Minimal Postulates

The minimal postulates are a compact specification of ordinary nonrelativistic quantum mechanics. They identify the mathematical objects used for states, measurements, probabilities, closed-system dynamics, and composite systems. Once a concrete model supplies a Hilbert space, Hamiltonian, preparation, and measurement, the postulates turn those inputs into testable probability distributions.

This statement uses density operators and generalized measurements because that language includes pure states and projective measurements as special cases. The finite-dimensional, wave-mechanics, and density-matrix presentations unpack the same structure for different purposes.

Helpful background. Quantum states and the Born rule supply the state and probability language; unitary time evolution explains closed-system dynamics.

The Five Postulates at a Glance

1. States: a system is associated with a complex Hilbert space $\mathcal H$, and a state is represented by a density operator $\rho$ on $\mathcal H$. Pure states are rank-one projectors, equivalently rays in $\mathcal H$.
2. Measurements and observables: a measurement is represented by a positive-operator-valued measure, or by an instrument when postmeasurement states are required. Sharp observables are represented by self-adjoint operators and their spectral projectors.
3. Born rule: the probability of an outcome is the state–effect pairing $p(a)=\operatorname{Tr}(\rho E_a)$.
4. Closed-system dynamics: between interventions, an isolated system evolves unitarily; a Hamiltonian generates the unitary evolution.
5. Composition: distinguishable systems compose by tensor product, $\mathcal H_{AB}=\mathcal H_A\otimes\mathcal H_B$, with qualifications for identical particles and constrained sectors.

These statements form one convenient package, not the unique possible axiomatization. Their role and package-relative minimality are discussed in Why Postulates Matter.

Scope and Conventions

The statements below assume:

• ordinary nonrelativistic quantum mechanics with an external time parameter;
• complex Hilbert spaces and standard probabilistic measurement outcomes;
• a declared system boundary separating the modeled system from controls and records;
• normalized states unless explicitly called unnormalized;
• finite or countable outcome sets in displayed formulas.

Continuous observables require projection-valued or positive-operator-valued measures on measurable sets. Infinite-dimensional systems also require attention to operator domains, self-adjointness, and convergence. Those mathematical qualifications do not change the basic state–measurement–probability pairing, but they do prevent one from treating every operator as an everywhere-defined matrix.

The postulates specify a formal framework. They do not select the correct Hamiltonian for an atom, identify which detector event corresponds to which effect, justify an approximation, or settle an interpretation. See Assumptions and Scope for the domain of the package and What the Postulates Do Not Say for its deliberate omissions.

Postulate 1: States

To each quantum system one associates a complex Hilbert space $\mathcal H$. A general normalized state is represented by a positive trace-class operator $\rho$ on $\mathcal H$ with unit trace.

Thus

$$\rho\geq 0, \qquad \operatorname{Tr}\rho=1.$$

Positivity means

$$\langle\phi\rvert\rho\lvert\phi\rangle \geq 0$$

for every $\lvert\phi\rangle\in\mathcal H$. It is the condition that ultimately prevents Born-rule probabilities from becoming negative.

A pure state is represented by a rank-one projector

$$\rho_\psi = \lvert\psi\rangle\langle\psi\rvert, \qquad \langle\psi\mid\psi\rangle=1.$$

Multiplying the vector by a global phase changes no projector:

$$\lvert\psi\rangle \longmapsto e^{i\alpha}\lvert\psi\rangle, \qquad \rho_\psi \longmapsto \rho_\psi.$$

The physical pure state is therefore a ray, not a particular normalized vector. Density operators also include mixed states and reduced states of entangled systems. The set of states is convex: if $\rho_j$ are states and $q_j$ is a probability distribution, then

$$\rho = \sum_j q_j\rho_j$$

is a state. The coefficients may describe classical randomization among preparation procedures, but a given density operator can generally have more than one ensemble decomposition. The density operator, rather than a preferred decomposition, controls all predictions made within the formalism.

The state postulate says what object carries the probability-relevant information assigned to a preparation. It does not say how a laboratory preparation is calibrated or whether the state is ontic, epistemic, relational, or something else. Those are additional physical and interpretive questions.

For the canonical development, see Quantum States and Density Operators.

