From chemistry-request@ccl.net Fri Jul 26 09:55:24 1991 Date: Thu, 25 Jul 91 15:23:54 CDT From: tripos!victor@uunet.UU.NET (Vic Lewchenko) To: uunet!ccl.net!chemistry@uunet.UU.NET Subject: MOPAC6.0 geometry optimization Status: R We have started using MOPAC 6.0 for geometry optimization and have noticed that we get different results from those using 5.0. In nearly all cases MOPAC 6.0 terminates the minimization at a higher energy structure than 5.0 when using the same input file. Also with 6.0 we encounter messages such as "GRADIENT TEST NOT PASSED, BUT FURTHER WORK NOT JUSTIFIED", where using MOPAC 5.0 the minimization is successfull and results in a structure with a significantly lower heat of formation than obtained with MOPAC 6.0. Using the keyword PRECISE improves the results with 6.0, but greatly increases the CPU time to get nearly the same result as 5.0 without this keyword. The documentation for 6.0 states that it is no longer necessary to use the PRECISE keyword. We also tried setting GNORM = .01, this also improved results. We are running MOPAC on both Silicon Graphics and Vax hardware and we are using the official versions of MOPAC distributed through QCPE. The following is an example of an input file that produces the behavior described above: ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ XYZ T=99999 NOINTER tst5_A.dat " " 0006 -1.3582 0001 -1.4394 0001 0.4157 0001 0000 0000 0000 0006 -0.0714 0001 -1.0474 0001 -0.2776 0001 0000 0000 0000 0006 0.7145 0001 -1.9545 0001 -0.8768 0001 0000 0000 0000 0006 0.3077 0001 0.4189 0001 -0.2700 0001 0000 0000 0000 0006 1.6649 0001 0.6331 0001 0.4640 0001 0000 0000 0000 0008 -0.7386 0001 1.2027 0001 0.3371 0001 0000 0000 0000 0001 -0.5690 0001 2.1375 0001 0.3029 0001 0000 0000 0000 0001 -2.2190 0001 -1.1108 0001 -0.1871 0001 0000 0000 0000 0001 -1.4123 0001 -0.9634 0001 1.4068 0001 0000 0000 0000 0017 -1.4681 0001 -3.1875 0001 0.6492 0001 0000 0000 0000 0001 1.6353 0001 -1.6462 0001 -1.3715 0001 0000 0000 0000 0001 0.4664 0001 -3.0151 0001 -0.8884 0001 0000 0000 0000 0001 0.3847 0001 0.7273 0001 -1.3270 0001 0000 0000 0000 0001 2.0466 0001 -0.3051 0001 0.9011 0001 0000 0000 0000 0001 1.5586 0001 1.3621 0001 1.2831 0001 0000 0000 0000 0001 2.4190 0001 1.0151 0001 -0.2421 0001 0000 0000 0000 MOPAC 6.0 terminates the minimization with "GRADIENT TEST NOT PASSED, BUT FURTHER WORK NOT JUSTIFIED" and FINAL HEAT OF FORMATION = -50.35035 KCAL MOPAC 5.0 terminates successfully with FINAL HEAT OF FORMATION = -54.59063 Has anyone else noticed this difference in behavior between 5.0 and 6.0? Should some of the MOPAC 6.0 convergence criteria be changed? Vic Lewchenko Tripos Associates, Inc. St.Louis, MO USA From chemistry-request@ccl.net Fri Jul 26 12:01:52 1991 Date: Fri, 26 Jul 91 11:44 EDT From: Subject: re: PC Version of Gaussian 90 To: chemistry@ccl.net Status: R >Please, can somebody let me know of the availability (if any) of the >Gaussian 90 package for PC platform. Also, does anybody know whether >molecular modelling program MOBY ver 1.4 has been released yet. >Thanks! >I.Novak, Dept.of Chemistry,National University of Singapore, > Singapore 0511 > > Since Gaussian 90 is a commercial software package, we must be circumspect about discussing details over the Computational Chemistry e-mail List at OSC, which prohibits the posting of commercial advertisements. Therefore, I ask that all inquiries regarding Gaussian 90 be sent to Gaussian, Inc., who will be happy to respond to your questions directly. Additionally, we at Gaussian, Inc. have assembled an e-mail list for those interested in news regarding the Gaussian series of programs. This mailing list will help us keep users updated on new versions of the program, improvements and enhancements to the program, as well as upcoming Gaussian workshops and user meetings. We invite any interested parties to subscribe to the list. To do so, please contact me at e-mail address or phone number(s) given below, and we will be happy to include your name on the list. We also welcome any comments or suggestions you might have. Sincerely, David J. Moses Phone - 412-621-2050 FAX - 412-621-3563 e-mail - moses@gaussian.com From chemistry-request@ccl.net Fri Jul 26 12:04:20 1991 Date: Fri, 26 Jul 91 11:55:13 -0400 From: zheng@retina.chem.psu.edu (Ya-Jun Zheng) Subject: Reply to "MOPAC6.0 geometry optimization" Status: R I did a test with AMPAC2.1 program. Without precise or gnorm, the calculated heat of formation is -54.545935 kcal, which is very close to the the value reported by Lewchenko using MOPAC5.0. In the AMPAC calculation, the geometry still need further optimization since the gradient is relatively large (about 3). I just want to say two things. First, MOPAC 5.0 and MOPAC 6.0 may use different optimization routines (e.g., DFP and BFGS). So we have to make sure same optimization routine is used. Second, the two results may corespond to two isomers. Please check the optimized geometries. By the way, if you are interested in chemical reactions, AMPAC2.1 is the method of choice. I will be glad to provide you with more information on this. Ya-Jun Zheng Penn State University From jkl@ccl.net Fri Jul 26 12:14:32 1991 Date: Fri, 26 Jul 91 12:16:48 -0400 From: jkl@ccl.net To: MOSES%CMCHEM.BITNET@OHSTVMA.ACS.OHIO-STATE.EDU Subject: Re: re: PC Version of Gaussian 90 Status: R As to the comment of Dave Moses from Gaussian Inc: Since Gaussian 90 is a commercial software package, we must be circumspect about discussing details over the Computational Chemistry e-mail List at OSC, which prohibits the posting of commercial advertisements. The HELPFILE specifically and explicitly encourages commercial software developers and vendors to answer queries, questions and help in solving user problems. This is an excript of the rules: Commercial Posting - commercial software advertisements posted by the people who develop or sell the software. Postings responding to user questions, as well as bug reports, requests for comments, and tips on efficient software use are not considered commercial postings and are highly encouraged. Commercial software reviews or evaluations conducted by people not associated with the software vendor/developer are also encouraged. Jan Labanowski jkl@ccl.net Your Brieftrager From chemistry-request@ccl.net Fri Jul 26 12:24:15 1991 Date: Fri, 26 Jul 1991 12:13 EDT From: "DOUGLAS A. SMITH" Subject: Re: Reply to "MOPAC6.0 geometry optimization" To: zheng@retina.chem.psu.edu Status: R The default minimization routine in both MOPAC 5.0 and 6.0 is BFGS. DFP has been discounted as a tremendously useful method. MOPAC 6.0 also has eigenvector following (EF), which was not in 5.0. Your comment concerning the necessity to check that all optimization methods etc. are the same is salient. I don't believe that the isomer possibility is too viable. We have done enough tests to believe that while you do find a stationary point at higher energy, it is typically very unrealistic as far as real chemistry (at least for our systems). Doug Smith