xvEPA
United States
Environmental Protection
Agency
Robert S. Kerr Environmental
Research Laboratory
Ada OK 74820 /
EPA/600/6-90/004
April 1990
Research and Development
Laboratory
Investigation of Residual
Liquid Organics from
Spills, Leaks, and the
Disposal of Hazardous
Wastes in Groundwater
EPA/600/6-90/004
r r r
-------
EPA/600/6-90/004
April 1990
LABORATORY INVESTIGATION
OF RESIDUAL LIQUID ORGANICS
from spills, leaks, and the disposal of hazardous wastes
in groundwater
by
John L. Wilson, Stephen H. Conrad, William R. Mason
William Peplinski, and Edward Hagan
Department of Geoscience
& Geophysical Research Center
New Mexico Institute of Mining and Technology
Socorro, New Mexico 87801
Cooperative Agreement: EPA CR-813571
Project Officer: Jerry Jones
Robert S. Kerr Environmental Research Laboratory
Office of Research And Development
U.S. Environmental Protection Agency
Ada, Oklahoma 74820
This study was conducted in cooperation with
the New Mexico Water Resources Research Institute,
Las Cruces, New Mexico, under Grant NMWRRI 1345648
ROBERT S. KERR ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
ADA, OKLAHOMA 74820
-------
NOTICE
The information in this document has been funded wholly or in part by the United States Environmen-
tal Protection Agency under Cooperative Agreement No. CR-813571 to the New Mexico Institute of Min-
ing and Technology, Socorro, New Mexico. It has been subject to the Agency's peer review and admin-
istrative review, and it has been approved for publication as an EPA document. Mention of trade names
or commercial products does not constitute endorsement or recommendation for use.
-11-
-------
FOREWORD
EPA is charged by Congress to protect the Nation's land, air and water systems. Under a mandate
of national environmental laws focused on air and water quality, solid waste management and the control
of toxic substances, pesticides, noise and radiation, the Agency strives to formulate and implement
actions which lead to a compatible balance between human activities and the ability of natural systems
to support and nurture life.
The Robert S. Kerr Environmental Research Laboratory is the Agency's center of expertise for inves-
tigation of the soil and subsurface environment. Personnel at the Laboratory are responsible for man-
agement of research programs to: (a) determine the fate, transport and transformation rates of pollut-
ants in the soil, the unsaturated and the saturated zones of the subsurface environment; (b) define
the processes to be used in characterizing the soil and subsurface environment as a receptor of pollut-
ants; (c) develop techniques for predicting the effect of pollutants on ground water, soil, and indige-
nous organisms; and (d) define and demonstrate the applicability and limitations of using natural pro-
cesses, indigenous to the soil and subsurface environment, for the protection of this resource.
This report presents residual saturation data, illustrations and models which help to better under-
stand the basic physical mechanisms controlling the movement, and especially the capillary trapping
of organic liquids in soils and ground water. Emphasis was on relating the various mechanisms to the
issues of contaminant movement, characterization and remediation.
Clinton W. Hall
Director
Robert S. Kerr Environmental
Research Laboratory
-111-
-------
ABSTRACT
Organic.liquids that are essentially immiscible with water migrate through the subsurface under the
influence of capillary, viscous, and buoyancy forces. These liquids originate from the improper disposal
of hazardous wastes, and the spills and leaks of petroleum hydrocarbons and solvents. The laboratory
studies described in this report examined this migration, with a primary focus on the behavior of the
residual organic liquid saturation, referring to that portion of the organic liquid that is 'trapped' by
capillary forces in the soil matrix. Residual organic saturation often constitutes the major volume of the
organic liquid pollution, and acts as a continual source of dissolved or vapor phase organics.
Four experimental methods were employed. First, quantitative displacement experiments using
short soil columns were performed to relate the magnitude of residual organic liquid saturation to fluid
properties, the soil, and the number of fluid phases present. Second, additional quantitative
displacement experiments using a long soil column were performed to relate the mobilization of residual
organic liquid saturation in the saturated zone to wetting fluid flow rates. Third, pore and blob casts were
produced by a technique in which an organic liquid was solidified in place within a soil column at the
conclusion of a displacement experiment, allowing the distribution of fluid phases within the pore space
to be observed. The columns were sectioned and examined under optical and scanning electron
microscopes. Photomicrographs of these sections show the location of the organic phase within the
porous soil matrix under a variety of conditions. Fourth, etched glass micromodels were used to visually
observe dynamic multi-phase displacement processes in pore networks. Fluid movement was recorded
on film and video tape.
We found that the spatial distribution and saturation of organic liquid within the porous media
depends on a variety of factors, including: (1) the fluid properties of interfacial tension, viscosity, and
density; (2) the soil structure and heterogeneity; (3) the number of fluid phases present; and (4) the
fluid flow rates. Photomicrographs on a pore scale show that the residual organic liquid appears as
blobs, films, rings, and wedges of microscopic size, depending on these factors. The size, shape, and
spatial distribution of these blobs, films, rings and wedges affects the dissolution of organic liquid into the
water phase, volatilization into the air phase, and the adsorption and biodegradation of organic
components. These four processes are of concern in the prediction of pollution migration and the design
of aquifer remediation schemes.
Large amounts of residual organic are trapped as isolated blobs in the saturated zone. Smaller
amounts are 'trapped' as interconnected films, rings and wedges in the vadose zone, where the
movement and distribution of organic liquids is much more complex. Residual saturations are very
sensitive to soil textural heterogeneities. Even minor amounts of clay in an otherwise sandy soil may play
a significant role. In the saturated zone residual saturations are largely independent of fluid properties.
The rate of initial invasion of a non-wetting organic liquid may influence 'irreducible water saturations',
and subsequent residual organic liquid saturations.The term 'irreducible saturation' is misleading,
because the water phase is still interconnected by a water film and can be drained. In the vadose zone,
the residual organic liquid saturation, which is interconnected by a similar film, may also be drained and
is probably sensitive to the air flow rates in vacuum extraction and similar remedial schemes. The organic
liquid film becomes a population of non-connected coalesced lenses floating at the water-air interface,
when non-spreading organic liquids are involved. Residual saturation in the saturated zone can by
mobilized by increasing groundwater velocities or reducing interfacial tensions with surfactants. The
former is impractical while the later is potentially feasible, at least for partial mobilization.
- iv -
-------
TABLE OF CONTENTS
PAGE
ABSTRACT iv
TABLE OF CONTENTS v
LIST OF FIGURES viii
LIST OF TABLES xvi
LIST OF ABBREVIATIONS AND SYMBOLS xvii
ACKNOWLEDGMENTS xix
Section 1 INTRODUCTION 1
Nature Of The Problem 1
Scope Of Previous Work 3
Motivation For This Study 4
Objectives 5
Experimental Approach 6
Organization Of This Report 7
Section 2 CONCLUSIONS 8
The Saturated Zone 8
The Vadose Zone 14
Experimental Approach 15
Section 3 RECOMMENDATIONS 17
Issues For Future Research 17
Improvements In Experimental Equipment & Procedure 23
Section 4 CHARACTERIZING EXPERIMENTAL FLUIDS AND SOILS 27
Fluid Characterization 27
Soil Characterization 30
Section 5 SHORT COLUMN EXPERIMENTAL METHODS 43
Fluids And Soils 44
Experimental Apparatus 44
- v -
-------
Soil Packing 49
Deairing 50
Saturated Zone Experiments 51
Vadose Zone Experiments 54
Possible Sources Of Error 58
Limitations Of The Apparatus And Technique 63
Section 6 LONG COLUMN EXPERIMENTAL METHODS 64
Long Column Apparatus 64
Fluids And Soils 68
Column Packing And Degassing 68
Measuring Absolute Permeability, Relative Permeability And Saturation 70
Possible Sources Of Error 73
Limitations Of The Technique 75
Section 7 PORE AND BLOB CAST EXPERIMENTAL METHODS 76
Column Design 76
Fluid And Soil Characterization 77
Saturated Zone Experimental Procedure 81
Vadose Zone Experimental Procedures 87
Possible Sources Of Error 91
Limitations Of The Technique 95
Section 8 MICROMODEL EXPERIMENTAL METHODS 97
Micromodel Construction 97
Micromodel Experimental Procedure 103
Limitations Of The Technique 106
Section 9 SATURATED ZONE RESULTS AND DISCUSSION 108
Review Of Capillary Trapping Phenomena In Porous Media 110
Micromodel Flow Visualization Of Two Phase Displacement
And Capillary Trapping 124
Capillary Trapping And Residual Saturation In An Unconsolidated Soil:
The Sevilleta Sand 130
Residual Saturations For Various Organic Liquids 143
Residual Saturation For Different Soils 153
Influence Of The Initial Rate Of Organic Liquid Invasion 159
Influence Of Water Flow Rate On Residual Organic Mobilization .... 162
Organic Liquid Movement And Capillary Trapping
In A Heterogeneous Porous Media 168
- vi -
-------
Section 10 VADOSE ZONE RESULTS AND DISCUSSION 194
Review Of Capillary Trapping Phenomena In Porous Media 195
Micromodel Flow Visualization Of Three Phase Displacement
And Capillary Trapping 198
Capillary Trapping And Residual Saturation In An Unconsolidated Soil:
The Sevilleta Sand 208
Micromodel Visualization Of Capillary Trapping
Of A Non-spreading Organic Liquid 218
REFERENCES 224
APPENDIX A:
Quantitative Two-phase Residual Saturation Results,
With Styrene As The Organic Phase 236
APPENDIX B
Videotapes Of Micromodel Experiments 239
'APPENDIX C
Saturation Curves And Processed Data For The
Short Column Sevilleta Sand Experiments 240
APPENDIX D
Raw Data From The Short Column Experiments 261
- vn -
-------
LIST OF FIGURES
FIGURE TITLE PAGE
1-1. Migration pattern for an organic liquid more dense than water (left),
and less dense than water (right)
4-1. A simple experiment illustrating the relatively low volatility of Soltrol,
compared to water 29
4-2. Setup for the organic liquid and water capillary pressure - saturation
relationship 31
4-3. Contact angle measurement on a clean, smooth solid surface 33
4-4. Generic representation of relative permeabilities versus saturation for
the long column experiments 35
4-6. SEM photomicrographs of Sevilleta sand 37
4-5. Particle size analysis for three of the soils used in this study 37
4-8. A typical organic liqujd-water capillary pressure-saturation curve used
to determine wettability, in this case for Soltrol-130 in Sevilleta sand. 39
4-7. Typical Sevilleta sand capillary pressure-saturation curves 39
4-9. Relative permeability vs water saturation for water and Soltrol in the
Sevilleta soil 40
4-10. Relative permeability vs water saturation for water and Soltrol in the
Llano soil 41
4-11. Water saturation versus capillary pressure for Soltrol draining water
from Palouse loam 42
5-1. The short column apparatus 45
5-2. Air entry test for bottom end cap filter and seal 47
5-3. Saturated Zone Test Step 1: Organic liquid flood into a water saturated
column 52
5-4. Saturated Zone Test Step 2: Waterflooding at low velocity to reduce
the organic liquid to its residual saturation 53
5-5. Vadose Zone Test Step 1: Water being drained with air under an
applied suction 55
5-6. Vadose Zone Test Step 2: Organic liquid flood in a column already
drained by air 56
- viii -
-------
FIGURE TITLE PAGE
5-7. Vadose Zone Test Step 3: Organic liquid drained by air 58
5-8. Temperature range and its effect on the accuracy of results 60
6-1. Long column experimental setup 65
6-2. Long column construction details 67
7-1. Exploded view of the TFE Short Column 78
7-2. Viscosity of initiated styrene vs. time 79
7-3. Experimental setup of a styrene flood 84
7-4. Intermediate-wetting phase flood 90
7-5. Cross-section of styrene flooding into a water-saturated column with
organic wet walls: a) early time; b) late time 94
8-1. Pore-network pattern for the homogeneous model 98
8-2. Mirror construction 99
8-3. Mirror with enamel removed to reveal copper surface 100
8-4. Copper surface coated with Kodak Thin Film Resist (KTFR) 100
8-5. Pore-network pattern exposed with UV light onto coated copper surface. 101
8-6. Pore-network pattern exposed on the resin coating 102
8-7. Copper and silver layers under the pore network pattern removed to
reveal the underlying glass plate 102
8-8. SEM photomicrograph of the cross-section through a typical pore
within a micromodel 103
8-9. Photograph of network pattern showing the capillary barrier built into
one end of a micromodel 104
8-10. Two-phase micromodel experimental set-up 105
8-11. Three-phase micromodel experimental set-up 106
8-12. Pore-network pattern for the 'aggregated' model 107
8-13. Pore-network pattern for the heterogeneous 'stringer' model 107
9-1. Schematic of residual organic liquid trapped in the saturated zone .... 108
- ix -
-------
FIGURE TITLE PAGE
9-2. Two sketches illustrating fundamentals: a) cohesive forces acting on a
molecule inside a fluid and at its interface with another, immiscible fluid;
and b) hydrostatic equilibrium of two fluid phases in contact with a solid
phase 112
9-3. Capillary rise in a slim tube 113
9-4. Effect of pore aspect ratio on organic liquid trapping in a tube of
non-uniform diameter (after Chatzis et at., 1983) 114
9-5. Wetting fluid displacing a non-wetting fluid from a circular, high aspect
ratio pore under strongly wet conditions (after Wardlaw, 1982) 115
9-6. One fluid displacing another from a circular, high aspect ratio pore,
under intermediate wetting conditions (after Wardlaw, 1982) 115
9-7. Final condition after an advancing fluid displaced a retreating fluid from a
rough-walled pore under intermediate wetting conditions (after Wardlaw,
1982) 116
9-8. Sketches illustrating trapping mechanisms using the pore doublet model
(after Chatzis et al., 1983) 117
9-9. Relationship between residual saturation and capillary number for
sandstones and glass beads 122
9-10. Residual saturation in uniform glass beads due to variable capillary
number entrapment of the continuous non-wetting phase (dashed line),
and due to mobilization of non-wetting phase originally trapped at a low
capillary capillary number (solid line) 123
9-11. Micromodel study. In the upper photo (a) Soltrol displaced water from
the left (the top of the model) to the right (the bottom of the model),
yielding a residual (irreducible) wetting phase saturation. In the lower
photo (b) Soltrol was displaced by water from the right (the bottom
of the model) to the left (the top) yielding a residual non-wetting
residual saturation 125
9-12. Detail from Figure 9-11 showing conditions following the displacement
of the water by Soltrol (a. upper photo), and at residual non-wetting
phase saturation (b. lower photo). The area is located just below the
very center of the model 126
9-13. Detail from Figure 9-11 showing conditions following the displacement
of the water by Soltrol (a. upper photo), and at residual non-wetting
phase saturation (b. lower photo). The area is located near the top
of the model, just to the right of the centerline 127
9-14. A second experiment in the homogeneous micrpmodel, depicting
conditions at the end of the organic liquid invasion-compare to
Figure 9-11 a 128
9-15. Photomicrographs of (a) a singlet blob occupying one pore body in
the upper photo, and (b)a doublet blob occupying two pore bodies
and a pore throat in the lower photo 129
- x -
-------
FIGURE TITLE PAGE
9-16. Photomicrograph of a complex blob as observed in the micromodel. .. 130
9-17. Correlation of maximum Soltrol saturation (triangles), and residua!
Soltrol saturation (squares), to porosity in the Sevilleta sand 134
9-18. Sevilleta sand pore cast photomicrographs of (a) a singlet blob
occupying one pore body, in the upper photo, and (b)a doublet blob
occupying two pore bodies and a pore throat, in the lower photo 135
9-19. Sevilleta sand pore cast photomicrograph of a variety of blobs
including some that are complex and branching 136
9-20. Sevilleta sand blob cast photomicrographs of (a) non-branching
blobs, and (b) branching blobs 137
9-21. Photomicrograph of Sevilleta sand pore cast covering many pores. ... 139
9-22. SEM photomicrograph of many blob casts from the Sevilleta sand. ... 139
9-23. The spatial distribution of a single-component residual organic
liquid undergoing dissolution as a function of time when the local
equilibrium assumption is invoked 142
9-24. The spatial distribution of a single-component residual organic liquid
undergoing dissolution as a function of time when a local equilibrium
between the fluid phases is not reached. A dispersed zone forms
and grows until a steady state is reached 142
9-25. Residual organic liquid saturation as a function of the maximum
organic liquid saturation 147
9-26. Residual organic saturation for tested organic liquids in the Sevilleta
sand 148
9-27. Residual organic saturation as a function of interfacial tension (IFT). . . 149
9-28. Residual organic saturation as a function of non-wetting phase
viscosity 152
9-29. Residual organic saturation as a function of non-wetting phase density. 152
9-30. Residual organic saturation for Soltrol in tested soils 156
9-31. Residual organic saturation for Soltrol, as a function of organic carbon
content in different soils 157
9-32. Residual organic saturation for Soltrol, as a function of porosity in
different soils 158
9-33. Residual organic saturation for Soltrol, as a function of water saturated
hydraulic conductivity in different soils 158
- xi -
-------
FIGURE TITLE PAGE
9-34. Homogeneous model. In the upper photo (a) Soltrol displaced water from
the left (the top of the model) to the right (the bottom of the model), at
1.5 ml/min yielding a residual (irreducible) wetting phase saturation. In the
lower photo (b) Soltrol was displaced by water from the right (the bottom
of the model) to the left (the top), also at 1.5 ml/min, yielding a residual
non-wetting residual saturation 160
9-35. Residual organic saturation, in short and long columns, for Soltrol in the
tested soils. The error bars represent sample standard deviations for the
short column experiments 163
9-36 Relationship between Soltrol residual saturation and capillary number,
for Sevilleta sand and Llano sand 164
9-37. Relationship between Soltrol residual saturation and capillary number,
for Sevilleta sand and Llano sand 164
9-38. Hydraulic gradients required to initiate blob mobilization in porous
media of various permeabilities, for organic liquids of various
interfacial tensions 166
9-39. Recovery of residual saturation as a function of permeability and
hydraulic gradient for an interfacial tension of 10 dyne/cm 167
9-40. Non-wetting fluid near a material boundary: a) moving from coarse to
fine and encountering a capillary barrier; and b) moving from fine to
coarse and encountering a 'capillary end effect1 resulting in rivulets
of non-wetting flow across the boundary 170
9-41. Aggregated model. In the upper photo (a) Soltrol displaced water at a
rate of 0.075 ml/min, from the left (the top of the model) to the right
(the bottom of the model), yielding a residual (irreducible) wetting
phase saturation. In the lower photo (b) Soltrol was displaced by water
at the same rate, from the right (the bottom of the model) to the left
(the top) yielding a residual non-wetting residual saturation 172
9-42. Aggregated model detail from Figure 9-11, showing conditions
following the displacement of the water by Soltrol (a. upper photo),
and at residual non-wetting phase saturation (b. lower photo).
The area is located just below the very center of the model 173
9-43. Aggregated model detail from Figure 9-11, showing conditions following
the displacement of the water by Soltrol (a. upper photo), and at
residual non-wetting phase saturation (b. lower photo). The area is
located near the top of the model, just to the right of the centerline. 174
9-44. Aggregated model. In the upper photo (a) Soltrol displaced water from
the left (the top of the model) to the right (the bottom of the model),
at 1.5 ml/min yielding a residual (irreducible) wetting phase saturation.
In the lower photo (b) Soltrol was displaced by water from the right (the
bottom of the model) to the left (the top), also at 1.5 ml/min, yielding
a residual non-wetting residual saturation 175
- xn -
-------
FIGURE TITLE PAGE
9-45. Soltrol draining a horizontally-aligned 'coarse lens' micromodel from the
left. The photos show fluid distributions as a) the Soltrol was part way
through the model, and b) once the Soltrol had advanced completely
through the model 178
9-46. Water imbibing into the horizontally-aligned 'coarse lens' micromodel from
the right. The photos show fluid distributions as a) the water was part way
through the model, and b) once the water had advanced completely
through the model 180
9-47. Water imbibing into the vertical 'coarse lens' micromodel from the right
(the bottom of the model). The photos show fluid distributions after the
water had advanced completely through the model 182
9-48. Photograph of residual organic liquid saturation (shaded light) in a
heterogeneous sand pack. Water was flooded from right to left at a low
rate. Notice the high organic liquid saturation in the coarse lenses. ... 183
9-49. Photograph of residual organic liquid saturation (shaded light) in another
heterogeneous sand pack. Water was flooded from right to left. A high
rate of flow produced sufficient force to displace some organic liquid
from the coarse lenses 183
9-50 Random lenses of permeability k2 in a matrix of permeability k, 186
9-51 Uniform, parallel lenses of permeability kz in a matrix of permeability k,:
a) side view of several lenses, and b) cut-away view of one lense. ... 187
9-52 Pressure profiles in soil i for fluids A and B, with fluid A as a) the
wetting fluid, and b) the non-wetting fluid, i =1,2 coarse or fine 189
9-53. Pressure profiles for fluids A and B, with fluid A as the wetting fluid,
for a) the fine matrix and b) the more coarse lense 189
9-54. Critical flow rates needed to displace organic liquid from coarse lenses
as a function of permeability in the coarse lens (top), and in the fine
matrix (bottom) 191
10-1. Schematic of residual organic liquid trapped in the vadose zone 194
10-2. Diagram of spreading potential for a drop of organic liquid floating on the
air (gas)-water interface (after Adamson, 1982, and others). The water
is wetting, the air is non-wetting, and the organic liquid is intermediate
wetting 197
10-3. A conceptual plot of residual saturation for the wetting fluid (solid line)
and the non-wetting fluid (dashed line), as a function of the ratio of the
sum of viscous and buoyancy forces, to capillary forces 198
10-4. Initial vadose zone condition, with water drained by air to residual
(irreducible) water saturation 200
10-5. Detail of Soltrol invasion into a different vadose zone model. The
Soltrol was advancing by filling pores and by film flow 201
- xiii -
-------
FIGURE TITLE PAGE
10-6. Detail of steady state conditions after the Soltrol invasion into the
vadose zone model 201
10-7. Steady state conditions after the Soltrol invasion into the vadose zone
model 203
10-8. Detail of steady state conditions after the Soltrol has been drained by
air from the vadose zone model 204
10-9. Steady state conditions after the Soltrol has been drained by air from
the vadose zone model 205
10-10. Steady state conditions in a micromodel after a) Soltrol has been
drained by air from a vadose zone model, and b) Soltrol has been
displaced by water in a saturated zone model 206
10-11. Detail a thin organic liquid film located between the gas and water.
The photo represents steady state conditions after the Soltrol has
been drained by air from the vadose zone model 207
10-12. A photomicrograph of a pore cast thin section from the simulated
three-phase system in the Sevilleta sand. The middle of the photo
depicts a pore body filled with non-wetting phase (blue or dark grey).
Above it is thick 'film' of intermediate wetting phase (white or light grey),
that is 'smiling' into a pore throat. The pore throat is otherwise filled
with wetting fluid (red or light grey). Shown at 100X magnification. . .. 212
10-13. A photomicrograph of a pore cast thin section from the simulated
three-phase system in the Sevilleta sand. The middle of the photo
depicts a pore body filled with non-wetting phase (blue or dark grey).
It is surrounded by an intermediate wetting film (white or light grey),
that is 'smiling1 into the pore throat on the right, and filling most of
the pore throat to the left. Shown at 100X magnification 213
10-14. A photomicrograph of a pore cast thin section from the simulated
three-phase system in the Sevilleta sand. The middle of the photo
depicts a small pore body filled with non-wetting phase (blue or dark
grey). It is surrounded by an intermediate wetting film (white or light
grey), that is 'smiling' into the pore throats on the right, left and
below. The pore throats are filled with the wetting fluid (red or black
in this photo). Shown at 100X magnification 214
10-15. Inferred distribution of fluids in the vadose zone for the Sevilleta sand,
using Soltrol-130 as the organic liquid in individual short column
experiments. The dry zone data is taken from Table 10-1, while the
transition zone data is taken from Table 10-3. Results from
experimental trials 5 and 6 are suspect for reasons discussed in
the text 216
10-16. A photomicrograph of non-spreading PCE in a micromodel 220
10-17. A photomicrograph of non-spreading PCE in a micromodel. This is
a close-up to the photo shown in Figure 10-16 221
10-18. A photomicrograph of Soltrol in a micromodel. The geometry is
similar to that depicted for PCE in Figure 10-16 222
- xiv -
-------
FIGURE TITLE PAGE
C-1. Soltrol-water saturation curve for SW trial 7 241
C-2. Soltrol-water saturation curve for SW trial 8 242
C-3. Soltrol-water saturation curve for SW trial 9 243
C-4. Soltrol-water saturation curve for SW trial 10 244
C-5. Soltrol-water saturation curve for SW trial 11 245
C-6. Soltrol-water saturation curve for SW trial 12 246
C-7. Soltrol-water saturation curve for SW trial 13 247
C-8. Soltrol-water primary drainage curves for SW trials 7-13, minus trial 8. 248
C-9. Soltrol-air saturation curve for SA trial 1 249
C-10. Soltrol-air saturation curve for SA trial 2 250
C-11. -Comparison of Soltrol-air primary drainage curves for SA trials 1 & 2. . 251
C-12. Air-water saturation curve for AW trial 1 252
C-13. Air-water saturation curve for AW trial 2 253
C-14. Air-water saturation curve for AW trial 3 254
C-15. Comparison of air-water primary drainage curves for SA trials 1-3. ... 255
- xv -
-------
LIST OF TABLES
TABLE TITLE PAGE
1-1. Migration pattern for an organic liquid more dense than water (left),
and less dense than water (right) 2
4-1. Measured properties of fluids used in experiments 28
4-2. Relationship between wettability measurement methods 34
7-1. Properties of fluids used in pore and blob cast visualization
experiments 80
7-2. Absolute viscosities of selected organic liquids 85
9-1. Soltrol residual saturation and other measurements in Sevilleta sand,
for three temperature dependent categories 132
9-2. Summary of Soltrol / Sevilleta sand saturated zone results 133
9-3. Summary of kerosene / Sevilleta sand saturated zone results 144
9-4. Summary of gasoline / Sevilleta sand saturated zone results 144
9-5. Summary of n-decane / Sevilleta sand saturated zone results 145
9-6. Summary of p-xylene / Sevilleta sand saturated zone results 145
9-7. Summary of PCE / Sevilleta sand saturated zone results 146
9-8. Average values for different organic liquids in the Sevilleta sand
saturated zone experiments 146
9-9 The interfacial tension of some priority pollutants with water 151
9-10 Summary of Soltrol / Traverse City soil saturated zone results 154
9-11. Summary of Soltrol / Llano soil saturated zone results 154
9-12. Average values of measured properties and saturations for different
sandy soils, in the saturated zone experiments run with Soltrol 156
9-13. Long column data for two different sandy soils run with Soltrol 163
9-14. Measurements of bulk residual organic saturations in two heterogeneous
packings of the Sevilleta sand. The sand was divided into a coarse and
a fine fraction, and the coarse fraction was packed into the column as
cylindrical lenses within a matrix of the fine fraction 184
10-1. Results from the vadose zone column experiments. Soltrol-130 was
used as the organic liquid and Sevilleta sand served as the soil 209
10-2. Relative density differences and interfacial tensions in the vadose
zone and saturated zone 211
10-3. Results from vadose zone column experiments performed to examine
the saturation distributions in the transition zone between the saturated
zone and the vadose zone. The media was Sevilleta sand and the
organic liquid was Soltrol 215
A-1. Two-phase TFE column residual saturation results; styrene was used
unless otherwise noted 237
C-1. Numerical values of measured saturations, pressures, and
temperatures, (continued on next four pages) 256
- xvi -
-------
LIST OF ABBREVIATIONS AND SYMBOLS
g or g gravitational constant
k intrinsic permeability of the porous media
km relative permeability for the organic liquid phase
&„, relative permeability for the water or wetting phase
n porosity
QW specific discharge in the water phase
rt radius of a capillary tube
2 elevation
A gross cross-sectional area of, eg, a column
Hc capillary head ( = PC /Qg)
IFT interfacial tension between two fluid phases = a
Jw hydraulic gradient in the water phase
Kw saturated hydraulic conductivity for water
Ms mass of soil
Mw mass of water
NB Bond number; represents the ratio of gravitational forces to viscous forces for a
multi-phase flow situation
Nc capillary number; represents the ratio of capillary forces to viscous forces for a
multi-phase flow situation
PC capillary pressure
Pnw non-wetting phase pressure
P0 organic liquid phase pressure
PW water or wetting phase pressure
Q total discharge
R radius of curvature
Sa air saturation
S0 organic liquid saturation
Sor residual organic liquid saturation
- xvii -
-------
S'or maximum, low capillary and Bond number, two-phase organic liquid saturation
Sw water or wetting phase saturation
S^ residual water or wetting phase saturation
SM irreducible water saturation = Sw
V volume
Vp pore volume
Vs soil volume ,
i i
Vt total volume
Vv void volume
7 surface tension between air and water
0 contact angle
f*o dynamic viscosity in the organic liquid
ft* dynamic viscosity in the water phase
Qb soil bulk density
Qo organic liquid density
Qs particle density
Qw water density
a interfacial tension between two fluid phases
oa air-water interfacial tension = surface tension, 7
oao air-organic liquid interfacial tension
Om, = o»« organic liquid-water interfacial tension
A(> density difference between two fluids
AP a pressure drop
2 spreading coefficient
- xvin -
-------
ACKNOWLEDGMENTS
This work described in this report was sponsored by the R. S. Kerr Laboratory, Office of Research
And Development, U.S. Environmental Protection Agency, Ada, Oklahoma, under Grant Number EPA
CR-813571-01-0, and the New Mexico Water Resources Research Institute, Las Cruces, New Mexico,
under Grant Number NMWRRI 1345648. Additional financial support was provided by the Research
Branch, Water Resources Division, U.S. Geological Survey, Menlo Park, California.
The authors would like to acknowledge the assistance of: Mary Graham, who introduced us to the
techniques of manufacturing etched glass micromodels; Robert Mace, who ran the short column Llano
soil experiments; and Jamine Wan, who took some of the photomicrographs. Dr. Norman Morrow of New
Mexico Tech's Petroleum Research Recovery Center freely shared his understanding of similar problems
encountered in both petroleum reservoir engineering and soil physics. Many of the experiments
described in this report are progeny of earlier experiments performed by Dr. Morrow and others in
petroleum research laboratories.
The authors would also like to acknowledge the comments of our various reviewers for the
Environmental Protection Agency, the New Mexico Water Resources Research Institute, and Steve
Conrad's Ph.D. Committee. These reviewers include Fred Phillips, Dan Stephens, Norm Morrow, Jack
Parker, Danny Reible and Robert Bowman.
- xix -
-------
SECTION 1
INTRODUCTION
NATURE OF THE PROBLEM
Many hazardous waste sites, and most leaking underground storage tanks, involve non-aqueous
phase organic liquids (e.g., Burmaster and Harris, 1982; Chaffee and Weimar, 1983; Convery, 1979;
EPA, 1980,1982,1983; Feenstra and Coburn, 1986; J.R.Roberts et al., 1982; Jercinovic, 1984; Maugh,
1979 ; McKee et al., 1972; Villaume, 1988; Williams and Wilder, 1971). Usually released at or near the
surface, these organic liquid contaminants move downward through the vadose zone toward the water
table. Migrating as a liquid phase separate from the air and water already present in the vadose zone,
some of the organic liquid is immobilized within the pore space by capillary forces. The remainder
passes on, and if the volume of organic liquid is large enough it eventually reaches the water table. If it is
less dense than water the organic liquid spreads laterally along the water table (see right side of Figure
1-1). If the organic liquid is more dense than water, it continues to move downward into the saturated
zone (the left side of Figure 1-1). In both cases the organic liquid usually migrates down-gradient with
the ambient groundwater flow, although dense organic liquids may migrate in other directions as they
encounter dipping barriers. In the saturated zone, which is mostly below the water table and includes the
capillary fringe, more organic liquid is immobilized by capillary forces (Schwille, 1967, 1981, 1984, 1988;
van Dam, 1967; de Pastrovich et al., 1979; Schiegg, 1980; Wilson and Conrad, 1984; Albertson et al.,
1986; Schiegg and McBride, 1987). Here the immobilized organics remain as small, disconnected
pockets of liquid, sometimes called 'blobs', no longer connected to the main body of organic liquid.
The immobilized volume is called the 'residual oil saturation' in petroleum reservoir engineering
(Taber, 1969; Morrow, 1979; Chatzis ef a/., 1983; Anderson, 1987b), and is measured as the volume of
organic liquid trapped in the pores relative to the volume of the pores. Organic liquid at residual
saturation can occupy from 15% to 50% of the pore space in petroleum reservoir rocks under conditions
that are equivalent to those in the groundwater saturated zone (Melrose and Brandner, 1974). At a spill
or hazardous waste site the entire volume of organic liquid can be exhausted by this immobilization,
although if the volume of organic liquid is large enough, it continues to migrate down-gradient where it
becomes a threat to the safety of drinking water or agricultural water supplies (Schwille, 1967,1981; de
Pastrovich ef al., 1979). This report refers to the immobilized organic liquid as 'residual organic liquid'.
As described in detail in sections 9 & 10 of this report, the actual spatial distribution of the residual
saturation within the pore space is completely different in the vadose and saturated zones.
The organic liquid phase is sometimes referred to as being immiscible with water and air. Although
that expression is used here, it is important to realize that small concentrations of the various
components of the organic phase volatilize into the air phase and dissolve into the water phase. A 'halo'
of dissolved organic components precedes the immiscible phase in its migration (Figure 1-1). Even
when the so-called immiscible organic liquid has been immobilized by capillary trapping, the passing
groundwater dissolves some of the residual. In effect, the organic liquid phase acts as a continuing
-------
hazardous waste site
ground surface
capillary fringe
——J
water table
leaking tank
vapor
ih<
organic
VADOSE
ZONE
floating organic
liquid
residual
organic
liquid
saturation
dissolved
organic
SATURATED
ZONE
« dense organic liquid
,nmt mini IIIIHI ttmn innii
in IHIIII IIIIHI IHIHI IIIIIH mi
IIIIIH IIHIII IIHIII imm IIHIII inim mini IIIIIH mini IIIIIH IHIIII mini IHIIII mini mini IIIIIH IIIIIH IHHII IIHIII IIIIHI in
1111 Illllll milll IHIIII IIHIII IIIIIH Illllll IIIIHI IIHIII Illllll Illllll HHIII IIHIII IIIIHI IHHII HHIII IHIIII Illllll Illllll IHIIII IIHIII
IHIIII IIIIIH IIIIIH IIIIIH limit IHIIII Illllll IIHIII IIHIII IHHII Illllll 11111111111111 IIIIIH Illllll HHIII 1111111 Illllll Illllll 1111111 III
III Illllll Illllll IIIIHI IHIIII Illllll Illllll Illllll IHHII III
Illllll IIIIIH limit HHIII IHHII Illllll Illllll Illllll limil
m iwm IIIIIH IIHHI IIHIH mmi HHIII IIIIIH IIHIII m
FIGURE 1-1. Migration pattern for an organic liquid more dense than water (left), and less dense
water (right).
source of dissolved organic pollutants (eg, Tuck et al., 1988). Similarly, in the vadose zone, the residual
organic liquid that volatilizes into the air phase migrates by gaseous diffusion and advection, becoming a
source of organic components to air or water pollution, and a possible explosion hazard. In large spills
and leaks it is apparent that most of the liquid organic remains as a liquid, some is volatilized, and a little
is dissolved. However small in volume, the volatilized or dissolved components are usually the ones that
cause problems. The liquid organic phase acts as a reservoir of additional organic to replenish the air
and water phases with dangerous and/or toxic material. Clearly, the source of the dissolved or gaseous
organic constituents — the liquid organic phase — must be removed or isolated in order to restore a
polluted aquifer.
There is no wholly effective mechanism to remove the residual organic liquid. Waiting for the residual
to dissolve can take several decades. In the vadose zone, induced volatilization may help reduce the
residual volume for lighter organics, but is not effective for heavier ones (Burris et al., 1986). Engineered
- 2 -
-------
removal is usually attempted hydraulically, by sweeping the organic liquid out with water, or biologically,
by encouraging the consumption of the organic constituents by the soil microbial community. This last
process, biodegradation, is the focus of current research and several recent restoration efforts. It is
seldom tried alone, for the microbes generally consume only the dissolved organics. Moreover, some
organic chemicals are extremely resistant to biodegradation. PCB's, for example, may biodegrade very
slowly, or not at all in the subsurface (J.R.Roberts ef al., 1982). Hydraulic sweeps remain a major
component of any attempt to remove organic liquids although, commonly, hydraulic sweeps fail to
remove all the liquid organic phase, often leaving a significant quantity of residual organic liquids behind
(Wilson and Conrad, 1984). There is, of course, another removal option often used for small pollution
events: excavate the site and dispose of or treat the contaminated soil. For large sites this alternative is
unfeasible. Since there is no panacea for the removal of organic liquids, containment is often adopted as
part of a restoration strategy. Hydraulic containment (e.g., Wilson, 1984), often in combination with
structural barriers such as a slurry wall, is becoming standard practice.
SCOPE OF PREVIOUS WORK
Development of improved technologies to clean up organic pollutants depends in large part on
developing an ability to understand and predict the migration of liquid, vapor, and dissolved organics.
Liquid organics move through a water and sometimes air filled porous soil, as a separate phase, under
the influence of viscous, gravity, and capillary forces. Dissolved organics move in the water phase and
are subject to advection, dispersion, biodegradation, and adsorption onto soil particles. Organic vapors
in the air phase are subject to similar mechanisms. A few of these major transport mechanisms are fairly
well understood today, principally those associated with the behavior of dissolved organics. McCarty et
al. (1981) and P. V. Roberts et al. (1982) have reviewed the progress of this research.
In contrast, the organic liquid phase transport mechanism has been virtually ignored by the research
community in the United States, although it has been the subject of empirical studies in Europe (e.g.,
Albertsonef a/., 1986; Schwille, 1967, 1981, 1984, and 1988; van Dam, 1967; Schiegg, 1980). Recently,
however, American researchers have obtained some laboratory results. Convery (1979) ran gravity
drainage experiments on a long column to relate organic liquid retention in the vadose zone with grain
size and sorting. Eames (1981) used a short soil core centrifuging method to measure residuals in the
vadose zone. Eckberg and Sunada (1984), and Ferrand ef a/. (1986) used gamma radiation attenuation
and bulk soil electrical resistivity to measure three-phase fluid saturations at various times and at various
elevations above a water table following a simulated petroleum spill. The experimental procedure
allowed a petroleum 'spill' to be tracked as it moved through the vadose zone to the water table. Gary ef
al. (1989) performed experiments to test the ability of multiphase flow theory to predict the infiltration
and redistribution of wetting and non-wetting fluids. They met with limited success. Lenhard and Parker
(1988b) used theoretical three-phase saturation-pressure relationships to estimate the volume of oil in
soils given observed fluid levels in monitoring wells.
Some simple numerical simulations of multi-phase transport have been developed. These focus on
immiscible transport of continuous phases. Residual organic liquids, trapped by capillary forces, are
often ignored, although they are sometimes treated as a source of dissolved contamination. This
- 3 -
-------
research effort mirrors the state of the art of petroleum engineering's 'black oil' models. A few
researchers (notably Baehr and Corapciaglu, 1984 and 1987; Corapciaglu and Baehr, 1987; Abriola and
Pinder, 1985a,b; and Finder and Abriola, 1986) have looked into interphase transfer, including the
volatilization and solution of organic components, using computer simulations. This again reflects the
state of the art in petroleum engineering, where so-called compositional models are used to examine
enhanced recovery techniques.
Parker ef a/. (1987) have proposed a model to estimate the functional relationships between fluid
pressures, saturations, and permeabilities of two- or three-phase porous media systems, and these
functional relationships have been implemented in a multi-phase numerical flow model (Kuppusamy ef
a/., 1987). The model has since been extended to include the effects of hysteresis and non-wetting
phase trapping (Parker and Lenhard, 1987; and Lenhard and Parker, 1987a). The results of concurrent
laboratory work were used to validate the model (Lenhard and Parker, 1987b,1988a,1989; and Lenhard
et a/., 1988).
Petroleum engineering's long history of research into improving recovery from petroleum reservoirs
may be applied to rehabilitating fresh-water aquifers polluted by organic liquids. Through over forty years
of experimentation, petroleum engineering has amassed considerable expertise in multi-phase
transport, the mechanics of oil phase capillary trapping, and oil recovery. To date, relatively little of this
technology has been applied to recovering organic hazardous wastes and petroleum hydrocarbons
released in the near-surface environment. The petroleum literature on residual oil saturation is reviewed
in papers by Anderson (1988), Chatzis et al. (1983), Melrose and Brandner (1974), andTaber (1969). In
groundwater hydrology we too are concerned with the capillary trapping of residual saturation, and with
its removal. However, unlike petroleum engineers, we are also concerned with the mechanisms that
initially brought the 'oil' into the aquifer in the first place. In the 'oil patch' that is the province of
petroleum geologists, and it involves issues that are quite different than ours. Consequently, we can
expect little help from the oil patch on these mechanisms.
MOTIVATION FOR THIS STUDY
Residual organic liquid saturation often constitutes the major volume of the organic pollution, and
acts as a continual source of dissolved or vapor phase organics(Wilson and Conrad, 1984). In particular,
there is a need to understand how the residual organic liquid is trapped and how it can be hydraulically
mobilized or otherwise removed. As shown in Sections 9 and 10 of this report, the residual organic liquid
appears to form blobs, films, wedges and rings of microscopic size, depending on the presence of other
fluids, the pore geometry, the surface wetting of the solids, and soil heterogeneity. The term wetting
refers to the relative affinity of the solid surface for the available fluids. Water is normally the wetting fluid
in most soils. Organic liquid is normally non-wetting relative to water, and wetting relative to soil gas. The
size, shape, and spatial distribution of these blobs, films, wedges and rings affects the dissolution of
organic liquid into the water phase, volatilization into the air phase, and the adsorption and
biodegradation of organic components. The presence of residual organic liquid also affects the relative
permeability versus saturation curves used in numerical simulation codes of fluid movement and
pollution migration. A paucity of experimental results regarding these issues makes site characterization
- 4 -
-------
conjectural, predictive modelling unreliable, and remediation design of organic liquid leak or waste sites
less effective than might be possible.
OBJECTIVES
The goal of this study was to better understand the basic physical mechanisms controlling the
movement, and especially the capillary trapping, of organic liquids in soils and groundwater. Emphasis
was on relating the various mechanisms to the issues of contaminant movement, characterization, and
remediation. This broad goal was broken down into two sets of specific research objectives, addressing
issues relevant to the saturated and vadose zones, respectively:
The Saturated Zone
Assuming that water is wetting and the organic liquid is non-wetting, our research objectives for
saturated zone conditions were to:
•, conduct a literature review of basic concepts, including non-wetting phase
capillary trapping and mobilization mechanisms, and petroleum experience;
•2 conduct experiments that permit the visualization of two-phase fluid flow and
capillary trapping, and record the visualizations on film and videotape;
•3 perform a detailed study of two phase flow capillary trapping and non-wetting
phase residual saturation in a typical unconsolidated soil, testing the
hypothesis that its behavior can be predicted from previously published results
from the petroleum engineering literature;
•4 compare non-wetting phase residual saturations for various organic liquids,
testing the hypothesis that residual saturation is largely independent of organic
liquid composition for expected conditions in hydrology;
•5 compare non-wetting phase residual saturations for various soils, testing the
hypothesis that residual saturations should be similar in soils that have a similar
grain size distribution;
•e investigate how the rate of initial invasion of a non-wetting organic liquid may
influence irreducible water saturations and, later, organic residual saturations;
•7 investigate the possible hydraulic mobilization of non-wetting phase residual
organic liquid, by increasing groundwater velocities, testing Wilson and
Conrad's (1984) conclusion that this is largely an unrealistic aquifer
remediation alternative unless interfacial tensions are reduced significantly;
and
•s test the hypothesis that porous media heterogeneity can dominate
displacement and trapping mechanisms.
- 5 -
-------
The Vadose Zone
Our research objectives for vadose zone conditions were to:
«! conduct a literature review of basic concepts, including capillary trapping
mechanisms, mobilization issues, and petroleum experience;
•2 conduct experiments that permit the visualization of multi-phase fluid flow and
capillary trapping, testing the hypothesis that spreading organic liquids
typically form a film between the water and air phases;
•3 perform a detailed study of capillary trapping and residual saturation in a typical
unconsolidated soil, testing the hypothesis that organic liquid residual
saturations are significantly lower in the vadose zone, than they are in the
saturated zone; and
•4 conduct experiments that permit the visualization of capillary trapping for
non-spreading organic liquids, testing the hypothesis that non-spreading
organic liquids behave differently than non-spreading organic liquids.
EXPERIMENTAL APPROACH
The problem was approached experimentally in four ways:
1. Quantitative displacement experiments using short columns were
performed to relate the magnitude of residual organic liquid saturation to fluid
and soil properties, and to the number of fluid phases present (i.e., both
saturated and vadose zone conditions).
2. Quantitative displacement experiments using long columns were performed
under two-phase saturated zone conditions, yielding water and organic liquid
relative permeabilities. In these experiments, reductions of residual organic
saturation were correlated to the pressure gradient applied in hydraulic
sweeps, and the potential for hydraulic mobilization of residual 'blobs' was
investigated.
3. Pore and blob casts were produced for saturated zone conditions by a
technique in which the organic liquid was solidified in place within a soil column
at the conclusion of a displacement experiment, allowing the distribution of
organic liquid to be observed. The polymerized organic phase was rigid and
chemically resistant. Following polymerization, the water phase was removed
and replaced by an epoxy resin. The solid core, composed of soil, solidified
styrene (the organic phase), and epoxy resin (the water phase), was cut into
sections to show the organic liquid phase in relation to the soil and the water
phase. The sections were photographed under an optical microscope.
Although polymerization only gave a 'snapshot' of the displacement process,
- 6 -
-------
it offered the advantage of seeing organic liquid in its 'natural habitat' (i.e.
within a soil) as compared to that observed in etched glass micromodels.
Sometimes, instead of replacing the water with epoxy resin, the solid matrix of
the soil column was dissolved with hydrofluoric acid, leaving only the hardened
organic liquid. The solidified organic phase was then observed under a
scanning electron microscope (SEM) and photographed. For vadose zone
conditions, styrene and epoxy liquids were sequentially applied, drained and
hardened in an attempt to simulate proper fluid distributions above the water
table. The resulting pore casts were photographed under an optical
microscope.
4. Etched glass micromodels were used to observe dynamic multi-phase
displacement processes. Micromodels provide two-dimensional networks of
three-dimensional pores. They offer the ability to actually see fluids displace
one another in both a bulk sense and also within individual pores. Although
displacements are known to be dependent upon a variety of factors, this report
describes micromodel experiments that focused on only three: (1) the fluid
flow rate, (2) the presence of heterogeneities, and (3) the number of fluid
phases present. The experiments were photographed and videotaped.
To interpret the experiments in heterogeneous material, we also developed a new but simple theoretical
model of multiphase flow and capillary trapping. The model is based on the interplay between viscous
and capillary forces.
ORGANIZATION OF THIS REPORT
Following this introduction are two sections (2 & 3) that summarize the report conclusions and
recommendations. The next five sections (4 through 8) detail fluid and soil characteristics, and the
experimental methodology, used for each of the experimental approaches outlined above. These
sections contain detailed information that may be used by future investigators wishing to verify or extend
the results of this study. The reader more concerned with results than methods can probably skip them.
The last two sections (9 & 10) describe experimental results for saturated zone and vadose zone
conditions, respectively. These sections contain a large number of photomicrographs that visualize
multiphase flow and residual saturation.
- 7 -
-------
SECTION 2
CONCLUSIONS
The intent of this study has been to better understand the basic physical mechanisms controlling the
movement, and especially the capillary trapping, of organic liquid pollutants in soils and groundwater.
This goal was pursued by observing organic liquid behavior in laboratory soil columns and etched glass
micromodels, built to represent various conditions in the subsurface. Fluid movement observed in these
studies is explained in terms of three basic forces acting on the fluid phases: capillary forces, viscous
forces associated with flow, and the effect of gravity.
Below we review our conclusions for the saturated and vadose zone experiments. The most
interesting series of observations has illustrated how the interplay between capillarity and soil
heterogeneities profoundly influence the migration and trapping of an organic phase in a multi-phase
system. We end with several conclusions concerning experimental procedures.
We strongly recommend that the reader peruse the photographs in Sections 9 and 10, in order to
appreciate these conclusions
THE SATURATED ZONE
• As organic liquid moves through the saturated zone, a significant portion is left behind,
trapped by capillary forces. In water wet soils this residual organic liquid saturation is
trapped as discontinuous 'blobs'. Photomicrographs taken of micromodel pores, blob
casts and pore casts indicate that trapped blobs form a variety of shapes: from spherical
shaped singlets occupying one pore, to complex, branching multi-pore blobs.
• Capillary trapped blobs are trapped by two different mechanisms. 'Snap-off creates
singlets, while by-passing creates doublets and much more complex, branched shapes.
By-passing is also the major mechanism when soil heterogeneities are involved.
• Large amounts of residual organic are trapped by capillary forces. We observed residual
saturations that ranged from 14% to 30% in unconsolidated sands, and possibly higher in
heterogeneous sand packs. If these estimates are typical for most organic liquids and
sandy soils, then there is a tremendous storage capacity for organic liquid pollutants in the
saturated zone. For example, expressed in terms of volumetric retention (Equation 9-5),
the Sevilleta sand has the capacity to store over 90 liters of organic liquid per cubic meter of
soil, in the saturated zone. A single 10,000 gallon spill of an organic liquid could be
absorbed in about 420 m3 of the saturated soil. This volume corresponds to a cube of soil
with sides only 7.5 meters in length. Even for soil residual saturations of only a third of this,
the capacity is still large.
-------
• The size and shape distribution of the trapped blobs can influence interphase transfer.
For large, branched blobs, in which perhaps only the ends of the ganglia are exposed to the
bulk flow of water, dissolution of soluble components may be limited by diffusion through
thin films of water lining pore walls. From our review of photomicrographs we've concluded
that:
o Mass transfer coefficients used in the mathematical models of partitioning often
employ the analogy of an equivalent spherical blob (e.g., Pfannkuch, 1984).
Certainly singlets fit this model, but an appropriate definition of equivalent sphere
size has yet to be proposed for the population of more complex and branched
blobs.
o The position of a blob within the pore space has a strong influence on mass
transfer between phases. If a large portion of the blob surface is in contact with
only a thin film of water, as certainly seen in our pore casts and micromodels, then
the transfer rate may be limited by advection or diffusion in the film. Videotaped
micromodel observations indicate that the flowing water moves around the blobs,
through the unoccupied pores, with little water movement in the films to help
advect organic components away.
o As components of the organic phase partition from it, a blob gets smaller, and the
residual organic saturation is reduced over time. Water phase relative permeability
increases.
o Mass transfer may or may not be governed by local equilibrium. When groundwater
velocities are high enough or the dissolution kinetics are slow enough, the local
equilibrium assumption no longer holds. In this sense the rate of mass transfer
between phases strongly depends upon the distribution of the residual saturation.
For large, single component, branched blobs mass transfer can be limited by a
diffusional process. Substantial portions of these blobs are only in contact with thin
water films along the walls of pores. Diffusion through the films to the main body of
passing groundwater can be the rate-limiting step in solubilizing organic liquid into
the water phase. If the flowing water-phase dispersed zone is large enough, or the
zone of trapped organics is relatively small, the solubility limit of the organic liquid
in water may never be reached.
o Dissolution from a multi-component organic phase is more complex. For large
branched blobs or by-passed regions of multi-component liquids, the mass transfer
rate for a given component can become limited by rate of diffusion of that
component within the organic phase to the organic/water interface. Since the major
portion of the transfer may be occuring at the ends of these tortuous blobs,
intra-blob diffusion may be small.
- 9 -
-------
The size and shape distribution of the trapped blobs can influence microorganisms. Our
review of photomicrographs has lead us to conclude that:
o Microorganisms attached to the pore wall, at the water-solid interface, experience
an environment that depends on their location. In blob-filled pore throats, the
organisms have ready access to organic components that partition into the aqueous
phase from the nearby blob surface. However, these organisms may not have a
ready supply of other substrates, which can only diffuse (or flow) slowly through the
thin water film that lines the pore throat. In blob-filled pore bodies that have a
surrounding water film, the same situation applies, as it does for microorganisms
attached to the blob surface in these regions of thin aqueous films. In the other
pores, occupied by flowing water, wall-attached microorganisms are readily
exposed to available dissolved substrates, subject only to upstream substrate
re-supply.
o Migration of microorganisms is also probably influenced by the spatial distribution of
blobs. It should be difficult for a seed population to find its way into regions with
thin aqueous films because of their low flow rates and tortuous diffusion paths. In
any event, most organisms would probably not thrive in these stagnant regions
because of the substrate re-supply problem and the possibility of toxicity due to
locally high concentrations of dissolved organics.
Prediction of residual saturation levels in a given soil is uncertain. A wide variety of
residual organic liquid saturation measurements were made on the Sevilleta sand, an
unconsolidated, aeolian sand. Its texture is much more like glass beads than sandstone,
both of which had been previously investigated by petroleum engineers. We examined the
hypothesis that, because of its similar texture, the Sevilleta sand should behave more like
the glass beads. Uniform glass beads exhibit a residual saturation of 14-16%. The Sevilleta
mean residual saturation was 27.1%, almost 10% greater than expected. Singlet blobs are
common in glass beads. Photomicrographs of Sevilleta blob and pore casts revealed a
much larger population of complex, branching blobs. Residual saturations in the Sevilleta
sand did not behave as expected.
Residual saturations should, but may not be similar in soils that have a similar texture
and grain size distribution. We examined residual saturations in a variety of sandy soils.
Two soils of very similar texture and grain size (the Sevilleta sand and the Traverse City
sand) yielded significantly different residual saturations (27.1% and 17.6%, respectively).
Two soils with somewhat different textures and grain sizes (the Traverse City sand and the
coarser, less uniform Llano sand) had very similar residual saturations ( 17.6% and 15.8%,
respectively).
- 10 -
-------
• Even minor amounts of clay or silt in a soil may play a significant role in the observed
residual saturation. The Sevilleta sand may had 2% clay and silt, by weight. The Traverse
City and Llano sands were essentially clay and silt free. We hypothesize that the clay and silt
content of the Sevilleta was responsible for the high residual saturation and complex,
branching blobs that we observed. One possible mechanism is clay swelling in the pore
throats, leading to increased by-passing. Another explanation is that the fine material may
have settled into microlayers during the wet column packing procedure. This type of
heterogeneity could lead to increased water by-passing of the coarser material. If this latter
explanation is true, then this experiment demonstrates that even minimal soil structure and
heterogeneity can have a dramatic influence on behavior.
• Models of organic liquid movement employ prescribed values of residual organic liquid
saturation as a soil property. It would be convenient to be able to estimate residual
saturations from more primitive soil properties, such as grain size distribution (see, eg, Soil
and Celia, 1988) .These experiments indicate that textural considerations alone may lead
to unreliable estimates and erroneous models of residual saturation.
• Fine-grained, water-wet soils (which do not shrink in the presence of organics) can
serve as an effective barrier to organic liquid movement in the subsurface. This was
indirectly demonstrated by our inability to inject an organic phase into the Palouse loam.
• Under low capillary and Bond number conditions residual saturation in a given soil is
largely independent of organic liquid composition. We measured residual saturations for
a variety of multi-component and single component organic liquids: Soltrol-130, gasoline,
kerosene, p-xylene, n-decane, and tetrachloroethylene (PCE). In repeated quantitative
short column experiments we found no correlation between residual saturation, and
viscosity, density, or interfacial tension for the six organic liquids tested. The residual
saturation was virtually identical for all of the organic liquids.
• Residual saturations appear to be relatively insensitive to fluid properties and very
sensitive to soil properties (and heterogeneities). This is true only under low capillary and
Bond number conditions.
• At a particular contamination site it may seem reasonable to directly measure residual
saturation of whatever organic liquid was spilled. We recommend that tests be conducted
with an ideal fluid, using uncontaminated soil from the site. Unless some odd wetting
behavior is anticipated, or unless some interaction between fluids, or between fluids and the
solid is expected, it is probably preferable to chose a fluid which has:
1. a sufficient density difference with water;
2. low solubility;
3. low volatility; and,
4. low toxicity.
- 11 -
-------
The easiest fluids for us to use were Soltrol and decane.
• The rate of initial invasion of a non-wetting organic liquid may influence 'irreducible
water saturations' and, later, organic residual saturations. The amount of organic liquid
that is ultimately trapped is strongly dependent on the nature of original emplacement.
• The term 'irreducible water saturation' is misleading. The term is often used to represent
the wetting phase residual saturation. However, even at residual saturation the wetting
phase is continuous, and is composed of an interconnected network of films, pendular
rings, and wedges. This is a considerably different situation than the discontinuous 'blobs'
pictured in this report for the non-wetting phase residual. The wetting phase liquid can
move through its interconnected network, draining the films and rings, and reducing the
residual wetting phase saturation. If there is no barrier to organic phase migration, and the
organic phase pressures cannot build-up, then by-passed pockets of water, constituting a
major portion of the residual water saturation, may not drain. The residual water saturation
is not a single 'irreducible' value, but depends on boundary conditions.
• Residual organic liquid can be mobilized by increasing groundwater velocities. Residual
organic liquid blobs are easier to mobilize for lower interfacial tensions, higher permeability
soils, and in the lab. They are harder to mobilize for higher interfacial tensions, lower
permeabilities, and in the field where there are severe practical constraints on the hydraulic
gradient. Although it may be possible, intentionally or by accident, to mobilize some of the
residual, it is difficult to get it all.
• The groundwater velocities and capillary numbers needed for mobilization make
hydraulic remediation of contamination unrealistic in the saturated zone. Several
schemes have been published in the literature and implemented in the field for hydraulically
sweeping organic liquids from polluted aquifers. These schemes are presumably meant to
sweep out the continuous organic liquid, knowingly leaving behind the residual. More often it
seems that naivete prevails, and many designers assume that as long as ground water is
flowing toward a collection system, eventually all of the organic liquid will make it. No matter
how long one waits, unless gradients are increased above the critical level, none of the
residual will be hydraulically removed (Wilson and Conrad, 1984). The critical velocity and
capillary number(see Equation 9-6) needed to initiate mobilization is much higher than the
sandstone values used by Wilson and Conrad (1984) in their study of hydraulic remediation.
Despite their low estimate they concluded that hydraulic floods are unrealistic remediation
alternatives. We amplify that conclusion.
• The interfacial tension (IFT) must be reduced significantly to allow hydraulic schemes to
perform adequately. Surfactant floods hold promise of reducing the IFT sufficiently to allow
some mobilization of the residual. However, unless the IFT is reduced to the point of
emulsification of the organic liquid, it is doubtful that more than a small proportion of the
organic liquid can be mobilized under field conditions.
- 12 -
-------
Porous media heterogeneities dominate displacement and trapping mechanisms. We
presented flow visualization results from micromodels representing 1) an aggregated soil
(strings of interconnected macropores or fractures, separating clumps of micropores), and
2) a soil with discontinuous lenses (lenses of coarse pores within an otherwise
homogeneous model). The character of these displacements were contrasted with
homogeneous displacements. The discontinuous stringers experiments were duplicated in
short column studies, yielding both pore casts and quantitative measurements of trapping
in a heterogeneously packed sand column. These experiments demonstrate that:
o The saturation and spatial distribution of organic liquid found behind an
advancing front of organic liquid depends upon the combined effects of soil
heterogeneity and capillarity.
o Organic liquid selectively travels through the coarser and more permeable
portions of heterogeneous aquifers, by-passing finer-grained regions. Organic
liquids can be expected to move quickly through aquifers which have
interconnected coarse layers or fractures. In New England groundwater consultants
commonly distinguish between gasoline leaking from underground tanks in
unconsolidated, glacial deposits, and leaks in ledge or bedrock. In the
unconsolidated deposits anecdotal evidence suggests that much of the gasoline is
trapped by capillary forces, with very limited and slow migration of the liquid
gasoline. The gasoline tends to appear in nearby brooks dissolved in the
groundwater discharge. In crystalline rock terrain the gasoline moves quickly and
far, with observable liquid gasoline discharges to nearby brooks.
o Following the recovery of free liquid organic, the residual saturations left behind
in interconnected fractures or macropores tend to be smaller than in a
homogeneous porous material, but can be expected to extend over a larger
portion of aquifer.
o At typical aquifer flow velocities, capillary forces can relegate the flow of water
to finer-grained regions, by-passing the coarser organic liquid-filled regions.
Corroborating results were obtained from micromodels, column experiments and a
simple mathematical model.
o Following the recovery of free liquid organic, the residual saturation left behind
in a heterogeneous aquifer, that is composed of disconnected coarse lenses,
tends to be larger than in a homogeneous aquifer, but can be expected to
extend over a smaller portion of aquifer.
Increased recovery of organic liquids from coarse lenses may be attained by increasing
the pumping rate. This principle was demonstrated in the micromodel, soil columns, and
mathematical theory. However, the theory predicts that should velocities get high enough,
water will prefer to move through the coarse lenses and by-pass organic liquid in the fine
matrix. This suggests that varying the pumping rate, over long periods of time, may have a
beneficial effect.
- 13 -
-------
• The effects of the lens pattern of heterogeneity were incorporated into a simple
mathematical model. The model was qualitatively validated by the column and
micromodel experiments. The model represents a balance of viscous and capillary forces
on a large spatial scale, and can be expressed in terms of an effective capillary number.
