A small scale silicone membrane reactor similar to the one described by Revsbech (1989) was used to provide a controllable environment in terms of H₂ supply (Figure 1). This design deviates substantially from the silicone-tubing membrane reactors used by Wang et al. (2013) and Díaz et al. (2015) as it was decided to exploit a geometry that was possible to analyze with microsensors and allowed reliable mathematical analysis. The reactor consisted of two cylindrical chambers (3 and 6 cm length) consisting of Plexiglas (i.d. 4 cm) separated by a silicone rubber membrane. The membrane was made by coating a stainless steel mesh (mesh size 1 mm) with silicone rubber (Silastic RT adhesive 732, Down coring, Midland, MI, USA), resulting in a total thickness of 1.5 mm. The membrane provided a stable base for the substrate while simultaneously being highly permeable to gasses (Robb, 1968). The chamber below the membrane was used as a reservoir of gasses diffusing up into the upper compartment. The bottom was closed by a rubber stopper and two holes in the sides (8 mm), closed with rubber stoppers, allowed the continuous flushing with a humidified gas flow of 100 mL/min through hypodermic needles of the chamber with a gas mixture controlled by a gas mixer (Brooks Instruments D.V., Holland). To ensure anaerobic conditions in the upper chamber, the upper opening, except for a small opening for microsensor insertion, was closed with Parafilm while it was continuously flushed with humidified Argon or an Ar:CO₂ mixture (∼50–100 mL/min). Maintaining a higher density of the flushing gas than the surrounding air was essential to avoid turbulent mixing of atmospheric air into the chamber. The anoxic conditions were confirmed by an oxygen microsensor with the tip placed in the upper headspace.

A recently improved H₂ microsensor (Nielsen et al., 2015) was used to measure the distribution of H₂ diffusing from the silicone membrane. The new sensor has no interference from H₂S, which can be present at relatively high concentrations (up to mM range) in both raw and digested cattle manure. The size of the tip (70–100 μm) allowed the measurement of profiles with a high spatial resolution of 100 μm. The relatively long diffusion path from the exterior to the internal anode results in a negligible effect of variations in turbulence and diffusive properties in the analyzed medium. The H₂ sensors applied thus exhibited a signal difference between vigorously stirred and stagnant medium of <3% but a response time of about 30 s for 90%. pH microprofiles were also measured in some of the experiments. The pH microsensors had a pH-sensitive tip length of 150 μm, allowing for a similar vertical resolution as the H₂ measurements. Hydrogen gas and pH were measured simultaneously in some of the experiments (see below) by the use of one sensor of each kind, measuring in the same reactor. The H₂ microsensor was connected to a pico-ammeter (PA2000, Unisense) whereas the pH microelectrode was connected to a custom-built mV-meter (Aarhus University) and to an external Ag/AgCl reference electrode positioned inside the chamber. Microsensors were mounted together in a computer-controlled micromanipulator (MC-232, Unisense) while the signals were recorded through an A/D converter (ADC-216, Unisense) using the Sensor TracePro software (Unisense).

Profiles of H₂ were modeled, using the numerical procedure described by Berg et al. (1998). The numerical fitting of the measured H₂ generate profiles with the spatial distribution of the H₂ consumption rates, being expressed as nmol cm^-3 s^-1. The obtained data were also depth-integrated to get the total rates being expressed as μmol cm^-2 h^-1. The diffusion coefficient in water was taken from the diffusion coefficient tables compiled by Unisense². The solubility of H₂ in water was calculated (Crozier and Yamamoto, 1974) to be 778 and 728 μmol L^-1 at 22 and 38°C, respectively. The effective diffusion coefficient in the sieved fresh manure was assumed to be equal to the one in water. The effective diffusion coefficient in the glass beads biofilm was calculated according to Revsbech (1989) being 0.63 times the value in the water.

The initial activity and development of an active H₂ consuming community was followed in fresh cow manure at room temperature (22°C). Cow manure was collected in a dairy cattle farm near Aarhus, Denmark, brought to the laboratory and sieved applying a mesh-size of 2 mm, to remove large particles that could damage the sensor. The manure was then added to the bioreactor described previously to get a layer of about 15 mm on top of the silicone membrane, ensuring the complete anoxic conditions at the bottom. The bottom chamber of the reactor was flushed with H₂ (100%) while the upper chamber was flushed with Ar. The system was left to equilibrate for 2 h after which H₂ microprofiles were measured every 5 h until a steady state was obtained as judged from several consecutive profiles.

