🔬 materials science

Capturing Nuclear Quantum Effects in Hydrogen Diffusion through MoS2 via Machine-Learning-Enhanced Path-Integral Simulations

This study employs machine-learning-enhanced path-integral simulations to demonstrate that nuclear quantum effects significantly lower free-energy barriers and induce a pronounced kinetic isotope effect for hydrogen diffusion in MoS2, while revealing that transport in twisted bilayer structures is strongly modulated by local stacking environments within moiré superlattices.

Original authors: Ismail Eren, Ege Yigit Erbil, Maria-Judith Caisachana-Lozada, Hossein Mirhosseini, Thomas D. Kühne, Agnieszka B. Kuc

Published 2026-06-25

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Ismail Eren, Ege Yigit Erbil, Maria-Judith Caisachana-Lozada, Hossein Mirhosseini, Thomas D. Kühne, Agnieszka B. Kuc

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 a sandwich made of two slices of bread (the layers of a material called MoS2) with a tiny crumb (a hydrogen atom) trying to squeeze through the gap in the middle. Scientists have long known that this "crumb" can move from one side of the sandwich to the other, which is useful for things like storing hydrogen fuel or separating different types of hydrogen.

However, there's a catch. For a long time, scientists calculated how this crumb moves by treating it like a tiny, solid marble rolling on a track. But hydrogen is so light that it doesn't behave like a marble; it behaves more like a fuzzy cloud of probability. It vibrates wildly and can even "tunnel" through walls it shouldn't be able to cross. This is called Nuclear Quantum Effects (NQEs).

This paper is about finally building a better map to see how this "fuzzy cloud" actually moves through the sandwich, and how it moves differently than its heavier cousin, Deuterium (which is like a marble with a lead weight attached).

Here is the breakdown of their discovery:

1. The "Smart" Simulator

To figure this out, the researchers couldn't use standard computer models because they are too slow and treat the hydrogen like a solid marble. Instead, they built a Machine-Learning Interatomic Potential (MLIP).

The Analogy: Think of this as training a super-smart AI assistant. They fed the AI thousands of perfect, high-speed snapshots of how atoms interact (from a very expensive, precise calculation method called DFT). Once trained, this AI could predict how atoms behave almost as accurately as the expensive method, but millions of times faster. This allowed them to run simulations that included the "fuzzy" quantum nature of the hydrogen.

2. The "Fuzzy" vs. The "Solid"

When they ran the simulation with the "fuzzy" quantum hydrogen (using a method called Path-Integral Molecular Dynamics), they found something surprising: The path was much easier.

The Analogy: Imagine trying to walk through a dense forest. If you are a solid person (classical physics), you have to push through every branch and bush, which takes a lot of energy. But if you are a ghost (quantum physics), you can phase right through the branches.
The Result: The "ghost" hydrogen found it much easier to jump between layers. The energy barrier (the hill it had to climb) dropped significantly—sometimes by nearly half. This means hydrogen moves much faster than scientists previously thought, especially at room temperature.

3. The Heavy Cousin (Deuterium)

They also tested Deuterium, which is just hydrogen with an extra neutron (making it heavier).

The Analogy: If hydrogen is a wobbly, fuzzy ghost, Deuterium is a slightly less wobbly, heavier ghost. It still has some "fuzziness," but not as much.
The Result: Because Deuterium is heavier, it doesn't tunnel through the barriers as easily as the lighter hydrogen. The difference in how easily they move is huge. The "hill" Deuterium has to climb is about 35 meV higher than hydrogen's. This creates a Kinetic Isotope Effect, meaning the material naturally acts like a sieve, letting the lighter hydrogen zip through while holding back the heavier deuterium.

4. Twisting the Sandwich (Twisted Bilayers)

The researchers didn't just look at flat sandwiches; they looked at sandwiches where the top slice is twisted slightly relative to the bottom one. This creates a giant, repeating pattern called a Moiré pattern (like the ripples you see when you overlap two window screens).

The Analogy: Imagine the gap between the bread slices isn't the same everywhere. In some spots (called "domains"), the bread slices line up perfectly, creating a wide, open hallway for the crumb to walk through. In other spots (called "nodes" or "solitons"), the bread slices are misaligned, creating a narrow, bumpy tunnel that is hard to get through.
The Result: The hydrogen doesn't move evenly. It loves the "wide hallways" (staggered stacking) and hates the "bumpy tunnels" (eclipsed stacking). By twisting the layers at different angles, you can create a map where hydrogen flows easily in some areas and gets stuck in others.

The Bottom Line

This paper proves that if you want to understand how hydrogen moves through these 2D materials, you cannot ignore its quantum "fuzziness."

Quantum effects make hydrogen move much faster than old models predicted.
Hydrogen and Deuterium move very differently, which is great for separating them.
Twisting the layers creates a complex landscape where hydrogen flows like water through a maze, speeding up in some spots and slowing down in others.

The authors conclude that by using their new "AI-assisted quantum simulator," we can now design better materials for hydrogen storage and separation by carefully choosing how to stack and twist these atomic layers.

