Skip to content

Proposal: AMF and mixed double counting for Full DFT+U #8281

Open

Description

@rimujiangshu

Background

I would like to extend the existing Full DFT+U implementation (dft_plus_u=1) with AMF and mixed double-counting options. This would allow comparisons between double-counting treatments within the same projector implementation.

Describe the solution you'd like

Add dftu_dc=fll|amf|mixed, with fll as the default. For mixed, require an explicit, finite dftu_alpha in [0,1]: 0 corresponds to AMF and 1 to FLL. Intermediate values would interpolate the local correction energy and its corresponding potential, rather than separate SCF results.

The initial scope would cover CPU/MPI LCAO, collinear fixed-structure SCF, and both Gamma-only and multi-k calculations. Dual and a new On-site projector would remain outside this proposal.

Task list only for developers

  • Notice possible changes of behavior
  • Explain the changes of codes in core modules of ESolver, HSolver, ElecState, Hamilt, Operator or Psi

Notice Possible Changes of Behavior (Reminder only for developers)

Omitting the new parameters or explicitly selecting fll would retain the existing FLL implementation and supported workflows. Restrictions on unsupported combinations would apply only to AMF/mixed requests, including mixed endpoints. No additional diagnostic calculations or files would be enabled by default.

Notice any changes of core modules (Reminder only for developers)

No response

Notice Possible Changes of Core Modules (Reminder only for developers)

No response

Additional Context

Research-branch CrI3 calculations have been checked for SCF convergence and expected FM/AFM magnetic order. However, an isolated implementation on current develop, upstream FLL regressions, and endpoint-equivalence tests remain outstanding. Research results do not establish upstream compatibility.

Would this scope and parameter interface be appropriate for a focused follow-up PR?

Task list for Issue attackers (only for developers)

  • Review and understand the proposed feature and its importance.
  • Research on the existing solutions and relevant research articles/resources.
  • Discuss with the team to evaluate the feasibility of implementing the feature.
  • Create a design document outlining the proposed solution and implementation details.
  • Get feedback from the team on the design document.
  • Develop the feature following the agreed design.
  • Write unit tests and integration tests for the feature.
  • Update the documentation to include the new feature.
  • Perform code review and address any issues.
  • Merge the feature into the main branch.
  • Monitor for any issues or bugs reported by users after the feature is released.
  • Address any issues or bugs reported by users and continuously improve the feature.

Activity

  1. ABACUS-PR-BOT commented on Nov 3, 2026

    @ABACUS-PR-BOT
    Collaborator

    Yes—the proposed scope is a reasonable basis for a focused PR, provided the AMF definition and existing implementation are identified explicitly.

    At f5315651166e, dft_plus_u=1 selects radius-adjustable localized projections, and hubbard_u is the simplified U_eff=U-J. This route already uses DFTU_onsite, so excluding a new On-site projector should mean retaining that existing construction. “Full” should not imply an implemented anisotropic U/J functional; the Liechtenstein/AMF enum entries are placeholders.

    For fixed input alpha, interpolate the local correction functional and its derivative at the same projected occupation. Specify the AMF spin-average and occupation-normalization conventions. Preserve the band-energy bookkeeping too: the active onsite kernel accumulates E_corr - Tr(V n), rather than the bare local functional.

    The remaining interface decisions are what to do with alpha outside mixed, and how AMF/mixed handles force/stress requests during SCF, restart/occupation initialization, Yukawa U and U-ramping. State supported combinations or reject them explicitly, while preserving existing FLL behavior and applying new-mode guards even at mixed endpoints.

    For acceptance, include independent energy–potential derivative checks at an interior alpha, with trace-changing and off-diagonal occupation perturbations, alongside endpoint and spin-factor tests, and same-input FLL/endpoint SCF comparisons across Gamma/multi-k and representative MPI layouts. Update INPUT metadata/help. These checks would establish more than convergence or expected magnetic order alone.

    This is source-based design feedback; no build or runtime check was performed.

  2. added
    DFT+UIssues related to DFT plus U function
    Features NeededThe features are indeed needed, and developers should have sophisticated knowledge
    on Nov 4, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    DFT+UIssues related to DFT plus U functionFeatures NeededThe features are indeed needed, and developers should have sophisticated knowledge

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions