Cu(111) surface energy

[physicalattraction]

I wanted to calculate the ionic and electronic structure of a Cu(111) surface. I started with a Cu bulk calculation, and found a = 3.637 A using a 13x13x13 Monkhorst Pack grid, similar as what most other people find.

I then made several surfaces, and froze some of the atoms, varied the K-grid a bit and varied the number of Cu layers and vacuum layers. All relevant input parameters and output is shown in:



There are some things that are striking. First of all is that the surface energy which I calculate is not converged and deviates from what I found in literature (also in the attached png, Crljen = Vacuum 71, 101, 2003, Da Silva = Surf. Sci. 600, 703, 2006). My surface energy is 2 to 3 times higher than reported values. Did someone also did this calculation before? What were your findings?

Second is that, when I release the atoms of both sides of the slab, the atoms contract really a lot (distance between layer 1 and 2, D12, takes values up to -8%), which is not realistic. How is this possible in this calculation? I set NSW = 30, it did not ionically converge after 30 steps.

Lastly, I thought that setting up a VASP calculation would not be too difficult after you have some experience. Now, however, I find the next strange thing in my calculations happening, which should be an easy calculation. Is there a better handbook available than the manual or the handouts which are sometimes referred to on this forum, preferably on line?

Thanks for any help.

[alex]

Hi there again,

please explain your codes in the figure.
Are you aware that symmetry of the slab changes with the number layers you are using? This means, you'll have different ground states and different total energies. Do you know the way of calculating the surface energy without putting too much error in, outlined in an article by Methfessel and Fiorino (or similar) in 1996.

cheers

alex

[panda]

Did you divide the energy by 2, because there will be vacuum on either side? You should make several surface layers, each with varying ground state configurations, as alex suggests. Finally, did you optimize a supercell first?

[physicalattraction]

Thanks for your answers. I had not heard about the article you mentioned (Fiorentini and Methfessel, J.Phys. Cond Matt. 8 (1996) 6252), I will try this method now to determine the proper surface energy and come back here with the results.

The code in my image are as follows:
slabs: T = free to move, F = frozen atom, V = vacuum, so 3T1F3T5V means a seven-layer slab, consisting of three free layers, one frozen layer, three free layers and five vacuum layers. L = layer, this means I don't know if it is frozen or free.
k-grid: MP = Monkhorst-Pack grid.
D12 = change of distance between first and second layer, compared to bulk layers, given in percentage. Analogous for D23 and D34 for deeper lying layers.

I did already divide the surface energy by 2. Panda, what do you mean with a supercell? Is this a very thick slab, or more atoms per layer?

[panda]

more atoms per layer, eg more than 50 atoms for the calculation