Unraveling Radical and Oxygenate Routes in the Oxidative Dehydrogenation of Propane over Boron Nitride

Oxidative dehydrogenation of propane (ODHP) is an emerging technology to meet the global propylene demand with boron nitride (BN) catalysts likely to play a pivotal role. It is widely accepted that gas-phase chemistry plays a fundamental role in the BN-catalyzed ODHP. However, the mechanism remains elusive because short-lived intermediates are difficult to capture. We detect short-lived free radicals (CH3•, C3H5•) and reactive oxygenates, C2–4 ketenes and C2–3 enols, in ODHP over BN by operando synchrotron photoelectron photoion coincidence spectroscopy. In addition to a surface-catalyzed channel, we identify a gas-phase H-acceptor radical- and H-donor oxygenate-driven route, leading to olefin production. In this route, partially oxidized enols propagate into the gas phase, followed by dehydrogenation (and methylation) to form ketenes and finally yield olefins by decarbonylation. Quantum chemical calculations predict the >BO dangling site to be the source of free radicals in the process. More importantly, the easy desorption of oxygenates from the catalyst surface is key to prevent deep oxidation to CO2.


■ INTRODUCTION
Propylene (C 3 H 6 ) is a crucial platform chemical in the petrochemical industry for organic and polymer synthesis. It is mainly produced by petroleum-derived steam cracking and fluid catalytic cracking. 1−3 The annual C 3 H 6 production was 130 Mt in 2019 and is projected to grow to 191 Mt by 2030. 4 The recent shift from petroleum-derived naphtha to shale gas feedstock greatly increases the availability of ethylene (C 2 H 4 ) and results in a gap between the supply of C 3 H 6 and rising global demand. 5,6 To alleviate the "propylene gap", the nonoxidative dehydrogenation of propane (C 3 H 8 ) was industrialized recently by Honeywell UOP (Oleflex) and ABB Lumus (Catofin). 7,8 However, this process suffers from (1) rapid accumulation of coke, which requires frequent catalyst regeneration, and (2) high energy need due to the endothermicity of the process. 9,10 Therefore, the oxidative dehydrogenation of propane (ODHP) represents a promising alternative, with an estimated energy saving of ca. 45% due to its exothermicity, as well as the prevention of coke formation in the presence of O 2 . 11−14 Transition metal oxide catalysts, such as vanadia species (VO x ), are able to activate C−H bonds in C 3 H 8 , boding well for ODHP performance. 15−18 However, the partially occupied d-orbitals in transition metal oxides interact with the reactive intermediates, binding them strongly to the catalyst surface and leading to overoxidation to CO and CO 2 , limiting the selectivity of the process. 11,19,20 In 2016, boron nitride (BN) emerged as a unique ODHP catalyst because of its high selectivity to C 3 H 6 and the prevention of CO x , especially CO 2 , formation. 21 Modified BN 22,23 and further boron-containing catalyst candidates, e.g., boron oxides (BO x ), 24−27 were widely studied to identify the active site. The similar reaction kinetics over BN and BO x were rationalized by surface oxygen functionalization on BN, generating BO x sites as the effective active site in ODHP. 28,29 The conversion was found to increase with higher dilution over boron-based catalysts, which was one of the peculiar kinetic features implying a gas-phase C 3 H 6 formation route. 30 Based on computational results and kinetic measurements, distinct surface-confined and gas-phase mechanisms were, thus, proposed over boron-based catalysts for ODHP. 31−33 So far, the only experimental evidence for the existence of a gas-phase route has been the observation of the methyl radical. 34 The gas-phase reaction mechanism and the possibility of free radical involvement beyond the methyl radical remain unknown, which makes it challenging to rationally optimize catalysts for the practical application of ODHP.
In this contribution, we shed light on the intermediates and products evolving in real time in ODHP over BN by operando synchrotron photoelectron photoion coincidence (PEPICO) spectroscopy. Operando PEPICO detects short-lived reactive intermediates, e.g., radicals, ketenes, and enols, desorbed from the catalyst surface 35,36 and, thus, provides direct experimental evidence to understand the ODHP reaction network over BN. A silica-supported vanadia catalyst (VO x /SiO 2 ) was also studied for direct comparison with the catalytic mechanism over BN. Temperature-dependent ODHP experiments in a tubular continuous-flow reactor coupled to the PEPICO endstation also provide temperature profiles of stable species and reactive intermediates. Surface-confined density functional theory (DFT) calculations on boron active sites and gas-phase G4 calculations reveal the formation mechanism of short-lived radical intermediates detected by operando PEPICO spectroscopy. A reaction mechanism for boron-catalyzed ODHP is proposed, encompassing coupled surface-confined and gasphase reaction steps.