Postulate 2: Measurements and Observables

A measurement with outcomes $a$ is represented probabilistically by positive effects $\{E_a\}$ that sum to the identity. If later states conditioned on the outcomes are needed, the measurement is represented by a quantum instrument $\{\mathcal I_a\}$.

For a finite or countable outcome set, a positive-operator-valued measure (POVM) obeys

$$E_a\geq 0, \qquad \sum_a E_a=I.$$

Each $E_a$ represents the event “the recorded outcome is $a$.” The complete set $\{E_a\}$ represents the measurement context. A single effect does not by itself specify what other outcomes were possible.

Sharp observables

A discrete sharp observable is represented by a self-adjoint operator

$$A = \sum_a aP_a,$$

where its spectral projectors satisfy

$$\begin{aligned} P_aP_b&=\delta_{ab}P_a,\\ \sum_aP_a&=I. \end{aligned}$$

The projectors $\{P_a\}$ form a projection-valued measure (PVM), a special kind of POVM with $E_a=P_a$. Degeneracy means that $P_a$ may project onto a multidimensional eigenspace. The outcome is then the eigenvalue $a$, not a unique eigenvector.

For a general self-adjoint operator with continuous spectrum, the spectral theorem writes

$$A = \int_{\mathbb R}\lambda\,dP^A(\lambda),$$

and the event that the result lies in a measurable set $\Delta$ is represented by $P^A(\Delta)$. The operator $A$ is a compact encoding of a sharp real-valued measurement; generalized measurements need not arise as the spectral measure of one system operator.

Instruments and disturbance

A POVM determines outcome probabilities but not the state left behind. A quantum instrument assigns to each outcome a linear, completely positive, trace-nonincreasing map

$$\mathcal I_a: \rho\longmapsto\mathcal I_a(\rho).$$

The sum $\sum_a\mathcal I_a$ must be trace preserving. Its probability effect is fixed by

$$\operatorname{Tr}\mathcal I_a(\rho) = \operatorname{Tr}(\rho E_a)$$

for every input state. If the outcome occurs with nonzero probability, the conditional state is

$$\rho_a = \frac{\mathcal I_a(\rho)} {\operatorname{Tr}\mathcal I_a(\rho)}.$$

For the ideal Lüders implementation of a projective measurement,

$$\mathcal I_a(\rho) = P_a\rho P_a.$$

This update is an additional statement about the measurement implementation. The projectors alone do not imply it: distinct instruments can have the same effects and therefore the same outcome statistics while disturbing the state differently.

See Observables, POVMs: First Encounter, and State Update Rule for the detailed treatments.

Postulate 3: The Born Rule

Given a state $\rho$ and a measurement effect $E_a$, the outcome probability is given by their trace pairing.

$$p(a\mid\rho,\{E_b\}) = \operatorname{Tr}(\rho E_a).$$

This trace pairing converts quantum data into an ordinary probability distribution. Positivity follows from $\rho\geq0$ and $E_a\geq0$, while normalization follows from POVM completeness:

$$\begin{aligned} p(a)&\geq0,\\ \sum_a p(a) &= \operatorname{Tr}\left( \rho\sum_aE_a \right)\\ &= \operatorname{Tr}\rho =1. \end{aligned}$$

For a pure state and a projective measurement,

$$p(a) = \langle\psi\rvert P_a\lvert\psi\rangle.$$

If $P_a=\lvert a\rangle\langle a\rvert$ is rank one, this becomes the familiar squared-amplitude rule

$$p(a) = \lvert\langle a\mid\psi\rangle\rvert^2.$$

For a sharp observable $A=\sum_a aP_a$, the expectation value is

$$\mathbb E_\rho[A] = \sum_a a\,p(a) = \operatorname{Tr}(\rho A),$$

for bounded $A$. An unbounded observable requires a finite absolute first spectral moment; interpreting the displayed trace as an ordinary operator product may require stronger conditions. See Trace rule. For a continuous sharp observable,

$$\Pr(A\in\Delta) = \operatorname{Tr}\!\left[ \rho P^A(\Delta) \right].$$

The Born rule does not choose the measurement, assign a postmeasurement state, or claim that one outcome rather than another must occur in an individual trial. It supplies the probability distribution conditional on a specified state and measurement model. Its detailed forms and consistency properties belong to Born Rule.

