History#
- 2026.11.12.3 – Quasi-restricted orbitals for open-shell species
Recipes can say which reference their correlated components use for open-shell species,
open_shell_reference: QRO(quasi-restricted orbitals) orUHF, and a component can override it. Both halves of a difference must use the same reference.The three tier-A recipes use QRO. Closed-shell species are not affected: their calculations, fingerprints and stored results are unchanged.
Needs orca-step 2026.11.12.2 or later for QRO.
- 2026.11.12.2 – Core-valence and relativistic corrections, and the tier-A recipes
Recipe components can be given their frozen core (named sets of frozen orbitals per element), a scalar-relativistic Hamiltonian with its nucleus model, and a decontracted basis, so recipes can add core-valence and scalar-relativistic corrections. A difference of two components must change one thing only, or the recipe is refused.
Three shipped recipes give accurate reference values after W4: CCSD(T)-F12 (H-Ar), CCSD(T) extrapolated from triple and quadruple zeta (H-Ar, Ga-Br), and the same with a decontracted relativistic correction for species with calcium. Their core-valence and relativistic corrections agree with W4’s to 0.13 and 0.01 kJ/mol on test reactions.
Components of one model chemistry with different settings are kept apart, in the job and the fragment store.
Needs orca-step 2026.11.12.1 or later for these settings.
- 2026.11.12.1 – Recipes that use the parts of an energy
Recipes can use the parts of a component’s energy:
part(c, "name")for the HF energy, the CABS singles, the correlation energies and the triples that the program reports, and parts derived from them (the CCSD-F12 correlation, the scaled triples (T*), …).hf_part(c)is the HF energy andcorr_part(c)the correlation alone, so a recipe can, for example, extrapolate the correlation energy and not the HF energy.A recipe that splits an F12 calculation into HF and correlation must account for its CABS singles, and the HF parts of components with different references cannot be combined unless the recipe allows it; such recipes are refused when they are read.
A stored energy without the parts a recipe needs is computed again; parts are never derived.
- 2026.11.12 – Recipes that combine several calculations run in every step
A composite recipe that combines several calculations now runs wherever a model chemistry is used: the Energy step, the Thermochemistry step’s CBH approach, and the others that run model chemistries through seamm-exec’s Evaluator. Each component runs as one batch of its own program over all the structures, the cheapest first, with that program’s own memory and cores.
A structure that fails one component is not sent to the later ones, and its result names the component that failed.
Every result carries the composite record: the recipe’s fingerprint and every component’s settings fingerprint, in the same form whether the recipe has one component or several, and whether a component was computed or found in a store.
A user’s or a job’s own recipe (in a
recipesdirectory of SEAMM’s data path) is found wherever its model chemistry runs.Needs seamm-exec 2026.11.12 or later for recipes with several components.
- 2026.11.11 – First release: composite methods as model chemistries
Composite methods are described by recipes, short YAML files that say how the energy is made from ordinary calculations: the geometry and frequency level, single points at that geometry, and a formula that combines them, with basis-set extrapolations, scaling and empirical terms.
Every recipe is a model chemistry,
SEAMM:CMP@<recipe>, for the steps that take one. Its components are ordinary calculations with their own settings fingerprints, so the fragment store reuses them.The recipes
r2SCAN-3c-HOandrevDSD-r2SCAN-3c-HOare shipped; they are the Thermochemistry step’squickanddefaultpresets.The Composite step itself does nothing yet and is not offered in the editor.