Journal of Colloid and Interface Science 280 (2004) 322333

www.elsevier.com/locate/jcis
Adsorption isotherm models for basic dye adsorption by peat in single
and binary component systems
S.J. Allen
a,
, G. Mckay
b
, J.F. Porter
b
a
School of Chemical Engineering, Queens University Belfast, Belfast, UK
b
Department of Chemical Engineering, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong, China
Received 1 April 2004; accepted 3 August 2004
Available online 7 October 2004
Abstract
Colored efuents from textile industries are a problem in many rivers and waterways. Prediction of dye adsorption capacities is important
in design considerations. The sorption of three basic dyes, namely Basic blue 3, Basic yellow 21, and Basic red 22, onto peat is reported.
Equilibrium sorption isotherms have been measured for the three single-component systems. Equilibrium was achieved after 21 days. The
experimental isotherm data were analyzed using Langmuir, Freundlich, RedlichPeterson, Tempkin, and Toth isotherm equations. A detailed
error analysis has been undertaken to investigate the effect of using different error criteria for the determination of the single-component
isotherm parameters and hence obtain the best isotherm and isotherm parameters which describe the adsorption process. The linear transform
model provided the highest R
2
regression coefcient with the RedlichPeterson model. The RedlichPeterson model also yielded the best
t to experimental data for all three dyes using the nonlinear error functions. An extended Langmuir model has been used to predict the
isotherm data for the binary systems using the single component data. The correlation between theoretical and experimental data had only
limited success due to competitive and interactive effects between the dyes and the dyesurface interactions.
2004 Elsevier Inc. All rights reserved.
Keywords: Adsorption; Isotherm; Langmuir; Freundlich; RedlichPeterson; Toth; Tempkin; Extended Langmuir; Error analysis
1. Introduction
Colored dye wastewater arises as a direct result of the
production of the dye and also as a consequence of its use in
the textile and other industries. There are more than 100,000
commercially available dyes with over 7 10
5
tonnes of
dyes produced annually [1]. It is estimated that 2% of dyes
produced annually are discharged in efuent from manufac-
turing operations, while 10% was discharge from textile and
associated industries [2]. Neglecting the aesthetic problem,
the greatest environmental concern with dyes is their ab-
sorption and reection of sunlight entering the water, which
interferes with the growth of bacteria, limiting it to levels
insufcient to biologically degrade impurities in the water.
*
Corresponding author.
E-mail address: s.allen@qub.ac.uk (S.J. Allen).
Color in efuents can cause problems in several ways: dyes
can have acute and/or chronic effects on exposed organisms
depending on the exposure time and dye concentration; dyes
are inherently highly visible, meaning that concentrations as
low as 0.005 ppm capture the attention of both the public
and the authorities [3]; dyes absorb and reect sunlight en-
tering water and so can interfere with the growth of bacteria
and hinder photosynthesis in aquatic plants [4]. It is evident,
therefore, that removal of such colored agents from aqueous
efuents is of signicant environmental, technical, and com-
mercial importance.
Adsorption is widely used in the removal of contaminants
from wastewaters. The design and efcient operation of ad-
sorption processes require equilibrium adsorption data for
use in kinetic and mass transfer models. These models can
then be used to predict the performance of the adsorption
contact processes under a range of operating conditions.
0021-9797/$ see front matter 2004 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcis.2004.08.078
S.J. Allen et al. / Journal of Colloid and Interface Science 280 (2004) 322333 323
The equilibrium isotherm plays an important role in pre-
dictive modeling for analysis and design of adsorption sys-
tems. The adsorption isotherm is also an invaluable tool for
the theoretical evaluation and interpretation of thermody-
namic parameters, such as heats of adsorption. An isotherm
may t experimental data accurately under one set of con-
ditions but fail entirely under another. No single model has
been found to be generally applicable, a fact which is read-
ily understandable in the light of the assumptions associ-
ated with their respective derivations. The current research
presents a method of direct comparison of the isotherm t of
several models to enable the best t and best isotherm para-
meters to be obtained. An accurate mathematical description
of equilibrium adsorption capacities, even if empirical, is in-
dispensable for reliable predictive modeling of adsorption
systems and quantitative comparisons of adsorption behav-
ior for different adsorbent systems or for varied conditions
within any given system [5].
Adsorption using peat is well established [6,7]. Peat is
known to have excellent ion-exchange properties similar to
those of natural zeolites [8]. The use of peat in water treat-
ment systems has received attention over the past 30 years
and peat lters and biolters currently offer an attractive
method of wastewater treatment. Peat is relatively inexpen-
sive and possesses characteristics which make it effective for
adsorption or ion-exchange operations. Peat is a low-grade
carbonaceous fuel containing lignin, cellulose, and humic
acids as its major constituents. These constituents are po-
tentially very useful from an adsorption standpoint as they
bear polar functional groups such as alcohols, aldehydes,
carboxylic acids, ketones, and phenolic hydroxides, which
can all be involved in chemical bonding with the adsorbed
pollutants. Peat has a polar character which lends itself to
the adsorption of specic materials such as metals and polar
organic molecules.
Leslie [9] studied the treatment of dyehouse efuent us-
ing a HussongCouplan wastewater treatment process based
on peat moss. Murray [10], Khader [11], and Allen [12]
discuss the equilibrium and kinetics of adsorption of color
removal from wastewater by peat. Data obtained by Smith
et al. [13,14] showed that peat treated with sulfuric acid
or a sulfuric/phosphoric acid blend possessed enhanced po-
tential for cation removal over untreated peats. It is sug-
gested [1517] that irreversible coulombic attraction and
ion exchange is the mechanism for basic dye adsorption by
adsorbents containing a high level of cellulose [18]. Other
adsorbents which can be used for basic dye adsorption in-
clude kudzu [19,20], peanut hulls [19], MDF waste [19],
carbon [21], bagasse pith [21], fullers earth [21], and sil-
ica [21]. A comparison of uptakes is presented in Table 1.
The present paper investigates the effect of the isotherm
type and the method used to derive its parameters on the
predictions of the parameters within the models. Sets of
isotherm parameters are determined using both linear re-
gressions of transformed isotherms as well as minimizing
a series of error functions.
Table 1
Comparison of basic dye uptake by different adsorbents
Adsorbent Dyes Approximate
uptake (g kg
1
)
Ref.