Postulate 4: Closed-System Dynamics

Between interventions, the state of a closed quantum system evolves by a unitary operator. For a time-independent system, a strongly continuous unitary group has a self-adjoint Hamiltonian generator.

From time $t_1$ to $t_2$,

$$\rho(t_2) = U(t_2,t_1)\rho(t_1) U(t_2,t_1)^\dagger,$$

where

$$U(t_2,t_1)^\dagger U(t_2,t_1) = I.$$

Unitary evolution preserves all defining state conditions. In particular,

$$\operatorname{Tr}\rho(t_2) = \operatorname{Tr}\rho(t_1),$$

and positivity is preserved because

$$\langle\phi\rvert\rho(t_2)\lvert\phi\rangle = \langle\chi\rvert\rho(t_1)\lvert\chi\rangle \geq0,$$

with

$$\lvert\chi\rangle = U(t_2,t_1)^\dagger\lvert\phi\rangle.$$

For finite matrices with a sufficiently regular time-dependent Hamiltonian, the propagator satisfies

$$\begin{aligned} i\hbar \frac{\partial}{\partial t} U(t,t_0) &= H(t)U(t,t_0),\\ U(t_0,t_0)&=I. \end{aligned}$$

For a time-independent Hamiltonian,

$$U(t,t_0) = \exp\!\left[ -\frac{i}{\hbar}H(t-t_0) \right].$$

Pure states obey the Schrödinger equation,

$$i\hbar \frac{d}{dt}\lvert\psi(t)\rangle = H(t)\lvert\psi(t)\rangle,$$

while density operators obey the von Neumann equation,

$$i\hbar \frac{d\rho}{dt} = [H,\rho].$$

For unbounded Hamiltonians, the differential equations hold on the appropriate domains and require regularity; the unitary propagator is the primary statement. A time-dependent self-adjoint operator at each instant does not by itself guarantee such a propagator.

The word closed matters. With an initially uncorrelated environment in a fixed state $\rho_E$ independent of the system input, tracing a joint unitary evolution gives a completely positive trace-preserving channel:

$$\rho_S' = \operatorname{Tr}_E\!\left[ U_{SE} (\rho_S\otimes\rho_E) U_{SE}^\dagger \right].$$

This representation connects open dynamics to the state, dynamics, and composition postulates; it does not make the reduced evolution unitary. See Unitary Time Evolution for the canonical closed-system treatment.

Postulate 5: Composition

For distinguishable systems $A$ and $B$, the Hilbert space of the composite system is the tensor product of the subsystem Hilbert spaces.

Thus

$$\mathcal H_{AB} = \mathcal H_A\otimes\mathcal H_B.$$

Independent preparations have product states

$$\rho_{AB} = \rho_A\otimes\rho_B,$$

but many joint states are not products. A mixed state can be correlated while still being a convex mixture of products,

$$\rho_{AB}=\sum_j p_j\rho_A^{(j)}\otimes\rho_B^{(j)}.$$

Such states are separable. Entangled states cannot be expressed in this form (or in its trace-norm closure in infinite dimensions). For a pure joint state, nonproduct and entangled are equivalent. The composition rule supplies the space in which these distinct kinds of correlation occur.

Subsystem states are recovered by partial trace:

$$\rho_A = \operatorname{Tr}_B\rho_{AB}, \qquad \rho_B = \operatorname{Tr}_A\rho_{AB}.$$

A measurement performed only on $A$ has effects $E_a\otimes I_B$. Its probabilities can be computed either from the joint state or from the reduced state:

$$\begin{aligned} p(a) &= \operatorname{Tr}_{AB}\!\left[ \rho_{AB}(E_a\otimes I_B) \right]\\ &= \operatorname{Tr}_A(\rho_AE_a). \end{aligned}$$

This identity is a basic compatibility check between composition and the Born rule: all local statistics are encoded in the reduced density operator.

For finite dimensions,

$$\dim\mathcal H_{AB} = (\dim\mathcal H_A) (\dim\mathcal H_B).$$

For identical particles, one does not interpret arbitrary tensor factors as permanently labeled particles. Bosonic and fermionic states occupy symmetric and antisymmetric sectors, respectively, and variable particle number is naturally described with Fock space. Gauge constraints, superselection rules, and effective subsystem choices can impose further qualifications. The simple tensor-product statement is therefore exact for declared distinguishable subsystems and a starting point, not the whole story, for every physical decomposition.

