Small bug fix for FP underflows appearing in systems with large CNs.

What's Changed

• Use trusted publishing for PyPI upload by @marvinfriede in #321
• Guard against FP underflow by @marvinfriede in #324
• Bump version to 4.0.1 by @marvinfriede in #325

Full Changelog: v4.0.0...v4.0.1

New major release for minor, but breaking API change and a couple new features.

Breaking Changes

• new_d4_model now requires an allocatable error_type as first argument.
  • old: call new_d4_model(d4, mol)
  • new (example can be found here)
    1. import: use mctc_env, only : error_type
    2. define: type(error_type), allocatable :: error
    3. call: call new_d4_model(error, d4, mol)
• All variables related to the Gaussian weights (gwvec(:, :, :), gwdcn(:, :, :), gwdq(:, :, :)) are now 3D arrays because the D4S model introduced pair-wise weighting parameters. Users that only interact with the get_dispersion subroutine are not affected.

New Features

• D4S model
• EEQ-BC partial charges
• Always use dynamic OMP scheduling for ATM term (improved parallel performance for large systems)

Detailed Changelog

• Support for latest Python by @awvwgk in #260
• Implementation of smooth D4S by @thfroitzheim in #264
• Update the CI by @thfroitzheim in #265
• Avoid redundant CI runs in PRs on push by @marvinfriede in #266
• Update and fix the dependencies in fpm by @thfroitzheim in #267
• Fix segfault in parallel CN gradient by @marvinfriede in #262
• Access parameters on Windows by @marvinfriede in #268
• Workarounds for Cray compilers by @marvinfriede in #269
• Try to use C compiler for finding libraries by @awvwgk in #271
• Fix zizmor security issues in GA by @marvinfriede in #272
• Add wB97X-D4 parameters by @marvinfriede in #276
• Fix bug in reference polarizability and add actinoid reference systems by @thfroitzheim in #275
• Allow compilation without OpenMP for custom LAPACK by @marvinfriede in #277
• Update citations by @marvinfriede in #278
• Update VASP linking instructions by @marvinfriede in #279
• Update VASP build instructions by @marvinfriede in #280
• Bump setup-fpm to v7 by @marvinfriede in #282
• Replace OMP reductions by thread-private variables + critical summation by @marvinfriede in #287
• Move coordination number to mctc-lib by @thfroitzheim in #283
• Replace Ubuntu 20.04 with 22.04 by @marvinfriede in #291
• fix accessing parameters in windows by @loriab in #292
• Exclude meson version 1.8.0 by @awvwgk in #293
• NVHPC for CMake build by @marvinfriede in #290
• Fix macOS workflows by @marvinfriede in #300
• Bump mctc-lib to 0.4.2 by @marvinfriede in #301
• Extend Fortran API docs by @marvinfriede in #299
• Print static polarizability by @marvinfriede in #302
• Fix typo by @marvinfriede in #303
• Remove AUR submodules by @awvwgk in #306
• Fix for custom libs on Cray by @marvinfriede in #309
• Fix wheels by @thfroitzheim in #310
• GCC 15 Workaround by @marvinfriede in #314
• Add support for EEQ-BC partial charges in DFT-D4 by @thfroitzheim in #311
• Fix badges by @marvinfriede in #318
• Remove old new_d4_model subroutine by @marvinfriede in #316
• Dynamic load balancing for ATM by @marvinfriede in #319
• Bump version to 4.0.0 by @marvinfriede in #317
• Add qmodel flag to man page by @thfroitzheim in #320

New Contributors

• @thfroitzheim made their first contribution in #264

Full Changelog: v3.7.0...v4.0.0

Changelog

• Command line option for functional listing by @marvinfriede in #219
• Add D4 parameters for RSHs by @marvinfriede in #220
• Add config for RTD by @awvwgk in #226
• Actinides by @marvinfriede in #227
• Update mamba provider by @marvinfriede in #228
• Actinide Update by @marvinfriede in #229
• Update interface for gradient for PySCF 2.4 by @awvwgk in #230
• Fix sorting functional names by @marvinfriede in #232
• r2SCAN double-hybrid and wr2SCAN range-separated hybrid parameters. by @lukaswittmann in #225
• Correction to Am and Cm electronegativities by @lukaswittmann in #233
• Add libxc functional names by @marvinfriede in #234
• Fix Ac parameters by @marvinfriede in #242
• Fix wheel workflow by @marvinfriede in #250
• Fix Python workflow by @marvinfriede in #251
• Fix passing atomic numbers from PySCF to D4 interface by @marvinfriede in #248
• Fix periodic PySCF calculations by @aizvorski in #254
• Treat Intel LLVM compiler as standard Intel compiler (#252) by @hongyi-zhao in #253
• Release version 3.7.0 by @awvwgk in #231

New Contributors

• @lukaswittmann made their first contribution in #225
• @aizvorski made their first contribution in #254
• @hongyi-zhao made their first contribution in #253

Full Changelog: v3.6.0...v3.7.0

Many thanks to Marvin Friede (@marvinfriede) and Eisuke Kawashima (@e-kwsm) for contributing to this release.

Changelog

• Update wheel workflow by @awvwgk in #187
• Provision CI from conda-forge by @awvwgk in #194
• PySCF: Capture dispersion energy in scf_summary by @awvwgk in #193
• Catch unsupported elements by @marvinfriede in #189
• Delay creation of API object to avoid memory leak by @awvwgk in #196
• Minor fixes by @e-kwsm in #198
• Replace merge statements for optional parameters by @marvinfriede in #202
• Make Python installation pure: false by default by @awvwgk in #210
• Move to meson-python for Python bindings by @awvwgk in #213

Full Changelog: v3.5.0...v3.6.0

Thanks to Ty Balduf (@TyBalduf), Sebastian Ehlert (@awvwgk), Eisuke Kawashima (@e-kwsm) and Daniel Mejia-Rodriguez (@dmejiar) for contributing to this release.

Changelog

• Add workflow for building wheels by @awvwgk in #155
• Add support for PySCF by @awvwgk in #156
• replace deprecated make_directory with file(MAKE_DIRECTORY) by @e-kwsm in #160
• Fix finding of MKL for Intel 2021 and newer by @awvwgk in #162
• Use meson-python to generate sdist for Python bindings by @awvwgk in #163
• Add documentation pages by @awvwgk in #164
• build(cmake): enable language C to use MKL as BLAS by @e-kwsm in #172
• Export API routines for MSVC compiler by @TyBalduf in #176
• ILP64 support by @dmejiar in #166
• test(C): do not use assert by @e-kwsm in #175
• test: enable C API test if WITH_API is true by @e-kwsm in #173
• Add the reference data base back by @awvwgk in #115
• Add CodeQL workflow for GitHub code scanning by @lgtm-com in #179
• Add Hessian support by @TyBalduf in #180
• Add dllexport for numerical hessian by @TyBalduf in #181
• Read .CHRG from the same directory as input file by @marvinfriede in #182
• Allow passing through values for dispersion model by @awvwgk in #184
• Add error testing for C-API by @awvwgk in #186

New Contributors

• @TyBalduf made their first contribution in #176
• @dmejiar made their first contribution in #166
• @marvinfriede made their first contribution in #182

Full Changelog: v3.4.0...v3.5.0

Created on 2022/11/28, 19:13 UTC by @TyBalduf with commit 2df5ee3.
This is an automated distribution of the latest dftd4 version. It contains the latest features and possibly also the newest bugs. Use with caution!
https://github.com/dftd4/dftd4/actions/runs/3567867705

Many thanks to Eisuke Kawashima (@e-kwsm) and Rodríguez Martínez (@ricrogz) for contributing to this release. Thanks to Yuri Victorovich (@yurivict) for packaging DFT-D4 for FreeBSD.

Changelog

• Use python3 for config/install-mod.py (#118)
• Add build instructions for CMake and fpm (#125)
• Better check for exceptional floating point values (#128)
• Restructure command line interface (#133, #153)
• Update dependency handling in meson (#136)
• Refactor Python API (#137)
• Do not link against OpenMP if WITH_OpenMP is set to FALSE (#139)
• Add information on FreeBSD port (#142)
• Allow more flexible adjustment of Python version (#145)
• Pin fetched dependency versions in cmake build (#146)
• Allow adjusting of module install subdirectory (#143)
• Update to CODATA 2018 (#151)
• Add type generic macro for deleting API objects (#152)
• Wheels (manylinux) available on PyPI (https://pypi.org/project/dftd4)

New parameters

• PBEsol, AM05, MN12sx (#132)
• HSE03, HSE06, HSE12, HSE12S, HSESOL (#134)
• r²SCAN hybrids (#141)

Full Changelog: v3.3.0...v3.4.0

Thanks to Bálint Aradi (@aradi) and Eisuke Kawashima (@e-kwsm) for contributing to this release.

• use np.ascontiguousarray in Python API to transfer arrays to C (#99)
• allow selection of reference charges when creating dispersion model (#100)
• update documentation (#103, #104, #116, #117)
• support CMake build system (#105, #110, #114)
• allow linking against custom linear algebra backend in meson (#106)
• simplify installation process for Python API with setuptools (#109)
• add compatibility layer for integration with Vasp (#113)

Feature release

• Allow pairwise resolved dispersion calculation
• Allow linking against MKL real time library
• Enable usage parameter data file in Python API
• Install module files in meson to include/dftd4/<compiler-id>-<compiler-version>/
• Convenience function to combine DFTD4 with other ASE calculators
• Pip can now be used to build extension module against existing DFT-D4 installation

Feature release of API improvements.

• Fortran API can now access dispersion properties (CN, charges, C6, α)
• C API can now construct dispersion model with customized parameters
• Python API supports now QCSchema and ASE integration