Journal of Membrane Science 266 (2005) 30–39
Chemical interactions between dissolved organic matter and
low-molecular weight organic compounds:
Impacts on membrane separation
Sarah K. Dalton, Jonathan A. Brant, Mark R. Wiesner ∗
Rice University, Environmental and Energy Systems Institute, 6100 Main Street, MS-317, Houston, TX 77005, USA
Received 14 January 2005; received in revised form 27 April 2005; accepted 6 May 2005
Available online 13 June 2005
Abstract
This work considers the hypothesis that the association of trace-level hydrophobic organic compounds with dissolved organic matter
(DOM), similar to macromolecular materials found in natural water systems, may influence the contaminant’s ability to permeate across
synthetic membranes. A batch dialysis system using semipermeable membranes in conjunction with a model lipid phase was used to explore
the impact of systematic changes in solution-chemistry on the permeability of four low-molecular weight organic compounds (LMWOCs) of
environmental concern: cyclonite, atrazine, naphthalene, and 2,4-dichlorophenol. Similar to previous studies, no correlation could be drawn
between molecular weight and contaminant permeability across the membrane for the four LMWOCs. However, contaminant transport was
observed to depend on the polarity and hydrophobicity of the LMWOCs. Moreover, the interactions between the organic compounds and
DOM varied as a function of solution chemistry (i.e., pH and divalent electrolyte concentration). These results demonstrate considerable
variability in the importance and the underlying mechanisms of interactions that may occur between LMWOCs and natural organic matter
during membrane separations.
© 2005 Elsevier B.V. All rights reserved.
Keywords: Organic matter; Membrane separation; Contaminant permeability
1. Introduction
Organic compounds, such as pesticides and herbicides,
exist as contaminants in many potable source waters, and in
turn pose a risk to human health if not removed. Additionally, the presence of these compounds in agricultural drainage
has significant impacts on surrounding ecosystems. In conventional water treatment systems, these compounds are
typically removed either through adsorption (e.g., activated
carbon) or oxidative processes (e.g., ozone) [1]. However,
these processes are plagued by a number of complications that
reduce their cost effectiveness and overall removal efficiency
[1]. Recently, membrane separation processes have emerged
as an attractive treatment option for removing organic contaminants, such as pesticides and herbicides from source
∗
Corresponding author. Tel.: +1 713 285 5129; fax: +1 713 348 5203.
E-mail address: [email protected] (M.R. Wiesner).
0376-7388/$ – see front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.memsci.2005.05.007
waters [1–5]. Membranes have specifically been identified
for removing organic contaminants as they do not require
the addition of costly chemicals and are relatively simple to
operate.
Membrane processes, particularly reverse osmosis (RO)
and nanofiltration (NF), have been shown to effectively
remove a wide variety of organic contaminants in potable
water treatment [1–5]. Nevertheless, the mechanisms by
which these compounds are removed are complex and are
as of yet not thoroughly understood. In membrane separations, removal is a function of both the characteristics of the
organic compound and the membrane [3,5]. Those characteristics that have been identified as being the most significant
in determining removal efficiency include: membrane molecular weight cutoff (MWCO), the molecular weight of the
compound, molecular size and shape, dipole moment, surface charge, and hydrophobicity of both the membrane and
compound.
S.K. Dalton et al. / Journal of Membrane Science 266 (2005) 30–39
Membrane molecular weight cutoff is commonly used to
determine its ability to reject a targeted compound(s). However, rejection of dissolved organics has been found to be
dependent on both the properties of the membrane and the
solute, and ultimately on the solution chemistry [4–7]. In fact,
it was proposed by Cho et al. [7] that the effective MWCO for
a membrane is a function of membrane and solute charge and
hydrophobicity. The value of using membrane MWCO as an
indicator for rejection of organics is less reliable for compounds with a molecular weight less than 250 g mol−1 ; such
compounds are identified as low-molecular weight organic
compounds (LMWOC), for which there is virtually no correlation between membrane removal efficiency and molecular
weight [5,6]. Rejection of these compounds is likely to be
highly sensitive to changes in functional group ionization,
specificity of interactions with the membrane surface, and
interactions with other dissolved compounds [4,6,7].
The presence of dissolved organic material has been
shown to significantly impact membrane performance, quantified by permeate flux and solute rejection [7–10]. Dissolved
organic matter (DOM) is a ubiquitous component of natural
water systems [11] that may function as an auxiliary phase
to alter the speciation and transport behavior of other organic
compounds [12,13]. In this way, organic compounds may
exist in either a freely dissolved phase or as a complex with
DOM [14]. Because solution chemistry (i.e., ionic strength
and pH) controls the charge and configuration (structure)
of organic macromolecules [8], the form and extent of the
contaminant–DOM interaction is a function of the nature of
the DOM, the properties of the contaminant, and the solution
chemistry [15]. It has previously been found that rejection
of organic compounds in some membrane-based monitoring
systems tends to increase in the presence of DOM [16,17].
Although this behavior can be generally attributed to a variety
of factors (e.g., changes size, shape, and surface chemistry),
the governing mechanism remains ill defined. It is therefore
necessary to assess the behavior and interaction of LMWOCs
with water treatment membranes in the presence of DOM.
The objective of this investigation is to characterize the
mechanisms that dictate the permeability or rejection of low-
31
molecular weight organic compounds (MW < 250 g mol−1 )
by a synthetic membrane. In this respect, the properties of
the contaminants, such as charge, polarity, and hydrophobicity, on their transport are accounted for. Furthermore, the
role played by DOM (fulvic and tannic acid) and solution
chemistry in these processes is considered and discussed.
