Appl Biochem Biotechnol (2014) 172:1699–1713 DOI 10.1007/s12010-013-0607-2
High Solids Enzymatic Hydrolysis of Pretreated Lignocellulosic Materials with a Powerful Stirrer Concept Daniel Ludwig & Buchmann Michael & Thomas Hirth & Steffen Rupp & Susanne Zibek
Received: 16 April 2013 / Accepted: 23 October 2013 / Published online: 19 November 2013 # Springer Science+Business Media New York 2013
Abstract In this study, we present a powerful stirred tank reactor system that can efficiently hydrolyse lignocellulosic material at high solid content to produce hydrolysates with glucose concentration > 100 g/kg. As lignocellulosic substrates alkaline-pretreated wheat straw and organosolv-pretreated beech wood were used. The developed vertical reactor was equipped with a segmented helical stirrer, which was specially designed for high biomass hydrolysis. The stirrer was characterised according to mixing behaviour and power input. To minimise the cellulase dosage, a response surface plan was used. With the empirical relationship between glucose yield, cellulase loading and solid content, the minimal cellulase dosage was calculated to reach at least 70 % yield at high glucose and high substrate concentrations within 48 h. The optimisation resulted in a minimal enzyme dosage of 30 FPU/g dry matter (DM) for the hydrolysis of wheat straw and 20 FPU/g DM for the hydrolysis of beech wood. By transferring the hydrolysis reaction from shaking flasks to the stirred tank reactor, the glucose yields could be increased. Using the developed stirred tank reactor system, alkaline-pretreated wheat straw could be converted to 110 g/kg glucose (76 %) at a solid content of 20 % (w/w) after 48 h. Organosolv-pretreated beech wood could be efficiently hydrolysed even at 30 % (w/w) DM, giving 150 g/kg glucose (72 %). Keywords Lignocellulose biomass . High solid content . Enzymatic hydrolysis . Stirred tank reactor D. Ludwig Evonik Industries AG, Rodenbacher Chaussee 4, 63457 Hanau-Wolfgang, Germany T. Hirth Institute of Interfacial Process Engineering and Plasma Technology IGVP, University of Stuttgart, Nobelstraße 12, 70569 Stuttgart, Germany T. Hirth : S. Rupp : S. Zibek (*) Fraunhofer-Institute for Interfacial Engineering and Biotechnology IGB, Nobelstraße 12, 70569 Stuttgart, Germany e-mail: [email protected] B. Michael Linde Engineering Dresden GmbH, Bodenbacher Straße 80, 01277 Dresden, Germany
1700
Appl Biochem Biotechnol (2014) 172:1699–1713
Introduction Lignocellulosic biomass is one of the most promising renewable resources that can be converted to bio-based chemicals, biofuels and bioenergy [1]. Especially its high carbohydrate content of about 60–75 % provides a huge potential to produce fermentable sugars without competing directly with food and feed industry [2]. Different types of resources are possible as feedstocks [3]. For example, waste products from forestry or agricultural industry like crop straw or special biomass plants like miscanthus or switchgrass can be used. The latter two can be grown on marginal land at high biomass production rates [4]. There is also the possibility to use softwoods and hardwoods. In contrast to sugar- and starch-based feedstocks like sugar beet or corn, lignocellulosic biomass has to be pretreated by physical or chemical methods to increase the cellulose conversion of the following hydrolysis process [5]. Cellulose hydrolysis is catalysed by a cellulase complex, which includes three main classes of glucoside hydrolases, namely, cellobiohydrolases (CBH), endoglucosidases (EG) and beta-glucosidases (BG) [6, 7]. CBH catalyses the hydrolysis from both polymer ends (reducing and nonreducing) producing cellobiose, whereras EG cleaves internal glucosidic bonds producing smaller polymers, and hence, new adsorption sites for the CBH. Finally, BG catalyses the hydrolysis of cellobiose to glucose, and hence, product inhibition of CBH and EG is reduced. Considering the different interfering reaction steps, the cellulase system shows a high degree of synergism [8]. From starch-based processes, it is known that the dry matter (DM) content should exceed 30 % (w/w) to realise an economic feasible conversion process [9, 10]. Therefore, the solid content should also be increased to >10 % (w/w) DM to realise an economic feasible process for the hydrolysis of lignocellulosic biomass [11–13]. By increasing the substrate concentration, the investment and production costs are decreased because smaller reactors and less heating or cooling energy are needed for the same sugar productivity. With higher substrate concentrations, the space-time yields are significantly enhanced and the high concentrated sugar hydrolysates will lead to more economic downstreaming processes. However, some disadvantages occur, when dealing with high biomass hydrolysis [14]. In general, the viscosity of the fibre suspension is increased dramatically. Consequently, mass and heat transfer problems will have a major drawback on the hydrolysis process. Usually, the sugar yield is decreasing with increasing solid content. Kristensen et al. summarised different high biomass hydrolysis experiments and could show that this tendency is valid for different feedstocks as well as for separate (SHF) and simultaneous saccharification fermentation (SSF) processes [15]. As a consequence, the reactor design will play an important role by realising an effective hydrolysis process. Besides some shaking flask or roller bottle experiments at a millilitre scale, only a few reactor concepts are investigated in literature, which can be applied as an industrial reactor concept. These reactor types can be generally divided into horizontal and vertical reactors. Jorgensen et al. introduced a horizontal five-chamber liquefaction reactor [14, 16, 17]. Each chamber was homogenised by three paddlers that were mounted on a horizontal rotating shaft. By using this reactor concept, the achievable glucose yields ranged from 60 at 20 % (w/w) DM to 35 at 40 % (w/w) DM after 96 h. It has to be mentioned that the yields could not be significantly increased when switching to a SSF process. Rosgaard et al. conducted batch and fed batch hydrolysis processes in a 2-L bottle (vertical system) with special six custommade impellers [18]. Steam-treated barley straw was used as substrate. They investigated the influence of different solid concentration up to 15 % (w/w) DM on glucose yield during batch and fed batch hydrolysis reactions. It could be pointed out, that no clear difference occurred between fed batch and batch processes up to 15 % (w/w) DM. The glucose yield was 80 % after 72 h. Hodge et al. compared a batch shaking flask experiment with fed batch experiments
Appl Biochem Biotechnol (2014) 172:1699–1713
1701
in a 7-L vertical stirred tank reactor equipped with special two marine impellers [19]. Acidpretreated corn stover was used as feedstock. The final accumulative solid concentration in the experiments was 25 % (w/w) DM. In this case, no clear differences between the batch shaking flask and the fed batch reactor systems appeared. To reach a maximum conversion of 80 %, the reaction had to be lasted for 288 h. As it can be seen from literature, only some technical reactor concepts have been investigated. In this report, we present a new vertical reactor concept based on a scalable stirred tank reactor system to realise a high biomass hydrolysis of lignocellulosic materials. By designing a special segmented helical stirrer (SHS), we could achieve an efficient homogenisation of the fibre suspension leading to an intensification of the hydrolysis reaction. In comparison to literature, we wanted to have glucose yields of >70 % within shorter reaction times (35 FPU/g DM at fibre concentrations of 70 % within 96 h at a DM of 8 % (w/w). As it can be seen from the literature, the investigated lignocellulosic materials could be efficiently hydrolysed at an optimised enzyme dosage giving at least 70 % yield after 48 h. Consequently, the chosen optimal enzyme dosages (30 FPU/g DM for wheat straw and 20 FPU/g DM for beech wood) represented a good compromise between high space-time yields and minimal enzyme dosage. Nevertheless, in the shaking flask reaction system, mass transport limitations occurred. Thus, shaking of the reaction system was insufficient for homogenising the slurry. For this purpose, rotating stirrers have to be employed. Therefore, we decided to use a conventional scalable stirred tank reactor with a special segmented helical stirrer. Before transferring the hydrolysis reaction into a scalable stirred tank reactor, the correlation between viscosity and fibre concentration was further investigated. Rheology of the Fibre Suspensions By increasing the fibre content, the viscosity of the suspension is influenced dramatically [18, 30]. To measure the relationship between fibre concentration and apparent viscosity rheological measurements were made. First, the viscosity respectively the shear stress in dependency of the shear rate (second) was measured at different fibre concentrations. In Fig. 3a, the viscosity curves for different wheat straw fibre slurries are exemplarily given. As it can be expected, the viscosity decreased by increasing the shear rate. Consequently, the fibre slurries showed a shear thinning behaviour, as it is known from polymeric fluids [22]. In this case, the
Appl Biochem Biotechnol (2014) 172:1699–1713
1707
Fig. 3 a Viscosity in dependency of the shear rate for different wheat straw slurries. b Viscosity in dependency of the fibre concentration for pretreated beech wood and wheat straw
long chains redirect themselves in the direction of the shear stress. This leads to a decrease of apparent viscosity. The same non-Newtonian fluid behaviour could also be observed for steam-treated barley straw [18], vinasse [31] or acid-treated corn stover [32]. In all these investigations, the suspensions showed a reduction of viscosity by increasing shear rate, indicating that Ostwald de Waele power law has to be considered for describing the rheological behaviour of the reaction fluid. The characteristic parameters for the power law are given in Table 3. The relationship between shear rate γ and agitator speed n were estimated with the relationship of Metzner and Otto [33]. γ ¼ k MO ⋅n
ð3Þ
For stirrers that work near the wall, the Metzner/Otto parameter can be approximated by 35 [34]. Consequently, the operating agitator speed of 80 to 100 rpm corresponded to shear rates in the range of 46 to 60 s−1. To measure the relationship between viscosity and fibre concentration a shear rate of 60 s−1 was applied. From Fig. 3b, it can be seen that with increasing fibre content, the viscosity of the slurries also increased. In the case of wheat straw, a significant swelling behaviour was observed. It is known that high contents of hemicelluloses lead to such a swelling behaviour [35]. Given this swelling, the apparent viscosity of the wheat straw suspension was much higher than the viscosity of the beech wood slurries. At 10 % (w/w) DM the wheat straw suspension had an apparent viscosity of 9 Pas, whereas the beech wood slurry showed a viscosity of 0.5 Pa s. By increasing the fibre concentration, the viscosity of the wheat straw suspension increased dramatically up to 19 Pa s at 17.5 % (w/w) DM. The viscosity for higher fibre concentrations (>17.5 %) could not be measured anymore because
Table 3 Characteristic parameters of the Ostwald de Waele power law for the investigated fibre suspensions at different DM content
Substrate Wheat straw
Beech wood
DM (%)
KpI
m
10
815.2
0.022
15
1,297.8
0.023
20
1,565.4
0.032
10 15
286.0 866.2
0.030 0.065
20
9,890.0
0.045
1708
Appl Biochem Biotechnol (2014) 172:1699–1713
the sample was instable because of the low water content at the fibre surface. In the case of the beech wood fibres, the viscosity only increased from 10 Pa s at 17.5 % (w/w) DM to 12.5 Pa s at 22.5 % (w/w) DM. Thus, a significant increase of viscosity could not be observed for beech wood. The experiments could prove that increasing substrate concentration will lead to a higher slurry viscosity. Higher viscosities will cause major heat and transfer problems [22]. The data suggested that the apparent viscosity can be very different for different feedstocks and pretreatment conditions. The pretreatment process and the resulting fibre composition, especially the hemicellulose content, will have a big influence of the rheological behaviour. Mixing Behaviour of the Segmented Helical Stirrer The magnitude of macroscopic mixing time in a real slurry system could be estimated was in the range of 150 to 250 s by a fluorescence tracer (see Fig. 4a). The mixing time was reduced by increasing the agitator speed. In comparison to the reaction time of the hydrolysis, the mixing time of the system was significantly shorter. Thus the reaction system can be seen as nearly homogenous during reaction. Enzyme that is dosed at the beginning of the reaction will be instantaneous distributed in the system. To our knowledge, this was the first approach to estimate the magnitude of mixing time in lignocellulosic fibre suspensions. Power Characteristics of the SHS The power consumption in a stirred vessel depends on geometric, operational and fluid specific parameter. From dimensional analysis the power characteristics can be written as [22, 23]: Ne ¼ f ðRe; D=d; H=D; h=d Þ
ð4Þ
The Newton number (Ne) and the Reynolds number (Re) for stirred vessels are traditionally defined as: Ne ¼
P ρn3 d 5
ð5Þ
Fig. 4 a Fluorescence signal in dependency of the mixing time for different agitator speeds. 15 % (w/w) beech wood slurry (700 g) was marked with 1 mL of 2 mg/mL 4-methylumbelliferone. Samples were taken at the bottom of the stirrer and analysed with fluorescence spectroscopy (λexcitation =340 nm, λemission =460 nm). b Power characteristics of the SHS in comparison with other stirring geometries taken from [36]
Appl Biochem Biotechnol (2014) 172:1699–1713
Re ¼
1709
nd 2 ρ η
ð6Þ
In the experiments, the geometric parameters were D/d = 0.9, H/D = 1.0, and h/d = 0.11. In Fig. 4b, the measured correlation between Ne and Re are given for Re = 1–105. For the comparison with other stirrer geometries, literature data from [36] are also given. As expected for Re < 100, an inverse relationship between Ne and Re could be observed. Consequently, the power input increases proportionally with n2, d3 and η. In the range Re < 40, the measured Ne numbers of the SHS are quite similar to a conventional helical stirrer. Only at Re > 40, the SHS shows higher power input in comparison to the conventional to helical stirrer. Consequently, the SHS showed even good power input characteristics at high Re numbers (>100) like it is known from Rushton turbines. Enzymatic Hydrolysis in the Stirred Tank Reactor From the shaking flask experiments, it could be stated out that optimal enzyme loading was 30 FPU/g DM for alkaline-pretreated wheat straw respectively 20 FPU/g DM for organosolvpretreated beech wood. In the case of alkaline-pretreated wheat straw, mass transport limitations occurred even at 13 % (w/w) DM. So the shaking flask system was not suitable for hydrolysing lignocellulosic material with higher DM content. By transferring the reaction into a stirred tank reactor equipped with a segmented helical stirrer (SHS), the reaction could be further intensified. The yield and the liquefaction time (=time to reach a liquid slurry) for both systems at 20 % (w/w) DM are given in Table 4. By using the stirred