i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 7 4 7 5 – 7 4 8 2

Avai lab le a t www.sc iencedi rec t .com

j ourna l homepage : www.e lsev ier . com/ loca te /he

Optimization of biohydrogen production by Clostridiumbutyricum EB6 from palm oil mill effluent using responsesurface methodology

Mei-Ling Chonga,*, Nor’ Aini Abdul Rahmana, Raha Abdul Rahimb, Suraini Abdul Aziza,Yoshihito Shiraic, Mohd Ali Hassana

aDepartment of Bioprocess Technology, Faculty of Biotechnology and Biomolecular Sciences, Universiti Putra Malaysia, 43400 Serdang,

Selangor, MalaysiabDepartment of Cell and Molecular Biology, Faculty of Biotechnology and Biomolecular Sciences, Universiti Putra Malaysia, 43400 Serdang,

Selangor, MalaysiacGraduate School of Life Sciences and System Engineering, Kyushu Institute of Technology, 808-0196 Hibikino 2-4, Wakamatsu-ku,

Kitakyushu-shi, Fukuoka, Japan

a r t i c l e i n f o

Article history:

Received 13 July 2008

Received in revised form

17 May 2009

Accepted 17 May 2009

Available online 23 June 2009

Keywords:

Biohydrogen

Clostridium butyricum EB6

Response surface methodology

POME

* Corresponding author. Tel.: þ603 89467590;E-mail address: meiling012001@yahoo.co

0360-3199/$ – see front matter ª 2009 Interndoi:10.1016/j.ijhydene.2009.05.088

a b s t r a c t

Clostridium butyricum EB6 successfully produced hydrogen gas from palm oil mill effluent

(POME). In this study, central composite design and response surface methodology were

applied to determine the optimum conditions for hydrogen production (Pc) and maximum

hydrogen production rate (Rmax) from POME. Experimental results showed that the pH,

temperature and chemical oxygen demand (COD) of POME affected both the hydrogen

production and production rate, both individually and interactively. The optimum condi-

tions for hydrogen production (Pc) were pH 5.69, 36 �C, and 92 g COD/l; with an estimated Pc

value of 306 ml H2/g carbohydrate. The optimum conditions for maximum hydrogen

production rate (Rmax) were pH 6.52, 41 �C and 60 g COD/l; with an estimated Rmax value of

914 ml H2/h. An overlay study was performed to obtain an overall model optimization. The

optimized conditions for the overall model were pH 6.05, 36 �C and 94 g COD/l. The

hydrogen content in the biogas produced ranged from 60% to 75%.

ª 2009 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights

reserved.

1. Introduction be produced by expensive methods, such as electrolysis and

Hydrogen gas is a clean and ideal alternative energy for the

future, therefore hydrogen gas production and utilization

have been highly investigated [1,2]. Hydrogen has a high

energy yield of 122 kJ/g, which is 2.75 times greater than

hydrocarbon fuels. This characteristic has made it a prom-

ising alternative fuel. However, there are major challenges

with using hydrogen gas as a fuel. Not only it is difficult to

transport and store, but also unavailable in nature and must

fax: þ603 89467593.m (M.-L. Chong).ational Association for H

steam reformation. For these reasons, biological hydrogen gas

production has sparked great interest, as it is less energy

intensive and can be combined with wastewater treatment

processes [3] via dark fermentation and photofermentation

[4]. The fermentation feedstocks most studied have been

municipal solid waste [5], food processing industry waste [6,7]

and dairy waste [8].

In Malaysia, palm oil extraction generates about 50 million

tons of palm oil mill effluent (POME) annually. POME contains

ydrogen Energy. Published by Elsevier Ltd. All rights reserved.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 7 4 7 5 – 7 4 8 27476

high chemical oxygen demand (COD) due to the lignocellulose

and hemicelluloses components of the material [9]. In our

previous study, we developed a closed-tank anaerobic system

to investigate COD removal efficiency [10]. During the acido-

genesis process, organic substances are converted into acid

and hydrogen gas. If the degradation is continued to the

methanogenesis process, methane gas is the end-product.