Technical Summary: Capturing Nuclear Quantum Effects in Hydrogen Diffusion through MoS₂ via Machine-Learning-Enhanced Path-Integral Simulations

Problem Statement
Hydrogen transport through layered two-dimensional (2D) materials, such as molybdenum disulfide (MoS₂), is critical for applications in hydrogen storage, fuel cells, and isotope separation. While experimental studies have demonstrated interlayer diffusion and isotope-selective transport in materials like MoS₂ and h-BN, accurate theoretical descriptions remain challenging. Previous theoretical investigations largely relied on classical treatments of atomic nuclei, neglecting Nuclear Quantum Effects (NQEs). However, light atoms like hydrogen exhibit pronounced quantum behaviors—including zero-point motion (ZPE), tunneling, and delocalization—even at room temperature. Furthermore, accurately modeling these systems requires capturing both van der Waals (vdW) interactions and hydrogen bonding, which necessitates advanced exchange-correlation functionals (e.g., r2SCAN+rVV10) in Density Functional Theory (DFT). The computational cost of combining such high-fidelity DFT with Path-Integral Molecular Dynamics (PIMD) and enhanced sampling for large, structurally complex systems (such as twisted bilayers) is prohibitive.

Methodology
To address these computational challenges, the authors developed a machine-learning interatomic potential (MLIP) based on the MACE framework, trained on a compact dataset generated using r2SCAN+rVV10 DFT. This MLIP was validated against DFT reference calculations for energetic stability and phonon band structures, showing good agreement.

The study employed a hybrid simulation framework combining:

Well-Tempered Metadynamics (WTMetaD): Implemented via PLUMED to calculate free-energy barriers ( $\Delta F$ ) for interlayer diffusion.
Path-Integral Molecular Dynamics (PIMD): Implemented via the i-PI package to explicitly incorporate NQEs. Nuclei were represented by ring polymers with $P=16$ beads for hydrogen (and $P=16$ for deuterium, ensuring convergence) at 300 K.
System Configurations: The study investigated high-symmetry bulk MoS₂ polytypes ( $H_h^h$ , $H_X^h$ , $R_M^h$ , and 3R phases) and twisted bilayer MoS₂ (TBL-MoS₂) with twist angles of $\theta = 3.89^\circ$ , $21.79^\circ$ , $38.21^\circ$ , and $56.11^\circ$ .
Isotope Analysis: Both protium (H) and deuterium (D) diffusion were simulated to quantify kinetic isotope effects (KIE).

Key Contributions

Development of an MLIP-PIMD Framework: The authors established a robust workflow that bridges the gap between DFT accuracy and the length/time scales required for quantum nuclear dynamics in complex 2D systems.
Quantification of NQEs in MoS₂: The study provides a systematic assessment of how quantum fluctuations alter diffusion landscapes across different MoS₂ polytypes.
Isotope-Selective Mechanism: The work offers atomistic insight into the mechanisms driving isotope separation, specifically quantifying the difference in free-energy barriers between H and D.
Moiré-Driven Diffusion Landscapes: The study maps hydrogen transport in twisted bilayers, revealing how moiré superlattices create spatially varying diffusion environments governed by local stacking (staggered vs. eclipsed) and soliton networks.

Key Results

Impact of NQEs on Diffusion Barriers: The inclusion of NQEs substantially lowers free-energy barriers for hydrogen diffusion compared to classical simulations. The reduction ranges from tens of meV to nearly 60 meV depending on the stacking configuration.
- For the most stable $H_h^h$ phase, the barrier drops from 122 meV (classical) to 64 meV (quantum).
- For the $H_X^h$ phase, the barrier decreases from 45 meV to 18 meV.
- This softening of the free-energy landscape leads to a significant increase in the self-diffusion coefficient ( $D$ ), with Quantum Correction Factors (QCF) reaching up to 9.43 for the $H_h^h$ phase.
Kinetic Isotope Effect (KIE): A pronounced KIE is observed. While deuterium exhibits higher barriers than hydrogen in classical simulations, the inclusion of NQEs amplifies this difference. The quantum free-energy barrier for protium is approximately 35 meV lower than that for deuterium (60 meV vs. 95 meV for $H_h^h$ ), indicating strong mass-dependent quantum diffusion.
Twisted Bilayer Dynamics:
- High-Angle Twists ( $21.79^\circ, 38.21^\circ$ ): These systems lack extended high-symmetry domains, resulting in diffusion landscapes dominated by intermediate configurations. Consequently, diffusion barriers are higher (~120 meV quantum) and diffusivities are lower (at least one order of magnitude lower than isolated high-symmetry phases).
- Low-Angle Twists ( $3.89^\circ, 56.11^\circ$ ): These systems form distinct moiré domains (staggered) and nodes/solitons (eclipsed). Hydrogen preferentially resides in staggered domains where interlayer distances are smaller and coordination is favorable.
- Spatial Variation: Diffusion barriers vary significantly across the moiré superlattice, with differences up to 80 meV between favorable staggered domains and unfavorable soliton/eclipsed regions. The $56.11^\circ$ system exhibits lower average barriers (~~150 meV) compared to the $3.89^\circ$ system (~~197 meV), attributed to the presence of metastable nodes and favorable $H_X^h$ domains.

Significance
The paper concludes that NQEs play a decisive role in hydrogen transport through layered MoS₂, even at ambient temperatures. Neglecting these effects leads to an overestimation of diffusion barriers and an underestimation of diffusivity. The study demonstrates that the magnitude of quantum enhancement is highly sensitive to the local stacking environment. Furthermore, the explicit inclusion of quantum nuclear dynamics is essential for accurately describing isotope-selective transport mechanisms. Finally, the research establishes that twist engineering in 2D materials can be used to tailor diffusion landscapes, creating spatially selective channels for hydrogen migration. The presented MLIP-PIMD framework is positioned as a transferable approach for investigating quantum transport in complex van der Waals heterostructures relevant to energy technologies.