Notes for Installing and Running
MN-GSM
Below is a summary of important notes associated with the installation and proper execution of MN-GSM. Notes in this section are updated on a regular basis, often after particulary useful suggestions from MN-GSM users, or after a new version of MN-GSM is released. Thus, the notes below are continually updated (useful suggestions should be communicated to the MN-GSM developers)
For a more detailed description of MN-GSM, as well as the notes described below, users are directed to the MN-GSM User's Manual, which is also available on this website.
Notes on Installation
1. On IBM platforms, the following environmental variable may have to be set in order to run MN-GSM:
setenv XLFRTEOPTS namelist=old:xrf_messages=yes:err_recovery=yes
This line can be added to the users .cshrc file, or any initialization file read during the login process.
Notes on Input File Structure
1. The Fortran namelist command looks for the first occurrence of the $CM2 keyword or the $MNGSM keyword in the input file, where the "$" is in column 2. Therefore, if either namelist is placed in the Gaussian title line beginning in column 2, this will be the line that will be read by the program.
2. No namelist keywords are case sensitive. However, character arguments provided as input (e.g., input for the JobName keyword) are case sensitive.
3. $CM2 and $MNGSM input can span several lines, provided that the input is terminated by a single $END.
Notes on Basis Sets
1. The keyword for the MIDI! basis set in Gaussian is midix. The keyword for MIDI!6D is midix 6d.
2. When using effective core potentials, the number of frozen electrons must always be specified in the .AtmP file.
3. Analytical gradients are only available for basis sets that use Cartesian d and f functions. Note that when the keyword "gen" is specified in a Gaussian input file, spherical d and f functions (5D, 7F) are used by default. Thus, when performing liquid-phase geometry optimizations with either SM5.42, SM5.43, SM6, or the generalized Born model with the gen keyword, the 6d keyword must also be specified. If the 6d keyword is not specified and liquid-phase geometry optimization is invoked, the program will stop, and print out the following error:
SYMSLV: T(1,1)=0
Notes on SCF Convergence
1. Accurate (i.e. SCF = Tight or SCF = (conver = N), where N is an integer greater than 4) should be used for all single-point calculations. Note that specifying SCF = (conver = 8) invokes the same SCF convergence criterion as specifying SCF = Tight. For more information on SCF convergence options in MN-GSM, see the section "SCF Convergence" in the manual
2. Only the DIIS SCF convergence method (which is the Gaussian default) is available to MN-GSM; SCF=QC is not supported.
3. If ΔEE is very small (e.g., less than 0.010 kcal/mol) then there is a possibility that the liquid-phase SCF calculation will not converge, leading to the following error message in the output:
Restarting incremental Fock formation.
>>>>>>>>>> Convergence criterion not met.
To solve this problem, one should carry out a separate calculation in the gas phase. Then, with the gas phase energy in hand, the following line should be added to the $MNGSM namelist for the calculation in solution:
IGAS=3 ETGAS=Gas-phase_Energy
These keywords tell MN-GSM to skip the gas phase calculation and begin obtaining the wave function in solution immediately.
Notes on Numerical Gradients
1. The Gaussian "Opt = CalcFC" causes the calculation of an analytical Hessian for HF and DFT methods. In order to use a liquid-phase numerical Hessian in a geometry optimization, the numerical Hessian must be calculated in a separate job, saved to the checkpoint file, and then read into the optimization job using Opt = ReadFC. This is illustrated in a test suite calculation discussed in the manual.
2. It has been determined that 10-5 angstroms is often the optimum step size in the evaluation of numerical gradients. The StepSize option of the Force keyword does not allow the user to input a step size smaller than 10-5 angstroms. The user can bypass this problem by using Iop(1 / 39 = -1), which instructs the code to read in a user-supplied step size (format D20.13) from the line following a blank line after the molecule input (see Example 5 of the Input and Output Examples section of the manual).
3. Numerical gradients can be used for geometry optimizations in the same fashion already described in the Gaussian manual for any method not having analytical gradients available.
Notes on Program Limits
1. The maximum number of atoms allowed in an MN-GSM calculation is 128, and the maximum number of basis functions allowed is 1000. If the user wishes to increase these limits, then the parameters MxDAWk MxSolWk, and MxNU in the MN-GSM specific subroutines (file MNl502.src) must be modified. (The memory used in file MNl701.src is dynamically allocated onto Gaussian's main work array, therefore there are no extra issues with the size of a particular system). If the user wishes to increase the maximum number of basis functions, he or she must set the value of MxNU to the new desired value. In addition, the user must adjust the size of MxSolWk according to the guidlines outlined in the manual. For a larger maximum value of the maximum number of atoms, MxSolWk must be adjusted according to the guidelines outlined in the manual, and the value of MxDAwk must be adjusted according to the guidlines outlined in the manual. Note that MxDAWk, MxSolWk, and MxNU appear in several places in MNl502.src, and they all must all be modified. Also, link 502 must be recompiled after these modifications are made (see the section in the manual entitled Installation and testing).
END OF NOTES