Real-Time Event Reconstruction and Analysis in CBM and STAR Experiments - IOPscience
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Journal of Physics: Conference Series
PAPER • OPEN ACCESS
Real-Time Event Reconstruction and Analysis in CBM and STAR
Experiments
To cite this article: Ivan Kisel and for CBM and STAR Collaborations 2020 J. Phys.: Conf. Ser. 1602 012006
View the article online for updates and enhancements.
This content was downloaded from IP address 46.4.80.155 on 07/01/2021 at 10:37
36th Winter Workshop on Nuclear Dynamics IOP Publishing
Journal of Physics: Conference Series 1602 (2020) 012006 doi:10.1088/1742-6596/1602/1/012006
Real-Time Event Reconstruction and Analysis
in CBM and STAR Experiments
Ivan Kisel
(for CBM and STAR Collaborations)
1
Goethe-University Frankfurt, Theodor-W.-Adorno-Platz 1, 60323 Frankfurt am Main,
Germany
2
FIAS Frankfurt Institute for Advanced Studies, Ruth-Moufang-Str. 1, 60438 Frankfurt am
Main, Germany
3
Helmholtz Research Academy Hesse for FAIR, Max-von-Laue-Str. 12, 60438 Frankfurt am
Main, Germany
4
GSI Helmholtz Centre for Heavy Ion Research, Planckstr. 1, 64291 Darmstadt, Germany
E-mail: I.Kisel@compeng.uni-frankfurt.de
Abstract. Within the FAIR Phase 0 program, the algorithms of the FLES (First-Level Event
Selection) package developed for the CBM experiment (FAIR/GSI, Germany) are adapted for
the STAR experiment (BNL, USA). Use of the same algorithms creates a bridge between online
and offline, which makes it possible to combine online and offline resources for data processing.
In this way, an express data production chain was created on the basis of the STAR HLT farm,
that extends the functionality of HLT in real time up to the analysis of physics. It is important,
that the express analysis chain does not interfere with the standard analysis chain. A particular
advantage of express analysis is that it allows calibration, production and analysis of the data
to begin immediately after they are collected. Therefore, the use of express analysis is beneficial
for BES II data production and helps to speed up scientific discovery by helping to obtain
results within one year after the end of data acquisition. The specific features of express data
production are presented and discussed as well as the results of online production and analysis,
such as real-time reconstruction of short-lived particles in the BES-II STAR environment.
1. Introduction
Within the framework of the Facility for Antiproton and Ion Research (FAIR) project, a large
international centre is being constructed to study the structure and fundamental properties of
matter. It will be a new generation accelerator complex that will provide unique opportunities
for detailed investigations in the most interesting areas of modern science: nuclear, hadron and
particle physics, atomic and anti-matter physics, high density plasma physics, and applications
in condensed matter physics, biology and bio-medical sciences [1].
In the Compressed Baryonic Matter (CBM) [2] experiment with heavy ions, the highest
baryon densities will be created, and the properties of super-dense nuclear matter will be
investigated in various extreme states that are similar to, for example, the conditions of matter
in the center of neutron stars, where matter is at the final stage of evolution before transition to
the black hole. The CBM experiment will thus complement the experimental heavy-ion program
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
36th Winter Workshop on Nuclear Dynamics IOP Publishing
Journal of Physics: Conference Series 1602 (2020) 012006 doi:10.1088/1742-6596/1602/1/012006
at the LHC accelerator complex at CERN where the properties of the hot matter similar to that
after the Big Bang are investigated.
The scientific program of the CBM experiment includes:
• explore properties of super-dense nuclear matter;
• search for in-medium modifications of hadrons;
• search for the transition from dense hadronic matter to quark-gluon matter;
• search for the critical endpoint in the phase diagram of strongly interacting matter;
• investigate the structure of neutron stars and the dynamics of core-collapse supernovae.
The experiment will measure rare and penetrating probes such as dilepton pairs from light
vector mesons and charmonium, open charm, multistrange hyperons, together with collective
KF
hadron flow and Particle:
fluctuations Reconstruction
in heavy-ion short-lived
collisions at rates Particles
up to 107 collisions per second.
