Effective design of semiconductor devices using evolutionary-based derivative free optimization

Designing semiconductor devices is a complex problem, which requires accurate models and effective optimizers. We have introduced a new methodology based on derivative-free optimization evolutionary algorithms, which aim to achieve a near-optimal design. We have tested these algorithms on various semiconductor devices using a drift-diffusion model; the experimental results show that CMA-ES and DE outperform the current state-of-the-art design algorithms.