Skip to content

Latest commit

 

History

History
878 lines (658 loc) · 38.1 KB

CHANGELOG.rst

File metadata and controls

878 lines (658 loc) · 38.1 KB

This post-release makes some changes to the source-distribution build process: - A bug is fixed in the version-numbering script; this particularly affected Windows - A copies of the unit tests and documentation were mistakenly included

in tarballs, making them excessively large. This is no longer present.
  • Bug fixes
    • Fixed a bug in the version numbering mechanism affecting builds from sdist on Windows and environments where Git is unavailable
    • Reduce size of sdist, removing docs and tests from archive. (Restoring similar size to pre-v1.4.0 releases.)
  • Maintenance
    • Source builds automatically tested on Windows as part of build/deployment process

This release includes some significant modernisation and maintenance, as well as new features and performance enhancements.

  • Requirements

    • Python 3.8, 3.9 is no longer supported
    • Python 3.12 is supported
    • importlib_resources backport is no longer required
    • toolz is a new requirement
    • Some other dependency requirements have been increased in order to simplify maintenance and testing:
      • Minimum version of numpy increased from 1.19.3 to 1.24.0
      • Minimum version of matplotlib increased from 3.2 to 3.8
      • Minimum version of Pint increased from 0.19 to 0.22
      • Minimum version of PyYAML increased from 3.13 to 6.0
      • Minimum version of h5py increaased from 2.10 to 3.6
      • Minimum version of threadpoolctl increased from 1.0 to 3.0.
  • Improvements

    • A "reciprocal_spectroscopy" Pint context is made available in the unit registry for tricky conversions between reciprocal frequency/energy units. It is not active by default but can be enabled with e.g.

      (10 * ureg("1 / meV")).to("cm", "reciprocal_spectroscopy")

      This can also help to avoid divide-by-zero issues when performing energy <-> wavenumber conversions.

    • A Spectrum2DCollection class has been added to euphonic.spectra, which shares many features with Spectrum1DCollection

      • In particular, the iter_metadata method is recommended when one needs to iterate over the collection metadata without copying the spectral data to new objects.
    • Both Spectrum1DCollection and Spectrum2DCollection have a .from_spectra() constructor with an "unsafe" option which bypasses some consistency checks on the component data. This should only be used when confident that these will be consistent, such as when iterating over an existing collection.

    • Performance optimisations have been made to the "item getter" for Spectrum1DCollection (and Spectrum2DCollection); it should now be significantly faster to access and iterate over the contained spectra.

    • A euphonic.writers.phonon_website module has been added with a function to export QpointPhononModes to appropriate JSON for use with the phonon visualisation website http://henriquemiranda.github.io/phononwebsite/

      From the command-line, this can be accessed with a --save-web-json argument to the euphonic-dispersion tool.

  • Bug fixes

    • Metadata strings from Castep-imported PDOS data are now converted from numpy strings to native Python strings.
    • Spectra from CASTEP .phonon_dos files are now imported with units of reciprocal energy (e.g. 1/meV)
  • Maintenance

