Module ani 

ANI (Accurate NeurAl networK engINe) machine-learning potentials.

Implements the ANI potential family for fast energy and force prediction with near-DFT accuracy using neural-network inference on Behler-Parrinello atomic environment vectors (AEVs).

Architecture: positions → neighbor list → AEVs → per-element NN → total energy. Forces are computed via analytical backpropagation through the AEV pipeline.

Re-exports

pub use ani_tm::compute_aevs_tm;

pub use ani_tm::is_ani_tm_supported;

pub use ani_tm::AniTmResult;

pub use api::compute_ani;

pub use api::AniConfig;

pub use api::AniResult;

Modules

aev

Behler-Parrinello Atomic Environment Vectors (AEVs).

aev_params

ANI-2x symmetry function parameter tables.

ani_tm

Extended ANI potential covering 25+ elements including transition metals.

api

Public API for ANI machine-learning potentials.

cutoff

Smooth cutoff functions for distance-dependent interactions.

gradients

Analytical gradient computation (forces) for ANI potentials.

neighbor

Cell-list spatial partitioning for efficient neighbor search.

nn

Pure-Rust feed-forward neural network for ANI atomic energy prediction.

weights

Binary weight file loader for pre-trained ANI models.