Mechanocatalytic hydrogenolysis of benzyl phenyl ether over supported nickel catalysts †

Mechanocatalysis is a promising approach for green, solvent-free biomass deconstruction and valorization. Here, the hydrogenolysis of benzyl phenyl ether (BPE), a model lignin ether, via ball milling is demonstrated over supported nickel catalysts at nominally room temperature and atmospheric hydrogen pressure. The hydrogenolysis reaction network closely follows that of solution-based reactions, with the primary products being toluene, phenol, and cyclohexanol. The mechanical energy during milling not only drives the chemical reactions but also activates the nickel by exposing fresh metallic surfaces from passivated particles, which replaces a thermal activation step. The hydrogenolysis rate is shown to be largely insensitive to the ﬁ nal nickel particle size, but reactivity of the oxide support can be enhanced during milling which contributes to carbon deposition. This work demonstrates the underlying chemistry necessary for mild lignin depolymerization using reductive mechanocatalysis.


Introduction
As society shis away from fossil-based fuels and chemicals, the necessity for complete utilization of biomass becomes ever more pressing.Lignin, a primary component of woody biomass, is an abundant renewable resource but is practically unutilized as a feedstock for chemicals and materials.Of the 60 million tonnes of lignin isolated annually, less than 2% is used for value-added products, while the rest is burned as a low value fuel. 1 Lignin is a heterogeneous, amorphous polymer that is primarily comprised of three propyl-phenolic monolignols (pcoumaryl alcohol, coniferyl alcohol, and sinapyl alcohol) linked together randomly by various C-O (b-O-4, a-O-4, 4-O-5) and C-C (b-b, b-5, 5-5) bonds. 2 However, the fractionation of lignin from the lignocellulosic matrix typically changes the structure of the polymer and increases the abundance of condensed (C-C) linkages. 3As the largest natural source of aromatic carbon, lignin is a promising, sustainable alternative starting point for historically benzene-toluene-xylene (BTX) derived chemicals if the polymer can be selectively depolymerized to monoaromatics.
Two of the most promising approaches to depolymerize lignin utilize hydrogen as either a cleaving or stabilizing reagent: hydroprocessing and reductive catalytic fraction (RCF). 4Hydroprocessing, generally performed with isolated lignin, combines hydrogenolysis of the linkages to produce monomers and hydrodeoxygenation (HDO) to upgrade monomers simultaneously. 5These reactions tend to produce monomer yields between 10 and 20 wt%, with harsher reaction conditions increasing monomer yields at the expense of product selectivity. 4RCF starts with complete lignocellulose and combines the isolation and depolymerization of lignin, where solvolysis formed monomers are stabilized before repolymerization can occur. 6Most reported monomer yields from RCF studies range from 10 to 30 wt% for sowood and from 30 to 60 wt% for hardwood feedstocks with very high selectivities of phenolics.Both hydroprocessing and RFC reactions employed noble metals, such as Pt, Pd, Ru, or Rh, or transition metals, primarily Ni and Cu, on either acidic or neutral supports.These reactions oen require elevated temperatures (130-450 °C), elevated hydrogen pressures (10-100 bar), and large quantities of organic solvents, such as methanol, dioxane, or THF. 4 The specialized equipment necessary for the combination of high hydrogen pressures and heated, ammable solvents, along with the energy demands for solvent recovery, currently stand as major barriers to scale up these processes. 7][10] The direct input of mechanical energy, such as collisions in a ball mill, can drive reactions between solid reactants and catalysts.][17] The multitude of ways that mechanochemical reactions can occur has resulted in a broad range of applications, such as alloying, 12 pharmaceuticals 18 and absorbents 19 synthesis, consumer plastics deconstruction, 20,21 and waste recycling. 22echanical energy, without added chemicals or catalysts, has long been used as a method for pulping wood and pretreating lignocellulose for conversion to fuels and chemicals. 23n recent years, the use of mechanocatalysis for biomass valorization has seen a rise in popularity, albeit with an emphasis on cellulose or unfractionated lignocellulosic biomass. 24,257][28] Additionally, the energy efficiency of the reaction has been shown to increase at least up to the kilogram scale while also maintaining greater than 90% yields of water-soluble oligosaccharides. 29Studies on the degradation of isolated lignin have been more limited.Early work focused on the homolytic cleavage of model compounds during ball milling. 30,31Mechanochemistry has since been utilized as a pretreatment method 32 or for alkaline 33,34 and oxidative 35 depolymerization.However, these approaches either rely on corrosive reactants (i.e., NaOH) or use a complex, multicomponent mixture of homogeneous catalysts, which may be difficult to separate and recycle.
