Comsol -leaderboard other pages

Topics

The many lives of RHIC

The proton–proton collider ISABELLE, at Brookhaven National Laboratory, was cancelled in July 1983 with its 3.8 km tunnel already constructed (CERN Courier July/August 2026 p22). Within the year, the nuclear-physics community had proposed reusing it for a quantum-chromodynamics (QCD) machine capable of producing the quark–gluon plasma (QGP) – a state in which quarks and gluons are no longer confined within protons and neutrons. Thus began, in 1991, the construction of the Relativistic Heavy Ion Collider (RHIC), which became the first collider designed specifically for heavy ions and the only one to collide polarised protons (see “Bird’s-eye” figure).

Bird’s-eye

When RHIC’s first gold ions collided in 2000, the BRAHMS, PHENIX, PHOBOS and STAR experiments observed strong collective motion and a suppression of high-transverse-momentum particles, evidence for matter at record energy density. Because the suppression could also have originated in the structure of the gold nuclei themselves, the four experiments turned to deuteron–gold (d–Au) collisions, in which no hot matter was expected to be created but any effects from the gold nuclei would still be apparent. By examining different aspects of the events, they observed no suppression and proved that the effect seen in gold–gold (Au–Au) collisions was due to jet quenching – the loss of energy by fast quarks and gluons in the dense QGP medium. The complementarity of their results was made explicit in August 2003, when the cover of Physical Review Letters displayed a plot from each (see “Back-to-back” figure). In 2005, the experiments published white papers on their collected measurements, which revealed that the matter created at RHIC behaved not as the anticipated weakly interacting gas, but as a strongly interacting, nearly perfect liquid, flowing with almost no internal friction. These remain among the most highly cited papers in nuclear physics.

Snakes in the rings

The RHIC spin programme, which collided beams of spin-aligned protons, was started in 2001, in response to the so-called “proton-spin crisis” – the 1988 finding by the European Muon Collaboration that quarks could only account for a fraction of a proton’s spin. Polarised collisions were not part of RHIC’s original design, and much of the required technology had yet to be demonstrated. A high-intensity optically pumped polarised ion source (OPPIS) produced beams with up to 83% polarisation. At the same time, partial Siberian snake magnets kept the proton spins aligned as the beam gained energy in the Alternating Gradient Synchrotron (AGS) without elaborate resonance-by-resonance corrections. At RHIC, two full Siberian snakes in each ring allowed protons to reach 255 GeV with little polarisation loss. Spin rotators provided longitudinal polarisation at the experiments, while proton–carbon and polarised hydrogen-jet polarimeters monitored the beam polarisation (CERN Courier October 2005 p15).

Back-to-back

After more than a decade of development of luminosity, polarisation and measurement techniques, the STAR and PHENIX collaborations decisively established that gluon polarisation contributes to the spin of the proton and, from parity-violating W decays, found an unexpected difference between the polarisations of the up and down antiquarks within a polarised proton. The observation of large transverse single-spin asymmetries at high energy and the exploration of their connection to the transverse-momentum-dependent structure of the proton will be a lasting legacy of RHIC. The latest datasets, taken with transverse polarisation, allow for the exploration of the three-dimensional spin-dependent structure of the proton, developing techniques that will be used at the future Electron–Ion Collider (EIC).

Stochastic cooling

Heavy ions in a stored beam scatter off each other, causing the beam size to grow and the collision rates to drop, so that the machine requires frequent refilling. To counteract this effect, stochastic cooling systems, which sense variations within the beam at one point of the ring and correct them with kicks at another, were installed in RHIC between 2006 and 2014, to act in both the transverse and the longitudinal directions. While stochastic cooling had previously been used at much lower energies for unbunched beams, RHIC was the first machine to apply it to bunched beams and the first hadron collider to cool both of its beams.

The system demonstrated its full capabilities during RHIC’s uranium–uranium (U–U) run in 2012 (CERN Courier October 2012 p17) and subsequently enabled a large increase in Au–Au luminosity, eventually reaching 44 times the design value for the average store luminosity (see “Cooler and brighter” figure). Stochastic cooling also allowed more ions to be stored. As a result, RHIC could operate with high luminosity near the “burn-off” limit, where almost all ions are lost through collisions with those in the opposing beam.

Cooler and brighter

The high luminosity, together with the silicon-detector upgrades operating in PHENIX from 2011 and in STAR from 2014, enabled next-generation studies of the QGP created at RHIC, including precise observations that charm quarks flow along with the nearly perfect liquid, measurements of electromagnetic probes to constrain its temperature and space–time evolution, and measurements of quarkonia as a major probe of deconfinement. The control over the beams that cooling brought also underpinned many of the later runs with novel beam combinations.

New ions, new questions

RHIC was designed for flexibility. Its two superconducting rings operated almost independently, coupled only by two dipole magnets in each of the six interaction regions, which brought the beams into collision and separated them again. The ion sources and injector chain could deliver beams ranging from polarised protons to uranium ions and switch rapidly between species for asymmetric collisions. Over its lifetime, the machine collided 12 different pairs of species, five of them asymmetric and three involving at least one polarised beam, at 18 centre-of-mass energies, for a total of 33 distinct combinations of species, energy and collision configuration (see “Thirty-three ways” figure). The STAR detector also operated in fixed-target mode, with gold beams at 13 energies from 3.85 to 100 GeV/nucleon.

Thirty-three ways

The ability to collide nuclei of varying sizes and shapes has opened a new window on their structure. Starting with the U–U run, the near-perfect liquidity of the QGP has been used with increasing precision to take snapshots of nuclear shape. The method works because collective flow encodes the geometry of the initial overlap region in the momenta of the emitted particles, allowing the shapes of the initial nuclei to be inferred. Comparisons of p–Au, d–Au and 3He–Au collisions extended this to small systems, providing evidence of fluid behaviour in small QGP droplets, while the isobar run took it to high precision by comparing 96Ru–96Ru with 96Zr–96Zr.

A quarter century of RHIC beam operations ended on 6 February 2026 with a replay of the 2021 O–O collisions, the only ion species run in common with the LHC

That two-month-long run used stochastic cooling and levelling to alternate between the two isobars daily, with 20-hour stores at constant luminosity. Beyond the shape differences of the two nuclei, the uniform data-taking conditions allowed for subtle measurements of charge versus baryon transport. These provide indications that topological configurations called baryon junctions, rather than valence quarks, carry baryon number in high-energy collisions. A quarter century of RHIC beam operations (see “Long runs” figure) ended on 6 February 2026 with a replay of the 2021 O–O collisions, the only ion species run in common with the LHC (CERN Courier May/June 2026 p10).

The phase diagram

Colliders normally aim for the highest possible collision energy. RHIC also pushed the other way, far below its nominal injection energy. Operating there is challenging, since larger beam sizes, increased superconducting magnet field errors, and stronger space-charge and intrabeam-scattering effects all reduce collision rates. The effort called for extensive optimisation of machine and beam parameters. To increase the luminosity at the lowest energies, the new Low Energy RHIC electron Cooler (LEReC) was built, in which a beam of electrons travelling alongside the ions draws heat from them.

Long runs

This was the world’s first electron cooler based on an electron beam accelerated by radio-frequency (RF) systems. Earlier coolers relied on electrostatic acceleration to produce the very small momentum spread required for effective cooling. RF acceleration can reach much higher energies, but produces a large momentum spread, requiring multiple RF correction stages before the beam can be used for cooling. By overcoming this limitation, the RHIC electron cooler established a technology critical for the EIC.

The QCD phase diagram, when viewed on the axes of baryon chemical potential and temperature, is potentially quite rich. At the highest RHIC energy, where the baryon density is small, lattice QCD calculations suggest that the phase transition from QGP to bound hadronic matter is a smooth crossover (CERN Courier September/October 2026 p41). At high baryon density, instead, effective models of QCD indicate that the transition should be first order. If a crossover and a first-order transition line both exist, they meet at a QCD critical point. Its signature would be a non-monotonic variation with collision energy of the event-by-event fluctuations in the number of net protons produced. The second phase of RHIC’s beam energy scan, enabled at its lowest energies by electron cooling, searched for it in collisions from 7.7 to 19.6 GeV, as well as in fixed-target collisions from 3 to 13.7 GeV. RHIC’s Beam Energy Scan has provided the world’s most systematic exploration of the QCD phase diagram at finite baryon density, including measurements of fluctuations, correlations and collective observables sensitive to the possible critical point and to the nature of the transition. Beyond the search for the critical point, the broad, high-statistics dataset has enabled precision studies of how the fluid responds to baryon doping – the excess of baryons over antibaryons that grows as the collision energy falls – and under what conditions fluidity breaks down.

Enter sPHENIX

The sPHENIX detector, which took over PHENIX’s hall in 2023 with a superconducting solenoid, latest-generation tracking detectors and calorimeters for measuring jets, required the highest possible luminosity in a limited vertex range of ±10 cm. To meet these requirements, a full crossing angle of 1 mrad was implemented, and the machine protection system was upgraded to safeguard the new silicon detectors close to the beam pipe.

Double exposure

In the 2025 run, sPHENIX took more than 200 PB of data, more than all previous RHIC raw datasets combined. High luminosity was also delivered to STAR, with the crossing angle switched between 0 and 1 mrad during the store (see “Double exposure” figure). While sPHENIX is still in the calibration phase for the bulk of its data, it has made standard-candle measurements of transverse energy and multiplicity in Au–Au collisions and has submitted for publication measurements of isolated photons and – for the first time at RHIC – fully corrected dijet momentum-imbalance measurements in p–p collisions. sPHENIX was designed to perform the full suite of jet-related measurements that have developed in the decades since the original jet-quenching discovery, such as fully reconstructed jets, jet substructure and jets containing heavy flavour. The statistical precision of the sPHENIX dataset will allow jet properties to be compared between RHIC and the LHC as a function of energy scale, to tease out the emergence of near-perfect fluidity from the interactions of quarks and gluons.

RHIC has combined sustained collider innovation with a constantly evolving scientific programme

RHIC has combined sustained collider innovation with a constantly evolving scientific programme. It leaves a unique legacy in relativistic heavy-ion and polarised-proton physics, accelerator science and technology, as well as detector and computing development. Its impact extends from fundamental research and broader applications to the future EIC, which will collide polarised electrons with polarised protons and with nuclei to map the gluons inside them. The machine will build directly on RHIC by reusing one of its superconducting rings, its tunnel and the hadron injector complex. Moreover, while beam operations have ended, the rich RHIC dataset will continue to be analysed for at least another decade.

The accelerator gap in cancer therapy

Cancer is responsible for nearly 10 million deaths a year – about one in six worldwide – and the number of new cases is expected to rise from roughly 20 million in 2022 to more than 35 million by 2050, driven by population growth, aging and the increasing prevalence of risk factors. Mortality, however, has been falling steadily. In the United States, for instance, cancer-related death rates dropped by as much as 34% between 1991 and 2022.

The trend hinges on access to modern technologies for early detection, diagnosis and treatment. Among them, particle accelerators take centre stage. Positron emission tomography (PET) relies on radioisotopes produced by particle accelerators to detect tumours with high sensitivity. Radiotherapy – one of the three main pillars of cancer treatment, along with surgery and chemotherapy – targets them with high-energy X-rays or charged particles, such as protons and heavier ions.

The large majority of medical accelerators are compact machines used for conventional X-ray radiotherapy, with about 17,000 units operating worldwide (see “Family tree” figure). Far fewer large units deliver protons or carbon ions for hadron therapy, while few small, low-energy electron accelerators are used for intraoperative radiation therapy. More than 1500 accelerators, typically small cyclotrons, are in operation worldwide to produce radioisotopes for imaging, therapy or the combination of the two, known as “theranostics” (CERN Courier May/June 2026 p30).

Behind these encouraging numbers, however, lies a striking geographical imbalance (see “Unequal doses” figure). Data from the Directory of Radiotherapy Centres of the International Atomic Energy Agency (IAEA) show that most of the radiotherapy equipment is concentrated in high- and upper-middle-income countries. While one of the reasons for this disparity is their higher average age, which makes cancer a more prominent cause of death, it is clear that low- and middle-income countries lack access to modern accelerator-based medical technologies.

Collateral damage

The workhorses of radiation therapy are X-ray linacs. They are compact, roughly one-metre-long linear electron accelerators reaching an energy between 5 and 25 MeV. Their electron beam impinges on a tungsten target, where it produces high-energy X-rays that destroy cancer cells. In most of them, the accelerator is of the “side-coupled” type, developed in the 1960s at the Los Alamos Meson Physics Facility (USA) for studies of nuclear and sub­nuclear matter. This compact design delivers a stable and reproducible beam by packing the accelerating cells closely and keeping the field between them steady. Although the accelerator at their core is relatively simple, radiotherapy linacs incorporate many sophisticated subsystems and require significant operating infrastructure and regular maintenance. Moreover, X-ray radiation attenuates only slowly as it passes through the body, and reaching deep-seated tumours requires a beam to traverse large portions of healthy tissues. Collimators and multi-angle delivery can limit this exposure, but some collateral damage is unavoidable.

Family tree

For this reason, protons or light ions have been proposed as an alternative to X-rays, taking advantage of the Bragg peak – the characteristic energy deposition of heavy particles in matter, which exhibits a sharp maximum at a well-defined depth (see “Stopping power” figure). To reach tumours deep within the body, however, these particles must enter with considerable energy, about 250 MeV for protons and 420 MeV per nucleon for carbon ions, the only two species currently licensed for clinical use. Because of the greater particle mass, proton- and ion-therapy accelerators are considerably larger and more expensive than electron linacs for X-ray production. While a treatment room equipped with a conventional X-ray linac has a footprint of about 50 m2, a single-room proton therapy system requires – shielding included – about 500 m2, and a carbon-ion therapy facility with three treatment rooms can occupy 10 times as much. While proton therapy is now commercially available, carbon–ion therapy is by contrast only accessible in custom-built centres equipped with large synchrotrons of more than 20 metres in diameter, only a few of which have a gantry for multiple-angle beam delivery. Two new pre-commercial carbon–ion therapy facilities are, nonetheless, under construction: a synchrotron-based one built by Hitachi for Mayo Clinic (Florida, US), and a cyclotron-based one built by IBA and Normandy Hadrontherapy in Caen, France.

Specialised conditions

The limited availability of accelerator-based therapies is related to their cost and complexity. High initial capital investment is followed by heavy operational expenses, often sustained through costly annual service contracts. The machines also require a clean and stable environment, particularly in terms of electricity supply, and both operation and maintenance call for highly specialised personnel. Many of these conditions are lacking in low- and middle-income countries, where there are documented cases of radiotherapy units donated to hospitals that were never installed, or that ceased operation when faults could not be repaired. These constraints are most severe for particle therapy. By the end of 2024, 126 proton and 17 carbon–ion therapy facilities were operating in 25 countries, growing by about eight and fewer than one per year, respectively. Two limitations beyond cost and complexity further contribute to this slow growth.

Unequal doses

The first stems from the Bragg peak itself, which is inherently precise but requires comparable or greater precision in tumour imaging and localisation, as well as in patient positioning and motion control – for example, to account for breathing during treatment. The second arises from the therapeutic goal of particle therapy, which is not to increase the dose delivered to cancerous cells with respect to conventional X-ray therapy, but rather to spare healthy organs. This, in turn, brings an improved quality of life and reduces the likelihood of complications or secondary cancers, which is the reason why particle therapy is recommended for paediatric patients. Quality of life, however, is difficult to quantify, requires long-term follow-up and, for part of the medical community, does not yet justify the higher cost of treatment. As a result, particle therapy is currently recommended only for a selected range of cancer types, although research – in particular through clinical trials – keeps expanding it. One area where the use of particles is fully justified is the treatment of radioresistant large solid tumours, which survive X-ray therapy and require particles with high linear energy transfer.

To increase access to particle therapy, progress is also needed on the clinical and pre-clinical side, as well as in the way particle beams are delivered to patients. Perhaps the most intriguing new direction is FLASH therapy. Recent research shows that delivering a radiation dose in milliseconds rather than minutes can spare healthy tissues, while producing the same effect on a tumour as conventional radiotherapy. Although still to be demonstrated clinically, this discovery has triggered intense work to understand the underlying biological mechanisms, as well as to adapt accelerator designs for possible future use with FLASH (CERN Courier November/December 2025 p45). The most promising particles for FLASH are high-energy electrons, for which the production of short high-intensity pulses is easier. Tests of FLASH extraction with proton and carbon–ion beams are also underway in several laboratories, with the aim of combining the advantages of the Bragg peak and the FLASH effect in the so-called “conformal FLASH”.

Stopping power

Studies at GSI in Darmstadt are exploring mixed helium–carbon beams for real-time imaging during treatment, using helium to traverse the body for imaging while carbon delivers the therapeutic dose. An alternative approach involves the use of radioactive ion beams, which can be visualised by scanners. Beyond proton and carbon, work is progressing on helium and heavier ions, to increase the effectiveness of the dose and the accuracy of its distribution. Vertical irradiation of seated patients is also being investigated as a way to avoid the cost and complexity of large proton and ion gantries, while the combination of ion therapy with immunotherapy could enhance and exploit the immune response that follows particle treatment.

The limited availability of accelerator-based therapies is related to their cost and complexity

Still, accelerators remain the most clearly defined technical components to optimise. The main challenges concern reliability, ease of operation, compactness, integration of new treatment modalities and, of course, cost. While reliability matters for all radiotherapy accelerators, it is particularly critical for expanding access to conventional X-ray therapy in low- and middle-income countries. A major effort to make radiotherapy linac design more robust is that of the STELLA collaboration (Smart Technologies to Extend Lives with Linear Accelerators), launched by the International Cancer Expert Corps together with CERN, UK scientific institutions and partners in African countries. STELLA aims to redesign radiotherapy linacs to make them more modular, serviceable and resilient, particularly for use in demanding operating environments where power instability, heat, dust and scarce spare parts can put conventional machines out of service for months at a time. The project is currently in the design/pre-engineering phase, aiming at a first prototype to be installed at the STFC Daresbury Laboratory in the UK.

Another avenue for increasing access to both X-ray and particle therapy lies in simplifying accelerator operation and integrating it more closely with treatment through machine learning and artificial intelligence (AI). Such tools can improve operational reliability by detecting anomalies early and guiding preventive maintenance, while enabling automated operation with minimal reliance on human operators. Coupling AI-enabled accelerator software with its clinical counterpart could also optimise the interface between accelerator and patient. Over the long term, this may reduce operating and maintenance costs enough to transform the particle-therapy business model.

Compact collaboration

A primary challenge in designing particle-therapy accelerators is to make them compact, yet able to integrate emerging treatment modalities. A promising example is the next generation of electron linacs, built to deliver the ultra-high dose that FLASH requires. Compact linac designs are being developed in collaboration with industry by several European laboratories, including CERN, which has adapted the Compact Linear Collider (CLIC) X-band technology for this purpose. These designs operate at different frequencies, reflecting a trade-off between compactness, cost and reliability.

Ring of helium

Synchrotrons are also being redesigned at CERN, together with partners interested in ion-therapy accelerators, under the Next Ion Medical Machine Study (NIMMS). Initially, NIMMS contributed to improving the design of a conventional carbon-ion synchrotron-based facility for SEEIIST (South East Europe International Institute for Sustainable Technologies), a planned cancer therapy and research centre in South East Europe. The next step was a simpler and more cost-effective accelerator: a compact, triangular synchrotron accelerating helium ions to treatment energies, with a total circumference of 18 metres (see “Ring of helium” figure). Helium is increasingly viewed as a natural evolution of particle therapy. Relative to protons, it scatters less in tissue, which sharpens the dose distribution, and has a higher linear energy transfer, while its nuclei fragment less than those of carbon. The proposed accelerator can treat deep-seated tumours using helium or protons, and head and neck cancers – which require less penetration – with carbon. Multiple-energy extraction and FLASH-type irradiation are possible, as well as the use of multiple beams for particle radiography.

Defining magnets

In parallel with synchrotron studies, NIMMS has also promoted the development of superconducting magnets for compact particle-therapy synchrotrons within the European projects HITRIplus and I.FAST. Three demonstrators of canted-cosine-theta (CCT) dipole magnets will soon be tested. In a CCT magnet, two layers of solenoid-like windings are tilted in opposite directions, so that their fields along the beam axis cancel while their transverse fields add to form the dipole. One demonstrator combines dipole and quadrupole functions, one is strongly curved, and one uses high-temperature superconductors (HTSs). The goal is to define a magnet that could be the basis of a compact superconducting synchrotron for carbon–ion therapy with a similar lattice and dimensions to the helium synchrotron equipped with conventional magnets. In a similar direction, Japan’s National Institutes for Quantum Science and Technology (QST), home of the HIMAC facility, is constructing a small superconducting synchrotron for carbon ions with a 30-metre circumference.

Superconductivity remains the principal route to further size reduction of ion-therapy accelerators

Superconductivity remains the principal route to further size reduction of ion-therapy accelerators, and here HTSs are particularly promising. The cost of HTS tapes is decreasing rapidly, and their higher operating temperature offers the opportunity to reduce both the capital cost and the electricity consumption of the cryogenic system (CERN Courier March/April 2026 p30). Teams at CERN are also considering the development of superferric magnets for medical applications, with an iron core and HTS windings. This relatively simple approach could enable low-cost, high-performance magnets whose field can be ramped as quickly as those of conventional synchrotrons, overcoming the slow cycling of traditional superconducting magnets that would limit the number of patients treated per day.

Every technology starts from a dream, and medical accelerators embody one of the oldest: to look inside the human body and cure disease without bloodshed. This fascinating and rapidly evolving field, which stands at the interface between medicine, biology, biophysics, physics and engineering, offers considerable prospects and opportunities to make that dream a reality. 

At the bend in the curve

A generation of physicists has never done research without arXiv. Where did it come from?

Steinn Sigurðsson

Steinn Sigurðsson arXiv came out of the custom of circulating paper preprints. In the 1980s, the major institutes would post the week’s preprints in an envelope, and because that cost money, they only sent them to their friends (see “Paper trail” figure). Harvard sent a single copy of each to Princeton or Berkeley, but not to every university, so there were literally fights over them. As a graduate student, I remember sitting in the hallway, waiting for a professor to finish with the preprints and toss them out.

Then string theory became extremely hot and people did not want to wait for results. Joanne Cohn, then a postdoc at the Institute for Advanced Study, was sending the TeX source of new papers around by email, but the list had grown cumbersome. In the summer of 1991, we were at the Aspen Center for Physics, and somebody said, “This is stupid. We have computers. Why don’t you automate it?” Paul Ginsparg, then at Los Alamos, was intrigued, and by the next day he had written the scripts. All of this happened before the web.

Peer review is often presented as a guarantee of scientific reliability. In practice, physicists read an article on arXiv first, and often only there. What responsibility does that place on the platform?

Steinn The early adopters were professional researchers, so arXiv ran on reputational risk. You put a paper on arXiv when you were ready, on the assumption that if you screwed up we would know, and if you screwed up a lot you were a screw-up.

Paper trail

Multiple studies over the years have found that most papers do not change much through peer review. Although, as I like to point out, adding the word “no” is a tiny change that is very substantive. On the theory side, it may also matter more to be interesting than correct. Many of the big breakthrough papers were concept papers with flaws, and the ones that survived are the ones whose flaws got cleaned up. That distinction is going to matter in the artificial intelligence (AI) world, where there will be a lot of papers that are correct but that nobody cares about.

Most researchers click submit, go to bed and find the paper online in the morning. What happens in between?

Steinn If you are an established author and your papers have gone through without problems, your next paper usually passes unless something raises a concern. We rely on trust. New authors get a closer look. Most of the time the moderator glances at the title and abstract, sees what it is about, often knows the people involved, and it is fine.

A smaller fraction needs follow-up and is put on hold. The paper may belong in a different category, fall outside arXiv’s scope, be completely deranged rubbish, or simply contain something weird that needs checking. And taking a look is not fun. It can be fascinating, but it is not enjoyable. The hold queue has to be cleared as fast as it fills, and that is where arXiv is breaking down, because the holds are building up faster than we can clear them.

Greg Morrisett

Greg Morrisett If I can add to that, there are also lots of well-meaning people submitting things that do not belong on arXiv. Students, for instance, post final projects that may be fascinating documents but are not research for the audience arXiv serves. Many people think arXiv is just a website, when it is really a community of some 270 volunteer moderators and 25 staff. Paul Ginsparg recognised early that the attention of the readers is what is valuable, and moderation protects that attention.

Steinn Exactly. If every introductory computer-science instructor told students to put their term projects on arXiv to learn how submitting to arXiv works, then arXiv would die.

On 1 July, arXiv left Cornell University to become an independent non-profit. Why?

Greg Paul moved from Los Alamos to Cornell in 2001 and brought arXiv with him. The university was never quite sure where to put it, because it is neither an institute nor a research centre. It started in the library, and eventually the provost handed it to me when I was dean of computing and information science. As a computer scientist, I valued arXiv immensely, so I was happy to take it on.

Meeting arXiv’s funding needs was always difficult, and we were limited in our ability to hire developers. I love universities, but they are not set up to run a production web service, an archive and a volunteer community. There was a good argument for staying. Universities endure, and that matters to something that is supposed to last. But there was also a risk that a future dean might not value arXiv as much as I did.

So we created an independent non-profit and moved the staff across from Cornell, with the Simons Foundation and others stepping up to fund it for its first three years. The aim is to give arXiv clearer leadership, more flexibility to hire and a structure built around its own mission, rather than leaving it as an unusual service inside a university.

What will users notice?

Greg Everybody worries about this. On 1 July, hopefully nobody noticed anything more than a change of colour, because we took the Cornell seal off the webpage. That was intentional, because the core mission is exactly the same. There were concerns that arXiv would suddenly be supported through advertising. Not at all.

There will be small changes that add capabilities. The HTML renderer we developed for accessibility turned out to be useful to everybody, and we want better tools for moderation and discovery. But we are not expecting the experience of submitting or reading a paper to change substantially.

Could somebody one day buy arXiv and turn it into something else entirely?

Greg A famous billionaire, who has since passed away, once tried to buy arXiv. We said no. The new organisation has a board drawn from the arXiv community, so that it understands the core mission and ethos. Who knows what the future holds, but no, I do not expect arXiv to sell out. Like any non-profit, it will have to live within its means.

Submissions are surging (see “Overtaken” figure), and AI now writes or assists some of them. From the inside, how much has that changed arXiv?

Steinn The short answer is that things are changing faster than we can deal with them, week by week. We went very quickly from many submissions that were mostly garbage slop to highly polished fluff, a mix of correct although trivial results, but recently including some increasingly substantive and even novel research results (see “Without genius or study” figure). My canonical example is a high-schooler who grabs a public dataset of X-ray images of cancer patients, runs a slightly modified open-source model and concludes it does 1% better at diagnosing cancer. It looks like a paper. It has a claim and a result, but it is done by someone who does not understand what they are doing.

At the same time, genuinely interesting results are arriving at a higher rate. Projects people had set aside, what I call graduate-student test projects, interesting and worth doing but not top of the queue, are being wrapped up in days by teams of AI agents. In modelling or theory, something that might once have been a chapter of a PhD thesis can now take anything from hours to a few days. You can do it live. I was sitting in a meeting the other day and asked Claude, “How about this as an alternative explanation of what the speaker just said?” Claude said, “No, there is this one problem.” So I asked Gemini, “Is Claude right?” Gemini said, “Yes, but there is a way around it.” There was a concept for a paper in 40 minutes.

Overtaken

Disruptive results from AI-assisted work are near. We have hit the bend in the curve. Our current mantra inside arXiv is that the only thing that scales like AI is AI.

AI is also breaking the alignment of incentives. Counting papers is going to become meaningless. If I took this afternoon off, I could probably send out five papers today. So we need to identify the papers on which other researchers actually build. We are going to use knowledge graphs and semantic analysis to see where ideas are going and what later work depends on.

The difficult part is that some of the mechanisms that work can also be retrogressive. Some journals are already moving towards systems where you need people to vouch for you before you can submit. That takes us back towards the 19th century and the idea that somebody has to say you belong to the club.

Greg, what can arXiv itself do?

Greg We were already writing more than people were reading, and that has now exploded. The scarce resource is the attention of readers, their eyeballs and their time. I do not think anybody yet has an answer to how readers should deal with that onslaught.

Without genius or study

One change we have made goes back to trust. An academic email address used to be a useful heuristic. Whoever had one was probably a researcher, so we let it fly. Since January, we have asked for endorsement of new authors, including those with academic addresses, to put some of that reputational pressure back into the system. It seems to have helped a little.

There are also things we can automate. We can scan for dangling references and other signs that something is wrong, then give moderators better evidence about what deserves a closer look. But the social incentives matter just as much. For 30 years, if you posted crap on arXiv, your reputation suffered and people trusted you less.

Endorsement works, but it is antithetical to one of our principles, which is to be open and accessible to the whole world. That tension will be a challenge no matter what.

This year, arXiv turns 35. How would you like to see it at 50?

Greg Still serving the communities and surfacing the research people want to read. I would love to see arXiv expand into adjacent areas, but that takes a willing community with the preprint ethos. You cannot simply add a category to a menu.

I would also like moderators to enjoy the role again. Tom Dietterich used to describe the excitement of having the great papers cross your desk first. We now get around 100 machine-learning papers a day, and no human can read them all and keep a sense of the field. Moderators need sophisticated automation, and readers need better tools to find papers, make connections, summarise and translate.

Paul’s taste in setting up arXiv also matters. LaTeX was not an obvious choice, but it means that papers from 30 years ago can still be rendered in new ways. He was also wise not to compete with journals. I would welcome more open journals overlaid on arXiv, with arXiv remaining the distribution layer and peer review sitting above it.

When you show the site to a non-scientist, they say, “This looks like 1990s HTML,” and we say, “Yeah, exactly.” We are not trying to be flashy. It serves the community well, and we do not want to mess that up.

Steinn I agree with Greg. I would add that arXiv is now a large repository of knowledge, but its primary purpose is to serve up new science. Somebody is going to be doing science, somebody is going to be parsing that science and somebody is going to use it to do more science. arXiv should be there in the middle, enabling that at speed and scale. 

A century along the fifth dimension

When doing physics, Oskar Klein could feel like Odysseus. Perhaps the most adventurous of his voyages came a century ago this year, when he proposed to merge the nascent quantum theory with an expanded, five-dimensional version of the general theory of relativity. Speculation, he explained, “drags many of us into a mental whirlpool not unlike the hydrodynamical one of Charybdis, from which the escape feels like a miracle. On the other hand, sticking too closely to [Scylla’s] facts […] may be equally deadly when using them as building stones for a theory.”

The tension between metaphysical speculation, beyond the observable, and a focus on hard empirical facts was already part of Klein’s earliest youth. He was born in 1894 near Stockholm, where his father, Gottlieb Klein, served as the rabbi of the city’s small Jewish community. Among his acquaintances was Svante Arrhenius – Nobel laureate and leading chemist – who asked Oskar to join his laboratory when he was still a teenager. A fellowship next offered him the opportunity to study in Copenhagen with Hans Kramers and Niels Bohr, who equally wavered between the concrete and the speculative (see “Quantum company” image). Klein defended his PhD thesis, “On the statistical theory of suspensions and solutions”, in Stockholm in 1921.

The unification problem

Klein’s 1926 publication studied Erwin Schrödinger’s successful novel wave equation – arrived at ultimately by foremost expanding on the empirical attainments of atomic spectroscopy – in the five-dimensional theory originally due to German mathematician Theodor Kaluza. Kaluza had seen that introducing an extra dimension could fit electromagnetism and gravity into a single mathematical object, encoding the geometry of a five-dimensional space–time, and thereby “unified” them: he could thus solve the “Unitätsproblem” that had recently begun preoccupying some of the vanguard of theoretical physicists. In his work, Klein pointed out that if that fifth dimension was curled up in a tiny circle (see “Close up” figure), this would account for both its unobserved status and the discrete nature of the electric charge (see “A circle and a tower” panel).

Quantum company

The most prominent member of that vanguard was Albert Einstein. Kaluza had already asked him to present his idea to the Berlin Academy in 1919, and have it published in its proceedings. A problem, however, gave both men pause. The force law for a charged particle contained, alongside the familiar gravitational and electromagnetic interactions, an additional term from a scalar field. In the case of the electron, this would be the leading term, which contradicted its observed behaviour. This convinced Einstein not to forward the work to the Academy, but by 1921 he had changed his mind.

Now deeply involved in unification himself, Einstein sought “non-singular” particle models: smooth field configurations rather than points where the equations break down. His ultimate goal was to offer an alternative route to explain quantum phenomena – top down from the mathematics of unification, so to speak, instead of bottom up from the empirical. In 1927, he repeated elements of Klein’s work in a paper of his own (and said as much), except for its quantum aspects. Likely, that was the material he wished to derive, rather than simply posit by incorporating a Schrödinger equation.

Close up

Einstein returned to five-dimensional unification a number of times, including to the question of whether this offered any non-singular particle solutions, and invariably with the intention to come up with an alternative to quantum mechanics. His case against the latter remained wedded to his agenda for unification, as expressed in a 1927 letter to Hendrik Antoon Lorentz (Vladimir Fock had presented ideas similar to Klein’s): “The quantum theory is completely Schrödingerized and has a lot of practical success because of it. But this cannot be the description of what really happens. It is a mystery. It turns out that the unification of gravity and Maxwell theory is achieved in a completely satisfactory manner by the five dimensional theory (Kaluza–Klein–Fock). I am curious what you have to say about that.”

After the 1920s, Einstein remained fully devoted to his attempts at a unified field theory – one reason why his later boss Robert Oppenheimer in 1935 privately called him “completely cuckoo”, forcefully expressing the dislike of many (see “Hidden dimension” image). Unlike in 1919, mathematical speculation was now leading Einstein’s way, rather than the latest empirical result on quantum particles. As he would say in 1933, in “looking for the mathematically simplest concepts, and the link between them […] lies the theorist’s hope of grasping the real in all its depth.” In a sense, his relation to the five-dimensional theory, from his first rejection to his later persistent and continued elaborations, illustrates his larger philosophical development, in his words of 1938, from “skeptical empiricism” into “someone who seeks the only trustworthy source of truth in mathematical simplicity.”

A circle and a tower

In general relativity, the gravitational field is described by the metric of space–time – the mathematical object that encodes its geometry. In 1919 Theodor Kaluza extended this from four to five dimensions, assuming that no physical quantity varies along the fifth coordinate. Seen from four dimensions, the five-dimensional metric then separates into three parts: the familiar space–time metric, an electromagnetic potential and a scalar field. Gravity and electromagnetism can therefore emerge from a single, higher-dimensional geometry.

Oskar Klein closed the fifth dimension into a circle, with a circumference of 0.8 × 10–30 cm. In quantum theory, a particle’s momentum is inversely proportional to its wavelength. Since a whole number of wavelengths must fit around the circle, momentum in the fifth direction can take only discrete values. Kaluza had conjectured that electric charge is related to motion in the fifth direction. In Klein’s construction, charge is quantised for geometric reasons.

To an observer in four dimensions, momentum in the fifth contributes to a particle’s mass. In the modern reading of Klein’s construction, every particle therefore acquires a “Kaluza–Klein tower” of ever heavier copies, one for each allowed integer momentum along the circle. Producing the lightest copy would require an energy inversely proportional to its size. For Klein’s tiny circumference, this is of order 1017 GeV, far beyond the reach of present experiments.

Smallness, however, is not the only way to hide dimensions. In 1998, for instance, Nima Arkani-Hamed, Savas Dimopoulos and Gia Dvali (ADD) proposed confining the SM particles to four dimensions, leaving gravity alone to spread, and dilute, through the extra ones. This would explain its apparent weakness and allow much larger extra dimensions, with possible signatures at the LHC. A year later, Lisa Randall and Raman Sundrum achieved a similar effect by warping space–time along a much smaller extra dimension. In the simple ADD model with two equal-sized extra dimensions, ATLAS and CMS at CERN have so far excluded radii above a few thousandths of a millimetre.

While in the following decades physicists did occasionally take up the Kaluza–Klein idea, it was only in the 1980s that it would again gain significant traction (CERN Courier July/August 2026 p22). Unlike those of Einstein, these proposals were fully on board with quantum theory. The tension between mathematical speculation and the empirical would, however, remain. The scientific context was different: in the preceding decade, an extensive communal effort had led to the formulation of the Standard Model (SM) of particle physics, and the revived interest in extra dimensions was part of attempts to unify it with gravity. Crucially, the novel higher-dimensional proposals incorporated supersymmetry, a new type of symmetry linking bosons and fermions introduced in the early 1970s.

The comeback of extra dimensions had started with what would become superstring theory, a supersymmetric framework whose basic entities are strings, rather than point particles. The mathematical structure of superstring theory had originated in attempts to model strong nuclear interactions in the late 1960s – notably at CERN, for instance in the work of Gabriele Veneziano – and was reinterpreted as a potential unified theory of particle physics and gravity in 1973 and 1974. Early on in this development, British physicist Claud Lovelace had demonstrated that superstring theory’s purely bosonic, non-supersymmetric predecessor could only be rendered consistent in 26 space–time dimensions – a result he referred to as “obviously unworldly.” Adding fermions and supersymmetry later brought the number down to 10. For a theory of nuclear interactions, situated in four-dimensional space–time, this was still in blatant disagreement with observation.

Compact thinking

For a unified theory with gravity, however, this was no longer necessarily the case. As elaborated in the mid-1970s by Joël Scherk, Eugène Cremmer and John Schwarz, if one conceived of the extra dimensions as real but unobservable – in the spirit of Oskar Klein’s idea – one could make sense of a 10-dimensional theory. Moreover, by making the extra dimensions of space close in upon themselves, the mechanism dubbed compactification, one could generate new gauge symmetries, and thus particles, in the resulting four-dimensional theory.

Hidden dimension

These and other scholars, including Bernard Julia, also began to use hidden extra dimensions in the study of supergravity, a supersymmetric extension of general relativity that was closely related to superstring theory; they formulated its 11-dimensional version in 1978. For some physicists, extra dimensions changed from a theoretical curiosity in conflict with experience to an opportunity for constructing a fully unified quantum theory of the fundamental interactions that could supersede the SM. Perhaps, these extra dimensions could be “real” after all.

A central figure in further advancing this perspective was Edward Witten (CERN Courier January/February 2022 p45), who was appointed professor at Princeton University in 1980. The following year, he published an attempt to construct a supersymmetric higher-dimensional theory unifying gauge theory and gravity, under the title “Search for a realistic Kaluza–Klein theory”. Witten, who had majored as an undergraduate in history in 1971 before turning to physics, devoted his paper’s first two pages to a brief historical introduction to what was then called “Kaluza–Klein theory”, including references to some of Einstein’s contributions.

Different attitudes

Witten argued that the recent developments had led to a different attitude to extra dimensions. Some considered them now “true, physical dimensions”, because the quantum field perspective offered solutions to some of the problems that had confronted those who had worked, like Einstein, in the “classical period” of unified field theories. For example, the scalar field that had initially frustrated both Kaluza and Einstein no longer stood in the way as its associated particle turned out to be practically undetectable when taking quantum corrections to its mass into account. Any motivation to introduce a fifth dimension to produce an alternative to quantum theory was also absent in the new work.

An important reason why interest in the original Kaluza–Klein theory had waned according to Witten had been its inability to produce a realistic model with testable predictions. Yet, the success of the SM made the goal to unify gauge theory and gravity so urgent that it was now warranted to revisit the idea. Moreover, the rise of new particles since the 1950s had tentatively increased Witten’s willingness to consider the additional – actually, infinite – new states that the proposal would entail. When considering the extra dimensions as real, compactification could then arise through spontaneous symmetry breaking, producing the observed four-dimensional theory – in line with the ideas of Cremmer, Scherk and Schwarz.

Short measure

Witten’s brief historical observations on the changing attitude toward Kaluza–Klein theory illustrated a broader change in the outlook of some theoretical particle physicists in the 1970s and 1980s. That is, with the rise of the SM, the immediate objective guiding their efforts was no longer primarily to account for new data, but instead to mathematically unify and thereby “explain” the accepted empirical descriptions. In that sense, their efforts echoed Einstein’s philosophical reorientation – and perhaps they had been equally persuaded by the temptation that Charybdis had presented to Klein.

In the mid-1980s, the theoretical work combining Kaluza–Klein theory and supersymmetry culminated in an increased popularity of superstring theory, supergravity, and the accompanying ideas of extra dimensions and compactification. Their actual existence has remained hypothetical, however. Searches for extra dimensions, from probes at sub-millimetre scales (see “Short measure” image) to recent results by the IceCube Neutrino Observatory (see “Francis Halzen wins the Nobel Prize”), have so far come up empty-handed, constraining their possible size. In any case, the number of physicists working with extra dimensions has increased quite considerably since the 1980s, and has remained substantial. Does this perhaps indicate a change in the attitude of part of the theory community toward the role of experimental knowledge? Does the growth in the effort to attain explanation by unification reflect an increase in its stature as a scientific value, at the expense of striving for novel empirical fact? Or is it simply a reflection of the intrinsically limited possibilities for experiment beyond the SM, into the realm of quantum gravity? One century on, the verdict on the fifth dimension is still out. Klein’s ideas, however, continue to provoke the minds of scientists and, increasingly, of the public at large.

Ukraine’s place in particle physics

In 1877, Ivan Pulyui published “About a radiometer or a light mill” in the Lviv journal Pravda – the earliest known physics article in Ukrainian. Nearly 150 years later, the 10th anniversary of Ukraine’s associate membership of CERN coincides with the 35th of its declaration of independence. These milestones – overshadowed by daily Russian attacks, among them one on the National Academy of Sciences of Ukraine on 28 September – provide an occasion to reflect on the contributions of scientists born in Ukraine to the field of particle physics.

The opening sentence of Georges Charpak’s autobiography is “Born in 1924, in a small town on the Ukrainian plain I was deeply marked by wanderings imposed upon me by geographical, political and cultural events.” These words also apply to many Ukrainians who became scientists despite wars, the Holodomor, the Holocaust, repeated purges of Ukrainian elites and at least six mass emigration waves – the most recent of which is still ongoing. At the same time, scientists from Ukraine helped to establish or led major particle-physics laboratories, with Edward Ginzton at SLAC, Maurice Goldhaber at BNL, Gersh Budker at the Institute of Nuclear Physics in Novosibirsk, Volodymyr Veksler at JINR, and Abram Joffe at the institutes of physics and technology in Kharkiv, Dnipropetrovsk and Leningrad. Through Gleb Wataghin and Marcos Moshinsky, they also played an instrumental role in the development of particle-physics communities in Brazil and Mexico, while Gregory Breit and George Kistiakowsky contributed to the Manhattan Project. George Gamow, who explained alpha decay as quantum tunnelling and laid the foundations of the hot Big Bang model, brought science to the wider public through his popular books. Matvei Bronstein, Moïse Haïssinsky and Stephen Timoshenko were pioneers of quantum gravity, modern radiation chemistry and engineering mechanics, respectively.

Tetiana Hryn’ova

A major milestone for particle physics in Ukraine was the establishment of the Ukrainian Institute of Physics and Technology in Kharkiv (UIPT, now KIPT) in 1928. It was there that Dmytro Ivanenko and Lev Landau – and later their students Oleksandr Akhiezer, Isaak Pomeranchuk, and Evgeniy and Ilya Lifshitz – laid the foundation of their schools of theoretical physics. Victor Weisskopf worked in Kharkiv, while Niels Bohr, Robert Van de Graaff and Paul Dirac came for visits. Type II superconductors were discovered in 1935 at UIPT’s cryogenic laboratory by Lev Shubnikov with colleagues Volodymyr Khotkevych, Georgy Shepelev and Yuri Rjabinin (CERN Courier November 2011 p17). Their success, however, led to a downfall. In the second half of the 1930s, at UIPT alone, 16 members of personnel were arrested for espionage, sabotage and “Ukrainian bourgeois nationalism”; seven were executed and one died in prison. Although Kharkiv was home to a 2 GeV electron accelerator between 1965 and 1990, a CERN delegation visiting in 1992 noted that “Few people we have met had had the opportunity to go abroad and our visit was for many the first direct contact with CERN.” For those who did make it to CERN, the experience was life-changing. Among them was Dmitry Volkov, who played a crucial role in the development of supersymmetry and supergravity with his KIPT colleagues Volodymyr Akulov and Vyacheslav Soroka.

Although Kyiv hosted the International Conference on High Energy Physics (ICHEP) in both 1959 and 1970, research carried out by Ukrainian scientists and institutions was rarely recognised internationally as such. Ukrainian-language scientific journals were closed, scientific institutions were systematically Russified, and many of their graduates were prevented from remaining in Ukraine to pursue their careers. Only a few scientists, most notably Mykola Bogolyubov, dared publish their research in Ukrainian. It is therefore hardly surprising that, writing in 1972, Weisskopf described Kharkiv, where he worked in 1933 and which was then the capital of Soviet Ukraine, as being in Russia. More troubling, in my opinion, is that, as recently as 2025, at the 4th International Symposium on the History of Particle Physics at CERN, research from scientific institutions in Kharkiv and Kyiv was presented under the title “Post-1980s Diaspora of Russian Particle Physicists”.

Strong foundations

Following the disintegration of the Soviet Union, one of the priorities of the government of independent Ukraine was to join CERN. This effort was driven by Ukrainian scientists and research institutes, whose achievements provided a strong foundation for the collaboration. Ludmyla Nagorna, from the Institute for Single Crystals in Kharkiv, introduced new techniques for growing large PbWO4 crystals, which are now used in the CMS and ALICE calorimeters. The advanced ultralight interconnection technologies, championed by Gennady Zinovjev from the Bogolyubov Institute for Theoretical Physics in Kyiv, found applications in the tracker of the ALICE experiment. Since the 1990s, Ukrainian scientists have contributed to a wide range of particle-physics experiments, in particular at B-factories, Fermilab, DESY and CERN.

As discussed extensively at the “Ukrainian contributions to particle physics: history and prospects” workshop held at CERN and in Kharkiv in January 2026, Ukraine’s associate membership of CERN marked not a beginning, but a continuation of a century-and-a-half-long scientific journey, reaffirming the role of its researchers in shaping the future of particle physics.

A shutdown in name only

The LHC completed its third run at the end of June. Is that when Long Shutdown 3 (LS3) began?

For the LHC, it was. It entered LS3 on 29 June, at exactly 6 a.m. But for some parts of the CERN complex, LS3 had started a year earlier. The first large activity, the removal of the target area of the former CERN Neutrinos to Gran Sasso (CNGS) beam facility, began in June 2025. ISOLDE then stopped in December 2025 for a complete renewal and upgrade of equipment several decades old and will take beam again in 2028. Starting early makes it possible to use the beam as soon as it returns, spreads the load on our resources and gives teams time to ramp up and train.

As with the other facilities, the LHC electrical circuits were locked out, so that people could access the equipment safely. Then its superconducting magnets, which operate at 2 K, had to be warmed up before the interventions could start. Bringing some 36,000 tonnes of metal to room temperature, one sector after another, took two and a half months. In parallel, we tested the machine’s mechanical and electrical integrity, looking for any non-conformity to fix for reliable operation in Run 4. Once the warm-up was complete, in September, the intensive removal of components began, to make room for the High-Luminosity LHC (HL-LHC).

What exactly is being removed, and what will replace it?

About 1.2 km of the LHC will be renewed, 300 m on each side of ATLAS and CMS. Along that stretch, the focusing magnets close to the experiments, which make the size of the beams as small as possible at the collision point to maximise the luminosity, will be removed together with other beamline equipment, such as collimators, valves and other magnets. The removal of the cables, cable trays and piping will follow, so that at some point this section of the tunnel will be completely empty. The magnets will then be replaced by new, higher-performing ones, or upgraded and put back in place in the last phase of LS3. The cryogenic line feeding them will also give way to one dedicated to the HL-LHC equipment, providing more cooling power.

That is a massive technical challenge. What about the people involved?

This huge effort covers the LHC and its experiments, the injector complex and its experimental areas, and a large part of the technical infrastructure, so literally every CERN site is touched. It involves CERN staff, collaborators and contractors, with more than a thousand extra people coming just for LS3. Those who operated the accelerators are now carrying out interventions or already preparing for the restart. And since the work will last years, with many people joining as students, graduates or new staff, we must also guarantee the continuity of expertise.

How does LS3 compare in scale with the two previous long shutdowns?

The first, in 2013 and 2014, was triggered by the need to consolidate the electrical splices, the connections between the LHC’s superconducting magnets. The second was mainly dedicated to upgrading the injectors’ chain, to prepare the beams needed for high-luminosity operation of the LHC, but it also covered the upgrades of ALICE and LHCb, the consolidation of the insulation of the diodes that protect the LHC dipoles and much of the underground civil engineering for the HL-LHC, which could not be done during LHC operation because of the vibrations. It started at the end of 2018 and was supposed to end in 2020, but it lasted longer than expected because of the COVID-19 pandemic. LS3 is larger than either, in volume and in extent, both in time and geographically. How much larger depends on the group, but for some the work is two or three times greater.

When did the planning start, and what did it involve?

The trigger for this third multi-year shutdown was the HL-LHC, with the upgrade of the LHC and of its detectors, mainly ATLAS and CMS, so that they can digest the increased luminosity. The installation has been under study for more than 15 years. Since the machines would stop anyway, it was the opportunity for other upgrades, like the complete renovation of the North Area, where some equipment is as much as 40 years old. Drawing on the first two long shutdowns, a coordination team was set up in 2021. It now has about 15 people, experts in planning, logistics and safety – which is our first priority – and representatives of the HL-LHC. We checked that every planned intervention was ready and that all were compatible, first for safety, minimising co-activity between teams working in the same place, and then for resources, which are limited.

Let’s talk about the schedule. The first estimate for the work in the LHC came to a little over 51 months, and it was brought down to 47. How did you manage that?

It came from the LS3 Readiness Review in 2024, where a bottom-up evaluation of all the activities in the LHC exceeded the allocated time. And that figure was already optimised. A reasonable target of 47 months was then set, and we switched to a top-down approach, with the hard objective of fitting into that window.

To achieve that, we reviewed in detail every single activity on the critical path, the sequence of tasks where any delay pushes back the end of the shutdown, and on the near-critical paths. We went to a very fine granularity, both geographically and in time, to maximise parallelism, which was one of the keys to gaining time. Another knob was double-shift work, used at an unprecedented level for key activities on the critical path, such as cabling and de-cabling. We will even use the night shift for transport, which restricts access, and for inspections such as radiography. Shift work, parallelism and the removal of any margin mean that the schedule is highly optimised. The adverse effect is that we have less flexibility to face unexpected events.

How fine did that granularity get? Can you give us an example?

I will give two. One important activity is drilling 28 vertical cores, 1 m wide and 8 m high, to connect the LHC tunnel with the new underground galleries built for the HL-LHC. The galleries were excavated during the second long shutdown, but the final connection has to be drilled during LS3. In the original schedule, drilling was supposed to be the only activity in that part of the tunnel. We divided the 250 m of tunnel into several smaller portions of a few tens of metres, to allow other civil-engineering or infrastructure work in parallel with the drilling.

The other example is in time. The critical path passes through Point 1 and Point 5, where ATLAS and CMS sit and where most of the HL-LHC work takes place. For some activities there, like the electrical lock-out at the start of LS3, we planned down to the day, or even the half-day. We did not do that for all activities, but that was the level of detail. Altogether, with the flexibility, commitment and experience of all the teams involved, and the support of CERN’s management, we just managed to fit into these 47 months.

Forty-seven months is still a long stretch…

It will be a marathon for the team [chuckles], that’s for sure, and far from plain sailing, hopefully with not too many hurdles. But it breaks down into phases common to most facilities. The first is testing, to identify any non-conformity or weakness that we must fix, followed by removal campaigns, to clear space for the improved or new equipment. For the LHC specifically, the drilling of the vertical cores connecting the LHC and HL-LHC worlds will be a key phase. After that, the new technical infrastructure will be installed, then the accelerator or experimental equipment, and finally all the superconducting systems will be cooled down. The last phase before operation is the hardware and beam (re)commissioning (see “Out of phase” figure).

You mentioned that ISOLDE will take beam again in 2028, and some of the people now working on LS3 will be back operating accelerators. How will that work?

It will be a real challenge for managing resources, because the same groups will be supporting operation and carrying out LS3 interventions. The accelerators always restart in order, from the source to the LHC, but after a winter break that takes a few weeks. This time the restart will span two years, from mid-2028 to 2030, which will be a first at CERN on such a scale. So there will be priorities to define, and arbitration. The HL-LHC comes first. It’s a must. By that I mean the project itself, the upgrade of the experiments and all the other projects needed for the LHC to enter the high-luminosity era efficiently and safely.

Out of phase

One of them is the access and alarm system, which dates from the LHC’s installation. The new one has to be ready by the end of 2028, because in the first part of 2029 we will start cooling down some LHC sectors to commission all the new equipment, and for that we will need a safe access system.

Another is the beam-screen treatment, which will deposit a thin layer of amorphous carbon on 12 km of beam screens, to keep the heat load from the beams within what the cryogenic system can handle.

Is it your team that sets those priorities?

We aim to optimise the schedule for efficient resource allocation, aligned with CERN’s priorities as defined by top management. In addition, I see our role as one of support and service for the groups and projects. They manage their resources and set their objectives, and we are there to give them the best conditions to reach those objectives safely. We also help with logistics, managing more than 40,000 square metres of requests for storage and site compounds – including underground, where space is scarcest. And we foster discussion between the people working on site, because teamwork, flexibility and understanding each other’s constraints are key to solving the everyday issues we encounter.

What is your typical day, if there is one?

That’s an interesting question. The great thing about this work is that it’s not routine, and we face very different issues. Sometimes we have to adapt and reshuffle work because we have discovered a non-conformity that needs fixing. Sometimes unexpected interventions have to be studied and integrated into the already heavy schedule, because we judge them important for the machines’ long-term robustness. If there is a fil rouge, it is the many meetings, presentations and exchanges with all the stakeholders, to understand their needs, their constraints and where there is room for flexibility. People at CERN are committed to their own work, but they also understand the global picture and the interest of the Organization.

How is the mood among the teams, now that the work has started?

For every machine, when operation stops, it is the end of an adventure for the physicists gathering data, and it can be a bit sad. But it is for the best, since it will bring improved machines. On our side, the team has prepared LS3 for years. In this first phase the feeling is that people are eager to start the actual interventions in the field. Seeing this become concrete makes for a nice atmosphere on – and under – the ground. Of course, there will be difficulties to overcome, but I am confident in the team spirit and the commitment of everyone, at all levels. LS3 has to be a success, to boost performance, improve reliability and broaden the physics programme. Everyone contributing to it is paving the way for the discoveries of the decades to come.

PSI captures positrons for the FCC

In July, an experiment at the Paul Scherrer Institute (PSI) in Switzerland produced, captured and transported positrons using a system developed for CERN’s proposed Future Circular Collider (FCC). The results from the PSI Positron Production (P3) experiment at SwissFEL, the laboratory’s X-ray free-electron laser user facility, show that the key elements of the FCC positron-source concept can operate together.

“These are the first positrons for the FCC project, so it’s an important milestone,” says Paolo Craievich from the PSI Center for Accelerator Science and Engineering, who is co-leader of the P3 project.

The electron–positron FCC (FCC-ee), recommended by the CERN Council as the preferred option for CERN’s next flagship project (CERN Courier July/August 2026 p7), would collide electrons and positrons in a 90.7 km ring at the Z pole, the WW threshold, the ZH production peak and the top-quark pair threshold. One of the major technical challenges for the FCC is its positron source. While positrons can be produced easily by firing electrons into a dense target, capturing them in large numbers and efficiently channelling them into a tight beam is much more difficult. “The positrons fly out at large angles, and typically many are lost,” says Craievich. “With our new technology, we should be able to significantly increase the positron yield.”

Capture magnets

At the Z pole, each of the two collider rings would hold 11,200 bunches, kept full by continuous top-up injection from a booster in the tunnel. The injector would supply the booster with pulses of four bunches at 100 Hz, with up to 5 nC per bunch, through a damping ring that reduces the transverse emittance of the positron beam by four orders of magnitude and a high-energy linac that takes it to 20 GeV. The yield of the positron source, the number of positrons captured for each electron sent to the target, sets the charge and power of the drive beam and the heat load on the target. The capture magnets of existing sources are pulsed normal-conducting magnets known as flux concentrators.

Each piece of technology is state of the art

In the P3 setup, electrons from the SwissFEL linac strike a tungsten-target system designed, produced and assembled by CERN, generating a shower of positrons travelling in many different directions. A superconducting magnet surrounding the target then captures as many of them as possible. To meet the stringent technical requirements of the FCC positron source, PSI researchers developed a compact sol­enoid made of high-temperature superconductors, which reaches a field up to 15 T.

The magnet’s five coils are wound from rare-earth barium copper oxide (REBCO) tape. The turns are not insulated from each other, so that if a spot in the tape loses its superconductivity, the current can cross to the neighbouring turn and flow around it, instead of heating the spot further and quenching the coil. The system is cooled to 15 K by two cryocoolers and has a warm bore 72 mm wide, in which the target sits 3 cm downstream of the centre, at 12.7 T. From there, the field decreases along the beam direction, converting the large angles of the positrons into a larger beam size that fits the aperture of the accelerating structures. Additional components guide, separate and measure the positrons. Radiofrequency cavities accelerate and bunch them so that they can be transported and, eventually, injected into a collider. The two cavities are S-band standing-wave structures with 40 mm irises, inside normal-conducting solenoids of 0.45 T. Simulations give a yield of about four positrons per electron, and up to seven with superconducting solenoids of 1.5 T in their place.

State of the art

The drive beam of 200 pC bunches at 6 GeV reached a screen in front of the target on 13 May. The solenoid was switched on at the end of June, the cavities were conditioned, and the first positrons were recorded in July on the broadband pickups, scintillating screens and Faraday cups along the line. “Each piece of technology is state of the art,” says Craievich. “For example, it would never have been possible in the past to reach such a high magnetic field without the new high-temperature superconducting solenoids.”

In the beam

Between May and August this year, proponents of the IDEA (Innovative Detector for e+e– Accelerator) detector concept for the FCC tested several prototypes of detector components using secondary beams from the Super Proton Synchrotron and the Proton Synchrotron.

Tracking the position of muons in the outermost layers of the FCC-ee detectors is essential, and one beam test explored a prototype based on micro-resistive WELL (micro-RWELL) technology, which offers a spatial resolution as fine as 50 μm. In parallel, a team working on the IDEA drift chamber, which would track and identify particles in the central part of the detector, analysed the performance of a technique called cluster counting that counts the primary ionisation clusters along a track in a helium-based gas mixture to separate particle species.

The FCC-ee physics programme also requires the energy of particle jets to be determined with a precision beyond that of conventional hadron calorimeters. This is driving the development of dual-readout calorimeters, which read out scintillation and Cherenkov light separately to measure the electromagnetic fraction of each hadronic shower. The beam tests at CERN offered the first opportunity to evaluate the full calorimeter concept under study for IDEA, a fibre prototype with about 10,000 silicon photomultipliers and a crystal electromagnetic section, both of which showed promise for future large-scale developments.

“My congratulations go to the P3 team and the many colleagues across PSI who made this achievement possible, together with their collaborators from CERN, IJCLab–Université Paris–Saclay and KEK,” says Michael Benedikt, FCC project leader at CERN. “This work also reflects the vital support of the Swiss CHART programme in helping to build the expertise and advanced technologies needed to turn the FCC vision into a technically credible project.”

The next step for the P3 team is to increase the beam intensity and optimise the system, with the aim of scaling up positron production to the levels required for the FCC. P3 runs with 200 pC bunches at 1 Hz, while the FCC-ee injector would send bunches of up to 5 nC to its target, four per pulse at 100 Hz. Validation of the positron-source yield is one of the deliverables of the FCC’s reference design phase, during which many other parts of the collider and its experiments are being developed. Six concepts for the FCC-ee detectors are being studied at institutes across the world, and this year prototypes for one of them were tested with beam at CERN (see “In the beam” panel). Beyond the FCC, the high-temperature superconducting magnet technology developed for P3 is also being investigated for applications in proton therapy and fusion research.

Francis Halzen wins the Nobel Prize

On 6 October, Francis Halzen of the University of Wisconsin–Madison was awarded the 2026 Nobel Prize in Physics “for decisive contributions to the IceCube Neutrino Observatory and the discovery of high-energy neutrinos of astrophysical origin”. The observatory, a cubic kilometre of instrumented ice beneath the South Pole, reported the first evidence for high-energy astrophysical neutrinos in 2013. It has since traced them to a flaring blazar, to the active galaxy NGC 1068 and to the plane of the Milky Way.

“The prize was a fantastic opportunity to advertise that this was a large project, and that it only happened because all the milestones we had to reach were overcome by some brilliant people who solved problems along the way,” says Halzen. He and colleagues at Madison – later joined by physicists at Berkeley – had proposed, around 1990, to freeze photomultipliers into the Antarctic ice sheet to catch the Cherenkov light from neutrino interactions. At the time, whether the ice was clear enough for this to work was unknown. “The real miracle, which nobody anticipated, was an absorption length of 100 metres and more for Cherenkov light. That’s why we are sitting here.”

The first South Pole array, the Antarctic Muon And Neutrino Detector Array (AMANDA), was deployed in the 1993–1994 Antarctic summer at depths between 800 and 1000 m, where air bubbles scattered the light too strongly for tracks to be reconstructed. The answer was to go deeper, betting that the bubbles would give way to clear ice, and AMANDA’s later strings hung mostly between 1500 and 2000 m. Between 2005 and 2010, a hot-water drill melted 86 holes into the ice, and a string of 60 optical modules was lowered into each, to hang between 1450 and 2450 m. There, IceCube records about 275 neutrinos a day from the Earth’s atmosphere. Two events of about 1 PeV, reported in 2013, stood out from that background and gave the first indication of an astrophysical flux. Nicknamed Bert and Ernie, they were joined by 26 more in the data from 2010 to 2012, and the sample ruled out a purely atmospheric origin at 4σ. A third year of data raised the significance to 5.7σ in 2014.

Work on the calibration of the ice and on the reconstruction software then helped to bring sources into view. One of them, the active galaxy NGC 1068, first emerged in 2020 and reached 4.2σ in 2022, with far fewer TeV gamma rays than neutrinos (CERN Courier January/February 2023 p10). The deficit points to the absorption of gamma rays in the dense region at the core of the galaxy. “The most exciting future lies in the synergy between observatories and multi-messenger astrophysics,” says Teresa Montaruli of the University of Geneva, member of the IceCube collaboration, “combining electromagnetic radiation, neutrinos and gravitational waves to explore the most extreme phenomena in the universe.” Beyond astronomy, atmospheric neutrinos lend themselves to other precision studies. “IceCube has measured neutrino cross sections through their absorption in the Earth, complementing measurements by experiments such as FASER,” says Montaruli.

