Putting density functional theory to the test in machine-learning accelerated discovery for transition metal chemistry

Professor Heather J. Kulik

Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA

Many compelling functional materials and highly selective catalysts have been discovered that are defined by their metal-organic bonding. The rational design of de novo transition metal complexes however remains challenging. First-principles (i.e., with density functional theory, or DFT) high-throughput screening is a promising approach but is hampered by high computational cost, particularly in the brute force screening of large numbers of ligand and metal combinations. In this talk, I will describe how automation[1-3], machine learning[4-6] autonomous tool development[7], sensitivity analysis, and multireference character detection[8] can all be used to simultaneously overcome limitations in cost and accuracy of DFT-based screening in open shell transition metal chemistry. I will describe the basic approach of our automated workflow, molSimplify[1-3], for automating calculations as well as recent extensions of machine learning (ML) models that avoid direct calculation of not just energetic or structural properties but also the likelihood of a calculation to succeed or a complex[7] to contain strong correlation[8]. I will describe how unifying databases shed light on the sensitivity of discovery outcomes to underlying DFT functionals. I will discuss possibilities for accelerating discovery with ML[9] with awareness of uncertainty[10] when sampling new regions of chemical space. Time permitting, I will describe how this powerful toolkit has advanced our understanding of metal-organic bonding in materials far-ranging from functional spin crossover complexes to open-shell transition metal catalysts and metal-organic frameworks.

Recording:

References

[1] Ioannidis, E. I.; Gani, T. Z. H.; Kulik, H. J. molSimplify: A Toolkit for Automating Discovery in Inorganic Chemistry. Journal of Computational Chemistry 2016, 37, 2106-2117.

[2] molSimplify lite web interface: https://molsimplify.mit.edu

[3] Kulik group github with molSimplify and MultirefPredict: https://github.com/hjkgrp

[4] Janet, J. P.; Kulik, H. J. Resolving Transition Metal Chemical Space: Feature Selection for Machine Learning and Structure-Property Relationships. The Journal of Physical Chemistry A 2017, 121, 8939-8954.

[5] Janet, J. P.; Liu, F.; Nandy, A.; Duan, C.; Yang, T.; Lin, S.; Kulik, H. J. Designing in the Face of Uncertainty: Exploiting Electronic Structure and Machine Learning Models for Discovery in Inorganic Chemistry. Inorganic Chemistry 2019, 58, 10592-10606.

[6] Janet, J. P.; Kulik, H. J. Machine Learning in Chemistry. ACS InFocus Book Series. 2020.

[7] Duan, C.; Janet, J. P.; Liu, F.; Nandy, A.; Kulik, H. J. Learning from Failure: Predicting Electronic Structure Calculation Outcomes with Machine Learning Models. Journal of Chemical Theory and Computation 2019, 15, 2331-2345.

[8] Liu, F.; Duan, C.; Kulik, H. J. Rapid Detection of Strong Correlation with Machine Learning for Transition-Metal Complex High-Throughput Screening. The Journal of Physical Chemistry Letters 2020, 11, 8067-8076. 

[9] Janet, J. P.; Ramesh, S.; Duan, C.; Kulik, H. J. Accurate Multiobjective Design in a Space of Millions of Transition Metal Complexes with Neural-Network-Driven Efficient Global Optimization. ACS Central Science 2020, 6, 513-524.

[10] Janet, J. P.; Duan, C.; Yang, T.; Nandy, A.; Kulik, H. J. A Quantitative Uncertainty Metric Controls Error in Neural Network-Driven Chemical Discovery. Chemical Science 2019, 10, 7913-7922.