taikan suehara, the 7th general meeting of the ilc physics working group, 2008/11/08 page 1 full...

Taikan Suehara, The 7th General Meeting of the ILC Physics Working Group, 2008/11/08 page 1

Full Simulation ofFull Simulation ofTau-pair/SUSY eventsTau-pair/SUSY events

(+ jet clustering)(+ jet clustering)

Taikan Suehara

ICEPP, The Univ. of Tokyo


• SUSY chargino/neutralino2 pair– Angular distribution

• Tau-pair• Jet clustering w/ vertex information

TopicsTopics


SUSY chargino/neutralino2 SUSY chargino/neutralino2 pair (Quick view)pair (Quick view)


• Point 5 neutralino (x02) / chargino– Chargino/neutralino decay to W/Z+LSP

• W/Z separation by jet invariant mass

SUSY parametersSUSY parameters


• Qqbar 4jet events, 130(chargino),30(neutralino) fb• 500 fb-1 with full simulation• Chargino/neutralino must be separated by

identifying daughter W/Z.• Main BG: SM WW -> 4 jets (10000 fb)

SUSY jet modeSUSY jet mode


Cut statistics (1)Cut statistics (1)Process SM 4jet

Geometry GLD GLD' LDC GLD GLD' LDC LDC'Cross section 10000 fbLuminosity 10 fb-1J et events 73310 72572 73263 13923 13930 13929 101905

MC 4jet events 36546 36206 36550 7948 7948 7948 -eL < 20GeV 47700 46591 46893 11604 11602 11586 83968jet angle <0.9 36748 35921 35937 9295 9287 9225 21419all jet > 5GeV 34935 34480 34436 8711 8706 8663 21413Evis < 300GeV 34935 34480 34436 8711 8706 8663 682

Chargino Neutralino

158 fb 29 fb500 fb-1

• # of SM events becomes comparable to chargino after these

cuts.• Full SM bg investigation is in preparation.

(hopefully shown in LCWS)


W/Z mass peak – best W/Z mass peak – best combinationcombination

Cut condition– Same cut as

previous slide– Best-W combination

for chargino events,Best-Z combinationfor neutralino events(minimum 2 combination, 1 width is set to 5 GeVfor both W and Z)

• Peak shift corrected (by adding 0.8 GeV to all jets).


22 distribution / mass cut distribution / mass cut


Cut statistics (2)Cut statistics (2)Process SM 4jet

Geometry GLD GLD' LDC GLD GLD' LDC LDC'Cross section 10000 fbLuminosity 10 fb-1J et events 73310 72572 73263 13923 13930 13929 101905

MC 4jet events 36546 36206 36550 7948 7948 7948 -Cuts after (1) 34935 34480 34436 8711 8706 8663 682

cos(theta_W)<0.8 24897 24595 24672 7069 7100 7068 170acop_w > 10deg 21293 20977 21043 5003 4985 4990 35(MC 4jet in it) 15980 15889 15903 - - - -W mass cut 5022 4908 4985 180 199 189 2

(MC 4jet in it) 4999 4897 4972 - - - -MC 4jet efficiency 13.68% 13.53% 13.60% - - - -

BG acception - - - 1.29% 1.43% 1.36% 0.00%Cuts after (1) 34935 34480 34436 8711 8706 8663 682

cos(theta_Z)<0.8 24648 24342 24425 7034 7108 7045 171(MC 4jet in it) - - - 4870 4946 4896 -Z mass cut 187 210 210 1308 1364 1346 8

(MC 4jet in it) - - - 1303 1360 1339 -MC 4jet efficiency - - - 16.39% 17.11% 16.85% -

BG acception 0.26% 0.29% 0.29% - - - 0.01%

Chargino Neutralino

158 fb 29 fb500 fb-1

• BG separation is efficient (see W/Z mass cut rows)• Slightly better BG separation performance in GLD

but almost within statistical fluctuations.

chargino

neutralino


• SUSY mass is determined by kinematics– Daughter W/Zs have characteristic energy

distribution• Fitting function:

– For neutralino: erf(left) x erfc(right)– For chargino: erf(left),

conbolution of linear and erfc for right• Generator-level energy distributions (by

Akiya):

Mass fit of chargino/neutralinoMass fit of chargino/neutralino

Chargino (W) Neutralino (Z)


Chargino mass fitChargino mass fit

MC cut and nocut shape is almost the same (wo left edge).No significant difference between geometries can be seen.


• Fitting function:3rd polynomial (4 param) (center) / 0 (edge) convoluted with a Gaussian with as linear function of energy (2 param)edge position: 2 param, total # of parameters = 8

• Cheat fitting:1. fix edge positions at true value, fit other 6 parameters.2. fix those 6 parameters and fit edge positions.

• No significant difference between geometries.

