Catalytic Performances of Sn-Beta Catalysts Prepared from Different Heteroatom-Containing Beta Zeolites for the Retro-Aldol Fragmentation of Glucose

Abstract

Beta zeolites with different heteroatoms incorporated into the lattice at two loadings (Si/M = 100 or 200, where M = Al, Fe, Ga, B) were hydrothermally synthesised and used as starting materials for the preparation of Sn-Beta using Solid-State Incorporation. ¹¹⁹Sn CPMG MAS NMR showed that various Sn species were formed, the distribution of which depended on the identity of the initial heteroatom and the original Si/M ratio. The final Sn-Beta materials (with Si/Sn = 200) were explored as catalysts for the retro-aldol fragmentation of glucose to α-hydroxy-esters in the continuous regime. Amongst these materials, B-derived Sn-Beta was found to exhibit improved levels of selectivity and stability, particularly compared to Sn-Beta catalysts synthesised from commercially available Al-Beta materials, achieving a combined yield of methyl lactate and methyl vinyl glycolate > 80% at short times on the stream. Given that B atoms can be removed from the Beta lattice in mild conditions without the use of highly concentrated acidic media, this discovery demonstrates that B-Beta is an attractive starting material for the future post-synthetic preparation of Lewis acidic zeolites.

1. Introduction

Lewis acidic zeolites have emerged as state-of-the-art catalysts for the conversion of biomass to commodity chemicals over the past decade, due to their high levels of activity and selectivity. Zeolites are crystalline microporous materials which can exhibit a wide array of catalytic properties depending on their precise composition (i.e., the choice of metal and metal loading) and topology [1,2,3,4]. In the context of biomass conversion, the most notable Lewis acidic zeolite is the stannosilicate catalyst Sn-Beta. Sn-Beta is a siliceous zeolite within which small amounts of Lewis acidic Sn(IV) atoms are isomorphously substituted into the three-dimensional lattice. The presence of isolated Sn(IV) atoms has been shown to endow the material with levels of Lewis acidity suitable for the conversion of biomass into various bio-based compounds of industrial interest [5,6,7]. Methyl lactate (ML) [8] and methyl vinyl glycolate (MVG) [9], for instance, are the main α-hydroxy-esters that can be produced following the cleavage of glucose in the presence of Sn-Beta (Scheme 1). These compounds are highly sought after due to their potential as monomers for the synthesis of bioplastics, or as precursors for surfactant production.

The synthesis of Lewis acidic zeolites such as Sn-Beta can be achieved by two generalised approaches. The first involves conventional hydrothermal synthesis (bottom-up), during which an appropriate metal precursor, such as SnCl₄, is added to the synthesis gel, resulting in the direct incorporation of the metal into the lattice during crystallisation at an elevated temperature and pressure. Alternatively, Sn-Beta can be synthesised by the post-synthetic treatment of an existing zeolite (top-down). In such cases, a previously prepared zeolite is treated to create vacant lattice sites [10,11], into which Sn can be incorporated in a subsequent synthetic step [12]. The advantages of such a top-down approach (also known as ‘demetallation–remetallation’) [13] include the avoidance of the long crystallisation times associated with hydrothermal Sn-Beta synthesis, and high degrees of synthetic flexibility in terms of choice of the metal and the final metal loading [14,15,16,17]. Amongst the top-down methods, Solid-State Incorporation (SSI) is an attractive method in which the desired heteroatom is introduced into the vacant tetrahedral framework sites by a combination of mechanochemistry and thermal treatment, thereby avoiding the generation of aqueous metal streams [18,19].

A common step in all demetallation–remetallation strategies is the preparation of a zeolite containing vacant lattice sites, suitable for the eventual hosting of the desired Sn heteroatom. This is typically achieved by pre-treating a parent zeolite material at an elevated temperature in concentrated acidic media, resulting in the removal of an original heteroatom [13,18,20,21,22]. In previous demetallation–remetallation studies, aluminosilicates have almost exclusively been employed as the parent zeolite, primarily due to their commercial availability and the ease of incorporating Al into the lattice of several zeolite structures. However, a variety of other metal-containing zeolites could also act as parent materials for the preparation of post-synthetic Sn-Beta. These include zeolites containing B, Fe and Ga, amongst others, which are also known to be prone to demetallation in appropriate conditions. Notably, such zeolites exhibit a large discrepancy between their ease of demetallation. In particular, B-derived zeolites have been shown to be particularly amenable to demetallation in very mild conditions, because boron is primarily trigonally coordinated, and is thus susceptible to nucleophilic attack by water [23,24,25]. Furthermore, the use of different heteroatom-containing zeolite materials as precursors for Sn-Beta synthesis could also provide access to final catalysts exhibiting different levels of performance. However, to the best of our knowledge, no study has yet investigated the catalytic properties of otherwise-analogous Sn-Beta samples prepared from hydrothermally synthesised zeolite parents containing different original heteroatoms.

As such, this study investigates the kinetic properties of a variety of Sn-Beta samples prepared by SSI, each prepared from different hydrothermally synthesised Al-, B-, Fe-, and Ga-Beta parent materials. The aim is to determine the impact of the use of different heteroatom-containing zeolites as parent materials for the top-down Sn-Beta synthesis. The characterisation of the synthesised catalysts is performed by means of X-ray Diffraction (XRD), porosimetry, Diffuse Reflectance Infrared Fourier Transform Spectroscopy (DRIFTS), Scanning Electron Microscopy (SEM), Inductively Coupled Plasma Mass Spectroscopy (ICP-MS), and ¹¹⁹Sn Magic Angle Spinning Nuclear Magnetic Resonance using the Carr-Purcell-Meiboom-Gill echo train acquisition method (¹¹⁹Sn CPMG MAS NMR). The kinetic performances of the resulting catalysts are evaluated for the continuous conversion of glucose to α-hydroxy ester compounds (ML and MVG), allowing the activity, selectivity, and stability of the series of catalysts to be determined in a rigorous manner.

2. Materials and Methods

2.1. Materials

Tetraethylammonium hydroxide (TEAOH, 35 wt.% in water), aluminium chloride (AlCl₃, ≥99.99%), ferric chloride (FeCl₃, ≥97%), gallium chloride (GaCl₃, ≥99.99%), boric acid (H₃BO₃, ≥99.5%), tetraethylorthosilicate (TEOS, ≥99%), potassium chloride (≥99%), and hydrofluoric acid (48 wt.% in water, ≥99.99%) were used for the synthesis of the Beta zeolites. All of the reactants were obtained from Merck Life Science UK Limited (Dorset, UK), and were used without any further treatment. Commercial Al-Beta (NH₄⁺-form, Si/M = 19) was purchased from Zeolyst^TM.

2.2. Catalyst Synthesis

The M-Beta zeolites were synthesised using fluoride ions as a mineraliser, adapting the procedure described in Ref. [14]. In a typical synthesis, 30.6 g TEOS was added to 33.4 g TEAOH under constant stirring at 450 rpm. After 60–90 min, one phase was observed and 0.734 mmol of the metal precursor (FeCl₃, AlCl₃, GaCl₃, and H₃BO₃) corresponding to an Si:M molar ratio of 200:1 was added and left stirring at 450 rpm until the adequate water ratio was reached, typically at 16 h. Afterwards, 3.3 g of HF was added to form a solid gel. The final molar ratios of the gels were 0.0050M:1.0SiO₂:0.54TEAOH:0.54HF:8.5H₂O, where M corresponds to the indicated metal. The gels were transferred into a stainless-steel autoclave with PTFE liners and placed into an oven held at a temperature of 140 °C for 7 days. The zeolites were then filtered and washed with water (2.0 L g⁻¹), and subsequently dried at 110 °C overnight. The removal of the residual Structure Directing Agent (SDA) was achieved by heating the synthesised material to 550 °C for 6 h at 2 °C min⁻¹ in static air. The samples were denoted ‘M-Beta’, with M corresponding to the metal being incorporated.

Al-Beta and B-Beta with Si:M molar ratios of 100 were prepared following the general procedure described above, albeit with the addition of 1.47 mmol AlCl₃ and H₃BO₃, respectively.

2.3. Post-Synthetic Modification of M-Beta Zeolites by Solid-State Incorporation

The post-synthetic SSI was conducted as described in Ref. [26]. For simplicity, and to provide a fair comparison for all of the materials, all of the parental Beta zeolites were demetallated in 13 M HNO₃ (20 mL g⁻¹ of zeolite) at 100 °C for 20 h. The demetallated samples were filtered, washed with water, and dried at 110 °C overnight. Subsequently, the incorporation of Sn was undertaken in order to fill the vacant lattice sites of the demetallated zeolites. This was achieved by grinding the demetallated materials with the required quantity of tin (II) acetate for 10 min. Thereafter, the solid mixtures were heated in a combustion furnace (Carbolite MTF12/38/400, Carbolite Gero , Sheffield, UK) to 550 °C (10 °C min⁻¹ ramp rate) in a nitrogen flow (60 mL min⁻¹). The catalysts were held at 550 °C for three hours before the gas flow was changed to air (60 mL min⁻¹), and this was maintained at the same temperature for an additional 3 h.

2.4. Catalyst Characterisation

X-ray diffraction patterns were obtained using a Panalytical X’Pert Pro diffractometer using a Cu Kα radiation (λ = 1.5406 Å) at 40 kV and 40 mA. Spectra data were recorded in the 5–80 2θ range (step size 0.0167°, time/step  =  150 s, total time  =  1 h). The nitrogen adsorption isotherms were determined on a Quantachrome autosorb iQ at 77 K. Prior to the analysis, the samples were degassed at 300 °C for 12 h. The specific surface area was calculated using the BET equation. The pore size distribution was calculated using the NLDFT method (the N₂ silica equilibrium transition equation). The vapour sorption was undertaken in the same porosimeter, using water as the solvent, within P/P₀ ranges of 0–0.7. The metal concentration in the zeolite was determined by ICP-MS using an Agilent 7900 ICP-MS (Agilent, Santa Clara, CA, USA) using an HF/HNO₃ matrix solution; approximately 50 mg of the catalyst was immersed in a solution composed of 75 μL HF (48%) and 250 μL ultrapure-grade HNO₃ (69%). SEM images were obtained in a Hitachi TM3030 electron microscope using an accelerating voltage of 15 kV and a back-scattering detector. Nuclear Magnetic Resonance (NMR) imaging was carried out in a Bruker Avance III HD 400 MHz NMR spectrometer. ¹¹⁹Sn NMR analyses were carried out at operating frequencies of 149.23 MHz. All of the analysis was run with a 10 kHz spin rate packed in ZrO₂ rotors. The ¹¹⁹Sn DE CPMG MAS NMR experiments were performed with the CPMG echo-train acquisition, as described in Refs. [27,28]. The spectra were measured by applying 100 echoes, using a 2.5 μs 90° pulse with 180° proton decoupling. The relaxation delay (t₁) used was 135 s with 512 scans. The tin chemical shifts were calibrated using tin (IV) oxide as a reference. DRIFT spectroscopy was performed in a Harrick praying mantis cell. The spectra were recorded on a Bruker Tensor II spectrometer over a range of 4000–600 cm⁻¹ at a resolution of 2 cm⁻¹.

2.5. Kinetic Studies

The retro-aldol fragmentation of D-glucose (Merck Life Science UK Limited, Dorset, UK, ≥99%) was performed under a continuous flow, in a plug flow stainless steel tubular (¼”, 4.1 mm ID) reactor [14]. The reactor was connected to an HPLC pump in order to regulate the reactant flow, and to allow operation at an elevated pressure. The catalyst (100 mg) was placed between two plugs of quartz wool and packed in the reactor with a 0.5 µm frit placed at the reactor outlet. The reactor was then immersed in a thermostatted oil bath at 160 °C. The pressure in the system was controlled using a backpressure regulator, set at the desired pressure (typically 20 bar). Periodic aliquots were taken from the reactor outlet and analysed using an Agilent 1260 Infinity HPLC coupled with a Hi-Plex Ca column and an ELS detector, where sorbitol was used as an external standard. Samples were also measured by GC-FID (Agilent 7820, 25 m CP-Wax 52 CB, Agilent, Santa Clara, CA, USA), for which biphenyl was used as an external standard.

2.6. Activity Definitions

$$\mathrm{Conversion}: \mathrm{X}(\%)=\frac{{\mathrm{mol}}_{{\mathrm{Glu}}_0}-{\mathrm{mol}}_{{\mathrm{Glu}}_{\mathrm{t}}}}{{\mathrm{mol}}_{{\mathrm{Glu}}_0}}\ast 100$$

(1)

$$\mathrm{Yield}: \mathrm{Y}(\%)=\frac{{\mathrm{mol}}_{{\mathrm{Product}}_{\mathrm{t}}}}{{\mathrm{mol}}_{{\mathrm{Glu}}_0}}\ast {100}_{\mathrm{Hexoses}} \mathrm{or} {50}_{\mathrm{ML}}$$

(2)

$$\mathrm{Selectivity}:\mathrm{S}(\%)=\frac{\mathrm{Product} \mathrm{Yield} \mathrm{Y}(\%)}{\mathrm{Conversion} \mathrm{X}(\%)}\ast 100$$

(3)

$$\mathrm{Levenspiel} \mathrm{deactivation} \mathrm{rate}:{ \mathrm{k}}_{\mathrm{d}} ({\mathrm{h}}^{-1})=\ln \left[\ln \left(\frac{1}{1-\frac{\mathrm{X}(\%)}{100}}\right)\right]$$

(4)

$$\mathrm{Turnover} \mathrm{frequency}:{ \mathrm{TOF}}_{\mathrm{ML} }({\mathrm{h}}^{-1})=\frac{{\mathrm{mol}}_{{\mathrm{ML}}_{\mathrm{t}}}}{\mathrm{time} \left(\mathrm{h}\right){\ast \mathrm{mol}}_{\mathrm{Sn}}}$$

(5)

$$\mathrm{Productivity}: \mathrm{P} ({\mathrm{h}}^{-1})=\frac{{\mathrm{m}}_{\mathrm{ML}+\mathrm{MVG}} \left(\mathrm{g}\right)}{{\mathrm{m}}_{\mathrm{Catalyst}}\left(\mathrm{g}\right)\ast \mathrm{time} \left(\mathrm{h}\right)}$$

