Release notes for v1.3


  • Weighted excitation generators for CCMC and FCIQMC as published in V. A. Neufeld, and A. J. W. Thom, arXiv:1808.05093 and A. A. Holmes, H. J. Changlani, and C. J. Umrigar, J. Chem. Theory Comput. 12, 1561 (2016) and S. D. Smart, G. H. Booth, and A. Alavi, unpublished.
  • Complex DMQMC.
  • CMake-based build system also available.
  • Ability to restart the state of the dSFMT RNG stream, allowing for restarted calculations to have the same Markov chain as single calculations. Enabled by default. Can be disabled in the restart options.
  • Quasi-Newton propagation for complex particles.
  • Shared memory MPI with the integrals.
  • Pyhande: reweighting plot to check for population bias as published in W. A. Vigor, J. S. Spencer, M. J. Bearpark, and A. J. W. Thom, J. Chem. Phys, 142, 104101 (2015).


  • write_frequency now is in units of report loops rather than Monte Carlo cycles.



Bug Fixes

  • Even selection weighting had an initialization bug affecting any calculation with more than one MPI process where one process has no single excitations. The effect is that for those cycles without singles, the even_selection probability for the singles will be uninitialized, and possibly not zero. For a sufficiently large calculation there will be a low probability that this is the case, and if there are ever single excitations on the processor, that number will be used, and will continue to be used if there are no singles at some later date. It is expected that there will be no notable effect after equlibration. Incorrect even selection weightings only affect the efficiency of the selection, and will not in general introduce a bias. Effects on systems without single excitations (e.g. UEG and Hubbard models) are undefined. The bug-fix changes Markov chains.
  • Even selection also had a bug that caused integer overflow when the number of MPI processes raised to the power of (coupled cluster level - 1) is bigger then a 32 bit integer.