Professional Documents
Culture Documents
Introduction
Metal Ion Binding to Hemic Humic substances play an important role in the behavior
Substances: Application of the and fate of metal ions in the natural environment. They
can control metal ion concentrations in soils and natural
Non-Ideal Competitive Adsorption waters and can affect the mobility of metals through soils
and aquifers (I, 2). The interaction between metal ions
Model and soil particles can also be controlled by humic sub-
stances. Humic substances are believed to play a major
MARC F. BENEDETTI,* *it't role in trace metal availability and toxicity to plants and
CHRIS J. MILNE,6 living organisms.
DAVID G. KINNIBURGH,6 Among the humic substances, the humic acids (HA) and
WILLEM H. VAN R I E M S D I J K , * ·+ AND fulvic acids (FA) are polydisperse mixtures of natural organic
LUUK K. KOOPAL" polyelectrolytes having different types of functional groups
Department of Soil Science and Plant Nutrition and to which ions can bind. Two major types of functional
Department of Physical and Colloid Chemistry, groups are of particular importance: carboxylic groups and
Wageningen Agricultural University, Dreijenplein 10,
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
6703 HB Wageningen, The Netherlands, and British Geological phenolic groups {3-7). Amino, sulfhydryl, and quinone
Survey, Wallingford, Oxon, 0X10 8BB, U.K. groups can also be present and may have strong interactions
with trace metal ions (1,3,4). However, since their number
is much smaller, their contribution to ion binding is also
generally small, although under certain conditions it can
The application of a new model to describe metal be substantial.
Downloaded via INRAE on October 2, 2020 at 19:30:49 (UTC).
ion binding by humic acids is discussed. Metal ion In natural systems, metal ion binding will be affected by
different processes and humic acid properties. Due to their
binding is always of a competitive nature since the
diverse functional groups, the humic acids behave as
proton is always present. Although of great practical
heterogeneous ligands (1,4-9), and this chemical hetero-
importance, the combination of a chemically hetero- geneity will be reflected in their ion binding properties.
geneous system with competitive binding poses The polyelectrolytic behavior of humic acids will also affect
difficult problems from both experimental and theoretical ion binding due to electrostatic interactions {1,5,6,9-12).
points of view. The new Non-Ideal Competitive Natural waters contain a variety of ions that will compete
to a greater or a lesser extent for the available binding sites.
Adsorption model (NICA model) used here is able to
account for the non-ideal binding to heterogeneous Protons are always present in aquatic systems, and they
too will compete with metal ions for binding to humic acids.
ligands. A good description of the binding of H, Ca, Cd, The proton concentration determines the degree of ioniza-
and Cu to a purified peat humic acid is achieved
tion of the functional groups as well as the charge of the
over a wide range of free metal ion concentrations (-2 humic acids. Multivalent ions such as Ca2+, Mg2+, Fe3+,
>
log Me2+ > -14) and pH (2 < pH < 10). The and Al3+ are known to bind to humic acids (1). In principle,
results show that binding of metal ions to humic acid all these ions will compete for the same sites and hence
is strongly influenced by the intrinsic chemical influence the binding of each other.
The complexity of such a system is great and has led to
heterogeneity of the humic material itself as well as
the development of various models {2,10,11,13-15). The
by ion-specific non-ideality. The results indicate
aim of forward modeling is to be able to describe ion binding
that copper competes much more efficiently with protons
at different pH’s and ionic strengths and over a wide range
bound to the phenolic type groups than calcium and of concentrations for the different ions. Models of ion
cadmium. binding by humic substances described in the literature
can be divided into two categories: those assuming a
discrete distribution of ligands and those assuming a
continuous distribution of heterogeneous ligands. The
discrete ligand approach can successfully describe metal
binding in the absence of competing ions (proton excluded)
within the range of the conditions of the calibrating titrations
{13). Although competition among ions can be described
by the discrete approach, because of the variable nature of
the natural ligands discrete models generally require extra
446 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 29, NO. 2, 1995 0013-936X/95/0929-0446Í09.00/0 1995 American Chemical Society
ligand sites in order to satisfactorily match the experimental
data (16). Whether this difficulty can be overcome by
models based on a continuous distribution is still a matter
of discussion (15). Dobbs et al. (17) and Susetyo et al. (14)
successfully used a continuous binding site model to predict
the effect of protons and Al on Eu binding to humics. Up
until recently, the main restriction with models based on
a continuous distribution is that it is assumed that the
different metal ions experience the same chemical het-
erogeneity (15, 18). When single metal ion adsorption
experiments are carried out using the same humic acid, it
is found however that the different metal ions do not
experience the same apparent heterogeneity. Competitive
adsorption cannot then be adequately described by the
"classical” (18) multicomponent models. Therefore, to
address this problem, Koopal et al. (19) developed a new
non-ideal competitive adsorption (NICA) model in which FIGURE 1. Calibration curve for Cu ISE produced by titrating Cu-
heterogeneity and component-specific non-ideality are (N03)2 with ethylenediamine. Free Cu activities were calculated
considered explicitly. explicitly for each point using a speciation model, the measured pH,
The NICA model is based on a continuous distribution and the total concentrations of the components. Two limits are
shown.
