Semiclassical Propagator Preview

The semiclassical propagator is the leading stationary-phase approximation to the quantum kernel. It says that, when actions are large compared with $\hbar$, the amplitude to go from $(q_i,t_i)$ to $(q_f,t_f)$ is organized by classical paths connecting those endpoints.

This page is a preview inside the Dynamics volume. It explains the idea, the main formula, and the pitfalls. The detailed method-level treatment belongs to Semiclassical Propagator, with separate references for the Van Vleck Determinant and Maslov Index.

What Is Being Approximated

The exact coordinate-space propagator is

$$K(q_f,t_f;q_i,t_i) = \langle q_f\rvert U(t_f,t_i)\lvert q_i\rangle.$$

It evolves wavefunctions by

$$\psi(q_f,t_f) = \int dq_i\, K(q_f,t_f;q_i,t_i)\psi(q_i,t_i),$$

with the measure appropriate to the configuration space. In path-integral language, the same kernel is written schematically as

$$K(q_f,t_f;q_i,t_i) = \int_{q(t_i)=q_i}^{q(t_f)=q_f} \mathcal Dq\, \exp\left[ \frac{i}{\hbar}S[q] \right].$$

Stationary phase expands this oscillatory integral around paths satisfying

$$\delta S[q_\gamma]=0,$$

with the endpoints fixed. These are classical trajectories.

Sum Over Classical Paths

In $d$ configuration-space dimensions, the leading Van Vleck-style expression has the form

$$K_{\rm sc}(q_f,t_f;q_i,t_i) = \sum_{\gamma:i\to f} \left( \frac{1}{2\pi i\hbar} \right)^{d/2} \lvert D_\gamma\rvert^{1/2} \exp\left[ \frac{i}{\hbar}S_\gamma - i\frac{\pi}{2}\nu_\gamma \right].$$

The sum is over classical paths $\gamma$ that start at $q_i$ at $t_i$ and arrive at $q_f$ at $t_f$. Each contribution contains:

Symbol	Meaning
$S_\gamma$	classical action along the path
$D_\gamma$	Van Vleck stability determinant
$\nu_\gamma$	Maslov index or caustic phase count
$e^{iS_\gamma/\hbar}$	rapidly oscillatory action phase

This is an amplitude sum, not a probability sum. If several classical paths connect the same endpoints, their contributions interfere.

Classical Action

For a classical trajectory $\gamma$, the action is

$$S_\gamma(q_f,t_f;q_i,t_i) = \int_{t_i}^{t_f} L(q_\gamma,\dot q_\gamma,t)\,dt.$$

As a function of endpoints, this is Hamilton’s principal function. Its endpoint derivatives generate the endpoint momenta:

$$p_f = \frac{\partial S_\gamma}{\partial q_f}, \qquad p_i = - \frac{\partial S_\gamma}{\partial q_i}.$$

These relations explain why the propagator phase is tied to Hamilton–Jacobi theory. In WKB language, the phase of a semiclassical wave is a classical action; in propagator language, the phase of each branch is the action of a classical path.

Fluctuation Determinant

Stationary phase does not only give the phase. It also gives a prefactor from Gaussian fluctuations around the classical path. Write

$$q(t) = q_\gamma(t)+\eta(t), \qquad \eta(t_i)=\eta(t_f)=0.$$

Then

$$S[q] = S[q_\gamma] + \frac12 \delta^2S_\gamma[\eta,\eta] + \cdots.$$

The linear term vanishes because $q_\gamma$ is classical. The quadratic fluctuation integral gives a determinant. In configuration-space endpoint variables, that determinant can be written as the Van Vleck determinant

$$D_\gamma = \det\left( - \frac{\partial^2S_\gamma} {\partial q_f\,\partial q_i} \right).$$

Intuitively, $D_\gamma$ measures classical stability: how a small range of initial momenta maps to final positions. When neighboring trajectories focus, the determinant signals that the isolated-trajectory approximation is approaching a caustic.