(6)

$$\mathrm{Weight} \mathrm{hourly} \mathrm{space} \mathrm{velocity}: \mathrm{WHSV}({\mathrm{h}}^{-1})=\frac{\mathrm{sugar} \mathrm{mass} \mathrm{feed} \mathrm{rate}\left(\frac{\mathrm{g}}{\mathrm{h}}\right)}{{\mathrm{m}}_{\mathrm{catalyst}} \left(\mathrm{g}\right)}$$

(7)

3. Results and Discussion

3.1. Catalyst Synthesis and Characterisation

A variety of analogous Beta zeolites were first prepared by hydrothermal synthesis according to the established procedures [29]. Within this series, four different metals were chosen as the heteroatom of each material. Specifically, Al-, B-, Ga- and Fe-Beta zeolites were prepared at a target Si/M molar ratio of 200. Additionally, two metal precursors (Al-, B-) were chosen as candidates in order to evaluate the impact of initial heteroatom loading in the parent zeolite; hence, they were also prepared at a target Si/M molar ratio of 100. In order to determine whether all of the parent Beta zeolites were successfully synthesised, XRD patterns were recorded for each sample following the removal of the residual SDA by calcination, and compared with the XRD pattern of a purely siliceous Beta zeolite (i.e., Si-Beta). As can be seen in Figure 1, the XRD patterns of the hydrothermally synthesised Beta zeolites were indistinguishable, and were identical in terms of crystallinity, confirming the successful synthesis of the Beta framework regardless of the choice of starting heteroatom and metal loading.

In order to further evaluate the properties of each zeolite, DRIFTS analysis was conducted (Figure 2). All of the synthesised zeolites presented vibrations typical of the Beta framework [30,31]. However, both B-Beta zeolites gave rise to additional vibrations at ca. 910–950 and 3734 cm⁻¹, indicative of Si–O stretching vibrations belonging to uncoupled SiO₄ tetrahedra [32] and internal Si–OH groups present in framework defects, respectively [33]. Further details regarding the presence and absence of vibrations in these materials are reported in Table S1.

In order to determine the impact of the choice of metal on the zeolite morphology, SEM analysis was performed for the Si/M = 200 zeolites (Figure 3). All four samples were of comparable morphology and size (ca. 10 μm), and possessed the typical trigonal bipyramidal structure of Beta zeolite when synthesised in fluoride media [34].

The nitrogen physisorption measurements and ICP-MS analyses were also in line with previous studies of Beta zeolites (

Table 1. Elemental and textural characteristics of the hydrothermally synthesised M-Beta zeolites, and the subsequent demetallation.

Catalyst	Si/M a	SBET b (m2 g−1)	Smeso/ext c (m2 g−1)	Vtotal d (cm3 g−1)	Vmicro e (cm3 g−1)
Fe-Beta 200 f	207	501	59	0.32	0.23
Ga-Beta 200 f	152	504	83	0.34	0.22
Al-Beta 200 f	220	463	87	0.34	0.20
B-Beta 200 f	204	468	13	0.32	0.22
Al-Beta 100 g	116	453	46	0.34	0.21
B-Beta 100 g	107	499	65	0.33	0.22

^a Experimental Si/M ratio measured by ICP-MS analysis. ^b Surface area values calculated using the BET method. ^c Mesoporous external surface area calculated using the t-plot method. ^d Total pore volume calculated using the t-plot method. ^e Micropore volume calculated from the total pore volume. ^f Theoretical Si/M ratio of 200. ^g Theoretical Si/M molar ratio of 100.

). All of the isotherms were typical of microporous Beta zeolites, exhibiting mainly a type I character, but presented some hysteresis in the higher P/P₀ range (Figure S1). The presence of some mesoporosity was confirmed both by Non-linear Density Functional Theory (NLDFT) pore size distribution (PSD) and the t-plot method, as shown in Table 1. All of the samples exhibited comparable Brunauer-Emmett-Teller (BET) surface areas (S_BET ≈ 450–500 m² g⁻¹) and micropore volumes (V_micro ≈ 0.20–0.23 cm³ g⁻¹), irrespective of the original choice of metal (Table 1). Elemental analysis (ICP-MS) showed that all of the parent zeolites contained the anticipated loading heteroatoms, corresponding to an Si/M molar ratio of 200, with the exception of Ga-Beta, which showed a lower Si/M ratio (152), indicating a higher-than-anticipated concentration of Ga in the zeolitic framework. Taken together, XRD, DRIFTS, N₂ physisorption, SEM and ICP-MS confirm that all of the zeolites were successfully synthesised, and possessed comparable porosity, topology, and composition, making them suitable parent materials for the SSI of Sn.

