INTERVIEW – As a boy from Amsterdam-East, he regularly attended the open days at the Nikhef particle physics laboratory. Now, computer scientist Stephen Swatman has won a prize from the ATLAS experiment at CERN for his thesis.
‘I think I came to Nikhef’s open day every year; I thought it was a magical place,’ says Stephen Swatman about his relationship with the institute. Although he did not study physics but computer science at the University of Amsterdam, he did return to Nikhef for his master’s thesis.
‘Literally across the Kruislaan. I worked on software for the reconstruction of particle tracks in ATLAS. That went well, and my supervisors pointed me towards CERN’s doctorate student programme. Six years ago, I was invited to join, moved to Geneva and was able to get started.’
Last autumn, Swatman obtained his PhD at the University of Amsterdam at the Institute of Computer Science. His thesis, Charged Particle Track Reconstruction Algorithms for Massive Parallel Systems, is one of the eight winners of the annual ATLAS thesis prize. The prizes will be awarded to the winners on 19 February.
Swatman’s research revolves around the software needed to derive traces of particles from collisions in the LHC accelerator from the signals in detectors such as ATLAS. ‘Simply put, you have a set of points, and the question is how to efficiently draw a line through them,’ says Swatman.
The challenge in the ATLAS experiment: millions of collisions per second, which generate a multitude of signals that need to be sorted and processed as quickly as possible. In his thesis, Swatman investigates whether a new type of processor could be useful for this, one that can perform many calculations simultaneously instead of one by one in ordinary CPUs.
Swatman: ‘Such GPU processors were originally developed for the gaming world, and are designed to calculate the colour and brightness of millions of pixels on a screen at the same time for the next moment. The question I am addressing is whether something like this can be used for reconstructing particle tracks in the ATLAS detector.’
At CERN, the LHCb experiment had already made the switch from regular to graphics processors and special software in the first step of data reconstruction, known as the trigger. Nikhef computer scientist Roel Aaij, once one of Swatman’s supervisors, was closely involved in that switch.
In his thesis, Swatman shows that massively parallel computing is indeed a viable option for track reconstructions at ATLAS. The technique is faster than using regular CPU processors, leads to more data that can be used for analysis, and all this at a lower cost and more energy efficient.
Swatman is now a postdoc with the ATLAS group and is helping with the analyses needed to decide whether and where GPUs will be used in the detector in the future. This question is on the agenda now that the LHC accelerator will be working with even more intense beams within a few years and there will be many more signals coming from the detector.
Swatman: ‘It is still too early to say what the outcome of the discussions will be. Whether parallel computing is the best choice or not depends very much on the exact requirements and criteria. We will see. The important thing now is to get the arguments as sharp as possible. We are working hard on that.’
VIDEO (in Dutch) Watch Stephen Swatman explain his thesis research here.
