A new approach to solving the cubic: Cardans solution revealed 1

RWD Nickalls 2 The Mathematical Gazette (1993); 77 (Nov, No 480), 354359 (jstor archive) http://www.nickalls.org/dick/papers/maths/cubic1993.pdf

Introduction

The cubic holds a double fascination since not only is it interesting in its own right, but its solution is also the key to solving quartics 3 . This article describes five fundamental parameters of the cubic (, , , and ), and shows how they lead to a significant modification of the standard method of solving the cubic, generally known as Cardans solution.
Y
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .... . . ............ ... ...... . . .. .. .. .. . . .. . .. .. . .. .. . .. . . .. . .. . . .. . . . . .. . . .. . . . . .. . . .. . . . . .. . .. . . .. . .. .. . . .. . .. .. .. . . .. .. . .. .. . ... . ... .. . .. ......... . ......... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . N . . . . . . . . .

yN

Figure 1: It is necessary to start with a definition. Let ( , ) be a point on a polynomial curve () of degree such that moving the axes by putting
1 This minor revision of the original article corrects typographic errors and incorporates some explanatory footnotes and a minor improvement to Figure 2. The original published version is available from http://www.jstor.org/stable/3619777. 2 Department of Anaesthesia, Nottingham University Hospitals, City Hospital Campus, Nottingham, UK. email: dick@nickalls.org [www.nickalls.org] 3 Nickalls RWD (2009). The quartic equation: invariants and Eulers solution revealed. The Mathematical Gazette; 93 (March), 66-75. (http://www.nickalls.org/dick/papers/maths/ quartic2009.pdf)

RWD Nickalls

The Mathematical Gazette (1993); 77, pp. 354359

= + makes the sum of the roots of the new polynomial () equal to zero. It is easy to show that for the polynomial equation + 1 + . . . + = 0 = /(). If () is a cubic polynomial then () is known as the reduced cubic, and is the point of inflexion. Now consider the general cubic = 3 + 2 + + . Here is /(3), and the point of symmetry of the cubic. Let the parameters , , , be defined as the distances shown in Figure 1. It can be shown, and readers will easily do this, that and are simple functions of namely 4 5 2 = 3 2 and = 2 3 , where 2 3 . 92 This result is found easily by locating the turning points. Thus the shape of the cubic is completely characterised by the parameter . Either the maxima and minima are distinct ( 2 > 0), or they coincide at ( 2 = 0), or there are no real turning points ( 2 < 0). Furthermore, the quantity 2 / is constant for any cubic, as follows 2 3 = . 2 2 = If a polynomial curve passes through the origin, then the product of the roots 1 , 2 , , 1 (excluding the solution = 0) is related to the product of the -coordinates of the turning points 1 2 1 by 1 2 1 = 1 2 1 , a result whose proof readers can profitably set to their classes, and which parallels a related but much more difficult result about the -coordinates of the turning points which we have discovered 6 .
Unfortunately in the original printed version was presented with a negative sign. Note that if denotes the slope at the point of inflexion , then = 3 2 . See also Thomas M llers interactive demonstration of how the parameters , , , , u influence the shape and location of the cubic at http://demonstrations.wolfram.com/ ParametersForPlottingACubicPolynomial/ 6 Nickalls RWD and Dye RH (1996). The geometry of the discriminant of a polynomial. The Mathematical Gazette, 80 (July), 279285 (jstor). (http://www.nickalls.org/dick/papers/ maths/discriminant1996.pdf)
5 4

The relationship 2 = 3 2 is a particular case of the general observation that

RWD Nickalls

The Mathematical Gazette (1993); 77, pp. 354359

Solution of the cubic

In addition to their value in curve tracing, I have found that the parameters , , and , greatly clarify the standard method for solving the cubic since, unlike the Cardan approach (Burnside and Panton, 1886), they reveal how the solution is related to the geometry of the cubic. For example the standard Cardan solution, using the classical terminology, involves starting with an equation of the form 3 + 31 2 + 31 + = 0, and then substituting = + (1 /) to generate a reduced equation of the form 3 + where = 1 2 1 3 + 3 = 0, 2 = 2 31 1 + 23 . 1

and

This obscures the fact that the reduced form of the cubic has the point on the -axis. Subsequent development yields a discriminant of the form 2 + 4 3 where 2 + 4 3 = 2 (2 2 61 1 + 43 + 43 32 2 ). 1 1 1 1 The problem is that it is not clear geometrically what the quantities and represent. However, by using the parameters described earlier, not only is the solution just as simple but the geometry is revealed.

