Skip to content

Atomistic Insights Into the Degradation of Inorganic Halide Perovskite CsPbI3: A Reactive Force Field Molecular Dynamics Study

Summary

CsPbI\(_3\) is an inorganic halide perovskite candidate for photovoltaics, but phase instability and decomposition under operational stimuli motivate atomistic models that go beyond harmonic lattice dynamics. This J. Phys. Chem. Lett. article introduces a Cs/Pb/I ReaxFF parametrization trained on PBE + DFT-D3(BJ) VASP data covering equations of state, atomic charges, formation energies, defect formation energies, and defect migration barriers for CsPbI\(_3\) and related precursors (CsI, PbI\(_2\)). Optimization uses a Monte Carlo ReaxFF optimizer in AMS 2020, starting from literature Cs/I parameters (Fedkin electrolyte–water set) and Pb parameters adapted from prior Pt-like training (Fantauzzi et al.), adjusted for Pb valency and mass. The local corpus PDF is a galley; cite the publisher VOR PDF for final figure labels when available. For the knowledge base, the paper matters as a worked example of QM → ReaxFF → reactive MD for a perovskite halide where both phase stability and decomposition are dominated by anion mobility and A-site disorder phenomena that are difficult to capture without bond rearrangement. The authors explicitly connect low-temperature structural fluctuations to loss of A-site registry and subsequent nonperovskite conversion, then use defect simulations to argue that iodine vacancies are especially effective at triggering PbI\(_2\)-like decomposition products in the trajectories analyzed.

Methods

1 — MD application. The paper reports reactive MD with the new Cs/Pb/I ReaxFF in LAMMPS-style workflows (see the full J. Phys. Chem. Lett. text) on 3D PBC perovskite and defected supercells. NVT and/or NPT segments, sub-1 fs timestep, and ps–ns trajectories (equilibration plus production) are specified in the article; Nose–Hoover-class thermostat; NPT Parrinello–Rahman or NVT-only barostat use must be read from the VOR Methods. Target temperatures in K appear in the phase-stability and defect studies. External electric field in the MD—N/A in the main stability narrative. Replica or metadynamics—N/A unless the SI adds a rare-event method.

2 — Force-field training. The ReaxFF is fit to PBE + DFT-D3(BJ) VASP data (equations of state, charges, defect formation and migration barriers, related precursor phases). Monte Carlo ReaxFF optimization in AMS 2020; parent parameter blocks are taken from literature Cs/I and Pb-related sets as the article states.

3 — Static QM / DFT. PBE + D3 dispersion on the VASP side; N/A GW or hybrid functionals for the training set in the main description.

4 — Galley. The corpus pdf_path is a galley; cite the version-of-record for final run parameters.

Findings

Mechanisms. The authors connect A-site (Cs) disorder and anharmonic lattice fluctuations to perovskite → nonperovskite trends in the ReaxFF sampling, and emphasize I-defect (vacancy and interstitial) mobility; I vacancies are highlighted as especially decomposition-promoting, with trajectories pointing toward PbI\(_2\)-like products in the scenarios they stress. ReaxFF addresses reactive ionics, not bandgaps or optical properties. Comparisons are to PBE+D3 DFT on defect and path energies where tabulated. Sensitivity to temperature and defect loading is central. Limitation: the ingested file is a galley; citable numbers should be taken from the VOR PDF and ## Limitations below.

Limitations

Galley PDF in corpus; electronic/optical properties are not modeled by ReaxFF. Transfer to mixed-halide or interface-rich devices requires additional training.

Relevance to group

Demonstrates ReaxFF extension into halide perovskites with explicit QM training and degradation diagnostics.

Citations and evidence anchors

Reader notes (navigation)

  • paper_keywords includes keyword:galley-or-proof-pdf.