Published May, 2012
Journal of Environmental Quality
EMERGING TECHNOLOGIES TO REMOVE NONPOINT P SOURCES FROM SURFACE WATER AND GROUNDWATER
Factors Controlling Phosphate Interaction with Iron Oxides
Liping Weng,* Willem H. Van Riemsdijk, and Tjisse Hiemstra
Factors such as pH, solution ion composition, and the presence
of natural organic matter (NOM) play a crucial role in the
effectiveness of phosphorous adsorption by iron oxides. The
interplay between these factors shows a complicated pattern
and can sometimes lead to controversial results. With the help
of mechanistic modeling and adsorption experiments, the net
macroscopic effect of single and combined factors can be better
understood and predicted. In the present work, the relative
importance of the above-mentioned factors in the adsorption of
phosphate was analyzed using modeling and comparison between
the model prediction and experimental data. The results show
that, under normal soil conditions, pH, concentration of Ca, and
the presence of NOM are the most important factors that control
adsorption of phosphate to iron oxides. The presence of Ca not
only enhances the amount of phosphate adsorbed but also changes
the pH dependency of the adsorption. An increase of dissolved
organic carbon from 0.5 to 50 mg L−1 can lead to a >50% decrease
in the amount of phosphate adsorbed. Silicic acid may decrease
phosphate adsorption, but this effect is only important at a very
low phosphate concentration, in particular at high pH.
Copyright © 2012 by the American Society of Agronomy, Crop Science Society
of America, and Soil Science Society of America. All rights reserved. No part of
this periodical may be reproduced or transmitted in any form or by any means,
electronic or mechanical, including photocopying, recording, or any information
storage and retrieval system, without permission in writing from the publisher.
J. Environ. Qual. 41:628–635 (2012)
doi:10.2134/jeq2011.0250
Posted online 13 Jan. 2012.
Received 14 July 2011.
*Corresponding author (liping.weng@wur.nl).
© ASA, CSSA, SSSA
5585 Guilford Rd., Madison, WI 53711 USA
628
SPECIAL SECTION
TECHNICAL REPORTS
E
nhanced primary production in surface water (i.e.,
eutrophication) has become a worldwide problem that
results in algal blooms, the development of harmful
cyanobacteria, fish kills, smell nuisance, and other detrimen-
tal consequences (Carpenter et al., 1998). In the past decades,
considerable reduction of P loss has been achieved thanks to
measures to control industrial release and sewage water treat-
ment. As a result, the relative importance of P release from
agricultural fields has increased in recent years (Barberis et al.,
1996; Chardon and Koopmans, 2005; Koopmans et al., 2007;
Sharpley et al., 1994).
Various strategies and measures have been proposed, tested,
and implemented to reduce release of P from agricultural fields,
including input reduction and leaching prevention (Kronvang
et al., 2005; Sharpley et al., 1994; Withers et al., 2000). In
agriculture fields, to guarantee sufficient nutrient supply for
crop production, a solution concentration of above 3 mg L−1
is required, and P concentrations as high as 30 mg L−1 can be
found in soil runoff (Withers et al., 2000). The P concentra-
tion in leachate from agricultural fields is much higher than the
target value for P in surface water (<0.1 mg L−1). One attrac-
tive approach to reduce leaching is to use P-fixing materials
that are directly mixed with soil or used to build a reactive
barrier (Codling et al., 2000; McDowell et al., 2008; Stout
et al., 2000). Iron-rich materials are good candidates to build
P-removing filters (McDowell et al., 2008; Zeng et al., 2004;
Zhang et al., 2009). Such materials can be metal iron (e.g.,
zerovalent iron) (McDowell et al., 2008; Sakadevan and Bavor,
1998) and iron oxides in various forms (Boujelben et al., 2008;
Pan et al., 2010; Zeng et al., 2004).
Phosphorus in soil drainage water consists of orthophos-
phate, organic P, and P associated with mineral colloidal particles
(adsorbed or precipitated). A permeable reactive filter/barrier
containing iron oxides can, in principle, remove all three forms
of P. The mineral colloidal P can be intercepted by filtration
effects, and orthophosphate and organic P can be immobilized
by adsorption reactions. The efficiency and effective lifespan
of a P barrier depends on various factors, including the basic
characteristic of the adsorptive material itself (e.g., the porosity,
reactive surface area, P-binding affinity, and binding capacity)
and environmental factors (e.g., pH, ion composition, the pres-
ence of other compounds, and water flow). The reactive groups
present on the surface of iron oxides can bind various cations
Dep. of Soil Quality, Wageningen Univ., P.O. Box 47, 6700 AA, Wageningen, The
Netherlands. Assigned to Associate Editor Ray Bryant.
Abbreviations: DOC, dissolved organic carbon; FA, fulvic acid; LCD, ligand and
charge distribution; NOM, natural organic matter.
and anions. The chemical reactions at the surface become more
difficult to predict the more the ions and compounds interact.
The complete set of chemical reactions that may take place in the
field situation leads to a very complex system.
In adsorption reactions, when equilibrium is maintained,
the relation between solution concentration of phosphate (C,
mol L−1) and the corresponding loading on the surface (Γ, mol
kg−1) can be expressed using the Langmuir model:
K 'C
Γ = Γ 'max [1]
1 + K 'C
where Γ′max is the apparent maximum binding capacity (mol
kg−1) and K′ the conditional affinity constant. For a reactive
barrier, the increase of the P loading equals the net amount of
P added per unit adsorptive material; this is equal to V(C0 –
C), where V (L kg−1) is the volume of solute drained expressed
per kg of adsorptive material, and C0 and C are the input and
output P concentrations, respectively. If the latter is small, the
average loading of the reactive material can be approached by
VC0. Under this condition, Eq. [1] can be written as
K 'C
Γ = VC 0 = Γ 'max [2]
1 + K 'C
Rearrangement of Eq. [2] results in
VC 0
C= [3]
(Γ 'max −VC0 ) K '
To maintain the C under a certain target value, the maxi-
mum amount of drainage that can be treated depends strongly
on both Γ′max and K′. In a mechanistic model, the maximum
binding capacity depends on the amount and type of iron
oxides in the material, the particle size (surface area), and acces-
sibility of the surface, i.e., Γmax is characterized by the mate-
rial. When this type of Langmuir model is derived empirically,
the apparent Γ′max also depends on factors that may promote
adsorption (like Ca2+) or decrease the adsorption due to com-
petition with natural organic matter (NOM). The conditional
affinity constant K′ depends on the type of reactive groups
present on the surface (a character of the material) and is also
strongly affected by the solution chemistry (e.g., pH and pres-
ence of other adsorbing compounds).
