Sie sind auf Seite 1von 10

chemical engineering research and design 1 0 9 ( 2 0 1 6 ) 495–504

Contents lists available at ScienceDirect

Chemical Engineering Research and Design

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

A review of the kinetics adsorption models


and their application to the adsorption
of lead by an activated carbon

L. Largitte a,∗ , R. Pasquier b


a COVACHIM M2E, Université des Antilles de la Guyane, Campus de Fouillole, BP 250, 97157 Pointe-à-Pitre Cedex,
France
b C3I, Université des Antilles de la Guyane, Campus de Fouillole, BP 250, 97157 Pointe-à-Pitre Cedex, France

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

Article history: 10 kinetic models are studied in detail and used after to model the adsorption kinetics of
Received 20 July 2015 lead on an activated carbon. All the kinetics parameters are determined: coefficients of
Received in revised form 28 January diffusion, rate constants and maximum lead adsorption, then, compared to the literature
2016 data. The data are better simulated by the nonlinear models in the order: simpli-
Accepted 5 February 2016 fied Elovich > Elovich > Bangham > pseudo order n > Boyd, Crank long times > pseudo order
Available online 11 February 2016 2 > Crank > Weber and Morris > Langmuir, pseudo order 1 > pore volume and surface diffu-
sion model by Leyva Ramos > simplified Crank short times. It is one of the first times that the
Keywords: kinetics of adsorption are fitted by the entire true Crank model. The kinetics of lead adsorp-
Adsorption tion by the activated carbon are best described by the Elovich equation: Q = (1/ˇ) Ln (˛ˇt)
Kinetics with ˛ equal to 0.035 mg g−1 s−1 and ˇ to 0.134 g mg−1 .
Modelling © 2016 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
Heavy metal removal

1. Introduction In this study, the adsorbent used is an activated carbon


from Coconut shells, the adsorbate is lead and one of the
An adsorption process includes one or more than one adsor- objective of this work is to study the kinetics of lead removal
bate which is fixed though physical or chemical bonds onto in aqueous environment by this activated carbon.
an adsorbent. Adsorption is widely used in wastewater treat- It is essential to describe precisely the kinetics of adsorp-
ment process (Foo and Hameed, 2009) to remove the pollutants tion (Gerçel and Gerçel, 2007), to compare the predicted
from water because this technique is simple, efficient, eco- adsorption parameters calculated by models to the experi-
nomically viable and socially acceptable (Sreejalekshmi et al., mental adsorbent behavior, in different adsorbent adsorbate
2009). systems and in varied experimental conditions (Mishra and
The adsorbents generally used are the activated carbons Patel, 2009; Shekinah et al., 2002).
because they possess large porous surface area, controllable The determination of the models adsorption parameters
pore structure, thermo-stability, interesting acid/base proper- permits to optimize the adsorption mechanism pathways,
ties and low cost if the preparation comes from byproducts. to express the dependence of the surface properties of the
The activated carbons are very efficient for the removal of a adsorbent to the sorption results, to determine the adsorbent
broad type of organic and inorganic pollutants dissolved in capacities and design effectively the adsorption systems (Ho
aqueous or in gaseous media (Li and Wang, 2009). et al., 1996; Thompson et al., 2001).


Corresponding author. Tel.: +33 590 590483403; fax: +33 590 590483403.
E-mail address: lucie.largitte@univ-ag.fr (L. Largitte).
http://dx.doi.org/10.1016/j.cherd.2016.02.006
0263-8762/© 2016 The Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
496 chemical engineering research and design 1 0 9 ( 2 0 1 6 ) 495–504

