w a t e r r e s e a r c h 5 1 ( 2 0 1 4 ) 1 1 3 e1 2 3

Available online at www.sciencedirect.com

ScienceDirect
journal homepage: www.elsevier.com/locate/watres

Boron removal by electrocoagulation and recovery

Mohamed Hasnain Isa a, Ezerie Henry Ezechi a, Zubair Ahmed b,*,
Saleh Faraj Magram b, Shamsul Rahman Mohamed Kutty a
a
b

Civil Engineering Department, Universiti Teknologi PETRONAS, 31750 Tronoh, Perak, Malaysia
Department of Civil Engineering, King Abdulaziz University, Jeddah, Saudi Arabia

article info

abstract

Article history:

This work investigated the removal of boron from wastewater and its recovery by elec-

Received 4 October 2013

trocoagulation and hydrothermal mineralization methods respectively. The experimental

Received in revised form

design was developed using Box-Behnken Model. An initial study was performed based on

9 December 2013

four preselected variables (pH, current density, concentration and time) using synthetic

Accepted 16 December 2013

wastewater. Response surface methodology (RSM) was used to evaluate the effect of pro-

Available online 27 December 2013

cess variables and their interaction on boron removal. The optimum conditions were obtained as pH 6.3, current density 17.4 mA/cm2, and time 89 min. At these applied optimum

Keywords:

conditions, 99.7% boron removal from an initial concentration of 10.4 mg/L was achieved.

Adsorption kinetics

The process was effectively optimized by RSM with a desirability value of 1.0. The results

Boron

showed that boron removal efficiency enhanced with increase in current density and

Electrocoagulation

treatment time. Removal efficiency also increased when pH was increased from 4 to 7 and

Hydrothermal mineralization

subsequently decreased at pH 10. Adsorption kinetics study revealed that the reaction

Produced water

followed pseudo second order kinetic model; evidenced by high correlation and goodness

Response surface methodology

of fit. Thermodynamics study showed that mechanism of boron adsorption was chemi-

Thermodynamics

sorption and the reaction was endothermic in nature. Furthermore, the adsorption process
was spontaneous as indicated by negative values of the adsorption free energy. Treatment
of real produced water using electrocoagulation resulted in 98% boron removal. The hydrothermal mineralization study showed that borate minerals (Inyoite, Takadaite and
Nifontovite) can be recovered as recyclable precipitate from electrocoagulation flocs of
produced water.
2013 Elsevier Ltd. All rights reserved.

1.

Introduction

Boron is an essential compound for the manufacture of

different products. Boron compounds are widely used in the
manufacture of glass, ceramics, high quality steel, catalysts,
cosmetics and flame retardants (Yilmaz et al., 2008a). Boron is
also an essential micronutrient for plants and is readily present
in the form of boric acid (H3BO3). Boron exists as undissociated

boric acid and borate ions in aquatic environment. The functions of boron in plants include degradation of carbohydrates,
sugar translocation, and hormonal action. Boron deficiency
causes stunted growth, yield loss and even death of plant
(Yilmaz et al., 2008b). High boron concentration in irrigation
water, however, can cause severe environmental problem
because boron compounds form complexes with heavy metals
present in soil and increase the potential toxicity of these

* Corresponding author. Tel.: 966 (0)2 6402000x68239; fax: 966 (0)2 6952179.
E-mail addresses: zkhan@kau.edu.sa, himatali@gmail.com (Z. Ahmed).
0043-1354/$ e see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.watres.2013.12.024

114

w a t e r r e s e a r c h 5 1 ( 2 0 1 4 ) 1 1 3 e1 2 3

complexes when passed to groundwater (Seki et al., 2006). High

boron concentration in surface water makes the water unfit for
consumption because boron has been shown to induce male
reproductive impediments in laboratory animals orally
exposed to boric acid and borax (Linder et al., 1990). The World
Health Organization (WHO, 2011) has set a guideline value for
boron concentration in drinking water as 2.4 mg/L.
Presence of boron in various industrial wastewaters, such
as produced water, can cause problems in wastewater reclamation and reuse. Produced water is water trapped in underground strata which is brought to the surface together with
oil and gas during drilling. Produced water is reportedly the
largest waste-stream of oil and gas exploration with an estimated 250 million barrels per day compared with about 80
million barrels per day of oil worldwide (Fakhruaul-Razi et al.,
2009). The composition of produced water differs from other
wastewater because produced water has been confined within
underground formations for a very long time (Ezechi et al.,
2012a). Produced water is being considered as a supplement
to limited freshwater resource especially in arid areas because
of its large production volume. One of the impediments to this
usage is the presence of boron at higher than permissible
concentrations. The large volumetric generation of produced
water would also suggest the potential for high amount of
boron recovery.
Boron removal from wastewater presents several challenges. Membrane process is a widely acceptable method for
wastewater treatment. However, studies have shown that
boron can diffuse through membranes in a non-ionic way,
similar to that of carbonic acid or water (Hou et al., 2010). The
use of selective ion exchange chelating resins has been shown
to be effective in boron removal (Kabay et al., 2004). Disposal of
the subsequently generated sludge and periodic regeneration
of resin, however, remain as major challenges. On the other
hand, conventional biological process only removes a small
amount of boron from wastewater due to its antiseptic nature
(Malakootian and Yousefi, 2009).
Electrocoagulation as a treatment process has been used in
the removal of various water contaminants. Process versatility, sludge reduction, minimal operator attention and ease
of operation are some of its advantages. The major action of
electrocoagulation depends on the ability of water particles to
respond to strong electric field in a redox reaction (Ezechi
et al., 2010b). Electrocoagulation involves three major mechanisms; formulation of coagulants by electrolytic oxidation of
sacrificial anodes, destabilization of the contaminants and
particulate suspension, breaking of emulsions and aggregation of the destabilized phases to form a floc (Babu et al., 2007).
The mechanism of aluminium oxidation during electrocoagulation is shown below (Balasubramanian et al., 2009).
Anode:
aq 3e

(1)

2H2 O l 2e / H2 g 2OH aq

(2)

Al s /Al

Cathode:

Aluminum forms polymeric species during oxidation of the

sacrificial anode. These polymeric species Al6(OH)3
15 ,

4
7
5
Al7(OH)4
17 , Al8(OH)20 , Al13O4(OH)24 , Al13(OH)34 , etc. transform
finally into Al(OH)3(s) according to the following simplified
equation (Ghosh et al., 2008).