2.1

New approach

Start with the usual form of the cubic equation () = 3 + 2 + + = 0, having roots , , , and obtain the reduced form by the substitution = + (see Figure 1). The equation will now have the form 7 3 3 2 + = 0, (1)

and have roots , , ; a form which allows the use of the usual identity ( + )3 3( + ) (3 + 3 ) = 0. Thus = + is a solution where = 2
7

and

3 + 3 = /.

( ) 23 /(272 ) /(3) +

RWD Nickalls

The Mathematical Gazette (1993); 77, pp. 354359

Solving these equations as usual by cubing the first, substituting for in the second, and solving the resulting quadratic in 3 gives 3 =
1 2 42 6 2 { }

and since 2 = 42 6 , this becomes 8 3 =

1 2 2 2 { }

(2)

When this solution is viewed in the light of Figure 1, it is immediately clear that Equation 2 is particularly useful when there is a single real root, that is when 2 > 2 . Contrast this with the standard Cardan approach which gives 3 =
} 1 { 2 + 4 3 , 23

which completely obscures this fact. The values of , , and 2 + 4 3 are therefore found to be = 2 , = 2 2 and
2 2 + 4 3 = 4 ( 2 ).

However, Equation 2 can be rewritten as

1 ( + ) ( ) = 2
3

If the -coordinate of a turning point is then let + = 1 and = 2 .

Our solution (Equation 2) can therefore be written as 3 =

} 1 { 1 2 . 2

Using the symbol 3 for the discriminant 9 of the cubic, we have

2 3 = 1 2 = 2 .
8

Providing = 0 then this is equivalent to (see also footnote 11)

{
=
9 3 3

2 1 2

}
.

Note that the product 1 2 of the y-coordinates of the turning points is known as the geometric discriminant of the cubic (see Nickalls RWD and Dye RH (1996)for URL see footnote 6). The classical algebraic discriminant (2 + 4 3 ) has the same sign as the 2 geometric discriminant since 2 + 4 3 4 ( 2 ) = 4 1 2 .

RWD Nickalls

The Mathematical Gazette (1993); 77, pp. 354359

Returning to the geometrical viewpoint, Figure 1 shows that the rest of the solution depends on the sign of the discriminant 10 as follows:
2 > 2 2 = 2 2 < 2

1 real root, 3 real roots (two or three equal roots), 3 distinct real roots.

These are now dealt with in order.

2.2

2 > 2

i.e. 1 2 > 0, or Cardans 2 + 4 3 > 0

Clearly, there can only be 1 real root under these circumstances (see Figure 1). As the discriminant is positive the value of the real root is easily obtained as 11

= +

1 2 + 2 2 (

1 2 2 2 (

2.3

2 = 2

i.e. 1 2 = 0, or Cardans 2 + 4 3 = 0

Providing = 0 this condition yields two equal roots, the roots being = , and 2. The true roots are then = + , + and 2. Since there are two double root conditions the sign of is critical, and depends on the sign of , and so in these circumstances has to be determined from =

. 2

If = = 0 then = 0, in which case there are three equal roots at = .

10 11 2 Since the sign reflects the relative magnitude of and 2 . The remaining two complex roots are given by 3 2 4 2 , = 2 2

where 2 = 1 (for derivation see: Nickalls RWD (2009). Feedback: 93.35: The Mathematical Gazette; 93 (March), 154156. http://www.nickalls.org/dick/papers/maths/cubictables2009. pdf). Note also that multiplying top and bottom by ( = 0) gives

{ } 2 3 = + + 1 + 2
2 and since = 2 3 then /2 = 3 , and hence

{ } 2 3 1 , 2
2

{ { } } 2 2 3 3 = + + 1 + 1 , 2 2
which highlights the significance of the ratio /. Thus when / = 1 we obtain = + 2. See footnote 13 for the case when 1 / 1.

RWD Nickalls

The Mathematical Gazette (1993); 77, pp. 354359

Y

. . . . . . . . 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . ......... . .. ..... . . ... . .. . .. .. . . . . .. . . .. . . . . .. . . .. N . . .. . . .. . . . ... .. . . ... ... . . ... . . . .......... . . . . . . .

.............. ....................... ....... .... .... ... .... ... ... .. .. ... ..... ... ... ... ... ... .. .. .. .. .. .. .. . ..... .. . ..... .. . .. . . ....... .. .. . .. ........ .. ..... .. . ..... .. .. . . .. . . . . . . . ... .. ... .. . . . .. .... . . . .. . .. . . .. .. . .. .. . . . .. . ... . ... . . .. . . ... ... .. . . .... ... . . . .. . . . .. . . . . ... . .. ... .. .... . . .. .... . . .. . . . .. . . . . . . . . . . . . . . .. . .. . ... . . .. . .. . .. .. . ... . . . . . . . . . .. . . . . . . . . . .. . . .. . . .. . . .. .. . .. . .. .. . .. . . .. . .. .. . .. . . .. .. ... ..... .. .. .. . .. ...... .. .. . . .... . ... .. . . . .... ... .. . .. . .. . . .. . ... .... . ... .... . ... .... . . .... . ..... . . .... ....... . .................... . . ............... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 2:

2.4

2 < 2

i.e. 1 2 < 0, or Cardans 2 + 4 3 < 0

From Figures 1 and 2 it is clear that there are three distinct real roots in this case. However, our solution requires that we find the cube root of a complex number, so it is easier to use trigonometry to solve the reduced form using the substitution 12 = 2 cos in Equation 1. This gives 2 3 4 cos3 3 cos + = 0, and since 2 3 = , this becomes cos 3 = . (3)
( )

12 Note that if we use instead the substitution = 2 sin , this leads to sin 3 = /, for which the condition = 0 is associated with = 0. Use of this form is described in Nickalls, RWD (2006). Vi`te, Descartes and the cubic equation. The Mathematical Gazette; e 90 (July), 203208 (jstor). (http://www.nickalls.org/dick/papers/maths/descartes2006.pdf).

RWD Nickalls

The Mathematical Gazette (1993); 77, pp. 354359

The three roots , and are therefore given by 13 = + 2 cos , = + 2 cos (2/3 + ), = + 2 cos (4/3 + ). These are shown in Figure 2 in relation to a circle, radius 2, centered above N. Note that the maximum between roots and corresponds to the angle 2/3. It is clear from Equation 3 that trigonometry can only be used to solve the reduced cubic when 1 +1, a point which is completely obscured by the corresponding classical form cos 3 = 2 () 2
3

13

Note that from footnote 11 (section 2.2) we have

{ { } } 2 2 3 3 = + 1 + + 1 . 2 2
Since for the three real-root condition 1 / +1 (see Figures 1 and 2) then

= +
3

1 + 2
2
3

2 , 2

and the substitution / = cos 3 then gives = + and hence = + (cos sin ) + (cos + sin ), i.e. = + 2 cos . Note that and can also be expressed as , = + 2 cos (2/3 ). Also, it is sometimes convenient to express the set of roots in the following succinct form: , , = + 2 cos (2/3 + ) , ( = 0, 1, 2).

cos 3 sin 3 +

cos 3 + sin 3,

RWD Nickalls

The Mathematical Gazette (1993); 77, pp. 354359

Example
3 72 + 14 8 = 0

Solve the equation 14 The parameters are

= 7/3, = ( ) = 07407, 2 = 7/9 and = |2 3 | = 13718.

2 Since < 2 , it follows (see Figures 1 and 2) that there are three distinct real roots, which are given by = + 2 cos ,

where

07407 = = 05399. 13718 So = 191066 , and the three roots are cos 3 = = = =
7 3 7 3 7 3

+2 +2 +2

7 7

9
7 9

cos 191066 = 4, cos 1391066 = 1, cos 2591066 = 2.

Conclusion

Finally, I would like to suggest that the usual Cardan-type terminology for cubics and quartics, though it has been used for hundreds of years, be abandoned in favour of the parameters , , , , which reveal to such advantage how the algebraic solution is related to the geometry of the cubic.

References
Burnside W. S. and Panton A. W. (1886). The theory of equations: with an introduction to the theory of binary algebraic forms. 2nd ed. Longmans, Green and Co., London. http://quod.lib.umich.edu/cgi/t/text/text-idx? c=umhistmath;idno=ACA7397

333

14 For another example see: Nickalls RWD (1996). A note on solving cubics. The Mathematical Gazette; 80 (November), 576577 (jstor). (http://www.nickalls.org/dick/papers/maths/ cubefink.pdf)