Multi-Solution Inverse Design in Photonics using Generative Model

 We have created an Inverse Design architecture that overcomes the non-uniqueness of nanophotonic structures by combining a genetic algorithm with a β Conditional Variational Autoencoder. This approach captures multiple solutions, providing flexibility in selecting design variables for specific requirements. Experimental validation involved designing variable thickness TiO2-SiO2 multilayer thin films in the visible range. The method provides a fast and robust photonics inverse design solution.