Computer Simulation of Colloids

Jrgen Horbach

Institut fr Materialphysik im Weltraum, Deutsches Zentrum fr Luft-
und Raumfahrt, Linder Hhe, 51147 Kln
lecture 1
introduction

lecture 2
introduction to the Molecular Dynamics (MD) simulation method

lecture 3
case study: a glassforming Yukawa mixture under shear

lecture 4
introduction to the Monte Carlo (MC) method

lecture 5
case study: phase behavior of colloid-polymer mixtures studied by
grandcanonical MC

lecture 6
modelling of hydrodynamic interactions using a hybrid MD-Lattice
Boltzmann method
Computer Simulations of Colloids

Introduction
(1) colloids as model systems: different interaction potentials and
hydrodynamic interactions


(2) correlation functions: describe the structure and dynamics of
colloidal fluids


(3) outline of the forthcoming lectures
(1) colloids as model systems: different interaction potentials and
hydrodynamic interactions


(2) correlation functions: describe the structure and dynamics of
colloidal fluids


(3) outline of the forthcoming lectures
colloids as model systems
size: 10nm 1m, dispersed in solvent of viscosity q
large, slow, diffusive optical access

can be described classically

tunable interactions, analytically tractable
theoretical guidance

soft shear melting

phenomena studied via colloidal systems:

glass transition, crystallisation in 2d and 3d, systems under shear,
sedimentation, phase transitions and dynamics in confinement, etc.

hard sphere colloids
confirmed theoretically
predicted crystallization
(of pure entropic origin)

study of nucleation and
crystal growth

test of the mode coupling
theory of the glass
transition
hard spheres






soft spheres
volume fraction u
volume fraction, hair length and density
tunable interactions I
charged spheres








super-paramagnetic spheres
B
magnetic field
number density, salt concentration, charge
tunable interactions II
entropic attraction in colloid-polymer mixtures
volume fraction of colloidal particles,
number of polymers, colloid/polymer
size ratio
tunable interactions III
colloid-polymer mixtures may exhibit a demixing transition which is
similar to liquid-gas transition in atomistic fluids

transition is of purely entropic origin
interaction between colloids due to the momentum transport through
the solvent (with viscosity )
hydrodynamic interactions
a
T k
D F
T k
D
i i
tq 6
v
B
0
B
0
= =

) , , , ( with ) (
1
v
1
2 1
) ( ) (
B
i N
N
j
N
j
N
ij
r r r r F r D
T k

=
= =
D
ij
: hydrodynamic diffusion tensor

decays to leading order ~ r
-1
(long-range
interactions!)
choose some effective interaction between the colloidal particles

but what do we do with the solvent?
simulation of a colloidal system
simulate the solvent explicitly

describe the solvent on a coarse-grained level
as a hydrodynamic medium

Brownian dynamics (ignores hydrodynamic
interactions)

ignore the solvent: not important for dynamic
properties (glass transition in hard spheres),
phase behavior only depends on configurational
degrees of freedom
(1) colloids as model systems: different interaction potentials and
hydrodynamic interactions


(2) correlation functions: describe the structure and dynamics of
colloidal fluids


(3) outline of the forthcoming lectures
pair correlation function
r
r small
i
n
p
(r) = number of pairs (i,j) with
] , [ r r r r
ij
A + e
ideal gas:
r r
N
r r r
N
r n A ~ A + =
2 3 3
p
4
2
] ) [(
3
4
2
) ( t
t

fluid with structural correlations:

) ( 4
2
) (
2
p
r g r r
N
r n A = t
density same at the gas ideal an for same the
origin in the atom arbitrary an from distance at density number average
) (
r
r g =
pair correlation function: hard spheres vs. soft spheres
soft sphere fluid: hard sphere fluid:

= =

= + =
i i j
ij
i i j
j i
r
r r
N
r r r
N
r g
2
4
) (
1
) (
1
) (
t
o
o

static structure factor

= =
= =
N
i
N
j
j q i
r q i r r r
1 1
FT
) exp( ) ( ) (

o
microscopic
density variable:
dr
qr
qr
r g r r r q i
N N
q S
N
m
N
n
n m q q
}

= =

+ = = =
0
2
1 1
) sin(
) 1 ) ( ( 4 1 )] ( exp[
1 1
) ( t


static structure factor:
ility compressib isothermal :
) 0 (
B
2
2
T
T
T k
N
N N
q S
k
k =

=
low q limit:
= =
V
N
r) (
homogeneous fluid:
1 ) ( = q S
large q:
typical structure factor for a liquid
1st peak at ~2/r
p
where r
p
measures the periodicity
in g(r)

dense liquid low
compressibility and therefore
small value of S(q0)

in general S(q) appropriate
quantity to extract information
about intermediate and large
length scales

