When two physicists from the University of California at Santa Barbara posted a paper called “Exact black string solutions in three dimensions” on a Web site at Los Alamos on 14 August 1991, a revolution in physics began. Since then, the pre-print server at the Los Alamos National Laboratory (LANL) has allowed researchers in physics and related disciplines to post and retrieve papers far faster than any other publishing method.

“It has had an enormous impact on the accessibility of the physics literature in terms of both reading other people’s papers and having your own papers read,” says Heath O’Connell, manager of the SPIRES database at the Stanford Linear Accelerator Laboratory. “It also paved the way for on-line journals.”

The physics community will mark the tenth anniversary of the “e-print” server this month during a meeting on strings in Aspen, Colorado, where Paul Ginsparg, a theoretical physicist at LANL, originally had the idea for xxx.lanl.gov. After the meeting, however, Ginsparg will move the server, now named arXiv.org (pronounced ar-chi-v), in a new direction. He will take up the post of professor of physics and computer science at Cornell University, where he earned his PhD 20 years ago. The server will be rehoused in the university itself.

“My hope is to see the research in communication resources represented by the arXiv continue to evolve and continue to serve as a model for other disciplines,” Ginsparg told Physics World. “In physics research, I’m hoping to pick up where I left off.” His non-arXiv work includes the chiral symmetry of the lattice and string theory.

Cornell regards the move as a double coup. “Ginsparg is an important hire because he is an excellent theoretical physicist and because his arXiv is transforming the way physicists interact with each other and with their scientific literature,” says Peter Lepage, head of physics.

Instant appeal

The arXiv is a fully automated electronic archive and distribution server for research papers. Authors can submit their papers to the site via the Web interface, by e-mail or by ftp. They can also update their submissions as they receive comments. Users access the papers through the Web or by e-mail.

The system has grown significantly since it began. Originally, recalls Ginsparg, “it ran on a NeXTstation computer in a colleague’s office with a 25 MHz processor and a 400 Mb hard drive”. It had a constituency of about 100 scientists, most of them particle physicists who were already used to exchanging paper pre-prints by post. Ginsparg anticipated 100 submissions per year.

Today, after several upgrades, the system serves the bulk of the physics community worldwide. Some two million visitors check the site each week and Ginsparg expects to receive 35 000 submissions from about 100 different countries this year. The scope of the site has also broadened, with the number of papers on condensed-matter physics rapidly overtaking that on particle physics.

“It has been an essential tool for superconductivity,” says David Cardwell from Cambridge University in the UK. “In my opinion it has enabled more coherent and efficient progress to be made by reducing repetition of experiments and by establishing correlation at an earlier stage.”

This year’s activity in superconductors based on magnesium-diboride compounds illustrates that value. “When we wrote our first papers on MgB₂ we would submit them to Physical Review Letters, post them on LANL, and then e-mail copies of the file to friends, colleagues, and/or researchers in the field,” recalls Paul Canfield from Iowa State University. “The LANL server rapidly became the focus of MgB₂ research. By the time the first paper was published there had been scores of papers posted on the server. The fact that the research community could exchange information so rapidly allowed for explosive growth of knowledge.”

Access all areas

The global nature of the interest illustrates another advantage of the server. “Primarily it has levelled the playing field, both geographically and hierarchically from graduates students up,” says Ginsparg. “There’s no privileged loop of researchers with advanced access to new material. It gives researchers everywhere the ability to decide when they want to go fully public with new results and makes those results simultaneously accessible to all interested.”

One might imagine that the arXiv would have succeeded at the expense of physics journals. In fact, says Ginsparg, “in physics this set of resources was established long before any of the publishers had discovered the Internet. It would have been a mistake for any publisher to alienate the established physicist clientele.” And while the American Physical Society (APS) initially worried about arXiv’s impact on its publishing, Ginsparg continues, “it’s clear now that we’re all trying to figure out how best to port physics communications infrastructure to the new technology”.

Conventional publishers have the obvious advantage of peer review. The arXiv, by contrast, simply screens e-mail domains. “We require a recognized affiliation,” says Ginsparg. “If someone from Caltech tried to submit a paper on perpetual motion, it would appear.” Iowa’s Canfield warns about this “slightly dark side” to the server’s use. “Several groups posted magnesium-diboride papers that were highly suspect, if not clearly wrong, that were then removed within several days for a variety of claimed errors,” he says. “This simply means that people who read the server have to be extra cautious about what to believe.”

Moving east

The arXiv will experience that future in its new location. Rumours in the physics community suggest that Ginsparg left LANL because it had not given him and the arXiv enough support. However, he sees the move as necessary to continue the server’s evolution. “The LANL research environment was essential to starting it and it wouldn’t have been possible had I been a university faculty member with too many other obligations,” he explains. “[Los Alamos] was a good place to incubate the project. But now it has achieved a level of maturity that makes it possible to institutionalize in a new and more appropriate setting.”

Although LANL will continue to provide some costs and services to arXiv, the site will lose a contribution from the US Department of Energy to its $300 000 annual cost. However, the server will continue to receive support from the National Science Foundation. Cornell University is providing some money and will also seek private funds.

As to future development, Ginsparg anticipates plenty of cross-fertilization with other projects at Cornell. “The Cornell group is strongly committed to electronic self-archiving,” says Stanford’s O’Connell. “Giving Ginsparg this official position should lay to rest any fears about the archive’s future.”