Characterization of spatial NAPL distribution, mass transfer and the effect of cosolvent and surfactant residuals on est...

MISSING IMAGE

Material Information

Title:
Characterization of spatial NAPL distribution, mass transfer and the effect of cosolvent and surfactant residuals on estimating NAPL saturation using tracer techniques
Physical Description:
xii, 136 leaves : ill. ; 29 cm.
Language:
English
Creator:
Cho, Jaehyun, 1963-
Publication Date:

Subjects

Subjects / Keywords:
Environmental Engineering Sciences thesis, Ph. D   ( lcsh )
Dissertations, Academic -- Environmental Engineering Sciences -- UF   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 2001.
Bibliography:
Includes bibliographical references (leaves 128-135).
Statement of Responsibility:
by Jaehyun Cho.
General Note:
Printout.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 028197855
oclc - 49279724
System ID:
AA00014257:00001


This item is only available as the following downloads:


Full Text










CHARACTERIZATION OF SPATIAL NAPL DISTRIBUTION, MASS TRANSFER
AND THE EFFECT OF COSOLVENT AND SURFACTANT RESIDUALS ON
ESTIMATING NAPL SATURATION USING TRACER TECHNIQUES












By

JAEHYUN CHO


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA














ACKNOWLEDGMENTS

During my four-and-a-half years of doctoral research at the University of Florida,

I received assistance from many people. I greatly appreciate their help and guidance and I

hold them in my heart. I would not be in the position to be writing this dissertation if it

were not for their assistance and support throughout the years.

Primary thanks go to my advisory committee chair, Dr. Michael Annable, who

guided me to this point with his encouragement, assistance, and support. I also gratefully

thank my other committee members: Dr. Suresh Rao, Dr. William Wise, Dr. Kirk

Hatfield, and Dean Rhue, for their teaching and substantial advice.

I would like to thank my previous fellows, Michael Brooks, Suzannah Kuhn,

Michael Vanvalkenberg; and my present lab partners, Rick Robert, Christina Martines,

and Diane Bonbehagen for their helpful support and advice in the lab or in the office. I

also thank Irene Poyer, Dr. James Jawitz, and Dr. Heonki Kim for their valuable

suggestions. I thank Dr. Booth for the many hours of exchanging knowledge and

assistance in the lab.

I especially thank my wife, Yoonjung, who was understanding and gave me

encouragement during my studies. I also thank my son, Michael. Obvious thanks go to

my parents and parents-in-law for their support from far away.









TABLE OF CONTENTS



ACKNOWLEDGMENTS...............................................ii

LIST OF TABLES.................................................vi

LIST O F FIGURES..........................................................vii

ABSTRA CT.................. ............................................. .......................x


CHAPTERS

1 INTRODUCTION.................. ....... ..................................1

2 CHARACTERIZATION OF PARTIAL NAPL DISTRIBUTION
BY COMBINED USE OF PARTITINING AND INTERFACIAL
TRACER: USING GRAIN SIZE AND NAPL DISSOLUTION

Introduction ............................ ................................................................... 8
Theoretical Background.................................................... 10
Adsorption Coefficients ................................... .......................... 10
Intefacial Area Estimate..................................................... 11
NAPL Saturation Estimate......................... .....................13
NAPL Morphology Index.......................... ........................... 13
Materials and Methods...................................................... 14
M materials ............................................................. 14
Interfacial Tension Measurement ................... ............. ........... .. 15
Miscible Displacement Tests............................................. ........... .. 15
Analytical Methods...................................................... 18
Results and Disscussion....................................................... ......................... 18
Estimation of Interfacial Adsorption and Partitioning Coefficients........ 18
NAPL-Water Interfacial Area and NAPL Saturation as a Function
of Pore Size............................................................................................. 20
NAPL-Water Interfacial Area and NAPL Saturation as a Function of
NAPL Dissolution.................................................. 27
NAPL Morphology Index............................................... ................... 30
C onclusions..................................................................... ............... ........... 33

3 INFLUENCE OF SPECIFIC INTERFACIAL AREA ON MASS
TRANSFER: NONEQUILIBRIUM PROCESS

Introduction ................................................................... ................ ........... 34



iii








Theoretical Backgroud.................................................. ............................. 36
Materials and Methods....................................................... 37
Experimental Methods......................... ........................... 38
Results and Disscussion.................................................. 38
Parameter Estimation................................... ......................... 38
Mass Transfer Coefficients ............................. .......................... 40
Influence of NAPL Reduction by Dissolution on Mass Transfer............ 46
Model Development..................................................... 48
Nonequilibrium Mass Transfer.......................................................... 54
C onclusions.................................................................... ................ .......... 55

4 ESTIMATING NAPL SATURATION USING PARTITIONING
TRACERS: INFLUENCE OF RESIDUAL COSOLVENTS

Introduction............................................ .................................................. 57
Theoretical Background................................................... 59
Materials and Methods....................................................... 60
M materials ............................................................. 60
Partition Isotherm Experiments ........................ ............. ........... ... 60
Solubility Experiments..................................................... 61
Miscible Displacement Experiments .......................................... ........... 61
Data Analysis................................................... ............................ 63
Results and Disscussion.................................................... ........................... 64
Effect of Cosolvent on Solubility ........................................................... 64
Effect of Cosolvent on Tracer Partitioning Isotherms......................... 68
Effect of Cosolvent on Tracer Retardation............................................ 70
Impact of NAPL Saturation in the Systems........................................... 75
C onclusions..................................... ............................................................... 77

5 INFLUENCE OF RESIDUAL SURFACTANT ON PARTITIONING
TRACERS

Introduction ........................................................................ ............................... 79
Surfactant Adsorption from Aqueous Solution onto Solid Matrix.......... 80
Surfactant Adsorption from Aqueous Solution onto Hydrophobic
Organic Chemicals........................................................................... 82
Partitioning of Hydrophobic Organic Chemical into Residual
Surfactants........................... .............................. 83
Materials and Methods..................................................... 84
M materials ..................................................................... 84
Batch Isotherm Tests.......................... ........................... 85
Miscible Displacement Tests......................... ........................ 86
Results and Disscussion.............................................................. 89
Effect of Surfactants on Batch Isotherms .............................................. 89
Effect of Residual Surfactants on Tracer Transport ............................ 94
Effect of Adsorbed Surfactants on the Solids Matrix.............................. 97








Demonstration through a Flushing Test............................................... 100
C onclusions................................. ................................................................. 103


6 Conclusions .................. ..................... ...................... 105

APPENDICES

A INTERFACIAL TENSION: SDBS SOLUTION-PCE WITH OIL-RED DYE.......109

B DERIVATION FOR MASS TRANSFER COEFFICIENT AT THE STEADY
STATE CONDITION............................. .................................110

C EQUILIBRIUM ISOTHERMS OF PARTITIONING TRACERS IN THE
PRESENCE OF COSOLVENTS WITH VARYING CONCENTRATIONS.........114

D EQUILIBRIUM ISOTHERMS OF PARTITIONING TRACERS IN THE
PRESENCE OF SURFACTANTS WITH VARYING CONCENTRATIONS.......121

REFERENCES................... ........................................ 128

BIOGRAPHICAL SKETCH..........................................................136










LIST OF TABLES


Tables pgge


1-1. Summary of tracer techniques.................................. ........................7

2-1. Experimental parameters and results for interfacial and partitioning
tracer tests conducted with various medium sizes...................................22

2-2. Experimental parameters and results of column tests conducted after
ethanol flushing......................................................................22

3-1. Mass transfer results and parameter values measured from column experiments
with various grain sizes..................................... ......................39

3-2. Mass transfer results with respect to reduction ofNAPL saturation/interfacial
area........ .............. ..... ............... ........... ........... 44

3-3. Mass transfer correlations...............................................................48

4-1. Comparison of estimated ac and be values from tracer solubility
and partition experiments for four tracers.............................................70

4-2. Parameter values observed from miscible displacement experiments
at low volume fractions of ethanol........................... ..................... 71

4-3. Comparison of parameter values observed from low and high NAPL (PCE)
saturation column tests using 2,4-dimethyl-3-pentanol as a partitioning
tracer...................................... ... ..... .. ....... ........... ... .. 76

5-1. Physical and chemical properties of surfactants used in the study................ 86

5-2. Partitioning coefficients (Kns) of alcohol tracers measured on various surfactant
contents (% by weight)....................................... .....................90

5-3. Parameter values observed from column miscible experiments with
PCE/DowFax 8390..............................................................94

5-4. Results observed from pre- and post-flushing tracer tests and PCE mass balance
from the flooding test with DowFax 8390 5%/AMA 80 3%/ NaCI 3%//
CaCI2 3% ................................... .............................. .................. 103












LIST OF FIGURES


Figures pBge

2-1. Experimental set-up................................ .... ......................... 16

2-2. Interfacial tensions measured at NAPL-SDBS solution interfaces.................19

2-3. Breakthrough curves of nonreactive tracer (bromide, symbol "x") and
interfacial tracer (SDBS, symbol "o") from columns with various grain
sizes as porous media. Shown are the plots on left-hand side without
NAPL and on right-hand side with residual NAPL (PCE)........................21

2-4. Geometric pore singlet models (Rhombohedral and Tetrahedral packing),
log-log linear fit, and measured data for anw-dso relationship......................... 24

2-5. NAPL-water interfacial area (anw) and trapped NAPL saturation (Sn) for
different porous media sizes.................. ........ ....................24

2-6. Breakthrough curves of nonreactive tracer (bromide, symbol "x") and
interfacial tracer (SDBS, symbol "o") from column tests after ethanol
flushing with various pore volumes (0, 4, 8, and 12 Pore volume)..............26

2-7. NAPL-water interfacial area (anw) as a function of remained NAPL
saturation (S,) after ethanol flushing...............................................28

2-8. NAPL-water interfacial area (anw) as a function ofNAPL dissolution..............28

2-9. Relationship of NAPL morphology index (HN) and porous medium size.
Shown are the symbols (data: *; fit data: solid line)..............................31

2-10. The HN estimated using HN-aw-Sn (Saripalli, 1997) and HN-dso-S, relationships:
Shown are also the predicted HN with respect to varying NAPL saturation
using the HN-dso-S, relationship ................................ ...................31

3-1. An example BTCs of non-relative tracer (Br) fitted to an analytical
solution to estimate logitudinal dispersion coefficient and dispersivity
values.......................... ...... ... ..........................40

3-2. Normalized PCE effluent concentration (C/Cs) versus pore water velocity.........41








3-3. Normalized PCE effluent concentration (C/Cs) versus specific interfacial
area................... ... ............ ... ...................41

3-4 Bulk mass transfer rate coefficients (KI) versus Reynolds number (Re) for various
grain sizes........................ ....... .... .... .. .............. 42

3-5. Bulk mass transfer rate coefficients (K,) versus specific interfacial area for various
velocities......... .......... ...... .............. ...... ........ .. .................42

3-6. Intrinsic mass transfer coefficients versus specific interfacial area for various
velocities..............................................................................45

3-7. Bulk mass transfer rate coefficients (Ki) as a function ofNAPL dissolution.
fD = (Sn* Sn)/ Sn* = 1- (Sn/ Sn) is volumetric fraction of dissolved
NAPL; S,* is the initial NAPL saturation; Sn is the NAPL saturation
remained after reduction of NAPL by dissolution..................................45

3-8. Intrinsic mass transfer coefficients (k4) as a function of NAPL dissolution .........47

3-9. Averaged intrinsic mass transfer coefficients for various velocities.................47

3-10. Comparison of modified Sherwood numbers (Sh') from this study with previous
investigation (dso = 0.073 cm, anw = 44 cm2/cm3, Sn = 0.14, 4 = 0.39,
U I= 1.2) ..... ............... ............. ..... ..... ....................... .... 51

3-11. Bulk mass transfer rate coefficients measured versus those predicted from the
developed model (Re/ anw/d5o model) for all data pooled together................51

3-12. Intrinsic mass transfer coefficients measured with those predicted from the model
(Re/Sc) for all data pooled together............................ .....................53

3-13. Damkohler number (Da = k,/v) with respect to Reynolds number for various
grain sizes.......................... ....... ..... ..... .............. 53

3-14. Damkohler number (Da = ki/v) with respect to specific interfacial area for various
grain sizes................................ ......... ...... .......................54

4-1. Relationship between tracer solubility (Sc) and cosolvent content (%, volume)
for (A) ethanol, (B) tert-butanol, and (C) isopropanol...............................65

4-2. Solubility of 2,4DMP with respect to Tert-butanol content (%, volume)............66

4-3. Relationship between tracer partition coefficient (Kn,) and cosolvent content
(%, volume) for (A) ethanol, (B) tert-butanol, and (C) isopropanol..............67

4-4. Comparison of the K,n values from batch tests to the Kcol estimated from column








tests for 4-methyl-2-pentanol, n-hexanol, 2-methyl-3-hexanol, and
2,4-dimethyl-3-pentanol in the ethanol/water system; Column test; m
Batch test................................ ................................72

4-5. Breakthrough curves of residual ethanol cosolvent and partitioning tracers
(n-hexanol and 2,4-dimethyl-3-pentanol) to display a degree of mixing,
explaining the effect of residual ethanol on the NAPL saturation estimation.
Initial residual ethanol content in the column is 10% (volume)...................74

5-1. S-shaped adsorption isotherm for an ionic surfactant on an oppositely
charged substrate.................... ................... ......... .................... 81

5-2. Schematic diagram of the experimental set-up.........................................87

5-3. Relatioship between tracer partition coefficients (Km) and surfactant [(A) AMA-
80, (B) DowFax 8390, and (C) Brij97] contents (%, weight).......................91

5-4. Comparison of the K,. values measured by batch tests to the Kol estimated by
column tests for (A) n-hexanol and (B) 2,4-dimethyl-3-pentanol in the PCE/
DowFax 8390 system .................................... ... ...................... ... 96

5-5. Results from column experiments with A) no surfactant, B) AMA 80,
C) Brij 97, and D) DowFax 8390 adsorbed surfactants and no NAPL.
Shown are BTCs for nonpartitioning tracer (methanol, plus symbols) and
partitioning tracer (n-hexanol, circles)....................... .......................98

5-6. BTCs of preflushing tracer test with only residual NAPL (SPCE = 0.16),
and post-flushing tracer test with residual NAPL (SPCE = 0.0125) and
surfactant (DowFax 8390)...................... .......... ..... .................101

5-7. BTC of removed PCE during the surfactant mixture flooding.....................102














Abstract of Dissertation Presented to the Graduate School
of the university of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

CHARACTERIZATION OF SPATIAL NAPL DISTRIBUTION, MASS TRANSFER
AND THE EFFECT OF COSOLVENT AND SURFACTANT RESIDUALS ON
ESTIMATING NAPL SATURATION USING TRACER TECHNIQUES
By

Jaehyun Cho

December, 2001


Chair: Michael D. Annable
Major Department: Environmental Engineering Sciences

The non-aqueous phase liquid (NAPL) specific interfacial area and morphology

as a function of pore size and NAPL dissolution were studied using the interfacial and

partitioning tracer techniques. The NAPL-water interfacial areas increased in a log-linear

fashion with decreasing grain sizes used as porous media, but did not show a clear trend

with residual NAPL saturation formed in the various grain sizes. At a given grain size,

however, the interfacial areas showed a proportional linear relation to a decrease in

NAPL saturation by dissolution. The NAPL morphology indices, which represent the

spatial NAPL distribution, increased exponentially with decreasing grain size.

Mass transfer across interfaces between NAPL and the aqueous phase was

investigated. The investigations consisted of the influence of specific interfacial area,

NAPL saturation, grain size, and aqueous phase velocity on mass transfer. Bulk mass

transfer rate coefficients increased with increasing the specific interfacial area as well as








NAPL saturation and pore velocity, and with decreasing grain sizes. Whereas, intrinsic

mass transfer coefficients were nearly independent of specific interfacial area and NAPL

saturation, but dependent on pore velocity. Reduction of NAPL saturation by dissolution

caused a linear decrease in the bulk mass transfer rate coefficients. Phenomenological

correlation models developed using dimensionless Sherwood number produced good

prediction for bulk and intrinsic mass transfer coefficients.

A series of batch and column tests was conducted to elucidate the influence of

residual cosolvents and surfactants on partitioning and transport of alcohol tracers

through soil columns containing residual tetrachloroethylene (PCE). Batch equilibrium

tests showed that as the volume fraction of cosolvents (<10%, vol.) increased,

partitioning coefficients for alcohol tracers linearly decreased for ethanol, linearly

increased for tert-butanol, and did not exhibit an evident relationship for isopropanol. The

column tests using ethanol as a residual cosolvent exhibited earlier partitioning tracer

breakthrough which caused an under-estimate of NAPL saturation. The under-estimates

of NAPL saturation were 1 to 10% lower than the actual saturation. Comparison between

low (0.5%) and high (15%) saturation columns revealed that the effect of residual

cosolvent was different depending on the amount of NAPL in the column.

