Fractionation of Lignocellulosic Residues Coupling Steam Explosion and Organosolv Treatments Using Green Solvent γ-Valerolactone

A two-step fractionation of lignocellulosic residues of Phragmites australis in its main components (cellulose-pulp, soluble hemicellulose sugars, and lignin) is described, based on the biomass-derived solvent γ-valerolactone (GVL). The solvent used is an excellent substitute for traditional organic solvents as it is not toxic, it is renewable, and it can be recycled after the extraction process. Prior the GVL-organosolv extraction process, a steam explosion pretreatment was performed in order to break up the tight lignocellulosic structure and partially depolymerise hemicellulose into soluble sugars, making lignin easier to be solubilised. Three common extraction techniques were compared: soxhlet, closed vessel microwave-assisted, and open vessel on a hotplate stirrer. The two-step approach resulted in a cellulose-rich solid, water-soluble hemicellulose sugars and lignin-rich GVL liquor which was further purified for lignin isolation. The two best resulting pulps presented a high cellulose content (75.47% and 78.68%) starting from 38.13% and a content of lignin down to 11.96% and 13.09% starting from 23.02%. Almost all hemicellulose was removed with a final content of 0.72% and 2.20% starting from 20.5%.


Introduction
The expected decrease in petroleum extraction raised interest in the development of efficient ways of using renewable carbon sources [1].One of the most abundant biological organic materials is lignocellulose and a consistent fraction of it is to be found in the form of agricultural, forestry, municipal, and industrial waste.This residual material not only causes environmental problems, such as pollution, health issues, and safety hazards, but also represents an important renewable source [2] suitable for being used as a substrate for lignocellulosic feedstock biorefineries [3].The feedstock used in the present work, Phragmites australis (Common reed) residues, is a good example of lignocellulosic waste from the crafting of roofs, baskets, and beach umbrellas.Phragmites australis is a perennial grass that grows in wetlands or near inland waterways.Due to its fast-growing properties, high biomass yield, and low requirement in nutrients and water, this arboreal variety is recognized as a promising feedstock [4].Lignocellulose is a complex matrix where cellulose, hemicellulose, and lignin are closely arranged making it resistant towards both biological and chemical treatments [2].Several pretreatment methods have been proposed in order to alter or fractionate this type of biomass and render its main components available for different conversion processes [5][6][7][8][9][10].Although organosolv (OS) process is at present rather expensive, it can provide some valuable products [11].Compared to other chemical pretreatment processes, such as dilute acid hydrolysis, where some components of the biomass are degraded, the OS process, as well as steam explosion (SE) pretreatment, fractionates the biomass so that most of its components are recoverable [12,13].Another advantage is the possibility of recycling the solvent while in other treatments the reagents become waste with environmental issues and disposing costs.Therefore, this process suits the biorefinery concept where every fraction of the feedstock needs to be separated and valorised in order to make the whole process sustainable [14].
Recently γ-valerolactone (GVL) has been proposed as an interesting molecule because of its renewable nature, as it can be obtained from levulinic acid, which is a degradation product of structural carbohydrates present in lignocellulose [15,16].
Several applications have been reported for GVL such as an additive for liquid fuels with properties similar to those showed by ethanol [16] and also as a precursor for the production of liquid alkanes [14][15][16][17].Additionally, GVL has been described as a platform molecule for carbon-based chemicals such as 1,4 pentanediol, 2-methyltetrahydrofuran (MTHF), and methyl pentenoate, amongst others, which can be further upgraded to liquid fuels, additives, or polymers [16][17][18][19][20][21].Moreover, GVL has been found to be a very promising solvent with a low melting point (−31 • C), a high boiling point (207 • C) and very low toxicity.It was recently employed as a solvent in catalytic conversion reactions [22][23][24] and also in the processing of lignocellulosic biomass [25,26].
In this article Phragmites australis residues were treated in a two-step approach, coupling SE pretreatment and the GVL-OS process, in order to obtain a cellulose-rich solid and a lignin-rich liquid phase that was further purified for lignin isolation.In this novel configuration the SE breaks down the lignocellulose structure [27] making lignin easier to be removed by the green solvent GVL.Three common extraction techniques were compared: soxhlet, closed vessel microwave-assisted, and open vessel on a hotplate stirrer.
Removing lignin from the SE pretreated solid enhances the digestibility of the substrate if it is to be used in the production of bioethanol [28,29] because of lignin's inhibitory effect due to the adsorption of enzymes.Additionally, higher ethanol yields are expected for a substrate with higher cellulose content.Another route could be the optimization of this two-step approach for pulping purposes.Furthermore, the isolated lignin and the sugars derived from hemicellulose, as well as other chemicals, like acetic acid and furfural, can be valorised through several applications listed elsewhere [28,[30][31][32].
Phragmites australis processing residues were employed as a lignocellulosic substrate for the whole experimentation and they were obtained from local handicrafts in the Lake Trasimeno area (Umbria, Italy).The starting composition of the raw material was determined by National Renewable Energy Laboratory (NREL) (Golden, CO, USA) characterisation procedure [33] and the results are summarised below in Table 1.

