2

The Cubic formula:

$ax^3+bx^2+cx+d=0$

With under the following conditions:

  1. $a \neq 0 $
  2. $a,b,c,d \in \Bbb{R} $

We can derive the following formula as a the root of $x$:

$u= \sqrt[3] {{bc\over 6a^2}-{d\over 2a}-{b^3\over 27a^3}+\sqrt{({bc\over 6a^2}-{d\over 2a}-{b^3\over 27a^3})^2+({c\over 3a}-{b^2\over 9a^2})^3}} $

$v= \sqrt[3] {{bc\over 6a^2}-{d\over 2a}-{b^3\over 27a^3}-\sqrt{({bc\over 6a^2}-{d\over 2a}-{b^3\over 27a^3})^2+({c\over 3a}-{b^2\over 9a^2})^3}} $

$x_1= u + v-{b\over 3a}$


But the proof of derivation of the formula mentioned above was only limited to 1 root. In addition, that formula had no complex conjugates. However as we know, there must be 2 other roots which includes complex conjugates in their formulas.

In reference, Wikipedia: Cubic equation also says that there should be 2 other roots at maximum.

So, altogether the 3 roots are:

$i= \sqrt {-1} $

$ \omega = -{1\over 2} + {{\sqrt 3}i\over 2} $

  • $x_1= u + v-{b\over 3a}$

  • $x_2= {\omega }u + {\omega}^2v -{b\over 3a} $

  • $x_3= {\omega }^2 u + {\omega}v -{b\over 3a} $

And unfortunately, I didn't find or know know the proof of any other 2 roots, i.e $x_2$ & $x_3$.

So would you please show me the proof of the other 2 roots of Cubic formula?

Note: Please, no synthetic division. I need proof by formula
  • 1
    Every general cubic can transform to reduced cubic $y^3-3py-2q=0$ by substituting $x=y-\dfrac{b}{3a}$.

    If once accept the identity $$a^{3}+b^{3}+c^{3}-3abc \equiv (a+b+c)(a+\omega b+\omega^{2} c)(a+\omega^{2} b+\omega c)$$ and replacing $a, b, c$ by $y, -u, -v$ respectively,

    $y^{3}-3uvy-(u^{3}+v^{3})=0 \implies y_{k}=u, \omega^{k}+ v, \omega^{2k}$ for $k=0,1,2$.

    Note that $\omega^4=\omega$ and see more in another answer.

    – Ng Chung Tak Aug 12 '20 at 03:09
  • Further tricks, try to simplify $u^3+v^3$ and $uv$ from your expressions first. – Ng Chung Tak Aug 12 '20 at 03:20
  • Wow, @NgChungTak That helps & I hope that should be illustrating the answer of this question. Thanks a lot – Prince Khan Aug 12 '20 at 03:40
  • You can take three different cube roots in each of the expressions for $u$ and $v$ giving nine possible pairings. Other solutions and comments give conditions which amount to a recipe for taking compatible cube roots (of which there are three pairs giving three solutions - choose the cube roots so that $uv$ is real). In the case where the coefficients are real and there is only one real solution, the other two solutions are complex conjugates of each other. – Mark Bennet Dec 20 '20 at 20:16
  • @MarkBennet, can you tell me the motivation or approach of Vieta's substitution? I need to know he got this: $y = z-\frac {p}{3z}$ – Prince Khan Dec 20 '20 at 20:20
  • The substitution you are talking about (and the letters vary) is just a translation of axis which makes the sum of the roots equal to zero. It gets rid of one of the coefficients and greatly simplifies both calculations and formulae. You simply reverse the translation at the end to get the roots of the original equation. – Mark Bennet Dec 20 '20 at 22:28
  • Your formulae for $u, v$ must be treated with caution. See my answer to https://math.stackexchange.com/q/2838797. The problem is that the square root occuring in these formulae may be imaginary in which case it is not all clear which of the three complex cubic roots should be taken. In fact, it turns out that $u$ and $v$ cannot be chosen indepedently. If you make a choice for, say, $u$, this uniquely determines $v$. You do not even need to perform a separate calculation for $v$. – Paul Frost Dec 21 '20 at 00:21

3 Answers3

4

Use polynomial long division to divide $ax^3+bx^2+cx+d$ by $x-x_1$. This will give you a quadratic, and from there you can find the two remaining roots.

You can do this systematically, using the usual steps for division. If the polynomial is relatively simple, then an alternative would be to write out $$ax^3+bx^2+cx+d = (x-x_1)(ax^2+px+q),$$ then start matching coefficients. For example, looking at $x^2$, we have $b = p-ax_1$, and so on.

