Guillaume Galliero, LFCR, Pau University,...
Transcript of Guillaume Galliero, LFCR, Pau University,...
INTERFACE DYNAMICS OF MODEL SYSTEMS:FROM LOCAL PROPERTIES TO BOUNDARY CONDITIONS
Guillaume Galliero, LFCR, Pau University, France
Collaborators :H. Hoang, H. Si Hadj Mohand, LFCR, Pau Univ.
D. Legendre, IMFT, Toulouse Univ.
16th May 2018Workshop on Modeling and Simulation of Interface DynamicsIMS - Singapore
MOTIVATIONS
2OIL AND GAS ISSUES
Quantifying accurately Multiphasic Fluid Dynamics in porous medium is crucial for the Petroleum Engineering
Various tools are available for that purpose but most of them are limited when employed to deal with “new” systems:
Complex resources : Sour/Acid Gases, Shales gas/oil … Complex conditions : Extreme HP/HT, Tight reservoirs …
Complex recovery methods: Smart Water, Polymer Flooding …
Data and more physically based models are required
But experiments are hard to achieve (HSE, HP …)
METHODOLOGY
4FROM NANO TO GIGA-SCALE: DIFFERENT TOOLS
ms
ms
ns
ps
fs
Å nm mm mm
Ab initioQuantum
Atomistic
Monte-CarloMolecular dynamics
Mesoscale
Lattice Gas, LBDPD, DSMC, KMC …
Time
Length
Continuum
Finite Difference/VolumeFinite ElementDiscrete Element …
Coarse graining
The choice is guided by time/length scalesDirect or Indirect Coupling between scales/tools is possible
Reservoir Simulations …
Pore/Core-scale modeling
NanoPore/interfaces
Molecular screening
Conservation Eqs.
Boltzmann Eq.
Newton Eqs.
Schrödinger Eq.
Adapted from Gubbins & Moore, IECR (2010)
5OUR GROUP STRATEGY
Develop/use molecular simulations codes to study/quantifyfinely fluid behavior at interfaces at the nanoscale
Thermophysical propertiesTension, Diffusion …
Physical mechanismsTransport, Osmosis …
Direct/indirect upscalingBoundary conditions …
g
qRqA
INTERFACE DYNAMICS:ISSUES IN A SIMPLE SYSTEM
7INTERFACE DYNAMICS AT THE NANOSCALE
g
E.g.: A nanodroplet of a simple fluid moved by body forces on a perfectly flat and rigid surface
Viscous stress (Slip ?)
Fluid Inhomogeneities? (Local density/viscosity ?)
Uncompensated Young Tension (Dynamic microscopic angle?)
Viscous stress (Boundary Conditions)
5 nm
FROM LOCAL PROPERTIES …
g
Fluid Inhomogeneities? (Local density/viscosity ?)
LOCAL FLUID TRANSPORT PROPERTIES ININHOMOGENOUS FLUIDS
H. HOANG, G. PIJAUDIER-CABOT (LFCR) AND F. MONTEL (TOTAL) Hoang and Galliero, PRE (2012), JCP (2013), JPCM (2014), PRE (2015)
10FLUID INHOMOGENEITIES ISSUES
L*
20 30 40 50 60
h*
2
4
6
8
Liquid Gas
Solid
Density of a “oil” drop in a nanopore
Low High
3 n
m
At solid interfaces, fluids are strongly inhomogeneous
(SFA, DFT …) Spatialy variable viscosity ?
Couette like simulations of a simple fluid (LJ) in a nanoslit
x*
-4 -2 0 2 40
0.2
0.4
0.6
0.8
1
bulk
=0.291
bulk
=0.408
bulk
=0.519
bulk
=0.625
p* = 0.5p* = 0.75p* = 1.25p* = 2
Density profiles
5 nm
11A VARIABLE VISCOSITY ?
The local viscosity computed by MD cannot be deduced from the local thermodynamic variables alone (T and )
Local Viscosity profile Local Viscosity profile
p = 20 MPa p = 80 MPa
-- ----
-- -- ---- -- -- --
---- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
x*
m
-4 -3 -2 -1 00
1
2
3
4
5
There are non local effects
MD simulationvdW model--
--
--
--
----
-- -- -- ----
-- -- --
--
-- -- -
- -- -- -- --
-- -- -- -- -- -
--- -- --
---- -
---
x*
m
-4 -3 -2 -1 00
0.2
0.4
0.6
0.8
P*=0.5 P*=2
The fluid viscosity is variable in space !
