Status report on analysis ofBR(KS )

A. Antonelli, M. Moulson,

Second KLOE Physics Workshop, Otranto, 10-12 June 2002

Notes on the decay KS

Branching ratio not well measured at present:

CPLEAR ’97 2.51.21.0

0.50.6×107

E621 ’96 4.81.1×107

PDG Avg. 3.21.21.0×107

Phenomenology & PT 2.4 0.7 × 107

Part of amplitude is CP violating:

L odd CP()I=0,2 = 1 Centrifugal barrierL even CP()I=1,3 =1 Violates CP

= A[KS ()CP]/A[KL ] CP– component extracted by integration over Dalitz plot, traditionally in interference measurements

BR[KS ()I=1] ~ BR[KS ] ~ 10-9

Unlikely to be observed at KLOE

General analysis considerations

0.34 events produced per pb-1:58 events in 170 pb-1 with 100% efficiency

Background rejection paramount:KL reconstruction needed for signal IDUse both KL crashes and vertices in DC

No reliable KS tag using DC vertices available:New EVCL algorithm to include KS events in ksl

stream starting from KS

Symmetric treatment of KL crash/vertexNot a tagged measurement in the usual KLOE sense

Acceptable to rely on MC for most efficiencies:KL crash studied in KS

KS additionally useful for normalization

Efficiency evaluation

MC does a miserable job at simulating KL crash:

Estimate KL crash efficiency from KS events in data

SLMC

SLdataLSMC KK

KKKK

|crash

|crash crash ,0

Useful mainly for estimating sensitivityFor the actual BR measurement:

crash ,/crash ,

crash ,/crash , 000

LSMCLS

LSMCLS

S

S

KKKKN

KKKKN

KBR

KBR

Fundamental assumption: KL crash detection probability cancels from ratio

Detection probability for KL vertex more likely to depend on KS decay mode, but ratio of efficiencies more amenable to simulation

Data and MC samples

Data: All 2002 dk0 DST’s as of 5 Jun 2002L dt, VLAB: 125.7 pb-1 46.0M total KL tagsUseful control sample (downscaled 50×)

MC kspppmp0, dedicated productionSignal 100K events

MC all_phid, 10 M eventsBackground Runs 81-160 (2.4 M events, DBV-8)

Runs 161-439 (7.6 M events, DBV-13)1.42M total KL tags, effective L dt 3.9 pb-1

Dedicated Ntuples made from DST’s and MC filesDST Processing rate: ~0.6 pb-1/CPU hourVolume of MC + data: 2.2 GB total

Control sample: KS

Unique vertex at origin, zero net charge:rxy < 5 cm, 20 cm < z < 20 cm

M MK < 10 MeV, P PK < 5 MeV

One track (or daughter of recognized kink) must analytically extrapolate to cluster:

d < 30 cm, E > 50 MeVAssociated cluster used to set event t0

(KL tag):

366 nb, using Ldatarec

436 nb, using = 3.1 b, KLtag(MC) = 60.7%

Event classification algorithm

Unique vertex at origin, zero net charge:rxy < 5 cm, 20 cm < z < 20 cm

Two prompt neutral clusters: (tR/c) < min(5, 2ns)

For each pair of clusters ( candidate):

• Close kinematics using vertex, mKS, m, and cluster directions

• Set t0 using clusters

• Search for KL crash in 20 cone

• If no KL crash, search for KL vertex in 20 cone

• Veto vertices using tracks in KS vertex tree in post-reconstruction analysis

If more than one candidate, correct candidate identified on basis of best alignment of KL direction and tagging momentum

Reconstruction of vertex

= 36.3%

Track momenta are low and anticorrelated

Vertex reconstruction efficiency dominated by acceptance

Reconstruction of clusters

E > 25 MeV = 73.3%

r > 60 cm= 96.6%

E1 + E2 >100 MeV <250 MeV = 99.98%

cos(12) > 75 = 98.6 %

Distributions are for clusters corresponding to MC truth, as reconstructed

Cut efficiencies shown evaluated in cascade

Expect overall 0 efficiency 70%

Find = 59.9%

Bug in tag!Cut on t instead of (tR/c)

Simulation of KL crash:

KS , MC KS , Data

spectra for MC and data completely different:

Evaluate KL crash efficiency from KS events in data

Stop decays?

