HIP: Hessian Interatomic Potentials without derivatives.

Hessian Interatomic Potentials (HIPs) predict molecular Hessians 10 to 100 times faster and more accurately than traditional methods in computational chemistry.

• Why it matters: Accurate Hessians are essential for many computational workflows but are costly to compute and scale poorly with system size, limiting efficiency and applicability.
• What they did: A deep learning model was developed that constructs SE(3)-equivariant, symmetric Hessians from irrep features using a graph neural network, trained on diverse molecular data.
• The result: HIP Hessians outperform existing approaches in tasks like transition state search and vibrational analysis, enabling faster, more memory-efficient computations and broadening practical applications.