Butanol production from lignocellulose by simultaneous fermentation, saccharification, and pervaporation or vacuum evaporation

Accepted Manuscript Butanol production from lignocellulose by simultaneous fermentation, saccharification, and pervaporation or vacuum evaporation Víctor Hugo Grisales Díaz, Gerard Olivar Tost PII: DOI: Reference:

S0960-8524(16)30908-7 http://dx.doi.org/10.1016/j.biortech.2016.06.091 BITE 16713

To appear in:

Bioresource Technology

Received Date: Revised Date: Accepted Date:

22 April 2016 22 June 2016 23 June 2016

Please cite this article as: Díaz, V.H.G., Tost, G.O., Butanol production from lignocellulose by simultaneous fermentation, saccharification, and pervaporation or vacuum evaporation, Bioresource Technology (2016), doi: http://dx.doi.org/10.1016/j.biortech.2016.06.091

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Butanol production from lignocellulose by simultaneous fermentation, saccharification, and pervaporation or vacuum evaporation Víctor Hugo Grisales Díaz* and Gerard Olivar Tosta a

Perception and Intelligent Control, Department of Electrical and Electronics Engineering

and Computer Science, Universidad Nacional de Colombia – Sede Manizales, Cra. 27 No. 64-60, Manizales, Colombia. *

Correspondence should be addressed to Victor Hugo Grisales Diaz:

[email protected] Abstract Techno-economic study of acetone, butanol and ethanol (ABE) fermentation from lignocellulose was performed. Simultaneous saccharification, fermentation and vacuum evaporation (SFS-V) or pervaporation (SFS-P) were proposed. A kinetic model of metabolic pathways for ABE fermentation with the effect of phenolics and furans in the growth was proposed based on published laboratory results. The processes were optimized in Matlab®. The end ABE purification was carried out by heat-integrated distillation. The objective function of the minimization was the total annualized cost (TAC). Fuel consumption of SFS-P using poly[1-(trimethylsilyl)-1-propyne] membrane was between 13.8-19.6% lower than SFS-V. Recovery of furans and phenolics for the hybrid reactors was difficult for its high boiling point. TAC of SFS-P was increased 1.9 times with supplementation of phenolics and furans to 3 gL-1 each one for its high toxicity. Therefore, an additional detoxification method or an efficient pretreatment process will be necessary. Keywords: Simulation process; membrane separation; heat-pump; intensification; kinetic model 1

Abbreviations Ji, flux nF, Kinetic exponential parameter of furans inhibition OA, oleyl alcohol Per, permeability PDMS, polydimethylsiloxane. PP, polypropylene. Psati, vapor pressure of compound i PTMSP, poly[1-(trimethylsilyl)-1-propyne] SFS-V, simultaneous saccharification, fermentation and vacuum evaporation SFS-P, simultaneous saccharification, fermentation, and pervaporation TAC, total annualized cost TIAC, total investment annualized cost TOA, trioctylamine. TOAC, total operational annualized cost tri, time of return on investment to, annual operational time xi, molar fraction of compound i wA, mass fraction of acetone wB, mass fraction of butanol wE, mass fraction of ethanol wW, mass fraction of water wH2,, mass fraction of hydrogen 2

wCO2, ,mass fraction of carbon dioxide [Phenolic], phenolic compounds concentration into the reactor [Furans], furans concentration into the ractor [Furmax], Maximum inhibitory concentration of furans [Phemax], Maximum inhibitory concentration of phenolic compounds [LimFurans], Limit concentration of furans non-inhibitory to fermentation γi, activity coefficient of compound i 1. Introduction Biobutanol produced by acetone, butanol and ethanol (ABE) fermentation is considered as a potential biofuel (Green, 2011). ABE fermentation is usually carried out by several mesophilic Clostridium bacteria, such as C. acetobutylicum, C. beijerinckii, C. saccharobutylicum and C. saccharoperbutylicum (Green, 2011). The main advantage of Clostridium strains is the ability to consume a wide variety of substrates such as glucose, sucrose, lactose, xylose, starch, or glycerol. Another type of substrate is lignocellulose, which is between 3 to 5 times more economical than traditional substrates (Qureshi et al., 2012). Due to this and to environmental concerns, biobutanol fermentation has been studied from several lignocellulosic substrates (Qureshi et al., 2012). When producing biofuel from lignocellulose, it is necessary to perform a pretreatment stage. The pretreatment converts the hemicellulose to xylose and decrease the cellulose crystallinity. The most common pretreatment methods are dilute acid, steam explosion and liquid hot water (Chandel et al., 2012). In these pretreatment methods, phenolic compounds, furans, and organic acid are produced. The ratio and concentration of these inhibitors depend on the selected pretreatment, biomass concentration and 3

pretreatment conditions. For example, furans and phenolic concentration for steam explosion and eucalyptus as raw material were of 0.6 and 1.3 g/l, and 1.4 and 2.6 g/l for a biomass composition of 0.067 and 0.25 w/w, respectively (Zheng et al., 2015). In enzymatic hydrolysis, glucose and xylose cause a strong inhibition at high-solids loading. Simultaneous fermentation and saccharification systems have been proposed to decrease the product inhibition (Qureshi et al., 2014). Integrated reactors with separation units decrease the inhibition, increasing the ABE productivity in more than 2 times (Mariano et al., 2012; Qureshi et al., 2014) with energy requirement between 6.3-17 MJ/kgABE (Qureshi et al., 2005). Gas stripping, pervaporation, liquid-liquid extraction, pertraction, adsorption and flash evaporation have been proposed for the purification process (Qureshi et al., 2005). These separation units are not 100% selective. Therefore, ABE is conventionally purified by distillation. The solids from vinasses can be concentrated after or before ABE purification by centrifugation, evaporation or filtration. Finally, concentrated solids are combusted for electricity and steam production or used as cattle feed. In this paper, a hybrid reactor with simultaneous fermentation, saccharification, and vacuum evaporation (SFS-V) or pervaporation (SFS-P) was simulated rigorously. The hybrid reactors were proposed to reduce the negative effect of butanol and monosaccharides in the fermentation and the saccharification, respectively. In this work, a kinetic model for ABE fermentation with the effect of inhibitors, furans and phenolic compounds, was proposed. The operation conditions were selected by minimization of the total annualized costs (TAC). Although several economic studies have been performed (Abdi et al., 2016) for integrated reactors, an economic optimization of SFS-P or SFS4

