Professional Documents
Culture Documents
Scales For Double-Diffusive Fingering in Porous Media: Groundwater Motions and The Associated Vertical Fluxes
Scales For Double-Diffusive Fingering in Porous Media: Groundwater Motions and The Associated Vertical Fluxes
Scales For Double-Diffusive Fingering in Porous Media: Groundwater Motions and The Associated Vertical Fluxes
g/m1
Fs~ • 1(•)2ffw'S'
dxdy-
o
8v
This order-of-magnitude estimateis independentof the tem-
peraturedifferencebetweenthe two mixed layers.This seem-
ingly obviousdeficiencycould be removedby making the
above analysisslightly more accurate.However, the several
other assumptions which would be still implicit in the model
argueagainstsucha refinement.For example,it is very likely
that mechanicaldispersionwill increasethe value of •. This
couldwell be the mostseriouslimitationon the aboveanaly-
sis.
I I -
10 -4
M2
10-2 It shouldbe noted that dimensionalargumentsmay in the
Fig. 1. The finger growth rate G as a function of the wave
future be productivein relatingsalt flux to known quantities.
numberM, for Q = 103,R = 10,andthediffusivity ratios• = 0, 0.01, But the number of independentparametersis large, and no
and 0.05. Seethe text for definitionsof the symbols. experimentsexistto suggesta way to organizetheseparame-
ters. For example,the constantsin the dimensionallyderived
resultsof Turner [1973] dependheavilyon molecularproper-
Typical (large) valuesof Gma x and Minax from Figure 2 then ties. These do not vary much in water. However, their soil
givea time scaleof the order of fractionsof a day and finger finger counterparts(e.g., k) can vary widely and so should
widths of the order of centimeters. The time scale seems small influencethe flux heavily.
comparedwith that for the presenceof, say, pesticidesin a Finally, the ratio of heat and salt fluxes,•, is also of some
field, and the spacescaleis certainlysmall comparedto the interest.This can be easilyderivedfrom equations(2): • ~ • 1
field width. Thus it seemsentirely possiblethat soil fingers + •(G + M2)•- •. For small• it variesmainlywithR: • ~ 0.1
contribute significantly to near-surfacevertical fluxes in the for R • 1; • • 1 for R • 1/•. Note that • is alwayslessthan 1.
saturatedzone. Their importancein other situations,suchas DISCUSSION
beneathlandfills,is alsosuggested.
It would be satisfyingto be able to compare double-
diffusivefingeringto the more well studied"viscousfingering"
VERTICAL SALT FLUX associatedwith mobility differencesacrossa sharp, moving
interface (see, for example, Woodin,] and Morel-Seytoux
The quantity of most practical importanceis usually the [1976]). Unfortunately,the assumptionsmade in the analyses
vertical flux of salt. (For example,this allows an estimateof of the two phenomenaare at this point so differentthat sucha
the residencetime of a pesticidein near-surface,saturated comparison is difficult. For example, almost all models of
soil.) Thus it is very tempting to try to estimate this flux, double diffusion rest on constant salinity and temperature
althoughwe do so with much hesitation,being without the gradients and also on constant viscosity and permeability
relatively rich experimental data set available for such esti- [Cornbarnousand Bories, 1975]. Even in water, where more
matesin the caseof oceanicsalt fingers.We will try to do this work has been done, analysesof sharp interfacesare in fact
to an order of magnitudeonly, only in the limit as •--• 0, and basedon the slow decayof originally sharp increasesof tem-
in a steadystate.Also, the standardand well-proven(water) perature and salinity acrossthe interface,by diffusion [Hup-
modelof two mixedlayers,separatedby a rather thin layer in pert and Manins, 1973]. On the other hand, modelsdescribing
which salt fingersoccur, is implicit in what follows.Whether viscousfingeringusea sharp interfaceand ignorethe diffusive
or not this is also a good model in porous media must be effectswhich would modify the interface,at least when con-
determinedby observation.The upper mixed layer has the sidering the stability of small disturbances.Moreover, the
uniform salinity S•; the lower has salinity S2. Then, in the movementof the interfaceis very important to the dynamics
abovenotation,S = (S• - S2)/2. of the process.
With salt diffusionneglected,the key equationsare those
describing momentumand heat conservation. In a steadystate
(4 = 0), and assumingthe same functional forms as above,
theseequationsare
WQ = •, -- 1 •'M2 -- -- W
Eliminating• gives
W = -(Q + l/M2) -•
We assume,following Schmitt[1979] and many others,that
the width of the fastest-growingfingers dominates in the
steadystate. SinceMinax growswithout limit as ß• 0, W ~
- 1/Q,which gives,disregardingthe minussign,w ~ karlSly. 103 Q lOG
(It shouldbe noted that this result is very similar, under the Fig. 2. Growth rates G (solid curves)and squaredwave numbers
M e (dashed
curves)
for x = 0.01and reasonable
rangesof R and Q.
appropriatetransformationfrom v to k, to that givenfor water The lines are isolinesspacedin powers of 10. A number by a line
by Turner [1973].) Then the downwardsalt flux for square denotesthe inversepowerof 10 appropriateto that line. For example,
fingersis 3 means10-3. Seethetextfordefinitions
of thesymbols.
1228 GREEN: DOUBLE-DIFFUSIVE FINGERING
However, an elementary consideration of the stationary, portant in some situations.The computedgrowth rates and
sharp interface may be instructive.The reasoningof Huppert finger widths are suchthat slow time and spacevariationsin
and Manins considersthe time changeof salinity and temper- the base state, and grain size of the medium, will not easily
ature gradientsnear the (horizontal) interface,in conjunction influencethe fingers.It is still unclear to what extent the seem-
with the classicalfree-boundaryRayleigh criterion for insta- ingly close analogy between soil fingers and oceanic salt
bility, in an admittedly ad hoc manner. By adapting this fingers can be pursued. For example, density-drivenconvec-
reasoningto a porous medium, one can argue that double- tive mixing of groundwater is probably much less energetic
diffusivefingerswill form at an originally sharpinterfacewhen than that of ocean water, so that layering may not occur as
readily. Thus modelsbasedon well-mixedlayersseparatedby
-•AT/•AS < (KT/Ks)3/2 a thin, fingeringregionmay not apply. On the other hand, the
internal-wave motions which can act to limit the vertical
Here AT and AS are the temperature and salinity increases length of salt fingersare not likely to be presentin groundwa-
acrossthe interface.(Note that • < 0, so that •A T/fiAS < 0.) ter.
Now considerviscousfingers.Sincethe fluid is the same on Horizontal dispersionmay well be the primary limit on the
both sides of the interface, interfacial tension is neglected.vertical coherence of double-diffusive fingers. The effect of
Then the criterion for viscousfingeringfor an interfaceat restsuch dispersion can be roughly estimated. For a Peclet
is 0•AT + BAS > 0, or -o•AT/fiAS < 1 [-Saffmanand Taylor, number of around 1 or greaterthe dispersioncoefficienttrans-
1958]. This simply saysthat the upper fluid is heavierthan theverseto the mean flow is found to be D • v•d/10,where v• is
lower fluid. the seepagevelocity w'/a and d is the (granular) porous
However, double-diffusivefingering is only possiblewhen medium particle diameter [List and Brooks, 1967]. The time
Kr/Ks > 1. In any such situation, as AS is slowly increased for salt to mix to the core of an upward moving finger (thus
from zero with AT fixed, double-diffusivefingeringwill occur destroyingthe finger's identity and greatly diminishingits
first. This comparisonis rather qualitative because,for exam- buoyancy) is [ • r2/D,wherer is thefingerhalfwidth.In this
ple, diffusionis neglectedin the Saffman-Taylormodel. How- timethefluidin thefingertravelsa distance L •,,•. œ.Then
ever, since Kr >>Ks in most circumstances,it seems that fingersshould remain coherentfor a vertical distanceof the
double-Oiffusivefingering is more likely to occur, especially order of L • 10r•/d, and the finger aspectratio shouldbe
along interfacesat rest.Nonetheless,effortsto provide a soun- about L/2r • 5r/d. The resultsgiven above suggestthat r >>d,
der basis for comparing the two phenomena are certainly in so that the fingers can be quite slender,in accord with the
order. model.
In a similar vein, one can ask how the growth rates studied This simpleestimateneedsimproving,along with the entire
in this paper changewhen 0•becomessmall. Now R < 1, and model. But it can inspire cautiousoptimism about the impor-
the situation is gravitationally unstable.Consider the limiting tance of double-diffusivefingering in porous media. Experi-
case0•= 0. Here the heat conductionequationalecouples from mental testing of the above analysisand resultsto guide fur-
the other two of equations (1). When w is eliminated from ther analyticalprogressseemvery much in order. As with salt
thesetwo equations,and the sametypesof solutionsassumed fingers,the completephenomenonis probably not amenable
as before, we find to analytical treatment, and satisfactorymodels need experi-
mental evidenceas support. Such experimentsappear more
difficult than with salt fingersbecauseof the lack of straight-
,•+-•- ,,t+ Ks
m2)
= forward flow visualizationtechniques.Hele-Shawexperiments
This givesa maximumgrowth rate '•mat rn = 0, whichis would certainlyseemto be calledfor, at least for preliminary
results.Suchexperimentshave beencarriedout on the closely
2•mk=1+ - 1 analogouscasewhere the upper fluid is colder and lesssalty
O'V O'2V
2 J than the lower fluid [Griffiths, 1981].
bility and salt fingersin a porous medium, Phys. Fluids, 15, 748-
T. Green, Department of Civil and Environmental Engineering,
753, 1972.
University of Wisconsin-Madison,1269 EngineeringBuilding, 1415
Turner,J. S.,Buoyancy
Effectsin Fluids,367 pp.,CambridgeUniver- Johnson Drive, Madison WI 53706.
sity Press,New York, 1973.
Wooding,R. A., Growth of fingersat an unstablediffusinginterfacein
a porous medium or Hele-Shaw cell, J. Fluid Mech., 39, 477-495,
1969. (ReceivedAugust25, 1983;
Wooding, R. A., and H. J. Morel-Seytoux,Multiphase fluid flow revisedMay 10, 1984;
throughporousmedia,Annu.Rev.Fluid Mech.,8, 233-274, 1976. acceptedMay 31, 1984.)