KKIAS-P17026
Light exotic Higgs bosons at the LHC
Shoaib Munir
School of Physics, Korea Institute for Advanced Study, Seoul 130-722, Republic of KoreaE-mail: [email protected]
Abstract.
Most models of new physics contain extended Higgs sectors with multiple Higgsbosons. The observation of an additional Higgs boson, besides the ∼
125 GeV ‘ h obs ’, willthus serve as an irrefutable evidence of physics beyond the Standard Model (SM). However,even when fairly light, these additional Higgs bosons may have escaped detection at the LargeElectron-Positron (LEP) collider, the Tevatron and the Large Hadron Collider (LHC) hitherto,owing to their highly reduced couplings to the SM particles. Therefore, in addition to thesearches based on the conventional production processes of these Higgs bosons, such as gluonor vector boson fusion, possible new search modes need to be exploited at collider experimentsin order to establish their signatures. We investigate here the phenomenology of pseudoscalars,with masses ranging from O (1) GeV to about 150 GeV, in the Next-to-Minimal SupersymmetricSM (NMSSM) and the Type-I 2-Higgs Doublet Model (2HDM) in some such atypical searchchannels at the LHC Run-II.
1. Introduction
In any new physics model with an extended Higgs sector, (at least) one of the neutral Higgsbosons should have a mass and signal rates in its various decay channels consistent with those ofthe h obs discovered at the LHC [1, 2]. This requirement, coupled with that of the satisfaction ofconstraints coming from precision electroweak (EW) and b -physics experiments, almost alwaysleads to the other two of the three neutral Higgs bosons being rather heavy in the minimalrealisation of Supersymmetry (SUSY). However, in the NMSSM, where the augmentation bya singlet scalar superfield results in a total of five neutral Higgs states, scalars H − (orderedin terms of increasing mass) and pseudoscalars A , , the singlet-like states can be fairly light(for a review, see [3]). A light A can prevent the over-closure of the universe when the lightestneutralino, χ , which is a crucial dark matter (DM) candidate, has a mass so low that neitherthe Z boson nor the h obs can contribute sufficienty to its annihilation.Without invoking SUSY, one can simply introduce a second Higgs doublet in the SM, with a Z -symmetry preventing the dangerous flavor changing neutral currents (for a review, see [4]).In a Type-I 2HDM (2HDM-I), this Z symmetry is imposed in such a way that all the fermionscouple only to one of the two Higgs doublets. The physical masses of the three neutral Higgsbosons can be taken to be the input parameters in this model. One can therefore assign anymasses to the scalars h and H (with m h < m H ) and the pseudoscalar A in order to study thephenomenological implications for different values of the other free parameters.We analyse some of the potential discovery channels of a light pseudoscalar at the LHC Run-II in these models. In the NMSSM we consider the production of A A and A Z pairs in thedecays of the heavier CP-even scalars [5] and of very light A and DM in the decays of the1 a r X i v : . [ h e p - ph ] A p r eavier neutralinos [6]. In the 2HDM-I we discuss the electroweak (EW) production of an hA pair with a combined mass less than that of the Z boson [7].
