Relating DFT MD simulations of high-pressure hydrogen to experiment.

There have been many experimental and theoretical studies of high
pressure hydrogen in recent years, but the linkage between them has
become increasingly weak. e.g. the controversial 2017 observation
of metallic hydrogen verified a 1935 theoretical
prediction, evidenced by comparing reflectivity based on a 1927 model and 1974
stability predictions. Theorists must be held culpable: calculations
of imaginary numbers, be they dielectric constants or phonon
frequencies, do not and have not make
experimental verification easy. Much calculation work avoids
presenting results in a form readily comparable with experiment.

I focus on the uncomfortable middle ground between theory and
experiment, in particular the timidity of theorists to challenge the heuritics used in interpreting data.

In principle, molecular dynamics provides the most direct way to simulate hydrogen fully anharmonically at finite temperature, essential for molecular-rotor phases I, IV and the liquid. But MD requires approximations constrained by computing time. Three key issues are the electronic exchange correlation, the nuclear dynamics,
and extraction of spectroscopic data.

We present large amounts of MD and PIMD data using various functionals. Our MD samples different crystal structures within picoseconds, effectively determining a PT phase diagram. For spectroscopic modes of known character, we extract frequencies and linewidths at various temperatures and
pressure. The key conclusion is that current DFT functionals give linewidths and qualitative peaks, but are not accurate enough to compare frequencies quantitatively. Worse, those functionals which agree well with QMC calculations perform very poorly in comparison with experiment. For metallisation, theoretical crystal
structures, pressures and even the character of the transition are strongly functional dependent.

Finally, we discuss extracting the Raman spectrum from quantum rotors using MD.