Light NMSSM Higgs bosons in SUSY cascade decays at the LHC

An interesting feature of the next-to-minimal supersymmetric standard model (NMSSM) is that one or more Higgs bosons may be comparably light (M_{H_i}<M_Z) without being in conflict with current experimental bounds. Due to a large singlet component, their direct production in standard channels at the Large Hadron Collider (LHC) is suppressed. We demonstrate that there are good prospects for observing such a light Higgs boson in decays of heavy neutralinos and charginos. We consider an example scenario with 20 GeV<M_{H_1}<M_Z and show that a large fraction of the cascade decays of gluinos and squarks involves the production of at least one Higgs boson. Performing a Monte Carlo analysis at the level of fast detector simulation, it is demonstrated how the Higgs signal can be separated from the main backgrounds, giving access to the Yukawa coupling of the Higgs to bottom quarks. Analyzing the resulting b\bar{b} mass spectrum could provide an opportunity for light Higgs boson discovery already with 5 fb^{-1} of LHC data at 7 TeV.


Introduction
The precise nature of the Higgs mechanism thought to be responsible for electroweak symmetry breaking remains unknown. To discover and study the properties of one or more Higgs bosons is therefore a challenge -and one of the major objectives -for the experiments at the running Large Hadron Collider (LHC). Most of the Higgs boson search strategies to date are designed to probe the Higgs sector of the Standard Model (SM), or its minimal supersymmetric extension (MSSM) [1,2].
There are several theoretically appealing arguments for weak-scale supersymmetry to be realized in nature: it solves the hierarchy problem of the SM Higgs mass, it enables gauge coupling unification, and with R-parity conservation it also provides a natural dark matter candidate. On the other hand, the realization of weak-scale supersymmetry in terms of the MSSM is not free of theoretical problems, such as the scale for the bilinear µ-parameter entering the MSSM superpotential with positive mass dimension. This parameter has no natural values besides M GUT or zero, while at the same time it must be close to the electroweak scale for a phenomenologically acceptable theory. To solve this problem in an elegant way the MSSM can be extended by a complex scalar singlet, giving the so-called next-to-minimal supersymmetric model (NMSSM). In this model an effective µ-term of the right size can be generated dynamically from supersymmetry-breaking operators. For a general introduction to the NMSSM we refer to the recent reviews [3,4].
The NMSSM is characterized by an enlarged Higgs and neutralino sector as compared to the MSSM, giving rise in particular to a richer Higgs phenomenology. While it is well known that in certain scenarios of the MSSM with complex parameters a light Higgs, with mass much below that of the Z boson, is unexcluded by the searches at LEP [5] and the Tevatron [6] (see [7,8] for recent reevaluations with improved theoretical predictions), such a scenario can occur even more generically in the NMSSM. In order to be compatible with the limits from the LEP Higgs searches, in particular the couplings of such a light Higgs state to gauge bosons must be heavily suppressed. As a consequence of the presence of a Higgs singlet, in the NMSSM such a situation happens whenever a light Higgs state has a sufficiently large singlet component.
The search for a heavier Higgs state with SM-like (or only moderately suppressed) couplings to gauge bosons is complicated in such a scenario by the fact that often the decay of this heavier Higgs state into a pair of lighter Higgses is kinematically open, giving rise to unusual decay signatures and to a large suppression of the standard search channels for a SM-like Higgs. It should be noted in this context that the observation of a decay of a heavier Higgs into a pair of lighter Higgses would provide an opportunity for gaining experimental access to triple-Higgs couplings, which are a crucial ingredient of electroweak symmetry breaking via the Higgs mechanism.
While in the NMSSM the case of a very light pseudo-scalar, M A 1 < 2m b , has found considerable attention in the literature, in particular in the context of "ideal" Higgs scenarios [9][10][11][12][13][14], we will focus in the following on scenarios with a light CP-even boson with 20 GeV < M H 1 < M Z . Within the MSSM the best known example of a light Higgs that is unexcluded by the present search limits is the "hole" in the coverage of the CPX benchmark scenario [15] for M H 1 ≈ 45 GeV and moderate values of tan β [5], see [8] for a detailed discussion of the dependence of the unexcluded parameter region on the choice of the various MSSM parameters. It will be difficult to cover this parameter region with the standard search channels at the LHC [16][17][18]. Various other (non-standard) search channels have been proposed which may provide additional sensitivity in the quest to close this "CPX hole" [16,[19][20][21][22][23][24][25][26].