Activated carbon BB3, BR22 448, 500 [21]
Bagasse pith BB3, BR22, 62, 67 [21]
Fullers earth BB3, BR22, 500, 460 [21]
Kudzu BB3, BR22, BY21 650, 220, 160 [19,20]
Peat BB3, BR22, BY21 550, 310, 660 [10,11]
Lignite BB3, BR22, BY21 730, 500, 675 [10,11]
Silica BB3, BR22 375, 24 [21]
Peanut hulls BB3, BR22, BY21 400, 130, 150 [19]
MDF BB3, BR22, BY21 115, 120, 110 [19]
Table 2
Physical characteristics of peat
Moisture content 98.1%
Ash weight (air-dried sample) 1.9% w/w
Surface area 32.5 m
2
g
1
Pore volume 0.085 cm
3
g
1
Elemental analysis
Carbon 49.0%
Oxygen 42.8%
Hydrogen 6.2%
Nitrogen 1.9%
Sulfur 0.1%
Metal analysis
(main metals detected) Ni
222
Fe
238
Mg
280
Ca
393
Al
396
Na
589
2. Experimental
The adsorbent used in the research was a medium-texture,
horticultural-grade sphagnum moss peat. The adsorbent was
milled using a hammer mill, air-dried, and then sieved into
a range of particle size ranges using Endecotts test sieves.
Some physical characteristics of the adsorbent are outlined
in Table 2.
Dyes are chemicals which on binding with a material
will give color to the material. Dyes are ionic, aromatic or-
ganic compounds with structures including aryl rings which
have delocalized electron systems. Basic dyes fall within the
class of polymethine dyes [22]. As such they are cationic
polymethinesbecause ionization in solution they possess
an overall positive charge and are therefore widely used
to bind to negatively charged substrates [23,24]. Aqueous
solutions of basic dyes were chosen as the principal adsor-
bates for this work. Three basic dyes, Basic red 22, Basic
blue 3, and Basic yellow 21, in single-component systems
were used during the batch adsorption studies.
The structure of each of the dyes, along with its com-
mercial name and manufacturer, as listed in the Colour in-
dex [25], is given in Fig. 1. The molar volumes of the dyes
are 337.6, 438.6, and 419.3 cm
3
/mol for the red, blue, and
yellow dyes, respectively [26]. The molecular diameters of
324 S.J. Allen et al. / Journal of Colloid and Interface Science 280 (2004) 322333
Fig. 1. Structure of the basic dyes.
the dyes are 1.13, 1.15, and 1.24 nm for the red, blue, and
yellow dyes, respectively [26].
Concentrations in aqueous solution were determined
spectrophotometrically in the visible range of the spectrum.
A PerkinElmer, Model 550S double-beam spectropho-
tometer with a 1-cm path length was used throughout this
work. Disposable plastic cuvettes (4 cm
3
volume) were used
to contain the dye samples and for all aqueous solutions of
dyes distilled, deionized water was used as the reference
sample. Initially, the wavelength at which the maximum
absorbance of each dye in aqueous solution occurred was
determined by scanning the light absorption curve. The max-
imum wavelengths
max
(nm) for the basic dyes studied are
417, 538, and 641 nm for yellow, red, and blue dyes, respec-
tively. Before the adsorbate concentration was analyzed in
solution, it was necessary to obtain a calibration curve for
each of the dyes in aqueous solution. The calibration range
was from zero to 1000 mg dm
3
. A 1000-mgdm
3
stock
solution was prepared and consecutively diluted to give a
series of dye solutions of known concentration. Using the
spectrophotometer, which was set at the
max
for the dye,
each solution was analyzed for optical density. Optical den-
sities were plotted against corresponding dye concentrations
and, once the optical density of the aqueous dye solution
was determined, the concentration of dye in solution could
be calculated.
Adsorption isotherms for this research were carried at
18 2

C using a bottle point procedure. A constant mass

of peat (0.05 g), of particle size 500710 m, was weighed
into each of twelve 60-ml dark glass powder jars. The dark
glass prevented UV light degradation of the color. A 1000-
mg/dm
3
stock solution of each dye was prepared and di-
luted to obtain twelve concentrations within the range 50
1000 mg/dm
3
. This range represents concentrations which
could arise in dilute efuents as well as in concentrated
dump solutions. It is also the range to give the isotherm
plateau. Aliquots of dye solution (50 ml) of specied con-
centration were then added to each of the preweighed sor-
bent samples. The glass jars were sealed using paralm and
screw-top lids and kept in a state of constant agitation using
an L.S.E. Scientic multishaker. Control experiments also
conrmed that there was no color degradation by biologi-
cal processes. Upon equilibration, the bottles were removed
from the shaker. Samples of 45 ml were then withdrawn
by syringe, ltered through 50-m micropore lter paper
supported in 25-mm syringe compatible lter holders, and
analyzed to determine the residual equilibrium liquid-phase
dye concentration.
Preliminary experiments over the range of twelve dye
concentrations indicated that adsorption equilibrium was
reached in 3 weeks. No further appreciable decrease in the
bulk dye concentration was observed for a time period of up
to 7 weeks. Based on these results, the duration of the ad-
sorption equilibrium isotherm experiments was chosen to be
3 weeks.
3. Results and discussion
3.1. Single-component isotherms
Successful representation of the dynamic adsorptive sep-
aration of solute from solution onto an adsorbent is depen-
dent upon a good description of the equilibrium separation
between the two phases. An equilibrium is established when
the amount of solute being adsorbed onto the adsorbent is
equal to the amount being desorbed. The equilibrium so-
lution concentration remains constant. Plotting solid-phase
concentration against liquid-phase concentration graphically
depicts the equilibriumadsorption isotherm. Several theories
of adsorption equilibrium, used to describe these isotherms,
are available in the literature [2732].
3.1.1. The Langmuir isotherm
The Langmuir isotherm theory assumes monolayer cov-
erage of adsorbate over a homogenous adsorbent sur-
face [27]. Graphically, a plateau characterizes the Langmuir
isotherm. Therefore, at equilibrium, a saturation point is
reached where no further adsorption can occur. A basic
assumption is that sorption takes place at specic homo-
geneous sites within the adsorbent. Once a dye molecule
occupies a site, no further adsorption can take place at that
site. In the following equation, K
L
and a
L
are the Langmuir
isotherm constants:
(1) q
e
=
K
L
C
e
1 +a
L
C
e
.
