Journal of Colloid and Interface Science 303 (2006) 214–223

www.elsevier.com/locate/jcis

Parsons–Zobel plots: An independent way to determine surface

complexation parameters?
Johannes Lützenkirchen ∗
Institut für Nukleare Entsorgung, Forschungszentrum Karlsruhe, Postfach 3640, D-76021 Karlsruhe, Germany
Received 1 June 2006; accepted 18 July 2006
Available online 24 August 2006

Abstract
Parsons–Zobel plots can in principle be used to estimate inner Helmholtz-layer capacitance values and electrochemical surface areas for
mineral particles. Their application to aqueous suspensions of various minerals has been documented in the literature. For the experimental data
used so far, the expected linear relationship between the overall and the diffuse-layer capacitances has been reported. The extracted values either
have not been used at all subsequently in a surface complexation model to describe the mineral surface charge versus pH curves, or were found
not to be suitable entirely for such purposes. In the latter case, the reported failure was not explained. In one part of the present paper, the Parsons–
Zobel plot concept is tested with data generated from a surface complexation model, for which the interfacial structure closely corresponds to
that assumed in the application of the Parsons–Zobel plot. From the analysis of the results it turns out that electrolyte binding and non-Nernstian
surface potential–pH curves more or less strongly affect the outcome of Parsons–Zobel plots. Despite the fact that the analysis in this paper is
restricted to iron(III) minerals only, it is concluded, in general, that the use of Parsons–Zobel plots with aqueous mineral suspensions to determine
inner Helmholtz-layer capacitances for subsequent application to surface complexation models cannot be recommended, since the reasons for
failure can be traced very nicely with applications to model-generated data. Such application requires the determination of further parameters,
and it was found that low electrolyte binding and Nernstian slopes should be imposed. Of these two issues, the more important is electrolyte
binding. For the surface complexation models, an inner Helmholtz-layer capacitance and weak electrolyte binding were required for a good fit to
experimental data. The values of the electrolyte binding constants required to achieve this end are in conflict with the assumptions of the Parsons–
Zobel plot (absence of specific adsorption). However, these parameters would not necessarily cause specific adsorption in terms of a classical
colloid chemistry definition (i.e., would not shift isoelectric points). The electrochemical surface areas were found to be in good agreement with
the value used to generate the data. Based on this, there is a potential for using the approach to determine surface areas in situ from titration curves.
Consequently, in a second part of the paper, Parsons–Zobel plots are applied to experimental data with the objective of determining electrochemical
surface areas in situ. Application to various sets of published experimental titration data for hydrous ferric oxide yielded consistently very large
electrochemical surface areas for fresh samples. This can be explained by very small particles and/or inclusion of substantial amounts of water in
the suspended particles. As would be expected, the electrochemical surface area for aged ferrihydrite was found to be substantially lower.
© 2006 Elsevier Inc. All rights reserved.

Keywords: Surface charge; Mineral suspensions; Surface complexation model; Parsons–Zobel plot; Specific adsorption; Nonspecific adsorption; Electrolyte
binding; Specific surface area

1. Introduction pacitance measurements, which are possible in such electrode

The Parsons–Zobel plot was originally applied to electrode
systems [1] to determine capacitance values and to evaluate
whether specific adsorption occurs at the electrode or not. Ca-
* Fax: +49 (0) 7247 82 3927.
E-mail address: johannes@ine.fzk.de (J. Lützenkirchen).
0021-9797/$ – see front matter © 2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcis.2006.07.037
systems, can be directly applied in the use of Parsons–Zobel
plots. This is usually not possible with mineral suspensions,
which are the focus of the present work. As indicated above,
the approach of Parsons and Zobel was originally advocated
as a means to detect “specific” adsorption of (electrolyte) ions.
Subsequently, the approach has been applied with the neces-
sary adjustments to suspensions of oxide minerals [2,3]. Here,
the purpose was to determine the (electrochemical) surface area
 J. Lützenkirchen / Journal of Colloid and Interface Science 303 (2006) 214–223
of the minerals and the inner layer capacitance of the min-
eral/electrolyte interface. From the point of view of surface
complexation modeling, additional information on in situ elec-
trochemical surface areas and a priori knowledge of specific
capacitance values pertaining to a simple but useful picture
of the mineral/electrolyte interface [2–4] would be very help-
ful.
Specific surface area is important in such models, since, to-
gether with the site density of surface functional groups, it will
determine the available surface sites in a given adsorption ex-
periment. In situ determined electrochemical surface areas may
be much more relevant to surface complexation studies than
surface areas obtained from gas adsorption studies on dry pow-
ders. Values for the latter may include certain problems, some
of which are summarized below:
– there is no unique standard method for the gas adsorption
procedure:
• different commercial or lab-built setups exist (e.g., one-
point, three-point, or full isotherm), which potentially
yield different values of the specific surface area and
generally cause a variation of data quality;
• different probe molecules exist, which allow a more or
less dense coverage of the surface, yielding (potentially
depending on issues like surface roughness) in different
values for the specific surface area of identical samples;
• even with one probe molecule such as nitrogen (as the
most frequently used probe), different orientations of the
molecule may be assumed, resulting in different specific
surface areas for one set of raw data;
– there is a need to degas the samples at higher temperatures:
• degassing for different times and at different tempera-
tures may influence results;
• exposure of minerals to higher temperatures may affect
their state to some extent;
– there is a danger of particles in the dry state being inacces-
sible to gas molecules:
• particles may stick together and part of the surface may
not be accessible to the probe molecules, e.g., gibbsite,
where a huge discrepancy has been reported between gas
adsorption results and AFM imaging [5];
• particles may stick to walls and part of the surface may
not be accessible to the probe molecules.
This summary justifies the question of whether an in situ
method would not be most relevant to any subsequent work
with suspensions. The EGME method [6] is such an approach.
The Parsons–Zobel plot concept would be an alternative way to
estimate suspension surface area in situ directly from titration
data, with the benefit of simultaneously yielding the specific ca-
pacitance value of the Stern layer.
Specific capacitance values are relevant to all surface com-
plexation models involving layers different from the diffuse
layer of mobile counterions. Thus the most realistic surface
complexation models are concerned. In particular, the basic
Stern model (BSM) includes only one such layer, and its capaci-
tance could be directly inferred from experimental data without
215
the need to fit titration data and to involve parameter interde-
pendencies of the capacitance with site density and/or stability
constants [7].
As a consequence, the Parsons–Zobel plot would allow con-
sistent determination of the capacitance value and an in situ
value for the specific surface area. These two parameters would
not be affected by the site density parameter, and parameter in-
terdependencies would be avoided. This would clearly be an
advantage, since at present the site density parameter cannot
be experimentally determined [8]. Even if there are approaches
to calculating and interpreting capacitance values for the triple
layer model, the inner layer capacitance currently largely re-
mains an adjustable (that is to say a fitted) parameter [9].
The Parsons–Zobel plot has been previously applied to min-
eral suspensions and was discussed in some detail by Trasatti
and co-workers [2,3]. Part of the discussions concerned particle
geometry, and the authors considered it necessary to introduce
the effect of particle geometry in the diffuse-layer capacitance.
They did not do likewise for the inner layer capacitance, which
is more difficult, since the thickness of the inner layer needs to
be known. Another part of the discussion concerned the Nerns-
tian behavior of the oxides, and the authors concluded that non-
Nernstian behavior would not affect the estimated electrochem-
ical surface areas. The Parsons–Zobel plot has been recently
re-discussed by Gunnarsson [4], and the conclusion was that its
application with experimental data to obtain parameters for sur-
face complexation models failed to yield satisfactory results for
several systems. For example the extracted Parsons–Zobel sur-
face areas were much higher than the measured BET surface
areas for hematite (factor of about 1.5, background electrolyte
NaNO3 ), magnetite (factor of about 4, background electrolyte
KNO3 ), and titanium oxide (factor of about 1.5, background
electrolyte KNO3 ). An explanation for the high electrochemi-
cal surface areas was based on problems with BET analysis on
dried particles as outlined above, and on the application of the
planar geometry and potential coagulation effects, in particular
at high ionic strength. As for the use of surface complexation
models, it is noteworthy that Gunnarsson [4] has used planar
geometry without electrolyte binding. This led to a good agree-
ment with data on goethite in NaClO4 as discussed in more
detail below. For hematite in NaNO3 the approach failed to
give an acceptable description of the surface charge data. It was
reported that inclusion of electrolyte binding and a significant
decrease of the capacitance was necessary to obtain a good fit
to the data. In some of the other data sets, it can be seen that
a good description was not possible with straightforward appli-
cation of Parsons–Zobel plot results either. This failure was not
fully explained.
In the present paper the Parsons–Zobel plot is applied to
model-generated data. For these, all surface complexation pa-
rameter values are known and the ability of the approach to
retrieve the known parameter values can be tested. One as-
sumption of Trasatti and co-workers [2,3] was that Nernstian
behavior of the mineral occurs in the vicinity of the point of zero
charge. This assumption can easily be tested with the gener-
ated data, since all potentials are obtained from the calculations.
Since the electrostatic model concept used in the calculations
216 J. Lützenkirchen / Journal of Colloid and Interface Science 303 (2006) 214–223
to generate the data (i.e., the BSM) is in principle consistent
with the one assumed in the Parsons–Zobel plot concept, a con-
clusion about the actual compatibility of the interfacial models
should be obtained. After the application to generated data, the
Parsons–Zobel plot is applied to experimental surface charge
data for hydrous ferric oxide (HFO). This system was chosen
because some speculation about the specific surface areas of
HFO can be found in the literature. The specific surface areas
typically measured on dry samples are about 250 m2 /g. How-
ever, in most modeling studies the value is set to 600 m2 /g
based on a recommendation by Dzombak and Morel [10]. The
system is considered as a good example of the disagreement
between gas adsorption measurements in the dried state and ac-
tual surface area in suspension and was, therefore, chosen here
to evaluate the Parsons–Zobel plot approach.
2. Theory
The overall (experimental) capacitance, Ctot,A in F/m2
(A indicates surface area), within the basic Stern model is given
by the equation
1/Ctot,A = 1/C1,A + 1/Cd ,
with the first r.h.s. term involving the Stern layer capacitance
and the second the diffuse-layer capacitance. The above equa-
tion can be related to the electrochemical specific surface area
(S in m2 /g) as follows:
S/Ctot,M = S/C1,M + 1/Cd .
The subscript M indicates that the overall and Stern layer ca-
pacitances are on a mass-specific basis, in F/g. The (surface-
specific) diffuse-layer capacitance at 298.15 K can be calcu-
lated from
Cd = εr ε0 κ
with εr the dielectric constant of water, ε0 the permittivity of
free space, and κ the inverse Debye length. This equation is de-
rived from the Gouy–Chapman equation either as the integral
or differential capacitance of the diffuse layer for sufficiently
low surface potentials at the head end of the diffuse layer (i.e.,
<26 mV). The condition of sufficiently low surface potential
holds in particular for the isoelectric point (IEP), which is in
general close to the point of zero surface charge. For aque-
ous solutions
 at 25 ◦ C, Eq. (3) results in Cd (IEP) (F m−2 ) =
2.285 Ic (mol dm−3 ) = 7 × 10−10 · κ (m−1 ), where Ic is the
ionic strength and κ the inverse Debye length. Note that the
diffuse-layer capacitance is always per area, whereas the Stern
layer and the overall capacitances may be per area or per mass.
Among the electrostatic interfacial concepts of the common
surface complexation models, the one that most closely corre-
sponds to that of the Parsons–Zobel plot approach is the basic
Stern model. That is why recently it was attempted to link
Parsons–Zobel plot results to this particular model. Within this
model C1 is the capacitance of the Stern layer and is related
to the measurable surface charge density, σ0 , and the unknown
interfacial potentials via
C1 = σ0 /(Ψ0 − Ψ1 ). (4)
In Eq. (4), Ψi is the interfacial potential in the adsorption
plane i. Plane “0” is the surface plane, whereas plane “1” corre-
sponds to the inner Helmholtz plane (IHP). For the basic Stern
model the potential at the IHP corresponds to the diffuse-layer
potential; i.e., Ψd = Ψ1 .
The application of Parsons–Zobel plots requires values for
the overall capacitance and the capacitance of the diffuse layer.
The overall (mass-specific) capacitance is obtained from the
first derivative of the mass-specific surface charge vs pH curve
at the pristine point of zero charge, (∂σ0 /∂pH)ppzc , in the fol-
lowing fashion:
Ctot,M = (∂σ0,M /∂Ψ0 )ppzc
= (∂σ0,M /∂pH)ppzc (∂pH/∂Ψ0 )ppzc . (5)
(∂σ0,M /∂pH)ppzc can be obtained from titration data. For
(∂pH/∂Ψ0 )ppzc in the original applications of Parsons–Zobel
plots to oxide suspensions, it was assumed that Nernstian be-
havior of the surface can be used. This has been pursued by
(1) Gunnarsson [4] based on work by Larson and Attard [11]. How-
ever, it is also possible to have another point of view based on
part of the available experimental data on oxides [12]. Over-
all, there is no agreement on the slope of surface potential pH
curves for oxide minerals, and therefore this is an inherent un-
certainty in the application of the Parsons–Zobel plot to oxide
minerals.
(2)
3. Data collection and data treatment
3.1. Generated data
Model data were generated with the speciation code ECO-
(3) SAT [13] based on goethite surface complexation parameters
given by Rietra et al. [14] for a range of sodium nitrate con-
centrations. Sodium nitrate was chosen because it is has been
shown that it does not shift isoelectric points of goethite (i.e.,
sodium nitrate is a nonspecifically adsorbing background elec-
trolyte, based on the common colloid chemistry criterion).
A mass-specific surface area of 100 m2 /g is a typical value for
this material, when it is prepared by slow titration of an iron(III)
solution, and this value was used in all calculations presented
here.
For all sodium nitrate concentrations considered, the output
from ECOSAT was used to calculate proton-associated (i.e.,
measurable) mass-specific surface charge in C/g as a func-
tion of pH. Furthermore, the potential in the surface plane was
retrieved from the output and plotted as a function of pH to
verify the assumption of Nernstian behavior inherent in the ap-
plication of Parsons–Zobel plots to mineral suspensions [2,3].
For a given sodium nitrate concentration, mass-specific surface
charge (in the surface plane) as a function of pH was spline fit
and the minimum derivative was used to calculate Ctot,M from
Eq. (5) (assuming Nernstian behavior in the very first step).
The results for the various ionic strengths were then plotted as
 J. Lützenkirchen / Journal of Colloid and Interface Science 303 (2006) 214–223
1/Ctot,M vs 1/Cd , calculated from Eq. (3). From linear regres-
sion S and C1,M were obtained; cf. Eq. (2). It was observed that
the minimum derivative of the charging curve coincided with
the point of zero charge. Recently, it has been suggested to use
derivatives of charging curves to determine the point of zero
charge [15], a procedure that is confirmed with generated data
here. Subsequently, the “true” model inherent surface potential
pH curves were used in the evaluations to replace the Nernst
assumption in Eq. (5).
To understand the importance of Nernstian behavior and of
electrolyte binding in a second step, the following parameter
variations were carried out, while keeping all other parameters
constant:
– variation of the electrolyte binding constant at the standard
capacitance value;
– variation of the capacitance value at the standard electrolyte
binding constant.
Increasing the electrolyte binding constant and the model ca-
pacitance value resulted in steeper charging curves and more
pronounced non-Nernstian behavior.
3.2. Experimental data
Different sets of HFO data were taken from Cox and
Gosh [16], Schwertmann and Fechtner [17], Swallow [18],
Hansen et al. [19], and Trivedi et al. [20]. They were trans-
formed to mass-specific surface charge vs pH curves and spline
fit. The minimum of the derivative was used to compute the
overall mass-specific capacitance assuming Nernstian behav-
ior. Its inverse was plotted vs 1/Cd as calculated from Eq. (3).
Linear regression yielded the values for the mass-specific inner
layer capacitance and the electrochemical specific surface area.
4. Results and discussion
4.1. Applications to generated data
The derivatives of the surface potential versus pH curves for
the model parameters according to Rietra et al. [14] are sum-
marized in Table 1. It can be noted that the slope of the surface
potential versus pH curves became more Nernstian when the
range of data was limited to values in the vicinity of the point
Table 1
Slope of surface potential versus pH curves for the goethite model of Rietra et
al. [14]
I in M
0.1
0.01
0.002
0.001
6 × 10−4
1 × 10−4
2 × 10−5
1 × 10−5
2 × 10−6
217
of zero charge (results shown in Table 1). Analysis over the
whole pH range yielded values with lower slopes. From the re-
sults in Table 1 it becomes obvious that Nernstian behavior is
approached when the ionic strength is lowered. However, true
Nernstian behavior was never reached, even with unrealistically
low values of ionic strength. The deviation between assumed
and actual behavior of the surface potential is an aspect already
discussed by Trasatti and co-workers [2,3] and is addressed here
with the objective of evaluating the applicability of Parsons–
Zobel plots for BSM parameterization.
The Parsons–Zobel plot for the model-generated data using
the complete set of parameters from Rietra et al. [14] for sodium
nitrate is shown in Fig. 1 and the outcome is summarized in
Table 2. This is compared to curves that would result directly
from the known input parameters using Eqs. (1) and (3). The
parameters derived from the treatment of the model-generated
data under the assumption of Nernstian behavior are 1.12 F/m2
for the capacitance of the Stern layer (compared to the value of
0.91 F/m2 used to generate the data) and about 104 m2 /g for
the specific surface area (compared to the value of 100 m2 /g
used to generate the data); cf. Fig. 1a. This could not be im-
proved by using the actual slopes of surface potential pH curves
(from Table 1); cf. Fig. 1b. Thus deviation from Nernstian be-
havior is not the reason for the obtained discrepancy in this
case. As noted by Trasatti and co-workers [2,3], deviation from
the ideal Nernst behavior is not expected to result in errors in
the estimated “electrochemical” surface area. Indeed, with the
generated data and a Nernst slope, a value close to the input
value is obtained (Table 2). Use of a different slope will just
parallel-shift the Parsons–Zobel plot and therefore not affect
the extracted electrochemical surface area (which depends on
the slope of the Parsons–Zobel plot only). However, an aspect
not discussed by Trasatti and co-workers [2,3] is the inherent
problem caused if the surface potential–pH slopes not only de-
viate from Nernstian behavior but also vary with ionic strength
(cf. Table 1). This will actually result in problems in estimat-
ing the electrochemical surface area and may explain the small
deviations obtained with the generated data. The deviations are
small since the changes in the slope of the potential–pH curves
are small. If the variation of the slopes with ionic strength were
to become larger, the error in the estimated electrochemical sur-
face area based on a constant Nernst slope would become more
important. Differences in activity coefficients at the surface and
in the bulk solution (apart from issues related to classical elec-
trostatic terms) may contribute to deviations from Nernstian
behavior and to the variations with ionic strength.
The electrochemical surface area furthermore pertains to the
Gouy–Chapman equation. It is referred to the same mass as the
Slope in mV
BET surface area, but due to the separation of the head end of
56.71
58.00
the diffuse layer, the electrochemical surface area should dif-
58.58 fer from the zero-plane surface area. It should be larger if the
58.73 zero-plane area is ideal. In the case of the flat-plane geometry of
58.82 the particles assumed here, this difference should not show up.
59.00 Though it might be of interest to investigate the issue in more
59.07 detail for ideal spherical or cylindrical geometries, the diffuse-
59.08
layer charge–potential relationship will usually simplify to the
59.09
limiting relationship for the differential capacitance of the dou-
218 J. Lützenkirchen / Journal of Colloid and Interface Science 303 (2006) 214–223

(a)

(b)

Fig. 1. Parsons–Zobel plot for the goethite system (generated data based on parameters by Rietra et al. [14] for a basic Stern model; original parameters: 0.91 F/m2 ,
100 m2 /g). (a) The Parson–Zobel plot based on Nernst behavior. (b) The actual potential–pH relationship as obtained from the data generation. The drawn lines and
triangles show values that would be directly calculated using the original parameters.

Table 2
Results of the application of Parsons–Zobel plots to the different parameterizations of surface complexation models
Comments on model parameters Nernst slope Actual slope
APZ in m2 /g C in F/m2 APZ in m2 /g C in F/m2
Original surface complexation parameters 104.3 1.12 104.4 1.25
Electrolyte binding constants decreased by two orders of magnitude 100.2 0.83 100.2 0.91
Electrolyte binding constants decreased by four orders of magnitude 100.1 0.83 100.1 0.91
Note. Original surface complexation parameters: 0.91 F/m2 , 100 m2 /g; electrolyte binding constant log K = −1; parameters from Rietra et al. [14] within a basic
Stern model. Variants reported here include the same model parameters except for electrolyte binding constants set to log K = −3 and −5, respectively.

ble layer, i.e., Eq. (3). So for theoretical data sets generated by
model parameters, no real changes are expected. For actual par-
ticles, however, discrepancies between measured surface areas
and electrochemical surface areas may be explained by the non-
ideal geometries of the real particles.
Another problem may occur in comparing results based on
Eq. (5) with a surface complexation model including Eq. (4).
While the former uses a differential capacitance, the latter is an
integral capacitance. Derivation of σ0 with respect to pH (based
on Eq. (4)) yields the following expression (the suffix ppzc is
 J. Lützenkirchen / Journal of Colloid and Interface Science 303 (2006) 214–223 219

omitted for simplicity in the following) for the surface specific Based on Eq. (18), Eq. (10) only holds if the charge density
capacitance: in plane 1 is zero (or sufficiently close to zero). This condition
will be approximately fulfilled for weak ion binding or at suf-
C1 = (∂σ0 /∂Ψ0 )/(1 − ∂Ψd /∂Ψ0 ). (6) ficiently low ionic strength. In other words, the Parsons–Zobel
Insertion into Eq. (1) and using Eq. (5) yield plot is not applicable to delivering BSM surface complexation
parameters in the case of too strong electrolyte binding.
1/Ctot = 1/(∂σ0 /∂Ψ0 ) This result is corroborated by model parameter variation
 
