Present status and future tests of the higgsino-singlino sector in the NMSSM

The light higgsino-singlino scenario of the NMSSM allows to combine a naturally small μ parameter with a good dark matter relic density. Given the new constraints on spin-dependent and spin-independent direct detection cross sections in 2016 we study first which regions in the plane of chargino- and LSP-masses below 300 GeV remain viable. Subsequently we investigate the impact of searches for charginos and neutralinos at the LHC, and find that the limits from run I do not rule out any additional region in this plane. Only the HL-LHC at 3000 fb−1 will test parts of this plane corresponding to higgsino-like charginos heavier than 150 GeV and relatively light singlinos, but notably the most natural regions with lighter charginos seem to remain unexplored.


Introduction
Despite considerable efforts the ATLAS and CMS collaborations have not (yet?) discovered supersymmetric particles at the LHC [1, 2]. The most stringent limits are on the masses of squarks and gluinos with strong interactions, but electroweakly interacting supersymmetric particles (charginos χ ± i and neutralinos χ 0 i ) can and have also been searched for. Among the electroweakly interacting supersymmetric particles are the higgsinos, the fermionic superpartners of the two scalar Higgs doublets H u and H d required in supersymmetry (Susy). The higgsinos include a chargino with a Dirac mass µ, and two neutral Majorana fermions which are part of the neutralinos. Charginos have been searched for at LEP, and masses below ∼ 100 GeV have been ruled out [3,4] implying a corresponding lower bound on |µ|.
µ is a supersymmetric mass term and generates a positive mass squared term µ 2 in the scalar potential for both SU(2) doublets H u and H d , both of which must have non-vanishing vacuum expectation values (vevs) in order to generate up-and down-type quark masses. At least for H u , the positive mass squared term µ 2 must be cancelled by a negative soft SUSY breaking mass term m 2 Hu such that µ 2 + m 2 Hu < 0 leading to −µ 2 − m 2 Hu ≈ M 2 Z /2. Such a cancellation becomes unnatural if each of these terms were much larger than M 2 Z , i.e. > ∼ (300GeV) 2 . This argument applies both in the MSSM [5][6][7] (see e.g. [8][9][10][11]) and in the NMSSM [12,13] (see e.g. [14,15]), where µ ≡ λ S is generated by the vev S of a scalar singlet S and λ is a higgsino-S Yukawa coupling.
The lightest neutralino, stable if R-parity is conserved and if it is the lightest supersymmetric particle (LSP), is a candidate for dark matter. Besides the higgsinos, additional neutralinos are neutral electroweak gauginos (bino and the neutral wino) and, in the NMSSM, the singlino, the supersymmetric partner of the scalar S. If one assumes Grand Unification of the gaugino masses lower bounds on the gluino mass M 3 imply also lower bounds on the bino and wino masses M 1 and M 2 , respectively. This assumption implies M 1 : M 2 : M 3 ≈ 1 : 2 : 6 depending somewhat on threshold corrections to the running masses and radiative corrections to pole masses. Then, if M 3 > ∼ 2 TeV (or more) is confirmed in the future, it is natural for the wino and bino masses M 2 , M 1 to be larger than the higgsino mass parameter µ.