2. Materials and methods
2.1. Organic compounds
Four LMWOCs were selected for study in this
investigation, representing a range of chemical characteristics: atrazine, 2,4-dichlorophenol, naphthalene,
and hexahydro-1,3,5-trinitro-1,3,5-triazine or cyclonite
(Table 1). Furthermore, the 2,4-dichlorophenol was studied
in both its neutral and ionized states, which are, respectively,
designated as DCPo and dcp− throughout this manuscript.
Atrazine, 2,4-dichlorophenol, and naphthalene were all
acquired from Sigma Aldrich, Inc. (St. Louis, MO), while
cyclonite was acquired from SRI International (Menlo Park,
CA). The reported purity for each compound was ≥98%;
stock solutions were protected from light exposure and were
stored in a refrigerator at 5 ◦ C. All compounds were used
in their 14C-radiolabeled form to facilitate more accurate
identification and analysis. The concentration of all radiolabeled contaminants in solution was determined through
liquid scintillation counting (Beckman LS6500, GMI,
Inc., Ramsey, MN). Solutions were prepared using doubly
deionized water (DDW) (resistance ≥ 10 m) supplied by a
Millipore (Billercia, MA) water purification system.
Based on the reported molecular weight for each compound cyclonite is the largest compound followed by atrazine,
2,4-dichlorophenol, and naphthalene. The polarity of each
compound is expressed in terms of its dipole moment (α).
Accordingly, cyclonite is the most polar molecule followed
by atrazine, 2,4-dichlorophenol, and naphthalene, which
is non-polar. A value of 3.7 for α has been reported for
dcp− [18], making it more polar than its neutral form
Table 1
Selected physical and chemical characteristics of the contaminant compounds used in this investigation
MW (g mol−1 )
Cyclonite
Atrazine
2,4-Dichlorophenol
Naphthalene
a
b
c
d
e
f
g
h
i
222.3
215.7
163.0
128.2
Dipole moment (D)
5.7c
2.8d
1.6e
0f
Solubilitya,b (mg L−1 )
42.3g
33.4
4500h
31.5
log KOW a,b
log KLW
0.87
2.56
3.08i
3.45
0.98
2.66
3.18
3.56
Mackay et al. [28,29].
Schwarzenbach et al. [30].
Dioxane at 25 ◦ C, from Calderbank and Pierens [31].
n-Hexane at 23 ◦ C, from Kavetskii and Bublik [32].
DCPo in benzene at 25 ◦ C, from Lumbroso et al. [33].
Benzene at 25 ◦ C, from Huang and Ng [34].
From Singh [35].
For neutral DCPo [28]; estimated as 5000 mg L−1 at pH 6.85 and 42,500 mg L−1 at pH 8.85 per correlation equation from Huang et al. [36].
For neutral DCPo ; Escher and Schwarzenbach [37] reported a value of −0.46 at pH 12 in presence of 100 mM K+ .
32
S.K. Dalton et al. / Journal of Membrane Science 266 (2005) 30–39
(DCPo ). The ability of each compound to interact with
water (i.e., hydrophobicity) is quantified using the equilibrium octanol–water partition coefficient (KOW ) [19]. Low
log KOW values correspond to hydrophilic compounds while
higher values are indicative of hydrophobic ones. Thus, from
Table 1, cyclonite is characterized as hydrophilic while dcp−
(log KOW = 1.9), atrazine, 2,4-dichlorophenol, and naphthalene are increasingly hydrophobic. The lipid–water partition coefficient for triolein (KLW ) was calculated from the
reported log KOW data for the non-ionic contaminant species
according to Chiou [20]. From Table 1, naphthalene has
the highest affinity for the triolein phase, followed by 2,4dichlorophenol, atrazine, and cyclonite.
2.2. Dissolved organic matter
Devitt and Wiesner [21] demonstrated that atrazine transport across semipermeable cellulose ester membranes was
most significantly impacted by the lower molecular fractions
(100 Da < MW < 500 Da), of a DOM surrogate, tannic acid,
which is in the range of fulvic acids. In the current study,
two types of DOM were: tannic acid (Sigma Aldrich, Inc.)
and soil fulvic acid (International Humic Substances Society, St. Paul, MN), which was fractionated from a natural
sample. Tannic acid in an attractive surrogate for the lower
molecular humic materials due to its availability, cost, and
controlled composition as well as its molecular weight distribution. Organic carbon contents were analytically determined
as 49.5% for tannic acid (r2 = 0.99) and 46.3% for the soil
fulvic acid (r2 = 0.99). DOM concentrations were measured
using either a total organic carbon (TOC) analyzer (Shimadzu
TOC-5050A, Torrance, CA) or a UV–vis spectrophotometer
(Hitachi U-2000, Tokyo, Japan), and are reported in units
of mg TOC L−1 . In this manner, DOM concentration was
determined using an established calibration curve (r2 = 0.99)
of DOM concentration versus TOC concentration or UV
absorbance strength at λ = 254 nm.
sodium azide) applied by the manufacturer. The tubing was
then stored until used in DDW at 5 ◦ C that was changed-out
regularly.
2.4. Batch dialysis set-up
Contaminant transport is studied in this investigation
using the semipermeable membrane device (SPMD) technique [22]. In the SPMD technique, organic contaminants
may diffuse across a membrane into a permeate solution
where it may then diffuse across a second membrane into
a model non-polar lipid solution (triolein). Because concentration and osmotic pressure difference serve as the driving
force for transport, this system allows for the study of contaminant/solute interactions in the absence of hydrodynamic
forces. Uptake of organics by the lipid phase is in accordance
with polymer permeability and equilibrium partition theory
and is therefore a function of the partitioning coefficient
between the membrane and water (KMW ); the partitioning
coefficient between the lipid and water (KLW ); and the
contaminant octanol–water partitioning coefficient (KOW )
[23].
