What CERN is for, when it isn't smashing protons
The web, the climate, and the habit of sharing data: three things CERN gave away, each with a Nature paper behind it.
Most people know CERN for one thing: the 27-kilometre ring under the French–Swiss border where protons are made to collide at nearly the speed of light. That is real, and it found the Higgs boson. But it is a narrow picture of what the place does. CERN is also where the web was invented, where one of the most careful experiments on clouds and climate is running, and where a way of sharing data was worked out that the rest of science has been borrowing ever since. Three stories, each with a Nature paper behind it.
The web was a side project
In March 1989 a CERN software engineer called Tim Berners-Lee wrote a memo titled Information Management: A Proposal. His boss scribbled "vague but exciting" on the cover. The problem he wanted to solve was a local one: thousands of physicists, on incompatible computers, losing track of each other's documents. His answer was a system of linked pages anyone could read with the same program. The first website went live on a NeXT machine at CERN in 1991.
The decision that mattered came two years later. On 30 April 1993, CERN put the web's software into the public domain, free for anyone to use, copy and change, with no licence fee and no permission needed. There were a few hundred websites at the time. Had CERN chosen to charge for it, or to keep it inside physics, the web as we know it would not exist. Every open-data portal, every preprint server, every public dataset Earth 1 links to runs on that one decision.
Clouds, cosmic rays, and the biggest open question in climate
The least certain number in climate science is not how much carbon dioxide warms the planet. It is how much clouds cool it, and how that is changing. Clouds form on tiny particles, aerosols, and the question of where new aerosols come from in clean air was, until recently, mostly guesswork.
CERN's CLOUD experiment attacks this with a stainless-steel chamber, three metres across, filled with air cleaner than anywhere on Earth, into which the laboratory's particle beam can fire a controlled stream of "cosmic rays". For fifteen years it has been measuring, one ingredient at a time, what makes particles form.
Its first Nature paper, in 2011, settled a long argument. Sulphuric acid and ammonia together make new particles, and ionising radiation does speed the process up, but not by nearly enough to explain what is seen in the lower atmosphere. The cosmic-ray-drives-climate idea, popular at the time, lost most of its footing. In 2016 the team showed that particles can form from plant vapours alone, with no sulphuric acid at all, which means the pre-industrial sky may have been cloudier than models assumed, and the cooling we have had from pollution smaller.
The most recent result, in Nature in December 2024, concerns isoprene, a hydrocarbon that forests breathe out in enormous quantities. Textbooks said isoprene could not make new particles. CLOUD showed that at the temperatures of the upper troposphere, below minus 30 °C, over tropical rainforests, it does, and quickly. That is a new source of cloud seeds, high above the Amazon and the Congo, that no climate model had in it.
None of this tells you the planet is warming faster or slower. It tells you the models are being corrected, measurement by measurement, in a steel can in Geneva.
The data problem, and what the rest of science borrowed
The Large Hadron Collider produces petabytes of data every year. No single computer centre could hold it, so CERN built a grid: roughly 170 computing centres in some 40 countries, sharing the storage and the processing as if they were one machine. It has run since before the LHC switched on.
Two things from that effort spread far beyond physics. One is the habit of publishing the data itself: CERN's Open Data portal releases collision data, with the software to read it, to anyone who wants it. The other is machine learning as a routine scientific tool. A 2018 review in Nature set out how particle physicists had come to rely on it to pick a handful of interesting collisions out of billions, long before "AI" was a word in every newspaper.
A word of care here. It is sometimes said that CERN's machine-learning models are lifted straight into climate and Earth-system models. We could not find that claim in Nature or anywhere peer-reviewed, and it is probably not true in that form; the models are built for very different data. What did travel is the practice: open data, open code, shared computing, and the expectation that a result comes with the means to check it. That practice is now ordinary in climate science, in astronomy and in genomics. It was not ordinary before.
Why Earth 1 cares
Earth 1 exists to make real science legible to anyone who wants to understand it. CERN is a useful case because it shows what that costs and what it pays. The web was given away. The CLOUD results, inconvenient for some and reassuring for no one, were published as found. The data is on a public server. That is the standard we try to hold ourselves to, and the one we think is worth asking of everyone else.
Sources
- Berners-Lee, T. Information Management: A Proposal, CERN, March 1989; the public-domain release of the web software, 30 April 1993: CERN, The birth of the Web
- Kirkby, J. et al. Role of sulphuric acid, ammonia and galactic cosmic rays in atmospheric aerosol nucleation. Nature 476, 429–433 (2011). doi:10.1038/nature10343
- Kirkby, J. et al. Ion-induced nucleation of pure biogenic particles. Nature 533, 521–526 (2016). doi:10.1038/nature17953
- Shen, J. et al. New particle formation from isoprene under upper-tropospheric conditions. Nature 636, 115–123 (2024). doi:10.1038/s41586-024-08196-0
- Radovic, A. et al. Machine learning at the energy and intensity frontiers of particle physics. Nature 560, 41–48 (2018). doi:10.1038/s41586-018-0361-2
- The Worldwide LHC Computing Grid: wlcg.web.cern.ch. CERN Open Data: opendata.cern.ch