Selective oxidative dehydrogenation of propane to propene using boron nitride catalysts

See allHide authors and affiliations

Science  23 Dec 2016:
Vol. 354, Issue 6319, pp. 1570-1573
DOI: 10.1126/science.aaf7885

Boron nitride catalysis

Propene is one of the highest-volume organic chemicals produced. Propene has mainly been made from naphtha, but changes in the global supply chain are creating shortages. Direct conversion from propane, a component of natural gas, via reaction with oxygen is an attractive alternative, but existing approaches produce a large fraction of unwanted CO and CO2. Grant et al. report that boron nitride, normally an unreactive material, has high selectivity to catalyze the production of propene (77%) and ethene (13%).

Science, this issue p. 1570


The exothermic oxidative dehydrogenation of propane reaction to generate propene has the potential to be a game-changing technology in the chemical industry. However, even after decades of research, selectivity to propene remains too low to be commercially attractive because of overoxidation of propene to thermodynamically favored CO2. Here, we report that hexagonal boron nitride and boron nitride nanotubes exhibit unique and hitherto unanticipated catalytic properties, resulting in great selectivity to olefins. As an example, at 14% propane conversion, we obtain selectivity of 79% propene and 12% ethene, another desired alkene. Based on catalytic experiments, spectroscopic insights, and ab initio modeling, we put forward a mechanistic hypothesis in which oxygen-terminated armchair boron nitride edges are proposed to be the catalytic active sites.

Selective oxidation technology plays a pivotal role in the modern chemical industry (1). The scientific challenge for partial oxidations is the prevention of consecutive overoxidation of the desired product, which is often more reactive than the parent substrate. The oxidative dehydrogenation of propane (ODHP) is a contemporary example of such a challenging reaction. Although propene is conventionally produced via steam cracking of large hydrocarbons in naphtha, the more recent trend to use shale gas as feeds to steam cracker units greatly increases the availability of low-cost ethene but also results in a gap between demand and supply of propene, as well as other higher olefins (2, 3). This disparity motivates the exploration of “on-purpose” propene technologies. Even though nonoxidative propane dehydrogenation is the emerging technology used today, ODHP has the potential to improve reaction efficiency over its nonoxidative counterpart because of favorable thermodynamics (exothermic and lower reaction temperatures) and enhanced catalyst stability (prevention of coke deposition on the catalyst surface). It is estimated that the potential energy savings for moving to ODHP would be ~45% of the energy consumption (4, 5). This gain in reaction efficiency, coupled with the facts that demand for propene is around 100,000 kt and its production consumes around 1 quad of energy, illustrates the potential impact of ODHP if it could be implemented on a large scale. The key scientific challenge that must be overcome remains, however, the prevention of the facile overoxidation of propene product into more thermodynamically stable CO and CO2 (COx).

To date, supported vanadium oxide catalysts show the most promising activity for ODHP, owing to the favorable redox properties of the active vanadium sites (6, 7). However, even after decades of research, propene selectivity remains too low, even at moderate propane conversion. As an example, at 10% propane conversion, the propene selectivity typically drops to less than 60% for such conventional catalysts. The lack of kinetic control identifies the need for the discovery of alternative materials with the ability to better control this partial oxidation (8).

Here, we present both hexagonal boron nitride (h-BN) and boron nitride nanotubes (BNNTs) as metal-free materials able to catalyze the ODHP reaction. Although graphene and fullerene materials are emerging as catalysts for partial alkane oxidations (911), BN materials, one of the “inorganic analogs” of graphene, have yet to be explored for their own catalytic activity. A supported vanadia on silica catalyst (V/SiO2) was used in this work to make direct comparisons with the catalytic performance of BN. These materials were loaded into a quartz tube reactor heated between 460 and 500°C under flowing propane, oxygen, and nitrogen as an inert carrier gas. Reaction parameters—such as temperature, catalyst mass, total gas flow-rate, and partial pressures of propane (Pc3h8) and oxygen (Po2)—were varied to observe changes to product distributions by sampling the reactor exhaust stream via online gas chromatography and mass spectrometry. Gas contact time with the catalyst is represented in this work as the inverse weight-hour-space-velocity {WHSV–1 [kg-catalyst s (mol C3H8)−1}, which was varied primarily by altering the total gas flow rate.

Use of BN materials results in extraordinary selectivity to propene, among the highest reported under ODHP conditions. For instance, h-BN afforded 79% selectivity to propene at 14% propane conversion (Fig. 1A). Meanwhile, the traditional V/SiO2 allows for a modest 61% propene selectivity at only 9% propane conversion (12). The selectivities obtained by using state-of-the-art ODHP catalysts (11, 1319) are compared in Fig. 1A. The decrease in propene selectivity with increasing propane conversion is indicative of the facile overoxidation of propene to COx. Comparisons between key reaction parameters of the referenced catalysts are included in table S1.

Fig. 1 Selectivity to propene and comparisons of product selectivity.

(A) Selectivity to propene plotted against propane conversion for ODHP, comparing previously reported data from representative catalysts to hexagonal boron nitride (h-BN) and boron nitride nanotubes (BNNTs). Open shapes indicate data from other works, cited within the figure. (B) Comparisons of product selectivity between V/SiO2 (Xc3h8 = 5.8%); h-BN (Xc3h8 = 5.4%); and BNNTs (Xc3h8 = 6.5%). Product selectivity is represented by colored bars. (C) Comparisons of propene productivity plotted as a function of C3H8 conversion between V/SiO2, h-BN, and BNNT. V/SiO2: 5 to 15 kg-cat s mol C3H8–1; h-BN: 15 to 40; BNNT: 2 to 5; T = 490°C, Po2 = 0.15 atm, Pc3h8 = 0.3 atm.

The entire product distribution further distinguishes BN materials from supported vanadia catalysts (Fig. 1B). When the supported vanadia catalyst is used, the main by-products are COx, which account for 33% of total product selectivity at 9% propane conversion. Conversely, when BN materials are used, the main by-product is ethene, a highly valuable olefin itself, rather than COx. The combined propene and ethene selectivity is 91% at 14% propane conversion using h-BN (fig. S1). We furthermore verified that the catalytic activity of the BN material remains stable for at least 32 hours on stream (fig. S2), validating the catalyst stability. Transmission electron microscopy images of the h-BN material before and after the ODHP reaction did not reveal any changes to the morphology of the material (fig. S3). Carbon balances for all reactions close within >98%.

The analogous product distributions for both h-BN and BNNTs suggest a similar reaction mechanism for these BN materials. However, BNNTs exhibit a rate of propane consumption (mol C3H8 kg-cat–1 s–1) more than one order of magnitude as high as that observed with h-BN (fig. S4). The higher activity of BNNTs at least partially reflects the higher surface area of BNNTs relative to h-BN (BNNT, 97 ± 5 m2 g–1 versus h-BN, 16 ± 1 m2 g–1) (20); however, the rate of propane consumption is more than three times as high with BNNT as with h-BN when normalized for surface area (BNNT: 3.6 ×10–7 mol C3H8 s–1 m–2 versus h-BN: 1.1 × 10–7 mol C3H8 s–1 m–2). This high reactivity and selectivity with BNNTs results in a substantial enhancement in the observed propene productivity (kg-C3H6 kg-cat–1 hr–1) (Fig. 1C), comparable to values deemed attractive for commercial implementation of this on-purpose propene technology (15, 21).

Further kinetic insights were obtained by studying the influence of reactant concentrations [partial pressure of O2 and C3H8 (Po2 and Pc3h8, respectively)] on the reaction rate. The inclusion of oxygen as a reactant is required for propane conversion using BN materials. The rate of propane consumption using h-BN indicates oxygen activation on the BN surface (Fig. 2A) and second-order dependence with respect to Pc3h8 (Fig. 2B). This kinetic behavior clearly distinguishes BN from traditional supported vanadia catalysts, which follow a Mars– van Krevelen mechanism (rate-determining substrate oxidation followed by fast reoxidation of the surface by oxygen) that typically leads to zero-order rate dependence with respect to Po2 and first-order rate dependence in Pc3h8 (22).

Fig. 2 Rates of propane consumption.

Rates of propane consumption using h-BN as a function of (A) Po2 (Pc3h8 constant at 0.3 atm) and (B) Pc3h8 (Po2 constant at 0.2 atm) fit with Eley-Rideal kinetics showing O2 adsorption and second-order dependence with respect to Pc3h8. Solid lines are a least-squares fit that takes into account all experimental data points at each respective temperature by using the displayed rate law.

It is surprising that BN, a material known for its high stability under oxidative conditions (23, 24), is catalytically active at all. So far it has been explored for its unique electronic, thermoelectric, and mechanical properties (2528). The combination of the interesting observations outlined in this work (i.e., improved selectivity to olefins and different reaction kinetics) points toward a fundamentally different reaction mechanism in conjunction with a different active site compared with other, previously studied and better understood catalysts for this reaction. Indeed, BN samples from various suppliers and containing different impurities (fig. S5) show almost identical catalytic performance (figs. S6 and S7), which indicates that said metal impurities in the material are unlikely to play a significant role.

To better understand the catalytically active parts of the material and to provide evidence for oxygen functionalization of the BN surface when exposed to ODHP reaction conditions, we used a combination of x-ray photoelectron spectroscopy (XPS), attenuated total reflectance–infrared (ATR-IR), and diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) measurements to characterize the material before (“fresh”) and after (“spent”) exposure to the ODHP reaction mixture. As a first step, we used XPS to monitor the surface concentrations of boron, nitrogen, and oxygen of h-BN (table S1) and observed an increase in the amount of surface oxygen in spent h-BN material, signaling oxygen functionalization. The same increase in surface oxygen content was not observed for h-BN exposed only to air at 490°C; this indicated that the hydrocarbon must be present for oxygen to functionalize the surface of h-BN. This observation was corroborated with ATR-IR measurements, which showed the emergence of a broad feature around 3200 cm–1 and a sharp signal at 1190 cm–1 for spent h-BN (fig. S8A), assigned to OH-stretching and B–O stretching vibrations, respectively (29, 30). These spectroscopic features were absent with fresh h-BN, as well as h-BN treated only under air at 490°C. Using DRIFTS, we saw more-resolved features appear at 3420 and 3250 cm–1 with spent h-BN (fig. S8B). BNNTs show a similar behavior, only with considerable intensity in the given spectral range already present for the fresh samples and a corresponding increased intensity for the spent materials (fig. S9).

