1 Introduction

In a theory with broken supersymmetry (SUSY) [19], the mass scale of the supersymmetric particles is undetermined. However, for SUSY to provide a solution to the hierarchy problem [1013] some of the new SUSY particles masses are typically required to be below about one TeV [14, 15], hence they could be within the reach of the LHC.

The scalar partners of the right-handed and left-handed chiral components of the top-quark state (\(\tilde{t}^{}_{\mathrm {R}}\) and \(\tilde{t}^{}_{\mathrm {L}}\) respectively) are among these particles. In many supersymmetric models, the large Yukawa coupling of the top quark to the Higgs sector makes the Higgs boson mass sensitive to the masses of the scalar top (referred to as stop in the following) states, such that, to avoid fine tuning, their masses are often required to be light. The \(\tilde{t}^{}_{\mathrm {R}}\) and \(\tilde{t}^{}_{\mathrm {L}}\) components mix to form the mass eigenstates \(\tilde{t}^{}_{1}\) and \(\tilde{t}^{}_{2}\), \(\tilde{t}^{}_{1}\) being defined as the lighter of the two. The scalar superpartner of the left-handed chiral component of the bottom quark (\(\tilde{b}^{}_{\mathrm {L}}\)) belongs to the same weak isospin doublet as the \(\tilde{t}^{}_{\mathrm {L}}\), hence they usually share the same supersymmetry-breaking mass parameter: a light stop can therefore imply the existence of a light scalar bottom. The lightest sbottom mass eigenstate is referred to as \(\tilde{b}^{}_{1}\).

The ATLAS and CMS collaborations have searched for direct production of stops and sbottoms  [1635] using about 4.7 \({\mathrm{fb}^{-1}}\) of data from the proton–proton collisions produced by the LHC at \(\sqrt{s} = 7\) TeV and 20 \({\mathrm{fb}^{-1}}\) at \(\sqrt{s} = 8\) TeV. These searches have found no evidence of third-generation squark signals, leading to exclusion limits in many SUSY models. The aim of this paper is to summarise the sensitivity of the ATLAS experiment to R-parity-conservingFootnote 1 [3842] models including the direct pair production of stops and sbottoms using the full \(\sqrt{s}=8\) TeV proton–proton collision dataset collected during Run 1 of the LHC.Footnote 2 The third-generation squarks are assumed to decay to the stable lightest supersymmetric particle (LSP) directly or through one or more intermediate stages. The analyses considered are those previously published by the ATLAS collaboration on the topic, together with new ones designed to increase the sensitivity to scenarios not optimally covered so far. A wide range of SUSY scenarios are studied by combining different analyses to improve the global sensitivity.

The paper is organised as follows: Sect. 2 briefly reviews the expected phenomenology of third-generation squark production and decay; Sect. 3 reviews the general analysis approach followed by the ATLAS collaboration for SUSY searches; Sects. 4 and 5 present the exclusion limits obtained in specific models by combining the results of several analyses. Two different types of models have been considered: simplified models, where the third-generation squarks are assumed to decay into typically one or two different final states, and more complex phenomenological supersymmetric models, where the stop and sbottom have many allowed decay channels. Conclusions are drawn in Sect. 6.

For the sake of brevity, the body of the paper provides no details of the ATLAS detector and object reconstruction, of the analyses used in the limit derivation, or of how the signal Monte Carlo simulation samples were generated. However, a comprehensive set of appendices is provided to supply additional information to the interested reader. Appendix  briefly summarises the layout of the ATLAS detector and the general principles used in the reconstruction of electrons, muons, jets, jets containing \(b\)-hadrons (\(b\)-jets), and the missing transverse momentum vector \(\mathbf {p}_{\mathrm {T}}^{\mathrm {miss}}\) (whose magnitude is referred to as \(E_{\mathrm {T}}^{\mathrm {miss}} \)). Appendix  discusses the analyses used to derive the exclusion limits presented in Sects. 4 and 5. The analyses that have already been published are only briefly reviewed, while those presented for the first time in this paper are discussed in detail. Appendix C provides further details of a combination of analyses which is performed for the first time in this paper. Finally, Appendix D provides details about the generation and simulation of the signal Monte Carlo samples used to derive the limits presented.

2 Third-generation squark phenomenology

The cross section for direct stop pair production in proton–proton collisions at \(\sqrt{s} = 8\) TeV as a function of the stop mass as calculated with PROSPINO  [43, 44] is shown in Fig. 1a. It is calculated to next-to-leading order accuracy in the strong coupling constant, adding the resummation of soft gluon emission at next-to-leading-logarithmic accuracy (NLO\(+\)NLL) [4547]. In this paper, the nominal cross section and its uncertainty are taken from an envelope of cross-section predictions using different parton distribution function (PDF) sets and factorisation and renormalisation scales described in Ref. [44]. The difference in cross section between the sbottom and stop pair production is known to be small [46], hence the values of Fig. 1a are used for both.

Fig. 1
figure 1

a Direct stop pair production cross section at \(\sqrt{s} = 8\) TeV as a function of the stop mass. The band around the cross section curve illustrates the uncertainty (which is everywhere about 15–20 %) on the cross section due to scale and PDF variations. b Illustration of stop decay modes in the plane spanned by the masses of the stop (\(\tilde{t}^{}_{1}\)) and the lightest neutralino (\(\tilde{\chi }^{0}_{1}\)), where the latter is assumed to be the lightest supersymmetric particle and the only one present among the decay products. The dashed blue lines indicate thresholds separating regions where different processes dominate

Searches for direct production of stops and sbottoms by the ATLAS collaboration have covered several possible final-state topologies. The experimental signatures used to identify these processes depend on the masses of the stop or sbottom, on the masses of the other supersymmetric particles they can decay into, and on other parameters of the model, such as the stop and sbottom left–right mixing and the mixing between the gaugino and higgsino states in the chargino–neutralino sector.

Assuming that the lightest supersymmetric particle is a stable neutralino (\(\tilde{\chi }^{0}_{1}\)), and that no other supersymmetric particle plays a significant role in the sbottom decay, the decay chain of the sbottom is simply \(\tilde{b}^{}_{1}\rightarrow b \tilde{\chi }^{0}_{1}\) (Fig. 2a).

Fig. 2
figure 2

Diagrams of \(\tilde{t}^{}_{1}\) and \(\tilde{b}^{}_{1}\) pair production and decays considered as simplified models: a \(\tilde{b}^{}_{1}\tilde{b}^{}_{1}\rightarrow b\tilde{\chi }^{0}_{1}b \tilde{\chi }^{0}_{1}\); b \(\tilde{t}^{}_{1}\tilde{t}^{}_{1}\rightarrow t\tilde{\chi }^{0}_{1}t \tilde{\chi }^{0}_{1}\); c three-body decay; d four-body decay; e \(\tilde{t}^{}_{1}\tilde{t}^{}_{1}\rightarrow c\tilde{\chi }^{0}_{1}c\tilde{\chi }^{0}_{1}\); f \(\tilde{t}^{}_{1}\tilde{t}^{}_{1}\rightarrow b\tilde{\chi }^{\pm }_{1}b\tilde{\chi }^{\pm }_{1}\); g \(\tilde{b}^{}_{1}\tilde{b}^{}_{1}\rightarrow t\tilde{\chi }^{\pm }_{1}t\tilde{\chi }^{\pm }_{1}\); h \(\tilde{b}^{}_{1}\tilde{b}^{}_{1}\rightarrow b\tilde{\chi }^{0}_{2}b\tilde{\chi }^{0}_{2}\). The diagrams do not show “mixed” decays, in which the two pair-produced third-generation squarks decay to different final states

A significantly more complex phenomenology has to be considered for the stop, depending on its mass and on the \(\tilde{\chi }^{0}_{1}\) mass. Figure 1b shows the three main regions in the \(m_{\tilde{t}^{}_{1}}\)\(m_{\tilde{\chi }^{0}_{1}}\) plane that are taken into account. They are identified by different values of \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})= m_{\tilde{t}^{}_{1}}\)-\(m_{\tilde{\chi }^{0}_{1}}\). In the region where \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})> m_t \), the favoured decay is \(\tilde{t}^{}_{1}\rightarrow t \tilde{\chi }^{0}_{1}\) (Fig. 2b). The region where \(m_W + m_b < \Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})< m_t \) is characterised by the three-body decayFootnote 3 (\(\tilde{t}^{}_{1}\rightarrow Wb \tilde{\chi }^{0}_{1}\) through an off-shell top quark, Fig. 2c). The region where the value of \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})\) drops below \(m_W + m_b\), sees the four-body decay \(\tilde{t}^{}_{1}\rightarrow b f f' \tilde{\chi }^{0}_{1}\), (where \(f\) and \(f'\) indicate generic fermions coming from the decay of an off-shell \(W\) boson, Fig. 2d) competing with the flavour-changing decayFootnote 4 \(\tilde{t}^{}_{1}\rightarrow c \tilde{\chi }^{0}_{1}\) of Fig. 2e; the dominant decay depends on the details of the supersymmetric model chosen [50].

