PII: S0043-1354(01)00127-0

Wat. Res. Vol. 35, No. 17, pp. 41594165, 2001

# 2001 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
0043-1354/01/$ - see front matter

DESORPTION OF DYE FROM ACTIVATED CARBON BEDS:

EFFECTS OF TEMPERATURE, pH, AND ALCOHOL
JIA-MING CHERN* and CHIA-YUAN WU
Department of Chemical Engineering, Tatung University, 40 Chungshan North Road, 3rd Sec.,
Taipei, 10451, Taiwan
(First received 16 June 2000; accepted in revised form 27 February 2001)
Abstract}The adsorption isotherms of yellow and red dye solutions onto granular activated carbon at
varying solution pHs (28), temperatures (15508C), and alcohol concentrations (020%) were
experimentally determined by batch tests and the Toth model was found to best t the adsorption
isotherm data for varying solution pHs, temperatures, and alcohol concentrations. The maximum
adsorption capacity was found to decrease with increasing solution pH and alcohol concentration and
could be predicted by the correlation equations obtained in this study. A correlation equation was also
obtained to account for the eects of solution temperature on the adsorption equilibrium constant. The
258C water was found to be a very poor regenerant for the carbon bed presaturated with the yellow dye
compared with 20% alcohol solution. A simple equation was derived, based on non-linear wave
propagation theory, to predict the desorption curves of activated carbon bed. Given presaturation
concentration, bed density and void fraction, and adsorption isotherm, the wave propagation theory
predicted the desorption curves quite satisfactorily. # 2001 Elsevier Science Ltd. All rights reserved
Key words}dye wastewater, activated carbon, adsorption isotherm, desorption curve, wave theory

NOMENCLATURE

a
b
c
C
Calc
DG
DH
DS
q
R
t
T
uo
uc
V
Vb
z
e
r

INTRODUCTION

correlation parameter dened in equations (8)

and (14)
correlation parameter dened in equations (8)
and (14)
correlation parameter dened in equations (8)
and (14)
dye concentration in mobile phase, kg/m3
alcohol concentration, %
Gibbs free energy change of adsorption, kJ/mol
Heat of adsorption, kJ/mol
Entropy change of adsorption, kJ/mol K
dye concentration in stationary phase, kg/kg dry
carbon
gas constant, atm-m3/mol K
time, h
solution temperature, K
linear velocity of carrier uid, m/h
concentration wave velocity, m/h
cumulative euent volume, m3
bed volume, m3
distance from the inlet of mobile phase, m
void fraction of bed, dimensionless
carbon bed density, kg/m3

*Author to whom all correspondence should be addressed.

Tel.: +886-2-25925252 ext 3489; fax: +886-2-25861939;
e-mail: jmchern@che.ttu.edu.tw

The discharge of textile wastewaters with high color,

high suspended solids and dissolved organics represents one of the major industrial pollutions in
Taiwan. In response to the increasingly stringent
discharge standards and discharge fee, textile industries in Taiwan must upgrade the performance of
their wastewater treatment facilities with primary
eort on removing contained chemical oxygen
demand (COD) and color. Due to its variation in
pH, temperature, ow rate and contained pollutants,
the textile wastewater usually requires physicochemical as well as biological treatments (An et al., 1996;
Liakou et al., 1997; Sarasa et al., 1998) to meet the
more stringent euent standards. An eective
method to remove color is to use an adsorbent such
as activated carbon to adsorb the dye molecules and
then remove the color from the wastewater.
Although many dierent adsorbents (Nassar et al.,
1995; Namasivayam et al., 1996; Ramakrishna and
Viraraghavan, 1997; Ho and McKay, 1998) were
tried to remove dyes from wastewaters, activated
carbon (Walker and Weatherley, 1997, 1998) was still
the most widely used adsorbent to this end. The
direct addition of powdered activated carbon to
remove dyes is quite ecient, but the carbon loss as
wasted sludge is rather signicant and causes another
problems. Therefore, granular activated carbon is

