nextuppreviouscontentsindex
MOLCAS manual:

Next: 8.19 GRID_IT Up: 8. Programs Previous: 8.17 genano

Subsections



8.18 geo


8.18.1 Description

The geo module handles geometry optimization in constrained internal coordinates[55]. The module is called automatically when the emil command > DO GEO is used and should never be called explicitly by the user. All input relevant for these type of geometry optimization should be supplied in gateway and all the relevant keywords are described in detail in the manual section for gateway.

The purpose of the geo module is to perform geometry optimization of several molecular fragments by keeping the fragments rigid and only optimize their relative position. The xyz-file of each fragment is supplied to the program by a separate coord input in gateway. Internal coordinates for the whole complex is then constructed and stored into the file $project.zmt of the directory $GeoDir which is a communication directory set up and used by the geo module. The internal coordinates are chosen so that only a maximum of six coordinates link each fragment. All coordinates within fragments are frozen and the optimization is only carried out for these linking coordinates. The geometry optimization is performed using a numerical gradient (and hessian if needed) and could in principle be used together with any energy relaxation method that is implemented into molcas. (Currently it only works with scf, mp2, rasscf, caspt2, chcc, cht3, and ccsdt but it is very simply to extend it to other modules.)

The module is intended to use in cases where one knows the geometry of each fragment and do not expect it to change much during the optimization.

Note that it is often advantageous to run a GEO job in parallel. The number of processors (cores) required can be easily evaluated by the following formula[55]:

\begin{displaymath}
N_{procs} = 2 + \frac{N_{var}\cdot\left(N_{var}+3\right)}{2},
\end{displaymath} (8.7)

where Nvar is the total number of active coordinates. In the simplest case of the Nfrag rigid polyatomic fragments, when only relative positions of fragments are going to be optimized, the Nvar parameter reads:
\begin{displaymath}
N_{var} = 6\cdot N_{frag}
\end{displaymath} (8.8)

In particular, for constrained two-fragment geometry optimization one needs 29 processes to run GEO job in a fully parallel manner.

8.18.1.1 Creating the z-matrix

The z-matrix is created by choosing the first atom in the xyz-file and put that in top of the z-matrix file (the example is a methane dimer:

H
The next atom is chosen as the closest atom not already in the z-matrix and in the same fragment as the first. The distance to the closest neighbor within the z-matrix is calculated and written into the zmatrix:

C
H 1.104408 0 1
The number zero means "do not optimize" and is appended to all coordinates within fragments and the 1 at the right hand side keeps track of which atom that were the closest neighbour. The next molecule is again chosen as the one not in the z-matrix but in the fragment. The distance between the third atom and its closest neighbour and the angle between the third atom, the closest atom and the second closest atom with the second closest in the middle is added.

C
H 1.104408 0 1
H 1.104438 0 109.278819 0 1 2
For the fourth atom a dihedral angle is added and after that all remaining atoms only have coordinates related to three neighbors. When all the molecule in the fragment has entered the z-matrix the closest atom in another fragment is taken as the next atom. Neighbors to atoms in that fragment is chosen from the same fragment if possible and then from the previous fragment. When two atoms from the second fragment has been added the z-matrix looks like this:

C
H 1.104408 0 1
H 1.104438 0 109.278819 0 1 2
H 1.104456 0 109.279688 0 -119.533195 0 1 2 3
H 1.104831 0 109.663377 0 120.232913 0 1 2 3
H 3.550214 1 75.534230 1 -133.506752 1 3 4 1
C 1.104404 0 78.859588 1 -146.087839 1 7 3 3
Note that coordinates linking fragments have a 1 after and will be optimized and that the last carbon has one neighbor within its own fragment and two within the previous fragment.

There are two special exceptions to the rules described above. Firstly, if it is possible neighboring atoms are chosen from non-hydrogen atoms. (When linking an atom to neighbors, not when choosing the next line in the z-matrix.) Secondly, dihedral angles are rejected if the angles between the three first or the three last atoms are smaller than 3 degrees since very small angles makes the dihedral ill-defined.


8.18.2 Dependencies

The GEO program requires that gateway have been run with either the keyword geo or hyper to setup internal communication files.


8.18.3 Files

8.18.3.1 Input files

Apart from the standard file Geo will use the following input files.
FileContents
RUNFILEFile for communication of auxiliary information.
Geo will also use internal communication files in the directory $GeoDir described in more detail in the next section.