2. Analysis methodology
For the first of our two NMSSM analyses, in order to perform a fast and efficient numericalscanning of the parameter space for obtaining A solutions spanning a wide range of masses, weadopted a model version with partial universality. In this version, unified parameters m , m / and A , corresponding to the scalar masses, gaugino masses and trilinear couplings, respectively,are input at the grand unification scale. The Higgs sector soft trilinear parameters, A λ and A κ ,though not unified with A , are also input at the same high scale. In contrast, the dimensionlessYukawa couplings λ and κ , and the parameters µ eff ≡ λs and tan β ≡ v u /v d , with v u and v d being the vacuum expectation values of the two Higgs doublets and s of the singlet, are defined atthe EW scale. These input parameters were fed into the NMSSMTools [8] program to calculatethe physical masses, couplings and branching ratios (BRs) of all the Higgs bosons. For the2HDM-I, the Higgs boson couplings and BRs were obtained for the scanned ranges of the inputparameters using the 2HDMC public code [9].During the scanning process, each point was first subjected to the basic theoretical conditionslike unitarity, perturbativity and vacuum stability. For every point fulfilling these conditions,we further required the H ( H ) in the NMSSM (2HDM-I) to have its mass and signal strengths,calculated with HiggsSignals [10], consistent with the latest available measurements for the h obs from the LHC (see, e.g., [11]). A successful model point was then tested for consistency of theremaining scalar, pseudoscalar and charged Higgs bosons with the exclusion limits from collidersearches, using the HiggsBounds program [12]. It was also required to satisfy the constraintson the most important b -physics observables, the predicted values for which were calculated fora point in each of the models using the SuperIso [13] program. In addition, for the 2HDM-Ipoints, the values of the oblique parameters S , T and U , calculated by 2HDMC, were checkedagainst exclusion limits from the experimental measurements [14]. In the case of the NMSSM,on the other hand, a point was rejected if the relic abundance due to the χ was computedby MicrOmegas [15] to be larger than the measurement by the PLANCK telescope [16]. Thescanned ranges of the free parameters in the two models are given in table 1.Table 1: Scanned ranges of the free parameters in the NMSSM (a) and the 2HDM-I (b). (a) NMSSM parameter Scanned range m (GeV) 200 – 4000 m / (GeV) 100 – 2000 A (GeV) − β λ κ µ eff (GeV) 100 – 2000 A λ (GeV) − A κ (GeV) − (b) m h (GeV) 10 – 2 M Z / m A (GeV) m h / M Z − m h ) m H ± (GeV) 90 – 150sin( β − α ) − .
25 – 0 m GeV m A sin β cos β tan β ( − . – − . β − α ) The second NMSSM analysis included here is dedicated to a very specific scenario in whichthe A and the χ both have masses O (1) GeV. Thus only a couple of benchmark points (BPs)corresponding to the general NMSSM, with all the input parameters lying at the EW scale, will2e discussed. For both these BPs, the role of the h obs is once again played by the H , and allthe relevant constraints noted above are satisfied.
3. Light pseudoscalars in the NMSSM
In the NMSSM, the tree-level mass-squared of the A is written (assuming negligible singlet-doublet mixing) as m A (cid:39) A λ √ s v λ sin 2 β + κ (2 v λ sin 2 β − sA κ ) , (1)where v ≡ (cid:113) v u + v d (cid:39)
246 GeV. Thus, m A can be varied with more freedom compared to themass of the singlet-like scalar, which is also indirectly constrained by the h obs measurements dueto the relatively stronger singlet-doublet mixing. By adjusting the trilinear couplings A λ and A κ , A can take a broad range of values without being in conflict with the experimental data. For this analysis, we restricted ourselves to m A <