4159

4160

Jia-Ming Chern and Chia-Yuan Wu

usually used in xed-bed adsorption processes

because of ease of operation and no carbon loss
problem.
In order to properly design and operate x-bed
adsorption processes, the x-bed dynamics, that is
the concentration proles and histories must be
known. Although various mass-transfer models
(Chatzopoulos and Varma, 1995; Wolborska and
Pustelnik, 1996) were used to predict the xed-bed
dynamics, a simple yet powerful tool that can be used
to predict the dynamics without tedious calculations
is the non-linear wave propagation theory (Helerich
and Klein, 1970). We have successfully applied the
wave propagation theory to predict the dynamics of
carbon beds for color removal from synthetic dye
wastewater (Chern and Huang, 1998, 1999; Chern
and Wu, 1999). In the prediction of the xed-bed
dynamics using either the mass-transfer models or
the wave theory, the adsorption isotherm plays a very
important role. For a favorable adsorption isotherm,
a suitable equation to t the adsorption data in the
concentration range of concern is crucial to predict
the desorption curve with the wave theory. This
paper continues to explore the adsorption isotherms
and the eects of adsorption isotherm on desorption
curves of dye/carbon systems at varying solution
temperatures and pHs in the absence and presence of
alcohol.
DESORPTION DYNAMICS

The theoretical basis of this study is directly built

upon the non-linear wave propagation theory developed by Helerich and Klein (1970). The desorption
dynamics of a single-component adsorption system
can be described by the following governing equations:
@q
@C
@C
uo e
0
r e
@t
@t
@z

q f C

where C is the concentration of the dye in the mobile

phase, q the concentration of the dye in the
stationary phase, r the carbon bed density, e
the void fraction of the bed, uo the linear velocity
of the carrier uid, t the operating time, and z the
distance from the inlet of the mobile phase. Equation
(1) is basically the unsteady-state mass balance of the
adsorbate while equation (2) is the so-called adsorption isotherm. The assumptions associated with
equation (1) are:
*
*
*
*
*
*

The adsorption reaches local equilibrium.

No chemical reactions occur in the column.
Only mass transfer by convection is signicant.
Radial and axial dispersions are negligible.
The ow pattern is ideal plug ow.
The temperature in the column is uniform and
invariant with time.

The ow rate is constant and invariant with the

column position.

The initial and boundary conditions associated with

equation (1) are
At t 0;

C CF

For t > 0;

for 0  z  L

at z 0;

C0

3
4

The initial condition states that the carbon bed is

saturated with the feed dye concentration in the
adsorption cycle while the boundary condition states
that the presaturated bed is regenerated with dye-free
solvent. Instead of directly solving the governing
equations, let us dene the concentration wave
velocity as


@z
5
uC 
@t C
to describe the column dynamics. Equation (5)
denes the velocity at which a given constant
concentration C travels through the carbon bed.
Combining equations (1) and (5) and using the chain
rule of calculus leads to
uC

uo
1 r=edq=dC

Equation (6) is used to calculate the concentration

wave velocity of a non-sharpening wave such as a
desorption wave for favorable-type isotherm. The
distancetime diagram can thus be plotted with a
spectrum of straight lines originated from the (0, 0)
point and the slopes calculated by equation (6), and
the corresponding desorption curve can be easily
obtained. The desorption curve in terms of the
cumulative euent volume can be calculated by the
following equations:
8
>
>
>
0  V  V1 ;
>
>
>
>
>
>
<
V1  V  V2 ;
>
>
>
>
>
>
>
>
>
: V  V2 ;

"

 #
r dq
V1 Vb e 1
;
C CF
e dC CF



r dq
;
0  C  CF
V Vb e 1
e dC

 

r dq
V2 Vb e 1
;
C0
e dC C0
7

where Vb is the bed volume. The dierentials on the

right-hand side of equation (7) must be evaluated at
concentrations of concern to predict the desorption
curve.
EXPERIMENTAL

The red, yellow, and mustard yellow acid dyes (Sumitomo Chemical Corp., Japan) with main ingredient of
triphenyalmethane were used to prepare the synthetic
wastewater for the tests in this study. The granular activated
carbon (Taipei Chemical Corp., Taiwan) with a size range
between 30 and 32 mesh was used. The activated carbon was

