Chemical Engineering Journal 229 (2013) 450–459

Contents lists available at SciVerse ScienceDirect

Chemical Engineering Journal

journal homepage: www.elsevier.com/locate/cej

Adsorbed solution theory based modeling of binary adsorption of

nitrobenzene, aniline and phenol onto granulated activated carbon
Ananda Jaysing Jadhav, Vimal Chandra Srivastava ⇑
Department of Chemical Engineering, Indian Institute of Technology Roorkee, Roorkee 247 667, Uttarakhand, India

h i g h l i g h t s

 Simultaneous adsorptive removal of three binary systems by granular activated carbon.

 Binary system studied: nitrobenzene–aniline, nitrobenzene–phenol and phenol–aniline.
 Single component isotherm modeling by Langmuir, Freundlich and Redlich–Peterson models.
 Binary adsorption data modeled by ideal and real adsorbed solution theory (IAST and RAST) models.
 Calculation of activity coefficients.

a r t i c l e i n f o a b s t r a c t

Article history: The modeling study on simultaneous adsorption of three binary systems (nitrobenzene–aniline (NI–AN),
Received 13 March 2013 nitrobenzene–phenol (NI–PH) and phenol–aniline (PH–AN)) onto granular activated carbon (GAC) in
Received in revised form 8 June 2013 aqueous solution were performed at 30 °C by conducting batch experiments. The single solute equilib-
Accepted 10 June 2013
rium adsorption data of nitrobenzene (NI), aniline (AN) and phenol (PH) were fitted with Langmuir, Fre-
Available online 19 June 2013
undlich and Redlich–Peterson model. The Redlich–Peterson and Freundlich model gave better fitting as
compared to Langmuir model for individual adsorption. The binary adsorption data were examined
Keywords:
and compared by using ideal adsorbed solution theory (IAST) and real adsorbed solution theory (RAST)
Ideal adsorbed solution theory
Real adsorbed solution theory
models. IAST model didn’t provide an acceptable prediction of binary data except for low liquid concen-
Granular activated carbon tration levels, as it undervalued AN and PH adsorption capacity and overvalued NI adsorption capacity.
Nitrobenzene This is due to the non-ideality of binary mixtures at high concentration levels in the solution. The RAST
Aniline model gave an excellent prediction of binary adsorption experimental data, thus it can be used as a reli-
Phenol able model for the design of industrial adsorption equipment.
Ó 2013 Elsevier B.V. All rights reserved.

1. Introduction methods for the removal of these chemicals from wastewater

Benzene derivatives such as nitrobenzene, aniline and phenol
are considered as toxic, carcinogenic and/or mutagenic agents.
These compounds are frequently used in chemical processes as
raw materials in the industries like plastic, paint, pharmaceutics,
pesticides, dyes, gasoline, rubber proofing, disinfectant, steel
industries, lubricating oils refinement, soaps or shoe polishes;
and as intermediates in the chemical synthesis industries [1–4].
They are introduced into the surface water from industrial efflu-
ents and their existence in wastewater even in very small concen-
trations is harmful to animals and human beings.
A number of technologies are available for removal of these
chemical from wastewater such as adsorption [5–9], photochemi-
cal reduction [10], oxidation by O3/UV processes [11,12] and bio-
degradation [13]. Adsorption is one of the most frequently used
⇑ Corresponding author. Tel.: +91 1332 285889; fax: +91 1332 276535.
E-mail address: vimalcsr@yahoo.co.in (V.C. Srivastava).
[9,14–15]. A considerable amount of work on the adsorption of or-
ganic chemicals has dedicated on the uptake of single solute by
various kinds of adsorbents such as granular activated carbons
(GACs), and polymeric adsorbents. However, it is necessary from
industrial point of view to study the interactive effect of the
adsorption of one compound on another during simultaneous
adsorption of toxic compounds.
The design of adsorption units essentially depends on adsorbent
adsorption capacity at equilibrium condition; and their perfor-
mance is influenced by the competitive interactions of the com-
pounds present in mixture. Hence, the assessment of the
adsorption capacity of a compound in the presence of other solutes
by suitable theoretical models is necessary for proper design of a
cost effective adsorption system [16]. Various multicomponent
adsorption models such as the modified competitive Langmuir
model [17], extended Langmuir model [18], extended Freundlich
model [19] and Sheindorf–Rebuhn–Sheintuch model [20] can be
used to describe the equilibrium characteristics of multicomponent

1385-8947/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.cej.2013.06.021
 A.J. Jadhav, V.C. Srivastava / Chemical Engineering Journal 229 (2013) 450–459 451

