Author Topic: relax problem  (Read 8503 times)

175116

  • Jr. Member
  • **
  • Posts: 13
  • Karma: +0/-0
relax problem
« on: February 13, 2009, 09:38:14 pm »
Hello! I am using Turbomole 5.10 and I have a problem with relax step during optimization. I have got an 43 atomic system (symmetry C1) and I need to optimize it using ridft/tpss/6-311++G(3df,3pd) with six torsion angles being fixed. I use a DEFINE programme and define six internal coordinates and mark them as fixed. I've used this script:

Code: [Select]
#!/bin/bash

x2t geometry.xyz > coord

define << EOF


a coord
i
idef
f tors 9 8 7 3
f tors 10 9 8 7
f tors 11 10 9 8
f tors 12 11 10 9
f tors 16 12 11 10
f tors 17 16 12 11


iaut


q

imet


*
b all 6-311++G(3df,3pd)
*
eht

-1

dft
on
func tpss

ri
on
m400

q
EOF

The optimization fails during the relax step with this warning (job.2 file) :

Code: [Select]
relax(t28) : TURBOMOLE V5-10 14 Jan 2008 at 15:57:28
 Copyright (C) 2008 University of Karlsruhe


    2009-02-13 19:39:41.035



                           this is  R E L A X


             conversion cartesian <---> internal coordinates
          coordinate and force constant update in optimizations

           by heinz schiffer, hans horn, christoph koelmel,
          marco haeser, michael ehrig, and reinhart ahlrichs

                          quantum chemistry group
                          universitaet  karlsruhe
                                  germany



------------------------------------------------------------------------
 $optimize

 optimization will be performed with respect to
 - CARTESIAN coordinates

------------------------------------------------------------------------
 $coordinateupdate

 upper limit for coordinate changes =   0.3000   
 interpolation/extrapolation has been enabled
 display optimization statistics for the last   5 cycles
 
 MODTRACE: no modules on stack

 Cannot do a Cartesian step with fixed internals
 relax ended abnormally
relax step ended abnormally
next step = relax


and this is the output from the jobex.out file:

Code: [Select]
convgrep will be taken out of the TURBODIR directory
OPTIMIZATION CYCLE 1
call define for new internals
OPTIMIZATION CYCLE 2
call define for new internals
run relax for a cartesian step
ERROR: Module relax failed to run properly - please check output job.2 and job.last for the reason

Can you please help me? I really do not know what is the problem. Thank you very much indeed.

uwe

  • Global Moderator
  • Sr. Member
  • *****
  • Posts: 405
  • Karma: +0/-0
Re: relax problem
« Reply #1 on: February 14, 2009, 09:56:38 pm »
Hi,

I am not sure what the problem is either, but it seems that the internal coordinate definition is either incomplete or gets linear dependent during optimization. jobex usually switches to Cartesian coordinates automatically in such cases, trying to re-define the internal coordinates after a couple of cycles. But in case of fixed internal coordinates, jobex can not decide which atoms to fix (without fixing too many degrees of freedom), so it stops.

Could you try to use internal redundant coordinates instead? So call 'ired' after defining the fixed internal coordinates instead of 'iaut'. This might help.

Regards,

Uwe

175116

  • Jr. Member
  • **
  • Posts: 13
  • Karma: +0/-0
Re: relax problem
« Reply #2 on: February 15, 2009, 10:56:06 am »
Hi,

I am not sure what the problem is either, but it seems that the internal coordinate definition is either incomplete or gets linear dependent during optimization. jobex usually switches to Cartesian coordinates automatically in such cases, trying to re-define the internal coordinates after a couple of cycles. But in case of fixed internal coordinates, jobex can not decide which atoms to fix (without fixing too many degrees of freedom), so it stops.

Could you try to use internal redundant coordinates instead? So call 'ired' after defining the fixed internal coordinates instead of 'iaut'. This might help.

Regards,

Uwe

Thank you very much Uwe! It seems that redundant coordinates usage is the key. Job number 4 is running right now and all six torsions appear to be fixed.