Re: Work function calculations and the PAO cutoff radius ( No.1 ) 
 Date: 2017/12/15 23:33
 Name: ChiCheng Lee <cclee.physics@gmail.com>
 Hi Konstantin,
Could it be the problem already exists in bulk calculation? For example, are the band structures of bulk consistent from different cutoff radii? If the band structure of Mo11.0 shows some ghost bands, we are lucky to know it is due to the over completeness. I have never calculated work function, but I guess adding some empty atoms in the vacuum could help you get more accurate values.
Cheers, ChiCheng

Re: Work function calculations and the PAO cutoff radius ( No.2 ) 
 Date: 2017/12/16 05:03
 Name: Konstantin Khromov <k.yu.khromov@gmail.com>
 An update to my question.
The original question was related to openmx version 3.7 and Mo11.0.pao coming with the package.
Now I've recalculated the work function using openmx version 3.8 and Mo11.0.pao provided with the new version, keeping all the parameters exactly the same. The work function now is about 12 eV, much better that 30 eV, but still too large.
I actually tried introducing empty atoms into the vacuum region for Mp9.0.pao, resulted in less than 0.2 eV difference from the case with no empty spheres.

Re: Work function calculations and the PAO cutoff radius ( No.3 ) 
 Date: 2017/12/16 07:26
 Name: ChiCheng Lee <cclee.physics@gmail.com>
 Hi Konstantin,
Have you checked the band structures of bulk or slab with different radii? Are they very different or similar? I was guessing the longer radius could give you a basis that is over complete. In that case, you should not trust the result.
Cheers, ChiCheng

Re: Work function calculations and the PAO cutoff radius ( No.4 ) 
 Date: 2017/12/16 08:36
 Name: T. Ozaki
 Hi,
I think that the case of 11 a.u. is due to the overcompleteness of basis functions which can be confirmed by an erratic band dispersion for the bulk calculation.
We have performed calculations of work function for metals before, and found that OpenMX well reproduces the values comparable to those by planewave methods. An example can be seen in the page 59 at
http://www.openmxsquare.org/video_lec/OpenMXHandson2014Oct10.pdf
Adding a single ghost layer reaches almost convergent result, while 'no ghost' underestimates the work function of Al(111) surface. For other metals, we confirmed a similar behavior.
Regards,
TO

Re: Work function calculations and the PAO cutoff radius ( No.5 ) 
 Date: 2017/12/16 19:54
 Name: Konstantin Khromov <k.yu.khromov@gmail.com>
 Hi, guys
Thanks a lot for your suggestions. I just want to clarify if a "ghost" is a synonym for an empty atom as described in the openmx manual?
Konstantin

Re: Work function calculations and the PAO cutoff radius ( No.6 ) 
 Date: 2017/12/16 21:29
 Name: ChiCheng Lee <cclee.physics@gmail.com>
 Hi Konstantin,
I would consider "ghost" as "not real". When I mentioned ghost bands, I meant the bands not belonging to the real band structure. The ghost layers means a layer not belonging to the real structure but just a basis.
Cheers, ChiCheng

Re: Work function calculations and the PAO cutoff radius ( No.7 ) 
 Date: 2017/12/16 23:39
 Name: Naoya Yamaguchi
 Hi,
I also calculated the work function previously for M(111) surfaces (M=Al, Cu, Ag, Au, Pd), and attach a slide about it to the following URL for reference: https://gist.github.com/Ncmexp2717/eceeac40c55153374fda4824e173720b/raw/7cfaa1fd89bf4418769777979b7a7d607d6c9dfe/slide_wf_reduced.pdf
I also confirmed that the values improves by using empty atom scheme for any case. And, I think two atomic layers are enough. In the calculations, only PAOs appropriate for bulk systems were used (e.g. 7.0 Bohr for Au). The calculated values were in agreement with experimental ones in the empty atom scheme (See the last page in the slides).
Regards, Naoya Yamaguchi
