Highly Selective Oxidation of Ethyl Lactate to Ethyl Pyruvate Catalyzed by Mesoporous Vanadia-Titania.

The direct oxidative dehydrogenation of lactates with molecular oxygen is a "greener" alternative for producing pyruvates. Here we report a one-pot synthesis of mesoporous vanadia-titania (VTN), acting as highly efficient and recyclable catalysts for the conversion of ethyl lactate to ethyl pyruvate. These VTN materials feature high surface areas, large pore volumes, and high densities of isolated vanadium species, which can expose the active sites and facilitate the mass transport. In comparison to homogeneous vanadium complexes and VO x /TiO2 prepared by impregnation, the meso-VTN catalysts showed superior activity, selectivity, and stability in the aerobic oxidation of ethyl lactate to ethyl pyruvate. We also studied the effect of various vanadium precursors, which revealed that the vanadium-induced phase transition of meso-VTN from anatase to rutile depends strongly on the vanadium precursor. NH4VO3 was found to be the optimal vanadium precursor, forming more monomeric vanadium species. V4+ as the major valence state was incorporated into the lattice of the NH4VO3-derived VTN material, yielding more V4+-O-Ti bonds in the anatase-dominant structure. In situ DRIFT spectroscopy and density functional theory calculations show that V4+-O-Ti bonds are responsible for the dissociation of ethyl lactate over VTN catalysts and for further activation of the deprotonation of β-hydrogen. Molecular oxygen can replenish the surface oxygen to regenerate the V4+-O-Ti bonds.

Introduction

Lignocellulosic biomass is attracting increased attention as a renewable carbon source for commodity chemicals.[1] Unlike the case with fuel applications, the high oxygen content and diversity of biomass-derived “platform molecules” make them suitable feedstocks for high-added-value chemicals.[2] Lactic acid (LA) and lactates are such platform molecules. They can be converted to several commodity chemicals, including acrylic acid, pyruvic acid, lactide, 1,2-propanediol, and acetaldehyde.[3] The direct catalytic air oxidation of LA is a promising route to pyruvic acid,[4] a key intermediate in the pharmaceutical, agrochemical, and food additive sectors.[5] Pyruvic acid occurs naturally as an intermediate product in carbohydrate and protein metabolisms in the body. Since it can be converted to carbohydrates, to fatty acids or energy, or to the amino acid alanine, it is a key intermediate in several metabolic processes. Pyruvic acid salts and esters (pyruvates) are used as dietary supplements. For example, calcium pyruvate is used as a fat burner in the food industry. Pyruvic acid is used as a starting material for the synthesis of pharmaceuticals, such as l-tryptophan and l-tyrosine, and for the synthesis of amino acids such as alanine and phenyl alanine. Its derivatives are also used to produce crop protection agents, cosmetic agents, and flavoring ingredients. Pyruvic acid is also a reagent for regeneration of carbonyl compounds from semicarbazones, phenylhydrazones, and oximes. Currently, pyruvates are still made via the energy-intensive pyrolysis of tartaric acid with stoichiometric KHSO4 as a dehydrating agent. Ideally, this would be replaced by direct oxidative dehydrogenation using molecular oxygen, giving water as the only byproduct. Even though this reaction is thermodynamically feasible, its selectivity is low, because LA is easily overoxidized.[6]

Several groups have studied this oxidation in the vapor and liquid phases. In vapor-phase reactions, various catalysts including iron phosphates and multicomponent mixed oxides were reported.[7−9] However, such high-temperature routes are energy intensive and can lead to the decomposition of pyruvates via decarbonylation or decarboxylation.[10] The side reactions can be suppressed in liquid-phase systems,[11] but the milder conditions typically require noble-metal catalysts for oxygen activation.[12] Hayashi et al. reported Pd–metal alloy catalysts on activated carbon for the synthesis of pyruvate in the presence of NaOH.[13] Ding and co-workers used bimetallic Pb–Pt supported on carbon materials for the synthesis of pyruvic acid in good yields, using an excess of LiOH to adjust the pH value.[14] Despite these achievements, it is still essential to develop a simple and cost-effective catalytic system for the oxidation of lactates to pyruvates with oxygen under mild conditions.

