🔬 materials science

Many-body benchmarking of DFT local-registry energetics in bilayer InSe

This study demonstrates that diffusion quantum Monte Carlo benchmarks reveal significant many-body effects in bilayer InSe that cause DFT to substantially underestimate local stacking-energy corrugation, thereby challenging the accuracy of standard DFT-based models for predicting structural and electronic properties in twisted layered materials.

Original authors: Jeonghwan Ahn, Abdulgani Annaberdiyev, Jovan Nelson, Nathaniel P. Stern, Hyeondeok Shin

Published 2026-07-03

📖 3 min read☕ Coffee break read

CC BY 4.0

Original authors: Jeonghwan Ahn, Abdulgani Annaberdiyev, Jovan Nelson, Nathaniel P. Stern, Hyeondeok Shin

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you have two sheets of a special, flaky material called Indium Selenide (InSe). When you stack these sheets on top of each other, the way they line up—like puzzle pieces fitting together—changes how they stick and behave. Scientists call this "stacking registry."

For years, researchers have used a popular computer simulation tool called DFT (Density Functional Theory) to predict how these sheets interact. Think of DFT as a very fast, but slightly blurry, map. It's great for getting a general idea of the terrain, but it might miss the tiny, crucial details of the hills and valleys.

The Problem: The "Blurry Map"
In this study, the authors looked at a specific type of InSe bilayer. The "blurry map" (DFT) suggested that three different ways of stacking the sheets (let's call them Stack A, Stack B, and Stack C) were almost identical in energy. It was like saying three different parking spots in a garage were equally easy to get into, with almost no difference in effort. The map also suggested that if you twisted the layers slightly, the energy wouldn't change much.

The Solution: The "High-Definition Camera"
To check if this map was accurate, the authors used a much more powerful and expensive tool called DMC (Diffusion Quantum Monte Carlo). If DFT is a blurry sketch, DMC is a high-definition, slow-motion camera that captures the exact physics of how electrons behave. It takes a lot more computing power, but it sees the truth.

The Big Surprise
When they used the high-definition camera, the picture changed completely:

The Parking Spots Aren't Equal: The three stacks that looked identical on the blurry map were actually very different. One stack was the clear winner (the best parking spot), another was a distant second, and the third was actually quite unstable. The energy difference between them was huge—like the difference between parking in a flat lot versus trying to park on a steep hill.
The "Twist" Matters More: Because the energy differences between the stacks are so much larger than the blurry map suggested, twisting the layers creates a much more dramatic effect. The material wants to rearrange itself into specific patterns (domains) much more strongly than the old models predicted.
Why the Difference? The authors looked at the "charge" (the electrons) to see why. They found that in the winning stack, electrons move in a very specific, helpful way that strengthens the bond. In the losing stack, the electrons move in a way that actually weakens the bond, even though the atoms look like they are in a similar position. The blurry map (DFT) smoothed over these subtle electron movements, making the stacks look the same when they are actually very different.

The Takeaway
This paper is a "reality check" for scientists studying twisted layers of materials. It shows that relying on the standard, fast simulation tools (DFT) can lead to a big underestimation of how much energy is needed to move between different stacking patterns.

In simple terms: The old maps said the terrain was flat and boring. The new, high-definition maps show it's actually a rugged landscape with deep valleys and high peaks. This means that when scientists design new materials by twisting layers, they need to account for these much stronger forces, or their predictions about how the material will shape itself might be wrong.

Technical Summary: Many-body benchmarking of DFT local-registry energetics in bilayer InSe

Problem Statement
Density functional theory (DFT) is the standard framework for modeling twisted bilayer materials, yet the accuracy of the underlying local stacking energetics remains uncertain. In layered materials like bilayer Indium Selenide (InSe), interlayer coupling is governed by a delicate balance of short-range repulsion, nonlocal electronic correlation, and interlayer hybridization. While DFT can describe structural trends, it often struggles to resolve the subtle energy differences between different lateral registries (stacking configurations), particularly when nonlocal van der Waals attraction and registry-dependent hybridization compete. This uncertainty is critical for twisted bilayers, where the moiré pattern samples a continuous range of local stacking environments. If the local registry energy landscape is inaccurately described (e.g., artificially flattened), predictions regarding structural relaxation, domain formation, and electronic reconstruction in twisted systems may be fundamentally flawed.