= (1 − ∂Ψd /∂Ψ0 )/(∂σ0 /∂Ψ0 ) + 1 −(∂σd /∂Ψd ) , (7) and subsequent application of Parsons–Zobel plots, with the
goethite model as a reference system. Variation of the (sym-
and rearranging results in metrical) electrolyte binding constants from log K = −1 to
−(∂σd /∂Ψd ) = −(1 − ∂Ψd /∂Ψ0 ) × (∂σd /∂Ψd ) −5 for otherwise constant parameters showed that the consis-
tency between input parameters and Parsons–Zobel plot para-
+ (∂σ0 /∂Ψ0 ) (8)
meters indeed improves. Excellent agreement is obtained for
and finally log K  −3 (Table 2). Using data-derived (actual) surface po-
tential versus pH slopes instead of Nernstian slopes does not
(∂σd /∂Ψ0 ) = −(∂σ0 /∂Ψ0 ) (9) contribute to improving the results at high electrolyte binding
or constants, whereas with lower electrolyte binding constants the
use of the actual potential–pH relationship leads to the correct
(∂σd /∂σ0 ) = −1. (10) capacitance value. Variation of the inner layer capacitance at
In these equations σd is the diffuse-layer charge. Equa- constant electrolyte binding corroborates this conclusion: al-
tion (10) must hold for the original Parsons–Zobel plot to be though a decrease of the capacitance makes the surface more
applicable in the context of the tested surface complexation Nernstian, it is found that this contribution to improving the ac-
model. Inclusion of electrolyte binding (in this kind of model curacy of the Parsons–Zobel plot parameters is not as important
usually in plane 1) yields the following equation based on over- as the contribution due to counterion binding (Table 2).
all particle charge neutrality: Electrolyte binding appears to be rather strong, as indicated
by shifts of IEPs at higher ionic strength or unsymmetrical
σd = −(σ0 + σ1 ), (11) charging curves. Shifts of IEPs at higher ionic strength appear
where σ1 is the charge density in the plane of electrolyte ad- to be a very common feature [21,22].
sorption. Corresponding to Eq. (10), differentiation of Eq. (11) The results may also be used to explain certain observa-
with respect to σ0 yields tions made by Gunnarsson [4]. In that work it was found that
Parsons–Zobel plots would yield parameters that could be used
(∂σd /∂σ0 ) = −1 − (∂σ1 /∂σ0 ). (12) in a straightforward manner to describe charging of goethite in
NaClO4 solutions (note in this context that in the BSM model
Combining Eqs. (10) and (12) yields
by Gunnarsson [4] no electrolyte binding was involved at all!).
(∂σ1 /∂σ0 ) = 0. (13) The explanation of the discrepancy between the outcome of
Parsons–Zobel plots with goethite data used in this work and
The derivative ∂σ1 /∂σ0 within the basic Stern model becomes those used by Gunnarsson [4] is that Gunnarsson’s titrations
more transparent by using the surface species in the planes of in NaClO4 solutions produced surface charge densities that are
adsorption; i.e., much lower than those corresponding to Rietra’s model in NaCl
   
σ1 = −B × SOH2 · · ·A– – SOH−1/2 · · ·C+ , or NaNO3 . In other words, the electrolyte binding in NaClO4
1/2
(14)
 1/2 
solutions is of minor importance. From Rietra’s titrations in var-
σ0 = B × SOH2 ious electrolytes it is known that the surface charge density is
     
+ SOH2 · · ·A– – SOH−1/2 – SOH−1/2 · · ·C+ lower in NaClO4 solutions than in NaCl or NaNO3 solutions.
1/2
 1/2    This was also discussed by Gunnarsson [4]. The results of the
= B × SOH2 – OH−1/2 − σ1 . (15) present work may also be used as an explanation for the failure
Combination of the two yields of the Parsons–Zobel plots to yield a description of the hematite
 data in NaNO3 used by Gunnarsson [4]. Gunnarsson indicated
1/2   
σ1 = B × SOH2 – SOH−1/2 − σ0 , (16) that involvement of electrolyte binding would improve the fit
with a simultaneous decrease of the capacitance value. This
where B is a factor relating molar concentrations of surface
corresponds directly to the parameter variation results obtained
species to surface charge densities. Equation (13) requires that
with goethite in this work. So the interpretation of the failure
  1/2    
(∂σ1 /∂σ0 ) = B × ∂ SOH2 – OH−1/2 ∂σ0 − 1 based on the results of this paper would be that electrolyte bind-
ing is too strong in the hematite system studied by Gunnarsson
=0 (17)
[4] for a Parsons–Zobel plot to be applicable, whereas Gunnars-
and finally son explained the failure by coagulation effects of the hematite
  1/2     during titration. Coagulation may be related to the strong ad-
B × ∂ SOH2 – SOH−1/2 ∂σ0 = 1. (18) sorption of the electrolyte ions.
220 J. Lützenkirchen / Journal of Colloid and Interface Science 303 (2006) 214–223
When Parsons–Zobel plots are to be used in parameteriza-
tion procedures for surface complexation models, the present
results suggest that an optimum parameterization might be a
sufficiently high capacitance value (to grant the observed pro-
ton uptake) combined with sufficiently low electrolyte binding.
Parsons and Zobel [1] implied that their methodology as ap-
plied to the mercury electrode holds in the presence of non-
specifically adsorbing ions. The same holds for mineral sus-
pensions. This has not received any attention in previous ap-
plications of Parsons–Zobel plots to mineral suspensions. The
term specifically or nonspecifically bound may unfortunately
lead to confusion. Traditionally, the terminology with mineral
suspensions has involved an operational definition, according
to which specifically binding electrolyte ions are able to shift
points of zero charge. From electroacoustic measurements on
various oxide minerals at high ionic strength with electrolytes,
which are “inert” at lower ionic strength, it is known that iso-
electric points shift [21,22]. Thus in terms of this definition it
might be sufficient to go up in ionic strength to show that the
monovalent ions are never “inert” over the whole concentration
range.
Parsons and Zobel [1] indicate for their electrode systems
that a linear Parsons–Zobel plot is an indication of nonspe-
cific binding. This clearly does not hold for mineral suspensions
based on the presented calculations, since linear plots were al-
ways obtained in the present examples. However, parameter
variation showed that with sufficiently strong electrolyte bind-
ing constants, the original parameters could not be retrieved.
4.2. Applications to experimental data
As has been seen in the treatment of generated data,
Parsons–Zobel plots yield good estimates for specific surface
areas, even if the slopes of surface potential–pH curves vary
with ionic strength. Thus, it was attempted to use the approach
with hydrous ferric oxide (HFO). Literature data [16–20] were
used to this end. The results are summarized in Table 3, in-
cluding the extracted capacitance and electrochemical surface
area values as well as the number of points used for the regres-
sion and the linear regression coefficient. The linear regressions
were usually very good. Interestingly for the data by Swallow
[18] no extreme value in the derivative was obtained for the
highest ionic strength, a result that would be expected since the
charging curves typically assume a linear pH-dependency at
sufficiently large electrolyte concentrations. The same was ob-
tained for a number of other data sets, which included data at
Table 3
A summary of results of the application of Parsons–Zobel plots to various HFO samples
Sample Electrolyte S in
from in titration m2 /g
[16] NaNO3 2000
[17] KNO3 2690
[18] NaClO4 1800
[19] NaClO4 525
[20] NaNO3 2650
a The derivative at the highest ionic strength did not yield an extremum for this data set. The linear regression coefficient is therefore meaningless and the results
are consequently to be considered with care.
ionic strength >0.1 M. These were therefore omitted from the
discussion.
In general, the Parsons–Zobel plots yielded very high spe-
cific surface areas of 2000 m2 /g and more. A notable exception
is obtained with the data by Hansen et al. [19] with a value
of around 500 m2 /g. This may be explained by the fact that
Hansen et al. aged their sample 9 days, whereas the other sam-
ples may rather represent fresh ferrihydrite. The particle size
of those latter samples must be extremely small or the den-
sity must be substantially affected by inclusion of water, if
the results from the Parsons–Zobel plots are to be meaningful.
The “electrochemical” surface area may also significantly differ
from BET or EGME areas, since it refers to the Gouy–Chapman
equation. In general the particles will be far from having a per-
fect geometry, but they will instead exhibit surface roughness.
To what extent this interferes with the Parsons–Zobel plot as-
sumptions cannot be discussed based on the simple models used
here, but it would be an interesting issue for theoretical consid-
erations. One might expect that the head end of the diffuse layer
would assume a more regular shape than an underlying rough
surface. It is clear that surface roughness creates a higher area
compared to a mean surface at a smooth model surface “zero”
plane. The separation of the location of the “electrochemical”
surface area from the surface would result in a higher value
for the “electrochemical” surface area, compared to the surface
area for the smooth model surface “zero” plane. Since all val-
ues refer to the same mass of sorbent, it is the interplay between
actual surface roughness and the extent of the separation and
the actual size of the particles, which will determine the ratio
between the surface area of the rough surface and the “elec-
trochemical” surface area. As for the probe molecules used in
BET measurements the dimensionality of the surface roughness
and the size of the probe molecules come into play. As a con-
sequence, it is not possible to give a general tendency, although
one would expect that the roughness effect might dominate. But
this is then a theoretical expectation, since the comparison must
presently be made through measurement with probe molecules
of finite size. The model involved in the interpretation always
assumes ideal geometries, and this may lead to inconsistencies
between measured values and model-input values.
From the application of the Parsons–Zobel plot it is ob-
served that the extracted capacitance values vary substantially.
The area specific capacitance depends on the slope of the mass-
specific surface charge versus pH curve around the pH of the
slope extremum and the extracted surface area. The combina-
C in Range of pH for Number of points Linear regression
F/m2 slope extremum used in regression coefficient
0.31 8.15–8.38 3 1.000
1.03 8.13–8.14 3 0.998
0.50 8.32–8.41 2a NA
0.18 6.71–7.00 3 0.979
0.12 6.54–7.44 3 1.000
 J. Lützenkirchen / Journal of Colloid and Interface Science 303 (2006) 214–223
Fig. 2. Comparison of the mass-specific surface charge of the HFO samples that were analyzed by Parsons–Zobel plots. Data from Schwertmann and Fechtner [17],
Cox and Gosh [16], Swallow [18], Hansen et al. [19], and Trivedi et al. [20].
tion of charging curves with a low slope and a high area results
in low capacitance values. This is not necessarily seen from
Fig. 2, where mass-specific surface charge versus pH curves are
shown for the different samples. The data by Hansen et al. [19]
clearly show the lowest charge, which may be attributed to the
extended aging time of their solid. The aging time for the sam-
ple by Sparks and co-workers [20] was two days, which is also
longer than what is reported for the fresh samples prepared by
Cox and Gosh [16] and Swallow [18]. Finally, it is interest-
ing to note from Fig. 2 that apparently the data for the fresh
preparations [17,18] and those for the two slightly aged sam-
ples [16,20] agree with each other but not with the other set of
data. The exact aging for the sample from Schwertmann and
Fechtner [17] is not known, but it was reported that the sample
was freeze-dried. The salts used as background electrolytes in
the titrations (cf. Table 3) do not seem to play a significant role,
although they usually do for goethite at 0.1 M electrolyte (i.e.,
charging is stronger in nitrate than in perchlorate media [14]).
Sparks and co-workers [20] actually discuss the discrepancy be-
tween the amount of charge on their samples and that on other
HFO samples. They also mention the difference of surface spe-
cific charge density between HFO and goethite. This difference
would dramatically decrease if the electrochemical surface ar-
eas obtained in this study from the Parsons–Zobel plots were
applied, i.e., if HFO surface areas increased by a factor of 3
to 4.
Such a huge difference between electrochemical and BET
surface areas has also been found by Gunnarsson [4], in partic-
ular for magnetite, and this was explained by sticking together
of small particles after drying (see Introduction).
Another interesting issue in the treatment of the HFO data is
that the extreme values of the derivatives varied more strongly
than expected with respect to their position on the pH scale.
Bourikas et al. [15] have suggested that this extreme should
correspond to the point of zero charge. Whereas a point of zero
charge around pH 8 for HFO [10] seems surprisingly well es-
221
tablished, the maxima of the derivatives of the more aged sam-
ples [19,20] are significantly lower than pH 8. Furthermore, for
the data by Sparks and co-workers [20], they vary over nearly
one pH unit as a function of ionic strength. Nevertheless, in
the respective papers the point of zero charge is fixed at pH 8.
This is done based on other assumptions/hypotheses (e.g., com-
mon intersection point). Rigorous determination of points of
zero charge from titrations is difficult, since common intersec-
tion points are also obtained in 2:1 electrolytes and are different
from those obtained in “nonspecific” electrolytes. Thus it is
highly preferable to apply some electrokinetic method to cor-
roborate the point of zero charge, in particular at low ionic
strength, where the resolution of microelectrophoresis increases
and electrolyte specific effects are minimized. The problem is
the low solid concentration in such experiments, which can be
avoided in electroacoustic measurements. An ill-defined point
of zero charge will strongly affect the parameterization of a
surface complexation model (e.g., in the case of Ref. [19] the
adsorption modeling for silicic acid and in the case of Ref. [20]
the adsorption modeling for lead), if the postulated metal ion or
ligand surface species is charged. Net transfer of charge to or
from the inner-surface plane will be affected by the electrosta-
tic term involving the surface potential, which depends on the
smeared-out surface charge, and this effect can easily amount
to orders of magnitude.
The capacitance values obtained are in all cases (except one)
quite low. From the tests on generated data, it is clear that these
values are not expected to be useful. The very low values would
imply that the thickness of the inner Helmholtz layer is rather
large, that the relative permittivity of that layer is rather small
(i.e., strong structuring of water), or both.
5. Conclusions
Parsons–Zobel plots have been tested on model-generated
data based on surface complexation parameters for goethite
222 J. Lützenkirchen / Journal of Colloid and Interface Science 303 (2006) 214–223
(within the BSM) and on experimental data on hydrous fer-
ric oxide (HFO). The application of the Parsons–Zobel plot to
model-generated data led to the conclusion that two issues may
prohibit the use of parameters extracted from this approach in
surface complexation models:
– Systems with significant electrolyte binding should not be
treated by the Parsons–Zobel-plot approach. In the present
case, only if electrolyte binding occurred with log K  −3,
Parsons–Zobel plots were able to reproduce the expected
parameters. For larger electrolyte binding constants, the
ion pairing interferes too strongly with the assumptions in-
herent in the Parsons–Zobel-plot approach. This has been
shown with the generic example based on parameter varia-
tion but became also evident from underlying equations.
– Nernstian slopes for surface potential versus pH curves of
relevant oxide minerals have been postulated but there is
no ultimate proof for their governing the oxide/water sys-
tems. As an example in the model-generated data, the ac-
tual slopes were not Nernstian and they also varied with
ionic strength. Decreasing strength of electrolyte binding
tended to create slopes closer to the Nernstian one. Note
that the calculations to generate the data involve model as-
sumptions, which await proof. However, the parameter set
by Rietra et al. [14] describes experimental surface charge
very well for a wide range of conditions.
The deviation from Nernstian behavior was found to be the
less important issue.
When Parsons–Zobel plots are used in the context of para-
meter estimation, it should subsequently be attempted to im-
pose sufficiently low electrolyte binding and Nernstian slopes.
