Mopac 5.09mn manual MOPAC 5.09mn by James J. P. Stewart, Ivan Rossi, Wei-Ping Hu, Gillian C. Lynch, Yi-Ping Liu, Yao-Yuan Chuang, Jiabo Li, Christopher J. Cramer, Patton L. Fast, and Donald G. Truhlar, based on MOPAC 5.0 by James J. P. Stewart Package revision date: October 18, 1999 Date of most recent manual update: October 18, 1999 This version of MOPAC is based on MOPAC 5.0, but it has been extensively modified at the University of Minnesota (i) to be portable, (ii) to contain additional capabilities, and (iii) to be suitable for interfacing with molecular dynamics programs like POLYRATE. With regard to (i) the code is now fully compliant with FORTRAN 77 except that it uses the INCLUDE extension, which is widely available, and it runs on supercomputers, Unix workstations, and PCs running Linux. The scope of this manual is to document the modifications and additions to the original MOPAC 5.0 code. To learn how to use MOPAC, please refer to the original MOPAC manual, which is distributed as part of the MOPAC 5.09mn package. The most important additional capabilities added to the code are inclusion of PM3 (as in MOPAC 6), flexible options for external modification of parameters (specific reaction parameters), class IV charges by CM2A and CM2P, the eigenvector following (EF) algorithm for geometry optimization of equilibrium structures and saddle points, and improved configuration interaction options. TABLE OF CONTENTS 1. USER AGREEMENT ................................................. 1 2. REFERENCES ..................................................... 2 3. REVISION HISTORY ............................................... 2 4. NEW CAPABILITIES ............................................... 4 4.1. NEW CONFIGURATION INTERACTION KEYWORD ..................... 4 4.2. THE EIGENVECTOR FOLLOWING METHOD FOR OPTIMIZATION ......... 4 4.2.1. Description of the Algorithm ....................... 5 4.2.2. New Eigenvector Following Keywords ................. 5 4.2.3. How to Fine Tune Optimizations ..................... 7 4.3. NDDO-SRP CALCULATIONS ..................................... 9 4.4. CHARGE MODEL 2 ............................................ 10 5. INSTALLATION ................................................... 10 5.1. COMPILING ................................................. 10 5.2. FILES INCLUDED IN DISTRIBUTION ............................ 11 6. TEST RUNS ...................................................... 12 6.1. RUNNING TEST RUNS ......................................... 12 6.2. DESCRIPTION OF TEST RUNS .................................. 13 1. USER AGREEMENT This code is supplied with the following conditions. The code, except for unmodified code from MOPAC-version 5.0 and the EF-related subroutines, is copyrighted by the authors who retain all rights for distribution. Persons receiving the code from the authors or from authorized distributors are Mopac 5.09mn Page 2 encouraged to share it with their co-workers but are asked not to redistribute it (in whole or in part, in modified or unmodified form, alone or as a part of another program) to third parties. Users may make additional copies for their own research use, including usage by coworkers, but are asked to retain the code name and version number, author names, copyright notice, and user agreement notice on the output file. Publications resulting from use of the MOPAC 5.09mn code should give the reference listed in section 2.0. 2. REFERENCES Reference for MOPAC 5.09mn: J. J. P. Stewart, I. Rossi, W.-P. Hu, G. C. Lynch, Y.-P. Liu, Y.-Y. Chuang, J. Li, C. J. Cramer, P. L. Fast, and D. G. Truhlar, MOPAC-version 5.09mn, University of Minnesota, Minneapolis, 1999. 