One of the best parts of being a physics PhD student is having the chance to broaden your knowledge by attending seminars and schools especially designed for helping you to be more efficient in your research. I was fortunate to have such an opportunity by attending the first CMS Physics Object School (POS) which took place from September 4th to 8th in Bari, Italy.
The school was organized by the Physics Department of the University of Bari and INFN and its aim was to help new CMS members to become familiar with the reconstructed physics objects and the detector’s performance, so that they can contribute effectively to the Physics Object Groups (POG) and the Data Performance Groups (DPG) of CMS.
Before attending the school we had to complete a set of pre-exercises, in order to have everything set-up and be ready to participate in the short and long exercises which would take place during the school. Personally, I found the set of the pre-exercises very helpful, as I got introduced to the CMS software (CMSSW), I set up my CRAB certificate and submitted my first jobs on CRAB. I also had the chance to learn about the Data aggregation system (DAS) and the Event Data Model (EDM) framework utilities.
The school’s schedule was an intense one-week course that consisted of several talks on the first day, two short exercises on different physics objects and one long exercise which would be presented in the context of a competition between the groups on the final day of the school.
The talks of the first day were focused on the motivation of the organizers to deliver such an event and the importance of training the new CMS members so that they can contribute to the physics analyses within the collaboration. Adbelhak Djouadi gave a theoretical talk on the motivation of the New Physics searches via searches of the H/W/Z/top sector and also through new particle searches with more data.
The rest of the first morning’s talks were focused on more technical topics, such as the talk of Lucia Silvestris on running the CMS detector, the talk of Gurpreet Singh Chanal on the physics performance and datasets and Marcellos Abbrescia’s talk on the CMS upgrade.
After lunch the short-exercise session started. The exercises were focused on the data performance of the different detector parts (DT, ECAL, HCAL, RPC, tracker) and there was also a group that would work on the physics performance and dataset. I decided to follow this last exercise, facilitated by Gurpreet Singh Chanal. In the beginning of the session there was a short tutorial on the need of the MC production and how to use DAS and McM to find details of the samples.
During the exercise time, we used a miniAOD file and learned how to look it up on DAS, find all the details (global tags, CMSSW release, etc.) and then look for it on McM. We further found the LHE, GEN-SIM and DIGI-Reco used to produced the miniAOD. We looked for the details on the x-section cuts on the generator level and the global tags in GEN-SIM and after we got familiar with using the McM we were asked to use the cmsdrivers and produce the LHE, GEN-SIM and DIGI-RECO requests that were used to produce a specific miniAOD file.
As for my second short exercise, I decided to follow the group of Drift Tubes (DT)/Cathode Strip Chambers (CSC) timing measurement, facilitated by Piotr Traczyk. The muon time measurement is important in order to determine the muon bunch crossing. The exercise was organized in three parts. In the first one we had to read the DT and CSC segments from the Event and write an ntuple containing the info of timing and the number of hits, while in the second part of the exercise we matched the segments to muons using the MuonSegmentMatcher method by looping over muons of the event and looking at segments attached to them. In the last part of the exercise we played around with the produced ntuples by applying ID or kinematic cuts and studying the behavior of the timing under these restrictions.
It was already Tuesday evening and it was time to move on to the long exercise. I chose to follow the muon exercise as its content was fully aligned with my work both on the SUSY analysis, but also on my service work on the L1 muon trigger. The muon exercise was facilitated by Daniele Trocino who had prepared a very detailed and well organized two-and-a-half-days program of tutorials and exercises.
In the beginning of the session we had an introduction on the muon object and its reconstruction in the CMS detector. As a first exercise we used the pat::Muon (reco) objects and extracted its main kinematic variables. We further matched the reconstructed to generator level and trigger objects by imposing geometrical matching criteria.
In the second part of the exercise we worked on muon identification by exploiting the main properties and quality variables of the muon tracks, such as the type of muon (global, tracker etc.), the number of hits in the muon chambers, the impact parameter, and others. These variables exhibit different behavior for muons coming from different sources (prompt, heavy flavor decay, light flavor decay, fakes) and thus we can use the distribution of these variables to distinguish between the different muon sources.
In the next step of the exercise we worked on the muon momentum scale. Distortions in the measurements of the muon momentum can be produced by an imperfect knowledge of the position and alignment of the tracker layers and the muon chamber, or the poor knowledge of the components and distribution of the material and the magnetic field configuration in the detector. We can thus use muon momentum calibration algorithms like Rochester, which use constraints on the mass of well known resonances (Z, J/ψ) in order to correct for the muon momentum.
Finally, the last step of the muon long exercise was focused on efficiency calculation using the Tag-And-Probe method, which utilizes well known resonances like Z that decays into two muons. We computed the MC-truth efficiencies by using a generator level muon in the acceptance region (pt > 20GeV, |η| < 2.4), matching reconstructed to generator level muons, and filling histograms with generator level particles at different selection stages.
The efficiencies were calculated from the ratio of the initial generator level muons when no selection cuts were imposed and the generator level particles after the selection cuts. We further calculated the efficiency using the Tag-And-Probe method on the reco muons and imposing kinematic cuts on the pt, η, isolation and single muon trigger cuts to select the tag muon. The probe was selected so that the tag and the probe muons are within a narrow window around the Z mass and are of opposite sign. If the probe was identified as a full muon it was set to be a passing muon and the efficiency could be calculated by dividing the number of probes by the number of passing probes.
After the end of the long exercise, all the physics object groups prepared a 15 minute talk in which they presented their work. Even though all the groups did an excellent job, the tau group won the competition and received the prize of a good luck charm from the region of Puglia. However, we all toasted and celebrated after the end of the presentations.
The Physics Object School in Bari was an amazing experience in every sense. We followed talks and tutorials fully related to our work, but we also got help on solving problems and using the CMS software for our analysis from the experts. Moreover, it was a great chance to meet other young people, also new to CMS, with whom we share the same interests but also the same struggles.
I think that it was a great opportunity for networking, improving computing skills, making some new friends, but also enjoying the good weather and amazing food of the Italian south. I fully recommend to all the CMS newcomers to attend the next CMS Physics Object School.
Until next time, arrivederci 🙂