If the third-generation squark decay involves more SUSY particles (other than the \(\tilde{\chi }^{0}_{1}\)), then additional dependencies on SUSY parameters arise. For example, if the lightest chargino (\(\tilde{\chi }^{\pm }_{1}\)) is the next-to-lightest supersymmetric particle (NLSP), then the stop tends to have a significant branching ratio for \(\tilde{t}^{}_{1}\rightarrow b \tilde{\chi }^{\pm }_{1}\) (Fig. 2f), or, for the sbottom, \(\tilde{b}^{}_{1}\rightarrow t \tilde{\chi }^{\pm }_{1}\) if kinematically allowed (Fig. 2g). The presence of additional particles in the decay chain makes the phenomenology depend on their masses. Several possible scenarios have been considered, the most common ones being the gauge-universality inspired \(m_{\tilde{\chi }^{\pm }_{1}} = 2 m_{\tilde{\chi }^{0}_{1}}\), favoured, for example, in mSUGRA/CMSSM models [5156]; other interpretations include the case of a chargino almost degenerate with the neutralino, a chargino almost degenerate with the squark, or a chargino of fixed mass. Another possible decay channel considered for the sbottom is \(\tilde{b}^{}_{1}\rightarrow b \tilde{\chi }^{0}_{2}\rightarrow b h \tilde{\chi }^{0}_{1}\) (Fig. 2h), which occurs in scenarios with a large higgsino component of the two lightest neutralinos.

Fig. 3
figure 3

Diagrams of \(\tilde{t}^{}_{2}\) decays considered as simplified models: a \(\tilde{t}^{}_{2}\tilde{t}^{}_{2}\rightarrow \tilde{t}^{}_{1}Z \tilde{t}^{}_{1}Z\); b \(\tilde{t}^{}_{2}\tilde{t}^{}_{2}\rightarrow \tilde{t}^{}_{1}h \tilde{t}^{}_{1}h \); c \(\tilde{t}^{}_{2}\tilde{t}^{}_{2}\rightarrow t\tilde{\chi }^{0}_{1}t\tilde{\chi }^{0}_{1}\). The diagrams do not show “mixed” decays, in which the two pair-produced third-generation squarks decay to different final states. The decay \(\tilde{t}^{}_{2}\rightarrow \gamma \tilde{t}^{}_{1}\) is not an allowed process

Despite the lower production cross section and similar final states to \(\tilde{t}^{}_{1}\), the heavier stop state (\(\tilde{t}^{}_{2}\)) pair production has also been studied: the search for it becomes interesting in scenarios where the detection of \(\tilde{t}^{}_{1}\) pair production becomes difficult (for example if \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})\sim m_t \)). The diagrams of the investigated processes are shown in Fig. 3.

Two types of SUSY models are used to interpret the results in terms of exclusion limits. The simplified model approach assumes that either a stop or a sbottom pair is produced and that they decay into well-defined final states, involving one or two decay channels. Simplified models are used to optimise the analyses for a specific final-state topology, rather than the complex (and model-dependent) mixture of different topologies that would arise from a SUSY model involving many possible allowed production and decay channels. The sensitivity to simplified models is discussed in Sect. 4.

More complete phenomenological minimal supersymmetric extensions of the Standard Model (pMSSM in the following [57]) are also considered, to assess the performance of the analyses in scenarios where the stop and sbottom typically have many allowed decay channels with competing branching ratios. Three different sets of pMSSM models are considered, which take into account experimental constraints from LHC direct searches, satisfying the Higgs boson mass and dark-matter relic density constraints, or additional constraints arising from considerations of naturalness. The sensitivity to these models is discussed in Sect. 5.

3 General discussion of the analysis strategy

The rich phenomenology of third-generation supersymmetric particles requires several event selections to target the wide range of possible topologies. A common analysis strategy and common statistical techniques, which are extensively described in Ref. [58], are employed.

Signal regions (SR) are defined, which target one specific model and SUSY particle mass range. The event selection is optimised by relying on the Monte Carlo simulation of both the Standard Model (SM) background production processes and the signal itself. The optimisation process aims to maximise the expected significance for discovery or exclusion for each of the models considered.

For each SR, multiple control regions (CR) are defined: they are used to constrain the normalisation of the most relevant SM production processes and to validate the MC predictions of the shapes of distributions of the kinematic variables used in the analysis. The event selection of the CRs is mutually exclusive with that of the SRs. It is, however, chosen to be as close as possible to that of the signal region while keeping the signal contamination small, and such that the event yield is dominated by one specific background process.

A likelihood function is built as the product of Poisson probability functions, describing the observed and expected number of events in the control and signal regions. The observed numbers of events in the various CRs and SRs are used in a combined profile likelihood fit [59] to determine the expected SM background yields for each of the SRs. Systematic uncertainties are treated as nuisance parameters in the fit and are constrained with Gaussian functions with standard deviation equal to their value. The fit procedure takes into account correlations in the yield predictions between different regions due to common background normalisation parameters and systematic uncertainties, as well as contamination from SUSY signal events, when a particular model is considered for exclusion.

The full procedure is validated by comparing the background predictions and the shapes of the distributions of the key analysis variables from the fit results to those observed in dedicated validation regions (VRs), which are defined to be orthogonal to, and kinematically similar, to the signal regions, with low potential contamination from signal.

After successful validation, the observed yields in the signal regions are compared to the prediction. The profile likelihood ratio statistic is used first to verify the SM background-only hypothesis, and, if no significant excess is observed, to exclude the signal-plus-background hypothesis in specific signal models. A signal model is said to be excluded at 95 % confidence level (CL) if the CL\(_s\) [60, 61] of the profile likelihood ratio statistics of the signal-plus-background hypothesis is below 0.05.

Several publications, targeting specific stop and sbottom final-state topologies, were published by the ATLAS collaboration at the end of the proton–proton collision run at \(\sqrt{s} = 8\) TeV, using a total integrated luminosity of about 20 \({\mathrm{fb}^{-1}}\). Each of these papers defined one or more sets of signal regions optimised for different simplified models with different mass hierarchies and decay modes for the stop and/or sbottom. A few additional signal regions, focusing on regions of the parameter space not well covered by existing analyses have been defined since then. All signal regions that are used in this paper are discussed in detail in Appendix , while Table 1 introduces their names and the targeted models. Each analysis is identified by a short acronym defined in the second column of Table 1. The signal region names of previously published analyses are retained, but, to avoid confusion and to ease the bookkeeping, the analysis acronym is prepended to their names. For example, SRA1 from the t0L analysis of Ref. [16], which is a search for stop pair production in channels with no leptons in the final state, is referred to as t0L-SRA1.

Table 1 Summary of the ATLAS analyses and signal regions used in this paper. Each signal region is identified by the acronym of the corresponding analysis followed by the original name of the signal region defined either in the published paper or in Appendix B.2. A dash in the signal region name column indicates that the analysis does not use the concept of signal region

4 Interpretations in simplified models

The use of simplified models for analysis optimisation and result interpretation has become more and more common in the last years. The attractive feature of this approach is that it focuses on a specific final-state topology, rather than on a complex (and often heavily model-dependent) mixture of several different topologies: only a few SUSY particles are assumed to be produced in the proton–proton collision – often just one type – and only a few decay channels are assumed to be allowed. In the remainder of this section, several exclusion limits derived in different supersymmetric simplified models are presented. Details about how the MC signal samples used for the limit derivations were produced are available in Appendix D.

4.1 Stop decays with no charginos in the decay chain

A first series of simplified models is considered. It includes direct stop pair production as the only SUSY production process, and assumes that no supersymmetric particle other than the \(\tilde{t}^{}_{1}\) itself and the LSP, taken to be the lightest neutralino \(\tilde{\chi }^{0}_{1}\), is involved in the decay. Under this assumption, there is little model dependence left in the stop phenomenology, as discussed in Sect. 2. The stop decay modes are defined mainly by the mass separation \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})\) between the stop and the neutralino, as shown in Fig. 1b. The corresponding diagrams are shown in Fig. 2.

Figure 4 shows the 95 % CL exclusion limits obtained in the \(m_{\tilde{t}^{}_{1}}-m_{\tilde{\chi }^{0}_{1}}\) plane by the relevant analyses listed in Table 1 and discussed in Appendix , or by their combination. A detailed discussion of which analysis is relevant in each range of \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})\) follows.