• This simple model suggests that it is feasible to develop effective properties or
equivalent homogeneous models of behavior in heterogeneous materials. Computer
simulation codes into which these equivalent homogeneous properties are inserted may not
be conventional codes. They will need to account for the partial desaturation of lenses, and
the effect that has on effective relative permeabilities and saturations. For example, the
effective permeabilities and residual saturations will certainly depend on the rate and
direction of flow. We believe that this is a typical consequence of dealing with effective
properties in non-linear systems. Certainly Yeh et al. (1985a,b,c), Mantoglou and Gelhar
(1987), and McCord et al. (1988a,b) encountered the same issue in their work on the
Richard's equation approach to water flow in the vadose zone.
THE VADOSE ZONE
• The movement of organic liquid through the vadose zone is more complex than organic
liquid movement in the saturated zone, especially because of the tendency for many
organics to spread. The organic liquid is of intermediate wettability between water, the
wetting phase, and the non-wetting gas phase (at least for the conditions examined here).
The propensity for a given organic liquid to spread between water and gas can be predicted
from the spreading coefficient, 2 = am-(aow + am), where the a 's represent IFT's.
Positive spreading coefficients lead to organic liquid films along the water-gas interface.
• A spreading organic liquid forms (1) a film between the gas and water phases, as well as
(2) pockets of organic liquid that replace gas in the pore bodies, and gas or water in the
pore throats.
• Films appear to be an important mechanism of organic liquid invasion into the vadose
zone. The organic liquid usually advances by displacing gas, and sometimes water, from
pore throats and bodies. It also advances through film flow, most easily seen in our
experiments when pockets of organic liquid began to accumulate deeper in a micromodel,
with no obvious connection to sources of organic liquid above. Some water is by-passed,
mostly in pore throats, as is some air, mostly in pore bodies. The resulting entrapped air
bubbles are essentially identical to the non-wetting phase blobs seen in the two-phase
saturated zone experiments.
• In the vadose zone the residual saturation of a spreading organic liquid consists of films,
pendular rings, wedges surrounding aqueous pendular rings, and filled pore throats.
There are a few isolated organic liquid blobs trapped in otherwise water filled pore bodies,
and the occasional gas bubble trapped inside an organic liquid blob.
- 14 -
-------
• Films of a spreading organic liquid maintain continuity of the organic phase. The
pockets of residual organic liquid trapped in vadose zone pore bodies or pore throats
communicate; they are not isolated blobs.
• Residual saturations are much lower in the vadose zone than they are in the saturated
zone. In the Sevilleta soil the vadose zone residual saturation is 9.1%, compared to 27.1%
for saturated zone conditions. The lower residual is explained by:
o continuity of the films during drainage;
o presence of a third, non-wetting gas phase;
o additional buoyancy forces due to the greater density contrast of the organic liquid
to air, than to water; and
o less capillarity due to the lower IFT between the organic liquid and gas, than
between the organic liquid and water.
• The capacity for storing or retaining organic liquids is much smaller in the vadose zone
than in the saturated zone.
• The films expose a large organic liquid surface area to the gas phase, making soil
venting of organics in the vadose zone an attractive remediation strategy. The rate of
mass transfer may be limited by diffusion within the film portion of a multi-component
organic liquid. The controlling variables are film thickness and tortuosity, and the geometry
of its connection to larger pockets of organic liquid.
• The formation of films also exposes a large organic liquid surface area to the water
phase, increasing the effective solubility of the organic. The effectiveness of this transfer
is severely limited by the water flow rate.
• The distribution of the films and pockets of organic liquid can influence
microorganisms. Our review of photomicrographs has lead us to conclude that:
o There are largely air filled pores, with a ready access to oxygen, but with only a
very thin double film of water and organic liquid. At residual saturation these films
would have a very limited ability to resupply needed nutrients to microorganisms.
o The environment of water filled pore throats and pendular rings may be controlled
by the lack of oxygen. The film of organic liquid should suppress oxygen transfer to
the water phase.
o We speculate that transient changes in fluid saturations would improve the
environmental conditions for aerobic bacteria.
• Biosurfactants are a by-product of biological activity, especially for organisms that
attach at the organic liquid-water interface. We hypothesize that these natural surfactants
could change interfacial tensions enough to reduce organic liquid saturations by drainage
(see Figure 10-3).
- 15 -
-------
• Non-spreading organic liquids do not form films, but rather coalesce into small lenses
that float at the water-gas interface. Non-spreading organics.e.g., halogenated organic
solvents, have a negative spreading coefficient.
• The residual saturation of a non-spreading organic liquid consists of floating lenses at
the water-gas interface, pendular rings, wedges surrounding aqueous pendular rings,
and filled pore throats. There is no interconnecting film.
• The soil venting of non-spreading organics in the vadose zone may not be nearly so
attractive a remediation strategy. The organic liquid surface area exposed to the gas
phase is severely reduced by the coalescence of the organic into lenses. The mass transfer
coefficient should be much lower.
• Biosurfactants that are a by-product of biological activity could change interfacial
tensions enough to alter a non-spreading organic liquid to become a spreading organic
liquid. This should improve mass transfer coefficients.
EXPERIMENTAL APPROACH
• Flow visualization techniques employing micromodels and styrene polymerization
provide useful tools with which to examine multi-phase displacement processes on both
pore and bulk scales.
• The similarity of the blobs trapped in micromodel experiments to those trapped in
two-phase pore casts and blob casts provide some confidence that the micromodels
provide a reasonable analogy to multi-phase flow in soils.
• We believe that micromodel results can be generalized to aquifer scales, as illustrated
by the heterogeneous aquifer with stringers (Section 9).
• Flow visualization approaches can be applied to other related pollutant transport
problems, for example: diffusion, hydrodynamic dispersion, and 'small-scale'
macro-dispersion; bacteria and colloid adhesion and migration; biotranstormation of
organic pollutants; adsorption of dissolved organic pollutants and the possible alteration
of soil wetting properties.
• New equipment must be designed and constructed in order to examine the movement
and trapping of organic liquids in low permeability soils.
- 16 -
-------
SECTION 3
RECOMMENDATIONS
Recommendations are divided into two groups. The first group deals more with the science of
multi-phase flow and recommends additional research. The second group looks at the experimental
procedure and equipment, and recommends improvements.
RECOMMENDED ISSUES FOR FUTURE RESEARCH
Saturated Zone Research
• Observe blob size distribution via imaging, Scanning Electron Microscope (SEM), and
Coulter counter. We had hoped to conduct a statistical analysis of size and shape blob
casts. However, two of the problems that we encountered led us to reconsider. First, the
preferential wetting of the TFE (Teflon™) column walls may have led to a final non-uniform
distribution of residual saturation over the column cross-section (see Section 7). Blob
populations taken from this sample might not be representative. Second, the samples of
blob casts we examined showed many broken casts (see photomicrographs in Section 9).
We believe that these problems can be overcome, and recommend that future research
efforts collect this statistical data. In particular we recommend the use of image analysis on
pore cast fat sections, that are sequentially polished to provide parallel sections into the
sample. These sections can be reconstructed on a computer into a three dimensional
model of the space that shows the relative locations of minerals and fluids. The fat sections
may prove too transparent to allow adequate discrimination of the phases. If so, parallel thin
sections can be prepared, but with a loss of material between the sections.
• Observe fluid distributions with pore and 'blob' casts for other saturations. The in situ
visualization work with pore and blob casts focused on residual non-wetting phase
saturations. These techniques should be applied to the visualization of fluid distributions at
other saturations.
• Investigate residual organic liquid saturations of other soils, with an emphasis on
correlation with texture & porosity. A combination of quantitative experiments, pore
casts, and blob casts should be employed to develop a detailed morphology of this issue,
which the experiments reported herein barely begin to address. We've observed two very
similar soils with dissimilar residual saturations, while other, dissimilar soils have almost
identical residual saturations. This is a disturbing result which requires additional
explanation.
- 17 -
-------
Determine the role of clay minerals on residual organic liquid saturations. We found our
highest residual saturations, and most complex blobs, in the Sevilleta soil. Of the sandy
soils we investigated, this was the only soil with any significant amount of clay and silt in it
(perhaps 2%). It is possible that the clay's swelling properties played a role in the high
residual saturation, although it may be due to micro-layering caused by the packing
procedure. We believe that the issue must be resolved. The large difference in residual
saturations (17% to 27%) has enormous practical implications. We recommend a careful
investigation of the role of clay minerals with an emphasis on possible packing induced
heterogeneities.
Further investigate capillary trapping and mobilization at high capillary numbers. We
expect that the correlation between residual saturation and capillary number depends on
the soil pore structure, and that it will vary from soil to soil. Our apparatus was not able to
very fully explore this relationship for mobilization and did not attempt to do so for initial
trapping. Micromodels should also be used to address these issues.
Investigate effects of dissolution on residual saturations. The position of a capillary
trapped blob within the pore space has a strong influence on mass transfer between
phases. If a large portion of the blob surface is in contact with only a thin film of water, as
certainly seen in our pore casts and micromodels, then the transfer rate may be limited by
advection or diffusion in the film. Preliminary videotaped micromodel observations indicate
that the flowing water moves around the blobs, through the unoccupied pores, with little
water movement in the films to help advect organic components away (Mason ef a/., 1988).
We suggest additional micromodel experiments using a fluorescent dye as a soluble
component, or fluorescent microspheres, with image enhancement to bring out the
concentration distribution in the water phase. We could then observe the flow of water on a
pore scale, especially through water films and wedges surrounding the blobs. Similar
techniques could be use to examine flow in the films of the vadose zone experiments.
Investigate wetting phase relative permeability as a function of reduced residual
saturation. Compositional models are numerical codes which simulate multi-phase flows
with interphase transfers such as solubilization. 'Black oil' models focus on more
sophisticated representation of the multi-phase flow, but at the expense of neglecting
compositional behavior. The two approaches are being integrated by various researchers,
allowing better predictions of the migration of organic pollutants. These models need
estimates of relative permeabilities to each of the fluids. In particular they need the wetting
phase relative permeability at residual saturation. When residual non-wetting phase
saturations are reduced by mobilization, volatilization, or solubilization the wetting phase
relative permeability increases [ &,»,(Sor) at 0 < Sor < S'or', see Sections 3 & 9]. Increases in
permeability result in either increased water flow under constant head boundary conditions,
or increased residence time under constant flux conditions. In either case an accelerated
rate of dissolution is the ultimate result. It is our expectation that the distribution of organic
liquid, after having undergone significant dissolution, will be much different than the
- 18 -
-------
distribution of an organic phase subjected only to hydraulic forces. Hence, it is
hypothesized that two systems, each having the same fluid saturations, will not have the
same fluid permeabilities due to different distributions of the fluids — because in each case
the saturations are achieved by a different process. Accurate measures of relative
permeabilities at reduced residual saturations may be as important as good estimates of
the mass transfer coefficients for predicting pollutant migration.
• Continue experiments and theoretical modeling of residual saturations caused by
heterogeneities. The effects of heterogeneities on multi-phase flow were explored in
Section 9. In effect, heterogeneities dramatically increase the residual saturation of organic
liquids and the accessibility of that saturation to the passing groundwater. Further laboratory
and theoretical research is needed on this issue. We suggest additional micromodel
studies, with quantitative measurements of the effluents and the saturations (using image
analysis), on a wide variety of heterogeneities. Sandbox experiments can then be
conducted on those heterogeneities identified by the micromodel studies as most
important. We also suggest that the simple mathematical model presented be extended.
Other mathematical models should be developed using simulation (eg, like that in Ababou
et a/., 1988) or spectral approaches (like that in Yeh et a/., 1985a,b,c; Mantoglou and
Gelhar, 1987; or Welty and Gelhar, 1989). Finally, field experiments on organic liquid
movement should pay special attention to this issue.
• Investigate the influence of surface wetting on multi-phase flow behavior and the
movement and capillary trapping of organic liquids. We usually assume that soils are
water wet or hydrophillic. However, wetting can be altered by physical, chemical or
biological means (Anderson, 1986a; Wilson, 1988). It appears likely to us that some of the
compounds present in organic waste solutions, or even in gasoline, can alter wettability.
Schiegg (1980) observed such changes in his laboratory study, and we saw some in ours
(although not under controlled conditions). The major chemical mechanism is adsorption,
particularly of polar oxygen-, nitrogen-, or sulfur-containing compounds (Anderson,
1986a; Morrow et al, 1986). In the laboratory organochlorosilane compounds are typically
adsorbed onto sandstone cores, glass beads, and etched-glass micromodels to alter
wettability in fluid flow experiments (Anderson, 1986a, 1988). The organosilane reacts with
the silica surface leaving its organic portion exposed, thus presenting a hydrophobia
surface to the pore space. Chemical modification of clays and other soils has recently been
investigated, with a focus on sorption behavior by hydrophobic surfaces in aqueous media
(Boyd ef al., 1987; Bouchard et al., 1987; Mortland et al., 1986). Although these changes
were made to alter sorption behavior it is reasonable to hypothesize that the altered
wettability will effect the behavior of fluids. Surfactants can also change wettability. In
addition to influencing the behavior of fluids, wetting disturbs chemical partitioning between
phases, and the diffusion/dispersion of chemicals within each fluid phase. In short, the
entire transport system is influenced by wettability. We suggest the following research
issues (Wilson, 1988):
- 19 -
-------
o Can soil wetting be altered in the natural environment and under antropogenic
stress? If so, how? Does wetting alteration require aging?
o Is wetting actually altered to any significant extent in the environment, or is this
simply an academic problem?
o Where is wetting altered? What is the spatial pattern and heterogeneity? What does
heterogeneous wetting look like through an SEM?
o Does wetting recover? If so, how? How long does it take?
o What are the effects and consequences of a change in wetting? In particular how
does it influence fluid flow and chemical transport including sorption? How does it
affect colloid transport? How does it affect bacterial growth and migration? How
does bacterial growth affect wetting?
o Can the consequences be mitigated? If so, how?
• Study the effects of large Bond numbers on residual saturations in the saturated zone.
When dense organic liquids move downward through the saturated zone gravity forces may
play a significant role in reducing the initial amount of capillary trapping, particularly in
coarse grained soils. Experiments should be conducted to explore whether this issue is
important and to determine the effect it has on residual saturations, and blob size and shape
distributions.
• Determine the conditions required for the onset of fingering as dense organics move
downward through the saturated zone. Many halogenated solvents are more dense and
less viscous than water — attributes which promote flow instabilities (gravity fingering) as an
organic phase percolates downward through an aquifer. Of special interest in this problem
are the roles of capillarity and heterogeneity. In the case of a non-wetting phase displacing a
wetting phase, capillary forces appear to promote fingering, while layering in the soil
transverse to flow appears to promote stability. Further micromodel, sandbox, and
mathematical models of these issues are needed.
• Employ micromodels to observe colloid transport in a pore network. In particular,
observe attachment-detachment of colloid particles to pore walls as a function of pore size
distribution, pore network pattern, wetting, colloid size and charge. Colloids could be made
observable through optical microscopes by using fluorescent microspheres to represent
them.
• Employ micromodels to observe bacterial transport and colonization. Micromodels with
fluorescent bacteria could be used to study bacterial motility, attachment, detachment, and
colonization (see, eg, Bitton and Marsell, 1979; Marsell,1985; Pringle and Fletcher, 1983;
Van Loosdrecht et al., 1987a,b).
- 20 -
-------
Vadose Zone Research
• Study the movement and ultimate distribution of non-spreading organic liquids;
compare to spreading organic liquids. In three-phase (air/organic/water) systems in
which the organic liquid is of intermediate wettability, some organics spread between the
water and air phases while others, such as halogenated solvents, do not. The propensity of
a given organic liquid to spread can be predicted from the spreading coefficient,
2 = c^ - (a™, + om) . In Section 10 we presented some photomicrographs from a
micromodel study which indicate that the spreading coefficient can have a profound
influence on how the organic phase moves and where (especially on a pore scale) it ends
up. This can be particularly important to remediation via soil venting. We suggest additional
quantitative and visualization experiments.
• Investigate the behavior of organic liquids in the transition zone between the vadose and
saturated zones. Determine the equilibrium saturation. Particularly for organic liquids
lighter than water, the low suction range in and just above the capillary fringe is where all the
" action" occurs as light organics reach the water table and begin to migrate laterally to form
a lens. Unfortunately, this is where the balance of forces governing movement for the three
fluid phases becomes much more complicated (and perhaps the least understood). Later
we suggest improvements in our apparatus to permit this experiment to be carried out.
Another approach would be to measure saturations in a sandbox model constructed in a
fashion that permits the lateral redistribution of the organic phase (eg, Schiegg, 1980).
Such experiments are being conducted at a variety of laboratories around the country.
• Explore the effect of heterogeneities on behavior in the vadose zone. This is simply an
extension of the work recommended above for two phase flow.
• Determine the conditions required for the onset of gravity fingering as liquid organics
move downward through the vadose zone. All liquid organics are more dense and viscous
than the gas phase in the vadose zone. The greater density promotes flow instabilities
(gravity fingering) as an organic phase percolates downward toward the water table.
Capillarity and heterogeneity play a significant role in this process. There is significant
literature on the two phase flow instability issue. There is little on three phase flow instability.
Further micromodel, sandbox, and mathematical models of these issues are needed.
• Employ micromodels to study the movement and behavior of colloids and
microorganisms, and the efficacy of soil venting. The first two items are simply
extensions of the work recommended above for two phase flow. The third item addresses
the issue of water-gas mass transfer coefficients as a function of pore size, pore
distribution, spreading coefficient, and fluid saturations.
- 21 -
-------
General Recommendations for Research
• Investigate mobilization via surfactant floods. Surfactant floods can be used in an attempt
to mobilize residual organic liquid saturations by reducing interfacial tensions and capillary
numbers, or simply by emulsifying the organic liquid. Both of these mechanisms should be
explored in micromodel and column experiments for typical combinations of organic liquids,
surfactants, and soils.
• Compare 2- and 3-phase flow in micromodels to results from cellular automata (CA).
Cellular automata has recently been used to mathematically model multiphase fluid flow on
a pore scale (D. Rothman, personal communication, 1988; B. Travis and K. Eggert,
personal communication, 1989). For example, two phase flow is represented by 'red' and
'blue' particles. The particles interact on a grid using rules that satisfy the conservation laws
of mass and momentum. Rothman invented other rules to describe the relative affinity of a
red particle for another red particle and similar rules for blue particles, allowing CA models
to simulate interfacial tensions. Red or blue properties can also be assigned to the walls, in
order to represent wetting and its alteration. We believe that CA can be extended to three
fluid phases. By changing the interfacial tension rules it may be possible to accurately
describe the movement of a spreading or non-spreading intermediate wetting phase
(organic liquid) on a pore scale in the vadose zone.
• Investigate 'irreducible' wetting phase saturation dependence on history, and
non-wetting phase velocity, and determine its relevance to contamination problems. At
residual saturation the wetting phase is continuous and is composed of an interconnected
network of films, rings and wedges. The wetting phase liquid can move through its
interconnected network, draining the films and rings, and reducing the residual wetting
phase saturation. Most models of pore pressure-saturation assume a fixed and known
'irreducible water saturation'. We recommend a study of the effect of the non-wetting
phase flow rate on the amount and distribution of the wetting phase residual. We expect that
the flow rate must vary over orders of magnitude before the change in residual is significant.
Further research is needed to determine this relationship, its importance to fluid flow, and
its relevance to conditions encountered in the field.
• Conduct field experiments to confirm laboratory observations. In particular, attempt to
obtain undisturbed soil samples containing residual organic liquids under both saturated and
vadose zone conditions. Using quick freezing techniques (see below), observe the
distributions of fluids on a pore scale. Make quantitative measurements of fluid saturations
using, for example, distillation. Compare results to laboratory observations.
- 22 -
-------
RECOMMENDED IMPROVEMENTS IN EXPERIMENTAL EQUIPMENT & PROCEDURE
General Issues for Soil Column Experiments
• Provide better control on soil sample variability. Most of our experiments were conducted
on the Sevilleta sand, which we selected as our standard soil for a variety of reasons (see
Section 3). This is an aeolian soil, and we were not always able to re-sample it so as to
obtain 'exactly' the same material. We believe that residual non-wetting phase saturations
may be more sensitive to this variability than some other soil properties (eg, porosity,
permeability). Consequently, great care should be taken in future work to insure that a
sufficient stock of material is on hand to complete all experiments. Naturally this material
should be properly handled and stored so as to minimize segregation and to avoid
undesired clumps and aggregates.
• Conduct mercury and nitrogen gas porosimetry on the soil samples, in order to yield
estimates of the actual pore size distributions for the soils tested in the various experiments
Residual Saturation Measurements - Short Column Experiments
• Modify the experimental procedure in order to observe and record capillary numbers, so
as to insure low capillary number conditions. This modification would insure that there is
no danger of comparing residual saturations taken under low capillary number conditions, to
those taken under conditions that would result in a lower initial residual saturation.
• Improve the apparatus, or develop a new apparatus, to handle low permeability soils like
the Palouse loam. We were not able to make significant measurements of residual
saturation for the Palouse loam because it was difficult to supply sufficient presure to force
organic liquid nito the soil. This problem can be circumvented with a high-pressure system,
including a high-pressure fluid delivery pump, a column built to withstand high pressures,
and a bottom membrane with a very high non-wetting phase entry pressure.
• Consider the development of a new apparatus capable of controlling pressures in all
fluid phases. Such equipment has been developed by others (eg, Lenhard and Parker,
1987b; A. Demond, personal communication, 1988), but is expensive to construct and
operate. To make a large number of measurements additional improvements to this type of
experimental equipment is warranted, motivated by both economics and science.
- 23 -
-------
• Improve the apparatus and procedure to permit observation of conditions in the
transition zone between the vadose and saturated zones. Results from our transition zone
experiments gave what we believe to be unreasonably high equilibrium organic liquid
saturations, reported in Section 10. During organic liquid drainage, water was unable to
re-imbibe into the soil. Under low applied suctions representing the capillary fringe and just
above, re-imbibing water would have displaced organic liquid resulting in lower equilibrium
organic saturations than the values we measured. The largest change in forces acting on
the organic liquid occurs in the capillary fringe, and just above, where the transition from the
vadose zone to saturated zone conditions is most pronounced. The apparatus mentioned
immediately above would eliminate this problem.
• Investigate the influence of different column packing procedures on residual
saturations. In particular, compare dry vs. wet packings on a variety of soils, with an
emphasis on the nature of packing induced heterogeneities. This type of experiment would
also be relevant to the many large scale sandbox experiments currently being planned or
conducted. This investigation is different than the typical packing study which does not
examine residual saturation.
Residual Orgariic Liquid Mobilization - Long Column Experiments
• Develop a long column with a water wet wall for future mobilization experiments. The
organic liquid wet PVC walls of our long column may have influenced the distribution and
amount of organic liquid in the column, prior to the water flood. The resulting residual
organic liquid saturations may have been biased. We recommend that future mobilization
experiments be conducted in a column with water wet walls. We initially designed a glass
version of the long column, using a stretched version of the glass short column. Pressure
taps in the column were designed so that they could be installed by a glassblower.
Concerned that the column would be too fragile, and that we would accidentally break these
glass taps off, we switched to the easier to work with PVC. As it turned out, we accidentally
broke off some of the PVC taps.
• Improve the apparatus and procedure so as to handle higher capillary numbers. We
found that the long column apparatus was unable to pump at a sufficient flow rate to achieve
significant mobilization for the soil we tested. This problem can be overcome by improving
the pump and plumbing to handle higher flow rates and pressures. However, for the soils we
examined this would mean Reynolds numbers greater than ten, an unlikely condition in the
field. Optionally, the wetting and non-wetting fluids could be selected so that their interfacial
tension (IFT) is significantly lower that that of the water and Soltrol used in our experiments.
- 24 -
-------
• Develop an experimental procedure to examine the effect of organic liquid aqueous
solubility on residual saturations, their mobility, and aqueous phase relative
permeability. Suppose a multi-component organic liquid is at residual non-wetting phase
saturation. Let each of the components have a significantly different solubility in the
aqueous phase. Then assume that the more soluble component completely partitions to the
aqueous phase, before there is any significant partitioning of the remaining components.
For example, consider a two component organic liquid composed of a mixture of isopropyl
alcohol and Soltrol. Then the production of alcohol at the end of the column is a measure of
the solubilization of the alcohol, and can be correlated to the reduction of residual saturation
of the mixture. As the residual saturation is reduced, the isolated blobs of capillary trapped
residual become smaller, and some of the residual is mobilized. Part of this mobilized
volume may coalesce down the column, while some may make it to the end of the column
where it is observed. The aqueous phase relative permeability is measured before and after
the solubilization. A three component mixture would provide a third point of data, or perhaps
the two component experiment can be rerun with different proportions of alcohol and Soltrol
to provide a much wide range of data.
In situ Visualization of Residual Saturation - Pore and Blob Casts
• Develop pore casts of the pore space. This will allow us to see into the actual pore space in
which the fluids move. Pore casts have been made for sandstones and carbonates by
injecting Wood's metal, epoxy, or styrene. In these rocks the matrix is interconnected and
can be removed with acid. The mineral grains of unconsolidated soils are not
interconnected. Only the exposed grains can be removed. A three-dimensional serial
reconstruction of the pore space can be made by sequentially sectioning the sample, using
acid to remove new grains as they appear.
• Develop a new short column with water wet walls for pore and blob casts. As described in
Section 7 we experienced some preferential flow of the styrene along the walls of the TFE
column used in these experiments. It is not clear to what degree this influenced the
visualization results. As the use of TFE for the column offered only minor advantages over
other choices, we recommend that water wet materials be employed in the future, for those
experiments using water as the wetting phase. TFE columns can still be employed when
styrene is considered the wetting phase.
• Investigate other monomers, initiators, and techniques for the saturated zone pore &
blob casts. We experienced several problems with styrene as the monomer for these
experiments. Once the inhibitor was removed its viscosity began to increase. This lead to an
acceleration of the experimental time table, with the result that the pore and blob casts may
be biased. The styrene also shrank upon polymerization. We suggest that additional efforts
be made to find fluids that can be hardened, but that have fewer problems than the styrene.
- 25 -
-------
• Reduce blob cast breakage. We were unable to perform a size and shape analysis on the
produced blob casts, partly because of breakage. We need to explore new materials and
handling procedures that will reduce the incidence of breakage.
• Investigate other monomers, initiators, epoxies, and techniques for the vadose zone
pore casts. We experienced several problems with fluids used in these experiments.
Principally this involved the changing viscosities of the fluids, which did not appear to have
always drained to equilibrium. This was especially a problem with the epoxy representing
the intermediate wetting phase. We also had air bubbles entrained in the epoxy. These
problems affected fluid distributions and interfaces in the pore casts. Additional efforts
could be directed toward finding a better combination of porous media, fluids, and dyes to
minimize these problems.
• Investigate freezing fluids in place with liquid nitrogen. Soils containing several fluid
phases can be quick frozen with liquid nitrogen, and viewed under a scanning electron or
optical microscope equipped with a cryogenic stage. An example of this approach is shown
in Gvirtzman et al (1987), for the case of frozen water and air in a spoonful of soil. Jamine
Wan of our group performed some preliminary experiments with a cryogenic on an optical
microscope, using Traverse City soil, undyed Soltrol and water. She found that keeping the
sample frozen is not very useful for samples containing two liquids. Instead, by slowly
bringing up the temperature until the Soltrol begins to melt, the distribution of fluids is briefly
visible and can be photographed. This technique holds promise for the visualization of fluid
phase distributions in laboratory and field collected samples. In the lab it avoids the artificial
nature of styrene and epoxy simulations of the various fluids, especially their increasing
viscosity with time.
• Improve thin sectioning of pore casts. We found it difficult to make thin sections of the
pore casts, whether done in local facilities or contracted out. The combination of an
unconsolidated soil and an infill of various plastics made it difficult to cut and polish the
sections without plucking or cracking the mineral grains. Improvements in this technology
would certainly aid in the production of better photomicrographs.
Visualization of Multi-phase Flow in Micromodels
• Improve the micromodel equipment and procedure to allow quantitative measurements.
Improved end reservoirs, incorporating oil- or water-wet etched channel porous
membranes provide a mechanism for controlling capillary end effects. Pressures can be
controlled and measured through these reservoirs. Etched pore cross-sections can be
estimated using measurescopes, epoxy casts, and other techniques. Saturations can be
estimated using a high resolution video camera, a frame grabber and image processing
techniques. These quantitative techniques could be combined with mathematical models at
the pore and network level (eg, Soil and Celia, 1988) in order to help explain the
observations and to validate the models.
- 26 -
-------
SECTION 4
CHARACTERIZING EXPERIMENTAL FLUIDS AND SOILS
In this section, characterization of the fluids and soils commonly used in several of the experiments is
presented. Predictably, the section is divided into two main parts — fluids, and soils. Within each part,
the methods used for characterization will be described, followed by the characterization results
themselves.
FLUID CHARACTERIZATION
Measurements of fluid properties such as viscosity, specific gravity, surface tension, and interfacial
tension were performed following procedures outlined by the American Society for Testing and Materials
(ASTM, 1986). Viscosity was measured with Cannon-Fenske routine viscometers according to ASTM
methods D445-83 and D446-85a. Specific gravity measurements were made as described by ASTM
method D1429-76 (pycnometer procedure). An adaptation of ASTM procedures D 1590-60 and D 971-82
were used to determine surface tension and aqueous-organic interfacial tensions, respectively, with a
Fisher Manual Model 20 tensiometer.
The surface tension procedure (D 1590-60) was slightly modified so that no centrifuging of the
samples was done. In Section 7 of the ASTM procedures, the method of glassware cleaning was
modified to follow the cleaning method of the interfacial tension procedure (D 971-82). The 100 ml
beakers were first rinsed with petroleum ether, followed by two rinses of methyl ethyl ketone
(2-butanone) and distilled water. A hot chromic acid solution was then used to remove any remaining
contaminants. Five rinses of distilled water followed by three rinses of distilled-deionized water
completed the procedure. The beakers were oven dried and allowed to cool to room temperature before
use. In Section 9, the procedure for cleaning the platinum-iridium ring was also modified to follow the
interfacial tension method. The ring was soaked in petroleum ether and methyl ethyl ketone after which it
was flamed in a bunsen burner. The ring was allowed to cool 3 minutes between flaming and use.
Measured fluid properties
The measured properties of all fluids used in our laboratory experiments (except for styrene and the
epoxies used in the blob and pore cast experiments) are shown in Table 4-1.
A 3000 ppm CaCI2 solution was used as the aqueous phase in all column experiments. Distilled,
de-ionized water was de-gassed by boiling. Enough calcium chloride dihydrate was added to the cooled
water to bring the concentration to 3000 ppm. The solution was stored under a vacuum to keep it
de-gassed. The properties of this aqueous phase are given in Table 4-1. The interfacial tensions of all the
organic liquids were measured against this fluid.
- 27 -
-------
liquid
aqueous
phase
Soltrol-130
kerosene
gasoline
n-decane
benzene
toluene
p-xylene
PCE
carbon
tetrachloride
specific
gravity
1.003±0.002
0.755 ±0.002
0.809 ± 0.002
0.733 ± 0.005
0.729 ±0.002 2'
0.877 204 t
0.864±0.002
0.858 ± 0.005
1.609 ±0.005
1.602±0.002
density
(g/cm3)
1.000±0.002
0.753 ±0.002
0.807 ± 0.002
0.731 ± 0.005
0.727 ±0.002 2'
0.87720 *
0.861±0.002
0.855 ±0.005
1.614 ± 0.005
1.599 ±0.002
kinematic
viscosity
(cst)
0.98±0.01
1.93 t 0.01
2.15 ± 0.01
0.66 + 0.01
1.25±0.01
0.745 «> *
0.68520 *
0.70 ± 0.01
0.54 ± 0.01
0.66 ±0.01 21
dynamic
viscosity
(cp)
0.98 ±0.01
1.45 ± 0.01
1.73 ±0.01
0.48 ± 0.01
0.91 ±0.01
0.652 2° t
0.59020 *
0.60 ± 0.01
0.87 t 0.01
1.05 ±0.01 21
interfacial
tension
with 0.3%
CaCI2
solution
(dynes/cm)
not applicable
47.8 ± 1.2
38.6 ± 1.2
22.9 ± 0.3
44. 5± 1.0
not determined
not determined
35.8 ±0.8
41.8 ± 0.7
32.9±0.82'
surface
tension
(dynes/cm)
72.0 ± 0.4
19.1 ± 0.3
26.8 ± 0.4
20.5 ± 0.3
24. 9 ±0.3
28.9 2ot
28.5 ± 0.5
28.4 20 T
31.7 20 *
28.2 ±0.321
t Weast, 1986
TABLE 4-1. Measured properties of fluids used in experiments. All measurements were taken at
temperatures between 22°C and 24°C except where noted by superscript,
eg, 20 refers to 20° C. Values with the T symbol were taken from Weast, 1986.
- 28 -
-------
•- Tap water
• - Soltrol
2.5-
Liquid 1.5 '<-
evaporated
in ml _
+0.5-
-0.5,
012 34 56
Elapsed time in weeks
FIGURE 4-1. A simple experiment illustrating the relatively low
volatility of Soltrol, compared to water.
Soltrol-130, a mixture of C10 to C13 isoparaffins produced by Phillips 66 Company, was the fluid most
commonly used as the organic phase in our experiments. It is a colorless, combustible liquid having a
mild odor, negligible solubility in water, and a relatively low volatility. For example, the simple experiment
documented by Figure 4-1 illustrates the low relative volatility of Soltrol to water. Equal amounts of
Soltrol and water were placed side by side in identical graduated cylinders and periodically checked
visually for liquid loss caused by evaporation. Soltrol's low toxicity, solubility and volatility, coupled with
its density contrast with water, made it an ideal organic liquid for our basic experiments. It was used in
most short column saturated zone experiments and all micromodel, vadose zone short column, and
saturated zone long column experiments. The properties of Soltrol as measured in the laboratory are
given Table 4-1.
Residual saturations were measured in short columns for a variety of other organic liquids and
compared to the Soltrol results (see Sections 9 & 10). Separate tests were conducted for n-decane,
p-xylene, tetrachloroethylene (PCE), gasoline, and kerosene. These liquids were selected to be
representative of several classes of organic liquids. Three are single-component liquids: n-decane was
chosen to represent straight-chain aliphatics, p-xylene was chosen to represent aromatics, and PCE
was chosen to represent halogenated hydrocarbons. Selection of a particular chemical to represent a
given class of chemicals was based on the objectives of minimizing solubility and volatility, while
maximizing the density contrast with water. Low volatility and solubility were desired to avoid competing
physical processes. A large density difference between the organic liquid and water was desired
because the soil columns were weighed, exploiting the density difference between the fluids, to measure
the fluid saturations. Larger density differences between the fluids led to more accurate measurement of
the saturations. Multi-component mixtures were represented by gasoline and kerosene, as well as
Soltrol-130. Gasoline was also represented by xylene for the lighter aromatic fraction and decane for the
less volatile fraction. The composition of gasoline and kerosene vary with the supplier and the season.
One batch of each was purchased locally for the tests. Care should be applied when generalizing our
results to other gasoline or kerosene mixtures.
- 29 -
-------
Gasoline components benzene and toluene, and the solvent carbon tetrachloride were
characterized, but not tested for residual saturation. The volatility or solubility was too high to permit
accurate experiments with present procedures.
SOIL CHARACTERIZATION
Four soils were used in these experiments. The soil that we have labled as the 'Sevilleta soil' was
used for the majority of the tests and was extensively characterized. The remaining soils were only
partially characterized. Quantitative soil descriptions include particle density, particle size analysis,
capillary pressure-saturation curves, and wettability tests. Soil particle densities were determined using
ASTM D584-83. Soil particle size distributions were measured using ASTM D422, and the saturated
hydraulic conductivities were measured by ASTM D2434. Porosities and bulk densities of the soils were
measured at the time of column packing and the methods used are described in Sections 5, 6, and 7, for
each experiment. Soil capillary pressure-saturation relationships were determined using equilibrium
methods (Vomocil, 1965). Wettabilities of the soils to water and the organic liquid were by the Amott
method (Amott, 1959) and the USBM method (Donaldson, 1969). Saturated hydraulic conductivity and
intrinsic permeability were determined using the steady flow method. Relative permeability was
determined for two of the soils using the unsteady state method of Jones and Roszelle (1978).
Capillary Pressure-Saturation Test
Soil capillary pressure-saturation relationships were determined using equilibrium methods
(Vomocil, 1965). Hanging column type equilibrium stepwise water-air, water-organic, and air-organic
displacement experiments were performed under both drainage and imbibition conditions to determine
the capillary pressure functions. The experiments were performed using prepared soil columns. The
effective column length was kept short (5 cm) in order to keep fluid saturations relatively constant along
the length of the column, and to optimize column weight for gravitimetric measurements. A longer
column would have necessitated the use of a balance with significantly lower accuracy. Standard
equilibrium procedures were altered for the water-organic liquid displacement experiments to
accommodate conditions involving two liquid phases. Figure 4-2 shows this modified set up. Tubing was
used to connect the tops of the burets to seal the system and reduce volatilization. The soil column was
prepared as described later in Section 5 for a water saturated column. The soil column was attached to
burets containing organic liquid (buret A) and water (buret B). The capillary pressure was increased
stepwise. At each step the system was allowed to equilibrate (about 24 hours were required). The
capillary pressure head was measured in terms of the wetting fluid, water, as:
- 30 -
-------
He =
Qo
+ QW h2
Qw
(4-1)
where: Hc = -Qw g Pc = capillary pressure head,
often called the suction head,
expressed in cm of wetting phase, in this case H20
Pc = capillary pressure = Pnon-wet - Pwe, = P0 - PW (pascals)
hi = height of the organic liquid level in buret A above the center of the column (cm)
ft2 = height of the water level in buret B below the center of the column (cm)
Qo - organic liquid density (g/cm3)
QW = aqueous fluid density (g/cm3)
organic
liquid
buret
buret A
middle of soil column
water buret
buret B
FIGURE 4-2. Setup for the organic liquid and water capillary pressure - saturation relationship.
- 31 -
-------
Fluid saturations were measured both gravimetrically (by weighing the column after equilibration had
been achieved in each step) and volumetrically (by measuring the volume change of the fluids in each
buret) . Gravimetrically, saturations were measured as :
AM
S0 = 1 - Sw (4-3)
where: Sw = water saturation (-)
SwP = water saturation at previous step (-)
AM = change in column mass from the previous step (g)
Ap = density difference between the fluids, Qw - Q0 (g/cm3)
Vp = pore volume of the column (cm3) (Section 5 describes how Vp was measured)
Column mass was measured using at least three independent weighings on a Mettler PE 1600 balance
with 0.01 g accuracy. Volumetrically, the measurements were based on:
Ay,
S, = Sp + —^ (4-4)
'p
where AV,- is the total change in volume of a fluid (/ = w or o) in the soil column as measured in the
respective buret.
Wettability Measurements
Wettability refers to the relative affinity of the soil for the various fluids — water, air, and the organic
phase. On a solid surface, exposed to two different fluids, it can be measured by contact angle (see
Figure 4-3). Melrose and Brandner (1974) claim that contact angles are the only unambiguous measure
of wettability. Several common methods for measuring contact angles are summarized by Adamson
(1982); often a contact angle cell is used in petroleum engineering studies (Craig, 1971). The contact
angle test is performed under brine. A drop of reservoir oil is placed between two polished crystals that
are representative of the reservoir rock (usually quartz or calcite) and contact angles are measured.
When surface active agents are present, it may take up to hundreds of hours of interface aging time to
reach equilibrium (Treiber et al., 1972). Hysteresis in contact angle measurements is common, so that
both advancing and receding angles are often measured. There is some question as to how
representative contact angle measurements are since they do not take into account the effect of surface
roughness, heterogeneity, and pore geometry (Anderson, 1986b). Contact angle measurements were
not used in this study. To measure the bulk wettability of our soils, we employed adaptations of two
methods commonly practiced on cores in petroleum engineering: the Amott test and the USBM method.
Both methods rely upon characteristics of organic/water capillary pressure-saturation curves to
determine the wettability of the porous media.
- 32 -
-------
a
water wet
/: ;: xxx ::: :: :: :: :;' ::/ X xxxxx xxxxXx
intermediate wet
organic liquid wet
" „ solid surface "„ ' /
FIGURE 4-3. Contact angle measurement on a clean, smooth solid surface.
The Amott (1959) test measures the wettability of soil as a function of the displacement properties
of the soil-water-oil system. Four displacement operations are performed on the soil, and the ratio of
spontaneous imbibition to forced displacement is determined for both the water and organic phases. The
technique is similar to that required to determine capillary pressure-saturation relationships for two fluid
phases. The four displacements are :
1) spontaneous displacement of organic liquid saturated core or column by water.
2) forced displacement of organic liquid saturated core or column by water.
3) spontaneous displacement of water saturated core or column by organic liquid.
4) forced displacement of water saturated core or column by organic liquid.
This test is based on the fact that the wetting fluid will spontaneously imbibe into the core, displacing the
non-wetting phase. When the displacement-by-water ratio (6W) approaches one and the
displacement-by-oil ratio (60) is zero, the core is water-wet. If the opposite is true, the core would be
oil-wet. The main problem with the Amott test is its insensitivity at near neutral wettability
(Anderson, 1986b).
The USBM test compares the areas under two fluid phase capillary pressure-saturation curves as a
measure of average wettability (Donaldson et al., 1969). When a soil is water-wet, the area under the
organic-displacing-water capillary pressure curve ( Aj) is larger than the water-displacing-organic curve
- 33 -
-------
water-wet
WETTABILITY
intermediate-wet
oraanic-wet
contact angle
minimum
maximum
Amott test
displacement-by-water ratio (8W)
displacement-by-oil ratio (50)
USBM wettability index, W
0°
60° to 75°
positive
zero
near 1
60° to 75°
105° to 120°
zero
zero
near 0
105° to 120°
180°
zero
positive
near -1
TABLE 4-2. Relationship between wettability measurement methods (after Anderson, 1986b).
( A2). In fact, for a strongly water-wet system, most of the water will spontaneously imbibe into the soil,
and the area under the water-displacing-organic curve will be very small. Since the work of fluid
displacement is proportional to the areas under the capillary pressure curve (Morrow, 1970), the USBM
wettability index in essence measures the ratio of work needed for organic phase to displace water, to
the work needed for the opposite displacement. The USBM wettability index, W, is given as:
W = log
(4-5)
Table 4-2 shows the wettability criteria for each of the three quantitative methods.
Relative Permeability
Relative permeabilities were measured in long columns (see Section 6). The columns were initially
water saturated. Organic liquid was drained into the column, reducing water to its wetting phase residual
saturation, S^. . This was followed by a water flood, or imbibition of water, to reduce the organic liquid to
its non-wetting phase residual saturation, Sor . Fluid outflows and pressure drops across the column
were monitored during these two displacement steps. A graphical version of the 'unsteady state
method', which is based on the Buckley-Leverett model of immiscible fluid displacement, was used to
calculate relative permeability-saturation curves for each fluid (Jones and Roszelle.1978). In the
drainage step the water saturation and the relative permeability to water, km . began at unity and were
reduced as organic liquid displaced water from the column. The solid lines in Figure 4-4 illustrate typical
results expected from this first displacement step. As the displacement continued, the water
approached residual saturation — at which time the relative permeability of the water approached zero.
In the imbibition step, as the water saturation increased, the relative permeability of the water also
increased. The saturation and permeability of the organic phase became reduced until the organic phase
was discontinuous and immobile within the pore space — i.e. this second displacement step caused the
- 34 -
-------
organic phase to be reduced to its residual saturation. The water relative permeability did not fully
recover to unity, because of the presence of residual organic liquid. The dashed lines in Figure 4-4
indicate typical results obtained from this displacement step.
Sevilleta Soil
The 'Sevilleta sand' is a well sorted, medium grained, aeolian sand, taken from a sand dune in the
Sevilleta Wildlife Refuge, located 15 miles north of Socorro, New Mexico. This sand was chosen for three
reasons. First, it is a uniform, homogeneous soil with a fairly high hydraulic conductivity, a very low
percentage of fine particles, and a very low organic content. These properties made the Sevilleta sand
easy to use during the development of the short and long column experimental techniques. Second, the
sand is easy to obtain. The Hydrology program maintains a field site at the location from which the soil is
taken. Third, the Sevilleta dune sand has been previously in several hydrologic studies here at Tech (eg.,
McCord et al. 1988a,b). This permitted a comparison of soil characterization results with those of
previous studies.
o
0)
0>
0)
o:
drainage: organic liquid displacing water
imbibition: water displacing organic liquid
0.5
Water Saturation
FIGURE 4-4. Generic representation of relative permeabilities versus saturation for the
displacements performed in steps 1 and 2 of the long column experiments.
- 35 -
-------
The particle density of this sand was determined to be 2.65 ± 0.02 g/cm from five replicate
measurements. Six sieve tests were conducted to measure the particle size distribution of the Sevilleta
soil. There was excellent agreement between tests with all curves falling essentially on top of one
another. The results of one test are presented in Figure 4-5, along with the results for two other tested
soils. The particle size distribution classifies the Sevilleta soil as a uniform medium grained sand, with a
median particle diameter of about 0.3 mm (300 microns) and a uniformity coefficient (d60/d10) of less
than 2.
An SEM photomicrograph of a sample of the sand is shown in Figure 4-6. The grains are sub-angular
to sub-rounded. The SEM picture shows a particularly angular grain. (Also note the web of graphite paint
that the grains are sitting on; this is an artifact of the SEM procedure.) A mineral characterization of the
sand indicates that it is composed mostly of quartz grains ( 72 ± 5% by number) , with lesser amounts of
feldspar( 1 1 ± 2%) and lithic fragments ( 1 6 ± 3%) . The lithics were generally much smaller in size than
the other mineral grains; they compose about 5% of the sand by volume. An organic carbon analysis was
conducted at North Carolina State University (courtesy of Dr. Cass Miller) and yielded an organic carbon
content of 0.02%, an almost negligible amount.
Eleven replicate measurements of water-saturated hydraulic conductivity (Kw ) and intrinsic
permeability (k) were conducted in a constant head permeameter with the following results:
Kw = (1.03 ± 0.20) x 1CT2 cm/sec
k = 1.03 x i(T7 cm2 - 104 darcys
The intrinsic permeability was calculated from saturated hydraulic conductivity by:
k = - (4-6)
Qw g
where: pw = dynamic viscosity of water (0.98 cp)
QW = density of water (0.997 g/cm2)
g = gravitational constant (981 cm/sec2)
Capillary pressure-saturation curves for all fluid pair combinations (air-water, air-organic, and
organic-water) were constructed with the Sevilleta soil. Soltrol-130 was used as the organic liquid phase
in all trials. The soil capillary pressure-saturation plots from data acquired during 12 experimental trials
include:
• seven organic-water capillary pressure curves, of which two curves have
drainage, imbibition, and secondary drainage cycles, three curves have
drainage and imbibition cycles, and two curves have the main drainage branch
only;
• two air-organic capillary pressure curves, one curve with drainage,
imbibition, and secondary drainage cycles, and one curve of the main
drainage branch only; and,
- 36 -
-------
Percent
Passing
O.O 0.2
1.6 1.8 2.0
0.4 O.6 O.8 1.O 1.2 1.4
Particle Diameter (mm)
FIGURE 4-5. Particle size analysis for three of the soils used in this study.
500 MICR
FIGURE 4-6. SEM photomicrographs of Sevilleta sand.
- 37 -
-------
• three air-water capillary pressure curves, of which two curves have the main
drainage and imbibition cycles, and the third curve has the main drainage
branch only.
For the air-organic trials, the soil column was packed under organic liquid. Examples of two of these 12
capillary pressure-saturation curves are shown in Figure 4-7. The remaining curves are presented in
Appendix C.
Organic liquid-water capillary pressure-saturation curves were used to determine the wettability of
the Sevilleta soil. A typical Soltrol-water capillary pressure-saturation plot (shown in Figure 4-8) was used
in conjunction with the Amott and USBM methods to determine the wettability of the Sevilleta soil. The
four displacement operations performed in the Amott test were:
1) spontaneous displacement of organic liquid by water (corresponding to a
capillary pressure of zero on curve 2); point A;
2) forced displacement of organic liquid by water (corresponding to the residual
organic liquid saturation); point B;
3) spontaneous displacement of water by organic liquid (corresponding to a
capillary pressure of zero on curve 3); point C; and,
4) forced displacement of water by organic liquid (corresponding to irreducible
water saturation or residual wetting phase saturation); point D.
Preferentially water-wet soils have displacement ratios, 6W = (SwA - SwD)/(SwB - SwD) , approaching
1.0 and displacement-by-organic ratios, 60 = (SwB - SwC)/(SwB - SwD) , of zero. Organic liquid-wet soils
give reverse results. The Sevilleta soil is strongly water-wet, not only to Soltrol, for which 6W = 0.98 and
80 = 0.0, but for all of the other organic liquids utilized as well.
The USBM test compares the areas under capillary pressure curves as a measure of wettability.
When a soil is water-wet, the area under the organic-displacing-water capillary pressure curve (shaded)
is larger than the area under the water-displacing-organic curve (black). In a strongly water-wet system,
most of the water spontaneously imbibes into the soil, and the area under the water-displacing-organic
curve is very small. This is the case for Soltrol in the Sevilleta soil, as shown in Figure 4-8.
Sevilleta sand relative permeability under drainage was measured during the long column test. The
results, shown in Figure 4-9, are not considered reliable.
Traverse City Soil
The Traverse City soil sample was supplied to us by EPA's Kerr lab from their biodegradation field
demonstration site in Traverse City, Michigan. The site is located at a U.S. Coast Guard Air Station
Superfund fuel spill. The soil is a clean, beach sand, composed primarily of sub-rounded to rounded
quartz grains, with some dolomite and igneous and metamorphic particles (Twenter, 1985). It has a
particle density of 2.65 ± 0.01 g/cm3 and a saturated hydraulic conductivity of 1.0 x 10~2 cm/s. An
organic carbon analysis conducted on soil samples from a nearby well indicated the soil had an organic
- 38 -
-------
100
80-
Suction
cm of 40—
water
Soltrol-Air Saturation Curve
(Drainage-Imbibition-Drainage)
Air-Water Saturation Curve
(Drainage-Imbibition)
Suction
cm of
water
' i r~rrTi
0 10 20 30 40 50 60 70 80 90 100
Water Saturation (%)
80-
60-
40-
20-
20
If AW Trial 1
4
t^
v^ * -
arwat . dat
0 10 20 30 40 50 60 70 80 90 100
Water Saturation (%)
FIGURE 4-7. Typical Sevilleta sand capillary pressure-saturation curves.
ao.o
ro.o-
60.0-
90.0-
Suction 40-0~'
cm of
water so.o-
20.0-
10.0-
0.0
-10.0
10 20 30 40 30 60 70 80 90 100
Water Saturation (%)
FIGURE 4-8. A typical organic liquid-water capillary pressure-saturation curve used to determine
wettability, in this case for Soltrol-130 in Sevilleta sand.
- 39 -
-------
Sevilleta soil
i —
0.9-
0.8-
0.7-
_
0.6-
-
relative °-5~
permeability
0.4-
.
0.3-
0.2-
0.1-
0
? I imbibition: water displacing Soltrol /
\ (curves inferred; drainage data only) /
V /
\\ /
S\ /
SOLTROL \ \ WATER \
\ \ /
\\-4- /
X \ + /
\ \ + /
\ \ + /
\ \ + I
\ / V /
X \ /
A \ /
/ \ \ /
/ \ \ /
data from \ \ /
drainage \ \ /
\ \ ?\
\ NXV Nv
\ _-/>y? \v
^*^^^=^*" \ \
r • i ' T - i ' N i ^1
0 0.2 0.4 0.6 0.8 1
water saturation
FIGURE 4-9. Relative permeability vs water saturation for water and Soltrol in the Sevill
soil. The data were determined using the unsteady state method (Jones £
Roszelle, 1978). The curves are inferred. These data and curves are i
considered reliable.
carbon content of about 0.01% (Twenter, 1985). Sieve analysis of Traverse City Soil is plotted in Figure
4-5. The Traverse City Soil is slightly coarser grained and less uniform than the Sevilleta Sand.
Llano Soil
The Llano soil sample was obtained from the escarpment of the Llano de Albuquerque, remnants of
a fluvial plain deposited and then eroded by the Rio Grande and the Rio Puerco, west of Belen, New
Mexico. The soil sample is a clean, coarse, fluvial sand. It has a particle density of 2.66 ± 0.01 g/cm3,
and a saturated hydraulic conductivity of 1.6 x 10"1 cm/s. Sieve analysis of the Llano soil is plotted in
- 40 -
-------
Figure 4-5. The sample is more coarse grained and much less uniform than either the Sevilleta or
Traverse City soils. Relative permeability, as measured and inferred during our long column test (see
Section 6) is shown in Figure 4-10.
Llano soil
i-
0.9-
0.8-
0.7-
0.6-
relative 0.5-
permeability
0.4-
0.3-
0.2-
0.1-
0-
drainage: Soltrol displacing water
imbibition: water displacing Soltrol
(curves inferred)
0
data from
imbibition or
waterflood
data from
drainage or
organic liquid
flood \
0.2
water saturation
FIGURE 4-10. Relative permeability /s water saturation for water and Soltrol in the Llano
soil. The data were determined using the unsteady state method (Jones and
Roszelle, 1978). The curves are inferred.
Palouse Loam
The Palouse loam soil is an agricultural soil from eastern Washington. It provides a good contrast to
the Sevilleta, Traverse City and Llano soils because of its much finer texture (15% sand, 80% silt, 5%
clay) and its higher organic carbon content, 1.5% (R. Bowman, personal communication). The
water-saturated hydraulic conductivity of the soil was measured to be 5 x ifr6 cm/s using a falling head
test, and the particle density was measured to be 2.67 ± 0.02 g/cm3.
- 41 -
-------
Figure 4-11 is a capillary pressure-saturation curve for the Palouse loam, constructed for a water
saturated being drained by Soitrol. The Soltrol did not readily drain the loam over the range of suctions
available in our laboratory using the short glass columns and applied vacuums. The maximum Soltrol
saturation reached was 11.7%, with an corresponding water saturation of 88.3%. Soltrol broke through
the capillary barrier at the end of the column at the highest suction. We were not able to develop the
imbibition portion of the curve, or the final estimate of the residual organic liquid saturation. Apparatus
improvements are needed to examine organic liquid behavior in this type of soil.
600-
Suction
(cm H2O)
540-
480-
420-
360-
300-
240-
180-
120-
60-
0
SOLTROL
Soltrol breakthrough
I I I I I I I I I
80 82 84 86 88 90 92 94 96 98 100
Water Saturation %
FIGURE 4-11. Water saturation versus capillary pressure for Soltrol draining water from the
Palouse loam.
- 42 -
-------
SECTION 5
SHORT COLUMN EXPERIMENTAL METHODS
Bench top short column experiments were performed to measure the residual organic liquid
saturations of soils under either vadose zone or saturated zone conditions, providing a direct means to
compare the amount of organic liquid trapped in the vadose zone to the amount trapped in the saturated
zone. The Sevilleta soil was used in most of these experiments. It is a uniform, medium-sized, quartz
sand. Soltrol-130, a mixture of C10 to C13 isoparaffins with negligible solubility in water, served as the
organic phase. The aqueous fluid used in the experiments was water with 3000 ppm calcium chloride
added to prevent dispersion of clays.
In other saturated-zone experiments kerosene, regular leaded gasoline, n-decane, p-xylene, and
tetrachloroethylene were used to directly compare the differences in residual saturations to different
classes of organic pollutants. Residual saturations were also measured for several soils under
saturated-zone conditions and these results were compared to the Sevilleta results and the published
results of petroleum reservoir engineers.
A short glass chromatographic column with threaded Teflon™ tetrafluoroethylene (TFE) endcaps
was used to contain the soil sample. The glass column had an inside diameter of 5 cm and an effective
internal length of about 5 cm. The column was kept short so that — especially in the vadose zone case —
saturations remained fairly constant over the length of the column. The short column also maximized the
accuracy of the gravitimetric measurements of saturation, since the greater weight of a longer column
would have required the use of a greater capacity but less accurate balance. Finally, the short column
allowed for easy packing with soil, and minimized deairing times. Water or oil wet membranes were used
on the lower endcap to minimize capillary end effects. A paper filter placed on the upper endcap
prevented clay-sized particles from leaving the column. Over the course of an experiment, the change in
column mass was used, in conjunction with the known density difference between the fluids, to measure
saturations. Soils and liquids were exposed only to glass, Teflon, and chemically resistant tubing during
the experiments.
In the saturated zone experiments, organic liquid was introduced into an initially water-saturated soil
sample until the so-called irreducible water saturation was reached. Water was then re-introduced into
the soil displacing most of the organic liquid, but leaving behind some discontinuous blobs of organic
liquid trapped by capillary forces. The fraction of the pore space occupied by these trapped blobs is the
residual saturation. The experiment represents a scenario in which organic liquid percolates into the
saturated zone and is, in turn, displaced by ambient groundwater flow, leaving behind a residual organic
liquid saturation.
In the vadose zone experiments, an initially water-saturated soil sample was drained with air under an
applied suction until an equilibrium was reached. The drained soil represented vadose zone conditions in
which the pore space is occupied by both air and water. As organic liquid was introduced (simulating a
spill or leak), it displaced air and water as it percolated through the soil. Organic liquid was then drained
- 43 -
-------
from the soil under an applied suction, simulating the downward movement of organic liquid as it passes
through the vadose zone toward the water table. Again, capillarity caused some organic liquid to
become trapped in the pore space.
In this section, we begin by describing the fluids and soils used in these experiments and the
rationale for our choices. (A detailed characterization of the fluids and soils used in the experiments is
presented in Section 4.) We then describe the short column apparatus, followed by detailed procedures
of how the apparatus was used to measure fluid saturations under two-phase (saturated zone) and
three-phase (vadose zone) conditions. We conclude the section by discussing the limitations of the
apparatus and possible errors that may occur when running these experiments.
FLUIDS AND SOILS (see Section 4)
The aqueous fluid used in all experiments was distilled, de-ionized, de-aired water with 3000 ppm
CaCI2 added to prevent dispersion of clays. Saturated zone residual organic liquid saturations were
determined for six organic liquids; three of the liquids were mixtures (regular leaded gasoline, kerosene,
and Soltrol-130), and three were single-component liquids (n-decane, p-xylene, and
tetrachloroethylene). Soltrol-130 was also used to develop the experimental procedure and to perform
the vadose zone short column experiments. The other organic liquids selected for use in the experiments
were chosen to be representative of several classes of organic liquid pollutants often present at landfills,
hazardous waste sites, and accidental spills.
Sevilleta sand was used to refine the test apparatus and procedure. To hold all variables but the
independent variable of interest constant, this soil was also used in all experiments comparing residual
saturations in the saturated zone to those in the vadose zone, and for all comparisons of residual
saturations between organic liquids. The Traverse City soil and Palouse loam were used in several trials
of the short column saturated zone experiment. The Llano soil was used in the long column (see Section
6).
EXPERIMENTAL APPARATUS
Each apparatus used in the short column experiments consisted of a short glass chromatographic
column with tetrafluoroethylene (TFE) endcaps (Figure 5-1), and associated plumbing. The glass
column and TFE endcaps were manufactured by Ace Glass, Inc. The glass column was specially
fabricated to our specifications: 5 cm inside diameter and 5 cm effective internal length between TFE
endcaps. The effective column length was kept short in order to keep fluid saturations relatively constant
along the length of the column. The endcaps were screwed into threaded ends on the glass column and
sealed against the column with o-rings.
A 5 mm thick fritted-glass disk with a 20 micron average pore diameter was placed into the taper of
the bottom endcap as a filter support. A water-wet Magna 66 nylon filter with a 0.22 micron pore
diameter, designed to allow water, but prevent organic liquid from leaving the column, was glued along
- 44 -
-------
5 cm
Nylon
Filter
Scrims
Column Side View
Bottom Endcap
Top Endcap
FIGURE 5-1. The short column apparatus, with blow-up views of the endcaps and filters.
the edges to the fritted-glass disk. A paper filter was placed over the nylon filter to protect the nylon from
abrasion by the soil.
A network of small channels, approximately 1 mm deep and 1.5 mm wide, were machined into the
surface of the top endcap, in a pattern similar to that shown in Figure 5-1. The grooves allowed for a
more uniform flow of fluids between the endcap and the soil. A paper filter was sandwiched in between
two polypropylene scrims and glued to the endcap. The paper filter kept fine soil particles from leaving
the column. The outer scrim kept the paper filter from tearing when the endcap was screwed down
against the soil. The inner scrim decreased clogging by preventing the paper filter from sagging back
into the grooves.
In vadose zone experiments, endcaps with organic-wet filters were used as part of the experimental
procedure (as described later in this section under Vaotose zone experiments'). Endcaps with
organic-wet membranes, which allow organic liquid to pass through but not water or air, were
constructed in one of two ways. By the first method, the endcaps were constructed identically to
water-wet endcaps, but an organic-wet TFE filter was glued in place instead of a water-wet nylon filter.
