What are generally referred to as the Fibonacci numbers and the method for their forma- tion were given by Virahanka (between A.D. 600 and X00). Gopala (prior to A.D. 1135) and Hemacandra (c. A.D. 1150). all prior to L. Fibonacci (c. A.D. 1202). Narayana Pandita (A.D. 13Sh) established a relation between his srftcisi~ci-pcrirLfi. which contains Fibonacci numbers

as a particular case. and “the multinomial coefficients.” ((1 IYX? Academic Prey. Inc.

Avant L. Fibonacci (env. 1202 ap. J. C.). Virahanka (entre 600 et X00 op. J. C.). Gopala (avant I I35 ap. J. C.). et Hemacandra (env. I IS0 ap. J. C.1 intruidisirent les nombrec de Fibonacci ainsi qu’une methode de les generer. Narayana Pandita t 1356 ap. J. C.) etablit une relation entre son .rco,ftrsi~tr-ptr,rXti. dont les nombres de Fihonacci sent un cas particulier. et les “coefficients multinomiaux.” ‘ce 1985 Academic Pre\. Inc

Die gewohnlich nach Fibonacci bezeichnete Zahlenfolge

wurden von Virahanka (zwischen 600 und 800 nach Christus), Gopala (vor 1135). und Hemacandra (urn 1150) angegeben, die alle frtiher als L. Fibonacci (urn 1202) lebten.

wie such deren Bildungsgesetz

Narayana Pandita (1356) fand Fibonacci-Zahlen als Sonderfall 0 1985 Academic Pres. Inc.

eine Beziehung enthalten

zwischen sind. und

seinen srimcisi&-p&o?, worin den “multinomialen Koeffizienten.”

die

The name of Leonardo of Pisa, also called Fibonacci (I 170-1250), is attached to the sequence 0, I, I, 2. 3, 5, 8, 13, . . . , in which the nth term is given by U,, = U,-, + U,,-? [Smith 1958, 2171. But the sequence was well known in India before Leonardo’s time. Indian authorities on the metrical sciences used this sequence in works on metric.

The basic units in Sanskrit prosody are a letter having a single mdtrd (mora or a syllabic instant) called laghu (light) and that having two morae called gurrl (heavy). The former is denoted by 1and the latter by S, and their role in metric is the same as that of I and 2 in combinatorics.

Meters in Sanskrit and Prakrit poetry mainly fall under the following three categories: Vurpz-urttus are meters (urrtas) in which the number of letters (uarnas) remains constant and the number of morae is arbitrary. Mdtrd-urttas are meters in which the number of morae remains constant and the number of letters is arbitrary. Finally, there are meters (called ~a~cr-u+~~) consisting of groups (ganas) of morae such as the afyti, the uaitdiyu, etc. In the latter type the number of morae in a group remains constant and the number of letters is arbi- trary. However, the number of morae in different groups will, in general. be different.

A meter of the type uarna-urttus having one letter has two variations, a ~UW and a lughu, and the expansion (called past&a) of such a meter is obtained by writing the latter below the former, thus:

A meter of this kind having two letters has four variations. The expansion of such a meter is obtained by writing, in order, each variation in the former expansion on tile left of a guru, and then the same on the left of a laghu:

ss IS SI I I.

The same process of expansion is followed for meters having more than two la tters. Thus, the expansion of a meter having n letters is obtained by writing, in ol-der, each variation in the expansion of a meter having (n – 1) letters on the left o’ a guru, and then the same on the left of a laghu.

A similar expansion process is applicable to mcitr&urttas. Here, the number of v,u-iations of a meter of one mora is 1and its expansion is written as I. The number 01‘variations of a meter of 2 morae is 2, and its expansion is written as

The number of variations of a meter of 3 morae is 3 and its expansion is obtained b.1 writing the variation of the meter of one mora on the left of a guru and then writing, in order, each variation in that of 2 morae on the left of a laghu (see Table I). The same process of expansion is followed for meters having more than 3 morae. Thus, the expansion of a meter of n morae is obtained by writing, in order, each variation in the expansion of a meter of (n – 2) morae, on the left of a guru, a,ld then, the same of (n – 1) morae on the left of a laghu. Table 1, in which the

TABLE1

I mora 2 morae 3 morae 4 morae 5 morae

PARMANAND

TABLE

(6)/SSI

(7) s 1s 1

(8)IIISI (91SS( 1

SINGH

HM 12

(I) sss

(2) I ) ss (3) Is 1s 14)s1/s (5)IIIIs

(wlslI

(11) I SI I I (12)SI I I I

~~3~llIlII

II

number of morae in a meter is written above and its expansion below, illustrates this rule.

The expansion of the meter having 6 morae, along with the serial number of each variation, is as given in Table II.

The expansion of m&r&vrttas just described corresponds to a partitioning of a number (the number being the number of morae in the meter), where the digits take on the values 2 and 1 and their order is relevant; the number of digits in a partition, however, is arbitrary.