See Tensor Products, Bipartite Systems, and Reduced States for details.

Worked Example: Creating and Measuring a Bell Pair

Consider two distinguishable qubits, $Q$ and $A$. The composition postulate gives

$$\mathcal H_{QA} = \mathbb C^2\otimes\mathbb C^2.$$

Prepare

$$\lvert\psi_0\rangle = \lvert+\rangle_Q\otimes\lvert0\rangle_A,$$

where

$$\lvert+\rangle = \frac{\lvert0\rangle+\lvert1\rangle} {\sqrt2}.$$

The initial density operator is

$$\rho_0 = \lvert\psi_0\rangle \langle\psi_0\rvert.$$

Apply a controlled-NOT unitary with $Q$ as control and $A$ as target. Closed-system dynamics gives

$$\begin{aligned} \lvert\Phi^+\rangle &= U_{\mathrm{CNOT}}\lvert\psi_0\rangle\\ &= \frac{ \lvert00\rangle+\lvert11\rangle }{\sqrt2}, \end{aligned}$$

and hence

$$\rho_{\Phi} = \lvert\Phi^+\rangle \langle\Phi^+\rvert.$$

Measure qubit $Q$ in the computational basis. The two joint effects are

$$\begin{aligned} E_0&= \lvert0\rangle\langle0\rvert \otimes I_A,\\ E_1&= \lvert1\rangle\langle1\rvert \otimes I_A. \end{aligned}$$

The Born rule gives

$$\begin{aligned} p(0) &= \operatorname{Tr}(\rho_\Phi E_0) =\frac12,\\ p(1) &= \operatorname{Tr}(\rho_\Phi E_1) =\frac12. \end{aligned}$$

If the measurement is implemented by the ideal Lüders instrument, the unnormalized outcome branches are

$$\begin{aligned} \mathcal I_0(\rho_\Phi) &= \frac12\lvert00\rangle\langle00\rvert,\\ \mathcal I_1(\rho_\Phi) &= \frac12\lvert11\rangle\langle11\rvert. \end{aligned}$$

After normalizing, outcome $0$ leaves $\lvert00\rangle$ and outcome $1$ leaves $\lvert11\rangle$. If the outcome is ignored, the state becomes

$$\rho' = \frac12\lvert00\rangle\langle00\rvert + \frac12\lvert11\rangle\langle11\rvert.$$

This final mixture has the same computational-basis correlations as the Bell state but lacks its coherence between $\lvert00\rangle$ and $\lvert11\rangle$. The distinction can be detected by other joint measurements.

A joint $X\otimes X$ measurement makes the lost coherence visible: its mean is $1$ for the Bell state and $0$ for the final mixture. The mixture is correlated but separable, illustrating why “not a product” is not sufficient to establish mixed-state entanglement.

Every postulate entered the calculation:

• the state postulate represented the preparation and final branches;
• the measurement postulate supplied effects and an instrument;
• the Born rule supplied outcome probabilities;
• the dynamics postulate supplied the controlled-NOT unitary;
• the composition postulate supplied the two-qubit state space and entanglement.

What “Minimal” Means Here

Minimality is relative to the chosen language. This package treats density operators, POVMs, instruments, unitary evolution, and tensor products as the most economical shared vocabulary for modern applications. Other presentations redistribute the same content:

• a pure-state introduction starts with rays and adds density operators later;
• a wave-mechanics presentation represents states by functions and observables by differential operators;
• a projective-measurement presentation introduces self-adjoint operators before POVMs;
• an open-systems presentation takes quantum channels as primitive;
• reconstruction programs replace textbook postulates with operational or information-theoretic principles and then derive Hilbert-space structure.

Some ingredients can be represented using others. A POVM can be realized as a projective measurement on a larger system, and a channel can be realized through unitary evolution on a larger system followed by a partial trace. Such representation theorems show compatibility among formulations. They do not remove the need to say which systems, environments, measurements, and accessibility assumptions define the model.

Ideal projective state update is not listed as an independent sixth postulate here. In the general formulation it is one particular instrument. A textbook may instead list “collapse” separately and restrict the measurement postulate to projectors. Those packages can agree on their common domain if they use the same instrument for the same experiment.

Exercises

1. Pure states are rays. Let $\lvert\psi'\rangle=e^{i\alpha}\lvert\psi\rangle$. Show that every POVM assigns the same probabilities to $\lvert\psi'\rangle$ and $\lvert\psi\rangle$.

Solution

The two vectors define the same density operator:

$$\begin{aligned} \rho_{\psi'} &= \lvert\psi'\rangle\langle\psi'\rvert\\ &= e^{i\alpha}\lvert\psi\rangle \langle\psi\rvert e^{-i\alpha}\\ &= \rho_\psi. \end{aligned}$$

Therefore, for every effect $E_a$,

$$\operatorname{Tr}(\rho_{\psi'}E_a) = \operatorname{Tr}(\rho_\psi E_a).$$

No measurement represented within the formalism can distinguish two representatives of the same ray.

2. Born-rule normalization. Let $\rho$ be a density operator and let $\{E_a\}$ be a POVM. Prove that $p(a)=\operatorname{Tr}(\rho E_a)$ is nonnegative and sums to one.

Solution

Because $\rho\geq0$, it has a positive square root $\rho^{1/2}$. Cyclicity of the trace gives

$$p(a) = \operatorname{Tr}\!\left( \rho^{1/2}E_a\rho^{1/2} \right).$$

The operator inside the trace is positive because $E_a\geq0$, so $p(a)\geq0$. Completeness gives

$$\begin{aligned} \sum_a p(a) &= \operatorname{Tr}\!\left( \rho\sum_aE_a \right)\\ &= \operatorname{Tr}(\rho I) =1. \end{aligned}$$

3. A qubit prediction. A qubit starts in $\lvert0\rangle$ and evolves under

$$U(\theta) = \exp\!\left( -\frac{i\theta}{2}\sigma_y \right).$$

It is then measured in the computational basis. Find the two outcome probabilities.

Solution

Using $\sigma_y^2=I$,

$$U(\theta) = \cos\frac{\theta}{2}\,I - i\sin\frac{\theta}{2}\,\sigma_y.$$

Since $\sigma_y\lvert0\rangle=i\lvert1\rangle$,

$$U(\theta)\lvert0\rangle = \cos\frac{\theta}{2}\lvert0\rangle + \sin\frac{\theta}{2}\lvert1\rangle.$$

The projectors are $P_0=\lvert0\rangle\langle0\rvert$ and $P_1=\lvert1\rangle\langle1\rvert$, so

$$\begin{aligned} p(0)&=\cos^2\frac{\theta}{2},\\ p(1)&=\sin^2\frac{\theta}{2}. \end{aligned}$$

Their sum is one, and the limiting cases $\theta=0$ and $\theta=\pi$ give certain outcomes $0$ and $1$, respectively.

4. Unitary preservation. Prove that $\rho'=U\rho U^\dagger$ is a density operator whenever $\rho$ is a density operator and $U$ is unitary.

Solution

Hermiticity follows from

$$(\rho')^\dagger = U\rho^\dagger U^\dagger = \rho'.$$

For every $\lvert\phi\rangle$,

$$\langle\phi\rvert\rho'\lvert\phi\rangle = \langle\chi\rvert\rho\lvert\chi\rangle \geq0,$$

where $\lvert\chi\rangle=U^\dagger\lvert\phi\rangle$. Thus $\rho'\geq0$. Finally,

$$\operatorname{Tr}\rho' = \operatorname{Tr}(U\rho U^\dagger) = \operatorname{Tr}(\rho U^\dagger U) =1.$$

All three defining properties are preserved.

5. Same POVM, different disturbance. For the qubit PVM $P_a=\lvert a\rangle\langle a\rvert$, compare the Lüders instrument

$$\mathcal I_a(\rho) = P_a\rho P_a$$

with the measure-and-prepare instrument

$$\mathcal J_a(\rho) = \operatorname{Tr}(P_a\rho) \lvert+\rangle\langle+\rvert.$$

Show that they have the same outcome probabilities but generally different conditional states.