3. REVISION HISTORY As explained in the MOPAC-version 5.0 manual, the first two digits (e.g., 5.0) of the MOPAC version number are constant for a given release of the program. Modified versions for a given release of MOPAC should have the third digit different from zero; for example, a modification of version 5.0 would result in a version with the version number 5.0x, where x is not equal to zero. Clearly this can result in multiple versions of 5.0x in different places; therefore we have added mn (Minnesota) to versions created at the University of Minnesota. This document provides the revision history for the University of Minnesota revisions leading up to this version. Version 5.0 Original distributed version by James J. P. Stewart. Version 5.01mn Modified version of MOPAC-version 5.0. The modifications were made by Minnesota Supercomputer Center Inc. for the Cray computers. Version 5.02mn Modification of version 5.01mn. In this version all the subprograms were returned to their original form, which used the include extension and the include file SIZES. This was done so that dimension changes would be much easier in MORATE. (In version 5.01mn the statements which made up the include file SIZES were explicitly included in all subprograms.) Mopac 5.09mn Page 3 Version 5.03mn Modification of version 5.02mn. The block data file was modified to include all the AM1 and PM3 parameters available in MOPAC-version 6.0. Save statements were added to all the subprograms so that this version of the code does not have to be compiled with either the static or the ev option on the various Crays. The include file SIZES was renamed SIZES.i and all include statements in the FORTRAN subprograms were modified accordingly; this change was made for portability purposes. Version 5.04mn Modification of version 5.03mn. The subprograms AM1 and MOLDAT have been modified so that the semiempirical parameters for all of the Hamiltonians (MINDO/3, MNDO, AM1, and PM3) can be modified by using the keyword EXTERNAL (In version 5.0 only the MNDO and AM1 Hamiltonian could be modified). Version 5.05mn (July 1994) J. J. P. Stewart, I. Rossi, W.-P. Hu, G. C. Lynch, Y.-P. Liu, and D. G. Truhlar Modification of version 5.04mn. Several new capabilities have been added, and the code has been modified to make it portable and ANSI FORTRAN-77 compliant. A bug which caused the program to crash when parameter MAXORB <= NMECI**2 has been corrected. A bug in ANALYT which caused PM3 analytical derivatives to malfunction has been corrected. The C.I.=(n,m) option and the Eigenvector Following optimization algorithm have been introduced as new capabilities. Subroutines H1ELEC, ANALYT, MULLIKEN, and AM1 (now renamed EXTPAR) have been modified to use the new NDDO-SRP pairwise parameters BETSS, BETSP and BETPP. Relevant to this new feature is the addition to the SIZES.i file of the new user-defined parameter MXATSP. Version 5.06mn (September 1997) J. J. P. Stewart, I. Rossi, W.-P. Hu, G. C. Lynch, Y.-P. Liu, and D. G. Truhlar Modification of version 5.05mn. DATA statements in the EXTPAR and ROTATE subroutines have been moved so that they are the last non-executable statements in the respective subroutine to adhere to standard Fortran 77. Dummy routines, date_dum.f and second_dum.f, and the makefile make.linux have been created to run MORATE under the Linux operating system. Version 5.07mn (December 1997) J. J. P. Stewart, I. Rossi, W.-P. Hu, G. C. Lynch, Y.-P. Liu, and D. G. Truhlar Modification of version 5.06mn. Files date_linux.f, second_liunx.f, second1.c are added to replace the dummy routines for Linux operating system. Some routines and makefiles for Sun, DEC, and HP workstations were added without testing. Version 5.08mn (September 1999) J. J. P. Stewart, I. Rossi, W.-P. Hu, G. C. Lynch, Y.-P. Liu, Y.-Y. Chuang, J. Li, C. J. Cramer, and D. G. Truhlar Modification of version 5.07mn. New files PARAMS.i and chgmp2.f are added to include class IV atomic charges calculated by Charge Model 2 (CM2). The subroutines CHRGE, WRTKEY, and WRITEMO are modified, and a new subroutine SCOPY is added to fromblas.f. Two new test runs test13.dat and test14.dat, are added for evaluating partial atomic charges by the CM2A and CM2P methods, respectively. Mopac 5.09mn Page 4 Version 5.09mn (October 1999) J. J. P. Stewart, I. Rossi, W.-P. Hu, G. C. Lynch, Y.-P. Liu, Y.-Y. Chuang, J. Li, C. J. Cramer, P. L. Fast, and D. G. Truhlar Modification of version 5.08mn. Tested make files have been provided for Compaq and Sun computers, and the subroutine WRTKEY has been modified to correct a floating point error on Compaq and Sun machines. The value of the parameter MAXDMP, defined in SIZEZ.i, was increased from 3600 to 36000. 