Docstoc

Hands on II Running Gaussian

Document Sample
Hands on II Running Gaussian Powered By Docstoc
					Hands-on II: Running Gaussian09




             David Tur, PhD
      Scientific Applications expert
            dtur@cesca.cat
        Hands-on II: Running Gaussian09


Working directory: /cescascratch/handson/gaussian
You can create your own directory inside
  /cescascratch/handson/users
  or work in your home
Examples at:
  • /usr/local/examples (archive.lsf examples)
  • /prod/G09/g09a2/tests/com (tests from Gaussian Inc.)
  • /cescascratch/handson/gaussian (Hands-on II examples)
Queu: curs
Clusters: obacs & cadi & prades
G03 versions: E.01,C.02, B.02
G09 version A.02
                   Hands-on II: Running Gaussian09

Example 1: Creating the first input file, H20 SP energy calculation h2o-sphf.dat
Input file can be created in two ways:
         • By hand: using local editor (VI, emacs, nedit...):
             • Coordenates file of H2O: h2o-zmatrix.dat and h2o-cartesianas.dat
           Hands-on II: Running Gaussian09

Example 1: Creating the first input file, H20 SP energy calculation
Input file can be created in two ways:
         • Using Molden:
            Hands-on II: Running Gaussian09
Example 1: Creating the first input file, H20 SP energy calculation
Input file can be created in two ways:
         • Using Molden:
           Hands-on II: Running Gaussian09

Example 1: Creating the first input file, H20 SP energy calculation
Input file can be created in two ways:
         • Using Molden:
            Hands-on II: Running Gaussian09

/home and /cescascratch are NFS , better alternative
running and writing at /tmp

Hands-on II. Example 2. Water dimer optimization
Running the job at /tmp, and when the job is finished
copy log and chk files at /home/whereeveryouneed
h2o-z-matrix-opthf-savechk.dat/lsf
runs a hf/6-31+G* optimization, saving the chk
h2o-z-matrix-spmp2-savedchk.dat/lsf
reading the chk of the previous calculations, performs
an mp2 single point calculation.
           Hands-on II: Running Gaussian09




Hands-on III. Example 3. Water dimmer frequencies
localizing an absolute minimum, and inspecting
thermochemical data:
h2o-dim-freq.dat
This job is done writing output files in the tmp
directory to be more efficient.
h2o-dim-freq.lsf
           Hands-on II: Running Gaussian09



Hands-on III. Example 4. Saving post-HF amplitudes
Run a job with small basis set, and save the
amplitudes:
g09a2-ccsdt-1.dat (ccsd(t)/cc-pvdz)
Reading amplitudes and run the job with a big basis
set:
G09a2-ccsdt-1.dat
                Hands-on II: Running Gaussian09

Different tips, suggestions or tricks to optimally run Gaussian using
CESCA facilities

Error: Error termination via Lnk1e in /prod/G03/g03e1/g03/l906.exe
                 Solution: You need more memory

Error: Something like “IUnit=1 –“ or “ Error termination in NtrErr”
                 Solution: You need more disk
Error: “SCF failure”
                 Solution 1: increase the number of steps
                 “scf=(maxcyc=300)”
                 Solution 2: try another algorithm “scf=qc”, much
                 slower but more 'efficient'
Hands-on II: Running Gaussian09




Thank you for your attention!!!
     QUESTIONS????


          David Tur, PhD
         Applications expert
          dtur@cesca.cat
Working with Gaussian09




 1.   Introduction
 2.   Preparing the input file
 3.   Running the program via a batch queue (LSF)
 4.   Examining and interpreting the output
 5.   From Gaussian03 to Gaussian09
 6.   Optimizing performance
Optimizing performance at CESCA


Different tips, suggestions or tricks to optimally run Gaussian
using CESCA facilities

  Type of calculation

  Hands-on Session:
   • Directories to be used
   • Cluster where the calculation is performed
   • Resources requested

  CESCA people is here to help you
Optimizing performance at CESCA




Take into account Performance and accuracy of the
method: i.e. Hybrid Functionals

  Cost similar to HF for medium-large systems
 Accuracy better than HF
 Accuracy for MP2 except for weakly-bound systems
 No as accurate as CCSD(T), CBS-QB3, etc.
 If B3LYP and CBS-4 agree, good check
Optimizing performance at CESCA


Take into account Performance and accuracy of the
method: i.e. Møller-Plesset Theory

 Generally a good hierarchy of models
    MP2 cheap
    MP4 good for most systems
    Series tends to oscillate
 Converged problems
    If HF a poor starting point
    If serious spin contamination
 Not exact for two-electron system
Optimizing performance at CESCA
Take into account Performance and accuracy of the
method: i.e. Compound Model Chemistries for
Thermochemistries

 Most accurate and expensive: W1U, CBS-APNO
  (~½ kcal error, 2 kcal worst case)
 Expensive but practical: CBS-QB3
  (~½ kcal error, 6 kcal worst case)
     Usually less expensive than G2 and avoids big failures of
     G2, G3 (e.g. SF6)
 Cheapest: CBS-4M (only recommended for minima
  (~3 kcal error, 20 kcal worst case)
  (If CBS-4M and B3LYP agree can have confidence)
Optimizing performance at CESCA




Initial Guess for Equilibrium Geometries

  GaussView, molden, molekel or other graphical interface
  Experiment
  Empirical force field calculations
  Semi-empirical MO calculations
  Lower level ab initio calculations
  Quantum chemical data bases
Optimizing performance at CESCA


Testing Minima

   Compute the full Hessian (freq from converged opt)
   Check the number of negative eigenvalues:
        -1 of more indicates a transition state of higher order
       saddle point
       Totally symmetric: a transition structure
       Non-totally symmetric: wants to break symmetry to reach
       some minimum
   If there are any negative eigenvalues, follow the associated
eigenvector to a lower energy structure
Optimizing performance at CESCA


Things to try when optimizations fail

 Number of steps exceeded
     Check for very flexible coordinates and/or strongly
    coupled coordinates
    Restart from a reasonable step and use CalcFC
 Maximum step size exceeded
     If it happens too often, check for flexible and/or strongly
    coupled coordinates
 Change in point group during optimization
 Check structure and/or use NoSymm
 Optimizing performance at CESCA

Things to try when Transition State Searches fail

   Too many negative eigenvalues of the Hessian during TS
optimization
       Follow the eigenvector with the negative eigenvalue
       Use Freq=Internal to see normal modes in internal
      coordinates
   No negative eigenvalues of the Hessian during a transition
structure optimization
      Try QST2 or QST3
      Relaxed scan along coordinate to loo k for highest
      energy (Opt=ModRedundant)

				
DOCUMENT INFO
Shared By:
Categories:
Stats:
views:103
posted:4/1/2011
language:English
pages:20