150 GeV, so that the production cross sectionfor the heavier Higgs bosons that decay into it did not get too suppressed kinematically. Afterperforming the parameter space scan to find model points satisfying all the imposed conditions,we carried out a dedicated signal (S)-to-background (B) analysis for the LHC with √ s = 14TeV. We first calculated the gluon-fusion production cross section for each of the H − using thepublic program SusHi [17]. The backgrounds coming from the pp → b , pp → b τ , pp → τ , pp → Z b and pp → Z τ processes were computed using MadGraph5 aMC@NLO [18]. Thehadronization and fragmentation of the signals and backgrounds was then done using Pythia8.180 [19] interfaced with FastJet [20].We used the two most dominant decay channels of the A , namely b ¯ b and τ + τ − . In the case ofthe b ¯ b decay, we employed also the jet substructure method [21], which gives enhanced sensitivityfor larger masses of the decaying Higgs bosons, by assuming one fat jet from boosted b -quarksinstead of two single b -jets. For three representative values of the accumulated luminosity atthe LHC, L = 30/fb, 300/fb and 3000/fb, we then estimated the signal cross sections whichwould give the statistical significance, S/ √ B , greater than 5 for a given mass of A in each ofthe various final state combinations.In figure 1(a) we show the cross section for the H → A A process. Also shown, in thisfigure and the subsequent ones, are the sensitivity curve(s), which assume BR( A → b ¯ b ) = 0 . b ¯ b pair and BR( A → τ + τ − ) = 0 . τ + τ − pair, corresponding to the bestfinal state combination for probing the given process. Here these curves are for the 2 b τ finalstate (one corresponding to two single b -jets and the other, showing an enhanced sensitivity forsmaller m A , to one fat jet) at L = 30/fb, and for the 4 τ final state at L = 3000/fb. In theframe (b) we show the H → A A cross section. Note that in the case of the H , the possibilityto reconstruct its mass (125 GeV) provides an important kinematical handle. We see that the2 b τ final state can be probed at the LHC with L as low as 30/fb, owing to the use of thejet substructure method, despite the fact that the H → A A decay is tightly constrained bythe h obs signal rate measurements at the LHC. However, the maximum A mass that can beaccessible in this channel is m H / ∼ . m A might beaccessible in the H /H → ZA decay channels. We therefore ignore these channels and turn tothe H for the production of heavier A .Figure 2(a) shows that the H → A A channel does not carry any promise. This is due to thefact that for such high masses of H ( ≥
400 GeV) the production cross section gets diminishedand, at the same time, other decays of H dominate over this channel. The sensitivity curve inthe figure corresponds to the 2 b τ final state for L = 3000/fb. Conversely, as seen in the frame(b), for the H → A Z channel, with the Z decaying into e + e − or µ + µ − states, a number of3 a) (b) Figure 1: Cross sections for (a) the gg → H → A A process, and (b) the gg → H → A A process, as functions of m A , for points obtained from the NMSSM scan. Taken from [5]. (a) (b) Figure 2: Cross sections for (a) the gg → H → A A process, and (b) the gg → H → A Z process, as functions of m A , for points obtained from the NMSSM scan. Taken from [5].points lie above the 2 b (cid:96) sensitivity curve for L = 300/fb. Again, the use of the fat jet analysis,along with a sizable H A Z coupling resulting from a significant doublet component in A , makean A lying in the ∼ −
100 GeV mass range discoverable in this channel.
Next we focus on the A with mass O (1) GeV in the NMSSM, which also contains fiveneutralinos, (cid:101) χ − , in total. The lightest of these states, given by the linear combination (cid:101) χ = N (cid:101) B + N (cid:102) W + N (cid:101) H d + N (cid:101) H u + N (cid:101) S , (2)of the gaugino ( (cid:101) B , (cid:102) W ), higgsino ( (cid:101) H u , (cid:101) H d ) and singlino ( (cid:101) S ) interaction eigenstates, is the DMcandidate when R -parity is conserved. The presence of the singlino fraction, Z s = | N | inthe (cid:101) χ , which is non-existent in the MSSM, leads to some interesting new possibilities in the4ontext of DM phenomenology. A look at the NMSSM neutralino mass matrix reveals that the[ M (cid:101) χ ] term, corresponding to the singlino eigenstate, is equal to 2 κs = 2 κµ eff λ . This impliesthat the singlino fraction in (cid:101) χ can be increased by reducing κ and/or µ eff and increasing λ .Since the mass of A also scales with κs , as noted above, a light A can naturally accompany alight singlino-like (cid:101) χ . This DM can thus undergo sufficient annihilation, via A in the s -channel,to generate the correct relic abundance of the universe.At the LHC, one of the main ways to probe the DM is in the decays of the heavier neutralinosand charginos. In particular, dedicated searches have been performed by both the CMS andATLAS collaborations [22, 23] for the pp → (cid:101) χ , + (cid:101) χ ± process which is followed by the decays (cid:101) χ , → Z + (cid:101) χ → (cid:96) + (cid:96) − + /E T and (cid:101) χ ± → W ± + (cid:101) χ → (cid:96) ± + /E T , where /E T implies missingtransverse energy. These searches have already put strong constraints on significant regions ofthe NMSSM parameter space, since the Z + (cid:101) χ decay channel is by far the dominant one of (cid:101) χ , .However, in the scenario with a very light singlino-like DM, the (cid:101) χ , → A + (cid:101) χ decay channel,although still subdominant, can become sizable for sufficiently large values of λ . The reason isthat the (cid:101) χ , (cid:101) χ and (cid:101) χ ± are predominantly higgsinos, since µ eff is much smaller than the gauginosoft masses M and M , in order to maximize the singlino fraction in (cid:101) χ while minimizing itsmass. The issue with the A + (cid:101) χ decay mode though, is that the main leptonic decay channel, A → µ + µ − , is highly suppressed, with its BR never exceeding 9%. Furthermore, the muonsthus produced are highly collinear and hence the isolation of this signal from the background isextremely challenging.We show here that, for the NMSSM parameter space points yielding O (1) GeV (cid:101) χ and A ,once the above complication can be overcome, the A + (cid:101) χ search channel can be more promisingthan the Z + (cid:101) χ one at the LHC. We refer to the former as the µ col channel and to the latteras the trilepton (3 (cid:96) ) channel. For this purpose, from the NMSSM parameter space, we choseBP1 such that the (cid:101) χ , → A (cid:101) χ decays are typically suppressed (both having BRs of 0.004),while BP2 has relatively enhanced respective BRs of 0.089 and 0.081. The BRs correspondingto the (cid:101) χ , → Z (cid:101) χ decays for both the points are in excess of 60%, while the BR ( A → µ + µ − )is 0.039 for BP1 and 0.087 (i.e., near its maximum possible value) for BP2. For these BPs wethen performed detector-level analyses of the two processes shown in figure 3. ˜ χ ± W ± ˜ χ ˜ χ / ˜ χ Z l ± ν l l + pp l − (a) ˜ χ ± W ± ˜ χ ˜ χ / ˜ χ A l ± ν l pp µ + µ − (b) Figure 3: Diagramatic representation of the processes containing, in the final state, (a) twoleptons coming form a Z boson, and (b) two collinear muons coming from an A .For the 3 (cid:96) channel, we first generated parton-level signal and background events at the 14TeV LHC for each of the BPs using MadGraph5 aMC@NLO and passed these to Pythia 6.4524] for hadronization. The most dominant irreducible backgrounds for this channel come fromthe di-boson, tri-boson and t ¯ tW/Z productions, all of which can have three or more leptons and /E T in the final states. To obtain the signal and background efficiencies, the ATLAS detectorsimulation was then performed with DELPHES 3 [25] via the CheckMATE program [26], whereinthe six distinct signal regions defined in the ATLAS search [23] have already been implemented.By multiplying the next-to-leading order cross sections for the signal process, calculated usingProspino [27], and the backgrounds with an assumed L = 300 fb − , we also obtained the numberof events for both in each of the signal regions.As for the µ col channel, in order to isolate the highly collimated muons, we employed thetechnique of clustering them together into one object, similar in concept to the constructionof a lepton-jet [28]. For applying this method, the signal events generated for BP1 and BP2were passed to Pythia 6.4 for hadronization and subsequently to DELPHES 3 for jet-clusteringusing Fastjet. Then, the µ col object was defined by requiring the transverse momentum, p T ,for each muon in the signal to be larger than 10 GeV, and the cut m µ ¯ µ < I sum < I sum being the scalar sum of the transverse momenta of all additional charged tracks, eachwith p T > . µ col and satisfying∆ R = 0 .
4. The main backgrounds, containing two collinear muons along with a third leptonand /E T , include W ( → (cid:96) ± v ) γ ∗ , Z ( → (cid:96) + (cid:96) − ) γ ∗ and W b ¯ b and Zγ ∗ . These backgrounds were alsogenerated with Pythia and Fastjet and subjected to certain cuts that maximize the isolation ofthe signal process from them, as explained in [6].The yield of each of the above analysis methods applied to the two search channels isquantified in terms of S/B , which is given in table 2 for the two BPs. For the 3 (cid:96) channel,this
S/B corresponds only to the signal region that gives the highest sensitivity, and we notethat it is slightly higher than the
S/B in the µ col channel for the BP1. For the BP2, however, the µ col analysis gives a considerably larger S/B than the 3 (cid:96) one, which is evidently a consequence ofthe sizable BR( (cid:101) χ , → A (cid:101) χ ) and BR( A → µ + µ − ). Thus dedicated searches in the µ col channelmay prove very crucial for the discovery of a very light DM in non-minimal supersymmetry atthe LHC. Note that while an estimation of the statistical significance would be a more realisticindicator of the strengths of the two signal processes compared to the S/B , it is not includedhere since there is no consistent way of treating the systematic uncertainties.Table 2: Measures of the strengths of the two analyses considered here, along with the masses,in GeV, relevant to them, corresponding to the two selected BPs.BP m (cid:101) χ m (cid:101) χ m (cid:101) χ m (cid:101) χ ± m A m H S/B (3 (cid:96) ) S/B ( µ col )1 1.00 189.1 − . − .
4. Scalar-pseudoscalar pair-production in the Type-I 2HDM
The Landau-Yang theorem [29, 30] prevents the contribution of an on-shell Z boson to thegluon-initiated production of a hA pair when the sum of their masses is smaller than m Z . The q ¯ q -initiated process, however, does not suffer from this limitation, and the cross section for hA pair-production can therefore get considerably enhanced due to a resonant Z boson in the s -channel. Our analysis of the Type-I 2HDM aimed at exploring this possibility, and hence theparameter space scan for this model also observed the condition m h + m A < m Z .In figure 4(a) we show the good points from the scan for which the Γ( Z → hA ) additionallylies within the 2 σ error on the experimental measurement of the total width of the Z boson,6 Z = 2 . ± . q ¯ q → hA process at the LHC with √ s = 13 TeV, calculated usingMadGraph5 aMC@NLO, which evidently grows as m h + m A gets smaller. Near the top leftcorner of the figure m A > m h , and we see a high density of points. The points disappear whenthe H → AA decay channel opens up (for m A < m H /
2) , potentially leading to a significantreduction in the signal strengths of H in the SM final states. The points start reappearing for m A <
35 GeV, near the bottom right corner of the figure, when the
HAA decay gets sufficientlysuppressed. But they disappear again for m A < m h /
2, where the h → AA decay channel,severely constrained by the LEP searches, is kinematically available.Figure 4(b) shows that the q ¯ q → hA production cross section at the 13 TeV LHC can exceedthe gg → hA one, calculated using [31], by a few orders of magnitude, reaching up to about 90pb. In table 3 we list the cross sections corresponding to the two production modes for the threeBPs for this model. The difference between the two cross sections is much more pronouncedfor the BP1, wherein m A < m h , compared to that for BP2 and BP3 with m h < m A . Onecan also note from the table that for BP1, Z ∗ A is the primary decay channel of h , with thedominant mode for the subsequent decay of the A being the b ¯ b pair. Thus Z ∗ b ¯ bb ¯ b , Z ∗ b ¯ bτ + τ − and Z ∗ τ + τ − τ + τ − could be the main signatures of interest. Similarly, for BPs 2 and 3 Z ∗ h is theprominent decay mode of A , so that the most common final states remain the same generally.For BP3 though, the highly fermiophobic h (owing to sin( β − α ) →
0) has a large BR into twophotons, which could make the Z ∗ γγγγ final state an important unconventional probe of thisscenario in the 2HDM-I.