tank reactor system, the yield after 48 h could be increased from 30 to 76 % for alkaline wheat straw. The liquefaction time could be significantly reduced from 60 to 8 h. Thus, the mass transfer problems could be diminished. In contrast to alkaline-pretreated wheat straw, the organosolv-pretreated beech wood was efficiently hydrolysed even in the shaking flask system. The transfer from the shaking flask system to the stirred tank reactor increased the glucose yield from 70 to 82 %. The liquefaction time was reduced from 16 to 1 h, so the beech wood is rapidly liquefied. For further characterisation of the reaction system, the substrates were hydrolysed at different DM content to maximise glucose concentration of the hydrolysate. The correlation between glucose concentration and initial solid content is shown in Fig. 5. Additionally, lines indicating the theoretical glucose yield for 80 and 100 % were included into the graphs. During the hydrolysis of alkaline-pretreated wheat straw mass transfer problems occurred at 30 % (w/w) DM. In this case the glucose concentrations were smaller in comparison to experiments at 20 % (w/w) DM. Maximal glucose concentration of about 110 g/kg (76 %) after 48 h was realised at 20 % (w/w). As mentioned above, wheat straw fibres showed significant swelling due to the higher hemicellulose content. This resulted in a higher viscosity of the slurry and
Table 4 Glucose yield after 48 h and liquefaction time of the different reaction systems for alkaline-pretreated wheat straw and for organosolv-pretreated beech wood at 20 % (w/w) DM Substrate Wheat straw Beech wood
Reaction system
Yield (%)
Liquefaction time (h) 60
Shaking flask
30
Stirred tank reactor
76
8
Shaking flask
70
16
Stirred tank reactor
82
1
1710
Appl Biochem Biotechnol (2014) 172:1699–1713
Fig. 5 Effect of time and initial DM on glucose generation during the hydrolysis of a alkaline-pretreated wheat straw and b organosolv-pretreated beech wood
thus mass transport problems occurred at DM > 20 % (w/w). As already mentioned in materials and methods section, the alkaline-pretreated wheat straw fibres could only be handled in batch mode up to DM of 20 % (w/w). To adjust a DM of 30 % (w/w), the rest of the fibres had to be fed to the reactor during the first 2 h. In the case of organosolv-prereated beech wood, DM up to 30 % (w/w) could be hydrolysed in batch mode, showing no significant regime of mass transport limitation. Thus higher glucose concentration could be reached with higher DM. The glucose yield after 48 h was 80 % for 20 % (w/w) DM and 72 % for 30 % (w/w). These yields corresponded to a sugar concentration of 120 g/kg respectively 150 g/kg. As described in literature, Y = 100 % can never been reached in high biomass hydrolysis [15]. This phenomenon cannot be explained by mass transport limitation, because even with very long reaction times (>128 h) complete conversion has not been detected by several authors [14, 19, 37, 38]. Moreover, irreversible binding of the enzymes to lignin is also not the main limiting factor, because it could be observed, that 100 % conversion could also not be achieved, when using pure cellulose [15]. One possible explanation could be the lack of cellulose binding sites for CBH at higher substrate conversion. Consequently, the maximal achievable yield Ymax for t ≥ ∞ is a characteristic value for every pretreated lignocellulosic material. In our case, Ymax was 80 % for both substrates. Mass transport limitation influences only the reaction time to reach Ymax. From literature only a few scalable reactor systems were investigated for high biomass hydrolysis. Comparing the data from literature (see Table 5), our developed scalable reactor concept leaded to higher glucose yields after 48 h using high fibre concentrations up to 20 % (w/w) DM giving hydrolysates with glucose concentration > 100 g/kg. In the case of alkalinepretreated wheat straw mass transfer problems occurred due to swelling of the fibre. The regime of mass transport limitations were in agreement with observations from literature. Interestingly, organosolv-pretreated beech wood could be hydrolysed successfully even up to 30 % (w/w) DM, showing the effectiveness of the developed reactor system.
Conclusions Enzymatic hydrolysis of lignocellulosic biomass at a high solid content (>10 % (w/w) dry mass) will be the main parameter to realise economic feasible processes to produce fermentable sugars within a lignocellulose biorefinery [11–13]. However, the current process
Appl Biochem Biotechnol (2014) 172:1699–1713
1711
Table 5 Comparison of the developed stirred tank reactor system with different reactor concepts from literature according to substrate concentration (DM), reaction time and corresponding glucose yield Substrate/pretreatment Reactor concept Wheat straw
Drum reactor with paddle stirrer
Steam extrusion Maize straw
Drum reactor with paddle stirrer
DM/% Y/% (time) Reference 10
80 (96 h)
20
62 (96 h)
40
45 (96 h)
[14, 16, 17]
20
62 (72 h)
30
47 (72 h)
Stirred tank with impeller
10
55 (72 h)
[39]
Barley straw Steam treatment
Stirred tank with impeller
15
68 (48 h)
[18]
Wheat straw
Stirred tank reactor with segmented helical stirrer
Own work
Diluted acid Softwood
[37]
SO2− explosion
NaOH Aufschluss Poplar wood Organosolv
Stirred tank reactor with segmented helical stirrer
15
80 (48 h)
20
76 (48 h)
30
45 (48 h)
15
80 (48 h)
20
80 (48 h)
30
72 (48 h)
Own work
development is not really focussing on high-biomass hydrolysis and the research is conducted at fibre concentrations < 10 % (w/w) DM. In this work, we presented a new reactor concept, which is based on a scalable stirred tank reactor. As lignocellulosic materials, we used alkaline-pretreated wheat straw and organosolvpretreated beech wood. By designing a special SHS geometry, we were able to effectively homogenise the viscous fibre suspensions. This could be proven by a pulse method with a fluorescent tracer. The mixing time was in the magnitude of minutes, whereas the reaction lasted for several hours. Rheological measurements showed, that the fibre suspensions showed shear thinning behaviour, which could be described by a power law. Generally, the viscosity increased by increasing the fibre content. Alkaline-treated wheat straw showed a higher hemicelluose content (21 % (w/w)) leading to a higher viscosity of the fibre suspension (19 Pa s at a solid content of 20 % (w/w)) because of swelling of the fibre. The SHS could also be characterised according to Ne–Re correlation. In the laminar flow region (Re < 100) respectively for high viscous media, the power input was comparable to a conventional helical stirrer. The typical inverse relationship between Ne and Re could be detected. So the power input P is proportional to n2, d3 and η. In the beginning of the hydrolysis experiments, the enzyme dosage was optimised by a central composite response surface plan. With the experimental data, an optimal enzyme to substrate ratio in the range of 20–30 FPU/g DM could be identified to reach at least a glucose yield of 70 % at fibre concentrations > 12 % (w/w) DM. By using the stirred tank reactor system the glucose yields could be significantly increased to 76–82 % after 48 h in comparison to the shaking flask system at a solid content of 20 % (w/w) DM. Given the swelling of the alkaline-pretreated wheat straw fibres, the fibre suspensions showed a higher viscosity (19 Pa s at a solid content of 20 % (w/w)). Therefore some mass transport problems occurred in the stirred tank reactor system at concentration of ≥20 % (w/w) DM. Using the developed stirred tank reactor system, alkaline-pretreated wheat straw could be optimally hydrolysed at a solid content of 20 % (w/w) to give 110 g/kg glucose (76 %) after 48 h.