Table 1 behavior (each of the adsorbates present in the mixture have

Comparison of individual adsorption equilibrium uptakes and yields of nitrobenzene, interactive effect).
aniline and phenol adsorption onto GAC.
The basic equation that represents the equivalence between the
Co (mmol/L) Ce (mmol/L) qe (mmol/g) Ad (%) chemical potential of liquid and adsorbed phases for each compo-
NI nent is expressed as follows:
0.2 0.002 0.020 98.98
0.4 0.006 0.040 98.55
C i ¼ C 0i ðT; P; wÞxi i ¼ 1; 2; . . . ; N ð1Þ
0.6 0.010 0.060 98.32
0.8 0.015 0.080 98.15 In case of non-ideal system Eq. (1) can be modified as follows
1.2 0.030 0.119 97.54
1.6 0.048 0.158 97.04
C i ¼ C 0i ðT; P; wÞxi ci i ¼ 1; 2; . . . ; N ð2Þ
AN
0.3 0.015 0.025 94.29 where N is the number of solutes present in a mixture, Ci is the equi-
0.5 0.045 0.049 91.69 librium liquid concentration of each solute, C 0i is the liquid phase
0.8 0.092 0.071 88.55
1.1 0.149 0.092 86.08
concentration which is in equilibrium with the resultant single-
1.6 0.283 0.133 82.46 component adsorption capacity (q0i ), at pressure (P), temperature
2.1 0.442 0.171 79.39 (T) and spreading pressure (w), ci activity coefficient of the adsorbed
PH component and xi is the mole fraction of each component on adsor-
0.3 0.017 0.025 93.56 bent, that can be expressed as:
0.5 0.043 0.049 91.98
0.8 0.078 0.072 90.24 qi
1.1 0.111 0.095 89.59 xi ¼ i ¼ 1; 2; . . . ; N ð3Þ
qT
1.6 0.190 0.140 88.08
2.1 0.291 0.183 86.32 The IAST models also assume that adsorbent is thermodynami-
cally inert and having constant area for adsorption per mole of sol-
adsorption. Basically they are the extensions of single-component ute for single or mixture of components. These assumptions are
isotherm equations. Ideal adsorbed solution theory (IAST) model expressed by following equation
and real adsorbed solution theory (RAST) better represent the mul-
ti-component adsorption behavior. Details of these theories are gi- 1 XN
xi
¼ 0
i ¼ 1; 2; . . . ; N ð4Þ
ven in Section 2. qT q
i¼1 i
In the present study, the Langmuir, Freundlich, and Redlich–
Peterson models were used for the fitting of experimental equilib- where qT is total surface loading on adsorbent i.e. sum of adsorbate
rium adsorption data of NI, AN, and PH. The IAST and RAST model adsorption capacity of adsorbent.
were used for the analysis of equilibrium adsorption isotherms for For non-ideal system the total surface loading at constant tem-
three binary systems namely nitrobenzene–aniline (NI–AN), nitro- perature can be calculated from the Gibbs-Duhem [25] relation as
benzene–phenol (NI–PH) and phenol–aniline (PH–AN) on to GAC follows
at 30 °C. To check the interactive effect of initial concentration of
X   !
one species onto the other, a number of experiments were carried 1 N
xi XN
@ ln ci ðwÞ
¼ þ xi i ¼ 1; 2; . . . ; N ð5Þ
out at different initial concentration ratios. The initial concentra- qT q0 i¼1
i¼1 i
@w T;xi
tion ratios for (NI–AN) system was 1:0.67, 1:1.3 and 1:2.6, for
(NI–PH) system was 1:0.67, 1:1.3 and 1:2.6, and for (PH–AN) sys- The spreading pressure for single as well as multi-component
tem was 1:0.5, 1:1 and 1:2. The RAST model is derived from IAST system is equal and it is derived from Gibbs adsorption isotherm,
model under the hypothesis of adsorbed phase does not follow as per following equation [22]:
the ideal behavior. The non-ideal behavior of the system was taken Z C 0i
q0i ðC 0i Þ 0
into account by including activity coefficient in to the basic equa- wi ¼ dC i i ¼ 1; 2; . . . ; N ð6Þ
tion of IAST that represents the equality between the chemical po- 0 C 0i
tential of liquid and adsorbed phases for each component. The where wi is the spreading pressure for individual components which
activity coefficient which is a function of spreading pressure was is dimensionless, q0i ðC 0i Þ is single-component adsorption isotherm.
calculated from experimental binary adsorption data. The IAST and RAST can be combined with single component
adsorption equations in order to calculate the spreading pressure.
2. Theory In this present work, Freundlich isotherm is incorporated into IAST
and RAST to predict the multicomponent adsorption isotherm. So
The models used and compared in present work are IAST and Eq. (6) can be modified by incorporating the Freundlich isotherm
RAST, these are multicomponent adsorption model developed by as follows:
Myers and Prausnitz [21] for gaseous adsorption and then modi- Z C 0i
fied for liquid solution by Radke and Prausnitz [22]. They provide q0i ðC 0i Þ 0 0n1
wi ¼ dC i ¼ ni K Fi C i i
i ¼ 1; 2; . . . ; N ð7Þ
a thermodynamically dependable and practical method for pre- 0 C 0i
diction of multicomponent adsorption system using single-com-
ponent isotherm data [16,23]. IAST model is best suitable for a where KFi and ni are Freundlich isotherm parameters for component
system which show non-interactive behavior. This model is based i.
on the assumptions that all component in a mixture behave as As per the IAST theory, at equilibrium the spreading pressure
ideal adsorbed solutes according to Raoult’s law which means for each individual as well as mixture of components is equal, as
that there is no interaction between adsorbent–adsorbate is con- per following equation
sidered [24] as well as the chemical potential in adsorbed phase
w1 ¼ w2 ¼ w3 ¼    ¼ wi ¼    ¼ wN ¼ w ð8Þ
and liquid phase is equal for each component therefore, the
spreading pressure is also constant for all components in a given The sum of all the mole or mass fraction in each phase must be
mixture. RAST is best suitable for a system shows antagonistic exactly equal to unity; for the solid phase
452 A.J. Jadhav, V.C. Srivastava / Chemical Engineering Journal 229 (2013) 450–459

X
N 2.3. Solution of RAST
xi ¼ 1 ð9Þ
i¼1
The solution of RAST model includes 3N + 1 equations (N equa-
The IAST model equation set (1), (3), (4), (7), (8), and (9) can be tion obtained from Eq. (2), N equation obtained from Eq. (7), N
used to predict multicomponent adsorption isotherms. The RAST equation obtained from Eq. (10) and one equation from Eq. (9)),
model equation set is completed by addition of equations that al- and 3N + 1 variables (N variables of xi, N variables of C 0i , N variables
lows the calculation of activity coefficient. of ci (T, p) and single variable w). The initializations for solution by
Newtons–Raphson method in case of RAST were similar to those of
IAST.
2.1. Calculation of activity coefficients
The goodness of the fit of experimental data and the prediction
accuracy of single component adsorption models, IAST and RAST
In this work the activity coefficient is calculated from known
models as well as Wilson’s activity coefficient equation were eval-
theoretical Wilson activity coefficient model as follows [26]
uated using statistical indices such as sum square error (SSE), root
ln ci ¼ ½1 mean square error (RMSE), Marquardt’s percent standard deviation
" ! !# (MPSD), average relative error (ARE) and coefficient of determina-
X
N X
N
xk Kki
 expðcwÞ 1  ln xj Kij  PN ð10Þ tion (R2) as follows:
j¼1 k¼1 j¼1 xj Kkj
SSE ¼ ðqe;cal  qe;exp Þ2 ð14Þ
where N is the number of components of the multicomponent sys- rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
tem and Kij , c are adjustable parameters. These parameters can be 1 XN
calculated by performing nonlinear regression on experimental RMSE ¼ ðqexp  qcal Þ2i ð15Þ
N i¼1
activity coefficient data; the calculation process is given in detail
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
8
as follows. u
u !2 9
<Xn qe;cal  qe;exp =
Starting with the calculation of spreading pressure for experi- u 1 XN
MPSD ¼ 100  t i¼1 :
ð16Þ
mental mixture (w); it can be calculated as ðN  pÞ j¼1 qe;exp ;
j i