Al

aq 3H2 O l/AlOH3 s 3H aq

(3)

The formed Al(OH)3 (s) appears as sweep flocs with large

surface area which increases its adsorption capacity and aids
in boron removal from solution. The formed flocs are separated from aqueous medium by sedimentation or flotation.
Considering that many landfill sites are filled up and
finding new landfill sites is difficult. Recovery of boron will not
only mitigate the adverse effect of boron in the environment
but also provide a means of producing boron compounds for
industrial use.
This study focuses on the use of electrocoagulation (EC) for
boron removal from aqueous solution and its recovery. The
specific objectives are: (a) to optimize EC removal of boron
based on significant operating parameters using response
surface methodology (RSM), (b) to study the boron adsorption
kinetics and thermodynamics, and (c) to determine the potential recovery of boron by hydrothermal mineralization (HM).

2.

Experimental method

2.1.

Characteristics of wastewater

A preliminary study was conducted with synthetic wastewater prepared with appropriate amount of boric acid (H3BO3)
and dissolved in 1 L distilled water to yield varying boron
concentrations of 10, 20 and 30 mg/L. The produced water was
collected from a local Crude Oil Terminal in Malaysia and was
characterized with atomic absorption spectrometry (AAS) and
ion chromatography (IC). A pH meter (Hach Sension 2 pH
meter) and a conductivity meter (Myron L conductivity meter)
were used to measure the pH and conductivity of the sample
respectively. The produced water characteristics are shown in
Table 1.

Table 1 e Produced water characteristics.

Parameter Concentrationa
Boron
pH

15
7.84

TSS
TDS
Conductivity
Turbidity
Aluminum

136
15,829
30,000 mS/cm
72 NTU
0.65

Iron
Chloride
Sodium
Calcium
Magnesium
Sulphate
Potassium
Hardness

1.66
7546
3952
357
600
168
284
957

Parameter

Concentration

Bromine
Total
phosphate
COD
BOD
Nitrite
Copper
Ammonia
nitrogen
TKN
Sulphate
Nitrate
Sulphide
Phenol
Total alkalinity
Zinc
Fluoride

31.2
12
1560
883
0.03
2.98
16.5
60.7
168
1.9
0.21
15
1546
0.04
0.61

All concentrations are expressed in mg/L unless stated

otherwise.

w a t e r r e s e a r c h 5 1 ( 2 0 1 4 ) 1 1 3 e1 2 3

2.2.

Electrocoagulation setup

The electrocoagulation setup consisted of a 500 mL beaker

with
six
aluminium
plate
electrodes
of
size
10 cm  1 cm  0.3 cm. Effects of different operating parameters (pH, current density, initial boron concentration and
treatment time) were investigated. The electrodes were connected to a digital DC supply characterized by the ranges of
0e3 A for current and 0e30 V for voltage. A digital ammetervoltmeter was used to regulate the current and voltage.
After each run of the experiments, the used aluminium electrodes were dipped in acetone solution for 10 min and rinsed
with deionized water and dried for 10 min at 105  C to remove
surface impurities before reuse. Samples were let stand for 2 h
to allow concentration of flocs prior to analysis for boron.
Boron concentration was analysed according to standard
method using carmine reagent. All results are average of three
analyses. Chemicals and reagents used were analytical grade
(Merck).

2.3.

Adsorption kinetics

Boron concentration for the adsorption study was varied in

the range 10e30 mg/L using a 500 mL beaker. Current density
12.5 mA/cm2, pH 7, temperature 308 K, and inter-electrode
spacing 0.5 cm were kept constant. Supernatant was
collected at different times and analyzed for residual boron
concentration. The amount of boron adsorbed at equilibrium
(qe) was calculated using the following equations:
qe C0  Ce

V
W

ITM
ZF

(4)

(5)

where C0 is initial boron concentration (mg/L), Ce is equilibrium boron concentration (mg/L), V is volume of sample (L), W
is mass of adsorbent (g), I is current (A), T is time (s), M is molar
mass of electrode, Z is number of electrons involved in the
redox reaction, and F is Faradays constant (C/mol).
The suitability of both pseudo first order and pseudo second order kinetic models was further evaluated using the chisquare (c2) represented as follows (Sundaram et al., 2008):


c2

exp

qe

 qcal
e
qcal
e

2
(6)

exp

where qe is experimental adsorption capacity at equilibrium

is calculated adsorption capacity at equilib(mg/g), and qcal
e
rium (mg/g).
The c2 test measures the goodness of fit between the
experimental equilibrium adsorption capacity and the calculated equilibrium adsorption capacity. The value of c2 for the
applicable model should be lowest. A good correlation coefficient and a low c2 indicate that the model is applicable.

2.4.

Adsorption thermodynamics

Thermodynamic parameters which include free energy

change (DG0), enthalpy of reaction (DH0) and entropy change
(DS0) can be used to deduce the mechanism of a reaction.

115

Observations were made at four different temperatures,

controlled using a thermostatic warm water bath, to determine these parameters. The thermodynamic constant was
evaluated using the following equations (Shen et al., 2009):
DG0  RTlnKc
0

ln Kc

DS
R

(7)

DH
RT

(8)

where Kc is distribution coefficient, R is thermodynamic gas

constant (8.314 J/mol.K), and T is temperature (K).

2.5.
(HM)

Boron recovery with hydrothermal mineralisation

The flocs produced during electrocoagulation were collected

after settling. They were transferred into an evaporating dish
and placed in the oven at room temperature for 24 h. The dried
flocs were kept in the desiccator for 20 min before grounded
and 2 g of the flocs were accurately weighed to recover boron.
40 mL of 3 M HNO3 was used to dissolve the flocs. The solution
was placed in an orbital shaker for one hour at 150 rpm to
enable complete dissolution. The pH of the solution was
adjusted to pH 10 with 1 M NaOH and 0.3 g calcium hydroxide
Ca(OH)2 was used as the mineralizer. Hydrothermal mineralization was conducted with a conventional oven for 2 h at
120  C. Thereafter, the solvent was collected for boron analysis and the precipitate was analysed using Simens diffractometer (Model D5000) with graphite monochromated Cu Ka
source operated at 40 kV and 40 mA. The XRD spectrum was
obtained at scanning angles (2q) ranging from 5 to 150 and at
scanning speed of 0.04 per second. The microstructure
properties were analyzed with scanning electron microscope
(SEM).

