4

In deriving his formula, Cardano arrives at the equation $y^3+py+q=0$. By substituting $y=\sqrt[3]{u}+\sqrt[3]{v}$, he gets the equation $(u+v+q)+(\sqrt[3]{u}\sqrt[3]{v})(3\sqrt[3]{u} \sqrt[3]{v} +p)=0$. Then he imposes that $u+v+q=0$ and $3\sqrt[3]{u} \sqrt[3]{v} +p=0$.

He "imposes".

My question: Why could he impose? Why isn't there any possible $(u, v)$ out there such that $u+v+q\neq0$ and $3\sqrt[3]{u} \sqrt[3]{v} +p\neq0$ but still $(u+v+q)+(\sqrt[3]{u}\sqrt[3]{v})(3\sqrt[3]{u} \sqrt[3]{v} +p)=0$?

Please help if you could-many thanks in advance!

Dick Grayson
  • 1,529
  • Like... it is just an impose... surely there will be infinite pairs of $(u,v)$ because of there is infinite number of ways to split a number into two, but he is going to find a special one – JetfiRex Jan 25 '22 at 01:36
  • 1
    Tthere exist $u,v$ such that $y = \sqrt [3] u+\sqrt [3] v$ and $u + v + s = 0$ and $3uv + p = 0.$ This does not mean that every choice of $u,v$ satisfying the first equation will satisfy the other two. – user317176 Jan 25 '22 at 01:40
  • 1
    @JetfiRexSo... are you suggesting that there could potentially be other solutions to a cubic equation that Cardano's formula fails to identify? – Dick Grayson Jan 25 '22 at 01:40
  • @DougM I see. I believe that you answered my question to JetfiRex. Thank you! – Dick Grayson Jan 25 '22 at 01:43
  • 1
    Are you satisfied with the accepted answer? I don't think that it answers your question. – José Carlos Santos Jun 02 '23 at 08:44
  • @JoséCarlosSantos The answer had already been alluded to in the comments, but I believe that Lee Mosher did answer the question. – Dick Grayson Jun 02 '23 at 14:26
  • 2
    It seemed to me that your question was about why every solution of the equation $y^3+py+q=0$ must be of the form $\sqrt[3]u+\sqrt[3]v$ for some complex numbers $u$ and $v$ such that $u+v+q=0$ and that $3\sqrt[3]u\sqrt[3]v+p=0$ and that the answer was in the reverse direction: whenever $u+v+q=0$ and $3\sqrt[3]u\sqrt[3]v+p=0$, then $u+v$ is a solution. I suppose that I misunderstood you. – José Carlos Santos Jun 02 '23 at 14:35
  • @JoséCarlosSantos No you didn't: on a second look, I realized that was indeed my confusion. Sorry the post has been a bit old. That said, I was reassured by the comments that the $u$ and $v$ so determined by the said imposition might not be ALL the solutions, which satisfied me. The answer, as you pointed out, did not quite address that confusion. – Dick Grayson Jun 02 '23 at 15:19
  • 1
    Yes, you were told that in the comments. But that is wrong. It is indeed true that if $y$ is such that $y^3+py+q=0$, then there are numbers $u$ and $v$ such that $y=u+v$, that $u^3+v^3+q=0$, and that $3uv+p=0$. If this still interests you, I can explain why in an answer. – José Carlos Santos Jun 02 '23 at 15:23
  • @JoséCarlosSantos I see-thank you for letting me know! And please do-I will deeply appreciate it. – Dick Grayson Jun 02 '23 at 15:25
  • 1
    See also https://math.stackexchange.com/a/1320330/589 – lhf Jun 02 '23 at 17:20
  • 1
    See also my answer here. – dxiv Jun 02 '23 at 20:17
  • 1

2 Answers2

5

If $p,q\in\mathbb C$, I will prove that any root of the polynomial$$x^3+px+q,\label{f}\tag1$$can be written as $u+v$, where $u,v\in\mathbb C$ are such that $u^3+v^3+q=0$ and that $3uv+p=0$, thereby proving that Cardano's formula always provides all roots of \eqref{f}.

