atlas-ის დეტექტორის გეომეტრიული აღწერის დამუშავება სიმულაციისა და რეკონსტრუქციის პროგრამული პაკეტებისთვის

PhD thesis of Niko TSUTSKIRIDZE entitled “Development of Geometry Descriptions of ATLAS Detector for Simulation and Reconstruction Software Packages” concern with the investigation of Data vs Monte-Carlo discrepancies caused by geometry descriptions of ATLAS detector. This research and results will have impact on Physics results published by the ATLAS collaboration while contribute to a reduction of systematic uncertainties. The ATLAS simulation is implemented for a deep and wide range investigation of physics processes from the event generator in a format that is identical to the output of the ATLAS detector data acquisition system. Both the simulated data and detector outputs are running through the same trigger and reconstruction packages. However data analyses shows discrepancies of simulated and real data for some regions of the detector. In several cases they are caused by inaccuracies of the detector geometry descriptions used in the simulation. It is possible to formulate 2 hypotheses about the origin of geometry inaccuracies: 1 – inaccuracy added by geometry transactions within the simulation software infrastructure and 2 – inaccuracies added by inadequate representation of geometry models compared with as-built geometry descriptions. The main research goal of the dissertation is the investigation and verification of above mentioned hypotheses. For that purpose existing ATLAS simulation workflow was modified be adding application which is capable to load geometry descriptions used inside the ATLAS simulation and descriptions designed for construction. Thus, it becomes possible for the first time to perform direct comparison of several descriptions from various platforms and different methods of programing. Verification of first hypotheses requires formation test examples for simulation test runs. ATLAS detector geometry was categorized and 3 classes of geometry objects were separated – geometry primitives, combined geometry objects and objects with typical geometry joining. Therefore 84 geometry objects were selected. Then, for given set of objects possible methods of programming were described and full set of simulation test cases were worked out with 15’675 examples. Then special selection criteria’s were developed and final 77 simulation test examples were selected from the full set of examples. On the next stage of research all 77 examples were running into ATLAS simulation test runs. Received results categorized and examples with failures grouped according to systematic investigation of nature of reasons caused it. As a result, it was – approved #1 hypotheses; found Boolean operation as a main contributor of inaccuracies; found cases of geometry descriptions programming where Boolean operations are correlated with Move/Rotation geometry transactions. For investigation of #2 hypotheses comparative analyses of simulation and “as-built” geometry descriptions of ATLAS detector Coils, were done. As-built 3D models were reproduced from 225 blue print drawings in CATIA. Also, the theoretical 3D models were exported from Geant-4 simulation platform to CATIA. The Coil assembly was disassembled into 21 sub assemblies and comparative analysis was done for each of them. Finally 11 tonnes of missing material were discovered in the theoretical descriptions and hypotheses #02 has been approved.