CUBIC AND QUARTIC FORMULAS

James T. Smith
San Francisco State University

Quadratic formula

You’ve met the quadratic formula in algebra courses: the solution of the quadratic
equation
ax 2 + bx + c = 0

with specified real coefficients a =

/ 0, b, and c is

b  b2  4 ac
x= .
2a
You can derive the formula as follows. First, divide the quadratic by a to get the
equivalent equation
b c
x2 + x+ =0.
a a
Now substitute x = y + d. You’ll choose d later so that the resulting equation is easy
to solve. Making the substitution, you get
b c
( y + d)2 + ( y + d) + =0.
a a
Work this out, ignoring some details that won’t be necessary:
b
y2 + 2 d y + d2 y + constants = 0
a
 b
y 2 +  2d   y + constants = 0.
 a
If the y coefficient were zero, you could move the constants to the other side and solve
for y by taking the square root. Thus you can find y easily if you let d = – b/(2a). Do
that and work out the details: the last equation displayed above becomes

2 b2 c
y – 2
+ =0
4a a

2013-08-25 14:45
Page 2 CUBIC & QUARTIC FORMULAS

b2  4 ac
y2 = .
4 a2
The quantity D = b2 – 4ac is called the discriminant of the quadratic: you can write
y = ± D /(2a). Finally, the desired solution is

b D
x=d+y=– ±
2a 2a
—the Quadratic Formula. If D < 0 then you can write x in complex form:

b D
x=d+y=– ± i.
2a 2a

Cubic formula

It’s possible to imitate this process to derive a formula for solving a cubic equation
ax 3 + bx 2 + cx + d = 0
with specified real coefficients a =
/ 0, b, c, and d. First, divide the cubic by a to get
the equivalent equation
b 2 c d
x3 + x + x+ =0.
a a a
Next, make a substitution x = y + e, and choose e so that the resulting equation is
easier to solve. Experience with the quadratic equation suggests that you might be able
to make one term of the cubic disappear, so that the resulting equation is like one of
these:
y3 + f y2 + g = 0 y 3 + py + q = 0.

Experimentation would show that you can reach the form on the left, but it doesn’t lead
anywhere. However, you can also get the one on the right, and it does help. In fact, if
you let e = – b/(3a) —that is, substitute x = y – b/(3a) —then you get, after consider-
able algebraic labor, the equation
y 3 + py + q = 0,
where
 b2 c 2 b3 bc d
p= 2
+ q= 3
– 2
+ .
3a a 27 a 3a a
CUBIC & QUARTIC FORMULAS Page 3

You’ve now reduced the problem of solving the original cubic to one of solving y3 + py +
q = 0 and setting x = y – b/(3a).

After centuries’ experience, mathematicians found a trick to solve this simpler

cubic. Set y = z – p/(3z) and consider the resulting equation. After considerable
algebraic labor, it comes out to

6 p3 3
z + qz – = 0.
27
This equation is quadratic in z 3. You can find z 3 by using the quadratic formula:
q
z3 = – ± D,
2
where
q2 p3
D= + .
4 27
This D is called the discriminant of the cubic equation. Taking the cube root, you get

q
3   D .
2
Together with the equations
p b
y=z– x=y–
3z 3a
this is called the cubic formula: it shows you how to compute the solution x of the
original cubic equation. Actually, the equation for z gives three complex cube roots for
each of the + and – signs, hence six different formulas for z. But when you substitute
these in the equation for y, at most three different y values will result, and the last
equation will thus give at most three distince roots x.

