Using the Box-Benkhen Design to Statistically Model Hydrogen Production During Glucose Fermentation in the Presence of Oleic Acid

Hydrogen (H₂) has been projected as a sustainable fossil fuel replacement and its production from non-conventional routes has gain a significant amount of research attention over the past several years.  Over the past several years, many governments have emphasized the need for prioritizing the development of biological processes such as biophotolysis and anaerobic fermentation to produce sustainable energy source.  These biological methods are able to produce hydrogen by enzymatic degradation of organic waste and agriculture residues [1]. Among the various biological routes of hydrogen production from organic compounds, the light-independent dark fermentation is considered economically more favorable due to its substrate non-specificity and faster reaction rates [2]. Moreover, recent studies have shown that the anaerobic process is more efficient than many biological processes in terms of net energy utilization [3].

In anaerobic dark fermentation complex organic substrates (such as hexose) are fermented to hydrogen and organic acids.  However, the process bottleneck is attempting to achieve the theoretical hydrogen yield (4 mol H₂/mol hexose) because of hydrogen consumption by hydrogen consuming microorganism [4]. Hydrogenotrophic methanogens (MB), sulfate-reducing bacteria (HSRB) and homoacetogens are the principal hydrogen consumers in anaerobic mixed microbial systems. The presence of hydrogen consumers causes a loss in electron equivalence by diverting electrons to cell biomass formation and facilitating incomplete oxidation of reduced by-products [5]. Due to lower free energy of hydrogen utilization for HSRB (-38.1 kJ/mol) compared to that of methanogens (-32.7 kJ/mol), HSRB are more efficient hydrogen consumers than methanogens.  However, HSRB also need at least a C₂-organic carbon source in addition to hydrogen for cell synthesis and they are often limited beyond a threshold substrate level [6].  Compared to homoacetogensis (-5 to -6 kJ/mol H₂), methanogenesis (-9 to -12 kJ/mol H₂) is energetically a more favorable route for hydrogen consumption.  In a competitive environment of hydrogen utilization, methanogens are able to outcompete homoacetogens [7]. Hence, increasing the hydrogen yield in mixed anaerobic fermentation by controlling the hydrogen-consuming methanogenic populations is of significant interest to many researchers

Methanogenic growth control in a mixed consortium can be effected either by modulating the operational conditions of the bioreactors or through addition of chemical growth inhibitors or both in combination. Koster et. al., reported inhibition of methanogenic population in a mixed consortium using long chain fatty acids (LCFAs) [8]. Lalman et. al., observed that the addition of LCFAs, such as linoleic acid (LA) (C18:2) and oleic acid (OA) (C18:1) resulted in the accumulation of acetic acid [9,10].  The inhibitory effect of different linoleic acid (LA) (C18:2) levels was examined by Chowdhury et. al. (2007) and these researchers reported that increasing the LA concentration decreases the methanogenic activity and diverted a large fraction of electrons towards hydrogen production [11]. Ray et. al. (2008) also reported that hydrogen production from a mixed mesophilic culture was favoured when LA was added under low initial pH conditions [12].  The study also reported improved hydrogen production on successive glucose injections. Evidently the choice of LCFA and favourable level of operating variables is of utmost importance for the enhancement of hydrogen yield in batch cultures. To date, no study which has assessed the effect of experimental variables and OA concentration on the hydrogen yield of mixed cultures.

Experimentally, the optimum levels of experimental settings can be located by a single factor optimization method. However, the maximization of hydrogen yield through a single factor optimization is often considered less advantageous than optimization using statistical experimental design. Among the available statistical designs, a fractional factorial design (FFD) is a common choice for investigating hydrogen yield. However, FFD is neither very accurate (lacks rotatability) nor economical (require large number of experiments) [13]. Alternately, response surface optimization using the Box-Benkhen design (BBD) has been seldom considered in optimization studies for fermentative hydrogen production [14]. Hence the objective of the present study is to maximize hydrogen production in a mixed mesophilic culture at 37^oC by assessing the effects of initial pH, OA concentration and number of glucose injection using a three factor three level BBD.

The experimental levels for each factor (Table 1) were selected based on literature values and the results from preliminary experiments.  The hydrogen-yield data from BBD experiments conducted in batch using mixed cultures were analyzed using Minitab 15 statistical software (Minitab Inc. State College, PA). Analysis of variance (ANOVA) and regression analysis was performed to derive a response surface model.  Three dimensional surface plots were used to investigate the factor effect on design surface and locate the optimum factors conditions for maximum hydrogen yield.

