Prediction of yields and composition of char from fast pyrolysis of commercial lignocellulosic materials, organosolv fractionated and torrefied olive stones

This study investigated the fast pyrolysis behaviour of torrefied olive stones, fractionated olive stones and lignocellulosic commercial compounds. Olive stones were reacted in a continuous industrial torrefaction unit. The olive stones were also fractionated into their main components in an organosolv reactor at temperatures from 170 to 190 C in both the presence and absence of an acidic catalyst. All samples were reacted in a wire mesh reactor at different temperatures (800–1150 C) and heating rates (400–1150 C/s), and the solid product was characterised for its yield, morphology, and elemental composition. The char yields from fast pyrolysis of commercially available cellulose, hemicelluloses, and lignin were compared with yields of fractionated olive stones. A model was developed to compare the measured yields of olive stones with the predicted yields using fractionated or commercial components. The presence of acid during fractionation had a stronger effect than the temperature, particularly on the lignin fraction. The fractionated lignocellulosic compounds provided more accurate predictions of the char yields of olive stones, as compared to the commercial lignocellulosic compounds. The fractionation at 180 C without acid catalyst gave the cellulose, hemicellulose, and lignin with highest degree of purity and resulted in the most accurate predictions of the experimental yields of olive stones. The results showed that interactions between the lignocellulosic components were not significant. The char yield of each fractioned compound and non-treated olive stones could be accurately predicted from the lignocellulosic content which has importance for biorefinery applications in which each fraction is used as a value-added