Fig. 4
figure 4

Summary of the ATLAS Run 1 searches for direct stop pair production in models where no supersymmetric particle other than the \(\tilde{t}^{}_{1}\) and the \(\tilde{\chi }^{0}_{1}\) is involved in the \(\tilde{t}^{}_{1}\) decay. The 95 % CL exclusion limits are shown in the \(m_{\tilde{t}^{}_{1}}\)\(m_{\tilde{\chi }^{0}_{1}}\) mass plane. The dashed and solid lines show the expected and observed limits, respectively, including all uncertainties except the theoretical signal cross-section uncertainty (PDF and scale). Four decay modes are considered separately with a branching ratio of 100 %: \(\tilde{t}^{}_{1}\rightarrow t \tilde{\chi }^{0}_{1}\), where the \(\tilde{t}^{}_{1}\) is mostly \(\tilde{t}^{}_{\mathrm {R}}\), for \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})> m_t \); \(\tilde{t}^{}_{1}\rightarrow Wb\tilde{\chi }^{0}_{1}\) (three-body decay) for \(m_W + m_b< \Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})< m_t \); \(\tilde{t}^{}_{1}\rightarrow c \tilde{\chi }^{0}_{1}\) and \(\tilde{t}^{}_{1}\rightarrow bff'\tilde{\chi }^{0}_{1}\) (four-body decay) for \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})< m_W + m_b\). The latter two decay modes are superimposed

\(\varvec{\Delta } \varvec{m} (\tilde{\varvec{t}}^{}_{\mathbf {1}}, \tilde{\varvec{\chi }}^{\mathbf {0}}_{\mathbf {1}})< \varvec{m}_{\varvec{W}}+\varvec{m}_{\varvec{b}}\) This kinematic region is characterised by the presence of two competing decays: the flavour-violating decay \(\tilde{t}^{}_{1}\rightarrow c\tilde{\chi }^{0}_{1}\) (Fig. 2e) and the four-body decay \(\tilde{t}^{}_{1}\rightarrow bff' \tilde{\chi }^{0}_{1}\) (Fig. 2d). Which one of the two becomes dominant depends on the model details, in particular on the mass separation between the stop and the neutralino, and on the amount of flavour violation allowed in the model [50]. Several analyses have sensitivity in this region of the \(m_{\tilde{t}^{}_{1}} - m_{\tilde{\chi }^{0}_{1}}\) plane. The monojet-like signal regions (tc-M1-3) dominate the sensitivity in the region with \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})\gtrsim m_b\), regardless of the decay of the stop pair, which goes undetected: their selection is based on the presence of an initial-state radiation (ISR) jet recoiling against the stop-pair system, which is assumed to be invisible. At larger values of \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})\), signal regions requiring the presence of a \(c\)-tagged jet (tc-C1-2) complement the monojet-like signal regions by targeting the \(\tilde{t}^{}_{1}\rightarrow c \tilde{\chi }^{0}_{1}\) decay. Limits on four-body decays can be set using signal regions which include low transverse momentum electrons and muons (t1L-bCa_low and WW).

The limits reported in Fig. 4 for these values of \(\Delta m\) all assume that the branching ratio of the stop decay into either \(\tilde{t}^{}_{1}\rightarrow c \tilde{\chi }^{0}_{1}\) or \(\tilde{t}^{}_{1}\rightarrow b f f' \tilde{\chi }^{0}_{1}\) is 100 %. However, this assumption can be relaxed, and exclusion limits derived as a function of the branching ratio of the \(\tilde{t}^{}_{1}\rightarrow c \tilde{\chi }^{0}_{1}\) decay, BR(\(\tilde{t}^{}_{1}\rightarrow c \tilde{\chi }^{0}_{1}\)), assuming that BR(\(\tilde{t}^{}_{1}\rightarrow c \tilde{\chi }^{0}_{1}\)) \(+\) BR(\(\tilde{t}^{}_{1}\rightarrow b f f' \tilde{\chi }^{0}_{1}\)) = 1. Two different scenarios, with \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})=10, 80\) GeV, are considered. The first compressed scenario is characterised by low-\(p_{\text {T}}\) stop decay products, and the set of signal regions which have sensitivity is the tc-M, independently of the decay of the stop. In the second scenario, the phase space available for the \(\tilde{t}^{}_{1}\) decay is larger, and the full set of tc-M, tc-C, t1L-bCa_low, t1L-bCa_med and WW-SR selections have different sensitivity, depending on BR(\(\tilde{t}^{}_{1}\rightarrow c \tilde{\chi }^{0}_{1}\)).

The cross-section limit is derived by combining the analyses discussed above. The SR giving the lowest expected exclusion CL\(_s\) for each signal model and for each value of BR(\(\tilde{t} \rightarrow c\tilde{\chi }^{0}_{1}\)) is chosen.

Fig. 5
figure 5

Upper limits on the stop pair production cross sections for different values of the BRs for the decays \(\tilde{t}^{}_{1}\rightarrow c \tilde{\chi }^{0}_{1}\) and \(\tilde{t}^{}_{1}\rightarrow ff'b\tilde{\chi }^{0}_{1}\). Signal points with \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})\) of 10 GeV (a) and 80 GeV (b) are shown. The limits quoted are taken from the best performing, based on expected exclusion CL\(_s\), signal regions from the tc-M, tc-C, t1L-bCa_low and WW analyses at each mass point. The blue line and corresponding hashed band correspond to the mean value and uncertainty on the production cross section of the stop as a function of its mass. The pink lines, whose darkness indicate the value of BR(\(\tilde{t} \rightarrow c\tilde{\chi }^{0}_{1}\)) according to the legend, indicate the observed limit on the production cross section

Figure 5 shows the result of these combinations. For \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})= 10\) GeV, the sensitivity is completely dominated by the tc-M signal regions, hence no significant dependence on BR(\(\tilde{t} \rightarrow c\tilde{\chi }^{0}_{1}\)) is observed. In this case, stop masses up to about 250 GeV are excluded. For \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})= 80\) GeV, the sensitivity is dominated by the tc-C signal regions at high values of BR(\(\tilde{t} \rightarrow c\tilde{\chi }^{0}_{1}\)). For lower values of BR(\(\tilde{t} \rightarrow c\tilde{\chi }^{0}_{1}\)), the “soft-lepton” and WW signal regions both become competitive, the latter yielding a higher sensitivity at smaller values of the stop mass. The maximum excluded stop mass ranges from about 180 GeV for BR\((\tilde{t} \rightarrow c\tilde{\chi }^{0}_{1})=25\,\%\) to about 270 GeV for BR\((\tilde{t} \rightarrow c\tilde{\chi }^{0}_{1}) = 100\,\%\).

\(\varvec{m}_{\varvec{W}}+\varvec{m}_{\varvec{b}} < \varvec{\Delta } \varvec{m} (\tilde{\varvec{t}}^{}_{\mathbf {1}}, \tilde{\varvec{\chi }}^{\mathbf {0}}_{\mathbf {1}}) < \varvec{m}_{\varvec{t}}\) In this case, the three-body decay of Fig. 2c is dominant. The signal regions that are sensitive to this decay are the dedicated signal region defined in the analysis selecting one-lepton final states (the t1L-3body) and the combination of several signal regions from the analysis selecting two-lepton final states, the t2L. The exclusion limits shown in Fig. 4 assume BR(\(\tilde{t}^{}_{1}\rightarrow bW\tilde{\chi }^{0}_{1}) = 1\). The WW signal regions are found to be sensitive to the kinematic region separating the three-body from the four-body stop decay region.

\(\varvec{\Delta } \varvec{m} (\tilde{\varvec{t}}^{}_{\mathbf {1}}, \tilde{\varvec{\chi }}^{\mathbf {0}}_{\mathbf {1}}) \sim \varvec{m}_{\varvec{t}}\) In this case, the neutralinos are produced with low \(p_{\text {T}} \), and the kinematic properties of the signal are similar to those of SM \(t\bar{t}\) production. Exclusion limits in this region were obtained by two analyses performing precision SM measurements. The first one is the measurement of the \(t\bar{t}\) inclusive production cross section \(\sigma _{t\bar{t}}\). Limits on \(\tilde{t}^{}_{1}\) pair production were already set in Ref. [65], which measured \(\sigma _{t\bar{t}}\) in the different-flavour, opposite-sign channel \(e\mu \). They were derived assuming a \(\tilde{t}^{}_{1}\) decay into an on-shell top quark, \(\tilde{t}^{}_{1}\rightarrow t \tilde{\chi }^{0}_{1}\). An extension of the limits into the three-body stop decay is discussed in Appendix B.1. For a massless neutralino, the analysis excludes stop masses from about 150 GeV to about \(m_t \). The limit deteriorates for higher neutralino masses, mainly because of the softer \(b\)-jet spectrum and the consequent loss in acceptance. The second analysis considered is that of the top quark spin correlation (SC) which considers SM \(t\bar{t}\) production with decays to final states containing two leptons (electrons or muons). The shape and normalisation of the distribution of the azimuthal angle between the two leptons is sensitive to the spin of the produced particles, hence it allows the analysis to differentiate between stop pair and \(t\bar{t}\) production. The limit obtained is shown in the bottom middle (dark orange) of the inset of Fig. 4. A small region of \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})\approx 180\) GeV is excluded with this measurement assuming a small neutralino mass.

\({\varvec{\Delta } \varvec{m} (\tilde{\varvec{t}}^{}_{\mathbf {1}}, \tilde{\varvec{\chi }}^{\mathbf {0}}_{\mathbf {1}})> \varvec{m}_{\varvec{t}}}\) In this kinematic region, the decay \(\tilde{t}^{}_{1}\rightarrow t \tilde{\chi }^{0}_{1}\) (see Fig. 2b) is dominant. The best results in this region are obtained by a statistical combination of the results of the multijet (t0L) and one-lepton (t1L) analyses. They both have dedicated signal regions targeting this scenario and the expected sensitivity is comparable for the two analyses. The number of required leptons makes the two signal regions mutually exclusive.

Fig. 6
figure 6

Combined exclusion limits assuming that the stop decays through \(\tilde{t}^{}_{1}\rightarrow t \tilde{\chi }^{0}_{1}\) with different branching ratios \(x\) and through \(\tilde{t}^{}_{1}\rightarrow b \tilde{\chi }^{\pm }_{1}\) with branching ratios \(1-x\). The limits assume \(m_{\tilde{\chi }^{\pm }_{1}} = 2 m_{\tilde{\chi }^{0}_{1}}\), and values of \(x\) from \(0\) to \(100\,\%\) are considered. For each branching ratio, the observed (with solid lines) and expected (with dashed lines) limits are shown

To maximise the sensitivity to the \(\tilde{t}^{}_{1}\rightarrow t \tilde{\chi }^{0}_{1}\) decays a statistical combination of the t0L and t1L signal regions is performed. The details of the combination are given in Appendix C and the final limit is shown in Fig. 4 by the largest shaded region (yellow). The expected limit on the stop mass is about 50 GeV higher at low \(m_{\tilde{\chi }^{0}_{1}}\) than in the individual analyses. The observed limit is increased by roughly the same amount and stop masses between 200 and 700 GeV are excluded for small neutralino masses.Footnote 5

A similar combination is performed to target a scenario where the stop can decay as \(\tilde{t}^{}_{1}\rightarrow t \tilde{\chi }^{0}_{1}\) with branching ratio \(x\) and as \(\tilde{t}^{}_{1}\rightarrow b \tilde{\chi }^{\pm }_{1}\) with branching ratio \(1-x\). Assuming gauge universality, the mass of the chargino is set to be twice that of the neutralino. Neutralino masses below 50 GeV are not considered, to take into account limits on the lightest chargino mass obtained at LEP [6670]. The exclusion limits are derived for \(x= 75, 50, 25\) and 0 %.Footnote 6 Regardless of the branching ratio considered, it is always assumed that \(m_{\tilde{t}^{}_{1}} > m_t + m_{\tilde{\chi }^{0}_{1}}\) and \(m_{\tilde{t}^{}_{1}} > m_b + m_{\tilde{\chi }^{\pm }_{1}}\), such that the two decays \(\tilde{t} \rightarrow t \tilde{\chi }^{0}_{1}\) and \(\tilde{t} \rightarrow b \tilde{\chi }^{\pm }_{1}\) are both kinematically allowed. A statistical combination, identical to the one described above, is used for \(x = 75\,\%\). For smaller values of \(x\), no combined fit is performed, as the sensitivity is dominated by the t1L analysis almost everywhere: rather either the t0L or the t1L analysis is used, depending which one gives the smaller expected CL\(_s\) value.

Figure 6 shows the result of the combination in the \(m_{\tilde{t}^{}_{1}}-m_{\tilde{\chi }^{0}_{1}}\) plane. The limit is improved, with respect to the individual analyses, by about 50 GeV for \(m_{\tilde{\chi }^{0}_{1}}= 50\) GeV and \(x = 75\,\%\). For other \(x\) values, the t1L analysis is used on the full plane, with the exception of the point at the highest stop mass for \(m_{\tilde{\chi }^{0}_{1}}= 50\) GeV at \(x=50\) and 25 %. Stop masses below 500 GeV are excluded for \(m_{\tilde{\chi }^{0}_{1}}<160\) GeV for any value of \(x\).

Fig. 7
figure 7

Summary of the ATLAS Run 1 searches for direct stop pair production in models where the decay mode \(\tilde{t}^{}_{1}\rightarrow b \tilde{\chi }^{\pm }_{1}\) with \(\tilde{\chi }^{\pm }_{1}\rightarrow W^{*} \tilde{\chi }^{0}_{1}\) is assumed with a branching ratio of 100 %. Various hypotheses on the \(\tilde{t}^{}_{1}\), \(\tilde{\chi }^{\pm }_{1}\), and \(\tilde{\chi }^{0}_{1}\) mass hierarchy are used. Exclusion limits at 95 % CL are shown in the \(\tilde{t}^{}_{1}-\tilde{\chi }^{0}_{1}\) mass plane. The dashed and solid lines show the expected and observed limits, respectively, including all uncertainties except the theoretical signal cross-section uncertainty (PDF and scale). Wherever not superseded by any \(\sqrt{s}=8\) TeV analysis, results obtained by analyses using 4.7 \({\mathrm{fb}^{-1}}\) of proton–proton collision data taken at \(\sqrt{s} = 7\) TeV are also shown, with the corresponding reference. The four plots correspond to interpretations of a the b0L and t1L soft-lepton analyses in two scenarios (\(\Delta m(\tilde{\chi }^{\pm }_{1}, \tilde{\chi }^{0}_{1}) = 5\ \mathrm{GeV}\) in light green and \(\Delta m(\tilde{\chi }^{\pm }_{1}, \tilde{\chi }^{0}_{1}) = 20\ \mathrm {GeV}\) in dark green), for a total of four limits; b the b0L, t1L and t2L analyses in scenarios with a fixed chargino mass \(m_{\tilde{\chi }^{\pm }_{1}} = 106\) GeV (dark green) and \(m_{\tilde{\chi }^{\pm }_{1}} = 150\) GeV (light green); c the t1L and t2L analyses in scenarios with \(m_{\tilde{\chi }^{\pm }_{1}} = 2 m_{\tilde{\chi }^{0}_{1}}\); d interpretations of the t1L, t2L and WW analyses in senarios with \(\Delta m \left( \tilde{t}^{}_{1}, \tilde{\chi }^{\pm }_{1}\right) = 10\ \mathrm {GeV}\)

4.2 Stop decays with a chargino in the decay chain

In the pMSSM, unless the higgsino–gaugino mass parameters are related by \(M_1 \ll \mu , M_2\), the mass difference between the lightest neutralino and the lightest chargino cannot be too large. The mass hierarchy \(m_{\tilde{\chi }^{0}_{1}} < m_{\tilde{\chi }^{\pm }_{1}} < m_{\tilde{t}^{}_{1}}\) is, hence, well motivated, leading to the decay chain shown in Fig. 2f.

If additional particles beside the stop and the lightest neutralino take part in the stop decay, the stop phenomenology quickly becomes complex. Even if the chargino is the only other relevant SUSY particle, the stop phenomenology depends on the chargino mass, on the stop left–right mixing, and on the composition of the neutralino and chargino in terms of bino, wino and higgsino states.

Figure 7 shows the exclusion limits obtained by the analyses listed in Table 1 and discussed in Appendix  if a branching ratio of 100 % for \(\tilde{t} \rightarrow b \tilde{\chi }^{\pm }_{1}\) is assumed. The exclusion limits are presented in a number of \(m_{\tilde{t}^{}_{1}}\)\(m_{\tilde{\chi }^{0}_{1}}\) planes, each characterised by a different hypothesis on the chargino mass. For all scenarios considered, the chargino is assumed to decay as \(\tilde{\chi }^{\pm }_{1}\rightarrow W^{(*)} \tilde{\chi }^{0}_{1}\), where the \((*)\) indicates a possibly virtual \(W\) boson.