Methodology
The authors employ Diffusion Quantum Monte Carlo (DMC) to provide a many-body benchmark for the local-registry energetics of bilayer InSe, comparing these results against a broad suite of DFT functionals.

DMC Calculations: Performed using the QMCPACK code within the fixed-node approximation. The study utilized single-determinant Slater–Jastrow trial wave functions with one-, two-, and three-body Jastrow terms. To mitigate finite-size effects, the authors employed twist-averaged boundary conditions and extrapolated energies to the thermodynamic limit using three supercell sizes. Charge density analysis used an extrapolated estimator to cancel leading-order trial-wave-function bias.
DFT Calculations: A comprehensive set of exchange-correlation functionals was tested, including various van der Waals-corrected functionals (vdW-DF, vdW-DF2, rev-vdW-DF2, optB88, optB86b), meta-GGAs (SCAN, r2SCAN) with nonlocal correlation, and dispersion-corrected PBE variants (D3, MBD).
System Scope: The study focused on five high-symmetry bilayer registries (AA, AB, AAr, ABr, ACr) which represent the local environments found in twisted bilayer InSe. Additionally, the authors mapped the full lateral sliding landscape and analyzed twisted bilayer structures at specific angles (e.g., 7.34° and 21.79°) to connect local registry physics to moiré-scale properties.

Key Contributions and Results

Failure of DFT to Resolve Near-Degenerate Stackings:
DFT predicts that three specific stackings—AB, AAr, and ABr—which share the same interfacial Selenium (Se) registry motif, are nearly degenerate, with energy differences within 1.5 meV/f.u. The ordering of these stackings is highly functional-dependent, with no single functional providing a robust hierarchy. Furthermore, DFT predicts that the charge-density responses for these three configurations are nearly indistinguishable.
DMC Reveals Significant Energy Separations:
The DMC benchmark overturns the DFT picture, revealing substantial energy separations driven by many-body effects:
- AAr (β-like) is identified as the most stable stacking with a binding energy of -70(3) meV/f.u.
- AB (ε-like) is the second most stable at -62(4) meV/f.u., separated from AAr by 8(5) meV/f.u.
- ABr, despite having a similar short interlayer distance and Se-interface motif to AAr and AB, is significantly less stable at -29(3) meV/f.u., creating a separation of 41(4) meV/f.u. from AAr.
- The total energy range between the most stable (AAr) and least stable (AA/ACr) registries is 60(7) meV/f.u. in DMC, compared to only 11–23 meV/f.u. across the tested DFT functionals.
Microscopic Origin via Charge Redistribution:
Analysis of the DMC charge density difference ( $\Delta\rho$ ) reveals that the energetic hierarchy is driven by registry-specific charge redistribution between the interlayer gap and the intralayer bonding network.
- In the stable AAr configuration, charge is depleted from the interlayer overlap region and redistributed into the intralayer In–Se and In–In bonding environments, reinforcing the layers.
- In the unstable ABr configuration, charge accumulates in the interlayer gap while depleting the intralayer bonding regions, a redistribution that destabilizes the system despite the short interlayer distance.
- DFT fails to capture this reversal, predicting similar, smooth charge rearrangements for all three stackings, thereby missing the physical mechanism that differentiates their stabilities.
Implications for Twisted Bilayer InSe:
The study demonstrates that DFT-based models for twisted bilayer InSe substantially underestimate the local stacking-energy corrugation. Because the twist-angle energetics and structural relaxation are derived from the local registry landscape, the flattened DFT landscape leads to an underestimation of the driving forces for moiré-scale structural relaxation and domain formation. Specifically, the large DMC energy contrast between favorable (AB-like) and unfavorable (AA-like) registries suggests that the energy scale driving reconstruction in twisted InSe is much larger than DFT predicts.

Significance
The paper claims that reliable modeling of twisted layered materials requires validating the local-registry energy hierarchy and its full energetic scale, rather than relying solely on relaxed geometries or the identity of the DFT energy minimum. The results indicate that DFT can artificially compress energy differences among locally similar registries, leading to a "flattened" landscape that misrepresents the many-body interlayer coupling. Consequently, the weak twist dependence often observed in DFT calculations should not be interpreted as an intrinsic physical property but as an artifact of the flattened registry landscape. This benchmark establishes that many-body effects are essential for accurately describing the interlayer physics of bilayer InSe and, by extension, other twisted layered materials where registry-dependent energetics dictate structural and electronic properties.

More like this