XLII .
A Method of extending Cardan 's Rule for resolving one Case of a Cubick Equation of this Form , x3 * qx = r , to the other Case of the same Equation , which it is not naturally fitted to solve , and which is therefore often called the irreducible Case .
By Francis Maseres , Esq. F. R. S. Cursitor Baron of the Exchequer .
Read July 9 , 1778 .
ARTICLE I. IT is well known to all persons conversant with algebra , that CARDAN'S rule for resolving the cubick equation x3-qx=r is only sitted to resolve it when rr/ 4 is equal to , or greater than , q3/ 27 , or when r is equal to , or greater than , 2q ~ q/ 3 ~ 3 , and that it is of no use in the resolution of the other case of this equation , in which r is of any magnitude less than 2q ~ q/ 3 ~ 3 .
For in this case rr/ 4 q3/ 27 becomes ( according to the usual language of algebraists ) a negative quantity , and consequently its square-root be1 comes comes impossible , and the expression given by CARDAN'S rule for the value of x ( which is either ~3~r/ 2 + ~~rr/ 4 q3 27 + q/ 3~3~r/ 2 + ~~rr/ 4 q3/ 27 or ~3~r/ 2 + ~~rr/ 4 q3/ 27 + ~3~r/ 2 ~~rr/ 4 q3/ 27 ) , involves in it the impossible quantity ~rr/ 4 q3/ 27 , and therefore is unintelligible and useless : or , according to what appears to me a more correct way of speaking ( who never could form any idea of a negative quantity , and never understand by the sign any thing more than the subtraction of a lesser quantity from a greater ) , the quantity rr/ 4 q3/ 27 becomes itself impossible , or the supposition that rr/ 4 is greater than q3/ 27 , ( which is one of the foundations of CARDAN'S rule ) , is no longer true , and consequently the rule itself , which is built upon it , can no longer take place .
2 .
Nevertheless it is possible , by the help of Sir ISAAC NEWTON'S binomial theorem , to extend this rule to this latter case , in which rr/ 4 is less than q3/ 27 , and which it is not of itself fitted to resolve ; or , to speak with more accuracy , it is possible to derive from the expression of the value of x given by CARDAN'S rule for the resolution of the equation x3 qx =r in the first case , in which rr/ 4 is greater than q3/ 27 , another expression somewhat different from the former , that shall exhibit the true value of x in the second case , in which rr/ 4 is less than q3/ 27 , provided it be not less than q3/ 2 X 27 , or q3/ 54 : and this without any mention of either impossible , or negative , quantities .
To show how this may be effected , is the design of the following pages .
3 .
That the whole of this matter may be seen at one view , it will be convenient to set forth the foundation and investigation of CARDAN'S rule for resolving the equation x3-qx=r , when rr/ 4 is greater than q3/ 27 ; which may be done as follows .
The Investigation of Cardan 's Rule for resolving the Cubick Equation x3* qx=r , when rr/ 4 is greater than q3/ 27 .
4 .
Previously to the investigation of this rule , it will be proper to make the following observations .
OBS .
1 .
In the equation x3-qx=r ( which is a proposition affirming that x3 is greater than qx , and that the excess is equal to r ) xx must always be greater than q , and x than ~ q. OBS .
2 .
While x increases from ~q ad infinitum , x3 will increase continually from q~ q ad infinitum , and qx will increase continually from the same quantity q~ q ad infinitum .
OBS .
3 .
Also , while x increases from ~q ad infinitum , the excess of x3 above qx will increase continually from nothing ad infinitum , without ever decreasing .
For , if we put x to denote the increment which x receives in any given time , either small or great , qx will be the increment which qx will receive in the same time , and 3x2x + 3xx2 + x2 will be the increment of x3 in the same time .
Now , since xx is always greater than q during the whole increase of x from being equal to ~q ad infinitum , xx Xx will be greater than qx during that whole increase .
Therefore , a fortiori , 3x2x + 3xx2 + x3 ( which is more than triple of xx X x ) will be greater than qx ; that is , the increment of x3 will be greater than the contemporary increment of qx during all the increase of x. Consequently the excess of x3 above qx , or the compound quantity x3-qx , will continually increase , without ever decreasing , while x increases from ~q to any greater magnitude .
OBS .
4 .
Since the compound quantity x3-qx increases continually at the same time as x increases ; and , when x is equal to 2~q/ ~3 , x3-qx ( = 8q~q/ 3~3 2q~q/ ~3 = 8q~q/ 3~3 6q~q/ 3~3 ) = 2q~q/ 3~3 , it follows that , if x is greater than 2~q/ ~3 , the compound quantity x3-qx will be greater than 2q~q/ 3~3 , and , if x is less than 2~q/ ~3 , the said compound quantity will be less than 2q~q/ 3~3 ; and , e converso , if the compound quantity x3-qx , or , its equal , the absolute term r , is greater than 2q~q/ 3~3 , the value of x will be greater than 2~q/ ~3 ; and , if x3 qx , or r , is less than 2q~q/ 3~3 , the value of x will be less than 2~q/ ~3 ; or , if rr/ 4 is greater than q3/ 27 , x will be greater than 2~q/ ~3 , and , if rr/ 4 is less than q3/ 27 , x will be less than 2~q/ ~3 .
OBS .
5 .
When r is greater than 2q~q/ 3~3 , or rr/ 4 is greater than q3/ 27 , and consequently ( by the last observation ) x is greater than 2~q/ ~3 , xx will be greater than 4q/ 3 , and xx/ 4 will be greater than q/ 3 .
But xx/ 4 is the square of x/ 2 .
Therefore when r is greater than 2q~4/ 3~3 , or rr/ 4 is greater than 3/ 27 , the square of half x will be greater than q/ 3 .
But ( by EUCLID'S Elements , Book II .
Prop. V. ) it is always possible to divide a line , as x , into two unequal parts in such a proportion that the rectangle under its parts shall be equal to any quantity that is less than the square of its half .
Therefore , when r is greater than 2q~q/ 3~3 , or rr/ 4 is greater than q3/ 27 , it is possible to divide the line , or root , x into two unequal parts of such magnitudes that their rectangle , or product , shall be equal to q/ 3 .
This observation is the foundation of CARDAN'S rule for the resolution of the equation x3-qx=r in the first case of that equation , or when r is greater than 2q~q/ 3~3 , or rr/ 4 is greater than q3/ 27 ; the investigation of which is as follows .
PROBLE M. 5 .
To resolve the Equation x3 * qx=r , when r is greater than 2q~q/ 3~3 , or rr/ 4 is greater than q3/ 27 .
SOLUTIO N. Since r is supposed to be greater than 2q~q/ 3~3 , and consequently ( by Obs. 5 .
) xx/ 4 is greater than q/ 3 , it is possible for x to be divided into two unequal parts of such magnitudes that their rectangle , or product , shall be equal to q/ 3 .
Let it be conceived to be so divided ; and let the greater of the two parts be called a , and the lesser b. Then will ab be = q/ 3 , and consequently 3ab will be = q , and 3ab xa+b will be =qxa+ b. Now , since a+b is equal to x , we shall have a3 + 3aab + 3abb + b3 = x3 , and qxa+b= qx .
Therefore x3 qx will be = a3 + 3aab + 3abb + b3 qxa+b= a3 + 3ab xa+b+ b3 qxa+ b ; that is ( because 3ab xa+b is =qxa+ b ) x3qx will be = a3 + b3 .
Therefore r ( which is = x3 qx ) will be = a3 + b3 .
But , since 3ab is = q , we shall have b= q/ 3a , and b3 = q3/ 27a3 .
Therefore a3 + b3 is = a3 + q3/ 27a3 , and r ( which is=a3 + b3 ) is = a3 + q3/ 27a3 .
Therefore ra3 is = a6 + q3/ 27 , and ra3 a6 is = q3/ 27 .
But ra3 a6 is the product of the multiplication of r a3 into a3 , which are together equal to r. Therefore ( by El .
II .
5 .
) ra3 a6 must be less than the square of half r , that is , than rr/ 4 , and consequently may be subtracted from it .
Let it , and its equal q3/ 27 , be so subtracted .
And we shall have rr/ 4 ra3 + a6 = rr/ 4 q3/ 27 .
Therefore the squareroot of rr/ 4 ra3 + a6 will be equal to ~ rr/ 4 q3/ 27 .
But the square-root of rr/ 4 ra3 + a6 is the difference of r/ 2 and a3 , that is , either r/ 2 a3 or a3 r/ 2 , according as r/ 2 or a3 is the greater quantity .
But it has appeared above that a3 and b3 together are equal to r ; and a is supposed to be greater than b , and consequently a3 is greater than b3 .
Therefore a3 must be greater , and b3 less , than r/ 2 .
Therefore a3 r/ 2 is the difference of a3 and r/ 2 , and consequently is the square-root of the quantity rr/ 4 ra3 + a6 .
Therefore a3 r/ 2 is =~ rr/ 4 q3/ 27 , and a3 is = r/ 2 +~ rr/ 4 q3/ 27 .
Consequently a is = ~3 r/ 2 +~ rr/ 4 q3/ 27 .
But b has been shown to be=q/ 3a .
Therefore b is= q/ 3~3 r/ 2 +~ rr/ 4 q3/ 27 ; and consequently a+ b , or x , is = ~3 r/ 2 +~ rr/ 4 q3/ 27 + q/ 3~3 r/ 2 +~ rr/ 4 q3/ 27 .
Q. E. I. 6 .
This expression may be rendered more simple by substituting the single letter s in it instead of ~ rr/ 4 q3/ 27 .
For then it will be ~3 r/ 2 +S+ q/ 3~3 r/ 2 + S. Synthetick Demonstration of the Truth of the foregoing Solution .
7 .
That this expression is equal to x in the equation x3-qx =r will appear by substituting it instead of x in the compound quantity x3-qx , which will thereby be seen to be equal to r , as it ought to be .
This may be done in the manner following .
Since x is = ~3 ~r/ 2 +S+ q/ 3~3 ~r/ 2 + S , or ~r/ 2 +S 1/ 3 + q/ 3 x ~r/ 2 +S 1/ 3 , we shall have x3 = r/ 2 +S+3X ~r/ 2 +S 2/ 3 x q/ 3 x ~r/ 2 +S 1/ 3 +3x ~r/ 2 +S 1/ 3 x qq/ 9 x ~r/ 2 +S 2/ 3 + q3/ 27 x ~r/ 2 +S= r/ 2 +S+qx ~r/ 2 +S 1/ 3 + qq/ 3 x ~r/ 2 +S 1/ 3 + q3/ 27r/ 2 + 27S ; and qx = qx ~r/ 2 +S 1/ 3 + qq/ 3 x ~r/ 2 +S 1/ 3 ; and consequently x3 qx = r/ 2 +S+ q3/ 27r/ 2 + 27S = r/ 2 +S+ q3/ 27r/ 2 + 54s/ 2 = r/ 2 +S+ q3/ 27r + 54s/ 2 = r/ 2 +S+ 2q3/ 27r + 54s .
Now ss , or rr/ 4 q3/ 27 , is 27rr 4q3/ 108 = 27rr 4q3/ 36 x 3 ; or , if we put mm = 27rr 4q3 , we shall have ss , or rr/ 4 q3/ 27 , = mm/ 36 x 3 , and S= m/ 6 ~3 .
Therefore 2q3/ 27r + 54s is = 2q3/ 27r + 54 x m/ 6 ~3 = 2q3/ 6 x 27 x ~3 xr+ 54m/ 6 ~3 = 12 ~3 x q3/ 6 x 27 ~3 xr+ 54m=2 ~3 x q3/ 27 ~3 xr+ 9m .
Therefore S+ 2q3/ 27r + 54s is=m/ 6~3 + 2 ~3 x q3/ 27 ~3 xr+ 9m = 27 ~3 x rm + 9mm + 36q3/ 6 x 27 x 3r +6x9x ~3 xm=3 ~3 x rm + mm + 4q3/ 54r +6 ~3 xm= 3 ~3 x rm + 27rr 4q3 + 4q3/ 54r +6 ~3 xm=3 ~3 x rm + 27rr/ 54r +6 ~3 xm= ~3 x rm + 9rr/ 18r +2 ~3 x m ; and r/ 2 + S+ 2q3/ 27r + 54s is = r/ 2 + ~3 x rm + 9rr/ 18r +2 ~3 xm= 18rr +2 ~3 x rm +2 ~3 x rm + 18rr/ 36r + 4~3 xm= 36rr +4 ~3 x rm/ 36r + 4~3 xm= r. But we have before shown that x3 q3 is = r/ 2 +S+ 2q3/ 27r + 54s .
Therefore x3 qx is = r , and consequently ~3 ~r/ 2 +S+ q/ 3~3 ~r/ 2 +S is the true value of x in the cubick equation x3 qx = r. Q. E. D. Two other Expressions for the Root of the foregoing Equation .
8 .
Two other expressions may be found for the root of this equation by resuming the investigation contained in Art .
5 .
The first of these expressions is ~3 ~r/ 2 ~ ~rr/ 4 q3/ 27 + q/ 3~3 ~r/ 2 ~ ~rr/ 4 q3/ 27 , or ( if we put ss , as before , = rr/ 4 q3/ 27 , ) ~3 ~r/ 2 S+ q/ 3~3 ~r/ 2 S. The other expression is ~3 ~r/ 2 +~ ~rr/ 4 q3/ 27 + ~3 ~r/ 2 ~ ~rr/ 4 q3/ 27 , or ~3 ~r/ 2 +S+ ~3 ~r/ 2 S. These expressions are to be found in the following manner .
Investigation of the said Expressions .
9 .
In Art .
5 .
we supposed the line x to be divided into to two unequal parts a and b , of which a was supposed to be the greater ; and we first found the value of the greater part a , and then determined that of the lesser part b from its relation to a , which is expressed by the equation 3ab = q. But we may with the same ease first determine the value of the lesser part b , and then derive from it that of the greater part a ; which would produce the first of the two expressions of the value of x mentioned in the last article .
This may be done as follows .
Since it has been shown in Art .
5 .
that r is = a3 + b3 , and 3ab is = q , and consequently a is = q/ 3b , and a3 to q3/ 27b , it follows that r will be = q3/ 27b3 + b3 .
Therefore rb3 is = q3/ 27 + b6 , and ( subtracting b6 from both sides ) rb3 b6 is = q3/ 27 .
Therefore ( subtracting both sides from rr/ 4 , than which they are evidently less ) , we shall have rr/ 4 rb3 + b6 = rr/ 4 q3/ 27 .
Therefore the square-root of rr/ 4 rb3 + b6 will be =~ ~rr~4 q3/ 27 .
But the square-root of rr/ 4 rb3 + b6 is the difference of the quantities r/ 2 and b3 , that is ( because b3 is the lesser part of a3 + b3 , or r , and consequently is less than the half of it , or r/ 2 ) , it is = r/ 2 b3 .
Therefore r/ 2 b3 is =~ ~rr/ 4 q3/ 27 , Therefore ( adding b3 to both sides ) r/ 2 will be = b3 + ~rr/ 4-q3/ 27 , and ( subtracting ~rr/ 4-q3/ 27 from both sides ) b3 will be=r/ 2-~rr/ 4-q3/ 27 .
Therefore b is =~3r/ 2-~rr/ 4-q3/ 27 , and a(=q/ 3b ) is q/ 3~3r/ 2-~rr/ 4-q3/ 27 , and consequently b+a , or a+b , or x , is =~3r/ 2-~rr/ 4-q3/ 27 + q/ 3~3r/ 2-~rr/ 4-q3/ 27 , or ( if we put ss=rr/ 4-q3/ 27 ) ~3r/ 2-s+q/ 3~3r/ 2-s Q.E.I. Synthetick demonstration of the truth of the foregoing expression .
10 .
Here again we may demonstrate synthetically , that this expression is equal to the true value of x in the proposed equation x3-qx=r , by substituting it for x in the left-hand side of that equation .
For , if we make that substitution , we shall find that the value of x3-qx thence arising will be equal to r. This may be done in the manner following .
If x is=~3r/ 2-s+q/ 3~3r/ 2-s , or r/ 2-s1/ 3+ q/ 3 x r/ 2-s 1/ 3 , we shall have x3=r/ s-s+3x r/ 2-s 2/ 3 x q/ 3 x r/ 2 s 1/ 3 +3x r/ 2 s 1/ 3 x qq/ 9 x r/ 2 s 2/ 3 + q3/ 27 x r/ 2 s= r/ 2 s+qx r/ 2 s 1/ 3 + qq/ 3 x r/ 2 s 1/ 3 + q3/ 27 x r/ 2 s , and qx =qx r/ 2 s 1/ 3 + qq/ 3 x r/ 2 s 1/ 3 , and consequently x3 qx = r/ 2 s+ q3/ 27 x r/ 2 s= r/ 2 s+ q3/ 27r/ 2 27s = r/ 2 s + q3/ 27r/ 2 54s/ 2 = r/ 2 s+ 2q3/ 27r 54s .
