Search for neutral MSSM Higgs bosons decaying into a pair of bottom quarks

A search for neutral Higgs bosons decaying into a b-bbar quark pair and produced in association with at least one additional b quark is presented. This signature is sensitive to the Higgs sector of the minimal supersymmetric standard model (MSSM) with large values of the parameter tan(beta). The analysis is based on data from proton-proton collisions at a center-of-mass energy of 8 TeV collected with the CMS detector at the LHC, corresponding to an integrated luminosity of 19.7 inverse femtobarns. The results are combined with a previous analysis based on 7 TeV data. No signal is observed. Stringent upper limits on the cross section times branching fraction are derived for Higgs bosons with masses up to 900 GeV, and the results are interpreted within different MSSM benchmark scenarios, m[h,max], m[h,mod+], m[h,mod-], light-stau and light-stop. Observed 95% confidence level upper limits on tan(beta), ranging from 14 to 50, are obtained in the m[h,mod+] benchmark scenario.


Introduction
The discovery of a Higgs boson with a mass around 125 GeV [1][2][3] marked a milestone for elementary particle physics. While the measured properties of the observed boson are in agreement with the expectations of the standard model (SM) with the current experimental precision, this particle could well be the first visible member of an extended Higgs sector, which would be a direct indication of new physics. Extended Higgs sectors are possible in various theoretical models, such as Supersymmetry [4][5][6][7], which relates fermionic and bosonic degrees of freedom and in consequence requires the introduction of additional Higgs bosons as well as a superpartner to each SM particle. The superpartners provide potential dark-matter candidates [8], and their contribution to quantum-loop corrections can lead to a unification of the gauge couplings at higher energies [9]. Moreover, the problem of the quadratic divergence of the Higgs boson mass at high energies [10] is solved naturally through cancellation of loop terms by the superpartners.
The minimal supersymmetric extension of the SM (MSSM) [5] contains two scalar Higgs doublets, which result in two charged Higgs bosons, H ± , and three neutral ones, jointly denoted as φ. Among the latter are two CP-even (h, H) and one CP-odd state (A). The recently discovered boson with a mass near 125 GeV might then be interpreted as one of the neutral CP-even states. Two parameters, generally chosen as the mass of the pseudoscalar Higgs boson m A and the ratio of the vacuum expectation values of the two Higgs doublets, tan β = v 2 /v 1 , define the properties of the Higgs sector in the MSSM at tree level. For tan β values larger than one, the couplings of the Higgs field to down-type fermions are enhanced relative to those to the up-type fermions. Furthermore, the A boson is nearly degenerate in mass with either the h or H boson. These effects enhance the combined cross section for producing these Higgs bosons in association with b quarks by a factor of ≈2 tan 2 β. The decay φ → bb is expected to have a high branching fraction (≈90%), even at large values of the Higgs boson mass [11].
Measurements at the CERN LHC in the φ → ττ decay mode [12][13][14][15] have lead to the most stringent constraints on tan β so far, with exclusion limits in the range 4-60 in the mass interval of 90-1000 GeV. Preceding limits had been obtained by the LEP [16] and Tevatron experiments [17][18][19]. Also the φ → µµ decay mode has been investigated [13,20]. Besides extending the MSSM Higgs boson search to an independent channel, the φ → bb decay mode is particularly sensitive to the higgsino mass parameter µ [21], and thus to the bottom quark Yukawa coupling. In the φ → ττ channel, the sensitivity to µ is much smaller due to a partial cancellation of the respective radiative corrections between the contributions to the production and decay processes [21]. Beyond the MSSM interpretation, lepton-specific two-Higgs-doublet models (2HDM) [22] may allow for enhanced couplings of down-type quarks relative to leptons. The bb decay mode is also relevant in the more general context of exotic resonance searches, motivated for example by dark-matter models involving mediator particles with a large coupling to b quarks [23,24].
Searches in the φ → bb decay mode have initially been performed at LEP [16] and by the CDF and D0 experiments [25] at the Tevatron collider. The first and so far the only analysis at the LHC in this channel has been performed by the CMS experiment, using the 7 TeV data, and set significantly more stringent bounds in the mass range 90-350 GeV [26].
In this article, the CMS search is extended by adding the data set comprising 19.7 fb −1 of proton-proton collision data, collected at a center-of-mass energy of 8 TeV, and by the use of a refined methodology. The higher integrated luminosity as well as the greater center-of-mass energy allow extension of the search up to a mass of 900 GeV.
The search is performed for neutral MSSM Higgs bosons φ with masses m φ ≥ 100 GeV that are produced in association with at least one b quark and decay to bb; an illustration of the signal process is given by the diagrams in Fig. 1. The signal is thus searched for in final states characterized by at least three b-tagged jets. No requirement of a fourth b-tagged jet is made, since its kinematic distributions extend significantly beyond the available acceptance, and the resulting signal efficiency would be very low. Events are selected by specialized triggers that identify b jets already at the online level. This is important to suppress the large rate of multijet production at the LHC. The analysis searches for a peak in the invariant mass distribution of the two b jets with the highest p T values, which are assumed to originate from the Higgs boson decay. The dominant background is the production of heavy-flavor multijet events containing either three b jets, or two b jets plus a third jet originating from either a charm quark, a lightflavor quark or a gluon, which is misidentified as a b quark jet. For the final limits, the results of the 8 TeV analysis are combined with the previous 7 TeV analysis [26].

