User Guide#
The XXXX plug-in …
Settings that depend on each other#
Which settings apply depends on others: UHF and MOZYME only for the Hartree-Fock
calculation, the CI settings only for CI calculations, the COSMO settings only with
COSMO, and the cell’s pressure and shear only when the cell is optimized. The step’s
dialogs show only the settings that apply with the current choices, and the same rules
are used when a flowchart is built or edited without the editor (seamm-flowchart or
SEAMM’s MCP server): a setting that would have no effect is refused, with the reason,
and a value that contradicts another is refused too. See “Flowcharts without the editor”
in SEAMM’s user guide.
The Structure to start from is shown in the dialog: by default the first sub-step
starts from the incoming structure and the others from the previous sub-step’s
(current); initial always uses the incoming one. Force Constants works with
MOZYME but does no follow-up calculation, so it does not offer that setting.
Rerunning a job#
MOPAC runs through SEAMM’s task layer, in the step’s directory so that its output can
be watched while it runs. The task layer keeps a record of each MOPAC calculation in
tasks/manifest.json in the step’s directory, with a tasks/mopac/DONE file once
it has finished.
When a job is run again in the same directory – after it was stopped, lost its node, or ran out of time:
a MOPAC step that had completed is skipped, as before: its
success.datfile says so, and its results are read from the files it left, whatever its input now;a MOPAC calculation that had finished but whose step had not completed (the job stopped just after MOPAC ended), and the Lewis structure calculation, are not repeated if their input is unchanged; their results are read from the files they left;
a calculation that did not finish – it failed, or the job was stopped while it was running – is run again, up to three attempts in all; after that the step stops with a message saying so. Deleting its entry in
tasks/manifest.json, or changing its input, lets it run again. If the job was killed outright, a MOPAC process from the earlier run may still be running; the rerun stops it first.
To recompute a completed MOPAC step, delete its success.dat.
MOPAC as a model chemistry#
Steps that set up a model chemistry and evaluate it at many structures – the
Energy step, the Dimer Builder’s contact search, Normal Mode Sampling,
LAMMPS QM-MD – use MOPAC through a Model Chemistry step (e.g. MOPAC:PM7), with
no settings in the MOPAC step itself. On this machine they keep one MOPAC running as an
MDI engine and pass it the structures
one after another, which suits MOPAC’s very short calculations. When the job’s target
sends its calculations to a cluster queue, each molecule instead runs there as a
separate MOPAC calculation, and a rerun reuses the ones that had finished.
The energy is the heat of formation either way, and the gradients agree. Periodic and open-shell structures always use the MDI engine; on a queue target where MOPAC is not installed on the job’s own machine, such a structure is reported as failed, with the reason, and the others finish.
For molecules the engine has the MOPAC program do each calculation (a single
point with gradients, MOPAC’s default SCF, started afresh at every structure), so
its energies are exactly those of a MOPAC calculation of the same structure. MOPAC’s
library, which the engine used before, runs every SCF with PULAY, and for some
molecules that converges to a solution far above the right one: PM6-ORG
penta-1,2,3-triene 449.9 instead of 421.5 kJ/mol, and the PM7 S2 triplet
up to 600 kJ/mol too high at some bond lengths. An optimization that meets such a
solution follows the wrong energy surface. For small molecules the program is as
fast as the library (about 35 ms a calculation for 11 atoms).
Periodic structures and MOZYME still use the library, which keeps the previous
structure’s density as the starting guess; the program cannot do them this way. In
the library, open-shell calculations each run in a separate process, because
MOPAC 23.2.5’s library aborts at random in some of them, and one whose process dies
is done by the program. The engine’s --scf option (auto, program or
library) overrides the choice.
Steps that take finite differences of the gradients, such as Vibrational
Analysis, ask the engine for tight convergence: each calculation then converges to
a relative tolerance of 0.001 instead of MOPAC’s default 1.0. With a warm start (the
library), the default tolerance leaves small errors that depend on which structure
came before, which finite differences magnify. For water’s anharmonic force
constants the noise was 447 cm-1 at 1.0, 22 at 0.01 (the tolerance
PRECISE gives), and 3.5 at 0.001.
Each calculation through the Model Chemistry carries a fingerprint of its settings: MOPAC’s version, the method, whether MOZYME is used, and the SCF tolerance. The charge and spin multiplicity belong to the structure and are not part of it. The MDI engine and the separate calculations on a queue give the same fingerprint for the same calculation, and steps that reuse results, such as Thermochemistry with its fragment store, use it to tell whether a stored MOPAC result came from the same calculation they need.
For a periodic structure the engine also returns the stress, as a pressure: positive when the system pushes outward (the opposite of MOPAC’s own tensile-positive stress). MOPAC’s model chemistries declare this, so steps that use the stress, such as the MBE step’s periodic levels, convert it correctly.