bug in PES scan function

[dvolkmer]

Hello,

there ssems to be a bug in the scan function of xtb:
when a dihedral angle shall be scanned which is defined by 4 non-adjacent atoms,
the scan will always end up at the initial coordinate.

As an example: the molecule tolane.xyz has a C:::C triple bond separating two phenyl groups
and a dihedral scan should recover the energy of the different rotamers.

My input is:

$constrain
force constant=0.5
$scan
dihedral: 2,3,9,10,0.0;0.0,360.0,72
$end

and for the xyz file:

24
00000001
C -1.211000 4.143000 -0.000000
C -1.213000 2.744000 -0.000000
C 0.000000 2.042000 0.000000
C 1.213000 2.744000 0.000000
C 1.211000 4.143000 0.000000
C 0.000000 4.842000 0.000000
C 0.000000 0.603000 0.000000
C -0.000000 -0.603000 0.000000
C 0.000000 -2.042000 0.000000
C -1.213000 -2.744000 0.000000
C -1.211000 -4.143000 0.000000
C 0.000000 -4.842000 0.000000
C 1.211000 -4.143000 0.000000
C 1.213000 -2.744000 0.000000
H -2.148000 4.685000 -0.000000
H -2.157000 2.212000 -0.000000
H 2.157000 2.212000 0.000000
H 2.148000 4.685000 0.000000
H 0.000000 5.925000 0.000000
H -2.157000 -2.212000 0.000000
H -2.148000 -4.685000 0.000000
H 0.000000 -5.925000 0.000000
H 2.148000 -4.685000 0.000000
H 2.157000 -2.212000 0.000000

Would be great if this issue could be fixed in this or the next release.
Kind regards, Dirk

[chselz]

Hello Dirk
thank you for your interest in xtb and sorry for the late follow up.

So for me it worked with this xcontrol file

$constrain
 force constant=0.5
$scan
 dihedral: 2,3,9,10,180.0; 180.0,540.0,72
$end

Hope that helps ☺️

[dvolkmer]

Dear Fabian,

thank you very much ! I have two remaining questions:
(i) regarding the usage of force constants:
is it possible to define different force constants for multiple constraints (and if so, how
should this be defined)
(ii) is any scan of parameters in xtb a sequential job which starts from the same (=first) reference position
or is it a scan which takes the result from the previous scan position as input for the next step ?

Best wishes, Dirk

[chselz]

Hi Dirk
as the force constant is a global parameter according to the documentation it is not possible to define different force constants
and regarding the second question, as far as i understand it it is a sequential job as in the geometry optimization call the restart logical is set to .false. however i don't think even if it used results from the previous scan position it would only speed up the first iteration of the geometry optimization cycle so the benefit would be not very huge.

[dvolkmer]

Dear Fabian,
for large and flexible molecules it makes a huge difference for torsion energy profiles if you start from the initial structure
or the structure of the preceeding optimisation step. For the latter the molecule can smoothly adapt all other internal degrees of freedom. This is the reason why some optimizers offer both options.