Chargino mass fit resultsChargino mass fit resultsinput unit

Chargino mass 215.41 ± 1.15 216.29 ± 1.55 214.99 ± 1.20 210.21 GeVLSP mass 121.59 ± 0.72 120.78 ± 0.89 120.40 ± 0.76 117.36 GeVChargino (cheat) 210.51 ± 0.56 210.67 ± 0.51 210.46 ± 0.52 210.21 GeVLSP (cheat) 117.59 ± 0.58 117.61 ± 0.47 117.49 ± 0.46 117.36 GeVWidth L (cheat) 3.59 ± 0.11 3.30 ± 0.08 3.46 ± 0.12 GeVWidth R (cheat) 3.79 ± 0.21 3.69 ± 0.17 3.67 ± 0.19 GeV

GLD GLD' J 4LDC


Neutralino mass fitNeutralino mass fit

Distribution in LDC is slightly broader (worse resolution).Chargino background might not be negligible for lower edge.


• Fitting function:Error function (left) x Complementary error function (right)Width of left and right is the same, # of parameters = 4

• Cheat fitting:Same as chargino

• J4LDC gives slightly larger width than other two.Corresponding fitting error is enhanced.

Neutralino mass fit resultsNeutralino mass fit resultsinput unit

Xn2 mass 214.61 ± 0.50 215.31 ± 0.46 214.04 ± 0.63 210.21 GeVLSP mass 120.51 ± 0.33 120.88 ± 0.32 120.40 ± 0.36 117.36 GeVXn2 (cheat) 214.06 ± 0.36 214.67 ± 0.36 214.14 ± 0.39 210.21 GeVLSP (cheat) 120.34 ± 0.31 120.78 ± 0.31 120.42 ± 0.34 117.36 GeVWidth (cheat) 7.42 ± 0.20 7.20 ± 0.22 8.04 ± 0.21 GeV

GLD GLD' J 4LDC


• Angular distribution should be checked to obtain SPIN information of the new particles.

• Helicity angle– Jet emission angle in W-rest frame– Charge-ID is not efficient so only |cos()|

information can be obtained.• Charge ID does not checked yet…

• Production angle– Can be extracted by 4-momenta of 2 Ws– Back-to-back is assumed– 2 solutions though

Angular distributionAngular distribution


LDC’ (Events by European side)LDC’ (Events by European side)

Helicity angle: Flat??Production angle: Flat??

Wrong spin treatment in LDC’? (need to be checked)


Comment for LCWSComment for LCWS• Daniela will talk at LCWS instead of

me.– They are trying kinematic fit for W/Z

reconstruction/separation.• Articles supplied by us:

– Cross section measurement by full BG scan by LDC’ SM sample

– Massfit result– Angular distribution

• To be discussed with DESY people…


Tau-pairTau-pair(Quick view)(Quick view)


• Event characteristics– Sqrt(s) = 500 GeV, including ISR/FSR– 2 jets,1 or 3 prongs/jet, very condensing

(taus in 500 GeV beam are highly boosted)– Cross section is large, ~2.3 pb-1

• Observables, total and differential cross section – AFB, forward-backword asymmetry– Apol, polarization asymmetry (paticularly in polarized beams)

• Background and Statistics– Background: bhabha (34000 pb), -> (1500 pb)– Signal: 20 fb-1 (~46000 events)– Bhabha: 0.2 fb-1 preselected full-simulated events -> 0.7 fb-1 generator info

Tau-pair eventsTau-pair events

Typical event


Decay modes in ADecay modes in Apolpol analysis analysis


Analysis flowAnalysis flow

PFO particles

JetsNo

Yes 8.6%

7.0%


1. 1 prong cutJets with >2 charged particle rejected.

2. Lepton vetoEvents containing e/s are rejected.(criteria is the same as AFB lepton-pair veto)

3. Energy cutJets with energy < 10 GeV rejected.(e// separation is inefficient in low energy)

4. Events with > 1 GeV neutral particles are rejected.In “tight cut” event with any neutrals are rejected.

-> -> selection cuts selection cuts


Selection performance between geometries(look at the 2nd row from the bottom)

• Efficiency: not so different• Purity: LDC’ > GLD > GLD’ > J4LDC

– -> mode (decay 2 is mis-reconstructed as single)might be the reason (larger is better)

– LDC’ has advantage due to high CAL granularity.