Vanadium is an abundant element that is often used as a catalyst in the selective oxidation of light alkanes and alcohols.[15] Yasukawa et al. studied the oxidation of ethyl lactate to pyruvate over various homogeneous vanadium compounds in a gas–liquid microflow system.[16] They found that vanadium oxytrichloride (VOCl3) gave the highest yield (31%) at room temperature, making it 5 times as active as V2O5.[17] Considering catalyst recovery and product purification, solid vanadium catalysts are preferable.[18] Moreover, the stability of unsupported vanadium compounds is an issue, as they are easily hydrolyzed in the presence of H2O produced during the reaction. Previously, we showed that titania is a suitable support for catalyzing the lactate to pyruvate reaction.[19] In principle, highly dispersed vanadyl species on anatase TiO2 are desired for an efficient oxidation. However, the density of isolated vanadyl species is limited by the titania surface area, and distributing the sites evenly on the surface is problematic at high loadings. In addition, these catalysts are also prone to leaching.[20] On the basis of our earlier experience with doped mixed oxides for the oxidative dehydrogenation of propane,[21] we hypothesized that doping vanadium ions into the mesoporous TiO2 lattice would solve these problems. Titania–vanadia hybrids have been made using various methods, including hydrothermal,[22,23] sol–gel,[24] electrochemical,[25] spray pyrolysis,[26] and coprecipitation.[27] Gopinath et al. studied wormhole-like mesoporous vanadium-doped titania in the oxidative dehydrogenation of ethylbenzene to styrene, attributing the high activity to vanadium atoms in the titania lattice.[28] However, in most of the cases, the vanadium sites embedded in the titania matrix were unavailable for catalysis. Moreover, mesoporous TiO2 typically has a low crystallinity, especially when soft template methods are used, leading to low stability.[29,30] Thus, we had to find some way to expose the active sites and shorten mass/energy diffusion pathways as well as improve the stability of the catalyst.

Here we report a one-pot synthesis of mesoporous vanadia–titania nanocrystals (meso-VTN) via the coassembly of vanadium and titanium precursors in the presence of an amphiphilic triblock copolymer as a templating agent. These new materials feature a large surface area and a high density of isolated vanadium species. These meso-VTN catalysts outperformed the homogeneous vanadium complexes and classical VO-TiO2. We also studied the role of various vanadium precursors in the formation of meso-VTN. Only NH4VO3, which has the VO3– anion, provided a strong electrostatic interaction with the Ti4+ cations, yielding an anatase-dominant structure. Increasing the vanadium loading also increases the number of active vanadium dopant sites in the titania lattice, which in turn improves activity and selectivity. We also ran in situ DRIFT spectroscopy studies and DFT calculations, which revealed the relation between the structure of the meso-VTN catalyst and its activity in the ethyl lactate oxidation.

Results and Discussion

Catalyst Synthesis and Testing

First, the coprecursors ammonium metavanadate (NH4VO3) and titanium isopropoxide were dissolved in acidic ethanol solution, and then the resultant mixture was assembled with F127 template via evaporation-induced self-assembly. Subsequent calcination removed the template (Figure S1), yielding a highly crystalline meso-VTN (denoted as NH4VO3@VTN). Three other meso-VTN materials were prepared similarly, starting from different vanadium precursors (VOSO4, VO(acac)2, and VCl3; see the experimental section in the Supporting Information for details).

We then studied the catalytic performance of these meso-VTN materials in the liquid-phase aerobic oxidation of ethyl lactate to ethyl pyruvate (Table ). In comparison to homogeneous vanadium complexes, all of the meso-VTN samples showed good catalytic activity and higher turnover frequencies (TOFs). Control reactions showed that the vanadium precursors themselves were active as oxidation catalysts, but not selective. In addition to the intrinsic disadvantages of homogeneous catalysts in catalyst recovery and product purification, the homogeneous vanadium compounds are unstable and are easily hydrolyzed by the water byproduct.[31,32] In contrast, our meso-VTN solid can be reused at least 10 times without losing activity (see recycling experiments below). The highest TOF of 118 h–1 was measured for NH4VO3@VTN.

Table 1

Oxidative Dehydrogenation of Ethyl Lactate with O2 to Ethyl Pyruvate over Various Catalystsa

entry	catalyst	conversn (%)b	selectivity (%)	yield (%)b	TOF (h–1)c
1	VCl3	63.9	12.3	7.9	0.5
2	VOSO4	63.4	32.9	20.9	1.4
3	VO(acac)2	79.9	64.1	51.2	5.7
4	NH4VO3	64.3	72.6	46.7	2.3
5	V2O5	13.9	61.8	8.6	0.3
7d	VCl3@VTN	20.8	85.1	17.7	61
8e	VOSO4@VTN	25.6	71.3	18.3	86
9f	VO(acac)2@VTN	28.8	78.5	22.6	97
10g	NH4VO3@VTN	34.6	89.4	30.9	118

Reaction conditions: amount of all catalysts 50 mg, temperature 130 °C, 1 atm of O2, 4 h, ethyl lactate 8.5 mmol (1.0 g), diethyl succinate (solvent) 2 mL.

Determined by GC using biphenyl as an internal standard.

Turnover frequency for ethyl pyruvate formation calculated as moles of ethyl pyruvate per moles of total vanadium per hour.