The influence of residual surfactants in aqueous and adsorbed phases on tracer

transport behavior was evaluated by estimating PCE saturation using partitioning tracers.

The batch equilibrium tests using residual surfactants ranging from 0.05 to 0.5% by

weight showed that as the concentrations of the surfactants increased, the partitioning

coefficients linearly decreased for Diphenyl oxide disulfonates (DowFax 8390), increased

for Polyoxyethylene (10) oleyl ether (Brij 97), and decreased slightly or exhibit no








observable trend for Sodium dihexyl sulfosuccinate (AMA 80). Results from column

tests using clean sand media with residual DowFax 8390 and PCE were consistent with

those of batch tests. In the presence of DowFax 8390 (less than 0.5% by weight), the PCE

saturations were underestimated by up to 20%. Adsorbed surfactants on a loamy sand soil

with strongly positive charged oxides without PCE showed false indications of PCE

saturation. Using no surfactant (background soil) gave a false PCE saturation of 0.0004

while AMA 80, Brij 97, and DowFax 8390 gave false PCE saturations of 0.0024, 0.043,

and 0.229, respectively.














CHAPTER I
INTRODUCTION

The contamination of soil and aquifers by organic contaminants is a major

problem caused by leaks, spills, and disposal of waste over long periods of time. A large

amount of organic substance has contaminated many sites and the threat to public health

has been recognized. Many of these contaminants exist as liquids immiscible with water.

The immiscible fluids serve as a long-term source contributing to groundwater

contamination. Most contaminants slowly dissolve and produce plumes that spread over

long distances from the source. Although the solubility of the contaminants is very slight,

it can affect groundwater quality both for drinking water and water supplied to

ecosystems. Characterizing these sources is a necessary primary step to cleaning up the

contaminants in an efficient and effective manner.

Characterization methods used for contaminant source zones can be divided into

three periods: before, during, and after remedial efforts. Before remediation,

characterization processes include estimating of saturation of liquids and distributing

contaminants in the target area. In the past, estimation has been typically achieved by

historical information, soil core samples, and inferences based on groundwater

concentration. These methods were used to estimate site characteristics through a partial

search of the target area. Given the difficulty and cost of complete site characterization,

the recently proposed partitioning and interfacial tracer techniques for characterization of

contaminant source in the aquifer can be useful tools.








Tracer Techniques

Tracer techniques can be divided into four categories; nonreactive tracers,

partitioning tracers, interfacial tracers, and biogeochemical tracers (Rao et al., 2000)

(Table 1-1). The partitioning tracer technique has been used in the petroleum industry to

estimate oil saturation and was initially developed by Cooke (1971) and Dean (1971). Jin

et al. (1995) proposed the application for characterization of contaminant source zones in

aquifers. The partitioning tracer technique, based on the travel time difference between

nonpartitioning and partitioning tracers, is a method to quantify the nonaqueous phase

liquids (NAPLs) trapped in a swept zone (Jin et al., 1995; Annable et al., 1995). The

applicability of this technique has been evaluated through field applications in previous

studies (Rao et al., 1997; Annable et al., 1998; Sillan et al., 1999; Jawitz et al., 1999). Use

of a network of multi-level samplers can provide information about the spatial location of

NAPL (Annable et al., 1994, 1998; Rao at al., 1997). Jawitz (1999) extended the method

to estimate NAPL spatial distribution using higher moments of partitioning tracer

breakthrough curves measured at a single well. Appropriate tracers and concentration

range should be carefully selected to minimize the time required to conduct the tracer

study (Annable et al., 1995) as well as error by nonlinear tracer partitioning (Wise, 1999).

Interfacial tracers were first developed by Saripalli et al. (for NAPL-water, 1997)

and Kim et al. (air-water, 1997). The first field application to estimate NAPL-water

interfacial area along with NAPL saturation took place at Hill AFB in 1996 (Annable et

al., 1998). The interfacial tracer technique is a method to estimate specific interfacial area

between NAPL and water in a swept zone (Saripalli et al., 1997, 1998; Kim et al., 1997,

1998; Annable et al., 1998). For the partitioning and interfacial tracer techniques, the








basic approach is the same since they are both based on travel time differences between

nonreactive and reactive tracers. Unlike partitioning tracers, however, interfacial tracers

are retarded because of tracer adsorption at interfaces between phases. The adsorption

isotherm for interfacial tracers can be determined using the Gibbs model, which is based

on a nonlinear correlation between tracer concentration and surface/interfacial tension.

Kim et al.(1997) reported that for gaseous interfacial tracers, a linear isotherm

approximation is valid at low concentration.

The combined tracer techniques (partitioning and interfacial tracers) makes it

possible to characterize spatial NAPL distribution in the subsurface (Saripalli et al., 1997;

Annable et al., 1998). The ratio of interfacial area to volumetric NAPL content, which is

defined as a contaminant morphology index, shows how the contaminant is distributed in

a porous medium (Saripalli et al., 1997). NAPL characterization can provide information

important in characterizing remedial performance efficiency. The efficiency often

depends on the contact area between NAPL and flushing agents. The contact area is

related to the effective NAPL-water interfacial area as well as NAPL morphology. The

interfacial area and morphology are characterized by capillary forces (Sarapalli et al,

1998). This shows that porous media size can act as a primary factor related to interfacial

area as well as the morphology of trapped NAPL.

In Chapter 2, the dependence of specific NAPL-water interfacial area and NAPL

morphology on pore size was investigated using the combined tracer techniques. Various

grain sizes of sand as porous media were used to measure NAPL saturation and

interfacial area. The observed results were correlated to evaluate their relationships with








respect to grain size. In addition, NAPL-water interfacial area was investigated as a

function of the reduced NAPL saturation by dissolution.

NAPL characterization process during a remediation effort is related to its

efficiency. The efficiency depends on contact area and time between NAPL and a

flushing agent as well as the type of flushing agents. We can assume that the contact area

is same as to an effective interfacial area between NAPL and aqueous phase. In a

chemical remediation work based on NAPL dissolution, it is especially important to

investigate the relationship of NAPL-water interfacial area and mass transfer rate to

predict the remediation efficiency. To date, most studies were limited in investigation of

overall mass transfer rate coefficient, which is defined as the product of a mass transfer

coefficient and the specific interfacial area because of the difficulty in measuring NAPL-

water interfacial area (Miller et al., 1990; Powers et al., 1992; Imhoffet al., 1997; Mayer

et al., 1999). In Chapter 3, mass transfer across NAPL-water interfaces was studied using

the NAPL-water specific interfacial areas measured in Chapter 2. It was related to how

the specific interfacial areas and other related factors (NAPL saturation, grain size, and

pore velocity) contribute to bulk and intrinsic mass transfer coefficients. In addition,

prediction models using dimensionless Sherwood number were developed and verified

through comparison with measured and predicted values. Nonequilibrium process in the

mass transfer can be also a critical factor to predict the remediaiton efficiency. It is a rate-

limited process related to physichemical properties such as flow velocity. The

nonequilibrium mass transfer here was described with the observed system parameters

such as NAPL-water interfacial area and pore velocity.








NAPL characterization after remediation efforts is related to performance

assessment of clean up. A primary purpose of the post-remediation works is

quantification of remaining and removed NAPL, which can be evaluated by the

partitioning tracer technique. After remedial flushing, some amounts of flushing agents,

such as cosolvent or surfactant, are likely to remain in the swept zone. The residual

flushing agents can modify the chemical properties of partitioning tracers such as activity

and solubility, and partitioning behaviors. The resulting effects can cause some errors on

NAPL saturation estimation. In Chapter 4, the influence of residual cosolvent was

investigated on NAPL volume estimation using partitioning tracers. Partitioning

coefficients and solubilities of alcohol tracers were measured in the presence of

cosolvents with varying concentration. The observed results were verified through

miscible displacement tests. Additionally, the dependence of the cosolvent impact on the

magnitude of residual NAPL saturation was evaluated on the NAPL saturation estimate.

Surfactants tend to remain adsorbed on the solid phase and dissolved in the

aqueous phase after surfactant flooding efforts. Above the critical micelle concentration,

surfactant molecules form micelles including a hydrophobic inner core in which the

alcohol tracers can partition. Also, adsorbed surfactants can act as a good adsorbent for

hydrophobic chemicals (Ko et al., 1998a), causing tracer retardation. These processes can

be a problem for NAPL saturation estimation based on retardation of partitioning tracers.

In Chapter 5, the influence of residual surfactants was investigated on NAPL saturation

estimation using partitioning tracers. Anionic and nonionic surfactants with varying

concentration were used to investigate influences on partitioning coefficients and

transports of alcohol tracers. A positively charged soil was also used to investigate the





6


effects of adsorbed surfactants on retardation of partitioning tracers. The observed results

were used to evaluate a false indication of NAPL saturation.












*'3NA 23uPSip3p-UOU
S1o0 01jo3 uoep.IBuS! s e"'a jm 2rp2,p 0o0 JO.Je uo0pMal s! 'y Jlqiumnu 3130ld s! d 'IwIsuoso 2W sSoIUosusuznp so! .m '(aA!sz "od 'suopnlos luOanUuo
prE lo oua p5 5 uo! pm0aumo no1m5 we 0 pug a :uoosUOI 1 a I e1 mei-'-0 V N VN s! :wm..Mdlm Jlnlosqr s! L l'uimsuoa seg s! U/ :rs.3 oru 5pj.n
pueos u!uopied S3umBPir 'L ptu )d siduoqns srp p "m 'j.mJ uoqe-pbul Ms s lU Opg J1po0 uDopdiosp0e leom1 ao q s! 'I :eare rpepaju! jaeo)-5I VN all s!
p' :uo nsmns 'VN aq01 sr! :ralorro 1m5 ltainulOA arp st *g '.issrop nrq jpos s! d lua!tsorao uoodjoslos n0 st Py U0IJaao03 uorlmOd iem-5-IdVN 0t s! "y


V/N
y-v i 0 =,,


olp J3
aPi lg.L

-pj 3


uoopdospe olrep.osJ


*,.a3oloqdjo- 'IdVN

(sma1td [EBis put 'oa AV)
SOW j5JB 11303i

rsoqou srqb Jnnomlm qgiq
:stiu3pns 3o!qo pe o- oo!Unv
510J10 [BralwIi


V/N [Ipo unoiepugaao


*e (Cs I) + a


*3'y Uj
.yd YS-






uo-pvip jo q- dun
suaripors oolouj

.uo0nqsp idY-HVN

(smwand Topmds pu 'OS-IAe)
uortmnsos Inpp. "IdYVN
ZZZ-uopgs
:Sloqo i pnmpsqns-l|Arpt

*ram 5 05m0om 1
S2fl1) 01U5l~f


(000o "I'e lo oeo moIj paidepe) sanbiuqrpal .I3eo jo AmusiouS nsI-Iq alj


iopJo-olg J10 -011z



d







Jo jeiqo.pnu



po0!qe 1q pipuOojd

'h0EAR) 505MOW

(Us. nsOB o02Lq)



SJ33BoJ 5UiaqXoSoig


IZpom uworup.il0









suoplanddV





pas nsjmzmL
P001 $002301














CHAPTER 2
CHARACTERIZATION OF SPATIAL NAPL DISTRIBUTION BY
COMBINED USE OF PARTITIONING AND INTERFACIAL TRACER
: USING GRAIN SIZE AND NAPL DISSOLUTION

Introduction



Characterization of NAPL introduced into the subsurface is difficult because of

complex intra- and inter-relationships between NAPL and soil properties. NAPL source

zone characterization requires knowledge of physio-chemical characteristics such as the

pore size distribution of a soil and the volume, shape, size, and spatial distribution of

NAPL. Capillary forces, which are results of the unbalanced forces from solid-liquid

adhesion and liquid-liquid cohesion, are inversely proportional to soil pore size. NAPL is

trapped in the pore space as isolated ganglia due to capillary forces. Thus, the pore size of

porous media can determine trapped NAPL saturation and morphology. The quantity and

distribution of the trapped NAPL is closely related to the ratio of pore body to pore neck

radii, which is called the aspect ratio. Homogeneous packs of uniform size media have a

low aspect ratio and NAPL blobs are nearly spherically shaped, single pore bodies,

whereas heterogeneous packs have a high aspect ratio and are irregularly shaped, and can

be connected by multiple pores (Chatzis et al., 1983; Morrow and Heller, 1985). Thus,

pore size and pore size distribution can be significant parameters for characterization of

trapped NAPL in porous media.








The primary consideration in trapped NAPL characterization is likely estimation

of the amount of residual NAPL and interfacial area between NAPL-water in the

subsurface. Residual NAPL can be quantified using the partitioning tracer technique

proposed by Tin et al. (1995). Its usefulness and effectiveness in measurements of NAPL

trapped in the subsurface have been reported in laboratory and field tests by previous

researchers (Jin et al., 1995; Annable et al., 1995, 1998a,b; Rao et al., 1997; James et al.,

1997; Sillan et al., 1998; Jawitz et al., 1999, 2000; Zhang et al., 2001; Wu et al., 2001).

Interfacial area between NAPL and water is also an important factor for characterizing

NAPL of distribution. Many approaches have been used to predict specific NAPL-water

interfacial area, such as geometric analysis (Gvirtzman and Roberts,1991),

thermodynamic models (Leverett, 1941; Bradford and Leij, 1997), pore singlet models

(Saripalli, 1997), and empirical models (Cary, 1994). These approaches were used to

predict the specific interfacial area through theoretical methods. Recently, the interfacial

tracer technique, using surface adsorption of surfactant, was developed to measure

interfacial areas between fluids: liquid-liquid (Saripalli et al., 1997); and liquid-air (Kim

et al., 1998, Rao et al., 2002). The interfacial tracer technique has been evaluated through

field and laboratory tests in several previous studies (Saripalli et al., 1997, 1998; Annable

et al., 1998; Kim et al., 1998, 1999; Rao et al., 2000). Because partitioning and interfacial

tracers can be used in tandem, characterization of spatial NAPL distribution is feasible.

The objective of this study is to characterize spatial NAPL distribution as a

function of the mean grain size and NAPL reduction by dissolution. NAPL-water

interfacial area and trapped NAPL saturation were measured using tracer techniques for

various media sizes as well as gradually reducing NAPL saturation. The functional








relationships between mean grain size (or NAPL dissolution) and NAPL-water interfacial

area (or NAPL saturation) were investigated and then evaluated in terms of spatial NAPL

distribution. Finally, the measured results were also used to develop an empirical model

to evaluate NAPL morphology as a function of grain size.

Theoretical Background

Adsorption Coefficients

Surfactants have been used as interfacial tracers because of their unique tendency to

accumulate at the interface between two immiscible fluids such as NAPL-water (Saripalli

et al., 1997; Annable et al., 1998; Kim et al., 1998). The interfacial tracer technique is an

indirect measurement: that is, the amount of adsorbed surfactant per unit interface area is

calculated from interfacial tension measurements. The adsorption of surfactant at the

NAPL-water interface can be described by the Gibbs adsorption equation (Rogen, 1989;

Adamson, 1982):



F 2R_ C dCIIy (2-1)
2RT dCJr


where r (mM/cm2) is the Gibbs surface excess of surfactant or the interfacial adsorption;

y (dynes/cm) is the interfacial or surface tension; R is the Universal gas constant (dyne-

cm/Mm OK); C is the bulk concentration of surfactant (mM/cm3); and T is absolute

temperature (*K). If the surfactant adsorption from aqueous bulk solution occurs in the

presence of a swamping constant amount of electrolyte (1:1 elctrolyte) containing a

common nonsurfactant counterion, the Gibbs equation (Eq. 2-1) is used without the factor

2 in the denominator (Rogen, 1989; Adamson, 1982).








The Gibbs equation (Eq. 2-1) can be used to determine the interfacial adsorption

coefficient, which describes equilibrium adsorption between the interface and the bulk

aqueous solution. The interfacial adsorption coefficient (K,) is expressed as (Saripalli et

al., 1997; Kim et al., 1997):



K-F -=- (2-2)
K Fo = (2-2)
Co 2RT Crc=co


where Co is the influent tracer concentration. Equation (2-2) requires knowledge of the

interfacial tension (y) as a function of aqueous concentration of surfactant (C) used as the

interfacial tracer. The y-C relationship can be obtained by measuring interfacial tension at

various surfactant concentrations. The y-C relationship can be approximated by nonlinear

adsorption regression with the Freundlich sorption model. Then, the Ki can be determined

from the slope of the y-C plot. Measurements of NAPL-water partitioning coefficients for

alcohol tracers were described in Chapter 4 and 5.

