Cubics in China

The solution of numerical higher equations for approximate values of roots has been known for a long time in China. It has been called the most characteristic Chinese mathematical contribution. That it was well developed in the work of the Sung algebraists has long been known, but it is possible to show that if the text of Han Jiuzhang [Suanshu], (Nine Chapter of the Mathematical Art) is very carefully followed, the essentials of the method are already there at a time which may be dated as of the -1st century.

``The Chinese also developed a similar method to find the cube root of a number and in so doing were able to solve a cubic equation of the form x3 + ax2 + bx = c, where a, b and c are positive. Through the method of finding a side of a square or cube with the rod numeral system, the Chinese invented a notation to represent quadratic and cubic equations and  were able to solve numerically a particular type of such equations. One could say that at this stage there were indications that arithmetic was forging into the yet to be known field of algebra.'' (AHES 47 (1) 1994, 1-51).


``The ancient Chinese performed the four fundamental operations on fractions and the extractions of square and cube roots as easily as we perform these operations using our numeral system. More important, in spite of the two thousand years' gap and the difference in media, the procedures that they used bear parallel resemblance to the ones we use.
[...] (AHES, 37 (4) 1987, 365 - 392).

It means that the principles and methods of extraction of square roots and cube roots as explained in Jiuzhang Suanshu have gradually evolved into a method of solving polynomial equations, a method known as "zeng cheng kai fang" (extraction of roots by addition and multiplication) in the Song period (9th -- 11th century), finally capable of dealing with equations of arbitrarily high degrees and coefficients positive, negative, and zero. Such numerical methods are indeed spigot algorithms which give a positive root of a polynomial equation digit by digit. In practice, each of the digits is determined by deliberation.

In Jiuzhang Suanshu and other Chinese texts, the digits obtained in this way are called `shang', the same character for the word `quotient'. This character `shang' (deliberation) captures the essence of the division process, (unless one insists on performing division by repeated subtraction).

``Among the problems in the Jiuzhang, one, the finding of a cube root, may be written
x3 = 1,860,867, and another, involving the square and the simple term,
x2 + 34x + 71,000 = 0. Naturally, these equations were not written as such, but set up with counting rods on a board. The uppermost line in the table is to contain the radical solution sought, the second line (shi) is the constant term before the operation, the third (fa, or later in the operation, ting fa) is for numbers obtained in its successive stages, the fourth (zhong) receives another number found at an intermediate state, and the fifth (jie suan) is the coefficient of x3 before the operation.

Numerical equations of degrees higher than the third occur first in the work of QIN Jiushao around 1245 AD. He deals very clearly with equations such as:

-x4 + 763,200x2 - 40,642,560,000 = 0.

 

The extraction of cube roots on the counting board.


Starting with say, 1860867. The digits are grouped by three digits from right to left

--- 1 860 867.

Make an estimate of some A such that A3 does not exceed 1, in this case take A=1. Subtract 1*106 from the number to obtain 860 867.

The next estimate of some B is made such that (3*A2*102+3*A*B*10+B2)*B does not exceed 860, in this case take B=2 ( because 364*2 = 728 while 369*3 = 1107 ). Continue in a similar fashion.

In this particular case, the third estimate gives C=3 and the procedure ends, yielding the answer 123. In general, the result can be calculated with an accuracy to within any decimal place. The mathematical basis is the identity
(10*A+B)3 = 1000*A3 + (300*A2+30*A*B+B2)*B.

 

 

A comparison of the method on the counting board (given in Jiuzhang Suanshu) is described below. It is remarkable how similar they are.

The extraction of cube root is carried out on a counting board, in five rows. The first row is for the root. In the second row one puts down the digits of the dividend (the number whose cube root is to be extracted). The third and fourth rows are left blank in the beginning.
They are respectively referred to as the divisor and the middle number. In the fifth (and bottom) row, under the UNITS digit of the dividend is a borrowed rod (jie suan).

The method begins by moving the borrowed rod to the left, jumping two places at a time. This has the effect of dividing the digits of the dividend into blocks of threes.

Suppose the number is not the cube of one of the digits 1, 2, ... 9. (The problem would then be trivial). Figure out the first quotient (digit of the root).