1712
Appl Biochem Biotechnol (2014) 172:1699–1713
Interestingly, the hydrolysis of organosolv-pretreated beech wood showed no significant mass transport limitations up to 30 % DM. So organosolv-pretreated beech wood could be efficiently hydrolysed even at 30 % (w/w) DM, giving 150 g/kg glucose (72 %). Consequently, we were able to design a new reactor system to efficiently produce lignocellulose hydrolysates with high sugar content. Acknowledgements Part of the work was financed by the “Bundesministerium für Ernährung”, Landwirtschaft und Verbraucherschutz (BMELV; FKZ: 22019309) within the project “Verbundvorhaben Lignocellulose Bioraffinerie (Phase 2)”.
References 1. Busch, R., & Hirth, T. (2006). Biotechnology Journal, 1, 770–776. 2. Wyman, C. E. (2007). Trends in Biotechnology, 25, 153–157. 3. Kamm, B., Gruber, P. R., & Kamm, M. (2007). Biorefineries—industrial processes and products. Weinheim: Wiley. 4. Fischer, G., Prieler, S. and van Velthuizen, H. (2005). Biomass and Bioenergy 28, 28, 119–132. 5. Mosier, N., & Wyman, C. (2005). Bioresource Technology, 96, 673–686. 6. Jørgensen, H., Kristensen, J. B. and Felby, C. (2007). Biofuels, Bioprod. Bioref., 1, 119–134, doi:110.1002/ bbb. 7. Lynd, L. R., & Weimer, P. J. (2002). Microbiology and Molecular Biology Reviews, 66, 506–577. 8. Walker, L. P. and Wilson, D. B. (1991). Bioresource Technology, 36(3), l–4. 9. Bayrock, D. P., & Ingledew, W. M. (2001). Journal of Industrial Microbiology & Biotechnology, 27, 87–93. 10. Thomas, K. C., & Ingledew, W. M. (1992). Journal of Industrial Microbiology, 10, 61–68. 11. Wingren, A., Galbe, M., & Zacchi, G. (2003). Trends Biotechnology Progress, 19, 1109–1117. 12. Sassner, P., & Galbe, M. (2008). Biomass and Bioenergy, 32, 422–430. 13. Zacchi, G., Skoog, K., & Hahn-Hägerdal, B. (1988). Biotechnology and Bioengineering, 32, 460–466. 14. Jørgensen, H., Vibe-Pedersen, J., Larsen, J., & Felby, C. (2007). Biotechnology and Bioengineering, 96, 862–870. 15. Kristensen, J. B., & Felby, C. (2009). Biotechnology for Biofuels, 2, 1–10. 16. Larsen, J., & Petersen, M. Ø. (2008). Chemical Engineering and Technology, 31, 765–772. 17. Felby, K., Larson, J. and Jorgensen, H. (2009) Patent US 7,598,069 B2. 18. Rosgaard, L., & Andric, P. (2007). Applied Biochemistry and Biotechnology, 143, 27–40. 19. Hodge, D. B., & Karim, M. N. (2009). Applied Biochemistry and Biotechnology, 152, 88–107. 20. Sluiter, A., Hames, B., Ruiz, R. and Scarlata, C. (2008). Determination of structural carbohydrates and lignin in biomass. Golden, CO: National Renewable Energy Laboratory 21. Xiao, Z., Storms, R., & Tsang, A. (2004). Biotechnology and Bioengineering, 88, 834–837. 22. Zlokarnik, M. (1999). Rührtechnik Theorie und Praxis. Berlin: Springer. 23. Kraume, M. (2003). Mischen und Rühren: Grundlagen und moderne Verfahren. Weinheim: Wiley. 24. Sluiter, A., B. Hames, B., Ruiz, R. and Scarlata, C. (2008). Determination of sugars, byproducts, and degradation products in liquid fraction process samples. Golden, CO: National Renewable Energy Laboratory 25. Mussatto, S. I., Dragone, G., Fernandes, M., Milagres, A. M. F., & Roberto, I. C. (2008). Cellulose, 15, 711– 721. 26. Arantes, V., & Saddler, J. N. (2011). Biotechnology for Biofuels, 4, 1–16. 27. Vlasenko, E., & Ding, H. (1997). Bioresource Technology, 59, 109–119. 28. Hodge, D. B., & Karim, M. N. (2008). Bioresource Technology, 99, 8940–8948. 29. Georgieva, T. I., & Hou, X. (2008). Applied Biochemistry and Biotechnology, 148, 35–44. 30. Lu, Y., & Wang, Y. (2010). Applied Biochemistry and Biotechnology, 160, 360–369. 31. Sørensen, H. R., Pedersen, S., & Meyer, A. S. (2006). Biotechnology Progress, 22, 505–513. 32. Viamajala, S., McMillan, J. D., Schell, D. J., & Elander, R. T. (2009). Bioresource Technology, 100, 925– 934. 33. Metzner, A. B., & Otto, R. E. (1957). AICHE Journal, 1, 3–10. 34. Todtenhaupt, E. K. and Zeiler, G. (2000). Handbuch der Rührtechnik. Schopfheim: EKATO Rühr- und Mischtechnik GmbH.
Appl Biochem Biotechnol (2014) 172:1699–1713
1713
35. Wan, J. Q., & Wang, Y. (2010). Bioresource Technology, 101, 4577–4583. 36. Zlokarnik, M. (1967). Chemie-Ing.-Techn. 39, 539–548 37. Hodge, D. B. and Karim, M. N. (2005). High-solids enzymatic saccharification of lignocellulosic biomass. In 27th Symposium on Biotechnology for Fuels and Chemicals, Denver 38. Mohagheghi, A., Tucker, M., Grohmann, K., & Wyman, C. (1992). Applied Biochemistry and Biotechnology, 33, 67–81. 39. Tengborg, C., Galbe, M., & Zacchi, G. (2001). Biotechnology Progress, 17, 110–117.