Théophile
  • 24,627
  • That's a nice idea, but however, I said proof by formula. Although I didn't mentioned the synthetic division, I'll edit my question so that no more confusion like this arose. – Prince Khan Aug 12 '20 at 02:44
  • 1
    @PrinceKhan You can easily derive that formula by carrying out division on the general form. – Théophile Aug 12 '20 at 14:12
  • Such as? I said I need proof BTW, everyone can do that arithmetically – Prince Khan Aug 12 '20 at 15:37
2

Starting with the "depressed" cubic:

$y^3 + py + q = 0$

If you don't have a diminished cubic you can substitute $x = y - \frac {b}{3a}$ which will eliminate the $bx^2$ term from the original cubic.

Next we do a similar substitution:

$y = z-\frac {p}{3z}$

$(z - \frac {p}{3z})^3 + p(z-\frac {p}{3z}) + q = 0\\ z^3 - pz + \frac {p^2}{3z} - (\frac {p}{3z})^3 + pz-\frac {p^2}{3z} + q = 0\\ z^3 - (\frac {p}{3z})^3 + q = 0\\ z^6 + qz^3 - (\frac {p}{3})^3 = 0\\ z^3 = -\frac {q}{2} \pm \sqrt {(\frac {q}{2})^2 + (\frac {p}{3})^3}$

We can choose $z^3 = -\frac {q}{2} + \sqrt {(\frac {q}{2})^2 + (\frac {p}{3})^3}$
with $\frac {1}{z^3} = \frac {-\frac {q}{2} - \sqrt {(\frac {q}{2})^2 + (\frac {p}{3})^3}}{-(\frac {p}{3})^3}$
Or, $-\frac {q}{2} - \sqrt {(\frac {q}{2})^2 + (\frac {p}{3})^3} = (-\frac {p}{3z})^3$

Note that $\omega^3 = 1$ has $3$ roots
$\omega_1 = -\frac 12 + \frac {\sqrt 3}{2} i\\ \omega_2 = -\frac 12 - \frac {\sqrt 3}{2} i = \omega_1^2 = \frac {1}{\omega_1}\\ \omega_3 = 1$

$z = \omega {\sqrt[3]{-\frac {q}{2} + \sqrt {(\frac {q}{2})^2 + (\frac {p}{3})^3}}}$

with $z$ equal to any of the values of $\omega$ above, and

$y = z - \frac {p}{3z} = \omega {\sqrt[3]{-\frac {q}{2} + \sqrt {(\frac {q}{2})^2 + (\frac {p}{3})^3}}} + \frac {1}{\omega}{\sqrt[3]{-\frac {q}{2} - \sqrt {(\frac {q}{2})^2 + (\frac {p}{3})^3}}}$

If $(\frac {q}{2})^2 + (\frac {p}{3})^3 >0 $ there is only one real root.
If $(\frac {q}{2})^2 + (\frac {p}{3})^3 <0 $ the complex terms ultimately cancel out and there are 3 real roots.

Doug M
  • 57,877
  • How did you changed that?

    $\frac {1}{z^3} = \frac {-\frac {q}{2} - \sqrt {(\frac {q}{2})^2 + (\frac {p}{3})^3}}{-(\frac {p}{3})^3}$

    – Prince Khan Aug 12 '20 at 10:03
  • 1
    $\frac {1}{z^3} = \frac {1}{-\frac {q}{2} + \sqrt {(\frac q2)^2 + (\frac{p}{3})^3}}\left(\frac {-\frac {q}{2} - \sqrt {(\frac q2)^2 + (\frac{p}{3})^3}}{-\frac {q}{2} - \sqrt {(\frac q2)^2 + (\frac{p}{3})^3}}\right) = \frac {-\frac {q}{2} - \sqrt {(\frac q2)^2 + (\frac{p}{3})^3}}{-(\frac p3)^3} $ – Doug M Aug 12 '20 at 10:28
1

By $a^{3}+b^{3}+c^{3}-3abc=(a+b+c)(a+\omega b+\omega^{2} c)(a+\omega^{2} b+\omega c)$;

and replacing $a, b, c$ by $x, -u, -v$ respectively,

$x^{3}-3uvx-(u^{3}+v^{3})=0 \implies x_{k}=u\, \omega^{k}+ v\, \omega^{2k}$ for $k=0,1,2$.

$u, v$ are known as resolvents.