4. NEW CAPABILITIES 4.1. NEW CONFIGURATION INTERACTION KEYWORD C.I.=(n,m) In addition to specifying the number of M.O.'s in the active space, the number of electrons can also be defined. In C.I.=(n,m), n is the number of M.O.s in the active space, and m is the number of doubly filled orbitals to be included in the active space. Examples: Keywords Number of M.O.s No. of Electrons C.I.=2 2 2 (1) C.I.=(2,1) 2 2 (3) C.I.=(3,1) 3 2 (3) C.I.=(3,2) 3 4 (5) C.I.=(3,0) OPEN(2,3) 3 2 (N/A) C.I.=(3,1) OPEN(2,2) 3 4 (N/A) C.I.=(3,1) OPEN(1,2) 3 N/A (3) Values for odd-electron systems are given in parentheses, following the number for an even number of electrons. N/A denotes not applicable. This modification is based on the MOPAC 6.0 code. 4.2. THE EIGENVECTOR FOLLOWING METHOD FOR OPTIMIZATION The Eigenvector Following code included in MOPAC 5.09mn is based on code developed by: Dr. Frank Jensen Department of Chemistry Odense University 5230 Odense Denmark and is taken from the MOPAC 7 code. Mopac 5.09mn Page 5 4.2.1. Description of the Algorithm The EF optimization routine used here is a combination of the original EF algorithm of Simons et al[1] as implemented by Baker[2] and the QA algorithm of Culot et al[3], with some added features (see RMIN, RMAX, and OMIN described below) for improving stability. [1] A. Banerjee, N. Adams, J. Simons and R. Shepard, J. Phys. Chem. 89 (1985), 52. [2] J. Baker, J. Comp. Chem. 7 (1985), 385. [3] P. Culot, G. Dive, V. H. Nguyen and J. M. Ghuysen, Theo. Chim. Acta 82 (1992), 189. The geometry optimization is based on a second order Taylor expansion of the energy around the current point. At this point the energy, the gradient, and some estimate of the Hessian are available. There are three fundamental operations in determining the next geometry based on this information: o find the ``best'' step within or on the hypersphere with the current trust radius. o possibly reject this step based on various criteria. o update the trust radius. For a minimum search the correct Hessian has only 3N-6 positive eigenvalues, where N is the number of atoms. For a Transition State (TS) search the correct Hessian should have exactly one negative eigenvalue, and the corresponding eigenvector should be in the direction of the desired reaction coordinate. The geometry step is parameterized as g/(s-H), where s is a shift factor which ensures that the step length is within or on the hypersphere. If the Hessian has the correct structure, a pure Newton-Raphson step is attempted. This corresponds to setting the shift factor to zero. If this step is longer than the trust radius, a P-RFO step is attempted. If this is also too long, then the best step on the hypersphere is made via the QA formula. Using the step determined, the new energy and gradient are evaluated. If it is a TS search, two criteria are used in determining whether the step is "appropriate". The ratio between the actual and predicted energy change should ideally be 1. If it deviates substantially from this value, the second order Taylor expansion is no longer accurate. If the ratio is outside the interval defined by the RMIN and RMAX limits, the step is rejected, the trust radius reduced by a factor of two and a new step is determined. The second criteria is that the eigenvector along which the energy is being maximized should not change substantially between iterations. The minimum overlap of the TS eigenvector with that of the previous iteration should be larger than OMIN, otherwise the step is rejected. 4.2.2. New Eigenvector Following Keywords EF The Eigenvector Following routine is an alternative to the default BFGS algorithm. This keyword invokes the Eigenvector Following routine to optimize Mopac 5.09mn Page 6 to a minimum energy structure. EF is particularly efficient for refining structures, when the gradient is already small. TS With the TS keyword, the Eigenvector Following routine is invoked to optimize to a transition state. The TS method is much faster and more reliable than either SADDLE or NLLSQ. TS appears to work well even with Cartesian coordinates. MODE=n MODE=n specifies that the nth Hessian eigenvector will be followed in the first step of an Eigenvector Following optimization. MODE=1 means the eigenvector with the lowest eigenvalue, MODE=2 the second lower, and so on. Note that the eigenvectors corresponding to translational and rotational motion, which have zero eigenvalue, are projected out of the Hessian and automatically renumbered as the last six eigenvectors. The next steps will be selected on the basis of the overlap between two following steps (see OMIN). If MODE=0 the eigenvector with the lowest eigenvalue will be followed, regardless of the overlap with the previous optimization step. The defaults are MODE=1 (TS keyword) and MODE=0 (EF keyword). HESS=n HESS=n specifies how the Hessian matrix will be calculated. HESS=0 The initial Hessian will be approximated as diagonal HESS=1 Calculate the Hessian using forward finite differences HESS=2 Read Hessian from disk HESS=3 Calculate the Hessian using central finite differences The default is HESS=0 for minimum optimization (EF keyword) and HESS=1 for transition state optimization (TS keyword) RECALC=n In an Eigenvector Following optimization, the RECALC=n keyword requests the program to recalculate the Hessian every n iterations. This is very effective but CPU-intensive. The Hessian will be recalculated using the method specified by HESS keyword. DMAX=n DMAX=n changes the value of the starting trust radius (in Angstroms) in Eigenvector Following optimizations. The default is DMAX=0.2 DDMIN=n DDMAX=m These keywords set the limits for trust radius (in Angstroms) in Eigenvector Following optimizations. The defaults are DDMIN=0.001 and DDMAX=0.3 (TS keyword) or DDMAX=0.5 (EF keyword). RMIN=n RMAX=m For an Eigenvector Following step to be accepted, the value of the ratio of the calculated energy to the predicted energy must be bracketed by the values of RMIN and RMAX. Default values are RMIN=0 and RMAX=4. Mopac 5.09mn Page 7 OMIN=n During transition state optimizations, the algorithm calculates the dot product between the previously followed direction and the Hessian eigenvectors. The new step will be along the direction defined by the eigenvector for which this dot product is maximum, if this value is greater than OMIN. The default is OMIN=0.8 (TS keyword) and OMIN=0.0 (EF keyword). See also keyword MODE IUPD=n IUPD=n selects the Hessian updating scheme in Eigenvector Following optimizations. IUPD=0 No updating IUPD=1 Powell updating scheme[4] IUPD=2 BFGS updating scheme[5] [4] M. J. D. Powell, Math. Prog., 1 (1971), 26. [5] R. Fletcher, Practical Methods of Optimization: Unconstrained Optimization, Vol. 1, Wiley, New York (1980). The defaults are IUPD=1 for transition state search (TS keyword) and IUPD=2 for minimum (EF keyword) RSCAL RSCAL scales the Eigenvector Following step to trust radius instead of using QA formula. The default is to use QA formula for scaling. NONR Skip Newton-Raphson step in Eigenvector Following optimizations. See section 2.2.1. 4.2.3. How to Fine Tune Optimizations RMIN and RMAX: The acceptance criterion for the optimization step is that the ratio of the calculated energy to the predicted energy should be larger than RMIN and lower than RMAX. If this ratio is outside this interval, the step is rejected, the trust radius reduced by a factor of two and a new step is predicted. Setting RMIN and RMAX close to one will give a very stable, but also very slow, optimization. Wide limits on RMIN and RMAX may in some cases give a faster convergence, but there is always the risk that very poor steps are accepted, causing the optimization to diverge. The use of the default values of 0 and 4 lead to a rare rejection of steps. This leads to a faster convergence, but occasionally leads to the acceptance of poor steps. If TS searches are found to cause problems, the first try should be to change the limits of RMIN and RMAX, narrowing the interval to 0.5 and 2, for example. Tighter limits like 0.8 and 1.2, or even 0.9 and 1.1, will almost always slow the optimization down