High Solids Enzymatic Hydrolysis of Pretreated Lignocellulosic Materials with a Powerful Stirrer Concept Daniel Ludwig & Buchmann Michael & Thomas Hirth & Steffen Rupp & Susanne Zibek
Received: 16 April 2013 / Accepted: 23 October 2013 / Published online: 19 November 2013 # Springer Science+Business Media New York 2013
Abstract In this study, we present a powerful stirred tank reactor system that can efficiently hydrolyse lignocellulosic material at high solid content to produce hydrolysates with glucose concentration > 100 g/kg. As lignocellulosic substrates alkaline-pretreated wheat straw and organosolv-pretreated beech wood were used. The developed vertical reactor was equipped with a segmented helical stirrer, which was specially designed for high biomass hydrolysis. The stirrer was characterised according to mixing behaviour and power input. To minimise the cellulase dosage, a response surface plan was used. With the empirical relationship between glucose yield, cellulase loading and solid content, the minimal cellulase dosage was calculated to reach at least 70 % yield at high glucose and high substrate concentrations within 48 h. The optimisation resulted in a minimal enzyme dosage of 30 FPU/g dry matter (DM) for the hydrolysis of wheat straw and 20 FPU/g DM for the hydrolysis of beech wood. By transferring the hydrolysis reaction from shaking flasks to the stirred tank reactor, the glucose yields could be increased. Using the developed stirred tank reactor system, alkaline-pretreated wheat straw could be converted to 110 g/kg glucose (76 %) at a solid content of 20 % (w/w) after 48 h. Organosolv-pretreated beech wood could be efficiently hydrolysed even at 30 % (w/w) DM, giving 150 g/kg glucose (72 %). Keywords Lignocellulose biomass . High solid content . Enzymatic hydrolysis . Stirred tank reactor D. Ludwig Evonik Industries AG, Rodenbacher Chaussee 4, 63457 Hanau-Wolfgang, Germany T. Hirth Institute of Interfacial Process Engineering and Plasma Technology IGVP, University of Stuttgart, Nobelstraße 12, 70569 Stuttgart, Germany T. Hirth : S. Rupp : S. Zibek (*) Fraunhofer-Institute for Interfacial Engineering and Biotechnology IGB, Nobelstraße 12, 70569 Stuttgart, Germany e-mail: [email protected] B. Michael Linde Engineering Dresden GmbH, Bodenbacher Straße 80, 01277 Dresden, Germany
1700
Appl Biochem Biotechnol (2014) 172:1699–1713
Introduction Lignocellulosic biomass is one of the most promising renewable resources that can be converted to bio-based chemicals, biofuels and bioenergy [1]. Especially its high carbohydrate content of about 60–75 % provides a huge potential to produce fermentable sugars without competing directly with food and feed industry [2]. Different types of resources are possible as feedstocks [3]. For example, waste products from forestry or agricultural industry like crop straw or special biomass plants like miscanthus or switchgrass can be used. The latter two can be grown on marginal land at high biomass production rates [4]. There is also the possibility to use softwoods and hardwoods. In contrast to sugar- and starch-based feedstocks like sugar beet or corn, lignocellulosic biomass has to be pretreated by physical or chemical methods to increase the cellulose conversion of the following hydrolysis process [5]. Cellulose hydrolysis is catalysed by a cellulase complex, which includes three main classes of glucoside hydrolases, namely, cellobiohydrolases (CBH), endoglucosidases (EG) and beta-glucosidases (BG) [6, 7]. CBH catalyses the hydrolysis from both polymer ends (reducing and nonreducing) producing cellobiose, whereras EG cleaves internal glucosidic bonds producing smaller polymers, and hence, new adsorption sites for the CBH. Finally, BG catalyses the hydrolysis of cellobiose to glucose, and hence, product inhibition of CBH and EG is reduced. Considering the different interfering reaction steps, the cellulase system shows a high degree of synergism [8]. From starch-based processes, it is known that the dry matter (DM) content should exceed 30 % (w/w) to realise an economic feasible conversion process [9, 10]. Therefore, the solid content should also be increased to >10 % (w/w) DM to realise an economic feasible process for the hydrolysis of lignocellulosic biomass [11–13]. By increasing the substrate concentration, the investment and production costs are decreased because smaller reactors and less heating or cooling energy are needed for the same sugar productivity. With higher substrate concentrations, the space-time yields are significantly enhanced and the high concentrated sugar hydrolysates will lead to more economic downstreaming processes. However, some disadvantages occur, when dealing with high biomass hydrolysis [14]. In general, the viscosity of the fibre suspension is increased dramatically. Consequently, mass and heat transfer problems will have a major drawback on the hydrolysis process. Usually, the sugar yield is decreasing with increasing solid content. Kristensen et al. summarised different high biomass hydrolysis experiments and could show that this tendency is valid for different feedstocks as well as for separate (SHF) and simultaneous saccharification fermentation (SSF) processes [15]. As a consequence, the reactor design will play an important role by realising an effective hydrolysis process. Besides some shaking flask or roller bottle experiments at a millilitre scale, only a few reactor concepts are investigated in literature, which can be applied as an industrial reactor concept. These reactor types can be generally divided into horizontal and vertical reactors. Jorgensen et al. introduced a horizontal five-chamber liquefaction reactor [14, 16, 17]. Each chamber was homogenised by three paddlers that were mounted on a horizontal rotating shaft. By using this reactor concept, the achievable glucose yields ranged from 60 at 20 % (w/w) DM to 35 at 40 % (w/w) DM after 96 h. It has to be mentioned that the yields could not be significantly increased when switching to a SSF process. Rosgaard et al. conducted batch and fed batch hydrolysis processes in a 2-L bottle (vertical system) with special six custommade impellers [18]. Steam-treated barley straw was used as substrate. They investigated the influence of different solid concentration up to 15 % (w/w) DM on glucose yield during batch and fed batch hydrolysis reactions. It could be pointed out, that no clear difference occurred between fed batch and batch processes up to 15 % (w/w) DM. The glucose yield was 80 % after 72 h. Hodge et al. compared a batch shaking flask experiment with fed batch experiments
Appl Biochem Biotechnol (2014) 172:1699–1713
1701
