# Changelog

## Sierra v1.5.0

• New OrbNet model "Denali"
• Adds support for charged molecules
• Adds support for Li, Mg, Ca, K, and Na ions and complexes
• Improved geometry optimization stability for all OrbNet methods
• Improved accuracy for small molecules and intermolecular complexes involving small molecules
• New feature: Thermodynamic properties
• New feature: Excited states via linear response can be run using the ExcitationInput
• Improved solvation model
• Solvation is now available for DFT with parameters optimized for the B97-3c functional
• Significantly improved electrostatics due to reparameterization of Generalized-Born method
• Solvation is now rotationally invariant
• Automatic selection of suitable parameter set
• Larger coverage of the periodic table
• Improved conformer workflow
• Total computation time approximately halved
• On average, resulting conformers are lower in energy
• Improved xTB robustness and speed for molecules with challenging electronic structure
• Improved IAO populations

• More robust algorithm
• Available for spin-unrestricted calculations
• API changes

• Denali is now the default OrbNet method when creating a method with OrbNetMethod(); “Fuji” can still be accessed by calling OrbNetMethod(model=”fuji”)
• time_step has been changed to step_size in StringInputDetails
• The NMR attributes stored in NMRProperty have been changed from a plain Array type to Dict[int, Array], where each key in the dictionary corresponds to an atom index in the molecule. This change has been performed to better account for the sparsity in NMRProperties where all atoms do not have corresponding attributes.
• The NMR attributes stored in NMRProperty now contain the coupling constants j_coupling and k_coupling, stored as a Dict[Tuple[int, int], Array] where the key is a pair of atom indices.