CPV measurements with Belle/KEKB Stephen L. Olsen Univ. of Hawai’i Feb 17, 2003 LCPAC meeting at KEK

B0B0 td  B0B0 V tb V cb KSKS J/  KSKS  V* 2  sin2  1 V tb V* tdtd V cb B0B0 B0B0 Sanda, Bigi & Carter: +  1 : interfere B  f CP with B  B  f CP (  ) V* td theory errors ~1% (aka sin2 

zz more B tags B - B B + B (tags) t   z/c βγ more B tags Now an established & well understood expt’l technique sin2  1 = 0.719±0.074±0.035

Belle & BaBar agree sin2  1 (Belle) =0.719±0.074± sin2  1 (BaBar) =0.741±0.067±0.033 sin2  1 (World Av.) =0.734±0.055 theory errors ~1% Agree on value, not name!! Agrees with SM

What’s next? sin2    shift to precision measurement mode high statistics better control of systematics measure other angles start with   measure sin2   in non-ccs decay modes sensitive to new physics

 2 (  ) from B  +   B0B0 B0B0 V* td tdtd V tb V V ++  ++  B0B0 +  V* 2 V 2 tdub  sin2  2 ub (aka sin2 

Must deal with “Penguin Pollution” i.e. additional, non-tree amplitudes with different strong & weak phases B0B0 ++  V tb V td * R q (  t)  1+q [A  cos(  m  t) + S  sin(  m  t)] q=+1  B 0 tag  1  B 0 tag direct CPV mixing-induced CPV

 t (ps) First results from Belle (Mar 02)  0.27   0.31 (stat.) (syst.)  S  =  1.21 A  = million B-meson pairs (42fb -1 ) 162 events in the signal region “Study of CPV Asymmetries in B 0   +  – Decays” PRL 89, (2002) Results indicate large CP asymmetries, outside of A 2 +S 2  1 allowed region

Outside physical region & some (~2  ) disagreement with BaBar

Changes since last March More data ! [85  10 6 B pairs (78 fb -1 )] Analysis improvements: better track reconstruction algorithm more sophisticated  t resolution function inclusion of additional signal candidates by optimizing event selection Thorough frequentist statistical analyses use of Monte Carlo (MC) pseudo-experiments based on control samples

 e + e -  qq (q=u,d,s,c) continuum background suppression  Event topology  Modified Fox-Wolfram moments  Fisher discriminants  Angular distribution  B flight direction  Combined into a single likelihood ratio  Select 2 regions for each flavor tag class  LR >  LRmin < LR  Event and time reconstruction (3) Flow Flavor tagging Vertex and  t Continuum suppression LRmin continuum     (MC) class 1class 2 class 3class 4 class 5class 6 B 0   +  – Selection

B 0      example ++ 

B 0   +  – candidates LR >  +  - : 57 K  : 22 qq : 406 total : 485 LRmin < LR ≤  +  - : 106 K  : 41 qq : 128 total : 275

Event and time reconstruction (4) Flow Flavor tagging Vertex and  t Continuum suppression The same algorithm as that used for sin2  1 meas. Resolution mostly determined by the tag-side vtx. B lifetime demonstration with 85 million B pairs Example vertices Vertex reconstruction B 0  D   , D*   , D*   , J/  K S and J/  K* 0 B 0 lifetime  0.018(stat) ps Time resolution (rms) 1.43ps (PGD02:  ps) B 0   +  – Selection

Time-dependent fit Unbinned maximum-likelihood fit (no physical-region constraint) 2 free parameters ( A , S   in the final fit  E-M bc dist. B 0  D   , D*   , D*   , J/  K S and J/  K* 0 Lifetime fit (single Gaussian outlier) The fit program reproduces our sin2  1 results

Reconstruction summary Now we are able to obtain A  and S   But let’s go through several crosschecks before opening the box. Established techniques for event selection background rejection flavor tagging vertexing time-difference (  t) fit In particular, background well under control Common techniques used for branching fractions,  m d,  B, sin2  1

B 0  K +  – control sample Positively-identified kaons (reversed particle-ID requirements w.r.t.  selection) total K  yield: 610 events LR > LRmin < LR ≤ 0.825

Mixing fit using B 0  K +    m d =0.55 ps  0.07 Consistent with the world average (0.489  0.008) ps -1 PDG2002 (OF  SF)/(OF+SF)