S.J. Allen et al. / Journal of Colloid and Interface Science 280 (2004) 322333 325
The Langmuir constants, K
L
and a
L
, are evaluated through
linearization of Eq. (1):
(2)
C
e
q
e
=
1
K
L
+
a
L
K
L
C
e
.
Hence by plotting C
e
/q
e
against C
e
it is possible to ob-
tain the value of K
L
fromthe intercept, which is (1/K
L
), and
the value of a
L
from the slope, which is (a
L
/K
L
). The the-
oretical monolayer capacity is Q
0
and is numerically equal
to (K
L
/a
L
). The Langmuir equation is applicable to homo-
geneous sorption, where the sorption of each molecule has
equal sorption activation energy. The equation is thermody-
namically consistent and follows Henrys Law at low con-
centrations. As C
e
becomes lower, a
L
C
e
is much less than
unity and q
e
=K
L
C
e
, which is analogous to Henrys Law.
3.1.2. The Freundlich isotherm
The Freundlich expression is an exponential equation and
therefore assumes that as the adsorbate concentration in-
creases so too does the concentration of adsorbate on the
adsorbent surface:
(3) q
e
=K
F
C
b
F
e
.
In this equation K
F
and b
F
are the Freundlich constants.
This expression is characterized by the heterogeneity factor,
b
f
, and so the Freundlich isotherm may be used to describe
heterogeneous systems [3335]. Theoretically, using this ex-
pression, an innite amount of adsorption can occur [28].
The Freundlich equation agrees well with the Langmuir
equation over moderate concentration ranges but, unlike
the Langmuir expression, it does not reduce to the linear
isotherm (Henrys law) at low surface coverage. Both these
theories suffer from the disadvantage that equilibrium data
over a wide concentration range cannot be tted with a sin-
gle set of constants [35].
To determine the constants K
F
and b
F
, the linear form
of the equation may be used to produce a graph of ln(q
e
)
against ln(C
e
):
(4) lnq
e
=ln K
F
+b
F
lnC
e
.
3.1.3. The RedlichPeterson isotherm
Redlich and Peterson [36] have proposed an empirical
equation, designated the three parameter equation, which
may be used to represent adsorption equilibria over a wide
concentration range [37],
(5) q
e
=
K
R
C
e
1 +a
R
C

e
,
where 1.
This equation reduces to a linear isotherm at low surface
coverage, to the Freundlich isotherm at high adsorbate con-
centration, and to the Langmuir isotherm when =1. How-
ever, the equation cannot be linearized for easy estimation
of isotherm parameters. The linearization of the expression
gives
(6) ln
_
K
R
C
e
q
e
1
_
=lna
R
+ lnC
e
.
Plotting log[(K
R
C
e
/q
e
) 1] against lnC
e
yields a straight
line of slope = and intercept = a
R
. However, plotting
Eq. (6) is not applicable because of the three unknown pa-
rameters contained within the equation. Therefore a mini-
mization procedure is adopted to maximize the correlation
coefcient R
2
between the theoretical data for q
e
predicted
from Eq. (6) and the experimental data.
Often, the method used to determine isotherm parameters
is to use linear regression with transformed variables. The
quality of the t of the experimental data to the isotherm
equation is assessed by the magnitude of the correlation
coefcient for the regression. In other words, the isotherm
giving an R
2
value closest to unity provides the best t.
3.1.4. The Tempkin isotherm
Tempkin and Pyzhev [38] considered the effects of
some indirect adsorbate/adsorbate interactions on adsorption
isotherms and suggested that because of these interactions
the heat of adsorption of all the molecules in the layer would
decrease linearly with coverage. The Tempkin isotherm has
been used in the following form:
(7) q
e
=
RT
b
(lnAC
e
).
Equation (7) can be expressed in its linear form as
(8) q
e
=
RT
b
lnA+
RT
b
lnC
e
,
where
(9) B =
RT
b
.
The adsorption data can be analyzed according to Eq. (8).
A plot of q
e
versus lnC
e
enables the determination of the
constants A and B. The constant B is related to the heat of
adsorption.
3.1.5. The Toth isotherm
The Toth isotherm [39], is derived from potential theory
and is applicable to heterogeneous adsorption. It assumes a
quasi-Gaussian energy distribution. Most sites have an ad-
sorption energy lower than the peak or maximum adsorption
energy:
(10) q
e
=
K
T
C
e
[a
T
+C
t
e
]
1/t
.
3.1.6. Single-component isotherm parameters
In relation to adsorption systems involving the extrac-
tion of large organic compounds and dyes from aqueous
solution the Langmuir and Freundlich isotherm expressions
are most commonly used [4043]. However, in adsorption
systems, the relative size of the sorbate molecule and the
326 S.J. Allen et al. / Journal of Colloid and Interface Science 280 (2004) 322333
molecular sieving effect of the sorbent have been shown to
affect the equilibrium of the system [4446]. Summers and
Roberts [47] found that larger dye molecules had the effect
of homogenizing the sorbent surface, such that access to
the microporous structure was restricted and adsorption was
limited.
The linearization of isotherm equations using such data
transformations will implicitly alter the error structure and
may also violate the error variance and normality assump-
tions of standard least squares [48,49]. This may explain
observations that linear Freundlich parameters produce
isotherms which t the experimental data better at low con-
centrations whereas those for the Langmuir expression tend
to t better at higher concentrations [50].
Nonlinear optimization techniques have been applied to
determine isotherm parameters [5154]. An error function is
dened to enable the optimization process to determine and
evaluate the t of the isotherm equation to the experimental
data. In the context of the current work, ve different error
functions were examined and in each case the isothermpara-
meters were determined by minimizing the respective error
function across the liquid-phase concentration range using
the solver add-in with Microsofts spreadsheet Excel (Mi-
crosoft Corporation, 1994).
(1) The sum of the squares of the errors (ERRSQ): This is a
widely used error function which has, however, one ma-
jor drawback. The function will result in the calculated
isotherm parameters providing a better t at the higher
end of the liquid-phase concentration range. This is be-
cause the magnitude of the errors and hence the square
of the errors will increase as concentration increases:
(11)
p

i=1
(q
e,calc
q
e,meas
)
2
i
.