A schematic of the SPMD test unit used in this investigation is depicted in Fig. 1. The dialysis system is composed
of three primary compartments corresponding to the feed,
permeate, and triolein (Sigma Aldrich, Inc.) solutions. The
purpose of the triolein or lipid solution is two fold. One is to
serve as a sink for the compound of interest, thereby maintaining a concentration gradient to facilitate transport from
the source (feed) compartment and allow for accurate characterization of the thermodynamic partitioning of the organic
compounds. Secondly, by accumulating the contaminant(s)
in the test system, mass balances on contaminant leaving the
feed compartment are facilitated. The feed (VT = 5 mL) and
triolein (VT = 5 mL) solutions are sealed in two dialysis membrane containers and immersed in the permeate solution. In
2.3. Membranes
Non-sterile Spectra/Por Biotech (Spectrum Laboratories,
Rancho Dominguez, CA) cellulose ester membrane tubing
was used in all dialysis experiments. Cellulose dialysis membranes are considered to be hydrophilic due to the presence of
surface hydroxyl groups. These membranes have a molecular
weight cutoff of 100 Da, and were chosen based on preliminary experiments in this work as well as the findings of Devitt
and Wiesner [21] where 52% (based on mass) retention of
the tannic acid was observed for these membranes. Because
the MWCO of the cellulose ester membrane is close to the
molecular weight of the organic compounds (Table 1) and
DOM evaluated in this work, subtle changes in rejection due
to association between the organic compounds and DOM
are better isolated. Prior to use, the membrane tubing was
thoroughly rinsed, then soaked in three successive batches of
excess DDW for 30 min to remove any residual biocide (0.1%
Fig. 1. Schematic of batch dialysis system used in dialysis experiments.
S.K. Dalton et al. / Journal of Membrane Science 266 (2005) 30–39
this set-up, contaminant transport was measured as a function
of time from the feed to the permeate and triolein solutions.
For the feed and triolein solutions the length of membrane
between the two closures was approximately 7 cm to yield a
total surface area of about 34 cm2 and a surface area-to-lipid
volume ratio of 6.7 cm−1 . A rigid wire was used to maintain
a constant separation distance between the feed and triolein
vessels so that they would not be in direct contact with each
other. The permeate solution (VT = 500 mL) had a fixed initial chemistry (i.e., ionic strength and pH) depending on the
conditions studied and was housed in a covered polystyrene
dialysis reservoir that was stirred for the specified dialysis
time at ambient temperature (22 ± 1 ◦ C).
Contaminant transport was studied as a function of three
different solution chemistries. In the first series, the suspension consisted of only the organic compound and is said to be
freely dissolved; in the second series the feed is composed of
a contaminant and a DOM source (either tannic or soil fulvic
acid); while in the third series, the suspension is composed
of the contaminant in the presence of DOM with a background electrolyte (calcium nitrate). For 2,4-dichlorophenol,
buffered solutions were conducted in pairs to allow comparison of this weak organic acid in its neutral (DCPo ) and ionized
(dcp− ) forms. To ensure that no less than 90% of the 2,4dichlorophenol mass was in the desired ionic state, the solution pH was buffered at least one unit above (ionized) or below
(protonated) the pKa for 2,4-dichloropehnol (pKa = 7.85),
with an upper limit of pH 9 imposed by the membrane
tolerance. Solutions were buffered using either archerite
(KH2 PO4 ) or sodium phosphate (Na2 HPO4 ) according to
standard methods. The suspensions were covered and stirred
at room temperature (22 ± 1 ◦ C) for 24 h to permit equilibration.
The experimental matrix included variations in the absolute and relative concentrations of the contaminant and DOM
added to the batch system via the feed solution, calcium
concentration, pH, and dialysis time. For the source solutions, initial contaminant concentrations ranged from approximately 0.01 to 0.1 mol L−1 , DOM concentrations ranged
from 0 to 70 mg TOC L−1 , and Ca(NO3 )2 was added at levels of 0 and 50 mg L−1 . In most cases, the contaminant to
DOM ratio (i.e., contaminant concentration/DOM concentration) was explored at three levels (0.7, 1.4, and 4.2 mol (g
TOC)−1 for a trio of dialysis times (24, 72, and 120 h).
Experiments to establish the standard retention of the freely
dissolved contaminant within the feed solution compartment
were conducted with an initial organic compound concentration in the feed solution of 0.093 M. For experiments containing DOM, initial concentrations in the feed solution were
0.027 M for the organic compounds and 20.1 mg TOC L−1
of tannic acid, corresponding to a compound to DOM ratio
of 1.4 mol (g TOC)−1 . An exception to this procedure was
used for cyclonite, where although the cyclonite to DOM
ratio was maintained (1.4 mol (g TOC)−1 ), the absolute concentrations were doubled to 0.054 M and 39.1 mg TOC L−1 ,
respectively, to improve detection. For studying the impact of
33
the calcium ion on organic transport, calcium concentrations
of 5 and 50 mg L−1 were used.
2.5. Analytical methods
Duplicate samples were collected from each of the three
compartments at various time intervals during the dialysis
period to measure the distribution of the organic compounds
in the system and to perform a mass balance at the end of
each experiment. Contaminant adsorption onto the membranes and other system components in the presence of DOM
was determined by performing a mass balance to calculate the
contaminant mass fraction in the aqueous and solid phases
at the end of the dialysis period. Here, the aqueous phase
is comprised of the feed, permeate, and triolein solutions,
while the solid phase is made up of the feed and triolein
membranes, and other system components depicted in Fig. 1
(i.e., membrane closures, stir bar, wire, and reservoir interior).
