Challenge:

Our main focus is meta-stable solid systems, which mean that their properties strongly depend on the reaction history. The main challenge is to find conditions that result in the desired phase with a desired properties. One exaple is Cu1+ that tends to oxidize to Cu2+ or reduce to Cu0 [1]. But simialrly, if we want to stabilize a glass, or nano-crystal, or a meta-stable intermediate, we need to have control an a navigation system that will tell us when to turn the heat on, to what temperature, and when to stop. To navigate in a multi-dimensional space especially when a new compound requires a particular structural state, one needs a smart navigation system.

Mission :

To navigate through the complex chemical state we proposed to use an active reaction control for targeting a desired product with a particular chemical signature [1]. In that work we manged to maximize the concentraion of a Cu+1 oxide by iteratively controlling the oxidation power. This current project comes to expand this work into the world of materials that live in the transition state and constantly shift from one to the other, such as glasses and phase change materials. With the help of Machine Learning tools, such as 'reinforcement learning', we intend to explore the infrastructure that is required for allowing one to target a desired strucutral state in glasses, so one can navigate towards a meta-stable structure, such as glasses in a particular strucutral form.

 Related Publications:  

[1] Active Reaction Control of Cu Redox State Based on Real-Time Feedback from in Situ Synchrotron Measurements

Rakita, Y., O'Nolan, D., McAuliffe, R. D., Veith, G. M., Chupas, P. J., Billinge, S. J. L. & Chapman, K. W., 4 Nov 2020, In: Journal of the American Chemical Society. 142, 44, p. 18758-18762 5 p.