## Firefly and PC GAMESS-related discussion club

Learn how to ask questions correctly

Re^5: How to localize Orbital

Neeraj Kumar
nkkchem@gmail.com

Dear Prof.
Your assistance help me a lot, i appreciate your support regarding my question. But one thing is not clear yet for me that how can i rotate orbital ?? like i have orbital no. 113(HOMO) and 114(LUMO) from which wana  major contribution from configuration 22221100000. But How  i can rotate these two orbital at 45 degree angle in my input. I mean what i should  need to write in input command in order to get rotation between two these twoo orbital.  Could you please suggest me.  Thanks
With best regards
Neeraj Kumar

On Thu Mar 18 '10 0:58am, Alex Granovsky wrote
----------------------------------------------
>Hi,

>there is nothing wrong with computations. Remember, the CASSCF
>wavefunction is invariant with respect to arbitrary rotations of
>orbitals forming the active space. Let me explain what happened
>using simpler model, namely (2,2) active space. CAS(2,2) is also
>called TC-SCF (Two Configuration SCF), and can be handled by GVB
>code as well. So, why is it called TC-SCF while there are three
>CSFs: (core)20, (core)02, and (core)11 ? The answer is that there
>is exactly one orbital rotation (orbital mixing) parameter in this
>case - and it can be selected so that the coefficient of (core)11
>contribution in the CI expansion will become identically zero.
>However, if you rotate this pair of orbitals by e.g., 45 degrees,
>the contribution of (core)11 CSF will be close to its maximum
>possible value.

>The same explanation is true in your case. Indeed, most likely
>you deal with the biradical state, however, using NOs you get
>the CSF expansion that has small contribution of (core)11 CSF.

>To maximize (core)11 contribution, you need to rotate these
>two orbitals. Unfortunately, Firefly cannot make this for
>you automatically. However, I doubt if this is really necessary -
>the wavefunction is invariant so that any observable properties
>should be identical regardless on the particular form of orbitals
>and SCF expansion.

>Hope this helps.

>Regards,
>Alex Granovsky
>
>
>
>
>
>
>
>
>
>
>
>On Tue Mar 16 '10 8:10am, Neeraj Kumar wrote
>--------------------------------------------
>>Dear Granovskky
>>                Thanks for your suggestion for localized orbital, i appreciate your help.I converged natural Orbital by using guess=moread and MCSCF as you have suggested. Now instead of singly excitation it gave me a more contribution from doubly excited stat.. which is something strange...My final contribution is

>> FINAL MCSCF ENERGY IS    -2555.6631563525 AFTER   2 ITERATIONS

>> -MCCI- BASED ON OPTIMIZED ORBITALS
>> ----------------------------------

>> STATE #    1  ENERGY =   -2555.663156353

>>      CSF      COEF    OCCUPANCY (IGNORING CORE)
>>      ---      ----    --------- --------- -----
>>        1    0.752319  22222000000
>>     3534   -0.054016  22202002000
>>     3538   -0.053914  22022002000
>>    47125   -0.618726  22220200000
>>I am expecting more contribution from term 22221100000 but this convergence  is not showing contribution from this term, i dont understand this part....Because my system has diradical character which has been shown by MCSCF convergence. Do you suggest me any alternative way to get major contribution from this 22221100000 term.
>>here is my input for NO
>> \$CONTRL SCFTYP=MCSCF RUNTYP=ENERGY MAXIT=200
>>  MULT=1 ICHARG=0 d5=1 EXETYP=RUN
>>  FSTINT=.TRUE. GENCON=.TRUE. \$END
>> \$SYSTEM TIMLIM=100000000000 MWORDS=400 NOJAC=100
>>  KDIAG=0 MKLNP=2 \$END
>> !\$BASIS GBASIS=n31 NGAUSS=6 NDFUNC=1 \$END
>> \$D5 d5=.TRUE. f7=.TRUE. \$END
>> \$SCF DIRSCF=.TRUE. DIIS=.TRUE. \$END
>> \$GUGEM PACK2=.TRUE. FASTCI=.TRUE. \$END
>> \$CIINP CASTRF=.TRUE. \$END
>> \$TRANS MPTRAN=2 DIRTRF=.T. AOINTS=DIST ALTPAR=.T. MODE=112 \$END
>> \$DRT GROUP=C1
>>      FORS=.TRUE.
>>      NMCC=108
>>      NDOC=5
>>      NVAL=6
>> \$END
>> \$GUGDIA NSTATE=1 ITERMX=100 \$END
>> \$MCSCF CISTEP=GUGA MAXIT=50 ENGTOL=1.0E-07 CASHFT=2.0 \$END
>> \$DATA
>>C1
>>\$VEC......
>>\$END

