3

The question: solve $$x^3-3x^2+4x-12=0$$ without using factoring (Cardano's method?)

So I first have to depress the equation so i make the substitution $x=z+1$. We know this is the substitition because it should be of the form $z-\frac{a_2}{3a_3}=z-\frac{-3}{3(1)}=z+1$. This then gives us

$$z^3+z-10=0$$

By cardano's method, we know that $p=1$ and $q=-10$. Thus we have that

$$1=-3ab \qquad -10=-a^3-b^3$$

Solving this system gives (i believe) for $a$ gives

$$a=\sqrt[3]{5\pm\frac{26\sqrt{3}}{9}}$$

$$b=\sqrt[3]{5\mp\frac{26\sqrt{3}}{9}}$$

and so

$$z=a+b=\sqrt[3]{5\pm\frac{26\sqrt{3}}{9}}+\sqrt[3]{5\mp\frac{26\sqrt{3}}{9}}$$

I have tried reducing this as best as I can, but I can not get anyone of the solutions.

If I were to factor the original equation, I should getg

$$x^3-3x^2+4x-12=x^2(x-3)+4(x-3)=(x^2+4)(x-3) \Rightarrow x=3, \pm2i$$

So where am I making my mistake?

Mando
  • 371
  • 3
    Note: $ \left(1\pm\dfrac2{\sqrt3}\right)^3=5\pm\dfrac{26\sqrt3}9$ – J. W. Tanner Aug 16 '20 at 17:56
  • @Mando I think it's better to ask, how we can solve this equation without using math? What do you think? – Michael Rozenberg Aug 16 '20 at 17:56
  • 2
    @MichaelRozenberg i don't understand what you mean? – Mando Aug 16 '20 at 17:58
  • Is the requirement just that you can't factor the equation, or that you must use Cardano's method? If not the latter, you can try using Vietta's formula applied to cubics –  Aug 16 '20 at 18:00
  • 1
    @J.W.Tanner, that doesn't seem trivial. But you must be thinking that the expression under the radical should be, if it is a real solution, a perfect cube. So you would say $$(m+n\sqrt{3}^3=m^3+3\sqrt{3}m^2n+9mn^2+3\sqrt{3}n^3=5+\frac{26\sqrt{3}}{9}$$ and solve for m and n? – Mando Aug 16 '20 at 18:01
  • Ah, The requirement is not to factor (easy to factor) but show Cardano's method works. – Mando Aug 16 '20 at 18:02
  • https://math.stackexchange.com/questions/3548756/finding-a-cubic-formula-for-roots-of-cubic-equations/3552398#3552398 – Arjun Aug 16 '20 at 18:13
  • 2
    I guess the remaining question is about denesting $$3a=\root3\of{135+78\sqrt3}$$ and its "conjugate" $3b$. Using the properties of the norm $N(u+\sqrt3 v):=u^3-3v^2$ we can observe that $N(135+78\sqrt3)=-27$ so the eventual cube root must have $N=-3$. This suggest the Ansatz $u+\sqrt3=(2+\sqrt3)^n\sqrt{3}$. A bit of trial and error shows that $n=1$ works. Observe that $N(2+\sqrt3)=1$, explaining why any power of it as a factor won't change the norm. – Jyrki Lahtonen Aug 16 '20 at 20:39
  • 2
    Justification and theory of denesting cube roots of quadratic irrationals is described in this answer by mercio - you need to know a few basic facts about quadratic number fields to keep up with everything. Bill Dubuque has broken the process into simple steps (for square roots only) here. – Jyrki Lahtonen Aug 16 '20 at 20:41
  • Work also many non-Cardano methods: $$2=15 - \frac{26}{\sqrt3} + \frac{52 (-45 + 26 \sqrt3)}{78 - 45 \sqrt3 + 3 \sqrt[3]{26 - 15 \sqrt3}}$$ – Dmitry Ezhov Aug 17 '20 at 09:37
  • ^ Should read "... Ansatz $u+v\sqrt3=(2+\sqrt3)^n\sqrt3$". Sorry about the missing $v$. – Jyrki Lahtonen Aug 18 '20 at 04:54

2 Answers2

5

Your calculations are correct, but it is necessary to complete the Cardano's method. Once you have calculated $a$ and $b$, the roots of the depressed cubic are as follows:

$$ \displaystyle z_{1}=a+b \\ {\displaystyle z_{2}=a\cdot \left(-{\frac {1}{2}}+i{\frac {\sqrt {3}}{2}}\right)+b\cdot \left(-{\frac {1}{2}}-i{\frac {\sqrt {3}}{2}}\right)} \\ {\displaystyle z_{3}=a\cdot \left(-{\frac {1}{2}}-i{\frac {\sqrt {3}}{2}}\right)+b\cdot \left(-{\frac {1}{2}}+i{\frac {\sqrt {3}}{2}}\right)} $$

Since in your case $a=1+2/\sqrt{3}$ and $b=1-2/\sqrt{3}$ (see below for the denesting procedure to obtain these values), the formulas give

$$z_1=2 \\ z_2=-1+2 i \\ z_3=-1-2 i$$

As $x=z+1$, you have

$$x_1=3 \\ x_2=2 i \\ x_3=-2 i$$


EDIT: as correctly stated in the comments, a key issue in applying Cardano's method is that, in some cases, there is the need to denest some cubic roots. This may sometimes be quite difficult. Some methods have been previously reported in the links provided in one of the comments. I would suggest a possible approach that sometimes works well for the radicand of the form $J+K\sqrt{n}$. The methods includes these steps:

