Bioresource Technology 173 (2014) 185–192
Contents lists available at ScienceDirect
Bioresource Technology journal homepage: www.elsevier.com/locate/biortech
Substrate milling pretreatment as a key parameter for Solid-State Anaerobic Digestion optimization J.-C. Motte, R. Escudié, J. Hamelin, J.-P. Steyer, N. Bernet, J.-P. Delgenes, C. Dumas ⇑ INRA, UR0050, Laboratoire de Biotechnologie de l’Environnement, Avenue des Etangs, F-11100 Narbonne, France
h i g h l i g h t s
g r a p h i c a l a b s t r a c t
The dynamic of Solid-State Anaerobic
Digestion of wheat straw was widely investigated. Fine milling increases the fermentation rates and the risk of digester failure. Reaction progress affects more microbial communities than particle size. Management of the soluble phase is essential for high-solid AD optimization.
a r t i c l e
i n f o
Article history: Received 17 June 2014 Received in revised form 1 September 2014 Accepted 4 September 2014 Available online 16 September 2014 Keywords: Dry anaerobic digestion Degradation dynamics Particle size Wheat straw Substrate soluble compounds
a b s t r a c t The effect of milling pretreatment on performances of Solid-State Anaerobic Digestion (SS-AD) of raw lignocellulosic residue is still controverted. Three batch reactors treating different straw particle sizes (milled 0.25 mm, 1 mm and 10 mm) were followed during 62 days (6 sampling dates). Although a fine milling improves substrate accessibility and conversion rate (up to 30% compared to coarse milling), it also increases the risk of media acidification because of rapid and high acids production during fermentation of the substrate soluble fraction. Meanwhile, a gradual adaptation of microbial communities, were observed according to both reaction progress and methanogenic performances. The study concluded that particle size reduction affected strongly the performances of the reaction due to an increase of substrate bioaccessibility. An optimization of SS-AD processes thanks to particle size reduction could therefore be applied at farm or industrial scale only if a specific management of the soluble compounds is established. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction Anaerobic digestion (AD) is a process developed at industrial scale for its ability to produce a biogas rich in methane from organic matters frequently considered as wastes (Mata-Alvarez et al., 2000). Among the solid wastes available for AD, agricultural residues are attractive because of the co-production of a digestate rich in nutrients that holds agronomic qualities that can be locally use as fertilizer (Karthikeyan and Visvanathan, 2012). Water
⇑ Corresponding author. Tel.: +33 4 68 42 51 76. E-mail address: [email protected] (C. Dumas). http://dx.doi.org/10.1016/j.biortech.2014.09.015 0960-8524/Ó 2014 Elsevier Ltd. All rights reserved.
content within the process is used to define two types of process configuration: wet digestion with a Total Solid (TS) content below 15% and the Solid-State Anaerobic Digestion (SS-AD) with a TS content higher than 20% (Mata-Alvarez et al., 2000). Because of a lower water content, SS-AD operation is generally associated with an increase of the organic loading rate, a reduction of the specific reactor volume, a decrease of specific energy consumption for heating the process and a simplification of the final digestate dewatering step (Karthikeyan and Visvanathan, 2012). However, SS-AD is not fully optimized and the performances are generally lower than those obtained with wet AD (Brown et al., 2012). At farm scale, agricultural solid wastes are often treated in batch SS-AD systems (Karthikeyan and Visvanathan, 2012). These
186
J.-C. Motte et al. / Bioresource Technology 173 (2014) 185–192
residues are constituted of biodegradable polymers, mainly cellulose and hemicelluloses, which accessibility and bioavailability are impaired by their bounds with the lignin known as a recalcitrant polymer (Monlau et al., 2013). Because of this complex structure, the limiting step of lignocellulose degradation is generally the hydrolysis step. Therefore, many researches were focused on lignocellulosic substrate pretreatment to improve their bioaccessibility (Monlau et al., 2013). Among them, size reduction (i.e. milling) is technologically appropriate for SS-AD at farm scale, since water addition and chemical compounds are not required (Barakat et al., 2013). The main effect of substrate grinding is to provide a greater access for microorganisms to the lignocellulosic tissues, by increasing the surface/volume ratio. Considering the size of wheat straw cells, from 0.5 to 3 mm of length and from 10 to 25 lm of diameter (Sun, 2010), the access to the central vacuole by the microorganisms, which constitutes a pre-required condition to maximize straw accessibility and its degradation, is favored by grinding bellow 1 mm. In wet AD, the size reduction generally improves the reaction kinetics and the degradation performances (Lindmark et al., 2012; Mshandete et al., 2006). However, the effect of particle size reduction in SS-AD is still poorly understood because of a lack of scientific research on this topic. Motte et al. (2013a) showed a strong risk of acidification for a fine milling, which was attributed to the increase of the soluble substrate fraction after milling (Tamaki and Mazza, 2010). However, in this study dynamic information were obtained only from the gas production, but not by direct metabolites analysis. In addition, acidification was favored by the low inoculation rate, i.e. a substrate/inoculum (S/X) ratio higher than 25 (in volatile solid (VS) basis). Because batch digesters are often performed at a S/X ratio of 3 in VS basis (Cui et al., 2011), the effect of particle size reduction in batch processes had to be investigated for this standard inoculation rate. The aim of this study was to understand how particle size reduction affects the SS-AD dynamics during the different steps of the biodegradation. To this end, the SS-AD dynamics was evaluated in terms of process performances (methane production and substrate conversion), fermentative activity (accumulation of metabolites) and microbial adaptation (bacterial and archae), thanks to an original sampling method that avoids bioreactor opening, and thus disturbances of the bioreaction. The reactions dynamics were followed by performing six samplings at different dates from the start-up, the initial hydrolysis and fermentation, the acids conversion into methane or their inhibition activity (if any), until the end of the reaction.
