Solid-State Nuclear Clock: Thorium-229 in Crystals

• Solid-state nuclear clocks are frequency standards that use nuclear transitions in a condensed-matter host, providing ultrahigh precision with sub-mHz natural linewidths.
• Experiments embedding Thorium-229 in CaF2 demonstrate controlled quadrupole splitting and temperature-dependent shifts, highlighting both measurement advances and host-induced challenges.
• These systems enable sensitive tests of fundamental constant variations and dark matter interactions while driving innovations in materials engineering for optimized clock performance.

Searching arXiv for papers on solid-state nuclear clocks and thorium-229. Search query: "solid-state nuclear clock thorium-229" A solid-state nuclear clock is a frequency standard in which the reference transition is a nuclear transition realized in a condensed-matter host rather than in an isolated atom or ion. In current experimental practice, the dominant implementation uses the low-energy isomeric transition of $^{229}$Th embedded in wide-band-gap crystals such as $\mathrm{CaF_2}$, $\mathrm{ThF_4}$, and related fluoride or sulfate hosts. The appeal is the combination of a laser-accessible nuclear resonance near the vacuum-ultraviolet, an intrinsic natural linewidth in the sub-mHz regime, and the possibility of interrogating macroscopic ensembles of nuclei in a compact platform. The same solid-state environment that enables large signal also introduces host-dependent frequency offsets, quadrupole splittings, magnetic broadening, strain sensitivity, and quenching channels, so the subject sits at the intersection of nuclear structure, AMO metrology, and defect-controlled solid-state physics (Derevianko et al., 9 Jun 2026).

1. Physical basis and metrological rationale

The $^{229}$Th clock transition connects the $5/2^+[633]$ ground state and the $3/2^+[631]$ isomeric state. The bare-nucleus transition energy is given as $8.272(22)\,\mathrm{eV}$, while solid-state measurements cluster near $8.4$ eV; one quadrupole-averaged solid-state frequency is $2\,020\,407\,384.335(2)\,\mathrm{MHz}$ (Derevianko et al., 9 Jun 2026). What makes the isomer exceptional is that its low energy is understood as a near cancellation of strong- and electromagnetic-interaction contributions,

$$\Delta E=\Delta E_{\rm strong}+\Delta E_{\rm EM} \approx 8.4~\mathrm{eV},$$

or equivalently,

$\mathrm{CaF_2}$0

This structure underlies the frequently emphasized sensitivity of the transition to variation of fundamental constants (Derevianko et al., 9 Jun 2026).

For metrology, the central attraction is the combination of optical frequency and extreme narrowness. The natural linewidth is sub-mHz, corresponding to a natural quality factor of order $\mathrm{CaF_2}$1 (Derevianko et al., 9 Jun 2026). Earlier nuclear-clock analyses framed the same advantage as the use of a nuclear radiative transition that is intrinsically less sensitive to field-induced systematic shifts than ordinary electronic transitions, while solid-state realization offers the possibility of interrogating a very large ensemble of nuclei and thereby improving short-term stability through $\mathrm{CaF_2}$2 scaling (Peik et al., 2015).

Solid-state realization changes the metrological problem. The realized clock frequency is not the bare nuclear frequency, but the transition frequency of a coupled electron–nucleus–lattice system. A plausible implication is that solid-state nuclear clocks cannot be characterized purely by nuclear structure data: host chemistry, defect topology, local electric-field gradients, and magnetic baths are part of the clock definition (Derevianko et al., 9 Jun 2026).

2. Hyperfine structure in crystals and the role of the host lattice

In a crystal, $\mathrm{CaF_2}$3Th does not appear as a free nucleus. In the canonical $\mathrm{CaF_2}$4 platform, $\mathrm{CaF_2}$5 substitutes for $\mathrm{CaF_2}$6 and is charge compensated by two interstitial $\mathrm{CaF_2}$7 ions, producing a defect complex whose microscopic geometry sets the local electric-field gradient (EFG) and therefore the spectroscopic structure (Higgins et al., 2024). Because both the $\mathrm{CaF_2}$8 ground state and the $\mathrm{CaF_2}$9 isomeric state carry electric quadrupole moments, the dominant crystal-field interaction is quadrupolar,

$\mathrm{ThF_4}$0

with principal EFG component $\mathrm{ThF_4}$1 and asymmetry parameter $\mathrm{ThF_4}$2 (Higgins et al., 2024).

