Dehydration Leads to Hydrocarbon Gas Formation in Thermal Degradation of Gas-Phase Polyalcohols

To understand the molecular mechanisms of hydrocarbon gas formation in biomass gasification, gasification of simple polyalcohols (glycerol, propylene glycol, and ethylene glycol) were studied at 400, 600, and 800 °C (residence times: 0.9–1.4 s) from the viewpoint of dehydration reactions that form aldehydes with various substituents as intermediates to produce hydrocarbon gases. The results were also compared with those of glyceraldehyde and dihydroxyacetone, which are reported to produce syngas (H2 and CO) selectively. All polyalcohols became reactive at 600 °C to form condensable products in 15.7–24.7% yields (C-based), corresponding to 33.9–38.4% based on the amounts of reacted polyalcohols. These condensable products, mostly aldehydes, act as gas-forming intermediates, because the polyalcohols were completely gasified at 800 °C (hydrocarbon gas contents: 20.3–35.3%, C-based). Yields of the intermediates bearing alkyl groups at 600 °C were proportionally correlated to the yields of hydrocarbon gases at 800 °C, suggesting that the alkyl groups are further converted into hydrocarbon gases via the fragmentation of acyl radicals. Dehydration reactions were suggested to occur in both heterolytic and radical mechanisms by theoretical calculations. Glyceraldehyde tended to fragment directly into CO and H2, instead of forming a dehydration intermediate. These results are informative for controlling the product gas composition in biomass gasification.


Introduction
Biomass gasification, which produces oxygenated (OX) (carbon monoxide (CO) and carbon dioxide (CO 2 )) and hydrocarbon (HC) (methane (CH 4 ), ethylene (C 2 H 4 ) and acetylene (C 2 H 2 )) gases along with H 2 , is a potential method for the sustainable production of biofuels and biochemicals. The producer gas can be used by gas turbines/engines to generate electricity. Petroleum and various chemicals can be produced over Fischer-Tropsch catalysts via syngas (CO + H 2 ). Contamination from hydrocarbon gases, however, should be eliminated for the Fischer-Tropsch process, because of the poisoning problem on the catalysts. Ethylene is an important industrial chemical currently produced from petroleum, and the improvement of the production in biomass gasification would promote replacing petroleum with biomass as the source of ethylene. Therefore, controlling the gas selectivity of biomass gasification is important to expand its usability.
For controlling the gas selectivity, the water-gas shift reaction [1,2] converts CO and H 2 O into H 2 and CO 2 under steam gasification conditions. These processes increase the H 2 content instead of CO from biomass. However, reports are limited for the reactions that produce hydrocarbon gases from biomass gasification. However, molecular mechanisms that produce hydrocarbon gases in biomass components gasification have not been well clarified.
Stein et al. [3] reported the ethylene formation from acrolein, and they explained the formation via vinyl radical formed by the α-scission of the acrolein radical. Fukutome et al. [4] compared the gas and coke forming reactions of eight cellulose-derived volatile intermediates in an ampoule reactor at 600 °C. They showed that the intermediates bearing methyl groups (acetic acid and hydroxyacetone) produce more hydrocarbon gas (mainly methane) than that formed using glycolaldehyde, formic acid, or furanic compounds, which do not have any methyl groups. They also reported that glyceraldehyde (Gald; an aldose) and 1,3-dihydroxyacetone (DHA; a ketose) were selectively gasified into syngas [5]. Therefore, the gas composition from cellulose gasification varies depending on the chemical structure of the volatile intermediates during cellulose gasification.
Levoglucosan (1,6-anhydro-β-D-glucopyranose) is a major volatile intermediate in cellulose gasification, because of the large amounts (69.3% and 52.7%, C-based under nitrogen and 7% oxygen/nitrogen conditions, respectively) detected during the pyrolysis in a flow-type reactor [6]. The gaseous levoglucosan is stable up to around 500 °C and starts to fragment at 600 °C into smaller compounds and finally into non-condensable gases [7], although it readily degrades into polymerization and dehydration products, including coke, in the molten state by cooling [8]. High reactivity of the molten levoglucosan has been explained by the intermolecular hydrogen bonding acting as acid and base catalysts [9][10][11].
The kinetic analysis of the decomposition of gaseous levoglucosan indicated the radical chain reactions acting as the rate-determining step [12]. In analyzing the fragmentation pathways of the Cand O-centered radicals of levoglucosan from the perspective of organic reaction mechanisms, an interesting hypothesis has been proposed; hydrocarbon gases (methane, ethylene, and acetylene) are produced by the cleavage of the substituents of aldehydes, produced by the dehydration reactions from the levoglucosan radicals [12]. This hypothesis suggests that the hydrocarbon gas formation is directly related to the extent of the progression of dehydration reactions that can occur in both heterolysis and β-scission of radical intermediates. However, contributions of these mechanisms have not been fully clarified due to the complexity of the gasification reactions of levoglucosan as an intermediate from cellulose and hemicellulose in biomass gasification. Consequently, simple polyalcohols: glycerol (Gly), propylene glycol (PG), and ethylene glycol (EG) in Figure 1 were used in this study. It is easier to identify the intermediate products and to discuss their gas-forming reactions by effectively utilizing the theoretical calculation results obtained by using density functional theory (DFT) conducted at MP4(sdq)/Aug-cc-pVTZ//DFT(M06-2X)/6-31+G (p,d) levels. The results are also compared with those reported for Gald and DHA [5], which selectively produce synthesis gas (CO and H2) instead of the hydrocarbon gases.