Now ss , or rr/ 4 q3/ 27 is = 27rr 4q3/ 108 = 27rr 4q3/ 36 x 3 .
Therefore if we put mm = 27rr 4q3 , we shall have ss = mm/ 36 x 3 , and s= m/ 6 ~3 .
Therefore 2q3/ 27r 54s is = 2q3/ 27r 54m/ 6 ~3 = 2q3/ 6 x 27 x ~3 xr 54m/ 6 ~3 = 12 ~3 x q3/ 6 x 27 x ~3 xr 54m =2 ~3 x q3/ 27 x ~3 xr 9m .
Therefore r/ 2 s+ 2q3/ 27r 54s is = r/ 2 m/ 6 ~3 +2 ~3 x q3/ 27 x ~3 xr 9m = r/ 2 27 x ~3 x rm + 9mm + 36q3/ 6 x 27 x 3r 54 ~3 xm= r/ 2 3 ~3 x rm + mm + 4q3/ 54r 6 ~3 xm= r/ 2 -3 ~3 x rm + 27rr 4q3 + 4q3/ 54r 6 ~3 xm= r/ 2 -3 ~3 x rm + 27rr/ 54r 6 ~3 xm= 54rr 6 ~3 x rm 6 ~3 x rm + 54rr/ 108r 12 ~3 xm = 108rr 12 ~3 x rm/ 108r 12 ~3 xm= r. But it has been before shown that x3 qx is = r/ 2 s+ 2q3/ 27r 54s .
Therefore x3 qx is = r ; and ~3 r/ 2 s+ q/ 3 ~3 r/ 2 s is the true value of x in the cubick equation x3 qx = r. Q. E. D. Investigation of the third expression of the value of the root x. 11 .
The third expression for the value of x , or the last of the two mentioned in Art .
8 .
to wit , ~3 r/ 2 +~ rr/ 4 q3/ 27 + ~3 r/ 2 ~ rr/ 4 q3/ 27 , or ~3 r/ 2 +s+ ~3 r/ 2 s , may be obtained as follows .
Since a3 + b3 is = r , it follows that b3 will be =r a3 .
But a3 is shown in Art .
5 .
to be = r/ 2 +~ rr/ 4 q3/ 27 .
Therefore r a3 is =r r/ 2 +~ rr/ 4 q3/ 27 = r/ 2 ~ rr/ 4 q3/ 27 .
Consequently b3 is = r/ 2 ~ rr/ 4 q3/ 27 , and b is = ~3 r/ 2 ~ rr/ 4 q3/ 27 , and a+ b , or x , is = ~3 r/ 2 +~ rr/ 4 q3/ 27 + ~3 r/ 2 ~ rr/ 4 q3/ 27 , or ( putting ss , as before , = rr/ 4 q3/ 27 , ) ~3 r/ 2 +s+ ~3 r/ 2 s. Q. E. I. Synthetick demonstration of the truth of the said third expression .
12 .
Here again we may demonstrate synthetically , that this expression is equal to the true value of x in the equation x3 qx = r , by substituting it for x in the lefthand side of the said equation .
For , if we make that substitution , we shall find that the value of x3 qx , thence arising , will be equal to r. This may be done in the following manner .
If x is = ~3 ~r/ 2 +s+ ~3 ~r/ 2 s , or ~r/ 2 + s~1/ 3 + ~r/ 2 s~1/ 3 , we shall have x3 = r/ 2 +s+3x ~r/ 2 + s~2/ 3 x ~r/ 2 s~1/ 3 +3x ~r/ 2 + s~1/ 3 x ~r/ 2 s~2/ 3 + r/ 2 s=r+3x ~r/ 2 + s~2/ 3 x ~r/ 2 s~1/ 3 +3x ~r/ 2 + s~1/ 3 x ~r/ 2 s~2/ 3 =r+ 3x ~r/ 2 + s~1/ 3 x ~r/ 2 + s~1/ 3 x ~r/ 2 s~1/ 3 +3x ~~/ 2 + s~1/ 3 x ~r/ 2 s~1/ 3 x ~r/ 2 s~1/ 3 = r+3x ~r/ 2 + s~1/ 3 x ~rr/ 4 ss~1/ 3 +3x ~rr/ 4 ss~1/ 3 x ~r/ 2 s~1/ 3 =r+3x ~r/ 2 + s~1/ 3 x ~rr/ 4 rr/ 4 + ~/ 27~1/ 3 +3x ~rr/ 4 rr/ 4 + q3/ 27~1/ 3 x ~r/ 2 s~1/ 3 =r+3x ~r/ 2 + s~1/ 3 x ~q3/ 27 ~1/ 3 + 3x ~q3/ 27~1/ 3 x ~r/ 2 s~1/ 3 =r+3x ~r/ 2 + s~1/ 3 x q/ 3 +3x q/ 3 x ~r/ 2 s~1/ 3 =r+qx ~r/ 2 + s~1/ 3 +qx ~r/ 2 s~1/ 3 .
And qx will be =qx ~r/ 2 + s~1/ 3 +qx ~r/ 2 s~1/ 3 .
Therefore x3 qx will be =r+qx ~r/ 2 + s~1/ 3 +qx ~r/ 2 s~1/ 3 qx ~r/ 2 + s~1/ 3 qx ~r/ 2 s~1/ 3 = r ; and consequently ~r/ 2 + s~1/ 3 + ~r/ 2 s~1/ 3 , or ~3 ~r/ 2 +s+ ~3 ~r/ 2 s~ , is the true value of x in the cubick equation x3 qx = r. 13 .
N. B. I do not remember to have seen these substitutions , or synthetical demonstrations of the truth of the expressions given by CARDAN'S rule , in any book of algebra .
14 .
An Example of the Resolution of a Cubick Equation of the aforesaid Form , x3 qx = r , by means of each of the three foregoing Expressions .
I will here insert a single example of a numeral equation of the foregoing form , x3 qx = r , resolved by each of the three expressions above-mentioned , in order to show that they will all three bring out the same number for its root .
Let it therefore be required to find the value of x in the cubick equation x3 3x = 18 .
15 .
In this equation q is = 3 , and r is = 18 .
Therefore ~q is = ~3 , and 2q ~q/ 3~3 is =2x3 ~3/ 3~3 = 2 , which is greatly less than 18 , or r. Therefore this equation comes under the above-mentioned rule , and may be resolved by either of the foregoing expressions .
Resolution of the equation x3 3x = 18 by the first of the said expressions .
16 .
The first of those expressions is ~3 ~r/ 2 +s+ q/ 3~3 ~r/ 2 + s , in which s stands for ~~rr/ 4 q3/ 27 .
Now , since q is = 3 , q/ 3 will be = 3/ 3 = 1 , and consequently q3/ 27 , or the cube of q/ 3 , will also be = 1 .
And , since r is = 18 , we shall have r/ 2 = 9 , and rr/ 4 = 81 , and consequently rr/ 4 q3/ 27 = 81 1= 80 ; that is , ss will be = 80 .
Therefore s is = ~80 = ~16 x ~5 = 4~5 ; and r/ 2 +s is =9+ 4 ~5 = 72 + 32 ~5/ 8 = 27 + 27 ~5 + 45 +5 ~5/ 8 ; and consequently ~3 ~r/ 2 +s is =3+ ~5/ 2 .
Therefore 3x ~3 ~r/ 2 +s is =3x ~3 + ~5/ 2 , and q/ 3~3 ~r/ 2 +s is =3x 2/ 3 x3+ ~5 = 2/ 3 + ~5 ; and ~3 ~r/ 2 +s+ q/ 3~3 ~r/ 2 +s is =3+ ~5/ 2 + 2/ 3 + ~5 =3+ ~5~ x3+ ~5 + 4/ 2 x3+ ~5 = 9+6 ~5 +5+ 4/ 2 x ~3 + ~5 = 18 +6 ~5/ 2 x ~3 + ~5 =6x3+ ~5/ 2 x ~3 + ~5 = 6/ 2 = 3 .
Therefore 3 is the value of x in the equation x3 3x = 18 .
And so we shall find it to be upon trial : for , if x is taken = 3 , we shall have x3 = 27 , and 3x =3x3= 9 , and x3 3x = 27 9= 18 .