While the mechanocatalytic hydrogenation of benzene over nickel in a vibratory ball mill was reported as far back as 1963, 36 studies on reductive mechanocatalysis with gaseous hydrogen have been sparse. 37The methanation of CO 2 over Ni/Fe/Ru on MgO catalysts 38 and LaNi alloys 39 and of CO over mixed FeCo and MgNi catalysts 40 have been demonstrated.2][43] Recently, the reduction of N 2 to NH 3 with H 2 was reported using a batch planetary ball mill 44 and in continuous-ow vibratory ball mills. 16,45erein, the mechanocatalytic hydrogenolysis of BPE was demonstrated under solvent-free, nominally ambient conditions (room temperature, 1 bar H 2 ) using supported nickel catalysts.7][48] Benzyl phenyl ether (BPE) is studied here as a model compound for lignin because of the similarity in C-O bond strengths between BPE and the b-O-4 linkage in lignin. 49dmittedly, BPE lacks the functional moieties of lignin (i.e., hydroxyl and methoxy groups), however the novelty and complexity of the interactions between the reagent, catalyst, and mechanical action necessitates a more simplied model system to draw meaningful insights.A process based on this approach would have a distinct advantage over traditional approaches by avoiding product-solvent separations.
Additionally, in this paper, the mechanocatalytic hydrogenolysis reaction network for BPE is elucidated.The in situ mechanical activation of the supported nickel catalysts, as well as recyclability and deactivation pathways, are investigated.Finally, the effects of the nickel support properties on the activity of the catalysts are probed.Insights gained on the catalytic system will aid in further developing processes for the solvent-free and mild depolymerization of lignin via reductive mechanocatalysis.

Catalyst synthesis
Three 5 wt% nickel catalysts were synthesized using amorphous silica-alumina (Ni 05 SiAl), silica gel (Ni 05 Si high ), and quartz (Ni 05 Si low ).The catalysts were prepared by incipient wetness impregnation with an aqueous Ni(NO 3 ) 2 solution.The wetted samples were dried at 105 °C in air overnight.The dried samples were reduced in 7% H 2 in Ar at 450 °C (ramp from room temperature at 10 °C min −1 ) for 8 h.The samples were exposed to and stored in air aer returning to room temperature.

Catalyst characterization
Inductively coupled plasma-optical emission spectroscopy (ICP-OES).Acid digestion and inductively coupled plasma optical emission spectroscopy (ICP-OES) was performed by the Renewable Bioproducts Institute.Sample aliquot was weighed into a predigested beaker.The beaker with sample was placed on a hot plate.Aer 5 mL of concentrate nitric acid was added, the contents were reacted for 30 min under elevated temperature (below boiling).This was followed by adding another 2.5 mL of concentrated HCl, and the mixture was reacted for another 30 min.The digested mixture was cooled to room temperature and was ltered through a grade 40 ashless Whatman lter paper.The ltering assemble was washed with de-ionized water and joined to the ltrate.The volume of the nal aqueous solution was recorded.The ICP-OES instrument was periodically tested for performance, which includes detector calibration, align view and wavelength calibration.In each batch of sample analysis, the instrument was calibrated with standards for the target elements, and a four-point calibration curve was established for each interested element.The aqueous sample prepared as stated above was used for ICP-OES analysis.