Concept:
π+
• Mother and daughter particles have the same state
vector and are treated in the same way
• Reconstruction of decay chains
Κ+ Figure • 1.KalmanAFiltersimulated
(KF) based central Au-Au
collision •atGeometry
25 AGeV energy with about 1000
independent
• Vectorizedin the CBM experiment.
charged particles
• Uncomplicated usage
p
Functionality:
Λ • Construction of short-lived particles
Ω+ Concept of KF Particle in CBM • Addition and subtraction of particles
• Transport
1. KFParticle class describes
collision at 25particles by: • Calculation of an angle between particles
3 KFParticle: Reconstruction of Vertices and Decayed
Simulated AuAu
Particles
AGeV
π+ r = { x, y, z, px, py, pz, E } • Calculation of distances and deviations
CBM is characterized by Position,
highdirection,collision
2
s2 C C rates, C large
3
•amount ofonproduced
Constraints particles,
mass, production non-length
point and decay
Λ̅ KΩ̅
C C C
+ + 6 x
xy xz xpx xpy xpz xE
2 7
State vector State vector momentum
6C
6 s y C C C C C 7
7 xy yz ypx ypy ypz yE
homogeneous
π+ Κ + magnetic fields
(r,C) and and energy6
a very 6C complex
C sz2 C
C = = 66C C C sp2 C C C 77
6
C
detector
C
7
C 7
7 • system.
KF Particle Event
Finder reconstruction
xz yz
is the
zpx zpy zpz zE
p̅ π+and time consuming task of the data analysis in modern high-energy physics
xpx ypx zpx x px py px pz px E
6 7
sp2y Cpy pz Cpy E 7
most complicated r = { x, y, z, px, py, pz,66CExp} Cyp
6C
y y Czpy Cpx py
4 xpz Cypz Czpz Cpx pz
7
Cpy pz sp2z Cpy E 7
5
Covariance matrix
experiments. Κ+ Itp is a key part of success in the CBM experiment Reconstruction
with of
CxE
updecays
CyE with neutral
to thousand
CzE C px E daughter
Cpy E Cpz E sE2
particles
Concept: by the missing mass method:
2. Covariance matrix contains essential information
per central collision (Fig. 1). An additional complication in CBM is its continuous data stream
• Mother and daughter particles have the same state
KFParticle Lambda(P, Pi); // construct
about tracking anti Lambda
and detector performance.
Λ represented
p
in form of3. time
Lambda.SetMassConstraint(1.1157); The method slices.
// improve This
vector and are treated in the same way
momentum
for mathematicallymakes and theusage
mass reconstruction
• Reconstruction of decay chains
correct of
π-
of such 4-dimensional data
Ω+ • Kalman filter based
Λ KFParticle
with timeOmega(K,
stamps Lambda);
and the covariance matrices
// construct
search isinteresting
for anti provided
Omega by thephysics KF Particle extremely difficult. All of the above
• Geometry independent
Ω+ PV -= (P; Pi; K); package based on the primary
// clean Kalman filter
• Vectorized
(KF) developed by
vertex
mentioned
PV += Omega;
makes necessary FIAS group to// 1,2
develop
addprimarily
Omegafor
fast and
CBMprimary
to the and efficient algorithms for data analysis and to
ALICE.
vertex
Σ-
• Uncomplicated usage
̅ ++
Ω Λ̅ K++
Simulated AuAu collision at 25 AGeV
Ω̅ Λ̅ K optimize them for
Omega.SetProductionVertex(PV); 4. Heavyon
running
̅̅ ππ+ (K; Lambda).SetProductionVertex(Omega);
+ Functionality:
mathematics
a
// modern
Omega is requires
fully fast and vectorized computer
high-performance
fitted
algorithms.// K, Lambda are fully fitted
cluster [3]. n
pp • Construction of short-lived particles
5. Mother
(P; Pi).SetProductionVertex(Lambda);
• Addition and
// p,daughter
and subtraction pi are fullyparticles
of particles fitted are KFParticle and
2. First Level Event are treated in the same way.
KFParticle Lambda(P, Pi); // construct anti Lambda
Lambda.SetMassConstraint(1.1157); // improve momentum and mass Selection • Transport
KFParticle Omega(K, Lambda); KF Particle provides
// construct anti Omega 6. aThesimple
naturaland and direct approach
simple to physics
interface
• Calculation of an angle between particles allowsanalysis
to (used in CBM, ALICE, STAR and sPHENIX)
PV -= (P; Pi; K);
PV += Omega;
The First Level Event Selection
// clean the primary vertex
// add Omega to the primary vertex
(FLES) package decay[4,
• Calculation of distances and deviations
reconstruct easily rather complicated chains.5] of the CBM experiment is intended to
Ivan Kisel, Uni-Frankfurt, FIAS, GSI
Omega.SetProductionVertex(PV); // Omega is fully fitted
• Constraints on mass, production point and decay length WWND, Puerto Vallarta, 02.03.2020 9 /20
reconstruct online the7.full
(K; Lambda).SetProductionVertex(Omega); // K, Lambda are fully fitted
The package
event istopologygeometry independent
• KF Particle Finder including and tracks
can be of charged particles and short-lived
(P; Pi).SetProductionVertex(Lambda); // p, pi are fully fitted easily adapted to different experiments.
particles. The FLES package consists of several modules: Cellular Automaton (CA) track finder,
KFParticle provides uncomplicated approach to physics analysis (used in CBM, ALICE and STAR)
1. KF Particle — S. Gorbunov, “On-line reconstruction algorithms for the CBM and ALICE experiments,” Dissertation thesis, Goethe University of Frankfurt, 2012,
kalman Filter (KF) track fitter, KF Particle Finder and physics selection. In addition, a quality
http://publikationen.ub.uni-frankfurt.de/frontdoor/index/index/docId/29538
V. Akishina, I. Kisel, Uni-Frankfurt, FIAS MMCP 2017, Dubna, 07.07.2017 11 /16
2. KF Particle Finder — M. Zyzak, “Online selection of short-lived particles on many-core computer architectures in the CBM experiment at FAIR,” Dissertation
thesis, Goethe University of Frankfurt, 2016, http://publikationen.ub.uni-frankfurt.de/frontdoor/index/index/docId/41428
20 July 2018
check module is implemented, STAR Collaboration Meeting
that allows to monitor and 2 /18
control the reconstruction process at
all stages. The FLES package is platform and operating system independent. The package is
portable to different many-core CPU architectures, vectorized using SIMD (Single Instruction,
Multiple Data) instructions and parallelized between CPU cores. All algorithms are optimized
with respect to the memory usage and the speed.
2
36th Winter Workshop on Nuclear Dynamics IOP Publishing
Journal of Physics: Conference Series 1602 (2020) 012006 doi:10.1088/1742-6596/1602/1/012006
2.1. Cellular Automaton (CA) track finder
The 4-dimensional (4D, space and time) Cellular Automaton (CA) track finder [5, 6] takes as
input hit measurements from the tracking detector in the form of a time-slice, which includes
time and spacial measurements. The track finding procedure starts with combining the hits
into triplets, combination of three hits on adjacent stations. The triplet structure was chosen,
since it allows to estimate the momentum of a particle, which could produce it. The triplets
with two common hits are combined into track candidates. The track candidates should survive
a dedicated selection based on the track length and calculated χ2 -value to be accepted to the
reconstructed tracks.
Input time information is used in the algorithm to the same extent and in similar manner
as it is done with the spacial coordinates. The same logic is used while constructing triplets:
the hits in the triplet should belong to the same particle, therefore they should correlate not
only in space, but also in time. The resulting track reconstruction efficiencies for the cases of
event-by-event analysis (so-called 3D analysis) as well as for the 4D case (with included time
measurement, as well as 3-dimensional spacial information) while reconstructing time-slices are
similar thus there is no efficiency degradation in the much more complicated case of time slices.
4D ofEvent
The same is valid for the speed the 4DBuilding at 10with
CA tracks finder MHzrespect to the 3D case.
2.2. Kalman Filter (KF) track fit
High precision of the parameters of particle trajectories (tracks) and their covariance matrices
is aHitsprerequisite
at high input rates for finding rare signal events among hundreds of thousands of background
events. Such high precision is usually
Hits 0.1 MHz obtained by using the estimation
Hits 1 MHz Hits 10 MHz
algorithms based on
the Kalman filter (KF) method. High speed of the reconstruction algorithms on modern many-
core computer architectures can be accomplished by: optimizing with respect to the computer
memory, in particular declaring all variables in single precision, vectorizing in order to use the
SIMD instruction set and parallelizing between cores within a compute node.
2.3. 4D event builder
From hits to tracks to events
(1) Hits 10 MHz (2) Tracks (3) Events
Figure 2. Reconstructed tracks in time slices clearly represent groups, which correspond to the
Reconstructed tracks clearly represent groups, which correspond to the original events:
original events 85%with 85%
of single of single
events, events,
no splitted no splitted
events, further events,
analysis with and final
TOF information event
at the building
vertexing stage is done at
the vertexing stage using TOF information.
