注：

1） 分析成功案例，探索每个步骤在计算过程中的作用，如输入及输出

参考文献：

Mostofi, A.A., J.R. Yates, Y.-S. Lee, I. Souza, D. Vanderbilt, and N. Marzari. wannier90: A tool for obtaining maximally-localised Wannier functions. Computer Physics Communications, 2008, 178(9): 685-699.

一、实例1

1.脚本

INCAR

################## step1: a DFT groundstate calculation  ###################

cat > INCAR << EOF
ISMEAR =  0
SIGMA  =  0.05
GGA    = PE
EDIFF  = 1E-8
EOF

# start calculation

$PARA -n $NP -machinefile .nodelists.$$ $EXEC  > STDOUT
 
cp OUTCAR OUTCAR-01
cp STDOUT STDOUT-01

################# step 2: obtain DFT virtual orbitals #####################

cat > INCAR << EOF
ISMEAR =  0
SIGMA  =  0.05
GGA    = PE
EDIFF  = 1E-8

ALGO = Exact
NBANDS  = 64
LOPTICS = .TRUE.
NEDOS = 2000
# you might try
#LPEAD = .TRUE.

EOF

# start calculation

$PARA -n $NP -machinefile .nodelists.$$ $EXEC > STDOUT

cp WAVECAR WAVECAR.LOPTICS
cp WAVEDER WAVEDER.LOPTICS
cp OUTCAR OUTCAR-02
cp STDOUT STDOUT-02

############step 3:  GW0 + WANNIER90  ###############################

cat > INCAR << EOF
ISMEAR =  0
SIGMA  =  0.05
GGA    = PE
EDIFF  = 1E-8

ALGO = GW0
LSPECTRAL = .TRUE.
NOMEGA = 50
LRPA = .FALSE.

NBANDS = 64
##VASP2WANNIER90
LWANNIER90=.TRUE.

NELM = 4

EOF

# start calculation

$PARA -n $NP -machinefile .nodelists.$$ $EXEC > STDOUT

cp OUTCAR OUTCAR-03
cp STDOUT STDOUT-03
cp  WAVECAR WAVECAR-03
cp  WAVEDER WAVEDER-03

2. wannier90.win文件内容 

 
num_wann=24

num_bands=36

# for GW uncomment
exclude_bands 37-64

Begin Projections
Sc:l=0;l=1;l=2
H: l=0
End Projections

dis_froz_max=9
dis_num_iter=1000
guiding_centres=true

#运算过程中程序以笛卡尔坐标的形式自动补全wannier90.win文件所需的晶格矢量及坐标以及kpoints数据

3. 运行VASP计算耗时：

grep  CPU  OUT*
OUTCAR:                  Total CPU time used (sec):   157266.953
OUTCAR-01:                  Total CPU time used (sec):       21.754
OUTCAR-02:                  Total CPU time used (sec):      167.016

OUTCAR-03:                  Total CPU time used (sec):   157266.953

建立band文件夹，将计算得到的wannier90.*文件拷入band:

   12947 Jun 20 19:54 wannier90.wout
232172022 Jun 20 19:55 wannier90.mmn
22239462 Jun 20 19:56 wannier90.amn
  837540 Jun 20 19:56 wannier90.eig
   25526 Jun 20 19:56 wannier90.win

4. 问题摘录：

once i re-run W90 as:
>
> /work/aha/giovanne/ESPRESSO/espresso-4.0.5/W90/wannier90.x FeSe_up
>
> i get the following error:
>
> ---------------------------------------------------------------------------------------
> Found a mismatch in FeSe_up.eig
> Wanted band  : 1 found band  : 21
> Wanted kpoint: 2 found kpoint: 1

Gianluca,

 I have to admit I'm a bit confused by this error. From your input  
files I don't see anything wrong - and I repeated your calculation  
(reducing the number of planwaves and kpoints to speed things up) and  
didn't reproduce the problem.

Here are a few things to check:

After running wannier90 with the -pp option the nnkp file is created.  

