I am calculating the polar surfaces of MgO. A slab alternantly arranged by a layer of Mg cation and a layer of O anion. So it will produce dipole moment in the slab and the energy will not converge. One of the solution is to add pseudoatom like pseudohydrogen on the Mg termination or O termination. But there are H, H.5, H7.5, H1.25 in VASP. pseudo hydrogen that could choose, but I don't know which one is proper. Can you give me some suggestion or give me another way to calculate polar surfaces
You have to count the number of electrons per bond. MgO crystallizes in a NaCl structure, that means each Mg (O) atom has six nearest neighbours. Therefore, each oxygen atom contributes 1 electron (of the six valence electrons) to each bond. On the other hand each Mg-atom contributes 1/3 electron (of the 2 valence electrons) to each bond.
That means a dangling bond of a Mg-atom have to be passivated by a H pseudohydrogen atom (i.e. an ordinary hydrogen atom), while a dangling bond of oxygen atom have to be passivated by a H.33 pseudohydrogen atom.
Dear Ms. Liu,
It looks that you are working at the {111} surfaces of a MgO with a NaCl-type structure. The MgO(111) surfaces are either a Mg- or an O-termination when you cleave it. .
If you want to get a stable {111} surface, you can also (a) use a domain model, that is by making unveven MgO (111) surface with both Mg/O at the top of the surfaces.
You can also (b) use the (OH) as terminations. Then you need to add one H on each surface O or one (OH) on each surface Mg .atom/ion. To make charge balance, you can use the normal H as suggested by Roman..
To deal properly with polar surfaces you would probably best use an AO-based periodic DFT code such as CRYSTAL or BAND (you're welcome to try out the latter for free), which can properly deal with 2D periodic boundary conditions
I understood that one can use a dipole correction in 3D plane-wave codes for treating dipolar surfaces in the slab-gap approach. This should probably be available in VASP and may be a better approximation than passivating your surface with (pseudo-)hydrogen atoms. I reckon that would diminish the surface polarization altogether. But to allow full polarization across a surface, one may be better of with a proper 2D periodic boundary conditions.
I think one should step back and ask what it is you want to understand and what you think the physical system looks like. Then you can decide what calculation is best (and the above technical comments are useful for that).
You say you want to study a polar surface of MgO. As you say, your layers are altenatively Mg and O which is polar. If you make this object very thick and somehow make the electrons frozen in place, it has a dipole per unit cell and thus an electric field inside which makes for a total energy diverging with thickness. This just says that polar surfaces reconstruct in some way and can't just "sit there" with no changes.
If you actually run an ab initio calculation of an MgO slab that is polar and you thicken the slab to simulate an actual thick material, the total energy *WILL* converge. This is because the electrons will not just sit in place on the atoms as in the bulk. They will move around to remove the dipole and field inside. Once the slab is thick enough (typically a few unit cells) then electrons will leave the anionic surface (valence band maximum) and move to the opposite cationic surface (conduction band minimum) and thus create a new dipole that cancels the ionic dipole due to the polar choice of layering. This happens automatically in the self-consistent calculation since it lowers the total energy. This is an electronic reconstruction. For a very thick slab, the amount of transferred electrons saturates to a value that exactly cancels the ionic dipole and gives zero field inside the slab.
In reality, electronic reconstruction is just one possibility. Some atoms can also be removed or added which is atomic reconstruction and this is *VERY* common on polar surfaces. For example, on the Mg terminated surface which is nominally positively charged, the removal of some Mg or addition of some O (or both) will effectively add negative charge. The opposite happens on the O surface. These atomic reconstructions (change of surface stoichiometry) can also remove the polar dipole. In practice, for insulating materials, the atomic reconstructions generally have lower total free energy than the pure electronic reconstructions.
Now for the question for you to answer: what are you trying to model? If you want to understand the surfaces phases of polar MgO surfaces let's say in a clean ultra-high vacuum system, you then should not be putting hydrogen on the surface or anything like that since you are trying to see what the actual surface is doing. You need to compute total energies for clean surfaces (ideal surfaces with electronic reconstruction only) as well as a large number of defective surfaces with various concentrations of missing Mg and O or added Mg and O. Then you compute the free energy with chemical potentials to make a phase diagram and see what happens: the surface state depends on the environment; for example, under high oxygen pressure (high oxygen chemical potential) the surface might have added O atoms while at low pressure it might not or might even have O vacancies. This method is well established in the literature.
Once you figure out what type of question you want to answer, then you can figure our the technique to do it. For examples
(1) With polar systems, if you are using plane wave methods and have a slab in vacuum with periodic boundary conditions, then you should try to as much as possible to use a dipole correction so that there are not fields in the vacuum (correct electrical boundary conditions).
(2) A slab usually has two surfaces let's call them A and B. If you are interested in what is going on with surface A but not B, then you have some freedom in dealing with B to make computations easier. For example, maybe you can make the slab thin and still get converged energies if you passivate the B surface with (pseudo)hydrogen. Or maybe you don't relax the atoms on the B surface and hold them on the layers above them to bulk coordinates to simulate the proper mechanical boundary conditions for surface A. Sometimes in polar systems, it is good to make A and B identical surfaces (a non-stoichiometric slab) and the added symmetry automatically gets rid of a polar field inside the material which is the correct electrical boundary condition going into the material from the surface assuming the material is thick.
- Badges
- Science topic
- Similar topics
- Analytical Chemistry
- Extraction