I'm not a QE user, but I develop a similar program (CASTEP) and I don't think there should be any significant differences in how you approach this situation.

When simulating a molecule in the gas-phase, you need to ensure your lattice unit cell is large enough that the molecules do not interact strongly with each other. (You would usually only put a single molecule in the unit cell, but it will still interact with its periodic images in neighbouring cells.)

In general, you need to try a range of different unit cell sizes to investigate the effects of the spurious inter-molecular interaction. It is difficult to know beforehand exactly how large your unit cell should be, since it depends on the molecule, but as long as it doesn't have a large (macroscopic) dipole moment then I would expect somewhere between 10 and 16 A to be enough separation -- so your unit cell would be the width of the molecule plus 10 to 16 A. I usually start investigating at 8 A to get a baseline, and converge from there -- look at energies and forces, and any properties of especial interest (e.g. DOS features).

When you do your geometry optimisation, make sure that you do *not* optimise the lattice vectors, just optimise the atomic positions. This will probably be faster too, since it won't need to compute the stress on the cell (and that's quite computationally demanding).

As a side-note, if you sample the Brillouin zone at the gamma-point you will need a larger separation compared to sampling at the (1/4,1/4,1/4) point, though the gamma-point calculation may be faster. This is because (a) the gamma-point maximises the constructive interference between the wavefunctions of neighbouring cells (they are all in phase), and you want to minimise the interference, but (b) at the gamma-point the wavefunctions are real (not complex) in real-space, so there is less work to do.

By the way, a simple cubic cell is the simplest to set up, but you may find something more isotropic (e.g. a hexagonal phase) is actually more computationally efficient.

Note that none of these considerations are specific to your molecule, so you can quickly see what kind of behaviour you get with something small and simple. I would suggest you try something like methane, and then again with methanol (to see the effect of electrostatics). These will be really quick, so you can do a very good job very quickly and get all your mistakes out of the way!

Hope that helps,

Phil Hasnip

I'm not a QE user, but I develop a similar program (CASTEP) and I don't think there should be any significant differences in how you approach this situation.

When simulating a molecule in the gas-phase, you need to ensure your lattice unit cell is large enough that the molecules do not interact strongly with each other. (You would usually only put a single molecule in the unit cell, but it will still interact with its periodic images in neighbouring cells.)

In general, you need to try a range of different unit cell sizes to investigate the effects of the spurious inter-molecular interaction. It is difficult to know beforehand exactly how large your unit cell should be, since it depends on the molecule, but as long as it doesn't have a large (macroscopic) dipole moment then I would expect somewhere between 10 and 16 A to be enough separation -- so your unit cell would be the width of the molecule plus 10 to 16 A. I usually start investigating at 8 A to get a baseline, and converge from there -- look at energies and forces, and any properties of especial interest (e.g. DOS features).

When you do your geometry optimisation, make sure that you do *not* optimise the lattice vectors, just optimise the atomic positions. This will probably be faster too, since it won't need to compute the stress on the cell (and that's quite computationally demanding).

As a side-note, if you sample the Brillouin zone at the gamma-point you will need a larger separation compared to sampling at the (1/4,1/4,1/4) point, though the gamma-point calculation may be faster. This is because (a) the gamma-point maximises the constructive interference between the wavefunctions of neighbouring cells (they are all in phase), and you want to minimise the interference, but (b) at the gamma-point the wavefunctions are real (not complex) in real-space, so there is less work to do.

By the way, a simple cubic cell is the simplest to set up, but you may find something more isotropic (e.g. a hexagonal phase) is actually more computationally efficient.

Note that none of these considerations are specific to your molecule, so you can quickly see what kind of behaviour you get with something small and simple. I would suggest you try something like methane, and then again with methanol (to see the effect of electrostatics). These will be really quick, so you can do a very good job very quickly and get all your mistakes out of the way!

Hope that helps,

Phil Hasnip

You should not use a periodic code for a molecule in the gas phase. Thats like trying to drive somewhere in an aeroplane. It can be done in principle, but why wouldn't you just use a car instead? To come back to your question: Use a molecular code instead, like e.g. ORCA. They are usually easier to operate, have more features and are much faster (for isolated molecules).

As said above, If you want to use a code based on periodic boundary conditions as QE (obviously, being based on plane waves) you can do it, but you have to test the convergence of the cell dimension with respect to the properties you are interested in. If you have a dipolar molecule, you can add a correction term. Also, use the command assume_isolated. Hope it helps