Skip to content

Conical intersection optimization, Normal Mode visualization, Improved Bends and Torsions

Choose a tag to compare

@eljost eljost released this 13 Dec 11:10

Addition

  • Calculator for Conical intersection optimization (type: conical), implementing the updated branching plane algorithm. See examples/opt/17_orca_conical_intersection.
  • Simple ursina-based script that can read and visualize pysisyphus HDF5 Hessians to animate imaginary modes (scripts/modes3d.py)
  • Dipole moment calculation in OpenMM calculator
  • New internal coordinates: BondedFragment and DummyTorsion; suitable for optimization of water interaction energies, as needed for parametrizing CGENFF.
  • Alternative implementations of bends and torsions, using atan2 instead of acos, that are numerically more stable. 2nd derivatives of Torsions are now calculated using the code-generated derivatives, starting from the atan2 expression.
  • Support for exit sign to cleanly shutdown calculations
  • perf driver to benchmark calculation-speedups with varying pal
  • Implemented Bofills TS-Hessian update hessian_update: ts_bofill
  • Support for ALPB solvent in XTB calculator
  • Frozen atoms can now be excluded from the internal coordinate setup algorithm (please see the documentation)

Bugfixes

  • pr_nto argument to OverlapCalculator was ignored
  • Don't try to call thermoanalysis, when it is not installed
  • IRC classes now respect out_dir argument
  • Requesting 2nd derivative of Rotations w.r.t. Cartesian coordinates led to a crash

Miscellaneous

  • Version string is now dumped to RUN.yaml
  • Improved Growing-Newton-Trajectory class
  • Improved printing of internal coordinates in pysistrj --internals
  • Relaxed scans are more robust
  • Speeded up some internal coordinates by avoiding calls to numpy methods
  • MO-coefficients can now be set in .gbw files
  • Initial dimer calculation can now be read from a HDF5 Hessian imaginary mode
  • Energy change is now reported in the first column of an optimization
  • Switched from plain down-scaling to shifted-Newton-step when proposed RFO-step is too big
  • Interpolation in internal coordinates is now more robust