And thus we see that the first of the three foregoing expressions , to wit , ~3 ~r/ 2 +s+ q/ 3~3 ~r/ 2 + s , has given us the true value of x in this equation .
Resolution of the same equation by the second and third of the foregoing expressions .
17 .
We are now to resolve the same equation x3 3x = 18 by means of the two other expressions , to wit , ~3~ r/ 2 S+ q/ 3 ~3~ r/ 2 S , and ~3~ r/ 2 +S+~3~ r/ 2 S. Now , since r is = 18 , and s has been shown to be =~ 80 , or 4~ 5 , we shall have r/ 2 S=94~5= 72-32 ~ 5/ 8 = 27 27 ~5+ 45 5~ 5/ 8 , and ~3~ r/ 2 S=3~ 5/ 2 .
Therefore 3~3~ r/ 2 S is =3X3~ 5/ 2 , and q/ 3 ~3~ r/ 2 S is = 3/ 3 x3~ 5/ 2 =3x 2/ 3 x3~5= 2/ 3 ~ 5 .
Consequently ~3~ r/ 2 S+ q/ 3 ~~~ r/ 2 S is = 3~ 5/ 2 + 2/ 3 ~5=3~5x3~5+ 4/ 2 x3~5=96~5+5+ 4/ 2 x3~5= 18 6~ 5/ 2 x3~5=93~ 5/ 3 ~5= 3x3~ 5/ 3 ~5= 3 .
Therefore x is = 3 , as it was found to be by the first expression .
18 .
The third expression ~3~ r/ 2 +S+~3~ r/ 2 S is in the present case =3+~ 5/ 2 +3~ 5/ 2 = 6/ 2 = 3 .
Therefore by this expression , as well as by both the former , the value of x in the equation x3 3x = 18 comes out to be 3 .
19 .
Note .
The foregoing method of resolving the cubick equation x3 qx = r , when r is greater than 2q ~ q/ 3 ~ 3 , or rr/ 4 is greater than q3/ 27 , and a like method of resolving the cubick equation x3 + qx =r ( which holds good in all cases , whatever be the magnitudes of q and r ) , are usually known by the name of CARDAN'S rules , because they were first published by him in his treatise of algebra , entitled , Ars magna , quam vulgo Cossam vocant , seu regulas Algebraicas , in the year 1545 , although , as he himself informs us , they were first found out by one SCIPIO FERREUS of Bononia .
See WALLIS'S algebra , Chap. XIII .
Of the second case of the cubick equation x qx = r ; in which r is less than 2q ~ q/ 3 ~ 3 , or rr/ 4 is less than q3/ 27 , and which cannot be resolved by CARDAN'S rule .
20 .
The remaining case of the cubick equation x3 qx = r , in which r is less than 2q ~ q/ 3 ~ 3 , or rr/ 4 is less than q3/ 27 , and which consequently cannot be resolved by the rules above-mentioned , has , upon that account , obtained amongst algebraists the name of the irreducible case : at least it is often called by the French writers of algebra le cas irreductible .
The object of the remaining pages of this paper is to show how , by the help of Sir ISAAC NEWTON'S famous binomial theorem , the foregoing solution of the other , or first , case of this equation may be , as it were , extended to this latter case , or , rather , may be made the means of discovering , by a very peculiar train of reasoning , another solution , that shall be adapted to it .
21 .
By the binomial theorem it appears that the cuberoot of the binomial quantity a+b ( in which a is supposed to be greater than b ) is equal to the following in1 1111 finite series , to wit , a13 + a3b/ 3a a3b2/ 9a2 + 5a3b3/ 81a3 10a3b4/ 243a4 + 22a3b5/ 729a5 11 154a3b6/ 6561a6 + 2618a3b7/ 137,781a7 &amp;c. or to a1/ 3 x the infinite series 1+ b/ 3a b2/ 9a2 + 5b3/ 81a3 10b4/ 243a4 + 22b5/ 729a5 154b6/ 6561a6 + 2618b7/ 137,781a7 &amp;c. or ( if we put the capital letters A , B , C , D , E , F , G , H , &amp;c. for the several numeral coefficients , 1 , 1/ 3 , 1/ 9 , 5/ 81 , 10/ 243 , 22/ 729 , 154/ 6561 , 26~8/ 137,781 , &amp;c. of the terms of the series , respectively , ) a1/ 3 x the infinite series 1+ 1Ab/ 3a 2Bb2/ 6a2 + 5Cb3/ 9a3 8Db4/ 12a4 + 11Eb5/ 15a5 14Fb6/ 18a6 + 17Gb7/ 21a7 &amp;c. in which series both the numerators and the denominators of the generating fractions , 2/ 6 , 5/ 9 , 8/ 12 , 11/ 15 , 14/ 18 , 17/ 21 , &amp;c. following the second term , increase continually by 3 , so that it will be easy for any one to continue the series to as many terms as he shall think proper .
22 .
In like manner the cube-root of the residual quantity ab is found by the same binomial theorem to be equal to the infinite series a1/ 3 a1/ 3b/ 3a a1/ 3b2/ 9a2 5a1/ 3b3/ 81a3 10a1/ 3b4/ 243a4 22a1/ 3b5/ 729a5 154a1/ 3b6/ 6561a6 2618a1/ 3b7/ 137,781a7 &amp;c. or to a1/ 3 x the infinite series 1 b/ 3a b2/ 9a2 5b3/ 81a3 10b4/ 243a4 22b5/ 729a5 154b6/ 6561a6 2618b7/ 137,781a7 &amp;c. or a1/ 3 x the infinite series 1 1Ab/ 3a 2Bb2/ 6a2 5Cb3/ 9a3 8Db4/ 12a4 11Eb5/ 15a5 14Fb6/ 18a6 17Gb7/ 21a7 &amp;c. in which series the numeral coefficients of the several terms are the same as in the series that expresses the cube-root of a+b , but the terms which involve the odd powers of b ( which in that series are marked with the sign + , or all added to the first term , ) are in this latter series marked with the sign - , and are all to be subtracted from the first term , as well as the terms which involve the even powers of b , which are to be subtracted from the first term in both serieses .
PROBLE M. 23 .
Let it now be required to resolve the first case of the cubick equation x3 qx = r , in which r is greater than 2q~q/ 3~3 , or rr/ 4 is greater than q3/ 27 , by means of an infinite series derived from the expressions given by CARDA N'S rule .
SOLUTIO N. We have seen in Art .
11 .
that , if ss be put = rr/ 4 q3/ 27 , the value of x in this equation will be = ~3~r/ 2 +s+ ~3~r/ 2-s .
For the sake of avoiding fractions , let e be put =r/ 2 .
And we shall have x=~3~e +s +~3~e-s .
But ( by Art .
21 .
) ~3~e +s is = e1/ 3 x the infinite series 1+ s/ 3e ss 9ee + 5s3/ 81e3 10s4/ 243e4 + 22s5/ 729e5 154s6/ 6561e6 + 2618s7/ 137,781e7 &amp; c ; and , by Art .
22 .
~3~e-s is = e1/ 3 x the infinite series 1 s/ 3e ss/ 9ee 5s3/ 81e3 10s4/ 243e4 22s5/ 729e5 154s6/ 6561e6 2618s7/ 137,781e7 &amp;c. Therefore ~3~e+s + ~3~e s is equal to e1/ 3 x the sum of these two serieses , that is , to e1/ 3 x the infinite series 2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp; c ; and consequently the root of the equation x3 qx =r is = e1/ 3 x the infinite series 2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp;c. ad infinitum .
Q. E. F. 24 .
Note .
This series must always converge , because ss , or rr/ 4 q3/ 27 , is always less than rr/ 4 , or ee .
And , when ss is considerably less than ee , or rr/ 4 q3/ 27 is considerably less than rr/ 4 , or rr/ 4 is very little greater than q3/ 27 , the convergency of the terms of this series will be sufficient to make it useful .
But in other cases , when rr/ 4 is much greater than q3/ 27 , ( as when it is triple , quadruple or quintuple of it , or still greater , ) the terms of this series will converge so slowly as to render it very unfit for practice .
And indeed in the most favourable cases it will , as I believe , be less convenient in practice than the expression ~3~e+s +~3~e-s , or ~3~r/ 2+ s +~3~r/ 2s , from which it was derived .
