A mechanistic model of fast pyrolysis of hemicellulose

Xiaowei Zhou a, Wenjun Li b, Ross Mabon b and Linda J. Broadbelt *a
aDepartment of Chemical and Biological Engineering, Northwestern University, 2145 Sheridan Road, Evanston, IL 60208, USA. E-mail: broadbelt@northwestern.edu; Fax: +1-847-491-3728; Tel: +1-847-467-1751
bCorporate Strategic Research, ExxonMobil Research and Engineering Company, 1545 US Route 22 East, Annandale, NJ 08801, USA

Received 9th November 2017 , Accepted 29th January 2018

First published on 29th January 2018


Hemicellulose is one of the major components of lignocellulosic biomass, which is an abundant source of renewable carbon on the Earth and has potential for the production of renewable drop-in transportation fuels and multiple commodity chemicals. In this work, a structure for hemicellulose extracted from corn stover was proposed to capture the experimentally characterized structural properties. A mechanistic model for hemicellulose pyrolysis was constructed based on the reaction family approach that we used for cellulose pyrolysis before. The model described the decomposition of hemicellulose chains, reactions of intermediates, and formation of a range of low molecular weight products (LMWPs) at the mechanistic level and specified rate constants for all the reactions in the network. Overall, 504 reactions of 114 species were included in the mechanistic model for fast pyrolysis of extracted hemicellulose. The mechanistic model closely matched experimental yields of various products with mass yield ≥1 wt%. Modeling results show that both the degree of polymerization and the polydispersity index of hemicellulose have an insignificant effect on the pyrolysis product distribution. Then, the mechanistic model of extracted hemicellulose is further extended to simulate the fast pyrolysis of native hemicellulose. Comparison of the model results showed that fast pyrolysis of native hemicellulose from corn stalk yielded more char, gaseous species, acetol, and much more acetic acid than that of extracted hemicellulose from corn stover, while yielding less 1,2-anhydroxylopyranose, 1,2;3,4-dianhydroxylopyranose and glycolaldehyde.



Broader context

Fast pyrolysis of lignocellulosic biomass offers potential for production of renewable liquid products that can be co-processed with vacuum gas oil using fluid catalytic cracking units or catalytically upgraded to transportation fuels and chemicals. Extensive studies have been conducted to understand the complexity of fast pyrolysis of biomass, especially for its three major components: cellulose, hemicellulose, and lignin. However, hemicellulose receives scarce attention compared to cellulose and lignin, while it deserves much more. A fundamental understanding of the reaction mechanism and kinetics of hemicellulose pyrolysis is crucial for better design, optimization and control of the process of biomass pyrolysis, and yet is lacking. In this work, the first mechanistic model of hemicellulose pyrolysis was developed based on the reaction family approach and continuous distribution kinetics. The model specifies the detailed pathways for the decomposition of hemicellulose polymeric chains, the reactions of intermediates, and the formation of various experimentally observed products. Extension of the model to fast pyrolysis of native hemicellulose was also demonstrated. Insight and information that the model provides lead to a better understanding of fast pyrolysis of hemicellulose at the mechanistic level. In a broader point of view, the fundamental modeling framework developed here can be further extended to unravel the chemistry and kinetics of thermal decomposition of both natural polymers (such as biomass) and synthetic polymers (such as waste plastics).

1. Introduction

Fast pyrolysis is a thermochemical technology that can break down lignocellulosic biomass into bio-oil (∼65–75 wt% yield of the original feedstock, including water) within seconds at moderate temperatures of typically ∼500 °C. Bio-oil potentially can be co-processed with vacuum gas oil in an existing fluid catalytic cracking process1–4 or catalytically upgraded to hydrocarbon fuels and multiple commodity chemicals.5–10 Fast pyrolysis of lignocellulosic biomass offers potential for production of renewable fuels and chemicals in a sustainable way. Tremendous efforts, especially in the past three decades, have been devoted to investigating fast pyrolysis of biomass, especially for the three major components of lignocellulosic biomass: cellulose, hemicellulose, and lignin. However, fast pyrolysis of hemicellulose (accounting for ∼20–35 wt% of dry biomass) is much less studied both experimentally and computationally than that of cellulose and lignin.11–13 Recently, we provided a critical review on hemicellulose pyrolysis in terms of composition and structural features of hemicellulose as well as fast pyrolysis experiments, the reaction mechanism and kinetic modeling.13

There are many inherent challenges in deepening the understanding of the thermochemistry of hemicellulose pyrolysis. The first challenge lies in the varieties of heterogeneous structures of hemicellulose. Unlike cellulose which is a homogeneous polysaccharide with a well-defined structure comprised entirely of β-1,4-linked D-glucose units, hemicellulose is a mixture of heterogeneous polysaccharides that are mainly composed of pentose and hexose sugar building blocks. As detailed in our previous work,13 the functional building blocks of hemicelluloses include pentoses (D-xylose and L-arabinose), hexoses (D-mannose, D-galactose, and D-glucose), hexuronic acids (4-O-methyl-D-glucuronic acid, D-glucuronic acid, and D-galacturonic acid), and acetyl groups, as well as small amounts of L-rhamnose and L-fucose. These functional groups can assemble into a range of various hemicellulose polysaccharides with diverse structures from linear to highly branched, such as xylans, mannans, xyloglucans, β-1,3;1,4-glucans, and galactans.13–15 The abundance, detailed sugar composition and chemical structure of these hemicellulose polysaccharides vary widely depending on the biomass sources.

Moreover, it is extremely challenging to obtain hemicellulose from biomass without damage to the native structure due to the complex nature of hemicellulose and its crosslinking with other biomass components.11,16–20 Therefore, model substances of hemicellulose, such as monomeric building blocks of xylose, arabinose, mannose and galactose, and commercial hemicellulose polysaccharides, like galactoglucomannan, glucomannan, and especially xylans, have been widely used to provide insight into and implications for hemicellulose pyrolysis.13 However, these studies based on model substances of hemicellulose might overlook the intrinsic structural features of hemicellulose.

Recently, extracted hemicelluloses from biomass, which may be close in composition and structure to the native form of hemicellulose, have been increasingly utilized as model compounds to study hemicellulose pyrolysis.11,18–23 However, the composition and structure of hemicellulose can be greatly affected by the severity of pretreatment and extraction conditions such as temperature, concentration of alkali or acid, and residence time.18–20,24–33 As a result, the product distribution from fast pyrolysis of extracted hemicellulose subjected to different extraction treatments can be very different. Overall, the current knowledge on fast pyrolysis of hemicellulose, especially for native hemicellulose, is still very limited.

Another inherent challenge is the complex nature of the fast pyrolysis process and resulting chemical speciation. Fast pyrolysis of hemicellulose polysaccharides involves formation of hundreds of intermediates and products through multiphase reactions of competing coupled pathways, which can be affected by many factors such as feedstock types, reaction conditions (e.g. temperature, pressure and residence time), and potential catalysis by indigenous inorganic materials. Moreover, the thermal decomposition of hemicellulose simultaneously yields both condensable and non-condensable gaseous products as well as solid residues called char. The condensable gaseous products are referred to as liquid phase product or bio-oil, which includes H2O, LMW alcohols, aldehydes, acids, furans and anhydrosugars. The non-condensable gaseous products include H2, CO, CO2 and light hydrocarbons such as CH4, C2H6 and C3H8. It is quite challenging to identify and quantify the complex mixture of pyrolysis products.

Therefore, experimental studies in the literature have primarily focused on the physical measurement of mass loss and the determination of global pyrolysis products based on phases, such as gases, bio-oil (aqueous phase product and organic phase product), and solid residue (char). Peng and Wu22 revealed that the thermal decomposition of hemicellulose during fast pyrolysis can be divided into four stages: (1) physical loss of water; (2) early pyrolysis stage where dehydration and cleavage of side chains of hemicellulose occur; (3) main pyrolysis stage where the sample loses most mass via reactions such as dehydration, decarboxylation, and decarbonylation; and (4) the last stage of charring. Generally, lumped product compositions of 20–30% char, 10–20% non-condensable gas, and 40–60% bio-oil are produced from hemicellulose pyrolysis.22,34–38

Recently, the Shanks group20 performed a careful analysis of the products from fast pyrolysis of hemicellulose extracted from corn stover using an advanced micro-pyrolyzer coupled to a GC-MS/FID system. The detailed chemical speciation including 27 final pyrolysis products was more accurately identified and quantified as compared to their previous experimental work.11 They also postulated a reaction scheme of hemicellulose pyrolysis, including competing pathways of depolymerization to sugars and anhydrosugars, dehydration to furan and pyran ring derivatives, and ring-breakage of furanose and pyranose to light oxygenated species.11 Wang et al.39 proposed a reaction scheme for xylan decomposition, in which the formation of all pyrolysis products except acetic acid required an acyclic D-xylose intermediate. Shen et al.40 also proposed a reaction scheme for the formation of 1,4-anhydro-D-xylopyranose, furfural, acetone, acetic acid, formic acid, CO2, CO, and methanol from fast pyrolysis of O-acetyl-4-O-methylglucuronoxylan.

These recent studies showed substantial progress towards unraveling the fundamentals of hemicellulose pyrolysis. However, no complete reaction network at the mechanistic level was reported in these studies. Moreover, none of these proposed decomposition networks/pathways has been validated via kinetic modeling or supported by quantum chemical calculation. Therefore, there is still little quantitative information that the speculated reaction pathways can provide to confirm or help understand the reaction mechanism of hemicellulose pyrolysis.

In the absence of mechanistic understanding, the kinetic models reported in the literature for hemicellulose pyrolysis are mainly built upon global kinetic schemes. The simplest models are one-reaction/stage models, in which the formation of volatiles and char from hemicellulose is described by a single first-order reaction or two parallel reactions with different rate constants. The most sophisticated global kinetic model to date for hemicellulose pyrolysis was reported by Ranzi et al.41 They used xylose polymer (C5H8O4)n as an approximate chemical structure of hemicellulose, which formed two intermediate species that experienced successive decomposition routes, eventually leading to the formation of xylose, H2, H2O, CO, CO2, HCHO, CH3OH, C2H5OH, and char with independent kinetic parameters. Besides, the formation of secondary gaseous products of CO2, CO and H2 from lumped pseudospecies in a solid matrix or melt phase was also included to address their experimental observation of CO2, CO and H2 released at higher temperature. Moreover, stoichiometric coefficients were utilized for different products to handle the different formation rates during pyrolysis. However, their model does not include several major products such as acetic acid, furfural, anhydroxylose and dianhydroxylose, which have been experimentally measured with substantial yields (up to 50 wt% in total).

Overall, these global kinetic models are able to explain the experimental observations and promote understanding of the kinetics of hemicellulose pyrolysis to some extent, such as the degradation rate of the starting materials, mass loss curves of hemicellulose during pyrolysis, and the yields of lumped products. Besides, the use of global kinetic models for the complex pyrolysis system simplifies data collection and analysis as well as the numerical implementation, which is attractive for many practical applications.

However, there are also many other factors that are obscured by global kinetic models. For example, the global kinetic models of hemicellulose pyrolysis are restricted to the specific starting materials and operating conditions in individual studies, and therefore have poor extendability to other applications. Besides, the majority of those global kinetic models were built based on the mass loss profile from the thermo-gravimetric analysis (TGA) of hemicellulose samples. Notably, pyrolysis of hemicellulose in TGA experiments with an insufficient heating rate (reactions under low temperature) and long residence time (secondary decomposition and condensation of volatiles) would result in a significantly different product distribution compared with typical fast pyrolysis under 500 °C. Furthermore, the global kinetic model is still based on the lumping strategy, by which reacting species are grouped into major products by phase, i.e. hemicellulose, active hemicellulose, volatiles (bio-oil and gas), tar (bio-oil), gas and char. Therefore, it cannot really capture the pyrolysis chemistry and predict the bio-oil composition at the molecular level.

Recently, substantial progress towards a fundamental understanding of thermal decomposition of hemicellulose was made by Klein and co-workers,42 who developed a molecular-level kinetic model for biomass gasification which included a sub-model for hemicellulose pyrolysis. This model includes two reaction pathways, hydrolysis and thermolysis, which break down hemicellulose polysaccharides (xylan) into monomeric units (xylose and anhydroxylopyranose). These species can then form light hydrocarbons via thermal cracking, decarbonylation, and enol–aldehyde tautomerization. The model also includes the formation of char, higher molecular weight species and heavy aromatics via Diels–Alder addition and dehydrogenation. This is a breakthrough in the modeling of hemicellulose pyrolysis at the molecular-level and in understanding the gasification chemistry. However, the sub-model for hemicellulose pyrolysis oversimplified the hemicellulose structure which is modeled as a linear xylan homopolymer. Furthermore, the model focused on the gasification chemistry, many relevant pyrolysis reactions such as dehydration were not included, and also detailed reaction pathways for pyrolysis products were not documented in their work.

Additional computational efforts have been put forth to shed light on the formation of typical pyrolysis products from thermal decomposition of monomeric model compounds of hemicellulose polysaccharides. Zhang et al.43 studied eight possible pathways for decomposition of xylose, in which the pathways leading to the formation of furfural and glycolaldehyde (GA) are favored over others. Tian et al.44 studied six possible reaction pathways associated with the formation of glycolaldehyde from xylopyranose and O-acetyl-xylopyranose. Huang et al.35,45 applied density functional theory (DFT) to study the decomposition pathways of O-acetyl-xylopyranose resulting in the formation of acetic acid, acetaldehyde, GA, acetone, CO, CO2 and CH4. Later on, Wang et al.46 confirmed that the formation of acetic acid via the breakage of O-acetyl groups can easily occur because of its lower energy barrier, and they also compared the decomposition pathways of xylose, mannose, galactose and arabinose, in which the mechanistic formation of furfural (2-furaldehyde) and 5-hydroxymethylfurfural (5-HMF) was also discussed. Note that these studies have only revealed a part of the overall map for decomposition of hemicellulose to various pyrolysis products, and thus the full picture still remains obscure.

A fundamental and mechanistic understanding of hemicellulose decomposition and product formation during fast pyrolysis is crucial to guide reactor design, development of catalysts for in situ and ex situ catalytic fast pyrolysis, process optimization, and product control. Mechanistic modeling is an effective tool in elucidating reaction mechanisms and quantifying competing reaction pathways for complex reaction networks,47,48 as demonstrated in our previous work for cellulose pyrolysis.49–53 In this work, the mechanistic modeling methodology and reaction family approach that we utilized for modeling of cellulose pyrolysis were applied to build a mechanistic model for fast pyrolysis of hemicellulose. The proposed mechanistic model was first validated against experimental data of fast pyrolysis of hemicellulose extracted from corn stover by Zhang et al.20 Insights into hemicellulose pyrolysis gained from the mechanistic modeling perspective are presented. Then, this mechanistic model was further extended to simulate fast pyrolysis of native hemicellulose. Key differences between extracted and native hemicelluloses during fast pyrolysis are discussed.

2. Mechanistic modeling

To build a mechanistic model for fast pyrolysis of hemicellulose, the structures and reaction chemistry of hemicellulose starting materials, intermediates and products need to be identified, and the associated kinetic parameters need to be specified. In this work, the mechanistic model is built based on the experimental work of Zhang et al.20 for fast pyrolysis of hemicellulose extracted from corn stover via hot-water treatment. Please see the Experimental section in the ESI for detailed information on the materials, methods, and procedures of experiments on fast pyrolysis of hemicellulose pyrolysis conducted by Zhang et al.20 Briefly, the experiment was performed under kinetically-limited and mineral-free conditions using a micro-pyrolyzer-GC/MS-FID system at 500 °C.20 As shown in Table S1 in the ESI, a total of 27 pyrolysis products were identified and quantified, which included 7.14 wt% anhydroxylopyranose, 17.76 wt% dianhydroxylopyranose, 12.85 wt% glycolaldehyde, 3.31 wt% methyl glyoxal, 1.15 wt% acetaldehyde, 2.20 wt% 2-furaldehyde, 1.20 wt% acetol, 14.98 wt% H2O, 6.02 wt% CO2, 1.72 wt% CO, and 9.44 wt% char.20 The yield of acetic acid was as little as 0.18 wt%, because acetyl groups were removed from hemicellulose during the hot-water treatment, as will be discussed in more detail later. This set of experimental data reported by Zhang et al.20 was used to rationalize the pyrolysis chemistry of hemicellulose and to validate the mechanistic model.

2.1. Structural model for extracted hemicellulose from corn stover

Characterization of the extracted hemicellulose by Zhang et al.20 showed monosaccharide composition of ∼92 mol% xylose, ∼6 mol% arabinose (molar ratio of arabinose to xylose ∼1:[thin space (1/6-em)]15), and 2 mol% of others, which indicated that arabinoxylan was the primary polysaccharide of hemicellulose extracted from corn stover. This is consistent with literature findings that arabinoxylans are the major hemicellulose polysaccharides of herbaceous biomass.15,18,19,22,54 Past efforts revealed that arabinoxylans have a β-1,4-linked xylose backbone with α-1,3-linked arabinose side groups. Besides, the xylose residues have acetyl groups attached at position 2 and/or position 3 in the native state of arabinoxylans, and the degree of acetylation (the molar ratio of acetyl groups to xylose in the backbone) can range from 0.1 to 0.7 depending on the biomass source and treatment methods utilized.14,15,40,55–60 However, the abundance of acetyl groups in the extracted hemicellulose by Zhang et al.20 is too low to be experimentally quantified due to the hot-water treatment, in which sulfuric acid could initiate hemicellulose depolymerization and cleave acetyl groups, thereby leading to a low yield of acetic acid (0.18%) from fast pyrolysis.

Moreover, sulfuric acid and acetic acid in the extraction process could also catalyze further depolymerization of hemicellulose, resulting in a lower degree of polymerization (DP) of extracted hemicellulose than native hemicellulose. However, the DP of arabinoxylan and the distribution of arabinose units in arabinoxylans were not determined in the experiments by Zhang et al.20 On the other hand, it was reported that the DP of 4-O-methylglucoronoxylan from woody biomass is about 200 for hardwood and 100 for softwood.61 Sun and Tomkinson62 reported DP ranging from 70 to 130 for hemicellulose extracted from wheat straw, an agricultural residue. High performance size exclusion chromatography analysis by Gu and Catchmark63,64 revealed that the main peak of the molecular weight distribution of commercial xylan was ∼10[thin space (1/6-em)]000 Da, corresponding to DP of ∼70. Besides, Pereira et al. reported that the average DP of arabinoxylans ranges from 50–185.65

Based on the above considerations, we proposed a structure of arabinoxylan that has a β-1,4-linked xylopyranose backbone with α-1,3-linked arabinose side groups to capture the compositional and structural features of hemicellulose extracted from corn stover. As shown in Fig. 1, the proposed structure has an average DP of 75 and a molar ratio of arabinose to xylose of 1[thin space (1/6-em)]:[thin space (1/6-em)]15. Moreover, a series of structures with different DP (Scheme S1 in the ESI) but all with molar ratio of arabinose to xylose of 1[thin space (1/6-em)]:[thin space (1/6-em)]15 were examined later in this work to evaluate the effects of DP on the pyrolysis product distribution.


image file: c7ee03208k-f1.tif
Fig. 1 Proposed structure for extracted hemicellulose from corn stover.

2.2. Reaction mechanism

Past efforts demonstrated that a concerted mechanism is more kinetically favorable than radical and ionic mechanisms and offers better alignment with experimental findings of carbohydrate pyrolysis.50,51,66–69 Therefore, in this work, the mechanistic model of hemicellulose pyrolysis is based on concerted reactions. Similar to cellulose pyrolysis, hemicellulose pyrolysis also yielded anhydrosugars, dianhydrosugars, furanic compounds, gaseous products CO2 and CO, water, char, and low molecular weight species as the major products from the polymeric starting material with a β-1,4-linked xylose backbone structure. Moreover, researchers have speculated that the decomposition of hemicellulose and the formation of products follow similar reaction pathways as those of cellulose pyrolysis.11 Given the similarity in the polymeric structures of cellulose and hemicellulose, we hypothesized that hemicellulose decomposes via similar reaction pathways as those of cellulose fast pyrolysis.

Therefore, our previous modeling framework for cellulose pyrolysis49–53 was applied here to rationalize the reaction mechanism of hemicellulose pyrolysis with further incorporation of reaction pathways revealed by recent quantum chemical calculations on hemicellulose model compounds. Briefly, the structure of a chain species can be expressed as a combination of a left-end group connected with a right-end group and/or a mid-group. Reaction families such as glycosidic bond cleavage, initiation, end-chain initiation, thermohydrolysis, dehydration, and mid-chain dehydration can occur to chain species to generate new chain species with different mid-groups or end-groups with the same or reduced chain length and release low molecular weight (LMW) intermediates and products. The LMW species can then undergo further decomposition pathways such as ring-opening/closing, dehydration, retro-aldol, and char formation to form a range of final LMW products. Overall, 504 reactions of 114 species were included in the mechanistic model for fast pyrolysis of extracted hemicellulose, which included 74 reactions of 50 LMW species. More details about species and reaction chemistry are described as follows.

2.2.1. Reactions of hemicellulose chains. The reactions of hemicellulose chain species are further divided into two types based on end-groups and mid-groups. Scheme 1 depicts the concerted reactions of arabinoxylan and its derived-chain species involving end-groups. End-chain initiation of the starting material arabinoxylan can occur at either end of the hemicellulose chain, yielding a shorter chain and a dimeric intermediate species (reactions I and II). End-chain initiation can further proceed to form shorter chains with a xylose non-reducing end (reaction III) or an anhydroxylopyranose end (reaction IV), releasing dimeric intermediate species. The model includes the formation of 1,2-anhydroxylopyranose from polymer chain decomposition via end-chain initiation (reaction V) and depropagation (reaction VI). The formation of xylose via thermohydrolysis (VII) was included, which is a key reaction for the formation of glucose in cellulose pyrolysis as reported in our previous work.49–53 Once xylose is formed, it can undergo a range of reactions to form other LMW products. Dehydration of end-groups (reactions VIII and IX) that forms anhydro- and dianhydro-sugar-end chains was also included in the model. Then, end-chain initiation of these anhydro- and dianhydro-sugar-end chains further produces a shorter chain with a 1,2-anhydroxylopyranose end and dianhydroxylopyranose (reactions X and XI), a major product of hemicellulose pyrolysis.
image file: c7ee03208k-s1.tif
Scheme 1 Decomposition mechanisms of hemicellulose chains involving end-groups.

Scheme 2 depicts the concerted reactions of arabinoxylan and its derived-chain species involving mid-groups. As shown in Scheme 2, initiation (reactions XII, XIII, and XIV) can occur to any glycosidic bond on the xylan backbone, forming shorter chain species with different end-groups. Another type of reaction that occurs for arabinoxylan and its derived chains is mid-chain dehydration (XV), which is an important reaction in cellulose fast pyrolysis as reported in our previous work.50–53 This is verified by the experimental observation of formation of significant amounts of water as well as unidentified anhydro- and dianhydro-xylose products.20 Mayes et al.68 revealed that the likelihood of forming a molecule of water by losing a hydroxyl group connected with C1–C4 and HO6 from glucose via dehydration follows the order: C1 < C3 < C2 < C4 (molecular structure and atom numbering scheme of glucose are shown in Fig. S1, ESI). The xylose unit of arabinoxylan shares almost the same environment as the ring-segment of the glucose unit of cellulose. The hydroxyl group at xylose C1 (molecular structure and atom numbering scheme of xylose are shown in Fig. S1, ESI) no longer exists in the midgroups of arabinoxylan since arabinoxylan has a backbone of xylose residues connected by β-(1,4) glycosidic linkages (C1–O1–C4). Therefore, dehydration occurring within arabinoxylan chains most likely favors the formation of water by losing a hydroxyl group at C3, and thus our model includes the mid-chain dehydration (reaction XV) of xylose units forming a 3,4-anhydroxylopyranose-containing mid-chain. The chain with a dehydrated mid-group subsequently undergoes concerted midchain glycosidic bond cleavage (GBC) and yields a 1,2;3,4-dianhydroxylopyranose-end chain and a xylopyranose end chain (reaction XVI) or a 1,2-anhydroxylopyranose-end chain and a 3,4-anhydroxylopyranose-end chain (reaction XVII). The resulting chains with end-groups then undergo further decomposition as described in Scheme 1.


image file: c7ee03208k-s2.tif
Scheme 2 Decomposition mechanisms of hemicellulose chains involving mid-groups.
2.2.2. Reactions of LMW intermediates and products. Overall, the reaction network developed here is rationalized based on the experimental work of Zhang et al.,20 isotopic labeling studies of Paine III et al.,70–73 and quantum chemical calculations of Seshadri and Westmoreland66 as well as Mayes et al.,67–69 and by applying pyrolysis chemistry and reaction families of LMW species reported in our previous mechanistic modeling of cellulose pyrolysis.47,49–53Fig. 2(a) shows all the end-groups and mid-groups used for representing hemicellulose-derived polymeric chain species, while Fig. 2(b) shows the 50 LMW products tracked in the current model. Besides, Scheme 3 shows the reactions of LMW intermediates generated from decomposition of polymeric chain species and the formation of various LMW products during hemicellulose pyrolysis. Note that this mechanistic model includes the formation pathways for all the major products with mass yield >1 wt% reported in Zhang et al.20
image file: c7ee03208k-f2.tif
Fig. 2 (a) Identity of hemicellulose and derived chains based on end-groups and mid-groups; (b) list of 50 LMW species tracked in the model.

image file: c7ee03208k-s3.tif
Scheme 3 Reactions of LMW species.

2.2.2.1. Anhydroxylopyranose and dianhydroxylopyranose. As shown in reactions 1 and 3 of Scheme 3, the initiation to α-L-arabinofuranose-(1,3)-β-D-xylopyranose and α-L-arabinofuranose-(1,3)-β-D-1,2-anhydroxylopyranose yields 1,2-anhydroxylopyranose and 1,2;3,4-dianhydroxylopyranose, respectively. Zhang et al.20 observed a significant amount of anhydroxylopyranose (AXP) and dianhydroxylopyranose (DAXP) in their experiments. However, there was an experimental difficulty associated with the identification of the exact forms of AXP and DAXP. Therefore, they reported the pyrolysis products as AXP, DAXP 1, DAXP 2, other DAXP 1, and other DAXP.20 In this work, the mechanistic model not only rationalizes the reaction pathways for the formation of various AXP and DAXP species, but also implies their most likely structures. Apart from the formation of 1,2-anhydroxylopyranose through glycosidic bond cleavage reactions, it can also be produced from xylose dehydration (reaction 21). Note that both anhydroxylopyranose and dianhydroxylopyranose species in enol form could undergo enol–keto tautomerization reactions to form more stable keto tautomers.

In this model, xylose generated from chain decomposition serves as an intermediate for the formation of a range of intermediates and major products through various unimolecular reactions in the condensed medium. As shown in Scheme 3, β-D-xylopyranose in the chair conformation (4C1) can undergo dehydration to form 4,5-anhydroxylopyranose (reaction 6), 3,4-anhydroxylopyranose (reaction 13), and 1,2-anhydroxylopyranose (reaction 21). The anhydroxylopyranose species can undergo further dehydration to give dianhydroxylopyranose or break into smaller molecules like glycolaldehyde (GA), acetol, and methylglyoxal via acyclic intermediates through ring-opening routes. 4,5-Anhydroxylopyranose can undergo further dehydration to form 1,2;4,5-dianhydroxylopyranose (reaction 12) or undergo a ring-opening route to form acyclic smaller molecules through retro aldol reaction. 1,2;3,4-Anhydroxylopyranose can be formed from dehydration of 1,2-anhydroxylopyranose and 3,4-anhydroxylopyranose. Reactions 32 and 33 represent the transformations between β-D-xylopyranose and its acyclic isomer D-xylose via ring-opening/ring-closing. Quantum chemical calculations showed that the ring-opening reaction of β-D-xylopyranose to acyclic D-xylose by breaking the C1–O bond is feasible and should easily occur under pyrolysis conditions.43,44,46 Wang et al.39 reported that D-xylose and its derivatives are key intermediates for the formation of C1–C3 molecules. This is consistent with 13C isotopic labeling study by Paine III et al.71 and the DFT calculations by Seshadri and Westmoreland66 and Mayes et al.67–69 They all demonstrated that dehydration, enol–keto tautomerization and especially retro aldol were the concerted reactions by which acyclic intermediates could break down into small molecules during fast pyrolysis.


2.2.2.2. Furfural. Furfural is a major product of pyrolysis of hemicellulose and xylose as widely reported in experimental studies in the literature. Both experiments and quantum chemical calculations indicate that pentose (xylose, arabinose) tends to produce furfural, whereas hexose tends to form 5-HMF (glucose, mannose, galactose).74–76 DFT calculations reveal that the formation of furfural from xylopyranose starts with the formation of D-xylose via a ring-opening mechanism. Two major pathways have been reported for the formation of furfural from D-xylose. One pathway is D-xylose undergoing a cyclization/dehydration reaction followed by dehydration reactions to give furfural,73 and the other pathway is two-step dehydration of D-xylose at the C3 and C4 positions followed by a rate-determining cyclization/dehydration step.43,46 The first pathway is faster than the second, the as evidenced by 13C isotopic labeling study of glucose pyrolysis by Paine III et al.,71 DFT calculations of Seshadri and Westmoreland,66 and our previous work50 on cellulose pyrolysis. Therefore, the model includes the reaction pathways (32)–(40)–(41)–(42) shown in Scheme 3 for the formation of furfural from β-D-xylopyranose.
2.2.2.3. Acetol. Acetol has been reported as a product from pyrolysis of various types of biomass hemicellulose. The 13C-labeling study by Ponder and Richards77 reported that acetol was predominantly derived from contiguous terminal carbons with the acetol methyl group being a terminal carbon of the source carbohydrate. Scheme 3 provides a pathway (13)–(14)–(16)–(17) for the formation of acetol from C3–C4–C5 of xylose through dehydration, ring-opening of anhydroxylopyranose, enol–keto tautomerization of the C5 acyclic intermediate and retro aldol reaction, with concomitant formation of glyoxal (reaction 17).
2.2.2.4. Methylglyoxal. The 13C isotopic labeling study by Paine III et al. revealed that the dominant labeling patterns of C3 compounds pyruvaldehyde and methylglyoxal were derived from contiguous terminal carbons, and glyceraldehyde is invoked as an intermediate for the formation of methylglyoxal during fast pyrolysis.72 As shown in Scheme 3, pathways (32)–(34)–(35)–(36) and (32)–(46)–(47)–(48)–(36) depict the formation of methylglyoxal via the enol–keto tautomerization of pyruvaldehyde, which can be generated from D-xylose through retro aldol and dehydration reactions. The third pathway (24)–(26)–(29) represents the formation of methylglyoxal from 1,2-anhydroxylopyranose through ring-opening followed by enol–keto tautomerization and retro aldol reactions. All these three pathways involve retro aldol chemistry and co-formation of GA.
2.2.2.5. Glycolaldehyde. Glycolaldehyde is a major product of fast pyrolysis of hemicellulose. In this model, besides the formation of GA along with methylglyoxal explained above, several other routes are also included. The pathway (37)–(38) in Scheme 3 yields two molecules of glycolaldehyde from C1–C2 and C3–C4 fragments of D-xylose, respectively. It involves retro aldol reaction of acyclic D-xylose forming formaldehyde and erythrose, followed by retro aldol reaction of erythrose to form GA and 1,2-ethylene diol, and enol–keto tautomerization of 1,2-ethylene diol to GA (reaction 39). This is a well-recognized and major pathway for the formation of GA during carbohydrate pyrolysis,66,71,78 which also produces formaldehyde. Retro aldol reaction is generally very competitive under pyrolysis conditions due to its low activation energy. Another pathway (6)–(7)–(9)–(11) is minor for GA formation, starting with dehydration of xylopyranose, followed by ring-opening and enol–keto tautomerization, and ending with retro aldol reaction. The last pathway (32)–(46)–(47)–(48) involves the dehydration of D-xylose, followed by enol–keto tautomerization, and a final retro aldol step to release GA.
2.2.2.6. Acetaldehyde. Paine III et al.70 revealed that 95% of the acetaldehyde is generated by unimolecular chemistry. Reaction (45) represents the formation of acetaldehyde through retro aldol conversion of anhydroxylose, involving a C3 fragment co-product.
2.2.2.7. Formaldehyde. Past isotopic labeling studies71,79 showed that formaldehyde was primarily produced from the end-carbon of acyclic intermediates during fast pyrolysis; for example, over 65% of formaldehyde is generated from C6 of D-glucose in glucose pyrolysis. Reaction (37) represents the formation of formaldehyde from C5 via the intermediate D-xylose.
2.2.2.8. Char, CO, CO2, water, and other LMW products. As noted in our previous work for cellulose pyrolysis,49–53 the formation of char is an extremely complicated process which usually involves dehydration, cross-linking, and bimolecular condensation reactions of the volatile and condensed phase species. However, the detailed mechanistic steps leading to the formation of char have not yet been fully understood.80,81 On the other hand, experimentally it is observed that the formation of char was often associated with the formation of light gases such as CO, CO2 and H2. The previous models used global reactions which produce char (tracked as carbon, C), CO2, CO and H2 in stoichiometric proportions to approximate char formation during fast pyrolysis.41,82,83 In this work, the same simplification strategy is adopted to track char formation in hemicellulose pyrolysis through global reactions 52–74 shown in Scheme 4.
image file: c7ee03208k-s4.tif
Scheme 4 Global reactions for char formation from various dehydrated species.

2.2.2.9. C6 products. It is worth noting that the current model does not include the formation of 5-HMF, levoglucosan–pyranose and levoglucosan–furanose, which indeed have been identified and quantified by Zhang et al.20 There are two main reasons. First, our previous model on cellulose pyrolysis49–53 suggests that these C6 compounds should be predominantly converted from C6 sugars such as glucose. However, for simplification purposes, this model did not take into account minor C6 monosaccharides (composition <2%) to build the hemicellulose structure, which is predominantly responsible for the formation of C6 pyrolysis products. Secondly, those C6 products have very low mass yield (<1 wt%), out of the range considered for kinetic model build-up here. Similarly, the formation of minor furanic products was not included either.
2.2.2.10. Acetic acid. Acetic acid has been widely reported to be a major product of pyrolysis of native hemicellulose. DFT calculations on the decomposition of O-acetyl-xylopyranose by Tian et al.44 suggested that the dissociation of O-acetyl groups forming acetic acid is kinetically favored. Besides, it has been speculated that the cleavage of acetyl groups attached to the backbone of hemicellulose polysaccharides is the primary route for the formation of acetic acid from hemicellulose pyrolysis based on many experimental studies.35,39,44,46,84–87

However, the pyrolysis of extracted hemicellulose by Zhang et al.20 yielded only 0.18% acetic acid, most likely due to the removal of acetyl groups from hemicellulose during acidic hot-water treatment. Although it has been reported that there might be other routes for the formation of acetic acid via ring-opening of monosaccharides (intermediates or starting materials),39,46,88 there are no detailed reactions and kinetics available in the literature. Moreover, experimental data have indicated that this route may require a higher activation energy than the primary route.46 Therefore, the formation of acetic acid was not included in the model for extracted hemicellulose.

Finally, it is worth noting that: (1) biomass pyrolysis is extremely complex, and many minor products are still not identified and quantified by existing techniques, and therefore are not included in the model; (2) in order to manage the model complexity, the current model has not included mechanistic pathways for species with mass yield <1 wt% from Zhang et al.;20 (3) the current model does not necessarily cover the complete reaction chemistry for existing species, and even not all the reactions are fully mechanistic, for example, char formation. (4) However, the mechanistic model developed here is general and flexible enough to incorporate any additional details whenever new mechanistic understanding emerges.

2.3. Specification of reaction rate constants

Specification of reasonable reaction rate constants is vital for the development of a kinetic model at the mechanistic level.47 According to the reaction family approach,89 the kinetic parameters for hemicellulose pyrolysis, involving glycosidic bond cleavage, dehydration, ring-opening/closing, retro aldol, enol–keto tautomerization and char formation, were adapted from our previous modeling work of cellulose pyrolysis49–53 without or with limited adjustment. These rate parameters were primarily derived from quantum chemical calculations of Mayes et al.69 and Seshadri and Westmoreland,66 which had been validated for fast pyrolysis of glucose, cellobiose, maltohexaose and cellulose over the temperature range 400–600 °C.49–53 Note that, however, activation energy (Ea) and frequency factor (A) values estimated using quantum chemical calculations have uncertainties. For example, the mean unsigned error associated with the B3LYP hybrid functional for activation barriers has been shown to be 4.8 kcal mol−1.90

Table 1 shows the rate parameters for different reaction families as well as their references. Almost all reaction rate constants are cited from literature, while only a couple are fitted to capture the product yields. It should be pointed out that the same parameters are used for all reactions in a particular reaction family in order to reduce the number of parameters for estimation.

Table 1 Arrhenius rate parameters of different reaction types used in fast pyrolysis of hemicellulose
Reaction type Reaction indexa A, s−1 or M−1 s−1 E a , kcal mol−1 Ref. A Ref. Ea
a Reaction indices refer to specific reactions in Schemes 1–4. b Activation energies are based on quantum chemical calculations, unless otherwise mentioned. c A and Ea are experimentally determined.
Hydrolysisc VII 1.00 × 1014 34 93 93
Initiation XII, XIII, XIV, 1, 3 2.00 × 1015 53.5 67 67
End-chain initiation I, II, III, IV, V, X, XI 2.00 × 1015 51.5 67 67
Depropagation VI 1.50 × 1015 51.5 67 67
Mid-chain GBC XVI, XVII 2.00 × 1014 53.5 67 67
Dehydration VIII, IX, XV 5.00 × 1015 57.0 50 68
2, 12, 13,19, 21, 22 5.00 × 1015 57.0 50 68
4, 41 7.75 × 1012 38.9 68 68
5, 42 5.00 × 1015 56.5 68 68
6, 30 6.00 × 1015 59.0 50 68
10, 18, 27, 40, 43, 46, 49 6.00 × 1014 50.5 68 68
35 1.00 × 109 29.0 92 92
Ring-opening/closing 7, 14, 24 9.00 × 1012 46.1 Fit 68
8, 15, 25 2.00 × 1010 42.1 Fit 68
32 3.00 × 1013 47.0 68 68
33 7.00 × 1010 34.0 68 68
Retro aldol 11, 17, 34, 38, 45, 48, 51 5.00 × 1011 39.2 66 66
29, 37 8.00 × 1011 39.7 66 66
Enol–keto tautomerization 9, 16, 20, 23, 26, 28, 31 1.00 × 1013 46.8 69 69
44, 47, 50 1.00 × 1013 46.8 69 69
36, 39 5.00 × 1015 57.0 66 66
Char formationc 52–74 6.50 × 1010 40.0 83 83


The Ea values of all the concerted chain reactions that involve glycosidic bond cleavage (GBC) are based on DFT calculations of Mayes and Broadbelt.67 For example, the model used Ea of 53.5 kcal mol−1 and A of 2 × 1015 s−1 for all of the initiation reactions. These values are in line with the reaction barrier of 52 ± 1 kcal mol−1 for the GBC of methyl β-D-glucoside reported by Hosoya et al.91 Kinetic parameters of 1,2-dehydration of glucopyranose were applied here to 1,2-dehydration of xylopyranose. The model used an Ea of 57 kcal mol−1 and A of 5.0 × 1015 s−1 for 1,2-dehydration reactions. The same kinetic parameters are also applied to similar dehydration reactions occurring within carbohydrate chains (reactions VIII and IX). A higher Ea of 59 kcal mol−1 is used for reactions 6 and 30 since the dehydration of xylopyranose by losing the hydroxyl group attached to C4 is more difficult. While the dehydration of acyclic species requires a much lower activation energy, Mayes et al.68 reported an Ea of 50.5 kcal mol−1 for 1,2-dehydration of D-glucose, which had been applied to the acyclic C6 and C5 species in our previous model. Here, we applied the same Ea of 50.5 kcal mol−1 and A of 6.0 × 1014 s−1 for 1,2-dehydration of acyclic C5 species. This is also consistent with the Ea of ∼54 kcal mol−1 for dehydration of D-anhydroxylose reported by Zhang et al.43 Lower barriers are also involved in the dehydration to a furan ring. The kinetic parameters for the dehydration of β-D-fructose to 5-HMF reported by Mayes et al.68 have been applied to similar dehydration reactions leading to the formation of species with a furan ring (reactions 4, 5, 41, and 42), such as furfural. Ea and A for the dehydration of a C3 fragment (reaction 35) are based on DFT calculations of Shen et al.92 The Ea values of ring-opening/closing of anhydroglucopyranose were applied here to anhydroxylopyranose without adjustment, while the A values were fit to match experimental data reported by Zhang et al.20 The interconversion of β-D-xylopyranose and D-xylose utilized the same kinetic parameters of ring-opening/closing of β-D-glucopyranose as in our previous model. The kinetic parameters for various retro aldol reactions are based on work of Seshadri and Westmoreland.66 They revealed that retro aldol condensation is a favorable unimolecular reaction due to its low activation energy ranging from 30 to 40 kcal mol−1. Ea of 39.2 kcal mol−1 and A of 5.0 × 1011 s−1 were used for retro aldol of acyclic C5 species, breaking into C2 and C3 species. Values of Ea of 39.7 kcal mol−1 and A of 8.0 × 1011 s−1 were used for retro aldol reactions forming methylglyoxal (reaction 29) and formaldehyde (reaction 37) directly from acyclic C5 species. The kinetic parameters for tautomerization reactions are based on quantum chemical calculations of Seshadri and Westmoreland66 and Mayes et al.69 The model utilized Ea of 46.8 kcal mol−1 and A of 1.0 × 1013 s−1 for tautomerization of C5 species and Ea of 57.0 kcal mol−1 and A of 5.0 × 1015 s−1 for tautomerization of smaller compounds. For thermohydrolysis and char formation, the rate parameters estimated through experiments83,93 were utilized.

2.4. Modeling framework and solution approach

Sections 2.1–2.3 have specified all polymeric chains and LMW species, as well as detailed reaction networks and their associated rate coefficients, forming a complete reaction mechanism for hemicellulose pyrolysis. The next step in the modeling framework is to generate mass balance equations based on the complete reaction network and accompanying approximations, which will be solved numerically to predict species profiles. In this model, the starting hemicellulose structure, pyrolysis temperature, and reaction residence time are the key independent input variables to predict all details for fast pyrolysis of hemicellulose.

The approach of continuous distribution kinetics94–98 was adopted here to describe and simplify the decomposition dynamics of polysaccharide chain species. The concentration of polymer chains is modeled as a function of chain length and pyrolysis time, resulting in integral-differential rate equations.94,95 Moment operations were employed to simplify and transform them into ordinary differential equations (ODEs).94,96,97

Based on the reaction mechanism in Schemes 1 and 2, the model includes three different types of reactions for hemicellulose and derived polymeric chains, (i) unimolecular random cleavage of polymeric chains, (ii) unimolecular unzipping of LMW species from the chains, and (iii) bimolecular thermohydrolysis. For example, the initiation reaction XIV in Scheme 2 falls under category (i).

 
image file: c7ee03208k-t1.tif(1)
As represented in eqn (1), this reaction involves the cleavage of a glycosidic bond at any random position in the arabinoxylan backbone C of chain length x, resulting in the formation of a shorter chain terminated by a 1,2-anhydroxylopyranose (AXP) end CAXP of chain length xx′, and another polymer chain C′ of chain length x′. The rate of this reaction is linearly dependent on chain length, kinit(x) = kinit × x.98 The even distribution of products along all x′ ≤ x is given by the random stoichiometric kernel, ΩRand(x′,x) = 1/x.98 The rate equations, after applying moments on the initial population balance equation, are given as follows,
 
image file: c7ee03208k-t2.tif(2)
 
image file: c7ee03208k-t3.tif(3)
where the moments C(0), C(1) and C(2) denote molar concentration, mass concentration, and second moment, respectively. Number average (Mn) and weight average molecular weight (Mw) of hemicellulose and derived chain species are given by C(1)/C(0) and C(2)/C(1), respectively. The Saidel–Katz approximation was utilized to close the moments for the third moment C(3), as shown in eqn (4).95
 
image file: c7ee03208k-t4.tif(4)
The rate equations associated with all the reactions of the polymer chains have been described in our previous work,49 while detailed derivation of these equations can be found elsewhere.94,95

All the LMW species were assumed to react according to mass action kinetics with rate laws based on elementary steps and rate coefficients calculated using the Arrhenius equation, k = A·exp(−Ea/RT).

A semi-batch reactor model that tracks the reduction in volume of the melt phase caused by the vaporization of LMWPs was established, in which various polymeric chains, dimeric species, sugars as intermediates and char were accounted for in the melt phase, while all other LMWPs were assumed vaporized into the vapor phase at ∼500 °C. The modeling codes were developed using programming languages Perl and C++ to automatically generate the differential and associated algebraic equations, which were solved using DDASAC numerical algorithms.99 Please see the ESI for detailed information and structure of the codes.

3. Results and discussion

3.1. Model validation for fast pyrolysis of extracted hemicellulose

To validate the mechanistic model, the yields of pyrolysis products predicted by the model are compared with experimental results from fast pyrolysis of hemicellulose extracted from corn stover involving hot-water treatment, which was reported by Zhang et al.20

As shown in Fig. 3, the model closely matched the experimental yields of major pyrolysis products reported by Zhang et al.,20 including acetaldehyde, methyl glyoxal, glycolaldehyde, acetol, 2-furaldehyde, dianhydroxylopyranose, anhydroxylopyranose, char, CO, CO2 and H2O. However, it does not capture well C6 and some other minor products with mass yield <1 wt%, as shown in Table S1 in the ESI. Zhang et al.20 pointed out that there were a couple unidentified compounds, AXP, DAXP1 and DAXP2, showing at relatively high yields of 7.14 ± 0.62 wt%, 2.83 ± 0.23 wt% and 13.34 ± 0.20 wt%, respectively. Our mechanistic model implies that AXP, DAXP1 and DAXP2 are most likely the keto forms of 1,2-anhydroxylopyranose (1,2-AXP), 1,2;4,5-dianhydroxylopyranose (1,2;4,5-DAXP) and 1,2;3,4-dianhydroxylopyranose (1,2;3,4-DAXP), respectively.


image file: c7ee03208k-f3.tif
Fig. 3 Comparison of model yields with experimental results of fast pyrolysis of extracted corn stover hemicellulose at 500 °C.

The strikingly good prediction for hemicellulose pyrolysis based on reaction families and their associated kinetic parameters from cellulose pyrolysis49–53 verifies our hypothesis that hemicellulose decomposes via similar reaction pathways as cellulose during fast pyrolysis. This not only further testifies to the extendibility of the mechanistic model, but also demonstrates that the rate parameters utilized for the various reactions are robust enough to capture the fundamental kinetics for fast pyrolysis of biomass carbohydrates.

3.2. Effect of DP of extracted hemicellulose

The structure, composition, DP and PDI of hemicellulose extracted from biomass vary widely depending on the biomass source, extraction treatment and analysis method.13,18,19 To study the effects of DP and structure of extracted hemicellulose on the pyrolysis product distribution, we proposed a series of different structures for extracted hemicellulose from corn stover, as shown in Scheme S1 in the ESI. Modeling results based on different initial DPs are summarized in Table S1 in the ESI.

Fig. 4 depicts the comparison of experimental data reported by Zhang et al.20 with model yields of major products from fast pyrolysis of hemicellulose at 500 °C with varying initial DP ranging from 15 to 135. Note that the initial PDI of hemicellulose was set to 1.58. It was shown that the yields of acetaldehyde, acetol, furfural, 1,2;4,5-DAXP, CO, CO2, and char are much less affected by the change in initial DP of hemicellulose than that of 1,2;3,4-DAXP, 1,2-AXP, water, methylglyoxal, and GA. This is because the decomposition of hemicellulose chains via end-chain initiation and dehydration can directly form 1,2;3,4-DAXP, 1,2-AXP and water, while GA and methylglyoxal were indirectly affected by the formation of 1,2-AXP and water, which affects the formation of xylose through thermohydrolysis. Moreover, it was shown that the yields of all these major products were only slightly changed as the initial DP was reduced from 135 to 30, while substantially changed when DP was further reduced to 15, especially for 1,2;3,4-DAXP, 1,2-AXP, GA and methylglyoxal. The model with an initial DP of 15 yields the largest deviation from the experimental results reported by Zhang et al.,20 indicating that the DP of 15 fails to represent the structure of extracted hemicellulose.


image file: c7ee03208k-f4.tif
Fig. 4 Comparison of model yields with experimental results of fast pyrolysis of extracted corn stover hemicellulose with different DP at 500 °C.

Varying the structure by changing the position of α-1,3-linked L-arabinose appearing on the xylan backbone does not have a significant effect on the pyrolysis product distribution. Fig. 5 shows the evolution of the overall weight loss of the initial samples during fast pyrolysis. The overall weight loss profiles correspond to the evolution of the yield of the total melt phase species, including unconverted hemicellulose polymer chains and their derived dimeric and sugar intermediates, and char. It can be seen that the total yield of melt phase species reached a constant value within 3 s, corresponding to complete thermoconversion of all the carbohydrates and their derived intermediates. Although the kinetics of individual pathways that lead to the formation of LMWPs respond differently, the reduction in initial DP of hemicellulose from 135 to 30 shows a negligible effect on altering the kinetics of the overall mass loss during fast pyrolysis, while hemicellulose with a DP of 15 deviates from the other entries more noticeably. Again the modeling results indicate that the DP of extracted hemicellulose does not have significant effect on the pyrolysis product distribution when it is greater than 30. It also reflects that the mechanistic model provides a much deeper understanding of the molecular speciation than lumped models that are built based on the overall mass loss kinetics.


image file: c7ee03208k-f5.tif
Fig. 5 Time evolution of overall mass loss predicted by the model of hemicellulose with different initial DP during fast pyrolysis at 500 °C.

3.3. Effect of polydispersity index (PDI)

A range of 1.0–2.0 has been reported for the polydispersity index (PDI) of xylans from various sources in the literature. For example, Wang et al.18 reported PDI values of 1.58 and 1.78 for native hemicellulose from agricultural corn stalk and rice straw, respectively. PDI of hemicelluloses isolated by Sun et al. from wheat straw were in the range of 1.34–1.91.100 The PDI of minimally damaged hemicellulose extracted from woody biomass ranges from 1.5 to 2.0. Simkovic et al. reported xylan obtained from beech sawdust has a PDI value of 1.6,101 while a commercial sample has a much lower PDI of 1.06.63,64

The effects of PDI of hemicellulose on its product distribution from fast pyrolysis were examined by varying the initial PDI value of extracted hemicellulose from 1.0 to 2.5. The results are summarized in Table S2 in the ESI. As seen in Fig. 6, varying the PDI of extracted hemicellulose with the initial DP set to 75 gives similar pyrolysis product distributions. Although the yields of GA, 1,2;3,4-DAXP, 1,2-AXP, and water are more sensitive to the PDI of hemicellulose than other products, overall, the yields of all the products change subtly. This indicates that the initial PDI value of extracted hemicellulose does not affect the pyrolysis product distribution significantly.


image file: c7ee03208k-f6.tif
Fig. 6 Comparison of model yields with experimental results of fast pyrolysis of extracted corn stover hemicellulose with different initial PDI at 500 °C.

Note that experiments of Zhang et al.20 did not measure the DP and PDI of the extracted hemicellulose due to experimental difficulties. However, model predictions show that the yields of pyrolysis products are not sensitive to the DP and PDI, which in turn suggests that the experimental accuracy in measuring the DP and PDI of hemicellulose or the severity of extraction that affects the DP or PDI of extracted hemicellulose would not matter in terms of the yields of pyrolysis products.

3.4. Effect of chain decomposition into monomeric, dimeric, and trimeric intermediates

Here we propose a strategy that further simplifies the decomposition of polymeric chain species by directly yielding dimeric or trimeric intermediates, which subsequently lead to monomeric and LWM intermediates and products. The effects of this strategy on the product distribution of fast pyrolysis were explored by modeling fast pyrolysis of extracted hemicellulose with DP = 45 and DP = 135 (Scheme S2 in the ESI), and the results are shown in Fig. 7, with more details listed in Table S3 in the ESI. As seen, the product distributions via a mechanism involving hemicellulose decomposition into monomeric, dimeric, or trimeric intermediates are quite similar. This suggests that the strategy of decomposing chain species into dimeric or trimeric intermediates maintains the modeling fidelity while significantly reducing the model complexity and numerical challenge.
image file: c7ee03208k-f7.tif
Fig. 7 Comparison of model yields with experimental results of fast pyrolysis of extracted corn stover hemicellulose based on models that handle differently the decomposition of chain species into monomers, dimers, and/or trimers.

3.5. Products profiles and reaction time scale of hemicellulose pyrolysis

Apart from predicting the final yields of the pyrolysis products, one of the advantages of a mechanistic model is that it is able to track the evolution of the product distribution of individual species and lumped fractions of interest during the entire duration of fast pyrolysis. As shown in Fig. 8, the mass fraction of polymeric species sharply decreased at the initial stage of fast pyrolysis, leading to a rapid accumulation of sugar intermediates and di-/anhydrosugars. The yield of sugar intermediates, dominated by xylose, reached a maximum of 30.9 wt% at 1.3 s via thermohydrolysis, and then underwent a sharp decrease due to their consumption reactions. Concomitantly, the resulting products of C2–C4 LMWPs, furanic compounds, gases and char started to increase at a high rate. The yield of the di-/anhydrosugar lump, which included anhydrosugars, dianhydrosugars and dehydrated sugars, continued increasing mainly due to the dehydration of xylose and reached its peak value at 2.3 s as sugar intermediates and polymeric species were consumed. This suggests that the fast pyrolysis should be terminated before 2.3 s for the purpose of maximizing the yield of sugars or anhydrosugars from pyrolysis. The decreased yield of anhydrosugar intermediates leads to the continued formation of C2–C4 LMWPs via retro aldol reactions and the increased yields of gases and char through char formation reactions.
image file: c7ee03208k-f8.tif
Fig. 8 Time evolution of lumped fractions including polymeric species, sugar intermediates, anhydro- and dianhydrosugars, furanic compounds, gases, char, and C2–C4 LMWPs during fast pyrolysis of extracted hemicellulose at 500 °C.

Looking into the evolution dynamics of individual products during fast pyrolysis (Fig. 9), it can be observed that 1,2-AXP was produced at the highest rate and reached its peak value within 1 s of pyrolysis, indicating that 1,2-AXP was primarily generated from chain decomposition and also that the formation of 1,2-AXP during fast pyrolysis was facile. The decrease in yield of 1,2-AXP was attributed to dehydration to dianhydroxylopyranose and retro aldol reaction forming GA and methylglyoxal. Competing with these consumption pathways is the tautomerization of the enol form of 1,2-AXP into the more stable keto isomer as evidenced by the plateau of concentration profile after 3.5 s. As seen in Fig. 8, 1,2;3,4-DAXP showed a sharp increase in yield before 2.3 s, because it can also be directly produced from decomposition of hemicellulose chains besides dehydration of intermediate xylose. Intermediate xylose was formed mostly via thermohydrolysis and then converted to its major final products, such as furfural, GA, acetaldehyde, char, CO and CO2, leading to the rapid increase for yields of these corresponding products within 1–3 s. The yields of 1,2-AXP and 1,2;3,4-DAXP slightly changed after 3.5 s, and at the same time, the yields of 1,2;4,5-DAXP, furfural, acetol, and acetaldehyde increased slowly. The yields of char and gases continued to increase as reaction proceeded, which suggests that the fast pyrolysis should be terminated at 3.0 s in order to achieve the maximum yield of bio-oil and to mitigate the secondary degradation and charring of bio-oil into gases and char. Most LMWPs reached the plateau within 4.3 s. Therefore, a time taken for complete conversion of hemicellulose during fast pyrolysis at 500 °C corresponds to 4.3 s because of complete conversion, which suggests that the fast pyrolysis should be terminated at this point to reduce operating cost. Overall, the learning from these species' profiles reflects another advantage of the mechanistic modeling that provides information and insights useful for process engineers to better design and control the pyrolysis process.


image file: c7ee03208k-f9.tif
Fig. 9 Time evolution of various LWMPs during fast pyrolysis of extracted hemicellulose at 500 °C.

