Fermentative Hydrogen Production by Molasses; Effect of Hydraulic Retention Time, Organic Loading Rate and Microbial Dynamics

© 2012 Mariakakis et al., licensee InTech. This is an open access chapter distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Fermentative Hydrogen Production by Molasses; Effect of Hydraulic Retention Time, Organic Loading Rate and Microbial Dynamics


Introduction
The generation of hydrogen by biological means is not energy intensive compared with the conventional thermochemical techniques, since the operating temperature and pressure are not very high. As raw materials organic waste streams can be used that can be considered as a renewable resource (Vijayaraghavan & Mohd Soom, 2006).The method of the dark fermentation has certain advantages compared with the other biological processes. In contrast to bio-photolysis and photo fermentation, the process needs no solar radiation, but the required energy is supplied by the organic substrates and hence the process is not interrupted during the night. Moreover, the production rate of the H2 of the fermentative bacteria in comparison with the other biological processes is greater (Kumar et al., 2000;Nath et al., 2005).
The different process parameters that are relevant for hydrogen production have been surveyed (Li & Fang, 2007;Wang & Wan, 2009) and include the type of substrate, nutrient concentration, the inoculum, pH, reactor configuration, hydraulic retention time (HRT), organic loading rate (OLR). Carbohydrate-rich substrates are the most suitable for fermentative H2 production systems (Hawkes et al., 2002;Kapdan & Kargi, 2006;Meherkotay & Das, 2008;Ueno et al., 2007) seeded with saccharoclastic microorganisms, They are able to break down organic substances via the Embden-Meyerhof pathway resulting to different metabolic products depending on the type of microorganism and the environmental conditions driving their catabolism (Hallenbeck, 2009).
The relevant microbial groups for the fermentative hydrogen production groups are clostridia and enterobacteria (Hallenbeck, 2005;Hawkes et al., 2007). Both groups were repeatedly experimentally confirmed as major hydrogen producers (Valdez-Vazquez & Poggi-Varaldo, 2009 Enteric bacteria are gram-negative rods, facultative aerobic, with relatively simple nutrient requirements and can not form spores (Schmauder, 1992). Among the species that can produce H2, are Escherichia (E. coli), Proteus (P. vulgaris), Enterobacter (E. aerogenes). Enteric bacteria ferment sugars to a variety of end products such as acetate, formate, lactate, succinate, ethanol, CO2 and H2. Hydrogen is produced according to equation 1 (Li & Fang, 2007). The maximum possible hydrogen yield by this pathway is 2 mol H2 per mol hexose. In experiments with intestinal bacteria rather half of this value has been found (Hallenbeck, 2005).
Clostridia are spore forming, gram-positive bacteria (Schmauder, 1992). Through sporulation they can survive for example dehydration, heat and large changes in pH. Clostridial catabolism includes a variety of reactions and hence fermentation endproducts such as acetate, acetone (C. pasteurianum), butyrate (C. butylicum), butanol (C. acetobutylicum) or caproic acid (C. kluyveri) (Schmauder, 1992). Hydrogen, using hexose as a substrate can be produced by two pathways with acetate and butyrate as end-products as equations 1 and 2 describe (Hallenbeck, 2005;Hawkes et al., 2007;Li & Fang, 2007). That fact that clostridia can produce higher amount of hydrogen makes them more attractive and hydrogen systems aiming at their growth must be strived. In experiments with mixed cultures yields between 1.5 mol H2/mol hexose and 2.5 mol H2/mol hexose were achieved (Wang & Wan, 2009). In practice the highest yield is achieved when the catabolism is driven through a mixed acetate butyrate according to equation 4 (Lengeler, 1999). This is because hexose can be also metabolized by hydrogen neutral fermentation pathways with lactic acid, ethanol, acetone, butanol as end-products, or a portion of the substrate is consumed for the production of biomass, which theoretically can be 18.5% and 14.5% of the theoretical one through the acetate and butyrate hydrogen producing pathway respectively (Aceves-Lara et al., 2008). More over, a part of hydrogen that is already produced may be consumed by certain microorganisms as homoacetogens (equation 5) with acetate as end-product (Dworkin et al., 2006), or propionic acid bacteria (equation 6) (Li & Fang, 2007), or sulfate reducing bacteria (equation 7). The co-existence of microorganisms other than hydrogen producing that compete for substrate has been observed in many hydrogen producing systems (Hawkes et al., 2007;Hung et al., 2011aHung et al., , 2011bLi et al., 2011). A major part of them belongs to the lactic acid bacteria (LAB) distinguished to heterofermentative LAB, which produce lactic acid together with CO2 (equation 8) and minor quantities of ethanol and acetic acid and the homofermentative LAB, which produce mainly lactic acid (equation 9) (Martinko & Clark , 2009). Through the monitoring of the metabolites of carbohydrate solely is not always possible to determine the metabolic pathays used by the bacteria, since many clostridia are capable of secondary fermentation. C. propionicum can for instance metabolize lactic acid for the production of propionate and acetate (equation 10), while some homoacetogens utilize ethanol and CO2 yielding acetate (equation 11). In some cases, hydrogen can be also produced by secondary fermentation C. kluyveri can utilize ethanol and acetate yielding hexanoic acid and molecular hydrogen by the oxidation of reduced ferredoxin (Fd) (equation 12) and C. tyrobutyricum can transform lactate and ethanol to butyrate (equation 14) (Martinko & Clark, 2009). It is therefore obvious that biological hydrogen production systems are complicated in terms of biological processes and microbial species involved.