Solution

For the Lüders instrument,

$$\operatorname{Tr}\mathcal I_a(\rho) = \operatorname{Tr}(P_a\rho P_a) = \operatorname{Tr}(\rho P_a).$$

For the measure-and-prepare instrument,

$$\begin{aligned} \operatorname{Tr}\mathcal J_a(\rho) &= \operatorname{Tr}(P_a\rho) \operatorname{Tr} \left( \lvert+\rangle\langle+\rvert \right)\\ &= \operatorname{Tr}(\rho P_a). \end{aligned}$$

Thus both instruments have effects $P_a$. When $p(a)>0$, the Lüders conditional state is $\lvert a\rangle\langle a\rvert$, whereas the second device prepares $\lvert+\rangle\langle+\rvert$ regardless of $a$. Outcome statistics alone therefore do not determine disturbance.

6. Local statistics of a Bell state. For

$$\lvert\Phi^+\rangle = \frac{\lvert00\rangle+\lvert11\rangle}{\sqrt2},$$

compute the reduced state of the first qubit and use it to find the probabilities of measuring either $\sigma_z$ or $\sigma_x$ on that qubit.

Solution

Expanding the projector and tracing the second qubit removes the cross terms:

$$\rho_Q = \operatorname{Tr}_A \left( \lvert\Phi^+\rangle \langle\Phi^+\rvert \right) = \frac{I}{2}.$$

For any rank-one qubit projector $P$,

$$\operatorname{Tr}\!\left( \frac{I}{2}P \right) = \frac12.$$

Therefore both outcomes of $\sigma_z$ occur with probability $1/2$, and both outcomes of $\sigma_x$ occur with probability $1/2$. The joint state is pure, but each local state is maximally mixed.

7. Identical-particle qualification. Two identical spinless particles each have a two-dimensional one-particle space spanned by $\lvert0\rangle$ and $\lvert1\rangle$. Find the dimensions of the symmetric and antisymmetric two-particle subspaces.

Solution

The symmetric subspace has basis

$$\lvert00\rangle, \qquad \frac{\lvert01\rangle+\lvert10\rangle}{\sqrt2}, \qquad \lvert11\rangle,$$

so its dimension is three. The antisymmetric subspace has the single basis vector

$$\frac{\lvert01\rangle-\lvert10\rangle}{\sqrt2},$$

so its dimension is one. The full labeled tensor product has dimension four, but identical bosons or fermions occupy only the appropriate symmetry sector.

8. Diagnosing an incomplete prediction. A calculation specifies a normalized state $\rho$ and a Hamiltonian $H$, then claims to predict the probability of a detector click without giving any detector operator. Which ingredient is missing, and why can it not be inferred from $\rho$ and $H$ alone?

Solution

The calculation lacks a measurement effect $E_{\mathrm{click}}$, or a fuller instrument if the post-click state matters. The state and Hamiltonian determine the modeled preparation and evolution, but many physically different detectors can be applied to the same evolved state. Their effects give different probabilities:

$$p(\mathrm{click}) = \operatorname{Tr} \left( \rho(t)E_{\mathrm{click}} \right).$$

Neither $\rho(t)$ nor $H$ selects the detector calibration or outcome event. That information belongs to the measurement model.

References

• P. Busch, P. J. Lahti, J.-P. Pellonpää, and K. Ylinen, Quantum Measurement, Springer, 2016 — modern mathematical treatment of observables and measurement.
• P. A. M. Dirac, The Principles of Quantum Mechanics, 4th ed., Oxford University Press, 1958 — classic ray, observable, and transformation language.
• K. Kraus, States, Effects, and Operations: Fundamental Notions of Quantum Theory, Springer, 1983 — effects, instruments, and quantum operations.
• M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information, 10th anniversary ed., Cambridge University Press, 2010 — finite-dimensional states, operations, measurements, and composite systems.
• A. Peres, Quantum Theory: Concepts and Methods, Kluwer, 1995 — operationally careful treatment of states, measurements, and composition.
• M. Reed and B. Simon, Methods of Modern Mathematical Physics, Vol. I: Functional Analysis, revised and enlarged ed., Academic Press, 1980 — operator-theoretic foundations needed in infinite dimension.
• J. von Neumann, Mathematical Foundations of Quantum Mechanics, Princeton University Press, 1955 — Hilbert-space, spectral, and measurement foundations.