Global and local billiard level dynamics
Michael Barth · Ulrich Kuhl · Hans-Jürgen Stöckmann
MotivationConductance through a system is related to the sensitivity of eigenvalues on the change of an external parameter.[E. Akkermans and G. Montambaux (1992)] Rescaling of eigenvalues En and parameter X, ![]() leads to universal behavior (Δ: mean level spacing), independent of the kind of perturbation. [B.D. Simons and B.L. Altshuler (1993)] These predictions are tested with two types of perturbations: |
||||||
ExperimentThere is a one-to-one correspondence between the stationary Schrödinger equation and the Helmholtz equation.Measurement of eigenfrequency spectra of billiard-shaped microwave cavities ==> Experimental approach to quantum chaos. Level dynamics measurements:
|
||||||
Global level dynamics
|
||||||
Local level dynamicsWhether the level dynamics is global or local depends on δ=kD (wavenumber: k = 2 π ν / c, diameter of the disk: D).a) 0.35 < δ < 0.65 (D=5 mm, 3.4GHz < ν < 6GHz) b) 1.4 < δ < 2.6 (D=20 mm, 3.4GHz < ν < 6GHz) c) 5.1 < δ < 5.9 (D=20 mm, 12.5GHz < ν < 14.5GHz)
|
||||||
InterpretationMovable disk can be interpreted as a perturber probing the field in the resonator.In two-dimensional billiards a metallic perturber leads to a negative frequency shift proportional to the field intensity at the position X of the perturber. ==> The eigenvalue velocity reads ![]() where α is a constant depending on the geometry of the perturber. Under the assumption that chaotic wavefunctions can be described by a random superposition of plane waves [M.V. Berry (1977)], ψ and ∇ψ are uncorrelated, and Gaussian distributed [V.N. Prigodin et al. (1995)]. This leads to the new velocity distribution ![]() with β = A / (2 α k), A is billiard area. The red curves correspond to this modified Bessel function (also called MacDonald function). With 1 / β² = 〈 ( dEn / dX )² 〉 and Δ=4π/A (mean level spacing), the rescaled parameter x is obtained as ![]() Result: x is not universal for local level dynamics! One should use xbar = kX instead. The autocorrelation function is obtained as: ![]() ![]() If the autocorrelation function is plotted as a function of the rescaled parameter xbar, a new universal behavior is found.
|
||||||
Literature:
|


















