Re: CCL:Crystal structure vs geometry optimised structure

Hello,
 Thanks to everybody who sent a reply to my query (see below) on whether
 it is better to use crystal structures or DFT-optimised geometries for DFT
 calculations (energy, orbitals, TD-DFT).
 Overall, the general consensus seems to be that although it may be interesting
 to calculate and compare the energies of the two structures, frequency
 calculations
 may only be performed at a minimum (of the same level of theory).
 One point which was not addressed was whether TD-DFT calculations must also
 be performed at a minimum. Anybody care to comment?
 Noel
 Original Query:
 **************
 Hello CCLers,
 I intend to carry out DFT calculations on inorganic complexes. In some cases
 I have crystal structures available.
 Is it better to use crystal structures as a starting point for energy and
 frequency calculations, or to use DFT-optimised structures? (where the
 DFT-optimised
 structure is in excellent agreement with the crystal structure)
 Thanks,
 Noel O'Boyle
 Han Vos Research Group,
 Dublin City University.
 http://www.dcu.ie/~chemist/Staffpages/han_vos.htm
 **************
 Replies:
 --------------
 It depends on what kind of crystal structures you are using, if they
 are for proteins,
 you will have to accept an experimental accuracy of 0.1-0.3 Angstroms;
 for smaller
 inorganic complexes you might be better off with the accuracy.
 If the complexes are not too big, I'd use both as starting point, see
 how much they are
 alike, and calculate the freqs for both structures, and see which ones
 agree better with
 experiment.
 Marcel Swart <swart..at..chem.vu.nl>
 --------------
 If the minimized structure is indeed in "excellent" agreement with the
 crystal structure, it is probably better to work from there. However, in
 my
 recent experience, you have to be very careful with the minimized
 structures, as they may have lost some subtleties of the crystal structure.
 You may want to do a test with the crystal structures you do have. Calculate
 the molecular propertied you are interested in for the crystal and minimized
 structures. If they are not terribly different than I would use the
 minimized ones. If they are very different, than you have to give some
 thought about why they are different and what that is telling you about
 the
 system.
 Joslyn Kravitz
 University of Michigan
 "Joslyn Kravitz" <jyudenfr..at..umich.edu>
 --------------
 For doing frequency calculations, you will want to use the DFT
 optimized structures, as the calculation won't be very meaningful
 otherwise. You should only be calculating frequencies at a minimum.
 If your DFT structures are in excellent agreement with your crystal
 structures, you are probably better off using the DFT optimized
 structures for energy comparisons as well. Of course, if in certain
 cases your DFT optimized structures were fundamentally different from
 the crystal structures, the DFT energy comparison might not make much
 sense if you are trying to examine the energetic differences between
 your (crystal) samples.
   Hope that helps - John
 John Bushnell <bushnell..at..chem.ucsb.edu>
 --------------
 I have not worked with crystal structures of inorganic compounds but have
 used crystal structures of organic compounds from the Cambridge
 Crystallographic Database. The older structures often have to be treated
 with caution. First, the coordinates of hydrogen atoms are missing. Many
 modeling programs allow one to add the missing hydrogens. After that I
 usually perform a geometry reoptimization. With organic compounds, I'll
 stop at the HF level with a 3-21G* basis set and then perform single-point
 calculations of the energy at higher levels of theory such as DFT.
 Wayne E. Steinmetz
 Coordinator of Molecular Biology
 Carnegie Professor of Chemistry
 Woodbadge Course Director
 Chemistry Department
 Pomona College
 645 North College Avenue
 Claremont, California 91711-6338
 USA
 phone: 1-909-621-8447
 FAX: 1-909-707-7726
 Email: wsteinmetz..at..pomona.edu
 WWW: pages.pomona.edu/~wsteinmetz
 --------------
 Hello,
 It is probably not a good idea to use experimental geometries for freq calc's,
 because
 freq calc's are strictly valid only at the geometry obtained by the same
 method as the
 freq calc. Thus a B3LYP/6-31G* freq calc should be done on a geometry obtained
 by a
 B3LYP/6-31G* optimization. Since your freq jobs are likely to be about as
 long or longer
 than any geom optimizations, you probably won't lose very much time by doing
 the
 optimizations first.
 Another problem with crystal structures is that they may differ significantly
 > from the
 isolated molecule structures that I presume you will use for your freq jobs
 (you won't
 use a crystal lattice as input, I guess); but you took account of this by
 saying "where
 the DFT...is in excellent agreement..."
 It would be interesting to do some freq calc's on both DFT-opt and experimental
 structures and see how much they differ. But conventional wisdom is that
 the DFT-opt
 structures are the ones to use.
 If you didn't want freqs, just relative energies of isomers, you could do
 single-point
 calculations (much faster than optimizations) on the experimental structures.
 Again, it
 might be interesting to compare relative energies based on the experimental
 and the
 DFT-opt structures.
 E. Lewars
 elewars <elewars..at..trentu.ca>
 --------------
 Hi Noel
 Before carrying out a frequency calc it is essential that you have optimised
 the geometry at the same level of theory. Therefore you can start with
 either your crystal structure or your previously optimised structure.
 Regards
 Roma
 "Roma Oakes" <r.e.oakes..at..btconnect.com>
 --------------
 Hai,
 It is better to use x-ray straucture. Because, u need to compare the dft
 results with experiments. In DFT, u need to chose a correct functional for
 ur calculation. If u use the x-ray structure, u can validate the correct
 functional by comparing with x-ray co-ordinates.
 Cheers, Mahesh
 Mahesh Sundararajan,
 C/O. Prof. Ian. H. Hillier
 Ph.D Student, Theoretical & Computational Chemistry Group,
 Department of chemistry,
 Oxford Road, The University of Manchester.
 Manchester - M13 9PL, UK
 07734321159, 0161 4480531
 --------------
 Hmm... I am not sure I understand you correctly....
 You want to perform energy and frequency calculations only?
 which structure (crystallographic or computational) is better for energy
 calculations depends on your problem
 but I do really not know how to use crystallographic geometry (without
 geometry optimization???) in case of frequency calculations
 Of course if you want to obtain only "real" frequencies
 <zborowsk..at..chemia.uj.edu.pl>
 --------------
 Dear Noel
 There have been some papers -- eg by Buntine at University of Adelaide
 and one by my co-workers -- in which the difference between
 the energies of the (optimised) gas-phase structures
 and those of the crystal structures are interpreted as crystal
 packing energy costs.
 eg:
 J. Beckmann, K. Jurkschat, M. Sch|rmann, D. Dakternieks, A. E. K. Lim
 and K. F. Lim, "Comparison of the flexibility of eight-membered
 tetrasiloxane and stannasiloxane rings: A crystallographic and
 computational study", Organometallics, 2001, 20 (24), 5125-5133
 <http://pubs3.acs.org/acs/journals/doilookup?in_doi=10.1021/om010620o>;
 .
 Kieran
 ------------------------------------------------------------
   Dr Kieran F Lim Biol. and Chemical Sciences
     (Lim Pak Kwan) Deakin University
   ph: + [61] (3) 5227-2146 Geelong VIC 3217
   fax: + [61] (3) 5227-1040 AUSTRALIA
   mobile phone: 0438 350 259 (within Australia)
   mailto:lim..at..deakin.edu.au http://www.deakin.edu.au/~lim
 ------------------------------------------------------------