Interfacial Area Estimate

The interfacial tracer technique is based on measuring retardation of the

interfacial tracer relative to a nonreactive tracer. It can be used as a method appropriate

for NAPL-water phases with the following four assumptions: (1) the interfacial tracer

adsorbs at the NAPL-water interface but does not partition or dissolve into the NAPL; (2)

the adsorption culminates as a monolayer coverage; (3) each tracer molecule occupies a

known molecular area; and (4) the tracer solution does not alter the properties of the

NAPL. The interfacial tracer is retarded due to adsorption at the NAPL-water interface

during its travel through a porous medium under steady water flow. The retardation factor








(Rif) of interfacial tracer with respect to the non-reactive tracer is defined as (Saripalli et

al., 1997):


R -'-l+asfK, pK
R" ,f. 1 a K (2-3)
,ur 0 9 0 (


where the subscripts ift and nr describe interfacial and nonreactive tracers, respectively;

K, (cm) is the interfacial adsorption coefficient; Ow is the volumetric water content; K,

(mL/g) is the tracer adsorption coefficient onto the solid matrix; p is the bulk density

(g/cm3). For a step-input tracer experiment, / can be obtained by integrating the area

above the breakthrough curve measured for continuous tracer input:



SCo
S ( )dt (2-4)


where Co is the influent tracer concentration; C is the effluent tracer concentration. For a

pulse-input tracer experiment, p can be obtained by the first, normalized, temporal

moment adjusting for the finite pulse duration of to:


f= tCdt (2-5)
to (2-5)
E Cdt 2



The NAPL-water interfacial area can be estimated by arranging equation (2-3):



(R, ,-1-pKJ,/O) (2-6)
K,










Note that K, = 0 if the tracer adsorption onto the solid matrix is negligibly small.

NAPL Saturation Estimate

Partitioning tracers are retarded due to interaction with NAPL during travel

through a porous medium under steady water flow. The retardation factor of a

partitioning tracer (Rp) is a measure of the tracer mobility. The R, is obtained by dividing

the first temporal moment of a partitioning tracer by the normalized first temporal

moment of a non-partitioning tracer. The Rp is given by Jin et al.(1995):


p 1 S,K., pKd
R (1 S= ) ,, (2-7)



where the subscripts p denotes partitioning tracer; K,, is the partitioning coefficient

between NAPL and water, which is determined by the slope of the regressed curve

between partitioning tracer concentrations distributed into NAPL and water phases; Kd

(mL/g) is the tracer distribution coefficient onto the solid matrix. Partitioning tracer tests

have generally been performed using a pulse input experiment and the p can be obtained

by equation (2-5). When sorption of the tracer to the background soil is negligible (Kd =

0), the NAPL saturation (Sn) is can be estimated by rearranging equation (2-7):


S = (2-8)
(R 1 + K, )



NAPL Morphology Index

As NAPL migrates through the subsurface, the trapped NAPL saturation can be

influenced by capillary forces and ambient groundwater flow conditions. The retained








NAPL can exist in several forms, such as single large or small blobs, doublets or

extensive ganglia, and films sorbed on the porous media. The morphology of NAPL

trapped can be different depending on pore size as well as NAPL saturation. As NAPL

saturation increases in a unit pore space, the blobs of trapped NAPL are larger and

thicker. This results in a decrease of interfacial area per unit NAPL volume. The ratio of

interfacial area to volumetric NAPL content provides information on how the NAPL is

distributed within the pore space. Saripalli at al. (1997) proposed the ratio of NAPL-

water interfacial area to NAPL saturation as a NAPL morphology index (HN):


S = a.
SN (2-9)



where anw is the NAPL-water interfacial area per unit volume of porous media; Sn is the

NAPL saturation or volume of NAPL per unit volume of pore space; 0 is the soil

porosity. A high value in HN indicates that NAPL is spread out, thus providing large

contact area between the NAPL and aqueous phase. In contrast, small value in HN

indicates that the NAPL is locally isolated in small pools.



Materials and Methods



Materials

Sodium dodecylbenzenesulfonate (SDBS, Aldrich, 85%) and potassium bromide

(KBr, Fisher, 98%) were used as interfacial and non-reactive tracers, respectively, for

interfacial tracer tests. Tetrachloroethylene (PCE, Acros, 99 %), colored red with Oil-red-








O dye (< 1 x 10.2 mM, Fisher) was used as a NAPL for all column experiments. A suite

of partitioning tracers was selected to examine the effects of varying retardation factors.

Methanol (Fisher, 98 %) was used as a non-partitioning tracer while 3,3 dimethyl-2-

butanol (3,3DMB) (Acros, 99+ %) and n-hexanol (Acros, 98 %) were used as partitioning

tracers. Silica sand with various particle sizes (0.165 1.425 mm) was used as porous

media in all column experiments. The sand was sieved, washed with water, oven-dried at

105"C for 24 hr, and baked at 5500C for 24 hr. The baked sand was washed several times

with concentrated nitric acid to remove metals, washed with water and dried in oven for

24hr, before use as porous media. All experiments were conducted at room temperature

(23 1 C).

Interfacial Tension Measurement

NAPL-water interfacial tensions (y) were measured at various SDBS

concentrations ranging from 0 to 2 mM (0 697 mg/L) (Appendix A). Glass

crystallization vessels (50 mm diam. x 35 mm depth) were cleaned for IFT measurements

using the procedure of Wilson et al. (Wilson et al., 1990). The interfacial tensions (IFT)

between the PCE with 0.01 mM Oil-red-O dye and SDBS solution at various

concentrations were measured using a Du Nuoy ring tensiometer (Fisher Scientific Model

21 Tensiomat). The measured interfacial tensions were used to estimate the adsorption

coefficient of the interfacial tracer (SDBS).

Miscible Displacement Tests

The columns used for all miscible displacement experiments were glass columns,

2.54 cm in diameter and 5 cm in length (HPLC column, Kontes company) with Teflon

end pieces and fine wire mesh screen. The packing and assembling procedures of



























A :Bromide Solution
B : SDBS Solution
C : Water
D : HPLC Pump
E : Switching Valve
F : Waste
G: Sand /PCE Column
H : Sand Column
I : Adjustable Adapter
J : Sampling Vial


A B C


Figure 2-1. Experimental set-up








columns were accomplished under water to avoid trapped air, which can cause an error

due to the air-water interface.

Two series of miscible displacement tests were performed for this study. The

schematic diagram of the experimental set-up is shown in Figure 2-1. The first series of

miscible displacement tests was conducted using the columns packed with various sizes

of clean sand to investigate NAPL-water interfacial area, NAPL saturation, and

morphology as a function of pore size. Each column was packed with each sand size

(10/20, 20/30, 30/40, 40/60, and 80/100 mesh) up to 2.5 cm in length using an adjustable

adapter (Figure 2-1), and cleaned by passing 20 pore volumes (PV) of degased DI water.

Before injecting NAPL (PCE) for all the columns, both non-reactive tracer (Br) and

interfacial tracer (SDBS) tests (step-input) were conducted to measure background

adsorption of interfacial tracer on to sand, producing tracer retardation. The Br and

SDBS concentration used in this study were approximately 50 mg/L. Then, PCE was

introduced in the water-saturated columns using a syringe pump at gradually increased

water flow rates (0.1, 0.2, 0.5, 1, 3, 5, and 7 mL/min) until no water was observed in the

outlet of the columns. Some of the PCE was then displaced with water through both up-

and down-flow modes to create a homogeneous distribution of trapped PCE. Residual

saturation of PCE was achieved by gradually increased flow rates (0.1, 0.2, 0.5, 1, 3, 5,

and 7 mL/min steps). Non-reactive tracer and interfacial tracer tests were separately run

with residual NAPL to estimate NAPL-water interfacial areas applying step-inputs until

the relative concentration (C/Co) reached unity. Then, partitioning tracer tests (pulse

input) were conducted to measure the residual PCE volume or saturation.








The second series of miscible displacement tests was conducted to investigate anw,

HN, and Sn as a function of NAPL dissolution. A column packed with 20/30 mesh sand

from the previous series was used and prepared for each miscible displacement

experiment by flushing a 4 pore volume (PV) 50 % (by volume) ethanol solution through

both ends of the column. Each flushing resulted in reduction of about 20 % of the total

residual PCE. No free phase PCE was observed in the effluent during flushing. This

indicates that NAPL reduction was by dissolution, not mobilization. The columns were

then flushed with 5 PV of water. Non-reactive, interfacial, and partitioning tracer tests

were conducted next using the procedures described for the first series of experiments. A

HPLC pump (Shimadzu 500) was used to inject water and tracer solutions at a steady

flow rate (0.5 mL/min).

Analytical Methods

Samples from partitioning tracer tests were collected in 2 mL vials with a 100 tL

insert and analyzed for alcohols by Gas Chromatography (Perkin Elmer GC, Autosystem

XL) with a flame-ionization detector. Br (205 nm) and SDBS (225 nm), from interfacial

tracer experiments, were analyzed using a high performance liquid chromatography

(HPLC) system (Perkin Elmer series LC 200) with UV detector.



Results and Discussion



Estimation of Interfacial Adsorption and Partitioning Coefficients

The interfacial adsorption coefficient (K,) was estimated from a relationship of

NAPL-water interfacial tension and tracer bulk concentration based on Gibbs adsorption




















50
30
y=-7.57x+ 11.11

E 40
040
I \ 10


30
c -25 -1.5 -05 0.5 1.5
0
Ln(C
I-
20 20

20

10

0 '

0
0 0.5 1 1.5 2
SDBS concentration in solution (mM)


Figure 2-2. Interfacial tensions measured at NAPL-SDBS solution interfaces








equation (2-2). The interfacial tensions (y) measured at various PCE-tracer solution

interfaces are shown in Figure 2-2. The y values decreased in log-linear fashion with

increasing surfactant (SDBS) concentration. The relationship is

y =a lnC (2-10)



The observed a and f were 11.1 and 7.575, respectively, with a coefficient of

determination (R2) of 0.99. Using the Equation 2-2 with Co (= 0.14 mM) and T (= 297

K), the estimated K, value was 1.07x10"3 (cm). Note that the K, value increases with

decreasing tracer concentration (Co), causing an increase of the tracer retardation due to

interfacial adsorption. Additionally, the Co for the tracer solution should be below the

CMC.

The partitioning coefficients (K,,) for the alcohol tracers were measured in batch

isotherm tests. The tracer partition equilibrium data obtained from batch tests were fit

using a linear isotherm to obtain the K,,. The observed K,, values for 3,3 DMB and n-

hexanol were 4.8 and 6.8, respectively, with R2 of 0.998. The K,, values along with

retardation factors measured from column tests were used to estimate residual NAPL

saturation.

NAPL-Water Interfacial Area and NAPL Saturation as a Function of Pore Size

NAPL-water specific interfacial areas and NAPL saturations were investigated

through columns with various mean grain sizes ranged from 0.15 to 1.42 mm. First, tracer

experiments for all the columns in the absence of NAPL were conducted to measure the

background adsorption of the interfacial tracer onto sands used as porous media and the

column assembly. BTCs for non-reactive (Br) and interfacial (SDBS) tracers are shown











o x a do -o
Sx o x
0.8x 10/20 mesh x o 10/20 mesh
0.6 o x o
x x o
0.4 x xo

00 o xo
0.2 Xo

0 --o- -


x 0
0.81 xxx---- a 11 XXo 8 o 8


x8 x 20/30 xo 20/30
0.6 xo xo
a x
0.4x x o

0.2 x x o

0 xo

XX x xo
0.8 Kx o x o
x o

0.6 xo
xo x
0.4 x
x x
0.2 x o


o- .... xo
0

1 x o- g x x o 0 S "O 0

x 0
0.6 xo x o

0.4 0

xoo
o X
0.2 a

x x-- 0 ,--." x-x- -0---


0.8 x o x o 0 0

0.6 x0 xx

0.4 x
xo x o
0.2 xo x o

0 _. .. x
0 1 2 30 1 2 3


Pore Volume Pore Volume


Figure 2-3. Breakthrough curves of non-reactive tracer (bromide, symbol "x") and
interfacial tracer (SDBS, symbol "o") from columns with various grain
sizes as porous media. Shown are the plots on left-hand side without NAPL
and on right-hand side with residual NAPL (PCE).








Table 2-1. Experimental parameters and results for interfacial and partitioning tracer tests conducted with
various medium sizes.


Media dso 0 R(back) R(back+if Rif K K, anw RP Sn HN
(mm) cm'/g 103(cm) (cm2/cm3) DMB n-hexanol (cm1')

10/20 1.425 0.38 1.136 1.216 1.080 0.021 1.07 28 1.63 1.88 0.116 570
20/30 0.725 0.34 1.125 1.263 1.138 0.017 1.07 44 1.75 2.07 0.137 793
30/40 0.513 0.37 1.124 1.254 1.130 0.018 1.07 44 1.68 1.97 0.124 832
40/60 0.338 0.35 1.115 1.342 1.227 0.016 1.07 74 1.86 2.25 0.154 1129
80/100 0.165 0.34 1.119 1.462 1.343 0.016 1.07 110 1.89 2.25 0.155 1670





Table 2-2. Experimental parameters and results of column tests conducted after ethanol flushing.


Media Flushing 08, R R(back+it) Rit K, anw R, Sn fD HN
(ethanol 50%) 103(cm) (cm2/cm3) DMB n-hexanol (cm')

20/30 OPV 0.34 1.125 1.263 1.138 1.07 44 1.75 2.10 0.14 0 793
20/30 4PV 0.36 1.125 1.237 1.112 1.07 38 1.56 1.81 0.11 0.23 876
20/30 8PV 0.37 1.125 1.218 1.093 1.07 32 1.41 1.59 0.08 0.41 976
20/30 12PV 0.40 1.125 1.192 1.067 1.07 25 1.25 1.37 0.05 0.62 1178








on the left-hand side in Figure 2-3. The BTCs indicated that SDBS used as interfacial

tracer was retarded due to adsorption onto the sand or column assembly. The measured

background retardation factor (Rback) for all the columns were in range of 1.12 to 1.14 and

the corresponding adsorption coefficients (K,) were estimated to be from 0.016 to 0.021

(cm3/g) (Table 2-1). The observed background retardations are likely due to metal oxides

(e.g., FeO3, 0.02%; Al203, 0.06%; TiO2, 0.01%, etc., US. Silica Company, 2001) present

in the clean sand used. Even though the sands were washed with nitric acid to remove the

metal oxides, some likely remain. The background partitioning of alcohol tracers onto the

porous media (clean sand) was not observed, and thus the Kd in the Equation 2-7 was set

equal to zero.

Second, tracer experiments in the presence of residual NAPL (PCE) with Oil-red-

O dye (0.1 x 103 mM) were conducted to estimate the anw as well as the Sn. Since PCE

saturated water was used during all the column experiments, reduction of tracer

retardation by loss of PCE mass can be ignored (less than 1%). In addition, since the

concentration of SDBS solution (= 0.14 mM) used was much less than the CMC (about

1.2 mM), loss of PCE mass due to enhanced solubilization or mobilization during all the

experiments did not occur. Table 2-1 summaries the measured Ri,, aw, Rp, Sn, and HN

values along with column parameters. BTCs for Br and SDBS are shown on the right-

hand side in Figure 2-3. The degree of retardation of SDBS relative to Br was evident on

the sand sizes used as porous media. The measured retardation factors (Rback+ft) were

subtracted from the earlier measured Rback to evaluate the retardation of SDBS by NAPL-

water interface. Here we assume that two retardation domains are independent. The

resulting Ri, data were then used to calculate anw using the Equation 2-6. The measured











160 [


{u i----------------
0 0.4 0.8 1.2 1.6
dso

DATA Rhombohedral Tetrahedral ------ Log-Log fit


Figure 2-4. Geometric pore singlet models (Rhombohedral and Tetrahedral
packing), log-log linear fit, and measured data for a,w dso relationship.


120

100

80
2
60


i 40

20

0
0.116 0.137 0.124 0.154 0.1545
(10/20) (20/30) (30/40) (40/60) (80/100)
NAPL Saturation (Sn)

Figure 2-5. NAPL-water interfacial area (anw) and trapped NAPL saturation (Sn) for
different porous media sizes.








anw values (74 to 110 cm2/cm3) for grain size ranging in 0.15 to 0.425 mm agreed

reasonably with 77 to 91 cm2/cm3 for similar sand size (0.15 to 0.35 mm) observed in an

earlier study (Saripalli et al, 1998). The following empirical function was fit to the anw

-dso data:

In (a,,) = 3.54 -0.635 In (dso) R2 = 0.966 (2-11)



The anw increased exponentially with decreasing grain size used as packing media. This

implies that the NAPL ganglia trapped in a finer media are smaller in size but the number

of the ganglia increases exponentially, causing a corresponding increase in the anw. Such

observation is likely to be more obvious by comparing to a previous prediction model.

