Version 18.5.1 of BoltzTraP2

    This is a small bugfix release.

    Bug fixes:
        - Include Eigen in the PyPI package.

Version 18.4.2 of BoltzTraP2

This is a small bugfix release.

Bug fixes:
    - Fix a problem with unit conversion when loading group velocities
    from VASP data (thanks to Gyu Hyeong Kim for reporting).
    - Update the README.

Version 18.4.1 of BoltzTraP2

New features:
    - Experimental ABINIT support.
    - Replace C++11 code with old-style C++ to increase compiler
    compatibility.
    - Replace custom C++ matrix routines with the Eigen library.
    - NMake some error messages more informative.

Bug fixes:
    - Use scipy.linalg.lstsq() instead of the numpy equivalent, to
    make the band reconstruction more robust.
    - Set the right default scattering model for the "dope" command
    (thanks to gabo13279 for the report).
    - Fix bandana(), which was dropping one band too many (thanks to
    Manuel Zingl for the report).
    - Fix a bug in the calculation of RH when custom scattering  rates
    were used (thanks to Manuel Zingl for the report).
    - Replace some dot products with einsum() to get around some
    out-of-memory errors on Mac, at the cost of some performance.

Other notes:
    - Compiling with CLang on macOS seems to introduce numerical
    errors in the band reconstruction. We are investigating this issue
    (apparently related to the Accelerate Framework) and will provide
    a fix as soon as possible. In the meantime, macOS users are encouraged
    to compile BoltzTraP2 with GCC. CLang on Linux is known to work.

Version 18.3.2 of BoltzTraP2

This version fixes a bug in v18.3.1 whereby the default scattering
model was not correctly set.

Version 18.3.1 of BoltzTraP2

Implement a new "dope" subcommand in the interface

    - Change the command-line syntax for temperatures. Now they
      are specified as a list of floats of ranges in the form
      "start:end:delta".
    - Use the same syntax for doping in the new command.
    - Change the name of refine_mu0 to solve_for_mu, which is now
      used to fit the chemical potential in more general cases.
    - Change the criterion for moving the chemical potential to the
      center of a gap.
    - Add a new "refine" boolean parameter to the function, which
      lifts the constraints that the returned value is a histogram
      bin.
    - Use refined estimates of the intrinsic chemical potential across
      the board.
    - Use the same value of N (relative to the neutral atom) in all
      "old-style" output files.

BoltzTraP2, version 18.1.2

BoltTrap2, version 18.1.1

BoltzTraP2, version 17.12.1

BoltzTraP2 release v17.11.6

BoltzTraP2 v17.11.3

BoltzTraP2, version 17.11.2