Re: [Wien] different MLD for bcc structure for magnetic equivalent directions M001, M010 and M100

2017-12-01 Thread Peter Blaha 
I'm pretty sure the problem is connected with the tetrahedron method, 
but also the presence of van Hove singularities and a k-mesh, which is 
not converged.


a) As mentioned before, for all scf quantitites there is no problem if 
one uses TEMP, but already the scf cycle differs when using TETRA.


b) At the end, this is not a big surprise. Checking case.kgen shows that 
although the number of IBZ k-points is the same for both cases, the 
number of tetrahedra (and their volume) is not. So clearly, the BZ of 
these 2 cases are decomposed into a different set of tetrahedra, even 
when you have the same number of k-points.


c) Naturally, from this it follows that all integrals over the BZ are 
"different" (density, or the DOS and of course also joint), except when 
the results are converged with respect to k-mesh.


d) I verified quite extensively that the matrix elements form optic obey 
the expected symmetry.


e) I don't think that one can say that the BZ integration with 
decomposition 001 or 100 is more or less accurate, therefore the only 
thing which can help is to go to enourmous k-meshes until the quantity 
you are looking for is converged (and then hopefully you obtain 
identical results). After all, a 30x30x30 mesh is still a "lousy" mesh 
for metallic optics (refer to Al).


f) I've done a 100x100x100 unshifted mesh. Also I've looked at the plot 
without broadening. Clearly, your broad peaks at 5 and 6 eV are in fact 
very sharp resonances, coming from some stupid van Hove singularities. 
Such things are always very difficult to converge. In any case, my 1M 
calculation has now the same sign (and also close magnitudes) for the 
peaks at 5 and 6 eV (in contrast to the 30x30x30 calculation). In fact 
all oscillations below 1 eV are now also correctly resolved and 
"identical", only below 0.1 eV the magnitudes are still not identical 
(but the sign is the same).


g) In order to do that, you probably don't need to do scf with these 
meshes. Just do a scf cycle with eg. TEMP 0.001 and a good k-mesh.
The switch to TETRA 101.1 in case.in2c. Run kgen with better and better 
meshes, and only lapw1/lapwso/lapw2/optic/joint in the proper sequence


h) Maybe ??? an alternative is to use a doubled P cell. But I'm not sure 
if the results converge quicker with the k-mesh. Agreement between 001 
and 100 directions does NOT necessarely mean that the results are 
converged with respect to k-mesh !!



On 11/28/2017 12:36 PM, Jaroslav Hamrle wrote:

Dear Laurence,

thank you for your detailed answer.

I have tried all your suggestions,
- I changed case.in0 with increased oversampling by factor two (new
parameters LUSE 26 and IFFTfactor 4)
 start of case.in0 ---
TOT  XC_LDA (XC_PBE,XC_PBESOL,XC_WC,XC_MBJ,XC_REVTPSSS)
NR2V  IFFT  26  (R2V)
   24   24   244.00  1min IFFT-parameters, enhancement factor,
iprint
 end of case.in0 ---

- I also tried to impose strong convergence criteria to be -cc
0.0001 -ec 0.0001

However, in both cases, the ghost MLD remains practically identical as
when using my default values (default Fe + convergence -cc 0.1 -ec
0.01)

Also, final :PUP 'Current' parameters remained practically the same for
all the calculations (by about 2 digits), being like (for M001)
PW CHANGE HKL  Current   ChangeResidue
:PUP001:  000  2.10719649E-02  5.090E-10 -7.530E-09
:PUP002:  0   -1   -1  3.67084665E-04 -7.004E-10 -3.572E-10
:PUP003:  1   -10  1.82124988E-04 -3.669E-10 -2.211E-10
:PUP004:  00   -2 -1.87471938E-03  6.192E-11  7.254E-10
:PUP005:  0   -20 -3.75090680E-03  1.125E-10  1.378E-09
:PUP006:  1   -1   -2 -3.46372731E-03  1.026E-10  1.378E-09
:PUP007:  1   -2   -1 -6.92804324E-03  2.418E-10  2.776E-09
:PUP008:  0   -2   -2 -7.14014083E-04  8.677E-11  2.032E-10
:PUP009:  2   -20 -3.57036516E-04  4.298E-11  1.121E-10
:PUP010:  0   -1   -3  2.62159615E-04  9.199E-11 -1.588E-10
:PUP011:  0   -3   -1  2.62289930E-04  9.844E-11 -1.555E-10
:PUP012:  1   -30  2.62219244E-04  9.133E-11 -1.551E-10
Unfortunately, all reflections seems to be allowed for bcc (H+K+L is
even), forbidden reflections of bcc are (H+K+L=odd), so I can not see
how they get close to zero ;-)) But the idea is excellent.

Thank you again and with my best regards

Jaroslav

On 27/11/17 15:27, Laurence Marks wrote:

Let me clarify slightly my comment about symmetry -- as I realized the
explanation (I think) and can also suggest something that might help.

First, concerning symmetry the explanation is I believe simple. If the
problem has a real symmetry operation such as inversion which is being
removed, then the Jacobian at the solution has zero's for charge
disturbances that break this symmetry. Because of this noise due to
numerical accuracy has a large effect, and almost certainly one has to
tighten the convergence criteria particularly -cc. You can monitor
this by

Re: [Wien] different MLD for bcc structure for magnetic equivalent directions M001, M010 and M100

2017-11-28 Thread Laurence Marks 
No "0001" at the end (Google autocorrect strikes again).

_
Professor Laurence Marks
"Research is to see what everybody else has seen, and to think what nobody
else has thought", Albert Szent-Gyorgi
www.numis.northwestern.edu

On Nov 28, 2017 6:35 AM, "Laurence Marks"  wrote:

> A better response after more coffee.
>
> You might have found a simple (& therefore useful) case that highlights a
> symmetry issue that I think is present, but have never been able to pin
> down. The key now is to try and find some way to eliminate it, since it
> could be in too many places (e.g. lapw[0-2], lapwso, sumpara, mixer). I
> suggest sticking with LDA since this should eliminate most VXC issues.
> Thoughts.
>
> a) Please check case.struct and ensure that all the positions are 0.0 or
> 0.5 and similarly for translations, with no "000" at the end.
> b) Try using "x kgen -fbz"
> c) Try using TEMPS rather than TETRA.
> d) Check the compiler options, and use -O1 -mp1 and perhaps add (at the
> end) -noftz.
> e) Test compiling lapw1 with -DoldScalapack. I may have the name/case
> wrong, do a "grep -e ifdef *.F" in SRC_lapw1. (This changes with version.)
> f) Try regressing your mkl version if you can.
> g) Wait for Peter/Fabien to add ideas (or test themselves).
>
> _
> Professor Laurence Marks
> "Research is to see what everybody else has seen, and to think what nobody
> else has thought", Albert Szent-Gyorgi
> www.numis.northwestern.edu
>
> On Nov 28, 2017 5:36 AM, "Jaroslav Hamrle" 
> wrote:
>
>> Dear Laurence,
>>
>> thank you for your detailed answer.
>>
>> I have tried all your suggestions,
>> - I changed case.in0 with increased oversampling by factor two (new
>> parameters LUSE 26 and IFFTfactor 4)
>>  start of case.in0 ---
>> TOT  XC_LDA (XC_PBE,XC_PBESOL,XC_WC,XC_MBJ,XC_REVTPSSS)
>> NR2V  IFFT  26  (R2V)
>> 24   24   244.00  1min IFFT-parameters, enhancement factor,
>> iprint
>>  end of case.in0 ---
>>
>> - I also tried to impose strong convergence criteria to be -cc
>> 0.0001 -ec 0.0001
>>
>> However, in both cases, the ghost MLD remains practically identical as
>> when using my default values (default Fe + convergence -cc 0.1 -ec
>> 0.01)
>>
>> Also, final :PUP 'Current' parameters remained practically the same for
>> all the calculations (by about 2 digits), being like (for M001)
>> PW CHANGE HKL  Current   ChangeResidue
>> :PUP001:  000  2.10719649E-02  5.090E-10 -7.530E-09
>> :PUP002:  0   -1   -1  3.67084665E-04 -7.004E-10 -3.572E-10
>> :PUP003:  1   -10  1.82124988E-04 -3.669E-10 -2.211E-10
>> :PUP004:  00   -2 -1.87471938E-03  6.192E-11  7.254E-10
>> :PUP005:  0   -20 -3.75090680E-03  1.125E-10  1.378E-09
>> :PUP006:  1   -1   -2 -3.46372731E-03  1.026E-10  1.378E-09
>> :PUP007:  1   -2   -1 -6.92804324E-03  2.418E-10  2.776E-09
>> :PUP008:  0   -2   -2 -7.14014083E-04  8.677E-11  2.032E-10
>> :PUP009:  2   -20 -3.57036516E-04  4.298E-11  1.121E-10
>> :PUP010:  0   -1   -3  2.62159615E-04  9.199E-11 -1.588E-10
>> :PUP011:  0   -3   -1  2.62289930E-04  9.844E-11 -1.555E-10
>> :PUP012:  1   -30  2.62219244E-04  9.133E-11 -1.551E-10
>> Unfortunately, all reflections seems to be allowed for bcc (H+K+L is
>> even), forbidden reflections of bcc are (H+K+L=odd), so I can not see
>> how they get close to zero ;-)) But the idea is excellent.
>>
>> Thank you again and with my best regards
>>
>> Jaroslav
>>
>> On 27/11/17 15:27, Laurence Marks wrote:
>> > Let me clarify slightly my comment about symmetry -- as I realized the
>> > explanation (I think) and can also suggest something that might help.
>> >
>> > First, concerning symmetry the explanation is I believe simple. If the
>> > problem has a real symmetry operation such as inversion which is being
>> > removed, then the Jacobian at the solution has zero's for charge
>> > disturbances that break this symmetry. Because of this noise due to
>> > numerical accuracy has a large effect, and almost certainly one has to
>> > tighten the convergence criteria particularly -cc. You can monitor
>> > this by looking at the :PUPXXX values in case.scfm and look how well
>> > the forbidden reflections have converged to zero.
>> >
>> > Second, do not be surprised about numerical issues. While the
>> > calculations are done in double precision, there are many large sums
>> > and in some cases double sums, and also numerical
>> > integrations/differentiation. Any large sum or numerical
>> > integration/differentiation in general reduces the numerical accuracy.
>> > Hence even though double precision has an accuracy of 1D-15 the sum
>> > may only be accurate to 1D-10 or even 1D-7. Also, the Intel ifort
>> > compiler will reduce the numerical accuracy for speed if one is not
>> > careful.
>> >
>> > One thing that may help is to increase the oversampling in

Re: [Wien] different MLD for bcc structure for magnetic equivalent directions M001, M010 and M100

2017-11-28 Thread Peter Blaha 

I looked into the two directories which you provided in a previous mail.

Why do you have 16 and 8 symm.ops in the corresponding struct files ??? 
 Definitely you should also have in the 100 case 16 sym.ops.
I can therefore also see a vastly different number of k-points for the 2 
scf files.


From you first email, I thought the problem is already in lapwso (or 
the scf cycle), since you mentioned that already eigenvalues differ.


Personally, I found it already quite difficult to converge the 100 and 
001 case to the same total energy using TETRA (there is no problem using 
eg. TEMP 0.001). Depending on the k-mesh and the fermi-method I get 
small oszillations (at low :DIS) on the mycroRy level and I would 
probably have to go to larger meshes .


This problem manifests itself in
grep :NEC01 case.scf   (normalization not identical or oszillates ...) and
grep 'CHA ' case.scf

My two tests with 30x30x30 unshifted meshes and TETRA 101., RKMAX=7 gives:
:CHA  : TOTAL VALENCE CHARGE INSIDE UNIT CELL =   8.072612
:CHA  : TOTAL VALENCE CHARGE INSIDE UNIT CELL =   5.927386
 13.98

:CHA  : TOTAL VALENCE CHARGE INSIDE UNIT CELL =   8.072635
:CHA  : TOTAL VALENCE CHARGE INSIDE UNIT CELL =   5.927359
 13.94
Obviously it should always be exactly 14.0 (and this is fulfilled with 
TEMP).
Using TEMP 0.0001 and tight convergence it is possible to get all 
eigenvalues of a 100 and 001 calculations identical to 0.01 mycroRy (not 
milliRy !!).
Of course, you have to be careful with this comparison, sind the same 
k-points will NOT have the same eigenvalues, but symmetry-equivalent k 
have: For instance for my test mesh, the 4th and 65th k-points are 
equivalent, ...:


E(001)(k=  4 3 3 021) =
E(100)(k= 652118 321)

Conclusion:
i) If k-space integration works, there is no problem whatsoever in the 
scf cycle with/without SO, different symmetry, .
ii) TETRA is dangerous in these comparisons, since the tetrahedra are 
built differently for different reciplocal lattices and in particular if 
there are subtle crossings at EF, the (non-)linear extrapolation may 
lead to some artefacts. They are very small, but since SO for Fe is a 
small effect, ...




Later on, you focussed more on the optics (since this is your primary 
interest) and claim, it has nothing to do with SO ???, only with 
symmetry breaking.  Maybe you are right, it is nothing than the 
integration in joint using the weights from lapw2, which makes the 
Problems. At least manual inspection of some momentum matrix elements, 
which should be identical (or permutated) for equivalent k-pionts 
indicates that they obey symmetry. 
I checked your case.in2c files, and in both you have TETRA 0.000 as 
Fermi method. As mentioned in the UG for optic one should NOT use the 
Blöchl-method (with non-linear corrections), but use  TETRA  100.1

and the original tetrahedron method.
Unfortunately, this does not fix the problems. Together with your 
reports that with a larger k-mesh the problems get smaller, it points to 
a problem in k-space integration in joint.


--

I'll continue to look into that problem.


On 11/27/2017 03:07 PM, Jaroslav Hamrle wrote:

Dear all,

thank you for your comments:

1)


Did you use a Gamma centered k mesh (and enough k points)


I have checked that the same inequality in MLD, appears both when
k-points are shifted or not shifted. So, influence of shift of k-points
can be ruled out.

I have 30x30x30 k-points, which should be enough. When using 46x46x46
k-mesh, the MLD inequality is reduced by about factor two, but still
present. Even when using very fine k-mesh (90x90x90) for optical
calculation, the inequality persists.


2)


In some cells  shifting the k-point origin with MSR1a leads to
slightly unbalanced forces which are hard to converge to the "right"
symmetric result. If the forces are slightly off, this is an
indication that the density is also slightly off. Exactly why this
occurs I do not know, I suspect very soft modes associated with
numerical errors in finite arithmetic.


Well, for me it is hard to believe, that the problem can originates from
numerical error. The MLD inequality creates sort-of ghost peaks in MLD
spectra which for bcc Fe are very stable at positions at 4.8 and 6 eV
for different calculation details (as different k-mesh, presence/absence
of spin-orbit, shifted/non-shifted k-mesh). Under all those changes in
the calculations, the position of ghost peaks remains very stable, just
their amplitudes varies. Also, if problem would be just numerical one,
why ghost peaks are not present in simple cubic or fcc calculations?


3)



It is instructive to repeat the calculation without SO and see how big
the difference between sig_xx and

Re: [Wien] different MLD for bcc structure for magnetic equivalent directions M001, M010 and M100

2017-11-28 Thread Laurence Marks 
A better response after more coffee.

You might have found a simple (& therefore useful) case that highlights a
symmetry issue that I think is present, but have never been able to pin
down. The key now is to try and find some way to eliminate it, since it
could be in too many places (e.g. lapw[0-2], lapwso, sumpara, mixer). I
suggest sticking with LDA since this should eliminate most VXC issues.
Thoughts.

a) Please check case.struct and ensure that all the positions are 0.0 or
0.5 and similarly for translations, with no "000" at the end.
b) Try using "x kgen -fbz"
c) Try using TEMPS rather than TETRA.
d) Check the compiler options, and use -O1 -mp1 and perhaps add (at the
end) -noftz.
e) Test compiling lapw1 with -DoldScalapack. I may have the name/case
wrong, do a "grep -e ifdef *.F" in SRC_lapw1. (This changes with version.)
f) Try regressing your mkl version if you can.
g) Wait for Peter/Fabien to add ideas (or test themselves).

_
Professor Laurence Marks
"Research is to see what everybody else has seen, and to think what nobody
else has thought", Albert Szent-Gyorgi
www.numis.northwestern.edu

On Nov 28, 2017 5:36 AM, "Jaroslav Hamrle" 
wrote:

> Dear Laurence,
>
> thank you for your detailed answer.
>
> I have tried all your suggestions,
> - I changed case.in0 with increased oversampling by factor two (new
> parameters LUSE 26 and IFFTfactor 4)
>  start of case.in0 ---
> TOT  XC_LDA (XC_PBE,XC_PBESOL,XC_WC,XC_MBJ,XC_REVTPSSS)
> NR2V  IFFT  26  (R2V)
> 24   24   244.00  1min IFFT-parameters, enhancement factor,
> iprint
>  end of case.in0 ---
>
> - I also tried to impose strong convergence criteria to be -cc
> 0.0001 -ec 0.0001
>
> However, in both cases, the ghost MLD remains practically identical as
> when using my default values (default Fe + convergence -cc 0.1 -ec
> 0.01)
>
> Also, final :PUP 'Current' parameters remained practically the same for
> all the calculations (by about 2 digits), being like (for M001)
> PW CHANGE HKL  Current   ChangeResidue
> :PUP001:  000  2.10719649E-02  5.090E-10 -7.530E-09
> :PUP002:  0   -1   -1  3.67084665E-04 -7.004E-10 -3.572E-10
> :PUP003:  1   -10  1.82124988E-04 -3.669E-10 -2.211E-10
> :PUP004:  00   -2 -1.87471938E-03  6.192E-11  7.254E-10
> :PUP005:  0   -20 -3.75090680E-03  1.125E-10  1.378E-09
> :PUP006:  1   -1   -2 -3.46372731E-03  1.026E-10  1.378E-09
> :PUP007:  1   -2   -1 -6.92804324E-03  2.418E-10  2.776E-09
> :PUP008:  0   -2   -2 -7.14014083E-04  8.677E-11  2.032E-10
> :PUP009:  2   -20 -3.57036516E-04  4.298E-11  1.121E-10
> :PUP010:  0   -1   -3  2.62159615E-04  9.199E-11 -1.588E-10
> :PUP011:  0   -3   -1  2.62289930E-04  9.844E-11 -1.555E-10
> :PUP012:  1   -30  2.62219244E-04  9.133E-11 -1.551E-10
> Unfortunately, all reflections seems to be allowed for bcc (H+K+L is
> even), forbidden reflections of bcc are (H+K+L=odd), so I can not see
> how they get close to zero ;-)) But the idea is excellent.
>
> Thank you again and with my best regards
>
> Jaroslav
>
> On 27/11/17 15:27, Laurence Marks wrote:
> > Let me clarify slightly my comment about symmetry -- as I realized the
> > explanation (I think) and can also suggest something that might help.
> >
> > First, concerning symmetry the explanation is I believe simple. If the
> > problem has a real symmetry operation such as inversion which is being
> > removed, then the Jacobian at the solution has zero's for charge
> > disturbances that break this symmetry. Because of this noise due to
> > numerical accuracy has a large effect, and almost certainly one has to
> > tighten the convergence criteria particularly -cc. You can monitor
> > this by looking at the :PUPXXX values in case.scfm and look how well
> > the forbidden reflections have converged to zero.
> >
> > Second, do not be surprised about numerical issues. While the
> > calculations are done in double precision, there are many large sums
> > and in some cases double sums, and also numerical
> > integrations/differentiation. Any large sum or numerical
> > integration/differentiation in general reduces the numerical accuracy.
> > Hence even though double precision has an accuracy of 1D-15 the sum
> > may only be accurate to 1D-10 or even 1D-7. Also, the Intel ifort
> > compiler will reduce the numerical accuracy for speed if one is not
> > careful.
> >
> > One thing that may help is to increase the oversampling in case.in0
> > for VXC, both that of the PW's and of the CLMs. A standard test is to
> > use LDA and see if the problem goes away, since oversampling is much
> > less relevant for this.
> >
> > Of course your problem may have nothing to do with any of this
>
> ___
> Wien mailing list
> Wien@zeus.theochem.tuwien.ac.at
> https://urldefense.proofpoint.com/v2/url?u=http-3A__zeus.
>

Re: [Wien] different MLD for bcc structure for magnetic equivalent directions M001, M010 and M100

2017-11-28 Thread Laurence Marks 
Ignore my last email, I had the wrong selection rules -- too early, not
enough coffee.

_
Professor Laurence Marks
"Research is to see what everybody else has seen, and to think what nobody
else has thought", Albert Szent-Gyorgi
www.numis.northwestern.edu

On Nov 28, 2017 5:58 AM, "Laurence Marks"  wrote:

> Interesting. I assume that this is with -so. Does the sum :PUP+:PDN obey
> the fcc selection rules? (You can also look at the PW in case.clmsum &
> case.valup/dn.)
>
> _
> Professor Laurence Marks
> "Research is to see what everybody else has seen, and to think what nobody
> else has thought", Albert Szent-Gyorgi
> www.numis.northwestern.edu
>
> On Nov 28, 2017 5:36 AM, "Jaroslav Hamrle" 
> wrote:
>
>> Dear Laurence,
>>
>> thank you for your detailed answer.
>>
>> I have tried all your suggestions,
>> - I changed case.in0 with increased oversampling by factor two (new
>> parameters LUSE 26 and IFFTfactor 4)
>>  start of case.in0 ---
>> TOT  XC_LDA (XC_PBE,XC_PBESOL,XC_WC,XC_MBJ,XC_REVTPSSS)
>> NR2V  IFFT  26  (R2V)
>> 24   24   244.00  1min IFFT-parameters, enhancement factor,
>> iprint
>>  end of case.in0 ---
>>
>> - I also tried to impose strong convergence criteria to be -cc
>> 0.0001 -ec 0.0001
>>
>> However, in both cases, the ghost MLD remains practically identical as
>> when using my default values (default Fe + convergence -cc 0.1 -ec
>> 0.01)
>>
>> Also, final :PUP 'Current' parameters remained practically the same for
>> all the calculations (by about 2 digits), being like (for M001)
>> PW CHANGE HKL  Current   ChangeResidue
>> :PUP001:  000  2.10719649E-02  5.090E-10 -7.530E-09
>> :PUP002:  0   -1   -1  3.67084665E-04 -7.004E-10 -3.572E-10
>> :PUP003:  1   -10  1.82124988E-04 -3.669E-10 -2.211E-10
>> :PUP004:  00   -2 -1.87471938E-03  6.192E-11  7.254E-10
>> :PUP005:  0   -20 -3.75090680E-03  1.125E-10  1.378E-09
>> :PUP006:  1   -1   -2 -3.46372731E-03  1.026E-10  1.378E-09
>> :PUP007:  1   -2   -1 -6.92804324E-03  2.418E-10  2.776E-09
>> :PUP008:  0   -2   -2 -7.14014083E-04  8.677E-11  2.032E-10
>> :PUP009:  2   -20 -3.57036516E-04  4.298E-11  1.121E-10
>> :PUP010:  0   -1   -3  2.62159615E-04  9.199E-11 -1.588E-10
>> :PUP011:  0   -3   -1  2.62289930E-04  9.844E-11 -1.555E-10
>> :PUP012:  1   -30  2.62219244E-04  9.133E-11 -1.551E-10
>> Unfortunately, all reflections seems to be allowed for bcc (H+K+L is
>> even), forbidden reflections of bcc are (H+K+L=odd), so I can not see
>> how they get close to zero ;-)) But the idea is excellent.
>>
>> Thank you again and with my best regards
>>
>> Jaroslav
>>
>> On 27/11/17 15:27, Laurence Marks wrote:
>> > Let me clarify slightly my comment about symmetry -- as I realized the
>> > explanation (I think) and can also suggest something that might help.
>> >
>> > First, concerning symmetry the explanation is I believe simple. If the
>> > problem has a real symmetry operation such as inversion which is being
>> > removed, then the Jacobian at the solution has zero's for charge
>> > disturbances that break this symmetry. Because of this noise due to
>> > numerical accuracy has a large effect, and almost certainly one has to
>> > tighten the convergence criteria particularly -cc. You can monitor
>> > this by looking at the :PUPXXX values in case.scfm and look how well
>> > the forbidden reflections have converged to zero.
>> >
>> > Second, do not be surprised about numerical issues. While the
>> > calculations are done in double precision, there are many large sums
>> > and in some cases double sums, and also numerical
>> > integrations/differentiation. Any large sum or numerical
>> > integration/differentiation in general reduces the numerical accuracy.
>> > Hence even though double precision has an accuracy of 1D-15 the sum
>> > may only be accurate to 1D-10 or even 1D-7. Also, the Intel ifort
>> > compiler will reduce the numerical accuracy for speed if one is not
>> > careful.
>> >
>> > One thing that may help is to increase the oversampling in case.in0
>> > for VXC, both that of the PW's and of the CLMs. A standard test is to
>> > use LDA and see if the problem goes away, since oversampling is much
>> > less relevant for this.
>> >
>> > Of course your problem may have nothing to do with any of this
>>
>> ___
>> Wien mailing list
>> Wien@zeus.theochem.tuwien.ac.at
>> https://urldefense.proofpoint.com/v2/url?u=http-3A__zeus.the
>> ochem.tuwien.ac.at_mailman_listinfo_wien=DwIGaQ=yHlS04Hh
>> Braes5BQ9ueu5zKhE7rtNXt_d012z2PA6ws=U_T4PL6jwANfAy4r
>> nxTj8IUxm818jnvqKFdqWLwmqg0=bUtQ2-rp6rZ_Fb4eLkE_Ye7S7bN3X4
>> CtfdUqQURN5sE=yMgpsPYg5Q_bN3I5bhvkGyvxgmLx4CLkO4HDCa3ZMM4=
>> SEARCH the MAILING-LIST at:  https://urldefense.proofpoint.
>>

