Hi, I just wanted to add my experience using some of these options. I have found that running a single optimization using MD.VariableCell True virtually always finds the same optimized structure as doing a two-step (frozen cell -> variable cell) optimization. This is true as long as the starting coordinates are reasonably accurate.
In my understanding, if you use the Monkhorst_Pack k-grid during variable cell optimization, the number of k-points won’t change, but their effective density will change a little bit. For example, if the unit cell shrinks during optimization, the same number of k-points will reflect a less dense k-grid. Possibly one approach would be to optimize the system with a reasonable guess for the number of k-points necessary and then, when you have the optimized coordinates, run a single energy calculation with a larger number of kpoints to see if the energy/forces/strain change significantly. Lastly, I’ve found that the MD.TypeOfRun Broyden is the most efficient optimization algorithm. The default settings for it seem to work fine, too. Anyway, as I said, these are just my experiences, so they certainly won't be helpful for every situation. I thought I would share though, just in case they might be helpful for someone else. Have a nice week everyone, Mike On Jul 21, 2014, at 8:10 AM, Максим Арсентьев <ars21031...@gmail.com> wrote: > When you perform internal optimization with MD.VariableCell false, cell > parameters are fixed, so if you further perform full opt. with this option > "true" cell parameters will be the same. So in both these cases cell > parameters are the same and you should chose the same kgrig_Monkhorst_Pack > block and Briollin zone will not change when you transfer from internal to > full geom. optim. > > > 2014-07-21 15:38 GMT+04:00 Kanhao Xue <xuekan...@gmail.com>: > It is strange argument. When the cell parameter changes, the Brillouin zone > will change, therefore the Monkhost-Pack K points will change their > locations. However, one is just concerned with the rough density of K points > in this scheme. Why one cannot use Monkhorst-Pack? > > > On Mon, Jul 21, 2014 at 11:48 AM, 邵德喜 <dxshao...@gmail.com> wrote: > I was told that when MD.VariableCell is setted True, then we can not use > block kgrid_Monkhorst_Pack option . > Thanks very much . > > > 2014-07-21 15:31 GMT+08:00 Максим Арсентьев <ars21031...@gmail.com>: > > You right, you should perform 2 structural relaxations. Before doing this do > calculations without geom. optim. with different density of kpoints by > "%block kgrid_Monkhorst_Pack" and see convergence of kpoint density on > forces. This also should be checked for meshcutoff - convergence in forces is > very important for geom. optim. > > > 2014-07-21 6:59 GMT+04:00 邵德喜 <dxshao...@gmail.com>: > > Thanks very much. > Do you mean that I should do structual relaxation twice?First with > MD.VariableCell False,then with MD.VariableCell True? > If I could just do structual relaxation once with MD.VariableCell True? And > if " kgrid Monkhorst Pack" option can be reserved at the same time? > Yours Sincerely! > Dexishao > > > 2014-07-21 4:55 GMT+08:00 Максим Арсентьев <ars21031...@gmail.com>: > > Dear, Dexi! > > You can vary residual strain by MD.MaxStressTol. > Use CG, relax with MD.VariableCell False first, then input coordinates > obtained to a new file with MD.VariableCell True. > Note you may need to shift some atoms manually before geometry optimization: > if all atoms are in 2D plane, then all of them will remain in plane after > geom. optim. because z forces are zero, so you need to shift any atom to > unlock z forces and avoid such a false state. > > > 2014-07-20 4:01 GMT+04:00 邵德喜 <dxshao...@gmail.com>: > > Dear everyone: > I really want to know how to get a structure wuthout strain.For example > ,Let's consider graphene ,I get the structure from materials studio,and use > it as initial one for siesta structual relaxation. > If in the relaxation set some options as follows: > MD.TypeOfRun CG/Verlet > MD.VariableCell True > ... > At the same time ,I replace " kgrid Monkhorst Pack" option with k-grid cutoff. > When the relaxation is done,I want to ask if the final structure after > relaxation is > one structure without strain? > In general ,how to get a structure without strain is what I care. > Thanks in advance. > Dexi Shao > > > > -- > Best wishes, > Maxim Arsent'ev, D.Sc.(Chemistry) > Laboratory of research of nanostructures > Institute of Silicate Chemistry of RAS > > > > > -- > Best wishes, > Maxim Arsent'ev, D.Sc.(Chemistry) > Laboratory of research of nanostructures > Institute of Silicate Chemistry of RAS > > > > > -- > > Sincerely yours, > Kanhao Xue > > Laboratoire de Réactivité et Chimie des Solides (LRCS) > UMR CNRS 7314 > Université de Picardie Jules Verne > 33 Rue Saint Leu, 80039 Amiens Cedex, France > > > > -- > Best wishes, > Maxim Arsent'ev, D.Sc.(Chemistry) > Laboratory of research of nanostructures > Institute of Silicate Chemistry of RAS > > > -- > Best wishes, > Maxim Arsent'ev, D.Sc.(Chemistry) > Laboratory of research of nanostructures > Institute of Silicate Chemistry of RAS >