8.18.3.2 Geo communication files

When the emil command > DO GEO or the keywords zonl or zcon are used a directory $GeoDir is created (by default in the input-directory: $CurrDir/$Project.Geo). This directory is used to store files related to geometry optimization and z-matrix generation.

FileContents
$project.zmt

The file with the z-matrix as described above.

general.info

A file used for storing general info about the geometry optimization, it is human readable with labels. The file is automatically setup by the program.

example:

Number of iterations: 3
Number of atoms: 8
Number of internal coordinates: 6
Internal Coordinates:
2.159252 99.560213 123.714490 99.612396 -179.885031 -123.791319
Displacement parameters:
0.150000 2.500000 2.500000 2.500000 2.500000 2.500000
Coordinate types:
badadd
Most of the lines are self-explanatory, coordinate-type is one character for each internal coordinate to optimize with b=bond, a=angle and d=dihedral, displacement parameters is the coordinates defined by hyper.

disp????.info

A file that contains all displaced coordinates. A new instance of the file is created automatically for each geometry step.

example:
disp0001.xyz 2.16485 99.52171 123.71483 99.57425 -179.88493 -123.79124
disp0002.xyz 2.17548 101.22524 124.91941 100.50733 -179.12307 -123.14735
disp0003.xyz 2.16485 101.22952 124.92244 100.50968 -179.12116 -123.14573
disp0004.xyz 2.17145 98.46360 125.21123 100.73337 -178.93851 -122.99136

$project.disp????.energy

A one line file that simply state the current energy for the displacement. If several energies are calculated this will first contain a scf energy and is then updated with the mp2 value when that calculation is finished. The file is written by a small addition to the add-info files and currently collects energies from scf, mp2, rasscf, caspt2, chcc, cht3, and ccsdt. If it should be used with new energy relaxation methods this must be added manually.

$project.final????.xyz

An xyz-file with coordinates for all fragments after XXXX geometry optimizations. This is just for the benefit of the user and should probably be replaced with the same output as created by slapaf, opt.xyz-file for optimized geometry and molden-file for history. Some internal history would still be needed for building more advanced geometry optimization algorithms and convergence criteria though.

$project.geo.molden

A molden file with information about the geometry optimization. The file could be browsed in molden or LUSCUS. The last energy-value is set to zero since the file cannot be created after the last energy calculation and need to be inserted by the user.


8.18.4 input

The general input structure of a geo-calculation looks like this:

&Gateway
[keywords to modify fragment position (frgm,origin)]
coord=fragment1.xyz
...
coord=fragmentN.xyz
[regular keywords + keywords to modify ]

>> DO GEO
&Seward
[energy relaxation methods with any of their keywords]
>> END DO

Both the keywords used to translate and rotate xyz-files (frgm and origin) and the keyword to modify the behaviour of the optimization (hyper, and oldz) is described in more details in the gateway section of the manual.

Here is an example input to calculate the relative orientation of two methane molecules:

>>  export  GeoDir=$CurrDir/$Project.GEO
&Gateway
Coord=$MOLCAS/Coord/Methane1.xyz
Coord=$MOLCAS/Coord/Methane2.xyz
Group=c1
basis=aug-cc-pVDZ
hyper
0.15  2.5  2.5

>>  Do  Geo

&Seward
CHOLESKY  HIGH

&SCF

&MP2

>>  End  Do
In order to run the job above in a task-farm parallel mode, one just needs to slightly modify the original input:

*activating  an  TaskFarm's  interface  in  GEO/HYPER
>>  export  MOLCAS_TASK_INPUT=NEW
*specifying  number  of  cores/cpus  available  for  parallel  execution
>>  export  MOLCAS_NPROCS=2
>>  export  GeoDir=$CurrDir/$Project.GEO
&Gateway
Coord=$MOLCAS/Coord/Methane1.xyz
Coord=$MOLCAS/Coord/Methane2.xyz
Group=c1
basis=aug-cc-pVDZ
hyper
0.15  2.5  2.5

>>  Do  Geo

&Seward
CHOLESKY  HIGH

&SCF

&MP2

>>  End  Do

*setting  the  TaskMode's  execution  model  to  MPP
*each  energy/displacement  is  computed  in  paralel
*by  using  2  cores
>>>  EXPORT  MOLCAS_TASKMODE=1
*actual  execution  of  TaskFarm  via  EMIL
>>>  UNIX  SERIAL  $MOLCAS/sbin/taskfarm


next up previous contents index
Next: 8.19 GRID_IT Up: 8. Programs Previous: 8.17 genano