-> -> selection results selection resultsGeometry

eff. purity eff. purity eff. purity eff. purityNO cut 100.00% 10.89% 100.00% 10.88% 100.00% 10.90% 100.00% 10.90%1+1 jet 67.87% 11.06% 66.49% 11.07% 71.39% 11.23% 72.50% 11.70%

opening angle>170deg 30.01% 11.05% 29.83% 11.05% 30.38% 11.12% 30.43% 11.20%AFB cut 25.20% 11.98% 25.07% 11.98% 25.23% 12.10% 25.17% 12.11%1 prong 25.17% 14.55% 25.06% 14.57% 25.22% 14.69% 25.16% 14.61%

J et energy cut 24.32% 14.50% 24.24% 14.54% 24.36% 14.66% 24.34% 14.58%e,mu veto 23.32% 24.26% 22.88% 24.02% 23.00% 24.53% 23.59% 23.98%

No gamma cut 21.29% 85.73% 21.37% 83.58% 21.43% 80.84% 21.16% 88.50%No gamma cut (tight) 20.54% 86.89% 20.56% 84.57% 20.66% 81.95% 20.42% 89.22%

GLD J 4LDC LDC'GLD'


AApolpol calculation ( calculation ( mode) mode)

Stat errorin 500fb-1

Pol estat shift estat shiftGLD 47.17%± 4.54% 1.25% -7.01% 54.89%± 4.67% 1.28% -4.49%GLD' 49.45%± 4.52% 1.25% -9.76% 52.11%± 4.64% 1.28% -7.65%J 4LDC 49.14%± 4.60% 1.28% -12.41% 52.20%± 4.68% 1.30% -10.28%LDC' 52.72%± 4.30% 1.22% -5.46% 57.95%± 4.49% 1.27% -3.25%GLD -25.62%± 4.77% 1.35% -6.20% -25.41%± 5.23% 1.48% -7.58%GLD' -24.04%± 4.79% 1.36% -9.23% -23.33%± 5.18% 1.47% -9.81%J 4LDC -28.57%± 4.88% 1.38% -7.58% -27.73%± 5.22% 1.48% -9.63%LDC' -18.93%± 4.63% 1.33% -6.57% -19.11%± 5.12% 1.48% -6.15%

Apol (count) Apol(linear fit)

eL(80%)

eR(80%)

Value shift due to the mode BG

Statistical error is almost the same for all geometriesValue shifts are larger in GLD’/J4LDC due to the lower purity.

Values obtained bysignal-only events!


1. 1 prong cut2. Lepton veto3. Energy cut (jet energy must be > 10 GeV)Above are same as -> cuts4. Events with > 10 GeV from neutrals (in total) are

selected.5. Mass of is reconstructed, must be within 200 MeV

from actual mass (770 MeV).6. Mass of p0 is reconstructed with neutral particles.

If # of neutrals >=3, nearest (in angle) two are combined until 2 particles are left.Application of this cut is discussed later.

-> -> selection cuts selection cuts


• Clear difference observed in invariant mass distributions.– LDC’s best, larger is better in Jupiter geometries.– Mark confirmed the granularity affects the mass distributions.

• Three candidates in mode selection– No 0 mass cut, 0 cut with left edge included / excluded

and and 00 reconstruction reconstruction


• 3rd row from bottom: used as “no 0 mass cut”.• 2nd row from bottom: used as “0 mass cut”.

– Events with single neutral are survived with this cut.• Most bottom row: used as “tight 0 mass cut”.

– Events with single neutral are eliminated with this cut.• Clear difference by geometries:

LDC’s the best, bigger is better in Jupiter’s.

-> -> selection results selection resultsGeometry

eff. purity eff. purity eff. purity eff. purityNO cut 100.00% 25.36% 100.00% 25.35% 100.00% 25.35% 100.00% 25.26%1+1 jet 66.69% 25.33% 65.54% 25.43% 69.26% 25.35% 70.31% 26.30%

opening angle>170deg 29.46% 25.28% 29.29% 25.29% 29.65% 25.24% 29.63% 25.28%AFB cut 24.63% 27.28% 24.45% 27.22% 24.30% 27.11% 24.43% 27.25%1 prong 23.30% 31.38% 23.10% 31.30% 23.02% 31.19% 23.07% 31.06%

J et energy cut 23.14% 32.15% 22.96% 32.10% 22.87% 32.00% 22.95% 31.87%e,mu veto 22.08% 51.22% 21.86% 51.14% 21.67% 51.14% 21.97% 50.64%

> 1 GeV gamma 19.07% 65.83% 18.49% 65.44% 17.96% 65.19% 19.69% 65.54%570<mRho<970 12.70% 83.38% 12.05% 81.80% 11.26% 81.39% 12.77% 85.71%

mPi0<200 10.41% 88.71% 9.81% 86.77% 8.95% 85.90% 9.73% 89.84%0<mPi0<200 5.31% 92.30% 4.32% 90.32% 3.72% 90.48% 6.38% 93.88%

GLD GLD' J 4LDC LDC'


• Clear difference between eL and eR observed.

• Distribution is degraded due to the cut effects.

-> -> , , ->-> distribution (1) no distribution (1) no 00 cutcut

Ppol vs dist. calc


• Number of signal is about a half.• Difference between geometry enhanced.

– J4LDC is not realistic with this cut?• Background is quite low, negligible level.

-> -> , , ->-> distribution (2) tight distribution (2) tight 00 cutcut


• Combined information of -> and -> decay can be used in this method.

Obtaining P(Obtaining P() value) value

Physics Letters B, 235 (1990) 198


AApolpol calculation ( calculation ( mode) mode)

Pol estat shift estat shiftGLD 34.06%± 4.26% 1.17% -2.68% 34.53%± 6.78% 1.86% -1.66%GLD' 38.66%± 4.30% 1.19% -3.59% 42.62%± 7.36% 2.04% -1.10%J 4LDC 34.86%± 4.47% 1.24% -4.24% 36.30%± 8.24% 2.29% 0.79%LDC' 35.62%± 4.13% 1.17% -3.36% 36.81%± 6.05% 1.72% -0.99%GLD -28.33%± 4.87% 1.37% 4.91% -30.89%± 8.32% 2.35% 3.70%GLD' -30.87%± 5.00% 1.42% 3.67% -34.26%± 9.36% 2.66% 0.88%J 4LDC -35.34%± 5.38% 1.52% 2.53% -36.45%± 11.18% 3.16% -1.90%LDC' -32.70%± 4.89% 1.41% 2.89% -32.46%± 7.86% 2.27% -0.49%

Apol (nopimasscut) Apol(wpimasscut)

eL(80%)

eR(80%)

Stat errorin 500fb-1

Value shift due to the mode BG

Statistical errors are larger in GLD’/LDC, esp. with m0 cut.Value shift is smaller than mode, negligible with m0 cut.

Values obtained bysignal-only events!


• First talk in SUSY/NP session• Basically the same as Cambridge

– No significant progress since then…• Full SM scan and cross section

measurement will be newly prepared.• If possible, improvements of

polarization measurements are included.

• Theoretical view by (Y.)Okada-san will also be presented.

LCWS planLCWS plan


Jet clusteringJet clustering


• Goals– Reducing misclustering in multi-jet final states

• ttbar, ZHH, ttH etc…– Improving b/c-tagging efficiency

• Method– Finding secondary vertices and use them as ‘jet-

core’– Combining traditional ‘y’ method

(‘y’ definition might be changed from Durham or Jade)

• Remaining jet-cores from light quarks• Fragment association

OverviewOverview


Ttbar sampleTtbar sample


• First vertex search was implemented.– Criteria

• Obtaining crossing points (PCAs) of charged tracks(Either is required to be off-vertex; tracks are treated as simple lines)

• Finding neighbor PCA-pairs as ‘vertex-candidates’(PCA distance: < 100um, distance between pair: <20um)

• Remove vertex candidates nearer than 200 um from IP.• Associate tracks passing near the vertex candidates• Treat them as ‘jet-core’

– Reconstructed vertex position is good but efficiency is currently low, need tuning

• New definition of ‘y’ was applied.• First clustering algorithm was implemented.

Current statusCurrent status


‘y’ distribution should be independent of jet energy.

New definition of ‘y’New definition of ‘y’


• bb/cc/uds in Jupiter, no ISRYY distributiondistribution


SampleSample


• Quite hopeful.• Reconstructed vertex position is ‘too’

matched to MC truth quark position• Should improve

– Vertex finder• Low efficiency – rescue 2-track vertex

– Clustering• Vertex information can be respected more

• Submitted to TIPP09

CommentsComments


• Sim/Reco session (approved??)• Centered on concept of new clustering• Some event samples can be shown• Improvements in vertex finding should

be realized next week.• Comparison between Durham and ours

will be prepared in ttbar events– ‘Goodness’ of jet clustering– Performance of flavor tagging (if

possible…)

Plan in LCWSPlan in LCWS


BackupBackup


• PFOを energy順に並べる– 見つかった Vertexは Energy orderにして先頭に insertする

•先頭は無条件に core–あと 2nd vertexがあるやつ…未実装

•見つかったやつとの y が 0.2以下は除外– 0.2は設定可能

•繰り返し• Core+子分で 1GeV以下は排除

– 1GeVは設定可能

CoreCore探し探し


• いったん core以外は忘れる ( 解放 )• PFOを energy orderにする• 各 PFOと coreとの y を計算し最小のものを採用

– 最適ではないか？• Jet数を変えていく

– 最初は coreを全部使う (10-30くらいある )– Coreを1 つ減らす ( 全部の場合を試す )

• それぞれで loglikelihoodの合計を計算• 最小のパターンを採用

– Coreを減らしていったときの最適 likelihoodの変化から true jet数を予測 ( どうやるか )

Fragment associationFragment association

taikan suehara, the 7th general meeting of the ilc physics working group, 2008/11/08 page 1 full...

Documents