The appearance and change of the different spectroscopic features furthermore indicates that a site is created by exposure of the material to propane and oxygen and that, in line with the higher intensity of this feature, a larger number of these sites are present in the BNNT sample, and the exposure to both propane and oxygen leads to the creation of those sites. To elucidate the origin of these vibrations, we performed ab initio density functional theory (DFT) calculations using the Vienna ab initio simulation package (VASP) (31, 32). All computational details are given in the supplementary materials. We focused on a single BN sheet and calculated vibrational frequencies for a set of OH-terminated zigzag and armchair edges. All structures and corresponding vibrational frequencies >3000 cm–1 are given in fig. S10. When comparing the modeled wavenumbers to the experimental measurements, we find that especially the armchair edge leads to features in the studied range, the feature around 3420 cm–1 corresponding to a single OH-stretch (displayed in movie S1) and the feature around 3250 cm–1 associated with a concerted stretching vibration (displayed in movie S2).

From the combination of catalytic activity, spectroscopic data, and DFT calculations and in line with the observed oxygen-dependence of the kinetics, we therefore propose that an oxygen molecule bonded to one B and one N [an oxygen-terminated armchair edge of BN (>B–O–O–N<)] acts as active site for the ODHP reaction. This site shows similarities to work in the semiconductor literature focusing on oxygen-terminated armchair edges of BN (28), as well as the proposed active sites of graphene and fullerene materials for similar oxidations (10, 11). These >B–O–O–N< sites can be viewed as inorganic peroxide species able to perform oxidation reactions. Starting from our modeled active site (namely, >B–O–O–N<), we then calculated key intermediates for the ODHP reaction. We suggest that the dehydrogenation initiates by the abstraction of a hydrogen atom from a secondary carbon of propane by the >B–O–O–N< sites, breaking the O–O bond while forming a B-OH species and one nitroxyl radical (Fig. 3C). This reaction is reminiscent of an alkyl hydroperoxide reaction proposed earlier to explain the formation of radicals in homogeneous autoxidation systems (RO–OH + H–R′ → RO + HOH + R′) (33). However, in the case of the surface-catalyzed reaction, we anticipate a rebound step where the secondary propyl radical reacts rapidly with the nitroxyl radical, forming a surface-stabilized propyl intermediate [namely, >N–O–CH(CH3)2] (Fig. 3D).

Fig. 3 Proposed radical rebound mechanism for hydrogen atom abstraction from propane to the O-terminated armchair BN edge site.

(A to D) Our model suggests that this O-terminated edge site (one-dimensional, 1D) may lead to the observed favorable olefin selectivity. This is drastically different from a 0D site (e.g., V/SiO2), which most likely creates a reactive propyl radical (34), or a 2D site [e.g., Cu(111) surface] where diffusion of adsorbates is unrestricted (35), both prone to overoxidation.

We consider that the observed high olefin selectivity may largely be due to the stabilization of the propyl radical by the nitroxyl radical site. Indeed, the one-dimensional (1D) nature of this edge avoids the creation of a highly reactive propyl radical (typical of 0D single-site catalysts) (34), and also the overoxidation of the adsorbed species (typical of 2D surface catalysts) is suppressed (35). The proposed intermediate structures and energies of the overall catalytic cycle are included in fig. S11. We envision that a second abstraction of a hydrogen atom from a primary carbon follows another radical rebound, and creates a di-propoxyl intermediate. Desorption of propene and the reorganization of hydrogen atoms along the edge form water as a side product. The desorption of water is followed by oxygen addition to regenerate the >B–O–O–N< active site. Similar surface reorganization of hydroxyl groups in the presence of oxygen was proposed for related carbon nanofilament catalysts for oxidative dehydrogenation of ethylbenzene to form styrene (36). All steps in this process, except for desorption of propene, are exothermic. The second-order rate dependence with respect to Pc3h8 suggests that two propane molecules are required to generate two molecules of water, in line with the overall stoichiometry of the reaction. The desorption of these water molecules forms BN edge vacancies that allow for unique O2 activation, which explains the influence that the surface coverage of adsorbed oxygen has on the rate of propane consumption.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S14

Tables S1 and S2

Movies S1 and S2

References (3740)

References and Notes

  1. Acknowledgments: We thank the Wisconsin Alumni Research Foundation (WARF) for funding through the WARF Accelerator Program. We thank S. Stahl and J. Dumesic for their helpful feedback when reviewing this manuscript. J.T.G., C.A.C, J.V., A.C., and I.H. are inventors on patent application U.S. 15/260,649, submitted by the WARF, that covers BN as catalysts for ODHP and other related reactions. All data are reported in the main paper and supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article