Measurements

• b-jet mode:Light˜band˜ttogether with enhanced heavy flavour production (due to Higgsino couplings) lead to many b quarks in SUSY decay chains. This feature can be exploited to suppress QCD background. This search mode requires four jets, out of which at least two are b-tagged, andE_T^miss.

7.4.1 Measurements

Once a signature consistent with SUSY has been established, the experimental focus will be to reconstruct the sparticle mass spectrum, to constrain the model parameters, and to measure the spin of the new (s)particles. In RPC models, sparticle decay chains cannot be fully resolved since the LSPs escape detection. As a consequence, edge positions rather than mass peaks in invariant mass distributions are measured and fitted.

One suitable sparticle decay chain is

˜

q_L→χ˜⁰₂q(→`˜<sup>±</sup>`^∓q)→χ˜⁰₁`<sup>+</sup>`⁻q (7.2) leading to final states containing two opposite-sign electrons or muons, hard jets and missing energy. These characteristics ensure a large signal to background ratio. The kinematic endpoint in the dilepton invariant mass distribution is a function of the sparticle masses involved. If the sleptons are heavier than the χ˜⁰₂ then the decay proceeds through the three body channel χ˜⁰₂ →

˜

χ⁰₁`<sup>+</sup>`⁻. In this case the invariant mass distribution is non-triangular in shape with an endpoint equal to the difference of the mass of the two neutralinosm^edge_`` =m_χ˜0

2 −m_χ˜0

1 [140,141]. If at least one of the sleptons is lighter than the χ˜⁰₂ then the two-body decay channel χ˜⁰₂ → `˜<sup>±</sup>`^∓ →

˜

χ⁰₁`<sup>+</sup>`⁻dominates. The dilepton invariant mass distribution is triangular with a sharp edge at the endpoint:

m^edge_`` =m_χ˜0 2

vu ut1−

à m_`˜ m_χ˜0 2

!2s 1−

µm_χ˜0 1

m_`˜

¶2

. (7.3)

A measurement of the dilepton endpoint thus gives a handle on the masses of the two lightest neutralinos and any sleptons that are lighter than χ˜⁰₂. Figure 7.10 shows the dilepton invariant mass distribution for two SUSY benchmark models. The SM background has been reduced by subtracting opposite flavour lepton pairs. The fitting function is a triangular distribution (7.3), smeared with a resolution function. To determine the masses of all sparticles involved in the decay chain (7.2), further invariant mass distributions involving a jet are used. In addition, several other sparticle decay chains give further kinematic endpoints. As an indication of the precision with which the LSP mass can be reconstructed, the χ˜⁰₁ mass, as obtained from endpoints of the lepton+jets edges and the dilepton edges, is found to be88±60 GeV(SU3) and62±126 GeV (SU4) for an integrated luminosity of1fb⁻¹ and0.5fb⁻¹, respectively [26]. The trueχ˜⁰₁ masses are118 GeV(SU3) and60 GeV(SU4).

Once enough edge positions have been measured, model parameters can be constrained. In the initial phase a limited number of measurements with rather large uncertainties would only allow to fit SUSY models with few parameters. In an optimistic scenario in which SUSY has been discovered and first endpoints have been measured with early data, i.e. an integrated luminosity of

m(ll) [GeV]

0 20 40 60 80 100 120 140 160 180 200 -1Entries/4 GeV/ 1 fb

-10

-1Entries/4 GeV/ 0.5 fb

-40

Figure 7.10:Left: Distribution of invariant mass after background subtraction for the SU3 benchmark point with an integrated luminosity of 1 fb⁻¹. Right: the same distribution is shown for the SU4 benchmark point and an integrated luminosity of 0.5 fb⁻¹. The line histogram is the SM contribution, while the points are the sum of SM and SUSY contributions. The fitting function is superimposed and the expected position of the endpoint is indicated by a dashed line. (Taken from Ref. [26])

1fb⁻¹, it is shown in Ref. [26] that most of the5mSUGRA parameters can already be constrained (for the two studied benchmark points).

Assuming mSUGRA, the achievable precision in cosmological parameters derived from LHC data for300fb⁻¹was evaluated in Ref. [142], The neutralino relic densityΩ_χh², calculated using MICROMEGAS [143], is obtained with a precision of ∼ 3%. A more general approach was considered in Ref. [144] where dark matter properties are estimated assuming the MSSM. The error onΩ_χh²is found to be∼9%, again for300fb⁻¹of data.

A method to measure the spin of SUSY particles at the LHC has been proposed in Ref. [145]. This method exploits the fact that angular distributions in sparticle decays lead to charge asymmetry in lepton-jet invariant mass distributions. It was shown with simulation studies that the asymmetry distributions are sufficient to determine the correct spin of some SUSY particles, and reject other possibilities.

Chapter 8

Monte Carlo Simulations

The generation of specific physics processes along with the detailed simulation of the detector response is a crucial part of modern high-energy physics. It is important for the design of de-tectors, to estimate the physics reach, for a detailed understanding of the data, and for precision measurements.

The first LHC physics runs, at a reduced luminosity (initially10³¹cm⁻²s⁻¹) and10TeV centre-of-mass (CM) energy, are foreseen for winter of 2009. The SUSY studies presented in this thesis are therefore solely based on simulated data. In this chapter the employed simulated datasets are introduced. They were generated as part of the so-called ATLAS computing system com-missioning (CSC) programme [26]. The author’s contribution was the production of the25×25 mSUGRA signal grid (described in Section8.3.2), including the generation of the SUSY spectra, and the whole production chain (generation, simulation) making full use of the computing grid.

Section8.1briefly describes the Monte Carlo event generators. Section8.2gives an overview of the ATLAS detector simulation. Finally, all simulated signal and background samples, which are relevant for the following SUSY analysis, are detailed in Section8.3.

8.1 Monte Carlo Generators

The ATLAS software framework provides interfaces¹to most general-purpose Monte Carlo (MC) event generators, including: PYTHIA [111], HERWIG [146], Sherpa [147], AcerMC [148], ALP-GEN [149], MadGraph/MadEvent [150], and MC@NLO [151,152]. In addition to these, further generators are available for the generation of more specific processes.

Parton-level MC generators are configured to use HERWIG/JIMMY or PYTHIA for the hadro-nisation and the underlying event modelling. In the former case, HERWIG is employed for the hadronisation and the JIMMY program [153] (versions 4.2 and 4.31) to simulate the

underly-1The event generator interfaces provide mechanisms to feed the generated particle-level events into the ATLAS simulation software package.

103

ing event. The model parameters of the underlying event were tuned to published data from the Tevatron and other experiments, as described in Ref. [58] and references therein.

The specialised TAUOLA package [154] (version 2.7) is utilised for the simulation ofτ-lepton de-cays. The radiation of photons from charged leptons is also treated separately, using the PHOTOS QED radiation package [155] (version 2.15).

Wherever available, MC generators and tools are taken from the LHC computing grid generator-services sub-project. Furthermore, a common definition of particle masses is used among all generators, e.g. all simulated datasets employed in the present work were generated with a top quark mass of175GeV.

Parton distribution functions (PDFs) are linked into all MC generators using the Les Houches accord PDF interface library (LHAPDF) [156]. All datasets employed in the present work used the PDF sets [157,158,159] CTEQ6L for leading order (LO) MC event generators, and CTEQ6M for the next-to-leading order (NLO) MC event generator MC@NLO.

A brief description of the employed MC generators is given in the following, details about the datasets (generator filter settings, production cross sections etc.) are further described in Sec-tion8.3.

ThePYTHIA MC event generator [111] is used for the simulation of QCD jets. The new imple-mentation of parton showering, commonly known asp_T-ordered showering, is used, as well as the new underlying event model where the phase-space is interleaved/shared between initial-state radiation (ISR) and the underlying event.

TheHERWIGMC event generator [146,160] is employed for the simulation of SUSY processes.

The pre-generated input tables (SUSY particle masses, and branching ratios) for these processes are provided by ISAJET and ISAWIG [130]. HERWIG is also used to generate electroweak bo-son pair samples (W W, ZZ, W Z). The simulation of the underlying event is performed with JIMMY [153].

The ALPGENMC event generator [149] (version 2.05) is used for several processes: W andZ boson production in association with up to five jets, andt¯tproduction with up to three additional jets. ALPGEN calculates the exact matrix elements of multiparton hard processes in hadronic collisions, at leading order in QCD and electroweak interactions. The parton multiplicity in the matrix element (N = 1 to 6) has to be specified before running ALPGEN, i.e. the simulation of the physics process is sliced in N-partons samples. HERWIG and JIMMY are inserted for the hadronisation and simulation of the underlying event, respectively. In order to perform the parton-showering and matrix element matching ALPGEN implements the MLM [161] technique.²Since exclusive matching is applied, the matched samples (each with N-partons) can be added, and the inclusive sample is obtained after summing up all N-partons samples. Also the total (inclusive) cross section is given by the sum over all cross sections, each multiplied by its MLM matching efficiency.

TheMC@NLOevent generator [151,152] (version 3.3) was employed for the production of the

2The MLM matching technique prevents double counting of parton emission either through the matrix element or the parton shower. The procedure vetoes events where the parton shower generates jets that have already been generated by the ALPGEN matrix elements.