A key part of modern-day experimental Physics is striving to find ways to automate processes in experiments. To paraphrase Jake Taylor, writing for the Harvard Belfer Center, the incremental improvement of experiments is done by researchers looking at the apparatus they have and seeing what can be automated. This becomes somewhat of a research loop that occurs in tandem with improvements in experiments’ results.
Like with any set of processes in business that are to be automated, it is the repetitive tasks that are focussed on first as they are the easiest for a robot or machine to carry out. In Physics this is often the fabrication of materials used in experiments. A prime example of this is the early experiments performed on graphene in the mid to late 2000s. At first, the only way to isolate few-layer graphene samples was by a method called mechanical exfoliation. This was done by applying adhesive tape to a sample of graphite by hand and tearing it off with just the few-layer graphene sticking to the tape. Naturally, this process has now been automated by the introduction of more technologically advanced techniques.
Taylor characterises this as just the first step in the decision-making process of automating an experiment, with the process outlined in the above paragraph as the affected group in an experiment developing an initial automation solution.
Just like in business, once the initial automation solution has been prototyped, the next step is to devise a number of tests and checks to ensure a reliable final product. This must include subjecting the automating system to adverse situations by passing it undesirable or unwanted data/materials etc. Taylor also writes that this can take the form of ‘both software and hardware sandboxing’, creating an isolated test environment, ensuring performance won’t do any harm before the solution is used in a full experiment.
The final requirement of introducing automation to experiments is that they must continue to yield results that advance the scientific goal. This means that developers must continually check the system to prove it produces the results desired. This kind of dynamic differs slightly from a business setting as in science, all affected parties of the automation are “developers” themselves, whereas in business someone having part of their repetitive job automated is not a “developer” in the same sense.
To conclude, a recent example of novel automation within Physics research was performed by M. Connett-Brown et al. at the University of Exeter in 2021. The group were able to automate the process of fitting Raman Spectra of Graphene flakes given many data sets. This was done in order to determine layer number and thickness for use in electronic applications. They plan to expand the scope of their investigation in the coming years.
Would you like to exchange ideas with us on the subject of digital transformation and process automation without obligation? Let’s talk!