Документ взят из кэша поисковой машины. Адрес оригинального документа : http://zebu.uoregon.edu/~uochep/seminars/seminars-02/snowmass_talk.pdf Дата изменения: Fri Aug 16 21:24:57 2002 Дата индексирования: Tue Oct 2 12:19:18 2012 Кодировка:

SM HIGGS BRANCHING RATIO MEASUREMENTS AT A LINEAR COLLIDER



e

+e,

!

ZH

! p
bb;

s

=500

GeV



Snowmass 2001 Session P1 Working Group 2 J. Brau, C. Potter, and M. Iwasaki University of Oregon
1


PARAMETERS
We assume for this study: ps =500 GeV Linear Collider R Luminosity dtL = 500 fb,1 250 fb,1 running with 80 right polarized electrons 250 fb,1 running with 80 left polarized electrons 115, 120, 140 and 160 GeV Standard Model Higgs boson masses

2


DATA SIMULATION
Pandora v2.1 Monte Carlo M. Peskin includes: Polarized beams Beamstrahlung Initial state radiation Interface to Tauola and Pythia M. Iwasaki: decay Parton shower Hadronization

3


DETECTOR SIMULATION
NLD Large Detector Con guration: Vertex Detector: 5 m res., rin =1:2cm Central Tracker: 25-200 cm Electromagnetic Calorimeter: 200-250 cm Hadronic Calorimeter: 250-374 cm 3 T Magnetic Coil Muon Detector: 450-650 cm NLD detector simulation implemented on Root C++ libraries M. Iwasaki

4


EVENT SELECTION
We select for e+e, ! HZ ! l+l, l = e; Reconstruct all lepton pair masses in an event Select pair with mass closest to mZ Calculate recoil mass Apply cuts on masses: jmZ , ml+l,j 10 GeV mH , 10 GeV mrecoil mH + 20 GeV Include hadronic Z decays by scaling signal up by a factor of 4 D. Strom, LEP II experience
1200 350 1000 300

250 800

200 600 150 400 100

200 50

0 60 70 80 90 100 110 120

0 100 120 140 160 180 200 220

Signal event reconstructed Z and recoil mass distributions.
5


SIGNAL
e
+

Z Z
*

e

-

H

Cross sections for e+e, are in fb with 80 left Mode 115 H ! b b 5:9 H ! WW ? 0:68 H ! c c 0:24 H! + , 0:62 H ! gg 0:41 H ! ZZ ? 0:050

! ZH with Z ! l+l, l = e; polarized electrons. 120 140 160 3:5 1:5 0:24 0:74 2:4 5:8 0:14 0:064 0:0099 0:38 0:17 0:027 0:27 0:16 0:033 0:08 0:34 0:19
6


BACKGROUND
Approximately 29 31 36 39 115 120 140 of signal events pass the mass selection cuts and then subjected to decay mode cuts. A small fraction of backgrounds also pass the cuts. mary backgrounds, with cross sections for left,right larizations are: e+e, ! W +W , 14300; 1700 fb e+e, ! qq 16000; 11000 fb e+e, ! ZZ 560; 340 fb e+e, ! t t 740; 400 fb 160 are Prip o-

The most pernicious of these is e+e, ! ZZ , especially for the lighter Higgs cases. Therefore the Higgs mass is reconstructed using tracks and unassociated clusters and cuts are made at the Higgs decay mode level.
7


CUT-BASED DECAY MODE TAGS
For H ! + , : reconstructed Higgs mass inconsistent with Z mass low track multiplicity 6 For H ! WW ? ! 2 jets : high momentum lepton in event 10 GeV high momentum lepton is isolated Econe 10 GeV For H ! WW ? ! 4 jets : force event to 4 jets best jet pair must satisfy jmW , mjj j 10 GeV jet algorithm ycut value y32 0.04 thrust in Higgs frame 0.88
8


CUT-BASED TAGS CONT.
For H ! b : b force event to 2 jets calculate mp with ZVTop D. Jackson, impl. T. Abe require mp 2 GeV for at least one jet
t t

For H ! c : c force event to 2 jets tag jet charm if mp 2 GeV, Nsig 10, pjet=pkin 0.45 require no jet tagged as beauty, at least one jet tagged as charm, and neither jet contains tertiary vertices
t

For H ! gg : require no tags from preceding modes neither jet has secondary vertices no high momentum leptons 1 GeV
9


NEURAL NETWORK STRUCTURE AND TRAINING
In order to optimize these results, the parameters and their cut values were used as inputs to a neural network. The neural network has 14 input units one for each parameter, 15 hidden units, and 6 outputs one for each decay mode. It is fully connected and uses standard back propagation as its learning algorithm. To speed and perhaps improve the training, the parameters were mapped to the interval 0,1 by the map p 7! 1 , exp ,p=pcut2 ln 2. For each set parameters in an event H ! X , training asked the network to ouput a 1 for the H ! X output unit and a 0 for the other output units.

