Correlations within the Non-Equilibrium Greens Function

Method

M.H. Mahzoon1,a) , P. Danielewicz1,2 and A. Rios3

1
Department of Physics and Astronomy, Michigan State University, East Lansing, MI 48824-1321, USA
2
National Superconducting Cyclotron Laboratory, Michigan State University, East Lansing, MI 48824-1321, USA
3
Department of Physics, University of Surrey, Guildford, Surrey GU2 7XH, United Kingdom

a)
[email protected]

Abstract. Non-equilibrium Greens Function (NGF) method is a powerful tool for studying the evolution of quantum many-body
systems. Different types of correlations can be systematically incorporated within the formalism. The time evolution of the single-
particle Greens functions is described in terms of the Kadanoff-Baym equations. The current work initially focuses on introducing
the correlations within infinite nuclear matter in one dimension and then in a finite system in the NGF approach. Starting from
the harmonic oscillator Hamiltonian, by switching on adiabatically the mean-field and correlations simultaneously, a correlated
state with ground-state characteristics is arrived at within the NGF method. Furthermore the use of cooling to for improving the
adiabatic switching is explored.

Introduction

The ultimate goal of this project is to investigate nuclear collisions following the Non-equilibrium Greens Function
(NGF) method within space-time representation. To initialize a many-body system, the correlated ground-state needs
to be constructed, static under real-time evolution. The latter can be ensured by employing the same infrastructure
for correlating the system as for the evolution, and by adiabatically switching on the correlations together with the
mean field. We originally trap the finite system in a harmonic oscillator potential and then adiabatically switch off that
potential while ramping up the mean-field in combination with a Gaussian residual interaction. The conversion of the
interactions can be aided with a transient cooling friction term.

Kadanoff-Baym Equations

One-body information within a many-body system is accounted for in terms of the one-body density matrix. In the
mean-field approximation, the dynamics of a many-body system can be cast in a self-consistent and self-contained
manner in terms of that one-body density matrix. Provided that the effects of short-range correlations are short-lived,
the dynamics of a many-body system can be approximately described in a self-contained manner in terms of the
single-particle Greens functions that are two-time extensions of the density matrix, cf. [1]:
G< (x1 t1 ; x10 t10 ) = ih (x10 t10 ) (x1 t1 )i , (1)
G> (x1 t1 ; x10 t10 ) = ih(x1 t1 ) (x10 t10 )i . (2)
Here, (x t) annihilates a particle at location x at time t and the expectation value hi is with respect to an uncorrelated
state at some starting time t0 . With space and time arguments lumped together as (x1 t1 ) 1, to simplify notation, the
Greens functions can be shown to satisfy the Kadanoff-Baym (KB) equations [2]:
~2 2
Z Z t1 h
d1 > (1 1) < (1 1) G (1 10 )
i
i~ + G (1 10
) = dx HF (1 1) G
( 110
)+
t1 2m x1 1
2


t0
Z t10 (3)
d1 (1 1) G> (1 10 ) G< (1 10 ) ,
h i

t0

~2 2
Z Z t1
d1 G> (1 1) G< (1 1) (1 10 )
h i
i~ 0 + = +
0 0

G (1 1 ) dx G (1 1) HF (1 1 )
t1 2m x1 1
02


t0
Z t10 (4)
d1 G (1 1) > (1 10 ) < (1 10 ) .
h i

t0

Here, HF and are the self-energies that can be approximated in terms of the single-particle Greens functions G .
The self-energy HF accounts for the most direct mean-field effects and the energies incorporate the effects of
correlations. When the self-energies are consistently approximated, the KB equations preserve the conservation laws
of particle number, momentum and energy.
In the past, we explored the mean-field dynamics in the context of employing the KB equations for describing nuclear
collisions [3]. We examined arriving at the mean-field ground-state through adiabatic switching and collisions of
nuclear slabs. Here, we investigate the process of switching on the correlations for infinite matter and for slabs.