Previous research has showed hydrogen production using

POME sludge as the inoculum is promising [11]. Single strain

cultivation, using locally isolated Clostridium butyricum EB6,

was successfully carried out in POME.

Hydrogen gas production from waste has been extensively

studied. Researchers reported that hydrogen gas production

was dependent on the pH, organic loading rate and tempera-

ture [12–14]. Results indicated that pH was crucial to hydrogen

gas production, due to its effect on the hydrogenase enzyme

[15] and its metabolism [16]. The reported optimum pH varied

with the substrate used, but it was always within the range of

pH 4.5–7.0 [17]. Temperature effect on the maximum substrate

utilization rates has also been studied in clostridia [13].

Results of previous studies have shown that the substrate

concentration significantly affected the hydrogen gas

production and the metabolic product distribution [18].

The response surface methodology (RSM) has been

proposed to study the influence of identified parameters and

their individual and interactive effects. This study used a well-

designed strategy, based on RSM theory, to perform a number

of planned experiments and analyze the responses statisti-

cally [19]. The objective of the current study was to identify the

individual and interactive effects of pH, temperature and COD

of POME on hydrogen gas production using C. butyricum EB6.

2. Materials and methods

2.1. Microorganism and culture medium

The bacterium used in this study was C. butyricum EB6.

Previous studies have proved that this strain is capable of

producing hydrogen gas via fermentation [20]. The bacterium

was preserved in 20% glycerol at �20 �C. Bacterium stability

was routinely examined under a microscope. The Hungate

tube cultivation technique was adapted to culture C. butyricum

EB6 in an aerobic environment.

2.2. Cultivation medium and reactor set-up

For inoculum preparation, a stock culture (10 ml in Hungate

tube) was heat-shocked at 90 �C for 90 s. After the culture

cooled, it was transferred to a 250 ml modified serum bottle

with 100 ml of reinforced clostridial medium. Reinforced

clostridial medium consists of (g/l): meat extract, 10; peptone,

5; yeast extract, 3; Dþ glucose, 5; starch, 1; sodium chloride, 5;

sodium acetate, 3; L-cystenium chloride, 0.5; agar, 0.5. The pH

of the medium was initially adjusted to 6.5 and sparged with

nitrogen before sterilization. The inoculum was then incu-

bated at 37 �C for 18 h without shaking. Hydrogen production

experiments were conducted in a 3L fermenter (B. Braun,

Germary). 1000 ml of fresh POME were added to the fermenter

at varying pH, temperature and COD levels. Prior to their use,

all media were sparged with nitrogen for 20 min to ensure

anaerobic conditions. The pH of the medium was controlled

using 1 N NaOH. The agitation rate was constant (200 rpm) for

all experiments.

2.3. Experimental design

Optimization of hydrogen production using POME was carried

out using a central composite design. The pH, temperature

and COD of POME were selected as independent variables with

the range of 5.3–6.7, 32–42 �C and 60–100 g COD/l, respectively.

These ranges were selected based on previous research

[11,13,14,21]. The variable, Xi, was coded as xi according to

equation (1) such that X0 corresponded to the central value:

xi ¼Xi � X�i

DXi; where i ¼ 1; 2;3;/k; (1)

where xi is the dimensionless coded value of an independent

variable, Xi is the actual value of an independent variable for

the ith test, Xi* is the actual value of an independent variable

at the center point and DXi is the step change.

The experimental design was analyzed using response

surface methodology (RSM); a collection of mathematical and

statistical techniques used to model and analyze problems in

which a response of interest is influenced by several variables.

The objective is to optimize this response [19]. The general

form of the second degree polynomial equation is:

Yi ¼ bo þXk

i¼1

bixi þXk

i¼1

biix2i þ

Xk

i¼1

Xk

j¼1

bijxixj (2)

where Yi is the predicted response, xixj are input variables

which influence the response variable Y, bo is the offset term,

bi is the ith linear coefficient, bii is the ith quadratic coefficient

and bij is the ijth interaction coefficient. The secondary order

polynomial coefficients were calculated using Design Expert

software version 7.0.0 (Stat-Ease Inc.).

2.4. Analytical method

The exit gas was analyzed using a Shimadzu GC-8A gas

chromatograph (GC) with a thermal conductivity detector.

The carrier gas was nitrogen and the column was packed with

Porapak Q (80/100 mesh). Temperatures at the stainless steel

column and injection/detector point were 50 �C and 100 �C,

respectively. A standard hydrogen curve was plotted using

standard hydrogen gas. Total carbohydrate in POME was

analyzed using phenol sulfuric acid method (standard method

by APHA)

2.5. Kinetic modeling

The cumulated hydrogen production in the batch experi-

ments followed the modified Gompertz equation [22] (equa-

tion (3)).

H ¼ P exp

�� exp

�Rmaxe

Pðl� tÞ þ 1

��(3)

H is the cumulative hydrogen production (ml), l the lag time

(h), P is the hydrogen production potential (ml), Rmax is the

Table 1 – Experimental results for Pc and Rmax for thecomposite design.

Run Actual value Pc (ml H2/gcarbohydrate)

Rmax (mlH2/h)

X1:pH

X2:temperature

X3:COD

1 6.0 37 80 297.9 780

2 5.3 42 60 206.2 230

3 6.0 37 80 296.0 885

4 6.0 37 80 310.6 754

5 6.0 37 120 279.0 839

6 6.0 37 80 284.0 910

7 6.0 37 80 283.2 885

8 5.3 32 100 262.4 421

9 6.0 37 40 200.2 635

10 6.7 32 60 195.5 383

11 6.7 42 100 170.6 646

12 5.3 42 100 227.3 269

13 5.3 32 60 216.4 150

14 6.0 37 80 297.9 783

15 7.4 37 80 160.8 383

16 6.7 42 60 125.1 929

17 6.7 32 100 212.4 550

18 6.0 27 80 168.0 171

19a 4.7 37 80 – –

20a 6.0 47 80 – –

Pc (ml H2/g carbohydrate)¼Hydrogen production.

Rmax (ml H2/h)¼Hydrogen production rate.

a Experiments were excluded from the design as it is an extreme

condition for biohydrogen production from POME using Clostridium

butyricum EB6.

1000

0

2000

3000

4000

5000

6000

Tot

al a

ccum

ulat

edhy

drog

en g

as (

ml)

0 2 4 6 8 10 12 14 16 18 20

Time (hour)

0

10

20

30

40

50

60

70

80

a

c

Fig. 1 – Batch hydrogen fermentation of Clostridium butyricum E

Profile of hydrogen production (ml), (b) Profile of biogas producti

total carbohydrate.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 7 4 7 5 – 7 4 8 2 7477

maximum hydrogen production rate (ml/h) and e is

2.718281828. The P, Rmax and l values for each batch were

estimated using the nonlinear estimation function in STA-

TISTICA (version 6).

Hydrogen gas production was calculated using gas

sampled from the bioreactor headspace and the gas compo-

sition was measured. The total volume of biogas produced at

each time interval is given by equation (4).

VH;i ¼ VH;i�1 þ CH;i

�VG;i � VG;i�1

�þ VH

�CH;i � CH;i�1

�(4)

VH,i and VH,i � 1 are cumulative hydrogen gas volumes at the

current (i) and previous (i� 1) time intervals, VG,i and VG,i � 1

the total biogas volumes in the current and previous time

intervals, CH,i and CH,i � 1 the fraction of hydrogen gas in the

headspace of the bottle measured using gas chromatography

in the current and previous intervals, and VH the total volume

of headspace in the reactor [23].

3. Results and discussion

In all experiments, the biogas produced contained hydrogen

(60–75%) and carbon dioxide (25–40%). No methane gas was

detected. All cumulative hydrogen production data fitted

equation (3) well, with R2> 0.99.

