QMCPACK v4.1.0 released

[Paul R. C. Kent]

On behalf of all the developers and contributors, I am happy to announce the new v4.1.0 release of QMCPACK. Thanks in particular to the new contributors. The release contains improved support for calculating forces, better GPU support, updated build recipes and documentation for many machines, and a broad range of other useful improvements. As always, feedback and requests are welcome, either here, on GitHub, or by direct email.

The release is available from https://github.com/QMCPACK/qmcpack/releases and via https://qmcpack.org. Over the coming days we will also update QMCPACK's spack package.

-- Paul Kent

[4.1.0] - 2025-04-30

This release is recommended for all users and includes improved support for calculating forces, better GPU support, updated build recipes and documentation for many machines, and a broad range of other useful improvements.

• Fast forces with multideterminant wavefunctions #5398
• Significant improvement in AMD GPU performance with better memory handling #5339. Note also updated Frontier instructions #5399. Use of LIBOMPTARGET_AMDGPU_MAX_ASYNC_COPY_BYTES is currently recommended.
• Ongoing performance optimization of LCAO / Gaussian basis set GPU evaluation e.g. #5357, #5342
• Updated build recipes for numerous machines, including NERSC Perlmutter #5464 , ALCF Aurora #5465, and OLCF Frontier #5456.
• HPCToolkit start/stop API support. #5361
• VMC runs advise how to achieve load balance if the run is unbalanced #5412
• History method removed from RotatedSPOs for simplicity and fewer code paths #5414
• 3-body Jastrow ionic derivatives are now analytic, not based on finite differences #5380
• PySCF LCAO to spline converter now compatible with NumPy v2 #5364
• Updated PR and contributor guidance e.g. #5432. Please read https://qmcpack.readthedocs.io/en/develop/developing.html#github-pull-request-guidance
• The wiki and other external documentation is being coallesced in the manual. e.g. Our release process is now in the manual #5443.
• Ongoing refactoring to eventually enable unified real/complex builds and to reduce code duplication.
• Significantly improved reliability of statistical tests. Please report test failures.
• Numerous minor bug fixes.

Known issues:

• Backflow wavefunctions are not working for ab initio (vs model) Hamiltonians in the batched drivers #5459
• The energy density estimator is incorrect #5468
• AFQMC utilities and converters are not compatible with modern PySCF #4653

NEXUS

• Nexus: Documentation for command line options such as --status_only and --progress_tty #5393, see https://nexus-workflows.readthedocs.io/en/latest/user-scripts.html#nexus-command-line-options
• Nexus: Support for stochastic reconfiguration and orbital rotation / orbital optimization #5386
• Nexus: Grand-canonical twist averaging (GCTA) documentation and examples #5316
• Nexus: Support for magnetization density estimator #5346
• Nexus: Support for several new CPU clusters, e.g. BES-MS machine #5416, ORNL Baseline #5371
• Nexus: Support orbital images for plotting/debugging of orbitals #5448
• Nexus: Support QE HDF5 charge density results #5427
• Nexus: Job options are now supported, e.g. to set custom mpirun options for thread binding #5255, #5413
• Nexus: Update force keyword handling #5405
• Nexus: Compatibility with spglib >=v2.5.0 #5429