Introduction+to+GAMESS

=Using GAMESS=

Written by TPP 3/29/2012, correcting and extending the previous GAMESS page from the summer of 2011
Register for GAMESS here, it's free and all you need is an email address.

Compiling GAMESS on OSC
After downloading GAMESS, you want to extract it, and place the resulting folder somewhere convenient, like maybe just in the home directory, or perhaps in a $HOME/Programs folder. Now follow these simple instructions:

code setenv GMS_PATH /nfs/04/ucn1093/gamess setenv GMS_TARGET linux64 setenv GMS_FORTRAN ifort setenv GMS_IFORT_VERNO 10 setenv GMS_MATHLIB mkl setenv GMS_MATHLIB_PATH /usr/local/intel/mkl/10.0.3.020/lib/em64t setenv GMS_MKL_VERNO 10 setenv GMS_DDI_COMM mpi setenv GMS_MPI_LIB impi setenv GMS_MPI_PATH /usr/local/intel-mpi/ code
 * cd /PATH/TO/GAMESS
 * ./config and answer the questions, OR...
 * Copy this install.info file (save it as install.info) and make the necessary changes to compiler versions and directories, etc
 * 1) !/bin/csh
 * 2) compilation configuration for GAMESS
 * 3) generated on opt-login04.osc.edu
 * 4) generated at Thu Mar 29 16:16:16 EDT 2012
 * 1) machine type
 * 1) FORTRAN compiler setup
 * 1) mathematical library setup
 * 1) parallel message passing model setup
 * cd ddi
 * ./compddi
 * Lots of output, save to file by appending "> compddi.log" to the above command
 * cd ..
 * ./compall
 * Very long wait here, but assuming you've made all the fixes in the above install.info file OR chose your config options correctly this should work
 * When you see this string "--- done with all compilations " you've just compiled GAMESS successfully
 * ./lked gamess 01
 * Creates a link for the GAMESS executable with version number 01, this can of course be 00, 01, 02, 03...
 * Allows user to compile several different versions of GAMESS (i.e., serial version, OpenMPI version, gfortran and atlas, Intel-based compilers)

Compiling GAMESS on a Local Machine

 * cd /PATH/TO/GAMESS
 * ./config
 * Answer the questions accordingly
 * cd ddi
 * ./compddi
 * Lots of output, save to file by appending "> compddi.log" to the above command
 * mv ddikick.x ..
 * You may or may not need this command, if the executable ISN'T there, you don't need to do this step
 * cd ..
 * ./compall
 * Very long wait here, but assuming you've made all the fixes in the above install.info file OR chose your config options correctly this should work
 * ./lked gamess 01
 * Creates a link for the GAMESS executable with version number 01, this can of course be 00, 01, 02, 03...
 * Allows user to compile several different versions of GAMESS (i.e., serial version, OpenMPI version, gfortran and atlas, Intel-based compilers)

Extra Notes for Memory Management:

 * The memory requirements for GAMESS can exceed the maximum allowed value designated by your OS, be sure to change them using the following command:
 * sudo /sbin/sysctl -w kernel.shmmax=10000000000
 * The above command allows GAMESS to use 10 GB worth of RAM (this will not error if you have less than 10 GB of RAM)
 * This command can of course be aliased and placed into the ~/.bashrc or ~/.login files
 * alias setmax='sudo /sbin/sysctl -w kernel.shmmax=10000000000'
 * Requires you type "setmax" every time you restart the PC
 * Adding the command above to the ~/.login file should set this parameter automatically every time the PC is restarted
 * The ddi README has a more 'permanent' fix in it as well, consult this README if this is the desired means by which to augment the maximum amount of memory a program can use.

Energy Decomposition Analysis using Localized Molecular Orbitals
GAMESS is able to perform a localized molecular orbital energy decomposition analysis which attempts to partition the interaction energy of a series of monomers (in GAMESS up to 10) into several components: electrostatics, polarization, exchange, and repulsion. This can be especially useful in understanding the interaction of these monomers, like the effect of charge transfer in the interaction of a series of monomers. In this analysis, charge transfer contributions would be lumped in with the polarization contribution. We've conjectured CT and polarization have no clear "cut off" point and so energy decomposition schemes which do actually define some means by which to partition CT from polarization should be taken with a grain of salt. This isn't to say the schemes and methods are wrong, only that they have some inherit limit to their accuracy. In fact, we are considering using some of these methods in our work. In the wise words of our advisor, we simply don't have the ability to perform "God's quantum mechanics."

Here's an example GAMESS input for the HF LMOEDA on F- with a single water: code $contrl scftyp=rhf runtyp=eda ICHARG=-1 $end $basis gbasis=ACCQ $end $guess guess=huckel $end

$lmoeda matom(1)=1,3 mcharg(1)=-1,0 mmult(1)=1,1 $end $contrl nosym=1 $end $system mwords=10 timlim=123456 $end

$data F-H2O LMOEDA test C1 F 9.0  0.00000000  0.00000000  0.00000000 O 8.0  4.59151010  0.00000000  0.00000000 H 1.0  2.59320927 -0.05105963  0.00000000 H 1.0  4.92758949  1.77407118  0.00000000 $end

code