3.

Statistical methods and data analysis

Response surface methodology was employed to determine

the optimum levels of process parameters. RSM uses a
collection of mathematical and statistical techniques to
analyse the effects of several independent variables on the
response. It is often used in process design, improvement and
optimization. The methodology is practical as it employs
experimental data and thus includes the interactive effects of
variables on the overall process performance. Box-Behnken
design was established with the help of the Design Expert
6.0.7 software for statistical design of experiment and data
analysis. The four significant process variables (Yilmaz et al.,
2008b) considered in this study were: pH (A), current density
(B), initial boron concentration (C) and time (D) as shown in
Table 2. Synthetic wastewater was used to determine the
optimum treatment conditions. The total number of experiments in this study was 29 including five replicates at the
centre point for the estimation of error. A second-order
polynomial model (Equation (8)), using Design Expert software, was fitted to the experimental data obtained according
to the Box-Behnken design; where Y is the response, Xi and Xj
are variables, bo is a constant coefficient, bj, bjj and bij are
interaction coefficients of linear, quadratic and the second

116

w a t e r r e s e a r c h 5 1 ( 2 0 1 4 ) 1 1 3 e1 2 3

Table 2 e Independent variables of the Box-Behnken

design.

Table 3 e Response (boron removal) values for different

experimental conditions.

Level pH Current density

Initial boron
Treatment
(mA/cm2)
concentration (mg/L) time (min)

Run no. pH Current Initial boron Treatment

Boron
density conc. (mg/L) time (min) removal (%)
(mA/cm2)

1
0
1

4
7
10

6.25
12.5
18.75

10
20
30

30
60
90

order terms respectively, k is the number of studied factors

and ei is the error. The coded values of the process parameters
in Eq. (9) could be determined by Eq. (10) where Xi is the
dimensionless coded value of the ith independent variable, xi
is the uncoded value of the ith independent variable, x0 is the
uncoded value of the ith independent variable at the center
point and Dx is the step change value between low level (1)
and high level (1) (Zhang et al., 2011).
Y b0

k
X
j1

Xi

bj Xj

k
X
j1

bjj X2j

XX

bij Xi Xj ei

(9)

i <j

xi  x0
Dx

(10)

Analysis of variance (ANOVA) was used for analyses of the

data to obtain the interaction between the independent (process) variables and the dependent variable (response). The
quality of fit of the polynomial model was expressed by R2 and
its statistical significance was examined by the F-test. Model
terms were evaluated by the P-value (probability) with 95%
confidence level. Three dimensional (3D) plots were obtained
for boron removal. The response values for the different
experimental condition are shown in Table 3.

4.

Results and discussion

4.1.

Analysis of variance (ANOVA)

The analysis of variance (ANOVA) of regression parameters of

the predicted response surface quadratic model for boron
removal efficiency after excluding the insignificant model
terms are shown in Table 4. The model F-value of 42.25 and a
low probability value (Prob > F < 0.0001) indicate that the
model is significant. Values of Prob > F less than 0.05 indicate
that the model terms are significant. ANOVA results for the
response surface quadratic model are summarized in Table 5.
Adequate precision compares the range of the predicted
values at the design points to the mean prediction error. Its
value greater than 4 is desirable and confirms the applicability
of the model for navigation of the design space (Zinatizadeh
et al., 2007). The adequate precision of 23.088, in the present
case, shows that the model is acceptable. The R2 value of
0.9769 is in reasonable agreement with the model adjusted R2
value of 0.9538 and predicted R2 value of 0.8740. The agreement is desirable for a good fit of a model (Mohajeri et al.,
2010). The R2 value shows that the process can explain about
97% of the model output. A significant lack of fit suggests that
there may be some systematic variation unaccounted for in
the hypothesised model. The lack of fit in this study is not

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29

10
4
7
7
7
4
7
7
7
7
7
7
4
7
10
7
7
10
10
4
7
4
7
10
7
7
7
10
4

6.25
12.5
12.5
12.5
18.75
12.5
12.5
12.5
18.75
18.75
12.5
6.25
12.5
12.5
12.5
18.75
6.25
12.5
12.5
18.75
12.5
6.25
6.25
18.75
6.25
12.5
12.5
12.5
12.5

20
10
30
20
20
20
20
20
20
10
30
20
20
20
20
30
30
30
20
20
10
20
10
20
20
10
20
10
30

60
60
30
60
90
30
60
60
30
60
90
30
90
60
30
60
60
60
90
60
90
60
60
60
90
30
60
60
60

26
50
55
76
94
37.5
74
73
68
97
82.7
49.5
50
71.5
26.5
87.5
52
33.6
44
53
98
34
67
43.5
64
78
70
41
42.3

significant which is good for the model. In this study, A, B, C,

D, A2, C2 are significant model terms. Insignificant model
terms have limited influence on the model and were excluded.
Based on the results, the response surface model constructed
in this study for predicting boron removal efficiency was
considered reasonable. The final regression model (secondorder polynomial equation) in terms of coded factors is
expressed in Equation (11):
B removal% 72:90 4:35A 12:54B  6:49C 9:85D
 33:53A2 4:32C2

(11)

The suitability of the selected model to provide adequate

approximation of the real system is also confirmed by the
diagnostic plots. Such plots include normal probability plots of
the studentized residuals and the predicted versus actual

Table 4 e Analysis of variance result for significant model

terms.
Sum of DF Mean square F value Prob > F
squares
Boron
A
removal B
C
D
A2
C2

227.07
1887.52
505.70
1164.27
7292.14
120.40

1
1
1
1
1
1

227.07
1887.52
505.70
1164.27
7292.14
120.40

11.13
92.50
24.78
57.05
357.35
5.90

<0.0049
<0.0001
<0.0002
<0.0001
<0.0001
<0.0292

117

w a t e r r e s e a r c h 5 1 ( 2 0 1 4 ) 1 1 3 e1 2 3

Table 5 e ANOVA result for response surface quadratic model.

Boron removal

Model
Residual
Lack of fit
Pure error

Sum of squares

DF

Mean square

F value

Prob > F

12069.28
285.69
264.49
21.20

14
14
10
4

862.09
20.41
26.45
5.30

42.25

<0.0001

4.99

0.0676

Significant
Not significant

Standard deviation 4.52, R2 0.9769, R2 adj 0.9538, Pred R2 0.8740, Adequate precision 23.088, Press 1556.58.

value plot. These plots are used to judge the adequacy of a

model. Fig. 1 shows the normal probability plot for the studentized residuals for boron removal. Studentized residuals
represent normal probability plots where the residuals follow
a normal distribution in which case the points will follow a
straight line. Some scattering is expected even with normal
data. It can be deduced from Fig. 1 that the data is evenly
distributed. As shown in Fig. 2, the predicted and actual values
are in good agreement.

4.2.

Boron removal efficiency

The three dimensional (3D) response surface plots (Fig. 3) of

the quadratic model were generated by the Design Expert 6.0.7
software and utilized to assess the interactive effect of the
independent variables on the response. In Fig. 3(a), the 3D
response surface plot was developed as a function of initial pH
and current density at initial boron concentration 10 mg/L and
reaction time 90 min. From the plot, removal efficiency
increased from pH 4 to 7 and decreased towards pH 10. Between pH 7e8, removal efficiency was near constant. As
mentioned by (Bayramoglu et al., 2004), at pH 4e8, Al3 and
OH ions generated by electrodes react to form various
monomeric and polymeric species that finally transform into
insoluble amorphous Al(OH)3 (s) through complex polymerisation. Above pH 10, the highly soluble monomeric AlOH
4

Fig. 1 e Normal probability plot of the studentized

residuals.

anion concentration increases at the expense of Al(OH)3 (s).

However, the solubility of aluminum hydroxide is less at pH
6e8 (Emamjomeh and Sivakumar, 2009). In the present work,
the removal efficiency was highest (98%) at pH 7. Electrocoagulation acted as a pH neutralizer at alkaline pH in this
study. At pH 10, the final pH (8.8) was observed to be lower than
the initial pH. However, at initial pH 4, the final pH (4.4) was
observed to be higher than the initial pH. The increase in final
pH at acidic condition has been attributed to the increase in
hydrogen evolution at the cathode while the decrease in final
pH at alkaline condition has been attributed to the generation
of an alkalinity consumer (AlOH)4 (Vik et al., 1984).
The collision between particles, release of coagulants and
amount of coagulants generated at the electrode are
controlled by the electric current. From Fig. 3(a), removal efficiency increased when current density was increased. Increase in current density reduced treatment time. However,
ohmic heating at high current density increases sample
temperature, therefore it may not be feasible to increase
current density beyond 12.5 mA/cm2 in this study. In addition,
energy consumption increases with increased potential. The
optimum current density was observed as 12.5 mA/cm2.
In Fig. 3(b), the 3D response surface plot was developed as a
function of initial boron concentration and time at current
density 10 mA/cm2 and initial pH 8. Increase in initial boron
concentration was found to cause a decline in the removal
efficiency. This can be attributed to the fact that the amount of

Fig. 2 e Predicted versus actual values.

118

w a t e r r e s e a r c h 5 1 ( 2 0 1 4 ) 1 1 3 e1 2 3

corresponding to pH 6.3, current density 17.4 mA/cm2, initial

boron concentration 10.4 mg/L, and reaction time 89 min;
whereas a removal efficiency of 98% was obtained from the
experiment. It is, therefore, evident that the model is
adequate for prediction of boron removal using
electrocoagulation.

4.4.

Adsorption kinetics

To understand the adsorption kinetics of boron using electrocoagulation, four different adsorption models were
evaluated.

4.4.1.

Largegren pseudo first order kinetics

The linearized form of pseudo first order equation is shown

below (Boparai et al., 2011).





k1 t
log qe  qt log qe 
2:303

(12)

where qe and qt are the amount of boron adsorbed at equilibrium (mg/g) on Al(OH)3 and at any time (t) respectively, and k1
(min1) is the calculated pseudo first order rate constant of
adsorption.
If the adsorption follows the pseudo first order kinetics, a
plot of log (qe  qt) versus t should be linear. qe and k1 are
calculated from the intercept and slope of the plot of log
(qe  qt) versus t respectively. As shown in Fig. 4, the data did
not completely conform to a linear plot. The points deviated
from the straight line. Though the correlation coefficient was
high, the calculated equilibrium adsorption capacity qcal
e
deviated from the experimental equilibrium adsorption caexp
pacity qe and the chi-square was high. This implies that
the adsorption did not completely follow the pseudo first
order kinetics. The kinetic constants of pseudo first order kinetics is shown in Table 6.

4.4.2.

Pseudo second order kinetics

The linearized form of the pseudo second order equation is

represented below (Ho and McKay, 1998).
Fig. 3 e Three dimensional surface plots of boron removal.

t
1
t

qt k2 q2e qe

(13)

metal ions generated at the same current density for a low

boron concentration was insufficient for solutions of higher
boron concentration. It can be seen from Fig. 3(b) that increase
in time resulted in increase in removal efficiency, due to
prolonged interaction between the coagulant (Al(OH)3) and the
target constituent (boron).

4.3.

Optimization

Numerical conditions optimization for boron removal was

carried out using the Design Expert software. The desired goal
for each operational condition (pH, initial boron concentration, treatment time and current density) was chosen within
the range while the response (boron removal) was defined as
maximum to achieve the highest performance. The programme combines the individual desirabilities into a single
number, and then searches to maximize this function. The
model predicted boron removal efficiency of 99.7%

Fig. 4 e First order kinetic plot of different concentrations.

pH 7, current density 12.5 mA/cm2, Inter-electrode spacing
0.5 cm.

119

w a t e r r e s e a r c h 5 1 ( 2 0 1 4 ) 1 1 3 e1 2 3

Table 6 e Adsorption constants for first and second order kinetics.

Co (mg/L)
10
20
30

exp

qe

(mg/g)

6.09
11.22
16.68

Pseudo-first order
qcal
e (mg/g)

R2

c2

k2 (g/mg min)

qcal
e (mg.g)

R2

c2

0.078
0.029
0.037

4.45
6.56
14.06

0.918
0.898
0.957

0.62
3.31
0.49

0.017
0.084
0.037

6.80
11.90
17.86

0.998
0.99
0.989

0.08
0.04
0.08

where qe and qt are the amount of boron adsorbed at equilibrium (mg/g) on Al(OH)3 and at any time (t) respectively, and k2
is the calculated pseudo second order rate constant of
adsorption.
If the adsorption kinetics follows the pseudo second order,
the plot of t/qt versus t should be linear. The qe and k2 can be
calculated from the slope and intercept of plot of t/qt versus t.
As shown in Fig. 5, the plot of t/qt versus t was linear for all
concentrations studied. The correlation coefficient was high,
the chi-square was very low and the calculated equilibrium
adsorption capacity qcal
e was in agreement with the experiexp
mental equilibrium adsorption capacity qe . This implies
that the adsorption of boron followed a second order kinetics
as shown in Table 6.

