Brief introduction of the results
Finding highly active and low-cost electrocatalysts for hydrogen evolution reactions (HER) is critical for the development of sustainable energy, but it remains a major challenge. Calculated based on density functional theory,Huang Shiping from Beijing University of Chemical Technology, Zhang Shengli from Nanjing University of Science and Technology, Guo Xiangyu from National University of Singapore, etcA novel diatomic catalyst (HDAC) with a metallic and non-metallic hybrid center embedded in G-CN is reported.
Calculation method:
All spin-polarized DFT calculations are performed in the Vienna Ab Initio Simulation Package (VASP) based on the projected augmented wave method, and the commutative correlation potential is described by the Perdew Burke Ernzerhof (PBE) functional in the Generalized Gradient Approximation (GGA). To describe the Van der Waals (VDW) interaction, the authors adopted the Grimme's DFT-D3 method and used 3 3 1 and 6 6 1 Monkhorst-pack K point grids for structural optimization and electronic property calculations, respectively. In addition, the authors set the truncation energy to 520 eV, and the convergence criteria for energy and force to 10 5 eV and 0., respectively01 evå−1。To avoid periodic layer interactions, the team set the thickness of the vacuum layer to 15. The solvation effect is described by an implicit solvation model embedded in VASPSOL, where the dielectric constant of water is 784。
Results & Discussion
Fig.1. Model mechanism, binding energy, and AIMD simulation
As shown in Figure 1a, the authors selected three possible sites as binding sites for nm atoms, including the central site (ho) in the cavity, the site binding to two adjacent n atoms (i-2n), and the binding site to three adjacent n atoms (i-3n). As shown in Figure 1b, the binding energies of nm(b, c, si, p, s) and g-cn are 5., respectively78、−5.60、−5.17、−3.81 and 190 EV, indicating that these systems are thermodynamically stable. To further verify their kinetic stability, the authors performed ab initio molecular dynamics (AIMD) simulations at 500 K. The energy fluctuations and structural deformations after 5 ps simulations (see Fig. 1c,d) confirm that these materials are thermodynamically stable.
Fig.2 Adsorption configuration, potential energy surface and PDOS
After structural optimization, the authors found that the hydrogen atoms adsorbed on top of nm and moved slightly towards the cavity (see Figure 2A) with a distance of 1 between H and nm g-cn09-1.49. Strong adsorption with H on the original G-CN (the adsorption energy at the Ho site is 439 eV), the introduction of nm can generate new adsorption sites to reduce the adsorption intensity of H. The δgh* values of nm(b,c,si,p,s) g-cn (see Figure 2b) are respectively. 25 and 105 EV, indicating that the adsorption strength of H at the non-metallic site is adjustable. Analysis of H-adsorbed PDoS (see Figure 2Cg) shows hybridization of H-S orbitals and Bc Si P-P orbitals around the Fermi level.
Fig.3. Model structure, formation energy and dissolution potential
As shown in Figure 3a, the authors considered all 3D, 4D, and 5D metal atoms except for toxic and radioactive elements (a total of 26 metal atoms) to produce a total of 130 M-nm G-CN HDACs. As shown in Figure 3b, 128 systems (except ZN-S G-CN and AG-S G-CN) have negative formation energies, indicating that these structures are thermodynamically stable. With the exception of m-p(m=zn,y) g-cn and m-s(m=mn,zn,ag) g-cn, 125 of them have positive dissolution potential, which indicates that they have excellent electrochemical stability. Based on stability criteria, the authors screened 125 HDACs to further explore their electronic properties and catalytic activity.
Fig.4 Potential energy surface and volcanic curve
As shown in Figure 4, 25 of the 125 HDACs have δgh* values at -019 to 019 EV, which includes M-B G-CN (M = ZN, PD, AU), M-C G-CN (M = TI, PD, AG, IR), M-SI G-CN (M = TI, CR, MN, CO, NI, ZN, RH, AG, IR, AU), M-P G-CN (M = FE, IR, AU) and M-S G-CN (M = TI, FE, NI, NB, RE). In addition, the δgh* values of the 11 catalysts were comparable to those of PT(111) catalysts, i.e., PD-B G-CN, TI-C G-CN, IR-C G-CN, CR-SI G-CN, MN-SI G-CN, CO-SI G-CN, RH-SI G-CN, AU-SI G-CN, IR-P G-CN, FE-S G-CN, and NI-S G-CN, respectively02, −0.03, −0.07, 0.05, −0.03, 0.07, 0.08, 0.01, 0.04, 0.08 and 008 ev。In addition, δgh* is 0The Au-Si g-cn of 01EV is located near the peak of the volcanic map and exhibits excellent catalytic activity. Ten HDACs located in the shaded region also had good catalytic activity, including PD-B G-CN, TI-C G-CN, IR-C G-CN, CR-SI G-CN, MN-SI G-CN, CO-SI G-CN, RH-SI G--CN, IR-P G-CN, FE-S G-CN and Ni-S G-CN.
