Effect of mixing

Given the experimental conditions (see “Methods” section) and the poor solubility of H₂ in the aqueous phase, the optimal mixing conditions yielding the most efficient delivery of H₂ from the gas phase had to be determined. The reaction vessels were incubated in an orbital shaker at various mixing speeds (rpm). Figure 1 indicates that there is an optimum value for this parameter; in our arrangement, it was 150–160 rpm. It is noteworthy that at higher mixing rates CH₄ production decreased sharply in contrast to earlier observations at thermophilic temperature [19]. In all subsequent experiments, the shaker was set at 160 rpm. It is evident that this mixing rate is valid under our conditions and henceforth was applied consistently in order to limit the number varying parameters. In other systems, the optimal mixing conditions should be determined individually. The main conclusion from these experiments was that the mixing that yields optimal H₂ utilization may not be the maximum achievable mixing rate.

Optimization of H₂ dosage

Next the optimal daily H₂ dosage was established. Various volumes of H₂ were therefore injected into the batch reactors, which were treated identically in all other known aspects. The batch fermentations were started by adding 0.3 g of α-cellulose as substrate for AD according to the VDI (Verein Deutscher Ingenieure, protocol [25]. H₂ gas was injected every day and the consumption of H₂ was followed by gas chromatography. Cumulative CH₄ evolution curves are plotted in Fig. 2. CH₄ production proceeded steadily for 7–8 days in the control reactors, which received no daily H₂ dosage, but from day 12 practically no gas evolved. In total, 6.2 ± 0.54 mmol of CH₄ was generated from the residual biogas potential of the sludge and added α-cellulose substrate. 1.62 mmol of this quantity originated from the sludge and 4.58 mmol from the α-cellulose substrate. The biochemical CH₄ potential of α-cellulose is 4.71 [26] and therefore all of the added substrate was consumed by the community and was converted to CH₄. Addition of a daily 0.81 ± 0.16 mmol of H₂ gas into the headspace of the batch reactors dramatically increased the CH₄ production (Fig. 2). The GC measurements revealed that all of the injected H₂ was completely consumed by the microbes within 24 h. In separate experiments, it was established more precisely that under these conditions all the H₂ had vanished from the headspace after 16 h and CH₄ evolution started at hour 2 following H₂ injection (data not shown). A new dosage of H₂ was dispensed consistently every 24 h. Increasing the total H₂ load to 43.00 ± 1.43 mmol resulted in a somewhat faster initial CH₄ production, but the cumulative-specific CH₄ production was lower than in the case of adding 24.42 ± 0.81 mmol of H₂ in the same period of time. In line with this observation, H₂ started to accumulate in the headspace on day 14 and from day 17–18 CH₄ production ceased. On further increase of the overall H₂ injection volume to 55.69 ± 1.85 mmol, i.e., 1.86 ± 0.38 mmol H₂ day⁻¹, even less cumulative-specific CH₄ was yielded. In these reactors, H₂ build-up in the headspace started sooner, i.e., on day 10 and CH₄ evolution stopped completely on day 13. Overall, these results indicated that the system utilized the α-cellulose substrate within 7–8 days and the microbial community sustained its H₂ conversion activity for an extended period of time if the daily H₂ injection did not exceed 0.81 ± 0.16 mmol of H₂ (Table 1). The concentrations of organic acids were determined every week. Acetate levels increased significantly by the end of the experimental period. 3.43 mM acetate accumulated by the end of the experiment in the reactors receiving 55.69 ± 1.85 mmol of H₂, which exceeded the recommended threshold, but apparently this alone did not explain why CH₄ evolution stopped in the reactors loaded with higher daily H₂ injections (Fig. 3). The pH had increased considerably by the end of the 4-week experiments (Fig. 4), indicating a severe loss of the bicarbonate buffering capacity of the inoculum sludge. It is noteworthy that the pH also shifted by 1.1 units in the control reactors which were not fed with H₂. In order to employ the same protocol, these vessels were also degassed and filled with N₂ gas every day. It is therefore likely that the daily replacement of the headspace prompted a gradual desorption and loss of dissolved CO₂ and caused a shift in the bicarbonate buffering system [27, 28]. The pH increased even further, i.e., beyond pH = 9, which is a critical upper limit for the methanogenesis [29]. A similar exhaustion of the buffering capacity upon H₂ addition was noted in previous reports [19, 20]. The system could apparently tolerate high pH fairly well when 0.81 ± 0.16 mmol of H₂ was the daily dosage, but started to inhibit CH₄ biosynthesis on day 13 and 10 upon addition of daily 1.43 ± 0.28 or 2.86 ± 0.38 mmol of H₂, respectively. In this experimental set-up, it was not possible to determine the time points when the inhibitory pH range was attained. The results indicated that the likely reason for the obstruction of CH₄ formation was the limiting buffering capacity of the system due to the low bicarbonate concentration. The optimal amount of daily H₂ dosage in this system was within the range of 0.8–1.5 mmol of H₂; further experiments should determine the exact value.

