Operando setup for hydrodeoxygenation of 4-propylguaiacolThe hydrodeoxygenation of 4-propylguaiacol (4PG) was investigated at the VUV beamline of the Swiss Light Source by detecting intermediates and products using operando PEPICO spectroscopy, which combines mass spectrometry and photoelectron spectroscopy. The acquired photoion mass-selected threshold photoelectron spectra (ms-TPES) enable the isomer-selective assignment of elusive and stable species in complex reaction systems32,33,34. The PEPICO endstation and reactor setup used in this work are shown in Fig. 2. We introduced H2 and 4PG over the catalyst in three distinct experiments: solvent-free, as well as co-feeding isooctane or tetrahydrofuran (THF, oxygen-containing) as the solvent. A mixture of H2, solvent, and 4PG vapor passed through a heated microreactor containing a commercial Ru/C catalyst bed at ~1.5 bar pressure. Intermediates and products exiting the microreactor expanded into high vacuum (10−4 mbar) and formed a molecular beam (MB), in which reactive collisions are quickly suppressed. After skimming, the MB entered the ionization vacuum chamber (10−6 mbar) and was ionized by monochromatic vacuum ultraviolet (VUV) synchrotron light. The photoions and electrons were detected in delayed coincidence, which enabled us to record mass spectra (MS) and photoion mass-selected threshold photoelectron spectra (ms-TPES) of all products and intermediates desorbing from the catalyst surface. This dataset is unique as it allows for the isomer-selective identification of products and elusive intermediates, which are otherwise difficult to detect and assign by alternative analytical techniques.Fig. 2: Experimental setup.Schematic of the operando PEPICO spectroscopy setup for 4PG hydrodeoxygenation in various solvent systems.Hydrodeoxygenation of 4-propylguaiacol in a solvent-free environment and in isooctane4PG hydrodeoxygenation (HDO) was first conducted solvent-free. Ru/C catalyst (10 mg) was pretreated in Ar for 30 mins at 200 °C in the microreactor. After cooling the catalyst down to 40 °C, we introduced 4PG vapor in a 10 sccm H2 flow. During the first 1 h (Fig. 3a), no discernible 4PG or HDO product photoionization mass spectral (MS) signal could be observed. Note that hydrogen clearly exited the reactor as observed by the vacuum pressure increase upon turning on the flow, but spectra were recorded below the H2 ionization energy of AIE = 15.426 eV35. Only 4PG peak was seen at m/z 166 prior to 4 h, indicating the saturation of the Ru/C catalyst with 4PG but without formation of other products (Fig. 3b). After 5–10 h time on stream, three major peaks at m/z 172, 142, and 124 began to emerge and gradually intensify. After 10 h at 40 °C, we raised the reaction temperature stepwise to 200 °C (Fig. 3c). At 80 and 120 °C, the product distribution remained like that observed at 40 °C. However, a new peak at m/z 126 showed up at 150 °C and became more prominent as the temperature reached 200 °C. Further peaks, such as at m/z 98 and 114, also emerged. To identify these newly formed species, we recorded their photoion mass-selected threshold photoelectron spectra (ms-TPES) and calculated the adiabatic ionization energies of the potential spectral carriers (Table S1). The peak at m/z 172 could be assigned to 2-methoxyl-4-propylcyclohexanol (2MPC), which is the product of complete benzene ring hydrogenation of 4PG (Fig. 1). The peak at m/z 142 is assigned to ionization of neutral 4-propylcyclohexanol (4PC, m/z 142), which also yields the peak at m/z 124 after dissociative ionization by water loss at the incident VUV energy of 10.5 eV. To provide evidence for this assignment, we recorded reference mass spectra (Fig. S1) and ms-TPES (Fig. 3d) of 4PC. The formation of 4PC is attributed to demethoxylation of 2MPC, which yields methanol as co-product as evidenced by the ms-TPES of m/z 32 (Fig. S2a). The peak at m/z 126 can be assigned to propylcyclohexane (PC) based on its ms-TPES, formed after dehydroxylation of 4PC (Fig. 3e). Since PC is only formed above 150 °C, dehydroxylation of 4PC to PC exhibits a slightly higher activation energy than 4PG ring