Materials
Ethylene glycol, propylene glycol, and glycerol, which were used for pyrolysis trials, were purchased from Nacalai Tesque, Inc. (Kyoto, Japan). Table 1 summarizes some physicochemical data of these polyalcohols. These chemicals were used as received, without further purification.

Materials
Ethylene glycol, propylene glycol, and glycerol, which were used for pyrolysis trials, were purchased from Nacalai Tesque, Inc. (Kyoto, Japan). Table 1 summarizes some physicochemical data of these polyalcohols. These chemicals were used as received, without further purification.

Pyrolysis
A flow-type two-stage tubular reactor ( Figure 2) used in this study is fully described in our previous paper [5]. Sample (15 mg) put in a ceramic boat was placed at the center of the evaporator, and the air inside the reactor made of quartz glass tube was replaced with a nitrogen flow at 400 mL min −1 . The nitrogen flow was continuously supplied during the pyrolysis and for additional 2 min after the pyrolysis to recover the volatile products. After the pyrolyzer temperature was pre-set to 400, 600, or 800 • C, the evaporator was heated to 120 • C and held for 5 min, then further heated to 200 • C at 16 • C min −1 and held for 5 min. This process completely volatilized the sample and led it to the pyrolyzer. The residence times of gaseous sample in the pyrolyzer were 1.4 s (400 • C), 1.2 s (600), and 0.9 s (800 • C), which were settled to be near residence times of actual gasifications [13,14]. After the pyrolysis, the furnace cover was open and cooled by an air flow. The volatile products were passed through a gas washing bottle containing an oximation reagent (hydroxylammonim chloride (NH 2 OH•HCl), 20 mg) in dimethylsulfoxide (DMSO)-d 6 (2.0 mL), where aldehydes and ketones were converted into the corresponding oxime derivatives. The non-condensable gases were recovered into a gas bag.

Pyrolysis
A flow-type two-stage tubular reactor ( Figure 2) used in this study is fully described in our previous paper [5]. Sample (15 mg) put in a ceramic boat was placed at the center of the evaporator, and the air inside the reactor made of quartz glass tube was replaced with a nitrogen flow at 400 mL min −1 . The nitrogen flow was continuously supplied during the pyrolysis and for additional 2 min after the pyrolysis to recover the volatile products. After the pyrolyzer temperature was pre-set to 400, 600, or 800 °C, the evaporator was heated to 120 °C and held for 5 min, then further heated to 200 °C at 16 °C min −1 and held for 5 min. This process completely volatilized the sample and led it to the pyrolyzer. The residence times of gaseous sample in the pyrolyzer were 1.4 s (400 °C), 1.2 s (600), and 0.9 s (800 °C), which were settled to be near residence times of actual gasifications [13,14]. After the pyrolysis, the furnace cover was open and cooled by an air flow. The volatile products were passed through a gas washing bottle containing an oximation reagent (hydroxylammonim chloride (NH2OH•HCl), 20 mg) in dimethylsulfoxide (DMSO)-d6 (2.0 mL), where aldehydes and ketones were converted into the corresponding oxime derivatives. The non-condensable gases were recovered into a gas bag.

Product Analysis
The condensed substance attached on the reactor tube was washed with the DMSO-d6 solution in the gas washing bottle, and then the resulting solution was directly measured by proton ( 1 H) nuclear magnetic resonance (NMR) spectroscopy using a Brucker AC-400 (400 MHz) spectrometer. Yields of the products were determined by the comparison of the peak areas with those of the internal standard (2-furoic acid) which was added to the NMR sample solution.

Product Analysis
The condensed substance attached on the reactor tube was washed with the DMSO-d 6 solution in the gas washing bottle, and then the resulting solution was directly measured by proton ( 1 H) nuclear magnetic resonance (NMR) spectroscopy using a Brucker AC-400 (400 MHz) spectrometer. Yields of the products were determined by the comparison of the peak areas with those of the internal standard (2-furoic acid) which was added to the NMR sample solution.
Except for acetylene, non-condensable gases were determined by micro GC (Varian CP-4900) with the two columns system (MS5A (10 m) and PoraPLOT Q (10 m)). Acetylene was determined by The details of the chromatographic conditions are described in our previous paper [5].

Computational Methods
Calculations were made with Gaussian 09 software package [15]. The geometry optimizations and vibrational frequency calculations were conducted with DFT (M06-2X) and 6-31+G(d,p) basis sets. The energies of optimized geometries were calculated with MP4(SDQ). Single point energy calculation to estimate bond dissociation energy (BDE) were conducted with the CCSD(T) method and the AUG-cc-pVTZ basis sets. Activation energy (E a ) and BDE include zero-point energies. In frequency analysis, one imaginary mode was observed for transition state, but no imaginary modes for the reactants, products, and precomplexes. The intrinsic reaction coordinate calculations were also conducted to ascertain the transition state connecting the reactant and product.