in a 7-L vertical stirred tank reactor equipped with special two marine impellers [19]. Acidpretreated corn stover was used as feedstock. The final accumulative solid concentration in the experiments was 25 % (w/w) DM. In this case, no clear differences between the batch shaking flask and the fed batch reactor systems appeared. To reach a maximum conversion of 80 %, the reaction had to be lasted for 288 h. As it can be seen from literature, only some technical reactor concepts have been investigated. In this report, we present a new vertical reactor concept based on a scalable stirred tank reactor system to realise a high biomass hydrolysis of lignocellulosic materials. By designing a special segmented helical stirrer (SHS), we could achieve an efficient homogenisation of the fibre suspension leading to an intensification of the hydrolysis reaction. In comparison to literature, we wanted to have glucose yields of >70 % within shorter reaction times (35 FPU/g DM at fibre concentrations of 70 % within 96 h at a DM of 8 % (w/w). As it can be seen from the literature, the investigated lignocellulosic materials could be efficiently hydrolysed at an optimised enzyme dosage giving at least 70 % yield after 48 h. Consequently, the chosen optimal enzyme dosages (30 FPU/g DM for wheat straw and 20 FPU/g DM for beech wood) represented a good compromise between high space-time yields and minimal enzyme dosage. Nevertheless, in the shaking flask reaction system, mass transport limitations occurred. Thus, shaking of the reaction system was insufficient for homogenising the slurry. For this purpose, rotating stirrers have to be employed. Therefore, we decided to use a conventional scalable stirred tank reactor with a special segmented helical stirrer. Before transferring the hydrolysis reaction into a scalable stirred tank reactor, the correlation between viscosity and fibre concentration was further investigated. Rheology of the Fibre Suspensions By increasing the fibre content, the viscosity of the suspension is influenced dramatically [18, 30]. To measure the relationship between fibre concentration and apparent viscosity rheological measurements were made. First, the viscosity respectively the shear stress in dependency of the shear rate (second) was measured at different fibre concentrations. In Fig. 3a, the viscosity curves for different wheat straw fibre slurries are exemplarily given. As it can be expected, the viscosity decreased by increasing the shear rate. Consequently, the fibre slurries showed a shear thinning behaviour, as it is known from polymeric fluids [22]. In this case, the
Appl Biochem Biotechnol (2014) 172:1699–1713
1707
Fig. 3 a Viscosity in dependency of the shear rate for different wheat straw slurries. b Viscosity in dependency of the fibre concentration for pretreated beech wood and wheat straw
long chains redirect themselves in the direction of the shear stress. This leads to a decrease of apparent viscosity. The same non-Newtonian fluid behaviour could also be observed for steam-treated barley straw [18], vinasse [31] or acid-treated corn stover [32]. In all these investigations, the suspensions showed a reduction of viscosity by increasing shear rate, indicating that Ostwald de Waele power law has to be considered for describing the rheological behaviour of the reaction fluid. The characteristic parameters for the power law are given in Table 3. The relationship between shear rate γ and agitator speed n were estimated with the relationship of Metzner and Otto [33]. γ ¼ k MO ⋅n
ð3Þ
For stirrers that work near the wall, the Metzner/Otto parameter can be approximated by 35 [34]. Consequently, the operating agitator speed of 80 to 100 rpm corresponded to shear rates in the range of 46 to 60 s−1. To measure the relationship between viscosity and fibre concentration a shear rate of 60 s−1 was applied. From Fig. 3b, it can be seen that with increasing fibre content, the viscosity of the slurries also increased. In the case of wheat straw, a significant swelling behaviour was observed. It is known that high contents of hemicelluloses lead to such a swelling behaviour [35]. Given this swelling, the apparent viscosity of the wheat straw suspension was much higher than the viscosity of the beech wood slurries. At 10 % (w/w) DM the wheat straw suspension had an apparent viscosity of 9 Pas, whereas the beech wood slurry showed a viscosity of 0.5 Pa s. By increasing the fibre concentration, the viscosity of the wheat straw suspension increased dramatically up to 19 Pa s at 17.5 % (w/w) DM. The viscosity for higher fibre concentrations (>17.5 %) could not be measured anymore because
Table 3 Characteristic parameters of the Ostwald de Waele power law for the investigated fibre suspensions at different DM content
Substrate Wheat straw
Beech wood
DM (%)
KpI
m
10
815.2
0.022
15
1,297.8
0.023
20
1,565.4
0.032
10 15
286.0 866.2
0.030 0.065
20
9,890.0
0.045
1708
Appl Biochem Biotechnol (2014) 172:1699–1713
the sample was instable because of the low water content at the fibre surface. In the case of the beech wood fibres, the viscosity only increased from 10 Pa s at 17.5 % (w/w) DM to 12.5 Pa s at 22.5 % (w/w) DM. Thus, a significant increase of viscosity could not be observed for beech wood. The experiments could prove that increasing substrate concentration will lead to a higher slurry viscosity. Higher viscosities will cause major heat and transfer problems [22]. The data suggested that the apparent viscosity can be very different for different feedstocks and pretreatment conditions. The pretreatment process and the resulting fibre composition, especially the hemicellulose content, will have a big influence of the rheological behaviour. Mixing Behaviour of the Segmented Helical Stirrer The magnitude of macroscopic mixing time in a real slurry system could be estimated was in the range of 150 to 250 s by a fluorescence tracer (see Fig. 4a). The mixing time was reduced by increasing the agitator speed. In comparison to the reaction time of the hydrolysis, the mixing time of the system was significantly shorter. Thus the reaction system can be seen as nearly homogenous during reaction. Enzyme that is dosed at the beginning of the reaction will be instantaneous distributed in the system. To our knowledge, this was the first approach to estimate the magnitude of mixing time in lignocellulosic fibre suspensions. Power Characteristics of the SHS The power consumption in a stirred vessel depends on geometric, operational and fluid specific parameter. From dimensional analysis the power characteristics can be written as [22, 23]: Ne ¼ f ðRe; D=d; H=D; h=d Þ
ð4Þ
The Newton number (Ne) and the Reynolds number (Re) for stirred vessels are traditionally defined as: Ne ¼
P ρn3 d 5
ð5Þ
Fig. 4 a Fluorescence signal in dependency of the mixing time for different agitator speeds. 15 % (w/w) beech wood slurry (700 g) was marked with 1 mL of 2 mg/mL 4-methylumbelliferone. Samples were taken at the bottom of the stirrer and analysed with fluorescence spectroscopy (λexcitation =340 nm, λemission =460 nm). b Power characteristics of the SHS in comparison with other stirring geometries taken from [36]
Appl Biochem Biotechnol (2014) 172:1699–1713
Re ¼
1709
nd 2 ρ η
ð6Þ
In the experiments, the geometric parameters were D/d = 0.9, H/D = 1.0, and h/d = 0.11. In Fig. 4b, the measured correlation between Ne and Re are given for Re = 1–105. For the comparison with other stirrer geometries, literature data from [36] are also given. As expected for Re < 100, an inverse relationship between Ne and Re could be observed. Consequently, the power input increases proportionally with n2, d3 and η. In the range Re < 40, the measured Ne numbers of the SHS are quite similar to a conventional helical stirrer. Only at Re > 40, the SHS shows higher power input in comparison to the conventional to helical stirrer. Consequently, the SHS showed even good power input characteristics at high Re numbers (>100) like it is known from Rushton turbines. Enzymatic Hydrolysis in the Stirred Tank Reactor From the shaking flask experiments, it could be stated out that optimal enzyme loading was 30 FPU/g DM for alkaline-pretreated wheat straw respectively 20 FPU/g DM for organosolvpretreated beech wood. In the case of alkaline-pretreated wheat straw, mass transport limitations occurred even at 13 % (w/w) DM. So the shaking flask system was not suitable for hydrolysing lignocellulosic material with higher DM content. By transferring the reaction into a stirred tank reactor equipped with a segmented helical stirrer (SHS), the reaction could be further intensified. The yield and the liquefaction time (=time to reach a liquid slurry) for both systems at 20 % (w/w) DM are given in Table 4. By using the stirred tank reactor system, the