■ RESULTS AND DISCUSSION ODHP as a Function of Temperature over BN and VO x /SiO 2 . The temperature-dependent ODHP performance over commercial BN and homemade VO x /SiO 2 catalysts is compared in Figure 1. Over the BN catalyst, the main products are propene, ethene, and methane at temperatures below 560°C with the propane conversion lower than 21% (Figure 1a). Small amounts of CO and butene isomers (C 4 H 8 ) are also found at 550−560°C. When increasing the temperature so that the C 3 H 8 conversion rises to above 80%, the selectivity to C 3 H 6 decreases with increased formation of CO, CO 2 , and C 4 H 8 . In contrast, the main products over a VO x /SiO 2 catalyst are C 3 H 6 and CO x (Figure 1b), the latter being predominant at high C 3 H 8 conversion. This indicates that overoxidation can be curtailed using the BN catalyst, especially at low C 3 H 8 conversion. As a result, the total selectivity to hydrocarbons (C 3 H 6 , C 2 H 4 , and CH 4 ) over BN is significantly higher than over VO x /SiO 2 at a comparable C 3 H 8 conversion. For example, the selectivities to C 3 H 6 , C 2 H 4 , and CH 4 are 61.6%, 18.7%, and 19.7%, respectively, at a C 3 H 8 conversion of 2.9% over a BN catalyst (Figure 1c), which corresponds to essentially full selectivity to hydrocarbons. Increasing the C 3 H 8 conversion to 9.7%, the selectivity to C 3 H 6 , C 2 H 4 , and CH 4 is still ca. 92%, while 0.8% C 4 H 8 and 8.3% CO are also formed over BN (Figure 1d). This can be compared with only 27.3% selectivity to C 3 H 6 with dominant production of CO and CO 2 already at 5.3% conversion over VO x /SiO 2 ( Figure  1e).
The selectivity to CO 2 is always low over BN, and CO is the main overoxidized product at high temperatures, as opposed to the VO x /SiO 2 catalyst, where CO 2 dominates. Another difference between the BN and VO x /SiO 2 catalysts is that small amounts of C 4 H 8 isomers are only observed over BN. The product distributions over BN and VO x /SiO 2 catalysts ( Figure 1) are consistent with previous results. 34,37 However, the origin of the high selectivity to C 3 H 6 /C 2 H 4 and the inhibition of overoxidation to CO x (especially to CO 2 ) over BN have remained unclear so far and are the subject of this study.
Operando PEPICO Spectroscopy. Operando synchrotron PEPICO spectroscopy 38,39 was utilized to detect stable products and elusive ODHP intermediates over BN and VO x /SiO 2 catalysts. The PEPICO setup is described in Figure  S1. In brief, a gas mixture of C 3 H 8 , O 2 , and Ar is fed into the preheated catalyst bed, and the continuous gas flow, with reactants, intermediates, and final products, expands from the reactor into high vacuum, forming a molecular beam. Molecular beam sampling freezes out the chemistry and suppresses quenching. The molecular beam travels through the skimmer and crosses the monochromatic vacuum ultraviolet beam in the ionization region. Soft photoionization yields photoelectrons and -ions, which are detected in delayed coincidence. Ion mass analysis yields photoionization mass spectra and, combined with electron kinetic energy analysis, allows us to plot photoion mass-selected threshold photoelectron spectra (ms-TPES) to identify the spectral carrier(s) of individual m/z peaks isomer-selectively. 40,41 Mass spectra and ms-TPES are first compared at 600°C based on a temperature-programmed surface reaction (TPSR) of adsorbed propane (C 3 H 8 ) on BN from 500 to 700°C ( Figure  S2). During ODHP operation, the m/z 42 peak dominates over both BN and VO x /SiO 2 , while the C 2 H 4 (m/z 28) and CH 4 (m/z 16) signals are much more intense over BN than over VO x /SiO 2 (Figure 2a Figure  1a as a final product. To the best of our knowledge, this is the first time that oxygenated species such as ketenes are observed as ODHP intermediates, likely because ketene and methylketene are unstable and evade detection by GC/MS even if present in the effluent. 42,43 This illustrates the advantage of operando PEPICO in detecting elusive intermediates to provide mechanistic insights into catalytic mechanisms. 40 The m/z 58 peak can be ascribed to propen-m-ol isomers (m = 1, 2, or 3, C 3 H 6 O) and propionaldehyde (Figures 2g and S4c), also newly observed in ODHP. Acetaldehyde and its  S5). In addition to propionaldehyde and acetaldehyde, formaldehyde (HCHO) is also detected by its ms-TPES at m/z 30 ( Figure S4b). The oxygenates (C 1−3 aldehydes, C 2−3 ketenes, and C 2−3 enols) found over BN represent minor peaks compared to C 3 H 6 , C 2 H 4 , and CH 4 . However, the desorption of these intermediates from the catalyst surface, as implied by their detection, is crucial in preventing overoxidation (vide infra). Finally, the ion peaks at m/z 15 and m/z 41 detected over BN (Figure 2c) can be unambiguously assigned to methyl (CH 3 • ) and allyl (C 3 H 5 • ) radicals, respectively, based on the corresponding ms-TPES (Figure 2h and i).