significantly, but they may be necessary in some cases. OMIN: OMIN has been designed for ensuring that the nature of the TS mode only Mopac 5.09mn Page 8 changes gradually, specifically the overlap between two successive geometrical displacement should be higher than OMIN. While this technique at first appears very promising, it may cause problems when the Hessian is updated. As the updated Hessian in each step is only approximately correct, there is an upper limit on how large the TS mode overlap between steps can be. To understand this, consider a series of steps made from the same geometry (e.g. at some point in the optimization), but with steadily smaller step sizes. The update adds corrections to the Hessian to make it a better approximation to the exact Hessian. As the step size becomes small, the updated Hessian converges toward the exact Hessian, at least in the direction of the step. The overlap between TS modes does not converge toward 1, but rather to a constant value which indicate how good a guess the first approximate Hessian was to the exact Hessian. It appears that an updated Hessian in general is not of sufficient accuracy for reliably rejecting steps with TS overlaps much greater than 0.80. The default OMIN of 0.80 reflects the typical use of an updated Hessian and allows fairly large changes to occur, and should be suitable for most uncomplicated systems. If problems are encountered with many step rejections due to small TS mode overlaps, try reducing OMIN, maybe all the way down to 0. This most likely will work if the TS mode is the lowest Hessian eigenvector, but it is doubtful that it will produce any useful results if a high lying mode is followed. Note that the only way to turn off the step rejection criteria is to give suitable values to RMIN, RMAX, and OMIN, e.g. the choice of RMIN=-100 and RMAX=100 effectively inhibits step rejection. Similarly setting OMIN=0 disables step rejection based on large changes in the structure of the TS mode. MODE: The algorithm has the capability of following Hessian eigenvectors other than the one with the lowest eigenvalue (see MODE keyword explanation) toward a TS, using the keyword MODE. It is always more difficult to make such higher- mode following. Ideally, as the optimization progresses, the TS mode should at some point become the lowest eigenvector. Care must be taken during the optimization, however, that the nature of the mode does not change suddenly, leading to optimization to a different TS than the one desired. Note that during TS optimizations, the default value MODE=1 means that mode following is active (See OMIN and MODE keyword explanation). This means that the TS MODE=1 will be followed, and in some cases this may eventually change to some higher mode, causing the optimization to fail. To turn off mode following, and thus following at every step the mode with lowest eigenvalue, set MODE=0. Remember that following modes other than the one with the lowest eigenvalue toward a transition state indicates that the starting geometry is not a good guess of the transition state one. In most cases it is better to further refine the starting geometry, than to try following high-lying modes. There are cases, however, where it is very difficult to locate a starting geometry which has the desired Hessian, and higher-mode following may be useful. Otherwise, if RECALC=1 the TS mode overlap does converge toward 1 as the step size goes toward zero, and in this cases there is no problems following high lying modes. HESS and RECALC: In certain very rigid systems or in some transition state optimizations, the initial default Hessian may result to be too approximate. In this case the Mopac 5.09mn Page 9 algorithm cannot find an acceptable step larger than DDMIN, so the optimization terminates after only a few cycles with the ``TRUST RADIUS BELOW (DDMIN value)" warning long before the stationary point is reached. In such cases RMIN could be set to some negative value, thereby allowing steps which increase the energy. But in general it is more useful to increase the precision of the Hessian calculation using the HESS keyword. In some very difficult cases it is even necessary to recalculate the Hessian every few iterations, using the RECALC keyword. Unfortunately setting RECALC to low values is very expensive in terms of computer time, but if used in conjecture with OMIN=0.90 (or possibly an even higher value) and maybe with tighter values of RMIN and RMAX, it represents an option for locating transition structures that otherwise might not be possible. 