Deep divefrom Episode 8New physics & chemistry24 October 202623 min

Why the Thorium Clock Crystal Glows in the Dark

A deep dive into one paper: what was known before, what the researchers did, what the numbers say and what they don't.

The paper: Defect Dynamics and Isomer Quenching in 229Th:CaF2 Crystals (arXiv preprint)

Chapters

Transcript

The script the two AI voices read from. The times are worked out from the script, not heard, so a line can start a few seconds early or late.

Read the whole episode

Cold open

Host 1You heard the trailer — this is the whole story. Welcome to the Hastrup Review of Science — where artificial intelligence reads real science, so you don't have to pretend you understood the paper. Today, one paper about the crystal that the first nuclear clocks are built from, and why it keeps glowing when nobody asked it to.

Host 2The paper is by Ming Guan and eighteen co-authors from Okayama University, the Technical University of Vienna, Kyoto University and the SPring-8 synchrotron in Japan, posted to the arXiv on the first of October twenty twenty-six, so it is a preprint. They measured every way a thorium-doped calcium fluoride crystal gives off light, across a wide range of temperatures. They then found that the temperatures at which trapped electrons are released line up with an odd temperature dependence in how fast thorium nuclei lose their excitation.

Host 1So it is a materials paper wearing a clock's badge.

Host 2Fairly put. It does not build a clock or improve one; it characterises the crystal that several clocks are built from.

Why this field exists

Host 1Start with the clock. Why would anyone want a clock that ticks on a nucleus?

Host 2As established background: every precise clock counts the oscillations of something with a very sharply defined frequency. Today's best clocks use transitions of electrons in atoms such as strontium or ytterbium, probed with lasers. Electrons sit on the outside of the atom, so stray electric and magnetic fields nudge them, and much of clock engineering is about shielding them from that.

Host 1And the nucleus is better hidden.

Host 2It is roughly a hundred thousand times smaller than the atom and wrapped in its own electrons, so it should be far less sensitive to its surroundings. The problem is that nuclear transitions normally need thousands to millions of electron volts, which means gamma rays, and there are no lasers for gamma rays.

Host 1Except thorium.

Host 2Thorium-two-twenty-nine has an excited state, called the isomer, only about eight point three six electron volts above the ground state. That corresponds to vacuum ultraviolet light at about one hundred and forty-eight nanometres, which lasers can reach.

Host 1How long has that been known?

Host 2The paper's reference list tells the history. A very low-lying state was first inferred from gamma-ray spectroscopy in nineteen seventy-six, a nuclear clock based on it was proposed by Ekkehard Peik and Christian Tamm in two thousand three, and an energy estimate near seven point six electron volts followed in two thousand seven.

Host 1Seven point six. So the number moved.

Host 2It did, and that uncertainty is why it took so long, because you cannot hit a narrow line with a narrow laser if you do not know where to look. In twenty twenty-three an experiment at CERN's ISOLDE facility observed the isomer's own light from inside a crystal. As background, in twenty twenty-four laser excitation followed, first by the Vienna group that grew these crystals and soon after at JILA in Colorado.

Host 1And now there are working clocks.

Host 2According to this paper, solid-state thorium clocks have been operated in three laboratories, and their frequencies compared. The transition energy used throughout is eight point three five five seven electron volts.

What was known before

Host 1Why put the thorium in a crystal at all, instead of trapping single ions?

Host 2Both routes are pursued. Trapped thorium ions are clean but few; doping thorium into a crystal that is transparent to vacuum ultraviolet light gives you an enormous number of nuclei at once, which beats down the statistical noise, and no laser cooling is needed.

Host 1And calcium fluoride is the favourite host.

Host 2By far the most studied. Its band gap is about twelve electron volts, well above the isomer's eight point three six, so the clock light can get out, and the thorium sits as a four-plus ion in place of a calcium.

Host 1Four-plus in a two-plus slot. That sounds like it needs balancing.