However , that it may appear that this series will exhibit the root of the equation x3-qx=r truly , if we will take the necessary pains of computing it , I will here subjoin one example , and no more , of the resolution of a cubick equation of that form by means of it , having taken care to chuse such numbers for q and r as shall make rr/ 4 be but little greater than q3/ 27 , and consequently shall give us only a small number for the fraction ss/ ee , by the continual multiplication of which the terms of the series are generated .
An example of the resolution of a cubick equation of the aforesaid form , x3-qx=r , in the first case of it , in which r is greater than 2q ~q/ 3 ~3 , or rr/ 4 is greater than q3/ 27 , by means of the expression e1/ 3 x the infinite series 2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp;c. obtained in Art .
23 .
25 .
Let it be required to resolve the equation x3 300x = 2108 by means of the infinite series e1/ 3 x ~2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp;c. obtained in Art .
23 .
by the help of Sir ISAAC NEWTON'S binomial theorem .
Here q is = 300 , and r is = 2108 .
Therefore 2q~q/ 3~3 is =2x 300 x ~300/ 3 x ~3 =2x 100 x ~300/ ~3 =2x 100 x 10~3/ ~3 =2x 100 x 10 = 2000 , which is less than 2108 , or ~ .
Therefore this equation comes under the case of CARDAN'S rule , and consequently may be resolved by means of the infinite series e 1/ 3 x ~2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp;c. if that series has been justly derived from the third expression of the value of x given by CARDAN 's rule .
26 .
Now , since r is = 2108 , r/ 2 , or e , will be = 1054 , and rr/ 4 , or ee , will be = 1,110,916 .
And , since q is = 300 , q/ 3 will be = 100 , and q3/ 27 , or the cube of q/ 3 , will be = 1000,000 ; and consequently ss , or rr/ 4 q3/ 27 , will be ( = 1,110,916 1000,000 ) = 110,916 .
Therefore , the fraction ss/ ee is = 110,916/ 1,110,916 =.0998 .
Therefore s4/ e4 is =.0998~2 =.009,950 , and s6/ e6 is =.0998~3 =.000,992 ; and 2ss/ 9ee is = 2/ 9 x.0998 =.1996/ 9 =.022,177 ; and 20s4/ 24 e4 is = 20/ 243 x.009,950 =.199,000/ 243 =.000,818 ; and 308s6/ 6561e6 is=308/ 6561 x.000,992 =.305,536/ 6561 =.000,046 ; and consequently , 2ss/ 9ee + 20s4/ 243e4 + 308s6/ 6561e6 is =.022,177 +.000,818 , +.000,046 =.023,041 ; and 2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 is=2-.023,041 = 1.976,959 .
But e is = 1054 .
Consequently , e1/ 3 , or ~3e , is = ~31054=10.1768 .
Therefore e1/ 3 x the series 2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp;c. is = 10.1768 x 1.976,959 = 20.119,116 .
Therefore the root of the equation x3 300 x= 2108 is = 20.119,116 .
Q. E. I. 27 .
This value of x is true to five places of figures , the~ more accurate value of it being 20.119,053 , as will easily appear by prosecuting it to three or four more places of figures by Mr. RAPHSON'S method of approximation .
28 .
That 20.119 is very nearly equal to , but somewhat less than , the true value of x in the equation x3 300x = 2108 , will appear by substituting it instead of x in the left-hand side of that equation .
For , if we take x= 20.119 , we shall have xx = 404.774,161 , and x3 = 8143.651,345,159 , and 300x = 6035.700 ; and consequently , x3 300x = 8143.651,345,159 , 6035.700 = 2107.951,345,159 , which is somewhat less than 2108 , or the accurate value of x3 300 x in the proposed equation x3 300x = 2108 .
Therefore , 20.119 must be nearly equal to , but somewhat less than , the accurate value of x in that equation .
29 .
It appears therefore from this example , that this expression , e1/ 3 x the infinite series 2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp;c. does truly exhibit the root of the equation x3 qx =r in that case of it which falls under CARDAN'S rule , or in which r is greater than 2q~q/ 3~3 , or rr/ 4 is greater than q3/ 27 .
30 .
I now proceed to consider the problem which is the principal object of this paper , which is to show how from the series e 1/ 3 x ~2 2ss 9ee 20s4/ 243s4 308s6/ 6561s6 &amp;c. we may derive another series , differing from it only in the signs of some of the terms , by which the equation x3 qx =r may be resolved in that other case of it which does not come under CARDAN'S rule , and in which r is less than 2q~/ 3~3 , or rr/ 4 is less than q3/ 27 : and this without any mention of either impossible or negative quantities .
PROBLE M. To resolve , by means of an infinite series derived from the infinite series e 1/ 3 x ~2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp;c. the second case of the cubick equation x3-qx=r , in which r is less than 2q~q/ 3~3 , or rr/ 4 is less than q3/ 27 .
SOLUTIO N. 31 .
We have seen that in the first case of the eequation x3-qx=r , in which rr/ 4 is greater than q3/ 27 , the product duct of e 1/ 3 into the series 2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp;c. ad infinitum , is equal to the root x. Now there are two different ways of computing this series , which ( though not equally short and convenient in practice ) are nevertheless equally just and true : and therefore they must both produce the same result for the value of the series .
The first way of computing it is the common one , which consists of the following processes ; to wit , first , to compute the quantities rr/ 4 and q3/ 27 , as was done in the foregoing example , art .
26 , where rr/ 4 was found to be = 1,110,916 , and q3/ 27 to be 1000,000 ; 2dly , to ~ q3/ 27 from rr/ 4 , in order to get the quantity ss , which is equal to their difference , and which in the foregoing example was 110,916 ; 3dly , to divide ss by ee , so as to obtain the value of the fraction ss/ ee ; as in the foregoing example we found the fraction 110,916/ 1,110,916 to be =.0998 ; 4thly , to compute the powers of the value found for the fraction ss/ ee ; as in the foregoing example we computed those of .0998 , and found its square to be.009,950 , and its cube to be.000,992 ; 5thly , to multiply ss/ ee , and its powers s4/ e4 , s6/ e6 , &amp;c. into the co-efficients 2/ 9 , 20/ 243 , 308/ 6561 , &amp;c. respectively , as in the foregoing example we multiplied .0998 into 2/ 9 , and.009,950 into 20/ 243 , and.000,992 into 308/ 6561 , and found the products to be.022,117,.000,818 , and.000,046 ; and 6thly , to subtract all the products so obtained from 2 the first term of the series .
This is the common and the proper way of computing the series 2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp;c. when we want to make use of it in practice .
But it may also be computed in another manner , which may be described as as follows .
Instead of ss insert the compound quantity rr/ 4 q3/ 27 itself , to which ss is equal , in all the terms of it .
And it will be thereby converted into the following series , to wit , 2 2/ 9ee x ~r/ 4-q3/ 27 20/ 243e4 x rr/ 4 q~/ 27 2 308/ 6561e6 x rr/ 4 q3/ 27 3 &amp;c. or ( because ee is =rr/ 4 , and consequently e4=r4/ 16 , and e6=r6/ 64 ) 2 8/ 9rr x rr/ 4 q~/ 27 320/ 243r4 x rr/ 4 q3/ 27 2 19712/ 6561r6 x ~r/ 4 q~/ 27 3 &amp;c. or 2 8/ 9rr x rr/ 4 q3/ 27 320/ 243r4 x 4/ 16 2rrq~/ 4 x 27 + q6/ 27 x 27 19712/ ~ x r6/ 64 3~4q3/ 16 x 27 + 3rrq6/ 4 x 27 x 27 q9/ 27 x 27 x 27 &amp;c. or ( putting the Greek letters ~ , ~ , ~ , &amp;c. for the numeral co-efficients 8/ 9 , 320/ 243 , 19712/ 6561 , &amp;c. respectively ) 2 ~/ rr x rr/ 4 q3/ 27 ~/ r4 x r4/ 16 2rrq3/ 4 x 27 + q6/ 27 x 27 ~/ r6 x r6/ 64 3~4q3/ 16 x 27 + 3rrq6/ 4 x 27 x 27 q9/ 27 x 27 x 27 VOL. LXVIII .
5Z &amp;c. &amp;c. or 2 ~/ 4 + ~q3/ 27rr ~/ 16 + 2~q3/ 4 x 27rr ~q6/ 27 x 27 x r4 ~/ 64 + 3~q~/ 16 x 27rr 3~q6/ 4 x 27 x 27r4 + ~q9/ 27 x 27 x 27r~ &amp;c. which consists of a much greater number of terms than the series 2 2ss/ 9ee 20s4/ 243e4 308s6/ ~ &amp;c. from which it is derived , and in which many of the terms are much more complicated than in that former series .
