GUÍA BÁSICA DE FUNCIONAMIENTO

In the SUSY model used for the present analysis, the mass meu of the up- (eur,ecr) and the mass med of the down-type (edr,esr) right handed squarks differ by approximately 10 GeV:

m_eu =med + 9.6 GeV (5.17)

As this difference is small compared to the total mass and the production cross section of the up type squarks is almost four times higher than the one for down type squarks, the present analysis is unable to distinguish between these two squarks families. Instead the template fit will result in an average of their masses, weighted with their respective cross sections: σeu σed = 3.82 (5.18) mq_e= σeumeu+σedmed σeu+σed (5.19) For the template generation a similar mass splitting between up- and down-type squarks of exactly 10 GeV was used. The templates were generated in steps of 3 GeV for the range of 1050 GeV to 1250 GeV, assuming that the approximate mass region of the

squarks is already known from another source like theM_C method or an independent experiment like the LHC:

meu =med + 10 GeV ={1050,1053, . . . ,1248}GeV. (5.20) To minimize influences on the final result through statistical fluctuations of the template distributions, each of these templates was generated with 50000 events which corresponds to an integrated luminosity of R dtL= 33.6 ab−1_{. As computation power}

was limited, these 3.3 million events could not go through the entire processing chain. The full detector simulation and reconstruction were omitted and replaced by including detector effects on generator level.

This included the rejection of particles in the very forward region _|cosθ_|>0.995 and low momentum particles p < 100 MeV to account for the detector acceptance region. As a second step jet clustering was performed using the same algorithm as for the fully simulated data earlier in this analysis (kt algorithm with R = 0.7). Detector

effects were included by smearing the energy of the resulting jets with a Gaussian. The width of this Gaussian is the assumed jet energy resolution σJet and was obtained by

comparing the fully simulatedMC distribution with a single template close to the actual squark masses using a Kolmogorov-Smirnoff test, after smearing the jet energies of the template with different resolutions of:

σJet ={0,1,2,3,3.5,4,4.5,5,7.5,10}%. (5.21)

These results are shown in Figure 5.14. For better readability only the extreme (no smearing, smearing of 10%) and the optimal cases are shown. The result closest to the full simulated data was the template with a smearing ofσJet = 4.5%. This is compatible

with the expected resolution of PandoraPFA of 3.5% to 4.0% RMS90 for TeV scale jets.

A few examples of the templates are shown in Figure 5.15.

The templates do not include any kind of background, as the remaining background contribution after the rejection with transverse momentum cut, Boosted Decision Trees and subtraction of the remaining background with a parametrized fit is expected to be negligible in the case of fully simulated data. In addition the smearing only included the jet energies, which is sufficient for the correct reproduction of the MC distribution with detector effects. But the smearing misses the influence on other observables like the number of particles or the correct particle ID. This information, however, is used by the Boosted Decision Trees for the event classification, and as a consequence the classification performance on the templates is worse than for full simulation.

Hence the selection based on the Boosted Decision Trees and the subtraction of the remaining, Gaussian shaped, background events as described in subsection 5.5.3 are not performed for the generation of the templates. However, as the cut on the transverse momentum ofpt,miss <600 GeV has a significant influence on the shape of

the MC distribution, it is included in the template generation.

The templates do not include overlayed events ofγγ _→hadrons from beam induced background, since with the usage of thekt algorithm as jet finder their influence is assumed to be negligible.

[GeV]

C

M

0 500 1000 1500 2000

Entries / 50 GeV

0 0.02 0.04 0.06 0.08 distribution C M Full simulation 0.0% smearing 4.5% smearing 10.0% smearing

Figure 5.14: The MC distribution obtained from generator level information without full detector simulation, but with smeared jet energy assuming a Gaussian jet energy resolution of 0, 4.5 and 10 per cent. For comparison a fully simulated sample is shown in black.

[GeV]

C

M

500 1000 1500

Entries / 20 GeV

0 0.01 0.02 0.03 0.04 0.05 Templates C M = 1043.8 GeV s m = 1142.8 GeV s m = 1241.8 GeV s m

Figure 5.15: TheMC distribution of 3 example templates, with their respective integral normalized to unity. For better readability the error bars are omitted.

The template construction method might introduce a bias towards the final result. To correct for this bias, the templates need to be calibrated. This is done by applying the entire squark extraction method as described in the following section to an independent, high statistics sample of fully simulated signal data without Standard Model background, but including the beam induced background ofγγ →hadrons. The result obtained is:

mqe= 1127.9 GeV±1.35 GeV (5.22)

This value is slightly higher than the input of mq ,_ein = 1123.7 GeV. This bias was

removed by recalibrating each template to a new squark mass which is 4.2 GeV smaller than its respective original value.

In document Diseño del campus virtual de la ESFOT en 3d (página 71-79)