Documente Academic
Documente Profesional
Documente Cultură
rm
Geochimim et Cosmochimiecr Acta Vol. 54, pp. 1535-1553
Copyright 8 1990 Pcrgamon PICSYpk. Rintcdin U.S.A.
Ahatract-This paper presents a new conceptual approach to interpreting titration curves of metal com-
plexation by physically and chemically heterogeneous natural complexants such as humic acids, clays,
complete soils, or sediments. The physico-chemical and analytical difficulties encountered with such
systems are reviewed by comparison with a system containing only a few simple ligands, followed by
discussion of the new approach on the same basis. It is shown that interpretation of heterogeneous
complexant properties necessitates a preliminary transformation of experimental raw data into a function
sufficiently normalized so as to allow comparison of results obtained under different conditions. A nor-
malized function called a Site Occupation Distribution Function (SODF) and its potential usefulness is
described here. The SODF is a readily computable function which relates the complexation buffer intensity
of the system to the differential free energy of the complexation sites present. Its major interest is that it
enables one to obtain both a rigorous mathematical description of the complexant properties (even when
highly heterogeneous) at the macroscopic level and, in certain cases, an estimation of the molecular-scale
behavior of particular site types.
The relationship of the SODF to other distribution functions proposed in the literature is discussed
and applications are exemplified using simulated and real natural systems. In particular, its utility is
discussed in detail for ( 1) discriminating between different site types (major, minor, dominant, back-
ground), (2) evaluating the degree of heterogeneity of an unknown complexant system, (3) estimating
the nature and true thermodvnamic constants of complexes, and (4) yielding a rigorous definition of
“complexation capacity.”
anthropogenic activities. More particularly, a large number Cl ~, amino acids) through groups of complexants made up
of observations and important questions point to the im- of relatively small-sized compounds having similar properties
portance of understanding the speciation of trace metals in and often with several sites per molecule (e.g., fulvic acids,
natural aquatic environments. Examples include peptides) to high molecular weight compounds or particles
having large numbers of sites per physical entity (e.g.. humic
( 1) the demonstration that, in laboratory experiments, bio-
substances, polysaccharides, hydrous metal oxide surfaces,
logical response is often correlated with the free aquo
clays). Prediction of the fraction of total metal complexed
cation activity (e.g., Cu*+. Cd’+, Pb’+), and not with
with each of the different complexant sites present would
total metal concentration;
require complete knowledge of all physico-chemical param-
(2) the need to understand total metal partitioning between
eters controlling metal speciation. Particularly important in
dissolved, colloidal, and particulate phases in order to
such a description are the complexation equilibrium con-
predict metal migration through groundwater and pore- , 1
status.
water systems;
At present, most of the equilibrium constants for metal
(3) questions concerning the relationship between biological
complexes with natural materials (humics, clays, hydrous
uptake and metal association with other compounds or
metal oxides, soils and sediments, etc.) are either unknown
phases;
or too empirical to be of predictive use. A procedure for
(4) the dynamics of free aquo ion activity response to changes
determining sound physico-chemical parameters from raw
in total metal loading or pH (i.e., the metal ion buffer
experimental titration data is needed, but this poses many
capacity of the system).
difficulties. Various interpretation models have been proposed
The speciation of a given metal is defined as the partitioning ( SPOSITO, 1986; BUFFLE, 1988) which make assumptions
of the total metal present in a particular system among all concerning the nature of the reactions involved. Although
possible chemical forms through reactions with all available such models yield adjustable reaction parameters, the chem-
complexant sites (Fig. 1). Complexants are considered here ical meaning of these parameters seems questionable when
to be all compounds, particulate or dissolved, organic or in- the complexity of natural systems is considered. Indeed, any
organic, which can combine with the metal cation ( BUFFLE. CIpriori model which takes this complexity into account must
1988). They range from the most simple ligands (e.g., OH -, include so many independent and non-directly measurable
DISSOLVED
cobIPLExEs
\
FIG. 1. Metal complexation by_dissolved _ and particulate complexants in natural systems determines metal ion
buffering response, biological uptake, and transport.
Theory of metal complexation 1537
DISCRETESYSTEM CONTINUOUS
SYSTEM
-6
-3og $2 O
-12 -10 -6 -6 -I -2 0
log IMI
Experimental
-..-..-..-..-..-..-..-..-..-..-..-..-... .-..-..-..-..-.._
Theoretical
12 10 6 6 4 2 0 12 10 6 6 4 2 0
log Ki log K
l2 lo a10:K4 * ’
FIG. 2. Comparison of various representations of data for simulated discrete and continuous complexant systems.
Discrete system: three ligands having log K, = 4, 6, and 8; mole fractions, Axi = 0.33 for each. Continuous systems:
simulated by 27 “background” sites spaced at 0.5 log K intervals along a Sips’ SADF (r = 0.5; log Km = 1); three
discrctc “dominant” sites are superimposed on the background at log Ki = 4,6, and 8 with CF = 15, 10 and 20 times
that of the background (CF = concentration ratio of the minor dominant site with respect to the background at a
given log K value). (SeeSec.111.1for more details concerning the continuous system and Table of Symbols for
definitions.)
(3) The minimum detectable concentration of site types gest the utility of differentiating the site types present in com-
varies across the analytical window and is a function of plexant systems into those which are amenable to classical
K value spacing; as the spacing decreases, the difference interpretation approaches and those for which another tack
in relative site concentrations should increase; otherwise must be taken. The former are the major sites and the latter
the influence of the site having the lower concentration the minor sites. This site distinction is discussed in greater
is “swamped” by that present in a much higher relative detail in Sec. IV. 1.
concentration ( ALTMANN and BUFFLE,1988 ) .
(4) The experimental error affects the minimum amplitude 1.3. A Proposed Approach
of the structural features in the titration data which can
Since interpretation of complexation by natural hetero-
be detected and, therefore, the sites which can be mea-
geneous complexants cannot be based on any a priori physic
sured and characterized.
chemical model (step 2’ in Fig. 3), a deductive approach
Given the complexity of natural system briefly described must be taken to develop a realistic model based on experi-
above, the difficulties likely to be encountered in interpreting mental observations (steps 2a, 2b in Fig. 3). This implies
their complexation behaviour using classical, discrete site ap- accumulation of a large number of experimental titration
proaches should be apparent. These considerations also sug- curves obtained under a wide range of different conditions
Theory of metal complexation 1539
r-----l
Formulate
quantitative
model
(3) Distribution Function), DEF (Differential Equilibrium
Function), SODF ( Site Occupation Distribution Function).
