Re: [SIESTA-L] << two set of charge calculations >>

[Emilio Artacho]

oops, sorry Camps, my lecturing you was not pertinent!

E

On Jun 20, 2022, at 10:12 PM, I. Camps 
mailto:ica...@gmail.com>> wrote:

I completely agree with Tamas and Emilio BUT my question is not related to 
which charge calculation scheme is "better".

My question is that in my calculations, two different sets of data of the same 
type of charges are appearing in the output file, instead only one for each. I 
have two outputs for Hirshfeld and two outputs for Voronoi charges.

[]'s,

Camps


On Mon, Jun 20, 2022 at 5:02 PM Emilio Artacho 
mailto:e.arta...@nanogune.eu>> wrote:
Tamas’s reply is correct, I just want to add a reminder of the fact
that atomic charges have a fundamental definition problem and none of
the proposals gives the ‘good’ answer. This is a direct consequence
of its responding to an ill-posed question: how many electrons ‘belong’
to a given atom (or can be assigned to it). It is perfectly defined if the atoms
are infinitely separated from each other, but not otherwise.

It is clear, however, that concepts like charge transfer etc are useful
in chemistry and very much support chemical analysis and intuition.
Atomic charges schemes (when used sensibly) are valuable. Just remember
to use them with care (qualitatively, trends etc). There are good comparative
studies assessing their reliability in various chemistry situations.

There are situations for which the question can be rephrased
into something physically well defined (see e,g, the Born effective
charges, or other questions relating to dielectric polarisation).

One can also find claims in the literature for a particular scheme to be
the ‘right’ one. To my mind they all rely on arbitrary choices, which can
be more or less sensible or well motivated, but still arbitrary (as Tamas
says, some depend on the basis set choice while other do not, for
instance).

best

Emilio

On Jun 19, 2022, at 2:47 PM, Tamas Karpati 
mailto:tkarp...@gmail.com>> wrote:

Dear Camps,

Please note that an argument is going on for decades about how to
calculate atomic charges. Different methods/schemes give different
results, each is giving better/worse results for different
applications. It is recommended to check how well each performs at
your actual problem and choose which one is to be used. Also
remarkable is the basis set dependence of atomic charges, consider
this a parameter to be calibrated.

Regards,
 t

On Fri, Jun 17, 2022 at 10:02 PM I. Camps 
mailto:ica...@gmail.com>> wrote:

Hello Alberto,

Here it is the info about the SIESTA version:

Siesta Version  : siesta-max-R3--710-676-597
Architecture: unknown
Compiler version: ifort (IFORT) 19.1.1.217 20200306
Compiler flags  : mpifort -fPIC -O2 -march=core-avx2 -axCore-AVX512 -fp-model 
precise
PP flags: -DFC_HAVE_ABORT -DF2003 -DMPI -DCDF -DNCDF -DNCDF_4 
-DNCDF_PARALLEL 
-I/cvmfs//soft.computecanada.ca/easybuild/software/2020/avx2/MPI/intel2020/openmpi4/netcdf-fortran-mpi/4.5.2/include
Libraries   : libncdf.a libfdict.a -Wl,-Bstatic -Wl,--start-group 
-lmkl_scalapack_lp64 -lmkkl_blacs_openmpi_lp64 -lmkl_intel_lp64 
-lmkl_sequential -lmkl_core -Wl,--end-group -Wl,-Bdynamic -lnetcdff
PARALLEL version
NetCDF support
NetCDF-4 support
NetCDF-4 MPI-IO support

And here is the output section:

siesta: Final energy (eV):
siesta:  Band Struct. =   -8272.290139
siesta:   Kinetic =   19960.524774
siesta:   Hartree =  151423.860682
siesta:   Eldau   =   0.00
siesta:   Eso =   0.00
siesta:Ext. field =   0.00
siesta:   Enegf   =   0.00
siesta:   Exch.-corr. =  -11180.064205
siesta:  Ion-electron = -320401.282309
siesta:   Ion-ion =  129282.468462
siesta:   Ekinion =   0.00
siesta: Total =  -30914.492596
siesta: Fermi =  -4.212218

siesta: Stress tensor (static) (eV/Ang**3):
siesta: 0.0001260.00   -0.00
siesta: 0.000.000101   -0.49
siesta:-0.00   -0.49   -0.016465

siesta: Cell volume =   7672.635004 Ang**3

siesta: Pressure (static):
siesta:SolidMolecule  Units
siesta:   0.5895  0.5941  Ry/Bohr**3
siesta:   0.00541292  0.00545494  eV/Ang**3
siesta:   8.67254766  8.73987328  kBar
(Free)E+ p_basis*V_orbitals  =  -30859.763440
(Free)Eharris+ p_basis*V_orbitals  =  -30859.763491
spin moment: S , {S} =0.0   0.0   0.0   0.0

siesta: Electric dipole (a.u.)  =0.000.0432460.00
siesta: Electric dipole (Debye) =0.010.1099190.00

Hirshfeld Net Atomic Populations:
Atom #Qatom  Species
10.149  B
20.149  B
30.149  B
40.149  B
5   -0.149  N
...
  155   -0.149  N
  156   -0.149  N
  1570.149  B
  1580.149  B
  1590.149  B
  1600.149  B

Voronoi Net 

Re: [SIESTA-L] Total energy

[Francisco Garcia]

Thanks  for your response Prof. Artacho!

