Changing Coordinate Axes: Bases and Basis Transformations (Quantum Computing Column, Part 017)
> Prerequisites: Part 012 (the state sphere and "vertical collapse"), 014 (the H-gate table), 016 (the Born rule). > Next up: Part 018, "The Necessity of Reversibility."
Suppose a new student joins your class and you want to record them in a roster. Two ways to ask:
- Ask about gender: "Boy or girl?"
- Ask about glasses: "Wears glasses or not?"
- |+⟩ = half |0⟩ plus half |1⟩ (amplitudes in phase) — exactly the output of the H gate acting on |0⟩ (row one of the table in Part 014);
- |−⟩ = half |0⟩ minus half |1⟩ (amplitudes out of phase) — the output of the H gate on |1⟩.
- Disanalogy 1: The class's classifications (gender, glasses) don't convert, but you can ask each separately, in any order; a quantum system's two bases cannot be asked in sequence — the first question's collapse has already rewritten the state (Part 002), so the second question is asked of the new, collapsed state. Quantum "axis changing" changes the object itself.
- Disanalogy 2: The way of asking is a human choice, independent of the person asked; in quantum mechanics, choosing a basis not only changes the answer distribution — which components the state even "contains" differs across bases (|0⟩ has one component in the up-down basis, two in the left-right basis). A state's "contents" are relative to a basis; there is no absolute self-inventory.
Both questions cover the same group of people, and everyone has an answer under each — but the two classifications don't convert into each other: knowing someone is a boy tells you nothing about whether they wear glasses. To determine both, you must ask twice — while the person being asked never changes; only the way of asking changes.
In the world of qubits, "way of asking" has a formal name: a basis. The central claim of this post is the hidden backbone of all quantum algorithm design:
> Measurement is not a single action but a family of actions — you choose a basis, you choose a way of asking; and no state can give a definite answer to both ways of asking simultaneously.
Beyond "Up-Down": There Is Also "Left-Right"
Part 012 said: measurement makes the arrow on the sphere fall vertically, landing on |0⟩ or |1⟩. We also planted a hint: "vertical is not the only measurement direction." Time to deliver.
Besides the "up-down" basis of |0⟩ and |1⟩, there is an equally legitimate "left-right" basis, whose two members are written |+⟩ and |−⟩. They are not new particles or new bits — just two other directions on the same sphere — two diametrically opposite points on the equator (per Part 012: latitude sets the ratio, longitude sets the phase; the left-right basis members are exactly the states with "fifty-fifty superposition, same/opposite phase"):
See it? The H gate is the translator between the "up-down" basis and the "left-right" basis — feed it |0⟩, |1⟩ and it outputs |+⟩, |−⟩; feed those back in (it applies twice to return home, as Part 014 noted) and you get |0⟩, |1⟩ again. One table, translation in both directions.
One State, Two Ways of Asking, Two Fates
Now run the Born rule (Part 016) in both bases and watch the same state "play to the audience":
State is |0⟩ (north pole). Asked in the up-down basis: 100% outcome 0 — it lives at the north pole, beyond doubt. Asked in the left-right basis: |0⟩ contains |+⟩ and |−⟩ in equal halves (decompose it into left-right coordinates), so 50% plus, 50% minus — purely random.
State is |+⟩ (equator's front point). Asked in the left-right basis: 100% plus. Asked in the up-down basis: 50/50 random.
The same state, undisturbed, gives a definite answer to one way of asking and pure noise to the other. The single-qubit version of uncertainty lives right here: the up-down and left-right questions are incompatible — any state can be definite for at most one of them. This is not an engineering shortcoming of measurement; it is a geometric inevitability that "north-south coordinates and east-west coordinates cannot be converted into each other" (Part 012: latitude and longitude are independent degrees of freedom).
This also repays an old debt: Part 003 said the Bell experiments ruled out the "glove camp" — the elegance of those experiments lies precisely in rotating through multiple bases when asking. If the world had only the up-down basis, the glove story could still hold together; it is exactly the move "switch basis and ask again, yet the correlations stay locked" that closes off every escape route of the glove model. Bases and basis transformations are standard weaponry in quantum mechanics' court judgments.
The Axis-Changing Dance Inside Algorithms
Now we can expose a universal pattern of quantum algorithms — it will reappear throughout this column, so let's plant the sign now:
> Change axes → do work in the new axes → change back.
The H gate sends the state into the "left-right" viewpoint (creating superposition is only its first face), where certain interference choreographies impossible in the up-down basis can be done; then you change back to read out. The quantum Fourier transform of Part 036 — the heart of Shor's algorithm — is essentially a supercharged axis change: it moves the entire "distribution of amplitudes" into the "frequency" coordinate system, making periodic structure emerge from noise. When a quantum circuit baffles you, look for its axis-changing dance first, and things will usually click.
Part 021 will give the H gate its own biography — today you only need to take away a new perspective: the H gate is not "a gate that creates randomness"; it is "a gate that changes the way of asking." Creating superposition is merely the first application of changing axes.
Dismantling the Analogy
What the class-roster analogy borrowed, and where it breaks:
Luggage to Take Home
1. A basis = a way of asking in measurement: |0⟩/|1⟩ are up-down, |+⟩/|−⟩ are left-right; the H gate is the translator between the two — its real job is changing axes; creating superposition is a byproduct. 2. No state can be definite for two incompatible bases at once: |0⟩ is 100% in the up-down basis, purely random in the left-right basis, and vice versa — uncertainty here is geometry, not mysticism. 3. The algorithmic pattern "change axes → work → change back" runs through all quantum algorithms; the QFT is its enhanced version (Part 036), and the Bell experiments were settled by rotating bases (Part 003).
Hook for the next stop: the amplitude-ratio table (Part 014) allows arbitrary numbers in its cells — but quantum gates face a qualification list, whose hardest rule is "no losing information." Why does the universe forbid information loss at the bottom level? And what trouble does an information-losing gate cause? Part 018, "The Necessity of Reversibility."
*(Original post tagged: quantum computing, from coins to error correction, basis transformation, uncertainty.)*