Parts | Days | Selection | Search | Updates | Downloads | Help

HK: Fachverband Physik der Hadronen und Kerne

HK 39: Computing II

HK 39.3: Talk

Wednesday, March 30, 2022, 14:30–14:45, HK-H5

Using Neural Network regression to describe the expected energy loss in the ALICE TPC in Run3 — •Christian Sonnabend for the ALICE collaboration — Physikalisches Institut, Universität Heidelberg

The ALICE experiment at CERN uses the largest Time Projection Chamber (TPC) built to date to identify particles that are created in collisions at the LHC. Particle identification is done by simultaneous measurement of the specific energy loss (dE/dx) and momentum (p) of the traversing particles, and comparison to the expected energy loss described by a Bethe-Bloch function. However, in practice, the expected dE/dx cannot be described by a simple one-dimensional function, but several effects have to be taken into account. E.g. the inclination angle of a particle track has an effect on the charge deposited in a given region of the TPC readout, thus changing its dE/dx signal (η-correction (pseudorapidity)). In order to correct for such effects, fits to a multidimensional parameter space consisting of e.g. p, η, multiplicity or particle mass are performed to adjust the expected dE/dx signals of the tracks.

With the application of Machine Learning in particle physics, new methods can be exploited to extract such functional forms. Thus, a variety of neural network fits to data are conducted to investigate their performance and compare their ability to describe deviations of the expected energy loss from an input Bethe-Bloch parametrisation in a multi-dimensional space.