JHEP02(2017)051
In the MSSM, the neutral higgsinos (nearly mass degenerate with the higgsino-like charginos) would then play the rôle of the LSP. However, a nearly pure higgsino LSP is known to be an imperfect candidate for dark matter: unless very heavy with a mass beyond 1 TeV, it annihilates too fast in the early universe implying a too small relic density [16]. (The situation might be remedied introducing additional axions as the dominant component of dark matter in the MSSM [9,11]. Alternatively, a non-thermal higgsino production in the early universe can be invoked; this scenario is severly constrained by limits from direct dark matter detection experiments [17].) In the NMSSM it is natural for the singlino to be quite light; the singlino (Majorana) mass parameter 2κ S (where κ is a singlino-Singlet Yukawa coupling) might even vanish for κ → 0 in which case a small singlino mass results only from mixing with the neutral higgsinos. Still, such a mostly singlino-like LSP is good dark matter candidate [18] since its relic density can be reduced to the observed value via various processes like a mostly singletlike pseudoscalar in the s-channel, chargino exchange in the t-channel or co-annihilation. Furthermore a mostly singlino-like LSP can easily satisfy limits from direct and indirect dark matter detection experiments; its spin-independent direct detection cross section can even fall below the neutrino floor [19,20].
It is the purpose of the present paper to investigate in how far this attractive scenario within the NMSSM is tested by the most recent direct dark matter detection experiments PICO-2L [21], LUX [22,23] and PandaX-II [24,25], in combination with searches at the LHC for electroweakinos at run I and, in the future, by the high luminosity (HL) LHC.
At the LHC, charginos (higgsino-like or wino-like) can be pair produced, or in association with neutralinos χ 0 i . The lightest chargino can decay via possibly virtual W ±( * ) bosons or, if kinematically possible, via sleptons. The search channels are typically missing transverse energy E miss T plus two leptons (from leptonic decays of two W ± bosons in the case of chargino pair production), or three leptons from pp → W ( * ) → χ ± 1 + χ 0 2 with χ ± 1 → W ( * ) + χ 0 1 , χ 0 2 → Z ( * ) + χ 0 1 and leptonic decays of both W ( * ) and Z ( * ) . This E miss T + three lepton channel has typically the farthest reach in mass. Another search channel is W H (where H is the Higgs boson of the Standard Model), where one assumes a χ 0 2 decay of the form χ 0 2 → H + χ 0 1 , still concentrating on the leptonic W ( * ) decay. Previous studies of constraints on, phenomenological aspects and future prospects of the light singlino scenario in the NMSSM have been performed in [15,20,. Here we proceed as follows: in the plane defined by the chargino (charged higgsino) mass M χ ± 1 < 300 GeV and the (mostly singlino-like) LSP mass M χ 0 1 < M χ ± 1 we look for regions which are or will be definitely excluded by present dark matter searches and/or searches at the LHC. To this end we use the code NMSSMTools 5.0.1 [59,60] with NMSDECAY [61] to compute the spectrum, couplings and branching fractions, and fix the less relevant soft Susy breaking parameters to quite pessimistic values: the squark masses of the first two generations to 3 TeV, the slepton masses to 1 TeV (disregarding a possible contribution to the muon anomalous magnetic moment), the gaugino masses M 3 , M 2 and M 1 to 2.1 TeV, 700 GeV and 350 GeV, respectively, only A top and the stop masses are varied to ensure a Standard Model-like Higgs scalar with a mass of about 125 GeV. The NMSSM specific Yukawa couplings λ and κ, the soft Susy breaking terms A λ and A κ , µ and tan β are varied JHEP02(2017)051 for given points in the M χ ± 1 − M χ 0 1 plane. Actually M χ ± 1 defines essentially µ, and M χ 0 1 the ratio κ/λ.
The remaining parameters are still varied in order to find the strict boundaries in the above defined plane. These originate from a dark matter relic density consistent with the WMAP/Planck value Ω DM h 2 = 0.1187 [62,63], allowing for a 10% theory error, and from upper limits on spin-dependent and spin-independent direct detection cross sections from PICO-2L [21], LUX [22,23] and PandaX-II [24,25]. The dark matter relic density and the spin-dependent and spin-independent direct detection cross sections are computed with help of micrOMEGAS 3 [64].
In addition various other phenomenological constraints are required to be satisfied. These are notably constraints from LEP on the invisible Z width, from searches for charginos and associate production of neutralinos by DELPHI [65] and OPAL [66], and from searches for MSSM Higgs bosons [67] (which are relevant for a lighter mostly singletlike scalar). Relevant constraints from the LHC stem from the ATLAS and CMS combinations of Standard Model-like Higgs boson production and decay rates [68], which constrain possible exotic decays into pairs of NMSSM-like scalars, pseudoscalars or singlinos with masses below ∼ 60 GeV. The lower limit on the combined signal rates into the W W/ZZ channels is actually more relevant than the direct searches for exotic decays. All these constraints are implemented in NMSSMTools 5.0.1.
In the next section we describe the model and, in more detail, the constraints from dark matter in the various regions of the M χ 0 1 − M χ ± 1 plane. In section 3 we describe our simulations of signals at the LHC, their comparisons with constraints from run I and prospects for the High Luminosity LHC (based essentially on CheckMATE [69][70][71]). We conclude in section 4 with a summary and outlook.