This is the only way to assure consistency between the sur-
face complexation model and the assumptions inherent to the
Parsons–Zobel plots. The present study explains observations
by Gunnarsson [4], which indicate that Parsons–Zobel plots do
not generally work as independent approaches to determine pa-
rameters (capacitance values) for the basic Stern model (even
without inclusion of electrolyte binding). Conceptionally, in
surface complexation models, such as the basic Stern and other
multilayer models, electrolyte ions actually bind to surface sites
as so-called outer-sphere surface complexes as point charges.
Sometimes, and particularly in the case of temperature stud-
ies, rather strong electrolyte binding constants are required [23]
to fit experimental data. Consequently, in its original version
the Parsons–Zobel plot will not be a realistic option to estimate
capacitance values for the basic Stern model in general, par-
ticularly since it is known that electrolyte binding is related to
the nature of the electrolyte and that of the surface [24–27].
Structure-breaking and -making properties of the electrolyte
ions and the surface with respect to the water molecules will af-
fect both the interlayer capacitance (via distance and dielectric
constant) and potentially even the electrochemical surface area.
Attempts have recently been made to incorporate water struc-
ture into surface complexation models [28]. These attempts
currently do not include the more elaborate, but qualitative ex-
planations by Dumont et al. [24–27].
The estimates for specific surface areas obtained from the
application of Parsons–Zobel plots to experimental data for
HFO has resulted in unexpectedly high values. This agrees with
findings by Gunnarsson [4]. With the generated data, the es-
timated specific surface areas were also higher (only slightly,
though) than the input model parameter values, but those re-
sults suggested that quite reasonable agreement with expected
values can be obtained. The major obstacle here was found to
be a variation of the surface potential vs pH curves with ionic
strength.
Based on the reported results, at present it seems that the use
of the conventional Parsons–Zobel plots is not a general means
of independently estimating parameters. Although the present
work is restricted to iron(III) minerals only, it is concluded from
the results that the approach can be used, but entails certain
ranges for remaining unknown parameters (which need to be
fitted to experimental data) and in a strict sense also for the
surface potential–pH curves. The approach yielded reasonable
values for the electrochemical surface area with the generated
data, but unexpectedly high values for most HFO samples com-
pared to recommended literature values [10]. Although this
might be explained by assuming very small particles in com-
bination with a low particle density and although it would place
the surface charge density of HFO in a range similar to that of
goethite, the values for electrochemical surface areas obtained
by Parsons–Zobel plots have to be considered with caution,
one obstacle being that the slope of surface potential versus
pH curves is usually not accessible. It might be expected that
the recent approach by Kallay et al. [29] might resolve this is-
sue. Surface potential–pH curves are measurable (single-crystal
electrodes) by that approach, so that from σ0 (pH) experimental
functions (potentiometric acid–base titrations of concentrated
suspensions) one may experimentally determine dσ0 /dΨ0 .
Partial incompatibility of the Parsons–Zobel plot capaci-
tances with surface complexation models, which was previ-
ously noted [4], could be explained in the present paper. It is
due to the fact that the assumption of nonspecific adsorption of
the electrolyte ions is a question of definition. In colloid chem-
istry the term is often connected to observations of shifting iso-
electric points. This notion is not necessarily compatible with
the application of Parsons–Zobel plots. The electrolyte binding
constants commonly encountered in basic Stern or triple-layer
models are too high compared to the notion within the Parsons–
Zobel plot.
Acknowledgments
The author is grateful for comments and suggestions from
Professor Kallay on a previous version of this manuscript prior
to submission. The author also gratefully acknowledges the
comments by an anonymous reviewer, which were very help-
ful in clarifying certain parts of the work.
References
[1] R. Parsons, F.G.R. Zobel, J. Electroanal. Chem. 9 (1965) 333.
[2] A. Siviglia, A. Daghetti, S. Trasatti, Colloids Surf. 7 (1983) 15.
 J. Lützenkirchen / Journal of Colloid and Interface Science 303 (2006) 214–223
[3] S. Ardizzone, S. Trasatti, J. Electroanal. Chem. 251 (1988) 409.
[4] M. Gunnarsson, Ph.D. thesis, Göteborg University, Göteborg, Sweden,
2002.
[5] J. Rosenqvist, P. Persson, S. Sjöberg, Langmuir 18 (2002) 4598.
[6] D.L. Carter, M.D. Heilmann, C.L. Gonzalez, Soil Sci. 100 (1965) 356.
[7] J. Lützenkirchen, J. Colloid Interface Sci. 210 (1999) 384.
[8] J. Lützenkirchen, J.F. Boily, L. Lövgren, S. Sjöberg, Geochim. Cosmo-
chim. Acta 66 (2002) 3389.
[9] D.A. Sverjensky, Geochim. Cosmochim. Acta 69 (2005) 225;
D.A. Sverjensky, Geochim. Cosmochim. Acta 65 (2001) 3643.
[10] D.A. Dzombak, F.M.M. Morel, Surface Complexation Modelling Hydrous
Ferric Oxide, Wiley, New York, 1990.
[11] I. Larson, P. Attard, J. Colloid Interface Sci. 227 (2000) 152.
[12] L. Bousse, N.F. De Rooij, P. Bergveld, IEEE Trans. Electron Devices 30
(1983) 1263.
[13] A. Keizer, W. van Riemsdijk, ECOSAT, vers. 4.7, Wageningen University,
2001.
[14] R.P.J.J. Rietra, T. Hiermstra, W.H. van Riemsdijk, J. Colloid Interface
Sci. 240 (2001) 384.
[15] K. Bourikas, C. Kordulis, A. Lycourghiotis, Environ. Sci. Technol. 39
(2005) 4100.
223
[16] C.D. Cox, M.M. Gosh, Water Res. 28 (1994) 1181.
[17] U. Schwertmann, H. Fechtner, Clay Miner. 17 (1982) 471.
[18] K.C. Swallow, Ph.D. thesis, Massachusetts Institute of Technology, Cam-
bridge, MA, 1978.
[19] H.C.B. Hansen, T.P. Wetche, K. Raulund Rassmussen, O.K. Borggaard,
Clay Miner. 29 (1994) 341.
[20] P. Trivedi, J.A. Dye, D.L. Sparks, Environ. Sci. Technol. 37 (2003) 908.
[21] M. Kosmulski, J.B. Rosenholm, Adv. Colloid Interface Sci. 112 (2004)
93.
[22] W.N. Rowlands, R.W. O’Brien, J. Colloid Interface Sci. 175 (1995)
190.
[23] M.L. Machesky, D.J. Wesolowski, D.A. Palmer, K. Ichiro-Hayashi,
J. Colloid Interface Sci. 200 (1998) 298.
[24] F. Dumont, A. Watillon, Discuss. Faraday Soc. 52 (1971) 324.
[25] F. Dumont, D. Van Tan, A. Watillon, J. Colloid Interface Sci. 55 (1976)
678.
[26] F. Dumont, J. Warlus, A. Watillon, J. Colloid Interface Sci. 138 (1990)
543.
[27] F. Dumont, S. Contreras, M. Diaz y Alonso, Anal. Quim. 91 (1995) 635.
[28] T. Hiemstra, W.H. van Riemsdijk, J. Colloid Interface Sci. 301 (2006) 1.
[29] N. Kallay, Z. Dojnovic, A. Cop, J. Colloid Interface Sci. 286 (2005) 610.