Nevertheless , since the compound quantity rr/ 4 q3/ 27 is equal to ss , the insertion of it instead of ss in the terms of that former series cannot alter its real value , though it will make it much more difficult to compute .
It must therefore be true of the new and complicated series 2 ~/ 4 + ~q3/ 27rr ~/ 16 + 2~q3/ 4 x 27rr ~q6/ 27 x 27~4 ~/ 64 + 3~q3 16 x 27rr 3~q6/ 4 x 27 x 27r4 + ~q9/ 27 x 27 x 27r6 &amp;c. as well as of the former series 2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp;c. that , if it be multiplied into e 1/ 3 , or ~3 e , or ~3r/ 2 , the series thereby produced will be equal to the value of x in the equation x3-qx=r , or that , if the said series be cubed , and also multiplied into q , and from its cube the product of its multiplication into q be subtracted , the remainder will be equal to r , or rather will approximate to the value of r , because , as the quantity substituted for x in the compound quantity x3-qx is only a part of an infinite series that is equal to x , it is impossible that the said remainder ( which is produced by that substitution ) should be accurately equal to the whole value of r. 32 .
By the help of this observation we may from the foregoing series 2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp;c. which , being multiplied into e 1/ 3 , or the cube-root of r/ 2 , expresses the value of x in the equation x3-qx=r , in the first case of that equation , when rr/ 4 is greater than q3/ 27 , deduce another series resembling the former in the composition of its terms , but differing from it in the signs to be prefixed to some of them , that will likewise ( if multiplied into e 1/ 3 , or the cube-root of r/ 2 ) express the value of x in the second case of the same equation , in which rr/ 4 is less than q3/ 27 , and which cannot be resolved by CARDAN'S rules .
This may be done as follows .
33 .
If in this second case of the equation x3-qx=r we subtract rr/ 4 from q3/ 27 , and call the remainder ss ( as we before put ss for the opposite difference rr/ 4 q3/ 27 ) and then ~ the powers of ss , to wit , s4 , s6 , s8 , s10 , and also the correspondent powers of its value q3/ 27 r4/ 4 , to wit , q3/ 27 rr/ 4~2 , q3/ 27 rr/ 4~3 , q3/ 27 rr/ 4~4 , q3/ 27 rr/ 4~5 , &amp;c. the even powers of the difference q3/ 27 rr/ 4 , to wit , q3/ 27 rr/ 4~2 , q~/ 27 rr/ 4~4 , &amp;c. will consist of the very same terms , or the same powers , products , and multiples of the two original quantities rr/ 4 and q3/ 27 , and with the same signs + and prefixed to them , as were before contained in the even powers of the opposite difference rr/ 4 q3/ 27 , when rr/ 4 was greater than q3/ 27 .
Thus , for example , the square of rr/ 4 q3/ 27 in the former case was r4/ 16 2rrq3/ 4 x 27 + q6/ 27 x 27 ; and in the present case the square of q3/ 27 rr/ 4 is q6/ 27 x 27 2q3rr/ 27 x4+ r4/ 16 , which consists of the same terms , and with the same signs prefixed to them , as were contained in the square of rr/ 4 q3/ 27 , and differs from it only in the order in which the extreme terms r4/ 16 and q6/ 27 x 27 are placed .
And the same observation is true concerning all the other even powers of the opposite differences rr/ 4 q3/ 27 and q3/ 27 rr/ 4 .
Also the odd powers of the difference q3/ 27 rr/ 4 , to wit , q3/ 27 rr/ 4 itself , and q~/ 27 rr/ 4~3 , q3/ 27 rr/ 4~5 , &amp;c. will consist of the same terms , or of the same powers , products , and multiples of the two original quantities rr/ 4 and q3/ 27 , as were contained in the same odd powers of the opposite difference rr/ 4 q3/ 27 , when rr/ 4 was greater than q3/ 27 .
But the signs prefixed to the said terms will be contrary to those which were prefixed to them in the former case .
Thus , the cube of rr/ 4 q3/ 27 in the former case was r6/ 64 3r4q3/ 16 x 27 + 3rrq6/ 4 x 27 x 27 q9/ 27 x 27 x 27 ; and the cube of q3/ 27 rr/ 4 in the present case is q9/ 27 x 27 x 27 3q6rr/ 27 x 27 x4+ 3q3r4/ 27 x 16 r6/ 64 , which consists of the same terms as are contained in the cube of rr/ 4 q3/ 27 : but they are placed in a contrary order to that in which they stood in the former case ; and the signs that are prefixed to them are contrary in every term to what they were before .
And the same is true of all the other odd powers of these opposite differences of rr/ 4 and q3/ 27 .
34 .
It follows , therefore , that if ss be put for q3/ 27 rr/ 4 in this latter case of the equation x3-qx=r , in which rr/ 4 is less than q3/ 27 , the even powers of ss , to wit , s4 , s8 , s12 , s16 , &amp;c. will represent , or be equal to , the same powers , products , and multiples of the two original quantities rr/ 4 and q3/ 27 in the present case as they represented in the former case , when rr/ 4 was greater than q3/ 27 , and ss was made to stand for rr/ 4 q3/ 27 ; and the several terms represented by the said even powers of ss will have the same signs + and prefixed to them respectively in this second case as in the first case .
And it likewise follows that the odd powers of ss , to wit , ss , s6 , s10 , s14 , &amp;c. will also represent , or be equal to , the same powers , products , and multiples of the two original quantities rr/ 4 and q3/ 27 , as in the former case ; but the signs + and prefixed to the several terms represented by the said odd powers of ss will be contrary to what they were before .
35 .
If therefore in this second case of the equation x3 qx = r , in which rr/ 4 is less than q3/ 27 , we put ss = q3/ 27 rr/ 4 , the series 2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp;c. will represent , or be equal to , a system of terms , derived from the two original quantities rr/ 4 and q3/ 27 , that will be the very same in point of composition , that is , will be the very same powers , products , and multiples of rr/ 4 and q3/ 27 , as the terms that were represented by it in the former case , in which rr/ 4 was greater than q3/ 27 : but the terms so represented will not all have the same signs + and prefixed to them as they had before ; but those terms in the said system , which are represented by the terms of the series 2 ~/ 9ee 20s4/ 243c4 368s6/ 6561e6 &amp;c. which involve the even powers of ss , to wit , 20s4/ 243e4 , &amp;c. will have the same signs prefixed to them as they had before when ss stood for rr/ 4 q3/ 27 ; and those terms of the said system which are represented by the terms of the said series which involve the odd powers of ss , to wit , 2ss/ 9ee and 308s6/ 6561e~ , &amp;c. will have contrary signs to those they had before .
Consequently , if we change the signs of those terms in the series 2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp;c. which involve the odd powers of ss , to wit , the terms 2ss/ 9ee and 308s6/ 6561e6 &amp;c. the new series thereby produced , to wit , 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. will represent , or be equal to , a system of terms which will not only be the very same in point of composition ( or will be the same powers , products , and multiples of the two original quantities rr/ 4 and q3/ 27 ) , as those which were represented by the series 2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp;c. in the former case , but will also be connected with each other in exactly the same manner by the signs + and - : that is , by Art .
31 .
the said new series will represent , or be equal to , the following system of terms , to wit , 2 ~/ 4 + ~q3/ 27rr ~/ 16 + 2~q3/ 4 x 27rr ~q6/ 27 x 27r4 r/ 64 + 3~q3/ 16 x 27rr 3~q6/ 4 x 27 x 27r4 + ~q9/ 27 x 27 x 27r6 &amp;c. But it has been shown ( in Art .
31 .
) that , if this system of terms be multiplied into 1/ e3 , or the cube-root of r/ 2 , and the series thence produced be cubed , and also multiplied into q , and from its cube the product of its multiplication into q be subtracted , the remainder thereby obtained will be ( nearly ) equal to r. Therefore , if the series 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. ( which represents , or is equal to , the said system of terms , when rr/ 4 is less than q3/ 27 , and ss is made = q3/ 27 rr/ 4 , ) be multiplied by 1/ e3 , or the cube-root of r/ 2 , and the series thence produced be cubed , and also multiplied into q , and from the cube of the said series the product of its multiplication into q be subtracted , it will follow that the remainder thereby obtained will be ( nearly ) equal to r ; that is , the product of the multiplication of 1/ e3 , or the cube-root of r/ 2 , into the infinite series 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. is equal to the value of x in the equation x3 qx =r in the second case of it , when rr/ 4 is less than q3/ 27 .