The CMS detector
The central feature of the CMS detector is a superconducting solenoid of 6 m internal diameter, providing a magnetic field of 3.8 T. Within the field volume, the inner tracker is formed by a silicon pixel and strip tracker. It measures charged particles within the pseudorapidity range |η| < 2.5. The tracker provides a transverse impact parameter resolution of approximately 15 µm and a resolution on p T of about 1.5% for 100 GeV particles. Also inside the field volume are a crystal electromagnetic calorimeter, and a brass and scintillator hadron calorimeter. Muons are measured in gas-ionization detectors embedded in the steel flux-return yoke, in the pseudorapidity range |η| < 2.4, with detector planes made using three technologies: drift tubes, cathode strip chambers, and resistive-plate chambers. Matching muons to tracks measured in the silicon tracker results in a p T resolution between 1% and 5%, for p T values up to 1 TeV. Forward calorimetry extends the coverage provided by the barrel and endcap detectors up to |η| < 5. A detailed description of the CMS detector, together with a definition of the coordinate system used and the relevant kinematic variables, can be found in Ref. [27].
teria to suppress the impact of instrumental noise and misreconstruction. Contributions from additional proton-proton interactions within the same bunch crossing (pileup) affect the jet momentum measurement. To mitigate this effect, charged particles associated with other vertices than the reference primary vertex are discarded in the jet reconstruction, and residual contributions (e.g. from neutral particles) are accounted for using a jet-area based correction [31]. Jets originating entirely from pileup interactions are identified and rejected based on vertex and jetshape information [32]. Jet energy corrections are derived from simulation, and are confirmed with in situ measurements of the energy balance in dijet and Z/γ+jet events [33].
For the offline identification of b jets, the combined secondary vertex (CSV) algorithm [34] is used. This algorithm combines information on track impact parameters and secondary vertices within a jet in a single likelihood discriminant that provides a good separation between b jets and jets of other flavors. Secondary-vertex reconstruction is performed with an inclusive vertex search amongst the tracks associated with a jet [35].
Simulated samples of signal and background events, also referred to as Monte Carlo (MC) samples, were produced using the PYTHIA [36] and MADGRAPH [37] event generators and include pileup events. The response of the CMS detector is modeled with GEANT4 [38]. The MSSM Higgs signal samples, pp → bbφ+X with φ → bb, were produced at leading order in the 4-flavor scheme with PYTHIA version 6.4.12. The p T and η distributions of the leading associated b jet are in good agreement with the next-to-leading order (NLO) calculations [39]. The multijet background from quantum chromodynamics (QCD) processes has been produced with PYTHIA, while for tt +jets events the MADGRAPH event generator was used in its version 5.1.5.11. For all generators, fragmentation, hadronization, and the underlying event have been modeled using PYTHIA with tune Z2 * . The most recent PYTHIA 6 Z2* tune is derived from the Z1 tune [40], which uses the CTEQ5L parton distribution functions (PDF) set, whereas Z2* adopts the CTEQ6L [41] PDF set.

Trigger and event selection
A major challenge to this analysis is posed by the huge hadronic interaction rate at the LHC, and it is addressed with a dedicated trigger scheme, designed especially to suppress the QCD multijet background. Only events with at least two jets in the pseudorapidity range of |η| ≤ 1.74 are selected. The leading jet (here and in the following the jets are ordered by decreasing p T ) is required to have p T > 80 GeV, while the subleading jet must have p T > 70 GeV. Furthermore, the event is only accepted if the absolute value of the difference in pseudorapidity between any two jets fulfilling the p T and η requirements is less than or equal to 1.74. The tight online requirements on the angular variables of the jets are introduced to reduce the trigger rates while preserving the signal significances in the probed mass range of the Higgs bosons. At the trigger level, b jets are identified using an algorithm that requires at least two tracks with high 3D impact parameter significance to be associated with the jet. At least two jets within the event must meet the online b tagging criteria to be accepted by the trigger. The efficiency of the jet-p T requirements in the trigger are derived from the data with zero-bias triggered events. The online b tagging efficiencies relative to the offline b tagging selection are obtained from simulations of QCD events generated with PYTHIA and scaled to account for the different b tagging efficiencies between data and simulation. The total trigger efficiency for events satisfying the offline selection requirements detailed below ranges from 46-62% over the Higgs boson mass range of 100-900 GeV.
The offline selection requires events to have the two leading jets within |η| ≤ 1.65 to be fully within the pseudorapidity windows of the trigger, and the third leading jet within |η| ≤ 2.2.
The three leading jets must also pass p T thresholds of 80, 70 and 20 GeV, respectively. In addition, the two leading jets must have a pseudorapidity difference of |∆η 12 | ≤ 1.4, because the QCD multijet background increases significantly with respect to the expected signal with increasing |∆η 12 |. A minimal pairwise separation of ∆R > 1 between the three leading jets, where ∆R = √ (∆η) 2 + (∆φ) 2 and ∆η and ∆φ are the pseudorapidity and azimuthal angle differences (in radians) between the two jets, is imposed to suppress background from b quark pairs arising from gluon splitting.
In the following, "triple-b-tag" and "double-b-tag" samples are introduced, which play crucial roles in the analysis. The triple-b-tag sample is the basis for the signal search. It is defined by requiring all three leading jets to satisfy a tight CSV b tagging selection requirement at a working point characterized by a misidentification probability for light-flavor jets (attributed to u, d, s, or g partons) of about 0.1% at an average jet p T of 80 GeV. The typical corresponding efficiency for b jets is about 50-60% in the central pseudorapidity region. The total number of events passing the trigger and offline selections is approximately 69 k.
The double-b-tag sample plays a key role in the estimation of the multijet background. In this selection, only two of the three leading jets must pass the tight CSV b tagging requirement. The total number of double-b-tag events remaining after the trigger and offline selections is about 2.4 M. While this definition does not explicitly exclude the triple-b-tag events, the potential signal contribution is negligible due to the size of the QCD multijet background in the double-b-tag sample, and a veto would lead to distortions in the background model described in Section 5.
An additional flavor-sensitive quantity, the secondary vertex mass sum of a jet, ΣM SV,j , is introduced to further improve the separation between jets of different flavor on top of the CSV b tagging requirement. It is defined as the sum of the invariant masses calculated from the tracks forming secondary vertices inside a jet, and thus provides additional separation power. The extension of the signal mass range compared to the previous 7 TeV analysis implies that the jets can have larger p T , with the consequence that b-tagged jets from background events have a higher probability to contain two heavy flavor quarks instead of at most one. This can occur for example if a very energetic gluon splits into a pair of b or c quarks with a narrow opening angle. For this reason, b and c quark pairs merged into the same jet, labeled as "b2" and "c2", respectively, are treated separately from the cases of unmerged b and c quarks, labeled "b1" and "c1", respectively.
The subsequent analysis will use the secondary vertex mass information to categorize events and to build background templates. Therefore, the secondary vertex mass sums of the three leading jets are combined into a condensed event b tagging estimator, X 123 . The construction of this estimator is shown in Table 1. Each selected jet j, where j is the rank of the jet in order of decreasing p T , is assigned an index B j , which can take one of the four possible integer values from 0-3 according to its secondary vertex mass sum value, as shown in Table 1 (left). For jets with no reconstructed secondary vertex, B j is also set to zero. The definition of these index regions is motivated by the population of the secondary vertex mass sum by the different jet flavors. From the three indices B 1 , B 2 , and B 3 , a combined event b tagging variable X 123 is constructed as shown in Table 1 (right). By definition, the event b tagging variable X 123 can assume nine possible values ranging from 0 to 8. The events are then categorized according to the value of X 123 , with the rationale of having sufficient statistics in each bin. The signal is searched for in the two-dimensional spectrum formed by the invariant mass of the two leading jets, M 12 , and the event b tagging variable X 123 .

Background model
The main background for this analysis originates from QCD multijet production, with at least two energetic jets actually containing b hadrons, and a third jet that passes the b tagging selection but possibly as a result of a mistag. Since this type of background cannot be accurately predicted by MC simulation, it is estimated from the data using control samples. The chosen method is similar to the one used in Ref. [42]. The background is modeled by a combination of templates, which are constructed from the double-b-tag sample. Only the shape of these background templates is relevant, since the normalization will be determined by the fit to the data.
Three categories of events are distinguished in the double-b-tag sample, which are denoted as xbb, bxb and bbx depending on whether the jet with the highest, second-highest or thirdhighest p T is exempt from the b-tag requirement. In this notation the three jets are referred to in order of decreasing p T .
From these three double-b-tag categories, background templates are constructed by weighting each untagged jet with the b tagging probability according to its assumed flavor. In the template nomenclature, the convention is to indicate the assumed flavor with a capital letter, and it can be one of the five options Q, C1, B1, C2, and B2, where Q refers to light quarks or gluons, while C1 and C2 refer to a jet with a single charm quark and a pair of charm quarks, respectively. Similarly, B1 and B2 refer to jets assumed to contain a single bottom quark and a pair of bottom quarks, respectively. The total number of templates is therefore 15. Each background template is a binned distribution in the two-dimensional space spanned by M 12 , the dijet mass of the two leading jets, and the event b tagging variable X 123 . For the construction of each template, each event is weighted with the b tagging probability corresponding to the assumed flavor of the untagged jet. This weight accounts for the effect of the b tagging discriminant threshold. The b tagging probability for each flavor is determined with simulated QCD multijet events, where the flavor selection is based on Monte Carlo truth information. Data/MC scale factors for the b tagging efficiencies are applied where appropriate [34]. Since the b tagging efficiency has a characteristic dependence on p T and η for each flavor, the weighting results in different shapes of the M 12 distributions. The X 123 dimension of the templates is modeled in the following way: In a given M 12 bin, an event can contribute to different X 123 bins depending on the flavor of its jets and its kinematics. For the two b-tagged jets, the secondary vertex mass sum information is taken as measured. For the untagged jet, each of the four possible values of the secondary vertex mass sum index is taken into account with a weight according to the probability that a jet with given flavor, p T and η will assume this value. These probabilities, parametrized as a function of the jet p T and η, were determined from simulated events, and validated in control data samples.
Two additional corrections are applied to the templates. The first correction addresses a contamination in the double-b-tag sample from non-bb events at the level of a few percent. This  contamination is estimated directly from the data using a negative b tagging discriminator [34] constructed with a track counting algorithm based on the negative impact parameter of the tracks, ordered from the most negative impact parameter significance upward. A second correction is required since the online b tagging patterns are different in the double-and the tripleb-tag samples. In double-b-tag events, the two online b tags usually coincide with the offline b tags, while in triple-b-tag events the online b tags can be assigned to any two-jet subset of the three leading jets. The correction is computed from simulated QCD multijet events, and is applied in the form of additional weights to the events in the double-b-tag sample.
Similarity in shape between some templates leads to unnecessary redundancy. For this reason, similar templates are combined using a χ 2 -based metric to guide the decisions. The relative weights in a combination are taken from MC. In the cases where one of the two leading jets is untagged, and the flavor assumption is the same, e.g. Qbb, and bQb, the templates are combined, resulting in a merged template (Q,b)b = Qbb + bQb. By analogy, also (C1,b)b, (B1,b)b, (C2,b)b, and (B2,b)b are obtained. The resulting set of ten templates still shows many similarities. For this reason, (B1,b)b, (B2,b)b, and (C2,b)b are combined into a single template; bbB1, bbB2, and bbC2 into a second; and bbC1 and bbQ into a third. The total number of templates to be fitted in combination to the data is thus reduced to five, , and bb(C1+Q). The projections of the M 12 and X 123 variables are shown in Fig. 2 for these five background templates.
Beyond QCD multijet production, top-quark pair (tt) events pose the largest potential background to the signal topology. The requirement of three b-tagged jets reduces this background substantially, since only two highly energetic b-tagged jets are expected from the decays of the top quarks. However, one of the W bosons can decay into a cs pair, and the c jet can be mistagged as b jet. Using the tt Monte Carlo sample, the tt contribution is found to be relatively small; the number of tt events passing the selections of the double-and triple-b-tag datasets it estimated to be about a factor of 70 smaller than the total amount of data in these samples. The invariant mass spectrum from tt is very similar to the one from the QCD multijet background, and does not show any narrow peaks. Since the tt events contribute to the double-b-tag sample, they are also taken into account in the background model. of the signal templates for three different Higgs boson masses are shown in Fig. 3. The shape of the mass distribution is dominated by the experimental resolution and the combinatorial background. The natural width expected for a MSSM Higgs boson in the considered mass and tan β region is negligible in comparison with the detector resolution. At a mass of 500 GeV and tan β = 50, for example, the natural width of the mass peak is found to be 13 GeV, which is only ≈14% of the RMS of the reconstructed mass distribution. The X 123 distributions show little variation with the MSSM Higgs boson mass; they reflect the triple-b-quark signature.

Signal efficiency
The signal efficiency for each MSSM Higgs mass point is obtained from the simulated data sets. The efficiency of the kinematic trigger selection has been derived with data from control triggers and is applied by weighting. Scale factors to account for the different b tagging efficiencies in data and MC [34] are also applied. The efficiency ranges between 0.17 and 6.38 per mille and peaks around 300 GeV. The detailed mass dependence is shown in Appendix A. The decrease of the efficiency for masses beyond 300 GeV is due to the degradation of the b tagging efficiency at high jet p T . For masses around 300 GeV the kinematic selections give rise to an efficiency of approximately 0.12, which is reduced to approximately 0.0065 when triple b tagging is required.

Fitting procedure
The overall two-dimensional distribution in the variables M 12 and X 123 is fitted by a model combining the background templates and optionally a signal template. A binned likelihood technique is used. The relative contribution of each template is determined by the fit. The systematic uncertainties are represented by nuisance parameters that are varied in the fit according to their probability density functions.

Systematic uncertainties
The following systematic uncertainties in the expected signal and background estimates affect the determination of the signal yield and/or its interpretation within the MSSM.
Uncertainties in the yields of the signal contributions include the uncertainty in the luminosity estimate [43], the statistical uncertainties in the signal MC samples, and the uncertainties of the relative online b tagging corrections. Also taken into account are the QCD renormalization and factorization scale (µ r , µ f ) uncertainties, the uncertainties due to the parton distribution functions (PDF) and the strong coupling constant α s , and the uncertainties in the underlying event and parton shower modeling, which all only affect the translation of the signal cross section into tan β in the MSSM interpretation. The impact of these uncertainties on the signal acceptance is not significant.
The rate as well as the shape of the signal contributions are also affected by the uncertainties in the trigger efficiencies, the jet energy scale, the jet energy resolution, and the pileup modeling, as well as the scale factors for the b-tag efficiency, the mistag rate, and the secondary vertex mass scale. The last three also affect the shapes of the background templates (recall that only the shape is relevant for the background templates). The statistical uncertainty in the template shape, due to the limited size of the double-b-tag sample and due to the uncertainty in the offline b-tag efficiencies and mistag rates, are propagated into the templates and accounted for in the fitting procedure. Additional systematic uncertainties in the shapes of the backgroundtemplates arise from the impurity of the double-b-tag sample and the online b tag correction to the templates. The sources and types of systematic uncertainties and their impact on the expected limit are summarized in Table 2.

Background-only fit
In the first step, an unconstrained fit is performed without inclusion of a signal template, involving a linear combination of the background templates only. Results are shown in Fig. 4 and benchmark scenario of the MSSM with m A = 350 GeV, tan β = 30, and µ = +200 GeV is overlayed, scaled by a factor 10 for better readability. In addition, the ratio of data to the background estimate is shown at the bottom. Table 3. The template-based background model describes the data well within the uncertainty of the template fits with a goodness-of-fit of χ 2 /N dof = 207.9/209, where N dof is the number of degrees of freedom, corresponding to a p-value of 0.51. As expected, the fit is dominated by templates involving triple b-jet signatures, whose fitted total contributions amount to ≈82%.

Combined fit of signal and background templates
In the second step, a signal template is included together with the background templates in the fit, with the relative fractions of signal and background templates allowed to vary freely. The fit is performed for all considered Higgs boson masses from 100 to 900 GeV. None of the fits shows any significant signal excess. Results for a Higgs boson mass of 350 GeV are shown in Fig. 5 and Table 3. At this mass point, the highest fluctuation in the fitted Higgs boson production cross section is observed, corresponding to a local significance of approximately 1.5 standard deviations. The goodness-of-fit is χ 2 /N dof = 205.2/208, corresponding to a p-value of 0.54.