Re: [Wien] different MLD for bcc structure for magnetic equivalent directions M001, M010 and M100

2017-11-28 Thread Laurence Marks 
Interesting. I assume that this is with -so. Does the sum :PUP+:PDN obey
the fcc selection rules? (You can also look at the PW in case.clmsum &
case.valup/dn.)

_
Professor Laurence Marks
"Research is to see what everybody else has seen, and to think what nobody
else has thought", Albert Szent-Gyorgi
www.numis.northwestern.edu

On Nov 28, 2017 5:36 AM, "Jaroslav Hamrle" 
wrote:

> Dear Laurence,
>
> thank you for your detailed answer.
>
> I have tried all your suggestions,
> - I changed case.in0 with increased oversampling by factor two (new
> parameters LUSE 26 and IFFTfactor 4)
>  start of case.in0 ---
> TOT  XC_LDA (XC_PBE,XC_PBESOL,XC_WC,XC_MBJ,XC_REVTPSSS)
> NR2V  IFFT  26  (R2V)
> 24   24   244.00  1min IFFT-parameters, enhancement factor,
> iprint
>  end of case.in0 ---
>
> - I also tried to impose strong convergence criteria to be -cc
> 0.0001 -ec 0.0001
>
> However, in both cases, the ghost MLD remains practically identical as
> when using my default values (default Fe + convergence -cc 0.1 -ec
> 0.01)
>
> Also, final :PUP 'Current' parameters remained practically the same for
> all the calculations (by about 2 digits), being like (for M001)
> PW CHANGE HKL  Current   ChangeResidue
> :PUP001:  000  2.10719649E-02  5.090E-10 -7.530E-09
> :PUP002:  0   -1   -1  3.67084665E-04 -7.004E-10 -3.572E-10
> :PUP003:  1   -10  1.82124988E-04 -3.669E-10 -2.211E-10
> :PUP004:  00   -2 -1.87471938E-03  6.192E-11  7.254E-10
> :PUP005:  0   -20 -3.75090680E-03  1.125E-10  1.378E-09
> :PUP006:  1   -1   -2 -3.46372731E-03  1.026E-10  1.378E-09
> :PUP007:  1   -2   -1 -6.92804324E-03  2.418E-10  2.776E-09
> :PUP008:  0   -2   -2 -7.14014083E-04  8.677E-11  2.032E-10
> :PUP009:  2   -20 -3.57036516E-04  4.298E-11  1.121E-10
> :PUP010:  0   -1   -3  2.62159615E-04  9.199E-11 -1.588E-10
> :PUP011:  0   -3   -1  2.62289930E-04  9.844E-11 -1.555E-10
> :PUP012:  1   -30  2.62219244E-04  9.133E-11 -1.551E-10
> Unfortunately, all reflections seems to be allowed for bcc (H+K+L is
> even), forbidden reflections of bcc are (H+K+L=odd), so I can not see
> how they get close to zero ;-)) But the idea is excellent.
>
> Thank you again and with my best regards
>
> Jaroslav
>
> On 27/11/17 15:27, Laurence Marks wrote:
> > Let me clarify slightly my comment about symmetry -- as I realized the
> > explanation (I think) and can also suggest something that might help.
> >
> > First, concerning symmetry the explanation is I believe simple. If the
> > problem has a real symmetry operation such as inversion which is being
> > removed, then the Jacobian at the solution has zero's for charge
> > disturbances that break this symmetry. Because of this noise due to
> > numerical accuracy has a large effect, and almost certainly one has to
> > tighten the convergence criteria particularly -cc. You can monitor
> > this by looking at the :PUPXXX values in case.scfm and look how well
> > the forbidden reflections have converged to zero.
> >
> > Second, do not be surprised about numerical issues. While the
> > calculations are done in double precision, there are many large sums
> > and in some cases double sums, and also numerical
> > integrations/differentiation. Any large sum or numerical
> > integration/differentiation in general reduces the numerical accuracy.
> > Hence even though double precision has an accuracy of 1D-15 the sum
> > may only be accurate to 1D-10 or even 1D-7. Also, the Intel ifort
> > compiler will reduce the numerical accuracy for speed if one is not
> > careful.
> >
> > One thing that may help is to increase the oversampling in case.in0
> > for VXC, both that of the PW's and of the CLMs. A standard test is to
> > use LDA and see if the problem goes away, since oversampling is much
> > less relevant for this.
> >
> > Of course your problem may have nothing to do with any of this
>
> ___
> Wien mailing list
> Wien@zeus.theochem.tuwien.ac.at
> https://urldefense.proofpoint.com/v2/url?u=http-3A__zeus.
> theochem.tuwien.ac.at_mailman_listinfo_wien=DwIGaQ=
> yHlS04HhBraes5BQ9ueu5zKhE7rtNXt_d012z2PA6ws=U_
> T4PL6jwANfAy4rnxTj8IUxm818jnvqKFdqWLwmqg0=bUtQ2-rp6rZ_Fb4eLkE_
> Ye7S7bN3X4CtfdUqQURN5sE=yMgpsPYg5Q_bN3I5bhvkGyvxgmLx4CLkO4HDCa3ZMM4=
> SEARCH the MAILING-LIST at:  https://urldefense.proofpoint.
> com/v2/url?u=http-3A__www.mail-2Darchive.com_wien-
> 40zeus.theochem.tuwien.ac.at_index.html=DwIGaQ=
> yHlS04HhBraes5BQ9ueu5zKhE7rtNXt_d012z2PA6ws=U_
> T4PL6jwANfAy4rnxTj8IUxm818jnvqKFdqWLwmqg0=bUtQ2-rp6rZ_Fb4eLkE_
> Ye7S7bN3X4CtfdUqQURN5sE=uMgLUPSM3EVs3UYAH3vo4bu8OK5CAe8g8QuKu7zc7wc=
>
___
Wien mailing list
Wien@zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
SEARCH the MAILING-LIST at:

Re: [Wien] different MLD for bcc structure for magnetic equivalent directions M001, M010 and M100

2017-11-28 Thread Jaroslav Hamrle 

Dear Laurence,

thank you for your detailed answer.

I have tried all your suggestions,
- I changed case.in0 with increased oversampling by factor two (new 
parameters LUSE 26 and IFFTfactor 4)

 start of case.in0 ---
TOT  XC_LDA (XC_PBE,XC_PBESOL,XC_WC,XC_MBJ,XC_REVTPSSS)
NR2V  IFFT  26  (R2V)
   24   24   24    4.00  1    min IFFT-parameters, enhancement factor, 
iprint

 end of case.in0 ---

- I also tried to impose strong convergence criteria to be -cc 
0.0001 -ec 0.0001


However, in both cases, the ghost MLD remains practically identical as 
when using my default values (default Fe + convergence -cc 0.1 -ec 
0.01)


Also, final :PUP 'Current' parameters remained practically the same for 
all the calculations (by about 2 digits), being like (for M001)

PW CHANGE H    K    L  Current   Change    Residue
:PUP001:  0    0    0  2.10719649E-02  5.090E-10 -7.530E-09
:PUP002:  0   -1   -1  3.67084665E-04 -7.004E-10 -3.572E-10
:PUP003:  1   -1    0  1.82124988E-04 -3.669E-10 -2.211E-10
:PUP004:  0    0   -2 -1.87471938E-03  6.192E-11  7.254E-10
:PUP005:  0   -2    0 -3.75090680E-03  1.125E-10  1.378E-09
:PUP006:  1   -1   -2 -3.46372731E-03  1.026E-10  1.378E-09
:PUP007:  1   -2   -1 -6.92804324E-03  2.418E-10  2.776E-09
:PUP008:  0   -2   -2 -7.14014083E-04  8.677E-11  2.032E-10
:PUP009:  2   -2    0 -3.57036516E-04  4.298E-11  1.121E-10
:PUP010:  0   -1   -3  2.62159615E-04  9.199E-11 -1.588E-10
:PUP011:  0   -3   -1  2.62289930E-04  9.844E-11 -1.555E-10
:PUP012:  1   -3    0  2.62219244E-04  9.133E-11 -1.551E-10
Unfortunately, all reflections seems to be allowed for bcc (H+K+L is 
even), forbidden reflections of bcc are (H+K+L=odd), so I can not see 
how they get close to zero ;-)) But the idea is excellent.


Thank you again and with my best regards

Jaroslav

On 27/11/17 15:27, Laurence Marks wrote:
Let me clarify slightly my comment about symmetry -- as I realized the 
explanation (I think) and can also suggest something that might help.


First, concerning symmetry the explanation is I believe simple. If the 
problem has a real symmetry operation such as inversion which is being 
removed, then the Jacobian at the solution has zero's for charge 
disturbances that break this symmetry. Because of this noise due to 
numerical accuracy has a large effect, and almost certainly one has to 
tighten the convergence criteria particularly -cc. You can monitor 
this by looking at the :PUPXXX values in case.scfm and look how well 
the forbidden reflections have converged to zero.


Second, do not be surprised about numerical issues. While the 
calculations are done in double precision, there are many large sums 
and in some cases double sums, and also numerical 
integrations/differentiation. Any large sum or numerical 
integration/differentiation in general reduces the numerical accuracy. 
Hence even though double precision has an accuracy of 1D-15 the sum 
may only be accurate to 1D-10 or even 1D-7. Also, the Intel ifort 
compiler will reduce the numerical accuracy for speed if one is not 
careful.


One thing that may help is to increase the oversampling in case.in0 
for VXC, both that of the PW's and of the CLMs. A standard test is to 
use LDA and see if the problem goes away, since oversampling is much 
less relevant for this.


Of course your problem may have nothing to do with any of this


___
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

Re: [Wien] different MLD for bcc structure for magnetic equivalent directions M001, M010 and M100

2017-11-27 Thread Laurence Marks 
Let me clarify slightly my comment about symmetry -- as I realized the
explanation (I think) and can also suggest something that might help.

First, concerning symmetry the explanation is I believe simple. If the
problem has a real symmetry operation such as inversion which is being
removed, then the Jacobian at the solution has zero's for charge
disturbances that break this symmetry. Because of this noise due to
numerical accuracy has a large effect, and almost certainly one has to
tighten the convergence criteria particularly -cc. You can monitor this by
looking at the :PUPXXX values in case.scfm and look how well the forbidden
reflections have converged to zero.

Second, do not be surprised about numerical issues. While the calculations
are done in double precision, there are many large sums and in some cases
double sums, and also numerical integrations/differentiation. Any large sum
or numerical integration/differentiation in general reduces the numerical
accuracy. Hence even though double precision has an accuracy of 1D-15 the
sum may only be accurate to 1D-10 or even 1D-7. Also, the Intel ifort
compiler will reduce the numerical accuracy for speed if one is not careful.

One thing that may help is to increase the oversampling in case.in0 for
VXC, both that of the PW's and of the CLMs. A standard test is to use LDA
and see if the problem goes away, since oversampling is much less relevant
for this.

Of course your problem may have nothing to do with any of this

On Mon, Nov 27, 2017 at 8:07 AM, Jaroslav Hamrle 
wrote:

> Dear all,
>
> thank you for your comments:
>
> 1)
>
> Did you use a Gamma centered k mesh (and enough k points)
>
>
> I have checked that the same inequality in MLD, appears both when k-points
> are shifted or not shifted. So, influence of shift of k-points can be ruled
> out.
>
> I have 30x30x30 k-points, which should be enough. When using 46x46x46
> k-mesh, the MLD inequality is reduced by about factor two, but still
> present. Even when using very fine k-mesh (90x90x90) for optical
> calculation, the inequality persists.
>
>
> 2)
>
> In some cells  shifting the k-point origin with MSR1a leads to slightly
> unbalanced forces which are hard to converge to the "right" symmetric
> result. If the forces are slightly off, this is an indication that the
> density is also slightly off. Exactly why this occurs I do not know, I
> suspect very soft modes associated with numerical errors in finite
> arithmetic.
>
>
> Well, for me it is hard to believe, that the problem can originates from
> numerical error. The MLD inequality creates sort-of ghost peaks in MLD
> spectra which for bcc Fe are very stable at positions at 4.8 and 6 eV for
> different calculation details (as different k-mesh, presence/absence of
> spin-orbit, shifted/non-shifted k-mesh). Under all those changes in the
> calculations, the position of ghost peaks remains very stable, just their
> amplitudes varies. Also, if problem would be just numerical one, why ghost
> peaks are not present in simple cubic or fcc calculations?
>
>
> 3)
>
>
> It is instructive to repeat the calculation without SO and see how big the
> difference between sig_xx and sig_yy (for any M direction) is then.
>
>
> I tried to calculate various combinations of structure, (with/without SO
> or sp) using full (non-magnetic) bcc symmetry, or bcc symmetry reduced by
> presence of magnetization (i.e. it means reduction of symmetry + new k-mesh
> as generated by initso).
> In case of spin-polarized calculations without SO, the ghost MLD peaks
> appear when going from full bcc symmetry to bcc symmetry reduced by
> magnetization.
>
>
>
> nosp+noso sp+noso sp+so
> bcc full symmetry without magnetization OK OK X
> bcc symmetry reduced by magnetization ? ghosts ghosts
> (fcc or simple cubic) reduced by magnetization
> ? OK OK
>
>
>
>
> Therefore it seems to me that the ghost MLD peaks appear when symmetry is
> reduced in the bcc structure. It seems that SO coupling is not important in
> this problem.
> It is the lower symmetry itself, which creates the MLD inequality (ghost
> MLD peaks).
>
> For example, can there be some small problem with generation of k-mesh or
> related symmetry in bcc+magnetization case?
>
>
> Thank you for your help
> With my best regards
> Jaroslav
>
>
>
>
>
>
>
>
> On 26/11/17 18:51, Laurence Marks wrote:
>
> I will third the comment that not using a shifted cell might be important
> (might). In some cells  shifting the k-point origin with MSR1a leads to
> slightly unbalanced forces which are hard to converge to the "right"
> symmetric result. If the forces are slightly off, this is an indication
> that the density is also slightly off. Exactly why this occurs I do not
> know, I suspect very soft modes associated with numerical errors in finite
> arithmetic.
>
> N.B., Wien2k is quite good with these numerical errors. I've noticed that
> Vasp calculations that collaborators have done often have much larger
>

Re: [Wien] different MLD for bcc structure for magnetic equivalent directions M001, M010 and M100

2017-11-27 Thread Jaroslav Hamrle 

Dear all,

thank you for your comments:

1)


Did you use a Gamma centered k mesh (and enough k points)


I have checked that the same inequality in MLD, appears both when 
k-points are shifted or not shifted. So, influence of shift of k-points 
can be ruled out.