So if I have a situation like below for a large metallic system in which
FreeEng is lower than Etot, should I use Etot? Specifically, if I want to
plot the E-V curve from single point runs to obtain the equation of state I
should use Etot instead of FreeEng? Or if I want to compute the cohesive
energy, Etot should be used instead of FreeEng?

siesta: Program's energy decomposition (eV):
siesta: Ebs =  -859.157108
siesta: Eions   =  9568.777238
siesta: Ena =   277.651800
siesta: Ekin=  4247.415878
siesta: Enl = -1698.956164
siesta: Eso = 0.00
siesta: Edftu   = 0.00
siesta: DEna=   162.636813
siesta: DUscf   =26.981770
siesta: DUext   = 0.00
siesta: Exc = -6715.533109
siesta: eta*DQ  = 0.00
siesta: Emadel  = 0.00
siesta: Emeta   = 0.00
siesta: Emolmec = 0.00
siesta: Ekinion = 0.00
siesta: Eharris =  -13268.580292
siesta: Etot=-13268.580250
siesta: FreeEng =-13269.911499

On Tue, Jun 21, 2022 at 2:07 PM Emilio Artacho 
wrote:

>
> On 20 Jun 2022, at 17:16, Francisco Garcia 
> wrote:
>
> Dear users,
>
> In metallic systems with a fairly sizable electronic smearing temperature
> T, is it accurate to claim that
>
> (i) in a single point calculation, the free energy is the representative
> energy of the system (due to the addition of -TS to the total energy U)
>
>
> No, the -TS term for the phononic entropy is missing.
>
> It is the free energy defining the finite temperature equilibrium for the
> purely electronic
> problem for fixed external potential (fixed nuclei), as in Mermin’s finite
> -T DFT.
>
>
> (ii) in a variable cell optimization, the enthalpy is the representative
> energy of the system (due to the addition of the PV term to the energy).
>
>
> Yes, it is the free energy of the system (minimum defines equilibrium) for
> T=0.
>
> Emilio
>
>
> Thanks!
>
>
> --
> SIESTA is supported by the Spanish Research Agency (AEI) and by the
> European H2020 MaX Centre of Excellence (http://www.max-centre.eu/)
>
>
> --
> Emilio Artacho
>
> Theory Group, Nanogune, 20018 San Sebastian, Spain, and
> Theory of Condensed Matter, Department of Physics,
> Cavendish Laboratory, University of Cambridge, Cambridge CB3 0HE, UK
>
>
>
>
>
> --
> SIESTA is supported by the Spanish Research Agency (AEI) and by the
> European H2020 MaX Centre of Excellence (http://www.max-centre.eu/)
>

-- 
SIESTA is supported by the Spanish Research Agency (AEI) and by the European 
H2020 MaX Centre of Excellence (http://www.max-centre.eu/)

Re: [SIESTA-L] << two set of charge calculations >>

[Nick Papior]

The 2nd data output relates to your ldos range. So they are effectively not
equivalent. The first is the final scf, the 2nd ldos.

I think this is the case, but I am not 100 sure (not at my computer to
check).

On Tue, 21 Jun 2022, 22:00 I. Camps,  wrote:

> I completely agree with Tamas and Emilio BUT my question *is not* related
> to which charge calculation scheme is "better".
>
> My question is that in my calculations, two different sets of data of the
> same type of charges are appearing in the output file, instead only one for
> each. I have two outputs for Hirshfeld and two outputs for Voronoi charges.
>
> []'s,
>
> Camps
>
>
> On Mon, Jun 20, 2022 at 5:02 PM Emilio Artacho 
> wrote:
>
>> Tamas’s reply is correct, I just want to add a reminder of the fact
>> that atomic charges have a fundamental definition problem and none of
>> the proposals gives the ‘good’ answer. This is a direct consequence
>> of its responding to an ill-posed question: how many electrons ‘belong’
>> to a given atom (or can be assigned to it). It is perfectly defined if
>> the atoms
>> are infinitely separated from each other, but not otherwise.
>>
>> It is clear, however, that concepts like charge transfer etc are useful
>> in chemistry and very much support chemical analysis and intuition.
>> Atomic charges schemes (when used sensibly) are valuable. Just remember
>> to use them with care (qualitatively, trends etc). There are good
>> comparative
>> studies assessing their reliability in various chemistry situations.
>>
>> There are situations for which the question can be rephrased
>> into something physically well defined (see e,g, the Born effective
>> charges, or other questions relating to dielectric polarisation).
>>
>> One can also find claims in the literature for a particular scheme to be
>> the ‘right’ one. To my mind they all rely on arbitrary choices, which can
>> be more or less sensible or well motivated, but still arbitrary (as Tamas
>> says, some depend on the basis set choice while other do not, for
>> instance).
>>
>> best
>>
>> Emilio
>>
>> On Jun 19, 2022, at 2:47 PM, Tamas Karpati  wrote:
>>
>> Dear Camps,
>>
>> Please note that an argument is going on for decades about how to
>> calculate atomic charges. Different methods/schemes give different
>> results, each is giving better/worse results for different
>> applications. It is recommended to check how well each performs at
>> your actual problem and choose which one is to be used. Also
>> remarkable is the basis set dependence of atomic charges, consider
>> this a parameter to be calibrated.
>>
>> Regards,
>>  t
>>
>> On Fri, Jun 17, 2022 at 10:02 PM I. Camps  wrote:
>>
>>
>> Hello Alberto,
>>
>> Here it is the info about the SIESTA version:
>>
>> Siesta Version  : siesta-max-R3--710-676-597
>> Architecture: unknown
>> Compiler version: ifort (IFORT) 19.1.1.217 20200306
>> Compiler flags  : mpifort -fPIC -O2 -march=core-avx2 -axCore-AVX512
>> -fp-model precise
>> PP flags: -DFC_HAVE_ABORT -DF2003 -DMPI -DCDF -DNCDF -DNCDF_4
>> -DNCDF_PARALLEL -I/cvmfs//
>> soft.computecanada.ca/easybuild/software/2020/avx2/MPI/intel2020/openmpi4/netcdf-fortran-mpi/4.5.2/include
>> Libraries   : libncdf.a libfdict.a -Wl,-Bstatic -Wl,--start-group
>> -lmkl_scalapack_lp64 -lmkkl_blacs_openmpi_lp64 -lmkl_intel_lp64
>> -lmkl_sequential -lmkl_core -Wl,--end-group -Wl,-Bdynamic -lnetcdff
>> PARALLEL version
>> NetCDF support
>> NetCDF-4 support
>> NetCDF-4 MPI-IO support
>>
>> And here is the output section:
>>
>> siesta: Final energy (eV):
>> siesta:  Band Struct. =   -8272.290139
>> siesta:   Kinetic =   19960.524774
>> siesta:   Hartree =  151423.860682
>> siesta:   Eldau   =   0.00
>> siesta:   Eso =   0.00
>> siesta:Ext. field =   0.00
>> siesta:   Enegf   =   0.00
>> siesta:   Exch.-corr. =  -11180.064205
>> siesta:  Ion-electron = -320401.282309
>> siesta:   Ion-ion =  129282.468462
>> siesta:   Ekinion =   0.00
>> siesta: Total =  -30914.492596
>> siesta: Fermi =  -4.212218
>>
>> siesta: Stress tensor (static) (eV/Ang**3):
>> siesta: 0.0001260.00   -0.00
>> siesta: 0.000.000101   -0.49
>> siesta:-0.00   -0.49   -0.016465
>>
>> siesta: Cell volume =   7672.635004 Ang**3
>>
>> siesta: Pressure (static):
>> siesta:SolidMolecule  Units
>> siesta:   0.5895  0.5941  Ry/Bohr**3
>> siesta:   0.00541292  0.00545494  eV/Ang**3
>> siesta:   8.67254766  8.73987328  kBar
>> (Free)E+ p_basis*V_orbitals  =  -30859.763440
>> (Free)Eharris+ p_basis*V_orbitals  =  -30859.763491
>> spin moment: S , {S} =0.0   0.0   0.0   0.0
>>
>> siesta: Electric dipole (a.u.)  =0.000.0432460.00
>> siesta: Electric dipole (Debye) =0.010.1099190.00
>>
>> Hirshfeld Net Atomic Populations:
>> 

Re: [SIESTA-L] << two set of charge calculations >>

[Tamas Karpati]

Dear Camps, Emilio,

True, I overlooked the point in the original question (my apologies!)
but thought that the doubling of charge prints is simply due to a spin
polarized calculation.

I was wrong, as that doubling also appears in case of closed shell
singlets' outputs. What is more, the two lists are quite different.

Conclusion: I don't know the reason (only I can think of the initial
and final data of an iterative charge calculation but it's just an
idea).

Question: Could someone clarify this?