V including a rigorous kinetic model and reactor model has been not studied in literature. The pretreatment conditions were not investigated in this work. For this reason, simulations were performed for several inhibitor concentrations. In addition, energy and economic comparison of hybrid reactors and batch process were performed. The recovery was achieved through a heat-integrated distillation process. 2. Process model Simulation of simultaneous fermentation, saccharification, and recovery were performed in Matlab 2015a®. The kinetic model of metabolic pathways of glucose and xylose consumption by Clostridium saccharoperbutylacetonicum N1-4 proposed by Shinto et al., 2008 and Shinto et al., 2007 was modified to reduce the error prediction in butanol and acetone production (supplementary info). The effect of inhibitors from pretreatment on biocatalyst was reported by Zheng et al., 2015 at laboratory level. This effect was performed at the same temperature, biocatalyst, and pH that fermentations with glucose and xylose as substrate (Shinto et al., 2008). However, nutrient supplementation was not performed by Zheng et al., 2015. A maximum inhibition parameter on growth was proposed for each furans and phenolic compounds. The Matlab functions used to carried out the parametric minimization was ‘ga’ and ‘fminsearch’. The parameters of kinetic model were adjusted in two stages. In the first, the parameters of kinetic model, without supplementation of inhibitors from pretreatment, were adjusted to the experimental data reported by Shinto et al., 2008 and Shinto et al., 2007. In the second, the exponential parameters of furans and phenolic inhibition were adjusted to those reported by Zheng et al., 2015. An additional parameter for furans inhibition [LimFurans] with respect to phenolic compound 5

inhibition was proposed because furans at low concentrations (between 1 to 2 g/l) did not cause inhibition in growth of Clostridium strains (Zhang and Ezeji, 2014):  [ Phenolic ]   [ Furans ]  In = 1 −  1 −  [Phemax ]   [ Furmax]  

nF

(1)

If [Furans]<[LimFurans], nF=0 The saccharification kinetic model was proposed by Kadam et al., 2004. In this work, the term substrate refers to the sum of cellulose and xylose. Fed ratio of cellulose/lignin/xylose was fixed to 2/1.5/1 in nominal conditions. The reactor model can be observed in Supplimental Material. SFS-V and SFS-P were simulated with in situ recovery. The fermentation time was 500 h. Reactor started as a batch (30 g-cellulose/l). The bleeding, the continuous feed, and butanol separation were initiated after of 25 h of operation. This operational condition for SFS-P and SFS-V reduces the concentration of solids into the reactor because the fermentation starts at a low concentration of cellulose. While the cellulose feed to reactor (high solid load) is consumed, the cellulose concentration into the reactor reduces with respect to feed. The effect of furan and phenolic compounds in the economic optimization was studied by feeding furfural and vanillin into the reactor. In this paper, inhibitors from pretreatment refers to the sum of the concentrations of furans and phenolic compounds in the feed. In the optimization, the maximum substrate concentration in continuous feed was 180 g/l. However, additional minimizations to substrate concentrations higher than 180 g/l were performed. In the simulations with supplementation of inhibitors, it was supposed that the substrate concentration from pretreatment was 180 g/l. Therefore, if the substrate concentration was lower than 180 g/l, substrate, lignin and inhibitors from

6

hydrolyzed were diluted to reduce the concentration of inhibitors from pretreatment. Water dilution cost was disregarded. The parameters of UNIQUAC model were taken from Aspen Plus V7.3®. Only butanolwater binary parameters were taken from Fischer and Gmehling, 1994. The isentropic effectiveness of the compressor was 0.75. Compression system was performed with a maximum of four stages. Pressure drop in condensation was fixed to 0.03 bar. Heat reaction was assumed null. The energetic evaluation was calculated assuming an efficiency in steam and electricity production of 0.9 and 0.33, respectively. As an example, if the steam and the electricity requirement were 0.9 and 0.33 MJ/kg-ABE, respectively, the fuel consumption was 2.0 MJ-fuel/kg-ABE. The simulations were performed with a minimum approach temperature of 10 ºC. The working volume of each reactor was fixed to 80%. The maximum volume of each reactor was 1000 m3. The overall heat transfer coefficient were 284, 568 and 852 W/s/m2 for partial condensation in CO2-liquid evaporation, liquid heat-liquid heat, and vapor condensation-liquid evaporation, respectively. The minimizations were performed with Matlab 2015a® functions ‘ga’ and ‘fminsearch’. The objective function in the optimization was the TAC (US$/kg-ABE): TOAC + TIAC TAC =

tri

(2)

FABE ⋅ to

TOAC and TIAC were the total operation annualized costs and the total investment annualized cost, respectively. In the optimization, The TIAC includes the costs of reactor, condenser, compressors, and membrane of reaction system. The TOAC includes the cost of cooling-water used in condensers, electricity cost, substrate cost,

7

and enzyme. FABE, to and tri were ABE total flow, annual operational time (8150 h) and return on investment time (3 years). Investment equipment functions were taken from Douglas, 1988. Low pressure steam, cooling water, electricity, enzyme costs were 2.18 $/ton-steam (3 atm) (Mussatto et al., 2013), 0.06 $/ton-water, 0.1 $/kWh and 4.24 $/kgprotein, respectively. Production flow was 4000 kg-butanol/h. In this paper, only the cost of hybrid reactor and ABE recovery by distillation were evaluated. The material of all equipment used in this work was stainless steel, except for the compressor that was carbon steel. 2.1.Evaporation A heat pump was proposed in SFS-V because non-condensables (CO2 and H2) must be compressed to atmos. pressure. Additionally, a heat pump increases the energy efficiency of SFS-V. The heat pump allowed using the condensation heat of compressed vapor to give energy to its evaporation (see figure in supplementary info). Then, the total energy requirement of SFS-V was due to compression work. The hybrid scheme was simulated in Aspen Plus V7.3® with a stoichiometric reactor (RStoic), and flash separator (Flash2) to validate the SFS-V model developed in Matlab®. The optimization variables were dilution rate, enzymes load, vacuum pressure, substrate concentration and work at each stage of compression. Filtration can be used to increase the ABE productivity of integrated reactors (Abdi et al., 2016). However, filtration increase the solids content into the reactor. This can cause insufficient mixing or an increase in energy requirements for mixing in stirredtank reactors, particularly above of 20% of solids (Kristensen et al., 2009). Due to this effect, filtration was not used in this work. 8

