Article Cite This: J. Am. Chem. Soc. 2019, 141, 10722−10728
pubs.acs.org/JACS
Stable and Active Oxidation Catalysis by Cooperative Lattice Oxygen Redox on SmMn2O5 Mullite Surface Yongping Zheng,†,§ Sampreetha Thampy,† Nickolas Ashburn,† Sean Dillon,† Luhua Wang,† Yasser Jangjou,‡ Kui Tan,† Fantai Kong,† Yifan Nie,† Moon J. Kim,† William S. Epling,‡ Yves J. Chabal,† Julia W. P. Hsu,† and Kyeongjae Cho*,† †
Department of Materials Science and Engineering, University of Texas at Dallas, Richardson, Texas 75080, United States Department of Chemical Engineering, University of Virginia, 102 Engineers’ Way, Charlottesville, Virginia 22904-4741, United States
Downloaded via NOTTINGHAM TRENT UNIV on July 21, 2019 at 14:02:03 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
‡
S Supporting Information *
ABSTRACT: The correlation between lattice oxygen (O) binding energy and O oxidation activity imposes a fundamental limit in developing oxide catalysts, simultaneously meeting the stringent thermal stability and catalytic activity standards for complete oxidation reactions under harsh conditions. Typically, strong O binding indicates a stable surface structure, but low O oxidation activity, and vice versa. Using nitric oxide (NO) catalytic oxidation as a model reaction, we demonstrate that this conflicting correlation can be avoided by cooperative lattice oxygen redox on SmMn2O5 mullite oxides, leading to stable and active oxide surface structures. The strongly bound neighboring lattice oxygen pair cooperates in NO oxidation to form bridging nitrate (NO3−) intermediates, which can facilely transform into monodentate NO3− by a concerted rotation with simultaneous O2 adsorption onto the resulting oxygen vacancy. Subsequently, monodentate NO3− species decompose to NO2 to restore one of the lattice oxygen atoms that act as a reversible redox center, and the vacancy can easily activate O2 to replenish the consumed one. This discovery not only provides insights into the cooperative reaction mechanism but also aids the design of oxidation catalysts with the strong O binding region, offering strong activation of O2, high O activity, and high thermal stability in harsh conditions.
■
INTRODUCTION Heterogeneous oxidation catalysis is of pivotal importance to many aspects of modern society, from chemicals synthesis to energy production and pollutant remediation.1 Over the past few decades, great success has been achieved in oxidation catalyst development based on transition metal oxides (TMOs).2,3 However, most of them deal with industrial partial oxidation processes where high selectivity is essential. Complete oxidation catalysis, especially in harsh conditions for abatement of fossil fuel emissions (e.g., CO and NOx), which requires both high activity and high thermal stability, heavily relies on expensive platinum group metals (PGMs).4 To meet increasing global environmental and energy demands, searching for earth-abundant TMOs as alternative complete oxidation catalysts is still imperative.5 One of the main challenges in designing alternative TMOs that match PGMs’ catalytic performance in harsh conditions is the poor stability of most active TMO catalysts, as their surfaces suffer from reconstruction and deactivation under hydrothermal aging.6−8 Fundamentally, this frequently observed surface degradation phenomenon is caused by the underlying catalytic mechanism at oxide surfaces, which is © 2019 American Chemical Society
closely related to the stability of lattice oxygen. A general surface oxidation catalytic mechanism of TMOs was originally proposed by Mars and van Krevelen in 1954, and a substantial amount of related research work was done in late 20th century.9−11 In this mechanism, lattice oxygen on oxide surfaces has been widely accepted as the oxidizing agent responsible for oxidation catalysis through a sequence of reaction steps at the same oxygen site.5 Typically, a lattice oxygen is consumed during the oxidation reaction by creating a surface oxygen vacancy (VO) (e.g., VO-O + NO → VO-ONO→ VO + NO2), and subsequently, this VO is later refilled by the dissociation of adsorbed O2 on the oxide surface; this process is known as the classic Mars−van Krevelen (MvK) reaction pathway.9 In this sense, the relatively weak binding of lattice oxygen on oxide surfaces benefits the oxidation activity, which is, however, also detrimental to the surface stability due to the weak lattice framework, leading to surface degradation reactions. This is the intrinsic reason that most active TMO catalysts have poor durability in harsh conditions. Received: March 27, 2019 Published: June 18, 2019 10722
DOI: 10.1021/jacs.9b03334 J. Am. Chem. Soc. 2019, 141, 10722−10728
Article
Journal of the American Chemical Society
Figure 1. Surface characterization. (a) SEM, (b) TEM, (c) HRTEM, and (d) LEIS spectrum of SmMn2O5 oxide. (e) Simulated atomic structure of the (010) surface (a−c plane).
■
RESULTS Investigating the Stability of Lattice Oxygen on the Surface. Mullite-phase SmMn2O5 oxide powders were prepared by a co-precipitation method17,18 and characterized by X-ray diffraction (Supplementary Figure 1a) and X-ray photoemission spectroscopy (Supplementary Figure 1b). The diffraction pattern shows that the sample is well crystallized and in pure phase as previously reported.18 To determine the crystal orientation of exposed surfaces, we performed scanning electron microscopy (SEM) and high-resolution transmission electron microscopy (HRTEM) measurements. The SEM image in Figure 1a and the TEM image in Figure 1b clearly show a plate-like morphology of SmMn2O5 oxide particles with a prevailing surface orientation, which is consistent with the fact that cleavage in the [010] direction is dominant for mullite.19 The spacing of lattice fringes, 1.95 and 2.47 Å, in the HRTEM image (Figure 1c) belongs to SmMn2O5 (202) and (300) planes, respectively. These results, taken together, corroborate that the (010) surface perpendicular to the (202) and (300) planes is most likely exposed. To further examine the exposed atomic composition, low-energy ion scattering spectroscopy (LEIS), an extremely surface sensitive technique, was performed. The LEIS spectrum in Figure 1d identifies that the exposed atomic ratio between Sm and Mn is around 0.5, indicating that the exposed (010) surface is Mn rich, as shown in Figure 1e. First of all, we need to validate the rationale that stable SmMn2O5 catalysts should have relatively strong binding of lattice oxygen on the surface. Although density functional theory (DFT) with a generalized gradient approximation (GGA) functional describing the exchange−correlation energy is fairly accurate in determining energy-related material properties such as binding energy, it is notorious that DFT has difficulty in describing strongly correlated electrons in TMOs. To overcome this difficulty, we use the recently developed improved DFT+U method by using temperatureprogrammed surface lattice oxygen desorption (O2-TPD)
Ideally, a stable and active TMO catalyst requires a relatively strong binding of lattice oxygen on the surface to suppress surface phase transition and deactivation and simultaneously requires facile O2 dissociation on VO, as well as active lattice oxygen on the surface for oxidation reactions. Unfortunately, these conflicting requirements are difficult to satisfy at the same time through the classic MvK reaction pathway on TMOs at the same lattice oxygen sites. There are strong correlations among lattice oxygen activity, binding energy, and O2 activation, which are ubiquitous on TM-based catalysts for different reactions.2,3,12−14 Typically, a relatively strong binding of lattice oxygen results in strong O2 activation on VO, but low lattice oxygen activity for oxidation of reactants. Such a correlation seems like an insurmountable obstacle against designing an ideal complete oxidation catalyst based on TMOs satisfying both stability and activity requirements. However, recent work on Mn-mullite RMn2O5 (R= Sm, Gd) compounds,15−17 a new class of complex oxide catalysts for NO oxidation in diesel exhaust, challenges this conventional wisdom based on the classic MvK mechanism. We now designate the classic MvK mechanism as sequential MvK reaction on a single oxygen site in contrast to cooperative MvK reaction on multiple oxygen sites, which we will explain as the newly identified mechanism on a mullite surface, simultaneously satisfying high stability and activity requirements. It was reported that RMn2O5 oxides have not only high thermal durability but also superior catalytic activity over commercial PGM-based catalysts and corresponding perovskite RMnO3 oxide catalysts.15,17 The open intriguing question here is how the commonly accepted correlations are not applicable to mullite-phase oxides, or whether there is a new mechanism other than the classic MvK reaction pathway. In this paper, we present unambiguous theoretical and experimental evidence of a cooperative MvK reaction on multiple oxygen sites to elucidate this unusual catalytic performance on the SmMn2O5 surface. 10723
