Supported Models#

ALCHEMI Toolkit ships wrappers for several machine-learning interatomic potentials (MLIPs) and classical force fields. Every wrapper implements the BaseModelMixin interface and declares its capabilities via a ModelConfig.

For a step-by-step guide on wrapping your own model, see the Models: Wrapping ML Interatomic Potentials.

Machine-Learned Potentials#

Neural-network potentials that learn interatomic interactions from quantum mechanical reference data.

Wrapper	Energies	Forces	Stresses	PBC	Needs PBC	Autograd Forces	Extra Inputs	Neighbor Fmt.
`MACEWrapper`	✓	✓	✓	✓	✗	✓	—	COO
`AIMNet2Wrapper`	✓	✓	✓	✓	✗	✓	charge	MATRIX
`UMAWrapper`	✓	✓	✓	✓	✗	✓	charge, spin	COO
`DemoModelWrapper`	✓	✓	✗	✗	✗	✓	—	—

Physical / Classical Models#

Analytical force fields and correction terms based on known physical functional forms.

Wrapper	Energies	Forces	Stresses	PBC	Needs PBC	Autograd Forces	Extra Inputs	Neighbor Fmt.
`LennardJonesModelWrapper`	✓	✓	✓	✓	✗	✗	—	MATRIX
`DFTD3ModelWrapper`	✓	✓	✓	✓	✗	✗	—	MATRIX
`PMEModelWrapper`	✓	✓	✓	✓	✓	✗	charges	MATRIX
`EwaldModelWrapper`	✓	✓	✓	✓	✓	✗	charges	MATRIX

Note

PipelineModelWrapper is excluded from the tables above because its capabilities are synthesized at runtime from the sub-models it composes (see PipelineModelWrapper).

Model Composition#

Models can be combined using the + operator for simple additive composition, or the explicit PipelineModelWrapper API for dependent pipelines with shared autograd groups and inter-model wiring. See Models: Wrapping ML Interatomic Potentials for full details.

# Simple: sum energies, forces, stresses
combined = mace_model + dftd3_model

# Advanced: dependent pipeline with shared autograd
from nvalchemi.models.pipeline import PipelineModelWrapper, PipelineGroup, PipelineStep

pipe = PipelineModelWrapper(groups=[
    PipelineGroup(
        # AIMNet2 outputs "charges"; Ewald requires "charges" — auto-wired.
        steps=[aimnet2, ewald],
        use_autograd=True,
    ),
    PipelineGroup(steps=[dftd3]),
])

References#

If you use any of the model wrappers provided by ALCHEMI Toolkit, please cite the original publications for the underlying methods.

Model	Citation
MACE	Batatia, I. et al. “MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields.” Advances in Neural Information Processing Systems (NeurIPS), 2022. openreview.net/forum?id=YPpSngE-ZU
MACE-MP-0 (foundation)	Batatia, I. et al. “A foundation model for atomistic materials chemistry.” arXiv:2401.00096, 2023. doi:10.48550/arXiv.2401.00096
AIMNet2	Anstine, D. M., Zubatyuk, R. & Isayev, O. “AIMNet2: a neural network potential to meet your neutral, charged, organic, and elemental-organic needs.” Chem. Sci. 16, 10228–10244, 2025. doi:10.1039/D4SC08572H
DFT-D3(BJ)	Grimme, S. et al. “A consistent and accurate ab initio parametrization of density functional dispersion correction (DFT-D) for the 94 elements H-Pu.” J. Chem. Phys. 132, 154104, 2010. doi:10.1063/1.3382344
	Grimme, S., Ehrlich, S. & Goerigk, L. “Effect of the damping function in dispersion corrected density functional theory.” J. Comput. Chem. 32, 1456–1465, 2011. doi:10.1002/jcc.21759
Lennard-Jones	Jones, J. E. “On the Determination of Molecular Fields.” Proc. R. Soc. Lond. A 106 (738), 463–477, 1924. doi:10.1098/rspa.1924.0082
Ewald Summation	Ewald, P. P. “Die Berechnung optischer und elektrostatischer Gitterpotentiale.” Ann. Phys. 369 (3), 253–287, 1921. doi:10.1002/andp.19213690304
Particle Mesh Ewald	Darden, T., York, D. & Pedersen, L. “Particle mesh Ewald: An `Nlog(N)` method for Ewald sums in large systems.” J. Chem. Phys.* 98 (12), 10089–10092, 1993. doi:10.1063/1.464397