2.2.Pervaporation SFS-P can be observed in the Fig. 1. Membrane fouling can be significant in in-situ operation (Hu et al., 2015). However, membrane fouling was not considered. Therefore, an ideal removal of ABE by pervaporation was studied. Permeance and selectivity were assumed constant in the simulations (Baker et al., 2010). In the literature, experiments are conducted at pressures lower than 2 mbar. Assuming that the vacuum pressure was null, and knowing the flux, the compositions, and temperatures at which the experiments were carried out, the permeance of the membrane can be determined by the next equation: Peri =

Ji

(3)

( xi ⋅ γ i ⋅ Psati )

Some membranes with the highest permeances or selectivities reported in the literature can be found in Table 1. Selectivity of solvents was equal to butanol selectivity (Vane and Alvarez, 2013). Only CO2, H2, and water selectivities were assumed as 0.033, 0.033, and 1, respectively. Therefore, the hydrophobic membranes were considered highly nonselective to CO2 and H2. Additionally, the effect of gas selectivity of the membranes in energy requirement and TAC was studied. Membrane lifetime was three times lower than conventional equipment. In this work, a low membrane installation cost was used (100 $/m2, including module costs). CO2 compression was performed in multiple stages (Fig. 1). The condensers operated with cooling water at 30 ºC. A heat pump was not proposed due to the low pressure of permeate and the high membrane selectivity. The condensation heat of distillation was applied for the evaporation of the reaction system (Fig. 1). Therefore, the fuel consumption of the hybrid reactor was due only to the work of compressor given that the

9

vaporization heat of SFS-P was lower than the condensation heat of distillation. The variables of optimization were the feed flow, permeate pressure, enzymes load, membrane area, substrate concentration and compressors work. 2.3.Distillation ABE production and purification by distillation was simulated in Aspen Plus V7.3® optimal reactor conditions (vacuum pressure of reactor and yield of acetone, butanol, and ethanol). The reactor system was simulated in Aspen Plus V7.3® using a stoichiometric reactor follow of a flash at vacuum pressure. The model base of simulations was UNIQUAC. CO2 and H2 were defined as Henry components. Acetone, butanol and ethanol purity achieved in simulations were 0.99, 0.9 and 0.997, respectively. ABE recovery was 0.98. The heat-integrated distillation process used in this work was a double-effect distillation system with four separation columns (Fig. 2). The distillation system was proposed to reduce the energy requirement of SFS, SFS-V and SFS-P processes. The temperature of decantation-1 and decantation-2 was 40 °C. The decanter was simulated with liquid-liquid equilibrium parameters from Aspen Plus V7.3®. An intermediate condenser in the column C1 (Fig. 2) was used to supply the energy requirement of low-pressures columns (C1-LP and C2 columns). The feed stages of C1LP were three and one for the liquid from the decanter-1 and the fermentation, respectively. The feed stages of C1 were 17, 21, 22, 25 for liquid from fermentation, decanter-1, and decanter-2 and vapor from C3, respectively. The feed stage of C2 and C3 were 16 and 1, respectively. The energy integration for the feed from fermentation was performed 10

with the bottoms of the distillation column. Additional information of sizing of distillation system can be observed in supplementary info. In SFS-V and SFS-P, the gases from distillation system and hybrid reactor were mixed with the bleeding stream to reduce the water requirements of the stripping system. The noncondensables from this mixer (CO2 and H2) were sent to the stripping column (10 trays) to recover the carried ABE. 3. Results and Discussion 3.1.Kinetic model The kinetic model of the proposed metabolic pathway of all metabolites had a correlation coefficient of 0.93 and an error in the final concentration of butanol and acetone of 3.4 and 8.3%, respectively. Mean error of the kinetic model with the effect of inhibitors, from pretreatment, in butanol and ABE production with respect to experimental data (Zheng et al., 2015) were 13% and 7.9%, respectively. The error in the model was comparable to experimental variability of fermentations from hydrolyzed reported by Zheng et al., 2015 (Fig. 3). The maximum concentration of phenolics ([Phemax]) and furan ([Furmax]) that inhibits the growth in the minimization were 2.9 g/l and 4.1 g/l, respectively. The limit of furan concentration at which furans did not inhibit the growth, [LimFurans], was 0.87 g/l. Exponential coefficient, nF, in furans inhibition was 3.4. 3.2.Optimization of simultaneous fermentation and hydrolysis (SFS) The optimal operational conditions of SFS were 52 h, 17.5 g-protein/kg-cellulose and 45 g-cellulose/l for fermentation time, total enzyme ratio and cellulose concentration, respectively. The following results were achieve in the simulation of SFS: productivity 11

of 0.36 g-ABE l-1h-1; an ABE yield of 0.28 g-ABE/g-substrate; and acetone, ethanol and butanol concentrations of 5.1, 1.2 and 12.4 g/l. Substrate cost (0.06 $/kg-ABE), enzyme cost and reactor cost represented 34%, 20%, and 45% of the total annualized cost (TAC), respectively. 3.3.Optimization of SFS-V The simulations of SFS-V with Matlab 2015a® were compared with those performed in Aspen Plus V7.3® using glucose as feed, for model validation. Glucose conversion, vacuum pressure, and compressor work were fixed in Aspen Plus® from optimal conditions (glucose was fed only at 180 g/l with a cost of 0.06 $/kg-glucose). Butanol concentration into the reactor, ABE in the vapor phase, total compressor work and evaporation heat in contrast to Matlab® simulation, had an absolute variation percentage of 0.7%, 0.9%, 4.8%, and 4.4%, respectively. In optimal conditions, the ABE productivity, ABE yield, energy requirement and TAC feeding glucose or xylose as substrate to SFS-V were 1.8 g/l/h, 0.33 g/g, 4.1 MJfuel/kg-ABE and 0.36 $US/kg-ABE or 1.1 g/l/h, 0.37 g/g, 4.7 MJ-fuel/kg-ABE and 0.42, respectively. The substrate concentration in the feed was 180 g/l. It is important to remark that the productivity of xylose was 1.6 times lower than that of the glucose. Productivity has been proposed as the minimization objective function (Mariano et al., 2010). However, the TAC increased in 27% using ABE productivity as the objective function instead of the economic one (glucose feed). Additionally, the energy requirement increased 1.5 times and an ABE productivity of 6 g/l/h was achieved. Due to the high cost of enzymes and the low economic performance to achieve the highest

12

productivity, an economic optimization must be necessary using a lignocellulosic substrate. Butanol/acetone/ethanol ratio in optimal conditions feeding only glucose was 9.7/5.5/1. Moreover, ABE ratio using a lignocellulosic substrate (lignocellulose cost: 0.06 $/kgsubstrate) was 12.5/3.8/1. Experimental butanol/acetone production of Clostridium saccharoperbutylacetonicum N1-4 reported by Shinto et al., 2007 using a glucose concentration of 9.5 g/l and 53.5 g/l were 4 and 2, respectively. Experimental butanol/acetone ratio using xylose are between 4 and 6.1 (Shinto et al., 2008). Therefore, the obtained butanol/acetone ratios using glucose (1.8) and lignocellulose (3.3) were realistic in contrast with experimental data. TAC and compressor work in optimal conditions were between 0.43 and 0.84 $/kg (Fig. 4 (a)), and 5.5-8.9 MJ-fuel/kg-ABE (Fig. 4 (b)) for a substrate cost between 0.03 and 0.15 $/kg. The yield and reactor productivity increased from 0.24 to 0.35 g-ABE/g-substrate (Fig. 4 (a)) and from 0.9 to 1.02 g-ABE/L/h in optimal conditions, respectively (Fig. 4 (c)). Compressor work increased with respect to increasing of subtrate cost because the vacuum pressure was reduced in the optimization. It increased the velocity of solvents recovery. Therefore, a higher yield and productivity can be achieved. The vacuum pressure was reduced from 0.056 to 0.047 bar. One simulation for a substrate cost of 0.15 $/kg-ABE was performed with the optimal conditions of substrate cost of 0.03 $/kg-ABE. TAC was 0.92 $/kg-ABE. It was 9.2% higher than the TAC obtained with optimal conditions at 0.15 $/kg-ABE. The reduction of 9.2% in TAC was achieved with an increase of work requirement of 1.6 times.

13

TAC of SFS-V with respect to batch process (simultaneous fermentation and saccharification) at optimal conditions (substrate cost of 0.06 $/kg (Table 2)) was reduced in 13.5%. The heat pump achieved a coefficient of performance (COP) between 4.8 and 6.5 (heat requirement of evaporation divided by compression work). This low COP was due to the high CO2 and H2 compositions in the vapor phase. The pressure drop in condensation was studied. An increase in pressure drop in each stage of condensation from 0.03 to 0.12 bar with a substrate cost from 0.3 to 0.15 $/kg-ABE, increases the TAC and the energy requirements between 4.8-5.9% and 9.5-21%, respectively. 3.4.Optimization of SFS-P External units at higher temperatures than the reactor were not studied in this work for the following two reasons: 1) It increases the equipment number that should be used, like biomass recirculation or immobilization units and heat exchangers, 2) The heating before of pervoration has higher energy requirements due to the low solvent concentrations into the reactor. For example, the energy consumption to heat 10 ºC a stream with a butanol concentration of 5 and 10 g/l were 8.7 and 4.4 MJ/kg-butanol, respectively. Several membranes have been reported for butanol recovery (Claes et al., 2012; Dong et al., 2014; Fadeev et al., 2001; Huang and Meagher, 2001; Liu et al., 2011; Matsumura et al., 1992; Thongsukmak and Sikar, 2007) (Table 1). In this work, a composite membrane of silicalite and poly(1-trimethylsilyl-1-propyne) (PTMSP) was selected for the optimization due to its high permeance and selectivity. Trioctylamine/polypropylene (TOA/PP) was chosen for its high selectivity (11). The membrane of polydimethylsiloxane (PDMS) and ceramic support was optimized because it had an intermediate permeability and selectivity (1.9) (Table 1). 14

In the literature, several membranes with lower selectivity than one have been reported for butanol recovery by pervaporation. This means that theses membranes have a lower separation factor than the relative volatility of butanol in water. Therefore, pervaporation must not be used because the energy requirement will be higher than the direct evaporation at vacuum. As an example, Pervatech produces a membrane with a separating layer of PDMS (1 µm thickness) and porous polyimide support (approximately 200 µm). This membrane was evaluated for butanol recovery (Van Hecke et al., 2016). The separation factor of butanol of this membrane is between 14 and 20 (Van Hecke et al., 2016), while the relative volatility of butanol at 0.5 and 3 wt% is between 23 and 27. Experimentally, TOA/PP (Thongsukmak and Sikar, 2007) and PDMS (Dong et al., 2014) membranes were operated for 300 and 200 h without significant changes in performance, respectively. However, the stability of the silicalite-PTMSP membrane is unknown. Long term testing of stability are required for this membrane because difficult polymer synthesis and membrane stability issues are an impediment in the upscaling of the manufacturing of PTMSP membranes (Dubreuil et al., 2013). However, the PTMSP membrane was studied in this work for its high performance (Table 1). Membrane area was increased in the minimizations from 6900 to 10000 m2 with an increase in substrate cost from 0.03 to 0.15 $/kg (Fig. 4 (c)). The TAC of SFS-P with silicalite-PTMSP, TOA/PP and PDMS membranes was 0.57, 0.73 and 0.66 $/kg-ABE, respectively (substrate cost: 0.06 $/kg-ABE (Table 2)). It is very interesting to note that the membrane area required in optimal conditions using TOA/PP was null due to its low permeability (Table 1). Therefore, optimum conditions of TOA/PP membrane was equal 15

to a non-integrated reactor in semi-continuous mode. Membrane area of PDMS and enzyme load were 3.1 and 1.3 times higher than PTMSP. However, total compressor work of PTMSP and PDMS membranes in optimal conditions were similar (Table 2). The productivity with SFS-P and PDMS was 27% lower than PTMSP (Table 2). In the literature, the energy requirement of pervaporation is associated to the low total vaporization heat due to its high selectivity (Oudshoorn et al., 2009). However, compression work of CO2 was the most important item in the calculation of energy requirement. The high butanol fraction in the vapor phase of high selective membranes and the low-temperature of fermentation decrease the driving force of pervaporation. Therefore, low pressure in permeate (4.9 mbar) was needed. Compression energy was 5.3 MJ-fuel/kg-ABE, while vaporization heat of pervaporation was 2.0 MJ/kg-ABE (PTMSP membrane). TAC for PTMSP-membrane was between 0.9 and 7.4% higher than SFS-V (Fig. 4 (a)). Additionally, work requirement of SFS-P was between 22.8 and 27.5% lower than SFS-V. The enzyme load of SFS-P (between 19.8 and 24 g-protein/kg-butanol and PTMSPmembrane (Fig. 4 (c))) was higher than the SFS-V (between 15.9 and 17.6 g-protein/kgbutanol) due to the membrane area cost and the higher selectivity. A higher selectivity reduced the productivity of the reactor because there is less water in the permeate. Therefore, higher bleeding was required. This reduced the residence time of the substrate. For the same reason, the optimal substrate concentration was the maximum fixed in the simulation (180 g/l). The compositional changes for seasonal variation and varieties for a specific lignocellulosic material are significant. For example, the glucan composition of 16

