Van-der-Waals exchange-correlation functionals and their high pressure and warm dense matter applications

This paper evaluates various exchange-correlation functionals, identifying r2SCAN as the most accurate for reproducing hydrogen molecular properties, to determine their suitability for modeling the molecular-to-metal transition in warm dense hydrogen under high pressure.

The Gist

Imagine you are trying to predict how a crowd of people behaves in a giant, pressurized room. Sometimes they stand alone as individuals, sometimes they hold hands in pairs (molecules), and sometimes the pressure gets so high that they let go and become a chaotic, fluid mass (metal).

This paper is a scientific "taste test" to figure out which set of rules (called exchange-correlation functionals) best predicts how these hydrogen "people" behave when they are squeezed together in Warm Dense Matter—a state of matter found inside giant planets like Saturn or in fusion experiments.

Here is the breakdown of their investigation using simple analogies:

1. The Problem: Choosing the Right Rulebook

Scientists use computer simulations to model this matter. To do this, they need a "rulebook" (a mathematical formula) to describe how electrons interact. There are many rulebooks available:

• PBE: The standard, reliable rulebook everyone uses.
• vdW (Van-der-Waals) functionals: Specialized rulebooks designed to account for very weak, long-distance "ghostly" attractions between neutral objects (like how a balloon sticks to a wall after rubbing it on your hair).
• r2SCAN & HSE06: Newer, more complex rulebooks.

The big question was: Do the specialized "ghostly attraction" rulebooks (vdW) actually make the predictions for high-pressure hydrogen better?

2. The Lab Tests: Checking the Basics

Before simulating the whole crowd, the authors tested the rulebooks on simple, isolated scenarios to see which one was most accurate.

• Test A: The Handshake (Bond Length & Energy)
  They looked at two hydrogen atoms holding hands.

  • Result: The standard PBE rulebook said the handshake was a bit too loose (atoms too far apart). The vdW rulebooks said the handshake was too tight (atoms too close).
  • The Winner: The r2SCAN rulebook got the handshake distance and strength almost perfect.
  • The Twist: The HSE06 rulebook was also very good, but r2SCAN was the champion for this specific test.

• Test B: The "Ghostly" Hug (Interaction between two molecules)
  They placed two hydrogen molecules near each other to see if they felt that weak "ghostly" attraction.

  • Result: Here, the vdW rulebooks failed. They predicted the molecules would hug too tightly and too far apart.
  • The Winner: Surprisingly, the standard PBE and the HSE06 rulebooks predicted the "ghostly hug" much better than the specialized vdW rulebooks.
  • The Loser: The r2SCAN rulebook completely missed the hug; it didn't see the attraction at all.

3. The Big Squeeze: Simulating the High-Pressure Room

Now they simulated the whole crowd under high pressure and heat (Warm Dense Matter).

• The "Ghostly" Effect is Tiny: The authors found that the "ghostly" attraction between molecules is incredibly weak (like a whisper) compared to the energy required to break the molecules apart (like a shout).
• The Conclusion: Because the "ghostly" force is so weak, it doesn't actually change how the crowd behaves in the high-pressure room. Whether you use the vdW rulebook or the standard PBE rulebook, the resulting "crowd behavior" (pressure, density, structure) looks almost identical.
• Why did vdW rulebooks change the results before? The authors realized that previous studies using vdW rulebooks predicted a different "phase transition" (when the crowd lets go of their hands) not because they saw the "ghostly hug" better, but because they overestimated how strong the handshake was. They thought the molecules held on too tightly, so it took more pressure to break them apart.

4. The Final Verdict

The paper concludes that for warm dense hydrogen:

1. The "Ghostly" forces don't matter much: The weak attraction between molecules is too small to change the big picture of how the material behaves under extreme pressure.
2. The Handshake matters most: The most important thing is getting the strength of the molecular bond right.
3. The Best Rulebook: Since r2SCAN got the handshake (bond length and energy) perfect, even though it failed to see the "ghostly hug," the authors suggest it is the best choice for simulating this material. It gets the most critical part right, whereas the specialized vdW rulebooks get the handshake wrong.

In short: The authors tried to find a special tool to measure a tiny whisper (vdW forces) in a noisy room. They found that the whisper is too quiet to matter. Instead, they realized the standard tools were actually better at hearing the loud shout (the molecular bond), and one new tool (r2SCAN) heard the shout perfectly. Therefore, they recommend using the tool that hears the shout best, rather than the one designed to hear the whisper.

Technical Summary

Technical Summary: Van-der-Waals Exchange-Correlation Functionals and Their High Pressure and Warm Dense Matter Applications

Problem Statement
Recent simulations of dense hydrogen have relied heavily on Density Functional Theory (DFT) to model the molecular-to-metal transition and the liquid-liquid phase transition (LLPT). While advanced exchange-correlation (xc) functionals, particularly those incorporating non-local van-der-Waals (vdW) corrections, have been employed to improve agreement with experimental LLPT curves, the physical justification for their superiority in warm dense matter (WDM) regimes remains debated. The energy scale of vdW interactions (few meV) is orders of magnitude smaller than molecular dissociation energies (~4.5 eV) or ionization energies (13.6 eV). Consequently, it is unclear whether the inclusion of vdW corrections significantly alters the hydrogen equation of state (EOS) or if observed shifts in the LLPT line are merely artifacts of how different functionals describe fundamental molecular properties like bond lengths and dissociation energies.

