The effects of dispersion damping and three-body interactions for accurate layered-material exfoliation energies

This study demonstrates that incorporating three-body Axilrod–Teller–Muto interactions into the XDM dispersion model, particularly with the new Z-damping function, yields the most accurate exfoliation energies for layered materials using semi-local density-functional theory to date.

The Gist

Imagine you have a stack of sticky notes. If you try to pull one off the top, it comes off easily. But if you try to pull a single sheet out from the middle of a thick pad, it's much harder because the sheets are holding onto each other.

In the world of materials science, scientists study "layered materials" (like graphite in your pencil or special 2D semiconductors) that act just like that stack of sticky notes. They want to know exactly how much energy it takes to peel a single layer off the bulk material. This is called the exfoliation energy.

To predict this energy, scientists use a powerful computer simulation tool called Density-Functional Theory (DFT). Think of DFT as a super-advanced calculator that tries to figure out how atoms behave. However, these calculators have a blind spot: they are terrible at calculating "invisible glue."

The Invisible Glue: London Dispersion

The layers in these materials don't stick together with strong chemical bonds; they stick together via London dispersion forces. You can think of these as tiny, fleeting magnetic whispers between atoms. They are weak, but when you have billions of atoms, they add up to a strong hold.

Standard computer models often ignore these whispers, leading to terrible predictions (like thinking the layers would fly apart instantly). To fix this, scientists add "corrections" to the math. This paper is about testing two specific ways to add this invisible glue to the calculation.

The Two Main Characters: BJ vs. Z Damping

The paper compares two different "rules" for how strong this invisible glue should be when atoms are very close together.

1. The BJ Rule (Becke-Johnson): This is the old, established rule. It works well for many things, but the authors found it sometimes gets a bit too excited when atoms are close, making the glue stronger than it really is. It's like a friend who hugs you a little too tight.
2. The Z Rule (Z-damping): This is a newer, smarter rule. It was designed to be more careful, especially with certain types of atoms (like alkali metals). It acts like a friend who knows exactly how much pressure to apply—firm but not crushing.

The Analogy: Imagine you are trying to stack a tower of Jenga blocks.

• BJ is like using a glue that sets too fast and too hard, making the tower wobble or collapse because the blocks are stuck too tightly in the wrong spots.
• Z is like using a glue that sets at the perfect speed, holding the blocks just right.

The paper found that while both rules are good, the Z rule is slightly more consistent and reliable, especially when dealing with a wide variety of materials.

The Third Character: The Three-Body Effect (ATM)

Here is where it gets interesting. Most calculations only look at how two atoms interact (Pairwise). But in a crowded room (or a layered material), what happens between three people matters too.

The paper introduces a Three-Body Interaction (called the ATM term).

• The Analogy: Imagine three friends standing in a circle. If they are in a straight line, they pull together nicely. But if they stand in a triangle (like the atoms in these layered materials), they actually push each other apart slightly.
• In these layered materials, the atoms are arranged in triangles. The "Three-Body" effect acts like a repulsive force that pushes the layers slightly apart.

Why does this matter?
Because the "glue" (dispersion) was often calculated to be too strong (overbinding), adding this "push" (the three-body effect) balances the equation perfectly. It's like realizing you added too much salt to a soup, so you add a tiny bit of lemon juice to balance the flavor.

The Big Discovery

The researchers tested these methods on a "benchmark" of 26 different layered materials (including graphite and others).

1. The Old Way: Without these corrections, the computer thought the layers would fall apart.
2. The "Glue" Only Way: Adding just the dispersion correction made the layers stick together, but often too tightly.
3. The Winning Combo: The most accurate results came from using the Z-damping rule (or the BJ rule) PLUS the Three-Body push.

This combination gave the most accurate prediction of how much energy is needed to peel a layer off, matching the "gold standard" of physics (called RPA) better than any previous method using standard computer models.

The Takeaway for Everyone

This paper is like a recipe refinement for a chef trying to bake the perfect cake (the layered material).

• They realized the old recipe (standard DFT) was missing an ingredient (dispersion).
• They tried two different brands of that ingredient (BJ vs. Z).
• They realized that simply adding the ingredient made the cake too dense, so they added a tiny pinch of a third ingredient (the three-body effect) to lighten it up.

The Result: They found the perfect recipe. Now, scientists can use this "recipe" to design new, better materials for batteries, electronics, and lubricants with much higher confidence, without needing to run incredibly expensive and slow supercomputer simulations.

In short: They figured out how to stop the computer from hugging the atoms too tight and how to account for the fact that atoms in a crowd push each other away slightly, leading to a much more accurate map of how these amazing 2D materials behave.

Technical Summary

1. Problem Statement

Accurate prediction of the properties of layered materials (e.g., graphite, hexagonal boron nitride, transition-metal dichalcogenides) relies heavily on the correct description of London dispersion interactions. While Density-Functional Theory (DFT) is the standard computational tool, standard functionals often fail to capture these weak van der Waals forces.

• The Challenge: Existing dispersion corrections (like DFT-D3 and XDM) use damping functions to prevent unphysical divergence at short interatomic distances. However, the choice of damping function (Becke–Johnson vs. Z-damping) and the inclusion of many-body effects (specifically the three-body Axilrod–Teller–Muto or ATM term) have not been systematically benchmarked for layered materials.
• Specific Gaps:
  • Previous studies suggested that pairwise dispersion corrections often overbind layered materials, leading to inaccurate exfoliation energies.
  • The performance of the newer XDM(Z) damping function (which resolves overbinding in alkali metals) had not been tested on the LM26 benchmark (a set of 26 layered materials).
  • The impact of the repulsive three-body ATM term on layered materials using XDM corrections was unexplored.