varieties of sugarcane bagasse 05TG004 and 05TG008 were 34.1 and 40.7 % (Benjamin et al., 2014), a deviation of 16%. Due to this, the effect of xylose composition in lignocellulose material was studied in this paper. The effect of feed xylose composition can be observed in the Fig. 6. An increase of xylose concentration in the substrate reduces the TAC (Fig. 6 (a)) and increases the ABE yield (Fig. 6 (b)). The energy requirements increased from 3.8 to 4.9 MJ-fuel/kgABE to achieve at higher yield when the xylose ratio was increased from 0.6 to 1 gABE/g-substrate. The TAC was increased 4.5 and -5.3% with a reduction of 20% and -20% with respect to the base case (xylose composition of 0.33 g-xylose/g-substrate). The CO2 and H2 that permeated through the membrane were 66.6% and 68.2% of the total produced, respectively. When the selectivity of the gases was reduced from 0.033 to 0.001, the permeating CO2 reduced from 66.6% to 2.2%. This effect reduced the energy requirement of compression and TAC in 47% and 9.5%, respectively. In contrast, when the selectivity of the membrane was increased from 0.033 to 1 the energy requirement and TAC under optimal conditions increased in 4% and 11%, respectively. This was caused because 99.8% of CO2, instead of 66.6%, permeates through the membrane. The substrate concentration effect in SFS-P was studied without supplementation of inhibitors from pretreatment. In continuous operation (500 h), the TAC without distillation cost of SFS-P was higher than SFS when substrate concentration was lower than 120 g/l. The TAC of SFS-P including the distillation cost were not reduced in relation to the SFS process for substrate concentrations lower than 100 g/l.

17

On the other hand, the TAC of hybrid reactor was reduced from 0.56 to 0.48 USD/kgABE with an increase of substrate concentration from 180 to 350 g/l. However, inhibitors are produced in the pretreatment in a lignocellulose substrate and high solids can increase the energy requirement of mixing or insufficient mixing can be achieved. Additionally, a higher concentration of inhibitors are produced in the pretreatment at higher substrate concentration (Zheng et al., 2015). Pervaporation or vacuum evaporation was an inefficient way of detoxification. Therefore, it was necessary to reduce substrate concentration to decrease the inhibitors concentration in the feed (Fig. 5 (b)). SFS-P was economically comparable to SFS-V with the inhibitors from pretreatment (Fig. 5 (a)). TAC of SFS-P increased in 8.9% and 45% for inhibitors concentration in hydrolyzed supplementation of 2 and 4 g/l, respectively. Inhibitors from pretreatment increased the TAC 1.88 times with inhibitors concentration of 6 g/l (substrate cost: 0.06 $/kg-ABE). Compression work increased from 5.1 to 9.4 MJ-fuel/kg-ABE for inhibitors supplementation from 0 to 6 g/l for SFS-P with PTMSP (Fig. 5 (b)). The recovery of inhibitors was difficult due to the high boiling point of phenolic compounds and furans. In all cases, SFS-V achieved a higher yield, lower productivity, and higher energy requirements than SFS-P (Fig. 5). The furfural concentration in all simulations under optimal conditions was lower than 0.87 g/l. Therefore, there was not a negative effect of furans in biomass. However, it is important to remember than furfural concentration was reduced mainly by water dilution. The vanillin and furfural recovered in the vacuum phase was between 4.3-

18

6.2% and 23-35%, respectively. A stability studied must be performed due to the strong effect of inhibitors from pretreatment. 3.5.Energetic analysis and distillation cost The fuel consumption and TAC of distillation system for non-integrated reactor were 8 MJfuel/kg-ABE and 0.12 US$/kg-ABE, respectively. The unrecovered acetone was obtained mainly in the non-condensable gas. In the stripping column, the used cooling water was 2 kg-water/kg-ABE. Unrecovered ethanol and butanol were founded in vinasses. Only 0.1% of butanol was unrecovered. In the literature, energy requirement for distillation has been reported between 18 and 21 MJ/kg-ABE (Mariano et al., 2011; Qureshi et al., 2005; Sánchez-Ramírez et al., 2015). A low energetic consumption for an industrial distillation system has been reported by Jilin Cathy Industrial Biotech, which is 6-7 kg-steam/kg-butanol (Xue et al., 2013). In this case, fuel consumption was between 8.4-9.8 MJ-fuel/kg-ABE assuming an efficiency of steam production of 0.9 and an ABE ratio of 3:6:1. Therefore, the energy requirement achieved in this work by the distillation system proposed was between the lowest energy requirements reported for distillation. The lower fuel requirement obtained for distillation in this work was due to the intensive heat integration and the efficient scheme of distillation process. Due to the low ABE concentration of the non-integrated reactor, in the preheated of the distillation columns was achieved the highest heat integration of distillation process (14 MJ/kg-ABE), while the heat recovery in the condensation-reboiler, RC-1 (Fig. 3), was 3.7 MJ/kg-ABE. The energy requirement of distillation for SFS-V and SFS-P was around 4.1 MJ-fuel/kgABE (Table 2). Fuel consumption was 48.75% lower than conventional process because 19

ABE concentration increased in 4.7 times (optimal conditions: substrate cost 0.06 $/kg). In consequence, the separation cost of distillation was 39.6% or 45% lower for SFS-V or SFS-P, respectively. Distillation cost of SFS-P was 0.065 $/kg. It was 8.8% lower than SFS-V because a lower requirement of cold water (condenser heat was used to supply the heat required by SFS-P). Total fuel requirement of SFS-P was 9.3 MJ-fuel/Kg-ABE (optimal conditions at 0.06 $/kg-substrate). Fuel consumption was 16.3% higher than heat-integrated distillation system and 13.8-19.6% lower than SFS-V. The fuel consumption of distillation was reduced from 4.1 to 3.6 MJ-fuel/kg-ABE when ABE yield increased (optimal conditions at 0.15 $/kg-substrate). However, total energy requirement increased from 9.3 to 10.1 MJfuel/kg-ABE for higher vaporization and gas compression. Similar energy consumption has been reported for pervaporation in the literature, 9 MJ/kg-ABE (Groot et al., 1992) and 14 MJ/kg-butanol or 10.9 MJ/kg-ABE (calculated in this work with C. beijerinkii BA101 solvent yield) (Qureshi et al., 2005). However, the implication of compression work of CO2 or condensation in these reports is not clear. Energy requirement obtained for SFS-V in this work was 1.6 times lower than that reported by Mariano et al., 2011 (18 MJ-fuel/kg-ABE; calculated with energy efficiencies from this paper). Compression work reported by Mariano et al., 2011 is 5.5 MJ/kg-butanol or 3.7 MJ/kg-ABE. In this work, the energy requirement for compression was between 1.7 and 2.9 MJ/Kg-ABE, under optimal conditions at a substrate cost between 0.03 and 0.15 $/kg-ABE. In addition, compression work was lower than the reported by Mariano et al., 2011 due to different assumptions (ABE yield, ABE ratio, ABE concentration into reactor and pressure drop), optimization and more efficient compression system (4 compressors 20

instead of 2). The energy requirement of distillation (5 columns system) reported by Mariano et al., 2011 is 6.5 MJ-fuel/kg-ABE, 37% higher than the energy requirement obtained in this work. 4. Conclusions The less profitable option and with less fuel requirement was SFS. The only membrane that made SFS-P competitive with respect to SFS-V was PTMSP. The TAC of SFS-P (PTMSP) was between 0.85 and 7.4% higher than SFS-V. The membrane fouling was not studied. Therefore, SFS-V is preferable, from an economical and operational perspective, to be used in industrial scale. SFS-P reduced the energy requirement of SFS-V between 13.8-19.6%. An additional detoxification method or an efficient pretreatment system must be used to reduce the inhibitors concentration from pretreatment to less than 2 g/l at substrate loads > 180 g/l. Acknowledgements Authors thank the Colombian Administrative Department of Science, Technology and Innovation (COLCIENCIAS) for the financial support that made possible this work. Appendix. Supplementary data References 1. Abdi, H.K., Alanazi, K.F., Rohani, A.S., Mehrani, P., Thibault, J., 2016. Economic comparison of a continuous ABE fermentation with and without the integration of an in situ vacuum separation unit. Can. J. Chem. Eng. 94, 833–843. doi:10.1002/cjce.22461 2. Baker, R.W., Wijmans, J.G., Huang, Y., 2010. Permeability, permeance and selectivity: A preferred way of reporting pervaporation performance data. J. Memb. Sci. 348, 346– 352. doi:10.1016/j.memsci.2009.11.022 21

3. Benjamin, Y., Görgens, J.F., Joshi, S. V, 2014. Comparison of chemical composition and calculated ethanol yields of sugarcane varieties harvested for two growing seasons. Ind. Crops Prod. 58, 133–141. doi:10.1016/j.indcrop.2014.04.010 4. Chandel, A.K., Antunes, F.A.F., de Arruda, P.V., Milessi, T.S.S., da Silva, S.S., de Almeida Felipe, M. das G., 2012. Dilute Acid Hydrolysis of Agro-Residues for the Depolymerization of Hemicellulose: State-of-the-Art, in: da Silva, S.S., Chandel, A.K. (Eds.), D-Xylitol. Springer Berlin Heidelberg, Berlin, Heidelberg, pp. 39–61. doi:10.1007/978-3-642-31887-0_2 5. Claes, S., Vandezande, P., Mullens, S., De Sitter, K., Peeters, R., Van Bael, M.K., 2012. Preparation and benchmarking of thin film supported PTMSP-silica pervaporation

membranes.

J.

Memb.

Sci.

389,

265–271.

doi:10.1016/j.memsci.2011.10.035 6. Dong, Z., Liu, G., Liu, S., Liu, Z., Jin, W., 2014. High performance ceramic hollow fiber supported PDMS composite pervaporation membrane for bio-butanol recovery. J. Memb. Sci. 450, 38–47. doi:10.1016/j.memsci.2013.08.039 7. Douglas, J.M., 1988. Conceptual design of chemical processes, McGraw-Hil. ed. New York. 8. Dubreuil, M.F.S., Vandezande, P., Van Hecke, W.H.S., Porto-Carrero, W.J., Dotremont, C.T.E., 2013. Study on ageing/fouling phenomena and the effect of upstream nanofiltration on in-situ product recovery of n-butanol through poly[1(trimethylsilyl)-1-propyne] pervaporation membranes. J. Memb. Sci. 447, 134–143. doi:10.1016/j.memsci.2013.07.032 9. Fadeev, A.G., Selinskaya, Y.A., Kelley, S.S., Meagher, M.M., Litvinova, E.G., 22

Khotimsky, V.S., Volkov, V. V, 2001. Extraction of butanol from aqueous solutions by pervaporation through poly(1-trimethylsilyl-1-propyne). J. Memb. Sci. 186, 205–217. doi:10.1016/S0376-7388(00)00683-9 10. Fischer, K., Gmehling, J., 1994. P-x and .gamma..infin. Data for the Different Binary Butanol-Water Systems at 50.degree.C. J. Chem. Eng. Data 39, 309–315. doi:10.1021/je00014a026 11. Green, E.M., 2011. Fermentative production of butanol—the industrial perspective. Curr. Opin. Biotechnol. 22, 337–343. doi:10.1016/j.copbio.2011.02.004 12. Groot, W.J., van der Lans, R.G.J.M., Luyben, K.C.A.M., 1992. Technologies for butanol recovery integrated with fermentations. Process Biochem. 27, 61–75. doi:10.1016/0032-9592(92)80012-R 13. Hu, S., Guan, Y., Cai, D., Li, S., Qin, P., Karim, M.N., Tan, T., 2015. A novel method for

furfural

recovery via

gas

stripping

assisted

vapor

permeation

by a

polydimethylsiloxane membrane. Sci. Rep. 5, 9428. doi:10.1038/srep09428 14. Huang, J., Meagher, M.M., 2001. Pervaporative recovery of n-butanol from aqueous solutions and ABE fermentation broth using thin-film silicalite-filled silicone composite membranes. J. Memb. Sci. 192, 231–242. 15. Kadam, K.L., Rydholm, E.C., McMillan, J.D., 2004. Development and validation of a kinetic model for enzymatic saccharification of lignocellulosic biomass. Biotechnol. Prog. 20, 698–705. doi:10.1021/bp034316x 16. Kristensen, J.B., Felby, C., Jørgensen, H., 2009. Yield-determining factors in highsolids enzymatic

hydrolysis of lignocellulose.

doi:10.1186/1754-6834-2-11 23

Biotechnol.

