ZrO₂/Li₂ZrCl₆ Heterostructures: Li⁺ Transport Insights

• The paper reveals that integrating ZrO₂ nanoparticles into Li₂ZrCl₆ dramatically enhances Li⁺ conductivity through interfacial amorphization and optimized space-charge effects.
• Machine learning–driven molecular dynamics simulations accurately model the interface, quantifying Li⁺ diffusion improvements by over one order of magnitude.
• The study provides practical design guidelines for electrolyte composites that balance ZrO₂ volume and interface orientation to maintain sufficient Li⁺ reservoirs for optimal conduction.

ZrO₂/Li₂ZrCl₆ heterostructures are engineered composites in which ionically insulating zirconia (ZrO₂) nanoparticles are integrated with lithium zirconium chloride (Li₂ZrCl₆) solid electrolytes. These systems have emerged as critical architectures to enhance lithium-ion (Li⁺) conductivity, a persistent bottleneck in halide-based all-solid-state batteries. Detailed machine learning-driven molecular dynamics studies have provided atomistic insight into the interfacial phenomena responsible for the remarkable conductivity boost observed upon introduction of ZrO₂ at the Li₂ZrCl₆ interface, elucidating the structural, dynamical, and mechanistic underpinnings of Li⁺ transport in these materials (Xu et al., 30 Jan 2026).

1. Atomistic Construction of ZrO₂/Li₂ZrCl₆ Interfaces

Representative interface models are constructed between low-mismatch facet pairs of ZrO₂ and Li₂ZrCl₆, focusing on minimizing lattice strain (≤1%) and maximizing interfacial relevance:

ZrO₂ Facet & Phase	Li₂ZrCl₆ Facet & Phase	Interface Character
Cubic Fm3m(111)	ideal-P31c(001)	ZrO₇ vs. fully occupied LiCl₆/ZrCl₆ layers
Tetragonal P4₂/nmc(101)	ideal-P31c(101)	Z-directional Li⁺ tunnels
Cubic Fm3m(221)	a-phase P3m1(001)	ZrO₄/⁵ vs. disordered LiCl₆ octahedra
Monoclinic P2₁/c(010)	a-phase P3m1(201)	ZrO₄ vs. mixed LiClₓ/ZrClₓ

Interface structures are generated with the InterOptimus package, sampling relevant Miller indices and surface terminations under constraints of interface area ≤200 Ų and strain <1%. Supercells are commensurately matched in-plane, periodic along all axes, with cell dimensions of 40–50 Å along the surface normal. These protocols ensure accurate representation of the physical interface without vacuum artifacts or excessive artificial strain (Xu et al., 30 Jan 2026).

2. Machine-Learned Interatomic Potential Parameterization

To enable large-scale simulation of complex interfaces, force fields are constructed using neuroevolution potential (NEP):

• Radial cutoff: 6 Å; Angular cutoff: 5 Å
• 8 radial and 8 angular descriptor channels using 12 basis functions
• Single hidden layer with 50 neurons; short-range ZBL repulsion activated for $r_{ij}<1.2$ Å

An active learning workflow, employing a “query-by-committee” with parallel NEP models, targets high-uncertainty configurations (force variance >0.3 eV/Å) via farthest-point sampling for DFT labeling. The training corpus expands from ~4,931 DFT-annotated structures (covering both bulk and interface geometries of various phases and polymorphs) to ≈12,000 after 25 iterations. Achieved errors on the final set are: energy RMSE 5.6 meV/atom, force RMSE 219.2 meV/Å, virial RMSE 17.4 meV/atom, and ≈10.0 meV/atom on a hold-out interface set (Xu et al., 30 Jan 2026).

3. Molecular Dynamics Protocols and Mobility Quantification

Molecular dynamics simulations employ the trained NEP model within GPUMD:

• NPT ensemble (1 bar) with Bussi–Donadio–Parrinello thermostat and stochastic barostat
• Time step: 1 fs; total simulation times: up to 5 ns (bulk, ≤500 K), up to 2 ns (>500 K), and 20 ns for interfaces
• Temperatures: 300, 400, 500 K (bulk benchmarking up to 700 K)
• Center-of-mass motion suppressed every 10 fs

Li⁺ diffusivity is calculated from the mean-squared displacement (MSD):

$$\mathrm{MSD}(t) = \frac{1}{N}\sum_{i=1}^{N}\langle|\mathbf{r}_i(t)-\mathbf{r}_i(0)|^2\rangle$$

Resulting Li⁺ diffusion coefficients and conductivities are extracted:

$D = \lim_{t\to\infty} \frac{1}{6t}\mathrm{MSD}(t)$

$\sigma = \frac{nq^2 D}{k_\mathrm{B}T}$

where $n$ is the Li⁺ number density and $q$ the elementary charge (Xu et al., 30 Jan 2026).

