Semiclassical scaling and classical limits

We use the mass-one Hamiltonian $P_h=-h^2\Delta/2+V(x)$, where h lies in (0,1] and V is real and smooth. A self-adjoint realization and its domain must be specified when discussing evolution or spectra; a displayed differential expression alone is not a self-adjoint operator. Typical confining potentials provide discrete eigenvalues, whereas free motion on Euclidean space has continuous spectrum. The small parameter can arise from nondimensionalizing physical constants or from rewriting a high-energy problem at a fixed scaled energy.

The classical principal symbol is kinetic plus potential energy, with momentum ξ of order one even when the ordinary Fourier frequency is of order one over h. For a phase exp(iS/h), differentiation creates the factor ∇S/h, while multiplication by h in hD cancels that large denominator. This balance is the reason the kinetic energy remains meaningful in the limit. A family with small amplitude but fixed wavelength exhibits a different effect and may have no nontrivial normalized classical state.

Define the unitary semiclassical Fourier transform with the negative exponential and normalization shown below. Its inverse has the positive exponential. On Schwartz functions, transforming hD_j u gives multiplication by ξ_j, so a phase exp(ix·ξ₀/h) carries momentum ξ₀. This phase alone is not square integrable on Euclidean space; a smooth envelope or a coherent Gaussian makes a genuine normalized state. Keeping the normalization fixed prevents factors of h and two pi from drifting in later Wigner and trace formulas.

A localized wave can have spatial width of order sqrt(h) and momentum width of the same order while retaining unit mass. The corresponding phase-space cell has area of order h in each conjugate coordinate pair. This does not assert that a quantum state has a sharp position and momentum simultaneously: both widths remain positive at each fixed h. A classical point can nevertheless emerge as a weak limit when those widths shrink together, as the later measure construction makes precise.

Classical Hamilton equations describe the leading transport of suitable observables or limiting measures, not pointwise convergence of a rapidly oscillating wavefunction. The wavefunction can converge weakly to zero while its quadratic expectations converge to nonzero phase-space data. Different subsequences can also have different limits. One must state the observable class, normalization, localization and time interval before claiming a classical approximation, rather than treating the phrase classical limit as a single universal convergence theorem.

The harmonic oscillator supplies exact checks for these distinctions. Its normalized ground state becomes spatially narrow and has an energy tending to the minimum of the classical symbol. Highly excited states with index growing as h decreases can instead stay near a positive classical energy. Holding the eigenstate index fixed and holding energy fixed are therefore different limiting regimes. Later quantization rules will explicitly separate regular positive-energy turning points from the bottom of a well.