See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/5854078

Production of activated carbon from coconut shell: Optimization using

response surface methodology

Article  in  Bioresource Technology · August 2008

DOI: 10.1016/j.biortech.2007.09.042 · Source: PubMed

CITATIONS READS

189 6,208

5 authors, including:

Animesh Dutta
University of Guelph
143 PUBLICATIONS   2,766 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

OMAFRA_UOG View project

CIDA Project View project

All content following this page was uploaded by Animesh Dutta on 19 March 2019.

The user has requested enhancement of the downloaded file.

 Available online at www.sciencedirect.com

Bioresource Technology 99 (2008) 4887–4895

Production of activated carbon from coconut shell: Optimization

using response surface methodology
M.K.B. Gratuito a, T. Panyathanmaporn b, R.-A. Chumnanklang b,
N. Sirinuntawittaya b, A. Dutta a,*
a
Energy Field of Study, School of Environment, Resources and Development, Asian Institute of Technology, P.O. Box 4, Klongluang,
Pathumthani 12120, Thailand
b
National Metal and Materials Technology Center (MTEC), 114 Thailand Science Park, Paholyothin Road, Klongluang, Pathumthani 12120, Thailand

Received 16 May 2007; received in revised form 11 September 2007; accepted 15 September 2007
Available online 13 November 2007

Abstract

The production of activated carbon from coconut shell treated with phosphoric acid (H3PO4) was optimized using the response
surface methodology (RSM). Fifteen combinations of the three variables namely; impregnation ratio (1, 1.5, and 2); activation time
(10, 20, and 30 min); and activation temperature (400, 450, and 500 C) were optimized based on the responses evaluated (yield, bulk
density, average pore diameter, small pore diameter, and number of pores in a unit area). Pore diameters were directly measured from
scanning electron microscope (SEM) images. Individual second-order response surface models were developed and contour plots were
generated for the optimization analysis. The optimum range identified for impregnation ratio was from 1.345 to 2, while for the activa-
tion time was from 14.9 to 23.9 min. For the activation temperature it was from 394 to 416 C. The optimum points are 1.725, 19.5 min,
and 416 C, respectively. The models were able to predict well the values of the responses when the optimum variable parameters were
validated as proven by the generally acceptable values of the residual percentages. Direct characterization of the pores using the SEM
was found to be a good technique to actually see the pores and get actual measurements. Additionally, RSM has also proven to be a good
tool in optimization analysis to get not only optimum production condition points but ranges, which are crucial for the flexibility of the
production process, as well.
 2007 Elsevier Ltd. All rights reserved.

Keywords: Activated carbon; Chemical activation; Scanning electron microscope; Optimization; Response surface methodology

1. Introduction of money for procurement. Furthermore, besides being

Activated carbon can be produced from different raw
carbon resources like lignite, peat, coal, and biomass
resources such as wood, sawdust, bagasse, and coconut
shells (Ioannidou and Zabaniotou, 2006). However, the
abundant supply of coconut shell as a waste-product from
the coconut oil and desiccated coconut industry makes pro-
duction of activated carbon from this material more finan-
cially viable since using grain or coal as raw materials for
activated carbon will require manufacturers extra amount
*
Corresponding author. Tel.: +662 524 5403; fax: +662 524 5439.
E-mail address: duttaa@ait.ac.th (A. Dutta).
an amorphous form of carbon that can absorb many gases,
vapors, and colloidal solids, coconut shell activated
carbons are advantageous over carbons made from other
materials because of its high density, high purity, and vir-
tually dust-free nature. These carbons are harder and more
resistant to attrition.
Production of activated carbon can either be through
physical or chemical activation. In physical activation,
the material is carbonized under inert atmosphere and
then activated at high temperature using either steam or
carbon dioxide as the activating reagent while in chemical
activation, the precursor is treated with chemicals to
help with the initial dehydration. In most cases, chemical