Look at the cubic formula in more detail. When D $ 0, you can select one of the
two real square roots ± D , then find three cube roots z = z 0 , z1 , and z 2 of
– q/2 ± D as follows. Let z 0 be the real cube root, then z1 = z 0 and z 2 = 2 z 0 ,
where  and 2 are the two complex cube roots of 1:

1 3 1 3
=– + i 2 = – – i.
2 2 2 2
You can use simple algebra of complex numbers to find the corresponding values y
and x.
Page 4 CUBIC & QUARTIC FORMULAS

If you get z = 0 in this process, then (q/2)2 = D, hence p = 0. In that case,

you’re just solving the equation y3 + q = 0, which has one root y = 0 if q = 0, and
three roots y = – qa otherwise. In the former case, D = p = q = 0. If D = 0 but p = /
0, then two of the y values coincide, and the equation has distinct single and double
real roots.

The situation is more interesting when D < 0, for then D is imaginary, and
you have to find the cube roots z of the complex number
q
A=– + D i .
2
In this case,
q2 p3
0>D= + ,
4 27
so p < 0. Next,
2
q2 p3
 q
 
2
*A* =    +
2
D = –D=–
 2 4 27
3
 p 2
*A* =   .
3
 
Now you can write A = *A* cis , where
q

Re A 2 3 3q
cos  = = 3 = .
A 2p p
 p 2

 
 3
Given coefficients a to d, you can compute p and q from the previous equations,
calculate cos 2, then determine 2. Now let

p
r= 3 A = ,
3
so that De Moivre’s formula yields the three values

z= 3
A = r cis ,
where

  
= , + 120E, + 240E.
3 3 3
CUBIC & QUARTIC FORMULAS Page 5

The details of the final calculation of y and x are interesting, too. You’ll get
p p
y=z– = r cis  –
3z 3r cis 
p p
= r cis  – (cis )–1 = r cis  – cis(– )
3r 3r
p
= r [cos  + i sin ] – [cos(– ) + i sin(– )]
3r

 p  p
= r   cos  + r   i sin  .
 3r   3r 
But
p p  p
r+ = + = 0,
3r 3 p
3
3
hence all three y values are real! You have

 p
y = r   cos 
 3r 
for the three values of R given earlier. In this case, where D < 0, you’ve found the real
roots of a real cubic equation through use of complex numbers and trigonometry. It’s
now known that some such ‘detour’ through the complex numbers is necessary to find
a formula for these roots!

Example cubics

1. To solve ax 3 + bx 2 + cx + d = 0 with a, b, c, d = 4, 3, 2, 1 follow the derivation

of the cubic formula, setting
b 1
x=y– =y–
3a 4

 b2 c 5
p= 2
+ =
3a a 16
Page 6 CUBIC & QUARTIC FORMULAS

2 b3 bc d 5
q= 3
– 2
+ =
27 a 3a a 32
p 5
y=z– =z–
3z 48z
b 1
x=y– =y–
3a 4
q2 p3 25
D= + = .
4 27 3456
Since D > 0, there are one real and two conjugate complex roots. Compute the real root
as follows:

q 45  20 6  0.1906230
z= 3   D = 3 .  .
2 576  0.5464530
Both z values give the same value of
5
y=z– . – 0.35583
48 z
1
x=y– . – 0.60583.
4
Substituting this x value in the left-hand side of the original equation yields a value of
about –1×10–6 —acceptable accuracy. The complex roots are computed from the other
two cube roots:
 1 3 
z . 0.1906280    i  . – 0.0953115 ± 0.1650844i
 2 2 
5
y=z– . 0.17792 ± 0.63827i
48z
1
x=y– . – 0.07208 ± 0.63827i.
4
Substituting these x values in the left-hand side of the original equation yields values
with norms of about 2×10–4 i —acceptable accuracy.

2. To solve y3 + py + q = 0 with p, q = –2, 1 follow the derivation of the cubic

formula, to get
CUBIC & QUARTIC FORMULAS Page 7

q2 p3 5
D= + =– .
4 27 108
Since D < 0, there are three real roots, determined as follows:

3 3q
cos  = . – 0.91856
2p p
 . 156.716E

  
= , + 120E, + 240E . 52.239E, 172.239E, 292.239E
3 3 3

p 2
r= =
3 3
 p
y = r   cos  . 1.00000, –1.61803, 0.61803
 3r 
The first value of y suggests that y = 1 is an exact root of the cubic. That’s easy to
verify: in fact,

y 3 – 2 y + 1 = ( y – 1)( y 2 + y – 1) .