The TFE filters, purchased from Gelman Sciences, had an average pore diameter of 0.5 microns. By the
- 45 -
-------
second method, an organic-wet ceramic disk was simply glued into place in an endcap, in place of the
fritted-glass disk, the membrane, and the paper filter. Before it was glued in place, an initially water-wet
ceramic disk was treated with an organosilane compound to change its wetting to organic-wet. To
change the wetting, the disks were first soaked in chlorotrimethylsilane for 2 hours, then immersed in
toluene for 1/2 hour, and finally rinsed thoroughly in methanol for 1 hour. A similar procedure was used
by Lenhard and Parker (1987), and previously by many experimenters in petroleum engineering (see
review by Anderson, 1987a). The disks were one-half bar, high-flow ceramic disks, purchased from Soil
Moisture Equipment Corporation. The ceramic disks were found to maintain their organic-wet properties
much better if they were stored in wetting (organic) fluid when not in use. Because the experimental
procedure called for these endcaps to be screwed into place against the soil, the TFE membrane could
become damaged by abrasion from the soil grains. Use of the much more rugged ceramic disks
alleviated this problem, but since the disks offered a much higher resistance to flow, the experiments
proceeded at a much slower pace.
Devcon 2-ton epoxy was used to glue filters and disks on all endcaps except those used in
experiments employing PCE as the organic liquid. Since PCE dissolved the epoxy, silicon sealant was
used to attach filters to the endcaps in any experiment's involving the use of PCE.
Filter Testing
For the saturated and vadose zone experiments to run properly, it was essential that filters maintain
their integrity. That is, the nylon filters must allow only water and not organic liquid or air to penetrate
under experimental conditions; while the TFE or treated ceramic filters must allow organic liquid to pass,
but not water or air. Capillary forces held the wetting phase in the pores of the filter, allowing only that
wetting phase to pass through. Non-wetting fluids could not pass through the filter, as long as the
non-wetting phase entry pressure of the filter was not exceeded.
An air-entry test was used as a simple means to test the integrity of a filter and its glued seal to the
endcap. The wetting fluid (water or organic liquid) was pulled through the filter and endcap under suction
using a vacuum pump (see step 1 in Figure 5-2), until the filter and any void space in the end reservoir
behind the filter was completely saturated. The suction on the vacuum pump was set to a performance
standard and run for a couple of minutes at that level. The endcap was removed from the wetting fluid
with the vacuum suction continuing (see step 2 in Figure 5-2). If air breached the filter (seen as air
bubbles in the tubing leading out of the endcap within 1 to 2 minutes), then the filter failed the air-entry
test and was not used. Filters which allowed no air to penetrate them passed the air-entry test and were
used in experiments.
The performance standards were set with a margin of safety so that filters could be assured of
maintaining integrity under experimental conditions for any combination of wetting and non-wetting fluids
(water-air, organic liquid-air, water-organic liquid). For example, about 60 cm of water suction was
needed to bring the Sevilleta sand close to irreducible water saturation. This suction was doubled to 120
cm (9 mm Hg) to ensure a margin of safety. To account for the fact that air was not always the
non-wetting phase, the integrity test standard was corrected by the ratio of interfacial tensions. For
Soltro! this correction was:
- 46 -
-------
vacuum
air bubbles
visible here??
wetting liquid
saturated
endcap
STEP 1 STEP 2
FIGURE 5-2. Air entry test for bottom end cap filter and seal.
120 cm H20 (water/soltrol) x — = 185 cm H20 (water/air) ~ 14 cm Hg (5-1)
where: owa =73.6 dyne/cm = water/air interfacial tension = water surface tension
awo = 47.8 dyne/cm = water/Soltrol interfacial tension
The performance standard for this particular air-entry test was set at 14 cm Hg.
Column Volume Measurement
Accurate measurement of the column volume was important because it was used in subsequent
calculations of bulk density, porosity, pore volume, and fluid saturations. To measure the column
volume, the column was weighed empty, filled with water, and weighed again. The difference between
the two weights divided by the density of water yielded the column volume, Vc • Allowances were made
for volume inside the column occupied by filters, the grooves in the top endcap, and how tightly the
column was screwed together to give what we call the 'effective column volume' ( Vce), the volume
occupied by soil and the fluids of interest. What follows is a detailed description of the procedures used
to accurately measure the effective column volume.
The mass of the empty column (Me) was determined by weighing the core body, endcaps with
valves, and o-rings using a Mettler PE 1600 balance with 0.01 g accuracy. The bottom endcap was
weighed while still full of water from the filter test, and the top endcap was weighed before the filters were
glued into place. The column was then assembled. The bottom endcap was screwed into the glass
column and tightened in place using a bench vise. The top endcap was only hand tightened into the glass
- 47 -
-------
column. The top endcap and glass column were marked to show this reference alignment. The column
was filled with water and reweighed. The column volume was calculated as:
Mw. - Me
y< = —i—^ (5-2)
where: MWl = mass of the water-filled column (g)
Qw = density of water (g/cm3)
Packing a column with soil was an imprecise science. The column might have been slightly over-filled
with soil one time, slightly under-filled the next. The top endcap could not be screwed down to the same
degree in an over-filled column resulting in a slightly larger volume inside the column. Conversely, a
slightly under-filled column might have been tightened down somewhat further than the reference,
resulting in a slightly smaller column volume than the reference volume. To account for this variability in
column volume from one packing to the next, a correction factor for endcap tightening was desired. To
accomplish this, the valve on the top endcap was opened and the previously hand-tightened endcap was
tightened in a bench vise. As the column volume was made smaller by the tightening, water squirted out
through the opened valve. A new mark, aligned with the glass column mark, was made on the top
endcap. The offset length between marks ( L,. ) was recorded, the column was reweighed, and the new
mass (MW2) was recorded. The column tightening correction factor ( C, ) was calculated as:
w w
C< = L o (5-3>
i*c (Jw
The volume of the top endcap (including the volume of the grooves, the connecting hole in the
endcap, and the valve) was measured in a manner similar to the method used to measure the gross
column volume. The endcap was weighed empty, the endcap was filled with a liquid of known density,
and the column was reweighed. Similar to before, the mass of the full endcap minus the mass of the
empty endcap divided by the liquid density provided a measure of volume which could be occupied by
fluid in the endcap. Because the endcap was made of Teflon, which is oil wet, an organic liquid did a
much better job than water of wetting the surface and filling the grooves.
The volume of the filter, scrims, and glue on the top endcap reduced the volume of the column
available to sand and fluids and also needed to be accounted for. The volume of the paper filter and the
polypropylene scrims was calculated by simply multiplying the cross-sectional area by the measured
thickness. The endcap, filter, and scrims were weighed prior to gluing the filter and scrims onto the
endcap. After the glue had hardened, the endcap was reweighed. The difference between the two
weights was the weight of the glue, and when divided by the glue density yielded the glue volume. The
density of hardened glue was determined using the ASTM D584-83.
The effective column volume was determined by accounting for the volume in the top endcap behind
the filters (including grooves, connecting hole, and valve), the volume of the paper filter and scrims on
- 48 -
-------
the top endcap, the volume of glue on the top endcap, and the correction for how tightly the top endcap
was screwed in place. So we calculated the effective column volume as:
Vce = K " ^groove ~ ^filter ~ Vglue ~ C, Lc (5-4)
where: Vce = effective column volume (cm3)
Vc = original column volume (cm3)
groove = volume of the top endcap grooves (cm3)
Vfute,. = volume of the paper filter and polypropylene scrims (cm3)
Vgiue = volume of the glue (cm3)
C, = column tightening correction factor (cm2)
Lc = length shy of mark (cm)
This procedure may seem to be somewhat elaborate, but was easy for an experienced technician to
master. By measuring all of the volumes suggested by equation 5-4 we reduced experimental error and
provided repeatable results. Some of the corrections in this equation would have proven less significant
in a longer column, but that would have required the use of a larger capacity balance to handle the
greater weight of the column. The available larger balances were much less accurate than the Mettler PE
1600 balance.
SOIL PACKING
The glass column was packed by hand in small increments under about 1 cm of water. A previously
determined mass of soil was gently poured into a vertically oriented column. The water was controlled by
a buret connected to the column through the bottom endcap. The soil was tamped into place using a
modified metal spatula (bent flat and turned 90 degrees at one end). The column was packed to a depth
of about 5 cm, reaching the top end, just at the base of the upper column threads. The soil was carefully
leveled and the top endcap was screwed into place. If the top endcap alignment mark was offset from
the mark on the glass column, the distance was measured as the length shy of mark (Lc). The effective
column volume was calculated according to equation 5-4.
Occasionally, several attempts at sealing the top endcap needed to be made before the top endcap
was successfully screwed into the column and the o-ring was seated properly. After the initial attempt,
soil often had to be added or removed from the column before a tight o-ring seal was obtained and the
endcap could be tightened in a vise. Any soil removed from the column was oven dried overnight and
combined with the leftover soil. The mass of soil in the column, Ms, was simply the original mass of soil
minus any leftover soil.
- 49 -
-------
The mass of soil packed into the column was used in determining the bulk density, Qb , the pore
volume, Vp , and porosity, n , of the soil pack:
M,
Qb = V^ (5-5)
M5
VP = V" ~ — (5-6)
fj
Ms
n = 1 — (5-7)
where: Ms = mass of soil in the column (g)
QS = particle density of the soil (g/cm3)
Upon completion of packing, the column was reconnected to the buret tubing to test the top
endcap's o-ring seal for water leakage. 0-ring leakage was tested by closing the top endcap valve,
opening the lower endcap valve and buret valve, and raising the input buret height, thus increasing the
pressure on the o-ring seal. It was usually immediately apparent if the seal leaked, requiring the endcap
to be refitted. A good o-ring seal was obvious, with the black o-ring appearing flat against the glass
column wall. Only rarely did the seal leak when the top endcap was tightened down using a bench vise.
On the other hand, when the top endcap was hand tightened, o-ring seal leakage was more common.
After the column had been tested for leakage, silicon sealant was placed between the outside of the
glass column and the top endcap. This prevented any change in column mass due to evaporation of any
water accidentally trapped in the threads between the glass column and top endcap.
In two of the experimental trials, the column was packed with coarse lenses within a finer matrix.
Sevilleta soil was split into two fractions for these trials, which were packed dry. The dry packing
procedures used were identical to those used to to achieve heterogeneous packings in the pore cast
experiments. A description of the procedures can be found in Section 7.
Some of the Palouse loam short columns were also packed under dry conditions. Wet packing was
difficult in this material. This problem is discussed at the end of this section under the topics 'possible
sources of error' and 'limitations'.
DEAIRING
The column was flooded with 20 to 30 pore volumes of deaired water in order to remove any gas
bubbles trapped in the porous matrix or plumbing. Entrapped gas dissolved into the deaired water and
was carried out of the column. After every 100 ml of water added, the column was disassembled from its
plumbing and weighed. The column gained mass as entrapped air was removed from the column and
replaced by water. The column mass eventually stabilized as all entrapped air was removed.
Operationally, we used three consecutive unchanging measurements of column mass to indicate the
- 50 -
-------
complete removal of entrapped air. The deairing of the soil column took anywhere from 5 to 10 days for a
sandy soil (eg, Sevilleta, Llano, Traverse City), and up to 3 weeks for an initially dry silty-clay soil
(Palouse loam). The mass of the column at this point was called the initial column mass ( M, ). The
water-saturated column could now be used for capillary pressure-saturation tests (see Section 4), the
saturated zone experiment (see below and Section 9), or the vadose zone experiment (see below and
Section 10).
SATURATED ZONE EXPERIMENTS
Saturated zone residual organic liquid saturation was determined by introducing organic liquid into a
water saturated packed soil column (organic liquid flood), then displacing the organic liquid by water
(water flooding). These experiments simulated conditions beneath the water table.
Step 1: Organic Liquid Flooding
Once the column had been packed with soil and any entrapped air removed, organic liquid was
flooded into the column simulating the movement of organic liquid into the saturated zone. To maintain a
stable displacement front, the column was flooded from top to bottom for the case of an organic liquid
lighter than water (Figure 5-3), or flooded from below when an organic heavier than water was used. The
organic liquid was injected under sufficient head so that the water saturation was reduced to the
so-called irreducible water saturation, but the head was kept low enough to prevent organic liquid from
breaking through the nylon filter on the bottom endcap.
All organic and aqueous liquids were loaded into the burets by a siphoning procedure to reduce
aeration of the fluids. To avoid contact with hazardous vapors, p-xylene, tetrachloroethylene, and
gasoline were loaded into burets with a vacuum suction under a fume hood. The openings at the top of
the burets were sealed to each other by a stopper and tubing arrangement (not shown in Figure 5-3) to
close the system and limit emission of fumes.
The organic liquid flooding was continued for at least 24 hours or until the column weight stabilized,
indicating that the fluid saturations had reached equilibrium. A stabilization of fluid levels in both the
inflow and outflow burets also indicated that the system had reached equilibrium and fluid saturations had
stabilized. Fluid production at the bottom endcap was observed to ensure that only water was produced
(verifying that the nylon filter had maintained its integrity and no organic liquid had breached the filter).
After stabilization, the column was detached from the burets and the column mass, M2, was measured
(using at least three independent weighings on a Mettler PE 1600 balance with 0.01 g accuracy). The
fluid saturations were determined employing the density difference between the organic liquid and water:
Ml-M2
S° -^^ ^
Swr = 1 - S0 (5-9)
- 51 -
-------
where: S0 = Organic liquid saturation (-)
A@ = Density difference between water, Q* , and the organic liquid, Go ,(g/cm3)
Swr = Residual water saturation (same as irreducible water saturation, Sm ) (-)
Modifications to the procedure were made in an attempt to introduce organic liquid (Soltrol) into the
Palouse loam soil. Because the soil was fine-grained and water wet, the Soltrol did not readily enter the
column. Compressed air was used to increase the pressure behind the Soltrol and force it into the
water-saturated soil column. One tubing line was connected to an air compressor and split with a tee.
One split line went to a mercury manometer board to record the pressure and the other line was
connected to the buret containing Soltrol, through a stopper inserted on top of the buret. In this manner,
a capillary pressure of about 500 cm of water was induced on the Palouse loam. Still, we had great
difficulty pushing Soltrol into the soil.
organic
liquid
organic
liquid
enters,
7
i
water
exits
water
7
FIGURE 5-3. Saturated Zone Test Step 1: Organic liquid flood into a water
saturated column.
- 52 -
-------
organic
liquid
organic
liquid
exits
Syringe pump
FIGURE 5-4. Saturated Zone Test Step 2: Waterflooding at low velocity to reduce the organic
liquid to its residual saturation.
Step 2: Water Flooding
Following the organic liquid flood, the column was slowly flooded with water in order to displace the
organic liquid. The sandy soils (eg, Sevilleta, Llano, Traverse City) were flooded at a rate of about
0.3 ml/min. The Palouse loam was flooded at a much slower rate.The water flood was intended to
simulate the action of ambient groundwater displacing the organic liquid and leaving behind trapped
organic liquid at residual saturation. As shown in Figure 5-4, the column was flooded from bottom to top
to promote a stable displacement of the less dense organic liquid. When an organic liquid denser than
water was used, the column was flooded from above, displacing the denser fluid out the bottom. A
syringe pump, or a carefully controlled and monitored buret flow was used to push the water through the
column at a low velocity to make sure the displacement proceeded under low Bond number and low
capillary number conditions (see Section 9 for definitions of Bond and capillary numbers). The waterflood
continued until no additional organic liquid was produced (as indicated visually through the transparent
- 53 -
-------
tubing and by an unchanging column mass). Introduction of about nine pore volumes of water through
the column was found to be sufficient to reach residual organic liquid saturation for the sandy soils. At the
conclusion of the water flood, the column was weighed, the mass (A/3) was recorded, and the residual
organic liquid saturation was determined:
MI - M3
Sw = 1 - Sor (5-11)
where: Sor = Residual oil saturation (-)
Sw = Final water saturation (-)
VADOSE ZONE EXPERIMENTS
The vadose zone experiments were designed to represent organic liquid trapping above the water
table where air, water, and organic liquid are simultaneously present. The simulated residual water and
organic liquid saturations were obtained under equilibrium conditions, at an equivalent height above the
water table. To achieve the aforementioned saturations, water was drained from an initially water
saturated, de-aired soil column under an applied suction. Then organic liquid was introduced into the top
of the column. Finally, the organic liquid was drained from the column under the same applied suction.
These experiments were performed only with Soltrol in the Sevilleta sand.
Step 1 : Water Drainage
Beginning with a de-aired, entirely water saturated column of known mass (Ml ), a suction was
placed on the column, draining water through the nylon filter and out the bottom endcap (see Figure
5-5). Air entered through the top endcap, replacing the drained water in the soil's pore space. The
water-wet nylon filter allowed water, but not air to pass through the bottom of the column. The suction
was applied by lowering the water level in the buret beneath the column in the same fashion as when
determining water/air capillary pressure-saturation functions. Figure 5-5 shows the configuration used
for applying a suction Ah to the column. The elevation difference Ah was measured from the middle of
the soil column to the water level in the buret in centimeters of water. Equilibrium was reached when the
water level in the buret stabilized and the column weight no longer changed.
Upon reaching equilibrium, the column was weighed and the mass (M4) was recorded. Air
saturation was a function of the suction applied and was measured at a particular equilibrium suction as :
_
, initial ~
Qw
, initial ~ ~~ a, initial
(5-13)
- 54 -
-------
air enters
water
exits
water
FIGURE 5-5. Vadose Zone Test Step 1: Water being drained with air under an
applied suction.
where: Saiinl,ia, = air saturation following water drainage (-)
Sw.Mtiai = water saturation following water drainage (-)
The density of water, Qw, was used in the denominator of equation (5-12), instead of the density
difference between fluids as in previous equations. The density of the air is negligible, and in any case
the measurements were made on a balance exposed to air at room temperature and pressure.
The saturations found by this method were the fluid saturations for the soil under equilibrium
conditions Ah centimeters above the water table. One reason the effective column length was kept short
(5 cm) was to ensure reasonably constant fluid saturations along the length of the column.
Step 2: Organic Liquid Flooding
Once the water had been drained from the column, organic liquid was introduced to simulate the
effect of organic liquid percolating through the vadose zone. Before proceeding with the organic liquid
flood, the bottom endcap with the water-wet filter was removed from the column, and replaced with an
- 55 -
-------
identical endcap with an organic-wet TFE filter or organic-wet ceramic disk in place of the nylon filter. The
TFE filter or ceramic disk was used as a vehicle through which to introduce the organic liquid into the
column. In a later step it was used to pull organic liquid from the column under suction. To change
endcaps, the column was inverted and the bottom endcap was unscrewed from the soil column. Any soil
grains clinging to the paper filter on the bottom endcap were gently brushed back into the column. The
replacement endcap with a TFE or ceramic filter was screwed down tightly in place to ensure good
contact with the soil. Prior to installation, the TFE or ceramic filter was tested for integrity and the end
reservoir behind the filter was saturated with organic liquid. The column was then reweighed (Afs) to
account for any difference in the mass of the endcaps. Switching endcaps appeared to cause little
disruption of the soil packing.
The column was once again attached to the burets (see Figure 5-6). Organic liquid was introduced
through the organic-wet bottom endcap. Because water is more dense and air is less dense than the
organic liquid used in the experiments, the column was turned horizontally to inhibit density instabilities
that might occur in a vertical displacement. Organic liquid was pushed through the column under low
capillary number conditions until no more water or air was produced (verified by a stable column mass,
organic liquid
organic
liquid
enters
water,
air, and
organic
liquid
exit
D/7
FIGURE 5-6. Vadose Zone Test Step 2: Organic liquid flood in a column already drained by air.
- 56 -
-------
M6 ). This step usually required about 300 ml of organic liquid to be pushed through the soil column
before the mass stabilized, and took 2 to 3 days to complete.
Once the fluid saturations had equilibrated, the volume of water produced from the soil pack was
measured. The volume of water produced was measured by collecting the outflow (water and organic
liquid) in a TFE flask and withdrawing the water from the bottom of the flask with a syringe. The water, as
well as a small amount of organic liquid pulled into the syringe, was then injected into a graduated
cylinder to determine the volume of the water produced (Vwp). The new water saturation was then
calculated as:
y
^w ~~ ^wtmitiai ~" .,W~' v
VP
With the water saturation known, the organic liquid saturation (S0 ) and the air saturation (Sa ) could
be found using mass measurements of the column before (M5) and after (M6) the organic liquid flood:
_ M6-M5-QW Vwp
° ~ !T~v (5 5'
(Jo 'p
Sa = 1 - Sw - S0 (5-16)
Step 3: Organic Liquid Drainage
In the final step of this process, organic liquid was drained from the column under an applied suction.
The organic liquid saturation was reduced to an equilibrium saturation for a given height above the water
table. For sufficiently large suctions, this represents the residual organic liquid saturation in the vadose
zone. This residual remains interconnected, unlike the saturated zone's disconnected blobs. Figure 5-7
shows the set-up for organic liquid drainage. It was very similar to the set-up used earlier in step 1, for
draining water from the soil column (Figure 5-5). The drainage was a stable displacement, in which air
entered from the top displacing organic liquid downward through the lower organic-wet filter. The
elevation difference, Ah, for this step was equal to the Ah used in the previous water drainage step. For
instance, if the suction used in the water drainage step was 70 cm of water, in the organic liquid drainage
step the applied suction was 70 cm of organic liquid. Although in each case the fluid column in the buret
was dropped the same distance beneath the soil column to induce suction, the capillary pressures
induced in each of these steps was not the same. The difference in capillary pressures was scaled by the
relative densities of the water and organic liquid phases.
- 57 -
-------
The displacement proceeded until the organic liquid and air saturations equilibrated (the water
saturation remained unchanged during this step). The equilibrated column was weighed (M7 ) and the
final organic liquid, water and air saturations were determined:
A/7 - MS - QW vwp
Qo Vp
(5-17)
(5-18)
POSSIBLE SOURCES OF ERROR
Several sources of experimental error were associated with performing multi-phase flow
experiments to determine residual organic liquid saturations. Possible sources of error included:
• incomplete removal of entrapped gas,
air enters
organic
liquid
exits
T
Ah
1
Q/7
organic liquid
FIGURE 5-7. Vadose Zone Test Step 3: Organic liquid drained by air.
- 58 -
-------
• changes in fluid densities due to changes in laboratory temperature,
• capillary end effects,
• lack or loss of filter integrity,
• faulty seals in the system leading to leakage or evaporative losses,
• trapping of fluids in the outflow end reservoir,
• packing variability (from one soil packing to the next),
• dry packing,
• soil sample variability,
• limited column length, and
• error associated with the precision limits of the measuring devices.
What follows is a brief discussion of each of these sources of error.
Entrapped Gas
As soil was packed into a soil column, some gas was trapped in the pore space. For a water-wet
system, trapped gas tends to reside in pores which would probably otherwise trap organic liquid (see for
instance, Kyte et al., 1956, and our photographs in Section 10). Residual saturations measured in the
presence of trapped gas would be expected to be lower than residuals measured under strictly
two-phase conditions.
Entrapped gas presents other problems because its saturation may not remain constant over the
duration of the experiment; some of it may dissolve into the liquid phases. When gravimetric
measurements are used to determine fluid saturations, the loss of entrapped gas over the course of the
experiment would indicate a lower residual saturation than is actually present (for an organic liquid less
dense than water).
Entrapped gas was eliminated by flushing the column with several pore volumes of degassed water.
As degassed water moved through the column, trapped gas dissolved into the water phase and was
carried from the column prior to beginning the experiment. The column gained weight as the pore space
previously occupied by gas was then occupied by water. All entrapped gas was assumed to be removed
once the column weight stabilized.
Variable Laboratory Temperature
The worst and most persistent problem encountered during this project was our inability to control
temperature variations in our laboratory. As the temperature in the laboratory fluctuated, so too did the
densities of the fluids used in experiments, introducing error into gravimetric determinations of fluid
saturations. In addition, we had observed fluctuations in column weight that were larger than could be
- 59 -
-------
accounted for by density effects alone. It is believed that as the temperature rose, dissolved gasses
came out of solution and were trapped in the soil — this gas may have subsequently re-dissolved as the
temperature fell. This problem was especially acute when liquids containing volatile components, such
as gasoline, were used.
Large temperature fluctuations in the laboratory over the course of an experiment led to poor
reliability of the results. Measured residual saturations of Soltrol in Sevilleta sand are plotted against
laboratory temperature variation in Figure 5-8. The temperature range in this figure represents the
difference between maximum and minimum observed room temperatures recorded during an
experiment. Measurements were made periodically in the early experiments, while the others were
recorded on a strip chart. This figure illustrates the large variation of residual saturations that were
measured when temperatures were not held constant over the course of an experiment. The
experiments conducted with small temperature variations (< 2°C) show much less scatter than
experiments conducted under less controlled temperature conditions.
To remedy the problem of fluctuating temperatures during the experiments, a constant temperature
cabinet was constructed. Experiments performed within the constant temperature box displayed a
noticeably smaller variation of residual saturations than experiments performed prior to completion of
the box.
Residual
Saturation
36
34
32
30
28
26
24
22
on
4
• 4
/
• •
• •
•*
•
•
1
»
•
•
•
• =
complete
d experin
lent
•
2468
Temperature Fluctuation (°C)
10
12
FIGURE 5-8. Temperature range and its effect on the accuracy of results.
- 60 -
-------
Capillary End Effects
Capillary end effects can become important sources of experimental error in multiphase fluid
experiments. An example of capillary end effect, for Soltrol displacing water in a micromodel, is shown
and discussed in Section 9. Filters that allow only a single phase to pass through reduce these effects.
For example, a water wet filter at the discharge end of the short column prevents the non-wetting phase,
the organic liquid, from breaking through. A similar filter, for air advancing into a water filled micromodel,
is demonstrated in Section 10. The water and oil wet filters used in the short column appear to have
performed at least as well. Inspection of saturated zone pore casts constructed with the TFE columns
(Section 7) and water wet filters revealed a minimal capillary end effect, extending less than 2mm into
the soil. A longer column would have minimized this effect even more. However, there are other
problems with long columns, including packing, deairing, and accurate recording of column mass, as
discussed earlier.
Filter Integrity
If filter integrity was not maintained, end effects were not eliminated and the non-wetting fluid could
have become trapped in the reservoir behind the filter. In our column-flooding experiments, each filter
was tested for integrity prior to use, and paper filters were glued above them to reduce abrasions from
the soil which could have caused leakage. For experiments run at low flow rates, ceramic disks were
sometimes used instead of filters because of their greater durability.
Leaking Seals
The endcaps were sealed to the column using o-rings. Each apparatus was pressure tested for
leakage prior to use.
Outlet End Reservoir
In these experiments we assume that no organic liquid was trapped in the outlet end reservoir (top
endcap), although this assumption was probably not entirely true. By keeping the outlet end reservoir
small — less than 1% of the column pore volume — the effects of trapping within the reservoir were
considered negligible.
Packing Variability
Variable soil packing from one experimental trial to the next could have lead to some variability in
measured residual saturations. Determination of soil bulk density provided a good measure of the
'tightness' of each packing, but says little about the uniformity of a hand tamped column. Packing the
columns under water may have produced some small-scale layering in the Sevilleta sand. The very small
fraction of clay sized particles in this soil formed a suspension in the water ponded above the soil surface
during packing. The presence of such small particles explains the need for the paper filter at the outlet of
- 61 -
-------
the column, in order to prevent clay migration out of the column, which would have effected the
gravitimetric methods. Some of these particles may have preferentially settled out as thin layers
between packing lifts. Our soil characterization efforts did not verify their presence. There is no reason to
suspect a similar packing problem with the Traverse City or Llano soils, both of which appeared to be
washed and without similar clay sized fines.
Dry Packing
The Palouse loam results reported in Section 9 were packed under dry conditions. Dry packing
increased the length of time required to de-air the column by two to three weeks. Cracks in the soil,
formed when the top endcap was tightened down, were a common problem. These cracks appeared to
heal when the column was wetted. No cracks were observed for the dry heterogeneous Sevilleta soil
packs. Pore casts of these packs, made using the techniques of Section 7 and shown in Section 9, also
gave no evidence of cracks.
Soil Sample Variability
The soils used in these experiments were natural materials collected from field sites in New Mexico
and Michigan. The New Mexico soils (Sevilleta and Llano) were collected from outcrops. The Michigan
soil (Traverse City) was subsurface sample taken by EPA's Kerr Lab staff with an auger rig coring device.
The initial batch sample of Sevilleta soil was insufficient in volume to complete the studies described in
this report. Several additional batch samples were collected. It is probable that these samples, each of
which were mixed and split in the standard way, differed somewhat in composition from each other and
from the original. The sieve analyses, pore pressure-saturation curves, and other soil characteristics
were not repeated for each sample, although they should have been. The small amount of variability of
measured residual saturation, for tests conducted with appropriate temperature control, indicates that
the variation from sample to sample was not significant. Never-the-less, this issue should be kept in
mind when reviewing the data for the Sevilleta soil. The small sample sizes for the other two soils limited
the number of tests that could be performed.
Short Column Length
The column was kept short so that — especially in the vadose zone case — saturations remained
fairly constant over the length of the column. The short column also maximized the accuracy of the
gravitimetric measurements of saturation, since the greater weight of a longer column would have
required the use of a greater capacity but less accurate balance. The Mettler PE 1600 balance had an
accuracy of 0.01 g. A higher capacity balance had an accuracy of 0.1 g. The short column allowed for
easy packing with soil, and minimized deairing times. Never-the-less the short length increased the
importance of minimizing capillary end effects, and making accurate measurements of fluid mass stored
in the end reservoirs.
Measurement Error
Measurement error, due to the precision limits of measuring devices such as balances, were
estimated and propagated through the sequence of calculations. For example, calculations of fluid
4
- 62 -
-------
saturations were dependent upon measures of fluid densities, soil particle density, total soil weight, total
column volume, and measures of the column mass at several points in the progress of an experiment.
All these measures have some error (or uncertainty) associated with them and these errors are routinely
propagated through the calculations used to determine fluid saturations. We used a worst case error
approach in which we assumed all errors were additive. Since some measurement errors would be
expected to cancel each other out, this worst case approach was a conservative estimate of the total
measurement error.
LIMITATIONS OF THE APPARATUS AND TECHNIQUE
The experiments with Palouse loam pointed out some of the limitations of our apparatus. Because of
the fine-grained nature of the soil, it was difficult to supply sufficient pressure to force organic liquid into
the soil. We used pressure equivalent to more than seven meters of Soltrol to overcome the entry
pressure for a non-wetting phase. Even if we had been able to inject the organic liquid with sufficient
pressure, we might well have also exceeded the entry pressure of the nylon filter on the bottom endcap,
resulting in failure of the filter. In order to accommodate fine-grained soils, the apparatus would have to
be re-designed as a high-pressure system, including a high-pressure fluid delivery pump, a column built
to withstand high pressures, and a bottom membrane with a very high non-wetting phase entry pressure.
Our inability to inject an organic phase into the Palouse loam is not a total loss. It indicates that
fine-grained, water-wet soils (which do not shrink and crack in the presence of organics) can serve as an
effective barrier to organic liquid movement in the subsurface.
Another limitation of the experimental technique occurred in the vadose zone experiments. Several
trials of the vadose zone experiment were run using high applied suctions to measure the saturation of
organics left behind a front of organics percolating through the vadose zone high above the water table.
Additional vadose zone experiments were performed over a range of applied suctions. These
experiments were run to give an idea of the saturation distributions that can be expected in the transition
zone between the saturated zone and the vadose zone. Results from these transition zone experiments
gave what we believe to be unreasonably high equilibrium organic liquid saturations, reported in Section
10. The limitation of our experimental procedure was that, during the organic liquid drainage step, water
was unable to re-imbibe into the soil. Under low applied suctions representing the capillary fringe and just
above, re-imbibing water would have displaced organic liquid resulting in lower equilibrium organic
saturations than the values we have measured. The largest change in forces acting on the organic liquid
occurs in the capillary fringe and just above where the transition from the vadose zone to saturated zone
conditions is most pronounced. Unfortunately, the experimental procedure used to measure
three-phase saturations over this range was found to be inappropriate. This low suction range is
important particularly for organic liquids lighter than water because it is the zone in which gravimetrically
light organic liquids spread, forming a lens on the water table.
- 63 -
-------
SECTION 6
LONG COLUMN EXPERIMENTAL METHODS
Organic liquid residual saturation in the saturated zone can be reduced by mobilizing some of the
blobs through an increase in hydraulic gradients and water flow rate (Wilson and Conrad, 1984). This
concept has long been known in petroleum reservoir engineering and involves increasing the viscous
forces of the groundwater flow above a threshold value needed to overcome the capillary forces trapping
the larger blobs (see, eg., Anderson, 1987b; Chatzis and Morrow, 1981; Chatzis et al., 1984,1988;
Morrow, 1979; Morrow etal., 1988;Taber, 1969). Alternatively, the interfacial tension between the water
and the organic liquid can be decreased below an equivalent threshold (see e.g., Tucket al., 1988). This
is the principle behind surfactant floods in enhanced oil recovery (eg., Taber, 1981). Reduction of
residual saturation by mobilization is also of interest in aquifer remediation studies.
The saturated zone hydraulic mobilization of the residual saturation of Soltrol-130 was correlated to
capillary number for the Sevilleta and Llano soils, both unconsolidated aquifer materials. Experiments of
this kind had previously been conducted in consolidated petroleum reservoir cores (eg., Chatzis and
Morrow, 1981; Chatzis et al., 1984,1988) and in glass beads (eg., Morrow et al., 1988), but not for
unconsolidated materials typical of shallow aquifers that are particularly susceptible to organic pollution.
The capillary number represents a ratio of viscous (flowing) forces to capillary forces.
Each experiment began with a long water-saturated column, into which organic liquid was injected,
followed by the injection of water under low capillary number conditions to reduce the organic liquid to its
residual saturation. Absolute and relative permeabilities were measured during these floods. After the
water flood the organic liquid was reduced to residual saturation under low capillary number conditions.
In the last step the water flow rate through the column was increased incrementally. Above a critical flow
rate (i.e., a critical capillary number), the residual organic liquid saturation in the column became
reduced as the force of water flow began to overcome the capillary forces holding the organic liquid in
place. By measuring the reduction of residual saturation versus flow rate and pressure drop, a
correlation was constructed relating the mobilization of trapped organic liquids to capillary number. At
each residual organic liquid saturation, the relative permeability to water was measured under
steady-state conditions.
A long column was used in these experiments in an attempt to maximize the accuracy of pressure
gradient measurements. The long column also minimized end effects, but at the cost of increased
difficulty in packing, longer de-gassing times, and lower accuracy gravimetric saturation measurements.
The long column also allowed us to make measurements of relative permeability during the flooding
stages using the unsteady state method (Jones and Roszelle, 1978).
LONG COLUMN APPARATUS
The set-up for the long column experiment is shown in Figure 6-1. The column itself was constructed
from a 105 centimeter long, 2.5 inch I.D., schedule 80 PVCpipe, and could accommodate soil packed to
- 64 -
-------
a
3
0)
E
*c
a>
a
c
E
^
o
o
O)
c
o
(O
UJ
a:
- 65 -
-------
about 100 cm in length. The column endcaps were made of tetrafluoroethylene (TFE) and sealed with
o-rings. Construction details for one end of the column are shown in Figure 6-2. Each end of the column
was of identical construction. PVC and TFE were chosen for the column because of their non-reactivity in
this experiment, and ease of machining. There is some possibility that the organic liquid phase might
preferentially wet PVC, but there was no evidence for this in the experimental results. Residual water and
organic liquid saturations were equivalent to those observed under similar conditions in the short glass
columns. (The short TFE columns described in Section 7 did show effects of preferential organic liquid
wetting; see Appendix A.)
The long column endcaps were constructed of TFE rod. Part of each rod was machined to just fit
inside the PVC cylinder. Grooves to accommodate two o-rings were machined into the sides of the each
endcap. The inner face of each endcap had radial and concentric grooves machined into it to allow
better fluid flow between the soil sample and the 1/16 inch endcap center hole. A polypropylene scrim
and a paper filter were glued to each inner face. The paper filter kept fine soil particles from leaving the
column and the polypropylene scrim kept the paper filter out of the grooves. Each endcap had an
aluminum plate screwed into the back. Four threaded rods passed through the aluminum plate and
through an aluminum ring mounted on the column, providing a means of attaching the endcaps,
essentially bolting them on. Nupro plug valves were threaded through the aluminum plates behind each
endcap and were sealed against the TFE endcap with o-rings. Near either end of the column, three
pressure transducers were mounted into the sidewall of the column. Each transducer was screwed into
the female end of a 1/4 inch NPT male-female PVC valve. A porous ceramic disk was glued to the male
end of the valve, and the valve was threaded into the sidewall of the column. The column sidewall was
thickened where the transducers were mounted by machining a short length of 3 inch PVC pipe to just fit
over the column and by gluing it in place. The thicker column wall allowed enough threads to be tapped
into the sidewall to hold the transducers, and the thicker wall provided a stop for the aluminum ring on the
column.
Fluids were injected into the column using a Pulsa 680 diaphragm pump, and the flow rate was
monitored with a Gilmont float-type (rotameter) flowmeter. The water phase was prepared as described
in Sections 4 & 5. Only water passed through the pump and flowmeter. When organic liquid was to be
injected into the column, valve A in Figure 6-2 was closed and valve B opened, so that water entered the
organic liquid reservoir displacing organic liquid from the reservoir and into the column. While not in use,
water was stored under vacuum to keep it de-gassed. When pumping from the water reservoir, a helium
atmosphere was maintained above the water to help keep the water as de-gassed as possible. (Helium
gas has a very low solubility in water.)
Three pairs of Omega PX-800 series pressure transducers were tapped into the PVC sidewall at
either end of the column to measure pressure drops across the length of the column. One set of 20 psi
transducers measured pressures in the water phase; another 20 psi set measured pressures in the
organic phase; and the third set, having a 200 psi capacity, measured water phase pressures under high
pressure conditions — pressures exceeding the 20 psi capacity of the other set of water phase
transducers. The Omega PX-800 series pressure transducers were selected for use in these
experiments because of their high accuracy and low weight.
- 66 -
-------
FIGURE 6-2. Long column construction details.
- 67 -
-------
Porous ceramic disks were built into the sidewall of the column, in contact with the soil. Water-wet
ceramic disks functioned similar to tensiometers, allowing the transducers to measure pressures in the
water phase. The reservoirs between the water-wet ceramic disks and the transducers were filled with
distilled water instead of the calcium chloride aqueous solution to reduce corrosion of the pressure
sensing diaphragms in the transducers1. For transducers which were to measure pressures in the
organic liquid phase, the wettability of the ceramic disks was changed to organic-wet prior to installation
(using the same organosilane procedure described in Section 5). Organic liquid filled the small
reservoirs between the organic-wet ceramic disks and the transducers. Ball valves were placed in the
reservoirs. When closed, the valves prevented the low-pressure transducers from being exposed to
damagingly high pressures. The organic liquid transducers were difficult to de-gas, and consequently
their response time was excessive. They were not essential to any of the experiments, and were
disconnected.
A 10 volt DC power supply served as input to the transducers. Full-scale output from the transducers
was 100 millivolts. A Metrabyte DASH-16 analog-to-digital interface board, installed in a CompuAdd
Standard PC-AT personal computer, transformed analog signals from the transducers to digital data
which could be stored in the computer. The signal from the transducers was boosted with an instrument
amplifier to meet the requirements of the interface board.
The mass of fluids produced from the column was measured using a Mettler PM 11 balance which
has a 11,000± 0.1 gram capacity. Since the densities of each fluid phase and the flow rate through the
column at any given time were known constants, the mass of fluid produced was used to monitor the
production of both fluid phases from the column as a function of time. Also, by employing mass balance,
the displacement data could be used as an independent means of measuring saturations in the column
— a check on the standard gravimetric method of measuring column saturations.
The Lotus 'Measure' software package was used to record readings from the balance and
transducers at specified intervals and to download the data into a Lotus 1-2-3 worksheet.
FLUIDS AND SOILS (see Section 4)
The column was packed with either Sevilleta or Llano soil . A 3000 ppm CaCI2 solution served as the
water phase, and Soltrol-130 was used as the organic phase because of its very low solubility in water.
Soil and fluid properties are reported in Section 4.
COLUMN PACKING and DE-GASSING
The long column was packed under procedures similar to the short column packing methods
described in Section 5. The column was aligned vertically with the bottom endcap in place at the lower
1. Any osmotic pressure generated by using distilled water in these small reservoirs and the CaCI2
solution in the column was negligible compared to the water pressures existing in the column.
- 68 -
-------
end. The bottom endcap was connected to a buret which supplied CaCI2 solution to the column. During
packing, the head in the buret was maintained so that the soil was packed into the column under
approximately 2 cm of water to reduce the amount of entrapped air in the column. Oven-dried soil was
poured through a funnel into the column and was tamped into place. Once the column was full, the top
endcap was bolted down tightly against the soil. The effective length of the soil in the column varied
slightly from one experiment to the next depending on just how much soil was packed into the column.
The mass of the soil packed into the column was recorded.
The procedures for removing entrapped air from the long column were similar to those described in
Section 5. At the conclusion of the packing procedure, entrapped air typically occupied from 5 to 7
percent of the pore space. To remove the entrapped air, the column was flushed with de-gassed CaCI2
solution at a slow rate. The column gained weight as the entrapped gas solubilized into the water phase
and was removed from the column. The column was weighed periodically, and once the weight had
stabilized the column was assumed to be de-gassed.
At the conclusion of the de-gassing procedure several useful quantities were calculated. The pore
volume of the column, Vp (in cm3), the volume occupied by soil in the column, Vs (in cm3), and the total
column volume, V, (in cm3), were determined by:
(6-2)
V, = VP + Vs (6-3)
where: Ml - total column mass at the conclusion of de-gassing (g)
Ms = mass of soil (g)
Mc = apparent weight2 of the empty column (g)
QW = density of the water phase (g/cm3)
Qs = particle density of the soil (g/cm3)
The bulk density of the soil, Qb (in g/cm3), the porosity, n , and the effective length of the soil in the
column, le (in cm), were determined from:
** (6-4)
2. Apparent column weight is the weight of the column with only the main cavity of the column empty.
To measure this weight, all component pieces of the column were weighed separately and summed
to obtain a total weight. The endcaps were weighed when water filled and the core body was
weighed with the sidewall pressure-tap reservoirs filled with fluid.
- 69 -
-------
(6-5)
(6-6)
where: A = internal cross-sectional area of the column (18.55 cm2) . Finally, since the effective length
of the column was a little longer than the distance between the pressure transducers, a pressure
correction coefficient, C, was calculated,
C=^ (6-7)
h
where: /, = distance between pressure transducers (94.75 cm).
This correction factor was used to provide a linear estimate of the pressure drop across the entire length
of the soil pack in the column . The effective pressure drop across the column , AF, , was calculated from
the measured pressure drop, AF, as follows:
AFe = C AF (6-8)
MEASURING ABSOLUTE PERMEABILITY, RELATIVE PERMEABILITY, AND SATURATION
In these experiments, the absolute permeability, k, was defined as the intrinsic permeability
measured under fully water-saturated conditions. Following column de-gassing, the absolute
permeability was determined by injecting water into the horizontal column at a constant rate and
measuring the pressure drop across the column as well as the outflow of water. Darcy's law was used to
solve for the permeability, k. Rearranging Darcy's law gives:
» xl /^ w ^l {c. n \
k=^J- (6-9)
A AF
where: Q = flow rate (in cm3/min)
AF = pressure drop
A/ = /, = length between pressure taps
A = gross cross-sectional area of the column.
For the horizontal column, where the dynamic viscosity of water, fiw , was 0.98 cP; the length, A/, was
94.75 cm; the area, A, was 18.55 cm2; the pressure drop, AF, was measured in psi; and permeability,
k, was in cm2; this becomes:
* (in cm2) = 1.208 x lO'8-^- (6-10)
- 70 -
-------
Because the data collection system was automated, the pressure drop, the flow rate, and hence the
permeability were measured every 15 seconds. The absolute permeability and error reported for each
experimental trial was the mean and standard deviation of these measurements.
Step 1 : Organic Liquid Displacing Water
In this first displacement step, organic liquid displaced water from the soil column reducing the water
saturation to its irreducible value. During this process the relative permeability of both water and organic
liquid were measured as a function of fluid saturations using the unsteady-state method.
Referring again to Figure 6-1 on page 65, water was initially injected into the column (valve A open,
valve B closed) until the desired flow rate had been established (as determined from the flow meter) .
Once the flow rate had stabilized, valve A was closed and valve B opened to allow organic liquid to enter
the column. During organic liquid injection, pressure data and the mass of fluid outflow from the column
were continuously recorded. The pressures of each fluid phase were measured at both ends of the
column. A pressure drop in the organic phase significantly different from the pressure drop in the water
indicated the presence of capillary pressure gradients across the length of the column, violating an
important assumption of the unsteady-state method. The experiments were run at sufficiently high flow
velocities that capillary pressure gradients were negligible. The cumulative mass of the produced fluids
were also continuously measured. Cumulative mass of the outflow, taken together with the flow rate and
the fluid densities, was used to determine the cumulative production of each fluid. Cumulative mass
measurements taken at early time (when only water was being produced) were used to more accurately
quantify the flow rate. The entire contents of the organic liquid reservoir (3 liters or about 4.5 pore
volumes) were injected into the column. Relative permeabilities (km and kro) were calculated using
the unsteady-state method graphical procedure of Jones and Roszelle (1978).
At the conclusion of the initial organic liquid injection, the column was weighed to determine the
average organic liquid saturation, s0, in the column:
- _Ml-M2 6_11
where M2 was the total column mass following organic liquid injection, and &Q (QW - QO) was the density
difference between fluids. This gravimetrically determined organic liquid saturation was compared with
the average organic liquid saturation calculated from the material balance:
— v -V
S0 = "•" °'°"' (6-12)
"p
where: V0iin = the volume of organic liquid injected into the column (cm3)
V0 out = the volume of organic liquid produced from the column (cm3)
In each case Sw = 1 - S0 .
- 71 -
-------
Additional organic liquid was injected into the column until the production of water ceased (and the
column weight stabilized). The organic liquid and water saturations were again determined
gravimetrically and the water was assumed to have reached wetting phase residual saturation, the
so-called 'irreducible saturation', Swi . At Swl the relative permeability to organic liquid, kro , was easily
measured under the steady-state procedures identical to those used for measuring the absolute
permeability, k. In this case:
Qo fj.0 M
'
Step 2: Water Displacing Organic Liquid
In step 2, water displaced organic liquid, reducing the organic liquid to its residual saturation. The
relative permeability of water and organic liquid were measured as a function of fluid saturations during
this step as well. In general, the procedures for this step were the same as the procedures outlined for
step 1. Once again, pressure data and the mass of fluid outflow from the column were continuously
recorded during the injection of water into the column and relative permeabilities were calculated from
these displacement data.
At the completion of waterflooding, the column was weighed to gravimetrically determine the
residual organic liquid saturation, sor '•
A0 Vp
where M3 was the total column mass following the injection of water. At Sor the relative permeability to
water, km , was easily measured under the same steady-state procedures as those used in step 1. In
this case:
A APW k
At the conclusion of step 2, the column contained organic liquid at residual saturation.
Step 3. Mobilization Experiment: Hydraulicallv Reducing the Residual Saturation
In the hydraulic mobilization experiment, the flow rate of water through the column was increased
incrementally. Above a critical flow rate, the residual saturation in the column was reduced as the force
of water flow began to overcome the capillary forces holding the organic liquid in place. For any given
flow rate above the critical rate, the injection of three pore volumes of water was found to be sufficient to
stabilize the organic liquid saturation at its new (reduced) level. After each incremental increase of the
- 72 -
-------
flow rate (after three pore volumes of water had been injected at that particular flow rate), the column
was weighed and the organic liquid saturation was determined using equation (6-14). By measuring the
reduction of residual saturation versus flow rate, the mobilization of trapped organic liquids was
correlated to capillary number. Then the flow rate was reduced to the same low value used in step 3 of
the solubilization experiment. The relative permeability to water at the new residual saturation was
measured under steady-state conditions using equation (6-15).
POSSIBLE SOURCES OF ERROR
Several sources of experimental error were associated with the long column multi-phase flow
experiments. Possible sources of error included:
• incomplete removal of entrapped gas,
• changes in fluid densities due to changes in laboratory temperature,
• capillary end effects,
• lack or loss of filter integrity,
• faulty seals in the system leading to leakage or evaporative losses,
• packing variability (from one soil packing to the next),
• soil property variability (from one collected field sampling to the next),
• error associated with the precision limits of the measuring devices,
• long column length,
• long column horizontal orientation,
• long column wall organic liquid-wetting,
• non-Darcian flow, and
• assumptions in the unsteady state method for the calculation of relative
permeability.
Many of these issues were discussed in Section 5 in the context of the short glass column
experiments. Below is a brief discussion of the last five issues, which are of particular concern to the long
column experiments. We also review capillary end effects.
Capillary End Effects (see discussion in Section 5 and micromodel examples in Section 9)
The outlet reservoir of the long column lacked the oil-wet or water-wet filters employed to minimize
capillary end effects. Consequently we expect that the long column had a capillary end effect on the
downstream end. The micromodel experiments described in Section 9 exhibit this effect, which occurs
- 73 -
-------
when the initial organic liquid flood reaches the outlet reservoir. The effect is minimized at higher flow
rates because viscous forces dominate capillary forces. The long column displacements were run at
high flow rates, a necessary condition for the unsteady-state method of calculating relative permeability.
The end effect should have been limited to a region less than two cm deep into the column at these flow
rates.
Problems Arising from the Length of the Long Column
As we stated earlier, a long column was used in these experiments in an attempt to maximize the
accuracy of pressure gradient measurements. The long column also minimized the impact of capillary
end effects. The cost for these advantages was increased difficulty in packing, longer de-gassing times,
and a lower accuracy gravitimetric saturation measurements. The large weight of the column required
the use of a Mettler PM 11 balance, which has a 11,000 gram capacity, but with only ± 0.1 g accuracy.
Long Column Horizontal Orientation
The saturated zone short column experiments were oriented vertically, which should have resulted
in a uniform saturation profile across each column. The long column experiment was oriented
horizontally, and thus it is possible that the effects of gravity could have caused a (vertical) saturation
profile across these columns. However, the saturation profiles presented in Appendix C demonstrate
that the saturation variation across the five centimeter diameter column should be negligible.
Column Wall Wetting
The long column walls were constructed of PVC. There is some possibility that the organic liquid
phase might preferentially wet PVC, but as we stated earlier, there was no evidence for this in the
experimental results. Residual water and organic liquid saturations were equivalent to those observed
under similar conditions in the short glass columns. (The short TFE columns described in Section 7 did
show effects of preferential organic liquid wetting; see Section 7 and Appendix A.) Future research
efforts should more closely examine this issue.
Non-Darcian Flow
At the higher flow rates in the experiment estimated Reynolds numbers were in the range of one to
ten. It is possible that the flow may have approached the limit of the Darcy regime. Flows at higher rates,
necessary to more fully mobilize the residual non-wetting phase, would certainly have been outside this
regime for the tested interfacial tensions.
Unsteady State Method of Calculating Relative Permeability
The unsteady-state method allowed us to make measurements of relative permeability during the
flooding stages (using the graphical procedure of Jones and Roszelle, 1978). This method assumes
- 74 -
-------
essentially one dimensional, incompressible flow in a water-wet soil. For the conditions of this
experiment it also requires very high flow rates, so that the fluid saturations can be assumed to be
distributed uniformly over the cross-section. It provides a 'first cut' approximation of the relative
permeability function. We decided to take advantage of this rough approach to relative permeabilities,
although the long column experiment was not originally designed for this purpose. The flow rates must
not be high enough to compromise the assumption of low capillary number initial conditions.
LIMITATIONS OF THE TECHNIQUE
The major limitation of this experiment was the limited capacity of the pump. As described in Section
9, we could not achieve the flow rate necessary to fully mobilize residual Soltrol in the Sevilleta soil. The
Llano soil (see Section 4) was then obtained. It is much coarser. For a given pressure (or head) drop the
capillary forces should be weaker in it than in the Sevilleta. Even so, the pump was barely able to mobilize
Soltrol in the Llano. This experience reconfirms the difficulty of hydraulically mobilizing residual organic
liquid saturations in saturated natural soils. The experiments should be re-run in the future with a higher
capacity pump.
Through the addition of a surfactant we could have attempted to mobilize the Soltrol by reducing the
interfacial tension. We could have used an organic liquid with an intrinsically lower interfacial tension.
Finally, we could have replaced the pump with one of higher capacity. We did not pursue the first two
options because of the additional cost in manpower and effort necessary to redesign the experiment and
characterize the new fluids. The cost of a satisfactory pump for the second option was beyond the
resources of this project. All three options should be considered for future work.
- 75 -
-------
SECTION 7
PORE AND BLOB CAST EXPERIMENTAL METHODS
Bench-top short column experiments were performed with polymers and epoxy resins to visualize
non-aqueous phase organic liquid movement and capillary trapping in conditions simulating the
saturated and vadose zones. A 3000 ppm CaCI2 solution and styrene monomer, respectively, modeled
the water and organic fluid phases in the saturated zone experiments. The styrene was then polymerized
and, in some cases, the column was dissected for visual inspection of the organic liquid phase
distribution with a scanning electron microscope. The polymerized organic phase was rigid and
chemically resistant. At residual saturation the polymerized styrene was disconnected into individual
ganglia or blobs. These hardened styrene objects were referred to as 'blob casts'. In other experiments,
the water was replaced with dyed Tra-Bond® 2114 epoxy resin, after the styrene had been polymerized.
After the epoxy hardened, the column was cut into sections and examined under an epifluorescent
optical microscope. These sections were referred to as 'pore casts'. For vadose zone conditions, dyed
styrene and two epoxy liquids were sequentially applied, drained, and hardened in an attempt to simulate
fluid distributions of water, organic liquid, and air. The resulting pore casts were photographed under an
epifluorescent optical microscope.
The technique of using styrene or epoxy to represent wetting or non-wetting fluids in indurated rock
and bead packs has been used previously by petroleum engineers (see, eg., Chatzis et al.1983, 1984,
1988; Chatzis and Morrow, 1982; Yadov et al, 1984; McKeller and Wardlaw, 1988), but so far little work
has been done using this technique in unconsolidated soils. Styrene's immiscible behavior with water, its
low viscosity, and its ability to harden, or polymerize, while in contact with water make it useful for
simulating non-aqueous phase liquid behavior in soils.
This section first addresses the design of the specially fabricated tetrafluoroethylene columns used
in these experiments. Next is a review of the methods selected for characterizing the fluids, especially
the styrene and epoxies, and the soil. The characterization results are summarized. The experimental
procedures for saturated zone conditions are then explained, including comments on pre-processing
the styrene, packing the column, and dying and hardening the styrene and epoxy. This is followed by a
description of the experimental procedures used in the vadose zone simulation experiments.
COLUMN DESIGN
Each of the three columns used to hold the soil sample in the styrene/epoxy resin impregnation
experiments were constructed from a 5 cm long tetrafluoroethylene (TFE) cylinder (5 cm I.D., 6.5 cm
O.D.), and two aluminum-TFE endcaps with 3000 psi valves. A schematic of a typical TFE column is
shown in Figure 7-1. TFE was chosen for the columns because of its non-reactivity, ease of initial
machining and final column dissection, and ability to stand up to the temperatures used to polymerize
the styrene. The TFE walls were organic liquid-wet and this caused some difficulty, as described later.
Both endcaps were made of virgin TFE rod (7.6 cm in diameter, 3.8 cm long) with 0.6 cm thick aluminum
- 76 -
-------
plate screwed to the backs. An external groove was cut in the cylinder into which fit a 10 cm diameter
aluminum split ring. Four threaded rods passed through the aluminum plates and split ring and provided
a means of attaching the endcaps, essentially bolting them on. Four o-rings, one on each end of the
cylinder and one on each of the endcaps, securely sealed the column and created a purely mechanical
method of holding filters in place.
The inner face of the endcaps had radial and concentric grooves machined into them to allow better
fluid flow between the soil sample and the 1/16 inch endcap center hole. In addition, a round fritted glass
disc (50 mm diameter, 5 mm thick) was fitted on top of the groove pattern of the lower endcap as a
support for filters. Nupro® plug valves were threaded through the aluminum plates and sealed against the
TFE rod with Viton® o-rings.
A Magna 66 nylon filter with pore diameters of 0.22 jim, held in place by an o-ring, was used on the
bottom endcap. It acted as a semi-permeable membrane; when water-wet the filter would allow water to
pass but not a non-wetting phase, such as styrene. The nylon filter was covered with a paper filter to
prevent abrasion against the sand. A paper filter alone was used on the upper endcap to prevent fine soil
particles from leaving the column. In experiments which required epoxy drainage, Millipore's 0.2 ^m
Durapore® (polyvinylidene difluoride) filters were used in place of the water-wet nylon filters.
FLUID AND SOIL CHARACTERIZATION
Although styrene's immiscible behavior with water, its low viscosity, and its ability to harden while in
contact with water, make it useful for simulating non-aqueous phase liquid movement through soils, it
does have several drawbacks. For example, it initially has a low viscosity, but once initiated the viscosity
increases over time. This indicates that polymerization begins as soon as initiator is added, albeit
slowly (see Figure 7-2). Styrene's volume also shrinks by about 17% as it hardens (Boyer, 1970). This
must be noted when making optical measurements. However, the interfacial tension of the styrene with
CaCI2 solution was found to remain constant over time (35.3 dynes/cm) even when initiator was added.
Dyes were added to the styrene in order to improve its visibility, but they changed fluid properties.
For example, 9,10-diphenylanthracene (maximum adsorption 260 -qm) was added to styrene (0.6% by
weight as recommended by McKeller and Wardlaw, 1988) causing the styrene to fluoresce blue under
ultraviolet light. Addition of the dye caused the interfacial tension (IFT) with CaCI2 solution to decrease
from oow = 35.3 ± 0.3 to 30.9 ± 3.0 dynes/cm, and also caused viscosity changes, as shown by Figure
7-2. Other dyes tried, such as oil blue N (IFT aow of styrene + dye = 20.2 ± 0.5 ; surface tension ooa of
styrene + dye = 31.8 ± 0.3 dynes/cm) and phthalocyanine blue did not work well enough to justify
further characterization.
Tracon's Tra-Bond® 2114 epoxy was used in several experiments, in addition to styrene, to
represent the wetting or non-wetting phase. Rhodamine B (maximum adsorption 543 t|m), a red-orange
fluorescent dye, was dissolved in benzyl alcohol and added to the Tra-Bond® 2114 epoxy, but no
characterization was done beyond that stated in Table 7-1. The benzyl alcohol caused a dramatic
decrease in the resin's viscosity but only a slight decrease in its surface tension. In the three-phase
- 77 -
-------
FIGURE 7-1. Exploded view of the TFE Short Column.
- 78 -
-------
VISCOSITY OF INITIATED STYRENE VS. TIME
B.U
7.0
6.0
5.0
ABSOLUTE
VISCOSITY 4.0
(cP)
3.0
2.0
1.0
n n
Solid - styrene and benzoyl peroxide
Dashed - styrene, benzoyl peroxide,
and 9,10-diphenylanthracene
_••*'
^^MM^
/'
^^••HMM
-w
— •*
^^
/
7
/
cf»*"r
/
/
0.0 6.0 12.0 18.0 24.0 30.0 36.0 42.0 48.0
ELAPSED TIME - HOURS
FIGURE 7-2. Viscosity of initiated styrene vs. time.
experiments benzyl alcohol (40% by weight) was added to the Tra-Bond ® 2114. This reduced the
viscosity to about 50 cP and decreased the surface tension to 38 ±1 dyne/cm.
Several other epoxy resins were considered but none performed well enough in preliminary tests, to
be used in actual experiments. Shell Chemical's Epon® 8132 epoxy, Dow Chemical's D.E.R ® 324, and
Polysciences' Ultralow® epoxy underwent tests for viscosity, the ability to harden in the presence of
water, compatibility of the liquid epoxy with other hardened epoxies, and dye solubility. All the liquid
resins were soluble with styrene.
Shell's Epon 8132, when mixed with a hardening agent (DETA or diethylenetriamene) in a ratio of
100:13 had a viscosity of approximately 600 centipoises. The viscosity can be decreased with the
addition of benzyl alcohol. To test the resin's resistance to moisture during the curing process, several
drops of the Epon 8132 mixture were covered with water and left to harden. Several of the resin's
components separated, when the drops came into contact with the water, and the resin failed to harden.
Liquid Epon 8132 did not react with any of the other hardened epoxies. Addition of benzyl alcohol to the
resin, as a thinner, however, did result in some surficial tackiness of the other hardened epoxies.
Observations of hardened samples showed that the Epon resin would dissociate over time. Small pock
marks on the surface, filled with an amber fluid, were observed to form over a period of months. This
phenomenon may have been due to water vapor, or other chemical vapors, in the atmosphere reacting
with the epoxy. For this reason no tests of dye solubility were performed.
- 79 -
-------
liquid
aqueous-
phase
styrene
Tra-Bond®
2114
Ultra-low
resin
specific
gravity
1.003 ±0.002
0.906 ±0.002
1.204 (factory)
1.072 ± 0.002
density
(g/cm3)
1.000 ± 0.002
0.903 ±0.002
1.2 (factory)
1.069 ±0.002
kinematic
viscosity
(cst)
0.98 ±0.01
0.89 ±0.01
545 (factory)
18.7 (factory)
dynamic
viscosity
(cP)
0.98 ± 0.01
0.81 ±0.01
597 (factory)
20.0 (factory)
interfacial
tension
with 0.3%
CaCI2 solution
(dynes/cm)
not applicable
35.3±0.3
< 2
not determined
surface
tension
(dynes/cm)
72.0 ± 0.4
31.9 ±0.3
40.9 ±0.5
36.6 ± 0.6
TABLE 7-1. Properties of fluids used in pore and blob cast visualization experiments. All
measurements were taken at 23 °C.
Dow Chemical's D.E.R. 324, when mixed with Henkel's Versamid® 150, had a viscosity comparable
to Shell's Epon 8132, approximately 600 centipoises. Again the viscosity could be decreased using
benzyl alcohol as a solvent. The D.E.R. 324 performed better than the Epon 8132 when tested for
hardening in the presence of water. A thin film of epoxy, in contact with the plastic container, hardened,
but the majority of each drop remained a white viscous gel. Liquid D.E.R. 324 did not react with any of the
other hardened epoxies. The solubility of fluorescent dyes in D.E.R. 324 was low; neither
9,10-diphenylanthracene, rhodamine B, nor coumarin 6 would dissolve to any great extent in the resin.
The resin did take on the dye's color faintly, but large clumps of dye would remain isolated in the liquid.
Interestingly, the undyed D.E.R. 324 resin fluoresced a light blue under ultraviolet light.
Polysciences' Ultralow® is a multi-component resin with a viscosity of 20 centipoises. It does not
harden in the presence of moisture. Due to its low viscosity Ultralow was a prime candidate for use in the
three phase simulation, but it was found that Ultralow reacted with, and softened, hardened styrene. All
three dyes were readily soluble in the resin. During the resin's curing process, however, Ultralow
darkens to a dense amber color. The coloring tended to dominate any dyes dissolved in the resin.
A 3000 ppm CaCI2 solution was used as the aqueous phase in all the saturated zone experiments.
Distilled, de-ionized water was de-gassed by boiling. Enough calcium chloride dihydrate was added to
the cooled water to bring the concentration to 3000 ppm. The solution was stored under a vacuum to
keep it de-gassed.
Measurements of fluid properties such as viscosity, density, surface tension, and interfacial tension
were performed following procedures outlined in Section 4. The Sevilleta sand was used in all of the
successful experiments described in this report. Its properties are also reported in Section 4.
- 80 -
-------
SATURATED ZONE EXPERIMENTAL PROCEDURE
Packing and Degassing
Both homogeneous and heterogeneous sand-pack experiments were performed in the TFE
columns. The packing and degassing procedures for the homogeneous case are reviewed first, followed
by the heterogeneous case. Equations are then presented for calculating pore volume, bulk density, and
bulk porosity, after packing.
Before packing a column with soil to be used in a homogeneous saturated zone experiment, the
lower endcap with filters was tightly bolted to the TFE cylinder, and the cylinder and endcap were
attached to an 18 cm Hg vacuum. Then the column was inverted and placed in a large beaker of CaCI2
solution. Liquid was drawn through the filters, fritted glass disc, and associated plumbing to saturate
them with water (see, eg., Figure 5-2). When air bubbles were no longer observed, the column was
removed from the vacuum and the filters checked visually for integrity. The water-wet filters should have
prevented the flow of the non-wetting air phase. If the filters did not allow air flow the whole column was
dried and weighed. This was called the empty column weight, Me. Mettler PE1600 and PM11 scales, with
1600 ± 0.01 gram and 11000 ± 0.1 gram capacities respectively, were used for the gravimetric
measurements in these experiments.
Next, a buret filled with CaCI2 solution was attached to the bottom endcap and the solution was
allowed to fill the column to a depth of 2 cm. Oven dried Sevilleta sand, which had been weighed
previously, was then poured and packed into the column, taking care to keep the packing surface below
the solution surface. Lab spatulas, bent to a 90 degree angle, were used to pack the soil down. The
process of filling the column with CaCI2 solution and then packing soil was repeated until the solution level
was approximately 2 cm from the top of the cylinder. At that point capillary rise of the solution was used
to saturate the soil/sand as much as possible, since there was nothing to prevent overflow of liquid from
the column. Sand packed above the end of the cylinder was carefully scraped off with a lab spatula and
collected together with spilled sand. After drying, it was combined with unused sand. The mass of soil in
the column, Ms, was simply the difference between the initial soil mass minus any leftover. 0-rings and a
paper filter were placed on the cylinder to seal and prevent the escape of clays and other fines,
respectively, and the upper endcap was lowered onto the column and bolted on.
Some trapped air remained on or between soil particles, especially near the top of the column, and
had to be removed. The column was slowly flooded with degassed CaCI2 solution to solubilize and
remove the trapped air. The solution was introduced through the bottom endcap and out the top endcap.
Periodic gravimetric measurement of the column was used to determine whether or not an equilibrium
had been established. When the column reached equilibrium all the entrapped air had been removed
from the column and the plumbing (see similar procedure for the short column, Section 5). Twenty five
pore volumes ("800 ml) of degassed CaCI2 solution usually solubilized the trapped air and removed it
from the column.
The heterogeneous sand pack column experiments were performed using the Sevilleta sand, split
with a size 50 sieve (» 296 microns) into coarse and fine portions, approximately 45% and 55% by mass,
- 81 -
-------
respectively. The finer fraction was used as a matrix to surround 3 stringers or lenses, composed of the
coarser fraction. The stringers were roughly circular in cross-section. The major difference in packing
technique between the homogeneous and the heterogeneous cases was that the heterogeneous
columns were packed dry. The TFE cylinder and water saturated bottom endcap were assembled as
described earlier, in the homogeneous packing procedures, but no CaCI2 solution was allowed to flow
into the column while it was being packed. This was to preserve the paper forms which were used to
separate the coarse stringers from the fine sand.
The three paper cylinders were constructed with diameters calculated so that the sum of their
cross-sectional areas would equal approximately 45% of the cross-sectional area of the column. This
was done so that any given cross-section would roughly have the same percentage of grain sizes as a
cross-section through a homogeneous column.
A 6 to 7 mm thick layer of the fine sand was placed on the bottom of the column assembly and the
paper cylinders were pushed down into it. This held the forms upright. Then the fine sand was carefully
poured and packed around the forms, so as not to collapse them, until the level of the sand was
three-fourths of the way to the top of the column. At that point the coarse sand was packed, using a lab
spatula, into the paper cylinders. They were filled to approximately 6 to 7mm from the top of the column.
The rest of the matrix was then filled with the fine sand. The three paper forms were slowly pulled out of
the column and any remaining volume was filled with the fine sand.
CaCI2 solution was then pushed upwards, through the bottom endcap, wetting the sand. Settling of
the sand was observed, and more fine sand as added to the top of the column. When the sand-pack was
totally wetted, excess sand was removed from the top of the column and the upper endcap was
attached, as described earlier. Degassed CaCI2 solution was flooded through the column until a
gravimetric equilibrium was attained. These same methods were used for heterogeneous packings in
the quantitative short glass column experiments (see Section 5).
The mass of water in both homogeneous and heterogeneous packed columns was determined from:
Mw = Ml-Me-Ms (7-2)
where: Mw = mass of water in the column (g)
MI = de-gassed, water saturated, sand packed column mass (g)
Me = mass of the empty column (g)
Ms = mass of soil in the column (g)
Estimates for pore volume, soil volume and total effective column volume were also calculated from
these gravimetric measurements:
v = *^ (7-3)
P Qw
- 82 -
-------
(7-4)
VT=VP + VS . (7-5)
where: Vp = pore volume of the column (cm3)
V, = volume of soil sample in the column (cm3)
VT = total volume of the column (cm3)
Qw = density of water (g/cm3)
Qs = particle density (g/cm3)
Finally, the bulk density, Qb , and bulk porosity, n , of the soil pack were determined:
« = 1 - ~~ (5-7)
Qs vce
where: Ms = mass of soil in the column (g)
Qs = particle density of the soil (g/cm3)
The bulk density and porosity formulae represent a bulk or mass average value for the heterogeneous
columns. The actual volumes, densities, and porosities of the fine matrix and coarse stringers were not
individually measured.
Styrene Preprocessing
Styrene monomer is commonly sold containing an inhibitor to prevent polymerization during
transport and storage. Distillation of styrene has been the preferred method to remove the inhibitor, but
the distillation process is quite involved due to styrene's volatility. In order to simplify the laboratory
procedure, inhibitor removal columns developed by Aldrich Chemical Company (cat.# 30,632-0) were
used to remove the 4-tert-butylcatechol inhibitor from the styrene.
Inhibited styrene, held in a separatory funnel, was slowly dripped into the column and then collected
in a beaker or flask at the bottom of the column. Each inhibitor removal column has the capacity to
remove inhibitor, at 15 ppm, from up to 4 liters of styrene. The uninhibited styrene was weighed and
benzoyl peroxide, 1% by weight, was added as an initiator. Benzoyl peroxide was chosen as an initiator
because it was found to preserve the water wetness of the soil sample (Chatzis and Morrow, 1 984) . Dyes
were then added to the styrene.
Styrene Flooding
The initiated, dyed styrene was transferred from the flask to a 100 ml buret which was attached to the
upper endcap of the de-gassed column. To eliminate the possibility of air entering the column, the
- 83 -
-------
Styrene Flood
In
Water Out
FIGURE 7-3. Experimental setup of a styrene flood.
column was iirst inverted and the tubing attached while a slow stream of styrene flowed from the buret.
The column was righted and the styrene elevated to a head of approximately 1 meter above the column.
Valves on the water and styrene burets, and those on the column, were opened and styrene flowed into
the column, as illustrated in Figure 7-3. During the flooding, the head on the column decreased as
styrene left the buret. Head drops of 30 to 40 cm, depending on the column, were common. Water,
displaced by the styrene, passed through the nylon filter and left the column via the lower endcap. The
displaced water was collected in a flask and later discarded.
Experiments with Soltrol-130 have shown that the column-buret system actually required close to 48
hours before residual or irreducible water saturation (Swr or Sm ) was reached. The time dependent
nature of styrene's viscosity, however, required the experiment to be completed within 24 hours of
initiation. Experiments running longer would be of questionable value since the viscosity would be
- 84 -
-------
increasing. Besides, many organic liquids of interest generally have viscosities close to that of water
(see Table 7-2). Attempting to displace a more viscous fluid with a less viscous fluid (i.e., displacing
viscous styrene with less viscous water) can lead to results different from those obtained from a
displacement experiment where the ratio of viscosities is near one. The ratio of the non-wetting fluid
viscosity to the wetting fluid viscosity is often referred to as the 'mobility ratio' in petroleum reservoir
engineering. High, or adverse, mobility ratios can lead to increased trapping of the non-wetting phase in
saturated zone experiments caused by viscous instabilities.
After 18 hours the valves on the burets and column were closed. The outflow tubing and buret were
checked to see that no styrene was produced. Styrene in the outflow line indicated a leak in the filter
which would invalidate the experiment. The upper (styrene) tubing was carefully removed, noting any
spills and cleaning them off the column; the column was inverted and the lower (CaCI2 solution) tubing
was removed. The TFE column was weighed and the mass recorded as M2. The fluid saturations were
calculated as:
MJ-,
(7-8)
V
SW=1-S0 (7-9)
where: S0 = oil saturation (%)
Sw = water saturation (%)
Vp = pore volume (g/cm3)
AQ = density difference between fluids (g/cm3)
The results of these calculations are given in Appendix A.
Water Flooding
Styrene was drained from its buret and the other buret was filled with de-gassed CaCI2 solution. The
column, at initial organic saturation, was re-attached to the two burets, the water buret to the lower
benzene
toluene
m-xylene
gasoline
soltrol-130
kerosene
0.65
0.59
0.62
0.48
1.45
1.73
TABLE 7-2. Absolute viscosities of selected organic liquids at 20° C in Centipoise. Data from
laboratory measurements (section 4) and Weast, 1986.
- 85 -
-------
endcap and the styrene buret to the upper endcap. The CaCI2 solution buret was raised to a head of
approximately 1 meter and all the valves in the system were opened.
CaCI2 solution entered the lower endcap and flowed upward, displacing the less dense styrene.
Connected, or continuous styrene, left the column and no water was produced until approximately
60-70% of a pore volume had passed. For a relatively short period of time after water breakthrough
occurred, both styrene and water were produced from the column. A total of 6-7 pore volumes of CaCI2
solution were flushed through the column in order to ensure equilibrium between the aqueous phase and
the organic residual.
The burets were disconnected from the column, as described above, and the column was weighed
again. This mass was recorded as M3. The residual organic saturation was then calculated as
_
"~
Calculated residual styrene saturations for the Sevilleta sand are reported in Appendix A.
Styrene Polymerization
The styrene residuals were hardened by placing the column in a pressure vessel, pressurizing it to 80
psig and then heating the vessel at 85 degrees centigrade for 40 hours.
The pressure vessel was a 40 cm tall aluminum cylinder (I.D. 10 cm) with 1 .3 cm thick walls. The
bottom was sealed with a flanged aluminum plate welded to the outer wall. The top was sealed with an
o-ring held between the top cap and the face by 10 threaded studs set into the cylinder face. A Nupro®
plug valve, with associated fittings, was threaded into the top cap.
Eight hundred ml of degassed CaCI2 solution were poured into the vessel. The column was placed
into the pressure vessel with the valves on both ends open and enough CaCI2 solution was added to
cover the column with liquid. The liquid acted to prevent nitrogen from entering the sand pack and also
as a diffusion barrier for any oxygen that may be present in the vessel atmosphere.
After the column was sealed into the pressure vessel, a 60 cm Hg vacuum was applied to the head
space to evacuate as much atmosphere as possible, then a dry nitrogen source was connected to the
plug valve and the vessel was pressurized to 80 psig. Checking for leaks in the seal was done by
submerging the vessel under water. If no leaks were detected, the vessel was placed in a laboratory
oven and heated.
Observation of Styrene Residuals
To complete the saturated zone simulation, either the styrene blobs were removed from the sand
matrix or the water/wetting phase was replaced by an epoxy. The first method produced individual
styrene residual 'blob casts', whose shapes and sizes were viewed optically or with a scanning electron
- 86 -
-------
microscope. The second method yielded polished epoxy slabs, the 'pore casts', in which the
relationships between soil grains, the wetting phase, and the non-wetting phase were studied.
The styrene blob casts were removed from the matrix by dissolving the sand grains with acids. A
small portion of the styrene-sand pack was carefully removed from the column with a spatula, placed in a
TFE beaker, and dried. The dry sample was covered with one of several concentrated acids and allowed
to dissolve for several days after which the sample was filtered through a TFE filter, washed with water,
and covered with another acid. This process was repeated with concentrated hydrofluoric acid, sulfuric
acid, nitric acid, hydrochloric acid, phosphoric acid, and chromic acid. This combination of acids
dissolved essentially all of the matrix and left the resistant styrene blob casts, as well as some insoluble
inorganic residue. Several styrene blob casts, as photographed with a scanning electron microscope,
are shown in Section 9 (Figures 9-20 and 9-22).
To construct the pore casts, the column was taken out of the pressure vessel, and the top endcap
was removed. The column and endcap were placed back in the oven at 75-80 degrees centigrade and
allowed to dry for 48 hours. The dried column was reassembled and attached to a pressure vessel filled
with 0.5% by weight rhodamine B dyed Tra-Bond® 2114 epoxy resin. Resin was forced through the soil
sample from the bottom of the column with air pressure. Four or five pore volumes of resin were forced
through the column in order to remove as much air as possible. Time was of the essence, since the
Tra-Bond epoxy would generally harden within 1 hour, depending on the mix of resin to hardener.
Twenty-four hours were allowed for the Tra-Bond epoxy to properly set. The consolidated core was then
removed from the column by cutting longitudinally through the teflon sleeve in two spots, and peeling the
teflon away.
Rock saws owned by the New Mexico Bureau of Mines and Mineral Resources were used to section
the core into 6 or 7 approximately seven rnm thick discs. One face on each disc was covered with a thin
layer of undyed Tra-Bond® 2114 epoxy to fill in holes left by plucked grains, and create a solid surface for
polishing. To remove air bubbles trapped under the epoxy, each disc was placed in a vacuum dessicator
for 5 minutes under a 60 cm Hg vacuum. When the epoxy layer dried, the disc was cut on a Bureau of
Mines trim saw into 9 pieces: 8 edge pieces and one rectangular middle piece. The smaller pieces were
easier to polish and hence provided better optical surfaces. Lap (grinding) wheels were used to grind the
epoxy layer down to a flat surface.Two-hundred-twenty and 400 grit powders were used to remove the
majority of the epoxy coating, and 10 minutes of polishing with 14.5 jo.m and 9.5 jam grit provided the final
surface. Some of these pieces were sent to a commercial firm to be processed into thin sections.
Photomicrography was done using one of the two in-lab Zeiss SR stereoscopes, or New Mexico
Tech's Petroleum Recovery Research Center's Nikon Opti-Phot epifluorescent microscope system.
Examples of these photomicrographs are shown in Figures 9-18, 9-19, and 9-21 of this report.
VADOSE ZONE EXPERIMENTAL PROCEDURES
Attempting to simulate organic liquid transport in the vadose zone using styrene and epoxies
presented several difficulties. In the vadose zone, there are three fluid phases: water, air, and the
organic liquid. In these visualization experiments, all three phases obviously could not be fluid
- 87 -
-------
simultaneously since most epoxies and styrene monomer are miscible. It was decided to simulate the
three fluid phases by a sequence of two phase experiments in which the wetting phase of each
experiment would be hardened. Styrene was used as the first wetting phase, and nitrogen was used to
drain the styrene to a wetting phase residual saturation. Assuming this saturation would not change with
the presence of an intermediate wetting phase, the styrene was hardened. Epoxy was then flooded into
the column and drained with gas to simulate the intermediate wetting phase. The remaining void space
was filled with a second epoxy to simulate the gas phase.
This approach raised a question about surface energies. Would differences between solid-liquid
wetting relationships and the liquid-liquid interfacial tension cause serious differences between the
simulation and reality? The styrene, acting as a wetting phase, was drained to some residual saturation
and hardened. The next intermediate-wetting phase, epoxy, interacted with the immobile 'wetting
phase' as it was flooded through the column and subsequently drained. In order to decrease the surface
energy of the system, the second phase should have spread out in a thin film over the wetting phase
whether the 'wetting phase' is solid or liquid. A liquid at residual wetting saturation would be relatively
immobile as it is trapped by an interaction of capillary, viscous, and buoyancy forces. It may not be as
immobile as a solid, yet simple wetting experiments have shown that a non-wetting liquid will not displace
a wetting liquid from a surface; rather, a thin film of the wetting phase will exist between the solid surface
and the intermediate or non-wetting phase (Morrow, 1974). So in this sense, a solidified wetting phase
may be a reasonable approximation of a fluid wetting phase, as long as flow in the wetting phase is not
significant (Amaufule and Handy, 1982). Granted, there are differences between a liquid propagating
over a solid surface versus a liquid surface, but the final results will be similar: a minimization of surface
energy.
Wetting Phase
Initially, due to its excellent wetting characteristics, styrene was chosen to represent the
wetting/aqueous phase. A column was packed with Sevilleta sand, under dyed initiated styrene,
following procedures described previously in the saturated zone methods. Then it was drained with air.
Polymerization of the styrene failed, however, because of the air in the sand pack. Oxygen in the air
attached to the free radical chain of the styrene, effectively stopping the polymerization process. The
sand pack was left as a gooey, viscous mess.
Polysciences' Ultralow® epoxy, dyed with coumarin 6 (maximum absorption 458 T|m), was tried in an
attempt to bypass the styrene problem. Packed as the wetting phase, it hardened sufficiently to hold the
sand together, but the dye, in the thin layers and pendular rings of the epoxy, did not fluoresce
sufficiently for easy viewing. We decided to try styrene again, in the absence of oxygen.
An I2R (Instruments for Research and Industry) glove bag was purchased, inside of which a column
could be packed in a nitrogen atmosphere. The inert nitrogen atmosphere made it possible for the
styrene chains to polymerize unhindered by oxygen molecules. The glove bag was connected to a tank
of dry nitrogen, inflated, and purged to remove any air. Sand, dyed initiated styrene, packing materials,
the pressure vessel, and the column with a nylon filter were put into the bag and the bag was purged two
more times. A vacuum line with a reservoir outside the bag to collect styrene and tubing attached to a
-------
buret were punched through the side of the bag. With these lines, styrene could be removed from, or
introduced into, the bag,
Using the vacuum line, styrene was pulled through the bottom endcap of the column to remove
atmosphere from behind the nylon filter. Styrene, which had passed from the column to the vacuum
reservoir, was transferred to the buret, which in turn was attached to the lower endcap of the column.
This source of styrene was used to pack the column in a method similar to that described earlier, for the
saturated zone TFE column.
After the packing of the column was completed, a paper filter was placed over the sand and the
upper endcap was bolted on. Approximately five pore volumes of styrene were pushed from the buret
upward through the column and out to the vacuum reservoir. Nitrogen trapped in the sand pack was
removed in this manner, The hydraulic gradient between the buret and the reservoir provided the
impetus for styrene flow, so the vacuum was not used. After the styrene flooding, the column was
considered saturated. The term 'considered saturated' is used because no gravimetric measurements
were made to determine if an equilibrium was attained. Nitrogen may still have been present within the
the sand pack or in the lower endcap, but costs of styrene and dye prevented a more thorough
de-gassing procedure.
Draining the column to a residual wetting phase saturation was done by applying tension to the sand
pack by lowering the styrene buret below the column. Still within the glove bag, the vacuum line was
removed from the upper endcap. Styrene drained through the nylon filter to the buret while nitrogen
entered the column through the upper endcap. Initially, the styrene-air interface was lowered to 50 cm
below the column, and the column allowed to equilibrate. After 1 hour, the buret was lowered another 25
cm and allowed to equilibrate for 2 hours.
The residual styrene was polymerized following the method described in the saturated zone
experiment. A difference in procedure to note is that the column was placed into the pressure vessel
while still within the glovebag, so that the nitrogen atmosphere was maintained. Also, no water was
present within the pressure vesse! in which to submerge the column, only nitrogen gas.
Intermediate Wetting Phase
In most experiments, after the styrene had been polymerized, the TFE column was disassembled
and cleaned. Hardened styrene clogged the fritted glass disc, froze the plug valves, and filled the
grooves in the endcaps. Soaking the valves and endcaps in toluene dissolved the styrene. The fritted
glass disc was discarded.
A new fritted glass disc and new filters were used to reassemble the column. In the first experiment,
a nylon filter was used at the lower end of the column, as in the two phase simulation. It was found that
Tra-Bond® 2114 and Polysciences' Ultralow® epoxy did not wet the nylon sufficiently to prevent air
breakthrough during drainage, so the nylon filter was eliminated from further consideration. Further
experiments and a search through the literature led to the purchase of Millipore's 0.2 jam Durapore
(polyvinylidene difluoride) filter. With a water-wet air entry value of 50 psig, Durapore filters provided
better, although not perfect, epoxy drainage.
- 89 -
-------
Ultralow epoxy dyed with coumarin 6 was used in a first attempt to simulate an intermediate-wetting
phase. The time required to flood and drain a column, 1 to 1.5 hours, seemed to make the choice
obvious. Ultralow had no time constraints and it required heating to cure, while the Tra-Bond hardened
after approximately 1 hour at room temperature. Unfortunately, the Ultralow proved to be incompatible
with the dyed styrene. In preliminary tests, the Ultralow softened the hardened styrene but not to the
point of destruction. While flooding the Ultralow through the sand-styrene pack, however, the epoxy
leached the 9,10-diphenylanthracene dye from the styrene, leading to a homogenization of dyes and an
overall decrease in intensity of both of the dyes. This failure led to an increase of interest in the Tra-Bond
resin.
Tra-Bond® 2114 resin with rhodamine B dye, 0.5% by weight, and benzyl alcohol, 40% by weight,
were pulled under a vacuum into a small stainless steel pressure cylinder. The upper end of the cylinder
was removed from the vacuum and reattached to a source of dry air with a pressure of 40 psig. The lower
end of the pressure cylinder was connected with tubing to the bottom end of the reassembled column,
as shown in Figure 7-4. A short length of tubing led from the upper endcap of the column to a drainage
beaker. The applied air pressure forced resin upwards through the lower endcap displacing air.
Approximately one pore volume of the moderately viscous resin was forced through the column in a 45
minute period.
Draining the column was accomplished by turning off the air source, removing the upper drain
tubing, and attaching a vacuum line to to pressure cylinder. Thirty cm Hg of tension were applied to the
Dry Air
To Drain
FIGURE 7-4. Intermediate-wetting phase flood.
- 90 -
-------
column and this drained approximately 50% of the intermediate-wetting phase volume. The column was
allowed to drain until the excess resin in the drainage beaker showed signs of hardening. This was about
1 hour after the hardener was added to the resin. At that time, the valves on the column were closed,
and the pressure cylinder was disconnected from the system. As the hardening time of an epoxy is
dependent on the volume of the epoxy, the fluid within the column was still semi-liquid when the tension
was removed.
It took 24 hours for the epoxy to cure after which the column was disassembled and cleaned. Benzyl
alcohol was used to flush epoxy from the pressure bottle and to dissolve epoxy from within the endcaps.
Non-Wetting Phase
The third fluid phase to be added to the column represented the non-wetting air phase. Again
Tra-Bond epoxy was used, but this time it was dyed with coumarin 6. Preliminary experiments indicated
that the epoxy resin's liquid phase would not react with its solid phase, so no problems of dye leaching
were anticipated, and none were encountered.
As no drainage of the third phase was required, this step was the easiest of all. The dyed epoxy was
pulled into the stainless steel pressure cylinder, as described earlier, and flooded into the column
through the bottom endcap. After breakthrough occurred, the excess resin was collected in a beaker
and discarded. Forty-eight hours were required for the epoxy to cure.
When the core had cured sufficiently, the teflon cylinder was cut away and the core was prepared for
observation as 'pore casts', as described previously for the saturated zone experiments.
POSSIBLE SOURCES OF ERROR
Several sources of experimental error were associated with the pore and blob cast experiments. All
of those errors listed and discussed in Section 5 apply here. However, recall that with pore and blob
casts we mainly seek qualitative visual pictures of the residual saturation, rather than quantitative
measurements. Possible sources of additional error are:
• TFE Column Weight,
• high styrene density,
• error propagation,
• styrene viscosity changes,
• effects of time constraints on saturated zone experiments,
• TFE column wettability
• styrene shrinkage,
• movement of styrene blobs after hardening,
- 91 -
-------
• breakage of styrene blob casts,
• heterogeneous sand packing,
• fluid properties as a function of the presence of dyes,
• time constraints on wetting phase saturation in the vadose zone experiment,
• incomplete drainage of the intermediate wetting phase saturation in the vadose
zone experiment, and
• finite thickness of pore cast sections.
What follows is a brief review of these issues.
TFE Column Weight
The pore and blob cast experiments produced quantitative results as well as results pertaining to flow
visualization. Unfortunately, the experimental methods made these quantitative results too inaccurate to
be included with the main body of saturated zone residual saturation data presented in Section 9.
Perhaps the largest source of error in the styrene experiments was the mass of the TFE column itself.
The column weighed much more than the glass short-columns used in the quantitative experiments. It
had to be weighed, during the gravimetric determinations of saturation, on the high capacity Mettler PM
11 balance, which has an accuracy of 0.1 grams. The glass short-columns, in contrast, were weighed on
a Mettler PE 1600 balance, with an accuracy of 0.01 grams. The difference in accuracy of the scales
propagated through the error calculations and led to much larger error bars on the data obtained during
the styrene experiments.
Styrene Density
Another factor contributing to the lower quantitative accuracy in these results was the density of the
styrene. In comparison to Soltrol-130, styrene is the denser — 0.90 versus 0.75 g/cm3. The larger
density of styrene led to a smaller value of Ag>, the difference between the density of water and the
organic phase, which was plugged into the equations for organic saturation, S0 , and residual saturation,
Sor . The smaller value of Ap led to greater uncertainty in the saturation measurements.
Error Propagation of the Residual Saturation Measurements
The glass short column experiments residual saturations commonly had errors of ±2-3%, while the
styrene experiments had errors of ±6-8%. When we considered that 6-8% was sometimes up to 50% of
the residual saturation measurement, it seemed prudent to consider these less accurate results
separately. The TFE column residuals are presented in Appendix A.
Styrene Viscosity
Styrene's low viscosity increases over time, once it is initiated (see Figure 7-2). This led us to run
the experiments more quickly than we desired, in order to avoid the adverse effects of a high viscosity
organic liquid.
- 92 -
-------
Effects of Time Constraint
In the saturated zone experiments styrene saturations, S0 , commonly reached values of 70 ± 6%.
In experiments with Soltrol-130 in the short glass columns, S0 values were commonly around 85% (see
Section 9). Clearly the water saturation at this point was greater than its residual wetting value (Swr , =
irreducible water saturation, Swl). Observed residual styrene saturations for the Sevilleta sand were
generally 15 ± 7%. These values are significantly lower than those observed with Soltrol-130 in Sevilleta
sand filled short glass columns (about 27%; see Section 9).
We hypothesized that the time constraint put on the experiment by styrene's ever increasing
viscosity did not allow enough time for the column to come to an equilibrium condition during the styrene
flood. Therefore, the maximum organic saturations were consistently lower than those obtained with the
glass short-columns. This led to consistently lower organic residual saturations after the water flood.
Several experiments were performed to investigate this phenomenon and are also reported in Appendix
A (see Table A-1). In one experiment, trial 0-6, Soltrol-130 was flooded into and drained out of a TFE
column following procedures dictated by the viscosity time constraints of the styrene procedure. The
residual saturation obtained matched those commonly found when performing experiments with styrene
in the TFE column (19%). Another experiment, trial 0-5, was also performed using Soltrol-130 in the TFE
column, but followed procedures described in Section 5 of this report. Since sufficient time for the
system to come to equilibrium was allowed, a value of residual saturation was obtained (27%) which
closely matched those found when using Soltrol-130 in a glass column. These results, which are
presented in detail later, suggest that the differences in residual saturation values were related not to
differences in fluid characteristics, but to the amount of time the liquid/soil system had in which to
equilibrate.
TFE Column Wettability
The wettability of the column walls was also found to affect the amount of time required to reach an
equilibrium during the organic liquid flood. That is, how well an organic liquid wetted the column side wall
affected how the liquid travelled through the column. TFE, which formed the walls of the column in these
styrene experiments, was preferentially wet by the organic phase, whereas the glass-walled columns
used in the short column experiments were preferentially wet by water.
In an attempt to study a non-wetting phase front advance in the TFE columns, an experiment was
conducted in which only one-third of a pore volume of styrene was introduced into a water-saturated TFE
column. The experiment was halted at that point and the column was heated to polymerize the styrene.
When the column was dissected, all the styrene was found around the edges of the sand-pack in contact
with the column walls or endcaps (see Figure 7-5b). The central core of the sand-pack had not yet been
contacted by styrene. This result indicated that because styrene wet TFE, but not the sand, the
advancing styrene followed the walls preferentially (see Figure 7-5a). Subsequent drainage of water
from the center of the column was slowed because the styrene had already achieved a high saturation at
the bottom of the column, interfering with water drainage. Due to the time constraints imposed on the
styrene experiments, the slower drainage of water resulted in lower maximum organic saturations which
contributed to lower residual saturations.
- 93 -
-------
Styrene
a.
b.
water
out
water
FIGURE 7-5. Cross-section of styrene flooding into a water-saturated column with organic wet
walls: a) early time; b) late time.
If this non-uniform displacement was typical for styrene, it is probable that the resulting spatial
distribution of polymerized sytrene blobs may not faithfully simulate the residual organic distributions in
the glass columns and perhaps in the field. This problem is one reason why statistical analyses of blob
size and shape using blob and pore casts may be premature for unconsolidated soils. We suggest the
development of a water-wet column in future blob and pore cast experiments.
Shrinkage of Styrene Volume after Polymerization
Styrene's volume also shrinks by about 17% as it hardens (Boyer, 1970). This must be noted when
making optical measurements of the hardened styrene blobs in the blob or pore casts.
Movement of Styrene During the Construction of Saturated Zone Pore Casts
In the saturated zone pore cast experiments, the water phase was replaced by an epoxy. It is likely
that each hardened styrene blob moved somewhat within the pore space surrounding it. Movement was
restricted by the proximity of the pore walls and, for complex, branching blobs, by their interweaving
within the pore network. There is a very minimal possibility that small hardened blob singlets may have
migrated to an adjacent pore during the epoxy flood. This possibility is inhibited by the large size of the
singlet blobs compared to the available pore throats.
Blob Cast Breakage
Although the styrene blob casts were chemically resistant, they were quite brittle and fragile. The
transfer of a blob from filter paper to a microscope slide for observation often resulted in the breakage of
- 94 -
-------
the blob. It was not clear that the blob casts recovered from the acid bath were wholly unbroken. Some
photomicrographs of broken blobs are presented in Section 9.
Heterogeneous Sand Packs
The heterogeneous sand pack column experiments were performed using the Sevilleta sand split
into two fractions. The geometry of the packing was not well controlled. The properties of the two
fractions was not measured. Future experiments should approach this issue more quantitatively.
Dyes and Fluid Properties
Dyes were added to the styrene in order to improve its visibility, but they changed fluid properties,
Addition of 9,10-diphenylanthracene dye caused styrene's interfacial tension (IFT) with CaCl2 solution to
decrease from oow - 35.3 ± 0.3 to 30.9 ±3.0 dynes/cm and also caused viscosity changes, as shown by
Figure 7-2.
Time Constraints in the Wetting Phase Saturation of the Three Phase Experiments
Styrene was used to represent the wetting phase saturation in the vadose zone experiments.
Because, once initiated, styrene's viscosity changes with time only about five pore volumes of styrene
were pushed through the column. The column was then considered saturated. The term 'considered
saturated' is used because no gravimetric measurements were made to determine if an equilibrium was
attained. Nitrogen may still have been present within the the sand pack or in the lower endcap, but costs
of styrene and dye prevented a more thorough de-gassing procedure.
Incomplete Drainage of Intermediate Wetting Phase in the Three Phase Experiments
The high viscosity of the epoxy prevented complete drainage of the resin in the three phase
experiments.
Finite Thickness of the Pore Cast Sections
Some of the pore casts were thin sectioned. These sections were not two dimensional sections
through the porous media. The styrene, epoxy, and quartz are all translucent to transparent. Fluorescing
materials (eg, styrene blobs) below the surface were visible at the surface. Consequently, blobs seen in
the pore casts are somewhat three-dimensional. Surface staining is a technique used in biology and
geology to insure that only the surface materials are being view. However, both styrene and epoxy take
up the surface dyes, so that it was not possible to use this approach to distinguish the different simulated
fluid phases. We suggest that future work further explore surface staining techniques to overcome this
problem.
LIMITATIONS OF THE TECHNIQUE
We had hoped to conduct statistical analysis of size and shape blob casts. However, two of the
problems that we encountered led us to reconsider. First, the preferential wetting of the TFE column
- 95 -
-------
walls may have led to a final non-uniform distribution of residual saturation over the column
cross-section. Blob populations taken from this sample might not be representative. Second, the
samples of blob casts we examined showed many broken casts (see photomicrographs in Section 9).
These two problems led us to conclude that the experimental procedure was not yet mature enough to
conduct the statistical study as part of this research project. We recommend future efforts in this
direction.
We represented the three phase saturations by a sequential application of fluids. These experiments
assumed that a liquid at residual wetting saturation would be relatively immobile as it is trapped by an
interaction of capillary, viscous, and buoyancy forces. It may not be as immobile as a solid, but a thin film
of the wetting phase should exist between the solid surface and the intermediate or non-wetting phase
(Morrow, 1974). A solidified wetting phase may be a reasonable approximation of a fluid wetting phase,
as long as flow in the wetting phase is not significant (Amaufule and Handy, 1982). Although we feel that
there is a similarity between a liquid propagating over a solid surface and a liquid propagating over a
similar immiscible liquid surface, this experimental approach is unproven. Other experimental
techniques should be developed to study pore scale fluid distributions in the vadose zone. Quick freezing
of very low temperatures is one such technique (see, eg., Gvirtzman ef a/., 1987).
- 96 -
-------
SECTION 8
MICROMODEL EXPERIMENTAL METHODS
Micromodels are physical models of a pore space network, created by etching a pattern onto two
glass plates which are then fused together. The pores have complex three dimensional structure,
although the network is only two dimensional. The advantage of performing multiphase flow experiments
using micromodels is that they give us the ability to actually see fluids displace one another both in a bulk
sense and in individual pores. Displacement photographs of the entire model allow examination of the
bulk displacement processes, while photomicrographs taken through an optical microscope permit
observation of details on a pore level. Etched glass micromodels provide an excellent method with which
to study the mechanisms controlling the transport and capillary trapping of organic liquids because the
structure of the pore network and the wettability of the system can be closely controlled.
Mattax and Kyte introduced the 'capillary micromoder in 1961 as a method to make detailed
observations of fluid interface movements. Their technique allowed them to precisely control the pore
geometry and its variability; factors difficult to control in bead-pack models. They created their models
by first mechanically scribing pore-network patterns on a wax-coated plate, and then by contacting the
exposed glass surface with hydrofluoric acid. Unfortunately, this method relied on the patience and
manual dexterity of the model maker. Davis and Jones (1968), Chatzis (1982), and McKellar and
Wardlaw (1982) improved the manufacturing process by adapting a common photo-etching procedure
to glass. By generating a pore-network pattern by photo-reproduction instead of by mechanical means,
they were able to manufacture large, complex models with pore sizes that approximated those found in
oil reservoir rocks. Their technique, similar to one used for making printed circuits in the electronics
industry, involved 'photographing' the desired patterns on glass plates coated with an
ultraviolet-sensitive resin and etching the plates with hydrofluoric acid. The micromodel construction
procedures described below were modifications of Chatzis's methods as well as those developed by
Eastman Kodak (1975, 1979).
Etched glass micromodels have been used to study a variety of petroleum and chemical engineering
problems. For example, Chatzis and Dullien (1983) used micromodels to investigate capillary trapping
during two phase flow in a pore doublet. Wardlaw (1982), Chatzis ef a/. (1983,1988), and many others
have examined displacement in complex two dimensional pore networks, using water flooding and
various enhanced recovery techniques.
This section first addresses the fabrication of glass micromodels used in these experiments. The
experimental procedures for saturated zone conditions are then explained, including comments on fluid
preparation. This is followed by a description of the experimental procedures used in the three-phase
experiments.
MICROMODEL CONSTRUCTION
The manufacture of micromodels was a difficult process, requiring an enormous investment of time
and effort to learn. However, once the technique was perfected, the whole construction process lasted
- 97 -
-------
only three or four days. A glass mirror, stripped of its protective enamel backing to reveal a copper layer,
was coated with a photosensitive resin. A transparency of a desired pore-network pattern was placed on
the coated mirror surface and was exposed with ultraviolet light. The unexposed resin beneath the
opaque portions of the pattern was removed with xylene. The copper beneath the pattern was removed
with nitric acid, and the glass beneath the copper was etched with hydrofluoric acid (HF). A mirror-image
pattern was etched on another piece of mirror, and the two etched halves were fused together in a muffle
furnace to form the completed micromodel.
Pattern Preparation
Pore-network patterns were created by modifying commercially available drafting pattern films with
drafting pens. A local photographer reduced each pattern to a standard size and transferred it and a
mirror image to plastic transparencies. The emulsion side of each transparency contacted the coated
glass plates; transparencies laid emulsion-side up on a plate allowed too much light to leak under the
pattern during exposure to UV light. The patterns included 'reservoirs' at each end of the network
through which fluids were added and removed in the completed micromodel. An example of a
pore-network pattern is shown in Figure 8-1.
FIGURE 8-1. Pore-network pattern for the homogeneous model.
- 98 -
-------
enamel
copper
silver
glass
FIGURE 8-2. Mirror construction.
Mirror Preparation
Ordinary mirrors are manufactured by coating a piece of glass first with a silver layer, then a copper
layer, and finally a protective enamel backing (see Figure 8-2). Mirrors were used in micromodel
construction as a matter of convenience: the copper layer provided a binding surface for the
photosensitive resin described in the next portion of this section, and the enamel backing served to
protect the copper during transport and storage. A less convenient alternative to mirror glass would have
been to coat plain glass with copper in a high-vacuum chamber by evaporation.
A 5 X 8 inch (12.7 X 20.3 cm) piece of mirror glass was prepared for coating with resin by first
placing it, enamel side up, in a hot 50% by weight solution of NaOH to remove the protective backing. The
solution was kept as hot as possible without actually boiling (approximately 90° C). During the next 5-10
minutes, the integrity of the enamel was tested by gentle scraping with teflon tongs. When the backing
scratched easily, the mirror was taken from solution and the enamel was removed from the plate by
gentle rubbing with a Viton-gloved hand under a stream of hot tap water. If the plate was left for greater
times in solution, the backing slid off easily with no rubbing necessary; however, for these longer soaking
times, the NaOH slightly corroded the copper beneath the enamel layer. If any enamel was left on the
plate after the rinse, the affected portion was reinserted into the solution for a short time and then
re-rinsed. Some experimentation was necessary to find a brand of mirror with enamel (Willard mirror
glass, for instance) that could be removed completely and easily. After the backing was removed, the
plate was rinsed with distilled water and dried in an 80° C oven (Figure 8-3).
Pattern Exposure and Development
Kodak Thin Film Resist (KTFR), an ultraviolet-sensitive resin, was used to transfer the pore-network
pattern to the mirror surface. In a darkened room, 1 part KTFR by volume to 2 parts xylenes were mixed.
A mirror plate stripped of its protective backing was held horizontally, copper side up, and coated with
- 99 -
-------
copper
silver
glass
FIGURE 8-3. Mirror with enamel removed to reveal copper surface.
approximately 10 ml of resist mixture. The plate was tipped in various directions to evenly distribute the
resin over the copper surface in a layer of uniform thickness (see Figure 8-4). The plate was tilted
vertically and allowed to air dry until the coating was no longer sticky to the touch (generally 20-30
minutes). If long HF etching times were expected, the plate was baked in an 80° Coven for 10 minutes to
help the resin adhere to the copper surface. The disadvantage of baking was that the resin was difficult to
remove in later steps. Unused coated plates were stored in a dark place.
After the coating was dry, the patterned transparency was placed emulsion-side down on the coated
mirror surface, covered with a clear piece of glass to ensure good contact between the pattern and the
surface, and placed under a 1600 microwatt per centimeter long-wave ultraviolet light source at a
distance of 24.5 cm. The assembly was exposed to the UV source for approximately 12 minutes as
illustrated in Figure 8-5.
Exposure times were found to be a function of the thickness of the resin coating, the intensity of the
light source, and the distance of the light source from the plate. A thick coating (an undiluted KTFR
photosensitive resin
copper
silver
glass
FIGURE 8-4. Copper surface coated with Kodak Thin Film Resist (KTFR).
- 100 -
-------
ultraviolet light source
clear piece of glass
patterned
transparency
KTFR coating
FIGURE 8-5. Pore-network pattern exposed with UV light onto coated copper surface.
mixture) better protected the non-pore areas from HF during the etching step but required greater
exposure times than for a thin coating. However, thin coatings made by diluting KTFR with xylene
reproduced fine details more faithfully and required smaller exposure times than for thick resist layers. It
was also found that exposure time decreased with a shortened distance between the light source and the
model, and with increased intensity of the light source. Exposure times were also affected by the age of
the resin: the older the bottle of KTFR, the longer the exposure times needed (it requires months of aging
on shelf to increase exposure times).
When the exposure was complete and while the room lights were still dim, the plate was removed
from under the UV light and the surface was sprayed with xylene. The plate was tipped back and forth for
about 1 minute to wash away the undeveloped resist representing the pore-network pattern (see
Figure 8-6). The plate was rinsed with warm tap water, then distilled water, after which the normal room
lighting was restored. If the pattern was not visible, more xylene was applied, and the plate was rinsed
again. The plate was shaken to remove excess water droplets and was placed in an 80° C oven for 10
minutes. The plate was removed and cooled before the next step.
Etching Copper
The cooled model was placed in a 50% by weight solution of HN03 for approximately 10 seconds, or
until the copper and silver layers unprotected by resist (the pore-network pattern) had dissolved to
reveal the underlying glass surface, as illustrated in Figure 8-7. The plate was rinsed quickly with cold tap
water and then with distilled water. After the plate had been dried in an 80° C oven, the pattern was
examined under a microscope for imperfections. Small undissolved portions of the network were
- 101 -
-------
resist removed to reveal copper layer
exposed resin
copper
FIGURE 8-6. Pore-network pattern exposed on the resin coating.
removed carefully with a dental tool or scriber. If necessary, the plate was re-dipped in the HN03, then
re-rinsed, to remove copper and silver left in the network after the first acid dip.
Etching Pattern in Glass
All areas of glass that were to remain unetched, such as the model edges and back, were coated
with excess resist mixture and allowed to dry. The model was placed pattern-side up in a tray of
concentrated HF for about 15 minutes. Longer etching times were used for models requiring deeper
pores. When the model was removed from solution, it was promptly rinsed in cold water, and the network
was scrubbed with a wire brush to remove siliceous deposits formed during etching. The resist was
removed with a razor blade, the copper and silver with HN03, and the model was washed with detergent,
rinsed with distilled water, and allowed to dry.
Model Assembly
A mirror-image micromodel half was produced by the above methods using a mirror-image
transparency. Inlet and outlet ports were drilled with a diamond drill bit in the reservoir areas of one of the
copper and silver layers removed
to reveal glass beneath network
exposed resin
copper
FIGURE 8-7. Copper and silvers layers under the pore network pattern removed to reveal the
underlying glass plate.
- 102 -
-------
FIGURE 8-8. SEM photomicrograph of the cross-section through a
typical pore within a micromodel.
plates. The two halves were aligned under a microscope, and cyanoacrylate glue was wicked in between
the plates from the edges to temporarily hold them together. The model was placed in a muffle furnace
and fused at 720° C for 15 minutes. Longer fusing times resulted in smoother, smaller pores; however, if
a model was left too long in the furnace, some of the pores closed and the network became
disconnected.
A completed micromodel has pores that are eye-shaped (see Figure 8-8). During multi-phase
displacements, the most wetting phase tends to preferentially fill the wedges on either side of the pore.
We believe that this mimics behavior in natural soils, in which the wetting phase tends to remain as
pendular rings at grain-to-grain contacts.
MICROMODEL EXPERIMENTAL PROCEDURE
In the two-phase micromodel experiments, as in the two-phase short column experiments, an initially
water-saturated and degassed micromodel was flooded with Soltrol at a prescribed rate to simulate the
movement of an organic liquid into the saturated zone. After the fluid saturations stabilized, the model
was flooded with water at low velocity. The organic liquid still in the model after water injection remained
as residual saturation 'blobs': immobile and disconnected pockets of organic liquid which have been
trapped by capillary forces. The two-phase experiments represented a scenario in which an organic
liquid percolated into the saturated zone, then was displaced by ambient groundwater flow, and finally
was left behind as residual organic liquid saturation.
- 103 -
-------
FIGURE 8-9. Photograph of network pattern showing the capillary barrier
built into one end of a micromodel.
In the three-phase experiments, an initially water-saturated micromodel was drained with air under
an applied suction. The magnitude of the applied suction determined the water saturation remaining in
the model. Organic liquid was then injected into the column under low-flow conditions, simulating the
infiltration of organic pollutants into the vadose zone. After equilibrium conditions were reached, the
organic liquid was again drained with air. The three-phase experiments represented a scenario where an
organic liquid percolated through the vadose zone to the water table, leaving behind trapped organic
liquid.
During flooding or drainage, a capillary end effect was sometimes observed at the bottom of the
model (see Figure 9-2a, for example). A cure to this problem was developed rather late in this study. The
pore-network pattern was altered so that it included a series of small pores in one of the end reservoirs,
as illustrated in Figure 8-9. These small pores served as a capillary barrier, a function equivalent to that
of the nylon filters described in Sections 5 and 7. Only some of the three-phase experiments were run
with these capillary barriers in place.
Fluid Preparation
The fluids used in the micromodel experiments were air, water, and Soltrol-130. The aqueous phase
was prepared by combining 10 milliliters of blue food color with one liter of distilled water. The water was
degassed. The organic phase was prepared by combining O.SO grams Oil Red 0 with one liter of
Soltrol-130, and then by straining the mixture through a coarse paper filter. The micromodels were
- 104 -
-------
cleaned with chromic acid, thoroughly rinsed with distilled water, and were saturated with the aqueous
phase prior to the experiments.
Two-Phase Experimental Procedure
The micromodels were initially saturated with water by connecting each one to a circulating pump
and a reservoir of the aqueous phase as shown in Figure 8-1 Oa. Water was then flushed through the
model until all entrapped air was removed. Then, Soltrol was injected at a prescribed rate into each
model with a Sage Instruments model 351 syringe pump. After the fluid saturations stabilized, the
syringe pump flow direction was reversed and the model was flooded with water. Capillary forces
trapped the remaining organic liquid as immobile, disconnected blobs.
Micromodels were oriented either vertically or horizontally during the two-phase experiments. For
vertically-positioned micromodel experiments, the syringe pump was connected to the top fitting of the
model and was flooded with Soltrol from the top to the bottom (Figure 8-1 Ob). In the horizontally-oriented
experiments, the syringe pump was arbitrarily connected to the left fitting, and Soltrol was flooded into
the model from left to right (Figure 8-1 Oc). The pore networks shown in Figure 8-1 and in Figures 8-12
and 8-13 were run vertically. The network in Figure 8-13 was run in both the horizontal and vertical
positions. Micromodels based on other patterns were prepared, but results from those experiments are
not presented in this report (see, eg, Conrad et a/., 1989).
Three-Phase Experimental Procedure
As in the two-phase experiments, the micromodels in the three-phase experiments were de-aired
with a circulating pump. With the model positioned vertically, the micromodel was drained with air under
a suction applied by the syringe pump, which had been connected to the bottom fitting (Figure 8-11 a).
After the syringe pump had been removed and re-attached to the model's top fitting, organic liquid was
micromodel
circulating pump
syringe pump
micromodel -
water
micromodel
(a) (b) (c)
FIGURE 8-10. Two-phase micromodel experimental set-up.
- 105 -
-------
air in
micromodel
flow
syringe pump
micromodel
(a) (b)
FIGURE 8-11. Three-phase micromodel experimental set-up.
introduced into the top of the model to simulate the infiltration of an organic liquid into the unsaturated
zone (Figure 8-11b). After equilibrium conditions were reached, that is, after the fluid saturations
stabilized, the syringe pump was re-attached to the bottom fitting and the micromodel was once again
drained with air.
Although we had a desire to control fluid pressures in each of the phases during this experiment, we
were not able to develop a cost effective apparatus to do so. We attempted to place various wetted
porous membranes in the model to control the pressures and eventually perfected this technique using
narrowly etched channels. This improvement came late in the project and we will exploit it in future work.
Glass bead pack micromodels, glued together with epoxy, would present no such problem, but it is more
difficult to make visual observations in them than in the etched glass micromodels. There is also the
issue of the competing wetting properties of the glass and the epoxy that holds the beads together.
LIMITATIONS OF THE TECHNIQUE
Micromodels are a flow visualization technique. They are not designed to provide quantitative
information. We were unable to measure pressures or saturations, or to control either one. Keep this
caveat in mind when reviewing the photographs and videotapes of the micromodel experiments.
It is possible to evolve the experimental procedure to allow semi-quantitative data to be collected.
Improved end reservoirs, incorporating oil- or water-wet etched channel porous membranes, provide a
mechanism for controlling capillary end effects. Pressures can be controlled and measured through
these reservoirs. Etched pore cross-sections can be estimated using measurescopes, epoxy casts, and
other techniques. Saturations can be estimated using a high resolution video camera, a frame grabber,
and image processing techniques. These quantitative techniques could be combined with mathematical
models at the pore and network level in order to help explain experimental observations and validate
theoretical models.
- 106 -
-------
FIGURE 8-12. Pore-network pattern for the 'aggregated' model.
FIGURE 8-13. Pore-network pattern for the heterogeneous 'coarse lens' model.
- 107 -
-------
SECTION 9
SATURATED ZONE RESULTS AND DISCUSSION
Figure 9-1 depicts the portion of the aquifer that includes the organic liquid residual saturation in the
saturated zone. For an organic liquid more dense than water (left below), an incoming slug of organic
liquid leaves behind a trail of capillary trapped residual as it makes its way downward toward the bottom
barrier, and then laterally along that barrier. Within and below the capillary fringe the trapped residual
organic shares the pore space only with water. In the saturated zone there is no air phase. The residual
organic liquid in the saturated zone is the subject of this section of the report.
Saturated zone residual is also found when an organic liquid less dense than water, a so-called
'floater', reaches the capillary fringe and water table (right below). The weight of the organic liquid
depresses the capillary fringe and water table, and then redistributes laterally. The water table rebounds,
3^
®
^L
^
ground surface
•A-
WIDOSE
ZONE
floating organic
liquid
residual
organic
liquid
saturation,1
in the
saturated
zone
SATURATED
ZONE
flow
"3SS dense organic liquid
imm mnn nnm nnm nnm mini mmi nnm mnn m..
m mini mini imm mini mini mini imm mnn mnn mini nnm mnn mini mini mnn mini mnn mini
mini limn mnn mini mini mint mini nnm nnm mini nnm mini nnm imm mnn
n mini mini mnn nnm mini nnm mini mnn nnm nnm nnm mini nnm mnn nn
mm nnm mnn mini nnm >"""': imm im :::.•:
n mnn mini n :; itnn mini nnm nnm nnm mini nnm mini mini mini nnm mini mini mi
N<<" :..,
-------
and some organic liquids are left behind, within the saturated zone that is defined by the capillary fringe
and below. A third possibility for the capillary trapping of residual organic liquids in the saturated zone is
also associated with a 'floater'. As the water table rises in response to recharge or other hydrologic
controls, some of the 'floater' is trapped and remains behind.
This section on organic liquid movement and capillary trapping in the saturated zone contains eight
main parts:
•1 review of basic concepts, capillary trapping mechanisms, petroleum
experience, and mobilization issues,
•2 flow visualization of displacement and capillary trapping in a micromodel,
•3 capillary trapping and residual saturation in an unconsolidated soil: the
Sevilleta sand,
•4 residual saturations for various organic liquids,
•5 residual saturations for various soils,
•e influence of the rate of initial organic liquid invasion on irreducible water
saturation and organic residual saturation,
•7 influence of the rate water flow rate on residual organic liquid mobilization, and
•a residual saturations in heterogeneous soils.
The first part is a brief review of background information, including a brief discussion of published
experiments on sandstones and glass beads. Next is the description of displacement and capillary
trapping in a homogeneous micromodel. It visually illustrates many of the issues that are raised in the
review. Third, we present the results of our study of the Sevilleta sand, including quantitative
measurements of Soltrol residual and photomicrographs of blob and pore casts. We examine the
hypothesis that the Sevilleta sand should exhibit a behavior more like that of glass beads than that of
sandstone. In the fourth part we test the hypothesis that residual saturation is largely independent of
organic liquid composition by making measurements for a variety of single and multi-component organic
liquids. Next, we examine residual saturations in a variety of soils and test the hypothesis that residual
saturations should be similar in soils that have a similar grain size distribution. Sixth, we investigate how
the rate of initial invasion of a non-wetting organic liquid may influence irreducible water saturations and,
later, organic residual saturations. Seventh, we study the possible hydraulic mobilization of residual
organic liquid by increasing groundwater velocities. We validate Wilson and Conrad's (1984) conclusion
that this is largely an unrealistic aquifer remediation alternative, unless interracial tensions are reduced
significantly. Eighth and last, we investigate the hypothesis that porous media heterogeneity can
dominate displacement and trapping mechanisms. We present flow visualization results from
micromodels representing 1) an aggregated soil (strings of interconnected macropores separating
clumps of micropores), and 2) a soil with discontinuous lenses (lenses of coarse pores within an
otherwise homogeneous model). The character of these displacements are contrasted with
homogeneous displacements. The discontinuous lenses experiments were duplicated in short column
studies, yielding both pore casts of the trapping in a heterogeneously packed sand column and
- 109 -
-------
quantitative measurements of increased trapping. The micromodel and column results are supported by
a new theoretical model. Based on the interplay between viscous and capillary forces, this model
explains why the trapping is strongly dependent on fluid flow rate when there are discontinuous lenses or
related heterogeneities.
The discussion of our results includes special paragraphs that address implications for various
aspects of groundwater contamination characterization and remediation. Not surprisingly, these
paragraphs are labeled 'Implications for ,,,-'.
REVIEW OF CAPILLARY TRAPPING PHENOMENA IN POROUS MEDIA
Capillary trapping of oil during displacement by water has been studied for years in the context of
petroleum recovery from oil reservoirs. A very few examples of this large literature are: Anderson, 1988;
Chatzis et a/., 1984, 1988; Craig, 1971; Hornof and Morrow, 1988; Melrose and Brandner, 1974;
Mohanty et a/., 1980; Moore and Slobod, 1956; Morrow, 1979;, 1894; Morrow and Songkran, 1981;
Morrow and Chatzis, 1982; Ng ef a/., 1978; Pathak et a/., 1982; Salathiel, 1973; Taber, 1969,1981; and
Yadav et a/., 1987. These authors are primarily concerned with the mechanisms of oil trapping during a
waterflood. Capillary trapping of residual oil leads to a reduction of economic oil recovery. Some
enhanced oil recovery techniques are largely aimed at reducing the amount of capillary trapped residual
by changing miscibility (eg, a C02flood) or interfacial tensions (eg, a surfactant flood). In groundwater
hydrology we too are concerned with the capillary trapping of residual saturation and with its removal.
The review given below focuses on these two issues. However, unlike petroleum engineers, we are also
concerned with the mechanisms that initially brought the 'oil' into the aquifer in the first place. In the 'oil
business' that is the province of petroleum geologists, and it involves issues that are quite different than
ours. Consequently, we can expect little help from petroleum reservoir engineers on these mechanisms.
There are three major forces acting in both oil recovery and organic liquid behavior in groundwaters:
capillary forces, viscous forces, and gravity or buoyancy forces. Capillarity is the result of the interplay
of cohesive forces within each fluid phase and the adhesive forces between the solid phase and each of
the fluids. The capillary force is proportional to the interfacial tension at the fluid-fluid interface and the
strength of fluid wetting to the solid surface, and inversely proportional to the pore size. Viscous or
dynamic forces are proportional to the permeability and to the pressure gradient, while buoyancy is a
gravitational force proportional to the density difference between the fluids. For multiple fluid phases in
an aquifer at typical aquifer flow rates, capillary forces often dominate over viscous and buoyancy
forces. As we shall see, the dominance of capillarity explains the trapping of residual organic liquid.
The review starts by assuming that there is an oil or organic liquid saturated porous medium (with
some residual water saturation), with the organic liquid being displaced by water. As the water flood
progresses through the medium some of the organic liquid is trapped by capillary forces and remains
behind as residual. We first review the fundamental concepts of interfacial tension, wettability, and
capillarity. Then we examine the two major mechanisms for capillary trapping when two fluid phases are
present: snap-off and by-passing (Mohanty et a/., 1980; Chatzis era/., 1983; Wilson and Conrad, 1984).
- 110 -
-------
We then review some of the published experimental measurements for residual oil saturations as
measured for sandstones and glass beads by petroleum researchers. Finally, we review the relative role
of viscous, gravity, and capillary forces in the mobilization of residual.
Fundamentals
Interfacial Tension (after Adamson, 1982) —
In the interior of a homogeneous fluid, a molecule is surrounded on all sides by other like molecules
exerting cohesive forces between one another. At the interface between two immiscible fluids however,
there are few if any like molecules across the interface. A molecule at the interface is attracted to
molecules of its own phase by a force greater than the force attracting it to molecules of the 'immiscible'
phase across the interface. The cohesive forces acting on a molecule inside a fluid phase, and on a
molecule at the interface (liquid-gas or liquid-liquid) between fluid phases are illustrated in Figure 9-2a.
This unbalanced force draws molecules along the interface inward and results in the tendency for the
fluid-fluid interface to contract. If the interface is stretched, it acts like an elastic membrane. The
restoring force seeking to minimize the interfacial area between the two immiscible fluids, is called the
interfacial tension, a . When encountered between a liquid and a gaseous phase, this same force is
called the surface tension, y.
Wettability —
Figure 9-2b illustrates a possible configuration of two immiscible fluid phases in contact with the solid
phase (walls) of a cylindrical tube. At the line of contact where the two fluid phases meet the solid phase,
both the cohesive forces within the fluids and the adhesive forces between the solid and each of the
fluids are at work. Suppose that one of these fluid phases is water (fluid 1), and the other is an organic
liquid (fluid 2). If the adhesive forces between the solid and the water phase are greater than the
cohesive forces inside the water itself and greater than the forces of attraction between the organic
phase and the solid, then the solid-water contact angle, 0 , will be acute and the water will 'wet' the solid
(Hillel, 1980, p.44). The contact angle provides the only direct measurement of wettability. As an
example of a contact angle, liquid water forms an acute contact angle 0 of about 25.5° on clean glass in
the presence of air. Water is a strong wetting fluid relative to air for this surface, but not necessarily for
other surfaces, such as TFE (tetrafluoroethylene). Further discussion of wettability can be found in
Section 4 of this report.
Capillarity —
As a result of the contact angle, a meniscus is formed between the fluid phases (Figure 9-2b); the
narrower the tube, the smaller the radius of curvature. Similar to the curvature produced by a pressure
difference across a membrane, the presence of curvature implies a pressure difference across the
fluid-fluid interface, called the capillary pressure:
2(7 2o cosO
PC = Pnw - Pw = a C = — = (9-1)
rc r,
- Ill -
-------
immiscible fluid 2
;„ ? '>V"""»'*".' - , ,- > "%f!s,
t s 9 f fjff ; ' / ^ ff s
f s s 'sssfttt ff, f f f * f f^Vf/f f& t
f ?, sSffjrS fVA f || |i/H 1 •"•• $$$$$£ff ffff ^flfiwSf ,Af,
•f f tfsff f f*t I IUIVJ I f SVfftff f fjff fVKVffjVf^f
,,„ , ,,*% 71
'c , icontact^. 6?
fluid 1K;
fluid-fluid interface
radius of
curvature
immiscible fluid 2
solid
FIGURE 9-2. Two sketches illustrating fundamentals: a) cohesive forces acting on a molecule
inside a fluid and at its interface with another, immiscible fluid (after Hillel, 1980);
and b) hydrostatic equilibrium of two fluid phases in contact with a solid phase
(after Melrose and Brandner, 1974).