0960-8524/$ - see front matter  2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biortech.2007.09.042
4888 M.K.B. Gratuito et al. / Bioresource Technology 99 (2008) 4887–4895
activation is preferred over physical activation as the latter
generally results to lower yields due to the mass loss asso-
ciated with oxidation (Toles et al., 2000). In addition,
chemical activation, which is commonly used for biomass
precursor mainly because it achieves higher yield and larger
surface areas, also requires lower operating and energy
costs as lower temperatures are used.
However, like in most production processes, the produc-
tion of quality activated carbon involves balancing the pro-
duction conditions to get the desired characteristics of the
output. More often than not, this balancing becomes com-
plicated as there are more than one characteristic that has
to be considered. Also, the desirable characteristics not
only refer to the requirements of the end-users such as high
surface area, appropriate porosity, and high bulk density
but also refer to the producer side. Higher yield obtained
at lower operating and energy cost is always desired but
careful balancing is a must to avoid overlooking other
characteristics. The goal of this study was to find the opti-
mum production conditions for making activated carbon
from coconut shell by simultaneously considering the
impregnation ratio, activation time, and activation temper-
ature. Desirable production outputs based on yield and
bulk density were considered as responses. Furthermore,
direct measurements of the pores using images from the
scanning electron microscope (SEM) were also taken as
responses.
As far as known to the authors, no study has been done
on optimization of the production of activated carbon
from coconut shells using the response surface methodol-
ogy (RSM) approach. Though RSM is a popular tool in
process optimization, its application in activated carbon
production is very rare. RSM has just recently been used
for the optimization of coconut husk (Tan et al., 2007),
Turkish lignite (Karacan et al., 2007) and olive-waste cakes
(Bacaoui et al., 2001). Additionally, the use of SEM images
for direct pore size measurements and the inclusion of its
results as a response for RSM are also new and unexplored.
2. Important parameters for activation
The succeeding sections will explain the complexity of
balancing the production conditions for the chemical acti-
vation as each independent variable has its own effect on
both physical and adsorption characteristics of carbons
as well as in production outputs such as yield and bulk
density.
2.1. Chemical/impregnation ratio
In chemical activation process, it is well known that the
impregnation ratio, the ratio of the weights of the chemical
agent and the dry precursor, is one of the variables that
have major effect on the characteristics of the final carbons
produced. The chemical agents used are dehydrating agents
that penetrate deep into the structure of the carbon causing
tiny pores to develop. Thus, aside from affecting the devel-
opment of the pores, particularly the size, it also affects the
resulting surface area as generally smaller pores will result
to larger surface areas. The surface area of untreated Arun-
do canes was observed to increase from 38 to 1151 m2/g
after chemical treatment using phosphoric acid (Verners-
son et al., 2002). The general trend for all precursors is that
as concentration is increased the surface area also increases
though an optimum concentration is evident in most cases.
Larger pores, which correspond to smaller surface area,
develop as more acids are used. When pores reach a partic-
ular size (in the range from mesopore to macropore), they
do not contribute to the surface area significantly (Lázaro
et al., 2007). For peach stones, it was found that an
increase of impregnation ratio resulted to an increase in
the volumes of micropores and mesopores (Molina-Sabio
et al., 1995) but for powdered peanut hulls the maximum
area was attained only at an impregnation ratio of 1 (Girgis
et al., 2002). Furthermore, only a 35% phosphoric acid
solution was determined ideal for sorghum as it was best
both for surface area and porosity development (Diao
et al., 2002). The degree of weakening of the sorghum grain
structure was also minimized, thus the hardness and bulk
density of the resulting product were not compromised.
Progressive development of porosity, especially of larger
pores, occur as more acid is incorporated until a limit is
reached, of which a larger excess leads to collapsed pores
likely due to structural weakness caused by the intensified
dilation. High ratios led to a reduction in pore volumes
and a marked decrease in surface area for Arundo cane
(Vernersson et al., 2002).
2.2. Activation duration
The duration of the activation has a significant effect on
the development of the carbon’s porous networks. The
time should just be enough to eliminate all the moisture
and most of the volatile components in the precursor to
cause pores to develop. Since the end of the volatile evolu-
tion marks the formation of the basic pore structure, acti-
vation should be limited up to that point. Longer durations
cause enlargement of pores at the expense of the surface
area. Also, the control of the activation time is of economic
interest since shorter times are generally desired as it
equates to reduction in the energy consumption. In a study
conducted for grain sorghum by Diao et al. (2002), initial
experiments of activation at 450 C resulted to smoke evo-
lution ceasing only after 8 min thus the minimum activa-
tion time was set at 10 min and 15 min was observed to
be the optimum for porosity development (Diao et al.,
2002). For the case of Arundo cane, zero activation time
which actually means that the sample has just reached
the desired activation temperature but was already exposed
to heat progression while attaining the desired temperature
yielded the highest surface area and micropore volume
(Vernersson et al., 2002). Maximum surface area was
attained at 45 min for the activation of rubber wood
sawdust and was reported to decrease with further increase
 M.K.B. Gratuito et al. / Bioresource Technology 99 (2008) 4887–4895
in activation time (Srinivasakannan and Zailani Abu
Bakar, 2004). Yield also appeared to be dependent on the
activation duration as it dropped upon reaching an opti-
mum point. Low activation time resulted to an incomplete
burn-off thus resulting in higher yields.
2.3. Activation temperature
The application of heat to an impregnated material fur-
ther accelerates the thermal degradation and the volatiliza-
tion process. This leads to development of pores, increase
of surface area and the subsequent mass loss. The selection
of the activation temperature is based on several factors
which include the type of precursor and the chemical agent
used. Activation temperature for different biomass precur-
sors range from 400 to 800 C (Diao et al., 2002) while for
coal-based materials can go as high as 900 C (Karacan
et al., 2007).
The optimum activation temperature for higher surface
area was found to be 450 C for coconut shells impregnated
with phosphoric acid (Laine et al., 1989) and 500 C for
rubber wood sawdust (Srinivasakannan and Zailani Abu
Bakar, 2004), though yield was at the lowest for the latter.
For impregnated acorns, 800 C offered the highest adsorp-
tion capacity as reported by Lafi (2001). Temperatures
lower than 500 C for grain sorghums produced micropo-
rous carbons but with small surface areas while tempera-
tures higher than 600 C yielded mesoporous carbons
with high surface areas (Diao et al., 2002). For Arundo
cane, lower activation temperature (400 C) was found to