VCl3@VTN (0.6 wt % V, determined by ICP analysis).

VOSO4@VTN (0.5 wt % V).

VO(acac)2@VTN (0.5 wt % V).

NH4VO3@VTN (0.6 wt % V).

To understand the high performance of NH4VO3@VTN, we characterized the catalysts. The scanning electron microscopy (SEM) images of the NH4VO3@VTN samples show highly uniform and nearly spherical morphology in large domains (Figure a). The VTN nanoparticles have a rough surface with an average size of ∼10 nm (see inset in Figure a). This size agrees well with the size obtained from the high-resolution transmission electron microscopy (HRTEM) studies (Figure b,c). HRTEM also confirmed the VTN uniformity. A large number of white dots can be observed by TEM over the entire nanostructure, indicating that the mesoporosity is well-dispersed in the VTN framework (see Figure S2 in the Supporting Information).[33,34] The corresponding selected-area electron diffraction (SAED) pattern confirmed a set of diffraction rings (inset in Figure b), in accordance with the crystalline anatase phase (JCPDS No. 21-1272). Figure d also shows the clear lattice fringes with an interplanar distance of 3.52 Å, matching well with the (101) planes of the anatase structure.[35] In addition, the high-angle annular dark-field scanning TEM (HAADF-STEM) images and corresponding elemental mapping further demonstrated the uniform distribution of Ti, V, and O atoms in the NH4VO3@VTN sample (Figure e–1h and Figure S3).[36,37]

Figure 1

(a) SEM image of NH4VO3@VTN. The insets give the particle size distribution (top right) and enlarged SEM image of NH4VO3@VTN nanospheres (bottom right). (b) Representative TEM image of NH4VO3@VTN and the corresponding SAED pattern (inset). (c) Magnified HRTEM image of NH4VO3@VTN. (d) Crystal plane indexing of NH4VO3@VTN. The inset shows the crystallite shape. (e) STEM image and (f–h) corresponding elemental mappings of Ti, V, and O in an NH4VO3@VTN sample.

Then nitrogen adsorption–desorption experiments for the various meso-VTN samples were performed (Figure a). All of the materials showed a typical type IV isotherm with an H1-type hysteresis loop, suggesting the presence of mesopores.[38]Table summarizes the textural parameters. The BET area and pore volume of NH4VO3@VTN were as high as 112 m2 g–1 and 0.33 cm3 g–1, respectively, comparable with those of pristine TiO2. When other vanadium precursors are used, the capillary condensation step becomes less steep and shifts to high relative pressure, indicating a gradual increase in pore size.[39] This trend agrees well with the results of pore size distribution (Figure b), where NH4VO3@VTN showed a sharp peak indicating uniform pore size, while the peaks of VOSO4@VTN, VCl3@VTN, and VO(acac)2@VTN became broad. This difference indicates the key role of vanadium precursors in the formation of meso-VTN. The increase in pore size can be attributed to the different strengths of electrostatic interaction between Ti and V precursors during the self-assembly process.[40,41]

Figure 2

Nitrogen adsorption–desorption isotherms (a) and BJH pore size distribution curves (b) of pure TiO2 and meso-VTN materials prepared from different V precursors.

Table 2

Textural Parameters of Meso-VTN Materials

sample	vanadium loading (wt %)a	anatase:rutileb	SBET (m2 g–1)c	pore volume, Vp (cm3 g–1)c	.
mesoporous TiO2		>100:1	121	0.2	7.1
NH4VO3@VTN	0.59	95:5	112	0.33	7.9
VOSO4@VTN	0.46	90:10	93	0.27	8.9
VO(acac)2@VTN	0.5	81:19	87	0.31	10
VCl3@VTN	0.6	41:59	82	0.23	8.8

Determined by ICP analysis.

The weight percentage of the rutile phase was calculated using the formula WR = 1/[1 + 0.884(AA/AR)], where AA and AR represent the XRD integrated intensities of anatase (101) and rutile (110) diffraction peaks.

Calculated on the basis of N2 sorption at 77 K.

Calculated from the BJH method.

X-ray diffraction (XRD) can verify the influence of V doping on the crystalline structure of meso-VTN. As shown in Figure a, no vanadia peaks were detected, suggesting a uniform distribution of vanadium in the VTN crystals, in agreement with the elemental mapping shown in Figure . Notably, the (101) diffraction peak shifts slightly toward a higher angle after the introduction of vanadium species (Figure b). This implies that the vanadium ions are incorporated in the titania lattice.[42] For NH4VO3@VTN, the XRD pattern showed the characteristic anatase titania peaks, with a trace amount of rutile. However, with other vanadium compounds as precursors, the intensity of the rutile diffraction peaks increased while the anatase fraction decreased. Notably, VCl3@VTN showed a rutile-dominant diffraction pattern. The anatase:rutile ratios, calculated on the basis of the integrated intensities of the anatase (101) peak and rutile (110) peak,[43] are shown in Table . The anatase content in meso-VTN decreases as follows: NH4VO3 > VOSO4 > VO(acac)2 > VCl3. This confirms that the vanadium-induced phase transition of meso-VTN from anatase to rutile depends strongly on the vanadium precursor. Figure c shows the Raman spectra of all V-doped VTN samples. The Raman characterization also upholds the XRD results. All of the samples showed a set of bands at 146 (Eg), 198 (Eg), 396 (B1g), 516 (A1g + B1g) and 638 (Eg) cm–1, which are assigned to the fundamental active modes of anatase TiO2.[28,44] No typical Raman peaks of VO species were detected, indicating the absence of crystalline vanadium oxides, also supporting the complete doping of vanadium into the anatase lattice.[45] In addition, all of the peaks of VTN samples became asymmetric in comparison to pure TiO2 (Figure d). This holds especially for VO(acac)2 and VCl3@VTN, where two weak bands at 445 and 610 cm–1 were observed, which can be assigned to the rutile TiO2 features.[43] Among those vanadium precursors, only NH4VO3 has the VO3– anion, which could provide a strong electrostatic interaction with the Ti4+ cations, showing a more anatase dominant structure. Note that both anatase and rutile titania are made up of TiO6 octahedra in tetragonal configurations.[46] The structural difference between them is caused by the stacking arrangement of these TiO6 units. The metastable anatase structure consists of edge-sharing TiO6 octahedra in zigzag stacking, while the stable rutile consists of both corner and edge-sharing TiO6 in the linear stacking.[47,48] Titanium oligomers can interact with vanadium precursors via electrostatic interaction during the self-assembly process.[49] The substitution of the lattice Ti by V ions distorts the anatase structure, thus promoting the transformation from anatase to rutile.

Figure 3

X-ray diffraction patterns (a, b) and Raman spectra (c, d) of blank-TiO2 and VTN materials prepared from different vanadium precursors. Due to the low loadings, VO species were not observed.

Temperature-programmed reduction studies showed that NH4VO3 forms monomeric VO4 units (Figure ). After vanadium was introduced into the titania lattice, the maximum hydrogen consumption (Tmax) shifted to lower temperature, in comparison with pure TiO2 and V2O5. Generally, the nature of the VO species affects the reducibility of V-doped TiO2. Monomeric VO4 species are more easily reduced.[50] Among the VTN samples, NH4VO3@VTN showed the lowest Tmax value, suggesting that NH4VO3@VTN forms more monomeric vanadium species. This may reflect the strong electrostatic interaction between VO3– and the Ti4+ precursors. We conclude that NH4VO3 is the best precursor, as it provides a strong electrostatic interaction with the Ti4+ cations, thereby suppressing the agglomeration of vanadium species (Scheme ).

Figure 4

H2-TPR profiles of TiO2, V2O5, and VTN materials with different V precursors.

Scheme 1

Illustration of the Synthesis Route of NH4VO3@VTN via Electrostatic Interaction between the VO3– and the Ti4+ Precursors

Factors Governing Activity and Selectivity

We examined the effect of reaction conditions on the aerobic oxidation of ethyl lactate with dioxygen over the NH4VO3@VTN catalyst. As shown in Figure , the selectivity of ethyl pyruvate was high at lower temperatures and decreased gradually with increasing temperature. Ethyl lactate conversion increased with the reaction temperature, with the highest yield to ethyl pyruvate obtained at 130 °C. Increasing the temperature further decreased the yield of ethyl pyruvate. We also plotted the selectivity of major products against ethyl lactate conversion and reaction time over 0.6%V-NH4VO3@VTN at optimized temperature. As shown in Figure a, the ethyl lactate conversion increased with the reaction time, reaching 70% after 8 h, with ethyl pyruvate as the main product together with ethanol, pyruvic acid, and some minor byproducts. The ethyl pyruvate selectivity gradually decreased from 91% to 58% with the reaction time on stream, as the amounts of both ethanol and pyruvic acid increased to 28% and 11%, respectively. Several minor byproducts including acetic acid and acetaldehyde were detected at a steady state with a selectivity of ∼2%. On the basis of these results, we propose a reaction network for the aerobic oxidation of ethyl lactate, involving oxidative dehydrogenation, hydrolysis, and decarboxylation steps (Figure b).

Figure 5

Effect of reaction temperature on the aerobic oxidation of ethyl lactate. Reaction conditions: 0.6%V-NH4VO3@VTN 50 mg, 1 atm O2, 2 h, ethyl lactate 8.5 mmol (1.0 g), diethyl succinate (solvent) 2 mL.

