This thesis presented a direct measurement of the decay width of the top quark with ATLAS data.
Due to the unique properties of the top quark, a large variety of precision measurements has been performed in the past few years in the field of top quark physics. The top quark decay widthΓt, however, is one of the fundamental properties which has not been measured with a high precision directly yet. An important motivation for such a direct measurement is a reduced dependence on Standard Model assumptions in comparison to indirect approaches which rely on SM predictions for partial top quark decay widths or single top t-channel cross-sections and presume certain branching ratios. A direct measurement is therefore based on far less assumptions leading to a potential sensitivity to BSM physics.
Nevertheless, performing such a direct measurement is very demanding which is why, in particular, indirect measurements of Γt have been published in the last years. Hence, a pivotal part of this thesis was devoted to the finding and the optimisation of a well-suited and sophisticated analysis setup to extract the decay width out of the given dataset and ease the effort of such a measurement.
The first part of this final chapter focuses on summarising the obtained results and concluding the thesis with an emphasis on the relevant and most challenging aspects of this analysis while the second part is dedicated to an outlook covering prospects for future direct measurements of the top quark decay width.
11.1 Summary of the Obtained Results
The performed measurement of the decay width of the top quark is based on LHC proton-proton collision data recorded with the ATLAS detector at a centre-of-mass energy ofp
s=8 TeV, corre-sponding to an integrated luminosity of 20.2 fb−1, and exploitst¯t events in the lepton+jets channel with one hadronically and one leptonically decaying top quark.
The measurement was realised by using a binned likelihood template fit to data which rests on two observables related to the hadronic and the leptonic decay branch of t¯t events to extractΓt. The templates that enter the likelihood fit were generated by a reweighting method. One observable originates from the hadronic decay branch while the other depends on quantities of the leptonically decaying top quark. The first observable with a good sensitivity toΓt ism`b, the reconstructed invariant mass of the system formed by thebjet associated with the leptonically decaying top quark and the charged lepton`. The second observable, which serves to significantly constrain systematic uncertainties in the combination withm`b, is∆Rmin(jb,jl), the angular distance between thebjet jbassociated with the hadronic top quark and the closest light jetjl from the hadronically decaying
W boson. The events entering the fit were split into events where the charged lepton is either an electron or a muon, into events where exactly one or at least two jets are tagged as originating from a bquark and into two pseudorapidity regions in order to obtain a region which suffers less from detector resolution effects and pile-up contributions. Thus, concatenated distributions composed of eight individual regions constitute the templates employed in the likelihood fit.
The fit to data yielded a decay width of
Γt=1.76±0.33(stat.)+−0.790.68(syst.)GeV=1.76+−0.860.76GeV
formt =172.5 GeV, which agrees well with SM predictions and represents the first direct decay width measurement of the top quark with the ATLAS detector.
The evaluated total uncertainty is by a factor of two smaller than the latest one published by the CDF Collaboration [14] and of a similar order of the preliminary result quoted by the CMS Collaboration[15]. Despite such a satisfactory uncertainty, the direct measurements are still less precise than indirect ones so that the currently achieved sensitivity is not sufficient yet to support or rule out alternative BSM models, predicting decay width values of the top quark which are different from the SM expectation. Due to the limited detector precision and resolution of objects used to define observables, the measured values constitute the best possible result currently reachable.
A first task of the analysis involved the development of the fit framework to conduct the binned likelihood template fit in one or two dimensions with one or two observables fitted simultaneously.
The performance and strength of the self-implemented code relying on dedicated ROOT and RooFit commands was examined in detail and the chosen fit setup was tested with its distinct features extensively, as summarised in Ch. 7. The fit method was validated successfully using calibration curves and pull distributions following the presumed behaviour, and very acceptable statistical uncertainties could be derived, indicating a good sensitivity toΓt.
The analysis challenges do not only comprise the implementation of the fit framework and its validation but also detailed studies on systematic uncertainties. According to the above quoted result ofΓt, the total uncertainty is dominated by systematic uncertainties. Consequently, a relevant part of the thesis was dedicated to understand and reduce dominant systematic uncertainties. As described in Ch. 8, a new treatment of the uncertainty related to the jet flavour composition was employed to decrease the size of the dominant JES uncertainty, and a separate section delineated the understanding of the large uncertainty due to ISR and FSR. Substantial studies served to investigate the impact of NLO and off-shell effects in the top quark decay and the impact of the top quark mass, which is both important for future measurements, as outlined in the next section below. Variations of the top quark mass of±0.5 GeV shiftedΓt by around 0.2 GeV while approaches to estimate the impact of the missing description of NLO and off-shell effects in the current MC samples yielded values of around or less than 0.5 GeV.
