Indian Journal of History of Science, 50.2 (2015) 227-244 DOI: 10.

16943/ijhs/2015/v50i2/48238

Nārāyaa’s Generalisation of Mātrā-vtta-prastāra

and the Generalised Virahāka-Fibonacci
Representation of Numbers
Raja Sridharan*, R Sridharan** and M D Srinivas***
(Received 15 September 2014)

Abstract
In his Vttajātisamuccaya (c. 600 AD), Virahāka discussed the prastāra or enumeration of
mātrā-vttas or moric metres, metres of value n consisting of long and short syllables, laghu and guru,
assigning value 1 to laghu and 2 to guru, the value of the metre being equal to the sum of the values of its
constituent syllables. Virahāka noted that the sakhyākas or number of rows in the prastāra is given
by a sequence of numbers, which were rediscovered much later in the 13th century by Fibonacci in a
different context. Similar problems of enumeration were considered in the context of tāla-prastāra in
music by Śārgadeva in his Sagītaratnākara (c. 1250 AD).
A general mathematical treatment of most of the combinatorial problems considered in the earlier
literature was given by Nārāyaa Paita in his work Gaitakaumudī (c.1356 AD). One of the problems
discussed by Nārāyaa is the prastāra of a general class of mātrā-vttas where, apart from the syllabic
units laghu and guru (of values 1, 2), there could be other syllabic units (such as pluta, etc.) of values
3,4,... , q. This is also a general form of tāla-prastāra, but it does not subsume the specific tāla-prastāra
considered by Śārgadeva in Sagītaratnākara, where the tāla-units have values 1, 2, 4 and 6.
The questions of naa and uddia, which consist of finding the row number associated with a
generalised mātrā-vtta and its converse, in relation to this prastāra, have been dealt with by Nārāyaa
Paita by means of a tabular form called unmeru. In this paper, we shall show that these naa and
uddia processes are indeed based upon a certain representation of each number uniquely as a sum of the
generalised Virahāka-Fibonacci numbers, i.e., sakhyākas associated with the generalised mātrā-vtta-
prastāra of Nārāyaa.
Key words: Generalised mātrā-vttas, Prastāra or enumeration, Generalised Virahāka-
Fibonacci numbers, Naa, Uddia, Unmeru, Virahāka-Fibonacci representation of numbers.

1. INTRODUCTION pratyayas introduced by Pigala provided a model

for many enumeration problems discussed by later
Indian combinatorics goes as far back as
Indian mathematicians (Sridharan, 2005).
Pigala (c. 300 BC), who introduced six
combinatorial tools (pratyayas) to study Sanskrit Calling the shorter syllable laghu (L) and
prosody. Among these is prastāra, an enumeration the longer syllable guru (G), Pigala enumerated
rule for generating all possible metric patterns, the 2n metres of length n consisting of Ls and Gs.
naa and uddia the processes of finding any In view of the fact that it takes twice as much time
metrical pattern given its row number and its to utter a long syllable (G) as it takes to utter the
converse, in the prastāra. The technique of short syllable (L), Indian prosodists also

*School of Mathematics, Tata Institute of Fundamental Research, Mumbai

**Chennai Mathematical Institute, Chennai
***Center for Policy Studies, 6 Balaiah Avenue, Luz Mylapore, Chennai 600004; Email: [email protected]
228 INDIAN JOURNAL OF HISTORY OF SCIENCE

considered the enumeration of mātrā-vttas, considered in earlier literature. After listing the
metres of value n consisting of long and short various pratyayas, Nārāyaa defines a number of
syllables, assigning value 1 to L and 2 to G, the paktis (sequences) and merus (tabular figures)
value of the metre being equal to the sum of the that are employed in various combinatorial
values of its constituent syllables. In his problems. Nārāyaa then considers different kinds
Vttajātisamuccaya (c. 600 AD), Virahāka of prastāras which generalise those considered
showed that the sakhyākas, or number of rows in prosody and music. He formulates the problem
in the prastāra, is given by a sequence of numbers, as one of enumerating the various possibilities
which were rediscovered much later in the 13th which arise when there are p slots or places
century by Fibonacci in a different context (sthānas) in which the q digits 1, 2,. .. , q, are
(Sridharan, 2006). placed, subject to various conditions.
In his treatise on music Sagītaratnākara, One of the problems discussed by
Śārgadeva (c. 1250 AD) discussed the Nārāyaa is the prastāra of a general class of
enumeration of both tānas (musical phrases) and mātrā-vttas where, apart from the syllabic units
tālas (musical rhythms). The tāna-prastāra laghu and guru (of values 1, 2), there could be
considered by Śārgadeva is essentially a other syllabic units (such as pluta, etc.) of values
procedure for enumerating the n! permutations of 3, 4,. .. , q. This is also a general form of tāla-
the symbols 1, 2, .., n. The tāla-prastāra is a prastāra, but it does not subsume the specific tāla-
generalisation of the mātrā-vtta-prastāra, where prastāra considered by Śārgadeva in
Śārgadeva considers all possible rhythmic forms Sagītaratnākara, where the tāla-units have
composed of the units druta, laghu, guru and pluta values 1, 2, 4 and 6. Before discussing the
with values 1, 2, 4 and 6 respectively (Raja prastāras considered by Nārāyaa, we first give
Sridharan et al., 2010). a brief description of mātrā-vttas and their
Prior to the seminal work Gaitakaumudī prastāra.
of Nārāyaa Paita (c. 1356 AD), most of the
work in combinatorics had its basis in practical 2. MA– TRA– -VR. TTAS –
AND THEIR PRASTARA
problems. Nārāyaa Paita was the first to treat In the case of vara-vttas or syllabic
these problems from a general mathematical metres, it is the number of syllables which are
viewpoint, generalising and unifying many of the fixed and the prastāra or the enumeration consists
earlier results. Specifically, Nārāyaa considered
of all possible metrical forms which are
prastāras of syllabic metres with more than two
characterised by a sequence of laghus and gurus
types of syllables generalising the work of Pigala,
having the same number of syllables. In the case
prastāras of permutations with repetitions
of mātrā-vttas or moric metres, it is the total value
generalising the work of Śārgadeva on svara-
of the metrical form that is fixed, when we assign
prastāra, and also prastāras of combinations
values of 1 mātrā for each laghu (L) and 2 mātrās
which were briefly considered earlier by
for each guru (G).
Varāhamihira (c. 550 AD) and Bhaotpala (c. 950
AD) (Raja Sridharan et al., 2012). In this paper The rule for the construction of the
we shall study Nārāyaa Paita’s generalisation prastāra for mātrā-vttas is stated explicitly in the
of the mātrā-vtta-prastāra. Vttajātisamuccaya of Virahāka (c. 600 AD) and
In Chapter XIII of Gaitakaumudī on is as follows:
Akapāśa (net of numbers), Nārāyaa Paita • The first row consists of all Gs if the total value
gives an abstract mathematical formulation is even and an L followed by all Gs if the total
encompassing most of the combinatorial problems value is odd.
 – – – – –
NARAYAN. A’S GENERALISATION OF MATRA-VR. TTA-PRASTARA 229

• Given any row in the prastāra, to generate the processes for the mātrā-vttas are indeed based
next row, scan from the left to identify the first on a certain representation of each number
G and place an L below that. The elements to uniquely as a sum of non-consecutive Virahāka-
the right are brought down as they are. Fibonacci numbers (Sridharan 2006). The
lagakriyā process for these vttas also leads to an
• The remaining mātrās to the left are filled by
interesting relation between binomial coefficients
Gs, and an L is added at the beginning if need
and the Virahāka numbers.
be to keep the total number of mātrās the same.
• Go on till the last row is reached which is 3. NA– RA– YAN. A’S GENERALISATION OF MATRA-
– –

formed by only Ls. VR. TTA-PRASTARA

–

For instance, the prastāra of vttas of 5- Nārāyaa presents a generalisation of the

mātrās is shown in Table 1: mātrā-vtta, which has, apart from laghu and guru
Table 1: Prastāra of vttas of 5-mātrās with values 1, 2, other elements also (like pluta
etc.) which have values 3, 4,. .. , q. Instead of using
laghus and gurus, Nārāyaa considers sequences
of variable length p (the number of places
(sthānas)) formed from the digits (akas) 1, 2,.. .
q where q is the highest or the last digit
(antimāka), such that their total value or sum
(yoga) n is fixed. In other words, each row of the
prastāra is an ordered partition of n in terms of 1,
2,.. ., q.
Virahāka also gave the rule for the Nārāyaa refers to this case as
sakhyā or the number of rows Sn in the prastāra
of vttas of value n, which is the following:
To generate all the variations (all possible
Sn = Sn–1 + Sn–2 sequences) composed of digits 1, 2,.. . q,
Taking S0= 1, and since S1 = 1, and S2 = 2, where both the total value or sum (n) of
the sequence, as also the highest digit (q)
the above relation generates the Virahāka
that is allowed, are fixed, while the
sequence (later rediscovered by Fibonacci): 1, 1, number of places (p) is variable.
2, 3, 5, 8, 13, 21, ...
If q = 2, we get the usual mātrā-vtta-
The other pratyayas such as naa, uddia prastāra formed with laghus and gurus which are
and lagakriyā (the last one being a method for now denoted by 1, 2 respectively. For instance,
finding the number of metres in the prastāra with the 5-mātrā-prastāra displayed earlier will appear
a given number of laghus or gurus) for mātrā- as shown in Table 2, in Nārāyaa’s notation.
vttas are worked out in later texts such as
Prāktapaigala (of unknown authorship In verses 79-80 of Chapter XIII of
composed around 12th century), Vāībhūaa of Gaitakaumudī on Akapāśa (net of numbers),
Dāmodara (c. 1550) and the commentary of Nārāyaa Paita states the rule for the
Nārāyaabhaa (c.1555) on Vttaratnākara. In enumeration of generalised mātrā-vttas as
fact, it can be shown that the naa and uddia follows:1
1
The verses of Gaitakaumudī have been cited from Kusuba T., Combinatorics and Magic-squares in India: A Study of Nārāyaa
Paita’s Gaita-kaumudī, Chapters 13-14, Ph.D. Dissertation, Brown University (Unpublished), 1993.
230 INDIAN JOURNAL OF HISTORY OF SCIENCE

Table 2: Prastāra with n = 5 and q = 2 This clearly, is a generalisation of the rule

of prastāra for the usual mātrā-vttas given by
Virahāka and others. Here, each prastāra is
characterised by both the total value n and the last
digit q. To illustrate this rule, Nārāyaa gives two
examples which are displayed in Tables 3, 4. The
first one is the prastāra of total value n = 7 and
highest digit q = 3. This is nothing but 7-mātrā-
prastāra with laghu, guru and pluta being the
syllabic elements. The second one is the prastāra
of total value n = 7 and highest digit q = 7.

Table 3: Prastāra with total value n = 7 and the highest

digit q = 3

Write the final number at the beginning

and fill out the sum of the numbers on
the left. After one has put the smaller
number below the first larger number, the
rest [of the numbers are brought down]
as above; the process [is repeated] till a
row all of whose entries are 1s is obtained.
This is the enumeration as declared by the
ancients well-versed in Bharata.
Nārāyaa’s rule of prastāra is essentially
Table 4: Prastāra with total value n = 7 and the highest
the following: digit q = 7
• In the first row of the prastāra, write from the
right to left as many last digits (q) as possible,
such that their sum does not exceed n. Follow
this procedure the same way with the
penultimate digit (q – 1), etc., till the sum of
all the digits is n.
• To go from any row of the prastāra to the next,
scanning from the left, identify the first digit
(say m) which is larger than 1. Below that,
place the digit less than that by 1 (i.e., m – 1).
To the right of it, bring down the digits as they
are from the row above.
• To the left, start from the highest digit q and
– – – . . –
so on (as in the first row) till the total sum of 4 . THE SAMASIKI-PANKTI AND THE SANKHYA OR
–
all the digits reaches the value n. THE NUMBER OF ROWS IN THE PRASTARA

• Repeat the process till the last row of the In order to discuss the various pratyayas
prastāra is reached, all of whose entries are 1. of the generalised mātrā-vtta-prastāra, Nārāyaa
 – – – – –
NARAYAN. A’S GENERALISATION OF MATRA-VR. TTA-PRASTARA 231

makes use of various sequences (paktis) and Below the sāmāsikī-pakti, write 0
tabular forms (merus) which he introduces right followed by 1. Then, the number above
the last number is added to the sum in
at the beginning of the chapter on combinatorics reverse order of those numbers which
(akapāśa). Of these, the most important is the occupy the places equal in number to the
sāmāsikī-pakti (additive sequence), which is a last digit. In this way, all the places in front
generalisation of the Virahāka-Fibonacci are to be filled. In the absence of places
sequence, and gives the sakhyā or the total equal in number to the last digit, the sum
of those present is written.
number of rows in the prastāra. In verses 13-14
of the Chapter on Akapāśa, Nārāyaa defines the The pātāla-pakti when the last digit is q
sāmāsikī-pakti as follows: is thus defined by the relations

Having first put down two ones, then

write their sum in front. And in their front In verses 78-79, Nārāyaa notes the
write the sum in reverse order of those significance of the sāmāsikī-pakti and the pātāla-
numbers which occupy the places equal pakti as follows:
in number to the last digit. In the absent
places equal in number to the last digit,
write in front the sum of them [that are
present]. In this way is formed the additive
sequence whose number of places is equal
to the sum plus one.
The end of the sāmāsikā sequence gives
The sāmāsikī-pakti when the last digit is the number of possible variations
q is thus defined by the relations (metrical forms of generalised mātrā-
vttas, or the number of rows in the
prastāra). The numbers of the sequence
taken in the reverse order, from the
penultimate element are the number of
variations ending in 1, etc. The numbers
of the pātāla sequence, which correspond
Clearly, when q = 2, Sn2 is nothing but the to the same last digit and the sum, when
taken in reverse order give the (number
Virahāka-Fibonacci sequence defined earlier. As
q of occurrences) of 1, etc. The sum of
we shall see below, Sn is the sakhyā or the total all these is the occurrence of all the
number of rows in the prastāra of total value or digits.
sum n with the highest digit being q.
Thus, the sakhyā or the number of rows
In verses 15-16, Nārāyaa also defines the in the prastāra is given by the last element of the
associated pātāla-pakti, the elements of which sāmāsikā sequence and the penultimate and other
give the number of times any digit occurs in the elements in the reverse order give the number of
prastāra, as follows: rows in the prastāra which end in digits 1, 2, etc.
Similarly the numbers in the pātāla sequence,
taken in the reverse order give the number of times
the different digits 1, 2, etc. occur in the prastāra.
The sum of all the elements of the pātāla sequence,
232 INDIAN JOURNAL OF HISTORY OF SCIENCE

gives the total number of all the digits occurring examples of prastāra displayed in Tables 3, 4. That
in the prastāra. We display below two examples is why, Nārāyaa refers to this case as aniyata-
of the sāmāsikī-pakti and pātāla-pakti given by sthāna, where the number of places (the number
Nārāyaa. of elements in each row of the prastāra) is not
fixed.
Table 5, displays the sāmāsikī-pakti and
pātāla-pakti for the prastāra with total value 7 Thus, for a generalised mātrā-vtta-
and the last digit 3. This is the generalised mātrā- prastāra, we need to answer the question as to
vtta, where we include apart from the syllabic how many rows in the prastāra (where the total
units laghu and guru (of values 1, 2), a third sum is n and the last digit is q), have length p.
syllabic unit pluta, which has value 3. Recalling Nārāyaa answers this question by means of the
the prastāra that we displayed earlier in Table 3, elements of the so called needle sequence (sūcī-
we can easily verify that 44 is the sakhyā or the pakti) or the arrow-head sequence (nārācikā-
number of rows in the prastāra; and that the last pakti), which are also put together in a fish-
24 rows end in 1, the previous 13 rows end in 2 tabular figure (matsya-meru). The elements Up,q(r)
and the first 7 rows end in 3. We can also verify of the needle sequence, as defined by Nārāyaa,
that the digit 1 appears 118 times in the prastāra, are essentially multinomial coefficients, the
2 appears 56 times, and 3 appears 26 times. The coefficients of various powers xr in the expansion
total number of digits appearing in the prastāra is of (1 + x + x2 + . . . + xq–1)p. Nārāyaa also gives an
26+56+118=200. interesting expression for the elements of the
sāmāsikī-pakti or the generalised Virahāka-
q
Table 5: The Sāmāsikī-pakti and pātāla-pakti for n = 7 Fibonacci sequence of sakhyākas Sn, as sums
and q = 3 of suitable elements of the sūcī-pankti Up,q(r).
We have summarised Nārāyaa’s discussion of
sūcī-pakti and its application to the study of the
generalised mātrā-vtta-prastāra in the Appendix.
Table 6: The Sāmāsikī-pakti and pātāla-pakti for n = 7
and q = 7 5. THE UNMERU AND THE NAS. T. A
AND UDDIS. T. A PROCESSES

In verses 84-86 of the Chapter on

Akapāśa, Nārāyaa introduces the following un-
Unlike in the case of vara-vtta-prastāra meru, which helps in carrying out the naa and
where the number of syllables is fixed to start with, uddia processes in a generalised mātrā-vtta-
in a mātrā-vtta-prastāra where only the total prastāra.
mātrā of the metric form is held fixed, the number
of syllables in different rows of the prastāra can
be different. This is clearly seen for example in
the example of 5-mātrā-prastāra that was
displayed in Table 1. In the same way, in the
generalised mātrā-vtta-prastāra as defined by
Nārāyaa, the lengths of the sequences of digits
The number of cells in each row, starting
that constitute different rows of a prastāra with one and increasing by one at each
(characterised by a given sum n and last digit q) step, are equal to the total sum plus one.
can be different as can be readily seen from the In the bottom row, the elements of the
 – – – – –
NARAYAN. A’S GENERALISATION OF MATRA-VR. TTA-PRASTARA 233

additive sequence (sāmāsikī-pakti) are

written in order. In the others, the
arithmetical sequence (caya-pakti 1, 2,
. .. , q, . . .) is written in the reverse (from
right to left). In the upper horizontal rows
they appear naturally. In those cells where
numbers larger than the last digit is to be Subtract the naāka (row number for
written, an omission is made. This figure which the variation or the corresponding
is called the lofty tabular figure (unmeru). sequence of digits is to be determined)
from the last number of the sāmāsikī-
The bottom row of unmeru has a number pakti (the sequence of sakhyākas or
of entries which is one more than total sum (n) the generalised Virahāka sequence).
and subsequent rows have one number less at each Then, subtract from the result the next
step. The bottom row is filled with the sāmāsikī- number and so on till no subtraction is
possible. The number in the cell which is
pakti or the generalised Virahāka sequence of the intersection of the row and column
sakhyākas. In the subsequent rows the numbers associated (with the subtracted number)
1, 2, .. . , q, are written from the right end. In Table is the first digit of the variation. The rest
7 and Table 8, we display two examples of unmeru of the elements of the row are similarly
given by Nārāyaa. obtained, by repeating the process with
unmeru.
The method is essentially as follows: First,
Table 7: Unmeru for n = 7 and q = 3 subtract the given row number from the last entry
of the bottom row of unmeru which consists of
the sequence of sakhyākas or the generalised
Virahāka sequence. Then, from what remains,
see if the penultimate number can be subtracted
and so on. In this way, amongst the numbers of
the bottom row, identify those which are patita
(numbers which have been subtracted) and apatita
(numbers which have not been subtracted). Then,
starting from right, the digit just above the first
apatita number is the last element of the desired
Table 8: Unmeru for n = 7 and q = 7
sequence. Now, go one cell above the top entry of
that column and move left till the intersection with
the column above the next apatita number is
reached. That digit is the penultimate element of
the desired sequence of digits. And so on.
Nārāyaa’s Example 1: To Find the 36th
row of the 7-mātrā-prastāra with highest digit 3
44 - 36 = 8, 8 - 7 = 1, and 1 - 1 = 0
Thus the second 1 and 7 are patita and the
rest are apatita-Sakhyākas. Note the entry 1 in
In verses 87-89, Nārāyaa presents a rather the first row above the last apatita 24. In the row
cryptic description how the naa process is carried above the topmost entry of the column of that
out with the help of unmeru: apatita, move left till you reach the column of the
234 INDIAN JOURNAL OF HISTORY OF SCIENCE

next apatita 13. Note the corresponding entry 1. the number (sakhyāka) of the
And so on. Thus the sequence of digits in 36th enumeration, one gets the row-number of
a given variation.
row of the prastāra is 21211.
The uddia process is indeed the reverse
Table 9: Example of naa process when n = 7 and q = 3 of the naa process as will be clear from the
following example:
Nārāyaa’s Example: To find the row
where 2212 appears in the 7-mātrā-prastāra with
highest digit 3:
From the right, we first identify the apatita
13 above which 2 appears. Then we move left in
the row above the topmost entry of the column of
that apatita till we get 1 and note the apatita
number 7. And so on.
Nārāyaa’s Example 2: To Find the 36th Row number = 44 - sum of the patitas =
row of the 7-mātrā-prastāra with highest digit 7 44 - (24+4+1) = 15
64 - 36 = 28, 28 - 16 = 12, 12 - 8 = 4 and 4 Table 11: Example of uddia process when n = 7 and q = 3
-4=0
Thus 16, 8 and 4 are patita and the rest are
apatita-sakhyākas. We find that the desired
sequence in the 36th row of the prastāra is 1141

Table 10: Example of naa process when n = 7 and q = 7

6 ALTERNATE METHOD FOR THE NAS. T. A

AND UDDIS. T. A PROCESSES

We here describe an alternative method for

the naa and uddia processes, by making use
of a correspondence between the various digits 1,
Nārāyaa’s description of the uddia 2,. .. , q, and associated sequences of patita and
process, presented in verse 90, is even more cryptic apatita sakhyākas. This method is similar to the
than his description of naa process: one discussed in works such as the commentary
of Nārāyaabhaa on Vttaratnākara for the usual
mātrā-vttas (with only laghus and gurus). Such
Those numbers in the (bottom row of)
a method was also used by, Śārgadeva for the
unmeru which were subtracted in the uddia process in the case of tāla-prastāra in
naa process, if these are subtracted from Sagītaratnākara (Raja Sridharan et al., 2010).
 – – – – –
NARAYAN. A’S GENERALISATION OF MATRA-VR. TTA-PRASTARA 235

Table 12: Signature of different digits as sequences of a and p

Write down the sequence of sakhyākas number is written above 1, two numbers above 2,
or the generalised Virahāka sequence for the and so on and q numbers are written above q. Omit
prastāra. For the given row-number, identify the number above 1, and also the first number
patita and apatita-sakhyākas and place the mark above each of the digits 2, 3, .. . , q, and sum the
a or p below each sakhyāka. From the right, rest. Subtract the sum from the last entry of the
scan the sequence of these marks and identify the sequence of sakhyākas (which is nothing but
corresponding element of the sequence of digits, the number of rows in the given prastāra), and
successively, via their signatures as given in Table the result is the row-number associated with the
12. given sequence of digits. We illustrate the above
process in Tables 15-18.
We illustrate the method with the
following examples which are displayed in Tables Table 15: To Find the row where 2212 appears in the 7-
13, 14. mātrā-prastāra with highest digit 3
Row-number = 44-(1+ 4+24) = 15
Table 13: To Find the 36th row of the 7-mātrā-prastāra with
highest digit 3

Table 16: To Find the row where 322 appears in the 7-mātrā-
prastāra with highest digit 3
Row-number = 44-(1+ 2+7+24) = 10

Table 14: To Find the 36th row of the 7-mātrā-prastāra

with highest digit 7

Table 17: To Find the row where 21211 appears in the 7-

mātrā-prastāra with highest digit 3
Row-number = 44-(1+7) = 36

We can also similarly devise an alternative

method for uddia process by using the above
correspondence between the various digits and Table 18: To Find the row where 1141 appears in the 7-
associated sequences of patita and apatita mātrā-prastāra with highest digit 7
sakhyākas. The procedure is as follows. Row-number = 64-(4+8+16) = 36

Write the sequence of sakhyākas (the

generalised Virahāka sequence) such that one
236 INDIAN JOURNAL OF HISTORY OF SCIENCE

One important feature that we notice in the The prastāra of rows of total value 0
above examples of naa and uddia processes, consists of one row with no symbols. The prastāra
is that they are based on certain representations of rows of total value 1 consists of a single row
of numbers as sums of the generalised Virahāka with the entry 1. The prastāra of rows of value 2
n
numbers {Sq}. is given in Table 19. This is obtained by appending
the symbol 2 to the prastāra of rows of value 0
For the case when q = 3, we have the
and 1 to the prastāra of rows of total value 1.
representations
8 = 1+7 Table 19: Prastāra of value 2

29 = 1 + 4 + 24
34 = 1 + 2 + 7 + 24
For the case when q = 7, we have the
representation The prastāra of rows of total value 3 is
given in Table 20. This is obtained by appending
28 = 4 + 8 + 16 the symbol 3 to the prastāra of rows of value 0
These are particular instances of a general and then appending the symbol 2 to the prastāra
result that every number can be decomposed of rows of value 1 and finally appending 1 to the
uniquely as a sum of the generalised Virahāka- prastāra of rows of total value 2.
Fibonacci numbers. We shall discuss this, and
Table 20: Prastāra of value 3
other mathematical aspects of the generalised
mātrā-vtta-prastāra, in the next section.
.
7. THE GENERALISED VIRAHA– NKA-FIBONACCI
REPRESENTATION OF NUMBERS
While discussing the mathematics More generally, we can construct the
underlying the generalised mātrā-vtta-prastāra prastāra of rows of total value n > 3 inductively
of Nārāyaa, we shall, for simplicity consider the as follows: Write down the prastāra of rows of
case q = 3, i.e., there are only three digits, 1, 2, 3, value n – 3 and append 3, then write down the
in each row of the prastāra. prastāra of rows of value n – 2 and append 2, and
finally write down the prastāra of rows of value
7.1 An inductive construction of the prastãra and n – 1 with 1 appended at the end. It follows by
the sakhyā rule induction on n that this is indeed the prastāra
We consider the symbols 1, 2, 3, having obtained by following Nārāyaa’s rule for
the same value as the numbers that these digits generalised mātrā-vttas total value n, and highest
represent. We write down all rows consisting of digit q = 3.
these symbols, such that the total value of each It follows that if Sn3 is the number of rows
row is n, the value of a row being defined to be in the prastāra of metres of value n, then S03 = 1,
the sum of the values of the constituent symbols S13 = 1, S23 = 2 and
in the row. A prastāra of metres of value n is a
.
table enumerating all such ordered partitions
(allowing repetitions) of n in terms of the numbers We shall now see how this also follows
1, 2, 3. from an argument that uses generating functions.
 – – – – –
NARAYAN. A’S GENERALISATION OF MATRA-VR. TTA-PRASTARA 237

Let us assume that a, b, c are non-commuting

variables and consider the formal expression
1 + (a + b + c) + (a + b + c)2 + ... 7.2 Representation of Numbers as Sums of
This formal sum will consist of monomials Generalised Virahāka- Fibonacci Numbers
a1 a2 . . . an, where each ai is a, b or c. Further, We first recall the following result for
because of non-commutativity we cannot permute Virahāka-Fibonacci numbers {Sn2}, which was
the various constituent symbols of the monomials. proved in an earlier investigation (Sridharan,
Now suppose that we assign values 1, 2, 3 to the 2006) on the mathematical theory of the usual
symbols a, b, c, and define the total value of a mātra-vtta-prastāra:
monomial to be the sum of the values of its Proposition 1: Any natural number n is either
constituent symbols. If we now specialize a = x, b equal to a Virahāka-Fibonacci number Sk2 or can
= x2, c = x3 in the formal sum, we obtain the be uniquely expressed as a sum of non-consecutive
following power series Virahāka-Fibonacci numbers {Si2}.
1 + (x + x2 + x3) + (x + x2 + x3)2 + ... Now let us consider the case of the
= [1 - (x + x2 + x3)]-1 generalised Virahānka-Fibonacci numbers Sn3. We
Now, the co-efficient of the power xn in have S 13 = 1, S23 = 2, S 33 = 4 and S n3 = Sn-1
3
+
the above power series is really the sakhyāka Sn-2 + Sn-3, n > 3, and we have the following:
3 3

Sn3, as it is equal to the number of monomials of Proposition 2: Any natural number n is either a
total value n in a, b, c, when we assign the values generalised Virahāka-Fibonacci number Sk3 or can
1, 2, 3 to the symbols a, b, c. Hence we have the be uniquely expressed uniquely as a sum of
equality generalised Virahāka-Fibonacci numbers Si3, no
1 + (x + x2 + x3) + (x + x2 + x3)2 + ... three of which are consecutive.
To prove this we first need the following Lemma:
Lemma 1: For all i > 0, we have
Therefore we obtain
2Si3 ≥ Si+1
3
,
[1 – (x + x2 + x3)]-1 [S30 + S31 x + S23 x2 + ...] = 1
with equality holding only when i = 1, 2.
By equating the coefficients of xn, we obtain the
relation Proof: We have to show that 2Si3 > Si3 + Si-1 3
+
3 3 3
S i-2. It is thus enough to show that Si > Si-1 +
S3i-2. But this follows from the fact that Si3 = S3i-1 +
S3i-2 +S3i-3 > S3i-1 + S3i-2.
Remark: In the case of the usual mātrā-vttas,
with laghus and gurus only, with values 1, 2 Remark: We note that the sakhyākas of the
respectively, the sakhyākas are the usual vara-vttas are Si = 2i which satisfy 2Si = Si+1.
Virahāka-Fibonacci numbers {S n2}, which The sakhyākas of ordinary mātrā-vttas, the
satisfy Virahāka-Fibonacci numbers Si2, also satisfy 2Si2
> S2i+1.
, and , for
Now, let n be a natural number. We define
2
These relations for S can be obtained
n the canonical decomposition of n as follows: We
following a similar argument as above as a choose the largest i1 such that S3i ≤ n. Then, we
1

consequence of the identity choose the largest i 2 such that S 3i ≤ n – S 3i.

2 1
238 INDIAN JOURNAL OF HISTORY OF SCIENCE

Continuing in this manner, we obtain a If p > t – 2; since p ≤ t – 1, we have p = t – 1; and

decomposition we thus have to prove that

which we refer to as the canonical decomposition

Now since no three consecutive {S3i } occur in the
of n.
decomposition, we have l < t – 2, and by induction
Since 2S3i > S3i+1, every S3i occurs only once
in the canonical decomposition. For, otherwise, if
S3i were to occur more than once in the canonical Hence,
decomposition of n, we would have at a certain
stage of the canonical decompositions, i as the
largest number such that

proving the lemma.

and that i also satisfies S3i≤ m – S3i . In other words,
2S3i ≤ m, which implies that S3i+1 ≤ m, contradicting Remark: The above lemma also holds for the
the fact that i is the largest number such that S3i ≤ sakhyākas of the vara-vttas Si = 2i, and also
m. for the sakhyākas of ordinary mātrā-vttas, the
Virahāka-Fibonacci numbers {S2i}, where in the
A similar argument can also be made to latter case we only need to impose the condition
show that three consecutive generalised that the decomposition does not contain
Virahāka-Fibonacci numbers {S3i} do not appear consecutive Virahāka-Fibonacci numbers.
in any canonical decomposition.
Now we state a proposition, which follows
Lemma 2: Let immediately from the above lemma.

Proposition 3: Let m be a natural number and

where, a < c < ...l < p < t are natural numbers with
no three of a, c, . . . , l, p, and t being consecutive.
Then

Proof: Under the hypothesis of the lemma, we be two decompositions of m such that both the
have to prove that decompositions do not contain three consecutive
sakhyākas {S3k }. Then the two decompositions
are the same. In particular, any decomposition of
Note that since p < t, we have p ≤ t – 1. It is enough m having no three consecutive sakhyākas {S3k }
to prove that is the canonical decomposition.
If we associate the (n + 1)-tuple (a0, a1,
a2,.. . , an), where each ai is either 0 or 1, with the
If p ≤ t - 2, we have, by induction number a0S30 + a1S31 + a2S32 + .. . anS3n, then we can
formulate the result regarding the existence and
uniqueness of the generalised Virahāka-
thus proving the result Fibonacci decomposition as follows:
 – – – – –
NARAYAN. A’S GENERALISATION OF MATRA-VR. TTA-PRASTARA 239

Proposition 4: There exists a bijection between value n, using symbols 1, 2, 3, starting from the
numbers 0, 1, 2,.. . , S3n – 1, and the (n + 1)-tuples last row (consisting of n ones), numbering the last
(a0, a1, a2,.. . , an), where each ai is either 0 or 1, row as 0, the penultimate row as 1, and so on. If
satisfying the following conditions: the rank of the row under our enumeration is t,
(i) a0 = 0, an = 0. then the numbering of the row in the usual
enumeration (starting from the top row and taking
(ii) No three consecutive ai’s are 1. its row-number as 1) is S3n – t.
The bijection is obtained by associating The process of naa involves writing
each number k (such that 0 ≤ k ≤ S3n – 1) with its down the row explicitly (without the aid of the
canonical decomposition. prastāra) given the row-number associated with
Remark: We can also prove that the above it.
assignment is a bijection by showing that the In a prastāra of value n :
cardinality of the set of all (n + 1)-tuples satisfying
i The last symbol of the m-th row is 1 if and
(i) and (ii) is S3n which can be done via a simple
argument using induction. only if m < S3n – 1, and the penultimate symbol
of the m-th row is the last symbol of the m-th
In the above discussion, for the sake of row in the prastāra of value n – 1 (where the
simplicity, we have considered the case of the last row consists of n – 1 ones).
generalised mātrā-vtta-prastāra of 3 digits, 1, 2,
ii The last symbol of the m-th row is 2 if and
3, or, to use Nārāyaa’s terminology, the case
only if S3n – 1 ≤ m < S3n – 1 + S3n – 2, and the
where the maximum digit q = 3. All the results
penultimate symbol of the m-th row is the last
which were demonstrated above for q = 3, can be
symbol of the (m – S3n – 1)-th row in the prastāra
easily generalised to the case where q is any
of value n – 2 (where the last row consists of
positive integer greater than or equal to 2. We shall
n – 2 ones).
here just state (without proof) the basic result on
the representation of any natural number uniquely iii The last symbol of the m-th row is 3 if and
as a sum of the generalised Virahāka-Fibonacci only if S3n–1 + S3n–2 ≤ m < S3n–1 + S3n–2 + S3n–3, and
numbers {Sqn} : the penultimate symbol of the m-th row is the
Proposition 5: Let q ≥ 2 be an integer. Then, every last symbol of the (m – S3n–1 – S3n–2)-th row in
the prastāra of value n – 3 (where the last row
natural number is either a generalised Virahāka-
consists of n – 3 ones).
Fibonacci number Skq for some k > 0, or can be
uniquely expressed as a sum of generalised Proceeding as above by induction, we can
Virahāka-Fibonacci numbers {Sq1, Sq2, S3q,.. .}, obtain each symbol of the m-th row of the prastāra
under the condition that the sum does not contain of value n.
any set of q consecutive generalised Virahāka- Given any integer, 0 < m < S3n–1, we have
Fibonacci numbers.
the canonical decomposition of m :
7.3 The Mathematical Basis of the Naa and m = a0S30 + a1S31 + a2S32 +... anS3n ,
Uddia Processes where, a0 = 0, an = 0, and the rest of the ai could
We again restrict ourselves, for the sake be 0 or 1, as discussed above. To this canonical
of simplicity, to the case of the generalised mātrā- decomposition, we associate the string (a0, a1, a2,..,
vtta-prastāra of the three digits, 1, 2, 3. We a n) of zeroes and ones, with a 0 = 0, a n = 0.
propose to enumerate the rows of the prastāra of Summarising the discussion above, we have:
240 INDIAN JOURNAL OF HISTORY OF SCIENCE

Proposition 6: In a prastāra of value n : We now briefly consider the converse

i If the string associated with m ends with (0, process uddia, where the objective is to
0), the last symbol of the m-th row is 1, and determine the row-number (in the enumeration
the penultimate symbol is given by the given above) given the sequence of symbols that
truncated binary sequence with the last zero occur in that particular row. Suppose that the row-
removed in the prastāra of value n – 1. number that we wish to determine is m and we
now describe a process to determine that.
ii If the string associated with m ends with (0,1,
0), then the last symbol of the m-th row is 2 Suppose that the last symbol in the given
and the penultimate symbol is given by the row is 1, then m < S3n–1 and therefore S3n–1 does not
truncated binary sequence with the end portion occur in the decomposition of m. If we delete the
(1, 0) removed in the prastāra of value n – 2. last symbol 1, we obtain a row in the prastāra of
value n – 1, where the row number t can be
iii If the string associated with m ends with (0, 1,
1, 0), the last symbol of the m-th row is 3 and determined by following the same process by
the penultimate symbol is given by the induction, and we have m = t.
truncated binary sequence with the end portion Suppose that the last symbol of the m-th
(1, 1, 0) removed in the prastāra of value n – row is 2, then S3n–1 occurs in the decomposition of
3. m. If we delete the last symbol 2, we obtain a row
The above proposition gives a way of of value n – 2, whose row-number t in the prastāra
writing down the m-th row of a prastāra of value of value n – 2, can be determined by following
n, given the canonical decomposition of m. Let the same process by induction. We then set m =
the canonical decomposition of m be given by S3n–1 + t.
m = a0S30 + a1S31 + a2S32 +... anS3n , If the last symbol of the m-th row is 3, then
S3
and S 3n–2 both occur in the canonical
with a0 = 0, an = 0, and the rest of the ai = 0 or 1. n–1
decomposition of m. If we delete the last symbol
We call the S3i where the ai = 0 as an apatita-
sakhyāka, and the S3i where the ai = 1 as a patita- 3, we obtain a row of value n–3, whose row
sakhyāka. The sequence of patitas and apatitas number t in the prastāra of value n–3, can be
corresponding to m determine the sequence of determined following the same process by
entries from 1, 2, 3, comprising the m-th row. The induction. We then set m = S3n–1 + S3n–2 + t.
above proposition actually gives the signature of The above process is exactly the same as
1, 2, 3, in terms of the apatitas (a) and the patitas the one discussed in the previous section, where
(p) as shown in Table 21 (this is a particular case one, two and three sakhyākas were written
of Table 12). above each of the symbols 1, 2, 3, and only the
second sakhyāka above 2, and the second and
Table 21: Signature of the digits 1; 2; 3 in terms of a and p
the third sakhyākas above 3, were summed
together. The total so obtained is the row-number
in the alternate enumeration (starting with row-
number 0 from the bottom of the prastāra) that
we have used in this section. The total subtracted
from S 3n gives the row-number in the usual
enumeration which starts with row number 1 from
the top of the prastāra.
 – – – – –
NARAYAN. A’S GENERALISATION OF MATRA-VR. TTA-PRASTARA 241

The above discussion can be generalised = q = 3, which is shown in Table 22. Clearly the
to the case where the largest entry in each row of elements of this sequence are nothing but the
the prastāra is any natural number q. Clearly, the coefficients of different powers of x in the
essential mathematical result that is at the heart expansion of (1 + x + x2)3.
of the naa and uddia processes is the canonical
decomposition of any positive integer uniquely If we look at the above definition of the
as a sum of the generalised Virahāka-Fibonacci Up,q(r), that it is the coefficient of xp+r in (x + x2 +
numbers {Sqi}. . . . + xq)p, we can easily see that this is also the
number of sequences of length p, formed out of
Appendix the digits 1, 2, .. . , q, such that the sum of the
The sūcī-pakti, matsya-meru and the number sequence is p + r. In other words, the number of
of rows of prastāra with p-sthānas rows of length p in the generalised mātrā-vtta-
prastāra characterised by the total sum n and last
In verse 21 of the Chapter on Akapāśa, digit q, is Up,q(n – p), where clearly we need to
Nārāyaa defines the sūcī-pakti (needle have [–qn] ≤ p ≤ n, (where [–qn] denotes the integral
sequence) or the nārācikā-pakti (arrow-head part of –qn), in order that the sum of a sequence of
sequence) as follows: length p composed of the digits 1, 2, .. . , q, may
have the sum n. We shall now see that Nārāyaa
also states essentially the same result while
identifying the number of rows of different lengths
Vaiśleikī-paktis measured by the last in the prastāra with the elements of the so called
digit, which are equal in number to the matsya-meru (fish tabular figure).
number of places are kept separately.
Their product is the sūcī-pakti (needle Table 22: Needle sequence for p = q = 3 by door-junction
sequence) or the nārācikā-pakti (arrow- method
head sequence)
If p is the number of places and q is the
final digit, then the arrow-head sequence is defined
by multiplying the vaiśleikī-pakti, which is of
the form 1, 1, 1, . . . ,1 (1s repeated q times),
multiplied by itself p times, by the door-junction
(kapāa-sandhi) method, which is nothing but the
algebraic method of multiplying keeping in mind
the different place-values (see Table 22).
Hence, the (r+1)-th element of the
sequence is a sum of multinomial co-efficients:
In verses 40-43 of the chapter on
Up,q(r) = Coefficient of xr in (1 + x + x2 + . . . + xq-1)p
Akapāśa, Nārāyaa defines the tabular figure
= Coefficient of xp+r in (x + x2 + . . . + xq)p called the matsya-meru associated with a given
In particular, for q = 2, Up,2(r) = C(p,r), sum n, last digit q, and also the number of places
i.e., in this case the elements of the arrow-head p, as follows:
sequence are nothing but the binomial coefficients.
Nārāyaa gives an example of how the
needle-sequence is obtained for the case when p
242 INDIAN JOURNAL OF HISTORY OF SCIENCE

Table 23: Matsyameru for n = 3, q = 3 and p = 3

Rows of cells are made beginning with

one cell and increasing each time by the
last digit minus one. The number of rows
is measured by the number of places plus
one. Leaving the first cell above,
horizontal rows equal in number to the In this matsya-meru, we will have [the
sum of numbers are to be made below the rows given by] the arrow-head and partial
second. One should place one in the first arrow-head sequences.
cell and, in each row below, write the sum
In fact, excluding the first row, the
in the reverse order of entries in the row
succeeding rows of the matsya-meru are filled with
above equal in number to the last digit. If
there are not as many numbers as the last elements of the arrow-head sequences U1,q(r),
digit, one should write the sum of the U2,q(r), and so on. The process of generating the
available digits. meru is actually based on the recurrence relation
Up,q(r) = Up-1,q(r - 1) + Up-1,q(r - 2) + ... + Up-1,q(r - q),
When we consider the prastāra which is
defined by the sum n and the last digit q, the which can be derived easily from the definition
number of rows in matsya-meru is equal to n + 1. of Up,q(r).
The first row has a single cell which has entry 1. Later in the same Chapter on Akapāśa,
The successive rows have their first cell moved while discussing the generalised mātrā-vtta-
to the right and each of them have (q – 1) cells prastāra, Nārāyaa seems to characterise the
more than the previous row. This later rule is matsya-meru in terms of just the sum n and last
followed till the number of rows is equal to the digit q without any reference to some fixed number
total number of places p + 1. What is really meant of places p. All the defining features of the matsya-
by this condition is not clear, though from the meru as described above are retained, except that
example given below we can see that the number both the number of rows and columns of the meru
of columns is not increased after this stage. The are now fixed at n + 1. This is clear from the
cells of the successive rows are filled as follows. example of matsya-meru, displayed in Table 24,
Each cell is filled by the sum of the entries in q which is merely characterised by the sum n = 3
cells of the row above which are just to the left of and the last digit q = 3. This form of the meru is
the given cell. presented by Nārāyaa while discussing an
example following verse 79 in the Chapter on
After defining the matsya-meru, Nārāyaa
Akapāśa.
displays the form of the meru shown in Table 23,
when the sum n = 3, the last digit q = 3 and the In verse 80 of the Chapter on Akapāśa,
number of places p = 3. Nārāyaa indicates how the matsya-meru can be
used in order to obtain the number of rows in the
While working out this example,
prastāra which have a given length:
Nārāyaa also notes that
 – – – – –
NARAYAN. A’S GENERALISATION OF MATRA-VR. TTA-PRASTARA 243

Table 24: Matsyameru for n = 3; q = 3 The elements of the arrowhead sequence

in each row are measured by the number
of places. The sum of each of these (rows)
separately becomes a geometric sequence
(guottara-pakti). The columns are
measured by the sum. The columns
summed separately become equal to the
elements of the sāmāsikī-pakti.
Here, Nārāyaa first mentions that the sum
of the sūcī-pakti (for a given value of q) form a
geometrical sequence. This is essentially the
The column (vertical row) at the end of relation
the matsya-meru, from that are known the
(number of) variations (or rows of the
prastāra for each number of places
(sthāna)...

