Deep Learning to Advance Higgs Boson Physics at the Large Hadron Collider

Mark Neubauer, University of Illinois at Urbana-Champaign

Usage Details

The quest to understand the fundamenta lbuilding blocks of nature and their interactions is one of the oldest and most ambitious of human scientific endeavors.  Facilities such as CERN’s Large Hadron Collider(LHC) represent a huge step forward in this quest. The LHC is the world’s most powerful particle accelerator and was designed to elucidate the nature of the fundamental building blocks of matter and the forces that govern their behavior by colliding beams of high-energy protons.  The discovery of the Higgs boson in 2012 (and Nobel Prize in 2013) and stringent constraints on many viable theories of physics beyond the Standard Model(SM) demonstrate the great scientific value of the LHC physics program. 

The University of Illinois is a key contributor to the ATLAS experiment, one of two general purpose LHC experiments. In prior work, we have succeeded in developing and validating our key applications on Blue Waters(BW).  We have integrated BW into our production processing environment and demonstrated our ability to run ATLAS production workflows on BW at large scale [1,2].  We have also developed and executed a number of machine learning applications using data from the ATLAS detector on the BW supercomputer, including development of novel methods to identify proton-proton collision events containing Higgs bosons decays.   The sophistication of the ATLAS detector and collision reconstruction software, combined with the complexity of the proton-proton collisions environment, place enormous demands on scientific computing resources.  This proposal will make use of the Blue Waters supercomputer to (1) utilize deep learning methods and software to identify and calibrate highly-boosted Higgs bosons at the LHC which would signal the existence of new physics at the energy frontier and (2) employ powerful and novel Matrix Element techniques for the analysis of LHC data to improve its sensitivity to new phenomena and therefore enhance its discovery potential.

The proposed activities make use of the unique capabilities of Blue Waters and advances in machine learning, container and work flow technologies, to substantially improve the sensitivity of searches for new phenomena a world’s largest collider.  They are also strengthened by leveraging three recently awarded NSF projects (IRIS-HEP, SCAILFIN, ASAML) and one recently awarded DOE project (FAIR4HEP) with which Neubauer (PI of this proposal) is involved.