Abstract
A major challenge in materials design is how to efficiently search the vast chemical design space to find the materials with desired properties. One effective strategy is to develop sampling algorithms that can exploit both explicit chemical knowledge and implicit composition rules embodied in the large materials database. Here, we propose a generative machine learning model (MatGAN) based on a generative adversarial network (GAN) for efficient generation of new hypothetical inorganic materials. Trained with materials from the ICSD database, our GAN model can generate hypothetical materials not existing in the training dataset, reaching a novelty of 92.53% when generating 2 million samples. The percentage of chemically valid (charge-neutral and electronegativity-balanced) samples out of all generated ones reaches 84.5% when generated by our GAN trained with such samples screened from ICSD, even though no such chemical rules are explicitly enforced in our GAN model, indicating its capability to learn implicit chemical composition rules to form compounds. Our algorithm is expected to be used to greatly expand the range of the design space for inverse design and large-scale computational screening of inorganic materials.
Highlights
Discovering new inorganic materials such as solid electrolytes for lithium-ion batteries is fundamental to many industrial applications
We propose the first generative adversarial generator loss and discriminator loss, which are defined as network model for efficient sampling of inorganic materials design space by generating hypothetical inorganic materials
The analysis shows that our generative generative adversarial network (GAN) can achieve much higher efficiency in sampling the chemical Variational autoencoder for evaluating GAN performance composition space of inorganic materials than the exhaustive enumeration approach
Summary
Discovering new inorganic materials such as solid electrolytes for lithium-ion batteries is fundamental to many industrial applications. Considering the huge space of doped materials with different mixing ratios of elements and many applications such as high-temperature superconductors, where six to seven component materials are common, the number of potential materials is immense Such combinatorial explosion calls for the need for more effective sampling approaches to search the chemical design space that employ existing explicit chemical and physical knowledge and implicit elemental composition knowledge embodied within known synthesized materials. To gain more efficient search, a variety of explicit chemical rules for assessing the feasibility of a given stoichiometry and the likelihood of particular crystal arrangements have been used in computational screening such as the Pauling’s rules (charge neutrality), electronegativity balance, the radius ratio rules[3], Pettifor maps[4] and etc Such approaches still fail to capture enough implicit chemical rules to achieve efficient chemical design space sampling
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have