The roots of the right-hand factor are (–1 ± 5 )/2, which agree with the other two
computed y values.

Quartic formula

Why stop with cubics? Why not apply the same method to a quartic equation

ax 4 + bx 3 + cx 2 + dx + e = 0
with real coefficients a =
/ 0, b, c, d, and e? First divide the quartic by a to obtain the
equivalent equation
b 3 c d e
x4 + x + x2 + x+ =0.
a a a a
Page 8 CUBIC & QUARTIC FORMULAS

The substitutions x = y – b/(2a) and x = y – b/(3a) worked for the quadratic and cubic;
try x = y – b/(4a) for the quartic. After a lot of algebra, you’ll get an equation of the
form
y 4 + py 2 + q y + r = 0

for certain values p, q, and r that you can compute from the original coefficients. Not
long after the trick was discovered that led to the solution of the cubic, mathematicians
discovered one that leads to the solution of this quartic. Write it as
y 4 = – py 2 – q y – r.
Now manipulate this equation, using a value z that will be determined later:
y 4 + 2 y 2 z 2 + z 4 = – py 2 – q y – r + 2 y 2 z 2 + z 4
( y 2 + z 2 )2 = (2z 2 – p) y 2 – q y + (z 4 – r).
Later, you’ll determine z so that the right-hand side of this last equation is ( f y + g)2
for some particular values f and g. Then you’ll have
( y 2 + z 2 )2 = ( f y + g)2
y2 + z2 = ±( f y + g)

 y 2  fy  ( z 2  g)  0, or
 2
 y  fy  ( z  g)  0.
2

Once z then f and g are found, you can solve for y by using the quadratic formula
on the last pair of equations.

Thus you have to find z, f, and g so that

(2z 2 – p) y 2 – q y + (z 4 – r) = ( f y + g)2 .
Consider the equations formed by setting each side of this equal to zero. The right-hand
equation (fy + g)2 = 0 would have just one root y = – g/f; thus the left-hand equation
(2z 2 – p) y 2 – q y + (z 4 – r) = 0
would have just one root also. But you can solve this equation for y by using the quadra-
tic formula, and for that formula to yield just one root its discriminant must be zero.
That is, you can find z, f, and g just when

q 2 – 4(2z 2 – p)(z 4 – r) = 0
– 8z 6 + 4 pz 4 + 8rz 2 + (q 2 – 4 pr) = 0
 w  z2

 8w  4 pw  8 rw  (4 pr  q )  0.
3 2 2
CUBIC & QUARTIC FORMULAS Page 9

With sufficient labor, you can solve the last equation by using the cubic formula, getting
at most three solutions w = w0 , w1 , and w2 . For each j, the equation
(2wj – p) y 2 – q y + (wj2 – r) = 0
has a single root
q
y= .
2(2w j  p)
Therefore, 2
 q 
(2wj – p) y 2 – q y + (wj2 – r) = (2wj – p)  y  
 2(2w  p)
 j 
2
 q 
=  y 2w j  p   = ( f y + g)2 ,
 2 2w j  p 

where
q
f= 2w j  p g=– .
2f
With these three values of f and g, solve the two quadratic equations

 y 2  fy  ( z 2  g)  0
 2
 y  fy  ( z  g)  0
2

for y. This yields as many as twelve possible y formulas, but at most four can have
distinct values. Finally, calculate the roots x = y – b/(4a).

