We are very happy to announce the release of a new update to ONETEP: Academic Release 8.2.
For details on how to obtain a copy, see https://onetep.org/code.
To find out what’s new, see the release notes below:
ONETEP 8.2 Academic Release
===========================
2026.08.21.
New features:
* A new SCF optimiser is available: trust-region limited-memory BFGS optimiser.
Implemented by Brad Ayers.
This provides a more robust method for optimising NGWFs, especially for
difficult systems that may stall near default convergence thresholds when
using the Fletcher–Reeves optimiser. It also improves reproducibility.
Activate with: ngwf_cg_type NGWF_LBFGS.
User documentation: https://docs.onetep.org/ngwf_optimisation.html.
Developer documentation (accessible only to ONETEP developers):
https://github.com/onetep-devel/onetep/pull/879.
* The rSCAN mGGA functional is now available and compatible with EMFT.
Implemented by Joseph Prentice.
Documentation: https://docs.onetep.org/mgga.html.
* Density Overlap Regions Indicator (DORI) has been added to the electron
localisation descriptors. Implemented by Brad Ayers.
Calculated during properties (do_properties T or task PROPERTIES).
Activate with:
eld_calculate T
eld_function DORI
Multiple descriptors can be stacked, e.g. “eld_function ELF DORI LOL”.
Developer documentation (accessible only to ONETEP developers):
https://github.com/onetep-devel/onetep/pull/761.
* Custom ionic masses can now be specified via %block species_mass.
Implemented by Harry Brough.
This allows overriding periodic-table defaults, for example to
study isotope effects. Only species listed in the block are modified,
those unspecified retain their default mass.
More details at: https://docs.onetep.org/variable_masses.html.
* Multiple improvements to the GPU port, expanding scope and optimising
performance. Implemented by Nicholas Hine, Jacek Dziedzic,
and Gianluca Seaford.
– GPU-accelerated NGWF gradient calculation (fast_ngwf_gradient T).
Details: https://docs.onetep.org/performance.html#user-fast-ngwf-gradient.
– GPU-accelerated operations on projectors
(triggers when “fast” options are enabled).
– GPU-accelerated PAW projector preconditioning (ditto).
– GPU-accelerated kinetic-energy integrals (ditto).
– GPU-accelerated kinetic-energy preconditioning
(triggers when fast_ngwf_gradient is enabled).
– GPU-accelerated diagonalisation via ELSI (experimental).
Details: https://docs.onetep.org/gpu.html#dense-linear-algebra.
– Optimised performance of the rod representation for NGWFs on GPUs
(transparent to users).
– Optimised performance of operations on trimmed boxes on GPUs (ditto).
– Reduced memory footprint of fast_density (ditto).
– Updated GPU documentation: https://docs.onetep.org/gpu.html.
Minor improvements:
* Improved consistency of EMFT calculations with conduction.
Implemented by Joseph Prentice.
* Extensions to NEB (NEB_UPDATE_METHOD keyword).
Implemented by Brad Ayers.
* Additional information on integrated charge and spin density in VERBOSE.
Implemented by Manuel dos Santos Dias.
* ONETEP now works correctly with Libxc versions 4, 5, 6, and 7.
Enable via compile-time flags: -DLIBXC4, -DLIBXC5, -DLIBXC6, -DLIBXC7.
Implemented by Jacek Dziedzic.
* Significantly improved TRM geometry optimiser (introduced in ONETEP 8.0).
Multiple bug fixes. Of note are:
* Correct dipole correction calculation.
* Workaround for NGWFs that touch exactly but do not overlap.
* Correct functioning of ngwf_cg_type NGWF_POLAK.
* Correct cold-smearing and Methfessel-Paxton smearing in EDFT.