XXVI .
A new Method of finding the equal Roots of an Equation , by Division .
By the Rev. John Hellins , Curate of Constantine , in Cornwall ; communicated by Nevil Maskelyne , D. D. F. R. S. and Astronomer Royal .
Read June 20 , 1782 .
THE following theorems are a production of juvenile years .
They were invented about twelve years ago , when algebra was my favourite study ; and one of them ( the first ) was published as a specimen of this method of extracting the equal roots of an equation about ten years ago .
Since that time my avocations have left me but very little leisure for improving any invention of this kind .
These theorems , then , are in their crude state ; however , such as they are , I flatter myself , they will afford an easier solution of equations that have equal roots than is generally known , and be acceptable to the ingenious algebraist .
THEOREM I. If the cubic equation x3 px2 + qx r=0 has two equal roots , each of them will be ( x ) = pq 9r/ 2pp 6q .
DEMONSTRATIO N. Call the three roots a , a , and b ; then , by the composition of equations we shall have x3 2a x2 + aa x aab = 0 , where b+ 2ab 2a +b= p , aa + 2ab = q , and aab = r ; which values being written in our theorem , we have x ( = pq 9r/ 2pp 6q ) = 2aaa + 4aab + aab + 2abb 9aab/ 8aa + 8ab + 2bb 6aa 12ab = 2aaa 4aab + 2abb/ 2aa 4ab + 2bb = a. Q. E. D. EXAMPLE I. If the equation x3 + 5x2 32x + 36 =0 has two equal roots , it is proposed to find them by the above theorem .
Here p= 5 , q= 32 , and r= 36 ; these values being written in the theorem , we have 5x 32 9x 36/ 2 x 25 6x 32 = 160 + 324/ 50 + 192 = 484/ 242 = 2 , which being written for x , the equation becomes 8+ 20 64 + 36 , which is evidently = 0 ; consequently 2 and 2 are roots of it .
Otherwise , 2 , the value of x given by the theorem , being written for it in the quadratic equation 3x2 + 10x 32 = 0 , the result is 12 + 20 32 = 0 .
Or , dividing the given cubic by the quadratic x2~ 2 , we have x2 4x + 4 ) x3 + 5x2 32x + 36 ( x + 9 ; therefore the three roots are 2 , 2 , and 9 .
EXAMPLE II .
Given x3 + 10/ 7x2 4000/ 9261 = 0 , an equation which has equal roots , to find them .
Here q= 0 , and the theorem gives 36000 x 49/ 200 x 9261 = 20/ 21 , which value being written for x the equation vanishes .
THEOREM II .
If the biquadratic equation x4 px3 + qx2 rx +s=0 has two equal roots , make A= 12r 2pq/ 3pp 8q , B= pr 16s/ 3pp 8q , C= 4B 2q/ 4A + 3p , and D= r/ 4A + 3p , and you will have x=D B/ A C. A synthetical demonstration of this theorem would be very long : the INVESTIGATION is as follows .
It has been demonstrated by the writers on algebra , that , if a biquadratic equation , as x4 px3 + qx2 rx +s= 0 , has two equal roots , one of them may be had from the equation 4x3 3px2 + 2qx r= 0 .
Multiply this equation by x , and the original one by 4 , and take the difference of the two , which will be px3 2qx2 + 3rx 4s = 0 .
Again , if this equation be multiplied by 4 , and the other cubic by p , and their difference taken , we shall have 3pp 8q x x2 + 12r 2pq xx+ pr 16s = 0 , or x2 + 12r 2pq/ 3pp 8q x+ pr 16s/ 3pp 8q = o , or x2 + Ax +B= 0 , putting A and B for the known quantities in the second and third terms .