Example quartic

To solve ax4 + bx3 + cx2 + dx + e = 0 with a, b, c, d, e = 2, 2, –3, –3, –4, follow the
quartic formula, dividing by a and setting
b 1
x=y– =y– ,
3a 4
to get the equation y 4 + py 2+ q y + r = 0, where p = –15/8, q = – 5/8, and r =
–443/256. Next, solve the equation 8w 3 – 4pw 2 – 8rw + (4 pr – q 2 ) = 0 by the cubic
formula, obtaining three solutions
w 0 . – 0.918531
w1 . – 0.00948425 + 1.30880i
w2 = w—.
1
Page 10 CUBIC & QUARTIC FORMULAS

With j = 0 compute
q
f= 2w j  p , g=– ,
2f
solve y 2 – f y + (wj – g) = 0 for two values y = y 0 and y1 , and set x k = yk – b/(4a) for
k = 0, 1 to get
x 0 . –1.74398, x1 . 1.43875.
Next, solve y 2 + f y + (wj + g) = 0 for two values y = y2 and y3, and set xk =
yk – b/(4a) for k = 2, 3 to get
x . – 0.347387 + 0.822439i,
2 x =— x .
3 2

With j = 1, 2 this process yields the same four roots x. Substituting these x values
back in the original quartic equation yields values with norm less than 2×10–10
—acceptable accuracy. Like the cubic examples considered earlier, these roots can be
expressed exactly with radicals and trigonometric functions. However, their expressions
would be so complicated that checking and using them would be impractical.

History

In 1505 Scipione de Ferro (1465–1526), Professor of Mathematics at Bologna,

discovered a method for solving certain cubic equations, but didn’t publicize it beyond
his students, preferring to use it secretly to establish himself as a problem solver. In 1535
one of the students, Antonio Maria Fior, challenged Nicolo Fontano of Brescia, nick-
named Tartaglia (stammerer), and the latter discovered the same method. Tartaglia was
a self-taught mathematics teacher who had already written the first serious treatise on
ballistics, and would later translate Euclid into Italian. He was pressured to reveal the
secret by Geronimo Cardano (1501–1575), a famous and infamous professor of mathemat-
ics, medicine, and roguery at Pavia and other North Italian universitites. On promise
of secrecy, Tartaglia showed him the method. But Cardano set to work elaborating it,
and soon his student Lodovico Ferrari had extended it to solve quartics. Cardano
published both in his Ars magna in 1545. This is one of several of his treatises typical
of the time: encyclopedias of everything, from occult descriptions of demons to natural
history to theoretical mathematics. Tartaglia became involved with Cardano and Ferrari
in a public priority dispute, and faded from the scene. Cardano became famous for
studies of syphilis and typhus, and for an autobiography (of a rogue and scoundrel). He
wrote one of the first books on probability, published posthumously. Cardano was
imprisoned in 1570 for the heresy of casting Christ’s horoscope; but the Pope rethought
the matter, released him, then hired him as papal astrologer!
CUBIC & QUARTIC FORMULAS Page 11

The Italians were handicapped by their lack of notation for variables. This was
provided by François Viète (1540–1603), a politician and lawyer from Brittany involved
with the Huguenot cause. He pursued mathematics as a hobby, especially while out of
office, and published privately a number of treatises in which he used an algebraic
symbolism similar to ours. The treatment of cubic and quartic equations given here is
essentially his.

The modern theory of roots of polynomials in general was not developed until the
middle 1600s by Descartes and others. The first complete treatment of cubic and quartic
equations was given by Euler in 1732. During the 1600s and 1700s, mathematicians
regarded extension of these methods to quintic and higher equations as a major open
problem, but met with no success. Not until 1800 did Gauss prove that every polynomial
has at least one complex root, and not until about 1830 did Galois, Abel, and others show
that roots of quintic and higher degree equations could not, in general, be found by the
familiar methods involving algebraic operations and extraction of roots.

Exercises

Find all roots of each of the following cubics. Verify each real root by substituting
it for x and calculating the left hand side of the equation.