Following their successful synthesis, all of the parental zeolites were demetallated in order to generate the vacant framework sites required for the incorporation of Sn [12]. Although some of the materials can be readily demetallated in mild conditions (particularly B-Beta, which is known to be hydrolytically unstable) [25,35,36], for comparative purposes, all of the parent zeolites described in Table 1 were demetallated by treatment in HNO₃, in accordance with previous studies (13 M HNO₃, 100 °C, 20 h) [13]. ICP-MS confirmed the successful extraction of all of the original heteroatom content in each sample, without major changes to the overall crystallinity of the material (Figure S2).

Following successful demetallation, the SSI of Sn was carried out by grinding the demetallated zeolites with the appropriate quantity of tin (II) acetate to yield a catalyst with a final metal loading 1 wt.% Sn (Si/Sn = 200), prior to heat treatment at 550 °C for a total of 6 h (3 h in N₂ flow followed by 3 h in air flow). The final post-synthetic catalysts were labelled as 1Sn-Beta_MX, where M denotes the original metal incorporated into the lattice, i.e., M = B, Al, Fe or Ga, and the X value represents the original Si/M molar ratio of the material. The XRD patterns of the final post-synthetic zeolites showed no changes of the overall crystallinity of the material following demetallation–remetallation (Figure S2). In order to provide further insight as to their respective Sn speciation, ¹¹⁹Sn Magic Angle Spinning Nuclear Magnetic Resonance using the Carr-Purcell-Meiboom-Gill echo train acquisition method (¹¹⁹Sn MAS CPMG NMR) was performed [27,28,37]. Spectra were measured in Direct Excitation (DE) mode, with a relaxation time sufficient to fully relax all of the Sn species in a quantitative manner (t₁ = 135 s) [38]. For comparison, a sample of 1Sn-Beta was also prepared from a commercially available aluminosilicate zeolite Beta (Zeolyst^TM, Si/M = 19, labelled 1Sn-Beta_COM19), which we have extensively characterised and benchmarked over recent years [14,38,39,40].

The ¹¹⁹Sn DE CPMG MAS NMR spectra of all of the samples presented a number of resonances between −600 and −800 ppm (Figure 4). In particular, all of the samples displayed a resonance characteristic of extra-framework SnO_x species at −600 ppm. Interestingly, out of the hydrothermally synthesised Beta zeolites with a target Si/M ratio of 200, the degallinated analogue exhibited a lower quantity of SnO_x species, which could be a consequence of the initial material having a larger-than-expected quantity of Ga in the lattice of the parental zeolite (Si/M = 154, see Table 1). Moreover, lower quantities of extra-framework Sn sites were also observed for the two zeolites prepared from parental materials with an initial Si/M ratio of 100 (1Sn-Beta_Al100 and 1Sn-Beta_B100, Figure 4 and Figure S3). In addition to the clear trends in terms of the initial Si/M ratio and quantity of extra-framework SnO_x, it is also clear that all of the synthesised samples presented higher quantities of SnO_x than 1Sn-Beta_COM19, indicating the incorporation of Sn to be less successful for the parental materials made by fluoride-route hydrothermal synthesis. This could be due to (i) the even lower Si/M ratio of the commercial material, resulting in a greater number of vacant framework sites following detmetallation; (ii) the larger crystallites generated by fluoride media preparation causing the clustering of Sn on the external regions of the material [41]; and/or (iii) differences in the chemical properties of the crystallites (e.g., hydrophobicity), impacting the incorporation mechanism [14,42].

Notably, a variety of other Sn signals resonating in the range of −650 to −737 ppm were also detected in the analysed samples. In general, resonances in this region can all be attributed to Sn atoms isomorphously substituted into the framework in a hydrated state [27,37,43]. However, the precise chemical shift and the relative intensities of these signals varied widely based on the choice of metal and the original Si/M ratio of the material (Al- and B-derived Beta zeolites), even though each material was synthesised at the same loading Sn and by the same preparation method (Figure 4).

Unfortunately, the relationship between the identity of the Sn of the active site(s) within the Beta framework and the chemical shift value observed by ¹¹⁹Sn MAS NMR is still not fully understood, and the scientific community is still divided on this matter [44,45,46,47]. However, it is generally appreciated that the different chemical shift values observed in this region arise from the presence of different Sn(IV) species, which could indicate: (i) Sn incorporated in different T-sites of the zeolitic framework; (ii) a different ratio of open or closed sites in the final catalyst; and/or (iii) different quantities of residual fluoride in some materials, which may perturb the Sn species and thus their associated chemical shifts [28,45]. As such, it can generally be concluded from Figure 4 that the identity and concentration of the initial heteroatom in the Beta framework ultimately results in the formation of slightly different active sites in the final Sn-Beta catalyst. However, it can be more specifically concluded that the original Si/M ratio of the parental zeolite is important for the maximisation of the ratio of framework to extra-framework Sn species, and that Sn-Beta catalysts derived from a commercial parental zeolite (i.e., Sn-Beta_COM19) appear to possess a lower quantity of extra-framework SnO_x relative to the counterparts synthesised by the fluoride route.