Aim of this work was to experimentally study the effect of HRT and OLR on biohydrogen production in terms of maximization of H2 yield, so as to optimize substrate utilization efficiency that contributes to the cost effectiveness of the process. Experiments were carried out in large-lab scale reactors of 30 L working volume. In this way, experience in the start-up procedure, selection of only H2 producing microorganisms and the stability of long term continuous operation without methanogenesis of such a set-up could be gained. This can further be used for the scale-up of bio-hydrogen production towards the final aim of commercial implementation. An attempt towards the clarification of the possible metabolic pathways and the involved microorganisms, with along as their possible interactions was also undertaken. The way that these microorganisms behave and interract with each other and their milieu is very important for the design of effective hydrogen producing systems. The understanding of these processes can help the designer to manipulate hydrogen production by the suitable variation of the process parameters.
With the use of molecular biological techniques it is possible to acquire better insights into such systems.

Experimental set-up
Dark fermentation experiments were conducted in two identical 40 L reactors made of borosilicate glass (QVF) with a working volume of 30 L heated at 37 °C ± 2 °C by a heating pipe. The content of the fermenter was homogenized by external recirculation with eccentric screw pumps (Netsch The HRT was independent of substrate dosing. It was regulated by the pumping (Prominent, Gamma/L) of tap water and automatic removal of excess mixed liquor by a peristaltic pump (Ismatec, MPC Standard) controlled by a water lever sensor (Endress & Hauser, Liquiphant). The tab water was stored in containers and was daily sparged with N2 in order to reduce the dissolved oxygen concentration bellow 1 mg/L. The produced biogas quantity was measured with a drum-type gas meter (Ritter, TG 05) and registered into the programmable controller. The produced gas was collected in gas bags (Lindte). In Figure 1 the experimental set-up is presented.

Reactor operation
The inoculum of the reactor has been acquired from the anaerobic digester of the Sewage Treatment Plant for Research and Education (LFKW) of the University of Stuttgart (Germany). It was diluted to 2% to 4% Total Solids (TS) concentration and sieved consecutively through 4 mm and 2 mm mesh size to prevent clogging of the tubing. It was then pretreated for 24 h at 105 °C, in order to kill the methanogenic bacteria. The dried sludge was pulverized and solved into tab water for the start-up of the system. As substrate sugar beet molasses acquired from a sugar factory in south Germany were used. In Table 2 the composition of molasses is presented. For the start-up of the system, the reactor was fed with 450 g of sucrose in a batch mode at pH 6.5 in order to enrich the biomass in H2producing microorganisms. Upon sucrose depletion continuous operation of the system was started. The pH was reduced to 5.5, a value that has been reported to be the optimum for continuous bio-hydrogen production (Mariakakis et al., 2011). The first phase of continuous operation aimed at the further selection of the biomass for hydrogen producing bacteria by application of high HRT and low OLR. The various experimental conditions tested during the continuous operation are presented in table 3. Their selection was based depending on the experimental progress as described in chapter 3.1. In many cases one of the two reactors had to be re-inoculated with seed sludge acquired by the other reactor. All phases had a minimum duration of 5 times the applied HRT. At phase DVII (

Analytical methods
The analyses of concern were determined according to the german standards (Deutsches Institut fuer Normung, 2002) and performed three times a week. These included; total solids (TS), volatile suspended solids (VSS), chemical oxygen demand (COD), a group parameter used for the detection of carbonaceous matter and nitrogen (in total and soluble form acquired after filtration through membrane with 0.45 μm pore diameter). Glucose, sucrose and lactic acid have been determined spectrophotometrically after enzymatic digestion by test kits according to the manufacturer's instructions (R-Biopharm). Gas Chromatography was used to analyze organic acids and alcohols. The sample was filtered through a 0.45 μm pore diameter filter and acidified with a 96% H2SO4 solution. Organic acids, ethanol and butanol were detected by GC (Perkin Elmer) mounted with a fused silica capillary (Varian) and using a flame ionization detector. Both the injection and capillary temperatures were set at 280 °C. Biogas composition was determined once daily after up-grading for particulate matter and water vapor removal by a gas analyzer (ABB, AO2020), equipped with an infrared detector for CH4 and CO2 and a thermal conductivity detector for H2.
DNA extraction, 16S rDNA of eubacteria and clostridia PCR amplification, DGGE analysis and sequencing were performed as previously described (Mariakakis et al., 2011).

Reactor operation
In figure 2, the reactor operation parameters, gas production and hydrogen yield performance for the experimental phases DI to DIII are presented.
