🔬 materials science

Temperature-driven structural phase transitions in SmNiO $_3$ : insights from deep potential molecular dynamics simulations

By employing deep potential molecular dynamics simulations on SmNiO $_3$ , this study demonstrates that the structural phase transition driving the metal-insulator transition is intrinsically temperature-induced and occurs spontaneously through collective lattice distortions, independent of electronic effects.

Original authors: Guoyong Shi, Fenglin Deng, Ri He, Dachuan Chen, Xuejiao Chen, Peiheng Jiang, Zhicheng Zhong

Published 2026-06-30

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Guoyong Shi, Fenglin Deng, Ri He, Dachuan Chen, Xuejiao Chen, Peiheng Jiang, Zhicheng Zhong

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

The Big Picture: A Material That Changes Its Mind

Imagine a material called SmNiO₃ (Samarium Nickelate) as a busy city made of tiny building blocks. This city has a special superpower: it can switch between being a "metal" (where electricity flows freely like a highway) and an "insulator" (where electricity is blocked like a road closed for construction).

Scientists have long known that this switch happens when the temperature changes. But there was a big mystery: What actually causes the switch?

Is it the electrons (the tiny particles carrying electricity) getting restless and forcing the buildings to change shape?
Or is it the structure (the buildings themselves) getting shaky from the heat and collapsing into a new shape, which then stops the electricity?

It's like trying to figure out if a crowd of people started running because they saw a fire (electrons causing the change), or if they started running because the floor started shaking (heat causing the structure to change). In this material, the two are so tightly linked that it's hard to tell which one started it.

The New Tool: A "Smart" Simulation

To solve this, the researchers used a clever trick. Instead of simulating the complex behavior of every single electron (which is like trying to track every person in the crowd individually), they used Machine Learning to create a "smart map" of how the atoms push and pull on each other.

Think of it like this:

Old Method (DFT): You try to simulate the whole city, including the people, the traffic, and the weather, all at once. It's accurate, but you can't easily separate the people from the weather.
New Method (This Paper): They trained a computer program (a "Machine-Learned Potential") to learn exactly how the buildings (atoms) interact with each other based on the old, complex data. Then, they ran a simulation where only the buildings move, ignoring the people (electrons) for a moment.

This allowed them to ask: "If we just heat up the buildings without worrying about the electrons, do they still change shape?"

What They Found: The Heat Does the Work

The results were surprising and clear: Yes, the heat alone is enough to make the change.

The "Breathing" Buildings: In the cold version of this material (the insulator state), the tiny building blocks (called NiO₆ octahedra) are arranged in a pattern where some are big and some are small. It's like a checkerboard of big and small houses. This is called "bond disproportionation."
The Heat Wave: As they turned up the virtual thermostat, the atoms started vibrating more. Imagine the buildings shaking so much that the big ones shrink and the small ones expand.
The Tipping Point: At 340 Kelvin (about 140°F or 65°C), the shaking became so intense that the big and small buildings became indistinguishable. They all became the same size. The "checkerboard" pattern disappeared, and the material switched to a uniform state (the metal state).

The Key Takeaway: The researchers showed that the structural change (the buildings shaking and merging) happens spontaneously just because of the temperature. The electrons didn't need to push first; the heat alone was enough to break the pattern.

Pressure: The Squeeze

The team also tested what happens if you squeeze the material (apply pressure).

Imagine squeezing a sponge.
They found that if you squeeze SmNiO₃, it becomes harder to make the switch. You have to heat it up less to make it change, or rather, the temperature at which it changes drops.
At high pressure, the "switching point" dropped from 340 K down to 225 K. This matches what other scientists have seen in real experiments, proving their simulation is reliable.

Why This Matters

The paper concludes that lattice dynamics (the movement and shaking of the atomic structure) are the primary drivers of this transition. While the electrons and the structure work together in the real world, this study proves that the structure is very sensitive to temperature on its own.

In short: The material doesn't need a complex electronic signal to switch from metal to insulator. It just needs to get hot enough to shake its atomic "skeleton" until it falls into a new shape. The researchers successfully isolated this "skeleton" behavior using a machine-learning tool, giving us a clearer picture of how these materials work.