N 2 -physisorption.Nitrogen physisorption measurements were performed using a Micromeritics ASAP 2020 physisorption analyzer.The samples (50-100 mg) were evacuated at 50 °C and 10 mmHg for 1 min and subsequently degassed at 110 °C and 100 mmHg for 4 h prior to the measurement.The measurement was performed at 77 K. Surface areas were calculated based on the BET method for a p/p 0 range between 0.025 and 0.30. 50-ray powder diffractometry (XRD).X-ray powder diffraction (XRD) patterns were obtained using an Empyrean diffractometer by Malvern Panalytical, which operates with CuKa radiation coming from a Bragg-BrentanoHD source with a wavelength of l = 1.5406Å.The mask was 4 mm, the divergence slit was xed at 1 4 °, the anti-scatter slit was 1°, and the soller slits were 0.04 rad.With this optics, an area of 138.7 mm 2 of the sample was scanned.The generator was operating at a tension of 45 kV and a current of 40 mA.The sample stage was a reection-transmission spinner, the detector was a PIXcel3D.Diffractograms were measured at incident angles from 2q = 15°to 100°with a step size of 0.0131303°and a number of steps of 6474.The time per step was set to 61.2 s, the scan speed was chosen to be 0.05471 s −1 .The sample was rotated during the measurement with a revolution time of 16.0 s.The detection limit was 0.5 wt%.Data analysis was performed using the HighScore Plus soware by Panalytical.
Scanning transmission electron microscopy (STEM).The powder sample was dispersed in ethanol, then dropped on carbon coated Cu grids aer ultrasonication for 5 min.A Hitachi HD2700 aberration-corrected scanning transmission electron microscope was used to record the high-angle annular dark-eld (HAADF) STEM images.The electron beam convergent angle a was 27 mrad and the HAADF detector collection angle b = 70-370 mrad.
Raman spectroscopy.Visible Raman spectroscopy was performed on a Renishaw Raman spectrometer using an excitation wavelength of l = 488 nm.The laser power was chosen between 1-10% of the total power (30 mW), the grating of the lter was 2400 lines per mm.Samples were analyzed with 50× magnication and an acquisition time between 5-20 s.The acquisition soware was WiRE 5.2.The spectra were collected using a CCD detector.The acquired spectra were deconvoluted using the peak analyzer function of the OriginPro 2020 (V9.7) soware.
Diffuse reectance infrared Fourier transform spectroscopy (DRIFTS).DRIFTS spectra were collected using a Nicolet 8700 FT-IR equipped with a liquid nitrogen cooled MCT/A detector and the Praying Mantis (Harrick Scientic) accessory with a high temperature reactor for in situ studies.Spectra were averaged at 64 scans with a resolution of 4 cm −1 .All spectra were treated using the Thermo Scientic Omnic Soware.A background of dried potassium bromide at 50 °C was used for all spectra.The sampling cup was lled with spent catalyst and then smoothed several times to ensure an even layer.The reactor was sealed and ushed with 100 sccm N 2 for 5 min to settle the sample bed.An initial spectrum of the pure milled catalyst was collected at 50 °C for spectral subtraction against the spent catalyst with species adsorbed.Spent catalysts were measured before at 50 °C and then ramped up to 700 °C at 10 °C min −1 , held for 30 min in 40 sccm N 2 and then cooled back down to 50 °C.All spectra were collected at 50 °C aer the sample had been heated to 700 °C.Control experiments were conducted on unmilled Ni 05 Si high and Ni05Si with physisorbed BPE.The catalyst was impregnated with BPE (dissolved in dichloromethane) via incipient wetness impregnation.DRIFTS spectra were collected for each control sample at 50 °C and aer heating to 700 °C.

Mechanocatalytic reactions
Mechanocatalytic reactions were performed using a modied Retsch MM400 vibratory ball mill with a 25 mL stainless steel milling vessel.The milling vessel was modied to allow for continuous gas ow.The effluent line was heated to 160 °C and the effluent gas was bubbled through methanol (∼10 mL with ∼0.03 g o-xylene as an internal standard) cooled to 4 °C to trap volatile products leaving the reactor.Unless otherwise specied, reactions were conducted with 1.00 g of catalyst, 0.20 g of BPE, two 15 mm stainless steel grinding balls, and a milling frequency of 20 Hz in a continuous ow of H 2 .The high catalyst to BPE ratio was necessary in order to keep the substrate a powder during milling.Due to the low melting point of BPE (40 °C), lower catalyst-to-reagent ratios resulted in liquefaction of the substrate and no observed reaction.This issue would not arise when milling polymeric materials, like lignin, and lower catalyst-to-reagent ratios can be used.Additionally, this type of system could be adapted to a continuous process at larger scales, allowing for more realistic reagent-to-catalyst ratios.The vessel was le unheated but milling induced an external temperature increase to ∼30 °C within the reactor.The vessel was purged with N 2 for 15 min and then with H 2 for at least 15 min before milling began.Aer milling, the vessel was purged with N 2 for 1 h to remove additional residual volatile products.