Motivated by the observation of oxygenates (aldehydes, ketenes, and enols) and radicals over BN, we further investigated the temperature-dependent product distribution over BN from 550 to 700°C. The corresponding photoionization mass spectra are shown in Figure S6, while the integrated signals are presented in Figure 3. With increasing temperature, the CH 3 • signal increases together with that of CH 4 , indicating that methane is likely formed by gas-phase H addition to CH 3 • (see below). Since CH 3 • is formed from C 3 H 8 , the observed C 2 H 4 from the counter fragment C 2 H 5 comes as no surprise. The C 3 H 5 • signal appears at 600°C ( Figure S7) and continues to increase with temperature, suggesting increasing C 3 H 6 decomposition at elevated temperatures. The rising C 3 H 8 conversion and C 3 H 6 activation affect the final C 3 H 6 signal in opposite ways and result in constant propene abundance when the temperature is increased further from 650 to 700°C (Figure 3). Ethenol, methylketene, and propen-m-ol signals all increase from 550 to 650°C, but become stable or even decrease from 650 to 700°C ( Figure  3), implying that they may also be converted further to other products. The m/z 56 signal taken at a photon energy of 9.5 eV, which can be assigned primarily to C 4 H 8 and, to a lesser degree, methylketene ( Figures S8 and S9), also increases  Table S1 for details. from 550 to 650°C and finally becomes relatively stable. The simultaneous increase of the C 4 H 8 signal together with CH 3 • and C 3 H 5 • from 550 to 650°C suggests a possible gas-phase CH 3 • + C 3 H 5 • reaction to yield C 4 H 8 . The relatively stable m/z 56 signal between 650 and 700°C is due to the simultaneous consumption of C 4 H 8 by dehydrogenation, yielding C 4 H 6 (m/z 54, Figure 3). CH 3 • is produced by desorption from the catalyst surface after propane activation, because C 3 H 8 cracking is not observed in the blank experiment ( Figure S3). It is worth noting that C 2 radicals, such as C 2 H 5 • , have not been observed. We consider the role of oxygenates in the ODHP mechanism and compare it with that in combustion. Notably, typical combustion temperatures are significantly higher than the ODHP reaction temperatures of 550−600°C. However, analogous reactions are likely to occur on ODHP catalytically, since the catalyst provides a dehydrogenation active site. 44 Vinyl alcohol (CH 2 �CH− OH) is less stable than its tautomer acetaldehyde 45 by ca. 42 kJ mol −1 and yet plays a role in the partial oxidation of hydrocarbons and alcohols. 46 The detection of C 2−3 enols in this work implies a possible low-temperature surface-mediated enol formation side reaction in BN-catalyzed ODHP. In contrast to VO x /SiO 2 , full oxidation to CO 2 is inhibited over BN, which is explained by only weakly surface-bound partially oxidized intermediates, which readily desorb into the gas phase and are detected. The observation of ketenes has mechanistic ramifications, as well. The presence of ethylketene (CH 3 CH 2 − CH�C�O) is evidenced by the m/z 70 ms-TPES ( Figure  S10) over BN. Ketenes have been widely reported as the key intermediates for olefin formation, 3,35,47,48 which indicates an oxygenate-driven route in BN-catalyzed ODHP and will be addressed computationally in the following section.