4. Interfacial Amorphization and Space-Charge Effects

Surface cleavage introduces undercoordinated ZrO₇ units at the ZrO₂ interface, causing a local electrostatic imbalance. Consequently, a space-charge region forms, modeled as:

$$\nabla^2 \phi(\mathbf{r}) = -\frac{\rho(\mathbf{r})}{\varepsilon_0\varepsilon_r}$$

where $\phi$ is the electrostatic potential, $\rho$ the charge density. This drives the spontaneous migration of Li⁺ ions from Li₂ZrCl₆ into the interface, inducing local amorphization with thickness circa 8 Å within ~2 ns of simulation time. The resulting non-crystalline layer is characterized by a high density of undercoordinated, geometrically distorted polyhedra, substantially differing from the bulk stoichiometry and coordination environment (Xu et al., 30 Jan 2026).

5. Li⁺ Coordination and Diffusion Mechanisms

The interface region displays markedly altered Li⁺ environments:

• In ideal-phase bulk, Li⁺ predominantly exhibits octahedral coordination (CN=6, volume 20–25 ų, CSM <3).
• a-phase bulk manifests increased populations of LiCl₅ (CN=5, 10–13 ų) and LiCl₄ (CN=4, 5–8 ų).
• Interfacial regions, for both ordered and amorphous phases, predominantly contain Li⁺ with CN <6, and highly elevated continuous symmetry measure (CSM up to ~30), indicative of severe geometric distortion.
• Hopping events (Li⁺ displacements >1.25 Å/10 ps) are concentrated in the amorphous interfacial layer (>0.08 hops/frame vs. ~0 in bulk ideal-phase). Normal (z-type) Li⁺ motion is confined to <15 Å, whereas in-plane (xy-type) conductivity extends >20 Å and, in the a-phase, >40 Å (Xu et al., 30 Jan 2026).

6. Quantitative Conductivity Enhancement

Room-temperature (300 K) bulk ionic conductivity benchmarks are:

Phase	$\sigma_{\mathrm{RT}}$ (mS/cm)	$E_a$ (eV)
a-Li₂ZrCl₆	0.803	0.29
ideal-Li₂ZrCl₆	0.066	0.56

At the interface:

• Ideal-phase interfaces: Li⁺ mobility within the amorphous region rises by more than 1–2 orders of magnitude over the bulk lattice; transport perpendicular to the interface is confined to <15 Å.
• a-phase interfaces: The intrinsically disordered environment enables Li⁺ excursions >30 Å along the z-axis and >40 Å in-plane, supporting long-range conduction pathways.
• Sites with depleted Li⁺ inventory, particularly at Li-poor terminations, exhibit suppressed hopping despite strong geometric disorder, underscoring the necessity of a local Li⁺ reservoir for interfacial conductivity gains (Xu et al., 30 Jan 2026).

7. Implications for Electrolyte Composite Design

Several concrete guidelines emerge for the rational design of high-conductivity heterostructures:

• Preference for a-Li₂ZrCl₆ over fully ordered (ideal) phase, leveraging its cation disorder and higher excess Li⁺ content to seed undercoordination at interfaces.
• The size and volume fraction of ZrO₂ inclusions must be optimized to maximize interfacial area and facilitate disorder-induced enhancement, counterbalanced against the risk of global Li⁺ depletion from the bulk.
• Interface orientation and termination engineering (e.g., ZrO₇ vs. mixed ZrO₃Cl₃) directly modulate the space-charge potential and local Li⁺ supply, offering a structural lever to tune fast conduction pathways.
• Avoidance of orientations leading to severe Li⁺ depletion at the interface is crucial for sustaining mobile-ion density and realizing full conductivity enhancement potential (Xu et al., 30 Jan 2026).

These findings establish that interfacial amorphization—driven by cleavage-induced space-charge formation—is the principal mechanism enabling conductivity boosts in ZrO₂/Li₂ZrCl₆ heterostructures. The resulting undercoordinated, highly distorted Li⁺ environments serve as rapid multidimensional transport channels, with the ultimate magnitude of enhancement governed by mobile-ion availability at the interface.