20/12/2011Christina Anna Dritsa1 The model: Input Charge generation The charge of the cluster is...
-
Upload
allison-barker -
Category
Documents
-
view
217 -
download
0
description
Transcript of 20/12/2011Christina Anna Dritsa1 The model: Input Charge generation The charge of the cluster is...
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 11
The model: InputThe model: InputCharge generationCharge generation
The charge of the cluster is taken by random sampling of the experimental distribution for 25 pixels
0 degrees incident angle
Charge sharing
The charge distribution among the pixels in the cluster is based on a 2D Lorentz distribution (derived from the 1D)
0 degrees incident angle
Landau:Accumulated charge on 25 pixels
The simulation of inclined particles is derived by this initial parameterization.
1: Charge generation
2: Charge sharing among pixels
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 22
The model: charge distributionThe model: charge distribution
Charge provided by Landau (25 pixels)Charge provided by Landau (25 pixels)– if needed: scale to the trajectory lengthif needed: scale to the trajectory length
The trajectory is divided in segmentsThe trajectory is divided in segments A Lorentz function corresponds to each segmentA Lorentz function corresponds to each segment
Illustration for inclined track
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 33
The modelThe model
L(xk,i,yk,i)
Charge on pixel i
Sum over segments (k) x,y-coordinates
of pixel i Pixel pitch
LorentzAmplitude
Lorentzwidth
x,y-coordinatesof segment k
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 44
EvaluationEvaluationAccumulated charge plot
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 55
EvaluationEvaluationAverage pixel multiplicityAverage pixel multiplicity
Good agreement between simulation and experimental data
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 66
Evaluation (qualitative): shapeEvaluation (qualitative): shapeSimulation Experimental data
Asymmetry is not sensor feature.Reflections in readout cable=> Not simulated.
Intermediate summary CBM aims to explore the QCD phase diagram
with rare probes Requirement for high intensities and performing
vertex detector (MAPS sensors) Collision pile up and delta electrons increase
substantially the hit density Precise simulation of detector response is crucial Detector response model developed and
successfully tested with experimental data
Perform D0 → K- π+ measurement feasibility study
Open questions Evaluate open charm performance accounting
for realistic sensor response and delta electrons.
Is pile up tolerable?
What is the impact of particle identification in open charm performance?
Approach: simulation– test different assumptions on pile up– test different assumptions on particle identification
Monte Carlo Transport Code(Geant3/Geant4)
Detector response models
Event Generation
Thermal model
UrQMD
beam particles
Simulation chain
Generate D0 , T=300 MeV, σY=1
nuclear coll. Au+Au , 25 AGeV
Delta electron generation due to passage through target
Simulate interaction of particlesthrough matter, MF etc
Simulate realistic detector response
Simulation chain (reconstruction level)
Hit Finding
Track Finding(Cellular Automaton)
Track fitting (Kalman filter)
Primary and secondary vertex fitting
Analysis code
Reconstruction of particle impact point
Association of hits belonging to the same particle trajectory
Fit the particle track and provide charge sign and momentum
Provides coordinates and uncertainties of vertex position
Reject the maximum of backgroundwhile keeping the maximum of signal
Definition of pile up What is a pile up of N collisions?
– Several collisions occur within a readout cycle of the detector
In simulation: Accumulate hits from N collisions– 1 central Au+Au at 25 AGeV– N-1 minimum bias– 100xN beam ions (delta electrons)
Pile up occurs only in the MVD layers
Generation of high statistics
Generation of high statistics 1 D0→π+K- per 106 collisions: requires highly efficient background
rejection (better than 10-9)
Need high statistics to test background rejection
Use event-mixing like technique:– During the last step (analysis):– Combine opposite charge particles originating from different
nuclear collisions.– Gain CPU time ( no need for GEANT simulation, track
reconstruction)
Increase statistics by factor N (= number of collisions simulated)
Statistics reached: ~108 collisions (from ~104)
time consuming
Background rejectionBackground rejection
High statistics background is generated.
How to efficiently eliminate it?
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 1515
Cut optimisationCut optimisation Criterion to define optimum cut value:
– Maximise significance S/sqrt(S+B)
Method: Use multidimensional analysis in which the significance is maximised using simultaneously all cuts+ Fast and user friendly- may converge to local maxima => careful usage
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 1616
The CBM – MVD The CBM – MVD
MVDSTS
RICHTRD
TOFECAL
PSD
Detector integration: IKF, Frankfurt
Station Z (cm) Rinner [mm] Router [mm]
Mat. Budget
1 5 5.5 25 0.3% X0
2 10 5.5 50 0.5% X0
Results: cluster mergingMerged cluster (MC)
MVD station
Unambiguous clusters (UC)
What is the fraction of unambiguous clusters as a function of the hit density (collision pile up) ?
f=UC/all
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 1818
Invariant mass distributionsInvariant mass distributions
open cuts open cuts
Background Signal
final cuts final cuts
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 1919
S and B calculationS and B calculation
max
min
)(~ m
m
mGaussS max
min
)(~ m
m
mExpoB
S
signalNorm bgNorm
B
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 2020
ResultsResults S/B efficiency
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 2121
SummarySummary Feasibility of open charm Feasibility of open charm
measurements was investigated measurements was investigated based on newly developed detector based on newly developed detector response model for MAPSresponse model for MAPS
Delta electrons were accounted forDelta electrons were accounted for Different assumptions on event pile Different assumptions on event pile
up and PID capabilities were madeup and PID capabilities were made
ConclusionConclusion The digitiser reproduces the response of
MAPS sensors within 10% in terms of… For an event pile up above 5 substantial
cluster merging is observed Additional counting statistics for a moderate
pile up is cancelled out by reduced sensitivity.
CBM remains sensitive to open charm with S/B between on ~0.1 and ~3 depending on pile up and particle identification assumptions.
Expect better results for collisions at 35 AGeV
Outlook Improve cluster finding algorithms
Adding a 3rd MVD station might improve sensitivity of CBM (master thesis C.Trageser, IKF)
Digitiser describes also partially depleted and irradiated MAPS (master thesis M.Domachowski, IKF)
Expect improvement on sensor development: new CMOS processes allow approaching ~µm
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 2424
Additional slidesAdditional slides
Simulation setup CBMROOT simulation framework
– root based framework
Detectors:– MVD for vertex reconstruction– STS for track and momentum reconstruction – Thickness of silicon detectors : 300 µm– TOF is modelled with ideal proton identification.
Tracking-vertexing– Cellular automaton and Kalman filter
Realistic detector response (MVD , STS)
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 2626
Strategy of D0 reconstructionStrategy of D0 reconstruction Reconstruct the invariant mass by
combination of all opposite charge pairs
Apply selection cuts– Define cuts– Optimise
Evaluate performance (S/B, significance…)– Estimate Signal, Background– Use scaling factors where needed
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 2727
MVD detector geometryMVD detector geometry Thickness of sensors – GeometryThickness of sensors – Geometry
– 11stst MAPS at 5 cm is 300 MAPS at 5 cm is 300 µµm thickm thick– 22ndnd MAPS at 10 cm is 500 MAPS at 10 cm is 500 µµm thickm thick
Station Z (cm) Rinner [mm] Router [mm]
Mat. Budget
1 5 5.5 25 0.3% X0
2 10 5.5 50 0.5% X0
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 2828
Assumptions on collision rateAssumptions on collision rate CBM year: CBM year: 55··10106 6 ss ≈ 2 months ≈ 2 months Assumed sensor time resolution: Assumed sensor time resolution: ttintint = 30 = 30 µµss
Collision rateCollision rate(interactions/s)(interactions/s)
Collisions/yearCollisions/year(mbias)(mbias)
No pile upNo pile up 3 3 ··10 10 44 1.51.5··10101111
Pile up NPile up N N x 3 N x 3 ··10 10 4 4 N x 1.5N x 1.5··10101111
Pile up 5Pile up 5 1.5 1.5 ··10 10 55 7.57.5··10101111
Event generators Nuclear collisions: UrQMD used for the final state phase
space distributions of hadrons for Au+Au collisions at 25 AGeV.
Delta electrons are generated with GEANT by the passage of beam particles through the target.
D0 signal generated with thermal model and embedded in Au+Au collisions. Production multiplicity taken from:– HSD : 2 x 10-4 – SHM : 3.7 x 10-5
Due to low production multiplicities, event mixing-like technique is used to generate high statistics for background.
Data Processing and Data Levels
Event generatorUrQMD, HSD, user defined, ...
Transport (VMC)GEANT3, GEANT4, FLUKA, ...
Detector Response
Reconstruction
Analysis
CB
MR
OO
T
Sim
ulat
ion
(MC
)
GEN
MC
RAW
ESD
Data Level
Experiment DAQ
CBM Software Workshop, Ebernburg, 8 November 2011 30Volker Friese
20/12/201120/12/2011 Christina Anna DritsaChristina Anna Dritsa 3131
Background rejectionBackground rejection
π+
K-
D0
PVSV
Cuts: e.g. impact parameter, vertex position, quality of vertex…
Cut effect
Explain pile up
٭ ٭ ٭ ٭٭
٭٭ ٭ ٭
٭
Explain pile up
٭ ٭ ٭ ٭٭
٭٭ ٭ ٭
٭٭ ٭٭ ٭
٭٭٭٭
٭٭٭٭ ٭٭٭٭٭٭
٭ ٭٭
٭٭٭٭٭٭ ٭٭٭٭٭
٭
Detector geometry MVD for vertex reconstruction STS for track and momentum
reconstruction PID assumptions…