UK researchers have taken the first step towards building a robotic chemist that can invent new materials and work out how to manufacture them.
Scientists working on the Brunel University-led project have developed a system that automatically experiments with feedstocks to create new substances with specified properties.
They hope this technology could one day dramatically speed up the processes by which new molecules, nanoparticles and drugs are created while simultaneously developing scalable manufacturing processes that minimise waste.
‘The technology won’t put chemists out of a job but rather automate the more laborious aspects of their work’
Dr Harris Makatsoris, Brunel University
‘The vision is a search engine for materials,’ said Dr Harris Makatsoris, project leader and head of Brunel’s Advanced Manufacturing and Enterprise Engineering (AMEE) research group.
‘You put in the properties you want and there is the infrastructure including the physical system to try out what can be done,’ Makatsoris told The Engineer. ‘You end up with your sample, your structure and your manufacturing recipe. That’s the 20-year vision.’
The prototype system takes feedstocks suggested by the user and experiments with them in a miniature flow reactor – a device for carrying out chemical reactions in a continuous stream rather than in batches – until it produces a new material that has the desired properties.
Computer algorithms allow the machine to learn from each experiment and design the next one, running through many iterations just as a human chemist would to find the right conditions to produce the right material.
The researchers, who are part of the EPSRC Directed Assembly Grand Challenge Network led by Brunel and Bath universities, have so far used the system to create organic molecules similar to those used in drugs, inorganic polyoxometalate nanoparticles, and porphyrin molecules that are expensive to synthesise because they are currently made in small quanitities.
Using a flow reactor allows the experiments to be carried out in short succession and provides a ready-made manufacturing method that can easily be scaled up. ‘Once you hit the sweet spot you can let the machine run or take your recipe to a larger scale flow reactor,’ said Makatsoris.
The research team is now planning to work with industrial partners to experiment with devising liquid food formulations, lubricants and fuels. They also hope that future versions of the system may work by designing specific molecules.
‘There are all sorts of other things we have to consider in order to enhance this capability and actually do material design in the proper sense,’ said Makatsoris. ‘But right now we measure the property, such as the colour or the binding site, and once we measure it we can direct the system towards this.’
One experiment the researchers performed in collaboration with Glasgow University was to use the machine to create nanoparticles that produced different coloured solutions.
‘Depending on the conditions and how you manage them, you can end up with one type of molecule that is a fullerene-shaped [spherical] that shows as brown,’ said Makatsoris.
‘On the other end of the space there is a donut shaped molecule that appears to be blue. When we were doing the search we were able to switch from the brown to the blue by specifying the colour of the solution.’
Makatsoris also insisted that the technology wouldn’t put chemists out of a job but rather automate the more laborious aspects of their work. ‘What flows into the tubes, provided the conditions are correct, is up to the chemist who sets up the experiment,’ he said.
‘Then, of course, there is the opportunity for scientists to do all sorts of analysis and feed that data into computer models to reconstruct what is happening. This is the step we’re looking towards.’
This article was amended on 11 April 2013 to clarify that the project is part of the EPSRC Directed Assembly Grand Challenge Network led by Brunel University, not the EPSRC Dial-a-Molecule Grand Challenge Network led by Southampton and Leeds universities.