CCL:G: Easy convergence of Diffuse functions

 Sent to CCL by: "David  Hose" [anthrax_brothers:hotmail.com]
 Hi Ya,
 The addition of diffuse functions to the basis set is well known to cause
 problems with the
 convergence of the calculations.  The most common fix is to use a better quality
 guess for the
 initial wavefunction in the second calculation.
 Assuming that you are using Gaussian, the simplest way is to optimise the
 geometry of the
 molecule at say HF/6-31G* level of theory.  Duplicate the resulting checkpoint
 point file.  Then set
 up the second (single point?) calculation, at say the HF/6-31+G* level of
 theory, with
 Guess=Checkpoint included in the route selection, ensuring that the %chk points
 to the previous
 calculation checkpoint file.  On occasions, it is better if the previous
 calculation included a
 frequency calculation as the superior hessian allows the subsequent calculation
 converge more
 easily.
 There are some other approaches that can be used for difficult cases, but the
 above method
 typically works.
 I suggest that you check out David Young's book Computational Chemistry: A
 Practical Guide for
 Applying Techniques to Real World Problems.  A great source of handy hints and
 tips.  There is a
 version of his book here on CCL.  The section of converge can be found here.
 http://www.ccl.net/cca/documents/dyoung/topics-orig/converge.html
 Regards,
 Dave.
 _____________
 Sent to CCL by: "Mr Shabbir" [shabbir]^[nenu.edu.cn]
 Dear all
 I want to share some thing about optimization. I optimized several systems with
 6-31G* basis set
 but when I use diffuse function (+) the same previously optimized system at
 (6-31G*) fails to
 converge by using diffuse function. How to treat the diffuse function especially
 in optimization? Is
 there any special keyword or way that can make the convergence easy with diffuse
 functions?
 Regards
 Mr.Shabbir
 shabbir::nenu.edu.cn