D Results from E789 - MJL - 2/10/93 Data-taking: - Trigger: 2/4 matrix + matrix-up + matrix-dn + nx1.nx4 + susd + h-high(~45 GeV) + processor vertex cut ( zvtx>.04" & b>2mils ) - Targets: 900A Be - 150um x 1.5mm 900A Au - 150um x 1.5mm 1000A Be - 100um x .8mm 1000A Au - 100um x .8mm typical targetting fractions: ~0.3 - Rates (per event): Target Luminosity Silicon DC1 DC2 DC3 -------- --------- ------- ------- ------- ------- 900A Be 9.54e36 5.7-7.5 6.3-9.9 4.2-4.5 6.4-7 900A Au 9.54e36 7.9-11.9 4.8-8.7 3.5-3.7 4.9-5.1 1000A Be 5.61e36 5.9-11.2 6.7-10.6 4.5-4.8 7.8-8.3 1000A Au 2.12e37 7.0-12.6 5.2-9.9 4.3-4.5 6.1-6.8 - Target Motions and Silicon Alignment: +/-0.3" in X - 239 usefull tapes written with about 382e6 events and about 3970 reconstructed D0 -> K Pi events. Analysis: - Cuts and Requirements: Pass1: hadron trigger bit (trig 9) < 3 mubits (i.e. < 3 muon/station-4 counters fired along track) FASTSI impact parameter > 1.5 mils distance of closest approach <= .07" |y-interaction| <= .025" (primary interaction point) #yhits >= 6 and #uvhits >= 6 (in silicon vertex detector) |xtgt| <= 2", |ytgt| <= .05" zvtx >= .10" (900A); .08" (1000A) b1,b2 >= .003" (impact parameters) Pass2: y@176" and y@236" >= 4.5" Ndeltab >= 4 (about b1-b2 >= .006") Nyint <= 80 (about .008") Ntau >= 6.25 (about .4 ps) and various Ntau cuts from ntuples - Peak Fitting: Fitting K+Pi- and K-Pi+ simultaneously with the following three lineshapes: Gaussian peak for the correct PID KPi peak Wide gaussian (25 MeV sigma as in Monte Carlo) for wrong PID KPi Polynomial background (4 parameters) Peak area of the K+Pi- peak in both the K+Pi- and K-Pi+ mass spectra are constrained to eachother and visa versa. Typical resolutions: 5-6 MeV (900A) and 7-9 MeV (1000A). - Number of D -> KPi's obtained: Ntau > 6.25 9.5 12.5 15.5 ---- ---- ---- ---- 900A Be 839 668 484 323 900A Au 763 521 323 259 1000A Be 180 1000A Au 2065 1226 817 595 (some errors in Ntau calculation are being corrected) - Analysis done on four HP9000/730's with a pass1 analysis taking up to 30 hours per tape. - Spill based cuts (cutspil): selected target check using amon/sem ratio magnet current check for sm3 and sm12 Monte Carlo: - Complete simulations of the detector except for: no random tracks in the drift chambers (only random hits) only a very trivial simulation of the calorimeter - Acceptance and vertexing efficiencies (for Ntau=9.5) Acceptance Analysis/vertexing eff. ---------- ----------------------- 900A Be 0.00214 0.119 900A Au 0.00232 0.115 1000A Be 1000A Au 0.00129 0.122 (Be and Au differ only because of different beam X positions) Cross Sections and normalization factors: - Trigger Efficiency: determined by analyzing prescale (TFI) triggers with same analysis as used for D events including silicon tracking but with no vertex cuts. Then cut on D mass region and look at whether these events fired the h+h- trigger and it's components: 900A Be 1000A Au bits required cnts eff. cnts eff. ------------- ---- ---- ---- ---- all events 408 1.00 5408 1.00 nx1nx3 377 0.92 4740 0.88 susd 344 0.84 4076 0.75 muup.mudn 387 0.95 5051 0.93 above 3 together 297 0.73 3354 0.62 h+h- trigger 172 0.42 2459 0.46 cal thr. eff. 0.57 0.73 - Calorimeter threshold efficiency: (see above table) The calorimeter threshold efficiency is higher for 1000A since the threshold was not changed between 1000A and 900A but the hadron energies dropped. Simulations with a trivial threshold cut including smearing of the energy of the pair by the calorimeter resolution show that you can get these kind of efficiencies with a threshold of 45 GeV and a resolution of 3/sqrt(E). - K and Pi decay in flight: This is in the Monte Carlo where the appropriate decay for each momenta is used. For the average momenta a simple hand calculation shows the following survival fractions: KK ~.53 KPi ~.69 PiPi ~.93 - Hodoscope efficiencies: Prescale muon analysis of 1000A Au data where the TFI was 3/4muLR (all the 900A data was 4/4muLR or 4/4LR) with