Ivan Kisel, Uni-Frankfurt, FIAS, GSI WWND, Puerto Vallarta, 02.03.2020 8 /20
After all tracks are found and their parameters are reconstructed, the tracks are grouped
into events. This is done by clustering tracks based on their time parameters in the area of the
target. The left distribution of Fig. 2 shows hits within a time slice for 107 interaction rate. One
can see that the traditional grouping of hits into events at this stage is impossible. The track
distribution at the middle shown against the same hits displays grouping of tracks belonging to
the same event. The right distribution with different colors shows different clusters of tracks,
close in time in the target area. One can see that already at this stage it is possible to build
events with efficiency more than 85%. The task of event building is finalized at the stage of
3
36th Winter Workshop on Nuclear Dynamics IOP Publishing
Journal of Physics: Conference Series 1602 (2020) 012006 doi:10.1088/1742-6596/1602/1/012006
searching the primary vertex, where it is possible to additionally use the proximity of tracks in
space, as well as more accurate time measurements of the TOF detector.
2.4. KF Particle Finder — a package for reconstruction of short-lived particles
Today the most interesting physics is hidden in the properties of short-lived particles, which
are not registered, but can be reconstructed only from their decay products. A fast and
efficient KF Particle Finder package [4, 7], based on the Kalman filter (hence KF) method,
for reconstruction and selection of short-lived particles is developed to solve this task. A search
of more4 than 100KF
decay channels
Particle has been
Finder currently
forFinder
Physics implemented
Analysis and(Fig.Selection
3).
KF Particle block-diagram
Charged particles: e±, µ±, π±, K±, p±, d±, 3He±, 4He±
Neutral particles: νµ, ν̅µ, π0, n, n̅ , Λ, Λ̅, Ξ0, Ξ̅0
Dileptons Open-charm Strange particles Hypermatter
Charmonium K0s → π+ π-
Open-charm K+ → µ+ νµ Hypernuclei
J/ψ → e+ e-
particles {Λn} → d+ π-
J/ψ → µ+ µ- K- → µ- ν̅µ Ξ- → Λ π-
D0 → K- π+ {Λ̅n̅ } → d- π+
K+ → π+ π0 Ξ̅+ → Λ̅ π+
Low mass D0 → K- π+ π+ π- {Λnn} → t+ π-
K- → π- π0 Ξ- → Λ π- Σ+ → p π0
vector mesons D̅ 0 → K+ π- {Λ̅n̅ n̅ } → t- π+
Ξ̅+ → Λ̅ π+ Σ̅- → p̅ π0
ρ → e+ e- D̅ 0 → K+ π+ π- π- Λ → p π- 3 H → 3He π-
Λ
Ω- → Λ K- Σ0 →Λγ
ρ → µ+ µ- D+ → K- π+ π+ Λ̅ → p̅ π+ Λ ̅
3 H → 3He π+
Ω̅+ → Λ̅ K+ Σ̅0 → Λ̅ γ
ω → e+ e- D- → K+ π- π- Σ+ → p π0 4 H → 4He π-
Λ
Ω- → Λ K- Ξ0 → Λ π0
ω → µ+ µ- Ds+ → K+ K- π+ Σ̅- → p̅ π0 Λ ̅
4 H → 4He π+
Ω̅+ → Λ̅ K+ Ξ̅0 → Λ̅ π0
ϕ → e+ e- Σ+ → n π+
4 He → 3He p π-
Λ
Ds- → K+ K- π- Ω- → Ξ0 π-
ϕ → µ+ µ- 4 He → 3He p π+
Λ ̅
Λc+→ p K- π+ Σ̅- → n̅ π- Ω̅+ → Ξ̅0 π+ 5 He → 4He p π-
Λ
Λ̅c- → p̅ K- π+ Σ- → n π- 5 He → 4He p π+
Λ ̅
Gamma Σ̅+ → n̅ π+
γ → e+ e-
Gamma-decays Strange resonances
π0 → γ γ Double-Λ
η →γγ hypernuclei
4
Ξ*0 → Ξ- π+ ΛΛH → 4ΛHe π-
4 H → 3 H p π-
K*+ → K0s π+ Ξ̅*0 → Ξ̅+ π- ΛΛ Λ
Light mesons Ω*- → Ξ- K- π+ K*+ → K+ π0 5 H → 5 He π-
ΛΛ Λ
and baryons Open-charm K*- → K0s π- 4 He → 5 He p π+
Ω̅*+ → Ξ̅+ K+ π- K*- → K- π0 ΛΛ Λ
resonances Σ*+ → Λ π+ K*0 → K0 π0
π+ → µ+ νµ D*0 → D+ π- Σ̅*- → Λ̅ π- Σ*0 → Λ π0
π- → µ- ν̅µ D̅ *0 → D- π+ Σ*- → Λ π- Σ̅*0 → Λ̅ π0
ρ → π+ π- D*+ → D0 π+ Σ̅*+ → Λ̅ π+ K*0 → K+ π- Ξ*- → Ξ- π0 Heavy multi-
Δ0 → p π- D*- → D̅ 0 π- Ξ*- → Λ K- K̅ *0 → K- π+ Ξ̅*+ → Ξ̅+ π0 strange objects
Δ̅0 → p̅ π+ Ξ̅*+ → Λ̅ K+ ϕ → K+ K-
{ΛΛ} → Λ p π-
Δ++ → p π+ Λ* → p K-
Δ̅-- → p̅ π- Λ̅* → p̅ K+ {Ξ0Λ} → Λ Λ
( mbias: 1.4 ms; central: 10.5 ms )/event/core
23 March 2017 Maksym Zyzak, 29th CBM Collaboration Meeting, Darmstadt 3 /15
Ivan Kisel, Uni-Frankfurt, FIAS, GSI WWND, Guadeloupe, 28.03.2018 !10 /18
Figure 3. Block diagram of the KF Particle Finder package. The particle parameters, such as
decay point, momentum, energy, mass, decay length and lifetime, together with their errors are
estimated using the Kalman filter method.
