Hi Longhua,

I'm also trying to make siesta-ldau working, but it seems that the code is 
quite experimental and not really optimized.
In my case it works only in serial mode and it takes much longer to converge.
You may try to set:

DM.NumberPulay        10
MeshCutoff            300.0 Ry
DM.MixingWeight         0.1
ElectronicTemperature 500.0 K

best,
Roberto

On Tue 25 Sep 2012, Longhua Li wrote:
> Dear all,
> 
> Recently, I tested the ldau version (serial run) by using the example of
> Fe_LDAU. The calculation seems to be infinite. Because the output file is
> stopped at the following message even after 3 hours running.
>  "siesta:    1    -1109.1768    -1117.0271    -1117.0271  0.2519  3.7921
> timer: Routine,Calls,Time,% = IterSCF        1       2.691  77.51
> elaps: Routine,Calls,Wall,% = IterSCF        1       2.720  72.93
>  DM norm before normalization:    15.9999999999536
>  Qtot:    16.0000000000000
> LDAU: Adding Hamiltonian contribution"
> 
> The calculation is still running, but no more messages.
> 
> Did anyone encounter this situation?
> 
> Best,
> Longhua
> 
> P.S. I use the same arch.make to compile the LDAU version. The 3.1 version
> is all right.
> 

Responder a