[Octopus-users] Time propogation problems

Tue Feb 9 10:25:36 WET 2010

Hi,

>
> I'm performing my ground state calculations under the PBE GGA  exchange correlation potential. However how do I run my time propogation run under the LDA? At first I changed the XCFunctional back to the LDA potentials. However when I did a time propagation without any perturbation the energy wasn't stable. When I left the XCFunctional as the GGA that I ran the ground state calculation with, the time propogation energy was stable and the calculation ran faster.
> I can understand why the energy under an LDA propagation wouldn't be stable, after all the ground state calculated using the GGA won't be the ground state of the LDA, and so it would therefore relax to the ground state over time. Is what I'm doing TDLDA? In which case why is the propagation under the GGA running faster than the LDA?

The calculation runs faster if you use the lanczos exponential scheme,
because the eigenvalue that you are propagating is an eigenvalue of
the Hamiltonian, and the Lanczos iteration scheme convergens in that
case immediately. This happens if you used functional X for the gs and
functional X for the propagation, and no kick (it would happen with
LDA also).

>
> My other problem is that I'm attempting to output the projections of my time-dependant run using a Delta kick.
> Do I perform this calculation by setting TDOutput=8 or 256? Also will this allow me to get a frequency-dependant projection, after I use oct-cross-section?
> Regardless of this, whenever I run my td calculation with  TDOutput= 8 or 256 I always get the following error:

Do not use numbers when defining variables in the inp file. Use the
names instead ("TDOutput = multipoles + energy" or whatever), because
the meaning of the numbers is bound to change over versions, and the
variable names not. In any case, probably there is a bug in the
calculation of the projections.

Alberto.

>
>
> **************************** FATAL ERROR *****************************
> *** Fatal Error (description follows)
> *--------------------------------------------------------------------
> * From node =    0
> *--------------------------------------------------------------------
> * Could not load restart/gs
> **********************************************************************
>
> When my TDOutput=1, or when it's left to default, it runs fine. My ground state calculation also restarts fine, so it appears to just be a problem with the td run for some outputs only.
> Are there some other variables that I need to set in order for the correct data to be stored in the restart folder for these runs? My inp file is as follows.
>
>
> calculationmode=td
>
> units=2
> %species
> 'Au'|196.9666|spec_ps_psf|79|2|0
> %
>
> extrastates=10
>
> XYZCoordinates='Au.ANI'
>
>
> ConvAbsDens=1e-6
> EigensolverFinalTolerance=1.e-13
> EigensolverInitTolerance=1.e-7
> BoxShape=3
>
> Radius=3.5
>
> SpinComponents=1
> Spacing=0.20
> TypeOfMixing=2
> What2Mix=1
>
> FromScratch=True
>
> periodicdimensions=0
>
> XCFunctional=gga_x_pbe_r + gga_c_pbe
> XCKernel=lda_x+lda_c_pz_mod
>
> EMEta=0.10
> LRConvAbsDens = 1e-3
> LinearSolverTol = 1e-5
> Mixing=0.2
> ParallelizationStrategy=1 + 2
> EigenSolverMaxIter=200
>
> MaximumIter=100
>
> TDDeltaStrength=0.01
> %TDPolarization
>  0.01868 |  0.99958 | -0.02206
>  -0.63010 | -0.33635 | -0.69989
>  0.91751 | -0.35555 | -0.17820
> %
> TDPolarizationDirection=1
> TDPolarizationEquivAxes=3
>
> %TDPolarizationWPrime
> -0.276478173280 |  -0.323634804281 |   0.904886917421
> %
>
> TDExponentialMethod=lanczos
> TDExpOrder=100
>
> SmearingFunction = fermi_dirac
> Smearing = 0.01
>
> tmax=30
> TDEvolutionMethod= aetrs
> TDTimeStep =0.010
> TDMaximumIter=tmax/TDTimeStep
> TDOutput=256
>
>
> Again, thanks for your help,
>
> Bob
>
>
>
> ------------------------------
>
> Message: 4
> Date: Wed, 03 Feb 2010 10:00:43 +0000
> From: Micael Oliveira <micael at teor.fis.uc.pt>
> Subject: Re: [Octopus-users] Time propogation problems
> To: octopus-users at tddft.org
> Message-ID: <4B69494B.5050506 at teor.fis.uc.pt>
> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
>  Just a couple of things I forgot to mention in my previous mail. In
> the APE input file, you should decrease the value of ConvAbsDens,
> because it is defined in a slightly different way than in octopus and
> 1e-4 is a bit too large (the default value should be OK). In the gold
> clusters paper we did no mention the cut-off radii used for the
> pseudopotential, but I guess that could be useful for your tests. So
> here there are:
>
> %PPComponents
>   0  | 2.47
>   1  | 2.98
>   2  | 2.00
> %
>
>
>  Micael
>
> On 02/02/2010 11:58 PM, Robertson Burgess wrote:
>> Dear Octopus Users,
>>
>> I've just ran the time propogation using the PSF psuedopotential that comes with Octopus. My absorption cross section now matches well with the calculation I did earlier that did not use symmetry. My non-physical results must therefore be originating from my psuedopotential and not my symmetry calculations.
>>
>> Here is the input that I'm using to generate my pseudopotentials. I'm attempting to follow the methods outlined in Engel et. al. Phys Rev B 63, 125121 (2001) which were also used in the paper by Castro et. al.  JCP 129, 144110 (2008):
>>
>> Title = "Au"
>> CalculationMode = pp
>> Verbose = 30
>>
>> WaveEquation = dirac
>>
>> NuclearCharge=79
>>
>> ConvAbsDens = 1e-4
>> Mixing=0.3
>> MaximumIter=300
>>
>> PPOutputFileFormat=upf
>>
>> PPScheme = tm
>>
>> %PPComponents
>>   0 | 2.2
>>   1 | 3.0
>>   2 | 1.2
>> %
>>
>> %Orbitals
>>   "Xe"
>>   4 | 3 |  7 | 7
>>   5 | 2 |  5 | 5
>>   6 | 0 |  1 | 0
>>   6 | 1 |  0 | 0
>> %
>>
>> Any help in letting me know what I'm missing from this would help immensely.
>>
>> Sincerely,
>> Robertson Burgess
>> University of Newcastle, Australia
>>
>> _______________________________________________
>> Octopus-users mailing list
>> Octopus-users at tddft.org
>> http://www.tddft.org/mailman/listinfo/octopus-users
>>
>
>
>
> ------------------------------
>
> _______________________________________________
> Octopus-users mailing list
> Octopus-users at tddft.org
> http://www.tddft.org/mailman/listinfo/octopus-users
>
>
> End of Octopus-users Digest, Vol 66, Issue 3
> ********************************************
> _______________________________________________
> Octopus-users mailing list
> Octopus-users at tddft.org
> http://www.tddft.org/mailman/listinfo/octopus-users
>

-- 
===============================
Dr. Alberto Castro

Instituto de Biocomputación y Física de Sistemas Complejos (BIFI),
Corona de Aragón 42,
50009 Zaragoza, España.

Temporarily at:

Departamento de Física Teórica, Facultad de Ciencias,
Universidad de Zaragoza,
Pedro Cerbuna 12,
50009 Zaragoza, España.

Tel:   +34 976761000 3332
Fax:  +34 976761264
skype: alberto_c_barrigon
email: acastro at bifi.es
===============================