\( \varvec{\Delta } \varvec{m} (\tilde{\varvec{\chi }}^{\varvec{\pm }}_{\mathbf {1}}, \tilde{\varvec{\chi }}^{\mathbf {0}}_{\mathbf {1}})= {\mathbf {5}, \mathbf {20}} \,{\mathbf {GeV}}\) This scenario assumes that the difference in mass between the lightest chargino and the neutralino is small (Fig. 7a), which is a rather common feature of models where, for example, the LSP has a large wino or higgsino component. Two hypotheses have been considered, with \(\Delta m(\tilde{\chi }^{\pm }_{1}, \tilde{\chi }^{0}_{1}) = 5\) GeV and \(\Delta m(\tilde{\chi }^{\pm }_{1}, \tilde{\chi }^{0}_{1}) = 20\) GeV. For both, the complete decay chain is \(\tilde{t}^{}_{1}\rightarrow b \tilde{\chi }^{\pm }_{1}\rightarrow b f f' \tilde{\chi }^{0}_{1}\), where the transverse momenta of the fermions \(f\) and \(f'\) depend on \(\Delta m(\tilde{\chi }^{\pm }_{1}, \tilde{\chi }^{0}_{1}) \) and on the stop mass, given the dependency on the chargino boost. If \(\Delta m(\tilde{\chi }^{\pm }_{1}, \tilde{\chi }^{0}_{1}) = 5\) GeV, the fermions have momenta too low to be efficiently reconstructed. The observed final state then consists of two \(b\)-jets and \(E_{\text {T}}^{\text {miss}} \). This final state is the direct target of the b0L signal regions. For \(\Delta m(\tilde{\chi }^{\pm }_{1}, \tilde{\chi }^{0}_{1}) = 20\) GeV, the signal efficiencies of the b0L signal regions decrease because of the lepton and jet veto applied. The t1L signal regions with soft leptons, instead, gain in sensitivity, profiting from the higher transverse momentum of the fermions from the off-shell \(W\) decay produced in the chargino decay.

\(\varvec{m}_{\tilde{\varvec{\chi }}^{\varvec{\pm }}_{\mathbf {1}}} =\mathbf {106}, \mathbf {150} \,{\mathbf {GeV}}\) This scenario (Fig. 7b) assumes a fixed chargino mass. The SR yielding the lowest expected exclusion CL\(_s\) for this scenario depends on the value of \(\Delta m(\tilde{\chi }^{\pm }_{1}, \tilde{\chi }^{0}_{1}) \). For \(\Delta m(\tilde{\chi }^{\pm }_{1}, \tilde{\chi }^{0}_{1}) <\) 20 GeV, the b0L signal regions provide the best sensitivity; for larger values of \(\Delta m(\tilde{\chi }^{\pm }_{1}, \tilde{\chi }^{0}_{1}) \), the t1L and t2L signal regions provide better sensitivity because of the same mechanism as in the \(\Delta m(\tilde{\chi }^{\pm }_{1}, \tilde{\chi }^{0}_{1}) = 5, 20\) GeV scenario above. The exclusion extends up to about 600 GeV for small values of \(\Delta m(\tilde{\chi }^{\pm }_{1}, \tilde{\chi }^{0}_{1}) \). A region of the parameter space with \(m_{\tilde{t}^{}_{1}}\) up to about 260 GeV and \(m_{\tilde{\chi }^{0}_{1}}\) between 100 GeV and \(m_{\tilde{\chi }^{\pm }_{1}}\) is not yet excluded.

\(\varvec{m}_{\tilde{\varvec{\chi }}^{\varvec{\pm }}_{\mathbf {1}}} = {\mathbf {2}\varvec{m}}_{\tilde{\varvec{\chi }}^{\mathbf {0}}_{\mathbf {1}}}\) Inspired by gauge-universality considerations, the third scenario (Fig. 7c) is characterised by a relatively large \(\Delta m(\tilde{\chi }^{\pm }_{1}, \tilde{\chi }^{0}_{1}) \). The t2L signal regions dominate the sensitivity for \(m_{\tilde{t}^{}_{1}} \sim m_{\tilde{\chi }^{\pm }_{1}}\). The sensitivity of the dedicated t1L-bC is dominant in a large region of the plane, and determines the exclusion reach for moderate to large values of \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})\).

\(\varvec{\Delta } \varvec{m} (\tilde{\varvec{t}}^{}_{\mathbf {1}}, \tilde{\varvec{\chi }}^{\varvec{\pm }}_{\mathbf {1}}) = {\mathbf {10}}\, {\mathbf {GeV}}\) The fourth scenario (Fig. 7d) assumes a rather compressed \(\tilde{t}^{}_{1}-\tilde{\chi }^{\pm }_{1}\) spectrum. The region at low \(m_{\tilde{t}^{}_{1}}\) and large \(m_{\tilde{\chi }^{0}_{1}}\) is characterised by low mass separations between all particles involved, and it is best covered by the t1L-bCc_diag, the t1L soft lepton, and the WW signal regions. At larger values of the stop mass, the leptons emitted in the \(\tilde{\chi }^{\pm }_{1}\) decay have larger \(p_{\text {T}} \), and the t2L signal regions provide the best sensitivity.

Fig. 8
figure 8

Exclusion limits assuming that the stop decays through \(\tilde{t}^{}_{1}\rightarrow b + \tilde{\chi }^{\pm }_{1}\rightarrow b + W^{(*)} + \tilde{\chi }^{0}_{1}\) with branching ratio of 100 % assuming a fixed stop mass of \(m_{\tilde{t}^{}_{1}} = 300\) GeV. The region below the purple line and above the blue line, indicated by a light shading, is excluded

\(\varvec{m}_{\tilde{\varvec{t}}^{}_{\mathbf {1}}}=\mathbf {300} \,\mathbf {GeV}\) The final scenario considered is one where the stop mass is fixed at 300 GeV, and the exclusion limits are expressed in the \(m_{\tilde{\chi }^{\pm }_{1}}\)\(m_{\tilde{\chi }^{0}_{1}}\) plane. In the case of the compressed scenario, corresponding to a small mass difference \(\Delta m(\tilde{\chi }^{\pm }_{1}, \tilde{\chi }^{0}_{1})\), the fermions from the \(W^{(*)}\) decay can escape detection and only the two \(b\)-jets and \(E_{\mathrm {T}}^{\mathrm {miss}}\) would be identified in the final state. Thus, the b0L signal regions are expected to have a large sensitivity in this case, while for larger values of \(\Delta m(\tilde{\chi }^{\pm }_{1}, \tilde{\chi }^{0}_{1})\), the lepton can be observed, yielding a final-state signature investigated by the t1L soft-lepton signal region. A combination of the b0L and t1L signal regions is performed by choosing, for each point of the plane, the SR giving the lowest CL\(_s\) for expected exclusion. The result, reported in Fig. 8, shows that a large portion of the plane is excluded, with the exception of a region where the mass separations between the \(\tilde{t}^{}_{1}\), the \(\tilde{\chi }^{\pm }_{1}\) and the \(\tilde{\chi }^{0}_{1}\) are small.

Summarising, in the simplified models with \(\tilde{t}^{}_{1}\rightarrow b \tilde{\chi }^{\pm }_{1}\rightarrow b W^{(*)} \tilde{\chi }^{0}_{1}\), stop masses up to 450–600 GeV are generally excluded. Scenarios where \(\Delta m(\tilde{t}^{}_{1}, \tilde{\chi }^{0}_{1}) \) is small are particularly difficult to exclude and in these compressed scenarios, stop masses as low as 200 GeV are still allowed (Fig. 7b). A small unexcluded area is also left for a small region around \((m_{\tilde{t}^{}_{1}},m_{\tilde{\chi }^{\pm }_{1}},m_{\tilde{\chi }^{0}_{1}})= \left( 180,100,50\right) \) GeV (Fig. 7c), where the sensitivity of the analyses is poor because the signal kinematics are similar to SM \(t\bar{t}\) production.

4.3 Limits on pair production of \(\tilde{t}^{}_{2}\)

Although the pair production of \(\tilde{t}^{}_{1}\) has a cross section larger than that of \(\tilde{t}^{}_{2}\), and although the decay patterns of the two particles can be similar, it can be convenient to search for the latter in regions where the sensitivity to the former is limited. This is the case, for example, in the region where \(\Delta m(\tilde{t}^{}_{1},\tilde{\chi }^{0}_{1})\sim m_t \) of Fig. 4, where the separation of \(\tilde{t}^{}_{1}\) pair production from SM top quark pair production is difficult. The t2t1Z and t2t1h analyses are designed to detect \(\tilde{t}^{}_{2}\) pair production in this region of the \(m_{\tilde{t}^{}_{1}}-m_{\tilde{\chi }^{0}_{1}}\) plane, followed by the decays \(\tilde{t}^{}_{2}\rightarrow \tilde{t}^{}_{1}Z\) and \(\tilde{t}^{}_{2}\rightarrow \tilde{t}^{}_{1}h\). The Higgs boson \(h\) is assumed to have a mass of 125 GeV and SM branching ratios.

