CP-violating MSSM Higgs at Tevatron and LHC
aa r X i v : . [ h e p - ph ] O c t CP-violating MSSM Higgs at Tevatron and LHC
Siba Prasad Das and Manuel Drees AHEP Group, Institut de F´ısica Corpuscular – C.S.I.C./Universitat de Val`enciaEdificio Institutos de Paterna, Apt 22085, E–46071 Valencia, Spain, Bethe Center for Theoretical Physics and Physikalisches Institut, Universit¨at Bonn,Nussallee 12, D–53115 Bonn, GermanyE-mail: [email protected], [email protected]
Abstract.
We analyze the prospect for observing the intermediate neutral Higgs boson ( h ) in its decayto two lighter Higgs bosons ( h ) at the presently operating hadron colliders in the framework ofthe CP violating MSSM using the PYTHIA event generator. We consider the lepton+ 4-jets+ E/ T channel from associate W h production, with W h → W h h → ℓν ℓ b ¯ bb ¯ b . We require two orthree tagged b -jets. We explicitly consider all relevant Standard Model backgrounds, treating c -jets separately from light flavor and gluon jets and allowing for mistagging. We find thatit is very hard to observe this signature at the Tevatron, even with 20 fb − of data, in theLEP–allowed region of parameter space due to the small signal efficiency. At the LHC, a priorihuge SM backgrounds can be suppressed by applying judiciously chosen kinematical selections.After all cuts, we are left with a signal cross section of around 0.5 fb, and a signal to backgroundratio between 1.2 and 2.9. According to our analysis this Higgs signal should be viable at theLHC in the vicinity of present LEP exclusion once 20 to 50 fb − of data have been accumulatedat √ s = 14 TeV.
1. Introduction
The Minimal Supersymmetric Standard Model (MSSM) [1] requires two Higgs doublets, leadingto a total of five physical Higgs bosons. At the tree level, these can be classified astwo neutral CP–even bosons ( φ and φ ), one neutral CP–odd boson ( a ) and two chargedbosons. In the presence of CP violation, the three neutral Higgs bosons can mix radiatively[2, 3]. The mass eigenstates h , h and h with m h < m h < m h can then be obtainedfrom the interaction eigenstates φ , φ and a with the help of the orthogonal matrix O αi ,( φ , φ , a ) Tα = O αi ( h , h , h ) Ti , which diagonalizes the Higgs boson mass matrix. O depends onvarious parameters of the SUSY Lagrangian.Due to this mixing, the Higgs mass eigenstates are no longer CP eigenstates. Moreover,the masses of the Higgs bosons, their couplings to SM and MSSM particles, and theirdecays are significantly modified [3]. For example, the Higgs boson couplings to pairs ofgauge bosons are scaled by g h i V V relative to the SM. These couplings can be expressed as g h i V V = cos β O φ i + sin β O φ i , where tan β is the ratio of Higgs vacuum expectation values(VEVs). The magnitude of g h W W is directly related to the production process studied in thispaper.In the absence of mixing between neutral CP–even and CP–odd states the LEP experimentswere able to derive absolute lower bounds of about 90 GeV on the masses of both the lighterP–even Higgs and the CP–odd boson [4]. However, in the presence of CP violation, the LEPexperiment were not able to exclude certain scenarios with very light h . In this “LEP hole” h is dominantly a CP–odd state with almost vanishing coupling to the Z boson. One then has tosearch for Zh or h h production. In part of the LEP hole, these cross sections are suppressedby the rather large h mass. Moreover, h → h h decays lead to quite complicated final states,which often yield low efficiencies after cuts. One LEP–allowed region has m h less than 10 GeV,so that h → τ + τ − is dominant; in the other, m h ∼ −
50 GeV so that h → b ¯ b is dominant. m h lies between slightly below 90 and slightly above 130 GeV. Scenarios with even lighter h are excluded by decay–independent searches for Zh production [4]. If m h is much above 130GeV, the CP–odd component of h becomes subdominant, so that the cross section for Zh production becomes too large. Finally, the LEP hole occurs for tan β in between 3 and 10 [4].We analyze the prospect for observing a signal for the production of neutral Higgs bosons in thesecond of these LEP allowed regions.