Figure 6

(a) Plots of product selectivity against ethyl lactate conversion and reaction time. Reaction conditions: 0.6%V-NH4VO3@VTN 50 mg, 1 atm O2, temperature 130 °C, ethyl lactate 8.5 mmol (1.0 g), diethyl succinate (solvent) 2 mL. (b) Proposed reaction pathway for the aerobic oxidation of ethyl lactate.

We also studied the effect of vanadium loading on the catalytic activity by modifying the NH4VO3 precursor concentration for a series of NH4VO3@VTN materials. These catalysts are denoted as xV-NH4VO3@VTN, where x represents the weight ratio (wt %) of vanadium. The corresponding porosity analysis and textural parameters are given in Figure S4 and Table S1 in the Supporting Information. Figure a shows the selectivity–conversion curves of these NH4VO3@VTN catalysts. Pure meso-TiO2 (0% V) showed some conversion, but selectivity to pyruvate was low. We think that the basic OH groups on the TiO2 surface can hydrolyze lactate and pyruvate to ethanol and lactic/pyruvic acid. When vanadium was introduced in TiO2, the resulting VTN catalysts showed high ethyl pyruvate selectivity, reaching 95% at low ethyl lactate conversion. At higher ethyl lactate conversions, the selectivity of ethyl pyruvate tended to increase with an increase in the V loading (from 0.2% to 2%V).[9] This indicates that vanadium acts as the active site in lactate-to-pyruvate reaction, and at higher vanadium loadings, the competing hydrolysis is suppressed. To check this, we ran control experiments on the aerobic oxidation of ethyl lactate over 0.2 V%-NH4VO3@VTN, using molecular sieves 3 Å (MS-3A) as a dehydrating agent (Figure S5). Addition of MS-3A gave a much higher selectivity of 80% to ethyl pyruvate (at 51.6% conversion), in comparison to 56% (at 50.2% conversion) in the absence of MS-3A (Figure S5a). Without the dehydrating agent, the amount of byproducts was higher: e.g., a selectivity to ethanol of 22% instead of 10% (Figure S5b). When the V loading was increased to 3%, the difference between 2% and 3% almost leveled up. To better understand this trend, we plotted the ethyl pyruvate yield against reaction time (Figure b), which confirmed that 2%V-NH4VO3@VTN gives the highest yield. Indeed, the crystalline V2O5 can be seen in the XRD of the 3%V NH4VO3@VTN sample, suggesting that some of the vanadium is aggregated as V2O5 (see Figure S6 in the Supporting Information). Thus, increasing the vanadium loading also increases the number of active vanadium dopant sites in the titania lattice, which in turn results in improved overall activity and selectivity.

Figure 7

Selectivity to ethyl pyruvate plotted against conversion (a) of NH4VO3@VTN on varying the vanadium loading from 0 to 3 wt % and (b) the corresponding time-resolved yield profile of ethyl pyruvate. Reaction conditions: catalyst 50 mg, 1 atm O2, 130 °C, ethyl lactate 8.5 mmol (1.0 g), diethyl succinate (solvent) 2 mL.

Further insight into the reaction can be gained by analyzing structure–activity relationships. Interestingly, we found that the TOF of NH4VO3@VTN was inversely proportional to the vanadium loading (Figure a), showing a close correlation with the ratio V4+/ (V4+ + V5+) (derived from the V 2p spectrum in Table ). In particular, the amounts of exposed V4+ ions are correlated with catalytic performance. The similar ionic radii of Ti4+ (0.61 Å) and V4+ (0.58 Å) enable the doping of the latter into the titania lattice (V5+ is much smaller, 0.54 Å).[45] To understand this behavior at the atomic level, we built a simple model of anatase titania (Figure b) and performed density functional theory (DFT) calculations of the substitution of V4+ and V5+ for Ti4+ in the anatase lattice (a more detailed description of the simulations is provided in the Supporting Information). We found that exchanging the surface Ti with V4+ on anatase TiO2 corresponds to a lower energy in comparison to that of V5+, in line with previous reports.[51]

Figure 8

(a) Relationship among turnover frequency (TOF) for ethyl pyruvate formation, V4+/(V4+ + V5+) (atom %), and loading of vanadium catalysts. Reaction conditions: NH4VO3@VTN 50 mg, 1 atm O2, 1 h, ethyl lactate 8.5 mmol (1.0 g), diethyl succinate (solvent) 2 mL. (b) Model of Anataste VTN. Atom colors: Ti, blue; O, red; V, yellow.

