Status of Run 8 FMS w analysis

1
Status of Run 8 FMS w analysis
STAR
Andrew Gordon Analysis Meeting at MIT July 8, 2009
2
Outline
Run 8 inclusive pions
Data extraction for omegas and comparisons of
GEANT and Data
Fast simulation and additional cuts
dAu comparisons and future work
3
Inclusive pions can be simulated over nearly
entire range
Simulation is PythiaGSTAR
Trigger simulation Includes simulation of high
tower trigger using realistic gains to calculate
ADCs per tower.
4
PT dependence of Run 8 inclusive pions
FMS data reaches PT 6 GeV
Mgg in bins of PT
XF-PT locus of data
Akio Ogawa, CIPANP 2009
5
PT dependence of Run 8 AN
Indications that AN persists to PT 5 GeV.
Positive XF
Negative XF
J.Drachenberg, Spin 2008, arXiv0901.2763
Akio Ogawa, CIPANP 2009
6
FMS acceptance allows azimuthal AN dependence to
be measured and extends data to higher XF
Run 8 data
Run 8 data
P(blue)45.53.3
L6.2 pb-1 in plot
Anf(f) versus ltcos fgt for positive (blue beam)
and negative (yellow beam) XF
AN as a function of xF integrated over the FMS
acceptance.
Important confirmation of previous data
Plots from Nikola Poljak, for STAR collaboration,
Spin-dependent Forward Particle Correlations in
pp Collisions at ?s 200 GeV,
hep-ex/0901.2828, to be published as Spin 2008
conference proceedings.
7
Enhanced areal coverage of FMS
allows new measurements
Run 5 FPD
8
Motivation to search for w signal
Spin asymmetries The w is spin-1. The Artru
string fragmentation model predicts that a spin-1
particle should have opposite asymmetry from a
scalar particle, if the observed asymmetries
result from the Collins effect. Observing a
negative asymmetry would provide strong evidence
for the Collins effect.
AN for w
Mass shift in dAu relative to pp The Run 8
data allows a comparison of pp data with dAu.
Shifts in the mass or width of the w would
provide evidence of the partial restoration of
the chiral symmetry in a denser hadronic medium.
Spectral shifts
RdAu Can compare results from this resonance to
previous scalar particle measurements.
Yield ratios
References
Artru and Dzyzewski, hep-ph/9805463
Brown and Rho, PRL 66, 21 (1991)
STAR collaboration, PRL 97, 152302 (2006)
9
Cut set 1 Use BBC to require w photon neutral
w decays accessible in the FMS through p0g
(BR9) decay channel.
Track photon associated with w decay back to
small tiles of BBC and require fewer than 5
counts in BBC.
Four west small tile BBC phototubes, zero
suppressed
MIP peaks 18 counts
Suppresses charged hadrons that looks like
photons in FMS.
Forces photons to be isolated from charged tracks.
Removes many real events by requiring that the
photon did not convert in material before BBC
(below 1 X0).
Forces photons into inner tiles and thus lower PT.
Kinematic cuts
Look at all clusters with Egt3 GeV instead of 6.
Look in window 30ltE(triple)lt35 and
2.0ltPT(triple)lt2.5 GeV
10
Spin-sorted (on blue beam) mass distributions
Right
Left
Large acceptance asymmetries
Calibration did not use data in this low cluster
energy region. Non-linear corrections need to be
calculated.
We need simulation in this region of phase space.
There is a hint of a negative asymmetry
11
Set 2 High PT w selection
See A. Gordon, Moriond 2009 Proc., arXiv0906.2332
Look at all triples of clusters with Egt6 GeV.
Apply fiducial cuts of 1/2 cell from all module
boundaries.

• Kinematic cuts to reduce QCD background (real p0
  decays with a third EM-rich hadronic cluster in
  the FMS)
• PT(triplet)gt2.5 GeV
• E(triplet)gt30 GeV
• PT(photon cluster)gt1.5 GeV
• PT(p0)gt1 GeV.

