The Postulates of Quantum Mechanics

Every article in this track has been playing by rules never written down in one place. The wave-mechanics chapter used them: wavefunctions normalized, observables as operators, energies off an eigenvalue menu, probabilities as squared amplitudes. The formalism chapter has been proving pieces: states are vectors in a Hilbert space, Hermitian operators have real eigenvalues and orthogonal eigenstates, expansions converge. This article is where the contract gets signed. Quantum mechanics, in its entirety, is a short list of postulates — five or six sentences, depending on how you count. One names the object that is a physical state. One names the observables. One says what values a measurement can return, one attaches probabilities, one describes what measurement does to the state, and one governs everything between measurements. That is the whole theory: hydrogen, chemistry, semiconductors, and quantum computers are these sentences plus arithmetic. Reading the list has a second payoff exams love: knowing which statements are axioms and which are theorems. "Probabilities sum to one" is not a postulate — completeness forces it. "Norms survive time" is not a postulate — Hermiticity of the Hamiltonian does it. This article states each postulate, points at the article owning its machinery, derives the two consistency results that hold the set together, and draws the axiom-versus-theorem boundary.

Key takeaways

P1 (state): a physical state is a unit ray $|\psi\rangle$ in a complex Hilbert space — global phase is unphysical.
P2 (observables): measurable quantities are Hermitian operators with complete eigenbases.
P3 (outcomes): a measurement of $\hat A$ can return only an eigenvalue of $\hat A$ — the spectrum is the menu.
P4 (Born rule): the probability of $a_n$ is $|\langle a_n|\psi\rangle|^2$ — developed quantitatively in the next article.
P5 (collapse): right after the measurement, the state is the normalized projection onto the outcome's eigenspace.
P6 (dynamics): between measurements, $i\hbar\,\partial_t|\psi\rangle = \hat H|\psi\rangle$ — deterministic, reversible, norm-preserving.
Consistency is earned: total probability 1 requires complete eigenbases, and predictions are basis-independent — both derived here.
Key idea: the postulates are the complete interface between Hilbert-space mathematics and laboratory numbers — six sentences; everything else in quantum mechanics is a theorem.

The core insight

A physical state is a unit ray $|\psi\rangle \in \mathcal H$ ; observables are Hermitian operators $\hat A$ ; measuring $\hat A$ yields an eigenvalue $a_n$ with probability $|\langle a_n|\psi\rangle|^2$ , collapsing the state onto the outcome's eigenspace; between measurements the state evolves unitarily under $i\hbar\,\partial_t|\psi\rangle = \hat H|\psi\rangle$ .

Where this lives today

The postulates are the interface layer of a technology stack. Quantum-computing textbooks open by restating them for finite dimensions — states in $(\mathbb C^2)^{\otimes n}$ , gates as unitaries, readout as projective measurement — so a quantum algorithm is choreography written in P1–P6, and hardware teams calibrate against these rules: preparation fidelity (P1), gate unitarity (P6), readout statistics (P4). The axioms are also under live interrogation: interference tests bound deviations from the Born rule's exact square (the next article's triple-slit story), and "collapse models" modifying P5/P6 are hunted in matter-wave interferometry. Ninety years in, the six sentences have survived every test — itself one of the most remarkable experimental facts in physics.

The physical picture

Think of the postulates as the specification sheet of a machine that converts preparations into predictions. Feed in a state (a unit vector), choose a question (a Hermitian operator), and the machine returns the possible readings (eigenvalues) with a probability attached to each (squared overlaps). Pressing the measurement button does two things at once: it prints one reading, and it overwrites the input — the state that comes out is the eigenstate matching the printed reading. Between button presses, an internal clockwork — the Schrödinger equation — rotates the state vector smoothly and reversibly.

Figure

The quantum machine — six postulates as one pipeline from preparation to prediction

Notice what the picture lacks: any dial showing "the actual value of $\hat A$ before measurement." The state fixes probabilities, not hidden readings — and the wave chapter's interference experiments (recall Double-Slit Experiment) rule out secretly attaching definite values. Notice also the machine's two different ways of changing a state: evolution (P6) smooth, deterministic, reversible; measurement (P5) abrupt, probabilistic, irreversible. The uneasy coexistence of these two update rules is the measurement problem, flagged in the FAQ and treated in earnest in the entanglement chapter.

The postulates, stated crisply

Each postulate is one sentence plus its equation; the machinery lives in an article this chapter has built or will build shortly. This is the assembly point — where the pieces are declared to be the theory.

Postulate 1 — states. The complete description of a system is a unit vector in a complex Hilbert space, vectors differing by a global phase describing the same state:

|\psi\rangle \in \mathcal H, \qquad \langle\psi|\psi\rangle = 1, \qquad |\psi\rangle \sim e^{i\theta}|\psi\rangle \qquad (1)

where: $\mathcal H$ is the Hilbert space whose three axioms were built in Hilbert Space and State Vectors, and the ray equivalence $\sim$ is that article's global-phase card (derived there; recalled, not re-proved, here). Packed into this one line: superposition is vector addition, normalization is unit length, and relative phases inside a sum are measurable while the overall phase never is.

Postulate 2 — observables. Every measurable quantity is a Hermitian operator on $\mathcal H$ whose eigenstates form a complete orthonormal basis:

\hat A = \hat A^\dagger, \qquad \hat A|a_n\rangle = a_n|a_n\rangle, \qquad \sum_n |a_n\rangle\langle a_n| = \hat 1 \qquad (2)

where: Hermiticity is the mirror property of Hermitian Operators and Observables, whose three theorems — real eigenvalues, orthogonal eigenstates, real expectations — are why this postulate is reverse engineering rather than decree: nothing but the Hermitian family delivers real, distinguishable, statistically consistent readings. The completeness clause is the resolution of identity of Completeness Relation and Resolution of Identity, and Derivation A below shows the probability interpretation dies without it.

Postulate 3 — outcomes. A measurement of $\hat A$ returns one of its eigenvalues $a_n$ , and nothing else — however exotic the state, the readout comes off the spectrum. Eigenvalue theory, degeneracy, and the discrete-versus-continuous distinction belong to Eigenvalues and Eigenstates (continuous spectra done properly at Continuous Spectra and Dirac Delta Normalization). The wave chapter cashed this postulate constantly: energy measurements on a well returned $E_n$ off the menu, never $1.7\,E_1$ .

Postulate 4 — the Born rule. If $\hat A$ is measured in normalized state $|\psi\rangle$ , the probability of outcome $a_n$ (nondegenerate case) is the squared overlap with its eigenstate:

P(a_n) = |\langle a_n|\psi\rangle|^2 \qquad (3)

where: $\langle a_n|\psi\rangle$ is the probability amplitude — the complex overlap Cauchy–Schwarz keeps inside the unit disk (Inner Product and Norm of States built that guarantee). The quantitative development — degenerate outcomes and the projector form, interference of amplitudes, continuous densities — is owned by the next article, Born Rule and Measurement Probabilities; here the rule takes its seat among the axioms.

Postulate 5 — collapse. Immediately after a measurement that returned $a_n$ , the state is the normalized projection of $|\psi\rangle$ onto the eigenspace of $a_n$ :

|\psi\rangle \;\xrightarrow{\text{result } a_n}\; \frac{\hat P_n|\psi\rangle}{\sqrt{\langle\psi|\hat P_n|\psi\rangle}} \qquad (4)

where: $\hat P_n$ is the projector onto the $a_n$ -eigenspace (Projection Operators owns projector theory). Two consequences: an instant re-measurement returns $a_n$ with certainty, and the superposition is destroyed. The finer mechanics — degenerate outcomes, partial measurements — belong to Measurement and Wavefunction Collapse, next in this chapter.