After the start-up of the reactor for a week, the HRT and OLR were reduced to 4 d and 11.6 g sucrose / (L•d) respectively. After about 15 d a continuous H2 production could be established (phase DI), which stabilized after approximately 30 d. The average soluble COD concentration in the reactor was app. 60 g/L and the average H2 yield 2.47 mol H2/mol hexose.

Figure 2.
Operation parameters, gas production, hydrogen yield performance and production rate of relevant metabolites during phases DI to DIII. Arrows indicate the sampling dates for microbial population analyses Table 3. Experimental phases, operating parameters and average concentrations and yields of produced gases and metabolites On the 41 st day of operation the OLR was increased from 11.6 g sucrose / (L•d) to 24.7 g sucrose / (L•d) (phase DII), which caused an increase to the residual sucrose concentration in the reactor during the first days. The biomass was not able to metabolize the total amount of substrate. Over the total 35 d of operation of this phase H2 in the produced biogas gradually decreased and after 70 d no H2 was detected. The fact that biogas production was sustained, indicated biological activity, but not towards H2 production. In phase DIII the HRT was reduced to 2 d with the OLR retained unchanged. The concentration of soluble COD decreased gradually from 100 g/L to 60 g/L due to the higher dilution rate. Nevertheless, H2 production could not be restored and the reactor operation was terminated.
After one month of operation the excess sludge from reactor a (R-a) was used to inoculate reactor b (R-b) (phases D1 to D7). In figure 3 the reactor operation parameters, gas production and hydrogen yield performance for the experimental phases D1 to D7 together with the data from phase DI for continuity and comparison reasons are presented. Biogas and hydrogen production started immediately upon seeding of the reactor at HRT of 3 d and OLR of 11.6 g sucrose / (L•d). Hydrogen production was stable, but the yield was lower than 1 mol H2/mol hexose. The concentration of soluble COD did not exceed 45 g/L. Increase of the OLR at 24.7 g sucrose / (L•d) lead to steep increase of the soluble COD and COD-sucrose concentration above 120 g/L and 60 g/L respectively and a decrease of the COD concentration of the metabolites. It seems that in the beginning the biomass concentration was not sufficient to metabolize the whole amount of sucrose, which on its turn accumulated in inhibitory levels as demonstrated by the cease of biogas and hydrogen production and the reduction of the COD-organic acids. It has been reported (Hafez et al., 2010) that glucose can become inhibiting for residual concentrations above 20 g/L. This problem could be overcome by the reduction of HRT to 1 d (phases D3 and on). H2 production was restored immediately and was maintained until the termination of the experiment due to time constraints. Biogas and hydrogen production exhibited fluctuations over time. These can be generally attributed to the experimental procedure followed, which affected the actual HRT and OLR. The dosing of water and substrate was stopped everyday for different periods each time, in order to be prepared as described in Materials and methods. Stronger fluctuations were related mostly to the failure of substrate dosing due to tube clogging. In all phases, except phases D4 and D6, the H2 yield was equal or lower than 1 mol H2/mol hexose (table 3). In phase D4 it reached 1.53 mol H2/mol hexose for HRT of 1 d and OLR of 36.1 g sucrose / (L•d) and in phase D6 1.31 mol H2/mol hexose for for HRT of 0.5 d and OLR of 36.1 g sucrose / (L•d).
During the whole R-b operation, which together with the time period of phase DI reached 180 d, no methane was detected. Methanogenesis could be inhibited through the thermal pre-treatment of the seed sludge and the selected operation parameters were sufficient for hindering the proliferation of the methanogens in the system. In our previous work for which no pre-treatment of the seed sludge was carried out methanogenesis could be only be inhibited for 120 d (Mariakakis et al., 2011). Sucrose in higher concentrations could be detected only in phases D2 and D7 resulting in average substrate degradations of 71.5% and 83.1% respectively. In D2 sucrose accumulated in the beginning of the phase as results of the long HRT and the OLR of 24.7 g sucrose / (L•d). At D7 sucrose could not be degraded due to the high OLR of 69.6 g sucrose / (L•d) and the short HRT of 0.25 d.
The excess sludge from phase D3 of R-b was used to re-inoculate R-a (phases DV and DVI). In phase DVII the reactor was inoculated with excess sludge of phase D6. In Figure 4 the reactor operation parameters, gas production and hydrogen yield performance for these experimental phases together with the data from phase D3 for continuity and comparison reasons are presented.  Biogas and hydrogen production started also in these cases immediately after seeding of the reactor. Due to the OLR set at 47.1 g sucrose / (L•d), the average soluble COD concentration reached 61 g/L. H2 production (phase DV) yieldied 0.43 mol H2/mol hexose. The decrease of the OLR to 36.1 g sucrose / (L•d) and increase of HRT to 2 d in phase lead to further increase of the soluble COD to an average concentration of 98 g/L. Hydrogen yield was slightly improved and reached 0.63 mol H2/mol hexose. Biogas and hydrogen production during phase DVII were instable mainly due to malfunctions of the substrate dosing. The hydrogen yield for addition of Fe 2+ achieved was 1.68 mol H2 / mol hexose, which corresponds to an increase of approximately 28 % in comparison to phase D6.
