Desalination 186 (2005) 57–64

Adsorption of basic dye onto palm kernel shell activated carbon:

sorption equilibrium and kinetics studies
A. Jumasiaha, T.G. Chuaha*, J. Gimbonb, T.S.Y. Choonga, I. Aznia
a
Department of Chemical and Environmental Engineering, Faculty of Engineering, Universiti Putra Malaysia,
43300 UPM Serdang, Selangor DE, Malaysia
Tel. +60 (3) 8946 6288; Fax: +60 (3) 8656 7120: email: chuah@eng.upm.edu.my
b
Faculty of Chemical and Natural Resources Engineering, KUKTEM, Bandar MEC 25200 Kuantan, Pahang, Malaysia

Received 1 November 2004; accepted 4 May 2005

Abstract
Adsorbents prepared from palm kernel shell, an agricultural waste product, were used to remove a dye, Basic
Blue 9, from an aqueous solution in batch mode at a constant temperature of 28EC. The sorption kinetics and
equilibrium of basic dye onto palm kernel shell activated carbon (PKSAC) were studied. The isotherm data were well
described by the Redlich–Peterson isotherm model, with constants obtained from non-linear regression. The sorption
kinetics are well described by the pseudo-second-order kinetic model. These studies suggested that PKSAC could be
used as low-cost alternatives in wastewater treatment for dye removal.

Keywords: Palm kernel shell; Activated carbon; Basic Blue 9; Isotherm; Kinetics; Adsorption

1. Introduction
Among the different pollutants of an aquatic
ecosystem, dyes are a large and important group
of industrial chemicals with over 700,000 tons of
waste produced annually [1]. Considering both
volumes discharged and effluent composition,
wastewater from the textile industry was declared
as one of the major sources of wastewater in
ASEAN countries [2]. Various techniques have
been employed for the removal of dyes from
wastewater (e.g., [1,2]). Liquid phase adsorption
*Corresponding author.
using activated carbon is one of the most effec-
tive treatment methods for removing a wide vari-
ety of dyes present in the wastewater. Although
activated carbon is highly effective, its running
costs are also high [3]. Therefore, locally avail-
able and abundant material should be found as a
source for cost-effective activated carbon.
As a leading palm oil producer, Malaysia has
an abundant supply of palm kernel shell. In 2002
alone, the amount of palm shell generated was
approximately 4.3 million tons, and this creates a
huge disposal problem [4]. Realising the scale of
this waste, several studies were initiated to utilise

0011-9164/06/$– See front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.desal.2005.05.015
58 A. Jumasiah et al. / Desalination 186 (2005) 57–64

palm shell as the raw material for activated Table 1

carbon production. The possible use of palm shell Product specification of palm kernel shell activated
as an activated carbon precursor was studied by carbon
several researchers [5–7], and they concluded that
a good quality product can be obtained. The Parameter Typical ASTM test
value standard
application of palm kernel shell activated carbon
(PKSAC) as carbon molecular sieves in N2 gas Iodine no., mg/g 950–1150 D4607-94
purification has been reported by Jaan and Ani Butane activity, % 20–30 D5742-95
[8], as an adsorbent in the bleaching process in CCl4 activity, % 55–70 D3467-94
palm oil refining by Mak [9], and as an adsorbent Ball pan hardness, % 88–96 D3802-79/94
in copper (II) removal by Issabayeva et al. [10]. Ash content, % 5 max. D2866-94
The purpose of this work was to obtain the Apparent density, g/cm3 0.42–0.52 D2854-96
Moisture, % 5 max. D2867-95
equilibrium relationship and a model that de-
pH 9–11 D3838-80-94
scribed the contact time of the adsorption of the Particle size 90 D2862-92
dye Basic Blue 9 (BB9) on PKSAC as adsorbent. distribution, %
These two pieces of information are important for BET surface area, m2/g 1088
the design, operation and optimisation of an
adsorption system.
The PKSAC was donated by K.D. Technology
Sdn. Bhd., in a granular form (<0.15 mm). The
2. Experimental PKSAC physical properties are listed in Table 1.

2.1. Materials
BB9 was purchased from Aldrich Chemical,
and it was used as received without further
purification. It has a maximum wavelength of
651 nm. Stock solutions were prepared by dis-
solving accurately weighed 1.000 g (+0.0005 g)
of dye in 1 dm3 of distilled water. The stock
solutions were then properly wrapped with alumi-
nium foil and stored in a dark place to prevent
direct sunlight, which may cause decolourisation.
The molecular structure of BB9 is shown in
Fig. 1.
Fig. 1. The molecular structure of Basic Blue 9
(C16H18N3ClS.3H2O).
2.2. Procedure
Equilibrium data were obtained by adding
0.1 g of activated carbon into 100 mL dye solu-
tion in 250 mL conical flasks (Pyrex, England).
Dye solutions of seven different initial concen-
trations were used (100, 200, 300, 400, 500, 600,
and 700 mg/L). The conical flasks were covered
properly with aluminium foil to avoid evapora-
tion of the dye solution. A shaking process was
carried out for 72 h in an incubator shaker
(Protech, Malaysia) operating at 150 rpm and
28EC (±2EC). Contact time studies confirmed that
equilibrium was reached after 72 h. The solutions
were then centrifuged (IEC Centra CL2, USA) at
2500 rpm for 5 min, and the clarified supernatant
solutions were carefully decanted to be analysed
using a UV/VIS double-beam spectrophotometer
(UNIKON 933, Italy) at a maximum wavelength
of 651 nm. Dye solutions were diluted if the
reading of spectrophotometer exceeded 0.7. The
final concentration of the solution was then
 A. Jumasiah et al. / Desalination 186 (2005) 57–64
determined from the calibration curve. All experi-
ments were carried in duplicate, and the average
values were reported. The adsorption of BB9 was
found to be pH dependent, with maximum
adsorption at pH 7.0.
The dye concentration at equilibrium, qe
(mg/g), can be calculated from
(1)
where C0 (mg.dm!3) is the initial dye concen-
tration in liquid phase, Ce (mg.dm!3) is the dye
concentration in liquid phase at equilibrium, V
(dm3) is the total volume of dye solution and m
(g) is the mass of sorbent.
2.3. Batch contact time study
Batch kinetic studies were carried out to
examine the dynamic behaviour of PKSAC as an
adsorbent on the removal of BB9. The experi-
ments were carried out at four different initial
concentrations ranging from 20 to 150 mg/L at a
predetermined adsorbent mass. A series of experi-
ments were performed by shaking the adsorbent
with 1000 mL dye solution at a constant tem-
perature of 28EC at pH 7.0. A sample of 1 mL
was carefully withdrawn using a digital micro-
pipette (ThermoLabsystems, Finland) every
3 min for the first 30 min and every 5 min for the
next 30 min. For the next 60 min, samples were
taken every 10 min and eventually every 15 min
for the last 60 min. All samples were diluted with
distilled water and analysed immediately.
3. Results and discussion
3.1. Isotherms and model fitting
The equilibrium isotherms of BB9–PKSAC
were analysed using three models, namely Lang-
muir, Freundlich and Rendlich–Patterson. An
accurate isotherm is important for design pur-
59
poses. Linear regression is commonly used to
determine the best fitted model, and the method
of least squares has been widely used for obtain-
ing the isotherm constants.
Langmuir, the best known and the most often
used isotherm, can be represented by the
expression:
(2)
where KL (dm3/g) and aL (dm3/mg) are Langmuir
isotherm constants. The Langmuir isotherm
implies formation of monolayer coverage of
adsorbate on the surface of the adsorbent. The
ratio of KL/aL gives the monolayer saturation
capacity. Eq. (2) can be rearranged into linear
form:
(3)
By plotting [(Ce)/(qe)] vs. Ce, the Langmuir
constants KL and αL can be obtained from the
slope and intercept of the plot, respectively.
The Freundlich equation is an empirical equa-
tion employed to describe heterogeneous systems:
(4)
where KF is the Freundlich constant (dm3/g) and
1/n is the heterogeneity factor. The Freundlich
isotherm describes reversible adsorption and is
not restricted to the formation of the monolayer.
The linear form of Eq. (4) can be obtained by
taking logarithms of both sides of it:
(5)
A plot log qe vs. log Ce yields a straight line,
which permits the determination of 1/n and KF
from the slope and intercept.
60 A. Jumasiah et al. / Desalination 186 (2005) 57–64

The Redlich–Peterson (R–P) isotherm is an
empirical isotherm incorporating three para-
meters. It combines elements from both the
Langmuir and Freundlich equations, and the
mechanism of adsorption is a hybrid and does not
follow ideal monolayer adsorption:
which has the advantage that all isotherms were
compared on the same abscissa and ordinate,
could be used to determine the best fitted model
for isotherm. The Chi-square value is given as:
(9)