Postulate 6 — dynamics. Between measurements the state evolves according to the Schrödinger equation,

i\hbar\,\frac{\partial}{\partial t}|\psi(t)\rangle = \hat H\,|\psi(t)\rangle \qquad (5)

where: $\hat H$ is the Hamiltonian. This is the abstract-ket form of the law postulated in Time-Dependent Schrödinger Equation — a postulate judged by its scorecard, not a theorem. Because $\hat H$ is Hermitian, evolution preserves norms: the wave-language proof lives at Conservation of Normalization, and the equivalent statement — $e^{-i\hat H t/\hbar}$ is unitary — is developed at Unitary Operators and Transformations and in the dynamics chapter. One footnote: for composite systems the state space is the tensor product of the parts' spaces. That postulate does no work in this chapter, but it is where entanglement lives, and the entanglement chapter opens with it.

Which claims are axioms — and which are theorems

Here is the census, with each claim's true status and owning article. Rule of thumb: structure is postulated; consistency is proved.

claim	status	who proves it
states are unit rays in $\mathcal H$	postulate (P1)	Hilbert Space and State Vectors
global phase is unobservable	theorem (P1 + P4)	phase card, Hilbert Space and State Vectors
observables are Hermitian	postulate (P2)	motivated by Hermitian Operators and Observables
real eigenvalues, orthogonal eigenstates	theorem (P2)	Hermitian Operators and Observables
outcomes come from the spectrum	postulate (P3)	Eigenvalues and Eigenstates
$P(a_n) = \lvert\langle a_n\vert\psi\rangle\rvert^2$	postulate (P4)	Born Rule and Measurement Probabilities
probabilities sum to one	theorem (P1 + P2 + P4)	Derivation A below
predictions are basis-independent	theorem	Derivation B below
collapse onto the eigenspace	postulate (P5)	Measurement and Wavefunction Collapse
$i\hbar\,\partial_t\lvert\psi\rangle = \hat H\lvert\psi\rangle$	postulate (P6)	Time-Dependent Schrödinger Equation
norm conserved in time	theorem (P2 + P6)	Conservation of Normalization; unitarity later
$\langle\hat A\rangle = \langle\psi\vert\hat A\vert\psi\rangle$	theorem (P4)	Expectation Values and Variance

The table is worth memorizing in one compressed sentence: six postulates, and every "obviously true" bookkeeping fact you rely on — probabilities summing to one, norms surviving time, basis freedom — is a theorem earned from them.

Figure

Classical and quantum mechanics side by side — what each theory calls a state, a question, and an answer

Derivation ATotal probability equals one — and why the eigenbasis must be complete

We prove that the Born probabilities of Postulate 4 always form a valid probability distribution — and, more pointedly, that this consistency is purchased by the completeness clause of Postulate 2. The strategy is two passes over the sum $\sum_n P(a_n)$ : first with a complete eigenbasis, where the resolution of identity collapses the sum to the norm; then with one eigenvector deliberately removed, where a concrete state exhibits a probability deficit. The punchline: without a complete eigenbasis, P4 is not a probability statement at all.

Step 1. Write the total probability over all outcomes and expand each term as a sandwich:

\sum_n P(a_n) = \sum_n \langle\psi|a_n\rangle\langle a_n|\psi\rangle = \Big\langle\psi\Big|\Big(\sum_n |a_n\rangle\langle a_n|\Big)\Big|\psi\Big\rangle

Why the regrouping is legal: the inner product is linear in the ket slot — the insertion craft of Completeness Relation and Resolution of Identity.

Step 2. Apply completeness. By Postulate 2's clause $\sum_n|a_n\rangle\langle a_n| = \hat 1$ , the operator in the middle is the identity:

\sum_n P(a_n) = \langle\psi|\hat 1|\psi\rangle = \langle\psi|\psi\rangle = 1

Why this is Parseval in lab clothes: the identity behind the Pythagoras card of Hilbert Space and State Vectors, now saying outcome probabilities exhaust certainty.