The NMSSM and the singlino relic density
We consider the NMSSM with the scale invariant superpotential where the dots denote the Yukawa couplings of the superfieldsĤ u andĤ d to the quarks and leptons as in the MSSM. Once the scalar component of the superfieldŜ develops a vev S ≡ s, the first term in W NMSSM generates an effective µ-term with Subsequently the index eff of µ will be omitted for simplicity. µ generates a Dirac mass term for the charged and neutral SU(2) doublet higgsinos ψ u and ψ d . Additional charginos are the charged winos with a mass term M 2 . As stated in the introduction we assume M 2 = 700 GeV which leads to little mixing < ∼ 3% between the charged higgsinos and winos.

JHEP02(2017)051
Of particular importance will be the neutralino sector. Altogether the symmetric 5 × 5 mass matrix M 0 in the basis Henceforth the LSP is identified with χ 0 1 . Another important rôle will be played by the singlet-like scalar and pseudoscalar Higgs masses. The CP-even sector comprises three physical states which are linear combinations of the real components (H dR , H uR , S R ). The (3,3) element of the 3 × 3 CP-even mass matrix M 2 S reads in this basis it corresponds essentially to the mass squared of the mostly singlet-like eigenstate. Another eigenstate must correspond to a Standard Model-like Higgs boson H 125 with its mass ∼ 125 GeV, and nearly Standard Model-like couplings to quarks, leptons and gauge bosons. A third MSSM-like eigenstate has a mass of about 2 µ(A λ +κs) sin 2β . In the regions of the parameter space of interest here we always find that the mostly singlet-like eigenstate is the lightest CP-even scalar H 1 , the Standard-Model-like Higgs boson H 125 is the second lightest CPeven scalar H 2 , and the MSSM-like state is the third CP-even scalar H 3 .
The CP-odd sector consists in the linear combinations of the imaginary components (H dI , H uI , S I ). The (3,3) element of the 3 × 3 CP-odd mass matrix M 2 P reads in this basis again it corresponds essentially to the mass squared of the mostly singlet-like eigenstate.
Other eigenstates are the electroweak Goldstone boson, and an MSSM-like eigenstate again with a mass of about 2 µ(A λ +κs) sin 2β . The masses of the MSSM-like Higgs bosons are bounded from below to > ∼ 300−400 GeV due to constraints from b → sγ on the charged Higgs boson whose mass is similar to the ones of the CP-even and CP-odd neutral scalars. Subsequently the lightest mostly singlet-like CP-odd eigenstate will be denoted by A 1 .
From eqs. (2.3)-(2.5) one can derive the sum rule [31] which relates, up to modifications by mixing, the singlet-like neutralino, CP-even and CPodd Higgs masses. Notably for sizeable tan β (i.e. small v d ) and not too large A λ and Yukawa couplings λ and κ the last term in eq. (2.6) is negligible.

JHEP02(2017)051
In the following we consider, for reasons of naturalness as discussed in the introduction, a µ parameter < ∼ 300 GeV -below the bino and wino masses -and hence a dominantly singlino-like LSP χ 0 1 (singlino for short). Assuming a standard thermal history of the universe with a singlino in thermal equilibrium at temperatures above its mass, its annihilation rate must be sufficiently large such that its relic density today complies with the WMAP/Planck value Ω DM h 2 = 0.1187 [62,63]. Various processes can give a large enough annihilation cross section: • Annihilation via a pseudoscalar in the s-channel. For singlino masses M χ 0 1 below µ as assumed here this pseudoscalar is the singlet-like A 1 with its mass given in eq. (2.5) (up to a small shift through mixing). For M χ 0 1 below ≈ 100 GeV, M A 1 should be about 2 × M χ 0 1 such that the annihilation cross section is enhanced by the s-channel pole (depending on κ and the mixing of A 1 with the MSSM-like SU(2)doublet pseudoscalar which induces its couplings to quarks and leptons). For M χ 0 • Annihilation into a pair of W bosons via (higgsino-like) chargino exchange in the t-channel. This t-channel process is strong enough to be dominant only for singlino masses above ∼ 100 GeV.
• Annihilation via the Z boson in the s-channel if the singlino mass is about half the Z mass.
• Annihilation via the Standard Model-like Higgs boson H 125 in the s-channel if the singlino mass is about half the Higgs mass.
• Coannihilation with higgsinos is possible in principle, but strongly limited by the constraints discussed in the introduction and below.
Annihilation via A 1 ∼ S I in the s-channel with a pseudoscalar mass about twice the mass of the singlino is constrained by eq. (2.6): replacing M 2 S I ,S I by ∼ 4M 2 ψ S ,ψ S eq. (2.6) becomes leading to a negative CP-even scalar mass squared if the terms ∼ v u v d are neglected. This conclusion is avoided if one takes into account that the physical masses are obtained through diagonalization of the mass matrices and shifted by mixing. In the cases of the scalars this effect would be contraproductive: for small M 2 S R ,S R below (125 GeV) 2 , mixing with the Standard-Model-like or MSSM-like Higgs bosons would decrease the eigenvalue of M 2 S corresponding to M 2 H 1 even further. Likewise, mixing in the CP-odd sector among the singlet-like and the heavier MSSM-like states reduces the physical mass of the mostly singlet-like state A 1 , aggravating the consequences of the sum rule (2.6) if M A 1 is required to be about twice as large as the singlino mass.

