Predicting the Catalytic Activity of Surface Oxidation Reactions by Ionization Energies

Abstract

Open AccessCCS ChemistryRESEARCH ARTICLE1 Aug 2020Predicting the Catalytic Activity of Surface Oxidation Reactions by Ionization Energies Shenjun Zha†, Zhi-Jian Zhao†, Sai Chen, Sihang Liu, Tao Liu, Felix Studt and Jinlong Gong Shenjun Zha† Key Laboratory for Green Chemical Technology of Ministry of Education, School of Chemical Engineering and Technology, Tianjin University, Collaborative Innovation Center of Chemical Science and Engineering, Tianjin 300072 , Zhi-Jian Zhao† Key Laboratory for Green Chemical Technology of Ministry of Education, School of Chemical Engineering and Technology, Tianjin University, Collaborative Innovation Center of Chemical Science and Engineering, Tianjin 300072 , Sai Chen Key Laboratory for Green Chemical Technology of Ministry of Education, School of Chemical Engineering and Technology, Tianjin University, Collaborative Innovation Center of Chemical Science and Engineering, Tianjin 300072 , Sihang Liu Key Laboratory for Green Chemical Technology of Ministry of Education, School of Chemical Engineering and Technology, Tianjin University, Collaborative Innovation Center of Chemical Science and Engineering, Tianjin 300072 , Tao Liu Key Laboratory for Green Chemical Technology of Ministry of Education, School of Chemical Engineering and Technology, Tianjin University, Collaborative Innovation Center of Chemical Science and Engineering, Tianjin 300072 , Felix Studt Institute of Catalysis Research and Technology, Karlsruhe Institute of Technology, 76344 Eggenstein-Leopoldshafen Institute for Chemical Technology and Polymer Chemistry, Karlsruhe Institute of Technology, 76131 Karlsruhe and Jinlong Gong *Corresponding author: E-mail Address: [email protected] Key Laboratory for Green Chemical Technology of Ministry of Education, School of Chemical Engineering and Technology, Tianjin University, Collaborative Innovation Center of Chemical Science and Engineering, Tianjin 300072 https://doi.org/10.31635/ccschem.020.201900096 SectionsSupplemental MaterialAboutAbstractPDF ToolsAdd to favoritesDownload CitationsTrack Citations ShareFacebookTwitterLinked InEmail Developing a descriptor to understand the reactivity of a catalyst is critical in achieving the rational design of heterogeneous catalysts. Ideally, the descriptor should be simple, predictive, as well as applicable to diverse types of reactions. This paper describes the development of a descriptor that could meet such ideal requirements based on its element-specific fundamental property, ionization energy. Our results indicated that ionization energies could be utilized to describe successfully the adsorption energies of oxygen (O*) and hydroxyl (OH*) groups on various materials. Moreover, we constructed a bond formation scheme to parse this phenomenon. The numerical value of ionization energy of the surface dopants determined its the position of d band in energy, and thus, the filling of the antibonding state after the adsorbate–surface interaction, indicating that the ionization energy could act as a descriptor that correlates with the adsorption energy. The application of this descriptor was employed successfully in a selective oxidation reaction of methane to methanol, which revealed the potential of the ionization energy to possess the capability to predict and modulate the reactivity of other surface oxidation reactions. Download figure Download PowerPoint Introduction In the field of heterogeneous catalysis, enhancing the activity and selectivity of a catalytic material toward the desired product is key to an efficient realization of the processes.1 Due to the complex structures of many heterogeneous catalysts, the rational design of new catalyst materials is extremely challenging, and a large number of empirical rules have been developed to guide the discovery of optimum catalysts.2,3 As early as the 1920s, Sabatier4 introduced the magnitude of the interaction strength between adsorbates and the catalyst surface as an essential descriptor to achieve high catalytic performance. Measuring the interaction strength experimentally is rather challenging; however, recent advances in density functional theory (DFT) calculations has enabled the determination of the interaction strength between reaction intermediates and catalyst surfaces.5–8 Moreover, DFT studies rationalize the Sabatier principle through linear scaling relationships, where it was shown that the adsorption energies and transition state energies of reaction intermediates are a function of another adsorption energy.8–11 Other descriptors that have been employed successfully are based on their electronic (e.g., the d-band center12,13) or geometric structures (e.g., the generalized coordination number14,15) of the catalytic surfaces. The utilization of reactivity descriptors