>>Thanks
>>With warm regards
>>Neeraj Kumar
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>

>>On Sat Mar 13 '10 11:49pm, Alex Granovsky wrote
>>-----------------------------------------------
>>>Hi,

>>>most likely you do not need localized orbitals at all...
>>>I'm afraid they will mix up coefficients of CI expansion
>>>rather than sum up to the single leading term.

>>>What I'd suggest is to look at CI expansion based on natural orbitals.
>>>You can find them in the punch file - just provide them
>>>using guess=moread and run CI or MCSCF. Note you do not
>>>actually need to converge MCSCF from the scratch - it will
>>>converge in a couple of iterations as NOs are already optimized.

>>>By the way, NOs themselves tend to be rather local in the space -
>>>you may want to look at them using e.g. ChemCraft. To some extent,
>>>transformation to NOs can be considered as a kind of orbital localization procedure. It also tends to make CI expansion more compact.

>>>Hope this helps.

>>>Regards,
>>>Alex Granovsky
>>>
>>>
>>>
>>>
>>>
>>>On Thu Mar 11 '10 6:46pm, Neeraj Kumar wrote
>>>--------------------------------------------
>>>>Can anybody help me.For localization of orbital in MCSCF
>>>>convergence....
>>>>
>>>>
>>>>
>>>>
>>>>On Thu Mar 11 '10 0:24am, Neeraj Kumar wrote
>>>>--------------------------------------------
>>>>>Hello
>>>>>    i am new to firefly, I have converged (10,11) active space for my
>>>>>model(having 45 atom) using MCSCF.And My system is closed shell and it
>>>>>has C1 sym. My overall wave-function has complicated form because it
>>>>>is multi reference. Now i came to now that this wave-function
>>>>>substantially simplified by a transformation to localized
>>>>>orbitals( because this operation does not effect the the overall wave-
>>>>>function).My model has total 113 occupied orbital. So my wavefunction
>>>>>has contribution from following major term {113(HOMO) and 114(LUMO)}.
>>>>> 0.555943 (22222000000) + 0.380879 (22221100000) -0.250496
>>>>>(22220200000)
>>>>>So instead of three term, i wana single dominant term with more weight
>>>>>in the total wave function
>>>>>What I dont know how to use \$Local and moreover its unclear from
>>>>>manual. If any body has experienced with transformation to localized
>>>>>or if you have any input file please send me.
>>>>>Here is the input for my system which is converged by  MCSCF.
>>>>> \$CONTRL SCFTYP=MCSCF RUNTYP=ENERGY MAXIT=200
>>>>>  MULT=1 ICHARG=0 d5=1 EXETYP=RUN
>>>>>  FSTINT=.TRUE. GENCON=.TRUE. \$END
>>>>> \$SYSTEM TIMLIM=100000000000 MWORDS=300 NOJAC=100
>>>>>  KDIAG=0 MKLNP=2 \$END
>>>>> \$D5 d5=.TRUE. f7=.TRUE. \$END
>>>>> \$SCF DIRSCF=.TRUE. DIIS=.TRUE. \$END
>>>>> \$GUGEM PACK2=.TRUE. FASTCI=.TRUE. \$END
>>>>> \$CIINP CASTRF=.TRUE. \$END
>>>>> \$TRANS MPTRAN=2 DIRTRF=.T. AOINTS=DIST ALTPAR=.T. MODE=112 \$END
>>>>> \$DRT GROUP=C1
>>>>>      FORS=.TRUE.
>>>>>      NMCC=108
>>>>>      NDOC=5
>>>>>      NVAL=6
>>>>> \$END
>>>>> \$GUGDIA NSTATE=1 ITERMX=100 \$END
>>>>> \$MCSCF CISTEP=GUGA MAXIT=50 \$END
>>>>> \$DATA

Sat Mar 20 '10 0:41am