w ¼ nT K FT C 1=n
T
T
ð11Þ  
N  
100 X qe;cal  qe;exp 
where KFT and nT are Freundlich isotherm parameters and ARE ¼   ð17Þ
P N i¼1  qe;exp 
C T ¼ 2i¼1 C i is the total solute concentration in liquid phase. CT i

can be calculated by adding the equilibrium liquid phase concentra- PN

tion (Ci) of both the solute present in the binary system. Then the i¼1 ðqe;exp  qe;cal Þ2i
R2 ¼ PN PN ð18Þ
Freundlich isotherm parameters, nT and KFT are calculated using
P i¼1 ðqe;exp  qe;cal Þ2i þ 2
i¼1 ðqe;exp  qe;cal Þi
regression analysis with experimental qT ¼ 2i¼1 qi values calculated
similarly by adding the equilibrium solid phase concentration (qi) of where qe,exp and qe,cal are experimental and calculated adsorbed
both the solute present in the binary system. This process is applied phase concentration respectively, N is the number of measurements
for each three binary system and calculated spreading pressure for and p is the number of parameters in the model.
experimental mixtures. Since at equilibrium the spreading pressure The experimental solid phase concentration (qe) (mg of adsor-
is constant for each component in mixture as shown in Eq. (8); so bate/g of adsorbent) and experimental percentage adsorption yield
single-solute liquid phase concentration (C 0i ) for each individual (Ad) for each individual component were calculated using follow-
compound can be calculated by performing back calculation and ing expressions:
putting Freundlich isotherm parameters for corresponding compo- ðC o  C e Þ  V
nent in Eq. (7). Finally from available experimental data on Ci and qe ¼ ð19Þ
W
xi, the activity coefficients (ci ) can be calculated by using Eq. (2).
The RAST model equation set (2), (3), (5), (7)–(10), (and) (11) can ðC o  C e Þ
be used for non-ideal system. Adð%Þ ¼  100 ð20Þ
Co
where Co is the initial liquid phase concentration of adsorbate
2.2. Solution of IAST
(mmol/g), Ce is the residual liquid phase concentration of adsorbate
(mmol/g), V is the volume of the adsorbate containing solution (L)
The activity coefficient in case of IAST was taken as equal to
and W is the mass of the adsorbent (g).
unity. The solution of IAST model involves 2N + 1 equations (N
equation obtained from Eq. (1), N equation obtained from Eq. (7)
and one equation obtained from Eq. (9)), and 2N + 1 variables (N 3. Materials and experimental
variables of xi, N variables of C 0i and single variable w). These equa-
tions can be solved by the Newtons–Raphson method. The initial 3.1. Adsorbent and its characterization
estimates were taken as follows:
GAC was supplied by Zeo-Tech Adsorbents, India. This GAC was
q
ðxi Þinitial ¼ PN i ð12Þ washed thoroughly with distilled water and then it was dried at a
i¼1 qi temperature of 105 °C for 72 h and stored in an air free container.
Proximate analysis was carried out according to Bureau of Indian
Ci Standards, detailed procedure is given in IS: 1350 (Part I) [27]. Bulk
ðC 0i Þinitial ¼ ð13Þ density was determined by using MAC bulk density meter. X-ray
ðxi Þinitial
diffraction analysis of GAC was performed on Phillips (Holland) dif-
and w was taken close to 0.00001 N/m. Once the conversion oc- fraction unit (Model PW 1140/90), using copper target with nickel
curred, the adsorbed phase concentration can calculated using as filter media, and radiation maintained at 1.542 Å. Goniometer
Eqs. (3) and (4). speed was maintained at 1°/min. The specific surface area was
 A.J. Jadhav, V.C. Srivastava / Chemical Engineering Journal 229 (2013) 450–459
measured using an ASAP 2010 Micromeritics instrument and by
Brunauer–Emmett–Teller (BET) method. Liquid nitrogen was used
as cold bath (77.15 K). The functional groups present in GAC were
determined using FTIR spectrometer (Thermo Nicolet, Model Mag-
na 760) at room temperature.
3.2. Sorbates and chemicals
All the used chemicals were of analytical reagent grade. Nitro-
benzene and phenol were purchased from S.D. Fine Chemicals,
Mumbai while aniline was supplied from E. Merck (India) limited,
Mumbai. Stock solutions of nitrobenzene, aniline and phenol were
prepared using double-distilled water. All solutions were prepared
fresh whenever required.
3.3. Batch adsorption studies
For single component systems, initial concentration (Co) was
varied from 0.2 to 1.6 mmol/L for NI, 0.3 to 2.1 mmol/L for AN
and 0.3 to 2.1 mmol/L for PH and the adsorbent dosage was kept
at m = 10 g/L. Each three binary systems were prepared in three
set of different ratios such as for (NI–AN) system is 1:0.67, 1:1.3
and 1:2.6, for (NI–PH) system is 1:0.67, 1:1.3 and 1:2.6, and for
(PH–AN) system is 1:0.5, 1:1 and 1:2 keeping Co 0.2, 0.4, 0.6, 0.8,
1.0, 1.2, 1.4, 1.8 and 2.0 mmol/L for NI and 0.3, 0.5, 0.8, 1.1, 1.3,
1.6, 1.9, 2.1 and 2.7 mmol/L for AN and PH. The adsorbent dose
was kept at m = 10 g/L.
For each experimental run, 0.1 L of known concentration of
nitrobenzene, aniline, phenol and their binary mixture were taken
in a 0.25 L glass stoppered conical flask having 1 g of GAC. The
constant shaking at the rate of 150 rpm for 8 h was provided
using temperature-controlled orbital shaker (Remi Instruments,
Mumbai) kept at 30 °C. For single component system, the residual
solution concentrations of nitrobenzene, aniline and phenol were
analyzed by UV–vis spectrophotometer (model Lambda 35, Per-
kin–Elmer, Germany) at wavelengths of 268, 230 and 269 nm
respectively. For binary systems, the residual concentration of
components from the mixture were quantified by using partial
least square regression method and the wavelength selection
was based on the genetic algorithm optimization method. The ge-
netic algorithm has been used for selection of wavelength in
simultaneous spectrophotometric quantitative analysis of nitro-
benzene, aniline and phenol in a binary mixture using partial
least-squares calibration method. Due to the interference in the
spectrum of each compound by the spectrum of other compound,
direct determination of residual concentration of nitrobenzene,
aniline and phenol in a binary mixture after adsorption by using
wavelength of maximum absorbance is not possible. The absorp-
tion spectra of binary solutions containing nitrobenzene, aniline
and phenol in various concentrations were recorded in the range
from 200 to 340 nm in the intervals of k = 1 nm at 141 wave-
length. For calibration, 16 samples each binary system containing
nitrobenzene, aniline and phenol were prepared in double-dis-
