[gmx-users] triclinic option

[Erik Lindahl]

Makoto Yoneya wrote:

>Hi Erik:
>
>Thanks for your quick response.
>The triclinic I mean the MD cell with
>a .ne. b .ne. c and alpha .ne. beta .ne. gamma in the ordinary crystal
>manner.
>I don't know whether these a, b, c, alpha, beta, gamma correspond to
> box_xx box_yy box_zz box_xy box_xz box_yz in GROMACS.
>

The Gromacs parameters are the coordinates of the cell unit vectors in
nanometers. To remove redundancy we assume that the first vector is
along the x axis (i.e. the coordiantes are [box_xx,0,0]), the second
lies in the xy-plane specified by the coordinates [box_xy,box_yy], while
the third vector has the coordinates [box_xz,box_yz,box_zz].

>
>At the same time, I'd like to do MD simulations something like
>crystal phase transition e.g., from orthorhombic (a .ne. b. ne. c and
> alpha .eq. beta .eq. gamma .eq. 90deg) to triclinic (see above).
>According to your response, should I use aniisotropic pressure option
>with six compressibilities and three reference pressures, not the triclinic
>pressure option?.
>If so, what the situations to use this triclinic pressure option?
>
The triclinic pressure option doesn't exist anymore - we removed it
since it made much more sense to control things with the
compressibilities instead (this allows more control: you can for
instance allow very small shape changes by using small but non-zero off
diagonal elements).

If you want to do accurate simulations involving shape-changes you might
want to use parinello-rahman pressure coupling, but a word of warning:
we haven't tested it on crystal transitions, so start by checking the
results for some simple system. Nos? & Klein did some work on this years
ago...

Cheers,

Erik