LecturesintheLattice

BoltzmannMethod

PauloCesarPhilippi
1

Professor
MechanicalEngineeringDepartment
FederalUniversityofSantaCatarina
88040900FlorianopolisSCBrazil

1
philippi@lmpt.ufsc.br
Preface

ThissetoflecturesisintendedtobeafirstcourseintheLatticeBoltzmannMethod(LBM)for
undergraduateandgraduatestudents,whowanttounderstandthetheoreticalfundamentals
behindthemethod.
Historically,thelatticeBoltzmannequation(LBE)hasitsoriginfromlatticegasBooleanmodels
andwasintroducedbyMcNamaraandZanetti[1],in1988,byreplacingtheBooleanvariables
inthediscretecollisionpropagationequationsbytheirensembleaverages.
Higuera and Jimenez [2] proposed a linearization of the term derived from the Boolean
models,recognizingthatthisfullformwasunnecessarilycomplexwhenthemainpurposewas
toretrievethehydrodynamicequations.
Following this line of reasoning, Chen et al. [3] suggested replacing the collision term by a
single relaxationtime term, followed by Qian et al. [4] and Chen et al. [5], who introduced a
model based on the celebrated kinetictheory idea of Bhatnagar, Gross, and Krook (BGK), [6],
but adding rest particles and retrieving the correct incompressible NavierStokes equations,
withthirdordernonphysicaltermsinthelocalspeed,u.
The BGK collision term describes the relaxation of the distribution function to an equilibrium
distribution. This discrete equilibrium distribution was settled by writing it as a secondorder
polynomialexpansionintheparticlevelocity
i
,withparametersadjustedtoretrievethemass
density,thelocalvelocity,andthemomentumfluxequilibriummoments,whicharenecessary
conditionsforsatisfyingtheNavierStokesequations.
Until some years ago LBM was mostly restricted to isothermal flows. Thermal lattice BGK
schemes included higherorder terms in the equilibrium distribution function, requiring to
increase the lattice dimensionality, i.e., the number of vectors in the finite and discrete
velocity set
i
, i= 0, . . . ,b 1. Simulation of the thermal lattice Boltzmann equation was
hamperedbynumericalinstabilitieswhenthelocalvelocityincreases.Thefirstthermallattice
Boltzmann models were introduced in about 1990 and there are several reasons that may be
conjectured for their failure in handling nonisothermal flows. Thermal lattice Boltzmann
models were firstly treated by Alexander et al. [7], who extended the Qian et al. [4] second
order equilibrium distribution to a thirdorder model for solving some thermohydrodynamic
problems,resultinginagoodagreementwhencomparedwithanalyticalsolutions.McNamara
andAlder[8],foundasetof13and26restrictionsthatthisexpansionmustsatisfytoretrieve
the correct advectiondiffusion macroscopic equations, respectively, in two and three
dimensions. Nonlinear deviations in the momentum and energy equations, in the model of
Alexander and coworkers, were found by Chen et al. [9], who introduced a fourth order
polynomial expansion into the equilibrium distribution, fitting adjustable parameters. These
authors used combinations of square lattices for satisfying the restrictions imposed by the
ChapmanEnskoganalysisandfounda16velocitylatticeintwodimensionsandalatticewith
41velocitiesinthreedimensions.
With the exception of McNamara and Zanettis unconditionally stable LBE, [1], all the above
modelshavestabilityissues.
In these studies the equilibrium distribution was written as finite expansions in the local
velocitywith freeparametersthatwereadjustedto satisfysomemainrestrictionstoretrieve
thefulladvectiondiffusionequations.Consequently,thereisnoformallinkconnectingtheLBE
totheBoltzmannequation.
This connection has been first established by He and Luo [10] who directly derived the LBE
fromtheBoltzmannequation forsomewidelyknownlatces D2Q9,D2Q6,D2Q7,D3Q27by
thediscretizationofthevelocityspace,usingtheGaussHermiteandGaussRadauquadrature.
Excludingtheabovementionedlattices,thediscretevelocitysetsobtained
bythiskindofquadraturedonotgenerateregularspacefillinglattices.
Shan et al.[11] and Philippi et al.[12], in 2006, reopened the prospect of using the lattice
Boltzmannmethodtosimulatenonisothermaland/orhighKnudsennumberflowsthroughthe
directresolutionoftheBoltzmannequation,whentheyestablishedasystematiclinkbetween
the Kinetic Theory of Gases and the lattice Boltzmann methods (LBM), determining the
necessary conditions for the discretization of the velocity space in different orders of the
Knudsen number. The lattices obtained through the method proposed by these authors, a
prescribedabscissasquadrature,provedtobestableinflowsoverawiderangeofparameters
by the use of the highorder lattice Boltzmann schemes, leading to velocity sets which, when
used in a discrete velocity kinetic scheme, ensures accurate recovery of the highorder
hydrodynamic moments and assuring increasingly higher order of isotropy of the lattice
tensors.
It is important to stress that the purpose of a continuous kinetic equation is not to solve the
full Boltzmann equation itself, which in most cases is, in fact, unknown, but to consistently
retrieve the macroscopic equations describing the behavior of a physical system. Therefore,
numerically solving a kinetic equation, the framework of LBM, must, firstly, be thought as a
methodforsolvingaphysicalproblem,withsomeadvantageswithrespecttoclassicalCFDdue
toitsintrinsicLagrangiannature.Inthereversesense,duetoitsmesoscopicnature,akinetic
equation enables to reveal, or to put in evidence, the influence of a number of molecular
processesonthemacroscopicbehaviorofaphysicalsystem.
With the advent of ever faster computers, mesoscopic particle models, together with
numerical simulations, provide scientists with a very powerful tool to investigate new
challenges in complex flow problems, both at the fundamental and the applied levels. The
recent years have witnessed many promising advances with the development of new
theoreticalframeworksandapplicationsofthemesoscopicparticlemodels.
Nowadays LBM subjects range from the fundamentals of the kinetic theory and quantum
transport to applied subjects, including some ones such as such as: magnetohydrodynamics;
quantum lattice Boltzmann models; unstructured lattice Boltzmann equation (ULBE); non
Newtonian flows; molecular dynamics and lattice Boltzmann approaches for the analysis of
precursor films; droplet spreading on solid surfaces; biopolymer translocation; microfluidics;
ballistic aggregation and fragmentation; granular flows; combustion; hybrid methods that
combine the lattice Boltzmann model with traditional finite difference techniques; boundary
conditions in the LBM framework; high performance computing with graphic processing unit
(GPU).
References
[1]G.R.McNamaraandG.Zanetti,Phys.Rev.Lett.61,2332,1988.
[2]F.J.HigueraandJ.Jimenez,Europhys.Lett.9,663,1989.
[3]S.Chen,H.Chen,D.Martinez,andW.H.Matthaeus,Phys.Rev.Lett.67,3776,1991.
[4]Y.H.Qian,D.dHumires,andP.Lallemand,Europhys.Lett.17,479,1992.
[5]H.Chen,S.Chen,andW.H.Matthaeus,Phys.Rev.A45,R5339,1992.
[6]P.Bhatnagar,E.Gross,andM.Krook,Phys.Rev.94,511,1954.
[7]F.J.Alexander,S.Chen,andJ.D.Sterling,Phys.Rev.E47,R2249,1993.
[8]G.McNamaraandB.Alder,PhysicaA194,218,1993.
[9]Y.Chen,H.Ohashi,andM.Akiyama,Phys.Rev.E50,2776,1994.
[10]X.HeandL.S.Luo,Phys.Rev.E56,6811,1997.
[11]P.C.Philippi,L.A.HegeleJr.,L.O.E.dosSantos,andR.Surmas,Phys.Rev.E73,56702,2006.
[12]X.Shan,X.F.Yuan,andH.Chen,J.FluidMech.550,413,2006.

Tableofcontents

Lecture01TheBoltzmannequation
Lecture02Theequilibriumsolution
Lecture03LBMdiscretization
Lecture04Boltzmannequationfornonidealfluids
Lecture05Propertiesoftherepulsionterm
Lecture06Macroscopicequations
Lecture07Phasetransitions
Lecture08Velocitydiscretization
Lecture09Attributingvolumetothemolecules
Lecture10LBMvariables

Lecture 1: The Boltzmann equation

Ludwig Boltzmann (1844 1906)
Matter is composed by molecules. The kinetic theory tries to understand the
macroscopic properties of uids from the properties of their molecules: molec-
ular mass, electrical properties, shape parameters, the mean free path and so
on. Although the atomic theory of matter begun in Greece with Leucippus
1
,
Democritus, and Epicurus
2
, the modern kinetic theory was born with the works
of Daniel Bernouilli
3
who theoretically demonstated the Boyles law, 1\ = ctc,
under a constant temperature and showed that the pressure is proportional to
1
Leucippus was one of the earliest Greeks to develop the theory of atomism the idea
that everything is composed entirely of various imperishable, indivisible elements called atoms
which was elaborated in greater detail by his pupil and successor, Democritus. Aristoteles
and Theophrastus explicitly credit Leucippus with the invention of Atomism. Around 440 BC
Leucippus founded a school at Abdera, which his pupil, Democritus, was closely associated
with. His fame was completely overshadowed by that of Democritus, who systematized his
views on atoms. Extracted from Wikipedia.
2
Epicurus (341 BC 270 BC) was an ancient Greek philosopher and the founder of the
school of philosophy called Epicureanism. Only a few fragments and letters remain of Epi-
curuss 300 written works. Much of what is known about Epicurean philosophy derives from
later followers and commentators.
For Epicurus, the purpose of philosophy was to attain the happy, tranquil life, characterized
by ataraxia, peace and freedom from fear, and aponia, the absence of pain, and by living a
self-sucient life surrounded by friends. He taught that pleasure and pain are the measures
of what is good and evil, that death is the end of the body and the soul and should therefore
not be feared, that the gods do not reward or punish humans, that the universe is innite and
eternal, and that events in the world are ultimately based on the motions and interactions of
atoms moving in empty space. Extracted from Wikipedia.
3
Daniel Bernoulli (1700 1782) was a Dutch-Swiss mathematician and was one of the
many prominent mathematicians in the Bernoulli family. He is particularly remembered for
his applications of mathematics to mechanics, especially uid mechanics, and for his pioneering
work in probability and statistics. Bernoullis work is still studied at length by many schools
of science throughout the world. Extracted from Wikipedia.
1
the square of the mean speed of the molecules. John James Waterston
4
showed
that the temperature of a gas is directly related to the kinetic energy of the mole-
cules. Rudolf Clausius
5
introduced the concept of the mean free path. James
Clerk Maxwell
6
introduced the velocity distribution function and established a
transport theory based on the molecular properties of a gas. At the end of 19
||
century, the kinetic theory received a strong impulsion with the work of Ludwig
Boltzmann
7
who establised an evolution equation for the distribution function
and demonstrated the second law of thermodynamics for a system of particles.
In the following the Boltzmann equation is derived.
Let )(

r ,

, t)

be the number of particles with velocities between

and

that are found at time t in the elementary volume

r between

r e

r +

r . Consider now the development of )(

r +

r ,

, t +t)
in a Taylor series around (

r ,

, t),
4
John James Waterston (1811 1883) was a Scottish physicist and a pioneer of the kinetic
theory of gases. A Waterstons paper submitted to the Royal Society was rejected. Some years
after Waterstons death, Lord Rayleigh (Secretary of Royal Society at that time) managed to
dig it out from the archives of the Royal Society. Finally, Watersons paper was published in
the Philosophical Transactions of the Royal Society in 1892. Extracted from Wikipedia.
5
Rudolf Julius Emanuel Clausius (1822 1888), was a German physicist and mathematician
and is one of the central founders of the science of thermodynamics. By his restatement of
Sadi Carnots principle known as the Carnot cycle, he put the theory of heat on a truer and
sounder basis. His most important paper, On the mechanical theory of heat, published in
1850, rst stated the basic ideas of the second law of thermodynamics. In 1865 he introduced
the concept of entropy. Extracted from Wikipedia.
6
James Clerk Maxwell (1831 1879) was a Scottish physicist and mathematician. His most
prominent achievement was formulating classical electromagnetic theory. Maxwell also helped
to develop the MaxwellBoltzmann distribution. Maxwell is considered by many physicists
to be the 19th-century scientist who had the greatest inuence on 20th-century physics. His
contributions to the science are considered by many to be of the same magnitude as those
of Isaac Newton and Albert Einstein. In the millennium polla survey of the 100 most
prominent physicistsMaxwell was voted the third greatest physicist of all time, behind only
Newton and Einstein. On the centennial of Maxwells birthday, Einstein himself described
Maxwells work as the "most profound and the most fruitful that physics has experienced
since the time of Newton." Extracted from Wikipedia.
7
Ludwig Eduard Boltzmann (1844 1906) was an Austrian physicist famous for his founding
contributions in the elds of statistical mechanics and statistical thermodynamics. He was
one of the most important advocates for atomic theory at a time when that scientic model
was still highly controversial. Boltzmanns most important scientic contributions were in
kinetic theory, including the Maxwell-Boltzmann distribution for molecular speeds in a gas. In
addition, Maxwell-Boltzmann statistics and the Boltzmann distribution over energies remain
the foundations of classical statistical mechanics. They are applicable to the many phenomena
that do not require quantum statistics and provide a remarkable insight into the meaning of
temperature. Much of the physics establishment did not share his belief in the reality of
atoms and molecules a belief shared, however, by Maxwell in Scotland and Gibbs in the
United States; and by most chemists since the discoveries of John Dalton in 1808. He had
a long-running dispute with the editor of the preeminent German physics journal of his day,
who refused to let Boltzmann refer to atoms and molecules as anything other than convenient
theoretical constructs. Only a couple of years after Boltzmanns death, Perrins studies of
colloidal suspensions (19081909), based on Einsteins theoretical studies of 1905, conrmed
the values of Avogadros number and Boltzmanns constant, and convinced the world that the
tiny particles really exist. Extracted from Wikipedia.
2
Figure 1: Attraction-repulsion eld around molecules.
)(

r +

r ,

, t + t) = )(

r ,

, t) + (0
|
)) t
+(\
r
))

r +
_
\

x
)
_

+ ..., (1)
or, dividing by t with t 0,
(0
|
)) +

(\
r
)) +

q
_
\

x
)
_
= lim
|!0
)(

r +

r ,

, t + t) )(

r ,

, t)
t,
(2)
where

q =

!

|
is an acceleration.
Therefore, if at time t + t the particles that were in

r , with velocity

are, presently, in

r +

r transported with

and with the velocity

+

due to the acceleration

q , the above expresion is null,
0
|
) +

(\
r
)) +

q (\

)) = 0. (3)
Nevertheless, there are (0
|
))
+
col

t particles in the volume

r , which
acquired the velocity

in the course of the time interval t because they
have collided with other particles and some of the particles in

r , lost, in this
time interval, the velocity

due also to the collisions they suered with other
particles.
So
3
0
|
) +

(\
r
)) +

q (\

)) = (0
|
))
+
col
(0
|
))

col
. (4)
In Boltzmanns model, the particles behave like billard balls, each collision
producing a sharp change in their velocity. A molecule may be thought as a
material point with a force eld around it. This force eld has a strong repulsion
kernel. When two molecules have a frontal collision, they will experience this
strong repulsion when their electronic orbits begin to intercept with themselves.
Around this repulsion kernel, molecules have a soft attraction eld produced by
electrostatic forces (Figure 1) .
Considering o to be a length related to a parameter possible to be identied
with the molecular diameter, the two particles will experience a repulsion force
when their centers are at a distance equal to o (Figure 2). At this distance
the attraction and repulsion elds canceal themselves but the kinetic energy of
the bullet with respect to the target particle is not null and the center of the
bullet particle will penetrate into the o-sphere, where it will be frained along
the radial coordinate.
Figure 2: Two particles will experience a repulsion force when their centers are
at a distance less or equal to o.
This radial frainage will produce a deviation in the bullet trajectory depicted
in Figure (3). Using the label 1 for the bullet particle, the relative velocity of the
bullet with respect to the target before the collision is

q =

when the
target and bullet have the velocities

and

1
, respectively, or,

q
0
=

0
1

0
4
when, as it is the case shown in Figure (3), the target and bullet have the
velocities