1. x 3 – 6x – 6 = 0
2. 3x 3 – 6x 2 – 2 = 0
3. x 3 – 6x + 2 = 0
4. x 3 + x 2 – 2x – 1 = 0

Acknowledgement

Thanks to Singapore student Jessica Ng for finding two serious typographical errors in
the previous version of this note.
Page 12 CUBIC & QUARTIC FORMULAS

Solutions

1. To solve x 3 + px + q = 0 with p = q = – 6, set x = y = z – p/(3z) = z + 2/z.

The equation then becomes
8
z3 + –6=0
z3
z 6 – 6z 3 + 8 = 0
(z 3 – 4)(z 3 – 2) = 0,
hence z 3 = 4 or z 3 = 2. Select z 3 = 4 (it makes no difference which alternative you
pick). Then z = z 0 , z1 , or z 2 , where

z0 = 3
4 z1 = z 1  –,
z2 = z0
and x = x 0 , x1 , or x 2 , where
2
x0 = z0 + = 3
4 + 6
4 . 2.84732
z0
2 2 –  1 3  2  1 3 
x1 = z1 + = z0 +  = z0    i +   i
z1 z0  2 2  z0  2 2 
1 2 3 2
=   z0   
2 z0  2 
 z0   i =
z0 
x0
2
+
2
3 3
464 i  
. –1.42366 – 0.28361i
x2 = —
x1 . –1.42366 + 0.28361i.

2. To solve 3x 3 – 6x 2 – 2 = 0, set x = y – (– 6)/(3@3) = y + 2/3. The equation

becomes y 3 – (4/3) y – 34/27 = 0. Now set y = z + (4/3)/(3z) = z + (4/9)z, and the
equation becomes z 3 + (64/729)z 3 – 34/27 = 0, i.e. 729z 6 – 918z 3 + 64 = 0. By the
quadratic formula, z 3 = 32/27 or 2/27. Take the former—it makes no difference which.
Then z = z 0 , z1 , or z 2 , where
3
2 –,
z0 = z1 = z 1  z2 = z0
3
and x = xj = yj + 2/3,
1 5
4 2 3 2 3
y0 = z0 + =
9z0 3
CUBIC & QUARTIC FORMULAS Page 13

4 – 1 3 13
y1 = z 0  +
9z0
 = – y0 +
2 6
5
2 2 3 i  
y2 = —
y1 .
Thus
x 0 . 2.14490
x1 . – 0.072452 – 0.55278 i
x 2 . – 0.072452 + 0.55278 i.

3. To solve x 3 + px + q = 0, where p = – 6 and q = 2, set x = z – p/(3z) = z

+ 2/z. The equation becomes
8
z3 + +2=0
z3
z 6 + 2z 3 + 8 = 0

z 3 = –1 ± 7 i = r cis 

r= 17 = 2 2
1 2
cos  = – =–
r 4
 = 0 , 0 + 2, or 0 + 4
 2
0 = cos–1    . 1.93216
 4 
 
z= 3
r cis = 2 cis
3 3
2  2
x=z+ = 2 cis +
z 3 
2 cis
3
   
= 2 cis + 2 cis    = 2 2 cos
3  3 3
. 2.2618052, – 2.6016777, or 0.33987722.

4. To solve x 3 + x 2 – 2x – 1 = 0, set x = y – a. The equation becomes

3
y – (7/3) y – 7/27 = 0. Now set y = z + 7/(9z), and the equation becomes
Page 14 CUBIC & QUARTIC FORMULAS

343 7
z3 + 3
– =0
729z 27
7 3 343
z6 – z + =0
27 729
7 21 3
z3 =  i = r cis 
54 54
3
72
r= 3
3
7
3
r =
3
7
cos  =
14
 = 0 , 0 + 2, or 0 + 4

7
 0 = cos–1 . 1.38067
14
 7 
z = 3 r cis = cis
3 3 3
7 7  7
y=z+ = cis +
9z 3 3 
3 7 cis
3

7  7   2 7 
= cis + cis    = cos
3 3 3  3 3 3
1
x=y– . 1.2469792, –1.8019369, or – 0.4450425.
3