Unconverted BPE and non-volatile products were measured aer each reaction by removing the product-catalyst mixture from the milling vessel and sonicating in methanol (∼10 mL) for 30 min.Roughly 90% of the sample mass could be recovered from the vessel.The remaining ∼10% were le adsorbed on the vessel and balls or in the valve port of the ball mill.Aer sonication, o-xylene was added (∼0.03 g) as an internal standard.The yield of volatile products over the course of the reaction was measured by taking aliquots (∼0.05 mL) of the methanol trap.
Product yields were determined via gas chromatography using a Varian 450-GC with a fused silica column (Supelco, SPB-1, 60 m × 0.25 mm × 1.0 mm) and a ame ionization detector (FID).The GC-FID was equipped with a PolyArc reactor (Activated Research Company), which allowed for calibration-free quanti-cation of the products.
The long-term stability of the Ni 53 SiAl catalyst was studied through a series of recycle experiments.The catalyst was milled for 2 h during the initial cycle and 4 h for the second and third cycles.Aer being washed with methanol to recover unconverted BPE and non-volatile products, the catalyst was vacuum ltered and dried under vacuum at room temperature.Similar to the mechanocatalytic reactions with the fresh catalyst, 1.00 g of the recycled catalyst was used.The same experimental procedures were used as the reaction with fresh catalyst.Aer vial transfers and ltration, only ∼65% of the spent catalyst mass could be recovered (∼0.65 g).So, spent catalyst samples of several different runs were combined for each recycle stage.For example, to produce enough sample to run a third cycle, two second cycle run were performed, which required four fresh catalyst runs.

Mechanocatalytic reaction kinetics
In order to semi-quantify the activity and the deactivation of the catalysts and the rate of carbon loss, the reaction is treated as a simplied system with only two rst order reactions.The rst is the hydrogenolysis of BPE to toluene and the second is the coke formation from BPE.Since the toluene yield is based on products detected in the condenser, it is assumed that all the produced toluene is found in the condenser and that toluene does not participate in the coking reaction.Initially, the mechanocatalytic hydrogenolysis reactions were t using this simplied kinetic model (eqn (1)-( 4)), where X BPE is the conversion of BPE, k H is the hydrogenolysis rate constant, m Ni is the mass of nickel, k C is the coking rate constant, and Y tol is the yield of toluene in the condenser.Eqn (3) and ( 4) are the analytical solutions to the integration of eqn ( 1) and (2), respectively.Since X BPE was measured for several time points for fresh Ni 53 SiAl, eqn (3) was t to solve for k For experiments where the full product yields were only measured once and the conversion of BPE was incomplete, k H $m Ni + k C were be determined by eqn (3) to yield eqn (5) and using the nal conversion.k H was t using Solver in Excel to minimize the error between the model and the toluene yield from the condenser.Due to the observed induction period in the condenser yield, a delay term was also t.For Ni 53 SiAl-cycle 2, which was run only once and had nearly 100% conversion of BPE, calculating k C from eqn (5) would have resulted in high uncertainty and was instead approximated with the k C of Ni 53 -SiAl-cycle 3.
To assess the deactivation of Ni 53 SiAl, a rst-order deactivation term was included in the model (eqn ( 6)), where k D is the deactivation rate constant.No analytical solution to eqn (S1), (S2), and (S6) † exists, so the equations were t numerically.A Matlab script based on the built-in functions fmincon (error minimization) and ode45 (differential equation solver) was used to determine k D and the initial hydrogenolysis rate constant (k H,0 ).An average k C was used and assumed constant.1b-d).Energy dispersive spectrometry (EDS) mapping of Ni 53 SiAl showed particles uniformly coated in nickel (Fig. 1e), while the low nickel loading catalysts had discrete nickel particles (Fig. 1f-h).Ni 05 SiAl, Ni 05 Si high , and Ni 05 Si low had initial particle sizes of 18.7 nm, 15.1 nm, and 46.7 nm, respectively (Fig. S4 †).The high abundance of nickel on Ni 53 SiAl prevented the measurement of enough particles to determine initial particle size.

