Dijet Production in Polarized Proton-Proton Collisions at 200 GeV

Matthew Walker

STAR

Outline
Motivation Experimental Overview Introduction to Dijets in STAR Cross Section Analysis Asymmetry Analysis Conclusions

STAR

Matthew Walker

Theoretical Motivation
1 1 = + Lq + G + Lg Where does the spin of the 2 2

proton come from?

An early calculation suggested quark spins contribute 60%

STAR

Matthew Walker

Theoretical Motivation
1 1 = + Lq + G + Lg 2 2

Early data suggestive of lower quark contribution.

Ashman et al. (EMC) Phys Lett. B 206:364 (1988)

STAR

Matthew Walker

Theoretical Motivation
More precise measurements have found is about 30%
1 1 = + Lq + G + Lg 2 2

Filippone and Ji, arXiv:hep-ph/0101224v1 (2001)

STAR

x
5

Matthew Walker

he 2 = 1 range, modern unpolarized global analysis it issituation. However for u, they of 2 = 1 between s of uncertainty, in suggesting something close to the ideal customary to consider instead only iation in variation in order of a 2%. This is a very of the example of how the 2 = 1 does not 2 of the 2 as conservative estimates good range of uncertainty. nd a 5% an unaccounted source of uncertainty: dependence of between the available sets of u and d resemble a parabola xpected in the ideal framework, the the dierences 2 on the rst moments of fragmentation

es 3a and 3b). The KKP curves are shifted upward almost six units relative to those from KRE, due to the nce in 2 of their respective best ts. Although this means that the overall goodness of KKP t is poorer than d and u seem to be more tightly constrained. The estimates for d computed with the respective best ts se and within the 2 = 1 range, suggesting something close to the ideal situation. However for u, they only p allowing a variation in 2 of the order of a 2%. This is a very good example of how the 2 = 1 does not q 0.4 g o apply due to an unaccounted source of uncertainty: the dierences between the available sets of fragmentation ns.
x!uv x!g

Theoretical Motivation
0.2

1 1 = + L + G + L 2 2

Look to the gluon!

x(!u+!u) x!d G(Q2 ) =

0
x!g

0.4

1 0

g(x, Q2 )dx

x!uv

-0.2 0.06
x!s

0.2

Unfortunately, x!d current DIS data x(!d+!d) x!d cannot constrain it very well.
v

KRE (NLO) 0.04 KKP (NLO) unpolarized 2 KRE "min +1 0.02 2 KRE "min +2%
10-2

-0.2
1

x!u
STAR

x!d

x!s

-1 100

0.06 0.04 0.02

D. de Florian et al., Phys. Rev. D71, 094018 (2005).

KRE (NLO) -0.02 KKP (NLO) unpolarized 2 KRE "min +1 -0.04

2

Matthew Walker

Theoretical Motivation
ALL d f1 f2 h aLL = = d f 1 f 2 h
aLL
1

qg qg

0.75 0.5 0.25

gg gg
qq qq

f1

f2

0 -0.25 -0.5 -0.75 -1 -1

qq qq

gg q q
-0.8 -0.6 -0.4 -0.2 -0 0.2 0.4 0.6 0.8 1

cos*

STAR

Matthew Walker

!u

Inclusive Results
x!d

0.04 0.02

DSSV DNS KRE DNS KKP

Inclusive data from RHIC has been included into a global analysis.
0 -0.02

DSSV !" =1 DSSV !" =2%

2

With RHIC data

-0.04

!s

x!g

0.3

0.2

0.1

Substantial improvement for 0.05 < x < 0.2, but large uncertainties at low x

-0.1

GRSV maxg GRSV ming

10
-2

-0.2

10

-1

10

-2

x
STAR

x x

10

-1

1
D. de Florian et al., Phys. Rev. Lett. 101 (2008) 072001

Matthew Walker

Why Dijets?
Jet

Reconstructing multiple physics objects (di-jets, photon/jet) provides information about initial parton kinematics

Jet

STAR

Matthew Walker

Outline
Motivation Experimental Overview Introduction to Dijets in STAR Cross Section Analysis Asymmetry Analysis Conclusions

STAR

Matthew Walker

10

Experimental Setup

RHIC produces polarized proton beams up to 250 GeV in energy and polarization up to 60%

STAR

Matthew Walker

11

STAR Detector
!=-1 BEMC !=0 !=1

Blue

BBC

Yellow

TPC

West East

Tai Sakuma

Not shown: Zero-degree calorimeters, time-of-ight, polarimeters

STAR

Matthew Walker

12

Outline
Motivation Experimental Overview Introduction to Dijets in STAR Cross Section Analysis Asymmetry Analysis Conclusions

STAR

Matthew Walker

13

Jet Terminology
Jet

detector

particle

Detector Effects

STAR

Matthew Walker

parton

Parton Branching, Hadronization, Underlying Event

14

Finding Jets

Jet Patch Trigger:

1x1 in x patch of towers in the BEMC (400 towers)

Midpoint Cone Algorithm with SplitMerge Cone Radius: 0.7

STAR

Matthew Walker

15

Dijet Reconstruction

Dijet Cuts

Jet || < 0.8 Jet Detector || < 0.7 Dijet |3-4| < 1.0 Back-to-back in Asymmetric pT cut

max(pT3,pT4) > 10 min(pT3,pT4) > 7

STAR

Matthew Walker

16

Dijet Kinematics
Jet 3
1 x1 = (pT 3 e3 + pT 4 e4 ) s 1 x2 = (pT 3 e3 + pT 4 e4 ) s

M=

x1 x2 s x1 3 + 4 = ln x2
Jet 4

STAR

Matthew Walker

17

Outline
Motivation Experimental Overview Introduction to Dijets in STAR Cross Section Analysis Asymmetry Analysis Conclusions

STAR

Matthew Walker

18

2006 Cross Section

Y Y

Correction factors:

C = Cvert Cdet Cvert: Vertex Acceptance Cdet: Correction for detector effects calculated from simulation
19

STAR

Matthew Walker

Data/Simulation Run 6

M=

x1 x2 s

Invariant mass distribution sensitive to trigger and detector parameter agreement

STAR

Matthew Walker

20

Data/Simulation Run 6

x1 3 + 4 = ln x2

Pseudorapidity distribution sensitive to vertex distribution and trigger effects.

STAR

Matthew Walker

21

2006 Cross Section

105 pp @ 200 GeV Cone Radius = 0.7 max(pT) > 10 GeV, min(pT) > 7 GeV 104 -0.8 < ! < 0.8, |!!| < 1.0 |!!| > 2.0 d3!/dMd!3d!4 [pb/GeV]
STAR

Jet

Preliminary

103

! Ldt = 5.39pb!1 !

d 3!
102

g
10 STAR Run-6
Systematic Uncertainty

Theory
NLO pQCD + CTEQ6M

1 30

Had. and UE. Corrections

q
70 80 90

40

50

60 Mjj [GeV]

STAR

Matthew Walker

parton

dMd!3d!4

particle

detector

Dijet Cross Section

22

Outline
Motivation Experimental Overview Introduction to Dijets in STAR Cross Section Analysis Asymmetry Analysis Conclusions

STAR

Matthew Walker

23

Asymmetry Measurement
ALL d 1 N RN = = ++ + RN + d PB PY N

++

Asymmetry formula:

N++: like sign dijet yields N+-: unlike sign dijet yields

2009 Data: 10.3 pb-1 analyzed from Run 9 Signicant increase in data size over previous years and small increase in polarization

R: relative luminosity P: polarization

STAR

Matthew Walker

24

Asymmetry Formula
ALL,j =
k k

jk ( jk (

++ f ill PB,f ill PY,f ill (Nf ill,k ++ 2 2 f ill PB,f ill PY,f ill (Nf ill,k

+ Rf ill Nf ill,k )) + Rf ill Nf ill,k ))

The value of ALL in a bin j is given by the above formula

jk are the matrix elements for the unfolding

There are different polarizations and relative luminosities for each RHIC ll.

STAR

Matthew Walker

25

Why do we need unfolding?

Purpose of unfolding - undo smearing caused by Detector effects, eg: Double counting electrons Hadronic response Lost tracks Energy Leakage Falling spectrum Used a matrix unfolding scheme based on G. DAgostini, NIM A 362 (1995), p. 487.

STAR

Matthew Walker

26

Why do we need unfolding?

Consider a specic bin of true (or unfolded) invariant mass. The contributions to this bin are signicant from multiple detector mass bins. The red bin has the same bin boundaries; about 50% of the yield comes from other bins Signicant contributions come from off-diagonal elements of the unfolding matrix

Contributions to corrected bin 48.83 < Mparticle < 64.15 Unfolded Yield 160 140 120 100 80 60 40 20 0 20 30 40 50 60 70 80 90 100 110 Reconstructed Invariant Mass (GeV/c2)

STAR

Matthew Walker

27

2009 Simulation

New simulation needed due to changes in detector Previous simulation effectively integrated 5.3 x 10-4 pb-1 Goal for new simulation: 1 pb-1

Use ltering to reduce CPU usage by factor of 300 and disk usage by factor of 500 Corresponds to about 1 year with a standard allocation on STAR resources

Solution: use cloud computing resources

STAR

Matthew Walker

28

Outline

Motivation Experimental Overview Introduction to Dijets in STAR Cross Section Analysis Asymmetry Analysis

Using cloud computing to simulate STAR (aside)

Conclusions

STAR

Matthew Walker

29

What is cloud computing?

Virtualization

Applications OS

Applications OS Virtualizing Software

Applications OS

Hardware Hardware

Virtual Machine

STAR

Matthew Walker

30

What is cloud computing?

Computing is treated like electricity.

There is a qualitative difference from grid computing.

STAR

Matthew Walker

31

STAR Cloud Computing Projects

date Facility 2009, March Amazon EC2 2009, November Amazon EC2 GLOW Madison 2010, February 430 1 130 Uni Wisconsin Clemson Uni, 2010, July 1000 1 17,000 SC Magellan 2011, February 20 6 or 7 600+ NERSC # of # jobs/ CPU calendar output VMs VM days days (TB) 100 1 500 5 0.3 10 1 or 2 1 1 0.01 0.6 20 20+ 0.1 7 1

GLOW NERSC Amazon EC2 STAR

Clemson

STAR

Matthew Walker

32

Clemson Palmetto Conguration

Job Manager Client Worker nodes

Node Manager

STAR

Matthew Walker

Virtualized
33

Cloud Results

Over 12 billion events generated by PYTHIA, ltered to allow only 36 million to undergo detector simulation (GEANT3), and 10 million through full reconstruction
N Machines

Took over 400,000 CPU hours Expansion of 25% of STAR computing resources Would have taken over 1 year on normal allocation Largest physics simulation on cloud, largest STAR simulation
Matthew Walker

1400 1200 1000 800 600 400 200 0

Available Machines Working Machines Idle Machines

Jul17

Jul24

Jul31 Date

STAR

34

Outline
Motivation Experimental Overview Introduction to Dijets in STAR Cross Section Analysis Asymmetry Analysis Conclusions

STAR

Matthew Walker

35

Data/Simulation Run 9
STAR Run 9 Data Preliminary
Normalized Yields
10
6

Data Simulation
10
5

p+p Jet + Jet + X s = 200 GeV

10

10 104

R cone = 0.7 -0.8 < < 0.8 | | < 1.0

10
4

10

| | > 2.0
0.2 0.4 0.6 0.8 1 0 1 0.8 0.6 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 0.2 0.4 0.6 0.8 1
34

20

30

40

50

60

70

80

90 100

-1 -0.8 -0.6 -0.4 -0.2 0 1 0.8 0.6 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8

0.1

0.2

0.3

0.4

0.5

0.6

(Data-Simu)/Simulation

1 0.8 0.6 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 20 30 40 50 60 70 80 90 100

2

STAR

Preliminary

-1 -1 -0.8 -0.6 -0.4 -0.2 0

-1 0

0.1

0.2

0.3

0.4

0.5

0.6

Invariant Mass (GeV/c )

cos(*)

STAR

Matthew Walker

36

Uncertainties
ALL,j =
k k

jk ( jk (

++ f ill PB,f ill PY,f ill (Nf ill,k ++ 2 2 f ill PB,f ill PY,f ill (Nf ill,k

+ Rf ill Nf ill,k )) + Rf ill Nf ill,k ))