Simulation of KL crash:

SMC

SMC

K

K

|Crash

|Crash 0

Relative efficiency: Lower cut on Upper cut on

Simulation of KL crash same for both KS decays

t0 correct in each case

SMC

Sdata

SMC

K

K

K

|Crash

|Crash

|Crash 0

Overall efficiency:

0.22

75

0.2

Simulation of KL crash: LS

SMC

SMC

K

K

|Crash

|Crash 0

SMC

Sdata

K

K

|Crash

|Crash

SMC

Sdata

SMC

K

K

K

|Crash

|Crash

|Crash 0

Overall efficiency:

cos-1 0.998.1°

LSLS

LS

Relative efficiency:

Significant influence from reconstruction of KS vertex

Relative efficiency:

Extent/position of KL crash in MC not as bad as timing

Preselection efficiency: KL crash

EfficiencyNo cuts on

LS

Default cuts on LS

present w/ bug fixed

vertex, MC 36.3 36.3 36.3

reconstruction, MC 59.1 59.1 69.6

Crash and KS MC 2.6 1.1 1.3

Crash given KS MC 15.1 7.2 7.2

Crash given KS data 31.7 25.6 25.6

Overall 5.5 3.9 4.7

FV for KL vertex analysis

Fiducial volume for KL vertex: 23.2%selects region in which reconstruction efficiency high and constantplays important role in eliminating backgrounds (mainly KK)

30 cm < rxy < 150 cm 100 cm < z < 100 cm

MC(K

L v

erte

x | K

S

)

MC(K

L v

erte

x | K

S

)

Preselection efficiency: KL vertex

EfficiencyNo cuts on

FV,LS

Default cuts on FV,LS

present w/ bug fixed

vertex, MC 36.3 36.3 36.3

reconstruction, MC 59.1 59.1 69.6

Vertex given KS MC 22.3 13.7 13.7

Vertex given KS data 21.8 13.3 13.3

Overall 3.5 1.9 2.2

Vertex efficiencies include acceptance and BR for KL decays with vertices

Principal backgrounds

Emphasis on KL crash sample for background elimination studiesCuts applied to KL vertex sample (almost) as an afterthought

MC background Crash Vertex

No cuts 70 439

/FV and LS cuts 3 54

Composition of MC background in KL crash events:26 KK events

• Both K, K both reconstructed and make vertexOne K or its daughter makes an unassociated crash clusterOther K decays to , 2 ’s fake at origin (no prompt ’s in event)

• One K reconstructed, other decays to before DC44 KS events

• Conversions on DC wall: vertex forms “fish” and tracks cross at origin• Dalitz decays: ee

Crash background studied by using vertex sample and/or relaxing , LS cuts

Background eliminationA

ngle

bet

wee

n

and

: c

os

Angle between in rest frame: cos

SignalKS

KK

Cuts made:|cos | < 0.85

cos < 0.85

Signal2646 2252 Crash3545 2922 Vertex

Background70 11 Crash439 69 Vertex

Background elimination

E1 = (E1 – E1close)/(E1)

E2 =

(E

2 – E

2clos

e )/

(E2)

Signal (crash & vertex)Background to vertexBackground to crash

Cuts made:E1

2 + E22 < (2.5)2

Signal2252 2103Crash2922 2731Vertex

Background11 3Crash69 28Vertex

and cuts applied

Status of analysisKL crash

Efficiency: 3.2%Events expected: 1.4 Events found: 11Events in MC bkg: 0

Efficiency: 1.5%Events expected: 0.6 Events found: 114Events in MC bkg: 6

KL vertex

Mmiss Mmiss

Summary and outlook

Bug in EVCL tag a disappointment, but:

• KL crash sample included in ksl stream by KL crash tag

• KL vertex sample in 2001 data useable for study, at least

Only ~10% loss in efficiency

Presumably amenable to simulation

• Fixed version of tag ready for DBV-14

2002 data taken so far may be reprocessed at some point anyway

Summary and outlook

Many additional handles on background:

KS Prompt cluster multiplicityEliminate candidates that cross on DC wall

KK More work on analytical expansion of vertex treeUse ADC information

Much additional work on physics backgrounds to do

Existing MC samples of KK

Dedicated MC generation of KS with ee and/or conversions

Work on noise and machine background not yet started

Prospects seem good for eliminating background entirely