Saripalli (1997) proposed a pore singlet model to predict the specific NAPL-water

interfacial area as:

For a rhombohedral packing,



4.19S,2'3 (2-12)
r


For a tetrahedral packing,


5.02 S 2'3
anw=- (2-13)
r



where r is the mean radius of porous media. As shown in Figure 2-4, the trend in the

measured aw-dso relationship agrees with the predicted values using the pore singlet

models. The measured Sn = 0.11-0.15 were lower than typical S, = 0.15-0.2. It may be

















xx o
x 0
x
o After flushing 0 PV
x S= 0.14
x
x


x o
xo
X x








X 0
x o

o After flushing 8 PV
o S,= 0.08
x
o
x
x
o
'x
- -x-o i


Pore Volume


1 X o-o-
x o
.8 x o
o After flushing 4 PV
.6 o S= 0.11
x


x 0o
.2
x o

0 _--.-x : o.






1 x oX xox xoxAox Ox"
x
8 x
x o After flushing 12PV
6 x S= 0.05
0
4 x


2 X


0 x-oo
ol .- .________


Pore Volume


Figure 2-6. Breakthrough curves of nonreactive tracer (bromide, symbol "x") and interfacial
tracer (SDBS, symbol "o") from column tests after ethanol flushing with various
pore volumes (0, 4, 8, and 12 Pore volume).








due to the method used in this study to create residual NAPL, i.e., two directional

flushing (up and down modes) forming a homogeneous spatial NAPL distribution

(typical methods use one directional flow/flushing). The observed Sn values increased in

three discrete regions with respect to coarse, medium, and fine grain sizes rather than a

constantly increasing fashion. The Sn was correlated with the corresponding anw in Figure

2-5. The result shows that a,w can be modeled better using the d5o than Sn. For example,

while the Sn values for 0.725 and 1.42 mm dso were nearly the same as 0.154 and 0.155,

respectively, the corresponding an, values showed larger deviations of 74 and 110

cm2/cm3. This indicates that at a given NAPL saturation the specific interfacial area can

vary greatly depending on the pore sizes. Saripalli (1997) reported that for a medium

textured sand, the interfacial area was predicted to be a constant or independent of the

degree of saturation, but for a finer sand, the interfacial area increased exponentially with

decreasing water saturation. Moreover, interfacial area decreased sharply with increasing

NAPL saturation at Sn > 0.4. This suggests that NAPL-water specific interfacial area is

closely related to porous media size rather than residual NAPL saturation.

NAPL-Water Interfacial Area and NAPL Saturation as a Function of NAPL Dissolution

NAPL-water interfacial area as a function of residual NAPL reduction by

dissolution was evaluated using the column packed with 20/30 mesh sand (dso = 0.725

mm) used earlier. Sets of tracer experiments were conducted following flushing with a 4

PV ethanol solution (50 % by volume). BTCs for non-reactive (Br) and interfacial

(SDBS) tracers are shown in Figure 2-6. Retardation of the interfacial tracers with respect

to the non-reactive tracers were evident. The observed results, along with column

parameters, are given in Table 2-2. NAPL-water interfacial areas decreased in a linear














4.


I*


Sn
Data Rhombohedral ----Tetrahedral ------- Linear-fit


Figure 2-7. NAPL-water interfacial area (aNw) as a function of remaining NAPL
saturation (Sn) after ethanol flushing: Pore singlet models, linear fit, and measured data.


0 0.2 0.4 0.6 0.8 1

Volumetric Fraction of Dissolved NAPL (fD)
Data --Linear fit

Figure 2-8. NAPL-water interfacial areas (aNw) as a function of NAPL dissolution.








fashion with decreasing NAPL saturation by dissolution (Figure 2-7). The measured anw

-Sn data were fit to the following function:



a, = 226.3S,, +13.55 r2 0.998 (2-14)



Equation 2-14 can be written as a general expression for anw in terms of volumetric

fraction of dissolved NAPL (fD):



a., = a n f r2 =0.998 (2-15)



wherefo = (S,' Sn) / Sn, = 1 (Sn / Sn) is volumetric fraction of dissolved NAPL; S,* is

the initial NAPL saturation; S, is the NAPL saturation remaining after reduction of

NAPL by dissolution; ao is the initial interfacial area before reduction of NAPL by

dissolution; n is the dissolution factor, which is determined by the slope of the anw -fo

regression line. The n value observed for 20/30 mesh sand (dso = 0.725 mm) is 30.4 with

the regression coefficient (r2) of 0.997 (Figure 2-8). As shown in Figure 2-7, although

some deviation in the overall absolute values was observed, the decreasing trend and rate

in the measured anw-S, relationship agreed with the pore singlet model prediction of

Saripalli (1996). The linear decrease in NAPL saturation suggests a linear decrease in the

radius of the trapped NAPL blobs by dissolution. This suggests that the reduction of

NAPL saturation during dissolution causes no significant change in the spatial NAPL

distribution. This observation is important for NAPL distribution characterization in the

presence of slow dissolution processes, given that simple monitoring of NAPL








dissolution can be used to predict changes in interfacial area. Note that the empirical

model proposed in the Equation 2-15 can only be applied to predict the NAPL-water

interfacial area with respect to NAPL reduction by dissolution. Recall that the

experiments in this study were based on homogeneous NAPL distribution and porous

media. Therefore, the proposed aw-fD model may not be appropriate for heterogeneously

distributed NAPL. In the case of heterogeneous NAPL with varying pore size

distribution, the smallest ganglia are dissolved first followed gradually by the larger

blobs. This results in an exponential decrease in the anw by NAPL dissolution.

NAPL Morphology Index

The NAPL saturation and interfacial area data observed earlier can be combined

to calculate the morphology index to provide better NAPL characterization. The

morphology index (HN), which is the ratio of interfacial area to volumetric NAPL

content, was calculated using Equation 2-9. The measured HN values increased

exponentially with decreasing mean grain size (dso) (Figure 2-9). The HN-ds5 data were fit

by the following empirical function:



ln(H ) = 6.489 0.496 In(dso) r = 0.981 (2-16)



The fact that the HN values between a coarser sand (d5o = 1.425) and finer sand (dso =

0.165) represent a wide range from 646 to 2079 indicates that the NAPL in the finer sand

exists in much more spread out fashion. This implies that porous media size is a primary

factor in characterizing spatial NAPL distribution, as determined by the magnitude of HN.








2000




1500
E


1000




500


0 0.3 0.6 0.9 1.2 1.5
dso (mm)

Figure 2-9. Relationship of NAPL morphology index (HN) and porous medium size.
Shown are the symbols (data: *; Fit data: solid line).


2000



1500



E. 1000



500



0


0 Hn-anw-Sn
x x Hn-d50-Sn
0 Sn=0.15
---Sn=0.2
....... Sn=0.3
.. Sn=0.4

x X
o

......




0 0.4 0.8 1.2 1.
d5o (mm)


Figure 2-10. The HN estimated using HN-anw-Sn (Saripalli, 1997) and HN-dso-S,
relationships. Shown are also the predicted HN with respect to
varying NAPL saturation using the HN-dso-S. relationship.









The exponential trend is similar to the anw-dso relationship described earlier. Based on this

relationship, the HN function (Eq. 2-9) proposed by Saripalli et al. (1997) can be written

in the terms of d50 instead of anw using the earlier measured a,w-dso functional

relationship. An approximate modification of Eq. 2-11 yields:

a, = 34.4.dso- (2-17)



Substituting the Equation 2-14 into Equation 2-9, the morphology index is


34.4 -ds0o-
H, 34 (2-18)
S,


As shown in Figure 2-10, the estimated HN using Equation 2-18 agreed well with the HN

using Equation 2-9. This suggests that the proposed model provides reasonable prediction

of HN. For varying Sn (0.15, 0.2, 0.3, and 0.4) for all grain sizes used in this study, the

predicted HN values using the HN-dso-S, relationship (Eq. 2-18) are shown in Figure 2-10.

The resulting prediction shows that at a given grain size, HN decreases with increasing the

S, and at a given saturation the HN increases exponentially with decreasing the d50

Application of the empirical model for HN must be carefully considered, however, for

characterization of a zone with a heterogeneous media size distribution. Annable et al.

(1998) showed that the HN values could be vastly based on the heterogeneity of the

media. In the light of the difficulty in measuring specific interfacial area, the HN

expressed as a function of d5o is useful for better understanding of spatial NAPL

distribution. The observed results not only supply useful information for prediction of








spatial NAPL distribution but also can provide important information for planning

remedial strategies.



Conclusions

This chapter presents NAPL characterization using partitioning and interfacial

tracers. A series of laboratory column experiments were conducted to investigate NAPL-

water interfacial area, NAPL saturation, and morphology as a function of grain size and

NAPL dissolution, which can affect spatial NAPL distribution. The specific interfacial

area increased exponentially with decreasing mean grain size of the porous media,

whereas the NAPL saturation showed slight increase in three discrete regions of coarse,

medium, and fine grain sizes. This indicates that the interfacial area has a functional

relationship with mean grain size rather than NAPL saturation. The specific interfacial

area as a function of NAPL dissolution was evaluated at a grain size. The specific

interfacial area decreased linearly with decreasing in NAPL saturation by dissolution.

This suggests that a typical slow dissolution process causes no significant change in the

spatial NAPL distribution, but a decrease in the radius of the trapped NAPL blobs by

dissolution. It suggests that change of specific interfacial area by NAPL dissolution can

be predicted by monitoring the effluent NAPL concentration. The morphology index

(HN), which describes the degree of spatial NAPL distribution, also increased

exponentially with decreasing grain size. As a result, porous media size is an important

parameter for characterizing spatial NAPL. Additionally, this study showed that the

combined tracer techniques could be more useful tool for spatial NAPL characterization.













CHAPTER 3
INFLUENCE OF SPECIFIC INTERFACIAL AREA ON MASS
TRANSFER: NONEQUILIBRIUM PROCESS


Introduction

Nonaqueous phase liquid (NAPL) introduced in subsurface is trapped by capillary

forces that are related to the properties of the soil and the pore geometry. The trapped

NAPL releases mass into the bulk aqueous phase by dissolution, acting as a potential

long-term source of contamination. Removal of trapped NAPL is limited by mass transfer

processes. An understanding of mass transfer process, therefore, plays an important role

in efficient removal of the trapped NAPL and prediction of long term behavior.

Interphase mass transfer, which occurs in the vicinity of phase interfaces, by

diffusion and advection is a function of the driving force and interfacial area between the

two phases of concern. Mass transfer processes from solid or liquid spheres to a flowing

fluid phase have been investigated over the last three decades. Nernst (Sherwood et al.,

1975) is credited with early work on mass transfer: A stagnant film model, which is based

on solute mass diffusion through a thin, stagnant boundary layer of fluid surrounding a

solid particle (Sherwood et al., 1975). An application to NAPL-aqueous phase mass

transfer in porous media was discussed by Pfannkuch (1984) and has been advanced by

numerous studies (Hunt et al., 1988; Miller et al., 1990; Powers et al., 1991, 1992; Geller

and Hunt, 1993; Imhoff et al., 1997; Cunningham et al., 1997; Soerens et al., 1998; Zhu

and Skyes, 2000).








The mass transfer process between NAPL and the aqueous phase has been

described using the local equilibrium assumption (Abriola and Pinder, 1985; Miller et al.,

1990; Parker et al., 1990). Local equilibrium implies that if the concentration of

contaminant in one phase is known, the concentration in other phase at the same spatial

location can be described by equilibrium partitioning relationships (Abriola and Pinder,

1985). Powers et al.(1991), however, showed that the local equilibrium assumption may

not be valid at higher aqueous fluid velocity. Other investigations have shown that

nonequilibrium mass transfer is important in such cases (Borden and Kao, 1992; Geller,

1990; Hunt et al., 1988; Powers et al., 1992). Nonequilibrium can be expected in 1) rate

limited exchange of mass from NAPL to aqueous phase (Hunt et al., 1988; Webber et al.,

1991; Powers et al.,1992), 2) flow bypassing around the NAPL contaminated region

(Soerens et al., 1998), and 3) nonuniform NAPL distribution (Soerens et al., 1998) and

nonuniform flow of aqueous phase due to heterogeneity of soil (Abriola and Pinder,

1985). These nonequilibrium processes have a significant role in fate and transport of

contaminants and remedial strategy.

To date, most studies related to NAPL/aqueous mass transfer processes have used

a lumped coefficient defined as the product of a mass transfer coefficient and the specific

interfacial area (Miller et al., 1990; Powers et al., 1991, 1992; Imhoff et al., 1997;

Soerens et al., 1998). To date, use of a lumped bulk mass transfer coefficient rather than

an intrinsic mass transfer coefficient is done due to the difficulty in measuring specific

interfacial area. Some studies have used a pore structure model closely related to specific

interfacial area to explain the effect of the interfacial area on mass transfer rate (Hunter et







al., 1988; 1990; Powers et al., 1992). However, there have been no studies using

measured specific interfacial area between NAPL and the aqueous phase.

The primary objective of this study was to illustrate how the specific interfacial

area influences the mass transfer process. Bulk and intrinsic mass transfer coefficients

were investigated as a function of specific interfacial area, pore velocity, grain size, and

NAPL saturation. The impact of NAPL (or interfacial area) reduction by dissolution on

mass transfer was also investigated. Models using dimensionless Sherwood number to

predict steady state mass transfer rates were developed and evaluated by comparing

measured and predicted values.



Theoretical Background



The mass transfer rate in a NAPL/aqueous system is a function of the driving

force and the interfacial area between phases. A first-order mass transfer is often used to

describe the mass flux of a solute between the phases:

I dm
J k, (C, -C) (3-1)
A-, dt

where J is mass flux from the non-aqueous phase to the aqueous phase [M/L2T]; k,, is the

intrinsic mass transfer coefficient [LT']; Anw is the interfacial area between NAPL and

water [L2 ]; Cs is the equilibrium concentration of the solute in the aqueous phase [M/L ];

C is the concentration of the solute in the bulk aqueous phase [M/L3]. Equation (3-1) can

be written in terms of the rate of change of mass per unit volume of porous media:

1 dm dC A
1 kV(C, -C) = a, k,(C -C) = K,(C, -C) (3-2)
Vdt dt V







where V is the bulk volume of soil [L3 ]; an, is the specific interfacial area [L2/L3]; KI =

aw-ki is the bulk mass transfer rate coefficient [L/T]. A steady state mass transport

equation can be written incorporating the mass flux equation (Equation 3-2). The

governing equation is:

aC a2C ac
+K,(C-C,)=D, -v- (3-3)
at Lx' ax

For the following boundary conditions:

C (0, t) =0; = 0 (3-4)
ax

The solution is (Carslaw and Jaeger, 1959; Van Genuchten and Alves, 1982):

C(x) 1 v v2+ 4D, K,
1-exp- x] (3-5)
C, 2DL

where DL is the longitudinal dispersion coefficient [L2/T]; v is the pore water velocity (

q / () [L/T]; q is the Darcy velocity [L/T]; ] is the porosity; x is the length of the column

[L]. This can be solved for Kj:



1 2D C
K, = [(v -2- ln(l ))2 -v2] (3-6)
4D x C,

The intrinsic mass transfer coefficients (ki) can be determined using the measured a,w:

k = K (3-7)


Materials and Methods








Experimental Methods

Experiments to determine mass transfer coefficients were conducted following the

interfacial and partitioning tracer tests in Chapter 2. The schematic diagram of the

experimental set-up is shown in Figure 2-1. The experiments were conducted at various

steady flow rates (0.5, 1, 2, 3, 5, and 7 ml/min). All the experiments were prepared by

passing 10 pore volumes (PV) of DI water to reach at a steady state condition. PCE

samples from the experiments were collected in 2 ml vials with a 100 Rl insert and

analyzed using a high performance liquid chromatography (HPLC) system (Perkin

Elmer series LC 200) with UV detector (230nm). A HPLC pump (Shimadzu 500) was

used to inject water and all experiments were conducted at room temperature (23 1 C).



Results and Discussion



Parameter Estimation

An analytical solution to the advective-dispersive equation for a saturated

homogeneous porous medium was used to estimate longitudinal dispersion coefficients

(Freeze and Cherry, 1997). For all grain sizes used, non-reactive tracer (Br-) experiment

data at a Darcy velocity, q = 0.1 (cm/min), were fit to the analytical solution to obtain the

longitudinal dispersion (DL) and longitudinal dispersivity (aL). For high velocities in this

study, the coefficients of longitudinal dispersion were estimated from

DL = aL+ Dd (3-8)

where Dd is the coefficients of bulk diffusion; v is the pore water velocity in the flow

direction (cm/min). Here, the coefficient of bulk diffusion can be negligible since it is










Table 3-1. Mass transfer results and parameter values measured from column experiments with various grain sizes.