Systematic

Non-longitudinal effects: 0.025 x ALL Polarization: 8.8% relative Relative Luminosity: R = 1x10-3 Jet energy scale: Vary inputs from different detectors Theory Scenario dependent trigger efciencies: Test unfolding under different scenarios
37

STAR

Matthew Walker

Run 9 Asymmetry
East - East and West - West Barrel
A LL A LL 0.08 0.06
MC GS-C(pdf set NLO) 2009 STAR Data Systematic Uncertainties

East Barrel - West Barrel

A LL 0.08 0.06 Scale uncertainty GRSV std DSSV 0.08 0.06

Full Acceptance

p + p jet + jet + X

s = 200 GeV

0.04 0.02 0 -0.02 20

0.04 0.02 0

0.04 0.02 0 -0.02 30 40 50 60 70 80 20 30

8.8% Polarization Scale Uncertainty Not Shown

30 40 50 60 70 80

-0.02 20

STAR
40

Preliminary
50 60 70 80

M [GeV/c2]

M [GeV/c 2]

M [GeV/c2]

East

West

East

West

+
=-1 =0 =1 =-1 =0 =1

STAR

Matthew Walker

38

Kinematic Sensitivity
East Barrel - East Barrel
10 10 10 10
5

East Barrel - West Barrel

10 10 10 10
5

10 1 10
-1

10 1 10
-1 -1

x1: 20.0 < M < 30.0 x2: 20.0 < M < 30.0 x1: 70.0 < M < 80.0 x2: 70.0 < M < 80.0

10 -2 10

-2

10

10 -2 10

-2

10

-1

10

-1

x 10

STAR

Preliminary

-1

20

p + p jet + jet + X
30 40 50

x1 x2
20 30 40 50 60 70 80 90 100 2 Invariant Mass (GeV/c )

s = 200 GeV

60 70 80 90 100 2 Invariant Mass (GeV/c )

STAR

Matthew Walker

39

Conclusions

pQCD has been validated as a framework to understand the spin of the proton, based on the measurement of the dijet cross section at STAR. It is now possible to place constraints on the shape of g(x) for the rst time using the measurement of ALL with dijets from 2009 data, which show the rst non-zero ALL. Over the x-range were probing, the gluon spin contribution to the proton spin is small Stay tuned for the impact of this data on a global analysis and measurements probing lower x....

STAR

Matthew Walker

40

Backup

STAR

Matthew Walker

41

Inclusive jets
ALL 1 N RN = ++ + RN + PB PY N
++ +

STAR

Matthew Walker

42

mple of how the 2 = 1 does not ge of uncertainty. the available sets of u and d resemble a parabola x!u he rst moments of fragmentation 0.04 six units relative to those from KRE, due to the hat the overall goodness of KKP t is poorer than 0.02 tes for d computed with the respective best ts e to the ideal situation. However for u, they only 0 very good example of how the 2 = 1 does not 0.4 erences between the available sets of fragmentation
-0.02

Impact
x!d
2 2

0.04

0.02

Inclusive data from RHIC has been included into a global analysis. DSSV
0 -0.02

DNS KRE 0.2 -0.04 Without RHIC data DNS KKP

DSSV !" =1 DSSV !" =2%

With RHIC data

-0.04

0.04

x!s 0
-0.2

0.4

x!g

0.3

x!g
0.02

0.2

0.2
0.1

0.06 0
0

KRE (NLO) 0.04 -0.02 KKP (NLO) unpolarized 2 KRE "min -0.04 +1 0.02 2 KRE "min +2%
10-2

Substantial improvement for 0.05 < x < 0.2, but large uncertainties at low x

-0.1

-0.2
-1 1 0.06 10

GRSV maxg GRSV ming

10
-2

-0.2

x!s

0 10 10-1 -2

-0.02 et al., Phys. Rev. D71, 094018 (2005). D. de Florian 0.04

KKP (NLO) unpolarized 2 KRE "min +1 -0.04 0.02 2 KRE "min +2%

x KRE (NLO)

x x

10

-1

1
D. de Florian et al., Phys. Rev. Lett. 101 (2008) 072001

STAR

Matthew Walker

43

2006 Cross Section

Systematic Uncertainty

1.0

Theoretical Uncertainty

! Ldt = 5.39 pb!1 !

(Data - Theory) / Theory -0.5 0.0 0.5

Data-theory Comparison
of Dijet Cross Section pp @ 200 GeV Cone Radius = 0.7 max(pT) > 10 GeV, min(pT) > 7 GeV -0.8 < ! < 0.8, |!!| < 1.0, |!!| > 2.0

STAR Run-6
Theory: CTEQ6M NLO pQCD Had. UE. Corrections

-1.0

30

40

50

60 Mjj [GeV]

70

80

90

STAR

Matthew Walker

44

2006 Cross Section

Systematic Uncertainties:

Luminosity: 7.6 % normalization uncertainty Jet Energy Scale: 20-50% Pile-up: 1% Timebin acceptance: 3%

STAR

Matthew Walker

45

2006 Asymmetry
0.08

Run 6 Longitudinal double helicity asymmetry Systematic uncertainties show effects on trigger efciency from different theory scenarios Scale uncertainty (8.3%) from polarization uncertainty not shown
ALL

0.06

0.04

Dijet ALL pp @ 200 GeV Cone Radius = 0.7 max(pT) > 10 GeV min(pT) > 7 GeV -0.8 < ! < 0.8, |!!| < 1.0 |!!| > 2.0

Data Run-6 Sys. Uncertainty

STAR Preliminary

0.02

0.00
GRSV STD DSSV GRSV !g = 0 GRSV !g = ! g

-0.02

! Ldt = 5.39pb!1 !
50 Mjj [GeV] 60 70 80

30

40

STAR

Matthew Walker

46

Correlation Measurements

1 x1 = (pT 3 e3 + pT 4 e4 ) s 1 x2 = (pT 3 e3 + pT 4 e4 ) s

M=

x1 x2 s x1 3 + 4 = ln x2

STAR Collaboration PRL 100 (2008) 232003

STAR

Matthew Walker

47

Relative Luminosity

Compare the calculations of R from two different detectors (ZDC and BBC) Check unphysical asymmetries

1 (N ++ + N + ) R(N + + N ) AL,Y ellow = PY (N ++ + N + ) + R(N + + N ) 1 (N ++ + N + ) R(N + + N ) AL,Blue = PB (N ++ + N + ) + R(N + + N ) 1 N ++ RN A++ = PB PY N ++ + RN 1 N + RN + A+ = PB PY N + + RN +

STAR

Matthew Walker

48

False Asymmetries
Raw A LL
0.1

AYellow
0.1

ABlue
0.1

0.05

0.05

0.05

-0.05

-0.05

-0.05

-0.1 20 30 40 50 60 70 80 Invariant Mass (GeV/c 2 )

-0.1 20 30 40 50 60 70 80 Invariant Mass (GeV/c 2 )

-0.1 20 30 40 50 60 70 80 Invariant Mass (GeV/c 2 )

A Like-sign
0.1

A Unlike-sign
0.1

Consistent with zero

0.05

0.05

-0.05

-0.05

-0.1 20 30 40 50 60 70 80 Invariant Mass (GeV/c 2 )

-0.1 20 30 40 50 60 70 80 Invariant Mass (GeV/c 2 )

STAR

Matthew Walker

49

Scenario Dependent Efciencies

ALL,j =
k( k(
A LL,pol - ALL,unpol
0.04

++ f ill PB,f ill PY,f ill (jk Nf ill,k ++ 2 2 f ill PB,f ill PY,f ill (jk Nf ill,k
DSSV

+ Rf ill jk Nf ill,k )) + Rf ill jk Nf ill,k ))

0.03

GRSV STD GRSV Zero

0.02

0.01

-0.01

-0.02

20

30

40

50

60 70 80 Invariant Mass (GeV/c2)

STAR

Matthew Walker

50

Jet Energy Scale Uncertainty

Contributions to uncertainty come from the two detectors

Neutral Energy Uncertainty - 2.1 %

BEMC Scale Uncertainty - 1.9 % BEMC Efciency Uncertainty - 1 % Track Scale Uncertainty - 2% Track Finding Efciency Uncertainty - 5%