The addition of Fe 2+ enhanced hydrogen production. From table 3 it can be seen that biomass production rate reached the highest value of 161 g/d comparable only with that of phase D7 that was acquired for double as much OLR. For comparison at phase D6, biomass production rate was only 94 g/d. It seems that this nutrient that is important for H2 production as will be explained in chapter 3.2 was limiting for the biomass growth, since it was not included in high enough concentrations in the nutrient solution with which the molasses solution was supplemented.
In Table 4 the hydrogen yield of other works in comparison to the best acquired by this work are presented. In most of the cases slightly better results were acquired by the other researchers. In a semi-scale reactor operated at HRT of 0.25 d and pH 4.5 hydrogen yield up to 1.86 mol H2/mol hexose was achieved (Ren et al., 2006). Lay et al. (2010) maximized the yield for HRT of 0.5 d and pH of 5.5 like in this work, but achieved a somewhat lower yield of 1.35 mol H2 / mol hexose, even though the reactor set-up was considerably smaller permitting for better control of operation. Aceves-Lara et al (2008)

Microbial metabolism
From Figures 2 to 4 and Table 3 it can be seen that the major metabolites for all tested parameters were lactic, acetic, and butyric acid together with ethanol. Propionic and hexanoic acid were produced in low quantities and only in same cases. Pentanoic acid and butanol were not detected at all. Acetate and butyrate are typical for H2 production by mixed acid fermentation of sucrose as already mentioned in introduction. Lactate can be produced by either lactic acid bacteria or by bacteria of the genus Clostridium via a pathway that does not promote H2 production (Hiligsmann et al., 2011).
During phase DI lactate was produced in less quantity than acetate and butyrate with the later being produced in almost the same proportion indicating a mixed acid fermentation, even though the theoretical yield of 2.6 mol H2 / mol hexose was not reached (equation 4). The increase of the OLR in phases DII and DIII has lead to gradually increasing lactate production rate with simultaneous increase of acetate into a lesser extent and decrease of butyrate. The higher substrate availability forced the biomass to shift its metabolism from mixed acid fermentation to lactate fermentation. The fact that biogas production also declined is an indicator for homolactic fermentation by LAB, during which only lactate is produced (equation 9). In phases DV and DVI lactic acid production could not be solely due to homolactic fermentation, since biogas was produced. Higher lactic acid production than acetic and butyric acid production was observed in the phases for which HRT and OLR were equal or higher than 1 d and 24.7 g sucrose / (L•d) respectively. The exact mechanism of lactic acid production can be explained either by the co-existence of LAB (Hafez et al., 2009) or by the metabolic shift of the hydrogen producing clostridia (Minton & Clarke, 1989). In all cases though, high lactic acid production was combined to an increase of the OLR (Oh et al. 2004;Kim et al., 2006;Oh et al., 2004;Hafez et al., 2009) and caused a diminution of hydrogen yield. The effect of lactic acid as a metabolite in hydrogen systems has not yet been clarified. It has been reported to be promoting to hydrogen production at low concentrations and inhibiting at high concentrations. In a work , an increase of the hydrogen yield combined with the complete degradation of the externally added lactic acid in concentrations up to 3 g/L was observed. In another work, Kim et al. (2012) also observed an increase of 22% in hydrogen yield when lactic acid up to 8 g/L was added to batch fermentors operated at pH of 4.5, and a reduction when the concentration was raised at 18 g/L. The corresponding undissociated form of the lactic acid, which is the potential inhibitor ( van Ginkel & Logan, 2005) was 21 mmol/L and 45 mmol/L at pH 4.5 according equation 15. In this work, the highest lactic acid concentration was 35 g/L (9 mmol/L undissociated lactic acid at pH 5.5) and was only reached temporarily in the beginning of phase D3 without any obvious long-term negative influence on the hydrogen process, like in phase DIII. It seems that lactic acid was not the inhibition factor, but another substance that was not monitored.
For phases DI to DIII and D1 to D3 during which OLR lower or equal to 24.7 g sucrose / (L•d) was applied, acetate and ethanol were produced in almost the same rates. A ratio of EtOH:HAc equal to 1:1 has also been proposed for clostridia as described by equation 1 for enteric bacteria, when hydrogen is evolved only through the conversion of Acetyl-CoA to pyruvate, yielding 2 mol H2 / mol hexose (Minton & Clarke, 1989). For higher loadings the observed ethanol production rate diminishes either due to a metabolic shift of the biomass or due to its consumption. Ethanol has also been detected in other hydrogen producing systems operated at pH 5.5 and various HRT and OLR (Gavala et al., 2006;Karadag & Puhakka, 2010a;Kim et al., 2006;Shen et al., 2009) as a product either of enterobacterial (Hallenbeck, 2005), clostridial hydrogen production (Akutsu et al., 2009;Lin & Lay, 2004), or heterolactic bacteria (Kandler, 1983). In the case of clostridia, it has been suggested that ethanol is produced during the late growth phase during which no hydrogen is produced, while H2 production is favored during the exponential growth phase, during which the organic acids are produced (Nath & Das, 2004), yielding at the end of a batch fermentation a ratio of 1:1. By the adjustment of short HRT in CSTR it is possible to maintain the bacterial population in the exponential growth phase. However, the adjustment has to be suitable so that the biomass concentration in the reactor can be also maintained in concentrations that are suitable for high substrate conversion rates.
Propionic acid was produced in phases D3 and D4. Sucrose fermentation to propionic acid is a sink for hydrogen and according to equation 6 for each mol of propionic acid produced 1 mol of hydrogen is consumed. The derived hydrogen consumptions for phases D3 and D4 correspond to 15% and 19% of the total produced hydrogen respectively.
Hexanoic acid was mainly produced during phases DI, DII, DVI and D2. The HRT of all these phases was equal to or longer than 2 d. The production of hexanoic acid can only be explained by a possible secondary fermentation of C. kluyveri as described by equation 12 and 15.
The addition of Fe 2+ in phase DVII did not influence the production rates of acetic acid in comparison to phase D6. On the other hand, the production rate of lactic acid was significantly reduced from 1230 mmol/d to 329 mmol/d corresponding to approximately 72% and ethanol was not produced anymore. Parallel, the production rate of butyrate increased approximately by 27%. The overall bacterial metabolism was shifted to butyrate fermentation and biomass growth as described in 3.1. This is an indication that lactic acid and ethanol production in the phases with relative short HRT (<1 d) was mainly due to the clostridial metabolism. They contributed more than 2/3 to the total lactic acid production. Iron is very important to hydrogen production, which is produced when the simple reaction of Eq. 16 takes place. This reaction is catalyzed in clostridia by a dimetallic iron only [FeFe]-hydrogenase, which receives protons by the reduced form either of ferredoxin or of NADH (Vignais & Billoud, 2007). Under iron limitation the activity of hydrogenase is also limited (Valdez-Vazquez & Poggi-Varaldo, 2009), pyruvate can not be degraded through the pathways leading to hydrogen, but fermentation is shifted towards lactic acid production (Minton & Clarke, 1989).

Species description and influence of HRT and OLR
The sample times for the phylogenetic analysis and the affiliated dominant species present in each experimental phase are indicated in figures 2 to 4. In figures 5 and 6 the PCR-DGGE profiles of the Eubacteria and Clostridium species are presented. The investigation on eubacterial 16S rDNA (table 5) showed mostly lactic acid bacteria of the Phylum Firmicutes as closest relatives to the detected DGGE bands. The analysis of 16S rDNA for clostridia showed repeating DGGE patterns (figure 5). In many cases several single bands in one lane showed the same sequence and could be affiliated to one Clostridium species, though it was possible to allocate several bands to different clostridia species. Most of the bands could be affiliated to the species C. butyricum, C. tyrobutyricum and C. ljungdahlii (table 5).
Carnobacterium sp. AT12 was the only Eubacterium species found in the seed sludge after the thermal pre-treatment and before start-up (table 6). It is a non-spore forming, facultative anaerobic and heterofermentative lactic acid bacterium (de Vos, 2009), which previously belonged to the genus of Lactobacillus (Dworkin et al., 2006). Even though this bacterium is no spore forming, it was able to survive the thermal pre-treatment. After the addition of substrate it was replaced by other species of LAB. The members of this genus are facultative anaerobic, mesophilic, non-spore forming, obligately saccharoclastic with complex nutritional requirements. Their optimum pH for growth is between 5.5 and 6.2, but they can also grow for pH lower than 5. L. fermentum is obligately heterofermantative. It degrades hexose to equimolar quantities of lactic acid, CO2 and acetate or ethanol via the 6phosphogluconate pathway.  Table 5 L. delbrueckii subsp. bulgaricus on the other hand is obligately homofermentative and degrades hexose via the Embden-Meyerhoff pathway to lactic acid (de Vos, 2009). Through the phases DI to DII species of Lactobacillus established a stable community in the reactor. At phases DI and DII Olsenella species were also detected, while at phase DIII Sporolactobacillus has been identified (table 6). Species of the genus Sporolactibacillus are facultative anaerobic and obligately homofermentative. They can form spores, which are resistant to heating at 80 °C for 10 min (de Vos, 2009). They can grow only on fermentable carbohydrates at pH greater than 4.5. Their optimum growth temperature is 30 °C (Dworkin et al., 2006). For Olsenella sp. oral taxon 809 there are no metabolic data available, since it has not been cultivated. In general the species genus Olsenella is non-spore forming, homofermentative and strictly anaerobic. The end-products of sucrose fermentation consist mainly of lactic acid and to a lesser extent of acetic acid (Olsen et al., 1991). The presence of these bacteria can explain the increasing production rate of lactate and acetate indicating an overall shift of the bacterial metabolism in the system to lactic acid fermentation and more specific to homolactic fermentation as derived by the decreasing biogas production. They compete with the hydrogen producing bacteria for substrate.
