Firefly and PC GAMESS-related discussion club

 
Learn how to ask questions correctly  
 
 

 

Thomas
thomaspijper@hotmail.com

To the best of my knowledge the TZV basis should be used with Cartesian d functions. So, it is not necessary to specify the D5 option (at least, as long as only s, p, and d functions are used).


Kind regards,
Thom



On Sat Dec 22 '12 9:30pm, Alex Granovsky wrote
----------------------------------------------
>Dear Matthieu,

>As to Mulliken and Lowdin population analyses, they do not work
>reliable for basis sets with lots of diffuse functions.
>In particular, the basis set you are using contains lots of (unneeded)
>diffuse functions and is quasi linearly dependent.
>This makes both Mulliken and Lowdin population analysis almost senseless.
>This also requires extra precision throughout all calculations
>to avoid some artifacts and numerical instabilities caused by these
>dependencies

>As to XMCQDPT2 properties and transition dipoles, you need to take
>into account that the computed values are the zero-order QDPT
>properties. This means that they include only a part of all effects
>due to interaction with the secondary space. Namely, they correctly
>include effects causing the rotation and intermixing of zero
>order states (i.e. CASCI vectors) within the model space but do
>not include the first order correction to wavefunction (i.e. the
>part of perturbed states which belongs to the secondary space).
>Thus these properties should be considered as the approximations
>to the "true" XMCQDPT2 properties.

>As to calculations with 10 and 20 states, it is easy to understand
>why you observed the picture you have reported.

>E.g. if you look at eigenvectors of the effecive hamiltonian for NSTATE=10 calculations, you'll see:

>

 EIGENVECTORS OF THE EFFECTIVE HAMILTONIAN

                      1          2          3          4          5

               -286.840923-286.681744-286.644400-286.613064-286.607064

    1            -0.992457   0.000006   0.048901  -0.000006   0.011915
    2             0.000003   0.925328  -0.000055  -0.367836  -0.000087
    3             0.049198   0.000078   0.943674   0.000100  -0.163528
    4             0.000007  -0.228667   0.000129  -0.657996  -0.000181
    5             0.015554   0.000044   0.215843   0.000090  -0.041182
    6             0.030684  -0.000013   0.168242  -0.000288   0.980470
    7             0.000000   0.253590   0.000003   0.631977   0.000230
    8             0.000009  -0.164838  -0.000112  -0.179840  -0.000024
    9            -0.015825  -0.000113   0.176117  -0.000128  -0.031814
   10             0.105713  -0.000006  -0.034310   0.000015  -0.095325

>that S1 state obtained at CASSCF level at PT2 level strongly
>interacts with S3, S6, and S7 states. Similarly, the zero-order
>S2 state is intermixed with S4, S5, and S8 states. These are
>the effect you cannot capture in NSTATE=3 computations. You really
>need to perform calculations with extended effective Hamiltonians
>to capture these important effects.

>A few additional comments. With Firefly version you are using,
>the weights of states in constructing averaged density matrices
>will be equal for all nstate=10 or 20 states (as indicated in the
>output):

>

 *** WEIGHTS OF STATES IN COMPUTING AVERAGED DENSITY MATRIX (AVECOE)
 -----------------------------------------------------------------------
                   1            2            3            4            5
     1  1.000000D+00 1.000000D+00 1.000000D+00 1.000000D+00 1.000000D+00
 -----------------------------------------------------------------------
                   6            7            8            9           10
     1  1.000000D+00 1.000000D+00 1.000000D+00 1.000000D+00 1.000000D+00
 ---------------------------------------------------------------------
--


This is most likely not exactly the calculations you want to perform.

You thus need to either set all other elements of avecoe and wstate arrays of $xmcqdpt2 group to zero explicitly, or to use the following syntax:

 wstate(1)=1,1,1,-0
 avecoe(1)=1,1,1,-0
>Next, the irot=1 settings correspond to the XMCQDPT2' variant 
>of theory as designated in my paper on XMCQDPT. Do you have some 
>good reasons to use this approximation to the exact theory? If not, 
>I'd recommend leave this option intact. 
>Finally, you may need to use D5 option with the basis set you have 
>selected (it is most likely usually used as spherical basis set, 
>I do not remember exactly at moment).
>I'm attaching the corrected input file for your reference. 
>Hope this helps.
>Kind regards,
>Alex Granovsky
>
>
>
>
>
>
>
>
>On Wed Dec 19 '12 4:45pm, Matthieu Sala wrote
>---------------------------------------------
>>Dear Firefly users, 
>>I am studying the photochemistry of the aniline molecule (Ph-NH2) using the XMCQDPT2 method, and I have trouble to obtain accurate values for the oscillator strengths, partial charges and dipole moments in the two first pipi* excited states.
>>In a previous CASSCF study, the first pipi* state has been assigned as a local excitation while the second as a charge transfer state with ca. 0.2 electron transfered from the N atom to the C atom at the opposite part of the ring. 
>>Moreover, accurate values (in comparison with experiments) for the oscillator strength has been reported at the SAC-CI, TD-CAMB3LYP and EOM-CCSD levels of theory. Experimental values are 0.028 for the 1pipi* and 0.144 for the 2pipi* states. 
>>I use a minimal active space of 8 el in 7 orb (3pi, 3pi* and the N lone pair) to describe excited pipi* states. I use a SA3-CASSCF reference wavefunction.
>>Then I performed XMCQDPT2 calculations with increasing size of the model space (3, 10 and 20). 
>>I have several problems. First Mulliken and Lowdin charge analysis give very different results (I think the Lowdin is more reasonnable).
>>Second the dipoles, oscillator strength and partial charges change a lot when the model space size is increased. For NSTATE=3 in XMCQDPT2, the results are quite close from what we can expect (although the oscillator strength for the 1pipi* state is significantly underestimated).
>>But when NSTATE is increased, the results change a lot. The oscillator strengths are greatly overestimated and the first pipi* state seems to acquire a charge transfer character.
>>Could someone tell me what am I doing wrong ?
>>Any help would be greatly appreciated.
>>I attach the inputs and output of the XMCQDPT2 calculations 
>>(all merged in one text file).
>>Best regards.
>>Matthieu Sala

[ This message was edited on Sat Dec 22 '12 at 11:07pm by the author ]