Biofuels 2,

11.

17. Liu, X., Li, Y., Liu, Y., Zhu, G., Liu, J., Yang, W., 2011. Capillary supported ultrathin homogeneous silicalite-poly(dimethylsiloxane) nanocomposite membrane for biobutanol recovery. J. Memb. Sci. 369, 228–232. 18. Mariano, A.P., Costa, C.B.B., de Angelis, D.D.F., Filho, F.M., Atala, D.I.P., Wolf Maciel, M.R., Maciel Filho, R., 2010. Optimisation of a fermentation process for butanol production by particle swarm optimisation (PSO). J. Chem. Technol. Biotechnol. 85, 934–949. doi:10.1002/jctb.2383 19. Mariano, A.P., Keshtkar, M.J., Atala, D.I.P., Maugeri Filho, F., Wolf Maciel, M.R., Maciel Filho, R., Stuart, P., 2011. Energy requirements for butanol recovery using the flash fermentation technology. Energy & Fuels 2347–2355. 20. Mariano, A.P., Qureshi, N., Maciel Filho, R., Ezeji, T.C., 2012. Assessment of in situ butanol recovery by vacuum during acetone butanol ethanol (ABE) fermentation. J. Chem. Technol. Biotechnol. 87, 334–340. doi:10.1002/jctb.2717 21. Matsumura, M., Takehara, S., Kataoka, H., 1992. Continuous butanol/isopropanol fermentation in down-flow column reactor coupled with pervaporation using supported liquid membrane. Biotechnol. Bioeng. 39, 148–156. doi:10.1002/bit.260390205 22. Mussatto, S.I., Moncada, J., Roberto, I.C., Cardona, C. a., 2013. Techno-economic analysis for brewer’s spent grains use on a biorefinery concept: The Brazilian case. Bioresour. Technol. 148, 302–310. doi:10.1016/j.biortech.2013.08.046 23. Oudshoorn, A., van der Wielen, L.A.M., Straathof, A.J.J., 2009. Assessment of Options for Selective 1-Butanol Recovery from Aqueous Solution. Ind. Eng. Chem. Res. 48, 7325–7336. doi:10.1021/ie900537w 24. Qureshi, N., Bowman, M.J., Saha, B.C., Hector, R., Berhow, M. a., Cotta, M. a., 2012. 24

Effect of cellulosic sugar degradation products (furfural and hydroxymethyl furfural) on acetone-butanol-ethanol (ABE) fermentation using Clostridium beijerinckii P260. Food Bioprod. Process. 90, 533–540. doi:10.1016/j.fbp.2011.09.002 25. Qureshi, N., Hughes, S., Maddox, I.S., Cotta, M.A., 2005. Energy-efficient recovery of butanol from model solutions and fermentation broth by adsorption. Bioprocess Biosyst. Eng. 27, 215–222. doi:10.1007/s00449-005-0402-8 26. Qureshi, N., Singh, V., Liu, S., Ezeji, T.C., Saha, B.C., Cotta, M. a, 2014. Process integration for simultaneous saccharification, fermentation, and recovery (SSFR): production of butanol from corn stover using Clostridium beijerinckii P260. Bioresour. Technol. 154, 222–8. doi:10.1016/j.biortech.2013.11.080 27. Sánchez-Ramírez, E., Quiroz-Ramírez, J.J., Segovia-Hernández, J.G., Hernández, S., Bonilla-Petriciolet, A., 2015. Process Alternatives for Biobutanol Purification: Design and Optimization. Ind. Eng. Chem. Res. 54, 351–358. doi:10.1021/ie503975g 28. Shinto, H., Tashiro, Y., Kobayashi, G., Sekiguchi, T., Hanai, T., Kuriya, Y., Okamoto, M., Sonomoto, K., 2008. Kinetic study of substrate dependency for higher butanol production in acetone–butanol–ethanol fermentation. Process Biochem. 43, 1452–1461. doi:10.1016/j.procbio.2008.06.003 29. Shinto, H., Tashiro, Y., Yamashita, M., Kobayashi, G., Sekiguchi, T., Hanai, T., Kuriya, Y., Okamoto, M., Sonomoto, K., 2007. Kinetic modeling and sensitivity analysis of acetone-butanol-ethanol production. J. Biotechnol. 131, 45–56. 30. Thongsukmak, A., Sikar, K.K., 2007. Pervaporation membranes highly selective for solvents present in fermentation broths. J. Memb. Sci. 302, 45–58. 31. Van Hecke, W., Vandezande, P., Dubreuil, M., Uyttebroek, M., Beckers, H., De Wever, 25

H., 2016. Biobutanol production from C5/C6 carbohydrates integrated with pervaporation: experimental results and conceptual plant design. J. Ind. Microbiol. Biotechnol. 43, 25–36. doi:10.1007/s10295-015-1717-3 32. Vane, L.M., Alvarez, F.R., 2013. Hybrid vapor stripping-vapor permeation process for recovery and dehydration of 1-butanol and acetone/butanol/ethanol from dilute aqueous solutions. Part 1. Process Simulations. J. Chem. Technol. Biotechnol. 88, 1436–1447. doi:10.1002/jctb.4087 33. Xue, C., Zhao, X.Q., Liu, C.G., Chen, L.J., Bai, F.W., 2013. Prospective and development of butanol as an advanced biofuel. Biotechnol. Adv. 31, 1575–1584. doi:10.1016/j.biotechadv.2013.08.004 34. Zhang, Y., Ezeji, T.C., 2014. Elucidating and alleviating impacts of lignocellulosederived microbial inhibitors on Clostridium beijerinckii during fermentation of Miscanthus giganteus to butanol. J. Ind. Microbiol. Biotechnol. 1505–1516. doi:10.1007/s10295-014-1493-5 35. Zheng, J., Tashiro, Y., Wang, Q., Sakai, K., Sonomoto, K., 2015. Feasibility of acetone–butanol–ethanol fermentation from eucalyptus hydrolysate without nutrients supplementation. Appl. Energy 140, 113–119. doi:10.1016/j.apenergy.2014.11.037