S(q) can be measured in
scattering experiments
van Hove correlation function
generalization of the pair correlation function to a time-dependent
correlation function
self part:

=
k l
l k
r t r r
N
t r G )]) 0 ( ) ( [ (
1
) , (

o

=
k
k k
r t r r
N
t r G )]) 0 ( ) ( [ (
1
) , (
s

o
proportional to the probability that a particle k at time t is separated by
a distance |r
k
(t) - r
l
(0)| from a particle l at time 0
intermediate scattering functions
coherent intermediate scattering function:
) (
) , (
) , (
)]) 0 ( ) ( [ exp(
1
) 0 ( ) (
1

) exp( ) , ( ) , (
q S
t q S
t q F
r t r q i
N
t
N
r d r q i t r G t q S
k l
l k q q
=
= =
=

}
= =
k
k k
r t r q i
N
r d r q i t r G t q F )]) 0 ( ) ( [ exp(
1
) exp( ) , ( ) , (
s s

incoherent intermediate scattering function:
can one relate these quantities to transport processes?
F(q,t) for a glassforming colloidal system
0.010 0.015 0.020 0.025 0.030
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
= 0.620
= 0.581
= 0.54
S(q)

q [nm
1
]
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
1
10
2
10
3
0.0
0.2
0.4
0.6
0.8
1.0
= 0.513
= 0.620
= 0.581

S
(
q
,
t
)
/
S
(
q
)
t/s
Eckert, Bartsch, Faraday Discuss. 123, 51 (2003)
diffusion constants
diffusion process
self-diffusion interdiffusion
tagged
atoms in I
concentration
differences
in I and II
homogeneous distribution of
tagged particles in I+II
homogeneous distribution of
red and blue atoms in I+II
s s
s

A =
c
c
D
t
A AB
A
c D
t
c
A =
c
c
calculation of the self-diffusion constant
Einstein relation:
dt
r t r
D
t
2
)] 0 ( ) ( [
lim
2
tag tag
s

=

Green-Kubo relation:
}

=
0
tag tag s
) 0 ( v ) ( v
3
1
t t d D
v
tag
(t): velocity of tagged particle at time t
the two relations are strictly equivalent!
Computer Simulations of Colloids

Molecular Dynamics I:

How does it work?
Outline
MD how does it work?

numerical integration of the equations of motion

periodic boundary conditions, neighbor lists

MD in NVT ensemble: thermostatting of the system

MD simulation of a 2d Lennard-Jones fluid:
structure and dynamics

Newtons equations of motion
classical system of N particles at positions




U
pot
: potential function, describes interactions
between the particles

simplest case: pairwise additive interaction between point particles





solution of equations of motion yield trajectories of the particles, i.e.
positions and velocities of all the particles as a function of time

N i r
i
,..., 1 }, { =

i
i
i i
F
r
U
r m

=
c
c
=
pot

= >
= =
1
1
pot
) (
N
i
N
i j
j i ij ij
r r r r u U

microcanonical ensemble
consider closed system with periodic boundary conditions, no coupling
to external degrees of freedom (e.g. a heat bath, shear, electric fields)

particle number N and volume V fixed microcanonical ensemble

momentum conserved, set initial conditions such that total momentum
is zero

total energy conserved:

const.
pot kin
= + = U E E
simple observables and ergodicity hypothesis
potential energy:

= >
= =
1
1
) (
1
N
i
N
i j
ij
r u
N N
U
u
kinetic energy:
T k
m
N N
E
e
N
i
i i
B
1
2
kin
kin
2
v 1
= = =

=
pressure:

>
+ =
i i j
ij ij
r F
dV V
T Nk
p

1
B
here < ... > is the ensemble average; we assume the ergodicity hypothesis:

time average = ensemble average
a simple model potential for soft spheres
c o o c + = ] ) / ( ) / [( 4 ) (
6 12
r r r u
WCA potential:
=1, =1

cut off at r
cut
=2
1/6
,
corresponds to minimum
of the Lennard-Jones
potential
a simple problem
N = 144 particles in 2 dimensions:

starting configuration: particles
sit on square lattice

velocities from Maxwell-Boltzmann
distribution (T = 1.0)

potential energy U = 0 due to
d > r
cut
how do we integrate the equations of motion for this system?
L=14.0
d>r
cut
the Euler algorithm
Taylor expansion with respect to discrete time step t
t
m
t F
t t t
t t t r t t r
i
i
i i
i i i
o o
o o
) (
) ( v ) ( v
) ( v ) ( ) (