Here, it is easily seen that the variations of mtitrii-vrttus form the sequence of numbers which are now called Fibonacci numbers. For, the numbers of variations of meters having I, 2, 3, 4, 5, 6, . . . morae are, respectively, I, 2. 3, 5, 8, 13, . . . , and these are the Fibonacci numbers. It is also observed that the method for finding the numbers of variations of miitrd-vrftus leads to the general rule, U,, = U,,-, + U,,-? for the formation of Fibonacci numbers. Thus it can be safely concluded that the concept of the sequence of these numbers in India is at least as old as the origin of the metrical sciences of Sanskrit and Prakrit poetry.

EARLY DEVELOPMENTS: ACARYA PICJGALA AND ACARVA BHARATA

&rya Piligala is the first authority on the metrical sciences in India whose writings indicate a knowledge of the so-called Fibonacci numbers. In his commen- tary, the Vedbthadipika on ~ksuwinukramtmi. SadguruSisya writes that .&rya Pixigala was a younger brother of .&%rya P@ini [Agrawala 1969, lb]. There is an alternative opinion that he was a maternal uncle of Panini [Vinayasagar 1965, Preface, 121. The period during which Pgnini was active has been discussed by several scholars (such as A. A. Macdonell, A. Weber, G. A. Grierson, T. Gold- stucker, and others) who have, for different reasons, placed him between 700 B.C. and A.D. 100. Agrawala [ 1969, 463-4761, after a careful investigation in which he considered the views of earlier scholars. has concluded that Ptinini lived between 480 and 410 B.C.

According to YBdava [Sinharay 1977, 1051, a commentator belonging to the 10th century, Pidgala’s rule “miSrau ca” (i.e., “and the two mixed”)’ is also meant for the expansion of mat@vrttas, and m&r&vrttas should be expanded by combining the expansions of two earlier meters with a gum and a fughu, respectively.

I Throughout the paper the translations of the texts are by the author of the paper

Hvl 12 FIBONACCI NUMBERS IN INDIA 233

According to Yadava [Sinharay 1977,411 and another commentator, Halayudha [Madhusudana 1981, 461, who also belonged to the 10th century, Pirigala’s rule “r;au gantamadhyridirlaSca” gives the variations of a gana (i.e., a group of le, ters, the group having 4 morae) in an tiryd meter. Following these commenta- tors the rule may be translated as “Two gurus, a guru (in) the end, the middle, the baginning and (all) laghus.” Here, the variations have been stated in the order in w lich they are obtained by the expansion rule given above.

Acarya Bharata is the next authority on the metrical sciences whose writings indicate a knowledge of the so-called Fibonacci numbers. Various dates between ICO B.C. and A.D. 350 have been assigned by different scholars to NdtyaSastru of Al:&ya Bharata. Though the work has not come down to us intact, Kane [1961, 4C-471, after discussing the various possibilities, assigns the work a date earlier than A.D. 300 and not much older than the beginning of the Christian era.

According to Abhinavagupta. whose literary activities ranged between A.D. 980 arrd 1020 [Kane 1961, 2431, &%rya Bharata’s rule [Shastri 1975, 11921 for the e, pansion of m&r&vrttus is “. . . miSruu cetyupi mdtrikau.” This may be trans- 1ar:edas “. . . ‘And the two mixed’ is also (meant) for mtitrcr‘f-vrttus).” Again, it is to the process of mixing the expansions of two earlier meters with gwus and hohus, respectively, that this quotation refers.

In another Hindu work, Vi+r.zudharmotturu Purina, a chapter on Sanskrit pros- only indicates a knowledge of the so-called Fibonacci numbers.

According to Biihler [Gairola 1960, 23 I-2321 the Vi+pdhurmotturu Put+~u was probably written in Kashmir during the seventh century A.D. Verse 16 (Pt. 3. C~iap. iii) of the Pun+4 [Shah 19581contains the following rule for the expansion of mtitrii-vrttus: “. . . mtitrticchundbstarhaiva hi. rrktuvurr$k+wam (Pru) c(~hando hhavedekavivurjitah . . . ,” which may be translated as ‘I. . . Leaving one in the beginning (the process of expansion) of mdtr&v~ttas is similar to (that 00 uarrp-ucttas, as stated above.” We have already seen that this process of e>.pansion of m&r&urttas is similar to that of vuryu-vrttus. However, the expan- sirIn of a m&r&vrtta (having one mora) differs from that of a vurrpurttu (having one letter).

SO-CALLED FIBONACCI NUMBERS IN INDIA PRIOR TO A.D. 1200: ACARYA VIRAHAtiKA, GOPALA AND ACARYA HEMACANDRA

Actirya Viraharika is the first authority on the metrical sciences who explicitly gave the rule for the formation of numbers of variations of m&r&-vrttas (the so- called Fibonacci numbers).

Velankar [1962a, Introduction, xxi-xxv] has given an extensive analysis of the a6tivities of Actirya VirahBrika, concluding finally that &%rya Virahtirika lived sGme time between the sixth and eighth centuries. An English translation of Vlzlankar’s translation (Sanskrit) of the rule (Prakrit) [1962a, 1011follows: “The v&riations of two earlier meters being mixed, the number is obtained. That is a direction for knowing the number (of variations) of the next mat@-vrttu).”