Charged Energy Uncertainty - 5.4 %

STAR

Matthew Walker

51

BEMC Energy Scale

Data
Physics Object
EM Shower Physics

Simulation
Physics Object
GEANT Shower

Energy in Detector
Sampling, Optics, Electronics

Energy in Detector
C1

ADCs
C2 C2

ADCs

Reconstructed Energy

Reconstructed Energy

STAR

Matthew Walker

52

Jet Energy Scale Uncertainty

Reconstruction Trigger Corrections
High; High

Unfolding Other Efciencies

High; High High; Nominal High; Low

High

High; Nominal High; Low

BEMC Max

Data

Nominal; High

Nominal; High Nominal Nominal; Low

Nominal

Nominal Nominal; Low

Yield

Low; High

Low; High Low; Nominal Low; Low

Low

Low; Nominal Low; Low

BEMC Min

STAR

Matthew Walker

53

Dijet Run 9 Projected

STAR

Matthew Walker

54

UNFOLDING
Unfolded Yield

Consider the true bin with 49 < M < 64 The spectrum at right represents the contributions to this true bin from each of the reconstructed mass bins The red bin is the contribution from the same bin in reconstructed mass The contributions from blue bins is ~50%

Contributions to corrected bin 48.83 < Mparticle < 64.15 160 140 120 100 80 60 40 20 0 20 30 40 50 60 70 80 90 100 110 Reconstructed Invariant Mass (GeV/c2)

There are important off diagonal components that must be considered

55

7 October 2009

Collaboration Meeting

UNFOLDING
Contributions to corrected bin 20.00 < Mparticle < 24.25 Unfolded Yield 1200 1000

Contributions to corrected bin 24.25 < Mparticle < 30.01

Unfolded Yield

Contributions to corrected bin 30.01 < Mparticle < 37.90

Unfolded Yield

1600 1400 1200 1000 800 600

1200 1000 800 600 400

800 600

400 200 0 20

400 200
30 40 50 60 70 80 90 100 110 Reconstructed Invariant Mass (GeV/c2)

200 0 20

Contributions to corrected bin 37.90 < Mparticle < 48.83 Unfolded Yield 600

60 70 80 90 100 110 Reconstructed Invariant Mass (GeV/c2) Contributions to corrected bin 48.83 < Mparticle < 64.15

0 20

30

40

50

60 70 80 90 100 110 Reconstructed Invariant Mass (GeV/c2) Contributions to corrected bin 64.15 < Mparticle < 85.92

30

40

50

Unfolded Yield

160 140 120 100

Unfolded Yield

18 16 14 12 10 8 6 4 2

500

400

300

80 60 40 20

200

100

0 20

30

40

50

60 70 80 90 100 110 Reconstructed Invariant Mass (GeV/c2)

0 20

30

40

50

60 70 80 90 100 110 Reconstructed Invariant Mass (GeV/c2)

Contributions to corrected bin 85.92 < Mparticle < 117.29 Reconstructed Invariant Mass (GeV/c2)

0 20

30

40

50

60

70

80

90

100

110

Here are the same plots for all of the bins The last bin has contributions from ONLY other bins
7 October 2009 Collaboration Meeting

Unfolded Yield

0.6 0.5 0.4 0.3 0.2 0.1 0 20

30

40

50

60 70 80 90 100 110 Reconstructed Invariant Mass (GeV/c2)

56

UNFOLDING
Method used based on G. DAgostini, NIM A 362 (1995), p. 487. Also used by (along with H1, ZEUS, HARP, and others): IceCube: arXiv:0811.1671 L3: arXiv: hep-ex/0507042 D0: arXiv: hep-ex/9807029 Use PYTHIA to populate the unfolding matrix A (in the naming convention of DAgostini) using the reconstructed invariant mass and the particle invariant mass Normalize so that A does not change the integral of the spectrum The following equation describes the matrix elements of A:

J(reconstructed bin j|particle bin i) ij = J(reconstructed bin j)

7 October 2009

Collaboration Meeting

57

2009 Simulation

Prepare a VM Image Start with a KVM image of Scientic Linux 5.3 Add ~50 additional required packages Install STAR libraries, ~2.5M lines of code Setup grid toolkit and credentials Install database server Setup scripts to interact with job manager Setup monitoring scheme Design HTTP based API for jobs to record messages in a database Write monitoring software

STAR

Matthew Walker

58

2009 Simulation

STAR

Matthew Walker

59