Sec. II. 1. summarizes the relationships between the discrete
(‘) values of classical complexation parameters and these con-
tinuous functions. It also discusses the relative degree of con-
ditionality of the latter, i.e., the extent to which they depend
on experimental data. Sets. II.2 to II.4 describe SADF, DEF,
and SODF. The SADF is the only rigorous distribution func-
tion representative of the complexation properties, but its
direct computation leads to severe artifacts. The SODF is
FIG.3. Approaches for interpreting simple ligand and heteroge-
based on the DEF, and it is a mathematically well-defined
neous complexant titration curves.
surrogate for the SADF.
plexation capacity (i.e., related to complexant concentration, the SADF and on that of the Differential Equilibrium Func-
an extensive variable). The complexation properties of nat- tion (DEF). the characteristics of which are shown in Fig. 4
ural systems will always be represented by functions of ex- and discussed below.
tensive vs. intensive parameters. But a number of different It is important to point out that DEF and SODF are con-
functions can be used, two extremes being the rigorously non- ditional for one or both of the following reasons in addition
conditional one, Cs,i = f( Ki) on the one hand, and the ex- to the “classical” conditionality factors (PH. T. and ionic
perimental data expressed as fl = g( [Ml) on the other. It is strength ) :
obvious that, for these two examples, the second will be more
conditional, i.e., will depend more on experimental conditions ( 1) Since the structure, and often composition, of most nat-
than the first. In particular, Cs,, and K,, in contrast to fl and ural complexants is unknown, the stoichiometry of the
[Ml, are independent of the metal to ligand ratio. The former metal coordinating groups is also unknown. As a con-
will, on the other hand, certainly necessitate mathematical sequence a complexing site can only be defined opera-
deconvolution with the accompanying high risk of artifact tionally as consisting of all functional groups involved
generation. In order to be able to make a valid comparison in the complexation reaction under the conditions ex-
of experimental results, a normalized representation (Fig. 3 ) isting at the titration point of interest ( ALTMANN and
should minimize the number of conditionality factors, which BUFFLE, 1988; BUFFLE, 1988). This definition of “site”
can only be done through comparison of many experimental does not make any reference to the chemical nature of
results obtained under appropriately varied conditions. The the coordinating group, and the corresponding stoichi-
normalization function proposed below, the Site Occupation ometry therefore becomes irrelevant. All metal-site re-
Distribution Function (SODF). has a level of conditionality actions are then treated mathematically on a 1 / 1 basis
somewhere between the log C’s,,vs. log Ki function (called a with no implication being made concerning the actual
Site Affinity Distribution Function; SADF), and the log /l nature of the coordinating groups. The coordinating
vs. log [M] function. The SODF is based on the concept of group nature. as well as the corresponding true stoichi-
.-:
-15 -12 -9 -6 -3 0
-12 9 6 12 9 6 3
?og KS0 log K”
differentiation
I .It_ integration
12 9 6 3 12 9 6 3
log Kp log I!
FIG. 4. (a-c) Steps involved in representing titration data for heterogeneous complexant systems. (d, e) Desired
thermodynamic description of the complexing system in terms of SADF or cumulative SADF ( =CADF). (The con-
tinuous system is the same as in Fig. 2; maximum values of [Ml,/{ S}, and Cs = 1 mole/kg and 1 mole/ 1, respectively.)
Theory of metal complexation 1541
ometry and all other equilibria between the site and the
other components of the system, will have to be deduced
from the changes observed in the above functions with
experimental conditions (step 2c in Fig. 3).
(2) The above definition of complexing sites means that the
corresponding complexation energy or equilibrium
“constant” will not be thermodynamic parameters but
rather an average of the free energies of all sites, weighted
-:12 -10 -8 -6 -4 -2 0
in a fashion dependent on the way in which they are
1% FI
computed. As has been discussed elsewhere ( ALTMANN
and BUFFLE, 1988; BUFFLE, 1988), the intensive param-
eter of the DEF, log K*, is a rigorous and direct means
of obtaining a mathematically well-defined and physico-
chemically meaningful value (see Sec. 11.3) for such a
complexation free energy.
It must be noted that CADF and SADF are complementary uncomplexed (i.e., having K, < 1/[M] and Hi+ 0).
functions. The DEF and SODF functions described below It may be shown (BUF’FLE et al., 1990) that this
are surrogates for the CADF and SADF, respectively. Most weighting factor causes K* to be controlled mainly
of the concepts presented in Sec. II are illustrated in Figs. 4 by the small fraction of sites which are near half-sat-
and 5 which show the above four functions calculated from uration at a given titration point. Nevertheless, because
a simulated titration data set. The ligand mixture used (see it always incorporates some influence from all sites
legend of Fig. 2a’-d’ and Sec. III. 1) was chosen ( 1) to simulate present, the K* value calculated at any given point in
characteristics of real natural materials and (2) to exhibit a titration obviously cannot be associated with only
several inflections or peaks. Although this is not frequently a single site’s K< value.
the case with natural systems, this situation was deliberately
chosen to exemplify the properties of these functions and (2) A significant advantage of K* is its direct and simple
computation from a continuous function fit of experi-
their potential for finding the properties of unknown systems.