Table 1:  Factors and levels selected for the experimental study

Anaerobic cultures utilized for this work were sourced from an ethanol manufacturing facility (Chatham, ON) and from a municipal wastewater treatment facility (
Chatham, ON).  The two inoculum were mixed in a 70:30 ratio (ethanol: municipal) in a semi-continuous reactor and maintained at 37±1°C.  The 4-l semi-continuous reactor (denoted as reactors A) with 2,000 mg×l^-1 VSS were prepared by diluting the mixed inoculum.  This reactor served as source of anaerobic mixed culture for all experiments.  A basal medium, adapted from Weigant et. al. was used for feed preparation and diluting the inoculum [15].  The experimental methods reported in this work were adapted from earlier studies published in the literature (9, 10, 16).  Each bottle was filled with 50 ml of basal media plus culture (2,000 mg mg.l^-1 VSS) in an anaerobic glove box (COY Laboratory Products Inc., MI) under a 70% N₂/30% CO₂ (Praxair
Inc.ON) atmosphere. A 50,000 mg.l^-1 OA stock solution was prepared using OA melted au bain-marie in hot NaOH and 100,000 mg×l^-1 glucose stock solution was prepared by dissolving glucose in MQ water.  The pH was adjusted with 1M HCl or 1M NaOH. All controls and culture for the experiment were prepared in triplicates in 160 ml serum bottles. The bottles were sealed in the anaerobic glove box with Teflon^â-lined silicone rubber septa and aluminum crimp caps. Each bottle was over-pressurizing with 20 ml of 70% N₂/30% CO₂ to avoid the formation of a negative pressure in the headspace during sampling. The bottles were agitated using an orbital shaker (Lab Line Instruments Model 3520, IA) at 200 rpm and maintained at 37±2^oC over the duration of the study. Headspace gas samples (25ml) were withdrawn at regular intervals and analyzed using a Varian-3600 (Varian,
Palo Alto, CA) gas chromatograph (GC) configured with a thermal conductivity detector (TCD) to measure the quantity of hydrogen produced.  A 2-m long x 2-mm I.D. Carbon Shin column (Alltech,
Deerfield, IL) was used for conducting the analysis.  The GC injector, detector and oven temperatures were set at 100^oC, 200^oC and 200^oC, respectively. The nitrogen carrier gas flow rate was set at 10 ml min^-1.  The detection limit for hydrogen was 0.0032 kPa.

Increasing OA concentration and low initial pH conditions were observed to enhance the hydrogen yield. Higher hydrogen yields were recorded with two consecutive glucose injections as reported in the literature for cultures receiving LA.  Based on the Box-Benkhen technique, a response surface model was developed for computing the hydrogen yield (mol H₂×mol glucose^-1)from glucose fermentation. The hydrogen yield in the response surface model was expressed in terms of the OA concentration, initial pH and number of glucose injections (three design variables). Significant interactions were observed to exist between the experimental factors.  The model was observed to have a 92.5% correlation (R-square of 85.5%) over the entire factor space under evaluation. A maximum hydrogen yield of 2.61 mol H₂×mol glucose^-1 was predicted with 2 g×l^-1 OA at an initial pH of 5.0 for two glucose injections. The observed hydrogen yield in experiments conducted near the computed optimum factor setting was 2.04 ± 0.27 mol H₂×mol glucose^-1.

References:

1.      Das et al. (2001). Int. J. Hyd. Energy. 26 (1), 13-28.

2.      Kargi et al. (2006). Enzym. Microb. Technol. 38 (5) 569-582.

3.      Hallenbeck et al. (2002). Int. J.  Hyd. Energy., 27 (11-12), 1185-1193.

4.      Manish et. al. (2008). Int. J. Hydrogen Energy. 33 (1), 279 - 286.

5.      Oh et. al. (2005). Water Res. 39 (19), 4673-4682.

6.      Liamleam et. al. (2007). Biotechnol. Adv. 25(5), 452-463.

7.      Conrad et. al. (1990). Arch. Microbiol. 155 (1) 94-98.

8.      Koster et. al. (1987). Appl. Env. Micbiol., 53(2), 403-409.

9.      Lalman et. al.(2000). Water Res. 34 (17), 4220-4228.

10.  Lalman et. al. (2001). Water Res. 35 (12), 2975-2983.

11.  Chowdhury et. al. (2007). J. Environ.
Eng. 133 (12) 1145 ? 1152.

12.  Ray et. al. (2008). J. Environ.
Eng. 134 (2), 110-117.

13.  Myer et. al. (2002). Response surface methodology: Process and product optimization using designed experiment, second ed., John Wiley and Sons,
New York, 343-350.

14.  Wang et. al. (In Press). Int. J. Hydrogen Energy. doi:10.1016/j.ijhydene.2008.10.008.

15.  Weigant et. al. (1985). Biotechnol. Bioeng. 27 (11), 1603-1607.

16.  Lalman et. al. (2002). Water Res. 36 (13), 3307-3313.