tilled water. The optimum wavelengths were selected out of
141 wavelengths using genetic algorithm optimization method.
The Partial least square calibration was done on selected absorp-
tion spectra, the model equations were developed for the predic-
tion of concentrations of nitrobenzene, aniline and phenol from
binary mixtures as follows:
(1) Model equations for individual components present in NI–
AN system
C NI ¼ 0:172 þ 38::767Aatk¼335 þ 37:328atk¼336
þ 36:066atk¼337 þ 34:735atk¼338
453
C AN ¼ 0:133 þ 0:939Aatk¼¼203 þ 1:475atk¼204  5:269atk¼205
 2:593atk¼206 þ 14:142atk¼223 ð22Þ
(2) Model equations for individual components present in
NI–PH system
C NI ¼ 0:219 þ 7:215Aatk¼228 þ 6:867atk¼289 ð23Þ
C PH ¼ 0:310 þ 35:789Aatk¼¼267  60:551atk¼295
 52:731atk¼296 þ 43:158atk¼314 þ 42:850atk¼315 ð24Þ
(3) Model equations for individual components present in
PH–AN system
C PH ¼ 0:090 þ 102:553Aatk¼271  65:307atk¼272  424:690atk¼293 þ 365:052atk¼294 þ
185:269atk¼334  340:550atk¼335  119:0atk¼336
ð25Þ
C AN ¼ 0:398 þ 5:460Aatk¼¼201  5:160atk¼202  2:288atk¼220
þ 1:814atk¼224 þ 2:152atk¼225 þ 2:630atk¼226 ð26Þ
whereas A is an absorbance at k (wavelength) nm.
The quantitative determination of components from their mix-
ture using partial least square regression method and the wave-
length selection using genetic algorithm optimization method
was also reported in literature [28–31]. The numerical calculations
were performed on MATLAB 7.8 software and MSExcel.
4. Results and discussion
4.1. Characterization of GAC
Moisture content, volatile matter, ash, fixed carbon and bulk
density of GAC were 3.60%, 28.30%, 2.64%, 65.46% and 591.13 kg/
m3 respectively. The isotherms of adsorption/desorption of N2 at
77.15 K on the GAC, X-ray diffraction pattern of GAC, and FTIR
spectroscopy of virgin GAC are shown in Fig. 1.
The BET surface area was found to be 591.06 m2 g1 whereas
Barrett–Joyner–Halenda (BJH) adsorption/desorption surface area
of pores between 17 Å and 3000 Å is 86.12/74.75 m2 g1. The single
point adsorption total pore volume of pores (d < 209.93 Å) is
0.901 cm3 g1, whereas BJH adsorption cumulative pore volume
of pores (17 Å < d < 3000 Å) is 0.052 cm3 g1. The average pore
diameter by BET method is found to be 21.15 Å, whereas the BJH
adsorption/ desorption average pore diameter is 24.02 Å/24.27 Å.
X-ray spectrum of GAC (Fig. 1b) reflected the amorphous nature
of GAC. Diffraction peaks corresponding to crystalline carbon were
not observed in GAC. FTIR spectra (Fig. 1c) reflected the presence of
both free and hydrogen bonded OH groups on the GAC surface was
pointed out by a broad band between 3100 and 3700 cm1 [32].
This stretching indicates the presence of both silanol groups (Si–
OH) and adsorbed water (peak at 3400 cm1) on the surface [33].
The 1360–1380 cm1 band specify the aromatic CH and car-
boxyl–carbonate structures [34]. The peak at 575 cm1 is due to
O–C–O scissors and the peak at 1620 cm1 shows the NH deforma-
tion. The presence of polar groups on the GAC surface provides the
significant cations exchange capacity to the adsorbents [35].
4.2. Adsorption of individual adsorbates
The equilibrium uptakes and the adsorption yields for NI, AN,
and PH were calculated using Eqs. (19) and (20), respectively,
and the results are shown in Table 1 and Fig. 2. It is observed that
as the initial concentration increases, the equilibrium uptake also
increases but adsorption yield goes on decreasing for NI, AN, and
ð21Þ PH. The adsorption capacity of the GAC was in the order
454 A.J. Jadhav, V.C. Srivastava / Chemical Engineering Journal 229 (2013) 450–459
Fig. 1. The Characterization of GAC (A) isotherms of adsorption/desorption of N2 at 77 K, (B) X-ray diffraction pattern, and (C) FTIR spectroscopy of unloaded GAC.
Fig. 2. Equilibrium adsorption isotherms for individual adsorption of nitrobenzene,
aniline and phenol onto GAC.
NI > AN > PH. This trend is according to the decreasing molecular
size of the compounds: NI (5.9 Å) [36] <AN (7.0 Å) [37] <PH
(7.92 Å) [38]. Also the solubility of NI, AN, and PH in water at
20 °C is 1900 mg/L, 36000 mg/L and 83000 mg/L, respectively.
Therefore, NI is more hydrophobic in nature and consequently it
gets more adsorbed onto GAC. The dielectric constant of NI
(34.8), AN (6.7) and PH (10) indicate that NI has strongest polarity
among all the adsorbates, and therefore, it can easily get adsorbed
on the GAC surface because of the polar interaction. These values
also indicate that the PH is more polar than AN, still adsorption
of AN is more than that of PH. This is due to the fact that AN is
slightly basic in nature whereas PH is slightly acidic; and because
of its basic nature, aniline gets better adsorbed on the acidic sur-
face of the adsorbent.
Surface functional groups present on the GAC also affect the
adsorption behavior of various adsorbates. Presence of acidic
groups increases the oxygen content, and thus increases the
adsorption of polar compounds such as NI. Considering the elec-
tron donor–acceptor parameter, NI is stronger electron acceptor
than AN, whereas PH is strong electron donor. Due to presence of
polar groups on the GAC surface, p–p interaction occurs between
NI (p acceptor) and polar group of GAC (p donor). The surface prop-
erties of adsorbent, hydrophobicity, polar interaction and p–p
interaction may be the important mechanisms affecting the
adsorption of NI, AN and PH from aqueous solution onto GAC. All
the possible mechanisms are more favorable for the adsorption
of NI onto GAC surface as compared to AN and PH.
4.3. Single-component adsorption isotherm modeling
The isotherm fitting parameters of Langmuir, Freundlich and
Redlich–Peterson isotherm models with experimental data of NI,
AN, and PH with statistical indices values are listed in Table 2. Fre-
undlich and R–P isotherm models better represent the experimen-
tal equilibrium sorption data than the Langmuir isotherm model.
Therefore, it is expected that the GAC has a heterogeneous surface
for the adsorption of NI, AN, and PH. The MPSD error values are
least and R2 values approach to one for the R–P model followed
by those for the Freundlich and Langmuir models. Since Freundlich
model for single component adsorption isotherm offers best fitting,
so it has been included in the IAST and RAST model formulation
(i.e. modification of Eq. (6) into Eq. (7)).
 A.J. Jadhav, V.C. Srivastava / Chemical Engineering Journal 229 (2013) 450–459 455

