Why Phase Space in Quantum Mechanics?

Classical mechanics describes a state by a point in phase space, while statistical mechanics describes uncertainty by a positive probability density over phase space. Quantum mechanics instead uses vectors or density operators on Hilbert space, with position and momentum represented by noncommuting operators.

Why return to phase space at all?

The answer is not that quantum mechanics secretly assigns simultaneous classical values to position and momentum. Phase-space quantum mechanics builds a translation dictionary:

$$\text{operators and density matrices} \quad \longleftrightarrow \quad \text{functions on }(x,p),$$

while modifying the classical rules enough to preserve noncommutativity and interference. The Wigner function represents states by a quasiprobability, the Weyl transform represents observables by symbols, and the star product replaces ordinary multiplication.

This formulation is useful because it places quantum and classical dynamics in closely comparable variables without identifying them.

Classical Phase Space

For $n$ classical degrees of freedom, local canonical coordinates are

$$z = \left( q^1,\ldots,q^n, p_1,\ldots,p_n \right).$$

An individual classical state is a point $z$ in a $2n$-dimensional phase space. Hamilton’s equations generate its trajectory:

$$\dot q^i = \frac{\partial H}{\partial p_i}, \qquad \dot p_i = - \frac{\partial H}{\partial q^i}.$$

For a classical statistical ensemble, a phase-space density $f(z,t)$ satisfies

$$f(z,t)\geq0, \qquad \int dz\,f(z,t)=1.$$

Expectation values are ordinary averages:

$$\langle a\rangle_{\mathrm{cl}} = \int dz\, f(z,t)a(z).$$

Hamiltonian flow preserves phase-space volume and gives the Liouville equation

$$\frac{\partial f}{\partial t} = \{H,f\}_{\mathrm{PB}},$$

using the sign convention adopted in this volume. The Phase Space page is the canonical home for this classical geometry, and Poisson Brackets develops the bracket.

Three classical ideas are bundled together here:

• states can be represented by points or positive densities on phase space;
• observables are ordinary real functions on that space;
• products of observables are ordinary pointwise products.

Quantum phase-space formulations cannot preserve all three statements unchanged.

Why There Is No Ordinary Joint Probability for Position and Momentum

In quantum mechanics,

$$[\hat x,\hat p] = i\hbar I.$$

The spectral measures of $\hat x$ and $\hat p$ therefore do not define one common sharp joint measurement. Position and momentum each have Born-rule distributions,

$$P_x(x) = \langle x\rvert\rho\lvert x\rangle,$$

and

$$P_p(p) = \langle p\rvert\rho\lvert p\rangle,$$

but these are distributions for different measurement contexts. Writing their product $P_x(x)P_p(p)$ does not reconstruct the quantum state and generally erases position–momentum correlations and coherence.

The obstruction is stronger than the uncertainty relation viewed only as a bound on variances. A universal positive joint density with exact position and momentum marginals, together with a classical rule for all observable products, would behave as though the noncommuting operators were compatible random variables. It could not reproduce the full operator algebra.

This does not forbid every joint or phase-space-like measurement:

• unsharp position and momentum can be jointly measured with a suitable POVM;
• positive phase-space distributions can be constructed after adding noise or smoothing;
• different experimental schemes can sample different ordered moments.

The price is that exact marginals, positivity, sharpness, or the ordinary product rule cannot all be retained simultaneously. The phase-space formulation makes that tradeoff explicit rather than hiding it.

Quasiprobability Distributions

A quasiprobability distribution is a function used to calculate quantum predictions in a probability-like way while relaxing at least one classical probability property. It may become negative, more singular than an ordinary function, or tied to a particular operator-ordering convention.

Several distributions are useful:

Representation	Main advantage	Main cost
Wigner function	real, normalized, exact $x$ and $p$ marginals	can be negative
Husimi $Q$ function	nonnegative and smooth	marginals are smeared; fine structure is lost
Glauber–Sudarshan $P$ representation	natural for normally ordered optical moments	may be highly singular or nonpositive as a distribution

These representations encode the same density operator when their transforms are well defined, but they organize information differently. “More positive” does not mean “more exact,” and negativity is not a mathematical defect to be repaired automatically.

Quasiprobabilities also depend on conventions. Fourier signs, factors of $2\pi\hbar$, scaling of quadratures, and operator ordering must be fixed before formulas are compared. The default conventions are collected in Phase-Space Conventions.

Wigner Function as a Bridge

For one degree of freedom, the Wigner function associated with a density operator $\rho$ is

$$\begin{aligned} W_\rho(x,p) &= \frac{1}{2\pi\hbar} \int_{-\infty}^{\infty}dy\, e^{-ipy/\hbar} \\ &\quad\times \left\langle x+\frac{y}{2} \middle| \rho \middle| x-\frac{y}{2} \right\rangle. \end{aligned}$$

It is real and normalized:

$$\int dx\,dp\, W_\rho(x,p) = 1.$$

Its marginals are exact:

$$\int dp\, W_\rho(x,p) = \langle x\rvert\rho\lvert x\rangle,$$

and

$$\int dx\, W_\rho(x,p) = \langle p\rvert\rho\lvert p\rangle.$$

The Wigner function is nevertheless not generally nonnegative. Interference between spatially separated wave packets, for example, produces oscillatory phase-space fringes that can cross below zero.

States alone are only half of the formulation. The Weyl Transform assigns an operator $A$ a phase-space symbol $A_W(x,p)$. With consistent conventions,

$$\operatorname{Tr}(\rho A) = \int dx\,dp\, W_\rho(x,p)A_W(x,p).$$

This looks like a classical ensemble average. The quantum structure moves into two places:

• $W_\rho$ need not be positive;
• operator multiplication becomes the noncommutative Star Product, not ordinary multiplication.

Thus the Wigner–Weyl formulation is not a classical approximation. It is an exact representation of quantum mechanics when the transform, domains, and conventions are handled consistently. Wigner Function owns the detailed properties and examples.

Classical Limit

The density operator obeys the Liouville–von Neumann equation

$$\frac{\partial\rho}{\partial t} = \frac{1}{i\hbar} [H,\rho].$$

Under the Wigner–Weyl transform, this becomes

$$\frac{\partial W}{\partial t} = \{H_W,W\}_{M},$$

where $\{\cdot,\cdot\}_M$ is the Moyal Bracket. For sufficiently smooth symbols varying on phase-space scales large compared with the relevant quantum cell,

$$\{H_W,W\}_M = \{H_W,W\}_{\mathrm{PB}} + O(\hbar^2).$$

The leading term is the classical Liouville flow. This makes phase space a precise language for asking when quantum dynamics resembles classical statistical dynamics.

The limit is not obtained by deleting every $\hbar$ from a formula. It can fail or become nonuniform when:

• the Wigner function has interference fringes on $\hbar$-dependent scales;
• the potential or symbol is not smooth enough for a derivative expansion;
• evolution creates fine phase-space structure;
• long times amplify small corrections;
• tunneling or other effects are nonanalytic in $\hbar$.

For Hamiltonians at most quadratic in $x$ and $p$, the higher Moyal derivatives vanish and the Wigner function follows the classical linear phase-space flow exactly. The state can still be fully quantum: uncertainty constraints, purity, and measurement statistics have not become classical merely because the transport equation has classical form.

The exact dynamics is developed in Phase-Space Dynamics. The detailed asymptotic comparison belongs to the planned classical-limit page and to Semiclassical Limit.

Applications

Phase-space methods are useful when the geometry of both position and momentum matters:

• Semiclassical dynamics: classical trajectories, stability, and quantum corrections can be compared in one set of variables.
• Quantum optics: coherent, squeezed, and nonclassical states have distinctive phase-space shapes.
• Continuous-variable information: first moments, covariance matrices, Gaussian operations, and noise channels are naturally organized in phase space.
• Quantum-state tomography: measured quadrature distributions can be used to reconstruct a Wigner function under an explicit inversion procedure.
• Transport and nonequilibrium dynamics: kinetic-looking equations reveal drift, spreading, interference, and quantum corrections.
• Quantum chaos: phase-space structure helps compare classical invariant sets with quantum evolution, while respecting finite resolution and long-time limits.
• Numerical work: grid and trajectory-based approximations can exploit phase-space locality, though oscillatory fine structure can be expensive to resolve.

Gaussian States and Wigner Functions develops the tractable Gaussian sector. Coherent States in Phase Space shows why displaced oscillator Gaussians follow classical-looking trajectories.