    • Cleared up unit-conversion-related warnings, de-cluttering the expected test suite output.
    • The Spectrum1DCollection class was significantly refactored to support addition of Spectrum2DCollection and improve maintainability.
    • Entire build system rework, migrating to pyproject.toml form with meson-python, cibuildwheel and removing versioneer to simplify future development and maintenance.
  • Requirements
    • packaging library added to dependencies.
  • Bug fixes
    • Fixed an error loading QpointPhononModes from JSON when there is a single q-point in the data
  • Improvements
    • When loading .castep_bin files, explicitly check the CASTEP version number and give a useful error message if this is < 17.1. (These files are missing information about the unit cell origins, and would previously cause an error with an unhelpful message.)
  • Maintenance
    • Compatibility fix for spglib 2.4 update: a new sanity-check in spglib raises TypeError when using empty unit cell and this needs handling when looking for high-symmetry labels
    • Compatibility fix for Numpy 2.0 update: avoid some broadcasting issues with array shape returned by np.unique
    • Update reference to scipy.integrate.simpson (scipy.integrate.simps is deprecated)
    • Filter out spglib deprecation warnings caused by SeeK-path.
  • Maintenance
    • Updated versioneer for compatibility with Python 3.12
    • In tests, avoid checking an attribute of 3D plots which is unreliable in recent matplotlib versions
    • Update readthedocs configuration to fix documentation builds
  • Requirements
    • Python 3.7 is no longer supported
    • Minimum version of scipy increased from 1.1 to 1.10
      • This requires numpy >= 1.19.5
    • Minimum version of matplotlib increased from 2.2.2 to 3.2.0
    • Minimum version of pint increased from 0.10.1 to 0.19
    • Minimum version of h5py increaased form 2.8 to 2.10
  • Improvements
    • Added variable-width broadening for 1-D and 2-D spectra. An arbitrary Callable can be provided relating the axis position to Gaussian or Lorentzian width parameter. euphonic-dos and euphonic-powder-map CLI tools accept polynomial coefficients as input. The broadening is implemented with the fast approximate interpolation method already available for adaptive broadening of DOS.
    • Added features to Spectrum classes
      • Added copy() methods returning an independent duplicate of data
      • Added __mul__ and __imul__ methods to Spectrum classes. This allows results to be conveniently scaled with infix notation * or *=
    • Added --scale parameter to euphonic-dos, euphonic-intensity-map, euphonic-powder-map to allow arbitrary scaling of results from command-line. (e.g. for comparison with experiment, or changing DOS normalisation from 1 to 3N.)
  • Bug Fixes:
    • Changed the masking logic for kinematic constraints: instead of requiring energy bin to _entirely_ fall within accessible range at Q-bin mid-point, unmask bins if _any_ part of energy range is accessible at this Q value. This gives much more intuitive behaviour, especially for narrow angle ranges.
  • Improvements

    • Added "prefer_non_loto" option to Castep .phonon file importers. When this is enabled, a block of q-points are encountered with splitting directions, and one q-point does not have a splitting direction, the data at this "exact" q-point is preferred and the other weights in the group are set to zero.

      This provides the intended behaviour of the Abins Castep parser and should give a reasonable result for Gamma-point only Castep calculations.

      The option is disabled by default, so existing scripts will not be affected.

  • Bug Fixes:

    • Allow color to be passed as an extra kwarg to plot_1d and plot_1d_to_axis. Previously this caused a TypeError.
    • Fix bug where Py_None was not incremented before returning from calculate_phonons() in the C-extension causing a deallocation crash
    • Support phonopy.yaml files from Phonopy versions >= 1.18, which have moved the data relating to dipole-dipole corrections. (i.e. Born effective charges, static dielectric tensor and a related unit conversion factor.)
  • Maintenance:

    • A deprecation in Numpy 1.25, which indirectly caused a test failure, has been addressed.
  • Improvements:
    • Euphonic now tests on Python 3.11
    • Euphonic now provides PyPI wheels for Python 3.11
  • New features:
    • You can now perform linear interpolation of phonon frequencies and eigenvectors with the Brille library using the new euphonic.brille.BrilleInterpolator object. This should provide performance improvements for large unit cells which require the dipole correction.
    • There is a new command-line tool euphonic-brille-convergence to assist with choosing the BrilleInterpolator.from_force_constants arguments to achieve the desired accuracy.
    • Brille interpolation can be accessed from the euphonic-powder-map tool using the new --use-brille, --brille-grid-type, --brille-npts and --brille-npts-density arguments.
  • New features:
    • There is a new function ForceConstants.from_total_fc_with_dipole to allow reading force constants from other programs which contain long-ranged dipole-dipole interactions.
  • Bug fixes:
    • Avoid occasional segmentation faults when using OpenBLAS, workaround for #191
    • Correctly read force constants from Phonopy with dipole-dipole interactions, see #239.
  • Changes:
    • Support for Python 3.6 has been dropped. This has also resulted in changes to the following dependencies:
      • numpy requirement increased from 1.12.1 to 1.14.5
      • scipy requirement increased from 1.0.0 to 1.1.0
      • pint requirement increased from 0.9 to 0.10.1
      • matplotlib requirement increased from 2.0.0 to 2.2.2
      • h5py requirement increased from 2.7.0 to 2.8.0
    • The following deprecated features have been removed:
      • The return_mode_widths argument in ForceConstants.calculate_qpoint_phonon_modes and ForceConstants.calculate_qpoint_frequencies has been removed
      • The eta_scale argument in calculate_qpoint_phonon_modes/frequencies has been removed
      • The alias command-line tool argument --weights has been removed
      • The alias arguments x_label, y_label, y_min and y_max to plot_1d/2d have been removed
      • The modes_from_file and force_constants_from_file functions from euphonic.cli.utils have been removed
      • Calling broaden on a Spectrum with uneven bin widths without specifying the method='convolve' argument will now raise a ValueError
    • DOS and PDOS calculated by the calculate_dos and calculate_dos_map methods of QpointPhononModes and QpointFrequencies, and QpointPhononModes.calculate_pdos are now calculated per atom rather than per unit cell (integrated area is 3 rather than 3*N_atom). This is to keep consistency with the structure factors calculated by QpointPhononModes.calculate_structure_factor which are calculated per atom.
    • The option average_repeat_points when importing q-point modes or frequencies from a CASTEP .phonon file with QpointFrequencies/QpointPhononModes.from_castep is now True by default. To recover previous behaviour set this to False.
  • New Features:
    • Kinematic constraints have been implemented for 2-D S(q,w)-like data.
      • A function euphonic.spectra.apply_kinematic_constraints(Spectrum2d, **kwargs) -> Spectrum2D is implemented which masks out inaccessible data, replacing it with NaN.
      • Both direct-geometry and indirect-geometry are supported, by using the appropriate argument to set incident or final neutron energy.
      • This function is exposed to the euphonic-powder-map tool, so these plots can be produced directly from the CLI.
      • Some parameters from real-world instruments are collected in the documentation for convenience.
    • There is a new function euphonic.util.convert_fc_phases, which converts a force constants matrix which uses the atom coordinates in the phase during interpolation (Phonopy-like), to one which uses the cell origin coordinates (Euphonic, CASTEP-like).
    • When importing q-point modes or frequencies from a CASTEP .phonon file, a new option (average_repeat_points=True) allows repeated entries (with the same q-point index) to be identified and their weights divided down by the number of entries. This option should give better statistics for sampling meshes that include the Gamma-point with LO-TO splitting.
  • Improvements:
    • Documentation on the shape and format of the force constants, and how to read them from other programs has been improved.
    • The euphonic.util.get_qpoint_labels function, which is called when importing band-structure data to identify and label significant points, primarily identifies these points by searching for turning-points in the band path. The function will now also pick up any q-point that appears twice in succession. This is a common convention in band-structure calculations and helps with edge-cases such as when the path passes through a high-symmetry point without changing direction. This may pick up some previously-missing points in band-structure plots generated with euphonic-dispersion and euphonic-intensity-map
  • Bug fixes:
    • Allow read of phonopy.yaml quantities in 'au' (bohr) units. Previously this was interpreted as an astronomical unit by Pint.
  • Improvements:
    • The euphonic-dos, euphonic-dispersion and euphonic-intensity-map command-line tools can now read files that don't contain eigenvectors, if eigenvectors are not required for the chosen options.
    • A new --save-json option is available for command-line tools which produce plots, this will output the produced spectrum to a Euphonic .json file.
    • There is now the option to use a fast, approximate variable-width broadening method when adaptively broadening dos:
      • Added new adaptive_method and adaptive_error arguments for calculate_dos which specify which adaptive broadening method to use (reference or fast) and an acceptable error level when using the fast method.
      • Fast adaptive broadening can be used in the euphonic-dos tool with the --adaptive-method and --adaptive-error arguments.
  • Changes:
    • euphonic.cli.force_constants_from_file and modes_from_file have been deprecated in favour of euphonic.cli.load_data_from_file.
    • Using Spectrum1D/1DCollection/2D.broaden on an axis with unequal bin widths is now deprecated, as broadening is performed via convolution, which is incorrect in this case. In the future, this will raise a ValueError. To broaden anyway, method='convolve' can be supplied, which will just emit a warning.
  • New Features:
    • New Spectrum1D.to_text_file and Spectrum1DCollection.to_text_file methods to write to column text files
    • An expanded and consistent set of styling options is made available for command-line tools that produce plots.
    • Consistent styling and advanced changes can be made using Matplotlib stylesheet files, either as a CLI argument or using matplotlib.style.context() in a Python script.
  • Improvements:
    • Internally, plot theming has been adjusted to rely on Matplotlib style contexts. This means user changes and style context are more likely to be respected.
    • Additional aliases for plot arguments in the command-line tools have been added, for example either --x-label or --xlabel can be used.
  • Changes:
    • x_label, y_label, y_min and y_max in euphonic.plot functions have been deprecated in favour of xlabel, ylabel, ymin and ymax respectively, to match the Matplotlib arguments they refer to, and to match other arguments like vmin, vmax.
  • Improvements:
    • Wheels are now provided with PyPI releases
    • Type hinting is now handled more consistently across different Euphonic classes and functions
  • Bug Fixes:
    • Will no longer raise a KeyError reading from phonopy.yaml if physical_unit key is not present, instead will assume default units
    • Can now read Phonopy BORN files where the (optional) NAC conversion factor is not present
  • Bug fixes:
    • The scaling of S(Q,w) as produced by StructureFactor.calculate_sqw_map was incorrect, and did not correctly scale with energy bin size (given its units are now length**2/energy). This has been fixed, and S(Q,w) scale has changed by a factor of (hartee to energy bin unit conversion)/(energy bin width magnitude). e.g. if using an energy bin width of 0.1 meV, the new S(Q,w) will be scaled by 2.72e4/0.1 = 2.72e5. The original structure factors can now be correctly recovered by multiplying S(Q,w) by the energy bin width.
  • Euphonic can now calculate neutron-weighted partial density of states, and has new Spectra features to handle PDOS data:
    • Added QpointPhononModes.calculate_pdos method
    • Added QpointFrequencies.calculate_dos_map method
    • New Spectrum1D.__add__ method, which adds 2 spectra together
    • New Spectrum1DCollection.__add__ method, which concatenates 2 collections
    • Enabled indexing of Spectrum1DCollection by a sequence
    • Added Spectrum1DCollection.group_by method, which allows grouping and summing spectra by metadata keys e.g. group_by('species')
    • Added Spectrum1DCollection.select method, which allows selection of spectra by metadata keys e.g. select(species='Si')
    • Added Spectrum1DCollection.sum method, which sums all spectra in a collection
    • Added -w={'coherent-dos','incoherent-dos','coherent-plus-incoherent-dos'} neutron-weighted PDOS options to euphonic-dos and euphonic-powder-map
    • Added --pdos options for plotting specific species PDOS to euphonic-dos and euphonic-powder-map
    • Deprecated --weights command-line argument in favour of --weighting for consistency with calculate_pdos
  • Improvements:
    • LICENSE and CITATION.cff files are now included in Euphonic's installation
    • Add ability to interactively change the colormap intensity limits in euphonic-powder-map
    • euphonic-optimise-dipole-parameter can now read from Phonopy sources
    • euphonic-optimise-dipole-parameter can now also be used for non-polar materials to get general per-qpoint timings
    • Dimensioned Euphonic properties (e.g. frequencies, cell_vectors) now have setters so can be set, previously this would raise an AttributeError
  • Changes:
    • The units of density of states as produced by calculate_dos have changed from dimensionless to 1/energy
    • The scaling of density of states has also changed. Previously the integration would sum to 1 (if the x_data were converted to Hartree units), now the integration will sum to 3N in the same units as x_data
    • StructureFactor.structure_factors have been changed to be in absolute units per atom (rather than per unit cell) so will have changed by a factor of 1/2*n_atoms, this formulation change has been reflected in the calculate_structure_factor docstring
    • The default unit of StructureFactor.structure_factors has been changed from angstrom**2 to millibarn
    • The unit of S(Q,w) as produced by StructureFactor.calculate_sqw_map has changed dimension from length**2 to length**2/energy. Also, as its unit is derived from the input StructureFactor object, its default units are now millibarn/meV
    • The eta_scale argument in calculate_qpoint_phonon_modes has been deprecated, dipole_parameter should be used instead.
    • This means the euphonic-optimise-eta script has been renamed to euphonic-optimise-dipole-parameter.
  • Improvements:
    • Added broaden method to Spectrum1DCollection
  • Changes:
    • The return_mode_widths argument in calculate_qpoint_phonon_modes has been deprecated in favour of return_mode_gradients. The mode widths can still be obtained from the mode gradients with util.mode_gradients_to_widths
  • Bug fixes:
    • Fixed memory leak when using the C extension and making multiple calls to calculate_qpoint_phonon_modes/frequencies
    • Fixed bug which resulted in incorrect energy bins being generated in euphonic-powder-map if units other than meV are used and --e-max and --e-min aren't specified
    • Use correct number of energy bins in euphonic-intensity-map, euphonic-powder-map and euphonic-dos. Previously only ebins - 1 bins were generated
  • New Features:
    • New Crystal.get_symmetry_equivalent_atoms method which uses spglib to get the symmetry operations and equivalent atoms under each operation
  • Improvements:
    • Added symmetrise argument to QpointPhononModes.calculate_debye_waller which will symmetrise it under the crystal symmetry operations. This means that there will no longer be a discrepancy between DebyeWaller calculated on a symmetry-reduced or full Monkhorst-Pack grid. By default, symmetrise=True
    • Added frequencies_min argument to calculate_debye_waller to exclude very small frequencies. This will also exclude negative frequencies. This improves on the previous behaviour which only excluded gamma-point acoustic modes, so would miss small/negative frequencies elsewhere
    • Loading the LAPACK libraries for the C extension now uses the interface provided by scipy for cython instead of loading directly from a DLL. The new method means we don't have to guess the DLL filename anymore!
  • Changes:
    • New dependency on spglib>=1.9.4
    • Fixed formula in calculate_debye_waller docstring to match actual implementation: moved 1/2 factor and added explicit q-point weights
  • New Features:

    • New command-line tool euphonic-powder-map allows generation and plotting of powder-averaged S(|q|,w) and DOS maps.
    • New QpointFrequencies object which allows storage of frequencies without eigenvectors, meaning that memory usage can be reduced if eigenvectors are not required.
    • StructureFactor now has a weights attribute and can be used to calculate DOS with calculate_dos and get dispersion with get_dispersion
    • Spectrum1D, Spectrum1DCollection and Spectrum2D objects have a new metadata attribute, see their docstrings for details
    • Euphonic can now read DOS/PDOS from CASTEP .phonon_dos files with Spectrum1D.from_castep_phonon_dos and Spectrum1DCollection.from_castep_phonon_dos
    • Adaptive broadening is now available for DOS, which can obtain a more representative DOS than standard fixed-width broadening. See the docs for details
    • Adaptive broadening can be used in the euphonic-dos tool with the --adaptive argument
  • Improvements:

    • Improved default behaviour for C extension use and number of threads:

      • By default the C extension will be used if it is installed
      • By default the number of threads will be set by multiprocessing.cpu_count()
      • The environment variable EUPHONIC_NUM_THREADS can be used to set a specific number of threads, which takes priority over multiprocessing.cpu_count()
      • fall_back_on_python argument has been removed and superseded by the default use_c=None behaviour
      • threadpoolctl.threadpool_limits is used to limit the number of threads used by numerical libraries in Euphonic C function calls, resulting in better overall performance
    • Command-line interfaces have been refactored, giving a more uniform set of options and clearer sections of related arguments on the interactive help pages.

