1 High precision study of charge asymmetry in K ±  3  ± decays by NA48/2 at CERN SPS Evgueni Goudzovski (JINR, Dubna) on behalf of the NA48/2 Collaboration: Cambridge, CERN, Chicago, Dubna, Edinburgh, Ferrara, Firenze, Mainz, Northwestern, Perugia, Pisa, Saclay, Siegen, Torino, Vienna Les Rencontres de Physique de la Vallée d’Aoste La Thuile, March 9 th, 2006

2 Overview Direct CP violation in K ±  3  decays; NA48/2 experimental setup; Measurement principle; Systematic effects; The preliminary result in the K ±  3  ± mode based on the full collected statistics; Status of the K ±  0  0  ± analysis; Conclusions. E. Goudzovski / La Thuile, March 9 th, 2006

3 DCPV in K 3  decays Kinematics: s i = (P K  P  i ) 2, i=1,2,3 (3=odd  ); s 0 = (s 1 +s 2 +s 3 )/3; u = (s 3 -s 0 )/m  2 ; v = (s 2 -s 1 )/m  2. Kaon rest frame: u = 2m K ∙(m K /3  E odd )/m  2 ; v = 2m K ∙(E 1  E 2 )/m  2. The two decay modes: BR(K ±  ±  +   )=5.57%; BR(K ±  ±  0  0 )=1.73%. “charged” “neutral” Matrix element: |M(u,v)| 2 ~ 1 + gu + hu 2 + kv 2 E. Goudzovski / La Thuile, March 9 th, 2006 “Charged” mode g =  ± |h|, |k| ~ Direct CP-violating quantity: the slope asymmetry A g = (g +  g  )/(g + +g  )  0 U V DCPV: a window to physics beyond the SM

4 Measurements before NA48/2 E. Goudzovski / La Thuile, March 9 th, 2006 ”Charged” mode ”Charged” mode (K ±  3 π ± ) Ford et al. (1970) at BNL: A g =(  7.0±5.3)  Statistics: 3.2M K ± HyperCP prelim. (2000) at FNAL: A g =(2.2±1.5±3.7)  Statistics: 41.8M K +, 12.4M K  Statistics: 41.8M K +, 12.4M K  Systematics due to knowledge of magnetic fields Systematics due to knowledge of magnetic fields Published as PhD thesis W.-S.Choong LBNL Berkeley 2000 Published as PhD thesis W.-S.Choong LBNL Berkeley 2000 ”Neutral” mode ”Neutral” mode (K ±  π ± π 0 π 0 ) Smith et al. (1975) at CERN-PS: A g 0 =(1.9±12.3)  Statistics: K ± TNF (2005) at IHEP Protvino: A g 0 =(0.2±1.9)  Statistics: 0.62M K ±

5 Theoretical expectations E. Goudzovski / La Thuile, March 9 th, 2006 |A g | Experimental precisions before NA48/2: SM SUSY & Newphysics Ford et al. (1970) HyperCP prelim. (2000) TNF (2005) “neutral” NA48/2 proposal [  A g ~10 -3 ] “charged” “neutral” Asymmetry in decay widths expected to be smaller than in Dalitz-plot slopes (SM: ~10 -7 …10 -6 ). Smith et al. (1975) “neutral” Some models beyond SM predict substantial enhancement partially within the reach of NA48/2. few … few SM estimates vary within an order of magnitude (few … few ).

6 NA48/2 beams setup E. Goudzovski / La Thuile, March 9 th, cm cm m He tank + spectrometer Front-end achromat Momentum selection Quadrupole quadruplet Focusing  sweeping Second achromat Cleaning Beam spectrometer (resolution 0.7%) ~7  ppp, 400 GeV K+K+ KK Beams coincide within ~1mm all along 114m decay volume focusing beams BM z magnet vacuum tank not to scale K+K+ KK beam pipe Simultaneous K + and K  beams: large charge symmetrization of experimental conditions Be target 2-3M K/spill (  /K~10),  decay products stay in pipe. Flux ratio: K + /K –  1.8 P K spectra, 60  3 GeV/c

7 The NA48 detector E. Goudzovski / La Thuile, March 9 th, 2006 Main detector components: Magnetic spectrometer (4 DCHs): 4 views/DCH: redundancy  efficiency; Δp/p = 1.0% %*p [GeV/c] Δp/p = 1.0% %*p [GeV/c] Hodoscope fast trigger; precise time measurement (150ps). precise time measurement (150ps). Liquid Krypton EM calorimeter (LKr) High granularity, quasi-homogenious;  E /E = 3.2%/√E + 9%/E % [GeV].  E /E = 3.2%/√E + 9%/E % [GeV]. Hadron calorimeter, muon veto counters, photon vetoes. Beam pipe