+ = +
+ = +
bad algorithm: does not recover time reversibility property of Newtons
equations of motion

unstable due to strong energy drift, very small time step required
the Verlet algorithm
consider the following Taylor expansions:
(2)
) (
2
1
) ( v ) ( ) (
(1)
) (
2
1
) ( v ) ( ) (
2
2
t
m
t t F
t t t t t r t r
t
m
t F
t t t r t t r
i
i
i i i
i
i
i i i
o
o
o o o
o o o
+
+ + + =
+ + = +

equations (1) and (2): velocity form of the Verlet algorithm

symplectic algorithm: time reversible and conserves phase space volume
addition of Eqs. (1) and (2) yields:
(3) )] ( ) ( [
2
) ( v ) ( v t t F t F
m
t
t t t
i i
i
i i
o
o
o + + + = +


implementation of velocity Verlet algorithm
do i=1,3*N ! update positions
displa=hstep*vel(i)+hstep**2*acc(i)/2
pos(i)=pos(i)+displa
fold(i)=acc(i)
enddo

do i=1,3*N ! apply periodic boundary conditions
if(pos(i).lt.0) then
pos(i)=pos(i)+lbox
else
if(pos(i).gt.lbox) pos(i)=pos(i)-lbox
endif
enddo

call force(pos,acc) ! compute forces on the particles

do i=1,3*N ! update velocities
vel(i)=vel(i)+hstep*(fold(i)+acc(i))/2
enddo

neighbor lists
force calculation is the most time-consuming part in a MD simulation
save CPU time by using neighbor lists
Verlet list:
update when a particle has
made a displacement
> (r
skin
-r
cut
)/2
cell list:
most efficient is a combination
of Verlet and cell list
simulation results I
c o t t o / with 0007 . 0 step time
2
MD MD
m t = =
initial configuration with u=0: after 10
5
time steps:
energies as a function of time
determine histogramm P(u) for the potential energy for different N
pressure
fluctuations of the potential energy
dotted lines: fits with
Gaussians

width of peaks is directly
related to specific heat c
V
per particle at constant
volume V
)
2
1 (
2
2 2
B
2
2
V
c
d
N
T dk
u u =
compare to canonical
ensemble
V
c
N
T k
e e
2 2
B
2
tot
2
tot
=
simulations at constant temperature: a simple thermostat
idea: with a frequency assign new velocities to randomly selected
particles according to a Maxwell-Boltzmann (MB) distribution
with the desired temperature

simple version: assign periodically new velocities to all the particles
(typically every 150 time steps)

algorithm:

(i) take new velocities from distribution

(ii) total momentum should be zero


(iii) scale velocities to desired temperature
) 2 / v exp( ) (v
2
, , i i
P
o o
=
i
i i
i i
Nm
m

,
, ,
v
v v
o
o o

i i
i i
m
T Nk
,
B
, ,
v
3
v v
o
o o
energies for the simulation at constant temperature
pair correlation function
pair correlation function
very similar to 3d

no finite size effects
visible
static structure factor
S(q) also similar to 3d

no finite size effects

low compressibility
indicated by small
value of S(q) for q0
mean squared displacement
t
t r
D
t
4
) (
lim
2

=
diffusion constant increases with increasing system size !?
long time tails in 3d
Green-Kubo relation for diffusion constant:
power law decay of velocity autocorrelation function (VACF) at long times
low density
high density
Alder, Wainwright,
PRA 1, 18 (1970)
Levesque, Verlet,
PRA 2, 2514 (1970)
2 / 3
t
}

=
0
) 0 ( v ) ( v
1
dt t
d
D
i i

cage effect in
dense liquid:
Alders argument
Stokes equation describes momentum diffusion:
current) mass e transvers , viscosity kinematic (
t t
2
t
V =
c
c
j j
t
j

q
v v
consider momentum in volume element V at t=0

volume at time t:
2 /
) (
d
t v
amount of momentum in original volume
element at time t
2 / 2 /
) /( VACF
d d
t t V

v o
d=2:
cally logarithmi diverges VACF
1
D t

refined theoretical prediction
1
) ) ln( ( VACF

t t
backflow pattern around
particle:
effective diffusion constants
larger system sizes required to check theoretical prediction !
literature
M. Allen, D. J. Tildesley, Computer Simulation of Liquids
(Clarendon Press, Oxford, 1987)

D. Rapaport, The Art of Molecular Dynamics
(Cambridge University Press, Cambridge, 1995)

D. Frenkel, B. Smit, Understanding Molecular Simulation: From
Algorithms to Applications
(Academic Press, San Diego, 1996)

K. Binder, G. Ciccotti (eds.), Monte Carlo and Molecular Dynamics
of Condensed Matter Systems
(Societa Italiana di Fisica, Bologna, 1996)
Computer Simulations of Colloids

Molecular Dynamics II:

Application to Systems Under Shear
with Jochen Zausch (Universitt Mainz)
shear viscosity of glassforming liquids
10
13
Poise
10
-3
Poise
dramatic slowing down of dynamics in a relatively small temperature range
F(q,t) for a glassforming colloidal system
0.010 0.015 0.020 0.025 0.030
0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
= 0.620
= 0.581
= 0.54
S(q)

q [nm
1
]
10
-6
10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
1
10
2
10
3
0.0
0.2
0.4
0.6
0.8
1.0
= 0.513
= 0.620
= 0.581

S
(
q
,
t
)
/
S
(
q
)
t/s
Eckert, Bartsch, Faraday Discuss. 123, 51 (2003)
glassforming fluids under shear
apply constant shear rate

Weissenberg number: ) (T W t

=
(T): relaxation time in equilibrium
hard sphere colloid glass: confocal microscopy
Besseling et al., PRL (2007)
central question: transient dynamics towards steady state for W >>1

problem amenable to experiment, mode coupling theory and simulation
drastic acceleration of dynamics
outline
simulation details: interaction model, thermostat, boundary conditions


dynamics from equilibrium to steady state?





dynamics from steady state back to equilibrium?
technical requirements
(1) model potential

(2) external shear field produces heat: thermostat necessary

(3) can we model the colloid dynamics realistically?

(4) how can we shear the system without using walls?
model potential: a binary Yukawa mixture
requirements: binary mixture (no crystallization or phase separation)

colloidal system (more convenient for experiments)

possible coupling to solvent: density not too high
7
c AA
AA AB AA BB
AA AB AA BB
10 ) ( off - cut , / 6
4 . 1 , 2
1 . 1 , 2 . 1
)] ( exp[ ) (

= =
= =
= =
=
o|
o|
o|
o| o|
o|
o k
c c c c
o o o o
o k
o c
r U
r
r
r U
equations of motion and thermostat
use of thermostat necessary + colloid dynamics solve Langevin equation

equations of motion:

dissipative particle dynamics (DPD):








Galilean invariant, local momentum conservation, no bias on shear profile

low (=12): Newtonian dynamics, high (=1200): Brownian dynamics

=
+ + = =
i j
ij ij ij i i i i
F F F p m p r ] [ /
R D

nce unit varia and mean zero with numbers random uniform :
25 . 1 for
25 . 1
1 ) (

) ( 2 , ) v )( (
AA
AA
R 2 D
ji ij
ij
ij
ij
ij
ij ij B ij ij ij ij ij ij
r
r
r w
t
r
r w T k F r r r w F
O = O
< =
O = =
o
o
o

Lees-Edwards boundary conditions

no walls, instead modified periodic
boundary conditions

shear flow in x direction,
velocity gradient in y direction,
free z direction

shear rate:


linear shear profile
L
x
/ v 2 =

) 2 / ( ) ( v
s
L y y =

further simulation details

generalized velocity Verlet algorithm, time step


system size: equimolar AB mixture of 1600 particles (N
A
=N
B
=800)


L=13.3 , density =0.675 , volume fraction 48%


at each temperature at least 30 independent runs


temperature range 1.0 T 0.14 (40 million time steps for production
runs at T=0.14)
3
AA

o
AA
o
AA
2
AA A
/ 0083 . 0 c o o m t =
self diffusion
B particles slower than
A particles

weak temperature
dependence of D for
sheared systems if
1 /
2
>> = D W o

along T = 0.14:
W ~ 100 - 1000
shear stress: equilibrium to steady state

>
+ =
i i j
y ij x ij y i i x i i xy
F r y m
L
] v )) ( v v ( [
1
, , , s ,
3
o
definition:
maximum around marks
transition to plastic flow regime

steady state reached on
time scale

time scale on which linear
profile evolves?
1 ~ t

0.1 ~ t

averaging over 250

independent runs
shear profile
velocity profile becomes almost linear in the elastic regime

maximum in stress not related to evolution of linear profile
structural changes: pair correlation function
pair correlation function not
very sensitive to structural
changes

consider projections of g(r)
onto spherical harmonics

structure and stresses
} }

c
c
=
c
c
=
0
22
3 kin
2
kin
)] ( Im[
) (
) (
) (
2
r g
r
r U
r dr A r g
r
xy
r
r U
r d
xy xy xy
o

o o

how is the stress overshoot reflected in dynamic correlations?
mean squared displacement
occurrence of superdiffusive regime in the transient plastic flow regime

superdiffusion less pronounced in hard sphere experiment ( Joe Brader)
EQ to SS: superdiffusive transient regime
y: gradient direction
z: free direction

measure effective exponent:
) log(
) ) , ( log(
w
w
2
t t d
t t r d
s

A
=
EQ to SS: Incoherent intermediate scattering function
transients can be described by
compressed exponential decay:
1 with
] ) / ( exp[ ) , (
s
>