I have 30x30x30 k-points, which should be enough. When using 46x46x46 
k-mesh, the MLD inequality is reduced by about factor two, but still 
present. Even when using very fine k-mesh (90x90x90) for optical 
calculation, the inequality persists.



2)

In some cells shifting the k-point origin with MSR1a leads to slightly 
unbalanced forces which are hard to converge to the "right" symmetric 
result. If the forces are slightly off, this is an indication that the 
density is also slightly off. Exactly why this occurs I do not know, I 
suspect very soft modes associated with numerical errors in finite 
arithmetic.


Well, for me it is hard to believe, that the problem can originates from 
numerical error. The MLD inequality creates sort-of ghost peaks in MLD 
spectra which for bcc Fe are very stable at positions at 4.8 and 6 eV 
for different calculation details (as different k-mesh, presence/absence 
of spin-orbit, shifted/non-shifted k-mesh). Under all those changes in 
the calculations, the position of ghost peaks remains very stable, just 
their amplitudes varies. Also, if problem would be just numerical one, 
why ghost peaks are not present in simple cubic or fcc calculations?



3)



It is instructive to repeat the calculation without SO and see how big 
the difference between sig_xx and sig_yy (for any M direction) is then. 


I tried to calculate various combinations of structure, (with/without SO 
or sp) using full (non-magnetic) bcc symmetry, or bcc symmetry reduced 
by presence of magnetization (i.e. it means reduction of symmetry + new 
k-mesh as generated by initso).
In case of spin-polarized calculations without SO, the ghost MLD peaks 
appear when going from full bcc symmetry to bcc symmetry reduced by 
magnetization.




nosp+noso   sp+noso sp+so
bcc full symmetry without magnetization OK  OK  X
bcc symmetry reduced by magnetization   ?   ghosts  ghosts
(fcc or simple cubic) reduced by magnetization
?   OK  OK




Therefore it seems to me that the ghost MLD peaks appear when symmetry 
is reduced in the bcc structure. It seems that SO coupling is not 
important in this problem.
It is the lower symmetry itself, which creates the MLD inequality (ghost 
MLD peaks).


For example, can there be some small problem with generation of k-mesh 
or related symmetry in bcc+magnetization case?



Thank you for your help
With my best regards
Jaroslav








On 26/11/17 18:51, Laurence Marks wrote:
I will third the comment that not using a shifted cell might be 
important (might). In some cells  shifting the k-point origin with 
MSR1a leads to slightly unbalanced forces which are hard to converge 
to the "right" symmetric result. If the forces are slightly off, this 
is an indication that the density is also slightly off. Exactly why 
this occurs I do not know, I suspect very soft modes associated with 
numerical errors in finite arithmetic.


N.B., Wien2k is quite good with these numerical errors. I've noticed 
that Vasp calculations that collaborators have done often have much 
larger symmetry breaking.


On Sun, Nov 26, 2017 at 11:39 AM, Karel Vyborny > wrote:


I suppose that this does not have to do (much) with centering the
mesh.
My guess based on other QMO calculations is that some contributions to
mat. els. of e.g. vx*vx from different parts of the BZ don't cancel
(numerically) even if they actually should.

It is instructive to repeat the calculation without SO and see how
big the
difference between sig_xx and sig_yy (for any M direction) is then.

Cheers,

Karel


--- x ---
dr. Karel Vyborny
Fyzikalni ustav AV CR, v.v.i.
Cukrovarnicka 10
Praha 6, CZ-16253
tel: +420220318459


On Sun, 26 Nov 2017, Fecher, Gerhard wrote:

> There was a recent discussion on magnetic anisotropy, With a
remark by Peter,
> Did you use a Gamma centered k mesh (and enough k points)
>
> Ciao
> Gerhard
>
> DEEP THOUGHT in D. Adams; Hitchhikers Guide to the Galaxy:
> "I think the problem, to be quite honest with you,
> is that you have never actually known what the question is."
>
> 
> Dr. Gerhard H. Fecher
> Institut of Inorganic and Analytical Chemistry
> Johannes Gutenberg - University
> 55099 Mainz
> and
> Max Planck Institute for Chemical Physics of Solids
> 01187 Dresden

--
Professor Laurence Marks
"Research is to see what everybody else has seen, and to think what 
nobody else has thought", Albert Szent-Gyorgi
www.numis.northwestern.edu  ;

Re: [Wien] different MLD for bcc structure for magnetic equivalent directions M001, M010 and M100

2017-11-26 Thread Karel Vyborny 
I suppose that this does not have to do (much) with centering the mesh. 
My guess based on other QMO calculations is that some contributions to 
mat. els. of e.g. vx*vx from different parts of the BZ don't cancel 
(numerically) even if they actually should.


It is instructive to repeat the calculation without SO and see how big the 
difference between sig_xx and sig_yy (for any M direction) is then.


Cheers,

Karel


--- x ---
dr. Karel Vyborny
Fyzikalni ustav AV CR, v.v.i.
Cukrovarnicka 10
Praha 6, CZ-16253
tel: +420220318459


On Sun, 26 Nov 2017, Fecher, Gerhard wrote:


There was a recent discussion on magnetic anisotropy, With a remark by Peter,
Did you use a Gamma centered k mesh (and enough k points)

Ciao
Gerhard

DEEP THOUGHT in D. Adams; Hitchhikers Guide to the Galaxy:
"I think the problem, to be quite honest with you,
is that you have never actually known what the question is."


Dr. Gerhard H. Fecher
Institut of Inorganic and Analytical Chemistry
Johannes Gutenberg - University
55099 Mainz
and
Max Planck Institute for Chemical Physics of Solids
01187 Dresden

Von: Wien [wien-boun...@zeus.theochem.tuwien.ac.at] im Auftrag von Jaroslav 
Hamrle [ham...@karlov.mff.cuni.cz]
Gesendet: Sonntag, 26. November 2017 12:12
An: wien@zeus.theochem.tuwien.ac.at
Betreff: Re: [Wien] different MLD for bcc structure for magnetic equivalent 
directions M001, M010 and M100

Hi Gerhard,

I know that due to SO, the electronic structure calculated for 100, 010
and 001 magnetization directions are different.

The problem I have is following:
I have three calculated electronic structures of bcc Fe, with
magnetizations along 001, 010 and 100.
Then, for any cubic structure, the permittivity tensor elements (ep_ij)
with the same relations with respect to the magnetization should be
equal in all three calculated structures.

For example,
symmetry clearly states that diagonal permittivity elements parallel to
magnetization direction must equal
ep_xx (for M=100) = ep_yy (for M=010) = ep_zz (for M=001).

My problem is, that for calculated bcc Fe they do not equal.
More specifically, they do not equal solely for bcc structures in
wien2k, with disagreement upto 1%.
For simple cubic and fcc structures they do equal, with tiny
disagreement upto 0.01%

Any help how to overcome this would be very helpful

Thank you and with my best regards

Jaroslav




On 25/11/17 14:13, Fecher, Gerhard wrote:

Hi Jaroslav,

with SO, 001 is not equivalent to 001 or 010, if the magnetisation is along 001
this you see easily from the changed symmetry after initializing SO (symmetso)

regards from Dresden

Ciao
Gerhard

DEEP THOUGHT in D. Adams; Hitchhikers Guide to the Galaxy:
"I think the problem, to be quite honest with you,
is that you have never actually known what the question is."


Dr. Gerhard H. Fecher
Institut of Inorganic and Analytical Chemistry
Johannes Gutenberg - University
55099 Mainz
and
Max Planck Institute for Chemical Physics of Solids
01187 Dresden

Von: Wien [wien-boun...@zeus.theochem.tuwien.ac.at] im Auftrag von Jaroslav 
Hamrle [ham...@karlov.mff.cuni.cz]
Gesendet: Freitag, 24. November 2017 16:36
An: wien@zeus.theochem.tuwien.ac.at
Betreff: [Wien] different MLD for bcc structure for magnetic equivalent 
directions M001, M010 and M100

Dear colleagues,


We have found non-physical asymmetry related with equivalent magnetization
directions, when calculating electronic structure for bcc Fe:

We want to calculate magnetic linear dichroism, MLD, defined as a
difference between diagonal permittivity element being parallel,
perpendicular to direction of magnetization, respectively.