Host 2It does; x-ray absorption work cited in the paper indicates the extra charge is compensated by additional fluorine ions squeezed in between lattice sites. More recent spectroscopy found four distinct thorium sites, with about the same radiative lifetime at each but a strongly site-dependent quenching efficiency. Which site the clock actually uses is still debated.

Host 1Quenching. Define it.

Host 2The isomer would normally decay by emitting its ultraviolet photon, slowly; as background, inside calcium fluoride that takes on the order of ten minutes. Quenching means something in the environment opens a faster, non-radiative exit, so the excited nuclei disappear sooner and without the photon you wanted to count.

Host 1And what in the environment does that?

Host 2Earlier work treated the crystal as a static holder. The authors' point is that it is not static: electrons and holes move around, defects capture them as so-called colour centres, and those later recombine and emit light.

The open question

Host 1So what was the open question going into this paper?

Host 2Earlier experiments, including one on the same crystal by an overlapping team, showed that x-ray irradiation shortens the isomer lifetime, and that the effect depends on temperature in a strange, non-monotonic way. The leading explanation is that mobile electrons do the quenching, coupling to the nucleus through electronic states inside the band gap.

Host 1Then you need to know where the electrons are stored and when they come loose.

Host 2Exactly, and nobody had measured that. The only systematic luminescence study of this material predates knowing the isomer energy, and no thermoluminescence study had ever been reported, so the trap depths that govern storage and release of charge had never been determined.

Host 1And there is a second, more practical question.

Host 2Yes. The crystal is radioactive, so it glows by itself all the time, and some clocks detect the isomer by counting its faint ultraviolet photons. Every bit of unwanted glow is background for those clocks, and its temperature behaviour had not been mapped.

What this paper did

Host 1So what did they actually do?

Host 2They took two crystals grown in Vienna and measured four kinds of light from them across a wide temperature range. One, called V zero five seven, has about eight times ten to the fifteen thorium ions per cubic centimetre and an activity of six hundred and seventy becquerel. The other, called X two, is about five hundred times more concentrated, at four times ten to the eighteen, with twenty thousand becquerel in a piece of about one cubic millimetre.

Host 1Four kinds of light. Walk me through them.

Host 2Radioluminescence is the glow driven by the crystal's own alpha and beta particles from thorium and its decay daughters. Photoluminescence is the glow while an external x-ray beam or laser hits it, and afterglow is what lingers for minutes after the beam is switched off. Thermoluminescence is light released when you heat the crystal and shake loose trapped charges.

Host 1And each one maps onto running a clock.

Host 2That is the authors' framing: preparation, excitation, detection and annealing. The radioluminescence never stops, the photoluminescence happens during excitation, the afterglow leaks into detection, and heating is how you reset the crystal.

Host 1Where did the x-rays come from?

Host 2The BL nineteen LXU beamline at SPring-8, a synchrotron in Hyogo, Japan. It delivers about ten to the ten photons per second at twenty-nine point two kiloelectron volts with a bandwidth of thirty millielectron volts, tuned to a nuclear resonance that populates the isomer indirectly. For the luminescence runs they quote about ten to the eleven photons per second.

Host 1And the cold part.

Host 2The crystal sat in a vacuum chamber on a holder cooled through a copper rod from a cryocooler, with heaters along the rod, giving a range from minus one hundred and ninety to plus three hundred degrees Celsius. Light was collected with a parabolic mirror and sent either to a spectrometer with a camera or to a pair of photomultiplier tubes.

Host 1Why two tubes?

Host 2One is solar-blind and sees roughly one hundred and thirty to one hundred and sixty nanometres, where the isomer photon lives. The other sees two hundred to five hundred and fifty nanometres and acts as a veto, watching the ordinary scintillation, while mirrors that reflect near one hundred and fifty nanometres filter out the rest.

Host 1Any practical surprises?

Host 2Ice. Below about minus one hundred degrees, residual gas freezes onto the crystal and absorbs the one-hundred-and-forty-eight-nanometre isomer light. They fixed it with a cold copper shield and a magnesium fluoride window, after which the far-ultraviolet count rate stayed flat at about thirty per second from plus two hundred and twenty to minus two hundred degrees.