  • set the cube root in the form $\sqrt[3]{J\pm K\sqrt{n}}$, with $J$ and $K$ integers;

  • assume that the radicand $A=J\pm K\sqrt{n}$ can be expressed as $(j\pm k\sqrt{n})^3$, with $j$ and $k$ rational numbers;

  • after expanding $(j\pm k\sqrt{n})^3$ and dividing its terms in two groups whose sums are equal to $J$ and $K\sqrt{n}$, use the resulting equations to determine $j/k$. This is the longer step, as it requires to search the rational roots of a new cubic equation using the rational root theorem, which sometimes may be tricky;

  • finally, determine the values of $j$ and $k$.


To better illustrate this method, let us try it for the specific case $\sqrt[3]{5+ \frac{26\sqrt{3}}{9}}$ (the same method can be used for the case where the radicand is $5-\frac{26\sqrt{3}}{9}$). Firstly, we have to set the radicand so that $J$ and $K$ are integers:

$$\sqrt[3]{5 + \frac{26\sqrt{3}}{9}}=\frac{1}{3} \sqrt[3]{135+ 78\sqrt{3}} = \frac{1}{3} \sqrt[3]{A} $$

Now let us hypothesize $A=(j+k\sqrt{3})^3$. Therefore

$$A= j^3+3\sqrt{3}j^2k+ 9jk^2+3\sqrt{3}k^3\\ =j(j^2+9k^2)+3k(j^2+k^2)\sqrt{3}$$

so that we can write

$$j(j^2+9k^2)=135\\ 3k(j^2+k^2)=78$$

Note that $j$ and $k$ have to be both positive. From the two equations above we have

$$78\cdot j(j^2+9k^2) =135\cdot 3k(j^2+k^2)$$

We have now to try to determine $j/k$. Dividing both members to $k^3$ and moving all terms to the LHS, we have

$$78\left(\frac{j}{k}\right)^3 - 405 \left(\frac{j}{k}\right)^2 + 702\left(\frac{j}{k}\right) - 405=0 $$

Setting $x=j/k$ and simplifying the coefficients, we get

$$26 x^3-135 x^2+234x-135=0$$

Using the rational root theorem, we can search for a rational root $p/q$ for the last equation, where the integer $p$ divides $135=3^3\cdot 5$ and the integer $q$ divides $26=2\cdot 13$. To speed up the search of a real root, it can be observed that for $x=1$ and $x=2$ the LHS gives $-10$ and $1$, respectively, so that the value of one real root has to be between $1$ and $2$. After few trials, we easily get $x=3/2$. The equation can then be rewritten as

$$\left(x-\frac 32\right)\left( 26x^2-96x+90\right)=0$$

from which we directly get that the other two roots are not real.

Since $x=j/k=3/2$, we can finally determine $j$ and $k$ by making the substitution $k=2j/3$ in the initial equations. For example, substituting in the equation $(j^2+9k^2)=135$, we have

$$j\left[j^2+9\left(\frac{2j}{3}\right)^2\right]=135$$ $$5j^3=135$$

and reminding that $j$ and $k$ are positive,

$$j=3$$

$$k=2$$

We can now conclude that

$$A=(3+2\sqrt{3})^3$$

so that the initial cubic root is

$$\sqrt[3]{5 + \frac{26\sqrt{3}}{9}}=\frac 13 \sqrt[3]{A}= \frac 13 \left(3+2\sqrt{3}\right)\\=1+\frac{2}{\sqrt{3}}$$

Again, it must be pointed out that this method works only in some cases (even when the rational $j$ and $k$ exist, the most important limiting step is the search of the rational root $x$, which as already stated can be very difficult).

Anatoly
  • 17,079
  • But, how do you get the simplified values of $a$ and $b$? You have to seek a rational root of a cubic equation by guessing at some point. – Oscar Lanzi Aug 16 '20 at 21:45
  • Thank for your comment. I edited my answer to better clarify this issue. – Anatoly Aug 17 '20 at 11:14
  • In the equation $\color{blue}{26}x^3-\color{brown}{135}x^2+\color{blue}{234}x-135=0$ we may render $\color{brown}{135}^2-(3)(\color{blue}{26})(\color{blue}{234})<0$, forcing the cubic function to be monotonic thus only one real root. A rational root must be the single real root between $1$ and $2$. – Oscar Lanzi Aug 17 '20 at 21:05
  • This is the most clear answer I've ever seen for this kind of problem.Thanks! – Arian Ghasemi Mar 05 '21 at 06:49
3

Apart from back-substituting $x=z+1$ to complete the solution process, you are not wrong. The casus irreducibilis is commonly described for cubic equations with three real roots, but a similar problem occurs when you have a rational root (and in this case it does not have to be all three roots). In effect, you can't simplify your radical expression to retrieve $z=2$ analytically; you have to guess the rational root beforehand (or make an equivalent guess involving another similarly structured cubic equation, as discussed in another answer).

When I put your expression for $z$ into a calculator I get $2.000000...$, which seems pretty close to your intended value of $z=2$.

Oscar Lanzi
  • 39,403