Computer simulation of highly compressed molecular hydrogen

Abstract

The first entry in the Periodic Table, hydrogen has the simplest electronic structure, but remains one of the most intriguing elements. At low temperatures and low pressures hydrogen crystallizes as an insulating quantum molecular solid [1]. At extreme pressure electrons would no longer remain stable localized in covalent molecular bonds and hydrogen would form a dense plasma [2]. Metallization of hydrogen fluid was recently discovered in dynamic hightemperature shock-compression experiments using two-stage light-gas gun [3] at pressures above 140 GPa and temperatures near 3000 K. Absence of closed atomic electronic shells makes hydrogen extremely compressible and stable in condensed phase. Softness of intermolecular repulsion in hydrogen is very important in statistical mechanical theory. It is just what makes hydrogen different from many other substances, and therefore the well-known and useful molecular models like hard spheres or dumbbells cannot be applied to hydrogen without essential modification. Diatomic structure of H2 molecule is stable even at extreme conditions. At lower temperatures hydrogen remains diatomic up to the highest pressure reached in experiments. High-energy barriers dividing molecular and metallic phases prevent the metallization transition at least up to 250 GPa. Hydrogen solid must be compressed more then ten times [1] before chemical bonds broke and electrons delocalize. Within that range of density intermolecular and intramolecular forces, as well as intermolecular distances and equilibrium bond lengths become comparable. Notable change of vibronic frequency [1] and squeezing of chemical bonds appear at high density. These make the conventional rigid-molecule models non-applicable to highly compressed hydrogen. Without doubt, ab initio path-integral Monte Carlo simulations [4] are able, in principle, to describe the behavior of dissociating hydrogen fluid at any density from pure molecular state up to fully ionized plasma. This approach, requiring all the performance of modern supercomputers, is promising but still semi-quantitative. Our approach is based on simple atom-atom potential model for non-rigid molecules (AAP-approximation) [7], shortly outlined in the next section. It provides a satisfactory description of principal Hugoniot and double-shock experimental data up to four-time compression of liquid hydrogen and deuterium (pressure range up to 40 GPa). At higher compressions predicted pressures become significantly higher and energies essentially lower