The evolution of the H₂ consumption rates showed the same pattern as bacterial growth and therefore, the rates were fitted to the Gompertz equation modified by Zwietering et al. (1990), adapted to the measured rates:

Where: R is the H₂ consumption rate, Ro is the initial rate, Rm is the maximum rate, μ_m is the maximum specific growth rate (h^-1), λ is the lag time (h) and A is the asymptote, being equal to Ln(Rm/Ro). The model was fitted to the experimental data by minimum least squares fitting, using the Solver command in Excel (Kemmer and Keller, 2010).

The development of an active H₂ consuming sludge matrix and the dependence on rates of CO₂ supply were investigated in sludge from a mesophilic anaerobic biogas reactor (hereafter referred as digestate) that was sieved through a 2 mm screen and mixed with glass beads (40–60 μm diameter, 3 M to provide a stable matrix for profile analysis. The digested manure is a liquid and turbulent mixing due to convection would make the results difficult to interpret. The mixture of digestate and glass beads was added to the bioreactor (the bottom chamber being flushed with 100% H₂) and the glass beads were allowed to settle down for 10 min. The supernatant was then removed with a pipette to form a 4–5 mm active layer above the membrane and the upper chamber closed with Parafilm. The thickness of the layer was reduced in order to reduce the capacity for CO₂ production of the system and then be more dependent of the external gasses manipulations. The bioreactor was then immersed in a temperature-controlled water bath at 38°C. A water saturated gas flow controlled by a gas mixer enabled changes in gas composition in both upper and lower chambers.

Two experiments with different types of gas supply manipulations were carried out:

An initially large H₂ penetration in the cattle manure suggested relatively low activity of the microbial community already present in the manure (Figures 2 and 3). H₂ microprofiles were similar during the first 15–25 h in both experiments conducted (Figure 3) showing H₂ diffusion from the silicone membrane into the first 3–4 mm of the manure. During this period, H₂ consumption rates were low (<0.25 nmol cm^-3 s^-1, equivalent to 0.55 m³ H₂ m^-3 manure d^-1) and extended along the 3–4 mm where H₂ could be measured. The integrated H₂ consumption rate was 0.22–0.30 μmol cm^-2 h^-1 (1.4–1.8 L m^-2 d^-1) (Figure 3) during this initial period. After this lag period, lasting for 14.4 to 27.5 h according to the Gompertz model, the H₂ profiles changed quickly, resulting in a progressive decrease in H₂ penetration and increase in rates of consumption (Figures 2 and 3). The H₂ consumption was finally concentrated in a thin layer of less than 0.5 mm, with maximum rates of 5.7–9.5 nmol cm^-3 s^-1 (corresponding to 12.5–20.9 m³ H₂ m^-3 manure d^-1). Assuming that the H₂ consumption rates were proportional to the populations density of methanogens it can be calculated from the Gompertz equation that the maximum specific growth rates were 0.06–0.07 h^-1. After 60 h, integrated consumption rates reached stable values of 1.11–2.05 μmol cm^-2 h^-1, corresponding to 6.8–12.5 L m^-2 d^-1 (Figures 2 and 3).

The development of an active H₂ consuming community was followed using digestate from a mesophilic anaerobic reactor. Contrary to the experiments with fresh cattle manure, the H₂ profiles showed a high rate of consumption already at the time of the first profile measurement 1 h after the set-up of the system (Figure 4). The H₂ diffusing out of the membrane was quickly consumed in a layer thinner than 0.5 mm with rates varying from 24.5 to 31.9 nmol cm^-3 s^-1 (53.9–70.2 m³ H₂ m^-3 digestate d^-1). The integrated rates were also relatively constant with values of 3.24–3.92 μmol cm^-2 h^-1 (19.8–24.0 L m^-2 d^-1) (Figure 4).

After 20 h of continuous H₂ supply thought the silicone membrane, and CO₂ supply from the upper headspace (Ar:CO₂ 70:30%), the CO₂ supply was stopped, resulting in a rapid decrease in H₂ consumption rates of the microbial community (Figure 4). The penetration of H₂ into the matrix thus increased from a layer thinner than 0.5 mm to more than 2.5 mm. After 6–7 h, the H₂ consumption seemed to be localized in the upper part of the matrix with maximum rates around 1.6 nmol cm^-3 s^-1 (3.5 m³ H₂ m^-3 digestate d^-1). The integrated rates were also stable, showing low consumption values of 0.39–0.47 μmol cm^-2 h^-1 (2.4–2.9 L m^-2 d^-1) (Figure 4). These rates may, however, be overestimates as the consumption determined from the change in profile slope in the upper part of the matrix may be an artifact due to measurements in a meniscus at the transition from liquid to gas (Figures 4 and 5).

Due to the clear effect of CO₂ supply on the activity of the digestate in terms of H₂ consumption, the effect of the CO₂ concentration on the measured rates was also investigated (Figure 5). To get dependence from the external CO₂ supply instead of the internal CO₂ stock, the system was initially supplied with a gas mixture of 71% H₂:29% N₂ to deplete the inorganic carbon in the system. Similar to the previous experiment with CO₂ starvation (Figure 4), the H₂ consumption rates decreased progressively, reaching relative stable and low values around 0.35 μmol cm^-2 h^-1 (2.1 L m^-2 d^-1) after 5–6 h (Figure 5). The CO₂ starvation also resulted in substantial changes in the pH values throughout all depths. The pH thus increased progressively from about 8.5 to 9.1 throughout the digestate (Figure 5).

A progressive increase in CO₂ concentration in the gas supplied to the chamber below the silicone membrane (keeping H₂ at 71%) resulted in a progressive increase in the H₂ consumption rates. After a period of several hours without CO₂, an increase of CO₂ to 2.9% produced a clear increase in the H₂ consumption rates, reaching a stable value of 0.5 μmol cm^-2 h^-1 (3.1 L m^-2 d^-1) after 5 h (Figure 5). The next increase in CO₂ % to values of 5.7% also produced a clear increase in the H₂ consumption rates to a value of 0.75 μmol cm^-2 h^-1 (4.6 L m^-2 d^-1). However, further increases did not produce any clear increase in H₂ consumption rates and a slight decrease could even be observed at the highest concentrations (Figures 5 and 6). The spatial distribution of the H₂ consumption was also modified. The H₂ consumption at the initial conditions (Figure 5) was mainly localized in a narrow layer of ca. 0.5 mm with a high activity of 5.9 nmol cm^-3 s^-1 (13.0 m³ m^-3 d^-1). However, the depletion and subsequent resumed CO₂ supply resulted in lower rates (maximum of 2.63 nmol cm^-3 s^-1 at 21.4% of CO₂) distributed in a broader area of 2–3 mm (Figure 5).

An increase of the CO₂ supply produced a substantial pH decrease in the layers closest to the membrane (Figures 5 and 6). Just 2.9% CO₂ thus reduced pH close to the membrane from a value about 9.2 during CO₂ depletion to 8.2. At 5.7% CO₂ the value further decreased to 7.7 and at 21.4% it reached a pH of 7. The pH in the upper parts of the matrix stayed at a relatively high level during the increase in CO₂ supply from below, due to the flushing of the headspace with pure argon.

In a standard methanogenic reactor, the decomposition of the various organic compounds results in acetogenesis with acetate (CH₃COO^-) and CO₂+H₂ being the final products of fermentation processes. These compounds are then converted to methane by the acetoclastic and hydrogenotrophic methanogenesis, respectively (Figure 7). Hydrogenotrophic methanogeneis is highly energetically favorable and the concentration of H₂ is usually maintained at very low values allowing for the syntrophic acetogenesis (Schink, 1988). However, the use of acetate is less energetically favorable and acetate is thus generally accumulating to higher concentrations. At low H₂ concentrations, the syntrophic acetate oxidation can also occur, converting acetate to CO₂+H₂, which then can be transformed into methane.

The localized supply of H₂ into an anaerobic reactor through a membrane is to a major extent modifying the balances and reactions taking place in the nearby microbial community. The high concentrations of H₂ close to the membrane inhibit syntrophic acetogenesis and syntrophic acetate oxidation, as these processes are no longer energetically favorable (Demirel and Scherer, 2008). Homoacetogenesis producing acetate from H₂ and CO₂ and hydrogenotrophic methanogenesis should then be dominant processes and consume all the H₂ supplied through the membrane. From Figures 2–5 it appears that complete depletion may occur over a distance as small as 0.5 mm. Among these two processes, the hydrogenotrophic methanogenesis is more energetically favorable and it has been shown to be the dominant process in reactors feed with H₂, outcompeting homoacetogenesis (Díaz et al., 2015). The H₂ consumption processes both use CO₂ resulting in an increase of pH (Figure 5), which could lead to inhibition of the processes.