2. Methods 2.1. Solid-state anaerobic digesters 2.1.1. Substrate preparation Organic wheat straw (Triticum aestivum) used as model substrate of lignocellulosic residues, was harvested in summer 2010 in southern region of France (Hérault). Biochemical characteristics of the raw straw were previously presented in Motte et al. (2013a). Three fractions of different straw particle sizes were obtained by using a cutting miller (RETSCHÒ SM100), the size of the grid determining the particle size of the substrate: a coarse fraction with a 10 mm grid, a medium fraction with a 1 mm grid and a fine fraction with a 0.25 mm grid. Because of technical limitations, the size distributions of coarse and medium fractions were measured with an analytical sieve shaker, while fine fraction was analyzed with a laser granulometer. Biochemical Methane Potential (BMP) of each fraction was evaluated according to the protocol described in Motte et al. (2014).
2.1.2. Experimental device SS-AD batch tests were carried out in reactors specifically designed for a sampling without perturbing the gas headspace (patent FR13 504493). The composition of the initial medium was prepared to obtain a S/X ratio of 3 (in VS basis) and a TS content of 22%. The inoculum was a leachate of digestate sampled in a full-scale solid-state digester plant treating municipal solid wastes. 2L-reactors were filled with 1 kg of media containing 160 g of straw, 825 mL of diluted inoculum, 2.6 g of bicarbonate buffer and 1.5 g of NH4Cl to adjust the C/N ratio at 40. The initial pH was alkaline at 8.6 ± 0.2. Small variations in the initial moisture content were recorded (0.05 ns >0.05 ns
188
J.-C. Motte et al. / Bioresource Technology 173 (2014) 185–192
Table 2 Final process performances for the reactors operated with different particle size fractions. Straw fraction g1 VS )
Methane production (NmL Metabolite accumulation (gCOD g1 TS ) pH Substrate conversion (%COD) Lignin + cellulose + hemicelluloses (% degradation)
Fine
Medium
Coarse
p-Value
13 ± 3 207 ± 2 5.5 ± 0.1 24.5% ± 0.7 14.3% ± 1.0
168 ± 6 8±4 8.2 ± 0.1 36.1% ± 0.1 37.8% ± 0.7
192 ± 25 7±1 8.4 ± 0.1 42.9% ± 5.2 34.6% ± 5.1
4.5e4⁄⁄⁄ 3.6e11⁄⁄⁄ 3.2e9⁄⁄⁄ 8.8e4⁄⁄⁄ 1.4e4⁄⁄⁄
around 26 g kg1. In comparison, a pH of 8.4 and a concentration of VFA of 1 g kg1 were measured for the medium and coarse fractions. This reaction interruption was thus attributed to a medium acidification as previously observed in literature (AbbassiGuendouz et al., 2012) and probably caused by the high amount of water-soluble compounds of the fine fraction (Motte et al., 2013a).
during the experiment (days 0, 3, 7, 14, 28, 62) (Fig. 1-a). Fig. 1b presents the methane composition during the reaction progress. Fig. 2 presents the process performances (methane production, VFA production, substrate conversion, pH and TS content) for the different sampling dates. According to these results, two types of process behaviors can be distinguished: functional conditions (coarse and medium fractions) and acidified one (fine fraction).
3.3. Process dynamics The daily biogas production rate of the three particle sizes reveals the process dynamics, thanks to six samplings, performed
3.3.1. Functional conditions During the first day, a peak of biogas production rate (Fig. 1-a) 1 occurred for medium and coarse fractions (18 and 23 NmL g1 , VS d
Fig. 1. Effect of straw particle size on the process dynamics during the first 30 days. (a) Daily biogas production. (b) Methane content. Samples dates are indicated as black triangles. Error bars are SD of three replicates.
