The Square Step: From Amplitudes to Probabilities
> Prerequisites: Episode 013 (amplitudes) and Episode 015 (the two-bit ledger). > Next episode preview: 017 "Changing the Coordinate Axes".
*From Coins to Error Correction - A Quantum Computing Column | Minimum Math Toolbox*
In the previous episode's ledger, the state (1/√2, 0, 0, 1/√2) appeared, and I said "the probability of |00⟩ is (1/√2)² = 1/2" - and you probably didn't stop to question it. This episode stops, because "the square step" is one of the strangest and deepest rules in quantum mechanics: the Born rule (another name-name; just terminology). This post answers three questions: Where does the square come from? Where did the cross terms go? Why a square specifically?
First, Lay Out the Accounts
The most general single-qubit superposition has two amplitudes: a·|0⟩ + b·|1⟩ - where a and b are complex arrows (Episode 011), each with a length and a direction.
When measured, what's the probability of getting 0? The Born rule in one sentence: P(0) = |a|² - the length of arrow a, squared.
Three straightforward corollaries, stated up front:
- Probability depends only on length, not direction - when measuring a qubit by itself, phase is invisible (the seed of "global phase doesn't matter" in Episode 019).
- a's length lies between 0 and 1, so its square is exactly a probability; total probability |a|² + |b|² = 1 is the bookkeeping version of the "welded length" statement from Episode 012 (the Bloch sphere).
- Amplitude 1/√2 → probability 1/2: "fifty-fifty" is called 1/√2 in amplitude-land - the origin of those strange numbers in the Hadamard gate table (Episode 014).
- Disanalogy one: geometric length doesn't care about context - any segment has a length. Amplitude length only makes sense for the combined account; the "length of a single path" cannot be independently measured (Episode 002: measurement is collapse). Length is a ledger property, not a physical one.
- Disanalogy two: in geometry, squaring means area; here, squaring is merely the bookkeeping act of "drop direction, keep magnitude," with no "area" imagery involved. Forcing it into an area picture only tempts you to hunt for "which plot of land," adding confusion.
So far, elementary-school arithmetic. The real question hides in the two-bit ledger.
Where Did the Cross Terms Go?
Two independent qubits: the first in a|0⟩ + b|1⟩, the second in c|0⟩ + d|1⟩. Expanding per Episode 015, the joint ledger has four cells: ac, ad, bc, bd. What is the probability of measuring "00"?
Two routes lead here. First, the formula directly gives it: |ac|² = |a|²·|c|² (the multiplicative property of complex lengths). This is exactly the elementary-probability answer: independent events multiply their probabilities - the classical rule survives intact as a special case inside the quantum rule.
Second, the clumsy "sum first, then square" method - which we must walk carefully, because entangled states only allow this route: the combined amplitude is ac, and the probability is |ac|². Identical to the digit.
Hmm - what happened to "change the order, change the world"? In the independent case, changing the order changes nothing - because each outcome is reached by only one path (|00⟩ can only arise from the a-paired-with-c combination). There are no arrows to add, so the square decomposes without distortion.
The key is entangled states: the ledger (1/√2, 0, 0, 1/√2) cannot be factored into a product expansion - but we can artificially change how we ask the question, making the same outcome reachable by multiple paths. That's Episode 017's plot; here's just one spoiler line: once multiple paths contribute amplitude to one outcome, "sum-then-square" and "square-then-sum" part ways. The missing piece is called the cross term, and it is the ledger identity of interference: the cosine of the angle between two arrows, times the product of their lengths.
Same direction: positive cross term, probability exceeds the sum of separate contributions (constructive interference). Opposite directions: negative cross term, probability falls below the sum, down to zero (destructive interference - Episode 013's "never happens"). What classical probability rules lack is precisely this term - the cross term is the entire difference between quantum and classical, and the ledger location of all the computational power of quantum algorithms.
Why a Square, Exactly?
An honest confession: "why a square?" has no deeper derivation - it is an axiom, backed by experiment. Physics has a few rules standing at the top of every derivation chain, and the Born rule is one of them. People have tried to derive it from deeper principles (some attempts are beautiful; we'll meet its shadow after Episode 046), but as of today the hardest evidence remains: compute with a square, and experiments all check out; compute with any other rule, and they don't.
Still, "squaring" is not arbitrary. It satisfies three constraints everyone accepts, and these nearly lock it in:
1. Probabilities must be non-negative - arrows can point positively or negatively (amplitudes can be inverted); the simplest way to turn that into a non-negative number is squaring (or taking a length). 2. Total probability conserved at 1 - the bookkeeping version of Episode 012's "welded length"; this rules out cubing and every other scheme that breaks conservation. 3. Two aligned arrows together must quadruple the probability, not double it - square each length, then add: two aligned arrows of length 1/2 each combine to length 1, probability 1. If you instead added the individual "intensities" directly, you'd only get 1/2. Experimentally, interference does give the quadrupling (the brightness of constructive fringes) - direct, hard evidence for "sum first, then square."
Put these three common-sense constraints together, and squaring is nearly the only surviving candidate - about the best treatment an "axiom" can get: not proven right, but hard to find an alternative.
Tearing Down the Analogy
Where the "arrow length" geometric analogy breaks down:
Luggage to Take Home
1. Born rule: probability = squared length of the combined amplitude; in the independent case it automatically reduces to the classical multiplication rule - classical probability is a special case of quantum bookkeeping, not its rival. 2. When multiple paths reach the same outcome, cross terms appear in the ledger (contributions from direction angles): positive means constructive, negative means destructive, zero means the outcome never happens - the cross term is interference's ledger identity, where quantum computing power hides. 3. The square is an axiom, not a derivation; yet the three experimental constraints - non-negativity, conservation, and interference quadrupling - nearly lock it in. Trust it, because every alternative fails.
Hook for the next stop: so far, measurement always "falls vertically" (the mine buried in Episode 012) - but who decreed up and down? Swap the coordinate system, and the same sphere answers questions completely differently; half the cleverness of quantum algorithms hides in changing axes. Episode 017: "Changing the Coordinate Axes."
*Tags: quantum computing, from-coins-to-error-correction, Born rule, cross terms*