Quantization vs Classical Limit

Quantization and the classical limit point in opposite conceptual directions.

Quantization starts with classical data and tries to construct a quantum model. The classical limit starts with a quantum model and asks when selected quantum predictions are well approximated by a classical description.

They are related, but they are not inverse functions. A good quantum model should reproduce its intended classical regime, but a classical model does not uniquely determine a quantum theory, and a quantum theory can have more than one useful classical approximation.

For a short myth-correction format, see Common Misstatements About the Classical Limit.

Helpful background. Poisson Brackets supplies the classical algebra; Canonical Commutation Relations introduces its quantum counterpart.

Quantization Constructs a Model; a Classical Limit Approximates It

The distinction is easiest to see as a map:

Direction	Starting point	Output	Typical question
Quantization	Classical phase space, observables, Hamiltonian, or action	Hilbert space, operators, commutation relations, quantum Hamiltonian, amplitudes	What quantum model should represent this classical system?
Classical limit	Quantum states, operators, dynamics, and measurements	Approximate classical variables, trajectories, probabilities, records, or actions	When does this quantum model look classical?

The Correspondence Principle says that a successful quantum theory must recover classical physics where classical physics is already accurate. It does not say that quantization is automatic or that the classical limit is a single theorem.

Quantization: From Classical Data to Quantum Model

In elementary canonical quantization, one begins with classical coordinates $q_i$ and canonical momenta $p_i$ satisfying Poisson brackets

$$\{q_i,p_j\}_{\mathrm{PB}} = \delta_{ij}.$$

The quantum model replaces them by operators $Q_i$ and $P_j$ satisfying the canonical commutation relations

$$[Q_i,P_j] = i\hbar\delta_{ij}I.$$

A classical Hamiltonian function $h(q,p)$ is then represented by a Hamiltonian operator $H$ that generates time evolution. For the familiar particle in a potential,

$$h(q,p) = \frac{p^2}{2m}+V(q)$$

becomes

$$H = \frac{P^2}{2m}+V(Q).$$

In the position representation this gives

$$H = - \frac{\hbar^2}{2m}\frac{d^2}{dx^2} + V(x).$$

This familiar example can make quantization look like a mechanical substitution rule. That impression is misleading. A real quantization also requires a Hilbert space, domains for unbounded operators, boundary conditions, symmetry representation, self-adjoint Hamiltonian, and a measurement interpretation for the chosen observables.

Spin gives a warning sign. Spin-$1/2$ is not obtained by taking an ordinary classical vector and applying $q\mapsto Q$, $p\mapsto P$. It requires a finite-dimensional representation of the rotation algebra. Quantum theory contains structures that may have only indirect or limiting classical analogues.

Classical Limit: From Quantum Model to Classical Behavior

The classical limit asks a different question. Given a quantum model, when can some of its predictions be replaced by a classical description?

For a wave packet in a smooth potential, one might ask whether

$$\langle x\rangle(t), \qquad \langle p\rangle(t)$$

follow approximately Newtonian motion. For a WKB state, one might ask whether the phase is controlled by a classical action. For a macroscopic detector, one might ask whether environmental decoherence makes interference between pointer alternatives locally negligible.

The Classical Limit page surveys these mechanisms. The Semiclassical Limit Overview focuses on action phases such as

$$\exp\left(\frac{i}{\hbar}S\right),$$

where stationary phase selects classical equations at leading order when $S/\hbar\gg1$.

This direction is not “unquantization.” It does not recover every microscopic detail of a quantum state as a classical object. It extracts a controlled approximation for specified states, observables, scales, and resolutions.

Why Neither Direction Is Unique

Quantization is not unique in full generality. Ambiguities can enter through:

• operator ordering when classical functions contain noncommuting variables;
• the choice of Hilbert space and domains;
• boundary conditions and topology;
• coordinate changes and canonical transformations;
• spin, internal degrees of freedom, and symmetry representations;
• path-integral measures and regularization choices;
• inequivalent field quantizations in systems with infinitely many degrees of freedom.

The classical limit is also not unique. The same quantum model can have different classical-looking regimes depending on which states and observables are chosen. A harmonic oscillator energy eigenstate, a coherent state, a thermal state, and a decohered oscillator pointer may all call for different classical descriptions.

Conversely, different quantum models can share the same leading classical equations while differing in their spectra, phases, spin content, anomalies, tunneling amplitudes, or higher-order corrections in $\hbar$. The leading classical equations are therefore not enough information to reconstruct the full quantum theory.

Operator Ordering Problem Preview

The simplest ordering problem appears when a classical product contains both $q$ and $p$. Classically,

$$qp=pq.$$

Quantum mechanically,

$$QP-PQ = [Q,P] = i\hbar I.$$

Thus $QP$ and $PQ$ are different operators even though they correspond to the same classical expression $qp$. Neither product is symmetric on the usual common test-function domain, so neither is automatically an observable for the real classical function $qp$. A symmetric choice is

$$\frac12(QP+PQ),$$

but that is a convention or principle to be justified, not a consequence of the classical expression alone.

More complicated expressions make the issue sharper. The classical function $q^2p^2$ might suggest $Q^2P^2$, $QPQP$, $P Q^2 P$, or a fully symmetrized operator. These formal expressions can differ by terms involving powers of $\hbar$. Their symmetry and self-adjoint domains must still be checked; simply listing orderings does not make them equally acceptable Hamiltonians.

In many elementary Hamiltonians, such as $p^2/2m+V(q)$ in Cartesian coordinates, the natural ordering is straightforward. In curvilinear coordinates, systems with constraints, position-dependent masses, gauge backgrounds, and field theories, ordering and regularization can carry physical content.