J.-C. Motte et al. / Bioresource Technology 173 (2014) 185–192
189
Fig. 2. Evolution of the samples characteristics during SS-AD reaction for the three particle size fractions. (a) Digestate characterisation – Fine fraction. (b) Substrate conversion – Fine fraction. (c) Digestate characterisation – Medium fraction. (d) Substrate conversion – Medium fraction. (e) Digestate characterisation – Coarse fraction. (f) Substrate conversion – Coarse fraction.
respectively). The production rate decreased rapidly during the 1 second and third day down to 5 and 7 NmL g1 . Meanwhile, VS d the methane content of the biogas (Fig. 1-b) was low because of a high CO2 production (around 90% of the biogas produced on day 1). On day 3, i.e. just after the first peak of biogas production, VFA production for both medium and coarse fractions reached about 45 gCOD g1 TS , whereas the methane production was about 15 gCOD g1 TS (Fig. 2). This rapid and high production of CO2 and VFA suggested the expression of fermentative activity at the beginning of the reaction (Shi et al., 2013). Since 7% of the substrate was transformed during this period, this initial production could be related to the degradation of the water-soluble compounds (10–13% of the initial straw fraction), which were the most accessible fraction of the substrate (Buffière et al., 2008). Despite an early start-up of the fermentative activity, the methanogenic communities were not inhibited as revealed by the biogas production rate which increased after the 3rd day (Fig. 1-a). This second phase of reaction, from day 3 to day 12, was characterized 1 by a biogas production rate of 12 and 19 NmL g1 for medium VS d and coarse particle size, respectively, This period was associated 1 with high daily methane production rates (7–9 NmL g1 ) and VS d an increase in methane content up to 50% (Fig. 1-b). The maximal biogas and methane production rates and the date of the peak were of the same order of magnitude as previous studies using similar inoculation ratio (Brown et al., 2012; Cui et al., 2011). On day 7 (peak of maximal methane production rate), a high concentration
1 of VFA was observed (70 gCOD g1 TS and 64 gCOD gTS for the medium and coarse fractions, respectively), associated to a slight decrease of pH (7.6 and 8.1, respectively) (Fig. 2). Despite the high VFA concentration, the alkaline pH allowed to avoid methane production inhibition since it favored the dissociated forms of VFA and reduced their inhibition potential. Three main metabolites (Fig. 3) were produced (acetic, propionic and butyric acids) as usually observed in the SS-AD literature (Motte et al., 2013b; Shi et al., 2013). Starting from day 8, the methane production rates of medium and coarse fractions decreased progressively until the end of the experiment on day 62. This progressive decrease of the reaction kinetics can be related to the variation of medium characteristics (Fig. 2). On day 14 (i.e., end of the maximal peak of methane production rate), a decrease of metabolites concentration down to 45 gCOD g1 TS was observed for both fractions. Fig. 3 shows a strong accumulation of propionic acid, which is known to be hardly degradable in AD (Vavilin and Lokshina, 1996). Meanwhile, around 35% of the final methane production was reached and around 20% of the substrate was converted. The overall VFA consumption was achieved on day 28 (residual concentration below 5 gCOD g1 TS ). As a consequence, part of the methane production observed between days 14 and 28 was associated with the degradation of propionic acid by microorganisms, which needed longer time to activate adequate degradation activity (Vavilin and Lokshina, 1996). At the end of the experiment (day 62), low concentrations of residual metabolites were reported, suggesting that the produced VFA were entirely
190
J.-C. Motte et al. / Bioresource Technology 173 (2014) 185–192
On day 4, a second peak of biogas production appeared with a 1 high maximum production rate of 27 NmL g1 . However, if a VS d fine milling enhanced the biogas production rate at this stage, it suddenly decreased after 4 days and was extremely low until the end of the experiment. In the meantime, the methane content in the headspace decreased slowly. This interruption in the biogas production was associated with a high VFA accumulation (140 gCOD g1 TS ) on day 7. The metabolites profile showed a high concentration of butyric acid (around 50% of the total VFA) and an accumulation of valeric acid (3.5 gCOD g1 TS ), which suggested a complex fermentative metabolisms. Until the end of the experiment, the medium presented a similar composition with acid pH at around 5.5, VFA accumulation higher than 150 gCOD g1 TS and no methane production. Meanwhile, the substrate conversion increased slowly (from 18% to 25%COD from day 7 to day 62) because of the production of other metabolites, such as ethanol, valeric and caproic acid. 3.4. Microbial communities structure
Fig. 3. Dynamic evolution of fermentative products for the three particle sizes. (a) Fine fraction. (b) Medium fraction. (c) Coarse fraction.
