Speaker
Description
This contribution summarises studies carried out in the frame of the DRD Calo Collaboration and the French-German project Calo5D on optimizing particle flow calorimetry with machine learning and timing with a fine granularity in 4D. All studies are based the ILD Detector concept as example and use the ILC Software or Key4Hep framework. In one study a Dynamic Graph Convolutional Neural Network (DGCNN) is applied. The results on jet energy resolution for jets with jet energies between 40-500 GeV are compared with the results of the well-known particle flow algorithm Pandora. Further results of the DGCNN with an without the use of timing are compared. Particle physics experiments at $e^+ e^-$ colliders that would implement big particle flow calorimeters are expected for the middle or late 2040s. Given the long lead time for calorimeter construction conclusions on the benefit of precise timing have to be provided until the end of this decade. The current studies are major ingredients for this goal.
At a more fundamental level we are also applying machine learning to optimise the calorimetric response to hadrons. To our knowledge we are the first that seek to combine signals in a realistic calorimetric system consisting of the ILD SiW ECAL and the analogue hadron calorimeter (AHCAL). Both are sampling calorimeters. A dual branch Convolutional Neural Network (CNN) has been developed exploiting the rich information provided by the granular calorimeters. An energy resolution of around $35\%/\sqrt{E} \oplus 3\%$ (or even better) using a regression method seems to be in reach. These results are competitive with those obtained by dual readout technologies while at the same time maintaining the high granularity needed for a successful particle flow up to highest energies. At the time the abstract is written precise timing is implemented into the study and first results are expected at the conference.
We will also report on detector optimisation studies varying the configuration of SiW ECAL layers of the ILD Detector Concept, to reach a higher precision on the low-energy photons energy and position measurements, dominating jet’s population and critical to specific final states..