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
 An experiment in a glass cup in the office:

How can I use this to teach surface tension effects?

Lian-Ping Wang
July 5, 2017
Poor man’s experiments in the office:
(1) Two layers of water sandwich an air layer
(2) A hanging drop below a wired water surface!

A pendant drop
A pendant drop
Pitts E, Inst. Maths Applies (1976) 17, 387-397. "The Stability of a Drop
Hanging from a Tube"; cited in "The chaotic dripping faucet." P. Coullet, L.
Mahadevan and C. Riera who give a nice dimensionless ODE for the shape

- Growth of micro-organisms
- Wire mesh is used to seal the water in a faucet when closing

P Coullet and L Mahadevan, and C.S. Rivera, Hydrodynamical

models for the chaotic dripping faucet. J. Fluid Mech. (2005), vol.
526, pp. 1–17.
Formulating the problem

Physical assumptions:
(1) Assume no flow;
(2) Surface tension is constant – consistent with (1);
(3) Axisymmetric;
(4) Neglect pressure change in the air;

Set up the mathematical formulation and boundary conditions

Figure out a solution strategy (Numerical? How?)

Code it

Examining the results

……
6
Example 3: The hanging droplet problem

z !
g
p0
θ x
! z=η(r)
n p0

Note that the pressure

above the surface is The free surface is defined as z = η r()
larger than the pressure
below the interface, so Governing equation
the pressure force is !
pushing down. This ( )
p − p̂ = γ ∇ ⋅ n
downward force is The surface equation is
balanced by the surface
tension force. S( r, z ) = z − η r () S < 0 below, S > 0 above
 The pressure below the interface p = pcompressed + ρ A gz
Theoretical formulation The pressure above the interface p̂ = p0 + ρW gz
air

p0 is the reference pressure at x → ∞ and z = 0

So at a location x, p − p̂ = !" pcompressed air − p0 #$ − ρW − ρ A gz ≈ Δpair − ρW gz
( )
∂r x ∂r y ! $
r = x2 + y2 , = , = , S = z − η #r = x2 + y2 &
∂x r ∂y r " %
! x y $ ! x y $
# −ηr ,−ηr ,1& # −ηr ,−ηr ,1&
! ∇S " r r % " r r %
n= = =
∇S ! x2 y2 $ 1+ ηr2
1+ ηr2 ## 2 + 2 &&
"r r %

! x$ ! y$
#ηr & #η x & 1 2
! ∂ " r% ∂ " r% ∂ 1
ηrr + ηr (ηr ) ηrr
κ = −∇ ⋅ n = + − = r −
∂x 1+ η 2 ∂y 1+ η 2 ∂z 3/ 2
1+ ηr2 1+ ηr2
r r (1+ η ) 2
r

! 1$ 2
1 1
" r% ( 2
#ηrr + ηr & 1+ ηr − ηr ) ( )η rr ηrr + ηr
r
+ ηr ( )
3

r η xx
= =
3/ 2 3/ 2 3/ 2
(1+ ηr2 ) (1+ ηr2 ) (
1+ η x2 )
ηrr + ηr
1
+ ηr ( )
3 1 axisymmetric planar
Δpair − ρW gη = γ r r
3/ 2
(1+ η )
2
r

Boundary conditions
ηr ( r = Rmax ) = tan (θ ) , η ( r = 0) = 0
In general, this nonlinear system has to be solved numerically.
Observations Rmax = The maximum radius on the top
1 3 1 hmax the overall height of the hanging droplet
ηrr + ηr
r
+ ηr ( ) r
Δpair − ρW gη = γ
3/ 2
R0 = the local radius of curvature at the bottom
(
1+ ηr2 )
Boundary conditions
ηr ( r = Rmax ) = tan (θ ) , ηr ( r = 0) = 0
In general, this nonlinear system has to be solved numerically.

2
At the tip region, we have !" R0 − η r − η0 #$ + r 2 ≈ R02
( () )
! $
1 r2 1 r4 & 1 r2
η r = η0 + R0 − R02 − r 2 ≈ η0 + R0 − R0 #1−
() − ≈ η0 +
# 2 R 2 8 R 4 &% 2R
" 0 0 0
1 r 1 1 31 1
⇒ ηr = 0, ηr =
r R r R
= , ηr
r
= 0, ηrr =
R
( )
0 0 0
1 3 1
ηrr + ηr
r
+ ηr ( ) r = η +η 1 + η 3 1 1 1 2
3/ 2 rr r
r r ( ) =
r R
+
R0
+0 =
R0
expected from physics!
( 1+ ηr2 ) 0

2γ 2γ 2γ − Δpair R0
⇒ Δpair − ρW gη = , Δpair + ρW ghmax = , hmax =
R0 R0 ρW gR0
Numerical solution (I)
2γ
Δpair + ρW ghmax =
R0
ρW ghnew = Δpair + ρW ghmax
del(Pair) does not
1 3 1
have net effect so I Δpair − ρW gη = γ
ηrr + ηr
r
+ ηr ( ) r
set to zero from (
1+ ηr2 )
3/ 2

now on. Boundary conditions

ηr ( r = Rmax ) = tan (θ ) , ηr ( r = 0) = 0
1 1 3 1 1
when r → 0, ηr = 0, ηr =
r R0
, (ηr ) r
= 0, ηrr =
R0

We can also set ζ = hmax + η , then

r : 0 → Rmax , ζ :0 → hmax
1.5
2 ! $3 % % (2(
d ζ 1 dζ 1 dζ 1 dζ *
2
+ + # & = !"Δpair + ρW g hmax − ζ #$'1+ ' * ( )
dr r dr r " dr % γ ' & dr ) *
& )
ζ r ( r = 0) = 0, ζ r ( r = Rmax ) = tan (θ )
1 1 3 1 1
when r → 0, ζ r = 0, ζ r =
r R0
, (ζ r ) r
= 0, ζ rr =
R0
Numerical solution (II)
r : 0 → Rmax , ζ :0 → hmax
1.5
3 ! ! 2$
d 2ζ 1 dζ 1 ! dζ $ ρ g #1+ # d ζ $
+ + # & = W h0 − ζ ( ) & &
dr 2 r dr r " dr % γ # " dr % &
" %
dζ d 2ζ dH r ()
Simplification: let H r ≡ () dr
⇒
dr 2
=
dr
dH r ( ) = −1 H ρW g
( r ) − 1r H ( r )
3 1.5

dr r
+
γ
(
(h0 − ζ ) 1+ H ( r ) )
dζ
dr
=H r ()

As a matching problem in r from r = 0:

2γ
()
1 assume hmax → R0 =
ρW ghmax
1 1 1 3 1
(2) start matching with (
H r = 0 = 0, ) r
()
H r =
R0
,
r
()
H r = 0, ζ rr =
R0

()
dζ r
() ()
3 march with very small dr, till ζ r = hmax , record the contact angle θ = arctan
dr
The final algorithm Code inputs:
dH r ( ) = −1 H ρW g hmax and physical parameters
( r ) − 1r H ( r )
3 1.5

