name: materialdiscovery

Accelerating material discovery with machine learning

.center[ :scale 80% ]

Accelerating material discovery with machine learning

Motivation

.center[

Hydrogen and methane storage — .cite[(Lawrence Zitnick et al., 2020)]

]

Renewable energy can be used to transform water into hydrogen or methane and back to electricity
However, current electrocatalysts are not sufficiently energy-efficient (35 % for round-trip AC to AC)

.references[

Lawrenece Zitnick et al. An introduction to electrocatalyst design using machine learning for renewable energy storage. arXiv:2010.09435, 2020. ]

Why machine learning?

Traditional electrocatalyst design

.context[Current electrocatalyst are only up to 35 % energy efficient]

.right-column-66[.center[ :scale 90% ]]

–

.left-column-33[ A relaxation of propane (C3H8) on a copper (Cu) surface.

.center[ :scale 100% ] ]

count: false

Why machine learning?

Traditional electrocatalyst design

.context[Current electrocatalyst are only up to 35 % energy efficient]

.right-column-66[.center[ :scale 90% ]]

.left-column-33[

Density Functional Theory is used to estimate the energy of a catalyst-molecule structure
DFT scales with $O(n^3)$ with the number of electrons
The calculations for one structure take hours or days
There are combinatorially many possible candidate materials ]

Why machine learning?

ML world model

.context[Physical models are computationally too expensive for fast discovery.]

.right-column-66[.center[ :scale 90% ]]

–

.left-column-33[

Data from physical models can be used to train ML-based approximators
ML models can be used to more rapidly evaluate candidate materials ]

–

Can we do better?

Why machine learning?

RL-based exploratory policy

.context[Using ML to only score candidate materials provides only linear gains.]

.right-column-66[.center[ :scale 90% ]]

–

.left-column-33[

We can train ML models to more efficiently search the space of candidate materials
An RL agent could exploit the structure of the search space ]

Promising results: GFlowNet (Bengio et al., 2021)

Accelerating scientific discovery

Summary

.right-column[.center[ :scale 90% ]]

.left-column[

World model: graph neural networks (GNN), capable of incorporating invariances and equivariances that preserve physical properties
Exploratory agent: RL-based algorithms capable of learning the structure of the world to propose diverse, high-reward candidates ]

.full-width[ .conclusion[These principles have applications in material discovery, drug discovery, causal reasoning, etc. and have the potential of pushing the boundaries of machine learning research.] ]