathematics up to the
fifteenth century’, Ind. Jour. Hist. Sc., 4, 86–98, 1969.
(14) R. C. Gupta, ‘An Indian form of third order Taylor series approximation
of the sine’, Hist. Mathematica, 1, 687–89, 1974.
(15) C. T. Rajagopal and M. S. Rangachari, ‘On an untapped source of me-
dieval Keralese mathematics’, Arch. for Hist. of Ex. Sc. 18, 89–101,
1978.
(16) B. B. Datta and A. N. Singh (Rev. by K. S. Shukla), ‘Use of calculus in
Hindu mathematics’, Ind. Jour. Hist. Sc., 19, 95–104, 1984.
(17) B. B. Datta and A. N. Singh (Rev. by K. S. Shukla), ‘Use of series in
India’, Ind. Jour. Hist. Sc., 28, 103–129, 1984.
(18) D. Pingree, ‘Power series in medieval Indian trigonometry’, in P. Gaeffke
and D. A. Utz Eds., Science and Technology in India, Vol. II, University
of Pennsylvania, Philadelphia 1985, pp. 25–30.
(19) C. T. Rajagopal and M. S. Rangachari, ‘On medieval Kerala mathemat-
ics’, Arch. for Hist. Sc., 35(2), 91–99, 1986.
(20) T. Hayashi, T. Kusuba and M. Yano, ‘Indian values of π derived from
Āryabhat.a’s value’, Historia Scientiarum, 37, 1–16, 1989.
(21) T. Hayashi, T. Kusuba and M. Yano, ‘The correction of the Mādhava
series for the circumference of a circle’, Centauras, 33, 149–174, 1990.
(22) Ranjan Roy, ‘The discovery of the series formula for π by Leibniz’, Gre-
gory and Nı̄lakan.t.ha, Math. Mag. 63, 291–306, 1990.
(23) D. Gold and D. Pingree, ‘A hitherto unknown Sanskrit work concerning
Mādhava’s derivation of the power series for the sine and the cosine’, Hist.
Scientiarum, 42, 49–65, 1991.
(24) K. Balagangadharan, ‘Mathematical analysis in medieval Kerala’, in Sci-
ence in the West and India: Some Historical Aspects, Ed. by B. V. Sub-
brayappa and N. Mukunda, Himalaya Publishing House, New Delhi 1995,
pp. 290–312.
Development of Calculus in India 285

(25) V. J. Katz, ‘Ideas of calculus in Islam and India’, Math. Mag. 68, 163–
174, 1995.
(26) T. Hayashi, ‘Calculations of the surface of a sphere in India’, Science and
Engineering Review of Dohshisha University, 37.4, 194–238, 1997.
(27) C. K. Raju, ‘Computers, mathematics education, and the alternative epis-
temology of the calculus in the Yuktibhās.ā’, Phil. East and West, 51,
325–362, 2001.
(28) D. F. Almeida, J. K. John and A. Zadorozhnyy, ‘Keralese mathematics:
Its possible transmission to Europe and the consequential educational im-
plications’, J. Nat. Geo. 20, 77–104, 2001.
(29) J. Bronkhorst, ‘Pān.ini and Euclid’, Jour. Ind. Phil. 29, 43–80, 2001.
(30) D. Bressoud, ‘Was calculus invented in India?’, College Math. J. 33, 2–
13, 2002.
(31) Kim Plofker, ‘Relations between approximations to the sine in Kerala
mathematics’, in G G Emch et al., Eds. Contibutions to the History of
Indian mathematics, Hindustan Book Agency, New Delhi 2005, pp. 135–
152.
(32) P. P. Divakaran, ‘The first textbook of calculus: Yuktibhās.ā’, Jour. Ind. Phil.
35, 417–443, 2007.
(33) D. Mumford, ‘Why I am a Platonist’, EMS Newsletter, Issue 70, Dec. 2008,
pp. 27–30.
(34) D. F. Almeida and G. G. Joseph, ‘Kerala mathematics and its possible
transmission to Europe’, in P. Ernest, B. Greer and B. Sriraman, Critical
issues in Mathematics Education, University of Montana, 2009, pp. 171–
188.
(35) D. Mumford, ‘Review of “Mathematics in India” by Kim Plofker’, No-
tices of the AMS, 57, 385–390, 2010.

Some Overviews of History of Calculus

(1) C. B. Boyer, The History of the Calculus and its Conceptual Development,
Dover, New York 1949.
(2) M. E. Baron, The Origins of the Infinitesimal Calculus, Pergamon, Oxford
1969.
(3) C. H. Edwards, The Historical Development of the Calculus, Springer,
New York 1979.
286 K. Ramasubramanian and M. D. Srinivas

(4) V. J. Katz, A History of Mathematics: An Introduction, 2nd Ed., Addison

Wesley, Reading Mass. 1998.

K Ramasubramanian
Cell for Indian Science and Technology in Sanskrit
Department of Humanities and Social Sciences
IIT Bombay, Mumbai 400076
E-mail: [email protected] and

M. D. Srinivas
Centre for Policy Studies
6, Baliah Avenue, Luz, Mylapore, Chennai 600 004
E-mail: [email protected]