Methods
Figure 1 summarises the workflow procedure followed in this experimentation.On the left of the figure are shown the main steps of the procedure i.e., SE and OS, while on the right and on the bottom are shown the secondary steps of separation, i.e., recovery, precipitation, washing, and drying.

Steam Explosion
The SE pretreatment was carried out in a pilot-scale reactor provided by Biochemtex (Tortona, Italy).The feedstock was pretreated at three different severity factors of LogR0 3.6, 4.0, and 4.4.The LogR0 value describes the intensity of the SE pretreatment as described by Overend et al. [34] according to Equation (1): where t is the time (min) and T is the temperature (°C).The output materials were divided in a water insoluble substrate (WIS), containing mainly cellulose and lignin, and a pretreatment liquor rich in hemicellulose sugars and degradation products.The WIS was washed in hot water (50 °C) and then treated for the GVL-OS extraction process.
Three extraction techniques were compared using GVL at a solvent/biomass ratio of 10:1.Soxhlet extraction (SOX) was performed in a traditional lab glass apparatus for six cycles and the solvent temperature measured in the top vessel was ≈90 °C.An open vessel extraction (OV) was conducted on a hotplate stirrer at 150 °C for 30 min and 120 min.The microwave-assisted extractions (MAE) were carried out in a Milestone ETHOS One (Sorisole, Italy) unit and the working conditions were 150 °C or 200 °C and 30 min or 120 min.The OV was not tested at 200 °C because it is close to the GVL boiling temperature.The operating conditions are summarised below in Table 2.

