Industrial-academic collaboration sheds light on polymer analysis

‘Although times have changed, we originally started out as a polymer analysis group, led for many years by Peter Schoenmakers’, says Bob Pirok, associate professor at the University of Amsterdam and group leader of CAST. ‘In that way we built up an extensive network through collaborations with the polymer industry.’

It was because of that network that Pirok was approached by Agilent Technologies, one of the world’s largest companies in the field of scientific instrumentation. ‘They have all manner of interesting “workhorses”, and they want to know how these can contribute to polymer analysis’, Pirok continues. ‘Polymers are a challenge for mass spectrometry, because very large molecules typically do not ionise well and are difficult to study.’ Agilent’s instruments have all sorts of functions and ionisation sources that Pirok’s group is keen to experiment with.

Pressure

In general, one can see that the polymer industry is under pressure from geopolitical fluctuations, tariff wars, and — not insignificantly — competition from China, which produces polymers far more cheaply. ‘In Europe there is also fairly progressive environmental legislation, which is not necessarily a bad thing, but it does put pressure on companies’, says Pirok. There is a great need for robust methods to better understand polymers and to outcompete rivals on the basis of knowledge. ‘It is therefore important for manufacturers to come forward with solutions. Agilent wants to demonstrate: this is what you can do with our equipment.’

‘Companies bring in technology and expertise, whilst academic freedom is preserved’

‘For us, it is a tremendous opportunity to work on method development with this advanced equipment and to see how our AI systems can extract even more from it’, Pirok continues. ‘We assist with publications, demonstrate what is possible, and identify which methods are useful for the polymer industry.’ Agilent is not involved in the substantive direction of the research programme, according to Pirok. ‘We therefore do not have to relinquish our scientific independence.’ In his view, the collaboration is a fine example of how the relationship between university and industry can work. ‘Companies bring in technology and expertise, whilst academic freedom is preserved. We are therefore always open to this kind of collaboration.’

AutoLC

Agilent’s Revident LC/Q-TOF system has since been integrated into the AutoLC network in Pirok’s laboratory, which is partly driven by AI and to which various systems are connected. ‘Our AI can operate all the instruments without human assistance’, the associate professor explains. ‘It has been programmed to identify interesting peaks in the data and then investigate these further on other instruments.’

AutoLC, which is not limited to LC but can also be applied to other separation techniques, is aimed at making the analyst’s work easier. ‘We present the chromatograms from our standard measurements to the AI. After the analyses, it can then try other methods to achieve a better separation.’ With polymers in particular, a great deal of trial and error is normally involved, but in this way AI can guide you to a good starting point from which to work further.

‘More complex techniques such as 2D LC/MS-MS then also come within reach’, says Pirok. ‘These normally take far too much time to set up yourself, but you can simply let the machine run day and night.’ Without people in the lab, it is important to know precisely what the machine is doing. The lab is therefore kitted out with all manner of coloured LEDs and control panels that allow you to monitor and correct AutoLC remotely. ‘You have to keep a close eye on it, because the AI has already nearly wrecked a few systems on occasion.’

AI winter

Humans will therefore always remain essential, according to Pirok. ‘Only humans can judge the value of a number that emerges from a method, and responsibility therefore remains with the human, never with the AI.’ Pirok does not believe that the day-to-day reality will change in any fundamental way. ‘It will change, however, for more complex techniques, even in less advanced labs. If you can automate those, you can use more sophisticated bells and whistles more easily. Take 2D LC, for instance. That has existed for thirty years, yet is still barely used in practice, even though everyone knows it can yield orders of magnitude more information. The effort and cost must therefore come down, and that is achievable through automation.’

What Pirok also wishes to emphasise is that AI can quickly become an unwarranted hype. ‘Yes, there is great potential for AI in automation, but we must not stumble into another “AI winter”. You see it time and again throughout the history of AI: in the fifties, chess had been cracked by AI, and people were dreaming wildly, whilst scientists were warning that many of those dreams might not yet be realisable.’ A great deal of funding was released, yet little came of it. The funding was withdrawn, and it was winter until the seventies. The same pattern repeated itself several times, and Pirok fears that another AI winter is currently on the horizon, precisely because the whole of society is hooked, whilst fundamental limitations still remain. ‘I think companies are too quickly and haphazardly looking to AI for fear of missing the boat. In my view, it is actually vitally important to exercise restraint, whilst not losing sight of the opportunities that genuinely exist.’