KIM_API.jl

KIM_API.jl

using KIM_API, StaticArrays, LinearAlgebra

# Create model function
model = KIM_API.KIMModel("SW_StillingerWeber_1985_Si__MO_405512056662_006")

# Define system
species = ["Si", "Si"]
positions = [
    SVector(0.    , 0.    , 0.    ),
    SVector(1.3575, 1.3575, 1.3575),
]
cell = Matrix([[0.0 2.715 2.715] 
               [2.715 0.0 2.715] 
               [2.715 2.715 0.0]])
pbc = [true, true, true]

# Compute properties
results = model(species, positions, cell, pbc)
println("Energy: ", results[:energy])
println("Forces: ", results[:forces])

Overview

KIM_API.jl provides both low-level and high-level interfaces to KIM-API, enabling Julia users to access the extensive collection of validated interatomic models available through the OpenKIM framework. Main purpose of this package is to provide a convenient way to integrate KIM models into Julia based MD simulators. Think of this as the Julia equivalent of the KIMPY package but with additional wrappers for ease of use.

Features

• High-level functional interface with KIMModel()
• Automatic species mapping and validation
• Default support for multiple unit systems (metal, real, SI, CGS, atomic)
• Memory-safe wrappers around KIM-API C functions
• Comprehensive error handling and validation

Quick Start

using KIM_API, StaticArrays, LinearAlgebra

# Create model function
model = KIM_API.KIMModel("SW_StillingerWeber_1985_Si__MO_405512056662_006")

# Define system
species = ["Si", "Si"]
positions = [
    SVector(0.    , 0.    , 0.    ),
    SVector(1.3575, 1.3575, 1.3575),
]
cell = Matrix([[0.0 2.715 2.715] 
               [2.715 0.0 2.715] 
               [2.715 2.715 0.0]])
pbc = [true, true, true]

# Compute properties
results = model(species, positions, cell, pbc)
println("Energy: ", results[:energy])
println("Forces: ", results[:forces])

Installation

Prerequisites

You must have the KIM-API C++ library installed on your system. Visit openkim.org for detailed installation instructions.

Easiest way to install the KIM-API is Conda:

conda create -n kim-api kim-api=2.4 -c conda-forge
conda activate kim-api
export KIM_API_LIB=${CONDA_PREFIX}/lib/libkim-api.so

Package Installation

FOr latest version:

using Pkg
Pkg.add(url="https://github.com/openkim/KIM_API.jl.git")

or

using Pkg
Pkg.add("KIM_API")

Environment Setup

If KIM-API is not in your system library path, set the environment variable:

export KIM_API_LIB="/path/to/libkim-api.so"

Supported Unit Systems

KIM_API.jl supports multiple unit systems commonly used in molecular dynamics:

You can also specify custom units by passing a named tuple of units during model creation:

custom_units = (length=KIM_API.A, energy=KIM_API.eV, time=KIM_API.fs, charge=KIM_API.e, temperature=KIM_API.K)
model_custom = KIM_API.KIMModel("SW_StillingerWeber_1985_Si__MO_405512056662_006",
                                 units=custom_units)

KIM Model Access

Visit openkim.org to browse the complete model database.

Architecture

The package is organized into several modules:

• libkim.jl: KIM-API library loading and initialization
• constants.jl: KIM-API constants, enumerations, and unit systems
• model.jl: Low-level model creation and management
• species.jl: Species handling and validation
• neighborlist.jl: Neighbor list generation and callbacks
• highlevel.jl: High-level user interface

Performance

KIM_API.jl is designed for high performance:

• Pre-computed species mappings minimize lookup overhead
• Efficient neighbor list generation using NeighbourLists.jl
• Zero-copy data passing to KIM-API where possible
• Minimal Julia-C FFI overhead
• Automatic memory management prevents leaks

Citation

Consider citing the original KIM-API paper:

@article{tadmor2011potential,
  title={The potential of atomistic simulations and the knowledgebase of interatomic models},
  author={Tadmor, Ellad B and Elliott, Ryan S and Sethna, James P and Miller, Ronald E and Becker, Chandler A},
  journal={Jom},
  volume={63},
  number={7},
  pages={17},
  year={2011},
  publisher={Springer Nature BV}
}

License

KIM_API.jl is released under the MIT License. See the LICENSE file for details.

Acknowledgments

• The OpenKIM project for providing the KIM-API framework
• The Julia community for excellent packages like StaticArrays.jl and NeighbourLists.jl
• Contributors to the KIM model database