4.4.3.

Intra-particle diffusion model

The linearized form of intra-particle diffusion equation is

represented below (Morris and Weber, 1963).
qt Kid t0:5 Ci

Pseudo-second order

k1 (min 1)

(14)

where Kid (mg/g min0.5) is a measure of the diffusion coefficient and Ci is the intra-particle diffusion constant (mg/g).
If intra particle diffusion is the rate limiting step, a plot of
fraction of solute adsorbed against the square root of time (t0.5)
should be linear, passing through the origin. Kid and Ci are
obtained from the slope and intercept of the graph respectively. Ci is directly proportional to the boundary layer thickness. From the result (Figure not shown), the plot of qt against
(t0.5) was observed to be linear but did not pass through the
origin. This could be due to the boundary layer effect. Additionally, the values of Ci (mg/g) for all boron concentrations

studied were positive. Kid (mg/g min0.5) and Ci (mg/g) of plot of

qt vs t0.5 were found to increase when the initial concentration
of boron was increased. This implies that intra particle
diffusion may not be the only transport mechanism, rather
there could be more than one mechanism involved in the
transport of boron. The adsorption constants obtained for
intra-particle diffusion are shown in Table 7.

4.4.4.

Elovich model

The modified form of Elovich equation is represented as

(Chien and Clayton, 1980).
qt

1
1
lnab lnt
b
b

(15)

where a is the initial adsorption rate (mg/g min) and b is the

desorption constant (g/mg) during any experiment.
The Elovich model does not predict any precise mechanism
but it is helpful in explaining predominantly, chemical
adsorption on highly heterogeneous adsorbents (Gupta and
Bhattacharyya, 2006). A plot of qt vs ln (t) should be linear.
The adsorption and desorption constants are calculated from
the slope and intercept of the plot of qt vs ln (t). When data was
fitted into the Elovich equation, the plot of qt vs ln (t) gave a
straight line (Figure not shown). The initial rate of adsorption
as calculated from the slope was found to increase and
desorption constant as calculated from the intercept was
observed to decrease as boron concentration was increased.
The correlation coefficient was also high. Increase in rate of
adsorption and decrease in desorption constant as concentration increases implies that the process is chemisorption.
The adsorption constant for Elovich model is shown in Table 7.

4.5.

Adsorption thermodynamics

The enthalpy change (DH0) and entropy (DS0) were calculated

from the slope and intercept of the plot of ln Kc versus 1/T
(Fig. 6). The thermodynamics constants obtained from the plot
of ln Kc versus 1/T is shown in Table 8. As shown in the Table,
the free energies were increasingly negative as temperature
was increased. Negative free energies (DG0) indicate that the

Table 7 e Adsorption constants for Intra-Particle

Diffusion and Elovich model.
Co
Intra-particle diffusion
Elovich model
(mg/L)
2
Ci
R a (mg/g min) b (g/mg) R2
ki
(mg/g min0.5) (mg/g)
Fig. 5 e Second order kinetic plot of different
concentrations. pH 7, current density 12.5 mA/cm2, interelectrode spacing 0.5 cm.

10
20
30

0.36
0.59
1.28

3.132 0.97
4.809 0.98
5.017 0.98

1.06
1.72
3.77

1.55
0.99
0.499

0.98
0.94
0.97

120

w a t e r r e s e a r c h 5 1 ( 2 0 1 4 ) 1 1 3 e1 2 3

adsorption was spontaneous for the temperature range evaluated and the degree of spontaneity of the reaction increases
with increasing temperature (Li et al., 2010). The positive value
of the standard enthalpy change (DH0 44.8 kJ/mol) implies that
the adsorption process is endothermic. Values of enthalpy
change less than 40 kJ/mol indicates that the process is
physiosorption while values above 40 kJ/mol indicate a
chemisorption process (Ma et al., 2011). Accordingly, in the
present case, the process is chemisorption. The positive value
of entropy change (DS0 125.04 J/mol) indicates increased
randomness of the solution interface during the adsorption of
boron on the electrocoagulant Al(OH)3.

4.6.

Process advantages and operational cost

Process evaluation and operational cost are two important

indices which determine the implementation of a technique.
In analysing these two indices, factors such as other treatment methods, environmental and health implications, electrical energy consumption, electrode consumption, use of
chemicals and sludge disposal are considered. One comparison of interest to this study is chemical coagulation which
follows the same pollutant removal mechanism as electrocoagulation. Their major difference is the mode of introduction of coagulants. Whereas coagulants are continuously
generated over an extended area of the anode material in
electrocoagulation, point addition of coagulants is done with
chemical coagulation. The freshly precipitated flocs generated
in electrocoagulation are more effectively dispersed resulting
in increased adsorptive removal of pollutants (Zhu et al.,
2005). In terms of floc separation, chemical coagulation is
associated with settling and electrocoagulation is characterised by both settling and flotation due to air bubbles
released at the cathode (Holt et al., 2002). During electrocoagulation, the smallest colloidal particles have a higher
probability of coagulation because the electric field sets them
in motion and produces a relatively low amount of sludge
compared to chemical coagulation (Pouet and Grasmick,
1995). Additionally, secondary pollution is mitigated by electrocoagulation because the rate of generation of coagulants is
regulated by the voltage and current while secondary pollution may occur at high chemical addition using chemical

Table 8 e Thermodynamic constants for boron

adsorption.
Temperature (K)
298
308
318
328

DG
(kJ/mol)

3.60
6.16
8.75
12.86

3.18
4.66
5.73
6.96

DH
(kJ/mol)

DS (J/mol.K)