Upper limits on cross sections times branching fractions
Cross sections are obtained from the fractions determined by the fit multiplied by the total number of data events after the selection in the signal region, and divided by the corresponding signal efficiencies (Section 6.2) and the integrated luminosity of 19.7 fb −1 .   In the absence of any significant signal, the results are translated into upper limits on the cross section times the branching fraction, σ(pp → bφ + X) B(φ → bb), of a generic Higgs-like state in the mass range 100-900 GeV. For calculations of exclusion limits, the modified frequentist construction CL s [44,45] is adopted using the ROOSTATS package [46]. The chosen test statistic, used to determine how signal-and background-like the data are, is based on the profile likelihood ratio. Systematic uncertainties are incorporated in the analysis via nuisance parameters and treated as pseudo-observables, following the frequentist paradigm. These uncertainties have been listed in Section 7.
The observed and the median expected 95% confidence level (CL) limits as a function of the Higgs boson mass are shown in Fig. 6 and listed in Table 5 in Appendix B. The 1σ and 2σ bands of the test statistic, including systematic uncertainties, are also shown.

Interpretation within the MSSM
The cross section limits shown in Fig. 6 are further translated into exclusion limits on the MSSM parameters tan β and m A . The cross sections obtained with the four-flavor NLO QCD calculation [47,48] and the five-flavor NNLO QCD calculation as implemented in BBH@NNLO [49] for b + h/H/A associated production have been combined using the Santander matching scheme [50]. The branching fractions were computed with the FEYNHIGGS [51][52][53][54] and HDE-CAY [55,56] programs as described in Ref. [11].
The observed and expected 95% CL median upper limits on tan β versus m A , together with the 1σ and 2σ bands, are shown in Fig. 7 (left). They have been computed within the traditional MSSM m max h benchmark scenario [57] with the higgsino mass parameter µ = +200 GeV. The observed upper limits range from tan β about 20 in the low-m A region to about 50 at m A = 500 GeV, and extend the existing measurement at 7 TeV [26] into the hitherto unexplored m A region beyond 350 GeV. The model interpretation is not extended to higher masses above 500 GeV because the theoretical predictions are not reliable for tan β much higher than 60.
While the cross section limits obtained from the 2011 and 2012 data cannot be combined directly due to the different center-of-mass energies, such a combination is possible for the modeldependent interpretation. The resulting upper limits on tan β versus m A from both data periods The aforementioned sensitivity of the φ → bb channel to the higgsino mass parameter µ is evident in Fig. 9, where the limit in the m mod+ h scenario is compared for different values of µ. The dependence is particularly pronounced at higher m A ; for example, the observed upper limit on tan β varies from 30 for µ = −500 GeV to beyond 60 for µ = +500 GeV for m A = 500 GeV. The limits are also listed in Table 12 in Appendix B.

Summary
A search for a Higgs boson decaying into a pair of b quarks and accompanied by at least one additional b quark has been performed in proton-proton collisions at a center-of-mass energy of 8 TeV at the LHC, corresponding to an integrated luminosity of 19.7 fb −1 . The data were taken with dedicated triggers using all-hadronic jet signatures combined with online b tagging. A selection of events with three b-tagged jets has been performed in the offline analysis. A signal has been searched for in the two-dimensional spectrum formed by the invariant mass of the two leading jets and a condensed event b-tag estimator.
No evidence for a signal is found. The observed distributions are well described by a background model constructed from events in which only two of the three leading jets are required to be b tagged. Upper limits on the Higgs boson cross section times branching fraction are obtained in the mass region from 100-900 GeV, thus extending the search to considerably higher masses than those accessed by the previous   The signal efficiencies are summarized in Table 4 and shown in Fig. 10 as a function of the Higgs boson mass.

Appendix B Exclusion limits
The model-independent 95% CL limits on σ(pp → bφ + X) B(φ → bb) are listed in Table 5 for different Higgs boson masses m φ . The 95% CL limits of (tan β, m A ) are listed in Tables 6 to 11 for different MSSM benchmark scenarios with µ = +200 GeV and for different values of µ in the m mod+ h scenario in Table 12. [GeV]