Update on ADS Sensitivity & Status report – ADS with B→D*K CP WG, 11 th May 2006 Mitesh Patel (CERN)

ADS Sensitivity Guy’s sensitivity study for the ADS modes made by estimating signal efficiencies/yields, values of relevant parameters and then looking at precision as a function of B/S, δ D K , δ D K3  : [CERN-LHCb ] –Assumed  TOT =0.5% for 2-body modes, B ± →D 0 K ±, D 0 →K , KK,  → ~2k events per year (B+ and B- together)suppressed modes → ~60k events per year favoured modes (Now confirmed by DC04 study) –Assumed  TOT =0.25% for 4-body mode, B ± →D 0 K ±, D 0 →K  → twice branching ratio hence yields as above (Still under study – Andrew) –Assumed :  = 60 o δ B = 130 o r B = 0.15→ may now be optimistic … r D K , r D K3  = cos (δ D ) known to ±0.20 (conservative estimate CLEO-C precision) –Tried full range of values for δ D K , δ D K3 , generate yields, fit for r B, δ B, δ D K , δ D K3  and 

Value r B =0.15 entirely consistent with information available at the time : Value of r B … ? B-B- B+B+ DK decays (~ 15 events; hint of asymmetry) D  decays (control channel; order 30 events) eg. Belle, hep-ex/ M BBbar BELLE measurements : –r B = 0.25±0.22 –δ B = 157±30 [hep-ph/ , – Dalitz analysis] –BR(suppressed) = (3.9±2.1)×10 -7 [hep-ph/ – search for ADS modes] BABAR measurements : –r B = 0.12±0.09 –δ B = 104±53 [hep-ph/ , – Dalitz analysis] We then assumed : r B = 0.15, δ B = 130 o, δ D K  =180 o (arb.) → BR(sup.) ~ 4.5×10 -7

Have now become aware of Belle’s updated ‘super-clean’ search for the ADS modes : Signal candidates they had previously have disappeared, without change in the signal acceptance Belle, hep-ex/ M BBbar i.e. 275M+111M

New situation from BELLE : Point that was taken by LHCb – r B = 0.15 R DK = (taking δ B =130,δ D K  =180 o ) Would expect 18 evts from 275M BB (cf. 15) and 28 evts from 386M BB (cf. 0) New best fit : r B = ± What happens to our sensitivity if we try UTfit’s new best fit value r B =0.077 … ?

r B =0.15r B =0.077 Clear that signal yields in suppressed modes go down a factor 2.5 but note a lot still hangs on what the strong phases are …

r B =0.15, w/o bkgrd, δ B = 130 o, δ D K  =180 o (arb.), 2fb -1 Taken from Guy’s note : r B =0.15, w/o bkgrd, δ B = 130 o, δ D K  =180 o (arb.), 2fb -1 Taken from my attempt at reproducing this study : Have not used quite same branching ratios/efficiencies but think that I have approx. compatible results

r B =0.077, w/o bkgrd, δ B = 130 o, δ D K  =180 o (arb.), 2fb -1 r B =0.15, w/o bkgrd, δ B = 130 o, δ D K  =180 o (arb.), 2fb -1 Now look at what difference changing to r B =0.077 makes : While precision at example point (δ B = 130 o, δ D K  =180 o ) is essentially unchanged, in general, (significant) decrease in statistics makes precision worse – but precision is not inherently poor

Add background to all modes with B/S=1 : Adding the background r B =0.077, B/S=1, δ B = 130 o, δ D K  =180 o (arb.), 2fb -1 r B =0.15, B/S=1, δ B = 130 o, δ D K  =180 o (arb.), 2fb w/o bkgrd : 4.1 w/o bkgrd : As might expect, decrease in signal yields with decrease in r B → background bigger effect

Different approach - use background levels from DC04 study NB : assumes that background and signal trigger efficiencies identical Background estimate B→ D 0 (K  )K : –bbar sample : 3 events in both sign combinations, wide mass window : 3/(2*10) = 0.15 event → 825 events/yr (2fb -1 = 1 year) –D 0  : favoured→ ~12500 events/yr suppressed→ ~12500*R D  BELLE = 44 events/yr ** Signal yield with r B =0.077, other parameters as before : Suppressed modes : 710 signal evts in 2-body : (cf. ~2000 events with r B =0.15) 530 B + → D 0 (K -   )K + B~ 870 i.e. B/S ~ B - → D 0 (K +   )K Favoured modes : unchanged - ~28k each B=12500 i.e. B/S =0.5 Make identical assumptions for K3  – probably overly optimistic ** BELLE have now measured R D  =BR(B→D sup  )/BR(B→D fav  )=3.5×10 -3

Treat D 0 →KK,  together : 4300 B + → D 0 (hh)K + B= ( )/2=4900 i.e. B/S ~ B - → D 0 (hh)K - r B =0.077, best-guess bkgrd, δ B = 130 o, δ D K  =180 o (arb.), 2fb -1 Background now a problem, will take another look – note that have assumed trigger efficiency, estimate based on only 3 events ! Up until now, assumed constraint from CLEOc on δ D K  has contributed nothing to the fit – now starts to look like it can help

