Questions about VCI rovibrational calculation

[Irikura, Karl K. Dr. (Fed)]

Dear Molpro experts,

 

I’m trying to simulate the IR absorption spectrum, including rotational envelopes, for a closed-shell, five-atom C2v molecule (MoO₂Cl₂).  (I’m able to use a ccsd/aug-cc-pVDZ-PP surface, but aug-cc-pVTZ-PP fails.)  I have two questions:

 

(1) Calculating the PES takes a long time.  Can the energy and dipole surfaces (from POLY) be saved to a file and read in a later calculation? 

(2) The calculation crashes after printing many similar lines about “Memory manager status”.  The last meaningful line is “Maximum rotational excitation:  JMAX =   199 (from Partition fun)”.  Is it merely that JMAX=199 is too large?

 

I will be grateful for any tips!

 

Thanks,

Karl

 

***,MoO2Cl2, CCSD/adz geom and VCI with combinations up to v=4

memory,2000,M;

 

gprint,orbitals,civector;

 

angstrom

geometry={

Mo

X  1 10.

O  1 moo  2 ao

O  1 moo  2 ao 3 180.

Cl 1 clmo 2 a  3 90.

Cl 1 clmo 2 a  3 -90.

}

 

MOO=                 1.72001638 ANGSTROM

AO=                 53.41700941 DEGREE

CLMO=                2.32987725 ANGSTROM

A=                 125.06887765 DEGREE

 

basis={DEFAULT=aug-cc-pvdz-pp,Cl=aug-cc-pv(d+d)z}

 

{hf}

ccsd

 

{optg,energy=1.d-08,gradient=1.d-06}

 

{frequencies,symm=auto}

put,molden,mpro.molden

 

label1

{hf

start,atden}

{ccsd; cphf,1}

 

{xsurf,start1D=label1,sym=auto; intensity,dipole=2}

{vibstate,combi=4}

poly

{vscf,info=1,pot=poly,print=1}

{vci,info=1,pot=poly,print=1;rovib,dump_ir="rovib.spec",temp=400,thrhotb=0.01}

put,irspec,irspec_rv.gnu

 

-------------------------------------------------------------------

Dr. Karl K. Irikura

National Institute of Standards and Technology

100 Bureau Drive, mail stop 8320

Gaithersburg, MD 20899-8320  USA

voice:  301-975-2510   fax:  301-975-3670

email:  karl.i...@nist.gov

http://www.nist.gov/mml/csd/informatics_research/index.cfm

-------------------------------------------------------------------

 

[rauhut]

Dear Karl,

once you switch to a multi-level scheme, you should be able to run the most important 1D terms with a triple-zeta basis. Now concerning your questions:

(1) Of course, just have a look at chapter 47 of the manual. The DISK directive (Restart capabilities) explains how to do it. Essentially you need a line like

      disk,where=home,dump='MoO2Cl2.pot' for dumping the surface and for restarting from it, you simply need to exchange the "DUMP" option by "EXTERN".

      The pot-file is in ASCII format and is easy to understand.

(2) Within your VCI calculation you request the calculation of very many combination bands (COMBI=4), which results in a huge basis for the subsequent 

      RVCI calculation. This is combination with a rather large value for Jmax will result in huge RVCI matrices and memory requirements. You have several

      options to fight this bottleneck: (a) You reduce the number of vibrational states either by using COMBI=3 or a lower value for UBOUND (in the VIBSTATE

      directive) or both. (b) Jmax has been determined from the partition function and usually is a pretty good estimate for a meaningful value for it. However,

      you can provide a value for Jmax manually and use lower values like 100. Most likely this will truncate some progressions too early, but you get an idea

      what to do. (c) Provide much more memory to this calculation (20 GByte ?), but of course the huge RVCI matrices will also need a lot of CPU time within

      the diagonalization and the subsequent calculation of billions of rovibrational transitions. Therefore, you better reduce the calculation.

Besides this, you lowered the threshold for hot bands, but did not request to compute them (HOTB=1).

Best wishes,

    Guntram