Importance of Initial-State Preparation

In the previous sections the KB equations which govern the time evolution of a many-body system were introduced.
Solving the latter equations yields one-body information about the system and also many-body provided the effects
of correlations are short-lived. In studying collisions, nuclear systems need to be initialized consistently with the KB
equations used to follow the collisions. Thus, if an uncorrelated initial state is naively followed with KB equations
incorporating correlations, then that system may spew particles and even explode and/or violently oscillate ahead of
any collision. In the previous work [3], we employed adiabatic switching to arrive at mean-field initial states consistent
with a mean-field version of the KB equations. Examples where adiabatic switching was used for constructing corre-
lated many-body states include Refs. [4, 5]. In [1], imaginary-time evolution was employed, requiring development
of a separate computational infrastructure, but equivalent to the adiabatic switching under proper circumstances.

Employed Interactions

In the Hartree-Fock part of the self-energy, we use a local Skyrme-type interaction such as in Ref. [3]. For the self-
energies , we employ the so-called self-consistent Born diagram illustrated in Fig. 1, yielding
Z
dp1 dp2
(p t; p0 t0 ) = V(p p1 ) V(p0 p2 ) G (p1 t; p2 t0 ) (p p1 t; p0 p2 t0 ) , (5)
2 2
where Z
dp1 dp2
(p t; p0 t0 ) = G (p1 t; p2 t0 ) G (p2 p0 t0 ; p1 p t) . (6)
2 2

In the semiclassical low-density limit, the self-energies represent phase-space feeding and depletion rates where
cross sections are described in the Born approximation in terms of the employed residual interaction. To describe
semi-quantitatively 3D rates within the 1D calculation, we employ a residual interaction modulated by a Gaussian:

V(p) =V0 2 p2 e p /4 ,
2 2

x2 x2 /2 (7)
!
V(x) =V0 1 2 2 e .

The constants V 0 and are adjusted to yield reasonable cross sections and quasiparticle content of single-particle
states in the ground state.

b


2
n p

p n n

n
(p) (p) 1

a

FIGURE 1. Direct Born diagram contributions to the self energy for neutrons (protons).

Adiabatic Transformation of Interactions

In the case of a finite system we generally start with one that is uncorrelated and trapped in a harmonic oscillator
potential U0 (x) = 12 m2 x2 . One way to choose the optimal frequency for the potential is by demanding that the
r.m.s. radius for the trapped matter is the same as for a self-bound uniform slab at normal density [3, 6]. In the case
when only the self-consistent mean field HF is active for the self-bound slab, the adiabatically converted potential
may be represented as
Ut (x t) = F(t) U0 (x) + [1 F(t)] HF (x t) . (8)
Here F(t) is a switching function that changes between 1 and 0 around some switching time t s .
In the literature there had been discussions of optimal choice of switching functions f for adiabatic switching [7]. If
the switching interval is from t0 to t1 , we ensure that F changes between 1 and 0 there by taking
f (t t s ) f (t1 t s )
F(t) = , (9)
f (t0 t s ) f (t1 t s )
and our standard function f is [3]
1
f (t) = (10)
1 + et/
where controls the pace of adiabatic switching.
Given sufficiently long switching time, desirable properties of an interacting initial state for a collision can be reached,
such as low net energy, sharply ending density distribution confirming low temperature, and little time dependence.
Due to the 2-time structure of NGF, though, a prolonged switching on of the interactions can be quite costly computa-
tionally. With this, it is of advantage to arrive at the desirable properties over as short time as possible, optimizing the
adiabatic path. In Ref. [7] different switching functions f were proposed that we test in Fig. 2 in the case of switching
of the pure mean field of Ref. [3] for a nuclear slab. We show there system size in terms of h|x|i in the case of the
same U0 and the same final mean-field. Short switching times lead to oscillating mean-field states with the amplitude
of oscillations diminishing as the pace of switching slows down. It is apparent that our original function ("this work"),
Eq. (10), performs well against other proposed switching functions [7], allowing to arrive at the desirable properties
of the interacting initial state in a shorter time than for the other proposed functions.