saturation and demethoxylation to PC over Ru/C. The peaks at m/z 98 and 114 were assigned to methylcyclohexane (MC) and 4-methylcyclohexanol (4MC), respectively (Figs. S2b, 3f), suggesting that removal of a propyl group can also take place at a higher temperature by deethylation.Fig. 3: Operando PEPICO for 4PG hydrodeoxygenation without solvent co-feed.a photoionization mass spectra as a function of time on stream at 40 °C. b Individual photoionization mass spectra from (a) shown for comparison. c Photoionization mass spectra at different reaction temperatures; the photon energy used in (a–c) is 10.5 eV; (d–f) ms-TPES of (d) m/z 142, (e) m/z 126, (f) m/z 114 at 150 or 200 °C. g proposed reaction routes in non-solvent and isooctane systems. Reference spectra and FC simulations are marked by arrows. Reaction conditions: 10 mg of Ru/C, 10 sccm H2, 4PG dipped in quartz wool was maintained at 70 °C, ~1.5 bar reactor pressure. (0.04% 4PG, 6% for isooctane, rest H2).Next, we investigated 4PG HDO in the presence of isooctane (Fig. S3). Under low reaction temperatures (80 or 100 °C), we observed the same products as in the solvent-free experiment: 2MPC at m/z 172 and 4PC at m/z 142 together with its dissociative ionization fragment at m/z 124 (Fig. S4). As the reaction temperature was increased, PC at m/z 126 emerged as the dehydroxylated product at 150 °C, becoming more pronounced at 200 °C. The signal of isooctane at m/z 99 and 114 coincided with the deethylated MC and 4MC signals (Fig. S4), rendering the identification of the deethylation reaction products in isooctane challenging. As supported by ms-TPES analysis (Fig. S5), we confirm that the reaction pathways are not substantially influenced by the introduction of isooctane, and the 4PG HDO reaction mechanism remains unchanged. In summary, the 4PG HDO mechanism exhibits three regimes both in solvent-free and isooctane experiments depending on the reaction temperature (Fig. 3g): (1) at low temperatures (40–120 °C), the benzene ring is saturated and subsequently demethoxylated; (2) at medium temperatures (150 °C) sequential dehydroxylation takes place; (3) while at high temperatures (>200 °C) a deethylation reaction is observed. All products are ring-saturated, highlighting the previously reported favorable planar adsorption configuration on noble metal catalysts, where the benzene ring lies parallel to the metal surface9. This adsorption configuration favors initial benzene ring saturation and makes it nigh impossible to preserve the inherent and valuable aromaticity of lignin in HDO.Hydrodeoxygenation of 4-propylguaiacol in tetrahydrofuranOxygen-containing solvents bind more strongly to noble metal surfaces than hydrocarbon solvents, possibly due to the oxygen lone pair similar to strong H2O adsorption on Ru36, often resulting in suppressed reaction rates due to the competitive binding of the solvent to the active sites24. To test whether solvent binding also affects the reaction mechanism, we conducted 4PG HDO experiments using tetrahydrofuran (THF, 72 amu), a solvent also derived from biomass, analogously to the non-solvent and isooctane experiments. Upon introducing H2 flow alongside 4PG and THF vapors at 80 °C, the 4PG signal at m/z 166 appeared already at 20 min time on stream (Fig. 4a), much faster than in the solvent-free (Fig. 3a) and isooctane systems (Fig. S4). This suggests more rapid surface saturation due to substantial coverage of the Ru sites with THF. Notably, no new species form even at extended reaction times at 80 °C (Fig. 4a, b), indicating that THF blocks the active sites and quantitatively inhibits 4PG HDO at this temperature. Upon raising the temperature to 150 °C, major peaks at m/z 136, 124, and 120 were observed (Fig. 4c). Further increasing the temperature to 250 °C did not significantly affect the peak at m/z 124, but peak intensities at m/z 136 and 120 continued to grow. A minor peak could be identified at m/z 142 and 172 at reaction temperatures of 150 and 200 °C. The m/z 124 peak arose from 4PC dissociative ionization, just as in the non-solvent and isooctane systems (Fig. S6). The lower intensity peaks at m/z 