2. Methodology

The authors conducted a comprehensive benchmark study using the LM26 dataset (26 layered materials with Reference Random-Phase Approximation (RPA) data).

• Computational Framework:
  • Software: Quantum ESPRESSO (using Plane-Wave/PAW basis sets) and FHI-aims (using Numerical Atomic Orbital basis sets).
  • Functionals: Semi-local GGA functionals (revPBE, PBE, B86bPBE) paired with dispersion corrections.
  • Dispersion Models:
    • XDM: Tested with both Becke–Johnson (BJ) and Z-damping functions.
    • D3: Tested with Zero-damping (0) and BJ-damping.
    • Three-Body Term: Both XDM and D3 variants were tested with and without the Axilrod–Teller–Muto (ATM) term.
• Theoretical Developments:
  • The authors introduced a new Z-damping function for the three-body ATM term ($f^Z_{ATM}$), derived as the product of three partial Z-damping functions. This ensures the correct united-atom limit behavior independent of dispersion coefficients.
  • Parameters for XDM(Z) were re-optimized for PAW calculations in Quantum ESPRESSO and NAO basis sets in FHI-aims.
• Metrics:
  • Exfoliation Energy ($E_{exfol}$): Energy required to separate a monolayer from the bulk.
  • Lattice Parameter ($c$): Interlayer spacing.
  • Error Metrics: Mean Absolute Error (MAE) and Mean Error (ME) compared to RPA reference data.

3. Key Contributions

1. First Benchmark of XDM(Z) on Layered Materials: The study provides the first assessment of the XDM(Z) damping function on the LM26 benchmark, comparing it directly against the established XDM(BJ).
2. Development of XDM(Z)+ATM: The authors derived and implemented a consistent three-body damping scheme for XDM(Z), allowing for the first time the inclusion of ATM effects with Z-damping.
3. Systematic Analysis of Damping and Many-Body Effects: The paper isolates the effects of damping functions (BJ vs. Z) and the inclusion of the ATM term across different functionals and basis sets.
4. Identification of Optimal Protocols: The study identifies specific combinations of functional, damping, and basis set that achieve state-of-the-art accuracy for layered materials at a semi-local computational cost.

4. Results

• Performance of Damping Functions:
  • XDM(Z) vs. XDM(BJ): XDM(Z) performs comparably to or slightly better than XDM(BJ) for layered materials. While average MAEs were similar (~4.9 meV/Ų), XDM(Z) showed a lower average absolute Mean Error (3.0 meV/Ų vs. 3.9 meV/Ų), indicating better systematic accuracy.
  • D3 Performance: D3 corrections generally overbind significantly. Only PBE-D3(0)+ATM approached the target accuracy (<5 meV/Ų), but XDM methods were generally more robust.
• Impact of the ATM Term:
  • The ATM term is repulsive for most layered materials due to the prevalence of ~60° angles between atomic triples spanning adjacent layers (similar to $\pi$-$\pi$ stacking dimers).
  • Including the ATM term consistently reduced the overbinding inherent in pairwise-only methods.
  • Best Performers: The inclusion of ATM with XDM yielded the lowest errors reported to date for the LM26 benchmark using semi-local functionals.
    • Top Methods: B86bPBE-XDM(BJ)+ATM and B86bPBE-XDM(Z)+ATM (using Plane-Wave/PAW basis sets).
    • Cost-Effective Methods: B86bPBE-XDM(BJ)+ATM and PBE-XDM(Z)+ATM using the lightdenser NAO basis set achieved excellent results due to error cancellation from basis-set superposition error (BSSE).
• Lattice Parameters:
  • While the ATM term significantly improved exfoliation energies, its effect on the $c$-lattice parameter was minor.
  • The best simultaneous accuracy for both energy and geometry was achieved with B86bPBE-XDM(BJ)+ATM and B86bPBE-XDM(Z)+ATM.
• Comparison to SCAN-rVV10: The semi-local methods with XDM+ATM rival or outperform the more computationally expensive non-local SCAN-rVV10 method for predicting exfoliation energies and lattice parameters.

5. Significance and Conclusions

• Accuracy: The study establishes that combining semi-local GGA functionals with XDM dispersion corrections (specifically using Z-damping) and the three-body ATM term provides the most accurate description of layered materials currently available at a low computational cost.
• Practical Recommendation:
  • For Geometry Optimization: The ATM term has a negligible effect on structural parameters ($c$-axis) and is computationally expensive for periodic systems. Therefore, it is recommended to use pairwise XDM for geometry optimization and apply the ATM term only as a final single-point energy correction.
  • For Energy Prediction: The B86bPBE-XDM(Z)+ATM (or BJ) protocol is recommended for high-accuracy exfoliation energy predictions.
• Theoretical Insight: The work confirms that three-body dispersion effects are non-negligible and often repulsive in layered systems, correcting the systematic overbinding of pairwise models. It also validates the utility of the Z-damping function, which offers a more physically consistent united-atom limit than BJ damping, particularly for s-block metals and layered materials.

In summary, this paper provides a definitive protocol for modeling layered materials, demonstrating that the inclusion of three-body interactions and the use of Z-damping in the XDM framework significantly enhances the predictive power of DFT for 2D materials.