Bioresource Technology 97 (2006) 1302–1307

Kinetic modeling of batch hydrogen production process

by mixed anaerobic cultures
Yang Mu, Gang Wang, Han-Qing Yu *

School of Chemistry, University of Science & Technology of China, Hefei, Anhui 230026, China

Received 7 April 2005; received in revised form 18 May 2005; accepted 26 May 2005
Available online 1 August 2005

Abstract

The kinetics of batch anaerobic hydrogen production by mixed anaerobic cultures was systemically investigated in this study.
Unstructured models were used to describe the substrate utilization, biomass growth and product formation in the hydrogen pro-
duction process. The relationship between the substrate, biomass and products were also evaluated. Experimental results show that
the Michaelis–Menten equation, Logistic model and modified Gompertz equation all could be adopted to respectively describe the
kinetics of substrate utilization, biomass growth and product formation. Furthermore, the relationship between the acidogenic prod-
ucts and biomass was simulated by Luedeking–Piret model very well. The experimental results suggest that the formation of hydro-
gen and the main aqueous products, i.e., butyrate and acetate, was all growth-associated.
 2005 Elsevier Ltd. All rights reserved.

Keywords: Anaerobic; Hydrogen; Kinetics; Mixed cultures; Model

1. Introduction substrate (Taguchi et al., 1995). Hydrogen was produced

at a rate of 1600 ml-H2/l/day and a yield of 2.14 mol-
Hydrogen is a clean source of energy. Combustion of H2/mol-hexose when a mixed culture was employed to
hydrogen produces no greenhouse gases, and has a high- ferment a 0.75% solution of soluble starch (Lay, 2000).
energy yield of 122 kJ/g, which is 2.75-fold greater than Kinetic models could be used to describe relationship
that of hydrocarbon fuels. Thus, using hydrogen as a among the principal state variables and to explain the
clean energy source seems to be promising (Nath and behavior of fermentation quantitatively. In addition, it
Das, 2004). The feasibility of applying fermentation of can provide useful information for the analysis, design
organic wastes to produce hydrogen has been widely and operation of a fermentation process. Kinetic models
demonstrated at various laboratories (Yu et al., 2002; are normally divided into two classes: structured and
Hussy et al., 2003; Nath and Das, 2004). Both pure unstructured one. Structured models take metabolic
cultures such as Clostridium sp. and mixed cultures of pathways into consideration and are generally compli-
anaerobic bacteria have been used to convert various cated. A structured model for acetone–butanol fermen-
carbohydrates to hydrogen in previous studies (Taguchi tations was established by Votruba et al. (1985). In
et al., 1992; Ueno et al., 1995; Lay, 2001). For example, this model the biochemical and physiological aspects
a pure strain of Clostridium could produce hydrogen in a of growth and metabolite synthesis were considered
continuous culture at pH 6.0 when xylose was used as a simultaneously, and the key fermentation rates were ex-
pressed and evaluated with regard to substrate con-
sumption and end-product inhibitory effects. On the
*
Corresponding author. Fax: +86 551 360 1592. other hand, the unstructured kinetic models are much
E-mail address: [email protected] (H.-Q. Yu). simpler than the structured ones (Ŏzilgen, 1998). In

0960-8524/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biortech.2005.05.014
 Y. Mu et al. / Bioresource Technology 97 (2006) 1302–1307 1303