In our analysis within the NMSSM we will investigate the prospects for the production of light Higgs bosons in cascade decays of heavy SUSY particles at the LHC. Such an analysis, where Higgs bosons are produced in association with -or in decays of -other states of new physics, is necessarily more model-dependent than the Higgs search in SMlike channels. On the other hand, investigating Higgs physics in conjunction with the production of other states of new physics offers additional experimental opportunities and may also be more realistic, in the sense that in order to extract a Higgs signal backgrounds both from SM-type and new physics processes have to be considered. In the case of the MSSM with real parameters, for a Higgs with a mass above the LEP limit for a SM Higgs of 114.4 GeV [27], a detailed experimental study for Higgs boson production in a SUSY cascade has been carried out by the CMS Collaboration [28], involving a full detector simulation and event reconstruction. These results, obtained for the benchmark point LM5, cannot be directly translated to the case of searches for a Higgs boson with mass far below M Z , since in the latter case the b jets resulting from the Higgs decay tend to be softer. Further phenomenological analyses of Higgs production in SUSY cascades in the MSSM with real parameters (and Higgs masses above the LEP limits) have been carried out in [29][30][31], with recent developments focusing on jet substructure techniques to identify highly boosted Higgs bosons and enhance the discovery significance [32,33]. The case of a lower mass Higgs has been considered in [23], and it has been pointed out that in the CPX scenario there is a significant rate for producing a light MSSM Higgs boson in SUSY cascades, but no simulation of signal and background events was performed. The potential importance of SUSY cascades to establish a signal for a light CP-odd Higgs in the NMSSM has been pointed out in [34].
We generalize and extend the investigations carried out in [23,34] by calculating the sparticle decay modes in a general NMSSM setting and performing a Monte Carlo simulation of the signal and the dominant background to the level of fast detector simulation. A simple cut-based analysis is performed to demonstrate that signal and background can be resolved in the bb+jets channel. The observation of the Higgs decay in the bb final state would be of interest also as a direct manifestation of the Higgs Yukawa coupling.
The outline of our paper is as follows: the next section begins with a brief recapitulation of the NMSSM, presenting the scenario with a light CP-even Higgs boson in some detail. In section 3, we describe the production of squarks and gluinos at the LHC and their eventual decay into Higgs bosons through electroweak cascades involving neutralinos and charginos. Section 4 describes a phenomenological Monte Carlo analysis of these cascade processes and contains the main results of this work in terms of kinematic distributions demonstrating the separation of signal and background. The conclusions are presented in section 5.

The Next-to-Minimal Supersymmetric Standard Model
In this section we review briefly the elements of the NMSSM which differ from the MSSM. Our conventions for the other sectors -that remain unchanged when going to the NMSSM -follow those of [35,36].

The Higgs sector
The Z 3 -symmetric version of the NMSSM is given by the scale-invariant superpotential whereΦ denotes a chiral superfield with scalar component Φ. The complex gauge singletŜ is a new addition with respect to the MSSM. To have a complete phenomenological model the soft SUSY-breaking terms must also be specified. These are extended by couplings of the singlet field, giving new contributions to the scalar potential The NMSSM Higgs potential, which is derived from the usual F -terms, D-terms and the soft-breaking potential given by Equation (2.2), allows for a minimum where the singlet develops a vacuum expectation value (vev) v s = S . This induces an effective bilinear term λ S H u · H d , thus providing a dynamical explanation for the µ parameter of the MSSM in terms of µ eff = λv s . Electroweak symmetry breaking (EWSB) proceeds similarly to the MSSM, and the two Higgs doublets are expanded around the potential minimum according to The upper left 4 × 4 submatrix is identical to the neutralino mass matrix in the MSSM. The neutralino masses can be diagonalized by a single unitary matrix N such that is real and positive with the neutralino mass eigenvalues in ascending order. Alternatively one can use a real mixing matrix N , and allow D to have negative elements. In this case the physical neutralino masses are given by |mχ0 i | and the neutralino couplings incorporate the additional phase shift on the neutralino fields.

The squark sector
We adopt a universal value M SUSY for the soft SUSY-breaking scalar mass parameters. This means that, for each squark pairq L ,q R of a given flavour, the mass matrix attains the form (2.12) Here m q is the mass of the corresponding quark, I q 3 the third component of the weak isospin, and Q q the electric charge quantum number. For the weak mixing angle we introduce the short-hand notations s W ≡ sin θ W and c W ≡ cos θ W . The off-diagonal elements of M 2 q are related to the soft trilinear couplings A q as X q = A q − µ eff cot β for up-type squarks, and X q = A q − µ eff tan β for the case of down-type squarks, respectively. The mass eigenstates (q 1 ,q 2 ) are obtained by a diagonalization of the mass matrix. A generic squark mass will be denoted Mq below.

Scenarios with light Higgs bosons
As mentioned above, we will focus in the following on the case where the lightest CP-even Higgs boson of the NMSSM, H 1 , has a mass much below M Z . The fact that such a light Higgs, possessing a heavily suppressed coupling to gauge bosons as compared to the Higgs boson of the SM, may be unexcluded by the current search limits is known already from the case of the MSSM with complex parameters [5,7,8]. In the NMSSM such a situation happens more generically, in particular also for the case where the SUSY parameters are real. If the mass eigenstate H 1 has a large component of the singlet interaction state φ s , its couplings to gauge bosons (and also to quarks) will be correspondingly suppressed. We will investigate the prospects for detecting such a light Higgs state through its production in SUSY cascades.
