CCL:G: Problems with oscillator strengths in TDDFT

 Sent to CCL by: Arvydas Tamulis [tamulis%a%mserv.itpa.lt]
 Dear Colleagues,
 Enclosed please find summary of your answers to my request:
 Problems with oscillator strengths in TDDFT.
 Thank you very much for your advises!
 Best regards,
 Arvydas Tamulis
 *********************************************
 Date: Wed, 4 Jul 2007 12:50:30 +0200
 
 From: Filipp Furche <filipp.furche-#-chemie.uni-karlsruhe.de>
 
 To: Arvydas Tamulis <tamulis-#-mserv.itpa.lt>
 Subject: Re: Problems with oscillator strengths in TDDFT
 Dear Professor Tamulis,
 Thanks for your message. I would recommend to re-do the calculations
 without symmetry, and see if the results are identical to the D3
 case.
 LanL2DZ is a small basis set. Reinhart Ahlrichs and Florian Weigend
 have recently developed triple and quadruple zeta basis sets for heavy
elements with small-core pseudopotentials [Phys. Chem. Chem. Phys. 7 (2005), 
 3297]. You may want to use those for comparison.
 I personally think it's possible that the oscillator strengths of the
 two lines are reversed in TDDFT. The relative intensities depend strongly
 on the coupling between the two states, which is very sensitive to
 inaccuracies of the functional but also to basis set error or solvent
 effects.
 If you are interested in an efficient TDDFT code I recommend TURBOMOLE:
 www.turbomole.com. I am attaching a review that contains some benchmark
 data.
 Sincerely,
 Filipp Furche
 PD Dr. Filipp Furche
 Institut fr Physikalische Chemie
 Lehrstuhl fr Theoretische Chemie
 Universitt Karlsruhe (TH)
 Kaiserstrae 12
 D - 76128 Karlsruhe
 Germanyphone: +49/721/608-7227
 fax  : +49/721/608-7225
 email: Filipp.Furche-#-chemie.uni-karlsruhe.de
 ************************************************
 Date: Wed, 04 Jul 2007 13:24:07 +0200
 
 From: Mark E. Casida <mark.casida-#-ujf-grenoble.fr>
 
 To: Arvydas Tamulis <tamulis-#-mserv.itpa.lt>
 Cc: Mark E. Casida <mark.casida-#-ujf-grenoble.fr>
 Dear Arvydas,
 As is often the case when questions are sent to me regarding real or
 supposed
 failures of TDDFT, the information you are sending me leaves me perhaps
 as confused as you are yourself. You tell me that you have done the same
 calculation with 2 program packages and gotten the same answer and then
 ask me what is wrong with the programs!!?? Permit me to say that I am
 confused. It would seem that this is just one more indication that the
programs are performing correctly and that the problem is in the exchange-correlation (xc) 
 functional. Oscillator strengths are more sensitive than are excitation
 energies.
 We talk about oscillator strengths from TDDFT in
 [CS00] Mark E. Casida and Dennis Salahub, J. Chem. Phys. 113, 8918 (2000).
 "Asymptotic correction approach to improving approximate
 exchange-correlation potentials: Time-dependent density-functional theory
 calculations of molecular excitation spectra"
I have attached the PDF file of the article. The point is that what you can expect to 
 get right in the TDDFT adiabatic approximation when Exc is exact is the
 polarizability.
 Assuming that the dynamic polarizability is right means that the global
 distribution of
 oscillator strengths is about right given the calculated excitation
 energies. However
nearby states (or states missed because of the adiabatic approximation) can lead to variations in the exact distribution of oscillator strengths. This is why in any method 
 (not just TDDFT) it is very difficult to calculate accurate absolute
 oscillator strengths.
 (By the way it is also very difficult to determine accurate absolute
 oscillator strengths
experimentally, which is why we were so careful in our article to chose good experimental 
 data.) You might want to just convolute your TDDFT stick spectrum with a
 suitable
broadening function and see how it compares with the experimental spectrum. 
 Below I will ask you a few other questions ...
 
 We have calculated spectrum and oscillator strengths of D3 symmetry
 complex [Ru(bpy)3]2+ using TDDFT B3LYP with the LanL2DZ
 basis set and IEFPCM or COSMO water solvent shell models in
 Gaussian03 or ORCA program packages. We have received two
 singlet-E lines: 456.18 nm (0.0140 oscillator strength) and
 435.61 nm (0.1373 oscillator strength).
 Experiment of [Ru(bpy)3]2+ derivatives gives two peaks:
 one at about 450 nm (intensity 1.0 arbitrary units) and the
 other near 430 nm (intensity 0.6 arbitrary units)
 therefore the values of our calculated intensities are reversible in
 comparison with experiment.
 Oscillator strength calculations done by G03 and
 ORCA are exactly the same wrong calculating with PBEPBE potential
 using 6-311G* basis set.
If G03 and ORCA give the same thing with the same basis and functional then 
 that is NOT wrong. Rather it is just what TDDFT gives you. And as you say
it is different from the official interpretation of the experiment (which to be perfectly rigorous would involve a vibronic calculation which was probably not done.) 
 
 My impression is that it is something wrong in B3LYP, PBEPBE potentials
 when
 calculating oscillator strengths by Gaussian03 and ORCA programs
 of the D3 symmetry complexes and their derivatives
 while the values of spectral lines are exactly good.
Your impression is probably right that the problem is the xc functional. But : 
 1) How did you make the correspondance between the calculated and measured
 spectra? Did only one set of peaks come out of the theory and only one set
 could be identified in the experiment?
 2) Why the association with D3 symmetry? Do you think that something is
 going on that has to do with degeneracy?
 
 It is interesting that even ZINDO gives the right intensities in
 comparison with experiment but the values of spectral lines are
 shifted strongly.
 
 3) What is the character of the ZINDO excitations? Do they have
 important charge transfer or 2-electron excitation character?
 ZINDO neglects Rydberg-type excitations. I wouldn't expect
 these to be too important in the species you are studying but
 do Rydberg-type excitations appear in your TDDFT calculations?
 
 Maybe you can advise some good program with DFT
potential/basis set for the calculations of right oscillator strengths 
 of
 
 the D3 symmetry complexes and their derivatives?
 
 The programs you are working with should be fine.
 
 In the case if you are far from the field of calculations of oscillator
 strengths in G03 or ORCA, would you please to forward my question to
 developers of program packages who are working on subroutine
 of calculations of oscillator strengths in G03, ORCA or people who
 possesses the good subroutine.
 
 If the programs are correct, then why should I bother the developers ... ?
 Or are you saying that developers of xc functionals should be contacted?
 
 For regular organic compounds the oscillator strengths obtained by
 TDDFT in G03 and ORCA are good.
 
 That is fine.
 
 More about our research you can see in our website
 www.itpa.lt/~tamulis/ and press on button
 For PACE project WEBsite
 
 Looks like very nice work.
 Best wishes,
 Mark
 Mark E. CASIDA                "Improving theory for better chemistry."
 Professeur, chimie theorique
Laboratoire de Chimie Theorique Departement de Chimie Molecularie (DCM, UMR 
 CNRS/UJF 5250)
 Institut de Chimie Moleculaire de Grenoble (ICMG, FR-2607)
 Universite Joseph Fourier (Grenoble I)
 301 rue de la Chimie, BP 53
 F-38041 Grenoble Cedex 9
 FRANCE
 tel: 04.76.63.56.28 (from/de France)
 011.33.4.76.63.56.28 (from/de Canada or/ou USA)
 mailto:Mark.Casida-#-UJF-Grenoble.FR
 web page:
 http://dcm.ujf-grenoble.fr/PERSONNEL/CT/casida
 ***************************************
 Date: Wed, 4 Jul 2007 09:41:52 -0500
 
 From: Jim Chelikowsky <jrc-#-ices.utexas.edu>
 
 To: Arvydas Tamulis <tamulis-#-mserv.itpa.lt>
 Dear Arvydas,
 Thank you for your interest in our work.
I believe you can show that the oscillator sum should work with TDDFT. As such, I believe it is likely that the codes you mention may not be converged or TDDFT may not describe transition metal states properly. I do not think it is a matter of which functional you use.
This said, I have not run either of these codes and cannot speak directly to the matter. (I do know with our code that we have computed static 
 polarizabilities with finite field differences and with TDDFT.  These
numbers agree to within 1%. This is a good test on the oscillator strength, 
 at least the sum of them.  If you could, you might try a similar test for
 the system of your (if it is feasible).)
I will forward your email to my former student, Igor Vasiliev. He may know more. 
 Best regards,
 Jim Chelikowsky
 **************************************
 Date: Thu, 5 Jul 2007 13:18:37 +0100
 
 From: Noel O'Boyle <baoilleach-#-gmail.com>
 
 To: tamulis-#-mserv.itpa.lt
 I've done this calculation too at the B3LYP/LanL2DZ level of theory.
 Two points:
 (1) Experiment is in solution. Try a solvent model (note that some
 solvent models have no effect on symmetric molecules, I seem to
 recall).
 (2) Experiment has a band comprising multiple overlapping transitions.
 Try convoluting the experimental results (with 40 or so calculated
 transitions) with something like GaussSum, and see what happens.
 Noel O'Boyle
 (Formerly Han Vos Research Group, DCU)
 *********************************************
 On Wed, 4 Jul 2007, Arvydas Tamulis tamulis_._mserv.itpa.lt wrote:
 
 Sent to CCL by: Arvydas Tamulis [tamulis.:.mserv.itpa.lt]
 Dear Colleagues,
 We have calculated spectrum and oscillator strengths of D3 symmetry
 complex [Ru(bpy)3]2+ using TDDFT B3LYP with the LanL2DZ
 basis set and IEFPCM or COSMO water solvent shell models in
 Gaussian03 or ORCA program packages. We have received two
 singlet-E lines: 456.18 nm (0.0140 oscillator strength) and
 435.61 nm (0.1373 oscillator strength).
 Experiment of [Ru(bpy)3]2+ derivatives gives two peaks:
 one at about 450 nm (intensity 1.0 arbitrary units) and the
 other near 430 nm (intensity 0.6 arbitrary units)
therefore the values of our calculated intensities are reversible in comparison with experiment. 
 Oscillator strength calculations done by G03 and
 ORCA are exactly the same wrong calculating with PBEPBE potential
 using 6-311G* basis set.
My impression is that it is something wrong in B3LYP, PBEPBE potentials when calculating oscillator strengths by Gaussian03 and ORCA programs 
 of the D3 symmetry complexes and their derivatives
 while the values of spectral lines are exactly good.
 It is interesting that even ZINDO gives the right intensities in
 comparison with experiment but the values of spectral lines are
 shifted strongly. Maybe you can advise some good program with DFT
 potential/basis set for the calculations of right oscillator strengths of
 the D3 symmetry complexes and their derivatives?
 In the case if you are far from the field of calculations of oscillator
 strengths in G03 or ORCA, would you please to forward my question to
 developers of program packages who are working on subroutine
 of calculations of oscillator strengths in G03, ORCA or people who
 possesses the good subroutine.
 For regular organic compounds the oscillator strengths obtained by
 TDDFT in G03 and ORCA are good.
 More about our research you can see in our website
 www.itpa.lt/~tamulis/ and press on button
 For PACE project WEBsite
 With best regards,
Arvydas Tamulishttp://www.ccl.net/chemistry/sub_unsub.shtmlConferences: http://server.ccl.net/chemistry/announcements/conferences/>;