of heterogeneous ligands in which the overall non-ideality
is divided into an intrinsic heterogeneity contribution (a
Cadmium and Calcium. Data for Cd and Ca binding
property of the humic substance) and ion-specific non- to PPHA are taken from Milne et al. (22). Metal ion
ideality contributions. In the NICA model, the polyelec- concentrations were measured potentiometrically using Cd
trolytic nature of ion adsorption to humic substances is or Ca ion-selective electrodes (ISE) to give ion binding
accounted for in a generic way by the non-ideality
isotherms at pH 4, 6, and 8 for Cd and at pH 6, 8, and 10
parameters. In literature, several attempts have been made for Ca. All experiments were performed in a 0.1 KN03
to model the electrostatic interactions more specifically.
For example, Tipping and Hurley (15) use an empirical background electrolyte.
Copper. Copper concentrations in solution were mea-
equation to relate charge and potential. Marinsky et al. sured potentiometrically using a sulfide-based solid-state
(10, 20, 21) employ a Dorman potential term, and De Wit ion-selective electrode (Orion 9429). This technique has
et al. (5, 6, 9), Bartschat et al. (12) and Milne et al. (22)
been used before for measuring Cu binding to humic
compute the electrostatic potential with the Poisson- materials with considerable success (24, 25). It was
Boltzmann equation, assuming a given geometry (e.g., demonstrated using ethylenediamine (en) titrations that
sphere or cylinder) for the shape of the FA or HA particles. linear response of Cu ISEs can be obtained for Cu activities
It is also possible to extend the NICA model to take into
as low as pCu 19 (26). Other authors have calibrated using
account specifically the ionic strength effects on both proton
standard solutions at higher Cu activities and have assumed
and trace metal binding, and we will return to this aspect
an extension of the calibration using linear (24) or poly-
in a future paper. The binding will be studied predomi- nomial extrapolation (25). Here the electrode performance
nantly at a single ionic strength of 0.1 M. We assume that was verified by titrating aqueous (en) (0.1 M) into Cu(N03)2
at constant ionic strength electrostatic interactions are of
(0.01 M). The total concentrations of the component
similar magnitude for the various divalent ions under
species and the measured pH were then fed into the
investigation. ECOSAT speciation model (27) to calculate the concentra-
The aim of this paper is to test the NICA model for a
tion and activity of free Cu2+ at each point in the titration.
wide range of competitive conditions using a single humic
acid sample. The sample used is a purified peat humic
Stability constants for complexes were taken from refs 28
and 29:
acid (PPHA) for which the proton-charging behavior has
previously been analyzed using an electrostatic diffuse
double layer model (7). Ca and Cd (22) binding data are Cu(en) ^ Cu2+ + en log K =
10.48
also available for the PPHA material, and new experimental
data for Cu binding are presented here. Hence, the model Cu(en)2 2=1 Cu2+ + 2en log K = 19.55
can be tested with H, Cd, Cu,and Ca data for the same
humic acid. In the following, we will first briefly describe H2(en) ^ H + H(en) log K = 6.848
the experimental procedures adopted and then describe
the model and its use for modeling proton and trace metal The calculated free Cu activities were plotted against the
binding. observed ISE emf readings to give a calibration curve (Figure
1) which was linear over the whole range of the calibration
Experimental Section titration from pCu 2 to pCu 12. For experimental data points
All experiments were performed on purified peat humic between pCu 12 and pCu 14, we assume that the linear
acid (PPHA), the purification and properties of which have response continues; we saw no evidence to the contrary
been described by Milne et al. (7). The initial concentration during the experiments. The fitted slope of the calibration
of the humic acid suspension was 5.28 mg of PPHA/g of (30.9 mV) is larger than might have been expected from
suspension. Potentiometric experiments were carried out Nemstian theory, but it was consistent and reproducible
using a fully automated titration system, which is described both in calibrations and during experiments with humic
in detail elsewhere (23). acid.
