Let $n>0$ be an even number, and let $z$ be an $n$'th root of a real number. Is $\bar z$ also an $n$'th root of this number?

My answer is yes. The way I solved this was to consider a complex number $z = a+bi$ on the (polar) form $z = re^{i\theta}$.

Solving $z^n = re^{i\theta}$, I get $z = \sqrt[n] r e^{{(i\theta+2\pi m)}/n}$, $0 \leq m \leq n-1$.

Since $n>0$ is an even number, we will always have an even number of solutions. If we then draw the solutions in the complex plane, we will always have that for each solution that $z$ that has an angle $\phi=\theta$, we will always find $\bar z$, also a solution with $\phi=-\theta$ on the complex plane.

However, my answer is more a geometric interpreration/realization than a rigorious algebraic proof.

I wanted to try to prove this in a more general algebraic form: My attempt was to assume $z=a+bi$ is an $n$'th root of a real number. We now want to show that $z=a-bi$ is also an $n$'th root of the same real number.

I tried setting up $(a+bi)^2 = a^2-b^2+2iab$ and $(a-bi)^2 = a^2-b^2-2iab$, noticing that the real part of these numbers are always the same, the imaginary part is different in sign.

Now it's easy to realize that $(2iab)^n=(-2iab)^n$ for all $n>0$ given that $n$ is even.

But I have not proved that if we have a complex number $z = a+bi$ (being an $n$'th root of a real number, there also exists a $\bar z = a-bi$ such that $\bar z$ is also an $n$'th root of the same real number.

If this were the case, shouldn't $(a+bi)^n=(a-bi)^n$ for all even $n>0$? This is obviously not the case, I can find $a$ and $b$ not satisfying this equation.

I guess I'm confused about what I really have to do to prove this.

  • 3
    $\begingroup$ A strongest result holds: Complex conjugate root theorem says that if $z$ is a solution for a given real polynomial $P$ then $\bar{z}$ is also a solution for $P$. In your case $P= x^n=a$ with a real number. $\endgroup$ – ALG Oct 18 '18 at 14:50
  • 6
    $\begingroup$ COnjugation is a "field automorphism", which is fancy-speak in particular for $\overline{z+w}=\overline z+\overline w$ and $\overline{ z\cdot w}=\overline z\cdot \overline w$. Therefore $\overline z^n=\overline{z^n}$ and $z^n=a$ implies $\overline z^n=\overline a=a$ $\endgroup$ – Hagen von Eitzen Oct 18 '18 at 14:54
  • 1
    $\begingroup$ Why does $n$ have to be even? $\endgroup$ – Doug M Dec 19 '18 at 21:01
  • $\begingroup$ No polar form, no Cartesian form, no polynomials, no parity of $n$, just the fact that, for every complex $z$, $$\overline{z^n}=\left(\bar z\right)^n$$ hence, if $z^n=u$ with $u$ real then $\bar u=u$ hence $$\left(\bar z\right)^n=\overline{z^n}=\bar u=u$$ $\endgroup$ – Did Dec 19 '18 at 21:16

It's really easy to prove it with the comments, but I don't think you're allowed to use these stronger results to resolve your problem.

So I think you should stay with your initial idea of using $z = re^{i\theta}$.

So, suppose:

  1. $z = re^{i\theta}$
  2. $\exists n \in \mathbb Z, z^n = r^ne^{ni\theta} = x$

Since $x^n \in \mathbb R$, you can conclude that $n\theta \equiv 0 \pmod \pi$

You also have:

$\bar{z} = re^{-i\theta}$


$\bar{z}^n = r^ne^{-ni\theta}$

From there, you can use the parity of n to prove $\frac{z^n}{\bar{z}^n}=\frac{x}{\bar{z}^n}=1$

Edit: As from [Hagen von Eitzen] comment, you don't even need n to be even

  • 1
    $\begingroup$ Polar forms are essentially a pain here. The "strongest" result one needs to solve this is that $\overline{zw}=\bar z\bar w$, and if the OP does not know that... well, then they should learn and understand and be able to use it, asap. $\endgroup$ – Did Dec 19 '18 at 21:19

Lemma: $(\overline z)^{2k} = \overline {(z^{2k})}$

Pf: If $z = a+bi$ then $(a+bi)^{2k} = \sum\limits_{j=0}^{2k}{2k \choose j}a^{j}b^{2k-j}i^{2k-j}$.

Now if $j$ is even, then $2k-j$ is even, and $i^{2k-j}$ is real, and ${2k \choose j}a^{j}b^{2k-j}i^{2k-j}$ is real.

If $j$ odd, then $2k-j$ is odd, and $i^{2k-j}$ is purely imaginary and ${2k \choose j}a^{j}b^{2k-j}i^{2k-j}$ is purely imaginary.

So $(a+b)^{2k} = \sum_{j=0; jeven}^{2k}({2k \choose j}a^{j}b^{2k-j}i^{2k-j}) + i\sum_{j=0; j odd}^{2k}({2k \choose j}a^{j}b^{2k-j}i^{2k-j-1})$

And so $\overline {(z^{2k})} = \sum_{j=0; jeven}^{2k}({2k \choose j}a^{j}b^{2k-j}i^{2k-j}) - i\sum_{j=0; j odd}^{2k}({2k \choose j}a^{j}b^{2k-j}i^{2k-j-1})$

Now if we expand out $(\overline z)^{2k}=(a-bi)^{2k}=\sum\limits_{j=0}^{2k}(-1)^j{2k \choose j}a^{j}b^{2k-j}i^{2k-j}$

Now $(-1)^j$ is $1$ if even and $-1$ if odd and $i^{2k-j}$ is real if even and imaginary if odd so:

$(\overline z)^{2k}=(a-bi)^{2k}=\sum_{j=0; jeven}^{2k}({2k \choose j}a^{j}b^{2k-j}i^{2k-j}) - i\sum_{j=0; j odd}^{2k}({2k \choose j}a^{j}b^{2k-j}i^{2k-j-1})=$



And that's it really. If $z$ is an $n$th root then $z^n = m\in \mathbb R$ so $z^n = m = m + 0i = m - 0i = \overline m = \overline {z^n} = (\overline z)^n$. So $\overline z$ is an $n$th root of $m$.

..... o Or, if it's not to geometric:

If $z = re^{i\theta}$ then $\overline z = re^{-i\theta}$ and and $z^n = r^ne^{in\theta}$ and $\overline z^n = r^ne^{-in\theta}$

If $z^n \in \mathbb R$ then $n\theta = k\pi $ for some integer $k$. And so $-n\theta = -k\pi$. Now $-k\pi \equiv k\pi \pmod {2\pi}$ so $(\overline z)^n = z^n$.


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.