Infinite Nuclear Matter

While working in 1D, we want to tie as closely as possible the properties of 1D systems to those of slabs in 3D, uniform
in two directions. Through the use of Hugenholtz-van Hove theorem, we have shown how to transcribe density and
energy between the systems. For the density the relation is
r 2 1/3
5 n0
n3D = n1D , = , (11)
3 62
 1.8
linear switching function
this work
polynomial switching function of rank 3
1.6

system size [fm]

1.2

0.8
0 20 40 60 80 100 120 140 160 180 200
time [fm/c]
FIGURE 2. Time evolution of the size of an A = 4 system for different types of switching functions, with the switching active in
the time interval of [50, 150] fm/c.

where n1D (x, t) = iG< (x t; x t), = 4 is spin-isospin degeneracy and n0 = 0.16 fm3 is the normal density.
In deciding on interaction parameters and exploring system properties, we first examine infinite isospin-symmetric
nuclear matter. We switch on the residual interactions adiabatically, just as in the case of mean field [3]. The residual
interactions in contribute to net energy, besides the mean-field interactions, in terms of the correlation energy.
Details, besides on interaction parameters depend on spatial grid employed in calculations, that introduces a high-
momentum cut-off. In production calculations we employ a mesh with the separation of dx = 0.67 fm. The mesh
implies what time steps can be used in the time integration, with a finer mesh requiring shorter time steps.
We supplement the residual interactions with mean field so that the net energy, consisting of kinetic, correlation and
mean-field energy, minimizes at n0 , with a sensible curvature around the minimum. That energy, in terms of points for
sample densities, and in terms of a line for a smooth interpolation of the results, is shown in Fig. 3. The numerically
relatively low energy values are tied to the fact that the energy in 1D excludes the kinetic energy associated with
transverse degrees of freedom, there in the 3D matter, but frozen here.

Finite Systems

Upon deciding on the combination of the residual interaction and the mean field, we proceed to studying the finite
systems that are started out in a harmonic trap. For efficient switching, the desired initial state for correlated evolution
should be arrived at, no matter what the details are of the starting Hamiltonian, at least within some range of those
details. Figure 4 shows the evolution of net energy during adiabatic switching when starting an A = 8 system in the
ground state of HO potential (filled orbitals N = 0 and 1) at different frequencies in the units of the nominally
optimal frequency . It is apparent that the systems arrive at about the same energy no matter how the systems are
started.
Near the minimum the energy per nucleon is a relatively forgiving observable and the geometric characteristics can
become more telling on the arrival to the ground state irrespectively of the starting conditions. In Fig. 5 we show
the evolution of A = 12 (N = 0, 1, 2) system size in terms of h|x|i, in the course of the adiabatic transformation of
interactions, for a couple of starting frequencies. It is seen that though = 1.5 yields a worse match between the
 0
Etot
5

10

15
E/A [MeV]
20

25

30

35

40
0.00 0.05 0.10 0.15 0.20 0.25
n [fm3]

FIGURE 3. Net energy per nucleon as a function of 3D density in cold correlated nuclear matter. Symbols represent values arrived
at sample densities and the line represents a smooth interpolation of the results.

initial and final sizes, than = , it allows to reach the desired final state a tad earlier with a more static final state.

Cooling Friction

Obviously, there may be more ways of transforming interactions between the initial and final states, leading to the
ground-state characteristics for the latter, than through a linear combination of the initial and final interactions [8, 9].
In particular, in [8] it was proposed to supplement the potential in the equations of motion by a term breaking the time-
reversibility and proportional to the time derivative of density in any particular representation. Figure 6 illustrates the
impact of a friction term proportional to time derivative of density in space representation, when gradually switching
on the interactions for an A = 4 (N = 0) system. Outside of the switching interval, the friction is turned off. It is
apparent that the friction allows to arrive at a state that is more compact and more static than without the friction.

The evolution of size in Fig. 6 with friction is complemented with Fig. 7 that shows the changes in the density profile
during adiabatic transformation of the interactions, on the left, and the evolution of the average density, on the right.
Finally, in Fig. 8 the evolution of net energy per nucleon is shown, for the A = 4 slab, together with the breakdown
of the energy in the kinetic and potential parts. The kinetic energy rises as a function of time, as occupations inside
the Fermi sphere drop down and momentum distribution develops a tail. The energy content in that tail depends on
resolution, though, that is dx.

In careful examination of Fig. 6 and the right panel in Fig. 7, a slow gradual change in the late-time size and average
density can be seen besides oscillations. These indicate that the system temperature is not completely zero. Artifacts
such as slight temperature and mild oscillations are something that one will have to live with in the collision simu-
lations, with a hope that they might not affect the reaction dynamics much if the reaction energy is elevated and the
reaction progresses rapidly.
 40

30 1.5
2
2.5
E/A [MeV] 20

10

-10

-20

-30
0 20 40 60 80 100
time [fm/c]
FIGURE 4. Net energy per nucleon as a function of time for the A = 4 system initialized in the ground state particle time evolution
for different initial HO potential frequencies .

