Molecular Dynamics for Devices and Boundary Conditions

[jalmeida]

Dear,

It is my first time trying to do Molecular Dynamics (MD) simulation and I have some questions that I did not find in details in the documentation and in the forum.

Please, Does ATK use always periodic boundary conditions in MD simulations, even for finite systems, e.g. molecules?

I tried to use MD for a device, so the electrodes and electrode repetitions are fixed automatically, right? Some atoms is going outside of the cell!

Is it better keep just the scattering region in a bulk configuration? Does cell size change in MD?

How can I choose a suitable unit cell? Should I increase the cell and keep the scattering region centralized or use a supercelll with the ATK tool "repeat cell" in all directions?

Thanks in advance 

[Anders Blom]

I'll be rather brief, maybe you have follow-up questions after that.

Yes, we always use periodic boundary conditions in MD, although you can of course have vacuum around a structure like a CNT etc; since the range of interatomic forces in classical potentials is usually quite short (electrostatic forces may have longer range) this effectively makes the system 1D.

For devices, however, since the electrodes and their repetitions (more correctly, the electrode extension) are indeed always fixed, it's sort of not periodic in Z, since the atoms cannot move there. Not sure what you observe about atoms leaving the cell, you'd have to show an example.

Cell size will change if you are using NPT, but I would use the device structure, because a small change in atoms doesn't make a difference, at least for forcefields. For DFT, you would have to think carefully about doing an MD simulation for such a large structure as a device. Keeping it as a device structure makes it much easier to do a DFT-NEGF calculation afterwards, without worrying about the cell size change or atoms in the electrode extensions moving.

Do not repeat in Z if it's a device, but otherwise yes, you usually need a relatively large supercell, else you are limiting the degrees of freedom the systems can explore.

[jalmeida]

Thank you very much for your reply!

So, if I understand correctly, if I want to perform the transmission spectrum for some MD configurations, it is better to do it already in the device configuration to maintain the correspondence between my scattering region and the electrode extension. And NPT is not recommended because it can also cause mismatch, right?

It seems to me that the best way to perform molecular dynamics in devices is to use the NVT ensemble in different calculations with different sizes only in the z direction of the scattering region, as described in this tutorial <https://docs.quantumatk.com/casestudies/md_landauer/md_landauer.html>. In my case, the system is a single-molecule junction, so the correct way to increase the cell size (like in the tutorial) is to increase the number of gold layers in the scattering region?

[Anders Blom]

Yes, but NPT will still work fine in the device configuration, since the constraints work to keep the device correct. The problem with NPT is just if you run the central region alone; I think one can still get it work right with the correct constraints but might require rigid body regions rather than fixed atoms, since fixed atoms (as I recall) means fractional coordinates are fixed, and if the volume changes, the interatomic distances do too. This is described a bit more in https://docs.quantumatk.com/tutorials/device_relaxation/device_relaxation.html#central-region-relaxation (look for the Rigid body constraint).

[jalmeida]

Thank you! I will do both ensembles and see how the structure evolves.