(2) The hybrid fractional error function (HYBRID): This er-
ror function was developed by Porter et al. [55] in order
to improve the t of the ERRSQ method at low con-
centration values. In this method each ERRSQ value is
divided by the experimental solid-phase concentration
value q. In addition a divisor is included as a term for
the number of degrees of freedom for the systemthe
number of data points minus the number of parameters
within the isotherm equation:
(12)
100
p n
p

i=1
_
(q
e,meas
q
e,calc
)
2
i
q
e,meas
_
i
.
(3) Marquardts percent standard deviation (MPSD) [56]:
This error function is similar to a geometric mean error
distribution which is modied to allow for the number
of degrees of freedom of the system [51,52]:
(13) 100
_

_
1
p n
p

i=1
_
(q
e,meas
q
e,calc
)
2
i
q
e,meas
_
i
_
.
(4) The average relative error (ARE) [57]: Here, the er-
ror function minimizes the fractional error distribution
across the entire concentration range:
(14)
100
p
p

i=1
_
(q
e,calc
q
e,meas
)
q
e,meas
_
i
.
(5) The sum of the absolute errors (EABS): In this error
function the approach is similar to the ERRSQ method.
Isotherm parameters determined using this method pro-
vide a better t as the magnitude of the errors increases
biasing the t toward the high-concentration data.
In all of the error methods it is assumed that both the
liquid-phase concentration and the solid-phase concentra-
tion contribute equally to weighting the error criterion for the
model solution procedure. Experimental investigations were
undertaken at m/V ratios of 1 g dm
3
. Hence the differ-
ence in the solid-phase concentrations reects the difference
in the predicted concentrations for both phases. The Lang-
muir isotherm parameters for single-component adsorption
as outlined earlier are presented in Table 3. Examples of the
Langmuir, Freundlich, and RedlichPeterson isotherms us-
ing the linear transform model are illustrated in Figs. 24.
Figs. 5 and 6 show the Tempkin and Toth isotherms. An in-
spection of the single-component equilibrium data indicates
a rapid and extensive uptake with liquid concentration by
the peat followed by a tailing off in the isotherm with an
apparent monolayer being achieved. It is clear that the peat
has a considerable afnity for the three basic dyes with ap-
proximate monolayer saturation capacities of 667, 556, and
313 mg g
1
for Basic yellow 21, Basic blue 3, and Basic red
22, respectively. This suggests that the Langmuir isotherm
should provide a reasonable description and analysis of the
data. The approach of determination of the isotherm para-
meters for the three dyes by linear regression appears to give
acceptable ts to the experimental data with respective re-
gression coefcients (R
2
values) close to unity as shown in
the linear transform (LTFM) columns in Tables 37. Over-
all, the RedlichPeterson isothermhas the highest R
2
values,
whereas the Freundlich and Tempkin values are consider-
ably lower. Purely from a comparison of the R
2
values, the
linearized RedlichPeterson and linearized Toth isotherms
would also be expected to provide a better t to the ex-
perimental data than the linearized Freundlich and Tempkin
isotherms.
Due to the inherent bias resulting from linearization, al-
ternative single-component parameter sets were determined
by nonlinear regression using the ve different error func-
tions as detailed earlier. The process of minimizing the re-
spective error functions across the experimental concentra-
tion ranges examined yields the isotherm parameters. The
isotherm parameters so obtained, together with the values of
the error measures for each isotherm, are presented in full
in Table 3 and the nal sum of the normalized error val-
ues (SNE) in Tables 47. The results tend to suggest that
S.J. Allen et al. / Journal of Colloid and Interface Science 280 (2004) 322333 327
Table 3
Single-component Langmuir isotherm parameters with error analysis, method/error function (parameter set)
Linear
transform
(LTFM)
Sum of the
squares of the
errors
(ERRSQ)
Hybrid
fractional
error
(HYBRID)
Marquardts
percentage
standard
deviation
(MPSD)
Average
relative
error
(ARE)
Sum of the
absolute
errors
(EABS)
Basic yellow 21
K
L
27.17 28.50785 29.73821 26.65989 27.45241 27.45248
a
L
0.040761 0.042601 0.045346 0.039995 0.040748 0.040749
R
2
(linear) 0.9871
ERRSQ 21854.72 21442.68 21781.02 22161.85 21535.5 21535.5
HYBRID 925.4524 907.4192 898.850 939.0334 920.9561 920.9513
MPSD 23.93572 24.25146 24.67206 23.9107 24.00456 24.00453
ARE 14.1986 14.37755 15.14896 14.39328 13.9960 13.996
EABS 393.4777 388.2409 410.8671 402.674 379.4826 379.481
SNE 4.836776 4.810845 4.940024 4.899315 4.772941 4.772928
Basic blue 3
K
L
54.95 67.03182 71.52186 74.72145 74.77079 69.56171
a
L
0.098901 0.129199 0.139445 0.14738 0.149515 0.138775
R
2
(linear) 0.9927
ERRSQ 25756.71 19969.22 20196.74 20745.57 21453.44 21554.62
HYBRID 1867.351 1664.504 1656.297 1661.176 1678.017 1672.196
MPSD 46.14848 45.21064 45.11305 45.0928 45.18128 45.1088
ARE 24.46388 21.17778 20.71077 20.51555 20.3195 20.35342
EABS 466.1996 359.7763 353.1369 354.1544 346.1035 346.103
SNE 54.283748 4.252741 4.270425 4.283562 4.284188
Basic red 22
K
L
12.95337 34.11096 50.12331 61.0745 59.87032 31.68383
a
L
0.041451 0.120618 0.185034 0.235598 0.221924 0.110025
R
2
(linear) 0.9798
ERRSQ 16984.91 9900.722 11034.23 13528.38 11887.2 10044.28
HYBRID 1401.072 613.3646 533.578 573.7164 553.7836 654.8342
MPSD 38.49018 23.13377 18.10363 17.0821 17.45065 24.41209
ARE 22.41024 14.6212 13.44936 13.17123 12.5671 14.84291
EABS 347.6343 281.5628 304.5666 340.2724 308.4765 276.983
SNE 5 2.89615 2.795968 3.016404 2.811141 2.958205
Fig. 2. Single-component Langmuir isotherms and experimental data de-
rived by linearization.
a lower absolute error value is obtained for the Redlich
Peterson isotherm and then the Langmuir isotherm and Toth
isotherm. This ties in quite well with the results from the
LTFM method. Once again, the Freundlich isothermexhibits
a tendency tends to yield higher error values, as does the
Fig. 3. Single-component Freundlich isotherms and experimental data de-
rived by linearization.