0 do v a DL C Sn an K; kv
mm (cm/min) cm (cm2/min) (mg/L) (cm2/cm3) (min-l) (cm/min)


10/20 mesh 0.38 1.43 0.3
0.38 1.43 0.5
0.38 1.43 1.0
0.38 1.43 1.6
0.38 1.43 2.6
0.38 1.43 3.7
20/30 0.34 0.73 0.3
0.34 0.73 0.6
0.34 0.73 1.1
0.34 0.73 1.7
0.34 0.73 2.9
0.34 0.73 4.0
30/40 0.37 0.51 0.3
0.37 0.51 0.5
0.37 0.51 1.1
0.37 0.51 1.6
0.37 0.51 2.7
0.37 0.51 3.8
40/60 0.35 0.34 0.3
0.35 0.34 0.6
0.35 0.34 1.1
0.35 0.34 1.7
0.35 0.34 2.8
0.35 0.34 4.0
80/100 0.34 0.17 0.3
0.34 0.17 0.6
0.34 0.17 1.2
0.34 0.17 1.7
0.34 0.17 2.9
0.34 0.17 4.1


0.103 0.027 119
0.103 0.054 97
0.103 0.108 78
0.103 0.162 66
0.103 0.270 50
0.103 0.378 45
0.059 0.017 147
0.059 0.034 143
0.059 0.068 129
0.059 0.102 118
0.059 0.171 103
0.059 0.239 91
0.055 0.015 150
0.055 0.030 144
0.055 0.060 132
0.055 0.090 124
0.055 0.149 107
0.055 0.209 95
0.036 0.010 150
0.036 0.021 147
0.036 0.041 146
0.036 0.062 143
0.036 0.103 138
0.036 0.145 136
0.028 0.008 150
0.028 0.016 150
0.028 0.032 149
0.028 0.048 149
0.028 0.080 149
0.028 0.112 146


0.116 28.3
0.116 28.3
0.116 28.3
0.116 28.3
0.116 28.3
0.116 28.3
0.137 44.3
0.137 44.3
0.137 44.3
0.137 44.3
0.137 44.3
0.137 44.3
0.124 44.4
0.124 44.4
0.124 44.4
0.124 44.4
0.124 44.4
0.124 44.4
0.154 73.7
0.154 73.7
0.154 73.7
0.154 73.7
0.154 73.7
0.154 73.7
0.155 109.6
0.155 109.6
0.155 109.6
0.155 109.6
0.155 109.6
0.155 109.6


0.17 0.006
0.23 0.008
0.31 0.011
0.38 0.013
0.43 0.015
0.53 0.019
0.48 0.011
0.75 0.017
0.93 0.021
1.10 0.025
1.36 0.031
1.53 0.035
0.76 0.017
0.73 0.017
0.95 0.021
1.17 0.026
1.39 0.031
1.56 0.035
0.76 0.010
0.93 0.013
1.70 0.023
2.19 0.030
3.03 0.041
3.88 0.053
0.84 0.008
1.52 0.014
2.55 0.023
3.67 0.033
6.33 0.058
6.30 0.058











0.8

0.6

0.4

0.2 fitted data
I measured data
0 --
0 5 10 15 20 25
Elapsed Time (min)

Figure 3-1 An example BTCs of non-reactive tracer (Br) fitted to an analytical
solution to estimate longitudinal dispersion coefficient and dispersivity
values.



small compared to DL. The estimated longitudinal dispersivities for porous media with dso

= 0.17-1.43 mm ranged from 0.03 to 0.1 cm. An example BTC is shown in Figure 3-1.

These longitudinal dispersivities were in a reasonable agreement with typical values (0.01

to 1 cm) found for longitudinal dispersivities from column experiments (Freeze and

Cherry, 1979). Additionally, NAPL saturation (Sn) and specific interfacial area (anw) used

in this study are the values measured using partitioning/interfacial tracers in Chapter 2.

The results are given in Table 3-1 and 3-2.

Mass Transfer Coefficients

A series of dissolution experiments was conducted to investigate mass transfer

between trapped NAPL in porous media and the aqueous phase. Normalized effluent

concentrations (C/Cs) as a function of pore water velocity (v = q / 0) are shown for

individual grain sizes in Figure 3-2. The effluent PCE concentration decreased with












8 xx

0 A0
o A


o od50=1.43
0 0 d50=0.73
0 A d50=0.51
x d50=0.34
-d50=0.17


0 1 2 3 4 5
Pore Water Velocity (v, cm/min)


Figure 3-2. Normalized PCE effluent concentration (C/Cs) versus pore
water velocity. (dso units mm)


1.2

1 I

0.8 o

S0.6 d50=1.43
O o d50=0.73
0 A d50=0.51
0.4
SX d50=0.34
-d50=0.17
0.2

0
0 -----------------i

0 20 40 60 80 100 120

Specific Interfacial Area (anw, cm /cm3)

Figure 3-3. Normalized PCE effluent concentration (C/C.) versus
specific interfacial area. (dso units mm)









8


6


E 4


2


0


10 20
Re


30 40


Figure 3-4. Bulk mass transfer rate coefficients (KI) versus Reynolds
number (Re) for various grain sizes. (dso units mm)


O q=0.1
0 q=0.2
A q=0.39
x q=0.59
-q=0.99
+ +q=1.38


x

0


0 50 100 15
Specific Interfacial Area (an, cm2 /cm3)


Figure 3-5. Bulk mass transfer rate coefficients (KI) versus specific
interfacial area for various velocities. (q units cm/min)


Od50=1.43
xx X d50=0.73
A d50 =0.51
0 d50 =0.34
X d50=0.17


0



fo a o
o 9 00


8


6



E 4


2



0








increasing pore water velocity and grain size. Typically the effluent concentrations

increased with increasing NAPL saturation, but some cases did not follow this trend

(Table 3-1). For example, although NAPL saturations for two different grain sizes (do =

0.17 and 0.34 cm) were nearly the same, the observed effluent concentrations were higher

for the finer grain size, dso = 0.17 cm, than for dso = 0.34 cm. In addition, effluent

concentrations for dso = 0.51 cm were slightly higher than those for d5o = 0.73 cm even

though NAPL saturation was larger. The effluent concentrations, alternatively, increased

consistently with increasing interfacial areas (Figure 3-3). This means that NAPL-water

mass transfer rate is well correlated with NAPL-water specific interfacial area rather than

residual NAPL saturation for soil with various grain sizes or heterogeneity.

The measured results were pooled together to calculate bulk mass transfer rate

coefficients (KY = kiaw) using Equation 3-6. The bulk mass transfer rate coefficients as a

function of Reynolds number, which represents pore velocity, are shown in Figure 3-4

and again as a function of specific interfacial area in Figure 3-5. It was evident that at the

smallest grain size (largest interfacial area) small increase in Reynolds number resulted in

very sharp increases in the bulk mass transfer rate coefficients. While at low interfacial

area (large grain size) the bulk mass transfer rate coefficients did not vary significantly

with pore velocity. At high interfacial area (small grain size) mass transfer coefficients

were very sensitive to pore velocity.

The bulk mass transfer rate coefficients were used to calculate intrinsic mass

transfer coefficients (ki) using specific interfacial area values. Figure 3-6 shows a plot of

observed intrinsic mass transfer coefficients as a function of specific interfacial areas.

The intrinsic mass transfer coefficients varied slightly or did not vary with increasing










Table 3-2. Mass transfer results with respect to reduction of NAPL saturation/interfacial area

Flushing 0 v DL C S. a., K, k,
(ethanol 50%) (cm/min) (cm2/min) (mg/L) (cm2/cm3) (min-') (cm/min)

OPV 0.34 0.29 0.017 147 0.137 44 0.48 0.011
0.34 0.57 0.034 143 0.137 44 0.75 0.017
0.34 1.15 0.068 129 0.137 44 0.93 0.021
0.34 1.72 0.102 118 0.137 44 1.10 0.025
0.34 2.87 0.171 103 0.137 44 1.36 0.031
0.34 4.02 0.239 91 0.137 44 1.53 0.035
4 PV 0.36 0.27 0.017 147 0.106 38 0.46 0.012
0.36 0.55 0.034 137 0.106 38 0.57 0.015
0.36 1.10 0.068 123 0.106 38 0.78 0.021
0.36 1.65 0.102 112 0.106 38 0.93 0.025
0.36 2.75 0.171 88 0.106 38 0.99 0.026
0.36 3.85 0.239 78 0.106 38 1.16 0.031
8PV 0.37 0.27 0.017 144 0.081 32 0.38 0.012
0.37 0.53 0.034 133 0.081 32 0.49 0.015
0.37 1.07 0.068 112 0.081 32 0.61 0.019
0.37 1.60 0.102 101 0.081 32 0.73 0.023
0.37 2.67 0.171 78 0.081 32 0.80 0.025
0.37 3.73 0.239 69 0.081 32 0.93 0.029
12 PV 0.40 0.25 0.017 127 0.052 25 0.19 0.008
0.40 0.49 0.034 114 0.052 25 0.29 0.012
0.40 0.98 0.068 89 0.052 25 0.36 0.015
0.40 1.48 0.102 73 0.052 25 0.40 0.016
0.40 2.46 0.171 60 0.052 25 0.51 0.020
0.40 3.44 0.239 54 0.052 25 0.62 0.025

* The dispersivity (a) used to estimate dispersion coefficients (DL) is 0.059 cm. The grain size used is 0.73 mm (do).









0.1


0.08


S0.06


0.04

0.02


0


Oq=O.1
Oq=0.2
Aq=0.39
Xq=0.59
+ -q=0.99
+ +q=1.38


+ x
x

o B 0


3 50 100 1i
Specific Interfacial Area (anw, cm2/cm3)


Figure 3-6. Intrinsic mass transfer coefficients versus specific interfacial area
for various velocities. (q units cm/min)


2
Oq=0.1

1.6 Oq=0.2
Aq=0.39
Xq=0.59
r 1.2 +

+ +8q=1.38
-90.8 A
a a +

0.4

0 ----------------
0 0.2 0.4 0.6 0.8
Fraction of NAPL removed by dissolution (fo)


Figure 3-7. Bulk mass transfer rate coefficients (KI) as a function of NAPL
dissolution. fD = (Sn* Sn)/ S* = 1 (Sn / Sn*) is volumetric fraction of
dissolved NAPL; Sn* is the initial NAPL saturation; S. is the NAPL
saturation remained after reduction of NAPL by dissolution.








specific interfacial area except for some data points at high velocity and high interfacial

area. Miller et al. (1990) reported that bulk mass transfer rate coefficients were extremely

sensitive at high effluent concentration. Therefore, it is likely that the observed large

intrinsic mass transfer coefficients at high interfacial areas can be explained by

uncertainty in system factors or potential experimental errors rather than the change of

interfacial area/NAPL saturation.

Influence of NAPL Reduction by Dissolution on Mass Transfer

Mass transfer as a function of NAPL (or interfacial area) reduction by dissolution

was investigated and the observed results are shown in Table 3-2. Here, all the

experiments were completed in a column that started at residual saturation (Sn = 0.14).

Each set of experiments was conducted following a 4PV ethanol (50 % by volume) flood,

that resulted in a NAPL reduction of about 20 %. It is assumed that a change in mass

transfer is only a result of NAPL saturation (or interfacial area) reduction. Figure 3-7

shows a plot of bulk mass transfer rate coefficients ((K) versus the fraction of removed

NAPL (fD). It is apparent that bulk mass transfer rate coefficients decrease with NAPL

reduction. The observed bulk mass transfer rate coefficients decreased in near linear

fashion with decreasing NAPL saturation. The KI values became less sensitive to pore

velocity with decreasing NAPL saturation. This observation agreed with earlier results.

These results imply that at higher NAPL saturation and interfacial area, alcohol flushing

efficiency is not sensitive to flushing velocity, while at lower NAPL saturation velocity

can be an important control on the efficiency of NAPL removal.

Figure 3-8 shows that while intrinsic mass transfer coefficients increase

consistently with increasing velocity, they change little with reduction of NAPL. This










0.08


0.06


E 0.04
0.02

0.02


0 0.2 0.4 0.6 0.8
Fraction of NAPL removed by dissolution (f,)

Figure 3-8. Intrinsic mass transfer coefficients (ki) as a function of NAPL
dissolution.


0.05


0.04
o
0.03 o x
E
x
a 0.02


0.01 a

0
0 0.5 1 1.5 2
V (cm/min)


Figure 3-9. Averaged intrinsic mass transfer coefficients for various
velocities. The symbol (o) and symbol (x) are the averaged ki
values from the experiments with various grain sizes and with
NAPL reduction by dissolution, respectively.


Oq=0.1
Oq=0.2
A q=0.39
X q=0.59
-q=0.99
+q=1.38


A A o
0







Table 3-3. Mass Transfer Correlations.


References Correlation model type

Miller et al. (1990) Sh' = 12 Re0.75,n0.6Sc0.5 lumped domain
Powers et al. (1992) Sh'= 57.7 Reo61d5o 64Ui41 lumped domain
Imhoff et al. (1997) Sh' = 1.34 Re0.750no.9Sc0.49 lumped domain

where Sh' (= KIds50/Dm) is the modified Sherwood number Miller et al., 1990); Re (
vpwdso/pw) is the Reynolds number, Sc (=pw/pwDm) is the Schmidt number; On is (= do6/dio) is the uniformity index; Dm is the molecular diffusivity in the aqueous phase;
pw is the aqueous density; gw is the aqueous viscosity.


indicates that intrinsic mass transfer coefficients can be independent of NAPL

saturation/interfacial area, but only dependent on velocity. The observed results were

averaged and compared with the averaged values from the experiments with various grain

sizes earlier in Figure 3-9. Results from two different experimental approaches agreed

well over the velocities tested.

Model Development

In general, models related to steady state mass transfer have been developed or

evaluated using a dimensionless Sherwood number (Sh). Numerous previous studies

(Table 3-3) have shown that the Sherwood number was an appropriate tool to explain or

correlate mass transfer process (Miller et al., 1990; Powers et al., 1991, 1992; Geller and

Hunt, 1993, Imhoff et al., 1997). In this present work, both modified Sherwood number

(Sh' = KIdso/Dm) (Miller et al., 1990) and Sherwood number (Sh = kjL/Dm), where L is a

characteristic length which can be expressed as mean grain size, dso, were used in model

development. The parameters used in the correlation were Reynolds number (Re),

Schmidt number (Sc), Sn, anw, and dso. Since the Sc was constant over these experiments,

an exponent of 0.5 was used for Sc as suggested by Miller et al., (1990) and Imhoff et al.







(1997). The data for parameters were log-transformed and fit to multiple linear models

within a SPSS statistical software package (SPSS for windows 8.0.0, SPSS Inc., 1997).

The fitting was performed by stepwise regression procedures to find the optimum

correlation combination of parameters. The best fit was measured with the sum of the

squares defined as


r2= log(Sh ,)- log(Sh. d) (3-9)
i-L log(Sh_,)

where n is the number of data points for each parameter used. The first model developed

used a modified Sherwood number which included lumped mass transfer rate coefficients

(KI = kanw). The model, which simultaneously incorporates all of the Re, Sc, Sn, anw, and

dso terms, did not offer meaningful correlation. Modified Sherwood number was also not

well correlated with other multiple parameter comibations, Re/Sn, Re/anw, and Re/dso/Sn

at the significance level. The best fit model in power law form is given as

Sh' = ,1 Re" d_52 Sc". (r = 0.92; 95 % CI) (3-10)

with 30 = 3.09 1.09
pl = 0.45 0.04
02 = 0.40 0.08


and the best fit model including anw in power law form is

Sh' = P' Re"' a,_2' d,3' (r = 0.92; 95 % CI) (3-11)

with 30' = 3.50 2.40
pl' = 0.45 0.04
12' = 0.49 0.39
13' = 0.70 0.25








Powers et al. (1992) reported that the inclusion of any parameters describing pore

structure such as the grain size distribution greatly increased r2 values and thus can be

used as representative measures of NAPL distribution and interfacial area. In this study,

the modified Sherwood number is also well correlated to grain size and specific

interfacial area. This indicates that the mass transfer rate can be correlated with pore size

and NAPL specific interfacial area as well as pore velocity. This developed model (Eq. 3-

10) along with the previous model (Powers et al., 1992) can be a useful tool since

measurement of NAPL-water interfacial area is difficult. When an irregular spatial

distribution of NAPL is considered, however, the model (Eq. 3-11), which includes the

specific interfacial area, can be more useful in predicting the mass transfer rate between

the phases than that using (Eq. 3-10).

Comparison of the models (Eq. 3-11) developed in this study with previous

investigations using the same modified Sherwood number (Miller et al., 1990; Powers et

al., 1992; Imhoff et al., 1997) are shown in Figure 3-10. The model (Eq. 3-11) was

closest to that of Miller et al. (1990) but between those of Powers et al. (1992) and

Imhoff et al. (1997). This deviation seems to be due to the difference in the measured

absolute values from experiments rather than the discrepancy in the developed models.