The objective of this study was to apply the most recent
developments in the understanding of ion and NOM adsorp-
tion to iron oxides to gain insight into the most relevant factors
that influence the performance of P-filters under field condi-
tions. The relative importance of pH, major cations (Ca2+ and
Mg2+), anions (HCO3− and SO42−), neutral species (H4SiO40),
and NOM in phosphate adsorption were analyzed using mech-
anistic modeling and the model prediction were compared
with data of an adsorption experiment. We show that modern
surface complexation models such as the CD-MUSIC model
(Hiemstra and Van Riemsdijk, 2006) can be applied to explain
and predict the net effects of various factors on the quality and
capacitance of a barrier.
Materials and Methods
Batch experiments were performed to study the effects
of pH, Ca, and NOM on phosphate adsorption to goethite.
www.agronomy.org • www.crops.org • www.soils.org
The goethite material was prepared based on the procedure
described by Hiemstra et al. (1989). The specific surface area
of the goethite material was 94 m2 g−1, as determined using
the BET-N2 adsorption method (Weng et al., 2005). Fulvic
acid (FA) was used to simulate NOM, and the FA material
was purified from a peat soil (Strichen Soil Assoc., Scotland)
following the International Humic Substance Society proce-
dures. The purified FA material contains 43% C, and its aver-
age molar mass is 0.68 kDa, which was measured using size
exclusion chromatography.
This study included three series of treatments: (i) no FA,
(ii) 150 mg L−1 FA (64.5 mg C L−1), and (iii) 300 mg L−1 FA
(129 mg C L−1). In all series, the suspension contained final
total concentrations of 2.7 g L−1 goethite, 0.01 CaCl2, and 0.48
mmol L−1 NaH2PO4. The solutions were added immediately
after each other in the order CaCl2, goethite, NaH2PO4, and
FA. For all series, acid or base solution (0.01 mol L−1 NaOH
and 0.05 mol L−1 HCl) was added to the suspension to adjust
the pH to selected values in the range of pH 3 to 7. During
the preparation, the gas-tight vessels used in the experiments
were flushed with N2 gas to minimize the influence of CO2.
The prepared suspensions were shaken at 20°C for 4 d and
centrifuged at 18,000 g for 30 min. The final pH in the super-
natant was measured using a pH meter. The P concentration
in the supernatant was measured with ICP–MS. The Ca con-
centration was measured with ICP–AES. The concentration of
dissolved organic carbon (DOC) was measured with a TOC
analyzer. The amounts of P, Ca, and FA adsorbed were derived
from the difference between the treatments and the blank (no
goethite). The small contribution of organic P (P in FA) to the
total soluble P measured is corrected using P content in FA
measured in a FA solution (0.16 μg mg−1) and DOC. The final
DOC concentration ranged from 15 to 75 mg L−1.
Modeling
CD-MUSIC Model
Iron (hydr)oxides can occur in nature in various crystal forms,
such as goethite (α-FeOOH), hydrous ferric oxide, or ferrihy-
drite, hematite (α-Fe2O3), and lepidocrocite (γ-FeOOH). All
these solid phases have a strong affinity for phosphate. From
a more fundamental point of view, much more is known with
respect to the adsorption of phosphate to goethite, whereas
much less information is available for the other phases men-
tioned. Goethite with varying crystal morphology shows similar
binding properties for phosphate (Torrent et al., 1990). It has
been shown (Hiemstra et al., 2010a) that phosphate desorption
data for various soils can be described well by assuming that the
reactive metal oxide phase in soils can be interpreted based on
the well calibrated adsorption model for goethite, although the
process may involve several metal oxide phases and clay miner-
als (Devau et al., 2009, 2011).
The surface of the iron oxides consist of a few different types
surface oxygen groups that differ in their degree of coordina-
tion with the iron in the solid. These oxygens have a variable
number of protons associated with them. Due to breaking
of bonds, surface groups with a coordination number that is
less than the maximum can occur. The formal charge of the
various surface groups can be derived using a simple Pauling
629
bond valence analysis or bond valence theory. In the Pauling
approach, one assigns a formal charge of 0.5 (3/6) per Fe bond
to the oxygen. This approach is used in the MUSIC (Multi Site
Complexation) model (Hiemstra et al., 1989), which, in com-
bination with the CD (Charge Distribution) model (Hiemstra
and Van Riemsdijk, 1996), forms the basis of ion adsorption
models for oxides. The CD-MUSIC model is based as much as
possible on available molecular level information with respect
to the surface speciation.
The application of the MUSIC model to various minerals
shows that sites that have a very high or very low value of the
formal charge are not present at the surface to any significant
extent. In other words, such surface species are thermodynami-
cally not very stable. For the singly coordinated surface oxygens
on goethite, the amount of the ≡FeO−1.5 site is insignificant,
and the major surface species are thus the ≡FeOH−0.5 and the
≡FeOH2+0.5 species, whereas for the triply coordinated oxygen
the relevant species are ≡Fe3OH+0.5 and ≡Fe3O−0.5. The doubly
coordinated oxygens have a high affinity for protons, which
leads to the formation of a stable uncharged species (≡Fe2OH0).
In the MUSIC model, the basic charging of the mineral goe-
thite is therefore assumed to be caused by the protonation and
deprotonation of the singly (≡FeOH−0.5) and triply (≡Fe3O−0.5)
coordinated surface oxygens:
≡FeOH−0.5 + H+ = ≡FeOH2+0.5 log KH,1 [4]
≡Fe3O−0.5 + H+ = ≡Fe3OH+0.5 log KH,2 [5]
Depending on environmental conditions such as pH, metal (hydr)
oxides can be positively charged, negatively charged, or uncharged.
Based on the structure analysis, the overall effective site
density of goethite is set at 3.45 site nm−2 for the singly coor-
dinated oxygen and 2.7 site nm−2 for the triply coordinated
oxygen (Table 1). The (refined) MUSIC model also allows for
an a priori estimate of the logK values of the various proton-
ation reactions for various minerals. In practice, the model for
goethite can be simplified by assuming that the proton affin-
ity constants for the two types of reactive sites are equal. The
intrinsic proton affinities (logKH,1 and logKH,2) follow in that
case directly from the point of zero charge of goethite (Table 1).
In the CD model, the charge of an adsorbing ion is dis-
tributed over two electrostatic planes. The advantage of this
approach is that the electrostatic energy component to the
binding is now a combination of two terms, one for each
plane. The CD approach is a practical way to account for the
fact that there is a gradient of the electrostatic potential over
the adsorbed ion. It is also an advantage that the value of the
CD parameter is linked to the structure of the surface com-
plex. The simplest electrostatic model that takes the structure
of the inner-sphere surfaces into account is a CD approach

Table 1. CD-MUSIC model parameters for goethite.†

PZC‡ Site density Capacitance
≡FeOH−0.5 ≡Fe3O−0.5 C1 C2
9.0
Surface species 3.45 nm−2 2.7 nm−2 0.93 F m−2 0.75 F m−2
Sites Charge distribution§
Ions logK
≡ FeOH−0.5 ≡ Fe3O−0.5 Δz0 Δz1 Δz2
≡FeOH2+0.5 1 0 1 0 0 1H+ 9.00
≡Fe3OH+0.5 0 1 1 0 0 1H+ 9.00
≡FeOH…Na+0.5 1 0 0 1 0 1Na+ −0.60
≡Fe3O…Na+0.5 0 1 0 1 0 1Na+ −0.60
≡FeOH…K+0.5 1 0 0 1 0 1K+ −1.61
≡Fe3O…K+0.5 0 1 0 1 0 1K+ −1.61
≡FeOHCa+1.5 1 0 0.31 1.69 0 1Ca2+ 3.23
≡FeOHCaOH+0.5 1 0 0.31 0.69 0 −1H+, 1Ca2+ −6.42
≡FeOH…Ca+1.5 1 0 0 2 0 1Ca2+ 1.8
≡Fe3O…Ca+1.5 1 0 0 2 0 1Ca2+ 1.8
≡(FeOH)2Mg+1.0 2 0 0.72 1.28 0 1Mg2+ 4.52
≡(FeOH)2MgOH0 2 0 0.72 0.28 0 −1H+, 1Mg2+ −6.78
≡FeOH2…Cl−0.5 1 0 1 −1 0 1H+, 1Cl− 8.55
≡Fe3OH…Cl−0.5 0 1 1 −1 0 1H+, 1Cl− 8.55
≡(FeO)2CO−1.0 2 0 0.62 −0.62 0 2H+, 1CO32− 22.01
≡(FeO)2CO…Na0 2 0 0.62 0.38 0 2H+, 1Na+,1CO32− 22.03
≡FeOH2…CO3−1.5 1 0 1 −2 0 1H+, 1CO32− 10.22
≡Fe3OH…CO3−1.5 0 1 1 −2 0 1H+, 1CO32− 10.22
≡(FeO)2Si(OH)2.−1.0 2 0 0.29 −0.29 0 1H4SiO40 5.85
≡(FeO)2SiOHOSi3O2(OH)7.−1.0 2 0 0.29 −0.29 0 1H4SiO40 13.98
≡(FeO)2SiOHOSi3O3(OH)6.−2.0 2 0 0.29 −1.29 0 −1H+, 1H4SiO40 7.47
≡(FeO)2PO2−2.0 2 0 0.46 −1.46 0 2H+, 1PO43− 29.72
≡FeOPO2OH−1.5 1 0 0.28 −1.28 0 2H+, 1PO43− 27.63
† Adapted from Hiemstra et al., 2007, 2010a.
‡ Point of zero charge.
§ Δz0, charge attributed to 0-plane; Δz1, charge attributed to 1-plane; Δz2, charge attributed to 2-plane.

630 Journal of Environmental Quality

with two electrostatic planes, a CD-Basic Stern model. The
electrostatic approach has recently been further refined by con-
sidering the fact that ions adsorbed as outer-sphere complexes
can have different sizes, which means the electrostatic energy
component for different size ions is also different (Hiemstra
and Van Riemsdijk, 2004; Hiemstra et al., 2004; Rahnemaie
et al., 2006). In this case, an Extended Stern model is required
(Hiemstra and Van Riemsdijk, 2006).
Phosphate ions can form surface complexes by reacting chemi-
cally with one or two singly coordinated surface groups (≡FeOH−
0.5
). The group that forms the bond is an oxygen ion in both cases
(Fig. 1). Originally, the adsorption behavior of phosphate was
described in the CD-MUSIC model using three surface species: a
monodentate, a bidentate, and a protonated bidentate. Recently,
Rahnemaie et al. (2007) reinterpreted the results using an exten-
sive data set and concluded that P adsorption can be described
using two inner-sphere complexes: the bidentate and protonated
monodentate. The formation reaction for the dominant bidentate
surface phosphate species can be written as:
2≡FeOH−0.5 + PO43− + 2H+ = (≡FeO)2PO22− + 2H2O [6]
and formation reaction for the less important monodentate
species can be written as:
≡FeOH−0.5 + PO43− + 2H+ = ≡FeOPO2OH−1.5 + 2H2O [7]
The mechanistic nature of the CD-MUSIC model allows pre-
diction of effects of pH and the synergistic and competitive
effects of cations and other anions. Model parameters for ions
that are most relevant in a soil environment are available in the
literature. The parameter set used in the following calculation
was adopted from Hiemstra et al. (2010b), except those for
silicic acid adsorption, which came from Hiemstra et al. (2007)
(Table 1).
Modeling Natural Organic Matter Effects
In natural systems, NOM is usually present. These organic
molecules also interact with iron oxides (Weng et al., 2006a)
and other soil minerals by forming an organic “coating” on these
mineral surfaces. It has been demonstrated that adsorption of
NOM strongly influences phosphate adsorption to goethite
(Bhatti et al., 1998; Borggaard et al., 1990; Hiemstra et al.,
Fig. 1. Representation of the structure of the protonated monodentate
(upper) and bidentate (lower) phosphate surface species on goethite
www.agronomy.org • www.crops.org • www.soils.org
2010a, 2010b; Weng et al., 2008). For modeling the adsorption
of NOM and the interaction between phosphate and NOM at
the surface of iron oxides, a simplified approach has been pro-
posed and used (Hiemstra et al., 2010b), in which the formation
reaction of a virtual NOM surface complex is defined as:
≡FeOH−0.5 + HNOM−1.0 ⇔ ≡FeNOM−1.5 + H2O [8]
The virtual NOM species can be used to model the appar-
ent effective NOM loading active in the phosphate compe-
tition. In the CD-MUSIC modeling, the overall charge z =
−1.5 of ≡FeNOM−1.5 is distributed in the interface, and this
charge contributes strongly to the competition with phosphate
(Hiemstra et al., 2010b). As shown in Hiemstra et al. (2010a),
for most soils studied, the calculated ≡FeNOM−1.5 loading is
between about 0.5 and 1.5 μmol m−2.
A more advanced approach to account for the presence of
NOM coating is the ligand and charge distribution (LCD)
model (Weng et al., 2007, 2008, 2006b). It is assumed that
the interaction between NOM and phosphate at the surface
of oxides can be simulated in the LCD model using a layer of
adsorbed FA, which is present and equally distributed between
the first and second Stern layer (Weng et al., 2008). Previous
study has shown that the FA particles can fit into the com-
pact part of the electrostatic double layer, whereas a large part
of the adsorbed humic acid is located beyond the electrostatic
double layer, in the diffuse double layer (Weng et al., 2008).
It was assumed that the carboxylic groups (RCOO−) of NOM
in the first compact layer can form innersphere complexes
(≡Fe1OOCR−0.5) with the singly coordinated surface sites on
goethite (Weng et al., 2006b). The corresponding charge is
distributed between the 0-plane and the 1-plane. The other
carboxylic and phenolic groups on the NOM can bind H+
and other cations like Ca2+. The reactions between the NOM
ligands and surface sites, protons, and other cations were cal-
culated in the LCD with the NICA model. It was assumed that
the NICA model parameters remain the same as for FA in the
solution (Milne et al., 2001, 2003).
Effects of Solution Chemistry
on Phosphorus Adsorption
The interaction of major cations and anions with goethite
(FeOOH) has been quantified systematically during the last
decade, resulting in a consistent set of parameters (Table 1)
that can now be used to quantify the major factors acting on
the phosphate binding in chemical barriers. These interactions
are explored for P-filter systems with and without the presence
of NOM.
In the Absence of Natural Organic Matter
The inorganic chemistry of natural aqueous solution is
dominated by a selective number of light elements. In non-
acid and oxidized systems, the main cations are Na+, K+, Ca2+,
and Mg2+, and the main inorganic anions are HCO3−, Cl−, and
SO42−. In addition, neutral silicic acid H4SiO40 is present at sig-
nificant concentrations. These major elements affect the bind-
ing of minor elements (e.g., heavy metals) and oxyanions (e.g.,
phosphate). The interaction of adsorbed ions is mainly of an
631
electrostatic nature (repulsion or attraction) and includes the
pH-dependent proton adsorption by the surface groups.
The interactions of cations and anions in multi-component
systems with iron oxides can be described with the CD-MUSIC
model, as explained above. Figure 2a gives the calculated sur-
face composition (Γ, in μmol m−2) of goethite in equilibrium
with a representative natural solution with [Na+] = 0.1 mmol
L−1, [K+] = 0.1 mmol L−1, [Ca2+] = 1 mmol L−1, [Mg2+] = 0.2
mmol L−1, [Cl−] = 2.5 mmol L−1, [H4SiO40] = 0.1 mmol L−1,
and [HCO3−] in equilibrium with atmospheric CO2 (4 × 10−4
bar) for pH 4, 6, and 8, respectively, as a function of the total
phosphate concentration in solution. The calculated loading of
specifically bound Na+, K+, and Cl− was not given in the figure
due to their insignificance as a result of a very low affinity.
The relative importance of solution species in their effect
on phosphate adsorption can be derived from their adsorbed
amount relative to that of phosphate (Fig. 2a). In Fig. 2a, both
scales are log scale. The calculation (Fig. 2a) shows that HCO3−/
CO32− and Mg2+ are relatively unimportant at the chosen ref-
erence composition of the equilibrium solution. Carbonate
adsorption increases as pH increases and decreases as phos-
phate concentration increases. The highest carbonate loading
(ΓCO3) is found in the system with the highest concentration of
HCO3− (pH 8) and lowest phosphate concentration, and ΓCO3
is less than 0.1 μmol m−2. The carbonate loading ΓCO3 decreases
with the increase of the phosphate concentration mainly due
to electrostatic effects that change outside the surface in the
1-plane due to the adsorption of both ions.
Adsorption of Mg2+ increases with pH, and the highest Mg
loading is for pH 8 (ΓMg < 0.1 μmol m−2). High pH is preferred
by cations because at high pH the particle is less positively or
more negatively charged. In the presence of phosphate, the
oxide particles as a whole are often negatively charged due to
the adsorption of the PO43− ions. This negative charge is attrac-
tive for Mg2+. However, despite the cooperative effect of phos-
phate, the Mg2+ loading starts to decrease at higher phosphate
concentrations in solution. This is due to the formation of dis-
solved MgHPO40 (aq), reducing the Mg2+ activity.
The behavior of Mg2+ contrasts with that of Ca2+ ions.
The loading of Ca2+ (ΓCa) increases with the phosphate con-
centration. The cooperative adsorption is due to electrostatic
interaction (attraction) between adsorbed phosphate (anion)
and calcium (cation). The adsorption of Ca2+ increases as pH
increases. Figure 2a shows that adsorption of Ca2+ becomes
significant at a relatively high pH (> pH 5) and above a phos-
phate concentration of 1 μmol L−1. Under these conditions, an

Fig. 2. The surface composition of goethite in equilibrium with K+, Na+, Ca2+, Mg2+, H4SiO40, Cl−, and HCO3− at concentrations representative for
natural soil solutions (see text) as a function of the phosphate concentration in solution for pH 4, 6, and 8 at PCO2 = 4 x 10−4 bar. Compositions
have been calculated with the CD-MUSIC model using the parameters of Table 1 (Hiemstra et al., 2007, 2010a). In the systems of Fig. 2a and 2b, no
natural organic matter is present, whereas in Fig. 2c and 2d, the natural organic matter loading is 1.66 μmol m−2. Figure 2a and 2b refer to the same
calculation, but the loading is given at the logarithmic and linear scales, respectively. Figure 2c and 2d refer to the same solution composition,
except a 10-fold higher H4SiO4 concentration (1.0 mmol L−1) was used in Fig. 2d than in 2c.

632 Journal of Environmental Quality

increase in Ca concentration promotes phosphate adsorption
to iron oxides.
Adsorption of silicic acid increases as pH increases. At a
given pH, the competition between phosphate and silicic acid
for the adsorption leads to a strong decrease of silicic acid
adsorbed when phosphate concentration increases (Fig. 2a).
Adsorbed silicic acid becomes significant at a low phosphate
concentration (<0.1 μmol L−1).
In summary, the calculation suggests that, in many natural
systems, the surface of iron oxides is dominated by adsorbed
PO43−, H4SiO4, and Ca2+. The influence of CO32− and Mg2+ on
phosphate adsorption is predicted to be insignificant. Silicic
acid (H4SiO40) may decrease phosphate adsorption, but this
effect is only important at a very low phosphate concentration,
in particular at high pH, as can be found in surface water.
In Fig. 2b, the surface composition is given at the linear
scale for the same solution conditions used for Fig. 2a. Because
the binding of Mg2+ and carbonate are very low (≤0.1 μmol
m−2), they were not plotted in Fig. 2b. At high phosphate con-
centrations, only Ca2+ and phosphate are relevant. The pH and
Ca concentration can be considered as most relevant in affect-
ing the phosphate adsorption to iron oxides. Only at very low
phosphate levels does the binding of Si become relevant.
In the Presence of Natural Organic Matter
Natural organic matter has a significant influence on
phosphate adsorption to iron oxides and may strongly influ-
ence the performance of the P filter. Using the CD-MUSIC
model (Table 1) and the virtual NOM approach (Eq. [8]), we
calculated the change of surface composition of goethite as a
function of the solution concentration of PO4 in the presence
of NOM. This calculation was performed for a representative
soil solution composition as used in the calculation in Fig. 2a
and 2b: [Na+] = 0.1 mmol L−1, [K+] = 0.1 mmol L−1, [Ca2+] =
1 mmol L−1, [Mg2+] = 0.2 mmol L−1, [Cl−] = 2.5 mmol L−1,
[H4SiO40] = 0.1 mmol L−1, and [HCO3−] in equilibrium with
atmospheric CO2 (4 × 10−4 bar) for pH 4, 6, and 8. The results
are shown in Fig. 2c. The difference between Fig. 2c and 2b is
the presence of adsorbed NOM species (1 virtual NOM nm−2
or 1.66 μmol NOM m−2). A second calculation is given in Fig.
2d for the same conditions but at a 10 times higher H4SiO4
concentration (1.0 mmol L−1).
To reduce the complexity of the figure, the adsorption of
carbonate and magnesium is not given. Comparing Fig. 2c
(with NOM) with Fig. 2b (no NOM), one can see that the
presence of NOM decreased phosphate and silicic acid adsorp-
tion and promoted Ca adsorption. The presence of NOM
decreased phosphate adsorption by 37 to 97% (compare Fig.
2c and Fig. 2b). The effect of NOM on Ca and phosphate
adsorption is stronger at a lower P concentration. In the pres-
ence of NOM, the surface loading of Ca is usually higher than
that of phosphate for soils of pH 6 to 8 under the Ca concen-
tration used (Fig. 2c, 2d). The Ca2+ adsorption is stimulated by
adsorbed NOM, whereas the phosphate loading is suppressed
by NOM, showing, respectively, the synergistic and the com-
petitive action of NOM on the adsorption of inorganic ions.
At pH > 5 and in the presence of NOM, phosphate adsorption
is more sensitive to the change of Ca concentration in the solu-
tion compared with that when NOM is absent.
www.agronomy.org • www.crops.org • www.soils.org
Similar to the case when no NOM was present (Fig. 2b),
silicic acid is dominantly adsorbed at low phosphate concentra-
tion in solution and at high pH value (pH 8) (Fig. 2c), as may
be found in surface waters. The high loading of Si is more pro-
nounced at a 10 times higher Si concentration (1.0 mmol L−1)
(Fig. 2d). This situation can be typical for aquifers. At an ele-
vated silicic acid concentration (1.0 mmol L−1) (Fig. 2d), com-
petition between silicic acid and phosphate plays an important
role at relatively high pH and at a phosphate concentration of
<10 μmol L−1. In fertile top soils, the phosphate concentration
is usually several micromolar or higher. At these concentrations,
phosphate is the most important oxyanion at the oxide surfaces
in acid and close to neutral soils (pH 4–6).
The role of SO42− in the surface speciation is not discussed
in detail here. Sulfate adsorption to oxides can be large at low
pH (pH < 5). The adsorption is significant in simple model
systems with goethite, sulfate, and a background electrolyte
(Geelhoed et al., 1997). However, in the presence of phosphate,
the anion competition determines the sulfate behavior. Because
sulfate is relatively weakly bound in comparison to phosphate,
phosphate adsorption removes adsorbed sulfate from the oxide
surface, while the phosphate adsorption remains largely unaf-
fected (Geelhoed et al., 1997). The presence of adsorbed NOM
further reduces the sulfate loading. Only at very acidic condi-
tions (pH < ?3–4), sulfate may be a relevant ion to be incor-
porated in surface complexation modeling of soil systems, such
as acid sulfate soils and acid mine tailings.
Another point that requires consideration in practice is the
contribution of organic P to the P release in the field. Organic
P adsorbs to iron oxides as well. Therefore, the adsorption of
NOM can diminish inorganic P adsorption via competition,
but it can also enhance P adsorption due to adsorption of
organic P.
Comparing Modeling and Experimental Adsorption Data
The above model simulation indicates that pH, Ca, and
NOM are the most important factors influencing phosphate
adsorption under the most common conditions in field. To
test the reliability of the model prediction for the effects of
these three factors, a series of batch adsorption experiments was
performed (see Materials and Methods). The results (Fig. 3,
symbols) can be compared with the model prediction (Fig. 3,
lines). The effect of the presence of adsorbed FA on phosphate
adsorption was calculated in this modeling using the LCD
approach. We can see from this figure that the CD-MUSIC
and LCD model predictions are in good agreement with the
data.
For comparison, the dissolved phosphate concentration in
the absence of Ca and NOM calculated using the CD-MUSIC
model (for an electrolyte background of 0.01 mol L−1 of NaCl)
is also given in Fig. 3. Compared with the situation when nei-
ther Ca nor NOM was present, adsorption of Ca promoted
phosphate adsorption (and therefore led to a lower soluble
phosphate), and this effect becomes significant when the pH
is not too low. In addition to the promotion of phosphate
adsorption, the presence of Ca changed the pH dependency
of phosphate adsorption as well. Due to a stronger adsorption
of Ca at increased pH, phosphate adsorption increases with
increased pH between pH about 4 to 7, whereas the trend
633
is the opposite in a background of Na or K. The presence of
FA decreased phosphate adsorption and led to strong increase
of phosphate in solution. The pH dependency of phosphate
adsorption in the presence of Ca and NOM shows a negative
relation in the range pH 3 to 7.
Implications
The above analysis shows that major factors determining
the phosphate adsorption to iron oxides are the phosphate con-
centration, the binding of NOM, the presence of Ca, and the
pH. This can be compared with the results found for a series
of soils. For these soils, the phosphate loading was measured
and the equilibrium concentration in 0.01 mol L−1 CaCl2 was
established, and pH and DOC concentration were measured
(Hiemstra et al., 2010b). Without scaling, no relation between
the PO4 loading on iron oxides (Γ) and the concentration of
dissolved phosphate (CPO4) is found. The reason is that the natu-
ral materials strongly differ in reactive surface area. Therefore,
scaling per unit reactive surface area is required. This surface
area can be approached experimentally using a new soil dilution
method in 0.5 mol L−1 NaHCO3 (Hiemstra et al., 2010a). After
scaling the PO4 loading (Γ) to mol m−2, the loading is related to
the logarithm of solution parameters. The strongest correlation
is with DOC (R2 = 0.66). The correlation is negative, point-
ing to the competitive character of DOC for phosphate. An
increase in NOM loading leads to a decrease in PO4 loading,
and vice versa (Hiemstra et al., 2010b). Considering the DOC–
PO4 interaction as an exchange process, the competition can be
described with a nonelectrostatic exchange equation:
K '(C PO4 / C DOC )n
Γ = Γ ' max
1 + K '(C PO4 / C DOC )n
where CPO4 is in mol L−1 and CDOC is in mg L−1. At a constant
DOC concentration, the isotherm has similarities with the
Fig. 3. Effect of the presence of fulvic acid (FA) (0, 150, and 300 mg L−1)
on the concentration of phosphate in a 0.01 mol L−1 CaCl2 solution
in a system with goethite (2.7 g L−1) having initially 0.48 mmol L−1
PO4 (experimental data points). The effects of FA on phosphate
adsorption were calculated using the ligand and charge distribution
approach (lines) applying the CD-MUSIC model parameters of Table
1. For comparison, the calculated soluble phosphate concentration
is given (thin line) for a system with 0.01 mol L−1 NaCl as background
(no Ca and FA).
634
Langmuir adsorption isotherm (Eq. [1]). A difference is the
presence of an exponent n at the concentration term. Modeling
shows that this term reflects the effect of electrostatics on the
shape of phosphate adsorption. The equation has been param-
eterized in this study using a large set of soil data (Hiemstra
et al., 2010a), for which the effective reactive surface area has
been determined, leading to K′ = 140 and n = 0.3 in the case
of Γ′max = 3 μmol m−2. A good description can be found (R2 =
0.69) for the data set of soils.
The practical equation (Eq. [9]) can be used for instance
to evaluate the possible influence of NOM on the adsorption
capacitance of a chemical barrier. In Fig. 4, the calculated effect
(using Eq. 9) of DOC on the PO4 adsorption in 0.01 mol L−1
CaCl2 is given for 0.5, 5, and 50 mg L−1 DOC. The variation in
DOC has a significant effect on amount of phosphate adsorbed.
In the micromolar range of phosphate concentration in solution,
an increase of DOC from 0.5 to 50 mg L−1 reduces amount of
phosphate adsorbed by about 50% (Fig. 4). This result shows
that the binding capacity of a reactive barrier depends on the
combination of CPO4 and CDOC in the solute. In addition, a reli-
able estimate of the reactive surface area is essential as a param-
eter (Hiemstra et al., 2010a) because the isotherms are expressed
per unit surface area, not per unit mass.
Conclusions
In summary, the above discussion shows that solution pH,
concentration of Ca, and the loading of NOM are the major
factors that may significantly influence the adsorption of phos-
phate to iron oxide materials in natural soil solutions at pH
4 to 8. Silicic acid may decrease phosphate adsorption, but
[9] this effect is only important at a very low phosphate concen-
trations, in particular at high pH, as can be found in surface
water. Natural organic matter and PO43− are strong competi-
tors, and adsorbed Ca2+ acts cooperatively on the binding of
these anions. The binding of NOM cannot be ignored when
attempting to understand the behavior of phosphate and other
Fig. 4. Effect of dissolved organic carbon (DOC) (0.5–50 mg L−1) on
the phosphate loading of the natural oxides in 0.01 mol L−1 CaCl2.