44.8

125.04

coagulation (Yildiz et al., 2008). Electrocoagulation sludge is

readily settleable and easy to de-water since it is composed of
mainly metallic oxides/hydroxide and its flocs tend to be
much larger, contain less bound water, have acid resistant
capacity, are more stable and are easily separated by filtration
(Avsar et al., 2007). Chemical coagulation is highly sensitive to
pH change with effective coagulation at pH 6e7 while electrocoagulation has a pH neutralization effect in a much wide
pH range (4e9). In literature, comparative reports favour
electrocoagulation over chemical coagulation (Avsar et al.,
2007; Zhu et al., 2005).
The overall cost was investigated to determine the feasibility of electrocoagulation compared to other processes. In
evaluating the overall cost, the following equation was used
(Olmez-Hanci et al.).
OPCost a  Cenergy b  Celectrode c  Csludge

(16)

where
Cenergy and Celectrode are amount of consumed electricity
(kWh/m3) and amount of consumed electrode material (kg
electrode/m3) respectively. Csludge (kg/m3) is the amount of
sludge generated during electrocoagulation. Unit prices used
in this study (for Malaysia market) were expressed as (a) unit
prices for electrical energy 35 cents Malaysia Ringgits (MYR)/
kWh (Tenaga National Malaysia, 2013), (b) unit price for electrode material (Bayramoglu et al., 2004) ($1.8; current exchange in MYR) RM 5.8/kg (c) Cost associated with sludge
handling and disposal 50 cents MYR/kg (MIDA, 2013).
The electrical energy consumption (Cenergy) is an important
parameter which defines the energy used up for a process. The
electrical energy consumptions at a constant potential of 2 V
was 1.2 kWh/m3, 2.4 kWh/m3 and 3.6 kWh/m3 for 6.25 mA/
cm2, 12.5 mA/cm2 and 18.75 mA/cm2 respectively as calculated using the equation below. Increase in current density
increased the electrical energy consumption and also
increased the overall cost.
ECC

Fig. 6 e ln Kc vs 1/T (K); pH 7, initial boron concentration

20 mg/L, current density 12.5 mA/cm2, inter-electrode
spacing 0.5 cm.

Kc

IUT
V

(17)

where
ECC energy consumption (kWh/m3); I current (A);
U voltage (V); T time (h); V volume (L)
The aluminum electrode consumption (AI consumption)
with a unit of (g Al/g of Boron) removed was investigated using
the Faraday law as expressed below. The electrode consumption increased from 0.4 g Al/g at 6.25 mA/cm2 to 0.81 g Al/
g at 12.5 mA/cm2. At 18.75 mA/cm2, there was further increase
in aluminum electrode consumption to 1.22 g Al/g making
current density and the corresponding voltages important
controls of operational cost.

121

w a t e r r e s e a r c h 5 1 ( 2 0 1 4 ) 1 1 3 e1 2 3

Al consumption

ItM
ZFW

(18)

where
I current in amperes (A); t time (s); M molecular
weight of the electrode material (Aluminum 26.98);
Z number of electrons involved in the redox reaction
(zAl 3); F Faraday constant (96,500 C/mol); W weight of
treated wastewater (g).
The amount of sludge generated at 6.25 mA/cm2, 12.5 mA/
cm2 and 18.75 mA/cm2 was 0.037 kg, 0.029 kg and 0.021 kg
respectively. Sludge generation is influenced by electric
charge. The unit cost of NaOH (50%) and H2SO4 (98%) used in
this study was about 0.018 cents MYR/kg. 0.3e0.5 ml of 1 M
NaOH and H2SO4 was used to control the pH of each experiment. The operational cost for the removal of boron in this
study at optimum conditions is RM 0.88/m3 as depicted in
Table 9. Electrical energy consumption was the largest
contributor to the total operational cost for boron removal.
The unit cost for the electrodialytic treatment of boron
containing wastewater was estimated at $1.27/m3 for a twostep process removing boron and salinity (Turek et al., 2007)
and MYR 6.33/m3 for adsorption-flocculation (Chong et al.,
2009). With an operational cost of MYR 0.88/m3 at optimum
condition, electrocoagulation has comparative advantage in
terms of cost over some boron removal processes.
In large installations, the operation is feasible using an
electrocoagulation reactor equipped with a secondary sedimentation tank. Installation of parallel perforated rhombus
shaped electrode materials into the reactor will enhance
movement of metal ions, improve coagulation, reduce the
number of electrodes, improve hydrogen bubble generation
and increase floatation of formed flocs. However, residual
aluminum concentration should be controlled by operating at
low electric potential and current. This could increase treatment time but will not affect operational cost because the
increased time is compensated by the low electrical energy
consumption (kWh/m3) and electrode consumption.

4.7.

Boron recovery

Real produced water containing 15 mg/L boron was used for

this phase of the study. Electrocoagulation treatment of the
produced water at pH 7, current density 12.5 mA/cm2, and
reaction time 90 min yielded a boron removal efficiency of
98%. The flocs thus produced were studied for boron recovery.
The electrocoagulation flocs were characterized with X-ray
fluorescence (XRF) and analysed with X-ray diffractogramme
(XRD) and Scanning Electron Microscope (SEM) during the
recovery of boron as a recyclable precipitate. The chemical

Fig. 7 e SEM result for floc precipitate after hydrothermal

mineralization at 120 C, 2 h, 0.3 g Ca (OH)2, 2 g
electrochemical coagulation flocs, pH 10.

composition of the flocs was obtained to be 2.9% B2O3, 1.3%

Al2O3, 1.09% Fe2O3, 14.3% CaO, 19.2% MgO, 28.4% NaO, 6.4%
SiO2, 2.4% K2O, and 24.01% loss on ignition. After hydrothermal mineralization of EC flocs, the results show about 1.1 mg/
L residue boron concentration indicating about 91.4% recovery. The SEM analysis of the obtained precipitates showed a
fibrous root like structure on an irregular beam shape basement as depicted in Fig. 7.
XRD investigation of the fibrous root like structure on an
irregular beam shape basement is shown in Fig. 8. The graph
showed a bragg reflections possessing broad humps and low
intensity which indicates that the analyzed phase is a short
range i.e. more amorphous and little crystalline. The chemical
speciation of this amorphous phase can be aluminum hydroxide or aluminum oxyhydroxide. This is suspected
because crystallization of Al hydroxides or oxyhydroxides is a
very slow process. It is reported that most Al hydroxides and
Al oxyhydroxides are found to be either poorly crystalline or
amorphous (Dixon and Weed, 1989).
Using Eva DiffraPlus indexing software in combination with
ICDD (International Center for Diffraction Data) database,
Inyoite
Nifontovite
(Ca3B6O6(OH)12$2(H2O),
(CaB3O3(OH)5$4H2O) and Takedaite (Ca3B2O6$2H2O) were
identified in the flocs. The three identified compounds are
hydrated calcium borate minerals. From the XRD shown in
Fig. 8, it can be concluded that Nifontovite could be the fibrous
root like structure identified from the SEM result while Inyoite
and Takadaite could be the irregular beam shape basement.

