Analysis of KSe decays Outline Motivations of the measurement Analysis scheme Ks tag Signal selection Efficiency estimates Systematic sources Uncertainty in efficiency determinations Uncertainty in the knowledge of signal and background shapes Where we stand (What has been understood and done) What remains Tommaso Spadaro INFN LNF
Motivations I – study of CPT violation 1) Measure charge asimmetry to test CPT invariance: CPT p-e+nHWK0 = a + b p+e-nHWK0 = a*- b* p+e-nHWK0 = c + d p-e+nHWK0 = c*- d* DS=DQ eS = eK + dK eL = eK - dK DSDQ A G(p-e+n) - G (p+e-n) G(p-e+n) + G (p+e-n) AS = 2(Re eK + Re dK + Re b/a - Re d*/a) AL = 2(Re eK - Re dK + Re b/a + Re d*/a) CPT dec. CPT mixing CP DSDQ e CPT SM: AS = AL KTEV ’02: AL = (3.322 0.058 0.047)10-3 NA48 ’03 AL = (3.317 0.070 0.072)10-3 (preliminary: hep-ex 0305059) AS: no existing measurement
Motivations II – DS=DQ rule violation 2) Measure BR(KSpen) to test of the validity of the DS=DQ rule: x = A(DSDQ )/A(DS=DQ ) x+ = (x + x)/2 c*/a x = A(DSDQ )/A(DS=DQ ) 110-3 BRS(pen)/tS 1 + 4Re x+= BRL(pen)/tL 710-3 810-3 SM: DSDQ transitions not allowed at tree level: |x| ~ GFmp2~10-7 CPLEAR ’98 Re x+ = (-1.8 4.1 4.5)10-3
Motivations III – extraction of Vus 3) Measure BR(KSpen) to extract Vus and test unitarity of 1st row of CKM matrix: |Vud|2 + |Vus|2 + |Vub|2 = 1 dVud / |Vud|0.05%, |Vub| negligible For Ke3 modes: K+e3 K0e3 K+m3 K0m3 E865 |Vus| f+ (0) K0p- exp th |Vus| = 0.2196 0.0019exp 0.0018th Comments: Old data (Chiang ’72) or PDG fit values: inclusiveness for Ke3g? New E865 measurement: agrees better with current values of Vud SM: From unitarity and |Vud|: |Vus|=0.22690.0021
Analysis scheme – Ks beam Simoultaneous KSKL production in a pure JPC = 1-- state: identification of a KS,L signals the presence of a KL,S (KL,S tag) KL “crash” = 0.22 (TOF) KS p-e+n KS tag through identification of KL interactions in EmC through a TOF technique: Neutral cluster in barrel, E>100MeV, 0.17<b<0.28 Efficiency ~ 30% (50% of KL’s decay before entering calorimeter) F KSKL is a two-body decay, tagged KS beam has definite momentum: KS angular resolution: ~ 1° (0.3 in f) KS momentum resolution: ~ 2 MeV
Analysis scheme – Signal selection 1) Two opposite charged tracks coming from IP, entering calorimeter, invariant mass at the vertex below KSp+p- peak: Mpp < Mcut TOF rejection of KSp+p- background: both tracks connected to EmC clusters, calculate dt(M) = Tcl – L/cb(M) 2) Test of pp mass hypothesis: |dt(Mp)1 - dt(Mp)2| > Tppcut 3) Test of ep, pe mass hypotheses: define ddtpe = dt(Me)2 - dt(Me)1 ddtep = dt(Me)1 - dt(Me)2 e-p+ |ddtpe| < Tpecut1 and ddtpe > Tepcut2 OR |ddtep| < Tpecut1 and ddtep > Tepcut2 e+p- Data MC pen
Analysis scheme – Counting signal events Use KS momentum from tag and PID from TOF, to test presence of neutrino: Emiss = ES - p1,2 + Me - p2,1 + Mp 2 Pmiss = |PS - p1- p2| 2001 data: L=170 pb-1 p-e+n c2/d.o.f = 1.5 N(p-e+n) = 3854 92 p+e-n c2/d.o.f = 1.1 N(p-e+n) = 3863 88 Data MC Emiss – pmiss MeV MeV Residual background due to KS pp, with p m before entering the tracking volume Fit data distribution to a linear combination of signal and background MC histograms Background MC histogram smeared with a gaussian, s = 2MeV
Efficiency estimates = To get BR: 1) normalize signal counts to KS p+p- counts in the same data set 2) correct for selection efficiencies = N(pen) N(pp) BR(KSpen) BR(KSp+p-) NSL eL(crash clust) epen(2 trks IP EmC) epp(2 trks IP EmC) epen(TCA T0 Trig) epp(T0 Trig) e(TOF) Rtag Rcosm Two methods to extract most of signal efficiencies: Selection relies on properties of p EmC clusters, better to estimate efficiencies from data: 1) extract single-particle probabilities from control samples; parametrize as a function of the relevant kinematic variables; weight MC by the ratio of data and MC efficiencies 2) directly measure efficiency in a sample of KL pen prompt decays, selected in evts with KSp+p- Signal efficiency 20% given the tag
Preliminary results from analysis* of year 2001 data BR(p-e+n) = (3.46 0.09 0.06)10-4 Branching fractions by charge: BR(p+e-n) = (3.33 0.08 0.05)10-4 Charge-independent analysis: BR(pen) = (6.81 0.12 0.10) 10-4 6 7 810-4 CMD-2 ’99 KL PDG average KLOE ’02 KLOE ’03 prelim. * http://www.infn.it/thesis/PS/253-Gatti-Dottorato.ps Test of the DS=DQ rule: Re x+ = (3.3 5.2 3.5)10-3 CPLEAR ‘98 KLOE ’03 prelim.