and nonlinear analysis, their application on the study of


Nomenclature lead adsorption by an activated carbon derived from Coconut
shells, the determination of the kinetics parameters and their
t time (s) comparison with the literature data and finally the selection
Q amount of adsorbate sorbed at instant t of the best model to fit the kinetics experimental data.
(mg g−1 )
Qmax maximum amount of adsorbate sorbed
2. Materials and methods
(mg g−1 )
Qr amount of adsorbate sorbed at t in the particle
2.1. Reagents
of adsorbent at a distance r (mg g−1 )
Qexp experimental amount of adsorbate sorbed at
The 69.4% HNO3 solution and the NaOH used to fix the pH,
instant t (mg g−1 )
the Pb(NO3 )2 which is the lead ions source and the NaNO3
Qmean mean of experimental amount of adsorbate
used to fix the ionic strength of the solution are AR grade and
sorbed at instant t (mg g−1 )
all obtained from Prolabo. All the solutions are prepared with
Qcalc amount of adsorbate sorbed at instant t pre-
deionized water.
dicted by the model (mg g−1 )
C0 initial concentration of adsorbate in solution
2.2. The activated carbon
(mg L−1 )
C concentration of adsorbate in solution at the
The activated Coconut shell carbon is a commercial carbon,
moment t (mg L−1 )
NC60, from the PICA Company. It is prepared by physical acti-
Cr concentration of adsorbate in the particle of
vation with carbon dioxide and water vapor. The activated
adsorbent at a distance r (mg cm−3 )
carbon sample, as received, is dried at 105 ◦ C for 2 days prior
k constant (Fick constant, simplification Crank
to the sorption experiments.
short times constant, Weber and Morris con-
stant, Bangham constant)
Ds diffusion coefficient at the surface (cm2 s−1 ) 2.2.1. Characterization
Dep diffusion coefficient in the pore volume 2.2.1.1. Texture. The texture of the activated carbon is charac-
(cm2 s−1 ) terized via adsorption of nitrogen at 77 K using a Micromeritics
kL external mass transfer coefficient (cm s−1 ) ASAP 2010 apparatus. About 0.30 g of sample is outgassed at
kad adsorption rate constant (L mg−1 s−1 ) 250 ◦ C for 48 h, prior to the adsorption measurements.
k ad pseudo adsorption rate constant (unit for The BET surface area (SBET ) (Brunauer et al., 1938), the DR-
order 1: s−1 , order 2: g mg−1 s−1 , order n: micropore volume (Vmi ) (Dubinin, 1960), the mesopore volume
mg1−n gn−1 s−1 , unit for Langmuir: s−1 ) (Vme ) and the DFT pore size distribution of the activated car-
kd desorption rate constant (s−1 ) bon are determined from the N2 adsorption isotherm. The
˛ initial sorption rate in the Elovich model mesopore volume is calculated as the volume of N2 adsorbed
(mg g−1 s−1 ) at P/P◦ = 0.99 minus the micropore volume and the density
ˇ constant related to the extent of surface cover- functional theory (DFT) pore size distribution is calculated by
age and activation energy for chemisorption in assuming a slit like geometry for the micropores (Olivier, 1995).
the Elovich model (g mg−1 )
 order of reaction to the adsorbate 2.2.1.2. Surface properties. The surface properties of the acti-
n order of reaction to the adsorbent in the pseudo vated carbon are characterized via the well-known Boehm
order n model or parameter in the Crank model method (Boehm, 1994). The pHPZC of the activated carbon
ϑ parameter in the Bangham model is determined using the potentiometric titration method as
S surface of adsorbate (cm2 g−1 ) described by (Noh and Schwarz, 1990). All the titrations for
V volume of adsorbate (mL) Boehm and pHPZC are performed on a 809 SM Titrino auto-
r distance in radial direction of adsorbent; matic titrator (Metrohm) and the maximum titrant dosing rate
0 < r < R (cm) selected is 0.25 mL min−1 . Duplicates are performed for each
D adsorbent dose (g L−1 ) sample and the relative standard error is equal or less than
ε void fraction in the adsorbent 3%.
P particle density of adsorbent (g cm−3 )
m weight of adsorbent (g) 2.3. Kinetic sorption studies
Ea activation energy (J mol−1 )
R gas constant (8.314 J mol−1 K−1 ) or Radius of the All glassware is presoaked in a 5% HNO3 solution, rinsed with
particle of adsorbent in the Crank model (cm) deionized water and oven dried. All the batch sorption tests
T temperature (K) and the absorption measurements are duplicated.
N number of experimental data In the operating mode, 1 g of activated carbon is put in
contact with 1 L of lead ion solution, at pH 5 (the adsorbent
dose D is equal to 1 g L−1 ). The flask is put in a stirred water
Over the years, a wide variety of kinetics models (Langmuir, bath maintained at a constant temperature of 30 ◦ C. The initial
Pseudo order 1, Pseudo order 2, Pseudo order n, Elovich, Crank, concentration of lead is 150 mg L−1 .
Boyd, Bangham, Weber and Morris, pore volume and surface At regular intervals, 5 mL of solution is removed from the
diffusion) have been formulated. flask, filtered through a Whatmann 32 filter paper and the con-
This paper presents an explained and detailed state of centration of the remaining lead in solution (C) is measured.
the art review of adsorption kinetics modelling by linear Note that the stirring is stopped before each removal and that
chemical engineering research and design 1 0 9 ( 2 0 1 6 ) 495–504 497

the total volume removed does not exceed 10% of the initial 3.1.1.2.2. Weber and Morris model (Weber and Morris,
volume of lead ion solution. 1963). Weber and Morris have studied the batch adsorption
The concentrations of lead are measured with an atomic of alkyl benzene sulfonates by activated carbon. They have
absorption spectrometer AA240FS from Varian. The amount shown that the plot of Q/Qmax versus t1/2 could be decomposed
of lead adsorbed at a time t, Q is calculated by the difference in two linear plots, the first one corresponding to the equation
between the initial (C0 ) and the instantaneous lead concentra-
tion (C): Q
= Cte × t1/2 (3)
Qmax
  C0 − C
Q mg g−1 = (1)
D and the second one to the equation

Qmax (mg g−1 ) is the maximum amount of lead adsorbed by Q


= Cte × t1/2 + Cte. (4)
the activated carbon. Qmax

3. Adsorption kinetics models Some authors suggest for the Weber and Morris model that
the first part of the plot also must be modelled by Eq. (4). They
explain the presence of the term Cte by the external mass
3.1. Kinetics models
transfer which is not negligible. The authors are not convinced
this is true as there is no apparent reason for the amount of
There are three steps in an adsorption process. First, the exter-
lead adsorbed at t = 0 to be different from zero.
nal mass transfer of the adsorbate from the bulk solution to
3.1.1.2.3. Bangham model. The Bangham model can be
the external surface of the adsorbent, followed by the inter-
understood as the generalization of the Weber and Morris
nal diffusion of the adsorbate to the sorption sites and finally
model or the application of the Freundlich and Helle (1939)
the sorption itself. Some models are based on the fact that the
model to kinetics. The equation is given in Table 1.
sorption is the rate limiting step in the adsorption process,
others suppose that the diffusion is the rate limiting step. So,
3.1.1.3. Complex diffusional model. More realistic models for
the fitting to the models permits elucidation of the adsorption
the solid phase can be used to describe the diffusion pro-
mechanism.
cess. However, their greater complexity makes their use
mathematically more difficult. An example of a complex dif-
3.1.1. Diffusional models
fusional model is the pore volume and surface diffusion model
The diffusional models assume that the diffusion is the rate
(PVSDM) suggested by Geankoplis and Leyva-Ramos (1985).
limiting step. They are divided in two parts, the external mass
It is a complex diffusional model which takes into account
transfer model and the internal diffusional models.
adsorption, external mass transfer and intraparticle diffusion.
It is assumed that the intraparticle diffusion is due both to
3.1.1.1. External mass transfer model. Mass transfer from the
pore volume diffusion and surface diffusion. All the equations
bulk liquid to the solid surface is assumed to be the slowest
needed to solve the model are given in Table 1.
step. It is characterized by a mass transfer coefficient. How-
ever, it seems implausible to the authors that for an adsorption
3.1.1.4. Adsorption models. In the adsorption models, the
process the approach of the adsorbate to the surface of the
adsorption is considered to be the slowest process. This is usu-
adsorbent is the slowest step. Moreover, the effect of trans-
ally the case when the adsorbate uptake on the adsorbent is
port is eliminated by mechanical mixing with the rapid and
of chemical nature (i.e., chemisorption).
constant stirring of the solution.
In the following equations S is the adsorption sites, M is the
adsorbate, C is the concentration of the adsorbate in solution,
3.1.1.2. Internal diffusional models. The mass transfer into the
, the order of reaction to the adsorbate, SM is the concentra-
interior of the particle, characterized by an intraparticle diffu-
tion of adsorbate bound to the sorbent, kad is the adsorption
sion coefficient, can be the slowest step.
rate constant and kd the desorption rate constant. The sorp-
3.1.1.2.1. Crank model (Crank, 1956). The equation
tion rate is equal to dQ/dt.
describing homogeneous diffusion of an adsorbate in an
3.1.1.4.1. Pseudo first order model (Lagergren, 1898).
adsorbent modelled as a sphere and assuming a constant sur-
The pseudo first order model is described by the non-reversible
face diffusivity Ds at all points in the particle is
equation:

