Statistical Mechanics of Liquids: Johan T. Padding Université Catholique de Louvain, Belgium

Statistical Mechanics of Liquids
Johan T. Padding
Université catholique de Louvain, Belgium
1
Simple liquids Colloidal liquids Polymeric liquids
near-spherical molecules liquid with embedded particles long, stringy molecules

structure on molecular scale structure also on nano/micro scale random coils / fibers / helices
water, light oils paint, milk, blood silly putty, actin, DNA solution
2
Statistical mechanics:
bridge between molecular and macroscopic world
• equilibrium properties
• intermolecular interactions (structure,pressure, ...)
many particles
• density, temperature • dynamical properties
(diffusion, viscosity, ...)
Statistical mechanics:
• does not predict individual molecular trajectories, but
• gives probabilities for large collections of particles, and
• yields measurable quantities as averages over ensembles
3
Chapter 1: Structure in liquids and gases
4
1.1Probability densitydensity
Probability
lity density
Consider a closed box of volume V filled with
• box volume V, with N identical (rigid) particles
a fluid (liquid or gas) consisting
of volume
of a large number N ofwith
V filled identical molecules.
a fluid (liquidFororourgas)
purposes we may assume
consisting
each molecule to be a rigid object with a given position and orientation.
identical molecules.
• define number density ρFor = our
N/V purposes we may assume
According to classical statistical mechanics, these positions and orientations
aarerigid object with
not completely random. a given Rather, position
at not too andloworientation.
temperature T ,1 the
• probability
statistical
probability density
for toencountering
mechanics,
density encounter
these system in configuration
positions
a certain and orientations
configuration of 3N position
and 3N 6N
orientation coordinates, in shorthand denoted by R 6N
, is given by
R
ndom. Rather, = {r , Ω
1 at , . .
1 not too. , r N , Ω low
N } temperature T , the 1
the Boltzmann distribution function:
is given by the a
encountering Boltzmann
certaindistribution function: of 3N position
configuration
6N
ordinates, in shorthand
1
!
Φ(R denoted
6N
)
" by R , is given by
P (R6N ) = exp − , (1.1)
ution function:Z kB T
6N
where• Φ(R 6N) is) the total potential
energy ofenergy of the configuration, kB = 1.38065×
! "
Φ(R is potential the configuration - contains interactions
p10−23−J/K = 8.617 ×, 10−5 eV/K is Boltzmann’s constant, and Z(1.1) is a normal-
kB T referred to as the configuration integral.
isation constant, 5
is
Jones
nes
state well described
formula,
formula,
energy ofLennard-Jones
the1two
Lennard-Jones by1 atoms
the formula, as
formula, 1 1
.Lennard-Jones
At At
largelarge distances
distances formula, 1
1.2
der
Eq.Waals
Pair interactions of spherical molecules
Eq. Eq.
!0 (rattraction
,the
(1.4).
r2)large
(1.4).
= van
!der
At At
(rder
large distances
large distances
Waals (1.4). 1At
attraction van 1) +
0distances
Waals !attraction
0 (rattraction
2r)=+ ϕ(r , r2 ). (1.2)
"(r)/#
the Waals
"(r)/#
"(r)/#
!r = ! 1
"(r)/#
r = !r = !
.ant.
theAtvan At
short short
der dis- dis-
isWaals
dominant.attraction At At short dis-dis-
Pair interaction between two neutral atoms
is dominant. short
"(r)/#
To
heHere
atoms focus
atoms !
repel on
repel
(r ) is
each the
each
tances the
tances essentials
the 0
0ground
atoms
the atoms statewe
repelrepel will
energy
each eachtreat
of
0 0 the
atom r =simplest
! in the case,
1 absence namely
of atom that 2, of
is dominant.
0 1 At short dis- 2
neutral
cause
se of of
the spherical
the
Pauli Pauli atoms. Suppose weϕ(r have just two atoms, fixed with their
and similarly
tances
of two
the
principle. The
rinciple.
nuclei at
other
atoms
positions
unperturbed
other
The
exclusion
because
for repel
!0because
exclusion r and
of
(r2 ).each
principle.
Sothe
of
principle.
ground r
the
,
Pauli
the
stateas
0term
The in
1.2
Pauli
The1.2 Pair
Fig. Pairinteraction
energies
1 , r2 ) is the correction to the sum
interactio
1.1.
vanof We
der the can write
atoms. Thisthe total
term is ground
also
other
of •
the because
one
the atom may pair
atom of
may
diameter
of the
atoms, 1 Pauli
fixed
of the atom
2
nuclei: maymay
van der WaalsWaals attraction
attraction van der
van Waals
der attraction
Waals attraction
diameter of the atom
state
ed
ascalled
exclusion
as the
the
energy
the
total ground
distance
distance
of σthe
interatomic
principle.
beσdefined -1 two
state as
atoms
interaction
-1energy
The
the distance
as σor interatomic
-1 -1 potential. Because of the
be defined as the distance σTo focus
To focus on the
on the
van essentials
der attraction we
essentials
Waals we wil
w
diameter
rotational
ese two of the
symmetry
interactions
two interactions where where atom
these of may
the
twotwo
these atoms,
interactions
interactions the interatomic potential only depends on
be
ancel
el
!
defined
out. out.0 (r 1 ,asr 2 )
the = ! 0 (r
distance 1 ) + σ! 0 (r 2 )-1 + ϕ(r 1 ,
neutralr ).
neutral
2 spherical
spherical atoms.
atoms. 2 2
Suppose (1.2)
Suppo
the distance rexactly 12 = 0.5
exactly |rcancel
cancel − r2out.
out.
1 0.5 | between
1.0 1.0 the
1.5 1.5two
0.5 0.5atoms,
2.0 2.01.0 1.0i.e.2.5 1.5
2.5 1.5 2.0 2.0 2.5 2.5
where these two interactions r/!nuclei at positions 1 and
r1 rand , as
Here ! (r ) is the ground state energy of atom 1 in the absence
0 1, r ) = ϕ(r ).
nuclei
r/! at positions r/! r/! 2r, 2as
of ratom 2,in
exactlyϕ(rcancel
1 2 out. is interatomic
12 interaction
0.5 / potential
1.0 1.5 2.0 (1.3)
2.5sum
and
well similarly
is well described
described for
by !
thethe
bydistance
distance is(r ). So the
Lennard-Jones
Lennard-Jones
0 well
is described
2 well described term state
byformula
formula
the
by state
ϕ(r ,
(seer energy
energy
) is
2Fig. theof
1.2):1.2):
(seeLennard-Jones
1Fig.
Lennard-Jones
the of
formula the
the (see
correction
formula two two
(seeFig.toatoms
atoms 1.2):asas
the
1.2):
Fig.
It is alsounperturbed
clear that because of its definition ϕ(∞) 2 r/!
= 0.atoms.
At finite distances,
of two ground state energies of the This term isthealso
• "Lennard-Jones potential describes both ! (r , r ) = ! (r ) + ! (r ) +
electrons
called
!
distance the
!"
σ
Van
# σin
12
is
der
one
12
well
# σ atom
interatomic
σ
6
$
described
Waals
" "
# # 6 will
$ !"!
attraction byfeel
interaction
σ 12
the
and
#
" the
# " electrons
or
σ #
" 6
σ Lennard-Jones
12 σ
$# $
6 ! in
interatomic (rthe
0 ,
0 formula r )
other
1 =
2
potential.
1 2 (see0 Fig. !
atom.
Pauli
(r
repulsion
0 )A+
1
Because !
1 1.2): 0 2 (r
classical
0of )2+
the
r)4"
= = 4" − −be ϕ(r) ϕ(r) .= 4"
.= following:
4" −the − charge . . distribution (1.4)
(1.4) (1.4)
(1.4
picture
rotational would
r r symmetry
short
the of the ratoms,
r repulsion
ranger Pauli r the in
r r interatomic potential only depends on
1 an atom is not
constant, butrfluctuates in time aroundHere ! (r ) is the ground state e
Here its ! (r
average. 0 ) 1is the
Consequently, ground the state
atom en
the distance = |r − r | between the two 0 atoms,
1 i.e.
(r)/¡
12 1 2
! $
has a fluctuating
" σ
dipole
# 12 "
moment σ which is and
# 6
zero similarly
on average.
r=m
for
The ! (r ).
instantaneous So the
ameter
eter " isϕ(r)
"the =
is the
The 4"
depthdepth
The of the
parameter
ofparameter
the −
"interaction
is"the
interaction is the .depth
depth
well,well,
of
and
and and
the
ofσ σ is
similarly
is
the the the
interaction diameter
diameter
interaction
0 well,
forwell,and
! σ
(r
and 0 is
σ). 2
the
is So
the (1.4)
diameterthe
diamete te
0 2
dipoles
.om.
TheThe ϕ(rin1 ,the
of
values
values of )"of
r2the =the
atoms,
ofand σratom.
ϕ(r
"atom.
and are
12 ).are
however,
σThe The rvalues
values influence
of "ofand
characteristic
characteristic σ of
for
" and
for eacheach
are two
each unperturbed
other
characteristic
σ atomic
are atomic in aforway
species.
characteristic
species. each
for each ground
which
atomicatomicmakes
species.(1.3)
stat
species
mple argon:
for for
eeach argon
argon For
"/k example
For
"/k = example
=
117.7 for
117.7 K argon
forK
and argon
dipole orient a little in the field ofcalled and
σ "/k
σ
= "/k
= of
= =two
117.7
0.3504
0.3504 117.7
nm, K unperturbed
nm, and
for
the other. K and
for σ =σ 0.3504
=
krypton
krypton This 0.3504
leads
ground
nm,
van nm,
to for the
der Waals
state
krypton
for krypton
attraction
so-
It is also clear
B
that
B
because of its
B B
definition ϕ(∞) the interatomic
-1 = 0. At finite distances, the interactio
4.0
164.0
The K krypton:
and
K σ
and
calledparameter "/k
=σ "/k
van der Waals
B = =
0.3827 =
164.0
0.3827 164.0
nm, K
" is theattraction
B and
nm, K
and and
σ
for
and
depth of the σ
= =
methane
for called
0.3827
0.3827
methane nm,
"/k
interaction
between two the
nm,
"/k
B and
= andinteratomic
148.9
=
Bwell,
neutral for
for148.9
andKmethane
methane
atoms. K is "/k
σ the
The Binteraction
"/k =
diameter
Bvan =
148.9 148.9
der K K
=electrons
.3783
0.3783nm. nm.inand
Noteoneand
Noteσ atom
=
thatσ =
0.3783
at
that will
0.3783
room
at nm.
room feel
nm. the
NoteNote
temperature that
temperature rotational
electrons
that
at
the at
room in
room
magnitudes
the the symmetry
other
temperature
temperature
magnitudes of of atom.
the the ofmagnitudes
A the
magnitudes atoms
classical of o
Waals attraction becomes stronger asrotational
of the atom. The values of " and σ are characteristic
the atoms get symmetry
0.5 forcloser to of
each
1.0 onethe
atomic
1.5 species.
another.
2.0 atoms, 2.5
epicture
same
the
For sameorder
At aexample
"
order
would are"
as the
certain point,
are
of of
the the
thermal
asbethe
for argonthe same
thermal
however,
sameorder
energy
following:
"/kB the
order
energy as
kB T
= 117.7
as
the the
k.B TThis
the the
thermal
thermal
.charge
atomsKwill
distance
energy
is important
This is
andrepel
energy
important
distribution
σ =each
k Tk
Bfor
0.3504 r. T .
forin
B12
other
=
This
nm,
This|r
is
an
r/m 1 is−
atom
for krypton
because
r
important
important 2 |
of the isbetwe
6 fo
for
not
structure to change dynamically under the influence of thermal fluctuations.
This is hardly allowed in a solid.
Pair sum
When dealingapproximation
with more than two spherical molecules, it is often as-
sumed that the total potential energy may be approximated as a sum of pair
interactions (in practice this is often a reasonable assumption):
• when >2 particles are present:
N
% −1 N
%
Φ (r1 , . . . , rN ) = ϕ(rij ). (1.5)
i=1 j=i+1
The double sum is constructed such that each pair interaction is counted
only• once.
in practice often a reasonable assumption
7
The radial distribution function g(r)
• molecular interactions cause correlations

in their positions
the radial distribution function g(r) by

• number of particles at a distance between
the spherical shell of thickness dr at a
r and r+dr from a given particle:
1. STRUCTURE IN LIQUIDS AND GASES
2N
4πr g(r)dr.
1. STRUCTURE IN LIQUIDS
V
the radial distribution function g(r) by equating the number of molecules in
the spherical shell of thickness dr• at
N/Va g(r) is average
distance r to number densityabove,
at
the radial distribution function g(r) by equating the number g(∞
According to our remarks of m
distance r from a given particle
N the spherical shell of given in Fig.
thickness dr at(1.3). We see
a distance r tothat g(r) =
4πr 2 g(r)dr. diameter σ. The first peak is (1.6) b
caused
V 2N • easy to see that
4πr g(r)dr. at distances where the potential has i
According to our remarksVabove, g(∞) = 1 and g(0) = 0. A typical g(r) is
than average. Consequently at distanc
given in Fig. (1.3). We see to
According thatourg(r) = 0 when
remarks above,r isg(∞)
smaller
= 1than
and the
g(0)molecular
= 0. A typi
diameter σ. Thegiven
firstinpeak
less particles than average.
Fig. is caused
(1.3). We by
see the
thatattractive part rofisthe
g(r) = 0 when potential;
smaller than8 the
at distances where the potential
diameter haspeak
σ. The first its minimum
is caused bythere
the are more particles
attractive part of the
less particles
P12 (r1 ,than r2 ) =average. d r3 . . . d rN exp − , (1.7)
we find Z kB T
Integrating the probability density for a configuration of N spherical parti-
cles, cf. PEq. (1.1), 3N
1over the
3 coordinates 3 forofg(r) allhave
particles ) except the first two,
" #
Φ(rparticle
! !
1.3.2
where P (r Statistical
12 (r
, r1 , )r 2=) is the dformulas
probability
r . . . ddensity
r exp to − , 1 at r 1 and particle
(1.7)
Statistical
we2 at
find 12 1 2
r2 . For convenience
Integrating the probability
formulas
Z
3
of notation
density
for g(r)
N
we write kB T
for a configuration of N spherical parti-
3N #$
" #
1 " Φ(r3N ) $except
! !
cles, cf. Eq. (1.1), over
1 the coordinates of all particles
Φ(r ) the first two,
! !
where P
PP 12 (r
(r
12 (r, 1,, rr
1 r2) = 2
! )) is
= the probability
d 3
3 r . . . density
d3
3
r exp to have
− particle , 1 at r 1 and particle
(1.7)
we find 12 Z d r 3. . .
3 d r N
N exp −
k T
$ . (1.8)
2 at•rintegrate
2 . For convenienceBoltzman Z distribution
of notationfunction we write over all
kBBTcoordinates$ except first 2
r1 =r,r2 =r!
3N #$
" #
1 probability Φ(r 3N ) $ 1 at r1 and particle
! !
where
Because P 12 (r
all1 , r 2 )
particles is 1 the
!
are 3
equal,
!
thisdensity
3
is equal to"
tohave
Φ(r
the particle
)
probability density P !
P12(r, (r1r, r)2=
! )= dr r3. .. .. . ddr rNexp
3 3 exp −− , 1j (1.7))
(r, r
2 of P
athaving
r2 .12Forparticle
convenience Z d of3 notation N we write
! k T
$ . (1.8)
Z1 at r and particle j at r . Thekprobability BT
B $ density
r1 =r,r 2 =r
! of having
!"
particle 1 at r and ! other particle
any at r equals 3N
#$
where P12 (r1 ,!r2 ) is1the probability Φ(rparticle) $$ 1 at r1 and particle!
!
3 density
3 to have
Because P12% all
(r,particles
r ) = aredequal, r3 . . .thisdisrequal N exp to− the probability density . P1j (r,(1.8) r)
2 at r2 . ForPconvenience (r, rZ !
) = of(Nnotation
− 1)P we rwrite
(r, !!
) k B T $
(1.9)
of having particle 1j 1 at r and particle12j at r . The probability density
r1 =r,r 2 =r
!
of having
!"
particle
• all 1particles
j" =at r! and
1
are 1any other
equal;3 prob.
particle
to have atparticle
r equals at r)and
3N
1 probability other atPr’:
$ anydensity
# $
Φ(r
! !
!
Because P% 1all
(r, particles
r ) = are dequal,
r . . . this d is
3
r equal
exp to − the . 1j (r,
(1.8) r )
12 3 N
$
! !!
ρg(|r
of having particle − r!1 at r and particle12j at !r). The k
Z |) = (N − 1)P (r, r BT
probability $ density of (1.10)
having
V P1j (r, r ) = (N − 1)P12 (r, r ) ! =r,r =r (1.9)
r !
1 2
particlej"=11 at r and any other particle at r equals
This %
Because isallequal
particles to the areprobability
equal, this isdensity equal to ofthehaving
probabilityparticledensity1 at r, (r, r! )
P1jwhich
is
of having1
simply 1/V
particle , !
! times
1 at the
and conditional
particle j at !!density
!
. The at !
,
probability
r which is
density ρg(|rof −
having
r !
|).
P (r, r )
ρg(|r1j− r |) = (N − 1)P12 =r (N − 1)P (r,
12(r, r )r r) (1.9)
(1.10)
particleVj"=11 at rby
Multiplying and N any we get other particle at r! equals
This is1% equal to !! the probability density ! of having particle 1 at r, which
ρ ρg(|r
2
g(|r − r
r |)
! ==
|) N(N (N − − 1)P1)P (r,
(r, r !r !)
). (1.10)
(1.11)
V 1/V , times the conditional density at r , which is ρg(|r −(1.9)
P (r, r ) = (N − 1)P (r, r )
12 !
is simply 1j 12
12 r! |).
j"=1
Multiplying
This is1 equal
We will need by to N the
this we get
probability
expression in the density
next subsection.of having particle 1 at r, which 9
! ! ! !
How do we get g(r) and what can we do with it?
• g(r) can be obtained from
• modern liquid theory
• computer simulation
• experimentally, through microscopy (for large particles) or scattering
• knowing g(r) we can calculate non-entropic thermodynamic functions, e.g.
• energy
• compressibility
• pressure
10
Once we know g(r), we can derive all non-entropic thermodynamic properties.
Energy
Relation between g(r) and energy
he simplest is the energy:
∞
3 1 N
!
int
U = U + NkB T + N dr4πr 2 g(r)ϕ(r). (1.12)
2 2 V 0
he first term originates from the internal