JHEP02(2017)051
In the neutralino sector, on the other hand, mixing reduces the mass of the singlino and allows it to be about half of the pseudoscalar mass for positive (albeit small) CP-even masses squared. Such mixing requires the term −λv u in the mass matrix (2.6) to be not too small. Altogether the scenario where annihilation via A 1 with M A 1 ∼ 2M χ 0 1 leads to a satisfactory relic density is characterized by -a light singlet-like CP-even scalar H 1 , -a non-negligible higgsino component of the dominantly singlino-like LSP χ 0 1 , -the latter requires a non-negligible value of λ ∼ 0.1 − 0.4 (larger for larger higgsino masses ∼ µ relative to M χ 0 1 ). Correspondingly this scenario is subject to the following constraints: -H 1 must satisfy constraints from Higgs searches at LEP, -the BR(H 125 → H 1 H 1 ) must remain below ∼ 10% in order not to reduce the Standard Model-like signal rates of H 125 below its present limits, -the spin-independent direct detection rate mediated by H 1 in the t-channel must satisfy the (M χ 0 1 dependent) bounds from LUX [23] and PandaX-II [24], -for M χ 0 1 + µ below ∼ 200 GeV limits from DELPHI [65] and OPAL [66] on Z * → χ 0 1 + χ 0 2,3 must be satisfied, -and, of course, the relic density must coincide with the WMAP/Planck value Ω DM = 0.1187 ± 10%.
As stated in the introduction, the NMSSM specific Yukawa couplings λ and κ, the soft Susy breaking terms A λ and A κ , µ and tan β are varied in order to satisfy all constraints if possible. Still, the combination of these constraints rules out regions in the M χ 0 1 − M χ Another forbidden region appears for M χ 0 1 too close to M χ ± 1 ∼ µ ∼ M χ 0 2,3 . There mixing between χ 0 1 and the higgsinos χ 0 2,3 cannot be avoided, and the higgsino component of χ 0 1 implies a spin-dependent direct detection rate via Z exchange in conflict with constraints from LUX [22] and PandaX-II [25]. (The importance of recent constraints from spindependent direct detection experiments on the viable parameter space of the NMSSM has recently been underlined in [58].) Also the relic density tends to be too small if χ 0 1 has a -6 - in the white regions are allowed, but at least one particular choice of parameter values gives a conflict-free good relic density.
Actually, due to the presence of the additional lighter mostly singlet-like Higgs boson H 1 in this scenario, the upper bounds on the spin-independent dark matter cross section mentioned above constrain the parameters everywhere: often the t-channel contribution of H 1 to the scattering amplitude has to be compensated at least partially by a t-channel contribution of H 125 of opposite sign. Furthermore, for M H 1 < ∼ 60 GeV, the coupling H 125 − H 1 − H 1 must be small enough in order to suppress H 125 → H 1 + H 1 decays as stated above, and LEP bounds from searches for a light Higgs boson must be respected. These constraints can become relevant all over the white regions.
Additional constraints depend on the dominant processes for the reduction of the relic density: 1 for a relic density below the WMAP/Planck value. Of course, fine tuning would be required in order to obtain a relic density within the present ∼ 2% accuracy of the combined WMAP/Planck measurements, but this cannot be blamed to the theory.

