stochastic methods beyond the independent particle picture denis lacroix ipn-orsay collaboration: s....
TRANSCRIPT
Stochastic methods beyond the independent particle picture
Denis Lacroix IPN-Orsay
Collaboration: S. Ayik, D. Gambacurta, B. Yilmaz, K. Washiyama,G. Scamps
Outline:
Mean-field dynamics: advantages, limitations and recent progress
Stochastic mean-field with initial fluctuations
Two-body effects through quantum jump
Quantum Monte-Carlo approach to the N-body problem
Deterministic approach beyond the single-particle picture
Nuclear Time-dependent Density Functional Theory
two-body
three-body
one-body
Mean-field: (DFT/EDF)
“Simple” Trial state:
Self-consistentMean-field
-Kim, Otsuka, Bonche, J. Phys.G23, (1997).-Nakatsukasa and Yabana, PRC71, (2005).-Maruhn, Reinhard, Stevenson, Stone, Strayer, PRC71 (2005).-Umar and Oberacker, PRC71, (2005).-Simenel, Avez, Int. J. Mod. Phys. E 17, (2008).-Washiyama, Lacroix, PRC78 (2008)
Current status
Typical examples of application
Courtesy C. Simenel
Ion-Ion collisionsCollective motion
System size particles
Interaction (hard-core, spin orbit, tensor, 2, 3-body… )
Recent progress
Relevant space
irrel
evan
t sp
ace
Mean-field
exact
Selection of few degrees of freedom
Find closed equation for relevant DOF
Mean-Field strategy
andActual trend: Inclusion of pairing
Y. Hashimoto and K. Nodeki, arXiv:0707.3083.B. Avez et al Phys. Rev. C 78, 044318 (2008).S. Ebata et al, Phys. Rev. C 82, 034306 (2010).I. Stetcu et al, Phys. Rev. C 84, 051309 (2011).G. Scamps et al, Phys. Rev. C 85, 034328 (2012).G. Scamps et al Phys. Rev. C 87, 014605 (2013).
(see G. Scamps talk)
Missing quantum effects
No tunneling
Pote
ntial
Ene
rgy
Surf
ace
Collective variable
Absence of quantum tunneling in collective spaceMissing quantum fluctuationIn collective space
Wrong dynamics close to a symmetry breaking point
Mean-field stays there
Mean-Field
Underestimation of dissipation
DL, Ayik, Chomaz, Prog. Part and Nucl. Phys. (2004)
Collective motionMean-field
Assié and Lacroix, PRL102 (2009)
Some deterministic Beyond Mean-Field Approach
Relevant space
irrel
evan
t sp
ace
Mean-field
exact
Enlarging the space of relevant DOF
Time Dependent Density MatrixTime Dependent Coupled ClusterIllustration: BBGKY
Difficulty
Some deterministic Beyond Mean-Field Approach
Relevant space
irrel
evan
t sp
ace
Mean-field
exact
Enlarging the space of relevant DOF
Time Dependent Density MatrixTime Dependent Coupled Cluster
Reproject the effect of irrelevant DOF
Extended TDHF
Theories beyond the mean-field
One Body space
<A1>
<A2>
<B>Exact evolution
Mean-field
Miss
ing
info
rmati
on
Y. Abe et al, Phys. Rep. 275 (1996)D. Lacroix et al, Progress in Part. and Nucl. Phys. 52 (2004)
Short time evolution
Approximate long time evolution+Projection
Correlation
with
Propagated initial correlation
Quantum zero point motion
Dissipation
projected two-body effect
MF Fluctuations
Dynamics beyond mean-fieldNon-Markovian effects
with
Non-Markovian master equation
Ave
rag
e p
osi
tio
n
Occupation number evolution
Occ
up
atio
n n
um
ber
s
DL, Chomaz, Ayik, Nucl. Phys. A (1999).
1D
Example: two interacting fermions in 1dimension
Difficulty: memory effect!
Starts to look like TDDMFT with memory
Strategy of stochastic methods tackling the N-body problem
Question: Is it possible to recover some of the quantum mechanics aspects by considering an ensemble of independent mean-field trajectories?
Quantum Monte-Carlo
Stochastic TDHF
Stochastic Mean-Field
Correlations that built up in time Direct NN collisions
Initial fluctuations
All Correlations
D. Lacroix and S. Ayik EPJA Review (in preparation)
We assume that the residual interaction can be treated as an ensemble of two-body interaction:
Statistical assumption in the Markovian limit :
Weak coupling approximation : perturbative treatment
Residual interaction in the mean-field interaction picture
Reinhard and Suraud, Ann. of Phys. 216 (1992)
GOAL: Restarting from an uncorrelated state we should:
2-interpret it as an average over jumps between “simple” states
1-have an estimate of
Markovian limit, quantum-diffusion and stochastic Schrödinger Equation
{t t+Dt
Rep
licas
Collision time
Average time between two collisions
Mean-field time-scale
Hypothesis :
Average Density Evolution:
Time-scale and Markovian dynamic
with
Initial simple state
One-body densityMaster equation
step by step
2p-2h nature of the interaction
with
Separability of the interaction
Dissipation contained in Extended TDHF is included
The master equation is a Lindblad equation
Associated SSE Lacroix, PRC73 (2006)
Dissipation: link between Extended TDHF and Lindblad Eq.