Free-Particle Check

For a free particle in $d$ dimensions with $T=t_f-t_i\gt0$,

$$S_{\rm free} = \frac{m\lvert q_f-q_i\rvert^2}{2T}.$$

The mixed second derivative is

$$- \frac{\partial^2S_{\rm free}} {\partial q_f\,\partial q_i} = \frac{m}{T}I_d,$$

so

$$D_{\rm free} = \left( \frac{m}{T} \right)^d.$$

With no caustic phase for $T\gt0$, the semiclassical kernel is

$$K_{\rm free}(q_f,T;q_i,0) = \left( \frac{m}{2\pi i\hbar T} \right)^{d/2} \exp\left[ \frac{i}{\hbar} \frac{m\lvert q_f-q_i\rvert^2}{2T} \right].$$

This is the exact free-particle propagator. Quadratic actions are special: the stationary-phase expansion stops after the Gaussian fluctuation integral.

Multiple Paths and Interference

For some endpoints and travel times, more than one classical path exists. Examples include motion on a circle, motion in a potential after turning points, and optical or mechanical focusing problems. The semiclassical propagator then contains a coherent sum:

$$K_{\rm sc} = \sum_\gamma A_\gamma e^{iS_\gamma/\hbar}.$$

The relative phases $S_\gamma/\hbar$ determine interference. A classical-looking path is not selected by probability alone; the quantum amplitude remembers all relevant stationary branches.

This is one reason semiclassics is richer than the slogan “classical path dominates.” Often the correct statement is: classical paths label the leading stationary contributions, and those contributions interfere as quantum amplitudes.

Maslov Phase Preview

The determinant in the prefactor has a phase as well as a magnitude. When a family of trajectories passes through a caustic or conjugate point, the naive square root changes branch. The Maslov index records the corresponding phase jump.

In the convention used here, a trajectory contribution carries

$$\exp\left[ - i\frac{\pi}{2}\nu_\gamma \right].$$

The integer $\nu_\gamma$ depends on the path and on the convention. The main practical rule is not to ignore it. Without the Maslov phase, semiclassical formulas fail across turning points and focal points even when their action phases and magnitudes look plausible.

Caustics and Uniform Approximations

A caustic occurs when the projection from a family of classical trajectories to configuration space becomes singular. In the Van Vleck formula this appears as a divergence of the isolated-trajectory prefactor.

The exact quantum propagator is not normally divergent at the caustic. The divergence means the local semiclassical representation has failed. Nearby classical branches must be treated together by a uniform approximation, such as an Airy approximation near a simple fold caustic.

Thus the Van Vleck formula is local in trajectory space. It is powerful away from caustics, but it is not a globally uniform approximation by itself.

Validity and Failure Modes

The leading semiclassical propagator is most reliable when:

• the relevant actions are large compared with $\hbar$;
• the contributing classical paths are isolated;
• the second variation is nondegenerate;
• the endpoints are away from caustics;
• the time is not so long that exponentially many unstable paths dominate the approximation;
• the observable or wave packet being propagated does not require unresolved fine interference beyond the approximation.

It can fail or require modification near turning points, caustics, separatrices, tunneling regions, singular potentials, hard boundaries, chaotic long-time propagation, and situations where complex saddles matter.

Relation to QFT Saddles

The same pattern appears in field theory. The path integral is over fields rather than particle paths, and stationary points are classical field configurations. Around a saddle one obtains

$$\text{amplitude} \sim e^{iS_{\rm cl}/\hbar} \times \text{fluctuation determinant} \times \text{phase data}.$$

Quantum mechanics is the clean finite-dimensional setting in which to learn the meaning of action phases, fluctuation determinants, and saddle interference before meeting instantons, solitons, gauge zero modes, and renormalized determinants in field theory.

Common Mistakes

• Treating the semiclassical propagator as a probability distribution over classical paths.
• Keeping the action phase while dropping the determinant needed for normalization and stability.
• Ignoring Maslov phases at caustics and turning points.
• Using the isolated-path formula exactly where the determinant diverges.
• Assuming there is always only one classical path between two endpoints.
• Confusing a useful saddle approximation with a literal statement that nonclassical paths do not contribute to the exact path integral.
• Forgetting that detailed semiclassical propagation has convention choices for determinants, phases, and endpoint variables.

Cross-Links

• What Is the Classical Limit? gives the broader classical-limit map.
• Hamilton–Jacobi Theory Preview derives the action phase, endpoint momentum rules, and stationary composition law.
• Stationary Phase gives the finite-dimensional asymptotic method.
• Quantum Chaos Preview explains how unstable periodic orbits and long-time path proliferation enter quantum-chaos diagnostics.
• Stationary Phase and the Classical Limit derives the time-sliced saddle equations and Jacobi fluctuation operator.
• Propagator Kernel defines the exact kernel being approximated.
• From Propagators to Path Integrals explains the time-slicing origin of the path integral.
• Free-Particle Propagator gives the exact kernel used as a check.
• Semiclassical Propagator gives the full method-level treatment.
• Van Vleck Determinant explains the stability prefactor.
• Maslov Index explains caustic phase bookkeeping.

References

• J. H. Van Vleck, “The correspondence principle in the statistical interpretation of quantum mechanics,” Proceedings of the National Academy of Sciences 14, 178, 1928.
• M. V. Berry and K. E. Mount, “Semiclassical approximations in wave mechanics,” Reports on Progress in Physics 35, 315, 1972.
• M. C. Gutzwiller, Chaos in Classical and Quantum Mechanics, Springer, 1990.
• M. Brack and R. K. Bhaduri, Semiclassical Physics, Westview Press, 2003.
• L. S. Schulman, Techniques and Applications of Path Integration, Wiley, 1981.
• R. G. Littlejohn, “The semiclassical evolution of wave packets,” Physics Reports 138, 193, 1986.

Exercises

1. Derive the endpoint momentum relations from the classical action.

Solution

For a classical path, the first variation of the action has only endpoint terms:

$$\delta S_\gamma = p_f\,\delta q_f - p_i\,\delta q_i.$$

If $S_\gamma$ is viewed as a function of the endpoints, then

$$\delta S_\gamma = \frac{\partial S_\gamma}{\partial q_f}\delta q_f + \frac{\partial S_\gamma}{\partial q_i}\delta q_i.$$

Comparing coefficients gives

$$p_f = \frac{\partial S_\gamma}{\partial q_f}, \qquad p_i = - \frac{\partial S_\gamma}{\partial q_i}.$$

2. Verify the free-particle determinant in $d$ dimensions.

Solution

For

$$S_{\rm free} = \frac{m}{2T} (q_f-q_i)\cdot(q_f-q_i),$$

one has

$$\frac{\partial^2S_{\rm free}} {\partial q_{f,a}\,\partial q_{i,b}} = - \frac{m}{T}\delta_{ab}.$$

Therefore

$$- \frac{\partial^2S_{\rm free}} {\partial q_f\,\partial q_i} = \frac{m}{T}I_d,$$

and

$$D_{\rm free} = \left( \frac{m}{T} \right)^d.$$

3. Explain why several classical paths can contribute to one propagator.

Solution

The propagator fixes endpoints and travel time, so the relevant classical problem is a boundary-value problem. Boundary-value problems can have multiple solutions. For example, on a circle a particle can reach the same final point by winding different numbers of times. Each solution is a stationary point of the action and contributes a semiclassical amplitude. The contributions must be summed coherently, so their phases can interfere.

4. What does a zero of the Van Vleck denominator signal?

Solution

In stability language, the Van Vleck determinant is related to the inverse of the map from initial momentum perturbations to final position perturbations. If that map becomes singular, nearby trajectories focus at the same final position. The isolated-trajectory prefactor diverges, signaling a caustic. The exact quantum amplitude is repaired by treating the coalescing branches together with a uniform approximation.

5. Why is the free-particle semiclassical propagator exact?

Solution

The free-particle action is quadratic in the path fluctuations around the classical path. In the stationary-phase expansion,

$$S[q_\gamma+\eta] = S[q_\gamma] + \frac12\delta^2S_\gamma[\eta,\eta],$$

with no higher fluctuation terms. Therefore the Gaussian fluctuation integral is not merely the leading approximation; it is the full path integral. The stationary-phase result equals the exact kernel.