ARTICLE IN PRESS

Available at www.sciencedirect.com

WAT E R R E S E A R C H 4 1 ( 2 0 0 7 ) 2 1 8 0 – 2 1 8 8

0043-1354/$ - see frodoi:10.1016/j.watres

�Corresponding autE-mail address: s

journal homepage: www.elsevier.com/locate/watres

Sorption and transport of acetaminophen, 17a-ethynylestradiol, nalidixic acid with low organic contentaquifer sand

Oranuj Lorphensria, David A. Sabatinib,�, Tohren C.G. Kibbeyb,Khemarath Osathaphanc, Chintana Saiwand

aNational Research Center for Environmental and Hazardous Waste Management, Chulalongkorn University, Bangkok, ThailandbSchool of Civil Engineering and Environmental Science, The University of Oklahoma, Carson Engineering Center, 202 West Boyd, Rm. 334,

Norman, OK 73019-1024, USAcDepartment of Environmental Engineering, Chulalongkorn University, Bangkok, ThailanddPetroleum and Petrochemical College, Chulalongkorn University, Bangkok, Thailand

a r t i c l e i n f o

Article history:

Received 23 July 2006

Received in revised form

20 January 2007

Accepted 23 January 2007

Available online 30 March 2007

Keywords:

Sorption

Column study

Acetaminophen

17a-ethynyl estradiol

Nalidixic Acid

Aquifer sand

Pharmaceuticals

nt matter & 2007 Elsevie.2007.01.057

hor. Tel.: +1 405 325 4273;abatini@ou.edu (D.A. Sa

a b s t r a c t

The sorption and transport of three pharmaceutical compounds (acetaminophen, an

analgesic; nalidixic acid, an antibiotic; and 17a-ethynyl estradiol, a synthetic hormone)

were examined by batch sorption experiments and solute displacement in columns of

silica, alumina, and low organic carbon aquifer sand at neutral pH. Silica and alumina were

used to represent negatively-charged and positively-charged fractions of subsurface media.

Column transport experiments were also conducted at pH values of 4.3, 6.2, and 8.2 for the

ionizable nalidixic acid. The computer program UFBTC was used to fit the breakthrough

data under equilibrium and nonequilibrium conditions with linear/nonlinear sorption.

Good agreement was observed between the retardation factors derived from column model

studies and estimated from equilibrium batch sorption studies. The sorption and transport

of nalidixic acid was observed to be highly pH dependent, especially when the pH was near

the pKa of nalidixic acid (5.95). Thus, near a compound’s pKa it is especially important that

the batch studies be performed at the same pH as the column experiment. While for ionic

pharmaceuticals, ion exchange to oppositely-charged surfaces, appears to be the dominant

adsorption mechanism, for neutral pharmaceuticals (i.e., acetaminophen, 17a-ethynyl

estradiol) the sorption correlated well with the Kow of the pharmaceuticals, suggesting

hydrophobically motivated sorption as the dominant mechanism.

& 2007 Elsevier Ltd. All rights reserved.

1. Introduction

Recently a number of pharmaceuticals have been detected in

rivers, lakes and groundwater (Alder et al., 2000; Daughton

and Ternes, 1999; Golet et al., 2002; Halling-SØrensen et al.,

1998; Heberer, 2002; Ternes et al., 2002). Classes of pharma-

ceutical compounds commonly detected in the environment

r Ltd. All rights reserved.

fax: +1 405 325 4217.batini).

include antibiotics, endocrine disrupters and other nonpre-

scriptive analgesic drugs.

Sorption is a major process affecting the fate and transport

of chemicals in groundwater. As discussed in our previous

paper (Lorphensri et al., 2006), important media property (e.g.,

PZC) and pharmaceutical properties (e.g., pKa, Kow) affect

sorption. In short, if the pH, pKa, and PZC are such that the

ARTICLE IN PRESS

WAT E R R E S E A R C H 41 (2007) 2180– 2188 2181

pharmaceutical and medium are oppositely charged then

adsorption by ion exchange can be expected as a dominant

mechanism; if the pH and pKa are such that the pharmaceu-

tical is neutral then sorption by hydrophobic partitioning to

the soil organic content can be anticipated.

While batch experiments are often used to quantify

sorption, certain phenomena are best elucidated in column

studies. For example, the column experiment has almost the

same solid to solution ratio as observed in aquifer systems,

the influent solution can be adjusted and maintained at

desired conditions, rather than changing over the course of a

batch experiment, and the potential for nonequilibrium

sorption can be evaluated.

In previous research we evaluated the sorption of select

pharmaceuticals on a range of media surfaces using

batch studies (Lorphensri et al., 2006). The objective of this

research is to extend our previous work by evaluating

select pharmaceuticals in continuous flow column studies,

thereby allowing us to confirm sorption processes identified

in the batch studies. In addition, this work extends our

previous research, which focused on pure mineral surfaces

and an organic medium, by studying adsorption on an

aquifer sand to evaluate adsorption processes for a hetero-

geneous natural medium. Moreover, to better understand

the role of adsorbent surface charge and pharmaceutical

ionization/protonation, column studies were conducted for

select pharmaceuticals for a range of pH values (ranging from

4 to 9), as was previously done in batch studies (Lorphensri

et al., 2006). Finally, this work provides a preliminary

assessment of the kinetics of sorption in column systems

(i.e., whether the transport exhibits equilibrium or nonideal

conditions).