Note that while bulk mass transfer rate coefficients increased with increasing effluent

concentration, modified Sherwood numbers decreased. Powers et al. (1992) conducted

experiments with trichloroethylene (TCE) and the resulting effluent concentrations were

lower than those using PCE in this study. This may result in larger modified Sherwood

numbers. Although Imhoff et al. (1997) used PCE in their experiments, the observed










1000



100



1 10






0.1


Figure 3-10. Comparison of modified Sherwood numbers (Sh') from this study with
previous investigation (dso= 0.073 cm, a,, = 44 cm2/cm3. Sn = 0.14, =
0.39, UI= 1.2)


0 2 4 6 8 10
Measured KI


Figure 3-11. Bulk mass transfer rate coefficients measured versus those predicted from
the developed model (Re/aw./dso model) for all data pooled together.


- This Work
- Powers et al. (1992)
-Miller e al. (1990)
- Imhoff e al. (1997)


c







effluent PCE concentrations were higher than those from this work. The higher effluent

concentration caused relatively smaller modified Sherwood numbers.

The Sherwood number (Sh = kdso/Dm), which includes the intrinsic mass transfer

coefficient (k,) was used to develop a model for the intrinsic mass transfer coefficient.

The best fit model in power law form is given as

Sh = P1 Re"' Sc'5 (r2 = 0.87; 95 % CI) (3-12)

with 0O" = 0.27 1.01
pi" = 0.57 0.04

This best fit model had relatively lower regression constants compared to those with the

modified Sherwood number. While the Sherwood number was not correlated to the Sn,

anw, and dso at a significant level, it was relatively well correlated to Re which includes

the velocity. As described earlier, considering that the intrinsic mass transfer coefficient

(ks) can be nearly independent of Sn, aw, and dso but dependent on flow velocity, the

correlated model seems to be reasonable. This model (Eq. 3-12) can be valuable in the

fact that it can be used to predict intrinsic mass transfer coefficients between NAPL and

the aqueous phase without measuring a specific interfacial area.

Validation of the correlated models was assessed by comparison with measured

data. The bulk mass transfer rate coefficients (Ki) predicted from the developed models

(Eq. 3-10 and Eq. 3-11) were compared with the measured values in Figure 3-11.

Although some data points scattered above the 1:1 line, most predicted values agreed

with the measured data. Average predicted values for Eq. (3-10) and Eq. (3-11) were 1.13

and 1.11 times higher than the measured values from experiments, respectively.

Comparison of measured and predicted values for intrinsic mass transfer coefficients (ki)

is shown in Figure 3-12. The predicted values were obtained using Eq. (3-11) and Eq. (3-









0.1

0.08

0.06

S0.04 x *




0 =
0 0.02 0.04 0.06 0.08 0.1
Measured k,

Figure 3-12. Intrinsic mass transfer coefficients measured with those predicted
from the model (Re/Sc) for all data pooled together.


Figure 3-13. Damkohler number (Da = ki/v) with respect to Reynolds number
for various grain sizes. (dso units mm)


Equilibrium Zone d50=1.43
k,/v > 0.02 0 d50=0.73
Re <3
Re <3 d50=0.51
x d50=0.34
A X d50=0.17


Nonequilibrium Zone

Nonequilibrium Zone








0.1





0.01





0.001


q0.1
Sq=0.2
Sq=0.39
I x q=0.59
..................................... q = 0 .99
x"' q=1.38


Zone2 : Near or Equlibrium
q/v > 0.02
Zonel : Nonequilibrium a, > 110


10 100 1000
a.. (cm'/cm')

Figure 3-14. Damkohler number (Da = ki/v) with respect to specific interfacial
area for various grain sizes. (dso units mm)


12). Average predicted values for Eq. (3-11) and Eq. (3-12) were 1.07 and 1.01 times

higher than average measured values, respectively. The small deviations indicate that the

developed models are valid as prediction models for mass transfer rate coefficients.

Nonequilibrium Mass Transfer

Equilibrium between NAPL and water reflects the conditions induced by

physichemical environments such as velocity (Powers et al., 1992), temperature (Imhoff

et al., 1997), contact time and concentration (Miller et al., 1990). The Damkohler number

(Da = K(dso/v = kianwdso/v = k/v), which characterizes the degree of nonequilibrium

between fluids, was used to investigate how system parameters contribute to

nonequilibrium conditions. Figure 3-13 shows Da (= k/v) versus Re, which incorporates

the velocity, for a range of grain size. The higher Da numbers indicate near equilibrium

conditions. That is, at ki/v greater than about 0.02 and Re less than about 3 in systems

with homogeneous grain size and NAPL distribution (present system conditions), the

equilibrium assumption is appropriate. It was clear, however, that at Re > 3 the Da

decreased log-linearly with increasing the Re. The decreasing trend with respect to








velocity was accelerated with increasing the grain size of porous media. Because specific

interfacial area can increase exponentially with decreasing grain size, the equilibrium

condition at different fluid velocity can be sensitive to the magnitude of interfacial area as

well as grain size (Figure 3-14 and 3-3). This means that above a critical interfacial area

(anw & 110 cm2/cm3 in present system conditions), effluent concentration did not vary

with increasing velocity. As NAPL-water interfacial area decreased, velocity can be an

important factor in determination of equilibrium conditions. In other words, below a

critical velocity (q 0.01 cm/min in present system conditions), mass transfer rate is

nearly independent of the magnitude of specific interfacial area. These results indicate

that above a critical interfacial area or below a critical velocity local equilibrium

assumption is achieved.



Conclusions



The effect of specific NAPL-water interfacial area, along with other system

parameters (e.g., grain size, velocity, and saturation), on mass transfer was investigated.

Bulk mass transfer rate coefficients were found to be directly related to specific

interfacial area as well as pore velocity and saturation. Particle size which was closely

related to the specific interfacial area was also an important parameter in determination of

bulk mass transfer rate coefficients. The bulk mass transfer rate coefficients can be

correlated with NAPL-water specific interfacial area rather than residual NAPL

saturation for soil with various grain sizes, and were more sensitive at high NAPL-water

interfacial area than at low interfacial area. However, intrinsic mass transfer coefficients





56

were nearly independent of the change of NAPL saturation and specific interfacial area,

but dependent on pore velocity. NAPL saturation (or interfacial area) reduction by

dissolution caused a linear decrease in bulk mass transfer rate coefficients, while it

caused no effect on intrinsic mass transfer coefficients. Correlation models consisting of

Reynolds number, Schmidt number, interfacial area, and grain size have been developed

to provide better understanding and prediction of intrinsic and bulk mass transfer

coefficients using both dimensionless Sherwood number and modified Sherwood

number. The developed models were verified throughout comparison of measured and

predicted data. The comparison revealed that the models were robust and produced

reasonable prediction for mass transfer rate.













CHAPTER 4
ESTIMATING NAPL SATURATION USING PARTITIONING TRACERS:
INFLUENCE OF RESIDUAL COSOLVENTS


Introduction


The partitioning tracer technique has been used to characterize residual saturation

and distribution of non-aqueous phase liquids (NAPLs) trapped in porous media

(Annable et al., 1998 a, b, 1995; James et al., 1997; Dwarakanath et al., 1999; Jin et al.,

1997). The technique is based on the differences in travel time of non-partitioning and

partitioning tracers through a NAPL source zone (Jin et al., 1995). The tracer technique

has been evaluated at both field (Annable et al., 1998a, b, 1995; James et al., 1997;

Dwarakanath et al., 1999; Jin et al., 1997) and laboratory (Jin et al., 1997, 1995; Pope et

al., 1994) scales. The technique has been mainly employed at sites associated with

aggressive in-situ remediation, such as cosolvent or surfactant flushing (Rao et al., 1997).

After a cosolvent flood, which often includes a water flood, some residual cosolvent will

likely remain in the swept zone following the effort to extract NAPL from the source

zone (Annable et al., 1998a). The residual cosolvent can affect the partitioning and

transport behavior of tracers used during a post-flushing tracer test.

In general, cosolvents present in the aqueous phase affect chemical

characteristics, such as solubility and sorption (hence, activity). In completely water-

miscible solvents, the solubility of hydrophobic organic chemicals (HOCs) increases in a

log-linear manner (Li et al., 1994; Pinal et al., 1990; Morris et al., 1988; Rubino et al.,








1987; Fu and Luthy, 1986; Yalkowsky and Roseman, 1986), and sorption decreases log-

linearly (Bouchard, 1998b; Hermann and Powers, 1998; Kimble and Chin, 1994;

Brusseau et al., 1991; Rao et al., 1990, 1985; Fu et al., 1986); as a result, the transport of

HOC through soils is less retarded (Rao et al., 1985; Bouchard, 1998a; Wood et al.,

1990). However, the behavior of chemicals can vary depending on type and composition.

For example, Coyle et al. (1997) reported that the solubility of HOCs, such as PCB-47,

PCB-153, and biphenyl, was depressed in the presence of partially miscible organic

solvents (PMOS), such as methylene chloride and chloroform.

As a result, in the presence of a residual cosolvent, an estimate of NAPL

saturation based on partitioning tracer test can be in error. The NAPL saturation (Sn) can

be under-estimated in the presence of cosolvents such as methanol, which cause

solubility enhancement of organic tracers. On the other hand, Sn can be over-estimated in

the presence of other cosolvents such as methylene chloride, which cause solubility

depression. Therefore, when a partitioning tracer test is conducted with a residual

cosolvent present, it is important to consider the cosolvent effects. Relatively little data,

however, are available on the effect of residual cosolvent on the partitioning tracer

technique.

The objective of this study was to investigate the influence of residual cosolvent

on the partitioning tracer technique for estimating Sn. Partitioning and solubility behavior

of alcohol tracers were investigated in the presence of residual cosolvents in batch

equilibrium tests. The results were verified through miscible displacement tests in packed

columns. It was also examined how the magnitude of residual Sn modifies the impact of

the cosolvent on the Sn estimates.








Theoretical Background



Addition of cosolvents generally increases aqueous solubility and decreases the

partitioning of HOCs; however, not every cosolvent increases the solubility of HOCs.

Most completely water-miscible organic solvents such as methanol, increase HOC

solubility in a log-linear (Li et al., 1994; Pinal et al., 1990; Morris et al., 1988; Rubino et

al., 1987; Fu et al., 1986; Yalkowsky and Roseman, 1986) fashion, since this type of

cosolvent results in a decrease of the activity coefficient. However, some partially water-

miscible organic solvents such as chloroform decrease HOC solubility in a log-linear

fashion (Coyle et al., 1997). One possible explanation is increasing activity of the PMOS

into the HOC phase or a solventing-out, in which dissolved PMOS can occupy a

significant portion of the water molecular space, rendering them unavailable for HOC

dissolution. The log-linear cosolvent model is one of several theoretical approaches for

predicting the organic chemical chemodynamics. While the log-linear model is applicable

over a large range of the volume fraction of cosolvent (f); at the low fc (0-0.3) range,

linear approximation may suffice (Pinal et al., 1990; Hermann and Powers, 1998) and the

relationship can be expressed

S = S w + a ,f (4-1)

where Sc is the solubility in the presence of a cosolvent; Sw is the aqueous solubility and

ac is the cosolvency factor; ac > 0 for cosolvent such as methanol which enhance HOC

solubility; ac < 0 for cosolvent such as chloroform that depress solubility; ac 0 for

cosolvents producing minimal change in solubility.

The solubility of HOCs is inversely related to its partitioning coefficient (Rao et

al., 1985). Therefore, as the volume fraction of cosolvent increases, the partitioning and







retardation of HOC decreases in a log-linear fashion Rao et al. (1985). However, over a

low fc range (fc < 0.3), it follows from Equation (4-2) that,

K = K,, ajf/ (4-2)

where the subscripts n, w, and c represent NAPL, water, and cosolvent respectively; KI

is the partitioning coefficient measured in presence of a cosolvent; Knw is the partitioning

coefficient measured in water; ac is an empirical constant which describes water-

cosolvent interactions.



Materials and Methods

Materials

A suite of tracers was selected to examine solubility and partitioning processes

over a range of retardation factors. Methanol (Fisher, 98 %) was used as a non-reactive

tracer, while 4-methyl-2-pentanol (4M2P) (Acros, 99+ %), n-hexanol (Acros, 98 %), 2-

methyl-3-hexanol (2M3H) (Acros, 98 %), and 2,4-dimethyl-3-pentanol (2,4DMP)

(Acros, 99+ %) were used as partitioning tracers in both batch and column experiments.

Ethanol (Spectrum Chemical, absolute 200 proof), Tertiary-butanol (TBA) (Aldrich, 99+

%), and Isopropanol (IPA) (Fisher, Electronic use) were used as cosolvents.

Tetrachloroethylene (PCE) (Acros, 99 %) was used as a NAPL for all batch and column

experiments. The laboratory temperature during the experiments was 23 1 oC. A 30-40

mesh silica sand (Ottawa) was used as porous medium in all miscible displacement

experiments; alcohol tracers adsorption to this solid matrix was determined to be

negligibly small.

Partition Isotherm Experiments







To assess tracer partitioning in solutions with low cosolvent concentrations (< 10

% by volume), a series of batch equilibrium experiments were conducted. Isotherms for

tracer partitioning into NAPL (PCE) were measured in aqueous/alcohol solution using

three cosolvent solutions: ethanol, TBA, and IPA. The f used in batch equilibrium

experiments were 0, 1, 3, 5, and 10 %. Each cosolvent solution was transferred to 100mL

volumetric flasks, and four alcohol tracers were added with varing concentrations.

Alcohol tracer mixtures prepared at each cosolvent fraction were added to an appropriate

amount of PCE in 25mL vials fitted with Teflon-lined screw caps. The vials were

tumbled end-over-end on a laboratory-rotator (Glas-Col model RD 4512) for 24 hr at

room temperature. At the end of the equilibrium period, alcohol mixtures in supernatant

solutions were analyzed by gas chromatography (Perkin Elmer GC, Autosystem XL).

Solubility Experiments

Tracer solubilities were measured using the cosolvent solutions referred to above.

The prepared solutions (20mL) were transferred into 25mL vials fitted with Teflon-lined

screw caps for each mixture. One alcohol tracer was added to each vial in an amount

more than 3 times the maximum solubility in water. The vials were placed on a rotator

for 24hr at room temperature. At the end of the equilibrium period, the vials were placed

upside-down for at least 6hr to allow the remaining separate phase alcohol to rise and

collect above the water. After equilibration was attained, a 5mL aliquot of liquid was

collected for analysis through the Teflon-lined septum using a glass gas tight 5mL

syringe (Hamilton).

Miscible Displacement Experiments







A series of column tests was conducted with various fractions of a residual

cosolvent (ethanol). The glass column used was 4.8 cm in diameter and 15 cm in length

(high performance liquid chromatography column from Kontes) with Teflon end pieces.

Two layers of fine wire mesh and plastic screen were placed inside of the Teflon end

pieces to allow for an even distribution of fluids and minimize column end effects.

Two different column packing methods were used for miscible displacement

experiments. In the first method, clean sand (30-40 mesh) was wet-packed in thin layers

under continuous vibration. Each layer was stirred and tamped with a rod before adding

new sand. The packed column was cleaned by passing 20 pore volumes of degased DI

water and weighed. Then, NAPL was introduced at a low flow rate (0.5 mL/min) using a

syringe pump. Some of the NAPL was subsequently displaced by injecting water.

Different levels of Sn trapped in the pores was achieved by gradually increased flow

rates (1, 3, 5, and 10 mL/min steps). In the second method, sand was pre-mixed with a

small amount of PCE to create very low residual saturations (Sn = 0.0056). The pre-

mixed sand was prepared by adding approximately 20 % of the saturated water content

(20mL) and the required amount of PCE (2 mL). The sand mixture was shaken and

stirred before and after adding PCE. The mixture was wet-packed as described above.

The initial residual PCE saturations of columns packed by both methods were also

determined using the partitioning tracer test.

The packed columns were oriented vertically and connected to a high

performance liquid chromatography (HPLC) pump (Perkin Elmer series LC 200 and

Gilson model 320) through an inert valve which allowed switching among three

reservoirs for the mobile phases. The first reservoir contained a mobile phase of solute-








free PCE saturated water to minimize the loss of residual PCE in the column, the second

a PCE-saturated cosolvent solution, and the third an alcohol tracer solution. The residual

cosolvent used in the column experiments was an ethanol/water binary mixture. The f.

used in the column experiments were 0, 1, 3, 5, and 10 % by volume.

The miscible displacement experiments were conducted with and without

residual cosolvent present. The column was initially prepared by passing 20 pore

volumes of degased-PCE-saturated water. A column with residual cosolvent was created

by flushing three pore volumes of a degased-PCE-saturated cosolvent solution. Effluent

breakthrough curves (BTCs) were measured under steady water flow condition with a

pulse-input boundary condition. Replicate column tests for each residual cosolvent were

conducted. R and Sn measured during each column test were used to estimate column-

based partitioning coefficients.

