Why quantum needs complex numbers
Every quantum speedup depends on amplitudes that can cancel each other out precisely. Real numbers can only cancel in one specific way. This lesson walks you through an experiment where you discover that limit yourself, and then see what a second dimension makes possible.
One secret behind every quantum speedup
Quantum computers can do some things dramatically faster than ordinary computers. The reason always comes down to the same phenomenon: interference.
Here is what interference means. In a quantum system, every possible answer to a calculation has a number attached to it called an amplitude. Amplitudes can do something that ordinary probabilities cannot: they can cancel each other out completely. A quantum computer is designed so that wrong answers cancel away and the right answer survives. That is how it homes in on the correct result.
This raises a quiet but important question: what kind of number does an amplitude need to be? Not every number system can produce the right kind of cancellation. In this lesson you will discover the gap yourself by running a small experiment. No prior knowledge is needed beyond what you have already read.
You will not be told the answer up front. Instead, you will try to produce perfect cancellation using only ordinary numbers, hit a wall, and then unlock a second dimension to see what becomes possible. By the end, complex numbers will not feel like a rule someone imposed on you. They will feel like something you genuinely needed.
Amplitudes have two parts, and real numbers only carry one
Think of an amplitude as an arrow. Every arrow has two things: its length (how large it is, which physicists call the magnitude) and the direction it points (called the phase). These two properties are completely independent. Changing one does not change the other.
When you measure a quantum system, the probabilities you observe come only from the length of the amplitude:
$$\text{probability} = |\text{amplitude}|^2$$
$$|\alpha|^2 + |\beta|^2 = 1$$
The direction (the phase) is invisible when you measure. But it controls interference. When two amplitudes combine and point the same way, they add (constructive interference). When they point in exactly opposite directions, 180° apart, they cancel completely (destructive interference). Any other angle gives partial cancellation.
A real number lives on a straight line. Its only direction is its sign: positive (pointing right, $0°$) or negative (pointing left, $180°$). Just two options, with nothing in between. Two real amplitudes can only ever be $0°$ or $180°$ apart. There are no other angles on a line.
Real amplitudes can cancel: $\frac{1}{\sqrt{2}} + \!\left(-\frac{1}{\sqrt{2}}\right) = 0$. That is genuine destructive interference. But it only works in one specific situation. Useful quantum algorithms need to rotate amplitude directions continuously by small, precise angles like $7°$ or $14.5°$ or $45°$. None of those angles exist on a straight line.
Phase and magnitude are completely independent. The numbers $\alpha = 0.6$ and $\alpha = 0.9$ have different magnitudes but the same phase (both positive, both pointing at $0°$). The numbers $\alpha = 0.7$ and $\alpha = -0.7$ have identical magnitudes but opposite phases ($0°$ vs $180°$). Real numbers can vary their magnitude freely, but their phase is frozen to just two options. That is the whole problem, and you are about to discover it directly.
The experiment: find where real numbers break
Picture an amplitude as an arrow (a phasor) drawn on a flat surface. The arrow's length is its magnitude. The direction the arrow points is its phase. Two arrows that point in exactly opposite directions (180° apart) cancel perfectly when added. Two arrows that point the same way add together.
With only real amplitudes, your arrow can only point right ($+$) or left ($-$). Below, you control two real amplitudes, $\alpha$ and $\beta$. The interference result $\alpha + \beta$ shows whether they add or cancel. Your goal is to find the combination that makes $\alpha + \beta = 0$ while keeping $|\alpha|^2 + |\beta|^2 = 1$. This second condition just means the total size stays at 1. Every quantum state must satisfy it.
Step 2: Flip the toggle to unlock the imaginary dimension. Set both $\alpha$ and $\beta$ to positive values (same sign), then use the new slider to rotate $\beta$ until you reach zero. Notice what the second dimension makes possible that the line could not.
A precise three-step argument
What you just found in the experiment can be stated as a clean logical argument. Here it is:
A real number can only point in two directions: right (positive, $0°$) or left (negative, $180°$). Those are the only two directions available on a straight line. Every other angle ($45°$, $90°$, $135°$, or anything in between) does not exist on the number line.
A well-known quantum search method (called Grover's algorithm) works by rotating amplitudes by a small angle at each step, around $14.5°$ for 16 items or $5.7°$ for 100 items. Another important technique splits phases into evenly spaced directions all around a full circle. None of these required angles are $0°$ or $180°$.
If your amplitudes are real numbers, you cannot represent a direction of $14.5°$. You cannot rotate by $5.7°$. Those directions simply do not exist on a line. Real numbers are not insufficient because of convention; they are insufficient because the geometry is wrong. You need a plane.
Left: real amplitudes are stuck on a line with only two directions. Right: 2D amplitudes can point anywhere on the unit circle, giving a continuous infinity of phases.
More precisely: a quantum amplitude has a magnitude $r$ (its length) and a phase $\theta$ (its angle). For real numbers, only $\theta = 0°$ and $\theta = 180°$ are possible. For a 2D number, $\theta$ can be any angle at all, giving a continuous circle of directions instead of just two isolated points on a line.
A 2D amplitude has two coordinates: $a$ along the familiar left-right axis, and $b$ along a perpendicular up-down axis. Its magnitude is $\sqrt{a^2 + b^2}$ (the distance from the center). Its direction angle is $\theta = \arctan(b/a)$. Rotating the amplitude means sweeping $\theta$ around the circle while keeping the length the same. Real numbers have nowhere to go except left or right, so they cannot rotate freely. A 2D number can rotate to any angle you choose.
The gap you found, in three precise statements
-
Phase is the hidden variable in every amplitudeAmplitude = magnitude × phase. Probabilities ($|\alpha|^2$) only reveal the magnitude. Phase determines interference. Real numbers give two phases: $+$ (0°) and $-$ (180°). That's the complete vocabulary.
-
Quantum algorithms need a continuous circle of phasesGrover rotates by $\arcsin(1/\!\sqrt{N})$. The QFT uses $e^{2\pi ik/N}$ for every $k$. Phase kickback shifts by arbitrary angles. None of these are 0° or 180°. Real numbers cannot represent them.
-
The solution is a plane, not a better lineAdding a perpendicular axis gives an amplitude a second coordinate. Direction becomes an angle: continuous, freely rotatable, and precisely controlled. Two real coordinates together form one 2D number. Infinite directions become possible.
-
$|\alpha|^2 + |\beta|^2 = 1$ survives in the planeWith 2D amplitudes, $|\alpha|^2 = a_R^2 + a_I^2$ (the squared distance from the origin). Normalization still holds and probabilities still sum to 1. The constraint does not change. Only the phase vocabulary expands from 2 to $\infty$.
Real numbers give two directions.
Quantum computing needs a full circle of directions.
A circle lives in a plane.
A plane needs a second axis.
That second axis has a name, and it is a surprising one.
- Nielsen, M. A. & Chuang, I. L. — Quantum Computation and Quantum Information, Cambridge University Press, 2000. §1.2: Quantum bits — postulate of state space and normalization.
- Preskill, J. — Lecture Notes for Physics 229, Caltech, 1998. Chapter 2: Foundations of quantum theory. Available online
- Feynman, R. P. — The Feynman Lectures on Physics, Vol. III, Chapter 1: Quantum Behavior. The double-slit argument for why phase must be continuous.