Additionally, the membranes and reservoir were rinsed with
methanol (MeOH) to measure the easily mobilized fraction
that is present on each surface. Overall, recovery of the 14 Clabeled contaminants in the experimental system was >98%
in all cases. Using permeate as an example, the mass fraction
of the organic compound attributable to the permeate compartment (Fp ) at the end of the dialysis period was calculated
according to:
Mp
Cp × V p
=
(1)
Fp =
Cf × V f
Mf
where C is the contaminant concentration, M the contaminant mass, V the volume of each respective solution, and the
subscripts f and p designate the feed and permeate solutions,
respectively. Using this approach it is possible to measure the
concentration of contaminant compounds in the various solutions, in addition to the amount of contaminant adsorbed to
the membrane materials [1]. This is significant for this study
in order to explain the mechanisms that govern the removal
of these compounds.
For each experiment, the mass fraction of contaminant
retained or rejected (R) by the feed membrane was calculated
as:
Cr − Cp
R=
× 100
(2)
Cf
where Cr is the contaminant concentration in the feed
solution. The net amount of contaminant remaining in the
feed solution after dialysis (Cr − Cp ) corresponded to the
portion of contaminant in the bulk aqueous phase that was
impermeable on the experimental timescale. On the contrary,
the fraction that diffused across the feed membrane during
the exposure period was operationally defined as “lipidavailable”.
To further characterize the impact of solution chemistry on
contaminant permeability as a function dialysis time a transport coefficient (Tc ) was calculated for each contaminant.
The transport coefficient is defined as the ratio of contam-
S.K. Dalton et al. / Journal of Membrane Science 266 (2005) 30–39
34
inant retention measured under a given set of conditions to
that measured for the freely dissolved contaminant. When Tc
is equal to unity it can be said that the respective chemical
conditions had no net effect on contaminant transport relative to its freely dissolved behavior. Conversely, a Tc greater
than unity suggests that the given chemistry enhanced retention (i.e., suppressed transport), while a value less than unity
indicates reduced retention (i.e., enhanced transport), relative
to the freely dissolved behavior.
3. Results and discussion
Contaminant transport in the experimental dialysis system is discussed in terms of retention by the feed membrane,
distribution, and adsorption within the system, and uptake
into the model lipid phase. Furthermore, these results are
analyzed in terms of the unique chemical properties and subsequent interactions for each contaminant with either tannic
or soil fulvic acid.
The mass fraction (Cr /Cf ) retained in the feed solution for
each contaminant as a function of time is reported in Fig. 2.
Additionally, transport coefficients (Tc ) calculated at t = 24,
72, and 120 h after the start of each experiment are reported
in Table 2 for each contaminant and three different solution
chemistries. Contaminant retention is reported in Fig. 2 for
four different solution chemistries: freely dissolved contaminant, in the presence of a DOM source (tannic or soil fulvic
acid at a contaminant to DOM ratio of 1.4 g mol−1 as TOC),
and in the presence of tannic acid and 5 mg L−1 calcium
nitrate. From Fig. 2, for each contaminant in the freely disTable 2
Calculated transport coefficients for the studied contaminant compounds
under varying solution chemistries
Contaminant
Additive
Transport coefficient (Tc )
24 h
72 h
120 h
Cyclonite
Tannic acid
Tannic acid and calcium
Soil fulvic acid
1.4
1.3
2.6
1.0
1.0
45.0
1.0
1.0
46.5
Atrazine
Tannic acid
Tannic acid and calcium
Soil fulvic acid
1.2
0.6
0.4
3.4
1.0
1.6
3.8
1.2
1.0
DCPo
Tannic acid
Tannic acid and calcium
Soil fulvic acid
1.7
1.7
2.6
1.8
1.2
3.5
1.5
1.2
2.2
dcp−
Tannic acid
Tannic acid and calcium
Soil fulvic acid
1.0
1.7
1.2
1.0
2.2
1.4
1.4
2.4
1.5
Naphthalene
Tannic acid
Tannic acid and calcium
Soil fulvic acid
1.2
1.5
1.9
1.1
1.4
1.6
1.1
1.6
1.6
A contaminant to DOM ratio of 1.4 g mol as TOC−1 was used in each of the
reported values. Calcium was added to the tannic acid and calcium solution at
a concentration of 5 mg L−1 as calcium nitrate. The reported values represent
the mean of no less than three measurements.
solved condition, naphthalene was retained by the feed membrane to the greatest degree over the 120 h test period followed
by dcp− , atrazine, DCPo , and cyclonite, respectively. With
the exception of naphthalene, greater than 98% of the initial
contaminant mass had permeated through the feed membrane
by t = 24 h, independent of the solution chemistry. Furthermore, steady-state conditions were reached by approximately
t = 24 h for nearly all of the contaminants, again independent of solution chemistry. Overall, no trend exists by which
an increasingly larger molecular weight (Table 1) results in
increased retention (Fig. 2).
From the system mass balance, the distribution of each
contaminant following 120 h of dialysis in the presence of
tannic acid is reported in Fig. 3. On average, approximately
90% of the contaminant mass resided in the aqueous phase
after 120 h of dialysis. Conversely, the portion sorbed to the
solid phase was on average approximately 10% of the initial
contaminant mass in agreement with earlier findings [26].
A comparatively weak label in combination with truncation
errors resulted in relatively large standard deviations (average coefficient of variation = 20%) and poor resolution for
cyclonite. However, the error associated with the remaining
compounds was more reasonable with an average coefficient
of variation of 10%.
Although each compound primarily remained in the aqueous phase the specific distribution of each compound was
variable (Fig. 3). For instance, naphthalene was retained in
the feed solution to a much higher degree than any of the other
compounds. Cyclonite, atrazine, and dcp− were all similarly
(±1 standard deviation) present in the permeate solution,
while naphthalene and particularly DCPo were present to a
lesser degree. This difference is attributed to the higher affinity for the triolein phase and greater retention by the feed
membrane for the hydrophobic DCPo and naphthalene compounds, respectively. Both cyclonite and naphthalene were
similarly present (16%) in the solid phase while atrazine,
DCPo , and dcp− were present to a lesser degree.
Retention of freely dissolved cyclonite (Fig. 2a) decreased
from approximately 0.5% at t = 24 h to about 0.02% at
t ≥ 24 h, indicating that it easily diffused across the feed
membrane. The addition of tannic acid to the feed solution enhanced cyclonite retention for the first 24 h (Tc = 1.4);
however, for t ≥ 24 h cyclonite retention closely followed the
freely dissolved behavior as virtually no cyclonite remained
in the feed solution. Similarly, the presence of tannic acid and
calcium initially enhanced cyclonite retention, but at t > 24 h
retention became more similar to that measured for freely dissolved cyclonite (Tc = 1). Soil fulvic acid seemed to increase
cyclonite retention at t ≥ 18 h (Fig. 2a), and resulted in a fourfold increase in Tc at 72 and 120 h (Table 2). Nevertheless,
for each of the solution chemistries studied here the cyclonite
was highly permeable to the membrane.
For atrazine (Fig. 2b), retention doubled in the presence
of tannic acid (Tc = 3.8 at t = 120 h) while inclusion of a small
amount of calcium reversed this trend (Tc = 1.2 at t = 120 h)
and largely restored atrazine’s freely dissolved behavior [21].
S.K. Dalton et al. / Journal of Membrane Science 266 (2005) 30–39
35
Fig. 2. Contaminant mass fraction retained by the feed membrane as a function of dialysis time for (a) cyclonite; (b) atrazine; (c) DCPo ; (d) dcp− ; (e)
naphthalene.
Increasing the calcium concentration by an order of magnitude to 50 mg L−1 (results not shown) did not produce results
significantly different from those obtained at the lower calcium concentration (5 mg L−1 ) (i.e., increased permeability
across the membrane) [21,24]. Soil fulvic acid increased
the initial atrazine permeability (Tc = 0.4 at t = 24 h) before
reaching a similar steady-state value as that measured for
the freely dissolved case. Therefore, in the presence of tan-
nic acid alone, atrazine permeability was reduced, while soil
fulvic acid initially (t < 24 h) increased it.
Similar to the observations for atrazine, tannic acid
enhanced DCPo retention (Fig. 2c), while the addition of
both tannic acid and calcium resulted in reduced retention
relative to that measured for tannic acid alone at t > 24 h
(Table 2). Soil fulvic acid had a greater impact than tannic acid on DCPo retention, as DCPo retention nearly tripled
36
S.K. Dalton et al. / Journal of Membrane Science 266 (2005) 30–39
Fig. 3. Distribution of each organic compound throughout the experimental system for the respective dialysis experiments ([C] = 0.028 M; [tannic
acid] = 20.1 mg TOC L−1 ; T = 20 ◦ C; t = 120 h). The mass fraction of each component was calculated using a mass balance of the 14 C-labeled contaminants in
the experimental batch system. The mean values are reported with their associated standard deviations (n = 3).
in the presence of soil fulvic acid. A subsequently different retention was observed for the negatively charged dcp−
compound (Fig. 2d). Retention of dcp− in the presence of
tannic acid did not deviate from that seen for the freely dissolved condition (i.e., Tc = 1) at t < 72 h. However, at t > 80 h
contaminant retention increased (Tc = 1.4 at t = 120 h), which
was similar to that calculated for DCPo . Therefore, over the
timescale studied here (t = 120 h) any difference in the interaction between DCPo and dcp− and tannic acid, as they
pertain to their rejection by the feed membrane, decreases
with time. However, dcp− retention virtually doubled in the
presence of both tannic acid and calcium for t > 10 h (Table 2).
This contrasts to what was seen for the neutral DCPo , where
retention only increased slightly (Tc = 1.2 at t = 120 h). Soil
fulvic acid only marginally, and to a lesser extent than for
DCPo , decreased (Tc = 1.5) dcp− permeability compared to
that measured for both the freely dissolved and tannic acid
conditions (Tc = 1.4).
Of the five contaminants, non-polar naphthalene exhibited the greatest overall retention in the feed solution, with
values ranging from 12 to 19% at t = 120 h (Fig. 2e). Tannic
acid alone weakly impacted naphthalene retention (Tc = 1.1
at t = 120 h); however, the addition of both tannic acid and
calcium resulted in a more substantial increase in retention (Tc = 1.6 at t = 120 h). Similarly, more naphthalene was
retained in the feed solution in the presence of soil fulvic
acid than for tannic acid (Table 2). Mass balance data for
naphthalene indicated that sorption to the membranes only
accounted for about 10% of its distribution in the batch system, so that resistance to mass transfer appears to partially
inhibit naphthalene transport into and across the membrane.
A mass balance of the system revealed that 8.5% of the naphthalene was adsorbed within the feed membrane matrix after
120 h of dialysis. This represented the highest sorption value
recorded for any of the contaminants examined in this work
(Fig. 3).
3.1. Impact of contaminant properties on transport
The role of contaminant polarity and hydrophobicity, as
well as the ratio of contaminant to DOM, are discussed in
more detail to elucidate their specific roles in determining
contaminant permeability to the membrane.
3.2. Role of contaminant polarity
In the aqueous phase, both protonated and ionized forms
of ionogenic compounds such as weak organic acids and
bases may be present at pH levels that are characteristic of
natural and engineered systems. Furthermore, for similarly
charged surfaces solute rejection increases as membrane and
solute ionization increases [7,9]; conversely, for oppositely
charged surfaces solute adsorption and permeability to the
membrane may increase as the magnitude of the surface
charge increases. Nevertheless, the vast majority of aqueous species and membrane surfaces are similarly negatively
charged under typical pH and ionic strengths making the first
scenario the more likely of the two possibilities. However,
very similar molecules, differing only by the placement of the
acidic group on the molecule, may be removed to a remarkably different degree although both are similarly charged or
dissociated. Thus, the ionogenic nature of the contaminant
must be considered in assessing its transport across a membrane. In general, when organic matter is dissolved in water
at neutral pH and low ionic strength its acidic functional
groups dissociate allowing for hydrogen bond formation with
other polar molecules, such as water. Additionally, charge
repulsion yields a configuration that makes these negatively
S.K. Dalton et al. / Journal of Membrane Science 266 (2005) 30–39
charged sites relatively accessible to organic compounds
exhibiting some polarity. However, as the concentration of
calcium ions is increased the DOM macromolecule assumes
a more compact configuration so that these functional moieties become primarily accessible only along the exterior of
the macromolecule. The significance of surface charge in
determining contaminant behavior in aqueous solutions is
evident from the distinctly different distributions of DCPo
and dcp− in the experimental system (Fig. 3). After 120 h,
more than three times as much contaminant was absorbed
into the lipid phase for the undissociated form (68.8% for
DCPo versus 17.5% for dcp− ). Thus, solution chemistry is
likely to play a role in determining the interaction between
the membrane and contaminant as well as those between the
DOM and the contaminant. The distinct behavior of atrazine
and dcp− in the presence of DOM appears to reflect such
electrolyte-mediated effects due to their unique charge characteristics (Table 1).
Atrazine is a moderately polar (α = 2.8 D), weakly basic
(pKa = 1.7) compound that is primarily present in natural and
engineered water systems in a charged state. Atrazine interacts with natural organic matter primarily through hydrogen
bonding. Resonant structures of the basic species indicate
that lone-electron pairs can be delocalized across the triazine
ring to generate a transient positive charge at either or both of
the side-chain nitrogens [21], which in turn, can interact with
anionic functional groups on DOM macromolecules. This
may explain the increased retention of atrazine by the feed
membrane in the presence of tannic acid, as atrazine forms
complexes within the tannic acid macromolecule, which is
more easily rejected by the membrane through a combination of charge repulsion and size exclusion (Fig. 2b). However, the results further suggest that atrazine ineffectively
competes with calcium ions for sorption sites on the DOM
macromolecule so that in the presence of both DOM and
calcium, the DOM takes on a more compact configuration.
Any remaining sites on the DOM for atrazine interaction are
likely to be exterior and highly reversible sites and therefore,
the atrazine remains freely dissolved. Thus, similar retentions
of the atrazine were observed in its freely dissolved phase and
in the DOM–calcium system (Fig. 2b).
Dissociation of 2,4-dichlorophenol’s hydroxyl group
(pKa = 7.85) yields the 2,4-dichlorophenolate anion, dcp− .
Whereas atrazine is positively charged, dcp− carries a net
negative charge. Consequently, in the absence of calcium,
the interaction between dcp− and tannic acid is inhibited by
charge repulsion between their respective anionic functional
groups [25]. In the presence of tannic acid alone, dcp− retention closely followed that of freely dissolved dcp− . However,
in the presence of both calcium and tannic acid, dcp− retention increased. The addition of calcium to the system may
affect dcp− /DOM interactions via a number of scenarios:
(i) charge shielding or neutralization of functional groups
on the tannic acid macromolecule, thus yielding a more
hydrophobic entity; (ii) association of calcium with dcp−
based on charge neutralization; or (iii) a combination of both,
37
in which calcium acts as a bridge to join dcp− and the tannic
acid macromolecule, which both physically hinders transport
across the membrane as well as decreases aqueous diffusivity
of the contaminant due to a significant increase in its apparent
molecular weight.
3.3. Role of contaminant hydrophobicity
Contaminant hydrophobicity also appears to play some
role in determining contaminant retention, however, the
relationship is not straightforward. For instance, both the
most hydrophobic (naphthalene) and the most hydrophilic
(cyclonite) contaminants had the highest and lowest retention
values, respectively. Furthermore, the increased retention of
naphthalene occurred despite it having the lowest molecular
weight (MW = 128 g mol−1 ), while cyclonite had the lowest retention despite being the heaviest (MW = 222 g mol−1 )
of the contaminants studied here (Table 1). Also, a higher
amount of the hydrophobic naphthalene was sorbed to the
feed membrane than the hydrophilic cyclonite, though the
difference was not as substantial as would be anticipated
based on the differences in hydrophobicity alone (Fig. 3).
Similarly, sorption to the membrane structure was greater for
the more hydrophobic DCPo (7.3%) compared to that for the
more hydrophilic dcp− (3.4%) (Fig. 3).
In contrast, both cyclonite and DCPo were similarly
retained by the feed membrane despite their differences in
hydrophobicity, molecular weight, and polarity. The retention of these two contaminants (Fig. 2a and c, respectively)
was virtually independent of changes in the feed solution
chemistry (e.g., the presence of DOM or calcium), suggesting
that association with DOM and/or calcium, or modifications
of DOM conformation and hydrophobicity through calcium
adsorption, did not impact transport of these two compounds
across the feed membrane.
3.4. Contaminant to DOM ratio
The retention of the five species into the lipid phase was
investigated for a series of contaminant to tannic acid ratios: 0,
0.7, 1.4, and 4.2 mol (g TOC)−1 . Retention of both cyclonite
and DCPo was independent of the contaminant to tannic
acid ratio studied, varying less than 0.5% for each species.
On the other hand, retention of atrazine, naphthalene, and
dcp− was more sensitive to the ratio employed. As the ratio
increased, contaminant retention decreased indicating that
the abundance of interaction sites on the organic matter may
be limiting. In support of this observation, at the highest contaminant to tannic acid ratio used (4.2 mol (g TOC)−1 ) (i.e.,
the lowest relative amount of tannic acid) the retention of
atrazine approximated its freely dissolved behavior: 1.4 and
1.5%, respectively.