Table 9 e Operational cost analysis for electrocoagulation.

Item

Energy consumption
Electrode plate consumption
Chemicals
Sludge disposal
Total

Cost (MYR/m3)

Operational parameter values

2

6.25 mA/cm

12.5 mA/cm

18.75 mA/cm

6.25 mA/cm

12.5 mA/cm2

18.75 mA/cm2

1.2 kWh/m3
0.4 g Al/g
10 ml
0.037 kg

2.4 kWh/m3
0.81 g Al/g
10 ml
0.029 kg

3.6 kWh/m3
1.22 g Al/g
10 ml
0.021 kg

0.42
0.0024
0.018
0.006
0.45

0.84
0.0049
0.018
0.014
0.88

1.26
0.0073
0.018
0.019
1.3

122

w a t e r r e s e a r c h 5 1 ( 2 0 1 4 ) 1 1 3 e1 2 3

PETRONAS (UTP), for graduate assistantship to the second

author. The authors, therefore, acknowledge with thanks for
the technical and financial support of DSR, KAU and UTP.

references

Fig. 8 e XRD pattern after hydrothermal mineralization at

120 C, 2 h, 0.3 g Ca (OH)2, 2 g electrochemical coagulation
flocs, pH 10.

The amorphous and poor crystalline nature of aluminium

flocs has been reported in literature. It is reported that
aluminum produces poor crystalline precipitates in the presence of natural organic matter (Masion et al., 1994). This could
be the reason why the XRD result is more amorphous since
produced water contains high concentration of organic
compounds.

5.

Conclusions

Presence of boron in produced water is one of the reasons that

hinder its reuse for purposes such as irrigation and drinking.
The electrochemical technique employed in this study was
able to reduce boron concentration to below the WHO
permissible level of 2.4 mg/L. Response surface methodology
was successfully applied to optimise boron removal from
aqueous solution. At optimal operating conditions of pH 7,
current density 12.5 mA/cm2, inter-electrode spacing 0.5 cm,
and treatment time 90 min, 98% removal of boron from produced water was achieved. The adsorption of boron followed
the pseudo second order rate kinetics. The thermodynamics
study revealed that the adsorption is chemisorption and
endothermic. The adsorption process showed an increased
dispersal of particles in the solution and the adsorption was
spontaneous. Attempt to recover boron as a recyclable precipitate from electrocoagulation flocs revealed that Inyoite,
Takedaite and Nifontovite can be recovered through hydrothermal mineralization of the electrocoagulation flocs. However, selective recovery of individual boron compounds from a
mixture of produced water electrocoagulation flocs, duration
of flocs settling, and flocs homogeneity is still a subject of
further investigation.

Acknowledgement
This work was funded by Deanship of Scientific Research
(DSR), King Abdulaziz University (KAU), Jeddah, under
research grant (no. 135-006-D1434) and Universiti Teknologi

Avsar, Y., Kurt, U., Gonullu, T., 2007. Comparison of classical

chemical and electrochemical processes for treating rose
processing wastewater. J. Hazar. Mater 148 (1e2), 340e345.
Babu, R.R., Bhadrinarayana, N.S., Sheriffa Begum, K.M.M.,
Anantharaman, N., 2007. Treatment of tannery wastewater by
electrocoagulation. J. Environ. Sci. (China) 19 (12), 1409e1415.
Balasubramanian, N., Kojima, T., Srinivasakannan, C., 2009.
Arsenic removal through electrocoagulation: kinetic and
statistical modeling. Chem. Eng. J. 155 (1e2), 76e82.
Bayramoglu, M., Kobya, M., Can, O.T., Sozbir, M., 2004. Operating
cost analysis of electrocoagulation of textile dye wastewater.
Separ. Purif. Technol. 37 (2), 117e125.
Boparai, H.K., Joseph, M., OCarroll, D.M., 2011. Kinetics and
thermodynamics of cadmium ion removal by adsorption
onto nano zerovalent iron particles. J. Hazar Mat. 186 (1),
458e465.
Chien, S.H., Clayton, W.R., 1980. Application of Elovich equation
to the kinetics of phosphate release and sorption in Soils1. Soil
Sci. Soc. Am. J. 44 (2), 265e268.
Chong, M.F., Lee, K.P., Chieng, H.J., Syazwani Binti Ramli, I.I., 2009.
Removal of boron from ceramic industry wastewater by
adsorption flocculation mechanism using palm oil mill boiler
(POMB) bottom ash and polymer. Water Res. 43 (13),
3326e3334.
Dixon, J.B., Weed, S.B., 1989. Soil Science Society of America Inc.
(SSSA).
Emamjomeh, M.M., Sivakumar, M., 2009. Review of pollutants
removed by electrocoagulation and electrocoagulation/
flotation processes. J. Environ. Manag. 90 (5), 1663e1679.
Ezechi, E.H., Isa, M.H., Kutty, S.R.M., 2012a. Boron in produced
water: challenges and improvements: a comprehensive
review. J. Appl. Sci. 12 (5), 402e415.
Ezechi, E.H., Isa, M.H., Kutty, S.R.M., 2010b. Removal of boron
from produced water by electrocoagulation. In: 10th WSEAS
International Conference on Environment, Ecosystems and
Development (EED12), Switzerland.
Fakhruaul-Razi, A., Pendashteh, A., Abdullah, L.C., Biak, D.R.A.,
Madaeni, S.S., Abidin, Z.Z., 2009. Review of technologies for oil
and gas produced water treatment. J. Hazar. Mater 170 (2aV3),
530e551.
Ghosh, D., Solanki, H., Purkait, M.K., 2008. Removal of Fe(II) from
tap water by electrocoagulation technique. J. Hazard Mater 155
(1), 135e143.
Gupta, S.S., Bhattacharyya, K.G., 2006. Adsorption of Ni(II) on
clays. J. Colloid Interface Sci. 295 (1), 21e32.
Ho, Y.S., McKay, G., 1998. A comparison of chemisorption kinetic
models applied to pollutant removal on various sorbents. Proc.
Saf. Environ. Prot. 76 (4), 332e340.
Holt, P.K., Barton, G.W., Wark, M., Mitchell, C.A., 2002. A
quantitative comparison between chemical dosing and
electrocoagulation. Coll. Surf. Physico. Eng. Asp. 211 (2e3),
233e248.
Hou, D., Wang, J., Sun, X., Luan, Z., Zhao, C., Ren, X., 2010. Boron
removal from aqueous solution by direct contact membrane
distillation. J. Hazard Mater 177 (1e3), 613e619.
Kabay, N., Yilmaz, I., Yamac, S., Samatya, S., Yuksel, M.,
Yuksel, U., Arda, M., Saglam, M., Iwanaga, T., Hirowatari, K.,
2004. Removal and recovery of boron from geothermal
wastewater by selective ion exchange resins. I. Laboratory
tests. React. Funct. Poly. 60 (0), 163e170.