Now multiply this equation by 4x , and take the first cubic from it , and we shall have 4A + 3p x x2 + 4B 2q xx+r= 0 , which being divided by 4A + 3p , and C and D put equal to 4B 2q/ 4A + 3p and r/ 4A + 3p respectively , gives x2 + Cx +D= 0 ; and this equation being taken from the other quadratic , there remains ACxx+BD= 0 ; consequently x=D B/ A C. Q. E. I. COROLLARY I. From the above investigation it appears , that one of the equal roots may also be obtained from either of these two quadratic equations , of which the first seems most eligible , as the co-efficients of it are less complex than those of the other : 3pp 8q x x2 + 12r 2pq xx+ pr 16s = 0 , and 4A + 3p x x2 + 4B 2q xx+r= 0 .
And these , when p= 0 , become 8qx2 + 12rx 16s = 0 , and 48r/ 8qx2 + 64s/ 8q 2q xx+r= 0 , or x2 3r/ 2qx + 2s/ q = 0 , and x2 + qq 4s/ 3rx q/ 6 = 0 .
COROL .
2 .
If both p and q vanish , then , from either of the quadratics we get x= 4s/ 3r , perfectly agreeing with the cubic px3 2qx2 + 3rx 4s = 0 , which , when p and q vanish , becomes 3rx 4s = 0 .
And this equation is of use ; because , in this case , the theorem fails , one of the divisors being = 0 .
COROL .
3 .
From the equation 4x3 3px2 + 2qx r= 0 , which , when p and q vanish , becomes 4x3 r= 0 , we also get x=3~ r/ 4 , another expression of the same value of x. COROL .
4 .
When r= 0 , D= 0 , and from the equation x2 + Cx +D= 0 , we have x= C. EXAMPLE I. If the equation x4 * 9x2 + 4x + 12 =0 has equal roots , it is proposed to find them .
Here p= 0 , q= 9 , r= 4 , and s= 12 ; and A becomes = 12 x 4/ 8 x9= -2/ 3 , B. .
= 16 x 12/ 8 x9= 8/ 3 , C. .
=4x -8/ 3 + 18/ 4 x -2/ 3 = 11/ 4 , D. .
= 4/ -8/ 3 = 3/ 2 , and D B/ A C= 3/ 2 + 8/ 3/ -2/ 3 + 11/ 4 = 18 + 32/ -8 + 33 = 50/ 25 = 2 , which being written for x , the equation becomes 16 36 +8 + 12 = 0 ; therefore 2 is one of the roots .
The same value of x may be discovered from either of the quadratic equations mentioned in corollary 1 .
The proper values of the co-efficients being written in the first of them , it becomes x2 2/ 3 x 8/ 3 = 0 , where one value of x is 1+~ 25/ 3 = 2 .
The other quadratic becomes x2 11/ 4 x+ 3/ 2 = 0 , one of whose roots is 11 +~ 25/ 8 = 2 .
EXAMPLE II .
It being known that the equation x4 x3 7x2 + 13x 6=0 has two equal roots , to find them .
Here p= 1 , q= 7 , r= 13 , and s= 6 ; and A= -142/ 59 , B= 83/ 59 , C= -1158/ 391 , D= 767/ 391 , DB= 12800/ 23009 , AC= 12800/ 23069 , and D B/ A C= 12800/ 12800 = 1 , one of the roots sought .
The same value of x may be found from either of the two general quadratic equations given in corollary 1 .
From the first of them we get one value of x= 71 ~144/ 59 = 1 .
And from the other , one value = 579 ~ ~ , which is also = 1 .
EXAMPLE III .
Given the equation x4 1/ 2 x+ 3/ 16 = 0 , in which two values of x are equal to each other , to find them .
By corollary 2 .
we have x=4x 3/ 16 ~3x 1/ 2 = 8/ 16 = 1/ 2 .
By corol .
3 .
x is =3~ 1/ 8 = 1/ 2 .
THEOREM III .
