Real numbers are all crammed onto a one-dimensional line — you can only go left or right. That line is too narrow. Inventing $i$ amounts to bolting an "up-down" dimension onto the number line. A complex number $a+bi$ is no longer a point on a line but an arrow in a plane: walk $a$ across, $b$ up.
"Imaginary" is the worst naming in the history of mathematics. These numbers are neither illusory nor any less "real" than negatives — they simply live in two dimensions instead of one. Negatives were once derided as "absurd numbers"; irrationals once terrified the Pythagoreans. Every expansion of the number system, humanity first resisted, then accepted. Complex numbers are just the next stop on that road.
$a$ is the real part (horizontal), $b$ the imaginary part (vertical). The modulus $|z|=\sqrt{a^2+b^2}$ is the arrow's length; the argument $\arg z$ is its angle $\theta$ from the positive real axis. In short, a complex number = a length + a direction.
The real line has a "hole": $x^2+1=0$ has no solution. The complex plane fills it. The Fundamental Theorem of Algebra says any degree-$n$ polynomial has exactly $n$ roots over the complex numbers — not one more, not one fewer. The reals are incomplete; the complex field is algebraically closed — every polynomial equation is solvable. Just by stepping from one dimension to two, the number system suddenly becomes whole. That completeness is itself a structural beauty.
The Fourier transform breaks a signal into frequencies, each a complex number (carrying both amplitude and phase). Circuit impedance, control-system stability (which side of the plane the poles fall on decides whether a system blows up), quantum states — all live in the complex plane. In deep learning, FFT-accelerated convolution and complex-valued features handling phase all run on this same plane underneath.
Raising a number to an "imaginary" power — $e^{i\theta}$ — sounds like nonsense. Reframe it: $e^x$ describes continuous growth, where the rate of growth is proportional to your current size. So what is $e^{i\theta}$? Multiplying by $i$ means "rotate the growth direction by 90°." Now growth no longer shoots outward but stays forever perpendicular to your current position — which is precisely the definition of going in a circle. So $e^{i\theta}$ neither explodes nor decays; it just rotates at constant speed around the unit circle, with $\theta$ the radians turned.
$\theta$ is the angle (radians); the real part is $\cos\theta$, the imaginary part $\sin\theta$ — exactly the coordinates of that point on the unit circle. Setting $\theta=\pi$ (a half-turn to $-1$) gives the immortal Euler identity $e^{i\pi}+1=0$, threading together the five most important constants.
The exponential speaks of "growth," trigonometry of "oscillation" — two seemingly unrelated crafts. Euler's formula says they are the same rotation, projected onto two different directions. Analysis ($e$), geometry (the circle and $\pi$), and algebra ($i$) are welded together in a single line. Feynman called it "our jewel, the most remarkable formula in mathematics."
It is the soul of Fourier analysis — any signal decomposes into a sum of $e^{i\omega t}$ terms, the mathematical bedrock of audio compression, image processing, and 5G. AC power and quantum wavefunctions rely on it too. The Transformer's positional encoding uses $\sin/\cos$ to order tokens — essentially mapping each position to a different angle on the unit circle.
You already know the geometry of real multiplication: times $2$ stretches by double, times $-1$ "flips to the opposite side." But look again at "flipping to the opposite side" — it is really a rotation by 180°. So a 90° rotation is multiplication by which number? Squaring it should equal a "180° turn," i.e. equal $-1$. The number whose square is $-1$ is exactly $i$. So $i^2=-1$ stops being mysterious: rotate by 90° twice and you have rotated 180° — a flip — which is times $-1$. The whole secret of complex multiplication is one phrase: multiply the moduli, add the angles.
Multiplying two complex numbers: the lengths (moduli $r$) multiply, the directions (arguments $\theta$) add. Multiplication $=$ scaling $+$ rotation, one operation bundling two geometric acts.
"Multiplication" is arithmetic, "rotation" is geometry — complex numbers make them one and the same, filling in the gap between algebra and geometry. Many results that once required grinding through trig identities now come almost for free: expand both sides of $e^{i(\alpha+\beta)}=e^{i\alpha}e^{i\beta}$ and the angle-addition formulas for $\cos(\alpha+\beta)$ and $\sin(\alpha+\beta)$ drop out in one line. De Moivre's theorem, likewise.
2D rotation in computer graphics is just multiplying by a unit complex number. Its three-dimensional cousin, the quaternion, is standard in game engines, robot pose, drone flight control, and spacecraft attitude — cheaper and steadier than rotation matrices, and free of "gimbal lock." The phase rotations in digital modulation/demodulation are the same trick.
Classical physics gets by with real numbers: position, velocity, temperature are all real. But in the quantum world a particle's state is complex — a so-called probability amplitude. Why insist on complex numbers; why won't real probabilities do? Because quantum states interfere: the amplitudes of two paths add like water waves, sometimes reinforcing (brighter), sometimes cancelling (darker). Adding real numbers only ever grows the total; to express "cancellation" you need a phase — and phase is exactly the argument of a complex number. The ability to cancel is innate to complex numbers.
The wavefunction $\psi$ is complex-valued; the only probability $P$ we ever observe is its squared modulus. When two paths superpose, $\psi=\psi_1+\psi_2$, expanding $|\psi|^2$ throws off a cross "interference term." That $i$ on the left of the Schrödinger equation cannot be removed — it is the skeleton of the equation.
Here complex numbers are no longer a "convenient computational trick" but a physical necessity. In 2021, several experimental teams, building on a generalized Bell inequality, proved that a real-number-only quantum mechanics cannot reproduce real observations — nature "rejected" the real version. A number that Cardano treated as a monster in the 16th century, branded "imaginary," turns out to be the writing language of the universe's deepest layer. It is the ultimate case of mathematics "invented from nothing" yet striking physical reality dead-on.
The foundation of all quantum technology. A qubit's state is a complex vector; quantum gates are unitary transformations — modulus-preserving "high-dimensional rotations." Quantum computing, quantum chemistry, lasers, semiconductors are all built on complex numbers. Machine learning even has complex-valued neural networks, built to handle audio, radar, and MRI signals that carry phase by nature.