 :  B =(1.42  0.14) ps K  :  B =(1.46  0.08) ps BG shape fit Lifetime measurements world average (PDG2002) (1.542  0.016) ps  background treatment is correct ! Very different bkgnd fracs

CP fits to the B  K  sample q=+1 q=  1 S K  = 0.08  0.16 A K  =  0.03  0.11 ( consistent with counting analysis) No asymmetry

Null asymmetry tests A =   S =  Null asymmetry

    fit results After background subtraction 5-50 Still see a large CP Violation! 5-50 Asymmetry with background subtracted

Fit results After background subtraction Asymmetry with background subtracted 5-50 A  =  0.27(stat)  0.08(syst) S  =  1.23  0.41(stat) (syst)  0.07 data points with LR > curves from combined fit result

Likelihoods & errors The probability for such small S  errors is ~1.2%  we use most probable errors from toy-MC ln(L) is not parabolic

Physical region A  2 + S  2 ≤ 1 Probability that we have a fluctuation equal to or larger than the fit to data (input values at the physical boundary) 16.6% [Note] prob. outside the boundary 60.1% (~independent of statistics) How often are we outside the physical region ? A  =  0.27(stat)  0.08(syst) S  =  1.23  0.41(stat) (syst)  0.07 Fit results:

Cross-checks Prev result A  S 

3.4  Evidence for CP violation in B 0   +  – (A ,S  ) CL regions

Constraining  2 | P/T| =  (Gronau-Rosner PRD65, (2002) S  A 

 2 (deg.)  (deg.) allowed regions Input values for  1 and |P/T|   1 =23.5  (sin2  1 =0.73)  |P/T| = 0.3  2 constraint w/o isospin analysis !  both A  and S  large less restrictive on    < 0 favored  no constraint on  at 3  Constraints on  2

 2 (deg.)  (deg.) |P/T| = 0.15 |P/T| = 0.30 |P/T| = 0.45  Consistent with theoretical predictions  Larger |P/T| favored (  1 = 23.5  ) |P/T| dependence Constraints on  2 (cont’d)

Constraints on  2 78  ≤  2 ≤ 152  22 (for: 0.15  |P/T|  0.45)

 1 dependence is small 78  ≤  2 ≤ 152  (95.5% C.L.)

Strategies for  3 D 0  CP V ub A max ~ 2R ~ 78 fb –1 47 CP-even evts 50 CP-odd evts A = 0.12 ± fb –1 :  A/A max  ~0.3 Gronau, London, Wyler D 0  CP V cb 3 2 KK KK

Strategies for  3 (cont’d) doubly Cabibbo-suppressed A max ~ 1; but rate is small 80 fb –1 : K+K+ M bc Only ~ 15 D o  evts, Cabibbo-suppressed D o  down by ~1/20 V ub Atwood, Dunietz, Soni V cb BDoBDo This strategy is very clean but requires lots & lots of data

Are there non-SM CPV phases?

Measure sin2  1 using loop-dominated processes: Example:,  ’, K  K  no SM weak phases SM: sin2  1 = sin2  1 from B  J/  K S unless there are other, non-SM particles in the loop eff

similar to  (g-2) well defined technique & target –theory & expt’l errors are well controlled –errors on SM expectations  are small (~5%) SM terms are highly suppressed –SM loops contain t-quarks & W-bosons –  effects of heavy non-SM particles can be large look for ppm effects look for pp1 effects (i.e.~100%)  (g-2):sin2  1 eff : SM loop particle:  SM loop particles: t & W lowest-order SM diagrams look for effects of heavy new particles in a well understood SM loop process

These channels are very clean & the techniques are understood Won’t reach experimental limits until ~100 x more data

sin2  1 eff results: (SM: sin2   =+0.72± 0.05) 2.2σ off (hep-ex/ )  PRD(r) 78fb -1  0.73 ± 0.66 B KSB KS S   ± ± 0.36 B  ’K S BK+KKSBK+KKS OK

CPV with Belle (summary)  1 well established – next: high precision measurements  2 1 st expt’l limits are established –interesting near future  3 just beginning non-SM phases search has begun – – 2.2  discrepancy seen in  K S –BaBar has seen a similar discrepancy in  K S

Conclusion We’ve accomplished a lot in CPV There is still a lot more to be done KEKB & Belle are up to the task