``Multiply it twice to the number represented by the borrowed rod; use this number as divisor (fa) to perform division in the first block. Treble this and leave it as ``fixed divisor'' (ding fa, in the third row). Once more, divide this divisor by the quotient and leave the result in the
next row, treble it as the middle number; again, put a borrowed rod in the bottom row''.

All these operations have the effect of

(i) removing the cube of the quotient from the first block, this first quotient is so chosen that its cube is just no greater than the three - digit number in the leftmost block),

(ii) putting 3 times the first quotient in the fourth row as ``the middle number'',

(iii) putting 3 times the square of the first quotient in the third row as divisor,

(iv) leaving the borrowed rod in the bottom row.

``Now, shift the fixed divisor one place to the right, the middle number two places, and the borrowed row in the bottom row three places.

This is illustrated in Problem IV.22 of Jiuzhangthis where the cube root of 1,937,541(17/27),
is calculated as . Ignoring the fractional part, 1 will be the first quotient, and following the method it will be as follows;

1 quotient
937,541 dividend
3 fixed divisor
3 middle number
1 bottom number


*****************

``Figure out the quotient again; multiply it to the middle number; multiply it twice to the bottom number; add these (two) to the fixed divisor''.

12 quotient
937,541 dividend
364 fixed divisor
6 middle number
4 bottom number


``Perform division with the fixed divisor. After the division, double the bottom number, add this, together with the middle number, to the fixed divisor''.


12 quotient
209,541 dividend
432 fixed divisor
6 middle number
4 bottom number


``Divide again (the fixed divisor) by the quotient not just the digit, but the number in the top row), and put the answer in the next row''.

12 quotient
209,541 dividend
432 fixed divisor
36 middle number
1 bottom number

``Shift the fixed divisor one place to the right, the middle number two places, and the bottom number three places''.

12 quotient
209,541 dividend
43,2 fixed divisor
36 middle number
1 bottom number

*************

At this point, the text indicates that ``if the division does not terminate, then the cube root cannot be extracted precisely. It is intended, of course, that the above process (of choosing a digit and adjusting the bottom, middle numbers and the fixed divisors) be carried out up to the block containing the units digit. Then it deals the case when the dividend has a fractional part. LIU Hui has remarked (in his commentary on the square root process) that this can be continued as long as required.

*************

In the present example, the next quotient is 4:

124 quotient
209,541 divident
43,2 fixed divisor
1,44 middle number
16 bottom number

_____________________

124 quotient
209,541 divident
44,656 fixed divisor
1,44 middle number
16 bottom number

_____________________

124 quotient
30,917 divident
46,128 fixed divisor
3,72 middle number
1 bottom number

_____________________

To work beyond the decimal point, one would shift the numbers in the three bottom rows as follows, and choose 6 for the next quotient:

124 quotient
30,917,000 divident
4,612,8 fixed divisor
37,2 middle number
1 bottom number

 

Wang Xiaotong (fl. 625) Mathematician and astronomer. Wrote Xugu suanjing (Continuation of Ancient Mathematics) of 22 problems. Solved cubic equations by generalization of algorithm for cube root.

According to J. Needham, Science and Civilization in China, vol 3, p. 65ff, the square and cube root algorithm already existed in the Chinese text: Chiu Chang Suan Shu (mathematical treatise in nine sections) by about 100 BC to 100 AD. The implementation on the abacus apparently occurred around the 13-th century (Yuan = Mongol Dynasty) when the abacus became the *calculator of choice* replacing the *counting rods*. The algorithm anticipated Horner's method.


The above method does of course depend on the formula:

 
     N = (a + b + c + d + ...)3
       = a3
         + (3a2 + 3ab + b2)b
         + {[(3a2 + 3ab + b2) + (3ab + 2b2)]
         + [3(a + b)c + c2]}c
         + {[3(a + b)2 + 3(a + b)c + c2]
         + [3(a + b)c + 2c2] + [3(a + b + c)d + d2]}d
         + ...

When extracting (cube) root, the above formula is reserved. At each step, each term (first a3, then (3a2 + (3a + b)b)b, etc.) is removed from the number N.

The above method requires more than one counting board to find the double or triple of a term (while keeping the table). Jia Xian improved the method so as to use only one counting board.

 



Back