The ABINIT Users Mailing List ( CLOSED )

["Jorissen Kevin" <Kevin.Jorissen@ua.ac.be>]

Dear D. Hamann, dear M. Verstraete,
 
thanks for your time and your valuable remarks.
Some clarifications and updates :
 
* My system and OS :
I'm running abinis v. 4.3.3  on a Intel(R) Xeon(TM) CPU 2.40GHz, with 
Linux version 2.4.27  (gcc version 3.3.3 (Debian 20040401)) #1 SMP

* I have uploaded the full log and output file of which I gave some extracts 
in my first e-mail (NiTi.log and NiTi.out, both gzipped) to the public abinit 
ftp server ftp.abinit.org/pub/incoming .
 
* I indeed had a typo in 'mk1mem' which led to large (but for my system 
affordable) memory use.
 
* I have reduced the input file according to D. Hamann's remarks.  Thank you 
especially for the remark about metals / insulators.  As for the 
rfphon/rfatpol : I did not remove them earlier (or reduce rfdir, as you 
propose) since I want to study other (lower symmetry) Ni-Ti compounds after I 
get the cubic NiTi working.  However, I have removed those keys now - without 
improvement, unfortunately (see new, reduced input file below).
 
* For this reduced input file, I have also tried to run the precompiled 
executables from the abinit web site, but they produce the same error as the 
executables I compiled myself.
 
* I now see in the output which routine crashes : 
  **** for the reduced input file :
jaws> tail log -n 20
 ETOT 24   1.8344069219448     4.073E-08 3.523E-16 6.247E-11
 At SCF step   24       vres2   =  6.25E-11 < tolvrs=  1.00E-10 =>converged.

Input/Output Error 153: Input file ended
   In Procedure: nselt3
        At Line: 331
      Statement: Unformatted READ
           Unit: 17
   Connected To: NiTi_KG
           Form: Unformatted
         Access: Sequential
Records Read   : 78
Records Written: 3
End of diagnostics
jaws>
 
  **** for the full calculation described in my first e-mail :
 ETOT 41   1.3755743821956    -3.610E-08 3.624E-16 8.450E-11
 At SCF step   41       vres2   =  8.45E-11 < tolvrs=  1.00E-10 =>converged.

Input/Output Error 154: Wrong length record
   In Procedure: nselt3
        At Line: 331
      Statement: Unformatted READ
           Unit: 17
   Connected To: NiTi_KG
           Form: Unformatted
         Access: Sequential
Records Read   : 15502
Records Written: 378
End of diagnostics
jaws>

 
****  Here I describe the reduced input file and its output :
jaws> more NiTisimple.in
#TiNi calculation of elastic constants
ndtset 2
# For reduction of memory demands
mkmem 0
mkqmem 0
mk1mem 0
mffmem 0
# Set 1 : initial self-consistency
kptopt1 1
tolvrs1 1.0d-19
iscf1 5
# Set 3 : response-function strain calculation
rfdir2 1 0 0
nqpt2 1
qpt2 0 0 0
rfstrs2 3
getwfk2 -1
kptopt2 2
tolvrs2 1.0d-10
iscf2 2
diemix2 0.5
#common input data
ecutsm 0.0
acell 3*5.7165
ecut 6
natom 2
nband 18
ngkpt 2 2 2
nshiftk 1
shiftk 0.5 0.5 0.5
nstep 200
ntypat 2
rprim 1 0 0 0 1 0 0 0 1
occopt 4
tsmear 0.02
typat 1 2
xred 0.0 0.0 0.0 0.5 0.5 0.5
zatnum 22 28
jaws>

The files NiTisimple.out and NiTisimple.log are attached.
 
