The Homogeneous Conversion Mechanism of Cellulose Pyrolysis Tar Under the Effect of Steam

Biomass accounts for the largest proportion of rural solid waste with high moisture content, which affects the thermal treatment process. This paper studied the effect of steam on the pyrolysis tar of microcrystalline cellulose (MCC) by a two-stage xed bed. The experiments had been carried out under different steam/feedstock mass ratios (S/F= 0, 0.8, 1.2, 1.6) when the rst stage was at 600 ℃ , and the second stage was at 800 ℃ . The tar content in the syngas was reduced effectively from 6.68% to 2.30% when the S/F addition was from 0 to 1.6. The steam could promote craking of compounds to form more stable compounds. To further study the removal mechanism of tar, the main tar component, phosphonic acid, (p-hydroxyphenyl-) , was investigated using density functional theory (DFT). It was concluded that intermediate product from cellulose pyrolysis was more likely to react with H 2 O and made against the production of phosphonic acid, (p-hydroxyphenyl-), consistent with the experimental results.


Introduction
Nowadays, there are still many agricultural-oriented countries and regions. Besides the kitchen garbage commonly found in municipal solid waste (MSW), there are also abundant agricultural and wood residues in these places. Given these wastes are always different as the various compositions and complex structures, it is meaningful to study the thermochemical reaction mechanism of biomass based on three major biomass components (cellulose, hemicellulose, and lignin). Among them, cellulose is the most abundant, accounting for 40-60% [1], and there have been many studies on it. For example, MCC(crystallinity index is approximately 80% [1]), a typical model compound, is always used to research cellulose pyrolysis characteristics [2][3][4][5].
For the thermal treatment of biomass waste, tar production can not be neglected, especially the high tar yield when biomass pyrolysis. To further reduce the adverse impact of tar on equipment and the environment, researchers have conducted fruitful studies by adding catalysts [5][6][7][8][9] and introducing steam [10][11][12]. In contrast to catalysts, adding steam is a more convenient method, whose calori c value of syngas [13] is usually 10-18 MJ/Nm 3 . Hence, the steam is often applied to promote tar cracking, although the steam could increase equipment heat loss [14].
On the other hand, the moisture content of biomass also impacts the operation of the facility. Compared with typical thermal treatment equipment, such as batch reactors, uidized beds, the xed bed has the advantages of simple design and operation with low investment and operating costs. But the xed beds running in towns and villages have some shortcomings due to the limited investment cost and simple design. For example, these devices do not have a drying system in front of the furnace because the moisture content of rural waste is lower than that of MSW. Nevertheless, the moisture in the garbage is heated and volatilizes into steam, which will stay in the furnace long enough and affect the whole reaction process. For biomass wastes with large moisture content, it is a non-negligible factor in the reaction process.
Some previous studies have shown that the reforming reaction between steam and pyrolysis/gasi cation products is conducive to increasing H 2 yield [15,16] and reducing the yield of tar [10,11]. Feng et al. [12] found that biomass gasi cation was optimal at 700-900℃. And under an inert atmosphere, thermal cracking of biomass tar had no apparent in uence when the temperature was less than 700℃. Moreover, some views suggest that tar can be cracked effectively by contact with coal char at temperatures above 800° C [11,17,18].
These results are encouraging, but the role of steam in the pyrolysis of biomass tar needs to be further con rmed in the absence of char, especially when the tar yield is the highest at 600 ℃ [19,20].
Gaussian is a comprehensive quantum chemistry package, which has been widely used in researching the molecular mechanism. The cellulose pyrolysis mechanism has been studied by using the Gaussian to analyze possible pyrolytic pathways [21][22][23][24]. Lu et al. [24] described this process as that the indigenous interior units, reducing end (RE end) and non-reducing end (NR end), initially form various characteristic chain ends and dehydrated units, which then evolve into different pyrolytic products. Huang et al. [25] used phenethyl phenyl ether (PPE) as a dimeric lignin model compound to investigate possible pyrolytic pathways by density functional theory methods at B3LYP/6-31G(d) level. Based on the unimolecular decomposition mechanism of lignin pyrolysis studies, Jiang et al. [26] proposed a new intermolecular interaction mechanism of lignin pyrolysis. This paper studied the yield and composition of tar formed from the cellulose pyrolysis with the effect of steam. In particular, this experiment was based on the MCC pyrolysis tar instead of a few representative modeling tar [27,28], which provided a more reliable result. Moreover, the experiment focused on the effect of steam content without considering the potential role of char. To further understand the conversion mechanism, Gaussian was also used to investigate the possible tar reaction paths.

Model compounds
To simulate the combustible biomass fraction of rural waste, the feedstock used here is MCC, a puri ed and partially depolymerized cellulose and accounts for about 70% of plant cellulose. The proximate and ultimate analysis of the feedstock is shown in Table 1, which were determined by 5E Series-MAG6700 Proximate Analyzer and 5E Series-CHN2000 Ultimate Analyzer, respectively. Before the experiment, MCC had been sieved to an average diameter of about 0.4 mm and then dried at 105℃ for 24 hours to reduce the interference of moisture content on the experimental results.

Experimental equipment and conditions
A scheme of the bench-scale plant used for the biomass wastes treatment process is shown in Fig 1. The apparatus mainly consists of a reverse L tubular reactor made of steel (outer diameter: 40 mm, inner diameter: 30 mm). There are two stages, with each length of 500mm. The feedstock was delivered into the pyrolysis reactor through a screw feeder installed in the tubular channel. After about 25 minutes of stirring and moving at a low rate, the raw material was entirely pyrolysis in an inert atmosphere. Then, the remaining residue was sent into the slag hole. Simultaneously, the gaseous products (including syngas and tar) entered into the second stage and reacted with saturated steam, as the suction of the vacuum pump at the end. After entirely react, all the gas components were pumped into a condenser to separate the moisture. The tar in the produced syngas could be absorbed entirely by the dichloromethane absorption in the scrubbing bottle after passing through a tar trap. The rest non-condensate gas was also treated safely by other means. Furthermore, heating belts (set at 180℃) cover the connecting section to reduce steam condensation before entering the second stage.
The experimental conditions are shown in Table 2. After carried out for 60 min of every continuous operation, stopped the feed and maintained the steam ow for another 20 min. Each experiment was repeated at least three times to ensure the reproducibility of the results.

Tar pretreatment and analysis
Before the tar analysis, there were a series of pretreatments: 1) transferred the mixture solution collected from the tar trap and washed it with 50 ml dichloromethane each time, repeating three times; 2) poured all the collected solutions, including the washing solution, into a separation funnel to separate moisture and organic solvents; 3) added anhydrous sodium sulfate to the organic solution to further remove the moisture; 4) separated the tar from the ltered organic solvent (a mixture of tar and dichloromethane) by rotating evaporators following the European standard (CEN/TS15439: 2006).
The composition of tar was measured by a GC-MS analyzer (7890B-5977A, Agilent Technologies Inc.) The injector and the transfer line were set at 270 and 250 °C, respectively. After 5 min at 50 °C, the oven raised to 270 °C at a rate of 15 °C/min and then maintained at 270 °C for 10 min. In the mass spectrometer, electron ionization (EI) energy was used for ionization. The ion source temperature was maintained at 200 °C. The volume of each injection was 0.2 μL, and the split ratio was 10:1. There are three parallel samples of each experiment condition, which have been measured by GC-MS, and the average results are shown in the Appendix.

Calculation methods
The optimizations of all the geometries of compounds (reactants, products) were performed at the DFT/ B3LYP level using the 6-31G(d)basis set. According to the possible reaction paths, the initial guess structure of transition states was located by TS QST2 with semi-empirical molecular orbital methods /PM6. Then, the transition states were calculated again by the TS method and were con rmed by exactly one imaginary frequency and IRC, whose calculations were employed at the same basic set of the DFT/ B3LYP method. Based on all the stable structures (reactants, intermediates, and products), using the 6-311G + + (d, p) basis set to complete the frequency and thermochemical analysis at 800 °C. After the zero-point energy correction (ZPE) and thermodynamic correction, the thermodynamic parameters, such as activation energy (reaction barrier), could be used to estimate the possible reaction paths.
Above all, the calculations were performed by Gaussian 09 [29]. And the Multiwfn3.8 [30]was also used to calculate the frontier orbital and Laplacian [31] to analyze the active sites.

Results And Discussion
3.1 Properties of tar As shown in Table 3, the yield of tar cuts back with steam increasing. It is consistent with earlier reported results on biomass pyrolysis gasi cation with steam [10][11][12]. However, this trend is not maintained: the reduction rate reaches the maximum at S/F=0.8, declined by one third from 6.68% during pyrolysis, and then it is gradually slowing down. Similarly, the H/C atomic ratio of tar reduces after the addition of steam and reaches the minimum at S/F=1.2. And the oxygen element in tar reduces signi cantly with the growth of steam, which indicates better tar quality in terms of calori c value and stability [32,33].

Effect of steam on tar components
According to the measurement result of tar products (shown in the Appendix), the major compositions have been quanti ed and summarized in Table 4.
Generally, O-heterocycles are highly corrosive to boilers, engines, and other equipment. Its yield is in uenced signi cantly by temperature. Baldwin et al. [34] summarized the variation of biomass tar with temperature changing and thought there were more O-heterocycles in the tar at 500~600 °C, such as furan and pentose. They also reported the tar, like the O-heterocycles, cracked distinctly with the rise of temperature. However, the O-heterocycles yield is also affected by steam. After the addition of steam, the proportion of total O-heterocycles increases, and forming two new derivatives (Appendix, 24, 21/43/71). It can be attributed to the fact that steam can facilitate the dissociation of these O-heterocycles, which are formed by the polysaccharide structure of MCC after a series of reactions, such as decarboxylation, decarboxylation, and so on [35,36]. In the reaction process, the compounds also crack into radical groups like the methyl radical to promote the substitution reaction and generate new derivatives.
Compared with complete pyrolysis, the steam stimulates the decline of open-chain compounds and increments alicyclic compounds in MCC pyrolysis tar signi cantly. However, the changing trend of alicyclic compounds is inversely proportional to the amount of steam. For example, with the continuous growth of steam, the cycloalkenes almost disappear after reaching the maximum value at S/F=0.8. Moreover, the rising rate of the cycloalkanes also slows down signi cantly. To be speci c, the increase in steam has two effects: some of the ole n, alkyne produced from MCC pyrolysis are further decomposed into noncondensing gases (such as CO, CO 2 , C1-C4) or may react with aromatic compounds to form new compounds. On the other hand, O/H/OH active radicals generated by steam can adhere with tar fragment [12,37], and produce many unsaturated open-chain compounds, which are likely to boost the forming of cyclohexene through the Diels-Alder reaction. At the same time, it is also accompanied by the hydrogenation reaction, which can further form cycloalkanes.
After the addition of steam, the proportion of aromatic compounds (the largest of the tar products) reduce slightly. For example, the yield of toluene is still over 10%, the same as during pyrolysis, while phosphonic acid, (p-hydroxyphenyl-) reduces and azobenzene disappears completely. In particular, the phosphonic acid, (p-hydroxyphenyl-) with multiple hydroxyls decreases by about 10% under the effect of steam; and this trend is more clearly with the increase of steam amount (S/F).
However, as shown in Fig 2, the trend of monocyclic aromatics compounds is enhancing with the amount of steam growing, and the yield of monocyclic aromatics compounds changes from 28.32% (S/F = 0) to 45.73% (S/F = 1.6). The amount of unsaturated monocyclic aromatics, which are the critical products in the pyrolysis-gasi cation process, is also proportional to steam. For example, alkenyl benzene increases signi cantly and reaches the maximum increments of 18.88% at S/F=1.6. However, it is challenging to generate polycyclic aromatic hydrocarbons (PAH) by the Diels-Alder reaction because the styrene is not a dienophile [38]. And the styrene is more likely to react with hydrogen radicals to form alkylbenzenes rather than cyclization and aromatization. It also can be proved by the increment of xylene and other alkylbenzenes in Fig 2. Signi cantly, the other alkylbenzenes increase considerably, which comes from 0.95% at S/F=0 to a maximum of 6.40% at S/F=1.2. Besides, monocyclic aromatics compounds, such as phenol, increase with the addition of steam and then slightly decrease at S/F=1.6, but all of these have a little uctuation. It indicates that phenols and its derivatives are stable, and only a part participates in the reaction [39,40]. Therefore, there are two possible pathways to reduce the yield of phenols: rstly, the phenols could be a precursor to form PAHs [41,42] , as shown in Fig 3; secondly, the reaction of substitution or polycondensation occurs.
Except for the toluene and phenol, naphthalene is also the most representative aromatic compound in biomass tar, whose molecular amount is the smallest among PAHs [43]. In combination with the previous analysis, naphthalene may be formed through a series of polymerization by the phenol precursor.
Naphthalene, unlike benzene, is more likely to participate in electrophilic substitution as its poor stability. Therefore, the naphthalene is inclined to react with the H radical to form naphthalene,1,2-dihydro-. Furthermore, steam can facilitate the formation of compounds with caged scaffolds (Appendix, 37/63, 38/90), but their yields are inversely proportional to steam. Furthermore, the anthracene derivatives (Appendix, 68/93) also increases with the amount of steam growing after the rst appearance at S/F=1.2.

The reaction mechanism of tar
Given the experimental results in 3.2, the Gibbs free energy of main tar components (yield > 4%) under pyrolysis conditions is summarized as listed in Table 5.
The results show that the phosphonic acid, (p-hydroxyphenyl-), yield =15.21%, has the lowest Gibbs free energy (-875.45883 Hartree), which means it has strong reactivity. And as shown in Table 4, phosphonic acid, (p-hydroxyphenyl-), is likely to react with steam, cracking and completely disappeared.
The CDD is widely used to predict the reaction sites [44], and the electrophilic reaction is more likely to happen when the value of CDD is smaller. As shown in Table 6, the O7 connected with a benzene ring is most likely to be attacked. The other atoms( C1, O11, C4, O10, H16, P8, and O9) are also possibly attacked, but the activity decreases in turn, which is consistent with the experimental results. Besides, there is a direct correlation between LBO and bond dissociation energy, which is also instrumental in predicting reaction sites. The LBOs between atoms are listed in Fig 5. Small LBOs of C4-C7, P8-O9, and P8-O10 indicate that the hydroxyl groups connected with C4 or P8 are probably to break. The following analysis will focus on the removal mechanism of the phosphonic acid group. Combined with the analysis of active sites, the possible reaction pathways are shown in Fig 6. R-1 and phosphonic acid may have a substitution reaction to form P-1 and H 2 O during cellulose pyrolysis (path 1). Moreover, R-2 is produced from cellulose pyrolysis, which can further react with OH radicals to form P-1 and H radical (path 2). But when the steam is added, the R-2 also probably reacts with H 2 O to form P-2 and phosphonic acid (path 3). Thus, there is the competition of OH radical and H 2 O to react with R-2. The Gibbs free energy of each possible reaction path is listed in eq. ( 1 )-( 3 ), all of these values are negative. It indicates that the reactions of different ways can be spontaneous. In particular, the Gibbs free energy of path 3 is -61.568kj/mol, which means that H 2 O is most likely to set a dominant position in competing with OH radicals for reacting with R-2. It also veri es the experimental result of cellulose pyrolysis tar that phosphonic acid, (p-hydroxyphenyl-) decreases sharply with the addition of steam.
The energy barriers of the three paths are listed in Fig 7, where also shows molecular structures (including reactants, transition states, products) involved in the reaction process. To further analyze the removal mechanism of the phosphonic acid group, this part will focus on the competition between H 2 O and OH radical for R-2 (path 2 and 3).
Path 2 is the reaction path of hydroxyl (O9-H17) attacking the P8-H19 bond of R-2 and then substituting the H19 atom. The bonds of O9-H17 and C1-P8 are slightly extended due to the mutual attraction between O9 and P8 atoms during the substitution of OH radical for H19 atom. At the same time, the H19 atom gradually separates from R-2 due to the offensive OH radical. When R-2 transforms to the transition state(TS5), the bond length of O9-H17 extends marginally from 0.96 Å to 0.981 Å, and the bond distance of the C1-P8 bond is stretched 0.05 Å to 1.85 Å, while the bond length of P8-H19 is lengthened from 1.410 Å to 1.488 Å. Besides, the angle between O9-P8 and P8-H19 is 58.93°, and the angle between P8-H19 and benzene ring plane is also 89.41°. All the changes of bond distance and angle promote the breaking of the P8-H19 bond and substitution of OH radical. Then P8 atom and O9 atom gradually form bond 7, the bond length shrinks to 1.61 Å (during the transition state that is 1.97 Å), and P8-H19 bond breaks, H19 gradually deviates from the benzene ring plane, C1-P8 bond also shrinks to 1.79 Å, nally forming a stable product system P-1.
Path 3 is the reaction process of H 2 O attacking the C1-P8 bond of R-2 and then substituting phosphonic acid. In the process of attacking the bond orbital, the H 2 O molecule gradually approaches the C1-P8 bond, and the bond length of H19-O18 stretches with the molecular angle expands. When the R-2 converters to transition state (TS3), the bond distance of H19-O18 extends to 1.44 Å, and the molecular angle extends from 109.50° to 113.69°. What is more, the bond length of C1-P8 is stretched from 1.80 Å to 2.08 Å, while the phosphonic acid group deviates from the benzene ring plane, and the C1-P8 bond forms an angle of 40.84 °with the benzene ring plane. All of these make for the substitution of the H atom. After that, the H19-O18 bond breaks to produce the H19 atom and OH radical (O18-H20). The H19 atom further attacks C1, and the P8 is also attacked by OH radical. In this case, the atom distance between C1 and H19, and between O18 and P8 is 1.22 Å and2.03 Å, respectively. And then, the C1-P8 bond breaks to forms C1-H19 and P8-O18 bond, which bond length shrinks to 1.09 Å and 1.62 Å, respectively. Finally, the compound forms a stable structure (P-2).

Conclusions
In summary, this paper addressed the effect of steam on cellulose pyrolysis tar in a two-stage xed bed. The primary findings are summarized as follows: (1) Adding steam can reduce the yield of tar contained in syngas from 6.68% to 2.30% availably. With the increase of steam, the element of oxygen in tar decreases while carbon increases.
(2) Steam mainly enhances the decomposition to form more stable compounds. When the cellulose pyrolysis, the phosphonic acid, (p-hydroxyphenyl-) can be  The conversion of compounds formed from the MCC pyrolysis tar with the effect of steam. ( the rst stage was 600℃, and the second stage was 800 ℃)