d3q Files

IMPORTANT: 
Please download the QE+D3Q snapshots from here: https://drive.google.com/open?id=1Yr8MaWTznXatnySgg3-yYgOlsJ_w7XDV

Separate release are not currently supported for lack of time and may not compile with the most recent versions of QE!


v 1.1.3
* Bugfix for pool: scattering of charge density was changend in QE 6.2 but not here, could give very bad (i.e. absurd) result when using pools parallelism

v 1.1.2
 * Code d3_tk.x in SMA mode can now cutoff phonons with a mfp longer than sample size, input variables casimi_length and casimir_dir now renamed to sample_length and sample_dir, as they can used for both approaches
 * phonon modes can be sorted according to eigenmode continuity
 * compatibility with qe 6.2

v 1.1.1
* D3 calculation of metallic systems has been improved, only full bands used for the Sternheimer equation: improved convergence.
* Added debug options to d3 and qq2rr, added documentation for tools.
* Added manual grid shift from input to thermal2 codes.

v 1.1.0
* the phonon HWHM was mistakenly used in the SMA and CGP calculations, instead of the FWHM, thanks to Francesco Macheda for reporting and helping with debugging
* version of d3q.x prior to 1.1.0 would save the complex conjugate of the D3 matrix (there is some freedom in the definition) which would result in poor result for most grid sizes (1x1x1 and 2x2x2 are fine). Now qq2rr contains a workaround that detects this case and takes the CC of the input D3 matrices
* openmp parallelization of d3_asr3.x allows for much faster sum rule, good scalability tested up to 4 or 8 CPUs
* the local part of the XC functional was always assumed to be Perdew-Zunger, this causes no error for PZ functional, a tiny error for PBE functional, possibly large error for other functionals.

v 1.0.7
* nasty bug in d3_tk.x, CGP calculation, wrong prefactor! Many thanks to Erpan Arkin for reporting and helping with the debug!
* added calculation of JDOS and RMS of atomic displacements to d3_r2q.x
* better threatmenet of LO-TO along paths
* support for random grids improved (should work in SMA)

v 1.0.6
* nasty bug in d3_qq2rr.x where the 3rd order matrix would not be read correctly has been fixed: PLEASE REGENERATE THE FORCE CONSTANT FILES!!

v 1.0.5
* tiny fix on the previous version (symmetrization removed for scattering modes as it is not necessary and does not respect detailed balance, i.e. possible funny results with small grids)
* fixed a bug when using MPI parallelism and isotopes (isotopes would be ignored on CPUs after the first one)

v 1.0.4
* one important change: quantities (linewidth, self-energy, scattering matrices, velocities) pertaining to degenerate phonon modes are now averaged out. This prevents broken symmetry when doing thermal conductivity calculations.

v 1.0.3
* fixed a case where a fuction result was used in an openmp reduce clause, causing segmentation fault on some intel compilers
* removed a mention of a non-existing unused subroutine, causing some compilers to rightfully complain
* not using vdCos and vdSin with intel compiler made easier (remove -D__HASVTRIG from thermal2/Makefile)
* openmp no more enable by default (it was actually unintended )

v 1.0.2
* fixed old style namelist in tutorial example input file (and fixed code to still read those)

v 1.0.1
* remove line from PROGRAM_sparse.f90 that prevented compilation (thanks S. Appalakondaiah for reporting)
* calculation of 3-body force constant now done in two passess (100x times faster)
* 3d order acoustic sum rule now imposed in-place (faster and much lower memory usage)

v 1.0
everything!