3.2. Kinetic Evaluation of the Synthesised Catalyst Materials

In order to evaluate the performance of the post-synthetic catalysts, and hence to identify the consequences of the different active site species present in the sample series (Figure 4), the post-synthetic 1Sn-Beta_MX series was evaluated for activity regarding the continuous conversion of glucose to α-hydroxy esters (MVG and ML) via retro-aldol fragmentation (Scheme 1). The 1Sn-Beta_MX samples were also compared to 1Sn-Beta_COM19, which has been used in our previous studies and thus provides context to the experimental trends. The initial experiments were performed in established literature conditions [18,39], using a feed of 1 wt.% glucose in methanol and a reaction temperature of 160 °C. A high Weight Hourly Space Velocity (WHSV = 4.75 h⁻¹) was employed in all of the reactions in order to achieve accurate kinetic data, and the reactions were initially undertaken for a period of two hours of operation in order to gain preliminary insight into the performance of each catalyst. Because many of the intermediate products formed in the retro-aldol fragmentation reaction are challenging to identify by chromatographic methods, the progress of the reaction was determined by measuring the quantity of glucose converted by each catalyst (Equation (1)), and by determining the yields of fructose, mannose, methyl lactate and methyl vinyl glycolate (Equation (2)), because these are known to be the most dominant products formed throughout the cascade reaction (Scheme 1). The overall carbon balance of the effluent solution gathered at the second hour of time on stream was initially used for comparative purposes (Figure 5).

Differences in the overall yield of products and reaction selectivity were evident within the 1Sn-Beta_MX series. The highest combined yields to retro-aldol products were observed for both B-derived catalyst materials, with 1Sn-Beta_B200 in particular exhibiting the highest carbon balance due to its enhanced yield of α-hydroxy ester compounds (a combined yield above 70% at 2 h of time on stream). In contrast, the remaining 1Sn-Beta_MX catalysts exhibited lower α-hydroxy ester yields of between 35–60%.

However, it is notable that all of the catalysts within the 1Sn-Beta_MX series presented much higher initial yields of, and selectivity to, α-hydroxy ester compounds, than those observed for 1Sn-Beta_COM19, despite all of the catalysts exhibiting a comparable degree of glucose conversion (approximately 90%). It is thus clear that despite possessing larger quantities of extra-framework SnO_x species (Figure 4), the 1Sn-Beta_MX catalyst materials were better-performing catalysts for the retro-aldol fragmentation of glucose than 1Sn-Beta_COM19 from an activity and selectivity perspective.

In order to acquire a more comprehensive evaluation of the overall catalytic performance and stability of the catalysts, glucose upgrading at 160 °C was performed in a continuous flow for up to 72 h on stream. We note that the operational conditions were established such that the quantity of substrate converted was sub-maximal at the initial hours of the reaction, thereby allowing intrinsic stability to be evaluated, i.e., to avoid an excess catalyst regime [48,49,50]. Figure 6 presents the time-on-stream data for each catalyst at identical conditions (temperature, flow rate, catalyst mass and, hence, WHSV). As can be seen, all of the catalysts broadly exhibited similar trends during continuous operation, showing an initially rapid rate of deactivation in the first 20 h of reaction, during which the glucose conversion decreased from ca. 95 to 50%, but with activity stabilising thereafter. The loss of activity over the initial reaction period was accompanied by corresponding decreases to the yields of the desired α-hydroxy ester products. The fructose yields increased in this timeframe due to the monosaccharide being an intermediate product in the overall reaction network, and after the first 20 h of time on stream fructose became the main reaction product in all cases.

However, although all of the catalysts showed relatively similar kinetic trends, more detailed analysis of the product distribution revealed that the choice of preparation methodology and starting heteroatom did result in some differences in the activity, selectivity, and stability of the catalyst. For example, Figure 7 presents the α-hydroxy ester (ML and MVG) selectivity as a function of glucose conversion for the various reactions reported in Figure 6. This figure provides a comparison of the combined α-hydroxy ester selectivity for each catalyst at all of the levels of conversion for the analysed materials. Of the catalysts investigated, the ones obtained by means of fluoride synthesis with Al- and B- as starting heteroatoms exhibited better selectivity to α-hydroxy ester compounds, with a selectivity value of approximately 50% obtained at 80% substrate conversion. Interestingly, 1Sn-Beta_COM19 showed much lower selectivity, displaying only 17% selectivity to the desired esters at the same glucose conversion (80%). This demonstrates how the catalyst synthesised by the commercial precursor is much more prone to catalysing different reaction pathways that are not easily detectable by common chromatographic analysis, resulting in a decreased level of carbon efficiency.