In the seed sludge three species of Clostridium could be identified. Only C. butyricum could be maintained in the system until the end of phase DIII (table 6). The rest of the initially present Clostridium species were replaced by others indicating a constant changing population, like in the case of (Huang et al., 2010). All of them are able to produce copious amounts of hydrogen gas by mixed acid fermentation with acetate and butyrate as their major end-products as already mentioned. Even though, hydrogen producing bacteria where identified during phases DI to DII, the deterioration of hydrogen production could not be hindered. The lactic acid bacteria could proliferate in the system. Also in the case of phases DVI, D1 and D2 (tables 6 and 7) the symbiosis of lactic acid together with hydrogen producing bacteria could be observed resulting to high lactic acid production rates and medium to low hydrogen yields. In all cases, the employed HRT was equal or higher to 2 d. It seems that long HRT contribute to the dominance of lactic acid over hydrogen producing bacteria. In the case of phases D3 to D7, high production rates of lactic acid could still be observed, but they were lower than that of acetic and butyric acid, while medium to higher hydrogen yields could be achieved. The HRT applied was equal to or lower than 1 d. Under these conditions the hydrogen producing metabolism could dominate over the lactic acid metabolism. After phase D3 and on C. tyrobutyricum became one of the dominant Clostridium species. After phase D5 and on Sporolactobacillus could not be found any more in the system. It seems that HRT lower than 0.5 d does not support its growth. In the work of Fang (2002) who investigated hydrogen production by untreated secondary sludge at pH 5.5, loading rate of 25 g sucrose / (L • d), HRT 0.25 d and temperature of 26 °C, 69.1% of the microbial clones were affiliated to Clostridium species and 13.5% to Sporolactobacillus racemicus, which was able to be retained in the system due to the granular sludge reactor configuration.
Analogously, Olsenella disappeared from the system at phase D7 for HRT of 0.25 d. Lo et al (2008) could detect Olsenella and C. butyricum in a hydrogen producing reactor with xylose as substrate at HRT of 2 hours. They also used granular sludge. In another work (Castelló et al., 2009), the acquired low hydrogen yield was justified by the presence of Olsenella along with Prevotella, Bulleidia, Mitsoukella and Selonomonas species, which consumed the substrate. In this work, C. tyrobutyricum became visible also through the eubacteria specific primers indicating a general proliferation of Clostridium species for vey short HRT that comes in agreement with the low doubling times of 30 min to 3 h that have been observed for most Clostridium species (Dworkin et al., 2006). Despite the dominance of hydrogen producing biomass over lactic acid bacteria, the hydrogen yield was relatively low due to the low sucrose degradation efficiency. The biomass concentration in the system was not sufficient to metabolize the whole amount of substrate. L. mucosae, an obligate heterolactic bacterium, could maintain itself in the system even for HRT as low as 0.25 d. These findings are in agreement with the observations of (Kim et al., 2012), who investigated hydrogen production by a lactate-type fermentation in a CSTR with working volume of 4 L, glucose as substrate and digester sludge as inoculum pre-treated with acid. For HRT of 1 d and OLR equal to 10 g / (L•d) they affiliated 65% and 35% of the total bacterial population to Eubacterium and Clostridium species respectively. When the OLR was increased to 40 g / (L•d) the proportion changed in favor to Eubacterium species to 72% and 28% respectively. For the same OLR but HRT equal to 0.5 d a balance between the two populations with 50% each was established.
C. ljungdahlii could be detected in the system during the phases DIV, DVI, DVII and D1 to D7. More than 90% of its strains can produce H2 during heterotrophic growth on glucose or fructose, while sucrose can not be utilized (de Vos, 2009). It can also grow autotrophically on H2 and CO2 or CO. It forms spores only rarely. The regulation of the diverse pathways of homoacetogens is still not understood (Dworkin et al., 2006). For the first case, extracellular enzymes released to the bulk liquid by other microorganisms that are able to degrade sucrose to glucose and fructose are required. In a mixed acid fermentation as described by Eq. 4, a ratio of HBu:HAc should have been obtained. In most of the cases this ratio is lower than one indicating that acetic acid is also produced through a pathway other than that of hydrogen production. It seems that C. ljungdahlii is consuming hydrogen. Clostridium tyrobutyricum strain SCTB125 61 JN650290

Effect of Fe 2+ addition
The addition of Fe 2+ at phase DVII had an influence on the bacterial population in comparison to the seed sludge (phase D6). L. casei, Sporolactobacillus from the LAB and C. butyricum became also dominant, while Olsenella was not any more detected. The presence of an extra lactic acid bacterium species though, contradicts the observed reduction of lactic acid production rate. The most probable explanation is that given in 5.3. The clostridial metabolism was shifted from lactic acid production, which is triggered under limitation of Fe 2+ (Dürre, 2005), to butyrate. Nevertheless, LAB could still influence the system as the production of lactic acid in phase DVII indicates, also for HRT as short as 0.5 d, not like in the case of Kim et al. (2006), who discovered only one lactic acid bacterium (Bacillus racemilacticus) and only during the phase with OLR = 60 g COD / (L • d), while for higher or lower OLR no such species was found, although lactate has been produced in all cases, indicating that lactate in the system was a product of Clostridium species metabolism, which dominated the system.