In the numerical analysis, we shall use a scenario derived from the "P4" benchmark point defined in [37]. This benchmark can be realized in models with non-universal Higgs mass parameters (m Hu = m H d ) at the scale of grand unification, and it is compatible with the data on the cold dark matter density. As originally defined, the P4 benchmark contains a very light CP-even Higgs boson (M H 1 = 32.3 GeV). In order to explore the full range M H 1 < M Z , we slightly modify the scenario to allow changing the value of M H 1 , with the remaining phenomenology essentially unchanged. To this end we set λ = 0.6 and allow A κ to take on values in the range 0 GeV < A κ < 300 GeV. 2 The soft SUSY-breaking parameters are defined directly at the SUSY-breaking scale, allowing us to consider a more Higgs sector parameters λ 0.6 κ 0.12 Soft scalar mass M SUSY = 750 GeV, 1 TeV  600 600 general spectrum for the remaining (non-Higgs) sectors of the theory. Values for the treelevel parameters in the Higgs sector and the soft SUSY-breaking parameters in the modified P4 scenario are specified in table 1. The two values given for M SUSY will be used later in the phenomenological analysis, while the values quoted in this and the next section (unless otherwise stated) have been evaluated for M SUSY = 750 GeV. The NMSSM Higgs masses are subject to sizable corrections beyond leading order [38][39][40][41][42][43]. In order to incorporate the most accurate predictions currently available [44], NMSSMTools 2.3.5 [45][46][47] is used to compute the Higgs spectrum. The resulting Higgs masses in the modified P4 scenario are shown in figure 1 as a function of the free parameter A κ . In the region with A κ 250 GeV the global minimum of the Higgs potential does not break the electroweak symmetry; hence these values will not be considered. The masses of two Higgs bosons show a dependence on A κ : the lightest CP-even Higgs plot the NMSSM scenario is compatible with the direct limits from Higgs searches. The light CP-even Higgs (M H 1 ≪ M Z ) is allowed due to a large singlet component, with |S 13 | 2 ranging from 0.9 for A κ = 0 GeV to |S 13 | 2 > 0.99 for A κ = 250 GeV. As a consequence, the couplings of H 1 to vector bosons are heavily suppressed, so that the cross section for production through Higgsstrahlung drops below the LEP limit. The pair production of A 1 H 1 is even further suppressed by the large singlet fractions of both H 1 and A 1 , while production of H 2 A 1 and H 2 Z are beyond the kinematic reach of LEP. The full mass ranges shown in the figure are also compatible with the constraints from B-physics implemented in NMSSMTools 2.3.5 [45][46][47], as expected when the charged Higgs boson is heavy [48,49].
The precise values obtained here for the heavy Higgs masses are M H ± ≃ 563 GeV, and None of the heavy Higgs bosons will play any role in the following. With the negative sign for the effective µ parameter, this model cannot be used to explain the observed deviation in the anomalous magnetic moment of the muon (see e.g. [50] for a review). However -since the considered value of tan β is rather low -the predicted value for (g − 2) µ at least stays close to that in the SM. The branching ratios of the three lightest Higgs states, H 1 , H 2 , and A 1 , are given in figure 2. As can be seen from this figure, the light singlet H 1 decays preferentially into bb, with BR(H 1 → bb) ≃ 90% over the full mass range. The subdominant decay into τ τ basically saturates the H 1 width. For lower values of A κ -where M H 1 90 GeV -we note a similar enhancement of BR(H 1 → γγ) compared to a SM Higgs with the same mass as recently discussed in [51]. The H 2 has a more complicated decay pattern, in particular for low A κ where H 2 → bb dominates and several competing modes (H 2 → τ τ , gg, W W ) each have a branching fraction around 10%. In this region H 2 is SM-like, and the same search strategies as devised for the SM Higgs (and the lightest MSSM Higgs boson in the decoupling limit) should apply. This situation changes radically when the channel H 2 → H 1 H 1 opens. When this is the case, the H 2 → H 1 H 1 mode becomes completely dominant. Finally, the lightest CP-odd Higgs A 1 decays predominantly into bb, with a large fraction going into the mode A 1 → H 1 Z when kinematically accessible.
In the neutralino sector the mass spectrum is independent of A κ (and M SUSY ) at lowest order, cf. equation (2.10), and therefore remains fixed at: GeV. There is a clear hierarchy in the mass parameters, which leads to a small mixing between the neutralinos. The heaviest neutralino is almost exclusively wino, andχ 0 4 is mostly bino. The intermediate mass statesχ 0 2 , andχ 0 3 are predominantly Higgsino, while the lightest neutralinoχ 0 1 is the singlino. The lightest neutralino is also the overall lightest supersymmetric particle (LSP) in these scenarios and thereby a candidate for cold dark matter.
3 Higgs production in the light H 1 scenario

Standard channels
The rate for direct production of a light singlet H 1 in gluon fusion, gg → H 1 , is proportional to its reduced (squared) coupling to quarks. Compared to a SM Higgs boson with the same mass, the dominant top loop contribution contains the additional factor |S 12 | 2 / sin 2 β. The size of |S 12 | 2 is limited from above by |S 12 | 2 ≤ 1−|S 13 | 2 , where S 13 is the singlet component. For M H 1 ≪ M Z , where |S 13 | → 1, the rate for this process gets heavily suppressed. The cross section for H 1 in weak boson fusion, involving the coupling of H 1 to gauge bosons, is similarly suppressed. In scenarios where M H 1 > M Z (corresponding to the mass range below the LEP limit on a SM-like Higgs which is unexcluded in the MSSM with real parameters) the suppression of gg → H 1 can be overcome by an increased branching ratio for H 1 → γγ [51].
For A κ 200 GeV in the modified P4 scenario, H 1 is light enough to be produced through the decay of the SM-like H 2 → H 1 H 1 , which can be dominant, see figure 2. The production of H 2 in standard channels is not suppressed. The resulting two-step decay chain leads to "unusual" final states for H 2 : 4b (about 82% of all decays), 2b2τ (17%), and 4τ (0.6%). These final states make it difficult to establish a Higgs signal, as it has been demonstrated, for instance, by the numerous attempts [52][53][54][55][56] to establish a "no-loose" theorem for NMSSM Higgs searches when decays of the SM-like Higgs into lighter Higgses are open.
Another possibility to produce H 1 in Higgs decays would be through the decay A 1 → H 1 Z. However, the singlet nature of A 1 in the modified P4 scenario leads to a suppression of A 1 production similar to that for H 1 , and this mode is therefore not likely to be accessible.
The direct production of the heavy Higgs bosons H 3 , A 2 , and H ± is in principle not suppressed with respect to the MSSM case, but at a mass close to 600 GeV and low tan β the observation of those states at the LHC will be difficult even at high luminosity. A large fraction of the heavy Higgs bosons in this scenario will decay into lighter Higgs bosons, neutralinos and charginos. A detailed investigation of these channels could possibly be of interest for a study assuming a very high luminosity at 14 TeV, but is beyond the scope of the present paper.
In summary, it will be problematic to produce and reconstruct the light H 1 in any of the standard channels proposed for Higgs production at the LHC. We shall focus instead on the possibility to produce H 1 in the decays of supersymmetric particles.

SUSY cascades
As discussed in the previous section, inclusive production of the heavier state H 2 with subsequent decay H 2 → H 1 H 1 may be difficult to observe at the LHC. However, the related process where a heavier neutralino decays into a lighter neutralino and a Higgs boson (and the corresponding mode of the decay of the heavier chargino) may offer better prospects. In fact, a light Higgs boson in the mass range below M Z may occur in a large fraction of cascade decays of heavier SUSY particles that are produced via strong interaction processes. The hard scale associated with the sparticle production can lead to event signatures which are more clearly separable from the SM backgrounds than those of inclusive Higgs production followed by a decay into a pair of H 1 states. The processes of interest areχ where H k (A k ) denotes any of the CP-even (CP-odd) Higgs bosons. As mentioned above we do not consider scenarios where the heavier H ± is produced in the cascades. The partial width for the neutralino decay (3.1) is given at tree-level by with a CP-even Higgs in the final state and for the decay into a CP-odd scalar. The Källén function τ (x, y, z) = (x − y − z) 2 − 4yz, and the coupling factors are and where the mixing matrices S ij , P ij , and N ij are defined in section 2.1. Equations (3.3)-(3.6) assume a real neutralino mixing matrix N ij and signed neutralino masses. Competing neutralino decay modes are into vector bosons,χ 0 For brevity we refrain from giving expressions for these (and the corresponding chargino decay modes) here; they can be found in [57]. A detailed analysis of the W ± mode is performed in [58]. Since the squarks and sleptons are assumed to be heavy, there are no open two-body decay modes of the neutralinos into the sfermion sector. Also slepton-mediated three-body decays -which can dominate over the two-body decays in certain scenarios -are numerically irrelevant for the same reason.
The branching fractions for the relevant decay channels have been computed at leading order using FeynArts/FormCalc [59,60] and a purpose-built Fortran code. 3 Results for the neutralino branching ratios in the modified P4 scenario are shown in figure 3. The decay modes ofχ 0 2 andχ 0 3 (upper row of figure 3) -which are both Higgsino-like -show similar patterns for large values of A κ . The dominant mode is alwaysχ 0 i →χ 0 1 Z with a branching ratio of about 50%, but the Higgs channels are also significant with BR(χ 0 15. An important point to note here is that the branching ratios ofχ 0 2 andχ 0 3 are quite insensitive to changes in M H 1 (A κ ). For the heavier neutralinos, χ 0 4 andχ 0 5 (lower row of figure 3), which also carry a larger gaugino fraction, the decay pattern is more complicated. Of largest interest for Higgs production is the sizable rate for χ 0 4 →χ 0 3 H 1 (once A κ is sufficiently large to make this decay mode kinematically possible), and the fact that direct decays ofχ 0 5 to the LSP are suppressed. This will lead to neutralino decay chains with intermediate (Higgsino) steps. Everything taken together, we can expect a large number of light Higgs bosons to be produced in neutralino cascade decays.
The light charginoχ ± 1 decays exclusively into the LSP and a W boson, while the corresponding decay channels for the heavier chargino χ ± 2 are shown in figure 4. Even if the dominant mode isχ ± 2 →χ ± 1 W ∓ , independently of A κ , there are several channels with a branching fraction of order 20% of interest for Higgs production. These include the modẽ χ ± 2 →χ ± 1 H 2 and the decays into intermediate-mass Higgsinos,χ ± 2 →χ 0 2,3 W ± . The heavier neutralinos and the heavy chargino, from which a Higgs could emerge as decay product, can either be produced at the LHC directly or in the decay of a heavier SUSY particle. The cross section for direct production of neutralino pairs is small, only O(fb) at √ s = 14 TeV, and the reach in these channels will be rather limited even for high luminosity. The large cross sections for production of strongly interacting sparticles (squarks and gluinos), on the other hand, are potentially more promising as a source of the heavier neutralino states and the heavier chargino. Exploiting cascade decays of this kind The color coding indicates the final state neutralino j = 1 (black), j = 2 (blue), j = 3 (magenta), j = 4 (green), or the chargino modẽ furthermore has the advantage that additional high-p T jets are produced, which facilitates triggering and event selection. We use Prospino to calculate the NLO cross sections for production of pp →gg, pp →qq, pp →qq, and pp →gq according to [63], with CTEQ6   [65] are also calculated and included in the analysis, but since they turn out to be significantly smaller than σ(pp →qq) they are not shown in the table. In order to give some indication of the expected change in the number of events for different scenarios, the results are presented for several values of the squark masses, Mq, and the gluino mass, Mg. The mass ranges are selected to respect the published limits from ATLAS [66][67][68][69][70] and CMS [71][72][73][74] based on the 2010 data. Taking into account also the most recent results [75,76], the Mq = 750 GeV case appears to be under some pressure. We present the results of our analysis below for the two cases M SUSY = 750 GeV and M SUSY = 1 TeV (the leading order squark masses are obtained from M SUSY through eq. (2.12), to which higher order corrections are then added). The nearly mass-degenerate squarks decay preferentially into the SUSY-EW sector. Direct decays into Higgs bosons (or Higgsinos) are negligible for squarks of the first two generations due to the small Yukawa couplings. In contrast to the MSSM, the neutralinos also have a singlino component to which no squark couples. The left-handed squarks decay mainly into the wino,q L →W 0 q,q L →W ± q ′ , while the right-handed squarks decay mostly to the bino,q R →Bq. Numerically, this leads to squark decay modes listed in table 3 for the case with a soft scalar mass of M SUSY = 750 GeV. The squark decay pattern for M SUSY = 1 TeV is qualitatively similar. 4 Since the gaugino components are largest in the two heaviest neutralinos, the neutralinos produced in the squark decays tend to give rise to cascade decays with several steps.
Finally, we note that the gluinos decay 'democratically' throughg →qq into all flavours, with rates governed only by the available phase space. 4 The main numerical difference is an increase of BR(ũL → q ′χ± 2 ) to 60% at the expense of a reduced BR(ũL → q ′χ± 1 ) = 3.9 × 10 −2 .  Table 3. Branching ratios for the first and second generation squarks into neutralinos and charginos in the modified P4 scenario with M SUSY = 750 GeV. Results for channels with a branching ratio below 10 −4 are not shown.

LHC analysis
In order to assess whether the process discussed in the previous section can be useful as a Higgs search channel at the LHC we perform a Monte Carlo simulation. Here we use as benchmark the modified P4 scenario with the two different settings for the soft scalar mass: M SUSY = 750 GeV and M SUSY = 1 TeV. The DR value of the gluino mass parameter is set to M 3 = 1 TeV. We select A κ such that M H 1 ≃ 40 GeV, which also affects M H 2 , M A 1 and the branching ratios in the two cases as discussed in section 3. We have chosen this value of M H 1 as an illustrative example of our scenario with 20 GeV < M H 1 < M Z and in order to make contact with the analyses of the "CPX hole" in the MSSM with complex parameters. Our results however depend only very mildly on the specific choice for M H 1 . The simulation results are presented below both for LHC running at centre-of-mass energies of 7 TeV and 14 TeV. The squark and gluino-induced cascades in general give rise to a final state with high multiplicities and several hard jets, as well as large missing transverse momentum due to the presence of the LSP at the end of each decay chain. The minimal signal cascades (defined to be those with at least one Higgs boson present) generated by the production of a single squark or gluino correspond tõ Equations (14 a) and (14 b) show the minimum number of light and heavy flavour (b-) jets expected in the signal. Each event contains production of a pair of sparticles and their associated jets, meaning that the full signature for production of at least one H 1 in the hadronic final state will be n j ≥ 2, n b ≥ 2. Since direct decays of the heavier (mainly gaugino) neutralinos into the singlino LSP are practically absent (cf. figure 3), most signal cascades will contain an intermediate Higgsino step which will add further particles in the final state. The typical jet multiplicity will also be higher due to additional QCD activity, in particular for gluon-initiated processes.