(6)
The sorption behaviour of BB9 onto PKSAC
was analysed using both linear and non-linear
where KR and aR are isotherm constants (dm3/mg) regression, and the results are summarised in
and β is an exponent which lies between 0 and 1. Table 2. For analysis of data using linear regres-
Eq. (6) can be linearised to give sion, the fitness of the model is often evaluated
based on the value of the correlation coefficient
(R2). Both Langmuir and R–P isotherms give very
(7) high R2 values. Despite this, the χ2 values for the
linear form of the Langmuir isotherm are very
large, suggesting that the isotherm constants of
By plotting vs. ln Ce, a straight
Table 2
line will be obtained with slope β and intercept
Isotherm constants for BB 9 sorption onto PKSAC
ln bj. The value of Kj is evaluated using a trial-
and-error method until a minimum deviation is Linear form Non-linear form
obtained, with the initial value of Kj estimated
from the value of KL. Langmuir
It is, however, well known that obtaining the KL /aL (mg g!1) 333.33 311.72
parameter of a non-linear equation using its linear aL (dm3 mg!1) 0.2273 24.417
form may introduce large errors. The preferred R2 0.9979 —
method of analysis is to perform non-linear SSE 3701.047
regression on the untransformed data. The non- χ2 14,945.47 19.145
linear regression could now be performed rather Freundlich
easily with the solver add-in function of Micro- KF (mg g!1) 194.94 205.0913
soft Excel. The isotherm parameters are obtained 1/n 0.0955 0.083919
by minimizing the sum of the errors square: R2 0.8011 —
SSE 5278.243
χ2 275.4571 31.89192
(8) Redlich–Peterson
KR /aR (mg g!1) 251.16 268.3872
The non-linear regression method requires initial aR 76.3781 37.64503
guesses to start the iterations. At least three dif- β 0.9543 0.968867
ferent initial guesses were performed to have R2 0.9996 —
confidence in the solution obtained. Ho [11] has SSE 3040.29
recently suggested that the Chi-square analysis, χ2 24.18827 18.91915
 A. Jumasiah et al. / Desalination 186 (2005) 57–64
the Langmuir isotherm obtained from linear
regression do not describe well the equilibrium
data. The isotherm constants are also obtained
using a non-linear regression by minimizing the
SSE.
The χ2 values of all the three isotherms are
low, implying that the experimental data could be
reasonably well represented by all three equa-
tions. The R–P isotherm has the lowest χ2 values,
suggesting that the R–P isotherm provides the
best fit to the experimental data. For the R–P and
Freundlich isotherms, the isotherm constants ob-
tained from both linear and non-linear analysis
are quite close. However, the isotherm constants
of the Langmuir equations obtained from linear
and non-linear analysis differ substantially. Based
on the χ2 values, the isotherm constants obtained
from non-linear regression provide a better fit to
the experimental data than the isotherm constants
from linear regression.
The maximum sorption capacity calculated
using constants from both linear and non-linear
regression are similar. However, the Langmuir
isotherm using constants from linear regression
poorly predicts the equilibrium data at low
concentrations. Therefore, it is suggested that the
equilibrium data be evaluated by both linear and
non-linear regression analysis to ensure correct
results. The isotherm constants obtained from
linear regression could be used as initial guesses
for the non-linear regression. Fig. 2 shows the
isotherm plots and the experimental data, with
isotherm constants obtained from non-linear
regression.
3.2. Sorption kinetics
The relationship between contact time and dye
uptake by PKSAC were studied through batch
kinetic experiments that were carried out at four
different initial concentrations ranging from 20 to
150 mg/L at predetermined adsorbent mass.
Fig. 3 shows that at a lower initial dye concen-
tration, i.e., 20 and 50 mg/L, the equilibrium was
61
Fig. 2. Equilibrium curves for sorption of BB9 onto
PKSAC.
reached within less than 40 min. As initial dye
concentration increased, a longer time was
needed to reach equilibrium and the adsorption
capacity was also found to be increased. In order
to determine the sorption kinetics of dyes, the
first-order and second-order kinetics models were
examined. The first-order rate expression of
Lagergren [11] based on solid capacity is
expressed as follows:
(10)
where qt is the amount of dyes adsorbed onto the
PKSAC (mg.g!1) at time t and k1 is the rate
constant of the first order (min!1). Integrating
Eq. (10) for the initial and end conditions t = 0 to
t = t and qt = 0 to qt = qt, and after some
rearrangement, a linear plot is obtained:
(11)
The rate constant for the first-order equation
was determined from the slope of the plot
log(qe!qt) vs. time. If first-order kinetics are
applicable to the system under study, the plot of
log(qe!qt) vs. time of Eq. (11) should give a
linear relationship. Further, the qe obtained from
the plot should also be close to the qe obtained
from experiment. Fig. 3 shows that after a short
62 A. Jumasiah et al. / Desalination 186 (2005) 57–64

Fig. 3. Linearised Lagergren first-order kinetics for BB9 sorption onto PKSAC.

