Get documents about "Using_Gaussian
Document Sample
Using Gaussian 03 at MCSR
Brian W. Hopkins
Mississippi Center for Supercomputing Research
19 February 2009
What We’re Doing Here
• General information: license, &c.
• Input file structure
• The g03sub utility
• Job naming
• Reading the output file
Gaussian 03
• “Starting from the basic laws of quantum mechanics,
Gaussian predicts the energies, molecular structures,
and vibrational frequencies of molecular systems,
along with numerous molecular properties derived
from these basic computation types. It can be used to
study molecules and reactions under a wide range of
conditions, including both stable species and
compounds which are difficult or impossible to
observe experimentally such as short-lived
intermediates and transition structures.” –
Gaussian.com
Site License
• Gaussian 03 is a commercial program.
• UM and MCSR have a complete site
license for Gaussian 03.
– available for all UM/MCSR computers
– source code and precompiled binaries
– latest revisions when they are released
– support
• through Brian: assist@mcsr.olemiss.edu
Parallel Versions & Licenses
• G03 has a native OpenMP
parallelization layer for SMP systems.
• For distributed-memory systems, G03
relies on Linda for parallelism.
• Linda is distributed under a separate
and more restrictive license.
– Ask about Linda-enabled G03 at
assist@mcsr.olemiss.edu
Installing Gaussian Locally
• The site license means that you can
build G03 on your lab computers if you
want it.
• Contact MCSR staff for media.
– Source code
– Binary versions
• Be aware that processor, OS and
compiler combinations are very
restrictive.
Using Gaussian at MCSR
• G03 is installed on all MCSR systems:
– Sweetgum (older SGI Origin 2000)
– Mimosa (cluster of P4 Linux boxes)*
– Redwood (SGI Altix 3700)
– Sequoia (SGI Altix XE hybrid cluster)
• Any MS researcher can get accounts
and use Gaussian.
• Accounts are set up to run G03 w/o
modification when they are created.
Doing Science with G03 at
MCSR
• Login
• Move to /ptmp/$USER
• Set up desired directory tree
• Build input files
• Submit jobs with g03sub
• Watch queue with qstat
• Monitor progress until job completes
G03, UNIX, and You
• All MCSR systems have some UNIX-
like operating system.
– IRIX on sweetgum
– openSuSe on mimosa
– SLES on redwood, sequoia
• The interface to G03 happens in text
files and the UNIX command line.
• MCSR staff offer training courses for
users new to UNIX environments.
Disk Space on MCSR
Systems
• We try to set up the filesystem
environment as uniformly as possible
• All systems have home areas (~),
permanent temporary areas (/ptmp) and
some form of scratch space (/tmp,
/work).
• The precise structure depends on the
machine.
~ and /ptmp
• Disk quotas in home areas are small.
– Files to customize the environment
• .cshrc
• .alias
– Some scripts, &c.
– Symbolic links to /ptmp
• /ptmp areas are larger
– Data storage
– Program code & executables
Building Input Files
%nprocl=4
%mem=900mb
%chk=methane.chk
#P MP2(FC)/6-31++G** 5D OPT(tight) SCF(tight) 5d
MP2/6-31++G** optimization of methane
0,1
C 0.0000000000 0.0000000000 -0.0000000318
H 0.0000000000 0.0000000000 1.0840336982
H 1.0220367932 0.0000000000 -0.3613445025
H -0.5110183966 -0.8851098265 -0.3613445025
H -0.5110183966 0.8851098265 -0.3613445025
Program Control Options
%nprocl=4 The number of processors
%mem=900mb The amount of memory needed
%chk=methane.chk The name of the checkpoint file
#P MP2(FC)/6-31++G** 5D OPT(tight) SCF(tight) 5d
MP2/6-31++G** optimization of methane
0,1
C 0.0000000000 0.0000000000 -0.0000000318
H 0.0000000000 0.0000000000 1.0840336982
H 1.0220367932 0.0000000000 -0.3613445025
H -0.5110183966 -0.8851098265 -0.3613445025
H -0.5110183966 0.8851098265 -0.3613445025
Requesting Multiple Processors
• Gaussian 03 as 2 different ways to
request multiple processors
– %nprocs=X is used to request X
processors of a shared-memory machine
(sweetgum, redwood, sequoia*)
– %nprocl=X is used to request X nodes of a
distributed memory machine (mimosa)
• In principle they can be used together,
but we don’t have any machines on
which that would be appropriate.
Memory Request
• Can be specified in words or bytes:
– KW, MW, GW
1W = 8B
– KB, MB, GB
• If unitless, the default unit is single
words.
• Different treatments for parallel jobs:
– If %nprocs, give the total memory for all
procs
– If %nprocl, give the memory for each node.
The Checkpoint File
• Stores information needed to restart the calculation
– Information the defines the current wavefunction and derivatives
thereof
– Most recent geometry
• Especially useful in geometry optimizations
– Restarting timed-out opts
– Speeding opts with the opt-freq-opt cycle
• If no path is given, the program will look for a file with
this name in the working directory
• If no such file exists:
– If nothing on the command line requires a checkpoint file, job
will start from scratch and create a new file
– If something on the command line (geom=check;
guess=read) needs the file, the job will crash.
The Command Line
%nprocl=4
%mem=900mb
%chk=methane.chk
#P MP2(FC)/6-31++G** 5D OPT(tight) SCF(tight) 5d The command line
MP2/6-31++G** optimization of methane
0,1
C 0.0000000000 0.0000000000 -0.0000000318
H 0.0000000000 0.0000000000 1.0840336982
H 1.0220367932 0.0000000000 -0.3613445025
H -0.5110183966 -0.8851098265 -0.3613445025
H -0.5110183966 0.8851098265 -0.3613445025
General Features
• Starts w/ #
• Next is 1-letter code to specify verbosity
• After that is long string to direct job
• Not case sensitive
• Job parameters can be in any order
– Best to choose a structure and stick with it
The Comment Line
%nprocl=4
%mem=900mb
%chk=methane.chk
#P MP2(FC)/6-31++G** 5D OPT(tight) SCF(tight) 5d
1 blank line
MP2/6-31++G** optimization of methane The comment line
1 blank line
0,1
C 0.0000000000 0.0000000000 -0.0000000318
H 0.0000000000 0.0000000000 1.0840336982
H 1.0220367932 0.0000000000 -0.3613445025
H -0.5110183966 -0.8851098265 -0.3613445025
H -0.5110183966 0.8851098265 -0.3613445025
The Comment Line
• Must be present
• Can’t be blank
• Will be echoed into output, and so can
serve as a sort of “label” for an output
file
• Usually more trouble than it’s worth
Charge, Multiplicity
%nprocl=4
%mem=900mb
%chk=methane.chk
#P MP2(FC)/6-31++G** 5D OPT(tight) SCF(tight) 5d
MP2/6-31++G** optimization of methane
0,1 Charge, muliplicity
C 0.0000000000 0.0000000000 -0.0000000318
H 0.0000000000 0.0000000000 1.0840336982
H 1.0220367932 0.0000000000 -0.3613445025
H -0.5110183966 -0.8851098265 -0.3613445025
H -0.5110183966 0.8851098265 -0.3613445025
Charge and Muliplicity
• Charges w/o sign will be considered
positive
• Program will automatically try to
reconcile structure, charge, multiplicity,
and reference (if specified)
– Failure will kill the job:
The combination of multiplicity 2 and 16 electrons is impossible.
Error termination via Lnk1e in /usr/local/apps/g03/l301.exe.
• Especially for open-shell systems, be
sure to check state after the job.
The Molecule
%nprocl=4
%mem=900mb
%chk=methane.chk
#P MP2(FC)/6-31++G** 5D OPT(tight) SCF(tight) 5d
MP2/6-31++G** optimization of methane
0,1
C 0.0000000000 0.0000000000 -0.0000000318
H 0.0000000000 0.0000000000 1.0840336982
H 1.0220367932 0.0000000000 -0.3613445025 Molecular structure
H -0.5110183966 -0.8851098265 -0.3613445025
H -0.5110183966 0.8851098265 -0.3613445025
Molecular Structure
• Can be given in various coordinate
systems:
– Cartesian
– Internal (Z-Matrix)
– Redundant Internal (Z-Matrix + additional
coords)
• Generally no requirement to orient
structure in any particular way, or to
define internals wrt symmetry elements
Symmetry Elements?
• Consider two structures:
O O
H 1 oh1 H 1 oh1
H 1 oh1 2 hoh1 H 1 oh2 2 hoh1
oh1 = 0.95 oh1 = 0.95 Recognized as de facto
hoh1 = 109.5 oh2 = 0.95
hoh1 = 109.5
C2v symmetry
• Both will be recognized by G03 as C2v;
optimizations will be automatically
constrained accordingly.
Blank Line At The End
• Required if molecule given as
Cartesians
• The only way for G03 to know that the
molecule is done.
• One of the most common problems for
new users of G03.
• Marked by this error message:
End of file in ZSymb.
Error termination via Lnk1e in /usr/local/apps/g03/l101.exe
Specialty Topics For Another
Time
• Custom basis sets
• Redundant coordinates
• Constrained opts & PES scans
• CP corrections
• Symmetry controls
Submitting Jobs With g03sub
• Because MCSR machines are shared by large
numbers of users, we use a batch system to control
the use of processors, memory, &c.
• The batch system is called PBS; users run jobs by
writing special scripts that are then submitted to a
queue.
• Due to the large volume of G03 jobs, we have
created a script that takes a simple command line
syntax and automatically generates and submits a
PBS job script.
The Basic G03sub Syntax
g03sub -n procs -m mem –t time <-d disk> file.ext
Where:
•procs is the number of nodes or processors requested
•mem is the amount of memory requested
•time is the amount of time requested
•disk (optional) is the amount of disk needed
•file.ext is the name of the job input file
Scratch Directories
• Certain G03 jobs will require scratch space to
read and write data during the job.
• This data is not kept after job completion.
• Performance of the scratch disk system can
be limiting factor in total job performance.
• All systems have default settings for scratch
disk that can be overridden as needed.
• Invoking the -d flag followed by an estimate of
the disk space needed by the job will help our
systems direct jobs to the best scratch
systems.
Consistency Checks
• G03sub automatically checks to make
sure some features of your job are
consistent and system appropriate:
– %nprocl vs. %nprocs
– total # of processors (limit 4)
– -n vs. %nproc
– -m vs. %mem
• Jobs that fail the checks will return
intuitive error messages
.out and .OUT
• A G03 job with input file file.ext is going
to create output files file.out and file.log
• To protect preexisting data, g03sub
scans the working directory and moves
files to safety:
– File.out file.OUT
– File.log file.LOG
• WARNING: There is no third file name;
further job submissions will overwrite
data.
Script Creation and Submission
• Once the consistency checks have been passed,
g03sub automatically creates a PBS script and
submits it.
• Accepted submissions will report job numbers:
7244.mimosa.mcsr.olemiss.edu
• Some submissions may be rejected by PBS itself,
usually because the user lacks the proper queue
access:
qsub: Job rejected by all possible destinations
Checking on a Job
• Once a job is submitted, it can be
checked on using qstat.
– qstat -u user
mimosa(no_blank_line)% qstat -u r1130
mimosa.mcsr.olemiss.edu:
Req'd Req'd Elap
Job ID Username Queue Jobname SessID NDS TSK Memory Time S Time
--------------- -------- -------- ---------- ------ --- --- ------ ----- - -----
7166.mimosa.mcs r1130 MCSR-2N DX101pcbZO 9291 2 2 400mb 12:00 R 00:42
7167.mimosa.mcs r1130 MCSR-2N DX002pcbZO -- 2 2 400mb 12:00 Q --
– qstat -f jobnumber
• returns all info for a job
The Job Naming Scheme
• Jobs are assigned a jobname by PBS
that’s different from the anme of the
input file.