0
and

0
1
.
Figure 3: Trajectory deviation of the bullet particle produced by the radial
frainage due to the repulsion forces.
The collision term takes account of these losses and gains in the particle
populations due to radical changes of the velocity

happening in a very small
time interval.
Considering that our reference is on a single target particle with velocity

,
this particle will change its velocity to

0
if it is hit by a bullet particle with
velocity

1
during the time interval t.
Consider a single bullet particle with velocity

1
and
q = [

q [ =

. (5)
Let the .-ordinate of our reference frame be oriented along the direction given
by the vector

q . The bullet particle will hit the target in the time interval t
when: a) the impact factor / is smaller than o (Figure 4); b) it is found in
the elementary volume /d-d/qt in this time interval (Figure 5), where - is the
azimuthal angle.
Inside the volume /d-d/qt there are
5
Figure 4: A bullet particle will hit the target when the impact factor / is smaller
than o
)
_

r ,

1
, t
_
d

1
. .

bullet part. per unit vol..
/d-d/qt
. .
,
volume in cyl. coord..
(6)
bullet particles.
Therefore, since there are )
_

r ,

1
, t
_

r target particles inside

r ,
during the time interval t, the number of target particles that loss the velocity

, during t will be,

(0
|
))

col
td

r
=
__
!
1
_
2t
:=0
_
c
b=0
)
_

r ,

, t
_
)
_

r ,

1
, t
_
/d-d/qd

1
_
t

r .(7)
considering all the possible

1
in the velocity space and all the impact para-
meters / = / (

q ) in all the azimuthal planes -,or

6
Figure 5: Elementary volume in cylindrical coordinates.
(0
|
))

col
=
__
!
1
_
2t
:=0
_
c
b=0
)
_

r ,

, t
_
)
_

r ,

1
, t
_
/d-d/qd

1
_
(8)
In the same way,
(0
|
))
+
col
=
__
!
1
_
2t
:=0
_
c
b=0
)
_

r ,

0
, t
_
)
_

r ,

0
1
, t
_
/d-d/q
0
d

1
_
(9)
where

0
,

0
1
must be such that

0
,

0
1

1
after the collision.It is shown
in Appendix A that q
0
= q. The Appendix also show how to calculate

0
,

0
1
from known vaules of

,

1
.
In its nal form the Boltzmann equation is written as,
0
|
) +

(\
r
)) +

q (\

))
=
_
_
_
!
1
_
2t
:=0
_
c
b=0
_
_
)
_

r ,

0
, t
_
)
_

r ,

0
1
, t
_
)
_

r ,

, t
_
)
_

r ,

1
, t
_
_
_
/d-d/qd

1
_
_
(10)
The acceleration

q is given by

q =

q
(t)
+

q
(|J)
(11)
7
where

q
(t)
is related to the external body forces and

q
(|J)
is the attraction
force per unit mass due to the long-range intermolecular interaction. In the
Boltzmann original model, developed for rareed gases, there is no long-range
action among the molecules,

q
(|J)
= 0, resulting in a model which equation of
state is the one for ideal gases, 1 = T.
In fact, in the classical approach, developed for rareed gases, the long-
distance attraction is included into the collision term, the molecules being free
of any interaction potential in the time interval between any two collisions,
when their centers are distanced by [

r
1

r [ o. In Appendix B it is shown
that, with some simplications, in the framework of the mean eld theory, the
intermolecular long range forces can be considered as in the above treatment.
8
Appendix A: Ballistics of binary collisions
8, 9
The idea is to nd the velocities

0
,

0
1
of, respectively, the target and bullet
particles before the collision that will retrieve

,

1
, after the collision. We
then start from

,

1
and will nd

0
,

0
1
in terms of the rst.
During the interaction, the target particle velocity

v evolves from

to

0
and the bullet particle velocity

v
1
evolves from

1
to

0
1
. Newtons second
law gives
d

r
dt
=

v ,
d

r
1
dt
=

v
1
(12)
and
:
d

v
dt
=
0
0r

r
r
, :
d

v
1
dt
=
0
0r

r
r
(13)
where

r =

r
1
.
From Eq. (13),
d
_

v +

v
1
_
dt
= 0 (14)
which expresses the preservation of momentum during the collision,

1
=

0
+

0
1
(15)
Now, by multiplying these equations by

v ,

v
1
,
d
dt
_
:
2
v
2
+
:
2
v
2
1
_
=
0
0r

r
r

_

v
1
_
=
0
0r

r
r

_
d (

r
1
)
dt
_
=
0
0r

r
r

_
d

r
dt
_
=
0
0

r

d

r
dt
=
d
dt
(16)
or
d
dt
_
:
2
v
2
+
:
2
v
2
1
+
_
= 0 (17)
8
For a complete treatment of the ballistic of binary collisions see C. Cercignani
Mathematical Methods in Kinetic Theory, Springer (1995)
9
Although important for understanding the collision details, this appendix is not a
main requisite for the remaining lectures and may be avoided by the students who are
mainly interested in learning LBM. The same comments for the Appendix B and C.
9
which expresses the conservation of the total energy inside a sphere of radius o.
Outside this sphere = 0 and

2
+
2
1
=
02
+
02
1
(18)
Let

c be the unit vector in the direction of

, Figure (),

= C

c (19)
or

0
=

+ C

c (20)
or

0
1
=

1
C

c (21)
Using these relations into Eq.(18)

2
+
2
1
=
2
+ 2C

c + C
2
+
2
1
2C

c + C
2
(22)
or
_

_
. .

c = C (23)
Therefore

0
=

+ (

c)

c (24)

0
1
=

1
(

c)

c (25)
Subtracting the two above equations

q
0
=

q 2 (

c)

c (26)
or
(

q
0
)
2
= (

q )
2
4 (

c)
2
+ 4 (

q

c)
2
= (

q )
2
(27)
meaning that the collision does not aect the norm of the relative velocity

q .
From Eq. (26)

q
0

q
. .

2
cos
=

q
. .

2
2(

c)
. .
cos 0

c (28)
or
cos c = cos 20 (29)
10
or
c = 20 (30)
The collision plane is given by -. So, considering the vector

q pointing
along the direction r, the direction given by the unitary vector

c will be given
by its coordinates along the axis r, j and ., Figure (3)
(cos -, sin0 cos -, sin0 sin-) (31)
Under a central eld the angular momentum is also preserved (Figure )
q
|
j = j
d
dt
j = j
2
d
dt
= ctc = q/ (32)
or
dt =
j
2
d
q/
(33)
Radial and tangential components of the relative velocity.
And from the energy conservation, Eq. (17), in the center of mass system
(see Remark 1),
:
4
_

v
1

v
_
2
+
=
:
4
_
.
j
2
+ j
2
.

2
_
+
=
:
4
q
2
(34)
11
The two above equations may be rewritten as
:
4
_
_
dj
dt
_
2
+ j
2
_
d
dt
_
2
_
+ =
:
4
q
2
(35)
or using dt =

2
J
b
,
:
4
_
_
q/
j
2
dj
d
_
2
+ j
2
_
q/
j
2
_
2
_
+ =
:
4
q
2
(36)
:
4
_
q
2
/
2
j
4
_
dj
d
_
2
+
q
2
/
2
j
2
_
+
:
4
q
2
= 0 (37)
dividing by

2
b
2

2
:
4
_
1
j
2
_
dj
d
_
2
+ 1
_
+
j
2
q
2
/
2
_

:
4
q
2
_
= 0 (38)
or
:
4
_
1
j
2
_
dj
d
_
2
_
+
:
4
+
j
2
q
2
/
2
_

:
4
q
2
_
= 0 (39)
or
_
j
2
q
2
/
2
_
:
4
q
2

:
4
_
j
2
_
d
dj
_
2
=
:
4
(40)
or
d
dj
=
q/:
1/2
j
2

_
1
4
_
n
4
q
2
_
1
_
b

_
2
_
(j)
_ (41)
Integrating this equation between the point where the bullet particle reaches
the sphere o, when = arcsin
_
b
c
_
and the point of closest approach, where
j = j
0
,
0 =
q/:
1/2
2
_

0
c
1
j
2

_
1
_
n
4
q
2
_
1
_
b

_
2
_
(j)
_dj + arcsin
_
/
o
_
(42)
where j
0
is the solution of
J
J
= 0, i.e., the solution of
:
4
q
2
_
1
_
/
j
_
2
_
(j) = 0 (43)
12
Therefore, for each

, the right hand side of the Boltzmann equation is
evaluated in accordance with the following steps:
i) / is an integration variable and so

1
. Then

q =

and Eq. (42)

can be used for nding the polar angle 0;
ii) The azimuthal angle - is also an integration variable and Eq. (31) can be
used for nding

c = (cos -, sin0 cos -, sin0 sin-) (44)

iii) the particle velocities before collision that will give

and

1
after
collision, are Eqs. (24) and (25)

0
_
/, -,

1
_
=

+ cos 0

c (45)

0
1
_
/, -,

1
_
=

1
cos 0

c (46)
iv) The distribution ) can be used for evaluating )
_

r ,

0
, t
_
)
_

r ,

0
1
, t
_
related to

and to the corresponding values of the integration variables /, -,

1
.
Remark
(:
1
+ :
2
) v
0
= :
1
v
1
+ :
2
v
2
q = v
2
v
1
(47)
Therefore when the masses are identical,
v
1
=
2v
0
+ q
2
v
2
=
2v
0
q
2
(48)
so
1
2
:v
2
1
+
1
2
:v
2
2
=
:
8
_
4v
2
0
+ q
2
+ 4v
0
q
_
+
:
8
_
4v
2
0
+ q
2
4v
0
q
_
=
:
4
_
4v
2
0
+ q
2
_
(49)
13
Appendix B: The Boltzmann equation derived
from the Liouvillle equation
In a liquid, the mean free path has the same order of magnitude than the
molecular diameter, multiple collisions are frequent and long-range interactions
are important. We begin with the Liouville equation.
Consider a mechanical system of N particles. Let
)
N
_

r
1
,

1
, ...

r

,

, t
_
, (50)
to be the joint probability of nding, at time t, dt the particle 1 at the position

r
1
, d

r
1
with velocity

1
, d

1
, the particle 2 at the position

r
2
, d

r
2
with
velocity

2
, d

2
and so on, until particle at the position

r

, d

r

with
velocity

, d

. The Liouville equation describing the dynamical evolution
of this system is given by
0
|
)

I
.0 !
r i
)
N
+

q
i
.0 !
i
)
N
= 0, (51)
where

q
i
is the acceleration due to the force acting on particle i,

q
i
=

q
t
i
+

=1
6=I

q
ij
. (52)
Force

q
t
i
is related to the force on particle i due to an external eld and

q
ij
is the force on particle i due to its interaction with particle ,,

q
ij
=
1
:
0(r
I
)
0

r
I
, (53)
where r
I
= [

r
I

r

[ and is the potential energy depending, only, on the
distance between particles i and ,.
Joint probability )

can be integrated in the phase space

r
2
,

2
, ...

r

,

to give the marginal probability )
1
of nding, at time t, dt the particle 1 at
the position

r
1
, d

r
1
with velocity

1
, d

1
)
1
_

r
1
,

1
, t
_
=
_
...
_
)

r
2
...d

r

d

2
...d

, (54)
considering that the probability )

gives a too detailed description of the sys-

tem, which is unnecessarily complex, since the dynamical evolution of an arbi-
trary, but, single particle, can be a reliable description of the whole mechanical
system of particles, when these particles cannot be individually labelled.
Integration of Eq. (51) gives for the temporal derivative
_
...
_
0
|
)

r
2
...d

r

d

2
...d

= 0
|
)
1
. (55)
14
Consider now the integral
_
...
_

I

I
.0 !
r i
)
N
d

r
2
...d

r

d

2
...d

=
_
...
_

1
.0 !
r 1
)
N
d

r
2
...d

r

d

2
...d

+
_
...
_

I=2

I
.0 !
r i
)
N
d

r
2
...d

r

d

2
...d

1
.0 !
r 1
)
1
+
_
...
_

I=2

I
.0 !
r i
)
N
d

r
2
...d

r

d

2
...d

1
.0 !
r 1
)
1
, (56)
because, each term

I
.0 !
r i
)
N
= 0 !
r i
_

I
)
N
_
, (57)
for i = 2, ..., :, since

I
and

r
I
are independent variables. In this manner,
when this term is integrated in

r
I
it gives i) the ux of )
N
outside the box
where the mechanical system is contained, which must be null or ii) something
proportional to )
N
_

r
1
,

1
, .. [

r
I
[ ....

r

,

, t
_
when there is no box
enclosing the system, which must be also null.
Proceeding in a similar term with the force term,
_
...
_

I

q
i
.0 !
i
)
N
d

r
2
...d

r

d

2
...d

=
_
...
_

I

q
t
I
.0 !
i
)
N
d

r
2
...d

r

d

2
...d

+
_
...
_

I,

q
i;j
.0 !
i
)
N
d

r
2
...d

r

d

2
...d

So
15
_
...
_

I

q
i
.0 !
i
)
N
d

r
2
...d

r

d

2
...d

=
_
...
_

q
t
1
.0 !
1
)
N
d

r
2
...d

r

d

2
...d

. .
=
!

e
1
.J !

1
}1
+

I=2
_
...
_

q
t
I
.0 !
i
)
N
d

r
2
...d

r

d

2
...d

. .
=0
+

,=2
_
...
_

q
1;j
.0 !
1
)
N
d

r
2
...d

r

d

2
...d

I=2

_
...
_

q
i;j
.0 !
i
)
N
d

r
2
...d

r

d

2
...d

. .
=0
. (58)
The term

,=2
_
...
_

q
1;j
.0 !
1
)
N
d

r
2
...d

r

d

2
...d

=2
0 !
1
_
...
_

q
1;j
)
2
_

r
1
,

1
,

r
2
,

2
, t
_
d

r
2
d

2
=
..
indistinguishable particles
( 1)
0 !
1
_
...
_

q
12
)
2
_

r
1
,

1
,

r
2
,

2
, t
_
d

r
2
d

2
(59)
After integration, considering ) = )
1
and changing the notation for the
target and the incident particles, the Liouville equation becomes, for large N,
0
|
) +

.0 !
r
) +

q
t
.0 !

)
= 0 !

_
...
_

q
12;j
)
2
_

r ,

r
1
,

1
, t
_
d

r
1
d

1
= 0 !

_
...
_
1
:
0([

r
1

r [)
0

r
)
2
_

r ,

r
1
,

1
, t
_
d

r
1
d

1
, (60)
which is a Boltzmann equation for the distribution function ), with a collision
term .
This collision term will be split in two collision terms
10
=
sJ
+
lJ
, where

sJ
is referred to short distance interactions, [

r
1

r [ < o and
lJ
to long
10
X. He and G. D. Doolen, J. Stat. Phys., 107, 1-2 (2002).
16
range interactions [

r
1

r [ o, o being the distance related to the molecular

diameter below which the attraction forces change to a strong repulsion among
the molecules.
Therefore
0
|
) +

.0 !
r
) +

q
t
.0 !

) =
sJ
+
lJ
, (61)
where

sJ
= 0 !

_ _
[
!
r 1
!
r [<c
1
:
0([

r
1

r [)
0

r
)
2
_

r ,

r
1
,

1
, t
_
d

r
1
d

1
,
(62)
and

|J
= 0 !

_ _
[
!
r 1
!
r [,c
1
:
0([

r
1

r [)
0

r
)
2
_

r ,

r
1
,

1
, t
_
d

r
1
d

1
.
(63)
Long-range term. Mean eld theory.
Consider, rst, the long-range collision term

lJ
=
0
0

.
_ _
[
!
r 1
!
r [,c
1
:
0([

r
1

r [)
0 (

r )
)
2
_

r ,

r
1
,

1
, t
_
d

r
1
d

1
.
(64)
By making the assumption that, for [

r
1

r [ o, the molecular chaos

prevails
)
2
_

r ,

r
1
,

1
_
= )
_

r ,

, t
_
)
_

r
1
,

1
, t
_
= ))
1
, (65)
one obtains

lJ
=
0)
_

r ,

, t
_
0

_ _
[
!
r 1
!
r [,c
1
:
0([

r
1

r [)
0

r
)
_

r
1
,

1
, t
_
d

r
1
d

1
=
0)
_

r ,

, t
_
0

.
_
[
!
r 1
!
r [,c
1
:
0([

r
1

r [)
0

r
:(

r
1
, t) d

r
1
=
1
:
0)
_

r ,

, t
_
0

.
0
0

r
_
[
!
r 1
!
r [,c
([

r
1

r [) :(

r
1
, t) d

r
1
. (66)
17
The integrand in the above equation is the mean eld, i.e., the potential
energy related to the integrated action of each of the : molecules placed in the
neighborhood, on a single molecule at the position

r ,

n
(

r ) =
_
[
!
r 1
!
r [,c
([

r
1

r [) :(

r
1
, t) d

r
1
. (67)
Since the acceleration

q
(|J)
is the gradient of the potential energy divided
by its mass,

q
(|J)
=
1
:
0
n
0

r
, (68)
we get

lJ
=

q
(|J)

0)
_

r ,

, t
_
0

, (69)
giving the same long-range term as previously obtained.
18
Appendix C: Short-range collision term derived
from Eq. (62)
11
Using the results of Appendix B, the integration of the Liouville equation
gives, after dropping out the subscript index for the target particles and using
the subscript 1 for the integration variables,
0
|
)
(1)
+

.0 !
r
)
(1)
+
_

q
(t)
+

q
(|J)
_
.0 !

)
(1)
= ( 1)
0 !

_ _
[
!
r 1
!
r [ <c

q !
r
!
r 1
.)
(2)
_

r ,

r
1
,

1
, t
_
d

r
1
d

1
. (70)
Similarly
0
|
)
(2)
+

.0 !
r
)
(2)
+

1
.0 !
r
)
(2)
+
_

q
(t)
+

q
(|J)
+

q !
r
!
r 1
_
.0 !

)
(2)
+
_

q
t
1
+

q !
r 1
!
r
_
.0 !
1
)
(2)
= ( 2)
0 !

_ _

q !
r
!
r 2
)
(3)
_

r ,

r
1
,

1
,

r
2
,

2
, t
_
d

r
2
d

2
( 2)
0 !
1

_ _

q !
r 1
!
r 2
)
(3)
_

r ,

r
1
,

1
,

r
2
,

2
, t
_
d

r
2
d

2
. (71)
The above equations and the related equations for )
(3)
, )
(4)
,... form the
BBGKY hierarchie ([1]-[7]) where the solution of each distribution )
(I)
depends
on the knowledge of )
(I+1)
.
For evaluating the short-distance collision term, consider the following as-
sumptions.
Assumption 1: Only binary collisions are considered
With this assumption Eq. (71) becomes:
0
|
)
(2)
+

.0 !
r
)
(2)
+

1
.0 !
r
)
(2)
+
_

q
(t)
+

q
(|J)
+

q !
r
!
r 1
_
.0 !

)
(2)
(72)
+
_

q
t
1
+

q !
r
!
r 1
_
.0 !
1
)
(2)
= 0 (73)
11
A key reference for the treatment given in this appendix is the Chapter 2 of Gilberto
Medeiros Kremer, An Introduction to the Boltzmann Equation and Transport Processes in
Gases, Springer (2010)
19
Assumption 2: The external and attractive forces are very small when com-
pared with the intermolecular forces during collision
0
|
)
(2)
+

.0 !
r
)
(2)
+

1
.0 !
r
)
(2)
+
_

q !
r
!
r 1
_
.0 !

)
(2)
+
_

q !
r 1
!
r
_
.0 !
1
)
(2)
= 0 (74)
or, equivalently, considering the time t to be the instant immediately after the
collision
)
(2)
_

r ,

r
1
,

1
, t
_
= )
(2)
_

t,

q !
r
t,

r
1

1
t,

q !
r 1
t, t t
_
(75)
where t is a time interval that is small with respect to the time interval between
any two collisions but large with respect to the collision time.
Assumption 3: Molecular chaos
)
(2)
_

r ,

r
1
,

1
, t
_
= )
(2)
_

t,

q !
r
t,

r
1

1
t,

q !
r 1
t, t t
_
= )
(1)
_

t,

q !
r
t, t t
_
)
(1)
_

r
1

1
t,

q !
r 1
t, t + t
_
(76)
where 0 < : < t.
Assumption 4:It is supposed that the distribution function does not vary be-
tween points

r and

r
1
at the begining of the collision process.
Consider now the time average of Eq. (70) given
)
(1)
=
1
t
_
|
0
)
(1)
d: (77)
0
|
)
(1)
+

.0 !
r
)
(1)
+
_

q
(t)
+

q
(|J)
_
.0 !

)
(1)
=
( 1)
t
_
|
0
_ _
[
!
r 1
!
r [ <c

q !
r
0 !

)
(2)
_

r +

(: t) ,

q !
r
(: t) ,

r
1
+

1
(: t) ,

1
+

q !
r 1
(: t) , t t + :
_
d

r
1
d

1
d:
20
or
0
|
)
(1)
+

.0 !
r
)
(1)
+
_

q
(t)
+

q
(|J)
_
.0 !

)
(1)
=
( 1)
t
_
|
0
_ _
[
!
r 1
!
r [ <c

q !
r
0 !

_
)
(1)
_

r +

(: t) ,

q (: t) , t + (: t)
_
)
(1)
_

r
1
+

1
(: t) ,

1
+

q
1
(: t) , t + (: t)
_
_

_
d

r
1
d

1
d:
or
0
|
)
(1)
+

.0 !
r
)
(1)
+
_

q
(t)
+

q
(|J)
_
.0 !

)
(1)
=
( 1)
t

_ _
_

_
_
|
0
0
s
_

_
)
(1)
_

r +

(: t) ,

q (: t) , t + (: t)
_
)
(1)
_

r
1
+

1
(: t) ,

1
+

q
1
(: t) , t + (: t)
_
_

_
_

_
d

r
1
d

1
d:
or
0
|
)
(1)
+

.0 !
r
)
(1)
+
_

q
(t)
+

q
(|J)
_
0 !

)
(1)
=
( 1)
t

_ _
_

_
_
|
0
0
s
_

_
)
(1)
_

r +

(: t) ,

q (: t) , t + (: t)
_
)
(1)
_

r +

1
(: t) ,

1
+

q
1
(: t) , t + (: t)
_
_

_
_

_
d

r
1
d

1
d:
=
( 1)
t

_ _
_

_
_
_
)
(1)
_

t,

q t, t t
_
)
(1)
_

1
t,

q
1
t, t t
_
_
_

_
)
(1)
_

r ,

, t
_
)
(1)
_

r ,

1
, t
__
_

_
d

r
1
d

1
21
Since t is very small

t -

1
t -

r (78)
t t - t (79)
the collision process being considered to only modify the velocities of the target
and bullet particles, respectively,

q t =

0
,before the collision to

,
after the collision and

q
1
t =

0
1
before the collision to

1
, after the
collision.
Therefore, we can rewrite Eq. (??) as
0
|
)
(1)
+

.0 !
r
)
(1)
+
_

q
(t)
+

q
(|J)
_
.0 !

)
(1)
=
( 1)
t
_ _
_
_
)
(1)
_

r ,

0
, t
_
)
(1)
_

r ,

0
1
, t
_
)
(1)
_

r ,

, t
_
)
(1)
_

r ,

1
, t
_
_
_
d

r
1
d

1
(80)
where

0
,

0
1
are the velocities of the target and bullet particles before the
colision (at time t t) that will give the velocities

,

1
after the collision.
When evaluating the right hand side of the above equation, the position of
the target particle

r is constant. Therefore when evaluating the integral we
can replace

r
1
by

r
1

r , by placing the system of coordinates on the center

of the target particle.
Writing the element of volume d

j = d (

r
1

r ) in cylindrical coordinates,
1
t
d

j =
dj
t
/d/d- = q/d/d- (81)
where

q is the velocity of the bullet particle with respect to the target particle.
When is very large 1 - and, considering
)
(1)
_

r ,

, t
_
= )
_

r ,

, t
_
(82)
as the number of particles that during the times between t and t+dt are expected
to be in the position between

r and

r + d

r with velocity between

and

+
d

,
0
|
) +

.0 !
r
) +
_

q
(t)
+

q
(|J)
_
.0 !

)
=
_ _ _
_
_
)
_

r ,

0
, t
_
)
_

r ,

0
1
, t
_
)
_

r ,

, t
_
)
_

r ,

1
, t
_
_
_
q/d/d-d

1
(83)
22
References
[1] N. N. Bogoliubov (1946). "Kinetic Equations" (in English). Journal of
Physics USSR 10 (3): 265274.
[2] N. N. Bogoliubov (1946). "Kinetic Equations" (in Russian). Journal of Ex-
perimental and Theoretical Physics 16 (8): 691702.
[3] N. N. Bogoliubov, K. P. Gurov (1947). "Kinetic Equations in Quantum
Mechanics" (in Russian). Journal of Experimental and Theoretical Physics
17 (7): 614628.
[4] J. Yvon (1935): Theorie Statistique des Fluides et lEquation et lEquation
dEtat (in French), Actes sientique et industrie. 203. Paris: Hermann.
[5] John G. Kirkwood (March 1946). "The Statistical Mechanical Theory of
Transport Processes I. General Theory". The Journal of Chemical Physics
14 (3): 180. http://dx.doi.org/10.1063/1.1724117.
[6] John G. Kirkwood (January 1947). "The Statistical Mechanical Theory
of Transport Processes II. Transport in Gases". The Journal of Chemical
Physics 15 (1): 72. http://dx.doi.org/10.1063/1.1746292.
[7] M. Born and H. S. Green (31 December 1946). "A General Kinetic Theory
of Liquids I. The Molecular Distribution Functions". Proc. Roy. Soc. A 188:
1018.
23
Lecture 2: The equilibrium solution
The Boltzmann equation
@
|
f +
!
(r
r
f) +
_
!
g
(t)
+
!
g
(|J)
_
(r

f) = ; (1)
is an integro-diferential equation and the term,
==
_
_
_
!
1
_
2t
:=0
_
c
b=0
_
_
f
_
!
x ;
!

0
; t
_
f
_
!
x ;
!

0
1
; t
_
f
_
!
x ;
!
; t
_
f
_
!
x ;
!

1
; t
_
_
_
bd"dbgd
!

1
_
_
; (2)
gives the net balance between the molecules that acquire the velocity
!
and
the ones that loss this velocity. In equilibrum conditions this balance must be
null, otherwise the distribution f would vary in the course of the time and in
the space. So,
f
_
!
x ;
!

0
; t
_
f
_
!
x ;
!

0
1
; t
_
= f
_
!
x ;
!
; t
_
f
_
!
x ;
!

1
; t
_
; (3)
or
lnf
_
!
x ;
!

0
; t
_
+ lnf
_
!
x ;
!

0
1
; t
_
= lnf
_
!
x ;
!
; t
_
+ lnf
_
!
x ;
!

1
; t
_
; (4)
meaning that
lnf = lnf
(tj)
= collisional invariant. (5)
The collisional invariants are
m = mass, (6)
m
!
= momentum, (7)
1
2
m
2
= kinetic energy. (8)
Therefore, any linear combination of these invariants is also a collisional
invariant,
lnf
(tj)
= A +
!
B
!
+ C
2
= a + b
_
!

!
c
_
2
; (9)
or
1
Figure 1: Maxwell-Boltzmann distributions for dierent values of a =
_
|T
n
f
(tj)
= e
o
e
b

!

!
c

2
= de
b

!

!
c

2
: (10)
Parameters d, b e
!
c are determined from
n =
_
f
(tj)
d
!
= number density of molecules, (11)
!
u =
1
n
_
f
(tj)
!
d
!
= mean molecular velocity, (12)
e
c,}
=
1
n
_
f
(tj)
1
2
m
_
!

!
u
_
2
d
!
= mean peculiar kinetic energy.(13)
In fact, for a uid in equilibrium the quantities
n =
_
fd
!
; (14)
!
u =
_
!

_
=
1
n
_
f
!
d
!
; (15)
must be constants and given by replacing the distribution f by the distribution
the system has at equilibrium, i.e., f
(tj)
.
The third integral deserves a more lengthy discussion. The mean kinetic
energy per molecule is given by
e
c
=
1
2
m

2
_
=
1
n
_
f
1
2
m
2
d
!
: (16)
Writting,

2
=
_
!

!
u
_
2
+ 2
_
!

!
u
_

!
u + u
2
; (17)
2
it can be easily seen that this kinetic energy has two components
e
c
_
J
molecule
_
=
1
n
_
f
1
2
m
_
!

!
u
_
2
d
!

. .
=t
c;f
+
1
2
mu
2
. .
advection energy
: (18)
The thermodynamic internal energy, e, per molecule
1
, is the sum of the the
peculiar kinetic energy e
c,}
, i.e., the energy due to the molecular random motion
of the molecules and the potential energy due to the intermolecular interaction
among these molecules,
e = e
c,}
+

n
; (20)
where is the intermolecular potential energy per unit volume.
For thermodynamic systems in equilibrium, the thermodynamic internal en-
ergy, e
1
, per mol, satises,
de
1
= c
u
dT
. .
energy due to molecular motion
+
_
T
_
@P
@T
_
u
P
_
dv
. .
intermolecular potential energy
: (21)
For ideal gases, with R=universal gas constant,
P =
RT
v
; (22)
the second term is null and,
de
1
= c
u
dT; (23)
meaning that solely the molecular motion contributes to the thermodynamic
energy. Nevertheless, for real uids, the thermodynamic energy must take the
intermolecular potential energy into account. For a uid with, e.g., a van der
Waals equation of state
P =
RT
v b

a
v
2
; (24)
the parameter "a" is related to the intermolecular forces and we get
1
We use a bar over the symbol e, e.g., e; ec; e
c;f
for denoting energy per unit molecule,
e =
e
m
; e =
ec
m
; e
c;f
=
e
c;f
m
for denoting energy per unit mass and e
M
; e
c;f;M
for denoting
energy per unit mol. Therefore e = ne is the thermodynamic internal energy per unit volume
and
e
c;f
= ne
c;f
=
Z
f
1
2
m

!