Host 1How did they do the thermoluminescence?

Host 2They held the X two crystal at minus one hundred and eighty degrees for sixteen hours, letting its own radioactivity fill the traps, then heated it at two tenths of a degree per second to plus one hundred, held there ten minutes to empty the traps, and cooled back down. The spectrometer only covers a one-hundred-nanometre window at a time, so the whole cycle was repeated nine times to stitch together ultraviolet to near-infrared. The radioluminescence recorded on the cooling leg was subtracted to isolate the heating signal.

Host 1Sixteen hours per window, nine windows.

Host 2By my arithmetic, not the paper's, that is close to a week for one map. Each glow peak's trap depth then came from fitting its rising edge with the standard exponential form for thermal release.

Host 1And the afterglow?

Host 2At the beamline they irradiated the X two crystal for one thousand seconds, switched off the beam, and recorded the veto tube for eighteen hundred seconds. Then they stepped the temperature by ten degrees and repeated, from minus one hundred and ninety to plus one hundred and ten. Each decay curve, with the steady self-glow subtracted, was fitted with three exponentials.

Results

Host 1Results. Start with the self-glow.

Host 2The radioluminescence shows bright bands near two hundred and seventy-five and three hundred and ten nanometres, a weaker one near four hundred, and a peak near nine hundred and fifty nanometres whose origin the authors say is unknown. It gets brighter as the crystal is cooled and nearly vanishes above one hundred degrees.

Host 1The paper keeps saying self-trapped exciton. What is that?

Host 2An electron and a hole bound together, with the lattice distorted around them; in calcium fluoride it is essentially a neighbouring pair of an electron trap and a hole trap that recombine within microseconds. At room temperature each burst's tail decays in about one microsecond, and the tail grows longer as the crystal gets colder.

Host 1And the temperature fit?

Host 2It needs two components, which is known for this material: a triplet with a quenching activation energy of about zero point three eight electron volts, and a singlet at about zero point one zero, with intensities of roughly one to seven, singlet to triplet. Between room temperature and one hundred degrees this glow drops by more than two orders of magnitude.

Host 1And the very short wavelengths?

Host 2That is Cherenkov light from fast beta particles, between one hundred and twenty and two hundred nanometres, right where the isomer photon is. It hardly changes with temperature, which is useful, because a steady background can be subtracted.

Host 1What about the glow under the beam?

Host 2At room temperature it looks like the self-glow, dominated by the two-hundred-and-eighty-nanometre band, and about a thousand times stronger than the first ten seconds of afterglow. The afterglow then fades by about a factor of fifty within five minutes and shifts toward longer wavelengths, with a doublet near four hundred and eighty to four hundred and ninety nanometres.

Host 1And when the crystal is hot?

Host 2At one hundred and fifty degrees the main band is quenched, the light shifts red, and after the beam goes off it falls by about a factor of ten thousand within ten seconds. The authors conclude that the hot-crystal emission is really thermoluminescence, deeper traps emptying almost as fast as they fill.

Host 1Did thorium add any colours of its own?

Host 2Not clearly. Thorium four-plus has a closed electron shell, unlike the rare-earth ions that make some calcium fluoride crystals glow brightly, and these spectra look like pure calcium fluoride, so thorium acts as an optical activator weakly if at all.

Host 1Now the centrepiece, the glow curve.

Host 2Heating from minus one hundred and eighty to plus one hundred degrees, they resolve eleven glow peaks, named P one to P eleven, and six emission bands. The peaks fall into three groups: P one to P four between minus one hundred and eighty and minus one hundred, P five to P seven from about minus eighty to plus ten, and P eight to P eleven from ten to one hundred degrees.

Host 1And the trap depths.

Host 2They extracted activation energies for five of the peaks, from zero point one one five electron volts for P one up to zero point five three five for P eleven. The two that matter in a minute are P four, at minus one hundred and five point six degrees, and P five, at minus fifty-five point six, with a depth of zero point three seven two electron volts.

