by Giles Strong

Guten Tag! Last week I took a short break from my secondment in Padova (which is going well, expect a post on it sometime) to attend a CMS data-analysis school hosted by The Deutsches Elektronen-Synchrotron (DESY) in Hamburg.

I’d gone with the hope of getting a proper and hands-on introduction to CMSSW, the centrally-managed software used by the CMS collaboration to access and analyse their data from the LHC. I got this, and a lot more!

The school began with registration and a day of presentations on various topics, covering: the current state of high-energy physics, the possibilities for future colliders and experiments, and the CMS experiment. Whilst these were interesting, I feel in retrospect that the time would have been better spent on the exercise tasks. They did, however, provide a nice motivation for our continued involvement in CMS and the field of HEP research.

The ‘school’ part of the school consisted of the aforementioned exercise tasks, which were divided into short and long exercises. Each of us took five short exercises, which ranged in length between two and three hours and aimed to introduce us to specific uses of CMSSW, such as b-tagging, electron and photon selection, and event displays. My selection consisted of: evaluating missing transverse-energy, statistics, muon selection, hadronic-tau selection, and jet selection. These were quite intense sessions, and I certainly have a lot of unfinished exercises to review if needs be.

wp_20160921_19_24_04_pro
The school meal was held in a lovely brewery in the old part of Hamburg

On day three we began our ‘long exercise’, which was designed to take a specific analysis, build it up from scratch, and follow it through to the very end, concluding with a presentation to the rest of the school on the last day.

My group was involved in a really interesting search channel, where two Standard Model Higgs bosons are produced by the decay of some BSM resonance, here a Heavy Higgs. One SM Higgs decays to a pair of b quarks, and the other to a pair of tau (τ) leptons. One τ decays hadronically, and the other to a muon (μ), a τ-neutrino, and a μ-neutrino. Now, where have I seen this before?

Ok, so discarding the BSM resonant-production mechanism, it’s the same process I’ve been looking into for the past six months; so far just with Monte Carlo generation and fast detector-simulation. Looking at full detector-simulation and collider data was going to be the next step.

For the first day we got to grips with the three main module-classes in CMSSW: producers, filters, and analyzers, and developed code to produce skimmed datasets for our signal and background MC data, and the ICHEP 2016 dataset. However, to save time, the organisers had already produced the skimmed samples.

On the second day, we began analysing the selected events and accounting for MC-collider mis-modelling, by reweighing events to account for differences in object-selection efficiency and pile-up simulation. We also applied a kinematic fit, which aimed to reconstruct the mass of the Heavy Higgs with better precision than a simple four-vector sum of the visible final-states and the missing energy (due to the neutrinos). It is done by divvying up the missing energy between the final-states under the hypothesis of H→hh→bbττ. Indeed, precision was much improved.

With some lovely, reweighted plots produced, we split into two groups for the afternoon session; one to look into inferring the QCD background contribution using a data-driven method, and the other to optimise cuts on the event selection to improve the ratio of signal-to-background in the fitted-mass plot for the Heavy Higgs.

20160923151941
Down in the old HERA tunnel.

Having convened, the last task was to feed the fitted-mass distributions for all processes into a tool called Higgs Combine, in order to extract limits on the production cross-section for a Heavy Higgs.

The morning of the third day was spent preparing a presentation to give to the other groups, summarising the various steps and reporting our results. Sadly I can’t report any figures or plots here due to them being unapproved and me not yet being a CMS author, but if you’re interested, we pretty much followed the analysis of this paper and got similar results.

The afternoon consisted of a tour of the DESY campus, to visit the various experiments and go underground to see the old HERA accelerator, and a talk on the perspectives for Higgs and new physics by Abdelhak Djouadi, a theorist who had made major contributions to the 2012 discovery of the Higgs boson.

Overall I found the school to be extremely useful, especially since I left having built up a basic frame-work with which I can eventually move my analysis on to use more accurate simulations and access collider data. My thanks to the organisers and guest speakers.

group-photo
Group photo at the end of the school.