The exclusion limits were first derived in a scenario in which the pair-produced \(\tilde{t}^{}_{2}\) decays either through \(\tilde{t}^{}_{2}\rightarrow Z \tilde{t}^{}_{1}\) with a branching ratio of 100 % (Fig. 3a), or through \(\tilde{t}^{}_{2}\rightarrow h \tilde{t}^{}_{1}\) (again with a branching ratio of 100 %; Fig. 3b). In both cases, the \(\tilde{t}^{}_{1}\) is assumed to decay through \(\tilde{t}^{}_{1}\rightarrow t \tilde{\chi }^{0}_{1}\), and its mass is set to be 180 GeV above that of the neutralino (assumed to be the LSP), which is the region not excluded in Fig. 4. The final state contains two top quarks, two neutralinos, and either two \(Z\) or two \(h \) bosons.

Fig. 9
figure 9

Exclusion limits at 95 % CL in the scenario where \(\tilde{t}^{}_{2}\) pair production is assumed, followed by the decay \(\tilde{t}^{}_{2}\rightarrow Z \tilde{t}^{}_{1}\) (blue) or \(\tilde{t}^{}_{2}\rightarrow \tilde{t}^{}_{1}h \) (red) and then by \(\tilde{t}^{}_{1}\rightarrow t \tilde{\chi }^{0}_{1}\) with a branching ratio of 100 %, as a function of the \(\tilde{t}^{}_{2}\) and \(\tilde{\chi }^{0}_{1}\) mass. The \(\tilde{t}^{}_{1}\) mass is determined by the relation \(m_{\tilde{t}^{}_{1}} - m_{\tilde{\chi }^{0}_{1}} = 180\) GeV. The dashed lines indicate the expected limit and the solid lines indicate the observed limit

Figure 9 shows the exclusion limits for the t2t1h and the t2t1Z analyses. In both cases, a limit on \(m_{\tilde{t}^{}_{2}}\) is set at about 600 GeV for a massless neutralino. In the case of a \(\tilde{t}^{}_{2}\) decay through a Higgs boson, the limit covers neutralino masses lower than in the case of the decay through a \(Z\) boson.

Fig. 10
figure 10

Exclusion limits as a function of the \(\tilde{t}^{}_{2}\) branching ratio for \(\tilde{t}^{}_{2}\rightarrow \tilde{t}^{}_{1}h \), \(\tilde{t}^{}_{2}\rightarrow \tilde{t}^{}_{1}Z\) and \(\tilde{t}^{}_{2}\rightarrow t \tilde{\chi }^{0}_{1}\). The blue, red and green limit refers to the t2t1Z, t2t1h and combination of t0L and t1L analyses respectively. The limits are given for three different values of the \(\tilde{t}^{}_{2}\) and \(\tilde{\chi }^{0}_{1}\) masses

The assumption on the branching ratio of the \(\tilde{t}^{}_{2}\) has also been relaxed, and limits have been derived assuming that the three decays \(\tilde{t}^{}_{2}\rightarrow Z \tilde{t}^{}_{1}\), \(\tilde{t}^{}_{2}\rightarrow h \tilde{t}^{}_{1}\) and \(\tilde{t}^{}_{2}\rightarrow t \tilde{\chi }^{0}_{1}\) (Fig. 3c) are the only possible ones. The limits are shown in Fig. 10 as a function of the three BRs, for different combinations of the \(\tilde{t}^{}_{2}\) and \(\tilde{\chi }^{0}_{1}\) masses. Three analyses have been considered: the t2t1Z, t2t1h and the combination of the t0L and t1L discussed in Sect. 4.1.Footnote 7 The three analyses have complementary sensitivities. Together, they exclude \(\tilde{t}^{}_{2}\) pair production with a mass of 350 and 500 GeV for \(m_{\tilde{\chi }^{0}_{1}} = 20\) GeV. A non-excluded region appears for \(m_{\tilde{t}^{}_{2}} = 500\) GeV if larger \(\tilde{\chi }^{0}_{1}\) masses are considered.

4.4 Sbottom decays

Under the assumption that no supersymmetric particle takes part in the sbottom decay apart from the lightest neutralino, the sbottom decays as \(\tilde{b}^{}_{1}\rightarrow b \tilde{\chi }^{0}_{1}\) with a branching ratio of 100 % (Fig. 2a). The final state arising from sbottom pair production hence contains two \(b\)-jets and \(E_{\mathrm {T}}^{\mathrm {miss}}\). The b0L signal regions were explicitly optimised to be sensitive to this scenario. In case of a mass degeneracy between the sbottom and the neutralino, the general consideration that the monojet-like tc-M selection is almost insensitive to the details of the decay of the produced particles still holds: the tc-M signal regions offer the best sensitivity for scenarios where \(m_{\tilde{b}^{}_{1}} \sim m_{\tilde{\chi }^{0}_{1}}\).

Fig. 11
figure 11

Observed (solid lines) and expected (dashed lines) 95 % CL limits on sbottom pair production where the sbottom is assumed to decay as \(\tilde{b}^{}_{1}\rightarrow b \tilde{\chi }^{0}_{1}\) with a branching ratio of 100 %. The purple lines refer to the limit of the tc analysis, while the blue lines refer to the b0L analysis

Figure 11 shows the limits of the tc and b0L analyses on the \(m_{\tilde{b}^{}_{1}} - m_{\tilde{\chi }^{0}_{1}}\) plane. The monojet-like (tc-M) SRs exclude models up to a value of \(m_{\tilde{b}^{}_{1}} \sim m_{\tilde{\chi }^{0}_{1}} \sim 280\ \mathrm{GeV}\). Sbottom masses are excluded up to about 600 GeV for neutralino masses below about 250 GeV.

If other supersymmetric particles enter into the decay chain, then multiple decay channels would be allowed. Similarly to the stop, the case in which other neutralinos or charginos have a mass below the sbottom is well motivated. The branching ratios of the sbottom to the different decay channels depend on the supersymmetric particle mass hierarchy, on the mixing of the left–right components of the sbottom, and on the composition of the charginos and neutralinos in terms of bino, wino, and higgsino states.

An exclusion limit is derived under the assumption that the sbottom decays with a branching ratio of 100 % into \(\tilde{b}^{}_{1}\rightarrow t \tilde{\chi }^{\pm }_{1}\) (Fig. 2g). The chargino is assumed to decay through \(\tilde{\chi }^{\pm }_{1}\rightarrow W^{(*)} \tilde{\chi }^{0}_{1}\) with a branching ratio of 100 %. The final state is a complex one, and offers many handles for background rejection: it potentially contains up to ten jets, two \(b\)-jets, and up to four leptons. The limits of Fig. 12a, shown in the \(m_{\tilde{b}^{}_{1}}-m_{\tilde{\chi }^{0}_{1}}\) plane, were obtained by using the three-lepton signal regions SS3L, either fixing the mass of the neutralino to \(m_{\tilde{\chi }^{0}_{1}}=60\) GeV or by making the assumption that \(m_{\tilde{\chi }^{\pm }_{1}}= 2 m_{\tilde{\chi }^{0}_{1}}\). In the two scenarios considered, sbottom masses up to about 440 GeV are excluded, with a mild dependency on the neutralino mass.

Fig. 12
figure 12

Exclusion limits at 95 % CL for a scenario where sbottoms are pair produced and decay as a \(\tilde{b}^{}_{1}\rightarrow t \tilde{\chi }^{\pm }_{1}\) with a BR of 100 % or b \(\tilde{b}^{}_{1}\rightarrow b\tilde{\chi }^{0}_{2}\) with a BR of 100 %. The signal regions used in a are the SS3L, and two different models are considered: a fixed neutralino mass of 60 GeV (in purple) or \(m_{\tilde{\chi }^{\pm }_{1}} = 2 m_{\tilde{\chi }^{0}_{1}}\) (in blue). The limits are shown in the \(m_{\tilde{b}^{}_{1}}\)\(m_{\tilde{\chi }^{\pm }_{1}}\) plane. The signal regions used in b are the g3b-SR-0j. A fixed neutralino mass of 60 GeV is assumed, and the limit is shown in the \(m_{\tilde{b}^{}_{1}}\)\(m_{\tilde{\chi }^{0}_{2}}\) plane

The last case considered is one where the pair-produced sbottoms decay through \(\tilde{b}^{}_{1}\rightarrow b \tilde{\chi }^{0}_{2}\), followed by the decay of \(\tilde{\chi }^{0}_{2}\) into a \(\tilde{\chi }^{0}_{1}\) and a SM-like Higgs boson \(h \) (Fig. 2h). The final state contains up to six \(b\)-jets, four of which are produced by the two Higgs bosons decays. Since multiple \(b\)-jets are present in the final state, the three-\(b\)-jets signal regions (g3b) are used to place limits in this model.

