xtb_step Phase D – File Drop Notes#

This drop addresses the six items from your last review:

Top-level xtb.py no longer sets self._metadata – you already have that fix in place.
energy_parameters.py now has "extra keywords" declared – you already have that fix; this drop additionally drops charge and multiplicity per items 3 below.
Net charge and spin multiplicity moved off the parameters and onto the configuration – now read via configuration.charge and configuration.spin_multiplicity in substep.base_xtb_args.
The Energy dialog now hides the solvent widget when solvation model == "none" and indents it under the solvation row when shown.
setup.py got a console_scripts entry for the installer – you already added that.
Results are now reported as a tabulated summary (Gaussian-style), with the table accumulated through the inheritance chain.

Files in this drop#

Replace existing#

xtb_step/energy_parameters.py

Drops charge and multiplicity from the parameter set. Keeps extra keywords, method, accuracy, solvation model, solvent, results. The class docstring explains the convention.

xtb_step/substep.py

base_xtb_args(P, configuration) now reads charge and multiplicity from the configuration:

charge = int(configuration.charge)
mult = int(configuration.spin_multiplicity)

Same as MOPAC’s mopac.py:run(). No other changes to this file.

xtb_step/metadata.py

Adds "format" strings to the scalar float results so the table can format them sensibly:

total / electronic / electronic energy and ZPE / G(T): .6f
HOMO / LUMO / gap: .4f
dipole moment: .4f
temperature: .2f

No structural changes – the same 17 results, same property names, same dimensionalities.

xtb_step/energy.py

Two visible changes:

description_text no longer mentions charge or multiplicity. Otherwise unchanged.
analyze now builds a tabulate-formatted summary with Property | Value | Units columns, takes an optional table= argument so subclasses can prepend rows, prints the result with a centered title, and gracefully handles missing results.
run reads extra keywords from P and appends them after the substep-specific args (so users can override / add to xTB’s CLI flags).

xtb_step/frequencies.py

The analyze method replaces the old _post_run_thermo_and_freqs. It re-parses the JSON / vibspectrum / thermo block, calls store_results for the new fields, and builds a single ordered table:

Frequency-count summary rows (“Number of frequencies”, “Imaginary frequencies”).
Energy / orbital / dipole rows (same set as Energy.analyze).
Thermochemistry rows (ZPE, H, T*S, G(T), total free energy, temperature).

The table is printed with a “Frequencies / Thermochemistry” title.

xtb_step/tk_energy.py

Dynamic dialog. The solvation-model widget is bound to a reset_energy_frame method (modeled on mopac_step.tk_energy). When the model changes, the frame is re-laid out: solvent appears in column 1 (indented) only when the model is not "none". The labels in each column are aligned with sw.align_labels separately so they read cleanly.

Not in this drop#

xtb_step/optimization.py: Unchanged from Phase B. It does not need to override analyze: when Energy.run() calls self.analyze(data, P=P) on an Optimization instance, Python’s MRO picks up the inherited Energy.analyze automatically. If you ever want optimization-specific rows (“Convergence: …”, “Max gradient: …”), the simplest pattern is to override analyze to add rows then call super().analyze(data=data, table=table, P=P).
xtb_step/xtb.py, xtb_step/__init__.py, xtb_step/setup.py: Use the versions you uploaded.
xtb_step/xtb_parameters.py: From the Phase C drop. With the working no-self._metadata-at-top-level approach you found, this file is no longer referenced by xtb.py and could be removed. If you remove it, also remove the from .xtb_parameters import xTBParameters line from __init__.py. (Leaving it in is harmless – the class just sits there unused.)

Open issues / risk register for this drop#

Configuration attribute names. I’m assuming configuration.charge and configuration.spin_multiplicity are the right attribute names on a molsystem Configuration, based on the MOPAC code I quoted. If a different attribute name is used in the version of molsystem you have, you’ll see an AttributeError and we’ll need to adjust.
Dialog widget API. self["solvation model"].combobox.bind(...) is the pattern from mopac_step.tk_energy. If the seamm-widgets version installed in your env exposes the underlying widget differently (some plug-ins use self[key].bind(...) directly, without .combobox), the bind will fail at dialog-creation time. If that happens, change smodel.combobox.bind to smodel.bind.
store_results called twice in Frequencies. Energy.run() calls store_results once with the JSON-derived data, then Frequencies.analyze calls it again with the freq + thermo additions. SEAMM’s store_results is idempotent for repeated calls with the same property name, but if you see duplicate property entries in the database, this is where to look.
Frequencies.analyze duplicates the Energy.analyze table-build loop. This is intentional but not pretty: I needed to interleave the freq-count rows BEFORE the energy rows but the thermo rows AFTER, which the simple “subclass adds then super() prints” pattern in Gaussian doesn’t quite handle. If we ever want a cleaner factoring, the right move is to add a helper method on Energy like _append_energy_rows(table, data) that Frequencies.analyze can call between its own pre- and post- sections.

What to test#

make lint && make install && make test – still passes.
Open the Energy substep dialog. The solvent widget should be hidden when “Implicit solvation” is “none”. Pick “ALPB” – solvent should appear, indented. Pick “none” again – solvent disappears.

Run a simple Energy job (e.g. methane). The job.out should contain a tabulated summary like:

                  xTB (GFN2-xTB) Results
╭─────────────────────────────────────┬───────────┬───────╮
│             Property                │   Value   │ Units │
├─────────────────────────────────────┼───────────┼───────┤
│           The total energy          │ -4.174678 │  E_h  │
│       The HOMO orbital energy       │ -13.4827  │   eV  │
│       The LUMO orbital energy       │   2.4567  │   eV  │
│         The HOMO-LUMO gap           │  15.9394  │   eV  │
│  The molecular dipole moment ...    │   0.0000  │ debye │
╰─────────────────────────────────────┴───────────┴───────╯

(Exact wording and number of rows depends on which keys xTB actually emitted in xtbout.json.)

Run an Optimization job. Same table; geometry updated.
Run a Frequencies job on water. Table should now include “Number of frequencies”, “Imaginary frequencies”, and the thermochemistry rows beneath the energy rows.
Run on O2 in its triplet state and on O2- (a stored configuration with charge -1, multiplicity 2). xTB should pick up the charge and multiplicity from the configuration without any per-step parameter editing – demonstrating the SEAMM convention for high-throughput.

Code style#

All files compile cleanly. All lines are <= 88 characters. I have not been able to run black --check here, so a first make format may produce small whitespace adjustments – those should be one pass and then stable.