Table 2
Isotherm parameters values for the adsorption of nitrobenzene, aniline and phenol from water by GAC.

Adsorbate KL (L/mmol) qm (mmol/g) SSE  105 RMSE MPSD ARE R2

qm K L C e
Langmuir isotherm: qe ¼ 1þK L C e
Nitrobenzene 41.11 0.21 22.5 0.006 10.2 7.13 0.987
Aniline 8.66 0.18 92.4 0.012 14.14 10.18 0.95
Phenol 4.76 0.29 27.9 0.007 8.94 6.09 0.988
KF (mmol/g)/(mmol/L)1/n) 1/n SSE  105 RMSE MPSD ARE R2

Freundlich isotherm: qe ¼ K F C 1=n

e
Nitrobenzene 1.2 0.659 4.6 0.002 3.46 2.34 0.998
Aniline 0.27 0.562 0.6 0.001 2.14 1.42 1
Phenol 0.45 0.708 1.9 0.002 1.99 1.58 0.999

KR (L/g) aR (mmol/g) b SSE  105 RMSE MPSD ARE R2

K R Ce
Redlich–Peterson isotherm: qe ¼
1þaR C be
Nitrobenzene 25.05 25.66 0.449 1.3 0.001 2.61 1.56 0.999
Aniline 8.40 31.32 0.489 0.8 0.001 2.12 1.22 0.999
Phenol 6.66 14.47 0.344 1.3 0.001 2.11 1.27 0.999

0.4 0.4
qeAN (mmol/g)
0.3 0.3
qeNI (mmol/g)

0.2 0.2

0.1 NI-AN (1:0.65) 0.1 NI-AN (0.38:1)

NI-AN (1:1.3) NI-AN (0.75:1)
NI-AN (1:2.6) NI-AN (1.5:1)
0.0 0.0
0.0 0.2 0.4 0.0 1.0 2.0
CeNI (mmol/L) CeAN (mmol/L)
(a) NI-AN binary system.

0.4 0.5

0.3 0.4
qePH (mmol/g)
qeNI (mmol/g)

0.3
0.2
0.2
0.1 NI-PH (1:0.65) NI-PH (0.38:1)
NI-PH (1:1.3) 0.1 NI-PH (0.75:1)
NI-PH (1:2.6) NI-PH (1.5:1)
0.0 0.0
0.0 0.2 0.4 0.6 0.0 0.5 1.0 1.5
CeNI (mmol/L) CePH (mmol/L)
(b) NI-PH binary system
0.5 0.5
0.4 0.4
qeAN (mmol/g)
qePH (mmol/g)

0.3 0.3
0.2 0.2
PH-AN (1:0.5) PH-AN (0.5:1)
0.1 PH-AN (1:1) 0.1 PH-AN (1:1)
PH-AN (1:2) PH-AN (2:1)
0.0 0.0
0.0 0.5 1.0 1.5 0.0 0.5 1.0 1.5
CePH (mmol/L) CeAN (mmol/L)
(c) PH-AN binary system
Fig. 3. Equilibrium adsorption data of individual compounds in various binary mixtures. (a) NI–AN system, (b) NI–PH system, and (c) PH–AN system.
456 A.J. Jadhav, V.C. Srivastava / Chemical Engineering Journal 229 (2013) 450–459