give rise to a microporous carbons, while higher tempera-
tures (500–550 C) resulted in carbons with larger pores
(Vernersson et al., 2002).
3. Imaging the pore structure of activated carbons
Activated carbons are traditionally characterized using
indirect methods such as gas adsorption. The Langmuir
and its extension, the BET equation, are commonly applied
to the adsorption isotherm to calculate the specific surface
area and further analysis can lead to obtaining the pore
diameters. Very few have characterized activated carbons
using direct methods such as utilizing actual images
obtained using high magnification microscopes. High-reso-
lution transmission electron microscope (HRTEM) was
used to have a visualized observation of pores in activated
carbon fibers (Endo et al., 1998) and anthracite (Lillo-Rod-
enas et al., 2004). The two-dimensional fast Fourier trans-
form (FFT) was used to carry out the frequency analysis to
obtain the pore size distribution. Image analysis of acti-
vated anthracite provided de-averaged data on fringes
length, interlayer spacings, and number of stacked layers
which were helpful in describing the structures of the pores
(Lillo-Rodenas et al., 2004). Scanning tunneling micros-
copy (STM) techniques were also used for comparative
investigation of different activated carbons of ultramicro-
4889
porous, supermicroporous, and mesoporous types (Paredes
et al., 2006). The morphologies of these carbons, which
include pore diameter and pore shape were analyzed using
the STM images. Scanning electron microscopy (SEM) was
used by Vernersson et al. (2002) to see the influence of acti-
vation temperature and impregnation ratio on the pore
structure, particularly the shape and diameter of the pores
of Arundo donax cane activated using phosphoric acid. The
surface morphology of the optimally prepared activated
carbon from coconut husk was also analyzed using images
from the SEM by Tan et al. (2007). The images obtained
were compared against images taken from the raw material
(raw coconut husk) wherein very little pores were seen on
the surface. The activated carbon showed many large pores
in honeycomb shape.
4. Methods
4.1. Preparation of the activated carbon
Clean, fiber-free, and soil-free coconut shells were milled
to reduce the size down to mesh 8 (2.4 mm diameter). Phos-
phoric acid solutions were prepared to the required impreg-
nation ratios of 1.0, 1.5 and 2.0, defined as the ratio of dry
weight of H3PO4 to the weight of the coconut shell based
on a study by Molina-Sabio et al. (1995). Ten grams of
the coconut shells were used per sample. Soaking time
was fixed at 12 h. Since most literatures reported that opti-
mum activation temperature for most biomass materials
generally falls between 400 and 500 C (Srinivasakannan
and Zailani Abu Bakar, 2004), temperatures of 400, 450,
500 C were evaluated for this study. Activation times of
10, 20, and 30 min were also assessed. These times were
based on an initial experimental run at 400 C, wherein
smoke evolution ceased after about seven (7) min. It was
presumed that at this time, moisture and most of the vola-
tiles were eliminated from the precursor. Thus, the mini-
mum time evaluated was set at 10 min. A steady supply
of nitrogen was provided for the whole activation time to
have an inert environment for the activation process.
Washing then followed to remove traces of acid in the acti-
vated carbon. The activated samples were repeatedly
washed with about 100 ml of distilled water. The acidity
of the wash liquor was monitored until the pH reading is
at 6–7. On the average, 7–8 washings were able to neutral-
ize the sample. The activated carbon was then again
washed with 0.1 M sodium hydroxide solution and finally
with distilled water. The washed activated carbon samples
were then placed in an electric oven with temperature set at
105 C for drying.
4.2. Experimental design and statistical analysis
A 33 (impregnation ratio, activation temperature, and
activation time) fractional factorial experimental design
based on Box and Behnken (Myers and Montgomery,
2002) with three center runs was used, giving a total of
4890 M.K.B. Gratuito et al. / Bioresource Technology 99 (2008) 4887–4895
15 experimental runs. The dependent variables (responses)
analyzed were yield, bulk density, average diameter of
common pores, average diameter of small pores, and the
number of pores per unit area.
4.3. Evaluation of the responses
The carbon yield was the ratio of the weight of the acti-
vated and dried carbon to the initial weight of the coconut
shell taken for activation, both based on dry basis and
expressed as percentage.
The bulk density was determined using the method of
Lima and Marshall (2005) by filling a 10-ml tube with
the sample, 1 ml at a time, capping and tapping to a con-
stant (minimum) volume. The bulk density was taken as
the ratio of the weight (in grams) and the volume (in cm3).
Representative samples from each experimental run
were used for the image analysis using the scanning elec-
tron microscope (SEM) of the National Metals and Mate-
rials Technology Center (MTEC) at Thailand Science
Park. Images at 500· magnification were taken for the dig-
ital image analysis using the Image Pro Plus version 5.1 for
windows. The average pore diameter, small pore diameter
and the total area in a plane occupied by the pores were
determined from the images. The average pore diameters
were taken from the normally distributed readings as
shown in a sample experiments in Fig. 1. The number of
pores per unit area was determined using the average diam-
eter of the pores and the total area covered by the pores.
The simple formula is shown below:
Total area of all pores in a plane
n¼
average area of the pores
12
10
8
No of obs
0
0 2
Fig. 1. Pore diameter readings shown in a normal distribution curve for a sample experiment.
4.4. Regression and optimization analysis
The resulting data were regressed to derive a suitable
equation for each response. All variable parameters and
their interactions were considered for a model for each
response. A statistical analysis software was used to solve
the coefficients of the second-order model with three vari-
ables for each response as shown below:
Y ¼ b0 þ b1 X 1 þ b2 X 2 þ b3 X 3 þ b11 X 21 þ b22 X 22
þ b33 X 23 þ b12 X 1 X 2 þ b13 X 1 X 3 þ b23 X 2 X 3 ð2Þ
where b0, b1, b2, b3, b11, b22, b33, b12, b13 and b23 are the
regression coefficients; X1, X2, and X3 are the coded inde-
pendent variables/regressor; and Y is the particular re-
sponse evaluated. Predicted values were solved from the
derived equations for each of the response. These values
were plotted to obtain contour plots that were used for
the optimization process. Boundary conditions were ap-
plied to the contour plots. For yield, the boundary condi-
tion was set at 50% (the region corresponding to lesser
values were not accepted). This value is higher than the
yields reported for carbonization and activation processes
of dried coconut shell which are around 32.46% (Bhattach-
arya et al., 1989) and 40% (FFTCAPR, 2004), respectively.
As for the bulk density, the boundary condition was set at
460 kg/m3, and the region corresponding to lesser values of
bulk densities were not accepted. Bulk densities of agricul-
tural-based raw materials are generally low. Activated car-
bons from sugarcane bagasse and pecan shells have bulk
densities around 470 and 440 kg/m3, respectively, (Ahmed-
na et al., 2000). For the average pore diameter, values high-
er than 10 lm were not accepted. For the small pore
ð1Þ
diameter, the boundary condition was 1.2 lm (1200 nm).
Exp 3 = 93*1*normal(x, 9.5044, 3.6204)
4 6 8 10 12 14 16 18
micrometer
 M.K.B. Gratuito et al. / Bioresource Technology 99 (2008) 4887–4895 4891

Table 1
Experimental raw data
I Ratio A Time A Temp Yield % (dry basis) BD (kg/m3) Average pore diameter (lm) Small pore diameter (nm) Number of pores
1 1 0 50.88 459.06 11.8628 578.02 85.28
1 1 0 54.30 521.99 15.2331 1410.95 22.74
1 1 0 51.97 474.62 9.5044 1203.60 74.64
1 1 0 51.20 477.43 9.0845 1597.25 54.88
1 0 1 52.82 485.71 6.6128 1011.90 101.41
1 0 1 40.09 432.81 10.3354 1084.17 236.99
1 0 1 54.52 478.38 8.4812 1354.20 105.44
1 0 1 52.01 486.92 9.8143 1135.25 105.95
0 1 1 54.40 456.98 5.8695 1175.20 323.26
0 1 1 48.33 439.00 10.0923 1140.83 49.54
0 1 1 50.91 450.26 9.8754 1110.00 158.58
0 1 1 52.14 503.92 12.0418 1117.80 86.64
0 0 0 50.43 468.99 6.7005 1330.45 151.26
0 0 0 50.67 465.99 6.4585 1316.16 162.99
0 0 0 50.46 463.69 6.2165 1301.88 174.73

The regions corresponding to lesser values of diameter
were not accepted. This characteristic of the activated car-
bon was included in the analysis so that the pore size will be
expressed in a range of value and not just a specific point.
This boundary condition set the range for an acceptable
diameter range, which was from 1.2 (corresponding to
the small pore diameter) to 10 lm (corresponding to the
average pore diameter). The boundary condition set for
the number of pores per unit area was 160. The regions cor-
responding to lesser values were not accepted. Since no
similar studies on direct characterization of pores using
SEM images were found in the literatures, this boundary
condition was decided based on the predicted value under
the center runs (Table 4). This value was also near to the
midpoint of the range of values from the experimental data
which is 173.
A relatively straightforward approach in optimizing the
responses based on the variable conditions was done. Con-
tour graphs showing the acceptable and unacceptable
regions for each response based on the boundaries
described above were overlayed. The optimum points for
the activation conditions were identified by the midpoints
of the range of values for the optimum region. Activated
carbons were again produced using the optimum activation
conditions for validation. The responses were again mea-
sured. The values of the response were also compared with
the predicted values from the correlations developed for
each response.
5. Results and discussion
Table 1 shows the results of the experiments conducted.
Code values for the variable parameters were used to facil-
itate regression with 1 as the minimum level and +1 as
the maximum level. The center runs (0, 0, and 0) were
repeated three times as they contribute to the estimation
of the quadratic terms in the model.
The values obtained for yield range from 40% to 55%,
averaging at 51%. This average was relatively high com-
pared to the reported average yield of 40% by the Food
and Fertilizer Technology Center for the Asian and Pacific
Region (2004). The bulk density values ranged from 433 to
522 kg/m3 averaging at 471 kg/m3. The values for the aver-
age pore diameter ranged from 5.87 lm to as high as
15.23 lm. The small pore diameters were also measured.
The smallest measured 578 nm while the biggest is at
1597 nm. As for the number of pores, it ranged from 23
to 323 pores per unit plane area.