allows an approximate description of the catalytic processes by reducing the high-dimensional reaction network. Interestingly, some newly emerged descriptors, for example, the generalized coordination number, require no additional DFT calculations to obtain the property of the descriptor, further speeding up the computational screening.14 Similar descriptors like the electronegativity of dopants, the band-gap energy, and the coordination-based φ have also been employed.16–18 However, such new types of descriptors are often limited to specific reactions and/or catalytic material, motivating us to develop a much more universal descriptor. Herein, we introduce the ionization energy as a potentially simple, predictive, and universal descriptor in the field of surface oxidation reactions. The ionization energy is defined as the minimum energy required to remove an electron from an isolated atom, ion, or a molecule in the gaseous phase, whereas surface oxidation could be viewed as a process containing charge transfer from the surface to an adsorbate.19 Inspired by the analogy between ionization and surface oxidation (losing electron), we developed scaling relationships between the ionization energy of surface metal atoms/ions and the adsorption energies of adsorbates. Experimental Methods We used the Vienna Ab initio simulation package (VASP) to perform quantum mechanics calculations with the Perdew–Burke–Ernzerhof (PBE) exchange-correlation function.20–22 Valence electrons were described using a plane-wave basis set with the cut-off energy of 400 eV. Meanwhile, core electrons were treated using the projector augmented wave (PAW) method.23 We employed the Monkhorst-Pack k-points grid to sample the Brillouin zone of the surface,24 and determined the electronic occupancies according to the Methfessel–Paxton scheme with an energy smearing of 0.15 eV, evaluating the total energies by extrapolating to zero broadenings. The dipole correction was included in all directions, and all the structures were optimized until each atomic force was less than the setting values. (All the detailed parameters are shown in Supporting Information Table S1.) Spin-polarized calculations were carried out for all varied materials and for the magnetic ordering of ACo2O4 was set according to the results of Chen et al.25–27 For the sake of simplicity, no Hubbard-U correction (DFT+U) was included, as we predicted that although the adsorption energy might change with the Hubbard-U correction, it was unlikely to break the relationship between the ionization energy and the adsorption energy.28 The zeolites were calculated using Bayesian error estimation functional-van der Waals correlation (BEEF-vdW) exchange-correlation functional.29,30 The catalysts were synthesized by metal ion exchange over the commercial zeolite MOR (mordenite; CBV10 A, SiO2/Al2O3 = 13, in sodium form), namely, M-MOR (M = Cu, Pd, Fe, Ag). For example, Cu–MOR was prepared by the following steps: 1 g of zeolite was stirred in 50 mL of 0.05 M solutions of Cu(NO3)2·3H2O (99%, J&K China Chemical Co. Ltd.) at 323 K overnight. Then the suspension was collected by centrifugation and washed six times thoroughly with water. The Cu–MOR precursor was dried overnight at 393 K in a drying oven; after that, it was calcined in the air at 773 K for 4 h using a heating ramp of 3 K min−1. Other metal-exchanged MOR zeolites were prepared using similar procedures. Subsequently, M–MOR catalysis was performed for methane oxidation reactions, and the products collected were analyzed on-line by an Agilent 7890 A gas chromatography (GC; Agilent Technologies Ltd., Santa Clara, CA, USA), as shown in Supporting Information Section S6. Results and Discussion Scaling relationship We started our study by obtaining the reference values of ionization energies as the reactivity descriptors from the CRC Handbook.19 Figures 1a and 1b show the correlation of the oxygen and hydroxyl binding energies with their respective ionization energies on the following transition metal oxide surfaces: rutile-type [RuO2(110), and IrO2(110), doped with 3d metals], perovskite-type [LaBO3(001), and SrBO3(001), B = 3d metals], and spinel-type oxides [ACo2O4(110), A = 3d metals].31 We observe that the O* adsorption energy scaled linearly with the corresponding ionization energy of the doped atom on all types of oxides. Also, we found that when the valence state of the doped atom was a+ (e.g., 4+ in RuO2, 3+ in LaBO3), the O* adsorption energy correlated with the (a + 1)th ionization energy, which could rationalize the occurrence of electron transfer from the surface to the adsorbate upon oxidation. Taking doped RuO2 as an example, the valence state of doped 3d metal was considered as 4+; thus, the adsorption energy correlated with the corresponding fifth ionization energy. Notably, in specific cases, it was better to correlate