Abstract

The formation of intensity patterns in transverse laser-beam profiles has been observed in various active and passive nonlinear optical systems.¹ Relatively few investigations, however, have included the vector (polarization) character of the optical field. We report on optical transverse effects based on the coupling of polarization components of copropagating polarized laser beams nearly resonant to the homogeneously broadened optical Dl transition of sodium vapor. In recent experiments we have observed² the predicted³ mutual deflection of two circularly polarized laser beams intersecting at a small angle, in a similar experimental setup, but with entirely superimposed beams at the entrance of a sodium-vapor cell, a spatial instability of the propagating laser beams within the cell occurs, and this leads to a sudden transverse splitting of the input beams into beams of pure circular polarization. The separation was measured at the cell output window for different input beam powers (Fig. 1) and was compared with predictions based on a Maxwell–Bloch-type beam-propagation model³ The method used to extract numerical results from this model decomposes the input laser field into an orthonormal set of Gauss–Laguerre modes, which usually describe free space propagation. The nonlinear interaction between laser beams and the medium can then be expressed by a set of ordinary differential equations for the amplitudes of the fundamental modes,⁴ At low input powers, experiments and calculations agree well, but the s-shaped part of the theoretical curve at higher input beam powers could not be observed in the experiment. This mode-decomposition approach produces results that are similar to those of the commonly used, more general split-step Fourier-transform procedure for numerical integration of the full partial-differenlial-equation system. Although the mode-decomposition method can be applied only for certain symmetric input field distributions, it has the advantage of being fast and of providing insight into the fundamental dynamics of the splitting process. Both methods have to introduce small symmetry-breaking initial conditions, which are naturally present in the experiment. In the mode-decomposition approach these asymmetries were achieved by adding small higher-order-mode amplitudes to the fundamental Gaussian input beam profile, whereas for the standard method a small separation of the input beams was used.

© 1993 Optical Society of America