[molpro-user] Strange limit on RASSCF sizes - uncovered bug?

Dr Seth OLSEN s.olsen1 at uq.edu.au
Wed Jun 22 11:21:43 BST 2005


Hi Molpro-users,

I have been trying how much I can get into an RASSCF calculation with
MolPro.  My system is 126 electrons on 26 atoms.  I am interested in the
excited state surface, so I have been doing 2-state averaged calculations.  

I have changed the mxshel parameter in cconf and ccconf1 to 25 and
mrslr=25 in cref.  So far so good.  With this compile I was able to get
up to an RASSCF(8(mxhole=2),2,5(mxparticle=2)).  That is RASSCF(18,15)
which includes a core with only 2 excitations out of orbitals below
'LUMO' and only 2 excitations into orbitals above 'HOMO'.  This seems to
work fine and the resulting energies & natural orbitals look good and
are quite informative.  However, when I attempt to increase the number
of virtuals in this calculation to RASSCF(18,16) (with the identical
restrictions below/above the HOMO/LUMO pair), the job crashes with an
error.  The error seems to indicate a programming bug, and I have not
seen the parameter referred to before (see below).  It seems from the
multi source code to be connected with the p-space threshold, but I am
not certain.  I have attatched the full output.  Below is the output of
MULTI:

***
 1PROGRAM * MULTI (Direct Multiconfiguration SCF)       Authors: P.J.
Knowles, H.-J. Werner (1984)     S.T. Elbert (1988)
 
 
 Number of closed-shell orbitals: 54 ( 54 )
 Number of active  orbitals:      16 ( 16 )
 Number of external orbitals:    198 (198 )
 
 State symmetry 1
 
 Orbital occupancy restrictions for symmetry 1
  0      2       0000000000222222
 14     16       2222222200000000
 
 Reference space for state symmetry 1
 ------------------------------------
 Sym  Shell      Conf.             CSFs
  1     1       1 -     112       1 -     112
  1     2     113 -    1096     113 -    1096
  1     3    1097 -    2848    1097 -    4600
  1     4    2849 -    3268    4601 -    6700
 
 Number of electrons: 18    Spin symmetry=Singlet     Space symmetry=1
 Number of states:     2
 Number of CSFs:    6700
 
 Storage for configuration lists  69005, coupling coefficients  25887,
address lists 501608
  
 Molecular orbitals read from record     2140.2  Type=MATROP/MATROP
 
 Wavefunction dump at record             2141.2
 
 Convergence thresholds  0.10E-01 (gradient)  0.10E-05 (energy) 
0.00E+00 (step length)
 
 Weight factors for state symmetry  1:    0.50000   0.50000
 
 Eigenvalues of redundancy matrix
   1- 10   -0.000000   -0.000000   -0.000000   -0.000000   -0.000000  
-0.000000   -0.000000   -0.000000   -0.000000   -0.000000
  11- 20   -0.000000   -0.000000   -0.000000   -0.000000   -0.000000  
-0.000000   -0.000000   -0.000000   -0.000000   -0.000000
  21- 30   -0.000000   -0.000000   -0.000000    0.000000    0.000000   
0.000000    0.000000    0.000000    0.000000    0.000000
  31- 40    0.000000    0.000000    0.000000    0.000000    0.000000   
0.000000    0.000000    0.000000    0.000000    0.000000
  41- 50    0.000000    0.000000    0.000000    0.000000    0.245680   
0.339661    0.354046    0.580233    0.598297    0.598680
  51- 60    0.600180    0.602238    0.628619    0.651691    0.652339   
0.653234    0.654234    0.655112    0.657345    0.692562
  61- 70    0.693732    0.694992    0.696263    0.697557    0.700114   
0.755021    0.850798    0.873288    0.989323    0.994008
  71- 80    0.995801    0.997020    1.008652    1.044258    1.048045   
1.052341    1.056994    1.061901    1.069910    1.205660
  81- 90    1.206340    1.208259    1.214000    1.258507    1.398027   
1.441443    1.456479    1.468920    1.471006    1.474704
  91-100    1.479232    1.485733    1.486020    1.492871    1.493966   
1.496204    1.767194    1.798176    1.807558    1.820422
 101-110    1.827833    1.829082    1.836017    1.836184    1.840252   
1.845778    1.862532    1.863722    1.869921    1.871433
 111-120    1.879861    1.886974    1.889899    1.894577    1.894940   
1.902603    1.903470    1.911190    1.913630    1.921397
 
