Speaker
Zbigniew Szadkowski
(University of Lodz)
Description
Neutrinos play a fundamental role in the understanding of the
origin of ultra-high-energy cosmic rays. They interact through
charged and neutral currents in the
atmosphere generating extensive air showers. However, the
very low rate of events potentially generated by neutrinos
is a significant challenge for detection techniques and requires
both sophisticated algorithms and high-resolution hardware.
A trigger based on an artificial neural network was implemented into the
Cyclone V E FPGA 5CEFA9F31I7, the heart
of prototype Front-End boards developed for testing new algorithms
in the Pierre Auger surface detectors.
Showers for muon and tau neutrino initiating particles at
various altitudes, angles and energies were simulated in CORSIKA and
OffLine platforms, giving a pattern
of ADC-traces in Auger water Cherenkov detectors. Several three-layer
neural networks were taught in MATLAB by simulated ADC traces for various energies,
angles, initialization points and types of primary particles
according the Levenberg-Marquardt algorithm.
Results show that the probability of an ADC traces generation is very low
due to a small neutrino cross-section.
Nevertheless, ADC traces for low-energy showers, if they occur, are relatively
short and can be analyzed by a
16-point input algorithm. We optimized the coefficients from MATLAB
to get a maximal range of potentially registered events, and for fixed-point
FPGA processing, to minimize calculation errors.
We noticed that there is not so significant discrepancy in a detection efficiency
for various networks, starting from 12-8-1 and going to much more complicated
and on the other side going to much simpler. It looks like a relatively
simple network could select on-line preliminary events for a future more detail
off-line analysis.
Offline simulation were performed mostly for 40 MHz as a basic pattern. 120 MHz Offline version
is in a development phase and these data should be treated as preliminary.
New sophisticated triggers implemented in
CycloneV E FPGAs with a large amount
of DSP blocks, and embedded memory running with 120 MHz sampling, may
support a discovery of neutrino events in the Pierre Auger Observatory.
We investigated also the pattern recognition of DCT coefficients generated from
ADC-traces in a trigger region. Unfortunately we did not observe a significant
improvement which would justify much sophisticated FPGA network structure.
Author
Zbigniew Szadkowski
(University of Lodz)
Co-authors
Dr
Dariusz Głas
(University of Łódź)
Dr
Krzysztof Pytel
(University of Łódź)
