Solution of the third degree equation

Ole Witt-Hansen 2018 (1966)

 The third degree equation 1

1. Polynomials of the third degree

Our purpose is to derive a solution formula for the roots of a third degree polynomial.

(1.1) P ( z )  a3 z 3  a2 z 2  a1z  a0

Where the coefficients are real numbers and a3 is non-zero, while z may be a complex number.

We know that a third degree polynomial always has one real root, due to the fact that a third
degree polynomial is a continuous function, which has both positive and negative values, and
therefore also must have the value 0.
A third degree polynomial can have 1, 2 or 3 real roots, but according to the fundamental theorem
of algebra, it has always 3 (complex) roots.

Our aim is to derive a formula for finding the 3 roots.

To begin with, we do a simple rewriting, as we remove the second order term, by setting z = w + a.

P ( w)  a3 ( w  a )3  a2 ( w  a ) 2  a1 ( w  a )  a0

P( w)  a3 ( w3  3w2a  3wa 2  a 3 )  a2 ( w2  2aw  a 2 )  a1 ( w  a)  a0 

(1.2)
 a3w3  (3aa3  a2 ) w2  (...) w  (...)
a
If we choose a   2 , the second order term disappears, resulting in a third degree polynomial
3a3
having the form:

(1.3) P ( w)  a3 ( w3  pw  q )  a3Q( w)

Where p and q may be calculated from (1.2). So if Q(w) has the roots w1, w2, w3, then Q(w) may
be written as:

Q( w)  ( w  w1 )( w  w2 )( w  w3 )
And thereby
P( z )  a3 ( z  w1  a)( z  w2  a)( z  w3  a)

To determine the roots in an arbitrary third degree polynomial, we may settle for determining the
roots in a polynomial of the form:

(1.4) P( z )  z 3  pz  q

We can assume that both p and q are non-zero, since otherwise the solution becomes trivial.
Making the substitution: z = u + v, we get:
 The third degree equation 2

P( z )  (u  v)3  p(u  v)  q  u 3  v3  3u 2v  3uv 2  p(u  v)  q

(1.5)
P( z )  u 3  v3  (3uv  p)(u  v)  q

Now for each z we may choose u and v, such that: u  v  z  uv   3p , since these equations
just imply that u and v are roots in the quadratic equation: x 2  zx  p
3
 0.

So as you probably remember from high school:

In the ordered and reduces quadratic equation, x 2  bx  c  0 the sum of the roots is equal to
minus the coefficient to x, and their product is equal to the last term in the equation.

For this specific choice of u and v, we find:

(1.6) P( z )  u 3  v3  q

And the condition that z =u + v is a root is therefore u 3  v3   q .

When two numbers which fulfil the conditions: u 3  v3   q and uv   3p it is equivalent to

fulfilling the conditions:
p3
(1.7) u 3  v3   q and u 3v 3   27

According to the theorem about the sum and the product of the roots in a quadratic equation, this
means that u 3 and v 3 are roots in the quadratic equation:

p3 q q 2 p3
(1.8) x 2  qx  0  x  
27 2 4 27
q 2 p3
If  < 0, however, the two roots should be replaced by:
4 27
q q 2 p3
(1.9) x i  
2 4 27

However we shall continue our writing as if:

q 2 p3 q 2 p3
  0 , but with the significance above, if  < 0.
4 27 4 27

We therefore proceed taking the third root of the solution (1.8), as if we were dealing with real
numbers.
u 3 q q 2 p3
(1.10)     
v 2 4 27

And thus finding the (possible complex) solutions: z = u + v

 The third degree equation 3

q q2 p3 3 q q2 p3
(1.11) z3       
2 4 27 2 4 27

To determine the other (possible complex) solutions, we shall recall the binome equation:

(1.12) zn  a

Where z  | z | (cos x  i sin x) and a  | a | (cos v  i sin v)

(1.13) z n  a  | z |n (cos nx  i sin nx)  | a | (cos v  i sin v)

If the direction angles for a are: v  p 2 , p  0, 1, 2... we get immediately:

| z | n | a | and nx  v, v  2 , v  4 , v  (n  1)2 

v 2
| z | n | a | and x p , p  0,1,2..., (n  1)
n n

So the complete solution becomes:

v 2 v 2
(1.14) | z | n | a | (cos(  p )  i sin(  p )) , p  0,1,2..., (n  1)
n n n n

For the equation z 3  a , where a is a real number, we therefore find:

2 2 4 4
(1.15) z  3 a, z  3 a (cos( )  i sin( ), z  (cos( )  i sin( )
3 3 3 3
Or
z  3 a , z  3 a ( 12  i 2
3
), z  3 a ( 12  i 2
3
)

It is seen that: ( 12  i 2
3
)  ( 12  i 2
) , so we put α = ( 12  i
3 2
2
3
) and α2 = ( 12  i 2
3
)
The three roots in the equation then becomes, (since   1 ): 3

(1.16) z  3 a , z   3 a , z 2 3 a

q q 2 p3
If we return to the equation: x     , which is equivalent with the two binome
2 4 27
equations:
q q 2 p3 q q 2 p3
(1.17) u3     and v3    
2 4 27 2 4 27

Then if both right hand sides are real, we find according to (1.16).
 The third degree equation 4

q q 2 p3 q q 2 p3 q q 2 p3
u0  3     u1    3     u2   2  3   
2 4 27 2 4 27 2 4 27

(1.18)
q q2 p3 q q2 p3 q q 2 p3
v0  3     v1    3     v2   2  3   
2 4 27 2 4 27 2 4 27

One might think that it should result in six solutions to the equation z = u + v, but we recall that
the solutions u and v must obey the equation: uv   3p . That this applies to u0v0 is seen from:

 q q 2 p 3  3 q q 2 p 3 
u0v0   3       
 2 4 27  2 4 27 
  
(1.19)
 q 2 3  2 3  3
3    q  p   q  q  p   3  p   p
 2 4 27  2 4 27  27 3


A simple calculation shows that the three solutions, which comply with the claim: uv   3p are:

(1.20) u0  v0 , u0   2v0 and  2u0  v0 , where   ( 12  i 2

3
)

This follows from  2  ( 12  i 2

)  ( 12  i
3 2
2
3
) and  3  ( 12  i 2
3
)( 12  i 2
3
)  14  34  1 .

Finally the solutions to the equation:

z 3  pz  q  0
Are then given by:

q q2 p3 3 q q 2 p3
z  
3      
2 4 27 2 4 27
q q2 p3 q q2 p3
(1.21) z     3    
2 3
 
2 4 27 2 4 27
q q2 p3 q q2 p3
z    
2
 3   
3 
2 4 27 2 4 27

Since every third degree equation: a3 z 3  a2 z 2  a1 z  a0  0 can be transformed into the equation
a
z 3  pz  q  0 by the substitution z  z  a with a   2 , we have proven that every third
3a3
degree equation has the solutions shown above.
 The third degree equation 5

Notice, however, we have used the square root, as if the argument was a positive number, but that
is not necessarily the case, which we shall look into now.

We shall then consider the 3 cases:

q 2 p3
1)   0 . Here we must have p < 0: As u0 and v0 we may apply the real value of 3
 q2 ,
4 27
and we therefore find, according to the solution formula above.

(1.22) z1  u0  v0  23  q2 , together wit the double root: z2  z3  3  q2

q 2 p3 q q 2 p3
2)  > 0. The numbers    are then real, and we may apply the real cubic
4 27 2 4 27
roots for u0 and v0. We then have the one real root z1  u0  v0 and the two complex
conjugate roots.
z2  u0   2v0  u0  v0 u v
(1.23)  2  i 3 0 0
z3   u0  v0  2 2
q 2 p3
3)  < 0. Here we must have p < 0. For u and v, we apply:
4 27
u3  q q 2 p3
    i   .
v3  2 4 27

It appears that the two roots are complex conjugates, and therefore have the same modulus.

2
 q   q 2 p 3 
2
p3
(1.24)        .
 2   4 27  27

We then get the solutions to u and v, as we notice that:

p3 p
3    u v.
27 3
If we put:
q
q q 2 p3 p3 
 i     (cos   i sin  ), where cos   2
2 4 27 27 p3

27
Then the applicable solutions to the equations becomes.

u3  q q 2 p3
(1.25)     i  
v3  2 4 27
Equal to:
 The third degree equation 6

u0  p  
   (cos 3  i sin 3 )
v0  3
u0  p
(1.26) 2 
  (cos  32  i sin  32 )
 v0  3
 2u0  p   4   4
   (cos 3  i sin 3 )
v0  3

So in this case we get three real roots!

p  p   2 p   4
(1.27) z1  2  cos , z2  2  cos , z3  2  cos
3 3 3 3 3 3

Determining the roots in a concrete example

To determine the numeric solutions to a specific cubic equation e.g. z 3  2 z 2  5 z  6  0 , (which
is seen to have the roots {-2,1, 3}), is rarely possible analytically, mostly because of the factor

cos .
3
 
To express cos by cos  one has namely to solve a 3. degree equation in cos .
3 3
For the equation: z 3  2 z 2  5 z  6  0 we find:

p  3( 23 ) 2  4( 23 )  5  43  83  5   193
q  ( 23 )3  2( 23 )2  5( 23 )  6  8
27
 89  103  6  8 2427 90  6  106
27
6 56
27

q

2   54 .
56
cos  
p3 ( 193 )3

27 27

Therefore there seem to be only one way to get the solutions, namely by numerical calculations.

Reference: Børge Jessen. Lectures on complex numbers. Mat 2. 1965 -1966. (Handwritten notes).