The isotherms have been calculated using Eq. 9, which has been cali-
brated on a large set of top soils with variable phosphate equilibrium
concentrations (~1–30 μmol L−1), DOC (~5–30 mg L−1), and pH (~4–7)
in 0.01 mol L−1 CaCl2.
Journal of Environmental Quality
oxyanions (e.g., B, As, Se, Cr, Mo, etc.) in natural systems.
The effects of NOM are stronger at a low phosphate concen-
tration. However, even at a micromolar concentration level of
phosphate, a 100 times increase in DOC can lead to a 50%
decrease in phosphate adsorption. Consequently, the soluble
phosphate concentration may increase by several orders of
magnitude because soluble phosphate is only a small fraction
of total phosphate. The mechanistic CD-MUSIC and LCD
models are useful tools in the understanding and prediction of
the net effects of combined factors.
References
Barberis, E., F.A. Marsan, R. Scalenghe, A. Lammers, U. Schwertmann, A.C.
Edwards, R. Maguire, M.J. Wilson, A. Delgado, and J. Torrent. 1996.
European soils overfertilized with phosphorus: 1. Basic properties. Fert.
Res. 45:199–207. doi:10.1007/BF00748590
Bhatti, J.S., N.B. Comerford, and C.T. Johnston. 1998. Influence of oxalate
and soil organic matter on sorption and desorption of phosphate onto
a spodic horizon. Soil Sci. Soc. Am. J. 62:1089–1095. doi:10.2136/
sssaj1998.03615995006200040033x
Borggaard, O.K., S.S. Jorgensen, J.P. Moberg, and B. Rabenlange. 1990.
Influence of organic-matter on phosphate adsorption by alu-
minum and iron-oxides in sandy soils. J. Soil Sci. 41:443–449.
doi:10.1111/j.1365-2389.1990.tb00078.x
Boujelben, N., J. Bouzid, Z. Elouear, A. Feki, F. Jamoussi, and A. Montiel.
2008. Phosphorus removal from aqueous solution using iron coated
natural and engineered sorbents. J. Hazard. Mater. 151:103–110.
doi:10.1016/j.jhazmat.2007.05.057
Carpenter, S.R., N.F. Caraco, D.L. Correll, R.W. Howarth, A.N. Sharp-
ley, and V.H. Smith. 1998. Nonpoint pollution of surface wa-
ters with phosphorus and nitrogen. Ecol. Appl. 8:559–568.
doi:10.1890/1051-0761(1998)008[0559:NPOSWW]2.0.CO;2
Chardon, W.J., and G.F. Koopmans. 2005. Phosphorus workshop. J. Environ.
Qual. 34:2091–2092. doi:10.2134/jeq2005.0001in
Codling, E.E., R.L. Chaney, and C.L. Mulchi. 2000. Use of aluminum-
and iron-rich residues to immobilize phosphorus in poultry litter and
litter-amended soils. J. Environ. Qual. 29:1924–1931. doi:10.2134/
jeq2000.00472425002900060027x
Devau, N., E. Le Cadre, P. Hinsinger, B. Jaillard, and F. Gerard. 2009. Soil
pH controls the environmental availability of phosphorus: Experimental
and mechanistic modelling approaches. Appl. Geochem. 24:2163–2174.
doi:10.1016/j.apgeochem.2009.09.020
Devau, N., P. Hinsinger, E. Le Cadre, B. Colomb, and F. Gerard. 2011.
Fertilization and pH effects on processes and mechanisms controlling
dissolved inorganic phosphorus in soils. Geochim. Cosmochim. Acta
75:2980–2996. doi:10.1016/j.gca.2011.02.034
Geelhoed, J.S., T. Hiemstra, and W.H. Van Riemsdijk. 1997. Phosphate
and sulfate adsorption on goethite: Single anion and competitive ad-
sorption. Geochim. Cosmochim. Acta 61:2389–2396. doi:10.1016/
S0016-7037(97)00096-3
Hiemstra, T., J. Antelo, R. Rahnemaie, and W.H. Van Riemsdijk. 2010a.
Nanoparticles in natural systems I: The effective reactive surface area of
the natural oxide fraction in field samples. Geochim. Cosmochim. Acta
74:41–58. doi:10.1016/j.gca.2009.10.018
Hiemstra, T., J. Antelo, A.M.D. Van Rotterdam, and W.H. Van Riemsdijk.
2010b. Nanoparticles in natural systems II: The natural oxide fraction at
interaction with natural organic matter and phosphate. Geochim. Cos-
mochim. Acta 74:59–69. doi:10.1016/j.gca.2009.10.019
Hiemstra, T., M.O. Barnett, and W.H. Van Riemsdijk. 2007. Interaction of si-
licic acid with goethite. J. Colloid Interface Sci. 310:8–17. doi:10.1016/j.
jcis.2007.01.065
Hiemstra, T., R. Rahnemaie, and W.H. Van Riemsdijk. 2004. Interpretation
of surface species from spectroscopy and CD modelling. Geochim. Cos-
mochim. Acta 68:A159.
Hiemstra, T., and W.H. Van Riemsdijk. 1996. A surface structural approach to
ion adsorption: The charge distribution (CD) model. J. Colloid Interface
Sci. 179:488–508. doi:10.1006/jcis.1996.0242
Hiemstra, T., and W.H. Van Riemsdijk. 2004. Interfacial distribution of
charge: Relationship between microand macroscopic ion adsorption
phenomena. Abs. Papers Am. Chem. Soc. 227:U1215.
