HAL Id: jpa-00212428
https://hal.archives-ouvertes.fr/jpa-00212428
Submitted on 1 Jan 1990
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of
sci-entific research documents, whether they are
pub-lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Direct simulation of a permeable membrane
U. Brosa
To cite this version:
1051
Short Communication
Direct simulation of
apermeable
membrane
U. BrosaHLRZ c/o KFA, Postfach 1913, D-517 Jülich, F.R.G.
(Reçu
le 27 mars 1990,accepté
le 3 avnil 1990) Abstract. 2014 Cellular automata are used tocompute flow thru a
permeable
membrane.Scattering
centers constitute the membrane. This is in marked contrast to the
approach
of classicalhydrody-namics which represents a membrane
by
aboundary
condition. With thescattering
centers we obtaindifferent, but more
plausible
resultsindicating
thatsimple
diffusion is thedominating
process in aporous
layer.
We have thus a case where cellular automata showsuperiority
over the classical meth-ods of theoreticalhydrodynamics.
1
Phys.
France 51( 1990)
1051-1053 1 er JUIN 1990,Classification
Physics
A bstracts 05.50-7.SSMLE
JOURNAL
DE
PHYSIQUE
Tome 51 N° 11 1 er JUIN 1990
Several
investigations
on theapplicability
of cellular automata on viscoushydrodynamics
have been donemeanwhile,
see e.g.[1 - 4].
In thesestudies,
solutions of the Navier-Stokesequation
available from other sources werecompared
withoutput
of cellular automata. It is clear now that cellular automata cangive
a truthfulpicture
ofliquid
reality
aslong
as certain limits are nottransgressed
[5].
However,
atpresent
those automata arerelatively
inefhcient so that their mainadvantage
consists inbeing
foolproof.
Probably
cellular automata will find their most relevantapplications
in themicroscopic-macro-scopic
research. From theviewpoint
of hydrodynamics,
cellular automata are based on a schematic model of the molecular motion. Hence one canstudy problems such as
fluctuations and localequi-libration,
and one can check now much thesemicroscopic
properties
show up in themacroscopic
behavior of the fluid.For this an
example
shall begiven right
now. It is flowthrough
a porous membrane. For amicroscopic
theory,
a membrane is alayer
ofscattering
centers which theliquid
has towriggle
thru. The
applications
ofmembranes,
however,
aremacroscopic.
We became aware of the
problem
in aprevious study
[6].
Thegeometry
is shown infigure
1: 1Bvo channels conduct Poiseuilleflow,
each on its own. Theproblem
would be trivial if it werenot for a membrane that connects the channels. The maximum velocities are v"zax at the inlets
of both channels but vmax -f-
Ovmax
and Vm02: -wmax
at the outlets of the upper and the lower1052
channel,
respectively.
Apart
of thefluid, therefore,
must pass the membrane. It turned out in[6]
that thecomputed velocity
fieldV(r)
depends
much on the way in which the membrane is taken into account. Theboundary-value
problem displayed
infigure
1 is hence a sensitive indicator of the treatment ofporosity.
Fig.
1. - Flow thru two channels connectedby
a membrane. All parts show flows withReynolds
number 500. The truput is50%,
i.e.làvmo:a:/vmo:a:1
= 0.5. We have solid walls in the middles, indicatedby
the fulllines, and mirror
planes
above and below,depicted by
the dashed lines. See the text for themeaning
of the four parts and the papers[6, 8]
for detailsconcerning
the geometry and thealgorithms.
In classical
hydrodynamics[7],
the behavior of the membrane is describedby
theboundary
con-ditionmeaning
that the membrane allowsonly
for fiowperpendicular
to its surface.Equation (1)
is tohold on the dotted line in
figure
1.However,
its correctness may be doubted.First, experts
inhydrodynamics
notice at once that(1)
isjust
theslip
boundary
condition on a vertical wall as used in thetheory
of idéalliquids.
Theapplication
ofslip
boundary
conditions toproblems
with viscous flowalong
solid bodies is known togive
misleading
results.Second,
there is no reasonwhy
the fluid should not cross the membrane in anoblique
direction.With cellular automata we can
easily
look into thisproblem.
All we have to know is that a cellular automatonopérâtes
on agrid.
Then we select some of the nodes in thevicinity
of the dotted line and let them becomepoints
at which the moleculesexperience
backward reflections.1053
The essence of the results is distilled in
figure
1. Part(a)
showsunhampered thruput.
Instead of themembrane,
there isjust
a hole. Due to inertia effects the fluid passes the holepreferentially
at its downstream
edge.
In Part(b)
theboundary
condition(1)
isput
into force. Itonly
makes the velocities vertical. As mentionedalready
above,
(1)
is anequation
for idéalliquids.
Hence we cannotexpect
a realimpediment
of the flow. Part(c)
shows whathappens
when every second node on the dotted line is made ascattering
center. We see now a much broader distribution of the velocities over the membrane. This isexpected
since thescattering
centers induce diffusion. Inaddition,
the velocities pass the membraneobliquely.
For the results exhibited in Part(d), only
every third node was used as
scattering
center. Theonly point
to be noticed here is the small difference with Part(c).
In addition 1 have modelled the porous membrane in several differentways. For
example,
1 took nodes notonly
from the dotted line but from astrip
around that line. Icomposed
nodes to form smallhexagons,
and so forth.Invariably pictures
as shown infigure
1(c)
and(d)
were obtained. This is an almost trivial result as it isalways plain
diffusion which features the flow in the porouslayer.
It is nevertheless remarkable that it takes so fewscattering
centersto
get
full diffusion.Part
(b)
offigure
1 shows resultspresented
first in[6].
Thejuxtaposition
with Part(c)
exhibits that we have in fact asignificant
modification.Finding
the same resultsby
computational
methods of classicalhydrodynamics
seems to bepossible.
However,
one has to discard theboundary
condition(1)
and must model the membraneby
alayer
of small but finite thickness. In thislayer,
inertia isdamped
so that the Navier-Stokesequation
can bereplaced
by
a diffusionequation.
As aboundary
condition between the porouslayer
and the freeflow,
continuity
of thevelocity
fieldought
bestipulated.
Such anarrangement
complicates
theapplication
of classical methods like finite elements so that cellular automata have anedge.
Acknowledgements.
l’m
grateful
to C. Küttner forsuggesting
theproblem
to me.Encouragement
by
D. Stauffer isappreciated.
Part of this work was doneusing
the new interactive facilities at the HLRZ.References
[1]
D’HUMIERES D. and LALLEMAND P.,Complex Systems
1(1987)
599.[2]
LIM H.A.,Phys.
Rev. A. 40(1989)
968;LIM H.A.,
Complex Systems
2(1988)
45.[3]
HAYOT E and RAJ LAKSHMI M.,Physica
D 40(1989)
415.[4]
DUARTE J.A.M.S. and BROSA U., J. Stat.Phys.
59(1990)
501.[4]
DAHLBURG J.P., MONTGOMERY D. and DOOLEN G.D.,Phys.
Rev. A 36(1987)
2471.[6]
BROSA U., KÜTTNER C. and WERNER U., Flow thru a porous membrane simulatedby
cellular automataand finite elements, submitted for