TÉCNICAS E INSTRUMENTOS DE RECOGIDA DE INFORMACIÓN

To study the characteristics of signal and background events, Monte Carlo samples are gen-

erated usingPYTHIA 6.4.21 [27], a leading-order (LO) parton-shower MC generator, with the

modified leading order MRST2007 [29] parton distribution functions (PDFs). The event gen- erator parameters are set according to the ATLAS MC09 tune [30], and the ATLAS detector

response is simulated using the GEANT4 program [31]. These samples are then reconstructed

with the same algorithms used for data. More details on the event generation and simulation infrastructure of the ATLAS experiment are provided in [32]. For the study of systematic uncertainties related to the choice of the event generator and the parton shower model, alter-

native samples are also generated withHERWIG6.5 [33].

To study background processes, two classes of samples are simulated. In the first, non- diffractive minimum bias events are generated and filtered by requiring at least 6 GeV of

transverse energy in a 0.18×0.18 region in η×φ at the truth particle level, which mimics

a calorimetric Level-1 trigger requirement. The events passing this filter, whose efficiency is around 5.3% in inclusive dijet events, are then fully simulated. This filter is found to be unbiased for transverse energies above 10 GeV. The equivalent integrated luminosity of this

sample, according to the effective production cross section returned byPYTHIA(including the

filter efficiency) ofσ= 2.58 mb, is 16 nb−1_.20

TheETspectrum of reconstructed fake candidates decreases rapidly above the filter thresh-

old. To ensure sufficient statistics for background studies a second class of samples, enriched in candidates with higher transverse energies, is used to study fake photon candidates with

reconstructed ET > 20 GeV. In these samples, all relevant 2→2 QCD hard subprocesses

are switched on, the transverse momentum of the hard-scattering products is required to be

greater than 15 GeV, and the same filter used for the minimum bias sample is applied. The filter has a higher threshold than in the minimum bias sample, with cuts at 17 GeV and 35

GeV.21 _{The sample with a 17 GeV minimum is found to be unbiased for transverse ener-}

gies above 20 GeV. Its equivalent integrated luminosity, according to the effective production

cross sectionσ= 0.99 mb computed withPYTHIA(taking into account also the filter efficiency,

8.6%), is 494 nb−1_{. The sample with a parton-levelp}

Tcut at 33 GeV and a truth-particle-jet

filter cut at 35 GeV helps extend the reach to higher transverse energies, and corresponds to

an integrated luminosities of 579 nb−1.

All of these QCD background samples contain “fake” photon candidates (typically from

π0 _{andη decays), as well as prompt photon signals produced by QED radiation emitted off}

quarks. The higher energy samples also contain direct leading-order contributions, either

from q q¯ → g γ or q g → q γ. Reconstructed prompt photon candidates are matched to

particles in the truth record of the event with a dedicated tool, which uses a combination of

ancestry information from the generator and ∆Rmatching to determine whether a candidate

is signal or background. All prompt photon contributions are removed from these samples when studying the background contribution.

For signal-only samples, two types of filters are used. The first class of prompt photon

samples consists of simulated leading orderγ-jet events, and contains only direct photons with

generated transverse momenta above some threshold (7, 17, 35 and 70 GeV thresholds are used

in these studies).22 _{The equivalent luminosities of these samples range from 71 nb}−1_{(for the 7}

GeV threshold) to 1 fb−1_{(for the 70 GeV threshold). The box-diagram hard subprocessgg→}

gγ is part of the next-to-next-to-leading order (NNLO) cross section and gives a negligible

contribution to the total prompt photon cross section compared to the other two subprocesses.

21_{Again, within ATLAS, these are known as JF17 and JF35.} 22_{These are called “unbinned photon+jet samples”.}

It is not included in the generated samples.

The second class of signal samples contains both direct photons and photons from QED radiation off quarks. The events generated for this class of samples are similar to those gen- erated for the study of the QCD backgrounds, but the filter applied before the full simulation only retains events that contain reconstructed photons matched to signal photons in the truth

record, with generated transverse momenta above some threshold, either 7 GeV or 17 GeV.23

The equivalent luminosities for those two samples are 206 nb−1 and 4.6 pb−1, respectively.

These samples are used to study the contribution to the prompt photon signal of photons from fragmentation or from the NLO part of the direct process and that are less isolated than those from the LO direct processes. In such studies, all direct LO photon contributions are removed from these samples in the analysis. The separation of the direct and brem/fragmentation components as described above is defined by the generator, and has no physical significance beyond leading order. In the analysis the LO-direct and NLO/brem contributions are typi- cally used separately, without relying on the generator for implicit assumptions about their relative rates.

Finally, for the efficiency and purity studies involving electrons from W decays, a pure

sample of W → eν events is used. The sample consists only of events with a final state W

decaying leptonically to an electron and a neutrino.