12
Mass distribution of all triples
Significant (10s) w?p0g signal seen in the data.
13
Df Distribution too narrow in PYTHIA
Dfdifference in azimuth in STAR coordinates
between p0 and g.
Comparison of Df distribution for data and
simulation
Df distribution too narrow in simulation (compare
black solid lines to blue dashed lines).
Replacing GSTAR and trigger simulation with
geometric acceptance only and using PYTHIA level
quantities (red triangles) does not change
distribution.
14
Df(p0g) for PYTHIA 6.2.22
CDF Tune A does not change distribution
Linear scale
Default PYTHIA 6.2.22 tune
PYTHIA quantities level only.
PYTHIA 6.2.22 with CDF Tune A parameters
Log scale
PYTHIA quantities level only.
Pythia 6.4.20 gave a narrower distribution with
default tune and with 6.2.22 tune.
Df(p0g)
15
Conclusions on Df distribution
PYTHIA Df distribution is tunable
PYTHIA needs to be tuned in this kinematic region.
One possibility is that the momentum components
perpendicular to the thrust axis (JT) need better
tuning in the forward region.
16
Backgrounds from simulation
Currently backgrounds are too high for a spin
analysis.
We do an association analysis with reconstruted
clusters and Pythia particles for triples in the
w mass region.
We find that the three clusters are 100
photons.
Perhaps do not need slow GEANT simulation in this
kinematic region.
Looking at the mother of the w decay photon
55 of backgrounds have p0 mothers
30 of backgrounds have h mothers




p0
h
Fake w


h
Fake w


Fake w

or



p0
p0
p0
Isolation-type cuts might help reduce these
backgrounds.
17
Fast Simulation
A fast simulation can help us tune the cuts and
would also make it possible to greatly extend
integrated luminosity of simulation.
Resources for the PythiaGSTAR simulation
We ran a Pythia preselection with 3.5 pass
rate.
83 nb-1 used 100,000 hours of CPU and 500
Gbytes
By varying offline random seeds (for example, the
smearing on the measured Z(BBC) vertex), we
increased the sample size relatively easily by a
factor of 3.
A bench-marked, fast simulation can greatly
reduce the resources needed. With linux farm,
1-2 pb-1 per day. Some of this increase is
increased efficiency of a photons-only
preselection.
18
Ingredients for fast simulator
Track all PYTHIA photons to face of FMS.
Smear x-y position by 10th of cell size.
Spread energy among cells based on transverse EM
shower shape.
After all photons have been accumulated, smear
energy in each FMS tower Currently using per
tower
s(E)/E 15/?E 2
Pass event to analysis code as if it were a
normal GEANT event and use offline gain
determinations to calculate ADCs and apply high
tower trigger.
19
Comparison of Fast simulator with GSTAR
Mass of two photons associated with p0 in triple.
Gaussiancubic fits to peak region give
Fast sim µ0.13720.0005 GeVs0.02190.0004 GeV
Fast sim
GSTAR
GSTAR µ0.1344 GeVs0.0202 GeV
Note data tends to be wider. This is partly
because large cell gains are less well
determined..
M(gg) (GeV)
20
Other variables to reduce backgrounds
RISOminimum transverse distance between extra
clusters and the w decay photon
Fast sim
GSTAR
Data
To reduce background where w decay photon is
really from a p0, search through all other
fiducial photon clusters in the FMS and calculate
nearest transverse distance.
The peak below 20 is from p0 decays.
RISO(cm)
21
Other variables to reduce backgrounds (cont)
Mpmass of extra cluster with each of the triple
clusters closest to 0.135 GeV/c2
Fast sim
Data
For w decay g
Instead of transverse distance, calculate the
pair mass that is closest to the pion mass, of
all the extra clusters in the event.
For p0 photon 1
Do this for the w decay photon and also for the
two photons from the identified w decay pion.
For p0 photon 2
p0 peaks are evident in all three distributions
Mp(with other clusters) (GeV/c2)
22
Other variables to reduce backgrounds (cont)
Mhmass of extra cluster with each of the triple
clusters closest to 0.548 GeV/c2
Fast sim
Data
RISOlt25 cm
For w decay g
Look at mass closest to the h mass after applying
a RISOlt25 cm cut.
RISOlt25 cm
For p0 photon 1
The h backgrounds are visible.
RISOlt25 cm
Perhaps isolation type cuts be improved by
looking at all energy in the FMS, instead of just
fiducial, fitted clusters
For p0 photon 2
Mh(with other clusters) (GeV/c2)
23
Update on dAu data
We have begun initial analysis of this data
Gain calibrations with neutral pions complete
Simulation needs to be generated
PYTHIA may be adequate for deuteron side (where
FMS sits)
May need to also compare to HIJING
For RdAu measurement
Use Hank Crawfords scaler boards for luminosity
tracking
Attempt to reproduce Run 3 RdAu measurements
(STAR collaboration, PRL 97, 152302 (2006)).
p
Extend to the w data.
24
Conclusions
A clear w signal is visible in the Run 8 pp data.
The fast simulator can be used to understand the
high PT triple data.
Additional cuts and better tuning on simulation
can improve signal to background.
To do
Continue simulation development
Tune the cuts.
Analyze the dAu data.