yield after 48 h could be increased from 30 to 76 % for alkaline wheat straw. The liquefaction time could be significantly reduced from 60 to 8 h. Thus, the mass transfer problems could be diminished. In contrast to alkaline-pretreated wheat straw, the organosolv-pretreated beech wood was efficiently hydrolysed even in the shaking flask system. The transfer from the shaking flask system to the stirred tank reactor increased the glucose yield from 70 to 82 %. The liquefaction time was reduced from 16 to 1 h, so the beech wood is rapidly liquefied. For further characterisation of the reaction system, the substrates were hydrolysed at different DM content to maximise glucose concentration of the hydrolysate. The correlation between glucose concentration and initial solid content is shown in Fig. 5. Additionally, lines indicating the theoretical glucose yield for 80 and 100 % were included into the graphs. During the hydrolysis of alkaline-pretreated wheat straw mass transfer problems occurred at 30 % (w/w) DM. In this case the glucose concentrations were smaller in comparison to experiments at 20 % (w/w) DM. Maximal glucose concentration of about 110 g/kg (76 %) after 48 h was realised at 20 % (w/w). As mentioned above, wheat straw fibres showed significant swelling due to the higher hemicellulose content. This resulted in a higher viscosity of the slurry and
Table 4 Glucose yield after 48 h and liquefaction time of the different reaction systems for alkaline-pretreated wheat straw and for organosolv-pretreated beech wood at 20 % (w/w) DM Substrate Wheat straw Beech wood
Reaction system
Yield (%)
Liquefaction time (h) 60
Shaking flask
30
Stirred tank reactor
76
8
Shaking flask
70
16
Stirred tank reactor
82
1
1710
Appl Biochem Biotechnol (2014) 172:1699–1713
Fig. 5 Effect of time and initial DM on glucose generation during the hydrolysis of a alkaline-pretreated wheat straw and b organosolv-pretreated beech wood
thus mass transport problems occurred at DM > 20 % (w/w). As already mentioned in materials and methods section, the alkaline-pretreated wheat straw fibres could only be handled in batch mode up to DM of 20 % (w/w). To adjust a DM of 30 % (w/w), the rest of the fibres had to be fed to the reactor during the first 2 h. In the case of organosolv-prereated beech wood, DM up to 30 % (w/w) could be hydrolysed in batch mode, showing no significant regime of mass transport limitation. Thus higher glucose concentration could be reached with higher DM. The glucose yield after 48 h was 80 % for 20 % (w/w) DM and 72 % for 30 % (w/w). These yields corresponded to a sugar concentration of 120 g/kg respectively 150 g/kg. As described in literature, Y = 100 % can never been reached in high biomass hydrolysis [15]. This phenomenon cannot be explained by mass transport limitation, because even with very long reaction times (>128 h) complete conversion has not been detected by several authors [14, 19, 37, 38]. Moreover, irreversible binding of the enzymes to lignin is also not the main limiting factor, because it could be observed, that 100 % conversion could also not be achieved, when using pure cellulose [15]. One possible explanation could be the lack of cellulose binding sites for CBH at higher substrate conversion. Consequently, the maximal achievable yield Ymax for t ≥ ∞ is a characteristic value for every pretreated lignocellulosic material. In our case, Ymax was 80 % for both substrates. Mass transport limitation influences only the reaction time to reach Ymax. From literature only a few scalable reactor systems were investigated for high biomass hydrolysis. Comparing the data from literature (see Table 5), our developed scalable reactor concept leaded to higher glucose yields after 48 h using high fibre concentrations up to 20 % (w/w) DM giving hydrolysates with glucose concentration > 100 g/kg. In the case of alkalinepretreated wheat straw mass transfer problems occurred due to swelling of the fibre. The regime of mass transport limitations were in agreement with observations from literature. Interestingly, organosolv-pretreated beech wood could be hydrolysed successfully even up to 30 % (w/w) DM, showing the effectiveness of the developed reactor system.
Conclusions Enzymatic hydrolysis of lignocellulosic biomass at a high solid content (>10 % (w/w) dry mass) will be the main parameter to realise economic feasible processes to produce fermentable sugars within a lignocellulose biorefinery [11–13]. However, the current process
Appl Biochem Biotechnol (2014) 172:1699–1713
1711
Table 5 Comparison of the developed stirred tank reactor system with different reactor concepts from literature according to substrate concentration (DM), reaction time and corresponding glucose yield Substrate/pretreatment Reactor concept Wheat straw
Drum reactor with paddle stirrer
Steam extrusion Maize straw
Drum reactor with paddle stirrer
DM/% Y/% (time) Reference 10
80 (96 h)
20
62 (96 h)
40
45 (96 h)
[14, 16, 17]
20
62 (72 h)
30
47 (72 h)
Stirred tank with impeller
10
55 (72 h)
[39]
Barley straw Steam treatment
Stirred tank with impeller
15
68 (48 h)
[18]
Wheat straw
Stirred tank reactor with segmented helical stirrer
Own work
Diluted acid Softwood
[37]
SO2− explosion
NaOH Aufschluss Poplar wood Organosolv
Stirred tank reactor with segmented helical stirrer
15
80 (48 h)
20
76 (48 h)
30
45 (48 h)
15
80 (48 h)
20
80 (48 h)
30
72 (48 h)
Own work
development is not really focussing on high-biomass hydrolysis and the research is conducted at fibre concentrations < 10 % (w/w) DM. In this work, we presented a new reactor concept, which is based on a scalable stirred tank reactor. As lignocellulosic materials, we used alkaline-pretreated wheat straw and organosolvpretreated beech wood. By designing a special SHS geometry, we were able to effectively homogenise the viscous fibre suspensions. This could be proven by a pulse method with a fluorescent tracer. The mixing time was in the magnitude of minutes, whereas the reaction lasted for several hours. Rheological measurements showed, that the fibre suspensions showed shear thinning behaviour, which could be described by a power law. Generally, the viscosity increased by increasing the fibre content. Alkaline-treated wheat straw showed a higher hemicelluose content (21 % (w/w)) leading to a higher viscosity of the fibre suspension (19 Pa s at a solid content of 20 % (w/w)) because of swelling of the fibre. The SHS could also be characterised according to Ne–Re correlation. In the laminar flow region (Re < 100) respectively for high viscous media, the power input was comparable to a conventional helical stirrer. The typical inverse relationship between Ne and Re could be detected. So the power input P is proportional to n2, d3 and η. In the beginning of the hydrolysis experiments, the enzyme dosage was optimised by a central composite response surface plan. With the experimental data, an optimal enzyme to substrate ratio in the range of 20–30 FPU/g DM could be identified to reach at least a glucose yield of 70 % at fibre concentrations > 12 % (w/w) DM. By using the stirred tank reactor system the glucose yields could be significantly increased to 76–82 % after 48 h in comparison to the shaking flask system at a solid content of 20 % (w/w) DM. Given the swelling of the alkaline-pretreated wheat straw fibres, the fibre suspensions showed a higher viscosity (19 Pa s at a solid content of 20 % (w/w)). Therefore some mass transport problems occurred in the stirred tank reactor system at concentration of ≥20 % (w/w) DM. Using the developed stirred tank reactor system, alkaline-pretreated wheat straw could be optimally hydrolysed at a solid content of 20 % (w/w) to give 110 g/kg glucose (76 %) after 48 h.