Table 2
The regression coefficients for each response
Yield Bulk density Average pore diameter Small pore diameter Number of pores
Estimate Standard error Estimate Standard error Estimate Standard error Estimate Standard error Estimate Standard error
b0 50.52 1.73 466.22 10.13 6.46 0.95 1316.6 126.54 162.99 46.40
b1 1.45 1.06 2.22 6.20 0.89 0.58 150.66 77.49 13.19 28.42
b2 0.37 1.06 15.49 6.20 1.11 0.58 142.29 77.49 26.24 28.42
b3 2.51 1.06 1.08 6.20 1.43 0.58 21.66 77.49 26.20 28.42
b11 0.01 1.56 12.73 9.13 2.15 0.86 54.14 114.06 60.33 41.83
b22 1.58 1.56 4.32 9.13 2.81 0.86 64.56 114.06 43.28 41.83
b33 0.65 1.56 8.00 9.13 0.20 0.86 115.64 114.06 34.79 41.83
b12 1.05 1.50 15.03 8.78 0.95 0.82 109.82 109.59 10.70 40.19
b13 2.56 1.50 15.36 8.78 0.60 0.82 72.80 109.59 33.77 40.19
b23 1.83 1.50 17.91 8.78 0.51 0.82 10.54 109.59 50.44 40.19
4892 M.K.B. Gratuito et al. / Bioresource Technology 99 (2008) 4887–4895