Gopala is the first author on metric who specifically mentions the numbers of

234 PARMANAND SINGH HM 12

variations of mdtr&u+zs (the so-called Fibonacci numbers) while commenting on this rule of Acarya Virahatika. According to Velankar [1962a, Introduction, xxvii] a palm leaf manuscript of Gopala’s commentary on Vrttajcitisamuccaya of Acarya Virahanka, handwritten between A.D. 1133 and 1135, is available at Jesalmere. The manuscript forms the basis of this edition of the V~ttujcStisamuccaya. Gopala discusses Virahanka’s rule in some detail:

Variations of two earlier meters [is the variation] of a m&r&urtta.

For example, for [a meter of] three [morae], variations of two earlier meters, one and two, being mixed three happens.

For [a meter] of four [morae], variations of meters of two morae [and] of three morae being mixed, five happens.

For [a meter] of five [morae] variations of two earlier [meters] of three morae [and] of four morae. being mixed, eight is obtained.

In this way, for [a meter] of six morae, [variations] of four morae [and] of five morae being mixed, thirteen happens. And like that, variations of two earlier meters being mixed, [varia- tions of a meter] of seven morae [is] twenty-one.

In this way, the process should be followed in all mrirrri-urttas. [Velankar 1962a, IOI]

Acarya Hemacandra, one of the greatest Jain writers, is the other authority on metric who specifically mentions the numbers of variations of m&r&urttus in his ChandonuStisana (c. 1150). He lived at Anhilvad Patan in Gujrat and enjoyed the patronage of two kings, Siddharaja and Kumarapala. Btihler states that Ac&rya Hemacandra wrote ChandonuSdsana and its commentary between A.D. 1142 and 1158 [Banthia, 1967, Preface, 16-27, 57-601. Acarya Hemacandra’s rule [Ve- lankar 1961, 2391 may be translated as follows: “Sum of the last and the last but one numbers [of variations] is [that] of the m&d-urtta coming next.” Acarya Hemacandra also comments on his own rule, as did Gopala on Virahanka’s rule, concluding with “Statement-l. 2, 3,5,8. 13, 21, 34 and in this way, afterwards” [Velankar 1961, 2391.

The rule for the formation of numbers of variations of mdtrcf-urttas continued to be given in works on metric even after A.D. 1200. Kauidarpana is such a work on metric which deals with the topic. Velankar advances reasons to establish that Kauidarpana was composed during the 13th century A.D. [1962b, Introduction, iv]. The author of Kaidurpana states a rule [Velankar 1962b, 67-681 for the forma- tion of numbers of variations of m&d-qrttas that is essentially the same as that given by Hemacandra.

Thus we see that the sequence of the so-called Fibonacci numbers resulted as a natural consequence of the variations of m&r&u$tas of I, 2, 3, . . . , morae.

THE PRAKRTA PAlr;JGALA

Prtikrta P&gala is another important work on metric that gives several rules regarding the so-called Fibonacci numbers. The period of composition of Prtikrtu Paiizgala has been discussed by Vyasa, who concludes that Pr&krta Paitigala was written during the first quarter of the 14th century A.D. [1962, 6-201. Probably the present form of Pnikrta Paiizgah is an amplification of an old work [Velankar 1%2a, Introduction, xiv].

HM 12 FIBONACCI NUMBERS IN INDIA

235

In connection with the so-called Fibonacci numbers, the author of Prdkrta Faihgala gives rules regarding naga and uddiga analyses of mitrd-v+as. These a-e the usual combinatorial problems discussed by the Hindus in metric and are rc:lated to measurement of morae and letters in verses. The aim of the nas!a a ralysis is to find n@a-bhedu (the unknown structure) of a particular variation of a given meter, the uddisttihka (the serial number of this variation from among all variations of the meter) being given. Conversely, the aim of the uddista analysis is to find uddist&iku of a variation amidst all variations of the meter, the nm!u- bbedu of the’variation being given.

The author of Prtikctu Puihgalu gives the following rule for finding nasfu-bheda ojrresponding to a given uddisfahku:

In na+(-bhedu) make all morae, kughus. Give [i.e.. write] the number equal to the pair of numbers, overhead. Omit [i.e., subtract] the numberasked [i.e., given]. Write the remainders after subtractions and omit the remaining numbers. Piligala NPga says that lagl~us of the parts takes laghus of others [i.e., of those lying ahead] to become gurus [and] writing thus, the writing [i.e.. the structure] comes immediately. [Vyasa 1959, 321

Let us consider the fourth variation of the meter having 6 morae. Making all niorae laghus and writing the number equal to the pair of numbers (i.e., 1,2, 3,5, 8. 13, . . .) above, we obtain

123.5813

lllll 1.

The number 4 (of the variation) is subtracted from 13, giving 9. Next the numbers written above are subtracted from 9, and then from the remainders, repeatedly. Thus we have 9 – 8 = 1 and 1 – 1 = 0. Laghus below the subtracted numbers (i.e., below 8 and 1) combine with the next laghus to give gurus. Other laghus rr:main as they are. This gives the structure of the desired fourth variation to be S11s (see Table II).

The author of Prdk$a P&gala gives the following rule for finding uddi@zka a)rresponding to a given naffa-bhedu: “Write the number equal to the pair of ru.tmbers, above. Omit all the numbers (except those) over gurus. Having written, bring the remainder. Know that as uddi@a” [Vyasa 1959, 311.

