## VMD-L Mailing List

**From:** Mayne, Christopher G (*cmayne2_at_illinois.edu*)

**Date:** Tue Feb 11 2014 - 14:06:30 CST

**Next message:**Boris Steipe: "Re: VMD can't read 1OXY"**Previous message:**Axel Kohlmeyer: "Re: Installation of VMD topotools"**In reply to:**Tim Lo: "Re: triple bond parametrization"**Messages sorted by:**[ date ] [ thread ] [ subject ] [ author ] [ attachment ]

Tim (& others),

I think I've made all of the necessary adjustments to the code to support optimization of linear elements within ffTK. In addition to the behind-the-scenes modifications, there are a few places users may notice differences relative to earlier "workarounds".

1) When writing the Gaussian input file to compute the hessian, ffTK automatically detects linear angles, updates the IC definition to "L" (instead of "A"), and removes dihedrals that span the linear section (as they are undefined).

2) When setting up the bond and angle optimization, the "Guess" routine treats the linear bend as an angle, which can now be included in the parameter optimization.

3) When setting up torsion scans using the "Read from PAR" button applies a filter to remove dihedrals that span linear bends, as they are undefined.

I've committed the changes to our CVS, and they should be incorporated into the next VMD 1.9.2 alpha release.

Regards,

Christopher Mayne

On Feb 7, 2014, at 7:30 PM, Tim Lo wrote:

*> Dear Christopher,
*

*>
*

*> Thank you for your solution. I tested my geometric optimization Gaussian job this week and it does not have any problem now. My observations were the same as yours.
*

*>
*

*> I look forward to the new version of ffTK which is compatible with linear molecules.
*

*>
*

*> Regards,
*

*> Tim
*

*>
*

*> On 2014/2/8 上午 07:33, Mayne, Christopher G wrote:
*

*>> Tim,
*

*>>
*

*>> Sorry it's taken me a while to get back to this. I ran some tests on acetonitrile today. Using standard cartesian definition, the Gaussian geometry optimization proceeded without error. Then after converting the linear angle from "A" to "L" and removing the dihedrals involved in the linear angle, the hessian calculation completed without error. If the linear angle is left out, the bond and angle optimization proceeds without error. When including the C-C-N angle in the optimization, ffTK threw an error. Without having looked into it too far, I believe that this is an error in how ffTK tries to interpret the "L" entry, and I'll have to look further into it.
*

*>>
*

*>> All of that said, it sounds like your particular error is originating from your modification of the Gaussian input file when running the initial geometry optimization, and is propagating to the hessian (ffTK tells Gaussian to pull the geometry used in the hessian calculation from the geometry optimization checkpoint file). so you may have to start over to correct the errors you're seeing. In the meantime, I'll try to fix the error associated with how ffTK reads in the "L" terms.
*

*>>
*

*>> Regards,
*

*>> Christopher Mayne
*

*>>
*

*>>
*

*>>
*

*>> On Feb 1, 2014, at 10:44 PM, Mayne, Christopher G wrote:
*

*>>
*

*>>> Tim,
*

*>>>
*

*>>> Make sure to cc VMD-L on this type of email so that others can learn from our discussion.
*

*>>>
*

*>>> Yes, the internal coordinate system is crucial for bond optimization. The short reasoning is because the IC system is defined to match molecular connectivity, hence the hessian maps directly to the desired distortions for the optimization. I will have to take a closer look and revisit some test cases when I'm back in the office on Monday.
*

*>>>
*

*>>> Christopher Mayne
*

*>>>
*

*>>> On Feb 1, 2014, at 9:51 PM, Tim Lo wrote:
*

*>>>
*

*>>>> Dear Christpoher,
*

*>>>>
*

*>>>> This has partially solved the problem and Gaussian can start the calculation. But at the end, the output file says "Error in internal coordinate system." When I click the "guess" button in the "Opt. Bonded" tab, there is an error message: "Couldn't read internal coordinate table!"
*

*>>>>
*

*>>>> Actually, at the beginning of the parametrization procedure in "Opt. Geometry", I also encountered this internal coordinate system error in the Gaussian output file. I just bypassed this problem by modifying the Opt part in the input file as "Opt=(Z-matrix)" or "Opt=(Cartesian)". I think the use of internal coordinate system is problematic for linear molecules. Must the internal coordinate system be used in "Opt. Bonded"? How to solve it or bypass it?
*

*>>>>
*

*>>>> Regards,
*

*>>>> Tim
*

*>>>>
*

*>>>> On 2014/2/2 上午 10:00, Mayne, Christopher G wrote:
*

*>>>>> I just noticed an error in my previous email. You should change "A" to "L" for the linear angle (NOT "LB").
*

*>>>>>
*

*>>>>> Sorry for any confusion,
*

*>>>>> Christopher Mayne
*

*>>>>>
*

*>>>>>
*

*>>>>> On Feb 1, 2014, at 7:02 PM, Mayne, Christopher G wrote:
*

*>>>>>
*

*>>>>>> Tim,
*

*>>>>>>
*

*>>>>>> I ran across this issue with a ligand I was parameterizing some time ago. Although I think I found a solution, I became distracted with something else more and never full tested the solution. If I recall correctly, you will need to make a few modifications to the hess.gau input file. You will need to find the linear angle and change it from "A" to "LB". This should tell Gaussian to treat it as a "linear bend" instead of an angle. You will also have to remove the dihedral terms that cover that angle, as linear dihedrals are undefined as far as Gaussian is concerned.
*

*>>>>>>
*

*>>>>>> If this solves the problem, please let me know and I'll update the code accordingly. If it doesn't solve the problem, let me know and I'll revisit my files to work it out.
*

*>>>>>>
*

*>>>>>> Regards,
*

*>>>>>> Christopher Mayne
*

*>>>>>>
*

*>>>>>>
*

*>>>>>>
*

*>>>>>> On Jan 31, 2014, at 8:11 PM, Tim Lo wrote:
*

*>>>>>>
*

*>>>>>>> I have a question about the parametrization of molecules involving a C-C triple bond. Take 1,2-diphenylethyne (ph-C-triple-C-ph) as an example. When I use FFTK to parametrize it, the Gaussian job of Hessian calculation (hess.gau) cannot be performed properly. From the error message in the Gaussian output file, I think, at least for the default hess.gau, Gaussian cannot deal with molecules involving a linear moiety. If so, how can I parametrize a molecule involving a C-C triple bond, esp. the angle energy terms?
*

*>>>>>>>
*

*>>>>>>> The CGenFF parameter file has 2-butyne as an example of C-C triple bond. There are indeed some angle energy terms including the sp-C (CG1T1) atom. Actually, how does one get these terms? There must be some ways to parametrize this kind of molecules but I cannot figure it out. Can anyone help? Thank you.
*

*>>>>>>>
*

*>>>>>>> Regards,
*

*>>>>>>> Tim
*

*>>>
*

*>
*

**Next message:**Boris Steipe: "Re: VMD can't read 1OXY"**Previous message:**Axel Kohlmeyer: "Re: Installation of VMD topotools"**In reply to:**Tim Lo: "Re: triple bond parametrization"**Messages sorted by:**[ date ] [ thread ] [ subject ] [ author ] [ attachment ]