The limit, derived as a function of \(m_{\tilde{b}^{}_{1}}\) and \(m_{\tilde{\chi }^{0}_{2}}\) assuming a fixed neutralino mass of \(\tilde{\chi }^{0}_{1}= 60\) GeV, is shown in Fig. 12b. Sbottom masses between about 300 and 650 GeV are excluded for \(\tilde{\chi }^{0}_{2}\) masses above 250 GeV.

5 Interpretations in pMSSM models

The interpretation of the results in simplified models is useful to assess the sensitivity of each signal region to a specific topology. However, this approach fails to test signal regions on the complexity of the stop and sbottom phenomenology that appears in a realistic SUSY model. To this extent, the signal regions are used to derive exclusion limits in the context of specific pMSSM models.

The pMSSM [57] is obtained from the more general MSSM by making assumptions based on experimental results:

  • No new source of CP violation beyond the Standard Model. New sources of CP violation are constrained by experimental limits on the electron and neutron electric dipole moments.

  • No flavour-changing neutral currents. This is implemented by requiring that the matrices for the sfermion masses and trilinear couplings are diagonal.

  • First- and second-generation universality. The soft-SUSY-breaking mass parameters and the trilinear couplings for the first and second generation are assumed to be the same based on experimental data from, e.g., the neutral kaon system [71].

With the above assumptions, and with the choice of a neutralino as the LSP, the pMSSM adds 19 free parameters on top of those of the SM. The complete set of pMSSM parameters is shown in Table 2.

Table 2 Description of the 19 additional parameters of the pMSSM model with a neutralino LSP

A full assessment of the ATLAS sensitivity to a scan of the 19-parameters space has been performed in Ref. [72]. Here, a set of additional hypotheses are made, to focus on the sensitivity to a specific, well-motivated set of models with enhanced third generation squark production:

  • The common masses of the first- and second-generation squarks have been set to a multi-TeV scale, making these quarks irrelevant for the processes studied at the energies investigated in this paper. This choice is motivated by the absence of any signal from squark or gluino production in dedicated SUSY searches performed by the ATLAS [62, 63, 7376] and CMS [29, 34, 7782] collaborations.

  • All slepton mass parameters have been set to the same scale as the first- and second-generation squarks. This choice has no specific experimental or theoretical motivation, and should be regarded as an assumption.

  • A decoupling limit with \(M_A =3\) TeV and large \(\tan \beta \) values (\(\tan \beta > 15\)) has been assumed. This is partially motivated by results of the LHC searches for higher mass Higgs boson states [83, 84].

  • For \(\tan \beta \gg 1\), the Higgs boson mass depends heavily on the product of the stop-mass parameters \(M_S = \sqrt{m_{\tilde{t}^{}_{1}} m_{\tilde{t}^{}_{2}}}\) and the mixing between the left- and right-handed states \(X_t = A_t - \mu /\tan \beta \) [85]. The stop sector is therefore completely fixed, given the Higgs boson mass, the value of \(X_t\) and one of the two stop mass parameters.Footnote 8

  • The trilinear couplings \(A_b\) in the sbottom sector are found to have limited impact on the phenomenology, and are therefore set to zero.

  • The gluino mass parameter \(M_3\) is set such to evade LHC constraints on gluino-pair production.

These assumptions reduce the number of additional free parameters of the model to the mass parameters of the electroweak sector (\(\mu , M_1, M_2\)) and two of the three third-generation squark mass parameters (\(m_{\tilde{q}L3}, m_{\tilde{t} R}, m_{\tilde{b} R}\)). All the assumptions made either have a solid experimental basis, or are intended to simplify the interpretation in terms of direct production of stops and sbottoms (as, for example, the assumption on the slepton mass parameters).

Three types of models have been chosen, that, by implementing in different ways constraints arising from naturalness arguments and the dark-matter relic density measurement, further reduce the number of parameters to be scanned over. They are described below, and summarised in Table 3 together with additional information on the most relevant production and decay channels.

Table 3 Details of parameters scanned in the three pMSSM models used for interpretations. The settings of additional parameters that are fixed for each model are also given together with the main production and decay channels targeted

Naturalness-inspired pMSSM The model is inspired by naturalness criteria, which require a value of \(\mu \) in the range of a few hundred GeV, favour stop masses below one TeV, place weak constraints on the gluino mass and give no constraints on the mass of other SUSY particles [86]. The exclusion limits are determined as a function of the higgsino mass parameter \(\mu \) and the left-handed squark mass parameter \(m_{\tilde{q}L3}\). The parameter \(m_{\tilde{q}L3}\) is scanned in the range \(350 \ \mathrm{GeV}< m_{\tilde{q}L3}< 900\ \mathrm{GeV}\). The parameter \(\mu \) is scanned in the range \(100 \ \mathrm{GeV}< \mu < m_{\tilde{q}L3}-150\ \mathrm{GeV}\), where the lower bound is determined by limits on the chargino mass arising from LEP [6670]. The right-handed stop mass parameter \(m_{\tilde{t} R}\) and the stop mixing parameter \(X_t\) are determined by choosing the maximal mixing scenario \(X_t/M_S = \sqrt{6}\) and by the requirement of having a Higgs boson mass of about 125 GeV. The other squark and slepton masses, as well as the bino mass parameter \(M_1\), are set to 3 TeV. The wino mass parameter \(M_2\) is set such that \(M_2 = 3\mu \). The gluino mass parameter \(M_3\) is set to 1.7 TeV.

With this choice of the model parameters, the spectrum is characterised by two light neutralinos \(\left( \tilde{\chi }^{0}_{1},\tilde{\chi }^{0}_{2}\right) \) and one chargino \(\left( \tilde{\chi }^{\pm }_{1}\right) \), all with masses of the order of \(\mu \), a light \(\tilde{b}^{}_{1}\) with a mass of the order of \(m_{\tilde{q}L3}\), and a light \(\tilde{t}^{}_{1}\) with mass of the order of \(m_{\tilde{q}L3}\) up to \(m_{\tilde{q}L3}\sim 700\) GeV (the constraint on \(M_S\) does not allow the mass of \(\tilde{t}^{}_{1}\) to increase beyond about 650 GeV). The production processes considered are direct pair production of \(\tilde{b}^{}_{1}\) and \(\tilde{t}^{}_{1}\) with similar masses. Because of the abundance of light higgsino states, many different decays can occur.

Well-tempered neutralino pMSSM The models are designed to loosely satisfy dark-matter thermal-relic density constraints (\(0.09 < \Omega _{\mathrm {c}} h^2 < 0.15\), where \(h\) is the Hubble constant), while keeping fine tuning (defined as in Ref. [87]) to less than 1 %. The exclusion limits are determined as a function of \(M_1\) and \(m_{\tilde{q}L3}\), or \(M_1\) and \(m_{\tilde{t} R}\), with \(\mu \sim - M_1\) in both cases to satisfy the dark-matter constraints through the presence of well-tempered neutralinos [88]. The constraints on the Higgs boson mass are satisfied in a way similar to the naturalness-inspired pMSSM model above. All other parameters are the same as in the naturalness-inspired pMSSM model. These models tend to have three neutralinos and two charginos with masses lower than \(\tilde{t}^{}_{1}\) or \(\tilde{b}^{}_{1}\), giving rise to a diverse phenomenology.

\(\mathbf{h}{/}\mathbf{Z}\) -enriched pMSSM These models are defined such that Higgs and \(Z\) bosons are produced abundantly in the SUSY particles’ decay chains. The assumption of \(M_1 = 100\) GeV ensures the presence of a bino-like neutralino LSP, while \(M_3 = 2.5\) TeV ensures that direct gluino production is highly suppressed compared to third-generation squark production. Two sets of models have been defined: in the first one, \(\mu \) and the right-handed sbottom mass parameter \(m_{\tilde{b} R}\) are scanned while keeping \(M_2 = \mu \), \(m_{\tilde{q}L3}= 1.2\) TeV, \(m_{\tilde{t} R}= 1.6\) TeV; in the second one, \(\mu \) and \(m_{\tilde{q}L3}\) are scanned while keeping \(M_2 = 1\) TeV, \(m_{\tilde{b} R}= 3\) TeV, \(m_{\tilde{t} R}= 2\) TeV. The former is dominated by sbottom pair production, while both sbottom and stop pair production are relevant for the latter. Stop mixing parameters are chosen with maximal mixing to satisfy Higgs boson mass constraints. In these models, the decays of the third generation squarks into the heavier neutralino states (\(\tilde{\chi }^{0}_{2}\) and \(\tilde{\chi }^{0}_{3}\)) are followed by decays to the lightest neutralino with the emission of a \(Z\) or a \(h\) boson. Typically the \(\tilde{\chi }^{0}_{2}\) (\(\tilde{\chi }^{0}_{3}\)) decays into a \(Z\) boson 30 % (85 %) of the times, and into a Higgs boson 70 % (15 %) of the times. The subsequent decays of the Higgs boson into \(b\)-quark pairs (happening with the same branching ratio as in the Standard Model) lead to final states rich in \(b\)-jets.

