Extracting bb Higgs Decay Signals using Multivariate Techniques

Smith, W Clarke ; /George Washington U. /SLAC

关键词: ALGORITHMS;    DECAY;    HIGGS BOSONS;    MASS RESOLUTION;    MONTE CARLO METHOD;    MULTIVARIATE ANALYSIS;    PRODUCTION;    STANDARD MODEL Experiment-HEP;    OTHER;   

DOI  :  10.2172/1049738
RP-ID  :  SLAC-TN-12-015
PID  :  OSTI ID: 1049738
Others  :  TRN: US1204576

学科分类：物理（综合）

美国|英语

来源: SciTech Connect

【 摘 要 】

For low-mass Higgs boson production at ATLAS at {radical}s = 7 TeV, the hard subprocess gg {yields} h{sup 0} {yields} b{bar b} dominates but is in turn drowned out by background. We seek to exploit the intrinsic few-MeV mass width of the Higgs boson to observe it above the background in b{bar b}-dijet mass plots. The mass resolution of existing mass-reconstruction algorithms is insufficient for this purpose due to jet combinatorics, that is, the algorithms cannot identify every jet that results from b{bar b} Higgs decay. We combine these algorithms using the neural net (NN) and boosted regression tree (BDT) multivariate methods in attempt to improve the mass resolution. Events involving gg {yields} h{sup 0} {yields} b{bar b} are generated using Monte Carlo methods with Pythia and then the Toolkit for Multivariate Analysis (TMVA) is used to train and test NNs and BDTs. For a 120 GeV Standard Model Higgs boson, the m{sub h{sup 0}}-reconstruction width is reduced from 8.6 to 6.5 GeV. Most importantly, however, the methods used here allow for more advanced m{sub h{sup 0}}-reconstructions to be created in the future using multivariate methods.