If your basis set is not converged, it is clear that the eigenvalues
will change (go down in energy) when the basis set is increased.
The cutoff is defined by RKMAX, i.e. we take all plane waves such that
|(k+K)| < KMAX. In a bandstructure plot k varies slightly, but it can
then happen that at a certain k, the maximal K is too large and gets
reduced. Then a "jump" in the bandstructure will appear.
In this description, "k" is the vector of the k-point, "K" is a full
reciprocal lattice vector.
PS: Yes, RKMAX=7 is of course too small for a 3d (Cu) element, while
RKMAX=7 is much too large for H. It is a compromise.
On our website www.wien2k.at at reg.user and faq there are some
suggestions for a "minimal" RKMAX for each element. You look into your
structure and identify the atom which has the smallest RMT. This atom
determines RKMAX according to the list given on this site. However,
these values are kind of "minimal" RKMAX values (quite good for eg.
optimization of atomic positions (-min), but for highest precision they
should be increased by 1-3.
Just a simple example:
You do Cu and need RKMAX=9 for full convergence.
You do CuOH with RMT=1.8; 1.2 and 0.6 for the 3 elements.
RKMAX=3 is counting for H; gives "effective RKMAX=6 for O and 9 for Cu.
Hence RKMAX=3 is fully converged for CuOH (when using RMTs according to
the suggestion of setrmt).
Am 23.01.2021 um 04:49 schrieb 巴拉比:
Dear professor Jianxin Zhu,
Thank you so much for your quick reply.
Enlarging RKmax solves the problem! I found I have to set RKmax=9 which
is pretty large compared to default 7 to completely eliminate all
sawtooth artifacts in bands.
But I still want to know why a smaller RKmax would cause eigenvalues to
be discontinuous (in essence hamiltonian is discontinuous) at several k
points.
best regards
------------------ Original ------------------
*From:* "A Mailing list for WIEN2k users" <jx...@lanl.gov>;
*Date:* Fri, Jan 22, 2021 11:57 PM
*To:* "A Mailing list for WIEN2k users"<wien@zeus.theochem.tuwien.ac.at>;
*Subject:* Re: [Wien] [EXTERNAL] Bands are often not smooth when zoom in
What rkmax value are you using? Usually a larger rkmax helps make the
bands smooth.
Jianxin
Sent with BlackBerry Work
(www.blackberry.com)
*From: *Wien <wien-boun...@zeus.theochem.tuwien.ac.at
<mailto:wien-boun...@zeus.theochem.tuwien.ac.at>> on behalf of: 巴拉比
<bal...@qq.com <mailto:bal...@qq.com>>
*Date: *Friday, Jan 22, 2021, 8:15 AM
*To: *wien <wien@zeus.theochem.tuwien.ac.at
<mailto:wien@zeus.theochem.tuwien.ac.at>>
*Subject: *[EXTERNAL] [Wien] Bands are often not smooth when zoom in
Dear wien2k developers,
I found bands obtained from wien2k are often not smooth and have many
small sawtooth bumps especially when you zoom in. I can reproduce this
kind of band artifacts in version 19.2 and 17.1, on differnet computers
and link to different version of MKL. I'll take simple copper as an
example to show how I obtain the bands:
the structure file I use is as:
blebleble
F LATTICE,NONEQUIV.ATOMS: 1 225 Fm-3m
RELA
6.872726 6.872726 6.872726 90.000000 90.000000 90.000000
ATOM 1: X=0.00000000 Y=0.00000000 Z=0.00000000
MULT= 1 ISPLIT= 2
Cu1 NPT= 781 R0=0.00005000 RMT= 2.0000 Z: 29.0
LOCAL ROT MATRIX: 1.0000000 0.0000000 0.0000000
0.0000000 1.0000000 0.0000000
0.0000000 0.0000000 1.0000000
48 NUMBER OF SYMMETRY OPERATIONS
1 0 0 0.00000000
0 1 0 0.00000000
0 0 1 0.00000000
1
-1 0 0 0.00000000
0-1 0 0.00000000
0 0 1 0.00000000
2
-1 0 0 0.00000000
0 1 0 0.00000000
0 0-1 0.00000000
3
1 0 0 0.00000000
0-1 0 0.00000000
0 0-1 0.00000000
4
0 0 1 0.00000000
1 0 0 0.00000000
0 1 0 0.00000000
5
0 0 1 0.00000000
-1 0 0 0.00000000
0-1 0 0.00000000
6
0 0-1 0.00000000
-1 0 0 0.00000000
0 1 0 0.00000000
7
0 0-1 0.00000000
1 0 0 0.00000000
0-1 0 0.00000000
8
0 1 0 0.00000000