Hydrogenolysis reaction network
The mechanocatalytic reactions were conducted in a vibratory ball mill (Retsch, MM400) reactor, modied to allow for the continuous ow of H 2 (Scheme 1).The catalyst and BPE were placed together in the milling vessel without additional pretreatment.While the vessel was not externally heated or pressurized, the milling caused the external vessel temperature to rise slightly to ∼30 °C.The reactor effluent was bubbled through methanol cooled to 4 °C to collect volatile products, and the methanol solution was sampled periodically to determine the yield of toluene during milling as a proxy for the extent of reaction.Gas chromatography was used to quantify products collected in the methanol trap and washed off the catalyst aer reaction.The majority of the hydrocarbon products (toluene, benzene, cyclohexane) were recovered in the methanol trap, while oxygenated products (phenol, cyclohexanol) remained in the milling vessel and were collected from the catalyst wash.2][53][54] At early milling times, the main products were toluene (1a) and phenol (1b), which are formed by hydrogenolysis of the benzyl ether bond.Benzyl alcohol (8) was not detected, which is unsurprising since the bond dissociation energy of the benzyl ether bond (56.4 kJ mol −1 ) 55 is much lower than that of the phenyl ether bond (95.7 kJ mol −1 ). 56Yields of hydrogenated ethers (4 & 5) passed through maxima (∼1%) at 60 min, suggesting that hydrogenolysis of these compounds occurred, but did not signicantly contribute to the formation of the singlering products.The formation of methyl cyclohexane (2) and cyclohexanol (3) predominantly occurred through the hydrogenation of toluene and phenol, respectively.Phenol hydrogenation to cyclohexanol was established as a reaction path based on the maximum in the phenol yield aer 60 min.Low methylcyclohexane yields (<1%) can be explained by the removal of toluene vapors from the reactor prior to be being hydrogenated or a lower affinity of toluene to the active sites for hydrogenation reactions on the catalyst.Benzene (6) and cyclohexane (7) were detected aer 60 min and 120 min, respectively.Their formation is attributed to hydrodeoxygenation of phenol and cyclohexanol.Since benzene and cyclohexane were only found in the condenser, benzene is expected to volatilize before hydrogenation can occur.Alkylation reactions and heavier products have been reported in thermocatalytic systems. 52,57,58The decreasing carbon balance with increasing milling time could be attributed to the formation of similar species.Over the course of the reaction, the yield of toluene consistently exceeds the combined yields of phenol and cyclohexanol, meaning the adsorption of phenolates on the catalyst surface is another source of carbon loss.Overall, the nickel catalyst behaves similarly in mechanochemical and thermochemical systems.Thus, it may be possible to extrapolate the results of other thermocatalytic reactions to predict the performance of mechanocatalytic processes for lignin conversion.
A series of experiments was conducted to explore the catalytic transfer hydrogenolysis 59,60 of BPE using isopropanol (IPA) in place of H 2 (ESI Section S2.2 †).IPA was added to the milling vessel (125 mL) and the reaction was conducted in a continuous ow of N 2 .Here, a 2.4% yield of toluene and a 1.6% yield of phenol was measured aer 180 min.In principle, IPA can work as a hydrogen transfer agent for mechanocatalytic hydrogenolysis, but additional optimization and study would be necessary.

Mechanocatalyst activation
Before traditional thermocatalytic reactions, nickel must typically be reduced in situ at elevated temperatures (i.e., activated) to remove passivating oxide layers or reduce nickel oxide particles to the metal altogether.Here, thermal pretreatment of the passivated nickel catalysts was not necessary as milling in a reductive environment activated the catalyst.Initially, the nickel particles of the Ni 53 SiAl catalyst were encapsulated by a layer of NiO (Fig. 3a-c).However, thermally oxidized Ni 53 SiAl (Fig. S4 †) showed signicantly less activity (Fig. 3d).Attrition during milling is expected to remove the nickel oxide layer, liberating the active metallic nickel.Particularly when milling hard and brittle material, localized hot spots haven been shown to form 13 which may produce temperatures high enough to directly reduce the NiO to metallic Ni in the presence of H 2 .The mechanochemical reduction of the NiO could explain why the rate of toluene production increased during milling with the oxidized Ni 53 SiAl.Importantly, the potential local thermal energy is not sufficient to break the BPE ether bond homolytically.Control experiments were conducted with BPE milled with Ni 53 SiAl in N 2 and milled with just silica-alumina in H 2 , and no conversion was detected in either case.Thus, mechanical energy both drives the hydrogenolysis reaction and activates the catalyst, removing the need for high temperature pretreatments and allowing for a fully ambient system.

Catalyst recycling and deactivation
The accumulation of toluene in the condenser allows for determining a lower limit of the hydrogenolysis rate.The toluene yield prole in the condenser during milling with fresh Ni 53 SiAl for 120 min (Fig. 4a cycle 1) closely matched the progression of the toluene yield during the full product analysis (Fig. 2b).An induction period for toluene accumulation in the condenser of about 10 min was observed, which could be attributed to a catalyst activation period.However, at 7.5 min, the full toluene yield was 7% (compared to just 1% for the condenser toluene yield), suggesting the delay is an artifact of the reactor system rather than, for example, a period needed to mechanically activate the nickel.
To evaluate the longer-term stability of the catalyst under such intense conditions, the Ni 53 SiAl catalyst was recycled twice.Aer each reaction, the catalyst was washed with methanol to analyze the complete product prole, and then ltered and dried under vacuum overnight at room temperature.The toluene yields for each cycle (Fig. 4a) showed continuous deactivation of the catalyst.In the second cycle, the toluene yield attened aer three hours of milling, trending toward a maximum toluene yield of 65%.During the third cycle, the toluene yield increased over the whole 4 h experiment, but only a 30% yield was achieved.Similar to the rst cycle, an induction period of about 10 min was observed in the second and third cycle.As expected from the toluene yield proles, the full product proles showed nearly 100% conversion of BPE by the end of cycle 2 and incomplete conversion (65%) by the end of cycle 3 (Fig. 4b).Deactivation of the nickel was also observed in the decreased selectivity of cyclohexanol to phenol between cycle 2 (69%) and cycle 3 (45%).Finally, signicant carbon loss was observed for each cycle.
The carbon balance suggests carbonaceous deposits form as undesired side reactions with increasing milling time.Raman spectroscopy was performed on the spent catalyst aer each cycle to qualitatively detect coke deposits (Fig. 4c).Three bands in the Raman spectra of spent catalysts were detected: the Gband (1580 cm −1 ; ideal graphitic lattice vibrations), the D 1band (1350 cm −1 ; vibrations of graphene layer edges), and the D 2 -band (1610 cm −1 ; surface graphene layers).These three  S1.† bands are indicative of polyaromatic coke. 61At longer milling times, the relative intensity of the D-bands increased (compared to the G-band), which is supported by deconvolution and tting of the bands (Table S2 †).This increase of the D-bands and the decrease of the G-band suggest that the coke transformed to a more disordered and amorphous structure, which is expected with extended milling times. 62Additional changes to the catalyst between cycles were investigated as well.Agglomeration of nickel particles was not observed in STEM-EDS images of the spent catalysts (Fig. S6 †) and no loss of nickel was detected between cycles (Fig. S7 †).Additionally, signicant oxidation of the nickel, due to handling in air, was not observed in XRD between each cycle (Fig. S8 †), although this does not rule out the formation of amorphous oxide phases.These results suggest that the deactivation occurs predominantly through the formation of adsorbed phenolates, which further react to form polyaromatic coke, resulting in blockage of active sites rather than mechanical degradation of the catalyst due to milling.
Phenolates and catecholates are well known to strongly bind to catalysts, 63 so the higher functionality of lignin is expected to exacerbate the formation of carbon deposits.The HDO of guaiacol, a more functionalized model compound, was briey studied (ESI Section S4 †).Aer 180 min, catechol, phenol, and cyclohexanol were detected at similar yields as HDO products during the hydrogenolysis of BPE (<1%), while the carbon balance was slightly lower (62% compared to 77%).Future work will focus on adapting reaction engineering strategies, such as simultaneous hydrolysis, 64 to desorb phenolates or the use of redox active catalysts, 65,66 to recover adsorbed products and prevent further side reactions.
Generally, the energy dose is a key descriptor for mechanochemical kinetics. 21,67However, the mechanochemical conditions were not varied in this study, so a simplied kinetic model is suitable here to provide a more quantitative description of the catalyst activity and deactivation.The system was approximated by two reactions: hydrogenolysis of BPE (k H ) and deposit formation (k C ), both rst order with respect to BPE (eqn (1)-( 5)).Initially, the conversion of BPE and yield of toluene for each cycle were t independently to determine k H and k C (Fig. S5 †).Both cycle 1 and cycle 3 had similar rates of coke formation (k C,1 = 0.17 h −1 ; k C,3 = 0.16 h −1 ), suggesting the activity of the catalyst toward coke formation is constant across all three cycles.Due to the high nal conversion in cycle 2, calculating k C,2 directly would result in high uncertainty.Thus, the value for k C,3 was used during the tting.The results of this kinetic model conrm the deactivation of the catalyst toward hydrogenolysis with a decrease in the rate constant between each cycle (k ).By the third cycle, the catalyst retained only a tenth of its initial activity.Next, all three cycles were t simultaneously, using an average k C of 0.163 h −1 , with a rst order deactivation term introduced to the hydrogenolysis rate constant (eqn ( 6)).These ts were overlaid on the toluene yields in Fig. 5a.This t shows an initial rate constant (k H,0 ) of 2.5 h −1 g Ni −1 and that the catalyst deactivated at a rate of 34% per hour of milling (k D = 0.34 h −1 ).The good agreement between the observed activity of the catalyst and the deactivation model suggests a continuous mechanism is occurring during milling instead of discrete deactivation occurring between cycles.

Inuence of catalyst support
To explore the inuence of the support properties on the reactivity, three catalysts with 5 wt% nickel loadings were synthesized using varying supports: a high surface area silica-alumina (499 m  reaction.The reduction of the accessible nickel surface area by roughly one order of magnitude only corresponded to a 33% lower rate constant.This resistance to deactivation from agglomeration could have several explanations.The agglomeration may be indicative of more effective mechanical activation of nickel particles or larger nickel particles may more efficiently convert mechanical energy into chemical energy.The pores of the high surface area supports can also allow for nickel particles to be effectively shielded from direct mechanical forces.However, the similarity in the rate constants for the three catalysts suggests porosity does not signicantly inuence mechanical activation.While the hydrogenolysis rates were similar among the three catalysts, the production of undesirable side products was much higher on the high surface area catalysts.Reactions with either Ni 05 SiAl or Ni 05 Si high resulted in carbon balances of just 30%, while the reaction with Ni 05 Si low had a carbon balance of 82% (Fig. 5b).The three spent catalysts were analyzed with Raman spectroscopy to determine the nature of the carbon deposits (Fig. 5c).Spent Ni 05 SiAl exhibited the same characteristic bands for polyaromatic coke as the spent Ni 53 SiAl (Fig. 4c).The spectrum of spent Ni 05 Si high did not show polyaromatic coke, despite a similar decit in the carbon balance.Only a single band at 1596 cm −1 was observed in the two spent silica supported catalysts, suggesting that the deposits are predominantly mono-aromatic species. 68,69The deposits on Ni 05 Si high and Ni 05 Si low were further analyzed using in situ DRIFTS (Fig. 5d).The region between 1400 and 1600 cm −1 corresponds to ring vibrations of aromatic species. 70Aer heating the samples to 700 °C, the presence of aromatics remaining on the Ni 05 Si high catalyst was much greater than the residue on the Ni 05 Si low or nonmilled Ni 05 Si high impregnated with BPE (Fig. S10 †).This indicates chemisorption or encapsulation of the remaining aromatic species onto the Ni 05 Si high surface.
The higher surface area of the support is most likely responsible for the discrepancy in carbon balances, as opposed to catalytic activity for undesirable side reactions.2][73] Due to a larger abundance of surface Si-O bonds in the high surface area support, it is plausible that ruptures result in more reactive sites compared to the low surface area support.The higher density of active sites from rupturing can thus be responsible for more adsorbed species and a lower carbon balance on the Ni 05 Si high .Additionally, mechanochemical conditions could transform the catalyst in a way that leaves carbonaceous materials encapsulated in closed pores.A similar mechanism is expected to occur with Ni 05 SiAl, but with the inherent acid sites of the silica-alumina also catalyzing polyaromatic coke formation. 74,75Future work should focus on mitigating these side-reactions by optimizing milling intensity and developing supports that are less prone to rupture.Soer material, like activated carbon, should ex and bend under the mechanical impact, and harder material, like zirconium oxide, should be more resilient to rupture.