„= /
Alog K,
,/log Kt) d(log Kt) (2)
(Kitlcri
u!
Qi,t Qmax,l ^
where /(log K,) is the distribution function of the affinity Xc^ic,.)""'11 + {£(^)nu}pi
constant, 0¿,¡ is the local adsorption isotherm, i.e., the
binding to a group of identical sites (32), and Alog K¡ is the
range of log K¡ to be considered. This equation can be Qmax,2--(5)
solved analytically for certain distribution functions in
1 +
combination with certain local isotherms. For a multi-
component system, Koopal et al. (19) have derived the
following nonideal competitive adsorption expression for where the subscripts 1 and 2 relate to the first and second
0¡,t by using the extended Henderson-Hasselbalch as the peak of the affinity distribution.
non-ideal local isotherm in combination with a Langmuir- Figure 2 illustrates schematically the general procedure
Freundlich (LF) type affinity distribution function (quasi- which is used to derive the various parameter values of the
Gaussian) by Sips (33): NICA model for a multicomponent set of ion binding data.
First, monocomponent binding data are used in order to binding to fulvic acid as measured by Saar and Weber (25)
derive a value of m, describing the combined effect for one can be described perfectly in this way (19).
component (m,· = n¡p). For this purpose, the proton binding
as a function of pH in indifferent electrolyte is used. The Results
salt level is chosen to be the same as what is used for the
Experimental Data. Proton. The proton data are discussed
study of the metal ion binding. A relatively high salt level extensively elsewhere (7), but since the proton charging
may be chosen to suppress the effects due to electrostatic curve is used to derive the proton parameters the major
interaction. In the first step, the value of the average proton results are summarized here. Surface charge versus pH
affinity constants, the total number of sites, and the value curves for the PPHA measured at different ionic strength
of m.H for both peaks can be established. In the second could be modeled reasonably well with the master curve
step, the value of p can be derived using data from a metal (surface charge versus surface pH) approach (7, 9). Het-
binding experiment in which metal binding is studied as erogeneity analysis (35,36) of the data suggested an affinity
a function of the metal concentration at different pH’s.
distribution with two peaks. Since the proton binding curve
This enables p to be calculated from both the metal and to PPHA at 0.1 KNO3 is used to obtain the proton
the proton data and at the same time K¡ and n, can be
parameters (Ku mu, and Qmax) for each of the two peaks
,
derived for the metal. The value of is given by mu I p. of the affinity distribution, the experimental binding curve
For subsequent metal ions, only K¡ and n¡ have to be derived and the derived affinity distribution are shown in Figure 3,
since p is already known. Therefore, each new metal ion
panels a and b, respectively.
species requires only two new parameters for each affinity Metals. The symbols in Figure 4 present the Cd binding
distribution. For the extended NICA (eq 5), the same
data, log[Cdads]> as a function of free cadmium concentra-
procedure applies, but now the constants for both groups tion, log [Cd2+], for pH 4,6, and 8. The binding is strongly
of sites should be determined.
To summarize, for the extended NICA model, six pH-dependent as previously seen by Saar and Weber (25)
for fulvic acids. Cadmium binding increases with increasing
parameters are derived from the proton data. These consist
of two median affinity constants, two site densities, and pH, although the separation between the pH 4 and 6 and
the pH 6 and 8 curves in the log-log plot is not the same.
two mH values corresponding to the widths of the two
In Figure 5, the measured Ca binding is given for pH 6,
apparent affinity distributions. Then for the first metal, i,
six additional parameters have to be derived: again two 8, and 10, by the symbols, as before; results are plotted as
median affinity constants for metal i, plus two p values and log[Caadsl versus log[Ca2+]. As for Cd, the binding is strongly
two n, values. Finally, each additional metal requires four pH-dependent with more Ca bound to the humic acid at
further parameters: two median affinity constants and two higher pH. ·
values for n¡. The total number of parameters needed to The Cu binding data (log[Cuadsl versus log[Cu2+]) are
describe the interaction of the proton and one metal with shown by the symbols plotted in Figure 6. Results show
humic, 12; is moderately high but is within the range used similar features to the other two metals, but Cu is clearly
by other authors, e.g., 10 parameters (34) or 13 parameters more strongly bound than Ca or Cd. The extent of binding
(15), and reflects the complex heterogeneous nature of is comparable to that found for Cu binding to Suwannee
humic materials. However, if within the range of the River fulvic acid and natural DOC (34, 37).