Experimental Study of Gas Formation via Fragmentation of Intermediates
Gly, PG, and EG were recovered almost quantitatively at 400 • C and became reactive at 600 • C, where the recoveries were 32.0%, 43.2%, and 53.7%, respectively. The 24.7%, 21.8%, and 15.7% (C-based) of the condensable products were identified for the pyrolyzates from Gly, PG, and EG at 600 • C, respectively, which correspond to 36.3%, 38.4%, and 33.9% (C-based) against the amounts of degraded polyalcohols, respectively. Most of the remaining components were non-condensable gases. Because polyalcohols were completely converted into the non-condensable gases by increasing the temperature to 800 • C, these condensable products are the gas-forming intermediates from the polyalcohols. Therefore, a comparison of the chemical structures of condensable intermediates (600 • C) with the final gas composition (800 • C) would be valuable to understand the gasification pathways of polyalcohols.
The compositions of the non-condensable gases produced from Gly, PG, and EG at 800 • C are summarized in Figure 3, compared with those reported for Gald-DHA (61/36, mol/mol) and DHA [5]. The gas yields (82.2-97.4%, C-based) are also shown in the figure. Although the major gaseous products were syngas, the hydrocarbon gas yields were different depending on the chemical structure of the polyalcohol. Gald-DHA (61/36) and DHA produced syngas selectively with the limited amounts of HC gases (2.9-3.0%, C-based) and CO 2 , whereas the contents of the HC gases (mainly methane and ethylene) reached 20.3%, 35.3%, and 22.8% (C-based) from Gly, PG, and EG, respectively.  Discussion of the non-condensable gas formation with the composition formulas of polyalcohols provides insight into the gas-forming reactions. As shown in Figure 4, the CHO ratio of Gald and DHA (C 1 H 2 O 1 ) is ideal for the production of equimolar amounts of CO and H 2 . However, PG (C 1 H 2.7 O 0.6 ) has insufficient oxygen for the production of oxygenated (OX) gases (CO and CO 2 ); therefore, PG produced larger amounts of HC gases (35.3%, C-based). Gly (C 1 H 2.7 O 1 ) and EG (C 1 H 3 O 1 ) also produced reasonable amounts of HC gases at 20.3% and 22.8% (C-based), respectively, although these polyalcohols have sufficient oxygen to produce CO. Because the yields of CO 2 were low from Gly and EG, some of the oxygen atoms must be lost as water from these polyalcohols. Consequently, gas composition is not merely determined by the elemental composition, indicating that the gas forming reactions are not equilibrated but controlled kinetically under the present experimental conditions. Dehydration may compete with the fragmentation reactions to form non-condensable gases.
Energies 2020, 13, x FOR PEER REVIEW 5 of 17 also produced reasonable amounts of HC gases at 20.3% and 22.8% (C-based), respectively, although these polyalcohols have sufficient oxygen to produce CO. Because the yields of CO2 were low from Gly and EG, some of the oxygen atoms must be lost as water from these polyalcohols. Consequently, gas composition is not merely determined by the elemental composition, indicating that the gas forming reactions are not equilibrated but controlled kinetically under the present experimental conditions. Dehydration may compete with the fragmentation reactions to form non-condensable gases. Occurrence of dehydration reaction converts polyalcohols into enols, which are further rearranged into carbonyl compounds ( Figure 5A). As summarized in Figure 6, carbonyl compounds, particularly aldehydes, are the major components of condensable products. These results support the occurrence of dehydration reactions during polyalcohol gasification, although the chemical compositions are different depending on the polyalcohol type. Gald-DHA (61/36, mol/mol) and DHA gave formaldehyde (FA), glycolaldehyde (GA), glyoxal (GO), and methylglyoxal (MeGO) as the major components. Gly and PG produced larger amounts of C3 products, which include hydroxyacetone (HA) and acrolein (ACR) from Gly, and propionaldehyde (Pald) and allyl alcohol (AllyOH) from PG. EG, a C2 polyalcohol, formed acetaldehyde (AA) and FA selectively.  Occurrence of dehydration reaction converts polyalcohols into enols, which are further rearranged into carbonyl compounds ( Figure 5A). As summarized in Figure 6, carbonyl compounds, particularly aldehydes, are the major components of condensable products. These results support the occurrence of dehydration reactions during polyalcohol gasification, although the chemical compositions are different depending on the polyalcohol type. Gald-DHA (61/36, mol/mol) and DHA gave formaldehyde (FA), glycolaldehyde (GA), glyoxal (GO), and methylglyoxal (MeGO) as the major components. Gly and PG produced larger amounts of C 3 products, which include hydroxyacetone (HA) and acrolein (ACR) from Gly, and propionaldehyde (Pald) and allyl alcohol (AllyOH) from PG. EG, a C 2 polyalcohol, formed acetaldehyde (AA) and FA selectively.
