Thanks for your rapid reply to my question and your suggestions. According to Axel’s suggestion, we have tried to use fix npt command (fix press/berendsen is not suitable for Triclinic system) to change the box, when we employ metadynamics with cell vector as CVs to study the phase transition of CL-20. No obvious phase transition has been observed, despite 200 metasteps (200ps). Of course, this may require further adjustment of the metadynamics parameters. But now, I’m worried about the rationality of using nPT ensemble in metadynamics with cell vector as CVs. Because in metadynamics theory, when we use cell vector as CVs, for each meta step, the lattice vector is adjusted by the width of the added Gaussian function, and the direction of adjustment is determined by the Gaussian potential and the original Gibbs potential, which is obviously different from NPT adjusting the lattice vector. Moreover, according to Pablo’s reply, there is a bug in the off diagonal component of the virial, thus, metadynamics may give an incorrect cell vector direction of adjustment.
Can fix colvars change the box directly, or i need to wait for the new fix plumed version?
Hello Jidong, no, fix colvars cannot use the unit cell vectors as collective variables.
It wouldn’t be a big effort to implement in Colvars that functionality. But this methodology has been developed and improved in Parrinello’s lab and more broadly in the labs of the people who develop and maintain PLUMED. So it’s never been a top priority for Colvars to just replicate that functionality without adding anything new
I or any of the other Colvars developers would be happy to take a second look at this issue, if appropriate. But the quickest course of action for you would be to wait for a patch by Pablo or any of the other people involved in Plumed.
Thanks for your rapid reply to my question and your suggestions. According to Axel’s suggestion, we have tried to use fix npt command (fix press/berendsen is not suitable for Triclinic system) to change the box, when we employ metadynamics with cell vector as CVs to study the phase transition of CL-20. No obvious phase transition has been observed, despite 200 metasteps (200ps). Of course, this may require further adjustment of the metadynamics parameters. But now, I’m worried about the rationality of using nPT ensemble in metadynamics with cell vector as CVs. Because in metadynamics theory, when we use cell vector as CVs, for each meta step, the lattice vector is adjusted by the width of the added Gaussian function, and the direction of adjustment is determined by the Gaussian potential and the original Gibbs potential, which is obviously different from NPT adjusting the lattice vector.
I would not worry about this, because the same principle also applies to atom positions. There the metadynamics calculations yield a biasing force which is added to the atom’s regular forces and then they require time integration of the positions to see the change. That time integration, however, is always present in an MD simulation. With cell dimensions the difference is that usually you do not do time integration of the box unless you have a variable cell time integrator. The additions to the virial are equivalent to the biasing forces on atoms in that respect.
Also, a simulation duration of 200ps is not very long. Phase transitions are not always straightforward and may require a significant bias. If that is the case here, is impossible to say without having specific knowledge about your project and your system.
Moreover, according to Pablo’s reply, there is a bug in the off diagonal component of the virial, thus, metadynamics may give an incorrect cell vector direction of adjustment.
In order to get a better sense for what are suitable parameters, I would suggest you try a simulation of a simpler system where the phase transition is easier to achieve (and ideally without tilting the cell) and use that to practice how to monitor and influence the process.