DOI: 10.1021/jacs.9b03334 J. Am. Chem. Soc. 2019, 141, 10722−10728
Article
Journal of the American Chemical Society peaks to calibrate the Hubbard U values,20 which enables us to obtain a reliable calculated surface chemistry. On the mullite (010) surface, we found three types of lattice oxygen, α, β, and γ, in different coordination environments as labeled in Figure 2a, with calculated binding energies of −1.19, −2.13, and
structure of O 2p bands of the corresponding lattice oxygen, which explicitly associates the amount of unoccupied 2p states of lattice oxygen α, β, and γ (α > β > γ) with their binding strengths on the (010) surface (α < β < γ), revealing the fact that lattice oxygen with a weaker binding energy is more electron deficient. Surface oxygens on TMOs are reported to be electron deficient and redox active due to coordinative undersaturation,21 so that lattice oxygen α and β are responsible for NO adsorption by accepting extra electrons, while γ oxygen is inactive due to its almost fulfilled 2p band. Given that the free energy of NO oxidation by O2 (NO + 1 /2O2 → NO2) is around −0.56 eV, according to the NIST thermodynamic database,22 and that the most weakly bound lattice oxygen α (Oα) is calculated to have a binding energy of −1.19 eV, the lattice oxygen is too strongly bound to directly oxidize NO into NO2 through the sequential MvK reaction on the single oxygen site by creating an oxygen vacancy. This result theoretically confirms the presence of relatively strong binding of lattice oxygen on the mullite surface, consistent with the experimentally observed stability of mullite surfaces in recent work,15,16 and it also predicts that mullite surfaces cannot turn NO into NO2 if there is no other reaction pathway. Probing the Reactivity of Lattice Oxygen on the Surface. To corroborate the DFT results, we performed in situ infrared (IR) spectroscopy and NO-temperature-programmed desorption (TPD) measurements to probe the reactivity of lattice oxygen on the surface. The oxidized sample is exposed to the NO gas at 40 °C and characterized by IR measurements. The IR absorption bands at 1542, 1293, and 1026 cm−1 growing with NO exposure as shown in Figure 3a (bottom) correspond to nitrite or nitrate species on oxide surfaces,23−25
Figure 2. Surface lattice oxygen characteristics. (a) Lattice oxygen atoms on the simulated (010) surface with different coordinative structures. (b) Electronic structures of the corresponding O 2p bands. The vacuum level (0.0 eV) is set as the reference, and unoccupied states are above the Fermi level.
−2.80 eV, respectively (see Methods for a detailed description of the calculations). These lattice oxygen species were assigned to the three distinct O2 desorption peaks observed in the O2TPD experiment reported in recent work, and the remaining lattice oxygen species were found to be stable against desorption until collapse of the (010) surface at very high temperature, above 800 °C.20 Figure 2b shows the electronic
Figure 3. Surface lattice oxygen reactivity measurement. (a) IR spectra of SmMn2O5 under 1% NO/He exposure growing with time and after evacuation at 40 °C. (b) NO-TPD profile of oxidized SmMn2O5 after NO exposure at 40 °C. The desorption peaks are fitted with Gaussian functions. (c) Configurations of NO adsorption on top of lattice oxygen at different NO coverages (θ) and the coverage-dependent NO and NO2 desorption barriers. 10724
DOI: 10.1021/jacs.9b03334 J. Am. Chem. Soc. 2019, 141, 10722−10728
Article
Journal of the American Chemical Society
Figure 4. NO oxidation mechanism under an O2 environment. (a) Simulated NO oxidation reaction pathway under an O2 environment through cooperative oxygen redox. The energy values on the inserted profile are in eV, and the red balls depict oxygen atoms from O2 gas. (b) Differential IR spectra with in situ heating under He flow and (c) 10% O2 in He flow of SmMn2O5 after 1% NO/He exposure and He flush at 40 °C. The differential spectra at 40 °C are the absorbance differences after He or 10% O2 in He flush for 15 min, while those at elevated temperatures are the absorbance differences between current and previous temperature. Each temperature is held for 15 min.