- 112 -
-------
where: Pc = capillary pressure
Pw = pressure in the wetting phase (fluid 1 in the figure)
Pnw = pressure in the non-wetting phase (fluid 2 in the figure)
a = fluid-fluid interfacial tension
C = curvature of the fluid-fluid interface
rc = radius of curvature
0 = contact angle, measured through the wetting fluid
rt = radius of capillary tube
The most common example given to illustrate capillarity is that of the capillary rise of water in a
straight thin tube, or straw (see Figure 9-3). The pressure difference between the wetting and
non-wetting fluid at the interface inside the tube causes the water to rise into the tube, above the level of
the free surface.
These same forces operate on the pore scale in saturated, and unsaturated, porous media. The
capillary fringe is a notable example. Also, the interplay of the various wettabilities and interfacial
tensions of different fluids, and the capillary forces they give rise to, leads to trapping of fluids within
pores, as the fluids migrate through, or drain out of, a system. These capillarity induced trapping
mechanisms are discussed below.
Capillary Trapping Mechanisms
When two fluid phases are present, and the non-wetting fluid is being displaced by the wetting fluid,
there are two major mechanisms for capillary trapping: snap-off and by-passing.
air
the non-wetting fluid
Pa
free surface >
^ t t St ' % f •
Pa
fen
,0i ,,,..
' :
' ' '
P - P
•* nw — *• a
P
-------
Snap-off —
Snap-off occurs as non-wetting fluid in a pore is displaced from a pore body into a pore throat.
The mechanism strongly depends on wettability and the aspect ratio — the ratio of pore-body diameter to
pore-throat diameter (Wardlaw, 1982).
Consider the case of water displacing an organic liquid from a tube with a non-uniform pore
diameter, as shown in Figures 9-4 and 9-5. The walls of the tube are smooth and strongly water wet. The
water contact angle is acute, the water-organic liquid interface is curved, and the water phase 'wicks'
along the pore wall. In high aspect ratio pores, the pore throats are much smaller than the pore bodies
(Figure 9-4a and Figure 9-5). When the thin layer of water phase reaches the exit pore throat, a large
blob of organic liquid still remains in the pore (Figure 9-5). Snap-off occurs as the water continues
through the exit throat leaving behind the now disconnected singlet blob. In a sequence of high aspect
ratio pores, a singlet blob is trapped by snap-off in each pore (Figure 9-4a). Pores in a pack of uniform
sized glass beads have a high aspect ratio, explaining the prevalence for singlet blobs observed by
Morrow and Chatzis (1982) and Chatzis et al. (1988). For low aspect ratio pores, in which the pore
throats are almost as large as the pore bodies, the organic fluid can be completely displaced, as shown
in Figure 9-4b.
a. high aspect ratio pores (snap-off):
pore
body
pore
body
pore
body
pore
body
wetting
fluid
non-wetting
fluid
pore
throat
pore
throat
pore
throat
b. low aspect ratio pores (no snap-off):
wetting
fluid
non-wetting
fluid
FIGURE 9-4. Effect of pore aspect ratio on organic liquid trapping in a tube of non-uniform
diameter (after Chatzis et al., 1983).
- 114 -
-------
0 = 10°
pore throat
flow
wetting fluid
flow
non-wetting fluid
pore body
FIGURE 9-5. Wetting fluid displacing a non-wetting fluid from a circular, high aspect ratio p<
under strongly wet conditions (after Wardlaw, 1982).
Trapping is a function of wetting and contact angle as well as pore geometry. The combined
effect of contact angle and pore geometry control the curvature of a fluid-fluid interface and determine
the potential for snap-off (Wardlaw, 1982). Figure 9-6 depicts an interface with a 90 degree contact
angle passing from a pore throat through a high aspect ratio pore body with smooth walls. The
intermediate contact angle of 90° causes the curvature of the interface to remain relatively small. As the
interface reaches the exit throat, little organic phase remains in the pore and no trapping occurs
(Wardlaw, 1982). In rough walled pores, there is probably some trapping of the retreating phase in the
0 = 90°
pore throat
pore throat
flow
FIGURE 9-6.
advancing fluid
flow
retreating fluid
pore body
One fluid displacing another from a circular, high aspect ratio pore, under
intermediate wetting conditions (after Wardlaw, 1982).
- 115 -
-------
asperities along the wall (see Figure 9-7). Photomicrographs of the Sevilleta sand (Figure 4-6) and its
pore casts (shown later in this section) indicate that this material had smooth walls. In any event all of the
micromodels, and the blob and pore cast experiments, described in this report were run under strongly
water wet conditions.
By-passing —
The pore doublet model has been used to describe trapping mechanisms on a microscopic
scale (Rose and Witherspoon, 1956; Moore and Slobod, 1956; Chatzis and Dullien, 1983; and Chatzis et
al., 1983). A pore doublet consists of a tube which splits into two pores, one generally narrower than the
other, and then rejoins. The pore doublet concept is used here to describe organic phase trapping by the
mechanism referred to as 'by-passing'.
Figure 9-8 depicts a wetting phase (water) displacing a non-wetting organic phase in a pore
doublet under several different circumstances. The pore walls are smooth and strongly water wet. Figure
9-8a demonstrates circumstances under which no trapping occurs. The advancing water phase enters
the narrower pore opening first (stage 1). In each pore, the total pressure drop driving flow is the sum of
the capillary pressure and the dynamic pressure drop caused by flow (Moore and Slobod, 1956). For the
pore doublet to have any physical meaning, the flow rates (and dynamic pressure drop) should be low
enough to approximate aquifer conditions. On a pore scale, under such conditions, capillary forces are
much larger than the dynamic viscous forces. Capillarity then controls advance of the wetting front
causing water to fill the narrower pore (Chatzis and Dullien, 1983). The water-organic interface remains
stable at the entrance to the wider pore (stage 2). When the water reaches the downstream node (where
the pores rejoin), it forms a stable meniscus with the organic liquid because the cross-section at the
downstream node is greater than at the entrance to the wider pore. In instances where a stable meniscus
can be maintained at the downstream node, water can then push organic out of the wider pore (stage 3).
pore throat
pore throat
advancing fluid
flow
pore body
retreating fluid
trapped in asperities
FIGURE 9-7.
Final condition after an advancing fluid displaced a retreating fluid from a
rough-walled pore under intermediate wetting conditions (after Wardlaw, 1982).
- 116 -
-------
The menisci rejoin at the downstream node (stage 4). In this case, the water has displaced the organic
liquid completely from the pore doublet and no trapping has occurred.
The displacement sequence for the pore doublet shown in Figure 9-8b illustrates the
by-passing mechanism of trapping. As before, the water enters the narrower pore first. However, as
water reaches the downstream node, it does not stop because no stable interface is formed (Figure
9-8b, stage 2). The organic liquid in the wider pore has become disconnected from the main body of
organic liquid and is now unable to drain from the pore. This liquid has become 'by-passed' by the
advancing water (stage 3).
Figure 9-8c uses a pore doublet model to demonstrate both snap-off and by-passing trapping
mechanisms. Once again, water enters the narrower pore first (stage 1). Due to the high aspect ratio in
the narrower pore, snap-off occurs (stage 2). Water continues to move through the narrower pore and
through the downstream node. No stable meniscus is formed and organic liquid in the wider pore is
by-passed (stages 3 and 4).
While the pore doublet model allows the organic liquid to be by-passed in at most one pore, in
a porous medium the organic phase in several pores may be collectively by-passed leaving an organic
blob which extends over the several pores. In contrast, blobs trapped by snap-off extend over one pore
a) no trapping :
stage 1
stage 2
b) trapping via by-passing :
stage 3
note that due to the
configuration of the
downstream node,
a stable interface is
not formed here.
stage 4
stage 1 stage 2
c) snap-off at top, by-passing below :
stage 3
STABLE
stage 1
stage 2
stage 3
stage 4
STABLE
FIGURE 9-8. Sketches illustrating trapping mechanisms using the pore doublet model (after
Chatzis et al., 1983)
- 117 -
-------
only. As pore aspect ratio decreases, the proportion of organic liquid trapped by by-passing, relative to
snap-off, increases, and blobs become larger and more complex. Soil or rock heterogeneity also
encourages trapping through by-passing, as will be shown later in this section (see also Chatzis et al.,
1983). Wettability, pore aspect ratio, and soil (or rock) heterogeneity are important factors influencing
trapping.
Quantitative Measurements of Residual Saturation
Quantitative measurements of residual non-wetting phase saturation have been made in many
petroleum reservoir cores and in glass bead packs ( see, eg.: Anderson, 1988; Chatzis et al., 1983,
1984, 1988; Hornof and Morrow, 1988;; Mohanty et al., 1980; Moore and Slobod, 1956; Morrow, 1979;,
1894; Morrow and Songkran, 1981; Morrow and Chatzis, 1982; Morrow et al. ,1988). In one type of
experiment a core or pack is saturated with a wetting phase. Then the wetting phase is displaced by a
non-wetting phase, reducing the wetting phase to its so-called irreducible saturation. Finally the
non-wetting phase is displaced by wetting phase at a low flow rate (low capillary number) to yield a
residual non-wetting phase saturation. Typically, water is the wetting phase and a hydrocarbon is the
non-wetting phase. (Our experiments are essentially the same; see Sections 5,6,7 & 8.)
Non-wetting phase saturation is measured as the volume of the non-wetting phase per unit
void volume, measured over a representative elementary volume of the porous media. If we assume
that the non-wetting phase is an organic liquid, this definition becomes:
o _ 'organic liquid f9~2)
0 ~ V
* voids
where the "o" indicates the organic liquid. The residual saturation at which the organic liquid becomes
discontinuous, as it is trapped by capillary forces, is defined by:
« _ 'discontinuous organic liquid (9-3)
v voids
where the "r" indicates residual. In two phase flow the wetting phase saturation is defined by:
Sw = 1.0 - S0 (9-4)
In the saturated zone of an aquifer we can usually assume that water is the wetting phase. A measure of
residual saturation often used in organic liquid pollution studies is the volumetric retention (e.g.,
de Pastrovich, 1979; Schwille, 1967):
R - liters °f residual organic liquid s x „ x 1Q3 (9.5)
cubic meters of soil
- 118 -
-------
where n is the soil porosity.
In strongly water-wet Berea sandstones, Chatzis and Morrow (1981) found residual organic liquid
saturations, S^ , between 27% and 43%. Chatzis et at. (1988) found that the majority of the blobs in
Berea sandstones were 'singlets' or 'doublets' occupying two pore bodies and the pore throat in
between. The largest complex blobs, which were often branched, tended to be no larger than 10 pore
bodies in size. Fifty percent of their blobs had an 'equivalent diameter' in the range of 30 to 130 microns,
consistent with measured pore sizes in Berea sandstone (Wardlaw and Taylor, 1976).
In water-wet, uniformly-sized glass beads, Morrow and Chatzis (1982), Chatzis et al. (1983), and
Morrow etal. (1988) found residual organic liquid saturations, Sor , between 14 and 16%. Bead size did
not seem to make any difference as long as the wetting phase velocity was low (low capillary number).
Most of the blobs were trapped as singlets (58%), much more so than in sandstone. Once again the
largest blobs measured on the order of 10 pore bodies long. Although by number there were more
singlets than any other blob size, they held less than 15% of the volume of the residual saturation (see
also Figure 6 in Wilson and Conrad, 1984). Blobs spanning five or more pore bodies held at least 50% of
the residual, even though they constituted less than 24% of the blob numerical population.
When they sintered the glass beads (consolidated the beads by fusing them together in a furnace),
Morrow ef al, (1988) found residual organic liquid saturations, Sor , between 20% and 30%, although they
did not feel that these numbers were reliable. The sintered beads should have better simulated the
sandstone cores. The reported saturations are consistent with this hypothesis.
Residual Organic Mobilization
Consider the residual saturation of organic liquid shown in the stable pore doublets, to the right of
Figure 9-8b and 9-8c. If the pressure gradient and Darcy velocity of the wetting phase are increased
sufficiently, some of the trapped blobs begin to move because of an increase in the viscous force
(Hornof and Morrow, 1988). This process is called mobilization. The pressure gradient must be
sufficiently high to squeeze the blobs through pore throats. At the leading edge of the blob, organic liquid
is displacing water (drainage), while at the trailing edge water is displacing organic liquid (imbibition).
The dynamic pressure difference required to support mobilization is proportional to the difference in
these drainage and imbibition capillary pressures (Melrose and Brandner, 1974).
A critical element in mobilization is the length of the blob in the direction of flow. For a fixed dynamic
pressure gradient the longer blobs are easier to mobilize because a greater pressure difference can be
established across them. The amount of residual saturation reduction by hydraulic mobilization can
strongly depend on how the trapped organics are distributed on a pore scale. If, for instance, many of
the trapped blobs are relatively long and complex ganglia formed by by-passing, they can be more easily
mobilized than if the majority of blobs are singlets resulting from snap-off. Once mobilized, blobs do not
always maintain their size. The larger mobilized blobs can break up into smaller blobs, with a significant
fraction being only temporarily mobilized (Chatzis et al., 1984). Also, as moving blobs collide with other
blobs, the blobs may coalesce (Ng et al., 1978). If substantial coalescence occurs, a bank of mobilized
organic liquid may form. The factors controlling blob break-up and coalescence are not, as yet, well
understood.
- 119 -
-------
Mobilization can be aided by a difference in density between the water and the organic liquid. If the
organic liquid is lighter than water, the blob will be buoyant and tend to rise vertically. If the blob is heavier
than water it will tend to sink. The vertical migration of the blob due to the density difference is resisted by
the capillary forces which trapped the blob in the first place. If the hydraulic gradient can be aligned so
that viscous and buoyant forces reinforce each other, their combined force may be sufficient to induce
mobilization.
There are many factors that influence initial trapping and subsequent mobilization: the geometry of
the pore network; fluid-fluid properties such as interfacial tension and density ratio; fluid-soil interfacial
properties which determine wetting behavior; and the applied water phase pressure gradient and its
alignment with gravity (Morrow and Songkran, 1981). When only two fluid phases are present these
factors can be incorporated into two dimensionless numbers. The ratio of viscous forces to capillary
forces is known as the capillary number, Nc , while the ratio of gravity forces to capillary forces is the
Bond number, NB '.
Ncl - ^- (9-6)
".' - (9-7,
where: k = intrinsic (absolute) permeability of soil
VPW = water phase pressure gradient
a = interfacial tension between the fluids
AQ = density difference between fluids
g = acceleration of gravity
R = representative grain radius
The superscript 1 in Ncl refers to the version of the dimensionless number. This model of the capillary
number assumes that the hydrostatic forces are negligible. The ratio of forces represented by the
capillary number can also be given in terms of the Darcy velocity in the water phase (see, e.g., Taber,
1981):
(9-8)
where: q» = specific discharge (darcy velocity) of the water phase
fj.» = viscosity of water
This version of the capillary number inherently accounts for relative permeability and the gravitational
(hydrostatic) portion of the driving force — the V(QW g z) term in the expansion of qw :
k k
q» = — — V( Pw + QW g z ) (9-9)
f^w
- 120 -
-------
where: k^ = relative permeability for the water phase
z = elevation of the point of interest above the datum
Nc2 can be obtained from Ncl by adding in the gravity term and multiplying by the relative permeability of
water. Wilson and Conrad (1984) defined a related capillary number for groundwater situations:
QW g /„ Kw Jw fj.w
(9-10)
a
P
where: Jw = hydraulic gradient in the water phase, VI — — + z
Q»S
Kw = water-saturated hydraulic conductivity of the soil,
Hw
This definition of the capillary number does not contain the relative permeability built into (9-8) , but does
contain the gravity term neglected in (9-6). When hydrostatics are negligible, (9-10) reduces to (9-6).
The Bond number also has at least one other version. It can be given in terms of intrinsic permeability
instead of a characteristic grain size:
(9-11)
a
Morrow and Songkran (1981) used the Kozeny-Carmen equation (Carmen, 1937) to determine an
approximate relation between the two forms of the Bond number for unconsolidated media of relatively
uniform grain size. This second version of the Bond number may be more convenient to use for soils
having a fairly wide grain size distribution — which makes selection of a representative grain size for use
in /vV rather arbitrary. Although buoyancy forces can aid mobilization, the following discussion will focus
on viscous forces and the capillary number, referring the reader to work by Morrow1 and to discussions
later in this section and in the next section concerning the effects of significant gravity forces.
There is a minimum viscous force needed to mobilized trapped residual. Once the viscous force
exceeded this critical value the residual saturation decreases. This process can be plotted in terms of
capillary number, as shown in Figure 9-9. In the figure, the residual saturation is normalized by its initial
value, S'or . This is, for example, the residual saturation measured in cores, packs, or columns, under
low capillary number conditions (for example, 27 to 43% in sandstones, and 14 to16% in uniform size
glass beads) . Above the critical capillary number, A^ , viscous forces begin to overcome the capillary
forces resulting in a reduction of the residual saturation. Equations (9-6) and (9-10) reveal that for a
prescibed gradient the critical capillary number is much easier to exceed in a large grained soil where the
intrinsic permeability, k, is high.
1. Morrow, 1979; Morrow and Songkran, 1981; Morrow et al., 1985; and Hornof and Morrow, 1987.
- 121 -
-------
Chatzis and Morrow (1981,1984) performed a large number of experiments using sandstone cores
to explore this correlation between capillary number, Ncl , and organic liquid saturation, sor • First a
residual organic liquid saturation, S'or , was established in the cores under low Bond and capillary number
conditions. Then the gradient across the core was increased in a stepwise manner, and at each step
residual organic liquid production was observed. A typical 5or/5^r vs. AC1 correlation curve for the
sandstone is shown as the left-hand curve in Figure 9-9. The critical capillary number at which motion
was initiated was typically AC1 = 2 x 10"5 for sandstones. AC' denotes the capillary number necessary
to displace all of the blobs; AC'1 « 1.3 x 10~3 for sandstones. Fifty percent of the residual organic liquid
is removed when Ncl ~ 3 x 10"4 . The left hand correlation in Figure 9-9 holds for a wide variety of
sandstones and organic fluids, provided that the organic liquid is the non-wetting fluid.
Similar experiments were performed for glass beads by Morrow and Chatzis (1982) and Morrow et at.
(1988). Using a variety of sizes of uniform beads and various organic liquids they obtained another
correlation, in terms of Sor/S'or vs. Nc2 , which is shown as the solid line in Figure 9-10. It has been
translated to Ncl so that it can be shown as the right-hand curve in Figure 9-9. It looks nothing like the
sandstone curve. A significantly larger capillary number (i.e. more gradient) is required to mobilize the
blobs in uniform glass beads. This is explained by the very large aspect ratio of the bead pack pores
relative to sandstones and the dominance of capillary trapping by snap-off. Sintering the beads (melting
them together slightly) reduces the aspect ratio and produces a curve much closer to the sandstone
curve , as shown by the x's in Figure 9-10 (Morrow and Chatzis, 1982; Morrow et a/., 1988). In a similar
Reduced
Residual
Saturation
0.8-
0.6-
0.4-
0.2-
typical
bead-pack
curve
Mobilization of non-wetting
fluid trapped at low
capillary numbers
typical
sandstone
curve
10
-6
TT
10
-5
10
-4
1(
3
10
,-2
Capillary Number, N* —
FIGURE 9-9. Relationship between residual saturation and capillary number for sandstones and
glass beads. The sandstone curve is from Chatzis and Morrow (1981,1984); the
bead-pack curve is based on work reported in Morrow and Chatzis (1982), and
Morrow et al. (1988).
- 122 -
-------
study using air, Morrow and Songkran (1981) found a that critical Bond number (NBl) of about 0.005 was
needed to reduce trapping in glass beads.
Gravity and viscous forces also play a strong role in the initial capillary trapping of the organic liquid. If
the wetting phase velocity is high enough, or if the density difference between fluids is large enough, the
trapping mechanism is not as efficient (see, eg., Mohanty et a/., 1980), and less residual is trapped in
the first place (Chatzis and Morrow, 1981,1984; Morrow and Songran, 1981; Morrow et at., 1988).The
results depicted in Figure 9-9 were developed for very low capillary and Bond number original
displacements. The capillary number was then increased to mobilize the residual. Contrast this to the
dashed curve Figure 9-10, which depicts glass bead experiments in which the original rate of
displacement was varied over a much wider range. In this graph the reference residual saturation, s'or <
is that measured for a low capillary number original displacement. Initially displacing the continuous
non-wetting fluid at a high rate, with the wetting fluid, significantly reduces the residual. Put another way,
it is easier to avoid trapping residual non-wetting phase than it is to mobilize it afterward (Morrow et al.,
1988). In sandstones the difference is not nearly so dramatic (see, eg., Chatzis and Morrow,
1981,1984).
In groundwater situations we may see a similar behavior with dense organic liquids as they move
downward through the saturated zone, particularly in coarse grained soils (where k is high). The Bond
number may be high enough to reduce the amount of initial entrapment.
Reduced
Residual
Saturation
1 -
0.8-
0.6-
0.4-
0.2-
Mobilization of non-wetting
fluid trapped at low
capillary numbers
•>
Trapping of X
non-wetting fluid at >
various capillary
numbers
Sintered Glass Beads
Morrow ef a/., 1988
10"6
-TTTT1 I ' I I I 1M|
io-5 io-4
Capillary Number,
FIGURE 9-10. Residual saturation in uniform glass beads due to variable capillary number
entrapment of the continuous non-wetting phase (dashed line), and due to
mobilization of non-wetting phase originally trapped at a low capillary capillary
number (solid line). After Morrow and Chatzis (1982) and Morrow et al. (1988).
- 123 -
-------
MICROMODEL FLOW VISUALIZATION OF TWO PHASE DISPLACEMENT
AND CAPILLARY TRAPPING
Flow visualization techniques in a homogeneous micromodel illustrate the process of organic liquid
invasion of the saturated zone and subsequent displacement by flowing groundwater — resulting in
trapped organic phase being left behind. It visually illustrates many of the issues that were raised in the
review. Later, these micromodel results are compared to micromodel results for higher flow rates,
heterogeneous media, and the vadose zone (in Section 10).
Organic Liquid Invasion into a Water Saturated Homogeneous Micromodel (Wilson et al. 1988)
A water-wet etched glass micromodel experiment (Figures 8-1 & 9-11) serves as good example of
the displacement process of organic liquid movement into a water-saturated homogeneous soil. The
water-saturated micromodel was oriented vertically and flooded from the top with an organic liquid, red
dyed Soltrol, at a relatively slow rate of 0.096 ml/min using the techniques described in Section 8. The
steady state condition, at the end of the organic liquid invasion, is shown in the upper photographs of
Figures 9-11 through 9-13. Figure 9-11a depicts the entire model at equilibrium, while Figures 9-12a &
9-13a depict close-ups. In these photos the red dyed Soltrol appears dark gray; the water was not dyed.
The Soltrol continues to flow through the model at the rate of 0.096 ml/min, but the Soltrol and water
saturations have stabilized. The saturation pattern remained the same when the flow was cut off
afterwards. Figure 9-12 is an area located just below the very center of the model. Figure 9-13 shows an
area located very near the top of the model and slightly to the right of centerline. Overall, the
displacement of water was fairly efficient, except at the bottom of the model (right side of photo) where a
capillary end effect came into play (see Figure 9-11 a). This end effect is discussed later, under the topic
of heterogeneous porous media. A videotape showing the displacement dynamics of this experiment is
available (Mason et al., 1988; see Appendix B).
The residual or irreducible water saturation occupies several by-passed pockets of the pore
network, as well as films, rings, and wedges in individual pores (see, eg., Amaefule and Handy, 1982;
Dullien, et al., 1986; Chatzis, et al., 1988). The term 'film' refers to the film of wetting fluid covering all
water wet glass surfaces; 'ring' refers to the pendular rings of wetting fluid found in many of the pore
throats; and 'wedge' refers to the wetting fluid in the narrow crevices of the the pores, where the glass
plates meet (these films, rings, and wedges are similar to those that are observed in natural porous
media - natural pores are not circular, nor are micromodel pores).
The micromodel experiments illustrated in this report depicts conditions that we were able to
reproduce by repeating the experiment. An example is shown in Figure 9-14. The photograph in this
figure can be compared to Figure 9-11 a and depicts the results of an identical experiment where Soltrol
has advanced into the micromodel, displacing water. The character of the saturation pattern is similar for
both experiments; only the details vary. The two-phase flow field appears to be a stochastic process,
with the same mean behavior, but a different detailed realization for each experimental replication.
- 124 -
-------
a.
b.
FIGURE 9-11. In the upper photo (a) Soltrol displaced water from the left (the top of the
model) to the right (the bottom of the model), yielding a residual
(irreducible) wetting phase saturation. In the lower photo (b) Soltrol was
displaced by water from the right (the bottom of the model) to the left (the
top) yielding a residual non-wetting residual saturation. Soltrol was dyed
red and appears dark grey; the water was not dyed. The photos record
steady state flow conditions at the end of the displacements.
- 125 -
-------
b.
top
flow
FIGURE 9-12. Detail from Figure 9-11 showing conditions following the displacement of
the water by Soltrol (a. upper photo), and at residual non-wetting phase
saturation (b. lower photo). The area is located just below the very center of
the model.
- 126 -
-------
a.
top
flow
b.
top
How
FIGURE 9-13.
Detail from Figure 9-11 showing conditions following the displacement of
the water by Soltrol (a. upper photo), and at residual non-wetting phase
saturation (b. lower photo). The area is located near the top of the model
just to the right of the centerline.
- 127 -
-------
Displacement of Organic Liquid by Water in a Homogeneous Micromodel (Wilson et al. 1988)
In the next step water was displaced upward through the micromodel at the same rate, pushing much
of the Soltrol out, but leaving behind a capillary-trapped, residual Soltrol saturation. These
displacements were also captured on videotape (Mason et al., 1988). When the model reached a new
steady state, additional photos were taken. The lower photos in Figures 9-11b through 9-13b depict the
residual organic liquid left behind. Water continued to flow through the model as the photos were taken.
The saturation pattern did not change when the flow was cut off. As seen in these photos, the residual
organic liquid saturation in these strongly water-wet models is composed of disconnected blobs and
ganglia which are fairly evenly distributed throughout the model and appear to occupy up to 30% of the
pore space.
Microscopic Inspection of Blob Size and Shape in Micromodels
Figures 9-15 and 9-16 present 'pore scale' close-ups of typical blobs taken from similar experiments
conducted in the same micromodel.The water was dyed blue in Figure 9-15 and may appear as a light
grey. Figure 9-15a is a photomicrograph of a blob trapped in a 'pore body'. The surrounding 'pore
throats' are filled with water, the wetting fluid. This blob is referred to as a 'singlet' and is usually trapped
by snap-off. The trapped singlet is roughly the size of the pore body. Figure 9-15b depicts a 'doublet', a
blob occupying two pore bodies and the pore throat between. Although many blobs have shapes like this
FIGURE 9-14. A second experiment in the homogeneous micromodel, depicting conditions at
the end of the organic liquid invasion-compare to Figure 9-11 a .
- 128 -
-------
a.
b.
FIGURE 9-15. Photomicrographs of (a) a singlet blob occupying one pore body in the
upper photo, and (b)a doublet blob occupying two pore bodies and a
pore throat in the lower photo. The water is dyed blue (light grey); the
Soltrol is dyed red (dark gray).
- 129 -
-------
41
FIGURE 9-16. Photomicrograph of a complex blob as observed in the micromodel.
singlet and doublet, some are more complex and extend over a number of pores bodies and the
connecting pore throats, such as the micromodel blob shown in Figure 9-16. This branched blob was
presumably by-passed. Other examples of both simple and complex blob shapes can be seen in the
upper photos of Figures 9-11b through 9-13b. These more complex shapes include a number of
branched blobs with more than two 'ends'. The population of blobs indicates that both by-passing and
snap-off were operable in the micromodel experiment.
CAPILLARY TRAPPING AND RESIDUAL SATURATION IN AN UNCONSOLIDATED SOIL: the
SEVILLETA SAND
Earlier we reviewed the results of some of the experiments conducted by petroleum engineers to
study capillary trapping and residual saturation in reservoir cores and glass bead packs. Few similar
studies have been carried out for unconsolidated natural soils. Using a strongly water-wet Sevilleta sand
for the media and water and Soltrol for the fluids, we made quantitative measurements of Soltrol residual
(see Sections 4 & 5). In other experiments we used styrene as the organic phase, which we hardened in
place and photographed (see Section 7).
The photomicrograph of the unconsolidated Sevilleta sand (Figure 4-6) and the sand grain size
distribution curve (Figure 4-5) appear to resemble those of (unconsolidated) uniform glass beads, more
than those of (consolidated) sandstone. Thus we hypothesize that sand should exhibit a behavior more
like glass beads than like sandstone. That is, the snap-off mechanism of trapping should be more
common in the uniform Sevilleta sand than it is in a sandstone, and the residual saturation should be
dominated by singlet and doublet blobs. There should be relatively few large and complex blobs. Berea
- 130 -
-------
sandstones have residual saturation values in the range 27 to 43%, while uniform glass beads have
values of 14 to 15%. If our hypothesis is valid we would expect the Sevilleta sand to exhibit a residual
saturation between 15 and 25%, with a probable value of 18-20%.
Quantitative Measurements of Residual Saturation In Soil Columns
Quantitative measurements of residual saturation were made in short soil columns, packed
with Sevilleta sand, as described in Section 5. Briefly, in each experiment an initially water saturated soil
column was first flooded with Soltrol to simulate the movement of an organic liquid into the saturated
zone. The injection pressure was held low enough to prevent the Soltrol from passing through a
water-wet filter at the lower endcap of the column. This boundary condition is different than that used in
the micromodel experiment, and reduced the possibility of capillary end effects such as observed in
Figure 9-11 a. After the fluid saturations stabilized, the column was flooded with water at a low velocity.
Six pore volumes of water were found to be more than sufficient to reach a stable Soltrol residual
saturation. As documented in Section 5, fluid saturations were determined gravimetrically using the
density difference between the fluids.
The results are presented in Tables 9-1 and 9-2 in terms of the residual organic liquid (Soltrol)
saturation, Sor , and the maximum organic liquid saturation, S0 , measured when water was at (or near)
the residual (irreducible) saturation, swr • These measurements correspond to the micromodel Soltrol
saturations depicted in Figure 9-11b and 9-11 a, respectively. Water saturations and organic retention
can be calculated from equations (9-4) and (9-5). Included in Tables 9-1 and 9-2 are measured values
of sample porosity, n, and bulk density, Qb. These measurements provide additional information for
ancillary calculations, and their variability is a measure of experimental control over the packing.
Many of the earlier Soltrol experiments were run without the benefit of a constant temperature
cabinet. This is noted in the second column of Table 9-2, and the difference between maximum and
minimum temperatures observed during the course of the experiment is given in the last column. As
discussed in Section 5, those experiments with large temperature fluctuations provided less reliable
results. In particular, it can be seen from the Table 9-2 that the two experiments with the largest
temperature fluctuations also had the largest estimated porosity and extreme estimates of residual
saturation. Sample statistics for the twenty-two experiments are presented in Table 9-1, with data
divided into three categories: a) all experiments; b) thirteen experiments conducted with good
temperature control (temperature range At < 2°C); and c) nine experiments conducted with poor
temperature control (At > 2°C). Error estimates, using a worst case error approach in which all
estimated errors were assumed additive and propagated through the calculations, are also given in
Table 9-2 for porosity, bulk density, maximum Soltrol saturation, and residual Soltrol saturation.
Comparing these estimates to the sample standard deviations suggests that the thirteen temperature
controlled experiments account for the known and tractable experimental errors. The estimate of
residual saturation taken from these thirteen reliable experiments is Sor = 27.1 +1.7 %. This estimate is
slightly lower than our earlier published estimates (Wilson, et al. 1988), which were biased by the
temperature fluctuations of the first experiments.
- 131 -
-------
all 22
experiments
good temp.
control
13
experiments
poor temp.
control
9
pvnprimpnt<:
porosity
(%)
34.3 ± 1.2
33.9 ± 0.6
34.9 ± 1.7
bulk density
(g/cm3)
1.741 ± 0.033
1.752 ± 0.016
1.724 ± 0.044
maximum organic
liquid saturation
(%)
84.7 ± 3.4
85.1 ± 2.8
84.0 ± 4.1
residual organic
liquid saturation
(%)
28.0 ± 3.8
27.1 ± 1.7
29.3 ± 5.4
TABLE 9-1. Soltrol residual saturation and other measurements in Sevilleta sand, for three
temperature dependent categories (sample mean + sample standard deviation).
Correlation with porosity-
The earlier discussion on mechanisms focused on capillary trapping as a function of pore
structure, among other factors. Each experiment recorded in Table 9-2 included measurements of
porosity and bulk density. Presumably, these measurements give a measure of pore structure as
controlled by the density of soil packing. A tendency for increased trapping with decreased porosity has
been reported in the petroleum literature (Morrow et al., 1988). This is believed to be related to the fact
that for more heavily cemented petroleum reservoir rocks, the pore throats (from which organic liquid is
easily displaced by water) make up a smaller percentage of the total pore volume (Chatzis et al., 1983).
Lower pore connectivity, another attribute of low porosity (heavily cemented) media, is also suspected
of contributing to increased trapping (Pathak et al., 1982). Although the Sevilleta is an unconsolidated
soil, we were concerned that we too might see some systematic change in residual saturations with
change in porosity from one packing to the next. Figure 9-17 presents a plot of maximum organic liquid
saturation and residual saturations for the 13 best samples in Table 9-2. In these results, there appears
to be no discernible correlation between porosity and either the residual or maximum organic liquid
saturations. Although the porosities varied somewhat, this variation did not seem to affect the
measurements of saturation in any systematic way. The lack of correlation between Sor and n is
probably due to the small range over which the porosities varied and the unconsolidated nature of the
soil.
Photomicrographs of Blob Casts and Pore Casts
Photomicrographs of polymerized styrene blobs embedded in epoxied Sevilleta soil pore casts, are
shown in Figures 9-18 and 9-19. These pore casts were constructed in short TFE columns of Sevilleta soil
using the techniques described in Section 7. Compare the singlet and doublets shown in Figure 9-18,
from the soil column, to those in Figure 9-15, from the micromodel. The similarity between blob shapes
in the two different media gives confidence that micromodels can reasonably be used to simulate the
- 132 -
-------
trial
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
in
box
?
no
no
no
no
no
no
no
no
no
no
no
no
no
yes
yes
yes
yes
yes
yes
yes
yes
yes
porosity
(%)
33.8
36.5
34.0
34 1
33.2
34.8
38.6
34.6
34.6
34.3
34.1
34.1
33.7
34.9
34.2
34.7
33.7
34.1
34.1
32.8
33.1
33.2
est.
error
(%)
0.6
0.5
0.5
0.5
0.6
0.5
0.9
0.5
0.5
0.5
0.6
0.6
0.5
0.5
0.6
0.6
0.5
0.6
0.6
0.6
0.6
0.5
bulk
density
(g/cm3)
1.754
1.683
1.750
1.747
1.769
1.729
1.626
1.732
1.733
1.741
1.747
1.746
1.756
1.726
1.745
1.730
1.758
1.747
1.746
1.781
1.774
1.771
est.
error
(g/cm3)
0.008
0.007
0.007
0.008
0.008
0.007
0.017
0.007
0.008
0.008
0.009
0.009
0.008
0.007
0.010
0.010
0.006
0.009
0.011
0.010
0.009
0.007
organic liquid saturation (%)
maximum
84.6
82.2
77.5
85.7
92.7
84.1
78.6
80.2
84.7
87.5
87.0
82.6
83.5
87.0
85.8
88.2
83.8
85.6
86.0
88.3
83.0
84.1
est.
error
3.5
3.0
3.1
3.5
3.9
3.2
3.7
3.0
3.5
3.6
3.9
3.6
3.3
3.4
3.9
3.9
3.2
3.7
4.0
4.1
3.7
3.4
residual
29.3
32.5
24.6
29.3
34.7
32.7
21.6
37.0
29.3
28.6
26.9
24.3
23.1
27.9
25.0
28.5
25.4
27.1
29.2
27.6
24.6
26.6
est.
error
1.7
1.5
1.4
1.7
2.0
1.6
1.5
1.7
1.7
1.8
1.8
1.6
1.4
1.6
1.7
1.8
1.4
1.7
1.9
1.9
1.7
1.7
temp.
range
(°C)
1.0
5.0
1.0
1.0
3.0
2.0
11.0
3.0
2.0
3.0
1.5
2.8
2.8
0.3
0.8
0.4
0.3
0.4
0.9
1.0
1.0
1.1
TABLE 9-2. Summary of Soltrol / Sevilleta sand saturated zone results.
- 133 -
-------
IUU-
Percent
Saturation
90-
80-
70-
60-
50-
40-
30-
20-
10-
0-
A A ~
A A A A ± A A
A ~
Maximum Organic Liquid Saturation
S0 - 1 - Swr
—
—
—
• • •*'" • • "~
Residual Organic Liquid Saturation, Sor —
—
32
33 34
Porosity, Percent
35
FIGURE 9-17. Correlation of maximum Soltrol saturation (triangles), and residual Soltrol saturation
(squares), to porosity in the Sevilleta sand.
immiscible displacements taking place in the subsurface (Conrad et al., 1989). Figure 9-19 shows a
variety of blobs in a similar pore cast. Some of these are singlets, while others are much more complex
and have branches. The pore casts also reveal the position of the blobs within the pore space and their
position relative to the sand grains. In the pore throats, there appears to be only a thin film of water
between the blob and the sand grain 'pore wall'. The micromodel mimics this behavior (Figure 9-15 and
9-16). For pore bodies occupied by organic liquid, sometimes there is a difference in the observed
behavior found in the micromodels and the pore casts. Blobs in the micromodel appear to occupy most
of each pore body they are found in, again leaving only a film of water between the blob and the 'pore
wall'. Blobs in the Sevilleta pore casts sometimes share the pore bodies with significant amounts of
water (see Figure 9-18b), perhaps due to greater surface roughness and irregularities in the pore walls
of the Sevilleta sand. This is certainly not always the case (see Figure 9-19). Note that no attempt had
been made to achieve similitude between the micromodel network and Sevilleta sand. One other factor
to keep in mind: the styrene shrinks slightly as it polymerizes.
SEM photomicrographs provide a three-dimensional view. Views of polymerized styrene 'blob casts'
from another Sevilleta soil column are shown in Figure 9-20a. The rough spots on the surface of these
blobs is probably due to the SEM coating process. Once again there is a typical singlet (upper left) and a
doublet (lower left). The singlet is almost spheroidal, with a length of 200 microns and a diameter of 100
or more microns. The doublet includes one pore body of roughly 100 microns in diameter and another
that appears to be almost 200 microns in diameter. At the pore throat, the doublet is only 20 microns in
diameter. The two more complex shapes involve four (upper right) and six (lower right) pore bodies. The
pore bodies of these more complex blob shapes appear to have a diameter of roughly 100 microns,
- 134 -
-------
a.
b.
FIGURE 9-18. Sevilleta sand pore cast photomicrographs of (a) a singlet blob occupying
one pore body, in the upper photo, and (b)a doublet blob occupying two
pore bodies and a pore throat, in the lower photo. The styrene non-wetting
phase is dyed blue (lightest grey); the epoxy wetting phase is dyed green
(light gray). The solid grains show as dark colors, or in the case of
translucent minerals, as a different shade of green (medium grey).
- 135 -
-------
FIGURE 9-19. Sevilleta sand pore cast photomicrograph of a variety of blobs including
some that are complex and branching. The styrene non-wetting phase is
dyed blue (lightest grey); the epoxy wetting phase is dyed greenish yellow
(light gray). The solid grains show as dark colors, or in the case of
translucent minerals, as a different shade of green (medium grey).
while the typical throat diameter is less than 50 microns. Contrast these diameters with the grain size
distribution of the very uniform Sevilleta soil, as given in Figure 4-5. The mean grain diameter is 300
microns. Few grains are larger than 500 microns or smaller than 100 microns. The typical pore body blob
diameter of 100 microns is smaller than the mean grain size. The blob lengths depend on the number of
pore bodies involved. The singlet in Figure 9-20a is 200 microns long, the doublet bends through a length
of 400 microns, the quadruplet is 800 micros long, and the contorted sextuple! is at least 800 microns
long (part of the sextuple! is sticking up into the photo, and thus its entire length is not readily apparent).
Several more complex, branching blob casts, taken from the same column, are shown in Figure 9-20b.
These branched blobs often have holes in them and 10 or more pore bodies. Presumably, the larger
holes were partially filled with a sand grain when in the column. Perhaps the smaller holes contained only
the 'pointed edge' of a grain.
At the beginning of this project, we anticipated taking pore casts and performing a statistical size
and shape analysis. We referred earlier to similar analyses which have been conducted on blob casts
taken from oil reservoir sandstone cores and glass bead packs (Morrow and Chatzis, 1982, Chatzis et
al., 1983, and Chatzis et al., 1988). As discussed in Section 7, the pore casts taken from the Sevilleta
- 136 -
-------
a.
b.
FIGURE 9-20. Sevilleta sand blob cast photomicrographs of (a) non-branching blobs, and
(b)branching blobs.
- 137 -
-------
soil columns appeared to be too fragile for quantitative analysis using a Coulter Counter or even an
image analyzer. Many of the larger casts broke into several pieces when handled. The right end of the
blob cast depicted in the lower right photo of Figure 9-20b is broken. This complex blob of roughly 15
pore bodies was even bigger before it broke. The branched blob in the lower left also appears to have
broken ends. Breakage of this kind would have severely biased a size analysis. The petroleum research
samples must have also suffered from the same problem, and their results may be biased. A second
approach would be to use image analysis on the sectioned epoxied pore casts. This proved beyond the
resources of this project, but is strongly recommended in future research efforts. Also recommended is
mercury and nitrogen gas porosimetry of the soil samples in order to yield estimates of the actual pore
size distribution.
Our size and shape analysis has been limited to visual inspection of the blob population through the
microscope. Figure 9-21 is a photomicrograph of a Sevilleta sand pore cast, showing many blobs. Figure
9-22 is an SEM photomicrograph of a random sample of Sevilleta blob casts. These photographs
document what we saw through the microscope. Many, if not most, of the Sevilleta blobs are larger than
a singlet or doublet. A closer inspection of the pore cast, with a stereoscope, shows that most of the
blobs, which appear on the surface to occupy only one or two pores, actually extend down into the
sample. Blobs commonly wind down and around into the pore network forming 'ghosts' at depth when
viewed through the relatively transparent quartz sand. Figure 9-22 may also be somewhat misleading in
reference to the size of the blobs. It must be remembered that many of the blob casts in the
photomicrograph are broken. Even if the complex blobs do not outnumber the singlets and doublets,
they clearly contain almost all of the residual saturation volume. Overall, it appears that the more
complex blobs dominate the Sevilleta sand.
Discussion of Sevilleta Sand Experiments
We hypothesized that Sevilleta sand would behave more like glass beads than like sandstone. The
evidence does not support this hypothesis. The observed residual Soltrol saturations of 25-29% were
much higher than the range of 18-20% that we anticipated, and the residual blobs were more complex.
The Sevilleta sand appears to behave more like sandstones than glass beads. The complexity of the
blobs may even exceed that of sandstones. We won't propose an explanation for this behavior until later
when we contrast the results from the Sevilleta sand and other soils.
The high residual saturation level found in Sevilleta sand has important implications on aquifer
contamination characterization and remediation. However, caution should be exercised when
generalizing the results of experiments on only one soil.
Implications for aquifer contamination -
Twenty-seven percent of the Sevilleta sand pore space is occupied by immobile,
discontinuous, blobs of Soltrol. If this estimate is typical for most organic liquids and sandy soils there is a
tremendous storage capacity of organic liquid pollutants in the saturated zone. Expressed in terms of
volumetric retention (9-5), the Sevilleta sand has the capacity to store over 90 liters of Soltrol per cubic
meter of soil. For example, a single 10,000 gallon spill of an organic liquid could be absorbed in about
- 138 -
-------
FIGURE 9-21. Photomicrograph of Sevilleta sand pore cast covering many pores.
FIGURE 9-22. SEM photomicrograph of many blob casts from the Sevilleta sand.
- 139 -
-------
420 m3 of saturated soil. This volume corresponds to a cube of soil with sides only 7.5 meters in length.
Even if typical soil residual saturation is only half of this, and similar to that of glass beads, the capacity is
still large.
Implications for biotransformation —
There are several implications of these observations on the study and practice of biotransformation
of organic components. Microorganisms attached to the pore wall at the water-solid interface
experience an environment that depends on their location. In blob-filled pore throats, the organisms
have ready access to organic components that partition into the aqueous phase from the nearby blob
surface. However, these organisms may not have a ready supply of other substrates, which can only
diffuse (or flow) slowly through the thin water film that lines the pore throat (see Figures 9-18b, 9-19,
9-12b, 9-13b,9-15,and 9-16). In blob-filled pore bodies that have a surrounding water film, the same
situation applies, as it does for microorganisms attached to the blob surface in these regions of thin
aqueous films. In the other pores, occupied by flowing water, wall-attached microorganisms are readily
exposed to available dissolved substrates, subject only to upstream substrate re-supply and biological
consumption.
Migration of microorganisms is also probably influenced by the spatial distribution of blobs. It should
be difficult for a seed population to find its way into regions with thin aqueous films because of their low
flow rates and tortuous diffusion paths. In any event, most organisms would probably not thrive in these
stagnant regions because of the substrate re-supply problem and the possibility of toxicity due to locally
high concentrations of dissolved organics.
Implications for phase partitioning-
Blob size and shape influence the partitioning of organic liquid components to the aqueous phase.
Mass transfer coefficients used in the mathematical models of this partitioning often employ the analogy
of an equivalent spherical blob (e.g., Pfannkuch, 1984). Certainly singlets can be represented by a
spherical model, but it is less clear that this model works for the multiple pore-body elongated blobs of
Figure 9-20a , or branched blobs of Figure 9-20b . These large complex blobs may be fewer in number
than the simple shapes, but they hold the majority of the liquid organic volume. Their surface area to
volume ratio is also greater. For multi-component organic liquids it is easier for these more complex
shapes to 'leach out' a lighter organic component than for a spherical blob of equal volume (this
equivalent sphere may be much larger than a real pore). But break these complex shapes into many
'model' spherical singlets of realistic size, say 100 microns diameter for the Sevilleta sand, and the
model leach rate increases dramatically. Coupled with this issue is the fact that there is a distribution of
blob sizes and shapes in any sample (see Figures 9-21 and 9-22). An appropriate definition of
equivalent sphere size has yet to be proposed, although work is progressing in that direction (eg, Powers
ef a/., 1988).
The position of a blob within the pore space also has a strong influence on mass transfer between
phases. If a large portion of the blob surface is in contact with only a thin film of water, as certainly seen in
our pore casts and micromodels, then the transfer rate may be limited by advection or diffusion in the
film. Videotaped micromodel observations indicate that the flowing water moves around the blobs,
- 140 -
-------
through the unoccupied pores, with little water movement in the films to help advect organic components
away (Mason et a/., 1988). This initial observation requires additional experimental confirmation.
As components of the organic phase partition from it, a blob gets smaller, and the residual organic
saturation is reduced over time. Let's consider the solubilization of a one-component residual organic
liquid. Simple compositional models of interphase mass transfer usually assume that a local equilibrium
is reached between the phases. As groundwater passes through the zone of residual saturation, trapped
organic liquid is dissolved by the water. In the case where a local equilibrium is in fact reached, only
blobs at the extreme upstream end of the zone are dissolved. By the local equilibrium assumption, as
soon as water comes into contact with organic liquid the solubility limit of the water with respect to the
organic is reached. With time, blobs at the upstream become dissolved and completely disappear. Blobs
not at the upstream end are not dissolved at all because passing water has already been completely
saturated with organic liquid. If we look spatially at this equilibrium dissolution process as a function of
time, we hypothesize that organic liquid saturation maintains a sharp front at the upstream end (Figure
9-23a), as does the dissolved concentration in the water phase (Figure 9-23b).
Often, when groundwater velocities are high enough or the dissolution kinetics are slow enough, the
local equilibrium assumption no longer holds. In this case, the saturation distribution may look like that
shown in Figure 9-24. A dispersed zone forms where traveling groundwater has not yet reached its
solubility limit for the organic liquid. The dispersed zone is the region in which there is a reduced residual
saturation (Q
-------
a.
Oor
Organic
Liquid
Saturation
Direction of Groundwater Flow
"• solubility
limit
Concentration
of Dissolved
Organics
n
J
/
J
f
J
r .:•... . ,..;.
FIGURE 9-23.
at t,
at t,
Direction of Groundwater Flow
The spatial distribution of a single-component residual organic liquid undergoing
dissolution as a function of time when the local equilibrium assumption is invoked.
Notice that a sharp front is maintained at the dissolution front for both the organic
liquid saturation, (a), and for the concentration of the organic dissolved in the water,
(b).
Organic
Liquid
Saturation
Direction of Groundwater Flow
FIGURE 9-24. The spatial distribution of a single-component residual organic liquid undergoing
dissolution as a function of time when a local equilibrium between the fluid phases is
not reached. A dispersed zone forms and grows until a steady state is reached.
- 142 -
-------
RESIDUAL SATURATIONS FOR VARIOUS ORGANIC LIQUIDS
If organic liquids are trapped under low capillary number conditions, then the amount and distribution
of the residual saturation should be similar for a wide variety of organics liquids. This is the principle used
in simulating Soltrol by styrene in the pore and blob casts. In particular, we expect that, below threshold
capillary numbers, residual saturation is independent of interfacial tensions and organic liquid viscosity.
Similarly, we expect that for the low Bond number conditions of our experiments, the residual saturation
is independent of organic liquid density. We tested these hypotheses by running additional quantitative
experiments in the Sevilleta sand short columns, using kerosene, gasoline, p-xylene, PCE, and
n-decane for the organic liquid.
The organic liquids selected for use in these experiments were chosen to be representative of
several classes of organic liquid pollutants often present at landfills, hazardous waste sites, and
accidental spills. Gasoline leakage from thousands of storage tanks and spills, incurred during
transportation and distribution of the fuel, are responsible for a large number of the groundwater
contamination incidents reported today. Gasoline and its individual Components (as represented by
xylene for the lighter aromatic fraction and decane for the less volatile fraction) were considered for
investigation because of gasoline's abundant use and misuse. Kerosene is very similar in composition to
the aviation fuel used in all military aircraft and long-distance passenger-carrying aircraft. Groundwater
contamination problems have occurred from spills of aviation fuel at airfields and military bases around
the country. PCE was one of two chemical carcinogens, TCE was the other, found most often in
groundwater according to a report compiled in 1980 by the EPA Office of Drinking Water (Burmaster &
Harris, 1982). PCE was also chosen because it represented the class of organic liquids that are denser
than water.
The experimental results are presented in Tables 9-3 through 9-8. The eight kerosene experiments
presented in Table 9-3 were run early in the project without the benefit of a constant temperature
cabinet. The five gasoline experiments presented in Table 9-4 were conducted under good temperature
control,as were the five n-decane experiment shown in Table 9-5. The six p-xylene tests summarized in
Table 9-6 were conducted under a fume hood, but the temperature was fairly well controlled. Although
the smaller density difference between water and the p-xylene resulted in greater error propagation for
organic liquid saturations, the actual sample standard deviation was relatively small in comparison. The
PCE tests described in Table 9-7 were also conducted under a fume hood with relatively good
temperature control. A tabulation of the raw data from all these experiments can be found in Appendix D.
A summary of all the tests is given in Table 9-8. It is apparent that the soil packing was very
consistent from the tests of one liquid to those of another. The soil dependent properties, porosity and
bulk density, remained nearly constant over the most of the experiments. The porosities for the PCE
experiments, however, were slightly but consistently lower than all the other experiments because the
operator of these experiments tightened the top endcap on the column more than normal as an added
precaution against leakage. This slightly lower soil porosity appeared to have no influence on the results.
We were concerned about the variability of the maximum organic saturations shown in the tables,
and whether this variability would substantially affect the residual saturation results. In running the
- 143 -
-------
Trial
1
2
3
4
5
6
7
8
Avg*.
porosity
(%)
33.1
33.0
33.7
33.2
33.8
33.7
33.3
35.0
33.6
error
(%)
0.4
0.4
0.4
0.5
0.5
0.4
0.3
0.5
0.6*
bulk
density
(g/cm3)
1.765
1.769
1.751
1.765
1.747
1.750
1.760
1.721
1.754
error
(g/cm3)
0.002
0.003
0.003
0.003
0.003
0.002
0.001
0.008
0.014*
Organic liquid saturation (%)
maximum
73.3
52.8
78.9
70.1
73.8
73.1
80.1
80.9
72.9
error
2.0
1.6
1.6
2.0
2.1
2.0
2.1
2.8
9.0*
residual
28.1
28.3
27.3
25.3
23.8
23.3
25.4
29.0
26.3
error
0.9
0.9
0.9
0.8
0.8
3.2
0.8
1.5
2.1*
Temp.
range
(°C)
5.0
4.0
3.0
3.0
3.0
5.0
2.0
7.0
* Average = sample mean Istandard deviation
TABLE 9-3. Summary of kerosene / Sevilleta sand saturated zone results.
Trial
1
2
3
4
5
Avg.
porosity
(%)
34.3
33.1
32.6
32.1
33.5
33.1
error
(%)
0.7
0.6
0.6
0.6
0.6
0.8*
bulk
density
(g/cm3)
1.742
1.772
1.786
1.799
1.762
1.772
error
(g/cm3)
0.011
0.008
0.008
0.009
0.009
0.022*
Organic 1
maximum
75.2
80.9
83.1
84.5
81.2
81.0
iquid s
error
6.0
6.2
6.3
6.5
6.2
3.6*
aturation
residual
26.9
29.3
31.0
26.9
28.8
28.5
(%)
error
2.6
2.8
2.9
2.6
2.7
1.7*
Temp.
range
(°C)
0.8
1.3
0.6
1.1
0.4
* Average = sample mean ± standard deviation
TABLE 9-4. Summary of gasoline / Sevilleta sand saturated zone results.
- 144 -
-------
Trial
1
2
3
4
5
Avg.*
porosity
(%)
34.7
33.2
34.2
32.2
33.5
33.6
error
(%)
0.6
0.6
0.6
0.6
0.6
1.0*
bulk
density
(g/cm3)
1.732
1.776
1.743
1.797
1.763
1.762
error
(g/cm3)
0.010
0.010
0.010
0.010
0.010
0.026*
Organic I
maximum
77.0
72.3
78.8
82.4
82.5
78.6
iquid s
error
3.6
3.6
3.9
4.1
3.9
4.2*
aturation
residual
28.6
24.6
25.1
26.7
24.6
25.9
(%)
error
1.8
1.8
1.9
1.8
1.6
1.7*
Temp.
range
(°C)
2.6
1.1
1.5
2.1
2.0
-—
Average = sample mean ±standard deviation
TABLE 9-5. Summary of n-decane / Sevilleta sand saturated zone
results.
Trial
1
2
3
4
5
6
Avg*.
porosity
(%)
34.1
33.8
32.8
32.6
34.0
33.8
33.5
error
(%)
0.6
0.6
0.6
0.6
0.5
0.5
0.6*
bulk
density
(g/cm3)
1.747
1.755
1.781
1.786
1.748
1.754
1.762
error
(g/cm3)
0.009
0.011
0.009
0.009
0.008
0.008
0.017*
Organic
maximum
62.8
79.6
75.4
78.2
78.2
78.6
75.5
liquid
error
8.1
11.1
9.8
10.0
9.7
9.7
6.4*
saturation
residual
24.3
25.7
21.3
20.1
26.5
21.9
23.3
(%)
error
4.0
5.2
3.9
3.6
4.2
3.7
2.6*
Temp.
Range
(°C)
2.6
1.1
1.5
2.1
1.3
0.6
—
Average = sample mean ±standard deviation
TABLE 9-6. Summary of p-xylene / Sevilleta sand saturated zone results.
- 145 -
-------
Trial
1
2
3
4
Avg.
porosity
(%)
32.5
33.0
32.2
32.4
32.5
error
(%)
0.6
0.6
0.6
0.6
0.3*
bulk
density
(g/cm3)
1.788
1.776
1.796
1.791
1.788
error
(g/cm3)
0.008
0.009
0.010
0.010
0.009*
Orgar
maximum
69.5
85.0
78.9
86.2
79.9
lie liquid
error
3.1
3.9
3.8
4.2
7.6 *
saturation
residual
27.5
25.7
26.1
28.8
27.0
;%)
error
1.4
1.4
1.5
1.6
1.4*
Temperature
range (°C)
0.5
1.7
0.9
1.3
...
TABLE 9-7.
* Average = sample mean ±standard deviation
Summary of PCE / Sevilleta sand saturated zone results.
Fluid
Soltrol
(13 trials)
Kerosene
(8 trials)
Gasoline
(5 trials)
n-decane
(5 trials)
p-xylene
(6 trials)
PCE
(4 trials)
all liquids
porosity
(%)
33.9 ± 0.6
33.6 ± 0.6
33.1 ± 0.8
33.6 ± 1.0
33.5 ± 0.6
32.5 ± 0.3
33.5 ± 0.8
bulk density
(g/cm3)
1.752 ±0.016
1.754 ±0.014
1.772 ±0.022
1.762 ±0.026
1.762 ±0.017
1.788 ±0.009
1.761 ±0.020
maximum organic
liquid saturation
(%)
85.1 ± 2.8
72.9 ± 9.0
81.0 ± 3.6
78.6 ± 4.2
75.5 ± 6.4
79.9 ± 7.6
79.5 ± 7.1
residual organic
liquid saturation
(%)
27.1 ± 1.7
26.3 ± 2.2
28.5 ± 1.7
25.9 ± 1.7
23.3 ± 2.6
27.0 ± 1.4
26.4 ± 2.4
Average = sample mean ± standard deviation
TABLE 9-8. Average values for different organic liquids in the Sevilleta sand saturated zone
experiments.
- 146 -
-------
correlation coefficient = 0.257
residual
organic
liquid
saturation
31-
29-
27-
25-
23-
21-
19-
5
+
A A * ^~ v *
X Soltrol-130
A kerosene
+ gasoline
^ p-xylene
• PCE
• n-decane
± x g A
A J4A X *x
* .A
A
^
*
0 55 60 65 70 75 80 85 9(
maximum organic liquid saturation (%)
FIGURE 9-25. Residual organic liquid saturation as a function of the maximum organic liquid
saturation.
experiments we avoided applying high suctions for fear of exceeding the breakthrough pressure of the
bottom, semi-permeable membrane. As a consequence, we were not always certain that the water had
been completely reduced to irreducible saturation and that the maximum organic saturation had been
achieved. A plot of maximum organic liquid saturation versus residual saturation for all the experiments is
shown in Figure 9-25. Visual inspection of the scatter plot does not reveal much correlation between
maximum organic saturation and residual saturation, but least-squares regression analysis reveals a
slight positive correlation.
Discussion of Experiments with Various Organic Liquids
The relationship between residual saturation and the type organic liquid is graphically displayed in
Figure 9-26. The average residual organic liquid saturation, 5or , was relatively uniform, varying from a
low value of 23.3% for p-xylene to a high value of 28.5% for gasoline, and similar to the value for Soltrol,
27.1%. The average residual saturation overall the fluids was 26.4%. All of the fluids were tested against
one another for a statistically significant difference in the sample means using the student's t-test.
Although the differences in residual saturations between the fluids appear to be small, the p-xylene
results were found to be statistically different from each of the other fluids at the 97.5% confidence
interval (a = 0.025).
P-xylene was the most soluble of the organic liquids tested ("0.20 g/l). In each experiment, we
flushed at least six pore volumes, or about 250 ml of water through the column to reach residual
saturation. If we assume the water was flooded slowly enough so that the effluent water had reached
equilibrium with the xylene, then as much as 0.05 g of xylene could have been removed from the column
by dissolution, resulting in an underestimation of the residual saturation by as much as 1.0%. The
reduction of xylene's residual saturation by dissolution may account for the significant difference in
residual saturation between it and the other fluids tested. The residual saturations of kerosene and
- 147 -
-------
30
25-
20
Percent
Residual 15
Saturation 1Q
Soltrol Kerosene Gasoline n-decane p-xylene PCE
FIGURE 9-26. Residual organic saturation for tested organic liquids in the Sevilleta sand.
gasoline were also found to have statistically different sample means. At the 95% confidence interval (a
= 0.05), the difference was just barely significant (the calculated value of t = 1.96 was greater than the
critical value of t = 1.78). This result may stem from some bias introduced by the poor temperature
control for the kerosene experiments. None of the remaining combinations showed significant
differences in residual saturations between fluids at the 97.5% confidence interval.
Influence of Interfacial Tens/on-
These same residual saturation results are again plotted in Figure 9-27 as a function of the interfacial
tension (IFT) between water and the organic liquid (see Section 4 for IFT data). There appears to be no
discernible correlation between IFT and So,. , just as we had hypothesized. The capillary forces
(proportional to the IFT) so outweigh the other forces which may act to reduce trapping (buoyancy and
viscous forces) that halving the interfacial tension — from an IFT of 47.8 for Soltrol to only 22.9 for
gasoline — has no effect on the amount of organic liquid trapped. Indeed, at typical aquifer flow rates
IFT's may have to be reduced by at least an order of magnitude or more before reaching the critical
capillary number needed to cause any reduction in trapping. This can be illustrated in the following
analysis.
Consider again the pore scale trapping mechanisms discussed earlier in this section. If we look at
these mechanisms in a quantitative way, using — as an example — the flow rates of these experiments
together with some observed characteristics of the Sevilleta sand, we can attempt some prediction of
the reduction of interfacial tensions required to begin to lower residual organic saturations.
Let's consider the snap-off mechanism first. Wardlaw (1982) performed flow visualization
experiments in glass micromodels using a single pore and pore throat to examine snap-off as a function
of several properties including interfacial tension. An illustration of the pore and pore throat pair used in
Wardlaw's experiments is given in Figure 9-5. For strongly water-wet systems, snap-off occurred for all
fluid pairs studied even though the interfacial tensions ranged from 480 dyne/cm down to 0.1 dyne/cm.
- 148 -
-------
30
28
Percent 26-
Residual
Saturation
24-1
22-
20
Soltrol
gasoline
PCE
kerosene
p-xylene
decane
20
25
30
35
40
45
50
Interfacial Tension
(dynes/cm)
FIGURE 9-27. Residual organic saturation as a function of interfacial tension (IFT). The error bars
represent the sample standard deviations taken from Table 9-8.
Trapping by snap-off was found to be insensitive to IFT down to 0.1 dyne/cm, and it is expected that IFT's
would have to be reduced still much further to avert the Maine's jump instability that leads to snap-off.
Although it appears that snap-off cannot be averted, perhaps for low enough IFT's the blob could
become mobilized immediately after the Maine's jump caused its formation. Oh and Slattery (1979) in a
theoretical study looked at the pressures required to mobilize blobs in a periodically constricting tube
(similar to that shown in Figure 9-4). The critical dimensionless pressure needed for mobilization in a
strongly water-wet pore with an aspect ratio of about four was:
AP r,
2 a
= 1.03
(9-12)
where: A.P = Pressure drop across the blob
r, - Radius of the pore throat
a = Interfacial tension
As an example, we looked at the interfacial tension needed to mobilize the singlet blob pictured in Figure
9-18a. With a flux rate in our column experiments of about 2.5 x 10~4 cm/s, we estimated the pressure
drop across the blob to have been about 3.7 dyne/cm2. The pore throat radius was estimated to be
about 50 microns and the aspect ratio to be 4. Using these estimates, the interfacial tension would need
to be less that 0.01 dyne/cm to induce mobilization of the blob.
- 149 -
-------
Now let's consider the by-passing mechanism. To prevent trapping in a pore doublet like the one
shown in Figure 9-8, the velocities of the interfaces traveling through each pore would have to be about
equal. That is, either the flow rate needs be increased or the capillary forces reduced (by reducing the
IFT) so that the water does not preferentially travel through the smaller pore of the doublet. To evaluate
this, we used the equation presented by Moore and Slobod (1956) for computing competing velocities in
a pore doublet:
4'G« + f .._. j_
Vj n r22
~ 4/G«~i r\ r (9-13)
V2
where: vf = Velocity through pore /
/ = Length of the pores
Q = Total flow rate through the doublet,
-------
PRIORITY POLLUTANT
carbon tetrachloride
PCE
benzene
chlorobenzene
ethylbenzene
toluene
phenol
o-chlorophenol
naphthalene
INTERRACIAL
TENSION
(dyne/cm)
45.0
47.5
35.0
37.4
38.4
36.1
39.3A(40°C)
42.25B
28.8B
TABLE 9-9
A — Lyman et al., 1982
B — Weast, 1986
The interfacial tension of some priority pollutants with water at 20°C. The data were
obtained from Girifalco and Good (1957) unless otherwise noted.
a decrease in S,^ as the density of the fluids approach one (as the density difference between the
organic and water approaches zero). This is the opposite effect one would expect had buoyancy played
a role. This is probably a coincidental effect of the reduced measurement accuracy as QO approaches
Qw
A larger density difference increases the buoyancy forces, but in this case we are well below the
threshold for gravity induced reduction of trapping. However, had we been testing a very coarse soil — a
gravel soil perhaps — we may not have been able to disregard buoyancy effects. PCE was the only
organic liquid tested having a density greater than water. Even so, the residual saturation measurements
do not appear to have been affected. The PCE experiments were run with the column upside down from
the other experiments to prevent any density induced instabilities. It is important to realize that on an
aquifer scale, as a dense organic liquid percolates downward through the saturated zone, density
induced fingering may well develop. This is most likely to occur in a layered soil, where a fine material
overlies a coarse material. It may be possible to exceed critical Bond numbers in this situation.
Summary-
Residual saturation has been shown to be invariant with respect to fluid properties, at least over the
range of fluid properties for common organic pollutants, and for the low capillary and Bond number
conditions of these experiments.
Implications for measurements of residual saturation-
Although it seems reasonable to directly measure residual saturations of whatever organic liquid was
spilled at a particular contamination site, it might be better to chose an ideal fluid with which to run
residual saturation experiments. Unless some odd wetting behavior is anticipated, or unless some
- 151 -
-------
30-i
28-
Percent 26
Residual
Saturation
24
22
20
gasoline
PCE
p-xylene
Soltrol
decane
kerosene
Dynamic viscosity
in cp
FIGURE 9-28. Residual organic saturation as a function of non-wetting phase viscosity. The error
bars represent the sample standard deviations taken from Table 9-8.
30-
Percent
Residual
Saturation
28-
26-
24-
22 J
20-
gasoline
T
Soltrol
kerosene
decane
p-xylene
0.6
I
0.8
I
1.0
I
1.2
PCE
1.4
I
1.6
Density (g/cm3)
FIGURE 9-29. Residual organic saturation as a function of non-wetting phase density. The error
bars represent the sample standard deviations taken from Table 9-8.
- 152 -
-------
interaction between fluids or between fluids and the solid is expected, it is probably preferable to chose a
fluid which has:
1. a sufficient density difference with water;
2. low solubility;
3. low volatility; and
4. low toxicity.
The easiest fluids for us to use were Soltrol and decane.
Implications for organic liquid movement-
Under low capillary and Bond number conditions the capillary trapping of different organic liquids
should be essentially the same. As we'll see below, the texture of the soil and its heterogeneity generally
have far more control on movement and trapping, than does the composition of the organic liquid.
RESIDUAL SATURATION FOR DIFFERENT SOILS
Different soils have different pore sizes and structures. If the structure of two soils is the same, say
two uniform glass beads of different diameters, then the capillary behavior of the soils will be identical
when scaled by capillary and Bond numbers. If the structure is different, then the soils may not be
scaled. For example, one soil structure may lead to capillary trapping of the non-wetting phase by
snap-off, while another soil may be dominated by a by-passing mechanism. We examined capillary
trapping in four soils: three sandy soils and a clay loam.
The three sandy soils that we investigated were of different depositional origin. The Sevilleta sand
was aeolian, the Traverse City sand was a beach deposit, and the Llano soil was a fluvial deposit.
Never-the-less, these soils were all fairly similar (see Section 4, especially Figure 4-5), and should have
had much the same type of pore network. Thus we hypothesized that they should show similar capillary
trapping behavior, and residual organic saturation levels. Recall that we expected "...Sevilleta sand to
exhibit a residual saturation between 15 and 25%, with a probable value of 18-20%." The actual Sor
value was 25 to 29%. We also expected an Sor value of 18 to 20% for the other two soils, under low
capillary and Bond number conditions. However, our experience with Soltrol had us doubting, this
hypothesis before the experiments. We were not sure of what to expect from the clay loam.
The Palouse clay loam was an agricultural soil with a much finer texture than the sandy soils. As
shown in Figure 4-11, we had great difficulty draining the water from the loam with Soltrol. At high
suctions the maximum Soltrol saturation reached was 11.7%, with a corresponding water saturation of
88.3%. At this point air broke into the column and we were not able to obtain an imbibition curve or
residual Soltrol saturation. From Figure 4-11 we subjectively estimated that the residual Soltrol
saturation for Palouse loam would have been less than 10%, but this is an unreliable estimate. The
saturations reported in Figure 4-11 may not have been at equilibrium.
The results of our experiments on Traverse City and Llano soils using Soltrol are shown in Tables
9-10 and 9-11. All experiments were conducted under good temperature control conditions. The
- 153 -
-------
Trial
1
2
3
4
Avg.
porosity
(%)
34.4
35.7
36.0
33.9
35.0
error
(%)
0.6
0.6
0.6
0.6
1.0*
bulk
density
(g/cm3)
1.738
1.705
1.696
1.753
1.723
error
(g/cm3)
0.009
0.010
0.009
0.010
0.027*
Organic
maximum
85.1
84.6
86.6
88.0
86.1
iquid s
error
7.3
6.8
6.9
7.2
«
1.5
aturation
residual
17.0
15.8
19.3
18.4
17.6
(%)
error
2.4
1.9
2.2
2.1
W
1.5
Temp.
range
(°C)
0.7
1.9
1.7
1.1
....
Average = sample mean ±standard deviation
Table 9-10 Summary of Soltrol / Traverse City soil saturated zone results.
Trial
1
2
3
4
5
Avg.
porosity
(%)
37.2
37.5
37.5
38.1
37.2
37.5
error
(%)
0.6
0.6
0.5
0.5
0.6
0.4*
bulk
density
(g/cm3)
1.665
1.656
1.656
1.640
1.664
1.656
error
(g/cm3)
0.009
0.009
0.008
0.008
0.009
0.010*
Organic 1
maximum
90.2
91.9
92.2
88.0
90.2
90.5
iquid s
error
3.7
3.8
3.6
3.4
3.7
*
1.5
aturation
residual
18.5
15.8
14.4
13.9
16.5
15.8
(%)
error
1.4
1.3
1.2
1.1
1.3
1.8*
Temp.
range
(°C)
1.6
1.3
1.8
1.4
1.7
....
* Average = sample mean ±standard deviation
Table 9-11. Summary of Soltrol / Llano soil saturated zone results.
- 154 -
-------
average residual saturation of Soltrol in the Traverse City soil was 17.6%, while it was 15.8% in the Llano
soil. These saturations were slightly lower than the range that we had initially hypothesized for this type of
soil (18-20%) and almost 10% lower than the residual saturation found in the Sevilleta soil. As shown in
Figure 9-30 and Table 9-12, this 10% difference is significant. In contrast, the minor S^ difference
between the Traverse City and Llano soils is not significant.
Discussion of Experiments with Different Soils
We hypothesized that the three sandy soils would behave more like glass beads than sandstone. The
evidence supports this hypothesis for the Traverse City and Llano soils. The observed residual Soltrol
saturations of 14-19% were in the low end of the range of 18-20% that we had anticipated, and only
slightly higher than the values observed for uniform glass beads.
A lower mean S^ value was observed for the Llano (15.8%) than for the Traverse City (17.6%). This
difference is not statistically significant. We would expect the less uniform soil (Llano) to experience
more trapping through by-passing and thus to have a higher residual saturation. This was not the case in
these experiments. The coarser material (Llano) would have had a higher capillary number during the
imbibition (water flooding) phase of the experiment because of its larger k value. Thus it is possible that
the capillary number was too high during the water flood, and less Soltrol was originally trapped. The data
qualitatively support this possibility, but not statistically. Our estimates for capillary numbers in these
experiments are well below the threshold for this condition.
The grain size curves for the Sevilleta and Traverse City soils are almost identical, while the Llano
curve indicates a coarser and less uniform material (see Figure 4-5). Why are residual saturations so
different between two almost identical sands (Sevilleta and Traverse City) and so much alike in sands
with much less similar grain sizes and distributions. Perhaps it is more appropriate to ask, 'Why are
residuals in the Sevilleta sand so high?'. After all, one would expect a uniform, unstructured sand like the
Sevilleta to behave in a manner similar to glass beads — as the Traverse City and Llano soils presumably
did.
We hypothesize that while packing the columns with Sevilleta soil some small-scale layering
developed. The Sevilleta sand contains a small but not negligible fine clay fraction not present in the
Traverse City or Llano sands. This fine fraction is not readily observed in Figure 4-5, which reports the
results of a dry sieve analysis. The clay and silt content may have been 2% by weight and could have
played an important role in the formation of fine layers in the column. We hypothesize that a fine layer
settled at the top of each lift of the packed column. The thickness of the layers was presumably variable,
depending on the time between lifts, and whether the packing process was interrupted. We did not
actually observe layering, either in the sides of the short glass columns or in the pore casts
(unfortunately, we did not cut the pore casts longitudinally, which might have shown the layers if they
were present). The layers would have caused the column to be heterogeneous. The effect of
heterogeneities on multi-phase flow are explored in detail later in this chapter. In effect,
heterogeneousness can dramatically increase residual saturation. Unfortunately, we developed this
hypothesis too late in the project to be able to test it. None of the heterogeneities explored later
correspond to this layering hypothesis. We recommend that future research efforts explore this issue
- 155 -
-------
Fluid
Sevilleta
(13 trials)
Llano
(4 trials)
Traverse City
(6 trials)
all sands
porosity
(%)
33.9 ± 0.6
37.5 ± 0.4
35.0 ± 1.0
35.5 ± 1.8
bulk density
(g/cm3)
1.752 ±0.016
1.656 ±0.010
1.723 ±0.027
1.71 ±0.05
maximum organic
liquid saturation
(%)
85.1 ± 2.8
90.5 ± 1.5
86.1 ± 1.5
87.2 ± 2.9
residual organic
liquid saturation
(%)
27.1 ± 1.7
15.8 ± 1.8
17.6 ± 1.5
20.2 ± 6.1
Average - sample mean ^standard deviation
TABLE 9-12. Average values of measured properties and saturations for
different sandy soils, in the saturated zone experiments
run with Soltrol.
30
25-
20
Percent
Residual 15
Saturation
ID-
S'
Soltrol
Sevilleta Traverse City
Llano
Palouse
FIGURE 9-30. Residual organic saturation for Soltrol in tested soils. The error bars represent
sample standard deviations, except for the Palouse loam, where the error bar
is a subjective indication of the poor quality of that experiment.
- 156 -
-------
Percent
Residual
Saturation
ou-
25-
20-
15-
10-
5-
0-
O.C
T Soltrol
4 Sevilleta
Traverse City • j
f Llano
? <
1 Panoche
)01 0.010 0.100 1 1C
Organic Carbon Content (%)
FIGURE 9-31. Residual organic saturation for Soltrol, as a function of organic carbon content in
different soils. The error bars represent the sample standard deviations taken from
Table 9-12. The organic carbon contents are of unknown uncertainty.
further. Another hypothesis is that clay swelling obstructed pore throats, and increased the propensity
for by-passing. This hypothesis alos arose too late in the project to allow the appropriate clay mineralogy
tests.
Figures 9-31 through 9-33 explore the relationship of observed residual Soltrol saturation and
various soil properties: organic carbon content, porosity and permeability. Organic content may effect
the wetting of the soil and thus would influence the trapping mechanisms. All the soils tested were
strongly water wet. Thus we expect that the organic material was aggregated in a few locations within the
pore space, rather than spread thinly over the mineral surfaces. As shown in Figure 9-31 the soil organic
carbon content did not appear to have any systematic influence on capillary trapping. However, these
experiments were not designed to elucidate such an influence. Porosity is a measure of pore structure,
and thus may show a correlation to residual organic liquid saturation. Figure 9-32 indicates that the lower
porosity material had the higher residual saturation. Presumably, the by-passing mechanism is more
common in the lower porosity material (see, eg., Chatzis et a/., 1983; Pathak ef a/., 1982). We are not
convinced that the trend shown in the figure is reliable. For example, the Sevilleta may have been packed
with layers, whereas we do not believe layering occurred for the other two soils. The Llano sand had the
least uniform grain distribution, but the highest porosity. This is not what one would expect for consistent
packings from material to material. The correlation of porosity to residual saturation needs to be
explored in future research. The influence of permeability on residual Soltrol saturation is shown in
- 157 -
-------
Percent
Residual
Saturation
30
25-
20
15
10-
5-
i
Sevilleta
i
Traverse City
Palouse not shown
32
33
34
35
Porosity
36
Soltrol
h
L|ano
37
38
FIGURE 9-32. Residual organic saturation for Soltrol, as a function of porosity in different soils. The
error bars represent the sample standard deviations taken from Table 9-12.
30
25-
Percent 20~
Residual
Saturation 15
10
Traverse City
T Soltrol
9 Sevilleta
1 Llano 4
Palouse not shown,
at 5x10-6cm/s
0.001
0.010
0.100
Hydraulic Conductivity
in cm/sec
FIGURE 9-33. Residual organic saturation for Soltrol, as a function of water saturated hydraulic
conductivity in different soils. The error bars represent the sample standard
deviations taken from Table 9-12.
- 158 -
-------
Figure 9-33 (as represented by water saturated hydraulic conductivity). If the waterflood was conducted
under low capillary and Bond number conditions then we would expect that permeability had no
influence. That appears to be the case.
Implications for measurements of residual saturations-
Even though two soils have very similar texture, their residual organic saturations may be
significantly different if they have different structures. When using packed columns to measure Sor in
the laboratory, it is appropriate to pack the columns with the same soil structure found in the field. If this
is impossible, undisturbed field samples are preferred. Looking at it another way, eliminating
heterogeneities from laboratory soil packings does not necessarily provide a good model of behavior in
the field, where similar heterogeneities are produced by natural depositional processes.
Implications for organic liquid movement and trapping-
Our inability to inject an organic phase into the Palouse loam demonstrates that fine-grained,
water-wet soils (which do not shrink in the presence of organics) can serve as an effective barrier to
organic liquid movement in the subsurface.
Implications for modeling organic liquid movement-
Models of organic liquid movement employ prescribed values of residual organic liquid saturation as
a soil property. It would be convenient to be able to estimate residual saturations from more primitive soil
properties, such as grain size distribution (see, eg, Soil and Celia, 1988). These experiments indicate
that textural considerations alone may lead to unreliable estimates and erroneous models.
INFLUENCE OF THE INITIAL RATE OF ORGANIC LIQUID INVASION
Most models of multiphase flow presume that the residual wetting and non-wetting phase saturations
are fixed (non-functional) soil properties that can be measured and prescribed (see, eg, Lenhard and
Parker, 1987a, 1987b, 1988, 1989). If that is the case, then these residual saturations should be
independent of the rate of fluid flow. We investigated this hypothesis using the homogeneous
micromodel pictured in Figure 8-1 and formerly used in the experiments depicted in Figures 9-11
through 9-13. As you'll see, the hypothesis and the common conceptual model it supports was
inappropriate for the modeled conditions.
Residual Water Saturation
The homogeneous micromodel was rerun at a 'fast' flow rate of 1.5 ml/min, approximately 15 times
higher than that for the experiment depicted in the upper photo of Figure 9-11. The steady state
condition at the end of the Soltrol invasion is shown in the upper photo of Figure 9-34a. Close-ups are
shown in Wilson ef a/. (1988). The Soltrol displacement of water was found to be slightly more efficient
than before (compare to Figure 9-11 a), with a decreased residual water saturation. The faster
displacement's larger viscous forces partially overcame capillary forces resulting in fewer by-passed
pockets of water and possibly less water held in wedges. The capillary end effects were also largely
overcome.
- 159 -
-------
a.
b.
FIGURE 9-34. Homogeneous model. In the upper photo (a) Soltrol displaced water from
the left (the top of the model) to the right (the bottom of the model), at 1.5
ml/min yielding a residual (irreducible) wetting phase saturation. In the
lower photo (b) Soltrol was displaced by water from the right (the bottom
of the model) to the left (the top), also at 1.5 ml/min, yielding a residual
non-wetting residual saturation. Soltrol was dyed red and appears dark
grey; the water was not dyed. The photos record steady state flow
conditions at the end of the displacements.
- 160 -
-------
This experiment illustrates the misleading nature of the term 'irreducible water saturation' often used
to represent the wetting phase residual saturation. Even at residual saturation the wetting phase is
continuous and is composed of an interconnected network of films, rings, and wedges (see, eg.,
Amaefule and Handy, 1982; Dullien, et a/., 1986; Chatzis, et a/., 1988) This is a considerably different
situation than the discontinuous 'blobs' pictured in this report for the non-wetting phase residual. The
wetting phase liquid can move through its interconnected network, draining the films and rings, and
reducing the residual wetting phase saturation. Reducible wetting phase residual saturation is seldom
considered in agriculturally oriented soil physics where the non-wetting phase, air, is usually assumed to
be at static equilibrium. In petroleum geology and reservoir engineering, where the non-wetting phase is
crude oil, it can be important although it is usually ignored.
Implications to fluid flow model/ng-
Several recent experiments provide some idea of the correlation between non-wetting phase flow
and wetting phase residual for reservoir rocks. The wetting phase residual appears to be a continuous
function of non-wetting phase flow rate, although the flow rate apparently must vary over orders of
magnitude before the change in residual is significant (Handy, personal communication, 1988; also see
earlier work of Amaefule and Handy, 1982, and Dullien, et a/., 1986). The short column experiments
described earlier in this report were not designed to investigate this correlation quantitatively.The
micromodel experiments shown in this report and in Dullien ef a/. (1986) provide ample visual evidence
that an increase of the non-wetting phase flow rate leads to a decrease of the wetting phase residual.
Further research is needed to determine this relationship, its importance to fluid flow, and its relevance
to conditions encountered in the field. In particular, most models of pore pressure-saturation assume a
fixed and known 'irreducible water saturation'.
Residual Organic Liquid Saturation
The lower photo of Figure 9-34b depicts the residual Soltrol saturation at the completion of a upward
displacement by water at the same 'fast' rate of 1.5 ml/min. The residual Soltrol saturation looks similar
to that observed after the slow experiment (Figure 9-11 b), despite the larger amount of organic liquid in
storage before the water displacement (compare Figures 9-11a and 9-34a). One possible explanation
for this increased efficiency is that there were larger viscous forces involved in the fast displacement.
Perhaps a more likely cause is the spatial distribution of residual water and organic liquid in storage after
the organic liquid advance. In the slow rate experiment the relatively large by-passed water zones
remaining after the advance (see Figure 9-34a) provided pathways for the upwardly advancing water to
by-pass large zones of organic liquid.
Implications for organic liquid movement, trapping, and modeling-
The amount of capillary trapped organic liquid, and therefore the efficiency of hydraulic aquifer
remediation activities, may depend strongly on the rate at which the organic liquid pollutant originally
invaded the aquifer.
- 161 -
-------
INFLUENCE OF WATER FLOW RATE ON RESIDUAL ORGANIC MOBILIZATION
Earlier in this section we reviewed published evidence that sufficiently high rates of flow in the wetting
phase can reduce the amount of capillary trapped non-wetting phase (eg., Melrose and Brandner, 1974;
Ngefa/., 1978; Chatzis and Morrow, 1981; Chatzis et a/., 1984; Hornof and Morrow, 1988; Morrow et a/.
1988). This mobilization of the residual non-wetting phase can be correlated to capillary number (see,
eg, Figures 9-9 and 9-10) and Bond number (Morrow and Songkran, 1981 ;Morrow ef a/. 1988). Aquifer
remediation schemes may depend in whole, or in part, on the mobilization of the trapped organic liquid
by increasing the effective capillary number (Wilson and Conrad, 1984). Mobilization correlations exist
for sandstones (Figure 9-9) and glass beads (Figure 9-10), but not for natural soils. We expect that the
correlation depends on the soil pore structure, and that it will vary from soil to soil. Understanding this
natural soil correlation would help define under what conditions it is possible to hydraulically mobilize
trapped organic liquids — and when it is not possible. It would also help prevent inadvertent mobilization
of organic liquids during laboratory experiments.
We investigated the mobilization of organic liquids in natural soils by mobilizing capillary trapped
Soltrol in the Sevilleta and Llano soils. We hypothesized that, because soil texture considerations (see
the grain size curves, Figure 4-5 ), the mobilization correlations for the Sevilleta and Llano soils should
look more like that of glass beads than like that of sandstones (see Figures 9-9 and 9-10). Given our
experience with low capillary number residual saturations in these soils (see Table 9-12), we also
expected that the Llano soil correlation would be the closest to the glass bead correlation.
The long column apparatus and procedure used in these experiments are described in Section 6.
One experiment was performed on each soil. The soil packed columns were water saturated, drained
with Soltrol, and then flooded with water under fairly low capillary number conditions. Once an initial
residual Soltrol saturation was established the column flow rate was increased incrementally, and the
residual saturation was re-measured at each step. Mobilization was indicated by a lower residual
saturation. Table 9-13 and Figure 9-35 summarize the initial conditions for the mobilization portion of
the experiments. The initial residual Soltrol saturations were within the range measured in the short
column experiments.The mobilization results are given in Figure 9-36. We were not able to effectively
mobilize a significant portion of the residual Soltrol saturation in either column due to limited pump
capacity (see Section 6).
Figure 9-36 contrasts our mobilization results with the published correlations for sandstones and
uniform glass beads. The Llano data follows the glass bead correlation up to the capillary number tested.
The Sevilleta soil data deviates from the uniform glass bead curve at a critical capillary number,
N*c2 = q*fiw/a, of 8 x 10~5 . This critical capillary number is about 40 times higher than the value
observed for sandstone (see the figure), but perhaps about one fourth of the value for glass beads. In
fact, the Sevilleta data appears to follow the pattern observed by Morrow ef a/. (1988) for sintered glass
beads (beads fused partially together in a furnace).
The observed data is consistent with our hypothesis. Both soils behaved much like glass beads, with
the Llano soil being the closest. The hypothesis is not confirmed, however, because of a lack of data at
higher capillary numbers
- 162 -
-------
Soil
Sevilleta
(1 trial)
Llano
(1 trial)
Notes:
porosity
(%)
37.2 (33.9)
37.3 (37.5)
bulk density
(g/cm3)
1.66 (1.752)
1.67 (1.656)
water saturated
hydraulic
conductivity (cm/s)
1.9 (1.0) x ID'2
17 (16) X 10'1
Apparently, the Sevilleta column was not
packed at maximum density.
maximum organic
liquid saturation
(%)
72.8 (85.1)
81.1 (90.5)
residual organic
liquid saturation
(%)
26.0 (27.1)
17.2 (15.8)
Drainage with Soltrol was halted
before the maximum Soltrol sat-
uration had fully stabilized.
TABLE 9-13. Long column data for two different sandy soils run with
Soltrol. The numbers in parentheses refer to the average
values for the short column experiments (taken from
Table 9-12).
30-i
25-
20
Percent
Residual 15_
Saturation
10
5-
short
columns
Sevilleta
long
column
Llano
Soltrol
FIGURE 9-35. Residual organic saturation, in short and long columns, for Soltrol in the
tested soils. The error bars represent sample standard deviations for the short
column experiments.
- 163 -
-------
0.8-
Reduced °-6 ~
Residual
Saturation
c /a- 0.4 -
0.2-
typical
correlation for
uniform
un-sintered
glass beads
typical
sandstone
correlation
Sevilleta sand, Sor = 26.0 %
Llano sand, 5» = 17.2 %
Sintered Glass Beads,
Morrow et al., 1988
0-J
io-6
-i-' '"I
io-5 io-4
Capillary Number,
FIGURE 9-36 Relationship between Soltrol residual saturation and capillary number, Nl , for
Sevilleta sand and Llano sand. Its compared to sandstone data from Chatzis and
Morrow (1981,1984) and bead-pack data from Morrow and Chatzis (1982), and
Morrow et al. (1988).
1 -
Reduced
Residual
Saturation
0.8-
0.6-
0.4-
0.2-
correlation for original \
trapping of continuous \
organic phase in glass \
beads ^
• Sevilleta sand, 5^ = 26.0 %
+ Llano sand, S'or = 17.2 %
correlation for
mobilization of
trapped blobs
in glass beads
10
-6
i ' ' ' ''"|
IO-5 IO-4 10~3
Capillary Number, N* =
10
,-2
a
FIGURE 9-37. Relationship between Soltrol residual saturation and capillary number, N?, for
Sevilleta sand and Llano sand. Its compared to correlations for un-sintered, uniform
glass beads developed by Morrow and Chatzis (1982), and Morrow et al. (1988).
- 164 -
-------
We also hypothesized that the initial residual saturations might be artificially low because these
saturations were established at fairly high initial capillary numbers (@ N? = 9 x 10~* and 9 x 10"5,
respectively, for the Sevilleta and Llano sands). If this were true, then we would expect the initial residual
saturations to be lower than the values observed for the low capillary number short columns. This was not
the case, as shown in Table 9-13. A weaker form of evidence is the comparison to the glass bead
correlation for initial trapping at elevated capillary numbers, which is shown as the dashed curve in Figure
9-37. We see no evidence in this comparison for lower initial saturations. We have no reason to believe
that the initial saturations were artificially low.
In summary, it appears that hydraulic mobilization in these two fairly uniform, unconsolidated sands
may be much like that of glass beads. Glass beads have a large critical capillary number, making
mobilization very difficult.
Implications for hydraulic removal of residual saturation —
Several schemes have been published in the literature and implemented in the field for hydraulically
sweeping organic liquids from polluted aquifers. These schemes are presumably meant to sweep out the
continuous organic liquid, knowingly leaving behind the residual. More often it seems that naivete
prevails, and many designers assume that as long as ground water is flowing toward a collection system,
eventually all of the organic liquid will make it. No matter how long one waits, unless gradients are
increased above the critical level, none of the residual will be hydraulically removed (Wilson and Conrad,
1984).
Remediation schemes attempting to hydraulically remove capillary trapped residual by waterflooding
can remove some of residual organic liquid whenever the capillary number exceeds the critical value:
Nc > N*c (see Figure 9-9). All of it will be removed when Nc > N"c". Figures 9-38a and 9-38b are plots of
water phase hydraulic gradient, J, necessary to initiate mobilization in a soil of intrinsic permeability, k.
The curves are taken from Wilson and Conrad (1984), who based them on the sandstone N\ capillary
number correlation of Figure 9-9 and the sandstone value of the critical capillary number,
N'cl =2 x 10~5 . Various organic-water interfacial tensions, o , are shown.
For example, PCE has a density of 1.62 g/cm, greater than that of water, and is likely to sink in the
saturated zone. It has an interfacial tension of 47.5 dyne/cm. In a fine gravel, with k = 10~5 cm2, Wilson
and Conrad (1984) estimated that a gradient of J = 0.1 is necessary to start some of the blobs moving.
This is a steep but not unreasonable gradient. In a medium sand with k = 1(T7 cm2, the necessary
gradient is a very steep 7=10. Figure 9-39 is their plot of percentage residual organic liquid recovered
as a function of gradient and soil permeability for an interfacial tension of 10 dyne/cm. The top curve
there and in Figure 9-38a is the 100% recovery curve for the same organic liquid. Other organic liquids
can be examined by simply multiplying these results by cr/10. For the fine gravel example, which
required a gradient of only 0.1 to begin to mobilize PCE blobs, Wilson and Conrad (1984) estimated that
the gradient must be increased to 47.5 times as large, or J = 4,75, in order to 'get it all'. Such a large
gradient is improbable at a remediation site, but is not uncommon in a laboratory experiment. In the
medium sand (where k is 100 times smaller) it would take a gradient of J = 100 x 4.75 = 475 to remove all
of the residual PCE, clearly an impossible feat at a remediation site.
- 165 -
-------
(a)
irf
KT1
N** = 1.3 x 1(T3 "
. NC = 2 x 10
l mud I mud
l l linift i i inn
US' 10* Itf* Iff4 KT* Kf' Kf XT
k (cm2)
0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0
k (cm2) X 106
FIGURE 9-38. Hydraulic gradients required to initiate blob mobilization in porous media of various
permeabilities, for organic liquids of various interfacial tensions. A critical capillary
number of NC = 2 x 10~5 was used. Plot (a) uses a log-log scale, while (b) is plotted
using linear coordinates. The upper curve in (a) represents the gradient necessary
for complete removal of all organic liquid with an interfacial tension of 10 dynes/cm.
Constructed from sandstone data presented in Figure 9-9. From Wilson and Conrad
(1984).
Even these predictions are optimistic. We've found that the critical capillary number for two sandy
soils is forty to a hundred times higher than the sandstone values used in these figures and calculations.
If we take the lower value of forty, then it would take a gradient of J = 40 x 0.1 =4 just to begin to mobilize
the PCE from the fine gravel, and a gradient of 19 to 'get it all'. Clearly, it is not practical to mobilize
residual organic liquid in the saturated zone using hydraulic means.
In summary, residual organic liquid is easier to mobilize for lower interfacial tensions, higher
permeability soils, and in the lab. It's harder to mobilize for higher interfacial tensions, lower
permeabilities, and in the field where there are severe practical constraints on the hydraulic gradient.
Although it may be possible, intentionally or by accident, to mobilize some of the residual, it is difficult or
impossible to get it all.
Implications for remediation using surfactants and bioremediation-
The only practical way to improve mobilization of the residual will be to lower the interfacial tension
using surfactants. Surfactants may already be present at a contamination site, either as part of the
disposed waste or spilled fuel or as a by-product of biological growth. Surfactants can also be injected,
although there is some difficulty getting the injected surfactant to actually make contact with the blobs.
- 166 -
-------
FIGURE 9-39. Recovery of residual saturation as a function of permeability and hydraulic gradient
for an interfacial tension of 10 dyne/cm. Based on sandstone data. From Wilson and
Conrad (1984).
Finally, the biological community can be stimulated to consume some of the dissolved organic,
producing more bio-surfactant as a by-product.
Even if surfactants reduce the interfacial tension a factor of ten or more, we cannot expect
dramatically increased mobilization. We believe that, in the saturated zone, surfactant concentrations
have to be high enough to emulsify the organic liquid in order for them to have a significant impact.
Implications for improving predictions of transport and fate —
Accurate measures of relative permeabilities at reduced residual saturations will be important in
reliably predicting pollutant movement. As the organic phase dissolves, its saturation becomes reduced,
and the permeability to water increases. Increases in permeability result in either increased water flow
under constant head boundary conditions or increased residence time under constant flux conditions. In
either case, an accelerated rate of dissolution is the ultimate result. Having good estimates of the
relative permeabilities may be as important as good estimates of the mass transfer coefficients for
predicting pollutant migration.
- 167 -
-------
ORGANIC LIQUID MOVEMENT AND CAPILLARY TRAPPING
IN A HETEROGENEOUS POROUS MEDIA
No soil is truly homogeneous. Soils exhibit spatially variable properties that are a result of their
original deposition and subsequent diagenesis. The importance of this variability on the flow of fluids and
the transport of contaminants has only recently been accepted by most groundwater hydrologists.
Gelhar (1986), Gutjahr (personal communication, 1988), and Dagan (1986) have reviewed the literature
on this topic. Conspicuously absent from these reviews is any discussion of multiphase flow, other than
water percolation in a soil with air at a uniform pressure. This is the so-called Richard's equation
approach to two phase flow. Following this approach Yeh ef a/. (1985a,b,c), Mantoglou and Gelhar
(1987), and Abobou ef a/. (1988) have developed geostatistical theoretical and computer simulation
models that demonstrate the importance of soil stratification on moisture movement in the vadose zone.
At high capillary pressures the finer textured layers and lenses of soil absorb water through capillary
forces, while the coarse layers remain relatively dry and impervious. The net result is a hypothesis that
the anisotropy of water permeability is a function of saturation. At low water saturations the fine material
is more permeable to water than the coarse material, and water has a preference for horizontal flow
along the beds due to capillary forces. At high saturations the coarse material has the greater water
permeability, and gravity forces tend to cause vertical flow. This hypothesis has been verified in the field
and the laboratory by Stephens and Heermann, 1988; Mattson era/., 1988; McCord ef a/., 1988a,b;and
others. A similar observation was made by Schwille (1988) using a hydrocarbon as the wetting phase.
Petroleum reservoir engineers have largely ignored spatial variability, although that has started to
change. The 1989 SPE Reservoir Simulation Conference had several sessions devoted to the topic,
which reservoir engineers refer to as 'reservoir heterogeneity' (Lake and Carroll, 1989; also Moissis, ef
a/., 1989; Ewing, etal., 1989). Most of this petroleum related work is focused on miscible displacement
involving the advection and dispersion of chemical components in a single non-homogeneous fluid
phase. Earlier, Chatzis ef a/. (1983) briefly examined the effect of reservoir heterogeneity on immiscible
displacement at the pore network level using heterogeneous glass bead packs. They observed that
during water floods the spatial variation of capillary properties can lead to large scale by-passing of
oil-filled, coarse bead zones. This observation is consistent with the vadose zone water infiltration
studies showing air by-passed in the coarse lenses and layers.
We hypothesized that the spatial variation or heterogeneity of soil properties has a dominant effect
on the movement and capillary trapping of organic liquids . We tested this hypothesis using flow
visualization of two phase flow in micromodels, representing three different types of spatial variability.
The character of these displacements was contrasted with homogeneous displacements. The first type
of heterogeneity concerned the boundary of a fine material with a coarse material. Fluid flow was from
one to the other. This type of heterogeneity was simulated as a by-product of our homogeneous
micromodel experiments. It was represented by the boundary between the pores and the end reservoirs
in the micromodels. The second heterogeneity type was classified as an 'aggregated soil', with strings
of interconnected macropores penetrating a homogeneous matrix of micropores. The macropores can
also be viewed conceptually as fractures through a consolidated porous rock. The micropores represent
the porous matrix of that rock. For example, Figure 8-12 is an illustration of the micromodel pattern used
- 168 -
-------
to represent an aggregated soil. The pore network was identical to that of the homogeneous model
presented earlier(Figure 8-1), with the pores that had been selected to be macropores simply enlarged
in size.The third type of variability that we investigated concerned stringers or lenses of coarse porous
material embedded in a matrix of fine porous material. A geological example of this heterogeneity is a
fluvial sediment, where gravel lenses are buried in a sand matrix. Figure 8-13 is an illustration of the
micromodel pattern used to represent this type of variability. Once again, the pore network was identical
to the homogeneous model, but the pores in the coarse zones were enlarged in size. The aggregated
media and stringer heterogeneous micromodels had bi-modal pore size distributions, with one peak
associated with the coarse lenses or macropores, and the second peak associated with the fine matrix or
micropores. The aggregated soil heterogeneity was simulated by the micromodel only; there were no
short column experiments. The coarse lenses were simulated in micromodels and in short glass and TFE
columns filled with Sevilleta soil, yielding both pore casts of the trapping in a heterogeneously packed
sand column, and quantitative measurements of altered trapping. A new theoretical model was
developed to explain the observed behavior. The theoretical model is based on the interplay between
viscous and capillary forces in a conceptually simple heterogeneous media.
In each laboratory experiment the heterogeneous sample was first saturated with water and then
subjected to flooding with the organic liquid (Soltrol or styrene). After the experiment equilibrated, the
organic liquid was displaced by water, resulting in a residual organic liquid saturation that depended on
heterogeneity and fluid flow rates.
Multiphase Movement Across an Interface Between Two Porous Media
Consider an organic liquid displacing water in a water wet porous media. If the displacement
advances toward an interface of two different porous media, the behavior of the fluids at the interface will
depend on capillarity. Suppose that the displacement is from a coarser material toward a finer material.
If the capillary pressure does not exceed the organic liquid 'entry value' into the fine matrix, the
displacement will cease (see sketch in Figure 9-40a). This is the principle behind the fine pore sized
capillary barriers at the end of our short column experiments and some of our micromodels (see, eg,
Figure 8-9). These experiments seek uniform equilibrium conditions in the coarser material upstream of
the interface. If, on the other hand, the displacement is from a finer material into a coarser material, the
behavior is quite different. The result is a capillary end effect (Figure 9-40b), such as that demonstrated
by the homogeneous micromodel shown earlier in Figure 9-11a.
The capillary end effect is what one would find in a fine sand layer overlying a coarse sand or gravel
layer. In Figure 9-11 a the micromodel pore network represents the upper fine sand layer. The bottom
model reservoir mimics the lower gravel layer. Once the organic liquid breaks though to the lower layer
along one, or perhaps two flow paths, the remaining organic liquid front retreats slightly. This leaves a
few capillary-trapped blobs in the zone at the bottom of the model (right side of Figure 9-11 a). The
continuing flow of organic liquid takes place through the one or two connected flow paths which offer less
resistance than the capillary forces in the many remaining water-saturated pores (Figures 9-11 a and
9-40b).
Had a capillary barrier been used at the bottom boundary of the micromodel, to reduce the capillary
end effect by preventing the organic phase from passing out, we believe that the residual water
- 169 -
-------
saturation seen in Figure 9-11a would have been much smaller. Less water would have been trapped in
the by-passed pockets and the wedges. This hypothesis has not yet been tested. Recently Dullien et
al.(1986) built a heterogeneous micromodel with a checkered pattern of coarse and fine pore zones.
They embedded the outlets in two of the fine zones, in effect simulating a capillary barrier at the outlet of
the model. Their micromodel experiments, and related sandstone core experiments, appear to confirm
the hypothesis that residual wetting phase saturation in the upstream material depends on the relative
texture of the downstream material.
Implications for organic liquid migration & its model ing-
The capillary end effect in the micromodel illustrates the significant impact that media heterogeneity
can have on the migration of an organic liquid phase. This end effect is not desired in some laboratory
experiments which seek uniform equilibrium conditions. However, in nature conditions are not uniform,
and the choice of a micromodel (or numerical model) boundary condition depends on what the operator
desires to simulate. The use of a capillary barrier boundary, as in the soil columns, may be arbitrary. In
effect, the freely draining condition of the micromodel experiment illustrated in Figure 9-11 a could be
more representative of conditions in aquifers. If there is no barrier to organic phase migration, and the
coarse
flow stoppe
fine (eg, silt)
flow stopped
coarse (eg, gravel)
flow continues
FIGURE 9-40. Non-wetting fluid near a material boundary: a) moving from coarse to fine and
encountering a capillary barrier; and b) moving from fine to coarse and
encountering a 'capillary end effect' resulting in rivulets of non-wetting flow
across the boundary
- 170 -
-------
organic phase pressures cannot build-up, then the by-passed pockets of water constituting a major
portion of the residual water saturation will not drain. In effect this hypothesis suggests that the residual
water saturation is not a single 'irreducible' value, but depends on conditions. This is the second reason
why we question the use of a single irreducible water saturation value for a numerical model or simulator.
Multiphase Movement in an Aggregated Porous Media (Wilson et al., 1988)
Multiphase flow in an aggregated porous media presents a different picture. Here, the coarse
material is represented by the interconnected macropores. The fine material or micropore matrix is
isolated between the macropores, as illustrated in Figure 8-12 for the micromodel.
The drainage of a water filled, and water wet aggregated micromodel media is shown in Figures 9-41
through 9-43. Soltrol invaded vertically downward into the model at a 'slow rate' of 0.075 ml/min. The
experiment was then repeated with a 'fast rate' of 1.5 ml/min, as shown in Figure 9-44. The model was
strongly water wet. Figures 9-41 a, 9-42a, and 9- 43a (upper photo in the figures) depict the steady state
conditions in the model at the end of the slower organic liquid invasion. The close-ups are focused on the
same area depicted earlier for the homogeneous model (Figures 9-11 through 9-13). The steady state
condition at the end of the fast displacement is shown in Figure 9-44a. This experiment is also available
on videotape (Mason, et al., 1988; see Appendix B).
The steady state saturations are similar in character for both rates. Capillary forces caused the
organic liquid to preferentially travel through the strings of macropores, almost completely by-passing
the water filled micropores. Because very little organic liquid entered the micropore matrix, the organic
liquid traveled across the model much more quickly than it did for the same injection rate in the
homogeneous model. This by-passing of the water within the aggregates resulted in a much higher
residual water saturation and a lower maximum organic liquid saturation. Under 'slow' conditions
essentially all of the micropores were by-passed, while 'fast' flow conditions permitted some organic
liquid penetration of the micropores due to the capillary pressures generated by significantly larger
viscous forces. The 'fast' flow rate's larger viscous forces also account for reduction in capillary end
effects at the lower end of the model.
The lower photos in Figures 9-41 b to 9-42b, and Figure 9-43b, depict the residual organic liquid left
behind after 'slow' and 'fast' upward water floods, respectively. In each case, the residual non-wetting
saturation largely consisted of by-passed strings of organic liquid left behind in the macropores. Because
very little organic liquid initially penetrated into to the aggregates, very little organic liquid was
subsequently trapped in them, resulting in much smaller residual organic liquid saturations than were
found in the homogeneous model. Perhaps coincidentally, the sweep efficiency (the amount of organic
liquid recovered divided by the amount of organic liquid originally in place) was actually about the same
for the two models. Displacement of organic liquid from the aggregated micromodel was no more
efficient than displacement of organic liquid from the homogeneous model, but the observed residual
saturations in the aggregated model were much lower because the amount of organic liquid originally
emplaced this model was so much less than for the homogeneous model.
Increasing the flow rate had a relatively minor effect on the residual saturations, yet there was some
difference observed between the two experiments run in the aggregated micromodel. The 'fast'
- 171 -
-------
a.
top
flow
b.
FIGURE 9-41. Aggregated model. In the upper photo (a) Soltrol displaced water at a rate
of 0.075 ml/min, from the left (the top of the model) to the right (the
bottom of the model), yielding a residual (irreducible) wetting phase
saturation. In the lower photo (b) Soltrol was displaced by water at the
same rate, from the right (the bottom of the model) to the left (the top),
yielding a residual non-wetting residual saturation. Soltrol was dyed red
and appears dark grey; the water was not dyed. The photos record steady
state flow conditions at the end of the displacements.
- 172 -
-------
a.
top
flow
b.
top
flow
\ - 'I ' ' - ." - • ^r'- /-<
•i" \ • - "*?;.".-SJ?
fc ; V.- - ' "--; /F;
|r *\
•' - ' '3' - .. .• :• ',; 'tiH
FIGURE 9-42. Aggregated model detail from Figure 9-11, showing conditions following
the displacement of the water by Soltrol (a. upper photo), and at residual
non-wetting phase saturation (b. lower photo). The area is located just
below the very center of the model.
- 173 -
-------
a.
top
flow
b.
top
flow
FIGURE 9-43. Aggregated model detail from Figure 9-11, showing conditions following
the displacement of the water by Soltrol (a. upper photo), and at residual
non-wetting phase saturation (b. lower photo). The area is located near the
top of the model, just to the right of the centerline.
- 174 -
-------
a.
b.
FIGURE 9-44. Aggregated model. In the upper photo (a) Soltrol displaced water from the
left (the top of the model) to the right (the bottom of the model), at 1.5
ml/min yielding a residual (irreducible) wetting phase saturation. In the
lower photo (b) Soltrol was displaced by water from the right (the bottom
of the model) to the left (the top), also at 1.5 ml/min, yielding a residual
non-wetting residual saturation. Soltrol was dyed red and appears dark
grey; the water was not dyed. The photos record steady state flow
conditions at the end of the displacements.
- 175 -
-------
experiment led to a slightly higher residual with more trapping within the aggregates. When the organic
liquid advanced into the model at a 'fast' rate it was able to penetrate into a portion of the aggregates.
Later, during the water flood, some of this organic was left behind, even though the water flood occurred
under high flow rate conditions with significant viscous forces. The low flow rate water flood led to slightly
lower residual organic saturations, not because it was more efficient, but because there was less organic
to be removed. In the field the 'slow' rate of organic liquid advance might correspond, for example, to a
slow leak from an underground storage tank, while the 'fast' rate might be associated with a spill from a
tanker accident.
Implications for spill migration rates and travel distances —
In New England groundwater consultants commonly distinguish between gasoline leaking from
underground tanks in unconsolidated, glacial deposits, and leaks in ledge or bedrock. In the
unconsolidated deposits anecdotal evidence suggests that much of the gasoline is trapped by capillary
forces, with very limited and slow migration of the liquid gasoline. The gasoline tends to appear in nearby
brooks dissolved in the groundwater discharge. In crystalline rock terrain the gasoline moves quickly and
far, with observable liquid gasoline discharges to nearby brooks. The aggregated micromodel appears
to provide a good analogue to the fractured bedrock situation.
Implications for aquifer remediation —
Two-phase flow experiments conducted in an aggregated micromodel demonstrate that the
saturation and spatial distribution of organic liquid found behind an advancing front of free product
depends on the combined effects of soil heterogeneity and capillarity. The amount of organic liquid that
is ultimately trapped within a unit volume of aquifer is strongly dependent on how much organic liquid was
originally emplaced within that volume.
Spilled organics can be expected to move quickly through aquifers which have interconnected
macropores or fractures. The organic phase travels preferentially through large pores and fractures
by-passing smaller pores in the matrix of these dual porosity systems. The residual saturations left
behind, following the recovery of free product, tend to be comparatively low, but can be expected to
extend over a much larger portion of the aquifer.
In this aggregated micromodel, most of the residual organic liquid was trapped as by-passed strings
in the macropores. For field sites involving aggregated or cracked soils, or fractured rock, the
implications are that it will be difficult to hydraulically remove all of the organic liquid from the cracks and
fractures. Since real fractures tend to be planar features, rather than the simple linear cracks shown in
this micromodel, multiphase flow in the field is considerably more complex than in laboratory
micromodel experiments (see, e.g., the videotape on multiphase flow in a single fracture by Wilson et
a/., 1988).
Multiphase Movement in a Heterogeneous Aquifer Containing Lenses and Stringers
Stringers or lenses of coarse porous material embedded in a matrix of fine porous material represent
a type of spatial heterogeneity found in clastic sedimentary deposits. For example, gravel bars are often
- 176 -
-------
buried in fluvial deposits. The bars are the lenses, and the surrounding sand and silt is the matrix. Figure
8-13 is an illustration of the micromodel pattern used to represent this type of variability. The pore
network was identical to the homogeneous model (Figure 8-1), but the pores in the coarse zones were
enlarged in size. The coarse lenses were also simulated in short glass and TFE columns filled with
Sevilleta soil, yielding both pore casts of the trapping in a heterogeneously packed sand column, and
quantitative measurements of altered trapping. A new theoretical model is presented that explains the
observed behavior.
Drainage of water by organic liquid-
Soltrol drained a water-filled, horizontally-held micromodel containing a number of coarse lenses.
The lenses were oriented parallel to the flow direction. The Soltrol injection was relatively slow, at 0.096
ml/min. Figure 9-45 illustrates the process.A videotape is also available (Conrad et a/., 1989, see
Appendix B). During the experiment, the front of Soltrol advanced in the finer pore matrix until it
encountered a lens, then slowed while most of the incoming organic liquid preferentially traveled through
that lens. Capillary forces resisting Soltrol entry were lower in the larger pores of the lens (see top photo,
Figure 9-45a). When the lens was full, the front in the fine pores picked up speed again until another lens
was encountered. If the front encountered more than one lens at a time, it fed both of them until one was
full and then fed only the second. The front in the fine pore matrix always advanced from the rear.
Although a full lens might serve as a source for a new front in the fine pores, movement always
proceeded slower than movement of the front through fine pores further behind. Due to the presence of
the discontinuous lenses, progress of the advancement was very unsteady, with the fine matrix front
decelerating and accelerating as lenses were encountered. This unsteady flow is a macroscopic analogy
to the Haines' jumps seen on a pore level during the advancement (Haines1 jumps on a pore scale are
graphically recorded on the videotape by Mason ef a/. ,1988). Favorable mobility (a more viscous fluid
displacing a less viscous one) played a stabilizing role in this displacement. The final steady state fluid
distribution is shown in the bottom photo, in Figure 9-45b. The distribution did not change when the
Soltrol flow was cut off. Wetting phase residual (irreducible) saturation is found in both the coarse and
fine pore regions. Because of the role of the lenses in the advancement, there are zones of fine material,
located between closely spaced coarse lenses, where the water was largely by-passed, leaving a large
residual wetting phase saturation. We also ran models with coarse lenses that extended across the
length of the model. The fine regions were almost entirely by-passed, similar in behavior to the
aggregated model.
The horizontal experiment was repeated in the same heterogeneous model but at a faster flow rate
of 1 ml/min. As above, the front of organic liquid advanced in the finer-pore matrix until it encountered a
lens, then slowed. Even though there was a slight difference in the way the model filled with organic
liquid, it did not make much difference in the residual water saturations behind the front in each model.
The only difference was that at the faster rate, there was less of a capillary end effect.
Implications for spill migration rates and travel distances —
Organic liquid selectively travels through the coarser portions of heterogeneous aquifers. Lenses of
coarse material embedded in a matrix of fine material (e.g. gravel in sand; sand in silt) will influence the
rate and direction of movement of an organic liquid, and create a 'fingering' or 'dispersion' of the
- 177 -
-------
a.
flow
b.
flow
FIGURE 9-45. Soltrol draining a horizontally-aligned 'coarse lens' micromodel from the left. The
photos show fluid distributions as a) the Soltrol was part way through the model,
and b) once the Soltrol had advanced completely through the model.
- 178 -
-------
location of the immiscible displacement front. The front will 'finger' because of the heterogeneity and the
competition between viscous and capillary forces. This fingering is a function of the length and width of
the lenses, and their pore size contrast with the matrix.
For media with continuously varying properties, such as we often represent geostatistically, the
parameter equivalent to the lens length is the correlation length or the range of the covariance or
variogram. There is a current controversy over when 'fingering' due to heterogeneity, sometimes called
'macrodispersion', dominates over fingering due to viscous or density instabilities that are caused by a
contrast in viscosity or density across the immiscible interface. Most of the recent work on this topic has
focused on miscible displacement (Araktingi ef a/., 1988; Moissis, ef a/., 1989; Ewing, ef a/.; Welty and
Gelhar, 1987,1989). The effect of capillarity in immiscible displacement has been considered by some to
be similar to local hydrodynamic dispersion operating during miscible displacement, in that the
transverse mixing caused by either processes tends to limit viscous or gravity instabilities, or
'macrodispersion', induced by heterogeneities. While this notion may hold true for imibition (the
advance of a wetting phase), the opposite appears to be true for drainage (the advance of a non-wetting
phase). During any drainage process capillarity effectively limits transverse mixing and enhances the
growth of heterogeneity, viscous, or gravity induced fingers. We have seen for both the aggregated and
coarse lens micromodels that the advancing organic liquid has been constrained by capillarity to
traveling preferentially through coarse regions, and that this 'macrodispersion' was not mitigated by any
transverse capillary mixing process.
Water flooding-
Figure 9-46a shows the fluid distributions at the end of a displacement of the Soltrol by water for the
model depicted in Figure 9-45. The water was injected at a rate of 0.096 ml/min (from the right of the
photo). The water front moved in the fine pore matrix, splitting as it migrated around each of the coarse
lenses filled with organic liquid. At this slow rate, capillary forces dominated over viscous forces to the
extent that water was preferentially imbibed into the fine pore matrix, even though it was of lower
permeability. When the wetting front reached the downstream end of a lens, it closed back together and
trapped the non-wetting fluid in the lens via by-passing. Very little, if any, of the non-wetting Soltrol was
displaced from the coarse lenses. Typical pore-level capillary trapping of blobs in the fine matrix
occurred, but it was of much smaller in scale than the by-passing of the Soltrol in the lenses. Figure
9-46a shows the equilibrium condition at the end of the displacement. The total residual organic liquid
saturation was significantly higher for this heterogeneous model than it was for the homogeneous model
(Figure 9-11b) due to large-scale by-passing of the organic liquid in the coarse lenses.
Figure 9-46b shows the model after Soltrol had been displaced by water at a fast rate of 1 ml/min. In
contrast to the slow rate displacement, a significant amount of organic liquid has been swept from the
coarse lenses, even though the initial condition was essentially the same. The Soltrol was trapped on the
downstream end of the lenses. Sufficient viscous forces were generated by the fast flow rate to partially
overcome the capillary forces which held Soltrol in the lenses at the slow rate. The effect of flow rate
upon displacement was much more dramatic for this model than for either the homogeneous model
(Figure 9-11) or the aggregated model (Figure 9-41). Since the viscosity of water is lower than that of
Soltrol, this displacement was somewhat unstable. Some fingering was observed as water moved
through the model, which could have been due to the instability or the heterogeneity. The fingering
- 179 -
-------
a.
now
b.
flow
FIGURE 9-46. Water imbibing into the horizontally-aligned 'coarse lens' micromodel from the
right. The photos show fluid distributions after the water had advanced completely
through the model for a) a slow displacement rate (0.096 mm/min), and b) a fast
displacement rate (1 mm/min).
- 180 -
-------
resulted in some by-passing in both the fine and coarse zones. Many common organic liquid pollutants
are less viscous than water, however. For these fluids, the viscous instability experienced under high
flow rates in this experiment would not be expected.
Implications for aquifer remediation —
At typical aquifer flow velocities, capillary forces can relegate the flow of water to finer-grained
regions, by-passing the coarser organic-filled regions. The result can be poor recovery of organic liquids
as high residual organic liquid saturations are left behind. Increased recovery of organic liquids from
heterogeneous aquifers may be attained in some cases by increasing the pumping rate. However, it
appears to be possible that at high enough flow rates the water would preferentially move through the
coarse lenses, possibly by-passing organic liquid in the fine matrix and again leading to poor recovery.
Finally, sampling a coarse stringer and failing to find it effectively saturated with organic liquid does not
eliminate the possibility of large scale trapping. The sample may be taken from the upstream end of a
stringer, while the major body of trapped organic is found toward the downstream end.
Buoyancy Effects-
In a variation of the 'coarse lens' experiment, Soltrol was advanced downward into the vertically-held
micromodel. The Soltrol injection rate, as before, was relatively slow at 0.096 ml/min. The drainage
process was virtually identical to that in the horizontally-held case, except that buoyancy forces played a
role in stabilizing the displacement front by acting in the direction opposite to flow.
Water was then imbibed into the model from the bottom of the vertically-oriented model at a rate of
0.096 ml/min. The end result is depicted in Figure 9-47. Notice that a significant amount of Soltrol was
displaced from the coarse lenses even though the flow rate was low. In the absence of sufficient viscous
forces, buoyancy forces were generated by displacing the less dense Soltrol from below with the more
dense water phase. Buoyancy partially overcame the capillary forces which had previously held organic
liquid in the lenses at this flow rate.
Implications for aquifer remediation —
Capillary trapping in heterogeneous materials is a function of both buoyancy and viscous forces.
Whenever possible remediation schemes should be designed so that buoyancy forces operate in your
favor.
Residual saturations in short columns packed with coarse sand lenses-
This 'coarse lens' experiment was repeated in several short TFE columns using styrene as the
organic liquid phase (see Section 7). Three cylindrical lenses were created by splitting the Sevilleta sand
into two fractions with a size 50 sieve. The fine fraction was used to pack the major portion of the column,
with the coarse fraction used to construct the three lenses, roughly 3.5 cm long and about 2 cm in
diameter. The column was held vertically, and styrene was flooded downward slowly. Later, water was
injected at the bottom and displaced styrene upward at a relatively slow rate. The residual saturation of
styrene was hardened at the end of the water displacement, and pore casts were constructed by
replacing the water phase with epoxy. A longitudinal section of the column through one of the lenses is
shown in Figure 9-48. The coarse lenses contained a much greater saturation of styrene than the
- 181 -
-------
FIGURE 9-47. Water imbibing into the vertical 'coarse lens' micromodel from the right (the
bottom of the model). The photos show fluid distributions after the water had
advanced completely through the model.
surrounding fine matrix, validating the micromodel results. The residual saturation trapped in the fine
matrix was observed to be approximately the same as that observed in the earlier homogeneous
experiments. Even though this column was oriented vertically, buoyancy forces were small and had little
effect on the results. The greater density of styrene, along with much smaller pore sizes in the sand
pack, resulted in a much smaller ratio of buoyancy forces to capillary forces than was encountered in the
vertically-oriented micromodel.
Another experimental trial was conducted in a similar column, but this time the column was oriented
horizontally during the organic displacement step, and the displacement was conducted at a much faster
rate. A longitudinal section of this column through a lens is shown in Figure 9-49. Notice that some
organic liquid was displaced from the coarse lenses, again corroborating the results from the
corresponding micromodel experiment. By-passed styrene still remained in the down-gradient end of the
stringer, but the residual saturation in the up-gradient end was reduced to small disconnected blobs
similar to that found in the matrix of finer pores.
Two vertical glass short column experiments were also run with coarse lenses, and quantitative
measurements were made of final Soltrol residual saturation. The columns were packed with three
lenses, similar to the heterogeneous pore cast experiments. The quantitative results are given in
Table 9-14.The observed 'bulk' maximum and 'bulk' residual Soltrol saturations represent volumetric
- 182 -
-------
FIGURE 9-48. Photograph of residual organic liquid saturation (shaded light) in a heterogeneous
sand pack. Water was flooded from right to left at a low rate. Notice the high organic
liquid saturation in the coarse lenses. The core is 5 cm long.
FIGURE 9-49. Photograph of residual organic liquid saturation (shaded light) in another
heterogeneous sand pack. Water was flooded from right to left. A high rate of flow
produced sufficient force to displace some organic liquid from the coarse lenses.
This core is 5.8 cm long.
- 183 -
-------
sample
1
2
porosity *
(%)
38.2 ±0.4
37. 6 ±0.4
volumetric
percent
coarse/fine
29 / 71
31 / 69
bulk maximum organic
liquid saturation
(%)
85.1 ±2.5
85. 4 ±2. 4
bulk residual organic
liquid saturation *
(%)
31.0 ±1.4
31.2 ± 1.5
* measured value ± propagated error
TABLE 9-14. Measurements of bulk residual organic saturations in two heterogeneous packings
of the Sevilleta sand. The sand was divided into a coarse and a fine fraction, and the
coarse fraction was packed into the column as cylindrical lenses within a matrix of
the fine fraction.
averages over the entire column pore space. The observed 'bulk' maximum Soltrol saturation is
consistent with that observed in the homogeneously packed Sevilleta soil columns. If both the coarse
and fine zones had the same maximum organic saturation, and if no water imbibed into the coarse lenses
during the upward water displacement, then the residual Soltrol saturation in the lenses should not have
changed from this value of roughly 85%. In the fine matrix surrounding the lenses the Soltrol saturation
was presumably reduced to no less than 15%, the Sor values reported for packings of uniform glass
beads (Morrow et a/., 1988). Dry packing of these columns could have avoided any micro-layering, such
as we have suggested is responsible for the higher Sor values in the more common wet packed
columns. Accounting for the relative volume of each soil fraction in the heterogeneous columns, an
estimate for the expected bulk residual Soltrol saturation can be made. If one assumes that little to no
drainage of styrene occurred in the stringers, the measured 'bulk' residual saturation can be compared
with a theoretical value:
estimated bulk residual saturation
- (normalized lense volume) x S0 + (normalized matrix volume) x Sor
= 0.3 x 0.85 + 0.7 x 0.15
= 0.36
Had the typical wet packed Sevilleta sand value of 27% been used for the matrix residual, instead of 15%,
then this estimate would have been higher. The actual bulk residual Soltrol saturation in the
heterogeneous columns was only 31%, greater than the 27% found in the homogeneous columns, but
also lower than the lowest 'expected' value of 36%. Perhaps the coarse lenses actually occupied less
volume than calculated. It is also likely that some mixing occurred between the coarse lenses and the
fine matrix when the packing forms used to make the lenses were removed from the column, thereby
reducing the effective volume of the coarse lenses.
- 184 -
-------
These quantitative results are somewhat inconclusive. On one hand, residual saturations were
increased by the presence of coarse lenses in agreement with both the micromodel and pore cast
results, but the observed residual saturations were less than the theoretically predicted value.
Mechanisms For Trapping via By-passing In Heterogeneous Porous Media
The basic mechanism for capillary trapping via large scale by-passing has been recently recognized
in both the petroleum and hydrologic literature as we reviewed earlier in the introduction to these
experiments on heterogeneity. The displacing wetting fluid migrates around pockets, or lenses, of the
larger or coarser pores, which are preferentially occupied by the non-wetting fluid. Although this basic
mechanism is recognized, there appears to be no literature that discusses the incomplete displacement
of non-wetting fluid from finite sized lenses, as observed here for the larger viscous and gravity force
displacements.
To better understand the mechanisms, consider a simple conceptual model of the aquifer with
porous lenses of intrinsic permeability k2 embedded in a matrix of permeability hi, as illustrated in
Figure 9-50. The pore space of this binary media is initially filled by fluid B, one of two immiscible fluid
phases. The other fluid is designated fluid A. Fluid A is injected at a known flow rate q from the left, and
displaces fluid B from the pore space. There may also be an initial pore level residual saturation of fluid A.
The pore space may be strongly wet by either of the two fluids, or it may be neutrally wet. For example,
Soltrol would be represented by fluid A as it advanced into a glass micromodel saturated with water, fluid
B. When water displaced Soltrol, fluid A would be the water and fluid B the Soltrol. The capillary
properties of the matrix and lenses are assumed to be correlated with their permeability. The spatial
pattern and efficiency of the displacement depends on the spatial statistics of the lenses (size, shape,
frequency), wetting, permeabilities, capillary properties, and the flow rate.
For this simple mathematical analysis a greatly simplified geometrical and fluid mechanical
conceptual model is adopted, as shown in Figure 9-51. The lenses are assumed to be roughly
rectangular in shape with length / and width w. Over the cross-section taken normal to the flow, relative
areas of each flow zone can be calculated. The relative area of the lenses is a2, while the relative area of
the matrix is alt such that:
QI + a2 = 1.0 (9-14)
When the front hits the lenses, the total flux will be divided into flow around the lenses, q\ , over the
area al, and flow through the lenses,
-------
FIGURE 9-50 Random lenses of permeability k2 in a matrix of permeability k,.
The model is base on the premise that displacement occurs as a sharp front. On each side of the
front only one of the two fluids is at greater than residual saturation. To the right this would be fluid B,
while to the left it would be fluid A. Then the flux rates are given by Darcy's law, which for flux in area a, (/'
= 1,2) is:
k,kr Vp,
fi I
(9-16)
where it is the flux over the solution domain, / , in material /, which has intrinsic permeability k, • The
relative permeability for each fluid is kr, and the viscous pressure drop is ApAi (as distinguished from
the total pressure drop, A/>, or the capillary pressure, pc). For the sake of simplicity, several
assumptions are made.
• uniform viscosity, the same in both fluids ( fiA = /UB ) — this assumption is
easily relaxed for a more sophisticated analysis;
• uniform relative permeability, within each the fluid ( eg.,krA = constant );
• each fluid has the same relative permeability( krA = krB ) — this assumption is
also easily relaxed for a more sophisticated analysis;
• pressure in each material of the domain is uniformly varying in the direction of
flow— ie, constant pressure gradient— except for a pressure jump at the
immiscible front caused by capillary forces;
• pressure at each end of the domain is constant and fixed; there is a known total
pressure drop from one end of the domain to the other
(ie, A/J = Api - A/>2 ):
- 186 -
-------
a.
k2
flow
fluid A r fluid B
fluidB
solution domain
b.
FIGURE 9-51 Uniform, parallel lenses of permeability k2 in a matrix of permeability k,: a)
side view of several lenses, and b) cut-away view of one lense.
- 187 -
-------
• there is no longitudinal smearing of the front by capillary forces — ie, the front is
sharp; and
• there is no transverse smearingxof the front by capillary forces in the vicinity of the
lenses; this assumption is very accurate except for the case of very high flow
rates where fluid A is the wetting fluid.
From these assumptions it follows that, within each material, the pressure gradient across the domain
will be constant except for a pressure jump at the immiscible front caused by capillary forces, as
sketched in Figure 9-52.
For the case that we are interested in a wetting fluid displaces a non-wetting fluid, and the lenses
with permeability k2 represent the coarse material. In this case the pressure profiles in each material
look something like that shown in Figure 9-53. Note that the total pressure drop across the domain must
be the same for both materials (ie, Ap! = Ap2), from which it follows that:
By rearranging (9-1 7) in terms of
get:
(9-17)
, substituting the result into (9-16) , and solving in terms of q\ , we
=
Now, solving (9-16) in terms of
and substituting the result into (9-18):
=
By rearranging (9-1 5) in terms of
-------
a.
b.
0
non-wetting fluid
0
X
FIGURE 9-52 Pressure profiles in soil / for fluids A and B, with fluid A as a) the wetting fluid, and b)
the non-wetting fluid. / =1,2 coarse or fine.
a. fine
Pi
0
fluid A
wetting fluid
b.coarse
0
P2
0
front
X
fluid A
flow
wetting;
fluid i
fluid B
0
X
FIGURE 9-53. Pressure profiles for fluids A and B, with fluid A as the wetting fluid, for a) the
fine matrix and b) the more coarse lense.
- 189 -
-------
where rt is the radius of the capillary. By the Kozeny-Carman equation, R, ~ 18 Jk, for random packings
of equal spheres (where Rt is the sphere radius). If media of well-sorted spheres is assumed where r is
of the same order as R:
la cos0 2a cos 9 a cosO
Pc>= —TT - ~^r - TTT (9-23)
Finally, by substituting this relation into equations (9-20) and (9-21) and by merging these equations into
ratio form, we get:
_
*!* > fc JT,
(9_24)
When — > 0 , water will flow into the coarse lenses displacing organic liquid.
To find the critical flux, q , needed to initiate the flow of wetting fluid into the non-wetting fluid filled
coarse lenses, equation (9-24) is set equal to zero and solved in terms of q :
_. _ o cos0 MrCl-Qz) / 1 1
q =
This is really an effective critical capillary number, necessary to begin to reduce the macroscopic
residual non-wetting phase saturation caused by the heterogeneity:
}
91
From equation (9-25), one can see that the critical flux needed to initiate displacement of non-wetting
organic liquid from coarse lenses is inversely proportional to the length of the lense, /, and the relative
area occupied by the lenses, a2, but is directly proportional to the capillary force, CT cos 0.
Figure 9-54 shows the relationship between the critical flux and the permeabilities. When *i = k2, the
system is homogeneous, and the critical flux is zero. Displacement occurs at the same rate in both
materials. When the lenses are composed of coarser material, £2 is larger than k\, and it takes the critical
flux to initiate displacement of non-wetting fluid from the lenses. For very coarse lenses, and large
- 190 -
-------
0
for a fixed k\
fens permeability
matrix permeability
FIGURE 9-54. Critical flow rates needed to displace organic liquid from coarse lenses as a
function of permeability in the coarse lens (top), and in the fine matrix
(bottom).
values of k2 ,this critical flux increases toward an asymptotic value that depends on the permeability of
the matrix:
a cos9 /kikr(l-a2)
9/ii
; for large k2 (9-27)
or, expressed as the critical capillary number necessary to begin to reduce the macroscopic residual
non-wetting phase saturation caused by the heterogeneity:
a
.for large k2 (9-28)
Note that this critical effective capillary number is zero if the contact angle is near to 90c
- 191 -
-------
The fine matrix permeability, #1, can vary between zero and £2. Toward either of these end points,
the critical flux approaches zero, in one instance as ^i becomes impermeable, and in the other instance
as the system approaches homogeneity. In between these extremes the critical flux reaches a maximum
at kl = kz/4 , in which displacing organic liquid from the coarse becomes most difficult (see Figure
9-54) .
For flux rates exceeding the critical value, non-wetting fluid is displaced from each coarse lense,
starting at the upstream end and proceeding through the lense. When the fine matrix wetting front
reaches the downstream end of the lense, the wetting fluid front coming around each side is presumed
to join, isolating the remaining non-wetting fluid that has not yet been displaced from the coarse lense.
This non-wetting fluid is trapped on the downstream end of the lense. The proportion of the lense not
swept by water is described by equation 9-24:
o2
Proportion of lense left unswept - — (9-29)
-------
of varied length and a somewhat random pattern. We strongly recommend further model development
with appropriate experimental validation. The model and experiments should test a variety of geometrical
heterogeneity patterns and material property contrasts. We especially recommend examining the case
of fine lenses imbedded in a matrix of coarser material.
Implications for modelling multiphase flow-
The simple mathematical model developed here suggests that it is feasible to develop effective
properties, or equivalent homogeneous models of behavior in heterogeneous materials. The simulation
codes into which these equivalent homogeneous properties are inserted may not be conventional
codes. They will need to account for the partial de-saturation of lenses, and the effect that has on
effective relative permeabilities and saturations. For example, the effective permeabilities and residual
saturations will certainly depend on the rate and direction of flow. We believe that this is a typical
consequence of dealing with effective properties in non-linear systems. Certainly Yeh et a/.
(1985a,b,c), Mantoglou and Gelhar (1987), and McCord era/. (1988a,b) encountered the same issue in
their work on the Richard's equation approach to water flow in the vadose zone.
Discussion of Multiphase Movement and Capillary Trapping in Heterogeneous Media
The experimental results and simple mathematical model confirm our hypothesis that the spatial
variation or heterogeneity of soil properties has a dominant effect on the movement and capillary
trapping of organic liquids. The impact of this issue on movement and trapping far exceeds that of the
influence of organic liquid composition or local soil texture.
- 193 -
-------
SECTION 10
VADOSE ZONE RESULTS AND DISCUSSION
Figure 10-1 depicts the portion of the aquifer that includes residual saturation in the vadose zone,
above the capillary fringe, where an organic liquid shares pore space with both gas and water. Whether
an organic liquid is more (left) or less (right) dense than water, it leaves behind a trail of capillary trapped
residual in the vadose zone as it moves downward toward the capillary fringe. This movement is much
more complex and diverse than in the saturated zone, especially on a pore scale, primarily because of
the presence of the gas phase. A dropping or fluctuating water table can further complicate the situation.
ground surface
capillary
/fringe
\«^
residual
organic
liquid
saturation,
in the
vadose
zone
floating organic
liquid
SSS dense organic liquid
IIP mmi
in mini mi
limn mini mmi iiniii mini
m limit mini mini mmi mini mini mmi mini m
ni mini mini mini mini mini mini mini mini mini mini mini mini mnn iitwi nniii mini mmi mmi mm mini mini <» ''»»»m n mi mini in
m mm mm mm mm BWH mm mm mm \mm mm mm m/» vm mm mm\ \\m\\ mtw mm \w\\\ \\\ rnf*tf l" mm wm mm mm
mini mini mini mini mint mmi mmi mini mini mini mini mnn mmi mini mini mnn mini imtu mini mini i w w r\ nm mnn mnn mini mi
tti \mm mm mw,\ vm w
FIGURE 10-1. Schematic of residual organic liquid trapped in the vadose zone.
- 194 -
-------
This section on organic liquid movement and capillary trapping in the saturated zone contains four
main parts:
•T review of basic concepts and assumptions,
•2 flow visualization of three-phase displacement and capillary trapping in a
micromodel,
•3 capillary trapping and organic liquid residual saturation in an unconsolidated
soil: the Sevilleta sand, and
•4 micromodel visualization of three-phase capillary trapping of a non-spreading
organic liquid.
The first part is a brief review of three-phase flow concepts and assumptions. It describes some of what
we know about organic liquid behavior in the vadose zone. Second is a description of three-phase
displacement and capillary trapping as observed in a homogeneous micromodel. It visually
demonstrates some of the important behavior issues for so-called 'spreading' organic liquids. We
examine the hypothesis that the organic liquid spreads out as a film along the gas-water interface. Third,
we present the results of our study of spreading organic liquid behavior in the Sevilleta sand, including
photomicrographs of three-fluid-phase pore casts. We further examine the film issue and investigate
the hypothesis, posed below, that organic liquid residual saturations are significantly lower in the vadose
zone than in the saturated zone. Finally, we briefly investigate the hypothesis that the behavior of
non-spreading organic liquids is significantly different than that of spreading liquids.
REVIEW OF CAPILLARY TRAPPING PHENOMENA IN POROUS MEDIA
At the beginning of Section 9 we presented a review of basic concepts that focused on two fluid
phases and drew from the experience of petroleum engineers with pore scale behavior. In the vadose
zone there are three fluid phases present: gas (air), organic liquid, and water. Despite its relevance to
gas-oil-water reservoirs, petroleum engineers have not yet thoroughly addressed pore scale behavior
of three fluid phases. Soil scientists and hydrologists have studied three-phase gas-oil-water flow in
column and sand tank experiments (eg., Schiegg, 1980; Eckberg and Sunada, 1884; Ferrand et a/.,
1986; Lenhard and Parker, 1987a,1988a,1989; Schwille, 1988; Wilson era/., 1988; Gary era/.,1989),
but with little emphasis on pore scale phenomena. Some have simplified the system and examined two
phase imbibition of organic liquids into dry soils (see, eg., Ammozegar et al, 1986; Kia, 1988). In that
case the organic liquid is assumed to be the wetting fluid, and the problem is essentially similar to that of
water imbibition into a dry soil. Below we review some of the basic issues and assumptions regarding the
movement and capillary trapping of organic liquid during three-phase flow.
Recall that there are three major forces acting in both oil recovery and organic liquid behavior in
groundwaters: capillary forces, viscous forces, and gravity or buoyancy forces. Capillarity is the result
of the interplay of cohesive forces within each fluid phase and the adhesive forces between the solid
phase and each of the fluids. The capillary force is proportional to the interfacial tension at the fluid-fluid
interface and the strength of fluid wetting to the solid surface, and inversely proportional to the pore size.
- 195 -
-------
Viscous or dynamic forces within a phase are proportional to the media permeability, the fluid phase
relative permeability, and the fluid phase pressure gradient. Buoyancy is a gravitational force
proportional to the density difference between the fluids.
The additional third phase in the vadose zone is the gas phase. Gas has a much lower density than
either of the two other fluids, so that gravity (buoyancy) forces play a more significant role here than in
the saturated zone. Gas is almost always the non-wetting phase in the vadose zone, so that it takes the
role that the organic liquid usually takes in the saturated zone.
Many organic liquids appear to be of intermediate wettability in typical aquifer materials; that is,
non-wetting relative to water, but wetting relative to gas. The organic phase, depending on whether it
encounters gas or water, can then display either wetting or non-wetting behavior, or both. Many organic
liquids also have low internal cohesion and will spread, presumably forming a film between the water
phase which, because of capillarity, preferentially occupies the smallest pores, and the gas phase which
preferentially fills the largest pores.
This film should interconnect the pockets of organic liquid which even at residual saturation should be
largely continuous. This is the usual assumption in mathematical models (see, eg, the model of Parker
and Lenhard, 1987; Parker et al, 1987; Lenhard and Parker, 1987a, 1988, 1989). Because the residual
organic liquid is interconnected by films, the term capillary trapped may be misleading. With some
exceptions described below, the organic liquid is not actually trapped in isolated blobs. This vadose zone
residual saturation has been referred to by other researchers as the organic phase 'retention' for a soil.
The propensity to spread can be measured by the 'spreading coefficient' (eg, Adamson, 1982):
2 = oaw-(oow +a^) (10-1)
where: 2 = spreading coefficient
oaw = air (gas) - water interfacial tension (the surface tension, y)
OM = organic liquid - water interfacial tension
om = air (gas) - organic liquid interfacial tension
Intermediate wetting organic liquids that tend to spread as a film have a positive spreading
coefficient, H . In this case the surface tension between the gas and water, aaw , exceeds the sum of the
interfacial tensions between the organic liquid and the two other fluids, aow & oao . Figure 10-2 depicts the
force balance. Most of our experimental work focused on spreading organic liquids.
Despite the presence of the intermediate wetting phase organic liquid film, it is realistic to
hypothesize that there are portions of the pore space where there would be no continuous non-wetting
gas phase. Here the intermediate wetting phase would act like a non-wetting phase and become trapped
by capillary forces in the same way we discussed in Section 9. Discontinuous blobs of organic liquid
should tend to occur in these portions.
Some organic liquids have more internal cohesion, as measured by a negative spreading
coefficient, Z (eg, halogenated hydrocarbons such as carbon tetrachloride and PCE). In this case the
- 196 -
-------
air (gas)
a —
niAj
organic liquid
lens
interface
water
ow
solid
FIGURE 10-2.
Diagram of spreading potential for a drop of organic liquid floating on the
air (gas)-water interface (after Adamson, 1982, and others). The water is
wetting, the air is non-wetting, and the organic liquid is intermediate wetting.
sum of the interfacial tensions between the organic liquid and the two other fluids, oow & oao , exceeds the
surface tension between the gas and water, am (see Figure 10-2). These organic liquids will not spread
as films. On a flat water surface (the water-gas interface) non-spreading liquids tend to coalesce into
lenses that float on the surface (much as depicted in Figure 10-2). When the water-gas interface is
within a porous media we don't know what behavior to expect.
In the previous section (Section 9) we introduced micromodel experiments which suggested that the
wetting phase residual saturation may depend on non-wetting phase flow rate. This suggestion was
amplified by the work of Amulfule and Handy (1982), Dullien ef a/. (1986), and Chatzis ef a/. (1988),
which we discussed on page 1 61 . In effect, any increase in either viscous forces, as represented by flow
rate, or buoyancy forces can cause a change in the wetting phase residual saturation. The process is
illustrated in Figure 10-3, in which residual saturations are plotted as a function of the ratio of viscous
plus buoyancy forces, to capillary forces, (Fv + Fb)/Fc (ie., capillary plus Bond numbers). A typical
non-wetting phase residual saturation mobilization curve is also illustrated. This represents water
displacing organic liquid in the saturated zone, such as we described in Section 9 (see, eg, Figures 9-9 &
9-10). There is a critical value of (Fv + Fb)/Fc for the non-wetting residual, Sor , below which viscous
and buoyancy forces are small compared to capillary forces, and the non-wetting phase residual
saturation is at a maximum value of S'or . In contrast, the wetting phase residual saturation is a smooth
function of (Fv +Fb)/Fc (Wilson et a/., 1988; Handy, 1988, personal communication). There is no
critical combination of forces necessary to begin reducing wetting phase residual, because the wetting
phase is interconnected by films and can always drain.
- 197 -
-------
residual non-wetting
phase critical value
FIGURE 10-3. A conceptual plot of residual saturation for the wetting fluid (solid line) and the
non-wetting fluid (dashed line), as a function of the ratio of the sum of viscous and
buoyancy forces, to capillary forces.
In the vadose zone, when the organic liquid spreads as a intermediate wetting phase film, we can
expect somewhat similar behavior. That is, because the residual organic liquid saturation is largely
interconnected, it should be a smooth function of the force ratio (Fv +Fb)/Fc . Buoyancy forces are
proportional to the organic liquid's density contrast with gas. Higher density organic liquids will result in
greater (negative) buoyancy forces and smaller residual organic liquid saturations. Significant gas phase
flows will increase the viscous forces and also reduce residual organic film saturation. Although such
velocities may not normally be expected in the vadose zone, they may be common in vacuum extraction
and other similar in situ volatilization approaches to remediation.
We hypothesize that the organic liquid residual saturation in the vadose zone is smaller than it is in
the saturated zone. The presence of gas as the non-wetting phase, filling the pore bodies, is the main
basis for this hypothesis. There is also a propensity for greater buoyancy forces because of the density
contrast with gas. Conversely, in the saturated zone, the combined buoyancy and viscous forces are not
usually great enough to overcome the capillary forces and the residual saturation is at its maximum.
MICROMODEL FLOW VISUALIZATION OF THREE PHASE DISPLACEMENT
AND CAPILLARY TRAPPING
Micromodel flow visualization techniques in homogeneous media illustrate the scenario of a large
slug of organic liquid percolating vertically downward into the vadose zone. Later, it's drained by air as it
- 198 -
-------
continues downward on its migration toward the water table. This scenario was simulated by first
saturating a water-wet etched glass micromodel with water. Then, a vadose zone condition was created
by draining the water with air to residual (so-called irreducible) water saturation. An organic liquid,
Soltrol, invaded downward into the experimental apparatus, and finally the Soltrol itself was drained by
air. The residual Soltrol left behind in the vadose zone was observed visually. Soltrol has a positive
spreading coefficient and was an intermediate wetting fluid in this experiment. This experiment allowed
us to investigate the hypothesis that the intermediate wetting organic liquid phase spreads out as a film
between the water and air.
Creating the Initial Vadose Zone Condition
The homogeneous micromodel with the capillary barrier (see Figures 8-1 and 8-9) was used in this
three-phase experiment. The model, oriented vertically with the barrier at the bottom, was imbibed with
blue dyed water from the bottom and then drained from the top with air to achieve the initial vadose zone
condition (see Section 8). The water was introduced and drained from a buret, under a suction of about
15 cm H20, just under the air entry pressure of the capillary barrier. The steady-state condition of the
entire model, at the end of drainage, is shown in Figure 10-4. In black and white photos the water
appears gray; the air is colorless. This steady-state condition represents the portion of the vadose zone
well above the capillary fringe (see Figure 10-1), where water saturation is near its so-called irreducible
value. The water fills some of the pore throats, an occasional by-passed pore body, and forms a film
around the air-filled pores. Because the model had fairly large pores and a relatively uniform pore size
distribution, not much suction was needed to achieve a fairly complete drainage. Unlike the results
obtained from the micromodels used in the saturated zone experiments, the presence of the capillary
barrier at the bottom of the model prevented a capillary end effect. (Capillary end effect was discussed
in Section 9 for the case of Soltrol advancing into the same model without a capillary barrier under
water-saturated conditions.) This photograph can be compared to Figures 9-11a and 9-34a. Here air
was the non-wetting phase; there it was Soltrol.
Organic Liquid Invasion
In the next step red dyed Soltrol invaded downward into the drained micromodel at a relatively slow
rate of 0.096 ml/min. Air and perhaps some water was pushed ahead of the Soltrol, which primarily
advanced by displacing air, in bulk, from the interconnected air-filled pore bodies and pore throats. The
Soltrol largely bypassed water that had previously been 'trapped' in pore throats. At the bottom of the
model sufficient pressure was achieved to exceed the air entry pressure of the capillary barrier, thus
allowing air to escape.
During this displacement, some pockets of organic liquid were developed ahead of the main front of
advancing organic liquid. At first glance, we were puzzled as to where this fluid had come from, but
closer examination revealed that it had traveled through thin films of organic phase formed between the
air and water phase. Figure 10-5, a photograph take from a videotape of an earlier experiment,
illustrates this process (Mason et al, 1989; a similar videotape is available for viewing, see Appendix B).
The blue water and red Soltrol dyes were deeper in color in this experiment. In black and white the Soltrol
- 199 -
-------
FIGURE 10-4. Initial vadose zone condition, with water drained by air to residual (irreducible)
water saturation.The water was dyed blue (light grey), and the air was not dyed.
- 200 -
-------
FIGURE 10-5. Detail of Soltrol invasion into a different vadose zone model. The Soltrol was
advancing by filling pores and by film flow. Soltrol was dyed red (dark grey),
water was dyed blue (light grey), and the air was not dyed.
FIGURE 10-6. Detail of steady state conditions after the Soltrol invasion into the vadose zone
model. Soltrol was dyed red (dark grey), water was dyed blue (light grey), and
the air was not dyed.
- 201 -
-------
appears as a very dark grey, and the water is a lighter grey. The initial water saturation was also greater
than the irreducible value. A finger of Soltrol, filling the pores, penetrates the frame from the top. In the
Soltrol-filled pore moving off to the left, the Soltrol attached to the water coated walls as a film, with air
filling the center of the pore. Soltrol was flowing in this film, which leaves the frame and then re-enters
just below. Soltrol then moved off toward the bottom of the frame as film flow. To the right center of the
frame is a film bridging an air-filled pore body between two Soltrol-filled pore throats. Rapid
advancement of Soltrol through film flow is apparently a form of preferential flow.
The steady-state condition for the entire model at the end of the organic liquid invasion is shown in
Figure 10-7. A more detailed photo is shown in Figure 10-6. In black and white the Soltrol appears as a
dark gray and the water appears as a lighter gray. Notice in the detailed photo that the Soltrol filled many
pore throats and formed a film between the air and the water. This film penetrated somewhat into the
water-filled pore throats, almost 'smiling' at the observer. In short, water-filled pore throats the 'smiles'
from each end almost touched, and the water phase pendular ring was clearly visible. Water was also
present as a wetting phase film between the Soltrol and the solid surface. Figure 8-8 is an SEM
photomicrograph of a typical eye-shaped micromodel pore cross-section. The wetting phase 'film'
actually occupied the wedges on either side of the pore. The saturation and distribution of the water
phase remained almost entirely unchanged when compared to the initial vadose zone condition.
The steady state condition depicted in Figure 10-7 can be compared to the two phase flow
experiments. The non-wetting air phase was trapped in isolated bubbles or blobs, similar to those seen
for the saturated zone organic phase (as shown in Figures 9-11 b, 9-12b, 9-13b, 9-15, and 9-16). Here
air was the non-wetting trapped phase; there it was Soltrol. Because of the presence of air in the vadose
zone (in addition to organic liquid and water), the maximum organic liquid saturation achieved was not as
great as in the saturated-zone case.
Drainage of Organic Liquid by Air
In the final step of the experiment, Soltrol was drained from the model by air, representing the
continued percolation of the organic phase downward toward the water table. The Soltrol was drained
from the model under a small applied suction. This suction is believed to have been insufficient to
completely drain the organic phase to its lowest possible saturation, but because Soltrol has a much
lower surface tension than water, higher applied suctions would have resulted in air breaking through the
capillary barrier. We refer to the resulting Soltrol saturation as 'residual saturation'.
The final distribution of fluids within the entire model is shown in Figure 10-9, while the close-up
photos in Figure 10-8 illustrate the more complex nature of organic liquids left behind in the vadose zone.
Comparing Figure 10-9 to Figure 9-11 b we note that the amount of organic phase retained in the vadose
zone was much less than the residual saturation left behind in the saturated zone, as we expected. The
distribution of fluids is also much more complex. The micromodel photographs in Figure 10-10 further
illustrate these points. The lower photo shows residual saturation obtained under two-phase conditions in
a small section of a micromodel, representing saturated zone conditions. The upper photo shows
residual saturation in the identical portion of the same micromodel under three-phase conditions,
representing vadose zone conditions.
- 202 -
-------
FIGURE 10-7. Steady state conditions after the Soltrol invasion into the vadose zone model.
Soltrol was dyed red (dark grey), water was dyed blue (light grey), and the air
was not dyed.
- 203 -
-------
Inspection of Figures 10-8 and 10-9 reveals that residual organic liquid is distributed in several
different ways in the vadose zone. In particularly dry portions of the pore space, it is found in pore throats
as pendular rings, in small pore bodies (see lower right portion of Figure 10-8), and as a film between the
air and water phases. The film helps fill the wedges of the pore cross-section (see Figure 8-8) and
penetrates as 'smiles' into the water-filled pore throats, surrounding water-filled pendular rings. This
film can be quite thin, as shown in Figure 10-11, taken from another micromodel. Sometimes organic
liquid was trapped, similar to the two-phase case, as blobs within the water phase. We found that this
kind of trapping is more likely to occur when the initial water content is relatively high or when infiltrating
water follows the organic liquid (eg, rainfall infiltration following a spill event). Occasionally we saw a gas
bubble trapped inside one of these organic liquid blobs. However, due to its intermediate wetting
properties and its tendency to spread, organic liquid in the vadose zone is most commonly retained in
between the water and air phases.
In contrast to organic liquid blobs trapped in the saturated zone, the residual organic phase in the
vadose zone remains more or less continuous, interconnected by the ubiquitous films (with the
exception of any organic liquid that might have been trapped as blobs entirely within the water phase).
We refer to this organic liquid as 'residual' and 'trapped', but these terms must be used with care. As we
discussed in reference to Figure 10-3, the intermediate wetting phase saturation is a continuous function
of applied viscous and gravity forces; it can always be reduced further, albeit slowly.
FIGURE 10-8. Detail of steady state conditions after the Soltrol has been drained by air from the
vadose zone model. Soltrol was dyed red (dark grey), water was dyed blue (light
grey), and the air was not dyed.
- 204 -
-------
FIGURE 10-9. Steady state conditions after the Soltrol has been drained by air from the vadose
zone model. Soltrol was dyed red (dark grey), water was dyed blue (light grey),
and the air was not dyed.
- 205 -
-------
a.
b.
FIGURE 10-10. Steady state conditions in a micromodel after a) Soltrol has been drained by air
from a vadose zone model, and b) Soltrol has been displaced by water in a
saturated zone model. The Soltrol was dyed red (grey), and the water was dyed
either green (top photo; dark grey) or blue (bottom photo; light grey).The air
was not dyed.
- 206 -
-------
FIGURE 10-11. Detail a thin organic liquid film located between the gas and water. The photo
represents steady state conditions after the Soltrol has been drained by air from
the vadose zone model. Soltrol was dyed red (dark grey), water was dyed blue
(light grey), and the air was not dyed.
The final distribution of Soltrol in this micromodel — although complicated by the presence of water —
was not dissimilar to drained wetting phase distributions described by Hillel (1980), Gvirtzman et al.
(1987), and others. In this experiment, due to having a relatively low initial water saturation, and because
the organic phase readily spread on the water/air interface, the organic phase acted much like a wetting
phase both in its character of migration and its final distribution within the pore space.
Discussion
Our results support the hypotheses that the organic liquid forms a film between the water and gas,
and that organic liquid saturations are much lower in the vadose zone than in the saturated zone.
Implications for modelling multi-phase flow -
Most continuum mathematical models of organic liquid migration assume that the organic liquid
phase is continuous and forms a layer between the water and air phases (for one example of this model
see Parker et al., 1987; Lenhard and Parker, 1987a, 1989 ; Parker et al., 1987; Parker and Lenhard,
1987). Within the context of this experiment that assumption appears valid.
Mathematical percolation network models of multi-phase flow (eg, Soil and Celia, 1988) often
assume that only one phase can occupy a given pore body or pore throat at any given time and neglect
- 207 -
-------
films and film flow. In our experiments we have commonly observed two and three phases
simultaneously occupying a single pore, and we've seen that film flow can play an important role in the
movement of organic liquids. The wetting and intermediate-wetting phases remain self-connected
through the films, which maintain a continuity for drainage or imbibition that cannot be captured using a
traditional percolation network approach. Future versions of these network models should explore this
issue.
Implications for phase partitioning-
As discussed earlier in the saturated zone section, the pore-scale distribution of the organic phase
influences the partitioning of organic liquid components. Because the organic phase is almost always
less wetting than water but more wetting than air, it remains in direct contact with both the air and water
allowing for both solubilization into the water phase and especially volatilization into the air phase. The
formation of thin organic liquid films between the water and air increases the surface area of the organic
phase, enhancing the propensity for inter-phase partitioning of organic components. Soil venting of
organics in the vadose zone has become an attractive remediation strategy because volatile organics
partition easily to the air phase. The thin film in Figure 10-11 provides a good example of the large
organic/air and organic/water interfaces generated by the organic phase's tendency to spread.
However, the films are not of uniform thickness. They interconnect organic liquid filled 'smiles' and pore
throats. In a multicomponent organic liquid, mass transfer rates could be limited by the diffusion of the
more volatile components to the interface, from the 'smiles' and pore throats. The potential rate of
transfer would then be distributed non-uniformly along the gas-organic liquid interface.
CAPILLARY TRAPPING AND RESIDUAL SATURATION IN AN UNCONSOLIDATED SOIL:
the SEVILLETA SAND
We carried out three-phase flow experiments on an unconsolidated natural soil, the Sevilleta sand,
using air, water, and Soltrol for the fluids (see Section 4). The Sevilleta sand was strongly water-wet, and
Soltrol was the intermediate wetting fluid. We made quantitative measurements of Soltrol residual (see
Section 5), for conditions representing the vadose zone well above the capillary fringe (see Figure
10-1), where water saturation is near its so-called irreducible value. In other experiments we used
styrene as the organic phase, which we hardened in place and photographed (see Section 7) .Through
these two sets of experiments we further examined the film issue and investigated the hypothesis that
organic liquid residual saturations are significantly lower in the vadose zone, than in the saturated zone.
Finally, we measured residual Soltrol saturations for conditions representing the vadose zone just above
the capillary fringe.
Quantitative Measurements Of Residual Saturation In Three-phase Soil Columns
Quantitative measurements of residual Soltrol saturation were made in short soil columns, as
described in Section 5. Ten trials of the column experiments were successfully completed, representing
vadose zone conditions far above the water table. Briefly, in each experiment an initially water saturated
soil column was first drained with air under an applied suction to create an initial 'vadose zone' condition.
- 208 -
-------
After the water saturation stabilized, an organic phase was flooded into the column, simulating the
infiltration of organic pollutants through the vadose zone. After the fluid saturations had equilibrated, the
organic liquid was drained under an applied suction. Once the column re-equilibrated, the residual
organic liquid saturation was measured.
We believe that the residual organic liquid saturation left behind in the vadose zone may be a function
of saturation history. In our column studies, we chose one simple but realistic saturation history to
directly compare residual organic liquids found trapped in the saturated zone to those found in the
vadose zone. Owing to the saturation history dependence of residual saturation, the following
operational definition for residual saturation in the vadose zone was used:
vadose zone residual saturation — The organic liquid saturation obtained by
injecting organic liquid into a water- and air-filled porous medium — in which
water is already at its irreducible water saturation — until the fluid saturations
stabilize, followed by drainage of the organic liquid with air until the organic
liquid becomes immobile and can be reduced no further.
A summary of the quantitative vadose zone results is presented in Table 10-1. In the first step of
each experiment, an initially water-saturated column was drained to create unsaturated conditions. The
column was drained under sufficient suction to reach the asymptotic portion of the capillary retention
curve where the sand was almost completely drained and the water saturation became relatively
insensitive to changes in capillary pressure. Water saturations in this region of the retention curve
approached the so-called irreducible water saturation, SWI . The water saturation established in this
drainage step remained constant throughout the duration of the experiment. The air saturation after this
initial drainage step was: 100% - Swl. The average water saturation over the ten experimental trials was
19.8%, thus the average air saturation was 80.2%.
Next, Soltrol was injected into the column simulating the movement of organic liquid percolating
through the vadose zone toward the water table. Soltrol displaced only air from the column. Again,
because the water saturation was so low (at or near its 'irreducible' saturation), the water was essentially
immobile and none was displaced from the column. At the conclusion of this step the Soltrol had reached
its maximum saturation (listed in Table 10-1). The maximum organic saturation reached in these vadose
zone experiments was significantly lower than the maximum saturation reached in the saturated zone
column experiments (66.0% versus 85.1% on average) due to to the presence of a third phase — that is,
entrapped air occupied pore space that otherwise would have been occupied by Soltrol in a two-phase or
saturated zone system (see Figure 10-6).
The residual or entrapped air saturation at the conclusion of the Soltrol injection was calculated as:
100% - S0 - Sw . The mean residual air saturation over the ten vadose zone experimental trials was
14.2%. Two-phase organic/air and water/air retention curves yield somewhat similar residual air
saturations (less than 20%; see Appendix C: Figures C-10, C-12 & C-13 and Table C-1). Since the
water saturation was unchanging in the vadose zone experiments, it might be reasonable to think of
organic liquid displacing air as being analogous to the final displacement step of the saturated zone
experiments where water displaced organic liquid leaving behind a residual saturation — in each case a
more wetting fluid displaced a less wetting one and some entrapped non-wetting phase was left behind.
- 209 -
-------
However, the entrapped air saturation in the three-phase experiments was found to be only about one
half the residual organic saturation of the saturated zone experiments (14.2% versus 27.1%). Some of
the air may have dissolved into the displacing organic phase, but more likely the displacement was more
efficient for some reason. Perhaps due to the very low viscosity of the air phase, the air was less likely to
become by-passed. One could think of this phenomena as the effect of an extremely favorable mobility
ratio operating on a pore scale. The presence of an intermediate wetting fluid, Soltrol, with a lower
interfacial tension to air could also have played a role (see Table 10-2). However, we saw similar results
in all two-phase experiments using air as the non-wetting phase (Appendix C). Perhaps the greater Bond
number of the displacement was sufficiently high to permit a greater efficiency. The low density of air
leads to a greater density difference (see Table 10-2) and greater gravity forces. However, these forces
were acting upward, whereas the displacement in the vadose zone experiment was horizontal (see
Figure 5-6). At any rate, it appears that residual air saturations created under two and three-phase flow
conditions were consistently smaller than residual organic liquid saturations created under two-phase
flow conditions.
In the final step of the three-phase experiment, organic liquid was drained from the column and
replaced by air. Sufficient suction was applied to reduce the organic liquid saturation as low as possible
Trial
1
2
3
4
5
6
7
8
9
10
Avg.
Vi
Suction *
(* 1 cm)
60
68
59
61
66
71
70
69
71
75
-
-
Temperature
Range
(°C)
4.0
3.6
5.0
4.3
2.3
4.3
2.1
1.9
2.6
2.0
-
-
Water Saturation
(%)
18.1 4- 1.9
19.8 i 1.9
22.8 4- 2.3
21.6 i 2.1
18.5 ± 2.1
20.3 ± 2.2
23.2 ± 2.2
18.3 ± 2.4
19.8 * 3.7
15.8 A 2.4
19.8
2.3
Maximum Organic
Liquid Saturation
(%)
62.1 4 1.6
65.5 i 1.8
63.9 4 2.1
78.3 i 2.2
64.5 i 1.9
66.2 ± 2.0
56.9 ± 1.8
65.2 ± 2.1
66.1 * 3.2
71.4 ± 2.2
66.0
5.6
Residual Organic
Liquid Saturation
(%)
8.3 * 0.4
12.0 i 0.5
9.7 * 0.6
11.3 ± 0.5
7.7 .* 0.4
7.9 A 0.5
12.2 ± 0.6
7.1 a- 0.4
9.1 4 0.7
5.5 i 0.4
9.1
2.2
TABLE 10-1.
cm of water were used during water drainage, cm
of Soltrol were used during organic liquid drainage.
Results from the vadose zone column experiments. Soltrol-130 was used as the
organic liquid and Sevilleta sand served as the soil.
- 210 -
-------
Zone
Vadose
Saturated
fluid pair
Soltrol-air
Soltrol-water
density
difference
(g/cm3)
0.75
0.25
Interfacial
tension
(dyne/cm)
19.2
48.7
TABLE 10-2. Relative density differences and interfacial tensions in the vadose zone and
saturated zone.
by purely hydraulic means. The results are presented in Table 10-1. The average measurement of
residual organic liquid saturation in the vadose zone was found to be 9.1 ± 2.2% — less than one third
of the value measured in the saturated zone.
The Soltrol residual for these three-phase experiments were not significantly different from the
minimum organic saturations obtained from air/organic drainage curves (compare with results in
Appendix C). These findings are somewhat inconsistent with those of other researchers, who generally
use the term 'organic retention'. Eckberg (1983) reported slightly higher organic retentions in two-phase
(organic/air) systems than for three-phase (water/organic/air) systems. Convery (1979) reported
two-phase retentions to be between 10% and 25% higher, while Hoag and Marley (1986) reported
retentions between 20% and 30% higher. However, these researchers measured average saturations
over the length of long soil columns in which the fluid saturations were not uniform. The actual
relationship depends on the soil and the fluid-fluid interfacial tensions.
Photomicrographs of Three-phase Pore Casts
Thin section photomicrographs of three-phase pore casts constructed from Sevilleta soil columns
are shown in Figures 10-12 to 10-14. These pore casts were constructed following the procedures and
techniques described in Section 7. Styrene was the wetting phase in these casts, while epoxies
represented the intermediate and non-wetting phases. The styrene was dyed red and appears grey or
even black in black-and-white photos. The intermediate phase appears white or light grey, while the
non-wetting phase was dyed blue and appears dark grey. Because these fluids were applied and
hardened sequentially, their distribution within the pore space only simulates three-phase conditions in
the vadose zone. The wetting phase simulated water, the intermediate wetting phase simulated the
organic liquid, and the non-wetting phase simulated the soil gas. These photomicrographs should be
compared to the micromodel results, particularly Figure 10-8. The observed behavior was similar.
Figure 10-12 shows a pore body filled with non-wetting fluid. The intermediate wetting phase forms
a thick film between this pore body and the adjacent pore throat, which is filled with the wetting phase.
- 211 -
-------
FIGURE 10-12. A photomicrograph of a pore cast thin section from the simulated three-phase
system in the Sevilleta sand. The middle of the photo depicts a pore body filled
with non-wetting phase (blue or dark grey). Above it is thick 'film' of
intermediate wetting phase (white or light grey), that is 'smiling' into a pore
throat. The pore throat is otherwise filled with wetting fluid (red or light grey).
Shown at 100X magnification.
- 212 -
-------
FIGURE 10-13.
A photomicrograph of a pore cast thin section from the simulated three-phase
system in the Sevilleta sand. The middle of the photo depicts a pore body filled
with non-wetting phase (blue or dark grey). It is surrounded by an intermediate
wetting film (white or light grey), that is 'smiling' into the pore throat on the
right, and filling most of the pore throat to the left. Shown at 10OX magnification.
- 213 -
-------
FIGURE 10-14. A photomicrograph of a pore cast thin section from the simulated three-phase
system in the Sevilleta sand. The middle of the photo depicts a small pore body
filled with non-wetting phase (blue or dark grey). It is surrounded by an
intermediate wetting film (white or light grey), that is 'smiling' into the pore
throats on the right, left and below. The pore throats are filled with the wetting
fluid (red or black in this photo). Shown at 100X magnification.
- 214 -
-------
The relative wetting of the phases can be deduced by the interfacial curvatures and the interface contact
angles with the surrounding grains. In this location the intermediate wetting phase film is very thick as it
'smiles' into the pore throat. These so-called 'smiles' represent a thick film of intermediate wetting fluid
'coating' a wetting phase pendular ring.
Figure 10-13 shows another pore body filled with non-wetting fluid. The intermediate wetting phase
occupies most of the pore throat on the left and forms a thick film between this pore body and the
adjacent pore throat on the right. In Figure 10-14 the intermediate wetting phase film is much thinner, as
it surrounds the non-wetting phase filled pore throat. In these two photos the non-wetting phase appears
to be isolated and discontinuous, but this impression is misleading. The thin sections are two dimensional
sections through a three-dimensional soil sample. Stereo microscopic inspection of pore cast fat
sections insured that these non-wetting zones were continuously connected.
The wetting and intermediate wetting films along the grain boundary of the pore body are not visible
at the magnification shown in the micrographs. Indeed, they are difficult to see at higher magnifications
because of grain roughness, thin section preparation limitations, small film thickness, and the leaching
and diffusion of dyes.
Quantitative Measurements Of Residual Saturation In Transition Zone
Six additional quantitative vadose zone column experiments were performed with Sevilleta sand and
Soltrol as the organic liquid. These experiments examined the saturation distribution in the transition
zone above the capillary fringe. Results are shown in Table 10-3, and all vadose zone results are plotted
in Figure 10-15. The left curve shows the water saturation versus equivalent height above the water
table; the right curve shows the total liquid saturation. The difference between the two curves is the
Soltrol saturation. The saturation curves were fitted by nonlinear regression (van Genutchten, 1980) to
the experimental data points (excluding the results from trials 5 and 6). The 'dry zone' marked on the
graph represents conditions far above the vadose zone.
For the Sevilleta soil, the difference between the two curves remains relatively constant for distances
more than 30 cm above the water table. In the transition zone below 30 cm the curves diverge as the
organic liquid saturation increases. Unlike the 'dry zone' experiments discussed above, the initial water
saturations in these transition zone experiments were higher, and some water was displaced along with
air as organic liquid moved through the columns. Particularly for experimental trials intended to simulate
conditions beneath 40 cm above the water table, the initial water saturations were quite high, and a
significant amount of that water was displaced as organic liquid was injected into the columns. In a
natural system, this type of displacement is to be expected as organic liquid approaches the water table.
If sufficient organic liquid reaches the water table, the water table may be depressed by the weight of the
organic phase as it forms a lens on the water (see, eg, right side of Figure 10-1). In time however, as the
organic lens spreads laterally, the water table will rebound and water (since it is the most wetting phase)
will re-imbibe back into the pore space above the water table displacing some organic liquid. Our
experimental technique did not allow the water to re-imbibe into the soil. Under low suctions
corresponding to the region in and just above the capillary fringe, this flaw in our experimental technique
lead to lower than reasonable final water saturations and higher than reasonable Soltrol saturations. Final
- 215 -
-------
trial
1
2
3
4
5
6
suction *
(cm)
50
48
41
32
22
19
Final Fluid Saturations (%)
Air
57.1 ±4.0
69.2 ±2.3
50.4 ±3.0
47.5 ±4.0
50.9 ±4.9
4.4±7.8
Water
32.4 ±2.5
26. 2 ±1.9
40. 1± 1.9
31.0 ±2.2
26.6 ±2.8
15.8±3.6
Organic Liquid
10. 5 ± 2.5
4.6±0.4
9.5 ± 1.1
21.5 ± 1.7
22.4 ±2.2
79.8 ±4.2
cm of water were used during water drainage, cm
of Soltrol were used during organic liquid drainage.
TABLE 10-3. Results from vadose zone column experiments performed to examine the saturation
distributions in the transition zone between the saturated zone and the vadose zone.
The media was Sevilleta sand and the organic liquid was Soltrol. Results from
experimental trials 5 and 6 are suspect for reasons discussed in the text.
'equilibrium' water saturations below 30% were measured from the two trials nearest the capillary fringe
(trials 5 and 6). A Soltrol saturation of nearly 80% was measured in trial 6, the experiment conducted with
the lowest suction. We believe these results are not indicative of conditions found in natural systems and
serve as examples of the limitations of the experimental technique in the low suction range above the
capillary fringe.
The largest change in forces acting on the organic liquid occurs in the transition zone above the
capillary fringe. Unfortunately, the procedure used in these experiments proved to be inappropriate for
measuring fluid saturations in this region. This low suction range is important, particularly for organic
liquids lighter than water, because it is the region in which these organic liquids spread laterally, forming
a lens on the water table (see Figure 10-1).
Discussion
Our quantitative results support the hypothesis that residual organic liquid saturations are much
lower in the vadose zone than in the saturated zone. Vadose zone column experiments using the
Sevilleta soil yielded an average value for residual Soltrol saturation (9.1%), approximately one-third as
large as the saturated zone residual Soltrol saturation (27.1%). The thin section photomicrographs
validate the probable presence of organic liquid films at the gas-water interface, and support our
micromodel observation that the distribution of fluids in the vadose zone is more complex than the
distribution found in the saturated zone.
As discussed earlier, we believe that the presence of a third and non-wetting phase, air, along with
the continuity of the films during drainage and increased buoyancy forces, helps account for the lower
- 216 -
-------
90-|
80
70
60
height above 5°
water table
(cm)
40-
30-
20-
10-
O-l
trials 5 & 6 were not
used in the regression
X
Dry Zone
X - water saturation
• - total saturation
Transition Zone
\
AIR
Capillary Fringe
ORGANIC ^
LIQUID
10 20 30 40 50 60 70
Liquid Saturation (%)
80
90
100
FIGURE 10-15. Inferred distribution of fluids in the vadose zone for the Sevilleta sand, using
Soltrol-130 as the organic liquid in individual short column experiments. The dry
zone data is taken from Table 10-1, while the transition zone data is taken from
Table 10-3. Results from experimental trials 5 and 6 are suspect for reasons
discussed in the text.
residual saturation in the vadose zone. Referring back to saturation-force diagram in Figure 10-3, the
residual saturation decreases as the ratio, (Fv +Fb)/Fc, is increased. Most organic pollutants
immiscible with water have a much greater buoyancy force in the vadose zone (larger Fb), due to the
large fluid density difference with air. We've also noticed that the interfacial tension between air and
organic in the vadose zone is also usually smaller than the tension between the organic liquid and water in
the saturated zone (smaller Fc). For example, Soltrol-130, the organic liquid used in the micromodel
experiments and column studies, has three times the buoyancy forces and 2.5 times less capillary
forces in the vadose zone than in the saturated zone (see Table 10-2). Under equivalent conditions
(i.e., same soil and packing), the ratio of forces, (Fv + Fb)/Fc , will be a much larger in the vadose zone
compared to the saturated zone.
Implications for aquifer contamination -
The capacity for 'storing', retaining or trapping organic liquids is much smaller in the vadose zone
than it is in the saturated zone. For organic liquids less dense than water (right side of Figure 10-1) this
- 217 -
-------
may be a misleading observation. After all, these 'floaters' can sweep through a large volume of the
vadose zone, between the surface and the water table, but can only sweep through that portion of the
saturated zone near the water table.
Implications for modeling multi-phase flow -
Models like that of Parker and Lenhard (1987) assume that three-phase
capillary pressure/saturation relationships can be approximated by two two-phase relationships. This
assumption, first proposed by Leverett in the early 1940's, maintains that water saturations in a
three-phase system are a function of the two-phase water/organic retention curve, while the total fluid
saturations (water + organic) are functions of the organic/gas curve. Implicitly, this assumption does not
allow water/gas interfaces to exist. Indeed, in our flow visualization experiments we see comparatively
few water/gas interfaces. So for the majority of common organic pollutants which tend to spread
between the air and water phases in the vadose zone, approximation of three-phase
capillary pressure/saturation relationships using two two-phase retention curves seems to be an
appropriate assumption. It aprears to preserve the physics and surface chemistry of the three-phase
flow system in addition to being computationally advantageous.
Implications for phase partitioning -
The rate of mass transfer from the organic liquid phase to the gas phase depends on the interfacial
surface area. In reviewing the micromodel results, we hypothesize that this rate of transfer can be
limited by diffusion within a multi-component organic liquid. The photomicrographs presented in Figures
10-12 to 10-14 suggest that this film can be very thin. The rate of diffusion would be limited by the film
thickness and tortuosity, and the geometry of its connection to larger pockets of organic liquid.
Implications for biotransformation -
Microorganisms tend to adhere to pore walls. There are a number of different environments that
organisms might colonize in the vadose zone. Some of these are revealed in the micromodel (Figure
10-8) and pore cast (Figures 10-12 through 10-15) photomicrographs. There are largely air filled pores,
with a ready access to oxygen, but with only a very thin double film of water and organic liquid. At residual
saturation these films would have a very limited ability to resupply needed nutrients. The environment of
water filled pore throats and pendular rings may be controlled by the lack of oxygen. The film of organic
liquid should suppress oxygen transfer to the water phase. We speculate that transient changes in fluid
saturations would improve the environmental conditions for aerobic bacteria.
Other implications for aquifer remediation - the surfactant effect
Biosurfactants are a by-product of biological activity, especially for organisms that attach at the
organic liquid-water interface. We hypothesize that these natural surfactants could change interfacial
tensions enough to reduce organic liquid saturations by drainage (see Figure 10-3).
MICROMODEL VISUALIZATION OF CAPILLARY TRAPPING
OF A NON-SPREADING ORGANIC LIQUID
On a flat water surface (water-air interface) a non-spreading organic liquid coalesces into a lens,
like that depicted in Figure 10-2. We hypothesize that this coalescence would interrupt the tendency of
- 218 -
-------
the organic liquid to spread out as a film in the vadose zone, leading to a different behavior than we had
observed for Soltrol. To investigate this hypothesis we repeated the vadose zone micromodel
experiment using PCE, a non-spreading organic liquid. The results were recorded on videotape (Mason
ef a/., 1989). The photos shown below were taken from that tape, and have a lower than desired
resolution. The PCE was not dyed, as the dyes reversed the spreading coefficient. The experiment looks
essentially the same in both color and black-and-white. Black-and-white photos are shown here.
The major result of this experiment is shown in Figures 10-16 and 10-17. The left side of Figure
10-16 shows an air-filled pore body. The air extends down into an air-filled pore throat. Above the pore
body and to its upper left are PCE-filled pore throats. A PCE 'smile' extents into the pore throat on the
right and abuts a pendular ring of water. There is no PCE on the other side of this pendular ring, which
directly contacts another air-filled pore body. Direct your attention to the PCE that surrounds the
air-filled pore body on the left. The PCE film extends downward along the side of the air pocket, and then
stops. Figure 10-17 is a close up of this truncated film. It comes to an abrupt end that could clearly be
seen through the stereo microscope, especially under different lighting conditions. Figure 10-18 is a
photograph taken from the Soltrol experiments and reproduced with similar black-and-white shading.
The Soltrol film shows no similar tendency to truncate.
In the PCE experiment notice the golf-ball dimples on the air-filled pore body. These 'dimples' do
not appear in the Soltrol experiment. They represent small lenses of PCE 'floating' on the air-water
interface within the pore (see, eg, Figure 10-2). These lenses would not occur if the PCE completely
surrounded the air-filled pore body. Recall the typical eye-shape of micromodel pores displayed in
Figure 8-8. The smallest dimension is from the 'floor' to the 'roof of the pore. We infer that the PCE film
truncates somewhere along the floor and roof. Beyond that point the PCE has coalesced into small
lenses along the air-water interface. In the Soltrol experiment stereo microscope observations revealed
that the Soltrol layer completely surrounded the air.
Our hypothesis was correct: a non-spreading fluid has a different behavior in the vadose zone.
Although it may fill or 'smile' into some pore throats, and may form or coat pendular rings, It does not
spread out as a film. The organic liquid coalesces into lenses and pockets. Large portions of the pore
space do not have a layer of organic liquid between the air and the water.
Implications for aquifer contamination -
We have not quantitatively investigated the amount of residual saturation expected for
non-spreading organic liquids in vadose zone soils, but we can speculate. Halogenated hydrocarbons
such as carbon tetrachloride and PCE are non-spreading. They are also much denser than water (see
left side of Figure 10-1). Greater density means a greater (negative) buoyancy force, and a lower
residual saturation (see Figure 10-3). The micromodel experiments indicate that non-spreading organic
liquids may have a greater tendency to by-pass portions of the pore space, leading to lower residual
saturations (consider the lack of PCE in right-hand air-filled pore in Figure 10-16). The experiments also
indicated the absence of an interconnecting film, leading to a lower residual saturation. In summary, we
hypothesize that non-spreading organic liquids have lower residual saturations in the vadose zone, than
do spreading organic liquids, (see Lenhard and Parker, 1987b)
- 219 -
-------
FIGURE 10-16. A photomicrograph of non-spreading PCE in a micromodel.
- 220 -
-------
FIGURE 10-17.
A photomicrograph of non-spreading PCE in a micromodel. This is
a close-up to the photo shown in Figure 10-16.
- 221 -
-------
FIGURE 10-18. A photomicrograph of Soltrol in a micromodel. The geometry is
similar to that depicted for PCE in Figure 10-16.
- 222 -
-------
Implications for modeling multi-phase flow -
Parker and Lenhard's (1987) model effectively assumes a zero or neutral spreading coefficient (see
Lenhard and Parker, 1987b). This somewhat contradicts their other assumption that the organic liquid
always forms a layer between the water and air. In any case, it is not clear whether or not this model can
handle the significantly different physics associated with a non-spreading organic liquid.
Implications for phase partitioning -
Coalescing organic liquids do not form films and have a much smaller contact area with the gas and
water phases. The respective mass transfer coefficients must be significantly reduced.
Implications for biotransformation -
Biosurfactants that are a by-product of biological activity could change interfacial tensions enough to
alter a non-spreading organic liquid to a spreading organic liquid. This should improve mass transfer
coefficients.
Other implications for aquifer remediation -
Induced volatilization, vacuum extraction, and similar vadose zone remediation schemes may be a
much less effective strategy for non-spreading organic liquids, than for spreading organic liquids,
because of reduced mass transfer coefficients.
- 223 -
-------
REFERENCES
Ababou, R., L. W. Gelhar, and D. McLaughlin. 1988. Three-dimensional flow in random porous media.
Ralph M. Parsons Laboratory, Tech. Rpt. 318. Massachusetts Institute of Technology Department of
Civil Engineering. 833pp.
Abriola, L. M., and G. F. Pinder. 1985a. A multiphase approach to the modeling of porous media
contaminated by organic compounds 1. Equation development. Water Resources Research, vol.21,
no.1, pp.11-8.
Abriola, L. M. , and G. F. Pinder. 1985b. A multiphase approach to the modeling of porous media
contaminated by organic compounds 2. Numerical Simulation. Water Resources Research, vol.21,
no.1, pp.19-26.
Adamson, A. W. 1982. Physical Chemistry of Surfaces, 4th edition. Wiley, New York.
Albertson, M., G. Baldauf, F. Birk, Th. Dracos, A. Golwer, H. Horauf, W. Kab, P. Muntzer, G.
Ruddig4er, H.O. Schiegg, F. Schweisfurth, F. Schwille, H. Tangermann, H.G. Van Waegeningh,
P. Werner, L. Zilliox, and L. Zwitting. 1986. Evaluation and treatment of cases of oil damage with
regard to groundwater protection, part 1, Scientific fundamental principles for understanding the
behavior of oil in the ground. LTwS-Nr. 20, Federal Office of the Environment, West Germay, english
translation available from Batelle Pacific Northwest Laboratory.
Amaufule J. 0., and L. L. Handy. 1982. The effect of interfacial tensions on relative oil/water
permeabilities of consolidated porous media. SPE Journal, vol.22, no.3, pp.371-381.
ASTM. 1986. Annual Book of ASTM Standards. American Society for Testing and Materials, Philadelphia,
PA.
Amott, E. 1959. Observations relating to the wettability of porous rock. AIME Transactions, vol.216,
pp.156-62.
Amoozegar, A., A.W. Warrick, and W.H. Fuller. 1986. Movement of selected organic liquids into dry
soils. Hazardous Waste & Hazardous Materials, vol.3, no.1, pp.29-41.
Anderson, W. G. 1986a. Wettability literature survey — part 1: rock-oil-brine interactions and the effects
of core handling on wettability. Journal of Petroleum Technology, vol.38, no.10, pp.1125-49.
Anderson, W. G. 1986b. Wettability literature survey — part 2: wettability measurement. Journal of
Petroleum Technology, vol. 38, no.11, pp. 1246-62.
Anderson, W. G. 1987a. Wettability literature survey — part 4: the effects of wettability on capillary
pressure. Journal of Petroleum Technology, vol. 39, no.10, pp.1283-1300.
Anderson, W. G. 1987b. Wettability literature survey — part 5: the effects of wettability on relative
permeability. Journal of Petroleum Technology, vol. 39, no.11, pp. 1453-68.
- 224 -
-------
Anderson, W. G. 1988. Wettability literature survey — part 6: the effects of wettability on water flooding.
Journal of Petroleum Technology, vol. 39, no. 12, pp. 1605-22.
Araktingi, U.G., D.C.Brock, and F.M.Orr. 1988. Viscous fingering in heterogeneous porous media.
presented at Fall 1888 Amer. Geophys. Union Meeting. Abstract in EOS, Transactions, American
Geophysical Union, vol.69, no.44, pp. 1204-5.
Baehr, A. L. 1987. Selective transport of hydrocarbons in the unsaturated zone due too aqueous and
vapor phase partitioning. Water Resources Research, vol.23, no.10, pp.1926-1938.
Baehr, A. L., and M. Y. Corapcioglu. 1984. A predictive model for pollution from gasoline in soils and
ground water, in proceedings of Petroleum Hydrocarbons and Organic Chemicals in Ground Water,
NWWA, Houston, TX, pp. 144-56.