1.1. Contaminant transport with sorption

In the absence of transformation processes, subsurface

contaminant transport is described by advection, dispersion

and adsorption processes according to the following govern-

ing equation (assuming equilibrium sorption) (Freeze and

Cherry, 1979):

RqCqt¼ D

q2Cqx2� n

qCqx

, (1)

where R [dimensionless] is the retardation factor; for linear

sorption isotherm R ¼ 1þ ðr=yÞKd and for Freundlich non-

linear sorption isotherm R ¼ 1þ ðr=yÞKfNCN�1; C [M/L3] is the

solute concentration; t [T] is time; D [L2/T] is the hydro-

dynamic dispersion coefficient; v [L/T] is the average pore

water velocity; r [M/L3] is the bulk density; y is the effective

porosity; Kd is the sorption coefficient; Kf is the Freundlich

sorption coefficient; and N is the Freundlich nonlinearity

parameter.

Nonequilibrium sorption may result from chemical none-

quilibrium or from rate-limited diffusive mass transfer

(e.g., film diffusion, retarded intraparticle diffusion, and

intra sorbent diffusion) as discussed in Grathwohl (1998),

van Genuchten (1981) and van Genuchten and Wagenet

(1989).

The following dimensionless equations represent the

transport of sorbing solutes under one-dimensional, steady

water flow in homogeneous porous medium (Brusseau, 1991)

bRqC�

qTþ ð1� bÞR

qS�

qT¼

1Pq2C�

q2X�

qC�

qX, (2)

ð1� bÞRqS�

qT¼ oðC� � S�Þ. (3)

In Eqs. (2) and (3), C�½¼ C=Co� is dimensionless aqueous

concentration; S� ¼ ½S2=ðð1� FÞKdÞ is the dimensionless sorbed

phase concentration; S2 [MM�1] is the sorbed-phase concen-

tration in the rate-limited domain; T½¼ vt=L� and X½¼ x=L� are

dimensionless relative pore volume and length, respectively;

P ½¼ vL=D� is the Peclet number which is a ratio of advective

flux versus dispersive flux; R½¼ 1þ rKd=y� is the retardation

factor; b½¼ ðyþ FKdÞ=ðyþ rKdÞ� is the dimensionless parameter

related to fraction of instantaneous retardation; o ½¼ ðk2ð1�

bÞRLÞ=v� is the Damkohler number, which is a ratio of

hydrodynamic residence time to characteristic time for

sorption; b and o specify the degree of nonequilibrium in

the system, which decreases as either of the two parameters

increase in magnitude; k2 [T�1] is first-order desorption rate

constant, v [LT�1] is the average pore water velocity; L [L] is the

column length; F [dimensionless] is the fraction of sorbent for

which sorption is instantaneous; x [L] is the distance, and

other parameters are as described above. Eqs. (2) and (3) form

the basis of the models included in the UFBTC package

(University of Florida, 1989), as used in this paper.

2. Materials and methods

Two pure sorbent materials, alumina (gAl2O3, point of zero

charge (PZC)�9), and silica gel (precipitated silica, PZC�2–4),

and an aquifer sand were used in this study. g-Alumina was

obtained from Aldrich Chemical Co., (Milwaukee, WI) and has

an average mesh size of 150 and a specific surface area of

155 m2/g. Silica gel, obtained from Aldrich Chemical Co., has a

mesh size of 35–60 and a specific surface area of 300 m2/g.

While Porapak P was previously evaluated in batch studies

column experiments were not conducted with Porapak P

(hydrophobic medium) due to the extremely high cost of this

GC packing material. The aquifer sand, composed mostly of

quartz (SiO2) with minor fraction of albite (NaAlSi3O8), biotite

(K(Mg,Fe)3(AlSi3O10)(OH)2) and muscovite KAl3Si3O10(OH)2,

was collected from a major shallow alluvial aquifer in the

recharge area of Central Thailand (Pamok District, Angthong

Province, approximately 100 km north of Bangkok Metropo-

lis). It was sieved through a size 40–80 mesh prior to use. The

organic carbon content is �0.08% (Walkley–Black Method) and

iron content is 1.4 ppm.

Three pharmaceuticals were evaluated in this research;

acetaminophen (analgesic with pKa (9.38) and log Kow (0.46)),

17a-ethynyl estradiol (synthetic hormone with pKa (10.4) and

log Kow (3.67)), and nalidixic acid (antibiotic with pKa (5.95),

log Kow pH5 (1.54), log Kow pH7 (0.47), and log Kow pH9 (�1.16)).

These pharmaceuticals have been identified in the environ-

ment, and have fundamental properties of interest to this

work and were studied in our prior batch studies (Lorphensri

et al., 2006). These pharmaceutical compounds were pur-

chased from Aldrich Chemical Co.

ARTICLE IN PRESS

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0.0 1.0 1.5 2.0 2.5

Pore Volume

C/C

o

Alumina

Silica

Aq. Sand

Tracer (NaNO3)

0.5

Fig. 1 – Column breakthrough curves of acetaminophen with

silica, alumina, and aquifer sand which all show retardation

factor (R)�1.

WAT E R R E S E A R C H 4 1 ( 2 0 0 7 ) 2 1 8 0 – 2 1 8 82182

While acetaminophen was prepared in Nanopure water,

due to their lower solubility nalidixic acid and 17a-ethynyl

estradiol were prepared in stock solutions with MeOH 99%

purity, purchased from Aldrich Chemical Co.; the amount of

MeOH was less than 1% in the actual samples and thus had

negligible impact on sorption. Hydrochloric acid and sodium

hydroxide were used in pH adjustment. Calcium chloride

obtained from Fisher Scientific was added to maintain a

constant ionic strength (0.01 M CaCl2 � 2H2O) for all tests.