The uddi+fahka may be found for the structure as given above. Writing the n.rmbers equal to the pair of numbers (i.e., 1, 2, 3, 5, 8, 13, . . .) above (and also b,:low gurus, in order) we obtain

13.5 8

SII s 2 13.

Subtracting the numbers over gurus (i.e., 1and 8) from the number of variations of the meter (i.e., from 13), 4 is obtained as the remainder; this is the desired uddis- fGzka of the variation whose structure is given (see Table II).

The next rule related to the so-called Fibonacci numbers, as given by the author oi“ Prtikrta Paihgalu, is meant for finding the serial numbers of variations of a

236 PARMANAND SINGH HM 12

meter having a particular number of gurus (or laghus) among serial arrangemenl of all variations of the meter. The rule may be translated as follows: “Keep the numbers like [those iu] uddisfu. Subtract the numbers, leftwards. Bring a single guru [for] single subtraction. Know double, triple gltrus [for] double, triple sub- tractions” [Vyasa 1959, 43-441.

To illustrate the rule, consider the variations of a meter having 6 morae, letting the variations be arranged serially (see Table II). Obviously, 13 gives the serial number of the variation having no guru (see Table II). The numbers of variations of meters having 1, 2, 3, 4, and 5 morae (i.e., 1, 2, 3, 5, and 8) are subtracted from the number of variations of the meter having 6 morae (i.e., from 13), in the reverse order. This gives 5, 8, 10, 11, and 12, which are the serial numbers of the varia- tions having a single guru, amidst all variations of the meter (see Table II). Again the same numbers (i.e., 1, 2, 3, 5, 8, . . .) are stibtracted from each of the remainders, separately. Leaving the numbers that appeared earlier as remainders gives 2, 3. 4,6. 7. and 9, which are the serial numbers of the variations having two grtrcts (see Table II). The process is repeated, giving 1, the serial number of the variation having three gurus (see Table II).

We have already seen the correspondence between partitions of a number (the number being the number of morae in the meter)-where the digits take on the values 2 and I, their order being relevant and the number of digits in a partition being arbitrary-and the expansion of mtitrti-urttus. Correspondingly, the rule gives the serial numbers of p-partitions of n (i.e., partitions containing p digits each, where nis the number of morae in the meter for p = n, n – 1,. . .amidst all partitions of n).

Still more interesting is the rule given in Prakrir by the author of Priikrta Paiizgala for the formation of matta-meru, or rniitrti-mew, (menc-like table for mcitrtiryttas where meru is the name of the imaginary mountain that is supposed to stand at the center of the earth). It gives a method for determining the numbers of variations (of a meter) having a definite number of gurus (or laghus) from among the variations of the meter. It also establishes a relation between the sequence of binomial coefficients and the (so-called) Fibonacci numbers, thus providing an alternative method for the formation of these numbers. On the basis of the Hindi commentary, the text [Vyasa 1959, 41-421 may be analyzed. Initially (two rows of) two cells (each) are formed. (Next, two rows of 3. 4, etc., cells each are formed.) Unity is (written) in the last cells of these (rows). Again unity is (written) in the first cells (of the first, the third, the fifth) . . . , rows and 2, 3, 4. . . , in the first cells of the second, the fourth, the sixth, . . . , rows, in order. Every other cell is filled by (the sum of) the number lying above (that cell or above that cell to the left, added to) the number above the (latter number and to its right). Mcitrti-meru for a meter of 7 morae and formed according to this rule is shown in Table III. Here unities have been kept in the last cells of all the rows. For an integer m, unities have been kept in the first cells of the (2/n – I)st rows, and m+1inthoseofthe(2m)throws,m=1,2,3, ….Thus,1hasbeenkeptineachof the first cells of the first. third, and fifth rows, while 2, 3, and 4 are in those of the

I-M 12

FIBONACCI

NUMBERS

IN INDIA

=Z

3

=5 =8 =13 =21

237

TABLE

I11

dvikala trihlu cafuskalu paticukala +YllU snptakalu

(2 morae) (3 morae) (4 morae) (5 morae) (6 morae) (7 morae)

sb:cond, fourth, and sixth rows. In the remaining cases, a cell of the (2m – 1)st row contains the sum of the number lying above it and to the left with the number in the cell above the latter ceil and to the right. Thus, for the second cell in the third n)w and the second and third cells in the fifth row, we have 2+ I = 3. 3+ 3= 6, and 4 + I = 5 as their entries; for a cell in the (2m)th row we use the sum of numbers in the cell just above it and in the cell above the latter cell and to the rtght. Thus the numbers in the second cell of the fourth row and the second and third cells of the sixth row are 3 + I = 4, 6 + 4 = 10, and 5 + 1 = 6.

It will be observed that the numbers in the cells of the (n – 11th row, from right to left, are the numbers of those variations of the meter of n morae having 0, 1, 2, . . . gurus, respectively, and the sum of the numbers in that row is the number of variations of the meter. This implies that these are the numbers of p-partitions of n,forp=n,n- I,. . . , and that their sum is the number of all partitions of n.

