This thesis presents a wide range of studies that go from energy reconstruction at the cell level in TileCal performance reconstruction of the E/T measurement in ATLAS, and finishing with the discovery of the Higgs boson decaying into a pair of W bosons. The reconstruction of the energy, as well as the time, in the TileCal cells is provided by the OF algorithm. These measurements are the inputs for object reconstruction algorithms which base their logic on the signal over noise ratio. In this light, the determination of the cell noise is crucial for an accurate event recons- truction in ATLAS. The impact of the TileCal noise constants has been evaluated through the performance of topoclusters since they clearly allow to identify anoma- lies in the TileCal noise. The results show the evidence of a coherent source of noise which is not properly described by the One-Gaussian approach used so far. This effect produces larger and wider structures for topoclusters in the first ATLAS data collected during 2008 and 2009 compared with the expectations from MC. The results motivated a new description of the TileCal noise constants using a Two- Gaussian method instead. The improvement using the Two-Gaussian description reduces the number of large topoclusters by a factor ∼10 in randomly triggered ATLAS events. The better description of the noise reduces the discrepancies between data and simulation as well. During these investigations, there were also found extremely energetic areas in TileCal. These hot spots are mainly originated by cells, probably affected by electronic damage, which poorly reconstruct the energy of the genuine signal. Bad cells producing hot spots were identified and properly treated at the detector operation level in order to ensure the quality of TileCal reconstruc- tion. In addition to the improvements on topoclusters, the Two-Gaussian method also leads to a better missing transverse momentum measurement. The RMS of the spectrum in data is reduced by a factor 2 and the contribution on the tails highly de- creases from 16% to 0.1%. All these results confirm and validate the Two-Gaussian description of the TileCal noise which has been used for the whole Run I ATLAS data reconstruction. The E/T measurement relies on the momentum conservation law in the transverse plane to the beam axis. For a specific process, this measures the unbalanced transverse momentum from all the particles in the final state, so it is sensitive to the presence of undetectable particles, such as neutrinos. The results on the peformance for the different E/T reconstructions developed in ATLAS are crucial as this measurement plays an important role in many analysis searches. The energetic measurements from the particles produced in the LHC collisions are taken from the ATLAS calorimetric system in the E T miss measurement. The E T miss measurement depends on the number of pile-up interactions since their final products may also deposit energy in the calorimeters. These extra energetic contributions are included in the ETmiss computation degrading the genuine measurement. The increasing pile-up environment at the LHC during 2012 motivated investigations on new approaches for improving the E/T reconstruction in ATLAS. Two pile-up suppressed alternatives based on track information and vertex association were developed: E T miss, STVF and E T miss,track, respectively. The former follows the calorimeter-based approach of the E T miss but scaling down the soft term and rejecting pile-up jets. The latter relies on the energy measured from well reconstructed tracks in the inner detector which are associated to the primary vertex of the event. The performance of the several approaches based on the object information to build the unbalanced transverse momentum have been evaluated in terms of resolution, scale and linearity. Results show that besides the better stability against pile-up of the E T miss, STVF and Emiss,track reconstructions, these approaches come with new features. For the E T miss, STVF, the poor modelling of tracks coming from pile-up interactions produces an under-calibrated soft term in MC. This results in discrepancies between data and simulation, specially in events without jets. In events with jets, the E T miss and E T miss,STVF perform very similar because the dominant component is the jet-term. The E T miss,track measurement is very robust against extra interactions since only tracks associated to the vertex of the hardest process are included. However, limited ID coverage and missing high-pT neutral particles lead to large degradation in the E T miss,track linearity and scale, specially in event topologies with high jet activity. Due to the variety on E/T reconstructions and their behaviours depend- ing on the event topology, the optimal measurement may be different based on the characteristics of the physics process to study. The second part of the thesis describes the strategy of the H→ W W (∗)→ lνlν analysis and reports the results using the complete ATLAS Run I data. This corres- ponds to about 25fb−1 at √s=7and8TeV collected with the ATLAS detector at the LHC. The Higgs boson decaying into a pair of W bosons benefits from a larger BR compared with other final states for a wide range of the Higgs boson mass. This makes the H→WW(∗)→lνlν analysis one of the most important channels for the Higgs boson search. However, this analysis suffers from high background contamination, which difficulties the distinction between the Higgs boson signal and other processes that may have the same reconstructed final state. In addition, the analysis is not sensitive to the Higgs boson mass due to the presence of the two neutrinos coming from the W bosons. These two facts define the strategy of the H→WW(∗)→lνlν analysis. The selection criteria should find an optimal compro- mise to be hard enough for rejecting as many background contributions as possible and, at the same time, soft enough for still keeping the Higgs boson signal. The anal- ysis selects events with exactly two high-pT well reconstructed leptons (electrons or muons)oppositelychargedandwithE/T measurementoriginatedbythefinalneutri- nos. In order to deal with different background contributions the events are divided by the number of jets as well as by the flavour of the two leptons. This separation al- lows to adequate the selection since the background composition is different in each category. In general, final states with same flavour leptons are mostly populated by Z/γ∗ background while events with different flavoured leptons are mainly originated by top quark processes. For the former, the analysis applies a combined requirement using several E/T reconstructions in order to further suppress Z/γ∗ contributions, for which non genuine E/T measurement is expected. The latter vetoes jets which are considered as produced by a b quark from reconstruction algorithms. In addition, the division on the jet multiplicity also allows