Computational Insights. Quantum chemical calculations were carried out for both surface-catalyzed and gas-phase reactions to reveal the origin of the reactive radicals and oxygenates detected over BN by operando PEPICO. Analogous theoretical calculations on ODHP over VO x -based catalysts have been carried out previously and, in light of propane overoxidation over VO x , are not discussed here further. 49−51 Although a crystalline B 2 O 3 (101) surface was used to simplify  the BN active surface, it is accepted that the active catalyst surface corresponds more closely to amorphous boron oxide in ODHP. 34 To simulate the real catalyst surface, ab initio molecular dynamics (AIMD) simulations were first carried out for the B 2 O 3 (101) surface at the high temperatures of 1500− 2000 K to achieve a disordered BO x surface within the tractable time window of 2−4 ps ( Figure S11). O 2 chemisorption on BO x forms peroxo-like >BO−OB< species preferentially, which is regarded as the main site for C 3 H 8 activation. 28,31 O 2 chemisorption is modeled on the amorphous BO x surface, and the adsorption configurations of C 3 H 8 on the >BO−OB< site are optimized ( Figure S12). After C−H cleavage at the >BO−OB< site (Figure 4a), C 3 H 8 * will transform into coadsorbed >BO−n-C 3 H 7 * and >BO−H* over a barrier of 0.31 eV (* represents adsorbed species). The formed n-C 3 H 7 * is strongly bound to the surface with nearly 3 eV desorption energy and will further dehydrogenate to yield C 3 H 6 * over a barrier of only 0.57 eV. Alternatively, C−C cleavage in C 3 H 8 on >BO−OB< can also occur ( Figure S13), but the cracked CH 3 * and C 2 H 5 * radicals are also strongly bound to the surface at desorption energies of 3.69 and 3.48 eV, respectively. This is much higher than the alternative C 2 H 4 formation energy of 1.01 eV ( Figure S14), which suggests that free radicals are not desorbed from the >BO−OB< site.
AIMD simulations of the BO x surface also yielded a >BO dangling site occurring at 1000 K ( Figure S15). This site was also proposed to be active in boron-catalyzed ODHP. 31 The C−C bond can be cleaved on the >BO dangling site (Figures  4b and S16). Here, C 3 H 8 * transforms into either >BO− C 2 H 5 */CH 3 * or >BO−CH 3 */C 2 H 5 *. The desorption energy of CH 3 * is only 0.18 eV from the former. For >BO−CH 3 */ C 2 H 5 *, calculations predict prompt C 2 H 5 * dehydrogenation to C 2 H 4 * ( Figure S17), implying that C 2 H 5 * is unlikely to desorb into the gas phase. Moreover, C 3 H 6 can also be activated on the >BO dangling site, readily forming gas-phase C 3 H 5 • with a desorption energy of 0.33 eV ( Figure S18). These results identify the >BO dangling site as the source for the experimentally observed CH 3 • and C 3 H 5 • radicals. In addition to the C−C cleavage route on the >BO dangling site, we also found a comparable C−H cleavage barrier toward n-C 3 H 7 * and i-C 3 H 7 * ( Figure S19). The adsorbed i-C 3 H 7 * intermediate will spontaneously transform into C 3 H 6 by terminal H abstraction ( Figure S20). In contrast, n-C 3 H 7 * may either transform into C 3 H 6 over a barrier of 0.17 eV or propagate into the gas phase with a desorption energy of only 0.39 eV ( Figure  S21). Despite surface-catalyzed n-C 3 H 7 * to C 3 H 6 conversion having a lower energy barrier, n-C 3 H 7 * desorption may effectively compete, especially at high temperatures. The propyl radical intermediate was, however, not detected experimentally in the gas phase, in contrast to CH 3 • and C 3 H 5 • . We first have to consider gas-phase radical chemistry to conclude whether this indicates nondesorption or a short gaseous lifetime of n-C 3 H 7 .
