Документ взят из кэша поисковой машины. Адрес оригинального документа : http://zebu.uoregon.edu/~uochep/talks/talks02/flavor-tagging.pdf Дата изменения: Thu Jan 23 21:56:55 2003 Дата индексирования: Tue Oct 2 12:06:07 2012 Кодировка: Поисковые слова: galactic collision

Flavor Tagging with a CCD Vertex Detector: Higgs BRs as a Benchmark

Chris Potter and Jim Brau University of Oregon Chicago Linear Collider Workshop January 8, 2002

J.Brau, Chicago,Jan 8, 2002

1


Outline
Review of Snowmass Study of Higgs Branching Ratio Measurement Investigation of Improved "Jet-level" NN tag Improved Event-level tags and new estimate of H cc precision

J.Brau, Chicago,Jan 8, 2002

2


Review of Snowmass Study
Higgs branching ratio measurement study: 500 GeV, 500 fb-1 MH = 115, 120, 140, 160, 180, and 200 GeV Pandora generator with ISR and beamstrahlung NLD Large detector w/ Fast Sim (LCD Root Tools) ZVTOP with neural net (LCD Root Tools)
J.Brau, Chicago,Jan 8, 2002

3


CCD Vertex Detector for the Future Linear Collider

~700,000,000 pixels standalone tracking w/ 5 barrels

J.Brau, Chicago,Jan 8, 2002

4


Event simulation
· · Pandora-pythia and Pythia v5.7 ­ beamstrahlung included and important Detector model : NLC L

e+ e- ZH H bb, , cc, gg, WW, e+ e- WW e+ e- e+ e- qq e+ e- tt

s = 500 GeV L = 500 fb-1 Analysis with Z l+ l- evts, scaled to Z qq (x 4)

Previous studies: Hildreth, Barklow, Burke, PRD49, 3441 (1994) I. Nakamura, K. Kawagoe, LCWS (1996) M. Battaglia, HU-P-264 (1999) G. Borisov, F. Richard, LAL-99-26 (1999) J.Brau,

Chicago,Jan 8, 2002

5


ZVTOP
· Vertex reconstruction is based on the SLD algorithm ZVTOP ­ D. Jackson, NIM A388, 247 (1997) Implemented in the ROOT based NLC software by T. Abe Provides secondary vertex reconstruction, and pt-corrected mass

· ·

M = pT + V2 + p

T

2

J.Brau, Chicago,Jan 8, 2002

6


Event Selection

J.Brau, Chicago,Jan 8, 2002

7


Neural Net Analysis
14 parameters have been defined to distinguish decay modes of the Higgs Boson, and the backgrounds. A neural net with 15 hidden units and 6 output units (one for each decay mode) was trained. Cuts on each of the 6 output units were determined for each decay mode to maximize S/S+B.
hSM cc

Parameters
plmax Econe mW(recon) thrustHiggs y32 ntracks mh(recon) Nsig nvert(jet 1) nvert(jet 2) mpt(jet 1) mpt(jet 2) pjet/pkin(jet 1) pjet/pkin(jet 2)

hSM bb hSM cc output
J.Brau, Chicago,Jan 8, 2002

8


Efficiency/Purity Curves from Neural Net

J.Brau, Chicago,Jan 8, 2002

9


Branching Ratio Errors (BR/BR)
s = 500 GeV, L = 500 fb-1

J.Brau, Chicago,Jan 8, 2002

10


Impact of Detector Parameters on BR Errors
MH = 140 GeV/c2 , s = 500 GeV, L = 500 fb-1
RINNER(cm) hit res (µm) H bb H H cc H gg H WW
*

1.2 5.0 3.8% 10% 46% 23% 3.5%

2.4 5.0 3.8% 10% 47% 22% 3.5%

1.2 3.0 3.8% 10% 42% 22% 3.5%
INNER

2.4 3.0 3,8% 10% 46% 22% 3.5%

1.2 4.0 3.8% 10% 42% 22% 3.5%

Mild dependence of charm to r

and hit resolution.

In this analysis, we are essentially tagging on one of the two possible jets. In an analysis in which one needs to tag multiple jets, the dependence should be stronger.
J.Brau, Chicago,Jan 8, 2002

11


Charm Backgrounds
There are two principal backgrounds to the charm measurement: the much larger bottom BR the ZZ* background

J.Brau, Chicago,Jan 8, 2002

12


Investigation of Jet Level Tags

· Five parameters to discriminate b and c ­ Largest 3D impact parameter sig in jet ­ Number of tracks with 3D-ip > 3
· these two jet-level parameters motivated by their effectiveness in TESLA study (R. Hawkings, LC-PHSIM2000-021)

­ Number of vertices in jet ­ Pt corrected mass ­ P(jet)

J.Brau, Chicago,Jan 8, 2002

13


Investigation of Jet Level Tags: Monojets (45 GeV)

Compares very well with results in TESLA study (R. Hawkings, LC-PHSIM-2000-021)

J.Brau, Chicago,Jan 8, 2002

14


Investigation of Jet Level Tags: Comparison

J.Brau, Chicago,Jan 8, 2002

15


New Jet Level Tag added to Event Level Tag
We have added the jet level tag described above to the event level tag: If an event tags as b or c, tag as H bb if at least 1 jet tags b tag as H cc if at no jet tags as b, and at least 1 tags as c Charm BR precision improves from 39% to 32% (at 500 GeV, 500 fb-1, Higgstrahlung only)

J.Brau, Chicago,Jan 8, 2002

16


Conclusions

H bb and e+e- ZZ* are significant backgrounds to H cc Improved H cc precision is obtained by adding the jet level maximum impact parameter, and jet level Nsig (39% -> 32%)

J.Brau, Chicago,Jan 8, 2002

17