Glucose, fructose, mannose, and 5-hydroxymethyl furfural were obtained from Aladdin Reagent Co., Ltd., (Shanghai, China). Tin (II) acetate (99%), methyl lactate (99%), Pyruvic aldehyde dimethyl acetal, magnesium nitrate (99.0%), commercial H-β zeolite (Si/Al = 12.5), HNO₃ (65%–68%) were supplied by Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). N₂ (99.999%) was purchased from Linde Industrial Gases.

X-ray powder diffraction (XRD) patterns were recorded on Rigaku UItima IV X-ray diffractometer with Cu Kα radiation source (40 kV and 30 mA) from 5 o to 60 o and a scan speed of 10^o/min. The adsorption-desorption isotherms of N₂ were obtained on a Micromeritics ASAP 2020 M instrument. The surface area was calculated via the Brunauer–Emmett–Teller (BET) equation, and the volume of pores was acquired via the single point method. X-ray photoelectron spectra (XPS) were recorded with a Quantum 2,000 Scanning ESCA Microprob instrument (Physical Electronics) using Al Kα radiation. The binding energy was calibrated by using the C1s photoelectron peak at 150 eV as a reference. Quantification of Mo and Mg elements was conducted using a Bruker S8 Tiger X-ray fluorescence (XRF) spectrometer. Fourier transform infrared (FTIR) spectra were collected on a Nicolet 6,700 instrument equipped with an MCT detector at a spectral resolution of 4 cm⁻¹. A self-supporting pellet made of the sample was placed in the flow cell and evacuated under reduced pressure at 623 K for 1 h. After cooling to 373 K, the samples were saturated with pyridine vapor and then evacuated. Subsequently, spectra were recorded at the same temperature in the 4,000–650 cm⁻¹ range by using the coaddition of 64 scans.

The reactions were conducted in a batch-type Teflon-lined stainless steel autoclave reactor with an inner volume of 50 ml. Typically, the glucose and catalyst were added to 20 ml of methanol. Then the autoclave was sealed, purged with N_2, and pressurized with N₂ to 2 MPa. Subsequently, the system was heated up to 373 K. After continuous magnetic stirring (800 rpm) for a given time, the autoclave was cooled to room temperature in cold water, and the liquid was obtained after filtration and analyzed by GC and HPLC. Sample products were first qualitatively analyzed using an Agilent GC-MS equipped with a 1701-ms column. The crude product was dissolved in a silylating agent (bis-(trimethylsilyl) trifluoroacetamide +1% trimethylchlorosilane). This solution was maintained at 65°C for 2 h to ensure the complete silylation which was subsequently injected into the GC-MS. Some peaks in Agilent HPLC were further identified using LC-MS. Quantitative analysis was then performed on an Agilent HPLC system equipped with RI and UV-Vis detectors and a Bio-Rad Aminex HPX-87H ion exclusion column (300 × 7.8 mm), using 0.005 M H₂SO₄ as the mobile phase at a flow rate of 0.6 ml min⁻¹. The column temperature and the detector were both set at 50°C. The number of varied products was determined using calibration curves generated with standard solutions.

The initial catalytic tests were performed over Sn-β catalysts and Mg-modified counterparts at moderate temperatures. As present in Figure 3, in the absence of Mg modification, fructose was the primary product and accompanied by the production of a small amount of PADA (Pyruvic aldehyde dimethyl acetal), and the yields of methyl lactate were below 10% at 100°C after 4 h over Sn-β catalysts. Comparatively, the Mg-modified Sn-β catalysts promoted the glucose conversion remarkably but still failed to increase the selectivity toward methyl lactate remarkably. The main product was also fructose, implying that the Sn sites alone can catalyze the glucose isomerization to fructose, but exhibited a rather limited catalytic activity for the retro-aldol reaction of fructose at 100°C.

The influence of Mo modification on the catalytic performance of the Sn-β catalyst is shown in Figure 4. Without Mo modification, merely a 5% yield of methyl lactate with a glucose conversion of 51.7% was achieved. By contrast, the glucose conversion was increased by above 30% over Sn-β loaded with Mo catalysts, accompanied by promotion in methyl lactate (MLA) yield to a significant extent. The MLA yield first increased and then decreased with the loading content of Mo, reaching a maximum of 16% at 3 wt% of loading content. The 3 wt% Mo enabled the MLA selectivity to increase from 10.3% to 18.9%, revealing that Mo⁶⁺ species promoted the retro-aldol condensation of fructose to C3 intermediates for the formation of MLA. However, Mo content exceeding 3 wt% resulted in a decrease in both MLA yield and selectivity, which is probably because the excessive Mo species blocked the pores of Sn-β and thus hindered the access of fructose to the Sn sites as reflected in Table 1. Surprisingly, 36.8% of selectivity and 35% yield of MLA were achieved over Mg, Mo co-modified Sn-β catalyst, outperforming the sole Mg or Mo modified counterparts. This implied that a synergistic effect exists between Mg and Mo promoters.

Given that the modification of Mg is capable of facilitating the MLA formation, the effect of Mg content on the MLA yield was further investigated. As present in Figure 5, with Mo content fixed at 3 wt% and Mg content range of 0.1%–2%, the maximum yield of MLA was attained at 0.5 wt% Mg. The higher content of Mg cannot further contribute to the MLA formation.

The catalytic performances at various reaction temperatures were also studied, and the results are shown in Supplementary Figure S1. The MLA yield was increased by elevating the reaction temperature in the range of 100–160°C over all the Sn-β catalysts, regardless of the modification. The highest MLA yield was achieved at 160°C over 3% Mo-0.5% Mg-Sn-β catalyst. This suggests that the kinetic-relevant retro-aldol condensation step is further favored at a higher temperature, against the competing side reactions. However, the structure of zeolite would be undermined after such severe reaction conditions, which limits its practical application (Hu et al., 2020). We did perform the recycling test to evaluate the durability of the catalyst at 100°C. After each run, the spent catalyst was recovered from the reaction mixture, which was subsequently washed with deionized water, dried overnight, and calcinated at 550°C for 5 h in a muffle furnace. Compared to the fresh catalyst, the spent catalyst after four cycles showed a slight loss of activity. This result suggests that the catalyst is relatively stable under moderate temperature conditions (Supplementary Figure S2).

To understand the catalytic function of distinct function sites in-depth, we conducted the conversion of different substrates in the presence of various catalysts within a short reaction time. As depicted in Table 2, upon glucose as the substrate, no MLA formed over the 3% Mo-β catalyst, indicating that the Mo species alone is unable to catalyze the glucose conversion to MLA. A 25% yield of fructose, accompanied by a 2% yield of MLA, was obtained over Sn-β suggests that the Sn species is capable of catalyzing the glucose isomerization but less for retro-aldol condensation of fructose. The latter is the kinetic-relevant step in MLA formation. The co-existence of Sn and Mo species enabled an increase in both glucose conversion and MLA yield, indicating that both species are responsible for glucose isomerization. Regarding the fructose substrate, a higher fructose conversion was observed over 3% Mo-Sn-β catalyst, compared to that over Sn-β catalyst. The PADA also formed during the reaction. These results clearly suggest that Mo species can effectively catalyze the retro-aldol condensation. Nevertheless, without the assistance of Sn sites, the MLA can hardly form. This is because the C3 intermediates (1,3-dihydroxyacetone and pyruvaldehyde) generated through the retro-aldol reaction of fructose still need to undergo a 1,2-H shift reaction prior to MLA formation, which has to be catalyzed by the isolated Sn sites featured of strong Lewis acidic property. Unfortunately, the Mo species does not possess the corresponding catalytic capacity, thus the employment of which alone can hardly produce any MLA. This speculation was further confirmed by the results using 1,3-dihydroxyacetone (DHA) as the substrate. A distinct difference exists in MLA yield and selectivity provided by Sn-β catalyst and 3% Mo-β catalyst, respectively. The former catalyst afforded a selectivity of 68.4% toward MLA, while merely 2.5% was obtained over the latter catalyst. The addition of 3% Mo into Sn-β even caused a notable decrease in MLA yield, further confirming the argument made earlier.

It should be noted that the amount of substrate used in the tests above is quite low, which may cause concern over its potential practical application. Hence, we also performed a series of tests with much larger ratios of substrate/catalyst amount. The results in Supplementary Figure S3 show that comparable MLA yields as earlier were obtained, even with the glucose amount increased to 0.25 g and the catalyst reduced to 0.16 g.

