Building interfaces in VNL

[star65]

Hello,

I am trying to build an interface of Li/Li3N with VNL. I followed the tutorial but I am not sure about the results of the interfaces. I couldn't find the best orientation of Li and Li3N in this interface in the literature. I think the interfaces I made looks weird.
Could you please comment on that? Another question is when I use for example Li(100) and Li3N(100) in the interface builder  the initial strain is around 6% which is high then  I change the set matching parameters and chose another point from the plot with lower strain. In this case the final structure of the slabs look different from what I used in the beginning and I think they are connecting with different directions. Is this process correct and which straining method is good to be used? I attached the figure of the input  generated with VNL.
These are the parameters when I use the strain on the Li metal (strain first surface)

E11 = 2.03 %  E22 = - 0.03% E12 = 0.00%   Mean Absolute Strain = 0.69%

I use this input for calculation in vasp. I am wondering when this interface is  fully relaxed, how can I measure the strain in the final structure?

Thank you in advance.

[Petr Khomyakov]

What exactly do you mean by "weird" regarding the structure you have enclosed? I see nothing obviously wrong with it.

I am not sure I fully understood your second question. The interface builder allows you to match two structures by rotating them with respect to each other (i.e., by choosing different shapes and dimensions of lateral unit cells). This process is automated and produces a set of interface structures with different strain configurations corresponding to particular choices of the lateral unit cells I have mentioned in the previous sentence. This is fully geometrical procedure, and it is up to you to decide on which ones suit your needs.

In general, one usually aims at structures with no or minimal strain. In practice, one may want to avoid dealing with huge structures, and in this case one should tolerate some strain in the interface structure. But that is really up to you to decide what is good for a particular problem of your study. For Li|Li3N, I would perhaps allow some strain on the metal side, i.e., in Li, and this is what you have done.  You may certainly consider an interface structure with less strain, but you will then likely need to enlarge your lateral unit cell size. 

If you only do ion optimization of your interface structure, the macroscopic strain will not change for the final structure. Otherwise, you have to use in-plane lattice parameters of the final interface structure to recalculate the strain tensor.

[star65]

Thank you for your explanations !

- Are the parameters that I added to my first post  good enough?

- How do I connect the way VNL build interfaces with the procedure explained in this paper: J. Mater. Chem., 2012, 22, 22063 , at page 22064 (The Interface model construction)

- In section 2.1.2, commensurability of two phases is explained. Is this the same way that VNL calculate the mismatch?(I mean in eq. 2)

- In fig. 1,  the steps leading to an interface model are shown.  Does VNL do all of these steps?  if yes,  how step 3  is done with VNL? 

[Petr Khomyakov]

- Are the parameters that I added to my first post  good enough?

I guess noone can say whether your parameters are good enough or not. You should actually study it by choosing other generated interface structures to find out how sensitive physical properties of your system to interface strain.

- How do I connect the way VNL build interfaces with the procedure explained in this paper: J. Mater. Chem., 2012, 22, 22063 , at page 22064 (The Interface model construction)

I would say that is the same procedure, except that the interface builder does not fully relax the interface structure; it may do some separation distance optimization using classical MD, provided there exist corresponding classical potentials. After the structure is built, you may however optimize it by means of other ATK tools. 

- In section 2.1.2, commensurability of two phases is explained. Is this the same way that VNL calculate the mismatch?(I mean in eq. 2)

As I said the interface builder provides geometrical construct of interfaces, so it does not use an equation such as Eq. 2, where there are some non-geometrical (energy) parameters. The interface builder uses mean strain = (e11+e22+e12)/3 as a measure for the interface matching quality. Having mean strain = 0 corresponds to  perfectly matched surfaces.

- In fig. 1,  the steps leading to an interface model are shown.  Does VNL do all of these steps?  if yes,  how step 3  is done with VNL?

See my reply to your first question. In step 3, the interface builder uses two surfaces (cleaved in a separate Builder plugin, see http://docs.quantumwise.com/tutorials/ag_au_interface/ag_au_interface.html for more details) to match them in different ways. 

[star65]

Thank you, Petr!

I guess noone can say whether your parameters are good enough or not. You should actually study it by choosing other generated interface structures to find out how sensitive physical properties of your system to interface strain.

I read in the literature that about 2 or 3% strain is ok. Mean value strain reported in VNL refers to that term, is this correct?

I know that this is a compromise between cell size and strain. However, I don't know when I make different interfaces and I want to compare their mechanical stability or report the strain in them whether I should take E11, E12 , E22 or mean value starin? (Considering strain on the Li metal slab)

 Some of these values are negative, what does that mean?

[Petr Khomyakov]

The VNL interface builder reports interface strain components and mean strain value (as specified in the previous post) for a chosen supercell structure.

I do not know what these 2-3% in the literature refer to. It may for example refer to the mismatch for the two surfaces matched with the minimal in-plane unit cells.  This is something one should evaluate first to see how lattice-matched your structures are; this can be done simply by comparing the in-plane lattice constants of the two surfaces. Some materials do not grow on top of each other if the lateral strain is too large.

I would say the individual components have real physical meaning, whereas their average is not uniquely defined, and more meant to give an overall idea of how far the interface is from being perfectly lattice-matched.   

Strain can be positive and negative, depending on whether the structure is compressed or stretched in certain directions.

[star65]

Thank you for your reply!

[Petr Khomyakov]

Please take a look the technical notes on how the Interface Builder, https://docs.quantumwise.com/technicalnotes/interface_builder/interface_builder.html#interface-builder.