Technical Summary: Temperature-Driven Structural Phase Transitions in SmNiO3

Problem Statement
Rare-earth nickelates (RNiO3) exhibit a sharp metal-insulator transition (MIT) accompanied by structural and electronic changes, specifically bond disproportionation (breathing-mode distortion of NiO6 octahedra) and charge disproportionation. A central challenge in the field is disentangling the driving forces of this transition: whether it is primarily driven by correlated electronic effects (charge instability) or lattice structural dynamics. Traditional Density Functional Theory with Hubbard correction (DFT+U) treats structure and electronic states in an unbiased, entangled manner, making it difficult to isolate the specific role of lattice dynamics in triggering the phase transition.

Methodology
To decouple structural and electronic responses, the authors employed a machine-learned interatomic potential approach combined with multiscale molecular dynamics (MD) simulations, using SmNiO3 as a prototypical system.

Data Generation: Training datasets were generated using spin-polarized DFT+U calculations (VASP, PBEsol functional, Ueff = 2.0 eV) on supercells representing both the low-temperature monoclinic ( $P2_1/n$ ) and high-temperature orthorhombic ($Pbnm$) phases.
Potential Training: A Deep Potential (DP) model was trained using the DP-GEN active learning framework. The model was trained on 1,429 configurations, achieving root-mean-square errors (RMSE) of 2.03 meV/atom for energy and ~80–85 meV/Å for forces, validating its accuracy against DFT+U.
Simulation: Large-scale MD simulations were performed on a 7,680-atom supercell using the Deep Potential Molecular Dynamics (DPMD) method within the isothermal-isobaric (NPT) ensemble. Crucially, this classical MD framework utilizes the potential energy surface derived from DFT but does not explicitly include electronic degrees of freedom during the simulation, thereby isolating the structural response to thermal fluctuations.

Key Results

Structural Evolution: The simulations successfully reproduced the bond disproportionation characteristic of the $P2_1/n$ phase at low temperatures (50 K), showing distinct volumes for expanded ("L") and compressed ("S") NiO6 octahedra.
Temperature-Driven Transition: As temperature increased, thermal fluctuations broadened the volume distribution of the octahedra. At 320 K, the distinction between "L" and "S" units was significantly suppressed. By 340 K, the volume distributions merged completely, indicating the disappearance of bond disproportionation and a spontaneous transition to the centrosymmetric $Pbnm$ phase. The reverse process (cooling) confirmed the emergence of disproportionation at 340 K.
Critical Temperature and Pressure: The simulated critical temperature ( $T_c$ ) for the structural transition is 340 K. The study further demonstrated that hydrostatic pressure modulates this transition; increasing pressure from 0 to 10 GPa monotonically suppresses $T_c$ from 340 K to 225 K, consistent with experimental trends in the RNiO3 family.
Comparison with Experiment: The authors note that their simulated $T_c$ (340 K) is lower than the experimental value (~400 K) and that the transition appears less sharp without the hysteresis typical of first-order transitions.

Significance and Claims
The paper claims that these findings provide atomistic insights into the structural evolution triggering the phase transition in RNiO3. By isolating the structure response from electronic effects, the authors demonstrate that the structural phase transition is intrinsically temperature-driven and occurs spontaneously via collective lattice distortions.

The authors conclude that lattice dynamics play a significant role in driving the phase transition and the subsequent MIT. They posit that while electronic effects are essential for the full description of the MIT, the structural transition itself can be understood as a temperature-driven phenomenon. The work suggests that a complete physical picture of the MIT in nickelates requires the cooperative behavior of both lattice and electronic mechanisms, offering a clearer perspective than studies where these effects remain entangled. The authors do not claim to have solved the full electronic mechanism but rather to have successfully isolated and characterized the structural component of the transition.

The Big Picture: A Material That Changes Its Mind

The New Tool: A "Smart" Simulation

What They Found: The Heat Does the Work

Pressure: The Squeeze

Why This Matters

More like this