∂Q DS ∂
 ∂Q 
2
= 2 r , (2) S + M → MS (5)
∂t r ∂r ∂r

r being the distance in the radial direction. The kinetic equation rests on five assumptions:
Crank gives an exact solution to this equation for the aver-
age concentration in the solid at any given time for the case 1. Sorption only occurs on localized sites and involves no
where the sphere is initially free of solute and the concentra- interaction between the sorbed ions.
tion of the solute at the surface remains constant (that is to 2. The energy of adsorption is not dependent on surface cov-
say in the absence of external mass transfer). erage.
For short times, or more precisely for Q/Qmax < 0.3, this 3. Maximum adsorption corresponds to a saturated mono-
equation may be simplified, idem for long duration (more layer of adsorbates on the adsorbent surface.
precisely for Q/Qmax > 0.85). The simplified Crank long times 4. The concentration of M is considered to be constant.
equation is also known as the Boyd equation (Boyd et al., 1947). 5. The metal ion uptake on the activated carbons is governed
The different Crank equations are given in Table 1. by a first-order rate equation.
498
Table 1 – Kinetic models of adsorption.
Process control Kinetic law Nonlinear form Linear form Linear plot

External mass transfer Fick equation – – –


  ∞    
Internal diffusion Crank (Q/Qmax ) = 1 − 6/ 2
n=1
1/n 2
exp −DS n  t/R 2 2 2
– –
Ln (Q/Qmax )   1/2 
 1/2   1/2  = (1/2) Ln t + Ln −1/2 − (1/2) Ds t/R2
Crank (short times) (Q/Qmax ) = 6 Ds t/R2 −1/2 − (1/2) Ds t/R2 –
 

chemical engineering research and design 1 0 9 ( 2 0 1 6 ) 495–504


Q/Qmax < 0.3
+ (1/2) Ln Ds /R 2
+ Ln 6
 1/2
Simplification Crank Q = kt1/2 with k = 6 Ds /R2 Ln Q = (1/2) Ln (t) + Ln (k) Ln Q = f(Ln t)
short times
Weber and Morris Q = kt1/2 + Cte – –
Bangham Q = ktϑ     Ln Q = ϑLn (t) + Ln (k)     Ln Q = f(Ln t)
Boyd, Crank long times: Q/Qmax = 1 − 6/2 exp −Ds 2 t/R2 Ln (1 − (Q/Qmax )) = Ln 6/2 − DS 2 t/R2 Ln (1 − cteQ) = f(t)
Q/Qmax > 0.85

Complex diffusional Geankoplis and V (dC/dt) = −mSkL (C − Cr |r=R ) – –


model Leyva-Ramos ε (∂Cr /∂t) + P (∂Q/∂t)
 

= 1/r2 (∂/∂r) r2 (Dep (∂Cr /∂r) + DS P (∂Q/∂r))



(∂Cr /∂t) =0
r=0
Dep (∂Cr /∂r) + DS P (∂Q/∂r) = kL (C − Cr |r=R )
t = 0 ; C = C0
t = 0 ; Cr = 0 ; 0 ≤ r ≤ R
Q(r) = f(Cr )
      
  
Adsorption Langmuir (1 site) Q = Qmax kad / kad + kd 1−e
− k +kd
ad
t Ln
1 − Q/Qmax kad / kad + kd Ln (1 − cteQ) = f(t)
= − kad + kd t
 
−k t
Pseudo order 1 Q = Qmax 1−e ad Ln (1 − (Q/Qmax )) = −kad t
     
Pseudo order 2 Q = kad Qmax
2
t/ 1 + kad Qmax t t/Q = (t/Qmax ) + 1/kad Qmax
2
t/Qt = f(t)
  
1/(n−1) 
Pseudo order n for n =
/ 0 Q = Qmax 1 − 1/ 1 + kad (n − 1) Qmax
n−1
t – –

Elovich Q = (1/ˇ) Ln (1 + ˛ˇt) Ln Q = Ln (1/ˇ) + Ln2 (1 + ˛ˇt) –

Simplification of Q = (1/ˇ) Ln (˛ˇt) Q = (1/ˇ) Ln (˛ˇ) + (1/ˇ) Ln t Q = f(Ln t)


Elovich: 1  ˛ˇt
chemical engineering research and design 1 0 9 ( 2 0 1 6 ) 495–504 499

Then, the sorption rate can be written as Then, the adsorption rate constant kad can be written, using
the Arrhenius equation, as:
Q 2
kad C
max −Q
= kad (Qmax − Q) (6)
 Ea 
Qmax kad = Cte × exp − = ˛ × exp (−ˇQ) (14)
RT

and the resulting equation after integration is given in Table 1.


with
3.1.1.4.2. Pseudo second order model (Ho et al., 1996; Ho  Ea 
˛ = Cte × exp −
and McKay, 1999). The kinetics of metal ion removal are 0
(15)
described by the pseudo second order model as follows: RT

3. The concentration of M is considered to be constant.


2S + M → M(S)2 (7) 4. The metal ion uptake on the activated carbons is negligible
before the exponential (equivalent to a metal ion uptake
The assumptions are almost the same as for the pseudo governed by a zero order rate equation).
first order model except the metal ion uptake on the activated
carbons is governed by a second-order rate equation. Then, the sorption rate can be written as:
Then, the sorption rate can be written as
˛ × exp (−ˇQ) (16)
Q 
−Q 2
max
kad C = kad (Qmax − Q)2 (8) with
Qmax

The resulting equation by integration is given in Table 1. ˛ = C ˛ (17)


3.1.1.4.3. Pseudo n order model for n different from zero
(Ritchie, 1977). The kinetics of metal ion removal can be and the resulting integration is given in Table 1.
described by the pseudo n order model. The equation is
3.2. Fitting the kinetics models to the experimental
data
nS + M → M(S)n (9)

All these models are tested to describe the sorption of lead


The assumptions are the same as for the pseudo first order on the activated carbon from Coconut shells using a home-
model except the last one as the metal ion uptake on the made program. The experimental sorption data are fitted
activated carbons is governed by a rate equation of order n. to the linear and the nonlinear form of the models. So,
Then, the sorption rate can be written as linear and nonlinear least squares analysis is applied to
Q −Q
n each set of adsorption data using the Ceres solver library
max
kad C = kad (Qmax − Q)n (10) [ht*tp://co*de.go*ogle.co*m/p/ceres-solver/].
Qmax The root mean square error (E) is calculated by the expres-
sion:
The result of the integration is given in Table 1.

