Synthesis and Optimisation in Optoelectronics

Automatic design of many complex optoelectronic structures such as photonic crystals and facet coatings can be performed using global multidimensional optimisation methods [1]. Using global methods such as Evolutionary Algorithms or Simulated Annealing a user defined figure of merit can be minimised over a large problem space, automatically obtaining a structure with desired operating characteristics.

To gain accurate results such methods require repeated calculation of numerical solutions which can lead to long runtimes. One way to remedy this is by the use of parallel computation to spread the simulation burden. Alternatively we can avoid intensive numerical simulations and employ approximate and semi-analytical methods where possible.

This project investigates how use of a hierarchy of simulation methods within device optimisation can be used to exploit model redundancy and obtain accurate device designs as efficiently as possible. Using convergence information gathered from the optimsation routine the tool will automatically trade-off solution accuracy for runtime, speeding up optimisation by avoiding slow numerical calculations until necessary. The code has been developed to take advantage of in-house parallel computing facilities, to enable higher throughput and faster discovery of optimal device parameters.

Our optimisation routine has been successfully applied to a number of practical optoelectronics design problems