Hoang & Galliero, PRE (2012)
12INTRODUCING NON LOCAL EFFECTS
Using the microscopic formulation of the momentum flux, shear viscosity is decomposed in two parts: t c
Translation Collisions
To introduce non-local effects, c is assumed to be a function of the density averaged over a typical length
2
2
1
x x s ds
0
c
bulk
xx x with e.g.
This method can be more accurate using more evolved weight functions and a perturbation scheme
Similar to Bitsanis et al. J. Chem. Phys. 1987
Hoang & Galliero, JCP (2013), JPCM (2014)
13NON LOCAL VISCOSITY ?
Effective viscosity can multiplied by an order of magnitude relatively to the bulk
Non local viscosity is well modeled by the simplest DFT
Can be combined with DGT/DFT to introduce fluid inhomogeneities …
Local Viscosity profiles Local Viscosity profiles
p = 20 MPa p = 80 MPa
----
--
--
----
-- -- -- ----
-- -- --
--
-- -- -
- -- -- -- --
-- -- -- -- -- -
--- -- --
---- -
---
x*
m
-4 -3 -2 -1 00
0.2
0.4
0.6
0.8
-- ----
-- -- ---- -- -- --
---- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
x*
m
-4 -3 -2 -1 00
1
2
3
4
5
MD simulationvdW modelNon-local model
P*=0.5 P*=2
Hoang & Galliero, PRE (2012)
TO SHEAR AT FLUID-FLUID INTERFACES …
g
Viscous stress (Slip ?)
VELOCITY SLIP AT FLUID-FLUID INTERFACES
Galliero, PRE (2010), Bugel et al. Micro&Nanofluid. (2011)
M. BUGEL, J.P. CALTAGIRONE (I2M, BORDEAUX, FRANCE)
16FLUID-FLUID SHEAR STRESS ISSUES
The partial slip (Navier B.C.) at fluid-solid interfaces is well known, but what about Fluid-Fluid interfaces ?
Hybrid and MD simulations of Fluid-fluid interfaces under shear
Slip s known to occur at polymer-polymer interfaces
Liquid-Gas Liquid-Liquid
17HYBRID SIMULATIONS OF A DIPHASIC COUETTE FLOW
Bugel et al., M&N (2011)
Hybrid MD-CFD simulations
Coupling on primary variables (v, T)Schwartz alternating method
Liquid Gas
At the liquid-gas interface a partial slip seems to occur !
Discrete
Continuous
18VELOCITY JUMP AT A LIQUID-LIQUID INTERFACE
x* position
0 10 20 30 40 50
i*
0.0
0.2
0.4
0.6
0.8
vz*
-0.4
-0.2
0.0
0.2
0.4
The velocity profile exhibits a “jump” at the interface
Density
Velocity
Lx 30 nm
Phase I & II have the same properties
but are weakly miscible (e12=kije, with kij<1)
Galliero, PRE (2010)
19INTERFACIAL VISCOSITY IN SIMPLE FLUIDS ?
x* position
6 8 10 12 14 16 18 20
vz*
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
kij = 0.125
kij = 0.375
kij = 0.625
vz*
Velocity jump for various miscibility
0.0 0.5 1.0 1.5
(b
ulk-
int)/
bu
lk
0.0
0.1
0.2
0.3
0.4
0.5
Interfacial viscosity
Phase I
Phase II
The interfacial viscosity is lower than the one deduced from profile (DGT?)
In a water wet nanopore (10 nm) the theoretical oil relative permeability is roughly two times higher when the partial slip is taken into account
The partial slip increases with the interfacial tension
Interfacial self-diffusion is probably inversely proportionally increased …
YOUNG TENSION AT THE CONTACT LINE …
g
Uncompensated Young tension (Dynamic microscopic angle?)
CONTACT LINE DYNAMICS:(NANO?)DROPLETS SPECIFICITIES
H. HOANG (LFCR), S. DELAGE-SANTACREU (LMAP, PAU, FRANCE)
R. LEDESMA, D. LEGENDRE (IMFT, TOULOUSE, FRANCE) Hoang et al., in preparation
22YOUNG STRESS ISSUES
Young stress occurs at the Contact Line but is it valid at the nanoscale?
MD simulations of nanodroplets on a perfectly flat and rigid surface
Equilibrium Dynamic (external force)
g
Lennard-Jones 2D droplets on LJ CFC solid surface for 0.4p <q < 0.85p
10 nm
23
LV
SVSL q
cos ( )z q
x
zx
-20 -10 0 10 20-0.2
-0.1
0
0.1
0.2
Local shear stress (zx)
z=0
z=5
Liquid GasGas
Young tension ?