2. Numerical analysis
In our analysis we took five different benchmark points, denoted by S1 through S5 [ m h =130-90GeV with m h =30 GeV] and two CPX . –scenario [CPX–1(2) where m h =102 (103) GeV with m h =36 (45) GeV, which can be realized for M H + = 131.8 GeV with tan β = 4.02 (4.39)] ofthe MSSM [5]. We calculated the spectrum and the couplings for these two benchmark pointsusing CPsuperH [3]. In our simulation we used the
PYTHIA v6.408 [6] event generator withthe
SLHA [7] interfacing option. We used
MadGraph/MadEvent v4.2.8 [8] for generating partonlevel SM backgrounds which were fed to
PYTHIA for showering. We set the renormalization andfactorization scale to Q = √ ˆ s and used CTEQ5L for the parton distribution functions (PDF).The signal arises from p ¯ p → W h → ℓν ℓ h h → ℓν ℓ b ¯ bb ¯ b , leading to ℓjjjjE/ T events, where ℓ = e or µ . The effective cross section for this signal topology can be expressed as, σ totsignal = σ SM ( p ¯ p/pp → W h ) × g h W W × Br ( h → h h ) × Br ( h → b ¯ b ) × Br ( W → eν e ) , (1)where g h W W is the h W W coupling in units of the corresponding SM value, W stands for W ± and the factor 2 is for ℓ = e and µ . This process has recently been studied in refs.[9, 10], usingparton–level analyses with quite promising results. We instead performed a full hadron–levelanalysis, including initial and final state showering as well as the underlying event. We take g h W W × Br ( h → h h ) × Br ( h → b ¯ b ) = 0 . √ s = 1 .
96 TeV. We haveused the toy calorimeter simulation (
PYCELL ) provided in
PYTHIA with the following criteria:calorimeter coverage is | η | < .
64; the segmentation is given by ∆ η × ∆ φ =0 . × .
098 whichresembles the CDF detector [11]; Gaussian smearing of the total energy of jets and leptons; acone algorithm with ∆R(j , j) = p ∆ η + ∆ φ = 0 . cellT , min ≥ . jetT , min ≥ . E T ; leptons ( ℓ = e , µ ) are selected withE ℓ T ≥ . | η ℓ | ≤ . E/ T ) from all observed particles.The tagging of b − jets plays a crucial role in our analysis. A jet with | η j | ≤ . E jT ≥ b − flavored hadron ( B − hadron), i.e. with ∆ R ( j, B − hadron) < .
2, isconsidered to be “taggable”. We assume that such jets are actually tagged with with probability ǫ b = 0 .
50 [12]. We find that our tagging algorithm agrees well with the t ¯ t analysis of CDF [13].We also modeled mistagging of non − b jets as b − jets, treating c − jets differently from those dueto gluons or light quarks with ǫ c = 0 .
10 and ǫ u,d,s,g = 0 . . − .
16 pb while the totalbackground is approximately 455 pb. We have displayed the raw number of events for signal (GeV) E v en t s / G e V S1S3S5ToB
Figure 1.
The four–jet invariant mass m j distribution after all cuts for signal scenariosS1, S3 and S5 and for the total background(ToB), requiring triple b − tag following thelast column of Table 1 at Tevatron. m (GeV) E v en t s / G e V S1S3S5ToB
Figure 2.