dr r
+
γ
(
(h0 − ζ ) 1+ H ( r ) )
Code outputs:
dζ
=H r () (1) Shape
dr
(2) Base diameter
As a matching problem in r from r = 0: (3) Base Contact angle
2γ
()
1 assume hmax → R0 =
ρW ghmax
1 1 1 3 1
(2) start matching with ζ ( r = 0) = 0, (
H r = 0 = 0, ) r
H r =()
R0
,
r
()
H r = 0, ζ rr =
R0
1 1
( )
H r0 = 0,
r
H r0 =( )
R0
r = ri + 0.5dr
&
( 1 1 3 ρ g 1.5 *
(
H ri+1 = H ri + dr '− H ri − !" H ri #$ + W
( ) ( ) ( )h0 − ζ ( ) ( )(
1+ H ri ( )) +
( r
) r γ (
,
1 !
ζ ( ri+1 ) = ζ ( ri ) + dr " H ri+1 + H ri #$
( ) ( )
2
()
dζ r
() ()
3 march with very small dr, till ζ r = hmax , record the contact angle θ = arctan
dr
A demo Fortran code: shape.f About 60 lines long
pi = 4.*atan(1.0) ! pi=3.14159.... 101 continue ! The Marching steps
g=9.8 ! gravity term1 = hh0/r
rho = 1000.0 ! water density term2 = hh0**3/r
gamma = 0.072 ! surface tension in N/m hh1 = hh0 + dr*( - term1 - term2
hmax = 0.005 ! in m . + g*rho/gamma*(hmax-zeta0)*(1.+hh0)**1.5 )
Rtip = 2.0*gamma/(rho*g*hmax) zeta1 = zeta0 + 0.5*dr*(hh0+hh1)
zeta0 = 0.0 write(10,*)r,zeta1,hh1,term1,term2
hh0 = 0.0 if(zeta1.gt.hmax)then
dr = 0.000001 theta = atan( (zeta1-zeta0)/dr )/pi*180.
c The first step go to 100
r0 = 0.0 end if
term1 = 1./Rtip r = r + dr
term2 = 0.0 r0 = r
hh1 = hh0 + dr*( - term1 - term2 hh0 = hh1
. + g*rho/gamma*(hmax-zeta0) ) zeta0 = zeta1
zeta1 = zeta0 + 0.5*dr*(hh0+hh1) go to 101
r = r0 + dr 100 continue
write(10,*)r,zeta1,hh1,term1,term2 write(10,*)'the base diameter in mm:', 2.0*r0*1000.
r0 = r write(*,*) 'the base diameter in mm:', 2.0*r0*1000.
hh0 = hh1 write(10,*)'theta in degree = ',theta
zeta0 = zeta1 write(*,*)'theta in degree = ',theta
c stop
end
Compile, run, and check the results:
A demo Fortran code: shape.f90 101 continue ! The Marching steps
term1 = hh0/r
real pi, g, rhi, gamma, hmax, Rtip, zeta0, dr term2 = hh0**3/r
real r0, term1, term2, hh0, hh1, zeta1, r, theta hh1 = hh0 + dr*( - term1 - term2 &
+ g*rho/gamma*(hmax-zeta0)*(1.+hh0)**1.5 )
pi = 4.*atan(1.0) ! pi=3.14159.... zeta1 = zeta0 + 0.5*dr*(hh0+hh1)
g=9.8 ! gravity write(10,*)r,zeta1,hh1,term1,term2
rho = 1000.0 ! water density if(zeta1.gt.hmax)then
gamma = 0.072 ! surface tension in N/m theta = atan( (zeta1-zeta0)/dr )/pi*180.
hmax = 0.005 ! in m go to 100
Rtip = 2.0*gamma/(rho*g*hmax) end if
zeta0 = 0.0 r = r + dr
hh0 = 0.0 r0 = r
dr = 0.000001 hh0 = hh1
! The first step zeta0 = zeta1
r0 = 0.0 go to 101
term1 = 1./Rtip 100 continue
term2 = 0.0
hh1 = hh0 + dr*( - term1 - term2 & write(10,*)'the base diameter in mm:', 2.0*r0*1000.
+ g*rho/gamma*(hmax-zeta0) ) write(*,*) 'the base diameter in mm:', 2.0*r0*1000.
zeta1 = zeta0 + 0.5*dr*(hh0+hh1) write(10,*)'theta in degree = ',theta
r = r0 + dr write(*,*)'theta in degree = ',theta
write(10,*)r,zeta1,hh1,term1,term2
stop
r0 = r end
hh0 = hh1
(1) All variables must be assigned a type
zeta0 = zeta1
(2) All explanation lines start with a !
(3) The line before the continuation line ends with a &
(4) More freedoms in placing the code on each line
 Compile, run, and check the results:

