| The depletion of fossil energy has made it an urgent requirement for the development of clean,environmentally friendly and renewable energy sources.Hydrogen energy is a very popular alternative energy source because of its high energy density,abundant reserves,and zero emission behavior.However,the storage of hydrogen is a bottleneck restricting the development of hydrogen energy.Magnesium hydride,as a solid hydrogen storage material with high hydrogen storage capacity,provides ideas for solving the hydrogen storage problem,but it is difficult to put it into practical use at present due to its poor thermodynamic properties and slow hydrogen uptake and release kinetics.Nanosizing and catalysis are two feasible ways to improve the hydrogen storage performance of magnesium hydrides.The first-principles method is difficult to handle the systems with too many atoms.To study the structural properties and dynamic properties of large-sized magnesium hydride clusters,as well as studying the catalytic effect of Ni doping in the clusters,we trained a neural network potential for Mg-H system,and used it to calculate the properties of magnesium hydride clusters with different sizes,compositions and temperatures,and an algorithm based on the joint of first-principles method and neural network potential was developed to simulate Ni-doped clusters.The main research processes and conclusions are briefly described as follows:A machine-learning(ML)interatomic potential for Mg-H system based on Behler-Parrinello approach was developed.Multiple sampling strategies were combined to obtain training samples that contain a variety of local atomic environments,including a global minimum searching algorithm coalescence kick(CK),ab initio molecular dynamics(MD)simulation method,and an adaptive sampline scheme.The calculations of the energies and forces of the crystal,surface,and cluster structures using the trained neural network at any Mg:H ratio matches the DFT results.The calculation of bulk properties and phonon dispersions,the simulation of H2 molecules,and the calculation of the potential energy surface(PES)for H2 dissociative adsorption on Mg(0001)surfaces confirm the accuracy of our neural network in calculating the physical and chemical properties of the Mg-H system.Through molecular dynamics simulations based on the ML potential,the structure of magnesium hydride clusters and the diffusion properties of hydrogen,as well as the role of size,composition,and temperature were studied.We find that the atoms in the MgnH2n clusters are mainly exist in a disordered way above 300 K.For MgnHm clusters,Mg/MgHx phase separation occurs when m<2n,and for a cluster with a diameter of about 4 nm,the Mg part of the cluster forms a hexagonal close-packed(hcp)nanocrystalline structure below 600 K.The results show that low temperature and large cluster size are conducive to the formation of ordered Mg nanocrystals.Also,the calculated diffusion coefficients reproduce the experimental values,and confirm an Arrhenius type temperature dependence in the range of 400 K to 700 K.The diffusion performance of hydrogen is proved to be independent of the hydrogen content of the clusters,that is,not affected by the hydrogen pressure of the system,which provides a reasonable explanation for the experimental phenomenon.Refer to the QM/MM method,a QM/NN calculation method based on the first-principles calculation and neural network potential was proposed to study the hydrogen release performance of Ni-doped magnesium hydride clusters.The spontaneous formation and desorption of hydrogen molecules at the Ni atom sites were observed.This proves that the Ni doping has a significant effect on improving the dehydrogenation kinetics of magnesium hydride nanocrystals.Molecular dynamics trajectories show that dehydrogenation occurs on Ni atoms,and a five-coordinated Ni structure is an intermediate in the dehydrogenation process.The presence of Ni provides a channel for the hydrogen release reaction,and the H atoms in the cluster can move to the direction of Ni through diffusion,which is beneficial to the improvement of the overall hydrogen release of the cluster.Ni weaken the Coulomb repulsion among surrounding H atoms,and shows anti-bonding interaction with H atoms at energies near the Fermi level,which is the internal reason that Ni activates the surrounding H atoms to promote their binding and desorption.Compared with the catalytic effect of monoatomic Ni,the effect of diatomic doping on the process of catalytic dehydrogenation reaction has little change,the mechanism of dehydrogenation is still the same.When three Ni atoms doping adjacently,because of the more desorption sites at the beginning of dehydrogenation,the reaction rate is significantly accelerated,but with the removal of a large number of hydrogen atoms,the vicinity of Ni atoms begins to protrude from the clusters,reducing the contact area with the clusters,and relatively reducing the ability of hydrogen to diffuse toward Ni.This inhibits the occurrence of further dehydrogenation. |
PDF Full Text Request