Re: CCL:crystal packing calculations and programs

 Hi Doug,
 Your question:
 >Can one reasonably start from a high quality ab initio calculation on a
 >single molecule (gas phase) and use that structure to pack a crystal?  If
 >one does, what is the likelihood of getting the crystal structure correct?
 Generated an interesting discussion on the CCL, and is closely related to
 Hans-Christoph's query:
 >        Based on X-ray data, I'm interested in intermolecular interaction
 >energies in order to rationalize specific packing effects.
 So I'd like to add my experience to the discussion.
 So far as correctly representing intermolecular interaction energies, this
 is not likely to be as simple as it seems.  In fact, depending on your
 system, molecular mechanics simulations are likely to work much better than
 quantum mechanics simulations, unless you include diffuse functions in the
 wavefuntions.
 As an extreme example, I've tried modeling some charge transfer compounds,
 where you might expect the intermolecular interaction  be dominated by the
 electronics, hence better modeled with an electronic method than with a
 force field.  If you try to find a stable geometry for the quinhydrone of
 tetramethyl benzoquinone, for instance, and start with the known X-ray
 crystal structure, both AM1 and PM3 attempted geometry optimizations tend to
 just blow the two molecules apart, as does MM2.  The Sybyl force field does
 find a stable geometry, but with a lateral displacement as well as the
 initial longitudinal displacement.
 The charge transfer complex between phenolthiazine (PTZ) and
 tetracyanoquinone (TCNQ) responds similarly: the X-ray crystal structure
 shows the two molecules to be about 3.4 A apart, but with AM1 or MP3
 calculations they just separate, and an INDO/1 geometry "optimization"
 bonds
 them together!  These calculations may handle the electronics, but they are
 otherwise seriously deficient in handling the long range intermolecular
 interactions.  I expect ab initio methods to have the same problem unless
 you include diffuse functions.
 Interestingly, if you model either system as an excited singlet, both AM1
 and PM3 find stable geometries with the molecules parallel and a reasonable
 separation for a charge transfer complex.  I'm not sure whether this is an
 artifact, or whether it may be a legitimate representation of what's going
 on in the crystal.  Does anyone have any thoughts on this?
 With the PTZ-TCNQ pair, MM2 does find a stable geometry with the two
 molecules parallel, and Sybyl not only finds a stable geometry, but gets
 them centered above each other as well.  This brings me to the second
 problem, that has already been pointed out by Mike  Kotelyanskii:
 >the molecular conformation in vacuum can be very different from it's
 >in the crystal, unfortunately I cannot come up with an examples right now
 The molecules are stabilized differently in a crystal than in an isolated
 complex, so that not only is a different conformation possible, but a
 different intermolecular alignment will be preferred as well.  In the case
 of the charge transfer compounds I just mentioned, the X-ray crystal
 structures show the molecules flat and parallel (as one would expect), but
 offset longitudinally relative to each other.  One half of each molecule
 overlaps with a molecule in the layer below it in the crystal, and the other
 half overlaps with a molecule above it!  Accordingly, you may be able to
 model an intermolecular interaction, but if you are going to reproduce the
 alignment observed in a crystal, I believe you will need to do as you suggest:
 >Or does one need to perform ab initio calculations in the periodic solid
 >(e.g., CRYSTAL 95 calculations) before performing crystal packing?
 Good luck.
 EC
 ---
 Ernest Chamot
 Chamot Labs, Inc.
 530 E. Hillside Rd.
 Naperville, Illinois 60540
 (630)637-1559
 echamot (- at -) xnet.com
 http://www.xnet.com/~chamotlb