1

The equation $ -4x^3+3x-v=0 $ has three solutions, according to WolframAlpha: https://www.wolframalpha.com/input?i=-4x%5E3+%2B+0x%5E2+%2B3x+-v+%3D0

$$x = \frac{1}{2} \left((\sqrt{v^2 - 1} - v)^{1/3} + \left(\frac{1}{\sqrt{v^2 - 1} - v}\right)^{1/3}\right)$$

$$x = \frac{1}{2} \left(\frac{-1+i\sqrt{3}}{2}(\sqrt{v^2 - 1} - v)^{1/3} - \frac{1+i\sqrt{3}}{2}\left(\frac{1}{\sqrt{v^2 - 1} - v}\right)^{1/3}\right)$$

$$x = \frac{1}{2} \left(-\frac{1+i\sqrt{3}}{2}(\sqrt{v^2 - 1} - v)^{1/3} + \frac{-1+i\sqrt{3}}{2}\left(\frac{1}{\sqrt{v^2 - 1} - v}\right)^{1/3}\right)$$

Leb be a $ \theta . $ We know the equation holds for $x=\sin\frac{\theta}{3}, v=\sin\theta .$ My question is, which of these three equations we have to choose so that $ \frac{\arcsin{v}}{3} =\arcsin x . $ It depends on $v$? $v$ is a real value between $ [-1,1]. $

  • 1
    The other solutions are $$x=\sin\frac{\theta\pm2\pi}3.$$ Why cannot you use those as solutions of the cubic? – Jyrki Lahtonen Sep 01 '24 at 16:44
  • Because if we choose $\theta=0^{\circ }$ , then none of the three work, exactly. How to tell a computer to show an exact expression of $ \sin(\theta /3) $ in terms of $ \sin (\theta) $ ? – Cătălin Bordea Sep 01 '24 at 18:02
  • Can you constrain $\theta$ to an interval? Otherwise you have the problem that $\theta$ and $\theta\pm2\pi$ give the same value for $\sin\theta$, but different values for $\sin(\theta/3)$. – Jyrki Lahtonen Sep 01 '24 at 18:07
  • $\theta\in (-90^{\circ} ,+90^{\circ}] . $ In my previous comment I made a mistake, sorry. The first of the general equations works, indeed, for $ \theta=0^{\circ} . $ However, if we change, say, $ \theta =45^{\circ} , $ then, the same equation doesn't hold. LHS would be $\sin 15^{\circ} , $ while RHS would be $ 1/\sqrt{2} $ . So, Cardano is useless? – Cătălin Bordea Sep 01 '24 at 18:44
  • At least for me Mathematica gives values in the range $[-1/2,1/2]$ using the bottom formula $$\frac{1}{2} \left(\frac{1}{2} \left(-1+i \sqrt{3}\right) \sqrt[3]{\frac{1}{-v+i \sqrt{1-v^2}}}+\frac{1}{2} \left(-1-i \sqrt{3}\right) \sqrt[3]{-v+i \sqrt{1-v^2}}\right)$$ when $v$ is in the interval $[-1,1]$. Is that what you want? – Jyrki Lahtonen Sep 01 '24 at 19:37
  • Thank you for your info. What I didn't realize is that Cardano's cube root is multi-valued. What I'm doing is some research on the complex unit circle, to see whether I can describe complex root functions as single valued functions, bypassing the notion of radians. I hope that's possible. We'll see. – Cătălin Bordea Sep 02 '24 at 09:43

2 Answers2

2

If you use the trigonometric method, the solutions are much simpler

$$x_0=\cos \left(\frac{1}{3} \cos^{-1}(-v)\right)$$ $$x_1=-\sin \left(\frac{\pi }{6}-\frac{1}{3} \cos ^{-1}(-v)\right)$$ $$x_2=-\sin \left(\frac{\pi }{6}+\frac{1}{3} \cos ^{-1}(-v)\right)$$

1

As noted in comments and in another answer, a nonzero complex number has three different cube roots, and simpler expressions for roots of the equation $-4\cdot x^3+3\cdot x-v = 0$ with $-1\le v\le 1$ can be obtained using trigonometric functions.

Verifying the expressions of roots

Let $\zeta = \frac{-1+\imath\cdot\sqrt{3}}{2}$; then $\zeta^2 = -\frac{1+\imath\cdot\sqrt{3}}{2}$, $\zeta^2+\zeta+1 = 0$, $\zeta^3 = 1$.

Let $\gamma = \sqrt{v^2-1}$ (there are two possible choices of $\gamma$) and $\alpha = \left(-v+\gamma\right)^{\frac{1}{3}}$ (there are three possible choices of $\alpha$); the choices of $\gamma$, $\alpha$ are made once for all occurrences. Then $\alpha^3+v = \gamma$, $\left(\alpha^3+v\right)^2 = \gamma^2 = v^2-1$, $\alpha^6+2\cdot v\cdot\alpha^3+1 = 0$.

Denote $\beta = \frac{1}{\alpha}$, then $$ \beta = \frac{1}{\alpha} = \frac{1-\left(\alpha^6+2\cdot v\cdot\alpha^3+1\right)}{\alpha} = -\left(\alpha^5+2\cdot v\cdot\alpha^2\right) = -\left(\gamma+v\right)\cdot\alpha^2 $$ and one can verify that $\beta^3+v = -\gamma$; that is, choosing another sign for $\gamma$ exchanges possible values of $\alpha = \left(-v+\sqrt{v^2-1}\right)^{\frac{1}{3}}$ with corresponding values of $\beta = \left(-v-\sqrt{v^2-1}\right)^{\frac{1}{3}}$.

Now the expressions in the question can be rewritten as follows: $$ \begin{align} x_1 = \frac{1}{2}\cdot\left(\alpha+\beta\right) &= \frac{1}{2}\cdot\left(\alpha-2\cdot v\cdot\alpha^2-\alpha^5\right) \\ x_2 = \frac{1}{2}\cdot\left(\zeta\cdot\alpha+\zeta^2\cdot\beta\right) &= \frac{1}{2}\cdot\left(\zeta\cdot\alpha-\zeta^2\cdot\left(2\cdot v\cdot\alpha^2+\alpha^5\right)\right) \\ x_3 = \frac{1}{2}\cdot\left(\zeta^2\cdot\alpha+\zeta\cdot\beta\right) &= \frac{1}{2}\cdot\left(\zeta^2\cdot\alpha-\zeta\cdot\left(2\cdot v\cdot\alpha^2+\alpha^5\right)\right) \end{align} $$

By directly substituting each of these three expressions, expanding and simplifying through repeated use of the identities $\zeta^2 = -\left(\zeta+1\right)$ and $\alpha^6 = -\left(2\cdot v\cdot\alpha^3+1\right)$, one can verify that $x_1$, $x_2$, $x_3$ are roots of the cubic equation $-4\cdot x^3+3\cdot x-v = 0$. It is also possible to check that expressions for $x_1$, $x_2$, $x_3$ refer to three different roots (when the cubic equation has three distinct complex roots).

A solution using trigonometric or hyperbolic functions

The "trigonometric" approach can be made more general by choosing between trigonometric and hyperbolic functions.

A cubic equation of the form $z^3+3\cdot p\cdot z^2+6\cdot q\cdot z+2\cdot r = 0$ with $p,q,r\in\mathbb{R}$ can be transformed into one of the following forms:

  • If $p^2=2\cdot q$ then $z=t-p$, $t^3=w=p^3-2\cdot r$: this case reduces to finding cube roots of $w$.
  • If $p^2>2\cdot q$ then $z=2\cdot k\cdot t-p$, $4\cdot t^3-3\cdot t=w=\frac{3\cdot p\cdot q-p^3-r}{k^3}$ where $k=\sigma\cdot\sqrt{p^2-2\cdot q}$, $\sigma^2=1$.
  • If $p^2<2\cdot q$ then $z=2\cdot k\cdot t-p$, $4\cdot t^3+3\cdot t=w=\frac{3\cdot p\cdot q-p^3-r}{k^3}$ where $k=\sigma\cdot\sqrt{2\cdot q-p^2}$, $\sigma^2=1$.

To solve the second and third cases above, one can check the sign of the term $\pm 3\cdot t$ and the absolute value of $w$ and choose an applicable identity from the following list:

  • $t=\cos\eta$, $4\cdot t^3-3\cdot t=w=\cos\left(3\cdot\eta\right)$, $-1\le w\le 1$
  • $t=\sin\eta$, $4\cdot t^3-3\cdot t=w=-\sin\left(3\cdot\eta\right)$, $-1\le w\le 1$
  • $t=\cosh\eta$, $4\cdot t^3-3\cdot t=w=\cosh\left(3\cdot\eta\right)\ge 1$
  • $t=\sinh\eta$, $4\cdot t^3+3\cdot t=w=\sinh\left(3\cdot\eta\right)\in\mathbb{R}$

Plotted in GeoGebra:

trig_hyp

mezzoctane
  • 2,241
  • 2
  • 8
  • 21