hcal%pre)cd)1% november2 ,2017 · conclusionsfrom pythia%analyses% •...

HCAL Pre-‐CD-‐1 Conceptual Design

Review November 2nd, 2017

Conclusions from Pythia Analyses •  Response from different detector configura4ons does not depend on R

9/20/17 Rosi Reed -‐ HCal 2

Songkyo Lee

Conclusions from Pythia Analyses •  Response from different detector configura4ons does not depend on R or strongly on pT


(GeV/c)T

GenJet p20 30 40 50 60

µ /

σ

0.0

0.2

0.4

0.6

0.8

1.0SIMULATIONsPHENIX

|< 0.6η R=0.4 Tower Jets, |tanti-kSteel HCALINSteel (frame) HCALINAl HCALINAl (frame) HCALIN

(GeV/c)T

Truth Jet p20 30 40 50 60

µ

0.4

0.6

0.8

1.0

1.2PYTHIA8 @ 200 GeV

T / truth jet p

Treco jet p

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 200.020.040.060.08

0.10.120.140.160.18 SIMULATIONsPHENIX

(GeV/c) < 25Ttruth15 < Jet p

|< 0.6η R=0.4 Tower Jets, |tanti-kSteel HCALINSteel (frame) HCALINAl HCALINAl (frame) HCALIN

T / truth jet p

Treco jet p

0 0.20.4 0.6 0.8 1 1.21.4 1.6 1.8 200.020.040.060.08

0.10.120.140.160.18

0.20.220.24


T / truth jet p

Treco jet p

0 0.20.4 0.60.8 1 1.2 1.41.61.8 200.020.040.060.08

0.10.120.140.160.180.2

0.22


PYTHIA8 pp @ 200 GeV

Raghav Elayavalli

Conclusions from Pythia Analyses •  Response from different detector configura4ons does not depend on R or strongly on pT

•  Response slowly degrades with removal of interac4on lengths with ac4ve readout – There is no cliff that we fall off of – inclusive observables are rela4vely insensi4ve

– Unfolding will start to become difficult – Minimal fragmenta4on bias will increase

• Will have further study


Jet in Heavy Ion Collisions •  Large, fluctua4ng background in heavy-‐ion collisions complicates jet analyses – Combinatorial jets – Changes Jet Energy Scale (JES) – Changes Jet Energy Resolu4on (JER) –  Jet finding Efficiency

•  Solu4ons involve – Background subtrac4on – Selec4on of “hard” processes – Unfolding


Jet in Heavy Ion Collisions •  Large, fluctua4ng background in heavy-‐ion collisions complicates jet analyses – Combinatorial jets – Changes Jet Energy Scale (JES) – Changes Jet Energy ResoluKon (JER) –  Jet finding Efficiency

•  Solu4ons involve – Background subtracKon – Selec4on of “hard” processes – Unfolding


These are the key aspects that pertain specifically to the HCAL design

Background SubtracKon

•  sPHENIX MIE, subtrac4on rou4ne à arxiv:1203.1353

•  We are con4nuing to work on this à 2 features not yet implemented: – Refinement of exclusion region defini4on in first itera4on step

– Flow es4ma4on / modula4on

•  It is important to note that the background is determined for each calorimeter layer separately!

7 Rosi Reed -‐ sPHENIX Collabora4on Mee4ng -‐ June 2017


arxiv:1203.1353

For now using R=0.2, pTreco > 25 GeV jets before subtracKon as exclusion regions”

No v2 modulaKon yet

Background SubtracKon •  Algorithm adjusts the addi4on to the JES by the UE, and requires: – Creates 0.1x0.1-‐towerized version of CEMC – Crea4on of towers from all 3 calorimeter subsystems

– Default jet reconstruc4on •  Es4mates UE contribu4on to towers in η=0.1 rings – Separate for each layer –  Creates new UE-‐subtracted tower containers

•  Jet reconstruc4on has been modified to handle nega4ve-‐E tower inputs


Jet Response •  Background smears reconstructed momentum –  Increases JER –  Depends on R and pT