A SHIMADZU UV-1601 spectrophotometer was used to

analyze acetaminophen, and nalidixic acid. At neutral pH, a

wavelength of 242 nm was used to analyze acetaminophen. A

wavelength of 258 nm was used to analyze nalidixic acid. 17a-

ethynyl estradiol was analyzed by HP 8452A Diode Array

Spectrophotometer together with 4 cm cell in order to

increase the path length and thus improve detection. A

wavelength of 280 nm was used to analyze 17a-ethynyl

estradiol. In pH dependent experiments, a wavelength of

258 nm was used to analyze nalidixic acid since the spectra

has been shown to be relatively independent of pH (Park et al.,

2000; Djurdjevic et al., 1995).

In addition to the organic solutes, sodium chloride and

sodium nitrate were used in solute displacement experiments

to characterize the hydrodynamics of water flow through the

packed columns. Chloride was analyzed using an Orion

chloride electrode (model 94-35A) and sodium nitrate was

analyzed by an inline UV spectrophotometer using a 300 nm

wavelength (DW-10-D Star Instrument).

2.1. Batch experiments

For aquifer sand, which was not evaluated in our previous

batch study (Lorphensri et al., 2006), sorption studies were

conducted using a constant mass of 3 g sand with 8 mL of

solution in 15 mL glass vials with Teflon-lined caps; for

additional details see Lorphensri et al. (2006).

2.2. Continuous flow experiments

Borosilicate glass columns (i.d. 2.5 cm; length of 15 cm,

Chromaflex) were homogeneously packed with either silica,

alumina, or aquifer sand for this research. A 20 mm porous

HDPE bed support served as a media support at the bottom of

the column. A peristaltic pump (Masterflex, L/S) was used to

establish a constant flow rate through the column. The pore

volume and the amount of sorbent were determined grav-

imetrically.

To minimize gas entrapment in the column, a degassed

electrolyte solution (0.01 M CaCl2 � 2H2O) was injected from

bottom to top of the column and purged until the pH of the

effluent equaled the influent. A miscible displacement

experiment was started by switching to a solution containing

0.01 M CaCl2 � 2H2O and pharmaceutical which had an initial

concentration of 10 mg/L. After completion of pharmaceutical

injection, the solute pulse was displaced with electrolyte

solution (0.01 M CaCl2 �2H2O) until complete pharmaceutical

elution was observed and the column was flushed for an

additional 24 h before the tracer pulse was injected. The

effluent solution was connected to an inline-UV spectro-

photometer (DW-10-D Star Instrument) or automatic fraction

collector (Pharmacia Biotech, model RediFrac). Flow rates of

�0.5 ml/min were used for column experiments, which

corresponds to pore-water velocities of on the order of

10 cm/h. The column runs were operated at room tempera-

ture (25 1C) with the flow maintaining a relatively neutral pH

(7.070.5) except where pH dependent studies were con-

ducted; in these studies pH values of 4.3, 6.2, and 8.2 were

used.

3. Results and discussion

Column breakthrough curves of acetaminophen for all media

show ideal breakthrough curve characteristics with retarda-

tion factors (R) of �1.0 (Fig. 1). These column breakthrough

results are consistent with the negligible sorption observed in

batch studies from previous research. With a pKa of 9.38

acetaminophen exists almost solely in the neutral form at

neutral pH; even in this neutral form it has very low

hydrophobicity (log Kow�0.46). These properties help explain

the low sorption to both charged media (i.e., silica, alumina).

In addition, both batch and column results of the aquifer sand

show similar behavior for acetaminophen. Fig. 1 also includes

the breakthrough of the conservative tracer NaNO3, with the

resulting dispersion coefficient summarized in Table 1; the

breakthrough demonstrates ideal behavior which validates

the absence of physical nonidealities such as flow bypassing.

Conservative tracer studies conducted for all other column

studies were also ideal in nature and are thus not shown—the

resulting dispersion coefficients are all summarized in Table 1.

Column breakthrough curves of 17a-ethynyl estradiol to

silica, alumina and aquifer sand show different character-

istics from what was observed with acetaminophen (Fig. 2). In

addition to retardation, the breakthrough curves also display

a nonsymmetric characteristic which includes sharpened

breakthrough curve and tailing desorption front. As men-

tioned above, the conservative tracer tests exhibited ideal

breakthrough characteristic, indicating that hydrodynamics

are not responsible for the asymmetry in Fig. 2. The break-

through of 17a-ethynyl estradiol with the silica and alumina

columns showed similar levels of retardation (R�1.2) whereas

the breakthrough with aquifer sand was further delayed

(R�4.5). The breakthrough in these columns is consistent

ARTICLE IN PRESS

Ta

ble

1–

Co

lum

np

ack

ing

con

dit

ion

sa

nd

resu

lts

of

mo

del

fitt

ing

for

the

tra

nsp

ort

ex

peri

men

ts

Ph

arm

ace

uti

cals

Med

iap

Hle

ngth

(cm

)B

ulk

den

sity

r(g/c

m3)

Po

rosi

tyPo

revelo

city

(cm

/h)

Dis

pers

ion

coef

fici

en

t(c

m2/m

in)

BT

Cs

sha

pe

RN

Fk 2

(1/h

)b

or2

AC

ES

ilic

a6.77

0.1

4.7

50.4

30.4

614.7

40.0

38

Eq

.1.0

0.9

9

AC

EA

lum

ina

6.57

0.1

5.1

01.1

80.5

210.4

90.0

07

Eq

.1.1

0.9

9

AC

EA

q.S

an

d6.67

0.1

4.8

01.5

90.4

315.7

40.0

15

Eq

.1.0

0.9

9

EE

2S

ilic

a6.77

0.1

4.7

50.4

30.4

017.0

30.0

19

No

nE

q.

1.2

0.6

00.1

01.9

90.8

50.1

0.9

0

EE

2A

lum

ina

7.67

0.2

5.0

01.2

0.4

611.1

70.0

26

No

nE

q.

1.2

0.6

00.1

01.2

40.8

50.1

0.9

5

EE

2A

qS

an

d6.57

0.1

10.0

01.5

30.3

818.8

10.0

24

No

nE

q.

4.5

0.6

00.4

90.1

00.6

0.1

0.9

6

NA

LS

ilic

a6.67

0.1

4.4

00.4

60.4

017.9

80.0

10

No

nE

q.

8.0

0.5

90.5

41.2

80.6

1.0

0.9

7

NA

LA

qS

an

d6.27

0.1

10.0

01.4

80.4

018.2

60.0

25

No

nE

q.

150.0

0.5

00.6

00.0

30.6

1.0

0.9

0

NA

LA

qS

an

d4.37

0.1

4.7

51.6

10.4

412.2

10.0

04

No

nE

q.

83.0

1.0

00.3

90.1

50.4

3.0

0.9

4

NA

LA

qS

an

d8.27

0.1

5.0

01.5

30.4

013.2

00.0

10

No

nE

q.

4.5

0.8

0.7

41.4

70.8

0.5

0.9

7

AC

E:

Ace

tam

ino

ph

en

,N

AL:

Na

lid

ixic

aci

d,

EE

2:

17a-

eth

yn

yl

est

rad

iol

Eq

.:E

qu

ilib

riu

mm

od

el,

No

nE

q.:

No

neq

uil

ibri

um

mo

del.

WAT E R R E S E A R C H 41 (2007) 2180– 2188 2183

ARTICLE IN PRESS

WAT E R R E S E A R C H 4 1 ( 2 0 0 7 ) 2 1 8 0 – 2 1 8 82184

with the batch sorption experiment (Table 2) which showed

no significant sorption to charged surface of alumina and

silica but more significant sorption to a hydrophobic medium

and aquifer sand. Since 17a-ethynyl estradiol is more hydro-

phobic (log Kow�3.67), and since with a pKa of 10.4 it exists

almost solely in neutral form at the neutral pH range studied,

hydrophobic partitioning is anticipated as the dominant

mechanism. The batch sorption isotherm of 17a-ethynyl

estradiol to aquifer sand is clearly nonlinear; the model

fitting for the breakthrough data (discussed below) also shows

nonlinear characteristic consistent with the batch study. The

aquifer sand is composed mostly of quartz and minor

minerals such as albite, biotite, and muscovite. While the

aquifer sand contains only 0.08% organic carbon, this appears

to dominate the sorption capacity of 17a-ethynyl estradiol to

the aquifer sand, consistent with the more significant

sorption to Porapak P as compared to the other media in the

batch studies (Lorphensri et al., 2006).

The breakthrough curve of nalidixic acid with silica shows

nonequilibrium characteristics (Fig. 3); i.e., sharpened break-

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 5 10 15 20 25

Pore Volume

C/C

o

Aq. sand

Silica

Alumina

Model

Fig. 2 – Breakthrough curves of 17a- ethynyl estradiol with

silica, alumina, and aquifer sand. They all show

nonequilibrium characteristic. The model fitting for aquifer

sand gave the retardation factor (R)�4.5, N�0.6, F�0.49,

k2�0.1, b�0.6, x�0.1.

Table 2 – Summary of the sorption coefficient (Kd) of acetaminsilica, Porapak P, and aquifer sand. The sorption coefficients w

Pharmaceuticals Sorbent

17a-ethynyl estradiol Aluminaa

Silicaa

Porapak Pa

Aq. Sand

Nalidixic acid Aluminaa

Silicaa

Porapak Pa

Aq. Sand

Acetaminophen sorption was not significant in all cases.a Previous work (Lorphensri et al., 2006).b n/s : not significant effKd calculated from equation effKd ¼ Kf Ce

N�1 usingc Using Kf ¼ 5.59� 10�3 and N ¼ 0.47.d Using Kf ¼ 8.32�10�5 and N ¼ 0.59.e Using Kf ¼ 2.43� 10�2 and N ¼ 0.69.

through and tailing in the elution curve. The conservative

tracer test for this column exhibited ideal breakthrough

characteristics (data not shown); thus, the nonideal transport

is sorption related rather than transport related. In order to

confirm nonequilibrium conditions, a flow-interruption was

conducted (Brusseau et al., 1997, 1989) on a column study of

nalidixic acid and silica for both sorption and desorption

fronts and at the plateau C/Co�1 (data not shown); the results

of these flow interruptions confirmed that nonequilibrium

sorption conditions were experienced (i.e., the breakthrough

concentrations decreased after a flow interruption as the

pharmaceutical had additional time to approach equilibrium

sorption). The model fitting for this breakthrough data also

show nonlinear characteristics as discussed below. Column

studies of nalidixic acid with alumina were not conducted

because the result of batch experiments suggested retarda-

tion factors on the order of 12,000 pore volumes. At neutral

pH, nalidixic acid is present in both neutral and anionic

forms. Therefore, both hydrophobic and electrostatic

sorption mechanisms are expected to be important. However,

ophen, 17a-ethynyl estradiol, nalidixic acid to alumina,ere normalized by specific surface area of sorbents

pH Kd (L/m2) r2

7.4270.01 n/sb

6.7170.02 n/sb

7.3470.17 4.40� 10�3 0.97

6.9170.14 2.38� 10�3c 0.96

7.3170.05 3.00� 10�2 0.99

6.6470.03 4.30�10�5d 0.99

6.7070.02 2.00� 10�4 0.99

6.6570.12 1.43� 10�2e 0.97

Ce ¼ 5 mg/l.

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 20 40 60 80 100

Pore volume

C/C

o

Silica

Model

Fig. 3 – Breakthrough curves of nalidixic acid with silica. It

shows nonequilibrium characteristic. The model fitting

gave the retardation factor (R)�8, N�0.59, F�0.54, k2�1.28,

b�0.6, x�1.0.

ARTICLE IN PRESS

EE2

NAL pH 6.2

Benzene

Naphthalene

PhenanthreneNAL pH 4.3

NAL pH 8.2

0.0

1.0

2.0

3.0

4.0

5.0

-1.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0

log Kow

log K

oc

Karickhoff, 1981

Bintein and Devillers, 1994

Fig. 5 – The relationship of log Koc ( where Koc ¼ Kd/foc) of

aquifer sand and log Kow from this study were shown along

with the log Koc estimation by 1 octanol/water partition

coefficient, log Kow (lower line, Karickhoff, 1981). The upper

line represents the work done by (Bintein and Devillers,

1994) which includes correction factor regarding ionization

process (pKa). Sorption of acetaminophen is not significant

and not shown. NAL, EE2 represent nalidixic acid, and 17a-

ethynyl estradiol respectively.

WAT E R R E S E A R C H 41 (2007) 2180– 2188 2185

electrostatic attraction had a greater influence on the sorp-

tion than did hydrophobic partitioning as evidence by high

sorption on positively charged alumina but low sorption to

hydrophobic medium. The sorption isotherm of nalidixic acid

was linear for alumina and Porapak P but nonlinear for silica

(Lorphensri et al., 2006).

The breakthrough curve of nalidixic acid with aquifer sand

(Fig. 4) shows the same characteristics as above with silica.

Since, the conservative tracer test for the aquifer sand

column exhibited ideal breakthrough characteristics (data

not shown), this suggests that nonideal transport of nalidixic

acid with aquifer sand column is sorption related rather than

hydrodynamics related. The sorption mechanism of nalidixic

acid to aquifer sand are expected to be the electrostatic

attraction between positively charged mineral surface and

negatively charged nalidixic acid, and hydrophobic partition-

ing to organic phase on the surface of sand. Although, the

main mineral composition is sand (silica) which normally has

negative-charged surface, the sorption of nalidixic acid to this

sediment is still high. Therefore, the sorption may be

contributed by a combination effect of positively-charged

mineral oxides (i.e., iron oxide) which appear to be present on

the silica grains according to the aquifer sand having brown

color which can be an evidence of iron oxide coverage. In

addition to iron oxide, aluminum oxide may be present in

aluminosilicate minerals such as feldspar, biotite, and

muscovite. Moreover the sorption of nalidixic acid may be

by partitioning to organic content, even though the amount of

organic carbon content is 0.08%. Thus, the resulting retarda-

tion factor of nalidixic acid to silica is �8 (see Fig. 3) whereas

the aquifer sand exhibits much larger retardation (R�150), as

shown in Fig. 4.

The correlation of pharmaceutical log Koc values with

1-octanol/water partition coefficient log Kow is shown in

Fig. 5 along with some aromatic hydrocarbon compounds

(i.e., benzene, naphthalene, phenanthrene) (Karickhoff, 1981).

A positive correlation assumes that compounds sorb to

natural media (Koc) by partitioning to organic fraction in the

media. The 17a-ethynyl estradiol agrees well with this

relationship for aromatic compounds (see Fig. 5). As men-

tioned previously, 17a-ethynyl estradiol is in its neutral form

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 200 400 600 800

Pore Volume

C/C

o

Aq. sand

Model

Fig. 4 – Breakthrough curves of nalidixic acid (pH�6.2) with

aquifer sand and nonequilibrium nonlinear model. The

model fitting gave the retardation factor (R)�150, N�0.5,

F�0.6, k2�0.03, b�0.6, x�1.0.

at neutral pH, hence the sorption is expected to be by

hydrophobic partitioning. The sorption of nalidixic acid,

which is partially/fully ionized in the neutral pH range, is

larger than predicted based on the estimate of Karickhoff,

1981). The other estimation (Bintein and Devillers, 1994)

shown in upper straight line, is based on the following

equation:

log Kp ¼ 0:92 log Kow þ 1:09 log foc þ 0:33 CFaþ 0:3, (4)

where CFa is log ð1=ð1þ 10pH�pKaÞÞ. This equation was devel-

oped based on 229 Kd values of 53 chemicals of both ionized

and nonionized form. The results for nalidixic acid are

parallel to the correlations of Bintein and Devillers (1994)

and Karickhoff (1981) indicating that hydrophobic partition-

ing is an important component of the nalidixic acid sorption.

The higher sorption for nalidixic acid at pH 6.2 where more

nalidixic acid is in the anionic form, demonstrates the

importance of speciation in the process.

3.1. pH dependent transport of nalidixic acid

Nalidixic acid is a very weak organic acid, with a pKa of 5.95.

This suggests that nalidixic acid can be present in both

neutral and anionic forms in the neutral pH range, and thus

the sorption capacity can also vary in this pH range. In order

to investigate the influence of pH on sorption and transport of

nalidixic acid with aquifer media, batch and column experi-

ments in the range of pH 4–9 were evaluated. The batch pH-

sorption profiles of nalidixic acid to aquifer sand are shown in

Fig. 6. The sorption magnitude gradually increases and

reaches it maximum at pH�6. The sudden drop of sorption

is shown after pH greater than 6 and reaches its minimum at

pH�9. The sudden drop at pH�6 may suggest that any metal

oxides present on the sand surface plays an important role in

decreasing the sorption capacity around pH 7 (PZC of iron

oxide (pH 7–9), (Cornell and Schwertmann, 1996). Column

studies were conducted at these pH values: 4.3, 6.2 and 8.2.

ARTICLE IN PRESS

0.00

0.05

0.10

0.15

0.20

3.0 4.0 5.0 6.0 7.0 8.0 9.0

pH

Kd [

L/g

]

0.0

0.2

0.4

0.6

0.8

1.0

1.2

Fra

ction o

f P

harm

aceutical

[Neutral] [Anionic]

Fig. 6 – pH-sorption profiles of nalidixic acid to aquifer sand

is shown along with the fraction of neutral and anionic

forms of nalidixic acid. Two lines represent the fraction of

neutral and anionic forms of nalidixic acid.

0.0

0.2

0.4

0.6

0.8

1.0

1.2

0 100 200 300 400

Pore Volume

C/C

o

pH 8.2pH 4.3pH 6.2model

Fig. 7 – Breakthrough curve of nalidixic acid with aquifer

sand at different pH. They all show nonequilibrium

characteristic. The retardation factors at pH 8.2, 4.3, and 6.2

are 4.5, 83, and 150, respectively.

0.1

1.0

10.0

100.0

1000.0

0.1 1.0 10.0 100.0 1000.0

R batch experiment

R c

olu

mn e

xperim

ent

EE2 - Aq

NAL - Si

NAL - AqpH 6.2

NAL - AqpH 8.2

NAL - AqpH 4.3

ACE - Si,Al,AqEE2 - Si,Al

ACE : AcetaminophenNAL : Nalidixic acidEE2 : 17�-ethynyl estradiolSi : SilicaAl : AluminaAq : Aquifer sand

Fig. 8 – Comparison of retardation factors derived from batch

sorption experiment and transport experiment.

WAT E R R E S E A R C H 4 1 ( 2 0 0 7 ) 2 1 8 0 – 2 1 8 82186

According to the combination of nalidixic acid speciation and

surface charges of sand, the breakthrough order of nalidixic

acid is expected to be in the order of pH 8.2 first, followed by

pH 4.3, and finally pH 6.2; this was in fact observed in the

column studies as shown in Fig. 7. As expected from batch

results, the greatest retardation (R�150) was observed for pH

6.2, and the least retardation (R�4.5) was found in column of

pH 8.2.

3.2. Batch-column comparisons and modeling results

The comparison of retardation factors from modeling and

calculated from batch studies are shown in Fig. 8. There is

generally very good agreement between retardation factors

derived from batch experiment and measured in column

experiments, which is very encouraging as it supports the use

of batch results to assess the field transport of pharmaceu-

tical compounds. Nonetheless, users should exercise caution

when using this approach for nalidixic acid, since the

sorption capacity of nalidixic acid varies significantly with

pH, especially for pH values near its pKa. Therefore, the

estimation from batch experiment must be performed at the

same pH condition of column experiment in order to obtain

good agreement with column experiment.

The computer program CXTFIT2 (Toride et al., 1989) was

used to simulate the local equilibrium assumption model for

tracer and acetaminophen experiment. CXTFIT2 is a non-

linear least squares curve fitting computer program that is

used to determined dispersion and retardation factor for

solute transport of one-dimensional experiment column

data. The hydrodynamic properties (i.e., dispersion coeffi-

cient) of the columns packed with the silica, alumina and

aquifer sand were inferred from breakthrough curves of the

conservative tracers (sodium chloride or sodium nitrate). The

shapes of tracer breakthrough curves were highly symmetric

with sharp fronts and breakthrough exactly one pore volume.

The results also indicate that the flow regime within all

column studied was dominated by advection.

The computer program UFBTC version 2.0 (University of

Florida, 1989) was used to simulate nonequilibrium assump-

tion model and/or nonlinear sorption. This program uses

finite-different numerical techniques to estimate the relative

concentrations at different time. The breakthrough curves for

pharmaceuticals were fitted with linear/nonlinear and equili-

brium/nonequilibrium model using the calculated average

pore water velocity and estimated D obtained from tracer test.

The F and k2 were calculated from the outputs of model fitting

parameters (i.e., retardation factor, b, o) (Table 1).

The two sites model was used to characterize sorption

nonequilibrium. It assumes that sorption occurs in two types

of domain: an instantaneous equilibrium type and rated-

limited type, with the kinetics of sorption in the latter domain

characterized by k2. The influence of sorption nonequilibrium

on organic contaminant transport has been recognized to be

important. Brusseau and Rao (1989) compiled and analyzed

extensive sorption kinetic database, revealing the existence of

an inverse log–log relationship between desorption rate

constants (k2) and corresponding equilibrium sorption con-

stants (Kd). They noted that the approach toward sorption

equilibrium was more constrained for solutes with reactive

functional groups (e.g., amino, phenoxyl, and carboxylic acid

groups), which was suggested to be the consequence of

specific interaction of solute functional groups on the specific

charge site on the sorbent. This is reflected by smaller k2 value

for a polar solute when compared with an equally sorptive

nonpolar solute.

The plot of log k2 and log Kd of nalidixic acid and 17a-ethynyl

estradiol are shown in Fig. 9. The parameters of nalidixic acid

ARTICLE IN PRESS

WAT E R R E S E A R C H 41 (2007) 2180– 2188 2187

at three different pH values (different levels of ionization) are

shown along with the relationships of hydrophobic and polar/

ionizable organic chemicals as compile by Brusseau and Rao

(1989). The results are seen to be consistent with and

intermediate to the results reported by Brusseau and Rao

(1989).

A linear relationship of F and ionized fraction of nalidixic

were shown in Fig. 10. It suggested the fraction of instanta-

neous sorption increased with the amount of ionized fraction

of nalidixic acid. At pH 4, the lowest fraction of ionization the

F was relatively low, while at higher pH (pH 6.2 and 8.2) and

higher ionized fraction, F was also relatively high. This may

suggest that the more ionized forms exist, the more can reach

the instantaneous sorption sites.

It has been suggested that the most probable cause of

sorption nonequilibrium for nonhydrophobic (polar/ionizable

organic chemicals) is chemical nonequilibrium and intrasor-

bent diffusion. Altfelder and Streck (2006) has stated that the

limitation of first-order model in describing the data mea-

sured in short term and long term kinetic sorption is that the

rate parameter and Freundlich coefficient are strongly time-

dependent, usually in studies involving column experiments.

Sabbah et al. (2005) also discuss how time-dependent sorp-

Hydrophobic

y = -0.668x + 0.301

Polar/Ionizable

y = -0.62x - 1.789

-6.0

-5.0

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

-4.0 -2.0 0.0 2.0 4.0

log Kd

log k

2

Nalidixic acid

17�- ethynyl estradiol

pH4.3

pH8.2

pH6.2

Fig. 9 – The inverse relationship of log K2 and log Kd of

nalidic acid at different pH values were plotted along with

the relationship of hydrophobic and polar/ionizable organic

compounds compiled by Brusseau and Rao (1989).

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0.0 0.2 0.4 0.6 0.8 1.0

Ionized fraction

F

pH 4.3

pH 6.2

pH 8.2

Fig. 10 – The linear relationship of ionized fraction of

nalidixic acid at different pH and F.

tion and desorption impacts the breakthrough and elution of

organic compounds, demonstrating that the extent of the

impact is related to the degree of nonequilibrium in the

system. These factors need to be considered when evaluating

and predicting nonequilibrium transport of organic contami-

nants.

4. Conclusions

The objective of this research was to compare previous batch

results with the sorption and transport of select pharmaceu-

ticals in continuous column flow studies and to assess the

role of kinetics in the sorptive transport in column system.

Sorption and transport of organic compounds in natural

geologic media has been shown to depend largely on the

organic fraction and composition of the medium. For

example, while the aquifer sand studied here had a low

organic content, significant sorption and retardation of

relatively hydrophobic 17a-ethynyl estradiol (log Kow of 3.67)

was still observed. In similar manner, acetaminophen, with

low hydrophobicity (log Kow of 0.46), exhibited virtually no

sorption and no retardation in aquifer sand studies. Nalidixic

acid, an anionic organic compound with pKa in neutral pH

range, exhibited strong sorption and large retardation to low

organic aquifer sand, much higher than expected based on

hydrophobicity (log Kow of 1.54). The sorption of nalidixic acid

was also observed to be highly dependent on pH, which

further supports electrostatic forces as dominant in the

sorption. This large retardation may be caused by iron oxide

appearing on the surface of sand grains, which at neutral pH

is expected to be oppositely charged to anionic nalidixic acid.

The column transports of nalidixic acid with silica and

aquifer sand showed nonideal characteristics (nonequili-

brium and nonlinearity). This nonequilibrium transport was

successfully described using dual domain nonequilibrium

transport equations. Comparing column results here with

previous batch studies demonsrates that retardation factors

derived from batch sorption studies are in good agreement

with prediction of pharmaceutical transport in column

systems. This is very encouraging as it supports the use of

batch results to assess the field transport of pharmaceutical

compound. However, the sorption and transport of ionizable

pharmaceutical compounds, such as nalidixic acid, is highly

pH dependent and can vary significantly with even small

changes in pH near the pKa, indicating that such studies

should be conducted and interpreted carefully. All these

processes are thus important to successfully assess the field

sorption and transport, and thus the fate and exposure, of

pharmaceutical compounds in the environment.

Acknowledgments

Funding for this work has been partly provided through the

United States Environmental Protection Agency Science to

Achieve Results (STAR) program, through Grant number

R829005 and partly through the National Research Center

for Environmental and Hazardous Waste Management (NRC-

EHWM) Program, Chulalongkorn University, Thailand.

ARTICLE IN PRESS

WAT E R R E S E A R C H 4 1 ( 2 0 0 7 ) 2 1 8 0 – 2 1 8 82188

Although the research described in this article has been

funded by the above-mentioned agencies, it has not been

subjected to the agencies’ required peer and policy review and

therefore does not necessarily reflect the views of the agency

and no official endorsement should be inferred.

R E F E R E N C E S

Alder, A., Golet, E., Ibric, S., Giger, W., 2000. Fate of fluoroquino-lone antibiotics during municipal wastewater treatment.Extended Abstr. Am. Chem. Soc. Natl. Meet. 40, 103–104.

Altfelder, S., Streck, T., 2006. Capability and limitation of first-order and diffusion approaches to describe long-term sorptionof chlortoluron in soil. J. Contaminant Hydrol. 86, 279–298.

Bintein, S., Devillers, J., 1994. QSAR for organic chemical sorptionin soil and sediment. Chemosphere 28, 1171–1188.

Brusseau, M.L., 1991. Influence of sorbate structure on none-quilibrium sorption of organic compounds. Environ. Sci.Technol. 25, 1501–1506.

Brusseau, M.L., Rao, P.S.C., 1989. The influence of sorbate–organicmatter interactions on sorption nonequilibrium. Chemo-sphere 18, 1691–1706.

Brusseau, M.L., Rao, P.S.C., Jessup, R.E., Davidson, J.M., 1989. Flowinterruption: a method for investigating sorption equilibrium.J. Contam. Hydrol. 4, 223–240.

Brusseau, M.L., Hu, Q., Srivastara, R., 1997. Using flow interrup-tion to identify factors causing nonideal contaminant trans-port. J. Contam. Hydrol. 24, 205–219.

Cornell, R.M., Schwertmann, U., 1996. The Iron Oxides. Structure,Properties, Reactions, Occurrence, and Uses. VCH, Weinheim,Germany.

Daughton, C., Ternes, T., 1999. Pharmaceuticals and personal careproducts in the environment: agents of subtle change?Environ. Health Perspect. 107, 907–938.

Djurdjevic, P., Jelic, R, Stankov, M., Veselinovic, D.S., Stankov, D.,1995. Complexation behavior of cadmium (II) ion withnalidixic acid in chloride medium. J. Serbian Chem. Soc. 60,403–410.

Freeze, R.A., Cherry, J.A., 1979. Groundwater. Prentice-Hall,New Jersey.

Golet, E.M., Alder, A.C., Giger, W., 2002. Environmental exposureand risk assessment of fluoroquinolone antibacterial agents in

wastewater and river water of the Gatt Valley watershed,Switzerland. Environ. Sci. Technol. 36, 3645–3651.

Grathwohl, P., 1998. Diffusion in Natural Porous, Media: Con-taminant Transport, Sorption/Desorption and DissolutionKinetics. Kluwer Academic Publishers, Boston, MA.

Halling-SØrensen, B., Nielson, S.N., Lanzky, P.F., Ingerslev, F.,Holten Lutzhoft, J., Jorgensen, S.E., 1998. Occurrence, fate andeffects of pharmaceutical substances in the environment—areview. Chemosphere 36, 357–393.

Heberer, T., 2002. Occurrence, fate, and removal of pharmaceu-tical residues in the aquatic environment: a review of recentresearch data. Toxicol. Lett. 131, 5–17.

Karickhoff, S.W., 1981. Semi-empirical estimation of sorption ofhydrophilic pollutants on natural sediment and soils. Che-mosphere 10, 833–846.

Lorphensri, O., Intravijit, J., Sabatini, D.A., Kibbey, T.C.G., Osatha-phan, K., Saiwan, C., 2006. Sorption of acetaminophen, 17a-ethynyl estradiol, nalidixic acid, and norfloxacin with silica,alumina, and a hydrophobic medium. Water Res. 40,1481–1491.

Park, H., Chung, K., Lee, H., Lee, J., 2000. Ionization and divalentcation complexation of quinolone antibiotics in aqueoussolution. Bull. Korean Chem. Soc. 21, 849–854.

Sabbah, I., Ball, W., Young, D., Bouwer, E., 2005. Misinterpretationin the modeling of contaminant desorption from environ-mental solids when equilibrium conditions are not fullyunderstood. Environ. Eng. Sci. 22 (3), 350–366.

Ternes, T.A., Meisenheimer, M., Mcdowell, D., Sacher, F., Brauch,H., Haist-Gulde, B., Preuss, G., Wilme, U., Zulei-Seibert, N.,2002. Removal of pharmaceuticals during drinking watertreatment. Environ. Sci. Technol. 36, 3855–3863.

Toride, N., Leij, F.J., van Genuchten, M.T., 1989. The CXTFIT codefor estimating transport parameters from laboratory or fieldtracer experiments, version 2.0. Soil Sci. Soc. Am. J. 53,1303–1310.

University of Florida, 1989. UFBTC, version 2-One dimensionalsubsurface transport model, finite difference technique.

van Genuchten, M.T., 1981. Non-equilibrium transport para-meters from miscible displacement experiments. ResearchReport 119. US Salinity Laboratory, Riverside, CA.

van Genuchten, M.T., Wagenet, R.J., 1989. Two-site/two regionmodels for pesticide transport and degradation: theoreticaldevelopment and analytical solutions. Soil Sci. Soc. Am. J. 53,1303–1310.