3.1. Overall performance of hydrogen productionfrom POME

The experimental results of all eighteen runs are summarized

in Table 1. Fig. 1 illustrates the biohydrogen production profile

for Run #4 at pH 6.0, 37 �C and 80 g COD/l. Fig. 1a and b repre-

sents hydrogen production and total biogas production, fitted

1000

0

2000

3000

4000

5000

6000

Tot

al b

ioga

s (m

l)

0 2 4 6 8 10 12 14 16 18 20

Time (hour)

12

14

16

18

20

22

24

26

28

Tot

alca

rboh

ydra

te (

g/L

)

b

d

B6 from POME at pH 6.0, 37 8C and 80 g COD/l (Run #4). (a)

on (ml), (c) Profile of hydrogen percentage, and (d) Profile of

Table 2 – ANOVA on model for Pc and Rmax.

Pc (ml H2/gcarbohydrate)

Rmax (ml H2/h)

Model probability <0.0001 <0.0001

Model F-value 27.29 32.34

R2 0.9685 0.9733

Adjusted R2 0.9330 0.9432

Adequate precision 14.2381 15.9990

Standard deviation 14.7828 65.3260

Lack of fit probability 0.0877 0.5363

Lack of fit F-value 3.92 0.82

3235

3740

42

60

7080

90

100

190

220

250

280

310

Hyd

roge

n pr

oduc

tion

(ml/g

car

bohy

drat

e)H

ydro

gen

prod

ucti

on(m

l/g c

arbo

hydr

ate)

Hyd

roge

n pr

oduc

tion

(ml/g

car

bohy

drat

e)

Temperature (deg C)

Temperature (deg C)

COD (g COD/L)

5.30

5.656.00

6.356.70

32

3537

40

42

160

198

235

273

310

pH

6.70100

213

236

259

281

304c

b

a

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 7 4 7 5 – 7 4 8 27478

to the modified Gompertz equation. The kinetic parameters

for total accumulated hydrogen gas (ml) were P: 3790 ml, Rmax:

885.8 ml H2/h and l: 7.14 h.

As shown in Fig. 1c, the percentage of hydrogen gas in the

bioreactor increased with time. Initially, there was a 7 h lag

phase, after which, the gas production increased greatly. The

total carbohydrate degradation began after the 4 h. During the

stationary phase of hydrogen gas production, 48% of total

carbohydrate in POME remained, suggesting that C. butyricum

cannot utilize all remaining carbohydrates. POME naturally

contains high amounts of cellulosic and lignocellulosic

material. C. butyricum in nature produces xylanase [24]. The

hemicellulose in POME can be readily utilized by C. butyricum

EB6, leaving behind unutilized cellulose.

3.1.1. Effects of pH, temperature and COD on Pc

(hydrogen production)In order to evaluate the effects of pH, temperature and COD on

Pc, the design matrix of experimental conditions with their

corresponding Pc values (Table 1) was subjected to regression

analysis. The mathematical model relating the production of

biohydrogen to the independent process variables of pH,

temperature and COD is given in the quadratic regression as

follows:

Pc ¼ 292:9� 23:1x1 � 22:7x2 þ 17:9x3 � 8:4x1x2 � 0:6x1x3

þ 0:5x2x3 � 24:5x21 � 45:6x2

2 � 14:8x23; (5)

where, Pc is the hydrogen production (ml H2/g carbohydrate),

and x1, x2 and x3 are coded values of independent variables.

The multiple correlation efficient, R2 and adjusted R2 were

evaluated. The adjusted R2 measures the amount of variation

about the mean explained by the model. It is adjusted for the

Table 3 – Summary of model terms.

Term p-value

Pc (ml H2/g carbohydrate) Rmax (ml H2/h)

X1-pH 0.0017 <0.0001

X2-temp 0.0019 0.0172

X3-COD 0.0013 0.0502

X1X2 0.1485 0.0048

X1X3 0.9125 0.0500

X2X3 0.9309 0.0061

X12 0.0005 <0.0001

X22 <0.0001 <0.0001

X32 0.0011 0.0862

COD (g COD/L)pH

5.305.65

6.006.35

6070

80

90

Fig. 2 – Three dimension surface graphs of the model for Pc

at the optimum point for each variable: (a) fixed pH at 6.0

(b) fixed COD at 80 g/l (c) temperature 37 8C.

number of model parameters relative to the number of points

in the design. The adjusted R2 of 93.3% for Pc is attributed to

the independent variables. The R2 for Pc is 96.85%, which

indicates a good agreement between experimental and pre-

dicted values.