Total CH4 production (mmol)	CH4 from α-cellulose (mmol)	Theoretical from α-cellulose (mmol)	Total injected H2 (mmol)	Theoretical CH4 from H2 (mmol)	Measured CH4 from H2 (mmol)	Difference
6.20 ± 0.54	4.58 ± 0.09	4.71	0.00 ± 0	0.00	0.00 ± 0.66	0.00
12.35 ± 0.44	4.63 ± 0.09	4.71	24.42 ± 0.41	6.10	6.10 ± 0.24	0.00
11.30 ± 0.50	4.61 ± 0.09	4.71	43.00 ± 1.02	10.75	5.08 ± 0.48	−5.67
10.33 ± 0.81	4.61 ± 0.09	4.71	55.69 ± 2.76	13.92	4.11 ± 0.75	−9.82

Effect of CO₂ addition

In the next series of batch fermentations, the inoculum originated from the same mesophilic industrial biogas plant, but at different points of time, and therefore small fluctuations of organic total solid content and microbial community composition should be taken into account when the results are subjected to direct comparison. The initial addition of α-cellulose was omitted in order to avoid any disturbing effect of the CH₄ generation from the substrate. The duration of these fermentations was extended to 80 days to test for sustainable CH₄ production. The reactors were supplied with the optimal daily dosage of 0.81 mmol of H₂ in order to check if the CO₂/bicarbonate buffering capacity was indeed the major limiting factor in the previous experiments [28, 30]. The daily CH₄ volumes measured in the headspace are plotted in Fig. 5. CH₄ evolution progressed steadily until day 28, but dropped sharply afterwards. A warning sign of system failure was noticed already on day 27, when measurable residual H₂ was detected in the headspace (Fig. 5; Table 2). As shock therapy, massive CO₂ injection (25 mL) was dispensed into the reactors following the daily dosage of H₂ on day 31 (Fig. 6). All of this CO₂ disappeared from the gas phase within 24 h, indicating that the system was indeed severely depleted of CO₂/bicarbonate. The same CO₂ treatment was repeated next day, which apparently restored the functional state of the system signaled by the build-up of residual CO₂ in the headspace (Fig. 6). The daily CO₂ dose was then gradually decreased to the stoichiometric volume, i.e., approximately 0.25 mol of CO₂/mol of H₂ per day. The system responded positively, as exhibited by the restoration of CH₄ production on day 32 accompanied by a gradual decrease of residual CO₂ levels in the gas phase. Daily CO₂ injection was stopped on day 41. H₂ accumulation commenced again almost immediately and was accompanied by the loss of CH₄-evolving ability from day 43, and therefore CO₂ injection (25 mL) was resumed on day 47. Detectable remaining CO₂ was noticed already on the next day and from this time on a daily dosage of 0.25 mol of CO₂/mol of H₂ of CO₂ was maintained until the end of the experiment. CH₄ production returned to the previous level, all of the injected daily H₂ and CO₂ were consumed within 24 h and this continued for an additional month. It is noteworthy that, except for pH bursts on days 31 and 45, the pH in both the control and H₂-fed reactors remained within the acceptable limit of pH ≤8.5 throughout the investigated period (data not shown).

