Ceilidh ... Re^3: Different results of SA-MCSCF calculations between Firefly 8.0 and 8.0.1

Firefly and PC GAMESS-related discussion club

Learn how to ask questions correctly

We are NATO-free zone

Re^3: Different results of SA-MCSCF calculations between Firefly 8.0 and 8.0.1

Panwang Zhou
pwzhou@gmail.com

Dear Alex,
Dear Thom,

Both the recommendations can resolve the problem.

Thanks again!

On Wed Jan 15 '14 2:04pm, Alex Granovsky wrote
----------------------------------------------
>Dear Panwang,
>Dear Thom,

>if I understand the problem correctly, it is related to state
>tracking. Namely, with Firefly v. 8.0.1 state tracking remaps state
>#2 to state #6 while with older versions of Firefly it does not.
>The reason of this is most likely as follows. The first set of
>reference CI vectors is prepared after first CI procedure i.e.
>using original input MOs. With Firefly v. 8.0.1, CASSCF orbitals
>are transformed to diagonalize current density matrix during first
>MCSCF step, while previous versions of Firefly used other
>canonicalization schemes. As old MOs come from older version of
>Firefly, they will be rotated at the first MCSCF iteration.
>As a result, starting MCSCF using old converged orbitals
>may result to false state remaps caused by a sudden changes in
>orbitals and hence CI vectors.

>There are several workarounds to this problem.

>First, one can resort to old style behavior as suggested by Thom.

>Second, one can converge a single point MCSCF with state tracking disabled
>to get new, rotated orbitals which can be used to start geometry
>optimization (with tracking) with Firefly v. 8.0.1. Note, these
>orbitals will be most likely bad starting guess for Firefly v. 8.0.0
>due to exactly the same problem described above.

>Hope this helps.

>Kind regards,
>Alex Granovsky
>
>
>
>
>
>
>
>1. By default, Firefly v. 8.0.1 uses different orbitals
>
>
>
>On Wed Jan 15 '14 1:01pm, Thomas Pijper wrote
>---------------------------------------------
>>Dear Panwang,

>>The first thing that comes to mind is that the SOSCF converger in Firefly 8.0.1 is slightly modified. Could you try to specify

>>$MCSCF OLDSTL=.T. $END

>>to see if this gives you results equal to those obtained with FF 8.0.0 beta 38?
>>
>>
>>Kind regards,
>>Thom
>>
>>
>>
>>
>>
>>On Wed Jan 15 '14 10:32am, Panwang Zhou wrote
>>---------------------------------------------
>>>Dear all,

>>>Recently I try to optimize excited state geometry of a molecule with State-Averaged MCSCF using the following input file:

>>>! File created by MacMolPlt 7.4.3
>>>!--------- One needs extra precision -------------------
>>> $MOORTH NOSTF=.T. NOZERO=.T. SYMS=.T. SYMDEN=.T. SYMVEC=.T. $END
>>> $DET CVGTOL=1D-7 $END
>>> $TRANS CUTTRF=1D-12 $END
>>> $MCSCF ACURCY=1D-7 ENGTOL=1.0D-12 $END
>>> $MCSCF NTRACK=2 $END
>>>!--------------------------------------------------------
>>> $CONTRL SCFTYP=MCSCF RUNTYP=OPTIMIZE EXETYP=RUN MAXIT=50 ICHARG=-1
>>> MULT=1 FSTINT=.T. GENCON=.T. INTTYP=HONDO NOSYM=1
>>> ICUT=11 ITOL=30 WIDE=.T. $END
>>> $SYSTEM MWORDS=300 TIMLIM=60000.0 KDIAG=0 NOJAC=100 $END
>>> $SCF DIRSCF=.T. FDIFF=.F. NCONV=8 $END
>>> $P2P P2P=.T. DLB=.T. $END
>>> $TRANS MPTRAN=2 DIRTRF=.T. AOINTS=DIST ALTPAR=.T. MODE=112 $END
>>> $MCSCF CISTEP=ALDET FULLNR=.F. SOSCF=.T. MAXIT=100 ISTATE=2 $END
>>> $DET NCORE=49 NACT=14 NELS=16 NSTATE=6 WSTATE(1)=1,1 PURES=.T. $END
>>> $BASIS GBASIS=N31 NGAUSS=6 NDFUNC=1 NPFUNC=1 DIFFSP=.TRUE. $END
>>> $GUESS GUESS=MOREAD NORB=359 $END
>>> $STATPT OPTTOL=1D-5 NSTEP=300 HSSEND=.T. $END
>>> $FORCE NVIB=2 VIBSIZ=0.005 $END
>>> $DATA
>>>Title
>>>C1
>>> C 6.0 -2.6087103932 -1.4314769433 0.0000292048
>>> C 6.0 -3.9204191980 -1.0434074131 -0.0001971732
>>> C 6.0 -4.2977549172 0.3624062254 -0.0003897880
>>>.....
>>>--- OPTIMIZED MCSCF MO-S --- GENERATED AT 11:57:37 18-JAN-2013
>>>Title
>>>E(MCSCF)= -719.6079219077, 11 ITERS, E(NUC)= 989.8338921639
>>> $VEC
>>> 1 1 2.80121665418580100D-06 1.80963473506224100D-05 -1.72772751620890100D-05 1.26171697312041000D-06 -1.05735486369750900D-09
>>>................