Methodology
The authors conducted a systematic benchmarking study comparing standard and advanced xc functionals using DFT and DFT-Molecular Dynamics (DFT-MD) simulations via the Vienna Ab-initio Simulation Package (VASP). The study utilized the Mermin formulation for thermal DFT and bare Coulomb pseudopotentials.

The functionals tested included:

• Standard: LDA, PBE.
• Non-local vdW: VDW-DF1, VDW-DF2, VDW-DF3 (optPBE-vdW).
• Meta-GGA: r2SCAN.
• Hybrid: HSE06.

To establish high-accuracy benchmarks, the authors employed:

1. Quantum Monte Carlo (QMC): Specifically initiator full configuration interaction QMC (i-FCIQMC) using the HANDE-QMC software to calculate near-exact interaction energies for isolated hydrogen molecule dimers.
2. Coupled Cluster (CCSD(T)): Used to generate complete basis set (CBS) corrections for the QMC results.

The investigation covered three distinct regimes:

1. Idealized Vacuum: Isolated hydrogen molecules (bond length, dissociation energy) and isolated molecule-molecule interactions (vdW energy minima) across various geometries.
2. Warm Dense/High Pressure: DFT-MD simulations of hydrogen fluids across a range of densities ($r_s$) and temperatures (100 K to 1500 K), spanning the molecular, interacting molecular, and pressure-dissociation regimes.
3. Structural and Electronic Analysis: Calculation of pair correlation functions, static and dynamic ion structure factors, effective bond lengths, and electronic density of states (DOS).

Key Contributions and Results

• Molecular Properties:

  • Bond Lengths: The r2SCAN functional provided the best reproduction of experimental bond lengths (deviation of 0.001 Å), followed closely by HSE06. Standard PBE and vdW functionals showed larger deviations (PBE overestimated by 0.1 Å; vdW functionals underestimated or slightly overestimated).
  • Dissociation Energies: PBE underestimated the dissociation energy by 0.21 eV compared to experiment. The vdW functionals overestimated it by a similar magnitude (0.25 eV). r2SCAN yielded the best agreement with a deviation of only ~0.07 eV.
  • Conclusion on Functionals: In terms of fundamental molecular properties, r2SCAN and HSE06 outperformed PBE and vdW functionals.

• Van-der-Waals Interactions:

  • In idealized two-molecule interaction scenarios, the vdW energy minimum (depth ~5 meV) was best reproduced by HSE06, which was nearly indistinguishable from QMC results. PBE was a close second.
  • Crucially, the vdW functionals (VDW-DF1, VDW-DF3) predicted minima that were too deep and located at too large a distance compared to QMC.
  • The r2SCAN functional failed to reproduce a vdW energy minimum entirely, lacking a proper description of these long-range effects. Consequently, r2SCAN was dropped from further WDM investigations in this study.

• Equation of State (EOS) and LLPT:

  • The study found that vdW functionals predict higher pressures in the LLPT region compared to PBE. The authors attribute this shift not to the direct inclusion of vdW forces, but to the fact that vdW functionals predict higher dissociation energies.
  • Higher dissociation energies require more pressure work to break molecules, thereby shifting the molecular-to-metal transition to higher pressures.
  • The energy scale of vdW interactions is negligible compared to the thermal energy at LLPT temperatures (e.g., 1500 K $\approx$ 130 meV vs. vdW energy $\sim$ 10 meV).

• Structural and Dynamic Properties:

  • Pair Correlation & Structure Factors: In the molecular regime, differences in the fine structure of molecular peaks were observed, correlating with the different bond lengths predicted by each functional. However, in the intermediate-to-long range (1–10 Å), the pair correlation functions and static structure factors showed negligible differences between functionals.
  • Dynamic Structure: The dynamic ion structure factors, which probe collective modes and sound speeds, showed no significant differences between PBE and vdW functionals. The collective behavior is dominated by mid- and long-range correlations, which are described similarly by all tested functionals.
  • Electronic Structure: The electronic density of states (DOS) was found to depend almost entirely on the ionic structure (snapshot geometry) rather than the specific xc functional used to generate the DOS. For a given ionic configuration, the band gap remained stable regardless of whether PBE, vdW, or other functionals were used to calculate the electronic properties.

Significance and Claims
The paper concludes that for warm dense hydrogen, the inclusion of non-local vdW corrections does not offer a fundamental advantage over standard functionals like PBE for describing the system's physics. The observed shifts in the LLPT line when using vdW functionals are primarily a consequence of their tendency to overestimate molecular dissociation energies, rather than a direct result of improved vdW force descriptions.

The authors assert that:

1. PBE remains a robust choice for warm dense hydrogen, particularly given its computational efficiency compared to hybrids like HSE06, and its ability to describe vdW interactions (in the context of molecule-molecule attraction) nearly as well as HSE06.
2. r2SCAN is the superior functional for calculating the EOS and LLPT if the decision is based solely on the accuracy of molecular bond lengths and dissociation energies, despite its failure to describe vdW minima.
3. vdW effects are negligible in the context of the hydrogen EOS at relevant WDM temperatures and pressures because the energy scale of polarization and neutral-neutral interactions is dwarfed by dissociation and ionization energies.

The study suggests that future improvements in understanding pressure dissociation should focus on electron-electron correlation functions and new quantum Monte Carlo methods rather than the selection of specific vdW-corrected xc functionals.