Best regards,
  t


On Tue, Jun 21, 2022 at 10:03 PM I. Camps  wrote:
>
> I completely agree with Tamas and Emilio BUT my question is not related to 
> which charge calculation scheme is "better".
>
> My question is that in my calculations, two different sets of data of the 
> same type of charges are appearing in the output file, instead only one for 
> each. I have two outputs for Hirshfeld and two outputs for Voronoi charges.
>
> []'s,
>
> Camps
>
>
> On Mon, Jun 20, 2022 at 5:02 PM Emilio Artacho  wrote:
>>
>> Tamas’s reply is correct, I just want to add a reminder of the fact
>> that atomic charges have a fundamental definition problem and none of
>> the proposals gives the ‘good’ answer. This is a direct consequence
>> of its responding to an ill-posed question: how many electrons ‘belong’
>> to a given atom (or can be assigned to it). It is perfectly defined if the 
>> atoms
>> are infinitely separated from each other, but not otherwise.
>>
>> It is clear, however, that concepts like charge transfer etc are useful
>> in chemistry and very much support chemical analysis and intuition.
>> Atomic charges schemes (when used sensibly) are valuable. Just remember
>> to use them with care (qualitatively, trends etc). There are good comparative
>> studies assessing their reliability in various chemistry situations.
>>
>> There are situations for which the question can be rephrased
>> into something physically well defined (see e,g, the Born effective
>> charges, or other questions relating to dielectric polarisation).
>>
>> One can also find claims in the literature for a particular scheme to be
>> the ‘right’ one. To my mind they all rely on arbitrary choices, which can
>> be more or less sensible or well motivated, but still arbitrary (as Tamas
>> says, some depend on the basis set choice while other do not, for
>> instance).
>>
>> best
>>
>> Emilio
>>
>> On Jun 19, 2022, at 2:47 PM, Tamas Karpati  wrote:
>>
>> Dear Camps,
>>
>> Please note that an argument is going on for decades about how to
>> calculate atomic charges. Different methods/schemes give different
>> results, each is giving better/worse results for different
>> applications. It is recommended to check how well each performs at
>> your actual problem and choose which one is to be used. Also
>> remarkable is the basis set dependence of atomic charges, consider
>> this a parameter to be calibrated.
>>
>> Regards,
>>  t
>>
>> On Fri, Jun 17, 2022 at 10:02 PM I. Camps  wrote:
>>
>>
>> Hello Alberto,
>>
>> Here it is the info about the SIESTA version:
>>
>> Siesta Version  : siesta-max-R3--710-676-597
>> Architecture: unknown
>> Compiler version: ifort (IFORT) 19.1.1.217 20200306
>> Compiler flags  : mpifort -fPIC -O2 -march=core-avx2 -axCore-AVX512 
>> -fp-model precise
>> PP flags: -DFC_HAVE_ABORT -DF2003 -DMPI -DCDF -DNCDF -DNCDF_4 
>> -DNCDF_PARALLEL 
>> -I/cvmfs//soft.computecanada.ca/easybuild/software/2020/avx2/MPI/intel2020/openmpi4/netcdf-fortran-mpi/4.5.2/include
>> Libraries   : libncdf.a libfdict.a -Wl,-Bstatic -Wl,--start-group 
>> -lmkl_scalapack_lp64 -lmkkl_blacs_openmpi_lp64 -lmkl_intel_lp64 
>> -lmkl_sequential -lmkl_core -Wl,--end-group -Wl,-Bdynamic -lnetcdff
>> PARALLEL version
>> NetCDF support
>> NetCDF-4 support
>> NetCDF-4 MPI-IO support
>>
>> And here is the output section:
>>
>> siesta: Final energy (eV):
>> siesta:  Band Struct. =   -8272.290139
>> siesta:   Kinetic =   19960.524774
>> siesta:   Hartree =  151423.860682
>> siesta:   Eldau   =   0.00
>> siesta:   Eso =   0.00
>> siesta:Ext. field =   0.00
>> siesta:   Enegf   =   0.00
>> siesta:   Exch.-corr. =  -11180.064205
>> siesta:  Ion-electron = -320401.282309
>> siesta:   Ion-ion =  129282.468462
>> siesta:   Ekinion =   0.00
>> siesta: Total =  -30914.492596
>> siesta: Fermi =  -4.212218
>>
>> siesta: Stress tensor (static) (eV/Ang**3):
>> siesta: 0.0001260.00   -0.00
>> siesta: 0.000.000101   -0.49
>> siesta:-0.00   -0.49   -0.016465
>>
>> siesta: Cell volume =   7672.635004 Ang**3
>>
>> siesta: Pressure (static):
>> siesta:SolidMolecule  Units
>> siesta:   0.5895  0.5941  Ry/Bohr**3
>> siesta:   0.00541292  0.00545494  eV/Ang**3
>> siesta:   8.67254766  8.73987328  kBar
>> (Free)E+ p_basis*V_orbitals  =  -30859.763440
>>