\frac{4}{49} + \frac{9}{49} = \frac{13}{49} \approx 0.265

and $73.5\%$ of outcomes are simply missing. No renormalization fixes this state-independently — a different $|\psi\rangle$ has a different deficit.

Step 4. Read the moral. Completeness of the eigenbasis is precisely the condition "every measurement has an outcome" — why P2 demands it, why the spectral theory supplying it is load-bearing, and why continuous spectra need their own careful treatment later in the chapter. Verified: numerically — complete-basis probabilities sum to 1 exactly for the $(2,3,6)/7$ state; the truncated basis leaves $13/49 = 0.2653$ ; and a random 4-dimensional Hermitian observable's full eigenbasis reproduces $\sum_n|c_n|^2 = 1$ to machine precision (verify script, 27 checks green).

Go deeper · Derivation B · 1 minPredictions do not care which basis you compute in

Derivation BPredictions do not care which basis you compute in

We prove that the numbers the postulates output — probabilities and expectation values — are invariant under a unitary relabeling of the entire description: state, observable, and eigenvectors together. This is the license behind every "work in whatever basis is convenient" instruction you have ever followed, and the formal reason one physics could be written as Heisenberg's matrices and Schrödinger's waves. The strategy is direct computation: transform everything by one unitary $\hat U$ and watch the $\hat U$ 's cancel in conjugate pairs.

Step 1. Set up the relabeling. Let $\hat U$ be unitary ( $\hat U^\dagger\hat U = \hat 1$ ) and define the primed description: $|\psi'\rangle = \hat U|\psi\rangle$ , $\hat A' = \hat U\hat A\hat U^\dagger$ . Then eigenvectors map along: if $\hat A|a_n\rangle = a_n|a_n\rangle$ , applying $\hat U$ gives $\hat A'\,(\hat U|a_n\rangle) = a_n\,(\hat U|a_n\rangle)$ — same eigenvalues, transported eigenvectors $|a_n'\rangle = \hat U|a_n\rangle$ . Why the spectrum is untouched: the eigenvalue sits outside the vectors; conjugation re-dresses the operator without renaming its menu.

Step 2. Born probabilities. Compute the primed amplitude:

\langle a_n'|\psi'\rangle = \langle a_n|\hat U^\dagger \hat U|\psi\rangle = \langle a_n|\psi\rangle

Why the cancellation is exact: unitarity is the definition $\hat U^\dagger\hat U = \hat 1$ — the relabeling preserves every inner product, hence every amplitude, hence every $P(a_n)$ .

Step 4. Read what was proved. Predictions depend on geometric relationships — overlaps, spectra — not on coordinates. Heisenberg's 1925 arrays and Schrödinger's 1926 waves were two languages for one geometry, and the isomorphism certifying it (Riesz–Fischer, told at Hilbert Space and State Vectors) predates both. Practical corollary: compute in whatever basis makes your problem diagonal. Verified: numerically — for a random 4-dimensional Hermitian observable, random state, and random unitary, all four Born probabilities and the expectation value match before and after relabeling to $10^{-12}$ (verify script).

Limiting cases

case	result	meaning
eigenstate input, $\|\psi\rangle = \|a_k\rangle$	$P(a_n) = \delta_{nk}$ : outcome certain	the probabilistic postulate contains determinism as a special case
measurement repeated immediately	same outcome, probability 1	collapse (P5) + eigenstate limit of P4: measurements are reproducible
$[\hat A, \hat H] = 0$ , eigenstate prepared	statistics frozen for all time	the formalism echo of Stationary States and Separation of Variables
classical limit (narrow packets, slow potentials)	$\langle x\rangle, \langle p\rangle$ obey Newton approximately	Ehrenfest's theorem — wave chapter
two-outcome system	all six postulates run in $\mathbb C^2$	the qubit: minimal quantum mechanics (Example 1)
composite systems	$\mathcal H = \mathcal H_1 \otimes \mathcal H_2$	tensor-product postulate — dormant here, central in the entanglement chapter

Common trapThe postulates say the wavefunction collapses because measurement disturbs the particle

Two errors in one sentence. First, the postulates say nothing about a mechanism — P5 is a rule for updating the state given the outcome, silent on how or why; "the apparatus pokes the system" is a folk story that fails quantitatively (interaction-free measurements exist, and the double-slit article's visibility analysis shows available information, not mechanical disturbance, kills fringes). Second, do not promote collapse into a dynamical process governed by P6 — the two update rules are structurally different, and their coexistence is an open interpretive question (the measurement problem), not a solved mechanism. For exams: state P5 as the update rule, cite reproducibility as its motivation, and avoid disturbance language.

Worked examples

Example 1 — every postulate once, in a two-level system (confidence builder). A system lives in $\mathbb C^2$ with orthonormal basis $|1\rangle, |2\rangle$ . It is prepared in $|\psi\rangle = 0.6\,|1\rangle + 0.8\,|2\rangle$ , and the observable measured is $\hat A$ with matrix $\begin{pmatrix} 2 & 1 \\ 1 & 2 \end{pmatrix}$ in units of eV. Run the full pipeline: menu, probabilities, mean, collapse, and re-measurement.

Step 1. P1 check. $0.6^2 + 0.8^2 = 1$ : legal state.

Step 2. P2 + P3 — the menu. $(2-\lambda)^2 - 1 = 0$ gives $\lambda = 3$ and $1$ eV, eigenvectors $|\pm\rangle = (|1\rangle \pm |2\rangle)/\sqrt2$ . Only 3 eV or 1 eV can ever be read out.

Step 3. P4 — probabilities. $\langle +|\psi\rangle = (0.6 + 0.8)/\sqrt2 = 0.99$ : $P(3\,\mathrm{eV}) = \boxed{0.98}$ ; $\langle -|\psi\rangle = -0.14$ : $P(1\,\mathrm{eV}) = \boxed{0.02}$ . Sum: 1. Mean over many runs: $3(0.98) + 1(0.02) = \boxed{2.96\ \mathrm{eV}}$ — equal to the sandwich $\langle\psi|\hat A|\psi\rangle$ (machine-checked), previewing the theorem owned by Expectation Values and Variance.

Step 4. P5 — collapse. Suppose the readout was 3 eV (it usually is). The state is now $|+\rangle$ — the 0.6/0.8 preparation is gone. An immediate re-measurement returns 3 eV with probability $\boxed{1}$ : reproducibility, as the eigenstate-input limit demands.

Step 5. P6 — clockwork. Let $\hat H = \hat A$ . The collapsed $|+\rangle$ is an eigenstate: it only collects a global phase — statistics frozen. The original superposition, left unmeasured, would have oscillated, its components accumulating relative phase $e^{-i(3-1)\,\mathrm{eV}\,t/\hbar}$ and returning to the initial ray every $T = 2\pi\hbar/(2\,\mathrm{eV}) = \boxed{2.07\ \mathrm{fs}}$ . Determinism between measurements, dice at them.

Example 2 — an energy measurement in the well, start to finish (exam level). An electron in a 1 nm infinite square well ( $E_1 = 0.376$ eV — the spectrum is Infinite Square Well's) is prepared in $\Psi = (\psi_1 + \psi_2 + \psi_3)/\sqrt3$ . Find the possible energy readings and their statistics; then the state right after a measurement returns $E_2$ , and what a second measurement gives.

Step 1. Menu and probabilities. Possible readings: $E_1 = 0.376$ , $E_2 = 4E_1 = 1.504$ , $E_3 = 9E_1 = 3.384$ eV — nothing else. Each amplitude is $1/\sqrt3$ : $P = 1/3$ each.

Step 2. Statistics. $\langle E\rangle = \tfrac{14}{3}E_1 = \boxed{1.755\ \mathrm{eV}}$ ; spread $\sigma_E = \sqrt{\langle E^2\rangle - \langle E\rangle^2} = \tfrac{7\sqrt2}{3}E_1 = \boxed{1.241\ \mathrm{eV}}$ — wide, because the outcomes are far apart. (Machine-verified.)