In addition to radicals, gas-phase reactive oxygenates are also detected by operando PEPICO. As guided by previous gasphase mechanistic insights on, for example, the role of enols in combustion, 44 we applied the G4 composite method to address possible gas-phase reaction pathways. The formation of enols and ketenes may proceed similarly to a combustion chemistry pathway. Enols, for instance, are likely produced via partial hydrocarbon oxidation on the catalyst surface in ODHP, followed by desorption into the gas phase. In the gas phase, the allyl + O 2 reaction affords formaldehyde (observed in Figure S4) and the H 2 C−C(�O)H radical, in an exothermic reaction (−2.75 eV) after an entrance barrier of 0.55 eV, 52 which is likely reduced in the presence of BN. The H 2 C−C(�O)H radical can either be dehydrogenated to yield ketene or hydrogenated to vinyl alcohol or acetaldehyde (m/z 44, Figures S5, S8), which were all observed in this study. On the one hand, we found computationally that C 2−3 enols (C 2 H 4 O, C 3 H 6 O) can be converted to C 2−3 ketenes (C 2 H 2 O, C 3 H 4 O) via dehydrogenation by, for example, a CH 3 • radical over a low energy barrier of 0.46 eV for C 2 H 2 O and 0.40 eV for C 3 H 4 O ( Figure S22). This H abstraction step of propen-1-ol by CH 3 • after CH 3 • desorption is included in Figure 4b for comparison. Since computational results do not rule out n-C 3 H 7 • desorption from the surface, we also investigate H abstraction from enols via nC 3 H 7 • , which was found to be associated with a similar energy barrier to that with CH 3 • radicals ( Figure S23) and could contribute to quenching n-C 3 H 7 • by C 3 H 8 formation. Hydrogen abstraction from C 2−3 enols by the resonantly stabilized C 3 H 5 • involves a high-energy transition state at ∼0.8 eV ( Figure S24); that is, the allyl radical is unlikely to participate in the first hydrogen transfer from enols. However, the second H-transfer step, to finally yield ketenes, is computed to be downhill, irrespective of the H-acceptor free radical. These observations imply that enols can easily dehydrogenate into ketenes in the presence of the Hacceptor radical species CH 3 • , n-C 3 H 7 • , or�as far as the second H-transfer step is concerned�C 3 H 5 • . However, n-C 3 H 7 • radicals are also H-donor agents. In fact, a barrier-free H-transfer reaction is predicted once an n-C 3 H 7 • radical collides with CH 3 • , C 3 H 5 • , or another n-C 3 H 7 • associated with the release of 2 to 3 eV of energy ( Figure S25), contributing to rapid quenching of n-C 3 H 7 • by H loss to C 3 H 6 and the associated formation of CH 4 , C 3 H 6 , or C 3 H 8 . At a 0.3 bar reactor pressure, the gas-phase collision frequency is on the order of 10 8 s −1 at 600°C, which implies prompt quenching of the n-C 3 H 7 • radical in the presence of H acceptors. Therefore, the absence of n-C 3 H 7 • radical detection can be ascribed (1) to the more favorable surface-confined n-C 3 H 7 * to C 3 H 6 reaction path compared to n-C 3 H 7 * desorption as well as (2) to its low expected lifetime in the gas phase under the reaction conditions and in the presence of H-acceptors. Since ketene and methyl radicals are both observed in the gas phase, and surface methyl and ethyl species are predicted ( Figure S13), ethylketene may also be formed by the methylation of methylketene or ethylation of ketene. The formation of ethylketene is proven by operando PEPICO (Figure S10). Methylketene and ethylketene are known as precursors for C 2 H 4 and C 3 H 6 via decarbonylation. 35 However, these processes can only take place over a high barrier (>3 eV) in the gas phase ( Figure S26). Therefore, ketene decarbonylation is surface catalyzed and proceeds over BN to form CO and olefins, both of which desorb from the surface easily. This also explains how CO is formed in the ODHP over BN. Additionally, the detection of gas-phase oxygenates only over BN suggests they are prone to desorb from the BN surface, which also prevents deep oxidation to CO 2 .
Reaction Mechanism. By combining operando PEPICO experiments and quantum chemical calculations, we have formulated a comprehensive ODHP mechanism over BN in Figure 5, involving coupled surface-confined and gas-phase reactions. Gas-phase radicals and reactive oxygenates are detected by operando PEPICO. Calculations predict that C 3 H 8 transforms into n-C 3 H 7 * on the >BO−OB< site (Figure 5a), which only desorbs as C 3 H 6 after further dehydrogenation. C 3 H 8 can also dissociate to CH 3 * and C 2 H 5 *. However, both are strongly surface-bound and do not desorb from the >BO− OB< site. On the >BO dangling site, n-C 3 H 7 * and i-C 3 H 7 * are both formed after C−H activation (Figure 5b). However, i-C 3 H 7 * yields C 3 H 6 instantaneously, and only n-C 3 H 7 * may desorb due to its low desorption energy of 0.39 eV. Yet, n-C 3 H 7 • is not detected in the gas-phase experimentally, which can be attributed to the more favorable surface reaction to C 3 H 6 and its short gas-phase lifetime in the presence of H acceptors. Additionally, >BO−C 3 H 6 * can be activated to form C 3 H 5 * on the >BO dangling site, followed by easy desorption to form the allyl free radical. C−C cleavage on the >BO dangling site has two different reaction routes, forming >BO− C 2 H 5 */CH 3 * or >BO−CH 3 */C 2 H 5 *. The former yields gasphase CH 3 • , and the latter will transform to C 2 H 4 . Thus, free radicals detected by operando PEPICO spectroscopy are derived from the >BO dangling site. Partial oxidation products can also be observed in the gas phase with BN as a catalyst (Figure 5c). The desorption of oxygenates from the catalyst surface indicates weak bonding, which is crucial in preventing catalytic oxygenation of oxygenates into CO and CO 2 .
After the desorption of radicals and oxygenates, gas-phase H-abstraction chemistry will be initiated (Figure 5c). Aside from the self-reaction to form C 2 H 6 , the CH 3 • free radical can also abstract an H from enols. Gas-phase dehydrogenation into ketenes with H-acceptors (e.g., CH 3 • , n-C 3 H 7 • , or, in the second step, C 3 H 5 • ) can take place readily based on G4 calculations ( Figures S22−S24). Meanwhile, CH 3 • , n-C 3 H 7 • , and C 3 H 5 • will form CH 4 , C 3 H 8 , and C 3 H 6 after abstracting an H. The formed methylketene and ethylketene can then undergo surface-assisted decarbonylation to form C 2 H 4 and C 3 H 6 ( Figure 5c). This new C 3 H 6 and C 2 H 4 formation mechanism is driven by methyl-and ethylketene in BNcatalyzed ODHP, which is observed for the first time. In addition to surface-bound propyl radicals (C 3 H 7 *) previously considered to be the only intermediate in C 3 H 6 formation, this establishes partially surface-oxidized oxygenates such as C 3 enols (propen-1-ol, C 3 H 6 O) as precursors to C 3 H 6 via a gasphase reaction sequence including dehydrogenation, methylation, and subsequent surface-confined ethylketene decarbonylation over the BN catalyst (Figure 5c). C 4 H 8 , only observed over a BN catalyst, can be ascribed to the gas-phase CH 3 • + C 3 H 5 • association reaction. Importantly, oxygenates desorb easily from the BN surface before deep oxidation to CO 2 , and ketene decarbonylation is the main CO source over BN. This explains why carbon monoxide is the main deeply oxidized byproduct over BN with negligible CO 2 formation. In contrast, CO 2 is dominantly formed over vanadium-based catalysts, and no oxygenates can be observed in the gas phase, due to strong binding to the surface and rapid catalytic deep oxidation. Additionally, quantum chemical calculations indicate that n-C 3 H 7 • could escape from the catalyst surface. If so, n-C 3 H 7 • will participate in gaseous reactions with H-donor or Hacceptor species to form C 3 H 6 or C 3 H 8 (Figure 5c). In this coupled surface and gas-phase route (Figure 5c), H-acceptor radicals and H-donor oxygenates (or n-C 3 H 7 • ) react via fast H migration. The selectivity to CO rises with temperature ( Figure 1a), which suggests a competitive advantage of the oxygenate route at high temperatures. Although CO is an undesirable ODHP product, it is still more useful than CO 2 , making the transformation of C 3 H 8 and O 2 to olefins and CO over boron-based catalysts alluring. Calculations also revealed the different role of chemistry at the >BO−OB< main and the >BO dangling sites. Such insights help tune the catalytic activity in a targeted way and may guide the rational design of boron-based catalysts with higher selectivity.

■ CONCLUSIONS
In addition to propyl as the main precursor of propene, we identify a new H-acceptor radical-and H-donor oxygenatedriven propene formation route by operando PEPICO spectroscopy. Partially oxidized enols dehydrogenate in the presence of H-acceptor radicals to ketenes, which then transform to olefins by decarbonylation. Ethylketene is the observed precursor of propene in this route, which is unique to BN and absent over a vanadium-based catalyst. Moreover, we report that free radicals are derived solely from the >BO dangling site, as they are strongly bound at the >BO−OB< main site. These results not only help explain the ever-elusive ODHP reaction mechanism on BN but also aid to design nextgeneration boron-based catalysts for enhanced ODHP performance.
The experimental methods, detailed PEPICO, and theoretical calculation data (PDF)