Exclusion limits for these pMSSM models are determined by combining many of the SRs defined for the searches discussed in this paper (t0L, t1L, tb,Footnote 9 t2t1Z, g3b, tc). For each set of parameters the individual 95 % CL expected limit is evaluated. The combined exclusion contour is determined by choosing, for each model point, the signal region having the smallest expected CL\(_s\) value of the test statistic for the signal-plus-background hypothesis.

Fig. 13
figure 13

Expected and observed 95 % CL exclusion limits for the naturalness-inspired set of pMSSM models from the combination t0L, t1L and tb analyses using the signal region yielding the smallest CL\(_s\) value for the signal-plus-background hypothesis. The dashed black line indicates the expected limit, and the yellow band indicates the \(\pm 1\sigma \) uncertainties, which include all uncertainties except the theoretical uncertainties in the signal. The red solid line indicates the observed limit, and the red dotted lines indicate the sensitivity to \(\pm 1\sigma \) variations of the signal theoretical uncertainties. The dashed and dotted grey lines indicate a constant value of the stop and sbottom masses, while the dashed light-blue line indicates a constant value of the neutralino mass

Figure 13 shows the exclusion limit for the naturalness-inspired set of pMSSM models based on the t0L, t1L and tb analyses. The t0L and t1L analyses have a similar expected sensitivity. These SRs were optimised assuming a 100 % BR for \(\tilde{t}^{}_{1}\rightarrow t \tilde{\chi }^{0}_{1}\) or \(\tilde{t}^{}_{1}\rightarrow b\tilde{\chi }^{\pm }_{1}\), while for these pMSSM models, the stop decays to \(\tilde{t}^{}_{1}\rightarrow t \tilde{\chi }^{0}_{1}\), \(\tilde{t}^{}_{1}\rightarrow b \tilde{\chi }^{\pm }_{1}\) and \(\tilde{t}^{}_{1}\rightarrow b \tilde{\chi }^{0}_{2}\) with similar branching ratios (and the sbottom to both \(\tilde{b}^{}_{1}\rightarrow b \tilde{\chi }^{0}_{1}\) and \(\tilde{b}^{}_{1}\rightarrow t \tilde{\chi }^{\pm }_{1}\)). The tb signal regions, discussed in detail in Appendix B.2.3, are designed to be sensitive to final states containing a top quark, a \(b\)-quark and missing transverse momentum and address such mixed-decay scenarios by requiring a lower jet multiplicity.

The signal regions that dominate the sensitivity are the tb, t0L-SRC1 and t1L-bCd_bulk at low values of \(m_{\tilde{q}L3}\), and tb, t0L-SRA1, t0L-SRA2 and t1L-tNbC_mix at intermediate and high values of \(m_{\tilde{q}L3}\). The excluded region for models with \(m_{\tilde{q}L3}\sim 900\) GeV and \(\mu \sim 150\) GeV is due to the saturation of \(m_{\tilde{t}^{}_{1}}\) at high \(m_{\tilde{q}L3}\) values: to satisfy the Higgs boson mass constraint requires \(M_S \sim 800\) GeV, hence \(m_{\tilde{t}^{}_{1}}\) at \(m_{\tilde{q}L3}\sim 900\) GeV is smaller than that at \(m_{\tilde{q}L3}\sim 800\) GeV. The large fluctuations of the observed limit with respect to the expected one are due to transitions between different signal regions providing the best expected exclusion in different regions of the plane.

Fig. 14
figure 14

Expected and observed 95 % CL exclusion limits for the pMSSM model with well-tempered neutralinos as a function of \(M_1\) and a \(m_{\tilde{q}L3}\) or b \(m_{\tilde{t}^{}_{\mathrm {R}}}\). The limit of a is obtained as the combination of the t0L, t1L, tb and SS3L analyses, while the t0L analysis is used for b. The signal region yielding the smallest CL\(_s\) value for the signal-plus-background hypothesis is used for each point. The dashed black line indicates the expected limit, and the yellow band indicates the \(\pm 1\sigma \) uncertainties, which include all uncertainties except the theoretical uncertainties in the signal. The red solid line indicates the observed limit, and the red dotted lines indicate the sensitivity to \(\pm 1\sigma \) variations of the signal theoretical. The dashed and dotted grey lines indicate a constant value of the stop and sbottom masses, while the dashed light-blue line indicates a constant value of the neutralino mass

Figure 14a, b show the exclusion limit obtained for the set of pMSSM models with well-tempered neutralinos as a function of \(m_{\tilde{q}L3}\) and \(m_{\tilde{t} R}\), respectively. In both cases, the exclusion is largely dominated by the t0L analysis. For Fig. 14a, the signal region dominating the sensitivity at low \(m_{\tilde{q}L3}\) is t0L-SRC1, while at higher \(m_{\tilde{q}L3}\) values t0L-SRA1 and t0L-SRA2 dominate the sensitivity. The drop in sensitivity at \(m_{\tilde{q}L3}= 410\ \mathrm{GeV}\), \(M_1 = 260\ \mathrm{GeV}\) is due to the opening of the \(\tilde{t}^{}_{1}\rightarrow t \tilde{\chi }^{0}_{2}\) and \(\tilde{t}^{}_{1}\rightarrow t \tilde{\chi }^{0}_{3}\) transition, kinematically suppressed for smaller values of the difference \(m_{\tilde{q}L3}- M_1\). Such decays introduce more intermediate states in the decay, effectively reducing the transverse momenta of the final state objects. The large fluctuations of the observed limit are again due to transitions between different signal regions. For Fig. 14b, the sensitivity is entirely dominated by the various t0L-SRC. The difference in sensitivity between these two scenarios is due to the presence of both a stop and a sbottom for small \(m_{\tilde{q}L3}\), while only a stop is present for low values of \(m_{\tilde{t} R}\).

Fig. 15
figure 15

Expected and observed 95 % CL exclusion limits for the set of \(h/Z\)-enriched pMSSM models as a function of \(\mu \) and a \(m_{\tilde{q}L3}\) and b \(m_{\tilde{b}^{}_{\mathrm {R}}}\). The limit of a is obtained as the combination of the t0L, g3b, t2t1Z and SS3L analyses, while the t0L, t2t1Z and tb analysis are used for b. The signal region yielding the smallest CL\(_s\) value for the signal-plus-background hypothesis is used for each point. The dashed black line indicates the expected limit, and the yellow band indicates the \(\pm 1\sigma \) uncertainties, which include all uncertainties except the theoretical uncertainties in the signal. The red solid line indicates the observed limit, and the red dotted lines indicate the sensitivity to \(\pm 1\sigma \) variations of the signal theoretical. The dashed and dotted grey lines indicate a constant value of the stop and sbottom masses, while the dashed light-blue line indicates a constant value of the neutralino mass

Finally, Fig. 15a, b show the exclusion limit obtained for the set of \(h/Z\)-enriched pMSSM models. These models yield large \(b\)-jet multiplicities to the final state through direct sbottom decays, top-quark decays and \(\tilde{\chi }^{0}_{2}\rightarrow h/Z \tilde{\chi }^{0}_{1}\). The exclusion is dominated by the t0L and g3b analyses for Fig. 15a and by and the t0L analysis for Fig. 15b.

More informations about the limits obtained, including the SLHA files for the points mentioned in Table 3, can be found in Refs. [89] and [90].

6 Conclusions

The search programme of the ATLAS collaboration for the direct pair production of stops and sbottoms is summarised and extended by new analyses targeting scenarios not optimally covered by previously published searches. The paper is based on 20 \({\mathrm{fb}^{-1}}\) of proton–proton collisions collected at the LHC by ATLAS in 2012 at a centre-of-mass energy \(\sqrt{s}\) = 8 TeV. Exclusion limits in the context of simplified models are presented. In general, stop and sbottom masses up to several hundred GeV are excluded, although the exclusion limits significantly weaken in the presence of compressed SUSY mass spectra or multiple allowed decay chains. Three classes of pMSSM models, based on general arguments of Higgs boson mass naturalness and compatibility with the observed dark-matter relic density have also been studied and exclusion limits have been set. Large regions of the considered parameter space are excluded.

We thank CERN for the very successful operation of the LHC, as well as the support staff from our institutions without whom ATLAS could not be operated efficiently.