ChemStatus

[ Up ] [ Current Status ] [ Previous Status ]

Computational Chemistry Status Report - 2nd Quarter

We have not received our final shipment of computers so there is nothing to add to the hardware story.

On the software front there is one significant development to report. The main program used in our Computational Chemistry course, Chem 765, is the commercial program Gaussian 94W (the Windows version of the UNIX ab initio program Gaussian 94). A significant limitation of this program is that there is no simple way to display the output graphically. We havewritten (and debugged) a Windows NT program that accepts a G94W output file from a frequency calculation and uses the structure and force constant data within it to calculate the infrared vibrational frequencies, vibrational modes, and absorption intensities for the molecule. These derived quantities can then be displayed as an animated 3D-projected vibration for any chosen mode or displayed as an infrared spectrum in any one of six conventional chart presentations. The charts can be printed on a laser printer.

Computational Chemistry Status Report - 1st Quarter

The goals of this project centered on expanded exposure of our chemistry students to modern computational chemistry. Although we are still at an early stage of the project (we have received less than 30% of the eventual computer power), we have made considerable progress. Most importantly, we have a much clearer understanding of our needs and how they can be met with the Intel contribution. To date we have received: one dual 200Mhz Pentium Pro and two dual 300MHz Pentium II's.
(1) The main challenge was to find a useful way to mix numerically intensive computations such as those done using the Gaussian 94 ab initio program with graphically intensive molecular mechanics programs such as PC Model. After some experimentation we hit upon setting up a dual-boot system using the commercial package System Commander. The first boot-up option is Windows NT, which handles the graphical needs. The second boot-up option is Linux, which deals with the computationally intensive needs. The version of Linux we installed, Red Hat 5.0, has the capability of directly writing files that can be read by NT. The delivered Pentium II workstations had inadequate hard-disk space for some of the anticipated large jobs; for these two machines we installed 8GB disks. The new computers will come with 8GB disks.

(2) The f2C conversion, compilation and installation of Gaussian 94 went smoothly with no major problems (a Gaussian feature for running linked successive jobs did not compile correctly, but we were able to create a batch file that simulated this feature). We compared the speed of this Linux installation of Gaussian on both the Pentium Pro and Pentium II units against the speed on the local IBM SP2 supercomputer. We did many speed tests, but a particularly interesting one is the widely used test job provided by Gaussian, Inc (their test job number 178, 2,4,6-trinitro-1,3,5-triamino-benzene). The times obtained were:
Cornell SP2 supercomputer 361 sec
200 Mhz Pentium Pro/Linux 1708 sec
300 Mhz Pentium II/Linux 1326 sec
The (typical) relative speed ratios of 4-5 provide strong justification for the Intel workstation approach. Even better, we were able to run a Gaussian ab initio job on the Pentium II that could not be run on the SP2 because it took longer than the maximum time allowed (currently 12 hr.) We have started using these machines in our graduate computer chemistry course, but it is too early to have any meaningful feedback. (3) Thus far we have not taken advantage of the second processor. We have purchased a Linux compatible FORTRAN compiler that reputedly recognizes a second processor. At the end of the semester, when time becomes available, we will attempt to reinstall the Gaussian 94 code.

Last modified on: 10/08/99