Towards the end it should report
 begin exclude_bands
  60
...

So you have 80 bands in your calculation, but 60 are excluded from the  
calculation of the overlaps.

After running pw2wannier90 the head of the amn file should read
  Created on 24Feb2010 at 22:44:33
          20           8          20
 #bands #kpoints (8 in my case) and #projections

And in the eig file
    1    1    5.863971456939
    2    1    6.677551952837
    3    1    6.806925448330
where the first number is band (1-20) the second is kpoint, and the  
final one band energy.

My tests were done with the latest releases of wannier90 (1.2) and  
pwscf (4.1.2). But I'm not aware of any changes that might cause the  
problem you describe. (still worth upgrading if you use older versions).

One suggestion is to try to make a small test case that reproduces the  
problem. Then you can quickly try it out with other versions of the  
code (or send it to the developers in you get stuck - we appreciate  
small test cases!).

 Jonathan

--
Department of Materials, University of Oxford, Parks Road, Oxford, OX1  
3PH, UK
tel: +44 (0)1865 612797

5. 问题处理

在上述提示下，运行wannier90.x  -pp wannier90

 vi wannier90.nnkp得到下述信息

begin exclude_bands
 35
 30

....

 64
end exclude_bands

the head of the amn file:

File generated by VASP: unknown system
         44         405          24

#bands #kpoints (405 in my case) and #projections

And in the eig file，

where the first number is band (1-20) the second is kpoint, and the  
final one band energy.

          1           1      -43.732552498481
          2           1      -43.712725598234
          .........
         41           1       52.987595149617
         42           1       53.123760150869
         43           1       53.123760150869
         44           1       54.312774388660
          1           2      -43.726736534895
          2           2      -43.707850087633
          3           2      -22.988288527517
          4           2      -22.743182384502
          ........
         41           2       52.147090111358
         42           2       52.197417859885
         43           2       52.832642529808
         44           2       53.484528425242
          ......//////
         ///////////////////////////////////
          1         405      -43.679325941305
          2         405      -43.679325941305
          3         405      -22.904528925483
          4         405      -22.904528925483
          ......
         41         405       48.172494291025
         42         405       49.116479330742
         43         405       49.336596793144
         44         405       49.336596793144

 【nm_bands数量必须大于num_wann数量，从计算的64条能带中选取8条(num_bands)，对这8条轨道采用wannier函数投影（wannier函数是实空间的函数）；exclude_bands不是必须的；Nbands必须大于num_wann，否则会存在误投影情形】

【-pp: This optional flag tells the code to generate a list of the required overlaps and then exit.
This information is written to the file seedname.nnkp.】

修改wannier90.win

num_wann=24

num_bands=44   #从之前的36修改为现在的44,

# for GW uncomment
exclude_bands 35-64   #从之前的37修改为现在的35

运行wannier90.x wannier90，得到

“可见，相比HSE,GW可以给出更大的‘band gap’

” We
can do hybrid HSE-06 calculations with VASP to get bigger band gaps (and we
do this standard practice nowadays for important systems). If GW is a little
more expensive than HSE-06 I think we should manage.
”

对于极化系统遇到的问题：

ls  -lart wannier*

7338342       Jun 21 10:19       wannier90.dn.amn      
276360       Jun 21 10:19       wannier90.dn.eig      
1822582       Jun 21 10:19       wannier90.dn.mmn      
7338342       Jun 21 10:19       wannier90.up.amn      
276360       Jun 21 10:19       wannier90.up.eig      
1822582       Jun 21 10:19       wannier90.up.mmn      
 16237       Jun 21 10:19       wannier90.win        
 13343       Jun 21 10:19       wannier90.wout        

问题1：

I'm using vasp2wannier (updated to wannier90 2.0 version) currently . The vasp2wannier funciton successfully generate wannier90.amn wannier90.mmn wannier90.win files required by the following wannier procedure . However, for a spin polorized system , the wannier reported error -- Error: Problem  opening input file wannier90.spn when runnig postw90.x (postw90.x is the new feautre of wanneir90 2.0). I really want to know whether vasp2wannier can generate spn files , if so , any additional compilation tricks needed in addition to

