Unlocking Faster Muon Data Analysis through Advanced Event Detection
Researchers at STFC Scientific Computing’s Ada Lovelace Centre (ALC) are delivering transformative advances in the processing and analysis of muon spin rotation, relaxation, and resonance (μSR) detector data. By developing innovative signal processing algorithms, this project addresses one of the biggest challenges facing next-generation muon spectroscopy: the ability to identify decay events accurately and in real time from massive streams of digitised detector data.
The Challenge
Muon spectroscopy is a powerful experimental method used to explore the properties of materials at an atomic scale. By implanting spin-polarised muons into a sample and monitoring their decay, scientists gain unique insights into superconductivity, energy storage devices, catalysis, and emerging quantum materials. These techniques are central to many areas of fundamental and applied science.
Modern experiments, especially at pulsed muon sources like ISIS Neutron and Muon Source, run under extreme data conditions. Each muon pulse can implant tens of thousands of muons within nanoseconds. As these muons decay, they emit positrons, which are detected as signals across hundreds of channels. Scaling up these measurements brings extraordinary scientific potential—but it also creates a computational bottleneck.
The forthcoming Super-MuSR instrument exemplifies this challenge. When it becomes operational in 2027, it will house 960 independent detector channels, up from just 64 on the current MuSR instrument. This leap in capability requires real-time digital signal processing capable of handling throughput in the order of one million events per second, ensuring experiments can run without data loss or long acquisition times.
The far higher counting rate of the new instrument will be applied to understanding materials like the superconducting tapes being used in nuclear fusion reactors, battery pouch cells used in electric vehicles and consumer electronics, and quantum decoherence in model systems.
The new signal processing methods trialled in this work will enable higher throughput experiments, with better quality data, over the full range of materials probed with muon spectroscopy, allowing facility users to access the full benefit of Super-MuSR for research on quantum matter, energy materials and beyond.
Rhea Stewart, Instrument Scientist, ISIS Neutron and Muon Source
Our Approach
The project aimed to design and validate event detection algorithms that improve on traditional analogue-based systems while preserving the precision required for μSR science. The algorithms need to identify true muon decay events reliably, filter out noise, and distinguish overlapping events that occur within extremely short time windows.
Key performance goals include:
- High temporal resolution – ensuring precise timing of decay events.
- Pulse-pair discrimination – reliably separating closely spaced events.
- Noise robustness – maintaining performance under challenging signal conditions.
- Computational efficiency – achieving near real-time speeds suitable for deployment in high-throughput environments.
What Has Been Achieved So Far
The ALC team conducted an extensive review of existing event detection strategies and then implemented and tested four core approaches:
- Fixed Threshold Discrimination – a baseline approach that is lightweight and fast.
- First-Derivative Thresholding – detects the rate of signal change for better timing accuracy.
- Smoothed Second-Derivative Thresholding – emphasises curvature changes to separate overlapping decay pulses effectively.
- Multiscale Pyramid Preprocessing – introduces a new way to denoise signals using hierarchical decomposition before event detection.
To evaluate these methods, the project used a dual testing strategy. Algorithms were benchmarked against simulated data with known ground truth to measure detection accuracy and were also validated on real detector outputs from the ISIS EMU and HiFi instruments and a prototype of the Super-MuSR system.
Results show that derivative-based and multiscale methods dramatically outperform traditional fixed-threshold detection at higher event rates. Compared to the baseline, they enable significantly higher usable count rates—up to a factor of four—allowing acquisition times to be cut by a similar margin. This improvement is especially valuable for experiments dealing with rare or time-sensitive phenomena, where every extra data point matters.
This work demonstrated that changing the signal processing should roughly double the rate at which the muon decays can be detected in each detector and also improve the time resolution of those detections by a factor of two. This means that we can use more of the available muon beam for the detectors we have.
Peter Baker, Instrument Scientist, ISIS Neutron and Muon Source
Technical Insight: Advanced Algorithmic Design
-
Laplacian-of-Gaussian Filtering – By computing a smoothed second derivative using a Gaussian-Laplacian approach, the algorithm becomes sensitive to local curvature, enhancing its ability to pick apart overlapping pulses. Smoothing suppresses high-frequency noise, yielding robust detection even under challenging conditions.
-
Pyramidal Multiscale Denoising – The project employs subdivision-based pyramidal transforms, offering a computationally efficient alternative to wavelet transforms. This hierarchical decomposition distinguishes meaningful signal features from random noise across different scales, improving signal-to-noise ratio while preserving sharp transitions crucial for accurate event timing. Unlike conventional wavelets, this approach is lightweight, making it suitable for real-time streaming and embedded systems.
Integrating these algorithms into the digital pipeline of next-generation detectors could unlock substantial performance gains. Faster processing means reduced experimental dead time, better data quality, and greater scientific throughput for large-scale facilities like ISIS.
Function fitting has proven challenging in this project, primarily due to its computational cost and the low signal-to-noise ratio of some the data. Nevertheless, advances in optimisation algorithms and hardware acceleration may enable more localised or hybrid fitting strategies. Embedding function fitting within the event-detection pipeline as a complementary refinement step to existing peak-finding methods could provide improved accuracy while mitigating its computational cost, making this a promising direction for future work
Boris Shustin, Computational Scientist, Scientific Computing
Future Directions
The next phase will focus on embedding these algorithms into firmware for Super-MuSR and exploring their scalability across different hardware platforms. This transition will ensure that the benefits seen in current tests translate into real-time operational environments. Looking ahead, similar approaches could be applied to other time-resolved experiments, amplifying the project’s impact beyond muon science.
DOI: 10.5286/stfctr.2026017
URL : ePubs
To find out more, please get in touch here.
ALC@stfc.ac.uk
The Ada Lovelace Centre is part of STFC Scientific Computing and is funded by UKRI