	Bassani et al. (2015)a		Present work
			No external CO2		External CO2 added
	Control	H2 added	Control	H2 added	Control	H2 added
Biogas composition (%)
CH4	69.7 ± 0.3	88.9 ± 2.4	17.71 ± 1.15	79.77 ± 2.31	0.00	95.53 ± 1.79
CO2	30.3 ± 0.3	8.8 ± 3.2	73.63 ± 3.61	17.71 ± 0.90	0.00	4.47 ± 1.34
H2	0	2.3 ± 1.8	0.0	2.51 ± 0.82	0.0	0.00
Gas production (mL L−1 h−1)
CH4	2.75 ± 0.58	4.17 ± 0.50	1.51 ± 0.07	6.78 ± 0.20	0.00 ± 0.00	6.21 ± 0.12
CH4 from H2	0.0	1.41	0.00	4.27	0.00	6.21
CO2	1.21 ± 0.25	0.42 ± 0.13	4.20 ± 0.21	1.51 ± 0.08	0.00 ± 0.00	0.29 ± 0.09
H2 injection rate (mL L−1 h−1)	0.00	8.00 ± 1.17	0.00b	22.66 ± 0.20b	0.00b	20.96 ± 0.23b
H2 consumption (mL L−1 h−1)	0.0	7.42 ± 1.08	0.00 ± 0.00	22.44 ± 0.19	0.00 ± 0.00	20.96 ± 0.02
H2 consumption (%)	0.0	92.7	0.0	99.06	0.0	100.00
pH	7.74 ± 0.16	8.17 ± 0.04	8.66 ± 0.19c	9.38 ± 0.11c	8.29 ± 0.04c	7.89 ± 0.20c
Organic acids (mM)
Acetate	nd	nd	nd	nd	0.33	1.48
Butyrate	nd	nd	nd	nd	0.00	0.04
Isovalerate			nd	nd	0.02	0.10

^a Mesophilic data

^b Estimated from daily dose

^c At the end of the experiment; nd not determined

Several deductions could be drawn from this series of tests. First, the system becomes depleted of CO₂ if semi-continuous H₂ feeding and daily degassing are administered to the fed-batch system. This phenomenon was manifested after about 1 month in our arrangement, where daily degassing and replacement of the headspace were included to retain the same protocol in the control and experimental reactors. Clearly daily degassing is not necessary in industrial setting. Second, the residual H₂ accumulation in the gas phase is a good early warning sign of upcoming system failure due to CO₂ exhaustion. Third, the microbial community participating in the CH₄ generation process recuperates quickly and completely even after repeated system failure if the process control is alerted in time. Fourth, the microbial community supplied only with H₂ and CO₂ upholds the pH within the normal operating range. Finally, stoichiometric administration of H₂ and CO₂ yields a practically complete conversion to pure CH₄ within 24 h under mesophilic conditions.

Effect of additional substrate addition

Next, it was tested whether the addition of α-cellulose affected the CH₄ productivity from H₂. Two series of experiments were designed and the duration of the experimental period was shortened in order to avoid any complication due to CO₂ depletion and concomitant pH elevation. In the first set of batch fermentations (Fig. 7), various amounts of α-cellulose were added only at the start of the experiments, and in the second series (Fig. 8) the addition of the same amount of α-cellulose was repeated every week. Daily replacement of the headspace with N₂ and the injection of 0.81 mmol of H₂ was maintained in all reactors.

There was no significant difference between the CH₄ productions from H₂ in the reactors receiving the substrate quantity recommended by the VDI [25] protocol as compared with those without substrate, i.e., the difference between the green and red curves in Fig. 7 correspond solely to the CH₄ produced from α-cellulose. This suggests that the addition of substrate at the beginning of the fermentation does not assist CH₄ evolution from H₂. Moreover, an inhibition of CH₄ productivity from H₂ was noted when the substrate load was doubled, i.e., upon the addition of 0.6 g substrate, 3.47 ± 0.08 mmol of CH₄ was formed from α-cellulose instead of the theoretical potential of 9.42 mmol of CH₄. It should be noted that the H₂ consumption rate remained unaffected by the substrate loading, i.e., the injected H₂ disappeared from the headspace within 24 h. The conversion efficiency of CH₄ formation from H₂ was estimated from the daily CH₄ levels in the headspace. The day-to-day values fluctuated considerably during the experimental period and achieved an average of 72 ± 25 %. The remainder of the H₂ may have been metabolized in alternative pathways, which are the subject of future studies.