!
u

2
d
!
(19)
is the peculiar kinetic energy per unit volume.
3
de
1
= c
u
dT +
a
v
2
dv: (25)
In both cases, the molal heat c
u
is a molecular parameter related to the
motion degrees of freedom of the molecules
2
,
c
u
= c
u,|:onsl
+ c
u,:o|
+ c
u,uIb
: (26)
For molecules with solely translational degrees of freedom
3
(our case)
c
u
= c
u,|:onsl
=
D
2
R; (27)
where D = space dimension.
2
With monatomic gases, thermal energy comprises only translational motions. Transla-
tional motions are ordinary, whole-body movements in 3D space whereby particles move about
and exchange energy in collisionslike rubber balls in a vigorously shaken container (see an-
imation here). These simple movements in the three X, Y, and Zaxis dimensions of space
means individual atoms have three translational degrees of freedom. A degree of freedom is
any form of energy in which heat transferred into an object can be stored. This can be in
translational kinetic energy, rotational kinetic energy, or other forms such as potential energy
in vibrational modes. Only three translational degrees of freedom (corresponding to the three
independent directions in space) are available for any individual atom, whether it is free, as a
monatomic molecule, or bound into a polyatomic molecule.
As to rotation about an atoms axis (again, whether the atom is bound or free), its energy
of rotation is proportional to the moment of inertia for the atom, which is extremely small
compared to moments of inertia of collections of atoms. This is because almost all of the mass
of a single atom is concentrated in its nucleus, which has a radius too small to give a signicant
moment of inertia. In contrast, the spacing of quantum energy levels for a rotating object
is inversely proportional to its moment of inertia, and so this spacing becomes very large for
objects with very small moments of inertia. For these reasons, the contribution from rotation
of atoms on their axes is essentially zero in monatomic gases, because the energy-spacing of the
associated quantum levels is too large for signicant thermal energy to be stored in rotation
of systems such small moments of inertia. For similar reasons, axial rotation around bonds
joining atoms in diatomic gases (or along the linear axis in a linear molecule of any length)
can also be neglected as a possible "degree of freedom" as well, since such rotation is similar
to rotation of monatomic atoms, and so occurs about an axis with a moment of inertia too
small to be able to store signicant heat energy.
In polyatomic molecules, other rotational modes may become active, due to the much higher
moments of inertia about certain axes which do not coincide with the linear axis of a linear
molecule. These modes take the place of some translational degrees of freedom for individual
atoms, since the atoms are moving in 3-D space, as the molecule rotates. The narrowing
of quantum mechanically-determined energy spacing between rotational states results from
situations where atoms are rotating around an axis that does not connect them, and thus
form an assembly that has a large moment of inertia. This small dierence between energy
states allows the kinetic energy of this type of rotational motion to store heat energy at
ambient temperatures. Furthermore (although usually at higher temperatures than are able
to store heat in rotational motion) internal vibrational degrees of freedom also may become
active (these are also a type of translation, as seen from the view of each atom). In summary,
molecules are complex objects with a population of atoms that may move about within the
molecule in a number of dierent ways (see animation at right), and each of these ways
of moving is capable of storing energy if the temperature is sucient. (from Wikipedia:
http://en.wikipedia.org/wiki/Heat_capacity)
3
See http://en.wikipedia.org/wiki/Degrees_of_freedom_(physics_and_chemistry)
4
Therefore, by integrating Eq. (25), we get for a van der Waals equation of
state,
e
1
(T; v) =
D
2
RT +
_
a
1
v
_
(in J=mol), (28)
plus a constant C (T
0
; v
0
) that depends only on the initial state.
This means that, in a molecular basis,
e
c,}
=
1
n
_
f
(tj)
1
2
m
_
!

!
u
_
2
d
!
=
D
2
kT (in J=molecule), (29)
where,
k =
R
N
; (30)
N being the Avogadro
4
number and k = Boltzmann constant, Also, for a van
der Waals equation of state,
e (T; n) =
D
2
kT + (an) (in J=molecule). (31)
Eq. (29) is, in fact, the denition of temperature for systems composed of
molecules with only translational degrees of freedom.
The intermolecular potential energy per unit volume becomes for a van der
Waal equation of state, = an
2
and is always negative. When the number
density of molecules increases, meaning a larger number of molecules in the same
volume, this energy increases in absolute value, meaning that the molecules are
4
The Avogadro constant is named after the early nineteenth century Italian scientist
Amedeo Avogadro, who, in 1811, rst proposed that the volume of a gas (at a given pressure
and temperature) is proportional to the number of atoms or molecules regardless of the nature
of the gas. The French physicist Jean Perrin in 1909 proposed naming the constant in honor
of Avogadro. Perrin won the 1926 Nobel Prize in Physics, in a large part for his work in
determining the Avogadro constant by several dierent methods.
The value of the Avogadro constant was rst indicated by Johann Josef Loschmidt who, in
1865, estimated the average diameter of the molecules in air by a method that is equivalent
to calculating the number of particles in a given volume of gas. This latter value, the number
density of particles in an ideal gas, is now called the Loschmidt constant in his honour, and
is approximately proportional to the Avogadro constant. The connection with Loschmidt is
the root of the symbol L sometimes used for the Avogadro constant, and German language
literature may refer to both constants by the same name, distinguished only by the units of
measurement.
Accurate determinations of Avogadros number require the measurement of a single quantity
on both the atomic and macroscopic scales using the same unit of measurement. This became
possible for the rst time when American physicist Robert Millikan measured the charge on an
electron in 1910. The charge of a mole of electrons is the constant called the Faraday and had
been known since 1834 when Michael Faraday published his works on electrolysis. By dividing
the charge on a mole of electrons by the charge on a single electron the value of Avogadros
number is obtained. Since 1910, newer calculations have more accurately determined the
values for Faradays constant and the elementary charge. (from Wikipedia)
5
strongly linked, When the number density decreases, ! 0. For an ideal gas,
the molecules are free from the intermolecular forces and = 0.
With these restrictions Eqs (11-13), give for the equilibrium distribution,
the Maxwell-Boltzmann distribution
f
(tj)
= n
_
m
2kT
_
1/2
e

(
!

!
u
)
2
2kT
m
: (32)
6
Lecture 03: Discretization using nite
dierences
Consider the Boltzmann equation in the form
0
t
) +
!
(r
x
)) = , (1)
where the term is given by
=
(sd)

!
q
(e)
+
!
q
(`d)

(r

)) , (2)
with the short distance repulsion term written as,

(sd)
=
Z
!
1
Z
2
"=0
Z

b=0
()
0
)
0
1
))
1
) /d-d/qd
!

1
. (3)
It is possible to make ) = )
(eq)
in the evaluation of r

) on the right hand

side of Eq. (2) without any eects on the macroscopic equations,
r

) = r

)
(eq)
=

!

!
n

kT
m
)
(eq)
. (4)
So we get a kinetic model for
=
(sd)
+

!
q
(e)
+
!
q
(`d)

!

!
n

kT
m
)
(eq)
. (5)
A kinetic model means that the Boltzmann equation is replaced by a model
that is able to retrieve the main, or some of the mains properties of the Boltz-
mann equation. In LBM, the more widely used model for
(sd)
is the BGK [1]
model
1

(sd)
=
)
(eq)
)
t
, (6)
1
"In 1951, Bhatnagar went to Harvard University, Cambridge as a Fullbright scholar for
two years. This handsome tall scholar from India was often mistaken in the University
corridors for a student. Once he took his place at the lecture rostrum, the students realized
that he was indeed a senior faculty. He lectured on mathematical theory of gases based on the
mathematically formidable book by S. Chapman and T.G. Cowling. At Harvard, he produced
two very important publications (1) a book Stellar Interiors jointly with D.H. Menzel and
H.K. Sen, published in the International Astrophysical Series and (2) a research paper in
Physical Review in 1954 which contained the famous BGK (Bhatnagar, Gross, Krook) model.
The Bolzmann equation, governing the evolution of a state of molecules in gases, contains an
extremely complicated integral term . Faced with the reality that the Boltzmann equation was
too dicult to handle due to this collision integral term, Bhatnagar, Gross and Krook used
their deep understanding of relaxation process of a swarm of molecules towards an equilibrium
state to replace this term by a much simpler term free, which has since been used as alternative
to the Boltzmann equation in solving problems in rareed gas dynamics, plasma physics and
the kinetic theory itself. " [2]
1
meaning that the collision term can be thought as a relaxation term towards
the equilibrium when a uid is in a non-equilibrium state.
Professor P.L. Bhatnagar
We will talk more about kinetic models in the following lectures. In this
lecture we focus our attention on LBM discretization and consider as a known
function

), )
(eq)

.
Discretization means to replace the entire continuous physical space, repre-
sented by the continuous variable
!
r by some points
!
r
i
and the entire velocity
space represented by the continuous variable
!
by some velocity vectors
!

i
,
i = 0, ..., :
b
1, for which the distribution ) will be calculated in dierent times
separated by a time interval t. Figure (1) represents a 37-velocity lattice found
by Philippi et al.[3], suitable for solving non-isothermal problems.
Figure 1: The D2V37 lattice, [3].
Therefore, considering a single velocity
!

i
from this set of :
b
velocities the
Boltzman equation can be written as
0
t
)
i
+
!

i
(r
x
)
i
) =
i
, (7)
2
where )
i
= )

!
r ,
!

i
, t

indicates the value of ) for

!
=
!

i
, i.e., the packet of
particles with velocity
!

i
that are found in the point
!
r at time t.
At the time t + c, this packet wil be,
)
i
(
!
r , t + t) = )
i
(
!
r , t) + t
0)
i
0t
+
1
2
(t)
2
0
2
)
i
0t
2
+ ..., (8)
and will be known from the value this packet had at time t, when all the time
derivatives
@fi
@t
,
@
2
fi
@t
2
, ...are known at time t.
This is not the case. Nevertheles, when t !0, it is possible to neglect all
the derivatives of order 2 and higher.
)
i
(
!
r , t + t) = )
i
(
!
r , t) + t
0)
i
0t
+O

(t)
2

(9)
giving
0)
i
0t
=
)
i
(
!
r , t + t) )
i
(
!
r , t)
t
+O(t) (10)
In the same way, for calculating the value )
i
in a point
!
r +
!
r , at the time
t + t
)
i
(
!
r +
!
r , t + t) = )
i
(
!
r , t) + t
0)
i
0t
+
!
r
0)
i
0
!
r
+O

(
!
r )
2
, (t)
2

.
(11)
When

!
r =
!

i
t, (12)
we get
)
i

!
r +
!

i
t, t + t

= )
i
(
!
r , t)+t

0)
i
0t
+
!

0)
i
0
!
r

+O

(
!
r )
2
, (t)
2

,
(13)
or

0)
i
0t
+
!

0)
i
0
!
r

=
)
i

!
r +
!

i
t, t + t

)
i
(
!
r , t)
t
+O(j
!
r j , t) .
(14)
And the discrete Boltzmann equation can be written as
)
i

!
r +
!

i
t, t + t

)
i
(
!
r , t) =
i
t +O

(
!
r )
2
, (t)
2

. (15)
The term
3
Figure 2: Illustration of the collision-propagation scheme in LBM. The amount
of particles in each direction is represented by the length of the vector pointing
in that direction.

t
j
!
r j
, (16)
is the Courant-Fridrich-Lewy number
2
, C11, and in the present numerical
scheme C11 = 1. This condition implis that at the time t + t the parti-
cles with velocity
!

i
will be found at the site
!
r +
!
r , at the time t + t.
This scheme is at the origin of the LB method, which i an explicite numerical
method when the interacton term
i
is evaluated at time t. Therefore, when the
external and intermolecular forces are neglected and the repulsion term
(sd)
is
modelled using a BGK relaxation term

i
=
)
(eq)
i
)
i
t
, (17)
the LB method proceeds in accordance with two steps (Figure 2).
2
In mathematics, the CourantFriedrichsLewy condition (CFL condition) is a necessary
condition for convergence while solving certain partial dierential equations (usually hyper-
bolic PDEs) numerically. (It is not in general a sucient condition.) It arises when explicit
time-marching schemes are used for the numerical solution. As a consequence, the timestep
must be less than a certain time in many explicit time-marching computer simulations, other-
wise the simulation will produce wildly incorrect results. The condition is named after Richard
Courant, Kurt Friedrichs, and Hans Lewy who described it in their 1928 paper, [4], [5].
For example, if a wave is crossing a discrete grid, then the timestep must be less than the
time for the wave to travel adjacent grid points. As a corollary, when the grid point separation
is reduced, the upper limit for the time step also decreases. In essence, the numerical domain
of dependence must include the analytical domain of dependence in order to assure that the
scheme can access the information required to form the solution.
4
a) Collision step: the distributions )
i
are recalculated in each site
!
r
i
using
the local information in this site
)
0
i
(
!
r
i
, t + t) = )
i
(
!
r
i
, t) +
)
(eq)
i
(
!
r
i
, t) )
i
(
!
r
i
, t)
t,t
. (18)
b) Propagation step: the new values of the distributions )
0
i
are propagated
along the i-directions
)
i

!
r +
!

i
t, t + t

= )
0
i
(
!
r , t + t) . (19)
References
[1] Bhatnagar, P.L., Gross, E.P and Krook, M., A model for collisional processes
in gases I: small amplitude processes in charged and in nutral one-component
systems, Phys. Rev. 94, 1954, 511-525.
[2] P. Prasad, 19th P.L. Bhatnagar Memorial Award Lecture -
2006, delivered at 72nd Annual Conference of the Indian
Mathematical Society at Jabalpur, 27 - 30 December, 2007
(http://math.iisc.ernet.in/~prasad/prasad/plbhatnagar.pdf)
[3] P. C. Philippi, L. A. Hegele, Jr., L. O. E. dos Santos, and R. Surmas, Phys.
Rev. E 73, 056702 (2006).
[4] R. Courant, K. Friedrichs and H. Lewy, ber die partiellen
Dixoerenzengleichungen der mathematischen Physik, Mathematische
Annalen, vol. 100, no. 1, pages 3274, 1928.
[5] R. Courant, K. Friedrichs and H. Lewy, "On the partial dixoerence equations
of mathematical physics", IBM Journal, March 1967, pp. 215-234, English
translation of the 1928 German original.
5
Lecture 04: The Boltzmann equation for non-ideal
uids
Consider the Boltzmann equation in the form
0
t
) +

0
x
) =
(sd)
+

q
(`d)

n
_
kT
m
)
(eq)
+

q
(e)

n
_
kT
m
)
(eq)
(1)
where

q
(e)
corresponds to the force per unit mass due to external body forces,

q
(`d)
to the intermolecular attraction among the molecules, which we will sup-
pose to be only dependent of the distante between the centers of the molecules
and,
(sd)
, to the collision term related to the short range repulsion forces (Fig-
ure 1 ).
Figure 1: Force eld around an isolated molecule.
The term
(sd)
should, then, be understood as related to the repulsion eld
inside the o-spheres of radius o and which centers are the material points used
to represent the molecules. In the surface of this sphere, this repulsion eld
is exactly equilibrated by the electrostatic attraction eld and the resulting
force is null (Figure 2). Therefore, in its trajectory inside the o-sphere of the
target particle, the bullet particle will be frained and deviated from the center
of the target, transforming part of its kinetic into potential energy, which is only
recovered when the bullet is again in the surface of the o-sphere in its exiting
trajectory
1
Figure 2: Repulsion o-sphere of the target particle, indicating the surface where
the two repulsion elds intercept with themselves.
The term

q
(`d)

n
_
kT
m
)
(eq)
(2)
related to the long-range interaction, deserves some discussion.
Consider a molecule labelled as 1 located in point

r
1
and all the molecules
labelled as 2 around it. The potential energy related to the interaction of all
molecules 2 with the single molecule 1 is gven by

1
=
_
[
!
x 2
!
x 1[>

(12)
:(

r
2
) d

r
2
(3)
where
(12)
is the potential energy (in Joules) related to the electrostatic inter-
action between the particle 1 and a single particle 2.
Considering that the interaction length is not too large, and that the number
of particles per unit volume : does not have strong variations, it is possible to
consider :(

r
2
) = :(

r
1
) inside the integration domain and we get,
2

1
= 2:(

r
1
)
1
2
_
[
!
[>

(12)
d

j
. .
a
= 2a:(

r
1
) (4)
Where a is a positive constant, because
(12)
< 0. This constant depends
solely on the electrostatic eld generated by the molecules being a molecular
property of the substance that is being analyzed.
The force

q
(`d)
on molecule 1 will thus be

q
(`d)
=
\
1
:
(5)
Therefore the Boltzmann equation takes the form
0
t
) +

0
x
) =
(sd)
. .
repulsion
+
(`d)
. .
intermolecular attraction
+

q
(e)

n
_
kT
m
)
(eq)
. .
external body forces
(6)
where

(`d)
=
2a
:
\:
_

n
_
kT
m
)
(eq)
(7)
represents the net gain of particles that acquire the velocity

due to the
intermolecular forces from the other particles around it.
Exercise: Discretize Eq. (6) considering a single discrete velocity

i
and
a Courant-Friedrich-Lewy number, C11 = 1. Propose a method to discretize
\: in 2D discrete spaces (hint: use Taylor series).
Appendix: Intermolecular attraction forces
1
Intermolecular forces are, essentially, electrostatic forces, since every mole-
cule is a site of positive and negative electrical charges. Coulombs forces
happen among ions, in accordance with
1
12
=
1
4c
0

