Pores for thought: The use of generative adversarial networks for the stochastic reconstruction of 3D multi-phase electrode microstructures with periodic boundaries

The generation of multiphase porous electrode microstructures is a critical\nstep in the optimisation of electrochemical energy storage devices. This work\nimplements a deep convolutional generative adversarial network (DC-GAN) for\ngenerating realistic n-phase microstructural data. The same network\narchitecture is successfully applied to two very different three-phase\nmicrostructures: A lithium-ion battery cathode and a solid oxide fuel cell\nanode. A comparison between the real and synthetic data is performed in terms\nof the morphological properties (volume fraction, specific surface area,\ntriple-phase boundary) and transport properties (relative diffusivity), as well\nas the two-point correlation function. The results show excellent agreement\nbetween for datasets and they are also visually indistinguishable. By modifying\nthe input to the generator, we show that it is possible to generate\nmicrostructure with periodic boundaries in all three directions. This has the\npotential to significantly reduce the simulated volume required to be\nconsidered representative and therefore massively reduce the computational cost\nof the electrochemical simulations necessary to predict the performance of a\nparticular microstructure during optimisation.\n