• G03sub assigns job names by
combining a two-character descriptive
code and the name of the input file
(truncated as needed)
7175.mimosa E132a11-15 r1221 0 Q MCSR-2N
A Quick Guide to the Naming
Convention
• The first character indicates the broad theoretical
method you’re using:
– H = Hartree-Fock
– D = Density Functional Theory
– P = Perturbation Theory
– I = Configuration Interaction Theory
– C = Coupled-Cluster Theory
– E = Empirical Theories
– X = Method not recognized by g03sub
• The second character indicates the order of the
derivative taken
– 0 = energy only
– 1 = gradient (opt, &c.)
– 2 = 2nd derivative (freq, &.)
– X = syntax not recognized by g03sub
Reading the Output File
• Once a job is finished, it will produce an output file
with the same root name as the input file and a .out
extension
• This file contains all the information produced by the
job:
– Orbital energies and occupations
– Reference and correlated energies
– Optimized geometries
– Vibrational frequencies
– Thermochemical data
• Be careful that you’re extracting the right information
from the file!
Some Important Output Features:
Energy Points
• The reference energy:
SCF Done: E(RHF) = -76.0307499791 A.U. after 10 cycles
Convg = 0.8565D-08 -V/T = 2.0020
S**2 = 0.0000
• Correlated energies:
\1\GINC-NODE4-1\SP\RCCSD-FC\6-31++G(d,p)\H2O1\BWHOPKIN\19-Feb-
2009\0\\#CCSD/6-31++G** 5D scf(tight)\\wah da
tah!\\0,1\O\H,1,0.95\H,1,0.95,2,109.5\\Version=IA32L-
G03RevE.01\State=1A1\HF=-76.03075\MP2=-76.2303342\MP3=-
76.2352666\MP4D=-76.2387725\MP4DQ=-76.2371357\MP4SDQ=-
76.2384984\CCSD=-76.2384892\RMSD=8.565e-09\Thermal=0.\PG=C02V
[C2(O1),SGV(H2)]\\@
Some Important Output Features:
Optimizations
• Convergence Information:
Item Value Threshold Converged?
Maximum Force 0.000013 0.000015 YES
RMS Force 0.000012 0.000010 NO
Maximum Displacement 0.000029 0.000060 YES
RMS Displacement 0.000027 0.000040 YES
• Final Geometry:
Final structure in terms of initial Z-matrix:
O
H,1,oh1
H,1,oh2,2,hoh1
Variables:
oh1=0.94306581
oh2=0.94306581
hoh1=107.088157
Some Important Output Features:
Frequency Calcualtions
• Imaginary modes:
Full mass-weighted force constant matrix:
Low frequencies --- -304.6041 -22.3249 -16.3689 -16.3688 0.0002 0.0004
Low frequencies --- 0.0008 996.8348 996.8348
****** 1 imaginary frequencies (negative Signs) ******
Diagonal vibrational polarizability:
0.4660701 0.4660701 0.2475471
• Thermochemical Data:
E (Thermal) CV S
KCal/Mol Cal/Mol-Kelvin Cal/Mol-Kelvin
Total 16.291 5.995 44.947
Electronic 0.000 0.000 0.000
Translational 0.889 2.981 34.608
Rotational 0.889 2.981 10.334
Vibrational 14.513 0.033 0.004
Q Log10(Q) Ln(Q)
Total Bot 0.280164D-02 -2.552587 -5.877549
Total V=0 0.121603D+09 8.084945 18.616274
Vib (Bot) 0.230447D-10 -10.637429 -24.493586
Vib (V=0) 0.100024D+01 0.000103 0.000237
Electronic 0.100000D+01 0.000000 0.000000
Translational 0.300432D+07 6.477746 14.915562
Rotational 0.404665D+02 1.607096 3.700475