The m&r&met-u also establishes a relation between the sequence of binomial coefficients and the (so-called) Fibonacci numbers. Let CJ,,denote the number of variations of a meter having n morae. Expressing n as 2m or 2m + I, according to v,hether it is even or odd, then, as is clear from the meru,

,,,

U,, = 2 n-rc, r=O

This gives another method for the formation of the so-called Fibonacci num- bers. Here, it is also seen that numbers in a cell of the m&r&meru are the binomial o>efficients (or their sum) formed according to a definite rule.

FURTHER DEVELOPMENTS: THE GANITA KAUMUD! OF NARAYANA PANPITA

The Ganita Kuumudi was the first mathematical work in which the ideas of the sl,-called Fibonacci numbers were developed further. From the colophon toward the end of the book we learn that the Gqitu Kuumudi was written by Ntirtiyana Pandita in A.D. 1356 [Dvivedi 1942, 4111. In Chapter 13 of the book the author defines sdmdsikd-puizkti (additive sequence). The (so-called) Fibonacci numbers a-e a particular case of this puizkti (sequence).

238 PARMANAND SINGH HM 12

Narayana’s sdmdsikd-parikti is his tool for the treatment of permutations, com- binations, partitions, etc.; it is defined with respect to “the greatest digit.” For a particular permutation, combination, or partition, we have the following corre- spondences between Naryana’s descriptions and contemporary language and no- tation: the digits from one to “the greatest digit” (= 9); the “number of places” is simply the number of digits (= p); the “sum of digits” (= n) is just that.

Narayana’s rule for the formation of s&mtisik&pui?kti may be translated as follows:

First keeping unity twice, write their sum ahead. Write ahead of that, the sum of numbers from the reverse order [and in] places equal to the greatest digit. In the absence of [numbers in] places equal to the greatest digit, write ahead the sum of those [in available places). Numbers at places [equal to] one more than “the sum of digits” happen to be “the scimcisikci- puitkri.” [Dvivedi 1942. 290-2911

Let v(9,r) denote the rth term of the sdmclsikd-parikti when “the greatest digit” is9. According to the rule, ~(9, I) = I and u(9,2) = 1. For other values of P,if 3I r 5q,u(q,r)=u(q,r- I)+u(q,r-2)+**.+u(q,2)+49,I);andif9<r,u(q,r)= u(q,r- I)+u(q,r-2)+.. . + ~(9, r – 9). The sequence is finite having (n + I) terms, n being “the sum of digits.”

For 9 = 2 we obtain the sequence I, I, 2, 3, 5, 8, . . . , which are the so-called Fibonacci numbers. Therefore, these numbers are a particular case of the s&m- Ckci-paizkti of Narayana Pandita.

The relation between the so-called Fibonacci numbers and the binomial coeffi- cients as contained in the Prdkrta P&gala has already been established with the aid of the figure of numbers, the mtitrd-meru. It has also been observed that the numbers in the cells of a mdtrci-meru are either the binomial coefficients or their sum, formed according to a definite rule. Similarly, Nrirayana Pandita established a relation between his stimcisikd-paizkti and the multinomial coefficients. Here, the relation is established with the help of another figure of numbers called matsya- mew (fish-mere). Like mdtr&meru, numbers in cells of a matsya-meni form “the sequence of multinomial coefficients,” or stici-paizkti (needle-like sequence).

S&f-puizkti is defined with respect to “the number of places” (= p) and “the greatest digit” (= 9). It is formed with the help of uais’feSi!i-paizkti (sequence of separated units) of measure 9, which consists of a sequence of 9 1’s. Narayana‘s rule for the formation of saci-parikti follows.

Put the vdle+-pahkfi of measure [equal to] the greatest digit. separately, in places equal to the number of places. Their [final] product is the slidpaizkti [needle-like sequence]. [Dvivedi 1942, 2951

Thus for p = 3 and q = 3 the uais’fesini-puizkri of measure 3 is 1, 1, 1. Keeping uaidesini-puizkti at 3 places and multiplying successively, we get 1, 3, 6, 7, 6, 3, and 1 to be the final products (see Table IV). These products make up the desired slici-puizkti when p = 3 and 9 = 3.

The uuis’le+i-puizkti of measure 9 (i.e., q I’s) are the coefficients of x’ in the polynomial1 +x+x2+…+xY-‘,r=0,1,. . . .9- I.Therefore,fromthe

HM 12

FIBONACCI NUMBERS IN INDIA 239

TABLE IV FORMATION OF SC-PA~JKTI

FORP=3ANDq=3

1,1,1 (1st place) 1,1,1 (2nd place)

1.1,1 1.1.1

1,1,1

1.2,3,2,1

l,l,l (3rd place)

1,2,3,2,1 1,.X3,2,1

1.2.3.2.1

method of formation of the saci-paizkti it is clear that when the greatest digit is 4 and the number of places is p, then the (r + I)st term of the sequence, say u(p, q, r), is the multinomial coefficient of x” in the expansion of (I + .x + x? + * . * +.w’)P, r=0,1,2,..‘,(y- I)p.

From the process of addition of the products in this method of obtaining the si’ci-pahkti, an alternative method of formation is also apparent. Let two such ssquences be formed with respect to the same greatest digit 4, but for a different number of places. Assuming that the number of places for the first sequence is p –

1 and p for the second, any term of the second sequence can be obtained from th’ose of the first. Consider the (r + 1)st term of the second sequence. If Y+ I I 9, then this term of the second sequence is equal to the sum of the first r terms of the first sequence. If, however, q< r + 1I (p – I)(q – 1)+ 1,this term isequal to the sum of 4 terms, up to the (r + I)st term of the second sequence. Finally, if (r + 1)> (p – I)(q – 1)+ I, then the (r + I)st term ofthe second sequence isthe

sq:rn of the terms of the first sequence that follow the (r – 4 + 1)st term. __

Narayana makes use of this method of formation of the sucl-paizkti (needle-like suquence) in the formation of his figure of numbers named matsya-meru. The figure is formed with respect to the greatest digit q and the sum of digits n. His rule fcr the formation of matsya-meru follows.

[Form] one [cell] initially [and then] lines of cells, [the cells in a line] increasing by “one less than the greatest digit,” till the measure of that is one more [than the sum of digits]. Leaving the first cell and starting below the second. form horizontal lines [of cells] equal to the sum of digits.

After that, putting [one] initially, keep unity in the [initial] stretched line. [Take cells] above one’s own cell in the reverse order and equal to the greatest digit. Write the sum of numbers [in them] in the lower cell. If there be an absence [of cells] equal to the greatest digit, [take] the sum of numbers as far as possible. Work, in order. [Dvivedi 1942, 304-3051

To illustrate the rule, Narayana considers the case when q = 3 and IZ = 7,

240

PARMANAND SINGH TABLE V

HM 12

MATSYA-MERU

`..-`

(’ The figure of mcr~swr-rnent given in the printed text seems to be

l-!–i

__-.~

p=0 p= I p=’

p=3 p=4 p=s p=6 p=J

vitiated. figure available script

for it does not shown here is based at Sampurnanand

satisfy the upon a

Sanskrit

rules [Dvivedi manuscript

University,

of

- Gtrnilrr

Varanasi - The Ktrrrmudi.

[Manu-

No. 35668/p. 7 recta].

obtaining the numbers of cells in a line as I, 3,5, and 7. By the above process he arrives at the matsya-mere as shown in Table V.

As for the relation between terms of a stimtisikd-pahkti and those of a stici- pahkti (i.e., multinomial coefficients), we see that for all values ofp and q, the first term of a samdsikti-pahkti is equal to the first term of a stici-paizkti, each being equal to 1. For other values ofp and q, we observe that v(q, 2) = ~(1, q, 0) and u(q, 3) = ~(2, q, 0) + ~(1, q. l), both sides of these equalities being equal to 1and 2, respectively. Similarly. u(q. 4) = ~(3, q, 0) + ~(2, q, 1) + . . . , and so on. In general. u(q, t + I) = u(t, q, 0) + u(t – 1, q, 1) + . + . + rr(t – k, q, k), where k = p + r – t and t I (p + r)ly < i + 1. A comparison of these relations with the matsya-meru (when the greatest digit is q) shows that the right sides correspond- ing to the rth relation yield the numbers in cells of the rth vertical line of the matsya-mew. Narayana’s rule for this case is precisely this: “Siimhsikti-pahkti is formed by the sum of numbers in vertical cells of that” [Dvivedi 1942, 3051. Thus we have another method for the formation of stimtisikti-pahkti, of which the so- called Fibonacci numbers are a particular case.

It has already been shown that the numbers of variations of mdtra-urttas are the so-called Fibonacci numbers. It has also been observed that the variations of mdtr&urttas correspond to a partitioning of a number where the digits take on the values 2 and 1 and their order is relevant. So, there exists a relation between the so-called Fibonacci numbers and such partitions. Since the former constitute a particular case of sdmdsikd-pahkti, there must exist partitions of a number corre- sponding to the scimdsikti-paizkti. Narayana discusses such partitions of a number

and establishes several relations between the two. For q 5 IO, his partitions contain —

all the digits between 1and q. Narayana’s rule for expansion of such partitions is HM 12 FIBONACCI NUMBERS IN INDIA 241

d:scribed next [Dvivedi 1942,339]. In the first partition the greatest digit is written ir all places and, if required, on their left that digit is written which when added y elds the given number. Starting from the beginning each time, the first available d git different from unity is replaced by the next lesser digit. Other digits on the right of the replaced digit are written as described above. Clearly, the digits on the left of a replaced digit are all 4, but they may be any lesser digit, if at the begin- n ng.

Narayana illustrated his rule with the help of two expansions in which n = 7 and p 5 7 [Dvivedi 1942, 3401. In the first expansion (see Table VI), q := 7, whereas in tile second expansion (Table VII), 4 = 3. In all cases we observe that the number oi‘partitions of n = u(q, n + I) and the number of partitions ending in I, 2, . . . are equaltou(q,n+I-t),wheret=I,2,….HereNarayana’s ruleis,“Last

nllmber of s&ziisik&paizkti is the number of variations. Starting from the penulti- rrate, numbers in the reverse order are the numbers of variations ending in 1, etc.” [Dvivedi 1942, 3351. Thus in the first example, where q = 7, the stimasikti- phkti is 1, I, 2,4,8, 16,32,64; since n = 7. the number of partitions of 7 is 64, and ttle numbers of partitions of 7 ending in I, 2, . . . , 7 are, starting from the pr:nultimate in the reverse order, 32, 16, . . . , I. Similarly, since y = 3 in the second example, the samcIsik&paizkti is I. I, 2. 4, 7, 13, 24, 44, and n = 7 implies tll at the number of partitions of 7 is 44 and the numbers of partitions of 7 ending in 1, 2, 3 are. starting from the penultimate in the reverse order, 24. 13, and 7. Here