Table 3

Surface Atom Ratios of NH4VO3@VTN Calculated from XPS Analysis

sample	V5+ (atom %)	V4+ (atom %)	OI (atom %)	OII (atom %)	OIII (atom %)	V4+/(V4+ + V5+)	OII/(OI + OII + OIII)
meso-TiO2			52.7	6	2.2		0.10
0.6%V-NH4VO3@VTN	1.13	1.32	38.05	17.1	6.83	0.54	0.28
1%V-NH4VO3@VTN	1.77	1.12	33.43	15.14	7.51	0.39	0.27
1.4%V-NH4VO3@VTN	2.52	1.35	39.52	14.37	4.08	0.34	0.25
2%V-NH4VO3@VTN	4.07	1.57	45.38	13.32	2.48	0.28	0.22

X-ray photoelectron spectroscopy (XPS) measurements show that the Ti 2p peaks of NH4VO3@VTN containing 0.6% V and 1% V are asymmetric, in comparison with pure TiO2 (Figure A, a–c). This indicates the presence of two types of Ti on the surface. The binding energy corresponding to Ti4+ (Ti 2p1/2, 464.8 eV; Ti 2p3/2, 459.1 eV) is shifted toward higher values. These results imply that Ti is replaced with V, decreasing the electron charge density of Ti4+.[52] This means that the neighboring lattice Ti (denoted as Ti(V)) of the vanadium dopant gives a high binding energy. For comparison, we prepared 0.6% VO/TiO2 using impregnation, where the Ti(V) peak was not detected in XPS spectra (Figure S7). This further confirmed the substitution of lattice Ti by vanadium. As the vanadium loading increases, the deconvoluted peaks corresponding to Ti(V) gradually disappear (Figure A, d and e). When more vanadium is incorporated in the TiO2 lattice, isolated V4+ will be gradually converted to polymerized VO, thus lowering the fraction of Ti(V).[53] This agrees also with the decrease in the ratio V4+/(V4+ + V5+) (Figure B and Table ). When the V loading is increased to 3%, the peaks of crystalline V2O5 can be seen in the XRD pattern (Figure S6). As shown in Figure C, the spectra of O 1s in NH4VO3@VTN can be deconvoluted into three peaks: the peak at 530.3 eV was ascribed to the O2– lattice oxygen (OI), and the peaks at ∼531.6 and ∼532.3 eV are ascribed to O22– and O– surface chemisorbed oxygen (labeled as OII and OIII).[54] The OII oxygens, which are located at the surface defects, are more active and therefore more easily reduced.[55,56] As vanadium ions with different valences replace Ti4+ ions in the lattice, the charge imbalance generates structural defects and additional oxygen vacancies.[57,58] This increases the number of mobile OII species.[59] In our case, the V4+/ (V4++V5+) ratio decreased from 0.54 to 0.28 when V loading is increased, and the OII/(OI + OII + OIII) fraction decreased simultaneously.

Figure 9

XPS studies showing high-resolution Ti 2p spectra (A), high-resolution V 2p spectra (B), and high-resolution O 1s spectra (C) of the NH4VO3@VTN with different V loadings: (a) 0% V (meso-TiO2); (b) 0.6% V; (c) 1% V; (d) 1.4% V; (e) 2% V.

Leaching of active species into the solution is a known problem in heterogeneous catalysis. To rule out the possibility of vanadium leaching, we ran a hot filtration experiment (Figure S8).[60] When the NH4VO3@VTN catalyst was filtered from the reaction mixture after 3 h, no further ethyl lactate conversion was observed. Moreover, the vanadium content of the filtrate was below the detection limit of ICP-AES analysis. For comparison, we prepared vanadia supported on titania with an identical surface vanadium content using impregnation (herein VO/TiO2; the vanadium content on the surface was determined by XPS). Mesoporous anatase TiO2 was impregnated with aqueous solutions of NH4VO3 and oxalic acid, followed by drying and calcining for 4 h at 500 °C. Control experiments showed that the filtrate of VO/TiO2 was still active in ethyl lactate oxidation (Figure S9), indicating that V species do leach into solution in the case of impregnation. We also determined the recyclability of NH4VO3@VTN. In each run, the catalyst was separated by simple centrifugation and then dispersed in water under ultrasonication for 1 h. As shown in Figure , the NH4VO3@VTN could be reused 10 times without significant loss of activity. The XRD and TEM analysis confirmed that the meso structure and anatase/rutile ratio were well preserved after recycling (Figure S10).

Figure 10

Kinetic plots of recycling tests of NH3VO4@VTN in the oxidation of ethyl lactate to ethyl pyruvate. Reaction conditions: 100 mg catalyst, 130 °C, 1 atm O2, 4 h, ethyl lactate 8.5 mmol (1 g), diethyl succinate (solvent) 2 mL.