查询：

.mmm

First line: a user comment, e.g., the date and time
Second line: 3 integers: num_bands, num_kpts, nntot
Then: num_kpts  nntot blocks of data:

 vi  wannier90.dn.mmn

File generated by VASP: unknown system
         24         245           8
   1    9    0    0    0

  -0.329224696146   -0.94147614834

问题2：

spinpolarized wannier90 calculations

http://www.democritos.it/pipermail/wannier/2012-March/000500.html

Dear all,Over the last few months I have been working on doped (Y, Gd, Sm, La, etc)and reduced ceria systems. MMy goal is to perform DFT calculations on thesecompounds and then extract the ionic forces and dipoles, which I will thenuse to fit an interionic potential model. To this end, I wanted to use theWCs to extract the dipoles. I have carried out this work using mainlyVASP/Wannier90. I also used CPMD (which can output the WCs) in some casesto check that the results are consistent. My motivation for usingVASP/Wannier90 is the implementation of the Hubbard (U) correction and theavailability of pseudopotentials for lanthanoid elements.The results I have obtained so far for yttrium-doped CeO2 andscandium-doped CeO2 have been consistent with each other using either DFTprogram. By this I mean that the forces and the dipole moments obtainedfrom the DFT codes show excellent agreement between them. HHowever, it iswhen I carry out simulations with elements that have f-electrons (Ce3+,Gd3+, Sm3+) that I start having some serious problems. IIn the case ofGd-doped ceria, when I visualize the wannier90_centres.xyz file from theWannier90 analysis of the data, I can see that the f-electrons that shouldhave been assigned to Gd, have instead been localized near the nucleus ofthe Ce atoms, in addition to the expected 8 valence electrons. I haveencountered similar problems with systems that include reduced cerium(Ce3+). In this case the single f-electrons from each Ce3+ are assigned toa single Ce atom (which according to the VASP output should not bereduced). These wannier centres are conspicuous among the others given thattheir spreads are typically higher (>4 Bohr^2) than those of all the otherWCs in the system (2.5 – 3 Bohr^2). In addition, the up and down spincomponents from the wannier90 calculation are the same, which, I think, isunexpected. Finally, a comparison with the results from CPMD is notavailable here because CPMD does not have the DFT+U method implemented. I am quite confident that the VASP calculations I performed are fine.Indeed, I have plotted the electronic density of states for saidcalculations from VASP and I found them to be in accordance with what wouldbe expected, i.e. in the case of Gd-doped ceria, the Gd f-states are deepin the valence band. The calculation also showed charges consistent with Cebeing 4+ and Gd being 3+ and the magnetization of the Gd ions is 7, whichis consistent with the literature.My question is if there is a specific set of keywords that I need to usefor these f-electron systems. To this end, I attach my wannier90.win,wannier90.wout, INCAR file from VASP. A plot of the EDOS for Gd-doped ceriais also available upon request. I would greatly appreciate some help withthis issue.Mario BurbanoINCARSYSTEM = CeO2Start parameter for this Run:NWRITE = 2LPETIM=F write-flag & timerISTART = 1 job : 0-new 1-cont 2-samecutICHARG = 1Electronic Relaxation 1PREC = normalENMAX = 500.00 eV Use no more than 400 for hybNELM = 150 Or use 150 depending on systemINIWAV = 1 random initial wavefunctionsEDIFF = 1E-05Ionic RelaxationEDIFFG = -0.01 stopping-criterion for IOMNSW = 0 number of steps for IOMNBLOCK = 1; KBLOCK = 40 inner block and outer blockIBRION = 2 ionic relax: 0-MD 1-quasi-New 2-CG!POTIM = 0.2!TEBEG = 1200ISIF = 2 stress and relaxationIWAVPR = 1 prediction: 0-non 1-charg 2-wave 3-combISYM = 0 0-nonsym 1-usesymLCORR = F Harris-correction to forcesISMEAR = 0LMAXMIX = 6!NPAR = 4ALGO = NormalLREAL = AROPT = 1E-04 1E-04 1E-04ISPIN = 2!MAGMOM = 63*0.0 2*7.0 30*0.0LORBIT = 12!NBANDS = 576!LWANNIER90 = .TRUE.LWANNIER90_RUN = .TRUE.LWRITE_MMN_AMN   = .TRUE.LASPH = .TRUE.LORBIT  =  12LDAU    =  .TRUE.   DFT+U ParametersLDAUL   = -1    3    3LDAUU   =  0    5   7LDAUJ   =  0    0   0LMAXMIX =  6LASPH = .TRUE.wannier90.winnum_wann = 394exclude_bands : 395-516translate_home_cell = Twrite_xyz = Tnum_iter = 10000!restart = defaultlength_unit = bohrconv_tol = 1E-05conv_window = 2use_bloch_phases = T

问题3：

Have you specified spin_component='up' (or spin_component='down' in the inputpp namelist)?

Technically, the spin polarization is implemented by doubling the number of k-points,
hence the pw2wannier code needs to know if only half of the k-points needs to be processed
at a time (this is th case in spin-polarized calculations).

If the problem persist or it is present in 4.1.2 and not in 4.1, please attach some sample
input file which generates such kind of problem.


Cheers from Miramare,

GS

回答：

Dear Gabriele,
thank you for your answer.
I could solve the problem. It works, if one uses nelup > neldw in the
pw.x calculation.

http://www.democritos.it/pipermail/wannier/2007-May/000068.html

1.

ear Brad

I think the problem is in the "projections" block of the input file, not
the energy windows. As you say, a 12 atom unit cell of silicon will need 24
maximally localised Wannier functions (MLWFs) to span the valence manifold
and this is the number of projections you should define in the projections
block.

What we've found is that the generation of MLWFs for the valence manifold
is very robust to the initial guess, so even specifying randomly centred
s-orbital projections is good enough. This can be done by setting

Begin Projections
random
End Projections

in the input file.

Cheers

Arash

2.

Thanks for the response Arash.  Setting the projections to random (they were
set to sp3) did indeed fix that issue.  I ran the wannier code and generated
the .nnkp file.  However, in the nnkp file it lists my excluded bands as
being 25-32 (which I had specified in the .win input file; perhaps
unnecessary because my window excludes them anyway) but also it lists band
#8 as being excluded, and I did not want or specify that.  Is there a reason
for this?  I'm worried that it may be causing me an error when I run the
Wannier code again because I get an error of "param_read: mismatch in
silicon.eig".  Inspecting the silicon.eig file it looks like it should be
okay, unless the code thinks that there are 23 bands (having excluded band
8), as there are 24 bands (for my 64 points) listed in the silicon.eig file.

I'm also interested in the bottom of the conduction band.  Do you have a
feel for how robust this is to the initial guess for the projections?  And
is it possible for me to specify sp3 projections without having 48 wannier
bands?  I'd like to use fewer bands if possible because it seems that I'm
running into memory allocation issues that prevent me from starting with a
denser k-grid. Thanks again for the help.

Brad

3.

------=_Part_149621_19952660.1180226097234
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

Aha, I see now that it simply first lists the number of excluded bands
before listing the specific numbers.  But I'm still wondering about why I'm
getting the error in the .eig file.  Thanks!

------=_Part_149621_19952660.1180226097234
Content-Type: text/html; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

Aha, I see now that it simply first lists the number of excluded bands before listing the specific numbers.&nbsp; But I&#39;m still wondering about why I&#39;m getting the error in the .eig file.&nbsp; Thanks!<br>

------=_Part_149621_19952660.1180226097234--

Omitting the 'exclude bands' portion in my input file apparently fixes my
issue with the .eig file.  I no longer get the error.  I don't know if what
I described before is the reason for that or not, but it works now.