The processes of syntrophic acetogenesis and homoacetogenetic oxidation can take place in the bulk reactor outside the zone of H₂ supply from the membrane. Both processes produce CO₂ and H₂ which can feed the CH₄ production by hydrogenotrophic methanogens. Acetogenesis and homoacetogenesis in the bulk reactor create good conditions for acetoclastic methanogenesis (Demirel and Scherer, 2008), producing CO₂ and thus contributing to the CO₂ supply of the high hydrogenotrophic activity near the membrane.

The structure and control of the activity near the membrane is thus regulated by the supply of H₂ and CO₂. The H₂ supply to the zone must be balanced with the CO₂ supply from local production and diffusion from nearby areas. An excess of H₂ supply as compared to the supply of CO₂ leads to the increase in pH which results in the inhibition of several microbial processes and an instability of the system (Wang et al., 2013; Bensmann et al., 2014). A control of H₂ supply and the maintenance of moderate pH values around 8 have been demonstrated to mediate an almost complete conversion of CO₂ to CH₄ in membrane reactors (98–99% CH₄ in the produced biogas) with stable microbial communities, without affecting the microbial degradation of organic matter (Wang et al., 2013), inferring that most of the reactor was still characterized by very low H₂ concentrations allowing for syntrophic acetogenesis (Wang et al., 2013). However, an additional external CO₂ supply could allow for higher amounts of methane being produced.

The use of small-scale biogas reactors is an increasing practice in livestock farms from developing countries (Pham et al., 2014). The digestion of animal manure at relatively low temperatures (25°C) has been proven to be enough for a modest biogas production (<1 m³ biogas/Kg volatile solids). However, the efficiency of the process is relatively low and start-up times longer than 1 month are needed (Zeeman et al., 1988; Alkaya et al., 2010; Pham et al., 2014). The injection of H₂ to untreated wet manure is, however, a different scenario. The methanogenic community is already present in the manure (Zeeman et al., 1988; Demirel and Scherer, 2008) and an injection of H₂ will boost the growth of the community. In our small reactor, we measured a H₂ consumption rate of 20 m³ H₂ m^-3 manure day^-1 (Figure 3) after 3 days. The H₂ consumption may, however, not result in a 100% conversion to CH₄. A large fraction (up to 70%) of the H₂ consumption has been reported to be used for growth of the methanogens and not for methane production during the start-up phase of a reactor (Díaz et al., 2015). The only about 60 h needed to obtain maximum activity in the raw manure (Figure 2) is a very short time period as compared to the start-up times needed for cattle manure reactors (Zeeman et al., 1988). Considering the integrated consumption rates as a proxy for the size of the microbial community, maximum values of 0.05–0.08 h^-1, corresponding to doubling times of 9–14 h of the hydrogenotrophic microbial community, were measured (Figure 3). These values are in the range of the optimal doubling times of several species of mesophilic methanogens (Demirel and Scherer, 2008). Our data illustrate that the growth rate of the methanogenic community in raw cattle manure is high and the long start up times for methanogenic digestors fed with cattle manure thus cannot be due to the methanogens.

The zone with measurable H₂ concentrations compared to the total thickness of the applied matrix (about 1.5 cm) was relatively low throughout the experiment. Even at the initial conditions, when H₂ penetrated more than 5 mm (Figure 3), a large fraction of the manure was net-producing CO₂, and the diffusional supply from this production must have prevented CO₂ limitation of the process near the membrane and buffered the potential increase in pH.

The start-up of an anaerobic reactor may require some time (weeks) but once established the populations of the different groups of fermenters, acetogens and methanogens are active and balanced (Griffin et al., 1998). As pointed by Díaz et al. (2015), the introduction of H₂ into an anaerobic reactor does not require the use of specific strains, and it is possible to use for instance unspecific anaerobic sludge inoculum to produce an active hydrogenotrophic methanogenic reactor. The relatively slow growth rates of acetogens and acetoclastic methanogens result in relatively long reactor start-up times before a stable methane production can be obtained (Luo and Angelidaki, 2013; Ghorbanian et al., 2014; Díaz et al., 2015). However, the supply of H₂ to an established anaerobic reactor results in the immediate activation of the hydrogenotrophic community, a fast consumption of H₂ and an increase in CH₄ production (Luo and Angelidaki, 2012; Wang et al., 2013; Xu et al., 2014; Hu et al., 2015). Our data confirm the predisposition of the microbial community present in a mesophilic anaerobic reactor to metabolize an external supply of H₂ (Figures 4 and 5), and also provide a relevant description of the microstructure of the process.

