ADDGRID tag impact

Bader charge density analysis

Moderator: moderators

Post Reply
eduard
Posts: 21
Joined: Tue Sep 02, 2008 1:40 pm

ADDGRID tag impact

Post by eduard »

Dear All,

I am trying to calculate the Bader charges for a molecule which binds via a carboxylate (OCO) group onto a Cu (011) surface.
To do this, I have first obtained the convergence for my system using VASP 5.2.2.

To calculate the Bader charges I have set in my INCAR file the tags:

LAECHG = TRUE
NGXF = 486
NGYF = 140
NGZF = 100

(The default values for NG(X,Y,Z)F were 486/70/50. This for a unit cell of: 35.000 X 5.142 X 3.636 A^3 ).

With these settings, the grids generated by VASP are:

dimension x,y,z ////// NGX = 288 ////// NGY = 42 //////// NGZ = 30
dimension x,y,z ////// NGXF= 486////// NGYF= 140 ////// NGZF= 100
support grid ///////// NGXF= 486 ///// NGYF= 140 ////// NGZF= 100

and the Bader charges obtained by using the last binaries:

Atom type ////// # ////////// X //////// Y /////////// Z //////// CHARGE //////// MIN DIST ////// ATOMIC VOL
--------------------------------------------------------------------------------------------------------------------------------------------------------------
Cu_2 //////////13 ////// 11.6954 ////// 1.2946 ////// 3.6362 ////// 11.0075 ////// 1.1817 ////// 13.3379
Cu_2 //////////14 ////// 11.6864 ////// 3.8668 ////// 3.6349 ////// 10.9996 ////// 1.1842 ////// 13.1022
Cu_1 //////////15 ////// 12.9327 ////// 2.5844 ////// 1.7901 ////// 10.7047 ////// 0.8945 ////// 14.1831
Cu_1 //////////16 ////// 12.9339 ////// 0.0171 ////// 1.8480 ////// 10.7065 ////// 0.8935 ////// 14.1787
C /////////////17 ////// 15.4285 ////// 3.8781 ////// 1.7612 ////// 2.4450 ////// 0.3646 ////// 5.6355
O /////////////23 ////// 14.8625 ////// 4.9736 ////// 2.0816 ////// 7.0849 ////// 0.7555 ////// 17.0275
O /////////////24 ////// 14.8514 ////// 2.7828 ////// 1.4879 ////// 7.0782 ////// 0.7540 ////// 17.1990

The Bader charges look very reasonable to me as they show that the Cu atoms from the first layer (surface layer, Cu_1 atoms in the table) and the C atom give charge to the O atoms. (Oxygen binds directly to Cu_1 atoms). The surface second Cu layer (Cu_2) seems unaffeected (in vacuum I found that Cu=11e, C=4e, O=6e).

As I have read that improved accuracy can be obtain if one consider also the LADDGRID tag, I have added this tag also to my INCAR.

However, this tag seems to have absolutely no impact on the VASP 5.2.2 grid or on Bader charges.

Finally, I have done a last test, considering the ADDGRID tag (no L here !!!). The presence of the ADDGRID tag leads to an increase dimension for the 3-rd grid. Thus, by setting this tag the grids become:

dimension x,y,z ////// NGX = 288 ////// NGY = 42 ////// NGZ = 30
dimension x,y,z ////// NGXF= 486 ////// NGYF= 140 ////// NGZF= 100
support grid ///////// NGXF= 972 ////// NGYF= 280 ////// NGZF= 200

(notice the increased dimension for the 3-rd grid).

The Bader charges modify now, but in a way which surprise me:

Atom type //// # //////// X /////////// Y /////////// Z ////// CHARGE ////// MIN DIST ////// ATOMIC VOL
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Cu_2 ////// 13 ////// 11.6954 ////// 1.2946 ////// 3.6362 ////// 10.8426 ////// 1.0838 ////// 13.1094
Cu_2 ////// 14 ////// 11.6864 ////// 3.8668 ////// 3.6349 ////// 10.7522 ////// 1.0654 ////// 11.9108
Cu_1 ////// 15 ////// 12.9327 ////// 2.5844 ////// 1.7901 ////// 11.4632 ////// 0.9075 ////// 19.4473
Cu_1 ////// 16 ////// 12.9339 ////// 0.0171 ////// 1.8480 ////// 11.2575 ////// 0.8982 ////// 17.6149
C ////////// 17 ////// 15.4285 ////// 3.8781 ////// 1.7612 ////// 2.5873 ////// 0.3576 ////// 6.2301
O ////////// 23 ////// 14.8625 ////// 4.9736 ////// 2.0816 ////// 6.9065 ////// 0.6229 ////// 14.4517
O ////////// 24 ////// 14.8514 ////// 2.7828 ////// 1.4879 ////// 6.8086 ////// 0.6322 ////// 16.1091


Thus, the two Cu atoms from the first layer (Cu_1) no longer give any charge to the oxygens (!!!) but rather receive charge from the Cu atoms from the layer beneath (Cu_2 atoms).

I would like to ask the more experience Bader developers and user to help me clarify:

1) the impact of the ADDGRID tag : is this a useful settings in doing Bader analysis ? Which is your experience in having it in the INCAR ?
2) - and just for curiosity - is the LADDGRID tag of any use ? In my case having it set or not in the INCAR file changes nothing.

Thank you very much for having read my post and for any help or hint !
Ed

PS. I used //// as column delimitator. I am sorry if it does not look good.
graeme
Site Admin
Posts: 2256
Joined: Tue Apr 26, 2005 4:25 am
Contact:

Re: ADDGRID tag impact

Post by graeme »

With the ADDGRID tag, it looks like the oxygen atoms still accept charge -- about 0.9 instead of 1.1 e.

You do have to find convergence of the charges with respect to grid density. Using a 2x finer grid is about the minimum for an accurate calculation, and going finer to check for convergence would be good. Perhaps you also don't need such a large vacuum along x! Reducing this will make the charge density files and the calculations less expensive.
Post Reply