Desorption of dye from activated carbon beds

repeatedly washed with deionized water to remove any
leachable impurities and adherent powder, then dried to
constant weight at 1108C for 24 h. In determining the
adsorption isotherms at varying temperatures, various
amounts of the washed and dried carbon (0.0501.000 g)
were added to 500-ml glass-stoppered asks, each lled with
250 ml of aqueous red or yellow dye solution. All the asks
were shaken in a temperature-controlled shaker (Hotech,
Model 706) at 150 rpm and the desired temperature until
adsorption equilibrium was reached. After ltration, the
solution volumes were measured and the dye concentrations
were calculated by the LambertBeer law with the
absorbency measured by a UV spectrophotometer (Milton
Roy, Spectronic Gene Sys 5). After the dye concentrations
in the solution were determined, the adsorbed amounts of
the dyes were calculated by mass balance. Buer solutions
were used to control solution pH at 2, 4, 6, and 8,
respectively, for the determination of the adsorption
isotherms at 258C. Ethanol alcohol was added to the
aqueous solution to determine the eect of the alcohol
concentration on the adsorption isotherm.
The column tests were carried out in a water-jacketed
glass column with an inside diameter and length of 1.74 and
60 cm, respectively. The activated carbon with known dried
weight was put into the column randomly. Deionized water
has used to wash the carbon in a upow fashion in order to
remove air bubbles and to rinse the carbon. Then, the bed
density and void fraction were determined before a column
test has begun. In an adsorption cycle, aqueous dye solution
was continuously fed to the top of the column until the
column was saturated with the feed dye concentration. The
saturated column was then regenerated with deionized water
or 20% alcohol solution at 3.5 BV/h owrate controlled by a
metering pump (Watson-Marlow, 302S/RL). The column
temperature for all the column tests was maintained at
25  18C. In all the column tests, the euent samples
were intermittently collected by a fraction collector
(ISCO, Retriever 500) and measured by the UV spectrophotometer.

4161

Fig. 1. Experimental and predicted desorption curves of the

yellow dye by water and by 20% alcohol solution at 258C.

RESULTS AND DISCUSSION

The granular activated carbon used in this study

has strong adsorption capability of the yellow and
red acid dyes. In order to make the xed-bed
adsorption process ecologically and economically
feasible, the saturated bed must be regenerated and
restored its adsorption capability. Therefore, the
desorption experiments of the carbon bed presaturated with 365 mg/L yellow dye solution were rst
performed and the experimental desorption curves
were compared with those predicted by equation (7).
Figure 1 shows such a comparison for using
deionized water and 20% alcohol as regenerants,
respectively. As is clearly shown in Fig. 1, 258C water
is a very poor regenerant compared with the alcohol
solution.
For water desorption, the wave theory predicts
that the concentrations of the rst 111 bed volumes
(BV) of the euents equal the feed concentration in
the adsorption cycle, and the euent concentration
gradually decreases after 111 BV as shown in Fig. 1.
For alcohol desorption, the wave theory predicts that
within the rst 6.2 BV the euent concentration is
25.5 times higher than the feed concentration in the
adsorption cycle as shown in Fig. 1. This phenom-

Fig. 2. Experimental and predicted adsorption isotherms of

the yellow dye in the absence and presence of 20% alcohol
at 258C.
Table 1. Freundlich adsorption isotherm parameters for the yellow
dye at 258Ca
q kCn
k
n
R2
a

0% alcohol

20% alcohol

551
0.129
0.992

245
0.172
0.983

q of unit mg/g; C of unit mg/L.

enon can be explained by the dierence of the

adsorption isotherms of the yellow dye with 0 and
20% alcohol.
Figure 2 shows the experimental and predicted
adsorption isotherms of the yellow dye with 0 and
20% alcohol at 258C, respectively. The Freundlich
model parameters associated with Fig. 2 are listed in
Table 1. As is clearly shown in Fig. 2, for the same
liquid-phase dye concentration, the adsorbed-phase

