Interim Report 1 (5/29/15) Wooram Seok (Advisors: Prof.

Myung Jhon, Wonyup Song,

Sesha Hari Vemuri)
Objectives/Goals
The objective of the past three weeks was to use various forms of the modified ShanChen multiphase (SCMP) model gathered from literature, compare them to see
advantages/disadvantages, and use the effective mass = ( ) found within the
pseudopotential term (Shan and Chen, 1993) to find density profiles across an axis given various
interaction strength coefficients (G) and analyze them in terms of stability and accuracy to
literature models and/or experimental data.
Methods
For the calculations (no visual post-processing yet), Sailfish, a basic CFD-LBM program
that had the SCMP model available, was used on Python 2.7.8 on Windows 7 64-bit processors.
The previous rendition only had the linear and the classic models for the effective mass,
represented by =

and =0 (1exp { 0 }) , respectively. A yuan-schaefer-vdw

model was added, representing the effective mass model described by Yuan and Schaefer (2007),
along with the inclusion of the van der Waals EOS. The Yuan-Schaefer model is seen here:
=

2(PRT )
RTG

Where P is given by a selected non-ideal EOS to improve the accuracy of the effective
mass (e.g.: R-K, vdW, CS, etc.)
Given the modified SC EOS for the LBM scheme:
P=RT +

G 2

P
P
and the conditions at the critical point ( ( )T =( 2 ) =0 ), Gcrit and crit were
T
calculated for each and density profiles across a an axis for a symmetric D2Q9 lattice domain
for various G < Gcrit were calculated and analyzed.
Results
Solving for the differential equations for the first two models (linear and classic), the Gcrit
for the classic model was shown to be -4 (also shown by Kuzmin 2009). Although Yuan and
Schaefer claim that their is independent of G values since they will cancel out in the SC EOS
calculations, Gcrit from the classic model was also used as an upper bound for this in the later
calculations.
Since Kuzmin 2009 already has density data for G from -4 to -9 for the classic model,
even lower G values were tried. However, stability/computation value errors were encountered at
density ratios of ~60 around |G| ~ 10 as the density gradient became steeper.
For the yuan-schaefer-vdw model, |G| ~ 200 created a memory error as the interface
width (area of largest density gradient change) decreased. Density ratios of ~1800 were
calculated before large inaccuracies/memory errors occurred.
Discussions/Questions
Given that basic phase separations such as water at ~100 oC have a density ratio of
~1700, the Yuan-Schaefer model can cover some basic fluid phase separation simulations
relatively accurately. However, various values of G had to continually tested until the limit was
found either by the program timing out/displaying an error or when it started to give
erratic/inaccurate velocities at the interface (AKA spurious currents in the literature).
Part of the problem may simply lie in the fact that a better workstation was not used for
calculations and thus a much smaller G caused time-outs before actual instability issues in the
scheme were seen. Another part may lie in the inherent viscosity contrasts of the different phases,
which are accounted for in multi-component schemes (Yang and Boek 2012) but not in the
single-component SC scheme.
Future Direction

Future work will be done on a better workstation so as to mitigate computational

limitations during simulations. The effects of spurious currents and viscosity effects will be
examined through research on multi-component SC models. The multi-range SC model,
designed to separate the effects of surface tension from the rest of the SC EOS in order to
prevent calculation instabilities at high densities, has not been examined and will be studied
through a literature survey.
Thermodynamic inconsistency (as well as the temperature-like effect of G) was also
ignored in the usage of these effective mass models, and must be optimized within the scheme
before attempting to study porous media.