SSE on single-particle state :
with
time (arb. units)
wid
th o
f the
co
nden
sate
mean-field
average evolution
Condensate size
N-body density:
1D bose condensate with gaussian two-body interaction
The numerical effort is fixed by the number of Ak
r
r(r)
(a
rb. u
nits
)
t=0t>0
mean-field
average evolution
Density e
volutio
n
Application to Bose-Einstein condensates
Strategy to construct a stochastic mean-field theory
Ayik, Phys. Lett. B 658, (2008).
MF
Collective phase-space Quantum fluctuations
The dynamics is described by a set of mean-field
evolutions with random initial conditions
Mean-Field theory at all time
Stochastic Mean-Field
at all time
Constraint:
The stochastic mean-field (SMF) concept applied to many-body problem
Ayik, Phys. Lett. B 658, (2008).
MF
Collective phase-space Quantum fluctuations
The dynamics is described by a set of mean-field
evolutions with random initial conditions
The average properties of initial sampling should identify with properties of the mean-field.
SMF in density matrix space
SMF in collective space
Description of large amplitude collective motion with SMFThe case of spontaneous symmetry breaking
Lipkin Model
e
See for instance : Ring and Schuck book Severyukhin, Bender, Heenen, PRC74 (2006)
p=1 p=2 … p=N
N=40 particlesExact dynamics
Mean-fieldis stationary
Description of large amplitude collective motion with SMFThe stochastic mean-field solution
Initial condition
One-body observables
ExactSMF
Lacroix, Ayik, Yilmaz, PRC 85 (2012)
Formulation in quasi-spin space
Description of large amplitude collective motion with SMFThe stochastic mean-field solution
Formulation in quasi-spin space
Initial condition
Fluctuations
Lacroix, Ayik, Yilmaz, PRC 85 (2012)
The stochastic mean-field (SMF) with pairing
Following the general strategy:
TDHFB with initial fluctuations
Setting the initial fluctuations
Quasi-particle occupation
Lacroix, Gambacurta, Ayik, PRC 87 (2013)
Requires to fix:
and
Illustration with the pairing Hamiltonian
Quasi-spin operators
: degeneracy
Occupation probabilities
Pair creation/annihiliationoperators
TDHFB equations
Illustration with the pairing Hamiltonian
Departure from the independent Particle picture
Quantum Fluctuations
weak coupling
strong coupling
Comparison between deterministic and stochastic methods
Strong coupling
Exact
Stochastic MF
Evolution of both One- and two-body densities
SMF might even be better than BBGKY hierarchy truncation
Open systems
One Body space
<A1>
<A2>
<B>
Exact evolution
Mean-field
Mis
sing
in
form
atio
n
Brownian motion
N-body
Towards Exact stochastic methods for N-body and Open systems
Environment
System(one-body)
(others)Environment
System
Exact stochastic formulation
More insight in mean-field dynamics:
Exact state Trial states
{The approximate evolution is obtained by minimizing the action:
Included part: average evolution
exact Ehrenfestevolution
Missing part: correlations
Environment
System
Complexself-interacting
System
Hamiltonian splitting
SystemEnvironment
One Body space
<A1>
<A2>
<B>
Exact evolution
Mean-field
Mis
sing
in
form
atio
n
Relevant degrees of freedom
The idea is now to treat the missing informationas the Environment for the Relevant part (System)
Mean-field from variationnal principle
<A1>
Exact evolution
<A2>
<A1A 2
>- <
A 1A 2
> MF
with
D. Lacroix, Ann. of Phys. 322 (2007).
…Mean-field
Mean-field level
Mean-field + noise
Theorem :One can always find a stochastic process for trial states such thatevolves exactly over a short time scale.
Valid for
orIn practice
Existence theorem : Optimal stochastic path from observable evolution
t>0Mean-field evolution:
x
t>0
Reduction of the information:I want to simulate the expansion with Gaussian wave-function having fixed widths.
t=0
with
Relevant/Missing information:
Relevant degrees of freedom
Missing information
Trial states
Coherent states
illustration: simulation of the free wave spreading with “quasi-classical states”
Stochastic c-number evolution from Ehrenfest theorem
Densities
with
Nature of the stochastic mechanics
with
the quantum wave spreading can be simulated by a classical brownian motion in the complex plane
x
x
tim
e
x
fluctuationsmean values
Guess of the SSE from the existence theorem
The method is general.the SSE are deduced easily
Ehrenfest theorem BBGKY hierarchy
D. Lacroix, Ann. Phys. 322 (2007)
Starting point:
with
Observables
Fluctuations
with
Stochastic one-body evolution
The mean-field appears naturally and the interpretation is easier
extension to Stochastic TDHFB DL, arXiv nucl-th 0605033
Occ
upati
on p
robabili
ty
time
two-level systemBosons
but…
the numerical effort can be reduced by reducing the number of observables
unstable trajectories
SSE for Many-Body Fermions and bosons
Stochastic MF
FluctuationDissipation
Stochastic TDHF
FluctuationDissipation
Exact QD
Everything
Mean-field
FluctuationDissipation
variational QD
Partially everything
Numerical issues
FlexibleFlexible Fixed Fixed
Approximate evolution
Summary, stochastic methods for Many-Body Fermionic and bosonic systems
Numericalinstabilities