Collapse Is Not a Rotation

Contents

1. The two maps
2. The shape of the second map
3. You have met this map before
4. The weights were never optional
5. Watching stops the world
6. The environment measures without reading
7. No frame can say when
8. Four ways to read the second map
9. What the structure fixes
10. The ledger
11. Exercises

Somewhere across the last eight lessons a habit formed. Change, wherever this series found it, kept turning out to be rotation. Multiplication by i was a quarter turn. The exponential was rotation run continuously. A frame was an angle; a boost was a rotation through an imaginary one. In Part Seven the habit became a theorem about quantum mechanics itself: the i in the Schrödinger equation is load-bearing precisely because it makes time evolution a rotation of the state — rigid, deterministic, reversible, exactly as strict as Newton. Nothing in that machine ever loses anything. Every turn can be turned back.

Quantum mechanics keeps the habit perfectly — right up to the moment somebody looks.

01The two maps

Von Neumann, writing the mathematical foundations of the theory in 1932, refused to blur this. He numbered the two ways a quantum state can change and treated them as separate axioms. Process 2 is the one this series has been building for eight lessons: the state rotates under the Schrödinger equation, continuously and causally. Process 1 is what a measurement does: the state jumps, discontinuously and statistically, to one term of the sum. He put the anomalous one first, because it is the anomaly.

Everything that makes the two maps different can be read off in one pass:

The title of this lesson is now a theorem, and a short one. A rotation is invertible: two different states before it are two different states after it, always, which is what makes the reversal possible. Collapse is many-to-one. A state that was 99 percent “up” and a state that was 1 percent “up” can both end the measurement as up, and once they have, no operation on the outcome tells you which one you started with. A many-to-one map cannot be a rotation — not because rotations are rare or special, but because invertibility is what the word means. Whatever collapse is, the one thing the mathematics rules out on line one is that it is more of the same motion.

A measurement always causes the system to jump into an eigenstate of the dynamical variable that is being measured.

P. A. M. Dirac, The Principles of Quantum Mechanics, 4th ed. (1958), p. 36

02The shape of the second map

So what is the second map, if not a rotation? It is the other fundamental thing a vector space knows how to do: drop a perpendicular. A projection P takes the state and keeps only its component along one chosen direction — the outcome that occurred — discarding the component perpendicular to it. Three properties define it, and each one is a sentence about laboratories.

P² = P says projecting twice is projecting once — which is Dirac’s sentence above wearing algebraic clothes: measure again immediately and you get the same answer with certainty, because the state is already lying along the outcome. The eigenvalues 0 and 1 say a projection asks a yes/no question — each direction either survives whole or dies entirely; there is no partial credit. And non-invertibility says the perpendicular component is not stored anywhere. It is not hidden, not encrypted, not folded into a phase. Within the formalism, after the projection, it does not exist.

Figure 01 One state, two maps: the rotation keeps the length, the projection spends it

• rotation: the whole state survives
• projection: one component survives

Two honest boundaries before this picture hardens. First, when an outcome is degenerate — many independent directions all counting as the same answer — the correct rule, written down by Gerhart Lüders in 1951, is to project onto the whole eigenspace and stop: coherence within the surviving subspace is preserved, not ground down to a mixture. Second, the projection postulate describes the ideal, repeatable, minimally disturbing measurement, and most real measurements are not that: the photon that announces itself at a detector is absorbed by the announcement. The modern framework for real instruments (positive operator-valued measures) generalizes the geometry without changing the moral: the textbook projection is the cleanest member of the family, not the whole family.

03You have met this map before

Here is the uncomfortable recognition. This series began, back in Part Three, with a projection — and called it the structure of experience. You only feel the real part. Taking the real part of a complex number is not a metaphorically projection-like operation; it is a projection, in exactly the sense of the last section. Re(x + iy) = x is idempotent — take the real part of a real part and nothing further happens — and it is many-to-one: every number on a vertical line in the complex plane has the same real part, and once you have taken it, no operation recovers which one you held. The map that defines what your body can feel and the map that ends a quantum measurement are the same shape.

So why did classical physics never have a measurement problem? If feeling is projection, why is there no classical collapse? Because classically, the projection is a viewpoint, not an event. When Part Three’s oscillator runs, the full complex object A·e^iωt keeps rotating regardless of what you feel; taking the real part changes your report, not the state. And because the rotation underneath never stops, the discarded component is not even lost to you: watch the real part alone for a stretch of time and the rotation replays everything into it — amplitude and phase can both be read off the felt history. The perpendicular component is out of view, not out of existence. Nothing was deleted, because the projection was never applied to the state. It was applied to your access.

Collapse is what you get when the projection stops being a viewpoint and becomes the new state. After the flash on the screen, the formalism does not say “the other term is out of view.” It says the other term is gone — the state is the projected vector now, and no amount of watching the survivor replays what was deleted, because the deletion is exactly what the survivor no longer carries. Part Four sharpened how much is deleted: the Born rule squares the amplitude, so measurement does not even keep the real part of the quantum state — it burns the entire angle. Every complex amplitude of the same length yields the same outcome statistics.

Figure 02 The deletion is total: four different states, one set of outcome statistics

Part Four called the Born rule “phase deletion” and left it as a fact about the formalism. The present lesson can now say it structurally: measurement is the one place in physics where a projection is applied to the state rather than to your view of it. The whole dispute over what collapse means — the dispute of the next four sections — is over whether that sentence describes the world or merely our bookkeeping.

04The weights were never optional

Whatever the second map is, it comes with numbers: this outcome with probability 0.75, that one with 0.25. It is natural to assume the numbers are a separate law — that God chose the geometry, then chose a probability rule to run on top of it, and could have chosen differently. In 1957 Andrew Gleason proved that this assumption is wrong, and the proof is the purest instance of this series’ slogan that quantum mechanics has to offer.

Suppose only this. Outcomes correspond to subspaces — directions and planes in the state space — and you want to attach to each subspace a probability. You require two things. One number per subspace: the probability attached to a direction does not depend on which measurement context you reached it through — which other orthogonal alternatives happened to be on the menu beside it. Additivity: across any complete set of mutually orthogonal alternatives, the numbers add to one. That is the entire demand. Nothing about physics, nothing about waves, nothing about |ψ|².

Gleason’s theorem says that in any state space of dimension three or more, every assignment satisfying those two requirements has the Born form: the squared-shadow rule, for some state ρ. There is no exotic alternative that a different universe might have used with the same geometry. The weights that the second map spits out are not legislation added to the structure; they are the structure’s only self-consistent way of being weighed. The circle did not get to vote on π, and the state space did not get to vote on the Born rule. Math is not modeling. Math is why.

Two boundaries, stated at full strength because this audience will find them anyway. The one-number-per-subspace premise is doing real work — it is a noncontextuality assumption, and denying it is a coherent (if costly) position; the theorem converts “the Born rule is arbitrary” into “the Born rule follows from context-independence,” which is a relocation of the mystery, not an execution of it. And dimension two escapes: for a lone qubit the subspaces are too sparse to constrain the measure, and non-Born assignments exist. Paul Busch showed in 2003 that extending “outcome” to the modern instrument framework restores the conclusion even there — at the price of a correspondingly stronger additivity premise. The theorem is magnificent; it is not magic.

05Watching stops the world

If the second map were just a story we tell about the first, you would not expect it to have consequences the first map cannot produce. It has them. The cleanest is what happens when you project often.

Take a two-level system being rotated by its own dynamics from state |0⟩ to state |1⟩ over one second — in laboratory language, a π pulse. Left alone, the transfer is certain: rotation is deterministic, and after the second the state is |1⟩. Now interrupt it: check “still in |0⟩?” at N evenly spaced moments. Each check is a projection, and each projection re-plants the state on the axis it just confirmed. The probability that every check finds the system still in |0⟩ — that it never leaves — is [cos²(π/2N)]^N, and that expression does something rotation alone can never do: it rises toward certainty as the checking gets denser.

Figure 03 The Zeno staircase: projection, applied often enough, beats rotation

Watched closely enough, the pot never boils: dynamics that guaranteed a transition are frozen by interrogation. And note carefully what did the freezing in the actual experiment, because a reader primed by the word “watching” will reach for a mind. The “checks” were laser pulses. No consciousness, no gaze, no observer in any psychological sense — any coupling strong enough to record which state the system is in does the work. The lesson of the Zeno effect is not that attention is magic. It is that the second map composes differently than the first: rotations compound smoothly, but a rotation diced by projections is a different dynamics with a different destination. (Whether the freezing strictly requires collapse, or can be told entirely in the language of strong coupling and the first map, is itself contested — which is this lesson’s whole subject in miniature.)

06The environment measures without reading

The obvious question has been waiting since the scene at the top: if states are sums and rotation preserves them, why have you never seen a sum? Not once. No half-silvered cat, no smeared pointer, no doubled moon. The twentieth century’s best answer is decoherence, and the discipline this series owes you is to state with equal clarity what it explains and what it does not.

A cat-sized object cannot be left alone. Air molecules, thermal photons, dust — each one scatters off it, and each scattering event carries away a little record of where the object was. Each record nudges the environment’s state a little further apart depending on which term of the sum it met: the environment is, without anyone designing it, performing the same which-state interrogation the Zeno laser performed — relentlessly, trillions of times a second, and never reading out. Interference between the terms — the experimental signature that a sum is a sum, Part Eight’s entire toolkit — has visibility equal to the overlap of those environmental records. And overlaps multiply. After N scattering events each preserving overlap c, the visibility is c^N: not zero, but falling like an avalanche.

Figure 04 Visibility of the sum, as the environment accumulates its records

The ψ-function of the entire system would express this by having in it the living and dead cat (pardon the expression) mixed or smeared out in equal parts.

Erwin Schrödinger, Naturwissenschaften 23 (1935), trans. J. D. Trimmer (1980)

So decoherence explains three things with real authority. Why interference is unfindable in practice for anything big: the phase information is not destroyed, it is dispersed into a trillion-particle environment you cannot reassemble. Why the surviving alternatives are the ones they are: the environment keeps copying position-like quantities, so position-like states are the stable ones — Wojciech Zurek’s einselection. And why the second map’s statistics look classical by the time you apply it. What decoherence does not explain — and its founders say so as plainly as its critics — is the flash. After every scattering event has been counted, the global state is still a sum, not a choice. Schrödinger’s sentence above remains exactly true of it; the cat terms have merely stopped interfering. Decoherence tells you why you will never catch the ghost. It does not tell you the ghost died. Something — the second map, or an interpretation of it — still has to get you from “the terms can no longer meet” to “one of them happened.”

07No frame can say when

If collapse is a physical event, it happens somewhere, and it happens somewhen. Part Five taught what “when” costs: simultaneity is a tilted slice, and events separated faster than light can connect have no frame-independent order. Now put those two lessons in the same room.

Take Part Eight’s entangled pair, one particle with Alice, one with Bob, far apart. Alice measures; Bob measures; the two events are spacelike separated. In the laboratory frame, Alice measures first — so her projection is the one that collapses the shared state, and Bob’s result is drawn from the already-collapsed distribution. But ride a frame moving at half the speed of light and the slices tilt: Bob measured first, his projection did the collapsing, and Alice received the fait accompli. Same experiment, same clicks, two incompatible histories of when the second map fired — and relativity’s verdict from Part Five is that neither frame is wrong, because there is no fact about the order for them to be wrong about.

Figure 05 Who collapsed whom? The question has no frame-independent answer

Two structural facts keep this from being a paradox, and both are already on the series’ books. First, the disagreement is invisible: Part Eight’s no-signalling result — nothing happens at Bob’s end — means every observable statistic is identical whichever collapse story you tell. The frames disagree only about the bookkeeping order of an unobservable. Second, this is not merely awkward but provable: Aharonov and Albert showed in 1981 that no collapse-at-an-instant can be made Lorentz covariant — a relativistic quantum state cannot carry a frame-independent answer to “has the collapse happened yet?”, and must instead be assigned slice by slice, frame by frame. (Covariant collapse theories do exist — Tumulka’s relativistic GRW attaches its randomness to pointlike “flashes” rather than to states — so the correct conclusion is not that relativity forbids collapse. It is that relativity forbids a frame-independent history of the state.)