the adsorption energy with the lower ionization energy because the valence state of the surface atom was lower than that of the bulk atom. However, in order to fulfill predictions faster, we still chose to use the ionization energy of the bulk atom directly instead of measuring precisely the valence state of surface atoms. Further, we found that the scaling relationships attained different slopes via OH* and O*.9,32 Figure 1 | A universal linear relationship between ionization energy and adsorption energy (a) O* and (b) OH*, for the adsorption on doped rutile, perovskite, and spinel oxides. (c) O* adsorption on RuO2 under different surface O* coverage. (d) O* adsorption on single-site alloy, M-N-C SAC, and cation-exchanged zeolite. (Detailed data of adsorption energies are displayed in Supporting Information Tables S2–S5). Download figure Download PowerPoint We investigated further the influence of coverage effects on oxide surfaces, in particular, those relevant to CO oxidation and water splitting (Surface models are illustrated in Supporting Information Figure S1).33–35 Figure 1c and Supporting Information Figure S2 showed that the scaling still existed with different intercepts for the various coverages, and the slopes were almost the same, which is attributable to the existence of similar surfaces. The varying intercepts were mainly due to the different O* coverage, and the higher the O* coverage, the smaller the O* adsorption energy. It has been shown previously that the O* chemisorption energies correlated with the third ionization energy of late transition metal oxides (RuO2, RhO2, PdO2, PtO2, OsO2, and IrO2)36; however, only six MO2(110)-type oxides were considered in the study, which limited the physical insights obtained from correlations. Therefore, we extended our investigations to metals, graphene, and zeolites. Examples of single-site alloys (3d metals/Pt), M-N-C SAC, (where M-N-C refer to 3d metals, and nitrogen-doped graphene, and SAC, refers to a single-atom catalyst), and cation-exchanged zeolite (3d metals exchanged MOR) are shown in Figure 1d. Interestingly, the O* adsorption energies still correlated with the corresponding ionization energy of all of these materials investigated. Charge transfer As mentioned above, there was an analogy between ionization and O* adsorption, both of which led to charge-transfer processes. Thus, we performed a Bader charge analysis to understand the relationship between ionization energy and adsorption energy in more detail. Figure 2a shows a linear relationship between the charge transfer and the fifth ionization energy, with the higher ionization energy corresponding to a smaller charge transfer, and hence, a weaker bond. We speculated that lower ionization energy made it more feasible to transfer electrons from the surface to the adsorbate, which typically led to stronger adsorption. More discussions can be found in Supporting Information Section S3. In principle, the asymmetry in electron distribution could also be a suitable descriptor to correlate with the binding energy. However, we presumed that the addition of DFT calculations to determine the charge distribution would lower the screening efficiency, compared with the usage of ionization energy descriptors. Figure 2 | (a) Charge transfer was used as a bridge to enable the understanding of the relationship between ionization energy and adsorption energy. (Taking doped RuO2 as an example). (b) Our two models used to explain the relationship between ionization energy and adsorption energy. Download figure Download PowerPoint Bond formation scheme To gain a deeper understanding of the relationship between ionization energy and adsorption energy, a bond formation scheme was constructed, as shown in Figure 3. Taking doped RuO2 as an example, the most relevant states determining the electronic structure of oxides are d electrons of the metals and p electrons of the oxygen. The density of states (DOS) projected onto the O 2p had almost the same shapes when RuO2 was doped with different metals from Sc to Zn, as illustrated in Supporting Information Figure S5. Thus, the effects of O 2p state in oxides were neglected at this point, shifting our focus to the d states only.37,38 As observed in Figure 3a, the d states of surface dopants show a trend of shifting down in energy when the dopant changed from Sc to Zn, and after interacting with the O* adsorbate, it gave rise to bonding and antibonding states, revealed by our crystal orbital Hamilton population (COHP) analysis data, and the integrated COHP hinted toward the bond strength.39 The results of the COHP analysis are illustrated in Figure 3c. We presumed that similar band structures and COHP curves would hold for all interactions between 3d metals and O*,37 with the only difference being the position of the antibonding state (the red part around Fermi level in Figure 3b). Indeed, as the dopant changed from Sc to Zn, the d states shifted down in energy with a consequence of the COHP moving down, which led to the filling of antibonding states (Figure 3c); while the filling of this state increased, it led to decrease in adsorption energy, comparable with the explanation derived from the d-band model.13. Figure 3 | (a) The d-projected DOS of the surface dopant of RuO2. (b) A bond formation scheme of O* on oxides to illustrate the relationship between ionization energy and adsorption energy. (c) The COHP between surface dopant and O* adsorbate. DOS, density of states; COHP, crystal orbital Hamilton population. Download figure Download PowerPoint To establish the link between the ionization energy and the adsorption energy, we introduced the crystal field theory into our analysis. When the dopant was changed from Sc to Zn, the ionization energy became higher, as shown in the left part of Figure 3b. To stabilize free cations, it was necessary to put them into a lattice to experience an electrostatic Madelung potential, during which the cations were surrounded by negatively charged oxygen anions, taking into consideration, the critical polarization that occurs whenever charges moved around in a lattice. Another effect specific to d orbitals is crystal field terms, likely to yield different crystal field stabilization energy values, displayed in Figure 3b. Finally, the effect that led to the broadening of the electronic levels due to overlap between ions, giving bands rather than localized atomic energy levels were observable. The splitting of the d band predicted by our DOS calculations was similar to that predicted by the crystal field model, which is a general feature of compounds with transition metals in octahedral coordination.40 Hence, based on the crystal field theory, the band-structure calculations were transformed from semi-empirical bands into quantitative bands. In this bond formation scheme, it was definite that higher ionization energy would ultimately lead to a downshift of the d band. After interacting with the energy level of oxygen, the lower d band led to a higher filling of the antibonding states, thus decreasing the O* adsorption energy, as observed in our calculations. Regarding other oxides like LaBO3, although the change of projected DOS of O 2p had a higher change in energy than that of doped RuO2, our bond formation model still could be briefly applied to these systems shown in Supporting Information Figures S6–S8. Descriptor application We selected oxidation of methane to methanol in transition-metal-exchanged zeolites as a specific model system to show how our established descriptor could be exploited in a fast search for new catalytic materials, due to its putative active site Cu–O–Cu structure.29,41,42 In general, the copper atoms could be exchanged with other transition metals, for example, Fe, which could change the energetics of the overall process.43 The entire reaction mechanism is illustrated in Figure 4, including the formation of the active center, CH4 activation, CH3OH formation, and CH3OH extraction. Two parts were identified to be critical for the realization of this reaction, that is, the formation of the M-O-M active center and the stability of intermediates (OH–CH3*, CH3OH*, and CH3*).44 Figure 4 | The relationship between the second ionization energy and the formation energy of the active center, M-O-M, OH–CH3*, CH3*, and CH3OH*. M-O-M, transitional heterodimetallic metal oxide active site. Download figure Download PowerPoint In our calculations, we found that all the formation energies correlated linearly with the second ionization energy of the corresponding metal, which meant that the ionization energy could serve as a reactivity descriptor for this process. Moreover, the formation of M-O-M active center was an oxidation reaction, while the other elementary steps were reduction reactions, which led to the opposite sign (negative) slopes of the correlation displayed in Figure 4, which also show that most of the M-O-M motifs (except Ag) could be generated easily in the activation process due to their negative formation energies. At the same time, the stability of all the reactive intermediates (OH–CH3*, CH3*, and CH3OH*) demonstrated a negative correlation with the second ionization energy, and OH–CH3*/CH3OH* was the most unstable intermediate. Therefore, an element with high second ionization energy, for example, Cu, Au, and Ag, was able to stabilize OH–CH3* more strongly, thus facilitating the entire reaction. This result is consistent with the theoretical and experimental observations of Cu-exchanged zeolites.44–46 Although the OH–CH3* formation energy was negative for Ag, that of the active site (Ag–O–Ag) was positive, indicating that Ag–O–Ag might be hard to form during the activation process. Note that the previously reported active site of Fe-exchanged zeolite47 was not M-O-M; thus, its catalytic property could not be evaluated by this current reaction mechanism. Accordingly, we synthesized Cu-, Pd-, Ag-, and Fe-exchanged MOR catalysts to test our predictions. The CH3OH formation rates followed the trend Cu–MOR > Pd–MOR > Fe–MOR ≈ Ag–MOR (Table 1), ideally, confirming the second ionization energies as a descriptor for these systems. More interestingly, only Pd–MOR and Cu–MOR could catalyze methane oxidation to methanol selectively (selectivity > 70%), while Fe–MOR facilitated the complete oxidation of methane to CO2, a fact that we explained by the failure to stabilize CH3* and CH3OH* in our established reaction system (Figure 4). Table 1 | Catalytic Performance of Metal-Exchanged MOR for Methane Selective Oxidation Initial rate/mmolCH3OH min−1·molM−1 Selectivity/% CH3OH CH3CH2OH CH3OCH3 C2H6 CO2 CO Cu–MOR 113.4 83.4 9.9 1.3 5.4 0.0 0.0 Pd–MOR 86.5 72.2 9.7 0.0 0.0 11.4 6.6 Fe–MOR 0.2 2.1 0.0 0.0 0.0 97.9 0.0 Ag–MOR 0.0 0.0 0.0 0.0 0.0 100 0.0 Conclusions We introduced the ionization energy of atoms and ions as a reactivity descriptor in heterogeneous catalysis. We showed that the ionization energies scaled linearly with the adsorption energies of the intermediates such as O* and OH* on a wide range of materials, including transition metal oxides, metals, graphene, and transition-metal-exchanged zeolites, thus providing a simple, predictive, and universal descriptor for their reactivity. Due to the simplicity of this descriptor, one could easily tailor the reactivity of a material, for example, by choosing another dopant with different ionization energy to tune the O* adsorption energy based on an established scaling relationship. We found that higher ionization energy led to a lower d band, which, in turn, led to an increased filling of the antibonding states, and hence, smaller adsorption energies; also, this descriptor is based mainly on electronic effect. To keep our descriptor simple, we did not place our focus too much on examining the local environment, such as geometric effects. In the future, we would make some efforts to make this descriptor more universal, for example, by taking into consideration the surface reconstruction and the nature of the chemical environment by combining all the lines into one line and extending our studies toward the elucidation of more heterogeneous catalytic reactions. Supporting Information Supporting Information is available. Conflict of Interest The authors declare no competing interests. Funding Information We acknowledge the financial supported from the National Natural Science Foundation of China (nos. 21525626, 21761132023, and 21676181), and the Program of Introducing Talents of Discipline to Universities (no. B06006) Acknowledgments We thank Prof. Jens Nørskov for stimulating discussions. References 1. Nørskov J. K.; Bligaard T.; Rossmeisl J.; Christensen C. H.Towards the Computational Design of Solid Catalysts.Nat. Chem.2009, 1, 37–46. Google Studt of over Google 3. C. the of by for Google Sabatier in Google J.; J. J.; Studt Bligaard T.; Nørskov J. the Sabatier to a of Google J. J.; Nørskov J. K.; Rossmeisl in on Google J.; J.; J. for Catalysis from Google J. J.; Rossmeisl J.; J.; Nørskov J. the Activity of for by the Surface Google J.; Studt Rossmeisl J.; Bligaard T.; Nørskov J. of Energies for on Google of A Google J. J. Studt Nørskov J. in Bond in Google J. the of the Google Nørskov J. Surface Science and and In in of Surface Science on Google J. J. of for with Google into Scaling by of Google J.; of CO2 and the from CH4 through of CO2 into Google as a Descriptor of Catalytic Activity for Oxidation over Google for a Design of Google of and CRC Google of for and a Google of the in Google J. K.; Google to the Google J.; J. for Google Chen J.; and at the A Google Chen J.; and of in the Google Chen J.; and Oxidation at the Google T.; Nørskov J. of in Activity on Google Nørskov J. K.; Studt into the Selective Oxidation of Methane to in Google J.; T.; Nørskov J. K.; Bligaard T.; for Surface with Bayesian Google by Google J. J.; Rossmeisl J.; Bligaard T.; Nørskov J. for Energies on and Google J. of Catalysis and Co. Google Rossmeisl J.; Nørskov J. of on Google J.; K.; of on an Google Gong into CO A of the on Google and A of in Google T.; J.; the Design of for Google Hamilton of Chemical in on Google Introduction to and Google J.; Oxidation of Methane by the on and Google J. of Methane to under over and Google J.; Tao of Methane to with a on the of Google Nørskov J. K.; Studt for the Selective Oxidation of Methane to Google in and for Google J. J.; in the in the Oxidation of Methane to Google K.; of Methane in Google Chen and of with to Experimental into CO2 Chemistry, Information Chemical formation thank Prof. Jens Nørskov for stimulating discussions. times

Topics & Concepts