Step 3. Collapse. The readout was $E_2$ . By P5 the state is now $\psi_2$ — one term survived, renormalized. The energy statistics changed discontinuously from a three-line spectrum to certainty.

Step 4. Re-measurement. $\psi_2$ is an energy eigenstate; P6 only dresses it with a phase no probability sees. A second energy measurement — now or an hour from now — returns $E_2$ with probability $\boxed{1}$ . (A position measurement in between would break this by rebuilding an energy superposition — the compatibility story of The Commutator and Compatible Observables.)

The man who ordered physics axiomatized

Origins · 1 minDavid Hilbert and how this was discovered

David Hilbert

1862–1943 · Göttingen

When Hilbert listed his famous problems at the 1900 Paris congress, the sixth was unlike the others: treat, by means of axioms, those physical sciences in which mathematics plays an important part — a demand that physics get the foundations he had just given geometry. For a quarter century it looked quixotic; then quantum mechanics arrived in his own town, mathematically chaotic, and Hilbert took it personally. His 1926–27 Göttingen lectures — prepared with his physics assistant Lothar Nordheim and a young John von Neumann — set out to find the axioms; the joint 1928 paper from them is the direct ancestor of this article's list. Hilbert's own toolkit fell short (the delta-function manipulations physicists loved had no rigorous home yet — the gap his quip that "physics is too difficult for physicists" was made for), but the program worked: within four years von Neumann's Mathematische Grundlagen delivered the axiomatization, built inside the very spaces that carry Hilbert's name. The sixth problem is still graded "partially solved" — but its quantum instalment is the standing counterexample to the idea that axiomatizing physics is empty formalism: the postulates are where every quantum technology's design document begins. (His other cameos — coining "spectrum"; the "Hilbert space" naming irony — are told at Position Operator and Hilbert Space and State Vectors.)

History

1899

— Hilbert's Grundlagen der Geometrie shows what an axiomatized physical-adjacent theory looks like

1900

— the sixth Paris problem asks for the axiomatization of physics

1925

— Heisenberg, Born and Jordan build matrix mechanics; Born writes Hilbert asking about the strange new arrays

1926

— Schrödinger's wave mechanics appears; two rival formalisms, one physics — begging for common axioms

1927

— Hilbert, von Neumann and Nordheim lecture on the foundations of quantum mechanics in Göttingen

1932

— von Neumann's Mathematische Grundlagen completes the axiomatization this article states

Going further

References. Griffiths & Schroeter, Introduction to Quantum Mechanics, 3rd ed., Ch. 3 (Formalism) · Sakurai & Napolitano, Modern Quantum Mechanics, 2nd ed., Ch. 1 · Shankar, Principles of Quantum Mechanics, 2nd ed., Ch. 4 (The Postulates — A General Discussion) · Cohen-Tannoudji, Diu & Laloë, Quantum Mechanics, Vol. 1, Ch. III · Townsend, A Modern Approach to Quantum Mechanics, 2nd ed., Ch. 1–3 · Nielsen & Chuang, Quantum Computation and Quantum Information, Ch. 2.