SERIAL

TABLE VI OF VARIATIONS

(I)

(2)

(3)

(4)

(5)

(6) 124

61 I51 241 1141 331 1231 2131 11131

(49) 511 (50) 1411 (51) 2311 (52) II311

(7) (8) (9)

(IO) (II) (12) (13) (14) (15) (16)

214 III4

43 133 223 1123 313 1213 2113 11113

(43) (44) (45) (46) (47) (48)

2221 11221 3121 I2121

21121 111121

7 16

52 142 232

(33)

(34)

(35)

(36)

(37)

(38)

(39)

(40)

(41) 421 (42) 1321

25 II5

II32 322 1222 2122

ARRANGEMENT

(17)

(18)

(19)

(20)

(21)

C21

(23)

(24) Ill22 (25) 412 (26) 1312 (27) 2212 (28) 11212 (29) 3112 (30) 12112 (31) 21112 (32) 111112

FOR n = 7. p I 7.

ANDY

= 7”

34

(53) (54) (55) (56) (57) (58) (59) (60) (61) (62) (63) (64)

3211 12211 21211

111211 4111 I3111 22111 112111 31111 121111 211111 1111111

” Omissions in variations numbering 4, 15, 49, and 54, found in [Dvivedi 1942. 3401, have been completed with the help of the manu- script of Gut+ Kaumudi [Manuscript No. 35668/p. I6 recta], referred to in Table V. Similarly, variation No. 27 has also been corrected. Variation No. 30 has been changed so that it satisfies the rules. 242

PARMANAND SINGH

TABLE VII

SERIAL ARRANGEMENT OFVARIATIONS FORn = 7, p 5 7.

HM 12

y = 3”

(23) 2131 (24) III31 (25) 1321 (26) 2221 (27) II221 (28) 3121

AND

12) 2122 13) 11122 14) 1312 15) 2212 16) 11212 17) 3112 18) 12112

(19) 21112 (20) I1III2 (21) 331 (22) 1231

UVariation

table on the basis of the manuscript of Gu$u KaumuJT [Manuscript No. 35668/p. 16 recta] referred to earlier. Variation No. 30 has been changed so that it satisfies the rules.

(1) 133 (2) 223 (3) 1123 (4) 313 (5) 1213 (6) 2113 (7) 11113 (8) 232 (9) II32

(10) 322 (II) 1222

(29) (30) (31) (32) (33)

I2121

21121 111121 2311

II311

(34) 3211 (35) 1221I (36) 21211 (37) 111211 (38) I3111 (39) 22111 (40) 112111 (41) 31111 (42) 121111 (43) 211111 (44) II I I I I I

No. I5 in [Dvivedi 1942, 3401 has been corrected in this

we also note that if CJ= 2, the partitioning is the same as that corresponding variations of rmitr&v~ffas, and the samtisika-pahkti (as observed earlier), become the so-called Fibonacci numbers.

The naga-bheda and uddis;dhka for such partitions are obtained with the help of unmeru (elevated-meru), another kind of figure of numbers. Narayana’s rule for the formation of unmeru is described next [Dvivedi 1942, 341-3421: One cell is formed in the first row. The number of cells increases by one in each subsequent row until the number of cells in a row is one more than the sum of the digits. The scimasikd-pahkti is written in the cells of the lowest row; natural numbers, in the reverse order, appear in the cells of the remaining rows. Naturally, the numbers in the columns are also natural numbers. The numbers greater than “the greatest digit” are omitted.

To illustrate his rule Narayana considered the case when n = 7 and y = 3, obtaining the figure shown in Table VIII.

Narayana’s rule for finding the naga-bheda corresponding to a given uddis- ghka follows.

Subtract the na+&~ from the last number of the sdmasikd-parikti. Subtract from the remainder, the numbers nearest to the remainders until no subtraction is possible. The digit obtained in the cell common to the horizontal line and the vertical line is the digit of the na+z[-bheda]. [Write] those numbers in a series. In this way, in [finding] nu+[-bheda], work this process dependent upon unmeru, repeatedly. [Dvivedi, 1942. 3431

The method seems to be similar to that for finding the naga-bhcdu corresponding to a given uddi+ihka as contained in the Prcikga Paihgala and meant for m&r& v+zs. In such a case the horizontal line (row) and the vertical line that determine a common cell are chosen as follows: In the beginning the row is just above the row containing the stimdsikd-pahkti. Later on it is just above the row containing

to the Hlvl 12 FIBONACCI NUMBERS IN INDIA 243

th: uppermost cell of the column whose digit has been already noted. The column is the column whose digit could not be subtracted. Moreover, the series is to be formed from right to left.

