TURBOMOLE Users Forum
TURBOMOLE Modules => Jobex: Structure Optimization and Molecular Dynamics => Topic started by: Matteo Guglielmi on February 19, 2009, 03:59:42 PM
-
Hello everyone,
I have to compute the vibrational frequencies for a pretty big system
(doubly-protonated gramicidin-S in gas-phase, 176 atoms) at
def2-TZVP basis level.
I've encountered a couple of problems though:
1) the automatic redundant internal coordinate command (ired) fails
completely crashing define right on the spot.
I've solved this problem using cartesian coordinates instead of internal,
2) the statistic run crashes as well without doing anything basically.
The only way I could "solve" the latter problem was to modify by hand
the control file taking as reference what a successful statistic run writes
into the control file of a completely different system, namely I've modified/
added the following few lines:
$last step statistics run for dscf
$statistics off
$2e-ints_shell_statistics file=metastase
$end
... using the 'metastase' file generated for the "different system".
Now the geometry optimization is running and I got quite
a few steps already... but the question is:
can I rely on the results I will obtain?
In other words, could a wrong/fake metastase file mess
up everything?
Thanks,
MG.
-
Hi,
In other words, could a wrong/fake metastase file mess up everything?
YES, a wrong task distribution can cause wrong results. In the statistics file a range of shells that have to be calculated by the clients are saved. If not all shells and basis functions are in that range, there will be missing contributions in your calculations and therefore you will get wrong results!
If ired fails, try to generate the internal coordinates in a different way (this will produce 'pure' internal redundant coordinates, while the default is to generate a mixture of internal Pulay coordinates and internal redundant ones):
Open your control file and add (see chapter 12.3.2 of the documentation of Turbomole 5.10 or chapter 14.2.3 in the documentation of Turbomole 6.0):
$redund_inp
metric 1
and restart define. Try to do ired now. If this still crashes, PLEASE send the coordinates to the Turbomole support! We need inputs that cause problems to make the program more stable.
Regards,
Uwe
-
$redund_inp
metric 1
and restart define. Try to do ired now. If this still crashes, PLEASE send the coordinates to the Turbomole support! We need inputs that cause problems to make the program more stable.
Regards,
Uwe
Now ired works but the statistic run keeps crashing:
+--------------------------------------------------+
| basis set information |
+--------------------------------------------------+
we will work with the 1s 3p 5d 7f 9g ... basis set
...i.e. with spherical basis functions...
type atoms prim cont basis
---------------------------------------------------------------------------
o 10 46 31 def2-TZVP [5s3p2d1f|11s6p2d1f]
n 12 46 31 def2-TZVP [5s3p2d1f|11s6p2d1f]
c 60 46 31 def2-TZVP [5s3p2d1f|11s6p2d1f]
h 94 8 6 def2-TZVP [3s1p|5s1p]
---------------------------------------------------------------------------
total: 176 4524 3106
---------------------------------------------------------------------------
total number of primitive shells : 66
total number of contracted shells : 1278
total number of cartesian basis functions : 3516
total number of SCF-basis functions : 3106
integral neglect threshold : 0.95E-11
integral storage threshold THIZE : 0.10E-04
integral storage threshold THIME : 5
symmetry group of the molecule : c1
the group has the following generators :
c1(z)
1 symmetry operations found
there are 1 real representations : a
maximum number of shells which are related by symmetry : 1
mo occupation :
irrep mo's occupied
a 3106 308
number of basis functions : 3106
number of occupied orbitals : 308
this is the statistics output provided by subroutine count
concerning the number of batches and integrals to be evaluated or stored .
the integral evaluation threshold is "scftol",
the integral storage threshold is "thize",
and the integral storage threshold concerning
the integral evaluation cost is "thime" .
using the roothaan upper bond for estimating the values of
two-electron integrals the number of contributing batches (integrals) is
for thr = 0.95E-11 : 29159 ( 1416494) millions,
approximate number of non-symmetry redundant batches (integrals) :
334323 ( 19195663) millions.