experimental conditions only one group of sites (e.g., the A comparison between the binding on PPHA of Cu, Cd,
carboxylic) is needed to account for the observed experi- and Ca at pH 6 and in a 0.1 KN03 background electrolyte
mental binding curves, the basic NICA (eq 3) can be used is given in Figure 7. The effect of pH on the binding (Figures
and only six parameters will be needed to model the binding 4—6) is larger for Cu than for Cd, and for Cd it is larger than
of a proton plus one metal. The pH-dependent cadmium for Ca. It should be noted that the slope of the different
{k1[H\}mi {k2[H\}m2
Q ~
+
QmaXi11 + {K.LHir Qmax'21 + {k2mm2
(6)
NICA
-10 -8 -6 -4 -2
log{[Cd2+]/mol l-1}
FIGURE 4. Cadmium binding to the PPHA as a function of !og[Cd2+] at pH 4,6, and 8 showing the comparison between experimental data
and the NICA model description.
2+
log{[Ca ]/mol 1_1}
FIGURE 5. Calcium binding to the PPHA as a function of log[Ca2+] at pH 6,8, and 10 showing the comparison between experimental data
and the NICA model description.
FIGURE 6. Copper binding to the PPHA as a function of log[Cu2+] at pH 4, 6, and 8 showing the comparison between experimental data
and the NICA model description.
of metal. The competitive nature of the NICA model fitting procedure (Figure 2). Initial estimates of the six
requires the assumption that Qmax,H Qmax,metal· Further-
=
parameters are needed before automatic optimization can
more, mi and m2 will be used to derive pi and p2 in the proceed. Although in numerical fitting procedure a “blind
multicomponent case (Figure 2). guess” could be made for the initial values for each
Modeling Ca, Cd, and Cu Data. Metal ions bound in the parameter, we prefer to derive initial estimates of the metal
diffuse double layer could produce differences between parameters graphically from the adsorption curves. In this
model results and the experimental data. However, it is way, failure of convergence in the fitting routine is
possible with a Donnan model {10,20,21) to estimate the minimized.
composition of the diffuse layer and to correct for elec- For the graphical analysis, the second peak in eq 5 is
trostatic effects. These matters will be discussed in a neglected. This simplification is the more reasonable the
separate article (39), but an important conclusion from these lower the pH. At constant low pH and low free metal ion
calculations is that non-specific binding is always lower concentration, it is assumed that the proton term in eq 3
than 1-10% for all metals (high values occurring for high is constant and always much larger than the metal term,
metal concentrations). A similar result has been found by i.e.
Tipping and Hurley (15) Therefore, we have decided to
.
FIGURE 7. Comparison of the relative strengths of Ca, Cd, and Cu binding to PPHA at pH 6.
the element under investigation, and yet it is clear from A pH-induced shift of the metal binding curves, expressed
our experimental results for Cd and Cu (Figures 4 and 6) as log-log plot now gives
that this is not often the case. Hence, as was previously
shown by Koopal et al. (19) with literature data from Saar nH(p 1)
-
experimental data.
Equation 3 can also be reduced to another simplified The interesting aspect of eq 14 is that it illustrates that the
form. This time the proton term is no longer assumed to pH effect on the metal ion binding may be dependent on
be constant, but it is still assumed to be much larger than the intrinsic heterogeneity (p) and on the non-ideality factor
the metal term at low metal concentrations and also Kn Ch for the proton (nH) and the metal ion (n¡). From measure-
is assumed to be larger than 1 i.e., eq 7b still applies. This ments of metal ion binding measured for different (con-
leads to stant) pH values, an estimate can be obtained of the term
tzh(p
—
pH (11)
graphical procedure outlined above are given in Table 2.
Initial values for p are calculated from p=ma/nn using the
where v is a new constant equal to Qmax , ^ . The fitted mu and estimated nH.
shift of the logarithm of the free metal ion concentration Since a bimodal affinity distribution was introduced in
for a given value of bound metal as a function of a change the NICA model, initial estimates of the parameters for the
452 " ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 29, NO. 2, 1995
Exchange Ratio (H/Cu) Sections A (using Cu to determine p) and B (using Cd to
determine p).