Aldehydes are known to fragment via acyl radicals ( Figure 5B) [16]. Aldehydic hydrogens have been suggested as the predominant sites for hydrogen abstraction by •OH, from experimental and theoretical investigations [17][18][19]. Pre-reactive complex formation between aldehyde and •OH has been proposed for this selective abstraction [19]. Although the reactivities of alkyl radicals such as •CH 3 remain unclear, the resulting acyl radical (•CR(=O)) is fragmented into CO and •R through the α-scission mechanism. Therefore, the final gas compositions vary depending on the chemical structure of •R, which is stabilized as R-H by abstraction of hydrogen or degraded further ( Figure 5C).
When •R is •H (case of FA), H 2 is produced through hydrogen abstraction reaction. As in the case of GO, the resulting •R (•CH=O) fragments into CO and •H through similar β-scission reaction. When R is hydroxymethyl or hydroxymethyne group (case of GA and Gald), the resulting •C(R")-OH would be converted into CO and H 2 via the aldehyde and enol intermediates. Accordingly, these intermediate aldehydes selectively produce syngas (CO and H 2 ). Selective syngas formation from Gald and DHA is explainable with these reactions, since DHA is suggested to be rearranged into Gald in the gas phase [5].
On the other hand, the acyl radicals with alkyl groups can form HC gases via the stabilization of the alkyl radical intermediates. Methyl radical (•CH 3 ) formed from AA would produce methane, and ethyl (•CH 3 -CH 3 ) and vinyl (•CH 2 =CH 3 ) radicals formed from Pald and ACR, respectively, can be converted into ethane and unsaturated hydrocarbons such as ethylene and acetylene.   Aldehydes are known to fragment via acyl radicals ( Figure 5B) [16]. Aldehydic hydrogens have been suggested as the predominant sites for hydrogen abstraction by •OH, from experimental and theoretical investigations [17][18][19]. Pre-reactive complex formation between aldehyde and •OH has been proposed for this selective abstraction [19]. Although the reactivities of alkyl radicals such as  Alkyl groups attached on ketonic (>C=O) and carboxylic (-COOH) groups are also reported to produce HC gases [4]. Based on such information, acetone (ACE), MeGO, HA, acetic acid, AA, methanol, Pald, ACR, AllyOH and ketene are picked up as the intermediates bearing alkyl groups, and the yields of these compounds from polyalcohols (%, C-based on the condensable products) are compared in Figure 7. The contributions of these intermediates are small for Gald-DHA (61/36) (16.0%) and DHA (8.7%), but greater for Gly (62.4%), PG (88.1%), and EG (34.9%). It is noted that these yields are proportionally correlated to the contents of the HC gases in Figure 3, supporting the hypothesis; the intermediates bearing alkyl groups produce hydrocarbon gases. Many of such intermediates can be produced by dehydration reactions as discussed below.
these yields are proportionally correlated to the contents of the HC gases in Figure 3, supporting the hypothesis; the intermediates bearing alkyl groups produce hydrocarbon gases. Many of such intermediates can be produced by dehydration reactions as discussed below.
Comparatively large amount of methane formation from EG is explainable with the higher yield of AA, and the greater yields of ethylene from PG and Gly are coincide with the greater yields of Pald and ACR, respectively. Although the HC gas yields are quite small from Gald and DHA, high contributions of methane in HC gas are explainable with the AA and MeGO formation.
Gald is reactive for gasification and directly converted into CO and H2 with a smaller contribution of dehydration reaction to form MeGO leading to the methane production. Ethane would be produced through coupling of two •CH3.
Therefore, the progression of the dehydration reactions, which lowers the O/C ratios of the intermediates from polyalcohols, is suggested to be a key process for the formation of hydrocarbon gases, even if the elemental compositions of polyalcohols are suitable for the production of syngas.