>>>The MOs are read from a previous converged SA2-MCSCF calculations with Firefly 8 Beta38.
>>>Surprisingly, with Firefly 8.0.1, the following energys are obtained for the first cycle:

>>>ITER TOTAL ENERGY DEL(E) LAG.ASYMM. SQCDF MICIT DAMP
>>> 1 -719.607832495 -719.607832495 0.000021 6.846E-06 5 0.0000
>>> Warning: state # 1 (# 1) may have been lost
>>> Warning: state # 2 (# 6) may have been lost
>>> Warning: remapping target state # 2 to converged CI state # 6
>>> ----------START APPROXIMATE SECOND ORDER MCSCF----------
>>> 2 -719.549772183 0.058060312 0.012006 8.024E-03 1 0.0000
>>> Warning: state # 1 (# 1) may have been lost
>>> Warning: state # 2 (# 6) may have been lost
>>> Warning: remapping target state # 2 to converged CI state # 6
>>> 3 -719.553559929 -0.003787747 0.003057 3.419E-03 1 0.0000
>>> Warning: state # 1 (# 1) may have been lost
>>> Warning: state # 2 (# 6) may have been lost
>>> Warning: remapping target state # 2 to converged CI state # 6
>>> 4 -719.554053578 -0.000493649 0.001052 6.190E-04 1 0.0000
>>> Warning: state # 1 (# 1) may have been lost
>>> Warning: state # 2 (# 6) may have been lost
>>> Warning: remapping target state # 2 to converged CI state # 6
>>> 5 -719.554137039 -0.000083461 0.000657 5.460E-04 1 0.0000
>>> Warning: state # 1 (# 1) may have been lost
>>> Warning: state # 2 (# 6) may have been lost
>>> Warning: remapping target state # 2 to converged CI state # 6
>>> 6 -719.554182119 -0.000045079 0.000366 1.857E-04 1 0.0000
>>> Warning: state # 1 (# 1) may have been lost
>>> Warning: state # 2 (# 6) may have been lost
>>> Warning: remapping target state # 2 to converged CI state # 6
>>> 7 -719.554198750 -0.000016631 0.000233 2.111E-04 1 0.0000
>>> Warning: state # 1 (# 1) may have been lost
>>> Warning: state # 2 (# 6) may have been lost
>>> Warning: remapping target state # 2 to converged CI state # 6
>>> 8 -719.554209319 -0.000010569 0.000144 4.957E-05 1 0.0000
>>> Warning: state # 1 (# 1) may have been lost
>>> Warning: state # 2 (# 6) may have been lost
>>> Warning: remapping target state # 2 to converged CI state # 6
>>> 9 -719.554212371 -0.000003052 0.000081 1.474E-05 1 0.0000
>>> Warning: state # 1 (# 1) may have been lost
>>> Warning: state # 2 (# 6) may have been lost
>>> Warning: remapping target state # 2 to converged CI state # 6
>>> 10 -719.554213261 -0.000000890 0.000048 1.384E-06 1 0.0000
>>>........
>>>
>>>and the MCSCF energy will become difficult to converge and the optimization will stop after two cycles due to MSCSCF energy convergence problems.

>>>Then I swith to Firefly 8, and the following results are obtained:

>>>ITER TOTAL ENERGY DEL(E) LAG.ASYMM. SQCDF MICIT DAMP
>>> 1 -719.607832495 -719.607832495 0.000021 6.846E-06 5 0.0000
>>> ----------START APPROXIMATE SECOND ORDER MCSCF----------
>>> 2 -719.607832500 -0.000000005 0.000006 6.892E-09 1 0.0000
>>> 3 -719.607832501 -0.000000002 0.000002 3.929E-09 1 0.0000
>>> 4 -719.607832501 0.000000000 0.000001 4.392E-10 1 0.0000
>>> 5 -719.607832501 0.000000000 0.000000 4.014E-10 1 0.0000
>>> 6 -719.607832501 0.000000000 0.000000 2.758E-11 1 0.0000
>>> 7 -719.607832501 0.000000000 0.000000 8.581E-12 1 0.0000
>>> 8 -719.607832501 0.000000000 0.000000 5.537E-13 1 0.0000

>>> --------------------
>>> LAGRANGIAN CONVERGED
>>> --------------------

>>> FINAL MCSCF ENERGY IS -719.6078325015 AFTER 8 ITERATIONS

>>>This is normal because I just increase convergence precision.

>>>So, what changes have been done in Firefly 8.0.1 makes this problem and which result is correct?

>>>Thanks in advance.

>>>Best Regards!

>>>

Thu Jan 16 '14 8:40am

This message read 908 times