In the package all registered particle trajectories are divided into groups of secondary and
primary tracks for further processing. Primary tracks are those, which are produced directly
in the collision point. Tracks from decays of resonances (strange, multi-strange and charmed
resonances, light vector mesons, charmonium) are also considered as primaries, since they are
produced directly at the point of the primary collision. Secondary tracks are produced by the
short-lived particles, which decay not in the point of the primary collision and can be clearly
separated. These particles include strange particles (Ks0 and Λ), multi-strange hyperons (Ξ and
Ω) and charmed particles (D0 , D± , Ds± and Λc ). After that tracks are combined according to
the block diagram in Fig. 3. The package estimates the particle parameters, such as decay point,
momentum, energy, mass, decay length and lifetime, together with their errors. The package has
a rich functionality, including particle transport, calculation of a distance to a point or another
particle, calculation of a deviation from a point or another particle, constraints on mass, decay
length and production point. All particles produced in the collision are reconstructed at once,
that makes the algorithm local with respect to the data and therefore extremely fast.
In addition, simultaneous reconstruction in the KF Particle Finder of different decay channels
of the same particle, including also decays with a neutral particle in the final state, makes it
4
36th Winter Workshop on Nuclear Dynamics IOP Publishing
Journal of Physics: Conference Series 1602 (2020) 012006 doi:10.1088/1742-6596/1602/1/012006
Clean Probes of Collision Stages
possible to calculate the efficiency of the reconstruction of rare particles and reliably estimate
their systematic errors.
×106 0
Ks σ = 3.6 MeV/c2 ×106 Λ σ = 1.6 MeV/c2 Λ σ = 1.4 MeV/c2
Entries
Entries
Entries
S/B = 42.6 S/B = 89.7 S/B = 8.73
5 -
1 K0s→π+π- Λ→ pπ 500
Λ→pπ+
0 0 0
0.5 0.6 1.1 1.2 1.1 1.2
minv {π+π-} [GeV/c2] minv {pπ-} [GeV/c2] minv {pπ+} [GeV/c2]
×103 -
Ξ σ = 1.9 MeV/c2
+
Ξ σ = 1.7 MeV/c2
-
Ω σ = 2.1 MeV/c2
Entries
Entries
Entries
20 S/B = 14.1 S/B = 10.7 200 S/B = 35.3
20
- - + - -
Ξ →Λπ Ξ →Λπ+ Ω →ΛK
10 100
10
0 0 0
1.3 1.4 1.3 1.4 1.6 1.7 1.8
- -
minv {Λπ } [GeV/c2] minv {Λπ+} [GeV/c2] minv {ΛK } [GeV/c2]
×106 -
Σ σ = 5.9 MeV/c2 ×103 Σ+ σ = 5.5 MeV/c2 ×103 Σ+ σ = 11.1 MeV/c2
Entries
Entries
Entries
S/B = 49.7 S/B = 7.21 20 S/B = 5.81
0.2 40
-
Σ →π-n Σ+→π+n Σ+→pπ0
0.1 20 10
0 0 0
1.1 1.2 1.3 1.1 1.2 1.3 1.1 1.2 1.3
minv {π-n} [GeV/c2] minv {π+n} [GeV/c2] minv {pπ0} [GeV/c2]
5M central AuAu UrQMD events at 10 AGeV with realistic PID
Ivan Kisel, Uni-Frankfurt,
Figure 4. The package provides clean probes of various stages WWND,
FIAS, GSI
FLES of thePuerto Vallarta, 02.03.2020
collision: results of14 /20
the search for short-lived particles are shown for 5M central AuAu UrQMD events at 10 AGeV
with realistic PID.
