The Nitrogen Buckyball: The “Most Expensive Substance on Earth” and the Atomic Clock It Was Meant to Shrink

Thomas Lee Abshier, ND | 10 October 2026

A Headline That Keeps Coming Back

Every year or two, a story circulates under some version of the title “The most expensive substance on Earth.” The substance is a nitrogen-filled fullerene, written N@C60, and the price quoted is about £110 million per gram. The story was first told in late 2015, when an Oxford spin-out company sold its first sample. The University of Oxford’s website revived it in October 2024, and it is still being passed around today.

The story makes three claims. The first is that this powder is the costliest material anyone can buy. The second is that it will let us build atomic clocks small enough to fit inside a phone. The third is that such clocks will make GPS accurate to a millimeter and so make driverless cars safe. The molecule itself is real and genuinely remarkable. The economics and the applications are a different matter. Following the physics from the molecule to the clock to the car shows what is true in the story, what was exaggerated from the start, and where the research actually stands ten years later.

What the Substance Is

A cage with an atom inside

Buckminsterfullerene, C60, is a closed cage of sixty carbon atoms arranged like the seams of a soccer ball: twelve pentagons and twenty hexagons. It was discovered in 1985 and is one of several forms of pure carbon, alongside diamond, graphite, graphene, and nanotubes. The cage is hollow, and its interior is just large enough to hold a single small atom.

An endohedral fullerene is one with something trapped inside the cage. The “@” in N@C60 means “inside,” so the name reads “nitrogen inside C60.” The molecule contains exactly one nitrogen atom, not several.

A free atom in a molecular box

What makes N@C60 extraordinary is that the nitrogen does not bond to the cage. Ordinarily a lone nitrogen atom is fiercely reactive and would grab the nearest carbon at once. Inside the fullerene it sits at the center, held there by the cage’s repulsion, and keeps the electronic structure of a free nitrogen atom in its ground state. That state has three unpaired electrons whose spins line up, giving a total electron spin of 3/2. The cage perturbs it so little that the visible and ultraviolet spectrum of purified N@C60 cannot be told apart from that of empty C60. In effect the molecule is an atom in a vacuum chamber only a nanometer across, and it can be handled as an ordinary powder or dissolved in a solvent. It is stable at room temperature and decomposes only when heated above roughly 130–150 °C.

Why the spin matters

An electron spin in a magnetic field has distinct energy levels, and transitions between them can be driven with radio or microwave signals. This is electron spin resonance. Most electron spins in solids lose their orientation quickly, because they are buffeted by neighboring atoms and vibrations. The nitrogen spin in N@C60 is shielded by the cage, so it keeps its orientation unusually long. A long-lived spin gives a narrow resonance line, and a narrow line defines a frequency precisely. That single property drives both of the molecule’s proposed uses: as a qubit in quantum computing research, and as the frequency reference in an atomic clock.

Why It Costs So Much

How it is made

N@C60 cannot be made by ordinary chemistry. The standard method, developed in Germany and refined at Oxford, fires a beam of low-energy nitrogen ions, about 40 electron-volts, at a film of C60 being deposited onto a cold target. Occasionally an ion passes through the cage wall and stays inside. The success rate is tiny: typically one filled cage for every ten thousand to one hundred thousand empty ones.

The product is a soot that is almost entirely ordinary C60. Separating the rare filled cages is the hard part. N@C60 and C60 differ in mass by only about 2 percent and are nearly identical chemically. They are separated by high-performance liquid chromatography, run in recycling mode so the mixture passes through the column again and again until the two slowly part. Reaching high purity takes many stages and weeks of instrument time.

The price is per gram, but nobody bought a gram

In 2015 Designer Carbon Materials, the Oxford spin-out founded by Dr Kyriakos Porfyrakis, sold its first sample to a consortium of UK and US researchers: 200 micrograms, for £22,000. That works out to £110 million per gram. The “per gram” figure is an extrapolation from a sample about one-fifteenth the weight of a snowflake. No one has paid £110 million for anything, and the price reflects the labor and machine time behind a first, tiny, high-purity batch rather than any scarcity of nitrogen or carbon.

Calling it “the most expensive thing on Earth” also needs a qualifier. Antimatter, by a NASA estimate, would cost tens of trillions of dollars per gram to produce, although none is for sale. N@C60 can fairly claim to be among the most expensive materials ever sold, priced per unit mass.

The price also matters less than it sounds. A spin-resonance clock would need only a minute quantity of the material, so a high price per gram does not by itself rule out a cheap clock. The real question is whether the clock works.

How an Atomic Clock Works

Locking an oscillator to an atom

Every clock counts the cycles of something that oscillates. A quartz watch counts the vibrations of a crystal, but the crystal’s frequency drifts with temperature and age. An atomic clock fixes this by steering an electronic oscillator so that its frequency matches a transition in an atom, which is the same for every atom of that kind everywhere. The oscillator sends a signal into the atoms, a detector measures how strongly they absorb it, and a feedback loop keeps the oscillator tuned to the center of the absorption line.

What makes a good reference

Two numbers set how well that works. The first is the transition frequency. The second is the width of the absorption line. Their ratio, the line’s quality factor, determines how finely the center can be located. A high frequency and a narrow line make a good clock. The reference must also be insensitive to its surroundings: stray magnetic fields, temperature, and pressure all shift atomic levels slightly, and every shift becomes a timing error.

Atomic clocks are not room-sized

The repeated claim that “typical atomic clocks are the size of a room” was not accurate even in 2015. The primary standards in national laboratories, and the record-setting optical clocks, do fill laboratory tables. But commercial atomic clocks have long come in rack-mounted and briefcase-sized units, and the chip-scale atomic clock, developed under a US defense program in the 2000s, has been a catalog product since about 2011. Microchip’s current model occupies under 17 cubic centimeters, roughly the size of a matchbox, and draws under 120 milliwatts. It is sold for GPS receivers, military radios, unmanned vehicles, and undersea sensors, with a short-term stability of about 3 parts in 10 billion over one second. It works by probing a vapor of cesium atoms in a tiny glass cell with a laser.

So the fullerene clock was never the first small atomic clock. Its promise was a simpler one: a reference that needs no vapor cell, no heater to keep the vapor dense, and no laser, operating entirely at radio frequencies in a solid at room temperature. That could, in principle, shrink power and cost further, toward something a phone could carry.

Nor was it ever going to be “the most accurate time-keeping system in the world.” That title belongs to optical lattice clocks. In 2024 a strontium clock at JILA in Boulder, Colorado, reported a total systematic uncertainty of 8 parts in 10^19, the best of any clock to date. A miniature clock trades accuracy for size, and the fullerene clock was always proposed as a portable device, not a primary standard.

The Fullerene Clock Tested

The clock transition

A spin resonance used as a clock must not shift when the magnetic field fluctuates, and ordinary spin resonances shift a great deal. The solution is a “clock transition”: a particular field at which two energy levels, mixed by the interaction between the electron spin and the nitrogen nucleus, bend so that the transition frequency momentarily stops changing with field. Using the isotope nitrogen-15, whose nucleus has spin 1/2, an Oxford team led by Reuben Harding and Edward Laird, with Porfyrakis among the authors, found and characterized such a transition. They reported it in Physical Review Letters in 2017. The clock transition lies near 38.6 megahertz at a field of about 0.8 millitesla, and they inferred a linewidth there of about 100 kilohertz. It was the first clock transition observed in a molecular spin system at room temperature, a real scientific result.

The verdict of the first full study

The same measurement allowed the first honest estimate of how good such a clock could be. Harding’s 2019 Oxford doctoral thesis, devoted to exactly this question, concluded that the predicted performance is very poor and that the outlook for fullerene-based frequency standards is poor. The reasons are physical, not a matter of engineering polish:

• Low frequency. The clock transition sits near 39 megahertz. The cesium reference in a chip-scale clock is near 9.2 gigahertz, more than two hundred times higher. The nitrogen-cage interaction that sets the fullerene’s frequency is simply small.
• Broad line. A 100 kilohertz line on a 39 megahertz transition has a fractional width of about 3 parts in a thousand. A chip-scale cesium clock locks to a line many thousands of times narrower in fractional terms.
• Coupling to the cage. The nitrogen is not perfectly isolated. Its spin is weakly coupled to the vibrations and motions of the cage, so the hyperfine constant, and with it the clock frequency, shifts with temperature. A free atom has no such degrees of freedom. Harding also measured the pressure dependence. It turned out too small to be used to cancel the temperature shift, though small enough that ordinary changes in air pressure would not matter.

In short, the feature that makes N@C60 special, a free atom held in a solid, is not quite complete, and the remaining coupling to the cage, together with the small hyperfine splitting, costs several orders of magnitude in clock quality relative to the miniature cesium clocks already on the market.

Work continues

That verdict did not end the program. Laird moved to Lancaster University, where his group, supported by the Florida company LocatorX and, since 2025, an Innovate UK program titled “A robust condensed-matter miniature atomic clock,” is building a benchtop fullerene clock. It aims to measure its stability directly and to look for better transitions or other endohedral molecules. That is the right way to settle the question: by building the clock and measuring its Allan deviation, the standard measure of clock stability. As of this writing no fullerene clock has been reported that matches existing chip-scale atomic clocks, and none is in a phone.

What a Clock Could and Could Not Do for GPS

How GPS actually finds you

Each GPS satellite carries atomic clocks and broadcasts its time and position. The receiver measures how long each signal took to arrive and multiplies by the speed of light to get a distance. Light travels about 30 centimeters in a nanosecond, so timing is everything. But the receiver does not need an accurate clock of its own. It treats its own clock error as a fourth unknown, alongside its three position coordinates, and solves for all four using signals from four or more satellites. An ordinary quartz oscillator therefore suffices for phone-grade positioning.

Where the errors really come from

The few-meter error of an ordinary receiver comes mainly from the atmosphere, which delays the signals by variable amounts, from signals bouncing off buildings, and from small errors in the satellites’ own orbits and clocks. None of these is fixed by a better clock in the receiver. Centimeter-level positioning is already available today, through techniques that track the phase of the radio carrier and use corrections from nearby reference stations. Those techniques do not depend on an atomic clock in the car.

A millimeter corresponds to about 3 trillionths of a second of light travel. Claiming that a receiver clock would bring GPS to millimeter accuracy confuses timing precision with positioning accuracy. The comparison of 2 meters with 1 millimeter on a country lane is also misleading, since a car needs to know its lane, a matter of tens of centimeters, and avoids oncoming traffic with cameras, radar, and lidar, not with satellite positioning.

Where a portable atomic clock does help

A good receiver clock still has real value. If the receiver’s time is known independently, one fewer satellite is needed for a fix, which helps in city canyons where the sky is partly blocked. A stable clock also holds time through signal outages, so the receiver can recover its fix quickly when it emerges from a tunnel, and it makes receivers harder to fool with jamming or false signals. Those are exactly the uses chip-scale atomic clocks already serve in military and infrastructure equipment. A clock alone cannot tell a car where it is inside a tunnel. That is the job of inertial sensors and wheel odometry. The case for cheap portable atomic clocks is resilience and holdover, not millimeter positioning.

The Open Questions

The physics of N@C60 is well established. Whether it can become a useful clock is not. Four questions remain.

Stability. Can a working fullerene clock reach the stability of existing chip-scale clocks? The 2019 analysis says the low frequency and broad line make this very hard. Only a measured Allan deviation from a real device will settle it.

Temperature. The cage’s coupling to the nitrogen spin makes the frequency temperature-dependent. A portable clock must either hold its temperature constant, which costs power and space, or find a way to cancel the shift.

Production. Making the material still depends on ion implantation with yields near one in ten thousand and on weeks of chromatography. Even a small quantity per clock requires a reliable and repeatable supply of very pure material.

Competition. The target keeps moving. Chip-scale cesium clocks keep improving in size, stability, and power, and other compact references are under development. A fullerene clock must beat what exists by the time it arrives, not what existed in 2015.

Conclusion

N@C60 is a beautiful piece of physics: a single nitrogen atom, still behaving as a free atom, held in a carbon cage a nanometer across and kept as an ordinary powder on a lab bench. Its price headline is real arithmetic built on a tiny first sale, and it says more about how hard the molecule is to purify than about its worth. The applications attached to it were overstated from the start. Atomic clocks were already the size of a matchbox, the most accurate clocks are optical lattice clocks, and a receiver clock does not bring GPS to the millimeter. The careful work that followed found a genuine clock transition and then an honest, discouraging estimate of how well it could keep time. The question is now in the right hands: build the clock and measure it. A substance famous for its price may end up most valuable for what it has taught about isolating a single atom inside matter.

---

References

• University of Oxford, “This powder is the most expensive thing on earth.” https://www.ox.ac.uk/research/started-in-oxford/powder-most-expensive-thing-earth
• University of Oxford, “Spin-out aims to make ‘bucky-balls’,” April 2014. http://www.ox.ac.uk/news/2014-04-15-spin-out-aims-make-bucky-balls
• ScienceAlert, “Scientists create world’s most expensive material, valued at $145 million per gram.” https://www.sciencealert.com/scientists-create-world-s-most-expensive-material-valued-at-145-million-per-gram
• A. Degeler, “Oxford company now selling endohedral fullerenes priced at £110 million per gram,” Ars Technica, December 2015. https://arstechnica.com/science/2015/12/oxford-company-now-selling-endohedral-fullerenes-priced-at-110-million-per-gram/
• P. Jakes, K. Dinse, C. Meyer, W. Harneit and A. Weidinger, “Purification and optical spectroscopy of N@C60,” Phys. Chem. Chem. Phys. 5, 4080 (2003). https://doi.org/10.1039/B308284A
• “Enrichment of N@C60 by high throughput HPLC processing,” Analyst (2014). https://pubs.rsc.org/en/content/articlehtml/2014/an/c4an00734d
• W. Harneit, “Spin quantum computing with endohedral fullerenes,” arXiv:1708.09298. https://arxiv.org/abs/1708.09298
• R.T. Harding et al., “Spin Resonance Clock Transition of the Endohedral Fullerene 15N@C60,” Physical Review Letters 119, 140801 (2017). https://arxiv.org/abs/1705.04817
• R. Harding, “Endohedral fullerenes as standards for portable atomic clocks,” DPhil thesis, University of Oxford (2019). https://doi.org/10.5287/ora-nyba0nare
• K. Porfyrakis and E. Laird, “To Build the World’s Smallest Atomic Clock, Trap a Nitrogen Atom in a Carbon Cage,” IEEE Spectrum 54, 34 (2017). https://spectrum.ieee.org/to-build-the-worlds-smallest-atomic-clock-trap-a-nitrogen-atom-in-a-carbon-cage
• Laird Group, Lancaster University, “On-chip atomic clocks.” https://wp.lancs.ac.uk/laird-group/research/on-chip-atomic-clocks
• Lancaster University, job description for the Innovate UK programme “ARC-MAC: A robust condensed-matter miniature atomic clock” (2025). https://hr-jobs.lancs.ac.uk/Upload/vacancies/files/12442/JobDescription%202025%20Atomic%20clocks.pdf
• Microchip Technology, CSAC SA65 Chip-Scale Atomic Clock datasheet. https://www.microchip.com/content/dam/mchp/documents/AERO/ProductDocuments/Brochures/00003478.pdf
• A. Aeppli, K. Kim, W. Warfield, M.S. Safronova and J. Ye, “Clock with 8×10⁻¹⁹ Systematic Uncertainty,” Physical Review Letters 133, 023401 (2024). https://arxiv.org/abs/2403.10664
• “The most expensive substance on Earth sells for $140 million per gram,” MSN (originally published 14 October 2024), the article reviewed here.