Comparing Tables 3 and 2, it follows that eq 14 gives an
excellent estimate of nH,i and thus also of pi, whereas
estimates based on eq 12 are much less accurate. It also
follows from Table 3 that the results obtained for the
parameters that pertain to the first peak are quite insensitive
for the choice of the metal used for the determination of
Pi. For the determination of p2, a large uncertainty is
observed when the cadmium data are used. This is easily
understood, since, as we will see later, the second peak
contributes relatively little to the total cadmium binding
for this data set. For a reliable estimate of the parameters
for the second peak, the use of the copper data instead of
the cadmium is therefore to be preferred. Therefore the
Exchange Ratio (H/Cd) set of parameters based on Cu (Table 3, Section A) were
used in order to calculate the model curves shown in Figures
4-6 for Cd, Ca, and Cu, respectively.
The description of the Cu data is very good (Figure 6)
over a wide range of free metal concentrations (10 pCu2+
units) and over a range of 4 pH units. The parameters
derived from the modeling show a strong non-ideal behavior
for Cu with respect to the carboxylic site (nCu,i = 0.52). An
even stronger Cu-specific non-ideality is found for the
second type of site since «cu.2 = 0.30. The phenolic sites
have a higher affinity for Cu than the carboxylic site since
log K2 > log k\. Both p values are small (pi = 0.47 and p2
=
0.68), pointing toward a strong intrinsic heterogeneity
for the peat humic acid (PPHA); since p\ < p2, our results
Exchange Ratio (H/Ca)
suggest that the carboxylic sites display a greater hetero-
geneity than the phenolic sites.
Good agreement between model predictions and the
measurements is also found for Cd over a wide range of
free metal ion concentrations (2 < pCd2+ < 10) and pH (4
pH < 8). The main difference in median affinity constants
between cadmium and copper is that the median affinity
constant for the second peak is much smaller for cadmium
than for copper, whereas for the carboxylic peak a much
smaller difference is observed. The non-ideality for cad-
mium is smaller than for copper since nca is closer to one
than «cu·
The agreement between model description and the
FIGURE 8. Measured and calculated proton metal exchange ratio measurements for Ca is somewhat less good as found for
for three metals (Ca, Cu, and Cd) as a function of log free metal Cd or Cu, but was nevertheless satisfactory. The calcium
concentration for pH 6 and 8. median affinity constants are smaller than those derived
for the other trace metals, which is in agreement with
second peak are also needed. We assumed, with the literature data, which shows smaller calcium stability
exception of the median binding constants, that these values constant values for the carboxylic and phenolic acid
were the same as those for the corresponding parameters complexes of various organic molecules compared with
of the first peak. This assumption was supported by the the other metals (28, 29). For calcium log ^ is slightly
small difference between mi and m2 found during the fitting smaller than log K2.
of the proton data (T able 1). Initial estimates of the median The parameters «h,i and nHl2> which account for the
binding constants for the metals were taken from literature non-ideal behavior of the proton toward the humic acid or
data (15). The final values of the model parameters were for a specific chemical heterogeneity of the humic toward
derived with the non-linear least squares program FIT the proton are also given in Table 3. They are quite different
for both types of sites. The carboxylic sites show very little
developed by Kinniburgh (41).
non-ideal behavior for protons since « , is almost 1, while
It is also possible to derive the values of pi and p2 for the phenolic sites uh,2 0.57 indicates a strong non-
=
common to all metals by fitting all of the metal data ideal behavior for the protons.
simultaneously. However, we decided to derive these values Proton/Metal Molar Exchange Ratio. Once all the
by fitting the Cu and the Cd data independently. This parameters are known for both the metals and the protons,
provides a simple sensitivity analysis of the model. After it is possible to calculate the proton metal exchange ratio
p was derived for the first metal, pairs of k¡, n¡ values were with the NICA model and compare the results with
derived for each of the subsequent metals. The results for experimentally measured data. The results (Figure 8)
the three metals (Cu, Cd, and Ca) are given in Table 3, indicate that the model predicts exchange ratios in the range
Parameters Derived from Proton Charging Data at Ionic Strength of 0.1 KN03 Using NICA Eq 6
log Ah,i m\ Qmax.i log Ah,2 rn2 Omax,2
4.601 ± 0.002 0.440 ± 0.001 2.48 ± 0.01 9.34 ± 0.01 0.39 ± 0.01 1.93 ±0.02
(see text for details), see Section A. Next, Cd was used first to derive
the weak acid groups contribute significantly to the binding
p and nH for both first and second peaks, see Section B. The residual
root mean square error (RMSE) defined as at higher pCd values. The importance of the binding to the
nd phenolic sites at high pH and low Cd concentrations (pCd2+
-
X)2
> 10) is
clearly demonstrated; less than 50% of the bound
£<X, Cd is associated with the carboxylic sites under these
RMSE = —-
nd -