The physics objects that are used in ATLAS, those that ultimately result from the collisions, are reconstructed at varying levels of quality. There are software tools that determine, after event collection, the quality of the object. There are many effects that influence the quality of an object, including the trustworthiness of the object’s identification and whether any detector subsystems experienced any issues that affected their performance during object detection.
For a data analysis, it is important to skip events with objects of quality that is too poor to trust.
A good example of objects that have a quality assessment is hadronic jets
(we will learn more about jets later in the tutorial). Jet reconstruction uses
considerable information from the calorimeter, which means that “bad” jets
might originate from noise in the calorimeter, from beam backgrounds (e.g.
the beams scraping against collimators upstream in the LHC), or from cosmic
rays, for example. Like electrons and muons, the trustworthiness of an identified
jet is noted by its “working point”, which spans from loose
to medium
to tight
, each
indicating progressively stricter requirements. While for leptons these
working points usually indicate our confidence that the candidate object
corresponds to a true lepton, as opposed to a hadron or some other particle,
for jets the working point usually indicates our confidence that the jet
is really from a proton-proton collision. You can read more about
jet cleaning here.
Many identification algorithms operate on a limited set of possibilities: is this reconstructed electron a true electron, or is it a pion, for example. For some analyses that involve unusual detector signatures (like long-lived particles), special care must be taken to ensure these algorithms are treating the signal as expected.
The default event cleaning that the CP Algorithms handles is the removal of
events that do not contain a primary vertex or have jets that fail loose
quality cuts.
To include this event cleaning in our job, add the following lines to the
config.yaml
file:
# Apply common event quality requirements
EventCleaning:
# prune events without a primary vertex,
# prune events with jets failing loose cuts
# etc.
runEventCleaning: True
Try running again and look for changes in your output. In particular, check
how many events are accepted by the VertexSelection
and JetCleaning
filters.
Once you are satisfied that your code is working correctly, save your progress by committing and pushing your code changes.