Pore Blocking by Phenolates as Deactivation Path during the Cracking of 4-Propylphenol over ZSM-5

: Cracking of propyl side chains from 4-propylphenol, a model compound for lignin monomers, is studied for a commercial ZSM-5 zeolite catalyst. The decline of 4-propylphenol conversion with time on stream can be delayed by co-feeding water. FTIR spectroscopy shows the formation of chemisorbed phenolates during reactions and signiﬁcant amounts of phenolics are detected by GC-MS of the extract from the spent catalysts. Thus, chemisorbed phenolates are identiﬁed as the main reason for deactivation in the absence of water. Regardless of the amount of co-fed water, substituted monoaromatics and polyaromatic species are formed. Comprehensive characterization of the spent catalysts including Raman and solid-state 27 Al NMR spectroscopy, and thermogravimetric analysis points to a combination of deactivation processes. First, phenolates bind to Lewis acid sites within the zeolite framework and hinder diffusion unless they are hydrolyzed by water. In addition, light oleﬁns created during the cracking process react to form a polyaromatic coke that deactivates the catalyst more permanently. Structure


Introduction
Bulk chemical production from biomass is one of the leading approaches to eliminating dependence on fossil carbon as feedstock for the chemical industry [1]. To improve the economics of biorefining, biomass utilization must take advantage of all available carbon [2,3]. Of the three major components of woody biomass (cellulose, hemicellulose, and lignin), the naturally aromatic structure of lignin makes it an attractive target for upgrading to chemicals [4]. To that end, research has focused on improving the extraction and conversion of lignin and its conversion into useful products for downstream upgrading [5][6][7][8][9][10]. The "lignin value prior to pulping" approach shows that limited modifications of industrial pulp and paper operations could advantageously generate a solvent-extracted lignin stream with sufficient quality for downstream processing [11]. Such a lignin could then be broken down by pyrolysis, acid digestion, or hydrogenolysis [4]. Alternatively, the lignin-first approach to second generation biorefinery operations uses reductive fractionation techniques to simultaneously extract and depolymerize the lignin from the native biomass [6,12]. These processes also generate a stream of cellulose and hemicellulose for further upgrading to ethanol or other sugar-based products [13].
Regardless of how lignin is extracted or broken down, lignin monomers are typically obtained as mixtures of similar aromatic oxygenates with a variety of different structures and functional groups. To convert such mixtures into useful replacements for petroleumderived additives and polymers, the fractionated lignin must be separated and upgraded. In most cases, chemical upgrading will involve some combination of hydrodeoxygenation and carbon chain cracking [4]. Both of these processes have been performed over zeolites or metals on zeolite support independently in the literature [14][15][16][17]. Recent work by Despite the significant practical impact of the deactivation problem, few have attempted to study this issue in detail. Often deactivation is attributed to coke formation on the catalyst surface, which can block access to micropores and active sites [23]. In their study of benzene conversion to phenol by photocatalysis over TiO2, Einaga et al. [26] hypothesized that the deactivation they observed was due to the formation of mono-oxygenate species on the surface which dimerize and eventually grow into deactivating polymers. Others have shown that some chemical structures, especially those involving phenolates, catecholates, and guaiacolates, can bind very strongly to surfaces [14,[27][28][29][30][31]. Popov et al. [32] suggested several possible binding modes for these phenolates on Lewis acids on alumina surfaces. Although usually known for their Brønsted acidic character, zeolites can also have a significant density of Lewis acid sites from extra-framework aluminum species [20]. We hypothesized the same types of surface species form during the upgrading of aromatics from lignin depolymerization, forming a kind of chemical "roadblock" that contributes to the fast deactivation of upgrading catalysts, as previously suggested for hydrodeoxygenation reactions [14]. Alternative explanations for the fast deactivation during propylphenol cracking include polymerization of propylene to form polyaromatic coke [33], and the leaching of active sites due to dealumination [20,21].
In this work, we aim to understand the cause of the fast deactivation of ZSM-5 during the cracking of propylphenol. Several common paths of zeolite deactivation are examined using a variety of techniques including FTIR spectroscopy, 27 Al MAS NMR spectroscopy, and thermogravimetric analysis. By fully understanding the root cause of this deactivation, engineering efforts can be focused on prevention to improve the productivity of next gen biorefineries.

Catalyst Characterization
The material properties of the ZSM-5 catalyst before the reaction were characterized by N2 physisorption, X-ray diffraction, and pyridine chemisorption followed by IR spectroscopy. Nitrogen physisorption measurements showed a BJH adsorption micropore Despite the significant practical impact of the deactivation problem, few have attempted to study this issue in detail. Often deactivation is attributed to coke formation on the catalyst surface, which can block access to micropores and active sites [23]. In their study of benzene conversion to phenol by photocatalysis over TiO 2 , Einaga et al. [26] hypothesized that the deactivation they observed was due to the formation of monooxygenate species on the surface which dimerize and eventually grow into deactivating polymers. Others have shown that some chemical structures, especially those involving phenolates, catecholates, and guaiacolates, can bind very strongly to surfaces [14,[27][28][29][30][31]. Popov et al. [32] suggested several possible binding modes for these phenolates on Lewis acids on alumina surfaces. Although usually known for their Brønsted acidic character, zeolites can also have a significant density of Lewis acid sites from extra-framework aluminum species [20]. We hypothesized the same types of surface species form during the upgrading of aromatics from lignin depolymerization, forming a kind of chemical "roadblock" that contributes to the fast deactivation of upgrading catalysts, as previously suggested for hydrodeoxygenation reactions [14]. Alternative explanations for the fast deactivation during propylphenol cracking include polymerization of propylene to form polyaromatic coke [33], and the leaching of active sites due to dealumination [20,21].
In this work, we aim to understand the cause of the fast deactivation of ZSM-5 during the cracking of propylphenol. Several common paths of zeolite deactivation are examined using a variety of techniques including FTIR spectroscopy, 27 Al MAS NMR spectroscopy, and thermogravimetric analysis. By fully understanding the root cause of this deactivation, engineering efforts can be focused on prevention to improve the productivity of next gen biorefineries.

Catalyst Characterization
The material properties of the ZSM-5 catalyst before the reaction were characterized by N 2 physisorption, X-ray diffraction, and pyridine chemisorption followed by IR spectroscopy. Nitrogen physisorption measurements showed a BJH adsorption micropore volume of 0.13 cm 3 /g ( Figure S1, Supplementary Materials). X-ray diffraction measurements indicated that the crystal structure was consistent with the MFI structure both before and after reaction ( Figure S2). SEM images of the catalyst post reaction clearly show pre-  Figure S3). Pyridine adsorption followed by transmission FTIR spectroscopy showed that the untreated ZSM-5 had a Brønsted acid site concentration of 400 µmol/g and a Lewis acid site concentration of 110 µmol/g. About a quarter of the pyridine on Lewis acid sites desorbed when the temperature was gradually raised to 450 • C, while essentially all Brønsted acid sites were strong enough to retain pyridine at this temperature ( Figure S4). These measurements are consistent with values found in the literature for this material [34].

Catalytic Cracking of 4-Propylphenol
The reactivity of 4-propylphenol over a commercial ZSM-5 catalyst in the presence of varying amounts of water matched well with previous reports in the literature ( Figure 2) [17,18,35]. In the absence of water, the catalyst quickly deactivated, approaching a steady state conversion of approximately 5-10% and phenol selectivity of around 40%. Interestingly, with even a small amount of water present, the conversion and phenol selectivity increased to 50% and 65%, respectively. Increasing the water content of the reactant stream from 3% to 9% increased conversion and selectivity, but further increase to 12% resulted in no measurable increase in performance. In the case of 9-12% water content, both the initial conversion and phenol selectivity were close to 100%, while the steady state conversion was 65% and steady state selectivity to phenol was 75%. Steady state operation with slow deactivation was reached after approximately 120 min on stream. The first order deactivation rate constants for all water contents are given in Table 1. There is a dramatic decrease in the deactivation rate constant between 0% water and 3% water followed by a less drastic but still significant decrease from 3% water to 12%. The steady state turnover frequency also decreased from 100 min −1 at 12% water to 41 min −1 at 3% water and 3.6 min −1 at 0% water ( Figure S5). volume of 0.13 cm 3 /g ( Figure S1, Supplementary Materials). X-ray diffraction measurements indicated that the crystal structure was consistent with the MFI structure both before and after reaction ( Figure S2). SEM images of the catalyst post reaction clearly show preserved MFI crystallites ( Figure S3). Pyridine adsorption followed by transmission FTIR spectroscopy showed that the untreated ZSM-5 had a Brønsted acid site concentration of 400 μmol/g and a Lewis acid site concentration of 110 μmol/g. About a quarter of the pyridine on Lewis acid sites desorbed when the temperature was gradually raised to 450 °C, while essentially all Brønsted acid sites were strong enough to retain pyridine at this temperature ( Figure S4). These measurements are consistent with values found in the literature for this material [34].

Catalytic Cracking of 4-Propylphenol
The reactivity of 4-propylphenol over a commercial ZSM-5 catalyst in the presence of varying amounts of water matched well with previous reports in the literature ( Figure  2) [17,18,35]. In the absence of water, the catalyst quickly deactivated, approaching a steady state conversion of approximately 5-10% and phenol selectivity of around 40%. Interestingly, with even a small amount of water present, the conversion and phenol selectivity increased to 50% and 65%, respectively. Increasing the water content of the reactant stream from 3% to 9% increased conversion and selectivity, but further increase to 12% resulted in no measurable increase in performance. In the case of 9-12% water content, both the initial conversion and phenol selectivity were close to 100%, while the steady state conversion was 65% and steady state selectivity to phenol was 75%. Steady state operation with slow deactivation was reached after approximately 120 min on stream. The first order deactivation rate constants for all water contents are given in Table 1. There is a dramatic decrease in the deactivation rate constant between 0% water and 3% water followed by a less drastic but still significant decrease from 3% water to 12%. The steady state turnover frequency also decreased from 100 min −1 at 12% water to 41 min −1 at 3% water and 3.6 min −1 at 0% water ( Figure S5).    TGA of spent catalysts showed three significant peaks in the derivative ( Figure 3). The first one below 150 • C is attributed to moisture loss from water adsorbed from the atmosphere during handling (not correlated with the water content of the reactant stream). A second small peak is observed around 250 • C and attributed to the desorption of bound surface species such as the chemisorbed molecules [36]. This peak is most prominent in the case with 0% water in the reactant stream. This sample also exhibited additional mass loss between 375 and 450 • C compared to the other catalysts. This indicates that both water (in this case supplied by adsorption from the atmosphere between experiments) and heat are required to drive off these species and that they are most plentiful in the sample which did not have water in the reaction stream. The largest peak, observed between 400 and 640 • C, is attributed to strongly chemisorbed species or recalcitrant carbon deposits. It has been shown that recalcitrant types of coke typically burn in this region, shifting to higher temperatures as they become more graphitic [37]. The total mass loss attributed to this region on a dry, coke free catalyst was approximately 6% in all cases. Based on these results, it appears that water does not significantly reduce the formation of carbon deposits.

Characterization of the Deactivated ZSM-5
TGA of spent catalysts showed three significant peaks in the derivative (Figure 3). The first one below 150 °C is attributed to moisture loss from water adsorbed from the atmosphere during handling (not correlated with the water content of the reactant stream). A second small peak is observed around 250 °C and attributed to the desorption of bound surface species such as the chemisorbed molecules [36]. This peak is most prominent in the case with 0% water in the reactant stream. This sample also exhibited additional mass loss between 375 and 450 °C compared to the other catalysts. This indicates that both water (in this case supplied by adsorption from the atmosphere between experiments) and heat are required to drive off these species and that they are most plentiful in the sample which did not have water in the reaction stream. The largest peak, observed between 400 and 640 °C, is attributed to strongly chemisorbed species or recalcitrant carbon deposits. It has been shown that recalcitrant types of coke typically burn in this region, shifting to higher temperatures as they become more graphitic [37]. The total mass loss attributed to this region on a dry, coke free catalyst was approximately 6% in all cases. Based on these results, it appears that water does not significantly reduce the formation of carbon deposits. To better understand the nature of these carbon species, Raman spectroscopy was performed on the most deactivated sample and compared to a deactivated ECat from a petroleum fluid catalytic cracking process ( Figure S6), which contained the typical bands of polyaromatic coke deposits [38]. The amount of observable coke species was negligible compared to the FCC sample, indicating that our ZSM-5 catalyst was essentially free from ordered or graphitic coke species. From this observation, we conclude that the deactivating species in this system are not polyaromatic and deactivation in the absence of water should not be attributed to hard coke.
Hydrofluoric acid digestion followed by DCM extraction and GC-MS analysis was performed to identify specific species responsible for deactivation in both the 0% water case and 12% water case, representing the most deactivated and least deactivated catalyst, respectively ( Figure 4). It is important to note that a negligible amount of DCM-insoluble coke was produced in both cases. Most of the DCM-soluble surface species were monocyclic aromatic compounds, primarily alkyl phenols and alkyl benzene, in both cases. In the case of 0% water, some bicyclic compounds, such as naphthalene and hydroxynaphthalene, were also observed. In a control experiment, diphenyl ether was subjected to the To better understand the nature of these carbon species, Raman spectroscopy was performed on the most deactivated sample and compared to a deactivated ECat from a petroleum fluid catalytic cracking process ( Figure S6), which contained the typical bands of polyaromatic coke deposits [38]. The amount of observable coke species was negligible compared to the FCC sample, indicating that our ZSM-5 catalyst was essentially free from ordered or graphitic coke species. From this observation, we conclude that the deactivating species in this system are not polyaromatic and deactivation in the absence of water should not be attributed to hard coke.
Hydrofluoric acid digestion followed by DCM extraction and GC-MS analysis was performed to identify specific species responsible for deactivation in both the 0% water case and 12% water case, representing the most deactivated and least deactivated catalyst, respectively ( Figure 4). It is important to note that a negligible amount of DCM-insoluble coke was produced in both cases. Most of the DCM-soluble surface species were monocyclic aromatic compounds, primarily alkyl phenols and alkyl benzene, in both cases. In the case of 0% water, some bicyclic compounds, such as naphthalene and hydroxynaphthalene, were also observed. In a control experiment, diphenyl ether was subjected to the same treatment as the spent catalysts. After digestion and DCM extraction, only diphenyl ether was observed in the GC/MS analysis.  In addition to understanding the nature of the surface species, the integrity of the zeolite frameworks of spent catalyst was examined by 27 Al MAS NMR spectroscopy (Figure 5). The samples were calcined and rehydrated prior to analysis to avoid an influence of carbonaceous deposits on the quadrupolar induced shifts and associated line broadening [39,40]. All catalysts showed peaks at 60 and 0 ppm representing aluminum atoms in tetrahedral and octahedral coordination, respectively. A shoulder was also observed at 45 ppm which indicates aluminum nuclei in a distorted tetrahedral or pentahedral environment [41,42]. Most aluminum nuclei in crystalline zeolites are present as tetrahedrally coordinated framework species with a limited contribution from octahedrally coordinated extra-framework species [43]. However, such octahedral aluminum can be formed by dealumination, especially at elevated temperature and in the presence of steam [43]. The deconvolution of the NMR spectra indicated a slight increase in the contributions from distorted, tetrahedrally coordinated Al species at the expense of Al in undistorted framework positions, with no apparent change in the contribution of octahedrally coordinated Al (Figure 5b,c). This suggests that the zeolite framework was only damaged by the reaction conditions to a minor extent, causing a distortion of local environments without significant dealumination. This distortion could also be caused by small amounts of residues that were not efficiently removed during calcination. It also indicates that the majority of active sites for the reaction were still present in all samples.  In addition to understanding the nature of the surface species, the integrity of the zeolite frameworks of spent catalyst was examined by 27 Al MAS NMR spectroscopy ( Figure 5). The samples were calcined and rehydrated prior to analysis to avoid an influence of carbonaceous deposits on the quadrupolar induced shifts and associated line broadening [39,40]. All catalysts showed peaks at 60 and 0 ppm representing aluminum atoms in tetrahedral and octahedral coordination, respectively. A shoulder was also observed at 45 ppm which indicates aluminum nuclei in a distorted tetrahedral or pentahedral environment [41,42]. Most aluminum nuclei in crystalline zeolites are present as tetrahedrally coordinated framework species with a limited contribution from octahedrally coordinated extra-framework species [43]. However, such octahedral aluminum can be formed by dealumination, especially at elevated temperature and in the presence of steam [43]. The deconvolution of the NMR spectra indicated a slight increase in the contributions from distorted, tetrahedrally coordinated Al species at the expense of Al in undistorted framework positions, with no apparent change in the contribution of octahedrally coordinated Al (Figure 5b,c). This suggests that the zeolite framework was only damaged by the reaction conditions to a minor extent, causing a distortion of local environments without significant dealumination. This distortion could also be caused by small amounts of residues that were not efficiently removed during calcination. It also indicates that the majority of active sites for the reaction were still present in all samples.  In addition to understanding the nature of the surface species, the integrity of the zeolite frameworks of spent catalyst was examined by 27 Al MAS NMR spectroscopy (Figure 5). The samples were calcined and rehydrated prior to analysis to avoid an influence of carbonaceous deposits on the quadrupolar induced shifts and associated line broadening [39,40]. All catalysts showed peaks at 60 and 0 ppm representing aluminum atoms in tetrahedral and octahedral coordination, respectively. A shoulder was also observed at 45 ppm which indicates aluminum nuclei in a distorted tetrahedral or pentahedral environment [41,42]. Most aluminum nuclei in crystalline zeolites are present as tetrahedrally coordinated framework species with a limited contribution from octahedrally coordinated extra-framework species [43]. However, such octahedral aluminum can be formed by dealumination, especially at elevated temperature and in the presence of steam [43]. The deconvolution of the NMR spectra indicated a slight increase in the contributions from distorted, tetrahedrally coordinated Al species at the expense of Al in undistorted framework positions, with no apparent change in the contribution of octahedrally coordinated Al (Figure 5b,c). This suggests that the zeolite framework was only damaged by the reaction conditions to a minor extent, causing a distortion of local environments without significant dealumination. This distortion could also be caused by small amounts of residues that were not efficiently removed during calcination. It also indicates that the majority of active sites for the reaction were still present in all samples.   27 Al MAS NMR spectra of deactivated ZSM-5 zeolites; (b) deconvoluted integral of the peak at 60 ppm representing 4-fold coordinated aluminum sites; (c) deconvoluted integral of the peak at 0 ppm representing 6-fold coordinated aluminum sites and the deconvoluted integral of the intermittent region representing aluminum sites neither fully tetrahedrally nor octahedrally coordinated. * denotes a sideband echo of the 60 ppm peak generated by the spinning