In an investigation (Karadag & Puhakka, 2010b) about the influence of Fe 2+ concentration in the range of 0.5 mg/L to 100 mg/L on a CSTR system fed with glucose and operated at HRT of 0.208 d, OLR of 43.2 g glucose /(L•d) and pH 5 an increase of 71% in hydrogen yield for Fe 2+ concentration of 50 mg/L was achieved, followed by a fermentation shift from ethanol type to butyric acid type like in this work. In two other works (Wang & Wan, 2008) and (Lee et al., 2001) investigating the influence of Fe 2+ concentration in batch experiments with vials, optimum concentrations of 350 mg/L and 352.8 mg/L were detected respectively. It seems that there is potential for the optimization of the quantity added to the system.

Symbioses in fermentative hydrogen production systems
There are only a few works that have investigated the bacterial population and the influence of HRT and OLR on hydrogen producing systems by molasses. Ren et al. (2007) studied the influence of pH on the microbial population structure of bio-hydrogen production by molasses in a 2.5 L CSTR seeded with sewage solids and operated at HRT of 0.25 d, OLR between 7 g COD / (L•d) and 30 g COD / (L•d) and at temperature of 35 °C. At pH between 5.5 and 6 mixed ethanol-butyrate fermentation was observed. In this work no ethanol could be detected for HRT of 0.25 d. In the reactor a co-existence of clostridia and LAB was observed like in this work. The bacterial population was dominated by C. pasteurianum, Lactococcus sp., Desulfovibrio ferrireducens together with uncultured species of Actinobacterium and Bacteroidetes. Chu et al. (2011) on the other hand, did not observe such a co-existence. They affiliated most clones found in a suspended sludge system treating fermented molasses with HRT raging from 0.33 d to 0.083 d to C. butyricum, Megasphaera sp. and Corynebacterium glutamicum. In the case of defined substrate, Kim et al. (2006) obtained the optimum hydrogen yield when a LAB (Bacillus racemilacticus) together with Clostridium species were dominant in the system. The dominance of the lactic acid bacteria Sporolactobacillus, Olsenella, along with hydrogen producing C. tyrobutyricum, C. butyricum and Clostridium sp strain S6 was observed in a Continuously Stirred Tank Reactor (CSTR) at pH 5.5 and HRT of 1 d yielding 1.24 mol H2/mol hexose (Wongtanet et al., 2007). The bacterial population examination confirmed in this case a symbiosis between the lactic acid and hydrogen producing bacteria, too. In another work (Ohnishi et al., 2010), in which kitchen waste containing lactate was used as substrate, lactic acid removal was also observed along with carbohydrate degradation during hydrogen production in an Anaerobic Sequencing Batch Reactor. The phylogenetic analysis affiliated the dominant bacteria to the genera of Lactobacillus, Selonomonas, Veillonella and Megasphaera, which belong to the phylum of Firmicutes, to Prevotella genus of the phylum of Bacteroidetes and to Atopobium and Bifidobacterium of the Actinobacteria phylum. Hydrogen production has been attributed, due to the absence of Clostridium species, to Megasphaera by simultaneous utilization of the carbohydrates and the lactate contained in the initial feed and produced by the LAB Lactobacillus present in the system and which in return as aerotolerant bacterium consumed residual oxygen for establishing a suitable milieu for the anaerobe Megasphaera and supplying it with lactate for hydrogen production.  Table 6. Allocation of the microbial species to the seed sludge and the operation phases DI to DVIIIe. Microbial population has been simplified to species level. Symbols: "o" no hydrogen production or consumption. "+/-" some strains can produce hydrogen. "+" hydrogen production and "-" hydrogen consumption In an other investigation (Hung et al., 2011b), it was also suggested that the facultative anaerobes of Streptococcus sp. and Klebsiella sp. found in the granular sludge of hydrogen producing fermentors seeded with untreated sewage sludge, maintained the strict anaerobic conditions required by the Clostridium species for hydrogen production. In the work of Kim et al. (2009), a CSTR seeded with thermally pre-treated sludge at 95 °C for 15 min was operaterd at OLR of 50 g COD / (L•d), HRT of 0.5 d, pH regulated at 5.3 and temperature at 35 °C. In this case also, a symbiosis of hydrogen producing clostridia with the lactic acid bacteria L. delbruecki and Lactococcus lactis was observed. Such an enhancing symbiosis can not be excluded that also takes place in the system of this work, so that the residual dissolved oxygen of the water added into the system could be removed. More over, the possibility of a secondary fermentation by C. tyrobutyricum present in most of the experimental phases and facilitated by the production of lactate by the LAB as described in our previous work (Mariakakis et al., 2011) and later demonstrated (Wu et al., 2012) has to be considered, too.