Steam Explosion
The SE pretreatment was carried out in a pilot-scale reactor provided by Biochemtex (Tortona, Italy).The feedstock was pretreated at three different severity factors of LogR 0 3.6, 4.0, and 4.4.The LogR 0 value describes the intensity of the SE pretreatment as described by Overend et al. [34] according to Equation (1): R 0 = te T−100 14.75 (1) where t is the time (min) and T is the temperature ( • C).
The output materials were divided in a water insoluble substrate (WIS), containing mainly cellulose and lignin, and a pretreatment liquor rich in hemicellulose sugars and degradation products.The WIS was washed in hot water (50 • C) and then treated for the GVL-OS extraction process.
Three extraction techniques were compared using GVL at a solvent/biomass ratio of 10:1.Soxhlet extraction (SOX) was performed in a traditional lab glass apparatus for six cycles and the solvent temperature measured in the top vessel was ≈90 • C.An open vessel extraction (OV) was conducted on a hotplate stirrer at 150 • C for 30 min and 120 min.The microwave-assisted extractions (MAE) were carried out in a Milestone ETHOS One (Sorisole, Italy) unit and the working conditions were 150 • C or 200 • C and 30 min or 120 min.The OV was not tested at 200 • C because it is close to the GVL boiling temperature.The operating conditions are summarised below in Table 2.
After the extraction procedure the insoluble solid was separated by a filtering crucible.It was then washed with clean solvent (same volume used for the extraction) in order to prevent re-precipitation of lignin [11] and then washed again with water at 50 • C. Finally, the substrate was dried and weighed.Lignin-rich GVL-liquor was concentrated by means of evaporation under vacuum in order to recover GVL solvent.The lignin dissolved into the concentrated GVL-liquor was precipitated by adding water using a ratio of 1:10 (v/v).The precipitate was recovered by centrifugation and supernatant was analysed for the characterisation of sugars.All the solids obtained were characterised following the NREL analytical procedures [33].The structural carbohydrates were quantified as their monomers, after acid hydrolysis with H 2 SO 4 , by Thermo Scientific Dionex Ultimate 3000 high-performance liquid chromatography (HPLC) (Thermo Fisher Scientific, Waltham, MA, USA) equipped with an Aminex Biorad HPX87H column (Bio-Rad, Hercules, CA, USA) and a refractive index detector.The acid insoluble residue (AIR) was quantified in a crucible after acid hydrolysis.Ash was quantified by burning the AIR in a muffle furnace at 575 • C. Acid insoluble lignin (AIL) was determined as the difference between AIR and ash.All the aqueous fractions were characterised by HPLC following NREL analytical procedures for determination of sugars [35].All the compounds that were not directly characterised, namely other components were quantified as the difference between the total mass of the sample and the mass of all the characterised components in the sample.
The lignin extracted (L ext ) was quantified as the difference between AIL content in the raw material and its residual content in the sample after organosolv extraction according to Equation (2): where RM is the mass of raw material (g); L RM is the content of AIL (wt %) in raw material; IS is the mass of insoluble solid after organosolv extraction (g); and L IS is the content of AIL (wt %) in IS.
The cellulose enrichment (C ent ) was determined according to Equation (3): where C fIS is the percentage of cellulose in insoluble solid after organosolv extraction and C fRM is the percentage of cellulose in raw material.
The hemicellulose removed H rem was quantified as the difference between hemicellulose content in the raw material and its residual content in the sample after organosolv extraction according to Equation ( 4): where RM is the mass of raw material (g); H RM is the content of hemicellulose (wt %) in raw material; IS is the mass of insoluble solid after organosolv extraction (g); H IS is the content of hemicellulose (wt %) in IS.

Results and Discussion
Three extraction techniques were assessed to outline some possible routes in the scaling up of the OS process.SOX was chosen as it is a standard reference extraction technique.The OV apparatus represents a scenario of atmospheric pressure working conditions while in the MAE the closed system allows working under pressure.
Table 3 displays the characterisation of the insoluble solids obtained after the SE and GVL-OS processes where C f , H f , L f , A f , and O f are the percentages of cellulose, hemicellulose, lignin, ash, and other components.

Comparison between Extraction Techniques; Effect of Time and Temperature
As shown in Figure 2 the L ext did not vary significantly depending on the extraction technique.In fact, for any block at the same LogR 0 value and extraction time the results are comparable.For example, at LogR 0 4.0 in the 120 min block the values are similar to each other and are all higher than those of the 30 min block suggesting that there is a small difference when using OV or MAE technique.On the other hand, at LogR 0 3.6 and 4.0 the effect of the residence time in the extraction is slightly more significant than the contribution of temperature which did not present a clear relationship with the L ext .
Energies 2017, 10, 1264 6 of 11 those of the 30 min block suggesting that there is a small difference when using OV or MAE technique.On the other hand, at LogR0 3.6 and 4.0 the effect of the residence time in the extraction is slightly more significant than the contribution of temperature which did not present a clear relationship with the Lext.Soxhlet extraction is not comparable in terms of temperature and time of the process because of its technical characteristics, where the temperature cannot be controlled and the extraction time depends on the number of cycles and is longer if compared to the other techniques assessed.Moreover the solvent is clean for every cycle.However, it showed good figures in terms of Lext and Cent (Figure 3).