This interaction splits the nuclear transition into five magnetic-dipole-allowed lines in $\mathrm{ThF_4}$3. High-resolution work resolved the four strongest transitions directly: $\mathrm{ThF_4}$4, $\mathrm{ThF_4}$5, $\mathrm{ThF_4}$6, and $\mathrm{ThF_4}$7, while the weaker $\mathrm{ThF_4}$8 line was inferred from a sum rule (Higgins et al., 2024). The important metrological distinction is between the center-of-gravity, or unsplit, frequency and the EFG-induced splittings. The former is the actual isomeric transition frequency; the latter encode the local crystal environment.

The host also generates an isomer shift, meaning a scalar shift of the unsplit nuclear frequency caused by the difference in nuclear charge distribution between the ground and isomeric states. In the modern formulation, the clock-frequency isomer shift depends on the electron density at the nucleus and must include self-consistent electronic relaxation, not only a first-order density-at-the-origin estimate (Perera et al., 26 Mar 2025). This is why host dependence is not a nuisance correction but a central design variable in solid-state nuclear clockwork.

3. Spectroscopic realization in $\mathrm{ThF_4}$9

Direct laser excitation of the $^{229}$0Th nuclear transition in a solid-state host was achieved in $^{229}$1 with a VUV frequency comb referenced to the JILA $^{229}$2Sr optical lattice clock. The measured unsplit transition frequency was

$^{229}$3

corresponding to the frequency ratio

$^{229}$4

for $^{229}$5Th embedded in $^{229}$6 at $^{229}$7 (Zhang et al., 2024). The same experiment resolved the nuclear quadrupole structure and identified six resonant fluorescence features, including the weak fifth line consistent with the quadrupole selection rules (Zhang et al., 2024).

The spectroscopy established both nuclear and host parameters. At $^{229}$8 K, the fitted asymmetry parameter was $^{229}$9, the quadrupole products were $5/2^+[633]$0 and $5/2^+[633]$1, and the extracted ratio of quadrupole moments was

$5/2^+[633]$2

Using $5/2^+[633]$3, the inferred EFG was $5/2^+[633]$4 (Zhang et al., 2024).

These advances did not eliminate solid-state broadening. The measured radiative lifetime in $5/2^+[633]$5 was reported as $5/2^+[633]$6 in the direct VUV-comb spectroscopy, implying an intrinsic linewidth on the order of $5/2^+[633]$7 Hz, while observed spectral widths in crystals were around $5/2^+[633]$8 kHz in earlier samples and around $5/2^+[633]$9 kHz in improved low-doping samples (Zhang et al., 2024, Girvin et al., 17 Nov 2025). This mismatch defines the central technical problem of the field: the nuclear transition is extraordinarily narrow in principle, but the usable solid-state resonance is presently set by disorder, defect-mediated EFG variation, and magnetic bath coupling rather than by radiative physics.

4. Temperature response, reproducibility, and systematic shifts

Temperature dependence is the best quantified systematic in the present $3/2^+[631]$0 platform. Measurements of the four strongest quadrupole-split lines at $3/2^+[631]$1 K, $3/2^+[631]$2 K, and $3/2^+[631]$3 K showed that the unsplit frequency decreases by $3/2^+[631]$4 kHz from $3/2^+[631]$5 K to $3/2^+[631]$6 K, while the EFG magnitude decreases by $3/2^+[631]$7 and the asymmetry parameter decreases by $3/2^+[631]$8 across the same interval. The fitted values were $3/2^+[631]$9 at $8.272(22)\,\mathrm{eV}$0 K, $8.272(22)\,\mathrm{eV}$1 at $8.272(22)\,\mathrm{eV}$2 K, and $8.272(22)\,\mathrm{eV}$3 at $8.272(22)\,\mathrm{eV}$4 K, with inferred EFGs of $8.272(22)\,\mathrm{eV}$5, $8.272(22)\,\mathrm{eV}$6, and $8.272(22)\,\mathrm{eV}$7, respectively (Higgins et al., 2024). The least temperature-sensitive line is $8.272(22)\,\mathrm{eV}$8, which shifts by only $8.272(22)\,\mathrm{eV}$9 kHz over $8.4$0 K, approximately $8.4$1 kHz/K, because the isomer shift and quadrupole shift partially cancel. For this line, achieving $8.4$2 fractional precision requires crystal temperature stability of about $8.4$3, corresponding to a frequency stability of about $8.4$4 mHz. The same analysis estimated other current systematics at smaller levels, including magnetic dipole broadening on the order of $8.4$5 Hz and second-order Doppler shifts around $8.4$6 Hz/K (Higgins et al., 2024).