The collaboration completed an upgrade of the detector in the 2025–2026 Antarctic summer. Six new strings, more densely instrumented than the original ones, went into its centre to calibrate the ice and to test the optical modules designed for a proposed extension, IceCube-Gen2. The project would take the observatory to about eight times its instrumented volume and add a radio array for the highest energies.

“This Nobel Prize recognises Francis Halzen’s decades of vision and persistence in making high-energy neutrino astronomy a reality,” says Erin O’Sullivan of Uppsala University, spokesperson of the IceCube collaboration. “We are proud to share this moment with Francis, and excited about what lies ahead for neutrino astronomy, particularly with IceCube-Gen2.”

Z bosons entangle in Higgs decays

The ATLAS and CMS collaborations have found evidence that the two Z bosons from a Higgs-boson decay are entangled. Using data from the second and third runs of the LHC, ATLAS rejects the hypothesis of a non-entangled pair at 4.7σ, and CMS excludes the only configuration in which the pair would not be entangled at more than 6σ. Both results rest on assumptions about the CP properties of the Higgs boson, in line with what the two collaborations have measured since its discovery.

The experiments had already observed quantum entanglement between top quarks and antiquarks produced in pairs. Since the top quark has spin ½, a measurement of its spin projection along any given direction has only two possible outcomes, +½ or –½. The Z boson, by contrast, is a massive particle with spin 1, so the corresponding measurement can give +1, 0 or –1. When a Higgs boson, which has spin zero, decays into two Z bosons – one of them off shell – their spin projections along the decay axis must cancel out, and remain correlated whichever common direction they are measured along.

Quantum tomography

Neither spin projection can be measured directly. Z bosons decay almost immediately, occasionally into pairs of electrons or muons that can reach the ATLAS and CMS detectors. Owing to the chirality of the weak interaction, which couples the Z boson differently to left-handed and right-handed particles, information on their spin passes to the leptons, and the directions of the four of them carry the spin state of the pair. Over many decays, their angular distributions allow the spin-density matrix of the two Z bosons to be reconstructed – a procedure known as quantum tomography.

CMS measured the matrix in 138 fb–1 of data at 13 TeV and 62 fb–1 at 13.6 TeV, including the degree of longitudinal polarisation of the two Z bosons. With a significance above 6σ, the data rule out purely longitudinal or purely transverse Z-boson configurations. “Assuming CP symmetry holds, that means the Z-boson pair is entangled. We also tested for CP violation in the decay and found the results to be consistent with CP conservation, justifying our assumption,” says Zhiyuan Huang of Johns Hopkins University. The main challenge, she explains, “was to extract the quantum information from the decay without letting it be distorted by potential differences in the Z-boson masses or by detector effects.” The same polarisation analysis can be applied to Z-boson pairs from other processes. “This study establishes a systematic framework for investigating spin correlations and quantum entanglement through polarisation measurements and moment methods,” says Leyun Gao of Peking University. “For example, future studies could investigate quantum entanglement in Standard Model (SM) ZZ production and vector-boson scattering (VBS) processes involving two Z bosons.”

Curiosity and collaboration

ATLAS combined its full Run 2 data set with 164 fb–1 of Run 3 data at 13.6 TeV. “There were only a couple of thousand events of this type produced at ATLAS during Run 2 and Run 3, and those are reduced further by acceptance and efficiency selections, leaving fewer than 500 Higgs-signal events for the ATLAS analysis,” says Alan Barr of the University of Oxford. Too few, that is, to determine all 80 parameters of the spin-density matrix. Assuming the Higgs boson to be a CP-even scalar, however, leaves the pair only one non-entangled state, and ATLAS tested the full four-lepton angular distribution against it and against the SM prediction. “The data were found to be consistent with the entangled hypothesis, the one predicted in the SM, but inconsistent, at the 4.7σ level, with the only non-entangled state,” says Barr. “In early 2023, we began exploring this idea within the ATLAS Higgs physics group, bringing together a small team of experimentalists and theorists,” remarks Theodota Lagouri of Yale University. “It is truly rewarding to see how our initial scientific curiosity, combined with the enthusiasm, close collaboration and dedication of young researchers, evolved into an exciting measurement of quantum entanglement in Higgs-boson decays.”

A question of qutrits

Both analyses describe the decay at tree level, with the two Z bosons as a pair of three-state quantum systems, or qutrits. Since the Higgs boson is lighter than two Z bosons combined, however, at least one of the pair is off shell. “Once we go beyond the tree-level description of the process, complications arise that can make both the interpretation in terms of two qutrits and the tomography more subtle,” says Fabio Maltoni of the University of Bologna and UCLouvain. In a paper posted in September, he and his colleagues trace the difficulty to the interference of photon- and Z-mediated contributions in the off-shell lepton pair, and show that a lower bound on the entanglement survives beyond leading order. “We argue that we reached an understanding of why these problems arise and provided a safe and robust definition of entanglement which, for practical purposes, is very close to what has been used by CMS and ATLAS,” he says.

The measurement is but one instance of a wider programme

For Maltoni, the measurement is but one instance of a wider programme. “In theories beyond the SM, these quantum properties can be modified,” he says. “Quantum observables may therefore provide additional sensitive tests for new physics by probing aspects of the dynamics that are complementary to more traditional observables. One example that is now being investigated is decoherence: the loss or degradation of quantum coherence and entanglement through interactions with additional degrees of freedom or with an effective environment.”

LZ records an intriguing event

On 16 June 2023, the LUX–ZEPLIN (LZ) experiment, 1.5 km underground at the Sanford Underground Research Facility (SURF) in South Dakota, recorded an event consistent with a nuclear recoil of about 250 keV, in a region of its data where known backgrounds predict almost none. The observation, reported on 1 September, emerged from 220 live days of data examined up to energies above those targeted by conventional searches, with a global significance against the background-only hypothesis of 2.6σ.

LZ uses seven tonnes of ultrapure liquid xenon to look for dark matter in the form of weakly interacting massive particles (WIMPs) from the galactic halo. When a particle scatters off a xenon nucleus, the recoil produces scintillation light and ionisation electrons. An electric field then drifts the electrons into the gas above the liquid, where they generate a second light signal through electroluminescence. The ratio of the two signals distinguishes a genuine recoil off a nucleus from a scatter off an atomic electron, which is what most backgrounds produce.

Conventional searches concentrate on recoils below about 100 keV, where a WIMP would scatter coherently off the nucleus as a whole. LZ’s large target mass allows it to investigate a broader range of possible interactions, including weaker couplings that would grow with the momentum transfer and so become more prominent at higher recoil energies. The recent study explored energies up to 270 keV.

“We found a single event remaining in the analysis, at around 250 keV,” says Richard Gaitskell, a physicist at Brown University and spokesperson for the LZ collaboration. “It would be consistent with dark matter interacting through a slightly more complicated coupling than usual, one that favours large momentum transfers. I should add that with a single event, a dark-matter interpretation is only one possibility.”

The collaboration considered numerous potential sources of background, including accidental coincidences between scintillation light and single ionisation signals, coherent scattering of atmospheric neutrinos, multiple scattering within the detector and high-energy neutrons. “The one that we found most likely, although rare, is multiple scattering,” says Gaitskell. “What we are concerned about is if one of the scattering vertices happened in a region of the xenon where no charge is collected.” In that case, the event would lose part of its ionisation signal and look more like a nuclear recoil. The experiment’s underground location and shielding, with the detector inside 229 tonnes of water and an active liquid-scintillator detector around the xenon, help suppress unwanted signals.

“We are at the level where the probability that the event is background leakage is expected to be about 1 in 200,” says Gaitskell, who – nonetheless – does not exclude a background of a kind not yet identified. “With one event, you have to hold all those possibilities in mind.”

Several theoretical interpretations have been proposed for the LZ event, including the higgsino, a candidate motivated by supersymmetric extensions of the Standard Model. To scatter off a nucleus at the rate needed to produce the LZ event, the higgsino would have to turn from its stable state into a second one, a few hundred keV heavier. Since only the fastest particles in the halo carry the energy to do so, most would never scatter, and the few that did would leave high-energy recoils.

“The higgsino model has really only one free parameter, the energy/mass splitting between the two neutral states, and once that is fixed, it fits the data fairly well,” says Tracy Slatyer, theoretical physicist at the Massachusetts Institute of Technology. “However, for a higgsino you would expect to see more events at higher energy in LZ. The details depend on the velocity distribution of dark matter, which is uncertain, but there is some tension there. Perhaps more importantly, as Maxim Pospelov and Harikrishnan Ramani pointed out, the dark matter in the higgsino model can also annihilate, so for the parameter choices that fit the LZ event you would expect a flux of high-energy neutrinos from the Sun, which has not been observed.” Variations of the model can evade these bounds. “Perhaps the simplest is just increasing its mass, although this makes it a bit more complicated to generate the correct dark-matter abundance,” says Slatyer.

“We now have more than three times as much data available,” says Gaitskell. “We can use that larger exposure to analyse in more detail whether additional events are occurring and whether their signature is consistent with dark matter.” Complementary searches may also be possible with other liquid-xenon experiments, including XENONnT at Gran Sasso, PandaX-4T and its 20-tonne successor at Jinping in China. Meanwhile, design studies are being carried out for a detector of 60 to 80 tonnes of xenon, referred to as XLZD.

bright-rec iop pub iop-science physcis connect