The ABINIT Users Mailing List ( CLOSED )

[Marek Veithen <Marek.Veithen@ulg.ac.be>]

Dear Wu, Clovis and all other ABINIT users who are interested in the
computation of non-linear optical properties,

there have been quite a number of emails on the forum during the last weeks
asking questions about the computation of non-linear properties and I think,
I should clarify a few points:

* I gave a presentation at the last ABINIT workshop where I explained 
what kind
  of non-linear optical properties can be computed and how this
  calculation can be done in practice. For those of you who did not 
attend the workshop,
  the slides of my presentation are available on the ABINIT wew site:

http://www.abinit.org/workshop/Presentations/veithen_1.pdf

  In addition, Xavier, Philippe and myself wrote 2 papers with the 
details of the implementation
  of the 2n + 1 theorem and the results we obtained for some 
ferroelectric oxides. One of them
  has been accepted for publication, the other one has just been 
submitted but the
  preprints of both are available on cond-mat [Note,  I submitted them 
today
  to the preprint server, so you will
  probably have to wait until monday before you can download them]:

http://arxiv.org/abs/cond-mat/0311240
http://arxiv.org/abs/cond-mat/0409067

* Concerning your comments about the negative sign of the non-linear optical
  susceptiblity in ZnO, I think it is perfectly possible to get a 
negative chi^(2)
  for some crystals. At the opposite to the linear dielectric 
susceptibility chi^(1),
  that is necessarily positive, the chi^(2) can either be positive or 
negative.
  You should also note, that the sign of chi^(2) you compute with ABINIT
  will depende on the orientation of your crystal axis.

* Finally, the problem Wu Rongqin mentioned yesterday, has already been
  reported by other people.  
  In order to compute a third-order energy derivative with respect to 
one or more
  electric fields, I have to use a finite difference expression to 
compute the derivative
  of a first-order wavefunction with respec to k. The way I implemented 
this
  finite difference expression is quite stable in most circumstances. 
Nevertheless, there
  are still some problems especially when you use a low number of 
k-points in your
  calculation (f. ex. Wu Rongqin used ngkpt 2 2 2). In most cases, it 
is sufficient
  to increase the number of k-points to circumvent this problem. In 
case of Wu's
  input file, the problem should disappear if you use at least ngkpt 4 4 4.

  In practical calculations on bulk crystals (at least those I did up 
to now),
  this issue is not restrictive at all. In fact, in order to obtain
  converged results, you have to use larger k-point grids than in a 
usual ground-state or
  linear response calculations. In my calculations on ferroelectrics, I 
typically
  used ngkpt 10 10 10 and for some semiconductors, I had to use ngkpt 
16 16 16.
  And for these large grids, there should be no problems at all.
  
I hope, these few informations will help all those of you who are 
interested in doing
non-linear response calculations. If you have still questions or 
suggestions,
don't hesitate to contact me. Feedback is alway welcome.

Marek

Clovis Darrigan wrote:

> Dear Wu and Abinitioners,
>
>     Actually for linear response, only the former can be obtained,
>     right? because in the visible range, nonlinear as well as linear,
>     is related to the properties of excitation states, which goes
>     beyond lda, rhgit? But as for very low frequency, the perturbation
>     only causes electronic polarization, right?
>
>
> About non-linear optical properties : my thesis work was on their 
> calculation by sum over states method for linear, first non-linear and 
> second non-linear order (i.e. : alpha, beta, gamma, or khi(1), khi(2) 
> and khi(3)). Informations I need in input are only energie and 
> momentum transitions between all occupied and virtual bands. All 
> effects (SHG, THG, Kerr, Pockel, IDRI...) can be calculated with 
> respect to frequency(ies).
>
> Two years ago (during a postdoctoral position), I made an interface of 
> my program with the excellent ABINIT, and it gave me first interesting 
> results. But I was not able to finish properly this work because a 
> hierarchical superior don't give out MY work when I quit the 
> laboratory...
>
>     I tried but I got a negative nonlinear coefficient for ZnO, is
>     this physical? I do not know why.
>
>
> For the first non-linear optical property, values can be negatives.
>
> I have a question about non-linear optical properties in v4.4. I read 
> it in the features but which effects will be available ?
>
> Sincerely,
>
>
> Clovis Darrigan


--
Marek VEITHEN
Universite de Liege
Institut de Physique, Bat. B5
Allee du 6 aout, 17
B- 4000 Sart Tilman
BELGIUM
                                                                               
Phone : ++(32) (0)4-366.36.12
Fax   : ++(32) (0)4-366.29.90
E-mail: Marek.Veithen@ulg.ac.be