In order to better probe the catalyst stability, the kinetic data presented in Figure 6 were converted to the Levenspiel function (Equation (4)) to calculate the rate of deactivation in each experiment (k_d, Figure 8). Overall, 1Sn-Beta_COM19 was shown to have the highest rate of deactivation (k_d = 0.048 h⁻¹), which was approximately 50–60% faster than the rates of deactivation determined for the 1Sn-Beta_MX catalyst series. Consequently, the catalyst derived from the commercial zeolite was both less selective (Figure 5) and less stable compared to the catalysts from the 1Sn-Beta_MX series. Additionally, a comparison of the α-hydroxy ester productivity of each catalyst (g_{ML + MVG} g_catalyst⁻¹ h⁻¹) was made (Figure 8). The higher α-hydroxy ester selectivity of 1Sn-Beta_B200, coupled with its slightly lower rate of deactivation, resulted in this catalyst presenting a higher productivity per gram of catalyst, which was almost twice as large as that obtained with 1Sn-Beta_COM19. Therefore, this material not only offers the possibility of employing a simpler demetallation procedure [23] but also delivers better performance in the process. The ML/MVG molar ratio was determined to be approximately 1.5–2 in all of the cases (Figure S4).

3.3. Effect of Water and Alkali Salts on the Kinetic Performance

The previous section demonstrated how improvement in the activity (productivity), selectivity and stability of post-synthetic Sn-Beta could be achieved by varying the choice and the loading of the heteroatom in the parent M-Beta zeolite. On a balance of these activity indicators [50], the best performance was achieved for Sn-Beta materials prepared from a parental borosilicate Beta zeolite, in particular 1Sn-Beta_B200. However, in the conditions described above, the overall yield of retro-aldol products (ML and MVG) obtained from 1Sn-Beta_B200 decreased from approximately 70–75% at short times on stream (<5 h) to less than 20% after 30 h on stream. Although these yields are comparable to previous reports addressing the catalytic conversion of glucose to α-hydroxy ester compounds, the selectivity and the lifetime of 1Sn-Beta_B200 towards the production of the desired products (ML and MVG) could be improved.

Towards this objective, previous studies in the literature have shown how slight changes in the reaction conditions can improve the selectivity performance of Sn-Beta for the retro-aldol fragmentation of glucose to α-hydroxy ester compounds, as well as the overall catalyst performance [51,52,53,54,55,56]. In particular, the addition of alkali salts to the reaction feed has been shown to increase methyl lactate selectivity during the retro-aldol fragmentation of glucose by a pronounced level [51,52,53,54]. In addition, we have shown that an increase in the stability of Sn-Beta catalysts derived from commercial aluminosilicate precursors can be achieved by adding small amounts (1–10% w/w) of water to the sugar/MeOH feed, mitigating deactivation under continuous flow reactions [37]. Hence, in order to gauge the potential performance of the 1Sn-Beta_MX materials in the presence of these promoters, the retro-aldol fragmentation of glucose over 1Sn-Beta_B200 was carried out with the addition of KCl (0.335 mol) and water (1:99 H₂O:MeOH v/v) to the reaction feed. The concentration of each additive was based on recent studies [39,52]. For these final studies, 1Sn-Beta_B200 was used as the catalyst, and its performance was compared to 1Sn-Beta_COM19 following similar perturbation with water and alkali salt. The non-perturbed data (i.e., the activity of both catalysts in the absence of water and alkali salt promoters) is presented in Figure 6.

Figure 9 presents the performance of 1Sn-Beta_COM19 (A) and 1Sn-Beta_B200 (B) for glucose conversion at 160 °C in the presence of alkali salt and water promoters. As can be seen, a drastic improvement in performance (glucose conversion and α-hydroxy ester yield) and stability was achieved for both catalysts following the addition of the alkali salt and water to the feed. In particular, the initial yields of ML and MVG increased from 24% and 12% to approximately 60% and 24%, respectively, following the addition of KCl and water to the reaction catalysed by 1Sn-Beta_COM19. The consequences of adding water and the alkali salt to the reaction feed were also evaluated for 1Sn-Beta_B200. As can be seen, the combined yield of ML and MVG increased from approximately 70% in the absence of the activity promoters to approximately 90% in the presence of water and alkali. The improved performance of 1Sn-Beta_B200 in the presence of water and KCl was maintained throughout the operational period, with combined yields of ML and MVG > 30% remaining even after 70 h of operation [53].