JHEP02(2017)051
-Z or H 125 in the s-channel: here M χ 0 1 has to be in the range M χ 0 1 ∼ 39 . . . 49 GeV (in the Z case) or M χ 0 1 ∼ 59 . . . 64 GeV (in the H 125 case) for a relic density below the WMAP/Planck value. The couplings of χ 0 1 to Z or H 125 are proportional to its higgsino component, and hence bounded by the limits on the spin-dependent dark matter cross section.
-χ ± 1 in the t-channel: as stated above, here M χ 0 1 is close to M χ ± 1 ∼ µ ∼ M χ 0 2,3 leading to mixing between χ 0 1 and the higgsinos χ 0 2,3 inducing potentially a too large spindependent direct detection cross section. The reduction of the mixing requires the more tuning the more one approaches the corresponding forbidden blue region.
The satisfaction of all of these constraints requires to tune the dimensionless parameters λ, κ and tan β with a precision of ∼ 1 − 2% relative to each other nearly everywhere in the white regions, the required tuning of the dimensionful parameters µ, A λ and A κ of ∼ 10 − 30% is weaker. A notable exception is the case of a light χ 0 1 with a mass below 5 GeV where the present constraints on the dark matter detection cross sections become weak or fade away: the remaining constraints from the relic density and from Z * → χ 0 1 +χ 0 2,3 production require only a relative tuning of ≈ 30%. However, near the boundaries to the blue regions the parameters have to be tuned more severely in order to satisfy the more and more stringent combinations of constraints (which become impossible to satisfy within the blue regions), hence NMSSM points near these boundaries are necessarily fine tuned.  [83][84][85][86][87][88][89][90] and run II [91][92][93]. We recall that the dominant search channels for charginos/neutralinos at the LHC are searches for 2 − 3 leptons and E miss T . The absence of significant excesses in these channels is typically interpreted by the ATLAS and CMS collaborations in terms of simplified models. These mostly assume, however, wino-like charginos χ ± 1 (and wino-like neutralinos χ 0 2 ) which have significantly larger production cross sections than higgsinos. Specific searches for higgsinos have been performed by ATLAS [72, 76, 78] and CMS [85] assuming a lightest SUSY particle (LSP) in the form of a massless gravitino [85], or the presence of a light stau [76, 78] or light sleptons [78]. Apart from these specific scenarios, light higgsinos are thus hardly constrained by searches at the run I of the LHC, see the analyses within phenomenologically viable versions of the MSSM or NMSSM in [33,[94][95][96][97][98][99][100][101][102][103][104][105][106][107][108][109].

Present and future constraints from the LHC
The discovery potential for wino-like χ ± 1 /χ 0 2 production at the future and high luminosity LHC has been estimated by ATLAS [110,111] and CMS [112] in the trilepton and W H channels, and seems quite promising at first sight with 95% CL exclusion contours reaching up to a wino mass of about 800 GeV for a light LSP at 300 fb −1 [110][111][112].

