Numeric Precision & Data Standards#

MolSysMT enforces strict numeric precision standards and array dimension invariants across all native data structures and converters.

Floating Point Precision Standards#

  • 32-Bit Floating Point (float32): Default precision for 3D coordinate arrays (n_structures, n_atoms, 3) and box vectors. float32 balances spatial accuracy (sub-picometer resolution) with optimal memory footprint and GPU hardware compatibility.

  • 64-Bit Floating Point (float64): Preserved for high-precision energy evaluations, double-precision numerical integrations, and physical unit conversions.

Dimension and ID Invariants#

  • Coordinate Arrays: Always shaped as (n_structures, n_atoms, 3) in nanometers.

  • Box Matrices: Always shaped as (n_structures, 3, 3) in nanometers.

  • Time Arrays: 1D NumPy arrays in picoseconds.

  • String Identifier Invariant: In native MolSysMT objects (such as molsysmt.Topology and molsysmt.MolSys), element ID fields (*_id) are stored strictly as normalized strings. Numeric inputs are automatically converted to string representations to maintain invariants across form adapters.