 Non-redundant active-active orbital rotations:
     1  2  3  4  5  6  7  8  9 10 11 12 13 14 15
  2
  3
  4
  5
  6
  7
  8
  9  +  +  +  +  +  +  +  +
 10  +  +  +  +  +  +  +  +
 11  +  +  +  +  +  +  +  +  +  +
 12  +  +  +  +  +  +  +  +  +  +
 13  +  +  +  +  +  +  +  +  +  +
 14  +  +  +  +  +  +  +  +  +  +
 15  +  +  +  +  +  +  +  +  +  +
 16  +  +  +  +  +  +  +  +  +  +
 
 Number of orbital rotations    14800     ( 864 Core/Active10692
Core/Virtual  76 Active/Active 3168 Active/Virtual)
 Total number of variables   =  28200
 
 
 ITER. MIC  NCI  NEG     ENERGY(VAR)     ENERGY(PROJ)   ENERGY CHANGE  
  GRAD(0)  GRAD(ORB)   GRAD(CI)     STEP       TIME
 
 
 ?PROGRAMMING ERROR: UNKNOWN PARAMETER THRPSP IN INPUT SET THRESH FOR
PROGRAM MULTI
 
 KNOWN PARAMETERS:
 ? Error
 ? Unknown parameter
 ? The problem occurs in get_inpf
 
 ERROR EXIT
 CURRENT STACK:      MAIN
 
*****

The parameter this error refers to, as above, seems connected to the
p-space threshold, but I am not certain.  I have not found the code line
that invokes the error yet, so I am not clear on what condition has been
made true to invoke the error.  In the relevant input I have not set the
p-space threshold, nor does it seem (from other tests) that the
threshold has any effect.  Presumably, these conditions would result in
other, more familiar errors anyway.  Is this a bug in the program?  Is
there another parameter which I must alter in order to make it work? 
Has my messing with the common variables mentioned above altered the
memory allocation scheme etc. in some fundamentally wrong way?  Should I
increase maxatoms?  I am running on a Sunfire v880 8cpu SMP enclosure. 
The memory requirements should be well within my limits.  Apart from the
parameters above, it is a standard tcgmsg parallel molpro install.  

I will continue probing this to see what I can do to bring it around,
and will post to the list if I can work it out.  However, if there is
someone on the list who is more familiar with the programming details of
molpro can lend some insight, that would be quite cool.  It seems that
molpro was written by more savvy coders than myself.

Cheers,

Seth

P.S.  I apologize for the brusqueness of yesterdays post: re the
geometry optimization error.  That was out of line & due to late-night
frayed nerves.  If the moderator wishes to strike it from the list, I
would not at all mind.  Upon review it (the error) seems to have come up
before.



ccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccms

Dr Seth Olsen, PhD
Postdoctoral Fellow, Computational Systems Biology Group
Centre for Computational Molecular Science
Chemistry Building,
The University of Queensland
Qld 4072, Brisbane, Australia

tel (617) 33653732
fax (617) 33654623
email: s.olsen1 at uq.edu.au
Web: www.ccms.uq.edu.au 

ccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccmsccms



-------------- next part --------------
A non-text attachment was scrubbed...
Name: m0car1816s2s.out
Type: application/octet-stream
Size: 52090 bytes
Desc: not available
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20050622/91c31dd7/attachment.obj>


More information about the Molpro-user mailing list