Figure 5 PDOS and COHP
As shown in Figure 5, there is a significant orbital overlap between the H-S orbital and the M-D orbital (NM-P orbital) in PDOS. The valence electrons of h interact with the active valence orbital of m (nm) to form bonding and antibonding orbitals. To quantitatively describe the bonding effect, the authors performed a COHP analysis and found that the ICOHP of M-H in PD-B G-CN, TI-C G-CN, IR-C G-CN, CR-SI G-CN, MN-SI G-CN, CO-SI G-CN, RH-SI G-CN, IR-P G-CN, and FE-S G-CN were -2., respectively67，-0.34，-1.57，-1.70，-1.77，-1.61，-1.47，-1.35 and -115 ev。The ICOHP of NM-H in PD-B G-CN, CR-SI G-CN, MN-SI G-CN, CO-SI G-CN, RH-SI G-CN and Ni-S G-CN were -0., respectively17、-0.68、-0.71、-0.67、-0.93 and -300 ev。
Fig. 6 Correlation coefficients, feature importance, RMSE and R2 for machine learning, DFT computation, and δgh* for machine learning**
As shown in Figure 6A, EM, ENM, DM-NM, DNM-H, H-M-NM, P, and HXF were used as the final features for screening ML models. As shown in Figure 6b, the importance of features in descending order is: hxf(0.).264）>θh-m-nm（0.240）>dnm-h（0.137）>εp（0.114）>dm-nm（0.098）>em（0.082）>enm（0.064）。These results show that the hydrogen affinity of metal atoms and the spatial configuration of hydrogen play a decisive role, and the coordination environment of m-nm synergistically influences the catalytic activity. HXF can reflect the electron gain loss capacity due to the strong correlation with NVA-M (see Figure 6A), which affects the hydrogen binding strength. The two spatial coordinates of H are determined by the interaction between M and NM, which includes H-M-NM and DNM-H, and strong interactions weaken the adsorption of H. As shown in Figure 6C, the authors compared the stability and accuracy of random forest regression (RFR), gradient enhancement regression (GBR), Gaussian process regression (GPR), support vector regression (SVR), and K-nearest neighbor regression (KNR). where RFR has the highest R2 (0.).89) and the smallest rmse (009), indicating that the RFR model is suitable for HER activity**. The ΔGH* values of RF** are shown in Figure 6D, and the ΔGH* values of M(sc, Ti, V, Cr, Fe, Zn, Ag, Ta, Re, IR)-SE G-CN and M(SC, TI, MN, FE, CO, Y, NB, MO, PD, TA, and IR)-TE G-CN are at -020 to 020 ev。Then, the authors performed DFT calculations on these 52 HDACs and confirmed that M(Fe,Re,Ir)-SE G-CN and M(SC,TI,MN,FE,NB,MO,IR,NB)-TE G-CN had good HER catalytic performance.
Conclusions and prospects
Studies have shown that the introduction of non-metallic atoms (B, C, SI, P, and S) in the vicinity of metal sites can achieve a unique charge transfer between them, which provides a diatomic center with a very different catalytic activity than its monoatomic counterpart. Among the 130 HDACs, PD-B, TI-C, IR-C, CR-SI, MN-SI, CO-SI, RH-SI, AU-SI, IR-P, FE-S, and Ni-S pairs are high-performance electrocatalysts with near-ideal adsorption strength for protons. Machine learning analysis can directly identify the key features that affect catalytic activity and establish a framework for rapid screening of unknown chemical spaces in HDACs. This work opens up a new avenue for the design and development of potential HER catalysts.
Bibliographic information
lihong zhang et.al, hybrid double atom catalysts for hydrogen evolution reaction: a sweet marriage of metal and nonmetal adv. energy mater. 2023
ht-tps: