Another common question on CCL…
The use of Internal Redundant coordinates (through the Opt=ModRedundant option) must not be overlooked! This option performes a geometry optimization at each step while maintaining the scanned variable constant, which is referred to as a Relaxed Potential Energy Surface (PES) Scan. Using internal coordinates becomes compulsory and a well-defined Z-Matrix is preferable. A common mistake is to define the variable to scan without taking into consideration all the other variables which depend from the first. This means that functional groups can be destroyed on each step and in the best case scenario the optimization algorithm will put it back together again at, of course, some unnecessary computational cost. The worst case scenario would consist of the molecule’s complete distortion from any physically achievable structure, making the calculation end with an error or by providing meaningless results. Therefore the use of wildcards is compulsory and it is illustrated below.
Lets say we want to calculate the energy barrier associated with the rotation of a dihedral angle on ethane. If we define the dihedral angle 2-1-5-8 (fig.1) to be scanned from 180.0 deg to 0.0 by 45.0 degrees increments then the second step of the scan would look distorted like the molecule on figure 2. Since the value of D[2-1-5-8] is kept constant until convergence then the algorithm (provided a decent level of theory is indicated when it comes to molecules more interesting than ethane) will increase the values of dihedral angles D[2-1-5-6] and D[2-1-5-7] until reaching the convergence criteria at a regular structure.
This would result in an unefficient method for computing the desired barrier. Hence, before defining the variable to be scanned we must select carefully, through the use of wildcards all other variables associated with it.
* 1 5 *
2 1 5 8 S #steps #increment
The first line will make all other dihedral angles around the C2-C3 bond to behave in the same way as D[2-1-5-8], steps is the number of increments in the variable to be performed from the starting geometry and increment is the step size to be taken in the variable (Note this values are always in integer format so a decimal point is always to be used! e.g. for our hypothetical calculation we could write the second line as 2 1 5 8 S 4 45.0 with which the dihedral angle would be scanned from 0.0 degrees to 180.0 degrees in 45.0 degrees steps) In this way, the second step of our hypothetical calculation would look like the molecule on figure 3, and the convergence criteria would be reached much faster.
This same procedure holds for scanning any other kind of geometrical variable. Common mistakes and error messages during geometry scans are associated to lack of memory, maximum number of steps reached or energy divergence due to inconsistent geometries; when analyzing for an error is always best to take a look at the optimization progress in order to find out at which step the molecule started to break apart.
Rigid PES are much easier to perform although they can be less inforative in terms of a real dynamic process. In this case the “scan” option is the one we want to use. The molecule must be defined in Z-Matrix format below of which a blank line must be placed and after that the following information
Variable as defined in Z-Matrix initial value Number of steps and Increment value
So a rigid PES scan for the previous example would look like
D1 180.0 3 60.0
If another variable is placed with the same info below this line then the program will perform all possible combinations while printing a summary with the energy at each conformation. Sometimes this summary doesn’t have enough space for the numbers to be printed and a set of stars ****** are shown where the energy is supposed to be. No panic the energy was calculated it just cannot be printed; you either have to use a visualization program such as GaussView in order to read the energy at each conformation or just browse directly -and patiently- through the output file
The intention of this blog was not to become a Gaussian 0x support forum but I’m glad that so many people -specially grad students- have found it helpful in their research.
As usual all comments and ratings are welcome