Theoretical Study of Intermediate Formation
Dehydration of the alcohols could proceed in the gas phase through heterolytic dehydration, homolysis of the C-OH bond, and β-scission of C-centered radicals as illustrated in Figure 8. However, very high bond dissociation energies of the C-OH bonds in polyalcohols (93.5-95.6 kcal mol −1 for Gly [5]) indicate that the homolysis pathway is improbable even at 600 °C, where Gly, PG, and EG degraded. Therefore, heterolysis and radical pathways were considered. Along with the  Comparatively large amount of methane formation from EG is explainable with the higher yield of AA, and the greater yields of ethylene from PG and Gly are coincide with the greater yields of Pald and ACR, respectively. Although the HC gas yields are quite small from Gald and DHA, high contributions of methane in HC gas are explainable with the AA and MeGO formation.
Gald is reactive for gasification and directly converted into CO and H 2 with a smaller contribution of dehydration reaction to form MeGO leading to the methane production. Ethane would be produced through coupling of two •CH 3 .
Therefore, the progression of the dehydration reactions, which lowers the O/C ratios of the intermediates from polyalcohols, is suggested to be a key process for the formation of hydrocarbon gases, even if the elemental compositions of polyalcohols are suitable for the production of syngas.

Theoretical Study of Intermediate Formation
Dehydration of the alcohols could proceed in the gas phase through heterolytic dehydration, homolysis of the C-OH bond, and β-scission of C-centered radicals as illustrated in Figure 8. However, very high bond dissociation energies of the C-OH bonds in polyalcohols (93.5-95.6 kcal mol −1 for Gly [5]) indicate that the homolysis pathway is improbable even at 600 • C, where Gly, PG, and EG degraded. Therefore, heterolysis and radical pathways were considered. Along with the direct dehydration, cyclic Grob fragmentation and pinacol rearrangement were considered for the heterolysis reactions that release water [20].
For the radical pathways, the reactions from the C-and O-centered radicals were considered, which are formed through hydrogen abstraction from the -C-H and -O-H moieties of polyalcohols, respectively [21]. Various radical species such as •OH and •CH 3 existing in the pyrolysis environment can abstract these hydrogens [22].
Energies 2020, 13, x FOR PEER REVIEW 8 of 17 direct dehydration, cyclic Grob fragmentation and pinacol rearrangement were considered for the heterolysis reactions that release water [20]. For the radical pathways, the reactions from the C-and O-centered radicals were considered, which are formed through hydrogen abstraction from the -C-H and -O-H moieties of polyalcohols, respectively [21]. Various radical species such as •OH and •CH3 existing in the pyrolysis environment can abstract these hydrogens [22].
In this section, heterolytic dehydration reactions were theoretically evaluated and discussed with the pyrolysis products obtained at 600 °C. Then, the products that could not be explained with the heterolytic reactions are discussed with the radical pathways that are reasonably drawn from the C-and O-centered radicals.
Bimolecular reactions are also considered, since bimolecular keto-enol tautomerization have been suggested for DHA ↔ enol ↔ Gald in our previous paper [5]. In this paper, pyrolytic degradation of Gald into MeGO + AA and GA + FA at 400 °C were explained with the unimolecular reactions, that is, dehydration via the six-membered cyclic TS of enol intermediate of Gald (Ea 35.5 kcal mol −1 ) and retro-aldol fragmentation of Gald (Ea 40.9 kcal mol −1 ), respectively. DHA was suggested to degrade via the enol and Gald. However, the values of Ea of the enolization steps of Gald and DHA were calculated to be too high (Ea 72.5 and 69.3 kcal mol −1 , respectively) to explain why these steps proceed at such a low temperature of 400 °C. In explaining the contradiction between experimental and theoretical calculation results, bimolecular mechanisms have been proposed for decreasing the values of Ea to 46.4 and 48.7 kcal mol −1 , respectively.
Nevertheless, bimolecular mechanism should be treated carefully, because the meeting of two molecules are necessary in the gas phase, which is not effective due to the increasing entropy. Accordingly, to conclude the bimolecular mechanisms, more systematic studies should be necessary. However, we consider that the bimolecular reactions would be possible under some special conditions such as in cluster and a product-water complex formed just after the dehydration reaction even in the gas phase.
Some molecules such as carboxylic acids and alcohols, both of which have the ability of the hydrogen-bonded complex formation, are known to exist as a cluster in the gas phase. For example, it is suggested that methanol exists as dimers, trimers, and oligomers in addition to the monomer in the gas phase [23][24][25][26][27][28][29].

Ethylene Glycol (EG)
EG produced AA and FA as the major condensable products. As shown in Figure 9, both direct dehydration (E-i, Ea 74.1 kcal mol −1 ) and pinacol rearrangement (E-iii, Ea 73.0 kcal mol −1 ) pathways can explain the formation of AA from EG, although the energy barriers are not small. The Ea (74.1 kcal mol −1 ) of the unimolecular dehydration decreased to 64.0 kcal mol −1 by assuming the bimolecular mechanism involving two EG molecules. In this section, heterolytic dehydration reactions were theoretically evaluated and discussed with the pyrolysis products obtained at 600 • C. Then, the products that could not be explained with the heterolytic reactions are discussed with the radical pathways that are reasonably drawn from the Cand O-centered radicals.
Bimolecular reactions are also considered, since bimolecular keto-enol tautomerization have been suggested for DHA ↔ enol ↔ Gald in our previous paper [5]. In this paper, pyrolytic degradation of Gald into MeGO + AA and GA + FA at 400 • C were explained with the unimolecular reactions, that is, dehydration via the six-membered cyclic TS of enol intermediate of Gald (E a 35.5 kcal mol −1 ) and retro-aldol fragmentation of Gald (E a 40.9 kcal mol −1 ), respectively. DHA was suggested to degrade via the enol and Gald. However, the values of E a of the enolization steps of Gald and DHA were calculated to be too high (E a 72.5 and 69.3 kcal mol −1 , respectively) to explain why these steps proceed at such a low temperature of 400 • C. In explaining the contradiction between experimental and theoretical calculation results, bimolecular mechanisms have been proposed for decreasing the values of E a to 46.4 and 48.7 kcal mol −1 , respectively.
Nevertheless, bimolecular mechanism should be treated carefully, because the meeting of two molecules are necessary in the gas phase, which is not effective due to the increasing entropy. Accordingly, to conclude the bimolecular mechanisms, more systematic studies should be necessary. However, we consider that the bimolecular reactions would be possible under some special conditions such as in cluster and a product-water complex formed just after the dehydration reaction even in the gas phase.
Some molecules such as carboxylic acids and alcohols, both of which have the ability of the hydrogen-bonded complex formation, are known to exist as a cluster in the gas phase. For example, it is suggested that methanol exists as dimers, trimers, and oligomers in addition to the monomer in the gas phase [23][24][25][26][27][28][29].