mental data without requiring knowledge of the total
molar site concentration, Cs. Indeed, K* can be written
11.3. Differential Equilibrium Function (DEF) ( BUFFLE, 1984)
(i) the complexing site composition (i.e., the form of the 11.4. The Site Occupation Distribution Function (SODF)
SADF) through the Ki, AXi terms. Sites having larger
K values or present in larger proportions have a stronger Any method of representing and interpreting titration data
influence than weaker complexing sites or those present of natural system cation complexation must be judged by its
in lower concentrations; capacity in answering the environmental questions raised in
(ii) the degree of occupation of each site type i, i.e., the the first part of this paper. These concerns can be grouped
factors 0i* ( 1 - 0i) in Eqn. ( 1). These terms vary con- depending on whether they deal with system behaviour at
tinuously during a titration, each one having a max- the microscopic or macroscopic level. The former category
imum value at and being symmetrical about the point of questions concerns true reaction stoichiometry and free
log [M] = -log Ki and falling off steeply for log [Ml energy while the latter, as mentioned previously, includes the
greater or less than -log Ki. This means that those overall cation buffer capacity of the system and its Bee metal
sites being actively titrated at a given point in a titmtion ion activity. What we would like to point out is that a well-
(e.g., with 0.05 5 0i I 0.95) will have a greater weight conceived normalized representation of the complexation
than those which are already near saturation (i.e., properties of a system (even if tied to specified conditions)
having Ki $ 1/[M] and t$ -, 1) or which are nearly provides a powerful means of addmssing both types of con-
Theory of metal complexation 1543
terns. First, it can be expected to allow prediction of the polyfunctional complexants giving titration responses char-
partitioning and buffering behaviour in a particular environ- acteristic of natural complexants, B’/( CM/ { S} ,) is usually
ment; in addition, observation of changes in the form of the close to 0.5 (BUFFLE et al., 1990) in which case log K* is
normalized representation can be correlated with variations approximately equal to -log [M] at any point in the titration.
in experimental conditions (e.g., pH, competing metal con- Under these conditions there is clearly a close analogy be-
centrations, ionic strength), providing valuable insight when tween
searching for and testing hypotheses concerning the nature
of complexation reactions. Given the above-mentioned dif- d(CMI{S}t) t=8) and _ d([Mlb/{S)t) (_*)
a’log[Ml d log K*
ficulty of extracting the true SADF from experimental data,
we propose here a function based on K*, therefore directly or between the functions
calculable from titration data, which mimics the form of the
SADF. This function is called the Site Occupation Distri- p =f(log [MI) and w* = g(log K*).
bution Function, SODF. This parallel is important given the central roles played by j3
As for all complexation representation functions, the Site and [M] in defining the behaviour of A4 in any system
Occupation Distribution Function (SODF) expresses a con- (Fig. 1).
centration (extensive variable) as a function of an identi@ing
intensity parameter. The intensity variable employed is
log K* for the reasons presented in the foregoing sections. III. CALCULATION OF SODF FROM SIMULATED
AND REAL SYSTEMS
The concentration parameter, since it must also be directly
calculable from titration data, is taken as the differential The utility of Site Occupation Distribution Functions will
change in total complexed metal concentration (normalized be exemplified in this section using both simulated and real
in terms of total complexant quantity, { S} ,) associated with data sets. The simulated data is used to demonstrate ( 1) the
a differential change in log K*. The dependent variable of similarity existing between the SODF and the actual SADF
an SODF is then defined as for a range of complexant systems and (2) that SODF struc-
ture (i.e., peaks) may greatly aid in elucidating the nature of
w,, = _ dt[Mlb/{S>t)
(3) the sites.
dlogK* ’
Examples in this section are restricted to conditions most
In the following we shall represent the SODF as often encountered in natural waters, i.e., 0 5 0.1, where 0
= [M]b/CS is the degree of site saturation. As explained in
log w* = f( log K* ).
Sec. IV (Fig. 8)) under these conditions complexant properties
The usefulwss of an !3ODF at the microscopic level derives are dominated by those of the minor sites. These sites can
from the mathematical analogy existing between the depen- be divided into ( 1) a large number of background site types,
dent variable of the SODF, w*, and that of the SADF. The whose concentration increases in a continuous manner in-
latter is rigourously defined as dCs/d log K (Sec. 11.2) which versely to their log K value, and (2) a few dominant sites,
clearly shows the analogy with o* (Eqn. 3). It has been shown whose concentrations are significantly larger than those of
(Sec. 11.2) that the dependent variable of SADF represents the background sites having similar K values. The former are
the concentration of sites having log K values in a particular responsible for the globally linear shape of the DEF and SODF
d log K interval. Similarly, o* is the concentration of metal of most natural complexants ( SFQSITO,1980; ALTMANNand
actually complexed to produce a small change, d log K*, in BUFFLE, 1988) (Fig. 7). The latter are responsible for the
the system. The important point is that, since K* is closely small deviations from linearity sometimes observed on these
related to K, it may be expected that SODF (log w* = f( log curves (Fig. 4b, c).
K*)) may be used as a good estimate of SADF (log (dCs/d
log K) = f(log K)) (Fig. 4c, d). Indeed, as we shall see in 111.1. Simulated Data
the next section, the SODF calculated from titration data
closely reflects the shape of the underlying SADF. III. 1.1. Simulating background and dominant sites
The usefulness of an SODF at the macroscopic level derives
from the analogy between w* and the overall metal buffer The behaviour of the background minor sites is simulated
intensity of the system, j3, defined as by considering that it is well represented, over a wide range
of log K, by a linear DEF, i.e.,
B = dtCd{st)) (4)
d log [MI log ([M]b/(S}t) = r-log K; - r-log K* (6)
(by analogy with the acid-base buffer intensity, &, = -dCH/ where I’ and Kz are constants. For conditions realistic for
d( pH) where dCH is the change in total acid concentration). trace metal complexation, the slope, r, generally falls near
Furthermore, under conditions where [M] < CM and there- the middle of the range 0.2 5 r 5 0.8. Conversion of DEF
fore CM - [M]b, it can be shown (see Appendix A) that showing this logarithmically linear behaviour into the cor-
responding SODF also results in a linear function
K* = 1 - w’/(cM/{s)t)l
(5) logo* = log[-d([M]b/{S},)/dlogK*]
[Ml *M’/(Gd~s),)l
where 8’ = 812.303. It can be demonstrated that for highly = log [2.303’r’[M]b/{S}t] (h)
I544 J. Buflle et al.
or. with Eqn. (6). MINEQL). In Figs. 2, 4-6, 8, and 9, this simulated titration
curve is used, in turn. to compute the corresponding DEF
logw* =log(2.303.T.(K;)“)- r.logK*. (7b) and SODF from Eqns. ( 2 ) and ( 3 ), to allow comparison of
It has been discussed elsewhere (ALTMANN and BUFFLE, these functions with the underlying CADF (Eqn. Bl 1 ) and
1988) that this type of behaviour implies that the corre- SADF ( Eqn. 9 ).
