Optimisation from transition state geometries: Difference between revisions
Jump to navigation
Jump to search
New page: == Introduction == An IRC calculation is a relatively expensive calculation in terms of cpu time. There exists a cheaper alternative: # Make a small displacement of the TS geometry along t... |
No edit summary |
||
| (4 intermediate revisions by 2 users not shown) | |||
| Line 1: | Line 1: | ||
go back to [[Main Page]], [[Computational Resources]], [[Computational Codes]], [[GAUSSIAN]] | |||
== Introduction == | == Introduction == | ||
An IRC calculation is a relatively expensive calculation in terms of cpu time. There exists a cheaper alternative: | An IRC calculation is a relatively expensive calculation in terms of cpu time. There exists a cheaper alternative: | ||
# Make a small displacement of the TS geometry along the forward and the backward direction of the eigenmode corresponding to the imaginary frequency. | # Make a small displacement of the TS geometry along the forward and the backward direction of the eigenmode corresponding to the imaginary frequency. (There are scripts available that do this process automatically) | ||
# From the two geometries created, start a geometry optimisation. | # From the two geometries created, start a geometry optimisation. | ||
# In order not to step over small hills on the energy surface, do this optimisation with a small step length. | # In order not to step over small hills on the energy surface, do this optimisation with a small step length. | ||
# A problem with this method is the risk of making a too large displacement from the TS geometry. If this happens, any minimum found at the end of the optimisation will ''not'' be connected to the TS geometry and will thus be useless in this context. | |||
# By looking at the maximum gradients of the first few steps, one can check if the displacement has been too large. | |||
# The gradients of the first few steps should be low and ''increasing'' for the successive steps. | |||
== Input example == | == Input example == | ||
<code><pre> | <code><pre> | ||
| Line 43: | Line 49: | ||
</pre></code> | </pre></code> | ||
== Comment | == Comment to the input example == | ||
# the shorter step length is ensured by including the option "maxstep=5" (the default is maxstep=30) in the keyword "opt" | |||
# when using such a small step length, the energy differences between two successive energy points can be very small. In order to compute this energy difference more accurately, it may be necessary to tighten the convergence criteria. In the example above scf(conver=10) is used (conver=8 is default). | |||
Latest revision as of 13:27, 29 July 2009
go back to Main Page, Computational Resources, Computational Codes, GAUSSIAN
Introduction[edit]
An IRC calculation is a relatively expensive calculation in terms of cpu time. There exists a cheaper alternative:
- Make a small displacement of the TS geometry along the forward and the backward direction of the eigenmode corresponding to the imaginary frequency. (There are scripts available that do this process automatically)
- From the two geometries created, start a geometry optimisation.
- In order not to step over small hills on the energy surface, do this optimisation with a small step length.
- A problem with this method is the risk of making a too large displacement from the TS geometry. If this happens, any minimum found at the end of the optimisation will not be connected to the TS geometry and will thus be useless in this context.
- By looking at the maximum gradients of the first few steps, one can check if the displacement has been too large.
- The gradients of the first few steps should be low and increasing for the successive steps.
Input example[edit]
%mem=720MB
%chk=prodOpt_b1.chk
#p scf=(conver=10) opt=(calcfc,maxstep=5,maxcycle=70) freq=internalmodes b3lyp/genecp nosymm
optimisation from TS
0 1
76 0 -0.983269 -0.024600 0.000009
6 0 -1.635059 1.808684 -0.000683
6 0 1.002562 0.806467 -0.000303
6 0 -0.600233 -0.734234 1.770918
8 0 -2.030038 2.894422 -0.001086
8 0 1.492402 1.913347 -0.000718
8 0 -0.307490 -1.181244 2.800662
6 0 -0.600203 -0.735596 -1.770350
6 0 -2.758976 -0.827633 0.000294
8 0 -0.307441 -1.183405 -2.799741
8 0 -3.815309 -1.292251 0.000450
1 0 3.461423 1.540051 1.193881
6 0 3.832917 0.518831 1.241612
7 0 3.465095 -0.180024 0.000068
1 0 4.925251 0.506760 1.356016
1 0 3.372253 -0.002424 2.083099
6 0 3.832916 0.517887 -1.242008
6 0 3.830772 -1.601406 0.000609
8 0 1.689587 -0.353491 0.000138
1 0 4.925248 0.505724 -1.356404
1 0 3.461427 1.539150 -1.195052
1 0 3.372251 -0.004009 -2.083100
1 0 4.923748 -1.699659 0.000649
1 0 3.414653 -2.076758 -0.888911
1 0 3.414651 -2.076082 0.890495
@/home/tfjermestad/basisSet/basisSet.txt/N
Comment to the input example[edit]
- the shorter step length is ensured by including the option "maxstep=5" (the default is maxstep=30) in the keyword "opt"
- when using such a small step length, the energy differences between two successive energy points can be very small. In order to compute this energy difference more accurately, it may be necessary to tighten the convergence criteria. In the example above scf(conver=10) is used (conver=8 is default).