[gmx-users] non-bonded parameters

[Chris Oostenbrink]

On Thu, 2005-11-17 at 07:40 +0000, nk yx wrote:

>> hi all,
>>     I am confused about the calculation of non-bonded LJ parameters in the 
>> ffG43a1nb.itp. According to the manual, the combination rule of C6 and C12 
>> is:
>> Cij(6)=(Cii(6)*Cjj(6))0.5
>> Cij(12)=(Cii(12)*Cjj(12))0.5
>> However,in the ffG43a1nb.itp file, the values seem not to be in agreement 
>> with the rule. For example,
>> O  C6=0.0022619536 C12=7.4149321e-07
>> OM C6=0.0022619536 C12=7.4149321e-07
>> OA C6=0.0022619536 C12=1.505529e-06
>> 
>> According to the rule, 
>> O OA   C6=0.0022619536 
>>        C12=(7.4149321e-07 * 1.505529e-06)0.5 = 1.0565697e-06 
>> but the value is 1.380375e-06 in the ffG43a1nb.itp.
>> In addition, C12 value between O and OM, and tween O and OA should be the 
>> same, but in the file, C12 is 1.38380375e-06 and 2.258907e-06,respectively.
>> Could anybody tell me why? Thanks in advance.
>  
>
> This is an oddity of the gromos force field. It doesn't work completely
> with combination rules. The numbers are correct though. For more info
> you will need to look up gromos documentation.

It is true that there are some interactions in the GROMOS force field that are not calculated using the geometric combination rules, but rather through the Lorenz-Berthelot combination rules. In the 43a1 parameter set, this is the case for the Chloroform and DMSO atoms. All other interactions are calculated from the geometric combination rules. The confusion arises because for every atomtype there are up to three different values defined for Cii(12)^0.5. There is then a table that specifies for the interaction between two types i and j, which of these parameters to take. In general: I) for regular interactions, II) for interactions between atom types that may form a hydrogen bond and III) for atomtypes that are involved in salt bridges. The physical background is that a stronger repulsion is needed if the atoms are strongly attracted by the electrostatics.

All this is a bit obscured if you only specify the readily calculated interactions, but as David wrote, the values in ff43a1nb.itp are indeed correct.

Chris



>> 
>> Regards,
>> nkyx
>