Programmable autonomous chemical robots for discovery and synthesis

Caramelli, Dario (2019) Programmable autonomous chemical robots for discovery and synthesis. PhD thesis, University of Glasgow.

Due to Embargo and/or Third Party Copyright restrictions, this thesis is not available in this service.

Abstract

The work presented in this thesis focuses on the development of a platform to explore chemical spaces for reaction discovery and a network of robots for automatic collaboration. We believe that in recent years scientific automated systems have become an invaluable asset in the laboratory, vastly improving the productivity and generally changing the approach to chemistry research in many fields. However, the current implementations are mostly for technical help in repetitive tasks or in optimisation of already known reactions. We envision a closed-loop platform pointed towards the unknown, able to automatically perform reactions, analyse them, and use the acquired data to navigate a chemical space and make discoveries.
The platform configuration went through a series of gradual improvements and expansions: it started with a single reactor and 4 reagents to finish with 6 parallel reactors, 20 reagents, an inert atmosphere line and three LEDs for photochemical reactions. As analysis equipment we used three benchtop instruments: NMR, MS and IR. All of them were automatically controlled and the data produced was processed using two different algorithms for reactivity assessment: a features extractor and a neural network. Chemical space exploration was also simulated using a neural network correlating the reaction parameters with the reactivity.
The system has been used for three applications of increasing complexity. The first one was a reaction optimization performed with two different approaches: a polynomial regressor model and a genetic algorithm. The second task was the exploration of a simple chemical space made of 6 organic molecules and a base. Lastly, the system explored a larger chemical space built to discover new photochemical reactions. By using real time analysis, it was possible to identify promising candidates and successfully discover a new multicomponent reaction, a new photochemical reaction, and re-discover various examples already known in literature. The most interesting reaction has also been investigated manually in order to test its robustness.
As a parallel project this thesis will also tackle the science reproducibility problem. We believe that chemistry automation will ensure a better preservation of scientific data, since the experiments’ parameters will be exchanged as defined robotic operations instead of ambiguous manual procedures. Expanding this concept one step forward we imagine a future where laboratory machines are interconnected through the internet, allowing automatic collaboration and data sharing. As a proof of concept we built a network of physically separated robots and showed how it is possible to use internet communication to search an azo-dye chemical space in a fraction of time as well as encoding and decoding information into a network of oscillating reactions.

Item Type: Thesis (PhD)
Qualification Level: Doctoral
Keywords: Discovery, automation, algorithm, machine learning, organic synthesis, data sharing, isocyanide, Twitter.
Subjects: Q Science > QD Chemistry
Colleges/Schools: College of Science and Engineering > School of Chemistry
Supervisor's Name: Leroy, Prof. Cronin
Date of Award: 2019
Embargo Date: 1 August 2022
Depositing User: Dr Dario Caramelli
Unique ID: glathesis:2019-74322
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 21 Aug 2019 12:18
Last Modified: 05 Mar 2020 22:33
Thesis DOI: 10.5525/gla.thesis.74322
URI: http://theses.gla.ac.uk/id/eprint/74322
Related URLs:

Actions (login required)

View Item View Item