transformed into biogas. During the overall experiment, the TS content of the media decreased of about 23% (from 22% to 17%) as previously indicated in the literature (Shi et al., 2013). The progressive increase of water availability during the substrate degradation was able to dilute the inhibitors, and to improve the mass transfer or substrate accessibility (Abbassi-Guendouz et al., 2012). 3.3.2. Acidified conditions The fine fraction was subjected to acidification. As for the functional condition, a peak of biogas production occurred the first day (Fig. 1-a), but was higher for the fine than for the medium and 1 coarse fractions (32 vs. 18 and 23 NmL g1 , respectively). Then, VS d after 3 days of experiments, methane production (18 ± 3 gCOD g1 TS ) was significantly higher than for the functional conditions (14 ± 1 and 15 ± 2 gCOD g1 TS , respectively). On day 3, a high VFA concentration (Fig. 2) of 70 gCOD g1 TS , mainly composed of acetic, propionic and butyric acids, was observed, leading to a slight decrease of pH (7.4). A similar process behavior was thus observed for fine, medium and coarse fractions at the beginning of the reaction.
The structures of microbial communities were analyzed dynamically to understand how particle size affected the microbial activity and the adaptation of microbial communities to specific process conditions. First, the detailed observations of initial and final bacterial CE-SSCP profiles are presented in Fig. 4-a. These profiles showed three main peaks that decreased drastically at the end of the reaction. This revealed a bacterial adaptation face to the substrate and the specific conditions of the SS-AD (AbbassiGuendouz et al., 2013). However, this adaptation was different depending on the particle size. Thus, on day 62, three abundant bacterial species were reported in both functional conditions, while two others were related to acidified conditions. The profiles of the fine fraction were characterized by a quite low Simpson diversity index of 3.2, compared to the efficient conditions characterized by a Simpson diversity index of 4.8. This demonstrated a simplification of bacterial communities for the acidified conditions. Meanwhile, the archaeal community structures were also evaluated (Fig. 4-b). The profiles of every condition, before and after degradation did not show any significant modification: Simpson diversity index was between 2.1 and 2.4 for each condition. Therefore, if an adaptation of the bacterial communities occurred during the degradation process, it was not the case for the archaeal communities. This can be explained by an adaptation of the hydrolytic and fermentative bacteria to the substrate (initially adapted to municipal solid waste), while archaeal communities did not required adaptation since they faced usual substrate (mainly acetic acid) and were already adapted to solid-state conditions. A Principal Component Analysis (PCA) of bacterial communities obtained from CE-SSCP fingerprinting patterns (Fig. 5) was performed to study dynamically the microbial adaptation. On the PCA the three replicates were analyzed per operating conditions and per sampling dates, and the closer the samples, the most similar bacterial compositions they present. The two first PCA axes explained 67.8% of the variance. On day 0, just after inoculation, the microbial structures of the three fractions were similar. A strong modification of the bacterial community structures was observed for all the conditions after only 3 days. Then, the bacterial community structures changed progressively according to the reaction progress and the efficiency of AD reaction. The bacterial communities of the functional conditions (i.e. coarse and medium fractions) followed a similar trend during the substrate degradation. The strongest changes occurred during the first 3 days suggesting that an early adaptation of microbial community appeared in SS-AD. The microbial community structure of fine fraction condition diverged from functional conditions after only 3 days, and this difference increased continuously but slowly with
J.-C. Motte et al. / Bioresource Technology 173 (2014) 185–192
191
time. In addition, CE-SSCP fingerprinting profiles on the archaeal communities did not indicate significant differences between the three fraction conditions, confirming that the failure of the system was caused by a too high fermentative activity. Therefore, the modifications of the microbial communities occurred very early in the process and drove the distinction between efficient conditions and inhibited conditions characterized by a higher accumulation of VFA. Thus, the first change in the bacterial community structure corresponds to the fermentation of the water-soluble compounds (similar for every conditions), while the second phase (different from acidified and efficient conditions) was related to the hydrolysis of the solid fraction of the substrate. Since medium and coarse fraction presented similar microbial communities, it could be concluded that particle size reduction affected indirectly the microbial adaptation through the substrate accessibility and particularly the proportion of water-soluble compounds. 3.5. Effect of straw milling
Fig. 4. Initial and final bacterial (16S) CE-SSCP profiles with the corresponding discriminant peaks. (a) Bacterial community. (b) Archaea community.
Fig. 5. Principal Component Analysis (PCA) of bacterial communities CE-SSCP. Circles represent the fine fraction (F), squares the medium fraction (M), and diamonds the coarse fraction (C).
