Course notes

Course Blackboard website:

https://bb.sustech.edu.cn/

Reference textbooks: check under File / Textbook

ZHAO Xiaozheng:
A demo: accessing the department computer

CHEN Hao:
Unix commands 常用命令大全(新手入门)
Unix OS & Fortran programming
 Xshell MAE Department
computer (A Unix ssh
computer)
Your Windows Xftp scp Your Mac
computer (1) Edit Fortran Codes Computer
(2) Run Fortran Codes
(3) Postprocessing
VcXsrv (4) ….
X11
Overview of fluid mechanics governing equations

Boundary conditions at fluid-particle and fluid-fluid

interfaces

Two Fortran demo codes

Numerical Fluid
analysis Mechanics

CFD

Computer
Science
The Navier-Stokes Equations

+
(
∂ρ ∂ ρU j
=0
) Two general methods to derive these equations
∂t ∂x j
+ ! .
∂U i ∂U i ∂P ∂ - %' ∂U i ∂U j 2 ! ( V
ρ + ρU j = ρ gi − + µ + − ∇ ⋅ U δij * + µ ∇ ⋅ U δij 0
∂t ∂x j ∂xi ∂x j -, '& ∂x j ∂xi D *
) 0/

µ = shear viscosity, µV = bulk viscosity or volume viscosity

Fluid-dynamics stress [Due to fluid motion]

% ∂U ∂U j 2 ! ( ! % 1 ! ( !
'
σ ij = µ ' i + − ∇ ⋅ U δij * + µV ∇ ⋅ U δij = 2µ ' Sij − ∇ ⋅ U δij * + µV ∇ ⋅ U δij
D * D
& ∂x j ∂xi ) & )

1 %' ∂U i ∂U j (*
D = spatial dimension, Sij ≡ +
2 '& ∂x j ∂xi *)
Total stress τ ij = −Pδij + σ ij

Governing equation valid for both laminar and turbulent flows

Kundu, Cohen, Dowling, Fluid Mechanics
Symmetry property: Galilean invariance

The Navier-Stokes equation is invariant under Galilean transformation

     
t, x,U  t, x + Vt,U + V

where V is a constant velocity

This ensures that the flow solution is independent of the inertial frame of reference

Any physical model (Reynolds stress model, SGS model, etc.) and numerical method
for the N-S equations must comply with the Galilean invariance
Energy equation (1st Law of Thermodynamics in differential form)
Primitive form (1), exactly as 1st Law of thermodynamics
# 1 & # 1 &
∂ % e + U 2 − g k xk ( ∂ % e + U 2 − g k xk (
$ 2 ' $ 2 ' ∂ ) + ∂ #% ∂T &(
ρ
∂t
+ ρU j
∂x j
=
∂xi *
( )
−Pδij + σ ij U j , + k
∂x j %$ ∂x j ('
Primitive form (2)
# 1 & # 1 &
∂% e + U 2 ( ∂% e + U 2 (
$ 2 ' $ 2 ' ∂ ) + ∂ #% ∂T &(
ρ
∂t
+ ρU j
∂x j
= ρ g jU j +
∂xi *
( )
−Pδij + σ ij U j , + k
∂x j %$ ∂x j ('

The Equation of State e.g., ideal gas P = ρ RT

P = f ( ρ, T,...) The Van der Waals equation of state

"1 %
( P + a ρ 2
) # ρ '& = RT
$ − b
Mechanical Energy equation
"1 % "1 %
∂$ U 2 ' ∂$ U 2 '
#2 & #2 & ∂ )
ρ + ρU j = ρ g jU j +U i *−Pδij + σ ij +,
∂t ∂x j ∂x j
Internal energy equation

∂e ∂e ! ! 1 ! $2 V
! 2 ∂
!
∂T
$
ρ
∂t
+ ρU j
∂x j " 3 %
(
= −P∇ ⋅ U + 2µ # Sij − ∇ ⋅ U δij & + µ ∇ ⋅ U ) +
∂x j
#k
# ∂x &
&
" j%
or
! ! $
∂e ∂e ∂ #k ∂T &
ρ + ρU j = −P∇ ⋅ U + ρε +
∂t ∂x j ∂x j# ∂x &
" j%
where the dissipation rate per unit mass (irreversible energy transfer)
! 1 ! $2 µV ! 2
ε = 2ν # Sij − ∇ ⋅ U δij & +
" 3 % ρ
(
∇ ⋅U )
Entropy equation – the 2nd law of thermodynamics
! " %
∂e ∂e ∂ $k ∂T '
ρ + ρU j = −P∇ ⋅ U + ρε +
∂t ∂x j ∂x j $ ∂x '
# j&

Applying Gibbs relation de = Tds − Pd 1 / ρ ( )

" %
Ds ε 1 ∂ $ ∂T '
= + k
Dt T ρT ∂x j # ∂x j '&
$

Then we can write

" %
Ds 1 ∂ $ k ∂T ' k 2 ε
Dt
=
$ '
+
ρ ∂x j # T ∂x j & ρT 2 (
∇T ) +
T
Therefore, the 2nd Law
Entropy production simply requires that
Entropy transfer
due to heat transfer due to finite temperature difference µ ≥0
& viscous dissipation µV ≥ 0
k ≥0
Viscosity of air as a function of temperature

Model 1: The Sutherland equation

𝑏𝑇 %.' 𝑘𝑔
𝜇= 𝑤ℎ𝑒𝑟𝑒 𝑏 = 1.458×1045 𝑆 = 110.4 𝐾
𝑇+𝑆 𝑚 ⋅ 𝑠 ⋅ 𝐾 <.'

Model 2: Power Law model

<.=
𝑇
𝜇 = 𝜇< 𝑤ℎ𝑒𝑟𝑒 𝑇< = 273𝐾,
𝑇<
𝑘𝑔
𝜇< = 𝜇 273 𝐾 = 1.71×104'
𝑚⋅𝑠

9
A demo Fortran code: viscosity.f Compile, run, and check the results:
C Compare two models of viscosity
real mu1, mu2, b, S, mu0, T, TK

b = 1.458e-6
S = 110.4
mu0 = 1.71e-5
T0 = 273. Results
do i = -5,10
T = 10.0*real(i) ! in C
C Convert to K
TK = T + 273.
mu1 = b*TK**1.5/(TK+S)
mu2 = mu0*(TK/T0)**0.7
write(10,100) T,TK,mu1,mu2
C Print out a data table
end do

100 format (2x,2F7.1,2(1PE15.5))

stop
end
Comments: Mesoscopic description of fluids

where molecular interactions are explicitly considered

The Boltzmann equation: mesoscopic kinetics description of fluid flow
! ! ! !
f ( v, x, t ) d 3v d 3 x : the number of molecules with
! ! !
microscopic velocity from v to v + d 3v
! ! !
and spatial location from x to x + d 3 x
! ! ! ! ! ! !
dN ( x, t ) = n ( x, t ) d 3 x = "# ∫ f ( v, x, t ) d 3v $% d 3 x

Bhatnagar-Gross-Krook (1954)
"∂ ∂ Fjext ∂ $ ! ! f − f (eq )
& + vj + ' f ( v, x, t ) = C12 ( f ), BGK model: C12 ( f ) = −
&# ∂ t ∂ xj m ∂ v j '% τ
change due to binary collisions of molecules

The collision operator C12 ( f ) conserves mass, momentum, and energy

! ! ! !
m ∫ f ( v, x, t ) d 3v = ρ ( x, t ) macroscopic density
! ! ! ! !
m ∫ v i f ( v, x, t ) d 3v = ρ ( x, t ) U i ( x, t ) macroscopic momentum density
1 ! !2 ! ! ! ! ! 3
m ∫ v − U f ( v, x, t ) d 3v = ρ ( x, t ) e ( x, t ) =ρ RT macroscopic internal energy
2 2
12
Macroscopic equation can be easily derived from the Boltzmann eqn
! F jext ∂ $
∂
# +v ∂
+ & f v! , x,t
!
= C12 f( ) ( )
j
#∂ t ∂ xj m ∂v j &
" %
! 1 ! ! 2
Taking moments with respect to m, mv ,
2
m v −U → ( )
∂ρ ∂ ρ U j
+
( =0
) The macroscopic models can be derived using the
∂t ∂xj Chapman-Enskog approximation. (To be discussed later)
∂ (ρ Ui )
+
(
∂ ρ U iU j ) = ∂τ ij
+ρ
F jext
Stress is the transport of momentum
∂t ∂xj ∂xj m due to random molecular movement
∂ ( ρ e)
+
(
∂ ρU j e ) = − ∂q j
+ τ ij
∂U i
∂t ∂xj ∂xj ∂xj
where Macroscopic, model Microscopic, exact
! , ∂U ∂U j 2 ! / ! ! !
( V
τ ij = −P + µ ∇ ⋅ U δij + µ ..
- ∂x)j
i +
∂xi
−
D
∇ ⋅ U δij 1 = −m ∫ vi − U i
1
0
( ) ( v j − U j ) f ( v , x,t ) d 3v
∂T 1 ! !2 ! ! !
qi = −k = m ∫ vi − U i ( v − U f v , x,t d 3v
) ( )
∂ xi 2
13
II.2. Boundary conditions
Fluid-solid interfaces
Fluid-fluid interfaces
Boundary condition on solid particles

Solid BCs: no-slip

Why no slip?
It no slip BC always correct?
Interface-resolving simulations for flows laden with solid particles
D~
u
⇢f = u + f~(~
rp + µr2 ~ x, t)
Dt
On the surface of k th particle,
~ (k) + ⌦
u|on the surface of k th particle = V
~ ~ (k) ⇥ ~
r

together with
~ (k) Z ⇣ ⌘
d V
m (k)
= p~n+~ n dA + body force + interaction force
⌧ ·~
~
dt S (k)

~ Z ⇣ ⌘
(k) d⌦
(k)
I = r⇥
~ n+~
p~ ⌧ ·~
~ n dA + ......
dt S (k)

- Energy is exchanged at the scale of particle diameter

- Viscous dissipation associated with the disturbance flows induced by particles
- Mixing by large-scale turbulent motion
- Interactions of disturbance flows 16
Interface-resolving simulations for flows laden with solid particles

Immersed solid method: Kajishima et al.

17
Visualization of vorticity
and particle distribution

Gao, Li, LPW, 2013, Lattice Boltzmann Simulation of Turbulent Flow Laden with
Finite-Size Particles, Computers & Mathematics with Applications, 65:194-210.
Viscous flow through porous media
• Compare flows in different porous media with experimental permeability data
Z

X Y
IBEAD
1
-1
120

100

80

60

Z
40

20

0
0
0
20
20
40
40
60
Y 60
80 80 X
100 100
120 120

randomly packed spherical beads Randomly packed ellipsoids

Blue: fluid, Green: solid sand grains

• Study transport and retention of colloids and nanoparticles

How about a fluid-fluid interface?

Computational Fluid Dynamics I!

Motivation and Goals! The Navier-Stokes Computational F
Motivation and G

Cavitation! Microstructure! equations can be Continuu

Evolving Heterogeneous

Bubbly Flow! applied tophases

of different each
Systems composed
ρ 0

phase, with by aa
and materials,
separated V! f
n!
different
whose set
sharp interface
location of f
x ( s,t f

changes with time!

fluid properties ρ , µ , k ,… 1 1 1

(density, viscosity, ρ , µ , k 2 2

conductivity,
Atomization! Splash! School of fish!
specific heat, etc.) Phase 1!

How about interfaces?

Gretar Tryggvason, 2010
Computational Fluid Dynamics I! Computational F
Motivation and Goals! Motivation and G
Boundary conditions at a fluid interface (I)
Typically, the interface is only of a thickness of a few molecules
so interface is viewed as a zero-thickness surface of a given surface tension γ
!
Interface equation: F( x,t ) = 0

Kinematic condition [Assume no mass diffusion / phase change at the interface]

u j ⋅ n j = û j ⋅ n j

[where the surface normal n points from phase 1 (u j ) to phase 2 (û j )]
 ∇F
n =±
∇F

Dynamic condition If viscous fluids on both side of interface, then no tangential slip
ui − (u j ⋅ n j ) ni = ûi − (û j ⋅ n j ) ni
A solid surface can be
viewed as a special case
Interface evolution ∂F 
+ u I ⋅ ∇F = 0 [The level set equation]
∂t
What is surface tension?
The intermolecular energy per unit surface area:
𝑆𝑖𝑧𝑒 𝑜𝑓 𝑚𝑜𝑙𝑒𝑐𝑢𝑙𝑒𝑠 = 𝑎
𝐶𝑜ℎ𝑒𝑠𝑖𝑜𝑛 𝑒𝑛𝑒𝑟𝑔𝑦 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑡𝑤𝑜 𝑚𝑜𝑙𝑒𝑐𝑢𝑙𝑒𝑠 = 𝑈~𝑅𝑇
1
𝑒𝑠𝑡𝑖𝑚𝑎𝑡𝑒 𝑜𝑓 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑚𝑜𝑙𝑒𝑐𝑢𝑙𝑒 𝑝𝑎𝑖𝑟𝑠 𝑝𝑒𝑟 𝑢𝑛𝑖𝑡 𝑠𝑢𝑟𝑓𝑎𝑐𝑒 𝑎𝑟𝑒𝑎 =
𝑎R
Then we have an estimate
1 1 𝑅𝑇
𝑠𝑢𝑟𝑓𝑎𝑐𝑒 𝑡𝑒𝑛𝑠𝑖𝑜𝑛 𝛾 = 𝑈 =
2 𝑎R 2𝑎R