•  Can decrease reconstruc4on efficiency

truthTp/reco

Tp

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Cou

nts

0

50

100

150

200

250

300

350

400

sPHENIX G4 Sim=0-4 fmbHijing Au+Au

dijet events = 30-35 GeVtruth

TpR=0.4,

truthTp/reco

Tp

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

Cou

nts

0

100

200

300

400

500

600

700sPHENIX G4 Sim

dijet eventsp+p = 30-35 GeVtruth

TpR=0.4,

pp

AA �  Jet energy Scale (JES)

�  Increased by background �  Subtrac4on method necessary �  Unfolding requires well behaved Response Matrix

Erin Bossard

Erin Bossard

10

ReconstrucKon Efficiency •  Truth jets are associa4on with closest reco jet within ΔR < Ran4-‐kT –  If this fails, the jet is not reconstructed

(GeV)truthTp

20 25 30 35 40 45 50 55 60

Rec

onst

ruct

ion

Effic

ienc

y

0

0.2

0.4

0.6

0.8

1

1.2

sPHENIX G4 Sim dijet eventsp+p

R=0.4R=0.3R=0.2

(GeV)truthTp

20 25 30 35 40 45 50 55 60

Rec

onst

ruct

ion

Effic

ienc

y

0

0.2

0.4

0.6

0.8

1

1.2

sPHENIX G4 Sim=0-4fmbHijing Au+Au

dijet eventsR=0.4R=0.3R=0.2

�  Larger jets and lower pT jets can be obscured by the fluctua4ng background

pp

AA

Erin Bossard

Erin Bossard

Rosi Reed -‐ sPHENIX Collabora4on Mee4ng -‐ June 2017

11

Jet Energy Scale •  For high momentum jets, the JES is

similar in pp and AA •  Defect in the JES at low-‐pT in AA à

suspect cause is the crude exclusion seed selec4on –  Low-‐pT (hard scakering) jets are

included in the es4mate of the background à over subtrac4ng

(GeV)truthTp

20 25 30 35 40 45 50 55 60

Jet E

nerg

y Sc

ale

00.10.20.30.40.50.60.70.80.9

1

sPHENIX G4 Sim dijet eventsp+p

R=0.4R=0.3R=0.2

(GeV)truthTp

20 25 30 35 40 45 50 55 60

Jet E

nerg

y Sc

ale

00.10.20.30.40.50.60.70.80.9

1

sPHENIX G4 Sim=0-4 fmbHijing Au+Au

dijet eventsR=0.4

R=0.3

R=0.2

pp

AA

�  Mean value of: pT,reco / pT,truth �  Determined by Gaussian fit

�  Uses default simula4ons �  Linear tower calibra4on à

assumed sampling frac4ons �  no addi4onal jet-‐level

calibra4on

Erin Bossard

Erin Bossard


12

Jet Energy ResoluKon •  In HI collisions JER increases with increasing R

•  Rapidly increases for pT,truth ~ average background

(GeV)truthTp

20 25 30 35 40 45 50 55 60

Jet E

nerg

y R

esol

utio

n

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35sPHENIX G4 Sim

dijet eventsp+pR=0.4R=0.3R=0.2

(GeV)truthTp

20 25 30 35 40 45 50 55 60

Jet E

nerg

y R

esol

utio

n

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35 sPHENIX G4 Sim=0-4 fmbHijing Au+Au

dijet eventsR=0.4R=0.3R=0.2

Erin Bossard

Erin Bossard

pp

AA

�  Width of pT,reco / pT,truth distribu4on �  Determined by Gaussian fit


13

T / truth jet p

Treco jet p

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 200.010.020.030.040.050.060.070.08 SIMULATIONsPHENIX


|< 0.6η R=0.4 Tower Jets, |tanti-k

Steel HCALINAl (frame) HCALIN

T / truth jet p

Treco jet p

0 0.20.4 0.6 0.8 1 1.21.4 1.6 1.8 200.020.040.060.08

0.10.120.14


T / truth jet p

Treco jet p

0 0.20.4 0.60.8 1 1.2 1.41.61.8 20

0.02

0.04

0.06

0.08

0.1 (GeV/c) < 35

Ttruth25 < Jet p

PYTHIA8 pp @ 200 GeV

Embedding into HIJING (R =0.4 jets)

•  Comparison of default and Al no readout – Versus pT,jet –  Inclusive


Raghav Kunnawalkam Elayavalli

Embedding into HIJING

•  Inclusive jet observables in HI collisions follow the same trend as in PYTHIA studies

•  For high pT jets, background has minimal effect

•  Addi4onal smearing due to UE added in quadrature with smearing from reduced IHCAL could reduce analysis window


(GeV/c)T

GenJet p20 30 40 50 60

µ /

σ

0.0

0.2

0.4

0.6

0.8

1.0SIMULATIONsPHENIX

|< 0.6η R=0.4 Tower Jets, |tanti-k

Steel HCALINAl (frame) HCALIN

(GeV/c)T

Truth Jet p20 30 40 50 60

µ

0.4

0.6

0.8

1.0

1.2PYTHIA8+HIJING 0-4fm @ 200 GeV

Outlook to CDR •  Current sPHENIX design similarity to LHC experiments has allowed reasonable knowledge transfer

•  There do not seem to be any gaping holes in the methods needed for compensa4ng for the background in HI collisions with respect to the design with any IHCAL configura4on discussed so far

•  However, a fluctua4ng background with a fluctua4ng detector response can be complicated.


Outlook to CDR •  New performance plots with respect to the development that has taken place – Regenerate plots of JES/JER vs R and pT with the latest geometry

•  Incorpora4on of JEWEL (a jet quenching model) will allow us to look at quenched versus non-‐quenched jets – Small difference between quark and gluon jets indicates this will not be a dominate effect

•  Benchmark background subtrac4on methods versus different calorimeter systems


Back Up


HCAL Descoping – Jet perspecKve •  What impact does each inner HCAL configura4on have on the JES and JER? – How much can we correct in unfolding?

•  Do the configura4ons increase the bias with respect to fragmenta4on? – Quark vs Gluon jets – High z jets – Dis4nguishing modifica4on of the fragmenta4on

•  How does this effect the background removal techniques in HI collisions?

•  We want to push the kinema4c reach to create jet probes which overlap with LHC kinema4cs


Inclusive Jet ResoluKon vs R

Different HCAL configura4ons have a minimal dependence on R for inclusive jets


Songkyo Lee

•  d


Songkyo Lee

We can study effects/configura4ons with one or two R choices

Inclusive Jet ResoluKon vs R

FragmentaKon

•  Inclusive jet performance shows minimal difference however, the impact on high-‐z or other “special” jets can be substan4al – Modifica4on of the fragmenta4on func4on is a key sPHENIX measure

– The ability to dis4nguish jet quenching effects from quark vs gluon will be important

•  sPHENIX is a next genera4on detector and should be able to make next genera4on jet measurements


Highest z Charged Hadron (π, K, p)


Default

JES à 1.29 corr JER à 13.0%

Al IHCal

More punch through?

Al IHCal NO READOUT

100%/√(50) = 14.1%

Red -‐ all jets Black -‐ 0.2 < z < 0.4 Blue -‐ z > 0.6

JES à 1.29 corr JER à 13.1%

JES à 1.36 corr JER à 15.1% Jamie

Nagle

Correc4on of 1.29 indicates a Calo scale issue

emfrac0 0.2 0.4 0.6 0.8 1

reco

10

20

30

40

50

60

70

reco:emfrac {truth>50.0}tp

Entries 13082Mean 0.2838Mean y 41.42Std Dev 0.2153Std Dev y 6.095

JES/EM fracKon (SS310) JES Correc4on of 1.29 •  CEMC energy scale set for EM

showers! For a jet with a high z π0 •  Ereco ~ Etruth As EM frac4on decreases, Ereco decreases •  IHCal inves4ga4ons as a

func4on of leading z hadron à walking up and down this curve

High peak in CEMC energy frac4on à “miscalculated” by using the EM energy scale to calculate the CEMC tower energy •  Suppresses the correla4on with

reco jet energy


Jamie Nagle

John Lajoie

Electron IdenKficaKon

•  An4protonsà main source of fake electrons at low pT

•  The worse case: low pT an4-‐protons steel frame example – Hadron rejec4on for 90% electron ID –  Just E/p cut: 5.11+/-‐ 0.26 –  2D cut: 5.05 +/-‐ 0.25


Sasha Lebedev

Set up – Jet Shape Observables •  Analysis code à developed to study jets with forward instrumenta4on:

•  Same PYTHIA 8 jets (R = 0.4, pT,jet > 50 GeV) •  3 Different reconstructed jets collec4ons –  Primary Par4cle Jets -‐ No muons, neutrinos –  Track Jets -‐ Tracks require ndf>60, χ2/ndf<1.5, DCA2D<0.1cm

–  Calorimeter Tower Jets -‐ Require tower energy > 100 MeV •  3 condi4ons – SS310, Al, and Al without readout •  “Matched” jets require track and tower jets match the primary jet within ΔR<0.3


Jet Shape in HCAL


Matched tower jets à Etower with respect to the jet axis

Inner HCAL Outer HCAL

Less energy in the inner HCAL and more in the outer HCAL with Al •  Look at normalized distribu4ons to see whether shape changes

John Lajoie

Jet Shape in HCAL


Matched tower jets à Etower with respect to the jet axis •  Normalize by area and integral to look at the shape

Small difference in the shape in the inner HCAL between SS and Al à A likle broader, at a scale < R


John Lajoie

Combined Jets •  Combine the tower and track jets to use the best informa4on from both –  Similar to par4cle flow à uses truth informa4on instead of a real clustering algorithm

•  Point each track jet cons4tuent into the calorimeter, sum up the energy it contributed to the tower jet in each calorimeter segment

•  Combine the tower and track jet informa4on: –  Expected E/p resolu4on calculated using measured quan44es –  If the ptrack resolu4on is beker than Etower resolu4on à add track to jet

–  If the Etower resolu4on is beker, rescale ptrack to Etower, add track to jet

–  Remaining Etower à“neutral energy” à add to jet 9/20/17 Rosi Reed -‐ HCal 29


Tower JES/JER – degrada4on is visible but small compared to the SS310 baseline:

Tower Jets

Combined Jets

JES ResoluKon

SS310 0.744 9.3%

Al 0.724 9.8%

Al (no IHCAL)

0.691 10.7%

<neutral frac4on> >~0.33, due to CEMC energy scale?

Jet Energy Scale

John Lajoie


Use the “black hole” to add back E within R=0.4 of the combined jet axis

Shape of the low side largely remains the same? •  Low side tail not due to leakage? •  Imperfec4ons in combina4on algorithm?

•  Jet axis resolu4on?

AccounKng for Leakage

John Lajoie

Default

Al IHCal

Al IHCal NO READOUT

Jet Flavor/FragmentaKon Bias I


light quarks

gluons

Light Quarks Gluons JES Resolu4on JES Resolu4on

SS310 0.749 9.4% 0.722 8.0%

Al 0.727 9.7% 0.702 9.0%

Al (no IHCAL)

0.694 10.7% 0.671 10.0%

•  Separate response for light quark (u+d) and gluon jets •  Require Eparton >50% Ejet

•  JES is a few % lower for gluon jets than for quark jets à JER is a bit beker

•  The difference between q/g jets is ~ to the difference between SS310 and Al.

Tower Jets John Lajoie

Embedding into HIJING

•  Chris has produced HIJING Files for embedding: –  0-‐4 fm: /sphenix/data/data02/rescope-‐2017-‐09-‐08/inner_hcal_al/sHijing/fm_0-‐4

– Other b ranges (0-‐12, 5-‐9, 9-‐11): /sphenix/data/data02/rescope-‐2017-‐09-‐08/inner_hcal_al/sHijing/fm_<b range>/

•  JER is degraded due to fluctua4ng background •  JES correc4on via background subtrac4on is necessary – Atlas itera4ve rou4ne subtracts layer by layer à How do descoping op4ons effect this?