Common Mistakes

• Treating phase space as the quantum state space rather than a representation space.
• Assuming a quantum particle has simultaneous hidden values $x$ and $p$ sampled from $W(x,p)$.
• Calling every nonnegative phase-space function a Wigner function.
• Interpreting Wigner negativity as a negative experimental frequency.
• Assuming negativity is necessary for every form of nonclassical behavior.
• Multiplying Weyl symbols pointwise when the corresponding operators are multiplied.
• Equating classical-looking evolution under a quadratic Hamiltonian with a classical state.
• Taking $\hbar\to0$ without identifying the physical scales held fixed.

Cross-Links

• Phase-Space Formulation
• Phase Space
• Poisson Brackets
• Classical vs Quantum Probability
• Canonical Commutation Relations
• Weyl Transform
• Wigner Function
• Star Product
• Moyal Bracket
• Phase-Space Dynamics
• Gaussian States and Wigner Functions
• Coherent States in Phase Space
• Phase-Space Conventions

References

• E. Wigner, “On the quantum correction for thermodynamic equilibrium,” Physical Review 40, 749–759 (1932).
• H.-W. Lee, “Theory and application of the quantum phase-space distribution functions,” Physics Reports 259, 147–211 (1995).
• M. Hillery, R. F. O’Connell, M. O. Scully, and E. P. Wigner, “Distribution functions in physics: fundamentals,” Physics Reports 106, 121–167 (1984).
• C. K. Zachos, D. B. Fairlie, and T. L. Curtright, eds., Quantum Mechanics in Phase Space, World Scientific, 2005.
• W. P. Schleich, Quantum Optics in Phase Space, Wiley-VCH, 2001.
• G. B. Folland, Harmonic Analysis in Phase Space, Princeton University Press, 1989.

Exercises

1. Compare a classical phase-space density with a Wigner function. Which properties are shared, and which classical property can fail?

Solution

Both are real functions on phase space and can be normalized:

$$\int dx\,dp\, f(x,p) = \int dx\,dp\, W(x,p) = 1.$$

Both can be used in phase-space integrals to calculate suitable expectation values. For the Wigner function, the observable must be represented by its Weyl symbol.

A classical probability density must satisfy $f(x,p)\geq0$ pointwise. A Wigner function can be negative. In addition, quantum operator products are represented by star products rather than ordinary pointwise products.

2. Why does multiplying the position and momentum Born distributions not produce the quantum joint state?

Solution

The product

$$P_x(x)P_p(p)$$

contains only the two separate marginals and imposes statistical independence between them. It discards position–momentum correlations and the phase information stored in off-diagonal density-matrix elements.

More fundamentally, $\hat x$ and $\hat p$ do not commute and have no common sharp spectral measure. Their separate Born distributions refer to incompatible measurement contexts, so their product is not a joint distribution supplied by the quantum state.

3. A minimum-uncertainty Gaussian wave packet has

$$\psi(x) = \frac{1}{(2\pi\sigma_x^2)^{1/4}} \exp\left[ - \frac{(x-x_0)^2}{4\sigma_x^2} + \frac{ip_0x}{\hbar} \right].$$

Its Wigner function is

$$W(x,p) = \frac{1}{\pi\hbar} \exp\left[ - \frac{(x-x_0)^2}{2\sigma_x^2} - \frac{2\sigma_x^2(p-p_0)^2}{\hbar^2} \right].$$

Verify the position marginal.

Solution

Integrate the momentum Gaussian:

$$\int_{-\infty}^{\infty}dp\, \exp\left[ - \frac{2\sigma_x^2(p-p_0)^2}{\hbar^2} \right] = \frac{\sqrt{\pi}\hbar} {\sqrt2\,\sigma_x}.$$

Therefore

$$\begin{aligned} \int dp\,W(x,p) &= \frac{1}{\sqrt{2\pi}\sigma_x} \exp\left[ - \frac{(x-x_0)^2}{2\sigma_x^2} \right] \\ &= |\psi(x)|^2. \end{aligned}$$

This Gaussian Wigner function is nonnegative, showing that positivity alone does not make the state classical.

4. Why does Wigner evolution under a quadratic Hamiltonian have the classical Liouville form?

Solution

The Moyal bracket expands as the Poisson bracket plus terms containing higher odd derivatives and higher powers of $\hbar$. If $H_W(x,p)$ is at most quadratic, all derivatives of $H_W$ of order three and above vanish. The correction terms therefore vanish identically:

$$\{H_W,W\}_M = \{H_W,W\}_{\mathrm{PB}}.$$

The transport equation is classically shaped, but the admissible Wigner functions still encode uncertainty, purity, and other quantum constraints.