The ½ is added as cutting though (ripping apart) a surface will create two surfaces

𝐽 1𝑚𝑜𝑙𝑒
𝑅𝑇 8.314 𝑚𝑜𝑙. 𝐾 × 6.022𝐸23 ×300𝐾 4×104R% 𝑚𝑁
𝑠𝑢𝑟𝑓𝑎𝑐𝑒 𝑡𝑒𝑛𝑠𝑖𝑜𝑛 𝛾~ R ~ ~ ~26
2𝑎 2×2.75R ×104R< 2×7.56×104R< 𝑚

The size of water molecule is estimated as 2.75 A = 2.75E-10.

The actual value of air-water surface tension is 72 mN/m.

https://www.fun-mooc.fr/ 22
Actual values of surface tension

23
Boundary conditions at a fluid interface (II)
More dynamic conditions
[Sharp-interface formulation]

Local force balance across the interface of area A

!
∫∫ (τ ij − τˆij ) n j dA + #∫ γ t dℓ = 0 where τ ij = −Pδij + σ ij
C

Apply the Stokes theorem

!
∫∫ (τ ij − τˆij ) n j dA + %& ∫∫ ∇ Sγ dA − ∫∫ γ n (∇ ⋅ n ) dA'( = 0
i
!
where ∇ S ≡ ∇ − ni ( n ⋅ ∇ )
! 1 1
Note (−∇ ⋅ n ) = κ = + , R1 and R2 are the principal radii of curvature
R1 R2

Notes:
Finally 1. Here the curvature is negative if the interface is concave and positive if the
! interface is convex, when viewed from the side of F < 0.
(τ ij − τˆij ) n j + ∇ Sγ − γ ni ( ∇ ⋅ n ) = 0 2. Both the radii of curvature and curvature it self can be positive or negative.
A little more differential geometry
Let an interface in 2D is defined as y = h ( x)
Define F ( x, y ) = y − h ( x )
Then the interface is F ( x, y ) = 0 interface is a contour line of F ( x, y )
! !
! ∇F ( ) = x( ) + j
x, y −h x i $ dh '
The normal of the interface is: n= h
&% x ≡ )(
∇F ( x, y ) 1+ ( h )
2 dx
x

It follows that the divergence is The normal points into the region above the
, 0 , 0 curve.
! ∂ . −hx ( x ) . ∂ . 1 .
∇⋅n = - 1 + - 1
∂x . 1+ ( h ) 2 . ∂y . 1+ ( h ) 2 .
/ x 2 / x 2
2 1 2hx hxx
−hxx 1+ ( hx ) + hx
2 1+ ( h ) 2
x
= 2
+0
1+ ( hx )
−hxx
= 3/2
$1+ ( h ) 2 '
% x (
25
The curvature !
dt
The curvature is written as κ=
ds
!
t = tangent unit vector= ( cosφ, sin φ )
φ = angle between the tangent and the horizontal (x axis)
!
dt ! d cos φ d sin φ $ ! dφ dφ $ dφ dφ
κ= =# , & = # −sin φ , cos φ &= sin 2 φ + cos 2 φ =
ds " ds ds % " ds ds % ds ds
dφ
We shall now define κ ≡
ds
dy Positive radius of
φ = tan −1 = tan −1 hx
dx curvature
dφ d tan −1 hx hxx
= = 2
dx dx 1+ ( hx )
x x
ds d 2 2 d 2 2
= ∫ ( dx ') + ( dy ') = ∫ 1+ ( hx (ζ )) dζ = 1+ ( hx ( x ))
dx dx 0 dx 0

dφ dφ / dx hxx
⇒ = =
ds ds / dx +1+ h 2 -3/2
, ( x) . 26
Sketch
The curvature is defined as
dφ
κ=
ds