Q. E. I. 36 .
This series 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. does not always converge , but only when ss is less than ee , or 93/ 27 rr/ 4 is less than rr/ 4 , or 93/ 27 is less than 2rr/ 4 , or rr/ 4 is greater than half q3/ 27 , or than q3/ 54 , though less than q3/ 27 .
And the nearer rr/ 4 approaches to q3/ 27 , the greater will be the swiftness with which this series will converge .
37 .
I will now add a few examples of the resolution of cubick equations of the aforesaid form , x3-qx=r , in the second case of those equations , in which r is less than 2q~q/ 3~3 , or rr/ 4 is less than q3/ 27 , by means of the infinite series e1/ 3 x~2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. found in Art .
35 .
in order to confirm the truth of the reasonings by which that series was obtained .
EXAMPLE I. 38 .
Let it be required to resolve the equation x3 50x = 120 by means of the said infinite series .
Here q is = 50 ; r is = 120 ; r/ 2 or e , is = 60 ; rr/ 4 , or ee , is = 3600 ; q3 is = 125,000 ; and q3/ 27 is = 125,000/ 27 = 4629 .
629,629,629 , &amp;c. which is greater than 3600 , or r/ 2 .
Therefore this equation cannot be resolved by CARDAN'S rule , but may by the expression e1/ 3 x the series 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. provided that series converges .
Now , since q3/ 27 is = 4629 .
629 , 629 , 629 , &amp;c. and rr/ 4 is = 3600 , we shall have ss = q3/ 27 rr/ 4 = 4629 .
629 , 629 , 629 , &amp;c. 3600 = 1029 .
629 , 629 , 629 , &amp;c. which is considerably less than 3600 , or ee ; and consequently the series will converge .
39 .
We shall therefore have ss/ ee = 1029.629,629,629/ 3600 &amp;c. =.286,00 ; and s4/ e4 =.081,796 ; and s6/ e6 =.023,393 ; and consequently 2ss/ 9ee =2 x.286/ 9 =.572/ 9 =.063 , 55 ; and 20s4/ 243e4 = 20 x.081,796/ 243 = 1.635,92/ 243 =.006,73 ; and 308s6/ 6561e6 = 308 x.023,393/ 6561 = 7.205,044/ 6561 =.001,098 .
Therefore 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 is = 2+.063,55-.006,73+.001,09=2.064,64-.006,73 = 2.057,91 .
And e1/ 3 , or ~3e , is = ~3~60 = 3.914,867 .
Therefore e1/ 3 x the series 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. is = 3.914 , 867 x 2.057 , 91 = 8.0564 ; that is , the root of the proposed equation x3-50x=120 is 8.0564 ; which is true in three places of figures , the error being in the fourth place of figures , or third place of decimal fractions , where the figure ought to be a5 instead of a 6 , the more accurate value of x in that equation being 8.055,810,345,702 , as may easily be found by Mr. RAPHSON'S method of approximation .
But 8.0564 , the value of x found by the foregoing process , is sufficiently near to its more accurate value 8.055,810 , &amp;c. to show the truth of the foregoing reasonings .
Their difference is only 6/ 10,000 parts of an unit , which is only the 13426th part of 8.055,810 , &amp;c. or the true value of x. 40 .
N. B. This equation x3-50x=120 expresses the relation between the diameter of a circle and three chords in it that lie contiguous to each other , and together take up a semicircle , and form a trapezium of which the diameter of the circle is the fourth side .
For if the three chords are called b , k and t , and the diameter of the circle is called x , the relation between them will be ex-bb pressed by the cubick equation x3 kk xx= 2bkt , which , -tt if the numbers 3 , 4 and 5 are substituted instead of the letters b , k , and t , will become x3-50x=120 .
See Sir ISAAC NEWTON'S Arithmetica Universalis , Edit .
2d .
1722 , page 101 .
EXAMPLE II .
41 .
Let it be required to find by means of the same series the root of the equation x3-x=1/ 3 .
Now in this equation q is = 1 , r is =1/ 3 , r/ 2 is = 1/ 6 , rr/ 4 , or ee , is = 1/ 36 , and q3/ 27 is = 1/ 27 , which is greater than 1/ 36 or rr/ 4 .
Therefore this equation cannot be resolved by CARDA N'S rule , but may by the series e1/ 3 x 2++ 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp;c. in case that series is a converging one .
Now , since q3/ 27 is = 1/ 27 , and rr/ 4 is = 1/ 36 , we shall have ss , or q3/ 27 rr/ 4 , = 1/ 27 1/ 36 = 36 27/ 27 x 36 = 9/ 27 x 36 = 1/ 3 x 36 , which is less than 1/ 36 , or ee , in the proportion of 1 to 3 .
Consequently the series 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. and the series e1/ 3 x2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561s6 &amp;c. will converge .
Therefore the equation x3 x= 1/ 3 may be resolved by the means of it as follows .
42 .
Since ss is = 1/ 3 x 36 , and rr/ 4 , or ee , is = 1/ 36 , we shall have ss/ ee = 1/ 3 =.333,333 , and s4/ e4 = 1/ 9 =.111,111 , and s6/ e6 = 1/ 27 =.037,037 , and consequently 2ss/ 9ee =2 x.333333/ 9 =.666666/ 9 = .074,074 , and 20s4/ 243e4 = 20 x.111,111/ 243 = 2.222,222/ 243 =.009,144 , and 308s6/ 6561e6 = 308 x.037,037/ 6561 = 11.407,396/ 6561 =.001,738 .
Therefore 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. is =2 +.074,074 -.009,144 + .001,738 = 2.075,812 -.009,144 = 2.066,668 .
And ~3e is = ~3~1/ 6 = 1/ ~36 = 1/ 1.817,121 .
Therefore e1/ 3 x the series 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. is = 1/ 1.817,121 x 2.066,668 = 1.13733 ; that is , the root of the proposed equation x3-x=1/ 3 is 1.137,33 ; which is true to four places of figures , the error being in the fifth place of figures , or the fourth place of decimal fractions , where the figure ought to be an unit instead of a 3 , the more accurate value of x being 1.137,158,164 , which differs from the value of it here found by less than.00017 , or 17/ 100,000th parts of an unit , which is less than the 6689th part of 1.137,158,164 , or the true value of x. EXAMPLE III .
43 .
Let it be required to find the root of the equation x3 5x = 4 .
Here q is = 5 ; r is = 4 ; r/ 2 , or e , is = 2 ; rr/ 4 , or ee , is = 4 ; q3 is = 125 , and q3/ 27 is = 125/ 27 = 4.629,629,629 , &amp;c. which is greater than 4 , or rr/ 4 .
Therefore this equation cannot be resolved by CARDA N'S rule , but may by the infinite series e1/ 3 x~2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. in case that series is a converging one .
Now , since q3/ 27 is 4.629,629,629 , &amp;c. and rr/ 4 is = 4 , we shall have q3/ 27 rr/ 4 , or ss , =.629,629,629 , &amp;c. which is less than 4 , or ee , in the proportion of about 6 to 40 , which is a pretty large proportion of minority , and much larger than the proportion of ss to ee in either of the former examples .
Consequently the series e1/ 3 x ~2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. will converge with a greater degree of swiftness than in either of those examples .
Therefore the equation x3 5x =r may be resolved by it as follows .
44 .
Here ss/ ee is =.629,629 , &amp;c. / 4 =.157,407 ; and consequently s4/ e4 is =.024,777 , and s6/ e6 is =.003,900 .
Therefore 2ss/ 9ee is =2 x.157,407/ 9 =.314,814/ 9 =.034,979 , and 20s4/ 243e4 is = 20 x.024,777/ 243 =.495,540/ 243 = 002,039 , and 308s6/ 6561e6 is = 308 x.003,900/ 6561 = 1.201,200/ 6561 =.000,182 , and consequently 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 is =2 +.034,979-.002,039+.000,182=2.035,161.002,039 = 2.033,122 .
And e1/ 3 , or ~3e , is = ~32 = 1.259,921 .
Therefore e1/ 3 x the series 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. is = 1.259,921 x 2.033,122 = 2.561,573 ; that is , the root of the proposed equation x3-5x=4 is 2.561,573 ; which is true to five places of figures , the error being in the sixth place of figures , or the fifth place of decimal fractions , where the figure ought to be a5 instead of a 7 .