FTIR Spectroscopic Analysis of Adsorbed Phenol and Spent Catalysts
Infrared spectroscopy was used to understand which surface species play a key role in the deactivation pathway. As a reference, phenol was dosed onto a self-supported ZSM-5 catalyst at 50 • C in the gas phase and allowed to equilibrate before being heated to 350 • C ( Figure 5 and Figure S7). A band was observed at 1597 cm −1 corresponding to the quadrant vibration of the aromatic ring [44]. Bands at 1495 and 1474 cm −1 can be assigned to semicircle vibrations of the aromatic ring with additional contribution from OH deformation in the band at 1474 cm −1 [44,45]. The band at 1353 cm −1 can be attributed to C-O-H bending together with ring vibration [45,46]. All band areas decreased when the temperature increased because phenol desorbed. A relative decrease in the intensity of the band at 1474 cm −1 was observed along with an almost complete reduction of the band at 1353 cm −1 to the baseline.
Because both bands contain contributions from the C-O-H bond system, the reduction of the relative band area at higher temperatures suggests either dehydroxylation or deprotonation of the phenol group. As these bands are known to increase with increasing hydrogen bonding of the phenol group to other surface species or surface silanols [45], their decrease at higher temperatures could also be attributed to increasing isolation of the phenols from zeolitic hydroxyl groups. However, because the bands of both terminal silanols and bridging Brønsted acidic OH groups were clearly visible in the IR spectrum (Figure 6b), the former static at 3744 cm −1 and the later shifting from a perturbed position at 3500 cm −1 to unperturbed at 3611 cm −1 as phenol concentration decreases, reduced hydrogen bonding due to a lack of binding partners seems unlikely. Thus, the most plausible explanation is the dissociation of the phenolic OH bond and chemisorption of the resulting phenolate at elevated temperature.
Catalysts 2021, 11, x FOR PEER REVIEW 6 of 17 of the intermittent region representing aluminum sites neither fully tetrahedrally nor octahedrally coordinated. * denotes a sideband echo of the 60 ppm peak generated by the spinning