Table 3 favorably adsorbed by GAC. The KF value indicates highest uptake

Statistical indices values between the experimental and the theoretically predicted qe of NI followed by that of PH and AN in that order.
by IAST and RAST for NI–AN, NI–PH and PH–AN binary systems.
The value of the R–P isotherm constant 0 6 b 6 1 shows favor-
System Model SSE RMSE MPSD ARE R2 able nature of adsorption. For NI, AN, and PH the b values were
NI–AN IAST 0.074 0.037 27.53 13.86 0.91 found to be 0.449, 0.489, and 0.344, respectively. Therefore, NI,
RAST 0.027 0.022 19.46 10.28 0.93 AN, and PH are adsorbed favorably by GAC.
NI–PH IAST 0.131 0.049 32.22 17.06 0.87
RAST 0.012 0.015 16.18 7.81 0.98
4.4. Binary adsorption isotherm modeling
PH–AN IAST 0.099 0.043 36.19 21.43 0.84
RAST 0.015 0.016 14.61 7.74 0.98
In this work, three sets of binary adsorption data of NI–AN, NI–
PH and PH–AN onto GAC have been analyzed by IAST and RAST.
The experimental liquid phase and solid phase concentration data
KF and 1/n are the Freundlich adsorption isotherm parameters for three binary system can be used for the finding of competitive
which indicate the adsorption capacity of adsorbent and adsorp- effect between NI, AN, and PH molecules for adsorption sites onto
tion intensity of adsorbate, respectively. The higher value of n indi- the GAC. Equilibrium adsorption data of individual compounds in
cates the nonlinear adsorption isotherm, and as n becomes greater three binary mixtures sets are plotted in Fig. 3. It may be seen that
than 10 then it is called rectangular or irreversible isotherm. The as the initial concentration of one compound increases, it results in
lower the value of 1/n represents the higher the affinity and the decrease in adsorption intensity of other compound. The results of
heterogeneity of the adsorbent sites. From Table 2 it is concluded IAST an RAST models are then compared on the basis of statistical
that the GAC shows the greatest heterogeneity for AN followed indices values calculated between the experimental and predicted
by that for NI and PH. Since 1/n < 1, all three components are adsorbed phase concentration values of nitrobenzene, aniline and

Fig. 4. Comparison of the experimental and calculated qe values for individual compounds in various binary mixtures. (a) NI–AN system, (b) NI–PH system, and (c) PH–AN
system.
 A.J. Jadhav, V.C. Srivastava / Chemical Engineering Journal 229 (2013) 450–459 457

phenol for three binary systems. These values are shown in Table 3.
The experimental and predicted adsorbed phase concentration val-
ues for nitrobenzene, aniline and phenol for three binary systems
are also plotted in Fig. 4. From Table 3 and Fig. 4, it is clear that
the RAST model provides the acceptable fitting in case of higher
as well as lower concentrations of nitrobenzene, aniline and phe-
nol for all three binary system than IAST.
The IAST model was unable to offer an adequate prediction of
equilibrium adsorption data for three binary systems, it underval-
ues the NI, AN, and PH adsorption capacity. It was observed that
when the initial concentration of solute increases, these binary
mixtures behavior becomes non-ideal in nature. The experimental
data was also analyzed with RAST model by taking into consider-
ation of non-ideal behavior of the system by introduction of activ-
ity coefficient. The experimental activity coefficient have been
calculated from experimental data of NI–AN, NI–PH and PH–AN
binary systems using Eq. (2) and reported in Fig. 5 as a function
of the experimental spreading pressure, calculated from Eq. (11).
From the experimental data, it was found that activity coeffi-
cient values for AN and PH are in their respective binary systems
are always less than one showing that the experimental adsorbed
phase concentration values are greater than the corresponding val-
ues predicted by IAST model. Conversely, activity coefficient values
for NI in are greater than unity, excluding few experimental test, in
both of its binary systems indicating that the experimental ad-
sorbed phase NI concentration values are lower than the corre-
sponding values predicted by IAST model.
Furthermore, it was observed that an increase in ratio of con-
centration of NI/AN for NI–AN system and NI/PH for NI–PH re-
sulted in an increase in activity coefficient of AN and PH with
spreading pressure in sharp crescent manner, respectively.
In case of PH–AN system, the values of activity coefficient for PH
increased with an increase in spreading pressure. The values of AN
show are almost constant with variation in spreading pressure for
PH–AN system having AN/PH ratio of 0.5 and 1. However, when the
AN/PH ratio is increased to 2, the activity coefficient values of AN
start decreasing with an increase in spreading pressure for smaller
values of spreading pressure (<0.5). For higher values of spreading
pressure (>0.6) activity coefficient values of AN become constant. It
may be pointed out that these values are experimental values
which have been calculated using the procedure discussed in Sec-
tion 2.1. Erto et al. [26] have also reported decrease in the values of
activity coefficient of tetrachloroethylene for simultaneous adsorp-
tion of tetrachloroethylene and trichloroethylene onto granular
activated carbon.
After determining experimental activity coefficients, the adjust-
able parameters (c, Kij) of Eq. (10) were determined by doing Fig. 5. Experimental activity coefficient as a function of spreading pressure at
regression analysis of experimental data represented such as different initial concentration ratio for three binary mixtures. (a) NI–AN system, (b)
spreading pressure (w), mole fraction (xi) and activity coefficient NI–PH system, and (c) PH–AN system.
(ci). For a binary system, Eq. (10) can be written as:
    
x1 x2 K21 Table 4
ln c1 ¼ ½1  expðcwÞ 1  lnðx1 þ x2 K12 Þ   ð27Þ
x1 þ x2 K12 x1 K21 þ x2 Statistical indices values and adjustable parameters of Wilson activity coefficient
model for three binary systems.
    
x1 K12 x2 System c K12 K21 SSE RMSE MPSD ARE R2
ln c2 ¼ ½1  expðcwÞ 1  lnðx1 K21 þ x2 Þ   ð28Þ
x1 þ x2 K12 x1 K21 þ x2
NI–AN 2.415 0.212 0.288 1.348 0.223 12.441 2.305 0.98
The regression analysis on Eq. (27) for the entire experimental NI–PH 3.102 0.218 0.227 2.039 0.275 16.06 2.975 0.93
PH–AN 2.074 2.019 0.215 0.904 0.183 19.69 3.647 0.91
data set for each binary system gives the activity coefficient of
one compound as a function of spreading pressure (w) and mole
fraction (xi). The adjustable parameters of Eqs. (27) and (28) for
all three binary systems were calculated by regression analysis A x1
c2 ¼  B c1 ð29Þ
using well known Newton’s iteration method. These values along x2 x2
with statistical indices are listed in Table 4. The activity coefficient
of other compound was calculated using the following analytical where A ¼ ðC T =C 02 Þ and B ¼ ðC 01 =C 02 Þ. It is observed that in Eq. (29),
equation derived from Eq. (1) with the assumptions that values of x1 and x2 are almost equal for each test at constant ratio
P
C T ¼ 2i¼1 C i and x1 + x2 = 1 [26]: and it is true for all three systems as shown in Fig. 6. It implies that
458 A.J. Jadhav, V.C. Srivastava / Chemical Engineering Journal 229 (2013) 450–459