 
Thank you for your help,
kind regards,
 
 
Kevin Jorissen
 
EMAT - Electron Microscopy for Materials Science   
(http://webhost.ua.ac.be/emat/)
Dept. of Physics
 
UA - Universiteit Antwerpen
Groenenborgerlaan 171
B-2020 Antwerpen
Belgium
 
tel  +32 3 2653249
fax + 32 3 2653257
e-mail kevin.jorissen@ua.ac.be
 

________________________________

Van: D. R. Hamann [mailto:drh@bell-labs.com]
Verzonden: ma 29-11-2004 17:29
Aan: forum@abinit.org
Onderwerp: Re: [abinit-forum] crash in NiTi rf calculations with abinit 4.3.3


Dear Kevin,

Beyond Matthieu's reply, I've noted a few things in your input file
that are possible causes of trouble.  First of all, you are doing
much too much for the elastic constants of this system.  A better
model is Test_v4/t64.in, an elastic constant calculation for Al.
NiTi is a metal, so Dataset 2, the d/dk calculation needed for the
piezoelectric constant of insulators, is useless.  For the same reason,
rfelfd and getddk should be eliminated from your RF dataset.  Since the 
NiTi structure is so simple, there is no atomic relaxation with strain
so rfphon and rfatpol are not needed.  Finally, since NiTi is cubic,
rfdir 1 0 0 will give you all the elastic constants.  You can still
process the DDB file through anaddb to convert the output units to GPa.

I'm now running my version of your input modified as above to see if
I get a crash, but it is probably going to take all day.  This turned
up a typo in your data: you had mk"L"mem 0 when I'm sure you wanted
mk1mem 0, which easily exhausted the memory on my laptop.  You also
might want prtwf 0 for the RF calculation.

Finally, if you move to 4.4, the convergence of metallic RF calculations
has been significantly improved and you can use iscf=5.

Don Hamann

Jorissen Kevin wrote: 

        Dear abinit users,
         
        I'm using abinit 4.3.3 to calculate response properties (I'm mainly 
interested in its elastic constants) of cubic NiTi.
        I am using the input file specified below, which is very similar to 
Test_v4/t67.in.
        The first two datasets seem to finish correctly, but an error occurs 
in the third dataset (I have run the calculation twice, obtaining the crash 
at exactly the same point).
         
        As my experience with abinit is limited, and I am using a perhaps 
more advanced feature of the code, I hope that you will be able to point out 
my error from the information given below.
         
         
         
         
         
         
        At the moment of the crash, I have the following NiTi_STATUS :
         
        jaws> more backupofcrash/NiTi_STATUS
         Status file, with repetition rate  49, status number  1535367
          Level abinit         : call driver
          Level driver         : call respfn
          jdtset  =    3
          Level respfn         : call loper3
          Level loper3         : call scfcv3
          respcase=   13
          Level scfcv3         : call vtorho3
          istep   =   41
          Level vtorho3        : call kpgstr
          isppol  =    1
          ikpt    =  124
        jaws>
        
        In the log file, in the section of dataset 3, I find hundreds of 
WARNINGS of the following type :
         
         cgwf3: WARNING -
          New trial energy at line       4 =  -1.332355E+03
             is higher than former: -1.332355E+03
         
        The two energies mentioned seem to be identical in all printed 
digits.  This warning has once been reported on the forum, but with no 
explanation following.
         
        There are also hdr_check warnings since the k-mesh has changed, but I 
assume these are normal.
         
         
        The end of the log file does not show anything suspicious :
         
        
        
         Non-SCF iterations; k pt #   50  k=  0.12500  0.29167  0.12500  band 
residuals:
          3.75E-19  3.27E-18  1.48E-18  3.73E-18  4.65E-18  6.01E-18  
1.67E-17  1.40E-17
          1.50E-18  7.75E-18  1.37E-17  1.58E-17  1.74E-17  1.93E-18  
9.45E-18  3.45E-17
         -1.00E-01 -1.00E-01
         vtowfk3 : number of one-way 3D ffts skipped in vtowfk until now =  
712913
         vtowfk3 : prtvol=0, 1 or 2, do not print more k-points.
         ETOT 41   1.3755743821956    -3.610E-08 3.624E-16 8.450E-11
         At SCF step   41       vres2   =  8.45E-11 < tolvrs=  1.00E-10 
=>converged.
        
        (END of the file)
         
         
        Files present at this stage of the calculation :
        jaws> lr
        -rwxr-xr-x  1 jorissen users         57 Oct 17 15:27 NiTi.files
        -rw-r--r--  1 jorissen users     305169 Oct 17 15:27 28ni.pspnc
        -rw-r--r--  1 jorissen users     305169 Oct 17 15:27 22ti.pspnc
        -rw-r--r--  1 jorissen users        777 Nov 25 17:27 NiTi.in
        -rw-rw-r--  1 jorissen users   99351466 Nov 25 19:21 NiTio_DS1_WFK
        -rw-rw-r--  1 jorissen users      15613 Nov 25 19:21 NiTio_DS1_DDB
        -rw-rw-r--  1 jorissen users          0 Nov 25 19:30 NiTi_1WF2
        -rw-rw-r--  1 jorissen users 1537398898 Nov 26 04:15 NiTio_DS2_1WF7
        -rw-rw-r--  1 jorissen users 1537398898 Nov 26 11:22 NiTio_DS2_1WF8
        -rw-rw-r--  1 jorissen users 1537398898 Nov 26 21:11 NiTio_DS2_1WF9
        -rw-rw-r--  1 jorissen users  224195158 Nov 27 00:29 NiTio_DS3_1WF1
        -rw-rw-r--  1 jorissen users  224195158 Nov 27 04:52 NiTio_DS3_1WF4
        -rw-rw-r--  1 jorissen users 1537398898 Nov 28 02:56 NiTio_DS3_1WF10
        -rw-rw-r--  1 jorissen users  223561126 Nov 28 02:57 NiTi_WFKQ
        -rw-rw-r--  1 jorissen users  223561126 Nov 28 02:57 NiTi_WFGS
        -rw-rw-r--  1 jorissen users        392 Nov 28 06:20 NiTi_STATUS
        -rw-rw-r--  1 jorissen users    8942556 Nov 28 06:20 NiTi_KG
        -rw-rw-r--  1 jorissen users     336050 Nov 28 06:20 NiTi.out
        -rw-rw-r--  1 jorissen users    8277211 Nov 28 06:20 log
        jaws>
        
         
         
         
         
         
        The NiTi.in file :
         
        jaws> more NiTi.in
        #TiNi calculation of elastic constants
        ndtset 3
        # For reduction of memory demands
        mkmem 0
        mkqmem 0
        mklmem 0
        mffmem 0
        # Set 1 : initial self-consistency
        kptopt1 1
        tolvrs1 1.0d-19
        iscf1 5
        #set 2 the ddk
        rfelfd2 2 # only derivative of the gs wf
        rfdir2 1 1 1
        nqpt2 1
        qpt2 0.0 0.0 0.0 #only the gamma point
        getwfk2 -1
        kptopt2 2
        iscf2 -3
        tolwfr2 1.0d-20
        # Set 3 : response-function strain calculation
        rfphon3 1
        rfatpol3 1 2
        rfelfd3 3
        rfdir3 1 1 1
        nqpt3 1
        qpt3 0 0 0
        rfstrs3 3
        getwfk3 -2
        getddk3 -1
        kptopt3 2
        tolvrs3 1.0d-10
        iscf3 3
        diemix3 0.5
        #common input data
        acell 3*5.7165
        ecut 76
        iscf 5
        natom 2
        nband 18
        ngkpt 12 12 12
        nshiftk 1
        shiftk 0.5 0.5 0.5
        nstep 200
        ntypat 2
        rprim 1 0 0 0 1 0 0 0 1
        occopt 4
        tsmear 0.02
        typat 1 2
        xred 0.0 0.0 0.0 0.5 0.5 0.5
        zatnum 22 28
        jaws>
        
        
         
        Thank you very much for your time,
         
         
         
         
        Kevin Jorissen
         
        EMAT - Electron Microscopy for Materials Science   
(http://webhost.ua.ac.be/emat/)
        Dept. of Physics
         
        UA - Universiteit Antwerpen
        Groenenborgerlaan 171
        B-2020 Antwerpen
        Belgium
         
        tel  +32 3 2653249
        fax + 32 3 2653257
        e-mail kevin.jorissen@ua.ac.be
         


-- 
D. R. Hamann                            Phone: 908-582-4454
Director, Theoretical Materials         Fax:   908-582-4702
  Physics Research (retired)            email: drh@physics.bell-labs.com
Bell Laboratories 
Lucent Technologies
700 Mountain Ave, Room 1D-371
Murray Hill, NJ 07974-0636 USA

<<winmail.dat>>