Table 3
Predicted values for each response
I Ratio A Time A Temp Yield, % (dry basis) BD (kg/m3) Average pore diameter (lm) Small pore diameter (nm) Number of pores
1 1 0 49.22 450.53 10.26 794.69 109.50
1 1 0 52.06 511.58 14.38 1298.91 35.64
1 1 0 54.22 485.03 10.36 1315.64 61.73
1 1 0 52.86 485.96 10.69 1380.59 30.66
1 0 1 53.47 485.18 7.68 944.58 143.06
1 0 1 43.34 452.29 11.73 1046.87 158.21
1 0 1 51.27 458.90 7.08 1391.5 184.22
1 0 1 51.36 487.45 8.75 1202.58 64.30
0 1 1 55.41 466.04 6.41 1025.86 257.38
0 1 1 46.74 428.06 10.30 961.47 104.10
0 1 1 52.50 461.21 9.67 1289.37 104.02
0 1 1 51.13 494.86 11.50 1267.14 152.52
0 0 0 50.52 466.22 6.46 1316.16 162.99
0 0 0 50.52 466.22 6.46 1316.16 162.99
0 0 0 50.52 466.22 6.46 1316.16 162.99

The results of the regression are shown in Table 2. The
predicted values for each response using the second-order
models are in Table 3. Figs. 2 and 3 show the superimposed
graphs used to derive the optimum production conditions.
The optimum regions and the optimum points, corre-
sponding to the midpoints of the optimum region, are tab-
ulated in Table 4. The overall optimum activation
temperature is 416 C. The optimum range for activation
temperature is from 394 to 416 C. In a published study
on coconut shell chemical activation using phosphoric acid,
the optimum temperature was found equivalent to 450 C
(Laine et al., 1989). This value is higher by about 34 C
to optimum temperature identified in this study. However,
the optimum temperature identified by Laine et al. (1989)
was only based on the surface area and not on multiple
responses like in this study.
The optimum range for impregnation ratio was from
1.345 to 2, while the overall optimum ratio is 1.725. It
was just sufficient to weaken the material’s structure and
not cause bigger pores that may not be able to trap gener-
ally smaller-sized contaminants in water stream applica-
tion. Additionally, the optimum impregnation ratio along
with the combination of the other activation parameters
did not cause development of bigger pore size that equates
to lower bulk density, an undesirable characteristic of acti-
vated carbons. Finally, the optimum ratio identified would
also allow financial savings as chemical agents are costly.
As for the activation time the optimum range was from
14.9 to 23.9 min, while the optimum point was determined
as 19.5 min. The minimum time required to eliminate all
the moisture and most of the volatiles in the precursor
which on the average was seven (7) min was met and the

Fig. 2. The optimum region and point based on the impregnation ratio and activation temperature.
 M.K.B. Gratuito et al. / Bioresource Technology 99 (2008) 4887–4895 4893

Fig. 3. The optimum region and point based on the activation time and activation temperature.

Table 4
The optimum regions and the optimum points
Optimum region Optimum point
Lower limit Upper limit
Coded value Actual value Coded value Actual value Coded value Actual value
I Ratio vs. A Time
I Ratio 0.41 1.295 0.47 1.735 0.03 1.515
A Time 0.72 12.8 min 0.33 23.3 min 0.20 18 min
I Ratio vs. A Time
I Ratio 0.31 1.345 1.2 2.1 0.45 1.725a
A Temp 1.04 398 C 0.32 434 C 0.68 416 Ca
A Time vs. A Temp
A Time 0.49 14.9 min 0.39 23.9 0.05 19.5 mina
A Temp 1.12 394 C 0.35 432.5 C 0.74 413 C
a
Overall optimum points.

identified optimum activation time was sufficient to allow
pore development due to the loss of volatiles and the sub-
sequent reorganization of the remaining structures. The
range identified indicated that it is not necessary to prolong
activation so much beyond the basic requirement, as doing
so would cause pores enlargement, which may be undesir-
able depending on the requirement of a specific activated
carbon application. Aside from that, longer duration also
obviously means higher energy costs as the high activation
temperature has to be maintained for a longer period of
time.
For the validation part, activated carbon was produced
under the optimum variable conditions (impregnation
ratio = 1.725, activation time = 19.5 min, and activation
temperature = 416 C). The responses were again mea-
sured and compared against the predicted values from
the second-order models previously developed. The results
are shown in Table 5. Except for the case of the average
pore diameter response, the models were able to predict
the responses well as evidenced by the generally acceptable
values of the percentage errors. The percentage error refers
to the deviation of the actual measurements from the pre-
dicted values. The smallest error was in the small pore
diameter measurement which valued just at 1.23%, while
the highest was for the average diameter pore reading
which is at 21.32%. The rest were below the 10% mark.
For comparison, two experimental runs outside the
identified optimum conditions were selected and used.
Table 6 tabulates the results of the non-optimum activated
carbons against the optimum output. It can be seen that
not all responses for the non-optimum carbons are within
the acceptable range. Clearly, yield (49%), bulk density
(450 kg/m3), small pore diameter (795 nm), and the number
of pores (1 1 0) of activated carbons produced under the
first set of conditions outside the range were not in the
acceptable regions. Only the average pore diameter barely
4894 M.K.B. Gratuito et al. / Bioresource Technology 99 (2008) 4887–4895

Table 5
The measured responses for the optimum activated carbon
Responses
Yield, % BD Average pore Small pore Number
(dry basis) (kg/m3) diameter (lm) diameter (nm) of pores
Predicted values using the correlationsA 53.65 467.06 6.38 1257.75 206.61
Optimum activated carbonB 51.59 487.36 5.02 1273.23 225.47
Acceptable boundary/range >50 >460 <10 >1200 >160
Residual (A  B) 2.06 20.3 1.36 15.48 18.86
% Error = |Residual| * 100/A 3.84 4.35 21.32 1.23 9.13
Remarks Acceptable Acceptable Acceptable Acceptable Acceptable

Table 6
Optimum against non-optimum activated carbons
I Ratio A Time A Temp Yield % BD Average pore Small pore Number
(min) (C) (dry basis) (kg/m3) diameter (lm) diameter (nm) of pores
Non-optimum 1 10 450 49.22 450.53 10.26 794.69 109.50
1 30 450 52.06 511.58 14.38 1298.91 35.64
Optimum 1.725 19.5 416 51.59 487.36 5.02 1273.23 225.47