the following table gives the number of stored batches (integrals)
as a function of THIZE and THIME :
THIZE : 0.10E-03 0.10E-04 0.10E-05
THIME *********************************************************************
45 125453113(2140938481) 210787651(-943132312) 317601173( 477187940)
40 158574434(**********) 272173717(-301990933) 416953810(1431999632)
35 177068236(**********) 302528565( 20307208) 461861626(1905994000)
30 207642475(-788802903) 353045112(1273233610) 537061599(-576176912)
25 221671832( 253943350) 375273363(**********) 569641646(1780892441)
20 352859984(1980597064) 608302234(1455575360) 934046548(1665842481)
18 399883220(**********) 693225395(**********) *********( -40676640)
16 498398082(1765200379) 857951642(-707352168) *********(**********)
14 553308319(-610647189) 950619278(**********) *********(**********)
12 735774064(-780083884) *********(**********) *********(-256128910)
10 735774064(-780083884) *********(**********) *********(-256128910)
9 789286383( 480484630) *********( 450886660) *********(**********)
8 *********( 625051784) *********(2058898984) *********(**********)
7 *********(1853510464) *********(1846851675) *********(1657356801)
6 *********(-874093192) *********(**********) *********(**********)
5 *********(-874093192) *********(**********) *********(**********)
4 *********(**********) *********(1301176129) 548279762(**********)
3 *********( -93208087) *********( 311779024) *********(-805201354)
2 *********(-241700812) 838641104( 702295475) *********(1757772149)
1 *********(**********) *********(**********) *********(**********)
for the actual (scftol,thime,thize)-combination
the file space required for the 2e integrals
is about ----> -10769273 kbytes <---- !
<dtasks> : still too many parallelized dscf tasks !
mxtask = ***
i,jshell = 885350 (running from *** --> 1)
costs already distributed : 2399130416439.5
(total costs : 4255397472978.0)
safety factor is 1.000
... and will be increased now to 1.874 !
<dtasks> : still too many parallelized dscf tasks !
mxtask = ***
i,jshell = 839 71 (running from *** --> 1)
costs already distributed : 2740167717434.0
(total costs : 4255397472978.0)
safety factor is 1.874
... and will be increased now to 3.097 !
<dtasks> : still too many parallelized dscf tasks !
mxtask = ***
i,jshell = 779547 (running from *** --> 1)
costs already distributed : 3101919300583.5
(total costs : 4255397472978.0)
safety factor is 3.097
... and will be increased now to 4.559 !
<dtasks> : still too many parallelized dscf tasks !
mxtask = ***
i,jshell = 709274 (running from *** --> 1)
costs already distributed : 3438449976621.5
(total costs : 4255397472978.0)
safety factor is 4.559
... and will be increased now to 6.098 !
<dtasks> : still too many parallelized dscf tasks !
mxtask = ***
i,jshell = 650643 (running from *** --> 1)
costs already distributed : 3640204126139.5
(total costs : 4255397472978.0)
safety factor is 6.098
... and will be increased now to 7.738 !
<dtasks> : still too many parallelized dscf tasks !
mxtask = ***
i,jshell = 600584 (running from *** --> 1)
costs already distributed : 3816003745491.9
(total costs : 4255397472978.0)
safety factor is 7.738
... and will be increased now to 9.403 !
<dtasks> : still too many parallelized dscf tasks !
mxtask = ***
i,jshell = 575561 (running from *** --> 1)
costs already distributed : 3893480362003.4
(total costs : 4255397472978.0)
safety factor is 9.403
... and will be increased now to 11.217 !
Loop overflow (safety factor); quitting
Try increasing mxtask (maximum number of integral tasks):
$parallel_parameters maxtask = <mxtask>
the default for mxtask is 400
MODTRACE: no modules on stack
<dtasks> : safety factor overflow - mxtask too small
dscf ended abnormally
I've then increased the number of maxtask to 10000
in the control file and finally got the metastase file!
Thanks Uwe!