Ethylene Glycol (EG)
EG produced AA and FA as the major condensable products. As shown in Figure 9, both direct dehydration (E-i, E a 74.1 kcal mol −1 ) and pinacol rearrangement (E-iii, E a 73.0 kcal mol −1 ) pathways can explain the formation of AA from EG, although the energy barriers are not small. The E a (74.1 kcal mol −1 ) of the unimolecular dehydration decreased to 64.0 kcal mol −1 by assuming the bimolecular mechanism involving two EG molecules.
To understand the natures of the bimolecular mechanisms in direct dehydration (E-i), the ∆G ‡ , ∆H ‡ , and T∆S ‡ values were calculated for EG (unimolecular), EG-EG, and EG-H 2 O at 25, 400, and 600 • C ( Table 2), and compared with those calculated for the keto-enol tautomerization of DHA (unimolecular), DHA-DHA, and DHA-H 2 O (Table 3 To understand the natures of the bimolecular mechanisms in direct dehydration (E-i), the ∆G ‡ , ∆H ‡ , and T∆S ‡ values were calculated for EG (unimolecular), EG-EG, and EG-H2O at 25, 400, and 600 °C (Table 2), and compared with those calculated for the keto-enol tautomerization of DHA (unimolecular), DHA-DHA, and DHA-H2O (Table 3     To understand the natures of the bimolecular mechanisms in direct dehydration (E-i), the ∆G ‡ , ∆H ‡ , and T∆S ‡ values were calculated for EG (unimolecular), EG-EG, and EG-H2O at 25, 400, and 600 °C (Table 2), and compared with those calculated for the keto-enol tautomerization of DHA (unimolecular), DHA-DHA, and DHA-H2O (Table 3   Formation of FA and other minor products, that is, GA, GO, and MeOH, cannot be explained with the heterolysis reactions, but with the radical pathways from the C-and O-centered radicals of EG as illustrated in Figure 10. FA and MeOH can form via the β-scission of the O-centered radical of EG (E-e). The formations of GA and GO are possible by the radical pathways E-a, E-d, and E-f and pathway E-g, respectively. Thus, radical pathways are necessary to explain the product formation from EG. AA can also form by pathway E-b. The influences of the bimolecular mechanisms on the dehydration reaction were much lower, because of the smaller reduction of the ∆H ‡ values (71.6-72. mechanisms. This result is concordant with the lower pyrolytic reactivity of EG than those of Gald and DHA. Formation of FA and other minor products, that is, GA, GO, and MeOH, cannot be explained with the heterolysis reactions, but with the radical pathways from the C-and O-centered radicals of EG as illustrated in Figure 10. FA and MeOH can form via the β-scission of the O-centered radical of EG (E-e). The formations of GA and GO are possible by the radical pathways E-a, E-d, and E-f and pathway E-g, respectively. Thus, radical pathways are necessary to explain the product formation from EG. AA can also form by pathway E-b. Formation of FA and other minor products, that is, GA, GO, and MeOH, cannot be explained with the heterolysis reactions, but with the radical pathways from the C-and O-centered radicals of EG as illustrated in Figure 10. FA and MeOH can form via the β-scission of the O-centered radical of EG (E-e). The formations of GA and GO are possible by the radical pathways E-a, E-d, and E-f and pathway E-g, respectively. Thus, radical pathways are necessary to explain the product formation from EG. AA can also form by pathway E-b. The radical intermediates are fragmented mainly through the β-scission reactions, which require the radical p-orbital and the σ-orbital of the C-X bond that is cleaved to lie in the same plane [30].

DHA-DHA (bi)
DHA + H 2 O (bi) Figure 10. Possible radical chain pathways from the C-and O-centered radicals of ethylene glycol (EG). The bond with a bold line, which is located at the β-position to the radical, is expected to be cleaved through β-scission reaction. GA, glycolaldehyde; GO, glyoxal; AA, acetaldehyde; and FA, formaldehyde.
The radical intermediates are fragmented mainly through the β-scission reactions, which require the radical p-orbital and the σ-orbital of the C-X bond that is cleaved to lie in the same plane [30]. This pathway leads to the stereoselective formation of cis enol from EG (reaction E-c in Figure 10). Therefore, the influences of these stereoisomers during the keto-enol tautomerization were studied further ( Figure 11). The cis enol (Ea 74.6 kcal mol −1 ) was more stable than the corresponding trans isomer (Ea 58.6 kcal mol −1 ) by assuming the unimolecular mechanisms. In the gas phase, -O-H would associate with the double bond π electron to accomplish the proton-transfer required for tautomerization to the aldehyde isomer. The hydrogen bond formed between the two OH groups in the cis enol make this transition difficult. Consequently, pathway E-f would compete with the hydrogen abstraction pathway E-g reading to the formation of GO. The intramolecular hydrogen bonding lowers the BDE of the -O-H bond in the cis isomer to 70.6 kcal mol −1 . The trans enol, which is not formed from the radical reactions of EG, may be transformed to GA through the unimolecular keto-enol tautomerization, because of the low E a of 58.6 kcal mol −1 .
hydrogen abstraction pathway E-g reading to the formation of GO. The intramolecular hydrogen bonding lowers the BDE of the -O-H bond in the cis isomer to 70.6 kcal mol −1 . The trans enol, which is not formed from the radical reactions of EG, may be transformed to GA through the unimolecular keto-enol tautomerization, because of the low Ea of 58.6 kcal mol −1 .
Based on the above discussion, both heterolysis and radical chain pathways are suggested to be involved in the gas-phase pyrolytic conversion of EG into gasifying intermediates. Cis isomer Trans isomer Figure 11. Influences of the stereochemistry of cis/trans enol on activation energy (Ea, kcal mol −1 ) of keto-enol tautomerization by assuming the unimolecular mechanisms, as calculated at the MP4(SDQ)//DFT(M06-2X) level. BDE: bond dissociation energy.