The practical rule is conservative: use the quantization procedure appropriate to the physical system, check symmetry and self-adjointness, and verify the intended classical regime. Do not treat $q\mapsto Q$ and $p\mapsto P$ as a universal algorithm.

Path to Canonical Quantization and QFT

For a finite number of canonical coordinate–momentum pairs, quantization is often introduced through

$$\{\,\cdot\,,\,\cdot\,\}_{\mathrm{PB}} \longrightarrow \frac{1}{i\hbar}[\,\cdot\,,\,\cdot\,].$$

That rule is a guide to the algebraic structure. “Finite number of pairs” does not mean finite-dimensional Hilbert space: finite matrices cannot satisfy $[Q,P]=i\hbar I$, since the trace of a commutator vanishes while $i\hbar\operatorname{Tr}I$ does not. The usual particle Hilbert space is infinite dimensional, with unbounded $Q$ and $P$ and specified domains.

The path-integral viewpoint starts from a classical action and assigns phases

$$e^{iS/\hbar}$$

to histories. It makes the classical limit especially transparent because stationary action appears as stationary phase. It also hides subtleties in the measure, normalization, coordinate dependence, and ordering. See Why Path Integrals? for the Core-adjacent motivation.

In field theory the same themes become more demanding. Canonical quantization promotes fields and conjugate momenta to operator-valued distributions, while path-integral quantization integrates over field histories. Gauge redundancy, regularization, renormalization, locality, and particle creation become structural issues rather than optional complications. The boundary with field theory is summarized in Relationship to Quantum Field Theory.

The historical phrase Second Quantization should not be confused with quantizing a theory twice. It usually means the Fock-space and creation-annihilation-operator language for variable particle number and many-body systems.

Worked Comparison: Harmonic Oscillator

Classical oscillator data:

$$h(q,p) = \frac{p^2}{2m} + \frac12m\omega^2q^2.$$

Canonical quantization gives

$$H = \frac{P^2}{2m} + \frac12m\omega^2Q^2, \qquad [Q,P]=i\hbar I.$$

Solving this quantum model gives the spectrum

$$E_n = \hbar\omega\left(n+\frac12\right).$$

The classical-limit question then runs the other way. For large $n$, the relative importance of the zero-point term decreases. For coherent states, expectation values oscillate like the classical oscillator while uncertainties remain controlled. For thermal states at high temperature, average energies approach the classical equipartition result. These are three different classical-looking statements extracted from the same quantum Hamiltonian.

The detailed quantum solution belongs in Quantum Harmonic Oscillator.

Common Mistakes

• Treating quantization and the classical limit as exact inverse operations.
• Assuming a classical Hamiltonian uniquely determines a quantum Hamiltonian.
• Replacing Poisson brackets by commutators without checking ordering, domains, and representation.
• Saying the classical limit is simply $\hbar=0$ rather than a controlled dimensionless regime.
• Expecting one classical description to cover all states of a quantum system.
• Confusing second quantization with a second physical application of quantization.

Exercises

1. On a common invariant domain for $Q,P$ with $[Q,P]=i\hbar I$, compare the two symmetric expressions

$$A=\frac12(Q^2P^2+P^2Q^2),\qquad B=PQ^2P.$$

Show that $A-B=-\hbar^2 I$, although both have classical expression $q^2p^2$. This is a formal ordering calculation; it does not settle their self-adjoint realizations.

Solution

The relation $[Q^2,P]=2i\hbar Q$ gives

$$Q^2P^2=B+2i\hbar QP,\qquad P^2Q^2=B-2i\hbar PQ.$$

Averaging yields

$$A=B+i\hbar(QP-PQ)=B-\hbar^2I.$$

The two symmetric orderings differ at order $\hbar^2$. Here the difference is only a constant energy shift if this expression is the Hamiltonian; that specific difference does not alter transition frequencies. More general ordering differences need not be constants.

2. For the harmonic oscillator, identify one statement that belongs to quantization and one that belongs to the classical-limit direction.

Solution

Quantization direction: start from

$$h(q,p)=\frac{p^2}{2m}+\frac12m\omega^2q^2$$

and construct

$$H=\frac{P^2}{2m}+\frac12m\omega^2Q^2$$

with $[Q,P]=i\hbar I$. Classical-limit direction: start from the quantum oscillator and show, for example, that coherent-state expectation values follow the classical oscillator equations.

3. Why does agreement with the same leading classical equations not prove that two quantum theories are identical?

Solution

The leading classical equations can ignore information that is quantum-mechanically important: operator ordering, spectra, spin representations, boundary conditions, tunneling amplitudes, phases, anomalies, and higher-order corrections in $\hbar$. Two quantum theories can therefore share a classical approximation while differing in measurable quantum predictions.

4. A derivation begins with the action $S[q]$ and writes an amplitude proportional to $e^{iS[q]/\hbar}$. Is this quantization, a classical limit, or both?

Solution

Writing amplitudes from a classical action is part of a quantization strategy, specifically the path-integral route. Taking the stationary-phase approximation of the resulting amplitude when $S/\hbar\gg1$ is a classical-limit argument. The same formula can therefore appear in both directions, but the question being asked is different.

References

• P. A. M. Dirac, The Principles of Quantum Mechanics, 4th ed., Oxford University Press, 1958.
• R. P. Feynman and A. R. Hibbs, Quantum Mechanics and Path Integrals, McGraw-Hill, 1965.
• B. C. Hall, Quantum Theory for Mathematicians, Springer, 2013.
• J. J. Sakurai and J. Napolitano, Modern Quantum Mechanics, 3rd ed., Cambridge University Press, 2020.
• R. Shankar, Principles of Quantum Mechanics, 2nd ed., Springer, 1994.
• J. von Neumann, Mathematical Foundations of Quantum Mechanics, Princeton University Press, 1955.