Searching for supersymmetry in the first LHC data with ATLAS

In recent years cosmological measurements have shown indications that the particles of the Standard Model (SM) can only account for 4% of the total energy content of the universe. The two other main contributions have been dubbed Dark Energy (74%) and Dark Matter (DM, 22 %). It is fascinating that the theory that is unsurpassed in precision of its predictions by any physical theory turns out to be so limited in scope. Now the remaining 96% of the energy content of the universe can be examined, and a working theory for its makeup can be developed. One theory predicting a particle that could be the main constituent of DM is Supersym- metry (SUSY). Although no evidence of this theory has been found so far, there is still a large region of the vast SUSY phase space that is unexplored. The ATLAS detector, one of four experiments on CERN's LHC, is well equipped to do just that. With the data delivered by the LHC during 2010, when two beams of protons were collided at a center of mass energy of 7 TeV, ATLAS can explore a new part of the SUSY phase space, and search for a good candidate DM particle. Finding SUSY in the ATLAS data is challenging for several reasons. First of all, the precise signature of SUSY in the pp collisions is unknown. Several models exist, each with their own properties and `smoking guns' that might be present in the ATLAS data. The signature of SUSY that was used in this analysis assumes that whatever supersummetric particle is created by the pp collision, it quickly decays to the lightest supersymetric particle, the LSP, which is assumed to be stable. Such a stable LSP would provide an excellent DM candidate. Since this LSP is subject to the weak interaction only, it will be seen in the detector as missing momentum transverse to the beamline. The second challenge is to use the missing p T in a collision as a tool for nding new physics. In order to do this the detector and the objects directly measured and reconstructed (jets, leptons) must be well understood. This is an experimental challenge that was tackled with good results by a large part of the ATLAS collaboration. The third challenge is distinguishing the SUSY signal from known SM processes. Before 2010, SM processes were not examined at such high energies, and uncertainties in the theory of the SM (e.g. uncertainties on the proton structure, and on nite order cross section calcu- lation) mean that the behavior of the SM at these high energies is not yet well understood. This thesis describes a data-driven method through which the SM behavior is evaluated while simultaneously searching for a SUSY signal. For the two main backgrounds to SUSY, t t and W +jets, a model is made of the distribution of missing transverse energy E miss T and the trans- verse mass M T . The model was built from probability density functions with parameterized shape, which reduced the model dependance on a specic Monte Carlo simulations. Taking advantage of the fact that a key signature of mSUGRA models combines high E miss T with high M T , the shape parameters are estimated in a t to a dataset of events with either low E miss T or low M T . With the mass of the three jets with the highest P ~p T ( M jjj ), a third distribution is added to the model which helps to distinguish the semileptonicaly decaying t t background. When the shape of the model has been estimated in a maximum likelihood t, the model can be extrapolated to a region with a high signal yield (thus at high E miss T and M T ), giving an estimate of the number of expected background events in this signal region. Here the shape of the control region, embracing the signal region from two sides with one arm extending to high E miss T and one to high M T gives a great benet to the method in enabling an accurate extrapolation. Although the method works best when the shape of each distribution is inde- pendent of the other two observables, in this thesis it will be demonstrated that the correlations between the observables can be accounted for by the model, and dealt with in the tting and extrapolation procedure. As the model is normalized to the number of events in the control region, contamination of SUSY events in this region will result in a higher estimate of the number of background events in the signal region, and thus in a reduced sensitivity to new physics. Unfortunately, once the threshold for producing SUSY events is reached, SUSY events will be present also in the low E miss T and low M T regions. This thesis will show that it is possible, at least for mSUGRA events, to create a simple Ansatz model for signal in the control region, which can in most cases accurately estimate the level of signal contamination. The main assumption that is made in creating the model is that the mass scale of most mSUGRA models to which the ATLAS detector is sensitive is much higher than for the SM background. The SUSY Ansatz model is not used in the extrapolation. This thesis presents two analyses that were performed using the so-called combined t method. One analysis is a proof of principle on a relatively large dataset containing simulated events for the SM background, combined with simulated signal events for a wide variety of possible mSUGRA models. A grid was dened in the phase space of mSUGRA parameters, which was used to explore the dierent behaviors of mSUGRA models. All these samples were created assuming a pp center of mass energy of 10 TeV. After developing and validating the dierent tools and aspects of the combined t method, it was applied on a somewhat smaller (in terms of integrated luminosity) dataset consisting of events measured in the 7 TeV pp collisions recorded by the ATLAS detector during the 2010 data taking period. This resulted in a limit on the number of non-SM events in a dataset with high E miss T and M T , that already exceeds the limit set by previous experiments. At the time of writing this abstract the 2011 dataset recorded by ATLAS is already 20 times larger than the one used in the nal chapter of this thesis. In the analysis of this new dataset the combined t method can be a valuable tool to nd evidence of supersymmetry or other new physics in the ATLAS data, or in the absence of an excess, extend the limit on new physics even further.