10


NEURAL NETWORK TOPOLOGY
1 15 30

0.000 2

0.192 16

0.038 31

0.000 3

0.013 17

0.996 32

0.351 4

0.003 18

0.000 33

0.528 5

0.170 19

0.000 34

0.730 6

0.000 20

0.000 35

1.000 7

0.102 21

0.000

0.706 8

0.000 22

0.631 9

0.870 23

0.500 10

0.000 24

0.500 11

0.000 25

0.272 12

0.807 26

0.000 13

0.003 27

0.765 14

1.000 28

0.000

0.000 29

0.013

State of the neural network for an event H ! c. c
11


NEURAL NETWORK OPTIMIZATION
The space C of all possible neural network output cut values is the unit cube in R6. Each point in C maps to signal S and background B for a given mode H ! X and thence to fractional branching ratio BR=BR = pS + B=S , purity p = S=S + B, and R e ciency = S= dtL. Minimizing pS + B=S for a particular mode mode H ! X is equivalent to nding the optimal set of neural network output cut values for H ! X . For a given mode H ! X , the boundary of the image of C under the p; map is the optimal purity e ciency curve. We sampled 6 and B for each mode in the cube on a S lattice with 10 points.

12


MISTAGS AND SIGNAL FOR 120 GEV CASE
The analyzed 500 fb,1 data sample is listed vertically. The number of signal event tags is listed horizontally. Sample H ! WW ? H ! b b H ! c c H! + , H ! gg H ! ZZ ? e+e, ! ZZ e+e, ! WW e+e, ! qq e+e, ! t t
WW ? 214 27.9 7.0 0.3 52.7 1.0 123.2 0 0 0 b b 12.7 1599 13.6 0 9.8 0.6 524.7 0 0 0 c c 3.3 59.7 29.3 0.3 3.0 0.1 38.6 0 0 0
+,

0.5 0 0.02 189.6 0 0 24.8 0 0 0

gg 98 13.9 12.2 0 112.8 0 161.1 0 0 0
13


PURITY EFFICIENCY PLOTS
H bb
purity purity

H cc 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.05 0.1 0.15 0.2 0.25 0.3 efficiency

0 0

0.05

0.1

0.15

0.2

0.25 0.3 efficiency

H gg
purity purity

H WW* 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.05 0.1 0.15 0.2 0.25 0.3 efficiency

0 0

0.05

0.1

0.15

0.2

0.25 0.3 efficiency

Purity vs. e ciency for the case mH = 120 GeV. The maximum possible e ciency is 0.31 due to mass cuts.
14


FRACTIONAL BRANCHING RATIO RESULTS
Listed below BR =B R. Mode H ! WW ? H ! b b H! + , H ! c c H ! gg H ! c + gg c are the fractional branching ratio errors 115 0.16 0.027 0.07 0.31 0.16 0.15 120 0.10 0.029 0.08 0.39 0.18 0.16 140 0.03 0.038 0.10 0.44 0.23 0.20 160 0.02 0.13 0.36 180 0.03 0.59 200 0.04 -

15


OTHER HIGGS BRANCHING RATIO STUDIES
Study HBB NK B BR BPI

ps GeV 500 300 350 350 500

R

dtL=f b,1 Mode P e, 50 ZH 0 50 ZH -0.95 500 ZH + H 0 500 ZH 0 500 ZH 0.8

H B B=M.D. Hildreth, T.L. Barklow and D.L. Burke, Phys. Rev. Lett., 49, 3441 1994 N K=I. Nakamura and K. Kawagoe, in Proceedings of the Workshop on Physics and Experiments with Linear Colliders, vol. II, World Scienti c, Singapore 1996. B=M. Battaglia, in Proceedings of the International Workshop on Linear Colliders LCWS99 1999. B R=G. Borisov and F. Richard, in Proceedings of the International Workshop on Linear Colliders LCWS99 1999. B P I=J. Brau, C. Potter and M. Iwasaki, in Proceedings of the Linear Collider Workshop LCWS2000 2000. 16


COMPARISON TO OTHER HIGGS BR STUDIES
The fractional branching ratio errors BR=BR from each study are shown in the table below. Here mH = 120 GeV. Mode H! H! H! H! H! H! HBB NK WW ? 0.48 b b 0.07 0.041 c c 0.80 gg + , 0.14 0.15 c + gg 0.39 c 0.17 B 0.054 0.024 0.083 0.055 0.06 BR 0.051 BPI 0.10 0.029 0.39 0.18 0.08 0.16

Given the di erent parameters assumed in each study, such a direct comparison may be misleading.
17


CONSISTENCY CHECK
The fractional branching ratio error BR=BR goes like R dtL,1=2. The former divided by the latter is plotted against the latter for the case mH = 120 GeV.
+1/2

( L)

90 80 70 60 50 40 30
Battaglia Nakamura/Kawagoe Brau/Potter/Iwasaki



BR/BR

H H H H H



WW bb gg cc + -

*

Hildreth/Barklow/Burke

20 10 0 0
Borisov/Richard

0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02 -1/2 ( L)

Broadly, the results are consistent though there is some discrepancy in the H ! c and H ! gg results. c
18


IMPROVING THE STUDY
By the end of Snowmass 2001, this study should be extended and improved in the following ways: Analyze higher Higgs mass cases. Confer with other authors to resolve di erences in results H ! c and H ! gg. c Consider how to apply this analysis to the light MSSM 0 in the decoupling limit. h

19