In Ch. 9, the focus was laid on finding an adequate analysis configuration in order to reduce
1 1 . 2 O U T L O O K
dominant systematic uncertainties to a significant extent. This included the choice of observables, the split of analysis regions by jet pseudorapidity or energy and further analysis cuts. Moreover, the optimisation of the fit setup was also carried out with respect to the KLFitter configuration utilised to reconstruct the t¯t events for this measurement. The comparison studies, taking the observable modelling into account, verified that the fit configuration with the two fit observables m`b and ∆Rmin(jb,jl) and the eight exclusive analysis regions yield the most promising results.
This substantiates the potential of these comprehensive studies to reduce systematic uncertainties and to optimise event reconstruction efficiencies.
The obtained results allow to deduce that the general analysis strategy with the underlying imple-mentation of a template fit is very well suited to access the top quark decay width. However, there is still room for possible modifications in future measurements which is addressed in the following.
11.2 Outlook
Future direct measurements of the top quark decay width will profit from the intensive studies performed in the course of the analysis presented in this thesis. The basic analysis strategy using a template fit to extractΓt will also be a valuable choice for a measurement using, for instance, ps=13 TeV data, corresponding to the current centre-of-mass energy of the LHC.
In due consideration of the obtained results, potential modifications are primarily useful in the context of the evaluation of systematic uncertainties. The approach of determining systematic uncertainties using pseudo-experiments was widely used in the field of top quark physics in analyses based atp
s=8 TeV. Latest measurements atp
s=13 TeV, however, applied profile likelihood fits with nuisance parameters as additional fit parameters to estimate systematic uncertainties directly during the fit. This method of constraining systematic uncertainties in the measurement may lead to a further reduction of the corresponding uncertainties. For each considered systematic effect, such a nuisance parameter is added. The size of the corresponding systematic uncertainty is adjusted by these parameters since the fitted values from the fit constitute the values which best fit the data.
Initial studies atp
s=8 TeV in order to constrain the radiation systematic uncertainty by such an approach with nuisance parameters were conducted and led to promising results.
With an updated description of the uncertainty due to ISR and FSR forp
s=13 TeV analyses, such a profile likelihood fit may even yield acceptable results for observables that suffered heavily from the radiation uncertainty in this measurement, as shown in Ch. 9.
Further prospects for a future measurement also include updated configurations of the event re-construction with KLFitter. Because of the above mentioned uncertainty due to ISR and FSR, the event reconstruction was carried out with exactly four jets in the event reconstruction step. With an updated description of the radiation uncertainty, the usage of at least five jets for the event reconstruction or further analysis cuts in combination with new KLFitter transfer functions for ps=13 TeV may be applicable and increase the reconstruction efficiencies further.
Additional optimisations regarding the choice of the observables, the number of utilised b-tag
regions or the t¯tdecay channel can be considered in the future as well. In the analysis presented here, analysis regions composed of events with exactly one b-tag were added as this constrained dominant systematic uncertainties to a higher degree. If future analyses are not affected in a similar way, aΓt measurement can be performed with events having at least twob-tagged jets to improve the purity of the sample by further reducing the fraction of background events. Apart from that, more sophisticated b-taggers and even c-tagging algorithms are and will be available for future measurements, facilitating the identification of those jets. The IBL, as the new innermost part of the ATLAS detector, will improve the corresponding efforts. Sincem`boffers a good sensitivity toΓt
while suffering less from systematic uncertainties, a fit configuration using the dileptont¯tchannel can be contemplated as well.
Special emphasis needs to be put on the impact of the top quark mass and the missing NLO precision comprising off-shell effects in the t¯t decay. For a more precise evaluation of those effects, it will be beneficial to have alternative MC samples with masses close to the current nominal value of mt = 172.5 GeV and an MC simulation that provides an NLO description of the t¯t decay, also accounting for off-shell effects. This will eliminate the need for the elaborate studies required to roughly estimate such effects for this measurement atp
s=8 TeV.
The consideration of such analysis items for future direct Γt measurements may lead to an im-provement of the precision reached in this thesis and may in particular allow for a more reliable evaluation of the impact of NLO and off-shell effects in the top quark decay and the influence of the top quark mass. The results shown in this thesis obtained after intensive studies and comparisons correspond to the currently best available estimate.
A significant fraction of fundamental studies and test setups established in the framework of this measurement can be used for future approaches. In this thesis, a template fit associated with well-motivated observables and adequate analysis regions was proven to provide an excellent opportunity to perform precision measurements and access underlying quantities like the top quark decay width although the limited detector resolution of objects needed to design potential observables translates into observable resolutions which are around one order of magnitude larger.