EvtGen at Belle

Akimasa Ishikawa

KEK, IPNS

21 Jan 2005

EvtGen Standard generator at Belle.

Both BBbar and continuum events are generated with EvtGen. Version 00-10-07++ is used. To be updated coming version.

EvtGen package is included in Belle library. Anybody can easily modify EvtGen for private purpose by checking out th

e source codes from Belle CVS. In order to load EvtGen from BASF (Belle AnalysiS Framework), shared ob

ject module “evtgen” is created.

basf<<EOF

path create main

path add_module main evtgen

…..

EOF

Decay Table We use modified DECAY.DEC and evt.pdl.

Only includes tree level bc decays. Other rare decays, which proceed through bu, s, d, annihilation or exch

ange diagrams, are separated to study rare decay backgrounds. Decay table is basically tuned with PDG and HFAG results and Theory.

Both exclusive and inclusive measurements. D0, D+ and Ds spectra are tuned by inclusive measurements.

Only possible at e+e- B-factory Total width difference between B0 and B+ is introduced.

ex. BF(B+Xl) /BF(B0Xl)=1.08 Some particles are added to evt.pdl

ex. higher N and charmed baryons and X(3872) assuming hc’.

Please find the decay tables at http://www.belle.kek.jp/~akimasa/decaytable/2003/

Branching Fractions 7 inclusive decay modes are considered. Determined with inclusive measurements

BF(Blepton X)=10.9% BF(BDs X)=10.5% etc.

BF B0 B+ Average

DXlnu 0.209 0.277 0.218

DXtaunu 0.024 0.026 0.025

ccX 0.024 0.026 0.025

DsX 0.101 0.109 0.105

DDX 0.087 0.083 0.085

DX 0.458 0.436 0.447

BaryonX 0.097 0.093 0.095

Total 1.000 1.000 1.000

total width difference

D0, D+ and Ds spectra Belle has measured inclusive D meson spectra.

B(BDsX)=11.7+-0.1(stat)+-1.1(syst)+-2.9(Ds) % B(BD0X)=66.3+-0.2(stat)+-2.1(syst)+-1.5(D0K) % B(BD+X)=24.0+-0.2(stat)+-0.9(syst)+-1.4(D+K) %

Before

After

Very Very Preliminary

D0 70.5%

D0 68.2%

D+ 28.3%

D+ 26.8%Ds fixed to 10.5%

Next Step inclusive charmonia and charmed baryons.

flavor specific D mesons and charmed baryons

Pythia, continuum production Rolf Seuster developed Pythia interface.

Now in EvtGen. Better representation for heavy flavor fragmentaion.

Charm fragmentaion tuning with D0 and Ds* mesons. We use Bowler function D0 prefers pure Bowler(rQ=1) but Ds* prefers pure Lund(rQ=0) rQ =1 gives minimum combined 2/ndf. We choose rQ =1.

This analysis was done before obsevations of Ds(2317) and Ds*(2457), so we need tocheck again.

Red : Bowler

Solid : Lund

Dot : data

D0

Ds*

Event shape Tuning of event shape is also performed.

19 event shape variables are put into and search for minimum by changing the parameters Q0, Q’: termination of parton shower, q: transverse momentum, QCD, Lund a and b.

Thrust, cosThurst, Thrust major and minor, oblateness, Planarity, heavy and light Jet Mass, Jet Broadening W and T, Sphericity, Aplanarity, R1, R2, R3, y23, y24, CPar, DPar

Belle EvtGen pythia

Q0 0.5 1.0 1.0

Q’ 0.3 0.3 0.8

q0.42 0.36 0.36

QCD0.3 0.3 0.3

A 0.32 0.3 0.3

B 0.62 0.58 0.58

Photos We only use photos for semi-leptonic and leptonic decays. However radiative effect is large especially two body decays, B. We need to turn on photos for all decay modes.

events K+K-

without photos 8000+-90 7577+-87

with photos 7474+-87 7282+-85

difference 6.6+-1.5% 3.9+-1.6%

B

E MbcE

Interference Weight Elisabetta Barberio, who is one of the photos authors, pointed out th

at interference weight for multi body decay Wmulti is not introduced in EvtGen except for two body particle-anti-particle decay, such as B

Wmulti should be introduced.

Speed, Random Generator Speed

EvtGen is about 2.5 times slower than qq98. 350Hz and 300Hz for BBbar and uds events on Pentium3 1266MHz, respecti

vely 30fb-1/day and 11fb-1/day. 33days and 90days to generate 1/ab MC (about 3x data we have). Need faster CPU or parallel production.

True random number generator board for parallel production Pick up thermal noise in semiconductor. Distribute random number via network. About 7M single precision random numbers/sec 800~900 random numbers to generate BBbar events. ~8kHz for BBbar events 600fb-1/day A dozen random boards are enough even at Super-B factory. Cost $6000/board

http://www3.toshiba.co.jp/power/techno/secret/random/index_j.htm

In japanese

Platform OS/compiler

Linux/gcc2,gcc3 We will/should use PC Linux to get enough computing power.

Solaris/gcc3, SunWS5.0 cc SunWS5.0 cc does not support template member function, so we cannot compile

EvtGen with it. Need to switch to Sun one studio 8/9. FreeBSD/gcc2, gcc3

Some institute at Belle used.

We want to try these compiler. PGI compiler Intel C/C++ Compiler

What is needed (my opinion) Decay models should allow that lower resonance decays to higher re

sonance. It is possible if both resonances have large width. Ex. K1(1270)K0*(1430).

Decay models for Higgs mediated BD and B decays. One of the inportant decay mode at Super B-factory.

Parity (and C-Parity?) should be defined in evt.pdl.

Summary EvtGen is standard generator at Belle. We improve the decay table by measuring D meson spectra.

Next target is charmonia and charmed baryons. Next to Next is flavor specific D mesons and charmed baryons.

We also tune the continuum production with D mesons and event shapes.

Photos should be applied to all decay modes. Interference weight should be introduced. EvtGen is not so fast but enough processing speed is possile with pa

rallel production with random generator allows

Back up

History

qq98

EvtGen

1999 2000 2001 2002 2003 2004 2005

~2000: qq98 for both signal/BG study. End of 2000: EvtGen is implemented to make a better flavor tagging

tool and study CPV in bccs. 2001: Observation of CPV. 2001~2003 some analyses used EvtGen for signal study (B, B

Kll, bul etc.). 2004: start background MC mass production with EvtGen.

Signal MC

Background MC

qq98

EvtGen