More information, please read “Fortran90 for Fortran77 Programmers” at

https://www.star.le.ac.uk/~cgp/f90course/f90.html

Fortran77
Fortran90
Fortran95
Fortran2000
Fortran77 vs Fortran90 What's New in Fortran90
What was wrong with Fortran77? • Free-format source code and many other
simple improvements.
• There are no dynamic storage facilities at all.
• Arrays as first-class objects, whole-array
• There are no user-defined data types or data
structures (except the COMMON block). expressions, assignments, and functions.
• It was too easy to make mistakes which the • Dynamic memory allocation; pointers to allow
compiler could not detect, especially when calling complex dynamic data structures to be
constructed.
procedures (subroutines or functions). A study of
• User-defined data types; existing operators
some 4 million lines of professional Fortran showed
» 17% of procedure interfaces were defective. can be overloaded (re-defined) or new ones
• Programs are less than 100% portable because of defined.
a few remaining platform-dependent features. • The MODULE - a new program unit which can
encapsulate data and a related set of
• Control structures are poor, so it hard to avoid
procedures (subroutines or functions). Can
using GOTOs and labels, often leading to spaghetti
code. implement classes and member functions for
• Archaic features left over from the punched-card object-oriented programming.
era: • New control structures: SELECT CASE,
CYCLE, and EXIT so labels and explicit jumps
Ø fixed-format lines,
are rarely needed.
Ø statements all in upper-case,
Ø variable names limited to 6-characters. • Recursive functions, generic names for
• In practice programmers are driven to use procedures, optional arguments, calls with
extensions to Fortran77 Standard - which reduce keywords, and many other procedure call
options.
portability.

https://www.star.le.ac.uk/~cgp/f90course/f90.html
 Choice of programming language?

Fortran. [ gfortran ]

C [ cc ]
I have all of them on my laptop.
Python

Java

We made a choice of using Fortran in this class.

It would be good that you learn at least two in your life.
18
Shape: Theory using my code Shape: poor man’s
experimentà

height base diameter contact angle

Observed
(mm) (mm) (degree)
1.0 11.74 12.60 base
2.0 10.88 25.55 diameter ~
5.0 10.54 43.26 12 to 13 mm
8.0 11.42 48.51
10.0 12.06 50.85
Edge slope is not
monotonic as seen
in the poor man’s
experiments
 An interesting observation

Physics / assumptions

Mathematical formulation à equations

Solving the equations numerically

Examining the results (plotting)

Comparing the physical results with observations

21
Summary
The simple model does provide reasonable predictions:
(a) The base size predicted is 10.5 to 12 mm, compared to 12 to 13 mm in the experiments.
(b) The base size has a minimum (10.5 mm) at certain height (5mm).
(c) The edge slope is non-monotonic

The wire mesh can alter local contact angle and affect the drop size.

The diameter of the drop scales with sqrt(4 gamma / rho g), which gives 5mm diameter. 5 mm is the
typical largest size of rain drops. In our experiment, we can alter gamma, rho, g to alter the size.

The air film pressure can effectively reduce the gravitational acceleration (namely, the differential
hydrostatic pressure is proportional to the density difference), one can try to design such a device.
This can be a little invention that can be used to produce larger drops, or control the size of drops.
Interface-resolving simulations for flows laden with liquid / gas particles

(A) N-S equations for phase 1

(B) N-S equations for phase 2
(C) Interface evolution equation

Initial conditions

Interface BCs: normal velocity, tangential velocity,

normal and tangential stress balance equation

Solid wall BCs: no-slip

23
A more efficient formulation: single-field or one-fluid formulation
[Sharp-interface one-fluid description] !
Identify each fluid by a marker function H ( x, t)
[Heaviside function]
!1 in fluid 1
H ="
# 0 Otherwise
The marker moves with the fluid
∂H !
+ U ⋅ ∇H = 0
∂t Tryggvason et al. JCP 169, 708-759
Define on-fluid fields: There are some subtle details
U" j = HU j + (1− H )Û j , P" = HP + (1− H )P̂ concerning the numerical realization
ρ" j = H ρ j + (1− H ) ρ! j , µ! j = H µ j + (1− H )µ̂ j of Heaviside and delta functions.
A single field equation is solved
[assume incompressible fluid, and constant surface tension γ ]
∂U! j
∂x j
=0 But sharp-interface cannot
be numerically resolved!
∂U! i ! ∂U! i ∂P! ∂ . ( ∂U! i ∂U! j +1 ! ! ! !
ρ!
∂t
!
+ ρU j
∂x j
!
= ρ gi − + 0µ! **
∂xi ∂x j /0 ) ∂x j
+ --3 + ∫ γ n (−∇ ⋅ n ) δ ( x − x f ) ds
∂xi ,32 S (t )
Then the details of delta
fcn implementation are
S(t) is the bounding contour of a differential interface
important.
This is the starting point for several numerical methods, including MAC
(marker-and-cell), VOF, level sets, and front tracking 24
Properties of the one-fluid formulation