In the next set of experiments, the reactors were fed with the same amount (0.3 g) of α-cellulose every week and the daily H₂ injection (0.81 mol H₂) was maintained. The aim was to test whether the microbial community remained intact for an extended period of time after the expiration of its residence time in the industrial AD facility and to see whether the metabolically active community facilitated the bioconversion of H₂ to CH₄. The cumulative CH₄ production increased almost linearly and the amount formed suggested an unchanged reaction rate for both α-cellulose and H₂ when the VDI protocol [25] was followed (Fig. 8). It is noteworthy that increasing the weekly α-cellulose load prompted a strong inhibitory effect. The collapse of the CH₄-forming activity was not associated with changes in pH. Without α-cellulose, the daily dosage of H₂ caused an increase of the pH into the dangerous zone, as observed earlier (Fig. 4), due to the depletion of the buffering capacity. Weekly supply of the substrate balanced the pH; the degradation of the α-cellulose apparently yielded enough CO₂ to maintain stable operation. Too much substrate, e.g., 0.6 g α-cellulose/week, shifted the pH to lower values, although it did not fall below 6.5, which is usually considered detrimental [29]. The accumulation of acetate increased dramatically upon substrate overloading (data not shown). This might have been the likely reason for the process inhibition. It is important to note that the H₂ conversion yields in this series of experiments were close to 100 %, which emphasizes the importance of the inoculum quality.

Comparison with previous works

Our approach to improve the P2B principle attempts to counteract the low solubility of H₂ in the aqueous environment by increasing the contact time of the gas and aqueous phases in a fed-batch fermentation arrangement. This has not been tested earlier.

There are four previous reports available to measure up against this approach. Lee et al. [24] used a fixed-bed reactor, while Reuter [32] developed several versions of a continuous stirred tank reactor (CSTR) design and scaled up the process to an industrial level. Both studies concluded that hydrogenotrophic methanogens in pure or mixed culture were markedly efficient catalysts and converted H₂ and CO₂ to CH₄ in surprisingly high yields and rates. Unfortunately, the published results from those studies contain limited data on process parameters to compare with the fed-batch system examined in the present study.

Two recent papers from the Angelidaki team [19, 22] also used CSTR reactors and reported promising results. Their thoughtfully designed and thoroughly documented reports provided data allowing the assessment with our study. Table 2 summarizes the results.

Besides the use of distinct reactor arrangements and sizes, i.e., fed-batch versus CSTR, several operational parameters differed in those studies from our set-up, e.g., inoculum composition and quality, substrate used for CH₄ generation, stirring mode and speed. Therefore, only the major tendencies and not the exact values are suitable for a rigorous comparison.

It was found that at high shaking speed the H₂ conversion process may not be limited by the gas–liquid mass transfer [19] at thermophilic temperature. In our experience, this observation could not be repeated under mesophilic conditions, and above 160 rpm CH₄ formation was inhibited (Fig. 1). It was concluded that the process in our system was critically limited by the mass transfer of H₂ at the gas–liquid interface. Hydrogenotrophic methanogens utilized the dissolved H₂ at a high rate, and therefore a concentration gradient developed between the liquid and gas phases, driving H₂ into the liquid compartment from the headspace as time advanced. It is likely that the fed-batch operation optimized the condition where the amount of H₂ transferred into the liquid phase was close to the amount consumed by the microbes. The data presented in Table 2 clearly indicate that this was indeed the case.