For the accurate value of x in this equation is 1+~17/ 2 , or 1+4.123,106/ 2 , or 5.123,106/ 2 , or 2.561,553 ; which differs from 2.561,573 , or the value of x found by the foregoing series , by only 20/ 1000,000th , or 2/ 100,000th , parts of an unit , or less than the 128,000th part of 2.561,553 , or the value of x itself ; which is a great degree of exactness .
45 .
Note .
That x , or the root of the equation x3 5x=r , is accurately equal to 1+~17/ 2 , will appear by substituting 1+~17/ 2 instead of x in the compound quantity x3-5x , and observing that it will make that quantity become equal to 4 .
For , if x is = 1+~17/ 2 , we shall have x3 =1+3x~ 17 +3x 17 + 17 x~ 17/ 8 = 52 + 20 x~ 17/ 8 = 13 +5~ 17/ 2 , and 5x =5+5~ 17/ 2 , and consequently x3 5x = 13 +5~ 17/ 2 55~ 17/ 2 = 8/ 2 = 4 .
Therefore 1+~ 17/ 2 is = x. Q. E. D. 46 .
These examples sufficiently prove that the expression e1/ 3 x the series 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. ( which we derived from the other series e1/ 3 x~2 2ss/ 9ee 20s4/ 243e4 308s6/ 6561e6 &amp;c. by the peculiar train of reasoning used in Art .
33 , 34 , and 35 , ) gives the true root of the cubick equation x3 qx = r qx =r in the second case of it , in which r is less than 2q ~q/ 3 ~3 , or rr/ 4 is less than q3/ 27 , and which therefore cannot be resolved by CARDAN'S rule .
I will , however , subjoin one more example to the same purpose ; which shall be that of the equation x3 63x = 162 , which both Dr. WALLIS and Mr. DE MOIVRE have resolved by extracting what they call the impossible cube-roots of the impossible binomial quantities 81 +~ 2700 and 81 ~ 2700 .
Now this equation may be resolved by the foregoing expression e1/ 3 x the series 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. in the manner following .
EXAMPLE 4 .
47 .
Let it be required to find the root of the equation x3 63x = 162 .
Here q is = 63 ; r is = 162 ; r/ 2 , or e , is = 81 ; rr/ 4 , or ee , is = 6561 ; q/ 3 is = 21 ; and q3/ 27 is = 9261 , which is greater than 6561 , or rr/ 4 .
Therefore this equation cannot be resolved by CARDAN'S rule , but may by the infinite series e1/ 3 x~2+ 2ss/ 9ee 24s4/ 243e4 + 308s6/ 6561e6 &amp;c. in case that series is a converging one .
Now , since q3/ 27 is = 9261 , and rr/ 4 is = 6561 , we shall have q3/ 27 rr/ 4 , or ss , = 2700 , which is less than 6561 , or ee , in the proportion of 100 to 243 .
Consequently the series 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6461e6 &amp;c. and the product of that series multiplied into e1/ 3 , or the series e1/ 3 x 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. will converge .
Therefore the equation x3 63x = 162 may be resolved by it as follows .
48 .
Since ss is = 2700 , and rr/ 4 , or ee , is = 6561 , we shall have ss/ ee = 2700/ 6561 = 100/ 243 =.411,522 , and s4/ e4 =.169,350 , and s6/ e6 =.069,691 , and consequently 2ss/ 9ee =2 x.411,522/ 9 =.823,044/ 9 = .091,449 , and 20s4/ 243e4 = 20 x.169,350/ 243 = 3.387,0/ 243 =.013,938 , and 308s6/ 6561e6 = 308 x.069,691/ 6561 = 21.464,828/ 6561 =.003,271 .
Therefore 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. is =2 +.091,449 , -.013,938 , + .003,271 &amp;c. = 2.094,720 , -.013,938 &amp;c. = 2.080,782 , &amp;c. And e1/ 3 , or ~3~ e , is =~3~ 81 = 4.326,749 .
Therefore e1/ 3 x the series 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. is = 4.326,749 x 2.080,782 , &amp;c. = 9.003,021 &amp; c ; that is , the root of the proposed equation x3 63x = 162 is = 9.003,021 , &amp;c. or somewhat less than 9.003,021 ; which is true to three places of figures , the error being in the fourth place of figures , or the third place of decimal fractions , where there ought to be a cipher instead of a 3 , because the accurate value of x in this equation is 9 , as will appear upon trial : for , if x be taken = 9 , we shall have x3= 729 , and 63x=567 , and consequently x3-63x ( =729 -567)=162 .
SCHOLIU M. 49 .
This resolution of the equation x3-63x=162 answers to Dr. WALLIS'S resolution of it by extracting the cube-roots of the impossible binomial quantities 81+~-2700 and 81-~-2700 , inasmuch as both resolutions are originally derived from CARDAN'S rule .
But the difference between them is , that the method here delivered is intelligible in every step of it , whereas Dr. WALLIS'S method treats of impossible quantities , or quantities of which no clear idea can be formed , in the whole course of the process , though it concludes with a result that is intelligible , by means of the equality of the impossible members of the two ultimate quantities 9/ 2+1/ 2~-3 and 9/ 2-1/ 2~-3 ( whose sum is equal to the value of x ) , and the contrariety of the signs + and - , which are prefixed to them .
The doctor 's method of finding 9/ 2+1/ 2~-3 and 9/ 2-1/ 2~-3 to be the cube-roots of the impossible binomial quantities 81+~-2700 and 81-~-2700 is only tentative .
But Mr. DE MOIVRE has given a certain method of finding the cube-roots of such quantities in all cases ; but not without the trisection of an angle , or finding ( by the help of a table of sines , or otherwise ) the cosine of the third part of a circular arc whose cosine is given ; by means of which trisection it is well known ( independently of CARDAN'S rule , or Mr. DE MOIVRE'S process ) that the second case of the cubick equation x3-qx=r ( in which rr/ 4 is less than q3/ 27 , may be resolved .
So that Mr. DE MOIVRE'S method of doing this business , though more perfect than Dr. WALLIS'S , does not seem to be of much use in the resolution of these equations .
And both methods are equally liable to the objection above-mentioned , of exhibiting to our eyes , during the whole course of the processes , a parcel of algebraick quantities , of which our understandings cannot form any idea ; though , by means of the ultimate exclusion of those quantities , the results become intelligible and true .
It is by the introduction of such needless difficulties and mysteries into algebra ( which , for the most part , take their rise from the supposition of the existence of negative quantities , or quantities less than nothing , or of the possibility of subtracting a greater quantity from a lesser ) , that the otherwise clear and elegant science of algebra has been clouded and obscured , and rendered disgusting to numbers of men of a just taste for reasoning ; who are apt to complain of it , and despise it , on that account .
And , doubtless , they have too much reason to do so , and to say , in the words of the famous Monsieur DES CARTES in his dissertation De Methodo , page II , Algebram vero , ut solet doceri , animadverti certis regulis et numerandi formulis ita esse contentam , ut videatur potius ars quaedam confusa , cujus usu ingenium quodammodo turbatur et obscuratur , quam scientia , qua excolatur et perspicacius reddatur .
If this complaint was just in DES CARTES'S time , there is certainly much more reason for it now .
50 .
The passage above alluded to in Dr. WALLIS'S algebra , is in the 48th chapter , pages 179 , 180 , of the folio edition at London in 1685 .
And Mr. DE MOIVRE'S method of extracting the cube-root of an impossible binomial quantity , as 81 +~ 2700 , or a+ ~ b , is published in the appendix to the second volume of professor SAUNDERSON'S algebra , pages 744 , 745 , 746 , 747 .
It is very ingenious , and shows that author 's great skill in the use and management of algebraick gebraick quantities .
See also on this subject CLAIRAUT'S Elemens d'Algebre , Part V. Section 9 .
pages 286 , 287 , 288 , and a paper of Monsieur NICOLE in the memoirs of the French Academy of Sciences for the year 1738 , pages 99 and 100 .
See also MACLAURIN'S algebra , Part I. the supplement to the 14th Chapter , pages 127 , 128 , 129 , 130 ; and the Philosophical Transactions , No. 451 .
51 .
If any gentleman should be inclined to compute the series 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 &amp;c. to more than four terms , he will find the first eight terms of it to be as follows , to wit , 2+ 2ss/ 9ee 20s4/ 243e4 + 308s6/ 6561e6 1870s8/ 59049e8 + 111,826s10/ 4,782,969e10 2,358,512s12/ 129,140,163e12 + 120,646,960s14/ 8,135,830,269e14 .

