Dissociation pathways and binding energies of (LiH)nLi+ and (LiH)nLi+3 clusters

Incorporating Electronic Information into Machine Learning Potential Energy Surfaces via Approaching the Ground-State Electronic Energy as a Function of Atom-Based Electronic Populations

Machine learning (ML) approximations to density functional theory (DFT) potential energy surfaces (PESs) are showing great promise for reducing the computational cost of accurate molecular simulations, but at present, they are not applicable to varying electronic states, and in particular, they are not well suited for molecular systems in which the local electronic structure is sensitive to the medium to long-range electronic environment. With this issue as the focal point, we present a new machine learning approach called "BpopNN" for obtaining efficient approximations to DFT PESs. Conceptually, the methodology is based on approaching the true DFT energy as a function of electron populations on atoms; in practice, this is realized with available density functionals and constrained DFT (CDFT). The new approach creates approximations to this function with neural networks. These approximations thereby incorporate electronic information naturally into a ML approach, and optimizing the model energy with respect to populations allows the electronic terms to self-consistently adapt to the environment, as in DFT. We confirm the effectiveness of this approach with a variety of calculations on Li_nH_n clusters.

Theoretical Study on the Small Clusters of LiH, NaH, BeH2, and MgH2

High-level ab initio molecular orbital theory is used to calculate the geometries, vibrational frequencies, atomic charges, and binding energies of the small clusters (LiH)(n), (NaH)(n), (BeH(2))(n), and (MgH(2))(n) (n = 1-4). For (LiH)(n) and (NaH)(n), there are planar cyclic structures when n = 2, 3. We have found the cubic structure T(d) in addition to the planar cyclic D(4)(h) when n = 4. The D(4)(h) is less stable than the T(d) geometry. For (BeH(2))(n) and (MgH(2))(n), when n = 3, there are three kinds of structures: chain C(2)(v), planar cyclic D(3)(h), and hat-like C(2)(v). The C(2)(v) geometry is more stable than the others. When n = 4, there are four kinds of structures: chain D(2)(h), cubic T(d), string-like C(2), and cubic transformation C(1). The most stable compounds in the families of (LiH)(n), (NaH)(n), (BeH(2))(n), and (MgH(2))(n) are cubic T(d), cubic T(d), chain D(2)(h), and string-like C(2) geometries, respectively, when n = 4. Calculated binding energies range from -24 to -37 kcal/mol for (LiH)(n) and --19 to -30 kcal/mol for (NaH)(n), (BeH(2))(n), and (MgH(2))(n). The hydrogen atoms in hydride clusters always have negative charges. The atomic charges of planar cyclic structures are weaker than those of cubic structures, and there is a tendency of reducing along with the increase of the cluster size. The vibrational frequencies of planar cyclic structures have consistent tendency, too. It indicates that the bond distance increases with the ionic character of the bond.