1. STAR coll. Meeting, March 26-31, BNL
Update on D0 analysis using full D0-vertex fit (µ-vertexing) and Silicon information KENT : J. Bouchet, J. Joseph, S. Margetis, J. Vanfossen
BNL : Y. Fisyak, V. Perevoztchikov
03/28/10
STAR coll. Meeting, March 26-31, BNL

2. STAR coll. Meeting, March 26-31, BNL
Outline
Since last analysis meeting :
Full D0-vertex fit code (TCFIT, micro-vertex code) has been evaluated/debugged.
Redo HIJING+D0 simulation study
‘real’ pT (not flat) spectra D0
Central Hijing Au+Au event
Rapidity not 
Started looking on Embedding (CuCu + AuAu)
FULL Pass/Production of Au+Au (run7) data
(~35 Million mbias events).
03/28/10
STAR coll. Meeting, March 26-31, BNL

3. STAR coll. Meeting, March 26-31, BNL
Outline-II
To recall, our strategy is :
Step 1 : pure D0 (see how signal parameter behave)
Step 2 : simulation : Hijing+1D0 (background behavior)
Step 3 : real Data
Step 4 : embedding (corrections/x-section/physics)
for each step, confirm that the microvertex code works as expected
and find usable cuts
03/28/10
STAR coll. Meeting, March 26-31, BNL

4. Secondary vertex fit methods used
Linear fit abandoned.
Helix swimming to DCA of the two track helices (V0-like) using the global track parameters to reconstruct helices (StPhysicalHelix) not saved .
Helix swimming to DCA of the two track helices (V0-like) using the parameters from StDcaGeometry : save full track information (covariance matrix) inside the vacuum (center of beam pipe).
Full D0/Helix Fit (TCFIT) with vertex constraint and full errors
Also a full Kalman D0-fit was tried but not significant gains in time etc
The combined info from points 3, 4 will allow momentum dependent cut using the full track information.
Least square fit of the decay vertex. In 2 body decay, combination of 2 tracks + addition of constraint(s) to impose ‘external knowledge’ of a physic process and therefore force the fit to conform to physical principles.
The Kalman fitter machinery allows the knowledge with high precision of tracks near the primary vertex (by taking into account the MCS due to the silicon layers).
03/28/10
STAR coll. Meeting, March 26-31, BNL

5. Long-ctau D0 evaluation
Each plot shows the correlation of the secondary vertex position from GEANT (y-axis) with 1 of the 3 methods investigated : TCFIT,global helix and DCA geometry helix (x-axis) for its 3 components.
TCFIT
GEANT
DCA X Y Z
J.Vanfossen
RECO
03/28/10
STAR coll. Meeting, March 26-31, BNL

6. ‘normal’ c D0 evaluation
TCFit does a bit better job than either helix swimming method.
The scatter along the x axis of the swimming methods can be attributed to low pt D0’s and daughter tracks that are close to being parallel or anti parallel. X Y Z
J.Vanfossen
03/28/10
STAR coll. Meeting, March 26-31, BNL

7. STAR coll. Meeting, March 26-31, BNL
Resolution plots
230µm
230µm
190µm
TFCIT makes a slightly better job in terms of resolution
270µm
260µm
250µm
270µm
260µm
250µm
J.Vanfossen
03/28/10
STAR coll. Meeting, March 26-31, BNL

8. STAR coll. Meeting, March 26-31, BNL
Simulation
Previously we have shown results from simulation (Au+Au Hijing +1D0) but it was with flat pT D0 and flat in centrality
We have studied the significance of the signal (invariant mass peak) as a function of cuts.
We have redone the simulation with real pT spectrum etc. With proper scaling we have build realistic, expected inv. mass plots
simulation
Example
M(D0)
simulation
L > 2
M(D0)
J.Bouchet
03/28/10
STAR coll. Meeting, March 26-31, BNL

9. STAR coll. Meeting, March 26-31, BNL
Real pT spectrum
Generate a pT distribution with a power law function : d2N/dpTdy = A[1 +pT/p0]n
with <pT> = 2p0/(n-3) for a better extrapolation at pT0
Ref :
Manuel Calderon (thesis)
Transverse momentum spectra of charged particle in ppbar collisions at sqrt(s) = 630 GeV, Phys. Letter B 366(1996)
Pick randomly values in this distribution.
changing n from 10 to 14 does not really modify the shape (which is what we want) of the pT distribution.
Parameters chosen : n=10, <pT>=1 GeV/c.
03/28/10
STAR coll. Meeting, March 26-31, BNL

10. STAR coll. Meeting, March 26-31, BNL
Check
Generated function
Values (here 10k events)
randomly picked
Mean <pT> is 1
03/28/10
STAR coll. Meeting, March 26-31, BNL

11. STAR coll. Meeting, March 26-31, BNL
Simulation settings
Run the BFC chain with geometry y2007g
Use the slow simulator
Library : SL08f
Au+Au central collisions : b= fm
Vertex position from Hijing event
D0 phase space :
0 < pT < 5 GeV/C power law
-1 < y< 1 flat
0 < phi < 6.28 flat
Check the result from Starsim : if D0 in fz file is correct.
03/28/10
STAR coll. Meeting, March 26-31, BNL

12. STAR coll. Meeting, March 26-31, BNL
Check .fz file
distributions look as expected
03/28/10
STAR coll. Meeting, March 26-31, BNL

13. STAR coll. Meeting, March 26-31, BNL
NEXT with HIJING:
increase statistics (up to now 6.5k events) to make conclusion
verify cut selection
normalize signal -> (pseudo) estimate x-section, produce ‘realistic’ mass plots
03/28/10
STAR coll. Meeting, March 26-31, BNL

14. STAR coll. Meeting, March 26-31, BNL
D0 embedding
Dataset : run 7
File:/eliza14/star/starprod/embedding/2007ProductionMinBias/D0_103_ /P08ic.SL08f/*.MuDst.root
(~ 4.8K Min-bias events)
The Embedded data contains SiliconHits, most of them have 0,1,and 2, but a very few has 3 SiHits.
The Multiplicity range goes from zero to ~1200
The pT distribution of D0s goes from 0 to 5GeV/c
Rapidity goes +/-1 (Gaussian like)
J.Joseph
03/28/10
STAR coll. Meeting, March 26-31, BNL

15. First look at Embedding with the microvertex code
Cuts used :
|PVz| < 10
TpcHits>20
Eta, +/-1
nSigmaPi|<2, |nSigmaK|<2
pTK > 0.1GeV/c, pTPi > 0.1GeV/c
|cos(*)|<0.6, chargeK<0 ,chargePi>0,
|etaD0|<1.85
Slength_tcfit < 1mm
Requirement of
silicon hits >0
No Signal
S/N ~
Silicon hits are important
Goal is to get familiar with environment, see what is useful and formulate our request
More plots (QA) at :
J.Joseph
03/28/10
STAR coll. Meeting, March 26-31, BNL

16. STAR coll. Meeting, March 26-31, BNL
Real Data
Run over run 7 data productionMinBias.
Sample is ~35 Million events/ ~55 vertices
QA plots done day by day :
Cuts (see next slide) chosen to speed the code.
03/28/10
STAR coll. Meeting, March 26-31, BNL

17. STAR coll. Meeting, March 26-31, BNL
Cuts used (real data)
EVENT level
triggerId : , ,
Primary vertex position along the beam axis : |zvertex| < 10 cm
Resolution of the primary vertex position along the beam axis: |zvertex|< 200µm
TRACKS level
Number of hits in the vertex detectors :SiliconHits>2 (tracks with sufficient DCA resolution)
Momentum of tracks p >.5GeV/c
Number of fitted TPC hits > 20
Pseudo-rapidity :||<1 (SSD acceptance)
dEdxTrackLength>40 cm
DCA to Primary vertex (transverse) DCAxy< .1 cm
DECAY FIT level
Probability of fit >0.1 && |sLength|<.1cm
Particle identification : ndEdx :|nK|<2, |nπ|<2
03/28/10
STAR coll. Meeting, March 26-31, BNL

18. STAR coll. Meeting, March 26-31, BNL
Multiple vertices
We (wrong) use all vertices in a given event.
As seen (plot left), events can have several vertices (with different positions and resolutions): event 31 has 8 vertices.
Issue (?) : we did not fill the rank of the vertex to make a selection on the best vertex.
Event 31
however there is a correlation btw the gRefMult (which we saved) and the rank of each vertices.
Cut is possible to select ‘only’ good (best) vertex.
Possible offline cut
03/28/10
STAR coll. Meeting, March 26-31, BNL

19. D0 signal in 2007Production mbias
Cuts(offline):
50µm< decaylength<400µm
trackDca<200 µm
dcaD0toPV<300µm
pTkaon>0.7GeV/c
pTpion>0.7GeV/c
Plot as a function of gRefMult
J.Joseph
03/28/10
STAR coll. Meeting, March 26-31, BNL

20. STAR coll. Meeting, March 26-31, BNL
50<gRefMult
S/N ~ 5.7
J.Joseph
03/28/10
STAR coll. Meeting, March 26-31, BNL

21. STAR coll. Meeting, March 26-31, BNL
50<gRefMult<500
S/N ~
J.Joseph
03/28/10
STAR coll. Meeting, March 26-31, BNL

22. STAR coll. Meeting, March 26-31, BNL
50<gRefMult<800
S/N ~
J.Joseph
03/28/10
STAR coll. Meeting, March 26-31, BNL

23. STAR coll. Meeting, March 26-31, BNL
50<gRefMult<1000
S/N ~
D0 signal remains stable. Not the case with low statistics.
An increase in the significance when including high multiplicities events is observed.
J.Joseph
03/28/10
STAR coll. Meeting, March 26-31, BNL

24. STAR coll. Meeting, March 26-31, BNL
Outlook
Microvertex code gives consistent results with GEANT and helix swimming.
Run for the first time on the almost entire run 7 data with microvertex code.
D0 signal remains stable under various cuts (good).
D0 signal seems to increase when high multiplicities are included (good).
Next step : try pT bins and make spectra.
Simulation : work in progress to increase stats. of real pT D0 in Hijing to make conclusion.
Investigation (near future) of other decays.
03/28/10
STAR coll. Meeting, March 26-31, BNL

25. STAR coll. Meeting, March 26-31, BNL
Back-up
03/28/10
STAR coll. Meeting, March 26-31, BNL

26. Constrained vertex fit
2 = ∑(yi0 - yi(x*))TV-1(yi0 - yi(x*)) + F
where :
x* : secondary vertex position
yi0 : track parameter of the original fit
y : track parameter after refit with knowledge
of the secondary vertex
V : covariance matrix of the track parameter
i : sum over tracks
F : constraint  f
f : physical process to satisfy K- π+
3d path length from primary vertex
to decay particle vertex
The constraint(s) is(are) added to the total 2 via Lagrange multiplier 
The minimum of 2 is then calculated with respect to the fit parameters x and with respect to  because the condition ∂2/∂  =0 required for the minimum correspond the the constraint equation f
03/28/10
STAR coll. Meeting, March 26-31, BNL

27. Check the MuDst file for the simulation (real pT)
after running the reconstruction, I have run MuKpi and checked the result
pT and the rapidity of reconstructed D0 look as the input
All combinations
K-π+
All combinations
K-π+
03/28/10
STAR coll. Meeting, March 26-31, BNL

28. Invariant mass D0 after selected cuts
Look at the entire production done ~35M events
Cuts :
siliconHits ==4
|cos*|<.6
|yD0|<.5
Slength > 200 µm
dcaXY(pion)*dcaXY(kaon)<0
|cospointing|>.9
Probability (of TCFIT) >75
W/o any background substraction, a ‘signal’ can be seen
J.Bouchet
03/28/10
STAR coll. Meeting, March 26-31, BNL

29. Resolution plots (Long)
There is no systematic shift in reconstructed quantities in all 3 directions.
The standard deviation of the distribution is flat at ~ 220 m , which is of the order of the resolution of (SSD+SVT).
220µm
210µm
200µm
210µm
230µm
210µm
200µm
230µm
220µm
J.Vanfossen
03/28/10
STAR coll. Meeting, March 26-31, BNL