Quantum Mechanics Without Leaving Phase Space

Status: an interactive companion to an independent pedagogical write-up. It reconstructs established Weyl–Wigner and Moyal theory; it does not claim a new quantization theorem or a solution to the measurement problem.

Classical mechanics gives us a seductive picture. A state is a point $(q,p)$ in phase space, an observable is a function $A(q,p)$, and Hamilton’s equations move the point along a trajectory. Quantum mechanics seems to throw that geometry away in favour of vectors and operators.

The Weyl–Wigner formulation asks a sharper question: how much of quantum mechanics can be written on the same phase space without pretending that the theory is classical? The answer is almost all of it. Operators become symbols, density matrices become quasi-distributions, traces become integrals, and commutators become deformed Poisson brackets.

The catch is where the quantum structure moves. It ends up stored in two unusual features: the state can be negative as a phase-space function, and the product of observables stops being ordinary multiplication.

The map must remember operator ordering

Take an operator $\hat A$ on one canonical degree of freedom. Its Weyl symbol is

$$A_W(q,p)=\int_{-\infty}^{\infty}dy\, e^{-ipy/\hbar} \left\langle q+\frac y2\right|\hat A \left|q-\frac y2\right\rangle.$$

Every piece of that integral is carrying weight. The midpoint $q$ and separation $y$ split the two position arguments of the kernel, while the Fourier variable $p$ supplies the conjugate momentum coordinate.

Simple operators behave as hoped:

$$(\hat q)_W=q,\qquad (\hat p)_W=p.$$

Trouble begins with products. Since $\hat q\hat p\ne\hat p\hat q$, both cannot map to the ordinary function $qp$. Weyl ordering treats the two symmetrically:

$$\left(\frac{\hat q\hat p+\hat p\hat q}{2}\right)_W=qp.$$

The symbol therefore encodes a particular ordering prescription. If you were taught quantization as putting hats on a classical expression, this is where that recipe runs out of instructions. Putting hats on $q$ and $p$ underdetermines the operator: $\hat q\hat p$, $\hat p\hat q$ and their symmetrization are three different operators sharing one classical expression, and the Weyl map has picked the third.

A quantum state can look like a distribution, but not a probability

For a density operator $\hat\rho$, define the Wigner function

$$W_\rho(q,p)=\frac{1}{2\pi\hbar} \int_{-\infty}^{\infty}dy\,e^{-ipy/\hbar} \left\langle q+\frac y2\right|\hat\rho \left|q-\frac y2\right\rangle.$$

Its normalization and marginals look exactly probabilistic:

$$\iint W_\rho(q,p)\,dq\,dp=1,$$ $$\int W_\rho(q,p)\,dp=\langle q|\hat\rho|q\rangle, \qquad \int W_\rho(q,p)\,dq=\langle p|\hat\rho|p\rangle.$$

Expectation values also become phase-space averages:

$$\langle\hat A\rangle =\operatorname{Tr}(\hat\rho\hat A) =\iint W_\rho(q,p)A_W(q,p)\,dq\,dp.$$

So why call $W$ a quasi-probability? Because it can be negative, and no apparatus has ever recorded an event a negative number of times. The negative regions are the price of placing noncommuting quantum information inside one real function on $(q,p)$ while still recovering the correct position and momentum marginals.

For a coherent Gaussian state, the Wigner function is positive. For excited number states and superpositions, it typically develops oscillatory negative lobes. Interference in Hilbert space has become signed fine structure in phase space.

Negativity is therefore informative, and easy to over-read in both directions. A positive Wigner function still leaves a state with plenty of nonclassical structure. What an experiment reconstructs is the state itself, assembled from many ordinary measurement statistics.

The star product is where noncommutativity lives

To preserve operator composition, symbols use the Moyal star product:

$$A\star B =A\exp\!\left[ \frac{i\hbar}{2} \left( \overleftarrow\partial_q\overrightarrow\partial_p -\overleftarrow\partial_p\overrightarrow\partial_q \right) \right]B.$$

The arrows say which factor each derivative acts on. Expanding the exponential reveals the deformation:

$$A\star B =AB+\frac{i\hbar}{2}\{A,B\}_{\mathrm P} -\frac{\hbar^2}{8}A\Lambda^2B+\cdots,$$

where

$$\{A,B\}_{\mathrm P} =\partial_qA\,\partial_pB-\partial_pA\,\partial_qB$$

is the Poisson bracket, and $\Lambda$ abbreviates the same bidifferential operator. For the coordinate functions,

$$q\star p=qp+\frac{i\hbar}{2},\qquad p\star q=pq-\frac{i\hbar}{2},$$

so

$$q\star p-p\star q=i\hbar.$$

That is the canonical commutation relation, translated exactly. Phase space remains ordinary as a set of points, while its algebra of observables has become noncommutative.

This is the central idea of deformation quantization. Instead of replacing classical functions by operators, one deforms the product of the functions. The parameter $\hbar$ measures the deformation. If that sounds like a change of bookkeeping rather than of physics, watch what happens at $\hbar\to0$: the star product collapses to ordinary multiplication and the bracket above collapses to the Poisson bracket, which is the classical limit arriving as a limit of the product itself.

Dynamics becomes a deformed Hamiltonian flow

The von Neumann equation

$$i\hbar\,\partial_t\hat\rho=[\hat H,\hat\rho]$$

maps to

$$\partial_tW_\rho =\{H_W,W_\rho\}_{\mathrm M},$$

where the Moyal bracket is

$$\{A,B\}_{\mathrm M} =\frac{A\star B-B\star A}{i\hbar}.$$

Only the odd powers of the bidifferential operator survive the antisymmetric difference. After division by $i\hbar$, the expansion therefore contains only even powers of $\hbar$:

$$\{A,B\}_{\mathrm M} =\{A,B\}_{\mathrm P}+O(\hbar^2).$$

This formula is often summarized as “quantum mechanics becomes classical when $\hbar\to0$.” That slogan hides the real condition. The corrections involve higher derivatives, so what matters is whether $A$ and $B$ vary slowly on the phase-space scale set by $\hbar$. Fine oscillations can make higher derivatives large enough to compete with the nominal powers of $\hbar$.

There is one exceptionally clean case. If the Hamiltonian is at most quadratic in $q$ and $p$, every third and higher derivative of $H$ vanishes. All quantum corrections in the Moyal bracket disappear, and

$$\{H,W\}_{\mathrm M}=\{H,W\}_{\mathrm P}$$

exactly. A harmonic-oscillator Wigner function is transported by the same linear symplectic flow as a classical phase-space density.

This does not make the oscillator fully classical. Its allowed states still obey uncertainty constraints and may have negative Wigner functions. The dynamical law can be classically shaped while the state space remains quantum.

Coarse-graining attacks the shortest scales first

A real apparatus has finite phase-space resolution. Model that limitation by convolving $W$ with a normalized Gaussian kernel. In Fourier language, convolution multiplies each wavevector component by a decaying Gaussian. High-frequency fringes are suppressed before broad positive lobes.

The model below isolates that mechanism. It is not fitted experimental data; it is a two-lobe quasi-distribution with an interference term. Increase the smoothing and watch the negative, rapidly oscillating part vanish first.

Phase-space microscope

Blur the fringes, not the normalization

resolution loss = 0.00

Gaussian coarse-graining strengthfine Wigner structurecoarse positive picture

position q →momentum p →

positivenegative

illustrative negative weight fraction9.11%

fringe visibility factor1.000

This is a dimensionless two-lobe model designed to expose the filtering mechanism, not a state-tomography dataset. Gaussian convolution damps the high-frequency interference term first. The Husimi Q-function is a specific physically normalized smoothing with guaranteed positivity.

This explains why a coarse observer can see an apparently classical density even when the fine description is quantum. It also explains why simply taking $\hbar$ to be numerically small is insufficient: a state can generate structure on correspondingly small scales.

The Husimi function provides a physically distinguished smoothing. For a coherent state $|\alpha\rangle$,

$$Q(\alpha)=\frac1\pi\langle\alpha|\hat\rho|\alpha\rangle\ge0.$$

It can be obtained by Gaussian smoothing of the Wigner function at the minimum uncertainty scale. Its positivity is guaranteed because it is an expectation value of the positive operator $\hat\rho$. The gain is a genuine probability density for coherent-state outcomes; the cost is resolution. The sharp features that distinguished some states have been averaged away.

What the phase-space picture does and does not establish

The formal classical limit is precise: pointwise multiplication and the Poisson bracket are the leading terms of a noncommutative deformation. The operational classical limit is subtler: finite resolution suppresses the small-scale structure on which quantum corrections and Wigner negativity can depend.

Neither statement, by itself, explains why one definite measurement outcome occurs. Coarse-graining is also not identical to environmental decoherence. Decoherence is a physical open-system process that transfers phase information into uncontrolled degrees of freedom; coarse-graining is a description of limited resolution or discarded detail. They can produce related mathematics, but they answer different questions.

The lasting value of the Weyl–Wigner formulation is that it makes the boundary visible. The same phase space supports both theories, and negativity and the star product mark exactly where the classical analogy fails.

For the historical foundations, see Wigner’s original phase-space construction, Groenewold’s analysis of quantization, and Moyal’s statistical formulation.

Read the full pedagogical write-up (PDF).