Square Root of Complex Number

The square root of a real number a0 is a number that gives a when multiplied with itself. For example, 4=2 because 22=4. However, we also have (2)2=4, so it seems like 4=2. To avoid getting 2=2, we require that a is not negative. Then all numbers that give a when multiplied with themselves are ±a, which is why ± appears in front of square root so often.

Let's generalize this to complex numbers.

With angles and lengths

We can define the square root of any complex number z by requiring (z)2=z, just like with real numbers. Because of how multiplication adds angles and multiplies lengths, this means that |z|2=|z|,2(angle of z)=angle of z. Solving for the length and angle of z, we get |z|=|z|,angle of z=angle of z2. As usual, if z=0, the angle equation is not needed because the length formula gives 0=0. Any nonzero complex number z0 has infinitely many angles, differing from each other by 2π, which can be written as all angles of z=arg(z)+k2π, where k is any integer. Then we have angle of z=arg(z)+k2π2=arg(z)2+kπ. If k is even, then +kπ adds some number of full turns to the angle and can be ignored. If k is odd, then +kπ adds some number of full turns, and also adds a half-turn π which means flipping to the opposite side of 0. So, we get two possible values of z, with equal lengths but on opposite sides of 0. In the below picture, the blue dots are the possible values of z.

The only thing left to do is to decide which of these should be called z. Because of the flipping, the other one will then be z, and ±z will appear commonly just like with real numbers.

Let z be any complex number. Then ±z are the only numbers that give z when multiplied with themselves.

Note that now we can put any complex number inside square root, including negative real numbers.

Examples:

The below picture shows how distinguisihing z and z works. Blue complex numbers can be written as z, with some complex number z, and red numbers are the corresponding z numbers.

Using the convention π<arg(z)π, the distinguishing of ±z becomes arg(z)(π2,π2]andarg(z)(π2,π2]. Now we get π<2arg(z)π. Because angles get added when multiplying complex numbers, 2arg(z) is an angle of (z)2=z. Another angle of z is arg(z), for which we also have π<arg(z)π. Because only one angle of z is between π and π, we must have arg(z)=2arg(z).

Using the convention π<arg(z)π, we have arg(z)=arg(z)2.

Broken rules

The rules zw=zwandzw=zw do not work for all complex numbers z and w. To show that the first rule doesn't work, we can choose z=w=1. Then we get 11=(1)2=1 and (1)(1)=1=1. For the second rule, we can choose z=1 and w=1. We get 11=1=i and 11=1i=ii(i)=i(1)=i. From these examples, it seems like the rules might work with additional ± signs, as in zw=±zwandzw=±zw. This is indeed true. To see why, notice that (zw)2=(z)2(w)2=zw. The only two numbers that give zw when squared are zw and zw, and because zw is one of them, we must have zw=±zw, or written slightly differently, zw=±zw. We can similarly derive the rule for division.

For all complex numbers z and w, we have zw=±zw. If w0, then zw=±zw.

For any positive real number p, we have angle of zp=angle of z, and because the angle of square root depends only on the angle of the complex number inside, we get angle of zp=angle of z. On the other hand, because p is also positive, we also have angle of zp=angle of z. By combining these results, we get angle of zp=angle of zp, so the ± in zp=±zp must be +, and it actually works without ±. That also works for p=0, because then both sides are zero.

If p is a real number and p0, then zp=zp.

Using this, we can for example calculate 5=(1)5=15=i5. We also get zp=z1p=z1p=z1p=zp for any positive p, where we used a/b=a/b with positive real numbers a=1 and b=p. We obviously can't allow p=0.

If p is a real number and p>0, then zp=zp.

Without angles and lengths

We find a formula for a+bi that doesn't require calculating the angle of a+bi. Let x+yi=a+bi. Then we have a+bi=(x+yi)2=x2+2xyi+i2y2=x2y2+2xyi. By comparing real and imaginary parts, we get {x2y2=a2xy=b. We already know how to calculate the square root of a real number (both negative and nonnegative), so we can assume b0. From the second equation, we see that then x and y are nonzero too. So, we can divide by x and we get y=b2x from the second equation. Let's plug that into the first equation: x2(b2x)2=ax2b24x2=a(x2)2b24=ax2(x2)2ax2b24=0 By applying the quadratic formula (TODO) with x2 instead of x, we get x2=(a)±(a)24(b24)2=a±a2+b22=a±r2, where r=a2+b2 is the length of a+bi. Because b is a nonzero real number, we have b2>0 and so r=a2+b2>a2=|a|. This means that ar is negative, regardless of whether a is positive, negative or zero. Because x2 is not negative, we must choose + and we get x2=r+a2. Because the point corresponding to the square root of a complex number cannot be on the left side of the imaginary axis, we must have x0. So, we get x=r+a2. We actually have x>0 because 2xy0. Now y=b2x has the same sign as b; it's positive if b is positive, and negative if b is negative. To actually find y, we can use x2y2=a, and we get y2=x2a=r+a2a=r+a2a2=ra2. So, we have y=±ra2, where ± is + when b is positive and when b is negative.

Let a+bi be a complex number with b0. Then a+bi=r+a2±ira2, where r=|a+bi|=a2+b2, and ± is + when b is positive and when b is negative.

If b=0, then we have r=|a| and the right side of the formula becomes |a|+a2±i|a|a2. If a is positive or zero, this is a±0, and if a is negative, this is 0±i|a|. So, if we choose + whenever b=0, this result also works in that case.

Example: sin2(θ) and cos2(θ)

Let θ be any angle, and let z be a complex number on the origin-centered unit circle with angle θ. Like in the derivation of de Moivre's formula, z2 is also on the unit circle with angle 2θ, and we have z=cos(θ)+isin(θ),z2=cos(2θ)+isin(2θ).

Because z=±z2, we have cos(θ)+isin(θ)=±cos(2θ)+isin(2θ). Now we can use the formula for a+bi with a=cos(2θ) and b=sin(2θ). Because z2 is on the unit circle, its length is r=a2+b2=1, and we get cos(θ)+isin(θ)=±(1+cos(2θ)2±i1cos(2θ)2). By comparing real and imaginary parts, we get cos(θ)=±1+cos(2θ)2sin(θ)=±1cos(2θ)2, where we combined the two ± signs into another ± sign for the imaginary part of the right side. Finally, to get rid of all ± signs and square roots, we can square both sides to get cos2(θ)=1+cos(2θ)2sin2(θ)=1cos(2θ)2.

For all angles θ, we have cos2(θ)=1+cos(2θ)2,sin2(θ)=1cos(2θ)2.