Hot and Dense QCD Matter shear viscosity of natures most ideal - - PowerPoint PPT Presentation
A data-driven approach to quantifying the Hot and Dense QCD Matter shear viscosity of natures most ideal liquid Unraveling the Mysteries of the Strongly Interacting Quark-Gluon-Plasma Steffen A. Bass A Community White Paper on the Future of
A Community White Paper on the Future of Relativistic Heavy-Ion Physics in the US
http://www.facebook.com/DukeQCD @Steffen_Bass
Molecule Nucleus Atom Proton/Neutron Quark
(plus anti-particles)
matter (u, d, e)
between quarks via exchange of gluons
Elementary Particles:
solid liquid gaseous by adding/removing heat, phase of matter can be changed between solid, liquid and gaseous Pressure plays an important role for the value of the transition temperature between the phases boiling temperature:
5
Ordinary Matter:
magnetic) interaction
phase-diagram
5
Ordinary Matter:
magnetic) interaction
phase-diagram Phases of QCD matter:
Big Bang the entire Universe was in a QGP state
matter allows to investigate the history of the Universe
temperatures and densities of the early Universe is by colliding beams of ultra-relativistic heavy- ions
The only way to heat & compress QCD matter under controlled laboratory conditions is by colliding two heavy atomic nuclei!
The only way to heat & compress QCD matter under controlled laboratory conditions is by colliding two heavy atomic nuclei!
two more experiments w/ Heavy-Ions:
ALICE experiment @ CERN:
The only way to heat & compress QCD matter under controlled laboratory conditions is by colliding two heavy atomic nuclei!
two more experiments w/ Heavy-Ions:
ALICE experiment @ CERN: typical Pb+Pb collision @ LHC:
characterize matter created in collision typical Pb+Pb collision @ LHC:
Brookhaven National Laboratory: Relativistic Heavy-Ion Collider
Brookhaven National Laboratory: Relativistic Heavy-Ion Collider
BRAHMS)
Brookhaven National Laboratory: Relativistic Heavy-Ion Collider
BRAHMS)
characterize matter created in collision
3+1D Hydro + Boltzmann Hybrid
3+1D Hydro + Boltzmann Hybrid
1x 10-23 s 10 x 10-23 s 30 x 10-23 s
1x 10-23 s 10 x 10-23 s 30 x 10-23 s
nuclei at 99.99% speed of light
1x 10-23 s 10 x 10-23 s 30 x 10-23 s
nuclei at 99.99% speed of light Quark-Gluon-Plasma
1x 10-23 s 10 x 10-23 s 30 x 10-23 s
nuclei at 99.99% speed of light Quark-Gluon-Plasma hadronic final state interactions
1x 10-23 s 10 x 10-23 s 30 x 10-23 s
nuclei at 99.99% speed of light Quark-Gluon-Plasma measurable (stable) particles in detector hadronic final state interactions
1x 10-23 s 10 x 10-23 s 30 x 10-23 s
nuclei at 99.99% speed of light Quark-Gluon-Plasma measurable (stable) particles in detector hadronic final state interactions non-equilibrium early time dynamics viscous fluid dynamics hadronic transport
Principal Challenges of Probing the QGP with Heavy-Ion Collisions:
t + p E ×
p, r, t) =
C( p, r, t)
microscopic transport models based
hybrid transport models:
p(t)− 2T v · ∆t+ (t)∆t
diffusive transport models based
a thermal medium
properties of the medium and a noise term representing random collisions
Tik = εuiuk + P (δik + uiuk)
3δik · u
ς δik · u
(viscous) relativistic fluid dynamics:
(plus an additional 9 eqns. for dissipative flows)
t + p E ×
p, r, t) =
C( p, r, t)
microscopic transport models based
hybrid transport models:
Each transport model relies on roughly a dozen physics parameters to describe the time-evolution of the collision and its final state. These physics parameters act as a representation of the information we wish to extract from RHIC & LHC.
p(t)− 2T v · ∆t+ (t)∆t
diffusive transport models based
a thermal medium
properties of the medium and a noise term representing random collisions
Tik = εuiuk + P (δik + uiuk)
3δik · u
ς δik · u
(viscous) relativistic fluid dynamics:
(plus an additional 9 eqns. for dissipative flows)
but mostly with an offset:
gets compressed and heated up Reaction plane
x z y
but mostly with an offset:
gets compressed and heated up Reaction plane
x z y
elliptic flow (v2):
preferential emission in the reaction plane
2nd Fourier coefficient of angular distribution: v2 Ø vRFD: good agreement with data for very small η/s
experimental data: π/K/P spectra yields vs. centrality & beam elliptic flow HBT charge correlations & BFs density correlations Model Parameter:
shear viscosity initial state pre-equilibrium dynamics thermalization time quark/hadron chemistry particlization/freeze-out
experimental data: π/K/P spectra yields vs. centrality & beam elliptic flow HBT charge correlations & BFs density correlations Model Parameter:
shear viscosity initial state pre-equilibrium dynamics thermalization time quark/hadron chemistry particlization/freeze-out
experimental data: π/K/P spectra yields vs. centrality & beam elliptic flow HBT charge correlations & BFs density correlations Model Parameter:
shear viscosity initial state pre-equilibrium dynamics thermalization time quark/hadron chemistry particlization/freeze-out
→collaboration with Statistical Sciences
Each computational model relies on a set of physics parameters to describe the dynamics and properties of the system. These physics parameters act as a representation of the information we wish to extract from RHIC & LHC.
