Physics Letters B 706 (2011) 3239
Contents lists available at SciVerse ScienceDirect
Physics Letters B
First observation of the decay B0s D0K 0 and a measurement of the ratioof branching fractions B(B
a r t i c l e i n f o a b s t r a c t
Article history:Received 18 October 2011Received in revised form 29 October 2011Accepted 31 October 2011Available online 4 November 2011Editor: L. Rolandi
The rst observation of the decay B0s D0K 0 using pp data collected by the LHCb detector at a centre-of-mass energy of 7 TeV, corresponding to an integrated luminosity of 36 pb1, is reported. A signalof 34.4 6.8 events is obtained and the absence of signal is rejected with a statistical signicance ofmore than nine standard deviations. The B0s D0K 0 branching fraction is measured relative to that ofB0 D00: B(B0s D0K 0)B(B0D00) = 1.480.340.150.12, where the rst uncertainty is statistical, the secondsystematic and the third is due to the uncertainty on the ratio of the B0 and B0s hadronisation fractions.
2011 CERN. Published by Elsevier B.V. All rights reserved.
Collider (LHC), start with the Vertex Locator, a silicon strip device
03doA theoretically clean extraction of the CabibboKobayashiaskawa (CKM) unitarity triangle angle can be performed usingme-integrated B DX decays by exploiting the interferencetween Cabibbo-suppressed b u and Cabibbo-allowed b cansitions . One of the most promising channels for this pur-se is B0 DK 0, where D represents a D0 or a D0 meson.1though this channel involves the decay of a neutral B meson,e nal state is self-tagged by the avour of the K 0 so that ame-dependent analysis is not required. In the B0 DK 0 decay,th the B0 D0K 0 and the B0 D0K 0 are colour suppressed.erefore, although the B0 DK 0 decay has a lower branchingaction compared to the B+ DK+ mode, it could exhibits anhanced interference.The Cabibbo-allowed B0s D0K 0 and B0s D0K 0 de-
ys potentially provide a signicant background to the Cabibbo-ppressed B0 D0K 0 decay. The expected size of this back-ound is unknown, since the B0s D()0K 0 decay has not yeten observed. In addition, a measurement of the branching frac-on of B0s D0K 0 is of interest as a probe of SU(3) breaking inlour suppressed B0(d,s) D0V decays [7,8], where V denotes autral vector meson. Thus, the detailed study of B0s D0K 0 isimportant goal with the rst LHCb data.The LHCb detector  is a forward spectrometer constructed to
easure decays of hadrons containing b and c quarks. The detec-r elements, placed along the collision axis of the Large Hadron
CERN for the benet of the LHCb Collaboration.In this Letter the mention of a decay will refer also to its charge-conjugate state.
that surrounds the pp interaction region with its innermost sensi-tive part positioned 8 mm from the beam. It precisely determinesthe locations of the primary pp interaction vertices, the locationsof the decay vertices of long-lived hadrons, and contributes tothe measurement of track momenta. Other tracking detectors in-clude a large-area silicon strip detector located upstream of the4 Tm dipole magnet and a combination of silicon strip detectorsand straw drift chambers placed downstream. Two Ring-ImagingCherenkov (RICH) detectors are used to identify charged hadrons.Further downstream an electromagnetic calorimeter is used forphoton detection and electron identication, followed by a hadroncalorimeter and a muon system consisting of alternating layers ofiron and gaseous chambers. LHCb operates a two stage trigger sys-tem. In the rst stage hardware trigger the rate is reduced fromthe visible interaction rate to about 1 MHz using information fromthe calorimeters and muon system. In the second stage softwaretrigger the rate is further reduced to 2 kHz by performing a set ofchannel specic selections based upon a full event reconstruction.During the 2010 data taking period, several trigger congurationswere used for both stages in order to cope with the varying beamconditions.
The results reported here uses 36 pb1 of pp data collected atthe LHC at a centre-of-mass energy
s = 7 TeV in 2010. The strat-
egy of the analysis is to measure a ratio of branching fractions inwhich most of the potentially large systematic uncertainties can-cel. The decay B0 D00 is used as the normalisation channel.In both decay channels, the D0 is reconstructed in the Cabibbo-allowed decay mode D0 K+; the contribution from the dou-bly Cabibbo-suppressed D0 K+ decay is negligible. The K 0is reconstructed in the K 0 K+ decay mode and the 0 in70-2693/ 2011 CERN. Published by Elsevier B.V. All rights reserved.i:10.1016/j.physletb.2011.10.073
LHCb Collaboration / Physics Letters B 706 (2011) 3239 33
the 0 + decay mode. The main systematic uncertaintiesarise from the different particle identication requirements andthe pollution of the B0 D00 peak by B0 D0+ decayswhere the + pairs do not originate from a 0 resonance. Inaddition, the normalisation of the B0s decay to a B
0 decay suffersfrom a systematic uncertainty of 8% due to the current knowledgeof the ratio of the fragmentation fractions f s/ fd = 0.267+0.0210.020 .
2. Events selection
Monte Carlo samples of signal and background events are usedto optimize the signal selection and to parametrize the probabil-ity density functions (PDFs) used in the t. Proton beam collisionsare generated with PYTHIA  and decays of hadronic particlesare provided by EvtGen . The generated particles are tracedthrough the detector with GEANT4 , taking into account thedetails of the geometry and material composition of the detec-tor.
B0 and B0s mesons are reconstructed from a selected D0 me-
son combined with a vector particle (0 or K 0). The selectionrequirements are kept as similar as possible for B0s D0K 0 andB0 D00. The four charged particles in the decay are each re-quired to have a transverse momentum pT > 300 MeV/c for thedaughters of the vector particle and pT > 250 MeV/c (400 MeV/c)for the pion (kaon) from the D0 meson decay. The 2 of the trackimpact parameter with respect to any primary vertex is required tobe greater than 4. A cut on the absolute value of the cosine of thehelicity angle of the vector meson greater than 0.4 is applied. Thetracks of the D0 meson daughters are combined to form a vertexwith a goodness of t 2/ndf smaller than 5. The B meson ver-tex formed by the D0 and the tracks of the V meson daughters isrequired to satisfy 2/ndf < 4. The smallest impact parameter ofthe B meson with respect to all the primary vertices is requiredto be smaller than 9 and denes uniquely the primary vertex as-sociated to the B meson. Since the B0 or B0s should point towardsthe primary vertex, the angle between the B momentum and theB line of ight dened by the line between the B vertex and theprimary vertex is required to be less than 10 mrad. Finally, sincethe measured z position (along the beam direction) of the D ver-tex (zD ) is not expected to be situated signicantly upstream ofthe z position of the vector particle vertex (zV ), a requirement of
(zD zV )/ 2z,D + 2z,V > 2 is applied, where z,D and z,V are
the uncertainties on the z positions of the D and V vertices re-spectively.
The selection criteria for the V candidates introduce some dif-ferences between the signal and normalisation channel due to theparticle identication (PID) and mass window requirements. TheK 0 (0) reconstructed mass is required to be within 50 MeV/c2(150 MeV/c2) of its nominal value . The selection criteria forthe D0 and vector mesons include identifying kaon and pion can-didates using the RICH system. This analysis uses the comparisonbetween the kaon and pion hypotheses, DLLK , which representsthe difference in logarithms of likelihoods for the K with respectto the hypothesis. The particle identication requirements forboth kaon and pion hypotheses have been optimised on data. Thethresholds are set at DLLK > 0 and DLLK < 4, respectively, forthe kaon and the pion from the D0. The misidentication rate iskept low by setting the thresholds for the vector meson daugh-ters to DLLK > 3 and DLLK < 3 for the kaon and pion respec-tively. In order to remove the potential backgrounds due to B0s D+s and B0 D+ with Ds K 0K and D K 0K ,vetoes around the nominal D and Ds meson masses  of15 MeV/c2 are applied. Monte Carlo studies suggest that thesevetoes are more than 99.5% ecient on the signal.
Finally, multiple candidates in an event (about 5%) are removedby choosing the B candidate with the largest B ight distance sig-nicance and which lies in the mass windows of the D0 and thevector meson resonance.
3. Extraction of the ratio of branching fractions
The ratio of branching fractions is calculated from the numberof signal events in the two decay channels B0s D0K 0 and B0 D00,
B(B0s D0K 0)B(B0 D00)
B0s D0K 0Nsig.
0 +)B(K 0 K+)
where the parameters represent the total eciencies, includ-ing acceptance, trigger, reconstruction and selection, and f s/ fd isthe ratio of B0 and B0s hadronisation fractions in pp collisions ats = 7 TeV. Since a given event can either be triggered by tracks
from the signal or by tracks from the other B hadron decay, abso-lute eciencies cannot be obtained with a great precision from theMonte Carlo simulation due to improper modelling of the genericB hadron decays. In order to reduce the systematic uncertaintyrelated to the Monte Carlo simulation of the trigger, the data sam-ple is divided into two categories: candidates that satisfy only thehadronic hardware trigger2 (TOSOnly, since they are Triggered Onthe Signal (TOS) exclusively and not on the rest of the event) andevents which are Triggered by the rest of the event Independentof the Signal candidate B decay (TIS). Approximately 6% of candi-dates do not enter either of these two categories, and are vetoedin the analysis. The B0 D00 signal yield is extracted separatelyfor the two trigger categories TOSOnly and TIS; the B0s D0K 0signal yield is extracted from the sum of both data samples. Theratio of eciencies are sub-divided into the contributions arisingfrom the selection requirements (including acceptance effects, butexcluding PID), rsel, the PID requirements, rPID, and the trigger re-quirements, rTOSOnly and rTIS . The ratio of the branching fractionscan therefore be expressed as
B(B0s D0K 0)B(B0 D00) =
B(0 +)B(K 0 K+)
where represents a correction factor for the non-0 contribu-tion in the B0 D00 decays.
The values of the eciency ratios are measured using sim-ulated events, except for rPID = 1.09 0.08 which is obtainedfrom data using the D D0 decay with D0 K+ whereclean samples of kaons and pions can be obtained using a purelykinematic selection. Since the event selection is identical for theD0 in the two channels of interest, many factors cancel out inrsel = 0.784 0.024 thereby reducing the systematic uncertain-ties. The values of the trigger eciency ratios, rTOSOnly = 1.20 0.08 and rTIS = 1.03 0.03, depend on the trigger congurationsand are therefore computed from a luminosity-weighted average.The quoted uncertainties reect the difference between data and
2 Events passing only the muon trigger on the signal candidate tracks are re-jected.
34 LHCb Collaboration / Physics Letters B 706 (2011) 3239
Fig. 1. The invariant mass distribution for the B0 D00 decay mode for the TOSOnly (left) and TIS (right) trigger categories with the result of the t superimposed. Theblack points correspond to the data and the t result is represented as a solid line. The signal is tted with a double Gaussian (dashed line), the partially reconstructed back-ground with an exponential function (light grey area) and the combinatorial background with a at distribution (dark grey area) as explained in the text. The contributionsfrom cross-feed are too small to be visible.
Monte Carlo simulation mainly caused by the energy calibration ofthe trigger.
The numbers of events in the two D00 trigger categories,NTOSOnly
B0D00 and NTISB0D00 , and N
sig.B0s D0K 0 are extracted from a
simultaneous unbinned maximum likelihood t to the data. In or-der to simplify the description of the partially reconstructed back-ground, the lower edge of the B meson mass window is restrictedto 5.1 GeV/c2 for the B0 D00 decay mode and to 5.19 GeV/c2for the B0s D0K 0 decay mode. There are four types of eventsin each category: signal, combinatorial background, partially re-constructed background and cross-feed.3 The signal B meson massPDFs for B0 D00 and B0s D0K 0 are parametrised for eachchannel using the sum of two Gaussians sharing the same meanvalue. The mean and width of the core Gaussian describing theB0 D00 mass distribution are allowed to vary in the t. Thefraction of events in the core Gaussian, 0.81 0.02, and the ratioof the tail and core Gaussian widths, 2.04 0.05, are xed to thevalues obtained from Monte Carlo simulation. In order to take intoaccount the difference in mass resolution for the B0 D00 andB0s D0K 0 decay modes, the value of the ratio of core Gaussianwidths
= 0.89 0.03 is xed from the Monte Carlo simu-lation. The mass difference between the means of the B0 and B0ssignals is xed to the nominal value .
Fig. 2. The invariant mass distribution for the B0s D0K 0 decay mode with theresult of the t superimposed. The black points correspond to the data and the tresult is represented as a solid line. The signal is tted with a double Gaussian(dashed line), the partially reconstructed background with an exponential function(light grey area), the combinatorial background with a at distribution (dark greyarea) and the cross-feed from B0 D00 (intermediate grey area) as explained inthe text.
In summary, 13 parameters are free in the t. Four shapeparameters are used, two for the signal and two for the par-tially reconstructed backgrounds. In addition, nine event yieldsThe combinatorial background mass distribution is modelleda at PDF and the partially reconstructed background is
rametrised by an exponential function; the exponential slopedifferent in the B0 D00 and B0s D0K 0 categories.
nce the number of B0 D00 decays is larger than that of D0K 0, the contribution from misidentied pions as kaons
om real B0 D00 has to be taken into account. The frac-ons of the cross-feed events, f D00D0K 0 = 0.062 0.031 and0K 0D00 = 0.095 0.047, are constrained using the resultsom a Monte Carlo study corrected by the PID misidenticationtes measured in data. The PDF for the cross-feed is empiri-lly parametrised by a Crystal Ball function , whose widthd other parameters are taken from a t to simulated events inhich B0s D0K 0 events are misidentied as B0 D00 andce versa; the width is xed to 1.75 times the signal resolution.r t...