The four–jet invariant mass m j distribution after all cuts for signal scenariosS1, S3 and S5 and for the total background(ToB), requiring triple b − tag following thelast column of Table 1 at LHC.and backgrounds in the first column in Table 1. The second column is the number of eventsafter applying the following basic acceptance cuts: N jet ≥ E j =1 − T >
10 GeV, | η j =1 − | < . N lepton ≥ E ℓT >
15 GeV, | η ℓ | < . E/ T ≥
15 GeV. Not surprisingly, t ¯ t is the main source ofbackground at this stage. We found that the suppressions by applying N b − tag ≥ t ¯ t . The signal contains more b quarks, but the t ¯ t background has much harder b jets, leading to larger tagging probabilities. However, this background can contain a third b jet only due to showering. Hence requiring N b − tag ≥ b − tagging probability,which lies between 1.3 and 3.6% depending on m h , also reduces the signal rate.We have also looked for invariant mass peaks to isolate the signal on top of a sizablebackground. In addition to the basic acceptance cuts (defined above) we demand that the signalcontains exactly (rather than at least) four jets. This reduces combinatorial backgrounds forHiggs mass reconstructions. Finally, we pick the jet pairing ( ij )( kl ) (with i, j, k, l ∈ { , , , } )that minimizes the difference m j i j j − m j k j l of di–jet invariant masses; in the absence of showeringand for perfect energy resolution, the signal would have m j i j j = m j k j l = m h . We then demandthat both m j i j j and m j k j l lie between 10 and 60 GeV, where the lower bound comes from therequirement that h → b ¯ b decays should be allowed, and the upper bound from the requirementthat h → h h decays should be open. We next require the four–jet invariant mass to liebetween 60 and 140 GeV; this covers the entire “LEP hole” in the MSSM Higgs parameterspace. Table 1 shows that after these cuts, the signal actually exceeds the total background.The results of the column labeled Eff3 in Table 1 have been obtained by including all eventsthat pass the other cuts (not related to tagging) and have at least three taggable jets, assigningeach event a weight given by its (mis)tagging probability. This greatly increases the statistics.We checked that this gives results that are consistent with the event rejection technique wheneverthe latter has good statistics; this is the case if at most one b − tag results from mistagging.The distribution of m j is shown in Fig. 1 for signal scenarios S1, S3 and S5 as well as for the able 1. Process column shows the signal benchmarks and the SM backgrounds, where j standsfor u, d, s, g and – for background process 7 and 8 – c . RawEvt is stands for the number of eventsproduced in the experiments (for backgrounds we applied basic pre–selections in the generatorlevel: p j,bT ≥ | η j,b | < . R ( jj, bb, bj ) ≥ . N acc is the number after the basicselection cuts, whereas N b is with at least three jets tagged as b − jets, allowing for mistagging.Eff3 is the number of events passing the selection cuts that contain exactly four jets and at least3 b − tagged jets; the numbers in parentheses represent the number of events with the inclusionof m pair and m j cuts. Finally, ToB is the total number of background events.Tevatron with 4fb − LHC with 10fb − Process RawEvt N acc N b Eff3 (+h1) RawEvt N acc N b Eff3(+h1)S1 38.11 11.09 0.77 0.49 (0.40) 1157 352.5 33.48 13.96(6.85)S2 51.59 13.85 0.83 0.51 (0.44) 1486 418.3 36.84 15.36(7.76)S3 68.91 16.58 0.83 0.54 (0.47) 1962 506.5 39.81 17.03(8.91)S4 94.76 19.88 0.87 0.54 (0.46) 2620 610.6 43.18 17.81(9.61)S5 133.6 23.92 0.86 0.52 (0.45) 3516 724.7 43.92 18.96(9.63)CPX-1 89.89 20.27 0.82 0.49 (0.43) 2509 600.2 40.07 17.25(9.07)CPX-2 87.56 22.46 0.84 0.53 (0.47) 2421 597.2 40.28 16.88(9.64) t ¯ t b ¯ bb ¯ bW ± b ¯ bbjW ± b ¯ bcjW − b ¯ bc ¯ cW ± b ¯ bjjW ± bjjjW ± jjjjW ± t ¯ tb ¯ b t ¯ tc ¯ c m j = m h at the partonlevel.By seeing the distributions of m pair (average of the two optimal pairing of di–jet invariantmasses) and m j (see Fig. 1) allow us define the final significance of the signal by countingevents that satisfy: 0 . m h ≤ m pair ≤ m h + 5 GeV and 0 . m h ≤ m j ≤ m h + 10 GeV. Wesee that requiring triple b − tags leads to very good signal to background ratio, of around 10 for m h = 30 GeV and slightly less for heavier h . However, we expect less than 2 signal eventsafter all cuts for a total integrated luminosity ( R L dt ) of 20 fb − . We tried also to see the effectfor two b − tag and found that the signal rate increase by a factor between 3.7 and 5 while thebackground increases by two orders of magnitude and S/ √ B is well below two.The significance defined in this way overestimates the true statistical significance of a doublepeak in the m pair and m j distributions somewhat, due to the “look elsewhere” effect: since m h and m h are not known a priori, one would need to try different combinations when looking forpeaks. However, given that we use rather broad search windows, there are probably only O (10)statistically independent combinations within the limits of the LEP hole. Note also that theignal rate is still quite small. Further kinematical cuts, which might slightly increase the signalto background ratio, are therefore not likely to increase the statistical significance of the signal.We are therefore forced to conclude that the search for W h → W h h → ℓνb ¯ bb ¯ b events at theTevatron does not seem promising, and turn instead to the LHC.Our analysis for the LHC follows broadly similar lines as that for the Tevatron. We simulateour signal and backgrounds at the LHC with √ s = 14 TeV. The PYCELL model is based on theATLAS detector [14] with calorimeter coverage | η | < .