Conclusions •  The differences between SS, Al and Al without read-‐out – Not large for inclusive jet observables

•  Including first look at embedded jets – Does not have much R dependence – Can maker for fragmenta4on/flavor analyses

•  JES difference for q/g ~ SS310/Al difference •  Need to to properly set the CEMC energy scale •  Combined jets show a proper PF algo might get the JES close to 1 – Leakage is s4ll an issue


Concerns/Thoughts •  We should combine 85% EMCal + descoped HCAL – Could simply cut on |ηjet|<0.85-‐R – We have an increase in sta4s4cs due to addi4onal years, but will lose on the acceptance

•  Need clear communica4on over default and other poten4al configura4ons – Understanding Calo/Tracking even for the default

•  Person-‐power as always •  Jet-‐trees will be produced for pp/AA – Help with person-‐power – Keep details consistent


Back-‐up


Some thoughts

•  Simula4ons have become increasingly realis4c – Great progress, and indica4ons are that simpler simula4ons were reasonable

•  However, addi4onal development both in the simula4ons and our understanding is vital – Even without descoping – We also do not know how more realis4c simula4ons will affect the response 9/20/17 Rosi Reed -‐ HCal 37

JES/EM fracKon (SS310) Reco jet energy vs. reco jet EM fracKon


The CEMC energy frac4on is peaked highà “miscalculated” by using the EM energy scale to calculate the CEMC tower energy. •  Suppresses the correla4on with reco jet energy.

John Lajoie

Jet Response vs z (π+,π-‐)

• 

•  Mean +/-‐ RMS •  Shiz in response visible here, some z dependence

•  But this not the full story


z =pTruthT ,particle

pTruthT , jet

Highest z Charged Hadron (π, K, p) R = 0.4

9/20/17 Rosi Reed -‐ HCal 40 Inclusive jets, jets with 0.2 < z < 0.4, and jets with z > 0.6.

Jamie Nagle

Combined Jets •  Combine the tower and track jets to use the best informa4on from both:

–  Analysis code includes bookeeping of par4cle energy contribu4on to calorimeter tower energy •  This uses truth informa4on (think of it as perfect clustering/cluster spli{ng)

–  Point each track jet cons4tuent into the calorimeter, sum up the energy it contributed to the tower jet in each calorimeter segment

–  Combine the tower and track jet informa4on: •  Expected energy/momentum resolu4on calculated using measured quan44es:

–  Track momentum resolu4on δpT /pT = 0.005 + (0.001*pT) –  Sort EM/hadronic par4cles by CEMC tower energy frac4on > 0.9 –  CEMC resolu4on 0.12/sqrt(E) for EM par4cles,

CEMC+HCAL resolu4on 0.15 + 0.7/sqrt(E) for hadrons •  If the track momentum resolu4on is beker than the tower resolu4on, add the track to

the combined jet •  If the calorimeter energy resolu4on is beker, rescale the track total momentum to match

the tower energy. –  Improves energy resolu4on but keeps improved poin4ng resolu4on of tracking (eta,phi)

•  Remaining tower energy azer all par4cles is “neutral energy” and is added to the combined jet.

–  This is similar to a par4cle flow algorithm, with the excep4on that it uses truth informa4on instead of a real clustering/cluster spli{ng algorithm

–  Gives an idea of what is the best you could possibly do, or how much informa4on is in principle available for you to take advantage of •  Leakage out the back is a loss of informa4on.


Jet Flavor/FragmentaKon Bias II


light quarks

gluons

Combined jets show a slightly different response for quarks and gluons (missing soz hadron tracks?)

Combined Jets

Comparison with “old” results

•  JES has shized downwards with new default – Old: 0.83+/-‐0.09 – New: 0.78 +/-‐ 0.10


Jet Response •  For high pT jets, it seem the default and the Al are not so different – On average

•  Looks similar to Ragav’s trend

•  Compare: – Default: 0.78 +/-‐ 0.10

– Al: 0.78+/-‐0.10 – Al no readout: 7.4 +/-‐0.11


Jet Shape II


For matched tower jets, look at the distribu4on of tower energies with respect to the jet axis. Normalize the Al distribu4on to the SS310 integral to look at the shape.


There is a small difference in shape for the inner HCAL, energy distribu4on pushed out a likle, but at a scale that is smaller than the jet reconstruc4on radius.

Larger Jet Radius


Combined jets, re-‐run with R=0.8 for comparison – likle change.

Combined Combined + BH

With the R=0.8 the BH seems to pick up a likle more energy.

η/φ ResoluKon


Nothing special here, this just shows the improved η/φ resolu4on of the combined just by using the tracking direc4on and the calorimeter energy.

hcal%pre)cd)1% november2 ,2017 · conclusionsfrom pythia%analyses% •...

Documents