8 NA48/2 data taking: completed E. Goudzovski / La Thuile, March 9 th, run: ~ 50 days 2004 run: ~ 60 days Total statistics in 2 years: K      +   : ~3·10 9 K    0  0   : ~1·10 8 Rare K ± decays: BR’s down to 10 –9 can be measured > 200 TB of data recorded A view of the NA48/2 beam line

9 Selected statistics E. Goudzovski / La Thuile, March 9 th, 2006 even pion in beam pipe 3.11x10 9 The final data sample: 3.11x10 9 events selected K + : 2.00x10 9 events  odd pion in beam pipe  M =1.7 MeV/c 2 Events  K  : 1.11x10 9 events

10  g measurement: fitting method E. Goudzovski / La Thuile, March 9 th, 2006 f(u,v) = n 1 + gu + hu 2 + kv (g+  g)u + hu 2 + kv 2 The “charged” mode: g =  ± |h|, |k| ~ analysis of 1-dimensional U spectra K + K  If K + and K  acceptances are made sufficiently similar,  g in general case  g can be extracted fitting the R(u,v) 2D-ratio R(u,v) with a non-linear function: f(u) = n∙(1+  gu) R 2 (u,v)=N + (u,v)/N  (u,v) R 1 (u)=N + (u)/N  (u) However, given the (small) slopes in the “charged” mode, R(u) 1D-ratio R(u) and linear function are sufficient approximations: (normalization is a free parameter)

11 Addressing the acceptance E. Goudzovski / La Thuile, March 9 th, 2006 Magnetic fields present in both beam line and spectrometer: This leads to residual charge asymmetry of the setup; Supersample Supersample data taking strategy: Beam line (achromat)Aweekly Beam line (achromat) polarity (A) reversed on weekly basis; Spectrometer magnetB ~daily~3 hours Spectrometer magnet polarity (B) reversed on a more frequent basis (~daily in 2003, ~3 hours in 2004) Example: Data taking from August 6 to September 7, 2003 B+B– B+ B–B– B–B– B–B– B–B– B–B– B–B– B–B– B–B– B–B– Supersample 1 Supersample 2 Supersample 3 12 subsamples 4 subsamples B–B– B–B– B–B– B–B– Achromat + Achromat – Week 1 Week 2 Week 3 Week 4 Week 5 Achromat +

12 Acceptance cancellation E. Goudzovski / La Thuile, March 9 th, 2006 Detector left-right asymmetry cancels in 4 ratios of K + over K – U-spectra: Z X Y Jura Salève Achromats: K + Up Achromats: K + Down B+ B–B– ’s Indices of R’s correspond to UD beam line polarity (U/D); SJ direction of kaon deviation in spectrometer magnet field (S/J). same deviation direction by spectrometer magnet in numerator and denominator; same deviation direction by spectrometer magnet in numerator and denominator; data from 2 different time periods used at this stage. data from 2 different time periods used at this stage. N(A+B+K+) N(A+B-K-) R US (u)= N(A+B-K+) N(A+B+K-) R UJ (u)= N(A-B+K+) N(A-B-K-) R DS (u)= N(A-B-K+) N(A-B+K-) R DJ (u)= within supersample

13 More cancellations E. Goudzovski / La Thuile, March 9 th, 2006 (2) Double ratio: cancellation of local beam line biases effects (slight differences in beam shapes and momentum spectra): R S = R US × R DS R J = R UJ × R DJ f 2 (u)=n∙(1+ Δg S  u) 2 f 2 (u)=n∙(1+ Δg J  u) 2 R = R US ×R UJ ×R DS ×R DJ f 4 (u)=n∙(1+  g  u) 4 (3) Quadruple ratio: both cancellations The method is independent of K + /K – flux ratio and relative sizes of the samples collected Δg = 2g  A g ≈ −0.43  A g [IMPORTANT: SIMULTANEIOUS BEAMS] (1) Double ratio: cancellation of global time instabilities (rate effects, analyzing magnet polarity inversion): [IMPORTANT: SIMULTANEIOUS BEAMS] R U = R US × R UJ R D = R DS × R DJ f 2 (u)=n∙(1+ Δg U  u) 2 f 2 (u)=n∙(1+ Δg D  u) 2 Normalization Slope difference

14 Beam spectra difference E. Goudzovski / La Thuile, March 9 th, 2006 Beam line polarity reversal almost reverses K + and K – beam spectra Systematic differences of K + and K – acceptance due to beam spectra mostly cancel in R U R D mostly cancel in R U ×R D Supersample 1 Supersample 2 SS 3 Systematic check: Reweighting K + events so as to equalize momentum spectra leads to a negligible effect  (Δg)=0.03  K+K+K+K+ K–K–K–K– (an example of cancellation)