( ) ( , )dRInt
LInt
x
zxInt xz x z x
EQUILIBRIUM: YOUNG TENSION (1/2)
At the contact line the (integrated) shear stress is macroscopicallyequal to the Young tension i.e.
At equilibrium, shear stress is localized at the Liquid Vapor interface and decreases when z increases (consistent with Young tension)
Irving-Kirwood/Method of Plane
24
z
[In
t-c
os(
q)]
0 2 4 6 8
0
0.2
0.4
0.6
z
[In
t-c
os(
q)]
0 2 4 6 8
0
0.1
0.2
0.3
0.4
z
0 2 4 6 80
0.5
1
1.5
2C=0.55
C=0.25
Young stress =
shear stress
Shear tension at the contact line differs from cos(q)!
But Young stress = shear stress at liq/vap interface above zY some molecular sizes (zy scales with q-1)
Shear minus Young tensions (q 0.85p)
EQUILIBRIUM: YOUNG TENSION (2/2)
How shear tension compares to Young tension ?
Shear minus Young tensions (q 0.4p)
Young stress =
shear stress
( ) ( , )dRInt
LInt
x
zxInt xz x z x
25DYNAMICS: SHEAR/YOUNG TENSION (qEq = p/2)
Even out of equilibrium Shear tension=Young tension
above z 5 !
z
In
t
0 2 4 6 8-0.2
0
0.2
0.4
0.6
Equil. State
Non-Equil. State (Ad)
Non-Equil. State (Re)
Shear minus Young tension (Ca=0.2)
cos(q)
In
t
0.1 0.2 0.3 0.4
0.1
0.2
0.3
0.4
Ca
Ca
Shear tension at z=5 vs Ca
CL dynamics (at z>5) driven by [cos (qNE)-cos(qEq)] +viscous stress (Ca)
z
[In
t-c
os(
q)]
0 2 4 6 8-0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
Young tension is compared with shear tension at different z and different Ca …
The influence of Ca on shear tension is not negligible
qR qA
Consistent with Qian et al. PRE 2003, Ren and E, PF 2007
26A SPECIFIC MICROSCOPIC PICTURE ?
The problem becomes more complex when an adsorbed layer (macromolecules) occurs …
In not too wetting cases, classical laws are really robust !
Classical (homogeneous) continuum breaks down (q-1)
Uncompensated Young Stress [cos (qNE)-cos(qEq)]
Viscous stress(Navier BC)
How upscaling MD results when dealing with wetting case (thin films …) ?
TO BOUNDARY CONDITIONS
LIQUID-LIQUID NANOCOUETTE FLOW:WHICH BOUNDARY CONDITIONS ?
H. HOANG H. SI HADJ MOHAND (LFCR)
D. LEGENDRE (IMFT, TOULOUSE, FRANCE)
29BOUNDARY CONDITIONS ISSUES
There exists many CL boundary conditions options but whichone is the most appropriate when all scales are solved ?
Comparison between CFD and MD simulation on a liq-liq Couette flow
Seminal works of Robbins, Koplik, Qian, Ren ...
Phase I & II (Lennard-Jones fluids) have the same properties but are non miscible
30CFD MODELS (JADIM)
Tested dynamic angle models :
Static : qd = qY
Generalized Navier BC (GNBC*) : cosqY
- cosqd=
Ucl
gB
cl
cosqY
- cosqd=
Ucl
g mv
l3exp
s(1+ cosqY)
nkBT
æ
èç
ö
ø÷Molecular Kinetic Theory (MKT**) :
f=14.3 (em)1/2/3
Navier-Stokes (NS) Equations +
Volume Of Fluids (VOF, interface)
Reference case : MD of Qian et al., Phys. Fluids, 2003
Imposed Slip length + Dynamic angle
Bcl=3.02 (em)1/2/3
Used slip length model (Navier): UW
- V = l¶U
¶nW
Mesh size < Slip length
where l=1.63
qd
Equilibrium angle: qY = p/2
+V
-V
*Qian et al. PRE (2003), Ren and E, PF (2007), **Blake and co-workers
NS Simulation – GNBC model
with normalized friction Bcl=3.02
Unstable case : the MKT “friction” is too large
NS Simulation – MKT model
with normalized friction f=14.3
31CONTACT ANGLE FRICTION CONTROLS THE STABILITY
NS - Static model NS – GNBC model
Too much slip Slip is modified by friction
-V
V
(friction Bcl=3.02)
32VOF-NS VERSUS MD: PREDICTIONS
No Contact Line model seems to be fully predictive when combined with VOF-NS
GNBC: Bcl=3.01 => Bcl=1.7
MKT: f=14.3 => f=1.7
Static: l=1.63 => l=1.2
Velocity profiles
33VOF-NS VERSUS MD: ADJUSTED MODELS
If adjusted, all CL models yield reasonable results !