3.1.1.4.4. Langmuir model (Langmuir, 1918). The kinet-  N
1
E=
ics of metal ion removal are described by the Langmuir model 2
(Qexp − Qcalc ) (18)
using this reversible equation: N
i=1

S + M  MS (11) N being the number of experimental points. The best model


describing the experimental data is the one with the lowest E.
The four assumptions are the same as for the pseudo first- The correlation coefficient of determination R2 is also cal-
order model but now the rate of desorption is not negligible culated as
compared to the rate of adsorption.
SSres
Then, the sorption rate can be written as R2 = 1 − (19)
SStot
Q −Q
  Q 
max
kad C − kd = kad (Qmax − Q) − kd Q (12)
Qmax Qmax

N
2
With the residual sum of squares, SSres = (Qexp − Qcalc ) ;
and the integrated equation is given in Table 1.
i=1
3.1.1.4.5. Elovich model (Elovich and Larinov, 1962). The (20)
kinetics of metal ion removal is described by the Elovich model
as follows:

N
2
With the total sum of squares, SStot = (Qexp − Qmean )
S + M → MS (5)
i=1
(21)
1. Sorption only occurs on localized sites and there is inter-
action between the sorbed ions.
2. The energy of adsorption increases linearly with the sur- The Fisher test (FR2 ) of two models seem close to each other
face coverage according to the law: in order to discriminate without any doubt the best model.
F is determined through the ratio of the residual sum of
Ea = Ea0 + RTˇQ (13) squares (SSres ) of the two models.
500 chemical engineering research and design 1 0 9 ( 2 0 1 6 ) 495–504

The P value (P value) is then calculated via the F cumulative


Table 2 – Characterization of the activated carbon.
distribution function (cdf):
Activated carbon
from coconut shells
P value = 1 − cdf(F, df1 , df2 ) (22)
SBET (m2 g−1 ) 1058
Vmi (cm3 g−1 ) 0.49
df1 , df2 is respectively the degree of freedom for each of the Vme (cm3 g−1 ) 0
two models. It is calculated as the number of experimental
Total of acid oxygenated surface groups 0.3
data less the number of parameters to estimate in the model.
(mmol g−1 )
Other error functions like E or R2 to compute the difference Carboxyl groups 0
between the experimental data and the predicted values are Hydroxyl groups 0.3
given in Foo and Hameed (2009). Lactone groups 0
Note that the modelling for the Crank model and the pore Total of basic oxygenated surface groups 0.9
volume and surface diffusion model suggested by Geankoplis (mmol g−1 )
pHPZC 9.2
and Leyva Ramos are explained with precision in other papers
(Largitte and Laminie, 2015; Largitte and Pasquier, 2016a).
Note that also the linear and nonlinear Boyd, Crank (long
times) modelling and linear pseudo order 1 modelling need to
fix a Qmax to begin the computation. Qmax is fixed at 52 mg g−1 ,
the next higher entire value than the experimental one.

Amount of lead adssorbed (mg.g-1)


50
4. Results and discussion
40
4.1. Characterization of the activated carbon short mes
30
long mes
The textural properties of the activated carbon have been
Weber and Morris
studied. The nitrogen adsorption/desorption isotherm for 20
Data
the activated Coconut shells carbon is shown in Fig. 1. The
isotherm is of type I according to the BDDT classification, 10

characteristic for microporous materials (Gregg and Sing,


0
1982).
0 1000 2000 3000 4000
The values of SBET , Vmi , Vme obtained from the isotherm are
Time (min)
listed in Table 2. The mesopore volume is negligible; the micro-
pore volume is in high agreement with such a high surface
Amount of lead adssorbed (mg.g-1)

50
area. The micropore size distribution of the activated carbon
is studied according to the DFT model. The DFT plot shows 40
that there are many micropores of size 6 Å, ´ close to the size of short
30 mes
the hydrated lead cation (Nightingale, 1959). long mes
The chemical surface properties of the activated carbon are
20 Bangham
also listed in Table 2.
The total number of acid groups on the surface of the
10
activated carbon is low (0.3 mmol g−1 ) compared to the total
number of basic groups (0.9 mmol g−1 ). The distribution of the
0
acidity is narrow because only alcohol groups are present. The 0 1000 2000 3000 4000
activated coconut shells carbon is a basic activated carbon Time (min)
with a pHPZC equal to 9.2.
Amount of lead adssorbed (mg.g-1)

50

40

30 short mes
long mes, Boyd
20 Crank
Data
10

0
0 1000 2000 3000 4000
Time (min)

Fig. 2 – Experimental kinetics adsorption (the points) and


Fig. 1 – Adsorption/desorption isotherm of N2 at 77 K on the (a) modelled kinetics adsorption by Weber, Morris
activated Coconut shells carbon. Pore size distribution of equations; (b) by Bangham equations and (c) by Crank
the activated carbon from Coconut shells by applying the equations using all the data and only the data at short
DFT method to the nitrogen adsorption isotherm at 77 K. times or long times.
chemical engineering research and design 1 0 9 ( 2 0 1 6 ) 495–504 501

The kinetic results are fitted according to the linear and


50 nonlinear form of the kinetics models. The results of the fitting
Amount of lead adsorbed (mg.g-1)

are given in Table 3.