e s t i m a t e
c a l c u l a t e p a r a m e t e r s calculate observables & compare to data
Model Parameters - System Properties
Experimental Data
Physics Model:
Each computational model relies on a set of physics parameters to describe the dynamics and properties of the system. These physics parameters act as a representation of the information we wish to extract from RHIC & LHC.
model-to-data comparison
Bayesian analysis
Model Parameters - System Properties
Experimental Data
Physics Model:
Posterior Distribution
parameter, integrating out all others
dependence between parameters
Physics Model:
Model Parameters - System Properties
Experimental Data
Gaussian Process Emulator
MCMC
(Markov-Chain Monte-Carlo)
weighted by posterior probability
Bayes’ Theorem
posterior∝likelihood × prior
data, given proposed parameters
after many steps, MCMC equilibrates to calculate events on Latin hypercube
brute force analysis:
brute force analysis:
Need to find techniques that cut down the cpu needed by at least a factor
brute force analysis:
Need to find techniques that cut down the cpu needed by at least a factor
−2 −1 1 2
Output
Random functions
1 2 3 4 5
Input
−2 −1 1 2
Output
Dashed line: mean Band: 2σ uncertainty Colored lines: sampled functions Conditioned on training data (dots)
Gaussian process:
maps inputs to normally distributed outputs
GP as a model emulator:
(Latin hypercube points)
Latin hypercube:
randomized, space-filling points (here: maximin Latin hypercube)
n: number of model parameters
Latin hypercube:
randomized, space-filling points (here: maximin Latin hypercube)
n: number of model parameters
Latin hypercube:
randomized, space-filling points (here: maximin Latin hypercube)
n: number of model parameters
this design:
to interpolate between points
reduce the computational footprint
Nukeserv: 100 TB array at Duke
local desktop at Duke:
compute cluster @ Duke:
XSEDE OSG submit node:
GridFTP protocol
Open Science Grid is used for:
cpu hours
Edison @ NERSC:
Duke QCD workflow:
simultaneously
be computed in 1 day
Edison @ NERSC:
Duke QCD workflow:
simultaneously
be computed in 1 day
Vector of input parameters: x=[p,k,w,(𝜃/s)min,(𝜃/s)slope,(𝜂/s)norm,Tsw]
Bayes’ Theorem: P(x★ |X,Y,yexp) ∝ P(X,Y,yexp| x★)P(x★)
⇒ initial knowledge of x★
⇒ probability of observing (X,Y,yexp) given proposed x★
⇒ probability of x★ given observations (X,Y,yexp)
Vector of input parameters: x=[p,k,w,(𝜃/s)min,(𝜃/s)slope,(𝜂/s)norm,Tsw]
Bayes’ Theorem: P(x★ |X,Y,yexp) ∝ P(X,Y,yexp| x★)P(x★)
⇒ initial knowledge of x★
⇒ probability of observing (X,Y,yexp) given proposed x★
⇒ probability of x★ given observations (X,Y,yexp)
Markov-Chain Monte-Carlo:
Prior: model calculations evenly distributed over full design space
1
1
Prior: model calculations evenly distributed over full design space
1
1
1
1
Posterior: emulator predictions for highest likelihood parameter values
l
l
β
parameter, integrating out all others
showing dependence between parameters
l
l
p≈0: IP-Glasma type scaling
β
parameter, integrating out all others
showing dependence between parameters
l
l
p≈0: IP-Glasma type scaling Tsw⩽Tc
β
parameter, integrating out all others
showing dependence between parameters
l
l
p≈0: IP-Glasma type scaling Tsw⩽Tc
β
temperature-dependent viscosities:
ubouum]
parameter, integrating out all others
showing dependence between parameters
and the smallest specific shear viscosity ever observed
level of sophistication that allows for the determination of the temperature dependent transport coefficients of the QGP
the feasibility and the success of the science!
Trento:
iEbE-VISHNU:
Computer Physics Communications in print, arXiv:1409.8164
UrQMD:
MADAI Collaboration:
Duke Bayesian Analysis Package:
this work has been made possible through support from
HPSS Capabilities:
Problem:
transfer data to/from NERSC
Scaling of analysis with # of observables:
and unnecessary in case of strong correlations
Scaling of analysis with # of observables:
and unnecessary in case of strong correlations
⇒ emulate each PC
Scaling of analysis with # of observables:
and unnecessary in case of strong correlations
⇒ emulate each PC this analysis:
to fall within 1 standard deviation
centrality:
maximum of the posterior probability distributions yield excellent agreement with data: