The ABINIT Users Mailing List ( CLOSED )

[Philippe Ghosez <Philippe.Ghosez@ulg.ac.be>]

Dear David,

static dielectric constant requires the knowledge of
- the optical dielectric constant and Born effective charges (Z*)
- the phonon eigenvectors and frequencies.

While the first quantities can be obtained together without rfatpol
the latter require computation of the dynamical matrix with
rfatpol ranging on all non-equivalent atoms.

So, although you can avoid rfatpol when computing Z*, you
cannot avoid it when looking at the static dielectric constant.

Best wishes,

Philippe.


On 06 Sep 2007, at 09:24, David Hendry wrote:

> Thank you all for your input.  Sorry, I'm still a little confused.  Is 
> it
> possible to calculate the static dielectric constant without having to
> specify rfatpol ranging over all ions?
>
> The v5 t23.in example includes the electric field as well as the 
> individual
> ionic perturbations.  I'm not sure how I would get the static dielectic
> constant without specifying rfatpol.
>
> Also, there could possibly be a Segmentation fault when running
> v5/Input/t23.in in parallel.  Abinip with 1 cpu is OK but Abinip with 
> 2 cpus
> produces the error below at the command prompt.
>
> #################################################
>
> [rapier614:05855] *** Process received signal ***
> [rapier614:05855] Signal: Segmentation fault (11)
> [rapier614:05855] Signal code: Address not mapped (1)
> [rapier614:05855] Failing at address: 0x2e6e7cc0
> [rapier614:05855] [ 0] /lib/tls/libpthread.so.0 [0x8598b0]
> [rapier614:05855] [ 1]
> /root/abinit-5.4.3/src/main/abinip(gradberry3_+0x3409) [0x836f445]
> [rapier614:05855] [ 2] 
> /root/abinit-5.4.3/src/main/abinip(vtorho3_+0xc170)
> [0x8344ce0]
> [rapier614:05855] [ 3] 
> /root/abinit-5.4.3/src/main/abinip(scfcv3_+0x353c)
> [0x832f978]
> [rapier614:05855] [ 4] 
> /root/abinit-5.4.3/src/main/abinip(loper3_+0xb4b6)
> [0x81fd02e]
> [rapier614:05855] [ 5] 
> /root/abinit-5.4.3/src/main/abinip(respfn_+0xfaeb)
> [0x8194c47]
> [rapier614:05855] [ 6] 
> /root/abinit-5.4.3/src/main/abinip(driver_+0x9147)
> [0x806c04f]
> [rapier614:05855] [ 7] 
> /root/abinit-5.4.3/src/main/abinip(MAIN__+0x5406)
> [0x80603ea]
> [rapier614:05855] [ 8] /root/abinit-5.4.3/src/main/abinip(main+0x3d)
> [0x805afd5]
> [rapier614:05855] [ 9] /lib/tls/libc.so.6(__libc_start_main+0xe3) 
> [0x88ce33]
> [rapier614:05855] [10] /root/abinit-5.4.3/src/main/abinip [0x805af11]
> [rapier614:05855] *** End of error message ***
> forrtl: error (78): process killed (SIGTERM)
> Image              PC        Routine            Line        Source
>
> abinip             0899F96B  Unknown               Unknown  Unknown
> abinip             0899EF8B  Unknown               Unknown  Unknown
> abinip             0897251A  Unknown               Unknown  Unknown
> abinip             0893C316  Unknown               Unknown  Unknown
> abinip             0893E992  Unknown               Unknown  Unknown
> libpthread.so.0    008598B8  Unknown               Unknown  Unknown
> abinip             087C7344  Unknown               Unknown  Unknown
> abinip             087BC267  Unknown               Unknown  Unknown
> abinip             08544D05  Unknown               Unknown  Unknown
> abinip             08358E16  Unknown               Unknown  Unknown
> abinip             0834827E  Unknown               Unknown  Unknown
> abinip             08340C9C  Unknown               Unknown  Unknown
> abinip             0832F978  Unknown               Unknown  Unknown
> abinip             081FD02E  Unknown               Unknown  Unknown
> abinip             08194C47  Unknown               Unknown  Unknown
> abinip             0806C04F  Unknown               Unknown  Unknown
> abinip             080603EA  Unknown               Unknown  Unknown
> abinip             0805AFD5  Unknown               Unknown  Unknown
> libc.so.6          006A5E33  Unknown               Unknown  Unknown
> abinip             0805AF11  Unknown               Unknown  Unknown
> mpirun noticed that job rank 1 with PID 5855 on node rapier614 exited 
> on
> signal 11 (Segmentation fault).
>
> ##################################################
>
>
> 3 and 4 cpus make the calculation hang at the end of Dataset 1 with the
> final log lines"
>
