Computational Design

With the proliferation of partial differential equation (PDE) simulation software, computation is emerging in nanoscience as a driver of new research. Yet computational design typically takes the form of either black-box stochastic optimization or few-parameter variations, due to the lack of efficient design tools. Development of new tools and techniques for design and control of wave systems is poised to simultaneously improve our physical understanding of wave physics as well as transform the application space. To achieve such prominence, computational design must be:
  • Fast. Adjoint-based methods, which are the backbone of high-speed optimization in fields from control theory and deep learning to aerodynamic wing design, promise to significantly improve design in electromagnetism.
  • Manufacturable. A key test for optimization tools and techniques will be the extent to which they can incorporate fabrication/synthesis constraints while retaining speed and flexibility.
  • Robust. The importance of a large design space is not to fine-tune thousands of parameters, but rather to find which 10 or 20 degrees of freedom (DOFs) are important. It is critical to avoid spurious local optima and to uncover the best-performing architectures. 
As computing power increases, simulation becomes faster and more accurate, and design takes on more importance in the scientific process. It is an exciting place to be!
Select publications:
  • Light trapping textures designed by electromagnetic optimization for subwavelength thick solar cells, V. Ganapati, O. D. Miller, and E. Yablonovitch, IEEE J. Photovoltaics 4, 175–182 (2014)
  • Adjoint shape optimization applied to electromagnetic design, C. M. Lalau-Keraly, S. Bhargava, O. D. Miller, and E. Yablonovitch, Opt. Express 21, 21693–21701 (2013)
  • Fundamental limits to extinction by metallic nanoparticles, O. D. Miller, C. W. Hsu, M. T. H. Reid, W. Qiu, B. G. DeLacy, J. D. Joannopoulos, M. Soljačić, and S. G. Johnson, Phys. Rev. Lett. 112, 123903 (2014)
  • Optimization of sharp and viewing-angle-independent structural color, C. W. Hsu, O. D. Miller, S. G. Johnson, and M. Soljačić, Opt. Express 23, 9516 (2015)