Let $\omega=-\frac12+\frac{\sqrt3}2i$ and note that the cube roots of $1$ are $1$, $\omega$ and $\omega^2$. Now, let $u$ be a cube root of $-\frac q2+\sqrt{\frac{q^2}4+\frac{p^3}{27}}$ and let $v$ be a cube root of $-\frac q2-\sqrt{\frac{q^2}4+\frac{p^3}{27}}$ (and note that this implies that $u^3+v^3+q=0$), chosen so that $3uv=-p$. This can always be done: if $u$ is any cube root of $-\frac q2+\sqrt{\frac{q^2}4+\frac{p^3}{27}}$ and $v$ is any cube root of $-\frac q2-\sqrt{\frac{q^2}4+\frac{p^3}{27}}$, then\begin{align}u^3v^3&=\left(-\frac q2+\sqrt{\frac{q^2}4+\frac{p^3}{27}}\right)\left(-\frac q2-\sqrt{\frac{q^2}4+\frac{p^3}{27}}\right)\\&=\frac{q^2}4-\left(\frac{q^2}4+\frac{p^3}{27}\right)\\&=\left(-\frac p3\right)^3,\end{align}and therefore $uv$ is one of the numbers $-\frac p3$, $-\frac p3\omega$ or $-\frac p3\omega^2$. If it is equal to $-\frac p3$, you are done. If $uv=-\frac p3\omega$, then use $\omega^2v$ instead of $v$ (note that $\left(\omega^2v\right)^3=v^3$) and you are done. Finally, if $uv=-\frac p3\omega^2$, then use $\omega v$ instead of $v$.

So, now you have these numbers $u$ and $v$ such that$$u^3=-\frac q2+\sqrt{\frac{q^2}4+\frac{p^3}{27}}\text{, that }v^3=-\frac q2-\sqrt{\frac{q^2}4+\frac{p^3}{27}},$$and that $3uv=-p$. Therefore, the number $s_1=u+v$ is a root of \eqref{f}. By the same argument, so are the numbers $s_2=\omega u+\omega^2v$ and $s_3=\omega^2u+\omega v$. So, if the numbers $s_1$, $s_2$ and $s_3$ are three distinct numbers, then, since we are dealing with a cubic polynomial here, there can be no more roots, and it is then proved that every root is given by Cardano's formula.

What if the numbers $s_1$, $s_2$ and $s_3$ are not distinct? Suppose, say, that $s_2=s_3$. We have\begin{align}s_2=s_3&\iff\omega u+\omega^2v=\omega^2u+\omega v\\&\iff u+\omega v=\omega u+v\\&\iff(1-\omega)u=(1-\omega)v\\&\iff u=v.\end{align}There are now two possibilities to be considered: $u=0$ and $u\ne0$.

If $u=0$, then, since $v=u$, $v=0$ too. So, $p=-3uv=0$ and $q=-u^3-v^3=0$. Therefore, the only root of \eqref{f} is $0$, but the only number provided by Cardano's formula in this case is also $0$.

Finally, assume that $u\ne0$. Then $s_1=u+v=2u$ and$$s_2=\omega u+\omega^2v=\left(\omega+\omega^2\right)u=-u.$$Since both $2u$ and $-u$ are roots of \eqref{f}, and $2u\ne-u$, \eqref{f} can be written as$$(x-2u)(x+u)(x-r)\label{g}\tag2$$for some $r\in\Bbb C$. But the coefficient of $x^2$ in \eqref{g} is $-u-r$, whereas the coefficient of $x^2$ in \eqref{f} is $0$, and therefore $r=-u$; in other words, the only roots of \eqref{f} are $2u$ and $-u$. So, again, there are no other roots of \eqref{f} besides those provided by Cardano's formula.

3

Think of it in words like this: "Let's suppose that $y$ is a sum of two numbers whose cubes sum to $-p$ and whose product is $-\frac{p}{3}$. Can we find those two numbers?" That's all that is going on.

More formally the $u$ and $v$ are free variables that have been introduced into the problem, in order to aid in solving for $y$. So one is free to impose any conditions on them that one wants to impose; hopefully, with a sufficiently clever set of conditions, one can solve for those free variables and from that compute $y$.

Lee Mosher
  • 135,265