When running a transition state search or geometry optimisation in orca, some of the quantites like forces in cartesian coordinates are only written to the *.opt file that is not mentioned in the manual. Here is a description of the file contents and formats.
The file is set up in sections where the labels of which are prefixed with the $ sign:
$orca_opt_file $trust $epredict $ediffsc $ctyp $coordinates $TSSearch $redundant_internals $energies $gradients $redundant_coords $redundant_gradients $hessian_approx $bmatrix
The contents of the “$coordinates” section is
$coordinates numframes numcoordinates cycle1_atom0_x cycle1_atom0_y cycle1_atom0_z cycle1_atom1_x ... ... cycle1_atomN_z cycle2_atom0_x cycle2_atom0_y ...
In orca, the geometry optimisation cycles start to count from 1, while atoms count from 0. The coordinates are given in Angstrom. Numframes gives the number of geometries in this file, numcoordinates is the number of cartesian coordinate components, i.e. three times the atom count.
The section “$energies” is given as
$energies numframes final_single_point_energy_cycle_0 final_single_point_energy_cycle_1 ...
All energies are given in Hartree.
Gradient information is given in the “$gradient” section as follows
$gradients numframes numcomponents cycle1_atom0_x cycle1_atom0_y cycle1_atom0_z cycle1_atom1_x ... ... cycle1_atomN_z cycle2_atom0_x cycle2_atom0_y ...
This follows the logic of the “$coordinates” section. All gradients are given in Hartree / bohr. Note that this is the actual gradient, so the force is the negative value of this (yes, cp2k, I am looking at you).
The other sections include current TS mode components ($TSSearch), definition of redundant internal coordinates ($redundant_internals), coordinates for all cycles in internal coordinates ($redundant_coords), gradients for all cycles in internal coordinates ($redundant_gradients), the current (!) approximation of the hessian matrix ($hessian_approx) and the Wilson B matrix.