Accelerated Materials Development for Manufacturing (AMDM)


Introduction

2020 10 14 AMDM 1(website)
There is a time-scale mismatch between materials R&D technology development from labscale to a product, and the relevant enablers such as government funding and student/postdoc residency. To solve this mismatch, we seek a solution to accelerate the R&D by a factor of >10X. The goal of the AMDM program is to enable this acceleration via high-throughput automation of experiments and characterization techniques, and machine learning optimization of experimental inputs with the aim of fastest convergence towards the optimum[1].


2020 10 14 AMDM 2(website)
Our team combines deep materials domain expertise with state-of-the-art machine learning models to achieve optimization, design of experiments as well as inverse design.

Capabilities

2020 10 14 AMDM 3(website)
  • We have built a high-throughput syringe pump flow setup that is capable of automated control of fluid flow and mixing of several components. The flow setup is integrated with several high-throughput characterization tools including hyperspectral UV-Vis imaging, in-line Raman tool and automated 4-point probe measurement. Datasets are optimized using machine learning forward prediction (e.g. Random forest, Neural network) and inverse design (e.g. Bayesian optimization) algorithms to quickly converge towards optimum experimental conditions.
  • We have also built a high-throughput chemical workstation that incorporates automated pipetting, reactor blocks, and spin-coating unit, together with UV-Vis and fluorescence plate reader. This extends our high-throughput synthesis capabilities to batch reactions in inert atmospheres and fabrication of thin films. The experiments are run automatically and robotically, guided by our machine learning algorithms.


Achievements

  • The machine learning workflow has been successfully demonstrated to predict cloud point transition temperatures in a class of poly(2-oxazolines) polymers, and new polymers have been designed and synthesized via inverse design.[2]
  • In collaboration with university partners our team is working on machine learning optimized synthesis of nanoparticles and the design of inorganic materials.[3,4]
  • We have also been working with an MNC to deploy the machine learning optimization workflow into their fabrication processes for functional devices. This builds on material descriptors-based optimization using machine learning in functional materials, devices and systems.[5,6]
  • In addition to academic and industrial projects, we have been actively involved in introducing machine learning to the community via workshops and machine learning courses.


References

  • [1] JP Correa-Baena et. al., “Accelerating materials development via automation, machine learning, and high-performance computing”, Joule 2018 2 (8), 1410-1420
  • [2] J. N. Kumar et. al., “Machine learning enables polymer cloud-point engineering via inverse design”, npj Computational Materials 2019, 5, 73.
  • [3] F Mekki-Berrada et. al., “Two-Step Machine Learning Enables Optimized Nanoparticle Synthesis”, chemxriv preprint 2020,  doi.org/10.26434/chemrxiv.12673742.v1
  • [4] Z. Ren et. al., “Inverse design of crystals using generalized invertible crystallographic representation”, arXiv preprint 2020 arXiv:2005.07609
  • [5] Y Xu et. al., “Machine learning-assisted cross-domain prediction of ionic conductivity in sodium and lithium-based superionic conductors using facile descriptors.” Journal of Physics Communications 2020 4, 055015
  • [6] A. Suwardi et. al., “Inertial effective mass as an effective descriptor for thermoelectrics via data-driven evaluation”, Journal of Materials Chemistry A 2019 7 (41), 23762-23769


Contacts

Prof. Kedar HIPPALGAONKAR, kedarh@imre.a-star.edu.sg
Dr. Jatin Kumar, kumarjn@imre.a-star.edu.sg
Dr. Lim Yee Fun, limyf@imre.a-star.edu.sg