to distinguish Higgs candidates as originated by gluon-gluon fusion (with zero or up to one jet) or vector boson fusion (at least two jets) production mechanisms. This distinction leads to better separate Higgs signal from the remaining backgrounds in each case by exploiting the differ- ences in dilepton kinematics and, when relevant, in jet based magnitudes. After all selection is applied, the transverse mass of the dilepton system and the E/T of the Higgs candidate events is used as final discriminants in a statistical test. Given the importance of simulating all background processes correctly, the analysis builds dif- ferent control regions to check the agreement between data and MC. The differences in the control regions are also inputs to the statistical procedure. This ensures that the likelihood fit includes them properly as associated uncertainties in the final results. The first results of the H→WW(∗)→lνlν analysis showed an excess of events over the expected background observed for mH = 125 GeV with a signal significance of 3.8 σ, for which the expectation is 3.7 σ. The best fit signal strength at that mass is μ = 1.01 ± 0.31. The expected VBF signal significance at mH = 125 GeV is 1.6 σ and the observation results in 2.5 σ. The first H→WW(∗)→lνlν results are consis- tent with the measurements from the H → γγ and H → ZZ → 4l searches. All ATLAS measurements from Higgs decaying into boson pair searches are combined allowing to observe an excess over the expectation with a local significance of 5 σ. After these first results several studies were focussing on optimising the selection of the H→WW(∗)→lνlν analysis in order to enhance the sensitivity of the search. The final optimised results mainly benefit from the development of a new E/T reconstruction, represented by the symbols E T miss,track,jetCorr or p T miss. This new reconstruction is based on the E T miss,track approach but replacing tracks by the calorimetric measurements of the objects associated to them and adding jets which are missing in the original E T miss,track computation. Although this may create a higher dependence with pile-up, the new approach still profits from pile-up rejection from the original track-based selection and with a much more accurate measurement in topologies with neutral particles in the final state. The results show that the p T miss is able to recover the resolution in events with jets while still maintains a good stability with pile-up and smaller tails in Z → ll process. Additional investigations using event topologies with genuine E/T also point to a more reliable measurements of the expected E/T when using the p T miss reconstruction. The strategy for optimising the E/T criteria in the H→WW(∗)→lνlν analysis is based on simulated final candidate events evaluated through the statistical likelihood fit. Given the composition and contribution of the different backgrounds depend on the final state , theE/T optimisation is evaluated in each analysis category. The different E/T measurements perform very similarly at the end of the event selection for eμ+μe final states. The low region of the spectrums are almost not populated since the main backgrounds, as well as the Higgs boson signal, are expected to have genuine E/T . In addition, the analysis requirements on mll and p T ll, which are correlated with the E/T measurement, sculpt the E/T shapes at the end of the selection. Hence, there are almost no differences in the expected significance values using any of the E/T reconstructions. However, the pmiss is preferred because of its better performance and resolution. A conservative threshold of 20GeV is used to deal with possible mis-measurements from multi- jets background in H+0j and H+1j analyses. Since the Higgs boson produced via VBF is typically characterised by two emerging quarks, the E/T measurement for the Higgs signal is expected to be smaller than in the ggF production mode. In this light, the VBF strategy does not apply any threshold on the E/T measurement since the low region of the spectrum is mainly populated by signal events. Final states with same flavoured leptons are affected by a huge Z/γ∗ contribution, so combining several E/T reconstructions achieves further Z/γ∗ rejection. In this case when the final state contains up to one jet, the requirement is done using the projected E/T,Rel magnitude for the Emiss and E T miss,track measurements. Investigations on the direction of the new pmiss conclude that the rejection power of the original E T miss,track is still higher. This is due to the fact that the latter tends to point to the mismea- sured jets, hence the E/T,Rel computation using E T miss,track benefits to highly reject 
the Z/γ∗ contribution. E T miss computed with the E T miss,track still provides the best 
significance. For the VBF-enriched analysis, however, the E/T,Rel magnitude may be biased because of the probability to randomly project the nominal measurement to any reconstructed jet. This points back to the usage of the p T miss measurement, complemented with a purely calorimeter-based E T miss threshold. The better p T miss performance can be exploited to also benefit other E/T -dependent quantities used in the H→WW(∗)→lνlν, as the mT. Results show a better resolution of the mT measurement obtained by using the p T miss in the computation. The usage of the p T miss in the mT leads to a better separation between the Higgs signal and the remaining backgrounds, specially for multi-jet and non-W W diboson processes. These optimal thresholds using the p T miss measurement increase the expected significance by 7% in the ggF-enriched analysis of the H→ W W (∗)→ lνlν search. The introduction of the pmiss for the transverse mass computation enhances the expected significance by 9%. For the VBF-enriched search, the overall improvement due to the optimised E/T selection is observed up to 14% in the expected significance results. Finally, the optimisation of the H→WW(∗)→lνlν analysis has been developed using the complete 8 TeV data sample. The main improvements rely on the introduction of more performant variables as the pmiss, new techniques for background estimation, and extensions of the Higgs signal phase space to enhance the sensitivity of the search. After the whole optimisation, the expected significance of the ggF production mode increases from 2.8 σ to 4.36 σ just in eμ+μe final states. For the VBF production mode of the Higgs boson the overall gain is up to 70% due to the BDT technique applied now for the H+2j category. The last H→WW(∗)→lνlν results using Run I ATLAS data are reported at mH = 125.36 GeV. There is an excess over background of 6.1 σ observed for the H→WW(∗)→lνlν analysis for which the SM expectation is 5.8 σ. Evidence of the VBF production mode is also obtained with a significance of 3.2 σ. All the measurements are consistent with SM Higgs boson expectations and state the observation of the Higgs boson decaying to WW∗ in ATLAS.