Tempkin model. In all cases, the red dye exhibits the best t
followed by the blue and yellow. Since each of the error cri-
teria produce different sets of isotherm parameters, an over-
all optimum set can be difcult to identify. To enable some
form of comparison of the error values, the results for each
328 S.J. Allen et al. / Journal of Colloid and Interface Science 280 (2004) 322333
Fig. 4. Single-component RedlichPeterson isotherms and experimental
data derived by linearization.
Fig. 5. Single-component Tempkin isotherms and experimental data derived
by linearization.
Fig. 6. Single-component Toth isotherms and experimental data derived by
linearization.
set were normalized and combined. The values of the er-
rors obtained for each error function for each set of isotherm
constants were divided by the maximum errors for that error
function to obtain asset of normalized errors. A comparison
of the scaled errors can be undertaken and hence an identi-
cation can be made of the isotherm constants which would
provide the closest t to the measured data. The gures in
bold type in Tables 37 represent the minimum error val-
ues and the optimum isotherm constants obtained using the
different error functions. For the Langmuir isotherm in Ta-
ble 3 the lowest value of the sum of the normalized errors
(SNE) is obtained using the EABS function for the yellow
dye and the HYBRID function for the red dye and blue dye.
The values of K
L
and a
L
in these columns are the optimum
isotherm constants for these dyes. For the Freundlich model
the HYBRID method yields the best t for this model in all
three cases. This error function was developed by Porter et
al. [55] in order to improve the t of the ERRSQ method at
low concentration values and appears to be applicable in this
case. For the red dye either the HYBRID or EABS method
appeared to be effective. The EABS method provides the
lowest error values for those isotherms with three parame-
ters or more complex groupings of parameters such as the
Tempkin, RedlichPeterson, and Toth models. In the EABS
error function the approach is similar to the ERRSQ method.
Isotherm parameters determined using this method would
provide a better t as the magnitude of the errors increased
biasing the t toward the high-concentration data. The va-
lidity of any comparison of error functions depends on any
dominant bias in the error functions and the data used to ob-
tain the parameter sets. The error functions selected in this
work provide a reasonably wide selection and the distribu-
tion of the experimental data does not give excessive bias to
either high or low concentrations.
It is suggested [1517] that irreversible coulombic at-
traction and ion exchange is the mechanism for basic dye
adsorption by adsorbents containing a high level of cellu-
lose [18]. Conversely, it is possible that coulombic repulsion
of an acid dye (negatively charged) can occur due to the cre-
ation of a negative charge on the adsorbent surface in the
presence of water [58]. There is a good correlation between
the content of unesteried polyuronic acids in the cell wall
of sphagnum peat and the cation exchange capacity. The ad-
sorption of cations occurs by the formation of complexes
with the carbonyl and nitrile groups in peat. The presence
of humic acids in peat is primarily responsible for its abil-
ity to adsorb metals and other cationic species. Adsorption
increases as the surface area of the organic substance in-
creases. This explains the increase in adsorption capacity af-
ter humic acids were extracted fromthe peat and lignite [59].
Increasing metamorphism in natural carbonaceous deposits,
i.e., peat > lignite > coal, can result in compaction, which
produces a decrease in the surface area and a corresponding
decrease in adsorption capacity. Peat is shown to be a par-
ticularly effective adsorbent for basic dyes but has a lower
capacity for acid dyes [6066].
3.2. Multicomponent isotherms
An extended Langmuir equation has been developed for
competitive multicomponent adsorption [67]. This model as-
sumes a homogeneous surface with respect to the energy
of adsorption, no interaction between adsorbed species, and
S.J. Allen et al. / Journal of Colloid and Interface Science 280 (2004) 322333 329
Table 4
Single-component Freundlich isotherm parameters with error analysis, method/error function (parameter set)
Linear
transform
(LTFM)
Sum of the
squares of the
errors
(ERRSQ)
Hybrid
fractional
error
(HYBRID)
Marquardts
percentage
standard
deviation
(MPSD)
Average
relative
error
(ARE)
Sum of the
absolute
errors
(EABS)
Basic yellow 21
K
F
65.7395 109.5223 71.8887 45.84603 51.67459 108.6377
b
f
0.4315 0.31864 0.400374 0.490939 0.446772 0.317956
R
2
(linear) 0.8609
SNE 3.774516 3.891851 3.422608 4.271745 3.951622 3.827488
Basic blue 3
K
F
99.43459 173.2889 155.0305 140.4162 138.2456 183.9086
b
f
0.31 0.194339 0.21544 0.23517 0.238275 0.18655
R
2
(linear) 0.8644
SNE 5 3.517311 3.48521 3.569177 3.589197 3.569698
Basic red 22
K
F
73.82519 90.53382 77.12983 64.96796 89.86263 95.55466
b
f
0.23 0.190974 0.218394 0.24907 0.190968 0.180186
R
2
(linear) 0.9024
SNE 4.084395 3.968157 3.905895 4.471877 3.940835 4.239594
Table 5
Single-component RedlichPeterson isotherm parameters with error analysis, method/error function (parameter set)
Linear
transform
(LTFM)
Sum of the
squares of the
errors
(ERRSQ)
Hybrid
fractional
error
(HYBRID)
Marquardts
percentage
standard
deviation
(MPSD)
Average
relative
error
(ARE)
Sum of the
absolute
errors
(EABS)
Basic yellow 21
K
R
42.8965 42.5571 34.89284 27.0352 28.27677 28.6733
a
R
0.1207 0.1207 0.084306 0.044371 0.045386 0.047917
0.887 0.8870 0.9143 0.9824 0.9866 0.9796
R
2
(linear) 0.991402
SNE 4.813684 4.773867 4.423551 4.510637 4.366689 4.348626
Basic blue 3
K
R
106.0796 105.2928 97.87177 93.8153 89.462 89.46204
a
R
0.3272 0.3272 0.289182 0.266514 0.232775 0.232771
0.9139 0.9139 0.9232 0.9306 0.9519 0.9519
R
2
(linear) 0.991898
SNE 5 3.411682 3.392184 3.385176 3.432787 3.432784
Basic red 22
K
R
127.3038 126.5776 106.3982 90.92814 76.55711 98.91877
a
R
0.9957 0.9959 0.803577 0.647599 0.562388 0.768273
0.8654 0.8654 0.8731 0.8842 0.8735 0.8659
R
2
(linear) 0.993644
SNE 4.64138 4.590114 4.483429 4.580528 4.680027 4.432419
that all sites are equally available to all adsorbed species:
(15) q
e
=
K
L,i
C
e,i
1 +

a
0
L,i
C
e,i
.
