[Pw_forum] constraining Cartesian lattice vector components

Nemeth, Karoly nemeth at aps.anl.gov
Sun Mar 16 02:41:06 CET 2014

Dear Quantum Espresso developers and users,

I was wondering whether it is possible to constrain individual Cartesian components of the lattice vectors during geometry optimization (vc-relax), at least for the ibrav=0 case. This would in principle be similar to constraining Cartesian atomic positions, however, it appears that this feature is not available in Quantum Espresso yet.

Also, it appears that during vc-relax the cell is slightly rotated. I would expect that the first two lattice vectors remain along the x axis and in the xy plane while the third one may have non-zero x, y and z components. In other words, if I start the optimization with the upper triangle of the cell_parameters matrix being all zero, I would expect that it stays so, and only the diagonals and the lower triangle will vary, which includes 6 numbers only. These six non-zero numbers represent the 6 lattice parameters in Cartesian space. Allowing all 9 numbers to vary in the cell_parameters matrix introduces redundancy that may lead to the rotation of the cell and slows down convergence.

I would greatly appreciate any advice in the above two (related) matters.



More information about the users mailing list