Provenance. All postulates and equations are standard textbook material (Griffiths Ch. 1–3, Sakurai Ch. 1, Shankar Ch. 4, Cohen-Tannoudji Vol. 1 Ch. III). A verifier should confirm: probability normalization via the resolution of the identity; basis-independence of predictions under unitary relabeling; both worked examples. For this edition, verification was machine-executed (numpy, 27 checks): the $(2,3,6)/7$ completeness-deficit demo ( $13/49 = 0.2653$ ); representation invariance to $10^{-12}$ ; unitarity of $e^{-i\hat Ht/\hbar}$ ; the two-level pipeline ( $P = 0.98/0.02$ , $\langle\hat A\rangle = 2.96$ eV both routes, period $2.07$ fs); the ISW three-mix statistics ( $\langle E\rangle = 1.755$ eV, $\sigma_E = 1.241$ eV at $L = 1$ nm). Per the chapter ownership map: the ray/phase derivation belongs to Hilbert Space and State Vectors, the Hermiticity theorems to Hermitian Operators and Observables, eigenvalue theory to Eigenvalues and Eigenstates, completeness theory to Complete Orthonormal Bases and Completeness Relation and Resolution of Identity, projector theory to Projection Operators, the Born rule's development to Born Rule and Measurement Probabilities, collapse mechanics to Measurement and Wavefunction Collapse, the expectation-value theorem to Expectation Values and Variance — this article states the axioms and proves only the two assembly-level consistency results.

Builds on: Hilbert Space and State Vectors · Dirac Bra-Ket Notation · Hermitian Operators and Observables · Inner Product and Norm of States

Leads to: Born Rule and Measurement Probabilities · Measurement and Wavefunction Collapse · Expectation Values and Variance · Completeness Relation and Resolution of Identity

Why six postulates? My textbook lists four (or five, or seven).

The physics is identical; authors package it differently. Some merge outcomes + probabilities + collapse into one "measurement postulate" (Shankar); some split the state postulate into "vector space" and "ray" clauses; some add the tensor-product postulate explicitly. What never changes: a state postulate, an observable postulate, a measurement block (menu, probabilities, update), and a dynamics postulate. Count the blocks, not the bullets — and on an exam, use your course's numbering.

Are the postulates derivable from something deeper?

Not currently — that is what "postulate" means operationally. There are serious reconstruction programs: Gleason's theorem gets the Born rule's form from weak assumptions given the Hilbert-space frame (next article), decoherence explains why collapse looks the way it does, and quantum-information axiomatizations derive the framework from operational principles. Each illuminates; none has displaced the standard list. The honest classroom position: the postulates are judged, like Newton's laws were, by an unbroken century of experimental victories.

Doesn't Postulate 5 contradict Postulate 6? Evolution is deterministic but collapse is random.

They have disjoint jurisdictions (P6 between measurements, P5 at them), so no formal contradiction — but the coexistence is uncomfortable, because an apparatus is itself atoms that should obey P6. Applying P6 to system-plus-apparatus produces an entangled superposition, not a definite outcome; where definiteness comes from is the measurement problem. Decoherence explains why superpositions become practically invisible; whether that dissolves the problem is contested — the entanglement chapter's foundations articles take it up. For calculations, the split framework works flawlessly.

Where do operators like x̂ = multiplication and p̂ = −iℏ d/dx fit into the postulates?

They are the representation of P2 for a particle on a line — specific operators for specific quantities, inherited from the wave chapter (Position Operator, Momentum Operator). The postulates say "some Hermitian operator," not which. Assigning operators to classical quantities is a separate craft (canonical quantization) guided by $[\hat x, \hat p] = i\hbar$ — see The Canonical Commutation Relation, later this chapter.

Is the state |ψ⟩ a property of one system or a recipe about many copies?

The postulates work either way: the Born rule is operationally about ensembles (frequencies over many identically prepared runs — the reading built at Born Statistical Interpretation), while collapse updates the description of a single run. Whether $|\psi\rangle$ is ontic (physical) or epistemic (a catalogue of expectations) is a live interpretation question (the PBR theorem constrains epistemic readings). Nothing in this chapter's calculations depends on the answer.

Why must probabilities come from the square of the amplitude and not some other power?

Consistency arguments pin the square down tightly: Gleason's theorem (dimension ≥ 3) derives it from non-contextual probability assignments on projectors, and experiment agrees to high precision — interference tests bound "higher-order" deviations at the sub-thousandth level. The next article develops the rule quantitatively and tells the experimental story; here, note only that the square is the least arbitrary of the axioms once the Hilbert-space frame is accepted.