|
t
|
q
t t q F
t
w
=0: 1.8 different for
Brownian dynamics?
EQ to SS: Newtonian vs. overdamped dynamics
same compressed exponential decay also for overdamped dynamics
change friction coefficient
in DPD forces:
dynamics Brownian" " : 1200
dynamics Newtonian : 12
=
=

shear stress: from steady state back to equilibrium

decay can be well described by stretched exponential with 0.7

stressed have completely decayed at
1 ~ t

SS to EQ: mean squared displacement

stresses relax on time scale of
the order of 300 (crossover in
t
w
=0 curve)

then slow aging dynamics toward
equilibrium
conclusions
phenomenology of transient dynamics in glassforming liquids under shear:
binary Yukawa mixture:
model for system of charged colloids, using a DPD thermostat and
Lees-Edwards boundary conditions

EQSS: superdiffusion on time scales between the occurrence of the
maximum in the stress and the steady state

SSEQ: stresses relax on time scale 1/ , followed by slow aging
dynamics

Computer Simulations of Colloids

Monte Carlo Simulation I:

How does it work?
Outline
MC how does it work?

calculation of via simple sampling and Markov chain sampling

Metropolis algorithm for the canonical ensemble

MC at constant pressure

MC in the grandcanonical ensemble

calculation of
first method: integrate over unit circle



second method: use uniform random numbers

direct sampling
Markov chain sampling
Werner Krauth, Statistical Mechanics: Algorithms and Computations
(Oxford University Press, Oxford, 2006)
calculation of : direct sampling
estimate from


algorithm:

N
hits
=0
do i=1,N
x=ran(-1,1)
y=ran(-1,1)
if(x
2
+y
2
<1) N
hits
=N
hits
+1
enddo

estimate of from ratio N
hits
/N
4
square
circle
t
=
A
A
calculation of : Markov chain sampling
algorithm:

N
hits
=0; x=0.8; y=0.9
do i=1,N
x=ran(-,)
y=ran(-,)
if(|x+x|<1.and.|y+y|<1) then
x=x+x; y=y+y
endif
if(x
2
+y
2
<1) N
hits
=N
hits
+1
enddo

optimal choice for range:
not too small and not too large

here good choice
max
=0.3
calculation of with simple and Markov chain sampling
simple sampling (N = 4000):
run N
hits
estimate of

1 3156 3.156
2 3150 3.150
3 3127 3.127
4 3171 3.171
5 3148 3.148
run N
hits
estimate of

1 3123 3.123
2 3118 3.118
3 3040 3.040
4 3066 3.066
5 3263 3.263
Markov chain sampling
(N = 4000,
max
= 0.3)
simple sampling vs. Markov chain sampling

} }
} }
=


= ~ =
N
i
i
y x dy dx
y x A y x dy dx
A A
N N
N
1
1
1
1
1
1
1
1
1 hits
) , (
) , ( ) , (
1
t
t

=
elsewhere 0
square inside 1
) , ( y x t
probability
density:

=
elsewhere 0
circle inside 1
A
observable:
simple sampling: A sampled directly through

Markov sampling: acceptance probability
p(oldnew) given by
]
(old)
new) (
, 1 [ min new) old (
t
t
= p
more on Markov chain sampling
Markov chain: probability of generating configuration i+1 depends
only on the preceding configuration i

important: loose memory of initial condition

consequence of algorithm: points pile up at the boundaries

max
shaded region:
piles of sampling points
detailed balance
consider discrete system: configurations
a, b, c should be generated with equal
probability
ies probabilit n transitio : ),...} ( ), ( {
,... , at be to system the of ies probabilit stationary : ),...} ( ), ( {
c a p b a p
b a b a

t t
detailed balance condition holds
) ( ) ( ) ( ) (
) ( ) ( ) ( ) (
a c p c c a p a
a b p b b a p a
=
=
t t
t t
detailed balance in the calculation of
one can easiliy show that
old) new ( new) ( new) old ( old) ( = p p t t
detailed balance: analog to the time-reversibility property of Newtons
equations of motion
acceptance probability for a trial move
]
old) (
new) (
, 1 min[ new) old (
t
t
= p
a priori probabilities
generalized acceptance criterion:
new) old ( new) old (
new) old (
ap

=
p p
P
moves x and y in square of linear size defines a priori
probability p
ap
(oldnew)


different choices possible
]
(old)
old) new (
new) old (
new) (
, 1 min[ new) old (
ap
ap
t
t

=
p
p
p
canonical ensemble
problem: compute expectation value
n Hamiltonia : ) ( ); ,......, , (
)] /( ) ( exp[
) ( )] /( ) ( exp[
) (
) (
2 1
) (
B
) ( ) (
) (
B
) ( ) (
) (
N
N
N
N N
N N N
N
r H r r r r
T k r H r d
r A T k r H r d
r A

=
}
}
simple sampling: sample random configurations

=
=

=
M
l
N
l
M
l
N
l
N
l
N
T k r H
r A T k r H
r A
1
B
) (
1
) (
B
) (
) (
)] /( ) ( exp[
) ( )] /( ) ( exp[
) (

samples mostly states in the tails of the Boltzmann distribution

) ( N
l
r

direct evaluation of the integrals: does not work!

idea of importance sampling
sample configurations according to probability
choose

=
=

=
M
l
N
l
N
l
M
l
N
l
N
l
N
l
N
r P T k r H
r P r A T k r H
r A
1
) (
B
) (
1
) ( ) (
B
) (
) (
) ( / )] /( ) ( exp[
) ( / ) ( )] /( ) ( exp[
) (

this can be realized by Markov chain sampling: random walk through

regions in phase space where the Boltzmann factor is large
) ( N
l
r

) (
) ( N
l
r P

)] /( ) ( exp[ ) (
B
) ( ) (
T k r H r P
N
l
N
l


, then

=
=
M
l
N
l
N
r A
M
r A
1
) ( ) (
) (
1
) (

importance sampling
choose transition probability such that
to achieve this use detailed balance condition
= M T k r H
Z
P
N
l
for )] /( ) ( exp[
1
B
) (
eq

possible choice
) (
) ( ) ( N
l
N
l
r r W

) exp(
) (
) (
or
) ( ) ( ) ( ) (
B
) ( ) (
1
) (
1
) (
) ( ) (
1
) (
1 eq
) (
1
) ( ) (
eq
T k
H
r r W
r r W
r r W r P r r W r P
N
l
N
l
N
l
N
l
N
l
N
l
N
l
N
l
N
l
N
l
A
=

=
+
+
+ + +

> A A
=
+
otherwise 1
0 )] /( exp[
) (
B ) (
1
) (
H T k H
r r W
N
l
N
l

yields no information about the partition sum
Metropolis algorithm
N. Metropolis, A.W. Rosenbluth, M.N. Rosenbluth, A.H. Teller, E. Teller, JCP 21, 1087 (1953)
old configuration
trial configuration
E = E(trial) - E(old)
E 0 ?
new config. =
trial config.
r <exp[-E/(k
B
T)] ?
new config. = old config.
yes
no
no
r uniform random
number 0 < r < 1
displacement move
A + =
old trial
r r
acceptance probability
] ) ( old))/ ( trial) ( ( [ exp min(1,
new) old (
B
T k E E
W

=
implementation of the Metropolis algorithm
do icycl=1,ncycl

old=int[ran(0,1)*N]+1 ! select a particle at random
call energy(x(old),en_old) ! energy of old conf.
xn=x(old)+(ran(0,1)-0.5)*disp_x ! random displacement
call energy(x(new),en_new) ! energy of new conf.

if(ran(0,1).lt.exp[ -beta*(en_new-en_old) ] ) x(old)=x(new)

if( mod(icycl,nsamp).eq.0) call sample ! sample averages

enddo
dynamic interpretation of the Metropolis precedure
: ) , (
) (
t r P
N

probability that at time t a configuration r

(N)
occurs during
the Monte Carlo simulation
rate equation (or master equation):
) ( ) , ( :
) , ' ( ) ' ( ) , ( ) ' (
) , (
eq
' '
r P t r P t
t r P r r W t r P r r W
dt
t r dP
r r

=
+ =


=
' '
eq eq
) ' ( ) ' ( ) ( ) ' (
r r
r P r r W r P r r W

equilibrium:
) (
) ' (
) ' (
) ' (
eq
eq
r P
r P
r r W
r r W

detailed balance condition one possible solution:

remarks
Monte Carlo measurements:
first equilibration of the system (until P
eq
is reached), before
measurement of physical properties
random numbers:
real random numbers are generated by a deterministic algorithm and
thus they are never really random (be aware of correlation effects)
a priori probability:
can be used to get efficient Monte Carlo moves!
Monte Carlo at constant pressure
system of volume V in contact with an ideal gas reservoir via a piston
V
V
0
- V
acceptance probability for volume
moves:
)}]
'
ln( ) ' (
)) , ( ) ' , ( (
1
exp{ , 1 min[
) new old (
B
) ( ) (
B
V
V
T k
N
V V P
V s E V s E
T k
W
N N
+

=

trial move:
] , [ interval from number random uniform with '
max max
V V V V V V A + A A A + =
grand-canonical Monte Carlo
system of volume V in contact with an ideal gas reservoir with
fugacity
V
V
0
- V
insertion:
))}] ( ) 1 ( (
1
exp{
1
, 1 min[
) 1 (
B
N E N E
T k N
Vz
N N W
+
+
= +
] exp[
B
T k
z

=
))}] ( ) 1 ( (
1
exp{ , 1 min[
) 1 (
B
N E N E
T k Vz
N
N N W

=
removal:
textbooks on the Monte Carlo simulation method
D. P. Landau, K. Binder, A Guide to Monte Carlo Simulations in
Statistical Physics
(Cambridge University Press, Cambridge, 2000)

W. Krauth, Statistical Mechanics: Algorithms and Computations
(Oxford University Press, Oxford, 2006)

M. Allen, D. J. Tildesley, Computer Simulation of Liquids
(Clarendon Press, Oxford, 1987)

D. Frenkel, B. Smit, Understanding Molecular Simulation: From
Algorithms to Applications
(Academic Press, San Diego, 1996)

K. Binder, G. Ciccotti (eds.), Monte Carlo and Molecular Dynamics
of Condensed Matter Systems
(Societa Italiana di Fisica, Bologna, 1996)
Computer Simulations of Colloids

Monte Carlo II:

Phase Behaviour of Colloid-Polymer

Mixtures
collaborations: Richard Vink, Andres De Virgiliis, Kurt Binder
depletion interactions in colloid-polymer mixtures
demixing transition possible into colloid-rich phase (liquid) and
colloid-poor phase (gas)
polymers cannot move into
depletion zones of colloids
with their center of mass

gain of free volume if there is
overlap of the depletion zones
of two colloids

effective attraction between
colloids (of entropic origin)
Asakura-Oosawa (AO) model

<
=
otherwise 0
2
) (
c
cc
R r
r u

+ <
=
otherwise 0
) (
p c
cp
R R r
r u 0 ) (
pp
= r u
in the following AO model for R
p
/R
c
= 0.8:

exhibits demixing transition into a colloid-poor phase (gas) and a
colloid-rich phase (liquid)

temperature not relevant; what is the corresponding variable here?
Outline
V
N
R R z
C
3
C C
3
P P p
3
4
,
3
4
r
t
q
t
q = =
phase diagram:
AO model for R
P
/R
C
= 0.8:

binodal

critical point (FSS)

interfacial tension

interfacial width

capillary waves
volume V, polymer fugacity z
P
and colloid fugacity z
C
fixed removal
and insertion of particles


problem 1: high free energy barrier between coexisting phases in the
two-phase region (far away from the critical point)
use successive umbrella sampling



problem 2: low acceptance rate for trial insertion of colloidal particle, high
probability that it overlaps with a polymer particles
use cluster move
grandcanonical MC for the AO model
problem: high free energy barrier between coexisting phases
successive umbrella sampling
] e
1
, 1 [ min
C
C
E
N
V z
A
A
+
+
=
|
cluster move
] e , 1 [ min
C
C
E
V z
N
A
A

=
|
try to replace n
r
n
P
polymers (A)
by a colloid (B):
reverse move (C+D):
sphere of radius and volume V

around randomly selected point

)! (
) ( )! (
r P
P P
r
n n
V z n
n
+
o
r
) ( )! (
)! (
P r P
P
n
V z n n
n
o

problem: low acceptance rate for trial insertion of a colloid cluster move
R.L.C. Vink, J. H., JCP 121, 3253 (2004)
phase diagram
estimate interfacial tension by = F/(2A) (Binder 1982)
critical point from finite size scaling
3D Ising universality class
finite size scaling
cumulant ratio
C C
2
2
with | | / q q = = m m m M
critical value

r
P,cr
0.766
interfacial tension
*(2R
C
)
2
as a function of the difference in the colloid packing
fraction of the liquid (L) and the vapor (V) phase at coexistence
DFT (M. Schmidt et al.)
yields accurate result
interfacial profile
|
|
.
|

\
|

+ = u
0
0
tanh ) (
w
z z
B A z mean field result:
strong finite size effects
(not due to critical fluctuations)
, L
x,y
=31.3
, L
x,y
=23.1
capillary wave theory
free energy cost of spatial interfacial fluctuations (long
wavelength limit)




result for mean square amplitude of the interfacial thickness



combination with mean-field result by convolution approximation



determination of intrinsic width w
0
by simulation not possible
( ) ) / ln( 2
1
2
cw
a L w

= t
a L w w ln ) 4 ( ln ) 4 (
1 1 2
0
2
+ =
(
(

|
|
.
|

\
|
c
c
+
|
.
|

\
|
c
c
= A
}
2
2
2
1
y
h
x
h
dxdy F
s
|
is CWT valid?
logarithmic divergence?
lateral dimension L
x
= L
y
large enough?
finite size scaling II
map on universal 3D Ising
distribution
account for field mixing
AO model belongs to 3D Ising
universality class
conclusions
discrepancies to mean-field results