which are stable upon evacuation at 40 °C, as shown in Figure 3a (top). There is no measurable adsorbed NO stretching band on the transition metal sites around 1800 cm−1 after evacuation,26 indicating NO molecules only interact strongly with lattice oxygen, forming surface NOx species. This is in line with the classic MvK mechanism, which suggests NO molecules preferentially react with surface oxygen. The NOTPD study in Figure 3b shows negligible NO2 generation, validating the idea that lattice oxygen on the mullite surface is reactive with NO, but unable to directly oxidize NO into NO2 by creating an oxygen vacancy through the sequential MvK reaction on single oxygen site. This experimental result substantiates the lattice oxygen binding energy calculations above that show the relatively strong binding of lattice oxygen leading to limited direct oxidation activity. To get molecular level understandings of the NO adsorption and desorption behavior, we calculated the most stable adsorption configurations with different NO coverages (θ) on the mullite (010) surface. The binding energy of NO on the exposed Mn site is calculated as −0.43 eV at 0 K, but becomes positive 0.21 eV at 40 °C under 40 Torr NO exposure when taking into account the entropy contribution of NO gas.22 This weak interaction makes NO adsorption on the Mn sites energetically unfavorable under the experimental condition, consistent with the observation that no stretching band of adsorbed NO on Mn sites (∼1800 cm−1) was detected in IR spectra (Figure 3a). In contrast, the adsorption of NO on top of Oα and Oβ is energetically favorable, as predicted by the O 2p band analysis. Figure 3c shows the coverage-dependent adsorption configurations and desorption barriers of NO on the mullite (010) surface. Figure 3c (ii) shows the chelating and bridging nitrites (NO2−) formed with full NO coverage (θ = 1) at low temperature, where O−N−Oα chelates the isolated Mn site by two oxygen atoms, and O−N−Oβ bridges between
two neighboring Mn sites by Oβ and N ions. The vibrational frequencies associated with chelating and bridging nitrites are calculated as νasym(NO2) = 1296 cm−1 and ν(NO) = 1573 cm−1, respectively, matching the IR absorption peaks in Figure 3a. With increasing temperature, NO molecules desorb first from bridging nitrites O−N−Oβ due to lower Oβ reactivity, responsible for the first NO desorption peak at 127 °C in the NO-TPD profile. Using microkinetic analysis described in the previous work,20 this desorption temperature corresponds to a 1.18 eV desorption barrier. The remaining chelating nitrites (θ = 0.5) tend to form more stable bridging nitrates (NO3−) with NO bonding to both Oα and Oβ as shown in Figure 3c (iii), and the desorption of NO from bridging nitrates requires a higher temperature, responsible for the second NO desorption peak at 221 °C, corresponding to a 1.43 eV desorption barrier. The rest of the bridging nitrates are further stabilized due to lower NO coverage (θ = 0.25), responsible for the highest desorption peak at 292 °C, corresponding to a 1.69 eV barrier. From Figure 3b we can see that the amount of NO desorbed at low temperature is about twice that desorbed in the second and third regions, well consistent with the simulated desorption scenario (θ = 1 → θ = 0.5 → θ = 0.25 → θ = 0). Nevertheless, the calculated NO desorption barriers deviate a lot from the NO-TPD-derived NO binding energy values, caused by the overbinding of gas phase and adsorbed NOx species calculated with the GGA-PBE functional (Supplementary Table 1).20 The NO and NO2 desorption barriers with different NO coverages were then corrected by considering the overbinding of NO, NO2, NO2−, and NO3− (Supplementary Table 2), shown in Figure 3c. We can see that the NO2 desorption barriers are always higher than NO, leading to negligible NO2 generation in the NO-TPD experiment, which, however, conflicts with the superior catalytic activity of SmMn2O5 catalysts for NO oxidation 10725
DOI: 10.1021/jacs.9b03334 J. Am. Chem. Soc. 2019, 141, 10722−10728
Article
Journal of the American Chemical Society
proposed in Figure 4a, rather than only replenishing the created lattice vacancies.
over commercial PGM-based catalysts during the performance test.15 Therefore, a new reaction mechanism other than the classic MvK reaction pathway must exist when lean NO is exposed to SmMn2O5 catalysts in an O2-rich environment. Cooperative Reaction Mechanism on the Mullite Surface. To unravel the underlying mechanism leading to high oxidation activity of strongly bound lattice oxygen on the mullite surface, we simulated the possible reaction pathway under the reaction conditions (400 ppm of NO with 10% O2 at 300 °C). At high temperature with a small partial pressure, the entropy contribution further stabilizes the gas-phase NO by −1.44 eV, leading to a low coverage of NO on the mullite (010) surface, so the following reactions are modeled at θ = 0.5 according to the NO binding strength at different coverages. Interestingly, we found that although the surface nitrate is difficult to directly desorb as NO2 by creating an oxygen vacancy with a large desorption barrier of 1.66 eV, a concerted rotation of bridging nitrate species can take place with simultaneous O2 adsorption. As shown in Figure 4a (ii), the Oα vacancy can be created as a transition state of this rotation with a small rotation barrier, 0.54 eV, and the adsorption of O2 on the Oα vacancy further stabilizes this transition state and increases the population of monodentate nitrate species after rotation. The subsequent dissociation and migration of adsorbed O2 along the path shown in Figure 4a (iii) with a small activation barrier, 0.65 eV, further transform the bridging nitrates into monodentate nitrates. Subsequently, the monodentate nitrates release NO2 with a much smaller desorption barrier, 1.08 eV, and the bare (010) surface is restored. The whole oxidation process is accomplished by cooperative oxygen redox, where Oβ functions as a reversible anionic redox center that accepts electrons through NO adsorption and then gives electrons through NO2 desorption, O2 stabilizes the critical rotation step by filling the Oα vacancy, while Oα functions as an oxidizing agent, turning NO into NO2 mediated by Oβ and O2 gas. The cooperative reaction mechanism described above explains the superior oxidation catalytic activity of mullite oxides in the reaction condition and answers the initial question of how the universal correlation can be broken. To provide extra experimental evidence of this new reaction pathway under an O2 environment, we measured the changes in IR spectra as a function of temperature of mullite oxides after NO exposure at 40 °C with in situ heating under pure He flow (Figure 4b) and 10% O2 in He flow (Figure 4c). The differential spectra in Figure 4b,c were measured after holding at each temperature for 15 min under the gas flow. If the classic MvK mechanism were operative, then NO2 would desorb first from the surface with oxygen vacancies being created, and O2 then would replenish the vacancies after lattice oxygen was consumed. As the reactions take place in a sequential way, O2 flow should not affect the desorption of NOx species from the surface. However, NO-TPD in Figure 3b shows that only NO desorbs from the surfaces; that is, no lattice oxygen vacancy was created, and Figure 4b,c clearly show that O2 strongly affects the desorption of NOx species. While little or no NOx species were desorbed in pure He flow at 40 °C (Figure 4b black curve), a majority of surface NOx species desorb under 10% O2 in He flow (Figure 4c black curve). Without O2, there is negligible desorption of surface NOx below 200 °C, and the majority do not desorb until the temperature is ramped up to 300 °C. This result unambiguously validates the fact that O2 is involved in the desorption reaction of surface NOx species as
■
CONCLUSIONS By integrating first-principles modeling and experiments, we demonstrated neighboring lattice oxygen pairs on the SmMn2O5 mullite surface working in synergy to circumvent the conflicting correlation between lattice oxygen stability and activity, resulting in an ideal complete oxidation catalyst with both high catalytic activity and high thermal stability in harsh conditions. The identification of cooperative lattice oxygen redox on transition metal oxide surfaces for boosting catalytic performance offers a new path for manipulating catalytic properties based on earth-abundant elements and overcoming the theoretical limits for heterogeneous oxidation catalysis.
■
METHODS
Ab Initio Calculations. Spin-polarized DFT simulation was performed using the Vienna ab Initio Simulation Package (VASP)27,28 with the projector-augmented wave (PAW) method.29 The plane wave cutoff energy was set to 450 eV, and the GGA with the semilocal Perdew−Burke−Ernzerhof (PBE)30 functional was adopted to describe the exchange−correlation interactions. The Brillouin zone was sampled with Monkhorst−Pack k-meshes with a spacing of 0.03 Å−1 for the crystal structure optimization. The criteria of convergence for energy and force were set to 10−4 eV and 0.01 eV/ Å, respectively. On-site Coulomb repulsion of Mn 3d electrons was corrected by the GGA+U method31−33 with Ueff = 4.5 eV according to the O2-TPD in previous work.20 The binding energies of lattice oxygen species were calculated with respect to gas-phase O2 at 0 K, and the overbinding of GGA in the O2 molecule (−1.36 eV) is corrected.34 The binding energy of lattice oxygen β was calculated after the desorption of lattice oxygen α, and the binding energy of lattice oxygen γ was calculated after the desorption of lattice oxygen α and β, and so on. The oxygen dissociation barriers were calculated using the climbing image nudged elastic band method.35 Structure Characterization. X-ray diffraction (XRD) measurement was performed in a Rigaku Ultima III diffractometer (40 kV/44 mA) with Cu Kα radiation (λ = 1.5406 Å). The pattern was collected over a 2θ range from 20° to 60° with a step size of 0.02° and a scan speed of 0.5°/min. The crystalline phase was identified by matching the measured diffraction pattern with the powder diffraction files. The X-ray photoelectron spectroscopy (XPS) measurement was carried out using a PHI5000 Versa Probe II (Physical Electronics Inc.). A monochromatic Al Kα X-ray source (1486.6 eV) was used, and the data were collected at a takeoff angle of 45° by a concentric hemispherical analyzer with pass energy of 23.5 eV and a step size of 0.2 eV. Peak fitting was performed as reported in our previous work.18 The TEM measurement was performed using a JEOL ARM200, Japan. The morphology of the SmMn2O5 sample was characterized using a Zeiss Supra 40 field-emission scanning electron microscope. Low-Energy Ion Scattering. LEIS measurement was performed in a custom-built ultra-high-vacuum chamber with a base pressure of
