Principal Investigator Barry Luther-Davies Project q04

Laser Physics Centre, Machine VP, CM

Research School of Physical Sciences and Engineering

Co-Investigators Xiaoping Yang and Jason Christou

Division of Materials Science and Technology, CSIRO, Clayton, Victoria and

Laser Physics Centre, Research School of Physical Sciences and Engineering

Dark Optical Solitons and Soliton Induced Waveguides

Photonics looks at mechanisms by which one beam of light is able to modify the behaviour of another, without intermediate electronics. By avoiding the conversion light to an electronic signal, photonic devices will able to operate much faster than their electronic counterparts. These devices will then be incorporated into ultra-fast optical computers and high bit rate communication systems.

In this project, photonic devices which utilise dark spatial solitons are being investigated. The dark soliton is a stable, controllable pattern of light which propagates in self-defocusing media and can be used to guide a secondary beam of light. It is this guidance, coupled with adjustability of dark soliton behaviour, that lays the foundation for dark soliton based photonic devices.

Previous work here has so far uncovered many novel and useful properties of dark soliton devices induced in planar structures. These investigations are now being extended to dark soliton devices in bulk, that is three dimensional, media. The bulk medium supports a dark soliton which is more commonly known as an optical vortex soliton. The guidance of a vortex is similar to that of an optical fibre, but the vortex may be manipulated where the optical fibre is static.

What are the basic questions addressed?

Vortices display a variety of interesting behaviours that have been noted throughout recent, relevant literature. We wish to find out which of these features will prove useful in photonic devices and if there are other useful behaviours. Theoretical and phenomenological models for vortex behaviour will be tested to find the parameters we can best use to effect and control vortices. Once a framework for understanding the vortex is obtained, experimental questions arise. How will deviations from ideal non-linear media or the ideal experimental components affect results obtained from models that assume ideal conditions? Simulation will play a role in answering all of these questions.

What are the results to date and the future of the work?

A number of simulations showing the splitting of high charge vortices under different conditions were carried out. It is conceivable that the splitting and recombination of vortices might form the basis for Y-junctions and coupling devices. Rotation of vortices and vortex pairs were simulated, showing potential as a type of rotary switching mechanism. Experimental work concerning vortex rotation has been carried out at the Laser Physics Centre, independently of this project, showing all the features of rotation observed in computations. Guiding of light in the vortex core has been demonstrated in simulation, and will hopefully proceed to experiment next year. Phase masks for generating vortices were made and their accuracy was tested computationally. It was found that for the coarse resolution of the experimental phase mask, the vortex field generated was very well behaved, displaying the high symmetry and phase accuracy required for experiment. Future work will focus on more specific device configurations and will test newly formed models of vortex behaviour.

What computational techniques are used and why is supercomputer required?

The beam propagation method is used to solve the equations for paraxial light propagation, as derived from Maxwell's equations. In terms of execution time, the beam propagation method is dominated by the two FFTs performed on a complex two dimensional array each iteration. The complex array represents the field cross section and each iteration propagates the cross-section a fixed distance. As this distance goes to zero, so does the computation approach the exact solution. Thus, for high accuracy over reasonable propagation distances the iteration count can be very large. The size of the complex array must also be very large, both for sampling accuracy and to avoid complications as features in the field approach the edge of the grid. On a scalar machine the two dimensional Fourier transform combined with the scalar nature of other matrix operations slows beam propagation computations to unworkable running times. Since each computational step comprising a single iteration is an operation on field matrix, the code is highly vectorizable and runs extremely efficiently on vector supercomputers. Run times for 10000 iterations using a 512 x 512 grid are only around 40 minutes with 97% vectorization.