Figure 3. Cellulose enrichment after SE and two-step approach (SE and GVL-OS). PAM is milled raw
Phragmites australis; PA3.6, PA4.0, and PA4.4 are WIS of Phragmites australis pretreated at increasing severity factor (LogR0 3.6, 4.0, and 4.4)  Soxhlet extraction is not comparable in terms of temperature and time of the process because of its technical characteristics, where the temperature cannot be controlled and the extraction time depends on the number of cycles and is longer if compared to the other techniques assessed.Moreover the solvent is clean for every cycle.However, it showed good figures in terms of L ext and C ent (Figure 3).those of the 30 min block suggesting that there is a small difference when using OV or MAE technique.On the other hand, at LogR0 3.6 and 4.0 the effect of the residence time in the extraction is slightly more significant than the contribution of temperature which did not present a clear relationship with the Lext.Soxhlet extraction is not comparable in terms of temperature and time of the process because of its technical characteristics, where the temperature cannot be controlled and the extraction time depends on the number of cycles and is longer if compared to the other techniques assessed.Moreover the solvent is clean for every cycle.However, it showed good figures in terms of Lext and Cent (Figure 3).Regarding the comparison between OV and MAE extractions, these both achieved comparable results.The advantage of working under atmospheric pressure is that there is no need to work under extremely tight conditions [11].On the other hand microwave-assisted biorefineries according to Sobhy and Chaouky [36] are expected to lead to improved process control and higher efficiency than conventional heating systems.
These preliminary tests laid the foundations for further optimization and scaling up of the extraction method where both atmospheric pressure systems or pressurised MAE systems represent a feasible path.

Severity Factor (LogR 0 ) Effect
According to the literature [37,38] the SE pretreatment disrupts the heteropolysaccharide-lignin network and depolymerises lignin by cleaving the β-O-4 aryl ether bonds resulting in an increased solubility in alkali and organic solvents.In Figure 2 the tests are divided by LogR 0 and confirmed that delignification increased by increasing the SE pretreatment severity.In fact, in the results for the milled sample, which did not undergo SE, the L ext values were the lowest.However, it can also be seen that the delignification rate reached a plateau between LogR 0 4.0 and LogR 0 4.4.In fact, as shown in Table 3 for some trials e.g., OVB4.0 and OVB4.4 the L ext values are similar, respectively, 75.98% and 77.61%, and in some cases, such as MD4.0 and MD4.4, the value is higher for the former, 75.52% and 74.89%, respectively.This is probably due to reduction of lignin solubility in harsh conditions and condensation of a glycosidic material, described as pseudo-lignin [39], which increases the mass of residual lignin into the insoluble solid after the GVL extraction.It is also likely that at a severity factor of 4.0 the total extractable lignin is already easy to solubilise meaning that it is not worthwhile to push the SE pretreatment to high conditions.Another interesting trend is the one showed by all the tests at LogR 0 4.4 which all showed similar L ext .This suggests that at LogR 0 4.4, the harsh steam explosion conditions altered the biomass structure greatly, so that lignin was easily extracted regardless of the experimental conditions used (time, temperature, and extraction technique) in the extraction step.
In order to evaluate the effect of coupling SE and GVL-OS processes, the cellulose enrichment into insoluble solid was plotted against the LogR 0 value together with the cellulose enrichment into WIS obtained by the sole SE (Figure 3).
It can be seen that the SE pretreatment not only makes lignin easier to solubilise but also enriches the substrate in cellulose.This effect increases with the LogR 0 because of the hydrolysis of the glycosidic bonds in hemicelluloses and, to a lesser extent, in the cellulose [38].In fact the samples showed an enrichment in cellulose due to the loss of hemicelluloses until LogR 0 4.0 and then a decrease in cellulose at LogR 0 4.4 because of the harsher conditions.This confirms that it might be counter-productive to work at high severity factors.The other curves in the graph showed a small increase in cellulose fraction for the milled sample.On the other hand, good enrichments were achieved for the SE pretreated samples, especially at LogR 0 4.0, where some tests reached cellulose overall enrichments above 100% (see Table 3) meaning that the cellulose fraction was doubled in the recovered solid.In this graph it is also confirmed that the effect of the severity factor (before reaching the plateau at 4.0) is the main contribution to the substrate fractionation.In fact all the curves representing the different extraction techniques and process variables showed the same trend and rather similar results.The C ent rose until LogR 0 4.0 whereas between this value and LogR 0 4.4 the gain was negligible.
It is also important to consider the effect of this process on hemicellulose.This structural polymer is more susceptible to hydrolysis compared to cellulose and lignin.Hence, the majority of hemicellulose is hydrolysed and solubilised in the form of monomeric and oligomeric sugars by the first SE step, whilst the second step is less aggressive towards the degradation of hemicellulose.These considerations are confirmed by the results displayed in Figure 4, which shows the removal of hemicellulose from the substrate in the worst and best case (OV 150 • C 30 min on the left and MAE 200 • C 120 min on the right).All the other results lie between these two.The majority of hemicellulose was removed in the SE step.Part of it was recovered in the pretreatment liquor (grey segment) while the rest was degraded into other substances (dark grey segment).The light grey segment represents the hemicellulose removed by the solvent extraction and it can be further recovered.For the samples which did not undergo SE (OVAm and MDm) the solvent extraction showed very poor removal of hemicellulose.The double treated samples showed increasing removals of hemicellulose with increasing LogR 0 reaching 98% at 4.4.However, at this LogR 0 most of the hemicellulose was degraded (92%).On the contrary, at LogR 0 3.6, 37% of hemicellulose was recovered in the pretreatment liquor, whilst only 13% of it was degraded.