For example, for two components 1 and 2:
(16) q
e
=
K
L,1
C
e,1
1 +a
0
L,1
C
e,1
+a
0
L,2
C
e,2
,
(17) q
e
=
K
L,2
C
e,2
1 +a
0
L,1
C
e,1
+a
0
L,2
C
e,2
.
The single-component parameter sets for the three dyes,
obtained using the LTFM and HYBRID methods, were sub-
stituted into Eqs. (16) and (17) to enable prediction of the
binary component isotherms using the extended Langmuir
isotherm. Examples of the model predictions compared with
experimentally measured results are presented in Figs. 710.
A summary of the R
2
values for the comparison of the
330 S.J. Allen et al. / Journal of Colloid and Interface Science 280 (2004) 322333
Table 6
Single-component Tempkin isotherm parameters with error analysis, method/error function (parameter set)
Linear
transform
(LTFM)
Sum of the
squares of the
errors
(ERRSQ)
Hybrid
fractional
error
(HYBRID)
Marquardts
percentage
Standard
deviation
(MPSD)
Average
relative
error
(ARE)
Sum of the
absolute
errors
(EABS)
Basic yellow 21
A 0.748211 0.7493 0.659603 0.599485 4428.147 4427.967
B 118.69 118.6912 121.4534 123.9125 15.68859 36.42257
R
2
(linear) 0.9645
SNE 4.640624 4.648125 4.262023 4.208684 4.456818 4.418202
Basic blue 3
A 1631.273 1631.079 5222.719 10147.8 4601.941 1500.012
B 35.987 35.9872 29.78419 24.15465 32.75441 37.63971
R
2
(linear) 0.8278
SNE 3.800548 3.800622 3.515037 4.512776 3.388382 3.969258
Basic red 22
A 5.8044 5.8046 4.951031 4.239376 3.455255 4.75042
B 36.635 36.6348 37.33943 38.26656 39.51944 37.90411
R
2
(linear) 0.9526
SNE 4.707511 4.707556 4.537478 4.535131 4.813903 4.510822
Table 7
Single-component Toth isotherm parameters with error analysis, method/error function (parameter set)
Linear
transform
(LTFM)
Sum of the
squares of the
errors
(ERRSQ)
Hybrid
fractional
error
(HYBRID)
Marquardts
percentage
standard
deviation
(MPSD)
Average
relative
error
(ARE)
Sum of the
absolute
errors
(EABS)
Basic yellow 21
K
t
896.9544 794.3424 738.9849 667.2781 703.0084 703.2677
A
t
3.6176 5.3143 9.381696 24.82604 18.55895 15.74947
T 0.5231 0.6269 0.7547 0.9978 0.9169 0.8842
R
2
(linear) 0.991135
SNE 4.539792 4.559929 4.149174 4.2883 4.111165 4.070148
Basic blue 3
K
t
896.9951 597.8487 570.5321 553.8524 526.0231 526.0212
A
t
0.6784 1.7296 2.284722 2.757224 3.683661 3.684526
T 0.2651 0.5366 0.6191 0.6800 0.7899 0.7900
R
2
(linear) 0.988567
SNE 4.998343 4.649591 4.603838 4.588046 4.610444 4.610433
Basic red 22
K
t
896.9950 439.0127 373.1405 332.4155 338.7539 399.3158
A
t
0.4183 0.6787 0.922701 1.233855 1.641035 0.865624
T 0.1433 0.2778 0.3605 0.4502 0.5023 0.3360
R
2
(linear) 0.992522
SNE 4.570612 4.223561 4.025737 4.18857 4.383459 4.004847
predicted isotherms with the measured values for each dye
combination is found in Table 8. Reduced values of equilib-
rium solid-phase dye concentrations of basic dyes adsorbed
onto peat from multicomponent solutions clearly indicated
the competition between constituent dyes for the available
surface area. Preferential uptake of dyes from the mixtures
was found to be related to the relative afnity of the dyes
for each particular adsorbent, as determined from single-
component isothermal analyses. However, results obtained
for the adsorption of binary solutions of Basic red 22 and Ba-
sic yellow 21 proved the exception. With this combination,
the presence of the yellow dye enhanced the afnity of the
red dye for the sorbents, such that molecules of red dye were
adsorbed to a greater extent than molecules of the yellow
dye. This is known as cooperative adsorption. Determina-
tion of the total adsorption capacity for the multicomponent
dye solutions showed that the maximum capacity had in-
creased on contact with dye mixtures. This suggests that the
S.J. Allen et al. / Journal of Colloid and Interface Science 280 (2004) 322333 331
Fig. 7. Extended Langmuir isotherm model and experimental data for Basic
blue 3 in binary solution with Basic red 22.
Fig. 8. Extended Langmuir isotherm model and experimental data for Basic
red 22 in binary solution with Basic blue 3.
Fig. 9. Extended Langmuir isotherm model and experimental data for Basic
blue 3 in binary solution with Basic yellow 21.
Fig. 10. Extended Langmuir isotherm model and experimental data for Ba-
sic yellow 21 in binary solution with Basic blue 3.
Table 8
Multicomponent isotherm R
2
values for the extended Langmuir isotherm
using LTFM and HYBRID isotherm parameter sets
Extended Langmuir R
2
Basic blue 3 R
2
Basic red 22 R
2
Yellow 21
LTFM-Blue/red 0.9232 0.7818
HYBRID-Blue/red 0.9084 0.7176
LTFM-Blue/yellow 0.8745 0.8867
HYBRID-Blue/yellow 0.8176 0.8175
LTFM-Red/yellow 0.8625 0.9017
HYBRID-Red/yellow 0.8044 0.8611
combination of dyes in solution had the effect of increas-
ing the afnity of the peat surface for dye adsorption, either
through an alteration of the overall charge within the system
or by the reorientation of adsorbed molecules. Alternatively,
molecular wedging effects may have been responsible for
the creation of new adsorption sites on the peat particle sur-
face.
In some cases it is evident that there is a good degree of t
using the extended Langmuir model. The results for the blue
dye in particular show evidence that the extended Langmuir
model can be applied to predict binary component isotherms
incorporating the parameter set for the LTFMmodel and also
the HYBRID model. The quality of t falls off for both the
yellow and red dyes. In the majority of cases the quality
of the t is much better at the lower levels of the equilib-
rium isotherms. The LTFM analysis is more biased toward
high concentration values. In the current adsorption systems
there are more data points in the higher concentration region,
so the LTFM analysis will tend to provide a better t. The
single-component isotherms for the three dyes have quite
different monolayer capacities, as suggested by the Lang-
muir isotherm. These are 660, 550, and 310 mg g
1
, respec-
tively, for the yellow, blue, and red dyes. There is clearly
quite a sizable difference. It is possible that the competi-
tion and interaction between the dyes would be signicant.
The extended Langmuir isotherm is based on the same as-
sumptions as the single-component Langmuir isotherm and
therefore the energy of adsorption between the specic dyes
and the surface sites is constant and equal and there should
be no interaction between the dyes and an equal competi-
tion for the adsorption sites. This is clearly not the case and
it must be assumed that there is signicant interaction and
competition occurring, resulting in the extended Langmuir
model failing to apply to this system. The results indicate
that the model makes a good attempt at predicting the higher
capacity dye in each dye pair but fails to predict the cor-
rect monolayer for the lower partner in each combination.
The dye with higher capacity appears to be preferentially
adsorbed by active sites. At lower liquid concentrations well
below the monolayer the solid-phase dye concentrations on
the peat are much closer for the three dyes, suggesting that
interaction and competition is less prevalent, resulting in the
extended Langmuir being able to predict the equilibrium
solid-phase concentrations.
332 S.J. Allen et al. / Journal of Colloid and Interface Science 280 (2004) 322333
4. Conclusions
The equilibrium adsorption of three basic dyes by peat
has been reported. The peat appears to be very effective at
removing the basic dyes from solution. The equilibrium re-
sults have been modeled and evaluated using ve different
isotherms and six different optimization and error functions,
including a linear transform model. The linear transform
model provides the highest R
2
regression coefcient for the
case of the RedlichPeterson isotherm. Using any of the
error functions for nonlinear optimization showed that the
Freundlich isotherm was better represented using the HY-
BRID error function, which compensates for low concentra-
tions by balancing absolute deviation against fractional error.
Also, the RedlichPeterson isothermyields the best t for all
three systems, using the other error functions. Some agree-
ment was observed between predicted extended Langmuir
output and experimental data for the binary systems. The
basic assumptions of the extended Langmuir model, namely
constant energy of adsorption, no interaction, and equal in-
dependent competition between species, appear not to hold
true in this application.
Appendix A. Nomenclature
a
L
Langmuir isotherm constant (dm
3
mg
1
)
a
L,i
Binary Langmuir isotherm constant (dm
3
mg
1
)
a
R
RedlichPeterson isotherm constant (dm
3
mg
1
)
a
T
Toth isotherm constant (dm
3
mg
1
)
A Tempkin isotherm constant
B Tempkin isotherm energy constant (dm
3
g
1
)
b Tempkin isotherm energy constant (J mol
1
)
b
F
Freundlich isotherm exponent
C
0
Initial liquid-phase concentration (mg dm
3
)
C
e
Equilibrium liquid-phase concentration (mg dm
3
)
K
F
Freundlich isotherm constant (dm
3
g
1
)
K
L
Langmuir constant (dm
3
g
1
)
K
R
RedlichPeterson isotherm constant (dm
3
g
1
)
K
T
Toth isotherm constant (mg g
1
)
n Number of isotherm parameters
p Number of data points
Q
0
Monolayer capacity of Langmuir equation (mg g
1
)
q
e
Equilibrium solid-phase concentration (mg g
1
)
q
e,calc
Calculated dye equilibrium solid phase concentra-
tion (mg g
1
)
q
e,meas
Measured dye equilibrium solid phase concentra-
tion (mg g
1
)
R Gas constant (J mol
1
K
1
)
T Temperature (K)
t Toth isotherm exponent
Exponent in RedlichPeterson isotherm

max
Wavelength of maximum absorbance (Nm)
References
[1] T. Robinson, G. McMullan, R. Marchant, P. Nigam, Bioresour. Tech-
nol. 77 (2001) 247.
[2] J.R. Easton, in: P. Cooper (Ed.), Colour in Dyehouse Efuent, Alden,
Oxford, 1995.
[3] J. Pierce, J. Soc. Dyers Colour. 110 (1994) 131.
[4] Y.M. Slokar, A. Majcen Le Marechal, Dyes Pigments 37 (1998) 335.
[5] F.L. Slejko, in: Adsorption Technology, Dekker, New York, 1985.
[6] P.A. Brown, S.A. Gill, S.J. Allen, Water Res. 34 (16) (2000) 3907.
[7] S.J. Allen, G. McKay, K.Y.H. Khader, Environ. Pollut. 52 (1987) 39.
[8] I. Belkevich, K.A. Gayduk, L.R. Christova, in: Proc. 5th Int. Peat
Congr., Pozna n, Poland, 1976, p. 328.
[9] M.E. Leslie, Am. Dyest. Rep. 63 (1974) 15.
[10] M.A. Murray, Adsorption and Desorption of Basic Dyes in the Aque-
ous Phase, Ph.D. thesis, Queens Univ. of Belfast, 1995.
[11] K.Y.H. Khader, Adsorption of Basic Dyes Using Peat and Lignite,
Ph.D. thesis, Queens Univ. of Belfast, 1988.
[12] S.J. Allen, Mathematical Modelling of the Adsorption of Dyes from
Efuent Using Peat, Ph.D. thesis, Queens Univ. of Belfast, 1981.
[13] E.F. Smith, P.F. McCarthy, B.M. Mark, J. Environ. Sci. Health Envi-
ron. Sci. Eng. 11 (1976) 179.
[14] F. Smith, P.F. McCarthy, T.C. Yu, B.M. Mark, J. Water Pollut. Cont.
Fed. 49 (1977) 633.
[15] J. Dufort, M. Ruel, in: Proc. 4th Int. Peat Congr. Otaniemi, Finland,
1972, p. 299.