>  Computing the polarization (Berry phase) for reciprocal vector:
>   0.50000  0.00000  0.00000 (in reduced coordinates)
>  -0.04708  0.04708  0.04708 (in cartesian coordinates - atomic units)
>  Number of strings:     4
>  Number of k points in string:    2
>
>
>
> Thank you all again,
>
> Regards,
>
> David.
>
>
>
> -----Original Message-----
> From: Philippe Ghosez [mailto:Philippe.Ghosez@ulg.ac.be]
> Sent: Wednesday, September 05, 2007 5:21 PM
> To: forum@abinit.org
> Subject: Re: [abinit-forum] 5.4.3 - better at static dielectric 
> constant?
>
> Dear Nicola,
>
> the computation of Z* in zero field for ALL the atoms in the unit cell 
> was
> already previously accessible by linear response from the ddk and 
> electric
> field perturbation ONLY : i.e. calculation of the optical dielectric
> constant is also  providing the Z* on all the atoms in the same run 
> using a
> non-stationary expression with no need to displace all the  
> uniquivalent
> atoms independently). Except if I missed something I do not think the 
> new
> implementation helps here.
>
> Now I agree that the calculation of Z* in finite field provides an
> interesting alternative way (to computation of  third-derivatives by 
> linear
> response) for the Raman efficiencies since the dchi/dtau can be
> alternatively written dZ*/dE and are now easily accessible by finite
> diffirence as shown in the paper of Xinjie.
>
> By the way, was this new feature implemented also for spin-polarized
> systems?
>
> Best wishes,
>
> Philippe.
>
>
> On 05 Sep 2007, at 05:57, Nicola Marzari wrote:
>
> > One side comment - it can become more efficient if the unit cell
> > contains many inequivalent atoms.
> >
> > Suppose you have 100 atoms - usually, you would need to displace each
> > of them, one by one, in the three crystallographic directions, and
> > calculate the resulting polarization (this can be done either with
> > linear-response theory, i.e. as the response to an infinitesimal
> > displacement, or with the Berry phase, as the response to a finite
> > one).
> >
> > Instead, you can apply a small electric field three times (in 3 non
> > equivalent directions), and you have right away the effective charges
> > for all the 100 atoms. Similar considerations apply, say, if you
> > wanted to calculate Raman couplings.
> >
> >
> >                                 nicola
> >
> >
> > Xinjie Wang wrote:
> > > Hi David,
> > > This feature (B.10) does not give more efficient/accurate 
> > > calculations
> > > of the static dielectric properties. It allows the calculation of
> > > electric-field-dependence of Born effective charges and dielectric
> > > tensor. You can read this paper, Phys. Rev. B 75, 115116 (2007),
> > > for details.
> > > Regards,
> > > Xinjie
> > > david.hendry@mandmd.com.au wrote:
> > > > Dear Abinitors,
> > > >
> > > > I'm interested in the static dielectric constant of
> > > > ceramics/insulators.  I notice the latest release notes of Abinit:
> > > > "B.10 Born effective charges and dielectric tensor calculations can
> > > > now be computed in a finite electric field." Does this feature or
> > > > any other new feature (from 5.3.4) allow a more efficient/accurate
> > > > calculation of the static dielectric properties?
> > > >
> > > > Sincerely,
> > > >
> > > > David.
> >
> >
> > --
> > ---------------------------------------------------------------------
> > Prof Nicola Marzari   Department of Materials Science and Engineering
> > 13-5066   MIT   77 Massachusetts Avenue   Cambridge MA 02139-4307 USA
> > tel 617.4522758 fax 2586534 marzari@mit.edu http://quasiamore.mit.edu
> --------------------------------------------------
> Philippe GHOSEZ
> Universite de Liege
> Institut de Physique, Bat. B5
> Allee du 6 aout, 17
> B- 4000 Sart Tilman
> BELGIUM
>
> Phone : ++(32) (0)4-366.36.11
> Fax   : ++(32) (0)4-366.36.29
> E-mail: Philippe.Ghosez@ulg.ac.be
>
> http://www.ulg.ac.be/phythema
> --------------------------------------------------
--------------------------------------------------
Philippe GHOSEZ
Universite de Liege
Institut de Physique, Bat. B5
Allee du 6 aout, 17
B- 4000 Sart Tilman
BELGIUM

Phone : ++(32) (0)4-366.36.11
Fax   : ++(32) (0)4-366.36.29
E-mail: Philippe.Ghosez@ulg.ac.be

http://www.ulg.ac.be/phythema
--------------------------------------------------