Introduction
The European Commission has established the Renewable Energy Directive that sets a binding target of 32% of renewables by 2030, as well as targets for a climate-neutral economy by 2050 [1]. In this landscape, carbon-neutral biomass is of crucial importance, since coalfired power plants can be retrofitted, and new biomass-fired plants can be commissioned. Moreover, biomass utilisation must be sustainable itself, and the minimisation of forest-intensive biomass and usage of biomass wastes is pressing. The differences between forest and waste biomass can be mitigated through torrefaction that energetically densifies the biomass and enhances its grindability and shelf-life. Torrefaction is a mild pyrolysis process that converts biomass into a more carbon-rich material with increased energy density and decreased oxygen content. Despite numerous previous studies on high temperature application of torrefied feedstock materials [2][3][4][5][6], few studies systematically investigate how the chemical and structural variance of torrefied biomass affects the product yield and composition.
Pyrolysis is an important step of most thermochemical processes involving biomass such as gasification and combustion. The conditions during pyrolysis largely influence the solid, liquid, and gas yields of the resulting bio-product, and a significant number of studies focus on the influence of parameters such as temperature and heating rate [7][8][9][10][11][12]. The reaction temperature is observed to be the most sensitive variable for prediction of product yields during slow and fast pyrolysis [13,14]. A recent review by Kan et al. [15] summarises the advances made on the effect of the operating parameters on the pyrolysis of lignocellulosic biomass.
The mechanisms and pathways occurring during pyrolysis of biomass have also been reviewed recently [16,17], with special emphasis given to the conversion and interaction of individual components (cellulose, hemicellulose, and lignin). As pointed out by Shen et al. [16], attention ought to be given to both sample preparation and interaction assessment in quantitative terms. The interaction between lignocellulosic biomass components has been investigated by Yu et al. [18] for a variety of biomass (oak, spruce, and pine), under both slow and fast pyrolysis conditions at temperatures below 600 • C. The authors observe interactions between cellulose and the other two components, but not between xylan and lignin, therefore stating that the yields of the original biomass could not be predicted from the individual components. Similarly, Couhert et al. [19] measure the gas yields from the flash pyrolysis of lignocellulosic biomass (beechwood, a mix of spruce and fir, rice husk, grass, and wood bark) and compare these yields with those from commercially available components (cellulose, xylan, and lignin), but they do not identify an additivity law that allows prediction of the yields during high-temperature pyrolysis and gasification. Previous studies [20,21] also show that the lignocellulosic compositional differences affect both the soot yield and oxidation reactivity and that the modelling accuracy can be improved by the consideration of the effects of primary and secondary reactions, as well as explicit mass and heat transfer during pyrolysis at temperatures above 800 • C. The results presented above highlight that the interaction of all lignocellulosic compounds during high-temperature pyrolysis is important for the prediction of yields and composition. Other studies [22][23][24] have also reported clear non-additive behaviour during pyrolysis of individual components. There are, however, some studies that claim that no interaction between components takes place for their specific fuels and pyrolysis conditions [25][26][27].
In addition, chemical pretreatment and fractionation method of feedstocks have a significant effect on the composition of lignocellulosic compounds and their further decomposition during pyrolysis and gasification [28]. Due to the difficulty in extracting the components of lignocellulosic biomass, most studies use commercially available cellulose, hemicellulose, and lignin, to investigate the interactions between components during pyrolysis (e.g. [24,[29][30][31]). However, a few studies use laboratory-extracted hemicelluloses [22,30]. Matsakas et al. [32,33] developed a hybrid fractionation method that combines organosolv and steam explosion pretreatment and allows retrieval of all three components from the wood and herbaceous biomass. Organosolv pretreatment is known as an effective method to fractionate biomass into cellulose, hemicellulose and lignin streams by using aqueous-organic solvent mixtures, with high solvent concentration (30-70%) at temperatures of 100-220 • C, with or without the addition of catalysts [34,35]. One of the main benefits of organosolv pretreatment is the isolation of high-quality lignin and cellulose fractions [36,37]. Another two advantages of the organosolv process are related to the relatively easy recovery and re-use of the commonly used organic solvents (such as ethanol or acetone) and improved mass transfer and dissolution of lignin in the presence of an organic solvent [38][39][40]. Previous research showed that organosolv pulps have bleachability and viscosity retention which are comparable to those of cellulose soda and kraft pulps [41]. The valorisation and CO 2 reactivity of organosolv fractionated lignins have been shown strongly affected by the type of feedstock and operating conditions of organosolv fractionation [42]. The literature is scarce on the yields and composition of products from pyrolysis of organosolv fractionated lignocellulosic compounds.
Since high heating rates and high temperatures, typical of fast pyrolysis, can be easily achieved in wire mesh reactors (WMR), these reactors are used recently to study the pyrolysis of various biomass feedstocks [11,12,[43][44][45]. Zhang et al. [43] compare yields and kinetics from pyrolysis of pellets of rice husk, straw, pine, and pine nut shell using both WMR and thermogravimetric analyser (TGA) and observe that the product yields remain constant for temperatures above 500 • C. Trubetskaya et al. [11,12,45] have extensively studied the pyrolysis of various biomass feedstocks using WMR under different conditions. In their studies, the authors conclude that the heat treatment temperature and presence of K affect the char yield more significantly than the heating rate and differences in the plant component levels, and that the presence of silicates in rice husk and wheat straw affect the morphology of the char. The authors observe that the influence of alkali on the char yields is more noticeable at heating rates characteristic of WMR. Specifically, the presence of K and Ca in herbaceous feedstocks resulted in catalytic effect and stronger cross-linking that prevented extensive melting of the chars during WMR pyrolysis.
Given the complexity and diversity of the chemical composition of biomass, the aim of the present study is for the first time to compare the char yields and composition from commercial and fractionated lignocellulosic compounds in high-temperature pyrolysis. The specific objectives are to: (1) use an organosolv process to fractionate olive stones into cellulose, hemicellulose and lignin, (2) measure the char yield of commercial and fractionated lignocellulosic compounds using a wire mesh reactor for fast pyrolysis, and (3) develop an empirical model to estimate the solid char yields from decomposition of raw and torrefied olive stones during high-temperature pyrolysis. Furthermore, the addition of an acid catalyst during fractionation is highlighted, and the effect on the yields and composition of the pyrolysis products is analysed. Pyrolysis performance highly depends on the reactor conditions. Thus, experiments are performed in a well-characterised wire mesh reactor (WMR) system to permit model validation. The novelty of this work relies on the finding of optimum fractionation conditions for olive stones based on component purity and high-temperature pyrolysis behaviour and on the comparison of the char yields and properties of fractionated compounds with commercially available samples.

Olive stones fractionation
Olive stones were fractionated to their main components: lignin, cellulose, and hemicellulose in an organosolv reactor at Luleå University of Technology following a laboratory procedure previously described [46]. More specifically, an air-heated multidigester apparatus was used, containing 2.5 L metallic cylinders. Olive stones were mixed with the solution of 60% v/v ethanol in water at a liquid to solid ratio of 10 (v/w) and placed in the cylinders. Treatment took place for 1 h at 170 • C and 180 • C in both the presence and absence of an acidic catalyst (1% w/ w biomass H 2 SO 4 ), and at 190 • C in the absence of acidic catalyst. At the end of the treatment time, the reactor was cooled to room temperature, and the pre-treated solids were removed from the slurry by vacuum filtration, washed with the 60% v/v ethanol solution and air-dried at room temperature until further use. The flow through liquor was collected and the ethanol was removed in a rotary evaporator. Ethanol removal resulted in reducing the solubility of lignin in the liquor, which was recovered by centrifugation (14,000 rpm, 29,416 × g, at 4 • C for 15 min) and air-dried at room temperature. The remaining aqueous solution, containing the solubilised hemicellulose, was dried in an oven at 40-50 • C to reduce the water content. The methodologies for characterisation of the fractionated cellulose, hemicellulose and lignin can be found in the supplemental material.

Commercial samples
Various lignocellulosic biomass samples were examined in this study. Samples comprised olive stones, both in raw and torrefied forms, and a variety of commercially available hemicelluloses, crystalline cellulose, and lignin. Washed olive pits (Olea europaea) were sourced from Spain and are a by-product of the olive oil industry where they are separated, crushed to < 1 mm and air-dried. Olive stones were torrefied during 24 h at 280 • C under nitrogen at the Arigna Fuels plant located in Leitrim, Ireland. A sample of the torrefied material was collected at the end of the experiment and held at ambient temperature in a desiccator, as described previously by Trubetskaya et al. [3]. The commercial hemicelluloses were wheat arabinoxylan, tamarind seed xyloglucan, larch arabinogalactan, and guar galactomannan (Megazyme International, US). In addition, crystalline cellulose and beechwood xylan (both from Sigma-Aldrich, US), and beechwood lignin (BOC Sciences, US) were used. The samples were used in different size cuts, and these were chosen to prevent loss of sample through the gaps in the wire mesh which were 40 µm, and to have uniform temperature throughout the particle. The particle size of the samples can be found in Table S1 of the supplemental material.

Fast pyrolysis in wire mesh reactor
A wire mesh reactor (WMR) was used to pyrolyse all samples under high temperature and high heating rate conditions. The WMR used herein comprised a wire mesh, conductive electrodes, a welding machine as power source, a thermocouple, a glass chamber, and a pressure gauge. This setup was based on the design by Gibbins and Kandiyoti [47]. Further details of the WMR used in this work can be found elsewhere [48]. Three different conditions were defined with plateau temperatures and heating rates as follows: T1 -800 • C and 400 • C/s, T2 -1000 • C and 800 • C/s, and T3 -1150 • C and 1150 • C/s, all under 1 atm of nitrogen in quiescent conditions. The residence time of 5 s was found sufficient to ensure the complete pyrolysis of the studied fuels, according to previous works by the authors [11,49]. The initial amount of sample was kept to 10 (±1) mg to minimise non-uniformities in heat distribution within the sample. Details of the methodology employed during WMR pyrolysis experiments can be found in the supplemental material. The char yield measurements were performed at least in triplicate and the results presented in Figs. 1-3 are the result of the arithmetic mean of all measured yields. The error bars displayed in Figs. 1-3 represent the 98% confidence interval limits, and the error was calculated to be within ± 4 wt% for all samples and conditions. The inaccuracy during measurement of char yields was mainly due to weighing errors. The experimental char yield, Y exp , which includes moisture and ash in both initial and char samples, was calculated using Eq. (1), where m f(mesh+char) is the final mass of the mesh and char, m f(mesh) is the final mass of the mesh after removal of the char, m i(mesh+sample) is the initial mass of the mesh and the raw sample, and m i(mesh) is the initial mass of the mesh.

Characterisation of raw feedstock and pyrolysis products
Ultimate and proximate analysis were performed for all raw samples and selected chars. Chars from condition T2 were characterised for their morphology using Scanning Electron Microscopy (SEM). The commercial and fractionated lignocellulosic samples were also analysed using an Attenuated Total Reflectance -Fourier Transform Infrared (ATR-FTIR) spectrometer. Details of the aforementioned analyses and procedures can be found in the supplemental material.

Modelling
Based on the experimental results, a model was developed to estimate the char yields from pyrolysis of biomass at conditions relevant to entrained-flow gasification. Previously, Couhert et al. [19] and Yu et al. [18] attempted to model pyrolysis gas yields based on the weight fraction additivity law using the yields obtained from isolated cellulose, xylan, and lignin feeds to predict the yields observed for pyrolysis of biomass sample of known composition. However, the model predictions   deviated significantly from the experimental results, suggesting that the usage of different components (i.e. fractionated components rather than commercial) may be required to capture the observed trends. The present model is based on the additivity law, and the measured char yield from pyrolysis of olive stones was compared with the predicted char yield using fractions from commercial or fractionated lignocellulosic compounds. In Eq. (2), Y i is the average product yield obtained from pyrolysis of the isolated lignocellulosic compound and α, β and γ are the mass fractions of lignocellulosic compounds in the untreated olive stones and torrefied olive stones, as shown in the supplemental material (Table S3). Table 1 shows the results of chemical composition of fractionated cellulose and hemicellulose. Fractionation of olive stones at 180 • C with and without an acid catalyst gave significantly higher cellulose content in the pre-treated solids compared to that from fractionations at 170 or 190 • C. Low amounts of residual hemicellulose and acid soluble lignin were detected in the fractionated cellulose at 180 • C treatment, with lignin contents of 14.3 and 15.7 wt%, demonstrating that treatment at 180 • C was the optimal to deliver pre-treated solids with high cellulose, low hemicellulose and low lignin contents. Introduction of the acid catalyst at 170 and 180 • C resulted in lower hemicellulose content, as is expected due to the acidic conditions that promote hemicellulose solubilisation. Further increase of the organosolv treatment temperature to 190 • C, resulted in severe treatment conditions that promoted degradation of cellulose (see Table 1). In general, the results showed that the mass balances could be accurately closed for the fractionated cellulose with greater than 85 wt% db.

Characterisation of fractionated lignocellulosic compounds
The hemicellulose fraction overall contained a low concentration of glucose (<2.6 wt%), and the major components were hemicellulosic sugars. Lignin was present mainly as acid soluble lignin, and the total lignin (both acid soluble and acid insoluble) was <9.9 wt% for all samples except for the sample at 170 • C without acid. The hemicellulose from extraction at 180 • C yielded the highest content in hemicellulosic sugars which was 50.6 or 45.9 wt% of xylose and mannose with the small amount of lignin varying between 6.9 and 9.9 wt% and traces of glucose which were below 2.6 wt%. The incorporation of acid at 170 • C produced a hemicellulose fraction (48.1 wt%) which was similar to those obtained by the fractionation without catalyst at 180 • C. At 190 • C, the mass balances could not be closed for hemicellulose likely because at higher temperatures a higher dissolution of feedstock into the liquid phase could occur, together with decomposition of the dissolved compounds to degradation products such as furfural, HMF and other light organics could evaporate during fractionation, as observed by Pu et al. [50].
The results in Table 1 showed that olive stone is a promising raw material to be used in organosolv fraction, as high purity fractions were obtained. The content of cellulose at 180℃ during olive stone fractionation was similar to that in pre-treated birch solids and slightly higher compared to the pre-treated spruce solids under the optimal conditions defined in those studies [32,33].
The fractionated lignin presented a high degree of purity with 4.1 wt % of hemicellulose and without detected cellulose sugars (see Table 2). The lignin fractionated at 180 • C without acid showed the highest purity with only 0.9 wt% of hemicellulose. Moreover, the amount of impurities in lignin from fractionation of olive stones was similarly low as in the lignin samples from birch and spruce. The size exclusion analysis showed a broad variation of the lignin molecular weight from 3670 to 7200 Da. In general, the molecular weight of the fractionated lignin samples with and without acid treatment was greater than that of the organosolv, Soda, Alcell, and Kraft lignins that varied from 726 to 4660 Da [51]. This highlighted the milder reaction conditions of the steam organosolv treatment in the present study. Overall, lower molecular weight fractions showed lower polydispersity, corresponding to previous results [52][53][54]. Additionally, the polydispersity index of lignin in the present study varied from 4.1 to 7.1 and thus, was greater than that of lignin samples reported in the literature (organosolv wheat: 2.0, organosolv poplar: 2.1, organosolv spruce: 2.2, Soda P1000: 3.5, Alcell: 3.3, and Indulin Kraft: 4.1) [51]. The greater polydispersity index suggests a broader molecular weight distribution and less uniformity of the polymer mixture [51].

Effect of feedstock
The char yields of all commercial samples are represented in Fig. 1. The yields of torrefied olive stones were slightly above those of the raw olive stones. Torrefaction led to a decrease in cellulose and hemicellulose contents while the lignin content appeared unaffected (see Table S3 in the supplemental material). Thus, the greater char yield from pyrolysis of torrefied olive stones was likely related to secondary char formation due to decomposition of cellulose, and hemicellulose.
The yields from the commercial hemicelluloses (with the exception of xylan), under condition T1, were within 5.7-6.6 wt% (Fig. 1). However, during condition T2 two groups were identified: xyloglucan and galactomannan with 7.1 and 8.1 wt%, respectively, and arabinoxylan and arabinogalactan, both with yields of ~ 12 wt%. Compared to the other hemicelluloses, xylan had a lower degradation temperature, an overall exothermal behaviour, and higher char formation [29], with yields of 19.8 and 17.2 wt% for conditions T1 and T2, respectively. As shown in supplemental material (Table S2), xylan showed the highest ash content (3.7 wt%) among the hemicelluloses. Since alkali metals are known to favour char formation, it is likely that the observed differences in higher char yield formation during xylan pyrolysis were related to the catalytic effect of alkali metals, confirming previous results [55,56].
The char yields of olive stones fractionated compounds obtained from condition T1 are presented in Fig. 2. For the components extracted without acid treatment, the cellulose and hemicellulose fractions from  170 and 180 • C had similar yields (~10 wt% and ~ 14 wt% for celluloses and hemicelluloses, respectively), but the char yield of lignin was greatly affected by the pre-treatment temperature (14.8 wt% for 170 • C and 29 wt% for 180 • C). Moreover, the char yields from pyrolysis of cellulose and hemicellulose using 180 • C non-acid and acid treatment were similar. However, the lignin char yield was greater from non-acid treatment at 180 • C (29 wt%) than from the acid fractionation (22.1 wt %). The char yields of each of the three components fractionated at 170 and 180 • C in the presence of an acid catalyst resembled each other, which showed the more important effect of the acid catalyst rather than the temperature during fractionation under these conditions. For the highest temperature, 190 • C, and the absence of acid catalyst, results showed cellulose having a lower char yield and hemicellulose and lignin with yields similar to each other. It should also be noted that the char yields of the hemicelluloses were always above those of the celluloses. Particularly for the case of treatment at 190 • C without acid, the yield of hemicellulose was three-fold that of cellulose. The amorphous phase in hemicellulose results in low thermal stability of hemicellulose and allows for further rearrangement reactions which promote higher char yields than during the pyrolysis of cellulose that mostly contains a more thermally stable crystalline phase, confirming previous results of Alen et al. [57]. Moreover, catalytic effects during char formation [40] may have also resulted in higher char yields for the hemicelluloses as it presented higher ash content as compared with the celluloses (Table S6 of supplemental material).
The highest char yield was observed during lignin pyrolysis. The high content of benzene rings led to strong depolymerization and crosslinking of lignin compounds during pyrolysis. In addition, the high alkali metal content in lignin could lead to the greater char yield [55,58].

Effect of temperature and heating rate
The char yields of raw and torrefied olive and fractionated compounds (180 • C acid and no acid) are represented in Fig. 3 for the three applied temperatures (800, 1000, and 1150 • C). In general, the char yield decreased slightly with an increase in temperature and heating rate. However, no significant effect attributable to either temperature (800-1150 • C) or heating rate (400-1150 • C/s) was investigated herein. The effect of temperature and heating rate during fast pyrolysis in a wire mesh reactor was investigated by Trubetskaya et al. [11] for softwood, hardwood, several types of straw, and rice husk. In line with the results obtained herein, the authors concluded that temperatures over 800 • C and heating rates above 600 • C/s had a negligible effect on the char yields. On the other hand, the char yields of crystalline cellulose, wheat arabinoxylan and larch arabinogalactan increased with the rise in temperature (800 to 1000 • C) and heating rate (400 to 800 • C/s), as seen in Fig. 1. The increase in char yield during pyrolysis of cellulose, arabinoxylan and larch arabinogalactan was attributed to secondary char formation. Formation of the main degradation compounds from arabinose, xylose, mannose and arabinitol during pyrolysis is known to give similar products [59]. However, pyrolysis of arabinoxylan, glucomannan and arabinogalactan is known to yield a greater amount of propanal-2-one and glycolaldehyde compared to xylan pyrolysis [29]. Thus, the increased polymerization and cross-linking of propanal-2-one and glycolaldehyde could lead to greater char yields with the increased heat treatment temperatures.

Char morphology and structural transformation
The morphology of selected chars obtained at 1000 • C was investigated using SEM-EDS. Selected images of chars are presented in Fig. 4 for raw and torrefied olive stones, and commercial hemicelluloses and lignin. Chars from both raw and torrefied olive stones displayed particles of similar size and morphology to the original fuel. The chars from torrefied olive stones appeared to be more porous than the chars from raw olive stones (see Fig. 4a and b). The chars from commercial hemicelluloses presented significant differences with respect to morphology and particle size. Arabinoxylan chars were long, fibrous and mostly nonporous, with the formation of scale-like particles which indicated partial melting, and EDS indicated the presence of Mg. Xyloglucan chars showed large cavities and micropores, and the presence of large flat structures indicated melting followed by repolymerisation and crosslinking. Galactomannan displayed large structures with micron-sized vesicles and micropores (see Fig. 4e2), and few of the particles underwent fragmentation into fibre-like ~ 100 µm length particles. Arabinogalactan chars were of two distinct sizes, likely due to fragmentation during pyrolysis. Regardless of the particle size, the surface of all chars was covered by calcium-rich ordinated structures ~ 10 µm in length (see Fig. 4f2). Xylan from beechwood presented some porous particles and needle-like structures rich in Na and Ca on the outer surface of the chars (see Fig. 4g2). Chars of beechwood lignin were nonporous, the surface of which displayed micron-sized vesicles of 5-10 µm (see Fig. 4h2).
The chars from the lignocellulosic compounds from 180 • C no acid (O1) and 180 • C acid (O2) fractionation are represented in Fig. 5. These chars showed significantly different morphologies. In the case of the fractionated cellulose, the fibres lost their integrity through the thermal degradation process with a much broader fibre dimension. The structure of the fractionated cellulose looked similar to that of an amorphous cellulose [24,60] that went through the formation of a liquid intermediate with bubbles formed from liquid boiling during pyrolysis [60][61][62]. Acid treatment resulted in chars which were more fluidised than nonacid treated chars, as a result of a more extensive melting during char formation. It is known that during fast pyrolysis, cellulose has a tendency to form a depolymerised liquid intermediate [63], and this greatly influences the morphology of the cellulose char. Micropores were observed for O1 cellulose, but not for O2 cellulose as a consequence of the extensive melting of the latter char which caused collapsing of micropores. Chars from hemicellulose with and without acid treatment displayed clear differences. Char from O1 hemicellulose presented large flat structures with a high degree of melting and formation of micropores and vesicles. As mentioned by Yu et al. [18], these vesicles are formed from the release of volatiles when the surface of the char is significantly melted, allowing bubbles to form and oftentimes burst into spherical micropores. O2 hemicellulose chars did not undergo extensive melting, and therefore cross-linking reactions were allowed to take place and resulted in fibrous non-coalesced particles with a few micropores on their surface. Both lignin chars (with and without acid treatment) showed smooth surfaces ( Fig. 5e1 and f1), which was a consequence of the formation of a liquid intermediate. Particularly acid-treated lignin (O2) displayed more coalesced and melted structures and a higher number of micropores and vesicles formed, as a clear consequence of the acid treatment that increased melting of lignin during pyrolysis and subsequent ejection behaviour [61]. Non-acid treated lignin showed fewer micropores and vesicles, and resembled the results reported by Hilbers et al. [24] for organosolv lignin.

ATR-FTIR
Original samples were analysed using ATR-FTIR to infer changes in functional groups. The results are presented in the supplemental material for olive stones (raw and torrefied) and commercial samples, and in Fig. 6 for the fractionated components of olive stones. The peak assignment and references for each band position are summarised in Tables S7 and S8 in the supplemental material. FTIR analysis of the fractionated olive stones components showed that mostly hemicelluloses showed a vibration band at 895 cm − 1 due to -CH stretching. The small size of the band at 895 cm − 1 could indicate that only a small amount of cellulose and hemicellulose achieved a crystalline structure by remaining mostly amorphous. All components gave peaks at 1033 cm − 1 (C-C, C-OH, C-H [64]). The C-O stretch at 1115 cm − 1 and the C-C and C-O stretch at 1215 cm − 1 were mainly observed for the lignins. Oppositely, the G ring stretching at 1242 cm − 1 was detected for the cellulose and hemicellulose fractions. The aromatic skeletal vibrations at 1424 cm − 1 and the C-H deformations in CH 2 and CH 3 were observed for all samples [64]. The aromatic skeletal vibrations S > G at 1594 cm − 1 were only detected for the lignins, whereas the S < G skeletal vibrations were detected for all components [65,66]. This indicated that small quantities of lignin remained present in cellulose and hemicellulose after olive stone fractionation. The C=O stretching band (1727 cm − 1 ) that was associated with the presence of carboxylic acids was only observed for the celluloses and hemicelluloses, as suggested by Marchessault and Liang [67]. The C-H stretching in the range 2911-2944 cm − 1 and the broadband at 3323-3370 cm − 1 from -OH and -NH stretching were presented for all samples [68]. In general, similar IR bands were obtained among cellulose, hemicellulose, and lignin.

Char elemental composition
The elemental composition of all chars obtained under condition T2 (1000 • C, 800 • C/s) is represented in Fig. 7 in the form of a Van Krevelen diagram. Chars from raw and torrefied olive stones showed similar elemental composition. The elemental analysis of commercial hemicelluloses showed variations in H/C and O/C composition. The H/C ratios of galactomannan and arabinoxylan were similar (0.25). The H/C of arabinogalactan was 0.31, and the O/C was lowest among the hemicelluloses (0.07) while the H/C of xyloglucan was the highest (0.37) among the hemicelluloses. Galactomannan (62% mannose) showed a greater O/C ratio (0.18) compared to arabinogalactan (85% galactose). This was due to the higher yields of 5-hydroxymethylfurfural and water from pyrolysis of d-mannose than from pyrolysis of d-galactose [69]. The elemental O/C ratios of arabinoxylan (50% xylose), xyloglucan (36% xylose) and beechwood xylan were similar due to the presence of xylose units in all three samples. Lignin from beechwood showed the highest H/C and O/C among all chars, due to its low carbon content of 27 wt%. This result, which differed significantly from the carbon content of lignin chars from softwood (79.6 wt%) and wheat straw (73.6 wt%) obtained by Trubetskaya et al. [20] under similar pyrolysis conditions, showed the strong effect of the lignin species on the composition and yield of the char.
The addition of an acid catalyst at 170 and 180 • C decreased cellulose O/C and H/C ratios by 60-80 % and 20-40%, respectively, as compared to the fractionation without acid catalyst. The ability of cellulose to interact with the acid catalyst depends on the ratio of amorphous and crystalline regions in the polymer [70]. However, the present results indicate that the extensive dehydration reactions during acid pretreatment were mostly affected by the treatment temperature greater than 160 • C and could, therefore, accelerate the penetration of acid into both crystalline and amorphous regions [71]. Small concentrations of acid (<1%) are known to be sufficient to interact with cellulose chains located at the surface of crystallites and lead to the formation of new hydrogen bonds [72,73]. When the pyrolysis temperature increases, the depolymerization of the dehydrated cellulose occurs, i.e. the splitting of the glycoside bonds produces the volatile products.
On the other hand, the addition of an acid catalyst during pretreatment at 170 and 180 • C led to an increase of 30% of H/C and O/C ratios in pyrolysis of hemicellulose. This was due to the catalytic influence of acids that can inhibit the formation of furfural during pyrolysis [74,75]. The fast hydrolysis of hemicellulose at temperatures >130 • C and retention times below 10 min could accelerate the conversion of cellulose to glucose and thus, decrease the formation of furfural [76].
The acid pretreatment showed a greater influence on the composition of lignin fractionated at the lowest temperature (170 • C) with the O/C ratio decreasing by 80% with the addition of acid. This was related to the increased solubilization of lignin during pretreatment in the temperature range 170-180 • C leading to an increase in carbon content and decrease in oxygen content of lignin char [77]. These results were in line with the carbon content of raw lignin (63 and 44 wt%, respectively for 170 • C with and without acid).
The elemental analysis of raw materials and fractionated lignocellulosic compounds showed that the sulphur content is below 0.2% w/w, whereas commercial lignin from beechwood (1.5% w/w) was rich in sulphur. However, the char yields of commercial lignin were twice higher than the char yields from pyrolysis of fractionated olive stone lignins (see section 3.2.1). This indicates that the presence of sulfuric acid during pre-treatment did not affect the yields of solid chars due to the higher char yield from sulphur rich commercial beechwood lignin compared to yields from pyrolysis of fractionated lignins.

Modelling
The predicted char yields were calculated using the fractionated compounds from 170, 180, and 190 • C and the results are compared with the measured char yield of olive stones and torrefied olive stones, as shown respectively in Figs. 8 and 9.
Only the chars obtained from fractionation at 180 • C (with acid or no acid) showed a relatively small difference to the measured char yield of olive stones for pyrolysis under conditions T1 and T2 (800 and 1000 • C). For condition T3 (1150 • C), however, 180 • C no acid samples presented a better fit as compared to the acid treated samples. Similar to raw olive stones pyrolysis, an accurate prediction of char yield from pyrolysis of torrefied olive stones ( Fig. 9) was achieved using the char yields of lignocellulosic compounds fractionated at 180 • C without any catalyst for all three temperature conditions. Thus, the overall trends of the model predictions for both raw and torrefied olive stones are in general agreement with the experimental data. Fig. 10 shows the predicted char yields of olive stones using the char yields from commercial lignocellulosic compounds for pyrolysis at 800 and 1000 • C. The cellulose used in the model was the crystalline cellulose from Sigma Aldrich used in this study. The hemicellulose char yield was varied using yields of xylan, arabinoxylan, xyloglucan, arabinogalactan, and galactomannan. The yields of lignin were also varied using the data for the lignin from beech wood, and softwood and wheat straw lignin from previous studies [20]. The results indicated that the char yield of lignin of beech wood overestimated the char yield of olive stones and torrefied olive stones. This was due to the greater concentration of alkali metals and sulphur in the lignin from beech wood than in the two other lignin samples (softwood and straw). Alkali metals are known to act catalytically in pyrolysis reactions enhancing the formation of solid char [56]. Due to a similar ash content and composition of softwood and wheat straw lignin samples, the influence on the predicted char yield of olive stones was small.
The highest yield of olive stones was obtained from a prediction that used the measured char yield of xylan. The variation in other hemicelluloses had a less significant impact on the predicted char yield of olive stones for both pyrolysis operating conditions at 800 and 1000 • C. Specifically for 800 • C, mixtures with any hemicellulose (except xylan) and softwood or straw lignin gave acceptable predictions of the yields of raw and torrefied olive stones. For 1000 • C, the same mixtures showed a good fit for raw olive stones, whereas for torrefied olive stones mixtures with xylan and softwood or straw lignin provided the most accurate predictions.
It was thus showed that the usage of commercial compounds for  Van Krevelen diagram of all studied WMR chars obtained at condition T2 (1000 • C, 800 • C/s). Note that crystalline cellulose is not represented since its low char yield did not allow retrieval of sufficient sample for analysis. Fig. 8. Predicted char yields using fractionated olive stones and measured char yields from raw olive stone pyrolysis at 800, 1000, and 1150 • C in the wire mesh reactor.
prediction of char yields may result in under or overpredictions. Compared to fractionated lignocellulosic compounds at 180 • C without acid catalyst, the use of char yields from commercial lignocellulosic compounds gave in general a less accurate prediction of olive stone char yield.