It has been reported that the use of varied solvents, such as methanol and ethanol, can make quite a notable difference over the yield of the desired products (Orazov and Davis, 2015). In our work, the solvents appeared not to affect the formation of alkyl lactate significantly upon glucose as the substrate but did cause a drastic change in the case of fructose, as exhibited in Figure 6. The yield of alkyl lactate is only about 25% in methanol solvent, while reached up to 61.3% in ethanol solution. Further experiments of fructose conversion over Sn-β and 3% Mo-β catalyst were conducted, respectively. The results in Supplementary Figure S5 show that in the presence of Sn species alone, the fructose conversions in methanol and ethanol were quite similar, while Mo species alone provided a much higher conversion of fructose in ethanol than that in methanol. It was then speculated that the Mo sites were more prone to be affected by the solvent, rather than the Sn sites. Mo species exhibited a lower catalytic performance in retro-aldol condensation of fructose in methanol. It may be attributed to the stronger binding of Mo sites with methanol, hindering the absorption and turnover of fructose on Mo sites.

The aforementioned results demonstrate the indispensable role of Mo and Mg over the modified Sn-β catalysts toward the moderate temperature conversion of C6 monosaccharide. Further experiments show that the physical mixture of 0.5% Mg-β and Sn-β only resulted in about 5% ELA, suggesting neither of them can effectively catalyze the retro-aldol condensation at the moderate temperature. Moreover, the combination of 0.5% Mg-β and 3% Mo-Sn-β could achieve an ELA yield of 54%, which is less than that (61.3%) of 0.3% Mo-0.5% Mg-Sn-β. This demonstrated that the MgO species should stay in close proximity to Mo or Sn sites to have a synergy effect and enhance the yield of ELA. A question then arises whether the Mg species have a synergy effect mainly with Mo or Sn sites in the working catalyst. Hence, we prepared Sn-β, 3% Mo-β and their Mg-modified counterparts to shed light on the intrinsic mechanism. Indeed, the physical mixture of Sn-β and 3% Mo-β in ethanol results in a relatively lower ELA yield at 52%, thanks to the absence of Mg species. The combination of 0.5% Mg-Sn-β and 3% Mo-β can improve the yield of ELA to 60%, comparable to that of 0.3% Mo-0.5% Mg-Sn-β. Thanks to the lack of catalytic ability for the retro-aldol condensation reaction, the addition of basic MgO in Sn-β could probably tune the Brønsted acid sites near the Sn sites, thus mitigating the side-reactions from fructose and intermediate products such as DHA/GLA during the conversion (Hu et al., 2020). However, the employment of Sn-β and 0.5% Mg-3% Mo-β can achieve the highest yield up to 70%, which can be further improved to nearly 82.8% with the extended reaction time (Supplementary Figure S4). This result is superior to most of the prior works reported in the literature, as compared in Supplementary Table S1. It appears that MgO species can cooperate more effectively with Mo-site during the rate-determining retro-aldol reaction step, thus significantly mitigating the side reactions. The weakened Brønsted acid sites near the Mo sites should play a role to abate the production of side products, similar to that of Mg-Sn-β. Pyridine-probed FTIR analysis could readily distinguish the Bronsted and Lewis acid sites. As shown in Supplementary Figure S6, the strength of Bronsted acid was weakened after adding Mg to the 3% Mo-Sn-β catalyst. Our earlier work shows that MgO species attached to the zeolite framework can have a synergy effect with the Lewis Sn site to directly enhance the retro-aldol condensation at a more severe temperature like 160°C. Through functioning as Lewis acid-base pair, MgO can facilitate the abstract of the proton from the OH groups of fructose and stabilize the deprotonated alkoxide that is subsequently subjected to retro-aldol condensation catalyzed by the strong Lewis Sn sites (Hu et al., 2020). It is speculated that a similar mechanism can take place between MgO and the nearby Mo sites in this work, where fructose can be readily deprotonated and stabilized for the subsequent retro-aldol condensation. Hence, the side reactions were further restricted to a very limited extent, leading to a much higher yield of ELA than the other cases listed in Table 3.

Based on the results above, a mechanism is proposed as Scheme 1 below. The ketonic oxygen of fructose adsorbs onto the Mo sites first, followed by the occurrence of retro-aldol condensation to generate 1,3-dihydroxyacetone (DHA) and glyceraldehyde (GLA). In this process, the framework O atom of Si-O-Mg, as the Lewis base site, assists to abstract the H atom of -OH attached to the C3 of fructose, while Mg²⁺ is the Lewis acidic site can stabilize the deprotonated alkoxide. A keto-enol tautomerization exists between DHA and GLA. The formed GLA undergoes dehydration and addition reaction with ethanol molecule, generating the hemiacetal compound 1. Subsequently, compound 1 is transformed into ethyl lactate via 1,2-H shift as catalyzed by Sn sites. The side reactions are significantly suppressed during the whole process, thanks to the addition of MgO species.