0 0 1 0.00000000
1 0 0 0.00000000
9
0-1 0 0.00000000
0 0 1 0.00000000
-1 0 0 0.00000000
10
0 1 0 0.00000000
0 0-1 0.00000000
-1 0 0 0.00000000
11
0-1 0 0.00000000
0 0-1 0.00000000
1 0 0 0.00000000
12
0 1 0 0.00000000
1 0 0 0.00000000
0 0-1 0.00000000
13
0-1 0 0.00000000
-1 0 0 0.00000000
0 0-1 0.00000000
14
0 1 0 0.00000000
-1 0 0 0.00000000
0 0 1 0.00000000
15
0-1 0 0.00000000
1 0 0 0.00000000
0 0 1 0.00000000
16
1 0 0 0.00000000
0 0 1 0.00000000
0-1 0 0.00000000
17
-1 0 0 0.00000000
0 0 1 0.00000000
0 1 0 0.00000000
18
-1 0 0 0.00000000
0 0-1 0.00000000
0-1 0 0.00000000
19
1 0 0 0.00000000
0 0-1 0.00000000
0 1 0 0.00000000
20
0 0 1 0.00000000
0 1 0 0.00000000
-1 0 0 0.00000000
21
0 0 1 0.00000000
0-1 0 0.00000000
1 0 0 0.00000000
22
0 0-1 0.00000000
0 1 0 0.00000000
1 0 0 0.00000000
23
0 0-1 0.00000000
0-1 0 0.00000000
-1 0 0 0.00000000
24
-1 0 0 0.00000000
0-1 0 0.00000000
0 0-1 0.00000000
25
1 0 0 0.00000000
0 1 0 0.00000000
0 0-1 0.00000000
26
1 0 0 0.00000000
0-1 0 0.00000000
0 0 1 0.00000000
27
-1 0 0 0.00000000
0 1 0 0.00000000
0 0 1 0.00000000
28
0 0-1 0.00000000
-1 0 0 0.00000000
0-1 0 0.00000000
29
0 0-1 0.00000000
1 0 0 0.00000000
0 1 0 0.00000000
30
0 0 1 0.00000000
1 0 0 0.00000000
0-1 0 0.00000000
31
0 0 1 0.00000000
-1 0 0 0.00000000
0 1 0 0.00000000
32
0-1 0 0.00000000
0 0-1 0.00000000
-1 0 0 0.00000000
33
0 1 0 0.00000000
0 0-1 0.00000000
1 0 0 0.00000000
34
0-1 0 0.00000000
0 0 1 0.00000000
1 0 0 0.00000000
35
0 1 0 0.00000000
0 0 1 0.00000000
-1 0 0 0.00000000
36
0-1 0 0.00000000
-1 0 0 0.00000000
0 0 1 0.00000000
37
0 1 0 0.00000000
1 0 0 0.00000000
0 0 1 0.00000000
38
0-1 0 0.00000000
1 0 0 0.00000000
0 0-1 0.00000000
39
0 1 0 0.00000000
-1 0 0 0.00000000
0 0-1 0.00000000
40
-1 0 0 0.00000000
0 0-1 0.00000000
0 1 0 0.00000000
41
1 0 0 0.00000000
0 0-1 0.00000000
0-1 0 0.00000000
42
1 0 0 0.00000000
0 0 1 0.00000000
0 1 0 0.00000000
43
-1 0 0 0.00000000
0 0 1 0.00000000
0-1 0 0.00000000
44
0 0-1 0.00000000
0-1 0 0.00000000
1 0 0 0.00000000
45
0 0-1 0.00000000
0 1 0 0.00000000
-1 0 0 0.00000000
46
0 0 1 0.00000000
0-1 0 0.00000000
-1 0 0 0.00000000
47
0 0 1 0.00000000
0 1 0 0.00000000
1 0 0 0.00000000
48
then I do step by step
1. init_lapw -b
2. run_lapw -in1new 3 # I use -in1new 3 to avoid the QTL-B value
warning message in the last few cycles of case.scf file
3. using xcrysden to generate a k path along GAMMA-W-X-L with 400 k points.
since the klist_band file is too long. I share it on
https://pastebin.com/ZaYuFMZE
4. x lapw1 -band
5. x lapw2 -band -qtl
finally, I use the data in case.qtl to plot bands
in [-10eV,10eV] range, it looks pretty normal as shared in below link
https://pasteboard.co/JKPC1tB.png
in [-2eV,-1eV] zoom in range, the look of bands becomes abnormal with
full of sawtooth bump as shared in link
https://pasteboard.co/JKPCB7Y.png
So I am wondering if this is a bug? Or what I have done wrong to have
this kind of bands(it appears in all my other bands calculation)? I
inspected the k path and found it is pretty fine. So I can not think of
a reason why the eigenvalue solutions would be discontinous at two
nearby k points.
best regards
_______________________________________________
Wien mailing list
Wien@zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:
http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
--
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-165300 FAX: +43-1-58801-165982
Email: bl...@theochem.tuwien.ac.at WIEN2k: http://www.wien2k.at
WWW: http://www.imc.tuwien.ac.at
-------------------------------------------------------------------------
_______________________________________________
Wien mailing list
Wien@zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:
http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html