the unstructured kinetic models microorganisms are CaCl2 50; MgCl2 Æ 6H2O 100; FeCl2 25; NaCl 10; CoCl2 Æ
usually considered to be a component or reactant in 6H2O 5; MnCl2 Æ 4H2O 5; AlCl3 2.5; (NH4)6Mo7O24 15;
the system. The unstructured kinetic models are the H3BO4 5; NiCl2 Æ 6H2O 5; CuCl2 Æ 5H2O 5; ZnCl2 5. Prior
most frequently employed for modeling microbial sys- to operation, the fermentor was purged with nitrogen gas
tems because they are simple, but are good enough for for 10 min to ensure anaerobic condition.
technical purposes (Chen et al., 2001; Hu et al., 2002; The pH of the mixed liquor was kept at constant by
Fujikawa et al., 2004). feeding 4 M NaOH or 2 M HCl solutions via respective
During the anaerobic hydrogen production from or- peristaltic pumps. The agitation rate in the fermentor
ganic wastes, in addition to hydrogen, carbon dioxide, was kept at 120 rpm. A biogas sampling port was in-
various volatile fatty acids (VFA), and sometimes alco- stalled between the meter and the reactor to allow direct
hols, are simultaneously produced. Thus, this is a com- biogas sampling with a syringe. A 10-ml liquor sample
plex multiproduct process. Depending on operating was taken from reactor at each given interval and was
conditions, the distribution of the acidogenic prod- analyzed. In this trail, the pH was chosen as 5.5, which
ucts varies substantially. Furthermore, some acidogenic was optimum for hydrogen production by mixed an-
products, e.g., acetate, butyrate, hydrogen, carbon aerobic cultures obtained by many previous studies
dioxide, may form more complex long-chain fatty acids (van Ginkel et al., 2001; Fan et al., 2004). The tempera-
or alcohols as the hydrogen is consumed (Yu et al., ture and substrate concentration was 35.0 C and
2004). In this case, it is almost impossible to establish 30 g l1, respectively. Moreover, the trial was replicated
structured models to describe such a multiproduct pro- at least three times and results averaged.
cess, but unstructured models might be a better option.
However, few studies have been carried out on unstruc- 2.3. Analytical methods
tured models to describe the kinetics of biological
hydrogen production by mixed anaerobic cultures. The amount of biogas produced was recorded using
Therefore, in this study unstructured models were water-replace equipment. The biogas contents were ana-
used to describe the kinetics of biological hydrogen pro- lyzed by a gas chromatograph (Lunan, Model SP-
duction from sucrose by mixed anaerobic cultures in a 6800A) equipped with a thermal conductivity detector
batch mode, including substrate utilization, biomass and a 1.5 m stainless-steel column packed with 5 Å
growth and product formation. In addition, relationship molecular sieve. The temperatures of injector, detector
between substrate, biomass and products was also eval- and column were kept at 100 C, 105 C and 60 C,
uated. The results from this study were expected to be respectively. Argon was used as carrier gas at a flow rate
helpful for understanding the behavior of anaerobic of 30 ml min1. The concentrations of VFA, including
hydrogen production process. acetate, propionate, and butyrate in the effluent were
determined by second gas chromatograph (Agilent,
Model 6890NT) equipped with a flame ionization detec-
2. Methods tor and a 30 m · 0.25 mm · 0.25 lm fused-silica capil-
lary column (DB-FFAP). The liquor samples were
2.1. Hydrogen-producing sludge first centrifuged at 12000 rpm for 5 min, and were then
acidified by formic acid and filtrated through 0.2 lm
The anaerobic microflora used in this study was ob- membrane and finally measured for free acids. The
tained from a full-scale upflow anaerobic sludge blanket temperatures of the injector and detector were 250 C
reactor treating citrate-producing wastewater located and 300 C, respectively. The initial temperature of oven
Benpu, China. Prior to use, the seed sludge was heated was 70 C for 3 min followed with a ramp of
at 102 C for 90 min to inactivate the hydrogentrophic 20 C min1 for 5.5 min and to final temperature of
methanogens and to enrich the hydrogen-producing 180 C for 3 min. Nitrogen was used as carrier gas with
bacteria as described by van Ginkel et al. (2001). a flow rate of 2.6 ml min1. Sucrose concentration was
measured using enthrone–sulfuric acid method (Koeher,
2.2. Experiments 1952). VSS concentration was measured according to
the Standard Methods (APHA, 1995).
Hydrogen production experiments were carried out in
a 5-l fermentor (Baoxin Biotech Ltd., Shanghai). One
liter hydrogen-producing sludge with volatile suspended 3. Results and discussion
solids (VSS) of 15.2 g l1 and 3 ml of nutrients solution
were added to the fermentor. The working volume of 3.1. Kinetics of substrate utilization
the fermentor was adjusted to 3.0 l with distilled water.
The solution in the fermentor was composed as follows The Michaelis–Menten equation was used to model
(unit in mg l1): NH4HCO3 2025; K2HPO4 Æ 3H2O 800; the substrate degradation as follows (Wu et al., 2002):
1304 Y. Mu et al. / Bioresource Technology 97 (2006) 1302–1307

