Closed stefanbringuier closed 1 year ago
Hi @stefanbringuier, thanks for the contribution! Sorry it took a minute to get to this. This is a neat work that you found. I was looking through the paper, the workflow seemed to be:
After completion of the automated procedures, crude reaction mixtures were purified using medium pressure liquid chromatography (MPLC) or preparative high-performance liquid chromatography (HPLC)
Single-molecule conductance measurements were performed using a home-built scanning tunneling microscope setup
Steps performed for a subset of molecules:
We further performed flicker noise analysis experiments for R6 at a solution concentration of 1βmM.
Electron transport calculations were performed using the nonequilibrium Greenβs function-density functional theory (NEGF-DFT) method via the Atomistix Toolkit package.
From what I could tell, only the Burke-type small-molecule synthesizer was fully automated (i.e., one step of the synthesis). I didn't see something that indicated automatic transfer to the liquid chromatography piece or the characterization equipment (STM-BJ), and I also didn't see an indication of closed-loop optimization (i.e., multiple iterations). Can you clarify?
@sgbaird Hmm, so you're correct, that the STM-BJ is not part of the automation nor used for closed-loop optimization. My understanding/read is the automation is focused on the synthesizability/confirmation of target molecules. So, the paper really has two parts: automated synthesis and single-molecule charge transport experiments. The first is "instruction-driven" while the second is human-performed characterization. The supplementary information in one of the references they give further supports that this is automated by instruction and final molecules are confirmed using in-situ(?) and ex-situ characterization, but there is no optimization since they follow known(?) synthesis recipes.
I guess for me this raises the question of whether this paper really falls within "self-driving labs"? Should papers considered include a decision-making step in order to be added to the list? I think so (maybe this was obvious, sorry) and therefore I would not actually add this paper. I'll let you decide, so feel free to close/reject if you agree. Thanks for the due diligence.
Maybe somewhere in contributing guidelines we could add a table or checklist to determine if a paper fits, something like:
Aspect | Self-Driving Lab | Automated Lab |
---|---|---|
Execution | Autonomously conduct experiments | Human-programmed experimental procedures |
Decision Making | Adapts based on outcomes | Limited to programmed tasks |
Scope | Capable of data-driven discoveries/optimization | Suited for repetitive high-throughput tasks |
Error Handling | Self-diagnosis, course correction | Requires human intervention |
Closing for now as I no longer think this fits scope for self-driving labs
Description
The added paper is about using automated synthesis to prepare a library of conjugated oligomers with systematically varied side chain composition followed by single-molecule characterization of charge transport. The project involved the development of a small-molecule synthesizer capable of parallel runs. The increased throughput allowed for the screening of large regions of chemical space in a single automated synthetic run. The level of autonomy used in the project is fully autonomous π§ͺπ¬ποΈπ».
Entry
Added the following entry to SDL Academic Research 2022 section:
π§ͺπ¬ποΈπ» | Using Automated Synthesis to Understand the Role of Side Chains on Molecular Charge Transport. Li, S.; Jira, E. R.; Angello, N. H.; Li, J.; Yu, H.; Moore, J.S.; Diao, Y.; Burke, M. D.; Schroeder, C. M. Nat. Commun. 2022, 13, 2102.