average energies
interaction of the molecules, the sec-
energy
internal energies
nd from the translations, and the third
of one from the
particular interactions. The average
molecule
of the molecules
otal potential energy equals 12 N with
timesall the
other molecules
average interaction of one partic-
lar molecule with all others; the factor 21 serves to avoid double counting.
translations
he contribution of all particles in a spherical shell of thickness dr at a dis-
of the molecules
ance r to the average interaction of one particular particle with all others is
πr 2 dr(N/V )g(r)ϕ(r). Integration finally yields Eq. (1.12).
Compressibility
he isothermal compressibility κT is defined as:
" #
1 ∂V
κT ≡ − (1.13)11
V ∂P T,N
4πr 2 dr(N/V )g(r)ϕ(r). Integration finally yields
Compressibility
Compressibility
The Relation between g(r)
isothermal compressibility κT and
isThe compressibility
defined as: [1/2]
isothermal compressibility κT is defined as:
" #
1 ∂V "
1 ∂V
#
•κ T ≡ − of isothermal compressibility: κT ≡ −
definition (1.13)
V ∂P T,N V ∂P T,N
From• compressibility
thermodynamics it istoknown
is linked From
that thermodynamics
spontaneous κ T can be linked
fluctuations it istoknown that κT ca
spontaneous
in number
fluctuations of particles
in the numberin of fluctuations
anparticles
open volumein an in the
V: open number
volume V ,ofsee
particles in an op
Fig. 1.4:
1. %STRUCTURE IN$ LIQUIDS AND 2 % GASES
$ 2%
#N$ ρkB T κT = (N − #N$)2 = N 2#N$−ρk B T 2κ,T = (N − #N$) = (1.14)
N − #N
$ % $
#N$
where the pointy brackets indicate a long time
where the pointy
conditions brackets
(in this indicate temperature
case constant a long time
manygives
average
T and
independent
or an
volume average
V ). over
From
configurations commensura
• assuming that g(r) is known, Eq. (1.11) us
many independent
Eq. (1.11) we configurations
obtain (where r
almost the needed expression:
commensurate
12 = |r 1 − r2 |): with the thermodynamic
10
! !
3 3 2
" 2#
10 d r1 d r2 ρ g(r12 ) = "N(N − 1)# = N − "N# . (1.15)
V V
We can use this to link the compressibility to the radial distribution function:
! ! ! !
"N# ρkB T κT = ρ d3 r1 ρ d3 r2 g(r12 ) + "N# − ρ d3 r1 ρ d3 r2
!V !V V V
12
3 3
nd
# the third from
conditions (in"this the#interactions.
case constant Thetemperature
average T and volume V ). From
1
N times the ∂V interaction of one partic-
1 average
Eq. (1.11)
κT ≡ −we
conditions
2 (in obtain
this (where
case r
constant12 = |r − r
temperature
1 2 |):(1.13)T and volume (1.13)
V ). From
s; T,N
the factor V
1 ∂P
serves to avoid double counting.
Eq. (1.11) 2 !obtainT,N
we (where r12 = |r1 − r2 |): " #
Relation between g(r) and compressibility [2/2](1.15)
!
cles
cs
From it in a spherical
3
isthermodynamics
known thatshell
3κ of
it2
canthickness
is be
known linked dr
that at
to κ a dis-
spontaneous 2
! d r1 ! d rT2 ρ g(r12 ) = "N(N − T can
1)# =be linked
N − to"N#
spontaneous
.
ction
mber of
fluctuationsofVone particular
particles
3in V in
the 3an particle
number open with all
2 of volume
particles V ,
in others
see
an Fig.
openis 1.4:
volume
" 2#
V−, "N#
see Fig. 1.4:
ration d
finally r d
1yields Eq.r ρ g(r
2 (1.12). 12 ) = "N(N − 1)# = N . (1.15)
$We canVuse2 %this$Vto2 $link the 2 compressibility
2% $ 2% to 2the radial distribution function:
• correct B byT adding and subtracting
%
(N − #N$#N$)ρk T =κ N= (N
− −
#N$ #N$)
, = N − #N$ ,
(1.14) (1.14)
We can use this to link the! compressibility ! to the radial distribution
! function:
!
where the 3 a 3 3 over! 3
"N#pointy
ckets indicate κbrackets
ρkBaTlong T time
= ρindicate
average
! d r 1or
ρlong
an
! d time
average
r 2
average
g(r over
12 ) + or
"N# an
− average
ρ ! d r 1 ρ d r2
many
ty κT independent
onfigurations
is
"N#defined Tas:
κ configurations
ρkcommensurate = ρ V 3 commensurate
with
d rtheρ thermodynamic
V 3
d r g(r with
) + the
"N# thermodynamic
− ρ V 3
d r ρ V 3
d r2
B T ! 1 ! 2 12 1
= ρ !V d3 r1 ρ !V d3 r2 (g(r12 ) − 1) + "N#V V
10
= ρ !V d3 r1 ρ !V d(1.13) 3
r2 (g(r12 ) − 1) + "N#
= ρ !V d3 r1 ρ !V d3 r (g(r) − 1) + "N# (1.16)
known that κT can be linked V to3 spontaneous
R3 3
= ρ d r1 ρ d r (g(r) − 1) + "N# (1.16)
f particles in an open volume VV , see Fig. 3 1.4:
Dividing by "N# we find R
2% $ 2% 2
Dividing
#N$) = by
N "N#
• dividing by − we
#N$ ,find
we find
! the compressibility (1.14) equation
3
ρk B
ndicate a long T κ = 1 + ρ
Ttime average or an
! d r (g(r) − 1)over
average . (1.17)
R3 3
ρkB T κT = 1 +
ations commensurate withρ thed thermodynamic
r (g(r) − 1) . (1.17)
R3
This so-called compressibility equation shows that the compressibility of a
fluid
This is intimately
so-called connected toequation
compressibility the radial shows distribution function of its con-
that the compressibility of a
stituent molecules. connected to the radial distribution function of its con-
fluid is intimately
13
stituent molecules.
Pressure
WeSTRUCTURE
1. will now consider the
IN LIQUIDS pressure
AND GASES of a fluid. If the density of the fluid is
We
not will now correlations
too high, consider the pressure
between threeoforamore
fluid.particles
If the may
density of the flu
be ignored,
not Relation
too high,
in which case Eq. between
correlations
(1.1) tells us g(r)
between
thatand pressure
thethree
radial or more particles
distribution functionmay be ign
is given
may
in
bywhichbe achieved
case Eq. by differentiating
(1.1) tells us the thatvirial
theequation to V :
radial distribution function is g
by • if density is not high,
! "
g(r) ≈ exp {−βϕ(r)} ∂P correlations
, V + Pbetween N be ignored. (1.18)
>2 particles may Then
= −NkB T B2 (T ) 2
∂V N,T V
whereg(r)ϕ(r)
1
≈ is exp
the
NkB T
{−βϕ(r)}
pair
! interaction, potential.
N
" Also for not N too high densities, (
−
the pressure +of a fluid is 1+toBa2good
(T ) approximation
= −NkB TgivenB2 (T )by 2the first two terms
where κT is the
ϕ(r) V pair interaction V potential. Also Vfor not too high dens
• atvirial
in the low density,
equation virial equation is an accurate expression for pressure
the pressure of a fluid $ is toρkaBgood approximation N given by the first two t
T κT = 1 − 2B2 (T ) .
% (1.20)
in thePvirial equation N V
V = NkB T 1 + B2 (T ) , (1.19)
Comparing the two expressions$ V the %
for compressibility, Eqs. (1.17) and (1.20),
we canPB write the second virial N
coefficient as a three-dimensional integral over
where V
• using = ) Nk
is T
called
B 1
the+ B
second
2 (T )
virial ,
coefficient. 3
2 simple mathematics (see Eq. 1.20) and the compressibility equation,link
(T Our goal now is to (
the pair interaction
B (T )wetoarrive
the radial ϕ(r):
distribution
at a microscopic
V
functionforg(r)
expression or pair virial
the second interaction ϕ(r). This
coefficient
2
3
where 3 B (T )
In principle
2 is called the
the 1virial equation
#
3
$ second
also contains higher order terms Our
−βϕ(r)
virial
% coefficient. in N/V goal
withnow
corre-is to
B2third,
sponding − etc,dvirial
(T ) =fourth, r e coefficients.
− 1 These
. become important at higher (1.21)
densities
B2 (T ) to the radial2 R3 distribution function g(r) or pair interaction ϕ(r).
than considered here.
3
TheInabove
principle the virial
equation equation
is important also contains
because it allowshigher order terms
us to calculate the in N/V with
pressure
sponding
of a fluidthird, fourth,
knowing onlyetc,the
virial coefficients.
pair interactionThese
ϕ(r)become
betweenimportant at higher
its constituent1114 den
than considered here.
In manyThetheories
pair of liquids the
interaction in hard
a hardsphere fluidfluid
sphere is usedis as a reference
given by system,
to which interparticle attractions
to The
which pairare added as attractions
interaction
interparticle ain perturbation.
a hard It
sphere
are isfluid
added therefore
asisagiven by
perturbation. I
to which interparticle attractions are added as a perturbation. It is therefore
useful to study the & radial distribution function, second virial coefficient and
useful
useful to study the ∞ radial r to
≤ σstudy
fordistribution the radial
& function, distribution
second function,
virial coefficient and second virial co
ϕ(r)
pressure of = sphere
a hard fluid. of a hard
pressure ∞ sphere
for r ≤fluid.σ (1.22)
Example: hard-sphere fluid [1/2]
0 for ϕ(r)
pressure of a hard sphere
The pair interaction inThe
r > σ=
fluid.
a hard
The pair interaction in a hard pairsphere fluid
0 fluidfor isis
interaction
sphere
given
r given
>
in by
σa hard
by sphere fluid is given by
At very low && densities the radial distribution function and second virial coef-
∞ for r ≤ σ
ϕ(r)
ficient =
are ∞
thereforeAt
for rvery
≤
given σ low
by densities
&
∞ the
for r radial
≤ σ distribution function
(1.22) and second
ϕ(r) =
• Pair interaction
0 for r > σϕ(r) = (1.22)
0 &forficient
r > σare therefore given
0 for r > σ by
0 for r ≤ σ &
AtAtvery
verylow
lowdensities
g(r) ≈
densities the
the radial
radial distribution
distribution0 function
for r
function ≤ σ and second
and secondvirial
virialcoef-
coef- (1.23)
ficient• g(r)
are at low
therefore density1
At
given for
very
by r
g(r) >lowσ≈ densities the#radial distribution function and secon
ficient are therefore given by 1% for r >σσ
1 given by 2 2 3 # σ
#
ficient 3 are
$ therefore
−βϕ(r)
B2 = 00−for d≤ rσσ e −11 = 2π $ drr = % πσ . (1.24)2
&&
r
#
g(r) ≈ 2for r ≤ 3 −βϕ(r) 3 (1.23) 2 3
• second virial
g(r) ≈ 11 coefficient
for r > B
σ = −
2and pressure
&1. d
at
STRUCTURE
0 for rlow
r e0
≤ density
IN
σ − 1
LIQUIDS = 2π
AND (1.23)
GASESdrr = πσ .
for r >g(r) σ ≈ 2 ## σσ1 0 3
3
According to 1Eq. 1 (1.19), and using % η1 =for6 πρσ r > σfor
2 the volume fraction of
##
33 $ −βϕ(r)
$ 2 2 3
BB
spheres,
2 ==
the −−
pressure dd r
r
According
ofee−βϕ(r) − 1 = 2π#
a hard to Eq.
sphere (1.19),
fluid drr
drrand
can
2 = πσ3 .
=be πσ
using . η
expressed = (1.24)
1 (1.24)
πρσ
as: 3
σ for the volume
The above expressions
2
22 are valid for not-too-high
1 00 3 $ −βϕ(r) densities.
3 At higher 2 3
#
3 % 6 2
nsities the probability to find B
spheres, the = hard
2 pressure
another − sphere ofda rhard
211πρσ
ein (near-)contact
− 1fluid
sphere = 2π can
with beadrr = πσas:
expressed
3
.
According to Eq. (1.19), and using η πρσ 33 for the volume fraction of
According P =
to ρk
Eq. T (1 +
(1.19), 4η)and . = for
ven hard sphere Bis higher than 1, and the 66pressure is higher than predicted the volume fraction 0of (1.25)
spheres,
spheres, the
thepressure
pressure of
of aa hard
hard
P = sphere
ρkto fluid
B TUsing can
can
(1 +(1.19),
4η) be
be expressed
. expressed as: calculated
one as:
the second virial coefficient
According alone. Eq. a computer and using η = 61 πρσ 3 for the volum
has
at 12 • pressure
the pressure at
formore
more general
general densities
densities (from
is givensimulations)
by: sphere fluid(1.25)
PP==ρk ρkBBTT(1(1++4η) spheres,
4η) .. the pressure of a hard can
(1.25) be expressed as
P 12
= 1 + 4η + 10η 2P+=18.365η ρk T
3
(1+ +28.24η
4η) .
4
+ 39.5η 5
+ 56.6η 6
+ ...
1212 ρkB T B
(1.26)
his is approximately 12 15
Example:
The above expressionshard sphere
are valid fluid
for not-too-high [2/2]
The above expressions are valid for not-too-high densities. At higher
densities. At higher
densities the probability to find another hard sphere in (near-)contact with a
densities
The the probability
above expressions to find
areanother
valid hardnot-too-high
for sphere in (near-)contact
densities. with
At a
higher
given hard sphere is higher than 1, and the pressure is higher than predicted
given hardthe
densities sphere is highertothan
probability find1,alone.
and thehard
another pressure is higher than predicted
by the second virial coefficient Using asphere in (near-)contact
computer with a
one has calculated
by thehard
second virialiscoefficient alone. Using apressure
computer one hasthan
calculated
that•
given Carnahan
the sphere
pressure and
for Starling
higher
more than
general1,noticed
and the
densities that
is the
given is
by:series
higher predicted
thatthe
by thesecond
pressure for more
virial generalalone.
coefficient densities
Usingis given by:
a computer one has calculated
P
that theP pressure for more
4η + 10η
2
2
general
+ 18.365η
3
3
densities + is given
28.24η
4
4 by:
= 1 + 4η + 10η + 18.365η + 28.24η + 39.5η + 56.6η + . .+
= 1 + + 39.5η
5
5
+ 56.6η
6
6
. ...
ρk
ρkPBBT
T
= 1 + 4η + 10η 2 + 18.365η 3 + 28.24η 4 + 39.5η 5 + 56.6η 6
+ ...
(1.26)
(1.26)
ρk T
is quite close to
B
This isis approximately
This approximately (1.26)
This is P P
approximately 2 2 3 3 4 4 5 5
==11++4η4η++10η
10η ++ 18η18η ++28η28η + 40η
+ 40η + 54η
+ 54η 6
+6. +
.. ... (1.27)
(1.27)
ρkρkBBTT
P 2 3
Extrapolating
Extrapolating =and
1and
+summing
4η + 10ηwe
summing +
we 18η
find
find + 28η 4 + 40η 5 + 54η 6 + . . . (1.27)
ρkB T
whichP can
P = 1and
Extrapolating
!be
∞∞
summed
summing
2 2 we
analytically
1
n find +1 η
+ +η η+
2 2 3to3
− η− η
η
= 1++ (n(n++3n)η 3n)η= = (1 − η)3 3 . . (1.28)
!
n
ρk T (1.28)
ρkB T
B n=1∞ (1 − η)
P n=1
!
2 n 1 + η + η2 − η3
This =Carnahan
1+ (nand+ 3n)η = equation for the. the (1.28)
This isisρkcalled
called
B T Carnahan and Starling’s
Starling’s equation
(1 − η) 3 for pressure
pressureof aofhard
a hard
sphere fluid. Monte Carlo simulations of hard sphere fluids have shown that
n=1
Eq. (1.28)
This is nearly
is called exact at
Carnahan andall Starling’s
possible volume fractions.
equation for the pressure of a hard
Eq. (1.28) is nearly exact at all possible volume fractions.
Eq. (1.28) is nearly exact at all possible volume fractions.
1.4 Scattering and the structure factor
1.4 Scattering and the structure factor
In the precious section we have linked the compressibility of a fluid to spon-
1.4
taneous
Scattering
In the precious section
fluctuations in the
and
we have the
linked
number thestructure
compressibility
of particles
factor
of a fluidMore
in a large volume. to spon-
taneous fluctuations
generally, in the number
density fluctuations in a fluidofcan
particles in a large
be described volume.
by means of theirMore 16
In the precious section we have linked the compressibility of a fluid to spon-
generally,
Fourier density fluctuations in a fluid can be described by means of their
components:
Scattering and the structure factor
kin
kout
• illuminate a fluid containing spherical particles by a wave (wavevector kin )
• X-ray or neutron scattering for atomic liquids
• small-angle X-ray, small-angle neutron or light scattering for colloids
• measured intensity in the direction kout depends on density fluctuations in

the fluid, as probed by the structure factor S(k), with k = kout − kin
N.B. intensity also depends on other factors such as wave polarization,

scattering strength and form factor, i.e. internal particle structure
17
1.4
1.4 Scattering and the
Scattering and the structure
structure factor
factor
In
Fourier
In the
the
taneous
precious
precious analysis
section we
section of
we have
density
have linked
linked the
fluctuations
thecompressibility
compressibilityofofaafluid
fluidtotospon-
spon-
taneous fluctuations
fluctuations in in the
the number
number ofof particles
particles inin aalarge
largevolume.
volume. More More
generally, density fluctuations
generally, density fluctuations in in aa fluid
fluid can
canbe bedescribed
describedby
bymeans
meansofoftheir
their
Fourier components:
• describe density fluctuations in fluid by means of Fourier components:
components:
11
""
ρ(r) = ρ + + d33 k ρ̂(k) exp {−ik · r} , (1.29)
3 d k ρ̂(k) exp {−ik · r} , (1.29)
(2π)3
(2π)
"
ρ̂(k) = dd33rr {ρ(r)
{ρ(r) − −ρ} ρ}exp
exp{ik{ik· ·r}
r}. . (1.30)
(1.30)
STRUCTURE IN LIQUIDS AND GASES
4 4
The•microscopic
corresponding microscopic
variable
variable variableto
corresponding
corresponding toaadensity
densityFourier
Fouriercomponent
componentisis
# NN
# $$
ere δ(r) = δ(x)δ(y)δ(z)
! is the three-dimensional Dirac delta-function.
" !
ρ̂(k) = d33rr δδ (r
(r−−rrjj))−
−ρρ expexp{ik
{ik· ·r}
r}, , (1.31)
(1.31)
is may be rewritten as
j=1
j=1
44 In N
In order to avoid
avoid overly
overly dressed
dressedsymbols,
symbols,3we
weuse
usethe
thesame
samesymbol
symbolfor
forthe
themacroscopic
"
! macroscopic
ρ̂(k) and
quantity
quantity = the microscopic
exp {ik · rvariable.
microscopic } − ρ InIn
jvariable. d general
rgeneral
exp {ik · r}
aa microscopic
microscopicvariable
variableAAmicr
micris an
is an
expression givenj=1
expression explicitly
explicitly in
in terms
terms ofof positions
positionsand/or
and/orvelocities
velocitiesofofthe
theparticles,
particles,
% % micrwhich
which
&&
after ensemble
after ensemble averaging
averaging
N yields
yields the
the corresponding
corresponding macroscopic
macroscopic quantity
quantity A,A,i.e.
i.e. AAmicr ==
micr
'
A. For example the microscopic density at r is given by ρ micr(r) = j δ(r −−rjr),j ),and the
! '
A. For example
= the microscopic density
exp {ik · r%j }micr
% − (2π)
&at 3r is given by ρ
ρδ(k). (r) = j δ(r and
(1.32)the
macroscopic density by ρ(r) = ρρmicr(r) ..
&
macroscopic density
j=1
by ρ(r) = (r)
Density fluctuations in a fluid can be measured experimentally by means

1313 18
∗
correlationand The
function division
of ρ̂(k) by
with N
its leads
complex to
scattering strength or form factor, a quantity
conjugate ρ̂ which
(k)but fori.e.
= ρ̂(−k), largescatte
generally enoug
the mean square of the density
pendent of fluctuation
system size with wave
(that is to vector
say, k, ismean
the a real square de
function ofmeasure correlation
the wavevector, called the functions of Fourier
structure factor S(k): components of t
grow linearly withofsystem size).itsThe structure factor give ∗
Relation between
correlation
1tion about structure
function factor
ρ̂(k) with and g(r)
complex
the structure of a fluid. It is essentially a Fo
conjugate ρ̂ (
the
S(k) ≡ mean ∗square of the density fluctuation with wave ve
#ρ̂(k)ρ̂ (k)$ . (1.33)
Nthe radial distribution function, as can be shown as follow
function of the wavevector, called the structure factor S(k)
• (static) scattering experiments # measure amplitudes of density fluctuations:
The division by N leads to a quantity which for large enough systems
N N
$is inde- "
2
1 ρ
!! "
pendent of system size
S(k) = 1 (that is to say, the mean
exp square
{ik · (rdensity
− r fluctuations
)} − d 3
r
∗ j k
grow linearly with S(k)system≡ N #ρ̂(k)ρ̂
size). The structure (k)$factor
. gives a lot of informa- N
tion about the structure of aNfluid. #
j=1 k=1
It is essentially a Fourier transform $ of
N N
• the
the radial distribution function, as1can ! be shown
structure factor S(k) is essentially
! as
a Fourier follows: of g(r)
transform
"
3
The division = by
1 +N leads to a quantity
exp {ik · which
(r j − rfor
k )} large
− ρ enough
d r
N j=1 k#=ρj2
#N N $
1
" "
S(k) = pendent of expsystem size (that is to3 say,3 the mean square den
!!
" "
{ik · (rj − r k )} − d r d r exp {ik · (r − r )}
N j=1 k=1 " N
grow #linearly with system 3 size). The structure factor gives
N ! = 1 + ρ d r [g(r) −"1] exp {ik · r} .
N
$
tion1 about the structure of a fluid.3 It is essentially a Fou
!
= 1+ exp {ik · (rj − rk )} − ρ d r exp {ik · r}
theNradial j=1 k#distribution function, as can be shown as follows
Comparison
"
=j
with Eq. (1.17) shows, perhaps surprisingly, t
= 1 + ρ ibility 3
of a−fluid
d r [g(r) 1] exp can
{ik · be
r} . obtained not only by compressing
(1.34)
# N N
$
2
1 !! ρ
" "
S(k)by = performing a scattering exp {ik experiment:
· (rj − rk )} − d3 r d
• Comparison
so you can measure
with Eq. the Ncompressibility
(1.17) shows, perhapsofsurprisingly,
j=1 k=1
a fluid without thattouching N
it!
the compress-
ibility of a fluid can beρk obtained
T κ not
= onlyS(k).
lim by compressing the fluid, but also
B T
# N N
$
by performing a scattering experiment:1 ! k→0!
"
3
19
= 1+ exp {ik · (rj − rk )} − ρ d r e
each molecule to be conditions (in thiswith
a rigid object casea constant
given position temperature T and volum
and orientation.
This may be rewritten as
Eq. statistical
(1.11) we obtain (where r12 positions
= |r1 − r2and |): orientations
According to classical mechanics, these
1
are not completely N random.
! Rather,
! at
" not too low temperature T , the
Summary = offorexp chapter
{ik ·1 rj }1−a2ρcertain
3 3 2 2
! " #
probability
ρ̂(k) density d r
encountering d r ρ g(r 3 )
d12r exp= "N(N
configuration − 1)# =
{ik · r} of 3N positionN − "N# .
V V 6N
and 3N orientation j=1coordinates, in shorthand denoted by R , is given by
We can use
the Boltzmann distribution
N
this to link the compressibility to the radial distribu
function:
• the structure ! of fluids is determined by the3!interparticle ! interactions !
= exp {ikρk·BrTj }κT"
!"N# − (2π)
= ρ ρδ(k).
d 3
r 1 ρ d 3
r 2 g(r 12 ) + "N# − ρ (1
d
6N
6N 1j=1 Φ(R ) V V V
P (R ) = exp − , ! ! (1.1)
Z kB T = ρ d 3
r ρ d 3
r2 (g(r12 ) − 1) + "N#
Density fluctuations in a fluid can beVmeasured 1
V
experimentally by me
• radial distribution
of scattering of light,function
neutrons, g(r): density
or X-rays at distance
!
(depending r from aon
! giventhe particle,
scale of intere
6N 3 3
wherenormalised
Φ(R ) is by thethe
total potential
average densityenergy
=
N/V. of
ρ the
Knowing d configuration,
r ρ d
1g(r), we canr (g(r) kB−=1)1.38065×
calculate +
e.g."N#
The scattered intensity also depends on details such as wave polariza
10−23 J/K = 8.617 × 10−5 eV/K is Boltzmann’s V constant, R3 and Z is a normal-
and scattering
• total strength
energy Dividing
isation constant, referred
or
to by
form
as "N#
factor,
we find
the configuration
but generally
integral.
scattering experim
measure correlation
The• positions
functions of Fourier components of the density.
pressure and orientations are not completely random because
! ∗
the to-
correlation
tal potential function ofρk
energy Φ containsρ̂(k)T κ with
terms
= 1 +its
which
ρ complex
d 3
depend
r (g(r) conjugate
on −the
1) . ρ̂ (k)
relative = ρ̂(−k),
positions
B T
the mean• compressibility
square of the density fluctuation3 with wave vector k, is a
and orientations of two or more molecules. The details R of such molecular
function
interactions of determine
• scattering theexperiments
wavevector,
This the probe
so-called called
precise Fourier the
structural structure
components
compressibility offactor
andequation
dynamical
density
showsS(k):
properties
that theofcompre
fluctuations a
fluid. fluid is intimately connected to the radial distribution functio
1 stituent∗molecules.
1 S(k) ≡ #ρ̂(k)ρ̂ (k)$ .
At very low temperatures the discreteness of the energy levels becomes apparent. In (1
N 20
that case the classical view needs to be replaced by a quantum mechanical one and other
Chapter 2: Time dependent properties of liquids
21
Time dependent
2.1 Time correlation functions
proper
liquids
Dynamic properties of liquids
In the rest of these lectures we will focus on time dep
liquids, i.e. their dynamics. Even when a fluid appears
• even when a liquid appearsscopically, microscopically
to be at rest macroscopically, the molecules are continually
microscopically the
sitions and
molecules are continually changing velocities.
their positionsMostand observable
velocities quantities therefo
2.1
and we needTime a waycorrelation
to characterise thefunctionsdynamics of the
• therefore most observable do
quantities
this by fluctuate
means ofintime timecorrelation functions. For a q
In the rest function
correlation of these islectures
definedwe as will focus on time dep
liquids, i.e. their dynamics.! Even T when a fluid appears
e.g. velocity of a 1
diatomic particle
scopically, microscopically
!A(t)A(0)" = lim the dτ molecules
A(τ + t)A(τare ).continuall
sitions(inand
a fluid) along
T →∞
velocities. its Tbond
Most direction quantities there
0observable
and
When wet need
= 0, awe waygettothecharacterise
average ofthe A2 .dynamics
When t of→th∞
do this bytomeans
unrelated A(τ ) of
andtimethe correlation
result will be !A"2 . From
functions. ForSch
a
• characterise the dynamics correlation
of fluctuations by time correlation functions
get !A(t)A(0)" ≤ {!A(t)A(t)" !A(0)A(0)"}1/2 = !A2 ". So
function is defined as
from !A2 " at t = 0 to !A"12 for ! T very large times.
A !A(t)A(0)" = lim
simple generalisation
T →∞ T 0
of dτ A(τ
Eq. (2.1)+ist)A(τ ).
! T
t = 0, we=getlimthe1 average
When!A(t)B(0)" of A 2
. When
dτ A(τ + t)B(τ ). t →
unrelated to A(τ ) and T →∞ theT result
0 will be !A"2 . From Sch
1/2
get
When B = A, as in Eq. (2.1), we speak about=autocorre
!A(t)A(0)" ≤ {!A(t)A(t)" !A(0)A(0)"} !A2 ". S
2 2 22
fromIn!A " at t = 0 to !A" for very large times.
equilibrium, the origin of time is irrelevant, which m
rrelation functions.
2 For a quantity
2 A, the time correlation function is defined as
from !A " at t = 0 to !A" do for
this very
by large times.
means of time correlation functions. ! For
T a quantity A
d as 1
T
A simple generalisation of Eq. (2.1)
correlation functionis is defined !A(t)A(0)"
as = lim dτ A(τ + t)A(τ )
T →∞ T 0
dτ A(τ + t)A(τ ). 1
! T (2.1) 1
! T
0
Properties of time-correlation functions
!A(t)B(0)" = lim
2 T →∞ T 0
dτ A(τ + t)B(τ
!A(t)A(0)" = lim ).
When
T →∞ T 0
t =dτ0,A(τ we + gett)A(τ (2.2) of A2 . Wh
the ).average
2
verage of A2. .TIME When t → ∞, A(t
DEPENDENT + τ ) will OF
PROPERTIES unrelated
be LIQUIDS to A(τ ) and the result will be !A" .
When
sult will B
be=!A" A,2 .asFrom
in Eq. (2.1), we
When
Schwarz’s t =speak0, we
inequality about
wegetget autocorrelation
the average of≤ A
!A(t)A(0)" 2
functions.
. When !A(0)A(0)"}
{!A(t)A(t)" t → ∞, A(t 1/2
+=
2 2
In
" !A(0)A(0)"} 1/2
• it is easy to see that
equilibrium, = !A the
". So, !A(t)A(0)" decays from
2 origin of time is irrelevant, which
unrelated to A(τ
decays ) and the
from result
!A " at t=0 to
2 means that !A(t
atwill
t = be
0 !A"
to !A".for
+ large
From
fors)B(s)"times
Schwarz’s
very ine
large times
orisvery from
independent
large which
times. of s. we obtain
Hence, the important
using a dot over property:
a quantity
A simple indicate1/2its= time
togeneralisation of
get !A(t)A(0)" ≤ {!A(t)A(t)" !A(0)A(0)"} !AEq.
2
". (2.1) is
So, !A(t)A(
Eq. (2.1)2.isTIME
of derivative, the DEPENDENT
following must
from
PROPERTIES
be
!A 2 true inOF
" at t = 0
LIQUIDS2
equilibrium:
to !A" for very large times.
• in equilibrium, the originPROPERTIES
of time is irrelevant, so
! T
!
2. TIME DEPENDENT
" !
OF LIQUIDS
" 1
T
d Ȧ(t + s)B(s) = − A(t
"A simple +
# s)Ḃ(s)
generalisation
" !A(t)B(0)"
of Eq. (2.1)# = islim dτ A(τ + t)B(τ )
(2.4)
T →∞ T 0
dτ A(τ + !A(t
t)B(τ
from + s)B(s)"
).
which we =
obtain Ȧ(t +
the s)B(s) (2.2)
important + A(t
property:+ s) Ḃ(s) = 0, (2.3)
0 ds When
! T
1 B = A, as in Eq. (2.1), we speak about
whichForfrom
gives
an whichthe we
autocorrelationobtain
important the important
property
!A(t)B(0)"
function =
this property:
lim
says that dτ A(τ + t)B(τ ).
), we speak about!autocorrelation " functions.! T →∞ In T"equilibrium,
0 the origin of15 time is irrelevan
of time is irrelevant,Ȧ(t!which+ s)B(s)
means "= −!A(t
# that !A(t + s)Ḃ(s)
+ s)B(s)" is independent
" of s. Hence, using a(2.4)
dot over a
d
using a dot over a quantity When
Ȧ(t + s)B(s)
to B
indicate
# = A,
= −itsA(t as
time in Eq.
+ s)Ḃ(s) (2.1), we speak about autocorrelation
(2.4) fun
"A(t)A(0)#In = 0, derivative, the following must be true in equili
(2.5)
equilibrium, the origin of time is irrelevant, which means that
#
dt
st be true in equilibrium: #
t=0
For an autocorrelation function
is independent this says that
of s. Hence, using d a s)B(s)"
dot over=a quantity
"
to indica
# "
• with this we can also derive
For
# an " autocorrelation # function this says that !A(t + Ȧ(t + s)B(s) +
Ȧ(t + s)B(s)i.e. any
+ A(t time derivative,
# = 0, the
+ autocorrelation
s)Ḃ(s) following
function
(2.3) starts dsout
must bewith
trueainhorizontal
equilibrium: slope.
d #
Fromd"A(t)A(0)#
Eq. (2.4) we # ## can
d = also
0, derive the useful
" relation # " (2.5) #
dt "A(t)A(0)# # # = 0,
t=0 !A(t + s)B(s)" = Ȧ(t + s)B(s) + A(t + s)(2.5) Ḃ(s) = 0,
dt # ds 15
t=0
d2 ! "
i.e. any 2time"A(tautocorrelation
+ s)B(s)# = − function
Ȧ(t + s)starts
Ḃ(s) out
. with a horizontal slope. (2.6)
dt time autocorrelation function starts out with a horizontal slope.
i.e. any
From Eq. (2.4) we can also derive the useful relation
From Eq. (2.4) we can also derive the useful relation
We will need this in section 2.3 when dealing with collective diffusion. For
2
d
an autocorrelation
2 function
!
this!Ȧ(t
says that
"
23
d
2
"A(t + s)B(s)# = − ˙ + s) Ḃ(s)
˙
". (2.6)
2.2 Self-diffusion and the velocity autocorre-
i.e. any time autocorrelation function is initially curved neg
lation function
Example: self-diffusion
Suppose we label some particles inside a very small region (a dot) in an
2.2 Self-diffusion and the velocity
otherwise homogeneous fluid, at time t = 0 at position r(0). When the dot,
a
lation
although on a macroscopic scale function
concentrated at r(0), is dilute enough on
a molecular scale, we may consider the concentration decay as due to the
self-diffusion of the separate
Suppose labelled
we labelparticles. The conditional
some particles probability
inside a very small regio
P (r, t) that a particle is at r at
otherwise time t, givenfluid,
homogeneous it was
at at
time t =at0time
r(0) t = 0, r(0)
at position
may then be obtained from Fick’s
although on alaw:
macroscopic scale concentrated at r(0), is d
• at t=0 we label some particles in a dot of fluid;
a molecular scale, we may consider the concentration deca
each particle
∂P (r, t) will self-diffusion
diffuse away via a different path
= D∇2 P (r, t), of the separate labelled particles. The condit
(2.8)
∂t P (r, t) that a particle is at r at time t, given it was at r(0
• define conditional probability P(r,t) that a labeled particle is at r at time t,
givenwith
together may
at r(0)
it wasthe
then be
at time t=0
boundary obtained from Fick’s law:
condition P (r, 0) = δ(r − r(0)). D is the self-
diffusion coefficient, which has units of length squared over time (m2 /s).
∂P (r, t) 2
The• Fick’s
mean law states
square that
displacement of the labelled
= D∇ P (r,particles
t), can be related to the
∂t � 2�
m
16 D = self-diffusion coefficient
together with the boundary condition P (r, 0)s =
δ(r − r(0)
diffusion coefficient, which has units of length squared ov
The mean square displacement of the labelled particles 24can
2. TIME DEPENDENT PROPERTIES OF LIQUIDS
Mean-square displacement and Einstein equation

