Can I use space group description and UFF?

GULP
1

I need to clarify, whatever I understand how GULP works:

  1. UFF needs correct bond orders for work
  2. GULP automatical bonds are of order 1
  3. Even if I set a molecule in asymetric unit cell with correct bonds, the whole unit cell is not populated by this molecules

As the result, there is no way to use UFF and automatic bonding. So I must use P1 space group for the system description and define bonds myself for whole unit cell (not only for asymetric part)…

Correct ?

How does this fit with eg. polymers of zeolites, when I need to defiine order of bonds crossing cell border ?

2

There are several ways you can use automatic bond finding to help with the set up of bonding, even if the bond orders aren’t all 1. The easiest way is:

  1. Run a single point calculation with automatic bond finding (i.e. molmec as a keyword, or similar)
  2. Write out a restart file using “dump connectivity” to create a version of your input that has the bonding listed.
  3. You can then edit the bonds by specifying their type (using the example from the help file “connect 1 2 double exocyclic 1 0 -1”)

Alternatively you can use Materials Studio where results have been validated against the original implementation of UFF.
NB: There is also the “bondtype” option which specifies the bond order between particular species, which allows the species types to be used to determine the default bond order between pairs of atoms when found by GULP.

3

Than you for your hints. My target is to make the process automatic so I will probably implement 1-3 in my code. It is similar to entering the structure in P1 from my side with all bonds pre generated by our code.
The “bondtype” keyword can be a nice addition to the individual force field files, but it give sense only for force field using this information (UFF).
Accordin Materials Studio - I can confirm the results of energy minimalization are consistent with original work (if force field atom types are correct). Unfortunately I had found on hundreds of structures the force field atom type identification in Materials Studio is incorrect. I do head to head comparison with OpenBabel type identification engine and it look OpenBabel results gives beter agreement with the atomtype description.

4

What would be helpful is if you post examples of where the MS and OpenBabel force field atom types are assigned differently and the reference data (pointer to the original paper?) that determines which is the correct assignment.

5

I had already posted in another thread names of 2 CSD CIF files generatin the issue + problem description. Some issues are clearly visible -C-O-H oxygen identified by MS as O_R = oxygen in aromatic cycle. You need to have MS for reproduction. Give me some time, I will make a list of more such cases. Generaly the correct assigment from articles is given in the headers of your .lib files and this definitions looks OK and corresponding to OpenBabel results. I work on gaff2-> UFF and gaff2->Dreading translation table and I am checking this against MS results, Open Babel results and the paper definiton of the types. So I see clerly the differences. The gaff2 atom types are over set of both Dreading and UFF so translation is possibe (if you restrict atoms to H,C,N,O,S,F,Cl,Br,I,F,P). I use AnteChamber engine for gaff2 types recognition and this looks reliable …

6

I had send more examples directly to BIOVIA people. I will update the discussion if I will get any response from them. The issue is with S_R,C_R,O_R,N_R types primary assigned when a non-resonant type should be used. It is directly visible (assignment on non resonant non aromatic parts of the molecules and clearly a bug, no Open Babel required) …

7

Thanks for sending through files to BIOVIA with the examples.

8

I had clarified this with BIOVIA. They believe an OH from a group attached to an aromatic cycle is an aromatic O_R oxygen … Their opinion "For hydroxyl groups attached to a resonant structure (e.g. phenol), one of the lone pairs can participate in the bonding, and oxygen has a slight sp2 character. " is a serious valid theory witch need to be eventually tested by results comparison with other methods … The same situation is with he other X_R assignments.

9

Thanks for reporting back. It sounds reasonable that O_R is a special atom type for O that can be partially conjugated to an aromatic system. Ultimately what matters is what was assigned in the original force field fitting and since I believe BIOVIA have the original UFF implementation then chances are it’s correct & so the assignment elsewhere is inconsistent with the original work.