[CP2K-user] [CP2K:22363] [CP2K 8.1 CDFT] Exact nonperturbative Becke/Hirshfeld weight export without entering outer SCF

[Thomas Kühne]

Dear Dayu,

> Am 30.07.2026 um 04:37 schrieb 'Dayu Zhu' via cp2k <cp2k at googlegroups.com>:
> 
> Dear CP2K developers and users,
> 
> I am following up with a much shorter version of my question.
> 
> For a fixed geometry, grid, and ATOM_GROUP definition, I need to export the exact CP2K Becke or Hirshfeld CDFT weight function without perturbing an already-converged electronic density.
> 
> Could you please clarify:
> 
> 1. Is there a stock CP2K input option, in version 8.1 or a newer release, that prints the weight-function grid without running the CDFT outer SCF?
No. CP2K 8.1, as well as the current version, provides CDFT%PROGRAM_RUN_INFO%WEIGHT_FUNCTION, but this print key is only reached from an active CDFT calculation. There is currently no stock builder-only or print-only input mode that avoids the electronic/CDFT SCF path

> 2. If not, would it be methodologically valid to add a minimal diagnostic hook that calls the existing CP2K weight builder and cube writer, and then exits before density integration, Kohn–Sham construction, constraint-potential application, and CDFT outer-SCF optimization?
Yes. For fixed geometry, cell, grid, atomic kinds and ATOM_GROUP definitions, the Becke and Hirshfeld weight functions are independent of the current molecular electron density. For SHAPE_FUNCTION DENSITY, the Hirshfeld builder uses isolated-atom reference densities rather than the converged molecular density

> A brief yes/no answer or a pointer to the relevant routine, example, or newer implementation would be very helpful.
> 
> The longer message below contains the source-level details, but no detailed review of those attachments is necessary to answer these two questions.
A clean diagnostic implementation could initialize the normal QS/PW grid environment, allocate the group weight grids as done in qs_ks_cdft_constraint, call the existing constraint-specific initialization and low-level builder, write the grids with cdft_print_weight_function, and then return through normal cleanup before scf_env_do_scf. In CP2K 8.1 the low-level builders and cube writer are private, so a small public wrapper or refactoring would be required.

For an unsubsampled grid, STRIDE 1 1 1 should be used. Please also note that the standard cube format is text-formatted and therefore does not preserve the full internal double precision.

Best regards, 

Thomas Kühne

> 
> Thank you very much for your time.
> 
> Best regards,
> Dayu Zhu
> 
> On Fri, 24 Jul 2026 at 13:12, Dayu Zhu <dayu.zhu at monash.edu <mailto:dayu.zhu at monash.edu>> wrote:
>> # CP2K 8.1 support request: exact nonperturbative Becke/Hirshfeld CDFT weight-grid export
>> 
>> ## Build identity
>> 
>> - CP2K version: `8.1`
>> - Executable revision: `git:0b61f2f`
>> - Container release-source HEAD: `0bfbd41a7402152da7b86a103d5d95a4184367d9`
>> - Release tag: `v8.1.0`
>> - Wrapper SHA256: `c786e567853dbe8fafb35a91ec25ba1ca8f5ea5c02d30f5cdf056719ce9bc2d1`
>> - Input XML SHA256: `728c139a1d87554490e67f0ab30f1f5c78f944dad5e87671894e1aa833f281f4`
>> - Container executable: `/cp2k/exe/local/cp2k.psmp`
>> - Container source root: `/cp2k/src`
>> 
>> The wrapper/account path is intentionally omitted. The hashes above and the
>> attached sanitized identity CSV are sufficient to identify the inspected
>> build without exposing an HPC account path.
>> 
>> ## Narrow technical need
>> 
>> We need exact CP2K 8.1 Becke and Hirshfeld weight grids for offline integration
>> of local magnetization from already-converged, unconstrained spin-density
>> cubes. We want to preserve the original density and avoid running CDFT or an
>> electronic SCF merely to obtain the geometric/partition weight grid.
>> 
>> This request concerns method semantics and a minimal diagnostic interface
>> only. It does not claim that stock CP2K is wrong, and it does not request
>> scientific interpretation of any material.
>> 
>> ## Distinctions used in this request
>> 
>> - **Exact weight construction:** running the original CP2K 8.1 Becke or
>>   Hirshfeld builder with fixed geometry, cell/PBC, grid, atom group, COEFF,
>>   atom-kind, and partition parameters.
>> - **Exact weight printing:** writing that already-constructed CP2K PW grid
>>   through the original `cdft_print_weight_function`/cube-writer path.
>> - **Nonperturbative measurement:** obtaining the weight without changing or
>>   reoptimizing the electronic density.
>> - **CDFT outer-SCF perturbation:** adding `weight * STRENGTH` to the
>>   Kohn-Sham potential and updating the electronic state/multiplier.
>> 
>> ## Exact source references inspected
>> 
>> | Routine | CP2K 8.1 source lines |
>> |---|---|
>> | `read_constraint_definitions` | `qs_cdft_utils.F:534-717` |
>> | `becke_constraint_init` | `qs_cdft_utils.F:100-449` |
>> | `becke_constraint_low` | `qs_cdft_methods.F:142-581` |
>> | `hirshfeld_constraint_init` | `qs_cdft_utils.F:966-1086` |
>> | `create_shape_function` | `hirshfeld_methods.F:141-224` |
>> | `hirshfeld_constraint_low` | `qs_cdft_methods.F:1153-1365` |
>> | `qs_ks_cdft_constraint` | `qs_ks_apply_restraints.F:62-128` |
>> | `cdft_constraint_integrate` | `qs_cdft_methods.F:772-954` |
>> | `cdft_constraint_print` | `qs_cdft_utils.F:1095-1249` |
>> | `cdft_print_weight_function` | `qs_cdft_utils.F:1257-1301` |
>> | `qs_ks_build_kohn_sham_matrix` | `qs_ks_methods.F:207-899` |
>> | `sum_up_and_integrate` | `qs_ks_utils.F:1195-1625` |
>> | `cdft_scf` | `qs_scf.F:934-1144` |
>> | `outer_loop_gradient` | `qs_outer_scf.F:102-174` |
>> | `outer_loop_update_qs_env` | `qs_outer_scf.F:386-430` |
>> | `cdft_control_release` | `qs_cdft_types.F:419-493` |
>> 
>> `cdft_print_hirshfeld_density` is not present in the exact `/cp2k/src` tree;
>> both weight types use the common `cdft_print_weight_function` route.
>> 
>> Our revision-pinned source audit indicates that the low-level Becke/Hirshfeld
>> builders construct weights before `cdft_constraint_integrate` reads the
>> current density. The stock weight print is reached only after `cdft_scf`
>> outer-loop initialization, although the print call precedes the first
>> constraint-potential addition in that Kohn-Sham build. We request confirmation
>> of the supported semantics and interface rather than asserting undocumented
>> API guarantees.
>> 
>> ## Questions
>> 
>> 1. For CP2K 8.1, is there a supported stock-input route to initialize and
>>    print exact Becke/Hirshfeld CDFT weight grids without entering the CDFT
>>    outer-SCF loop?
>> 
>> 2. For fixed geometry, real-space grid, atom group, COEFF, atom-kind, and
>>    partition parameters, are the Becke/Hirshfeld weight grids completely
>>    independent of the current electron density, current spin density, TARGET,
>>    STRENGTH, and current Lagrange multiplier?
>> 
>> 3. Does `PROGRAM_RUN_INFO/WEIGHT_FUNCTION` print exactly the same weight grid
>>    regardless of the constrained electronic solution, even though the print
>>    call occurs after outer-loop initialization?
>> 
>> 4. Is there an officially recommended internal routine or minimal driver for
>>    invoking the original CP2K 8.1 Becke/Hirshfeld weight builders and cube
>>    writer without running CDFT or electronic SCF?
>> 
>> 5. Would a minimal source hook that constructs the weight, writes it, and
>>    exits before `cdft_constraint_integrate`, constraint-potential building,
>>    electronic SCF, or CDFT outer-SCF optimization be considered semantically
>>    exact?
>> 
>> 6. For Hirshfeld CDFT in CP2K 8.1, what implementation limitations affect
>>    ENERGY-only local-magnetization integration, weight printing, forces,
>>    geometry optimization, and MD?
>> 
>> 7. Can multiple `ATOM_GROUP` weight grids, for example R0/R1/R2 and a separate
>>    HSO4 monitoring region, be generated in one invocation without activating
>>    coupled constraints?
>> 
>> 8. What metadata must be preserved to guarantee exact offline integration of
>>    `integral w(r)[rho_alpha(r)-rho_beta(r)] dr` against existing spin-density
>>    cubes, including grid origin, axes, stride, cell/PBC, atom order/kinds,
>>    partition settings, group/COEFF mapping, and numeric print precision?
>> 
>> ## Candidate minimal interface for review
>> 
>> After normal input, subsystem, `qs_env`, MGRID, and PW-pool initialization:
>> 
>> 1. allocate one weight PW grid per group using the stock lifetime pattern;
>> 2. call the original Becke or Hirshfeld `init` and low-level builder;
>> 3. call the original common weight cube writer;
>> 4. hash/write provenance metadata;
>> 5. release all temporary grids; and
>> 6. exit before density integration, Kohn-Sham potential construction,
>>    electronic SCF, or outer-SCF optimization.
>> 
>> No patch, build, input, or calculation is included with this request.
>> 
>> ## Lightweight attachments
>> 
>> - `CP2K81_CDFT_weight_evaluator_support_request.md`
>> - `CP2K81_CDFT_weight_evaluator_minimal_context.txt`
>> - `CP2K81_exact_source_identity_for_support.csv`
>> - `CP2K81_CDFT_weight_call_graph_compact.csv`
>> - `CP2K81_weight_function_dependency_matrix.csv`
>> - `CP2K81_weight_print_timing_audit.csv`
>> - `CP2K81_exact_weight_extraction_options.csv`
>> - `CP2K81_exact_weight_source_excerpts_compact.txt`
>> 
>> No coordinates, WFN, trajectory, cube, binary, full source tree, proprietary
>> data, or unrelated project details are attached.
>> 
> 
> 
> -- 
> You received this message because you are subscribed to the Google Groups "cp2k" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe at googlegroups.com <mailto:cp2k+unsubscribe at googlegroups.com>.
> To view this discussion visit https://groups.google.com/d/msgid/cp2k/CAJ1za7wZVwZXNb_h4Ty2z49-AhxGu-xBxv%3Dzfp32oH%2BWz0N-eQ%40mail.gmail.com <https://groups.google.com/d/msgid/cp2k/CAJ1za7wZVwZXNb_h4Ty2z49-AhxGu-xBxv%3Dzfp32oH%2BWz0N-eQ%40mail.gmail.com?utm_medium=email&utm_source=footer>.

-- 
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+unsubscribe at googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/EC955463-E3F8-41ED-B043-69A6298D2F85%40gmail.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20260802/d412d07a/attachment-0003.htm>