English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Global Phase Doesn't Matter: Relative Phase Is Everything

Forum topic · 小凯 · 2026-11-07

Summary

Episode 019 of the 'From Coins to Error Correction' quantum computing column explains the distinction between global and relative phase in quantum states. Multiplying all amplitudes of a state by the same complex number (a global phase) changes nothing physically measurable: probabilities depend on amplitude lengths squared, and interference cross-terms depend only on angles between amplitudes. States differing by a global phase are literally the same point on the state sphere. Relative phase, by contrast, governs interference: 0 degrees means constructive interference, 180 degrees destructive. The states |+> and |->, which differ only by a 180-degree relative phase, are indistinguishable in the computational basis (both 50/50 random) but fully distinguishable in the X basis. The post resolves a mystery from episode 014: the Z gate leaves pure basis states like |1> unaffected (only a meaningless global phase flip), but maps |+> to |-> by flipping only the |1> component's phase, acting as a single-qubit relative-phase adjuster that rotates the state half a turn along a meridian of the state sphere. Relative phase is likened to invisible ink: hidden in the wrong basis, revealed by interference in the right one, a motif behind algorithms like Simon's algorithm and phase estimation.

> Prerequisites: Episode 014 (Z gate), 016 (cross terms), and 017 (bases and axis changes). > Next: Episode 020, "The X/Y/Z Trio".

Global Phase Doesn't Matter: Relative Phase Is Everything

Let's reopen a cold case first.

Episode 014 introduced the Z gate but left a loose end: it multiplies the amplitude of the |1⟩ component by -1, yet no measurement probability budges — |0⟩ is still |0⟩, and a half-and-half superposition stays half-and-half. The hanging question: does this gate do nothing at all?

The answer: it does something big, and understanding what requires this episode's only new concept — the distinction between global and relative phase. This is the most subtle post of the second batch, and the watershed between "doing the math" and "designing algorithms": all the interference choreography in the algorithm episodes ahead stands on this concept.

First: something that cannot be measured

What happens if you multiply all amplitudes of a state by the same complex arrow (say, all by -1, or all rotated by 37 degrees)?

Episode 016's ledger gives the answer: probability is the squared length of an amplitude, and multiplying everything by the same number changes no lengths; the cross-terms in multi-path interference depend only on the angles between amplitudes — rotate everything together and the angles don't move. Every line of the ledger is unchanged.

Conclusion: a uniform rotation of everything does not physically exist. Two states related by a global phase are not merely similar — they are the same state. Episode 012's state sphere gave this a geometric identity: the same point on the sphere. The north pole |0⟩ has no east/west distinction; "the north pole facing east" is not a new location.

This is also why a choir transposing every voice together isn't a rearrangement: the whole choir moves from C to D, every part shifts, but the intervals between voices don't change at all — a melody's "content" lives in the intervals, not in absolute pitch. A quantum state's content lives in the relative relationships between amplitudes, not in absolute phase.

Next: something that decides life and death

Global phase is irrelevant; relative phase is everything.

In episode 016's cross-term formula, the real decision-maker is the angle between two arrows: 0 degrees means constructive interference and quadrupled probability; 180 degrees means cancellation and probability zero; 90 degrees changes nothing. Turn the angle a few degrees and the success of an interference shifts accordingly — relative phase is the string of interference; its tension directly decides which answer survives.

A sharp contrasting pair (met in episode 017, now viewed with fresh eyes):

  • |+⟩ = half |0⟩ plus half |1⟩ (relative phase 0 degrees)
  • |−⟩ = half |0⟩ minus half |1⟩ (relative phase 180 degrees)
  • Ask in the up/down basis: the two states are identical — pure 50/50 randomness. Ask in the left/right basis: one is 100% +, the other 100% − — utterly different states. What they differ by (a 180-degree relative phase) is real, distinguishable, manipulable physics — it just only shows up in the right basis.

    Phase thus earns its proper portrait: invisible ink — completely invisible in the wrong basis (all probabilities identical), but under interference with the right basis, every stroke appears. Countless moves in quantum algorithms amount to "write the information you want to read into relative phase, then reveal it with a suitable basis" — episode 035's Simon's algorithm and episode 041's phase estimation have it written right into their names.

    Now we close the case on the Z gate

    Back to the opening cold case — what the Z gate actually does:

    On the pure state |1⟩: it applies a global -1, with no physical consequence — the Z gate really does "do nothing" to it.

    On the superposition |+⟩: Z multiplies only the |1⟩ component by -1 — note, this time it is not global: the |0⟩ component is untouched while the |1⟩ component flips phase. The angle between the two arrows is wrenched by 180 degrees, and |+⟩ becomes |−⟩ on the spot. In the up/down basis you see no difference (still 50/50); in the left/right basis, everything changes.

    So the Z gate's precise identity: a single-qubit relative-phase adjuster — it rotates the state half a turn along a meridian of the state sphere (episode 012 said longitude = relative phase; today that coordinate gets its physical meaning). For some states it's a no-op, for others a fatal rewrite — one gate, two fates, and the criterion is whether the state contains relative phase that can be twisted.

    An engineering rule of thumb along the way (used many times later): since global phase is meaningless, algorithm designers only ever need to worry about relative phase. Two circuits differing only by a global phase are fully equivalent solutions — when checking interference, watch the pairwise angles between arrows and ignore any common rotation. This "lazy rule" saves quantum engineers countless gray hairs every year.

    Dismantling the analogy

    What the choir analogy borrowed — and where it breaks:

  • Dissimilarity one: pitch is a physical frequency; transposing really changes sound waves. A quantum global phase doesn't even change the wave — it is not "changed but inaudible," it physically never happens. The analogy holds only for "content lives in relative relationships."
  • Dissimilarity two: a choir's interval errors are audible to human ears; relative phase has no "direct ear" — it only develops in interference in a particular basis (episode 017's axis changes). Whether something is "hidden" or "revealed" depends on how you ask — far more counterintuitive than any auditory experience.

What to take home

1. Global phase doesn't exist: multiplying all amplitudes by the same complex number = the same state (same point on the sphere); a choir may transpose, a quantum state may not. 2. Relative phase is everything: the angle between amplitudes rewrites probabilities directly via cross terms — 0 degrees constructive, 180 degrees destructive; |+⟩ and |−⟩ differ only by this angle and are two completely different definite states in the left/right basis. 3. The Z gate = a relative-phase adjuster (half a turn along a meridian); phase is invisible ink — information written into phase must be developed with interference in the right basis, the mother motif of every phase-style algorithm ahead.

Hook for next stop: X (the switch) and Z (the phase flip) are in hand; one piece remains to complete single-qubit control — Y. Around which axis of the state sphere does each gate rotate, why exactly three axes, and how do they combine into any rotation on the sphere — episode 020, the finale of batch two: "The X/Y/Z Trio."

*(Original post includes an SVG title-card illustration with the caption "a slight difference, entirely different outcomes.")*

Tags

#quantum-computing#global-phase#relative-phase#z-gate#interference#phase-estimation#quantum-state-sphere

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178636853