The PT symmetric potential V0[cos(2πx/a) + iλ sin(2πx/a)] has a completely real spectrum for λ ≤ 1, and begins to develop complex eigenvalues for λ > 1. At the symmetry-breaking threshold λ = 1 some of the eigenvectors become degenerate, giving rise to a Jordan-block structure for each degenerate eigenvector. In general this is expected to give rise to a secular growth in the amplitude of the wave. However, it has been shown in a recent paper by Longhi, by numerical simulation and by the use of perturbation theory, that for an initial wave packet this growth is suppressed, giving instead a constant maximum amplitude. We revisit this problem by developing the perturbation theory further. We verify that the results found by Longhi persist to second order, and with different input wave packets we are able to see the seeds in perturbation theory of the phenomenon of birefringence first discovered by El-Ganainy et al.
The
PT-symmetric potential ${V}_{0}[\mathrm{cos}(2\ensuremath{\pi}x/a)+i\ensuremath{\lambda}\mathrm{sin}(2\ensuremath{\pi}x/a)]$ has a completely real spectrum for $\ensuremath{\lambda}\ensuremath{\leqslant}1$ and begins to develop complex eigenvalues for $\ensuremath{\lambda}>1$. At the symmetry-breaking threshold $\ensuremath{\lambda}=1$ some of the eigenvectors become degenerate, giving rise to a Jordan-block structure for each degenerate eigenvector. In general this is expected to result in a secular growth in the amplitude of the wave. However, it has been shown in a recent paper by Longhi, by numerical simulation and by the use of perturbation theory, that for a broad initial wave packet this growth is suppressed, and instead a saturation leading to a constant maximum amplitude is observed. We revisit this problem by explicitly constructing the Bloch wave functions and the associated Jordan functions and using the method of stationary states to find the dependence on the longitudinal distance
z for a variety of different initial wave packets. This allows us to show in detail how the saturation of the linear growth arises from the close connection between the contributions of the Jordan functions and those of the neighboring Bloch waves.