Baehr, A. L., and M. Y. Corapcioglu. 1987. A compositional multiphase model for groundwater
contamination by petroleum products 2, numerical solution. Water Resources Research, vol.23,
no.1, pp.201-13.
Bitton, G. and K. C. Marsell. (eds.) 1979. Adsorption of microorganisms to surfaces, Wiley, New York.
Bobek, J. E.,C. C. Mattax, and M. O. Denekas. 1958. Reservoir rock wettability — its significance and
evaluation. Trans. AIME, vol.213, pp.155-60.
Bouchard, D.C., R.M. Powell and D.A. Clark. 1987. Alkylammonium cation effects on aquifer material
sorptivity, Agronomy Abstracts, vol.79, no. 166.
Bowman, R.S., M.E.Essington, and G.A.O'Conner. 1981. Soil sorption of nickel: influence of solution
composition. SSSAJ, v.45, no.5, pp.680-5.
Boyd, S.A., M.M. Mortland and C.T. Chiou. 1987. Organo-clays as sorbents for organic contaminants,
Agronomy Abstracts, vol.79, no. 166.
Boyer, R. F., ed. 1970. Styrene Polymer, in Encyclopedia of Polymer Science and Technology, Bikales,
N. M., ed., vol.13, Interscience Publishers, New York, pp.128-447.
Brown, C. E., and E. L. Neustadter. 1980. The wettability of oil/water/silica systems with reference to oil
recovery. Journal of Canadian Petroleum Technology, vol.19, no.3, pp.100-10.
Burmaster, D. E., and R. H. Harris. 1982. Groundwater contamination: an emerging threat. Technology
Revue, vol.84, no.7, pp.50-62.
Burris, D. R., and W. A. Maclntyre. 1986. Solution of hydrocarbons in a hydrocarbon-water system with
changing phase compositon due to evaporation. Environmental Science and Technology, vol.20,
no.3, pp.296-299.
Callahan, M. A., and others. 1979. Water- Related Fate of 129 Priority Pollutants (2 volumes). EPA
Office of Water Planning and Standards (WH-553), Washington, D.C., EPA-440/4-79-029(a and b).
- 225 -
-------
Carman, P. C. 1937. Fluid flow through granular beds. Transactions: Institute of Chemical Engineers
(London), vol.15, pp.150-66.
Gary, J. W., J. F. McBride, and C. S. Simmons. 1989. Observation of water and oil infiltration into soil:
some simulation challenges. Water Resources Research, vol.25, no.1, pp.73-80.
Chaffee, W. T., and R. A. Weimar. 1983. Remedial programs for ground-water supplies contaminated
by gasoline, in proceedings of Third National Symposium on Aquifer Restoration and Ground-Water
Monitoring, National Water Well Association, pp.39-46.
Chatzis, I. 1982. Photofabrication technique of twol-dimensional glass micromodels. PRRC report no.
82-12, New Mexico Institute of Mining and Technology, Socorro, NM.
Chatzis, I., and N. R. Morrow. 1981. Correlation of capillary number relationships for sandstones.
paper SPE 10114, presented at 1981 SPE Annual Technical Conference and Exhibition, San Antonio,
TX.
Chatzis, I., and N. R. Morrow. 1984. Correlation of capillary number relationships for sandstone. SPE
Journal, vol.24, no.5, pp.355-562.
Chatzis, I., and F. A. L. Dullien. 1983. Dynamic immiscible displacement mechanisms in pore
doublets:-theory versus experiment. Journal of Colloid and Interface Science, vol.91, no.1,
pp.199-222.
Chatzis, I., N. R. Morrow, and H. T. Lim. 1983. Magnitude and detailed structure of residual oil
saturation. SPE Journal, vol.23, no.2, pp.311-25.
Chatzis, I., M. S. Kuntamukkula, and N. R. Morrow. 1984. Blob-size distribution as a function of
capillary number in sandstones, paper SPE 13213, presented at 1984 SPE Annual Technical
Conference and Exhibition, Houston, TX.
Chatzis, I., M. S. Kuntamukkula, and N. R. Morrow. 1988. Effect of capillary number on the
microstructure of residual oil in strongly water-wet sandstones. SPE Reservoir Engineering, vol.3,
no.3, pp.902-912.
Chilingar, G. V.,andT. F. Yen. 1983. Some notes of wettability and relative permeability of carbonate
reservoir rocks. Energy Sources, vol.7, no.1, pp.67-75.
Chiou, C.T., V.H.Freed, D.W. Schmedding and R. L. Kohnert. 1977. Partition coefficient and
bioaccumulation of selected organic chemicals. Environmental Science and Technology, vol.11,
no.5, pp.475-8.
Conrad, S. H., J. L. Wilson, W. R. Mason, and W. Peplinski. 1989. Observing the transport and fate of
petroleum hydrocarbons in soils and in ground water using flow visualization techniques, in
Proceedings, Symposium on Environmental Concerns in the Petroleum Industry, American
Association of Petroleum Engineers, Palm Springs, CA, May 10, 1989.
Convery, M. P. 1979. The Behavior and Movement of Petroleum Products in Unconsolidated Surficial
Deposits. M.S. Thesis, University of Minnesota.
- 226 -
-------
Corapcioglu, M. Y., and A. L. Baehr. 1987. A compositional multiphase model for groundwater
contamination by petroleum products 1. theoretical considerations. Water Resources Research,
vol.23, no.1, pp.191-200.
Craig, F. F. 1971. Reservoir Engineering Aspects of Waterfloading. SPE Monograph Series, no.3, Dallas,
TX.
Dagan, G. 1985. Stochastic modeling of groundwater flow by unconditional and conditional probabilities:
the inverse problem. Water Resources Research, vol.21, no.1, pp.65-72.
Dagan, G. 1986. Statistical theory of groundwater flow and transport: pore to laboratory, laboratory to
formation, and formation to regional scale. Water Resources Research, vol.22, no.9, pp.120S-34S.
Davis, J. A., and S. C. Jones. 1968. Displacement mechanisms of miscellar solutions. Jour. Pet. Tech.,
Dec. 1968, pp.1415-28.
Dean, J. A. (ed.). 1979. Lange's Handbook of Chemistry (12th edition). McGraw-Hill, New York.
Denekas, M. 0., C. C. Mattax, and G. T. Davis. 1959. Effect of crude oil components on rock
wettability. Journal of Petroleum Technology, Nov., pp.330-3.
de Pastrovich, T.L., Y.Baradat, R.Barthel, A. Chiarelli, and D.R. Fussell. 1979. Protection of
Groundwater from Oil Pollution. Report 3/79, CONCAWE, Den Haag, the Netherlands.
Donahue, D. J., and F. E. Bartell. 1952. The boundary tension at water — organic liquid interfaces.
Journal of Physical Chemistry, vol.56, p.480.
Donaldson, E. C., R. D. Thomas, and P. B. Lorenz. 1969. Wettability determination and its effects on
recovery efficiency. SPE Journal, vol.9, no.1, pp. 13-20.
Dreisbach, R. R. 1955-1961. Properties of Chemical Compounds (3 volumes). American Chemical
Society, Washington, D.C.
Dullien, F. A. L., F. S. Y. Lai, and I. F. MacDonald. 1986. Hydraulic continuity of residual wetting phase
in porous media. J. Coll. Sci., vol.109, no.1, pp.201-18.
Eames, V. 1981. Influence of Water Saturation on Oil Retention Under Field and Laboratory Conditions.
University of Minnesota M.S. Thesis.
Eastman Kodak Company. 1975. Decorating Glass Using Kodak Photosensitive Resists. Kodak
Publication No. P-245, 4 pp.
Eastman Kodak Company. 1979. Photofabrication Methods with Kodak Photoresists. Kodak Publication
No. G-184, 32 pp.
Eckberg, D. K.,andD. K. Sunada. 1984. Nonsteady three-phase immiscible fluid distribution in porous
media. Water Resources Research, vol.20, no. 12, pp. 1891-7.
EPA. 1979. Waste alert. EPA Journal, vol.5, no.2, p.12.
- 227 -
-------
EPA. 1980. Proposed Groundwater Protection Strategy. Office of Drinking Water, Washington, D.C.
EPA. 1982. Massive voluntary cleanup to help with hazardous waste removal. EPA Journal, vol.9, no.5,
pp.24-5.
EPA. 1983. Hazardous Waste Site Descriptions: National Priority List — Final Rule. Office of Solid Waste
and Emergency Response (WH-5624), Washington, D.C., HW-8.1.
Epstein, S. S., L. O. Brown, and C. Pope. 1982. Hazardous Waste in America. Sierra Club Books, San
Francisco, CA, 593pp.
Ewing, R. E., T. F. Russel, and L. C. Young. 1989. An anisotropic coarse-grid dispersion model of
heterogeneity and viscous fingering in five-spot miscible displacement that matches experiments
and fine-grid models, in proceeedings of Tenth Annual SPE Symposium on Reservoir Simulation,
February 6-8, Houston, TX, pp.447-465.
Faust, C. R. 1985. Transport of immiscible fluids within and below the unsaturated zone: a numerical
model. Water Resources Research, vol.21, no.4, pp.587-596.
Feenstra, S., and J. Coburn. 1986. Subsurface contamination from spills of denser than water
chlorinated solvents. Bull. Calif. Water Poll. Control Assoc. vol.23, no.4, pp.26-34.
Felsenthal, M. 1979. A statistical study of core waterflood parameters. Journal of Petroleum
Technology, vol.31, no. 10, pp. 1303-4.
Ferrand, L. A., P. C. D. Milly, and G. F. Pinder. 1986. Dual-gamma attenuation for the determination of
porous medium saturation with respect to three fluids. Water Resources Research, vol.22, no.12,
pp. 1657-1663.
Fowkes, F. M. 1967. Attractive forces at solid-liquid interfaces, in Wetting. Society of Chemical
Industry, London, pp.3-31.
Gaudin, A.M., A.F. Witt, and T.G. Decker. 1963. Contact angle hysteresis — principles and application
of measurement methods. Trans. AIME, vol.226, pp. 107-12.
Gelhar, L. W. 1986. Stochastic subsurface hydrology from theory to applications. Water Resources
Research, vol.22, no.9, pp.135S-145S.
Girifalco, L. A., and R. J. Good. 1957. A theory for the estimation of surface and interfacial energies,
I: derivation and application to interfacial tension. Journal of Physical Chemistry, vol.61, p.904.
Gvirtzman, H., M. Magaritz, E. Klein, and A. Nader. 1987. A scanning electron microscopy study of
water in soil. Transport In Porous Media, no.2, pp.83-93.
Hillel, D. 1980. Fundamentals of Soil Physics. Academic Press, New York, NY, 413p.
Hornof, V. and N. R. Morrow. 1988. Flow Visualiztion of the effects of interfacial tension on
displacement, SPE Reservoir Engineering, vol.3, no.1, pp.251-56.
- 228 -
-------
Jercinovic, D. E. 1984. Petroleum-Product Contamination of Soil and Water in New Mexico. Ground
Water/Hazardous Waste Bureau, New Mexico Environmental Improvement Division, EID/GWH-84/4.
Jones, S.C. and W.O.Roszelle. 1978. Graphical techniques for determining relative permeability from
displacement experiments. Journal of Petroleum Technology, May, pp.807-17.
Kia, S.F. 1988. Modeling of the retention of organic contaminants in porous media of uniform spherical
partricles. Water Research, vol.22, no.10, pp.1301-9.
Kobayashi, H., and B. E. Rittmann. 1982. Microbial removal of hazardous organic compounds.
Environmental Science and Technology, vol.16, no.3, pp.170A-183A.
Kuppusamy, T., J. Sheng, J. C. Parker, and R. J. Lenhard. 1987. Finite element analysis of multiphase
immiscible flow through soils. Water Resources Research, vol.23, pp.625-31.
Kyte, J. R., R. J. Stanclift Jr., S. C. Stephan Jr., and L. A. Rapoport. 1956. Mechanism of water
flooding in the presence of free gas. Trans. AIME, vol.207, pp.215-21.
Lake, L. W., and H. B. Carroll, Jr. - eds. 1986. Reservoir Characterization, Harcourt Brace Jovanovich,
Publishers, Orlando, FL., 659pp.
Lam, K. 1979. Three-phase gravity drainage, unpublished report, Petroleum Recovery Research
Center, New Mexico Tech, Socorro, NM.
Lenhard, R. J., and J. C. Parker. 1987a. A model for hysteretic constitutive relations governing
multiphase flow, 2. permeability - saturation relations. Water Resources Research, vol.23,
pp.2197-96.
Lenhard, R. J., and J. C. Parker. 1987b. Measurement and prediction of saturation - pressure
relationships in three phase porous media systems. Journal of Contaminant Hydrology, vol.1,
pp.407-24.
Lenhard, R. J., and J. C. Parker. 1988a. Experimental validation of the theory for extending two phase
saturation-pressure relations to three phase systems for onotonic drainage paths. Water Resources
Research, vol.24, pp.373-80.
Lenhard, R. J., and J. C. Parker. 1988b. Estimation of oil volumes in soils from observed fluid levels in
monitoring wells, presented at Fall 1888 Amer. Geophys. Union Meeting. Abstract in EOS,
Transactions, American Geophysical Union, vol.69, no.44, pp. 1212-1213.
Lenhard, R. J., and J. C. Parker. 1989. A model for hysteretic constitutive relations governing
multiphase flow, 3. refinements and numerical simulations. Water Resources Research, vol.25,
pp. 1727-36.
Lenhard, R. J., J. H. Dane, and J. C. Parker. 1988. Measurement and simulation of one-dimensional
transient three-phase flow for monotonic liquid drainage. Water Resources Research, vol.24,
pp. 853-63.
- 229 -
-------
Lyman, W. J., W. F. Reehl, and D. H. Rosenblatt (eds.). 1982. Handbook of Chemical Property
Estimation Methods. McGraw-Hill, New York.
Mantoglou, A., and L. W. Gelhar. 1987. Stochastic modeling of large-scale transient unsaturated flow in
stratified soils. Water Resources Research, vol.23, no.1, pp.57-68.
Marsell, K.C. (ed.) 1985. Microbial adhesion and aggregation, Springer-Verlag, KG, Berlin.
Mason, W. R., S. H. Conrad, and J. L. Wilson. 1988. Micromodel study of organic liquid advance into a
soil, presented at Spring 1988 Amer. Geophys. Union Meeting. Abstract in EOS,Transactions,
American Geophysical Union, vol.69, no. 16, pp.370.
Mason, W.M., W. Peplinski, J.L. Wilson and S.H. Conrad. 1989. Pore level flow visualizaion of three
phase fluid flow, presented at Spring 1989 Amer. Geophys. Union Meeting, Abstract in
EOS,Transactions, American Geophysical Union vol. 70.
Mattax, C. C., and J. R. Kyle. 1961. Ever see a water flood? Oil and Gas Jour., Oct.16, 1961,
pp.115-128.
Mattson, E. D., A. M. Parsons, D. B. Stephens, K. Black, and K. Flanigan. 1988. Field simulation of
waste impoundment seepage in the vadose zone, in proceedings of FOCUS on Southwestern Ground
Water Issues Conference, March 23-25, National Water Well Association, Dublin, OH.
Maugh II, T. H. 1979. Toxic waste disposal a growing problem. Science, vol.204, pp.819-23.
McCarty, P.L., M. Reinhard, and B. E. Rittmann. 1981. Trace organics in groundwater.
Environmental Science and Technology, vol.15, no.1, pp.40-51.
McCord, J. T., D. B. Stephens, J. L. Wilson. 1988a. Field-scale variably saturated flow and transport in
a sloping uniform porous media: field experiments and numerical simulations, in proceedings of
International Workshop on Validation of Flow and Transport Models for the Unsaturated Zone. May
23-26, Ruidoso, NM, New Mexico State University.
McCord, J. T., D. B. Stephens, and J. L. Wilson. 1988b. Field-scale variably saturated flow and
transport: the roles of hysteresis and state-dependent anisotropy, in proceedings of NATO
Advanced Study Institute on Recent Advances in Modeling Hydrologic Systems, July 9-22, Sintra,
Portugal.
McKee, J. E., F. B. Laverty, and R. H. Hertel. 1972. Gasoline in groundwater. Journal of the Water
Pollution Control Federation, vol.44, no.2, pp.293-302.
McKellar, M., and N. C. Wardlaw. 1988. A Method of Viewing "Water" and "Oil" Distribution in
Native-State and Restored-State Reservoir Core. AAPG Bulletin, V. 72, No. 6, pp. 765-771.
Melrose, J. C., and C. F. Brandner. 1974. Role of capillary forces in determining microscopic
displacement efficiency for oil recovery by waterflooding. Journal of Canadian Petroleum
Technology, vol.13, no.4, pp.54-62.
- 230 -
-------
Mohanty, K. K., H. T. Davis, and L. E. Scriven. 1980. Physics of oil entrapment in water-wet rock.
paper SPE 9406, presented at 1980 SPE Annual Technical Conference and Exhibition, Dallas, TX.
Moissis, D. E., C. A. Miller, and M. F. Wheeler. 1989. Simulation of miscible viscous fingering using a
modified method of characteristics: effects of gravity and heterogeneity, in proceedings of Tenth
Annual SPE Symposium on Reservoir Simulation, February 6-8, Houston, TX, pp.431-446.
Moore, T. F., and R. L. Slobod. 1956. The effect of viscosity and capillarity on the displacement of oil by
water. Producers Monthly, vol.20, pp.20-30.
Morrow, N.R. 1970. Physics and thermodynamics of capillary action in porous media. Industrial
Engineering Chem., vol.62, no.6, pp.32-56.
Morrow, N. R. 1979. Interplay of capillary, viscous and buoyancy forces in the mobilization of residual
oil. Journal of Canadian Petroleum Geology, vol.18, no.3, pp.35-46.
Morrow, N. R. 1984. Measurement and Correlation of Conditions for Entrapment and Mobilization of
Residual Oil. Report NMERDI 2-70-3304, New Mexico Energy Research and Development Institute,
Santa Fe, NM.
Morrow, N. R., P. J. Cram, and F. G. McCaffery. 1973. Displacement studies in dolomite with
wettability control by octanoic acid. SPE Journal, vol.13, pp.221-32.
Morrow, N. R., and B. Songkran. 1981. Effect of trapping and buoyancy forces on non-wetting phase
trapping in porous media, in Surface Phenomena in Enhanced Oil Recovery, D.O.Shah (ed.), Plenum
Publishing Corporation.
Morrow, N. R., and I. Chatzis. 1982. Measurement and Correlation of Conditions for Entrapment and
Mobilization of Residual Oil. Report DOE/BC/10310-20, Department of Energy.
Morrow, N. R., H. T. Lim, and J. S. Ward. 1986. Effect of crude-oil-induced wettability changes on oil
recovery. SPE Formation Evaluation, vol.1, no.1, pp.89-103.
Morrow, N. R., I. Chatzis, and J. J. Taber. 1988. Entrapment and mobilization of residual oil in bead
packs. SPE Reservoir Engineering, vol.3, no.3, pp.927-934.
Mortland, M.M., S. Shaobaio and S.A. Boyd. 1986. Clay-organic complexes as adsorbents for phenol
and chlorophenols, Clays & Clay Minerals, vol.34, no.5, 581-5.
Ng, K. M., H. T. Davis, and L. E. Scriven. 1978. Visualization of blob mechanics in flow through porous
media. Chemical Engineering Science, vol.33, pp. 1009-17.
Parker, J. C., R. J. Lenhard and T. Kuppusamy. 1987. A parametric model for constitutive properties
governing multiphase flow in porous media. Water Resources Research, vol.23, pp.618-24.
Parker, J. C., and R. J. Lenhard. 1987. A model for hysteretic constitutive relations governing
multiphase flow, 1. saturation - pressure relations. Water Resources Research, vol.23, no.12,
pp.2187-96.
- 231 -
-------
Pathak, P., H. T. Davis, and L. E. Scriven. 1982. Dependence of residual non-wetting liquid on pore
topology, paper SPE 11016, presented at 1982 SPE Annual Technical Conference and Exhibition,
New Orleans.
Pfannkuch, H. O. 1984. Determination of the contaminant source strength from mass exchange
processes at the petroleum-ground-water interface in shallow aquifer systems, in proceedings of
Fourth National Symposium on Aquifer Restoration and Groundwater Monitoring, NWWA, Columbus,
OH.
Pinder, G. F., and L. M. Abriola. 1986. On the simulation of nonaqueous phase compounds in the
subsurface. Water Resources Research, vol.22, no.9, pp.109s-19s.
Powers, S.E., Y. M. Chen, L. M. Abriola, and W. J. Weber. 1988. The significance of non-equilibrium
effects on interphase partitioning of organic contaminants in multiphase systems [abst.]. Eos,
Transactions, American Geophysical Union, vol.69, no.44, p. 1201.
Pringle, J.H. and M. Fletcher. 1983. Influence of substratum wettability on attachment of freshwater
baceria to soil surfaces, Applied Environmental Microbiology, vol.45, no.3, 811-7.
Roberts, J. R., J. A. Cherry, and F. W. Schwartz. 1982. A case study of a chemical spill:
polychlorinated biphenyls (PCBs) 1. history, distribution, and surface translocation. Water
Resources Research, vol.18, no.3, pp.523-34.
Roberts, P. V., M. Reinhard, and A. J. Valocchi. 1982. Movement of organic contaminants in
groundwater: implications for water supply. Journal of the American Water Works Association,
vol.74, pp.408-13.
Salathiel, R. A. 1973. Oil recovery by surface film drainage in mixed-wettability rocks. Journal of
Petroleum Technology, vol.25, pp. 1216-24.
Schiegg, H.O. 1980. Fundamentals, setups, and results of laboratory experiments on oil propagation in
aquifers. VAX-Mitteilung no. 43, Versuchsanstalt fur Wasserbau, Hydrologie und Glaziologie, ZTH,
Zurich, english translation available from Batelle Pacific Northwest Laboratory.
Schiegg, H.O. and J.F. McBride. 1987. Laboratory setup to study two-dimensional multiphase flow in
porous media, in proceedings of Petroleum Hydrocarbons and Organic Chemicals in Ground Water,
NWWA, Houston, TX, pp. 371-95.
Schwille, F. 1967. Petroleum contamination of the subsoil — a hydrological problem, in The Joint
Problems of the Oil and Water Industries (P.Hepple — ed.), Elsevier, Amsterdam, pp.23-53.
Schwille, F. 1981. Groundwater pollution in porous media by fluids immiscible with water, in Quality of
Groundwater (W. van Duijvenbooden et al. — eds.), Elsevier, Amsterdam, pp.451-63.
Schwille, F. 1984. Migration of organic fluids immiscible with water in the unsaturated zone, in Pollutants
in Porous Media: The Unsaturated Zone Between Soil Surface and Groundwater (B. Yaron, G. Dagan,
and T. Goldschmid — eds.), Springer-Verlag, New York, pp.27-48.
- 232 -
-------
Schwille, F. 1988. Dense Chlorinated Solvents in Porous and Fractured Media. Lewis Publishers,
Chelsea, Ml. 146pp.
Sharma, M. M., and R. W. Wunderlich. 1985. The alteration of rock properties due to interactions with
drilling fluid components, paper SPE 14302 presented at 1985 SPE Annual Technical Conference and
Exhibition, Las Vegas.
Soll.W.E. and M.A.Celia. 1988. A pore-scale model of three-phase immiscible fluid flow, presented at
Fall 1988 Amer. Geophys. Union Meeting. Abstract in EOS,Transactions, American Geophysical
Union, vol.69, no.44, pp.1189-90.
Stephens, D. B., and S. Heermann. 1988. Dependence of anisotropy on saturation in a stratified sand.
Water Resources Research, vol.24, no.5, pp.770-778.
Taber, J. J. 1969. Dynamic and static forces required to remove a discontinuous oil phase from porous
media containing both oil and water. SPE Journal, vol.2, pp.3-12.
Taber, J. J. 1981. Research on enhanced oil recovery: past, present, and future, in Surface Phenomena
in Enhanced Oil Recovery (D.O.Shah ed.), Plenum, New York, pp. 13-52.
Treiber, L. E., D. L. Archer, and W. W. Owens. 1972. A laboratory evaluation of the wettability of fifty
oil producing reservoirs. SPE Journal, vol.12, pp.531-40.
Tuck, D.M., P.R. Jaffe, D.A. Crerar, and R.T. Mueller. 1988. Enhancing recovery of immobile residual
non-wetting hydrocarbons form the unsaturated zone using surfactant solutions, in proceedings of
Petroleum Hydrocarbons and Organic Chemicals in Ground Water, NWWA, Houston, TX, pp.457-78.
van Dam, J. 1967. The migration of hydrocarbons in water-bearing stratum, in: The Joint Problems of
the Oil and Water Industries (P.Hepple — ed.}, Elsevier, Amsterdam, pp.55-88.
van Genuchten, M. Th. 1980. A closed-form equation for predicting the hydraulic conductivity of
unsaturated soils. Soil Sci. Soc. Amer. J., vol.44, pp. 892-8.
van Loosdrecht, M.C.M., J. Lykllema, W. Norde, G. Schraa, and A.J.B. Zehnder. 1987a. The role of
bacterial cell wall hydrophobicity in adhesion, Applied Environmental Microbiology, vol.53,
no.8,pp.1893-7.
van Loosdrecht, M.C.M., J. Lykllema, W. Norde, G. Schraa, and A.J.B. Zehnder. 1987a.
Electrophoretic mobility and hydrophobicity as a measure to predict the initial steps of bacterial
adhesion, Applied Environmental Microbiology, vol.53, no.8,pp.1898-901.
Villaume, J. F. 1985. Investigations at sites contaminated with dense non-aqueous phase liquids.
Groundwater Monitoring Review, vol.5, no.2, pp.60-74.
Vomocil, J. A. 1965. Porosity, in Methods of Soil Analysis, Part 1, C. A. Black (Ed.), American Society of
Agronomy, Madison, Wl, pp.299-314.
Wardlaw, N. C. 1982. The effect of geometry, wettability, viscosity, and interfacial tension on trapping in
single pore-throat pairs. Journal of Canadian Petroleum Technology, vol.21, no.3, pp.21-7.
- 233 -
-------
Wardlaw, N. C., and R. P. Taylor. 1976. Mercury capillary pressure curves and the interpretation of pore
structure and capillary behavior in reservoir rocks. Bulletin of Canadian Petroleum Geology, vol.24,
no.2, pp.225-262.
Weast, R. C. (ed.). 1981. Handbook of Chemistry and Physics. CRC Press, Boca Raton, Fla.
Welty, C. 1989. Stochastic analysis of viscosity and density variability on macrodispersion in
heterogeneous porous media. Ph. D. thesis, Dept. of Civil Engr., Mass. Inst. of Tech., Cambridge,
Ma.
Welty, C. and L.W. Gelhar, 1987. Stochastic analysis of the effects of viscosity variability on
macrodispersion in heterogeneous porous media, presented at Fall 1987 Amer. Geophys. Union
Meeting. Abstract in EOS.Transactions, American Geophysical Union, vol.68, no.44, pp.1189-90.
Welty, C. and L.W. Gelhar, 1989. Stochastic analysis of the effects of density variability on
macrodispersion in heterogeneous porous media, presented at Spring 1989 Amer. Geophys. Union
Meeting. Abstract in EOS, Transactions, American Geophysical Union, vol.70, no. 15, pp.340.
Williams, D. E., and D. G. Wilder. 1971. Gasoline pollution of a ground-water reservoir — a case
history. Ground Water, vol.9, no.6, pp.50-4.
Wilson, J. L. 1984. Double-cell hydraulic containment of pollutant plumes, in proceedings of Fourth
National Symposium on Aquifer Restoration and Groundwater Monitoring, NWWA, Columbus, OH.
Wilson, J. L. 1988. The role of wetting in environmental problems, in proceedings of Conference on
Fundamental Research Needs in Environmental Engineering, Assoc. Environ. Engr. Professors,
Washington, DC.
Wilson, J. L., and S. H. Conrad. 1984. Is physical displacement of residual hydrocarbons a realistic
possibility in aquifer restoration? in proceedings of Petroleum Hydrocarbons and Organic Chemicals
in Ground Water, NWWA, Houston, TX, pp.274-98.
Wilson, J. L., S. H. Conrad, E. Hagan, W.R. Mason, and W. Peplinski. 1988. The pore level spatial
distribution and saturation of organic liquids in porous media, in proceedings of Petroleum
Hydrocarbons and Organic Chemicals in Ground Water, NWWA, Houston, TX, pp.107-133.
Wilson, J.L., W. Cox and W.M. Mason. 1988. Flow visualization of two-phase flow in a fracture.
videotape presented at Spring 1988 Amer. Geophys. Union Meeting, Abstract in EOS, vol. 69, no.
16, p.353.
Yadav, G. D., F. A. L. Dullien, I. Chatzis, and I. F. Macdonald. 1987. Microscopic Distribution of Wetting
and Nonwetting Phases in Sandstones During Immiscible Displacements. SPE Reservoir Engineering,
vol. 2, may, pp. 137-147.
Yeh, T.-C. Jim, L. W. Gelhar, A. L. Gutjahr. 1985a. Stochastic analysis of unsaturated flow in
heterogeneous soils, 1, statistically isotropic medium. Water Resources Research, vol.21, no.4,
pp.447-456.
- 234 -
-------
Yeh, T.-C. Jim, L. W. Gelhar, A. L. Gutjahr. 1985b. Stochastic analysis of unsaturated flow in
heterogeneous soils, 2, statistically anisotropic media with variable a. Water Resources Research,
vol.21, no.4, pp.457-464.
Yeh, T.-C. Jim, L. W. Gelhar, A. L. Gutjahr. 1985c. Stochastic analysis of unsaturated flow in
heterogeneous soils, 3, observations and applications. Water Resources Research, vol.21, no.4,
pp.465-472.
- 235 -
-------
APPENDIX A:
QUANTITATIVE TWO-PHASE RESIDUAL SATURATION
RESULTS, WITH STYRENE AS THE ORGANIC PHASE
The pore and blob cast experiments described in Section 7 produced quantitative results, as well as
results pertaining to flow visualization. Unfortunately, the experimental methods made these quantitative
results too inaccurate to be included with the main body of two-phase residual saturation data presented
in Section 9. There were several reasons for the inaccuracies, as reviewed in Section 7.
Perhaps the largest source of error in the styrene experiments was the mass of the TFE column itself.
Another factor contributing to the lower accuracy in the results was the density of the styrene. In
comparison to Soltrol-130, styrene is the denser — 0.90 versus 0.75 g/cm3. The column was weighed,
during the gravimetric determinations of saturation, on the high capacity Mettler PM 11 balance, which
has an accuracy of only 0.1 grams. As a result of these problems the styrene experiments had saturation
errors of ±6-8%, while the short-column experiments had errors of only ±2-3%. When we considered
that 6-8% was sometimes up to 50% of the residual saturation measurement, it seemed prudent to
consider these less accurate results separately.
Another reason for excluding the styrene data from the other quantitative data was that the residual
saturations obtained were consistently lower than those obtained with the short-columns. In Section 7 we
hypothesize that time constraints, put on the experiment by styrene's ever increasing viscosity, did not
allow enough time for the column to come to equilibrium. The wettability of the column walls was also
found to affect the amount of time required to reach an equilibrium during the organic liquid flood, and
may have influenced the residual styrene saturations.
EXPERIMENTAL RESULTS
The first column in the Table A-1 describes the experimental column, and the experimental run
performed in it. There were three TFE columns constructed, two of which are represented here. For
example, experiment 0-6 was the sixth experiment performed in TFE column 0, and experiment 2-1 was
the first experiment performed in column 2. Column 1 was used exclusively for three-phase
experiments, which were non-quantitative, so it is not represented in this table. Porosity values in column
2, maximum organic liquid saturations, S0 , in column 3, and residual organic saturations, Sor , in
column 4, were calculated as described in Section 7.
Experiments 0-1 through 0-4 and 2-1 were homogeneous two-phase runs using the Sevilleta soil and
styrene. The Sor data set has an average value of 0.16 and a standard deviation of 0.04. In comparison
to the glass short column results (see Section 9), they are uniformly lower.
In experiment 0-6, Soltrol-130 was flooded into, and drained out of, a TFE column following
procedures dictated by the viscosity related time constraints of the styrene procedure. The residual
- 236 -
-------
Column
0-1 homogeneous
0-2 homogeneous
0-3 homogeneous
0-4 homogeneous
2-1 homogeneous
0-5 soltrol-130
0-6 soltrol-130
2-2 crushed tuff
0-8 heterogeneous
2-3 heterogeneous
Porosity (%)
32.3
33.5
33.5
31.7
34.6
33.4
34.1
38.7
34.5
37.1
So
0.710+ .091
0.667 ±.087
0.713 ±.094
0.722 ±.101
0.700 ±.086
0.764 ±.039
0.745 ±.038
0.9181.116
0.936+.107
0.896±.133
Scr
0.170 ± .065
0.108 ± .059
0.192 ± .065
0.157 ± .071
0.200 ± .060
0.265 ± .027
0.186 ± .025
0.295 +.082
0.263 + .072
0.407 ±.105
TABLE A-1. Two-phase TFE column residual saturation results; styrene was used unless
otherwise noted.
saturation obtained matched those commonly found when performing experiments with styrene in the
TFE column (19%). In experiment 0-5 Soltrol was again flooded into the TFE column but without time
constraints, following the glass short column procedures of Section 5. Since sufficient time was allowed
for the system to come to equilibrium, a value of residual saturation was obtained (27%) which closely
matched those found when using Soltrol-130 in a glass column. These results suggest that the
differences in residual saturation values were related not to differences in fluid characteristics, but to the
amount of time the liquid/soil system had in which to equilibrate.
Experiment 2-2 was a homogeneous two-phase run done with a crushed and sieved tuff. This was an
attempt to obtain blob casts that would photograph well on black and white film. The black tuff would
stand out from the hardened fluid phases. Other methods of creating black and white photographs were
successful so these casts were never used.
Experiments 0-8 and 2-3 were heterogeneous two-phase runs using Seviileta soil split into two
fractions, as described in Section 7 of this report. High organic saturations, as seen in these
experiments, indicated that the organic liquid displaced the aqueous phase from the larger pores found
in the stringers. This was expected, as a non-wetting fluid will move preferentially to larger pores in order
to decrease its surface to volume ratio, and therefore decrease the surface energy of the system. So
with the larger pores available, more non-wetting liquid can enter the system for a given energy level, or
head. The fact that the columns were packed dry may have also contributed to the higher S0 values. The
dry soil may have settled and compacted during the water saturation process, creating a large pore at
one end of the column. This one 'macro' pore could have filled with styrene and contributed to a high
organic saturation.
The SOT values presented represent 'bulk' residual saturations in that they are averaged over the
whole heterogeneous column. Observations of the core, after it had been cut on a rock saw indicate
however, that most of the residual styrene was trapped in the coarse stringers. A lesser amount,
- 237 -
-------
approximately the same as that observed in the earlier homogeneous experiments (16%), was trapped
in the finer matrix. The measured 'bulk' residual saturation can be compared with a theoretical value
(see Section 9):
Estimated bulk residual styrene saturation =
= (normalized lens volume) x S0 + (normalized matrix volume) x Sor
= (0.40 x 0.90) + (0.60 x 0.16) = (0.36 + 0.09) = 0.45 = 45%
Column 2-3's value of residual saturation, 41%, comes close to the theoretical value, while column 0-8 is
nowhere near the predicted value. This suggests that by-pass trapping may not totally exclude water
from the coarse stringers (refer to Section 9).
The differing S^ values of experiments 0-8 and 2-3 are explained by the flow rate of the water flood.
Experiment 0-8 was flooded with water as described in Section 7 of the report. That is to say, it was
flooded exactly as if it were a homogeneous two-phase experiment. The residual saturation value, 26%,
reflects the relatively larger volume of styrene trapped in the coarser sand stringers. Column experiment
2-3, in contrast, was water-flooded at a much lower rate, approximately 1 ml/minute. This flow rate was
controlled by pushing water through the column with a syringe pump. The low flow rate led to a situation
where almost 45.% of the styrene in the column was bypassed and trapped. In this case, the combination
of viscous and buoyancy forces were not strong enough to overcome the capillary forces holding the
styrene in place. Although some CaCI2 did flow through the stringers, the majority of the stringer's pore
space remained filled with styrene. For photographs of the styrene filled stringers, see Section 9 of this
report.
- 238 -
-------
APPENDIX B
VIDEOTAPES OF MICROMODEL EXPERIMENTS
Videotapes of the micromodel experiments described in Sections 9 and 10 are available on VHS
format tapes. They are treated as Open File Reports, Hydrology Progam, Department of Geoscience,
New Mexico Institute of Mining and Technology, Socorro, New Mexico 87801. The following videotapes
are currently (1990) available:
• Wilson, J.L., W. Cox and W.M. Mason. 1988. Flow visualization of two-phase
flow in a fracture, videotape presented at Spring 1988 Amer. Geophys. Union
Meeting, Abstract in EOS, vol. 69, no. 16, p.353; order as Hydrology Open File
Report 88-1.
• Mason, W.M., S.H. Conrad and J.L. Wilson. 1988. Micromodel study of
organic liquid advance in a soil, poster & videotape presented at Spring 1988
Amer. Geophys. Union Meeting, Abstract in EOS, vol. 69, no. 16, p.370, 1988;
order as Hydrology Open File Report 88-12.
• Mason, W.M., W. Peplinski, J.L. Wilson and S.H. Conrad. 1989. Pore level
flow visualizaion of three phase fluid flow, talk & videotape presented at Spring
1989 Amer. Geophys. Union Meeting, Abstract in EOS, vol. 70, no. 15; order
as Hydrology Open File Report 89-10.
• S.H. Conrad, W.M. Mason and J.L. Wilson. 1989. Two phase immiscible
displacement in a heterogeneous porous media. Order as Hydrology Open File
Report 89-11.
The tapes were originally recorded on Beta II format tape, and subsequently edited for distribution.
The VHS tapes are copies of Beta masters, and are several generations old. They are not of broadcast
quality.
There is a nominal charge for the tapes, to cover media and reproduction expenses. Please contact
the Hydrology Secretary at 505-835-5308 or 5307 for the current charge.
- 239 -
-------
APPENDIX C
SATURATION CURVES AND PROCESSED DATA FOR THE
SHORT COLUMN SEVILLETA SAND EXPERIMENTS
The folowing pages contain capillary pressure-saturation curves for the two-phase fluid pair
combinations tested with the Sevilleta soil. These include air-water, air-organic, and organic-water
curves. Soltrol-130 was used as the organic liquid phase in all trials. Following the curves, Table C-1
tabulates the numerical values of saturation and capillary pressure for each curve, and the temperature
of the experiment at the time of the measurement. Results from twelve experimental trials are
presented.
Seven Soltrol-Water (SW)Capillary Pressure Curves. Figures C-1 through C-8
Two of these curves have drainage, imbibition, and secondary drainage cycles, three curves have
drainage and imbibition cycles, and two curves have the main drainage branch only. Figure C-8 plots all
but one of the primary drainage curves together, illustrating the repeatability of the experiments. Trial 8
(Figure C-2) shows a similar behavior, but is not as consistent as the other curves. The SW trial numbers
corresponds to the numbers in Table 9-2.
Two Soltrol-Air (SA)Capillary Pressure Curves. Figures C-9 through C-11
The first curve has the main drainage branch only, while the second curve has drainage, imbibition,
and secondary drainage cycles. The primary drainage curves are compared in Figure C-11.
Three Air-water Capillary Pressure Curves. Figures C-12 through C-15
Two curves have the main drainage and imbibition cycles, while the third curve has the main
drainage branch only. The primary drainage curves are compared in Figure C-15.
- 240 -
-------
Soltrol-Water Saturation Curve (Drainage-Imbibition)
100
60-
Suction
(cm of H20) 40
-20
0 10
Water Saturation (%)
FIGURE C-1. Soltrol-water saturation curve for SW trial 7.
- 241 -
-------
Soltrol-Water Saturation Curve (Drainage-Imbibition-Drainage)
100
80-
60-
Suction
(cm of H2O) 40
0 10 20 30 40 50 60 70 80 90 100
-20
Water Saturation (%)
FIGURE C-2. Soltrol-water saturation curve for SW trial 8.
- 242 -
-------
Soltrol-Water Saturation Curve (Drainage-Imbibition-Drainage)
Suction
(cm of H20)
0 10 20 30 40 50 60 70 80 90 100
-20
Water Saturation (%)
FIGURE C-3. Soltrol-water saturation curve for SW trial 9.
- 243 -
-------
Soltrol-Water Saturation Curve (Drainage-Imbibition)
100
Suction
(cm of H20)
10 20 30 40 50 60 70 80 90 100
-20
Water Saturation (%)
FIGURE C-4. Soltrol-water saturation curve for SW trial 10.
- 244 -
-------
Soltrol-Water Saturation Curve (Drainage-Imbibition)
Suction
(cm of H20)
-20
60 70 80 90 100
Water Saturation (%)
FIGURE C-5. Soltrol-water saturation curve for SW trial 11.
- 245 -
-------
100-
80-
60-
Suction
(cm of H2O)
40-
20-
-20-
Soltrol-Water Saturation Curve (Drainage of Water)
SW Trial 12
I ^ [
10 20 30 40 50 60 70 80
Water Saturation (%)
I
90
100
FIGURE C-6. Soltrol-water saturation curve for SW trial 12.
- 246 -
-------
Soltrol-Water Saturation Curve (Drainage of Water)
100
Suction
(cm of H2O)
0 10 20 30 40 50 60 70 80 90 100
Water Saturation (%)
FIGURE C-7. Soltrol-water saturation curve for SW trial 13.
- 247 -
-------
Soltrol-Water Saturation Curves (Drainage only)
100-
80-
60-
Suction
(cm of H2O)
40-
20-
-20-
O A
X •
• SW Trial 7
• SW Trial 9
+ SW Trial 10
* SW Trial 11
A SW Trial 12
X SW Trial 13
—i r 1 1 1 1 1 1 \ 1 \ 1 i 1 i ] i ]i
0 10 20 30 40 50 60 70 80 90 100
Water Saturation (%)
FIGURE C-8. Soltrol-water primary drainage curves for SW trials 7-13, minus trial 8.
- 248 -
-------
100-
Soltrol-Air Saturation Curve (Drainage)
SA Trial 1
80-
60-
Suction
(cm of H2O)
40-
20-
-20-
10
I ' I
20 30
\
40
50 60 70 80
Soltrol Saturation (%)
90 100
FIGURE C-9. Soltrol-air saturation curve for SA trial 1.
- 249 -
-------
Soltrol-Air Saturation Curve (Drainage-Imbibition-Drainage)
100
Suction
(cm of H20)
20 30 40 50
-20
Soltrol Saturation (%)
90 100
FIGURE C-10. Soltrol-air saturation curve for SA trial 2.
- 250 -
-------
Soltrol-Air Saturation Data (Drainage)
100-
80-
60-
Suction 40
(cm of H2O)
20-
-20-
* SA Trial 1
• SA Trial 2
0 10 20 30 40 50 60 70 80 90 100
Soltrol Saturation (%)
FIGURE C-11. Comparison of Soltrol-air primary drainage curves for SA trials 1 & 2.
- 251 -
-------
Air-Water Saturation Curve (Drainage-Imbibition)
100
Suction
(cm of H2O)
100
Water Saturation (%)
FIGURE C-12. Air-water saturation curve for AW trial 1.
- 252 -
-------
Air-Water Saturation Curve (Drainage-Imbibition)
100
80-
60-
Suction
(cm of H2O)
-20
100
Water Saturation (%)
FIGURE C-13. Air-water saturation curve for AW trial 2.
- 253 -
-------
100-
80-
60-
Suction
(cm of H20)
40-
20-
-20
Air-Water Saturation Curve (Drainage)
AW Trial 3
arwata dat
-i|I|.|.|,|i|r-
0 10 20 30 40 50 60 70 80 90 100
T
Water Saturation (%)
FIGURE C-14. Air-water saturation curve for AW trial 3.
- 254 -
-------
100-
80—
60-
Suction
(cm of H20)
40-
20-
-20-
Air-Water Saturation Data (Drainage)
* AW Trial 1
A AW Trial 2
• AW Trial 3
-1I,IIIII,IIIII!|I|r-
0 10 20 30 40 50 60 70 80 90 100
Water Saturation (%)
FIGURE C-15. Comparison of air-water primary drainage curves for SA trials 1 through 3.
- 255 -
-------
TABLE C-1. Numerical values of measured saturations, pressures, and temperatures, (continued
on next four pages)
Saturation
SW Trial 7
Capillary Head Temperature
Saturation
SW Trial 8
Capillary Head Temperature
100
97.6
95.7
88
77.5
65
57
42.3
35.6
28
23
23.1
45.6
56.4
68.9
71.6
73.2
— Primary Drainage —
0
10.5
15.7
21.5
22.3
24.8
26
27.6
32.5
42.5
62.1
— Primary Imbibition —
30
15.7
12.7
7.2
-0.4
-5.7
26.5
26
25
24
23
24
24.5
25
27
28.5
31
31
31
28.5
18.5
19
20
100
97.6
26.5
22
26.7
49.9
66.6
68
69.3
68.3
68.1
63.9
43.5
28.7
26.3
23
22
— Primary Drainage —
0
35.1
49.6
60.9
— Primary Imbibition —
28.6
19.8
14
1.9
-10.6
— Secondary Drainage —
4.3
21.8
33.3
35.8
39.1
28.8
60.4
67.4
22
24
24.5
24
24.5
23
27
28
26
21.5
20.5
20
20
18.3
20
23
21.5
Note: the saturations are wetting phase saturations given in percent; 'capillary
heads' are capillary pressure heads given in equivalent cm of water (at 20°C);
and the temperature is the room/cabinet temperature in °C, measured at the
same time.
Note: The S designates Soltrol, W is water and A is air. Therefore SA is a Soltrol-air
experiment.
- 256 -
-------
TABLE C-1. Numerical values of measured saturations, pressures, and temperatures, (continued)
?atyi
100
80
58
41
26
23
18
16
16
17
41
59
68
68
70
70
41
31
21
19
15
•atic
.8
.8
.9
.2
.3
.6
.6
.9
.8
.7
.7
.6
.2
.7
.4
.5
.7
SW Trial 9
in Capillary Head T«
— Primary Drainage —
0
30
31.1
34.1
41.1
46.3
62.5
69.5
— Primary Imbibition —
40.8
27.3
20.5
15.6
8
-2
-8.6
— Secondary Drainage —
24
31.4
35.6
49.8
56.4
66.5
?mp(
24
24.
25.
24
24,
20
24
24.
25.
23
27
28
22
22.
21.
20.
20
20
18
20
20.
3r?
,5
,5
,5
,5
.5
5
,5
,5
5
SW Trial 10
Saturation Capillary Head Temperature
— Primary Drainage —
100 0 23
97.2 22.2 23.2
81 24.8 20.5
43.9 30.8 21.5
25.9 38 20.5
19 49.1 21.8
16.2 56.1 21.3
12.3 63.4 21.8
— Primary Imbibition —
16.8 41.3 21.8
17.6 34.1 20.8
25.2 26.4 23
39.7 13.9 20
68.8 12.4 22
69.5 -2.4 22.5
71.8 -9.3 19.5
- 257 -
-------
TABLE C-1. Numerical values of measured saturations, pressures, and temperatures, (continued)
Saturation
SW Trial 11
Caoillarv Head Temperature
Saturation
SW Trial 12
Capillary Head Temperature
100
99.
61.
38
24.
18.
15.
14.
15.
17.
33.
50.
73.
74.
3
2
4
9
8
7
4
7
6
8
5
2
— Primary
0
24.
25
29.
36.
48.
54.
61.
— Primary
38.
24.
19.
15.
7
-2.
Drainage —
3
5
1
8
5
6
Imbibition —
9
9
6
6
1
— Primary Drainage —
23.
21.
20.
21.
20.
21.
21.
21.
21.
20.
23
20
22
22.
2
6
5
5
5
8
3
7
8
8
5
100
86.7
69.6
45.8
31.6
24
19.2
17.3
0
23
26.3
30.9
36.5
48.4
57.3
64.3
24.
24.
26.
25.
26.
24,
27
25
5
5
,8
,8
,4
.8
.6
.6
Saturation
100
63.3
39.1
24.5
20.7
"18.4
17.3
16.4
SW Trial 13
Capillary Head Temperature Saturation
Primary Drainage —
0 24.5 100
23.4 26.8 99.7
27.2 25.8 59
32 26.4 19
40 24.8 17.9
48.5 27.6 12.2
59.3 25.6 8.1
64.1 28.2 7.8
SA Trial 1
Capillary Head Temperature
Primary Drainage —
0 23
15 21.5
26.8 22.4
28.5 22.8
39.8 22
46 22
52.5 22.6
58 22
- 258 -
-------
TABLE C-1. Numerical values of measured saturations, pressures, and temperatures, (continued)
SA Trial 2
AW Trial 1
Saturation
Capillary Head Temperature
Saturation
Capillary Head Temperature
— Primary Drainage —
100
100
76
44
20
16
8
8
.1
.9
.7
.1
.4
.4
0
9
14
17
21
25
32
41
.5
.5
21
21
21.
22
22
22
22
22
.3
.8
.2
.2
.2
— Primary Imbibition —
8
10
17
60
80
82
82
83
83
81
79
75
47
19
7
.4
.9
.7
.1
.3
.8
.4
.4
.4
.4
.4
.9
.1
.8
31
19
15
6
3
-1
-9
-15
— Secondary
-5
-0
9
11
15
20
30
.5
.7
.7
.4
.5
.1
.5
Drainage —
.5
.6
.2
.8
22
22
22
22
22
22.
22
22.
22.
22.
22.
23
23
23.
22.
.6
.5
,8
,9
7
5
5
100
99.
92.
53.
22
16.
14.
15
15.
18.
26.
42.
62.
75.
80
3
2
9
4
9
9
2
9
7
6
7
— Primary
0
9
32
43
58
83
99
— Primary
84
67
52
39
30
23
14
7
Drainage —
.7
.5
.1
.2
.1
Imbibition —
.6
.5
.7
.2
.2
.9
.7
26
23
25
25
24
18
16
15
18
24
24
26
26
22
21
.9
.3
.4
.8
.6
.6
.2
.6
.2
.8
.2
- 259 -
-------
TABLE C-1. Numerical values of measured saturations, pressures, and temperatures, (continued)
AW Trial 2
Saturation Capillary Head Temperature
— Primary Drainage —
AW Trial 3
Saturation Capillary Head Temperature
— Primary Drainage —
100
99.
98.
87.
65.
32.
25
22.
23.
26
32.
42.
51.
65.
76,
80
4
9
8
9
7
6
6
8
.4
.7
.8
.1
.1
0
10
21
31
47
62
80
94
— Primary
78
60
44
35
27
20
11
5
.2
.5
.9
.3
.7
.4
Imbibition —
.1
.1
.8
.1
.5
.4
.9
27
23.
24
25
25
18.
16.
15.
18.
24.
24.
26.
26.
22.
21
20
3
4
8
6
6
2
6
2
.8
,2
100
100
97.
92
89,
84
79
62
32
25
19
16
13
.4
.6
.4
.3
.1
.8
.1
.1
.6
.4
0
4.
14.
22.
27.
32.
36.
41
50.
55.
62.
69.
75
5
7
9
7
6
5
8
5
8
5
24.
24.
24
25.
24.
24
27
24.
24.
24.
26.
26
22,
9
3
1
6
6
4
,4
1
,9
- 260 -
-------
APPENDIX D
RAW DATA FROM THE SHORT COLUMN EXPERIMENTS
This appendix contains the Lotus 1-2-3 worksheets for the short glass column, quantitative
experiments described in Section 5. The results are summarized and discussed in Sections 9 and 10.
- 261 -
-------
I Xor«»^«»'r^'O*»**<»»^^**^r*»*
vCjoooooooo oooooooooooooo
*
vv I J3XPOOOOOOWV
n«t-t0nvwr<-«iNino«r4Onr>r«A
. o\oounoinwinolf-l«-tr*4r««4«-lr-«C«-l«-ir*r-l«4
J-^jsij 666666060660666666006
x^oooooooooooodoodooooo
>.
g^xoooooooo oooooooooooooo
VPOOOOOOOodoOodoOodoOOO
5 VP
-
I XN0OOOOOOOOOOOOOOOOOOOOO
I vpoooooooooooooooodoooo
«•» v/WWM'K'lwwrwiOK'jriNNNMrwwiNr*
UIVOOOOOOOOOOOOOOOOOOOOOO
UV>*^QOOOOOOOOOOOQOOOOOOOO
0» ^pdddddddo'ddddddddddddd
tAM'xlDOOOC
3i^s§ig
3OOOOOOOOOOOOOO
ooooooooooooooo
3OOOOOOOOOOOOOO
S"i;
I x>£ooooo6oo6666666666600
I +XPOOOOOOOoooooooooooooo
I X£>OOOOOOOOOOOOO OOOOOOOO
!E^(
-*5:
ooooooooooooooooooooo
11^333333335383333883333
1^3335
"^
t
I XJSPOOOOOOOOOOOOOOOOOOOOO
I +X ......................
I XPOOOOOOO oooooooooooooo
J
M I •*^nr*nru*nn
w\xpooooooooooooooooooooo
2 NjDoooooooooooooooooo'ooo
i-c«-)-«-i-r»
+ X£>OOOOOOOOOOOOOOOOOOOOO
xpooooooooooooooooooooo
•HX0OOOOOOOOOOOOOOOOOOOOO
xpooooooooooooooooooooo
X'-V^OOOOOOtAOOOOOOOOOOOOOO
• BN<»r-
XNJ3 OOOOOOOO OOOOOOOO 000 OO
v^ooooooooooooooooooooo
x—xjAin^m**^*
rH UW^^^tn^riA^
oox - - - -
><«*sJ3OOOOOOOOOOOOOOOOOOOOQ
•I
'666666660666666666666
M XPOOOOOOOOOOOOOOOOOOOOO
Xx^OOOOOOOOOOOO OOOOOOOO OO
•sgooooooooooooooooooooo
I v>«r4c«r«r«r»c»^»*'*^**'»'»^***»'*
x^pooooooooooo oooooooooo
>pddddo*ddddddddo*ddddddd
ce«o««
. _o«««om
\xoooo oooooooo oooooooo oo
+\POOOOOOOOOOOOOOOOOOOOO
Ns^ddddddddddddddddddddd
I f^OOOOOOOOOOOQiOOQQOOOOOO St^
—OX»«*« •»•• »•«»»• «»»«««»« t!«
oux •
><-*sj3000000000000000000000
oooo
••••••••••
•V.XJ30000000000000000OOOOO
Xpddddooddddddddddooooo
' -vOoOOOOrtOOOOOOOOOOOOOOO
-
>ooooooooooooo
^'.
l^«.«ffc HMIXMKWI
" >0OOOOOOOOOOOOOOO
«o
nfMM •WDiW^I
262
-------
X.VOOOOO
ixxooooo
• xooooo
i Oxrwomoi
•~rfMXfMWO>f*
otcx •
taxxooooo
w+x
< VPOOOO
I Bxto-vneoio
i «x0tcoutoiiH
— XfOCDOWOl
WXXQOOOO
• xooooo
Z
O
M
H
<
I
$
+ XOOOOO
XPOOOO
•
a t xooooo
oxxooooo
W« XOOOOO
M3X0OOOO
M 0X00000
0
M
8
O
M
g
+XOOOOO
XOOOOO
B
-HX
XXOOOOO
«x
BX
+ XOOOOO
XOOOOO*
oxooooo
•x. .....
O*NpOOOO
•O tTNrf-trH^t^t-
ja^
IPXOOOOO
i bxnuuaift
XXQOOOO
+XOOOOO
I
M V00000
'i
263
8 *c
-------
I '-XPOOOOO
xxpooooo
xpooooo
.—xoo^^^o
IOXrlHOOOr4
rf
I
Q i v^HHeoor^r*
M -v.X(0«HfftOO>O»
D +XP^O<-4OO
o \ ......
M XPOOOOO
5e H \UJr-r- r-r-r*
< -HX ......
O xxpooooo
cu «x
o ax
X
XXPHOOOO
+XPOOOOO
xpooooo
•O O»-w-)HiHr-*«-lr-4
.5^:
xxpooooo
txpooooo
xpooooo
xx«»u>«« **>**«
xpooooo
2
0
XX-^^M^f^r*
Xpooooo
i^
r$
is;
264
-------
I
1
s
M
k
P
1
I
xx-SSSSS laxVr......
9>xin»»rr^i»r«r« !«-*Uxr>*iAiA«0
p*xo o o o o >vxooooo
—$ ? $
X 0 X
13 X •-**-( *H •-* ^ J 9 X
gxuuuuo N x
3x3(1333 B ix««^«^
w&x gxxooooo
§x > xo oooo
o> u ^
iS X * t • • f Or4UXIAlAlAlAlA
P^SUSSl Q>-^00000
xtimmxi w 3 x
J«M« ?s 1
•^tttfS >^1'!Neo
^SSSS j *^-«-o
xot * at ot * t • «x* •* « o o
X« * * * • I «UX«>«*HH
>LLLLt 2i)>JJ_:jj
*$f?iff :*$
X* * * * O 3 X
X« N O H « *H X
^ " x-
xSSSSS l^nr,n««
ii^iiij,! x^sssss
imm ! §""'
^HHHHH 49x«SnCn
^^ xoo o'oo
^$SS8S8 ' ^ ^
5^?.?.?.?.?. -01 ^
"X1 <3$Ht,CT.l«
4Xr4Nn«iA T!o
3$ ^
H:
rr
M
X
-»x
XX-
u s^
F • x
S *>x
X
X
X
x
Js
+ xo
X
2 3xS
! «^0
h U x
s ^
Q I X*
8 *^°.
3 x-0
o ixo
8 i^R
3 S5;o
s a^
1
§
1 XO
» xo
x°
• -*XM
1SI5
*"$
i$-
xxo
+ xo
x •
xi°
• x:"
8$°
^
,$-
^^^
•s?^s
O U X •
-» X
?1
i xr-
>^*
xo
.^n
^8^
X»
1 3 ^
*-t
S -^x
P -25:
8 «x
* ft ft «
IM * w r-
« * m r-
oooo
* IA « w
oooo
OOOO
n « ^ m
N CO *>* r*
OOOO
OOOO
oooo
o n r» «
r« r* r^ r-
ssss
oooo
r» * M n
w r* t* r-
oooo
« A r« w
n « o *
*» * * ^
oooo
n « o «
Nor* N
»« »«
i n * ut
265
-------
-•5333388
VJ30000
•jj«V.HO»»«I
-X^wotfuin
IXtomioiDr.
IA X9>*0*OI0»
?
X^DOOOO
vooood
—| VHf-IHHiH
I XVpOOOO
I xpoood
uxpooon
! 1^53333
4lXTMn«onrt
tv,vpoooo
"^^o
VJ3OOOO
X
<~xboooo
jllxW>-iinNO
Ox.
x>^oooo
NJ3OOOO
SI N^>lv«««
V.XOOOOO
o +V
> XDOOOO
p*4 Ox-^r^K-r^K
oou-
Q
» X
D>—X>
COX
£
O
H
3
"^**«^
+VJDO
XX0OOOO
SU
o-
o <«•••••
*'WTT
xSSSSP!
I i Njnnrtmrt
I X>^oooo
I v^j'oooo
II
3^30000
iS ^
1-1 »s
I xr*
v>jo«»r*«
•NAOOOO
iXttr-oMO^
'SNnri'***
xooooo
OO
S
8
I
266
-------
X-xO OOOO
XO O O O O
" r* *
<— XX1* Ol ** « H
? X Ot U> r* f** Ot
V) XOi Ot « Ot Oi
jxxo oooo
> XO OOOO
XXO OOOO
xo oooo
A XN M N re N
0 I XO OOOO
o ^» _ _ _ — _
x+:
^o o o oo
I XOO 000
XXH H «H H H
+ xo oooo
7*$.,.»o
W £ XO N O « «
1 «o ot
~«\ X H in n co
£ X r-v i"* « r**
H V X«^ co «> in co
£ *x oooo
W \ H H H H
M X
I +*^
i xo oooo
! x"
I fcxr- « co inn
! £x*o* «'<•*«'<;
1 M XO A r- « Ot
! v^nSSSS
J_4j^HHHHH
— 1 XO O O H O
XX»-* H H O H
M -—X OOOO
9Bx- -;*r?r:
j"fHHO-
",t....
XXO OOOO
XOOOO 0
> ^-'XO OOOO
• X
M X
3 I Xv «t> v *o •*
g >5:0.^0.0.°
> xo oooo
> •-'XO OOOO
SB ?
H X
",l
XXO OOOO
xo oooo
• -— xo oooo
b^-
• X
1 $
» X
X
X
a IT
& Jj^-
C 9 x
W *X
1
i ^3^g
n a xC
P MX
•e S
« Nv
o i xa
i ^g
3 ^
u axp
////////////////
n«jxra
iNvsao
1 XP
+-x£
.^«
8S^S
"1
jli
+/~ porosity
V/////////////V////
ill1
X
•* n r* n
oooo
oooo
o co n v
i- »n en co
•H H H H
oooo
n co ot N
»% g» « r*
oooo
oooo
H «0 « O
J in « \D «o
I CO BO CO CO
oooo
i at ooio
I O O O O
oooo
is ?SS
! o o o o
40 OOO
oooo
j-wr- oo»
jjr irtjeo
oooo
oooo
^ *>
oo
I :;
xo oooo
» *^X»H *r 10 n o
• 2\* * *> f* r
9 xo* o'o*o*
S
X OOOO
> —X 0> « H<0
X •• •»
II
267
•J -H X
8 iJ^
-------
|