Data Analysis

All tracer partition equilibrium data obtained from batch tests with and without

cosolvents were fitted using a linear isotherm to obtain K, and Knw. The Retardation

factor (R) was calculated by moment analysis of effluent BTCs. The NAPL saturation

(S,) was estimated from the K, from batch tests, and the Re value measured in the

presence of cosolvent. The tails of BTCs were monitored until the injected relative tracer

concentration was less than 10-2. The measured data were log-linearly extrapolated up to

103. For comparison with batch test results, Kcol values were computed using Re and

actual Sn values (i.e., Sn estimated using partitioning tracers in the absence of cosolvent)

as follows:








K = (R 1)-(1-S,) (4-3)
S.



where Ko i is a partitioning coefficient estimated from a column test, Re is a retardation

factor obtained in the presence of cosolvent, and Sn is actual NAPL saturation estimated

by tracers without cosolvent. The Kcol was compared to the Knc from batch tests.



Results and Discussion



Effect of Cosolvent on Solubility

The solubilities (Sc) of three alcohols tracers were measured in the presence of

low concentrations (0 to 10 %) of cosolvents (ethanol, TBA, and IPA). The observed

results (Figure 4-1) revealed three types of behavior depending on the cosolvent:

enhancement of solubility (ethanol), reduction of solubility (TBA), and no observable

change (IPA).

The solubility enhancement with ethanol is in good agreement with other studies

(Li et al., 1994; Pinal et al., 1990; Morris et al., 1988; Fu and Luthy, 1986; Yalkowsky

and Roseman, 1986). For instance, Yalkowsky and Roseman (1986) reported that in

completely water-miscible solvents such as methanol, the solubility of HOCs increases

log-linearly. However, the measured results showed a proportionally linear relation. It

seems to be why the observed linear relation was attributed to the low concentration of

cosolvent used.

Solubility reduction, as observed with TBA, was also noted by others (Coyle et

al., 1997; Groves, 1988). Coyle et al. (1997) reported that the solubility of HOCs was







9
SA


E
6
5 I ---------- ---



5
0 2 4 6 8 10 12
Ethanol content (%, volume)

8
87- B
-"----__


5

4 ----
0 2 4 6 8 10 12
Tert-butanol content (%, volume)

9

E8
X x
SX x
X A A
E 6 [ II I I


0 2 4 6 8 10 12
Isopropanol content (%, volume)


Ml-Hexanol A 2-Methyl-3-Hexanol
X 2,4-Dimethyl-3-Pentanol


Figure 4-1. Relationship between tracer solubility (Sc) and cosolvent content
(%, volume) for (A) ethanol, (B) tert-butanol, and (C) isopropanol








20000


16000


E 12000


S8000 *


4000


0
0 10 20 30 40 50 60
Volumetric fraction of TBA (%)

Figure 4-2. Solubility of 2,4 DMP with respect to tert-butanol content (%, volume)


decreased in the presence of PMOS, such as methylene chloride, resulting from an

activity increase of the PMOS in the HOC phase (Coyle et al., 1997). Even though TBA

is a CMOS such as methanol or ethanol, at low concentration the solubility behavior of

the tracer alcohol in the TBA followed that of PMOS rather than the other CMOS.

Additionally, Figure 4-2 shows that the solubility of the alcohol tracer, 2,4 DMP, in TBA

solutions decreased linearly until about 25 % TBA, at which point, solubility increased

sharply with increasing concentration (> 25 %). The reason for the solubility reduction at

low concentration is not clear, however, one possible explanation is an activity increase

of TBA-cosolvent in the tracer-alcohol phase. Above 25 %, however, the TBA began to

partition into the alcohol, swelling it, and caused significant increases in solubility.

For the IPA-cosolvent system, the solubilities did not show any observable trend,

with increasing volume fraction of IPA (Figure 4-1C). This was different from results







30 A

20

10

0
0 .------------

0 2 4 6 8 10 12
Ethanol content (%, volume)

30

20

10

0
0 2 4 6 8 10 12
Tert-butanol content (%, volume)


30 x

20

10



0 2 4 6 8 10 12
Isopropanol content (%, volume)



4-methyl-2-pentanol U n-hexanol
A 2-methyl-3-hexanol X 2,4-dimethyl-3-pentanol


Figure 4-3. Relatioship between tracer partition coefficient (Kn,) and cosolvent content
(%, volume) for (A) ethanol, (B) tert-butanol, and (C) isopropanol







reported in the literature. Morris et al. (1988) reported that the solubility of naphthalene

increased log-linearly with increasing the fraction of IPA. Pinal et al.(1990) and

Valkenburg (1999) also showed that IPA at high level (range 0-80 %) is a strong

cosolvent in experiments with TCE and PCE. However, experiments in this study were

conducted at low concentrations (range 0-10 %). Thus, it can be concluded that the IPA

as a cosolvent has little affect on the solubility of the tracer-alcohol at low concentrations

(< 10 %).

Effect of Cosolvent on Tracer Partitioning Isotherms

The partition coefficients, K,,, determined from the batch isotherm data are

shown in Figure 4-3. Based on the observed Knc, three types of partitioning behavior

were also found in accordance with the property of each cosolvent added in the aqueous

phase.

In the presence of the ethanol, the measured Kn. values decreased for all alcohol

tracers used in this study, as the volume fraction of ethanol (fc,EtOH) increased (Figure 4-

3A). The Knc and fc,EoH values were well correlated in inverse-linear relationships with r2

2 0.95. The Knc values decreased up to approximately -15, 21, 18, and 18 % of each K,,

value measured in water, for 4M2P, n-hexanol, 2M3H, and 2,4DMP, respectively. Recall

that the solubilities of the tracer-alcohols increased in the presence of fc,EtOH (Figure 4-1).

This solubility increase supports well the decrease in observed partitioning since

they are inversely related to each other. Such results indicate that the residual ethanol-

cosolvent can affect estimations of S. using Equation 4-3. The S. has a functional

relationship between K.n and R. The R, which is a measure of partitioning tracer

mobility, becomes smaller or larger according to a degree of partitioning of the tracer into







NAPL. Reduced partitioning produces less retardation. In the presence of cosolvent such

as ethanol, if the Kw value determined in water is used, the S. would be under-estimated.

In the presence of the TBA-cosolvent, the Knc values increased with increasing

fraction of TBA (fc,TBA) (Figure 4-3B). The correlated results showed a linear relationship

between the Knc and fe,TBA (r2 = 0.86-0.92). The K.c values increased compared K.w up to

-35, 49, 24, and 16 % for 4M2P, n-hexanol, 2M3H, and 2,4DMP, respectively. This

partitioning increase has not been reported in the literature. However, we could expect a

partitioning increase based on the observed solubility reduction. The results may be

explained as the solventing-out effect referred by Coyle et al. (1997). The dissolved TBA

causes an increased partitioning of alcohol tracers by occupying the water molecular

space. In other words, an increase of cosolvent such as TBA, unlike the ethanol-

cosolvent, results in a decreasing activity, producing increased partitioning. This result

suggests that failing to consider the effect of residual cosolvent could result in

overestimation of Sn.

In the presence of isopropanol as a cosolvent, the partitioning coefficients did not

display an apparent trend (Figure 4-3C). This is consistent with the minimal affect on the

solubilities of the tracer-alcohols (Figure 4-1). This suggests that residual IPA-cosolvent

may not affect the tracer partitioning behavior or the estimate of S. at low concentration

(< 10 %).

Based on above observed results, at low concentration (5 10 %) the relationships

between Kc and fc can be divided into three types according to the properties of

cosolvents used. The equation to describe the relationships is as follows:








Table 4-1. Comparison of estimated a, and be values from tracer solubility and partition
experiments for four tracers

Ethanol TBA IPA

Tracers solubility" partition solubility" partition solubility" partition

4M2P nd -0.06(0.93) nd +0.11(0.88) nd -0.01(0.00)

n-hexaol +119(0.96) -0.14(0.99) -96(0.98) +0.27(0.84) +19(0.53) -0.02(0.16)

2M3H +119(0.99) -0.43(0.99) -83(0.96) +0.47(0.86) +34(0.41) -0.13(0.61)

2,4DMP +124(0.83) -0.50(0.95) -93(0.96) +0.36(0.83) +36(0.06) -0.17(0.38)


a ac values estimated by linear regression on solubility data for tracers-low cosolvents
systems
be values estimated by linear regression on partitioning coefficient data for tracers-low
cosolvents systems


Knc = KIw + be fc


be < 0 for ethanol
be > 0 forTBA
b e 0 for IPA

where, Knc is the partitioning coefficient measured in presence of a cosolvent, Knw is the

partitioning coefficient measured in water, be is an empirical constant which describes

water-cosolvent interactions with the subscript c designating cosolvent. The estimated be

values along with a were shown in Table 4-1.

Effect of Cosolvent on Tracer Retardation

Column miscible displacement tests were conducted to investigate the effect of

residual cosolvent on tracer partitioning and transport behavior. The displacement fluid














Table 4-2. Parameter values observed from miscible displacement experiments at low
volume fractions of ethanol

Tracers Ethanol, fc Rc Sn Sn errora Keolb Kc

4-methyl-2-pentanol 0 1.84 0.188 3.60 3.60
0.01 1.82 0.185 -1.54 3.53 3.50
0.03 1.80 0.182 -3.14 3.46 3.37
0.05 1.78 0.178 -5.44 3.36 3.23
0.10 1.74 0.170 -9.41 3.19 3.07

n-hexanol 0 2.47 0.184 6.56 6.56
0.01 2.47 0.183 -0.75 6.50 6.42
0.03 2.45 0.181 -2.00 6.40 6.13
0.05 2.41 0.176 -4.27 6.22 5.74
0.10 2.36 0.172 -6.82 6.02 5.18

2-methyl-3-hexanol 0 6.28 0.189 22.7 22.7
0.01 6.15 0.185 -1.98 22.1 22.3
0.03 6.09 0.183 -3.03 21.8 21.1
0.05 5.96 0.180 -5.03 21.3 20.4
0.10 5.87 0.177 -6.47 20.9 18.5

2,4 dimethyl-3-pentanol 0 7.13 0.189 26.3 26.3
0.01 7.10 0.189 -0.40 26.1 25.9
0.03 7.02 0.187 -1.43 25.8 23.9
0.05 6.86 0.183 -3.55 25.1 23.7
0.10 6.71 0.179 -5.57 24.5 21.5


a calculated by ((Sn measured with each f,) (S, with 0 % f )) / (S, with 0 % fc) x100
b Partitioning coefficients estimated from column tests in the presence of ethanol-
cosolvent
C Partitioning coefficients measured from batch tests in the presence of ethanol-cosolvent











4-methyl- 2-pentanol


y= -0.0402x+ 3.58
r= 0.99



y=-0.052x+ 3.55
r=0.96


0 2 4 6 8
Ethanol content (%, volume)


2-methyl-3-hexanol


0 2 4 6 8 10
Ethanol content (%, volume)


7.0

6.5

6.0

5.5

5.0


n-hexanol

y = -0.0556x + 6.55
r 2 =0.98



y=-0.14x+6.54
r2 =0.99


3.8

3.6 i

S3.4

3.2

3.0


2,4-dimethyl-3-pentanol


y=-0.186x+ 26.26
S= 0.97




y = -0.477x + 26.09
r2 = 0.96


0 2 4 6 8
Ethanol content (%, volume)


Figure 4-4. Comparison of the K. Values from batch tests to the KI estimated from column tests for 4-methyl-2-pentanol,
n-hexanol, 2-methyl-3-hexanol, and 2,4-dimethyl-3-pentanol in the ethanol/water system; Column test; m Batch
test.


0 2 4 6 8
Ethanol content (%, volume)







used is a binary ethanol/water solution, with fc ranging from 0 to 10 %. The results are

provided in Table 4-2. As expected, an increasing fc resulted in earlier breakthrough of

the tracers. Therefore, as described earlier using the Knw values measured in the absence

of cosolvent, lower retardation will result in lower S, estimation.

The observed Sn values at low fc were 1 to 10 % lower than the actual Sn,

measured using tracer in the absence of ethanol(Table 4-2). While the error is fairly

minor, its effect should not be neglected. The cosolvent effect can be larger depending on

the magnitude of residual Sn, as discussed later.

Partitioning coefficients based on column tests (Kcol) were calculated to compare

to the batch-measured Knc. The Kcol values (Table 4-2) were computed using the Re

values measured with fc,EtOH and the actual Sn into Equation 4-4. Regressing Kcol against

fc,EIOH yielded high coefficients of determination with 0.99, 0.98, 0.90, and 0.97,

respectively, for 4M2P, n-hexanol, 2M3H, and 2,4DMP. Figure 4-4 shows inverse-linear

relations in good agreement with the batch results. The Kcol values, however, were

consistently higher than the Knc values from batch tests. The slopes of the correlated

curves for column results were lower than batch tests by the factors of about 0.2, 0.6, 0.6,

and 0.6, respectively, for 4M2P, n-hexanol, 2M3H, 2,4DMP. The deviations between

column and batch results are a result of two processes: cosolvent dilution and the

difference in solute travel time. First, for deviation due to cosolvent dilution, recall that

the column miscible displacement tests were initiated with one pore volume of the

residual ethanol cosolvent without any external ethanol source. The displaced ethanol

solution was diluted with the injected alcohol tracer pulse and subsequent mobile phase,

PCE saturated water, which were cosolvent-free. This dilution can impact the partitioning










Low Sn (PCE): 0.0056


0 0.5 1 1.5
Pore Volume


High Sn (PCE): 0.15


0.001


0 2 4 6
Pore Volume


B 8-

6

4
x
xx 2 I
uJ
1 0
8 10


Methanol 1-Hexanol
x 2,4-Dimethyl-3-pentanol Resident Ethanol


Figure 4-5. Breakthrough curves of residual ethanol cosolvent and partitioning tracers
(n-hexanol and 2,4-dimethyl-3-pentanol) to display a degree of mixing,
explaining the effect of residual ethanol on the NAPL saturation estimation.
Initial residual ethanol content in the columns is 10 % (volume).


0.01


0.001


A
di AX
m ^x
Ix mX
x

J. U


10



6

4

2
U0
0


2 2.5







process by reducing local ethanol concentration. Second, deviation is caused by solute

travel time of the partitioning tracers compared to the displaced ethanol front, which is

essentially non-partitioning. At the tracer front, interaction with the residual cosolvent is

significant. However, as the experiment proceeds, the partitioning tracers transport

through the column is long, while the residual ethanol is rapidly displaced. Because of

this, the influence of the residual cosolvent is reduced. Therefore, in the presence of a

resident cosolvent without any additional source, the estimated Keol values from column

tests become higher than those from batch tests.

Impact of NAPL Saturation in the Systems

As noted above, in the presence of residual cosolvent, the difference in travel

times between the partitioning tracers and the cosolvent front can be a factor for

estimating the residual Sn using partitioning tracers. We hypothesize that the effect of a

cosolvent will be different based on the amount of Sn present in the column during a post-

flushing tracer test. To verify this, column experiments were performed with low and

high NAPL (PCE) saturation in the presence of ethanol-cosolvent (10 % by volume).

Effluent ethanol from the columns was monitored along with the partitioning

tracers, n-hexanol and 2,4DMP. The BTCs are given in Figure 4-5A and 4-5B for the low

and high NAPL saturations. The BTCs showed that the degree of mixing between

residual ethanol cosolvent and the injected tracers was different between the low and high

saturation columns. The BTCs of methanol and ethanol-cosolvent, which are non-

partitioning solutes, followed similar trajectory for both the low and high saturation,

whereas those of n-hexanol and 2,4DMP were very different for both cases as expected.








Table 4-3. Comparison of parameter values observed from low and high NAPL (PCE)
saturation column tests using 2,4-dimethyl-3-pentanol as a partitioning tracer

Low saturation column High saturation column

Cosolvent (%) 0%Ethanol 10%Ethanol 0% Ethanol 10% Ethanol


Retardation factor (Re) 1.15 1.12 5.78 5.51
Partitioning coefficient, Knc 26.38 21.7b (21.5)a 26.3" 24.8b(21.5)a
PCE saturation, Snc 0.0056 0.0046 0.154 0.147

Sn % error -17.4 -4.90

a Knc values measured by batch tests with the 0 % and 10 % ethanol-cosolvent.
b K o values estimated by column tests with residual 10 % ethanol; calculated using
Equation (4-3).
c calculated using Eq. (4-3); using the Re value obtained from each column test and the
KIw (26.25) from the batch test without cosolvent.