no silicon tracking. Events with at least one muon downstream track are kept. Then the hodoscope efficiencies are determined exluding the 15% edges of each hodoscope counter. A cut of Chi-squared per degree of freedom of 4 and number of drift chamber hits greater than 15 was made on the drift chamber tracks to assure good tracks. About six weak Y hodoscopes were found (worst about 66%) with the rest at or near 100% efficiency. These efficiencies were then used in the Monte Carlo calculations and are then part of the efficiency through the trigger requirement in the Monte Carlo. - Drift Chamber efficiencies: The final 900A Be D events are used. Full downstream+silicon tracks are used and the distribution of missed drift chamber planes is used to calculate the efficiencies of the drift chamber planes. Efficiencies are 90-95% with V1 and U2 having the lowest efficiencies. These efficiencies are then used in the Monte Carlo. - Silicon efficiencies: These are determined from the D dsts using the distribution of missing planes seen by the full silicon tracker. The efficiencies put into the Monte Carlo are not what you get out using this method for the similated data so the input efficiencies are tuned until the efficiencies obtained out of the analysis for the data and simulation agree. Efficiencies vary between 80 and 95%. - Processor efficiency: Processor force-thru events from most of data were analyzed with the same cuts used for the normal D analysis including vertex cuts. Then a mass cut to select events near the D mass was made. If we assume that these events are representative of the real D events (i.e. they have similar momentum resolution and vertex) then the processor efficiency is given by the number of these events which also pass the processor. efficiency lost due to hardware errors ------------------- --------------------------- 900A Be 200/270 = .74+/-.05 4 900A Au 250/356 = .70+/-.04 30 1000A Au 514/719 = .71+/-.04 ? 1000A Au (no proc. cut) 31/49 = .66 (real D's) same (all events at D mass) Simulation of the processor with Ron Jeppeson's simulation program (works with the regular Monte Carlo) using drift chamber and silicon efficiencies determined above yields about .79, not too far from the values measured above. Comparison of the simulation on real data with the hardware processors results show that for parts of the data the processor was not working properly (see the last column above). This presumably explains some of the remaining discrepancy between the simulation and the real processor. Cross sections: The factors needed to obtain a cross section from the number of counts are then: 900A Be 900A Au 1000A Au ------- ------- -------- protons/amon 7.1e7 6.24e6 1.18e7 nucleons/cm**2 1.67e23 1.74e24 9.31e23 acceptance 0.00214 0.00232 0.00133 vtx. eff. 0.119 0.115 0.122 proc. eff. 0.74 0.70 0.74 trig. eff. 0.42 0.42 0.46 amonsb 8.05e5 8.76e5 1.93e6 BR-KPi <----------- 0.0365 ------------> With these factors the resulting cross sections, A-dependence, and asymmetries are as follows, Ntau 900A cut sigBe sigAu alpha asymm. ---- --------- --------- ---------- -------- 6.25 21.6(1.3) 20.9(6.3) .975(.016) .78(.05) 9.5 24.5(1.7) 19.1(8.8) .919(.019) 1.01(.08) 12.5 27.1(1.6) 17.6(1.7) .861(.023) .94(.09) 15.5 29.7(12.4) 21.8(1.9) .900(.027) .91(.10) Ntau 1000A cut sigAu asymm. ---- --------- -------- 6.25 36.3(1.7) .70(.03) 9.5 28.8(2.0) .94(.05) 12.5 31.6(3.3) 1.01(.07) 15.5 36.2(1.9) 1.01(.15) Continuum - Study of continuum A dependence and size: The continuum level is substantially higher for 900A Au than for Be. Work is in progress to try to understand why this is so, with, e.g. Study of target h+h- fake downstream events with monte carlo. Study of D Dbar -> h+ h- fake downstream vertices using Pythia. Study of the target h+h- data A dependence using processor force-thru events. KK and PiPi - Consistant with estimates based on BR's, acceptance, and decay. Poor signal to noise at best. Systematics and Uncertainties: ?