GAMESS crash course?

I'm still learning about MO theory – and I thought that I would do some calculations with GAMESS to become more familiar with the concept. Even though I look forward to delving into the realm of quantum math in the not-so-distant future – it's very difficult to get a grip on all of the terminology. I'm using the GAMESS Avogadro extension to write some input files, but I don't know which options to use. Which options should I use if I want to study the orbitals of, say, shikimic acid? RHF, MP2, AM1 / MINI 6-31G(d)? Computer specs: 3.6GHz i5 with 16GB of RAM

• I strongly advice you to not enter this dangerous path. First things first. Learn theory, then apply it. Otherwise 99% of your calculations will be GIGO (garbage in, garbage out). Programs can crunch almost any garbage you input, but the output numbers won't have the meaning you expect. – Wildcat Sep 20 '14 at 18:52
• gamess-US source code comes with good enough example jobs (afaik, in test directory), though making it work is a bit of challenge. For learning purposes, i.e. molecules up to 16 atoms with 1-2 d-element atoms it is enough to use mp2/6-31G(d). Thouhg in case your job starts to crunch your hard drive, you should consider reducing amount of atoms. – permeakra Sep 20 '14 at 19:29
• @Wildcat I do agree that makes no sense of doing e.g. MP2 without knowing much about theory, but wouldn't discourage OP from experimenting. There is always a subset of calculations that makes sense to try, even with limited familiarity with the theory. E.g. doing Huckel or extended-Huckel calculations can make sense and has strong pedagogic value. I think we do our job best here, if we help OP to find those problems/calculations. Also, testing methods agains each other and seeing how they break down even "simple" cases like bond breaking in diatomic molecules, teaches a lot. – Greg Sep 21 '14 at 1:36
• @Greg, that was just my own opinion based on personal experience. I mean, I was there many years ago, and I would not recommend anyone to take the same road. – Wildcat Sep 21 '14 at 7:59

While I do think you should learn some theory, it's definitely possible to learn some practical computational chemistry through experimentation. And yes, there are a lot of abbreviations to follow.

One thing you should track while you explore is how long different calculations take. In general, more accurate quantum chemical methods are also more computationally costly. This means that they take longer, but also as systems get bigger, the cost rises faster. You might hear some people talk about "linear scaling" (i.e., double the size of the system, twice as much time) or $N^2$, $N^3$, etc.

A good resource, IMHO, is Molecular Modeling Basics by Jensen. It's available through CRC Press but Jan also blogs extensively. He uses GAMESS and Avogadro but also Jmol and other programs.

While the Avogadro project is working on some more detailed tutorials, I'd suggest starting some calculations with HF/6-31G* or B3LYP/6-31G* and see how things go for different molecules. I strongly suggest running some calculations and then consider how the results compare to experiment. Benchmarking is important. Computational models all have systematic and other errors, and for real work it's critical to run a set of molecules and carefully choose methods that work well.

Here are a few key terms to consider:

• Method: There are several types of methods.

• Semiempirical methods: These include AM1, PM3, RM1, (and PM6, PM7, and others that I don't think are in GAMESS-US). Such methods are fast, since they approximate some integrals and use empirical parameters. They also have an implied basis set which is usually very small. (Avogadro doesn't support viewing semiempirical calculations with GAMESS yet.)
• Wavefunction methods: These include HF and relatives. These are similar to typical quantum chemistry methods taught in school.

• MP2, MP3, MP4: These are perturbation methods that improve on Hartree Fock and introduce some level of electron correlation.
• CCSD, CCSD(T), etc.: These are "coupled cluster" methods that are highly accurate, but often slow, and do not handle large molecules.
• Density Functional: This includes another large set of methods, including BLYP, B3LYP, and newer methods like M06-X. Rather than approach things through the Schrödinger equation,

• Basis: Most methods (except AM1, RM1, PM3, etc.) also require a "basis set," a set of functions to describe the atomic orbitals. Wikipedia article There are a few families, including:

• Pople Type: 3-21G(d), 6-31G(d), 6-311+G(d), etc. These are older, more traditional basis sets. Generally 6-31G(d) or 6-31G* is considered minimally acceptable today. The "*" or (p,d,f) included at the end indicate that higher polarization is included on some or all atoms.
• Correlation Consistent: cc-pVDZ, aug-cc-pVTZ, etc. These are newer basis sets, designed to function better with methods that treat electron correlation. The "aug-" includes various diffuse functions, which are useful particularly for anions and electron density that is particularly, well, diffuse. (In Pople basis sets, these are indicated with "+".
• Thank you for your reply! Now I know where to start. One question though: to what extent is the hard drive used in QM calculations? I understand that temporary files are stored during the calculations, but do I have to worry about fragmentation or even exceeding my disk space for relatively small molecules like shikimic acid? – Ravaru Sep 21 '14 at 8:22
• @Ravaru, it depends on your HDD size. But if you have something around 1TB (more or less standard this days), you don't have to worry about exceeding the disk space. Fragmentation could be a potential problem, although, I don't think that it will considerably affect performance. – Wildcat Sep 21 '14 at 8:31
• Sorry, 1TB free? I think I have around 600GB free. Will that suffice? The data that takes up space is only temporary, right? – Ravaru Sep 21 '14 at 10:44
• @GeoffHutchison, ah, you mean this tradition I hate and never use. :D Besides, I doubt OP is aware of this computational chemistry notion of "heavy elements" which differs essentially from one often used in general chemistry. – Wildcat Sep 21 '14 at 12:03
• @GeoffHutchison, I prefer to use "light elements" for first-, second-, and third-period elements, i.e. for which non-relativistic approximation is usually a good one. – Wildcat Sep 21 '14 at 12:06