2
r
2
12

r
12
[

r
12
[
, (8)
1
References [2] and [1] are indicated for further reading.
3
when two charged particles
1
and
2
separated by a distance r
12
, where c
0
is the
electrical permittivity of vacuum equal to 8.8541910
12
(con|o:/)
2
,
_
:
2
_
.
The potential energy is given by

1 = \, (9)
giving

12
=
1
4c
0

2
r
12
(10)
Coulomb forces are inversely proportional to the square of the distance be-
tween any two ions and have a much longer interaction length when compared
to the other intermolecular forces. Asymmetric molecules such as H
2
0 and CO
2
have a permanent dipole, the geometric center of the electronic clouds does not
coinciding with the geometric center of the positive charges (Figure 3). Their
dipole moment vector is,

j = /

:
`
, (11)
where / is the distance that separates the charges and

:
`
is a unitary vector
giving the dipole direction and conventionally oriented from the negative to the
positive charge.
Figure 3: H
2
O and CH
4
molecules. Intermolecular forces are dependent on the
molecular shape.
Consider two dipoles which centers are separated by a distance r
12
, let

:
`;1
to be given in terms of the polar 0
1
and azimuthal, c
1
angles and

:
`;2
, in terms
4
of 0
2
and c
2
, Figure 4. When r
12
is much larger than the dipole lenghts, these
two dipoles will share a potential energy given by

12
=
1
4c
0
j
1
j
2
r
3
12
[2 cos 0
1
cos 0
2
sin0
1
sin0
2
cos (c
1
c
2
)] (12)
Figure 4: Calculation of the potential energy between two dipoles.
When c
1
= c
2
= 0, the dipoles are in a single vertical plane and the potential
energy becomes

12
=
1
4c
0
j
1
j
2
r
3
12
[2 cos 0
1
cos 0
2
sin0
1
sin0
2
] (13)
which attains a minimum when the dipoles are aligned, 0
1
= 0
2
= 0, the positive
end of dipole 1 being at the closest position with respect to the negative end of
dipole 2. Potential energy becomes a maximum, when 0
1
= 0 and 0
2
= .
Molecules are subjected to random uctuations in their translational motion
as they collide with other molecules. Kinetic energy, 1
c
, is exhanged among the
colliding molecules and the rotational and vibrational modes are excited from
the available translational energy. The potential energy, related to the elec-
trostatic attraction between any two molecules is converted into kinetic energy
during the approximation. Two molecules that are permanent dipoles have a
minimum in their mutual potential energy, < 0, when they are aligned. When
these molecules do have enough kinetic energy, i.e., when +1
c
_ 0, they will
rest aligned at their closest position. They are only uncoupled when they receive
enough energy to liberate them from the attractive electrostatic eld by, e.g.,
collisions with another molecules.
Collision frequency and, also, molecular kinetic energy are increased when
the temperature increases.In this manner, it may be supposed that the prob-
ability that a pair of dipoles have a potential energy
12
follows a Boltzmann
distribution
5
1 () = c

kT
(14)
In this case, considering all the possible directional congurations of this
pair, given by all the possible polar 0
1
, 0
2
and azimuthal angles, c = c
1
c
2
,
the ensemble average potential energy of this couple is given by

12
=
_

12
c

12
kT
d
_
c

12
kT
d
(15)
where
d = sin0
1
sin0
2
d0
1
d0
2
dc (16)
It results, neglecting higher order terms,

12
=
2
3
_
1
4c
0
_
2
j
2
1
j
2
2
r
6
12
/T
(17)
In this way, when the temperature T increases, the molecules have more
available kinetic energy and the congurations with a small potential energy,
related to a strong electrostatic attraction, are less frequent in the ensemble of
possible congurational states of the molecular system.
The ensemble average intermolecular force between any two polar molecules
separated by

r is given by
_

1
12
_
= 4
_
1
4c
0
_
2
j
2
1
j
2
2
r
7
12
/T

:
r
(18)
The dipole length and the dipolar moment of a polar molecule are subjected
to uctuations produced by electrostatic elds originated from other molecules.
In the same manner, a non-polar molecule will acquire a dipolar moment, j
ind
,
under the induction of an electrical eld,

1. We dene the molecular polariz-
ibility, c, by

j
ind
= c

1 (19)
giving a measure of the easiness a molecule becomes polarized, or changes its
dipolar moment, when subjected to an electrical eld
When a polar molecule, 1, is at a distance r
12
from a non-polar molecule, 2,
the electrical eld at 2 will be

1
2
=
1
2c
0
j
1
r
3
12

:
r
(20)
and the dipolar moment

j
ind
=
c
2
2c
0
j
1
r
3
12

:
r
(21)
6
Considering that the induced dipole 2 has always the same orientation of
the electrical dipole that is at its origin, the potential energy that is shared
by a polar 1 and a non-polar 2 molecule can be calculated using Eq. (13) for
0
1
= 0
2
= 0, giving

12
=
1
2
1
(2c
0
)
2
c
2
j
2
1
r
6
12
(22)
London forces
Symmetrical molecules do not have any permanent electrical dipole and the
attractive forces among these molecules was rst described by London in 1930
as due to very rapid uctuations of the geometric center of their molecular
electronic clouds. These uctuations originate temporary dipoles that are con-
stantly changing in magnitude and direction with a characteristic electronic
frequency i
0
that can be related to the refractive index : of the uid, when a
beam of polarized light with frequency i traverses it, following,
: = 1 +
c
i
2
0
i
2
(23)
where c is a constant.
Each time a molecule has a instantaneous dipolar moment j it induces a
dipolar moment in a neighbor molecule. The strenght of the interaction de-
pends on the polarizibility of the rst molecule, because the strenght of the
instantaneous dipole moment j depends on the loseness of the control that the
nuclear charges exert on the outer electrons and also on the polarizibility of the
second molecule, because that polarizibility depends on how readly a dipole can
be induced by another molecule. The resulting potential energy was found to
be, [2], [1],

12
=
3
2
1
(4c
0
)
2
c
1
c
2
r
6
12
1
1
1
2
1
1
+ 1
2
(24)
where 1 is the rst ionization potential (Figure 5), i.e., the work that must be
performed to remove one electron from the uncharged molecule, 1 - /i
0
. The
attractive forces among non-polar molecules are also called dispersion forces
considering the dispersion of light in the medium due to the oscilating dipoles.
Polar molecules also atract themselves with dispersion forces, but, for polar
molecules, the time average of their dipolar moment is dierent from zero (being
equal to its permanent dipolar moment).
When an apolar molecule is close to a polar molecule, the second molecule
induces a dipolar moment on the rst one and the rst molecule induces a
uctutation on the geometrical center of the electrons of the second molecule
and, consequently a uctuation on the permanent dipolar moment of this mole-
cule. The two molecules will attract themselves with a London dispersion force
superposed to a Debye induction force.
7
Figure 5: Ionization potential of some molecules
Summary
All these forces can be writen as

12
=
1
(12)
[

r
2

r
1
[
6
(25)
The Table below summarizes the several contributions to the potential en-
ergy.
References
[1] Prausnitz, J. M. , Lichtenthaler, R. N. & de Azevedo, E. G. 1999 Molec-
ular Thermodynamics of Fluid-Phase Equilibria, 3rd Edition, New Jersey:
Prentice Hall.
[2] A. W. Adamson, Physical Chemistry of Surfaces, John Wiley (1997).
8
Lecture 05: Properties of the repulsion term
Consider the integral
Z

(sd)

'

d
!

=
Z Z

f
0
f
0
1
ff
1

'

gbdbd"d
!

1
d
!
; (1)
for any function '

of the molecular velocity
!
.
Although
(sd)

is dierent from
(sd)

; the value of the integral
Z

(sd)

'

d
!
=
Z

(sd)

'

d
!

1
; (2)
and so
Z Z

f
0
f
0
1
ff
1

'

gbdbd"d
!

1
d
!

=
Z Z

f
0
f
0
1
ff
1

'

gbdbd"d
!

1
d
!
: (3)
On the other hand if we change
!

!

0
and, consequently,
!

0
1
,
Z
!

0
Z
!

0
1
Z
2
"=0
Z

b=0

ff
1
f
0
f
0
1

'

g
0
bdbd"d
!

0
1
d
!

0
=
Z
!

0
Z
!

0
1
Z
2
"=0
Z

b=0

f
0
f
0
1
ff
1

'

gbdbd"d
!

0
1
d
!

0
; (4)
because g = g
0
. Since the intermolecular eld is a central eld (Figure 1) we
can also show, (see Lecture 01, Appendix A) that d
!

0
1
d
!

0
= d
!

1
d
!
meaning
that the transformation
!

1
;
!
!
!

0
1
;
!

0
has a Jacobian, J = 1:Changing
!

1
;
!
by
!

0
1
;
!

0
is equivalent to invert the collision term and under a central eld
the gain of particles with velocities
!

1
;
!
when they have the velocities
!

0
1
;
!

0
before the collision is the same gain of particles with velocities
!

0
1
;
!

0
when
they have the velocities
!

1
;
!
before the collision.
Therefore
Z

(sd)
'

d
!
=
Z

(sd)
'

d
!

=
Z

(sd)
'

d
!
=
Z

(sd)
'

0
1

d
!
; (5)
1
g
g

b
gg
g g

b
Figure 1: Trajectory of the bullet under a central eld.
or
Z

(sd)
'

d
!
=
1
4
Z Z

(sd)

' + '
1
'
0
'
0
1

d
!
: (6)
When '

is a collisional invariant, '

= m; m
!
;
1
2
m
2
, we will get,
Z

(sd)
'

d
!
= 0: (7)
This property is important and is directly related with the preservation of
mass, momentum and kinetic energy in collisions.
It is easy to show that this property is satised by the BGK relaxation term

(sd)
=
f
(eq)
f

; (8)
meaning that, when '

is a collisional invariant,
Z
f
(eq)
f

'

d
!
= 0: (9)
2
Lecture 06: Macroscopic equations
Moments of the distribution )
For a given point
!
r , the distribution ) gives the amount of particles with
velocities between
!
and
!
+ d
!
. When this distribution is integrated in
the velocity space, we loss this detailed information and get statistical average
values, which are the macroscopic variables of interest. The rst moments of
the distribution ) are:
density , j =
_
:)d
!
, (1)
momentum per unit volume, j
!
n =
_
:
!
)d
!
, (2)
mean kinetic energy per unit volume, jc
c
=
_
1
2
:
2
)d
!
, (3)
Writing

2
=
_
!

!
n
_
2
+ 2
_
!

!
n
_

!
n + n
2
, (4)
we see that the kinetic energy has two parts,
jc
c
=
_
)
1
2
:
_
!

!
n
_
2
d
!

. .
=t
cj
+
1
2
jn
2
. .
advection energy
(5)
The rst part, jc
c,}
, is due to the random motion of the molecules around the
local velocity
!
n . This kinetic energy is dierent from zero even when the local
velocity
!
n = 0. This kinetic energy is called the mean peculiar kinetic energy
of the molecules and can be used to dene the thermodynamic temperature, T,
by requiring
1
:
_
)
1
2
:
_
!

!
n
_
2
d
!
=
1
:
_
)
(tj)
1
2
:
_
!

!
n
_
2
d
!
=
1
2
/T (6)
since the thermodynamic internal energy per molecule for an ideal gas, with
solely translational degrees of freedom in equilibrium is known, from statistical
mechanics
1
, to be
1
2
/T.
When ,
_
!

_
is a collisional invariant, ,
_
!

_
= :, :
!
,
1
2
:
2
, the mo-
ments
_
),
_
!

_
d
!
(7)
1
See the link: http://en.wikipedia.org/wiki/Degrees_of_freedom_(physics_and_chemistry)
1
are equilibrium moments
_
),
_
!

_
d
!
=
_
)
(tj)
,
_
!

_
d
!
(8)
in the sense that, in equilibrium, these are the only moments that show a macro-
scopic manifestation: density, local velocity (or momentum) and temperature.
We will, now, consider the moments of interest in non-equilibrium conditions.
The tensor

c{
=
_
(:
o
)
o
)d
!
(9)
represents the rate with which the momentum :
o
is transferred along the di-
rection , due to the molecules that move with a non-null velocity component
along this direction. In the SI units,
_

c{

=
|(
r
s
)
2
n
3
=
|(
r
s
2
)n
n
3
=

n
2
, rep-
resenting a force along the direction c per unit area, orthogonal to , (Figure
1)
Figure 1: Tensor
Writting

o
= (
o
n
o
)
_

o
n
o
_
+
o
n
o
+
o
n
o
n
o
n
o
(10)
it can be seen that

c{
=
_
:(
o
n
o
)
_

o
n
o
_
)d
!
+ jn
o
n
o
(11)
Now, let the distribution ) be decomposed in two parts,
) = )
(tj)
+ )
(ntj)
, (12)
considering the second part, )
(ntj)
, as a perturbation of )
(tj)
.
2
By dening,
!
C
o
=
!

!
n
o
_
2|T
n
_
1/2
(13)
we see that,
_
:(
o
n
o
)
_

o
n
o
_
)
(tj)
d
!

= ::
_
:
2/T
_
1/2
_
c

(r)
2
2!J
r
(
o
n
o
)
_

o
n
o
_
d
!

= ::
_
:
2/T
_
1/2
_
2/T
:
_T
2
+1
_
c
c
2
C
o
C
o
d
!
C
=
::

1/2
_
2/T
:
__
c
c
2
C
o
C
o
d
!
C
=
2:/T

1/2
__
c
c
2 C
2
3
d
!
C
_
. .
=

T2
2
c
oo
= 1
0
c
oo
(14)
where
1
0
= :/T (15)
is the thermodynamic pressure of an ideal gas.
Each molecule has its own velocity
!
. When the uid is at rest
!
n =
_
!

_
= 0, and :
o

o
indicates the momentum :
o
that is transported in the
,-direction by a single molecule. In equilibrium, since the MB distribution is
isotropic (independent of the direction), the rate with which the momentum c is
transferred into a direction dierent from c by the whole set of molecules is zero.
We have solely transfer of momentum in the same direction of this momentum
and this transfer is independent of the direction (a consequence of isotropy).
When these molecules transport momentum in the r-direction and nd a bar-
rier in their trajectory they return with the same r-velocity (elastic collisions)
indicating a force on the barrier given by the net balance of momentum along
the r-direction,
3
_
1
1
d
:
_
1
1
d

_
1
0
:
r

r
)
(tj)
d
r
. .
molcules moving to the right

_
1
1
d
:
_
1
1
d

_
1
0
:
r

r
)
(tj)
d
r
. .
molcules moving to the left
=
_
1
1
d
:
_
1
1
d

_
1
0
:
r

r
)
(tj)
d
r
+
_
1
1
d
:
_
1
1
d

_
0
1
:
r

r
)
(tj)
d
r
=
_
1
1
d
:
_
1
1
d

_
1
1
:
r

r
)
(tj)
d
r
= 1 = :/T (16)
Consider now the non-equilibrium moment,
_
:(
o
n
o
)
_

o
n
o
_
)
(ntj)
d
!
= t
oo
(17)
representing a diusive ux of momentum which appears in non-equilibrium
conditions. For a uid at rest,
!
n = 0, we get
t
oo
=
_
:
o

o
)
(ntj)
d
!
= 0 (18)
Because, in this case, t
oo
is an equilibrium moment
_
:
o

o
)d
!
=
_
:
o

o
)
(tj)
d
!
= :/Tc
oo
(19)
In 2D channel ows, for t
r
be something dierent from zero, the velocity
n
r
must vary along the direction j, in which case
t
r
=
_
:(
r
n
r
)
. .
co
_

_
. .
c
)
(ntj)
d
!
(20)
is related to the diusive transfer of the r momentum along the j-direction.This
gives rise to the concept of viscosity. Viscosity is caused by the transfer of
momentum between two planes sliding parallel to one another but at dierent
rates, and this momentum is transferred by molecules moving between the planes
(Figure 2). Molecules from the faster plane move to the slower plane and tend
to speed it up.
We get for the tensor
c{
, Eq. (11)

c{
= :/Tc
oo
+ t
oo
+ jn
o
n
o
(21)
4
Figure 2: Faster molecules in plane j transfer their r-momentum when moving
to plane j + dj, while slower molecules in the plane j + dj transfer their r-
momentum when moving to plane j. This gives rise to a frainage tangential
force in the plane j and to an acceleration force in the plane j +dj. When the
two planes move with the same velocity, t
r
= 0.
Another moment of interest is the one giving the ow of kinetic energy
transported by the molecules,
!
c =
_
)
1
2
:
2
!
d
!
(22)
Decomposing
!
=
!
n +
_
!

!
n
_
(23)
we get,
!
c =
!
n
_
)
1
2
:
2
d
!
+
_
)
1
2
:
2
_
!

!
n
_
d
!

=
_
jc
c,}
+
1
2
jn
2
_
!
n
+
_
)
1
2
:
_
_
!

!
n
_
2
+ 2
_
!

!
n
_

!
n + n
2
_
_
!

!
n
_
d
!
(24)
Decomposing ) = )
(tj)
+ )
(ntj)
5
_
)
1
2
:
_
_
!

!
n
_
2
+ 2
_
!

!
n
_

!
n + n
2
_
_
!

!
n
_
d
!

=
_
)
(tj)
1
2
:
_

_
_
!

!
n
_
2
. .
!0
+ 2
_
!

!
n
_

!
n
. .
!10
!
u
+ n
2
..
!0
_

_
_
!

!
n
_
d
!

+
_
)
(ntj)
1
2
:
_

_
_
!

!
n
_
2
+ 2
_
!

!
n
_

!
n
. .
!br
!
u
+ n
2
..
!0
_

_
_
!

!
n
_
d
!

= 1
0
!
n +t
!
n +
!
(25)
where
!
is the heat ow vector,
!
=
_
)
(ntj)
1
2
:
_
!

!
n
_
2
_
!

!
n
_
d
!
(26)
This vector indicates how much of kinetic energy related to the relative
mouvement of the molecules is transferred into a direction due to the relative
velocity of the molecules along that direction. This ux is diusive and for a
uid at rest it is
!
=
_
)
(ntj)
1
2
:
2
!
d
!
(27)
The term 1
0
!
n , related to the compression work and the term t
!
n will be
discussed later.
Macroscopic equations
Consider the Boltzmann equation
0
|
) +
!
0
r
) =
(sJ)
. .
repulsion
+
(|J)
. .
intermolecular attraction
+
!
q
(t)

_
!

!
n
_
|T
n
)
(tj)
. .
external body forces
(28)
where

(|J)
=
2a
:
r:
. .
!

(d)

_
!

!
n
_
|T
n
)
(tj)
(29)
Mass conservation
6
Multiplying the Boltzmann equation by : and integrating in the velocity
space, remarking that
_
:0
|
)d
!
= 0
|
_
:)d
!
= 0
|
j (30)
because the time t and the molecular velocities
!
are independent variables.
On the other hand,
_
:
!
0
r
)d
!
= r
r

_
:
!
)d
!
= r
r
(j
!
n ) (31)
since the molecular velocities are also independent of
!
r .
For the repulsion term
_
:
(sJ)
d
!
= 0 (32)
because : is a collisional invariant.
For the force terms
_
:
!
q
_
!

!
n
_
|T
n
)
(tj)
d
!
= 0
because
!
q depends only on the position, independently if
!
q is the external
body or the intermolecular force and the MB distribution is an even function of
!

!
n .
Therefore we get,
0
|
j +r (j
!
n ) = 0 (33)
which is the mass conservation equation.
Momentum balance equation
Multiplying the Boltzmann equation by :
!
and integrating in the velocity
space we will get
0
|
(jn
o
) + 0
o
_

c{
_
= j
!
q
(t)
+ j
!
q
(|J)
(34)
which is the momentum balance equation (do that as an exercise).
Replacing
c{
from Eq. (21)
0
|
(jn
o
) + 0
o
(jn
o
n
o
+ 1
0
c
oo
) = j
!
q
(t)
+ j
!
q
(|J)
0
o
t
oo
(35)
The term 1
0
=:/T corresponds to the thermodynamic pressure for a system
of material points without intermolecular attraction. The term
j
!
q
(|J)
= j
2a
:
r: = 2a:r: = r
_
a:
2
_
(36)
Therefore, Eq. (35) can be written as
7
0
|
(jn
o
) + 0
o
_
_
_jn
o
n
o
+
_
:/T a:
2
_
. .
=1
c
oo
_
_
_ = jq
(t)
o
0
o
t
oo
(37)
The term
1 = :/T
..
repulsive part
a:
2
..
attractive part
(38)
is the thermodynamic pressure for a system of material points with intermolec-
ular attraction forces.
The momentum balance equation may also be written in the form,
0
|
(jn
o
) + 0
o
(jn
o
n
o
) = jq
(t)
o
0
o
1 0
o
t
oo
(39)
or, using the mass conservation equation,
j0
|
(n
o
) + jn
o
0
o
(n
o
) = jq
(t)
o
0
o
1 0
o
t
oo
(40)
It should be observed that Eq. (38) is not still the van der Waals equation
of state,
1 =
RT
/

a

2
=
:/T
1 /:
a:
2
(41)
because our molecules are being represented by material points without volume,
which coresponds to the parameter / in the van der Waals equation of state.
8
Lecture 07: Phase transitions
The surface tension counteracts the weight of a small liquid drop.
When a liquid has a phase transition, the interface liquid-vapor is too much
thin to enable the density : to be considered constant through it, as we have
done in the Lecture 04, when calculating the potential energy on a point
!
r
1
to
to the interaction of a molecule in this point with all the molecules around it

1
=
_
j
!
r 2
!
r 1j,c

(12)
:(
!
r
2
) d
!
r
2
. (1)
Nevertheless, we can consider a Taylor series for :,
:(
!
r
2
) = :(
!
r
1
) +
0:
0r
o
(r
2o
r
1o
)
. .
=q

+
1
2
0:
0r
o
0:
0r
o
(r
2o
r
1o
)
. .
=q

(r
2o
r
1o
)
. .
=qo
+ ...,
(2)
giving,

1
=
_
j
!
r 2
!
r 1j,c

(12)
:(
!
r
2
) d
!
r
2
= 2:(
!
r
1
)
1
2
_
j
!
q j,c

(12)
d
!
j
. .
=o
+
0:
0r
o
_
j
!
q j,c

(12)
j
o
d
!
j
. .
=0
+
1
2
0:
0r
o
0:
0r
o
_
j
!
q j,c

(12)
j
o
j
o
d
!
j . (3)
1
The potential
(12)
depends solely on the norm of
!
j being an even function
of j
r
, j

and j
:
. Therefore
_
j
!
q j,c

(12)
j
o
j
o
d
!
j = c
oo
_
j
!
q j,c

(12)
j
2
3
d
!
j . (4)
On the other hand, in the same way as for the force parameter a, the term
_
j
!
q j,c

(12) q
2
3
d
!
j depends only on the electrical properties of the molecules,
being a constant
_
j
!
q j,c

(12)
j
2
3
d
!
j = 2
_
j
!
q j,c

(12)
j
2
6
d
!
j = 2i. (5)
So

1
= 2a:(
!
r
1
) ir
2
:(
!
r
1
) + ..., (6)
and the intermolecular attraction force will be given by
!
q
(|J)
=
r
1
:
, (7)
or
j
!
q
(|J)
= :r
1
. (8)
So
j
!
q
(|J)
= 2a:r: + :ir
_
r
2
:
_
. (9)
In this case, the kinetic equation will be written as
0
|
) +
!
0
r
) =
(:t)
. .
repulsion
+
(|J)
. .
intermolecular attraction
+
!
q
(t)

_
!

!
n
_
|T
n
)
(tj)
. .
external body forces
, (10)
with

(|J)
=
!
q
(|J)

_
!

!
n
_
|T
n
)
(tj)
=
_
2a
:
r: +
i
:
r
_
r
2
:
_
_

_
!

!
n
_
|T
n
)
(tj)
. (11)
The momentum balance equation will be also modied. Considering that,
2
:r
_
r
2
:
_
= :0
o
(0
o
(0
o
:))
= 0
o
[:(0
o
(0
o
:))] 0
o
(0
o
:) 0
o
:
= 0
o
[:(0
o
(0
o
:))] 0
o
[0
o
:0
o
:] + (0
o
:) 0
o
0
o
:, (12)
and
(0
o
:) 0
o
0
o
: =
1
2
0
o
[0
o
:0
o
:] , (13)
we get
j
!
q
(|J)
= 2a:r: + ir
_
:r
2
:
_
+
1
2
ir(r: r:) ir (r:r:)
= r
__
a:
2
+ i:r
2
: +
1
2
i(r: r:)
_

c i(r: r:)
_
. (14)
The momentum balance equation will thus have the form
0
|
(j
!
n ) +r. (j
!
n
!
n )
= j
!
q
(t)
. .
external forces
r1
s
..
scalar pressure
r t
. .
viscous forces
r

o
. .
surface forces
, (15)
where the scalar pressure is given by
1
s
= :/T
..
repulsion
a:
2
..
attraction
. .
1=thermodynamic pressure
i:r
2
:
1
2
ir: r:
. .
surface pressure
, (16)
and the surface tensor is given by

o = ir: r:. (17)

The surface forces have a role solely in the interface liquid vapor where the
gradients are strong.
3
Lecture 09: Velocity discretization
The velocity discretization is a critical step in deriving the lattice Boltzmann
equation from the continuous Boltzmann equation since it is intended, in this
step, to replace the entire continuous velocity space <
1
by a nite set of discrete
velocities
!

I
, i = 0, ..., :
b
1. In fact, it was shown, [1], [2], that the number :
b
of poles
!

I
and the poles themselves are dened by the order of the equilibrium
moments to be preserved in the discrete space.
A Chapman-Enskog analysis shows that the set of necessary conditions for
the correct thermo-hydrodynamic equations to be retrieved is given by assuring
that the discrete distributions )
i;eq
used in the LB equation satisfy,
n
b
1

I=0
)
i;eq
,
_
!

I
_
=
_
)
MB
,
_
!

_
d
!
, (1)
for all equilibrium moments h,i of interest, , = 1,
o
,
2
, ...
In a quadrature problem we want to nd the values r
I
inside a given interval
[a, /] and the weights \
I
in such a manner that
_
b
o
)(r)dr =
n

I=1
\
I
)(r
I
). (2)
This means that although we cannot preserve the function )(r) along the
entire interval [a, /], we expect that with suitable values of r
I
and \
I
it will be
possible to preserve the area below )(r), namely, the rst moment of )(r).
Considering the discretization as a quadrature problem, the discrete distribu-
tions )
i;eq
on the left hand side of the above equation are replaced by )
MB
_
!

I
_
,
i.e., by the values of the MB distribution evaluted at the pole
!

I
, multiplied by
a weight \
I
to be attributed to each velocity vector
!

I
, in order to satisfy the
quadrature condition
b1

I=0
\
I
c
m
2
i
2kT
0
_
2/T
0
:
_
1/2
)
MB
_
!

I
_
,
_
!

I
_
=
_
)
MB
_
!

_
,
_
!

_
d
!
, (3)
where the factor
_
2t|T0
n
_
1/2
was introduced for assuring the weights \
I
to be
dimensionless, T
0
is a reference temperature.
Therefore, as before, solving this quadrature problem is what is wanted,
leading to a set of discrete velocities
!

I
and weights \
I
in such a manner that
all the equilibrium moments of interest are preserved.
The MB distribution )
MB
_
!

0
_
is then developped in a series of Hermite
polynomial tensors H
(n)
:n
1
)
MB
_
!

0
_
=
_
1
2
:
/T
0
_
1/2
c

2
0
2
1

n=0
1
:!
a
(n)
:n
(:,
!
n
0
, ) H
(n)
:n
_
!

0
_
, (4)
where
!

0
=
_
|T0
n
_
1/2 !
is a dimensionless velocity,
!
n
0
=
_
|T0
n
_
1/2
!
n is a local
dimensionless velocity
=
T
T
0
1. (5)
The rst Hermite polynomial tensors are
H
(0)
= 1,
H
(1)
o
=
0,o
,
H
(2)
oo
=
0,o

0,o
c
oo
,
H
(3)
oo~
=
0,o

0,o

0,~

1
3
_
c
oo

0,~
+ c
o~

0,o
+ c
o~

0,o
_
,
H
(4)
oo~o
=
0,o

0,o

0,~

0,o

1
6
_
c
oo

0,~

0,o
+ c
o~

0,o

0,o
+ c
oo

0,o

0,o
+c
o~

0,o

0,o
+ c
oo

0,o

0,~
+ c
~o

0,o

0,o
_
(6)
+
1
3
(c
oo
c
o~
+ c
o~
c
oo
+ c
oo
c
o~
) , (7)
Using the orthogonality of the Hermite polynomial, the tensors parameters
a
(n)
:n
(:,
!
n
0
, ) can be found as
a
(n)
:n
=
_
/T
0
:
_
1/2
_
H
(n)
:n
_
!

0
_
)
MB
d
!

0
. (8)
The rst parameters in the Hermite development are
a
(0)
= :,
a
(1)
o
= :n
0,o
,
a
(2)
oo
= :(n
0,o
n
0,o
+ c
oo
) ,
a
(3)
oo~
= :
_
n
0,o
n
0,o
n
0,~
+
1
3
(n
0,~
c
oo
+ n
0,o
c
~o
+ n
0,o
c
o~
)
_
,
a
(4)
oo~o
= :
()
_
_
_
_
n
0,o
n
0,o
n
0,~
n
0,o
+
1
6

_
n
0,o
n
0,o
c
~o
+ n
0,o
n
0,~
c
oo
+ n
0,o
n
0,o
c
o~
+n
0,o
n
0,~
c
oo
+ n
0,o
n
0,o
c
o~
+ n
0,~
n
0,o
c
oo
_
+
1
3

2
(c
oo
c
o~
+ c
o~
c
oo
+ c
oo
c
o~
)
_
_
_
_
. (9)
2
Now, the Hermite expansion of the MB distribution is truncated by removing
all Hermite polynomial tensors H
(n)
:n
with order : . Indeed, it can be shown,
[1], [2], [3], that the number / of discrete velocities used in LBE is directly related
to the order of approximation of the nite sum )
()
MB;N
to the full MB distribution
)
()
MB
.
Therefore, solving a discrete LBE gives an approximation to the full kinetic
equation, whose accuracy can be as high as it can be achieved by increasing the
order of approximation of the Hermite expansion and, consequently, the number
/ of discrete velocities used in the representation of the continuous velocity space.
Nevertheless, since the main interest in numerically solving a LBE is to
describe the behaviour of a physical system whose macroscopic equations are
known, or supposed to be known, the expansion of the MB distribution in a
nite set of orthogonal polynomial tensors of order of approximation has the
important feature of preserving the moments

,
()
_
of order smaller or equal
than ,
_
)
MB;N
,d
!
=
_
)
MB
,d
!
. (10)
Using this property by replacing )
MB
by )
MB;N
in Eq. (3), the following
equation is obtained,
1
(2)
1/2
_
c

2
0
2
H
(n)
:n
_
!