For the low Sn in Figure 4-5A, the portion of ethanol-cosolvent BTC overlapped

with the frontal over about 60 % of the partitioning tracer BTCs (n-hexanol and

2,4DMP), indicating that over about 60 % of the n-hexanol and 2,4DMP traveled in the

presence of the ethanol-cosolvent through the column. On the other hand, for the high Sn

in Figure 4-5B, the overlapped portion between ethanol and the partitioning tracer BTCs

was minimal. This suggests that the affect of residual cosolvent should be more

significant in the low Sn column than in the high. As seen in Table 4-3, the Sn estimated

from the low saturation column in the presence of ethanol-cosolvent (10 %) was about 17

% less for 2,4DMP than the actual Sn estimated without ethanol-cosolvent. In the high

PCE saturation column, the measured Sn in the presence of 10 % ethanol was about 5 %

less for 2,4DMP than actual.

The Kco values were computed to evaluate the relative effect of residual ethanol-

cosolvent for both low and high Sn in the system. The Kco. values (= 21.7 for 2,4DMP)








estimated from the low S, column was very close to the Kn values (21.5) from the batch

test measured in the presence of same 10 % ethanol solution. However, the Kco, values (=

24.8 for 2,4DMP) estimated from the high saturation column revealed more deviation

than the Knc values (21.5). This suggests that the S, may determine the magnitude of the

cosolvent effect. Typically, post-flushing tracer tests have been conducted with less than

one percent of S, in the field (Annable et al., 1998 a, b). Such a low residual Sn can result

in greater cosolvent effect on Sn estimation using partitioning tracers. This suggests that

in such low S,, as a post-flushing tracer test is conducted, it is important to consider the

effect of residual cosolvent.



Conclusions

Based on the results of this study, residual cosolvent can impact the estimate of

NAPL (PCE) saturation using partitioning tracers. Batch test results indicated three

different effects on NAPL estimation dependent on the properties of cosolvents at the low

concentration (5 10 %, vol.) used in this study: under-estimation, over-estimation, and no

observable effect. Increasing the fraction of ethanol cosolvent resulted in a linear

decrease in tracer alcohol partition and an increase in solubility, suggesting an under-

estimation of NAPL saturation. In contrast to ethanol-cosolvent, increasing the fraction of

tert-butanol cosolvent results in a linear decrease in tracer alcohol partition and decrease

in its solubility, suggesting an over-estimation of NAPL saturation. In column tests, the

observed Sn values in the presence of ethanol-cosolvent were under-estimated by about 1-

10% with high NAPL Sn (0.15-0.18), whereas by about 17% with low NAPL Sn (0.0056).

The implication from this study is that the effect of cosolvent may be significant on





78

estimation of NAPL saturation using a post-flushing tracer test, which is conducted in

very low NAPL saturation. Consequently, a better understanding of the partition and

transport of chemicals according to cosolvent types at the low level would provide

valuable information for a better estimate of NAPL saturation.













CHAPTER 5
INFLUENCE OF RESIDUAL SURFACTANT ON PARTITIONING TRACERS

Introduction


Surfactants as micelle-forming surface active chemicals have been used as an

enhanced subsurface flushing technology, principally in the oil industry or at organic

liquid contaminated sites. The surfactants are classified, by the nature of their hydrophilic

head groups as anionics, nonionics, and cationics. The anionic and nonionic surfactants

have been good candidates as flushing agents since soil particles generally carry a

negative charge. Their effectiveness at cleaning up contaminated sites has been

demonstrated in previous studies (Abdul et al., 1990; Boving et al., 2000; Dwarakanath et

al., 1999; Jawitz et al., 1998; Pennell et al., 1993; Rhue et al., 1999; Shiau et al., 1994).

After remedial efforts and following a water flood, however, some amount of flushing

agent is likely to remain adsorbed on the solid phase, or dissolved (as monomer or

micelle) in the aqueous phase. The residual surfactants, which alter the partitioning

properties of hydrophobic organic chemicals (HOCs) in the flushed zone, must be

considered when the partitioning tracer technique (Pope et al., 1994; Jin et al., 1995;

Annable et al., 1995) is used to characterize residual NAPL saturation.

The adsorption of surfactant at the solid-liquid or liquid-liquid interface is

influenced by many factors: the nature of adsorbent structure (i.e., highly charged sites

and polar or nonpolar surface), the surfactant structure (i.e., ionic or nonionic and short or








long-chain), the aqueous solution condition (i.e., electrolyte content, pH, and

temperature, etc) (Rosen, 1989; Abu-Zreig, 1999).

Surfactant adsorption from aqueous solution onto solid matrix

The adsorption of surfactant onto the solid phase is related to its orientation,

concentration, and the energy change at interfaces. In general, the surface of natural soil

is negatively charged, but it often has positively charged sites which contains oxides such

as alumina. The adsorption occurs mainly by ion exchange, ion pairing, and hydrogen

bonding. The adsorption isotherm of ionic surfactant onto an oppositely charged sites is

typically S-shaped and reflects three or four distinct regions (Rosen, 1989; Holsen, 1991;

and Ko, 1998a) (Figure 5-1). In region 1, the ionic surfactant sorption is low due to

electrostatic repulsion and occurs mainly by ion exchange and pairing. Nonionic

surfactant adsorbs mainly onto the solid surface by hydrogen bonding. The adsorbed

surfactant molecules spread themselves on the surface, forming a monolayer. The charge

density or potential at the stem layer of the surface remains constant. In region 2, there is

a sharp rise in adsorption, resulting from lateral interaction of the hydrophobic moieties

of oncoming surfactant with those of the previously adsorbed surfactant. In region 3, the

slope of the sorption isotherm is reduced and then plateaued, indicating complete surface

coverage with a monolayer or bilayer of surfactant. This occurs mainly near the critical

micelle concentration (CMC). When the concentration is above the CMC, any surfactant

added to the solution will not increase significantly the sorption onto the solid phase and

the number of monomers in solution, but rather contributes to the formation of additional

micelles. The surfactant adsorption is also influenced by the pH and temperature of the

solution. As the pH of the aqueous phase decreases, the adsorption of anionic and
























4
33


0 2








Log Equil. Bulk Cone. of Surfactant

Figure 5-1. S-shaped adsorption isotherm for an ionic surfactant on an oppositely
charged substrate (adapted from Rosen, 1989)








nonionic surfactants increases while that of the cationic surfactant decreases. Also, the

solid phase becomes more positive, resulting from adsorption onto charged sites of

proton from solution. Ko et al. (1998b) and Holsen et al. (1991) showed that the sorption

of anionic surfactant (SDS) and nonionic surfactant (Tween 80) onto kaolinite increased

with decreasing solution pH. Regarding temperature, as the temperature of the solution

increases, even slightly, the sorption of ionic surfactant increases while that of nonionic

surfactant decreases.

Surfactant adsorption from aqueous solution onto hydrophobic organic chemicals

In a system containing hydrophobic adsorbents, surfactants mainly adsorb onto

the surface by dispersion forces (Rosen, 1989). The adsorption isotherms are generally of

the Langmuir type. The adsorbed surfactant tail groups initially are oriented parallel to

the hydrophobic surface of HOC with a slightly tilted or L-shape and the hydrophilic

head groups are oriented toward the aqueous phase. As adsorption of the surfactant

continues, the orientation of the adsorbed molecules is more perpendicular to the surface.

The aggregation of surfactant molecules begins to form a micelle which has a

hydrophobic interior where hydrophobic organic molecules can reside. John et al. (2000)

reported that a low aqueous surfactant concentration, the sorption coefficient of the

surfactant onto NAPL increases dramatically, but decreases with increasing surfactant

concentration above the critical micelle concentration (CMC). The surfactant adsorption

appears to level off near/above the CMC and this leveling off trend is attributed to the

formation of micelles which limit the amount of surfactant adsorption onto the NAPL

(Zimmerman, 1999; John et al., 2000).








The micelle formation significantly increases the aqueous solubility of organic

chemicals. This micellar solubilization is the process used for surfactant flooding to clean

up organic contaminants. In addition, an increase in the length of the hydrophobic group,

and neutral electrolyte addition which decreases the electrical repulsion between the

similarly charged ions and oncoming ions increase the efficiency and effectiveness of

adsorption. The adsorption rate on an HOC is faster with surfactants containing the

hydrophilic group in a central position because of greater a diffusion coefficient and

greater CMC (Rosen, 1989).

Partitioning of hydrophobic organic chemical into residual surfactants

In a soil/water system, while anionic surfactants generally repel the negatively

charged soil, nonionic surfactants adsorb onto the solid surface by hydrogen bonding. On

the other hand, if the soil contains positively charged sites such as metal oxides (i.e.,

aluminum oxide), the anionic surfactants can adsorb strongly onto the charged sites by

ionic exchange or pairing (Holsen et al., 1991; Smith et al., 1991; Rouse et al., 1993; Sun

and Jaffe, 1996). The adsorbed surfactants act as good hydrophobic sites to allow

partitioning of HOCs. Like the NAPL system, as the concentration of surfactants exceeds

the CMC, adsorption levels off and excess surfactants form micelles which have a

hydrophobic interior where hydrophobic organic molecules can reside (Holsen et al.,

1991; Ko et al., 1998a,b; Sun and Jaffe, 1996).

In the aqueous phase, HOC also partition into surfactant micelles or mononers,

causing a decrease in retardation. However, the retardation decrease is typically much

smaller than an increase due to partitioning into the adsorbed surfactant. Sun and Jaffe

(1996) reported that the partitioning of phenanthrene between the adsorbed dianionic








surfactant (DowFax) phases (monolayers and bilayers) and water was 5 to 7 times more

effective than that between the corresponding surfactants phases monomerss and

micelles) in the aqueous phase and water. This means that the adsorbed surfactant in

partitioning competition dominates the corresponding surfactant micelles and monomers.

Residual surfactant as a result of remedial activities can be a concern when

applying partitioning tracers (Rao et al., 1997; Annable et al., 1998a). The partitioning

behavior of tracers travelling through the porous media can be impacted by surfactants:

1) partitioning into the NAPL phase resulting in an increase in tracer retardation, 2)

forming micelles in the aqueous phase causing a decrease in tracer retardation, and 3)

adsorbing on the solid matrix resulting in increased tracer retardation.

As a result, the residual surfactants can give an erroneous NAPL indication and

limit the use of partitioning tracers. This study investigates the influence of residual

flushing agent (surfactants) upon residual NAPL volume estimation using partitioning

tracers. The objectives were to quantify the effect of residual surfactants on the

partitioning and transport behavior of alcohol tracers through batch isotherm and column

miscible tests. The effects of adsorbed surfactants on the solid matrix, including

positively charged sites, were also investigated. Finally, the observed effects of residual

surfactants were verified by a post-flushing column tracer test following a surfactant

flood.



Materials and Methods


Materials
Diphenyl oxide disulfonates (DowFax 8390) (Dow Chemical Co.), Sodium

dihexyl sulfosuccinate (AMA-80-I) (Cytec Industries Inc.), and Polyoxyethylene (10)








oleyl ether (Brij-97) (Uniqema Industries Inc.) were the surfactants used in this study.

Their physical and chemical properties are given in Table 5-1. DowFax-8390 is a di-

anionic surfactant with two negatively charged hydrophilic sulfonate heads and an alkyl

chain. AMA-80-I is a mono-anionic surfactant in a mixture of isopropanol and water, and

Brij-97 is a nonionic surfactant. Tetrachloroethylene (PCE) (Acros, 99 %) was used as a

NAPL for all batch equilibrium and column experiments. A suite of tracers was selected

to examine the effects of varying retardation factors. Methanol (Fisher, 98 %) was used

as a non-partitioning tracer while 4-methyl-2-pentanol (4M2P) (Acros, 99+ %), n-

hexanol (Acros, 98 %), and 2,4-dimethyl-3-pentanol (2,4DMP) (Acros, 99+ %) were

used as partitioning tracers. All experiments were conducted at room temperature (23 1

o C). A clean silica sand (30-40 mesh, Ottawa) and a clay-silt-sandy loamy soil (Dover

AFB site soil) including some metal oxides were used as porous media throughout the

miscible displacement experiments. The metal oxides in the soil were extracted by an

acid digestion method for soil (U.S. EPA SW-846 3050B) and quantified by an inductive

coupled plasma atomic emission spectrometry (Thermo Elemental series Enviro 36).

Batch Isotherm Tests

A series of batch equilibrium experiments was conducted to evaluate the behavior

of tracer partitioning in varying surfactant concentrations (_ 0.5 % by weight). The

partition isotherms were measured for three surfactant solutions (Brij 97, AMA 80, and

DowFax 8390) with varying concentration (0.0, 0.05, 0.1, 0.3, and 0.5 %). Each

surfactant solution was transferred to a 100 mL volumetric flask, and three alcohol tracers

were added. The tracer mixtures were combined with PCE in 25 mL vials fitted with

Teflon-lined screw caps. The vials were tumbled end-over-end on a Laboratory-rotator








Table 5-1. Physical and chemical properties of surfactants used in the study.


Surfactants Mono-anionic Di-anionic Non-ionic

Chemical name Dihexyl Diphenyl Oxide Polyoxyethylene (10)
Sulfosuccinate Disulfonates Oleyl Ether
Trade name Aerosol MA 80-1 DowFAX 8390 Brij 97
Chemical Structure C16H2907NaS C16H33 CI2 H70(SO3Na) C58H116021
Mol wt. 388 642 1149
Water Solubility Miscible Miscible Miscible
CMC (mM) 2.3a 3b
HLB 78.6c 12.4d

Notes : CMC is critical micelle concentration; mM is milliMolar.
a Lowe et al., 1999; b Rouse et al., 1993; 'Dow Chemical Co.; d Zou and Rhue, 1999.


(model RD 4512) for 24hr at room temperature. At the end of the equilibrium period, the

alcohol mixtures in the supernatant solutions were analyzed by Gas Chromatography

(Perkin Elmer GC, Autosystem XL) with a flame-ionization detector.

Miscible Displacement Tests

Three series of miscible displacement tests were performed for this study. The

schematic diagram of the experimental set-up is shown in Figure 5-2. The first series of

tests was conducted to evaluate the effect of residual surfactants in the aqueous phase. A

glass column with Teflon end pieces was used with a diameter of 4.8 cm and a length of

15 cm (high performance liquid chromatography (HPLC) column from Kontes). Two

layers of a fine wire mesh and a plastic screen were placed inside of the Teflon end piece

to minimize column end effects. The column was packed with pre-mixed sand containing

PCE. The pre-mixed sand was prepared by adding 20 mL of water and 2 mL of PCE to

750 g (- 300 cm3) of sand and homogenizing. The columns used for the residual

surfactant (DowFax 8390) tests were prepared for the miscible displacement experiments














M
15cm






.G H




5cm K



2.5cm


A : Tracer Reservoir
B :PCE Saturated Water Reservoir
C :Cosolvent or Surfactant Solution Reservoir
D : HPLC Pump
E : Switching Valve
F : Waste
G: Sand/PCE Column
H Sand /PCE/ Cosolvent or Surfactant Column
I : Dover Soil Column
J : Dover Soil/PCE Column
K : Dover Soil/ Surfactant Column
L : Dover Soil/ Surfactant/ PCE Column
M : Sampling Vial


A B C


Figure 5-2. Schematic diagram of the experimental set-up








by flushing 3 pore volumes (PV) of DowFax solution through the column. The mobile

phase and DowFax solution used PCE saturated water to minimize loss of residual PCE

in the column. The partitioning tracer tests were conducted both with and without

DowFax solution at varying concentrations (0.05, 0.1, 0.3, and 0.5 % by weight). The

retardation factor and PCE saturation measured from each column test were used to

estimate column based partitioning coefficients (Kcl).

The second series of miscible displacement tests was conducted to assess residual

surfactant adsorbed on the solid matrix without NAPL. The column used was a glass

column, 2.54 cm in diameter and 5 cm in length (HPLC column, Kontes) with Teflon end

pieces and a fine wire mesh and a plastic screen. Silt-sandy loamy soil (Dover AFB site

soil) was wet-packed and cleaned by passing 20 PV of degased DI water. The silt-sandy

loamy soil contained some metal oxides which can exchange with a negative charged

surfactant (i.e., DowFax 8390). Brij 97 (5 %), AMA 80(5 %), and DowFax 8390 (5 %)

solutions were used as surfactants. The column tests for adsorbed surfactants were

prepared by flushing 5 PV of each surfactant solution, followed by 5 PV of water through

the column. Partitioning tracer tests were conducted in the columns containing adsorbed

surfactant and the results were used to quantify false indications of PCE.

The third series of miscible displacement tests was conducted to demonstrate the

impact of the adsorbed surfactant using a surfactant flushing test. The column, media, and

packing method used were the same as those described for the second series of

experiments. Column test 1 was conducted using a surfactant mixture (5 % DowFax

8390/ 3 % AMA 80/ 3 % NaCI / 3 % CaC12, wt.) as the flushing agent, but without the

presence of PCE. The column was prepared by flushing 5 PV of the surfactant mixture,