One thorium site in four: the crystal detail that makes a solid-state nuclear clock buildable

A team led by Thorsten Schumm at the Vienna University of Technology has published a map, in Science, of the four lattice positions a thorium atom can occupy inside a doped crystal, and identified the one that supports a clean nuclear transition ^[5]. That single result converts the solid-state nuclear clock from a materials guessing game into an engineering target, because the other three sites were quietly poisoning the measurement ^[12].

The motivation is not accuracy alone. The best optical atomic clocks, which read time off electron transitions driven by laser light, would drift less than a second over the age of the universe ^[1]. They pay for that with an entire lab's worth of vacuum chambers and shielding to keep stray electric and magnetic fields out ^[2]. Moving the reference from the electrons to the nucleus, flipping a single neutron between quantum states ^[3], could improve precision by up to a factor of ten while letting the atom shield itself, and would allow the atoms to be embedded in a crystal, according to Schumm ^[4].

Only one isotope permits this at all. Flipping a neutron this way normally demands a gamma-ray laser, which does not exist; thorium-229 is the known exception, and in 2024 researchers showed the transition can be driven with ultraviolet light ^[6]. Andrei Derevianko of the University of Nevada, Reno, who was not involved, calls thorium-229 "a quirk of nature" whose transition energy is low enough to be reached by high-purity laser light ^[7].

The host material took fifteen years to work out: calcium fluoride doped with thorium-229 impurities, grown as transparent millimetre-sized cubes ^[8]. Schumm describes crystal growing as closer to cooking than physics, and says the finished piece looks like a scrap of glass carrying tens of thousands of hours of learning ^[9]. The catch the group ran into is the one that matters for hardware: a thorium atom in a bad internal position sees an uneven electric field, and that degrades the clock's timing ^[10].

The experiment to sort the sites was blunt and slow. The group illuminated the crystal with a custom ultraviolet laser at a fixed wavelength for 60 seconds, switched it off, watched for five minutes as excited nuclei decayed and re-emitted, then shifted frequency and repeated ^[11] - roughly six minutes of wall-clock time per frequency point ^[19]. Three of the four sites emitted at multiple wavelengths, the signature of an uneven local field; the fourth answered at a single wavelength ^[12]. Three quarters of the available sites are therefore unusable as a clock reference ^[18].

Outside assessments are supportive rather than hedged. Eric Hudson of UCLA, not involved, says the data is what the field has wanted to see for some time ^[13]; Ekkehard Peik, who heads the Time and Frequency Department at Germany's national metrology institute and was also not involved, calls it a very important result for solid-state nuclear clock research ^[14].

What to watch: Schumm's group has built early prototypes and patented a route to a chip-scale version ^[15], while a group led by Shiqian Ding at Tsinghua University has demonstrated a prototype with comparable results using a stronger laser and a lower thorium-229 concentration ^[16]. Schumm declines to say when nuclear clocks will beat the best atomic clocks ^[17]. The interesting number will be the first published stability figure from a crystal grown deliberately to favour the good site.