Governing factors of supports of ammonia synthesis in an electric field found using density functional theory Free

About 160 × 10⁶ tons of ammonia, an important chemical feedstock, is synthesized worldwide each year to produce synthetic fertilizers, fibers, and other nitrogen-containing compounds. Recently, ammonia has also been attracting attention as an excellent hydrogen carrier because of its high hydrogen content (17.6 wt. %) and characteristics that are amenable to transport and storage.¹ 

At present, most of ammonia is produced using the Haber–Bosch process with an iron-based catalyst.^2,3 Because of its thermodynamic and kinetic limitations, this process requires high temperatures and high pressures (20–40 MPa, 673–873 K).⁴ Therefore, only large-scale plants can use this process efficiently. If catalytic ammonia synthesis could be achieved under milder conditions with high reaction rates, then a novel, smaller-scale, on-site, and on-demand ammonia production method would be feasible.

Many researchers have tried to promote catalytic ammonia synthesis. Aika reported that Ru-based catalysts with alkaline promoters can synthesize ammonia at lower temperatures and pressures than Fe-based catalysts.^5,6 Recently, new catalyst supports with strong electron-donating properties such as rare earth oxides^7,8 and electrides^9–12 have been developed. They showed high ammonia synthesis rates. Furthermore, various ammonia synthesis methods using hydride support,^13,14 plasma,^15–19 an external direct current (DC) electric field,²⁰ electrolysis,^21,22 and photocatalysis²³ have been reported. However, ammonia synthesis rates achieved in the low-temperature region using these methods are still too low to use in practical applications.

We reported earlier that various catalytic reactions can be promoted by an application of an electric field to the catalyst bed.^24,25 The results indicated surface proton hopping as an important factor for the activation of catalytic reactions in the electric field.^25–28 By adopting this novel approach for ammonia synthesis, we demonstrated that Cs/Ru/SrZrO₃ exhibited high catalytic activity even under mild conditions. Using this catalyst, we achieved much higher catalytic activity (30 mmol g⁻¹ h⁻¹, 623 K, 0.9 MPa) than that achieved with conventional catalysts.²⁶ For this process, N₂ dissociation through N₂H (associative mechanism) is an important step. The trend for active metal differs from the conventional order without the electric field.^26,29,30 However, the support factors determining the ammonia synthesis rate in the electric field have not been investigated. Herein, we investigated the effects of dopants on SrZrO₃ (SZO) for ammonia synthesis in an electric field. The results show that Ba-doped and Ca-doped SZO (SBZO and SCZO) exhibited high ammonia synthesis activity, even at 423 K. Density functional theory (DFT) calculations revealed that Ba doping and Ca doping promote proton and electron donation from the support to Ru at the interface. These findings are important for the development of a highly efficient ammonia synthesis process using an electric field.

We prepared SrZrO₃ (SZO), Sr_0.875Ba_0.125ZrO₃ (SBZO), Sr_0.875Ca_0.125ZrO₃ (SCZO), SrZr_0.875Y_0.125O₃ (SZYO), and SrZr_0.875Al_0.125O₃ (SZAO) using a complex polymerization method.²⁹ First, citric acid monohydrate and ethylene glycol were dissolved in distilled water. Subsequently, a stoichiometric mixture of Sr(NO₃)₂, ZrO(NO₃)₂·2H₂O, Ba(NO₃)₂, Ca(NO₃)₂·4H₂O, Y(NO₃)₃·6H₂O, or Al(NO₃)₃·9H₂O was dissolved in the solution. The obtained solutions were dried and crushed into particles. The particles were then calcined twice at 673 K for 2 h and at 1123 K for 10 h.

Catalysts of 5 wt. % Ru/support (SZO, SBZO, SCZO, SZYO, and SZAO) were prepared using an impregnation method. First, Ru(acac)₃ was dissolved into acetone. Support powders were added to the Ru solution and were then stirred at room temperature for 2 h. Subsequently, solvents of the slurry were evaporated. The dried powders were reduced at 723 K for 2 h under Ar:H₂ = 1:1 (100 SCCM total flow). Finally, the catalysts were shaped into 355–500 µm grains.

The crystalline structures of supports were characterized using an X-ray diffractometer (Smart Lab III; Rigaku Corp.) with Cu Kα radiation operating at 40 kV and 40 mA. After activity tests, Ru particle sizes were observed using a field emission transmission electron microscope (FE-TEM) (JEM-2100F; Hitachi Ltd.) operating at 200 kV acceleration voltage. The samples were dropped on a collodion (NP-C₁₅; Okenshoji Co., Ltd.) after ultrasonic wave dispersion in ethanol. Energy dispersive X-ray spectrometry (STEM-EDX, HF-2200; Hitachi Ltd.) was used for elemental mapping.

X-ray photoelectron spectroscopy (XPS) measurements (VersaProbe2; Ulvac-Phi, Inc.) were performed with an Al Kα X-ray source. The binding energies were calibrated using C1s peaks at 284.8 eV. Pretreatment of the samples was conducted under the same conditions as those used for activity tests.

Ammonia synthesis rates were examined using a fixed flow type quartz reactor with 0.1 g of catalysts under the following conditions (0.1 MPa, H₂:N₂ = 3:1, total gas flow rate 240 SCCM). We have already confirmed that these conditions are suitable for determining kinetic parameters considering diffusion and kinetics. Two 2-mm-diameter stainless rods were inserted into the catalyst bed. Then, DC constant current (6 mA) was imposed using a power supply device, as shown in Fig. S1 of the supplementary material. A digital phosphor oscilloscope (TDS 2001C; Tektronix, Inc.) was used for the detection of the applied current and response voltage. The catalyst bed temperature was then increased with Joule heating by about 30 K. Therefore, the catalyst bed temperature was detected directly by attaching a thermocouple on the catalyst bed. Hence, the heat effects were confirmed and were negligibly small in this case. The catalyst was reduced at 723 K for 2 h under an ammonia synthesis atmosphere before the activity tests. The synthesized ammonia was trapped using iced distilled water. It was analyzed using ion chromatography (IC-2001; Tosoh Corp.).

Density functional theory (DFT) calculations were performed using the Vienna ab initio simulation package (VASP) 5.4.1.^31–34 We used projector-augmented wave (PAW) pseudopotentials including gradient-corrected reparameterized Perdew–Burke–Ernzerhof (RPBE) for all calculations.^35,36 Electron Kohn–Sham orbitals were expanded using a kinetic energy cutoff of 400 eV. Reciprocal space integration was done with numerical integration using a k-point grid: 3 × 3 × 3 k-points were set with the Monkhorst–Pack scheme for bulk calculations. The gamma point was used for surface calculations. The Gaussian smearing method was used for the electron occupation near the Fermi level. The electronic charge was evaluated using the Bader analysis.^37–40 

Bulk structures for SZO were optimized. The calculated lattice constants were used to form the SZO support part. The support was constructed as four layers of SZO with the top Sr-O (001) surface. Note that this surface termination with Sr was more stable than Zr-O (001) surface termination.⁴¹ A 4 × 4 supercell was chosen. For the metal loading catalyst models, some researchers placed metal nanoparticles on the support. Such a model is suitable for the elucidation of “nanosize effect” because the specific electronic states of the nanoparticles can be calculated. On the other hand, the structure of such particles will be changed after the optimization by DFT calculations.^42,43 In this paper, we would like to prevent such a specific phenomenon. The supported Ru particle has (1011) and (0001) facets (Fig. 1) by TEM observation. Hence, we focused on the (1011) and (0001) facets, and for avoiding the structural change of the Ru metal particle, we designed a Ru rod catalyst having a periodic structure with the (1011) and (0001) facets. During the geometry optimization of the Ru/SZO slab, the Ru rod was not forcedly distorted by relaxing the lattice constant while the supercell volume was constant, and the Ru-Ru bond length was slightly different from that of the Ru bulk calculation models in which Ru was not loaded over SZO. The errors were smaller than 7.0% even at the Ru-SZO interface, where the Ru structure was distorted by the Ru-O interactions. The system was expressed by repeated slabs separated by a 20 Å vacuum layer. The bottom two layers of the SZO were fixed during calculations. The N₂, H₂, and N₂H molecules were placed in a 10 × 10 × 10 Å cubic box. Energies were evaluated with gamma point calculations. The calculation models were visualized using BIOVIA Materials Studio.

Dopant effects on the ammonia synthesis rate in the electric field were examined closely. Figure 2 shows the ammonia synthesis rate achieved at 423 K during application of the electric field over each catalyst (6 mA, about 0.4 kV). The synthesis rate at 423 K without the electric field was almost negligible, as shown in Table S2 of the supplementary material. The activity trend in the electric field is depicted exactly in Fig. 2. The response voltages and mean particle sizes of Ru on each support are presented in Table I. According to Fig. 2 and Table I, Ba and Ca doped SZO (SBZO and SCZO) exhibited much higher activity than that of pristine SZO. The response voltages and mean particle sizes of Ru differed slightly among all catalysts. The results indicated that applying energy by the electric field and physical properties of loading Ru did not contribute to high activities of Ru/SBZO and Ru/SCZO. The morphologies of the respective supports were investigated using XRD measurements (Fig. S2 of the supplementary material). The results demonstrated that all supports maintained SZO structures in this doping amount range. Therefore, the chemical factors which scale the ammonia synthesis rate in the electric field were investigated using DFT calculations.

We performed DFT calculations to map out the reaction path at the Ru–support interface and to ascertain the dopant effects on the reaction energy. First, the N₂ dissociation mechanism at the Ru–SZO interface was regarded, assuming the rate-determining step in the electric field. Figure 3 portrays the optimized structures of the intermediate states. Table II shows the reaction energies (ΔE) defined as

where E(N_xH_y/Catalyst) represents the energy of the system with adsorbates, E(Catalyst) denotes the energy without adsorbates, and E(N₂ or H₂) stands for the energy of gaseous molecules.

The DFT calculations suggested that gaseous N₂ adsorbs over Ru as end-on N₂ (NN^*) at first, where the atom with asterisk (^*) denotes the adsorbed species. Reportedly, the adsorbed N₂ dissociates via the “dissociative mechanism,” without the electric field as portrayed in Fig. 4.^30,44,45 DFT calculations show that N₂ direct dissociation (^*NN^* → ^*N + ^*N) is an exothermic reaction even at the Ru-SZO interface, meaning that the “dissociative mechanism” proceeds. With the electric field, the potential of the proton over SZO would be changed and hydrogenation of N₂ using the proton would be activated. We detected NH₄⁺ using in situ DRIFTS (diffuse reflectance for infrared Fourier transform spectroscopy) only when the electric field was imposed.²⁶ Presumably, NH₄⁺ was formed from NH₃ and proton, suggesting that the reaction with the proton was activated in the electric field. This assumption was elucidated using D₂ isotope effects.²⁹ These findings suggest that the protons are activated in the electric field and that they play an important role in ammonia synthesis. We also found that the N₂ dissociation was activated by the existence of H₂ in the electric field. We assumed that N₂H is formed as an intermediate.^26,29 Consequently, we inferred that the main reaction path would change from N₂ direct dissociation (dissociative mechanism) to an “associative mechanism” by which N₂ dissociation proceeds via ^*NNH or ^*NN^*H. Figure 5 depicts a potential energy diagram related to this mechanism. The results show that once ^*NNH is formed, sequential hydrogenations (^*NNH → ^*NNH₂ → NH₃ + ^*N) proceed easily as presented in Fig. 5(a). In contrast, ^*NN^*H dissociates preferably into ^*NH + ^*N as presented in Fig. 5(b). The potential energy difference between ^*NN and ^*NNH (0.12 eV) is less than that between ^*NN^* and ^*NN^*H (0.19). In addition, ^*NN is more stable than ^*NN^*. Therefore, it is assumed that the ^*NNH formation proceeds preferably for the first hydrogenation. The rate determining step in the reaction path is ^*NNH formation.

The potential energy diagrams for the “associative mechanism” suggest that the decrement of ^*NNH formation energy will be the effects of Ba and Ca addition in the electric field. Therefore, the dopant effects on the N₂H formation from N₂ over Ru and the proton over supports (ΔE_{N2H formation}) were calculated as

where E(N₂H over Ru) is the energy of ^*NNH models (depicted in Fig. 3). Also, E(N₂ over Ru, H over support) is the energy of models depicted in Fig. 6. Doped models were expressed by substitution of Sr (Zr) at A1 or A2 (B1 or B2). The calculation results for dopant effects are presented in Table III.

