projected bands

Message

giacomo giorgi · #1 Post by **giacomo giorgi** » Wed Apr 14, 2010 3:21 pm

Dear all,
I want to check if the surface induces states in the gap of the bulk. In other words,　I want to see if my 18 atom slab of anatase TiO2 induces states in the gap of bulk anatase. How to compare the gap of the two systems? Is it sufficient to allign eigenvalues simply subtracting the Fermi energy, thus setting the Fermi level as zero for both systems? Must I allign them with the vacuum level (and thus also subtracting the alpha+ bet value)?

Please consider that I optimized the first ionic positions of the first two layers of TiO2 surface,　keeping frozen the bottom layers.

Could you kindly give me a hint?
Best,
G

#2 Post by **admin** » Wed Apr 14, 2010 3:35 pm

please align the states to the same vacuum level. But please do not use alpha+bet, but the constant energy level in the vacuum, as obtained by setting LVTOT = .True. and by running vtotav with the LOCPOT (written if LVTOT is set) file as input.

giacomo giorgi · #3 Post by **giacomo giorgi** » Wed Apr 14, 2010 3:57 pm

Dear administrator,
thanks a lot for the quick reply!
LVTOT must be set only in　the NON self consistent calculation. Correct?
Best,
G

alex · #4 Post by **alex** » Thu Apr 15, 2010 4:58 am

The only problem is that there is no vacuum level in a bulk calculation. You have to use thick slabs for that case ...

Cheers,

Alex

giacomo giorgi · #5 Post by **giacomo giorgi** » Thu Apr 15, 2010 7:06 am

Thanks Alex! Thus I deduce that it would take a lot of computational time. THICK SLAB+ Vacuum..... (how thick?)

If I properly understand a real comparison surface-bulk band structure seems not to be feasible

What about a DOS comparison between thin slab and bulk? Of course in this case I only focus my attention on Gamma.

Best,
G

#6 Post by **admin** » Thu Apr 15, 2010 10:13 am

1) yes, LVTOT is one of the parameters typically to be set in a post-processing run (ie one additional run without relaxing the ions after everything has been converged.
2) concerning the thickness of the slab in order to obtain bulk-like behaviour in the center layer(s): please have eg a look at the approaches suggested by
V. Fiorentini and M. Methfessel, J. Phys Cond Matt. 8 (36), 6525 (1996)
J.C. Boettger, PRB 49, 16798 (1994)

tlchan · #7 Post by **tlchan** » Thu Apr 15, 2010 5:56 pm

If there are deep localized levels, you can use them to align the two eigenvalue spectrums as well.

giacomo giorgi · #8 Post by **giacomo giorgi** » Sun Apr 18, 2010 1:42 am

Thanks to everybody. Let me see if I have properly understood. I add the correction (calculated with vtotav) along the normal to the surface for both my slab (say, z) and the bulk (thick slab). Then I align the corrected values of the Fermi Energy (or of deep localized levels). Correct? Thanks once more,
G

tlchan · #9 Post by **tlchan** » Mon Apr 19, 2010 6:52 pm

I would do the problem using one of the following methods:

1. Do a surface calculation, and a separate bulk calculation (using bulk, not thick slab). For the bulk calculation, find the average SCF potential (V_bulk). For the surface calculation, also find the average SCF potential (V_slab, for example using vtotav) around the center of the slab. The difference between V_bulk and V_slab is the shift between the eigenvalue spectrum of the bulk and surface calculations.

2. Do the same two calculations as above, but align the two eigenvalue spectrums using deep localized levels, if any.

3. If you have a very thick slab for your surface calculation, then the middle of the slab is bulk-like. Then you can do a projected density of states in the middle region. The projected density of states will have an energy gap, which corresponds to the bulk gap. You don't need to do a separate bulk calculation for this case, but you need a thick slab for the projected DOS to be accurate.

giacomo giorgi · #10 Post by **giacomo giorgi** » Wed Apr 21, 2010 12:04 pm

Dear Tlchan,
thanks for the very useful explanation.
I decided to follow the first way.

If I properly understood I take the averaged potential of one atom of the bulk (say, 1 O, their potential is equivalent of course).
Then I take the same potential for the innermost O in the slab.
Make the difference. And add this value to the spectrum of the slab.

I have one last doubt. Which value for the correction of the potential? the one in VLINE (I mean, the last value of Y column) or the value ( a very small number) I got as output when I run the vtotav program?
Sorry, since I am very very new with slab physics, so I feel like a dummy.

Thanks in advance,
G