Mechanistic Considerations

To study the reaction mechanism of lactate to pyruvate on the NH3VO4@VTN surface, we did in situ DRIFTS studies. Figure shows the results for the aerobic oxidation of ethyl lactate over 0.2%V-NH3VO4@VTN at 130 °C at different reaction times. Typical vibrational bands of ethyl lactate on NH3VO4@VTN were observed after ethyl lactate adsorption. The positive bands at 2987, 2943, and 2885 cm–1 were ascribed to νas(CH3), νas(CH2), and νs(CH3), respectively.[61,62] The corresponding C–H bending vibrations were also detected (δ(CH2) 1473 cm–1, δ(CH3) 1454 cm–1, and δ(CH3) 1324 cm–1).[61,63] The lactate carboxyl stretching vibrations were observed at 1562 cm–1 (νs(COO)) and 1425 cm–1 (νas(COO)), while the peak at 1740 cm–1 belongs to the lactate carbonyl ν(C=O).[61] Two additional peaks appeared at 1678 and 1658 cm–1, probably due to the C=O bonds coordinated with the VTN surface.[64] After 20 min, we observed two new shoulder peaks at 1867 and 1780 cm–1, which are related to the carbonyl stretching of the α-keto group of the pyruvate.[65] The intensity of pyruvate bands increased with the reaction time on stream; meanwhile, νs(COO) at 1562 cm–1 gradually shifted to 1590 cm–1, suggesting the formation of more ethyl pyruvate. Moreover, two weak bands at 1130 and 1217 cm–1 can be ascribed to the hydroxyl-related C–O vibrations of ethyl lactate (νlactate(C–O) and δlactate(C–O)), reflecting the OH deprotonation of ethyl lactate on VTN surface. The signals of νlactate(C–O) and δlactate(C–O) reached a maximum in the first 20 min, indicating that the lactate OH deprotonation plays a key role in this reaction. This is also confirmed by the new broad band at ∼3250 cm–1, which indicates the formation of adsorbed water on VTN surface.[65] Note that the V=O overtone band (∼2040 cm–1)[63] is well-preserved during the aerobic oxidation of ethyl lactate, while the V–O related bond (δ(V4+–O–Ti) ∼1370 cm–1)[66] is diminished in intensity. This indicates that the terminal V=O bonds are not involved in the ethyl lactate conversion, confirming the vital role of V4+–O related bonds.

Figure 11

In situ DRIFT spectra recorded during aerobic oxidation of ethyl lactate with air over 0.2%-NH3VO4@VTN catalyst at different time intervals (0, 20, 40, ..., 360 min) at 130 °C.

To complement the experimental results and the reaction pathways for the oxidative dehydrogenation of ethyl lactate to ethyl pyruvate, we built a periodic model of VTN by replacing lattice Ti with V4+ in the top layer of the (101) facet of anatase titania. This model was simulated using density functional theory (DFT) calculations as implemented in the CP2K package (see the Supporting Information for details). First, we studied the adsorption and dissociation of ethyl lactate (EL) on the VTN surface. Several possible adsorbed forms were considered (see Figure S11). Geometry optimization showed that EL can be stabilized on either surface titanium or vanadium atoms (Figure ). The hydroxyl oxygen coordinates to surface titanium atoms with a Ti···OH–R distance of 2.085 Å, in comparison with a V···OH–R distance of 2.126 Å. In both cases, the O–H bonds were elongated, from 0.974 Å to 1.039 and 1.025 Å, respectively. The adsorption energy of EL on Ti was higher than that of EL on V by 0.73 kcal/mol, in agreement with previous reports.[19]

Figure 12

Optimized geometries of two different models for ethyl lactate chemisorption on the VTN (101) surface: (a) Ti-type adsorption geometry, where ethyl lactate interacts with the Ti atom; (b) V-type adsorption geometry, where ethyl lactate interacts with the V atom. Atom colors: Ti, blue; O, red; V, yellow.

Then, activation of the hydroxyl group triggers the dissociation of EL via proton transfer to a nearby bridging oxygen atom (Obr). To simplify this process, we focused on two types of bridges: V4+–O–Ti (Obr-V) and Ti–O–Ti (Obr-Ti). Figure shows the geometries of two possible intermediates, (a) and (b), in this step. In both cases, the protonated EL is bonded to a titanium atom. Yet while in (a) the hydroxyl H transfers to a nearby V4+–O–Ti site, forming a H–Obr-V bond, in (b) the proton transfers to a Ti–O–Ti to form a H–Obr-Ti bond. The intermediate energies for the formation of (a) and (b) are −75.6 and −136 kcal/mol, respectively. This suggests that the V4+–O–Ti bonds are responsible for the dissociation of ethyl lactate over VTN catalysts, supporting the experimental result. In addition, the β-hydrogen of the protonated EL can interact with the V4+–O–Ti oxygen, increasing the C–Hβ bond length from 1.099 to 1.118 Å.

Figure 13

Representative structures of the two intermediates during proton transfer process: (a) hydrogen atom transfer to bridging oxygen atoms in V–O–Ti sites; (b) hydrogen atom transfer to bridging oxygen atoms in Ti–O–Ti sites.

Therefore, the simulations show us that (i) ethyl lactate adsorbs preferentially on the VTN surface via the coordination of hydroxyl oxygen to titanium, (ii) V4+–O–Ti bonds play a vital role in the dissociation of ethyl lactate, and (iii) the deprotonation of β-hydrogen is also activated by V4+–O–Ti bonds.

For most oxidative dehydrogenation reactions, the molecular oxygen activation involves either free radical or Mars–van Krevelen pathways.[67] To determine if any radical species is involved in this reaction, we introduced a number of free radical scavengers to the reaction mixture (BHT, p-benzoquinone, and tert-butyl alcohol; see Table S2). These free radical scavengers did not suppress the reaction completely, ruling out the formation of free radical intermediates (i.e., superoxide radical O2•–) in the bulk reaction mixture. To further study the role of molecular oxygen, we ran control experiments over NH4VO3@VTN, where the molecular oxygen was replaced by nitrogen. The reaction was limited without oxygen.[55,56] Thus, we can conclude that the aerobic oxidation follows a Mars–van Krevelen mechanism for VTN catalysts, wherein molecular oxygen can replenish the V4+–O–Ti bonds. A trace amount of ethyl lactate was converted to ethyl pyruvate in nitrogen initially, probably due to the oxygen still chemisorbed on the surface of VTN. This anaerobic process of lactate to pyruvate was also observed by in situ DRIFTS experiments at elevated temperatures under a helium atmosphere. The signals for the V4+–O–Ti bonds gradually decreased and shifted to higher wavelengths, accompanied by the formation of the characteristic peaks of pyruvate (Figure S12).

On the basis of the experimental and computational results, we can propose a probable mechanism (Scheme ). Starting with the pristine catalyst VTN, ethyl lactate adsorbs and then chemisorbs by the coordinated bond to titanium atoms a and further forms the transition state b. Then, the hydroxyl H atom transfers to near V4+–O–Ti sites, forming the Ti4+–O–substrate intermediate and a V4+–O–H bond (c). This is followed by β-hydrogen activation through an interaction with an adjacent V4+–O–Ti oxygen, giving the five-membered intermediate d. Ethyl pyruvate is then produced by β-hydrogen elimination, giving water as a byproduct as well as creating an oxygen vacancy (e, f). Finally, adsorbed oxygen replenishes the oxygen vacancy, regenerating the V4+–O–Ti bonds. In general, structural defects containing lower oxidation state cations are reported to be necessary to activate the oxygen. Here, we anticipate that V4+, because of its lower oxidation state, can activate oxygen better than V5+. This could be also connected with the fact that a linear relationship is observed between TOF and the amount of V4+ in the samples.

Scheme 2

Proposed Catalytic Cycle for the Oxidative Dehydrogenation of Ethyl Lactate to Ethyl Pyruvate in the Presence of Meso-VTN

Conclusions

We developed a one-pot strategy for the controllable synthesis of uniform mesoporous vanadia–titania nanoparticles (VTNs). NH4VO3 is the optimal vanadium precursor, forming primarily monomeric VO4 units, avoiding the agglomeration of polymeric vanadium species as well as the formation of a rutile structure to a great extent. In comparison to homogeneous vanadium compounds and the classical VO-TiO2 prepared by impregnation, the meso-VTN catalysts showed superior catalytic activity and selectivity for the ODH of ethyl lactate to ethyl pyruvate. In addition, V4+ as the major valence state was incorporated into the anatase TiO2, which can increase the availability of surface chemisorbed oxygen, resulting in high catalytic activity for aerobic oxidation of ethyl lactate. Hot filtration and recyclability tests confirmed that NH4VO3@VTN does not leach into solution and can be reused at least 10 times without loss of activity. In situ DRIFTS and DFT simulations show that ethyl lactate adsorbs preferentially on the VTN surface via the coordination of hydroxyl oxygen to titanium. The V4+–O–Ti bonds play a key role in the dissociation of ethyl lactate and further promote the deprotonation of β-hydrogen. In addition, molecular oxygen can replenish the surface oxygen to regenerate the V4+–O–Ti bonds. Thus, this work provides fundamental insights for developing further simple and cost-effective catalytic systems for highly efficient conversion of biomass derivatives to value-added chemicals under mild conditions.