MLD=epzz - (epxx+epyy)/2 for M001

MLD=epyy - (epxx+epzz)/2 for M010

MLD=epxx - (epyy+epzz)/2 for M100


Obviously, MLD calculated for different equivalent magnetization
directions should
be identical. But they are not, MLD calculated for 001 is different to
MLD calculated for 010 and 100 (MLD for 010 and 100 are identical).

In most cases, we used k-mesh 30x30x30, exgange LDA (choice 5), with
convergence criteria

runsp_lapw -so -cc 0.01 -ec 0.001 -s lapw1

and the convergence was reached.

* We tested this asymmetry also for fcc structures (Ni, Co, Co2MnSi). We
also
tested simple cubic structure (bcc Fe, defined  as a simple cubic
structure with two Fe atoms).  In all those cases, the asymmetry
disappears. On the other hand, it also appeared also in bcc Ni.
Hence, the asymmetry seems to be specifically related with
bcc structure.

* this asymmetry can be observed already in energy levels (files
case.energysoup). Hence, we think, the asymmetry is not a feature of
optics.
Namely, there is a very good agreement for energies for M010 and M100
(in example below difference is below 2e-7Ry), but much bigger
difference between energies for M001 and (M010,M100)  (in example below
max. difference is

Re: [Wien] different MLD for bcc structure for magnetic equivalent directions M001, M010 and M100

2017-11-26 Thread Fecher, Gerhard 
There was a recent discussion on magnetic anisotropy, With a remark by Peter,
Did you use a Gamma centered k mesh (and enough k points)

Ciao
Gerhard

DEEP THOUGHT in D. Adams; Hitchhikers Guide to the Galaxy:
"I think the problem, to be quite honest with you,
is that you have never actually known what the question is."


Dr. Gerhard H. Fecher
Institut of Inorganic and Analytical Chemistry
Johannes Gutenberg - University
55099 Mainz
and
Max Planck Institute for Chemical Physics of Solids
01187 Dresden

Von: Wien [wien-boun...@zeus.theochem.tuwien.ac.at] im Auftrag von Jaroslav 
Hamrle [ham...@karlov.mff.cuni.cz]
Gesendet: Sonntag, 26. November 2017 12:12
An: wien@zeus.theochem.tuwien.ac.at
Betreff: Re: [Wien] different MLD for bcc structure for magnetic equivalent 
directions M001, M010 and M100

Hi Gerhard,

I know that due to SO, the electronic structure calculated for 100, 010
and 001 magnetization directions are different.

The problem I have is following:
I have three calculated electronic structures of bcc Fe, with
magnetizations along 001, 010 and 100.
Then, for any cubic structure, the permittivity tensor elements (ep_ij)
with the same relations with respect to the magnetization should be
equal in all three calculated structures.

For example,
symmetry clearly states that diagonal permittivity elements parallel to
magnetization direction must equal
ep_xx (for M=100) = ep_yy (for M=010) = ep_zz (for M=001).

My problem is, that for calculated bcc Fe they do not equal.
More specifically, they do not equal solely for bcc structures in
wien2k, with disagreement upto 1%.
For simple cubic and fcc structures they do equal, with tiny
disagreement upto 0.01%

Any help how to overcome this would be very helpful

Thank you and with my best regards

Jaroslav




On 25/11/17 14:13, Fecher, Gerhard wrote:
> Hi Jaroslav,
>
> with SO, 001 is not equivalent to 001 or 010, if the magnetisation is along 
> 001
> this you see easily from the changed symmetry after initializing SO (symmetso)
>
> regards from Dresden
>
> Ciao
> Gerhard
>
> DEEP THOUGHT in D. Adams; Hitchhikers Guide to the Galaxy:
> "I think the problem, to be quite honest with you,
> is that you have never actually known what the question is."
>
> 
> Dr. Gerhard H. Fecher
> Institut of Inorganic and Analytical Chemistry
> Johannes Gutenberg - University
> 55099 Mainz
> and
> Max Planck Institute for Chemical Physics of Solids
> 01187 Dresden
> 
> Von: Wien [wien-boun...@zeus.theochem.tuwien.ac.at] im Auftrag von Jaroslav 
> Hamrle [ham...@karlov.mff.cuni.cz]
> Gesendet: Freitag, 24. November 2017 16:36
> An: wien@zeus.theochem.tuwien.ac.at
> Betreff: [Wien] different MLD for bcc structure for magnetic equivalent 
> directions M001, M010 and M100
>
> Dear colleagues,
>
>
> We have found non-physical asymmetry related with equivalent magnetization
> directions, when calculating electronic structure for bcc Fe:
>
> We want to calculate magnetic linear dichroism, MLD, defined as a
> difference between diagonal permittivity element being parallel,
> perpendicular to direction of magnetization, respectively.
>
> MLD=epzz - (epxx+epyy)/2 for M001
>
> MLD=epyy - (epxx+epzz)/2 for M010
>
> MLD=epxx - (epyy+epzz)/2 for M100
>
>
> Obviously, MLD calculated for different equivalent magnetization
> directions should
> be identical. But they are not, MLD calculated for 001 is different to
> MLD calculated for 010 and 100 (MLD for 010 and 100 are identical).
>
> In most cases, we used k-mesh 30x30x30, exgange LDA (choice 5), with
> convergence criteria
>
> runsp_lapw -so -cc 0.01 -ec 0.001 -s lapw1
>
> and the convergence was reached.
>
> * We tested this asymmetry also for fcc structures (Ni, Co, Co2MnSi). We
> also
> tested simple cubic structure (bcc Fe, defined  as a simple cubic
> structure with two Fe atoms).  In all those cases, the asymmetry
> disappears. On the other hand, it also appeared also in bcc Ni.
> Hence, the asymmetry seems to be specifically related with
> bcc structure.
>
> * this asymmetry can be observed already in energy levels (files
> case.energysoup). Hence, we think, the asymmetry is not a feature of
> optics.
> Namely, there is a very good agreement for energies for M010 and M100
> (in example below difference is below 2e-7Ry), but much bigger
> difference between energies for M001 and (M010,M100)  (in example below
> max. difference is 18e-6 Ry for band 5). Therefore it seems
> that this problem arises in either lapw0 or lapw1 for bcc structure.
>
> To demonstrate the difference, we show energy levels for the first
> k-point (in vicinity of the Gamma point shifted in  111 direction from
> the Gamma point):
>
> Fe30M001:
>
>0.E-01 0.E-01 0.E-01 155
> 18  8.0
>  1  -3.4390104377017581
>  2  -3.4064979309023942
>

Re: [Wien] different MLD for bcc structure for magnetic equivalent directions M001, M010 and M100

2017-11-26 Thread Jaroslav Hamrle 

Hi Gerhard,

I know that due to SO, the electronic structure calculated for 100, 010 
and 001 magnetization directions are different.


The problem I have is following:
I have three calculated electronic structures of bcc Fe, with 
magnetizations along 001, 010 and 100.
Then, for any cubic structure, the permittivity tensor elements (ep_ij) 
with the same relations with respect to the magnetization should be 
equal in all three calculated structures.


For example,
symmetry clearly states that diagonal permittivity elements parallel to 
magnetization direction must equal

ep_xx (for M=100) = ep_yy (for M=010) = ep_zz (for M=001).

My problem is, that for calculated bcc Fe they do not equal.
More specifically, they do not equal solely for bcc structures in 
wien2k, with disagreement upto 1%.
For simple cubic and fcc structures they do equal, with tiny 
disagreement upto 0.01%


Any help how to overcome this would be very helpful

Thank you and with my best regards

Jaroslav




On 25/11/17 14:13, Fecher, Gerhard wrote:

Hi Jaroslav,

with SO, 001 is not equivalent to 001 or 010, if the magnetisation is along 001
this you see easily from the changed symmetry after initializing SO (symmetso)

regards from Dresden

Ciao
Gerhard

DEEP THOUGHT in D. Adams; Hitchhikers Guide to the Galaxy:
"I think the problem, to be quite honest with you,
is that you have never actually known what the question is."