self-diffusion coefficient as follows: 2. TIME DEPENDENT PROPERTIES OF LIQUIDS
d 2 ∂P (r, t)
#
! 2 " 3
|r(t) − r(0)| = d r |r(t) − r(0)|
dt
self-diffusion
• averaging coefficient
over all as follows:
possible ∂t
# paths, we find:
d 3 ∂P 2(r,2t)
#
!
|r(t) − r(0)| 2 "
== D d r |r(t) −
d3 r |r(t) − r(0)| r(0)|
2 ∇ P (r, t)
dt ∂t
##
3 2 2
= D d
3 rP (r, t)∇ r 2
= D d r |r(t) − r(0)| ∇2 P (r, t)
= 6D, # (2.9)
3 2 2
= D d rP (r, t)∇ r
where we have used partial integration and the fact that P (r, t) and its
derivative are zero far from = r(0).
6D, For real fluid particles Fick’s law only(2.9) holds
for •large
for avalues
real fluid,
of t. 1 Fick’s law only holds at large times (after many collisions)
Integration of Eq. and(2.9)the
yields
where we have used partial integration factthe Einstein
that P (r, t)equation
and its
derivative are zero far from r(0). For real fluid particles Fick’s law only holds
for large 1 !1 2
D values
= lim of t. |r(t)
Integration
− r(0)| of. Eq. (2.9) yields the Einstein equation
"
(2.10)
t→∞ 6t
1 ! 2"
We may D= lim
transform |r(t)
this − r(0)| .to an expression involving the autocorrela-
equation (2.10)
Einstein
t→∞ 6t equation
tion of the velocity v = ṙ of a labelled particle:
We may transform this equation
# t to # an
t expression involving the autocorrela-
− r(0)|2v ==ṙ of a labelled
dt# ##
#v(t# ) · v(t## )$
! "
tion of the
|r(t)velocity dtparticle:
# 0t# # t0# !
t t
2" 25
== 2 dt dt dt## dt
# # ## # ) · v(t
# ## )$ ##
!
|r(t) − r(0)| #v(t #v(t ) · v(t )$
self-diffusion coefficient as follows: |r(t)
1 − r(0)|
=the D 3=
dofrPEq (r
We may transform for large
this equation values
to an expression of
dt t. Integration
involving autocorrela-
ansform this
1 ! equation
tion of the 2"
velocity
|r(t) − r(0)| . d ! v =
to ṙ
an
of a
expression
labelled particle:
# involving the autoc
2 ∂P (r,
D = lim (2.10)
t→∞ 6t 2" 3
velocity v =! ṙ of a labelled |r(t) # t −particle:
r(0)| = d r |r(t)
= −6D,
r(0)| = D
We Connection with to the = velocity dt autocorrelation
dt ∂t
# t
may transform this equation
|r(t) − r(0)| 2"
an expression #
dt involving ## #
)1 ##
the autocorrela-
#v(t · v(t )$
! 2 "
0 D = lim |r(t) 3 − r(0)| .
#
tion of the velocity v = ṙ of#a labelled #particle: 0 2
t where t we have = usedD d r
partial|r(t) −
integration
r(0)| ∇ 2
P
2 " t t # ## = #2 ##dt
# t !
#
# t t→∞ 6t
### #v(t# ) · v(t
## ## )$ = D
− r(0)| |r(t) − r(0)| == dt dtdtderivative
#v(t )dt #v(t ) · v(t )$
# #
! 2" # ## dt
0are zero far from r(0). For real
· v(t
0
)$ #
0 0
#0t 0 #t # t !
= D 1 d3
rP (r, t)∇ 2 2
r
= 2 # #
# t
We for
!
=# 2
## may
large
dt t dt #v(tt)! · v(t )$# dt transform
###values ## of this
# t.##
dt #v(t − t ) · v(0)$ equation
Integration oftoEq.
= 6 an(
0 0
= 2
0
# t
0
tion
# t! #
dt of
# t ## the velocity
#= 6D, ##
dt #v(t ) · !v(t )$
# t! v = ṙ of a labelled
= 2 dt# dt=##
#v(t #
− t## #
dt) · v(0)$ dτ #v(τ ) · v(t## )$
0 0 0
2
0
where 1we have used 2 partial
" i
# ! D =derivative lim |r(t) − r(0)| .
0 0
where
# t
we
# t!
have#! t used partial integration and t the
# tfact #
= 2 dt#t dτ=#v(τ
# t) · v(0)$ t→∞ 6t2 . "are zero far from r(0
#(2.11) #
derivative
#0 t 0
2
# are zero |r(t) − r(0)|
dτ (t − τ )
##far from
#v(τ ) ·
# r(0).
v(0)$ =
## For real fluid
1 particles dt dt
= 2 = for
2 large
dτ
dt
(t − τvalues
0
dt for #v(tlarge
of t.. 1 Integration
) #v(τ ) · v(0)$
− tvalues ) · v(0)$
0
of Eq. (2.9)
(2.11)
of t. Integrati
0 the
yields
The last step follows
0 0 We after 0may
partial transform
integration. Comparing this equationwith # to an
the tEinstein# tex !
equation (2.10), taking the limit for t find
→ ∞, we finally find
The•lastcomparing
step followswith afterthe
#Einstein equation,
tintegration.
tion of thet! we velocity v = ṙ of1 a labelled # pad
partial Comparing with the Einstein
= 2 dt
#
1
!
equation (2.10), taking the limit
=D = 21 dtdt# ∞forD t →# ∞, we finally find
= lim dτ #v(τ
!
|r(t) − ) ·D v(t
r(0)| = 2 ##lim
"
)$
. |r(t)
0 #
− r(0)
0
∞
t→∞
#v(t) · v(0)$ 6t. t→∞ 6t
# t# (2.12)
t# !
1 3 00 0 t# t
# ! 2 (2.12)
" ##
D=
3
dt #v(t) · v(0)$# t . |r(t) − r(0)| = dt # dt #v
this is the
0
is the We
TheGreen-Kubo may
Green-Kuborelationtransform
for thefor
relation this equation
Weself-diffusion
maycoefficient
self-diffusion
the to=an expression
transform
coefficient.
20 thisdt0equatio invold
The is the Green-Kubo = relation
2 forofdτ
tion the
the (tvelocity
− τ ) coefficient.
self-diffusion #v(τ
v = ) of· v(0)$
ṙ a labelled. particle:# t0 # t!0
At very short times the fluid particles tion
1
are not of the
yet moving velocity v
completely randomly. = ṙ of ta! #
For
0 t #
= 2 dt # # dt
1
At very short times thethey
fluid may
particles
# #
example, stillare
be not yet moving
trapped insidecompletely
a temporaryrandomly.
cage For
formed by their neighbours.
example, they may still be trapped inside a temporary cage formed by their t
neighbours. t 26
= ## 2 0 # dt 0 ## t d
# #
Fick’s law applies to time scales on which " the particles are diffusing freely.
! 2
Fick’s law applies to time scales on which the particles are diffusing freely. #
0 0
# t # t!
= 2 dt# dτ #v(τ
#0 t 0
Velocity autocorrelation in a Lennard-Jones
= 2
fluid
dτ (t − τ ) #v(τ
0
The last step follows after partial integration.

equation (2.10), taking the limit for t → ∞, w
∞
1
#
D= dt #v(t) · v(0)$ .
3 0
The is the Green-Kubo relation for the self-di

1
At very short times the fluid particles are not ye
example, they may still be trapped inside a temporar
Fick’s law applies to time scales on which the particle
reversed velocity
particle collides with another
particle in front of it
27
by a macroscopic law. We shall generalise this approac
In the previous section we made
regression use of the fact that the microscopic self-
hypothesis.
diffusion of a labelled particle
Considerin an
a liquid may for
observable large times
quantity be described
A having zero mean,
Onsager’s
by a macroscopicregression
law. We shallhypothesis
generalise this approach using Onsager’s
fluctuations, however, A will have a nonzero value at (alm
regression hypothesis.
Onsager’s regression hypothesis says that the decay of t
Consider an observable quantity A having zero mean, !A" = 0. Due to
fluctuations, ”the
large times and on macroscopic scales will be governed by
average
however, A will decay of fluctuations
have a nonzero is,
value at (almost) all instants.
Onsager’s at
regressionmacroscopic
hypothesis laws.that the decay of this fluctuation at
says
large times and on macroscopic scales,
large governed Notice that
times and onbymacroscopic scalesthewill
macroscopic
be governedlaws
by usually
the apply to non
corresponding
the corresponding macroscopic laws”
macroscopic laws. ations. In formula Onsager’s hypothesis says
Notice that the macroscopic laws=usually
!A(t)"A(0) apply to non-equilibrium situ-
A(0)α(t),
ations. In formula Onsager’s hypothesis says
where α(t) is determined by macroscopic laws. The avera
• macroscopic
!A(t)"A(0)laws [the ones that determine α(t)]
= A(0)α(t), (2.13)
average: it expresses the average time development of A,
usually apply to non-equilibrium situations
where α(t) is determined
A(0) at by tmacroscopic laws. The
= 0. Multiplying by average
A(0) and is aaveraging
conditional
over al
average: it expresseswethegetaverage time development of A, given that it was
• multiplying
A(0) at t = 0. A(0) and averaging
byMultiplying by A(0)over
andallaveraging
!initial conditions
over all initial conditions
we get a useful equilibrium expression: 2
"
we get !A(t)A(0)" = A α(t).
! 2"
!A(t)A(0)" = The
A averages
α(t). are now simple equilibrium averages.
(2.14) We s
use of Eq. (2.14) with two examples in the next two sectio
The averages are now simple equilibrium averages. We shall illustrate the
use of Eq. (2.14) with two examples in the next two sections. 28
18
fluctuations by means of their Fourier components, but now with an explicit
time dependence:
2.4.1 Decay of macroscropic density fluctuations
In the first example, we 1 will focus on the time dependence of density fluc-
Application 1: collective diffusion
!
3
ρ(r,
tuations int)a liquid.
= ρ +Following
3
dEqs.
k ρ̂(k, t) exp
(1.29) and{−ik · r}we
(1.30), , describe density
(2.15)
(2π)
fluctuations by means! of their Fourier components, but now with an explicit
ρ̂(k, t) =
time dependence: d3 r {ρ(r, t) − ρ} exp {ik · r} . (2.16)
• we’re interested in the
! time dependence of density fluctuations
1 3
ρ(r,Onsager’s
To apply t) = ρ +regression
3
d k ρ̂(k, t) expwe
hypothesis, {−ik
first· r} , to know how(2.15)
need a macro-
(2π)
scopic density fluctuation
! decays. Suppose at time t = 0 we prepare a fluid
system with
ρ̂(k, d3 r {ρ(r,sinusoidal
t) a=macroscopic t) − ρ} exp density
{ik · r}fluctuation.
. In the above equa-
(2.16)
tions this corresponds to a situation in which there is only one non-zero
• in order
Fourier to apply
component ρ̂(k,Onsager,
0): let’s first study the decay of a macroscopic
To apply Onsager’s regression hypothesis, we first need to know how a macro-
sinusoidal density fluctuation prepared at t=0
scopic density fluctuation decays. Suppose at time t = 0 we prepare a fluid
systemρ(r,
with a 1 sinusoidal density fluctuation. In the above equa-
macroscopic
0) = ρ + 3
ρ̂(k, 0) exp {−ik · r} (2.17)
(2π)to a situation in which there is only one non-zero
tions this corresponds
Fourier component ρ̂(k, 0):
According to Fick’s law, Eq. (2.8), the rate of decay of such a macroscopic
density fluctuation is1 determined by the so-called collective diffusion coeffi-
ρ(r, 0) = ρ +
cient D: (2π) 3
ρ̂(k, 0) exp {−ik · r}
? (2.17)
" # " #
∂ to1Fick’s law, Eq. (2.8), the rate of decay
According 2 1of such a macroscopic
ρ̂(k, t) exp {−ik · r} = D∇ ρ̂(k, t) exp {−ik · r} ,
density∂tfluctuation
(2π) 3 is determined by the so-called collective
(2π) 3 diffusion coeffi-
cient D: ∂ ρ̂(k, t) 2
exp {−ik · r} = −D(k)k ρ̂(k, t) exp {−ik · r}#, 29
∂
"
1 ∂t# "
1
2 $ 2
%
1
ρ(r, 0) = ρ + 3
ρ̂(k, 0) exp {−ik · r} (2.17)
(2π)
Decay of macroscopic density fluctuations
According to Fick’s law, Eq. (2.8), the rate of decay of such a macroscopic
density fluctuation is determined by the so-called collective diffusion coeffi-
cient D:
• Fick’s law applies here too, but now D is the collective diffusion coefficient
" # " #
∂ 1 2 1
3
ρ̂(k, t) exp {−ik · r} = D∇ 3
ρ̂(k, t) exp {−ik · r} ,
∂t (2π) (2π)
∂ ρ̂(k, t)
exp {−ik · r} = −D(k)k 2 ρ̂(k, t) exp {−ik · r} ,
∂t
2
$ %
ρ̂(k, t) = ρ̂(k, 0) exp −D(k)k t . (2.18)
Note• athat we have included

macroscopic densitythe possibilitydecays
fluctuation that theexponentially
collective diffusion co- times)
(at large
efficient
withdepends on thetime
a relaxation wavegiven
lengthby of the density disturbance, D = D(k).
Eq. (2.18) shows that a density fluctuation smoothens out with a relaxation
2 1
time τ (k) = 1/(D(k)k ). Large τ (k) k=(short wavelength)
2
fluctuations decay
rapidly, whereas relaxing small k (longD(k)k wavelength) fluctuations can take a
very long time. This is a consequence of the fact that relaxing a long wave-
lengthshort wavelength
inhomogeneity requires k) fluctuations
(large transport decay
of fluid rapidly
particles over large length
scales,long wavelength
which (small k) fluctuations decay slowly
is a slow process.
(relaxing a long wavelength inhomogeneity requires transport of19