To clarify this rule, we will find the naga-bheda when the uddi@hka is 36, n = 7, and q = 3. Here, the stimdsikti-puhkti will be I, I, 2,4,7, 13,24,44. Subtract- in: 36 from 44, the last number of the sdmrisikd-puhkti, 8 is obtained. The next nq mber of the sdmcisikci-pahkti to the left of 44 is 24, which cannot be subtracted from the remainder 8. The digit of the unmeru in the row above the one containing sdmdsikd-puhkti and in the same column as 24 is I, which is written separately. A/gain, the next number of the scfmdsikd-puhkti to the left of 24 is 13, which also cannot be subtracted from 8. The digit of the unmeru common to the column cdntaining I3 and the row which isjust above the uppermost cell in the column of 24 (for which 1has already been computed as the digit of the rrnmerrr) is I ; this value is situated to the left of the 1obtained earlier. Next, the number of the sdmdsikd- perhkti to the left of 13 is 7, and 7 subtracted from the remainder 8 yields 1. The next number of the scimtisikd-puhkti, to the left of 7, is 4, which cannot be sub- tnucted from the remainder 1. The digit of the unmeru common to the column of 4 add the row which is just above the column of 13 is 2, which is placed to the left of 11 (as noted above). Again, the number in the scim&ik&puhkti to the left of 4 is 2. winch cannot be subtracted from the remainder 1. The digit of the unmeru com- m on to the column of 2 and the row just above the column of 4 is I, which is placed

TABLE VIII

THE UNMERU FORH = TANDY= Y

a The figure in [Dvivedi 1942, 341 also con- tains some errors and has been corrected on the basis of the manuscript of Ganiza Kaumudi [Manuscript No. 35668/p. 16 versa], where the first vertical line of the elevated part is absent. The second vertical line, though absent in the book, appears in the manuscript. Both the book and the manu- script give the samasika-pankti as I, 2,3, . . for 1, 1,2, . . . (in the table). 244 PARMANAND SINGH HM I2

to the left of 21 I (determined in the preceding step). The number in the sdmdsikti- pahkti to the left of 2 is I and 1 – I = 0. Finally. the number in the sclm&ik& pahkti to the left of 1 is I, which cannot be subtracted from 0. The digit of the rrnmenr common to the column of this I and the row just above the column of 2 is 2, which is written to the left of 121I: thus the desired partition is determined to be 2121 I.

The uddi@ka corresponding to a given nas[u-bheda is obtained by a process that is the reverse of that meant for nasla-bheda [Dvivedi 1942, 3441.

ACKNOWLEDGMENTS

The author expresses his sincere gratitude to Dr. K. S. Shukla. (Retired) Professor. Lucknow University, for his guidance and help. The author also expresses his gratitude to the Vice-Chancellor

of Sampurnanand Sanskrit University. Varanasi, for making available to the author photocopies of the

manuscript

Agrawala. Banthia.

referred to in the paper.

REFERENCES

V. S. 1969. P@zinikrilina BhCirutuuur.~u (Hn.). Varanasi-I: The Chowkhamba Vidyabhawan.

K. (Tr.). 1967. Hemucundrticxir~u Jivnnucuritru (Hn.). Varanasi-I: The Chowkhamba Vi-

dyabhawan.

Dvivedi, P. (ed.). 1942. T/r, Gu+u Kuftnzudi hv Ndrciyu?cI Pu~~iru, Pt. Il. The Princess of Wales Sarasvati Bhavana Texts No. 57. Varanasi. Gairola. V. 1960. Suns& Sdhityu Xti S&nk.$m Itihdsu (Hn.). Varanasi-I: The Chowkhama VI- dyabhawan. Kane, P. V. 1961. Hisfoe of Sot&it porfks. Delhi-6: Motilal Banarasidas. Madhusudana. V. (ed.) 1981. Pitigulu Ntigu Virucimn ChonduhS6srrum (with a commentary by Ha- lgyudha). Parimala Publications, Shah. P. (ed.). 1958. The Visnudllurmoltur~~

33117.

Muni Benares

Delhi-l 10007: Shakti Nagar.

Purrinu, Vol. I. Pt. 3. Baroda: Gaekwad’s Oriental Series

Shastri.

No. cxxx. Oriental M. (ed. ). - Pt.

Ins&te. Bhtrrtr/cr

Virtrcitutn N~i/,vuS~is~rtrm Hindu University.

Yfiduuttpnt~ci.ic~~~~um. Ahhintrvcr~~rpitrprr~~ii~b~~inuucr-

hhdrufi.suhirum. Sinharay. H. Smith. D. E. Velankar. H.- D.

II. Varanasi: C~tunduhsrirruhhdsvurn

Calcutta-16: Dover.

Bombay: Bhartiya

The Asiatic

Samiti. Bhavan.

Society.

Institute. Oriental Re-

Hisroryv (ed.). 1949.

of Muthetnurics. Jrrwdtirnun.

Vol. Haritosam?&

Sk.

I.

New York:

- D.

- – 1962a.

ChundonrrS6sunu V~~rujLltisam~c~~~u~~~

of Hemucundru ofKuui

Rajasthan Bhuitu. Jodhpur:

Research Rajasthan

No. Bombay-7:

I.

Haritosh Vidya Oriental

Fibonacci numbers stolen from Indian mathematician

%d bloggers like this: