I was reading a paper in preparation for summer research. The paper evaluates two fragmentation methods (FMO and SMF) against ab initio results, and I’m learning a lot about fragmentation methods. However, one particular section stood out (and not without reason):
…Calculations of these total energies as well as all FMO calculations, were carried out using the General Atomic and Molecular Electronic Structure System (GAMESS) program package. All of the SMFA energies were evaluated by MAC and MAA using the GAUSSIAN03 program package. SRP and MSG did not use, nor would they consider using, the aforementioned program package.
Emphasis is not mine.
If you don’t have much experience with computational chemistry, the need for such a clarification arises from the fact that Pruitt (SRP) and Gordon (MSG) are banned by Gaussian. In essence, that means they can’t use any products developed by Gaussian, Inc. An anonymous (group?) created the website to “shed light on some practices of Gaussian, Inc. that can undermine basic scientific ideals”. And they’re not alone – the list of people banned by Gaussian are only the ones who have consented to appear on the website. I’m sure there are plenty of other people out there who either have not been contacted about appearing on the website or who have been recently banned since the website’s last update.
Since I’m but a measly soon-to-be-grad student, I am unable to afford a copy of Gaussian (see their prices here), and Google was not very helpful in finding a recent copy of the license agreement, I can’t directly and accurately quote the license agreement (is that even allowed?). However, my understanding is that you may be banned by publishing a comparison of timings between Gaussian and competing software, or if you contribute code for a competitor. To quote Mark Gordon, this is “the wrong way to advance the scientific enterprise”. Placing such restrictions on who can use Gaussian essentially prevents researchers from reproducing reliable results and it prevents a fair communication between chemists about software performance. That is not to mention that it hinders the development of new theories or insights into their applications by requiring that people who are banned start from zero (although, I don’t really know how bothersome that is).
Gaussian DID issue a response (silly.html), written by Michael Frisch, that points out some not-very-accurate points made by the Banned by Gaussian website. For example, the listing of “Who is banned? By University/Institution” on the Banned by Gaussian website gives the idea that an entire institution can be banned (meaning, nobody in that university can use Gaussian). This claim, Gaussian says, is simply not true.
In cases where there is directly competitive work going on at a University, we always offer that University a license which provides for access by anyone except members of the competitor’s research group. The statement that we refuse to license entire universities is an outright lie.
In fact, most of the sites listed on the website have numerous users of our software, including the current version, Gaussian 03. 10 of the 11 sites listed on the website have Gaussian 98, and 10 are already using Gaussian 03. The only remaining site, Iowa State, was offered a license for Gaussian 03 which would provide access for everyone outside of the one competing research group, but has elected not to return it*.
I don’t know how much I trust the language “Iowa State, was offered a license for Gaussian 03″, which sounds like Gaussian was extending the olive branch by offering free software. I can’t be too sure, though. I do believe, however, that the authors of the Banned by Gaussian website meant that the institutions listed have someone who is banned, but not that the institution as a whole is banned. I kinda figured that one out when I visited a grad school and met with one professor who was banned by Gaussian (because he develops code for QChem) and another professor who develops code for Gaussian. Maybe that’s a point against the Banned by Gaussian folks for not making that clear.
Other than that, other arguments in the response are somewhat irrelevant or weak.
Almost every calculation which can be run by Gaussian can also be done in some other package, albeit not as conveniently or as quickly. Thus, in general it is possible to directly reproduce the results without access to our software.
This is true to an extent. The keywords, however, are “conveniently” and “quickly.” A competent chemist could reproduce results of calculations done in Gaussian, but the ban by Gaussian is just a burden on the scientist. Furthermore, in the event that a researcher takes it into their hands to write code to verify the results, there are possibilities of rounding errors and logic errors (if the response is different, how will the researcher know his own code isn’t bugged or that the original author was mistaken?). However, I am left wondering about why in the paper I’m reading, SMFA calculations are being done in Gaussian instead of consistently keeping with GAMESS (a quick search for SMFA in the GAMESS manual returned 0 results).
Some Groups Choose Not to Distribute Their Code. If access to the specific software package were made a requirement for publication, then no paper using software written in a group and not distributed would be accepted. Do the authors of this website really advocate rejecting all papers from theoretical groups which choose not to distribute their programs?
This point is rather irrelevant. The main idea is that Gaussian (a widely used and commercial product) bans research groups from using their programs. This is different from researchers asking each other for native software for which they wouldn’t have to pay.
There is one last point that Gaussian’s response makes which I can entirely agree with:
Using Gaussian is Not Harmful
Finally, the instructions on “How to be banned by Gaussian” are downright silly. More importantly, even the misrepresentation on this website cannot manage to imply that applying our software to chemical problems and using or publishing the results would get anyone banned. In fact, the activities they describe are of interest primarily to the rival software package developers listed on the website and not to the thousands of chemists using our program as a tool in their research.
The Gaussian program is not bad. I’ve used it during my undergraduate years, and the experience is similar to using a different web browser. It has its strengths, it has its weaknesses, but it has the same final mission, which is to perform calculations to solve chemical problems. There are no subliminal messages, and Gaussian, Inc has not set up a war against competing software packages. Sure, being banned is an inconvenience, but there are plenty of people out there who choose to use Gaussian and there are those who aren’t banned but still prefer other packages. It all comes down to which package offer the most appropriate tools for research.
Lastly, I would like to point out that being banned by Gaussian isn’t the end of the line. The Banned by Gaussian website mentions that “having your research group or your university banned by Gaussian is recognition of your outstanding scientific achievements!”, and a commenter on reddit mentioned that banned groups show their pride about it. It worries me to see that, sometimes, people are afraid to post questions or answer questions on the internet for fear that their response will result in them getting banned (example 1 and example 2). This is truly when I see the system of collaboration in science fall apart, and I think it would be great that researchers could let go of that fear. I can’t tell how likely it would be for Gaussian to change the terms outlined in their software license, but (thankfully) it hasn’t kept people from practicing science.
So my message to everyone: Keep Calm and Calculate On.
*I do know that, in Iowa State, Mark Gordon is very much opposed to Gaussian’s restrictions. That explains the underline in the paper I mentioned at the opening of this blog (“SRP and MSG did not use, nor would they consider using, the aforementioned program package.” emphasis mine), and the possible “rejection” of the license Gaussian “extended” to them. In fact, rumor has it that MSG rejects peer-reviewing any paper that solely uses Gaussian. But that’s only a rumor, afaik. I’ve heard from a member in his group that this would be a fun topic to have with long chat him.