due to different affinity of the adsorbent towards various adsor-

bates and due to adsorbate–adsorbate interaction itself. In the
present study, GAC showed heterogeneous nature towards the
adsorbates as shown by better fitting of Freundlich isotherm as
compared to Langmuir isotherm for all the three adsorbates. The
adsorption capacity of the GAC was in the order NI > AN > PH.
The reasons for this behavior may be because of various properties
such as molecular size, dipole moment, and hydrophobicity, as dis-
cussed in earlier sections. The interaction parameter, K12 and K21,
in Wilson equation represent the effect of interaction of one adsor-
bate with respect to other. For NI–AN system, value of KAN,NI
(0.288) is greater than KNI,AN (0.212) indicating that the NI has
more affect on AN adsorption than that AN has on NI adsorption.
Analyzing the values of interaction parameters for NI–PH and
PH–AN systems in the similar way (Table 4), the trend of the effect
of one adsorbate on the other was found to be in order of
NI > AN > PH. This is same as the adsorption capacity of GAC vari-
ous adsorbates.

5. Conclusion

Freundlich and Redlich–Peterson isotherms showed very good

fits with the single-component experimental adsorption equilib-
rium data for NI, AN and PH. Therefore, Freundlich model was in-
cluded in the IAST and RAST model formulation. Because of non-
ideal behavior of all three binary systems, the adsorption data can-
not be predicted using the IAST model. The RAST model can cor-
rectly predict adsorption data for all three binary systems
because it takes into account the non-ideal behavior by introduc-
ing the activity coefficient term. The activity coefficient can be cal-
culated using simple Wilson activity coefficient model and
analytical equation, the parameters of these equations can be cal-
culated using regression on experimental activity coefficient data.
Based on statistical indices values, the simultaneous adsorption
phenomena of NI, AN and PH molecules onto GAC can be satisfac-
torily represented by the RAST model than IAST.
The affinity of GAC for NI molecule was greater than that for AN
and PH under the same experimental conditions. The interactive
effect of initial concentration of one species onto the other was
studied by varying the initial concentration ratio of the two com-
pounds. It was observed from experimental data that as initial con-
centration of one compound increases, the adsorption intensity of
Fig. 6. Experimental c2 as a function of experimental c1 for all tests at different
ratio. (A) NI–AN system, (B) NI–PH system, and (B) PH–AN system.
other compound goes on decreasing and it is true for all three bin-
ary systems. It was also found that when initial concentration of
compounds is high, then system shows non-ideal behavior. This
is also true for all the three binary systems. For AN and PH, the
Table 5 activity coefficient values in each system were always lower than
Statistical indices and parametric values of empirical Eq. (29) for three binary
systems.
unity, showing that experimental adsorption capacity was greater
than the corresponding value predicted by IAST model. It was also
System A B SSE RMSE MPSD ARE R2 observed in case of NI (in both NI-AN and NI-PH systems) that, the
NI–AN 0.491 0.005 0.151 0.075 6.943 1.286 0.98 ratios for which activity coefficient value is greater than unity,
NI–PH 0.916 0.701 0.134 0.071 6.644 1.231 0.78 experimental adsorption capacity is lower than the corresponding
PH–AN 0.948 1.049 0.133 0.071 13.231 2.451 0.88
value predicted by IAST model.