In any case, this symbiosis reduces the available substrate for hydrogen production. Furhtermore, in this work, it was not beneficial to hydrogen production in the phases with long HRTs, but deteriorated or even completely inhibited it. Lactobacilli possess the potential of inhibiting other microorganisms by different mechanisms. Their fermentation products consist mainly of lactic and acetic acid, which reduce the pH which on its side reduce their dissociation degree. In the presence of oxygen many species such as L. lactis and L. bulgaricus can produce H2O2, which is bacteriocidal for gram negative and bacteriostatic for gram positive bacteria (de Vos, 2009). They can produce bacteriocins, proteinaceous substances with bactericidal effect on microorganisms closely related to the producer. Most probable each Lactobacillus species has strains that can produce bacteriocins. Among the organisms that have been found in the present work L. casei and L. fermentum can produce bacteriocins that are inhibiting other lactobacilli (Dworkin et al., 2006). Nevertheless, it was demonstrated that bacteriocins can hinder non-closely related microorganisms, too. For instance, an inhibition effect of L. lactis on C. tyrobutyricum due to the excretion of the bacteriocin nisin Z was detected (Rilla, 2003) and of L. paracasei on C. acetobutylicum and C. butyricum (Noike, 2002). It can not be excluded that the lactobacilli present on this work exert a similar effect on the clostridia species. There are also some bacteriocin-like substances produced by lactobacilli, such as bulgarican produced by L. delbrueckii subsp. bulgaricus that can inhibit a wide range of non related pathogenic gram negative bacteria. L. delbrueckii, L. fermentum and L. casei are also capable of producing bacteriophages that can cause cell lysis (Dworkin et al., 2006). L. casei, L. fermentum and L. bulgaricus have been also identified in this work, so an edverse effect can not be excluded as in several other works has been suggested. It has been proposed that L. ferintoshensis and L. paracasei present along with Clostridium sp. and Coprothermobacter sp. in untreated digester sludge negatively influenced hydrogen production (Kawagoshi et al., 2005). The same was suggested for Sporolactobacillus sp. that was present in an anaerobic sequencing batch reactor seeded with heat pre-treated sludge and operated at pH 5.0 using sweet sorghum syrup as substrate at OLR of 25 g sugar/(L•d) (Saraphirom & Reungsang, 2011). They justified it by their ability to produce bacteriocins as Noike et al. (2002) suggested. None of these substances has been monitored, so a possible accumulation in concentrations that are inhibiting to the clostridia can only be determined by a general parameter as the COD.  Table 7. Allocation of the microbial species to the seed sludge and the operation phases D1 to D7. Microbial population has been simplified to species level. Symbols: "o" no hydrogen production or consumption. "+/-" some strains can produce hydrogen. "+" hydrogen production and "-" hydrogen consumption In most cases, heat treatment of the seed sludge was applied, but it was not sufficient to inhibit the growth of lactic acid bacteria, although they can not form spores and it has been demonstrated in vials that heat treatment between 50 °C and 90 °C is sufficient (Noike et al., 2002). For instance, L. delbrueckii and L. fermentum were able to be identified at batch experiments producing hydrogen seeded with activated sludge and digester sludge thermally pre-treated at 65 °C for 30 min Hafez et al., 2010). It seems that with increasing inoculum quantity, thermal treatment becomes less effective and it can not hinder the co-dominance and activity of lactic acid bacteria.

Conclusion
Hydrogen production by molasses could be successfully carried out in large lab-scale reactors for a period longer than 180 d and under variable combinations of OLR and HRT. The maximum H2 yield obtained was 1.53 mol H2/mol hexose for HRt of 1 d and OLR of 36.1 g sucrose/(L•d). Improvement of the hydrogen production yield of 28%was achieved by the addition of Fe 2+ to an end concentration of 1000 mg/L. In figure 7 the acquired hydrogen yields of all phases as a function of the operation parameters HRT and OLR, along with a suitable range of combination of these parameters, as determined in the current work, are presented. Combinations resulting to COD concentrations higher than 50 g/L (phase D3), was showed to be inhibitory to H2 production. Reason was not the undissociated form of acids, but most probably the production and accumulation of bacteriocidal or bacteriostatic substances, excreted by the LAB. The second line indicates the combination for which the process becomes unfavorable in terms of substrate utilization efficiency and hence can not be considered as cost effective. The applied seed sludge pre-treatment and reactor start-up methods were successful in enriching the biomass in hydrogen producing microorganisms and killing methanogenic microorganisms that are detrimental to H2 production. Nevertheless, H2 production has been carried out parallel to lactic acid metabolism, which was driven either by the presence of LAB, or by the hydrogen producing clostridia due to iron Fe 2+ limitation. A co-existence of clostridium species with lactic acid bacteria seems to be unavoidable, even for extensive pre-treatment of the seed sludge. Lactic acid bacteria influence the system primarily by consuming the substrate available for hydrogen production. The co-existence of clostridia and LAB though, seems to become beneficial to hydrogen production at HRTs in the range of 0.5 d to 1 d by supplying certain clostridia genera that are capable of performing secondary fermentation with substrate and/or by removing the residual dissolved oxygen from the system and hence establishing an appropriate milieu for the growth of clostridia. For the successful technical implementation of hydrogen production, for which process control is complex and oxygen in trace concentrations is to be expected, this symbiosis may be regarded as pre-requisite. The exact extent, to which the LAB contribution is beneficial and not adverse, requires the quantification of the biomass for the allocation of the metabolic products to specific microbial species and the monitoring of the concentrations of possible inhibiting substances. Figure 7. Suitable range of OLR-HRT combination for fermentative hydrogen production by molasses at pH 5.5