The use of the Kalman filter at all stages of particle reconstruction allows in many cases to
get rid almost completely of the combinatorial background and to obtain clean sets of particles,
which can serve as probes of various stages of the collision (Fig. 4).
2.5. Deep learning for quark-gluon plasma detection
In addition to the macroscopic inverse approach [8] we investigate the microscopic inverse
approach by using artificial neural networks to classify processes in heavy ion collisions. We
have created two types of neural networks: fully connected (FC) and deep convolutional (CNN)
neural networks. These networks were then used to identify quark-gluon plasma simulated
within the Parton-Hadron-String Dynamics (PHSD) microscopic off-shell transport approach
for central Au+Au collision at a fixed energy.
For FC networks we use a 64-neuron fully-connected hidden layer with batch normalization,
Leaky Rectified Linear Unit (LReLU) activation and dropout. The number of neurons is chosen
empirically and is fixed to allow comparison of FC neural networks with one, two and three
layers. Batch normalisation and dropout are used to reduce overfitting and therefore improve
overall performance. LReLU is used as it performs similarly to the most commonly used Rectified
Linear Unit (ReLU) activation function but avoids dead neuron issues.
The CNN consists of two three-dimensional convolutional layers, each followed by a max
pooling layer, and two sequential fully-connected layers.
5
theory.gsi.de/~ebratkov/phsd-project/
Input (28x20x20x20)
2-layer fully-connected network
Conv3D (32, 3x3x3, LReLU)
Input (28x20x20x20)
: PHSD model Max pooling (2x2x2)
Collision:
September 29, 2019 7:56 WSPC/INSTRUCTION FILE ”Deep learning for Flatten
Conv3D (64, 3x3x3, LReLU)
QGP ● Au+Au
detection”
QGP off on Nuclear
● Central 36th Winter Workshop QGP on
Dynamics
FC (64, bn, LReLU, dropout 0.5)
IOP Publishing
Max pooling (2x2x2)
5000 events 5000 events FC (2, Softmax)
● 31.2A GeV
Journal of Physics: Conference Series 1602 (2020) 012006 doi:10.1088/1742-6596/1602/1/012006
Flatten
QGP off QGP on
How to classify an event? FC (64, bn, LReLU, dropout 0.5)
8 F. Sergeev, E. Bratkovskaya, I. Kisel and I. Vassiliev FC (2, Softmax)
QGP off QGP on
CBM Collaboration Meeting, Kolkata, 01.10.2019 6 /12
Architecture Accuracy
Ivan Kisel, Uni-Frankfurt, FIAS, GSI CBM Collaboration Meeting, Kolkata, 01.10.2019 9 /12
1-layer ~80%
FC NN 5.
Figure Training ~80%
2-layer and validation accuracy
for the FC networks and the CNN.
3-layer ~75%
CNN >90%
Fig. 5. Goalaccuracy
Training and validation is to determine physical
for the FC networks properties
and the CNN. of QCD matter in real time
Ivan Kisel, Uni-Frankfurt,
The FIAS, GSI5
Fig. shows that the accuracy on the validation set rapidly increases WWND, Puerto
for allVallarta, 02.03.2020
four network 17 /20
4. Conclusion
architectures
The results obtained in ourupworktosuggest
the fifth epoch,
that raw whenhidden
data contains the patterns
rise slows down and the curves level off. At the
same
that allow time,network
the neural the precision on thean training
classifiers to discern set using
event simulated continues
the to go up until it reaches 100%, which
transport model with and without the quark-gluon plasma formation model. Out
suggests that overfitting occurs after the fifth epoch. Nevertheless, the fully-connected networks
of four architectures that included several fully-connected networks as well as a
reachneural
convolutional 80%network
precision while
the latter the
showed theconvolutional
best performance. neural network attains the best performance of more
than 90% accuracy.
Acknowledgments
3. Express
Fedor Sergeev reconstruction
is thankful to andStudent
the International Summer analysis
Programinat STAR
GSI-
FAIR for the opportunity to participate in the Summer School in 2019.
The STAR (Solenoidal Tracker At RHIC) experiment [9] at the RHIC (Relativistic Heavy Ion
The work was supported in part by the Helmholtz International Center for FAIR
(HIC forCollider) facility
FAIR), the Hessen Stateof the of
Ministry Brookhaven National
Higher Education, Laboratory
Research and the (BNL, USA) is designed to study
nuclearandmatter
Arts (HMWK), under
the Federal extreme
Ministry conditions
of Education of relativistic
and Research (BMBF), heavy ion collisions, including hadron
Germany.production and search for signs of quark-gluon plasma formation and its properties.
