1

Given a quadratic equation $\quad ax^2+bx+c=0\quad$ I'm looking for a trig equivalent to $$x = \frac{-b \pm \sqrt{b^2 - 4ac}}{2a}$$

the way I found one with help for a cubic equation $\quad mn^3-m^3n+D=0\quad$ where the final formula (that works like a charm) is:

$$n_0=2\sqrt{\frac{m^2}{3}}\cos\biggl({\biggl(\frac{1}{3}\biggr)\arccos{\biggl(-\frac{3\sqrt{3}D}{2m^4}\biggr)}\biggr)}$$ $$n_1=2\sqrt{\frac{m^2}{3}}\cos\biggl({\biggl(\frac{1}{3}\biggr)\arccos{\biggl(\frac{3\sqrt{3}D}{2m^4}\biggr)}\biggr)}$$ $$n_2=n_1-n_0$$

where $$\lfloor\sqrt[4]{D}\rfloor\le m\le \lceil\sqrt[3]{D}\space \rceil$$

Has anyone ever seen such a thing? Or, how would I go about constructing such a thing?

poetasis
  • 6,338
  • When you depress a quadratic equation you get an elementary one. – egreg Jan 16 '20 at 18:36
  • @egreg I never understood how to depress an equation. Can you walk me though it? And, how will that help me find a trig formula? – poetasis Jan 16 '20 at 18:41
  • 1
    You depress the equation $x^n+a_{n-1}x^{n-1}+\dots+a_0=0$ by the substitution $x=y-a_{n-1}/n$, which makes the sum of the roots $0$. – egreg Jan 16 '20 at 20:24
  • @egreg If I can do this with my limited knowledge, how does this help me find a trig solution? And, what is an elementary equation? It seems intuitive as being one of the $1^{st}$ degree but I want to make sure. – poetasis Jan 17 '20 at 17:03
  • What trig solution? If you have the equation $y^2=a^2$, there's not much to do. – egreg Jan 17 '20 at 17:06
  • @egreg I'm trying to compare the approach that can be taken for a quadratic to the one shown me for a cubic in hopes that I can, perhaps, understand how such could be applied to solve a quintic. – poetasis Jan 17 '20 at 17:13
  • 1
    Poetasis, is this what you want? I think that's the trig method for solving a quadratic. – Jyrki Lahtonen Jan 18 '20 at 08:24

1 Answers1

1

AFAICT the trig method for solving low degree polynomial relies on the properties of Chebyshev polynomials. Another starting point is that the solutions of $$ \cos n\alpha=\cos \beta $$ are $$ \alpha=\frac1n(\pm \beta+\frac k{2\pi})\tag{1} $$ with $k\in\Bbb{Z}$. This is an immediate consequence of $\cos\alpha=\cos\beta\Leftrightarrow \alpha=\pm\beta+k\cdot2\pi$.

Chebyshev polynomials come into play via the relation $$ T_n(\cos x)=\cos nx. $$

Let's start with $n=3$. Then $T_3(t)=4t^3-3t$, all according to the familiar triplication formula $$\cos 3\alpha=4\cos^3\alpha-3\cos\alpha.$$ If you can write a cubic equation in the form $$4x^3-3x=y,\tag{2}$$ then the substitutions $x=\cos\alpha$, $y=\cos\beta$ turn $(2)$ into $\cos3\alpha=\cos\beta$, and we are in business. It follows that $$ x=\cos\left(\pm\frac13\left(\arccos y+k\cdot2\pi\right)\right). $$ Because cosine is even, we can ignore $\pm$. Because cosine has period $2\pi$, it suffices to only include $k=0,1,2$.

The upshot is that a general cubic can be brought into the form $(2)$ by a linear subsititution. You first need to depress the quadratic term, and then you need to arrange the coefficients of the cubic and the linear terms have that $4:(-3)$ ratio. For example, with your cubic $$m^3n-mn^3=D$$ we easily see that the substitution $n=2mx/\sqrt3$ brings it to the form $$ \frac{8m^4}{3\sqrt3}x^3-\frac{2m^4}{\sqrt3}=D, $$ which is a scalar multiple of $$ 4x^3-3x=\frac{3\sqrt3 D}{2m^4}. $$

When $n=2$ we have $T_2(t)=2t^2-1$ meaning that the solutions of $$ 2x^2-1=y\tag{3} $$ are $$ x=\cos\left(\frac12\left(\arccos y+k\cdot2\pi\right)\right) $$ with $k=0,1$. Again, a linear substitution brings a general quadratic to the form $(2)$, first be depressing the linear term, and then scaling the variable linearly. In the case of a quadratic this is not usually done, because we have the simple quadratic formula.

Unfortunately with $n>3$ we lose the key ability to bring a general polynomial into the desired form involving a Chebyshev polynomial. We can find the solutions to a quartic of the form $$ 8x^4-8x^2+1=y $$ as $$ x=\cos\left(\frac14\left(\arccos y+k\cdot2\pi\right)\right) $$ with $k=0,1,2,3$, but the trickery with linear substitutions does not work (at least not alone). We can depress the cubic term, and fix the ratio of the coefficients of the quadratic and quartic terms, but we cannot kill the linear term. With $n=5$ and higher it becomes worse.

Jyrki Lahtonen
  • 133,153