0
_
H
(n)
:m
_
!

0
_
d
!

0
=
b1

I=0
\
I
H
(n)
:n
_
!

0,I
_
H
(n)
:m
_
!

0,I
_
, (11)
for all :, : smaller or equal than . By dening the inner product in the
discrete space
()
that maps the velocity space onto the real numbers as
(), q) =
n
b
1

I=0
\
I
)
_
!

0,I
_
q
_
!

0,I
_
, (12)
it can be seen that Eq. (11) requires that the norm and orthogonality of Hermite
polynomial tensors are preserved in the discrete space
()
for all :, : smaller
or equal than . This means that the discrete velocities
!

0,I
and weights \
I
must be chosen in such a manner as to satisfy these conditions.
In Philippi et al., [1], it is shown that when the discrete velocity space is
invariant under ,2 rotations and reections about the r, j and . axis, the
norm preservation of the Hermite polynomial tensors in
()
is a necessary and
sucient condition for the orthogonality among these functions.This is impor-
tant, since this property reduces the discretization problem to nd the weights
\
I
and poles
!

0,I
satisfying solely the norm restrictions, Eq.(11), when : = :.
3
If a set
!
c
I
, i = 0, ...., :
b
1, of lattice vectors is chosen, the discretization
problem reduces to nd the weights \
I
and a scaling factor , such that
!

0,I
=

!
c
I
, by solving the norm preservation condition, Eq. (11), with : = :.
Considering that the poles
!
c
I
are previously known this method was called
quadrature with prescribed abcissas. When the set
!
c
I
, i = 0, ...., :
b
1, is
regular, the lattice is space lling and enables the use of collision-propagation
schemes in the numerical solution of the LBE. When this set is not regular,
another numerical method, such as nite dierence, must be used. Surmas et
al., [5] and Shan, [3], show several discrete regular and non-regular velocity sets
in one, two and three dimensions, obtained with this method.
Finaly, from Eqs. (1) and (3), using a truncated expansion of )
MB
_
!

I
_
,
Eq (4) it is seen that the projections of the MB distribution along the discrete
poles
!

0,I
are,
)
i;eq
= \
I
c
m
2
i
2kT
0
_
2/T
0
:
_
1/2
)
MB
_
!

I
_
= \
I

n=0
1
:!
a
(n)
:n
(:,
!
n
0
, ) H
(n)
:n
_
!

0,I
_
(13)
= \
I
_
a
(0)
H
(0)
(
!
c
I
) + a
(1)
o
H
(1)
o
(
!
c
I
) +
1
2
a
(2)
oo
H
(2)
oo
(
!
c
I
) + ...
_
. (14)
Example: For 1D problems let, c
0
= 0, c
1
= 1 and c
1
= 1. Consider a
second order equilibrium distribution, with = 2
H
(0)
= 1, (15)
H
(1)
r
=
0r
= c
I,r
, (16)
H
(2)
rr
=
2
0r
1 =
2
c
2
I,r
1. (17)
The constants a, \
0
, \
1
are determined by solving
b1

I=0
\
I
_
H
(n)
:n
_
!

0,I
__
2
=
1
(2)
1/2
_
c

2
0
2
_
H
(n)
:n
_
!

0
__
2
d
!

0
, (18)
giving
4
\
0
+ \
1
+ \
1
=
1
(2)
1/2
_
1
1
c

2
0
2
d
0
(19)
\
0
(c
0
)
2
+ \
1
(c
1
)
2
+ \
1
(c
1
)
2
=
1
(2)
1/2
_
1
1
c

2
0
2

2
0
d
0
(20)
\
0
_

2
c
2
0
1
_
2
+ \
1
_

2
c
2
1
1
_
2
+ \
1
_

2
c
2
1
1
_
2
=
1
(2)
1/2
_
1
1
c

2
0
2
_

2
0
1
_
2
d
0
, (21)
and the discrete equilibrium distributions are
)
0;eq
= \
0
_
a
(0)
H
(0)
(
!
c
0
) + a
(1)
o
H
(1)
r
(
!
c
0
) +
1
2
a
(2)
rr
H
(2)
xx
(
!
c
0
)
_
= \
0
:
_
1 + n
0
c
0
+
1
2
(n
0
n
0
+ )
_

2
c
0
c
0
1
_
_
= \
0
:
_
1
1
2
_
n
2
0
+
_
_
, (22)
)
1;eq
= \
1
_
a
(0)
H
(0)
(
!
c
1
) + a
(1)
o
H
(1)
o
(
!
c
1
) +
1
2
a
(2)
oo
H
(2)
oo
(
!
c
1
)
_
= \
1
:
_
1 + n
0
c
1
+
1
2
(n
0
n
0
+ )
_

2
c
1
c
1
1
_
_
= \
1
:
_
1 + n
0
+
1
2
_
n
2
0
+
_ _

2
1
_
_
, (23)
)
1;eq
= \
1
_
a
(0)
H
(0)
(
!
c
1
) + a
(1)
o
H
(1)
o
(
!
c
1
) +
1
2
a
(2)
oo
H
(2)
oo
(
!
c
1
)
_
= \
1
:
_
1 + n
0
c
1
+
1
2
(n
0
n
0
+ )
_

2
c
1
c
1
1
_
_
= \
1
:
_
1 n
0
+
1
2
_
n
2
0
+
_ _

2
1
_
_
. (24)
Example:The D2Q9 lattice (Figure 1)
The Hermite polynomials are
5
Figure 1: The D2Q9 lattice
H
(0)
= 1, (25)
H
(1)
r
=
0r
= c
I,r
, (26)
H
(1)

=
0
= c
I,
, (27)
H
(2)
rr
=
2
0r
1 =
2
c
2
I,r
1, (28)
H
(2)

=
2
0
1 =
2
c
2
I,
1, (29)
H
(2)
r
=
0r

0
=
2
c
I,r
c
I,
. (30)
The constants a, \
0
, \
1
, \
2
are determined by solving
b1

I=0
\
I
_
H
(n)
:n
_
!

0,I
__
2
=
1
(2)
1/2
_
c

2
0
2
_
H
(n)
:n
_
!

0
__
2
d
!

0
, (31)
giving
6
\
0
+ 4\
1
+ 4\
2
=
1
(2)
1/2
_
1
1
c

2
0x
2
d
0r
_
1
1
c

2
0y
2
d
0
(32)
\
0

2
c
2
0,r
+ \
1

2
4

I=1
c
2
I,r
+ \
2

2
8

I=5
c
2
I,r
=
1
(2)
1/2
_
1
1
c

2
0x
2

2
0r
d
0r
_
1
1
c

2
0y
2
d
0
(33)
\
0

2
c
2
0,
+ \
1

2
4

I=1
c
2
I,
+ \
2

2
8

I=5
c
2
I,
=
1
(2)
1/2
_
1
1
c

2
0x
2
d
0r
_
1
1
c

2
0y
2

2
0
d
0
, (34)
\
0
_

2
c
2
0,r
1
_
2
+ \
1
4

I=1
_

2
c
2
I,r
1
_
2
+ \
2
8

I=5
_

2
c
2
I,r
1
_
2
=
1
(2)
1/2
_
1
1
c

2
0x
2
_

2
0r
1
_
2
d
0r
_
1
1
c

2
0y
2
d
0
(35)
\
0
_

2
c
2
0,
1
_
2
+ \
1
4

I=1
_

2
c
2
I,
1
_
2
+ \
2
8

I=5
_

2
c
2
I,
1
_
2
=
1
(2)
1/2
_
1
1
c

2
0x
2
d
0r
_
1
1
c

2
0y
2
_

2
0
1
_
2
d
0
, (36)
\
0
_

2
c
0,r
c
0,
_
2
+ \
1
4

I=1
_

2
c
I,r
c
I,
_
2
+ 4\
2
8

I=5
_

2
c
I,r
c
I,
_
2
=
1
(2)
1/2
_
1
1
c

2
0x
2

2
0r
d
0r
_
1
1
c

2
0y
2

2
0
d
0
, (37)
Only four of the six above equations are linearly independent, giving the
values of \
0
=
16
36
, \
1
=
4
36
, \
2
=
1
36
and the scaling factor =
_
3,2.
The D2Q13 and the next lattices are also able to run second-order models.
In these cases, the number of unknowns is greater than the number of dispos-
able equations, and several solutions will be available, satisfying the quadrature
problem.
Nevertheless, this lattice is not able to run full third-order models. Indeed,
when N=3, it is impossible to nd real positive values for , \
0
, \
1
, \
2
, \
3
7
Figure 2: Lattices D2Q17 and D2Q21
satisfying all the norm restrictions. This result is the same for the D2Q17 lattice
(Figure 2).
Considering the D2Q21 lattice, Figure 2), as a next candidate for third-order
models, there will be, in this case, seven unknowns , \
0
, \
1
, \
2
, \
3
, \
4
.\
5
for six norm restrictions,after eliminating identical equations. Letting be a
free variable, the system gives a solution with real positive roots when is inside
the interval 0.659836 and 1.16208. The values =0.659836 and =1.16208 are
roots of the polynomials W
0
() and W
3
() , respectively. Therefore, when the
value =1.16208 is chosen, W
3
=0 and the lattice loses an energy level, giving a
modication of the D2Q17 lattice, which has been named D2V17. The weights,
with six signicant digits, are W
0
=0.402005, W
1
=0.116155, W
2
=0.0330064,
W
3
=0, W
4
=0.0000790786, and W
5
=0.000258415.
Figure 3: Lattice D2V17 for third order models.
8
References
[1] P. C. Philippi, L. A. Hegele, Jr., L. O. E. dos Santos, and R. Surmas, Phys.
Rev. E 73, 056702 (2006)
[2] X. Shan, X.-F. Yuan, and H. Chen, J. Fluid Mech. 550, 413 (2006)
[3] X. Shan, Phys. Rev. E 81, 036702 (2010)
[4] S. Chapman and T. G. Cowling, The Mathematical Theory of Non-Uniform
Gases, Cambridge University Press (1970).
[5] R. Surmas, C.E. Pico Ortiz and P.C. Philippi, Eur. Phys. J. Special Topics
171, 8190 (2009).
9
Lecture 10: Attributing volume to the molecules
The kinetic equation
0
|
) +

0
r
) =
(sJ)
. .
repulsion
+
(|J)
. .
intermolecular attraction
+

q
(t)

n
_
|T
n
)
(tj)
. .
external body forces
, (1)
does not consider that the molecules have a volume /. The consequence is that
the equation of state is given by
1 = :/T
..
repulsion part
a:
2
..
attraction part
, (2)
without any volume contribution.
In LBM simulation it is important to consider this contribution to avoid the
colapse of mass
1
in certain points in the course of the simulation.
We can, nevertheless, correct the kinetic equation to satisfy the equation of
state we want. Along this line of approach, the repulsion term is corrected by
writting it as

(sJ)
=
(sJ)
b=0
+

1
_

n
_
)MB , (3)
where the correction term

1
_

n
_
)MB is written in such a manner as to
not aect the mass conservation equation, since
_
:

1
_

n
_
)MB d

= 0. (4)
The term
(:t)
b=0
represents the short-range repulsion term when the par-
ticles are considered without volume, / = 0, and must satisfy the following
conservation restrictions,
_
:
(sJ)
b=0
d

= 0, (5)
meaning that the mass is not aected in their interaction,
_
:

(sJ)
b=0
d

= 0, (6)
and
_
1
2
:
2

(sJ)
b=0
d

= 0, (7)
1
The collapse of mass happens in a given point, when the repulsion forces are weak and do
not equilibrate the attraction forces. When this happens, the numerical scheme loss stability.
1
related, respectively, to the conservation of momentum and kinetic energy in
short range interactions of material points.
Multipliying Eq. (3) by :

and integrating, using,

=
_

n
_
+

n , (8)
we get
_
:

(:t)
d

= :1
o
_
)MB (
o
n
o
)
_

o
n
o
_
d

= :/T

1 . (9)
Therefore, without considering the surface terms, the momentum balance
equation becomes
0
|
(j

n ) + 0
o
(jn
o
n
o
) = jq
(t)
o
\
_
:/T a:
2
_
0
o
t
oo
+ :/T

1 . (10)
Writting,

1 =
1
:/T
\(:/T) , (11)
we get
:/T (1 ) a:
2
=
:/T
1 /:
a:
2
, (12)
or
= 1
1
1 /:
=
/:
1 /:
. (13)
Finally, we can write the correction term as

1 =
1
:/T
\(:/T (/:)) , (14)
where
=
1
1 /:
, (15)
is a volume parameter taking the value = 1, when / = 0.
The full kinetic equation becomes then
0
|
) +

0
r
) =
(sJ)
. .
repulsion
+
(|J)
. .
intermolecular attraction
+

q
(t)

n
_
|T
n
)
(tj)
. .
external body forces
, (16)
2
where

(sJ)
=
(sJ)
b=0
. .
repulsion eld contribution

1
:/T
\(:/T (/:))
_

n
_
)MB
. .
volume contribution
, (17)
and

(o||)
=

q
(|J)

n
_
|T
n
)
(tj)
=
_
_
_
_
2a
:
\:
. .
bulk phase
+
i
:
\
_
\
2
:
_
. .
surface forces
_
_
_
_

n
_
|T
n
)
(tj)
(18)
Enskog, [1], was the rst who considered a correction in the Boltzmann
equation for taking the molecular volume / into account. His treatment is
given below, in the appendix and results in Eq. (75). For isothermal problems
Enskogs correction can be written as

(sJ)
=
(sJ)
b=0
. .
repulsion eld contribution

1
:/T
\(:/T (/:))
_

n
_
)MB
. .
volume contribution
(19)
which is the same as Eq. (17), with the volume parameter multiplying the
short range repulsion term.
1 Appendix: Enskogs correction to the short
range term
2
When the distance between the target and the incident molecules is very short,
we consider that the collision term is the Enskog collision term for dense gases.
Molecules are supposed to be rigid spheres with diameter o.
2
Chapter 16 of [1] is recommended as a reference for this part.
3
Considering )
_

r ,

0
, t
_
d

r d

0
to be the expected number of target mole-
cules in the position

r , d

r with velocities

0
, d

0
, the molecules
)
_

r + o

c,

1
0
, t
_
d

0
1
/
0
d-d/
0

0
1

dt (20)
will collide with the target molecules in the time interval dt and will rescue the
target molecules to the state

r ,

, when

0
=

c (

q ) =

c (q cos 0) , (21)

0
1
=

c (

q ) =

c (q cos 0) , (22)
where

q =

, (23)

c = (cos 0, sin0 cos -, sin0 sin-) , (24)

0 = arcsin
_
/
o
_
. (25)
4
The expected number of molecules that will be rescued in time dt to the
state

r ,

, due to such collisions, will be thus

r +
1
2
o

c
_
)
_

r ,

0
, t
_
)
_

r + o

c,

1
0
, t
_
/
0
d-d/
0

0
1

0
d

0
1
d

r dt
(26)
Where is an heuristic correction factor introduced by Enskog to take ac-
count of the nite volume eects of the populations )
_

r ,

0
, t
_
and )
_

r + o

c,

1
0
, t
_
.
Now,
d

0
d

0
1
= d

1
(27)
because the Jacobian of the transformation

,

0
,

0
1
is 1,
/
0
= / (28)
[

q
0
[ = [

q [ (29)
due to the symmetry in collisions with a central interaction potential.
In this manner, the collision term, Eq. (26) will be written as
5

r +
1
2
o

c
_
)
_

r ,

0
, t
_
)
_

r + o

c,

0
1
, t
_
/d-d/qd

1
d

r dt (30)
For the inverse collisions, this term will be written as

r
1
2
o

c
_
)
_

r ,

, t
_
)
_

r o

c,

1
, t
_
/d-d/qd

1
d

r dt (31)
and the short range collision term wil be written as

sJ
=
_ _ _
_
_

_

r +
1
2
o

c
_
)
_

r ,

0
, t
_
)
_

r + o

c,

0
1
, t
_

r
1
2
o

c
_
)
_

r ,

, t
_
)
_

r o

c,

1
, t
_
_
_
q/d/d-d

1
(32)
or, considering that
sin0 =
/
o
(33)
and
6
q/d/d- = o
2
q cos 0
. .
!
.
!
o
sin0d0d-
. .
J
!
o
= o
2
q .

