<div class="gmail_quote"><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">On Wed, Nov 19, 2008 at 2:14 PM, Zhao Lifeng <<a href="mailto:zhao.lf@gmail.com">zhao.lf@gmail.com</a>> wrote:<br>
<br>
><br>
> Dear all,<br>
><br>
> I encountered a problem in alchemistry free energy calculations:<br>
><br>
> In my simulations a group (A) is to be changed to (B) in a molecule, while<br>
> the atom numbers n(A) is not equal to n(B).<br>
> A scheme is to include both A and B in the same topology file and switch<br>
> on/off the B/A charges and vdw interactions.<br>
> And interactions between A and B were excluded in the simulations. However,<br>
> the vibration mode would not be correct<br>
> for both A and B state.<br>
><br>
> Another scheme is to change the corresponding atoms to zero mass dummy<br>
> atoms or from them. In this case, there will be<br>
> no interactions on the dummy atoms, either bonded or non-bonded<br>
> interactions.<br>
><br>
> I don't know if these schemes are correct and if there is any other more<br>
> reasonable schemes. In my simulations I prefered the<br>
> later one and thus the problems came.<br>
> In state of lambda=0 or =1, since there are dummy atoms with no force on<br>
> them, (sometimes with velocities, e.g. from<br>
> the last lambda simulation), the dummy atoms often bring forth errors,<br>
> for example, 1-4 distance too large, neighbor list error<br>
> and segmentation fault. Bond constraints seems work but it might not be<br>
> reasonable for heavy atoms.<br>
><br>
> Any suggestions would be appreciated.<br>
<br>
<br>
Hello,<br>
<br>
I have used schemes similar to the one you described and to circumvent the<br>
problems associated with growing or disappearing particles out of nothing I<br>
used soft-core interactions and manipulated the soft-core parameters to get<br>
"nice" dgdl curves. Hope this helps.<br>
<br>
Diana Lousa<br>
PhD student<br>
Protein Modelling Laboratory<br>
ITQB/UNL<br>
Oeiras, Portugal<br>
<br>
><br>
> Lifeng Zhao<br></blockquote></div><br><br>Thanks Diana. But softcore does not help for it only affect vdw interactions. <br>In my TI simulation case, the dummy atoms has no mass and the bond interactions <br>are turned off. (If bond interactions were not turned off there would produce <br>
a segmentation fault for forces on zero mass.) <br><br>Lifeng Zhao<br>