FTIR Spectroscopic Analysis of Adsorbed Phenol and Spent Catalysts
Infrared spectroscopy was used to understand which surface species play a key role in the deactivation pathway. As a reference, phenol was dosed onto a self-supported ZSM-5 catalyst at 50 °C in the gas phase and allowed to equilibrate before being heated to 350 °C (Figures 5 and S7). A band was observed at 1597 cm −1 corresponding to the quadrant vibration of the aromatic ring [44]. Bands at 1495 and 1474 cm −1 can be assigned to semicircle vibrations of the aromatic ring with additional contribution from OH deformation in the band at 1474 cm −1 [44,45]. The band at 1353 cm −1 can be attributed to C-O-H bending together with ring vibration [45,46]. All band areas decreased when the temperature increased because phenol desorbed. A relative decrease in the intensity of the band at 1474 cm −1 was observed along with an almost complete reduction of the band at 1353 cm −1 to the baseline.
Because both bands contain contributions from the C-O-H bond system, the reduction of the relative band area at higher temperatures suggests either dehydroxylation or deprotonation of the phenol group. As these bands are known to increase with increasing hydrogen bonding of the phenol group to other surface species or surface silanols [45], their decrease at higher temperatures could also be attributed to increasing isolation of the phenols from zeolitic hydroxyl groups. However, because the bands of both terminal silanols and bridging Brønsted acidic OH groups were clearly visible in the IR spectrum (Figure 6b), the former static at 3744 cm −1 and the later shifting from a perturbed position at 3500 cm −1 to unperturbed at 3611 cm −1 as phenol concentration decreases, reduced hydrogen bonding due to a lack of binding partners seems unlikely. Thus, the most plausible explanation is the dissociation of the phenolic OH bond and chemisorption of the resulting phenolate at elevated temperature. This hypothesis was further corroborated by reintroducing water vapor after treatment at high temperatures ( Figure S8). Adding water vapor to the system regenerated the relative intensities of the bands at 1474 and 1353 cm −1 . These bands then reduced in intensity again upon secondary heating to 300 °C, suggesting a reversible process which restores the phenol groups of the surface species. The reversible nature of this phenomenon in the presence of water may explain the fast and reversible deactivation observed during reaction. This hypothesis was further corroborated by reintroducing water vapor after treatment at high temperatures ( Figure S8). Adding water vapor to the system regenerated the relative intensities of the bands at 1474 and 1353 cm −1 . These bands then reduced in intensity again upon secondary heating to 300 • C, suggesting a reversible process which restores the phenol groups of the surface species. The reversible nature of this phenomenon in the presence of water may explain the fast and reversible deactivation observed during reaction.
It is also useful to directly compare the spectra of adsorbed model compounds to the deactivated catalysts to see if any patterns can be discerned in the surface species (Figure 7). The sharp band at 3744 cm −1 is attributed to ν(OH) of non-acidic surface silanols at the external surface or defects of the zeolite [47]. The slight decrease in intensity of this band after reaction indicates that some of these surface silanols are interacting with adsorbed species. The reduction of this band was less pronounced at high water content, which is consistent with hydrolysis of more accessible surface species on the external surface of the zeolite.
It is also useful to directly compare the spectra of adsorbed model compounds to the deactivated catalysts to see if any patterns can be discerned in the surface species ( Figure  7). The sharp band at 3744 cm −1 is attributed to ̅ ( ) of non-acidic surface silanols at the external surface or defects of the zeolite [47]. The slight decrease in intensity of this band after reaction indicates that some of these surface silanols are interacting with adsorbed species. The reduction of this band was less pronounced at high water content, which is consistent with hydrolysis of more accessible surface species on the external surface of the zeolite. Figure 7. Infrared spectra in the hydroxyl region of ZSM-5 after reaction with various water content (1-3), with gaseous phenol adsorbed (4) and post calcination (5) spectra included for reference.
In contrast, the band at 3612 cm −1 is attributed to free ̅ ( ) of Brønsted acidic groups that are predominantly within the micropores. [47] This group shifted to a lower wavenumber both after the addition of phenol and after the reaction, indicating that the acidic bridging hydroxides are no longer free but instead interacting with nearby organics. After six hours on stream, the deactivated catalysts exhibited different proportions of acidic hydroxides in the free and interacting state. In the case of the high water content reaction, all of the Brønsted acid sites are in the interacting state even after heating and evacuation, while in the absence of water more acid sites remain free. One possible explanation for this phenomenon is that, in the absence of water, diffusive barriers quickly form in the outer shells of the zeolite, preventing access to the inner acid sites and allowing them to remain in the free state.
In the C-H stretching region, pronounced aliphatic ̅ ( 2 ) and ̅ ( 3 ) stretching bands were observed between 2800 and 3000 cm −1 , with relatively smaller contributions from the ̅ ( ) vibrations of aromatic moieties at 3000-3100 cm −1 ( Figure S9). The ratio of ̅ ( 3 ) asymmetric stretching at 2964 cm −1 to ̅ ( 2 ) asymmetric stretching at 2932 cm −1 was approximately constant between all samples. Although not strictly quantifiable due do the different molar extinction coefficients of these vibrations, the consistency of this ratio indicated that the same proportions of alkyl groups are present in each sample. In other words, water does not appear to have an effect on the intrinsic rates of side chain cracking or propylene polymerization.
Significant changes between phenolated and deactivated catalysts were observed in the lower wavenumber part of the spectrum (Figure 8). New bands were observed at 1618, 1607, 1573, and 1512 cm −1 in the region typically attributed to aromatic ring vibrations [14]. In contrast, the band at 3612 cm −1 is attributed to free ν(OH) of Brønsted acidic groups that are predominantly within the micropores [47]. This group shifted to a lower wavenumber both after the addition of phenol and after the reaction, indicating that the acidic bridging hydroxides are no longer free but instead interacting with nearby organics. After six hours on stream, the deactivated catalysts exhibited different proportions of acidic hydroxides in the free and interacting state. In the case of the high water content reaction, all of the Brønsted acid sites are in the interacting state even after heating and evacuation, while in the absence of water more acid sites remain free. One possible explanation for this phenomenon is that, in the absence of water, diffusive barriers quickly form in the outer shells of the zeolite, preventing access to the inner acid sites and allowing them to remain in the free state.
In the C-H stretching region, pronounced aliphatic ν(CH 2 ) and ν(CH 3 ) stretching bands were observed between 2800 and 3000 cm −1 , with relatively smaller contributions from the ν(CH) vibrations of aromatic moieties at 3000-3100 cm −1 ( Figure S9). The ratio of ν(CH 3 ) asymmetric stretching at 2964 cm −1 to ν(CH 2 ) asymmetric stretching at 2932 cm −1 was approximately constant between all samples. Although not strictly quantifiable due do the different molar extinction coefficients of these vibrations, the consistency of this ratio indicated that the same proportions of alkyl groups are present in each sample. In other words, water does not appear to have an effect on the intrinsic rates of side chain cracking or propylene polymerization.
Significant changes between phenolated and deactivated catalysts were observed in the lower wavenumber part of the spectrum (Figure 8). New bands were observed at 1618, 1607, 1573, and 1512 cm −1 in the region typically attributed to aromatic ring vibrations [14]. The bands at 1618 and 1607 cm −1 were likely due to quadrant stretching of the 4-and 2substituted alkylphenols [44]. Interestingly, the phenol (monosubstituted) quadrant stretch was most visible after the reaction with 0% water. This finding is tentatively attributed to entrapment of products by fast pore plugging that would also lead to deactivation. The vibrations at 1573 and 1512 cm −1 were most prominent in the case of high water and decrease with increasing water content. The band at 1573 cm −1 is attributed to polyaromatic coke species, which formed to a larger extent when more water in the feed preserved the cracking activity for an extended time [14]. The band at 1512 cm −1 is assigned to the semicircle stretching mode of 4-propylphenol [44]. The bands at 1618 and 1607 cm −1 were likely due to quadrant stretching of the 4-and 2substituted alkylphenols [44]. Interestingly, the phenol (monosubstituted) quadrant stretch was most visible after the reaction with 0% water. This finding is tentatively attributed to entrapment of products by fast pore plugging that would also lead to deactivation. The vibrations at 1573 and 1512 cm −1 were most prominent in the case of high water and decrease with increasing water content. The band at 1573 cm −1 is attributed to polyaromatic coke species, which formed to a larger extent when more water in the feed preserved the cracking activity for an extended time [14]. The band at 1512 cm −1 is assigned to the semicircle stretching mode of 4-propylphenol [44]. Bands observed at 1467 and 1455 cm −1 are attributed to ̅ ( 3 ) asymmetric and symmetric deformations, respectively. The appearance of these bands in deactivated samples is consistent with the aliphatic nature of the side chain in propylphenol and the propylene product. The large band at 1381 cm −1 is attributable to ̅ ( ) of geminal methyl groups possibly formed by the reaction of propene with an aromatic ring to form an isopropyl side chain [48,49]. We also observe a decrease in the band associated with ( ) at 1350 cm −1 in all the samples, although due to interference from neighboring bands this may not be shown definitively.

Routes to Deactivation
Deactivation of zeolite catalysts is a well-known and studied phenomenon [38,[50][51][52][53][54]. In many processes for oil and gas conversion, the formation of carbonaceous deposits, such as polyaromatic coke species, is responsible for deactivation. These carbon deposits come in many forms and can deactivate the catalyst through several different routes including the covering of individual active sites, formation of transport barriers, and distortion of the zeolite framework [53]. In particular, at high temperatures and with highly unsaturated feedstocks, this coking can deactivate zeolites extremely quickly. Commercial FCC catalysts are fully deactivated by coke deposition after only a few seconds on stream at 500-550 °C and require a high temperature calcination step to regenerate [54]. Often the activity level after regeneration is not the same as the activity before the initial deactivation because the combination of heat and steam during the calcination steps can lead to Bands observed at 1467 and 1455 cm −1 are attributed to ν(CH 3 ) asymmetric and symmetric deformations, respectively. The appearance of these bands in deactivated samples is consistent with the aliphatic nature of the side chain in propylphenol and the propylene product. The large band at 1381 cm −1 is attributable to ν(CH) of geminal methyl groups possibly formed by the reaction of propene with an aromatic ring to form an isopropyl side chain [48,49]. We also observe a decrease in the band associated with δ(OH) at 1350 cm −1 in all the samples, although due to interference from neighboring bands this may not be shown definitively.

Routes to Deactivation
Deactivation of zeolite catalysts is a well-known and studied phenomenon [38,[50][51][52][53][54]. In many processes for oil and gas conversion, the formation of carbonaceous deposits, such as polyaromatic coke species, is responsible for deactivation. These carbon deposits come in many forms and can deactivate the catalyst through several different routes including the covering of individual active sites, formation of transport barriers, and distortion of the zeolite framework [53]. In particular, at high temperatures and with highly unsaturated feedstocks, this coking can deactivate zeolites extremely quickly. Commercial FCC catalysts are fully deactivated by coke deposition after only a few seconds on stream at 500-550 • C and require a high temperature calcination step to regenerate [54]. Often the activity level after regeneration is not the same as the activity before the initial deactivation because the combination of heat and steam during the calcination steps can lead to structural changes and dealumination of the zeolite [38]. During the conversion of oxygenates, which are common in biomass derived feedstocks, the formation of strongly chemisorbed deposits also needs to be considered [14,55,56].
In the present study, polyaromatic deposits were not observed in Raman spectra ( Figure S6) or extraction of deposits by HF digestion (Figure 4) and only a small amount of polyaromatic vibration was observed in the infrared spectra of spent ZSM-5 ( Figure 8). Therefore, a deactivation path similar to the oligomerization and aromatization of carbocations to coke during the FCC process can be ruled out as the primary source of deactivation over the course of 6 h on stream. However, this type of deactivation still appears to contribute to slower deactivation observed when the presence of water increases the turnover number per acid site of the catalyst. As water limits the fast deactivation, a larger number of reactive species form and polymerize into polyaromatic coking species that, eventually, deactivate the catalyst.
Polyaromatic coke is not the only way that zeolites can lose activity. In the presence of water, Si-O-Al bonds can be hydrolyzed leading to dealumination, loss of Brønsted acid sites and, consequently, deactivation of the catalyst [20,57]. It is plausible that this process could happen when water is cofed during a reaction at elevated temperature as in the present study. However, based on 27 Al MAS NMR spectra ( Figure 5), this process does not occur to a dramatic extent during 6 h on stream at 350 • C. While framework distortion was observed to a limited extent, the change in intensity of tetrahedrally coordinated aluminum species was not significant enough to explain the observed change in reactivity. However, dealumination may cause slow deactivation if the catalyst is used for much longer times on stream.
Because dealumination and coking are shown to not be the main reasons for the observed deactivation, the most plausible remaining theory is the formation of strongly chemisorbed or bulky molecules with moderate size that poison active sites and/or create diffusional barriers within the micropores of the catalyst. Considering that the maximum sphere diameter within the ZSM-5 structure (6.36 Å) [58] is only slightly larger than the kinetic diameter of a benzene ring (5.85 Å) [59], it is to be expected that rigidly chemisorbed species, such as phenolates, dimers or even other propylphenol isomers can create diffusional barriers that prevent reactants from reaching active sites that are located deeper in the pores. This mode of deactivation is supported by the observation of the OH stretching band of Brønsted acid sites that are not hydrogen bonding to any molecular species after the reaction in the absence of water (Figure 7, trace 3). Thus, the diffusional barriers form rapidly in the absence of water and prevent some active sites from ever interacting with the reactant. Given the possibilities of molecular species to create diffusional barriers in microporous systems, understanding the nature of possible pore blocking species is critical (vide infra).

Nature of Species Causing Pore Blockage
The formation of diffusional barriers in the present study ( Figure 2) and previous publications [17,18,35,55,60] can be explained by the formation of rigidly chemisorbed species or molecules that are too large to diffuse through the windows of the ZSM-5 structure. Given the sizes of these cages (6.36 Å compared to the kinetic diameter of pxylene at 5.85 Å and o/m-xylene at 6.8 Å) [58,59] even an isomerization of 4-propylphenol to 2-propylphenol or 3-propylphenol would create species that occupy a significant part of a cage that they cannot leave. However, the ability of water to remove the diffusional barriers allows for ruling out a major impact of isomerization products.
Liao et al. recently suggested that the formation of diphenyl ethers as the main cause for deactivation [35]. It is plausible that co-fed water would be able to hydrolyze these species forming products that are capable of diffusing out of the zeolite framework. At first sight the presence of diphenyl ether seems to be consistent with their observation of IR bands at 1580 and 1497 cm −1 . However, similar bands are also observed in the spectra of phenol and other monosubstituted aromatics. In the present study, the spectra of most spent catalysts contained a band at 1573 cm −1 , which is red shifted with respect to reference diphenyl ether, which typically shows a band between 1595 and 1580 cm −1 [61]. After reaction with a high water content, the band at 1573 cm −1 did appear, but this is equally well explained by hydrolysis of other surface species. Additionally, if this band was caused by diphenyl ether, one would expect lower band intensity on catalysts deactivated with more water in the reactant stream, which is opposite the observed trend. More importantly, diphenyl ethers were not detected in the DCM extracts of the spent catalyst after HF digestion, while a reference experiment indicated that diphenyl ether was not degraded during the workup of the sample. Thus, a significant impact of diphenyl ethers on the deactivation of the catalysts is unlikely.
Based on the IR spectra of spent catalysts and adsorbed phenol (Figure 8), it is concluded that significant amounts on phenolates form during reaction in the absence of water. These species strongly chemisorb and thus deactivate the catalyst. Interestingly, water had a minimal effect on the types of surface species observed after reaction (Figure 4), but it is capable of removing them. The most logical explanation for this is that phenolate species are readily hydrolyzed, so that the resulting phenols can be removed. It has been shown in the literature that phenols can strongly bind onto Lewis acid sites, such as extra framework aluminum species and defect sites of zeolites [14,45,60]. Brønsted acid sites are less likely to be affected by these species because the acidic proton should readily restore the phenol group instead of forming a surface ether.

Effect of Propylene Polymerization
The formation of polyaromatic coke is a common path of catalysts deactivation during processes involving light olefins [51]. The formation of these species involves acid catalyzed oligomerization and aromatization reactions. Their presence is indicated by a characteristic IR band around 1575 cm −1 [62,63]. Based on the IR spectra of deactivated catalysts (Figure 8), the band attributed to polyaromatic coke is most significant after the reaction in the presence of a high amount of water. It is suggested that water continuously removes diffusional barriers in the zeolite micropores to increase the turnover numbers of sites inside of catalyst particles. As these sites keep producing propylene, there is a certain probability for oligomerization and aromatization reactions to occur resulting in coke precursors and polyaromatic coke. These coke species will accumulate and eventually fully deactivate the catalyst in a similar way to FCC catalysts. It has been shown in the literature that these kinds of deposits significantly hinder diffusion through the micropores of the catalyst [64]. This blocks access to active sites and significantly increases the length of the diffusive path, decreasing overall activity [65,66].

Implications for Catalyst Design of Biorefining Processes
Catalyst deactivation is a serious threat to the profitability of future biorefining processes. As the present study shows, oxygen-containing aromatics are prone to deactivating zeolite catalysts by chemisorbing as phenolates or similar species (e.g., catecholates) that quickly prevent reactants from reaching active sites within the micropores. It has been shown that these phenolate species form primarily on Lewis acid sites [14,45]. Thus, limiting the Lewis acid content of zeolites, either through careful manufacturing or framework healing, can prevent this kind of adsorption [57]. In addition, the impact of phenolates and diffusional barriers in general is reduced in mesoporous or hierarchical zeolites. The wider channels in the hierarchical structure allow for fast transport into and out of the core of zeolite crystals so that each diffusional barrier blocks a smaller fraction of the micropore volume [19,67,68]. The impact of barriers such as phenolates will also be reduced in large pore zeolites, but these frameworks are less effective at imposing shape selectivity and will stabilize adsorbed reactants and transition states in a different manner. Moreover, such zeolites will also provide additional space for polyaromatic coke species to form in the cages, potentially decreasing the time between regenerations. In the present study, thermogravimetric analysis ( Figure 3) showed that harsh calcination conditions are not required to regenerate ZSM-5 zeolites, provided that the concentration of coke precursors remains limited.
The need to dry bio-oil streams before further upgrading has also been discussed in the literature [69,70]. However, this work shows the critical role water plays in preventing the adsorption of pore blocking species. In their study, Xu et al. claim that water affects the alkylation of phenol over HBEA in two ways, through the creation of additional Brønsted acid sites and through interaction with reactive intermediates [60]. NMR spectra collected in this study corroborate their conclusion that dilute water vapor does not have a profound impact on zeolite dealumination under reaction conditions ( Figure 5). This means that bio-oil does not need to be fully dried before reaction, as a small amount of water "contaminants" would prove beneficial as a cofeed. This insight points to opportunities for decreasing separation costs while simultaneously reducing the frequency of catalyst regeneration cycles.

Catalyst Characterization
Nitrogen physisorption data was collected on a Micromeritics ASAP2020 physisorption/chemisorption analyzer. Catalysts were degassed at 250 • C at a pressure of <2 µm Hg pressure for 6 h prior to analysis. The micropore volume was calculated based on the adsorption branch of the isotherm using the BJH method [71]. X-ray diffraction of calcined ZSM-5 was performed on a PANalytical Alpha-1 using a Cu K α source and an X'Celerator detector between 4 and 40 • 2θ. SEM imaging was performed on a Hitachi SU8010 ultra-high resolution scanning electron microscope at 3.0 kV. Acid site concentrations were determined by pyridine adsorption followed by FTIR spectroscopy. Pyridine was introduced to the sample at 0.1 mbar and allowed to equilibrate for 30 min at 150 • C. The pressure was then reduced below 10 −3 mbar for one hour to desorb physisorbed and weakly chemisorbed pyridine. The temperature was then increased first to 250 • C, then to 350 • C, and finally 450 • C under vacuum for one hour to desorb pyridine. Samples were then returned to 150 • C, and IR spectra were collected after each desorption step. The characteristic bands of pyridine adsorbed on Brønsted acid sites (1540 cm −1 ) and Lewis acid sites (1440 cm −1 ) were integrated, and these band areas were then converted to surface site concentration based on the Lambert-Beer equation, as is common in the literature [34,72]: where A BAS is the integrated absorbance of the band corresponding to pyridine bound to Brønsted acid sites in units cm −1 , S is the area density of the pressed catalyst wafer in units of g·cm −2 , and ε BAS is the molar extinction coefficient of pyridine bound to Brønsted acid sites. The molar extinction coefficients used in this study were 1.23 cm/µmol for the pyridinium ion bound to Brønsted acid sites and 1.73 cm/µmol for pyridine adsorbed on Lewis acid sites as reported by Tamura et al. [34].

Catalytic Cracking Reactions
Reactivity studies were performed in a custom packed bed reactor setup with online GC sampling. A 0.25 in. outer diameter stainless steel tube was used as the reactor. The temperature was controlled by using a Eurotherm 2416 temperature controller. The reactor was packed with quartz wool followed by 0.2 g of pressed ZSM-5 positioned in the center of the reactor. The catalyst particle size was 90-212 µm. An inert layer of 0.2 g silicon carbide (35-75 µm) was then added directly on top of the ZSM-5 to promote the evaporation of the liquid reactants and gas mixing. Liquid 4-propylphenol was fed through 1 /16 in. tube by a 500D Teledyne ISCO syringe pump directly onto the heated silicon carbide. Water was supplied upstream of the reactor by an 8 mL KDScientific Legato 100 syringe pump and vaporized over inert quartz wool.
Experiments were performed at 350 • C under a constant flow rate of 80 mL/min of nitrogen. The catalyst was heated at~10 • C/min and pretreated at the reaction temperature under inert flow of nitrogen for at least 30 min before liquid injection began. Effluent lines were heated to 300 • C to prevent condensation of products that were analyzed on stream by an Agilent 7890A gas chromatograph equipped with an HP-5 column and a flame ionization detector. Conversion and selectivity were calculated as: where, C 0,pph is the concentration of 4-propylphenol in the inlet, C f,pph is the concentration of 4-propylphenol in the outlet, C x is the concentration of the product of interest and C i is the concentration of each of the products containing aromatic rings. Deactivation rate constants were calculated assuming a first order decay model: where X is the conversion of 4-propylphenol at time t, X 0 is the initial conversion, and k d is the first order decay constant. More complex deactivation models were excluded so as to not overfit the data. After the reaction, the flows of all liquid streams were stopped to desorb loosely bound surface species, reactants, and products from the catalyst in the same carrier gas stream of 80 mL/min of N 2 for 30 min. The reactor was then cooled to room temperature overnight in static atmosphere, and the catalyst was collected the following morning for further analysis. ZSM-5 catalyst pellets were separated from SiC using a 90 µm sieve before being further characterized.

FTIR Spectroscopy of Adsorbed Phenol and Carbonaceous Deposits
Fourier transform infrared spectroscopy (FTIR) was performed using a Thermo Nicolet 8700 spectrometer with an MCT/A detector in transmission mode through a custom high vacuum cell. Catalysts were pressed into wafers (5-20 mg) and held in the beam path by a temperature-controlled heating coil and aluminum cups inside a high vacuum cell with ZnSe windows. Fresh ZSM-5 samples were pretreated at 450 • C high vacuum to drive off water and physisorbed species and cooled to 50 • C for data collection. Post-reaction catalysts were activated at 350 • C (i.e., reaction temperature) before collecting spectra at 50 • C. Spectra were collected as the sum of 64 scans at 4 cm −1 resolution. Phenol was introduced to untreated ZSM-5 in the gas phase at 0.2 mbar and 50 • C. After equilibration, the sample was evacuated before being heated to 350 • C to observe changes of adsorbed species. Water was then introduced at 50 • C, and the sample was heated to 300 • C.

Thermogravimetric Analysis
Thermogravimetric analysis was performed in a TA Instruments simultaneous DSC-TGA Q series 600 instrument. Samples between 5 and 10 mg were loaded into alumina cups and allowed to equilibrate in 20 mL/min of flowing dry air until the mass was stable, typically 10-20 min. The temperature was then ramped at 10 • C/min to 800 • C under the same flowrate of air and held at 800 • C for 30 min. Weight loss below 200 • C is attributed to moisture loss.

Raman Spectroscopy
Raman spectra were collected on a Renishaw Qontor inVia Raman spectrometer using a blue 488 nm excitation laser source with 0.3-3 mW of power. Samples were prepared by placing the spent catalyst onto a glass microscope slide and manually focusing the optical microscope. Spectra were acquired at 5× magnification in the range of 105-3200 cm −1 . Deconvolution was done as suggested by Sadezky et al. with D1-D4 bands at 1350, 1600, 1500, and 1200 cm −1 , respectively, and a G band at 1585 cm −1 [73].

Hydrofluoric Acid Digestion
HF digestion of deactivated ZSM-5 catalysts, followed by liquid-liquid extraction was performed following the procedure proposed by Guisnet and coworkers [74]. Deactivated catalyst (100 mg) was mixed with 10 mL of HF acid (40 wt% in water) and stirred at room temperature for 12 h in a polypropylene vessel to fully dissolve the zeolite. Liquid-liquid extraction was performed by adding 3 mL of dichloromethane (DCM) to the HF solution and thoroughly mixing. After settling, this mixture separated into a water phase and an organic phase. The soluble fraction of the surface deposits entered the lower DCM phase, while any insoluble deposits floated on top of the water phase. Both phases were then passed through a 0.2 µm nylon filter tipped syringe to collect any solid deposits. The DCM phase was then separated by pipetting, concentrated to 1 mL, and analyzed on a Shimadzu QP2010 GC-MS using an Agilent HP-5 column to identify soluble species. Filter tips were removed and allowed to dry before collecting any solid insoluble species, if present.

Solid State 27 Al NMR Spectroscopy
Samples were prepared for NMR spectroscopy by calcining at 550 • C for 6 h in dry air. Then, the samples were left in a closed vessel with a flask of water overnight to hydrate fully. 27 Al MAS NMR spectroscopy was performed on a 400 MHz Bruker Avance III instrument with a 9.4 T magnetic field. Approximately 0.1 g of catalyst was packed into 4 mm zirconia rotors and spun at 10 kHz before sampling. Scans were taken by 0.6 ms pulse followed by a 1 s decay time. The total number of scans for each sample was 2048. Integration was performed with DMfit2015 using the Q 1/2 mas peak model ( Figure S10).

Conclusions
In this work, we examine the mechanism of fast, reversible deactivation of ZSM-5 during the cracking of 4-propylphenol. We propose a two-step deactivation mechanism: First, phenols quickly bind to Lewis acid sites creating diffusional barriers to parts of the zeolite framework. Reactivity studies combined with the analysis of deposits on the catalyst showed that a sufficient amount of water in the feed stream allows for removing the phenolates by hydrolysis, which eliminates their negative impact on catalytic activity. If the impact of phenolates is sufficiently reduced, slow deposition of polyaromatic species derived from polypropylene eventually causes deactivation.
FTIR spectroscopic analysis of deactivated catalysts showed significant contributions from mono substituted aromatics as well as 2-and 4-disubstituted rings. This result is consistent with the proposed theory of strongly bound phenolates as the primary cause of deactivation. Secondary deactivation due to coke formation was most apparent when the cracking activity remained high over an extended time on stream. Further characterization of spent catalysts showed little structural degradation despite the fact that steam is known to harm zeolite frameworks. This suggests that future biorefineries need not include energy intensive pre-drying steps before further upgrading lignin oils.