In order to elucidate the effect of water and the alkali salt further, the initial carbon balance and the α-hydroxy ester selectivity of both catalysts (1Sn-Beta_COM19 and 1Sn-Beta_B200) were compared in the absence and presence of alkali and water (Figure 10A). In doing so, it is clear that both catalysts showed an increase in selectivity in the presence of water and alkali, which led to α-hydroxy-esters yields above 80% for both materials (Figure 10B), in contrast to yields in the region of 50–60% (for 1Sn-Beta_COM19) in the absence of water and KCl. Although both catalysts resulted in higher yields following the addition of water and the alkali salt, Figure 10A showed that the selectivity to α-hydroxy esters at the same rate of conversion was always higher for 1Sn-Beta_B200 throughout the entire reaction, both in the absence and in the presence of the alkali salt. This indicates that a greater quantity of substrate was converted into undesired products throughout the entire reaction with 1Sn-Beta_COM19, leading to a loss of valuable substrate that could otherwise have been recycled post reaction. Furthermore, the addition of alkali salts also significantly decreased the production of the intermediate fructose over 1Sn-Beta_B200 (Figure S5), whereas 1Sn-Beta_COM19 yielded a similar selectivity to fructose at the same level of conversion irrespective of whether alkali salts were present. Given the importance of the maximum carbon efficiency during biomass conversion, the improved selectivity of 1Sn-Beta_B200 throughout the entire reaction coordinate is notable.

In contrast, the stability of the material over 72 h of time on stream followed a slightly different trend. In fact, although the addition of water enhanced the stability of both catalysts, the enhancement in stability was greatest for 1Sn-Beta_COM19, which experienced only a 20% loss of activity of the reaction period in the presence of water. In contrast, 1Sn-Beta_B200 experienced a loss of 40% activity over the same 72 h of time on stream. This indicates that the beneficial role of water in the maximisation of stability is somewhat less effective for 1Sn-Beta_B200 than for 1Sn-Beta_COM19, which is likely due to the more hydrophilic characteristics of 1Sn-Beta_COM19 compared to 1Sn-Beta_B200, as evidenced by water sorption studies (Figure S6) [57,58,59]. The enhanced hydrophilicity of 1Sn-Beta_COM19 likely favours its interaction with water, and as a result ensures that the promotional effect of water in terms of maximising stability is observed for a longer period.

Based on the results reported in Figure 9 and Figure 10, it is clear that although the addition of water and alkali was beneficial for both 1Sn-Beta_COM19 and 1Sn-Beta_B200, the better performance of 1Sn-Beta_B200 makes it much less reliant on the use of activity promoters than the commercially-derived analogue. Moreover, the improved selectivity of 1Sn-Beta_B200 in all of the tested conditions is evident. Thus, it is clear that despite possessing a less homogeneous distribution of active Sn sites, and hence likely containing a lower quantity of active Sn sites per gram of catalyst, the catalyst derived from B-containing Beta zeolite is the most favourable catalyst for the retro-aldol fragmentation of glucose amongst the materials studied in this report.

4. Conclusions

This study focuses on how the choice of starting zeolite material impacts the preparation and performance of post-synthetic Sn-Beta catalysts prepared by solid-state incorporation for the retro-aldol fragmentation of glucose. In particular, four different metal-incorporated Beta zeolites (M-Beta, M = Ga-, Fe-, B-, Al-) were explored as parent materials for the top-down synthesis of Sn-Beta via demetallation–remetallation, and their ultimate performances were explored for the retro-aldol fragmentation of glucose to the α-hydroxy ester compounds methyl lactate and methyl vinyl glycolate. ¹¹⁹Sn MAS DE CPMG NMR, alongside XRD, IR, DRIFTS, ICP-MS and porosimetry analysis, demonstrated the successful synthesis of Sn-Beta in all of the cases, albeit with a wide array of different isomorphously-substituted Sn species being present in each sample.

The kinetic performance of each catalyst was evaluated in a continuous flow reactor, and revealed that all of the synthesised materials were active for this reaction, and hence, that different metal precursors can be employed as starting materials for the synthesis of Sn-Beta by SSI. The catalyst synthesised from the B-Beta starting material was shown to exhibit the best performance in terms of productivity and selectivity to the desired α-hydroxy-ester compounds.

Based on these findings, we can conclude that B-derived zeolites are a feasible alternative to Al-containing zeolites for SSI. This finding may be advantageous in the long term, because the hydrolytically unstable nature of borosilicate zeolites has been shown to result in their easier demetallation compared to their aluminosilicate analogues, potentially avoiding the use of highly acidic reaction solutions to remove the original heteroatom from the zeolite lattice during demetallation.

Along with identifying that B-containing materials exhibit a higher potential to act as parental zeolites for solid-state incorporation, this study also reveals broader guidelines for the successful synthesis of Sn-Beta catalysts. Namely, although ¹¹⁹Sn MAS DE CPMG NMR suggested that the 1Sn-Beta_MX series possessed a less favourable distribution of active sites than the catalyst derived from commercial Al-Beta (1Sn-Beta_COM19), this series nevertheless presented higher levels of activity and stability compared to the commercially-derived counterpart. Although this trend was somewhat reversed when activity promoters were added to the solution, this suggests that the different lattice properties achieved through the fluoride medium preparation (i.e., increased hydrophobicity, decreased defect site density, larger crystallite sizes) are more determining in terms of the achievement of high performance without resorting to activity promoters.