The methodology used in this research involved a static
batch system in which a single dosage of source solution
was applied prior to dialysis, thus representing a pulse of
contaminant under systematically controlled conditions. Fur-
38
S.K. Dalton et al. / Journal of Membrane Science 266 (2005) 30–39
thermore, the timescale of these experiments was on the order
of a few days. At this timescale, changes to the source solution appear to have a relatively small impact on contaminant
permeability; for instance, after 120 h retention as a function
of source solution differed by less than 3% for each contaminant (Fig. 2). In contrast, conventional membrane processes
operate with a relatively continuous source of contaminant
in a flow-through system over a considerably shorter contact period (e.g., minutes to hours). Under such conditions,
it is expected that the trends observed using the static batch
system in this work would be magnified and more highly contrasted. For instance, Devitt et al. [27] simulated flow-through
conditions using a pressure-driven dead-end filtration technique to demonstrate similar, though magnified trends for
atrazine in the presence of tannic acid.
4. Conclusions
As observed by previous investigators, the membrane
permeability of low-molecular weight compounds (<250
g mol−1 ) in the absence of dissolved macromolecules is
not directly related to molecular weight of the compounds.
The polarity/polarizability and hydrophobicity of these compounds appear to play a more important role in determining
rejection. However, in the presence of DOM, polarity and
hydrophobicity were not unique determinants of contaminant
behavior. Relatively hydrophilic or hydrophobic compounds
may, in some case display little to no affinity for DOM. Where
compounds interact with DOM, elements of the solution
chemistry that affect ionizable groups may influence contaminant transport in several ways. The presence of calcium may
enhance contaminant/DOM interactions and reduce transport
of contaminants across membranes in the presence of DOM
by increasing DOM hydrophobicity (e.g., naphthalene) or by
bridging functional groups (e.g., dcp− ).
Naturally occurring macromolecular material such as fulvic acids may reduce the transport of some contaminants
across synthetic membranes. Based on the model DOM systems used in this work, we speculate that when a reduction
in contaminant permeability does occur, the effect of DOM
should be relatively small (<5%). The pH and the presence
of divalent cations are important factors in determining the
degree to which permeability of a given contaminant may be
reduced in the presence of DOM. Solution chemistry appears
to play a greater role in influencing the kinetics of mass transfer than the equilibrium behavior, as the effects of DOM on
transport observed in this work tended to be smaller as membrane systems approached steady-state with solutions.
Acknowledgements
This work was supported primarily by a grant from the
Hazardous Substance Research Center-South and Southwest.
Additional financial support was provided by a National
Science Foundation Graduate Research Trainee Fellowship.
The authors gratefully and wholeheartedly thank Mark Hoff,
Sebastien Follet, Dr. Mason B. Tomson, and Dr. Jean-Yves
Bottero for their contributions to this research as well as the
European Centre for Geosciences and Environment Research
and Education (CEREGE) of the Centre National de la
Recherche Scientifique for their hospitality during the preparation of this manuscript.
References
[1] R. Boussahel, et al., Removal of pesticide residues in water using
the nanofiltration process, Desalination 132 (2000) 205–209.
[2] B. van der Bruggen, C. Vandecasteele, Removal of pollutants from
surface water and groundwater by nanofiltration: overview of possible applications in the drinking water industry, Environ. Pollut. 122
(2003) 435–445.
[3] B. van der Bruggen, et al., Influence of molecular size, polarity and
charge on the retention of organic molecules by nanofiltration, J.
Membr. Sci. 156 (1999) 29–41.
[4] J.A.M.H. Hofman, et al., Removal of pesticides and other micropollutants with cellulose–acetate, polyamide and ultra-low pressure
reverse osmosis membranes, Desalination 113 (1997) 209–214.
[5] C. Bellona, et al., Factors affecting the rejection of organic solutes
during NF/RO treatment—a literature review, Water Res. 38 (2004)
2795–2809.
[6] Y. Kiso, et al., Effects of hydrophobicity and molecular size on
rejection of aromatic pesticides with nanofiltration membranes, J.
Membr. Sci. 192 (2001) 1–10.
[7] J. Cho, G. Amy, J. Pellegrino, Membrane filtration of natural organic
matter: initial comparison of rejection and flux decline characteristics
with ultrafiltration and nanofiltration membranes, Water Res. 33 (11)
(1999) 2517–2526.
[8] A. Seidel, M. Elimelech, Coupling between chemical and physical
interactions in natural organic matter (NOM) fouling of nanofiltration
membranes: implications for fouling control, J. Membr. Sci. 203
(2002) 245–255.
[9] J. Cho, et al., Characterization of clean and natural organic matter
(NOM) fouled NF and UF membranes, and foulants characterization,
Desalination 118 (1998) 101–108.
[10] J. Cho, et al., Predictive models and factors affecting natural organic
matter (NOM) rejection and flux decline in ultrafiltration (UF) membranes, Desalination 142 (2002) 245–255.
[11] L.J. Shaw, et al., Bioavailability of 2,4-dichlorophenol associated
with soil water-soluble humic material, Environ. Sci. Technol. 34
(22) (2000) 4721–4726.
[12] R.L. Wershaw, The importance of humic substance–mineral particle complexes in the modeling of contaminant transport in
sediment–water systems, in: R.A. Baker (Ed.), Organic Substances
and Sediments in Water: Humics and Soils, Lewis Publishers,
Chelsea, MI, 1991, pp. 23–34.
[13] F.A.P.C. Gobas, X. Zhang, Interactions of organic chemicals with
particulate and dissolved organic matter in the aquatic environment,
in: J.L. Hamelink, et al. (Eds.), Bioavailability: Physical, Chemical,
and Biological Interactions, Lewis Publishers, Boca Raton, FL, 1994,
pp. 83–91.