period, the experimental data deviated consider- Table 3

ably from linearity. Even though the correlation Kinetic constants for BB9 sorption onto PKSAC
coefficient (R2) of the first-order equation is
Initial First-order Second-order
considerably high, the values of calculated qe
conc., ppm kinetic constants kinetic constants
from the first-order kinetics plot were far too
small compared to the experimental values. The 20 k1 = 0.0359 min!1 k2 = 9.225×10!3 g.
inapplicability of the Lagergren equation to mg!1.min!1
qe = 7.63 mg.g!1 qe = 20.66 mg.g!1
describe the kinetics of BB9 sorption was also R2 = 0.9301 R2 = 0.9986
observed by Annadurai et al. [12] for adsorption 50 k1 = 0.0393 min!1 k2 = 2.022×10-3 g.
of dyes using banana and orange peels. mg!1.min!1
The experimental data were also analysed qe = 31.06 mg.g!1 qe = 52.91 mg.g!1
using the pseudo-second-order model [13]: R2 = 0.9736 R2 = 0.9953

100 k1 = 0.0407 min!1 k2 = 5.68×10-4 g.

(12)
where k2 is the pseudo-second-order rate constant
of adsorption (g.mg!1 min!1). Integrating Eq. (12)
for the initial and end conditions t = 0 to t = t and
qt = 0 to qt = qt, and after some rearrangement, a
linear plot is obtained:
(13)
mg!1.min!1
qe = 90.32 mg.g!1 qe = 112.36 mg.g!1
R2 = 0.9906 R2 = 0.9899
If pseudo-second-order kinetics are applicable
to the system under study, the plot of t/qt vs. time
of Eq. (13) should give a linear relationship. The
second-order sorption rate constant and qe were
determined from the slopes and intercepts of the
plots t/qt vs. time for different adsorbent initial
 A. Jumasiah et al. / Desalination 186 (2005) 57–64
Fig. 4. Linearised pseudo-second-order kinetics for BB9 sorption onto PKSAC.
concentrations (Fig. 4). The correlation coeffi-
cients for the linear plots are very high (>0.98)
for all systems. The calculated qe values from the
pseudo-second-order model also are in good
agreement with the experimental qe values. This
suggests that the sorption system followed the
pseudo-second-order model. The values of kinetic
constants and qe values of BB9 sorption onto
PKSAC are summarised in Table 3.
4. Conclusions
The results of the present investigation show
that activated carbon prepared from palm kernel
shell is a potential useful adsorbent for the
adsorption of a basic dye, BB9. Equilibrium data
are well represented by the Redlich–Peterson
isotherm followed by the Langmuir and Freund-
lich models. The Langmuir monolayer saturation
sorption capacities are 311.72 mg/g. The chi-
square analysis provides better determination for
the best fitting isotherm. The adsorption kinetics
follows the pseudo-second-order kinetic model.
63
Acknowledgement
The authors are grateful to KD Technology
Sdn. Bhd., Malaysia, and the Malaysian Palm Oil
Board (MPOB) for support and assistance.
References
[1] A.L. Ahmad, W.A. Harris, M. Syafie and O.B. Seng,
J. Teknologi, 36 (2002) 31–44.
[2] S. Setamanit, P. Fongsatikul and J. Sivadechathep,
ASEAN Achievement and Future Directions in
Pollution Control. Ministry of Science, Technology
and Environment, Thailand, 2002, p. 33.
[3] T. Robinson, B. Chandran, G.S. Naidu and P. Nigam,
Bioresource Technol., 85 (2002) 43–49.
[4] A.N. Ma, Palm Oil Eng. Bull., 8 (2002) 24–26.
[5] M.Z. Hussien, R.S.H. Tarmizi, Z. Zainal, R. Ibrahim
and M. Bardi, Carbon, 34 (1996) 1447–1454.
[6] W.M.A. Wan Daud, W.A. Wan Shabuddin and
S. Mohd Zaki, Carbon, 38 (2000) 1925–1932.
[7] J. Guo and A.C. Lua, Mat. Chem. Physics, 80 (2003)
114–119.
[8] S.T. Jaan and F.N. Ani, Sep. Purif. Technol., 35
(2004) 47–54.
64 A. Jumasiah et al. / Desalination 186 (2005) 57–64

[9] S.M. Mak, The characterization of steam-activated
palm kernel shell activated carbons and the role of
porosity in adsorption and catalytic applications.
Master Thesis, Universiti Putra Malaysia, 2003.
[10] G. Issabayeva, M.K. Aruoa and M.N. Sulaiman,
Study on adsorption of Cu (II) on palm shell acti-
vated carbon, Proc. International Conference on
Environmental Management and Technology,
Putrajaya, Malaysia, 2003.
[11] S. Lagergren and K. Svenska Vetenskapsaka-
demiens, Vol. 24, Handlingar, 1898, pp. 1–39.
[12] G. Annadurai, R.S. Juang and D.J Lee, J. Haz. Mat.,
B92 (2002) 263–274.
[13] Y.S. Ho and G. McKay, Process Biochem., 34
(1999) 451–465.