vm S where X0 (g l1) is the initial microbial concentration (g-

v¼ ð1Þ
km þ S VSS l1).
where v (g g-VSS1 h1) is the specific substrate degra- Although the predictive power of Eq. (4) may be lim-
dation rate; vm (g g-VSS1 h1) is the maximum specific ited since it does not involve a substrate term (Panikov,
substrate degradation rate; km (g l1) is the dissociation 1995), the logistic model is a fair approximation of the
constant; and S (g l1) is the substrate concentration. microbial growth curve for the batch hydrogen produc-
Linearization of Eq. (1) gives: tion experiments in this study, due to constant values of
the initial substrate concentrations and the inoculation
1 km 1 1 volume. Fig. 2 shows the relationship between the
¼ þ ð2Þ
v vm S vm microbial concentration and cultivation time, including
both experimental data and fitted curve obtained based
Plotting 1/v against 1/S, a straight line was obtained on the logistic model. Correspondingly, the logistic
with an intercept of 1/vm and a slope of km/vm. This plot model parameters of kc and Xmax were estimated as
is shown in Fig. 1, from which vm and km were estimated 0.07 h1 and 9.46 g-VSS l1, respectively. The R2 value,
as 0.28 g g-VSS1 h1 and 13.5 g l1, respectively. The as a measure of the goodness of the fit, was about 0.954,
regression line had a correlation coefficient of 0.963, indicating excellent agreement of the model to the exper-
suggesting the applicability of Eq. (1). The km value imental data. Moreover, the measured Xmax, 9.14 g-
represents the substrate concentration required to VSS l1, was close to the estimated value by the Logistic
achieve 50% of the maximum specific substrate degrada- model, 9.46 g-VSS l1.
tion rate. This item could be used for adjusting the most
appropriate substrate concentration in the feed. 3.3. Kinetics of product formation

3.2. Kinetics of microbial growth In this study, the desirable gaseous product from the
acidogenesis of sucrose was hydrogen, whereas the main
The Logistic model (Eq. (3)) was adapted to describe aqueous products were butyrate and acetate. Therefore,
the kinetics of the microbial growth (Fujikawa et al., the formation kinetics of hydrogen, butyrate and acetate
2004): were focused. A modified Gompertz equation was em-
  ployed to model the formation of product (van Ginkel
dX X
¼ kcX 1  ð3Þ et al., 2001; Lin and Lay, 2004):
dt X max   
Rmax;i  e
where kc (h1) is the apparent specific growth rate; X (g- P i ¼ P max;i exp  exp ðki  tÞ þ 1 ð5Þ
P max;i
VSS l1) is the microbial concentration; and Xmax (g l1)
is the maximum microbial concentration. Integration of where i represents hydrogen, butyrate and acetate
Eq. (3) gives the following equation for microbial respectively; Pi is the product i formed per liter of reac-
concentration: tor volume at fermentation time t; Pmax,i is the potential
maximum product formed per liter of reactor volume;
X 0 expðk c tÞ Rmax,i is the maximum rate of product formed; is the
X ¼ ð4Þ
1  ðX 0 =X max Þð1  expðk c tÞÞ lag time to exponential product formed.

24 10

9
Microbial concentration

18
1/v (g-VSS h g-1)

(g-VSS l-1)

12 2
7 R = 0.954
y = 48.3x+3.58
2
R = 0.963
6
6

5
0.0 0.1 0.2 0.3 0.4 0 20 40 60 80 100
1/S (l g-1) Fermentation time (h)

Fig. 1. Linearized Michaelis–Menten equation plot for the degrad- Fig. 2. Logistic model plot for the growth of hydrogen-producing
ation of sucrose. microorganisms.
 Y. Mu et al. / Bioresource Technology 97 (2006) 1302–1307 1305

The fitted curves are shown in Fig. 3 and the param- 3.4. Relationship between biomass and products
eters estimated for the formation of various products
are summarized in Table 1. All correlation coefficients The relationship between biomass and the products
of non-linear analysis by Eq. (5) was over 0.990, sug- for the anaerobic hydrogen production by mixed anaer-
gesting that the modified Gompertz equation was able obic cultures could be simulated by the Luedeking–Piret
to adequately describe the formation of various prod- model:
ucts for anaerobic hydrogen production by mixed dP i dX
cultures. ¼ ai þ bi X ð6Þ
dt dt
where ai is the growth-associated formation coefficient
of product i; bi is non-growth-associated formation coef-
5000 ficient of product i.
Hydrogen This model was originally proposed for describing the
4000 lactic acid production by Lactobacillus delbrucckii
(Luedeking and Piret, 1959). The first term of Eq. (6),
P (ml l-1)

3000
i.e., ai dX/dt, is referred as the growth-associated forma-
2000 tion rate of product i. This implies that the growing cells
produce the product in constant proportion of their
1000 P = 4335*exp{-exp[201*e/4335*(9.3-t)+1]}
2
growth. Non-growth associated product formation term
R = 0.999 (biX) shows that all the microorganisms produce the
0
product i in a constant proportion of their concentra-
tion, regardless of the growth phase. Eq. (6) could be
10
Butyrate changed into:
8 dP i dX
¼ ai þ bi
X dt X dt
P (mg l-1)

6
or
4
dP i
¼ ai l þ b i ð7Þ
2 P = 9.30*exp{-exp[0.51*e/9.30*(8.5-t)+1]} X dt
2
R = 0.997 where dPi/dt/X is the specific formation rate of product i;
0
and l is the specific growth rate of the microorganisms.
1.0 A straight line could be obtained with an intercept of
Acetate bi and a slope of ai, if plotting dPi/dt/X against l. The
0.8 fitted plots are shown in Fig. 4 for the three products
(hydrogen, butyrate and acetate). The high correlation
P (mg l-1)

0.6
coefficients indicate that the Luedeking–Piret model
0.4 could properly describe the relationship between bio-
mass and product in the anaerobic hydrogen production
0.2 P = 0.92*exp{-exp[0.043*e/0.92*(9.5-t)+1]} process. The estimated values of a and b, indicating the
2
R = 0.999 dependence of product formation and biomass concen-
0.0
0 20 40 60 80 100 tration, are shown in Table 2. These values could be
Fermentation time (h) used to predict the productivity in the bioreactor design.
Since all b values were zero in Table 2, the formation of
Fig. 3. Modified Gompertz equation plot for the formation of
acidogenic products.
hydrogen, butyrate and acetate in the anaerobic hydro-
gen production was associate-growth.

3.5. Relationship between substrate and products

Table 1
Calculated results using the modified Gompertz equation for the Since the formation of product was associate-growth
formation of acidogenic products for the anaerobic hydrogen production by mixed anaer-
Product Pmax (g l1) Rmax (g l1 h1) k (h) R2 obic cultures, the correlation between product i and sub-
Hydrogen 4335 a
201b
9.4 0.999 strate could be expressed as Eqs. (8) and (9) (Yang et al.,
Butyrate 9.3 0.51 8.5 0.997 1988):
Acetate 0.92 0.043 9.5 0.999
a
Unit in ml l1.
dP i dS
¼ Y P i ð8Þ
b
Unit in ml l1 h1. dt dt
1306 Y. Mu et al. / Bioresource Technology 97 (2006) 1302–1307

30 5000
Hydrogen Hydrogen
4000
(ml g-VSS h )
-1 -1

20
dP/dt/X

P (ml l )
3000

-1
10 2000
y = 759x
2
R = 0.834 1000 y = 143x
2
0 R = 0.998
0
0.10
Butyrate 10
0.08 Butyrate
(g g-VSS h )
-1 -1

8
dP/dt/X

0.06

P (g l )
6

-1
0.04
y = 2.45x
4
0.02 2
R = 0.890
y = 0.35x
0.00 2 2
R = 0.995
0
0.008
Acetate
1.0
0.006
(g g-VSS h )
-1 -1

Acetate
dP/dt/X

0.8
0.004
0.6
P (g l )
-1

0.002 y = 0.15x
2
R = 0.964 0.4
0.000
0.00 0.01 0.02 0.03 0.04 0.2 y = 0.03x
2
-1 R = 0.999
µ (h ) 0.0
0 10 20 30
Fig. 4. Luedeking–Piret model plot for the relationship between the
-1
products and microorganisms. S0-S (g l )

Fig. 5. Plot for the relationship between the products and sucrose.
Table 2
Calculated results using the Luedeking–Piret model for the relation-
ship between the acidogenic products and biomass
Product a (g g-VSS1) b (g g-VSS1 h1) R2
a b
Table 3
Hydrogen 759 0 0.834 Calculated results from Eq. (8) for the relationship between the
Butyrate 2.45 0 0.890 products and substrate
Acetate 0.15 0 0.964
Product YP (g g1) R2
a
Unit in ml g-VSS1.
b
Unit in ml g-VSS1 h1. Hydrogen 143a 0.998
Butyrate 0.35 0.995
Acetate 0.03 0.999
a
Unit in ml g1.
dP i ¼ Y P i dS
Z Pi Z S
dP i ¼ Y P i dS ð9Þ
0 S0
successfully. The yields of hydrogen, butyrate and
acetate were estimated as 143 ml g1, 0.35 g g1 and
P i ¼ Y P i ðS 0  SÞ
0.030 g g1, respectively.
where Y P i is the yield of product i.
Solution of Eq. (9) is plotted in Fig. 5 and the esti-
mated YP values for hydrogen, butyrate and acetate 4. Conclusions
are summarized in Table 3. The high correlation coeffi-
cients in Table 3 suggest that the relationship between With the experimental results of batch hydrogen pro-
products and substrate could be simulated by Eq. (8) duction by mixed anaerobic cultures, unstructured mod-
 Y. Mu et al. / Bioresource Technology 97 (2006) 1302–1307 1307