Dr. Gerhard H. Fecher
Institut of Inorganic and Analytical Chemistry
Johannes Gutenberg - University
55099 Mainz
and
Max Planck Institute for Chemical Physics of Solids
01187 Dresden

Von: Wien [wien-boun...@zeus.theochem.tuwien.ac.at] im Auftrag von Jaroslav 
Hamrle [ham...@karlov.mff.cuni.cz]
Gesendet: Freitag, 24. November 2017 16:36
An: wien@zeus.theochem.tuwien.ac.at
Betreff: [Wien] different MLD for bcc structure for magnetic equivalent 
directions M001, M010 and M100

Dear colleagues,


We have found non-physical asymmetry related with equivalent magnetization
directions, when calculating electronic structure for bcc Fe:

We want to calculate magnetic linear dichroism, MLD, defined as a
difference between diagonal permittivity element being parallel,
perpendicular to direction of magnetization, respectively.

MLD=epzz - (epxx+epyy)/2 for M001

MLD=epyy - (epxx+epzz)/2 for M010

MLD=epxx - (epyy+epzz)/2 for M100


Obviously, MLD calculated for different equivalent magnetization
directions should
be identical. But they are not, MLD calculated for 001 is different to
MLD calculated for 010 and 100 (MLD for 010 and 100 are identical).

In most cases, we used k-mesh 30x30x30, exgange LDA (choice 5), with
convergence criteria

runsp_lapw -so -cc 0.01 -ec 0.001 -s lapw1

and the convergence was reached.

* We tested this asymmetry also for fcc structures (Ni, Co, Co2MnSi). We
also
tested simple cubic structure (bcc Fe, defined  as a simple cubic
structure with two Fe atoms).  In all those cases, the asymmetry
disappears. On the other hand, it also appeared also in bcc Ni.
Hence, the asymmetry seems to be specifically related with
bcc structure.

* this asymmetry can be observed already in energy levels (files
case.energysoup). Hence, we think, the asymmetry is not a feature of
optics.
Namely, there is a very good agreement for energies for M010 and M100
(in example below difference is below 2e-7Ry), but much bigger
difference between energies for M001 and (M010,M100)  (in example below
max. difference is 18e-6 Ry for band 5). Therefore it seems
that this problem arises in either lapw0 or lapw1 for bcc structure.

To demonstrate the difference, we show energy levels for the first
k-point (in vicinity of the Gamma point shifted in  111 direction from
the Gamma point):

Fe30M001:

   0.E-01 0.E-01 0.E-01 155
18  8.0
 1  -3.4390104377017581
 2  -3.4064979309023942
 3  -3.3508627657180750
 4  -3.2276472567243979
 5  -3.1955089683446780
 6  -3.1702455400854954
 7  -7.1658179115217727E-002
 8  -4.3723732810772589E-002
 9  0.37296762299903474
   10  0.37521967189559313

Fe30M010:

   0.E-01 0.E-01 0.E-01 155
18  8.0
 1  -3.4390110394480322
 2  -3.4064968725403300
 3  -3.3508644682352022
 4  -3.2276486274720977
 5  -3.1954902103327028
 6  -3.1702472318057655
 7  -7.1659013996950252E-002
 8  -4.3723316415832839E-002
 9  0.37296632778787425
10  0.37521816821120640
Fe30M100:

   0.E-01 0.E-01 0.E-01 155
18  8.0
 1  -3.4390109925234049
 2  -3.4064968346346225
 3  -3.3508643700301919
 4  -3.2276485335559135
 5  -3.1954901707639891

Re: [Wien] different MLD for bcc structure for magnetic equivalent directions M001, M010 and M100

2017-11-25 Thread Fecher, Gerhard 
Hi Jaroslav,

with SO, 001 is not equivalent to 001 or 010, if the magnetisation is along 001
this you see easily from the changed symmetry after initializing SO (symmetso) 

regards from Dresden

Ciao
Gerhard

DEEP THOUGHT in D. Adams; Hitchhikers Guide to the Galaxy:
"I think the problem, to be quite honest with you,
is that you have never actually known what the question is."


Dr. Gerhard H. Fecher
Institut of Inorganic and Analytical Chemistry
Johannes Gutenberg - University
55099 Mainz
and
Max Planck Institute for Chemical Physics of Solids
01187 Dresden

Von: Wien [wien-boun...@zeus.theochem.tuwien.ac.at] im Auftrag von Jaroslav 
Hamrle [ham...@karlov.mff.cuni.cz]
Gesendet: Freitag, 24. November 2017 16:36
An: wien@zeus.theochem.tuwien.ac.at
Betreff: [Wien] different MLD for bcc structure for magnetic equivalent 
directions M001, M010 and M100

Dear colleagues,


We have found non-physical asymmetry related with equivalent magnetization
directions, when calculating electronic structure for bcc Fe:

We want to calculate magnetic linear dichroism, MLD, defined as a
difference between diagonal permittivity element being parallel,
perpendicular to direction of magnetization, respectively.

MLD=epzz - (epxx+epyy)/2 for M001

MLD=epyy - (epxx+epzz)/2 for M010

MLD=epxx - (epyy+epzz)/2 for M100


Obviously, MLD calculated for different equivalent magnetization
directions should
be identical. But they are not, MLD calculated for 001 is different to
MLD calculated for 010 and 100 (MLD for 010 and 100 are identical).

In most cases, we used k-mesh 30x30x30, exgange LDA (choice 5), with
convergence criteria

runsp_lapw -so -cc 0.01 -ec 0.001 -s lapw1

and the convergence was reached.

* We tested this asymmetry also for fcc structures (Ni, Co, Co2MnSi). We
also
tested simple cubic structure (bcc Fe, defined  as a simple cubic
structure with two Fe atoms).  In all those cases, the asymmetry
disappears. On the other hand, it also appeared also in bcc Ni.
Hence, the asymmetry seems to be specifically related with
bcc structure.

* this asymmetry can be observed already in energy levels (files
case.energysoup). Hence, we think, the asymmetry is not a feature of
optics.
Namely, there is a very good agreement for energies for M010 and M100
(in example below difference is below 2e-7Ry), but much bigger
difference between energies for M001 and (M010,M100)  (in example below
max. difference is 18e-6 Ry for band 5). Therefore it seems
that this problem arises in either lapw0 or lapw1 for bcc structure.

To demonstrate the difference, we show energy levels for the first
k-point (in vicinity of the Gamma point shifted in  111 direction from
the Gamma point):

Fe30M001:

  0.E-01 0.E-01 0.E-01 155
18  8.0
1  -3.4390104377017581
2  -3.4064979309023942
3  -3.3508627657180750
4  -3.2276472567243979
5  -3.1955089683446780
6  -3.1702455400854954
7  -7.1658179115217727E-002
8  -4.3723732810772589E-002
9  0.37296762299903474
  10  0.37521967189559313

Fe30M010:

  0.E-01 0.E-01 0.E-01 155
18  8.0
1  -3.4390110394480322
2  -3.4064968725403300
3  -3.3508644682352022
4  -3.2276486274720977
5  -3.1954902103327028
6  -3.1702472318057655
7  -7.1659013996950252E-002
8  -4.3723316415832839E-002
9  0.37296632778787425
   10  0.37521816821120640
Fe30M100:

  0.E-01 0.E-01 0.E-01 155
18  8.0
1  -3.4390109925234049
2  -3.4064968346346225
3  -3.3508643700301919
4  -3.2276485335559135
5  -3.1954901707639891
6  -3.1702471600962974
7  -7.1658839213425571E-002
8  -4.3723135315494835E-002
9  0.37296642179994044
   10  0.37521826479385562

The difference in energy is (in micro-Ry):

left column E(001)-E(100), right column E(010)-E(100)

 1   0.5548-0.0469
 2  -1.0963-0.0379
 3   1.6043-0.0982
 4   1.2768-0.0939
 5-18.7976-0.0396
 6   1.6200-0.0717
 7   0.6601-0.1748
 8  -0.5975-0.1811
 9   1.2012-0.0940
 10 1.4071-0.0966

clearly, energies for M001 are different from those for M010 and M100.

* We checked also similar calculation by elk code. Here the
asymmetry in optical spectra is not present.
On the other hand, it is presented and equals in all wien2k versions 14-17.
Similar asymmetry is also present when testing equivalent directions of
types 110 and 111 directions in bcc Fe.

* calculations were done for k-point mesh 30x30x30, but the same
symmetry/asymmetry appears from k-mesh