Later work turned temperature from a dominant perturbation into an operating point. For the $8.4$7 line conventionally called line b, $8.4$8, the center frequency exhibits a minimum at

$8.4$9

where the first-order thermal sensitivity vanishes (Ooi et al., 1 Jul 2025). Around this point, line c, $2\,020\,407\,384.335(2)\,\mathrm{MHz}$0, serves as an in-situ thermometer because it changes by about $2\,020\,407\,384.335(2)\,\mathrm{MHz}$1 MHz over the explored temperature range, whereas line b changes by only about $2\,020\,407\,384.335(2)\,\mathrm{MHz}$2 kHz. Measuring line c to $2\,020\,407\,384.335(2)\,\mathrm{MHz}$3 kHz corresponds to about $2\,020\,407\,384.335(2)\,\mathrm{MHz}$4 K temperature uncertainty near $2\,020\,407\,384.335(2)\,\mathrm{MHz}$5, and $2\,020\,407\,384.335(2)\,\mathrm{MHz}$6 mK temperature reproducibility maps to about $2\,020\,407\,384.335(2)\,\mathrm{MHz}$7 mHz reproducibility on line b, or approximately $2\,020\,407\,384.335(2)\,\mathrm{MHz}$8 fractionally (Ooi et al., 1 Jul 2025).

Frequency reproducibility is now an experimental quantity rather than a projection. At $2\,020\,407\,384.335(2)\,\mathrm{MHz}$9 K, the weighted mean line-b frequency was reported as

$$\Delta E=\Delta E_{\rm strong}+\Delta E_{\rm EM} \approx 8.4~\mathrm{eV},$$0

with standard error $$\Delta E=\Delta E_{\rm strong}+\Delta E_{\rm EM} \approx 8.4~\mathrm{eV},$$1 kHz, corresponding to $$\Delta E=\Delta E_{\rm strong}+\Delta E_{\rm EM} \approx 8.4~\mathrm{eV},$$2 Hz or a fractional reproducibility of $$\Delta E=\Delta E_{\rm strong}+\Delta E_{\rm EM} \approx 8.4~\mathrm{eV},$$3 for two differently doped crystals over four months (Ooi et al., 1 Jul 2025). In an operating absorption-locked clock, the same b-line transition in two distinct crystals differed by only $$\Delta E=\Delta E_{\rm strong}+\Delta E_{\rm EM} \approx 8.4~\mathrm{eV},$$4 Hz, a fractional difference of $$\Delta E=\Delta E_{\rm strong}+\Delta E_{\rm EM} \approx 8.4~\mathrm{eV},$$5, showing that independently grown $$\Delta E=\Delta E_{\rm strong}+\Delta E_{\rm EM} \approx 8.4~\mathrm{eV},$$6 crystals can reproduce the nuclear reference at the $$\Delta E=\Delta E_{\rm strong}+\Delta E_{\rm EM} \approx 8.4~\mathrm{eV},$$7 level (Huang et al., 7 Jun 2026).

5. Readout strategies and clock operation

Solid-state nuclear clock design has evolved around the disparity between very long isomer lifetime and much shorter usable coherence time in a crystal. Early analyses of $$\Delta E=\Delta E_{\rm strong}+\Delta E_{\rm EM} \approx 8.4~\mathrm{eV},$$8 concluded that decoherence from magnetic coupling to surrounding nuclear moments rules out the commonly used Rabi or Ramsey interrogation schemes in the solid-state case and instead proposed stabilization based on counting fluorescence photons under continuous interrogation. Under optimized assumptions, that scheme projected a fractional instability level of $$\Delta E=\Delta E_{\rm strong}+\Delta E_{\rm EM} \approx 8.4~\mathrm{eV},$$9 within the solid-state approach (Kazakov et al., 2012).