Hiemstra, T., and W.H. Van Riemsdijk. 2006. On the relationship between charge
distribution, surface hydration, and the structure of the interface of metal hy-
www.agronomy.org • www.crops.org • www.soils.org
droxides. J. Colloid Interface Sci. 301:1–18. doi:10.1016/j.jcis.2006.05.008
Hiemstra, T., W.H. Van Riemsdijk, and G.H. Bolt. 1989. Multisite pro-
ton adsorption modeling at the solid-solution interface of (hydr)
oxides: A new approach. 1. Model description and evaluation of in-
trinsic reaction constants. J. Colloid Interface Sci. 133:91–104.
doi:10.1016/0021-9797(89)90284-1
Koopmans, G.F., W.J. Chardon, and R.W. McDowell. 2007. Phosphorus move-
ment and speciation in a sandy soil profile after long-term animal manure
applications. J. Environ. Qual. 36:305–315. doi:10.2134/jeq2006.0131
Kronvang, B., M. Bechmann, H. Lundekvam, H. Behrendt, G.H. Rubaek,
O.F. Schoumans, N. Syversen, H.E. Andersen, and C.C. Hoffmann.
2005. Phosphorus losses from agricultural areas in river basins: Effects
and uncertainties of targeted mitigation measures. J. Environ. Qual.
34:2129–2144. doi:10.2134/jeq2004.0439
McDowell, R.W., A.N. Sharpley, and W. Bourke. 2008. Treatment of drainage wa-
ter with industrial by-products to prevent phosphorus loss from tile-drained
land. J. Environ. Qual. 37:1575–1582. doi:10.2134/jeq2007.0454
Milne, C.J., D.G. Kinniburgh, and E. Tipping. 2001. Generic NICA-Donnan
model parameters for proton binding by humic substances. Environ. Sci.
Technol. 35:2049–2059. doi:10.1021/es000123j
Milne, C.J., D.G. Kinniburgh, W.H. Van Riemsdijk, and E. Tipping. 2003. Ge-
neric NICA-Donnan model parameters for metal-ion binding by humic
substances. Environ. Sci. Technol. 37:958–971. doi:10.1021/es0258879
Pan, G., L. Li, D.Y. Zhao, and H. Chen. 2010. Immobilization of non-point
phosphorus using stabilized magnetite nanoparticles with enhanced
transportability and reactivity in soils. Environ. Pollut. 158:35–40.
doi:10.1016/j.envpol.2009.08.003
Rahnemaie, R., T. Hiemstra, and W.H. Van Riemsdijk. 2006. A new structural
approach for outersphere complexation tracing the location of electrolyte
ions. J. Colloid Interface Sci. 293:312–321. doi:10.1016/j.jcis.2005.06.089
Rahnemaie, R., T. Hiemstra, and W.H. Van Riemsdijk. 2007. Geometry,
charge distribution, and surface speciation of phosphate on goethite.
Langmuir 23:3680–3689. doi:10.1021/la062965n
Sakadevan, K., and H.J. Bavor. 1998. Phosphate adsorption characteristics of
soils, slags and zeolite to be used as substrates in constructed wetland
systems. Water Res. 32:393–399. doi:10.1016/S0043-1354(97)00271-6
Sharpley, A.N., S.C. Chapra, R. Wedepohl, J.T. Sims, T.C. Daniel, and K.R.
Reddy. 1994. Managing agricultural phosphorus for protection of surface
waters: Issues and options. J. Environ. Qual. 23:437–451. doi:10.2134/
jeq1994.00472425002300030006x
Stout, W.L., A.N. Sharpley, and J. Landa. 2000. Effectiveness of coal combus-
tion by-products in controlling phosphorus export from soils. J. Environ.
Qual. 29:1239–1244. doi:10.2134/jeq2000.00472425002900040030x
Torrent, J., V. Barron, and U. Schwertmann. 1990. Phosphate adsorption and
desorption by goethites differing in crystal morphology. Soil Sci. Soc. Am.
J. 54:1007–1012. doi:10.2136/sssaj1990.03615995005400040012x
Weng, L.P., W.H. Van Riemsdijk, and T. Hiemstra. 2007. Adsorption of hu-
mic acids onto goethite: Effects of molar mass, pH and ionic strength.
J. Colloid Interface Sci. 314:107–118. doi:10.1016/j.jcis.2007.05.039
Weng, L.P., W.H. Van Riemsdijk, and T. Hiemstra. 2008. Humic nanoparticles
at the oxide–water interface: Interactions with phosphate ion adsorption.
Environ. Sci. Technol. 42:8747–8752. doi:10.1021/es801631d
Weng, L.P., W.H. Van Riemsdijk, L.K. Koopal, and T. Hiemstra. 2006a.
Adsorption of humic substances on goethite: Comparison between
humic acids and fulvic acids. Environ. Sci. Technol. 40:7494–7500.
doi:10.1021/es060777d
Weng, L.P., W.H. Van Riemsdijk, L.K. Koopal, and T. Hiemstra. 2006b. Li-
gand and charge distribution (LCD) model for the description of ful-
vic acid adsorption to goethite. J. Colloid Interface Sci. 302:442–457.
doi:10.1016/j.jcis.2006.07.005
Weng, L.P., L.K. Koopal, T. Hiemstra, J.C.L. Meeussen, and W.H. Van Riemsdijk.
2005. Interactions of calcium and fulvic acid at the goethite-water interface.
Geochim. Cosmochim. Acta 69:325–339. doi:10.1016/j.gca.2004.07.002
Withers, P.J.A., I.A. Davidson, and R.H. Foy. 2000. Prospects for controlling
nonpoint phosphorus loss to water: A UK perspective. J. Environ. Qual.
29:167–175. doi:10.2134/jeq2000.00472425002900010021x
Zeng, L., X.M. Li, and J.D. Liu. 2004. Adsorptive removal of phosphate from
aqueous solutions using iron oxide tailings. Water Res. 38:1318–1326
10.1016/j.watres.2003.12.009. doi:10.1016/j.watres.2003.12.009
Zhang, G.S., H.J. Liu, R.P. Liu, and J.H. Qu. 2009. Removal of phosphate
from water by a Fe-Mn binary oxide adsorbent. J. Colloid Interface Sci.
335:168–174. doi:10.1016/j.jcis.2009.03.019
635