− +

The right-hand rule applies

The relationship between the curvature
and the divergence of the normal
The curvature is written as 𝜙 is the angle of the tangential of
dφ the interface w.r.t., say, the x axis.
κ=
ds

dφ dφ / dx hxx
⇒ = =
ds ds / dx "1+ h 2 $3/2
# ( x) %

hxx !
κ= 3/2
= −∇ ⋅ n
"1+ ( h ) 2 $
# x %

!
κ = −∇ ⋅ n

28
2D example: A circle
F ( x, y ) = R 2 − ( x 2 + y 2 )
Note that F ( x, y ) > 0 inside
! ∇F ( x, y ) (−x, −y) = (−x, −y) = − cosθ , −sin θ
n= = ( )
∇F ( x, y ) 2
x +y 2 R
! ∂ x ∂ y
−∇ ⋅ n = +
∂x x 2 + y2 ∂y x 2 + y2
1 1 2x2 1 1 2 y2
= − 3/2
+ −
2
x +y 2 2 (x + y )
2 2 2
x +y 2 2 ( x 2 + y 2 )3/2
1 1
= = =κ
x 2 + y2 R

Notes: In this case, the curvature is positive since the

interface is convex, when viewed from the side of F < 0.
2D example: A ellipse
" x 2 y2 %
F ( x, y ) = 1− $ 2 + 2 '
#a b &
Note that F ( x, y ) > 0 inside
" x y %
$− 2 , − 2 '
! ∇F ( x, y ) # a b &
n= =
∇F ( x, y ) 2
" x % " y %
2 Again, the curvature is positive since the interface is
$ 2 ' +$ 2 ' convex, when viewed from the side of F < 0.
#a & #b &
x y
! ∂ a2 ∂ b2
−∇ ⋅ n = +
∂x 2 2 ∂y 2 2
" x % " y % " x % " y %
$ 2 ' +$ 2 ' $ 2 ' +$ 2 '
#a & #b & #a & #b &
1 1 x 2 y2 x2 y2
+ − − +
= a2 b2 + a6 b6 = b2 a 4 b4a2
2 2 3/2 3/2
" x % " y % +" x %2 " y %2 . +" x %2 " y %2 .
$ 2 ' +$ 2 ' -$ 2 ' + $ 2 ' 0 -$ 2 ' + $ 2 ' 0
#a & #b & -,# a & # b & 0/ -,# a & # b & 0/
! 1
[−∇ ⋅ n ] ( a, 0) = 2 → the local radius of curvature = b 2 / a
b /a
3D example: A spheroid
x 2 y2 z2
F ( x, y, z ) = 1− − − =0
a2 a2 c2
Find the curvature at location (a, 0, 0)
location (0, 0, c)

Note that F ( x, y, z ) > 0 inside

# 2 x 2 y 2z &
∇F ( x, y, z ) = % − 2 , − 2 , − 2 (
$ a a c ' Again, the curvature is positive since the interface is
convex, when viewed from the side of F < 0.
# a2 z &
% −x, −y, − 2 (
! ∇F ( x, y, z ) $ c '
n= =
∇F ( x, y, z ) # a2 z &
2

x 2 + y2 + % 2 (
$ c '
! ∂ x ∂ y a2 ∂ z
−∇ ⋅ n = + +
∂x # a2 z &
2 ∂y # a2 z &
2 c 2 ∂z # a2 z &
2
2 2 2 2 2 2
x + y +% 2 ( x + y +% 2 ( x + y +% 2 (
$ c ' $ c ' $ c '
3 2
# a2 & 2 # a2 & 2 # a2 z &
a2 x 2
+ y 2
+ % ( z %1+ ( ( x + y 2
) %$ c 2 ('
+ 2
2+ 2 2
c2 '
c $c ' $
= − 3/2
=
2 &2 + 2 &2 . + 2 3/2
2 2
# a z 2 2
# a z 2 2
# a2 z & .
x + y +% 2 ( -x + y + % 2 ( 0 -x + y + % 2 ( 0
$ c ' -, $ c ' 0/ -, $ c ' 0/
 3D example: A spheroid
Find the curvature at location (a, 0, 0)
location (0, 0, c)

2
$ a2 ' 2 $ a2 z '
&1+ 2 ) ( x + y ) + 2 & 2 )
2

! % c ( % c (
[−∇ ⋅ n ] = 3/2
* $ 2 '2 -
, x 2 + y 2 + & a 2z ) /
,+ % c ( /.