Best,
Brad

问题4：

http://www.democritos.it/pipermail/wannier/2013-December.txt

On 4 Dec 2013, at 08:46, ??? <hslk77 at hotmail.com> wrote:

> Dear all,
>  
> I'm new with wannier90. Now I'm trying to get the band structure from a spin polarized GW0 calculation by wannier90.
> By setting LWANNIER90=T, I got the wannier90 files as: wannier90.win, wannier90.wout, wannier90.up.mmn, wannier90.dn.mmn, wannier90.up.eig, wannier90.dn.eig. Now I want to ask how to run wannier90 to get the band structure? If I simply type: "wannier90.x wannier90.win", I get the error message as: "Error: Problem opening input file wannier90.mmn". Could anyone kindly tell me how to do this?

Wannier expects to find a set of files of the form <seedname>.win <seedname>.amn <seedname>.mmn <seedname>.eig
So you need to rename the files slightly to conform to this pattern. One option is

cp wannier90.win wannier90.up.win
wannier90.x wannier90.up

You will then get output files wannier90.up.wout etc.

问题：

你好：是这样的，一般系统中若有未配对电子，会提示你选择自旋极化的，所以对于你的问题，有可能一个缺陷有未配对电子，两个缺陷就没有了。而且你就算是选择自旋极化计算得到的态密度，仍然可以选择total而不是把上下自旋分开画那样的态密度图，这样应该可以与未选择自旋极化的比较两者态密度的变化了吧？

aylayl08(站内联系TA)

自旋极化貌似只是讨论磁性的时候用到，讨论带隙不涉及自旋不自旋的事，也就是说其不影响带隙。当然如果你体系上自旋下自旋电子数相等没有必要分了。
因此对于number of up-spin electrons is not equal to the
                             number of down-spins but spin_polarized=false
                             - consider setting spin_polarized=true.这个警告，不用理睬。有时候体系明明是上下自旋电子数相同，有时还会提示这个。看带隙直接看能带图就可以了，价带顶到导带底的距离，结果文件里也有数据。

如果考虑之，也不妨碍，我曾看过一点讨论转述如下：关于自旋极化的能带，一般是画出两幅图：majority spin和minority spin，经典的说，分别代表自旋向上和自旋向下的轨道所组成的能带结构。注意它们在费米能级处的差异。如果费米能级与majority spin 的能带图相交而处于minority spin 的能隙中，则此体系具有明显的自旋极化现象。而该体系也可称之为半金属。因为majority spin与费米能级相交的能带主要有杂质原子轨道组成，所以也可以此为出发点讨论杂质的磁性特征。

   最好可以参考一下相关文献尤其是博士论文会有详尽的解释。;);)

参数再解释：

When LPEAD=.TRUE., VASP will compute using the aforementioned finite difference scheme. The order of the finite difference approximation can be specified by setting the IPEAD-tag.

These tags may be used in combination with LOPTICS=.TRUE. (Sec. 6.72.1) and LEPSILON=.TRUE. (Sec. 6.72.4).

LRPA: local field effects on the Hartree level (RPA)

LRPA= .TRUE. |.FALSE.

Default: LRPA= .FALSE.

Usually local field effect are included on the Hartree level only (LRPA=.TRUE.). This means that cell periodic microscopic changes of the local potential related to the change of the Hartree potential are included.  If LRPA=.FALSE., however, changes of the Hartree potential and the exchange correlation potential are included. This usually increases the dielectric constants. The final values for the dielectric matrix can be found in the OUTCAR file after the lines.   MACROSCOPIC STATIC DIELECTRIC TENSOR (including local field effects in RPA (Hartree)) For LRPA=.FALSE. the dielectric matrix is written after the lines: MACROSCOPIC STATIC DIELECTRIC TENSOR (including local field effects in DFT) The dielectric constants without local field effects is always determined (irregardless of LRPA). The piezoelectric tensors and the Born effective charges as well as the ionic contributions the to dielectric tensor are only calculated for LRPA=.FALSE.