26

Tables Table 1. Membranes reported in the literature for the separation to butanol Temperature

Mass

Flux

Separation

Permeance

[°C]

fraction

[kgm-2h-1]

Factor

[molbar-1m2h-1]

25-70

0.01

0.06-1.0

52-70

159-334

2.3-3.2

50

0.05

9.5

106

4061

6.2

5-11.2

41.6-25

990-1009

1.9-1.3

Membrane

Selectivity

PTMSP (Fadeev et al., 2001) Silicalite-PTMSP (Claes et al., 2012) Silicalite-PDMS (Liu et

0.00280

al., 2011)

0.03

Silicalite-PDMS (Huang 30-70

0.01

0.06-0.61

86-93

167-240

3.8-4.3

40

0.01

1.28

42.9

1285

1.9

30

0.0095

0.08

180

29.1

8.4

54

0.015

0.053

240

47.2

11.2

and Meagher, 2001)

PDMS-ceramic support (Dong et al., 2014) OA/PP (Matsumura et al., 1992)

TOA/PP (Thongsukmak and Sikar, 2007) PP, polypropylene. OA, oleyl alcohol. TOA, trioctylamine. PDMS, polydimethylsiloxane. PTMSP, poly[1(trimethylsilyl)-1-propyne]

27

Table 2. Performance of the process optimized in this work (substrate cost 0,06 $/kg)

Process

TAC of reactor (US$/kgABE)

TAC distillation and stripping system (US$/kg-ABE)

TAC total (US$/kgABE)

Yield (gABE/gsubstrate)

Productivity (g-ABE/l/h)

Compressor work (MJfuel/kg-ABE)

Fuel consumption of distillation (MJ/kg-ABE)

SFS-E 0.57 0.071 0.638 0.29 1.06 6.7 4.1 SFS-P 0.57 0.065 0.640 0.24 0.92 5.1 4.2 (PTMSP) SFS-P 0.66 0.068 0.728 0.24 0.77 5.5 4.1 (PDMS) C-1 0.66 0.118 0.778 0.28 0.37 8.0 PDMS, polydimethylsiloxane; PTMSP, poly[1-(trimethylsilyl)-1-propyne]; simultaneous saccharification, fermentation and vacuum evaporation (SFS-V) or pervaporation(SFS-P).

28

Figures

Fig. 1. Simultaneous fermentation, saccharification, and pervaporation (SFS-P), optimal conditions to substrate cost of 0.06 USD/kg, the membrane used in SFS-P was poly[1(trimethylsilyl)-1-propyne] (PTMSP). C1 and C2 are the distillation columns of Fig. 2. Lo and So were lignin and substrate, respectively. W and w, work and mass fraction.

29

Fig. 2. Acetone, ethanol, and butanol purification (S-I) from SFS-P by double-effect distillation process

30

Concentration [g/L] (Model)

15

Butanol

ABE

+30% 10

-30% 5

0 0

5

10

15

Concentration [g/l] (Experimental)

Fig. 3. Parity diagram of butanol and ABE, experimental data (Zheng et al., 2015)

31

22

8,5

18

7,5

14

6,5

(a) Productivity [g-ABE/L/h]

0,06

0,09

0,12

0,15

0,4

12

0,3 9 0,2 6 0,1 SFS-V

SFS-P

0

Work [MJ-fuel/kg-ABE]

(b) Yield (g-ABE/g-substrate)

0,03

Memmbrane area [m2/1000]

9,5

3 0,03

0,06

0,09

0,12

0,15

1,1

1

1

0,8

0,9

0,6 SFS-V

TAC [$/kg-ABE]

(c) Enzyme [mg-protein/gbutanol]

SFS-V SFS-P

26

SFS-P

0,8

0,4 0,03

0,06

0,09

0,12

0,15

Subtstrate cost [$/kg]

Fig. 4. Effect of substrate cost in the performance to optimal conditions of simultaneous saccharification, fermentation and vacuum evaporation (SFS-V) or pervaporation (SFS-P), the membrane used in SFS-P was poly[1-(trimethylsilyl)-1-propyne] (PTMSP). Substrate cost: 0.06 $/kg-ABE

32

35 0,6 25

15

0,1 0

2

4

6

10

(b) Cellulose [g/l]

9 8

90

7 6 60

5 SFS-V SFS-P

4

30

Work [MJ-fuel/kg-ABE]

120

3 0

(a) Yield (g-ABE/g-substrate)

Productivity [g-ABE/L/h]

45

2

4

6

0,3

1,1

0,28 0,9

0,26 0,24

0,7

SFS-V SFS-P

0,22 0,2

TAC [$/kg-ABE]

(c) Enzyme [mg-protein/g-butanol]

1,1 SFS-V SFS-P

0,5 0

2

4

6

Inhibitor concentration in hydrolysed (g/l)

Fig. 5. Effect of inhibitors hydrolyzed in the performance optimal conditions of simultaneous saccharification, fermentation and vacuum evaporation (SFS-V) or pervaporation (SFS-P), the membrane used in SFS-P was poly[1-(trimethylsilyl)-1propyne] (PTMSP). Substrate cost: 0.06 $/kg-ABE

33

0,4

1

0,35

0,8

0,3

0,6

0,25

Base case

(a)

0,4 0 2

Work [MJ/kg-ABE]

+20%

0,2 0,2

0,4

-20%

0,6

0,8

1 0,75

+20%

1,8

Base case

0,6

1,6 0,45 1,4

(b)

1,2 0

Yield [g/g]

-20%

0,3 0,2

0,4

0,6

0,8

TAC [USD/kg-ABE]

Productivity [kg-ABE/l/h]

1,2

1

Xylose ratio (β)

Fig. 6. The effect in performance of xylose ratio (xylose/(xylose+cellulose))

34

Highlights •

CO2 compression was the most important item in energetic evaluation

•

Detoxification by pervaporation or vacuum evaporation was inefficient

•

A kinetic model with furans and phenolics effect was proposed

35