fluid particles over large length scales, which is a slow process) 30
for the
forcollective diffusion
2.4.2
the collective coefficient.
Microscopic
diffusion We have
coefficient. Weencountered
equationhave encountered fortheD(k) microscopic
the microscopic
variable corresponding
2.4.2
variable to a density
Microscopic
corresponding to a densityFourier
equation component
Fourier for
component D(k)
before,before,
see Eq.see (1.32).
Eq. (1.32).
We now add an Let us now
explicit time see if we can derive a microscopic (and equilibrium) equati
dependence:
We now add an explicit time dependence:
Let us for nowthe see if we candiffusion derive a$ microscopicWe (and equilibrium) equation
Microscopic equation for collective D [1/2]
2. TIME DEPENDENT
for the variable
collective
!
3
"
! #
collective
"PROPERTIES
diffusion
# coefficient.
corresponding
OF
coefficient.
LIQUIDS
to a density We have
$
Fourier
have
encountered
component
encountered
thebefore,
the microscop
microscopic
see Eq. (1.32
ρ̂(k, t)ρ̂(k,
variable= t) d=r d3 r δ (r −
corresponding to δ j (t))
ra(r −
density−
r ρ Fourier
(t)) exp
− ρ {ikexp · r}{ik
component. · r} .before,(2.19) see Eq. (2.19)
(1.32).
We now add j
an explicit jtime dependence:
We now
from which addweanobtain explicit j time
the dependence:
important property:
• microscopic expression !for density
" Fourier component: $
Eq. (2.14) states that the time
! autocorrelation # function $ of this of variable decays decays
"
Eq. (2.14) states that
ρ̂(k, t)the= time
"3 d !r # 3 autocorrelationδ (r − function
r−j (t)) − ρ {ikexp this variable
{ik · r} . (2.1
according to: ρ̂(k, t) = d r δ (r − r (t)) ρ exp · r}
! "
accordingȦ(t to: + s)B(s) = − A(t + s) j
Ḃ(s) (2.4)
j
j
∗ ∗ 2
% &
"ρ̂(k, t)ρ̂ (k, 0)# ∗ = "ρ̂(k, N 0)ρ̂ (k, 0)# ∗ exp −D(k)k % t . 2
& (2.20)
"ρ̂(k, t)ρ̂(2.14)
Eq. (k, 0)# = "ρ̂(k,
states
# that 0)ρ̂the(k,time0)# exp −D(k)k t function
autocorrelation . (2.20)
of this variable deca
For an autocorrelation = expfunction
{ik · r this
(t)} says
− (2π) that3
ρδ(k). (2.19)
To calculate D(k) according to:
we differentiate j
with respect to t and divide by −k 2 ; more- 2
To calculate D(k) wej=1 differentiate with respect to t and divide 2 by −k ; more-
over we shall dassume that k# is∗ small enough to set∗ exp {−D(k)k
# t} ≈ 122 for
• over
Onsager:we shall assume that ktimeis small enough to 0)#
set exp {−D(k)k t} ≈ 1decays
for
% &
Eq. (2.14)
all t of interest: "ρ̂(k,
states
"A(t)A(0)# that t)ρ̂
# the(k,
= 0)#
0, = "ρ̂(k,
autocorrelation 0)ρ̂ (k, exp
function −D(k)k
of this t .
variable (2.5) (2.2
allaccording dt to:
t of interest: #
t=0
1 calculate
To d D(k) we differentiate ∗with respect to t and divide by −k 2 ; mor
then − 2 "ρ̂(k, 1∗
d ∗
t)ρ̂ (k, 0)# ∗ = ∗"ρ̂(k, 0)ρ̂ (k, % 0)#∗ D(k)2 & (2.21)
i.e. !any k dt
time
"ρ̂(k,
over −we
t)ρ̂ 2(k,
shall"ρ̂(k,
autocorrelation
0)# = t)ρ̂
"ρ̂(k,
assume (k, 0)#(k,
function
0)ρ̂
that k =0)#
is "ρ̂(k,
starts
exp out
small 0)ρ̂
−D(k)k(k, 0)#
with
enough tD(k)
ato horizontal
. exp {−D(k)k
set (2.21)
slope.
(2.20)
2
t} ≈ 1 f
t 2 k dt
1 Fromall dtt of(2.4)
!Eq. 2 we∗ can also derive the useful
interest: ∗ relation
−To2 calculate1dτ 2 D(k) d τwe
"ρ̂(k, )ρ̂ differentiate
(k, 0)# ∗ = "ρ̂(k, with 0)ρ̂respect (k, 0)# to∗ D(k)
t and divide(2.22) by −k 2 ; more-
k −0 2 dt dτ 2 "ρ̂(k, τ )ρ̂ (k, 0)# = "ρ̂(k, 0)ρ̂ (k, 0)# D(k) 2 (2.22)
over we kd2 shall
0 assume
dt that
1 dk is! small ∗enough" to set exp {−D(k)k ∗ t} ≈ 1 for
Usingusing
Eq.
all (2.6) we
"A(t
t of interest: may + write−this
s)B(s)# 2= as
− "ρ̂(k,
k dtthis as Ȧ(t t)ρ̂
+ s) (k,
Ḃ(s)0)# . = we "ρ̂(k,write
can 0)ρ̂ (k,
this0)#
as D(k) (2.6) (2.2
Using Eq. dt2(2.6) we may write
1 d t ( d2 ∗
! t !
!1t ˙ '− − ˙∗2 "ρ̂(k, t)ρ̂ (k, 0)# ∗ = "ρ̂(k, 0)ρ̂∗ (k, 0)# D(k) ∗ (2.21)
dτ "ρ̂(k, τ ∗)ρ̂ (k, 0)# = "ρ̂(k, 0)ρ̂ (k, 0)# D(k) (2.2
'
dτ
We will ρ̂(k,
need1 τ )
thisρ̂
k 2(k,
in
dt 0) =
section "ρ̂(k,
2 2.3 0)ρ̂
when (k, 0)#
dealing
˙ k τ2 0)ρ̂˙ ∗ (k,dt0) = "ρ̂(k, 0)ρ̂∗ (k, 0)# D(k).
( D(k). with collective (2.23)
diffusion. For
0 k 2 dτ !2 t ρ̂(k, (2.23)
0
1k
an autocorrelation d function this ∗ says that ∗
−Using dτ
Eq. "ρ̂(k,
(2.6) we τ )ρ̂
may (k, 0)#
write =
this "ρ̂(k,
as 0)ρ̂ (k, 0)# D(k) (2.22) 31
It is now convenient k 0 to choose
2 dt # a specific
2 orientation for the wave vector, say
It is now 2convenient to choose a!specific " orientation for the wave vector, say
− 1 dτ 2 "ρ̂(k,d τ )ρ̂ (k,∗0)# = "ρ̂(k, 0)ρ̂ (k,∗0)# D(k) (2.22)
k2
− 02 dtdτ 2 "ρ̂(k, τ )ρ̂ (k, 0)# =Using "ρ̂(k, Eq.
0)ρ̂ (k, 0)# D(k)
(2.6) we may (2.22)this as
write
k 0 dt
Using Eq. (2.6) we may write this as
Using Eq. (2.6) we may write this as ! t
1 ˙' (
˙
! t ∗ ∗
1 dτ ρ̂(k, τ ) ρ̂ (k, 0) = "ρ̂(k, 0)ρ̂
Microscopic equation for collective D [2/2]
' (
! t ˙ ' τ )ρ̂˙ ∗ (k, 0) =("ρ̂(k, 0)ρ̂∗ (k, 0)# D(k). 2
dτ 2 1ρ̂(k, (2.23)
0 kdτ ˙ τ )ρ̂˙ (k, 0) = "ρ̂(k, 0)ρ̂ (k, 0)#
ρ̂(k, ∗ ∗ 0 D(k). k (2.23)
k 2
0
It is now convenient to choose a specific orientation It is now forconvenient
the wave vector,to say
choose a specific orie
the It is now convenient
z-axis: kspecific
= kêz . Then to choose a specific orientation for the wave vector, say
• the
choose
z-axis: k = kê orientation
z . Then for wavevector,
the say
z-axis: k = kê z:. Then
! t )# #
*
! t )# )* ∗
dτ #viz (τ )vjz (0) exp {ik (zi (τ ) − zj (0))}
! t =
# # (k,∗0)# D(k).
"ρ̂(k, 0)ρ̂
0 dτ i j viz (τ )vjz (0) exp {ik (zi (τ ) − zdτ j (0))} = "ρ̂(k,
v 0)ρ̂
(τ )v(k, 0)#
(0)D(k).
exp {ik (z (τ ) −
0 iz jz i
i j
0 i j (2.24)
(2.24)
• wethe
Using find the Green-Kubo
definition relation
of the structure for the
factor, Eq.collective
(1.33), we diffusion coefficient:
finally obtain:
Using the definition of the)structure factor, Eq. (1.33), we finally obtain: *
! ∞
D(k) =
1
dt
∞
1 # )#
v
Using
(t)v (0)
the
exp
definition
{ik (z (t) − z
of the
(0))} .
structure
* factor,
1 1
! # # iz jz i j
D(k) =S(k) 0 Ndt i j viz (t)vjz (0) exp {ik (zi (t) − zj (0))} ) .
S(k) 0 N ! ∞
i j
1 1 ##
(2.25)
compare this with the Green-Kubo relationD(k) for the= self-diffusion
S(k) 0
dt (2.25)
coefficient:
N
viz (t)v
This is the Green-Kubo
� � relation for the �
collective diffusion coefficient. i j
Thisself ∞
is the Green-Kubo
1 � relation for the collective diffusion coefficient.
D = viz (t)viz (0)
20 0 N i
20
This is the Green-Kubo relation for the coll
20
32
2.5 Shear 2.5 Shear viscosity
dynamics2.5 viscosity
Shear viscosity we will focus on transversal tr
In the second example,
through a fluid. Suppose the fluid velocity on a ma
hydrodynamics
on transversal transport 2.5.1
scribedof Macroscopic
momentum
by the fluid hydrodynamics
velocity field v (r). When two neig
2.5.1 Macroscopic 2.5.1 Macroscopichydrodynamics hydrodynamics
will
Application
velocity
focus
on a macroscopic
on
2:
transversalIn shear
elements
the
scale
secondviscosity
transport
is
moveexample,de-
with
of different
we
momentum will velocities,
focus on they will expe
transversal tran
mics
In When
). the second two example,
Inneighbouring
the second weexample,
will
fluid focusvolumewe on will
transversal
focus on transport
transversal of momentum
transport of
thethey fluidawill velocity proportional
through
on aa macroscopic a to
fluid. the area
Suppose of
scaleonisa de- the surface
fluid between
velocity on the a two
mac
through
es,
nsversal fluid.
transport Suppose
experience
through of a fluid.
momentum the
frictionfluid
Suppose velocity
force the fluid macroscopic
velocity on a scale is de-
macroscopic
y
• suppose the fluid has a macroscopic
Moreover,
scribed by even
the velocity
without
fluid (flow)
velocity field vmotion,
relative
field (r). When the volume
two neigh el
yscribed
field vby
between (r).
thethe When
fluid
two
scribed fluid
by
two
velocity neighbouring
volume
the field
fluid v (r).
elements.
velocity
fluid
When
field v
volume
two (r). neighbouring
When two fluid volume fl
neighbouring
y on a macroscopic scale elements
exchangeis de-move momentum with different
through velocities,
the motionstheyof, will
and exper
int
nt
on, velocities,
elements
the move
• neighbouring
volume they
with
elements
en two neighbouring fluid elementswill
fluiddifferent
move experience
elements
will velocities,
with exert
be ableforces
different
volume particles.
ato friction
they
on will force
each
velocities, experience
other they willa friction
experience forcea fr
proportional
constituent to the area of the surface between the two fl
proportional
the
otions surface
•
of, friction
and to
betweenthe
due
proportional area
to
interactions the of
velocity
tothe
two surface
fluid
differences
between,
the area between
volume
the
ofwithout
the the
elements.
surface two
betweenfluid volume
the two elements.
fluid volum
y will experience a friction
Moreover,
Allforce
the even
above forces relative
can motion,
conveniently the
be volume
summariz ele
Moreover,
ative motion,
• even the
momentum withoutvolume
exchange relative
elements
due motion,
to migration theof
will bevolume
able
particles elements will be able xto
to
en the two fluid Moreover, even
volumeexchange
elements.
Consider without momentum
a relative
surface motion,
through
element of the
thesize volume
motions
dA and elements
of, and
normal wil
inter
n̂.
exchange
ugh the momentum
motions of, through the motions of, and interactions between, the
ntly
e volume be •summarized
momentum
elements
exchange willinand
exchange
the
be
momentuminteractions
constituent
exerted
due
stress
able toby
totensor.
interactions
through
particles.
the
between,
fluid thebetween
below motionstheparticles
the of, andelement
surface interactions on thb
constituent particles.
of,
A and interactions
normal constituent
n̂. between,
Let dF bethe
particles.
All the the aboveforce forces can conveniently be summarized
• all these forces can conveniently be summarized
All the above forces can conveniently be summarized in the stress by
element. Then we define in
the the stress
stress tensor
tensor S̄ tensor.
ceconveniently
element on the be summarized
All fluid
the aboveforces
above
Consider the
a insurface
the
can stress
fluid conveniently
element tensor.of be summarized
size dA and in then̂.st
normal
Consider a surface element of size dA! and normal n̂. Let dF#be the_ force
summarized
tensor
of size
S̄ by dA
transformsinandthethe
Consider stress
normal
normal
a
exertedtensor.
surfacen̂.
dF of
by Let
a= the − dF
surface
element fluidbeS the
element
of size
below
n̂ dAforce
dA the= and normal
surface
−
"
S̄ · n̂ n̂.
element
dA,
dF = Let
S . on
^
n dAdF
the b
exerted by the fluid below the αsurface element αβ β on the fluid above α the fluid
normal
the to the
n̂.
surface force
Let
exerted dF
element exerted
be
by the
the
on
element. by
the theThen
force
fluid fluid
below
fluid below
above
we the the
definesurface
the the fluid element
stress ^ on the
n
tensor S̄ byfluid abo
element. surfaceThen we
element define
on the the
fluidstress
above tensor
the
β S̄
fluid by
element:
ment
#
n̂
he stresson
dA, the fluid
element.
tensor above
S̄ by Thenthe fluid
we define (2.26) the stress tensor S̄ by #
α !
by
"
where dFα =
and" −β run # S from
αβ n̂"β dA 1 =
to −
3 (orS̄ ·x,
n̂ y, dA,
and z).
!
dF
"α # = − S n̂
αβ β dA ! = −
α S̄ · n̂ α
dA, # dA α (2.26)
A = − S̄ · n̂ α dA, β dFα = − ManySαβ n̂β dA
fluids βcan= be− described· n̂ α dA,by assuming that the
S̄(2.26)
, y, and z).
A, (2.26)
of a part β which is independent of the flow velocity and
assuming
where α and thatβ the run stresswhere
from 1 to α3and
tensor β run
consists
(or x, y, from
and z). 1 to 3 (or x, y, and z).
linearly on the instantaneous derivatives ∂vα /∂xβ33.2 I
flowMany velocity where
fluids andcanaαbe and
part Many
β run fluids
which
described from
depends 1can
by assuming to be3 (ordescribed
x, y,
that theand byz).
stress assuming that the
tensor consists
Consider a surface element of size dA and normal n̂. Let
Consider a surface element of size dA and normal n̂. Let dF be the force
exerted by the fluid below the surface element on the fluid
exerted by the fluid below the surface element on the fluid above the fluid
element.
element. Then we define Thentensor
the stress we define
S̄ by the stress tensor S̄ by
Stress tensor
dFα = −
! for simple
dF=
Sαβ n̂β dA α −=−
" fluids
!
S̄ · n̂ αSdA,
# " #
αβ n̂β dA = − S̄ · n̂ α dA,(2.26)
β β
• simple fluids
where canβbe
α and rundescribed
from 1 α
where by
to andassuming
3 (orβx,run
y, and that
fromz). 1 stress tensor
to 3 (or x, y,consists
and z). of a
Many fluids can be described by assuming that the stress tensor consists
• part
of awhich is independent
part which
Many fluids
of the
is independent
can
flow
of the
be described
flowvelocity,
velocity and
by assuming that
and aa part which depends
the stres
of a part which
linearly on the instantaneous is independent
derivatives ∂vα /∂xβ .2 Inofhydrodynamics
the flow velocity
it is and a par
• part which
shown thatdepends
the most linearlyon
linearly
general onthe
stressthe instantaneous
tensor having thesederivatives
instantaneous /∂xβ .2 In hyd
∂vαreads
properties then
$ shown that % $the most & general
' stress
% tensor having these prop
∂vα ∂vβ 2
Sαβ = η + − P$ + η − κ ∇%· v $δαβ . & (2.27)
' %
∂xβ ∂xα ∂v3α ∂vβ 2
Sαβ = η + − P+ η − κ ∇ · v δαβ
Here δαβ is the Kronecker delta (1 if∂x αβ = β,∂x0αotherwise), η is3 the shear
viscosity, κshear bulk
is the bulk viscosity, and P the pressure.
Here δ pressure
is the
viscosity
Combining αβ
Newton’s equations of Kronecker
viscosity
motion delta (1
(expressing theiflaw
α of=conser-
β, 0 otherwise
vation of momentum) viscosity,
with theκlawis the bulk viscosity,
of conservation anditPis the
of mass, pressure.
possible to
derive the macroscopic Combining
Navier-StokesNewton’s
equation, equations of motion (expressing th
• N.B. for complex
vation of momentum) with the law of conservation
fluids such as polymers, the stress tensor may actually
of mass
D
ρm v = ∇ ·derive
S̄, the macroscopic Navier-Stokes (2.28)
equation,
depend on the
Dt history of fluid flow
D
where ρm = mρ is the mass density and D/Dt = v · ∇ + ∂/∂t is the total
ρm
derivative. The combination Dt
v = ∇ · S̄,
of Eq. (2.27) with Eq. (2.28), sometimes also
34
referred to as the Navier-Stokes equation, is rather formidable. Fortunately,
exchange momentum through the motions of, and interactions between, the
constituent particles.
All the above forces can conveniently be summarized in the stress tensor.
Consider a surface element of size dA and normal n̂. Let dF be the force
Navier-Stokes and Stokes equations
exerted by the fluid below the surface element on the fluid above the fluid
element. Then we define the stress tensor S̄ by
! " #
dFof
• law α = − Sαβ n̂β dA
conservation of = − S̄ · n̂ α dA,
momentum (2.26)
+ law of conservation of mass:
β
� �
where α and β run from ∂ 1 to 3 (or x, y, and z).
2. TIME mρ
Many v ·can
fluids
DEPENDENT ∇+ v = by
be described
PROPERTIES∇ assuming
· S̄
OF LIQUIDSNavier-Stokes
that equation
the stress tensor consists
∂t
of a part which is independent of the flow velocity and a part which depends
linearly on the instantaneous derivatives ∂vα /∂xβ .2 In hydrodynamics it is
• we that
many
shown focus
flow theon incompressible
fields
most of interest
general flow of
stress may
tensor besimple fluids
described
having these at small
assuming
properties velocities:
then that
reads the fluid is
incompressible. $ In that %
∂vα ∂vβ case$∇·v&= 0. Assuming
2
' % moreover that the velocities
Sαβ = and
are small, η that+ second− order
P + nonlinear
η − κ ∇ terms
· v δαβin. v may (2.27)
be neglected, we
∂xβ ∂xα 3
obtain Stokes equations for incompressible flow:
Here δαβ is the Kronecker delta (1 if α = β, 0 otherwise), η is the shear
viscosity, κ∂v
is the bulk viscosity, and P the pressure.
mρ
Combining = η∇2equations
Newton’s v − ∇P of motion (expressing the law of conser- (2.29)
∂t Stokes equations
vation of momentum) with the law of conservation of mass, it is possible to
· v = 0. Navier-Stokes equation,
∇ macroscopic
derive the (2.30)
D
Theseρmare vthe
=∇ hydrodynamic
· S̄, equations that we will use from here on.
(2.28)
Dt
Our aim is to derive a microscopic equation for the shear viscosity η. To 35
where ρ = mρ is the mass density and D/Dt = v · ∇ + ∂/∂t
3 is the total
equationsmρ that= weη∇ will
2 use from here Our
v − ∇P (2π)
on.3 aim is to derive a microscopic These
decays are
(2.29) equation
the
macroscopichydro f
∇ ∂t · v = 0. ∇ · v = 30. (2.30) (2.
ee how a wavelike velocity
icroscopic equation for the shear viscosity field this end,η. let Tous first see how a wavelike (2.30)
Our aim shows velocity
is that
to der t
aThese ∇ · vvelocity
wavelike = 0. field 3
decays macroscopically. Introducing this in the incompressibility one: (2.30) equat
are the hydrodynamic
These are the equations
hydrodynamic that we will
equations use from
that we here
will this
on.
use end,
from let
here us first
on.
(2.30) shows that the only possible 1wavelike velocity field is a transver
g(k,
These Decay
t) are
Our expaim theis of
{−ik aOurmacroscopic
to·derive
r}
hydrodynamic a microscopic
aim
one: to deriveequation
is equations athatwavelike
ρv(r, wefor
microscopic t)will (2.31)
=the velocity
shear
equation
use
3 (2π)
from
3
g(k, t) exp
viscosity
for
here the
on.field
{−ik
η. To [1/2]
shear · r}
viscosity
g(k, t) · k = η.0
1
this
exp {−ik end,
Our aim let us
· r}is to first
this see how a
end, aletmicroscopic
derive wavelike
us first see equation velocity
how(2.31) field
a wavelikefor thevelocity field
shear viscosity
3
η.ρv(r,
To t) = (2π)
this end, let us first see how g(k,
a t) · k = 0.velocity field3
wavelike (2.3
ly. Introducing this
1 in the incompressibility
decays macroscopically. equation Taking
Introducing this in the the divergen
• consider a wavelike velocity 1field
only ρv(r,
eroducing possiblet) =
this inwavelike
the
(2π) 3
g(k, t)
ρv(r,
Taking t) exp
=
velocity
incompressibility
the
{−ikg(k,
(2π)
· r} t) exp {−ik · r}
field
(2.30)
divergence
3 is
equationa
shows
of transversal
Eq. that
(2.29) the
and only
using
It(2.31)
decays
possible
Eq. (2.30)
is now
we
convenien
macroscopica
wavelike
get ∇ 2 (2.
P ve
=
1 velocity field. Supp
possible ρv(r, t) = velocity
wavelike 3
It is t)
g(k, expconvenient
field
now {−ik
is one: · r}to choose specific orientations for(2.31)
a transversal (2.30) the wave vector th
shows that a
decays macroscopically. (2π) velocity Introducingfield. Suppose
this in that
the the wave
incompressibilityvector is k = kêalong
oriented
equation
one: z , and that
the z-ax t
decays macroscopically. Introducing this in the incompressibility at the x and equat
z com
(2.30) shows that the k = kêpossible
only z , and that the flowvelocity
wavelike is along the field y-axis,
is a g(k, t) = gy (k, t)êy . Look
transversal
decays (2.30) shows
• formacroscopically.
an incompressible thatthe
fluid
Introducing theonlyonly
this g(k,
possible
in t) · incompressibility
possible
the = 0.
k(2.32)
wavelike velocityequation field is awith
Together transver
∇2 P=
one: at the x and z components of Eq. (2.29) this yields ∂P/∂x = ∂P/∂z
g(k, t) · k = 0.
wavelike one:
velocity
(2.30) shows that the Together field is
only possiblea transversal
with ∇ wavelike
2 (2.32)
one:
P = 0 thisvelocity also means field
thatis∂P/∂y component
a transversal
= 0. The onlyinremain Eq. (
Taking the 2
e one:
of Eq. (2.29) and using Eq. (2.30)
component in Eq. we
(2.29) get then ∇divergence
P = 0. of Eq. (2.29) and using E
reads
g(k, t) · k = 0. g(k, t) · k = 0. 2 (2.32)the∂vdivergen
Taking (2.
. (2.29)
to chooseand usingorientations
specific Eq. (2.30) we forget theis∇wave
It P =convenient
now 0.
vector and to choose specific orientatio
y
mρconvenient = η
seTaking
specific orientations for the ∂vwave ∂
vector
velocity
2
v y and
field. Suppose that It is now ∂t
2the(2.32)wave vector is
y
that g(k,
• e.g. the t) wave
·
choose k
the divergence= k vector
0.along of z is
Eq.and
mρ oriented
flow
(2.29) =along
and ηalong y
using the
Eq. z-axis,
(2.30) we get ∇ P = 0. 2
Taking the divergence ∂t of Eq.
∂z 2
(2.29) and using Eq. velocity
(2.30) we field.
∂gy ∇Supp
get P =
te Itthe
flow wave
is along
is now vector is
the y-axis,
convenient oriented
to g(k,∂g
choose along
t)y =k g=
specific the(k, z-axis,
kêt)ê
z , and
yorientations that
y . Looking
for thewave
the flow vector
is along and the
mρ y-axis, = g −
It isg(k,
nowof convenient
mρ =toandchoose
−ηk 2
g specific orientations k = kê
for2 Pthe=zwave , and that
∂tvector a th
sonents
along
Taking
velocity the
the
Stokes y-axis,
divergence
Eq.equation
offield. (2.29)
Suppose t)
then
this that=
Eq. g
gives
yields (k,
(2.29) t)ê
y ∂t ∂P/∂x
the at
wave .
the Looking
y vectorusing
x
= ∂P/∂z
y and
is Eq. z (2.30)
components
=
oriented 0. we
along getthe ∇
of Eq.
z-axis, 0.
(2.29) this y
velocity field. Suppose that the wave
! 2 vector " is oriented at the along x andthe z comp
z-a
ofIt is
0k this
= now
Eq.kê (2.29)
and
also convenient
taking
z , andmeansthis
the
that that
the to
yields choose
divergence∂P/∂x
∂P/∂y
flow is =specific
of =
along 0. ∂P/∂z
Stokes
Together
The
the orientations
only
y-axis, =with
equation 0. ∇
remaining
g(k, t) forP g=
η= the
2y (k,
wave
0 this
t)ê vector
also
. Looking and
means gy (k,that
t) = ∂P/
2 g
k = kê z , andgythat (k, t) the = flowgy (k,is0)alongexp − the y-axis,
k t . g(k, t) y Together
= g y (k,with
t)ê y .∇ P(2.3
Look =
29)velocity
also
at then
thesox we
means field.
and
reads get
that Suppose
∂P/∂y
z at
components =that0. too.
The
of the
Eq. wave
only
component
(2.29) vector
remaining
this is
in
yields oriented
Eq. mρ (2.29)
∂P/∂x along
= then
∂P/∂zthe reads z-axis,
= 0. = ∂P/∂z
the x and z components of Eq. (2.29) this yields component
∂P/∂x in Eq. (2 =
= kêz , with
nkTogether
reads and that 2
P the
= 0flow
∇Together this is
also along
2means thethat y-axis,∂P/∂y g(k, = t)0.=The gy (k,
only t)êremaining
y . This
Looking shows that am
with
This shows ∇ P = 0
that(2.29) this also
a macroscopic means that
transversal ∂P/∂y
2 velocity
= 0.
field =The
decaysonly remain
to zero w
2 at the
component x and in z components
Eq. (2.29) then of Eq.
reads this ∂v yields ∂P/∂x∂ vy = ∂P/∂z 0.
vy component a in
relaxation Eq. (2.29)
time τ then
(k) = y
reads
mρ/(ηk 2
). a relaxation ∂v time τ (∂
Together with ∇ 2
P = 0 this also means thatmρ∂P/∂y== 0.η The 2 only remaining mρ
y
= η
z 2 ∂t ∂z ∂t (2π) 36 −3
component ∂vyin Eq. (2.29) ∂ 2 vy 3then reads 3
The factor
∂vThe factor (2π) ∂ 2 v appears∂g fory the same reason as in Eq. (2.15), so that g(k, t) m
−3
ly.
It isIntroducing
now convenientthistoinchoose
the incompressibility
specific orientationsequation
for the wave vector and
eTaking
only possible
velocity the
field. wavelike
divergence
Suppose velocity
ofthat
Eq. (2.29)
the field
wave is a istransversal
andvector
using Eq. (2.30) along
oriented ∇2z-axis,
we getthe P = 0.
It
k= is kê
now convenient
z , and that thetoflow choose specific
is along the orientations
y-axis, g(k, t)for= the wave
gy (k, t)êy .vector
Lookingand
at theDecay
velocity and of
x field. a
z Supposemacroscopic
components thatoftheEq.wave(2.29)wavelike
vector velocity
is oriented
this yields ∂P/∂x along
= ∂P/∂zfield =[2/2]
the z-axis,0.
kTogether
= kêz , with ∇2 P the
and that = 0flow this is
also means
along thethat ∂P/∂y
y-axis, g(k,=t)(2.32)
0.=The only
gy (k, t)êremaining
y . Looking
component
at the x and in zEq. (2.29) thenofreads
components Eq. (2.29) this yields ∂P/∂x = ∂P/∂z = 0.
• Stokes’
Together with equation
∇ 2 is
P =using0 reduced also to
this Eq. means that ∂P/∂y 2 = 0. The only remaining
e of Eq. (2.29) and 2
(2.30) we get ∇ P = 0.
component
to choose ∂v in Eq.
specific
y ∂
(2.29) vy then reads
orientations for the wave vector and
mρ = η 2
ose that the ∂t wave vector ∂z is oriented along the z-axis,
∂g 2
∂vy ∂ v2y
e flowwhich is along
mρ may=the
be−ηk
ηrewritten
gy in
y-axis, terms
g(k, t) of gy (k, t)êy . Looking
=
∂t ∂z 2
onents of∂gEq. (2.29) this yields ! ∂P/∂x
η 2
" = ∂P/∂z = 0.
y 2
= 0 thisgmρ (k, t)means
yalso = g−ηk y (k, 0)
that y exp −
g∂P/∂y =mρ 0.k The
t . only remaining (2.33)
∂t
29) then reads !
η 2
"
gy (k, t)that= a gmacroscopic
This shows y (k, 0) exp transversal
− k t .velocity field decays to zero (2.33) with
2 mρ
avyrelaxation time τ (k) = mρ/(ηk 2 ).
2
∂zThis shows that −3 a macroscopic transversal velocity field decays to zero with
3 • a macroscopic wavelike
The factor (2π) appears forvelocity
the 2same field decays
reason as inexponentially
Eq. (2.15), so (at
thatlarge
g(k, times)
t) may
abe2relaxation time
withasa arelaxationτ (k)time= mρ/(ηk
determined ). by thefield.
shear viscosity and the density:
ηk gy viewed Fourier component of the velocity
3
The factor
! (2π)mρ−3 appears
" for the same reason as in Eq. (2.15), so that g(k, t) may
τ (k) = η
22 0)
(k,
be exp as −
viewed ηk 2k 2 tcomponent
a Fourier . of the velocity field. (2.33)
mρ
37
22
2.5.2 Microscopic equation for η
We will now derive a microscopic
2.5.2 (and for
equilibrium)
η equation for the shear
We will now derive Microscopic equation
a microscopic (and equilibrium) equation for the shear
viscosity. The microscopic variable corresponding
2. TIMEcorresponding
DEPENDENT PROPERTIES to LIQUIDS
OF g(k, t) is
viscosity. The microscopic variable to g(k, t) is
We will!now derive ! a microscopic (and equilibrium) equation for the shear
Microscopic equation for shear viscosity [1/3]
"
3
viscosity.
g(k, t) = g(k, t)
The
dr3 =
"microscopic
v d
(t)δ r (r −variable
vrj (t)δ
(t)) (r
exp −{ik rj (t))
corresponding
· r} exp {ik to ·g(k,
r} t) is
2.5.2 Microscopic equation !j for
j η
j
j " 3 "
g(k,"t) = = dvr(t)equilibrium)
vj (t)δ
exp {ik (r
· r −(t)} rj (t))
. expfor{ikthe · r}shear (2.34)
We will now derive= a microscopic (and
vj (t) exp {ikj · rjj (t)} . j equation (2.34)
• microscopic
viscosity. The microscopic expression
j
variable " jfor velocity field
corresponding to g(k, Fouriert) is component
Choosing
!
k along= z and vjg(t) exp {ik
along y, · rj (t)} . regression hypothesis, Eq. (2.14),
Onsager’s (2.34)
Choosing =alongd3zrand gvjalong (t)δj(ry,−Onsager’s
rj (t)) expregression
{ik · r} hypothesis, Eq. (2.14),
"
g(k, t) k
states thatstates the time thatautocorrelation
the time autocorrelation
j function offunction this variable of this variable
decays decays according
according
to Choosing
to " k along z and g along y, Onsager’s regression hypothesis, Eq. (2.14),
=
states#that vj (t)the exptime
{ik ·autocorrelation
rj (t)} . function of &
this
% variable (2.34)& decays according
∗#
$ # ∗
%
η$ η 2
#
gy (k,
j
to t)g ∗ (k, gy (k,
0) t)g
$
= y (k,
gy (k,0) 0)g =∗(k, gy (k,
0) 0)g
$
expy (k,−0) exp k 2 t .− mρ k t (2.35) . (2.35)
y y
Choosing k along z and g along y, Onsager’s regression mρ
hypothesis, % Eq. (2.14), &
η 2
•Following
Onsager: ∗ ∗
. samewe procedure...
# $ # $
states that theFollowing
timesame
the gy (k,the t)gsame
autocorrelation
analysis y (k,as 0) in =
analysis
function
going gyas(k,
of
from 0)ggoing
in
this (k,(2.20)
yvariable
Eq. 0)
from exp −(2.20)
Eq.(2.23),
decays
to k we
according tto obtain
(2.23), (2.35)
obtain
mρ
to ! t % &
! t
1 Following 1 the#
same analysis
∗
$ # % $ η ∗ & $ ηη
as in going from
%
Eq. (2.20)
&
to η 2 we obtain
(2.23),
dτ$t)ġ2 (k,
∗ #ġy (k, t)=ġy∗(k, 0) $0)g = (k,
∗ gy (k, 0)g2y (k, 0) − exp 2 −(2.36) k t . (2.36)
# $ #
# dτ ∗ ġy (k,
k 0) g y (k, 0) η exp mρ k t .
g0y (k, kt)g
2 (k, y
y !0 t 0) = gy (k, 0)gy (k, 0) exp −
y
mρ k t . mρ (2.35) % mρ &
1 # ∗
$ # mρ ∗ $ η
� η �2
From statistical Fromdτstatistical
mechanics
� 2
ġy (k, t) itġy (k,
mechanics
is �
0) =
known � it gy (k,
is
that known 0)gy#(k,
"v y,j �
that
= 0)0"vand
�y,j #exp
� =
"v 0
v−and# k"vt y,j.vy,k
= (2.36)
# =
Following the same k
0 analysis as in ∗ going from Eq. (2.20) to (2.23),mρ we y,j y,k
obtain mρ N k BT
kB T /mδsmall (kB Tk:/m)δ
jk #(equipartition
gyjk 0)g y (k, 0) ≈Intheorem).
(k,(equipartition
theorem). gthe
y (0, limit
∗
0)gy#(0, of 0)the
Insmall = klimit
this of
can vy,j
small
be vused
y,k = can be
k this
∗"v j&# k= 0 and "v v m# =
# # ∗ $ $
! t
1 # used From statistical
to0)g replace $ 0)#gyby mechanics
(k, 0)g it is known
(k, 0)$ by gy (0, 0)gTy/m. that (0,y,j0)Then = NkB T /m.y,jThen
∗
$ ∗
$ %
to replace gy (k, ∗ y (k, gy (0,∗y 0)gy (0,η0) = NkBη 2
y,k
dτ 2 ġy (k, k B Tt)/mδ
ġy (k,jk0) = gy (k,! 0)gytheorem).
(equipartition (k, 0) expthe−limitkoft small
In . (2.36) k this can be used
0 k 1replace g1y (k, 0)g
!
# ∞ ∞
1 #∗(k, 0) 1by# g∗y (0, 0)g
$ # mρ mρ $
$ ∗(0,∗ 0) =$ Nk T /m. Then
• so we η =find: η lim
to = limy mġy (k,
dtk→0 dt t)m 2
m
ġ ġ(k,
y y (k, 0) t)m.y ġy (k, 0) . B (2.37) (2.37)
B T k→0V 0kBit T isk known k "vy,j # = 0 and "vy,j vy,k # =
From statistical V kmechanics 2
!0 ∞ that
kB T /mδjk #(equipartition 1
theorem). In the 1limit# of small k ∗this can $ be used
All that remains All that η∗ =isremains
to $ write #lim
isoutto write dt∗ terms
the 2
m
out $ ġthe
ym(k,
ġ y
t)mt).
terms
(k, ġy (k,
m ġ0)
y (k, . t). Remembering
Remembering that (2.37)
that
V k T
to replace gy (k, 0)gy (k, 0) Bby gy (0,0 0)gy (0, 0) = NkB T /m. Then
k→0 k
ṙj = vj weṙjfind = vj we find 38
1 All that ! ∞remains is to write out the terms mġ (k, t). Remembering that
1 # $ y
used to replacetheorem).
(kB T /m)δjk (equipartition gy (k, 0)gy (k,In 0)
the by
limitgy (0,
of 0)g
smally (0,k0)this = Nkcan B Tbe
/m. Then
$ 1 # ∞
1
# ! $
∗ ∗
used to replace gy (k, 0)gy1(k, 0) by
! ∞ gy (0, 0)g (0, 0) = NkB T /m. ∗ Then
# $
η= lim1 # ydt 2 mġ∗y (k, t)m $ ġy (k, 0) . (2.3
η != ∞ limV kB T dt k→0 0mġy (k, k t)mġy (k, 0) . (2.37)
V kB T 2
k ∗
1 1 k→0
# 0 $
η=
V kBAll
lim All dt
T that 0remains
k→0
that2
mġy (k, t)m
remains is ġy (k,
to 0) .out the terms mġ(2.37)
write
k is to write out the terms mġy (k, t). Remembering that
ṙj = vj we find
y (k, t). Remembering th
All that remains ṙj = isvj towewrite

find out the terms mġ (k, t). Remembering that
y
N
ṙj = vj we find N "
• write out themterms mġy"
ġy (k, t) =
(k, t) = {mv̇y,j (t) + ikmvz,j (t)vy,j (t)} exp {ikzj (t)} ,
{mv̇y,j (t) + ikmv (t)v (t)} exp {ikz (t)} , (2.38)
(2.3
z,j y,j j
N j=1
" j=1
mġy (k, t) = {mv̇y,j (t) + ikmvz,j (t)vy,j (t)} exp {ikzj (t)} , (2.38)
part j=1 part ofcan
of which which can beinwritten
be written terms of in forces:
terms of forces:
N N N N "
part of which can be " written inmterms
"
v̇ expof forces:
{ikz }
" " (i)
=
"
F
(i)
exp {ikzj }
mv̇y,j exp {ikz y,jj } = j Fy,j exp {ikzy,j j}
N j=1 " N " j=1 i$=j
j=1
(i) j=1 i$=j
"
mv̇y,j exp {ikzj } = Fy,j exp N {ikzNj } " N " N
1 " " 1 (i) (i)
j=1 j=1 i$=j = = Fy,j [exp {ikz Fy,jj }[exp
− exp{ikz j } i−
{ikz }] exp {ikzi }]
N " N 2 j=1 i=1 2 j=1 i=1
1 " (i)
= Fy,j [exp N{ikz N j } −NexpN{ikzi }]
2 j=1 i=1 1 " " 1 (i) " " (i)
= ik = Fy,j (zj − zFi ).
ik y,j − zi ). k) (2.39)
(zj (small (2.3
N " N 2 j=1 i=1
2 j=1 i=1
1 " (i)
= ik Fy,j (zj − zi ). (2.39)
2 j=1 i=1 23
23
39
of aFigure
polymer 2.2:meltShear
(left: re-
(i) linear scale, modulus
laxation right: log-G(t) i on particle
Here Fy,j is the force in the y-direction exerted by particle
arithmic scale). melt (left:
of a polymer
the second step of Eq. (2.39) we have used Newton’s
linear scale, right: log-
principle of action
Here Fy,jreaction,
(i)
is the force Fy,j
(i)
in the = y-direction
(j)
−Fy,i . Inexerted the third
by particle step weparticle
arithmic
i on have
scale).j.usedIn the fact that k
be very
the second step ofsmall.Eq. (2.39) Collecting
we have used everything together,
Newton’s principle we may
of action and write
(i)
(i) (j)
reaction,
Here Fy,jF = −F y,i . In
y,j is the force in !the∞y-direction
third step we have by
exerted used the fact
particle i onthat k will
particle j. In
• collecting
be very small. step
everything,
Collecting
we get the
V everything Green-Kubo
together,
relation for the shear viscosity
the second of Eq. (2.39) we have used we may write
Newton’s principle of action and
η
(i) = (j) dt "σ yz (t)σ yz (0)# , (
reaction,VFy,j = ! ∞ −FkBy,iT . In 0 the third step we have used the fact that k will
be ηvery
= small. Collectingdt "σyz (t)σ yz (0)# ,
everything together, we may write (2.40)
kB T 0
where the
V
!microscopic
∞ stress tensor is defined as
where the ηmicroscopic
= stress
dt" "σyztensor
(t)σyz is defined
(0)# , as (2.40) $
kB T 0 1 N # N $
" # N # N 1 # (i)
where σ 1 is the
(t)# =
yzmicroscopic microscopic mv 1
stress
y,j
#(t)v tensor:
z,j (t)
(i) + F (t)(z j (t) − zi (t)) . (
σwhere
(t) =the mv (t)v
stress (t) +
tensor is F
defined (t)(z
as (t) − z (t)) y,j. (2.41)
yz
V V y,j z,j
j 2 y,j 2j
i=1 j=1
i
j" i=1 j=1
N N
$
1 # 1 # # (i)
σEq.
Eq. (2.40)yz (t) =the Green-Kubo
is(2.40) is mv
the y,j (t)v z,j (t) + for therelation
relation
Green-Kubo Fy,j (t)(z
shear j (t) −
viscosity.
for thezThe
i (t)) . (2.41)
above
shear viscosity. The a
V 2
equations also yield a jmicroscopic expressioni=1forj=1 the so-called shear relaxation
modulus,
equations also yield a microscopic expression for the so-called shear relax
Eq. (2.40) is the Green-Kubo relation for the shear viscosity. The above
modulus,
equations also
V yield momentum
a microscopic expression for momentum
the so-called shear relaxation
G(t) = "σyz (t)σyz (0)# . (2.42)
modulus, kB T transport V through transport through
G(t) V =
particle "σyz (t)σyz (0)#
motion .
particle interactions (
Macroscopically,
G(t) = G(t) "σ iskthe Tlinear
(t)σ stress relaxation in the system following
yzB yz (0)# . (2.42)
a sudden step %strain. kB T In agreement with Eq. (2.40) its integral is the shear
Macroscopically,
viscosity, η =
Macroscopically,
∞
G(t)dt.
G(t) is
G(t)
the linear
is the
stress
linear
relaxation
stress in
relaxation
the system
in the system follo
following
0
Thesea sudden
a sudden equations step
step %strain. strain.
are useful because
In agreement In agreement
they
with enable
Eq. (2.40) uswith Eq. (2.40)
toitsmeasure
integral the its
shear
is the integral is the
shear
∞
% ∞ viscosity in a simulation of a liquid without 40
relaxation modulus
viscosity,
viscosity, and
η = 0η G(t)dt.
= shear G(t)dt.
be very small. Collecting everything together, we may write
! ∞ where the microscopic stress tensor is defined as
V
η= dt "σyz (t)σyz (0)# , (2.40)
" N # N
kB T 0 1 # 1 # (i)
σyz (t) = mvy,j (t)vz,j (t) + Fy,j (t)(zj (t) − z
Example: shear relaxation modulus of polymer melt
where the microscopic stress tensor is defined as
V j
2 i=1 j=1
" Eq. (2.40)
N N is the Green-Kubo relation $ for the shear viscosit
1 # 1
equations
# #
also yield
(i) a microscopic expression for the so-called sh
σ (t) =
• shear relaxation
yz mv (t)v
modulus G(t)
y,j z,j (t) + is F (t)(z j (t) − zi (t)) . (2.41)
the (macroscopic, linear) stress relaxation
y,j
V j
2
modulus,
i=1 j=1
following a sudden step strain Vshear viscosity. The above
Eq. (2.40) is the Green-Kubo relationG(t) for the
= "σyz (t)σyz (0)# .
• we can
equations alsodetermine G(t) and expression
yield a microscopic forkthe
shear viscosity B Tfrom so-called shear relaxation
equilibrium simulations without
modulus, actually shearing!
Macroscopically, G(t) is the linear stress relaxation in the sys
V a sudden step %strain. In agreement with Eq. (2.40) its integr
∞
G(t) = "σyz (t)σyz (0)# . viscosity, η = 0 G(t)dt. (2.42)
kB T These equations are useful because they enable us to mea
Macroscopically, G(t) is the linearrelaxation stress relaxation
modulus in and theshear
system following
viscosity in a simulation of a
a sudden step %strain. In agreement with Eq.
actually (2.40)theitssystem,
shearing integralbutis rather
the shear by analyzing the spo
∞
viscosity, η = 0 G(t)dt. tuations in forces and velocities. As an example, in Fig. 2.2
These equations are useful because they enable
shear relaxation us to measure
modulus measuredthe in ashear
molecular dynamics s
relaxation modulus and shear viscosity melt ofinpolyethylene
a simulation chains
of a at 450 K.
liquid Note that in this parti
without
actually shearing the system, but the stress
rather bydoes not relax
analyzing theimmediately
spontaneous to fluc-
zero, but remains at a
tuations in forces and velocities. ofAsapproximately
an example, 3inMPa Fig.for2.2times
we melt of C400
between
show the5 and 50 ns. Such
polyethylene chains
indicative of temporary
shear relaxation modulus measured in a molecular dynamics simulation elasticity, which of
with a shear
is viscosity
typical
a of an entang
of
melt of polyethylene chains at 450liquid. K. Note that in this particular 0.2 Pa s (200 x water)
example
the stress does not relax immediately to zero, but remains at a plateau value
of approximately 3 MPa for times24between 5 and 50 ns. Such behaviour is 41
Onsager’s
over we shall regression
assume hypothesis
that k is small saysenough
that the decay
to set exp of this fluctuation
{−D(k)k 2
t} ≈ 1 for at
large
all t times and on macroscopic scales will be governed by the corresponding
of interest:
macroscopic laws. 1 d ∗ ∗
Notice that the − "ρ̂(k,
macroscopic t)ρ̂ (k, 0)#
laws = "ρ̂(k,
usually 0)ρ̂
apply (k,
to 0)# D(k)
non-equilibrium (2.21)
situ-
Summary of chapter 2
ations. In 1 formula
! t
2. TIME DEPENDENT
k 2 dt
d2Onsager’s
PROPERTIES
hypothesis
OF LIQUIDS
says
− 2 dτ 2 "ρ̂(k, τ )ρ̂∗ (k, 0)# = "ρ̂(k, 0)ρ̂∗ (k, 0)# D(k) (2.22)
k A(0)
!A(t)" 0 =dtA(0)α(t), (2.13)
• most microscopic Using
where Eq. (2.6)
quantities
α(t) we
is determined may write
fluctuate this as
byinmacroscopic
time (due tolaws. thermal fluctuations);
The average is a conditional
time correlation functions 1reveal their dynamics
! t
average: it expresses
' the average( time development of A, given that it was
˙ τ )ρ̂˙ ∗ (k, 0) = "ρ̂(k, 0)ρ̂∗ (k, 0)# D(k).
dτ 2 ρ̂(k, (2.23)
A(0) at 0t = 0. k Multiplying by A(0) and averaging over all initial conditions
• Onsager: the average
weIt get
decay of fluctuations obey macroscopic laws
is now convenient to choose a specific orientation for the wave vector, say
the z-axis: k = kê z . !Then2
" Figure 2.2: Shear re-
!A(t)A(0)" = A α(t). (2.14)
! t )# # * laxation modulus G(t)
The dτ averages areviznow (τ )v simple
(0) exp equilibrium
{ik (z (τ ) − averages.
z (0))} =ofWea polymer
shall
"ρ̂(k, 0)ρ̂ (k, melt
∗illustrate
0)# (left:
D(k). the
• using Onsager we derived Green-Kubo relations for transport linear
jz i j properties
scale, right: log-
use of Eq. i(2.14)
0 j with � two examples in � the next two sections.
� ∞
1 � arithmic scale). (2.24)
self-diffusion 18 D self
= viz (t)viz (0)
0 N
Using the(i) definition of thei structure factor, Eq. (1.33), we finally obtain:
Here Fy,j is the force in the y-direction exerted by particle i on particle j. In
the second step 1of Eq. (2.39) 1 we #have # used Newton’s principle of action and
) *
! ∞
collective diffusion
reaction,D(k) = = −F (j) . dt
(i)
Fy,j In the third viz (t)v
step we jz (0)
haveexpused
{ik (zi (t)fact
the − zjthat
(0))}k will
.
S(k) y,i 0 N i j
be very small. Collecting everything together, we may write
! ∞ (2.25)
V
shear viscosity This isη the = Green-Kubo dt "σrelation
yz (t)σyz (0)# , collective diffusion coefficient. (2.40)
for the
kB T 0
where
20 the microscopic stress tensor is defined as
" $ 42
# #N # N
Chapter 3: Brownian motion
43
Colloidal liquids
• colloidal liquids contain particles much larger than the solvent molecules
(typically 1 nanometer to 1 micrometer)
• Brownian motion is caused by random collisions with solvent molecules
44
stochastic force F(t). These forces are summarized in Fig.
of motion then read1 3.1 Friction and random f
Friction and random forces Consider

(3.1)
on a colloidal particle
a spherical colloidal particle of radius
ter and a micrometer) and mass M moving t
V + F. (3.2)
R(t). The colloidal particle will continuously c
• suppose
e divided a colloidal
all forces by the particle
mass M of the (radius a,Because
colloid. Consequently, mass M)onmoves
F(t) average the acolloid
along will collide mo
path R(t)
d the friction constant ξ is a frequency. on the back side, it will experience a system
velocity V, and directed opposite to its veloci
loid • colloid will collide on average more often ence on the front than on the back:
a random or stochastic force F(t). Thes
city systematic force proportional and opposite to its velocity = friction 1
e a
3.1 The equations of motion then read
op-
city
F dR
= V 3. BROWNIAN MOT
s F -!v v dt
uous 3. BROWNIAN
dV MOTION
sol- = −ξV + F.
dt
By solving Stokes
1 By solving Stokes’ equations (2.29), wit
Note that we have dividedofallthe sphere,
forces it c
by the mass
• we can find an expression for the friction by solving Stokes’ equations
25of
is an the sphere,
acceleration anditthe
can be shown
friction that
constant ξ is the
a freqf
for a sphere moving through a hydrodynamic solvent:
ξ = ζ/M =
ξ = ζ/M = 6πηs a/M,
Figure 3.1: A colloid
moving with velocity
solvent where
viscosity ηs of
is the
the sol
sh
where ηs is the shear viscosity
V will experience a Solving Eq.
Solving Eq. (3.2) yields 45 (3
friction force −ξV op-
s
of the sphere, it can be shown that the friction constant ξ is given
where η is the shear viscosity of the solve
s by
Solving
where ηs isEq.
the (3.2)
shear yields
viscosity of the solvent. Solving Eq. (3.2) yields
ξ = ζ/M
Solving Eq. = 6πηyields
(3.2) s a/M,
! t (3.3) ! t
−ξt −ξ(t−τ )
V(t) = V0 e + ! t dτ e F(t). (3.4)
Fluctuation-dissipation theorem [1/2]
−ξt −ξ(t−τ )
where ηs is the shear viscosity
0 of the solvent. V(t) = V 0 e + dτ e F(t).
V(t) = V0 e−ξt + dτ e−ξ(t−τ ) F(t). (3.4) 0
Solving Eq. (3.2) yields
where V0 is the initial 0velocity. We will now determine averages over all
possible realizations of ! t with the initial velocity aswhere
F(t), a V0 is To
condition. thethisinitial velocity. We will
where V0 is the initial velocity. We will now determine averages over all
end we V(t)
have = V
to e
make
0
−ξt
+
some dτ e −ξ(t−τ )
assumptions F(t). about the possible
stochastic realizations
force. In(3.4)
view of F(t), with the initi
possible realizations of F(t), with the initial velocity as a condition. To this
ofend
itswechaotic
have tocharacter,
make some
0
theassumptions
following assumptions seem
about the stochasticendtowe behave
force. Into
appropriate make some assumptions a
view
for • average
its
where
of V0 is the
average
chaotic evolution
properties: offollowing
initial velocity.
character, the velocity
We will V depends
now determine
assumptions onofcharacteristics
seem its
to bechaotic
averagesappropriate allof
overcharacter, the following ass
itsrandom
possible
for forceofF(t).
realizations
average properties: F(t),Suppose
with the initial the random
velocity force
asfor hasaverage
its
a condition. no memory properties:
To this
end we have to !F(t)"
make some = 0 assumptions about the stochastic force. In view (3.5)
of its!F(t) !F(t)"
"
)"V0 = the
· F(tcharacter,
chaotic C δ(t − t" ) assumptions seem to be appropriate
0 V0following (3.5)
!F(t)"
(3.6) = 0
"
for its!F(t) )"V0 = CV0 δ(t − t" )
· F(tproperties:
average !F(t) · F(t (3.6)
"
)"V0 = CV0 δ(t − t" )
where CV0 may depend on the initial velocity. Using Eqs. (3.4) - (3.6), we
where CV0 may!F(t)"
find depend=on0the initial velocity. Using Eqs. (3.4) - (3.6),(3.5) we
• what is the magnitude of the random forces, i.e. V0?
where C may depend on the initial velo
find !F(t) · F(t )"V = CV0 δ(t − !
" "
t )t (3.6)
0 find
!V(t)"V0 = V0 e−ξt + ! t dτ e−ξ(t−τ ) !F(τ )"V0
where CV0 may !V(t)" depend= onVthe 0 e −ξtinitial
+ 0 velocity.
dτ e Using
−ξ(t−τ )
!F(τ Eqs.
)" (3.4) - (3.6), we ! t
V0 V0 −ξt
find = V0 e −ξt 0 !V(t)"
(3.7)V0 = V 0 e + dτ
= V0 e−ξt ! ! t t (3.7) 0
2 −2ξt −ξ(2t−τ ) −ξt
!V(t) · !V(t)"
V(t)"VV0 = = VV0 0
ee−ξt + 2 ! t dτ−ξ(t−τ
+ dτ e e )
!F(τ V)"
0 · !F(τ )" V0 = V 0 e
!V(t) · V(t)"V00 = V02!e−2ξt +!20 0 dτ e−ξ(2t−τ ) V0 ·V!F(τ 0
)"V0 ! t
t
−ξt t 0 2 −2ξt
= V! e
0 t " −ξ(2t−τ −τ !)
!V(t) " · (3.7)
V(t)" = V 0e +2 d
+ dτ " t!dτt e−ξ(2t−τ −τ ! ) !F(τ ) · F(τ" )"V0
! V0
+ 2 0−2ξt dτ 0 dτ e −ξ(2t−τ ) !F(τ ) · F(τ )"V0 0
!V(t) · V(t)"V0 = V0 e0 + 02C "dτ e V0 · !F(τ )"V0 ! t ! t
V "
= V022!e−2ξt−2ξt −2ξt
#
+!CV000 " 1 − e−2ξt # . (3.8) + dτ dτ
= V0 e t + 2ξ
t 1−ξ(2t−τ
− e −τ ! ). (3.8) 0 046
+ dτ " 2ξ dτ e !F(τ ) · F(τ )"V0"
C "
0 0
where CV0 may depend=on V
thee−ξt
initial−ξt
velocity. Using Eqs. (3.4) - (3.6),(3.7)
we
find
= V0 e !
0 (3.7)
t ! t
2 −2ξt !
!V(t) · V(t)"V0 = V0 e 2t dτ e−ξ(2t−τ ) V−ξ(2t−τ
2+ −2ξt
2 )dτ e 0 · !F(τ) V)"V0
Fluctuation-dissipation theorem [2/2]
!V(t) · V(t)"V0 =−ξtV0 e 0 +−ξ(t−τ
!V(t)"V0 = V0 e! t + ! t dτ e 0
!F(τ )"V0 0 · !F(τ )"V0
" 0t t
−ξ(2t−τ −τ ! ) "
+ dτ dτ" e !F(τ ) · F(τ )"V0
! !
= V0 e−ξt
0 + !F(τ ) · (3.7)
−ξ(2t−τ −τ ! )
F(τ " )"V0
0
!dτ t
dτ e
0CV0 " 0 −ξ(2t−τ
VV0202ee−2ξt
−2ξt −2ξt )
#
!V(t) · V(t)"V0 == ++2 dτ1 C
e− e " V . 0 · !F(τ
# )"V0 (3.8)
2ξ
2 −2ξt V0
= V e
! t 0! t 0 + 1 − e−2ξt . (3.8)
The colloid is in thermalequilibrium
" with
2ξ
the −τ
−ξ(2t−τ solvent.
!) �" 2 � to3k
According BT
the
• colloid is in thermal equilibrium with the solvent, so V V0 =
+ dτ dτ e !F(τ ) · F(τ )"
equipartition
The colloid theorem, for large
is in thermal 0 t, Eq.
0
equilibrium (3.8) should
with be equal
the solvent. M
to 3kAccording
B T /M, to the
from which it follows that 2 −2ξt CV0
− e−2ξt
" #
equipartition theorem, = V0 efor large + t, 1Eq. (3.8) . should be equal to 3kB T /M,
(3.8)
• from this
from which it follows the
follows
k T fluctuation-dissipation
that
ξ
2ξ theorem
" B "
!F(t) ·isF(t
The colloid in )"thermal
=6 δ(t − t ). with the solvent. According to (3.9)
equilibrium the
M
equipartition theorem, kB Tt,ξ Eq. (3.8)
" for large " should be equal to 3kB T /M,
This is !F(t)
one · F(t )"
manifestation = 6
of the δ(t − t ).
fluctuation-dissipation theorem, which states (3.9)
from which it follows that M
that the systematic part of the microscopic force appearing as the friction is
This is one kBthe
T ξ correlation
manifestation
actually determined
"
!F(t) · F(t )" = 6 by δ(t of
− tthe
"
). fluctuation-dissipation
of the random force. theorem, which states
(3.9)
that the fluctuations
systematicare coupled
Mpart of the to dissipationforce appearing as the friction is
microscopic
26 actually determined by the correlation of the random force.
This is one manifestation of the fluctuation-dissipation theorem, which states
that the systematic part of the microscopic force appearing as the friction is
actually
26 determined by the correlation of the random force.
26
47
g Eq. (3.4) we get
3. BROWNIAN MOTION
# t # τ
V0 ! −ξt " −ξ(τ −τ ! ) 3. BROWNIAN MOTION
F(τ " ),
"
R0 + 1−e + dτ dτ e (3.10)
Einstein relation for the self-diffusion
ξ Integrating Eq. 0(3.4) we
0 get
3. BROWNIANcoefficient
MOTION
we calculateIntegrating
the mean Eq. (3.4)displacement
square we get # t # τ
! V0
" −ξt " −ξ(τ −τ ! ) "
R(t) = R0 + 1−e + # dτ t dτ
τ e F(τ ), (3.10)
Integrating Eq. (3.4) weB0 Tget!displacement...
ξV
#
2
2
% • from
V 0
!the
R(t) mean-square
= 2
−ξtR0 +
" 3k !
1 − e −ξt
"
+ 0
dτ
−ξt 0
dτ "
−2ξte
"−ξ(τ −τ !)
F(τ "
), (3.10)
) V0 = 2 1 − e + ξ 2 2ξt − 3 + 4e − e . (3.11)
fromξ which we V Mξ the" mean
calculate
# t #0 τ
square 0
displacement
0 −ξt " −ξ(τ −τ ! )
F(τ " ),
!
R(t) = R0 + 1−e + dτ dτ e (3.10)
from which
e t this$ becomes 2 % ξ we calculate the mean square displacement
V02 ! −ξt
0"
2 3k0
BT ! −ξt −2ξt
"
(R(t) − R0 ) V0 = 2 2 1 − e + 2
2ξt − 3 + 4e − e . (3.11)
from which
$
6kweTcalculate
2
% ξV0mean square
the ! " 3kB T
Mξ
−ξt 2 displacement
! −ξt −2ξt
"
2 (R(t) B − R0 ) V0 = 2 1 − e + 2ξt − 3 + 4e − e . (3.11)
%
− R0 ) = t, ξ Mξ 2 (3.12)
Mξ
$ For very 2large % t this
V0 !becomes
2
−ξt 2 3kB T
" ! −ξt −2ξt
"
(R(t) − R�0 ) V0 = 2 1�− e + 2ξt − 3 + 4e − e . (3.11)
For very large t this 6k
becomes
ξ 2 6k TB Mξ T 2
we get the limEinstein
$ (R(t) − equation
R02)% = forBthe self-diffusion
t ≡ 6Dt coefficient
t→∞ (R(t) − R0 ) = M ξt, (3.12)
For very large t this becomes
$ 2
% 6kB T
Mξ
(R(t) − R0 ) = t, (3.12)
BT ... we find the Einstein relation
Mξ for theforself-diffusion coefficient
, from$ which we2get % the 6k Einstein
T equation (3.13)
the self-diffusion coefficient
ζ (R(t) − R ) =
B
t, (3.12)
from which 0we get the Mξ Einstein equation for the self-diffusion coefficient
kB T2
e from
used "(R(t) D− =R0 ) #, = 6Dt and ζ = Mξ = 6πηs a. Notice that (3.13)
which we getkζBthe T Einstein equation for the self-diffusion coefficient
sion coefficientDD ,
= is independent of the mass M of the colloid. (3.13)
ζ D is independent of the mass of the colloid!
where kBhave
we T used "(R(t) − 2
0 # = 6Dt and ζ = Mξ = 6πηs a. Notice
)
D= , R (3.13) that
the
where weζ have used
self-diffusion coefficient
"(R(t) − DR is0independent
)2 # = 6Dt and ofζthe
= Mξ mass= M 6πηofs a.theNotice
colloid.
that
moluchowski and Langevin equation
the self-diffusion coefficient D is independent of the mass M of the colloid.
where we have used "(R(t) − R0 )2 # = 6Dt and ζ = Mξ = 6πηs a. Notice that 48
7)the
we self-diffusion
see that Smoluchowski
3.2 the coefficient
colloid loses D isitsindependent
memory
and ofLangevin
itsthe
of initial
massvelocity
M equation
of the colloid.
s chaotic character, the following assumptions seem to be appropriate
!F(t)" = 0
ts average properties: (3.5)
!F(t) · F(t" )"V0 = CV0 δ(t − t" ) (3.6)
!F(t)" = 0 (3.5)
ere!F(t)Overdamped motion?
F(t" )"depend
CV0· may V0 = on
C V the
0 δ(t initial
− t"
) velocity. Using Eqs. (3.4) - (3.6), we
(3.6)
d
re CV0 may depend on the initial! velocity. t
Using Eqs. (3.4) - (3.6), we
• reconsider
!V(t)" V0 the evolution
= V 0 e −ξt
+ of the colloidal
dτ e −ξ(t−τ )
!F(τ velocity
)"V0
! 0t
!V(t)"V0 == VV00ee−ξt −ξt
+ dτ e−ξ(t−τ ) !F(τ )"V0 (3.7)
0 ! t �
!V(t) ·•V(t)" = V 2 −ξt
= V00memory
e e−2ξt
+ 2 dτ e 3k
−ξ(2t−τ )B T
· !F(τ )"
1 (3.7)
colloidV0 loses of V 0 ≈ V 0 after τV= 0
t
! 0t
t
M ξ
!V(t) · V(t)"V0 = V02 e−2ξt +" 2 e−ξ(2t−τ )
! !
dτ −ξ(2t−τ −τ !V) 0 · !F(τ )"V"0
+ dτ 0dτ e −4 !F(τ ) −2 · F(τ )"V0
• traveled distance ! 0t / !radius 0t ≈ 10 ! − 10
+ dτ
2 −2ξt
" Cdτ V0 e"−ξ(2t−τ−2ξt
−τ # )
!F(τ ) · F(τ "
)"V0 (3.8)
= V0 0e +0 1−e .
2ξ
• on time scales on 2 −2ξtwhich CVparticle
0
" configurations
−2ξt
# change,
e colloidweis mayin thermal 0e
= Vequilibrium + 1 −the
with e solvent.
. (3.8)
restrict our attention 2ξ to the positions only, and the
According to
ipartition theorem,
average over forthe
large t, Eq. (3.8) should be equal to 3kB T /M,
velocities
mcolloid
which is in thermal
it follows that equilibrium with the solvent. According to the
partition theorem, for large t, Eq. (3.8) should be equal to 3kB T /M,
m which it follows
" kB T ξ
that
!F(t) · F(t )" = 6 δ(t − t" ). (3.9)
M
" kB T ξ "
!F(t) · F(t )" =
s is one manifestation 6 of δ(t
the − t ).
fluctuation-dissipation theorem, which (3.9)
states
M
t the systematic part of the microscopic force appearing as the friction is
is onedetermined
ually manifestation of the
by the fluctuation-dissipation
correlation of the random theorem, force. which states 49
the systematic part of the microscopic force appearing as the friction is
ery instant of time the particle is in thermal equ
ρm is where
the mass ρm density
is the mass of the colloid.
density of theTypical colloid. velocity,
values are l/a
Typical i.e.,≈the
values 10are particle
−2
l/a ≈velocity
10−2 is strongly dam
nanometre for asized colloid sized
nanometre and l/a ≈ 10and
colloid −4
for 10−4timescale.
l/aa≈micrometre for a sized A colloid
micrometreflux will sizedexist, given by
colloid
Smoluchowski equation [1/2]

er at room
in water temperature. We see that We
at room temperature. the see particles
that the have hardly have
particles moved hardly moved 1
time at possible
the time velocity
possible gradients
velocity have relaxedhave
gradients to equilibrium.
relaxed J(R, , t) = −D∇Ψ(R,
toRequilibrium.
0When WhenR0 ; t) − ζ Ψ(R, R
e interested
we areininterested
timescalesinontimescales which particle on which configurations change, we change, we
particle configurations
estrictmay our restrict
attentionour to attention
the spacetocoordinates,
the and The
space coordinates,
3. BROWNIAN first term
average
MOTION inthe
overaverage
and Eq. (3.15)
over the is the flux due to di
•TheQuestion:
ties.3. BROWNIAN time MOTION
velocities. howtime
development
The does ofa the
distribution
development distributionof the of colloidal
ofsecond
particles
distribution particles
term on theevolve
of isparticles
these fluxon inin thetime
these when gradie
“downhill”
they
cales istimegovernedare also
by thesubjected
Smoluchowski to an
scales is governed by the Smoluchowski equation. external
equation. potential
potential Φ(R), ? damped3. BROWNIAN
by the MOTION
friction coefficien
3.
3. BROWNIAN
BROWNIAN MOTION MOTION particular time must t, be
given
zero itandwasthe at distribution
R0 at t = 0. must It is
obeyassum
the
e Smoluchowski equation describes
The Smoluchowski equation thedescribes
time evolutionthe time of evolution
the proba-of the proba-
particular time t, given it was at Rery instant
at t = of time
0. It is the particle
assumed thatis inat thermal
ev- equilibrium with
densitybility
Ψ(R,density
R0 ; t) to is
Ψ(R,probability t) to density
findRa0 ; particle findataaparticle
0
velocity,
to find
particular
i.e.,
ata aparticle
position
particular
the particle
at
R position
at R at
a particular
position
velocity
at time
time
a t,, given it w
ery instant of time the particle is in thermal equilibrium Ψeq (R)
with C exp is
=respect to strongly
[−Φ(R)/(k
its damped
B T )] , on the
particular
particular time time t,t, given
given
given it it was
it was
was at at R
at 00 atat
at tt =
Rtimescale. = 0. 0.
A It
It is
flux assumed
iswill
assumed
exist, thatery
that
given at
at
by
instant
ev-
ev- of time the partic
velocity,
ery
ery instant
instant i.e., the
timeparticle
of time
of the particle
the velocity
particle inisthermal
is in
is stronglyequilibrium
thermal damped onwith
equilibrium the respect
with Smoluchowski
27 velocity,
respect to its
to i.e.,
27 the particle
its constant. velo
where C is a normalization Using this
timescale.
• in the
velocity,
velocity, Aoverdamped
i.e.,
i.e., fluxparticle
the
the will exist,
particle given
limit
velocity
velocity bystrongly
a flux
is
is J will exist,
strongly damped
damped givenonleads
on by Smoluchowski
the
the timescale.
Smoluchowski 1 equation
A flux will exist,Ing
J = 0, to the Einstein
J(R, R0, t) = −D∇Ψ(R, R0 ; t) − Ψ(R, R0; t)∇Φ(R). (3.13).
timescale. A
timescale. A flux
flux will
will exist,
exist, given
given by by 1 particles are generated or destroyed, ζ so
J(R, R0, t) = −D∇Ψ(R, R0 ; t) − Ψ(R, R0; t)∇Φ(R). (3.15)J(R, R0, t) = −D∇Ψ(R
ζ
The11first term in Eq.
J(R, R
J(R, R00,, t)
t) = = −D∇Ψ(R,
−D∇Ψ(R, R R00;; t)
t) − − Ψ(R, Ψ(R, R ∂ (3.15) is the(3.15)
R00;; t)∇Φ(R).
t)∇Φ(R).
flux due to diffusion of the
(3.15)
second ζ
ζ term is the Ψ(R,
flux in R
the 0 ; t) = −∇ · J(R,
“downhill” R0, t).direction
gradient
The first term in Eq. (3.15) is the flux due to diffusion∂tof the particle. The The
first term in Eq. (3.15) i
second term is flux
the due
flux to
in diffusion
the potential
“downhill” flux in downhill
Φ(R),
gradient damped
directiongradient
by
of the
the friction
external coefficient ζ. At equili
The first
The first term
term in in Eq.
Eq.
of (3.15)
(3.15)
the is the
is
particles the fluxflux due
due of to
to diffusion
diffusion
Combining
external of
of the
the
potential Eq.particle.
particle. second
(3.15) The
The
with term
the is the flux
above in th
equation
potential Φ(R), damped by the frictionmustcoefficient
be zero and ζ. theequilibrium,
At distributionthe mustflux obey the Boltzmann
second term
second term is is thethe flux
flux inin the
the “downhill”
“downhill” gradient gradient direction
direction
arrive at theof potentialequation
ofSmoluchowski
the external
the external Φ(R), damped by th
must be Φ(R), zero and the distribution mustcoefficient
obey the ζ. Boltzmann distributionmust
potential
potential Φ(R), damped
damped by
by the
the
• note that at equilibrium this re-confirms friction
friction coefficient
Ψ eq (R) ζ. At
At equilibrium,
equilibrium,
the Einstein
= C exp [−Φ(R)/(k the
the
relation: )]be
B Tflux , zero
flux ! and the distribu
∂ 1
must be
must be zero
zero and
and the
the
Ψeq (R) = C exp�[−Φ(R)/(k distribution
distribution must
must
T )] , obey
obey the
the Boltzmann
Boltzmann distribution
distribution
Ψ(R, R0; t) = (3.16)∇ · Ψ(R, R0; t)∇Φ(R
� B
where C is a normalization ∂t constant. Ψ eq (R)
Using = Cinexp
ζ this Eq.[−Φ(R
(3.15
Φ(R) kB T
where C ΨΨ (R)
(R)
(R) =
== CC exp
exp [−Φ(R)/(k
[−Φ(R)/(k
−
eq is a normalization constant.
eq
eq B
B TT )]
)]
J ,
,
= 0,
Using leads
this to
in the
= Einstein
D Eq. (3.15) equation
while (3.16)
(3.16)
setting (3.13). In general, we a
kB T ζ where C is a normalization c
particles are generated
lim Ψ(R, orassume
destroyed,
R0; t)Jthat = so
J =
where C
where 0, leads
C is to the
is aa normalization Einstein
normalization constant. equation
constant. Using (3.13).
Using this In
this ingeneral,
in Eq.
Eq. we
(3.15) while
t→0(3.15) while setting= 0,δ(R
setting noleads −R to0 ).the Einstein
50 e
particles are generated or destroyed, so ∂ particles are generated or des
in section must
potential an
2.3Φ(R),
when
be autocorrelation
zero dealing
and theby
damped withthefunction
collective
distribution
friction mustthis says
obey that
diffusion.
coefficient For
the
ζ. AtBoltzmann
equilibrium, distribution
the flux
n function
must be this
zero says and that
the distribution # must obey the Boltzmann distribution
2
Ψeq (R)d= C exp [−Φ(R)/(k T )]
! , " (3.16)
# B 2
#
2
"A(t)A(0)# # = − Ȧ < 0, (2.7)
Ψeq (R)
Ȧ2C = Cdt exp [−Φ(R)/(kt=0 B T )] , (3.16)
# ! "
Smoluchowski equation [2/2]
#
0)### =where
− < 0, (2.7)
is a normalization constant. Using this in Eq. (3.15) while setting
t=0
where JC = 0, aleads
isi.e. to the Einstein
normalization constant.equation
Using (3.13).
this InEq.
in general,
(3.15)we assume
while settingthat no
any time autocorrelation function is initially curved negatively.
J = 0,
correlation particles
leads
function arethe
to isgenerated or
Einsteincurved
initially destroyed,
equation (3.13).
negatively.so In general, we assume that no
particles
• inserting are ∂ generated
the flux equationor destroyed, in ansoequation for particle number conservation,
Ψ(R, R0; t) = −∇ · J(R, R0, t). (3.17)
∂ 2.2 ∂t Self-diffusion and the velocity autocorre-
iffusion ∂t
and
Ψ(R, R0the
; t) = −∇ velocity
· J(R, R0, t). autocorre- (3.17)
Combining Eq.lation
(3.15) with the above equation of particle conservation we
function
function...arrive
Combiningwe arrive
atEq. atSmoluchowski
the(3.15) the with
important Smoluchowski
the equation
above equation of particle equation:conservation we
arrive at the SupposeSmoluchowskiwe label equation
some !
particles inside a "
very small region (a dot) in an
some particles∂ inside a very
Ψ(R, R0homogeneous small
; t) = !∇ · fluid,
1
region
Ψ(R,at (a dot) in
; t)∇Φ(R) an +∇ · [D∇Ψ(R,
otherwise R0time t"= 0 at position r(0). 0 ; t)] the dot,
RWhen
neous fluid, ∂ at∂t time t = 0 at position 1 r(0).ζ When the dot,
Ψ(R,
althoughR0; t) on = a∇ at· r(0),Ψ(R,
macroscopic ; t)∇Φ(R)
R0scale + ∇ · [D∇Ψ(R,
concentrated at Ris0 ;dilute
t)]
acroscopic ∂t scale concentrated ζ is dilute enough on r(0), (3.18)enough on
we may consider a molecular
lim Ψ(R, ;
scale,
theRconcentration
t) =
we may
δ(R decay
− R
consider
). as due thethe
to concentration decay (3.18)
as due to the
(3.19)
0 0
e separate labelled self-diffusion
t→0 particles. of The
the separate
conditional labelled particles. The conditional probability
probability
lim Ψ(R, R0; t) = δ(R − R0 ). (3.19)
ticle is atThe P (r, t)
r atSmoluchowski
t→0 that a
time t, given equation particle is
it was atdescribesat r at
r(0) at time time t,
howt particle
= 0, given it was at
distributionr(0) at time t = 0,
functions
inedThe from
• this Fick’s
change
maymay
Smoluchowski law:
bethen
in time
viewed be
andobtained
equationas from Fick’s
is fundamental
adescribes
generalization howto the law:
ofnon-equilibrium
Fick’s
particle law
distribution statistical
functions me-
change chanics
in time of overdamped
and is fundamental particles such to the as non-equilibrium
colloids and polymers.statistical me-
D∇2chanics It overdamped
P (r, t), of can be ∂Pshown
(r, t)particles
(we will notasdocolloids
2such that here)and
(2.8)that the explicit equations
polymers.
= D∇ P (r, t), (2.8)
Itofcan
motionbe shown for the∂t(we particles,
will noti.e. do the
thatLangevin
here) that equations, which
the explicit lead to the
equations
Smoluchowski
of motion equation i.e.
for the Pparticles, are the Langevin equations, which lead to the
boundary condition
together (r,
with 0) =
the δ(r − r(0)). condition
boundary D is the self- P (r, 0) = δ(r − r(0)). D is the self-
Smoluchowski
nt, which has units dR equation are
of length squared 1 over time (m /s). 2
diffusion coefficient, = − ∇Φwhich + ∇Dhas + f units of length squared over(3.20) time (m2 /s).
displacement of dR the labelled
dt square 1 particles can be related to the
ζ displacement
The mean = − ∇Φ + ∇D + f of the labelled particles can be related to the51
(3.20)
dt"f(t)# =ζ 0 (3.21)
! "
∂ 1
Ψ(R, R0; t) = ∇ · Ψ(R, R0; t)∇Φ(R) + ∇ · [D∇Ψ(R, R0; t)]
∂t ζ
Langevin equations
(3.18)
lim Ψ(R, R0; t) = δ(R − R0 ). (3.19)
t→0
The Smoluchowski equation describes how particle distribution functions

• Smoluchowski
change inequation
time anddescribes how particle
is fundamental distribution functions
to the non-equilibrium change
statistical me-
chanics of overdamped particles such as colloids and polymers.
It can be shown (though we will not do that here) that the explicit equa-
• the corresponding explicit stochastic equations of motion for the particles
tions of motion for the particles, i.e. the Langevin equations, which lead to
are called the Langevin equations
the Smoluchowski equation are
dR 1
= − ∇Φ + ∇D + f (3.20)
dt ζ
"f(t)# = 0 (3.21)
"f(t)f(t" )# = 2D Īδ(t − t" ). (3.22)
where Ī denotes the 3-dimensional unit matrix Iαβ = δαβ . We use these
equations
• the Langevin in the next
equations arechapter
used asto derive the equations
a starting point of motion for a polymer.
for many soft matter simulations and theories;
in the next
28 chapter we will use them to build a theory of polymer dynamics
52
where C is a normalization V0 constant.
0 Using this in Eq. (3.15)0 while setting V0
Integrating Eq. (3.4) we get
J = 0, leads to the Einstein equation ! t (3.13). ! t 0 In general, we assume that no
particles are generated or destroyed, #sot" # τ −ξ(2t−τ −τ ! )
V0 ! +
−ξt
" dτ dτ e " −ξ(τ −τ ! ) !F(τ
" ) · F(τ " )"V0
R(t) = R0 + 1−e + dτ0 dτ e F(τ ), (3.10)
∂ ξ 0
0 0
Ψ(R, R0; t) = −∇ · J(R, 2
R 0 , t). CV0 "
−2ξt −2ξt
(3.17)
∂t
#
from which we calculate the V0 e square
= mean + displacement
1−e . (3.8)
2ξ
Combining Eq. (3.15) with the above equation of particle conservation we
2 !
• Brownian V "2 3kB T !
arrive
The at− R0 ) motion
the
colloid Smoluchowski
is
2 in of1colloids
thermal
0 equation is caused
equilibrium
−ξt withby collisions
the solvent.
−ξt with solvent
According
−2ξt molecules
to the
$ % "
(R(t) = 2
− e + 2
2ξt − 3 + 4e − e . (3.11)
equipartition
V0
theorem, ξ for! large t,Mξ Eq. (3.8) "should be equal to 3k T /M,
∂ 1 B
Ψ(R,it ; t) = that
R0follows ∇ · Ψ(R, R0; t)∇Φ(R) + ∇ · [D∇Ψ(R, R0; t)]
• the
from
For ∂t random
which
very large t thisforces
becomesare related
ζ to the friction experienced by a colloid
k
6k T ξ (3.18)
B
!F(t) · " 2
)" = 6 B
δ(t − t"
). (fluctuation-dissipation theorem) (3.12) (3.9)
$ %
(R(t) −F(tR )
lim Ψ(R, R0; t) = δ(R
0 = −t, R ). (3.19)
t→0 MξM 0
• orSmoluchowski
The
from
This equivalently
which
is one we theEinstein
get the
manifestation diffusion
equation is related
ofdescribes
equation
the how to self-diffusion
for the the friction
particle distribution
fluctuation-dissipation functions
coefficient
theorem, which states
change
that theinsystematic
time and ispart
fundamental to the non-equilibrium
of the microscopic statistical
force appearing as theme-
friction is
kB T
chanics of overdamped particles such as colloids and relation)
polymers.
actuallyD= determined
, by the correlation(Einstein
of the random force. (3.13)
It can be shown
ζ (though we will not do that here) that the explicit equa-
tions of motion for the particles, i.e. the Langevin equations, which lead to
26• Smoluchowski
where
the motion
we haveof used
colloidal particles
"(R(t)
equation R0 )2 # in
−are an external
= 6Dt potential
and ζ = Mξ = 6πηs a. Notice that
the self-diffusion
dR coefficient
1 D is independent of the mass M of the colloid.
= − ∇Φ + ∇D + f (3.20)
dt ζ
(Langevin equations)
3.2 Smoluchowski
"f(t)# = 0 and Langevin equation(3.21)
"f(t)f(t" )# = 2D Īδ(t − t" ). (3.22)
From Eq. (3.7) we see that the colloid loses its memory of its initial velocity
where
after Ī denotes the 3-dimensional unit matrix I velocity
= δαβmay
. Webe use
put these
& a time τ ≈ 1/ξ. Using equipartition its initialαβ equal 53
Intermezzo: hydrodynamic interactions
• note that Langevin equations neglect

hydrodynamic interactions:
when a colloid moves, it induces a solvent flow
which affects the motion of other colloids too
• example: sedimentation of micron-sized colloids

in a closed vessel under the influence of gravity
• hydrodynamic interactions may be neglected
• at very low colloid concentration, or
• when other interparticle interactions dominate
54
Chapter 4:
The dynamics of unentangled polymer liquids
55
Rouse chains can be expressed as
N & '2 N
kB T !Xk (t) · Xk (0)" ckB T
Polymer liquids
# #
G (t) = =
V p=1
!Xk2 " N + 1 p=
• long molecules consisting of

where c = N/V is the number density of beads.
N = 100 - 100.000 identical monomers
The Rouse model predicts a viscosity, at constant m
c and segmental friction ζ, proportional to N :
• characteristics of polymeric liquids:
∞ N
ckB T τ1 1 ckB T τ1 π 2
( #
• high shear viscosity η= dtG(t) ≈ 2
≈
0 N + 1 2 p=1 p N +1 2 6
• often highly elastic
(memory in shearThis has been
relaxation confirmed
modulus G(t)) for concentrated polymers with
Concentrated polymers of high molecular weight give d
• slow dynamics ing the importance of entanglements.
(low diffusion coefficient)
2
A somewhat stronger N dependence is often observed beca
important, the segmental friction coefficient increase with incre
56
40
Polymers
Ψ r are=long linear
N
ψ (r ) macromolecules made up of a large number of chem-
. (4.2)
Polymers are long linear macromolecules made up of a large number of chem-
i
ical units or monomers,
i=1 which are linked together through covalent bonds.
ical units or monomers, which are linked together through covalent bonds.
The number of monomers per polymer may vary from a hundred to many
The
Assumenumberfurther of that
monomers
the bond pervector
polymer may vary
probability from aψ (r
density hundred to many
thousands. We can describe the conformation of a polymer i by givingonly
) depends the
on Conformations of a polymer
thousands.
the length
positions
positionswe
We
of its
of write
canbond
of backbone
the describe vector the conformation
atoms.and Thehas zero mean,
positions
of a"rpolymer
of the
its backbone atoms. The positions of the remaining atoms then
i# = 0. For
remaining
by the
giving
atomssecondthe
then
moment
usually follow by simple chemical rules. So, suppose we have N +1 monomers,
usually follow(by simple chemical rules. So, suppose we have N +1 monomers,
with &N + 1 position vectors
with• N + 1 position
specified 3 N+1 vectors 2
r 2 = dby r r 2 ψ(r)backbone atom positions
'
≡b , (4.3)
R0 , R1 , . . . , RN .
R0 , R1 , . . . , RN .
where we have defined the statistical segment (or Kuhn) length b. Let
We then
or have
by N N bond
bond vectors
vectors
We then have N bond vectors
Ω (R; N) be the probability distribution function for the
4. THE end-to-end
DYNAMICS vector
OF UNENTANGLED POLYMERIC LIQU
R given 4. THE DYNAMICS OF UNENTANGLED POLYMERIC LIQUIDS
r1 that
= R1we − have
R0 , . .a. ,chain
rN =ofRN N
bonds,
− RN −1 .
r1 = R 1 − R )0 * N N N+, N −1 .
, . . . , r = R − R R
For
!example, one of
can For example,
measure one can measure
the be
end-to-end vector, the end-to-end
defined as vect
Much•Ω of
large the static
scale and dynamic behavior polymers can explained by
Much of(R;theN) = properties
static δandRdynamic− independent ,
ri behavior of of chemistry:
polymers can be explained (4.4)by
models which are surprisingly simple. This is possible because the r1 global,
N
models which are surprisingly simple. This is possible
i=1 N because the global,
large scale properties of polymers do not depend !on theRchemical details
!
large e.g.
scaleend-to-end
properties ofvector polymers do R = − R = Rri1.
Nnot depend on i . the chemical details
= − N 0
R R R 0 = r RN (
of the monomers, except for some species-dependent
where δ is the Dirac-delta function. The central limit theorem then states “effective” parameters.
R 0
i=1
of the monomers, except for some species-dependent i=1 “effective” parameters.
that for large enough N:
often has a Gaussian distribution
If the end-to-end vector If the end-to-end
is measured forvector
a largeisnumber
measured
29 of for a largeinnum
polymers am
.3/2 29
one will find that the distribution of end-to-end vec
3 one will find that 3R2 the distribution of end-to-end vectors is Gaussian and
- - .
Ω (R; N) = 2the expmean
root − squared
2
. the root meandistance
end-to-end squared (4.5)with
end-to-end
scales √ distance
the square scales
root ow
2πNb 2Nb " √of bonds, "R2 # ∝ N, irrespective of th
"
number
number of bonds, "R2 # ∝ N, irrespective of the chemical details. Th
WHY? a consequence of the a consequence
central limit of the central limit theorem.
theorem.
30
4.1.2 4.1.2
The central The
limit centraland
theorem limit theorem
polymer 57and
elastic
he end-to-end vector, defined as " √
d for a large number of polymers in a melt, 4. THE DYNAMICS numberOF UNENTANGLED
of bonds, "RPOLYMERIC 2 # ∝ N, irrespecLIQUID
ofIfend-to-end
the end-to-endvectorsvector is measured
is Gaussian and forthata large number of polymers
a consequencein a melt,
of the central limit theore
one will scales
distance find that
withthe thedistribution
square rootofofend-to-end
the vectors is Gaussian and that
(4.1) scales
For example, one canroot measure the end-to-end vector,
the
The central limit theorem
root mean squared
rrespective of the chemical
theorem.
" end-to-end
2
√
details. distance
This is
number of bonds, "R # ∝ N, irrespective of the chemical
with the square
4.1.2details. The
of the
Thiscentral
N
is limit theor
a consequence of the central limit theorem.
!
red for a large number of polymers in a melt, R = RConsider
N − R0 =a chain ri . consisting of N indepe
n of end-to-end vectors is Gaussian and that
theorem • if enough
and consecutive
polymer monomers
elasticity are mean that i=1
the orientation and length o
nd4.1.2
distance The scales central
with the square limit root of the and polymer
theorem others. A justification
elasticity will be given at$ the
combined into one segment, vectors If the end-to-end vector is measured for a large numbe
,independent
irrespective of bondthe vectors
chemicalr .details.By thisThis
we is density in configuration space Ψ
#
r N
m
itConsiderconnecting
a chain the segments
i are independent
one will find that the distribution of end-to-end vecto
engththeorem.
of each bondconsisting
is independentof N independent
of all bond vectors ri . By this we
mean (neglecting
that the long-range
orientation and excluded
length of volume)
the bond
each root mean
is squared
independent end-to-end
of √all
N
distance scales with
n at$ the end of this section. The probability number of bonds,
" # N2 #$ ∝ %
"R N,ψ irrespective of the c
Nothers. A justification
may thenand be written will
as be given at the end of this section. The Ψ probability
r = (r ) .
tr density
theoremi.e.inthe polymer
probabilityspace
configuration densityelasticity
Ψ r for may
# N$
segment-to-
a consequence
then be written as of the central i=1
i
limit theorem.
segmentbond
N independent bond vectors
vectors ri . factorizes:
By this we
N
(4.2) Assume further that the bond vector pro
length of# each bond
% is independent of all 4.1.2 The central limit theorem and p
Ψ rN =
$
ψ (ri ) . on the length of the bond vector and has
(4.2)
ven# Nat$ the end of this section. The probability
i=1 Consider a chainmoment we write
consisting of N independent bond
Ψ r may then be written as
• now suppose
tor probability density ψ (ri ) depends
depends only only on length
mean thatofthe
bond vector;
orientation ( and length of each bond
ndAssume
has zero further
mean, that
"r # the0.bond
= For vector
the probability density ψ (ri ) depends
define the statistical segment (or Kuhn) length b as
second 2 only3
r r#2 ψ(r) b2 , of this s
& '
i others. A r
justification =
will be dgiven at$ the≡ end
on the length of the bond vector and has zero mean, "ri # = 0. For the second
(4.2) density in configuration space Ψ rN may then be w
moment we write
where we have defined the statistical
• the
& 2 'central3 limit
( theorem then (4.3)states that for large N,
N irrespective of the
2 2 Ω (R;
% N) be the probability distribution
precise form
r = d r of ψ (ri ), depends
r ψ(r) ≡ bthe distribution
, of Ψ r will
N
= beψ Gaussian (4.3)
# $
ector probability density only (ri ) .
R given that we have a chain of N bond
and has zero mean, "ri # = 0. For the second i=1
stical segment (or Kuhn) length b. Let
where we have defined the statistical segment (or Kuhn) length b. Let
) * N
+,
bution function for the end-to-end vector Assume further that the bond vector probability !
58dens
Ω
N bonds,(R; N) be the probability distribution function for the Ω (R;
end-to-end N) =
vector δ R − r i ,
on the length of the bond vector and has zero mean, "
these segments become independent
R given that we have a chain of N bonds,
of each other.
So, irrespective of)the * preciseN form +,of the bond length distribution ψ(r), the
end-to-end vector will have! a Gaussian distribution with zero mean and a
Ω (R; N) =of the δ R − form ri of, the bond length distribution ψ(r), (4.4)
Polymer elasticity
So, irrespective
variance given by
end-to-end vector will
precise
have i=1
a Gaussian distribution with zero mean and a
the
! 2" 2
variance R =
given Nb
by
where δ is the Dirac-delta. function. The central limit theorem then states (4.6)
that for
! large
• end-to-end
2
" enough
vector
2 N: distribution
UsingR Ω=(R; NbN), . we can obtain an interesting insight in the thermody- (4.6)
namic behaviour of - a polymer .3/2 chain.- The2 entropy
. of a chain in which the
Using Ωvector
end-to-end (R; N),Rwe is3 kept
can obtain
fixed, 3R
an interesting
absorbing all insight ininto
constants the athermody-
reference
Ω (R; N) = exp − . (4.5)
namic
entropy,behaviour
is given by of 2πNb
a polymer chain. 2Nb
2 The entropy of a chain in which the
2
end-to-end vector R is kept fixed, absorbing all constants into a reference

2
entropy, is given by 3kR
30• entropy
S (R; N) and= free
kB lnenergy
Ω (R; N) of=a Spolymer
0 − 2chain
. in which R is kept(4.7) constant:
2Nb2
3kR
The freeS (R; N) =iskthen
energy B ln Ω (R; N) = S0 − 2
. (4.7)
2Nb
The free energy is then 3kB T R2
A = U − T S = A0 + 2
. (4.8)
2Nb 2
3kB T R
We see A that
=U− theT Sfree
=A 0 +
energy .
is related quadratically to the end-to-end(4.8) dis-
entropic spring with spring 2Nb constant
2
tance, as if the chain is a harmonic (Hookean) spring with spring constant
We
3kB Tsee
/Nbthat
2
. 3kthe
B Tfree
Unlike anenergy
ordinaryis related
spring, quadratically
however, the to the end-to-end
strength dis-
of the spring1 2
tance, kas=ifwith
increases thetemperature!
2chain is a harmonicThese (Hookean)
springs are springreferred
often toΦ(x)
with springas = kx
constant
entropic
2 Nb 2
3kB T /Nb . Unlike an ordinary spring, however, the strength of the spring
springs.
increases withintemperature!
Of course, a real polymer These springsconnecting
the vectors are often referred to as
consecutive entropic
monomers
springs.
do not take up random orientations. However, if enough (say λ) consecutive
59
Of course,
monomers areincombined
a real polymer
into onethesegment
vectors connecting consecutive
with center-of-mass monomers
position Ri ,
4.1.3 The Gaussian chain
4.1.3 Theestablished
Now we have Gaussian that chain
global conformational properties of polymers
Gaussian chain
are largely
Now
model
independent
we have
available,
of the
established
consistent
chemical
that
with
global details, we can start
conformational
a Gaussian end-to-end
from theofsimplest
properties
distribution.
polymers
This
are largely independent of the chemical details, we can start from the simplest
model is
model one in which
available, every bond
consistent with vector itself isend-to-end
a Gaussian Gaussian distributed,
distribution. This
• simplest
model is one !polymer
in which
3 model: every
every !bond
"3/2 bond
vector
3 2 is Gaussianly
" itself distributed
is Gaussian distributed,
ψ (r) = ! 2 "3/2exp !− 2 r ". (4.9)
2πb3 2b3
2
ψ (r) = 2
exp − 2
r . (4.9)
Such a Gaussian2πb 2b
chain is often represented by a mechanical model of beads
connected by harmonic springs, as in Fig. 4.2. The potential energy of such
Such a Gaussian chain is often represented by a mechanical model of beads
a chain
• is given by:
mechanical equivalent: beads connected
connected by harmonic springs, as in Fig. 4.2. by Thesprings
potential energy of such
a chain is given by: 1 # N
Φ (r1 , . . . , rN ) = k ri2 . (4.10)
2 i=1 N
1 #
Φ (r1 , . . . , rN ) = k ri2 . (4.10)
It is easy to see that if3k 2the Tspring constant is chosen equal to k = 3kB T /b2
B i=1
k=
the Boltzmann distribution b2 of the bond vectors obeys Eqs. (4.2) and (4.9). 2
It
Theis Gaussian
easy to see thatis ifused
chain the as
spring constant
a starting pointisfor
chosen equal model.
the Rouse to k = 3kB T /b
the Boltzmann distribution of the bond vectors obeys Eqs. (4.2) and (4.9).
The Gaussian chain is used as a starting point for the Rouse model.
4.2 Rouse dynamics of a polymer
4.2 Rouse
In the previous dynamics
section of athepolymer
we have introduced Gaussian chain as a model for
the equilibrium (static) properties of polymers. We will now adjust it such 60
R(t). The colloidal particle will continuously collide with the solvent molecules
Because on average the colloid will collide more often on the front side than
on the back side, it will experience a systematic force proportional with its
velocity V, and directed opposite to its velocity. The colloid will also experi-
ence a random or stochastic force F(t). These forces are summarized in Fig.
From statics to dynamics 3.1 The equations of motion then read1
dR
= V (3.1)
dt
dV
= −ξV + F. (3.2)
dt
• statics: Gaussian chain 1
Note that we have divided all forces by the mass M of the colloid. Consequently, F(t)
is an acceleration and the friction constant ξ is a frequency.
Figure 3.1: A colloid

moving with velocity
• dynamics: collisions with neighbouring
V will particles
experience a cause
friction force −ξV op-
posite to its velocity
F
• friction forces on each bead and random forces F -!v v
due to the continuous
bombardment of sol-
• random forces on each bead vent molecules.
• note that the resulting Rouse model 25
• neglects hydrodynamic interactions (important in dilute polymer solutions)
• neglects entanglements (important for very long polymers)
• applies best to melts of relatively short polymer chains (100 monomers)
61
velocity field like in a normal liquid is highly improbable, meaning
of as being caused by the motion of a chain relative to the rest of the material,
no hydrodynamic
which to a first approximation may be taken interactions.
to be at rest; propagation of a
velocity field like in a normal We liquid
will start withimprobable,
is highly a Gaussianmeaning
chain consisting
there are of N + 1 beads c
Langevininteractions.
no hydrodynamic equations
by N springs.for If wethe
focusRouse chain
on one bead, while keeping all other be
We will start with aweGaussian
see thatchainthe consisting
external field
of N +Φ 1inbeads
which that bead moves is gen
connected
by N springs. If we focus connections to itswhile
on one bead, predecessor andother
keeping all successor. We assume that each
beads fixed,
we •see thatbead
each the external thefield
feelsthe same
same Φ in which ζ,
friction
friction that
thatbead
itsmoves
motion is generated by
is overdamped, and that the
connections to its predecessor and D
coefficient successor.
= kB T /ζ Weisassume that eachofbead
independent the feels
position Rn of the b
the •same
for each bead,
friction the
ζ, that external
model for apotential
its motion isis called
is overdamped,
polymer andthe
that the diffusion
Rouse chain. According to Eq
generated
coefficient D = kby bonds
B T /ζ to predecessor
is independent of the and
position Rn of the bead. This
(3.22) the Langevin equations describing the motion of a Rouse ch
modelsuccessor
for a polymer is called the Rouse chain. According to Eqs. (3.20)-
dRthe
(3.22) the Langevin equations describing 0 3kofB Ta Rouse chain are
motion
= − (R0 − R1 ) + f0
dR0 3kB T dt ζb2
= − (R0 − R1 ) + f0 (4.11)
dt ζb 2 dR n 3k B T
= − (2Rn − Rn−1 − Rn+1 ) + fn
dRn 3kB T dt ζb 2
= − 2
(2R n − R n−1 − R n+1 ) + fn (4.12)
dt ζb dRN 3kB T
dRN 3kB T = − 2
(RN − RN −1 ) + fN
= − (R N − dtRN −1 ) + fN ζb (4.13)
dt ζb 2
"fn (t)# = 0
"fn (t)# = 0 " " (4.14)
"fn (t) fm (t
" )# = 2D Īδnm δ(t − t ).
"fn (t) fm (t" )# = 2D Īδnm δ(t − t ). (4.15)
(3N+3) coupled stochastic differential equations

4.3 Normal 4.3
mode Normal
analysis mode analysis
62
Equations
Equations (4.11) - (4.13) are (3N (4.11) - (4.13)
+ 3) coupled are (3Ndifferential
stochastic + 3) coupled
equa-stochastic differen
"f"fnn(t) fm(t(t" )#
(t)fm "
)# == 2D
2DĪδĪδ
nm δ(t
nm t" ).t" ).
δ(t−− (4.15)
(4.15)
4.3
4.3 Normal
Normal mode
mode analysis
analysis
Normal modes [1/3]
Equations
Equations(4.11)
(4.11)- -(4.13)
(4.13)areare(3N
(3N++3)3)coupled
coupled stochastic differential
stochastic equa-
differential equa-
tions.
tions. InInorder
ordertotosolve
solvethem,
them,wewewill
willfirst ignore
first ignorethethe
stochastic forces
stochastic fn and
forces fn and
try•specific
try first ignore
specific theofrandom
solutions
solutions ofthe forcesform:
thefollowing
following and
form:try a specific solution of the form
Rnn(t)
R (t)==X(t)
X(t)cos(an
cos(an++c).c). (4.16)
(4.16)
The this yields
The equations
equations ofofmotion
motionthenthenread
read
dXcos c = − 3k3k
dX BT T
B {cos c − cos(a + c)} X
cos c = − 2 {cos c − cos(a + c)} X (4.17)
(4.17)
dtdt ζbζb2
dX
dXcos(na + c) = − 3k3k BT T
B 4 sin2 (a/2)
2 cos(na + c)X (4.18)
dt cos(na + c) = − ζb 2
2
4 sin (a/2) cos(na + c)X (4.18)
dt ζb
dX 3kB T
dXcos(Na + c) = − 3k B T{cos(Na + c) − cos ((N − 1)a + c)}(4.19)X,
dt cos(Na + c) = − 2
ζb 2 {cos(Na + c) − cos ((N − 1)a + c)} X,
dt ζb
where we have used (4.19)
where 2 cos(na
we have+ used
c) − cos ((n − 1)a + c) − cos ((n + 1)a + c)
[make use of cos(x + y) = cos(x) cos(y) − sin(x) 2 sin(y)
= cos(na + c) {2 − 2 cos a} = cos(na + c)4 sin (a/2). (4.20)
2 cos(na + and
c) −sin(x +−
cos ((n y)1)a
= cos(x) sin(y)
+ c) − cos ((n + 1)a + c)cos(y) ]
sin(x)
= cos(na + c) {2 − 2 cos a} = cos(na + c)4 sin2 (a/2). (4.20)
33
33
63
4. THE DYNAMICS OF UNENTANGLED POLYMERIC LIQUIDS
Normal modes [2/3]
The boundaries
The boundaries of theofchain,
the chain, Eqs. (4.17)
Eqs. (4.17) and (4.19),
and (4.19), are consistent
are consistent wi
with Eq.
The boundaries of the chain,
(4.18) if weEqs. (4.17) and (4.19), are consistent with Eq.
choose
(4.18) if we choose
• consistent
(4.18) description of boundaries of chain if we choose 2
if we choose
cos c coscos(a
− c −2cos(a
+ c) +=c) 4 sin
= 2 (a/2)
4 sin cos
(a/2)
c cos c (4.21)
cos c − cos(a + c) = 4 sin (a/2) cos c (4.21)
2
cos(Na + c) − cos ((N − 1)a + c) = 2 4 sin (a/2) cos(Na + c),
cos(Na + c) −cos(Na
cos ((N + − c)
1)a−+cosc) ((N
= −4 1)a
sin2 + c) cos(Na
(a/2) = 4 sin + (a/2) cos(Na + c), (4.22)
c), (4.22)
which which
is is equivalent
equivalent to to
which is equivalent to
cos(a − c) = cos c
cos(a − c) = cos(a cos c − c) = cos c (4.23) (4.23)
cos cos+((N
((N 1)a ++1)a
c) +=c) cos(Na
= cos(Na + c). + c). (4.24)
cos ((N + 1)a + c) = cos(Na + c). (4.24)
We •find We We
find find independent
independent
we find independent solutions from
independent solutions from solutions
solutions from from
a − c = c a − ca − =cc= c (4.25) (4.25)

(N + 1)a + c (N = +p2π (N
1)a−++Na
1)a
c − +
=c,c p2π
= −p2π Na−−Na c, − c, (4.26) (4.26)
wherewhere where where
p is anp integer.p is
So an
is an integer p is an integer.
integer.
finally So So finally
finally
pπ pπ pπ pπ pπ pπ
a= ,c=a=a/2a== , c = ,a/2 c. ==a/2 = . . (4.27) (4.27)
N +1 N +2(N 1N ++11) 2(N +2(N1) + 1)
Eq. a Eq.
Eq. (4.16), with and (4.16),
c from
(4.16), with
with Eq.
a and a cand
(4.27), c from
decouples
from Eq.
the(4.27),
Eq. (4.27), set decouples
the setthe
of differential
decouples of set of differ
differential
equations. Toequations.
find equations.
the general solution
To
To find the to Eqs.
findgeneral
the (4.11)
general (4.15)
solution
solution to Eqs. we form
to Eqs.
(4.11) to a(4.15)
(4.11) to (4.15)
we form we af
linear combination ofcombination
all independent
linearlinear of allsolutions:
combination of all independent
independent solutions:
solutions: 64
where cos
p is((N
an integer.
+ 1)a + c) So finally
= cos(Na + c). (4.24)
pπ pπ
We find
a =independent , c = solutions
a/2 = from . (4.27)
N +1 2(N + 1)
a−c = c (4.25)
Eq. Normal
(4.16), withmodes a and c from [3/3]Eq. (4.27), decouples the set of differential
(N + 1)a + c = p2π − Na − c, (4.26)
equations. To find the general solution to Eqs. (4.11) to (4.15) we form a
wherecombination
linear p is an integer. So independent
of all finally solutions:
• so set pπ of differential equations pπ is decoupled by
a= , c! N � =
= a/2 "
pπ .� 1 # (4.27)
R = NX+ 12
+ X pπ2(N + 1)
cos 1(n + ) . (4.28)
Rn (t) = X(t) cos
n 0 p (n +
N + 12 ) 2
N + 1
Eq. (4.16), with a and c from Eq. (4.27), decouples the set of differential
p=1
The •factor
to find2To
equations. the general
infind
front solution,
theofgeneral we form
solution
the summation isaEqs.
to linearfor
only combination
(4.11) of of
to (4.15)
reasons all independent
we form a
convenience.
linearsolutions
Making use of by taking
combination p in the rangesolutions:
of all independent p = 0, ..., N
N N " #
1
!"
pπ pπ
1
# 1
Rn = X0 +cos 2 Xp cos (nδ+ ) . (0 ≤ p < 2(N + 1)), (4.28)
!
(n +N + ) 1= p0 2 (4.29)
N + 1 n=0 p=1N +1 2
The
we factor
may 2 inthis
invert front
to of the summation is only for reasons of convenience.
• this may be inverted to
Making use of
N " #
N1 !" pπ# 1
Xp1= ! Rpπn cos 1 (n + ) . (4.30)
N +cos 1 n=0 (n +N +) 1= δp0 2 (0 ≤ p < 2(N + 1)), (4.29)
N + 1 n=0 N +1 2
The equations
we may invert of motion
this to then read
$ %
dXp N 3kB T 2 pπ#
= − 4 pπ
sin Xp + Fp (4.31)
"
1 ! 2 1
Xp = dt ζb
Rn cos (n2(N
+ + ) 1)
. (4.30) 65
Rnn = X00 + 2 Xpp cos N + 1 (n +2 ) . (4.28)
p=1
p=1
N +1 2
The factor 2 in front of the summation is only for reasons of convenience.
The factor 2 in front of the summation is only for reasons of convenience.
Making use of
Rouse modes
Making use of
1 !N "
" pπ
#
1 #
N
1 ! cos pπ (n + 1) = δp0 (0 ≤ p < 2(N + 1)), (4.29)
N + 1 n=0 cos N + 1 (n +2 ) = δp0 (0 ≤ p < 2(N + 1)), (4.29)
N + 1 n=0 N +1 2
we•may
in summary,
invert thistheto equations are decoupled by transforming to
we may invert this to
N " #
1 ! pπ 1 #
Xp = 1
N Rn cos "
pπ (n + )1 . (4.30)
Xp = N + 1 n=0 Rn cos N + 1 (n +2 ) .
!
(4.30)
N + 1 n=0 N +1 2
The equations of motion then read
called the Rouse modes. The $ equations %of motion for the Rouse modes are
The equationsdX of pmotion then
3kB T read2 pπ
= − 2
4 sin $ Xp + Fp
% (4.31)
dt p
dX ζbB T
3k 2 2(N + pπ1)
= − 4 sin Xp + Fp (4.31)
#Fp (t)$
dt = 0 ζb 2 2(N + 1) (4.32)
#F (t)$
! = 02D (4.32)
p
#F0 (t)F0 (t )$ = Īδ(t − t! ) (4.33)
N 2D
+1
#F0 (t)F0 (t! )$ = D Īδ(t − t! )!
!
(4.33)
#Fp (t)Fq (t )$ = N + 1Īδpq δ(t − t ) (p + q > 0) (4.34)
N+ D 1
#Fp (t)Fq (t! )$ = Īδpq δ(t − t! ) (p + q > 0) (4.34)
34 N +1
34 where p, q = 0, ..., N
66
and is therefore itself a stochastic variable, characterised by its first and
second moments, Eqs. (4.32) - (4.34).
4.4Centre-of-mass motion
Rouse relaxation times and amplitudes
Eqs. (4.31) - (4.34) form a decoupled set of 3(N + 1) stochastic differential
• p = 0:each
equations, centre-of-mass motion
of which describes offluctuations
the the polymerand chain
relaxations of a nor-
mal mode (a Rouse� mode) of the Rouse chain. It
N is easy to see that X0 is
1 $
the position
X0 = of the polymerRn centre-of-mass RG = n Rn /(N + 1). The mean
N + 1 n=0
square displacement of the centre-of-mass, gcm (t) can easily be calculated:
% t
X0 (t) = X0 (0) + dτ F0 (τ ) (4.36)
0
(% t % t )
2'
dτ ! F0 (τ ) · F0 (τ ! )
&
gcm (t) = (X0 (t) − X0 (0)) = dτ
0 0
6D
= t ≡ 6DG t. (4.37)
N +1
So the diffusion coefficient DG of the centre-of-mass of the polymer scales
inversely proportional to the weight of the polymer chain. All other modes
the Rouse model predicts that the diffusion coefficient of a (unentangled)
1 ≤ p ≤ N describe independent vibrations of the chain leaving the centre-
polymer in a melt scales inversely with its molecular weight!
of-mass unchanged; Rouse mode Xp descibes vibrations of a wavelength cor-
responding to a subchain of N/p segments. In the applications ahead of us,
this is confirmed in experiments and computer simulations
we will frequently need the time correlation functions of these Rouse modes.
From Eq. (4.31) we get 67
%
SoSothe
the diffusion
diffusion Ncoefficient
= coefficient
+1 t ≡D 6D D
G Gof theofcentre-of-mass
t. G the centre-of-mass of the polymer of the polymer sca
scales
(4.37)
inversely proportionalN + 1to the weight of the polymer chain. All other modes
So the diffusion
nversely coefficientto
proportional DGthe of the centre-of-mass
weight of the of the polymer
polymer chain. scales
All other mo
So the
1≤p≤N
inversely
diffusion coefficient
describe independent
proportional
D
to the weight G of the centre-of-mass
vibrations
of the polymerof the chain of the
chain.leaving
polymer scales
the modes
All other centre-
1 inversely
≤ p ≤ unchanged;
N describe
proportional independent
to the weight of vibrations
the polymer of the
chain. chain
All otherleaving
modes the cent
1 ≤Internal
of-mass
p ≤ N describe motion Rouse
independent of
mode
the
X
polymer
descibes
vibrations
p of the chainchain
vibrations of a wavelength
leaving the centre- cor-
of-mass
of-mass unchanged;
1responding
≤ p ≤unchanged;
N to describe
a subchain Rouse
independent
Rouse of
modeN/p mode X
vibrations
Xsegments.
p descibesp descibes
of the
In
vibrations vibrations
the applications
chainof leaving
a of
the
ahead
wavelength acor-
wavelength
centre-
of us, c
of-mass
esponding unchanged;
totoa a
we will frequently
responding Rouse
subchain
need
subchain theofmode
time
N/p X p descibes
of correlation
N/p
segments.segments. vibrations
functions
In In of a wavelength
of the
the applicationsthese applications
Rouse
ahead modes.cor- ahead of
of us,
responding
From Eq. to a subchain
(4.31) we get oftime
N/pcorrelation
segments.functionsIn the applications ahead of us,
wewe • p = 1, ..., N: independent vibrations of the polymer chain,
will
will frequently
frequently needneed
the the time correlation functions
of these of
Rouse these
modes. Rouse mod
we will frequently need the time correlation functions of these Rouse modes.
From
From wavelength
Eq.
Eq. (4.31)
(4.31) wecorresponding
weget get t to a−(t−τ
subchain of N/p segments
%
From X Eq. (4.31) we get +% t dτ e
p (t) = Xp (0)e
−t/τ p )/τ p
Fp (τ ), (4.38)
0 t % −(t−τ
Xp (t) = Xp (0)e −t/τ +−t/τp % dτ e t−(t−τ)/τ)/τp
Fp (τ ), if p << N (4.38)
−t/τ −(t−τ )/τ
whereXXthe
pp(t)
(t) == (0)e
(0)e
characteristic
X Xp p +
relaxation
p p0
+ dτ e dτ
time τe
p is
p
F (τ
given
p ), Fp (τ ),
pby (4.38) (4.
NAMICS OFcharacteristic
UNENTANGLED 0
POLYMERIC 0+#−1τpLIQUIDS
where the 2
" relaxation
* time is given
2
by 2
ζb
where the characteristic 2 pπ
relaxation time τ p ζbgiven
is (N +by 1) 1
where the τp = characteristic
3k 2
ζbB T "
" 4 sin * relaxation +#−1 ≈
pπ+ 1) +#−1 ζb 3π
time 2 τ2p is 2given
(N k+B 1) T 1p 2
. by (4.39)
2 *2(N
τp = ζb2 4 sin pπ ≈ ζb2 (N + 1) 2 .
1 (4.39)
e equations
τ = 3k itB Tis 2 clear
4"sin 2that 2(N * the
+ lower
1) Rouse
≈+# 3π
−1 2 modes,
k B T p
2
2 which
. represent
2 p% (4.39)
3k ζb
The lastp approximation T is valid
2(N
2 +for pπ
1) large wavelengths,
3π 2 k ζb
T in (N
which
p 2 + 1)
case 1 N.
ith larger
Multiplying
The last =wavelengths,
τp approximation
is the B
(4.38) 4 sin
Eq.characteristic are
isbyvalid
also
Xrelaxation
(0)
for
slower
and
large time
taking modes.
of
wavelengths, ≈
mode
the
B The
p
average
in
relaxation
which over all
case p
time
. N.
possible
% (4.
3k T p
2(N + 1) 3π 2 k T p 2
west
The mode,
τ
last
realisations
Multiplying
1 is p
the =
approximation
of B 1,
relaxation
Eq.the(4.38) is often
random istime referred
valid
by force, of
Xp (0) we the
for to
slowest
large
andfind as the
mode
wavelengths, Rouse
(Rouse
in time
time
which
taking the average over all possible B τ
case
R .
) p % N.
2over all possible
w calculate
Multiplying
realisations
The last ofthe
Eq.the
approximation equilibrium
(4.38)
random &byforce,
X p expectation
(0)we and
find taking
2is valid for large wavelengths,
' values
the of
average X p , i.e.,inthe am- case p %
which
&Xtime
the
realisations p (t)of·correlation
Xthe p (0)' = & functions
random Xforce,
p' exp we (−t/τ
of p ) Rouse
the
find . modes are given by (4.40)
f the normal modes. To this end, first consider the statistical weight
Multiplying
&Xp (t) · Eq. Xp (0)' (4.38)
= &Xp by 2
expX p (0)p ) and
(−t/τ . taking the average(4.40) over all possi
guration&XR 0
(t), .· .X . , R
(0)' N =in XCartesian
2
'
exp (−t/τ coordinates,
p ) . find (4.40)
ealisations p of the p random
!
p force, we # 35
N 35
1 3 2'
& " 2
&X (t) · X (0)' = X exp (−t/τ
R0 , . . . , RNp ) = pexp − 2 p (Rn − Rn−1 )
p ). . (4.41)35 (4.
Z 2b n=1
68
of the these
From slowest mode, pit =
equations is 1, clearis often
that the referred
lower to as the
Rouse modes,Rouse timerepresent
which τR .
2
We now calculate the equilibrium expectation
motions with larger wavelengths, are also slower modes. The relaxation values of X p , i.e., the am-
time
From these
plitudes equations itmodes.
is clearTo that the lower Rouse modes, which represent
of the slowest mode, p = 1, is often referred to as the Rouse time τR . weight
of the normal this end, first consider the statistical
motions with largerRwavelengths, are also slower modes. The relaxation time
of
Rouse mode amplitudes
aWeconfiguration , . . . , R in Cartesian coordinates, 2
now calculate0the equilibrium N expectation values of Xp , i.e., the am-
of the slowest mode, p = 1, is often referred to as the Rouse time τR .
plitudes of the normal modes. To this end, ! N first consider the statistical
#
2 weight
We now calculate the 1equilibrium3 expectation values2 of Xp , i.e., the am-
"
of a configuration
P (R 0 , . . . , R RN0), .=. . , Rexp N in − Cartesian(R coordinates,
− Rn−1 )the. statistical weight
n consider (4.41)
plitudes of the normal modes. Z To this
! 2b Nn=1 2end, first
of a• configuration
statistical weight R0 , . (probability) 3of" a configuration
#
. .1, RN in Cartesian coordinates, 2
P
We can use(R , . . . ,
0 Eq. (4.28) R N ) = to expfind − (R
! the2 statistical n − R
weight
n−1 ) . a configuration
of (4.41) in
Z1 2b3 n=1 N
#
RousePcoordinates.
(R0 , . . . , RN )Since the transformation (Rn −toRthe Rouse. coordinates is a
" 2
= exp − 2 n−1 ) (4.41)
linear
We can transformation
use Eq. (4.28)from Z find
to one set the2bof orthogonal
statistical
n=1 coordinates
weight to another,inthe
of a configuration
Rouse coordinates.
corresponding Jacobian Sinceis the simply transformation
a constant. to The the Rouse coordinates
probability therefore isreadsa
We can use Eq. (4.28) to find the statistical weight of a configuration in
lineartransformation
transformationto fromRouse one coordinates
set
! of orthogonal is a linear
coordinates transformation,
to$another, the
Rouse coordinates. Since the transformation " to
N the Rouse coordinates is%a
#
so Jacobian
corresponding of transformation
Jacobian is1simply a 12 is a constant:
constant. The probability 2therefore pπ reads
linear Ptransformation
(X0 , . . . , XN ) from = one exp set − of2 (Northogonal
+ 1) coordinates
Xp · Xp sin to another, the .
Z
corresponding Jacobian is1 simply a12constant. "
! b
The
Np=1 probability therefore reads
$ 2(N + 1)
% #
2 pπ
P (X0 , . . . , XN ) = exp !− 2 (N + 1) Xp · Xp sin .
(4.42)
Z b N $2(N + 1) % #
1 12 p=1
"
2 pπ
Since Pthis
(X0is, .a. .simple
, XN ) =product
Z
exp of − independent
b2
(N + 1) Gaussians, Xp · Xp sinthe amplitudes .of
2(N + (4.42)
1)
this ismodes
the Rouse a product of independent
can easily be calculated: Gaussians, therefore
p=1
Since this is a simple product 2
of independent Gaussians, 2
the amplitudes of
(4.42)
the Rouse
& 2
'modes can easily b be calculated:(N + 1)b 1
Xp is=a simple product
Since this )≈
( of independent Gaussians,
2 2
. the amplitudes (4.43)
of
8(N + 1) sin 2
2 pπ 2π 2 p
the Rouse
& b
2 modes can easily be calculated:
' 2(N +1) (N + 1)b 1
Xp = ( )≈ 2 2
. (4.43)
8(N + 1) sin 22
pπ 2π p # 2 pN. We can now calculate
Again,& the2
' last approximation b is
2(N +1) valid (Nif+p1)b
when << N1
all kindsXpof = dynamic quantities ( of the ) ≈Rouse chain. 2 2
. (4.43)
Again, the last8(N + 1) sin2 2(N
approximation pπ
is valid when 2πp # N. p We can now calculate 69
+1)
Predictions of the Rouse model;
comparison with atomistically detailed
simulations of a C120 polyethylene melt
70
the orienta- chain under melt con-
polyethylene p
<R(t)R(0
nction of the R = R N − R 0 = 2 X p {(−1)
ditions (symbols), compared with
<R(t)R(0)>/<R >
0.8
2
p=1
a4.5C
the120 H
RouseCorrelation
242 model prediction of(solid 0.6
the end-to-end vector
derline).
melt con-J.T. Padding and W.J. Because the Rouse mode amplitudes de
mpared
TheTime with
first correlation
dynamic quantity
Briels, J. Chem. Phys. 114, 8685 weof
are the
interestedend-to-end
in is the nated
time by vector
p values which
correlation func- are extremely
0.6
tion
diction of the
(solid
(2001). end-to-end vector R. Notice that 0.4 write
0 1000 2000 3000 4000
g and W.J. N * + t [ps] N
"
p pπ " "
s. 114, R8685 = RN − R0 = 2 Xp {(−1) − 1} cos R. = −4 Xp ,
(4.44)
where the prime 0.4 at the p=1 summation sign indicates 2(Nthat + 1)only terms p=1
with odd
p should occur in0the sum. 1000
Then2000 3000 4000
Because the Rouse mode amplitudes t [ps]decay as p−2 , our results will be domi-
N 36 N
nated by p values
he summation which!
sign indicates are extremely
that
! only small
terms with compared
8b odd
2 1.0 to! N.! We 1 −t/τtherefore
write
sum. Then!R(t) · R(0)" = 16 !X p (t) · X p (0)" = 2
(N + 1) 2
e p
. (4.46)
p=1
π p=1
p
N
N " "
2 N
p 8b ! 1 Rouse model
!! R = −4 , (4.45)
<R(t)R(0)>/<R >
The characteristic Xdecay time at
!
large t is τ
−t/τ , which 0.8is proportional to (N +1) 2
.
2
6 !Xp (t) · Xp (0)" = 2 (N + 1) e 1p
. (4.46)
p=1 Figure 4.3 p=1
shows π that Eq. (4.46)
p=1
p 2
gives a good description prediction of the time
correlation function of the end-to-end vector of a real polymer chain in a melt
ay 36
time at large t
(provided the polymeris τ 1 , which is proportional
is not much longer than theto (N +1) 2
.
0.6 entanglement length).
that Eq. (4.46) gives a good description of the time simulation
the end-to-end vector of a real polymer chain in a melt
r is 4.6
not much Segmental
longer than the entanglement motion length).
0.4
0 1000 2000 3000 4000
In this section we will calculate the mean square displacements gseg (t) of the
t [ps]
talindividual
motion segments. Using Eq. (4.28) and the fact that different modes are
not correlated, we get for segment n 71
pπ pπ
a =4.3function
Figure
correlation shows ,c= that
of a/2
theEq. = (4.46) gives
end-to-end . a of
vector gooda realdescription of the
polymer chain (4.27)
in atime
melt
N +1 2(N + 1)
correlation function of the end-to-end
(provided the polymer is not much longer vector than
of a realthepolymer chain in
entanglement a melt
length).
Eq. (4.16),
(provided with a and
the polymer is notc from
muchEq. (4.27),
longer thandecouples the set oflength).
the entanglement differential
Segmental
equations. To find motion
the general [1/2] solution to Eqs. (4.11) to (4.15) we form a
4.6 Segmental
linear combination motion solutions:
of all independent
4.6 Segmental N
motion" #
In this section we will calculate the
! pπmean square 1 displacements gseg (t) of the
Rn = X0 + 2 Xp cos (n + ) . (4.28)
Inindividual
this section we will calculate
segments. Using Eq.the mean
(4.28)
N + 1 andsquare
the
2 displacements
fact that gseg (t)
different of the
modes are
p=1
individual segments.
not correlated, we getUsing Eq. (4.28)
for segment n and the fact that different modes are
not •correlated,
The mean-square
factor 2 we displacement
in front
get forof ofn segment
the" summation
segment n # for reasons of convenience.
is only
" 2 # 2
" (Ruse
Making n (t)of− Rn (0))# =
2 " (X0 (t) − X0 (0)) 2#
(Rn (t) − N
N
Rn (0)) = (X0 (t) −#X0 (0)) $ %
1+4N! "(X (t)pπ
! "
12# 2 pπ 1%
! " cos p − X(n p (0))
+2 # ) cos
=
$
δ pπ
p0
(n
(0 +
≤1 p )< .2(N + 1)), (4.47) (4.29)
N+4 + 1 p=1 (Xp (t)N−+X1p (0)) 2cos 2 N + 1
(n + ) . 2 (4.47)
n=0 N +1 2
p=1
we may invert
Averaging overthisall to
segments, and introducing Eqs. (4.37) and (4.40), the
Averaging overdisplacement
• mean-square
mean square all displacement
segments,of and ofintroducing
a"typicala typical
segment Eqs.
segment (4.37)
in the andmodel
(average
Rouse (4.40),
over is n)
all the
N #
mean square displacement 1 ! of a typical pπ segment 1 in the Rouse model is
Xp = 1 R n cos
N (n + ) #. (4.30)
gseg (t) N= + 11 n=0! N(t)+− 1 R (0)) 22
! "
N (R n n
gseg (t) = N + 1 "
n=0(Rn (t) − Rn (0))
2#
The equations of N motion
+ 1 n=0then N
read
!" #& $ %
=dX6D 4N3kB" TX#p & 21 − e pπ'p .
2 −t/τ
'
p Gt + ! (4.48)
= − 4
2 sin −t/τp X p + F p (4.31)
= dt 6DG t + 4 p=1 ζb Xp 1 − e2(N + .1)
2 (4.48)
#Fp (t)$ = 0p=1 (4.32) 72
37
2D
−20
gseg Rouse, !k exact ditions line
izontal (symbols). The dottedseg-
is the statistical and
10
10 gcm Rouse
dot-dashed
length lines are Padding
Rouse predic-
g(t) [nm ]
b2 . J.T.
2
ment and
10
0
10
1
10
2
10
3
10
4 tions for a chain with an infinite
10 W.J. Briels, J. Chem. Phys. 114,
5
−1 g
number of modes and for a finite
Segmental motion [2/2]
10 t [ps] g seg
cm
8685 (2001).
g Rouse, infinite N
seg Rouse chain, respectively. The hor-
g Rouse, ! exact
−2
seg
g Rouse
k
izontal line is the statistical seg-
Two limits may be distinguished. First, when t is very large,
10 cm
2 t ! τ1 , the first
ment length b 1 . J.T. Padding and
term• in goodEq. (4.48) will
agreement dominate,
with yielding
simulations 10
10 W.J. Briels, J. Chem. Phys. 114,
0 1 2 3 4 5
10 10 10 10 10
g(for displacements
seg (t) ≈ 6DG t t [ps] (t >!segment
τ1 ) . size)
8685 (2001). (4.49)
0
This is consistent with the fact that the polymer as 10
a whole
diffuses with
g(t) [nm ]
2
Two limits
diffusion may be distinguished.
coefficient DG . First, when t is very large, t ! τ1 , the first
termSecondly,
• in
longEq.time
(4.48)
when will
t #dominate,
limit: τ1 the sum yielding
over p in Eq. 10(4.48) ï1 dominates. If g seg
N ! 1 the relaxation times can be approximated by the right hand side of g cm

