Skip to content

Changes to energy-per-layer calculations for PFlow and UFO jets

Chris Malena Delitzsch requested to merge cdelitzs/athena:master-EMFrac into master

Updating the description to match the latest status and take into account the feedback that was received on this MR and in meetings.

Thanks to the trigger experts, an issue was discovered for EMFrac (and in general for the energy-per-layer calculations) for EMPFlow jets. The changes were introduced by !56971 (merged) when including the energy-per-layer information for UFO jets. UFO and PFlow jets use the same code to calculate the variables. The impact of the change in the aforementioned MR on EMFrac for PFlow jets is shown in the below figure:

EMFrac

After discussing this with the jet experts, e.g. in the Jet Definitions meeting (agenda) it was decided that we restore the behaviour for PFlow jets as the recommendations were derived with the blue distribution. This means that we also restore the behaviour in R22.

The difference between the red and the blue distributions is that for the blue distribution, the jet constituents (i.e. the FEs) are used directly to calculate the total energy and to retrieve the energy-per-layer. In the red distribution, the underlying cluster corresponding to the FE is used, i.e. the energies are extracted before the energy subtraction!

In addition, new variables (corresponding to the red distribution) will be calculated at the derivation level that are expected to have superior performance compared to those used for the blue distribution. This has to be confirmed. This approach was not used in R21 as the links to the clusters were broken and therefore the underlying cluster could not be retrieved. The new variables will be used to derive new recommendations using small-R UFO jets. These new variables will be only calculated for the moment in JETMX derivations until it's clear that they should be used for future recommendations.

This MR violates frozen Tier-0 policy as (1) EMFrac, HECFrac, PSFrac, FracSamplingMax, FracSamplingMaxIndex and EnergyPerSampling change significantly (all triggered by changes to EnergyPerSampling) and (2) the trigger count will change because the change in the energy-per-layer calculation affects the GSC calibration of jets and therefore the events passing jet trigger requirements. The reference files will be updated accordingly once the CI ran.

Summarizing here quickly the logic for the restored EMFrac:

PFlow jets (constituents are charged and neutral FEs):

  • The energy of charged FEs is added to the total energy
  • The energy of neutral FEs is added to the total energy and the energy-per-layeris extracted only for neutral FEs

For the new variables (with CaloBased in the name), for each FE, the underlying cluster is extracted and its energy is added to the total energy and its energy-per-layer used to calculate variables such as EMFrac. Double-counting is avoided by checking if the cluster has been used already in the calculation.

Tagging some experts: @mhodgkin @mswiatlo @schaarsc @mleblanc @camacho @jroloff @mspalla @haweber

Tagging also @calpigia for his information

Edited by Chris Malena Delitzsch

Merge request reports