In the present study, the effects of particle size reduction were mainly observed for a fine milling of the substrate (
Contents lists available at ScienceDirect
Bioresource Technology journal homepage: www.elsevier.com/locate/biortech
Substrate milling pretreatment as a key parameter for Solid-State Anaerobic Digestion optimization J.-C. Motte, R. Escudié, J. Hamelin, J.-P. Steyer, N. Bernet, J.-P. Delgenes, C. Dumas ⇑ INRA, UR0050, Laboratoire de Biotechnologie de l’Environnement, Avenue des Etangs, F-11100 Narbonne, France
h i g h l i g h t s
g r a p h i c a l a b s t r a c t
The dynamic of Solid-State Anaerobic
Digestion of wheat straw was widely investigated. Fine milling increases the fermentation rates and the risk of digester failure. Reaction progress affects more microbial communities than particle size. Management of the soluble phase is essential for high-solid AD optimization.
a r t i c l e
i n f o
Article history: Received 17 June 2014 Received in revised form 1 September 2014 Accepted 4 September 2014 Available online 16 September 2014 Keywords: Dry anaerobic digestion Degradation dynamics Particle size Wheat straw Substrate soluble compounds
a b s t r a c t The effect of milling pretreatment on performances of Solid-State Anaerobic Digestion (SS-AD) of raw lignocellulosic residue is still controverted. Three batch reactors treating different straw particle sizes (milled 0.25 mm, 1 mm and 10 mm) were followed during 62 days (6 sampling dates). Although a fine milling improves substrate accessibility and conversion rate (up to 30% compared to coarse milling), it also increases the risk of media acidification because of rapid and high acids production during fermentation of the substrate soluble fraction. Meanwhile, a gradual adaptation of microbial communities, were observed according to both reaction progress and methanogenic performances. The study concluded that particle size reduction affected strongly the performances of the reaction due to an increase of substrate bioaccessibility. An optimization of SS-AD processes thanks to particle size reduction could therefore be applied at farm or industrial scale only if a specific management of the soluble compounds is established. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction Anaerobic digestion (AD) is a process developed at industrial scale for its ability to produce a biogas rich in methane from organic matters frequently considered as wastes (Mata-Alvarez et al., 2000). Among the solid wastes available for AD, agricultural residues are attractive because of the co-production of a digestate rich in nutrients that holds agronomic qualities that can be locally use as fertilizer (Karthikeyan and Visvanathan, 2012). Water
⇑ Corresponding author. Tel.: +33 4 68 42 51 76. E-mail address: [email protected] (C. Dumas). http://dx.doi.org/10.1016/j.biortech.2014.09.015 0960-8524/Ó 2014 Elsevier Ltd. All rights reserved.
content within the process is used to define two types of process configuration: wet digestion with a Total Solid (TS) content below 15% and the Solid-State Anaerobic Digestion (SS-AD) with a TS content higher than 20% (Mata-Alvarez et al., 2000). Because of a lower water content, SS-AD operation is generally associated with an increase of the organic loading rate, a reduction of the specific reactor volume, a decrease of specific energy consumption for heating the process and a simplification of the final digestate dewatering step (Karthikeyan and Visvanathan, 2012). However, SS-AD is not fully optimized and the performances are generally lower than those obtained with wet AD (Brown et al., 2012). At farm scale, agricultural solid wastes are often treated in batch SS-AD systems (Karthikeyan and Visvanathan, 2012). These
186
J.-C. Motte et al. / Bioresource Technology 173 (2014) 185–192
residues are constituted of biodegradable polymers, mainly cellulose and hemicelluloses, which accessibility and bioavailability are impaired by their bounds with the lignin known as a recalcitrant polymer (Monlau et al., 2013). Because of this complex structure, the limiting step of lignocellulose degradation is generally the hydrolysis step. Therefore, many researches were focused on lignocellulosic substrate pretreatment to improve their bioaccessibility (Monlau et al., 2013). Among them, size reduction (i.e. milling) is technologically appropriate for SS-AD at farm scale, since water addition and chemical compounds are not required (Barakat et al., 2013). The main effect of substrate grinding is to provide a greater access for microorganisms to the lignocellulosic tissues, by increasing the surface/volume ratio. Considering the size of wheat straw cells, from 0.5 to 3 mm of length and from 10 to 25 lm of diameter (Sun, 2010), the access to the central vacuole by the microorganisms, which constitutes a pre-required condition to maximize straw accessibility and its degradation, is favored by grinding bellow 1 mm. In wet AD, the size reduction generally improves the reaction kinetics and the degradation performances (Lindmark et al., 2012; Mshandete et al., 2006). However, the effect of particle size reduction in SS-AD is still poorly understood because of a lack of scientific research on this topic. Motte et al. (2013a) showed a strong risk of acidification for a fine milling, which was attributed to the increase of the soluble substrate fraction after milling (Tamaki and Mazza, 2010). However, in this study dynamic information were obtained only from the gas production, but not by direct metabolites analysis. In addition, acidification was favored by the low inoculation rate, i.e. a substrate/inoculum (S/X) ratio higher than 25 (in volatile solid (VS) basis). Because batch digesters are often performed at a S/X ratio of 3 in VS basis (Cui et al., 2011), the effect of particle size reduction in batch processes had to be investigated for this standard inoculation rate. The aim of this study was to understand how particle size reduction affects the SS-AD dynamics during the different steps of the biodegradation. To this end, the SS-AD dynamics was evaluated in terms of process performances (methane production and substrate conversion), fermentative activity (accumulation of metabolites) and microbial adaptation (bacterial and archae), thanks to an original sampling method that avoids bioreactor opening, and thus disturbances of the bioreaction. The reactions dynamics were followed by performing six samplings at different dates from the start-up, the initial hydrolysis and fermentation, the acids conversion into methane or their inhibition activity (if any), until the end of the reaction.
2. Methods 2.1. Solid-state anaerobic digesters 2.1.1. Substrate preparation Organic wheat straw (Triticum aestivum) used as model substrate of lignocellulosic residues, was harvested in summer 2010 in southern region of France (Hérault). Biochemical characteristics of the raw straw were previously presented in Motte et al. (2013a). Three fractions of different straw particle sizes were obtained by using a cutting miller (RETSCHÒ SM100), the size of the grid determining the particle size of the substrate: a coarse fraction with a 10 mm grid, a medium fraction with a 1 mm grid and a fine fraction with a 0.25 mm grid. Because of technical limitations, the size distributions of coarse and medium fractions were measured with an analytical sieve shaker, while fine fraction was analyzed with a laser granulometer. Biochemical Methane Potential (BMP) of each fraction was evaluated according to the protocol described in Motte et al. (2014).
2.1.2. Experimental device SS-AD batch tests were carried out in reactors specifically designed for a sampling without perturbing the gas headspace (patent FR13 504493). The composition of the initial medium was prepared to obtain a S/X ratio of 3 (in VS basis) and a TS content of 22%. The inoculum was a leachate of digestate sampled in a full-scale solid-state digester plant treating municipal solid wastes. 2L-reactors were filled with 1 kg of media containing 160 g of straw, 825 mL of diluted inoculum, 2.6 g of bicarbonate buffer and 1.5 g of NH4Cl to adjust the C/N ratio at 40. The initial pH was alkaline at 8.6 ± 0.2. Small variations in the initial moisture content were recorded (0.05 ns >0.05 ns
188
J.-C. Motte et al. / Bioresource Technology 173 (2014) 185–192
Table 2 Final process performances for the reactors operated with different particle size fractions. Straw fraction g1 VS )
Methane production (NmL Metabolite accumulation (gCOD g1 TS ) pH Substrate conversion (%COD) Lignin + cellulose + hemicelluloses (% degradation)
Fine
Medium
Coarse
p-Value
13 ± 3 207 ± 2 5.5 ± 0.1 24.5% ± 0.7 14.3% ± 1.0
168 ± 6 8±4 8.2 ± 0.1 36.1% ± 0.1 37.8% ± 0.7
192 ± 25 7±1 8.4 ± 0.1 42.9% ± 5.2 34.6% ± 5.1
4.5e4⁄⁄⁄ 3.6e11⁄⁄⁄ 3.2e9⁄⁄⁄ 8.8e4⁄⁄⁄ 1.4e4⁄⁄⁄
around 26 g kg1. In comparison, a pH of 8.4 and a concentration of VFA of 1 g kg1 were measured for the medium and coarse fractions. This reaction interruption was thus attributed to a medium acidification as previously observed in literature (AbbassiGuendouz et al., 2012) and probably caused by the high amount of water-soluble compounds of the fine fraction (Motte et al., 2013a).
during the experiment (days 0, 3, 7, 14, 28, 62) (Fig. 1-a). Fig. 1b presents the methane composition during the reaction progress. Fig. 2 presents the process performances (methane production, VFA production, substrate conversion, pH and TS content) for the different sampling dates. According to these results, two types of process behaviors can be distinguished: functional conditions (coarse and medium fractions) and acidified one (fine fraction).
3.3. Process dynamics The daily biogas production rate of the three particle sizes reveals the process dynamics, thanks to six samplings, performed
3.3.1. Functional conditions During the first day, a peak of biogas production rate (Fig. 1-a) 1 occurred for medium and coarse fractions (18 and 23 NmL g1 , VS d
Fig. 1. Effect of straw particle size on the process dynamics during the first 30 days. (a) Daily biogas production. (b) Methane content. Samples dates are indicated as black triangles. Error bars are SD of three replicates.
J.-C. Motte et al. / Bioresource Technology 173 (2014) 185–192
189
Fig. 2. Evolution of the samples characteristics during SS-AD reaction for the three particle size fractions. (a) Digestate characterisation – Fine fraction. (b) Substrate conversion – Fine fraction. (c) Digestate characterisation – Medium fraction. (d) Substrate conversion – Medium fraction. (e) Digestate characterisation – Coarse fraction. (f) Substrate conversion – Coarse fraction.
respectively). The production rate decreased rapidly during the 1 second and third day down to 5 and 7 NmL g1 . Meanwhile, VS d the methane content of the biogas (Fig. 1-b) was low because of a high CO2 production (around 90% of the biogas produced on day 1). On day 3, i.e. just after the first peak of biogas production, VFA production for both medium and coarse fractions reached about 45 gCOD g1 TS , whereas the methane production was about 15 gCOD g1 TS (Fig. 2). This rapid and high production of CO2 and VFA suggested the expression of fermentative activity at the beginning of the reaction (Shi et al., 2013). Since 7% of the substrate was transformed during this period, this initial production could be related to the degradation of the water-soluble compounds (10–13% of the initial straw fraction), which were the most accessible fraction of the substrate (Buffière et al., 2008). Despite an early start-up of the fermentative activity, the methanogenic communities were not inhibited as revealed by the biogas production rate which increased after the 3rd day (Fig. 1-a). This second phase of reaction, from day 3 to day 12, was characterized 1 by a biogas production rate of 12 and 19 NmL g1 for medium VS d and coarse particle size, respectively, This period was associated 1 with high daily methane production rates (7–9 NmL g1 ) and VS d an increase in methane content up to 50% (Fig. 1-b). The maximal biogas and methane production rates and the date of the peak were of the same order of magnitude as previous studies using similar inoculation ratio (Brown et al., 2012; Cui et al., 2011). On day 7 (peak of maximal methane production rate), a high concentration
1 of VFA was observed (70 gCOD g1 TS and 64 gCOD gTS for the medium and coarse fractions, respectively), associated to a slight decrease of pH (7.6 and 8.1, respectively) (Fig. 2). Despite the high VFA concentration, the alkaline pH allowed to avoid methane production inhibition since it favored the dissociated forms of VFA and reduced their inhibition potential. Three main metabolites (Fig. 3) were produced (acetic, propionic and butyric acids) as usually observed in the SS-AD literature (Motte et al., 2013b; Shi et al., 2013). Starting from day 8, the methane production rates of medium and coarse fractions decreased progressively until the end of the experiment on day 62. This progressive decrease of the reaction kinetics can be related to the variation of medium characteristics (Fig. 2). On day 14 (i.e., end of the maximal peak of methane production rate), a decrease of metabolites concentration down to 45 gCOD g1 TS was observed for both fractions. Fig. 3 shows a strong accumulation of propionic acid, which is known to be hardly degradable in AD (Vavilin and Lokshina, 1996). Meanwhile, around 35% of the final methane production was reached and around 20% of the substrate was converted. The overall VFA consumption was achieved on day 28 (residual concentration below 5 gCOD g1 TS ). As a consequence, part of the methane production observed between days 14 and 28 was associated with the degradation of propionic acid by microorganisms, which needed longer time to activate adequate degradation activity (Vavilin and Lokshina, 1996). At the end of the experiment (day 62), low concentrations of residual metabolites were reported, suggesting that the produced VFA were entirely
190
J.-C. Motte et al. / Bioresource Technology 173 (2014) 185–192
On day 4, a second peak of biogas production appeared with a 1 high maximum production rate of 27 NmL g1 . However, if a VS d fine milling enhanced the biogas production rate at this stage, it suddenly decreased after 4 days and was extremely low until the end of the experiment. In the meantime, the methane content in the headspace decreased slowly. This interruption in the biogas production was associated with a high VFA accumulation (140 gCOD g1 TS ) on day 7. The metabolites profile showed a high concentration of butyric acid (around 50% of the total VFA) and an accumulation of valeric acid (3.5 gCOD g1 TS ), which suggested a complex fermentative metabolisms. Until the end of the experiment, the medium presented a similar composition with acid pH at around 5.5, VFA accumulation higher than 150 gCOD g1 TS and no methane production. Meanwhile, the substrate conversion increased slowly (from 18% to 25%COD from day 7 to day 62) because of the production of other metabolites, such as ethanol, valeric and caproic acid. 3.4. Microbial communities structure
Fig. 3. Dynamic evolution of fermentative products for the three particle sizes. (a) Fine fraction. (b) Medium fraction. (c) Coarse fraction.
