Understanding the mechanism of catalytic fast pyrolysis by unveiling reactive intermediates in heterogeneous catalysis

Catalytic fast pyrolysis is a promising way to convert lignin into fine chemicals and fuels, but current approaches lack selectivity and yield unsatisfactory conversion. Understanding the pyrolysis reaction mechanism at the molecular level may help to make this sustainable process more economic. Reactive intermediates are responsible for product branching and hold the key to unveiling these mechanisms, but are notoriously difficult to detect isomer-selectively. Here, we investigate the catalytic pyrolysis of guaiacol, a lignin model compound, using photoelectron photoion coincidence spectroscopy with synchrotron radiation, which allows for isomer-selective detection of reactive intermediates. In combination with ambient pressure pyrolysis, we identify fulvenone as the central reactive intermediate, generated by catalytic demethylation to catechol and subsequent dehydration. The fulvenone ketene is responsible for the phenol formation. This technique may open unique opportunities for isomer-resolved probing in catalysis, and holds the potential for achieving a mechanistic understanding of complex, real-life catalytic processes.

, and methyl-fulvenones (m/z = 106), along with literature reference spectra of the pure compounds. 6,7 For the latter species, no reference spectrum is available, and a FC simulation was not possible due to the pseudo-rotation of the methyl group upon ionization, leading to large amplitude motions. However, CBS-QB3 calculations revealed that two isomers contribute to the m/z = 106 ms-TPES, namely 1-, and 2-methyl-fulvenone. 13 C-labeling shifts the spectrum by one mass-to-charge unit, ruling out contributions from xylenes, which have the same unit molecular mass in the non-labeled experiments. The m/z = 108 ms-TPES consists of four different isomers of the composition C8H8O. Besides the three cresols (o-, p-, and m-methyl-phenol), the contribution of anisol (methoxy benzene) is revealed by the intense transition at 9.25 eV belonging to an electronically excited cation state.

Supplementary Note 1: Number of molecules per pulse, weighted hourly space velocity (W/F), and time-on-stream curves of guaiacol over H-USY
We have connected the pulsed valves to a gas mixing unit consisting of a gas tank, valves and tubing. The pressure in the gas tank changes during the experiment and we determine the number of molecules in view of the pressure change using the ideal gas law and the total tank volume: The number of molecules per gas pulse is obtained by dividing with the number of pulses, and was typically in the order of 1×10 16 Fig. 3 shows the time profile of the desorbed species as a function of time or amount of guaiacol feed into the reactor. Comparing the guaiacol signal strength of an empty versus H-USY-coated hot reactor, the signal drops by almost two orders of magnitude in the latter case, meaning that almost all guaiacol is adsorbed on the catalyst once the feed starts. At the beginning of the feed, the guaiacol signal remains fairly constant indicating a constant fraction of the pulse that passes through the reactor without being adsorbed on the catalyst surface. However, after around 150 min (> 25 μg guaiacol feed), the signal increases again, due to saturation of the catalyst surface by guaiacol or adsorbed reaction intermediates and products. In addition, deactivation of the catalyst can also contribute to the increasing guaiacol signal. Indeed, after 5-10 h time-on-stream, the catalyst turns strongly brownish or black indicating coking. Coking is known as one of the major deactivation mechanisms of the catalyst. 8,9 However, it also has positive effects, as discussed below.
Phenols and methyl-phenols, including anisole, are the major pyrolysis conversion products, and start to desorb after around 100 min (20 μg guaiacol feed), similar to the unreacted guaiacol. Since transalkylation often occurs under presence of methoxy groups it is obvious that methyl-phenols (and anisole) are yielded by methylation of phenol. 10 More apolar, and therefore less strongly bound non-oxygenated species, such as cyclopentadiene (c-C 5 H 6 ), methyl-cyclopentadienes, fulvene and benzene, desorb earlier from the reactive surface of the zeolite catalyst compared to the major polar products ( Supplementary Fig. 3, lower trace). The two ketenes, ethenone (H 2 C=C=O) and 6-fulvenone (c-C 5 H 4 =C=O) appear simultaneously with the nonoxygenated species. Catechol (m/z = 110) is the most strongly adsorbed species on the surface among the ones that can escape it, which may be explained by presence of two OH groups, both of which can form hydrogen bonds on the acid sites of the catalyst. Xylenols (m/z = 122) appear also at the same time as the major products, but at significantly smaller yields. The fact that similar molecules (phenol & cresols and cyclopentadiene & methylcyclopentadienes) with similar adsorption enthalpies on H-USY, desorb at almost the same time, also point to a desorption limited process. Thus, the TOS scans may not show the appearance and disappearance of intermediates exactly, and do not reflect the reaction pathways directly. However, they give information on the state and performance of the catalyst as a function of time. Mass spectrum taken at 10.5 eV using Na-USY @ 500°C, with 12 C-guaiacol as precursor. Apart from the much lower conversion compared to H-USY, the appearance of methyl radical (m/z = 15) along with cyclopentadienone (m/z = 80), both assigned based on their ms-TPES, is eye catching. Therefore, another radical pathway exists besides the acid catalyzed pathway driving transmethylation reactions and dehydration of catechol, as discussed in the main manuscript. This new pathway yields reactive species, such as methyl radicals. The reaction probably proceeds as follows.
After methyl abstraction, a hydrogen addition occurs yielding catechol, which can decompose in a Lewis acid catalyzed way yielding fulvenone (m/z = 92) and later phenol (m/z = 94). In addition, the hydroxylphenoxy radical (not observed) can lose hydrogen and decabonylate to cyclopentadienone (m/z = 80). Hydrogen addition on the other hand may be responsible for cyclopentenone formation (m/z = 82), which was also observed during the Na-USY experiments. The reaction steps are summarized in the lower part of the figure.

Supplementary Note 2: Additional Experiments: py-GC/MS with cyclopentadiene
Due to the presence of stable intermediates such as cyclopentadiene and methyl-cyclopentadiene in the py-iPEPICO setup, we have carried out a set of experiments to explain, why they evade detection in the py-GC/MS setup.
Firstly, the H-USY catalyst was loaded with guaiacol in the batch-type reactor and the helium flow rates through the reactor into the GC/MS were varied between 0.8 and 4.0 sccm. In addition, we have measured at several oven temperatures of the transfer line (175, 250 and 325 °C). Both changes did not lead to an enhancement of any five-membered ring species, but the conversion and selectivity remains identical.
Secondly, freshly prepared cyclopentadiene (from dicyclopentadiene at 180 °C) was directly injected onto the glass wool. Using the same conditions as in the experiments with H-USY and guaiacol, cyclopentadiene could be observed in the GC/MS system, without any further reaction. Only traces of the dimer, dicyclopentadiene, were detected.
Upon introducing H-USY catalyst, the amount of desorbed CP dropped by several orders of magnitude compared to the previous experiment, indicating strong adsorption on the catalyst. Besides benzene, toluene, xylenes, indane and naphthalene only trace amounts of cyclopentene, cyclopentane and methyl-cyclopentane have been detected, but the overall yield of CP is small. Increasing the concentration of CP on the catalyst did not lead to a higher CP signal, but to significant increase in the coke concentration, as indicated by color change of the used zeolite from brownish to black.
Thirdly, we wanted to investigate the lifetime and stability of cyclopentadiene in the reactive and catalytic environment. Thus, we have applied mixtures of guaiacol and cyclopentadiene (4:1) on the catalyst. Here, cyclopentadiene completely evades detection and we have found only stable intermediates and reaction products as compared to the experiments without CP. This observation proves that CP is stable enough to be transferred in the gas phase and to be detected by GC/MS, however, as soon as other reactants are present in the reaction mixture CP rapidly reacts further on the catalyst than being desorbed from the surface. The results confirm the difference in the two analytical setups: the py-iPEPICO has a very short residence time in the reactor and the catalyst, while the batch-type py-GC/MS has an increased residence time in the reactor (and on the catalyst) by several orders of magnitudes (20 seconds compared to a few µsec).
Supplementary Figure 6. Enthalpy diagram of the fulvenone hydrogen radical addition. After hydrogenation six isomers (1-6) of the composition C6H5O can equilibrate quickly and further stabilization can occur to yield either cyclopentadienyl (3 to 8) or phenoxy radicals (6 to 7). In addition the hydrolysis pathway of ketene to yield carbon dioxide and c-C5H6 is also depicted. Grey arrows denote effective reaction channels.