sponding SADF. i.e.. the function p(log K) = L/X/~ log K
(Sec. Il.?), follows a Sips distribution (SIPS, 1950).$ The III. 1.2. Comparison cfsimulated SADFs and SODFs
complete Sips distribution is a complicated mathematical
function which corresponds to the generalized Freundlich Figure 5a shows the working curves for three complexing
isotherm systems having different values for the parameter, r, deter-
e= (G*[W)’ mining the spread ofthe Sips contribution to the SADF. The
corresponding SODFs are shown on Fig. 5b, c, d ( solid lines)
1 + (K,,*[Ml)’
and can be compared with the SADF distributions (dotted
where K,,, and r are empirical parameters. The shape of the lines) used to generate the titration data. The total concen-
complete Sips distribution is similar to that of a gaussian trations of three sites (with log K, values of 4, 6. and 8) were
distribution, K,,, being the value of K at the distribution max- increased to simulate three dominant sites (see legend of Fig.
imum ( ALTMANN and BUFFLE, 1988). However, under the 5). Also shown is a hypothetical analytical window serving
restricting condition 0 5 0.1, valid for the background minor to indicate that only a portion ofthe complete system response
sites, Eqn. (8a) simplifies into the classical Freundlich iso- will be observable for any given measurement technique.
therm The principal point to be noted in Fig. 5b, c. d is that the
shape of the SODF emulates closely that of the SADF. In
0 = (K,,,.[nci])“. (8b) particular
The corresponding part of the Sips distribution is given ( 1) the slopes, I’. of both functions are the same;
mathematically by the simple expression (Appendix B) ( 2 ) the peaks allowing detection of the dominant sites are
clear;
log @(log K))= log =logO-I’.logK (9) (3) the values of log K* of the dominant site maxima are
close to the corresponding “true” values (the chosen log
where 0 is a constant related to Kt and r (Appendix B). K,‘s ) . Furthermore the distances between these maxima
From Eqn. (9), any value of AX corresponding to a par-
ticular A log K interval (e.g., between log K, and log Kb)is
easily computed, by using the chosen values of K,*and r,
and the expression
Ax =
s l0g.q’
-a-
IO 9 a 7 6 5 4 3
IV. PRACTICAL UTILITY OF DEF AND SODF
The DEF and SODF functions were derived to aid in mak-
ing rigorous interpretations of the complexation behaviour
Jog IK’ / I .mole-jl
dominant
discrete model based on a priori physico-chemical assump- CADF) can be estimated (Sec. IV.3.2) and incorporated into
tions (step 2’ in Fig. 3). Good examples of this approach equilibrium calculation models (Sec. IV.3.1.).
include the interpretation of metal complexation by the car-
boxylate and phenolate groups of fulvic compounds (MAR- IV.3.1. Usefulness of linear representations:Incorporation
INSKYet al., 1980, 1982) or by the =Me-OH groups of in predictive models based on Langmuir isotherms
silicon, iron, and aluminum oxyhydroxides ( SCHINDLERet
al., 1976; SCHINDLER,198 1; DAVISand LECKIE, 1978; HOHL Before discussing how the SADF and CADF of background
and STUMM, 1976; STUMM et al., 1980) synthesized under sites can be estimated from their complexation properties
laboratory conditions (thereby minimizing the contribution represented by linear DEF (Eqn. 6), the advantages of this
of chemical impurities and physical non-homogeneities). linear representation must be discussed and compared to the
Obviously, use of such models is not possible for the back- global parameters given by classical interpretations based on
ground continuum in the minor sites due to their low con- the empirical fitting of data by combinations of Langmuir-
centrations and close log K spacing. type isotherms. Such empirical parameters have been widely
used, the most popular ones being sets of a few ( 1 to 3)
coupled values for the global site concentration and equilib-
IV.2. Findllg the Nature and Thermodynamic K Values
rium “constants” for “pseudo-sites” as determined, for in-
of Dominant Minor Sites
stance, by a Scatchard plot (FISH et al., 1986). The use of
Classical interpretations of titration curves do not allow such parameters should be avoided for two reasons:
the thermodynamic K values of minor sites to be obtained, ( 1) It has been shown ( BUFFLE, 1984) that these global pa-
even if they are dominant, owing to a lack of significant in- rameters have no physico-chemical meaning and that
flections in titration curves (Fig. 8b). Observation of peaks the corresponding “pseudo-sites” have no relationship
on an SODF (Fig. 8d), on the other hand, enables estimation to real sites. The use of these empirical parameters pre-
of the nature and log K values of the corresponding dominant sented in the same form as the theoretical ones both in
sites as indicated below. the symbols used and in their mathematical application,
It should first be noted that the position of the peak of a can only confuse those workers entering the field and as
dominant site along the log K* axis depends only slightly on yet unaware of its difficulties.
the value of I, i.e., the characteristic of the background sites (2) Such a representation of complexation properties cor-
located in the same log K* range (Fig. 9a, b). Secondly, in responds to a summation of 1 to 3 Langmuir-type iso-
all cases, the log K* value is very near that of the thermo- therms. Since it is known that each isotherm can rep
dynamic K, K,,, of the dominant site (Fig. 9b), in the “nor- resent complexation properties over a maximum of 2
mal” range of I’ (0.2-0.8). Finally, the difference between decades of [M] at least 3 sets of couples (i.e., 6 param-
log K* and log K,,, is nearly independent of the position of eters) are necessary to cover the large complexation range
log K* with respect to the K value of the major sites, K, exhibited by natural heterogeneous complexants (e.g.,
(Fig. 9c). These characteristics show that normalization of lo-’ < [M] < low3 M, i.e., 6 orders of magnitude of
an experimental titration curve into an SODF allows a good
[MI).
estimation of Km, from the value of log K* at the peak max-
imum (by using the curves in Fig. 9b, c to make the necessary On the other hand, the metal binding qualities of a com-
corrections based on the value of l? given by the slope of the plexant displaying a roughly linear DEF function (Eqn. 6)
background sites below the peak). can be summarized equally well using only the two param-
The Ktllleobtained as described above is obviously subject eters necessary to fit this DEF with a straight line. Such a
to the same conditionalities as the SODF. By varying the fitting yields the slope, I, and the intercept on the log K*
experimental conditions to generate various SODF functions axis, log K,* , corresponding to log ([Ml b/ { $} 1) = 0 (Fig.