4.3. NDDO-SRP CALCULATIONS Neglect-of-diatomic-differential-overlap (NDDO) molecular orbital theory with specific reaction parameters (SRP) can be invoked in this program in the following way. An NDDO-SRP calculation, which can be used in conjunction with MNDO, AM1, or PM3 calculations, can be invoked by using the MOPAC keyword EXTERNAL=filename. The parameters adjusted specifically for the given reaction should be defined in the file given in EXTERNAL=filename according to the guidelines in the MOPAC 5.0 manual. In order to allow the wavefunction to be more flexible, three new parameters, BETSS, BETSP, and BETPP, have been defined and are recognized in an NDDO-SRP calculation. In standard calculations the so called one-electron resonance integral of NDDO theory is calculated as the overlap integral of the two atomic orbitals times a scaling factor, which is defined as the mean of two atomic BETA parameters. In MOPAC 5.09mn, however, every scaling factor can be specified separately for every pair of atomic species, specifying the BETxy parameters in the EXTERNAL file, as: BETxy atom1 atom2 value where: x = type of orbital of atom type atom1 (only S-type and P-type orbitals available) y = type of orbital of atom type atom2 (only S-type and P-type orbitals available) atom1 = chemical symbol atom2 = chemical symbol Example: BETSS C Br -17.557832 BETPP C Br -8.338239 BETSP Br C -13.559582 BETSP C Br -12.336489 Note that BETSP C Br is a different parameter from BETSP Br C. The value of BETxy should be in eV units. These parameters must be specified in the EXTERNAL file after all the other Mopac 5.09mn Page 10 standard parameters. See test run 10 for an example, and in particular see file TEST10.SRP in the mopac509/test directory. Setting BETSS for C and Br to the average of the standard BETAS for C and Br is equivalent to doing nothing. Reference for NDDO-SRP with independent beta values: J. C. Corchado, J. Espinosa-Garcia, W.-P. Hu, I. Rossi, and D. G. Truhlar, J. Phys. Chem. 99 (1995), 687-694. 4.4. Charge Model 2 (CM2) The keyword CM2 specifies that Charge Model 2 (CM2) should be used. CM2 parameterizations are available in MOPAC for either the AM1 or PM3 Hamiltonian. CM2 with AM1 is called CM2A or CM2/AM1; CM2 with PM3 is called either CM2P or CM2/PM3. When the CM2 keyword is chosen in addition to the AM1 or PM3 Hamiltonian, the Charge Model 2 charges and dipole moment will be be printed. CM2 results will replace the Mulliken charges in the FOR12 file with MOPAC input file format. CM2 is parameterized for H, C, N, O, F, Si, P, S, Cl, Br, and I. Reference for Charge Model 2: J. Li, T. Zhu, C. J. Cramer, and D. G. Truhlar, J. Phys. Chem. A 102, 1820-1831 (1998). 5. Installation 5.1. Compiling The MOPAC 5.09mn distribution package can be obtained from the University of Minnesota in two forms: 1)tarred (mopac509mn.tar) or 2) tarred and compressed (mopac509mn.tar.Z). The mopac509mn.tar.Z file can be uncompressed by typing the following: uncompress mopac509mn.tar.Z After uncompressing, the tar file, mopac509mn.tar, is created in the current working directory. The mopac509mn.tar file can be untarred by typing the following: tar xvf mopac509mn.tar This creates the directory mopac509 in the current working directory. The tar file remains in the current working directory. This file should be saved and backed up for future reference or re-installation. Building MOPAC 5.09mn involves two commands. First type the following command: cd mopac509/src Mopac 5.09mn Page 11 to change the working directory to the directory that contains all the source code. Then issue the command: make -f make.xxx mopac509 where xxx is machine dependent. The makefiles, located in the mopac509mn/src directory, distributed with MOPAC 5.09mn are make.compaq (to compile on Compaq workstations) make.cray (to compile on Cray supercomputers) make.dec (to compile on DEC workstations make.hp (to compile on HP workstations) make.ibm (to compile on IBM RS/6000 and SP systems) make.linux (to compile on LINUX systems) make.sgi (to compile on Silicon Graphics systems) make.sun (to compile on Sun workstations) This will create the executable file mopac509mn. After the successful compilation, to save disk space, type: make -f make.xxx clean (where xxx is replaced by cray, ibm, etc.) in order to remove all the object files. 5.2. FILES INCLUDED IN DISTRIBUTION Directory mopac509 doc directory containing the manuals src directory containing the source code test directory containing the test input files testo directory containing the test output files Directory doc mopac.doc the original MOPAC 5.0 manual (ASCII file) mopac509.doc this document (ASCII file) Directory src SIZES.i include file betsrp.f BETxy auxiliary routines date_compaq.f machine-dependent time routines date_dec.f date_hp.f date_ibm.f date_linux.f date_sgi.f date_sun.f dateclock.c second_ibm.f second_linux.f second_sgi.f Mopac 5.09mn Page 12 second1.c fromblas.f source code of the BLAS routines used m509_ef.f EF-related subroutines m509_ef_cray.f m509_mdep_cray.f MOPAC machine-dependent routines m509_mdep_ws.f m509_mod.f MOPAC routines modified at Univ. of Minnesota m509_unmod.f unmodified MOPAC 5.0 routines mopac_main.f MOPAC 5.09mn main program make.compaq makefile for Compaq workstations make.cray makefile for Cray supercomputers make.dec makefile for DEC workstations make.hp makefile for HP workstations make.ibm makefile for IBM RS/6000 and SP systems make.linux makefile for LINUX systems make.sgi makefile for Silicon Graphics systems make.sun makefile for Sun workstations Directory test testsuite.sh the Bourne shell script to run the full test suite test1.dat MOPAC test input test2.dat MOPAC test input test3.dat MOPAC test input test4.dat MOPAC test input test5.dat MOPAC test input test6.dat MOPAC test input test7.dat MOPAC test input test8.dat MOPAC test input test9.dat MOPAC test input test10.dat MOPAC test input test11.dat MOPAC test input test12.dat MOPAC test input test13.dat MOPAC test input test14.dat MOPAC test input Directory testo test1.out test2.out test3.out test4.out test5.out test6.out test7.out test8.out test9.out test10.out test11.out test12.out test13.out test14.out 6. Test runs 6.1. Running test runs To run the complete test suite for MOPAC 5.09mn, change the current working directory to mopac509/test and execute the shell script testsuite.sh as follows: testsuite.sh After some minutes, a group of files called test##.out will be created in Mopac 5.09mn Page 13 the mopac509/test directory. You can check the results with the reference test suite outputs, named test##.out that are in the mopac509/testo directory. The reference output files were created on an SGI Origin 2000 with R10000 processors running the IRIX 6.5 operating system. The testsuite.sh script also creates a text file called test.timing containing timing information about the test runs. You can compare them with the reference timings given in the table below. The full test suite has been run successfully on the following machines: Computer Processors Operating system Makefile used Compaq ES40 Alpha 500 MHz Tru64 4.0F make.compaq IBM SP Power3 AIX 4.0 make.ibm Pentium Intel Pentium Pro Linux kernel 2.0.36 make.linux SGI ORIGIN 2000 R10000 IRIX 6.5 make.sgi Sun Enterprise 4500 UltraSparc-II Solaris 2.7 make.sun The timings have been done using the UNIX time command (/bin/time). User + System CPU time (s) Compaq IBM Pentium SGI Sun Test 1 0.6 1.6 7.0 1.7 1.8 Test 2 0.4 1.1 0.1 1.1 1.2 Test 3 0.4 1.0 4.9 1.1 1.2 Test 4 0.1 0.2 0.6 0.2 0.1 Test 5 4.8 11.1 58.6 10.8 12.8 Test 6 0.7 1.2 7.0 2.5 1.8 Test 7 0.0 0.0 0.1 0.4 0.0 Test 8 0.1 0.3 1.2 0.4 0.3 Test 9 0.5 1.2 5.1 1.2 1.2 Test 10 0.2 0.2 0.8 0.3 0.3 Test 11 0.1 0.2 0.8 0.3 0.2 Test 12 0.3 0.7 3.2 0.7 0.7 Test 13 0.4 0.3 1.4 0.4 0.2 Test 14 0.8 0.4 1.8 0.5 0.3 6.2. Description of test runs The input and output files for the fourteen MOPAC 5.09mn test runs are included in the mopac509mn/test and mopac509/testo directories, respectively. Below is a brief description of each of the test runs. Test run 1. The EF, GRADIENTS, NOINTER, PRECISE, and SYMMETRY keywords are specified for calculating the energy and optimized geometry of methylcyclohexane. The input file is mopac509mn/test/test1.dat. The reference output file is mopac509mn/testo/test1.out Test run 2. The AM1, EF, GNORM, GRAD, SCFCRT, and SYMMETRY keywords are specified for calculating the energy and optimized geometry of morpholine. The input file is mopac509mn/test/test2.dat. The output file is mopac509mn/testo/test2.out. Note: GRAD is a short form of GRADIENTS and means the same thing. Mopac 5.09mn Page 14 Test run 3. The GRADIENTS, MNDO, PRECISE, and SYMMETRY keywords are specified for calcualting the energy and optimized geometry of methyl butanoate. The input file is mopac509mn/test/test3.dat. The output file is mopac509mn/testo/test3.out. Test run 4. The AM1, CHARGE, PRECISE, SYMMETRY, and TS keywords are specified for calculating the energy and optimized geometry for the transition state of the SN2 reaction of bromide methyl iodide. The input file is mopac509mn/test/test4.dat. The output file is mopac509mn/testo/test4.out. Test run 5. The PM3, PRECISE, SYMMETRY, and TS keywords are specified for calculating the energy and optimized geometry of the transition state for the Diels-Alder reaction of cyclopentadiene and MVK. The input file is mopac509mn/test/test5.dat. The output file is mopac509mn/testo/test5.out. Test run 6. The C.I., GRADIENTS, MECI, PM3, ROOT, SINGLET, and VECTORS keywords are specified for calculating the energy and optimized geometry for the first singlet excited state of acroleine. The input file is mopac509mn/test/test6.dat. The output file is mopac509mn/testo/test6.out. Test run 7. The 1SCF, ANALYT, GRADIENTS, PM3, PRECISE, and VECTORS keywords are specified for calculating the energy of formaldehyde with the C-O bond stretched. The input file is mopac509mn/test/test7.dat. The output file is mopac509mn/testo/test7.out. Test run 8. The AM1, CHARGE, C.I., EF, GRADIENTS, PRECISE, and VECTORS keywords are specified for calculating the energy and optimized geometry of the allyl cation. The input file is mopac509mn/test/test8.dat. The output file is mopac509mn/testo/test8.out. Test run 9. The AM1, GNORM, MODE, PRNT, RECALC, SCFCRT, and TS keywords are specified for calculating the energy and optimized geometry for the MeNC to MeCN transition state. The input file is mopac509mn/test/test9.dat. The output file is mopac509mn/testo/test9.out. Test run 10. The EXTERNAL, GNORM, LET, PM3, PRECISE, TS, and UHF keywords are specified for calculating the NDDO-SRP (in particular, PM3-SRP) energy and optimized geometry of the transition state of the NH3 + OH reaction. The input and NDDO-SRP files are in mopac509mn/test/test10.dat and mopac509mn/test/TEST10.SRP, respectively. The output file is mopac509mn/testo/test10.out. Test run 11. The 1SCF, AM1, CHARGE, C.I., GRADIENTS, MECI, PRECISE, and VECTORS keywords are specified for calculating the energy of the allyl cation. The input file is mopac509mn/test/test11.dat. The output file is mopac509mn/testo/test11.out. Test run 12. The GRAD, MINDO3, and PRECISE keywords are specified for calculating the energy and optimized geometry of thiophene. The input file is mopac509mn/test/test12.dat. The output file is mopac509mn/testo/test12.out. Note: GRAD is a short form of GRADIENTS and means the same thing. Mopac 5.09mn Page 15 Test run 13. The AM1, CM2, and SYMMETRY keywords are specified for calculating the CM2 charges, the energy, and the optimized geometry of 1,2-ethanediol. The input file is mopac509mn/test/test13.dat. The output file is mopac509mn/testo/test13.out. Test run 14. The CM2, PM3, and SYMMETRY keywords are specified for calculating the CM2 charges, the energy, and the optimized geometry of 1,2-ethanediol. The input file is mopac509mn/test/test14.dat. The output file is mopac509mn/testo/test14.out. ====END OF MOPAC 5.09mn MANUAL====