Discussion
The results of the present study showed that the fractionation conditions, particularly the acid catalyst, had a great effect on the char yields, morphology, and composition. At 180 • C, the acid catalyst addition resulted in lignin with more coalesced and melted structures, along with a decrease in the char yield, and subsequent less accurate model predictions for both raw and torrefied olive stones. For the cellulose and hemicellulose fractions, a relation between the melting degree and the char yield was not found as described further in this section. It can be concluded that the acid catalyst mostly influenced the pyrolysis behaviour of lignin and therefore had a clear effect on the model predictions (see Figs. 8 and 9).
The fractionated lignins were significantly different from the commercial beechwood lignin. The char yield of commercial beechwood lignin was high (54 wt%) compared with chars from commercial lignins obtained under similar conditions, namely softwood (34 wt%) and wheat straw (35 wt%) [20]. Moreover, the char yield of commercial beechwood lignin was approximately twice that of the fractionated lignins (54 vs. 22-25 wt%), leading to the overprediction of the char yields (see Fig. 10). Based on the comparison of the SEM images, the results showed that commercial beechwood lignin chars were highly cross-linked with only minor melting due to the high ash content (46 wt %), whereas the 180 • C fractionated lignin chars underwent clear melting (cf. Figs. 4h, 5, and f).
The commercial and fractionated celluloses demonstrated clear differences in char yields. Commercial cellulose, which has a high crystallinity, based on the FTIR analysis, gave a lower char yield than any of the fractionated celluloses (4.2 vs. 8.9 wt%). While the ultimate analysis of both commercial and 180 • C fractionated celluloses did not vary significantly (C content within 36-42 wt%), the fractionated celluloses were mostly amorphous based on the shape and allocation of fibres as seen in the SEM images, despite their high degree of purity (75-80% for 180 • C treatment). The influence of the degree of crystallinity on the char yields of celluloses, i.e. that less crystalline samples give higher char yields, was thus underlined, and should be considered during the fractionation of lignocellulosic compounds.
The fractionated celluloses displayed differences in morphology with acid treatment resulting in more fluidised chars. However, these morphological differences had no effect on the char yields which were 8.9 wt% for both acid treated and non-acid treated celluloses (condition T1). For the 180 • C acid treated cellulose, an intermediate compound (metaplast) appeared to form, but complete melting of the particles was not achieved due the occurrence of competing cross-linking. This is in line with the mechanism proposed by Bradbury et al. [78] according to which, the pyrolysis of cellulose is defined by means of two competitive reactions (char and gases formation, and tar and gases formation). For the 180 • C acid treated cellulose, an intermediate compound was apparently not extensively formed, as visible in Fig. 5 from the lack of melting of the chars, which suggested that a single-step reaction could be fitted.
The fractionated hemicelluloses had char yields similar to those of arabinogalactan and arabinoxylan. Even though these chars were morphologically different, they appeared to have similar degrees of cross-linking. Xylan was among the hemicelluloses with highest char yield (17-20 wt%). Observation of the SEM pictures showed a more extensive melting of galactomannan as compared to xylan for which cross-linking took place before extensive melting resulting in higher char yield.
The fractionation at 180 • C without acid catalyst gave the cellulose and hemicellulose with highest degree of purity (see Table 1) and resulted in the most accurate predictions of the experimental yields of raw and torrefied olive stones (see Figs. 8 and 9). On the other hand, fractionation at 170 • C and 190 • C without acid resulted in the least pure cellulose and hemicellulose and subsequent underestimation of the char yields of raw and torrefied olive stones. The data strongly suggested that the additive behaviour of the samples was mostly related to the degree of purity of its cellulose and hemicellulose. Moreover, the results show that interactions between the components were not significant and that the char yields of olive stones could be predicted from the pyrolysis of its individual components.

Conclusion
The novelty of this work relies on the fast pyrolysis of organosolv fractionated lignocellulosic compounds from olive stones, along with the comparison with commercially available samples. The results showed that the acid treatment of biomass during fractionation had a strong effect on the char yield and composition during pyrolysis, whereas the organosolv process temperature was a less preponderant  factor. Importantly, it has been shown for the first time that the fractionated lignocellulosic compounds at 180 • C without an acid catalyst, provided the most accurate prediction of raw and torrefied olive stone char yields, as compared to the commercial lignocellulosic compounds. The results obtained provide relevant information to understand the effect of the fractionation conditions on the purity of organosolv lignocellulosic compounds, and on the product yields and properties of chars obtained from fast pyrolysis which is of great importance for kinetic modelling and biorefinery applications for which the knowledge of the char yield of each component (cellulose, hemicellulose, and lignin) is required.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.