0, segmentation ∆ η × ∆ φ = 0 . × . , j) = 0 .
4. Calorimeter cells with E cellT , min ≥ . cellT , min ≥ . jetT , min ≥ . E T ;leptons ( ℓ = e , µ ) are selected if they satisfy E ℓ T ≥
20 GeV and | η ℓ | ≤ .
5. The jet–leptonisolation criterion and the missing transverse energy E/ T are adopted similar to our Tevatronanalysis.Only jets with | η j | < . b − jets. If the jet is “matched”to a b − flavored hadron, with ∆ R ( j, hadron) ≤ .
2, the tagging efficiency is taken to be 50%.If instead the jet is matched to a c − hadron, the (mis)tagging efficiency is taken to be 10%,whereas jets matched to a τ − lepton have zero tagging probability. All other taggable jets have(mis)tagging probability of 0.25%. These efficiencies follow recent ATLAS analyses [15, 16, 17].The cross-sections for the signal benchmarks lie between 0 . − . . · pb. We have displayed the raw number of eventsfor signal and backgrounds (again with the same generator–level cuts as of Tevatron) in thefirst column in Table 1. The second column is the number of events after applying the basicacceptance cuts: N jet ≥ E j =1 − T >
15 GeV, | η j =1 − | < . N lepton ≥ E ℓT >
20 GeV | η ℓ | < . E/ T ≥
20 GeV. They reduce the cross section by about a factor of 5 (3) for m h = 90 (130)GeV.The number of events (for R L dt =10 fb − ) passing the acceptance cuts and containing exactlyfour jets, at least three of which are tagged (adopting the same strategy like Tevatron), is givenby the Eff3 column of Table 1. Similar to Tevatron we require both jet pair invariant massesto lie between 10 and 60 GeV and m j to lie between 60 and 140 GeV. The m j distribution isshown in Fig. 2. After these cuts we are left with slightly less than one signal event and slightlyless than two background events per fb − of data. A 5 σ signal would then require almost 100fb − of data, more than the LHC is likely to collect during “low” luminosity running.We also checked that requiring a fourth b − tag reduces the signal cross section by anotherorder of magnitude or more. The signal rate then becomes so low that one would have to waitfor the high–luminosity phase of the LHC to accumulate enough events to reconstruct invariantmass peaks. We therefore stick to triple b − tag in our LHC analysis.We found that the background shows a peak in the m pair distribution between 30 and 40GeV, not far from the peak of the signal in the scenarios we consider. A tighter cut on m pair will nevertheless improve the signal–to–background ratio. Moreover, the four–jet invariant massdistribution (in Fig. 2) of the background peaks at large values, largely due to the contributionfrom t ¯ t production. At least for scenarios with h masses in the lower half of the “LEP hole”region a tighter cut on m j will therefore also improve the significance of the signal. We thereforeapplied the same “double–peak” cuts as at the Tevatron. The signal then always exceeds thebackground. Assuming an integrated luminosity of 60 fb − at the end of “low” luminosityrunning, we find a final statistical significance of at least 5 standard deviations, and a signalsample of some 30 events. . Conclusions We analyzed the possibility of observing neutral Higgs bosons at currently operating hadroncolliders in the framework of the CP violating MSSM. We explored the ℓjjjjE/ T channel withdouble, triple and quadruple b tag, focusing on the region of parameter space not excluded byLEP searches. We considered a large number of SM backgrounds and employed a full hadron–level Monte Carlo simulation using the PYTHIA event generator. We carefully implemented b − tagging, including mistagging of c − jets or light flavor or gluon jets. At the Tevatron, requiring3 b − tag, we can only expect about one signal event per 10 fb − of R L dt , on a background ofabout 0.3 events. If we require only 2 b − tag, the signal increases by a factor of about 4, butthe background increases by two orders of magnitude, making the signal unobservable. At theLHC we focussed on events with exactly four jets, cutting simultaneously on the average di–jetinvariant mass and the four–jet invariant mass and demanding at least 3 b -tags. We found asignal rate above the background, and a signal significance exceeding 5 σ for an R L dt of 60 fb − .One might be able to increase the S:B ratio even more by requiring 4 b − tags with softer taggingcriteria (enhancing mistag rate also), possibly simultaneously relaxing the requirement on thenumber of jets. This could be used to confirm the existence of a signal.We conclude that searches for W h production at the LHC with W → ℓν and h → h h → b ¯ bb ¯ b should be able to close that part of the “LEP hole” in parameter space where h → b ¯ b decays dominate. The details of this analysis can be found in our recent paper [18]. This work was partially supported by the Bundesministerium f¨ur Bildung und Forschung(BMBF) under Contract No. 05HT6PDA, by the EC contract UNILHC PITN-GA-2009-237920, and by the Spanish grants FPA2008-00319, CSD2009-00064 (MICINN) andPROMETEO/2009/091 (Generalitat Valenciana).
References [1] For introductions into supersymmetric extensions of the Standard Model in general, and the MSSM inparticular, see e.g. M. Drees, R.M. Godbole and P. Roy,
Theory and Phenomenology of Sparticles , WorldScientific, Singapore (2004); H.A. Baer and X.R. Tata,
Supersymmetry: from superfields to scatteringevents , Cambridge Press (2006).[2] A. Pilaftsis, Phys. Rev.
D58 (1998) 096010; Phys. Lett.
B435 (1998) 88; E. Accomando et al. ,arXiv:hep-ph/0608079.[3] J.S. Lee et al. , Comput. Phys. Commun. (2004) 283; Comput. Phys. Commun. , 312 (2009).[4] S. Schael et al. (ADLO Collab.) Eur. Phys. J. C (2006) 547; OPAL Collab., G. Abbiendi et al. , Eur.Phys. J. C27 (2003) 311; DELPHI Collab., J. Abdallah et al., Eur. Phys. J.
C38 (2004) 1; P. Bechtle etal. , Comput. Phys. Commun. (2010) 138.[5] M. Carena, J.R. Ellis, A. Pilaftsis and C.E.M. Wagner, Phys. Lett. B (2000) 155, hep-ph/0009212.[6] T. Sjostrand, S. Mrenna and P. Skands, JHEP , 026 (2006).[7] P. Skands et al. , JHEP , 036 (2004).[8] F. Maltoni and T. Stelzer, JHEP , 027 (2003).[9] K. Cheung, J. Song and Q.-S. Yan, Phys. Rev. Lett. (2007) 031801, hep-ph/0703149.[10] M. Carena, T. Han, G.Y. Huang and C.E.M. Wagner, JHEP , 092 (2008), arXiv:0712.2466 [hep-ph].[11] CDF
Collab., T. Aaltonen et al. , Phys. Rev. Lett. (2008) 091803;
CDF IIb
Collab., P. T. Lukens,
The CDF IIb detector: Technical design report , FERMILAB-TM-2198.[12] K. Hanagaki [D0 Collab.], FERMILAB-CONF-05-647-E; C. Neu [CDF Collab.], FERMILAB-CONF-06-162-E; T. Wright [CDF and D0 Collabs.], arXiv:0707.1712 [hep-ex]; M.T. Bowen et al. , Phys. Rev. D ,074016 (2005).[13] D. Acosta et al. , Phys. Rev. D (2005) 052003.[14] The Atlas Collab., CERN-LHCC-99-15, ATLAS-TDR-15 (1999).[15] G. Aad et al. [The ATLAS Collab.], arXiv:0901.0512 [hep-ex].[16] M. Lehmacher, arXiv:0809.4896 [hep-ex].[17] E. Alagoez et al.et al.