Exercise 13.2.17

Following Example 13.2.14, consider the equations x 3 + 3 x + 1 = 0 , and y = a + bx + x 2 .

(a)
Use Maple or Mathematica and Section 2.3 to eliminate x and obtain (13.26).
(b)
Show that coefficients of y 2 and y in (13.26) both vanish if and only if a = 2 and b 2 + b 1 = 0 .
(c)
The equation for y becomes trivial to solve when a = 2 and b = ( 5 1 ) 2 . We could then solve for x using y = a + bx + x 2 , but there is a better way to proceed. Note that x 3 = b x 2 ax + yx

follows from y = a + bx + x 2 . Furthermore, we can use y = a + bx + x 2 to eliminate the x 2 in the above equation. Then use x 3 + 3 x + 1 = 0 to obtain an equation in which x appears only to the first power. Solving this gives a formula for x in terms of y . The general version of this argument can be found in [Lagrange, p.223].

Answers

Proof.

(a)
We eliminate x between the two polynomials f = x 3 + 3 x + 1 , g = x 2 + bx + ( a y ) ,

with the resultant Res x ( f , g ) = det ( S ) , where

S = ( 1 0 1 0 0 0 1 b 1 0 3 0 a y b 1 1 3 0 a y b 0 1 0 0 a y )

We can obtain this determinant with Sage:

     R.<a,b,x,y> = QQ[]
     f = x^3 + 3*x + 1
     g = x^2 + b*x + (a-y)
     S = matrix(R,[[1, 0, 1,   0,   0  ],
                   [0, 1, b,   1,   0  ],
                   [3, 0, a-y, b,   1  ],
                   [1, 3, 0,   a-y, b  ],
                   [0, 1, 0,   0,   a-y]])
     R = S.det(); R

a 3 + 3 a b 2 b 3 3 a 2 y 3 b 2 y + 3 a y 2 y 3 6 a 2 + 3 ab + 12 ay 3 by 6 y 2 + 9 a 3 b 9 y + 1

But it is more easy to call the method "resultant" to obtain the same result:

     res = f.resultant(g,x);res

The list of coefficients of Res x ( f , g ) is given by

     l =[-res.subs(y=0)] +  [-res.coefficient(y^k) for k in range(1,4)]
     l

[ a 3 3 a b 2 + b 3 + 6 a 2 3 ab 9 a + 3 b 1 , 3 a 2 + 3 b 2 12 a + 3 b + 9 , 3 a + 6 , 1 ]

We find the equation(13.26):

y 3 + ( 6 3 a ) y 2 + ( 9 + 3 b + 3 b 2 12 a + 3 a 2 ) y + P ( a , b ) = 0 ,

where P ( a , b ) = a 3 3 a b 2 + b 3 + 6 a 2 3 ab 9 a + 3 b 1 .

(b)
The coefficient of y 2 vanishes if a = 2 , and then the coefficient of y vanishes if 0 = 9 + 3 b + 3 b 2 12 a + 3 a 2 = 3 b + 3 b 2 3 : b 2 + b 1 = 0

so b = 5 1 2 is a solution.

If we pick b = ( 5 1 ) 2 , then the above cubic equation becomes

y 3 + 5 2 5 25 2 = 0 ,

so

y 3 = 25 5 5 2 = 5 3 5 1 2 .

By the property of the resultant, if y is evaluated to y 0 = ω k 5 5 1 2 3 , k = 0 , 1 , 2 , ω = e 2 3 , then there exists a common root of f and g in , where

f = x 3 + 3 x + 1 g = x 2 + 5 1 2 x + 2 y 0 .
(c)
The Euclidean division of f by g gives x 3 + 3 x + 1 = ( x 2 + bx + a y ) ( x b ) + ( y + b 2 + 3 a ) x + 1 + ab by .

If x 0 is a common root of f , g , then the remainder is 0, so

x 0 = b y 0 ab 1 y 0 + b 2 + 3 a ,

and this gives a formula for the roots x 0 of f in terms of the roots y 0 of the resultant.

Since y 0 is an algebraic number of degree 3 over , and x 0 ( 5 , y 0 ) , there exists some polynomial p of degree 2 such that x 0 = p ( y 0 ) .

To find this more simple formula for x 0 , we search the gcd of f , g in the field ( 5 , 5 1 2 3 ) by the extended Euclid’s algorithm.

This is obtained with the following Sage instructions:

     K.<u>= QQ[sqrt(5)]
     R.<z> = K[]
     res = z^3 - (u-1)/2
     L.<w> = K.extension(res)
     A.<x> = L[]
     f = x^3 + 3*x + 1
     g = x^2 + (u - 1)/2 * x + (2 - u*w)
     gcd(f,g)

x + ( 1 2 5 + 1 2 ) w 2 w

We have obtained that

gcd ( f , g ) = x + 5 + 1 2 w 2 w ,  where  w 3 = 5 1 2 .

Since w 3 = 5 1 2 , so w 2 = 5 1 2 w 1 , 5 + 1 2 w 2 = w 1 .

Therefore the roots of f are

x 0 = w 5 + 1 2 w 2 = w w 1

We can write w = ω k 5 1 2 3 , k = 0 , 1 , 2 , then w 1 = ω 2 k 5 + 1 2 3 , where the cubic roots are chosen so that their product is real, equal to 1. Then

ω k 5 1 2 3 ω 2 k 5 + 1 2 3 , k = 0 , 1 , 2

are the roots of f . This is identical to the formulas obtained with Cardano’s formulas, with more sweat.

User profile picture
2022-07-19 00:00
Comments