4162

Jia-Ming Chern and Chia-Yuan Wu

concentration in the presence of 20% alcohol is much

lower than that in the absence of alcohol. In the
adsorption cycle, the carbon bed was rst saturated
with the 365 mg/L feed and the resulting saturation
concentration in the adsorbed phase was 1179 mg/g
(=551 3650.129). The introduction of 20% alcohol
as the regenerant dramatically increases the equilibrium concentration in the liquid phase to 9293 mg/L
[=(1179/245)1/0.172]. The wave of this high concentration gradually moves toward the bottom of
the bed and results in desorption curve as shown in
Fig. 1.
In summary, the Freundlich model ts the
adsorption isotherms in the absence and presence of
20% alcohol quite well; the wave theory also predicts
the desorption curves using water and 20% alcohol
quite satisfactorily. Therefore, if we know the
adsorption isotherms of the dye in various liquidphase conditions we can predict the desorption
curves and then determine the optimal regenerant
and regeneration strategy.
Since the type of isotherm has a signicant
inuence on the choice of regeneration conditions
and the adsorption isotherm model is indispensable
to the application of the wave theory, adsorption
isotherms in various liquid-phase conditions were
determined and many adsorption isotherm models
were tried to t the experimental data. The major
parameters aecting the adsorption isotherms investigated in this study were solution pH, temperature,
and alcohol concentration. For each set of adsorption isotherm data, various adsorption isotherm
models (LeVan et al., 1997), listed in Table 2, were
rst tried and the associated model parameters were
determined.
The adsorption isotherm models listed in Table 2
can be classied as two-parameter models such as
Langmuir and Freundlich models and three-para-

meter models. In general, the three-parameter models

tted the experimental data better than the twoparameter models. The Langmuir model, based on
monolayer, uniform and nite adsorption site
assumptions, did not t all the experimental data
well. The Freundlich model, an empirical equation
for heterogeneous surface, tted all the experimental
data quite well. Other models listed in Table 2 can
reduce to either the Langmuir or Freundlich model in
special cases. For example, if m 1, the Fritz
Schluender, RadkePrausnitz, Toth, and Sips models
reduce to the Langmuir model; if s 0, the UNILAN model reduces to the Langmuir models; for
high liquid-phase concentration, the FritzSchluender and RadkePrausnitz models reduce to the
Freundlich model.
Although most models listed in Table 2 except
the Langmuir model tted individual set of data
quite well, the associated model parameters varied
with the solution pH, temperature and alcohol
concentration randomly. Only the Toth model
gave good correlations for varying solution pHs,
temperatures and alcohol concentrations. In the Toth
model, N represents the maximum adsorption
capacity, K the adsorption equilibrium constant,
and m the dissociation parameter. The eects of
solution pH, temperature, and alcohol concentration
on the parameters N; K, and m will be discussed
separately.
The functional groups or adsorption sites of
activated carbon are inuenced by the solution pH.
Therefore, the maximum adsorption capacity may be
expected to vary with the solution pH while the
adsorption equilibrium constant and the dissociation
parameter remain unaected by the solution pH.
Figures 3 and 4 show the experimental and Tothmodel predicted adsorption isotherms at 258C and
varying solution pHs for the red and yellow dyes,

Table 2. Various adsorption isotherm models used to t the

experimental data
Model

Equation

Parameters

Langmuir

Freundlich

q kC n

FritzSchluender

KNC
1 NC m

K, N, m

RadkePrausnitz

KNC
1 KCm

K, N, m

Toth

UNILAN


N
1 KCes
ln
2s
1 KCe
s

K, N, s

Sips

NKCm
1 KCm

K, N, m

KCN
1 KC

K, N
k, n

NC
1=K Cm 1=m

K, N, m

Fig. 3. Experimental and Toth-model predicted adsorption

isotherms of the red dye at varying solution pHs.

Desorption of dye from activated carbon beds

Fig. 4. Experimental and Toth-model predicted adsorption

isotherms of the yellow dye at varying solution pHs.

4163

Fig. 5. Experimental and Toth-model predicted adsorption

isotherms of the red dye at varying solution temperatures.

respectively. The modied Toth model, whose parameters shown in Figs 3 and 4, used to predict the
adsorption isotherms at varying solution pHs is as
follows:
q

a bpH cpH2 C
1=K Cm 1=m

According to equation (8), the maximum adsorption

capacity is expressed as a quadratic function of the
solution pH. The choice of the pH-dependent
function is arbitrary without any physical meanings.
Since the simple linear function could not give
good t to all the data, the quadratic function was
used. The average absolute deviations for the
yellow and red dyes are 3.83% for a total of 38 data
points and 3.57% for a total of 36 data points,
respectively.
The Gibbs free energy change of the adsorption
process is related to the equilibrium constant by the
following equation:
DG
RT ln K

According to thermodynamics, the Gibbs free energy

change is also related to the entropy change and heat
of adsorption at constant temperature by the
following equation:
DG DH
TDS
Combining equations (9) and (10) leads to



DS DH
K exp

R
RT

Fig. 6. Experimental and Toth-model predicted adsorption

isotherms of the yellow dye at varying solution temperatures.

10

11

The maximum adsorption capacity and dissociation constant are not aected by the solution
temperature while the equilibrium constant depends
on the temperature. Figures 5 and 6 show the
experimental and Toth-model predicted adsorption
isotherms at varying solution temperatures for the

red and yellow dyes, respectively. The modied Toth

model, whose parameters also shown in Figs 5 and 6,
used to predict the adsorption isotherms at varying
solution temperatures is as follows:
q

NC
e
DS=RDH=RT C m 1=m

12

The average absolute deviations for the yellow and

red dyes are 3.97% for a total of 37 data points and
4.75% for a total of 37 data points, respectively. As
shown in Figs 5 and 6, the entropy changes of
adsorption are
0.93 and
5.3 J/mol K for the
yellow and red dyes, respectively. The heat of
adsorption are
4.44 and
6.58 kJ/mol for the
yellow and red dyes, respectively. These values of
adsorption heat seems to be lower than those

4164

Jia-Ming Chern and Chia-Yuan Wu

evaluated from the modied Langmuir isotherm

model:
q

NCe
DS=RDH=RT
1 Ce
DS=RDH=RT

13

The heats of adsorption evaluated from equation (13)

are
16.3 and
8.49 kJ/mol for the yellow and red
dyes, respectively. However, some of the experimental data cannot be tted by the modied Langmuir
model as well as by the modied Toth model. In
general, the heat of chemical adsorption (40120 kJ/
mol) is considerably greater than that of physical
adsorption (below 40 kJ/mol). Hence, our results
show that the adsorption of the yellow and red dyes
on the activated carbon is probably physical adsorption.
Similar to the eect of solution pH, the existence
of alcohol in the dye solution may compete the
adsorption site with the dyes and thus reduces the
maximum adsorption capacity while has no inuence
on the adsorption equilibrium constant and the
dissociation parameter. Figures 7 and 8 show the
experimental and Toth-model predicted adsorption
isotherms at varying alcohol concentrations for the
red and yellow dyes, respectively. The modied Toth
model whose parameters Figs 7 and 8, used to predict
the adsorption isotherms at varying alcohol concentrations is as follows:
q

2
a bCalc cCalc
C

1=K Cm 1=m

solution, we can expect from the wave theory and the

isotherms shown in Figs 3 and 4 that the dye
solution can be appropriately adsorbed at a low pH
and regenerated at a high pH. Similar results can also
be expected to adsorb dye solution at a low
temperature and regenerate the bed at a high
temperature. Theoretically, for given adsorption
isotherms and with the help of the wave theory, we
can easily develop suitable pH-swing or temperatureswing adsorptiondesorption cycles for the color
removal from textile wastewaters. However, in
practical operations the carbon bed capacity may
decrease gradually in the adsorptiondesorption
cycles due to some adsorption site occupied permanently by alcohol. A series of adsorptiondesorption
cyclic operations are being carried out using alcohol
and other solvents to investigate the economic
feasibility of removing dye wastewater using carbon
adsorption process.
CONCLUSIONS
*

14

Again, the empirical quadratic function of alcohol

concentration was used to evaluate the maximum
adsorption capacity. The average absolute deviations
for the yellow and red dyes are 4.94% for a total of
42 data points and 3.50% for a total of 40 data
points, respectively.
Similar to the desorption of the carbon bed
presaturated with the yellow dye by 20% alcohol

Fig. 7. Experimental and Toth-model predicted adsorption

isotherms of the red dye at varying alcohol concentrations.

