Knowing how long to scan for until you've got enough data is a tricky balance for users of neutron facilities. While longer scans can mean more reliable statistics, the gains become marginal and sometimes it can be better to stop and make the most of the beamtime by measuring another sample instead.
Jos Cooper from the ISIS neutron reflectometry group wondered if the optimal measurement time for an experiment could be quantified to advise users during their beamtime. After a few months finishing off a project started by a previous student, his industrial placement student James Durant was able to take on this challenge.
James has spent a year with ISIS, on placement from his computer science degree at the University of Warwick. He went into this experimental design project with the aim of designing a framework that could use applied statistics to provide information on experimental parameter uncertainties in near real time.
“The framework James has ended up developing actually does a lot more than we had planned," explains Jos; “by simulating experiments, and individual measurements, he has been able to see the effects of not only changing the instrument parameters, but also the way the sample is prepared."
For neutron reflectivity experiments, the aim of the experiment is to determine the properties of the layers being studied. James' framework quantifies the information in these properties, enabling the experimental conditions to be accurately optimised.
After developing the technique, and publishing the results, James then began work on applying it to different samples. “In one magnetism study, we worked out what conditions would be needed to prove that a magnetic moment was present, and then showed how long you would need to measure the sample for to reach a specific confidence level," says Jos. “In this example, we saw that if the sample had been made just 5 nm thicker, it would have taken an extra ten hours of beamtime to reach the same level of confidence."
Another article has been submitted to the Journal of Applied Crystallography that presents the application of the code. “I never expected to publish anything during my placement year," says James (pictured, right); “I have been very surprised to have multiple publications come out!"
Being able to provide users with this sort of information before they arrive will be incredibly useful for making the most of their valuable beamtime. In some cases, adding just a small extra layer to the sample can make the measurements up to ten times faster.
“We never expected the biggest gains to come from changes you can make before the sample even enters the beamline," explains James. “As well as this, if someone knows the basis of the model representing their sample, the framework can be used to determine how best to run the experiment."
This work has started on the ISIS neutron reflectometry instruments, but could be used more widely. By inputting different instrument profiles, it could be rolled out to reflectometry beamlines at other facilities, and Jos is working with ISIS staff in the SANS and muon groups to apply it to those techniques.
The initial study can be found online at DOI: 10.1107/S160057672100563X; The second study is currently under review, but available at https://arxiv.org/abs/2108.05605.