Eq. (29) shows the linear relationship between the activity coeffi-
cients for each constant ratio for all three systems. The linear
regression analysis on Eq. (29) for the entire experimental data
set gives the values of parameter A and B, these values along with
statistical indices are reported in Table 5. Now, using Eq. (29) with
known parameters, all the equations of the RAST models can be
solved to represent the experimental data. Overall, RAST model bet-
ter represented the experimental data as compared to IAST model
as shown in Fig. 4.
It is well known that the non-ideality of adsorption phenome-
non may be due to the heterogeneous nature of the adsorbent,
References
[1] A.A. Gurtena, S. Ucana, M.A. Ozlerb, A. Ayar, Removal of aniline from aqueous
solution by PVC–CDAE ligand-exchanger, Journal of Hazardous Materials 120
(2005) 81–87.
[2] S. Kumar, M. Zafar, J.K. Prajapati, S. Kumar, S. Kannepalli, Modeling studies on
simultaneous adsorption of phenol and resorcinol onto granular activated
carbon from simulated aqueous solution, Journal of Hazardous Materials 185
(2011) 287–294.
[3] W. Wei, R. Sun, J. Cui, Z. Wei, Removal of nitrobenzene from aqueous solution
by adsorption on nanocrystalline hydroxyapatite, Desalination 263 (2010) 89–
96.
[4] S. Rengaraj, S.H. Moon, R. Sivabalan, B. Arabindoo, V. Murugesan, Removal of
phenol from aqueous solution and resin manufacturing industry wastewater
 A.J. Jadhav, V.C. Srivastava / Chemical Engineering Journal 229 (2013) 450–459
using an agricultural waste: rubber seed coat, Journal of Hazardous Materials
89 (2002) 185–196.
[5] F. An, X. Feng, B. Gao, Adsorption property and mechanism of composite
adsorbent PMAA/SiO2 for aniline, Journal of Hazardous Materials 178 (2010)
499–504.
[6] J. Pan, B. Guan, Adsorption of nitrobenzene from aqueous solution on activated
sludge modified by cetyltrimethylammonium bromide, Journal of Hazardous
Materials 183 (2010) 341–346.
[7] Q. Wen, Z. Chen, J. Lian, Y. Feng, N. Ren, Removal of nitrobenzene from aqueous
solution by a novel lipoid adsorption material (LAM), Journal of Hazardous
Materials 209–210 (2012) 226–232.
[8] K. Lin, J. Pan, Y. Chen, R. Chenga, X. Xu, Study the adsorption of phenol from
aqueous solution on hydroxyapatite nanopowders, Journal of Hazardous
Materials 161 (2009) 231–240.
[9] S. Lin, R. Juang, Adsorption of phenol and its derivatives from water using
synthetic resins and low-cost natural adsorbents: a review, Journal of
Environmental Management 90 (2009) 1336–1349.
[10] O.V. Makarova, T. Rajh, M.C. Thurnauer, A. Martin, P.A. Kemme, D. Cropek,
Surface modification of TiO2 nanoparticles for photochemical reduction of
nitrobenzene, Environmental Science and Technology 34 (22) (2000) 4797–
4803.
[11] P.C.C. Faria, J.J.M. Orfao, M.F.R. Pereira, Ozonation of aniline promoted by
activated carbon, Chemosphere 67 (2007) 809–815.
[12] A. Latifoglu, M.D. Gurol, The effect of humic acids on nitrobenzene oxidation
by ozonation and O3/UV processes, Water Resources 37 (8) (2003) 1879–1889.
[13] L. Wang, S. Barrington, J. Kim, Biodegradation of pentyl amine and aniline from
petrochemical wastewater, Journal of Environmental Management 83 (2007)
191–197.
[14] X. Yang, Q. Guan, W. Li, Effect of template in MCM-41 on the adsorption of
aniline from aqueous solution, Journal of Environmental Management 92
(2011) 2939–2943.
[15] O.J. Hao, H. Kim, P.C. Chiang, Decolorization of wastewater: critical review,
Environmental Science and Technology 30 (2000) 449–505.
[16] A. Erto, A. Lancia, D. Musmarra, D, A modelling analysis of PCE/TCE mixture
adsorption based on ideal adsorbed solution theory, separation and
purification technology 80 (2011) 140–147.
[17] J.C. Bellot, J.S. Condoret, Modelling of liquid chromatography equilibria,
Process Biochemistry 28 (1993) 365–376.
[18] R.T. Yang, Gas Separation by Adsorption Processes, Butterworths, Boston, MA,
1987.
[19] W. Fritz, E.U. Schluender, Simultaneous adsorption equilibria of organic
solutes in dilute aqueous solutions on activated carbon, Chemical
Engineering Science 29 (1974) 1279–1282.
[20] C. Sheindorf, M. Rebhum, M. Sheintuch, A Freundlich-type multicomponent
isotherm, Journal of Colloid Interface Science 79 (1981) 136–142.
[21] A.L. Myers, J.M. Prausnitz, Thermodynamics of mixed-gas adsorption,
American Institute of Chemical Engineers 11 (1) (1965) 121–127.
[22] S.J. Radke, J.M. Prausnitz, Thermodynamics of multi-solute adsorption from
dilute liquid solutions, American Institute of Chemical Engineers 18 (4) (1972)
761–768.
459
[23] G. Mckay, K.K.H. Choy, J.F. Porter, Single and multicomponent equilibrium
studies for the adsorption of acidic dyes on carbon from effluents, Langmuir 20
(2004) 9646–9656.
[24] S.K. Papageorgiou, F.K. Katsaros, E.P. Kouvelos, N.K. Kanellopoulos, Prediction
of binary adsorption isotherms of Cu2+, Cd2+ and Pb2+ on calcium alginate
beads from single adsorption data, Journal of Hazardous Materials 162 (2009)
1347–1354.
[25] D.D. Do, Adsorption Analysis: Equilibria and Kinetics, Imperial College Press,
London, 1998. pp. 18–21.
[26] A. Erto, A. Lancia, D. Musmarra, A real adsorbed solution theory model for
competitive multicomponent liquid adsorption onto granular activated
carbon, Microporous and Mesoporous Materials 154 (2012) 45–50.
[27] IS 1350 Part I, Methods of Test for Coal and Coke, Proximate Analysis, Bureau
of Indian Standards, Manak Bhawan, New Delhi, India, 1984.
[28] L. Sratthaphut, N. Ruangwises, Genetic algorithms-based approach for
wavelength selection in spectrophotometric determination of vitamin B12 in
pharmaceutical tablets by partial least-squares, Procedia Engineering 32
(2012) 225–231.
[29] H. Khajehsharifi, E. Pourbasheer, Genetic-algorithm-based wavelength
selection in multicomponent spectrophotometric determination by PLS:
application on ascorbic acid and uric acid mixture, Journal of the Chinese
Chemical Society 55 (2008) 163–170.
[30] J. Ghasemi, A. Niazi, R. Leardi, Genetic-algorithm-based wavelength selection
in multicomponent spectrophotometric determination by PLS: application on
copper and zinc mixture, Talanta 59 (2003) 311–317.
[31] M. Acros, Julia, M. Ortiz Cruz, Belen Villahoz, A. Sarabia Luis, Genetic-
algorithm-based wavelength selection in multicomponent spectrometric
determination by PLS: application on indomethacin and acemethacin
mixture, Analytica Chimical Acta 339 (1997) 63–77.
[32] S.R. Kamath, A. Proctor, Silica gel from rice hull ash: preparation and
characterization, Cereal Chemistry Journal 75 (4) (1998) 484–487.
[33] M.M. Abou-Mesalam, Sorption kinetics of copper, zinc, cadmium and nickel
ions on synthesized silico-antimonate ion exchanger, Colloids and Surfaces A:
Physicochemical Engineering Aspects 225 (2003) 85–94.
[34] S. Ricordel, S. Taha, I. Cisse, G. Dorange, Heavy metals removal by adsorption
onto peanut husks carbon: characterization, kinetic study and modeling,
Separation and Purification Technology 24 (2001) 389–401.
[35] Y.S. Ho, C.C. Chiang, Y.C. Hsu, Sorption kinetics for dye removal from aqueous
solution using activated clay, Separation Science and Technology 36 (11)
(2001) 2473–2488.
[36] J. Reungoat, J.S. Pic, M.H. Manero, H. Debellefontaine, Adsorption of
nitrobenzene from water onto high silica zeolites and regeneration by
ozone, Separation Science and Technology 42 (7) (2002) 1447–1463.
[37] C.R. Maechling, S.J. Clemett, F. Engelke, R.N. Zare, Evidence for thermalization
of surface-desorbed molecules at heating rates of 108 K/s, Journal of Chemical
Physics 104 (21) (1996) 8768–8776.
[38] A.H. Norzilah, A. Fakhru-Razi, T.S.Y. Choong, A.L. Chuah, Surface modification
effects on CNTs adsorption of methylene blue and phenol, Journal of
Nanomaterials 55 (2011) 1–18.