This library is a work in progress.

Library for simulating atomically precise manufacturing.

Goals (for ab initio simulations):

• Designed for GPU acceleration with OpenCL
  • Perform all computations in FP32, with compensated summation when necessary.
• Real-space formalism
  • Removes FFTs, an $O(n^2\log{n})$ bottleneck.
  • Most DFT libraries (Gaussian, GAMESS, TeraChem) use the plane-wave/basis-set formalism. Basis sets add unnecessary complexity and obfuscate pathways to linear scaling.
• Linear scaling
  • Adaptive mesh refinement at the per-electron granularity, instead of the typical global granularity. This reduces the bottleneck from $O(n^2)$ to $O(n\log{n})$.
  • Linear scaling likely possible for insulators, with localized electrons.
• No pseudopotentials
  • Large-core pseudopotentials interact with the XC functional, polluting simulation results.
  • Similar to the issue with the AO basis: BSSE pollutes simulation results.

Also included (for semiempirical simulations):

• Swift bindings for xTB
• Fast matrix diagonalization kernel