10 20 30 40 50 60 m h [GeV] m H m A [ G e V ] Γ( Z → hA ) >δ Γ Z (cos( β − α ) =1) m h + m A >m Z h → AA σ σ BP 1BP 2BP 3 σ ( q ¯ q → h A ) [ pb ] (a) -4 -3 -2 σ ( gg → hA ) [pb]10 σ ( q ¯ q → h A ) [ pb ] m A [ G e V ] (b) Figure 4: (a) Successful scan points with Γ( Z → hA ) lying within the δ Γ Z at the 1 σ (lighter)and 2 σ (darker) levels. The color map corresponds to the total cross section for the q ¯ q → hA process and the three BPs have been highlighted in yellow. (b) QCD vs. EW production crosssections of the hA pairs, with the color map showing the mass of A . Taken from [7]. Acknowledgments
The author would like to thank his collaborators, Nils-Erik Bomark, Rikard Enberg, ChengchengHan, Doyoun Kim, William Klemm, Stefano Moretti, Myeonghun Park and Leszek Roszkowski,each of whom were involved in one (or more) of the analyses reviewed in this contribution.7able 3: Cross sections (in pb) for the gg - and q ¯ q -initiated hA pair-production, correspondingto the three BPs. Also given are the leading BRs of h and A for each BP.BP m h m A σ ( q ¯ q ) σ ( gg ) BR( h → Z ∗ A, b ¯ b, γγ, τ τ ) BR( A → Z ∗ h, b ¯ b, τ τ )1 54.2 33.0 41.2 1 . × − < . < .
01 0, 0.86, 0.072 22.2 64.9 34.4 7 . × −
0, 0.83, 0.03, 0.07 0.86, 0.12, 0.013 14.3 71.6 31.6 1 . × −
0, 0.60, 0.24, 0.07 0.90, 0.08, 0.01
References [1] Chatrchyan S et al. [CMS Collaboration] 2012
Phys. Lett. B et al. [ATLAS Collaboration] 2012 Phys. Lett. B Phys. Rept.
Phys. Rept.
J. High Energy Phys.
JHEP07(2015)002[6] Han C, Kim D, Munir S and Park M 2015
J. High Energy Phys.
JHEP04(2015)132[7] Enberg R, Klemm W, Moretti S and Munir S 2017
Phys. Lett. B Comput. Phys. Commun.
Eur. Phys. J. C et al. [CMS and ATLAS Collaborations] 2016 J. High Energy Phys.
JHEP08(2016)045[12] Bechtle P, Brein O, Heinemeyer S, Stal O, Stefaniak T, Weiglein G and Williams K E 2014
Eur. Phys. J. C Comput. Phys. Commun. (2009) 1579–1613[14] K. Olive et al.
Chin. Phys. C Comput. Phys. Commun et al. [Planck Collaboration] 2016.
Astron. Astrophys.
A13[17] Harlander R V, Liebler S and Mantler H 2013
Comp. Phys. Commun. et al.
J. High Energy Phys.
JHEP07(2014)079[19] Sjostrand T, Mrenna S, and Skands P Z 2008
Comput. Phys. Commun.
Eur. Phys. J. C Phys. Rev. Lett. et al. [CMS Collaboration] 2014
Eur. Phys. J. C no. 9 3036[23] Aad G et al. [ATLAS Collaboration] 2014 J. High Energy Phys.
JHEP04(2014)169[24] Sjostrand T, Mrenna S and Skands P Z 2006
J. High Energy Phys.
JHEP05(2006)026[25] de Favereau J et al.
J. High Energy Phys.
JHEP02(2014)057[26] Drees M, Dreiner H, Schmeier D, Tattersall J and Kim J S 2014
Comput. Phys. Commun.
Nucl. Phys. B [28] Falkowski A, Ruderman J T, Volansky T and Zupan J 2010 J. High Energy Phys.
JHEP05(2010)077[29] Landau L D 1948
Dokl. Akad. Nauk Ser. Fiz. no. 2 207[30] Yang C N 1950 Phys. Rev. J. High Energy Phys.