The ABINIT Users Mailing List ( CLOSED )

[Lingzhu Kong <kongl@physics.rutgers.edu>]

Thanks a lot, Martin!

Lingzhu

On Wed, 11 Nov 2009, Martin Stankovski wrote:

> Dear Lingzhu,
>
> It is possible to start from a density, but you need to take special
> measures to do so. An abinit SCF cycle by default does four
> conjugate-gradient minimisation steps per SCF cycle and it starts from
> random wavefuctions. That means that at the first step, abinit by default
> takes your "good" density, and does four minimisation steps on all the
> wavefunctions. Then abinit recalculates the density for the next step. At
> this point, having done only four conjugate gradient minimisations, the
> wavefunctions which produce the new density might not be very converged at
> all, and so in the next step your density is bad again.
>
> There are two easy options (if you have to use a density), either you make
> sure that abinit stays longer in the first step of the SCF cycle. This can
> be done with the two keywords:
>
> nline    -  which sets the maximum of how many conjugate gradient
> minimisation steps abinit performs inside each scf cycle
> nnsclo   -  sets number of loops inside each scf cycle
>
> Another oprion is to calculate all the wavefunctions you need with keeping
> the density frozen, by setting:
>
> iscf -2
>
> Then the number of iterations in the single non-scf cycle is set by nstep as
> usual. I would suggest looking up these keywords in the abinit
> documentation. A third alternative would be to change from the
> conjugate-gradient minimisation by setting a different, positive iscf, but I
> personally have little experience with that.
>
> Hope this helps,
> Best wishes,
> Martin Stankovski
>
>
> -----Original Message-----
> From: Lingzhu Kong [mailto:kongl@physics.rutgers.edu]
> Sent: den 10 november 2009 20:39
> To: forum@abinit.org
> Subject: [abinit-forum] Is irdden disabled in ABINIT?
>
>
> Dear ABINIT users,
>
> I tried both irdden and irdwfk to restart the calculation.
> It seems that irdden is not speeding the calculation at all although
> I started from the scf density. but irdwfk works. so is irdden
> not enabled? Thanks!
>
> Lingzhu
>
>
> No virus found in this incoming message.
> Checked by AVG - www.avg.com
> Version: 8.5.425 / Virus Database: 270.14.60/2495 - Release Date: 11/10/09
> 19:56:00