CCL:G: Low-progression Franck-Condon transitions

 Sent to CCL by: tianxiaohui/./zju.edu.cn
 Dear Julien,
 Thank you for your previous explanations.
 Some of my molecules show large displacements. I'd like to know could I use
 AH+zmatrix in Gaussian16 now? I did z-matrix frequency calculations in
 Gaussian16, but the TD result is still the same with AH+cartesian and I don't
 know why.
 Thank you for your kind help.
 > -----原始邮件-----
 > 发件人: "Tobias Kraemer Tobias.Kraemer=-=mu.ie"
 <owner-chemistry\a/ccl.net>
 > 发送时间: 2018-11-12 19:32:50
 (星期一)
 > 收件人: "Tian, Xiaohui "
 <tianxiaohui\a/zju.edu.cn>
 > 抄送:
 > 主题: CCL:G: Low-progression Franck-Condon transitions
 >
 >
 > Sent to CCL by: Tobias Kraemer [Tobias.Kraemer(a)mu.ie]
 > Dear Julien,
 >
 > Thank you for your additional explanations and suggestions. There is plenty
 of things to try out now. This will take some time,
 > If there a further questions I will turn to you again.
 >
 > Kind regards,
 >
 > Tobias
 >
 > Dr. Tobias Krämer
 > Lecturer in Inorganic Chemistry
 > Department of Chemistry
 >
 > Maynooth University, Maynooth, Co. Kildare, Ireland.
 > E: tobias.kraemer-.-mu.ie   T: +353 (0)1 474 7517
 >
 >
 >
 > > Sent to CCL by: "Julien Bloino" [julien.bloino:_:gmail.com]
 Dear Dr.
 > > Krämer,
 > >
 > > As commented by Dr. Götze, the likely reason for the small
 progression
 > > is a significant shift of one or more modes.
 > > The keywords are for G16 and should be inserted in the `ReadFCHT`
 > > section (Freq=ReadFCHT).
 > > You can see the shift vector with
 > > `Print=Matrix=K`
 > > Depending on the symmetry and the structural changes, you may improve
 > > the convergence by increasing the maximum number of quanta for the
 > > overtones (MaxC1) and 2-modes combinations (MaxC2):
 > > `Prescreening=(MaxC1=20,MaxC2=13)` (those are the default values.)
 > > This is rarely sufficient to fix the convergence issue and you may
 > > want to check the presence of low-frequency large amplitude modes
 > > (large shift of low-energy modes) and potentially exclude them as they
 > > are poorly treated with this model.
 > > You can do them with:
 > > `RedDim=Block`
 > > followed by the list of modes to exclude (the reference state is the
 > > lower state, so you will have to list the modes to exclude from the
 > > initial state, compatible with the definition of the shift vector K).
 > > Note that Gaussian will try to build a consistent set of modes (same
 > > number in each state) to exclude from the vibronic treatment. It has a
 > > safety check to stop if too many modes are selected this way compared
 > > to the initial list. You can force it by changing the value of
 > > `RedDim=BlockTol` The definition of the set of modes to exclude is
 > > based on the Duschinsky matrix, which can be printed with,
 > > `Print=Matrix=J` (we generally always print both J and K with
 > > `Print=Matrix=JK`)
 > > Be careful in the truncation as the model system obtained this way may
 > > not be representative of the full system anymore.
 > >
 > > To obtain a fully converged spectrum, you can use the time-dependent
 > > formalism instead of the sum-over-states one (the default in this
 > > case) with the option `TimeDependent` I would recommend to use first
 > > TI to setup your protocol (trying the options described above) and
 > > once a sufficient convergence is reached, use TD to obtain the full
 > > band-shape. Indeed, the breakdown of the Franck-Condon approximation
 > > has a direct impact on a TI calculations (low and slow convergence)
 > > but is difficult to detect within the TD framework (the spectrum is
 > > always fully converged by definition).
 > >
 > > Regarding ForcePrtSpectrum, the option (and all "advanced"
 options) is
 > > still there but I chose not to document it as it is a double-edged
 > > sword and could be misinterpreted. There are technically 2 separate
 > > checks but the one you mention will not be helpful in your case (at
 > > least in a first time).
 > > - the first test is on the overall convergence after 2-modes
 > > combinations. If it is below 20% (which is the case here), Gaussian
 > > will stop. You can override this with `Advanced=ForceFCCalc`
 > > - the second test is at the end of the calculations, before printing
 > > the spectrum. If the progression is below 50%, the spectrum is not
 printed.
 > > You can override this with
 > > `Advanced=ForcePrtSpectrum`
 > >
 > > Regarding the description of the potential energy surfaces, Gaussian
 > > supports AdiabaticHessian (AH, the default), AdiabaticShift (AS),
 > > VerticalHessian (VH, also noted VFC) and VerticalGradient (VG, aka LCM
 > > or IMDHO). From my understanding, a behavior similar to IMDHO-FA would
 > > be obtained in Gaussian with `VerticalHessian
 > > DataMod=Duschinsky=Identity` Simplified models (like VG) should be
 > > used with care. While it is easier to reach convergence with them,
 > > they can also misrepresent the actual system, leading to incorrect
 > > spectra. The validity of such approximation will depend on your
 > > system.
 > >
 > > I hope this will answer your questions regarding the progression and
 > > keywords.
 > >
 > > Best regards,
 > >
 > > Julien Bloino
 > >
 > > ------ Original Message ------
 > > > From: "Tobias Kraemer Tobias.Kraemer!A!mu.ie"
 > > > <owner-chemistry],[ccl.net>
 > > To: "Bloino, Julien " <julien.bloino],[gmail.com>
 > > Sent: 2018-10-30 03:36:55
 > > Subject: CCL:G: Low-progression Franck-Condon transitions
 > >
 > > >Dear Jan,
 > > >
 > > >
 > > >thanks for your reply. Sorry for being so unspecific in my post, I
 > > >thought this was a more generic error that could be solved more
 easily.