Integrating each term across a small control volume centered at

the interface
DU! i ∂P! ∂ * $ ∂U! i ∂U! j '- ! !
∫ ρ! Dt dV = ∫ ρ! gi dV − ∫ ∂x dV + ∫ ∂x ,,µ! && ∂x + ∂x ))// dV + ∫ γ n (−∇ ⋅ n ) δ (n) dV
δV δV δV i δV j + % j i (. δV
! ! !
0 = 0 − ΔPn + Δ (σ ij n j ) + γ n (−∇ ⋅ n )

And the Navier-Stokes equations are satisfied within each phase

25
Fundamental forces in multiphase flow
Volume forces: e.g., inertial force, buoyancy force, gravity force ~ L3

Surface forces: e.g., viscous stress, pressure force ~ L2

Line forces: e.g., surface tension force ~ L

Non-dimensional groups: typical for single-phase flows

inertial force ρ LU
Reynolds number Re = =
viscous force µ
pressure gradient Δp
Euler number Eu = =
inertial force ρU 2
inertial force U 2
Froude number Fr = =
gravity gL

If the flow has an intrinsic or external time scale τ

the time scale Uτ
Strouhal number Sl = = e.g., Flow around a cylinder
convective time scale L
 large à surface tension not important
Flows with fluid-fluid interfaces Small à surface tension is important

inertial force ρ LU 2
Weber number W e = =
surface tension force γ
buoyancy force Δρ gL2
Eotvos number Eo = =
surface tension force γ
= Bo Bond number
viscous force µU ρνU
Capillary number Ca = = =
surface tension force γ γ
ρd
density ratio Γ ρ =
ρc
µd
viscosity ratio Γ ρ =
µc Δρ
Velocity scale due to buoyancy ~ gL
ρ0
[e.g., rise velocity of a gas bubble]
Water droplets in air
Assume sphere
Drag ≈ weight
!1 ! 2$
2 $# π D & 1
C D # ρaW & ≈ ρW π D 3g
"2 % #" 4 &% 6

ρW 4 ! WD $
W ≈ Dg C D ~ 0.44 #103 < Re p = < 105 &
ρa 3C D " υ %
ρW 4 g ρW
W ≈ Dg , ρa = 0.0012 , = 840, υ = 0.15cm / s, g = 980cm / s 2
ρa 1.32 3 ρa
cm
kg.m / s 2
γ = 0.072 = 72g / s 2
m
0.0012.DW 2
D W Re p We =
Only estimates 72g / s 2 spherical
0.1mm 157cm/s 10.46 0.004
1mm 499.3cm/s 333 0.42
10mm 1579cm/s 10,527 41.55
deformed
 We <1

We >1

Reyssat et al., Shape and instability of free-falling liquid globules EPL,

80 (2007) 34005.
29
Summary
² The multiphase flow system is complex due to:
New physics (momentum, mass, and energy exchanges) at the interfaces
New scales at the interfaces à multiscale
Discontinuities at the interfaces
More governing parameters of the system

² However, governing equations and interface / boundary conditions at

microscale are known, so rigorous theoretical analysis can be conducted for
simple systems, and direct numerical simulations are possible for complex
systems.

² The interface / boundary conditions include kinematic and dynamics

conditions, and there are alternative ways to formulate the equations.

² Many more governing parameters in multiphase flow systems

30