'Artificial retina' could detect sub-atomic particles
- Published
The human eye has inspired physicists to create a processor that can analyse sub-atomic particle collisions 400 times faster than currently possible.
In these collisions, protons - ordinary matter - are smashed together at close to light speeds.
These powerful smash-ups could yield new particles and help scientists understand matter's mirror, antimatter.
The experimental processor could speed up the analysis of data from the collisions.
Published in the pre-print arXiv server, the algorithm has been proposed for possible use in Large Hadron Collider (LHC) experiments at Cern in 2020. It could also be useful in any field where fast, efficient pattern recognition capabilities are needed.
The processor works in a similar way to the retina's incredible ability to recognise patterns extremely quickly.
Snapshots in time
That is, individual neurons in our retinas are specialised to respond to particular shapes or orientations, which they do automatically before our brain is even consciously aware of what we are processing.
Cern physicist Diego Tonelli, one of a team of collaborators of the work, explained that the "artificial retina" detects a snapshot of the trajectory of each collision which is then immediately analysed.
These snapshots are then mapped into an algorithm that can run on a computer, automatically scanning and analysing the charged particle trajectories, or tracks. Exposing the detector to future collisions will then allow teams sift out the interesting events.
Data crunching
Speed is of the essence here. There are roughly 40 million collisions per second and each can result in hundreds of charged particles.
The scientists then have to plough through an incredible amount of data. It's spotting the deviations from the norm that may give hints of new physics.
An algorithm like this could therefore provide a useful way of crunching through this vast amount of data, in real time.
"It's 400 times faster than anything existing or foreseen for high energy physics applications. If implemented in a real experiment it will allow us to collect more interesting data more quickly," Dr Tonelli told the BBC.
Flavour physics
The LHC has been switched off since February 2013 but is due to begin its hunt for new physics in 2015 when the giant machine will once again begin smashing together protons.
As this happens, they break down and free up a huge amounts of energy that forms many neutral and charged particles. It's the trajectories of the charged ones that can be observed.
The new algorithm is not aimed at the type of physics used to find the famous Higgs boson, instead it's intended to be used for "flavour physics" which deals with the interaction of the basic components of matter, the quarks.
Commenting on the work, Tara Shears a Cern particle physicist from the University of Liverpool, said it could be extremely useful to automatically "give us most information about what we want to study - Higgs, dark matter, antimatter and so on. The artificial retina algorithm looks like it does this brilliantly".
"When our detectors take these snapshots of the collisions - to us that's like the picture that your eye sees and when your brain is scanning that picture and making sense of it, well we try and codify those rules into an algorithm that we run on computers that do the job for us automatically," Prof Shears told the BBC's Inside Science programme.
"When the LHC continues... we will start to operate with a more intense beam of protons getting a much higher data rate, and then this problem of sifting out what you really want to study becomes really really pressing," she added.
"This artificial retinal algorithm is one of the latest steps in our mission to [understand the Universe], and it's really good, it does the job vast banks of computers normally do."
The algorithm has been developed with the 2020 upgrade of the LHC in mind, which will have even more powerful collisions.
Follow Melissa on Twitter
- Published18 February 2014
- Published24 July 2013
- Published2 April 2013
- Published28 February 2013
- Published29 February 2012
- Published15 November 2011