Next Step: Isospin Dynamics

The next natural step in the progression of nuclear NGF investigations is the investigation of collective oscillations
in the presence and absence of correlations. The low-lying and relatively weakly damped nuclear oscillations are
those of protons against neutrons. For study of those oscillations in a systematic way, we need to describe the nuclear
symmetry energy in a controlled way. Contributions to the symmetry energy stem from the Fermi-gas kinetic energy,
the correlation energy and Hartree-Fock contribution. The Hartree-Fock contribution to the net symmetry energy is
simply
(nn n p )2
Z
EHF =
a
dx S HF (n) . (12)
n
The corresponding contributions to the Hartree-Fock self-energy are then

(nn n p )2
" #
p(n)
HF = S HF (n) = 2 S HF + 2 (nS 0HF S HF ) , (13)
n p(n) n

where = (nn n p )/n and the upper r.h.s. sign pertains to protons and lower to neutrons. The correlation contribution
stems from asymmetry in the blocking for protons and neutrons and the corresponding difference developing for
occupations of neutron and proton states, even when residual interaction is assumed to be isoscalar, cf. Eq. (7) and
Fig. 1. At a later stage this may need to be modified.

Conclusions

We have carried out first steps towards exploring the dynamics of correlated nuclear systems in NGF. We prepared
correlated infinite and finite nuclear systems in 1D through adiabatic switching. We explored the use of different
 3.6
1
1.5
3.4

system size [fm]

3.0

2.8

2.6

2.4

2.2

2.0
0 20 40 60 80 100
time [fm/c]
FIGURE 5. System size (in fm) as a function of time in the course of adiabatic transformation of interactions for different initial
HO potential frequencies in the units of optimal frequency from analytic considerations.

switching functions, initial conditions and cooling friction, arriving at reasonably stationary and cold states for initiat-
ing the dynamics of interest. In parallel, we developed a suitable combination of mean-field and residual interactions
for describing the initial states and the dynamics.

ACKNOWLEDGMENTS

The authors benefited from discussions with Brent Barker and Hao Lin. This work was supported by the U.S. National
Science Foundation under Grant PHY-1520971.

REFERENCES
[1] P. Danielewicz, Annals of Physics 152, 239304 (1984).
[2] L. P. Kadanoff and G. Baym, Quantum Statistical Mechanics (Benjamin, New York, 1962).
[3] A. Rios, B. Barker, M. Buchler, and P. Danielewicz, Annals of Physics 326, 12741319 (2011),
arXiv:1009.0215 .
[4] M. Tohyama, Progress of Theoretical Physics 92, 905908 (1994).
[5] A. Pfitzner, W. Cassing, and A. Peter, Nuclear Physics A 577, 753772 (1994).
[6] P. Ring and P. Schuck, The Nuclear Many-Body Problem (Springer, Berlin, 1980).
[7] M. Watanabe and W. P. Reinhardt, Physical Review Letters 65, 33013304 (1990).
[8] A. Bulgac, F. M. McNeil, R. J. Kenneth, and G. Wlazlowski, (2013), arXiv:1305.6891 .
[9] A. Patra and C. Jarzynski, arXiv:1707.01490 [cond-mat, physics:quant-ph] (2017).
 1.8

1.6
<x> [fm]

1.4

1.2

1.0 no friction
with friction

0 20 40 60 80 100 120
time [fm/c]

FIGURE 6. Effect of the supplemental cooling-friction term on the evolution of system size when transforming the interactions.

t = 0 [fm/c] 0.22 average density

0.20
density [fm3]

0.18
0.15

0.16
0.10

0.05 0.14

0.00
4 3 2 1 0 1 2 3 4 0 20 40 60 80 100 120 140
x [fm] time [fm/c]

FIGURE 7. Time evolution of the density in position space (left) and of the average density (right) for an A = 4 system during and
after the transformation of interactions.
 30

20
Total Energy
Kinetic Energy
10 Potential Energy
E/A [MeV]

-10

-20

-30
0 20 40 60 80 100 120 140
time [fm/c]

FIGURE 8. Evolution of net energy per nucleon for an A = 4 system, and its potential and kinetic components, during and after
the adiabatic transformation of interactions.