Host 1Only five of eleven got a depth?

Host 2Yes, the table leaves the narrower peaks blank. And for P eleven the authors say the background subtraction is uncertain, because charge was still trapped when cooling began.

Host 1Afterglow numbers.

Host 2Three components, with lifetimes near one, ten and one hundred seconds. At twenty-five degrees the fit gives about zero point nine, nineteen and sixty-six seconds; at minus one hundred and sixty-four, about five point seven, forty and three hundred and forty-four seconds, so everything slows down in the cold.

Host 1And it tracks the thermoluminescence.

Host 2Its amplitudes follow the same temperature regions as the glow peaks, which the authors read as both drawing on one reservoir of trapped charge. That, they argue, makes the afterglow, not the prompt flash, the optical fingerprint of the carriers that matter for quenching.

A word from our sponsor

Host 1Do you hold things at minus one hundred and eighty degrees for sixteen hours and still have nothing cold in your glass? Does your freezer contain one bag of peas and a deep sense of unrealised potential? Then you are in luck.

Host 1The Hastrup Review of Science is brought to you by ice cube trays. Silicone, stackable, and calmly committed to a single phase transition.

Host 2I filled mine last night. By morning every compartment had released exactly what it stored, at exactly the temperature I expected.

Host 1Ice cube trays are very reasonably priced and in stock right now. No synchrotron, no copper shield, no sixteen-hour soak.

Host 2Mine also came with a lid. I was not expecting it to manage the ice layer for me, but it did.

Host 1Ice cube trays. Because some traps are meant to be filled. Our referral link is in the show notes.

Traps and the nucleus

Host 1Back to the nucleus. How do the traps connect to quenching?

Host 2They compare with an earlier quenching experiment on the same X two crystal. In that model only free-moving electrons can quench an isomer, so the quenching rate is the density of mobile electrons times a cross section times their speed. The mobile electrons are those made by the x-rays minus those currently sitting in lattice traps.

Host 1So the traps protect the nucleus.

Host 2In this picture, yes; an electron stored in a lattice trap is not available to find a thorium site. The model then predicts that the isomer yield and the afterglow should share a temperature dependence.

Host 1What did the earlier experiment see?

Host 2Two oddities. Near minus sixty degrees, the isomer's lifetime under irradiation has an unexpected maximum and the number of excited nuclei goes up; from minus sixty to minus eighty that number drops to a minimum, then rises again below minus eighty.

Host 1And the glow curve explains it?

Host 2It lines up with it. The minus-sixty maximum coincides, within uncertainty, with peak P five at minus fifty-five point six, and the minus-eighty minimum sits between the P four and P five release regions. There the afterglow is small, which in the model means fewer trapped electrons and more free ones quenching the nucleus.

Host 1So a bumpy yield from smooth ingredients.

Host 2That is their first consequence: the cross section and the electron mobility each change smoothly with temperature, but the competition between trapping and capture at thorium gives a yield with bumps in it. The bumps sit where the crystal's traps fill and empty.

Why it matters

Host 1So what is the principle here?

Host 2That the host crystal is part of the clock. A nuclear clock was pitched as insensitive to its environment, and the frequency may well be, but how many nuclei you can excite and how long they survive depend on electrons wandering through a defect-filled lattice.

Host 1What changes for whom?

Host 2For groups that read the clock by fluorescence, counting isomer photons, this is a map of their background: self-glow cut by more than a hundred near one hundred degrees, an afterglow with one, ten and one hundred second components, and a temperature-flat Cherenkov floor. For theorists modelling quenching, it is the first measured list of trap depths in this material, between about zero point one and zero point five electron volts.

Host 1They also give operating advice.

Host 2Run near one hundred degrees if background is the main enemy, or at the zero-shift temperature of about minus seventy-seven degrees reported by another group if frequency is, with a copper shield against ice. Between cycles, hold at one hundred degrees for ten minutes to empty shallow traps, and roughly every one to two weeks anneal fully at four hundred degrees for several hours. Keep the vacuum at or below ten to the minus four pascal.

Host 1Two temperatures, two goals.

Host 2Yes, and the authors stress that the two criteria are independent. As my own observation, not theirs, minus seventy-seven sits right next to the minus-eighty dip in isomer yield from the earlier experiment, so anyone choosing that temperature may want this trap map at hand.

Host 1Does it square with the clocks already running?

Host 2Two checks. One clock uses the self-trapped exciton light to monitor laser power, and this paper's quenching curve supplies the calibration to do that across temperature. The same group reported background lingering after frequency scans and distorting line shapes, which matches the afterglow measured here.

Host 1And what does it not mean?

Host 2It does not mean thorium clocks are limited by this glow today. The authors note that the clocks demonstrated so far mostly read out by absorption, with signal rates about a thousand times higher than fluorescence, and the first closed-loop device runs at room temperature with on-off differential detection, where this background is not limiting.

Host 1And it does not solve quenching.

Host 2No. The microscopic mechanism is still open, no trap is tied to a specific defect or thorium site, and the link to the isomer is a correlation in temperature, not a demonstrated cause.

Caveats

Host 1Now the caveats, in full.

Host 2First, the central link is a match in temperature between this paper's glow peaks and quenching features from an earlier paper. Two features lining up with peaks in a region crowded with eleven of them is suggestive, not conclusive.

Host 1Second?

Host 2The authors stress that thermoluminescence measures carriers being released and recombining, not how many are stored at a given moment. Turning the glow curve into the actual trapped-electron density under x-rays needs a kinetic model of capture and release, which they do not provide.

Host 1Third.

Host 2Thorium occupies at least four sites in this crystal, so the spectrum here is a superposition over them. They cannot assign any trap to a specific defect from luminescence alone, and they name a site-resolved measurement as the natural next step.

Host 1Sample size?

Host 2Two crystals from one grower, and the key thermoluminescence and afterglow maps come from one of them, the highly doped X two. Crystals with different doping, growth or radiation history could hold different trap populations.

Host 1Anything shaky in the fits?

Host 2Some. Only five of eleven peaks got activation energies, P eleven's subtraction is admittedly uncertain, one emission band is unresolved, and the nine-hundred-and-fifty-nanometre peak is unexplained. The one-to-two-week anneal interval is a suggestion scaled from the activity, not something tested over weeks of clock operation.

Host 1And the status.

Host 2It is an arXiv preprint, not yet peer reviewed.

What to watch next

Host 1So what would move this forward?

Host 2A kinetic model that predicts the isomer yield versus temperature from these trap depths, rather than lining up features by eye. If it reproduces the minus-sixty bump and the minus-eighty dip, the trapping picture gets much stronger.

Host 1And experimentally?

Host 2Site-resolved luminescence on crystals of known site makeup, as the authors propose, and the same measurements on crystals with other doping levels. A direct test would be to deliberately fill or empty the traps, by heating or by light, and see whether the quenching rate moves the way the model says.

Host 1And for the clocks?

Host 2Watch whether fluorescence-readout clocks adopt the thermal cycling and annealing schedule, and whether their background and line shapes improve. That would be the practical verdict.

Bottom line

Host 1Bottom line?

Host 2This is careful materials characterisation: the first trap-depth spectrum and afterglow map of thorium-doped calcium fluoride, with eleven glow peaks, three afterglow timescales, and a hundredfold drop in self-glow by one hundred degrees. The match between trap release temperatures and the isomer's odd quenching behaviour supports the idea that stored electrons shield the nucleus, but it is a correlation still waiting for a kinetic model.

Sign-off

Host 1That is the deep dive: Defect Dynamics and Isomer Quenching in thorium-two-twenty-nine doped calcium fluoride crystals, by Ming Guan and colleagues, on the arXiv. The link is in the show notes.

Host 1Every Friday this same feed carries the roundup episode: five new papers as trailers, and you pick which become deep dives. This has been the Hastrup Review of Science.