NS+VOF describes well this nanoflow when all scales are solved
OUTCOMES
35OUTCOMES
Shear viscosity can be variable in space at the Contact Lineo Non locals effects on viscosity occur close to a solid surfaceo Interfacial region may lead to apparent fluid-fluid slip
And for future works …
Include non local density/viscosity …. introduce upscaling …
Macroscopic laws are very robust in liquids if all scales are solvedo Uncompensated Young tension dominant at the CL (but at z>0)o For Liq-Liq Couette flow, CL models works well when fitted
Sheared Liq-Gas
THANK YOU FOR YOUR ATTENTION !
AcknowledgmentsPau University
CNRSAnd TOTAL
All colleagues collaborating
EXTRAS
Boundary conditionsConfinement effectsWettability …
Contact anglesDiffusion coefficientsSlip length ...
38
MOLECULAR SIMULATIONS: WHAT FOR ?
Molecular ModelForce Fields
Molecular SimulationsMC,MD
“Exact” emerging properties/fields
Test/Development of Theories
Pseudo-experimental Data
, q, Pc, k,, D, V …
39
MOLECULAR SIMULATIONS: HOW IT WORKS ?
No assumption concerning the physical phenomena that may emerge!
MOLECULAR MODEL
PHYSICAL PROPERTIES (X)
Molecular Dynamics (Temporal evolution)
Monte Carlo (Statistical evolution)
n ( 106-108)configurations
X a physical property 105-107 timesteps (t 1 ns)
1
MD
X X t dtt
1
1
n
iMCi
X Xn
A set of N interacting
particles
Average over configurations
Average over time
Ergodic Theorem MC MD
X X
40
MOLECULAR DYNAMICS: HOW IT WORKS ?
At a time step t each centre of forces (atom or molecule) is characterised by its position r(t), its velocity v(t) and its acceleration a(t)
Newton’s Equation + Effective interaction potentials (V)
dt
vdmF i
i
j
ij
ij
ij
ij rr
VF ˆ
i
iiii
m
tFttvttrttr
)(
2
1)()()(
2
New position at t + t (various integrator)
i
Fij
vi
i
(t)
Fijvi
(t + t)
j
Explicit Scheme, “easily” parallelised (domain decomposition)Computation time N2 (can be reduced to N logN)
Quasi-experimental process (data with error bars ...)
41
FORCE FIELDS: THE CORE OF THE PROBLEM
From hard-sphere representation to a full atom (ab initio) description, but no ideal model
12 6
4LJUr r
e
Dispersive/repulsive interactions: usually a Lennard-Jones pot.but Mie/Exp pot. are used
e
ULJ
r
Polar interactions: isotropic or point charges
Internal degrees of freedom: Bonding, Flexion …
Mixtures: Combining rules (Lorentz-Berthelot …)
The choice is guided by the goal (pseudo-experimental vs understanding)
Non-Bonded
42
HOW COMPUTING MACROSCOPIC QUANTITIES ?
N
i
N
ij
ijijc
c
EV
P1
23
1rF
2
1
x
x
TN dxxPxP
Field properties
Sound velocity, …
Dynamic properties
A flux is imposed (heat, momentum, …) and the response is measured (Non-Equilibrium)
All physical properties are “mesurable” in the ad hoc ensemble
The property is deduced from the fluctuations (Equilibrium)
Momentum/Energy
Exchange
Equilibrium/Thermodynamic properties
The strain, velocity, temperature, pressure, concentration … fields are deduced from averages (over time or ensemble)
43METHODOLOGY
Extensive Molecular Dynamics simulations on well controlled Vapor/Liquid/Solid systems
Variable wetting properties C=0.25 (q 0.85p )
Lennard-Jones interactions
Static and migration under external force configurations
12 6
4Ur r
C
e
Fluid-Fluid: C=1; Fluid-Solid: C =0.25-0.55
q
C=0.55 (q 0.4 p )
Sessile Lennard-Jones droplets on a CFC flat rigid LJ solid
Static nanodroplet Moving nanodroplet
Vapor
Liquid
44EQUILIBRIUM : CONTACT ANGLE
The Contact angle is determined by a circle fit (using Gibbs dividing surface and extrapolated at z=0.5)
C
qp
0.2 0.3 0.4 0.5 0.6
0.4
0.5
0.6
0.7
0.8
0.9
ef-s=Cef-f
Contact Angle at z=0.5
The contact angle is proportionalto fluid-solid interactions
Consistent with literature
The contact angle is weakly size dependent (here R 5nm)
x
z
-15 -10 -5 0 5 10 150
5
10
15
20
x
z
-20 -10 0 10 200
10
20
30
C=0.25 (q 0.4p ) C=0.55 (q 0.85p )
45EQUILIBRIUM: YOUNG-LAPLACE EQUATION (1/2)
As well known Pz= cst everywhere but not Px (cf. = (Pz-Px)dz)
Vapor pressure profile (q p/2)
PzPx
When a curved fluid-fluid interface is present a pressure (capillary)
difference between the fluids exist
PLiqPVap
LV
Liq VapP PR
Liquid pressure profile (q p/2)
zL
oca
lP
ress
ure
0 2 4 6 8 10
-0.5
0
0.5
1
1.5
z
0 2 4 6 80
0.5
1
1.5
PzPx
Density
But what occurs at the nanometer scale ?
Young-Laplace Eq.
z
Loca
lP
ress
ure
0 2 4 6 8 10-0.05
0
0.05
0.1
PzPx
46EQUILIBRIUM: YOUNG-LAPLACE EQUATION (2/2)
, , 0.6 z Liq z VapR P P
0.6LV
Young-Laplace equation, LV = RP, is well respected for such nano-droplets (using normal pressure)
The curvature correction to LV (Tolman’s length) is negligible !
From liquid/vapor interfaces simulations one can compute surface tension Liquid Vapor
and from the droplet normal pressure
In 3D droplets, line tension is negligible as well … (works of Bresme …)
47EQUILIBRIUM: YOUNG-DUPRÉ LAW (1/2)
Contact angles vs C
LV
SVSL q
1cos SV SLY
LV
q
From Young –Dupré law
From circle fit
From the computation of all three , the Young angle can be estimated
At the contact line the three (Sol/Vap, Sol/Liq and Liq/Vap)
tensions compensate each otherscosLV Y SV SL q
Young-Dupré law
Young angle, qY, is not equal to the measured angle at z=0.5 !
C
qp
0.2 0.3 0.4 0.5 0.6
0.3
0.4
0.5
0.6
0.7
0.8
48EQUILIBRIUM: YOUNG-DUPRÉ LAW (2/2)
Why so: Px (and so ) is affected by the surface
over long-range (cf. vdW forces)
Finite size effects on the result when q 0 (zY > h)Macroscopically described by disjoining pressure (when h < zY)
Distance at which qY = q, noted zY,
Scales 1/qY
zYq Is there a distance from the solid surface at which qYD=q,
i.e. macro angle = micro angle?
Distance at which qY=q
1/qY
z Y
0.45 0.6 0.75 0.9 1.050
1
2
3
4
5
6
7
h
49DYNAMICS: DROPLET (qEq = p/2) UNDER AN EXTERNAL FIELD
Droplet under a « gravity » field Flow field (Ca=0.2)
C = 0.45qEq p/2
FExt
VC
entr
oid
0.001 0.002 0.003 0.004 0.005 0.006
0.02
0.04
0.06
0.08
Droplet velocity vs external force
A friction coefficient can bededuced (here x=F/V 0.07)
linear regime
By varying Fext, the capillary numberis changed
Ca=V/ 0.05-0.35
Slip length at Liq/Sol 1
Non linear response occurs for smaller q and high V
50DYNAMICS: ADVANCING/RECEDING CONTACT ANGLES (qEq = p/2)
x
z
-15 -10 -5 0 5 10
6
8
10
12
14
16
qAqR
Dynamic contact angles (at z=5) qA,R (at z=5) vs Ca
Ca
Abs(q
q
Eq)
0 0.1 0.2 0.3 0.40
0.05
0.1
0.15
0.2
0.25
0.3
0.35
CaA
bs(
q
q3 E
q)
0 0.1 0.2 0.3 0.40
0.2
0.4
0.6
0.8
1
A hydrodynamic behavior holds for both angles at z = 5 !
qR
qA
Even with nano-droplets (but not too low q), hydrodynamics seem to be well respected above zYs!
At both contact lines advancing and receding angles can be measured
Circle fit is stil valid
Hydrodynamics model predict that q3 Ca