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^

-------