Table 4 – Summary of optimize conditions for Pc and Rmax and experimental value.

Pc (ml H2/gcarbohydrate)

Rmax

(ml H2/h)Overall modeloptimization

Experimental value Experimental value

Optimum

condition

Low value Optimum

pH 5.69 6.52 6.05 5.5 6.05

Temperature

(�C)

36 41 36 32 36

COD (g/L) 92 60 94 94 94

Predicted value 306 914 Pc¼ 298, Rmax¼ 850 Pc¼ 254 (Std dev¼�12.65)a,

Rmax¼ 460 (Std dev¼�11.93)aPc¼ 290 (Std dev¼�5.66)a,

Rmax¼ 802 (Std dev¼�33.94)a

a Std dev¼ standard deviation.

Pre

dict

ed V

alue

s

120 140 160 180 200 220 240 260 280 300 320120

160

200

240

280

320

0 100 200 300 400 500 600 700 800 900 1000

Observed Values

0

200

400

600

800

1000

a

b

Fig. 3 – Relationship between predicted values and

observed values: (a) hydrogen production (ml H2/g

carbohydrate) and (b) hydrogen production rate (ml H2/h).

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 7 4 7 5 – 7 4 8 2 7479

The ANOVA study is documented in Table 2. The F-value

for a term compares a term’s variance with the residual

variance. The ANOVA result showed the computed F-value

was 27.29, which was much greater than the F0.05,9,8 (3.39),

which implied the model is significant at high confidence level

of 95%. The probability> F of <0.0001 means there was less

than a 0.01% chance that a model F-value could occur due to

noise. Lack of fit measures the variation of the data around the

fitted model. The lack of fit F-value compares the lack of fit

variance with pure error variance. If the model does not fit the

data well, this value will be significant; if the variances are

close to the same, the ratio will be close to one and it is less

likely that lack of fit is significant. The computed lack of fit F-

value of 3.92 was less than the F0.05,3,5 (5.41), therefore, the lack

of fit test was insignificant. The model fit the data well, with

a p-value of 0.0877; there was only 8.77% chance that this lack

of fit F-value could occur due to noise.

The p-values for the linear, interactive and quadratic terms

are shown in Table 3. The confidence level was set at 95%

(a¼ 0.05). The linear and quadratic coefficients of the three

variables showed a significant effect with p-values lower than

0.002. In short, all three variables influenced the Pc. This result

was comparable to results from previous research using

ananaerobic culture [25] and Enterobacter aerogenes [26]. The

coefficients for interactive effects were very high. However,

we could not overrule their effect on biohydrogen production

from POME by C. butyricum EB6 because of their individual

effect.

Fig. 2 illustrates the three-dimensional response surfaces

based on equation (5), by keeping one variable constant at its

center point, and leaving two variables within the experi-

mental range. As shown in Fig. 2, the response surface of Pc

displayed a clear optimum point which fell inside of the

boundary range. From the examination of the contour plot

with fixed COD (Fig. 2b) compared to the others (Fig. 2a and c),

there was a slight elongation sloping downward. The

hydrogen production, Pc, was slightly more sensitive to

changes in temperature than to pH. The contour plot with

respect to temperature to COD (fixed pH) and pH to COD (fixed

temperature) showed a round ridge running around the center

point, suggesting low correlation levels. This is confirmed by

the p-value of both being >0.05. The maximum predicted

response value for Pc, as estimated by equation (5), is 306 ml

H2/g carbohydrate with a pH of 5.69, temperature of 36 �C and

92 g COD/l (Table 4). Fig. 3a shows that actual experimental

values were close to the predicted using equation (5).

In many studies, the pH, temperature and substrate

concentration significantly affected the hydrogen gas

production during acidogenesis of the anaerobic digestion of

waste [27]. The optimum conditions found from this study

were similar to previously published results obtained with

mixed culture, mainly clostria sp.

3.1.2. Effect of pH, temperature and COD on Rmax

Similarly, the data for Rmax was analyzed based on ANOVA

and response surface to study the pH, temperature and COD.

The mathematical model relating to the production of bio-

hydrogen, with the independent process variables of pH,

temperature and COD, was given in the quadratic regression

as follows:

Rmax ¼ 829:37þ 184:95x1 þ 66:057x2 þ 37:63x3 þ 89:25x1x2

� 53:25x1x3 � 85:25x2x3 � 209:26x21 � 136:76x2

2 � 25:69x23

(6)

The R2 for Rmax is 97.33%, which indicated a good agreement

between experimental and predicted values. The ANOVA

32

3537

4042

6070

80

90

100

470

565

850

Hyd

roge

n pr

oduc

tion

rat

e (m

l/h)

Hyd

roge

n pr

oduc

tion

rat

e (m

l/h)

Hyd

roge

n pr

oduc

tion

rat

e (m

l/h)

5.30

5.65

6.00

6.35

6.70

3235

37

40

42

270

427.5

585

742.5

900

pH

pH5.30

5.65

6.00

6.35

6.70

6070

80

90

100

310

452.5

595

737.5

880

755

660

Temperature (deg C)

Temperature (deg C)

COD (g COD/L)

COD (g COD/L)

a

b

c

Fig. 4 – Three dimension surface graphs of the model for Rmax (a) fixed pH at 6.0 (b) fixed COD at 80 g/l 37 8C (c) fixed

temperature.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 7 4 7 5 – 7 4 8 27480

(Table 2) showed that the F-value for the model was 32.34,

greater than the tabulated F-value. This indicated that the

model term was significant, with p-value< 0.000. The pre-

dicted lack of fit F-value of 0.82 was less than the F0.05,3,5 (5.41),

showing the lack of fit test was insignificant. The model fit the

data well and the lack of fit p-value of 0.5363 shows there was

53.63% chance that a lack of fit F-value could occur due to

noise.

Fig. 5 – Overlay plot of hydrogen production and hydrogen

production rate in response to temperature and pH with

the COD fixed at 94 g COD/l. The optimum area is shaded in

grey.

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 7 4 7 5 – 7 4 8 2 7481

The p-values of the linear, interactive and quadratic terms,

shown in Table 3 with 95% confidence, showed that almost all

the terms were significant, except the quadratic terms for

COD. However, the effect of quadratic terms for COD cannot

be completely eliminated, because the p-value was within the

90% confidence range.

Fig. 4 illustrates the three-dimensional response surfaces

based on equation (6), by keeping one variable constant at its

center point, and leaving two variables within the experi-

mental range. As shown in Fig. 4, the response surface of Rmax

showed a clear optimum point which fell inside of the

boundary range. From the examination of the contour plot,

there was an elongation diagonally in both directions. This

showed that the interactions of temperature to COD (Fig. 4a),

Table 5 – Comparison of biohydrogen production with cited lit

Microorganism Substrate pH Temperature (�C) Subst

C. butyricum POME 6.05 36

Mixed culture Glucose 6 35

Mixed culture Glucose 5.5 36

C. butyricum Sucrose 5.5 37

Mixed culture Sucrose 5.5 37

Mixed culture POME 5.5 60

a g COD/l.

b ml H2/g carbohydrate.

c mol H2/mol glucose.

d mol H2/mol sucrose.

e ml H2/g COD/l.

f ml H2/L-medium.

temperature to pH (Fig. 4b) and COD to pH (Fig. 4c) were

significant. This was confirmed by the ANOVA result, with the

p-value lower than 0.05. The maximum predicted response

value for Pc, estimated by equation (6), was 914.08 ml H2/h

with the corresponding variables of pH 6.52, 41 �C and COD

60 g/l (Table 4). The experimental values were close to value

predicted using equation (6) (Fig. 3b). The hydrogen produc-

tion rates from this study were comparable to those found in

literature [28]. Additionally, this study confirmed that the RSM

approach is appropriate for optimizing the hydrogen produc-

tion from POME by C. butyricum EB6.

3.2. Process optimization

Hydrogen production (Pc) and hydrogen production rate (Rmax)

were each optimized at pH, temperature and COD of POME.

We aimed to then determine the correct combination of

variables to allow both responses to be optimized simulta-

neously. The best compromise can be determined visually by

superimposing overlaying responses on a contour plot [29].