/ d

/ (34)

sJ
=
_ _ _
_
_

_

r +
1
2
o

c
_
)
_

r ,

0
, t
_
)
_

r + o

c,

0
1
, t
_

r
1
2
o

c
_
)
_

r ,

, t
_
)
_

r o

c,

1
, t
_
_
_
o
2
q .

cd

cd

1
(35)
The Enskog correction factor is splited in two factors: a shielding factor,

s
, which reduces the probability of collisions and a collisions enhancement
factor,
t
.
Shielding Factor
s
.
Consider a target molecule with diameter o. An incident molecule with the
same diameter will shield the target molecule against collisions with another
incident molecules when its center O
1
is located at a distance o _ r < 2o from
the center O of the target molecule (Figure 1). The shielded surface is the
surface of a cup on the sphere of radius o centered at O.
When the center of the incident molecule is outside a sphere of radius 2o
there is no shielding. When o _ r < 2o, a portion of the spherical surface of ra-
dius o around the center O will be forbidden to another incident molecules.This
shielded area can be calculated as bellow.
7
Shielding area
d
s|ItlJ.
= o
2
sin,d,dc (36)
cos , =
j
o
(37)
o sin,d, = dj (38)

s|ItlJ
= 2o
_
c
r/2
dj = 2o
_
o
r
2
_
(39)
In a shell of thickness dr around O there are 4r
2
:(r)dr molecules and the
total probable area that is forbidden to the another molecules to collide with
the molecule which center is at O is given by
_
2c
c
2o
_
o
r
2
_
4r
2
:dr =
11
3

2
:o
5
(40)
In this manner since the total area of the sphere o is 4o
2
the probability
of collision of any incident molecule with the traget will be reduced by a factor
8

s
=
4o
2

11
3

2
:o
5
4o
2
= 1
11
12
:o
3
(41)
Enhancement factor
t
.
Each molecule has an associated sphere with volume
4
3
o
3
inside which
the center of the another molecules cannot lie. Considering that there are N
molecules in a given volume V, the remaining volume that is disposable for the
molecules to lie is
\
4
3
o
3
(42)
In this manner, due to this lack of freedom the probability of collisions will
increase by a factor

t
=
1
1
4
3
o
3
:
(43)
and the term giving the collision factor due to the nite volume of molecules
will be written as
=
1
11
12
:o
3
1
4
3
o
3
:
=
1
11
8
/j
1 2/j
= 1 +
5
8
/j + O(j
2
) (44)
9
where
/ =
2
3
o
3
,: (45)
Equilibrium and non-equilibrium solutions
For an uniform uid does not depend on the spatial coordinate

r and Eq.
(35) can be written as

sJ
=
_ _ _
_
_
)
_

r ,

0
, t
_
)
_

r + o

c,

0
1
, t
_

)
_

r ,

, t
_
)
_

r o

c,

1
, t
_
_
_
o
2
q .

cd

cd

1
= 0
(46)
which solution is the Maxwell-Boltzmann distribution,
)
tj
= :
_
:
2/T
_
1/2
exp
_
_
_
_

n
_
2
2|T
n
_
_
_ (47)
Considering then
)
_

r + o

c,

0
1
, t
_
- )
_

r ,

0
1
, t
_
+
0)
_

r ,

0
1
, t
_
0

r
o

c (48)
)
_

r o

c,

1
, t
_
- )
_

r ,

1
, t
_

0)
_

r ,

0
1
, t
_
0

r
o

c (49)

r +
1
2
o

c
_
- (

r ) +
1
2
\

c (50)

r
1
2
o

c
_
- (

r )
1
2
\

c (51)
in Eq(35)

r +
1
2
o

c
_
)
_

r ,

0
, t
_
)
_

r + o

c,

0
1
, t
_
=
_
+
1
2
o\

c
_
)
_

r ,

0
, t
__
)
_

r ,

0
1
, t
_
+\)
0
1
o

c
_
= )
_

r ,

0
, t
_
)
_

r ,

0
1
, t
_
+ (

r ) )
_

r ,

0
, t
_
\)
0
1
o

c
+
1
2
o\

c)
_

r ,

0
, t
_
)
_

r ,

0
1
, t
_
+O(2) (52)
we obtain
10

sJ
= (

r )
_ _
()
0
)
0
1
))
1
) o
2
q .

cd

cd

1
+(

r )
_ _
()
0
\)
0
1
+ )\)
1
)

co
3
q .

cd

cd

1
+
1
2
\(

r )
_ _
()
0
)
0
1
+ ))
1
)

co
3
q .

cd

cd

1
(53)
For evaluating the second term on the right of the above equation, we con-
sider ) = )
tj
and since
)
tj,0
)
tj,0
1
= )
tj
\)
tj
1
(54)
we obtain

_
)
tj,0
\)
tj,0
1
+ )
tj
\)
tj
1
_
= )
tj
\)
tj
1
\)
tj,0
1
)
tj,0
1
+ )
tj,0
)
tj,0
1
\)
tj
1
)
tj
1
= )
tj
\)
tj
1
\ln)
tj,0
1
+ )
tj
\)
tj
1
\ln)
tj
1
(55)
The third term on the right of Eq. (53) gives
1
2
\
_
)
tj,0
)
tj,0
1
+ )
tj
)
tj
1
_
= )
tj
)
tj
1
\ln = (56)
So, the second and third term of Eq. (53) give
)
tj
_ _
)
tj
1
\ln
_
)
tj
1
)
tj
0
1
_

co
3
q .

cd

cd

1
(57)
Now, with =
|
n
,
)
tj
1
= :
_
:
2/T
_
1/2
exp
_

C
2
1
2T
_
(58)
)
tj0
1
= :
_
:
2/T
_
1/2
exp
_

C
02
1
2T
_
(59)
and
)
tj
1
)
tj0
1
= :
2
_
:
2/T
_
1
exp
_

C
2
1
+ C
02
1
2T
_
(60)
so
\
_
ln
_
)
tj
1
)
tj0
1
__
= \
_
ln + ln:
2
+ ln
_
:
2/
_
1
lnT
1

C
2
1
+ C
02
1
2T
_
= \
_
ln
n
2
_
T
D

_
C
2
1
+ C
02
1
_
\
_
1
2RT
_

1
2RT
\
_
C
2
1
+ C
02
1
_
_
(61)
11
but
C
2
1
+ C
02
1
=
2
1
2
1o
n
o
+ n
o
n
o
+
02
1
2

0
1o
n
o
+ n
o
n
o
(62)
and
0
o
_
C
2
1
+ C
02
1
_
= 2
1o
0
o
n
o
+ 2n
o
0
o
n
o
2
0
1o
0
o
n
o
+ 2n
o
0
o
n
o
= 2
_
C
1o
+ C
0
1o
_
0
o
n
o
(63)
Eq. (61) will thus give,
\
_
ln
_
)
tj
1
)
tj0
1
__
= \
_
ln
n
2
_
T
D
+
_
C
2
1
+ C
02
1
_
1
2RT
2
\T
+
1
RT
_

C
1
+

C
0
1
_
\

n
_
(64)
Replacing this result into Eq. (57)

_ _
)
tj
)
tj
1
\ln
_
)
tj
1
)
tj
0
1
_

co
3
q .

cd

cd

1
= )
tj
_ _
)
tj
1
\
_
ln
n
2
_
T
D
+
_
C
2
1
+ C
02
1
_
1
2RT
2
\T
+
1
RT
_

C
1
+

C
0
1
_
\

n
_

co
3
q .

cd

cd

1
(65)
Velocities

C
1
and

C
0
1
are related by
3

C
0
1
=

C
1
(

q

c)

c (66)

C
0
1
+

C
1
= 2

C
1
(

c)

c (67)
_

C
0
1
_
2
=
_

C
1
_
2
2 (

c)
_

C
1

c
_
+ (

c) (

c) (68)
_

C
0
1
_
2
+
_

C
1
_
2
= 2
_

C
1
_
2
2 (

c)
_

C
1

c
_
+ (

c) (

c) (69)
These relations enable
4
Eq. (65) to be integrated in the variable

c, for a
xed

q with

c = q cos 0 (70)
3
See Lecture 1: Ballistics of collisions
4
see Section 16.8 of [1]
12

c = (cos -, sin0 cos -, sin0 sin-) (71)

d

c = sin0d0dc (72)
We can, then, integrate with respect to

1
, using

q =

(73)
with
d

q = d

1
(74)
The Enskog collision term can be nally written as

sJ
=
1ol|:.
/j)
tj
_

_
_

n
_
.
_
\ !
r
ln
_
j
2
T
_
+
1
1+2
_
(
2

1+2
2
_
\ !
r
lnT
_
+
2
1+2
_
2

( : \ !
r

n
+
_
(
2

1+2
2
_
\ !
r
.

n
_
_

_
(75)
where

( =

n
_
2|T
n
(76)
When the short and long range collision terms are considered the Boltzmann
equation will be written as
0
|
) +

.0 !
r
) +

t
.0 !

) =
sJ
+
1
:
0)
_

r ,

, t
_
0

.\ !
r
c
n
(

r ) (77)
References
[1] S. Chapman, T.G. Cowling, The Mathematical Theory of Non-Uniform
Gases, Cambridge (1970)
13
Lecture 10: Discrete kinetic equation in LBM
variables
In the following, the kinetic equation will be written in terms of dimensionless
LBM variables, starting from
0
|
) +
!
.0 !
r
) = . (1)
Consider
!

I
to be a given discrete velocity, t to be a time step and
!
r
to be the length separating any two sites along the direction given by
!

I
, in
such a way that
!
r =
!

I
t and meaning a CourantFriedrichsLewy (CFL)
number equal to 1. Using an Euler scheme for the temporal derivative and an
upwind scheme for the advective term, it results
)(
!
r +
!

I
t, t + t) )(
!
r , t) = c
I
, (2)
with errors of O(
!
r
2
, t
2
). Multiplying this equation by

i
n0
=
Vi
n0
c
m
2
i
2kT
0
_
2t|T0
n
_
1/2
,
the following LBE is obtained,
)

I
(
!
r +
!

I
t, t + t) )

I
(
!
r , t) =

I
t
:
0

I
, (3)
where )

I
(
!
r , t) =

i
n0
)(
!
r , t) =
}i(
!
r ,|)
n0
is the number density of particles with
velocity
!

I
, resident in the site
!
r , at time t, referred to a reference number
density :
0
.
In LBM, after the velocity discretization, dening the lattice speed
c =
/
t
=
_
/T
0
:
, (4)
the particle velocity
!

I
is related to the usual lattice vectors
!
c
I
, through,
!

I
=
!

0,I
_
/T
0
:
= (
!
c
I
)
_
c

_
= c
!
c
I
, (5)
where / is the lattice spacing and is the dimensionless scaling factor, i.e.,
a constant that depends on the lattice, found together with the quadrature
weights \
I
in the solution of the equations giving the norm preservation or
orthogonality conditions and written in capital letters to not confuse with the
force parameter a in the equation of state.
In the same manner, since,
!
n =
1
:

I
)
I
!

I
=
c
:

I
)

I
!
c
I
, (6)
where :

= :,:
0
The macroscopic velocity
!
n is then related to the dimensionless macroscopic
velocity
!
n

through
1
!
n = c
!
n

. (7)
In Eq. (3),

I
t
:
0

I
=

I
t
:
0

(sJ)
I
+

I
t
:
0

(|J)
I
,
where using a BGK model,

I
t
:
0

(sJ)
I
=
_

I
t
:
0
_

)
tj
_
!

I
_
)
I
t

I
t
:
0
_
1
:/T
r(:/T (/:))
_
!

!
n
_
)
tj
_
!

I
_
(8)
Now

I
:
0
)
tj
_
!

I
_
=
\
I
:
0
c
m
2
i
2kT
0
_
2/T
0
:
_
1/2

_
_
1
2
:
/T
0
_
1/2
c

2
0;i
2

n=0
1
:!
a
(n)
:n
(:,
!
n
0
, ) H
(n)
:n
_
!

0,I
_
. .
=}eq

!
i

_
=
1
:
0
_

_
\
I

n=0
1
:!
a
(n)
:n
(:,
!
n
0
, ) H
(n)
:n
_
!

0,I
_
. .
=}eq;i
_

_
=
)
tj,I
:
0
= )

tj,I
(9)
It is important to notice that
)
tj
_
!

I
_
=
number of particles
:
1
(:,:)
1
(10)
)
tj,I
=
number of particles
:
1
(11)
)

tj,I
= dimensionless (12)
In the same manner

I
:
0
)
_
!

I
_
= )

I
(13)
2
Consider now the term
_

I
t
:
0
_
1
:/T
r(:/T (/:))
_
!

!
n
_
)
tj
_
!

I
_
= t
1
:/T
r(:/T (/:))
_
!

!
n
_
)

tj,I
= t
c
:

T
T0
r
_
:

T
T
0
(/:)
_
(
!
c
I

!
n

) )

tj,I
=
/
:

( + 1)
r(:

( + 1) (/:)) (
!
c
I

!
n

) )

tj,I
=
r

(:

( + 1) (/:))
:

( + 1)
(
!
c
I

!
n

) )

tj,I
(14)
Therefore, Eq. (8) becomes

I
t
:
0

(sJ)
I
=
)

tj,I
)

I
t,t

r

(:

( + 1) (/:))
( + 1) :

(
!
c
I

!
n

) )

tj,I
(15)
Consider now

I
t
:
0

(|J)
I
= t
_
2a
:
r: +
i
:
r
_
r
2
:
_
_

_
!

!
n
_
|T
n
)

tj,I
(16)
The force parameter a is the units of [pressure] ,
_
:
2
0

=
|T0
n0
=
n
n0
|T0
n
=
n
n0
_
c
.
_
2
. So,
o
n
=
1
n0
_
c
.
_
2
a

t
_
2a
:
r:
_

_
!

!
n
_
|T
n
)

tj,I
= t
_
c

_
2
c (2a

r:

)
(
!
c
I

!
n

)
|T0
n
T
T0
)

tj,I
= /(2a

r:

)
(
!
c
I

!
n

)
( + 1)
)

tj,I
= 2a

(
!
c
I

!
n

)
( + 1)
)

tj,I
(17)
In the same manner, since the surface parameter is given by
r
n
=
|
2
n0
_
c
.
_
2
i

,
the term
3
t
i
:
r
_
r
2
:
_

_
!

!
n
_
|T
n
)

tj,I
= t
c/
2
:
0
_
c

_
2
i

r
_
r
2
:
_

(
!
c
I

!
n

)
|To
n
T
T0
)

tj,I
= /
3
i

r
_
r
2
:

(
!
c
I

!
n

)
( + 1)
)

tj,I
= i

_
r
2
:

(
!
c
I

!
n

)
( + 1)
)

tj,I
(18)
So, Eq. (16) becomes

I
t
:
0

(|J)
I
=
_
2a

+ i

_
r
2
:

__

(
!
c
I

!
n

)
( + 1)
)

tj,I
(19)
Finally, the Boltzmann equation can be written in LBM variables as
)

I
(
!
r +
!

I
t, t + t) )

I
(
!
r , t)
=
)

tj,I
)

I
t,t

r

(:

( + 1) (/:))
( + 1) :

(
!
c
I

!
n

) )

tj,I
+
_
2a

+ i

_
r
2
:

__

(
!
c
I

!
n

)
( + 1)
)

tj,I
(20)
4