1712
Appl Biochem Biotechnol (2014) 172:1699–1713
Interestingly, the hydrolysis of organosolv-pretreated beech wood showed no significant mass transport limitations up to 30 % DM. So organosolv-pretreated beech wood could be efficiently hydrolysed even at 30 % (w/w) DM, giving 150 g/kg glucose (72 %). Consequently, we were able to design a new reactor system to efficiently produce lignocellulose hydrolysates with high sugar content. Acknowledgements Part of the work was financed by the “Bundesministerium für Ernährung”, Landwirtschaft und Verbraucherschutz (BMELV; FKZ: 22019309) within the project “Verbundvorhaben Lignocellulose Bioraffinerie (Phase 2)”.
References 1. Busch, R., & Hirth, T. (2006). Biotechnology Journal, 1, 770–776. 2. Wyman, C. E. (2007). Trends in Biotechnology, 25, 153–157. 3. Kamm, B., Gruber, P. R., & Kamm, M. (2007). Biorefineries—industrial processes and products. Weinheim: Wiley. 4. Fischer, G., Prieler, S. and van Velthuizen, H. (2005). Biomass and Bioenergy 28, 28, 119–132. 5. Mosier, N., & Wyman, C. (2005). Bioresource Technology, 96, 673–686. 6. Jørgensen, H., Kristensen, J. B. and Felby, C. (2007). Biofuels, Bioprod. Bioref., 1, 119–134, doi:110.1002/ bbb. 7. Lynd, L. R., & Weimer, P. J. (2002). Microbiology and Molecular Biology Reviews, 66, 506–577. 8. Walker, L. P. and Wilson, D. B. (1991). Bioresource Technology, 36(3), l–4. 9. Bayrock, D. P., & Ingledew, W. M. (2001). Journal of Industrial Microbiology & Biotechnology, 27, 87–93. 10. Thomas, K. C., & Ingledew, W. M. (1992). Journal of Industrial Microbiology, 10, 61–68. 11. Wingren, A., Galbe, M., & Zacchi, G. (2003). Trends Biotechnology Progress, 19, 1109–1117. 12. Sassner, P., & Galbe, M. (2008). Biomass and Bioenergy, 32, 422–430. 13. Zacchi, G., Skoog, K., & Hahn-Hägerdal, B. (1988). Biotechnology and Bioengineering, 32, 460–466. 14. Jørgensen, H., Vibe-Pedersen, J., Larsen, J., & Felby, C. (2007). Biotechnology and Bioengineering, 96, 862–870. 15. Kristensen, J. B., & Felby, C. (2009). Biotechnology for Biofuels, 2, 1–10. 16. Larsen, J., & Petersen, M. Ø. (2008). Chemical Engineering and Technology, 31, 765–772. 17. Felby, K., Larson, J. and Jorgensen, H. (2009) Patent US 7,598,069 B2. 18. Rosgaard, L., & Andric, P. (2007). Applied Biochemistry and Biotechnology, 143, 27–40. 19. Hodge, D. B., & Karim, M. N. (2009). Applied Biochemistry and Biotechnology, 152, 88–107. 20. Sluiter, A., Hames, B., Ruiz, R. and Scarlata, C. (2008). Determination of structural carbohydrates and lignin in biomass. Golden, CO: National Renewable Energy Laboratory 21. Xiao, Z., Storms, R., & Tsang, A. (2004). Biotechnology and Bioengineering, 88, 834–837. 22. Zlokarnik, M. (1999). Rührtechnik Theorie und Praxis. Berlin: Springer. 23. Kraume, M. (2003). Mischen und Rühren: Grundlagen und moderne Verfahren. Weinheim: Wiley. 24. Sluiter, A., B. Hames, B., Ruiz, R. and Scarlata, C. (2008). Determination of sugars, byproducts, and degradation products in liquid fraction process samples. Golden, CO: National Renewable Energy Laboratory 25. Mussatto, S. I., Dragone, G., Fernandes, M., Milagres, A. M. F., & Roberto, I. C. (2008). Cellulose, 15, 711– 721. 26. Arantes, V., & Saddler, J. N. (2011). Biotechnology for Biofuels, 4, 1–16. 27. Vlasenko, E., & Ding, H. (1997). Bioresource Technology, 59, 109–119. 28. Hodge, D. B., & Karim, M. N. (2008). Bioresource Technology, 99, 8940–8948. 29. Georgieva, T. I., & Hou, X. (2008). Applied Biochemistry and Biotechnology, 148, 35–44. 30. Lu, Y., & Wang, Y. (2010). Applied Biochemistry and Biotechnology, 160, 360–369. 31. Sørensen, H. R., Pedersen, S., & Meyer, A. S. (2006). Biotechnology Progress, 22, 505–513. 32. Viamajala, S., McMillan, J. D., Schell, D. J., & Elander, R. T. (2009). Bioresource Technology, 100, 925– 934. 33. Metzner, A. B., & Otto, R. E. (1957). AICHE Journal, 1, 3–10. 34. Todtenhaupt, E. K. and Zeiler, G. (2000). Handbuch der Rührtechnik. Schopfheim: EKATO Rühr- und Mischtechnik GmbH.
Appl Biochem Biotechnol (2014) 172:1699–1713
1713
35. Wan, J. Q., & Wang, Y. (2010). Bioresource Technology, 101, 4577–4583. 36. Zlokarnik, M. (1967). Chemie-Ing.-Techn. 39, 539–548 37. Hodge, D. B. and Karim, M. N. (2005). High-solids enzymatic saccharification of lignocellulosic biomass. In 27th Symposium on Biotechnology for Fuels and Chemicals, Denver 38. Mohagheghi, A., Tucker, M., Grohmann, K., & Wyman, C. (1992). Applied Biochemistry and Biotechnology, 33, 67–81. 39. Tengborg, C., Galbe, M., & Zacchi, G. (2001). Biotechnology Progress, 17, 110–117.