AO model belongs to 3D Ising universality class

interfacial broadening by capillary waves

much more on the AO model in

R.L.C. Vink, J.H., JCP 121, 3253 (2004);
R.L.C. Vink, J.H., JPCM 16, S3807 (2004);
R.L.C. Vink, J.H., K. Binder, JCP 122, 134905 (2005);
R.L.C. Vink, J.H., K. Binder, PRE 71, 011401 (2005);
R.L.C. Vink, M. Schmidt, PRE 71, 051406 (2005);
R.L.C. Vink, K. Binder, J.H., Phys. Rev. E 73, 056118 (2006);
R.L.C. Vink, K. Binder, H. Lwen, PRL 97, 230603 (2006);
R.L.C. Vink, A. De Virgiliis, J.H. K. Binder, PRE 74, 031601 (2006);
A. De Virgiliis, R.L.C. Vink, J.H., K. Binder, Europhys. Lett. 77, 60002 (2007);
K. Binder, J.H., R.L.C. Vink, A. De Virgiliis, Softmatter (2008).
Computer Simulations of Colloids

Modelling of hydrodynamic interactions
collaboration: Apratim Chatterji (FZ Jlich)
simulation of colloids
colloids:
size 10 nm 1 m
mesoscopic time scales
solvent:
atomistic time (~ 1 ps) and
length scales (~ )
simulation difficult due to different time and length scales
coarse graining of solvents degrees of freedom
Langevin equation
i i i
i
f t F
dt
r d
M
, r 0 , c
2
2
) ( v

+ =
) ' ( ) ' ( ) ' , ' ( ) , (
, , with 0 ) , (
) (
, r
) (
, r
) (
r,
t t r r A t r f t r f
z y x t r f
i i
i
=
= =
o o o
o
o|
| o
o

0
2 T k A
B
=
Brownian particles of mass M :
f
r,i
uncorrelated random force with zero mean:
fluctuation-dissipation theorem:
many collisions with solvent particles
on typical time scale of colloid

systematic friction force on colloid
decay of velocity correlations
) 0 ( v ) ( v ) 0 ( v ) ( v ) ( v
) ( v
0 , r 0 x x x x x x
x
t t
dt
d
M f t
dt
t d
M = + =
) exp( ) 0 ( v ) ( v
0
t
M
t
x x


consider single Brownian particle of mass M :
exponential decay of velocity autocorrelation function (VACF)
correct behavior: power law decay of VACF due to local momentum
conservation
2 / 3
) 0 ( v ) ( v

t t
t x
generalized Langevin equation
)] , ( ) ( v [ ) , (
0 fric
t r u t t r F
i

=
idea: couple velocity of colloid to the
local fluid velocity field via frictional force

point particle:
problems:

determination of hydrodynamic velocity field from solution
of Navier-Stokes equation

rotational degrees of freedom colloid-fluid coupling?
) , ( t r u

lattice Boltzmann method I
: number of particles at a lattice node , at a time t, with
a velocity

discrete analogue of kinetic equation:



i
: collision operator

velocity space from projection of 4D
FCHC lattice onto 3D (isotropy)


moments of n
i
:
r

= = =
= H = =
18
1
18
1
18
1
, ,
i i i
i i i i i i
c c n c n u j n

) , ( t r n
i

i
c

) , ( ) , ( ) 1 , ( t r t r n t c r n
i i i i

A + = + +
lattice Boltzmann method II

=
= A
18
1
eq
] ) , ( [
j
j j ij i
n t r n L

)] 1 ( :
1 1
[
2
s
4
s
2
s
eq
c c c u u
c
c u
c
a n
i i i
c
i
i
+ + =


linearized collision operator:



equilibrium distr.:



recover linearized Navier-Stokes equations:




thermal fluctuations: add noise terms to stress tensor
j p
t
j
j
t

2
, V + V =
c
c
V =
c
c

lattice Boltzmann algorithm III
) , ( ) , ( t r t r n
i i

A +
) 1 , ( + + t c r n
i i

collision step: compute
propagation step: compute
colloid fluid momentum exchange
sphere represented by (66)
uniformly distributed points
on its surface
each point exchanges
momentum with surrounding
fluid nodes
velocity correlations at short times
give particle a kick and determine decay of its velocity
blue curves:
) exp( ) (
0
t
M
t g

=
red curves:
fluid velocity at
the surface of the
sphere
C
v
(t) and C

(t)
2 / 3
0
) (
12

= tv

M
A
2 / 5
) 4 (

= tv

tI
B
independent of R and
0
I: moment of inertia
thermal fluctuations ok?
linear response theory:
velocity relaxation of kicked particle in a fluid at rest =
velocity autocorrelation function of particle in a thermal bath
charged colloids
colloids in electric field