Event Generation
For the event generation, we use MadGraph/MadEvent 4.4.44 [77] to calculate the leading order matrix elements for pp →gg,qq,gq,qq,tt, andbb. The different event categories are weighted by the corresponding NLO cross sections to produce an inclusive SUSY sample. The resonance decay chains are then generated with PYTHIA 6.4 [78] using the NMSSM decay rates calculated above as input through the SUSY Les Houches accord [36]. The PYTHIA generator is also used to produce additional QCD radiation through initial-and final state parton showers, for parton fragmentation, and to generate multiple interactions for the underlying event. This produces fully dressed hadronic events which are passed through the fast simulation of the ATLAS detector performance implemented in the Delphes package [79]. 5 Hadronic jets are clustered using the anti-k T algorithm [80] with a jet radius measure of R = 0.4. Since for the lightest Higgs boson the decay to bb is favored, the probability η b to correctly identify jets originating from bottom partons (b-tagging efficiency) becomes a crucial quantity for the analysis. Based on [81] we parametrize this efficiency as a constant η b = 0.6 with respect to both the detector geometry and the jet energy scale. Only jets in the central tracking region |η| < 2.5 can be tagged. The rate for misidentification as a b-jet is assumed to be η c = 0.1 for charm jets, and η q = 0.01 for jets produced by light quarks and gluons. The actual tagging algorithm implemented in the Delphes simulation is not based on a particular experimental method to identify b-jets. The algorithm determines if a jet is close enough in ∆R to a "true" b parton. When this is the case, the efficiencies given above are applied to determine if the tagging is successful or not.

Backgrounds
Based on the event signature, SM production of tt with at least one hadronically decaying W boson (or additional jet activity) constitutes an irreducible background to the Higgs signal. We can a priori expect this to be the most important SM background since the scale for the SUSY-QCD processes is high (> 1 TeV). In principle there are other sources of background from production of W + jets (bb), Z + jets (bb), direct production of bb + jets, or from QCD multijets. The cross sections for these processes are large compared to the signal cross section, with QCD multijets the largest and thereby potentially the most serious. However, for QCD jet production to constitute a background to the Higgs signal simultaneously a double misidentification of heavy flavour jets and a large mismeasurement of the missing transverse energy is required. It is furthermore difficult to simulate this background reliably, since extreme kinematical fluctuations -or experimental effectswould be necessary to produce the signal-like events. A detailed study of the experimental effects would require a full detector simulation, which is beyond the scope of the present paper. However, the dominance of the tt background over other SM processes, such as W + jets or Z + jets, for our final state has also been demonstrated experimentally by the results from SUSY searches with b-jets and missing E T [68]. We therefore proceed under the assumption that the cuts devised to suppress the irreducible tt background will also be efficient for suppressing the other SM backgrounds as well.
For the normalization of the tt background we use the NLO cross section σ(pp → tt) = 902 pb ( √ s = 14 TeV) and σ(pp → tt) = 162 pb ( √ s = 7 TeV), computed with the HATHOR package [82] for m t = 173.3 GeV and MSTW2008 PDFs [83]. In this way a consistent NLO normalization is used for both the signal and background events. The tt background is generated in the same Monte Carlo framework as already described for the signal.
In addition to the SM backgrounds, the process we are interested in receives an important background from the SUSY cascade itself. Any final state containing two b-jets which do not result from an intermediate Higgs boson contributes to this background. Attempting to suppress the SUSY background events would require additional cuts that depend on the kinematics of the decay chains. This is something which may indeed be possible to devise once information on the supersymmetric spectrum has become available, but since we do not want to make any particular assumptions on the pattern of the SUSY spectrum, no selection will be applied aiming to reduce the SUSY background. Instead we will consider the inclusive bb mass spectrum directly after applying the cuts designed to reduce the SM background to determine if a Higgs signal can be extracted.

Event Selection
As a first step, we perform a preselection of the expected event topology, demanding n j ≥ 2, n b ≥ 2. All reconstructed jets are required to have a minimum p T ≥ 25 GeV. Figure 5 shows the p T distribution for the hardest jet in each event, comparing the inclusive SUSY events (with M SUSY = 750 GeV) to the tt background. We show the results for the two cases √ s = 7 TeV (left) and √ s = 14 TeV (right). In order to illustrate the effect of applying cuts to this variable, each histogram is normalized to unity. From figure 5 it is clear that the leading jet from the SUSY events has a much harder scale compared to the tt events. This can be understood as a result of the large boost obtained by the light  quark jets originating from squark decays. It can also be seen that there is only a minor scaling difference in the jet p T distribution between the 7 TeV and 14 TeV cases. The same is true for the second hardest (light) jet, for which the corresponding p T distribution is shown in figure 6. Similar differences between signal and background can be observed also for the third and fourth jet when they are present.