transformed into biogas. During the overall experiment, the TS content of the media decreased of about 23% (from 22% to 17%) as previously indicated in the literature (Shi et al., 2013). The progressive increase of water availability during the substrate degradation was able to dilute the inhibitors, and to improve the mass transfer or substrate accessibility (Abbassi-Guendouz et al., 2012). 3.3.2. Acidified conditions The fine fraction was subjected to acidification. As for the functional condition, a peak of biogas production occurred the first day (Fig. 1-a), but was higher for the fine than for the medium and 1 coarse fractions (32 vs. 18 and 23 NmL g1 , respectively). Then, VS d after 3 days of experiments, methane production (18 ± 3 gCOD g1 TS ) was significantly higher than for the functional conditions (14 ± 1 and 15 ± 2 gCOD g1 TS , respectively). On day 3, a high VFA concentration (Fig. 2) of 70 gCOD g1 TS , mainly composed of acetic, propionic and butyric acids, was observed, leading to a slight decrease of pH (7.4). A similar process behavior was thus observed for fine, medium and coarse fractions at the beginning of the reaction.
The structures of microbial communities were analyzed dynamically to understand how particle size affected the microbial activity and the adaptation of microbial communities to specific process conditions. First, the detailed observations of initial and final bacterial CE-SSCP profiles are presented in Fig. 4-a. These profiles showed three main peaks that decreased drastically at the end of the reaction. This revealed a bacterial adaptation face to the substrate and the specific conditions of the SS-AD (AbbassiGuendouz et al., 2013). However, this adaptation was different depending on the particle size. Thus, on day 62, three abundant bacterial species were reported in both functional conditions, while two others were related to acidified conditions. The profiles of the fine fraction were characterized by a quite low Simpson diversity index of 3.2, compared to the efficient conditions characterized by a Simpson diversity index of 4.8. This demonstrated a simplification of bacterial communities for the acidified conditions. Meanwhile, the archaeal community structures were also evaluated (Fig. 4-b). The profiles of every condition, before and after degradation did not show any significant modification: Simpson diversity index was between 2.1 and 2.4 for each condition. Therefore, if an adaptation of the bacterial communities occurred during the degradation process, it was not the case for the archaeal communities. This can be explained by an adaptation of the hydrolytic and fermentative bacteria to the substrate (initially adapted to municipal solid waste), while archaeal communities did not required adaptation since they faced usual substrate (mainly acetic acid) and were already adapted to solid-state conditions. A Principal Component Analysis (PCA) of bacterial communities obtained from CE-SSCP fingerprinting patterns (Fig. 5) was performed to study dynamically the microbial adaptation. On the PCA the three replicates were analyzed per operating conditions and per sampling dates, and the closer the samples, the most similar bacterial compositions they present. The two first PCA axes explained 67.8% of the variance. On day 0, just after inoculation, the microbial structures of the three fractions were similar. A strong modification of the bacterial community structures was observed for all the conditions after only 3 days. Then, the bacterial community structures changed progressively according to the reaction progress and the efficiency of AD reaction. The bacterial communities of the functional conditions (i.e. coarse and medium fractions) followed a similar trend during the substrate degradation. The strongest changes occurred during the first 3 days suggesting that an early adaptation of microbial community appeared in SS-AD. The microbial community structure of fine fraction condition diverged from functional conditions after only 3 days, and this difference increased continuously but slowly with
J.-C. Motte et al. / Bioresource Technology 173 (2014) 185–192
191
time. In addition, CE-SSCP fingerprinting profiles on the archaeal communities did not indicate significant differences between the three fraction conditions, confirming that the failure of the system was caused by a too high fermentative activity. Therefore, the modifications of the microbial communities occurred very early in the process and drove the distinction between efficient conditions and inhibited conditions characterized by a higher accumulation of VFA. Thus, the first change in the bacterial community structure corresponds to the fermentation of the water-soluble compounds (similar for every conditions), while the second phase (different from acidified and efficient conditions) was related to the hydrolysis of the solid fraction of the substrate. Since medium and coarse fraction presented similar microbial communities, it could be concluded that particle size reduction affected indirectly the microbial adaptation through the substrate accessibility and particularly the proportion of water-soluble compounds. 3.5. Effect of straw milling
Fig. 4. Initial and final bacterial (16S) CE-SSCP profiles with the corresponding discriminant peaks. (a) Bacterial community. (b) Archaea community.
Fig. 5. Principal Component Analysis (PCA) of bacterial communities CE-SSCP. Circles represent the fine fraction (F), squares the medium fraction (M), and diamonds the coarse fraction (C).
In the present study, the effects of particle size reduction were mainly observed for a fine milling of the substrate (