[14] S.M. Schrap, A. Opperhuizen, Relationship between bioavailability
and hydrophobicity: reduction of the uptake of organic chemicals
by fish due to the sorption on particles, Environ. Toxicol. Chem. 9
(1990) 715–724.
[15] I.H. Suffet, et al., Influences of particulate and dissolved material
on the bioavailability of organic compounds, in: J.L. Hamelink, et
al. (Eds.), Bioavailability: Physical, Chemical and Biological Interactions, Lewis Publishers, Boca Raton, FL, 1994, pp. 93–108.
S.K. Dalton et al. / Journal of Membrane Science 266 (2005) 30–39
[16] J.N. Huckins, et al., Semipermeable membrane devices (SPMDs) for
the concentration and assessment of bioavailable organic contaminants in aquatic environments, in: G.K. Ostrander (Ed.), Techniques
in Aquatic Toxicology, CRC-Lewis Publishers, New York, NY, 1996,
pp. 625–655.
[17] G.S. Ellis, et al., Evaluation of lipid-containing semipermeable membrane devices for monitoring organochlorine contaminants in the
upper Mississippi River, Environ. Toxicol. Chem. 14 (11) (1995)
1875–1884.
[18] P.L. Huyskens, et al., Dipole moments and complexation enthalpies
of 1-methyl- and 1,2-dimethylimidazole with various phenols in benzene solution. Angle between the moment of the bases and the
hydrogen bond, J. Phys. Chem. 84 (21) (1980) 2748–2751.
[19] J.N. Israelachvili, Intermolecular and Surface Forces, second ed.,
Academic Press Harcourt Brace Jovanovich, London, 1992, pp. 450.
[20] C.T. Chiou, Partition coefficients of organic compounds in
lipid–water systems and correlations with fish bioconcentration factors, Environ. Sci. Technol. 19 (1) (1985) 57–62.
[21] E.C. Devitt, M.R. Wiesner, Dialysis investigations of atrazine–
organic matter interactions and the role of a divalent metal, Environ. Sci. Technol. 32 (2) (1998) 232–237.
[22] J.N. Huckins, et al., Lipid-containing semipermeable membrane
devices for monitoring organic contaminants in water, Environ. Sci.
Technol. 27 (12) (1993) 2489–2496.
[23] Y.X. Wang, et al., Monitoriing priority pollutants in a sewage
treatment process by dichloromethane extraction and troleinsemipermeable membrane device (SPMD), Chemosphere 43 (2001)
339–346.
[24] C.E. Clapp, et al., A quantitative estimation of the complexation of
small organic molecules with soluble humic acids, J. Environ. Qual.
26 (1997) 1277–1281.
[25] F. De Paolis, J. Kukkonen, Binding of organic pollutants to humic
and fulvic acids: influence of pH and the structure of humic material,
Chemosphere 34 (8) (1997) 1693–1704.
[26] P.J. Woolgar, K.C. Jones, Studies on the dissolution of polycyclic
aromatic hydrocarbons from contaminated materials using a novel
dialysis tubing experimental method, Environ. Sci. Technol. 33 (12)
(1999) 2118–2126.
39
[27] E.C. Devitt, et al., Effects of natural organic matter and the raw water
matrix on the rejection of atrazine by pressure-driven membranes,
Water Res. 32 (9) (1998) 2563–2568.
[28] D. Mackay, W.-Y. Shiu, K.-C. Ma, 2,4-Dichlorophenol, in: D.
Mackay, W.-Y. Shiu, K.-C. Ma (Eds.), Illustrated Handbook of
Physical–Chemical Properties and Environmental Fate for Organic
Chemicals, vol. IV. Oxygen, Nitrogen, and Sulfur Containing Compounds, CRC-Lewis Publishers, New York, NY, 1995, pp. 351–
355.
[29] D. Mackay, W.-Y. Shiu, K.-C. Ma, Atrazine and pentachlorophenol,
in: D. Mackay, W.-Y. Shiu, K.-C. Ma (Eds.), Illustrated Handbook
of Physical–Chemical Properties and Environmental Fate for Organic
Chemicals. vol. V. Pesticide Chemicals, CRC-Lewis Publishers, New
York, NY, 1997, 67–75, 525–532.
[30] R.P. Schwarzenbach, P.M. Gschwend, D.M. Imboden, Environmental
Organic Chemistry, John Wiley and Sons, New York, NY, 1993, pp.
681.
[31] K.E. Calderbank, R.K. Pierens, Molecular polarisability. The conformation of N,N -dinitropiperazine and other cyclic nitroamines in
dioxan solution, J. Chem. Soc. Perkin Trans. II 6 (1979) 869–
871.
[32] V.N. Kavetskii, L.I. Bublik, Use of thin-layer chromatography for
determining the dipole moment of organic compounds, Russ. J. Phys.
Chem. 63 (4) (1989) 561–563.
[33] H. Lumbroso, J. Curé, C.G. Andrieu, The preferred conformations of
2,4-dichlorophenol, 2,4-dichloroanisole and their sulphur analogues,
J. Mol. Struct. 43 (1) (1978) 87–95.
[34] H.H. Huang, S.C. Ng, The molar Kerr constants and anisotropic
polarisabilities of certain sterically hindered aromatic nitrocompounds, J. Chem. Soc. B: Phys. Org. 5 (1968) 582–587.
[35] J. Singh, Natural and accelerated detoxification of RDX and atrazine
in contaminated soil and water. Ph.D. Dissertation, University of
Nebraska, Lincoln, NE, 1997.
[36] G.L. Huang, et al., Effects of pH on the aqueous solubility of selected
chlorinated phenols, J. Chem. Eng. Data 45 (3) (2000) 411–414.
[37] B.I. Escher, R.P. Schwarzenbach, Partitioning of substituted phenols
in liposome–water, biomembrane–water, and octanol–water systems,
Environ. Sci. Technol. 30 (1) (1996) 260–270.