reached the acceptable range. As for the second non-opti- References

mized sample, the average pore diameter and the number
of pores were way beyond what were required. On the
other hand, all responses for the optimized activated car-
bon were within the identified acceptable range.
6. Summary
This research was carried out to determine optimum
conditions for the production of activated carbon from
coconut shell based on output characteristics such as yield,
bulk density, average pore diameter, small pore diameter,
and number of pores per unit area. The last three responses
were directly seen and measured from images generated by
a scanning electron microscope (SEM). Direct character-
ization of the pores using the SEM was found to be a good
technique to actually see the pores and get actual measure-
ments that can then be used to analyze and the optimize
production process. The major findings are:
• The optimum range identified for impregnation ratio
starts from 1.345 to 2, while for the activation time it
was from 14.9 to 23.9 min. The optimum range for acti-
vation temperature is from 394 to 416 C. The optimum
points are 1.725, 19.5 min, and 416 C for impregnation
ratio, activation time, and activation temperature,
respectively.
• Upon validation of the optimum conditions, it was
found out that the correlations developed were able to
predict the values of the responses quite well as evi-
denced by the generally acceptable values of the percent-
age errors. The smallest error was in the small pore
diameter measurement which valued at just 1.23%, while
the highest was for the average diameter pore reading
which is at 21.32%. The rest were below the 10% mark.
Ahmedna, M., Marshall, M.E., Rao, R.M., 2000. Production of granular
activated carbons from select agricultural by-products and evaluation
of their physical, chemical and adsorption properties. Bioresource
Technology 71, 113–123.
Bacaoui, A., Yaacoubi, A., Dahbi, A., Bennouna, C., Phan Tan Luu, R.,
Maldonado-Hodar, F.J., Rivera-Utrilla, J., Moreno-Castilla, C., 2001.
Optimization of conditions for the preparation of activated carbons
from olive-waste cakes. Carbon 39 (3), 425–432.
Bhattacharya, S.C., Bhattacharya, S.P., Amin, N., Rongthonng, S., 1989.
Laboratory scale batch carbonization of selected residues. AIT–GTZ
Biocoal Project. Asian Institute of Technology, Bangkok, Thailand.
Diao, Y., Walawender, W.P., Fan, L.T., 2002. Activated carbons prepared
from phosphoric acid activation of grain sorghum. Bioresource
Technology 81, 45–52.
Endo, M., Furuta, T., Minoura, F., Kim, C., Oshida, K., Dresselhaus, G.,
Dresselhaus, M.S., 1998. Vizualized observation of pores in activated
carbon fibers by HRTEM and combined image processor. Supramo-
lecular Science 5, 261–266.
Food and Fertilizer Technology Center for the Asian and Pacific Region
(FFTCAPR). 2004-11-01 document (Processing of coconut shell into
activated carbon/charcoal) <http://www.agnet.org/library/pt/
2004003/>.
Girgis, B.S., Yunis, S.S., Soliman, A.M., 2002. Characteristics of activated
carbon from peanut hulls in relation to conditions of preparation.
Materials Letters 57 (1), 164–172.
Ioannidou, O., Zabaniotou, A., 2006. Agricultural residues as precursors
for activated carbon production – a review. Renewable and Sustain-
able Energy Reviews 11 (9), 1966–2005.
Karacan, F., Ozden, U., Karacan, S., 2007. Optimization of manufactur-
ing conditions for activated carbon from Turkish lignite by chemical
activation using response surface methodology. Applied Thermal
Engineering 27 (7), 1212–1218.
Lafi, W.K., 2001. Production of activated carbon from acorns and olive
seeds. Biomass and Bioenergy 20, 57–62.
Laine, J., Calafat, A., Labady, M., 1989. Preparation and characterization
of activated carbons from coconut shell impregnated with phosphoric
acid. Carbon 27 (2), 191–195.
Lázaro, M.J., Gálvez, M.E., Artal, S., Palacios, J.M., Moliner, R., 2007.
Preparation of steam-activated carbons as catalyst supports. Journal
of Analytical and Applied Pyrolysis 78 (2), 301–315.
 M.K.B. Gratuito et al. / Bioresource Technology 99 (2008) 4887–4895
Lillo-Rodenas, M.A., Cazorla-Amoros, D., Linares-Solano, A., Beguin, F.,
Clinard, C., Rouzaud, J.N., 2004. HRTEM study of activated carbons
prepared by alkali hydroxide activation of anthracite. Carbon 42, 1305–1310.
Lima, I., Marshall, W.E., 2005. Utilization of turkey manure as granular
activated carbon: physical, chemical and adsorptive properties. Waste
Management 25, 726–732.
Molina-Sabio, M., Rodriguez-Reinoso, F., Caturia, F., Selles, M.J., 1995.
Porosity in granular carbons activated with phosphoric acid. Carbon
33 (8), 1105–1113.
Myers, R.H., Montgomery, D.C., 2002. Response Surface Methodology:
Process and Product Optimization Using Designed Experiments,
second ed. John Wiley & Sons, Inc.,, New York.
Paredes, J.I., Martinez-Alonso, A., Hou, P.-X., Kyotani, T., Tascon,
J.M.D., 2006. Imaging the structure and porosity of active carbons by
scanning tunneling microscopy. Carbon 44, 2469–2478.
View publication stats
4895
Srinivasakannan, C., Zailani Abu Bakar, M., 2004. Production of
activated carbon from rubber wood sawdust. Biomass and Bioenergy
27, 89–96.
Tan, I.A.W., Ahmad, A.L., Hameed, B.H., 2007. Optimization of
preparation conditions for activated carbons from coconut husk using
response surface methodology. Chemical Engineering Journal, in
press. doi:10.1016/j.cej.2007.04.031.
Toles, C.A., Marshall, W.E., Johns, M.M., Wartelle, L.H., McAloon, A.,
2000. Acid-activated carbons from almond shells: physical, chemical
and adsorptive properties and estimated cost of production. Biore-
source Technolology 71, 87–92.
Vernersson, T., Bonelli, P.R., Cerrella, E.G., Cukierman, A.L., 2002.
Arundo donax cane as a precursor for activated carbon prepara-
tion by phosphoric acid activation. Bioresource Technology 83,
95–104.