In other words, the last column of the Nārāyaa then gives the following
matsya-meru gives the number of rows in the important relation between the coefficients Up,q(r)
prastāra with different lengths, or number of and the samasikī-pakti or the generalised
places. For the case where n = q = 3, the last Virahāka-Fibonacci sequence:
column of the matsya-meru (shown in Table 24)
is given by 6, 16, 15, 6, 1. These indeed are the
number of rows in the prastāra which are of length
3, 4, 5, 6 and 7 respectively. The sum of this last where t is such that (t – 1)q < n ≤ tq.
column of the meru is equal to 44 which is indeed
S 73, the sakhyā or the total number of rows in the The above relation, gives an important
prastāra characterised by n = q = 3. In fact, relation between the multinomial coefficients
the various column sums of the above matsya- Un–r,q(r) and the elements of the generalised
meru are 1, 1, 2, 4, 7, 13, 24 and 44, which are Virahāka-Fibonacci sequence {Snq}. It has also
nothing but the generalised Virahāka sequence an important interpretation in the prastāra of the
{S3i}. generalised mātrā-vttas which is the following.
The last property mentioned above is Since U n–r,q (r) is the total number of
indeed a particular case of a general relation generalised mātrā-vttas (sequences of 1, 2,. .. ,
between the elements of the nārācikā-pakti Upq(r) q), which are of length n – r and whose total value
(which are also entries of matsya-meru) and the or sum is n, the sum of all such numbers for various
sāmāsikī-pakti or the generalised Virahāka possible lengths, should clearly be equal to Snq,
sequence {Snq}, which is noted by Nārāyaa in which is the sakhyāka or the total number of all
verses 43-44 of the Chapter on Akapāśa, while the generalised mātrā-vttas of value n.
discussing the sums of the rows and columns of Remark: In mathematical terms, Nārāyaa’s
matsya-meru. matsya-meru is a part of the triangle of
multinomial coefficients. If we tilt this triangle in
a certain way and take the sum of the rows, we
get the generalised Virahāka-Fibonacci numbers.
In a similar manner, in case we tilt the Pigala-
244 INDIAN JOURNAL OF HISTORY OF SCIENCE

Pascal triangle and take the sum of the rows, we Sagīta-ratnākara of Niśaka Śārgadeva (c.1225): Ed.
get the standard Virahāka-Fibonacci numbers with commentary Kalānidhi of Kallinātha (c.1420), 2
Vols., Anandasram Sanskrit Series, Pune 1896-7. Ed.
(Raja Sridharan et al., 2012a)
with Kalānidhi of Kallinātha (c. 1420) and Sagīta-
sudhākara of Sihabhūpāla (c.1350), by Subramanya
BIBLIOGRAPHY Sastri S., Krishnamacharya V. and Sarada S., 4 Vols.,
Adyar Library, 1943-1986. Chapters 1-4, Ed. with
1. Source-works English translation and Notes by Premalata Sharma
Bhat-sahitā of Varāhamihira (c. 6th century): Ed, with and Shringy R. K., 2 Vols., Delhi, 1978, 1989.
the Vivti of Bhaotpala (c.967), by Sudhakara Vaī-bhūaa of Dāmodara (c.1550): Published in
Dvivedi, 2 Vols., Varanasi 1897. Ed, with the Vivti of Kavyamala Series, No 53, Bombay 1895.
Bhaotpala, by Dvivedi K. C., 2 Vols, Sampurnananda
Sanskrit University, Varanasi 1968. Vtta-jāti-samuccaya or Kaisiha-chanda (in Prākta) of
Virahāka (c.650): Ed. with commentary of Gopāla
Chanda-Sūtra of Pigalanāga (c.300 BC): Ed. and Tr. into (c.1000) by Velankar H. D., JBBRAS 1929, 1932. Ed.,
German in Weber A., Über die Metrik der Inder in A. with commentary by Gopāla, by Velankar H. D.,
Weber, Indische Studien, VIII Band, Harrwitz and Rajasthan Prachya Vidya Pratisthana, Jodhpur
Gofsmann, Berlin 1863; Rep. Hilberheim New York, 1962.
1973. Ed. with commentary Mtasa–jīvanī of
Halāyudha (c. 950), by Viswanatha Sastri, Bibliotheca
Indica, Calcutta 1871-1874. Ed. with commentary 2. Secondary works
Mtasa–jīvanī of Halāyudha (c. 950), by Kedaranatha Raja Sridharan, Sridharan R. and Srinivas M. D.,
Vasudeva Lakshmana Pansikara, Kavyamala Series Combinatorial Methods in Indian Music: Pratyayas
No. 91, Bombay 1908, Rep. Chowkhamba 2001. in Sagīta-ratnākara of Śārgadeva, in Seshadri C.S.
Gaita-kaumudī of Nārāyaa Paita (c.1356): Ed. by Ed., Studies in the History of Indian Mathematics,
Padmakara Dvivedi, 2 Vols., Varanasi 1936, 1942. Hindustan Publishing Agency, Delhi, 2010, pp. 55-
Chapters 13-14 edited and translated in Kusuba T., 112.
Combinatorics and Magic-squares in India: A Study Raja Sridharan, Sridharan R. and Srinivas M. D., Nārāyaa
of Nārāyaa Paita’s Gaita-kaumudī, Chapters Paita’s Enumeration of Combinations and
13-14, Ph.D. Dissertation (Unpublished), Associated Representations of Numbers as Sums of
Brown University 1993. Translated into English Binomial Coefficients, IJHS 47, 2012, 607-631.
by Paramanand Singh, Gaita Bhāratī 1998-
2002. Raja Sridharan, Sridharan R. and Srinivas M. D., Differences
in Style, but not in Substance: Ancient Indian Study
Nārāyaī of Nārāyaabhaa (c.1555) on Vtta-ratnākara of Combinatorial Methods, in Sujatha R. et al Ed.,
of Kedārabhaa (c.950): Ed. by Vaidyanatha Sastri Math Unlimited, Science Publishers, Enfield NH USA,
Varkala, Kashi Sanskrit Series, Varanasi 1927. Ed. by
2012, pp.451-462.
Madhusudanna Sastri, Krishnadasa Academy,
Varanasi 1994. Sridharan R., Sanskrit prosody, Pigala Sūtras and Binary
Arithmetic, in Emch G. G., Srinivas M. D. and
Prākta-Paigala (in Prākta, c.1300): Published in
Sridharan R. Eds., Contributions to the History of
Kavyamala, Bombay 1894. Ed with commentary of
Mathematics in India, Hindustan Publishing Agency,
Viśvanātha Pa–cānana, Vaśīdhara etc., by Chandra
Delhi, 2005, pp. 33-62.
Mohan Ghosh, Bibliotheca Indica, Calcutta 1902. Ed.
by Vyasa B. S., 2 Vols., Prākta Text Society, Varanasi, Sridharan R., Pratyayas for Mātrāvttas and Fibonacci
1959, 1962. Numbers, Mathematics Teacher, 42, 2006, 120-137.