Fig. 5 illustrates the overlay plot of temperature and pH. The

optimum region (shaded region) was identified for hydrogen

production (Pc) and hydrogen production rate (Rmax). The

optimum condition for the overall model was at pH 6.05, 36 �C

and 94 g COD/l. In order to validate the model, one point

within the optimum region and one point outside the region

were chosen. Experiments were performed to compare the

experimental result with its corresponding predicted value

(Table 4). The first experiment corresponded to the lower

value of pH (5.5) and temperature (30 �C), yielding Pc at 254 ml

H2/g carbohydrate and Rmax at 460 ml H2/h. The second

experiment corresponded to the optimum value of pH (6.0)

and temperature (36 �C), yielding Pc at 290 ml H2/g carbohy-

drate and Rmax at 802 ml H2/h. The accuracy of the optimum

conditions was checked by calculating the standard deviation

for each response. The experiment results were close to the

model prediction. Table 5 shows the comparison of hydrogen

gas production and the production rate with previously pub-

lished results. Hydrogen gas production and hydrogen

production rate results were found to be comparable with the

published data.

erature.

rate concentration Pc Rmax (ml H2/h) Reference

94a 315.8b 885.8 This work

10 1.43c 198.6 [26]

7 2.1c nv [17]

20 2.91d 160 [30]

7.5a 46.6e 74.7 [14]

nv 4708f 454 [11]

i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 4 ( 2 0 0 9 ) 7 4 7 5 – 7 4 8 27482

4. Conclusion

C. butyricum EB6 produced hydrogen gas from POME. The

optimized conditions were determined for hydrogen produc-

tion (Pc) and hydrogen production rate (Rmax), using variables

of pH, temperature and COD. It was found that pH, tempera-

ture and COD greatly influenced the hydrogen gas production,

as determined by ANOVA of these variables. The optimized

conditions for hydrogen production from POME using C.

butyricum were pH 6.05, 36 �C and 94 g COD/l, with the esti-

mated hydrogen production (Pc) at 298 ml H2/g-carbohydrate

and hydrogen production rate (Rmax) at 849.5 ml H2/h.

Acknowledgments

The authors are grateful for financial support from the

Ministry of Science, Technology and Innovation, Malaysia;

University Putra Malaysia; Kyushu Institute of Technology

(KIT) and Japan Society for Promotion of Science (JSPS).

r e f e r e n c e s

[1] Penner SS. Steps toward the hydrogen economy. Energy2006;31:33–43.

[2] Solomon Barry D, Banerjaee A. A global survey of hydrogenenergy research, development and policy. Energy Policy 2006;34:781–92.

[3] Nath K, Das D. Hydrogen from biomass. Curr Sci 2003;85:265–71.[4] Kapdan IK, Kargi F. Biohydrogen production from waste

materials. Enzyme Microb Technol 2006;38:569–82.[5] Lay JJ, Lee YJ, Noike T. Feasibiligy of biological hydrogen

production from organic fraction of municipal solid waste.Water Res 1999;33:2579–86.

[6] Van Ginkel SW, Oh SE, Logan BE. Biohydrogen gas productionfrom food processing and domestic wastewaters. IntJ Hydrogen Energy 2005;30:1535–42.

[7] Shin HS, Youn JH. Conversion of food waste into hydrogen bythermophilic acidogenesis. Biodegradation 2005;16:33–44.

[8] Collet C, Adler N, Schwitzguebel JP, Peringer P. Hydrogenproduction by Clostridium thermolacticum during continuousfermentation of lactose. Int J Hydrogen Energy 2004;29:1479–85.

[9] Sumathi S, Chai SP, Mohamed AR. Utilization of oil palm asa source of renewable energy in Malaysia. Renew SustainEnergy Rev 2008;12:2404–21.

[10] Yacob S, Shirai Y, Hassan MA, Wakisaka M, Subash S. Start-upoperation of semi-commercial closed anaerobic digester forpalm oil mill effluent treatment. Proc Biochem 2006;41:962–4.