In the CSTR work, H₂ was dosed on the basis of the available CO₂ from the coupled biogas production [22]. Although significant upgrading of the biogas was achieved, this stipulation limited the rate and amount of H₂ injection into the system. The goal in these investigations was to achieve maximal H₂ conversion yield. H₂ bubbles are difficult to retain in the aqueous system, and diffusers and very low purging rates therefore had to be applied to facilitate the dissolution of H₂ and its conversion to CH₄ during the short residence time of the gaseous substrate in the reactor. In the fed-batch configuration, the H₂ loading rate could be increased to 4 times that of the CSTR operational mode without the loss of H₂ (Table 2).

In the present study, mesophilic conditions were maintained. Bassani et al. [22] carried out their experiments at 35 and 55 °C under otherwise identical conditions. A significant improvement in CH₄ formation rate was noted at higher temperature. A similar effect can be expected in the fed-batch system; this will have to be established in future studies. A comparison between our mesophilic data with those obtained at thermophilic temperature indicates a 2.0 [19] and 2.7 [22] times higher CH₄ production rate from H₂ in the mesophilic fed-batch reactors as compared with the thermophilic CSTR, respectively.

The mesophilic process performance parameters of Bassani et al. [22] can be compared directly with our results reported under the “Effect of CO₂ addition” subtitle above. Two sections of stable operation in our experimental period were taken into account, i.e., the initial phase without external CO₂ addition between days 2 and 28 and the part when stoichiometric CO₂ and H₂ were injected daily (days 50–80) (Figs. 5, 6). To make a fair assessment, the residual CH₄ production in the control reactors (no H₂ added) should be taken into account.

The control samples in our work started at an unusually low CH₄/CO₂ ratio (Table 2), which could be due to the residual biogas potential of the inoculum and the fact that all H₂ was removed during initial degassing of the reactors. Therefore, the activity of the hydrogenotrophic methanogens was severely restricted until some H₂ became available from the fermentation of the residual, small amount of biomass. The situation changed dramatically in the reactors receiving H₂ injections and the system produced bio CH₄ of high purity, i.e., containing only 17.71 % CO₂.

There was a 6.5-fold increase in CH₄ yield from H₂ in the fed-batch system relative to the mesophilic CSTR experiments if a stoichiometric amount of CO₂ was added to both systems together with the H₂ (Table 2). Moreover, the fed-batch system operated at a 4-times higher H₂ load than the CSTR reactor. The H₂ consumption was above 90–100 % in both systems, indicating that the reaction was carried out very efficiently in both systems. The CSTR operation mode has its benefits and advantages, but apparently does not help overcome the low H₂ solubility problem, which seems to be the major bottleneck in the accomplishment of the P2B principle at mesophilic temperature.

As an added value, it should be noted that in the fed-batch system a considerable accumulation of acetate takes place without any observable sign of acidosis-related process failure (Fig. 9). The accumulation of acetate was probably due to the inhibition of acetoclastic methanogenesis and syntrophic acetate oxidation [33] by the high H₂ doses. Acetate is a valuable commodity [30, 34] and, if acetate can be recovered by a suitable technology from the reaction mixture, it would be a useful side-product of the fed-batch fermentation-based P2B technology.

The batch fermentation system

Organic acid analysis

Samples for organic acid analysis were taken from the liquid phase of the reactors. The samples were centrifuged (13,000 rpm for 10 min,) and the supernatant was filtered through PES centrifugal filter (PES 516-0228, VWR) at 14,000 rpm for 20 min. The concentrations of volatile organic acids were measured with HPLC (Hitachi LaChrome Elite) equipped with refractive index detector L2490. The separation was performed on an ICSep ICE-COREGEL—64H column. The temperature of the column and detector was 50 and 41 °C, respectively. The eluent was 0.01 M H₂SO₄ (0.8 mL min⁻¹).

Gas composition analysis

The gas composition of the reactor headspace was measured every day by GC. The CH₄ and H₂ contents were determined with an Agilent 6890 N GC (Agilent Technologies) equipped with an HP Molesive 5 Å (30 m × 0.53 mm × 25 µm) column and a TCD detector. The temperature of the injector was 150 °C and application was made in split mode 0.2:1. The column temperature was maintained at 60 °C. The carrier gas was Linde HQ argon 5.0, with the flow rate set at 16.8 mL min⁻¹.