If the sursolid equation x5 px4 + qx3 rx2 + sx t=0 has two roots equal to each other , and you make A= 15r 3pq/ 4pp 10q , B= 2pr 20s/ 4pp 10q , C= 25t ps/ 4pp 10q , D= 5B 3q/ 5A + 4p , E= 5C + 2r/ 5A + 4p , F= s/ 5A + 4p , G=B E/ A D , H=F+ C/ A D , I=B H/ A G , and K= C/ A G , then shall one of the equal values of x be =H K/ I G. The investigation of this theorem being altogether similar to that of the last , it is unnecessary to give it here .
The difference of equations being taken as in the investigation of theorem II .
it will appear , that one of the equal roots may also be had from any one of the following five equations , of which sometimes one , sometimes another , will be the most eligible .
1 .
5x4 4px3 + 3qx2 2rx +s= 0 .
2 .
px4 2qx3 + 3rx2 4sx + 5t = 0 .
3 .
x3 + Ax2 + Bx +C= 0 .
4 .
x3 + Dx2 + Ex F= 0 .
5 .
x2 + Gx +H= 0 .
It is obvious , that , when p vanishes , the work will be considerably shortened ; and when both p and q are wanting , though the above formula fails , yet the equal root may be easily obtained from the equation px4 2qx3 + 3rx2 4sx + st = 0 , which in that case becomes 3rx2 4sx + 5t = 0 .
Whenever s is wanting , F , in the second cubic above , will be = 0 , and consequently x may be found from the quadratic equation x2 + Dx +E= 0 .
But in any of these cases the equal root may be found by division .
However , the operation probably will not , in general , be so short as extracting the root of the quadratic ; I will therefore hasten to give an example or two of the use of the theorem .
EXAMPLE I. Given x5 + x3 x2 + 0.09433 = 0 , to find x , two values of it being equal to each other .
Here p= 0 , q= 1 , r= 1 , s= 0 , t= 0.09433 , and we get A= 1.5 F=0 B=0G= 0.2231 C=+ 0.2358 H= 0.1241 D=+ 0.4 I= 0.0972 E= 0.4238 K= 0.185 and x=H K/ I G= 0.48 .
The proper values of the co-efficients being written in the five equations before mentioned , and some of them divided by the co-efficient of the highest power of x , we have these four equations , in each of which one value of x is one of the equal ones sought : x3 *+ 0.6x 0.4 = 0 .
x3 1.5x2 *+ 0.2358 = 0 .
x2 + 0.4x 0.4238 = 0 .
x2 0.2231x 0.1241 = 0 .
Now the most eligible equation is the quadratic x2 + 0.4x 0.4238 = 0 , whose affirmative root is ~ 0.4638 0.2 = 0.4811 , agreeing with the value of x found above , but true to two places lower in the decimal .
EXAMPLE II .
To find the two equal values of x in the equation 64x3 20x2 +3 =0 .
The given equation being divided by 64 , we have x3 0.3125x2 + 0.046875 = 0 ; and then , from the first of the five equations given above , we get 5x4 0.625x = 0 , and x= 3~ 0.125 = 0.5 .
But from the second of the equations just mentioned , we have 0.9375x2 0.234375 = 0 , or x2 = 0.234375/ 0.9375 = 0.25 , and x=~ 0.25 = 0.5 .
From the foregoing few pages it is evident , that rules may be made for finding the equal roots of equations of more than five dimensions by division ; but the operations by them will , in most cases , be long and tedious .
It is obvious , however , that such equations may be depressed to any dimension the algebraist pleases .
It has indeed been supposed , that the number of equations that have equal roots is but small , and , consequently , that the chief use of the rules for finding their roots is to get limits and approximations to the roots of equations in general .
That use , it must be allowed , were it the only one , is sufficient to pay for investigating them .
But if the equations that have equal roots should hereafter be found not so few as has been generally received , then the use of the above theorems will become more extensive .
I beg leave to add , that this short essay is but a small part of a work , in which , if I should ever have leisure to put a finishing hand to it , something more on this subject may very probably appear .
In the mean while , I hope , this little piece will be candidly received by those who have more leisure and better abilities for studies of this kind .
Constantine , February 9 , 1782 .