Conclusions
The mechanocatalytic hydrogenolysis of benzyl phenyl ether, a model lignin ether, was demonstrated using cost-effective, abundant catalysts under ambient H 2 pressure, at room temperature, and without a solvent.The observed reaction network closely matches that for thermal hydrogenolysis of BPE, suggesting similar reactions can be expected to occur between mechanocatalytic and thermocatalytic reactions over metal catalysts.Milling not only supplied the mechanical energy to drive the reaction, but also provided metallic nickel surfaces in situ, avoiding the need for high temperature activation steps.Adsorption of phenolic species on mechanically activated domains of the catalyst support was identied as a major source for carbon loss, which provided guidance for the design of more robust mechanocatalysts.Overall, this works demonstrates the underlying chemistry necessary for developing a pathway for mild lignin depolymerization which may be able to avoid many of the safety, technical, and economic issues hindering the realization of lignin valorization at industrial scales.
Catalyst characterizationInductively coupled plasma optical emission spectroscopy (ICP-OES) analysis of the nickel loadings of the catalysts showed that the commercial nickel on silica-alumina (Ni 53 SiAl) sample contained 53.4 wt% Ni, while the synthesized Ni catalysts with silica-alumina, high surface area silica and low surface area silica (Ni 05 SiAl, Ni 05 Si high , and Ni 05 Si low ) had 4.42 wt%, 4.93 wt%, and 4.85 wt% Ni, respectively.The BET surface areas of Ni 53 SiAl, Ni 05 SiAl, Ni 05 Si high , and Ni 05 Si low were 85.3 m 2 g −1 , 499 m 2 g −1 , 515 m 2 g −1 , and 6.7 m 2 g −1 , respectively.Adsorption and desorption isotherms are shown in Fig. S1.† X-ray powder diffractometry (XRD) of Ni 53 SiAl showed the presence of both Ni and NiO, while only peaks corresponding to metallic Ni were observed in the diffractograms of Ni 05 SiAl, Ni 05 Si high , and Ni 05 Si low (Fig. S2 †).Scanning transmission electron microscope (STEM) images showed that the nickel particles of Ni 53 SiAl were encapsulated in NiO (Fig. 1a).Similarly, STEM showed outer layers of NiO on the nickel particles for Ni 05 SiAl, Ni 05 Si high , and Ni 05 Si low (Fig.

2 g − 1 ,
Ni 05 SiAl), high surface area silica (515 m 2 g −1 , Ni 05 Si high ), and low surface area silica (6.7 m 2 g −1 , Ni 05 Si low ).BPE hydrogenolysis was performed for 180 min with each catalyst.The yield of toluene over the course of the reaction (Fig. 5a) showed a constant increase for Ni 05 Si low , while the rate of toluene formation diminished with time for Ni 05 SiAl and Ni 05 Si high .The simplied kinetic model (eqn (4) and (5)) was applied to these reactions to determine the hydrogenolysis rate constant (ts shown in Fig. 5a).Ni 05 SiAl (k H = 0.81 h −1 g Ni −1 ) and Ni 05 Si low (k H = 0.83 h −1 g Ni −1 ) had similar hydrogenolysis rates, which were approximately 33% lower than the rate of Ni 05 Si high (k H = 1.21 h −1 g Ni −1 ).The nickel particle size had a weak correlation with the activity of the catalyst.Nickel particles observed using STEM-EDS on Ni 05 SiAl and Ni 05 Si low agglomerated during milling to a size on the order of 100 nm (Fig. S9 †), while the size of the nickel particles on Ni 05 Si high (15.1 nm) did not appreciably change over the course of the

Fig. 3
Fig. 3 (a) STEM image of a nickel particle from Ni 53 SiAl showing a NiO shell around a Ni core.Fourier transformation of a region of (b) the shell (blue ellipse) and (c) the core (red circle) showing d-spacings corresponding to NiO and Ni, respectively.(d) Yields of toluene in the condenser during mechanocatalytic hydrogenolysis of BPE over Ni 53 SiAl (red squares) and thermally oxidized Ni 53 SiAl (blue circles).1.00 g catalyst, 0.20 g BPE, 25 mL S. S. vessel, 2 × 15 mm S. S. grinding balls, 20 Hz, 60 sccm H 2 .