4. Extension to native hemicellulose

4.1. Mechanistic modeling for fast pyrolysis of native hemicellulose

Another feature of the mechanistic model is that the reaction mechanism is extendable. The excellent agreement of the model results with experimental data20 for fast pyrolysis of extracted corn stover hemicellulose motivates a confident extension of the mechanistic model to simulate fast pyrolysis of native cornstalk hemicellulose. This is essentially building up a new mechanistic model for fast pyrolysis of native hemicellulose but built on the same fundamental framework.

Just like extracted hemicellulose, the first step is to construct a structural model for native corn stover hemicellulose that meets the characterization properties, such as monosaccharide composition, molecular weight, and PDI. The properties used for reconstruction of native hemicellulose structure are from Wang et al.18 on characterization of minimally damaged corn stalk hemicellulose. They reported that using the neutral solvent dimethyl sulfoxide (DMSO) to extract hemicellulose from cornstalk is able to retain the acetyl groups on hemicellulose. They also observed that the resulting extracted hemicellulose has a more branched structure and is closer to native hemicellulose, and pyrolysis of the so-called “minimally damaged corn stalk hemicellulose” in TGA produced 17.35 wt% acetic acid.

The composition of monosaccharides for native corn stover hemicellulose was characterized by Wang et al.18 and are reproduced in Table S4 in the ESI. Briefly, the ratio of xylose to arabinose is about 7.35, the ratio of xylose to 4-O-methyl-glucuronic acid (UA) which is attached on position 2 of xylose units is about 6, the ratio of xylose to acetyl groups is about 4, and acetyl groups are attached on either position 2 or 3 of a xylose unit.

The average molecular weight of native hemicellulose is reported to be 37[thin space (1/6-em)]345 Da. Since our modeling results of extracted hemicellulose suggest that values of DP 30 or higher have an insignificant effect on the pyrolysis product distribution, we reduced the DP of the proposed structure of native hemicellulose to simplify the mechanistic model without compromising the model fidelity. Finally, we proposed a structure for native hemicellulose from cornstalk with a DP of 30 that matches the experimental monosaccharide composition,18 as seen in Fig. 10.


image file: c7ee03208k-f10.tif
Fig. 10 Proposed structure of native hemicellulose consists of 4 arabinose, 5 UA, 7 acetyl (attached to positions 2 and 3), and 30 xylose units.

Different from extracted hemicellulose shown in Fig. 1, both UA and acetyl groups are linked to the backbone of native hemicellulose structure, in addition to an increased abundance of arabinose. This is consistent with the findings reported by Brillouet et al.102 that acid hydrolysis led to a reduction in the amount of arabinose and uronic acid in hemicellulose extracted from wheat bran. In comparison with the reported native 4-O-methyl-glucuronoxylans in the literature, experiments by Wang et al.18 did not detect any ferulic acid groups, by which arabinoxylans can be covalently cross-linked with lignin.103–105 Therefore, the proposed structure of native hemicellulose does not contain ferulic acid groups. In fact, ferulic acid is a phenolic group and thus should be considered as a building block of lignin rather than of hemicellulose.

The next step is to rationalize the reaction mechanism and specify the rate parameters of each reaction. Based on the proposed structure, we applied reaction families (i.e. initiation, end-chain initiation, thermohydrolysis, dehydration, ring-opening/closing, retro aldol and enol–keto tautomerization) and their associated kinetic parameters to create a mechanistic decomposition network for fast pyrolysis of native hemicellulose. Furthermore, we used the strategy of decomposition of chain species into trimeric and dimeric intermediates to reduce the model complexity. The final mechanistic model included 315 reactions of 134 species, in which reactions of 102 LMW species were included. More details are referred to in Schemes S3–S8 (ESI).

Finally, the same computational framework based on continuous distribution kinetics was applied to build up the model and solve it for fast pyrolysis of native hemicellulose.

4.2. Model results of fast pyrolysis of native hemicellulose

Here we report the modeling results and explore additional effects of the abundance of acetyl groups and the structure of xylan on the product yields. Unfortunately, there is no detailed yields of products from experiment to compare with our modeling results. The main purpose of modeling native hemicellulose is to achieve a better understanding of fast pyrolysis of native hemicellulose and obtain more insights and information through our mechanistic modeling into the effect of structure on the product distribution.
4.2.1. Effects of the abundance of acetyl groups on the model yield of acetic acid. The abundance of acetyl groups on so-called “native hemicellulose” obtained in experiments is sensitive to the extent of pretreatment and the extraction process.18–20,25–30 In the native state, the xylose residues of heteroxylan backbones have acetyl groups attached at position 2 and/or position 3, and the degree of acetylation (the molar ratio of acetyl groups to xylose in the backbone) can range from 0.1 to 0.7 depending on the biomass source and treatment methods utilized.14,15,40,55–60 To understand the effects of the abundance of acetyl groups on the yield of acetic acid and other products, structures with varying amounts of acetyl groups were studied, shown in Scheme S9 in the ESI.Table 2 summarizes the modeling results.
Table 2 Comparison between model yields of acetic acid from hemicellulose with varying abundance of acetyl groups and experimental data
Acetyl groups on hemicellulose (structures referred to Scheme S9, ESI) Model yield of acetic acid, wt% Experimental yield of acetic acid, wt%
Wang et al.18 Zhang et al.20
7 acetyl to 30 xylose (Scheme 1) 9.91
11 acetyl to 31 xylose (Scheme 2) 15.33 ∼13 17.8
14 acetyl to 30 xylose (Scheme S9, ESI) 19.77


As shown, a greater abundance of acetyl groups attached on the hemicellulose backbone results in a higher model yield of acetic acid. This is consistent with the experimental observation that fast pyrolysis of extracted hemicellulose yields very little acetic acid, because the pretreatment and extraction processes cleave acetyl groups off the hemicellulose structure, while fast pyrolysis of native biomass and native hemicellulose yields significant amounts of acetic acid. Experimental comparison also shows that the hemicellulose structure with 11 acetyl groups attached on 31 xylose units gives the best match with the experimental yields of acetic acid from fast pyrolysis of native hemicellulose. Therefore, we adopted this structure of hemicellulose, seen in Fig. 11, for further analysis. Note that the yield of acetic acid from Wang et al.18 was obtained from TGA experiments while the yield of acetic acid from Zhang et al.20 was calculated from the pyrolysis of native holocellulose from corn stover.


image file: c7ee03208k-f11.tif
Fig. 11 Native hemicellulose structure consists of 4 arabinose, 5 UA, 11 acetyl (attached to positions 2 and 3), and 31 xylose units.
4.2.2. Effects of the structure of native hemicellulose on model yields of pyrolysis products. The O-acetyl groups could be attached through ester linkages at the C2 and/or C3 positions of xylose units of heteroxylans. Past characterization efforts for acetyl groups on hemicellulose have been limited to measuring the abundance of acetyl groups in hemicellulose by calculating the acetic acid yield from acid hydrolysis of hemicellulose samples. Only a few studies have targeted the characterization of positions where acetyl groups are attached to hemicellulose, and the relative amounts of different positions especially have yet to be thoroughly explored. For instance, some researchers reported that O-acetyl linkages are more frequently encountered at the C3 position than at the C2 position of xylose units of xylans,18,106,107 while others have speculated that the majority of O-acetyl groups are on the C2 position rather than the C3 position.40,108

To examine the effects of hemicellulose structure on model yields of pyrolysis products, a series of different structures of native hemicellulose xylan were proposed by changing the repeating units (structures V3 and V4 in Scheme S10, ESI) and the position on which the acetyl groups are attached on xylose units (structures V3, V5, and V6 in Scheme S11, ESI). These structures (V3–6) have the same monosaccharide composition, but come up with different product distributions during fast pyrolysis as summarized in Table S5 in the ESI.

The modeling results show that changing the repeating units from xylose–acetylxylose to acetylxylose–xylose does not have a significant effect on the yields of products (Fig. 12). However, the position where acetyl groups are attached to xylose does affect the yields of products, especially for 1,2:3,4-DAXP and acetol, as shown in Fig. 13. More acetyl groups attached on the C3 position yield more 1,2;3,4-DAXP and acetol. Because more acetyl groups attached at this position lead to increased formation of the intermediate 3-acetyl-xylose, decomposition of which would result in the formation of 1,2;3,4-DAXP and acetol, as shown in Scheme S6 in the ESI. Therefore, the relative amount of acetyl groups attached at C2 positions vs. C3 positions of xylose units is a critical structural feature of hemicellulose polysaccharides that should be carefully addressed in future characterization studies.


image file: c7ee03208k-f12.tif
Fig. 12 Comparison of model yields of major products from fast pyrolysis of native hemicellulose with different repeating units.

image file: c7ee03208k-f13.tif
Fig. 13 Comparison of model yields of major products from fast pyrolysis of native hemicellulose V3 (acetyl groups attached to positions 2 and 3), V5 (all the acetyl groups attached to position 3), and V6 (all the acetyl groups attached to position 2).
4.2.3. Comparison of fast pyrolysis of native hemicellulose with extracted hemicellulose. Fig. 14 shows the comparison between model results of fast pyrolysis of extracted hemicellulose and native hemicellulose. As seen, pyrolysis of native hemicellulose yielded significantly more acetic acid, char and gaseous products, but less anhydroxylopyranose and dianhydroxylopyranose.
image file: c7ee03208k-f14.tif
Fig. 14 Comparison of model yields of major products from fast pyrolysis of extracted corn stover hemicellulose and native hemicellulose from cornstalk.

This is because native hemicellulose has a significant amount of acetyl groups, which lead to the formation of acetic acid during fast pyrolysis. On the other hand, the extraction treatment removes the acetyl groups on extracted hemicellulose. Meanwhile, the cleavage of side chains of native xylan hemicellulose forms LMW products and anhydroxylopyranose units in the xylan chains. Further decomposition of those xylan chains with anhydroxylopyranose units leads to the formation of more anhydroxylopyranose/dianhydroxylopyranose-containing species, which eventually lead to the formation of more char than in the case of extracted hemicellulose. Furthermore, there is less water produced from native hemicellulose, because more anhydroxylopyranose/dianhydroxylopyranose units are generated from side-chain cleavage and glycosidic bond cleavage rather than dehydration reactions, as compared to extracted hemicellulose. This is due to the fact that there are fewer active sites for dehydration to occur to native hemicellulose-derived chains and species.

Finally, it should be pointed out that the current model of native hemicellulose xylan covers only 80% of the monosaccharide composition, ignoring galactose, mannose and glucose, which have been reported to be present in native hemicellulose.18,109,110 Future efforts on modeling other hemicellulose polysaccharides, including β-1,3;1,4-glucans, xyloglucans, mannans, galactomannan and galactoglucomannan, are necessary to achieve more complete understanding of hemicellulose pyrolysis. In fact, the modeling framework developed here could be easily extended to model fast pyrolysis of other hemicellulose structures.

5. Summary

In this work, we have developed, to the best of our knowledge, the first mechanistic model of fast pyrolysis of hemicellulose. The model includes a hemicellulose structure that was developed based on the experimentally characterized monosaccharide composition and associated linkages of extracted hemicellulose. The reaction pathways for the decomposition of hemicellulose and the formation of various pyrolysis products were rationalized based on the reaction family approach. The hemicellulose model tracks 114 species consisting of a wide range of hemicellulose-derived polymeric chains and LMWPs in 504 individual reactions, which include decomposition reactions of hemicellulose chains via initiation, depropagation, end-chain initiation, thermohydrolysis, dehydration, and the formation of LMW products via ring opening/closing, dehydration, retro aldol, and enol–keto tautomerization. Reaction rate constants were specified by Arrhenius parameters, and are mainly from our previous work of mechanistic modeling of cellulose pyrolysis.49–53 A computational framework based on continuous distribution kinetics was constructed to solve the mechanistic model.

The model is validated by comparing model yields of pyrolysis products with experimental data from fast pyrolysis of hemicellulose extracted from corn stover at 500 °C reported by Zhang et al.20 Strikingly good agreement verifies our hypothesis that hemicellulose decomposes via similar reaction pathways as cellulose during fast pyrolysis. Model results showed that the initial DP of extracted hemicellulose should be ≥30 to capture well the structural features and pyrolysis chemistry, while varying the initial PDI was showed to have a negligible effect on the pyrolysis product distribution.

It should be emphasized that the mechanistic model not only can provide information and insights at the molecular level that a lumped kinetic model cannot provide and that are difficult to obtain through experiment, such as dynamic product profiles and pyrolysis time-scale, but also can provide useful information to guide pyrolysis reactor design and process control.

The mechanistic model of fast pyrolysis of extracted hemicellulose from corn stover was extended to simulate fast pyrolysis of native hemicellulose from corn stalk. Comparison of the model results showed that fast pyrolysis of native hemicellulose xylan yielded more char, gaseous species, acetol, and much more acetic acid than extracted hemicellulose, while yielding lower formation of 1,2-AXP, 1,2;3,4-DAXP and GA. The model results showed that hemicellulose with more acetyl groups yields more acetic acid as product, which confirmed that biomass pretreatment altering hemicellulose structures could have a significant effect on the pyrolysis product distribution. Moreover, the relative amounts of acetyl groups attached at C2 positions vs. C3 positions of xylose units can greatly affect the pyrolysis product distribution.

Finally, this work not only testifies to the robustness and extendibility of the mechanistic model, but also demonstrates that the rate parameters utilized for the various reactions are robust enough to capture the fundamental kinetics for fast pyrolysis of biomass carbohydrates. The modeling framework can be further applied to unravel the decomposition kinetics and mechanism for fast pyrolysis of other hemicellulose polysaccharides, such as galactoglucomannan, xyloglucans and β-1,3;1,4-glucans, allowing for future advancement towards a mechanistic understanding of fast pyrolysis of native hemicellulose and even whole biomass.

Conflicts of interest

The authors declare no competing financial interest.

Acknowledgements

The authors are grateful for the financial support from ExxonMobil Research and Engineering Company, National Science Foundation (CBET-1435228), and the Institute for Sustainability and Energy at Northwestern (ISEN). The authors thank Dr. Jing Zhang at University of Colorado Boulder, Dr. Chang Geun Yoo at Oak Ridge National Laboratory, and Dr. Michael W. Nolte and Prof. Brent H. Shanks at Iowa State University for fruitful discussions and useful suggestions.

References

  1. G. Fogassy, N. Thegarid, Y. Schuurman and C. Mirodatos, Energy Environ. Sci., 2011, 4, 5068–5076 CAS .
  2. C. Lindfors, V. Paasikallio, E. Kuoppala, M. Reinikainen, A. Oasmaa and Y. Solantausta, Energy Fuels, 2015, 29, 3707–3714 CrossRef CAS .
  3. C. Wang, M. Li and Y. Fang, Ind. Eng. Chem. Res., 2016, 55, 3525–3534 CrossRef CAS .
  4. A. de Rezende Pinho, M. B. B. de Almeida, F. L. Mendes, L. C. Casavechia, M. S. Talmadge, C. M. Kinchin and H. L. Chum, Fuel, 2017, 188, 462–473 CrossRef .
  5. E. G. Baker and D. C. Elliott, in Research in Thermochemical Biomass Conversion, ed. A. V. Bridgwater and J. L. Kuester, Springer Netherlands, 1988, ch. 67, pp. 883–895 Search PubMed .
  6. D. A. Ruddy, J. A. Schaidle, J. R. Ferrell III, J. Wang, L. Moens and J. E. Hensley, Green Chem., 2014, 16, 454–490 RSC .
  7. R. K. Sharma and N. N. Bakhshi, Energy Fuels, 1993, 7, 306–314 CrossRef CAS .
  8. D. M. Alonso, S. G. Wettstein and J. A. Dumesic, Chem. Soc. Rev., 2012, 41, 8075–8098 RSC .
  9. J. Q. Bond, A. A. Upadhye, H. Olcay, G. A. Tompsett, J. Jae, R. Xing, D. M. Alonso, D. Wang, T. Zhang, R. Kumar, A. Foster, S. M. Sen, C. T. Maravelias, R. Malina, S. R. H. Barrett, R. Lobo, C. E. Wyman, J. A. Dumesic and G. W. Huber, Energy Environ. Sci., 2014, 7, 1500–1523 CAS .
  10. E. Taarning, C. M. Osmundsen, X. B. Yang, B. Voss, S. I. Andersen and C. H. Christensen, Energy Environ. Sci., 2011, 4, 793–804 CAS .
  11. P. R. Patwardhan, R. C. Brown and B. H. Shanks, ChemSusChem, 2011, 4, 636–643 CrossRef CAS PubMed .
  12. C. Dong, Z. Zhang, Q. Lu and Y. Yang, Energy Convers. Manage., 2012, 57, 49–59 CrossRef CAS .
  13. X. Zhou, W. Li, R. Mabon and L. J. Broadbelt, Energy Technol., 2017, 5, 52–79 CrossRef CAS .
  14. H. V. Scheller and P. Ulvskov, Annu. Rev. Plant Biol., 2010, 61, 263–289 CrossRef CAS PubMed .
  15. A. Ebringerová, Z. Hromádková and T. Heinze, in Polysaccharides I, ed. T. Heinze, Springer Berlin Heidelberg, 2005, ch. 136816, pp. 1–67 Search PubMed .
  16. K. Werner, L. Pommer and M. Broström, J. Anal. Appl. Pyrolysis, 2014, 110, 130–137 CrossRef CAS .
  17. R. Zeitoun, P. Y. Pontalier, P. Marechal and L. Rigal, Bioresour. Technol., 2010, 101, 9348–9354 CrossRef CAS PubMed .
  18. S. R. Wang, B. Ru, G. X. Dai, W. X. Sun, K. Z. Qiu and J. S. Zhou, Bioresour. Technol., 2015, 190, 211–218 CrossRef CAS PubMed .
  19. S. R. Wang, B. Ru, H. Z. Lin and W. X. Sun, Fuel, 2015, 150, 243–251 CrossRef CAS .
  20. J. Zhang, Y. S. Choi, C. G. Yoo, T. H. Kim, R. C. Brown and B. H. Shanks, ACS Sustainable Chem. Eng., 2015, 3, 293–301 CrossRef CAS .
  21. G. J. Lv, S. B. Wu and R. Lou, BioResources, 2010, 5, 2051–2062 CAS .
  22. Y. Peng and S. Wu, J. Anal. Appl. Pyrolysis, 2010, 88, 134–139 CrossRef CAS .
  23. Y. Y. Peng and S. B. Wu, Cellul. Chem. Technol., 2011, 45, 605–612 CAS .
  24. M. H. O'Dwyer, Biochem. J., 1940, 34, 149–152 CrossRef .
  25. H. L. Cheng, J. L. Li, Q. H. Feng, H. Y. Zhan and Y. M. Xie, BioResources, 2014, 9, 2671–2680 CAS .
  26. H. Hasanjanzadeh, S. Hedjazi, A. Ashori, S. Mandavi and H. Yousefi, Int. J. Biol. Macromol., 2014, 68, 198–204 CrossRef CAS PubMed .
  27. H. C. Hu, X. S. Chai, H. Y. Zhan, D. Barnes, L. L. Huang and L. H. Chen, Ind. Eng. Chem. Res., 2014, 53, 11684–11690 CrossRef CAS .
  28. J. V. Rissanen, H. Grenman, S. Willfor, D. Y. Murzin and T. Salmi, Ind. Eng. Chem. Res., 2014, 53, 6341–6350 CrossRef CAS .
  29. M. H. Sipponen, V. Pihlajaniemi, S. Sipponen, O. Pastinen and S. Laakso, RSC Adv., 2014, 4, 23177–23184 RSC .
  30. F. Lopez, M. T. Garcia, V. Mena, J. M. Loaiza, M. A. M. Zamudio and J. C. Garcia, BioResources, 2015, 10, 55–67 CAS .
  31. L. Desharnais, F. Du and N. Brosse, Ind. Eng. Chem. Res., 2011, 50, 14217–14220 CrossRef CAS .
  32. T. Song, A. Pranovich and B. Holmbom, Bioresour. Technol., 2013, 130, 198–203 CrossRef CAS PubMed .
  33. M. J. Taherzadeh and K. Karimi, Int. J. Mol. Sci., 2008, 9, 1621–1651 CrossRef CAS PubMed .
  34. T. T. Qu, W. J. Guo, L. H. Shen, J. Xiao and K. Zhao, Ind. Eng. Chem. Res., 2011, 50, 10424–10433 CrossRef CAS .
  35. J. Huang, C. Liu, H. Tong, W. Li and D. Wu, J. Fuel Chem. Technol., 2013, 41, 285–293 CrossRef CAS .
  36. D. J. Nowakowski, C. R. Woodbridge and J. M. Jones, J. Anal. Appl. Pyrolysis, 2008, 83, 197–204 CrossRef CAS .
  37. T. Hosoya, H. Kawamoto and S. Saka, J. Anal. Appl. Pyrolysis, 2007, 78, 328–336 CrossRef CAS .
  38. S. R. Wang, T. Liang, B. Ru and X. J. Guo, Chem. Res. Chin. Univ., 2013, 29, 782–787 CrossRef CAS .
  39. S. Wang, T. Liang, B. Ru and X. Guo, Chem. Res. Chin. Univ., 2013, 29, 782–787 CrossRef CAS .
  40. D. K. Shen, S. Gu and A. V. Bridgwater, J. Anal. Appl. Pyrolysis, 2010, 87, 199–206 CrossRef CAS .
  41. E. Ranzi, A. Cuoci, T. Faravelli, A. Frassoldati, G. Migliavacca, S. Pierucci and S. Sommariva, Energy Fuels, 2008, 22, 4292–4300 CrossRef CAS .
  42. S. R. Horton, R. J. Mohr, Y. Zhang, F. P. Petrocelli and M. T. Klein, Energy Fuels, 2016, 30, 1647–1661 CrossRef CAS .
  43. Z. Zhang, C. Liu, H. J. Li, J. B. Huang and X. L. Huang, Acta Chim. Sin., 2011, 69, 2099–2107 CAS .
  44. H. Tian, B. Hu, Y. Zhang, Q. Lu, C. Dong and Y. Yang, J. Fuel Chem. Technol., 2015, 43, 185–194 CAS .
  45. J. B. Huang, C. Liu, H. Tong, W. M. Li and D. Wu, Comput. Theor. Chem., 2012, 1001, 44–50 CrossRef CAS .
  46. S. R. Wang, B. Ru, H. Z. Lin and Z. Y. Luo, Bioresour. Technol., 2013, 143, 378–383 CrossRef CAS PubMed .
  47. R. Vinu and L. J. Broadbelt, Annu. Rev. Chem. Biomol. Eng., 2012, 3, 29–54 CrossRef CAS PubMed .
  48. X. Zhou, L. J. Broadbelt and R. Vinu, in Advances in Chemical Engineering, ed. M. V. G. Kevin, Academic Press, 2016, pp. 95–198 Search PubMed .
  49. R. Vinu and L. J. Broadbelt, Energy Environ. Sci., 2012, 5, 9808–9826 CAS .
  50. X. Zhou, M. W. Nolte, H. B. Mayes, B. H. Shanks and L. J. Broadbelt, Ind. Eng. Chem. Res., 2014, 53, 13274–13289 CrossRef CAS .
  51. X. Zhou, M. W. Nolte, B. H. Shanks and L. J. Broadbelt, Ind. Eng. Chem. Res., 2014, 53, 13290–13301 CrossRef CAS .
  52. X. Zhou, M. W. Nolte, H. B. Mayes, B. H. Shanks and L. J. Broadbelt, AIChE J., 2016, 62, 766–777 CrossRef CAS .
  53. X. Zhou, M. W. Nolte, B. H. Shanks and L. J. Broadbelt, AIChE J., 2016, 62, 778–791 CrossRef CAS .
  54. O. Theander, in Fundamentals of Thermochemical Biomass Conversion, ed. R. P. Overend, T. A. Milne and L. K. Mudge, Springer Netherlands, 1985, ch. 2, pp. 35–60 Search PubMed .
  55. J. Z. Zhang, T. J. Chen, J. L. Wua and J. H. Wu, RSC Adv., 2014, 4, 17513–17520 RSC .
  56. N. Obel, L. Neumetzler and M. Pauly, in The Expanding Cell, ed. J. P. Verbelen and K. Vissenberg, Springer Berlin Heidelberg, 2007, ch. 71, pp. 57–88 Search PubMed .
  57. D. B. Jordan, M. J. Bowman, J. D. Braker, B. S. Dien, R. E. Hector, C. C. Lee, J. A. Mertens and K. Wagschal, Biochem. J., 2012, 442, 241–252 CrossRef CAS PubMed .
  58. S. M. Carnachan and P. J. Harris, Plant Physiol. Biochem., 2000, 38, 699–708 CrossRef CAS .
  59. Y. Habibi, M. Mahrouz and M. R. Vignon, C. R. Chim., 2005, 8, 1123–1128 CrossRef CAS .
  60. L. M. C. Cordeiro, C. P. de Almeida and M. Iacomini, Food Chem., 2015, 173, 141–146 CrossRef CAS PubMed .
  61. R. C. Pettersen, The Chemistry of Solid Wood, American Chemical Society, 1984, ch. 2, pp. 57–126 Search PubMed .
  62. R. C. Sun and J. Tomkinson, Carbohydr. Polym., 2002, 50, 263–271 CrossRef CAS .
  63. J. Gu and J. M. Catchmark, Carbohydr. Polym., 2012, 88, 547–557 CrossRef CAS .
  64. J. Gu and J. M. Catchmark, Cellulose, 2013, 20, 1613–1627 CrossRef CAS .
  65. H. Pereira, J. Graça and J. C. Rodrigues, in Wood Quality and Its Biological Basis, ed. J. R. Barnett and G. Jeronimidis, Blackwell Publishing, Oxford, 2003, pp. 53–86 Search PubMed .
  66. V. Seshadri and P. R. Westmoreland, J. Phys. Chem. A, 2012, 116, 11997–12013 CrossRef CAS PubMed .
  67. H. B. Mayes and L. J. Broadbelt, J. Phys. Chem. A, 2012, 116, 7098–7106 CrossRef CAS PubMed .
  68. H. B. Mayes, M. W. Nolte, G. T. Beckham, B. H. Shanks and L. J. Broadbelt, ACS Sustainable Chem. Eng., 2014, 2, 1461–1473 CrossRef CAS .
  69. H. B. Mayes, M. W. Nolte, G. T. Beckham, B. H. Shanks and L. J. Broadbelt, ACS Catal., 2015, 5, 192–202 CrossRef CAS .
  70. J. B. Paine III, Y. B. Pithawalla, J. D. Naworal and C. E. Thomas, J. Anal. Appl. Pyrolysis, 2007, 80, 297–311 CrossRef .
  71. J. B. Paine III, Y. B. Pithawalla and J. D. Naworal, J. Anal. Appl. Pyrolysis, 2008, 82, 10–41 CrossRef .
  72. J. B. Paine III, Y. B. Pithawalla and J. D. Naworal, J. Anal. Appl. Pyrolysis, 2008, 82, 42–69 CrossRef .
  73. J. B. Paine III, Y. B. Pithawalla and J. D. Naworal, J. Anal. Appl. Pyrolysis, 2008, 83, 37–63 CrossRef .
  74. U. Raisanen, I. Pitkanen, H. Halttunen and M. Hurtta, J. Therm. Anal. Calorim., 2003, 72, 481–488 CrossRef .
  75. B. Girisuta, K. G. Kalogiannis, K. Dussan, J. J. Leahy, M. H. B. Hayes, S. D. Stefanidis, C. M. Michailof and A. A. Lappas, Bioresour. Technol., 2012, 126, 92–100 CrossRef CAS PubMed .
  76. R. J. van Putten, J. C. van der Waal, E. de Jong, C. B. Rasrendra, H. J. Heeres and J. G. de Vries, Chem. Rev., 2013, 113, 1499–1597 CrossRef CAS PubMed .
  77. G. R. Ponder and G. N. Richards, Carbohydr. Res., 1993, 244, 27–47 CrossRef CAS .
  78. G. N. Richards, J. Anal. Appl. Pyrolysis, 1987, 10, 251–255 CrossRef CAS .
  79. Y. Houminer and S. Hoz, Isr. J. Chem., 1970, 8, 97–98 CrossRef CAS .
  80. M. S. Mettler, D. G. Vlachos and P. J. Dauenhauer, Energy Environ. Sci., 2012, 5, 7797–7809 CAS .
  81. I. Milosavljevic, V. Oja and E. M. Suuberg, Ind. Eng. Chem. Res., 1996, 35, 653–662 CrossRef CAS .
  82. Y. C. Lin, J. Cho, G. A. Tompsett, P. R. Westmoreland and G. W. Huber, J. Phys. Chem. C, 2009, 113, 20097–20107 CAS .
  83. J. M. Cho, J. M. Davis and G. W. Huber, ChemSusChem, 2010, 3, 1162–1165 CrossRef CAS PubMed .
  84. A. Aho, N. Kumar, K. Eranen, B. Holmbom, M. Hupa, T. Salmi and D. Y. Murzin, Int. J. Mol. Sci., 2008, 9, 1665–1675 CrossRef CAS PubMed .
  85. S. D. Stefanidis, K. G. Kalogiannis, E. F. Iliopoulou, C. M. Michailof, P. A. Pilavachi and A. A. Lappas, J. Anal. Appl. Pyrolysis, 2014, 105, 143–150 CrossRef CAS .
  86. G. R. Ponder and G. N. Richards, Carbohydr. Res., 1991, 218, 143–155 CrossRef CAS .
  87. F. Shafizadeh, G. D. McGinnis and C. W. Philpot, Carbohydr. Res., 1972, 25, 23–33 CrossRef CAS .
  88. P. Williams and S. Besler, in Advances in Thermochemical Biomass Conversion, ed. A. V. Bridgwater, Springer Netherlands, 1993, ch. 60, pp. 771–783 Search PubMed .
  89. L. J. Broadbelt and J. Pfaendtner, AIChE J., 2005, 51, 2112–2121 CrossRef CAS .
  90. B. J. Lynch, P. L. Fast, M. Harris and D. G. Truhlar, J. Phys. Chem. A, 2000, 104, 4811–4815 CrossRef CAS .
  91. T. Hosoya, Y. Nakao, H. Sato, H. Kawamoto and S. Sakaki, J. Org. Chem., 2009, 74, 6891–6894 CrossRef CAS PubMed .
  92. C. Shen, I. Y. Zhang, G. Fu and X. Xu, Chin. J. Chem. Phys., 2011, 24, 249–252 CrossRef CAS .
  93. H. M. Pilath, M. R. Nimlos, A. Mittal, M. E. Himmel and D. K. Johnson, J. Agric. Food Chem., 2010, 58, 6131–6140 CrossRef CAS PubMed .
  94. Y. Kodera and B. J. McCoy, AIChE J., 1997, 43, 3205–3214 CrossRef CAS .
  95. T. M. Kruse, O. S. Woo, H. W. Wong, S. S. Khan and L. J. Broadbelt, Macromolecules, 2002, 35, 7830–7844 CrossRef CAS .
  96. M. Wang, J. M. Smith and B. J. McCoy, AIChE J., 1995, 41, 1521–1533 CrossRef CAS .
  97. B. J. McCoy, AIChE J., 1993, 39, 1827–1833 CrossRef CAS .
  98. B. J. McCoy and G. Madras, J. Colloid Interface Sci., 1998, 201, 200–209 CrossRef CAS .
  99. W. E. Stewart, M. Caracotsios and J. P. Sorenson, DDASAC Software Package Documentation, Dept. Chem. Eng., Univ. Wisc, Madison, 1994 Search PubMed .
  100. R. C. Sun, J. M. Lawther and W. B. Banks, Carbohydr. Polym., 1996, 29, 325–331 CrossRef CAS .
  101. I. Simkovic, J. Alfoldi and H. R. Schulten, Biomass Bioenergy, 1993, 4, 373–378 CrossRef CAS .
  102. J. M. Brillouet, J. P. Joseleau, J. P. Utille and D. Lelievre, J. Agric. Food Chem., 1982, 30, 488–495 CrossRef CAS PubMed .
  103. B. Y. Wang, Environmental biodegradation research focus, Nova Science Publishers, New York, 2007 Search PubMed .
  104. M. M. de O. Buanafina, Mol. Plant, 2009, 2, 861–872 CrossRef CAS PubMed .
  105. D. A. B. Martins, H. F. A. do Prado, R. S. R. Leite, H. Ferreira, M. M. de S. Moretti, R. da Silva and E. Gomes, Agroindustrial Wastes as Substrates for Microbial Enzymes Production and Source of Sugar for Bioethanol Production, 2011 Search PubMed .
  106. C. Altaner, B. Saake and J. Puls, Cellulose, 2003, 10, 85–95 CrossRef CAS .
  107. Q. Beg, M. Kapoor, L. Mahajan and G. Hoondal, Appl. Microbiol. Biotechnol., 2001, 56, 326–338 CrossRef CAS PubMed .
  108. S. Willfor, R. Sjoholm, C. Laine, M. Roslund, J. Hemming and B. Holmbom, Carbohydr. Polym., 2003, 52, 175–187 CrossRef CAS .
  109. H. V. Scheller and P. Ulvskov, Annu. Rev. Plant Biol., 2010, 61, 263–289 CrossRef CAS PubMed .
  110. A. Ebringerova, Z. Hromadkova and T. Heinze, in Polysacchardes I, ed. T. Heinze, Springer Berlin Heidelberg, 2005, pp. 1–67 Search PubMed .

Footnote

Electronic supplementary information (ESI) available. See DOI: 10.1039/c7ee03208k

This journal is © The Royal Society of Chemistry 2018