MRCI problem

[Timothy John Giese]

 >An MRCI calculation I am running hits a problem after few iterations
 >with an error message like:
 >> ?EXPECTATION VALUE AND EIGENVALUE NOT IDENTICAL STATE=1  EIG= ...

I have found myself in a similar situation.
The situation being the calculation of LiCl using SA-MRCI/aug-cc-pV5Z.
I receive the hereto error on an Origin 2000 processor
running IRIX64, but I do not receive the error on an
Intel P4 Xeon processor running Linux RedHat.

I have provided an example input file and information
regarding the computers below.

**Error occurs on this machine**
Version 2000.1 linked 28 Feb 2001 18:19:30
IRIX64-6.5/origin5(IP35)    64 bit Version 
                 DATE:  29-May-03         TIME: 21:48:57
More info at http://www.msi.umn.edu/origin

**Error does not occur on this machine**
Version 2000.1 linked 11 Nov 2002 16:12:51
Linux-2.4.18-27.7.xsmp/nf16(i686)    32 bit Version 
                 DATE:  29-May-03         TIME: 21:29:40
More info at http://ww.msi.umn.edu/netfinity

Sample file:
***,reset 
 
    proc runmrci
nogprint,variable
progx='mrci'
runmr
endproc
 

proc runacpf
nogprint,variable
progx='acpf'
runmr
endproc
 

proc runmr
hf
casscf;WF,20,1,0;STATE,2
istatx=0
spinx=spin
statx=state
symmx=symmetry
#symmetry=1
#state=1
#spin=1
spin=spinx(1)
state(1)=statx(1)
do isymx=1,#symmx
symmetry=symmx(isymx)
if(#statx.ge.isymx) state(1)=statx(isymx)
if(#spinx.ge.isymx) spin(1)=spinx(isymx)
ci;WF,20,1,0;REF,2;STATE,2,1,2;REFSTATE,2,1,2;OPTION,maxit=25,maxiti=50,nstati=2
save,6000.2+isymx
!saveresults
enddo
do jsymx=2,#symmx
do isymx=1,jsymx-1
mrci;trans,6000.2+isymx,6000.2+jsymx
enddo
enddo
spin=spinx
state=statx
symmetry=symmx
endproc

***,Input file generated by gabedit;
Gthresh,zero=1.d-27;      !Numerical zero
Gthresh,oneint=1.d-27;    !Threshold for one-electron integrals
Gthresh,twoint=1.d-27;    !Threshold for the neglect of two-elec. int.
Gthresh,preface=1.d-27;   !Threshold for test of prefactor in TWOINT
 

basis
Cl=aug-cc-pV5Z
Li=cc-pV5Z
end basis
 

memory,50,m
mem_dfock=50,m
 

rs=[4.0] 
                       do ir=1,#rs
   r=rs($ir)
     geometry={Li;Cl,Li,r}
     text,Li--Cl r=$r
     dip,,,-0.0005
     expec,dm,qm
     runmrci
     SHOW[6,8F24.18],r,energy,dmx,dmy,dmz
enddo

Cheers,
Tim Giese

University of Minnesota
giese025 at tc.umn.edu