With each cascade ending in the stable LSP, a large missing transverse energy / E T is expected for the signal events. This distribution is displayed in figure 7, and shows indeed that the SUSY distribution peaks at high / E T values ( 200 GeV). This is therefore an important discriminating variable to suppress the background from tt events, where the missing transverse energy is due to neutrinos from leptonic W decays. As already mentioned, a hard cut on / E T is also necessary to suppress the background from ordinary QCD multijet events and direct production of bb. A further advantage of the large / E T is that it can be used for triggering.
The final kinematical distribution we are going to consider is displayed in figure 8. It  Table 4. Number of events remaining after each step of the event selection at √ s = 7 TeV. The SUSY events are classified as signal or background based on the presence of (at least one) Higgs boson in the decay chain. The total number of generated events in the inclusive sample is arbitrary.
shows the separation in ∆R = (∆η) 2 + (∆φ) 2 between pairs of b-jets. For events with n b > 2 all possible combinations have been included. The signal distribution is seen to peak near the minimum separation of ∆R = 0.4 set by the jet measure, while the tt background prefers the b-jets to be more back-to-back and peaks at ∆R ∼ π.
The precise cuts applied -and their effect on the event selection -are shown for the SUSY events in table 4 (for the 7 TeV case) and table 5 (14 TeV). Table 6 gives the corresponding information for the SM tt background. Note that the number of generated events in these tables does not correspond to any particular luminosity, but is rather  selected to give adequate statistics for the event selection. The inclusive SUSY sample is split into signal and background, where the signal consists of the events containing at least one Higgs boson (as determined from Monte Carlo truth information). In the last row we give the accumulated total efficiencies of all the cuts. Looking first at table 4, we see that an efficiency of 5.5 × 10 −2 is obtained for the case with M SUSY = 750 GeV. This efficiency is more than doubled (0.12) for the case with M SUSY = 1 TeV, since the heavier squarks give harder jets as decay products which leads to more events passing the jet p T cuts. The larger boost given to the LSP at the end of the decay chain also leads to an increased / E T . The same qualitative features are visible at 14 TeV, as can be read off table 5. Due to the favorable signal statistics at 14 TeV, 6 we can afford slightly harder cuts on / E T and ∆R(bb) in this case, something which is also needed to maintain a good background suppression. One should therefore not be discouraged by the somewhat lower efficiencies recorded in this case (4.2 × 10 −2 for M SUSY = 750 GeV vs. 9.8 × 10 −2 for M SUSY = 1 TeV). The signal efficiencies can be compared to those for the tt background, given in table 6, which are at the 10 −5 level for both energies. It is clear from this table that the hard cuts on the jet p T and the / E T distribution are the most important handles available to suppress the background.
As discussed in the previous section, we do not apply any specific cuts to suppress the background from SUSY events that do not involve a Higgs boson. The numbers given in tables 4 and 5 show that nevertheless our event selection gives rise to an improvement also in the ratio of signal events over SUSY-background events. The largest difference in  selection efficiency between the SUSY signal and background arises from the typical number of b quarks produced in the two cases, which is larger for the events where Higgs bosons are produced, leading to a stronger reduction of the SUSY background by the jet multiplicity cut. The cut on ∆R also contributes to the difference. This cut has the pleasant "side effect" to enrich the SUSY sample in Higgs events since the jets resulting from H 1 → bb decays are more likely to show up for small ∆R than those from two unpaired b-jets. The same distributions are shown in figures 11 (for the LHC at 7 TeV) and 12 (for 14 TeV), but here with stacked histograms to more closely resemble "real" data. Here we have furthermore split up the inclusive SUSY sample into signal events (displayed in red), characterised by the presence of (at least) one Higgs boson in the decay chain, and the remaining SUSY background events (black). The latter constitutes an additional source of background besides the SM tt background (light gray). In figure 11 we see that the most striking feature is the H 1 peak. Although the tt background peaks at roughly the same position as the signal, the statistics of signal events should be sufficient for establishing a signal over the background. In the 14 TeV case, figure 12 illustrates the features observed already in figure 10. The  With the assumed statistics, no peaks are observed in the bb mass spectrum for the heavier Higgs bosons H 2 and A 1 . This is mainly due to the smallness of the branching ratios into the bb mode because of the open Higgs decay channels. Part of the difficulty in observing the heavier resonances is also a result of selecting the combination minimizing ∆R(bb) in configurations with multiple b-jets, which favors selection of the light H 1 .