One of the main problems previously reported by H₂ addition to reactors is the limitation of H₂ transfer to the reactor fluid (Luo et al., 2012; Luo and Angelidaki, 2013; Bensmann et al., 2014) by direct supply of H₂ gas to the reactor, resulting in substantial H₂ concentrations in the produced biogas. The use of gas permeable membranes for the H₂ supply makes it possible to have localized high H₂ concentrations and consumption rates (Figures 3 and 4) without massive ebullition of H₂ to the headspace. Laboratory scale membrane reactors have thus been shown to produce biogas with only trace concentrations of H₂ in the produced biogas and a high quality of up to 99% CH₄ (Wang et al., 2013). The performance of membrane reactors is, however, limited by the amount of H₂ that can diffuse out of the membranes, and membrane reactors must thus contain a large surface area of permeable tubing. Hollow fiber membranes may be applied (Wang et al., 2013; Díaz et al., 2015) instead of silicone tubing, but the cost of these membranes might prohibit full scale applications.

Addition of H₂ to anaerobic reactors can result in high rates of methanogenesis, but as the H₂ supply is not coming from a catabolic process with a parallel CO₂ production there may be a problem with CO₂ depletion and rise in pH. The interruption of the CO₂ supply as shown in Figures 4 and 5 resulted in an immediate decrease in H₂ consumption rate and an increase in pH (Figures 4 and 5). Although no external CO₂ was supplied, the small internal CO₂ production and CO₂/HCO₃^- pool were supporting H₂ consumption rates for a few hours, dropping to low values after 10 h. During this period, the H₂ consumption rates decreased progressively while the CO₂ depletion resulted in an increase of pH to values of about 9.2 (Figure 5), similar to the values measured in other reactors exposed to high H₂ loadings (Wang et al., 2013; Xu et al., 2014). This value corresponds to the pK_a of ammonium (9.25), which is present at high concentrations in anaerobic reactors and therefore at this pH will contribute to buffering compound in the reactor. Several microbial processes may be inhibited at such high pH, resulting in the stop of the anaerobic digestion and the destabilization of the reactor, finally leading to the dramatic decrease in H₂ consumption and CH₄ production (Wang et al., 2013). Especially the high concentration of free NH₃ at high pH values may be very inhibitory for both methanogens and syntrophic acetogens (Wang et al., 2015). The supply of CO₂ together with the H₂ through the silicone membrane following CO₂ depletion resulted in a partial recovery of the H₂ consumption rates (Figure 5). The optimal CO₂ content in the feeding gas producing the highest H₂ consumption rates was 5.7% (Figures 5 and 6), which seems to be far away for the theoretical 17.8% CO₂ needed to react with the 71% H₂ following a 1:4 stoichiometry for hydrogenotrophic methanogenesis. However, the CO₂ permeability in silicone membrane is about five times higher than for H₂ (Robb, 1968), resulting in higher fluxes through the membrane. Consequently, the CO₂ supplied by the membrane was enough to support the high biological demand from the hydrogenotrophic methanogenesis, decreasing the pH to values below 8 (Figures 5 and 6). pH values below 8 are considered as optimal for anaerobic digestion (Luo and Angelidaki, 2013; Wang et al., 2013) and allow for the highest CH₄ upgrade (Wang et al., 2013). Higher increases in the proportion of CO₂ did not increase the H₂ consumption rates (Figures 5 and 6), but decreased the pH of the matrix close to the membrane.

A balanced CO₂ supply, being external or internal, is crucial for maintaining reactor performance by addition of H₂ to a methanogenic reactor. The internal capacity of the reactor to produce CO₂, including the extraction from bicarbonate resulting in moderate pH increases, should thus not be exceeded to maintain stable microbial processes. Addition of H₂ to a methanogenic reactor through a membrane or by bubbling will probably mostly be used to upgrade the methane content of the biogas by reduction of the high CO₂ levels present in the slurry, but it may be an advantage to have the possibility for adding more CO₂ along with the H₂ to facilitate recovery in reactors temporally exposed to an overdose of H₂. The present study is from a membrane reactor, and methanogenic membrane reactors may be difficult to implement at full scale. The results illustrate, however, that H₂ may be depleted at a very short distance from a rising H₂ bubble as the penetration of H₂ was ∼0.5 mm from the membrane into the digestate. Apart from a gradual rise in bulk pH, a supply of H₂ to biogas reactors may thus only have local consequences on microbial transformations such as interspecies H₂ transfer.