The results demonstrate that the experimental ammonia synthesis rate in the electric field has good correlation with the calculated ΔE_{N2H formation} (Fig. 7). As ΔE_{N2H formation} decreases, the ammonia synthesis rate increases. This trend corresponds to the trend we considered using metal surface models.³⁰ That study found that the N₂H formation became advantageous by substitution of Sr with Ba and Ca. We considered the governing factors for ΔE_{N2H formation} over each support. Hydrogen desorbs from the support and N₂ over Ru accepts it during N₂H formation at the interface. Then, the N₂H adsorption energy and hydrogen adsorption energy were regarded as factors.

Figure 8 shows the correlation among ΔE_{N2H formation}, the N₂H adsorption energy (ΔE_N2H adsorption), and the hydrogen adsorption energy (ΔE_{H adsorption}). The N₂H adsorption energy was calculated using the gaseous N₂H energy. Smaller circles in Fig. 8 represent smaller N₂H formation energy. Circles become smaller as the N₂H adsorption energy increases, and the hydrogen adsorption energy decreases. The results suggest that the ability of the support to supply protons is necessary for ammonia synthesis in the electric field. The N₂H stability over Ru is also important, but the role of the support in this factor remains unclear. Therefore, Ru charge was considered for the factors which decide ΔE_{N2H adsorption}. Here, the Ru charge was represented by taking the average of Ru atom Bader charges for the Ru catalyst part. The results show that Ru was charged negatively by Ba and Ca addition (Table III). Reportedly, the lattice strain of perovskites influences oxide electronic structures.⁴⁶ Therefore, the lattice strain was regarded as the controlling factor for Ru charges. Table S3 of the supplementary material shows dopant effects on O and Ru charges before and after structural optimization after doping the cation into the optimized Ru/SZO model. The results show charge transfer by structural optimization. Using XPS, Ru charges were investigated experimentally (Fig. S4 of the supplementary material). No marked shift was observed for Ru 3d_5/2 spectra. Presumably, the importance of electron donation was limited at the Ru–support interface close to local strain derived from doping. The correlation between ΔE_{N2H adsorption} and the Ru average charge was investigated, as shown in Fig. 9. The results revealed that the N₂H adsorption becomes advantageous as the electron is supplied to Ru from the support. These results indicate that the electron-donating property plays a key role in providing N₂H stability over Ru.

In summary, the coexistence of proton donation and electron donation from catalyst supports is important for ammonia synthesis in the electric field. Electron donation is important for N₂H stability. For these reasons, amphoteric substances such as SZO exhibit high catalytic activity in an electric field.

Ru catalysts supported on SZO doped with Ba and Ca showed high ammonia synthesis rates in an electric field. To elucidate the dopant effects, N₂ dissociation via the “associative mechanism” at the Ru–support interface was considered using DFT calculations. The results revealed that sequential hydrogenation of N₂ (^*NN → ^*NNH → ^*NNH₂ → NH₃ + ^*N) proceeds preferably. The rate determining step in this reaction path is the first hydrogenation (^*NN → ^*NNH). Therefore, dopant effects on ^*NNH formation were considered. The results revealed that the ^*NNH formation energy decreased by Ba and Ca addition. Finally, the factors which determine the ^*NNH formation energy were investigated. The results showed the importance of the proton-donating ability of supports and N₂H stability over Ru. Electronic state analysis revealed that electron donation from supports to the supported Ru (basicity) affected the stability of N₂H over Ru. The coexistence of electron donation and proton donation from supports is necessary for a high ammonia synthesis rate in the electric field. Such a trend is specific in the reaction with the electric field, different from conventional heterogeneous catalysts.

See supplementary material for the apparatus for ammonia synthesis, XRD patterns, and FE-TEM images of catalysts, the bond length on the Ru/SZO slab model, experimental data of the ammonia synthesis rate with and without the electric field at a low temperature region, and dopant effects on O and Ru charges before and after optimization.

This work was supported by JST-Mirai (Grant No. JPMJMI17E5) and JSPS-KAKENHI (Grant No. 17H01344). This work was partly achieved through the use of supercomputer system at the information initiative center, Hokkaido University, Sapporo, Japan.