Notes on the calculation of an optimized geometry
starting with a Materials Studio file

Introduction

Materials Studio generates two sets of data that are used in constructing a single MOPAC data set.  These sets are (1) the geometry within the unit cell, and (2) the translation vectors.  The following instructions describe how to generate a MOPAC data-set from the data in a Materials Studio file.

All the files mentioned here can be downloaded from ZIP.

Conversion of geometry in the .car file to MOPAC format

Starting with the un-modified file "Data set for resin from Material Studio.car", use Excel to convert the geometry into a MOPAC format.  The result should look like "resin (intermediate).txt"

Conversion of  resin (intermediate).txt" to a simple MOPAC data set

Edit " resin (intermediate).txt" to insert the three lines:

Line 1: 0SCF HTML ZERO
Line 2: Resin. Check to verify that the geometry looks alright
Line 3:

These lines have the following meanings:

Line 1: Keyword 0SCF to suppress any real calculations - the geometry is unmodified, and no SCF is done.  A 0SCF job runs very fast.

Keyword HTML generates a HTML file that can display the geometry using JSmol.  If JSmol has not been installed, go to HTML for instructions.

Keyword ZERO is unique to MAKPOL.  This tells MAKPOL to use the Tv supplied, and to not modify them even if atom 1 is not at the origin.  Use ZERO for Materials Studio files, and don't use ZERO for cif files.

Line 2: A description of the job for the benefit of the user only.
Line 3: A second description line, here it is blank.

 

Edit the end of the file to add the three lines:

Tv     22.76620000      0.00000000      0.00000000 
Tv      0.00000000     22.76620000      0.00000000  
Tv      0.00000000      0.00000000     22.76620000 

 

These are the translation vectors from " Data set for resin from Material Studio.car "

Save the edited file as "MS resin.mop"

First check that the geometry in MOPAC format is correct

Run " MS resin.mop" using MOPAC.  Open the resulting HTML file with Netscape, and verify that it looks reasonable.  At this point there are unattached hydrogen atoms; that's okay - they are present because when the periodic boundary conditions are applied, they will join up with other fragments to form stable, neutral molecules.

Conversion of MOPAC data-set to construct discrete molecules

This conversion is useful for GUI work.  It is not essential for the MOPAC run. It uses program MAKPOL

Edit "MS resin.mop" to add keyword ZERO to the first line, and save the file as "make_resin.dat".

ZERO is a MAKPOL  specific keyword.  This will be deleted before the MOPAC data-set is generated. Keyword ZERO indicates that the supplied translation vectors are correct, and do not have to be modified to account for the non-zero coordinates of atom 1, an operation that has to be run when cif files are used.

Run "make_resin.dat" using MAKPOL, for example, at the command prompt run:

makpol resin

Note that the prefix "make_" is not typed.  This run will generate a MOPAC data-set "resin.mop".  That data set will be composed of discrete molecules, i.e., no unattached hydrogen atoms.  

Second check that the geometry in MOPAC format is correct

Run "resin.mop" using MOPAC.  The resulting geometry should look correct in all details, i.e., there will be no lose hydrogen atoms and no radicals.

Running a 1SCF calculation

Edit "resin.mop" to change 0SCF to 1SCF, and save the edited file as "resin 1SCF.mop"  Run this file using MOPAC.  It should run in less than 10 minutes, and give a Heat of Formation, DHf, of -4698.6 kcal.mol-1, or -81.0 kcal.mol-1, per formula unit.

Running a constrained geometry optimization calculation

Edit "resin.mop" to delete 0SCF and add GNORM=10, and save the edited file as "resin Constrained Geometry Optimization.mop"  Then edit the first three lines to change all the "+0" to "+1".  When this job was run using MOPAC, it took about 20 CPU hours, with 4 cores, this took 9 wall-clock hours.

Running an unconstrained geometry optimization calculation

Edit "resin.mop" to delete 0SCF and add GNORM=10, and save the edited file as "resin Unconstrained Geometry Optimization.mop"  Then either change all "+0" to "+1" or add keyword "OPT".  When this job was run using MOPAC, it took about 35 CPU hours, with 4 cores, this took 17 wall-clock hours.

Quality of results

There are three types of error in these calculations.  The input data-set might have errors; in the resin system described here there do not seem to be any errors, so this is not a source of concern in this particular case.  The theoretical method used is of definitely limited accuracy, so there will some concern, but again in simple organic systems of the type used here this is of minor importance.

The main error is likely to be from finite geometry optimization.  Geometry optimization of solids is much more difficult than for molecules, and as a result the geometry convergence criterion, GNORM=10, is much higher than the default of GNORM=1.0.  A side-effect of using GNORM=10 is that the run might stop before the energy has stopped dropping.

To correct this error is not difficult, but does require a new procedure.

Increasing the quality of results

This is done in three steps:

(1) Delete the keyword GNORM=n.  If no further action was taken, the resulting job would run for a very long time, many days, possibly several weeks.  This will not happen.  After the job starts, it should be monitored by the user, and at the appropriate time it can be stopped under user control.

(2) Start the job running.  Monitor the output to see what's happening to the energy.  Initially, the heat of formation will drop quickly as errors in the input geometry are corrected, then the rate of descent will decrease, and finally the heat of formation will start to oscillate in an unpredictable way.  In some runs, the DHf might rise dramatically, giving the impression that the run has gone bad.  This occurs when a problem with the optimization is detected.  During the attempt to overcome this problem, the DHf can rise, sometimes by many hundreds of kcal.mol-1. 

There are several ways to monitor the run.  The simplest is to open the output file using a simple editor, e.g., notepad, and look at the optimization - it's at the bottom of the output.  A slightly more sophisticated method is to use the UNIX command "grep"  To do this, open a command-prompt window, and use a command of the type:

tail -5000f "resin Constrained optimization.out" | grep CYCLE

(3) When an examination of the optimization shows that the DHf has stopped dropping, issue the MOPAC shut command.  This will shut down the MOPAC run in an orderly manner.  When the run stops, sometime in the next few minutes, edit the output file.  There will be a data-set at the bottom of the output file that holds the best - the lowest energy - geometry.  Delete everything else, then edit this new data-set to include the keyword 1SCF.  Save the resulting file, giving it a new name, e.g.,

"resin Constrained optimization 1SCF.mop"

then run this new file.  The results will represent the best geometry that can be obtained using MOPAC.