From chemistry-request@ccl.net Sat May 4 09:57:46 1991 Date: Sat, 4 May 91 09:20:44 EDT From: Mike Peterson / Monstergauss To: chemistry@ccl.net Subject: "Certified" MOPAC on Apollo/UNIX Status: R I have been trying to get the "certified" MOPAC 6.0 version working on our Apollo system. It is working reasonably well, but the following problems remain: 1) It must still be compiled with 'save' - it fails miserably if the stack is used (at least on the Apollo). 2) On many test cases from the "original" MOPAC 6.0 (and MOPAC 5.0), the ionization potentials differ by several % (e.g. 8.97 vs 9.29 for AlF), and orbital energies differ by up to 1.5 units (this is somewhat shocking). These cases all seem to run for 1 more SCF (1 more geometry?) than with previous versions of the program, and the "new" answers agree in almost all cases with the "new" validation suite output. I find such a difference quite disturbing, and wonder what this means for any existing results based on IP or orbital energies (which seem to be the main results affected). 3) The "acid" test posted here a few months ago (which caused the Cray version of the "original" MOPAC 6.0 to abort, and also caused the "original" MOPAC 6.0 to fail with a segmentation fault on the Apollo) now works just fine Here is the input file to try on your version: AM1 DOUBLET TS alpha, beta-unsat acid carboxyl radical for decarboxylation C 0.0000000 0 0.000000 0 0.000000 0 0 0 0 0.3262 O 1.2235372 1 0.000000 0 0.000000 0 1 0 0 -0.2483 O 1.2372997 1 118.704508 1 0.000000 0 1 2 0 -0.3070 C 2.1148597 1 157.902093 1 -174.337130 1 1 2 3 -0.0873 C 1.3006401 1 121.439811 1 -174.074926 1 4 1 3 -0.2313 H 1.0750946 1 98.120631 1 6.535395 1 4 1 3 0.2176 H 1.1001354 1 122.939376 1 2.612134 1 5 4 6 0.1720 H 1.1027606 1 120.905097 1 -178.797410 1 5 4 6 0.1580 The correct answer is supposed to be 3.19 KCAL; I get 3.205 Kcal on the Apollo. 4) The "keys.dat" validation suite input fails on the last step, where 'COMPFG' attempts to rewind unit 2 (which is not open at that point). Searching the source code reveals that unit 2 is opened and closed in the main program which is fine, but 'rotate.f' opens and writes to unit 2, and 'compfg.f' and 'dcart.f' rewind unit 2, but I can't find anywhere that reads the stuff written out in 'rotate.f'. The test input that fails is; POLAR DEBUG ANALYT GNORM=0.02 T=30M LINMIN EXTERNAL=PARAS.DAT PM3 & graph search dep External param's turn PM3 into AM1 HEAT: -31.4977 XX 0.000000 0 0.000000 0 0.000000 0 0 0 0 H 1.000000 0 0.000000 0 0.000000 0 1 0 0 -0.0011 C 1.106118 1 180.000000 0 180.000000 0 2 1 0 0.2923 H 1.106135 1 112.894927 1 180.000000 0 3 2 1 -0.0011 O 1.216395 1 123.560741 1 180.000000 0 3 2 4 -0.2901 XX 2.166435 0 150.770198 0 0.000000 0 3 2 4 0 0.000000 0 0.000000 0 0.000000 0 0 0 0 and the file 'PARAS.DAT' contains: USS H -11.396427 ZS H 1.188078 BETAS H -6.173787 ALP H 2.882324 FN11 H 0.122796 FN12 H 0.005090 FN13 H -0.018336 FN21 H 5.000000 FN22 H 5.000000 FN23 H 2.000000 FN31 H 1.2 FN32 H 1.8 FN33 H 2.1 USS C -52.028658 UPP C -39.614239 ZS C 1.808665 ZP C 1.685116 BETAS C -15.715783 BETAP C -7.719283 ALP C 2.648274 FN11 C 0.011355 FN12 C 0.045924 FN13 C -0.020061 FN14 C -0.001260 FN21 C 5. FN22 C 5. FN23 C 5. FN24 C 5. FN31 C 1.6 FN32 C 1.85 FN33 C 2.05 FN34 C 2.65 USS O -97.83 UPP O -78.262380 ZS O 3.108032 ZP O 2.524039 BETAS O -29.272773 BETAP O -29.272773 ALP O 4.455371 FN11 O 0.280962 FN12 O 0.081430 FN21 O 5. FN22 O 7. FN31 O 0.847918 FN32 O 1.445071 GSS H 12.848 GSS C 12.230 GSP C 11.470 GPP C 11.080 GP2 C 9.840 HSP C 2.430 GSS O 15.420 GSP O 14.480 GPP O 14.520 GP2 O 12.980 HSP O 3.940 I have commented out the 'REWIND 2' statement in 'compfg.f' - this allows the job to run to completion, but the final Heat of Formation is now -31.49514 (supposed to be -31.49771), and all subsequent results are "off" but "close", with the deviation getting larger as the higher polarizability moments are computed. This test case also fails on the "original" MOPAC 6.0 - it never even reads the external parameter file. 5) This is not a "problem", but running the 'geometry.dat' and 'oldgeo.dat' can require some fiddling of the 'T=' and/or 'DUMP=' options, especially on a fast machine. 6) There are missing commas in FORMAT statements in 'flepo.f' (near line 392) and 'writmo.f' (near line 650). Mike. -- Mike Peterson, U/Toronto Department of Chemistry E-mail: mgauss@alchemy.chem.utoronto.ca Tel: (416) 978-7094 Fax: (416) 978-8775 From chemistry-request@ccl.net Sat May 4 11:18:20 1991 Date: Sat, 4 May 91 10:55:38 EDT From: Mike Peterson / Monstergauss To: chemistry@ccl.net Subject: mopachelp on UNIX Status: R I have got the 'mopachelp' program to work on our Apollo's (UNIX). The following changes were made to 'help.f': 85a86,90 > DEBUG = .FALSE. > DEBUGL = .FALSE. > DEBUGN = .FALSE. > DEBUGO = .FALSE. > DEBUGP = .FALSE. 272c277 < WRITE ( *, '(A)' ) TEXT --- > WRITE ( *, '(A,$)' ) TEXT 274c279 < WRITE ( *, '( 1X, A)') TEXT --- > WRITE ( *, '( 1X, A,$)') TEXT where the first patch gives values to some undefined variables, and the second/third make the program prompt look better. In addition, the file 'MOPAC.HLP' must have all the keyword lines, identified by a digit in the first column then a blank, in upper case since 'mopachelp' converts all user input to upper case internally. Mike.