gseg (t) ≈ 6DG t (t ! τ1 ) . (4.49) g seg Rouse, infinite N
Eq. (4.39), the Rouse mode amplitudes can be approximated by the right g seg Rouse, ok exact
ï2
hand• is
This side of Eq. (4.43),
intermediate times:
consistent with and
the the
fact sum
that can
the be replaced
polymer as by
10 a anwholeintegral,
diffuses with g cm Rouse
diffusion coefficient 2b 2DG . ! ∞

1 " 2
#
−tp /τ1 0 1 2 3 4 5
g
Secondly,
seg (t) = (N + 1) dp 1 − e
when 2 t # τ1 the sum 2 over p in Eq. (4.48) dominates. If 10 10 10 10 10 10
π 0 p t [ps]
N ! 1 the relaxation 2 times can∞ be approximated
t by the right hand side of
2b 1
! !
# be 2 /τ
−t! papproximated
Eq. (4.39), the = Rouse
2
(N mode
+ 1) amplitudes
dp can
dt e 1 by the right
hand side of Eq. (4.43), π and the0 sumτ1can 0 be replaced by an integral,
t
2b22 (N + 1) 1!√∞ # 1
!
= 2b2 πτ1 1 dt √ −tp2 /τ1 #
"
gseg (t) = π 2 (N τ+ 1 1) 2 dp 0 2 1 − te#
$π %1/2
2 ! 0 p
12k
2 BT b ∞1/2 ! t
= 2b t 1 (τN # # t−t# τ11, N ! 1) .
! p2 /τ (4.50)
= 2
(N
πζ+ 1) dp dt e
π 0 τ1 0 73
At short times the mean 2 square displacement
! t of a typical segment is subdif-
where
ations also yieldthe microscopicexpression
a microscopic stress tensor for the is defined
so-calledasshear relaxation
ulus,
" N # N
$
1 # 1 # (i)
σyz (t) =
V mvy,j (t)vz,j (t) + 4. THE DYNAMICS Fy,jOF(t)(z j (t) − zi (t))
UNENTANGLED POLY
Stress
G(t) =
kB T
and viscosity
V yz (0)#
"σyz (t)σ j . 2 i=1 j=1 (2.42)
Eq. (2.40)
roscopically, G(t) isisthe 4.7
the linear Polymer
Green-Kubo
stress relaxation stress
relation forthe
in and
the viscosity
shear
system viscosity. Th
following
• remember
equations the also
Green-Kubo
yield a relations
microscopicfor viscosity
expression & shear for relaxation
the modulus
so-called shear re
dden step %strain. In agreement with Eq. (2.40) its integral
We will finally calculate the viscosity of a melt of Rouse cha is the shear
osity, η =modulus,
∞
G(t)dt. Green-Kubo relation Eq. (2.40). Generalizing Eq. (2.41), the
0
These equations are useful V because
microscopic theytensor
stress enable σ canus beto expressed
measure as the shear
G(t) = "σyz (t)σyz (0)# .
xation modulus and shear kB T viscosity!inN a simulation of a liquid without #
tot N"tot −1 "Ntot
ally shearing the system, but rather1 by analyzing the spontaneous fluc-
"
Macroscopically, G(t) σ =is−the linear Mi V stress
i Vi + relaxation(R ini −the
Rj )system
Fij , f
ions in aforces and velocities. As Van i=1
example, in Fig.
sudden step %strain. In agreement with Eq. (2.40) its integral is th
i=1
2.2
j=i+1
we show the
r relaxation modulus
• microscopic stress measured
∞
tensor
viscosity, η = 0 G(t)dt. is a in a molecular
collective property dynamics simulation of a
of polyethylene chains where
at 450Mi K. is the
Notemass that of inparticle i and Fij isexample
this particular the force tha
• sumsThese equations are useful because they enable us to measure th
stress does notrun over immediately
relax all segments oftoallzero,
exerting on polymer
particle i. butchains
remains in theatsystem
a plateau value
relaxation modulus At and sight,
first shearitviscosity
would be in
a a simulation
tremendous task oftoa calculat
liquid
pproximately
Rouse 3model:
• actually MPa dynamics
for timesofbetween
one chain 5uncorrelated
and 50 ns.with Such
that behaviour
of another is
shearing the system,
analytically becausebuttherather
sums in byEq. analyzing
(4.51) must thebespontaneo
taken ov
cative of temporary elasticity, which is typical of an entangled polymeric
• tuations
we can in forces
study ticles,
stress and
i.e. velocities.
relaxation over
of all
single As an
segments
chain ofexample,
and all
make polymer in chains
ensemble Fig. 2.2 we sy
in the
average s
d. why in real polymer systems
shear relaxation modulus measured in athe stress tensor
molecular is a collective
dynamics simulatp
• melt
overdamped: kinetic
Rouse
of polyethylene contribution
model, to stress
chainshowever,
at 450 thereK. may isbenoneglected
Note correlation
that in thisbetween the dy
particular
chain and the other, so one may just as well analyze the stres
the stress does not relax immediately to zero, but remains at a plate 74
a single chain and make an ensemble average over all initial
analytically because the sums in Eq. (4.51) must be taken over all Ntot par-
ticles, i.e. over all segments of all polymer damped,
chains in the
the polymer
system. stress
This is
is dominat
ticles, i.e. over all segments of all polymer chains in the system. This is
why in real polymer systems the stress tensor segments.
is a The property.
collective first (kinetic)
In part of Eq.
the
why in real polymer systems the stress tensor is a collective property. In the
Rouse model, however, there is no correlation Using Eqs.
between the (4.28) andof(4.51),
dynamics one the m
Rouse model, however, there is no correlation between the dynamics of one
chain Microscopic
and the other, so stress
one may tensor
just as wellof a
chain
analyze Rouse
in the
a specific
chain and the other, so one may just as well analyze the stress relaxation of
chain
stress configuration,
relaxation of neglecti
aa single
single chain
chain andand make
make an an ensemble
ensemble average
average overover all
all initial
initial configurations.
configurations.
N
Moreover, because the velocities of the polymer segments 1are
3kBusually
T "
over- − R ) (R
Moreover, because the velocities of the polymer segments
• the forces are due to the entropic springs connecting the segments:
σ = are usually
2
over-
(R n−1 n
damped,
damped, the the 1polymer
polymer
N stress is
stress is dominated
dominated by by the
the interactionsV b
interactions between
between the
the
� 3k B T (4.51) may then be neglected. n=1
segments.
segments. The first
− first (R
σ = The (kinetic)
n−1 − R
(kinetic) part
part of
n ) of 2
Eq.
Eq. (R n−R
(4.51) may ) be
n−1then neglected.N " N " N
V b 1 48kB T "
Using
Using Eqs.Eqs. (4.28)
(4.28) and
n=1 and (4.51),
(4.51), thethe microscopic
microscopic stressstress tensor of
= tensor of aa Rouse
Rouse Xp Xq
2
V b is n=1
chain
chain in
in a
a specific
specific configuration,
configuration, neglecting
neglecting kinetic
kinetic contributions,
contributions, is equal
equal p=1 toq=1
to
• rewrite in terms of Rouse modes $ % $
qπn qπ
11 3k T
N
N sin sin
3kB BT N +1 2(N + 1)
"
= (R − ) (R − )
"
σ = V b22
σ (Rn−1
n−1 − Rn R n) (Rn−1n−1 − Rn R n)
V b n=1 n=1 1 24kB T "
N $
2
N "
N N "
N NN = 2
N X X
p p sin
V pπ b
$ % $ %
11 48k
48k B
B
T """
T
" $ pπn
pπn
% $ pπ %
p=1
=
= V b22 X X sin sin
Xp Xq sin N + 1 sin 2(N + 1) ×
p q ×
V b n=1 p=1
n=1 p=1 q=1
q=1 N +1 2(N + 1)
$$ qπn % % $ % Combining this with the expression for
qπn
$ qπ
qπ
%
tudes, Eq. (4.43), this can be written mo
sin
sin N + 1 sin sin
N +1 2(N +
2(N + 1)1)
N
N
$ % N
11 24k
24k B
B
T
T
"
2
$
pπ
pπ
% 3kB T " Xp Xp
σ..= .
"
=
= V b22 N N XppX
X sin
Xpp sin 2 '(4.52)
& (4.52)
V b p=1
2(N
2(N +
+ 1)
1) V p=1
Xp 2
p=1 75
To obtain
with the onethe at shear
t = t viscosity
is thereforethe ensemble average must be taken over all
possible configurations at t = 0. Now, since the Rouse modes are Gaussian
variables, all the ensemble
! 2# N #
averages
" N products of an odd number of X ’s are
of
4. THE DYNAMICS OF UNENTANGLED 3k B T Xpx (t)XpyLIQUIDS
POLYMERIC (t)Xqx (0)Xqy (0) p
zero σ xy (t)σ
and (0) =
thexyensemble averages of products of $an even
% $ number of. Xp ’s(4.54)
can
Stress as arelaxation in ofaq=1melt ofofonly Rouse chains
%
V X 2
p X 2
be written sum of products p=1 averages twoq Xp ’s. For the even
The
termcorrelation
in Eq. (4.54) of the xy-component of the microscopic stress tensor at t = 0
we find:
To obtain the shear viscosity the ensemble average must be taken over all
with the one at t = t is therefore
• we
possible need averagesatoft = 0. Now, since the Rouse modes are Gaussian
!Xconfigurations
px (t) Xpy (t) Xqx (0) Xqy (0)" = !Xpx (t) Xpy (t)" !Xqx (0) Xqy (0)"
variables, all the ensemble ! averages"2 # N of Nproducts of an odd number of Xp ’s are
3kB T # + X!X px (t)X
px (t)pyX(t)X qx (0)X
qy (0)" !Xpyqy(t)(0)Xqx (0)"
zero and σxythe
(t)σensemble
xy (0) = averages of products of an $ even % $ number
% of .Xp ’s (4.54)
can
V + !Xpx X 2 2
XqX!X’s.
be written as a sum of products p=1 of averages
q=1 of(t) Xpqxtwo
only (0)"
p
py (t)
For Xthe
qy (0)"
even.
term in Eq. (4.54) we find: (4.55)
To •obtain
Rousethe modesshearare Gaussian
viscosity thevariables:
ensemble average must be taken over all
The !Xfirstpxconfigurations
possible four
(t) Xensemble
py (t) Xqx at
averages
(0)tX=qy0. equal
(0)"Now, = zero
!Xpxbecause,
since the
(t) X Rouse for !Xa Rouse
py (t)"modes qx (0) Xchain
are (0)"in
Gaussian
qy
equilibrium,
variables, allthere is no correlation
the ensemble averagesbetween
of+productsdifferent cartesian components.
!Xpx (t) Xqy (0)" !Xpy (t) XofqxX
of an odd number p ’s are
(0)"
The last
zero andtwo theensemble
ensembleaverages
averages areofnonzero
products only of when
an p = number
even q, since the
of XRouse
p ’s can
modes are mutually orthogonal. Using +the !X
fact (t) X
px that all (0)" !X
qx cartesian (t) X
py components
qy (0)" .
be written as a sum of products of averages of only two Xp ’s. For the even
are equivalent, and Eq. (4.40), the shear relaxation
1 modulus of a (4.55)
melt of
term�X in pxEq. (4.54) we find:
= �Xpx (t)X
Rouse chains (t)Xcanqx (0)�
be expressed as px (0)� δpq = 3 �Xp (t) · Xp (0)� δpq
The first four ensemble averages equal zero because, for a Rouse chain in
!Xpx (t) Xpy (t) X&qx (0) Xqy (0)" '=2 !Xpx (t) Xpy (t)" !Xqx (0) Xqy (0)"
equilibrium, there is# Nno correlation between different N cartesian components.
kB T !Xp (t) · Xp (0)" + !XckpxB(t) T Xqy (0)" !Xpy (t) Xqx (0)"
#
The lastG (t)
two=ensemble averages $ are % nonzero=only when pexp = q, (−2t/τ
sincepthe) , (4.56)
Rouse
V 2 N +(t) 1X
modes are mutuallyp=1 orthogonal.XpUsing the + fact
!X pxthat (0)" !Xpycomponents
allqxcartesian
p=1 (t) Xqy (0)" .
are equivalent, and Eq. (4.40), the shear relaxation modulus of a melt(4.55) of
where c = N/V is the
Rouse chains can be expressed as number density of beads.
76
TheThefirstRouse
four model
ensemblepredicts a viscosity,
averages equal atzero
constant
because, monomer
for a concentration
Rouse chain in
'2 V p=1
X 2
p N + 1 p=1
N
Xp (0)" ckB T #
2 where
% c ==N/V N + is1 the exp (−2t/τ
number p ) , (4.56)
density of beads.
p
Viscosity of a melt of Rouse chains
p=1
The Rouse model predicts a viscosity, at constant monomer concentration
cJ. and
Chem.segmental
Phys., Vol. 114,friction
No. 19, 15 ζ,
Mayproportional
2001 to N: J. T. Padding and W. J. Briel
ensity of beads.
viscosity, at ( constant
∞ monomer concentration
N
ckB T τ1 # 1 ckB T τ1 π 2 cζb2
ortional η =to N:dtG(t) ≈ 2
≈ = (N + 1). (4.57)
0 N + 1 2 p=1 p N +1 2 6 36
N
# 1 ckB T τ1 π 2 cζb2
This 2
≈
has been confirmed = for (N + 1).
concentrated (4.57)
polymers with low molecular weight. 2
p=1
p N +1 2 6 36
Concentrated polymers of high molecular weight give different results, stress-
ing the importance of entanglements. 2
entrated polymers with low molecular
• viscosity scales linearly with molecular weight.
molecular
2
A weight
weight
somewhat of stronger
the
give polymer
different results,
N dependence stress-
is often observed because the density and, more
important, the segmental friction coefficient increase with increasing N .
ments.
nce is•often
in good agreement
observed because with
the experiments
density and, and
more
40 �
fficient simulations
increase
Comparison with of unentangled
increasing
of the short-time N .shear polymer
behavior of the relaxation
melts t
FIG. 6. The long time behavior of the integral of the shear relaxation modu
determined from the atomic "line# and molecular "dotted line# η(t) ≡ dτ G(τ )
lus determined from the molecular stress tensor "line#. The dashed line is th
nsor. The inset shows the integral up to time t. After 1 ps the atomic Rouse prediction, Eq. "A13#, for N!13 and ! 1 !6.0
ecular integrals become identical.
0 ns. The inset shows th
difference ' $ (t) %see Eq. "14#& between measured and Rouse integrals fo
different sets of parameters. From top to bottom: "N!13, ! 1 !6.0 ns#, "N
!14, ! 1 !6.5 ns#, and "N!20, ! 1 !7.0 ns#. For the first two sets, the differ
N to get the best fit for short time behavior. This ence becomes constant after approximately 0.4 ns "dashed lines#. From thi
ure yields ! 1 !(6.0"0.5)ns and N!13"3, in agree- graph we are unable to determine which of the two sets of parameter
provides the best fit. 77
with the fit parameters obtained from the mean square
Ω (R; N) be the probability distribution function for the end-to-end vector
R given that we have a chain of N bonds,
) * N
+,
!
Ω (R; N) = δ R − ri , (4.4)
i=1
where δ isscale
• large the Dirac-delta
propertiesfunction. The are
of polymers central limit theorem
independent then states
of chemistry;
thatend-to-end
for large enough N: distribution is Gaussian
vector
- .3/2 - 2
.
3 3R
Ω (R; N) = exp − . (4.5)
2πNb2 2Nb 2
30• Rouse model = Gaussian chain + Langevin equations for segments;

normal mode analysis reveals characteristic scalings:
2
τ1 ∝ N longest relaxation time
DG ∝ N −1 diffusion of the centre-of-mass

η∝N shear viscosity
in agreement with experiments & simulations of unentangled polymer melts
78
The end
79

Statistical Mechanics of Liquids: Johan T. Padding Université Catholique de Louvain, Belgium

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Statistical Mechanics of Liquids: Johan T. Padding Université Catholique de Louvain, Belgium

Uploaded by

Copyright:

Available Formats

Statistical Mechanics of Liquids

near-spherical molecules liquid with embedded particles long, stringy molecules

• molecular interactions cause correlations

the radial distribution function g(r) by

• g(r) can be obtained from

• modern liquid theory

• experimentally, through microscopy (for large particles) or scattering

• knowing g(r) we can calculate non-entropic thermodynamic functions, e.g.

he first term originates from the internal

• illuminate a fluid containing spherical particles by a wave (wavevector kin )

• X-ray or neutron scattering for atomic liquids

• small-angle X-ray, small-angle neutron or light scattering for colloids

• measured intensity in the direction kout depends on density fluctuations in

N.B. intensity also depends on other factors such as wave polarization,

Density fluctuations in a fluid can be measured experimentally by means

Mean-square displacement and Einstein equation

The last step follows after partial integration.

The is the Green-Kubo relation for the self-di

Note• athat we have included

(relaxing a long wavelength inhomogeneity requires transport of19

All that remains ṙj = isvj towewrite

• Brownian motion is caused by random collisions with solvent molecules

Friction and random forces Consider

Smoluchowski equation [1/2]

The Smoluchowski equation describes how particle distribution functions

• note that Langevin equations neglect

• example: sedimentation of micron-sized colloids

• hydrodynamic interactions may be neglected

• at very low colloid concentration, or

• when other interparticle interactions dominate

• long molecules consisting of

end-to-end vector R is kept fixed, absorbing all constants into a reference

From statics to dynamics 3.1 The equations of motion then read1

Figure 3.1: A colloid

• note that the resulting Rouse model 25

• neglects hydrodynamic interactions (important in dilute polymer solutions)

• neglects entanglements (important for very long polymers)

• applies best to melts of relatively short polymer chains (100 monomers)

(3N+3) coupled stochastic differential equations

a − c = c a − ca − =cc= c (4.25) (4.25)

N ! 1 the relaxation times can be approximated by the right hand side of g cm

diffusion coefficient 2b 2DG . ! ∞

30• Rouse model = Gaussian chain + Langevin equations for segments;

DG ∝ N −1 diffusion of the centre-of-mass

in agreement with experiments & simulations of unentangled polymer melts

You might also like