followed by plotting the corresponding log K,,,,, as a function 11) . These two parameters provide an operational yet ratio-
of the varying parameter should allow one to obtain a non- nally normalized, twodimensional summary of the metal
conditional constant in the classical way. This may aid in binding behaviour of heterogeneous complexants over the
deducing the nature of the site based upon its value and its entire range of log [M] for which linear behaviour is observed.
change with pH or other parameters. In addition, log Kz and r are of direct practical utility. Kz
may be used to compare the complexing strength of different
IV.3. Finding the Parameters of Background Minor Sites complexants at equivalent values of [Mlb/ { S}, or of the
same complexant under different conditions. I can be shown
The sites comprising the background minor site continuum (BUFFLE et al. 1990) to be simply related to w* (Eqn. 3):
generally observed for natural complexants are especially
w*
important since they ( 1) represent most of the minor sites (lla)
and (2) serve to maintain the metal buffering action over r = [WLJ{~jt
many orders of magnitude of log [M] ( BUFFLEet al., 1990). and for minor sites, to the buffering intensity of the com-
This section describes a succinct overall representation of the plexant @(Eqn. 4):
complexant properties in the range dominated by the back-
ground site characteristics (Sec. IV.3.1). It also shows how B
( 1lb)
the true theromodynamic representation of these sites ( SADF, r = CM,/{S>, .
1548 J. Buffle et al.
In contrast to the empirical parameters obtained by the of the equilibrium constant, and such that the overall be-
above mentioned classical approaches, 1 has, therefore, the haviour of the system follows Eqn. ( 13). A mathematical
advantage of being linked to a well-defined property of the expression for the corresponding p( log K) function has been
system. In addition, K: and T allow realistic predictions of derived by SIPS ( 1950) and discussed by SPOSITO ( 1980 ).
complexation behaviour to be made outside the tested ex- Simpler derivations of this SADF and the corresponding
perimental range. This is an additional advantage over the CADF are given in Appendix B.
global empirical parameters discussed above, which have a In particular the CADF is expressed by
poor predictive capacity because they cannot adequately
i0gx = i0g(o/2.303.r) - r.i0gK (15)
simulate the fact that complexation properties of natural sys-
tems are buffered over wide ranges of log K values. which shows that CADF and DEF (Eqn. 6) are parallel
An often-mentioned limitation of representing complexant straight lines, thus justifying, as previously mentioned, the
properties by distribution functions is that they cannot be fact that the DEF is a good analogue for the CADF. The
readily incorporated into classical computer programs for vertical difference between the two lines is obtained by sub-
the calculation of a species distribution at equilibrium because tracting Eqn. (6) from Eqn. ( 15), for log K = log K*. By
these programs require site concentration/equilibrium con- making use of Eqns. ( 14) and (BS), one gets
stant couples as input data. This limitation however is not
applicable to linear DEF since the mathematical expressions
of the corresponding CADF and SADF are known (Sec.
IV.3.2: Appendix B). In particular, numerical values of the
function p( log K) = dx/d log K representing the SADF are
easily computed (Sec. IV.3.2) from the values of Kz and T
obtained from the DEF. As explained in Sec. III. 1, the p( log
K) function can then be split into small discrete A log K The interesting feature of Eqn. ( 16) is that it depends only
intervals which enable one to compute the corresponding on l’ and, in particular, is not dependent on K. This implies
mole fraction of sites AX (Eqn. 10). The (AX, log K) couples that it is independent not only of the nature of the chemical
can then be fed into classical predictive computer models, system, but also of the units used for the distributions, pro-
like MINEQL, for computing species distribution. viding, of course, that the DEF and CADF are expressed in
the same units. Indeed, in this case {S}, = Cs and the last
term of the right-hand side of Eqn. ( 16) is zero.
IV.3.2. Finding CADF and SADFfrom linear DEF Another important feature of the function represented by
Eqn. ( 16) is that it is only slightly dependant on I’, as is
For the linear DEF profiles often observed for background shown in Fig. 10. Thanks to this characteristic, it is easy to
sites, [Mlb/ { S}, is related to K* by Eqn. (6): estimate the theoretically meaningful CADF corresponding
to the experimentally obtained linear DEF: the two distri-
log ([M]J{S)J = I’slog K; - Telog K*. (6)
butions are parallel lines separated by the value given in Fig.
Furthermore, it has been shown that, for [M] 4 [M] b, K* 10. This value is most often close to -0.17 since, in the vast
can be expressed by Eqn. (5). By taking into account the majority of cases, 0.2 < T < 0.8. Note that in practice Cs is
definition of 0 (Eqn. 4) and the fact that, for linear DEF, I? generally not measurable. In this case, only the CADF ex-
=@/(2.303~C~/{S},)(BUm~etal.. 1990),Eqn.(5)can pressed by the cumulative concentration of sites per mass of
be rewritten as complexant (x * C’s/ { S} *) can be obtained from Eqn. ( 16 ).
i-r V. RELATlON BETWEEN DEF AND SODF, AND
K*=[M].r. ( 12)
GLOBAL COMPLEXANT PROPERTIES
Combining Eqns. (6) and ( 12) yields It has been shown in the previous sections that DEF and
SODF are good estimates of the rigorous thermodynamic
(13) functions (CADF and SADF) representing the complexing
properties of natural systems, and that they may aid in finding
the characteristics of both dominant and background sites.
with 0 < r < 1. Equation ( 13) is the description of the
They are more informative and preferable, for predictive
Freundlich isotherm (Eqn. fib) where 0 ( = [M]L,/CS) is re-
purposes, than the global empirical parameters classically
placed by [Mlb/ { S} t. Equations (8b) and ( 13) are related
used. However, since these latter are widely used, their re-
by the proportionality constant {S} JCs, which depends only
lationships to DEF and SQDF are worth addressing. The two
on the units chosen for ( S} t (e.g., g/l). Consequently, the
types of global parameters used most frequently for charac-
parameter K,,, in Eqn. (8b) is related to Kt and I’ by
terizing natural complexants are ( 1) some sort of average
K = fc*r {SJt “I- complexation “constant” (or distribution coefficient) and (2)
m
l-rcs. ( 1 (14) a single complexation capacity. Such single global average
values are most often meaningless since the K* of the SODF
The SADF (p( log K) = d x / d log K) corresponding to the for the above mentioned complexants extends over several
isotherm represented by Eqn. ( 13) is the probability distri- decades ( BUFIXE, 1988). A need exists, however, for a suc-
bution function for finding sites with any particular value K cinct means of classifying such complexants; we propose re-
Theory of metal complexation 1549
r
rameter more closely related to a normalized complex-
ation energy of the sites.
FIG. 10. Variation with r in the difference between the ordinate (3) A preferable alternative for this reference complexation
values of linear DEF (expressed by log 0) , and of the corresponding intensity parameter would be to employ K* and use the
linear CADF (log x), at K = K*. (See Sec. IV.3.2. and Eqn. 16 for concept of the SODF for defining a K$,-based CC, CC*.
details). Note that, for comparison with Eqn. ( 16), the ordinate is Considering the definition of w* given in Eqn. (3) and
equal to log x - (log [MlbI{S)t - log{S)JCd.
the schematic representation of Fig. 11b, CC* can be
defined by
m
placement of these global, mathematically ill-defined param- CC* = . a*- dlogK*. (17)
eters, by more meaningful ones based on K*. s b&Cl,.
This integration is done automatically in the calculation
V.l. Complexation Capacity
of the DEF curve from the titration working curve as
In the following discussion we are excluding site concen- shown in Fig. 4a, b. Any log K* on a DEF curve (log
tration estimates made from acid-base titrations, functional ([M]J { S} ,) vs. log K*; Fig. 1 lc) therefore could be
group analyses, tritium exchange, etc. We consider only CC taken as a specification for a particular value of CC*.
values measured by titrating the complexant by the metal of The advantages of using CC * rest in ( 1) its direct cal-
interest and defined as the total metal bound per unit mass culation from titration data, (2) its use of a normalized
of complexant as measured under some set of selected ex- binding intensity parameter, K*, as a reference, and (3)
perimental conditions. The purpose of measuring CCs is to the fact that the choice of the value of K$n to be used
allow inter-comparison of complexants based on their total for CC* can be made based on the shape of the SODF
complexing site concentrations. As discussed below (Fig. curve. This latter point is important since SODF enables
11b), this comparison has significance only if one of the fol- us to discriminate between different types of complexing
lowing two criteria is met: sites of the system (Sec. IV. 1).
FIG. 11. Comparison of different definitions of complex&on capacities: CC, (true total completing site concentration),
CC& (the complexing site ~oncen~~on corresponding to a particular value of [Ml, [Ml,; Fig. 1la) and CC* (the
complexing site concentration obtained for a particuhrr log K* value, log K& ; Fig. 11b). (a) Comparison of CC, and
CCM in a working curve. (b) Comparison of CC* with CC,. (c) Significance of CC,, CC,, Kz and r in a DEF.
to be occupied, some near their half&ration point, and some tration jumps are observed). This property of the background
close. to saturation. The composite behaviour yields both the sites is important ecologically since it helps living organisms
slope smoothness and the average value of I’ close to 0.5 as to adapt to chemical changes in their environment.
it is described in detail in BUFFLE et al. ( 1990). As a con-
sequence of this property, the complexant is able to “absorb” V-3. Heterogeneity Limits
the metal over a large concentration range (several orders of
m~itude) with only a gradual increase in the biologicahy The simulation used to compute the DEF and SODF
critical free metal ion concentration (i.e., no sharp concen- curves in this paper (in particular, Fii. 4b) assumes that there
Theory of metal complexation 1551
large _
log K*
i
c :8 -’
-3 -2
SO
,E1EI_
P
-5 -4
z
F -7 -6
-a -9 -10 -11 1 0 -1 -2 -3 -4 -5
logK*
FIG. 12. General shapes of log ( [Mlb/ { S},) = f(log K* ) curves. (a) Schematic curves for various degrees of
heterogeneity. Since each analytical method has a limited analytical “window,” only a portion of the overall curve will
be “seen.” (b) Cd adsorption on amorphous Fe oxyhydroxide (BENJAMIN and LECKIE,1981). (c) Cd, Cu, and Zn
adsorption on goethite in seawater medium ( BALESTRIERI
and MURRAY,1983)
is no upper limit in the log K* values for sites present in a entirely free. For highly heterogeneous complexants this
system. For real chemical systems, however, even extremely zone tends to be roughly linear with a slope in the vicinity
heterogeneous ones, there must exist some site with the max- of 0.5 ( BUF~E et al., 1990).
imum log K* value. If this is so then the leftmost portion of
the DEF curve must tend towards a vertical line (i.e., constant For most natural systems only one or two of the above
K* ) , One can, therefore, expect the shape of DEF curves for DEF zones may be observed because the analytical window
heterogeneous complexants to be as schematically shown in (Figs. 2, 12) is not sufficiently wide to allow determination
Fig. 12a. Three regions of these idealized curves can be iden- of the entire curve. This representation is of course largely
tified as conceptual, but one may note that curves having similar
shapes (Fig. 12b, c) have been observed in a few cases for
( 1) one which is asymptotically horizontal ([Ml,,/ { S} r + metal complexation with metal ozides (BENJAMIN and
Cc) corresponding to the region where complexation is LECKIE, 198 1; BALISTRIERIand MURRAY, 1983 ) .
controlled by major sites (minor sites being fully satu-
rated);
(2) one which is asymptotically vertical corresponding to VI. CONCLUSION
the state where even the strongest sites are largely free;
(3) an intermediate zone where the minor sites are partially This paper shows that DEF and SODF have useful poten-
occupied by the metal and the major ones are almost tialities for representing complexation characteristics of nat-
1552 J. Buffle et al.
Ural systems, as well as for predictive purposes. They are plexation at the oxide/water interface. II. Surface properties of
more informative and closer to the rigorous thermodynamic amorphous iron oxyhydroxide and adsorption of metal ions. J.
Colloid Interface Sci. 67, 90- 107.
description than the classically used global parameters. In
FILELLAM., BUFFLEJ.. and VAN LEEUWENH. P. ( 1990) The effect
particular, as opposed to the latter, they are free of any a of physico-chemical heterogeneity of natural complexants. I. Vol-
priori assumptions, making them particularly useful for ob- tammetry of labile metal-fulvic complexes. Anal. Chim .4cta. (in
jective comparison of complexation behaviours of different press )
systems. FISH W., DZOMBAKD. A., and MOREL F. M. M. ( 1986 ) Metal-
humate interactions. 2. Application and comparison of models.
The main limitation to this approach is the same as for Environ. Sci. Tech. 20, 676-683.
computation of any type of site affinity spectrum: it is linked GAMBLED. S. ( 1970) Titration curves of fulvic acid: The analytical
to the propagation of errors during the mathematical decon- chemistry of a weak acid polyelectrolyte. Canadian .I Chem. 48,
volution of the raw data into the DEF and SODF. Indeed, 2662-2669.
GAMBLED. S. and SCHNITZERM. ( 1973) The chemistry of fulvic
although K* is obtained by applying a simple equation (Eqn.
acids and its reactions with metal ions. In Trace Metal and Metal
2) free of any a priori assumption, its computation neces- Organic Interactions in Natural Waters (ed. P. C. SINGER), pp.
sitates derivation of the curve In (C,/ { S} t) = f(ln (Y). As 265-302. Ann Arbor Science Publ.
a result of this step, the error in the DEF, and particularly GAMBLED. S., UNDERDOWNA. W., and LANGF~RDC. H. ( 1980)
the SODF, may become large due to the errors in the exper- Copper (II) titration of fulvic acid ligand sites with theoretical,
potentiometric, and spectrophotometric analysis. Anal. Chem. 52,
imental measurements. Curve fitting techniques may be em- 1901-1908.
ployed to overcome this problem, but classical procedures GREENLANDD. J. and HAYES M. H. B. (eds.) ( 1978) The Chemistry
frequently introduce artifactual oscillations or production of ofSoil Constituents. J. Wiley & Sons.
apparent jumps or peaks which are transferred to and am- HOHL H. and STUMMW. ( 1976) Interactions of Pb’+ with hydrous
plified in the DEF and SODF. The criticalness of the data y-A1203. J. Colloid Interface Sci. 55, 28 l-288.
HUNSTOND. L. (. 1975), Two techniaues for evaluatina small mole-
fitting step cannot be overemphasized since any subsequent cule-macromolecule binding in complex systems. Anal. Biochem.
physico-chemical interpretation is entirely dependent upon 63,99-109.
it. This, however, is a purely mathematical problem, different LUOMAS. N. and BRYANG. W. ( 1981) A statistical assessment of
from the conceptual ones discussed here. A curve fitting pro- the form of trace metals in oxidized estuarine sediments employing
chemical extractants. Sci. Total Environ. 17, 165-196.
cedure addressing the above questions has been developed MARINSKYJ. A., WOLF A.. and BUNZLK. ( 1980) The binding of
for computing DEF and SODF and will be described else- trace amounts oflead( copper( cadmium(II), zinc(I1) and
where. calcium ( II) to soil organic matter. Taianta 27, 46 l-468.
MARINSKYJ. A., CURA S., and SCHINDLERP. W. ( 1982) A unified
physico-chemical description of the equilibria encountered in
Acknowledgments-This work was supported by the Swiss National humic acid gels. J. Colloid Interface Sci. 89, 4 12-426.
Foundation (Projects Nos. 2.07 l-O.86 and 2000.5.5 12). SCHINDLER P. W. ( 1981) Surface complexes at oxide-water interfaces.
In Adsorption ofInorganics at Solid-Liquid Interfaces (eds. M. A.
Editorial handling: J. I. Drever ANDERSONand A. J. RUBIN), pp. l-49. Ann Arbor Science Publ.
SCHINDLERP. W., FURST B., DICK R., and WOLF P. U. (1976)
Ligand properties of surface silanol groups. I. Surface complex
REFERENCES formation with Fe3+, Cu 2+1Cd2+. and Pb’+. J. Colloid Interface
Sci. 55, 469-475.
ALTMANNR. S. and BUFFLEJ. ( 1988) The use of differential equi- SHUMANM. S., COLLINSB. J., FITZGERALDP. J., and OLSOND. L.
librium functions for interpretation of metal binding in complex ( 1983) Distribution of stability constants and dissociation rates
ligand systems: Its relation to site occupation and site atfmity dis- among binding sites on estuarine copper-organic complexes: Ro-
tributions. Geochim. Cosmochim. Acta 52, 150% 1519. tated disk electrode studies and an affinity spectrum analysis of
BALISTR~ERJ L. S. and MURRAYJ. W. ( 1983)Metal-solidin&actions ion-selective electrode and photometric data. In Aquatic and Ter-
in the marine environment: Estimating apparent equilibrium restrial Humic Materials (4s. R. F. CHRISTMANand E. T. GJESS-
binding constants. Geochim. Cosmochim. Acta 47, 109 1- 1098. ING), pp. 349-370. Ann Arbor Science Publ.
BENJAMINM. M. and LECKIEJ. 0. ( 198 1). Competitive adsorbtion SIGGL. ( 1985) Metal transfer mechanisms in lakes; the role of settling
of Cd, Cu, Zn and Pb on amorphous iron oxyhydroxides. .I. Colloid particles. In Chemical Processes in Lakes (ed. W. STUMM), pp.
Interfce Sci. 83.410-419. 283-3 10. J. Wiley & Sons.
BU~E J. ( 1984) Natural organic matter and metal-organic inter- SIPSR. ( 1950) On the structure of a catalyst surface. J. Chem. Phys.
actions in aquatic systems. In Circulation of Metals in the Envi- 18, 1024-1026.
ronment (ed. H. SIGEL), pp. 165-221. Metal Ions in Biological SPIEGEL( 1968) Mathematical Handbook of Formulas and Tables.
Systems 18. M. Dekker, New York. Schaum’s outline series. McGraw Hill Book Company.
BUFFLE J. ( 1988) Complexation Reactions in Aquatic Systems: An SPOSITOG. ( 1980) Derivation of the Freundlich equation for ion
Analytical Approach. Ellis Honvood, Chichester. exchange reactions in soils. Soil Sci. Sot. Amer. J. 44,652-654.
BUFFLEJ. and ALTMANNR. S. ( 1987 ) Interpretation of metal com- SFQSITO G. ( 1986) Sorption of trace metals by humic materials in
plexation by heterogeneous complexants. In Aquatic Surface soils and natural waters. CRC Crit. Rev. Environ. Control 16, 193-
Chemistw: Chemical Processes at the Particle- Water Interface
” (ed. 229.
W. STUMM), pp. 351-383. J. Wiley & Sons. STUMMW. and MORGANJ. J., ( 1981) Aquatic Chemistry. An In-
BUFFLE J., ALTMANN R. S., and RLELLA M. (1990) The effect of troduction Emphasizing Chemical Equilibria in Natural Waters.
physico-chemical heterogeneity of natural complexants. II. The J. Wiley & Sons.
buffering action and role of their background sites. Anal. Chim. STUMMW., KUMMERTR., and SIGG L. ( 1980) A ligand exchange
Acta (in-press). model for the adsorption of inorganic and organic ligand at hydrous
CABANN S. E.. SHUMANM. S.. and COLLINSB. J. ( 1984) Metal- oxide interfaces. Croat. Chem. Acta 53,29 l-3 12.
organic bind&: A comparison of models. In Complexation of TANFORDC. ( 1961) Physical Chemistry of Macromolecules. J. Wiley
Trace Metals in Natural Waters (eds. C. J. M. KRAMER and J. C. & Sons.
DUINKER), pp. 165-179. Martinus Nijhoff/Dr. W. Junk Publish- TE~SIERA., CAMPBELLP. G. C., and B@.SONM. ( 1979) Sequential
ers, The Hague. extraction procedure for the speciation of particulate trace metals.
DAVISJ. A. and LECKIEJ. 0. ( 1978) Surface ionization and com- Anal. Chem. 51.844-85 1.
Theory of metal complexation 1553
THAKURA. K., MUNSONP. J., HUNSTOND. L., and RODBARDD. 11- B’l(chllI~~t)l
( 1980) Characterization of @and-binding systems by continuous (AZ)
K* = [Jw~w’I(chd~~~,) - [WCMI
affinity distributions of arbitrary shape. Anal. Biochem. 103,240-
254. where /3’ = ,8/2.303. For most natural systems, /3’/( C,/{ S},) is
TURNER D. R., VARNEY M. S., WHITFIELD M., MANTOURA close to 0.5 ( BUFFLEet al., 1990), so that under conditions of strong
R. F. C., and RILEY J. P. ( 1986) Electrochemical studies of copper complexation, where [M] 6 CH, Eqn. (A3) simplifies to Eqn. (5).
and lead complexation by fulvic acid. I. Potentiometric measure-
ments and a critical comparison of metal binding models. Geochim. APPENDIX B
Cosmochim. Acta SO.289-297.
An SADF ( = d x / d log K) is the probability, p( log K) , of finding
sites with any particular value, K, of the equilibrium constant.
GLOSSARY p( log K) also expresses the mole fraction, dx, of the sites whose K
valuesare within the interval d log K. The relation between 0 ( = [A&,/
Cs) and p( log K) is given by ( ALTMANNand BUFFLE, 1988)
COMPLEXANT: Any compound, organic or inorganic, dissolved,
colloidal, or particulate, containing one to several complexing sites.
HETEROGENEOUS COMPOUND: A compound presenting
physical and/or chemical heterogeneity (note: our use of “hetero-
B=
s m K.[M]
-.z, I + K.[M]
.p(logK).dlogK (Bl)
s-
which may also be written
geneous” is broader than in normal usage, i.e., systems containing
more than one distinct phase. Here an heterogeneous compound K.[M]
may be truly dissolved if it includes only chemical heterogeneity, 8= .p’(K).dK W)
-m I + K.[M]
e.g., fulvic compounds).
where
MAJOR SITES: One (or possibly a few) group(s) of sites which is/
are chemically homogeneous and which account for the large majority p’(K) = p(log K)/2.303. K. (B3)
of the sites of the complexant of interest.
Furthermore, mathematical tables (SPIEGEL, 1968 ) give
MINOR SITES All sites other than the major site type(s). For
heterogeneous complexants, this group accounts for most of the
chemical heterogeneity. sm
Xm/andx=
-m 1 + x”Ja”
m~ x”Ja” X(m-n). &
s-- 1 + x”/a”
*. a(m+l-n)
Background minor sites: The whole of the minor sites which, taken
together, produce a gradual change in the complexation parameter = sin [(m + 1)*/n]
(M)
(K, [M] ) without any predominant single site type.
whichisvalidforO<(m+ l)<n.Byreplacingn= l,x=K,and
Dominant minor site(s): A minor site present in a concentration [M] = I /a, in Eqn. (B4), and comparing the resulting right- and
much greater than neighboring sites, which dominates the com- left-hand sides of this equation with the right-hand sides of Bqns.
plexation properties of the system over a given restricted range of (8b) and ( B2), respectively, one obtains
log K(or log [Ml).
m=-F (B5)
SADF: Site Affinity Distribution Function (see Sec. II.2 and Figs. 2,
4). and
CADF: Cumulative Site Affinity Distribution Function (see Sec. II.2
and Figs. 2, 4). p’(K)=bsin[(l -r).*].(K,)r.K-(“r). (B6)
DEF: Differential Equilibrium Function (see Sec. II.3 and Fig. 4). Combining Bqns. (B6) and (B3) gives the SADF
SODF: Site Occupation Distribution Function (see Sec. II.4 and Fig. dx
4). -=p(logK)=O.K-r
dlogK
= 0.exp( -2.303. Felog K) (B7)
APPENDIX A
with
For systems where the complexant is titrated by M, i.e., the com-
plexant concentration is constant, d In {S},/d In (Y= 0, d lnCk( = d 8 = 2.303 - sin [ ( I - r). ST]
(Kn)“. (J38)
In C,/ { S},, and Eqn. (2) can be transformed into k