Part Five ended with a warning against relativism: frame-dependent does not mean unreal — components are frame-dependent, the invariant object is not. Apply the same audit here and notice what is on which list. The clicks and their correlations: invariant. The statistics: invariant. The moment the state collapsed: frame-dependent, unobservable, and provably impossible to covariantize. When a candidate physical process sits entirely on the frame-dependent side of Part Five’s ledger, the structure is not telling you it is unreal — but it is telling you exactly what evidence there could never be.

08Four ways to read the second map

…the shifty split of the world into ‘system’ and ‘apparatus’.

John Bell, “Against ‘measurement’”, Physics World 3(8), 1990

The formalism says: rotate, except when a measurement occurs; then project. Bell’s complaint was never that this fails — it never fails — but that “when a measurement occurs” is not a clause the structure itself can parse. Nothing in the mathematics marks certain interactions as measurements; the split between system and apparatus is drawn by hand, and can be drawn in different places without changing any prediction. Part Seven framed the standard escape routes as Maudlin’s trilemma. This lesson can now restate them — plus the one the trilemma’s framing tends to hide — by what each does to the second map.

1. Collapse is physics · GRW The second map is a real process with its own dynamics: every constituent suffers a spontaneous localization about once per 10¹⁾ seconds, width ~10⁻⁷ m. One atom almost never collapses; the 10²⁳ entangled constituents of a pointer collapse it in a microsecond. Note the fine print: the real process is not a sharp projection but a smooth Gaussian squeeze — nature’s version blurs the textbook’s map. Denies: the state always rotates. In principle testable, so far untested.
2. Collapse never happens · Everett Rotation is all there is. The measurement entangles apparatus, environment and observer into the sum; decoherence stops the terms from meeting; what looks like collapse is what a branch looks like from inside. The deleted component was never deleted — you are in it, too. Denies: single outcomes. Standing bill: why the Born weights, when everything happens.
3. Collapse is effective · Bohm The universal wave never collapses, but particles have actual positions, and the wave function that guides a subsystem — conditioned on the real configuration of everything else — collapses as a theorem, not a postulate. Grounded in fact, not credence; no branching either. Denies: the wave function is complete. Price: the guiding connections are non-local.
4. Collapse is inference · the epistemic reading The state update is what happens to a forecast when you look outside — conditioning, not dynamics. The textbook update rule reads this way with the state fixed by the preparation; QBism radicalizes it into personal probability. One clause is compulsory: unlike the weather, there was no fact already there that the state was ignorant of. Denies: the premise — that ψ describes the system rather than the bettor.

All four reproduce every confirmed prediction; GRW alone diverges in principle, in regimes not yet reached. The structure — and this is the point of listing them in its vocabulary — does not adjudicate. Anyone who tells you quantum mechanics has settled which reading is true is selling something. What the structure does instead is the subject of the last section.

09What the structure fixes

Step back and count what the mathematics has determined about collapse without once deciding whether collapse occurs. Its shape: if outcomes are subspaces, the update is projection — idempotent, yes/no, many-to-one; the only alternative to rotation the geometry offers. Its weights: Gleason — grant context-independence and additivity, and the Born rule is the only measure available; the probabilities were never a separate law. Its basis: einselection — the environment’s relentless unread measurement picks which alternatives are stable enough to be outcomes at all. Its discretion: no-signalling — whatever collapse is, it provably cannot carry a message. Its timelessness: Aharonov–Albert — whatever collapse is, it provably has no frame-independent moment of occurrence.

That is an extraordinary amount of legislation for a process that may not exist. And it is the series’ central claim, run in the only direction it runs: none of those five results was discovered by watching collapses and generalizing. Each was read off the structure — the geometry of subspaces, the composition of overlaps, the tilt of a slice — and the world then complied. The structure does not describe collapse. The structure is the thing collapse, if it happens, must be an instance of. What remains open is a single bit: whether. Whether the projection is applied to the state (GRW, at the price of new dynamics), to your slice of a branching state (Everett, at the price of probability), to your effective description (Bohm, at the price of non-locality), or to your forecast (the epistemic reading, at the price of the system itself). The measurement problem, stated in this series’ vocabulary, is that the mathematics has fixed everything about the second map except whether the universe runs it.

The standing hedge, owed in every part and doubly owed here: the physics is compatible with the claim that the world is this structure; it does not prove it. A committed realist about objects can accept all five results and hold that the structure merely describes how object-facts hang together. And one vaccination, administered deliberately: nothing in the five fixed items mentions a mind. The Zeno freeze was done by laser pulses; the environment measures without reading; Wigner’s suggestion that consciousness closes the circuit was a minority reading that Wigner himself later abandoned. If a book tells you quantum mechanics proves the observer creates reality, check whether it ever states the difference between a projection and a rotation. This lesson exists so that you can.

10The ledger

What is load-bearing

The two-maps distinction is von Neumann’s, verbatim, and it is exact: unitary evolution is deterministic, invertible and norm-preserving; the measurement update is stochastic, many-to-one and norm-shrinking, so it cannot be a rotation — that is a theorem about invertibility, not a slogan. P² = P is repeatability as algebra. Gleason’s theorem, the Zeno formula and its Itano confirmation, the impossibility of signalling by collapse, and the Aharonov–Albert impossibility of a covariant collapse history are all established results.

What is convention

Calling the second map “collapse” at all — the formalism says only “update.” The Lüders choice for degenerate outcomes, natural but chosen. And treating sharp projection as the measurement: the general instrument is a POVM, the photon at the screen is absorbed rather than projected, and exact repeatability holds only for ideal measurements of discrete quantities. The textbook map is the cleanest member of a family, promoted to spokesman.

Where the shorthand breaks

“The geometry forces the Born rule” stops at two boundaries: dimension two, where Gleason’s theorem is silent (Busch’s POVM extension restores it, on a stronger additivity premise), and the one-number-per-subspace assumption, which is noncontextuality doing real work, not a triviality. “Collapse is projection” breaks at GRW, whose real process is a smooth Gaussian squeeze. And Part Three’s bridge is an identity of shape, not of consequence: Re is a projection, but classical physics applies it to your access while measurement applies it to the state — the whole difference lives in that clause.

Where I would push back on myself

An Everettian will say this lesson anatomized a map that never fires — nine sections on the algebra of an artifact of self-location — and nothing in this lesson refutes them. Sharper still: Gleason fixes the form of the probabilities but not why there are outcomes to be probable; the hard core of the measurement problem — how “the terms can no longer meet” becomes “one of them happened” — is not solved here, only located. And my closing inversion assumes the structure legislates rather than summarizes; a determined instrumentalist reads all five “fixed” items as facts about our bookkeeping that the world merely tolerates.

11Exercises

1. The map, by hand For the state at 60° in Figure 01, write P for the projection onto |1⟩ as a 2×2 matrix, verify P² = P, and compute ‖Pψ‖. Then exhibit two different states that P sends to the same output — the two-line proof that no inverse exists.
2. The staircase Compute [cos²(π/2N)]^N for N = 4 and N = 50. Then use cos θ ≈ 1 − θ²/2 to show the survival behaves as 1 − π²/4N for large N, and check that against the computed values in Figure 03.
3. What the phase was for Show |0.6e^iφ|² is independent of φ. Then explain, using Part Eight’s two-slit rule, why φ was physically consequential five nanoseconds before the measurement. What kind of map preserves a quantity’s consequences while deleting the quantity?
4. The tilted slice With A = (0, −1.5) and B = (0.3, +1.5), use t′ = γ(t − βx) to find the β at which A and B are simultaneous, and verify that β = 0.5 gives the values in Figure 05. Why does no signal — and hence no collapse-borne influence — survive this ambiguity?
5. Argue the other side This lesson treated the second map as the puzzle. Steelman the Everett position that there is no second map: write the strongest paragraph you can arguing that rotation plus decoherence plus self-location accounts for everything in this lesson, including the Zeno staircase. Then state plainly what the position costs — where the Born weights come from when every outcome occurs — and whether you judge the price fair.