The adsorption isotherms of the yellow and red

dye solutions onto the granular activated carbon
at varying solution pHs, temperatures, and
alcohol concentrations were experimentally determined. Many adsorption isotherm models tted
individual set of data quite well, but the Toth
model was found to best t the adsorption
isotherm data for varying solution pHs, temperatures, and alcohol concentrations.
The activated carbon has a higher adsorption
capacity at lower pHs and lower alcohol concentrations. Correlation equations were obtained to
account for the eects of solution pH and alcohol
concentration on the maximum adsorption capacity of the activated carbon. The activated carbon
adsorbed more dyes at a lower solution temperature. A correlation equation was obtained to

Fig. 8. Experimental and Toth-model predicted adsorption

isotherms of the yellow dye at varying alcohol concentrations.

Desorption of dye from activated carbon beds

account for the eects of solution temperature on

the adsorption equilibrium constant. The entropy
changes of adsorption were found to be very low
and the low heat of adsorption showed that the
adsorption was physical adsorption, not chemical
adsorption.
The 258C water is a poor regenerant for the
carbon bed presaturated with the yellow dye
compared with 20% alcohol solution. Given the
presaturation concentration, bed density and void
fraction, and adsorption isotherm, the wave
propagation theory can predict the desorption
curves quite satisfactorily.
Desorption of saturated carbon bed with other
regenerants should be studied and repeated
adsorptiondesorption cycles with the optimal
regenerants should be investigated in the future to
make the carbon adsorption for color removal
ecologically and economically feasible.

Acknowledgements}This work was supported by the

National Science Council of Taiwan, Republic of China
(Grant No. NSC 88-2211-E-036-001).

REFERENCES

An H., Yi Q., Xiasheng G. and Tang W. Z. (1996)

Biological treatment of dye wastewaters using an
anaerobicoxic system. Chemosphere 33, 25332542.
Chatzopoulos D. and Varma A. (1995) Aqueous-phase
adsorption and desorption of toluene in activated carbon
xed beds: experiments and model. Chem. Eng. Sci. 50,
127141.
Chern J.-M. and Huang S.-N. (1998) Study of non-linear
wave propagation theory: 1. Dye adsorption by activated
carbon. Ind. Eng. Chem. Res. 37, 253257.

4165

Chern J.-M. and Huang S.-N. (1999) Study of non-linear

wave propagation theory: 2. Interference phenomena of
single-component dye adsorption waves. Sep. Sci. Technol. 34, 19932011.
Chern J.-M. and Wu C.-Y. (1999) Adsorption of binary dye
solution onto activated carbon: isotherm and breakthrough curves. J. CIChE 30, 507514.
Helerich F. G. and Klein G. (1970) Multicomponent
Chromatography: Theory of Interference. Marcel Dekker,
New York.
Ho Y. S. and McKay G. (1998) Sorption of dye from
aqueous solution by peat. Chem. Eng. J. 70, 115124.
LeVan M. D., Carta G. and Yon C. M. (1997) Adsorption
and ion exchange. In Perrys Chemical Engineers, Handbook 7th Ed, eds R. H. Perry, D. W. Green and J. O.
Maloney. McGraw-Hill, New York.
Liakou S., Pavlou S. and Lyberatos G. (1997) Ozonation of
azo dyes. Water Sci. Technol. 35, 279286.
Namasivayam C., Muniasamy N., Gayatri K., Rani M. and
Ranganathan K. (1996) Removal of dyes from aqueous
solutions by cellulosic waste orange peel. Bioresource
Technol. 57, 3743.
Nassar M. M., Hamoda M. F. and Radwan G. H. (1995)
Adsorption equilibria of basic dyestu onto palm-fruit
bunch particles. Water Sci. Technol. 32, 2732.
Ramakrishna K. R. and Viraraghavan T. (1997) Dye
removal using low cost adsorbents. Water Sci. Technol.
36, 189196.
Sarasa J., Roche M. P., Ormad M. P., Gimeno E.,
Puig A. and Ovelleiro J. L. (1998) Treatment of a
wastewater resulting from dyes manufacturing with
ozone and chemical coagulation. Water Res. 32,
27212727.
Walker G. M. and Weatherley L. R. (1997) Adsorption of
acid dyes on to granular activated carbon in xed beds.
Water Res. 31, 20932101.
Walker G. M. and Weatherley L. R. (1998) Fixed bed
adsorption of acid dyes onto activated carbon. Environ.
Pollut. 99, 133136.
Wolborska A. and Pustelnik P. (1996) A simplied method
for determination of the break-through time of an
adsorbent layer. Water Res. 30, 26432650.