15 Monte-Carlo simulation E. Goudzovski / La Thuile, March 9 th, 2006 Still Monte-Carlo is used to study systematic effects. Based on GEANT; Full detector geometry and material description; Local DCH inefficiencies simulated; Variations of beam geometry and DCH alignment are followed; Simulated statistics similar to experimental one. Example of data/MC agreement: mean beam K + data K  data K + MC K  MC K+K+ KK Due to acceptance cancellations, the analysis does not rely on Monte-Carlo to calculate acceptance Due to acceptance cancellations, the analysis does not rely on Monte-Carlo to calculate acceptance

16 Systematics: spectrometer E. Goudzovski / La Thuile, March 9 th, 2006 Time variations of spectrometer geometry: do not cancel in the result. Alignment is fine-tuned by scaling pion momenta (charge-asymmetrically) to equalize the reconstructed average K +,K − masses Maximumequivalent transverse shift: ~200  or ~120  or ~280  Sensitivity to DCH4 horizontal shift: |  M/  x|  1.5 keV/  m The effect of imperfect inversion of spectrometer field cancels in double ratio [simultaneous beams] Momentum scale adjusted anyway by constraining average reconstructed K masses to the PDG value Sensitivity to 10 −3 error on field integral:  M  100 keV Transverse alignment Magnetic field M3M3 M3M3 Max. effect in 2004  Much more stable alignment in 2004!

17 Systematics: beam geometry E. Goudzovski / La Thuile, March 9 th, 2006 Acceptance largely defined by central hole edge (R  10cm); Geometry variations, non-perfect superposition: asymmetric acceptance. Additional acceptance cut defined by a “virtual pipe” (R=11.5cm) centered on averaged reconstructed beam position as a function of charge, time and K momentum Y, cm Sample beam profile at DCH X, cm Beam widths: ~ 1 cm Beam movements: ~ 2 mm Y, cm “Virtual pipe” also corrects for the differences between the upper and lower beam paths 2mm 12% Statistics loss: 12% [Special treatment of permanent magnetic fields effect on measured beam positions]

18 Systematics: trigger E. Goudzovski / La Thuile, March 9 th, 2006 L2 trigger [online vertex reconstruction on DCH data] time-varying inefficiency (local DCH inefficiencies, tuning) 1–  = 0.06% to 1.5% u-dependent correction applied L1 trigger [2 hodoscope hits] small and stable inefficiency 1–  ≈ 0.9x10 -3 (no correction) Only charge-asymmetric trigger inefficiency dependent on u can bias the result Inefficiency, % L2 inefficiency vs time (2003 data) Beginning of 2003 run: L2 algorithm tuning Trigger efficiencies measured using control data samples triggered by downscaled low bias triggers Statistical errors due to limited sizes of the control samples are propagated into the result Inefficiency x10 3 L2 inefficiency vs u (normal conditions) U Max. inefficiency in 2004

19 Other systematics E. Goudzovski / La Thuile, March 9 th, MeV/c 2 No magnetic field correction correction Magnetic field corrected for Field map in decay volume: Y projection Decay volume: Z coordinate [Gauss] Residual effects of stray magnetic fields (magnetized vacuum tank, earth field) minimized by explicit field map correction Further systematic effects studied: Accuracy of beam tracking, variations of beam widths; Bias due to resolution in u; Sensitivity to fitting interval and method; Coupling of  decays to other effects; Effects due to event pile-up;  + /  – interactions with the material.

20 Supersamples collected E. Goudzovski / La Thuile, March 9 th, 2006 RunSupersampleDatesSubsamples K 3  events selected (millions) /06  25/ /08  20/ /08  03/ /09  07/ /05  07/ /06  07/ /07  19/ /07  01/ /08  11/ Total Supersample: a minimal independent self-consistent set of data (including all magnetic field polarities)

21  g fits in supersamples (quadruple ratios corrected for L2 trigger efficiency) SS0 SS1 SS2 SS3 SS4 SS5 SS6 SS7 SS8

22 Results in supersamples E. Goudzovski / La Thuile, March 9 th, 2006 RunSupersample  g x10 4 “raw”  g x10 4 L2-corrected  2 of the R 4 (u) fit (L2-corrected) ±1.4  0.8±1.8 30/26 1  0.4±1.8  0.5±1.8 24/26 2  1.5±2.0  1.4±2.0 28/ ±3.21.0±3.319/  2.8±1.9  2.0±2.2 18/ ±2.54.4±2.620/ ±2.15.0±2.226/ ±2.11.5±2.110/ ±2.30.4±2.323/26 Combined 0.7± ±0.7 Accepted by PLB; hep-ex/ [NEW]