els were used to describe the substrate utilization, bio- Lay, J.J., 2000. Modeling and optimization of anaerobic digestion
mass growth and product formation in this process. sludge converting starch to hydrogen. Biotechnol. Bioeng. 68, 269–
278.
The Michaelis–Menten equation, Logistic model and Lay, J.J., 2001. Biohydrogen generation by mesophilic anaerobic
modified Gompertz equation could respectively describe fermentation of microcrystalline cellulose. Biotechnol. Bioeng. 74,
the kinetics of substrate utilization, biomass growth and 280–287.
product formation very well, while Luedeking–Piret Lin, C.Y., Lay, C.H., 2004. Effects of carbonate and phosphate
model was able to simulate the relationship between aci- concentrations on hydrogen production using anaerobic sewage
sludge microflora. Int. J. Hydrogen Energy 29, 275–281.
dogenic products and biomass. The experimental results Luedeking, R., Piret, E.L., 1959. A kinetic study of the lactic acid
also suggest that the formation of hydrogen, butyrate fermentation: batch process at controlled pH. J. Biochem. Micro-
and acetate was all growth-associated. biol. Technol. Eng. 4, 231–241.
Nath, K., Das, D., 2004. Improvement of fermentative hydrogen
production: various approaches. Appl. Microbiol. Biotechnol. 65,
Acknowledgements 520–529.
Ŏzilgen, M., 1998. Kinetics of multiproduct acidogenic and solvento-
genic batch fermentations. Appl. Microbiol. Biotechnol. 29, 536–
The authors wish to thank the Natural Science Foun- 543.
dation of China (Grant No. 20377037), the National Panikov, N.S., 1995. Microbial Growth Kinetics. Chapman & Hall,
Basic Research Program of China (Grant No. London.
2004CB719602), and the Trans-Century Training Pro- Taguchi, F., Chang, J.D., Taguchi, S., Morimoto, M., 1992. Efficient
gram Foundation for the Talents, Ministry of Educa- hydrogen production from starch by a bacterium isolated from
termites. J. Ferment. Technol. 73, 244–245.
tion, China, for the financial support of this study. Taguchi, F., Mizukami, N., Yamada, K., Hasegawa, K., Saito-Taki,
T., 1995. Direct conversion of cellulosic materials to hydrogen by
Clostridium sp. strain No. 2. Enzyme Microbiol. Biotechnol. 7,
References 147–150.
Ueno, Y., Kawai, T., Sato, S., Otsuka, S., Morimoto, M., 1995.
APHA, AWWA, WEF, 1995. Standard Methods for the Examination Biological production of hydrogen from cellulose by natural
of Water and Wastewater, 19th ed. American Public Health anaerobic microflora. J. Ferment. Bioeng. 79, 395–397.
Association, Washington, DC. van Ginkel, S., Sung, S., Lay, J.J., 2001. Biohydrogen production as a
Chen, C.C., Lin, J.S., Chang, J.S., 2001. Kinetics of hydrogen function of pH and substrate concentration. Environ. Sci. Technol.
production with continuous anaerobic cultures utilizing sucrose 35, 4726–4730.
and the limiting substrate. Appl. Microbiol. Biotechnol. 57, 56–64. Votruba, J., Volesky, B., Yerushalmi, L., 1985. Mathematical model of
Fan, Y.T., Li, C.L., Lay, J.J., Hou, H.W., Zhang, G.S., 2004. batch acetone–butanol fermentation. Biotechnol. Bioeng. 28, 247–
Optimization of initial substrate and pH levels for germination of 255.
sporing hydrogen-producing anaerobes in cow dung compost. Wu, S.Y., Lin, C.N., Chang, J.S., Lee, K.S., Lin, P.J., 2002. Microbial
Bioresour. Technol. 91, 189–193. hydrogen production with immobilized sewage sludge. Biotechnol.
Fujikawa, H., Kai, A., Morozumi, S.A., 2004. New logistic model for Prog. 18, 921–926.
Escherichia coli growth at constant and dynamic temperatures. Yang, S.T., Tang, I.C., Okos, M.R., 1988. Kinetics and mathematical
Food Microbiol. 21, 501–509. modeling of homoacetic fermentation of lactate by Clostridium
Hu, W.C., Thayanithy, K., Forster, C.F., 2002. A kinetic study of the formicoaceticum. Biotechnol. Bioeng. 32, 797–802.
anaerobic digestion of ice-cream wastewater. Process Biochem. 37, Yu, H.Q., Zhu, Z.H., Hu, W.R., Zhang, H.S., 2002. Hydrogen
965–971. production from rice winery wastewater in an upflow anaerobic
Hussy, I., Hawkes, F.R., Dinsdale, R., Hawkes, D.L., 2003. Contin- reactor by using mixed anaerobic cultures. Int. J. Hydrogen Energy
uous fermentative hydrogen production from a wheat starch co- 27, 1359–1365.
product by mixed microflora. Biotechnol. Bioeng. 84, 619–626. Yu, H.Q., Mu, Y., Fang, H.H.P., 2004. Thermodynamic analysis on
Koeher, L.H., 1952. Differentiation of carbohydrates by anthrone formation of alcohols and high-molecular-weight VFA in acido-
reaction rate and color intensity. Anal. Chem. 144, 1231–1240. genesis of lactose. Biotechnol. Bioeng. 87, 813–822.