The business of quantum chemistry

The six scientists in this elegant old brick building in central Connecticut work on a project that is vital to modern chemistry, but would never get funded by a grant. They don’t have to worry about winning grants, though. They get all the funding they need from an international crowd of benefactors—their fellow chemists.

Gaussian’s offices are at 340 Quinnipiac Street in Wallingford, Connecticut. CREDIT: Kim Krieger

The first thing you notice when you step into the lobby is the sweeping staircase that embraces the front desk like a split river of dark wood. The art deco murals on the ceilings, the hushed air of a library, and the occasional glimpse of a figure hunched over a computer add to the university atmosphere. But this is no university. It’s the headquarters of Gaussian, a private company that develops quantum mechanical modeling and analysis software. These tools are critical for modern chemical research. But researchers don’t win grants to develop tools. Researchers win grants to solve chemical problems: to develop safer fluorination of pharmaceutical compounds, say, or find cheaper alternatives to precious metals catalysts. Any methods or modeling work gets done on the side.

The founders of Gaussian are keenly aware of this. The original software, Gaussian-70, was developed by John Pople’s group at Carnegie Mellon in the 1970s. Initially, it was given away for free. At the time it was groundbreaking, allowing chemists to model molecular interactions at a level of detail they could never have managed before. It also supported Pople’s mathematics research that applied the rules of quantum mechanics to chemistry, an accomplishment for which he eventually won the Nobel Prize in Chemistry.

But people started to make unauthorized changes to the software. Sometimes they broke it and then complained. And responding to the complaints got tiresome. So in the early 1980s, Gaussian’s authors copyrighted it. But as the software grew more capable and useful as a tool, it became less and less viable as a research hobby.

“The problem with [designing] free software is that people do the fun stuff. Not the hard, boring stuff, like quality testing and ease of use. That kind of thing tends to be neglected,” says Michael Frisch, president of Gaussian. The group decided to form a private company in 1987. Instead of tinkering with the software on the side whenever a new line of research required it, the software would be the research. And instead of relying on grant money for methods development (there isn’t any), they would be funded through revenue from the software.

The revenue model proved durable. Gaussian moved to its present building, the former headquarters of Wallace Silversmiths in Wallingford, Connecticut, in 2002. In 2013 the company’s flagship software, Gaussian09, can model the energies, structures, vibrational frequencies, and properties of molecules in a variety of chemical environments.

Someone who doesn’t work in computational chemistry might question how useful it is to be able to calculate molecular properties from the first principles of quantum mechanics. But recall your high school chemistry: Electrons are shifty. Maybe they’re here, maybe they’re there, maybe they’ve skipped town and gone to Disneyland. If you work in the macroscopic world, such details seem fussy, but materials researchers care deeply. They model molecules as colorful, knobby balloons outlining electrons’ likely locations. An unexpected kink in the balloon alters the properties of the molecule, from its shape to the energy at which it will decompose. So it is vitally important to get the models right.

But for a long time, modeling work at the molecular level was just too hard for systems of more than three particles. The laws of quantum mechanics that make it theoretically possible to calculate the balloons—really electron energy density functions—were formulated in the 1920s. But as Paul Dirac said at the time, “the difficulty lies only in the fact that the application of these laws leads to equations that are too complex to be solved.”

The equations haven’t gotten simpler, but our computers have gotten more powerful. And companies like Gaussian have emerged to develop the software.

Pesticides, combustion, and unknown compounds

Gaussian09, the current iteration, is powerful. With it, an industrial chemist developing a pesticide can model a molecule with a photo-sensitive bond that needs to last at least one hour under bright sunlight in order to kill the target pest, but should degrade within 24 hours to avoid broader environmental poisoning. Chemists testing alternative fuels for suitability in combustion engines might want to see which chemical reactions are likely during the combustion process, and Gaussian09 can help them. A researcher seeking to identify an unknown compound can use Gaussian09 to more quickly interpret the results of nuclear magnetic resonance spectroscopy.

The highly specialized nature of the product means you can’t just hire a clever programmer. You need a programmer who also has a doctorate in chemistry or physics, someone who actually understands quantum mechanics. Gaussian finds it hard to hire people from the United States. Researchers here tend not to be interested in underfunded chemical methods work, Frisch says. And despite the numerous open questions in the field—Gaussian cannot easily do relativistic modeling of heavy molecules, which would be useful to condensed matter physicists, for example—it’s not as if the company has gone on a hiring spree. It has six full-time scientific staff in addition to its sales and marketing team.

But Gaussian rarely loses a staff member. If someone leaves, it’s typically to join a university faculty. Some of the scientific staff hold university positions concurrently—Frish, for example, holds a faculty position at Wesleyan University in Middletown, half an hour north up the parkway.

If Gaussian found venture capitalist investors it could hire more people, but eventually it would have to go public in order to repay loans, and that does not appeal. Public companies have to demonstrate continuing growth in order to please investors. “If I compare my gripes with those I hear from colleagues in the public companies . . . I’d rather have my problems,” Frisch says. It’s a question of balance, he says, “between attacking problems that matter now and building the tools to attack problems that matter in the future.” The continuing growth that public software companies seek isn’t even a goal for Gaussian. Steady-state might be a better description. Because really, the revenue is simply a way to fund the research, a way to keep a little bastion of academia going and improving, providing reliable tools independent of the fickle fashions of academic funding.

Kim Krieger is an independent science writer. She has reported on science policy from Capitol Hill, energy from the floor of the New York Mercantile Exchange, and physics innovation everywhere.