Results
In order to obtain an estimate of the significance of the H 1 mass peak we have performed a Gaussian fit to the maximum of the distributions in figures 11 and 12. Table 7 lists the results extracted from the fit for the mean value M H and the 1 σ width ∆M H of the Gaussian peak. We find that the fitted central values reproduce well the correct H 1 mass for all cases (recall that the input mass used in our numerical simulation is M H 1 = 40 GeV).  tics available in the low energy running, and the more coarse binning in M bb required to observe the peak. The number of signal and background events in the peak region is obtained by integrating M bb over the interval [M H − ∆M H , M H + ∆M H ], corresponding to ±1 σ of the Gaussian distribution. As explained above, the combined background includes both the events from SM tt and the part of the inclusive SUSY sample containing no Higgs bosons in the cascades. The event numbers are combined into the ratios of signal/background (S/B) and S/ √ B given in table 7. We use S/ √ B as a simple illustration for the expected significance and in particular for comparing between the four example cases we consider here and with other theoretical studies using the same criterion. Clearly, claiming an actual discovery would require a more sophisticated statistical treatment. We regard it nevertheless as encouraging that a significance of S/ √ B > 5 is achieved for three of the four cases considered in

Summary and Conclusions
The NMSSM is both theoretically appealing as an extension of the SM and interesting phenomenologically, as its spectrum may contain Higgs bosons with mass much below the limits in the SM or the MSSM. We have investigated an NMSSM scenario with a light CP-even Higgs in the mass range 20 GeV < M H 1 < M Z . Scenarios like this may be missed with the standard Higgs search channels at the LHC, in particular due to a potentially large branching ratio of the heavier H 2 state, that has SM-like couplings to gauge bosons, into a pair of light Higgses. We have pointed out that there are good prospects for discovering such a light Higgs boson in SUSY cascade decays at the LHC.
We have performed a Monte Carlo simulation of the signal and the dominant background to the level of fast detector simulation, taking into account also background from other SUSY events that do not involve cascade decays containing a Higgs boson. For our numerical analysis we adapted the "P4" benchmark point proposed for the NMSSM, choosing M H 1 = 40 GeV as example value for the mass of the light Higgs. Production of squarks and gluinos via the strong interaction at the LHC may give rise to cascade decays involving heavy neutralinos and charginos decaying into lighter ones and a light Higgs. We have investigated the impact of various kinematical variables on discriminating between the inclusive SUSY signal (including events both with and without a Higgs boson in the cascade) and the SM background from tt production. A set of simple cuts has been devised that turned out to be efficient for establishing the inclusive SUSY signal. We did not assume any specific knowledge about the background from SUSY events without a Higgs in the cascades. Accordingly, besides favoring events containing the light H 1 by selecting the combination minimizing ∆R(bb) in configurations with multiple b-jets, we have not applied any particular cuts for suppressing the SUSY background.
Our results show that reconstruction of the decay of the light Higgs into bb may be feasible. Such an observation would be a direct experimental sign of the bottom Yukawa coupling, which is difficult to access in standard search channels. We have investigated two values of the soft SUSY-breaking parameter in the squark sector, M SUSY = 750 GeV and M SUSY = 1 TeV, while we set the gluino mass parameter to 1 TeV. A modest integrated luminosity of 5 fb −1 has been considered for LHC running both at 7 TeV and 14 TeV. We find a statistical significance for the H 1 mass peak of S/ √ B ≈ 4 for M SUSY = 1 TeV at √ s = 7 TeV. This significance increases to S/ √ B ≈ 8 for M SUSY = 750 GeV at 7 TeV and reaches a level of almost 30 for both values of M SUSY at 14 TeV. While the example values that we have chosen for M SUSY and the gluino mass are close to the current search limits from the LHC, the large statistical significance that we have found for the 14 TeV case indicates that there is certainly scope to extend our analysis to scenarios with heavier squarks and gluinos or to scenarios with reduced branching ratios of the neutralinos into Higgs bosons. Since the high-energy run of the LHC is not imminent, we leave a more detailed analysis of this reach for future work.
The results presented here have been obtained in a specific benchmark scenario, but it is easy to see that they are more generally applicable. First of all, the value M H 1 = 40 GeV used in our numerical analysis was chosen just for illustration. Our results are rather insensitive to the precise value of M H 1 . Since the production relies on the decay of heavier SUSY states, with branching ratios largely independent of M H 1 , the Higgs production rates remain similar for the whole mass range M H 1 < M Z . The event selection and signal identification through H 1 → bb proceeds along similar lines as we have discussed.
Concerning the settings of the other SUSY parameters, our results will be similar for other scenarios fulfilling a few simple criteria: Obviously, the neutralinos and charginos have to be sufficiently lighter than the squarks and gluinos in order to be produced at all in the cascade decays of the latter. The squark decays also provide the hard jets utilized in the event selection. With the present limits from the LHC searches on the masses of the gluino and the squarks of the first two generations this criterion is almost automatically fulfilled for any model of interest. Furthermore, the neutralino and chargino mass hierarchy and mixing character must be such that the squark decays go through heavier neutralinos or charginos, and the decays of the latter into a light Higgs and a lighter neutralino or chargino are open. Such a scenario is disfavored if the LSP is gaugino-like. In order to generate a sufficient number of Higgs bosons in the cascade decays, it is also advantageous for (at least one of) the gauginos to be heavier than the Higgsinos, so that an intermediate Higgsino decay step can be present. In the NMSSM such a situation can be realized quite easily if the LSP is singlino-like.
While the results presented in this paper are based on a rather simple-minded analysis, involving for instance just a fast detector simulation, we nevertheless regard them as very encouraging, motivating a further exploration of the potential for detecting a light non-SM type Higgs in SUSY cascade decays. In fact, there exists the exciting possibility that the discovery of a SUSY signal could go hand in hand with the discovery of one or more Higgs bosons.