 > > >You are right about the fact that the geometries of the ground and
 > > >excited state of this ZnPc complex differ (not too a large extend,
 > > >but obviously enough). The ground state is planar with D4h
 symmetry,
 > > >while the structure of the (1st) excited state converges to a
 > > >C2v-symmetric geometry (consistent with literature J. Chem. Phys.,
 > > >2015, 142, 094310). In fact the white paper by Barone
 > > >"Vibrationally-excited states in Gaussian09" mentions
 the distortion
 > > >of the excited state geometry away from a planar geometry in the
 > > >ground state can cause problems (and FC does not apply). However,
 > > >since the aforementioned paper in J. Chem. Phys. presents a FC
 > > >spectrum, I believe that it must still be possible to generate the
 > > >spectrum, and find a way around this issue. I should also mention
 > > >that by visual inspection the excited state geometry is not hugely
 > > >different from the ground state (but obviously large enough to
 cause
 > > >a problem). It seems in G09 one could force the plot of a spectrum
 > > >nonetheless, via FORCEPRTSPECTRUM. My question was also regarding
 a
 > > >range of other keywords that might be useful here
 > > >(MAXBANDS/MAXC1/MAXOVR..). So the question still stands, since I
 think it must be possible to solve this issue.
 > > >
 > > >
 > > >Nonetheless, I might try one of your suggestions as well, thanks
 for
 > > >pointing me in this direction.
 > > >
 > > >
 > > >Best,
 > > >
 > > >
 > > >Tobias
 > > >
 > > >
 > > >
 > > >
 > > >Dr Tobias Krämer
 > > >
 > > >Lecturer in Inorganic Chemistry
 > > >
 > > >Department of Chemistry
 > > >
 > > >Maynooth University
 > > >
 > > >[Maynooth University PNG Trans]
 > > >
 > > >Maynooth University, Maynooth, Co. Kildare, Ireland.
 > > >
 > > >E: tobias.kraemer|-|mu.ie T: +353 (0)1 474 7517
 > > >
 > > >________________________________
 > > >>From: owner-chemistry+tobias.kraemer==mu.ie|-|ccl.net
 > > >><owner-chemistry+tobias.kraemer==mu.ie|-|ccl.net> on
 behalf of Jan
 > > >>Götze jgoetze[]zedat.fu-berlin.de
 <owner-chemistry|-|ccl.net>
 > > >Sent: Saturday, October 27, 2018 4:25:23 PM
 > > >To: Tobias Kraemer
 > > >Subject: CCL:G: Low-progression Franck-Condon transitions
 > > >
 > > >
 > > >Sent to CCL by: =?UTF-8?Q?Jan_G=c3=b6tze?=
 > > >[jgoetze##zedat.fu-berlin.de] Dear Tobias,
 > > >
 > > >the data you provided only allow for limited analysis why your
 proble
 > > >occurs. In case you did not do any errors in preparation of your
 two
 > > >excited states, it appears that the minima of ground and excited
 > > >state are very distant from each other (such as groups rotating,
 > > >and/or normal modes differing strongly between ground and excited
 > > >state). For a large, planar, aromatic system like pc this is
 rather
 > > >unusual. As such, without further details on the molecular
 structure,
 > > >any additional help can only be guesswork.
 > > >
 > > >To obtain a preliminary spectrum quickly and often without
 problems,
 > > >I personally would suggest using a vertical TD approach, which
 might
 > > >be available in Gaussian16, or an IMDHO-FA as in ORCA. See for
 > > >example doi:10.1021/ct500830a
 > > >
 > > >Cheers,
 > > >Jan
 > > >
 > > >Am 26.10.2018 um 12:57 schrieb Tobias Kraemer
 tobias.kraemer[a]mu.ie:
 > > >>Sent to CCL by: "Tobias  Kraemer"
 [tobias.kraemer_._mu.ie] Hello
 > > >>everyone,
 > > >>
 > > >>I am interested in calculating vibrationally-resolved spectra
 in G16.
 > > >>The
 > > >>molecule in question is a phthalocyanine (pc) complex. I've
 followed
 > > >>the protocol detailed in the whitepaper by Barone et al.,
 however in
 > > >>the final step (generating the spectrum) an error occurs:
 > > >>
 > > >>
 > > >>       ==================================================
 > > >>                Calculations of Band Intensities
 > > >>       ==================================================
 > > >>
 > > >>   -- To: vibronic fundamental state --
 > > >>      Spectrum progression:    0.06%
 > > >>
 > > >>   -- To: single overtones --
 > > >>      Spectrum progression:    0.71%
 > > >>
 > > >>   -- To: combinations of  2 simultaneously excited modes --
 > > >>      Spectrum progression:    4.14%
 > > >>
 > > >>   ERROR: Low progression after class 2. Total convergence =
 4.1%.
 > > >>          The vibronic spectrum will likely be unreliable.
 Stopping.
 > > >>
 > > >>The whitepaper provides some possible causes, but I'd like to
 ask
 > > >>for some expert opinions here on CCL nonetheless. In the
 excited
 > > >>state optimisation I have included 6 states, of which the
 gradients
 > > >>for the first one are to be followed
 [TD=(Read,NStates=6,Root=1)].
 > > >>There are a good number of keywords listed on the Gaussian16
 webpage
 > > >>that relate to this type of calculation, and I'd appreciate
 some
 > > >>guidance on the above issue and possible ways around it.
 > > >>
 > > >>Thanks for your help, as always much appreciated.
 > > >>
 > > >>Kind regards,
 > > >>
 > > >>Tobias>