Propylene Glycol (PG)
Pald represents 54.8% of the condensable products from PG, followed by AllyOH, AA, FA, and other minor products (ACE, MeGO, and ACR) at 600 °C. Heterolysis and radical pathways, which were considered for the pyrolysis of PG, are summarized in Figures 12 and 13, respectively.
For unimolecular heterolysis reactions, Pald can form from three pathways, P-iv, P-v, and P-vii, with the values of Ea of 73.6, 68.3, and 76.6 kcal mol −1 , respectively. Pathway P-ii (Ea, 66.7 kcal mol −1 ) and pathways P-i (Ea, 76.1 kcal mol −1 ) and P-vi (71.7 kcal mol −1 ) can produce AllyOH and ACE, respectively. The lower values of Ea of 68.3 (P-v) and 66.7 kcal mol −1 (P-ii) are reasonable to explain why Pald and AllyOH are produced as the major products. The reactions forming ACE as a minor product (1.7%, C-based) have the greater values of Ea of 76.1 (P-i) and 71.7 kcal mol −1 (P-vi).
Bimolecular mechanisms lower the values of Ea of direct dehydration reactions P-i, P-ii, and Piv. As a result, the Ea of reaction P-ii forming AllyOH as the second greatest product decreased to the level of 57.4 kcal mol −1 . Furthermore, the isomerization from AllyOH to enol, which exhibited extremely high Ea (93.8 kcal mol −1 ) via a unimolecular mechanism (reaction P-iii), was reduced to 60.9 kcal mol −1 with the bimolecular mechanism. The ∆G ‡ , ∆H ‡ , and T∆S ‡ values calculated for the unimolecular and bimolecular isomerizations of AllyOH at 25, 400, and 600 °C are summarized in Table 4. The ∆G ‡ at 600 °C decreased from 95.7 (unimolecular) to 63.2 kcal mol −1 (bimolecular) due to the stabilization of the TS with a six-membered ring. Consequently, the preferential formation of Pald and AllyOH from PG is explainable with the pathways P-ii + P-iii followed by the keto-enol tautomerization, by assuming the bimolecular mechanisms. When the product water formed from reaction P-ii can be used for the bimolecular reaction P-iii before leaving the product AllyOH, the biomolecular reaction would proceed smoothly.
As discussed in the case of EG, the formations of AA, FA, and other minor products (HA, MeGO and ACR) cannot be explained with the heterolysis pathways. These products would produce through radical pathways ( Figure 13). AA and FA would form via the pathways P-i and P-k. The Based on the above discussion, both heterolysis and radical chain pathways are suggested to be involved in the gas-phase pyrolytic conversion of EG into gasifying intermediates.

Propylene Glycol (PG)
Pald represents 54.8% of the condensable products from PG, followed by AllyOH, AA, FA, and other minor products (ACE, MeGO, and ACR) at 600 • C. Heterolysis and radical pathways, which were considered for the pyrolysis of PG, are summarized in Figures 12 and 13 formations of other minor products are also explained with the pathways P-c, P-d, P-g, P-j, and P-l for HA and P-c for MeGO. ACR may form through oxidation of AllyOH. Pald, ACE and AllyOH would also form through P-a, P-f, and P-h, respectively. Therefore, radical pathways are also expected to contribute to the degradation of PG.  For unimolecular heterolysis reactions, Pald can form from three pathways, P-iv, P-v, and P-vii, with the values of E a of 73.6, 68.3, and 76.6 kcal mol −1 , respectively. Pathway P-ii (E a , 66.7 kcal mol −1 ) and pathways P-i (E a , 76.1 kcal mol −1 ) and P-vi (71.7 kcal mol −1 ) can produce AllyOH and ACE, respectively. The lower values of E a of 68.3 (P-v) and 66.7 kcal mol −1 (P-ii) are reasonable to explain why Pald and AllyOH are produced as the major products. The reactions forming ACE as a minor product (1.7%, C-based) have the greater values of E a of 76.1 (P-i) and 71.7 kcal mol −1 (P-vi). re 12. Possible heterolytic dehydration pathways for propylene glycol (PG). The values nex arrows represent the activation energies (kcal mol −1 ), as calculated at the MP4(SDQ)//DFT(M level. Pald, propionaldehyde; AllyOH, allyl alcohol; ACE, acetone; uni, unimolecular mechani bi, bimolecular mechanism.  Bimolecular mechanisms lower the values of E a of direct dehydration reactions P-i, P-ii, and P-iv. As a result, the E a of reaction P-ii forming AllyOH as the second greatest product decreased to the level Energies 2020, 13, 3726 13 of 18 of 57.4 kcal mol −1 . Furthermore, the isomerization from AllyOH to enol, which exhibited extremely high E a (93.8 kcal mol −1 ) via a unimolecular mechanism (reaction P-iii), was reduced to 60.9 kcal mol −1 with the bimolecular mechanism. The ∆G ‡ , ∆H ‡ , and T∆S ‡ values calculated for the unimolecular and bimolecular isomerizations of AllyOH at 25, 400, and 600 • C are summarized in Table 4. The ∆G ‡ at 600 • C decreased from 95.7 (unimolecular) to 63.2 kcal mol −1 (bimolecular) due to the stabilization of the TS with a six-membered ring. Consequently, the preferential formation of Pald and AllyOH from PG is explainable with the pathways P-ii + P-iii followed by the keto-enol tautomerization, by assuming the bimolecular mechanisms. When the product water formed from reaction P-ii can be used for the bimolecular reaction P-iii before leaving the product AllyOH, the biomolecular reaction would proceed smoothly.

Glycerol (Gly)
Gly formed HA, ACR, AA, GA, and FA, along with other minor products at 600 °C. The postulated heterolysis pathways as summarized in Figure 14 explain the formation of the major components except for GA. The pathway with the lowest Ea (66.7 kcal mol −1 ) with unimolecular mechanism is the pinacol rearrangement (G-v) giving FA and AA via the retro-aldol fragmentation (G-vi, Ea 41.7 kcal mol −1 ). Cyclic Grob fragmentation (G-i, Ea 68.8 kcal mol −1 ), which is a characteristic reaction of Gly, also forms AA and FA.
By assuming the bimolecular reactions, direct dehydration reactions G-ii (66.6 kcal mol −1 ) and G-iii (63.4 kcal mol −1 ) become more reactive to explain the formation of HA and ACR, respectively, more reasonably as the other major products.
All of the dehydration products listed in Figure 14 can be produced through radical reactions illustrated in Figure 15. Many of the radical reactions explain the formation of GA, a major product which is not expected to form by heterolysis reactions. Other minor condensable intermediates (MeGO and GO) can also be produced by the radical pathways.

Discussion for Syngas and Hydrocarbon Gas Production from Polyalcohol
Gas-forming pathways suggested from the results of the present investigation are summarized in Figure 16. Several heterolysis pathways are suggested to explain the formation of dehydration products from polyalcohols along with the radical chain pathways, where C-OH bonds are cleaved by the β-scission type reactions occurring for the C-centered radicals. On the other hand, many of the radical chain pathways particularly from the O-centered radicals give aldehydes/ketones that bear the hydroxymethyl and hydroxymethyne groups such as glyceraldehyde and glycolaldehyde, leading to the syngas production. Accordingly, relative efficiency of hydrogen abstractions from C-H and O-H bonds may affect the relative composition of syngas and hydrocarbon gas in addition to the occurrence of heterolytic dehydration reactions. The higher concentration of syngas in the gaseous products from polyalcohol gasification would be due to the greater contribution of the radical chain fragmentations via the O-centered radicals, although further systematic study is needed to conclude it.

sion for Syngas and Hydrocarbon Gas Production from Polyalcohol
rming pathways suggested from the results of the present investigation are su 6. Several heterolysis pathways are suggested to explain the formation of de The bond with a bold line, which is located at the β-position to the radical, is expected to be cleaved through β-scission reaction. ACR, acrolein; MeGO, methylglyoxal; Gald, glyceraldehyde; DHA, 1,3-dihydroxyacetone; HA, hydroxyacetone; GA, glycolaldehyde; GO, glyoxal; AA, acetaldehyde; FA, formaldehyde.
leading to the syngas production. Accordingly, relative efficiency of hydrogen abstractions from C-H and O-H bonds may affect the relative composition of syngas and hydrocarbon gas in addition to the occurrence of heterolytic dehydration reactions. The higher concentration of syngas in the gaseous products from polyalcohol gasification would be due to the greater contribution of the radical chain fragmentations via the O-centered radicals, although further systematic study is needed to conclude it.  Figure 16. Syngas and hydrocarbon gas formation pathways from polyalcohol. Figure 16. Syngas and hydrocarbon gas formation pathways from polyalcohol.

Conclusions
Thermal degradation of gas-phase simple polyalcohols (glycerol, propylene glycol, and ethylene glycol) was studied experimentally and theoretically, focusing on the formation and roles of condensable gas-forming intermediates at 400, 600, and 800 • C under nitrogen flow (residence times: 0.9-1.4 s), and the results were compared with those of glyceraldehyde. The following conclusions are obtained:

1.
All polyalcohols become reactive at 600 • C and produce condensable products in 15.7-24.7% yields (C-based), corresponding to 33.9-38.4% based on the amounts of reacted polyalcohols. Because the polyalcohols are completely converted into non-condensable gases at 800 • C, the condensable products observed at 600 • C act as gas-forming intermediates.

2.
Most of the condensable products are aldehydes, which can be formed from by dehydration and the following keto-enol tautomerization. Loss of some oxygen atoms as water is also suggested by the elemental compositions of gas and gas-forming intermediates.

3.
Yields of gas-forming intermediates (600 • C, mostly aldehydes) bearing alkyl groups are proportionally correlated to the yields of hydrocarbon gases (800 • C), suggesting the alkyl groups being converted into hydrocarbon gases.

4.
Fragmentation pathways of aldehydes via acyl radicals can explain the formation of characteristic hydrocarbons from polyalcohols (methane from ethylene glycol, ethylene and acetylene from glycerol and propylene glycol).

5.
Glyceraldehyde directly fragments into CO and H 2 with a smaller contribution of dehydration reaction to form methyl glyoxal leading to the methane production, which explains why glyceraldehyde produces syngas selectively. 6.
Theoretical calculations indicate heterolysis mechanisms for polyalcohol dehydration along with the contribution of radical chain mechanisms. 7.
By assuming the bimolecular mechanisms for heterolytic dehydration reactions, the formations of gas-forming intermediates are explainable more reasonably. 8.
These findings provide insight into upgrading gasification processes and controlling the gas compositions between hydrocarbons and syngas for sustainable utilization of biomass in biofuel and biochemical applications.