[16] G. McKay, V.J.P. Poots, J. Chem. Technol. Biotechnol. 30 (1980) 279.
[17] G. McKay, M. El-Geundi, M.M. Nassar, Water Res. 22 (1987) 1527.
[18] A. Boucher, X. Shen, R.G.J. Edyvean, Water Res. 31 (1997) 2084.
[19] R. Matthews, The Adsorption of Dyes Using Kudzu, Peanut Hulls and
MDF Sawdust, Ph.D. thesis, Queens Univ. of Belfast, 2003.
[20] S.J. Allen, Q. Gan, R. Matthews, P.A. Johnston, Bioresour. Technol. 88
(2003) 143.
[21] S.J. Allen, in: G. McKay (Ed.), Use of Adsorbents for the Removal of
Pollutants from Wastewaters, CRC Press, Boca Raton, FL, 1996, p. 59.
[22] H. Zollinger, Colour Chemistry: Syntheses, Properties and Applica-
tion of Organic Dyes and Pigments, second ed., VCH, Weinheim,
1991.
[23] L. Tinghui, T. Matsuura, S. Sourirajan, Ind. Eng. Chem. Prod. Res.
Dev. 22 (1983) 77.
[24] R.S. Juang, S.L. Swei, Sep. Sci. Technol. 37 (1996) 2143.
[25] American Association of Textile Chemists and Society of Dyers and
Colourists, Colour Index, third ed., vol. I.
[26] A.M. James, M.P. Lord, Macmillans Chemical and Physical Data,
Macmillan, London, 1992.
[27] I. Langmuir, J. Am. Chem. Soc. 40 (1918) 1361.
[28] H. Freundlich, Phys. Chem. 57 (1906) 384.
[29] A.V. Tvardovski, Bull. Russ. Acad. Sci. Div. Chem. Sci. 41 (4) (1992)
603.
[30] D.N. Misra, Surf. Sci. 18 (1969) 367.
[31] W. Fritz, E.U. Schlunder, Chem. Eng. Sci. 29 (1974) 1279.
[32] L. Jossens, J.M. Prausnitz, W. Fritz, E.U. Schlunder, A.L. Myer,
Chem. Eng. Sci. 33 (1978) 1097.
[33] H. Moon, W.K. Lee, J. Colloid Interface Sci. 96 (1983) 162170.
[34] B. Al-Duri, G. McKay, Chem. Eng. J. 38 (1988) 23.
[35] G. McKay, J. Soc. Dyers Color. 96 (1980) 576.
[36] O. Redlich, D.L. Peterson, J. Phys. Chem. 63 (1959) 1024.
[37] A.P. Matthews, W.J. Weber Jr., AIChE Symp. Ser. 73 (1976) 91.
[38] M.J. Tempkin, V. Pyzhev, Acta. Physiochim. URSS 12 (1940) 217.
[39] J. Toth, Acta Chem. Acad. Hung. 69 (1971) 311.
[40] G. McKay, J. Chem. Technol. Biotechnol. 32 (1982) 759.
[41] G. McKay, M.J. Bino, A. Altememi, Water Res. 20 (1) (1986) 435.
[42] M. Maekawa, M. Udaka, M. Sasaki, Y. Tujii, H. Yoshida, T. Kataoka,
M. Narrgo, J. Appl. Polym. Sci. 37 (1989) 2141.
[43] S.H. Kost, H.D. Brever, Int. J. Phys. Chem. 96 (2) (1992) 134.
[44] C.H. Giles, D. Smith, A. Huitson, J. Colloid Interface Sci. 47 (1974)
755.
S.J. Allen et al. / Journal of Colloid and Interface Science 280 (2004) 322333 333
[45] C.H. Giles, D. Smith, A. Huitson, J. Colloid Interface Sci. 47 (1974)
766.
[46] R.S. Summers, P.V. Roberts, J. Colloid Interface Sci. 122 (1988) 367.
[47] R.S. Summers, P.V. Roberts, J. Colloid Interface Sci. 122 (1988) 382.
[48] R.H. Myers, in: Classical and Modern Regression with Applications,
PWS-KENT, 1990, pp. 297298, 444445.
[49] D.A. Ratkowski, in: Handbook of Non-Linear Regression Models,
Dekker, New York, 1990.
[50] E. Richter, W. Schutz, A.L. Myers, Chem. Eng. Sci. 44 (8) (1989)
1609.
[51] A. Seidel, D. Gelbin, Chem. Eng. Sci. 43 (1) (1988) 79.
[52] A. Seidel-Morgenstern, G. Guiochon, Chem. Eng. Sci. 40 (2) (1993)
215.
[53] A. Malek, S. Farooq, AIChEJ 42 (11) (1996) 3191.
[54] A.R. Khan, T.A. Al-Bahri, A. Al-Haddad, Water Res. 31 (8) (1997)
2101.
[55] J.F. Porter, G. McKay, K.H. Choy, Chem. Eng. Sci. 54 (1999) 5863.
[56] D.W. Marquardt, J. Soc. Ind. Appl. Math. 11 (1963) 431.
[57] A. Kapoor, R.T. Yang, Gas. Sep. Purif. 3 (4) (1989) 187.
[58] McKay, G. Ramprasand, P. Mowli, Water Res. 21 (1987) 375.
[59] L.H. Ong, V.E. Swanson, Econ. Geol. 61 (1966) 1214.
[60] S.J. Allen, G. McKay, K.Y.H. Khader, Environ. Pollut. 52 (1988)
39.
[61] S.J. Allen, G. McKay, K.Y.H. Khader, J. Colloid Interface Sci. 126
(1988) 517.
[62] S.J. Allen, G. McKay, K.Y.H. Khader, Environ. Pollut. 56 (1989) 9.
[63] S.J. Allen, G. McKay, K.Y.H. Khader, J. Chem. Technol. Biotech-
nol. 45 (1989) 291.
[64] S.J. Allen, P.A. Brown, M.A. Murray, O. Flynn, J. Resour. Conserv.
Recycl. 11 (1994) 25.
[65] S.J. Allen, M. Murray, Dirasat 22 (1995) 423.
[66] S.J. Allen, L.J. Whitten, M.A. Murray, O.P. Duggan, P.A. Brown,
J. Chem. Technol. Biotechnol. 68 (1997) 442.
[67] J.A.V. Butler, C. Ockrent, J. Phys. Chem. 34 (1930) 2841.