Nottingham's robot lab picks its own next coating to test

Nobody at Nottingham decides what the robot tests next. The software does. Researchers at the University of Nottingham's Centre of Excellence in Coating and Surface Engineering have built a self-driving laboratory for coatings designed to survive extreme heat, European Coatings reported on 5 November. The work is published in *Engineering Applications of Artificial Intelligence*.

How it works

The set-up uses two robotic arms, a custom liquid-dispensing tool, other lab equipment and purpose-built software. It runs the whole sequence itself: measuring and mixing chemicals, processing samples in a high-temperature furnace, controlling the cooling and retrieving the finished sample.

Get Chem Data Digest every TuesdayThree-minute read. Free. One-click unsubscribe.

The clever bit is the loop. The system uses active learning, so the results from each round of experiments decide which compositions get tested next. Instead of working through a grid of recipes, it goes where the data says the best candidates are likely to be.

In early runs it dispensed liquids accurately during calibration, then prepared a precursor solution, fired it, cooled it and pulled out the sample with no manual help. It has already flagged several candidate pyrochlore materials for thermal barrier coatings, which need low thermal conductivity.

Who's behind it

The researchers named in the report are Professor Tanvir Hussain, Professor of Coatings and Surface Engineering, and Dr Mohammadreza Lalegani Dezaki, Research Fellow in Mechanical and Manufacturing Engineering. No industrial partner or funding figure is named in the report.

Why it matters for R&D software

Self-driving labs get a lot of airtime in pharma. This one is pointed squarely at materials and coatings, where development still leans on experienced formulators and a lot of trial and error. The bottleneck it attacks isn't the robot. It's the decision about what to try next, which is exactly the problem materials informatics and formulation AI platforms sell against.

For vendors, academic rigs like this matter for two reasons. They prove out the closed-loop workflow on hard, high-temperature chemistry. And they produce the people who know how to run one: researchers comfortable with robotics, experiment design and machine learning at the same time.

Commercial adoption in coatings will take longer than a single paper suggests. But the direction is clear, and the software layer that plans, records and learns from experiments is where most of the value sits.

What it means for hiring

In our view, researchers who combine lab automation, active learning and real materials chemistry are some of the scarcest hires in this market. Informatics vendors building closed-loop products will compete hard for them.
Selling autonomous experimentation into coatings companies needs solution scientists who can speak to formulators as peers. Expect demand for application scientists with both lab and data backgrounds.
University coatings centres are a useful early talent pipeline for vendors, especially for product and research engineering roles.

Share Ask Anna about this →