40
4.2.1. Comparison between the linear and the nonlinear
Data
30 models
Crank Globally, the order of magnitude of the root mean square error
20 Bangham E follows the same trend for the linear and the nonlinear mod-
Weber and Morris els except for the Boyd, Crank long duration model which has
10 Geankoplis the lowest E in the linear modelling and one of the worst in
the nonlinear modelling.
However, as the linearization is not possible for all the mod-
0
0 1000 2000 3000 4000 els, the comparison between the linear and nonlinear models
Time (min) is truncated.
Moreover, other authors have pointed that the linearization
bias the estimate of model’s parameter set (Ho, 2006). This
50
Amount of lead adsorbed (mg.g-1)

means that the use of nonlinear expressions must be favored.


Then, only the results obtained from nonlinear modelling will
40
be further discussed.

30
4.2.2. Comparison between the time range for the kinetic
modelling
20 Data Order 1, Langmuir
Order 2 Order n = 2.8 As it can be seen from Figs. 2–4, the whole models are com-
Elovich Simplified Elovich pared in the entire time range available except for the Crank
10 short duration and long duration models which are, by default,
limited to a short duration range. However, restricting the
0 study to the beginning or the latest stage of the process, as
0 1000 2000 3000 4000
it can be seen in Fig. 2, is not representative for the whole pro-
Time (min) cess and can lead to an under or overestimation of the kinetics
Fig. 3 – Experimental kinetics adsorption (the points) and parameters. Then, the best way to obtain an average is to take
modelled kinetics adsorption (the lines); (a) by the the data for the entire uptake process.
diffusional equations (b) by the adsorption equations. Consequently, it is better to use the Crank model in the
entire duration range instead of the Crank short times or long
times models in short duration ranges as can be seen in Fig. 2.
4.2. Application of the kinetics models This result can be generalized and the models describing the
entire time range of the process must be favored. This is the
The experimental uptake rate of lead by the activated carbon case for the Crank, Bangham, Leyva Ramos, Langmuir, Elovich
from Coconut shell is studied at pH 5 for an initial lead concen- and Pseudo order 1, 2 and n models.
tration of 150 mg L−1 and an adsorbent dose of 1 g L−1 at 30 ◦ C.
The sorption is quite rapid as can be seen in Figs. 2–4: 16% 4.2.3. Comparison between the kinetics parameters
of the lead is removed within the first two hours followed by Fig. 3 presents the kinetics data and the fitting plots for all
a slow approach to equilibrium. Equilibrium is reached after the diffusional and all the adsorption models and Table 3, the
50 h and the value of the experimental maximum amount of kinetics parameters obtained from the modelling.
lead adsorbed is 50.2 mg g−1 . According to the diffusional models, that is to say,
the Crank, Bangham, Leyva Ramos, Weber and Morris and
Geankoplis models, the diffusion coefficient (Ds , Dep ) or the
rate constant k can be calculated. The maximum amount of
50 lead adsorbed Qmax predicted by the model can also be calcu-
Amount of lead adsorbed (mg.g-1)

lated and compared to the experimental maximum amount


40 which is 50.2 mg g−1 .
Data As it has been previously assumed, the diffusion coeffi-
30 n = 2.8 cient at the surface Ds is different in the Crank long times
n=6 model (here, underestimated) compared to the entire Crank
20 model. However, the maximum amount of lead predicted by
n = 40
the Crank long times model (52 mg g−1 ) is closer to the exper-
imental result.
10
On the contrary, the pore volume and surface diffusion
model suggested by Leyva Ramos shows that when one
0
0 1000 2000 3000 4000 assumes only diffusion inside the pore volume, Ds is equal
Time (min) to zero and Dep /R2 is 7.5E − 03 s−1 , a value four times higher
than the one obtained for Ds /R2 in the simplified Crank short
Fig. 4 – Experimental kinetics adsorption (the points) and duration model (1.71E − 03 s−1 ).
modelled kinetics adsorption by Pseudo order n equation The simplified Crank model and the Weber and Morris
(the lines). model can be compared to each other because they both use
502 chemical engineering research and design 1 0 9 ( 2 0 1 6 ) 495–504

Table 3 – Kinetics modelling results.


Kinetics laws Nonlinear form Results for the E in the Results for the linear
nonlinear form linear form form
2
E R

Crank 2.600 0.923 Qmax = 46.8 mg g−1 ; – –


Ds /R2 = 4.74E − 06 s−1
Crank short times na na na na na
Q/Qmax < 0.3 (not
enough points)
Simplification Crank 10.469 na k = 0.140 mg g−1 s−1/2 = 15.674 Ln k = −1.708, then
 1/2
short times 6 Ds /R2 ; k = 0.181 mg g−1 s−1/2 ;
Ds /R2 = 2.86E − 03 mg2 g−2 s−1
Ds /R2 = 1.71E − 03 mg2 g−2 s−1
Weber and Morris: 2.694 0.917 k = 0.07 mg g−1 s−1/2 ; – –
kt1/2 + Cte Cte = 22.350 mg g−1
Bangham: kt 1.175 0.984 k = 4.639 mg g−1 s− ; ϑ = 0.198 1.288 Ln k = 1.398, then
k = 4.047 mg g−1 s− ;
ϑ = 0.210
Boyd, Crank long times 1.802 na Qmax = 52.05 mg g−1 ; 0.360 Ds /R2 = 1.61E − 06 s−1
Q/Qmax > 0.85 (only 4 Ds /R2 = 1.90E − 06 s−1
points)
Geankopolis and 6.774 0.479 Dep /R2 = 7.50E − 03 s−1 ; Ds = 0 cm2 s−1 ; – –
−1
Leyva-Ramos  −02 cm s 
kL = 4.60E
Langmuir (1 site) 3.785 0.837 Qmax kad / kad + kd = 45.76 – –
kad + kd = 1.03E − 04
Then, k ad Qmax = 45.13E − 04
Pseudo order 1 Qmax = 45.76 mg g−1 ; 10.795 k ad = 1.94E − 05 s−1
k ad = 1.03E − 04 s−1
Pseudo order 2 2.323 0.939 kad Qmax
2
= 0.0072; kad Qmax = 3.917 k ad Qmax 2 = 0.0041;
1.46E − 04 Qmax = 53.75 mg g−1 Then
Then, k ad = 2.96E − 06 g mg−1 s−1 ; k ad = 1.44E − 06 g mg−1 s−1
Qmax = 49.33 mg g−1

Pseudo order n 1.686 0.968 k ad = 1.27E − 07 mg1−n gn−1 s−1 ; – –


Qmax = 54.18 mg g−1
for n = 2.8
0.839 0.992 k ad = 7.16E − 14 mg1−n gn−1 s−1 ;
Qmax = 76.83 mg g−1
for n = 6
0.462 0.99758 k ad = 4.14E − 103 mg1−n gn−1 s−1 ;
Qmax = 331.76 mg g−1
for n = 40
Elovich 0.468 0.99751 (1/ˇ) = 7.524; ˛ˇ = 0.004 – –
Then ˇ = 0.133 g mg−1 ;
˛ = 0.033 mg g−1 s−1
Simplification of 0.461 0.99758 1/ˇ = −39.77; ˛ˇ = 7.44 0.461 1/ˇ = −39.77; ˛ˇ = 7.44
Elovich: 1  ˛ˇt Then, ␤ = 0.134 g mg−1 ; ˇ = 0.134 g mg−1 ;
␣ = 0.035 mg g−1 s−1 ; ˛ = 0.035 mg g−1 s−1

the parameters k and t0.5 , however this is not the case for the modelling because there are two unknown parameters and
Bangham model which uses a dependence of k and t0.2 . only one equation.
The parameter k is the double for the simplified Crank A method to determine the constants k ad and kd is to
model than for the Weber and Morris one (0.14 mg g−1 s−1/2 use Qmax exp = 50.2 mg g−1 or Qmax predicted by the Lang-
against 0.07 mg g−1 s−1/2 ). muir equilibrium model equal to 47.7 mg g−1 (Largitte and
The results obtained from the Weber and Morris modelling, Pasquier, 2016b). Knowing the sum of the constants k ad and
from the Crank long duration modelling and the Bangham kd equal to 1.03E − 04 and k ad Qmax equal to 45.13E − 04, k ad
modelling are in agreement with the literature data found for can be deduced. It is equal to 8.99E-05 (9.5E − 05) s−1 and kd to
the sorption of lead by activated carbons (Sreejalekshmi et al., 1.31E − 05 (8.4E − 06) s−1 .
2009; Li and Wang, 2009; Gerçel and Gerçel, 2007; Mishra and Considering that the value of k ad is higher than kd , one
Patel, 2009; Shekinah et al., 2002; Ho et al., 1996; Nadeem et al., can consider the adsorption to be easier than the desorption
2006; Wang et al., 2010; Ayyappan et al., 2005). on the surface of this activated carbon.
Concerning the adsorption models, Langmuir, Elovich, These results are near those of Machida et al. (2005) and
Pseudo order 1, 2 and n models, the pseudo adsorption rate Sekar et al. (2004).
constant or the sorption rate and sometimes the maximum Note that he knowledge of the ratio kad /kd obtained though
amount of lead adsorbed Qmax can be calculated. the fitting of the isotherm data by the Langmuir equilibrium
Note that the pseudo adsorption rate constant in the Lang- model can also permit to determine kad . According to Largitte
muir model (k ad ) and the desorption rate constant kd cannot and Pasquier (2016b), the ratio kad /kd is equal to 0.749. Solving
be calculated solely with the results of the Langmuir kinetics the equation yields kad equal to 9.8E − 06 (6.3E − 06) L mg−1 s−1 .
chemical engineering research and design 1 0 9 ( 2 0 1 6 ) 495–504 503

The kinetics are better described by the pseudo order n 55


model and the pseudo order 2 model than by the pseudo 1 y = 7.5474x - 10.173

Amount of lead adsorbed (mg.g-1)


50 R² = 0.98
model.
For the pseudo second order model, the pseudo rate con- 45
stant k ad is equal to 2.96E − 06 g mg−1 s−1 and the maximum
40
amount predicted by the model is 49.33 mg g−1 .
In the case of the pseudo order n model, a higher value of 35 y = 7.9469x - 12.069
R² = 0.992
n gives a better fitting (see Fig. 4 and Table 3), but produces 30
predicted maximum amounts that are very different from the
experimental one. The most coherent result is obtained for 25

n equal to 2.8, which gives a predicted maximum amount of 20


54.2 mg g−1 . 4 5 6 7 8 9
Many data are available on the pseudo rate constant for Ln(t+t0) With t0 = 1/(αβ)
the pseudo first order model (Sreejalekshmi et al., 2009; Li and
Fig. 5 – The two types of sorption sites implied in the
Wang, 2009; Gerçel and Gerçel, 2007; Mishra and Patel, 2009;
sorption process according to Elovich model.
Shekinah et al., 2002; Nadeem et al., 2006; Wang et al., 2010;
Ayyappan et al., 2005; Ibrahim et al., 2012; Inyang et al., 2012;
From these results, a simplified Elovich model can be put
Largitte et al., 2012, 2014; Ricordel et al., 2001), pseudo second
forward for the sorption of lead by the activated Coconut shells
order model (Sreejalekshmi et al., 2009; Li and Wang, 2009;
carbon. This means the rate limiting step in the process is the
Gerçel and Gerçel, 2007; Mishra and Patel, 2009; Nadeem et al.,
chemical sorption between the lead and the active sites of the
2006; Wang et al., 2010; Ibrahim et al., 2012; Inyang et al., 2012;
carbon and that the activation sorption energy increases with
Largitte et al., 2012, 2014] and pseudo nth order model (Ricordel
the process.
et al., 2001) which can be compared to our results.
So, the experimentally demonstrated observation of the
Concerning the Elovich model, there is only a minor dif-
sorption rate slowing down with time is explained by the
ference between the results of the Elovich model and the
Elovich model that explains, at first, the lower energy surface
simplified Elovich model. The initial sorption rate ˛ changes
sites of the carbon saturate, and then, the adsorption shifts to
from 3.3E − 02 to 3.5E − 02 mg g−1 s−1 and the constant related
the higher energy surface sites, resulting in a decrease of the
to the extent of surface coverage and activation energy for
sorption rate.
chemisorption, ˇ, from 0.133 to 0.134 g mg−1 . These results are
According to the Elovich model assumptions and to the
concordant with previous results from other authors (Li and
break of slope in the derivative Elovich plot (Fig. 5), there are
Wang, 2009; Wang et al., 2010; Ibrahim et al., 2012; Inyang et al.,
two types of sorption sites implying in this sorption process.
2012).
Two sorption sites, probably relative to the both ways of sorp-
The initial sorption rate can also be calculated from the
tion of lead cations on the adsorbent surface, one with the
pseudo second order model using the expression k ad Qmax 2
oxygen of the alcohol surface groups by complexation and the
and can be compared to the value ˛ obtained by the
other one with the ␲ electrons of the graphene layers of the
Elovich model. The result gives 7.2E − 03 mg g−1 s−1 and is
activated carbon by electrostatic interaction.
five times smaller than the value obtained from Elovich
Consequently, the kinetics of the adsorption of lead by the
(3.5E − 02 mg g−1 s−1 ).
activated carbon from Coconut shell can be described though
To conclude, the value of the maximum amount of lead
the equation:
adsorbed obtained for this activated carbon, from the experi-
ment or from the fitting, is really well compared to literature 1
Q= Ln (˛ˇt) (23)
data. This result is in agreement with the high surface area ˇ
of the activated carbon, the high micropore volume and the
presence of alcohol groups on its surface which can promote with ˛ equal to 0.035 mg g−1 s−1 and ␤ to 0.134 g mg−1 .
the complexation of lead to the carbon surface.
5. Conclusion

4.2.4. Comparison between the error values This paper aims to study in detail the different existing models
The values of the error E and R2 are presented in Table 3 and of kinetics adsorption and to apply them to the sorption of lead
Fig. 3 presents the fitting of the different diffusional and sorp- onto an activated carbon. The kinetics is studied at 30 ◦ C, ini-
tion models to the experimental data. The lowest value of E tial pH 5, initial lead concentration 150 mg L−1 and adsorbent
and the best coherent fitting of the models is for the simplified dose 1 g L−1 for an activated carbon from Coconut shells. The
Elovich model and the error increases in the order: sim- objectives are to explain, with precision, all the kinetics mod-
plified Elovich < Elovich < Bangham < pseudo order n < pseudo els, to determine all the kinetics parameters, to compare them
order 2 < Crank < Weber and Morris < Langmuir, pseudo order to the literature data and finally to find the model that can bet-
1 < pore volume and surface diffusion model by Leyva Ramos. ter describe the experimental results of adsorption kinetics of
According to these error results, Elovich is the best the activated carbon. The screening is made via the root mean
model which describes the experimental data. However, the square error and the correlation coefficient of determination.
E value for Bangham and Elovich are very proximate. So So, 10 mathematical models of adsorption are studied
a F test is made to choose undoubtedly the best model. and several kinetic parameters (coefficient of diffusion, rate
F = SS(Bangham)/SS(Elovich) = 6.30517; then, the calculated constant, sorption rate and maximum of lead adsorbed) are
P value is equal to 0.0087. The P value, inferior to 0.01, shows calculated.
that the Elovich model is statistically better than the Bangham The results show that the Crank, the Bangham, the pore
model with a false-rejection probability inferior to 0.01. volume and surface diffusion by Geankoplis and Leyva Ramos,
504 chemical engineering research and design 1 0 9 ( 2 0 1 6 ) 495–504

the Langmuir, the Elovich, the pseudo order 1, 2 and n models from aqueous solution by biochars derived from anaerobically
fit the experimental data quite well over a wide time range. digested biomass. Bioresour. Technol. 110, 50–56.
This is in contrast to the Crank short or long times (simplified Lagergren, S., 1898. Zur theorie Der Sogenannten adsorption
geloster stoffe, Kungliga Svenska Vetenskapsakademiens.
or not). So the first group of models cited must be preferred to
Handlingar 24, 1–39.
describe the kinetics of adsorption. It can also be noted that Langmuir, I., 1918. Adsorption of gases on glass, mica and
it is one of the first times, to our knowledge, that the Crank platinum. J. Am. Chem. Soc. 40, 1361–1403.
model is used to model all the kinetic data. Largitte, L., Gervelas, S., Tant, T., Dumesnil, P.C., Lodewyckx, P.,
In the case of this activated coconut shell carbon, the sorp- 2012. Influence of the surface properties of the Bois Carré
tion data are better simulated by the nonlinear models in the seeds activated carbon for the removal of lead from aqueous
order: simplified Elovich > Elovich > Bangham > pseudo order solutions. Eurasian Chem.-Technol. J. 14 (3), 201–210.
Largitte, L., Laminie, J., 2015. Modelling the lead concentration
n > pseudo order 2 > Crank > Weber and Morris > Langmuir,
decay in the adsorption of lead onto a granular activated
pseudo order 1 > pore volume and surface diffusion model by carbon. J. Environ. Chem. Eng. 3, 474–481.
Geankoplis and Leyva Ramos. Largitte, L., Pasquier, R., 2016a. The Crank Equation, from Theory
Based on the lowest value of the root mean square error, it to Practice. Bioresource Technology (Submitted, (a)).
can be said that the adsorption of lead on this activated carbon Largitte, L., Gervelas, S., Tant, T., Couespel Dumesnil, P.,
is best described by the simplified Elovich model. The kinetics Hightower, A., Yasami, R., Bercion, Y., Lodewyckx, P., 2014.
Removal of lead from aqueous solutions by adsorption with
correspond to the equation: Q = (1/ˇ) Ln (˛ˇt) with ˛ equal to
surface precipitation. Adsorption 20 (5–6), 689–700.
0.035 mg g−1 s−1 and ˇ to 0.134 g mg−1 .
Largitte, L., Pasquier, R., 2016b. A Review of the Adsorption
Isotherm Models and their Application to the Equilibrium of
References Lead Adsorption. Miccroporous and mesoporous materials
(Submitted, (b)).
Li, K., Wang, X., 2009. Adsorptive removal of Pb II by activated
Ayyappan, R., Carmalin Sophia, A., Swaminathan, K., Sandhya, S.,
carbon prepared from Spartina alterniflora: equilibrium,
2005. Removal of Pb(II) from aqueous solution using carbon
kinetics and thermodynamics. Bioresour. Technol. 100,
derived from agricultural wastes. Process Biochem. 40 (3–4),
2810–2815.
1293–1299.
Machida, M., Aikawa, M., Tatsumoto, H., 2005. Prediction of
Boehm, H.P., 1994. Some aspects of the surface chemistry of
simultaneous adsorption of Cu(II) and Pb(II) onto activated
carbon blacks and other carbons. Carbon 32, 759–769.
carbon by conventional Langmuir type equations. J. Hazard.
Boyd, G.E., Adamson, A.W., Meyers Jr., L.S., 1947. The exchange
Mater. B 120, 271–275.
adsorption of ions from aqueous solutions by organite
Mishra, P.C., Patel, R.K., 2009. Removal of lead and zinc ions from
zeolithes. II. Kinetics. J. Am. Chem. Soc. 69, 2836.
water by low cost adsorbents. J. Hazard. Mater. 168, 319–325.
Brunauer, S., Emmett, P.H., Teller, E., 1938. J. Am. Chem. Soc. 60,
Nadeem, M., Mahmood, A., Shahid, S.A., Shah, S.S., Khalid, A.M.,
309–319.
McKay, G., 2006. Sorption of lead from aqueous solution by
Crank, J., 1956. Mathematics of Diffusion. Oxford at the
chemically modified carbon adsorbents. J. Hazard. Mater. 138
Clarendon Press, London.
(3), 604–613.
Dubinin, M.M., 1960. The potential theory of adsorption of gases
Nightingale, E.R., 1959. Phenomenological theory of ion solution.
and vapors for adsorbents with energetically non uniform
Effective radii of hydrated cations. J. Phys. Chem. 63,
surfaces. Chem. Rev. 60 (2), 235–241.
1381–1387.
Elovich, S.Y., Larinov, O.G., 1962. Theory of adsorption from
Noh, J.S., Schwarz, J.A., 1990. Effect of HNO3 treatment on the
solutions of non-electrolytes on solid (I) equation adsorption
surface acidity of activated carbons. Carbon 28, 675–682.
from solutions and the analysis of its simplest form, (II)
Olivier, J.P., 1995. Modeling physical adsorption on porous and
verification of the equation of adsorption isotherm from
nonporous solids using density functional theory. J. Porous
solutions. Izv. Akad. Nauk. SSSR, Otd. Khim. Nauk. 2, 209–216.
Mater. 2, 9–17.
Foo, K.Y., Hameed, B.H., 2009. An overview of landfill leachate
Ricordel, S., Taha, S., Cisse, I., Dorange, G., 2001. Heavy metals
treatment via activated carbon adsorption process. J. Hazard.
removal by adsorption onto peanut husks carbon:
Mater. 171 (1–3), 54–60.
characterization, kinetic study and modeling. Sep. Purif.
Freundlich, H., Helle, W.J., 1939. Ubber die adsorption in
Technol. 24 (3), 389–401.
lusungen. J. Am. Chem. Soc. 61, 2–28.
Ritchie, A.G., 1977. Alternative to the Elovich equation for the
Geankoplis, C.J., Leyva-Ramos, R., 1985. Model simulation and
kinetics of adsorption of gases on solids. J. Chem. Soc.,
analysis of surface diffusion of liquids in porous solids. Chem.
Faraday Trans. 1: Phys. Chem. Condens. Phases 73, 1650–1653.
Eng. Sci. 40 (5), 799.
Sekar, M., Sakthi, V., Rengaraj, S., 2004. Kinetics and equilibrium
Gerçel, O., Gerçel, H.F., 2007. Adsorption of lead (II) ions from
adsorption study of lead(II) onto activated carbon prepared
aqueous solutions by activated carbon prepared from biomass
from coconut shell. J. Colloid Interface Sci. 279 (2), 307–313.
plant material of Euphorbia rigida. Chem. Eng. J. 132, 289–297.
Shekinah, P., Kadirvelu, K., Kanmani, P., Senthilkumar, P.,
Gregg, S.J., Sing, K.S.W., 1982. Adsorption, Surface Area and
Subburam, V., 2002. Adsorption of lead(II) from aqueous
Porosity. Academic Press, London.
solution by activated carbon prepared from Eichhornia. J.
Ho, Y.S., Wase, D.A.J., Forster, C.F., 1996. Removal of lead ions
Chem. Technol. Biotechnol. 77 (4), 458–464.
from aqueous solution using sphagnum moss peat as
Sreejalekshmi, K.G., Krishnan, K.A., Anirudhan, T.S., 2009.
adsorbent. Water SA 22, 219–224.
Adsorption of Pb(II) and Pb(II)-citric acid on sawdust activated
Ho, Y.S., McKay, G., 1999. Pseudo second order model for sorption
carbon: kinetic and equilibrium isotherm studies. J. Hazard.
process. Process Biochem. 34, 451–465.
Mater. 161 (2–3), 1506–1513.
Ho, Y.S., 2006. Second order kinetic model for the sorption of
Thompson, G., Swain, J., Kay, M., Forster, C.F., 2001. The treatment
cadmium onto tree fern: a comparison of linear and nonlinear
of pulp and paper mill effluent: a review. Bioresour. Technol.
methods. Water Res. 40, 119–125.
77, 275–286.
http://code.google.com/p/ceres-solver/
.
Wang, L., Zhang, J., Zhao, R., Li, Y., Li, C., Zhang, C., 2010.
Ibrahim, H.S., Ammara, N.S., Soylak, M., Ibrahim, M., 2012.
Adsorption of Pb(II) on activated carbon prepared from
Removal of Cd(II) and Pb(II) from aqueous solution using dried
Polygonum orientale Linn.: kinetics, isotherms, pH, and ionic
water hyacinth as a biosorbent. Spectrochim. Acta, A: Mol.
strength studies. Bioresour. Technol. 101, 5808–5814.
Biomol. Spectrosc. 96, 413–420.
Weber Jr., W.J., Morris, J.C., 1963. Kinetics of adsorption on carbon
Inyang M., Gao, Bin, Yao, Y., Xue, Y., Zimmerman, A.R.,
from solution. J. Sanit. Eng. Div., Am. Soc. Civ. Eng. 89, 31–60.
Pullammanappallil, P., Cao, X., 2012. Removal of heavy metals

Das könnte Ihnen auch gefallen