Conclusions
The coupling of steam explosion and organic solvent extraction with GVL showed good results in terms of delignification (75-78%) and cellulose enrichment (≈100%) of Phragmites australis.Pulps obtained presented a high cellulose content of up to 78.68% and a content of lignin and hemicellulose down to 11.96% and 0.50%, respectively.These results represent a promising starting point for a further optimisation of the process.
A steam explosion pretreatment of LogR 0 4.0 was sufficient to maximise the delignification by GVL-OS process, while no significant effects were observed in the milled feedstock.Anyway the SE conditions need to be further optimised, between LogR 0 3.6 and 4.0, to reduce the degradation of hemicellulose sugars.The GVL-OS process showed similar figures both at high temperature with short time (200 • C-30 min) and low temperature with long time (150 • C-120 min), meaning that SE pretreatment played a critical role in Phragmites australis fractionation.Regarding the extraction techniques assessed, all showed good figures of the variables investigated but the MAE seemed the less time consuming and the best in terms of conditions control.
The two-step process proposed suits the biorefinery concept where all the components of biomass are recovered and can be valorised and it can be applied for those applications where high cellulose content materials are needed, e.g., in the production of cellulosic ethanol or paper production.In fact the materials obtained can be suitable to be hydrolysed to obtain monomers with enzymatic cocktails, removing the inhibitory effect of both hemicellulose and lignin or, after further optimisation, used in the paper industry.

Figure 1 .
Figure 1.Flowchart of the steps followed in the experimentation.

Figure 2 .
Figure 2. Overview of Lext comparing different extraction techniques, temperatures, and residence times.

Figure 2 .
Figure 2. Overview of L ext comparing different extraction techniques, temperatures, and residence times.

Figure 2 .
Figure 2. Overview of Lext comparing different extraction techniques, temperatures, and residence times.

Figure 4 .
Figure 4. Comparison of Hrem between OV (a) and MAE (b) methods.The Hrem plot (a and b) is divided into three parts: the fraction recovered into the pretreatment liquor (recovered hemicellulose), the fraction extracted by OS treatment (extracted hemicellulose), and the fraction degraded by SE pretreatment.

Figure 4 .
Figure 4. Comparison of H rem between OV (a) and MAE (b) methods.The H rem plot (a and b) is divided into three parts: the fraction recovered into the pretreatment liquor (recovered hemicellulose), the fraction extracted by OS treatment (extracted hemicellulose), and the fraction degraded by SE pretreatment.

Table 2 .
Summary of operating conditions for samples that underwent to OS treatment.
Figure 1.Flowchart of the steps followed in the experimentation.

Table 2 .
Summary of operating conditions for samples that underwent to OS treatment.

Table 3 .
Characterisation of samples after two-step process SE and GVL-OS.C f , E f , L f , A f , and O f are the content of cellulose, hemicellulose, lignin, ash, and other components into the samples.