$ a2 '
&1+ 2 )
! % c ( 1 1
[−∇ ⋅ n ] (a, 0, 0) = = + 2 the two principal radii of curvature are a and c 2 / a
a a c /a

! 2 1 1
[−∇ ⋅ n ] (0, 0, c) = = +
a2 / c a2 / c a2 / c
Boundary conditions at a fluid interface (II)
More dynamic conditions
[Sharp-interface formulation]

Local force balance across the interface of area A

!
∫∫ (τ ij − τˆij ) n j dA + #∫ γ t dℓ = 0 where τ ij = −Pδij + σ ij
C

Apply the Stokes theorem

!
∫∫ (τ ij − τˆij ) n j dA + %& ∫∫ ∇ Sγ dA − ∫∫ γ n (∇ ⋅ n ) dA'( = 0
i
!
where ∇ S ≡ ∇ − ni ( n ⋅ ∇ )
! 1 1
Note (−∇ ⋅ n ) = κ = + , R1 and R2 are the principal radii of curvature
R1 R2

Finally
!
(τ ij − τˆij ) n j + ∇ Sγ − γ ni ( ∇ ⋅ n ) = 0
Boundary conditions at a fluid interface (III)
More dynamic conditions
The normal stress balance
!
( P̂ − P ) + (σ ij − σˆ ij ) n j ni − γ ( ∇ ⋅ n ) = 0
For still fluid
! $1 1 '
( ) (
P̂ − P − γ ( ∇ ⋅ n ) = P̂ − P + γ & +
R
% 1 R
)
)=0
2 (
[The Young-Laplace equation]

The tangential stress balance

(σˆ ij − σ ij ) n j ti = ∇ Sγ ⋅ ti
jump of shear stress across the interface = surface tension gradient
[Thermocapillary flow]

Thermal interface BCs can also be developed as needed.

 Example: a stationary bubble of radius R submerged in a static fluid

The curvature of the spherical surface is computed as

! ! 1 ∂ 2 2
∇ ⋅ n = ∇ ⋅ er = 2
r ∂r
( r ) =
R
r=R

so the normal stress jump condition is

2γ
( p − p̂) = κγ =
R

In the above, we used the divergence formula in spherical coordinates:

! 1 ∂ 2 1 ∂ 1 ∂
∇⋅F = 2
r ∂r
( r Fr ) +
r sin θ ∂θ
( Fθ sin θ ) +
r sin θ ∂φ
( Fφ )

Small bubbles are louder than large bubbles when they burst at a free surface.
 Example 2: static planar meniscus at an air-water interface (1)
z
! !
z=η(x) n g
θ x
The free surface is defined as z = η (x)

Governing equation
!
( p − p̂) = γ ∇ ⋅ n
The surface equation is
Note that the pressure S(x, z) = z − η ( x ) S < 0 below, S > 0 above
below the surface is less
The pressure below the interface p = p0 − ρW gz
than the pressure above the
interface, so the pressure The pressure above the interface p̂ = p0 − ρ A gz
force is pushing down. p0 is the reference pressure at x → ∞ and z = 0
This downward force is So at a location x, p − p̂ = − ( ρW − ρ A ) gz ≈ − ρW gz
balanced by the surface
tension force. Notes: In this case, the curvature is positive since the interface is convex, when viewed
from the side of S < 0.
Example 2: static planar meniscus at an air-water interface (2)
! ∇S (−η x ,1)
n= =
∇S 1+ η x2
! ∂ ηx ∂ 1 η xx 1 2η xη xx −η xx
κ = ∇⋅n = − + =− + ηx 3/2
+ 0 = 3/2
∂x 1+ η x2 ∂z 1+ η x2 1+ η x2 2 (1+ ηx2 ) (1+ ηx2 )
η xx η xx
− ρW gη = −γ → ρW gη = γ
2 3/2 3/2
(1+ η x ) (1+ ηx2 )
Boundary conditions
η x = − cot (θ ), η ( x → ∞) = 0
In general, this nonlinear system has to be solved numerically.

If we assume that the slope is small, η x << 1, then

ρW gη ≈ γ η xx
We have
η ≈ lc cot (θ ) e− x/lc
γ For air-water interface
where lc ≡ the capillary length
ρW g γ 0.072N / m
lc ≡ = = 0.27 cm
ρW g 1000kg / m 3 9.81m / s 2