      • It is now possible where appropriate to specify Monkhorst-Pack sampling with a single-parameter --q-spacing as an alternative to setting Monkhorst-Pack divisions. This approach will account for the size and shape of reciprocal-lattice cells.
    • Build process tweaks

      • On Linux, the build process will now respect a user-defined C-compiler variable CC.
      • On Mac OSX, the build process will now respect a user-defined C-compiler variable CC. Homebrew library paths will only be set if CC is empty and the brew command is available.

      These tweaks are intended to facilitate Conda packaging.

  • Breaking changes:

    • The --q-distance argument to euphonic-intensity-map has been renamed to --q-spacing for consistency with other tools.
    • Debye-Waller calculation in euphonic-intensity-map is now enabled by setting --temperature, which no longer has a default value.
    • Default Monkhorst-Pack meshes (i.e. [6, 6, 6] in euphonic-dos and [20, 20, 20] in sample_sphere_structure_factor()) have been replaced by default grid-spacing values.
    • The scaling of density of states has changed, due to a change in implementation
  • There have been some major changes and improvements to spectra, plotting and command line tools, including:
    • New command line tool euphonic-intensity-map for plotting weighted 2D Spectra e.g. Coherent neutron S(Q,w)
    • Existing command line tools euphonic-dispersion and euphonic-dos have been updated to also read force constants and Phonopy files. Arguments are also more consistent across tools so some may have changed, check the command line tool help for details.
    • New Spectrum1DCollection object for containing 1D spectra with a shared x-axis (e.g. phonon dispersion modes)
    • New plot_1d_to_axis and plot_2d_to_axis functions to allow plotting on specific axes
    • get_bin_centres and get_bin_edges utility functions on spectra
    • The ratio argument to plot_2d has been removed, it should no longer be required due to better management of relative axis sizes.
    • The btol argument to plot_1d has been removed, it is recommended to use Spectrum1D.split() or Spectrum1DCollection.split() instead.
    • The plot_dispersion function has been removed. It is now recommended to plot dispersion using plot_1d(QpointPhononModes.get_dispersion()). See docs for details.
  • Other changes:
    • Some of Euphonic's dependency version requirements have been changed, but can now be relied on with more certainty due to better CI testing. This includes:
      • numpy requirement increased from 1.9.1 to 1.12.1
      • matplotlib requirement increased from 1.4.2 to 2.0.0
      • pint requirement decreased from 0.10.1 to 0.9
      • h5py requirement decreased from 2.9.0 to 2.7.0
      • pyyaml requirement decreased from 5.1.2 to 3.13
  • Improvements:
    • yaml.CSafeLoader is now used instead of yaml.SafeLoader by default, so Phonopy .yaml files should load faster
    • Metadata __euphonic_version__ and __euphonic_class__ have been added to .json file output for better provenance
  • Bug fixes:
    • Fix read of Phonopy 'full' force constants from phonopy.yaml and FORCE_CONSTANTS files
    • Fix structure factor calculation at gamma points with splitting, see #107
    • Change broadening implementation from scipy.signal.fftconvolve to use scipy.ndimage functions for better handling of bright Bragg peaks, see #108
  • New Features:
    • Added weights as an argument to ForceConstants.calculate_qpoint_phonon_modes, this will allow easier use of symmetry reduction for calculating density of states, for example.
    • Modules have been added to support spherical averaging from 3D q-points to mod(q)
      • euphonic.sampling provides pure functions for the generation of points on (2D) unit square and (3D) unit sphere surfaces.
      • A script is provided for visualisation of the different schemes implemented in euphonic.sampling. This is primarily intended for education and debugging.
      • euphonic.powder provides functions which, given force constants data, can use these sampling methods to obtain spherically-averaged phonon DOS and coherent structure factor data as 1D spectrum objects. (It is anticipated that this module will grow to include schemes beyond this average over a single sphere.)
    • Added Crystal.to_spglib_cell convenience function
  • Changes:
    • The Scripts folder has been removed. Command-line tools are now located in the euphonic.cli module. The entry-points are managed in setup.py, and each tool has the prefix "euphonic-" to avoid namespace clashes with other tools on the user's computer. (e.g. euphonic-dos)
    • From an interactive shell with tab-completion, one can find all the euphonic tools by typing "euphonic-<TAB>".
    • Changed arguments for util.get_qpoint_labels(Crystal, qpts) to util.get_qpoint_labels(qpts, cell=None) where cell = Crystal.to_spglib_cell()
  • Bug fixes:
    • Correctly convert from Phonopy's q-point weight convention to Euphonic's when reading from mesh.yaml (see 7509043)
    • Avoid IndexError in ForceConstants.calculate_qpoint_phonon_modes when there is only one q-point (which is gamma) and splitting=True
  • New Features:
    • A system has been added for reference data in JSON files. These are accessed via euphonic.utils.get_reference_data and some data has been added for coherent scattering lengths and cross-sections. This system has been made available to the calculate_structure_factor() method; it is no longer necessary to craft a data dict every time a program uses this function.
  • Changes:
    • Fixed structure factor formula in docs (|F(Q, nu)| -> |F(Q, \\nu)|^2 and e^(Q.r) -> e^(iQ.r))
  • Bug fixes:
    • Fix 'born':null in ForceConstants .json files when Born is not present in the calculation (see c20679c)
    • Fix incorrect calculation of LO-TO splitting when reduce_qpts=True, as the 'reduced' q rather than the actual q was used as the q-direction (see 3958072)
    • Fix interpolation for materials with non-symmetric supcercell matrices, see #81
    • Fix interpolation for force constants read from Phonopy for materials that have a primitive matrix and more than 1 species, see #77
  • Breaking Changes:
    • There has been a major refactor, for see the v0.3.0 docs for how to use, or here for refactor details
    • Python 2 is no longer supported. Supported Python versions are 3.6, 3.7 and 3.8
  • New Features:
    • Euphonic can now read Phonopy input! See the docs for details.
  • Improvements:
    • Added fall_back_on_python boolean keyword argument to ForceConstants.calculate_qpoint_phonon_modes to control whether the Python implementation is used as a fallback to the C extension or not, see #35
    • Added --python-only option to setup.py to enable install without the C extension
  • Bug fixes:
    • On reading CASTEP phonon file header information, switch from a fixed number of lines skipped to a search for a specific line, fixing issue #23
    • Fix NaN frequencies/eigenvectors for consecutive gamma points, see #25
    • Fix issue saving plots to file with dispersion.py, see #27
    • Fix incorrect frequencies at gamma point when using dipole correction in C, #45
  • Bug fixes:
    • Add MANIFEST.in for PyPI distribution
  • Bug fixes:
    • Cannot easily upload C header files to PyPI without an accompanying source file, so refactor C files to avoid this
  • There are several breaking changes:
    • Changes to the object instantiation API. The former interface InterpolationData(seedname) has been changed to InterpolationData.from_castep(seedname) in anticipation of more codes being added which require more varied arguments.
    • Changes to the Debye-Waller calculation API when calculating the structure factor. The previous dw_arg kwarg accepted either a seedname or length 3 list describing a grid. The new kwarg is now dw_data and accepts a PhononData or InterpolationData object with the frequencies calculated on a grid. This is to make it clearer to the user exactly what arguments are being used when calculating phonons on the grid.
    • Changes to parallel functionality. The previous parallel implementation based on Python's multiprocessing has been removed and replaced by a C/OpenMP version. This has both better performance and is more robust. As a result the n_procs kwarg to calculate_fine_phonons has been replaced by use_c and n_threads kwargs.
  • Improvements:
    • The parallel implementation based on Python's multiprocessing has been removed and now uses C/OpenMP which both has better performance and is more robust
    • Documentation has been moved to readthedocs and is more detailed
    • Clearer interface for calculating the Debye-Waller factor
    • Better error handling (e.g. empty InterpolationData objects, Matplotlib is not installed...)
  • Bug fixes:
    • Fix gwidth for DOS not being converted to correct units
    • Fix qwidth for S(Q,w) broadening being incorrectly calculated