Trying to understand this derivation by Jean-Claude Arbaut, of why formula of quadratic form must hold also for complex variable with complex coefficents.

(derivation starts)

Complex resolution

Again, we start with the equation


Where $a\neq0$, and $a,b,c\in\Bbb C$.

Then, rewrite the equation


Now, a complex number $z$ has always a square root (two distinct ones if $z\neq0$, see for example here). We go on with a square root $\alpha$ of $b^2-4ac$, that is, $\alpha^2=b^2-4ac$, so that the equation can be written


It's a difference of squares, and we know that $A^2-B^2=(A+B)(A-B)$, so


So the roots are


Or if you prefer


Important note: there is no canonical way to write that $\sqrt{t}$ is a specific square root of $t\in\Bbb C$, hence this symbol is meaningless. However, since it appears with a $\pm$, we will always work with both roots, so using the symbol here is harmless. Nevertheless, it should be avoided.

(derivation ends)

First of all to get into rewritten form

$$ax^2+bx+c=0 \Rightarrow \left(x+\frac{b}{2a}\right)^2-\frac{b^2}{4a^2}+\frac{c}{a}=0$$ $$\Rightarrow \left(x+\frac{b}{2a}\right)^2-\frac{b^2}{4a^2}+\frac{4ac}{4a^2}$$

But now to me there seems to be 2 options. Either this step is not true or $a^2$ means here $aa$ instead of $aa^*$. Where $a^*$ is it's complex conjugate.

As $$\frac{a}{a^2}=\frac{a}{a^*a}=\frac{a}{a^*}\frac{a^*}{a^2}=\frac{1}{a^*}\ne \frac{1}{a}$$

What is the actual justification for this step?

As now if $a^2$ would mean $aa$ following step seems not to make any sense, if we assume the $b^2$ means $bb^*$. (I've before used this formula in physics, and correct answers have required $b^2=bb^*$, so that's why I assume this.)


Thus both of the ways I can think this with, seems to fail, so I assume there's something fundamental I've understood wrong now.

And also as last one the step of opening the difference of squares would require

$$\frac{\alpha}{2a}\frac{\alpha}{2a}=\frac{\alpha ^2}{4a^2}$$

Which to my knowledge implies $a^2=aa$ (again), but which seemed unconsistent with the step before.

  • $\begingroup$ $a^2$ always means $a\cdot a$. $a\cdot a^{\ast}$ (or more conventionally $a\cdot \overline{a}$) is $\lvert a\rvert^2$ - which only coincides with $a^2$ when $a$ is real. $\endgroup$ – Daniel Fischer Sep 6 '16 at 14:01
  • $\begingroup$ You say "Trying to understand (this) derivation" but there is nothing "behind" this "this"... $\endgroup$ – Jean Marie Sep 6 '16 at 14:07
  • $\begingroup$ @JeanMarie Whole derivation is in the post, separated by lines, could add link too though. $\endgroup$ – PyHop Sep 6 '16 at 14:28
  • $\begingroup$ @DanielFischer So the $b^2$ in the formula should actually also be treated as $bb$. Alright thanks, makes this consistent, and pushes problem elsewhere. $\endgroup$ – PyHop Sep 6 '16 at 14:36
  • $\begingroup$ All right, if nothing more can be found in the initial post... $\endgroup$ – Jean Marie Sep 6 '16 at 14:36

Seems unnecessarily complicated. For $z\in \mathbb C$ with $z\ne 0$ there are exactly 2 solutions $y_1,y_2$ to $y^2=z,$ with $y_1+y_2=0.$ And for $a\ne 0$ we have $u^2=z/4a^2 \iff u=y/2a$ where $y$ is a solution of $y^2=z.$... Also $y^2=0\iff y=0.$

So just take the real-quadratic-formula method verbatim: For $a\ne 0,$ $$ax^2+bx+c=0\iff (x+b/2a)^2=(b^2-4ac)/4a^2\iff x\in \{-b/2a+ y/2a: y\in S\}$$ where $S=\{y: y^2=b^2-4ac\}.$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.