Precision as fn of r B Guy demonstrated that, taking r B =0.15, are very robust to bkgrd – could cope with B/S=5 throughout and still get   ~6 o Why is precision so much worse… ? As r B decreases background has bigger and bigger effect – if took r B =0.15 and best-guess background   =6 o, same background at r B =0.077 gives   =9 o

Status report – ADS with B→D*K

A brief reminder : D*K has an extremely attractive feature –D*→D 0  0 – here the D* and D 0 have the same CP –D*→D 0  – here the D* and D 0 have opposite CP → relative 180 o offset to δ B in the expression for the rates If can distinguish the two decays → powerful additional constraint ! [Bondar and Gershon: hep-ph/ ] Last time, made first attempt to select D*→D 0  and looked at the background from D*→D 0  0 (have still not yet tried the reverse – reconstructing D*→D 0  0 ) Today show update on treating the D*→D 0  0 background and first look at the combinatoric background from the bbar sample (DC04, 20M evt)

D* Mass /MeV B ± → D*(D 0  0 )K ± B ± → D*(D 0  )K ± Have ~ 1 D 0  : 1 D 0  0 (Start with 1 : 1.6 ) B ± Mass /MeV  D* mass ~ 16 MeV Was requiring that we select all the particles from the D*→D 0  decay or the D 0 and one of the two  ’s from a D*→D 0  0 decay i.e. selected particles are associated to mc truth – no accidental/combinatoric  As first look, applied two-body ADS cuts, no additional requirement on  Making D* mass cut at 3  D* mass (  D* mass =16MeV), found could get : 1 D 0  : 1 D 0  0 Was concerned about efficiency : –B ± →D 0 (K  )K ± →  Sel/Gen = 3420*0.347/72k = –B ± →D*(D 0 (K  )  )K ± →  Sel/Gen = ~500*0.347/37k = Factor ~3.5 difference MCEffBuilder →  det diff. by factor 2

Talked to Vanya re: efficiency – seems that factor ~3 reduction in efficiency from searching for the photon ‘normal’ (?!) We have 50% x 0 before RICH2 Moreover, photons we are searching for have low energy : Have understood that calorimeter well calibrated for E T >>250MeV  energy /MeV  p T /MeV

Although in this case don’t have slow additional D* daughter like in D 0  ± decays, found was still better to cut on D*-D 0 mass difference : –  D*-D0 mass = 13.5 MeV –  D0 mass = 16 MeV Making cut on mass difference and optimising both this and the B mass window, find that with : 2.2  D*-D0 mass 1.7  B± (  B± = 25MeV ) mass cuts, can get : 3 D 0  : 1 D 0  0 … much healthier ! … but has cost yet more acceptance D*-D 0 Mass /MeV B ± Mass /MeV

Combinatoric bkgrd to D*  Values r B *,  B * not necessarily same as r B,  B but assuming they are : Factor 4.5 difference in rec. efficiency, D 0  BR suppressed by factor 2.7 cf. regular 2-body ADS modes → factor ~12 : –suppressed modes :530,180 B +, B - events → ~45, 15 events –favoured modes : 25k events in → ~2.1k → have to reduce the background from the 20M evt bbar sample to zero or we’re in big trouble … First look at the bbar : Find ~130 events in tight mass window ! Look pretty flat : So far only additional constraint on D*  analysis is on D*-D 0 mass difference, clear we need to add some cuts – further reduction in signal acceptance … B ± Mass /MeV

Looking through distributions of all of usual quantities for D* and  find only two where can get some sensible separation between signal and bkgrd : Making the cuts shown are still left with ~50 events (63 candidates) in wide mass window  energy /MeV D* SIPS

Use background categorisation tool to try understand these events : Have started to examine events in each of these categories : –From PV categories – clear pattern : D 0 or even D* from one B, presume  from PV attached to wrong vertex then makes correct masses -  energy resolution makes hard to confirm this by looking at true  energies –bbar category – D 0 or even D* from one B, often can’t find anything even close in energy/direction to any  from the other B Seems our poor energy resolution for these E~5 GeV  and the fact we assume the vertex is that of the D means we suck-in background Case and point : Low mass event – in fact looks like signal event to me, not sure why has been classed as low mass – for this event have the signal candidate, and 3 other candidates in which we ignore the real  and taking something random from elsewhere ! Total 30 Refln. 40 Part. rec. phys. 50 Low mass 60 Ghost 70 From PV 100 From diff. PV 110 bbar

Conclusions If r B < 0.15 precision suffers markedly – will take another look at the background to see if can improve matters D*  analysis progressing : –Looks like background from D*  0 can be made to have B/S~0.3 –Background from combinatoric sources formidable, under study