Preliminary results from analysis* of year 2001 data Charge asymmetry: AS = (19 17 6) 10-3 N(p-e+n)/e+ - N(p+e-n)/e- AS = N(p-e+n) /e+ + N(p+e-n)/e- KLOE KS K+e3 K0e3 K+m3 K0m3 |Vus| f+ (0) K0p- Extraction of |Vus|: E865 * http://www.infn.it/thesis/PS/253-Gatti-Dottorato.ps Preliminary KS pen (analysis of 170 pb-1) consistent with previous measurements
Uncertainty in efficiency determinations Systematic error from: 1) statistics of data and MC control samples 2) comparison between results from the two methods Analysis of data taken in year 2001, L170 pb-1 +??% due to radiative correction
Uncertainty in knowledge of signal and background shapes Not reliability of MC in reproducing background Emiss - Pmiss distribution MC bkg normalized to data in the range: Emiss- Pmiss(20-40) MeV Evts/MeV p-e+n p+e-n Data Y2001 – MC bkg Data Y2001 – MC bkg Emiss - pmiss (MeV) Dependence of the number of estimated counts on the higher limit of the fit range is the dominant source of systematic uncertainty Also a sizeable charge dependence of the background distribution...
Uncertainty in knowledge of shapes: TOF refinement Tried to correct MC for ratio of efficiencies data/MC, tracking, TCA, T0, Trigger, but it does not help much Different behavior of p+ and p- in EmC comparing data and MC: different paths inside the calorimeter (dlong) different dependence of the time delay (Tcl - Texp) at fixed path Ks decay vertex Cluster centroid dlong EmC DC Paritcle track Data - p+ on barrel, P=200 MeV Tcl-Texp (ns) Linear parametrization: Tcl-Texp = a + b dlong Fit separately in data and MC distinguishing: particle type: p+, p- , m+ , m- , e+ , e- dlong (cm) impact position: barrel, endcap momentum: 10 MeV slices
Uncertainty in knowledge of shapes: TOF parametrization In the end, get a parametrization with O(100ps) accuracy, merely using average values of the fitted parameters Data p+ MC p+ Tcl-Texp (ns) dlong (cm) dlong (cm)
Uncertainty in knowledge of shapes: inclusion of radiation Radiative modes: simulation at O(a), IR finite, without any cutoff in energy BR(KS ppg; Eg > 50 MeV) BR(KL peng; Eg > 30 MeV,qeg > 20) 0.26% 0.9% BR(KS pp(g)) BR(KL pen(g)) KS ppg KL peng ~0.26% ~3%
New estimate of the number of signal events – p-e+n Inclusion of events with one g in the final state and refinement of TOF estimates better significantly data/MC agreement for both signal and background
New estimate of the number of signal events – p+e-n Inclusion of events with one g in the final state and refinement of TOF estimates better significantly data/MC agreement for both signal and background
New estimate of the number of signal events – pen Inclusion of events with one g in the final state and refinement of TOF estimates better significantly data/MC agreement for both signal and background
New estimate of the number of signal events Improvement in fit stability as upper limit of fit range varies:
New estimate of the number of signal events The distance DPCA between tracks at PCA to IP discriminates background events with a p decaying to m before entering the tracking volume The fit correlation can be lowered cutting on DPCA, thus reducing the error on the signal counts
what has been understood... Conclusions – job done what has been understood... Necessary to include correct treatment of radiation: 1) At this level of statistical accuracy, radiation of photons is able to modify the signal shape 2) Radiation affects tracks acceptance for the signal at the 1% level (last-minute estimate), and for the normalizing sample at the few per-mil level (estimate included in published result for G(KSpp(g))/G(KS p0p0)) Necessary to calculate TOF including the contribution of time delay inside the EmC: 1) Main dependence of background spectra on details of pions simulation in EmC (dlong) is removed, better agreement of data and MC bkg spectra 2) Selection efficiency for background almost independent on charge ...and done... A selection in which TOF cuts have been loosened has been run on the entire data set, both for data and MC Result: total relative error on signal counts now at 1%, fit robust The 5 cut values (including DPCA) in the analysis have been varied independently in order to choose the optimal set
Conclusions – road map ...and what still remains Having chosen the optimal cut values: 1) Run over the control samples to estimate again the efficiencies both at single-particle level (efficiency estimate method 1) and at the whole-event level (efficiency estimate method 2) 2) Use the output to weight MC (efficiency estimate method 1) 3) Evaluate track acceptance both for signal and normalizing sample (efficiency estimate method 1) Check result stability along the data set, etc Explicitly try to identify events with a tagged photon in the final state
Spare slides
Vus from Kl3 decays |Vus|f+(0) K+e3 K0e3 K+m3 K0m3 KLOE KS KLOE KL K+e3 K0e3 K+m3 K0m3 |Vus|f+(0) Measurements for the extraction of Vus Canale BR(%) 103 l+ 103 l0 K+e3 4.870.06 27.81.9 K0e3 38.790.27 29.11.8 K+m3 3.270.06 3.31.0 49 K0m3 27.180.25 3.30.5 276 tL = (5.17 0.04) · 10-8 s tS = (8.9598 0.0007) · 10-11 s t = (1.2384 0.0024) · 10-8 s Preliminary KLOE results seem to confirm previous existing measurements of K0e3, K0m3 K+e3 will soon arrive… At KLOE: All BR’s will be measured at O(0.1%) Significatively improve on l+, l0, tL |Vus| = 0.2196 0.0019exp 0.0018th For Ke3 channel exp th f+ (0)= 0.961 0.008th K0p- 1.2% 0.2% 0.3% K+e3 0.7% 0.8% 0.3% K0e3