The Trigger Studies Group (TSG) had its first workshop of 2014 in LAPP Annecy, on Mon-Wed Apr 7-9. TSG workshops are intense events, attended by around 40 people, where questions to the speakers during the talks are encouraged and unlimited, and 20 slides reports often last more than one hour.

A very important report was presented by the L1 conveners, with a first study of a possible menu for one of our ambitious targets: luminosity of 1.4E34cm-2s-1 with 25ns bunch spacing, and average pileup of 40 interactions. In this scenario, both the luminosity and the cross section are a factor two higher than in 2012, resulting in a major challenge for our trigger. Studies are still preliminary and do not cover all the phase-space of the L1 triggers, as some disagreements between data and simulation have still to be resolved. But some trend is already clear: the planned “Phase 1” calorimeter trigger is essential and effective to keep the jet trigger rates under control using pile-up subtraction. At the same time, it will provide improved isolation for the electrons, which will be instrumental to keep the single and double-electron thresholds low enough not to spoil the electro-week signals. The tools to make more detailed studies of the individual seeds, and to devise new cross-seeds if needed, are being handed over to the PAGs; in the meantime, the L1 team will also consider lower luminosity scenarios that are foreseen at the start-up.
Moving forward to the HLT, the plan of TSG is to have the core of the object reconstruction improvements, as developed by the Physics Object Groups (POGs) in the last year, integrated in the CMSSW 7.1.0 release before the end of April. The goals are: to bring on-line reconstruction closer to the off-line one for better performance and maintainability, to reduce the CPU time at high Pileup and to converge on common algorithms that are shared by different paths. So a focus of the workshop was the review of the work of the POGs and their plans for integration.

• The E/gamma group has been successful in bringing offline reconstruction to the online, and recently has converged on an isolation algorithm that is based on Particle Flow clusters and regional iterative tracking limited to iterations 0,1,2. This provides optimal performance as a function of pileup, and at the same time improves on the CPU time requirement. Further improvements may be gained using a light version of the ECAL “rechit” timing reconstruction, to remove Out Of Time (OOT) pileup at the source.

• Similar work and progress was reported by the muon group: muon reconstruction has improved significantly, curing the inefficiency at high pileup that was observed in 2012. The approach to isolation is now similar to that of electrons and photons, with iterative tracking limited to iterations 0-2.

• The tracking POG made an extensive review of the algorithms, and will provide three different flavours of track reconstruction, for evaluation by all physics groups. The first is essentially the “2012 tracking”, with some technical improvements, and will be used as reference for studies. The second has a primary vertex constraint also in iteration 0, higher regional pT cuts in iteration 2 and will use only iterations 0-2 for particle-flow reconstruction and isolation. It is the fastest of the three, with the CPU time more than halved at high PU. The third is similar to the second, but with the first tracking iteration closer to its off-line version, providing a better efficiency at high eta at the expense of some of the gained CPU time.

• The JetMET group is focusing on the jet energy corrections and pileup subtraction. Good progress has been achieved at 50 ns bunch spacing, while the approach at 25 ns will depend on the upcoming HCAL reconstruction tailored to reduce the OOT contribution. Also the particle flow MET has been improved using noise filters. A common particle flow approach to object reconstruction is now the baseline at the HLT.

• The B-tagging and Taus POGs have more specific tasks, and are making sure that the choices made in tracking and vertexing are tuned adequately for their purposes.
Once the reconstruction of the various physics objects is integrated in CMSSW, all the physics groups will be requested to provide a final feedback on the performance. A key choice is the flavour of tracking that will be brought forward: it has to be the best compromise between CPU time, efficiency for the key prompt signals and quality of the tracks handed to particle flow reconstruction.

The second half of the workshop was dedicated to reports by the Physics Analysis Groups (PAGs), and was quite uniformly devoted to the L1 triggers needed to seed their HLT paths. The L1 trigger is clearly seen as the challenge to address, and the TSG will help coordinate the effort in these studies. The goal is to have a first version of a menu (both L1 and HLT) ready for the summer, in order to evaluate it in term of rates, bandwidth assignment and CPU time at HLT.
Many other tasks and challenges were not addressed at the workshop due to lack of time: we decided to leave out all technical aspects like e.g. the integration with the new DAQ2, the overhauling of the ConfDB and its GUIs, to focus on the menu developments. The next TSG workshop will take place on June 30-July 3 in Brussels, with the explicit goal of reviewing the first version of the trigger menu. A third workshop, aiming at a near-final version of the menu ready for the “production” version of CMSSW, will follow on October 27-30 in Strasbourg.

Many thanks are due to our Atlas colleagues who kindly hosted us in LAPP, and to our CMS secretaries that helped in the organization of the logistics. Above all, warm thanks to the several CMS members, in particular young students and post-docs, who produced the results and/or participated in the workshop, for their dedicated work and enthusiasm.

By Roberto Carlin