[11] Atif AAY, Fakhru’l-Razi A, Ngan MA, Morimoto M, Iyuke SE,Veziroglu NT. Fed batch production of hydrogen from palmoil mill effluent using anaerobic microflora. Int J HydrogenEnergy 2005;30:1393–7.

[12] Kim IS, Hwang MH, Jang NJ, Hyun SH, Lee ST. Effect of low pHon the activity of hydrogen utilizing methanogen inbiohydrogen process. Int J Hydrogen Energy 2004;29:1133–40.

[13] Zhang YF, Shen JQ. Effect of temperature and ironconcentration on the growth and hydrogen production ofmixed bacteria. Int J Hydrogen Energy 2006;31:441–6.

[14] Van Ginkel S, Sung SW, Lay JJ. Biohydrogen production asa function of pH and substrate concentration. Environ SciTechnol 2001;35:4726–30.

[15] Dabrock B, Bahl H, Gottschalk G. Parameters affectingsolvent production by Clostridium pasterurianum. ApplEnviron Microbiol 1992;58:1233–9.

[16] Lay JJ. Modeling and optimization of anaerobic digestedsludge converting starch to hydrogen. Biotechnol Bioeng2000;68:269–78.

[17] Fang HHP, Liu H. Effect of pH on hydrogen production fromglucose by a mixed culture. Bioresour Technol 2002;82:87–93.

[18] He GQ, Kong Q, Ding LX. Response surface methodology foroptimizing the fermentation medium of Clostridiumbutyricum. Lett Appl Microbiol 2004;39:363–8.

[19] Montgomery DC. Design and analysis of experiments. US:John Wiley; 2005.

[20] Chong ML, Raha AR, Shirai Y, Hassan MA. Biohydrogenproduction by Clostridium butyricum EB6 from palm oil milleffluent. Int J Hydrogen Energy 2009;34:764–71.

[21] O-Thong S, Prasertsan P, Intrasungkha N, Dhamwichukorn S,Birkeland NK. Optimization of simultaneous thermophilicfermentative hydrogen production and COD reduction frompalm oil mill effluent by thermoanaerobacterium-richsludge. Int J Hydrogen Energy 2008;33:1221–31.

[22] Lay JJ, Li YY, Noike T. The influences of pH and moisturecontent on the methane production in high solids sludgedigestion. Water Res 1997;31:1518–24.

[23] Logan BE, Oh SE, Van Ginkel SW. Biological hydrogenproduction measured in batch anaerobic respirometers.Environ Sci Technol 2002;36:2530–55.

[24] Marichamy S, Mattiasson B. Rapid production of cellulase-free xylanases by solventogenic clostridia from rumen.Enzyme Microb Technol 2005;37:497–504.

[25] Jo JH, Lee DS, Park D, Choe WS, Park MP. Optimization of keyprocess variables for enhanced hydrogen production byEnterbacter aerogenes using statistical methods. BioresourTechnol 2008;99:2061–6.

[26] Mizuno O, Dinsdale R, Hawkes FR, Hawkes DL, Noike T.Enhancement of hydrogen production from glucose bynitrogen gas sparging. Bioresour Technol 2000;73:59–65.

[27] Wang G, Mu Y, Yu HQ. Response surface analysis to evaluatethe influence of pH, temperature and substrateconcentration on the acidogenesis of sucrose-richwastewater. Biochem Eng J 2005;23:175–84.

[28] Mu Y, Wang G, Yu HQ. Response surface methodologicalanalysis on biohydrogen production by enriched anaerobiccultures. Enzyme Microb Technol 2006;38:905–13.

[29] Zinatizadeh AAL, Mohamed AR, Abdullah AZ, Mashitah MD,Hasnain Isa M, Najafpour GD. Process modeling and analysisof palm oil mill effluent treatment in an up-flow anaerobicsludge fixed film bioreactor using response surfacemethodology (RSM). Water Res 2006;40:3193–208.

[30] Chen WM, Tseng ZJ, Lee KS, Chang JS. Fermentativehydrogen production with Clostridium butyricum CGS5isolated from anaerobic sewage sludge. Int J HydrogenEnergy 2005;30:1063–70.