JHEP02(2017)051
First we have recast the run I analyses from [73-75, 79] for the light higgsino-singlino scenario of the NMSSM. (The exclusion reach of these analyses in the M χ 0 1 − M χ ± 1 plane is not superseeded by others.) As in the previous section we varied the NMSSM parameters for many values of M χ 0 1 and M χ ±
The result of this recasting was that no region in the M χ 0 1 − M χ ± 1 plane is conclusively excluded at the 95% level by the available 3-lepton + E miss T (and 1-lepton + H 125 + E miss T ) searches at run I, and this not only marginally. The reasons for this result are -the smaller production cross section for higgsinos compared to winos; -a further reduction of the higgsino production cross section due to mixing of the neutral higgsinos with the singlino; -sizeable branching fractions of the neutral higgsinos into the singlino-like LSP + the light singlet-like scalar H 1 (and, less prominently, the light singlet-like pseudoscalar A 1 ) notably if decays via Z and/or H 125 are kinematically forbidden. These decays escape detection in the W H final state due to the much smaller singlet-like Higgs mass.
Next we recast the prospects for chargino-neutralino searches at the HL-LHC at 3000 fb −1 into the higgsino-singlino scenario of the NMSSM. The analyses from [110] (which have similar exclusion/detection reaches as the ones in [111,112]) are implemented in CheckMATE 2 [71].
The relevant signal regions are again the ones dedicated to W Z and W H 125 final states, with Z or H 125 originating from χ 0 2 → χ 0 1 + Z/H 125 decays. Since the search for W Z requires two same-flavour opposite-sign leptons with an invariant mass close to M Z , it becomes sensitive to the light higgsino-singlino scenario of the NMSSM for figure 2 as a black line (below which the inequality is satisfied). Discovery of NMSSM points below this line is thus possible for selected points, since the production cross sections and branching fractions can be large enough for higgsino masses below 300 GeV despite their reduced couplings.
However, even a "3 σ discovery" is not necessarily guaranteed, amongst others for the same reasons as for the run I. Another reason are large cuts on the lepton p T and E miss T , which are required to cope with the larger instantaneous luminosity (and the larger background), see the corresponding cuts in [110][111][112]. So we ask again whether, varying all parameters within the imposed constraints, any region in the M χ 0 1 − M χ ± 1 plane can be conclusively excluded at the 95% level at the HL-LHC. The cuts lead to large enough acceptances only for mass differences M χ ± . The red region can be excluded at 95% CL at the HL-LHC at 3000 fb −1 using a recast of the prospects for chargino-neutralino searches by ATLAS [110]. The yellow region can be excluded according to a search proposed in CheckMATE 2 [71] including hadronic W decays. The blue region is the same as in figure 1.
The region in the M χ 0 1 −M χ ± 1 plane which can be excluded at the 95% CL level is shown in red in figure 2. In CheckMATE 2 [71] an additional search region is proposed, which is based on 2 leptons + E miss T +W had , the hadronic decays of W from the χ ± 1 → χ 0 1 + W ± decays. Due to the larger branching fraction of W had this search may cover a slightly larger region in the M χ 0 1 − M χ ± 1 plane, which is indicated in yellow in figure 2. The viability of this search channel remains to be confirmed by the experimental collaborations, however. In any case, even the HL-LHC at 3000 fb −1 seems unable to test the full M χ 0 1 − M χ ± 1 plane in the light higgsino-singlino scenario of the NMSSM.
In figure 2 we have indicated again in blue the regions which are excluded by the combination of constraints notably from the dark matter relic density, and the present constraints from direct dark matter detection. In the absence of signals in the future these latter constraints would become stronger, but we have checked that the boundaries would hardly move. The reason is that the LSP can remain dominantly singlino-like in most regions except for a singlino mass very close to the higgsino (chargino) masses; hence the conclusions would hardly change.
Although one observes a certain complementarity between the constraints from dark matter and the possible future limits at the LHC it is obvious that large regions in the M χ 0 1 − M χ ± 1 plane will remain unexplored. This concerns in particular well motivated and natural regions where M χ ± 1 < ∼ 150 GeV and M χ 0 1 < ∼ 5 GeV; such light dark matter candidates are difficult to discover via direct detection and the hope was that they are easier to discover at the LHC.

JHEP02(2017)051 4 Summary and conclusions
The light higgsino-singlino scenario of the NMSSM is probably the most attractive scenario for supersymmetric dark matter, since it allows to combine a small µ parameter with a good relic density and alleviated constraints from dark matter searches. First we have shown that this remains so after the updated constraints on spin-independent and spindependent cross sections in 2016. Then we have studied present and future constraints on this scenario from the LHC, and found that even the HL-LHC seems not able to test all possible realizations of this scenario. Possible ways to improve this situation could be: • To rely on the larger production cross sections of wino-like charginos and neutralinos, and the discovery of lighter higgsinos/singlino via cascade decays. Of course this depends on additional parameters like the wino-(and possibly slepton-) masses and branching fractions. Moreover the transverse momenta of the leptons and E miss T would again be reduced through multiple cascades.
• To try to become sensitive to smaller transverse momenta of the leptons and E miss T even at 14 TeV and large instantaneous luminosity by recording and analysing only a fraction of the events.
If such attempts turn out to be fruitless, this attractive scenario could be tested only at an electron-positron collider [115].  [80] ATLAS collaboration, Search for supersymmetry with two same-sign leptons or three leptons using 13.2 fb −1 of √ s = 13 TeV pp collision data collected by the ATLAS detector, ATLAS-CONF-2016-037 (2016).
[82] ATLAS collaboration, Search for supersymmetry with two and three leptons and missing transverse momentum in the final state at √ s = 13 TeV with the ATLAS detector, ATLAS-CONF-2016-096 (2016).