172, 142, and 124 indicated that the predominant 4PG HDO to 4PC route via 2MPC in the solvent-free and isooctane systems is suppressed in the presence of THF. In contrast, the formation of m/z 136 and 120 is more pronounced. Based on their respective ms-TPES (Fig. 4d, e), these two peaks were identified as 4-propylphenol (4PP) and propylbenzene (PB), respectively. This indicates that demethoxylation and dehydroxylation reactions take place prior to benzene ring saturation in the presence of THF, representing a new 4PG HDO reaction pathway (Fig. 4g). Even at temperatures as high as 250 °C, the reactant 4PG signal remained high, indicating lower reaction rates in THF, consistent with previous findings24.Fig. 4: Operando PEPICO for the 4PG hydrodeoxygenation in THF.a photoionization mass spectra as a function of time on stream at a reactor temperature of 80 °C. b Selected photoionization mass spectra from (a) at different times. c Photoionization mass spectra as a function of reactor temperature; the photon energy used in (a–c) is 10.5 eV; (d–f) ms-TPES of (d) m/z 136, (e) m/z 120, (f) m/z 126 at 250 or 225 °C. g Proposed reaction routes in 4PG HDO in the presence of THF. Franck–Condon simulated spectra are marked by arrows. Reaction conditions: 10 mg of Ru/C, 10 sccm H2, 4PG dipped on quartz wool was maintained at 70 °C, ~1.5 bar pressure. (0.04% 4PG, 21% THF, rest H2).While the signal at m/z 126 could be assigned to PC based on its ms-TPES (Fig. 4f), the peaks of the aromatic 4-propylphenol (4PP) and propylbenzene (PB) dominate the product distribution (Fig. 4c). Moreover, due to the much stronger peak intensity of 4PP compared to 2MPC, it is anticipated that, instead of 2MPC demethoxylation, 4PC mainly forms by the hydrogenation of 4PP in the presence of THF (Fig. 4g). Additionally, PB is also a feasible precursor for PC. However, due to suppressed benzene ring hydrogenation, much less PC forms even at 250 °C. These results imply that adding an inert solvent coordinating comparatively strongly to the active sites of the catalyst can preserve the aromaticity of 4PG HDO products by altering the reaction pathway at the cost of a reduced reaction rate. As seen in Fig. 4g, the 4PG HDO mechanism in THF can be categorized into two reaction regimes in the 150–200 °C temperature range: (1) the 4PG → 4PP → PB route leads to the formation of PB; while (2) the 4PG → 4PP → 4PC and 4PG → 2MPC → 4PC routes result in the formation of 4PC. Because of the large 4PP abundance, it is also the likely intermediate on the way to 4PC, as opposed to 2MPC like in the non-solvent and isooctane experiments. However, even at the increased reaction temperature of 250 °C, PC formation remains limited.Theoretical insights on 4PG and THF adsorptionTo gain insights into the effect of THF on 4PG adsorption and the dynamics on the catalyst surface, we employed density-functional theory (DFT) calculations and on-the-fly machine learning force field simulations. A three-layered p-(6×6)-Ru(0001) slab model was constructed to represent the Ru/C catalyst. We compared the distinct flat and tilted adsorption configurations of 4PG on Ru(0001) as well as the configurations with broken O—H bond (dissociative adsorption) (Fig. 5a, S7). Notably, the flat adsorption configuration of 4PG was found to be thermodynamically more favorable than the tilted mode, with molecular adsorption energies of −2.54 and −0.84 eV, respectively (Fig. 5a). In addition, dissociative adsorption is exothermic by −3.09 and −1.15 eV in the flat and tilted configurations, respectively, demonstrating the thermodynamic preference for O—H bond scission from 4PG on Ru. The adsorption energy of THF on Ru(0001) was −1.05 eV, comparable to the 4PG adsorption energy of −0.84 eV in the tilted configuration. Furthermore, considering each surface atom as one site, the flat or tilted adsorption configuration of 4PG occupies 10 or 4 sites, while an adsorbed THF molecule occupies 4 Ru sites, respectively. (Fig. 5a). We compared the adsorption energy per site and found that the interaction of THF with the Ru surface is favored over 4PG compared with either the flat or tilted molecular adsorption configuration of 4PG. Only the dissociative flat adsorption configuration of 4PG is comparable to THF with respect to the adsorption energy per site, suggesting that 4PG and THF compete for the Ru adsorption sites. Given the high THF/4PG molar ratio (~ 500) employed herein, Ru surface sites were predominantly covered by THF, decreasing the density of available sites and suppressing the adsorption of 4PG via the flat configuration. This also aligns with the lower conversion observed at the same reactor temperature when THF was co-fed in the stream. From a mechanistic perspective37, surface THF must first desorb to provide free Ru sites for 4PG adsorption prior to HDO. This requires three and one THF molecules to desorb to accommodate the flat and tilted 4PG adsorption, respectively (see Fig. 5a). The tilted adsorption conformation of 4PG is therefore more likely to form at high surface coverage because of dynamics. In contrast, the adsorption energy of isooctane per Ru site is only −0.19 eV (Fig. S8), significantly lower than that of 4PG and THF. As a result, the local steric hindrance by isooctane coverage is not expected to be significant, and 4PG readily outcompetes isooctane for the Ru sites. This explains why isooctane does not substantially influence the 4PG HDO mechanism, as the experimental observation shows that the ring-hydrogenated 2MPC (172 amu) is the first product in both the isooctane and the non-solvent system.Fig. 5: Adsorption configurations and MLMD simulations from flat and tilted 4PG configurations.a Top view of adsorption configurations of 4PG (flat/tilted and molecular/dissociative) and THF on the Ru(0001) site; the occupied Ru sites and adsorption energy (eV) per site for each configuration. b MLMD starting from flat adsorption configuration of 4PG in the presence of THF at 20 °C for 10 ps and 220 °C for 6 ps. c MLMD starting from tilted adsorption configuration of 4PG in the presence of THF at 20 °C for 35 ps and 220 °C for 10 ps. C and O atoms in THF are in light gray and pink, and in 4PG are in dark gray and red, respectively.Since the desorption–adsorption equilibrium, the HDO conversion, and selectivity strongly depend on temperature (see Fig. 4c), we employed on-the-fly machine learning force field molecular dynamics (MLMD) simulations, which significantly accelerated the simulation time at the DFT-level of accuracy38, to investigate the temperature effect on the THF and 4PG adsorption on Ru(0001). To study the adsorption of THF on the Ru catalyst, we performed MLMD simulations from a monolayer THF-covered Ru(0001) surface for 50 ps at three different temperatures (20, 120, and 220 °C), as depicted in Fig. S9. The results revealed that more surface sites became exposed as the temperature increased and THF desorbed, creating free binding sites for 4PG. This temperature-dependent change may explain the absence of 4PG HDO activity at 80 °C (Fig. 4a, b) and why the product formation is only observed at temperatures higher than 150 °C in the presence of THF.Under the assumption that the flat adsorption of 4PG favors the formation of 2MPC and the tilted adsorption is responsible for the 4PP reaction pathways, we constructed three initial configurations in which we considered 4PG co-adsorption with THF on the surface via the flat and tilted configurations, and 4PG in the gas phase among randomly distributed THF molecules (Fig. S10 at 0 ps). For the MLMD from the 4PG flat adsorption configuration (Fig. 5b, 0 ps), we observed that more THF molecules moved onto the Ru surface at 20 °C (Fig. 5b, 10 ps), but they tended to desorb from the surface at 220 °C (Fig. 5b), in agreement with the MD results of neat THF on Ru without 4PG. In contrast, the flat 4PG remained stable and less affected by THF, suggesting that 4PG in the flat adsorption configuration is indeed stable once formed. The stability of the flat configuration was further assessed using the MLMD for 500 ps at 120 °C starting from the flat adsorption configuration; we observed the O—H scission at 8.4 ps and the persistence of the dissociated flat configuration from 0 to 500 ps (Fig. S11). For the tilted adsorption configuration, 4PG moved away from the Ru surface at 20 °C at 35 ps (Fig. 5c) due to the comparable adsorption energy between tilted 4PG and THF and the latter being present in excess. When the MLMD was run at 220 °C for 10 ps, 4PG desorbed from the Ru surface along with THF, consistent with its lower adsorption energy compared to flat configuration (Fig. 5a). For MLMD simulations at a medium temperature of 120 °C (Fig. 6a), the tilted configuration was maintained for 19 ps. However, as more and more THF desorbed from the surface as time progressed, the tilted adsorption configuration eventually transformed into a flat one once the Ru sites for the flat adsorption of 4PG became available. Our results indicate a strong temperature dependence of the competitive 4PG adsorption in the THF atmosphere. At low T (20 °C), the surface as covered with THF, hindering the 4PG adsorption. However, 4PG in the tilted adsorption configuration was prone to desorb at high temperature (220 °C). At medium T (120 °C), the tilted adsorption configuration of 4PG was stable for a while, but it eventually transformed into a flat configuration once the surrounding Ru sites are liberated from THF.Fig. 6: MLMD simulations from tilted and gas phase 4PG configurations.Evolution of structure changes in MLMD simulation leading to the formation of flat 4PG configuration, starting from (a) tilted 4PG configuration at 120 °C and (b) gas phase 4PG configuration at 220 °C in the THF atmosphere.In our experiment, 4PG was introduced in the gas phase. Therefore, we also investigated this as a starting point for the simulation. We did not observe 4PG adsorption at 20 °C and 120 °C in the presence of THF (Fig. S12), indicating that 4PG adsorption on the surface is not favored and the reactions are unlikely to occur at low reaction temperatures, consistent with the experimental results shown in Fig. 4b, c. However, as more surface sites are exposed when the temperature increases to 220 °C, 4PG can be adsorbed. The process shows three stages of the evolution of the adsorption geometry, namely from tilted, to flat, and dissociated flat configuration (Fig. 6b). This indicates that 4PG adsorption becomes feasible once the catalyst has sufficient, even if transient, surface sites and it finally leads to the most stable flat adsorption configuration. As the temperature increases, the initially complete surface THF coverage decreases, exposing more and more Ru sites to 4PG adsorption. However, partial THF coverage still creates a steric hindrance to selectively block the flat adsorption configuration, requiring more Ru sites (Fig. 5a). This explains why 4PP and PB dominate the product distribution in Fig. 4c, when THF is present. Nonetheless, the tilted adsorption configuration eventually transitions into the flat one in the MD simulation once more Ru surface is exposed as the temperature increases. Therefore, the experimentally observed change in the HDO mechanism can be attributed to the competition between deoxygenation in the tilted configuration and the conversion of the tilted configuration to the flat one, which entails benzene hydrogenation. Accordingly, abundant H2 may accelerate the conversion of adsorbed 4PG in the tilted configuration before transforming into the flat adsorption configuration. The driving force for the chemical transformation is also that the dissociative adsorption is more exothermic than non-dissociative adsorption (Fig. 5a). The dissociative configuration has broken O–H bond, making C–O(*) bond much more stable than C–O(CH3). Thus, demethoxylation must occur prior to the dehydroxylation, resulting in the formation of 4PP and PB step by step. This also agrees with the reaction sequence that –OCH3 decomposition occurs earlier than C–OH bond scission from 2-methoxy-1,1′-biphenyl39 and guaiacol40. As the temperature increases, the blocking effect of THF becomes weaker, and the transformation from tilted to flat configuration is favored, resulting in a benzene ring adsorption on Ru sites via the d-π interaction, which promotes the further hydrogenation to 4PC and PC through 4PP and PB (Fig. 4c, g).