23 Time-stability & control quantities E. Goudzovski / La Thuile, March 9 th, Accepted by PLB; hep-ex/ Physics asymmetry (results consistent) 2004 [NEW] R LR (u)=R S /R J Control quantities cancelling in the result quadruple ratio components rearranged (smallness demonstrates: second order effects are negligible) R UD (u)=R U /R D Control of setup time-variable biases Control of differences of the two beam paths Monte-Carlo (reproduces apparatus asymmetries)

24 Systematics summary E. Goudzovski / La Thuile, March 9 th, 2006 Systematic effect Effect on  g  10 4 Spectrometer alignment ±0.1 Momentum scale ±0.1 Acceptance and beam geometry ±0.2 Pion decay ±0.4 Accidental activity (pile-up) ±0.2 Resolution effects ±0.3 Total systematic uncertainty ±0.6 L1 trigger: uncertainty only±0.3 L2 trigger: correction  0.1±0.3 Total trigger correction  0.1±0.4 Systematic & trigger uncertainty ±0.7 Raw  g 0.7±0.7  g corrected for L2 inefficiency 0.6±0.7

25 The preliminary result Ford et al. (1970) HyperCP (2000) preliminary 2003 final 2003 preliminary NA48/2 (results superseding each other)  g = (0.6 ± 0.7 stat ± 0.4 trig ± 0.6 syst )   g = (0.6 ± 1.0)  A g = (  1.3 ± 1.5 stat ± 0.9 trig ± 1.4 syst )  A g = (  1.3 ± 2.3)   g = (0.6 ± 0.7 stat ± 0.4 trig ± 0.6 syst )   g = (0.6 ± 1.0)  A g = (  1.3 ± 1.5 stat ± 0.9 trig ± 1.4 syst )  A g = (  1.3 ± 2.3)  A NEW RESULT based on the full statistics accumulated in 2003 and 2004 runs Measurements of A g More than an order of magnitude better precision that the previous measurements; Uncertainties dominated by those of statistical nature; Result compatible with the Standard Model predictions; The design goal reached! There is still some room to improve the systematic uncertainties. E. Goudzovski / La Thuile, March 9 th, 2006 A g  10 4

26 The “neutral” mode analysis E. Goudzovski / La Thuile, March 9 th, 2006  M =1.1 MeV/c 2 M(3  ), GeV/c 2 Events U |V| The final result with the 2003 sample (based on ~48  10 6 events) [a new result superseding the old one!] A g 0 = (1.8±2.6)  Quadruple ratios R 4 (u) Statistical precision in A g 0 similar to “charged” mode: Ratio of “neutral” to “charged” statistics: N 0 /N ± ~1/30; Ratio of slopes: |g 0 /g ± |  1/3; More favourable Dalitz-plot distribution (gain factor f~1.5). (will be presented at Moriond’06)

27 Conclusions E. Goudzovski / La Thuile, March 9 th, 2006 New NA48/2 results on direct CP-violating charge asymmetry in K ±  3  slopes: ~10 times better precisions than previous measurements, still dominated by statistical contributions; The NA48/2 design goal reached, however further improvements of the analysis possible. A g = (  1.3 ± 1.5 stat ± 0.9 trig ± 1.4 syst )  A g = (  1.3 ± 2.3)  A g = (  1.3 ± 1.5 stat ± 0.9 trig ± 1.4 syst )  A g = (  1.3 ± 2.3)  A g 0 = (1.8±2.2 stat ±1.0 trig ±0.8 syst ±0.2 ext )  A g 0 = (1.8±2.2 stat ±1.0 trig ±0.8 syst ±0.2 ext )  A g 0 = (1.8 ± 2.6)  A g 0 = (1.8 ± 2.6)  A g 0 = (1.8±2.2 stat ±1.0 trig ±0.8 syst ±0.2 ext )  A g 0 = (1.8±2.2 stat ±1.0 trig ±0.8 syst ±0.2 ext )  A g 0 = (1.8 ± 2.6)  A g 0 = (1.8 ± 2.6)  “Charged” mode K ±  3  ± Full data set, preliminary “Neutral” mode K ±  0  0  ± ~½ of data set, final result

28 Conclusions E. Goudzovski / La Thuile, March 9 th, 2006 Spare slides

29 Consistency of 2003 & 2004 results E. Goudzovski / La Thuile, March 9 th, 2006 Slope difference 2003:  g = (  0.7±0.9 stat ±0.6 trig ±0.6 syst )  = (  0.7±1.3)  :  g = (1.8±1.0 stat ±0.5 trig ±0.6 syst )  = (1.8±1.2)  Slope asymmetry 2003: A g = (1.6±2.1 stat ±1.4 trig ±1.4 syst )  = (1.6±2.9)  : A g = (  4.1±2.2 stat ±1.1 trig ±1.4 syst )  = (  4.1±2.8)  Probability=15% Raw (L2-corrected) 2003 and 2004 results are 1.4  (1.7  ) away