w a t e r r e s e a r c h 5 1 ( 2 0 1 4 ) 1 1 3 e1 2 3

Li, Q., Yue, Q.-Y., Su, Y., Gao, B.-Y., Sun, H.-J., 2010. Equilibrium,
thermodynamics and process design to minimize adsorbent
amount for the adsorption of acid dyes onto cationic polymerloaded bentonite. Chem. Eng. J. 158 (3), 489e497.
Linder, R.E., Strader, L.F., Rehnberg, G.L., 1990. Effect of acute
exposure to boric acid on the male reproductive system of the
rat. J. Toxicol. Environ. Health 31 (2), 133e146.
Ma, J., Jia, Y., Jing, Y., Yao, Y., Sun, J., 2011. Kinetics and
thermodynamics of methylene blue adsorption by cobalthectorite composite. Dyes Pigments 93 (1e3), 1441e1446.
Malakootian, M., Yousefi, N., 2009. The efficiency of
electrocoagulation process using aluminum electrodes in
removal of hardness from water. Iran. J. Environ. Health Sci.
Eng. 6 (2), 131e136.
Masion, A., Thomas, F., Tchoubar, D., Bottero, J.Y., Tekely, P.,
1994. Chemistry and structure of Al(OH)/Organic precipitates.
A small-Angle x-ray scattering study. 3. Depolymerization of
the Al13 polycation by organic ligands. Langmuir 10 (11),
4353e4356.
Malaysian Investment Development Authority (MIDA), 2013.
Scheduled Waste Treatment Rates. Available: http://www.
mida.gov.my/env3/index.php?pagescheduled-wastetreatment-rates.
Mohajeri, L., Aziz, H.A., Isa, M.H., Zahed, M.A., 2010. A statistical
experiment design approach for optimizing biodegradation of
weathered crude oil in coastal sediments. Biores. Technol. 101
(3), 893e900.
Morris, W., Weber, W., 1963. Kinetics of adsorption on carbon
from solution. J. Sanit. Eng. Div. 89 (2), 31e60.
Olmez-Hanci, T., Kartal, Z., Arslan-Alaton, I. Electrocoagulation of
commercial naphthalene sulfonates: process optimization
and assessment of implementation potential. J. Environ.
Manag. 99(0), 44e51.
Pouet, M.F., Grasmick, A., 1995. Urban wastewater treatment by
electrocoagulation and flotation. Water Sci. Technol. 31 (3e4),
275e283.
Seki, Y., Seyhan, S., Yurdakoc, M., 2006. Removal of boron from
aqueous solution by adsorption on Al2O3 based materials
using full factorial design. J. Hazard Mater 138 (1), 60e66.
Shen, X.E., Shan, X.Q., Dong, D.M., Hua, X.Y., Owens, G., 2009.
Kinetics and thermodynamics of sorption of nitroaromatic

123

compounds to as-grown and oxidized multiwalled carbon

nanotubes. J. Colloid Interface Sci. 330 (1), 1e8.
Sundaram, C.S., Viswanathan, N., Meenakshi, S., 2008.
Defluoridation chemistry of synthetic hydroxyapatite at nano
scale: equilibrium and kinetic studies. J. Hazard Mater 155
(1e2), 206e215.
Tenaga National Malaysia, 2013. Pricing and Tariff for Electricity.
Turek, M., Dydo, P., Trojanowska, J., Bandura, B., 2007.
Electrodialytic treatment of boron-containing wastewater.
Desalination 205 (1e3), 185e191.
Vik, E.A., Carlson, D.A., Eikum, A.S., Gjessing, E.T., 1984.
Electrocoagulation of potable water. Water Res. 18 (11),
1355e1360.
World Health Organization (WHO), 2011. Guidelines for Drinkingwater Quality. Available: http://whqlibdoc.who.int/
publications/2011/9789241548151_eng.pdf.
Yildiz, Y.I., Koparal, A.S., Keskinler, B., 2008. Effect of initial pH
and supporting electrolyte on the treatment of water
containing high concentration of humic substances by
electrocoagulation. Chem. Eng. J. 138 (1e3), 63e72.
 lu, R., KocakerUm, M.M.,
Yilmaz, A.E., Boncukcuog
 istan, E., 2008a. An empirical model for kinetics of
Kocadag
boron removal from boron containing wastewaters by the
electrocoagulation method in a batch reactor. Desalination
230 (1e3), 288e297.
Yilmaz, A.E., Boncukcuoglu, R., Kocakerim, M.M., Yilmaz, M.T.,
Paluluoglu, C., 2008b. Boron removal from geothermal waters
by electrocoagulation. J. Hazard Mater 153 (1e2), 146e151.
Zhang, H., Ran, X., Wu, X., Zhang, D., 2011. Evaluation of electrooxidation of biologically treated landfill leachate using
response surface methodology. J. Hazard Mater 188 (1e3),
261e268.
Zhu, B., Clifford, D.A., Chellam, S., 2005. Comparison of
electrocoagulation and chemical coagulation pretreatment for
enhanced virus removal using microfiltration membranes.
Water Res. 39 (13), 3098e3108.
Zinatizadeh, A.A.L., Mohamed, A.R., Mashitah, M.D.,
Abdullah, A.Z., Isa, M.H., 2007. Optimization of pre-treated
palm oil mill effluent digestion in an up-flow anaerobic sludge
fixed film bioreactor: a comparative study. Biochem. Eng. J. 35
(2), 226e237.

All in-text references underlined in blue are linked to publications on ResearchGate, letting you access and read them immediately.