Very important for RHIC is the possibility to collide ions, covering the range of baryon
References
chemical potential (µB ) from 20 to 420 MeV, which corresponds to a wide range of energies
1. K. Fukushima and T. Hatsuda, Rept. Prog. Phys. 74 014001 (2011).
2. The (so
CBM called Beam
collaboration. EnergyetScan,
(T. Ablyazimov al.), Eur.BES). The BES results once again confirmed evidence of QGP
J. Phys. A
√ 53, 60 (2017).
discovery in the upper RHIC energy sN N = 200 GeV. The results of the search for the critical
point and the first-order phase boundary have narrowed the region of interest to collision energies
√
below sN N = 20 GeV. Thus, the phase II of the Beam Energy Scan (BES-II), scheduled for
√
2019-2020 [10], covers the energy range sN N
36th Winter Workshop on Nuclear Dynamics IOP Publishing
Journal of Physics: Conference Series 1602 (2020) 012006 doi:10.1088/1742-6596/1602/1/012006
used in HLT and its integration into the official STAR repository for use in the standard physics
analysis is currently in progress.
BES-II: eXpress+Standard Data Production and Analysis
2019, 2020 DAQ Disk Tape
? „good“
tHLT = tDAQ + 1s HLT Tape
tHLT = tDAQ + 1s
xHLT oRCF
Disk xCalibration Disk
txCal = tDAQ + 1w
txCal = tDAQ + 1h
DB
30% 70%
txProd = txCal + 1w xProduction xProduction
txProd = txCal + 1d
CA Track Finder CA Track Finder
txPhys = txProd xPhysics xPhysics xStorage
txPhys = txProd
KF Particle Finder KF Particle Finder
Disk
tCal = tRun + 1m DB Calibration
Disk
tProd = tCal + 6m Production Tape
StiCA Track Finder
Disk
tRun tCal tProd tPhys
tBES-II = 2020 + 6 m + 1 h + 1 d + 3 m = 6÷9 m = 2020 Physics
time
tPhys = tProd + 6m
txPhys = txProd + 3m tBES-II = 2020 + 6 m + 1 m + 6 m + 6 m = 1÷2 y = 2020÷2021
PWG Standard,
tBES-II = 2020 + 6 m + 2 m + 1 y + 1 y = 2÷3 y = 2022÷2023 KF Particle, KF Particle Finder
Ivan Kisel, Uni-Frankfurt, FIAS, GSI WWND, Puerto Vallarta, 02.03.2020 19 /20
Figure 6. The HLT express and the standard data production and analysis workflows.
The use of the CA track finder and the KF particle finder in online extends significantly
the functionality of HLT (Fig. 6). The standard calibration, production and analysis remain
unchanged. HLT starts the calibration procedure as soon as data become available. The express
chain makes possible physics analysis of the data as soon as the calibration is reasonable. It
unifies approaches in extended (x)HLT and online (o)RCF to speed up the express workflow,
and combines high competence of xHLT and oRCF experts involved in online operation. In
addition, it provides physics working groups with instant and uncomplicated access to the data,
like picoDST etc.
With the express calibration and alignment one can reconstruct hyperons with high
significance and low level of background, as it is shown in Fig. 7. Hyperons are clearly seen
at all BES-II energies: 3, 3.2, 3.9, 7.7, 9.1, 14.5, 19.6, 27 GeV. In addition, high significance
allows extraction of spectra.
4. Conclusion
The CBM experiment with 107 input rate will require the full event reconstruction and physics
analysis of the experimental data online. As the same HPC farm will be used for offline and online
processing of experimental data, the main reconstruction and analysis algorithms will work both
offline and online. Errors and insufficient accuracy in online data processing, physics analysis
or selection of interesting collisions by the reconstruction algorithms will lead to complete loss
of all experimental data, since only the incorrectly selected data will be stored in this case.
Therefore only immediate comparison of the results of online analysis with the predictions of
theoretical models using ANNs can guarantee the proper operation of the whole experiment. It
has been demonstrated, that the core algorithms of the FLES package, the Cellular Automaton
for searching for particle trajectories (100 µs/core/track) and the Kalman Filter to estimate
their parameters (0.5 µs/core/track), have a very high level of intrinsic parallelism for their fast
736th Winter Workshop on Nuclear Dynamics IOP Publishing
Journal of Physics: Conference Series 1602 (2020) 012006 doi:10.1088/1742-6596/1602/1/012006
BES-II: xHyperons
200M AuAu events at 14.5 GeV, 2019 BES-II express production
6 6 3
Entries ×10 M = 1116.0 MeV/c2 σ = 1.7 MeV/c2 ×10 M = 1322.3 MeV/c2 σ = 2.1 MeV/c2 ×10 M = 1672.7 MeV/c2 σ = 2.2 MeV/c2
Entries
Entries
4
20 S/B = 15.1 S/ S+B = 7468.9 S/B = 9.16 S/ S+B = 1046.5 S/B = 3.62 S/ S+B = 73.9
- 0.5 - - - -
Λ→ pπ Ξ →Λπ Ω →ΛK
10 2
0 0 0
1.1 1.15 1.3 1.35 1.65 1.7 1.75
- -
minv {pπ-} [GeV/c2] minv {Λπ } [GeV/c2] minv {ΛK } [GeV/c2]
6 3 3
×10 M = 1116.0 MeV/c2 σ = 1.6 MeV/c2 ×10 M = 1322.4 MeV/c2 σ = 2.2 MeV/c2 ×10 M = 1672.9 MeV/c2 σ = 2.3 MeV/c2
Entries
Entries
Entries
2 S/B = 7.36 S/ S+B = 2290.1 S/B = 14.4 S/ S+B = 475.7 S/B = 7.35 S/ S+B = 52.8
100 + +
+
Λ→pπ Ξ →Λπ +
Ω →ΛK+
1
1
50
0 0 0
1.1 1.15 1.3 1.35 1.65 1.7 1.75
minv {pπ+} [GeV/c2] minv {Λπ+} [GeV/c2] minv {ΛK+} [GeV/c2]
• With the express calibration and alignment we reconstruct hyperons with high significance and low level of background.
Figure 7. Online
• Hyperons search
are clearly seen for
at all hyperons
BES-II energies:on 200M
3, 3.2, 3.9, 7.7,AuAu
9.1, 14.5, events at 14.5 GeV (2019 BES-II express
19.6, 27 GeV.
production).
• High significance allows extraction of spectra.
Ivan Kisel, Uni-Frankfurt, FIAS, GSI WWND, Puerto Vallarta, 02.03.2020 19 /20
and efficient implementation on many-core CPU/GPU architectures. The KF particle finder
package with more than 150 decay channels implemented (100 µs/core/decay) is a common
platform for offline physics analysis and for real-time express analysis at 107 interaction rate in
CBM.
Adaptation of the FLES algorithms within the FAIR Phase-0 program to the STAR
experiment with its excellent detector performance, high quality experimental data and a well
established reconstruction chain is the first and successful step in preparing the FLES algorithms
for reconstruction and analysis of CBM real data at Day-1. Use of the CA track finder and KF
particle finder developed in the CBM experiment can be beneficial for other experiments as the
experimental heavy ion physics becomes more and more challenging.
References
[1] FAIR — Facility for Antiproton and Ion Research. Green Paper. The Modularized Start Version. GSI. October
2009.
[2] CBM Collaboration, Compressed Baryonic Matter Experiment, Tech. Stat. Rep., GSI, Darmstadt, 2005; 2006
update.
[3] I. Kisel, Event reconstruction in the CBM experiment, Nucl. Instr. and Meth. A566 (2006) 85-88.
[4] I. Kisel, I. Kulakov and M. Zyzak, Standalone first level event selection package for the CBM experiment,
IEEE Trans. Nucl. Sci. 60 (5) (2013) 3703-3708.
[5] V. Akishina and I. Kisel, Online 4-dimensional reconstruction of time-slices in the CBM experiment, IEEE
Trans. Nucl. Sci. 62 (6) (2015) 3172-3176.
[6] V. Akishina, “4D Event Reconstruction in the CBM Experiment”, Dissertation thesis, Goethe university,
Frankfurt am Main (2017).
[7] M. Zyzak, “Online Selection of Short-Lived Particles on Many-Core Computer Architectures in the CBM
Experiment at FAIR”, Dissertation thesis, Goethe university, Frankfurt am Main (2016).
[8] I. Kisel, Event Topology Reconstruction in the CBM Experiment, J. Phys. Conf. Ser. 1070 (2018), 97.
[9] K.H. Ackermann et al. (STAR Collaboration), “STAR detector overview”, Nucl. Instr. Meth. A499 (2003)
624.
[10] STAR Collaboration, “Studying the Phase Diagram of QCD Matter at RHIC”, 01 June 2014.
https://drupal.star.bnl.gov/STAR/files/BES WPII ver6.9 Cover.pdf
8You can also read