The decisive shift from spectroscopy to clock operation came with continuous-wave VUV absorption locking in $\mathrm{CaF_2}$00. One implementation stabilized a continuous-wave $\mathrm{CaF_2}$01 nm VUV laser to a resolved nuclear transition in a solid-state host using frequency-modulation absorption spectroscopy and phototube photocurrent readout of transmitted VUV power. The VUV source, generated by resonantly enhanced four-wave mixing in cadmium vapor, produced about $\mathrm{CaF_2}$02 of $\mathrm{CaF_2}$03 nm light, with roughly $\mathrm{CaF_2}$04 delivered to the crystal. Using the weakly temperature-sensitive D-centre line b as the discriminator, the clock reached a fractional frequency instability of $\mathrm{CaF_2}$05, and the Allan deviation followed the expected $\mathrm{CaF_2}$06 scaling down to about $\mathrm{CaF_2}$07 (Huang et al., 7 Jun 2026).

A second implementation realized a stand-alone solid-state nuclear clock with active feedback based on continuous absorption spectroscopy in a millimeter-sized, room-temperature $\mathrm{CaF_2}$08 crystal. The feedback cycle could be as short as $\mathrm{CaF_2}$09 s, the measured linewidth was $\mathrm{CaF_2}$10, and the clock approached $\mathrm{CaF_2}$11 instability over $\mathrm{CaF_2}$12 day of continuous operation (Col et al., 3 Jun 2026). This work also used the clock output to search for periodic fluctuations and slow drifts in the nuclear transition energy on time scales between $\mathrm{CaF_2}$13 s and $\mathrm{CaF_2}$14 day, finding no significant periodic signals above the $\mathrm{CaF_2}$15 detection threshold and a drift slope of $\mathrm{CaF_2}$16, consistent with zero (Col et al., 3 Jun 2026).

Auxiliary control methods are becoming part of the clock architecture. Laser-induced quenching (LIQ) was demonstrated as a depumping method for the $\mathrm{CaF_2}$17Th isomer in $\mathrm{CaF_2}$18, achieving a threefold reduction in the isomer lifetime with $\mathrm{CaF_2}$19 mW of laser power and addressing the mismatch between a nuclear coherence time of about $\mathrm{CaF_2}$20 ms and a lifetime of $\mathrm{CaF_2}$21 s (Schaden et al., 2024). A distinct alternative is the internal-conversion-based solid-state clock concept, in which a thin $\mathrm{CaF_2}$22ThO$\mathrm{CaF_2}$23 layer is interrogated by a VUV comb and the nuclear excitation is detected through emitted electrons rather than radiative decay; in that scheme, a net scanning time of $\mathrm{CaF_2}$24 minutes over a $\mathrm{CaF_2}$25 eV uncertainty interval was estimated to be achievable, with projected clock performance comparable to a crystal-lattice nuclear clock but with a drastically simpler detection scheme (Wense et al., 2019).

6. Materials engineering, alternative hosts, and unresolved limits

Host selection has become a primary research frontier because host-dependent frequency offsets are calculable but non-negligible. A hybrid relativistic many-body plus periodic-DFT treatment predicted

$\mathrm{CaF_2}$26

$\mathrm{CaF_2}$27

and a bare-nucleus transition energy

$\mathrm{CaF_2}$28

with host-dependent valence-band isomer shifts constraining solid-state frequencies to an approximately $\mathrm{CaF_2}$29 MHz-wide window across candidate materials (Perera et al., 26 Mar 2025). This suggests that moving between hosts is a tens-of-MHz problem rather than a GHz problem, but it also means that transferability requires explicit control of local chemistry and defect occupancy.

Several alternative materials platforms target specific limitations of $\mathrm{CaF_2}$30. A spinless stoichiometric sulfate host, $\mathrm{CaF_2}$31, was proposed to remove fluorine nuclear-spin broadening while retaining a band gap of $\mathrm{CaF_2}$32 eV; the same study concluded that introducing $\mathrm{CaF_2}$33Th does not modify the material band gap nor introduce electronic states associated with nuclear quenching, and projected a clock instability of $\mathrm{CaF_2}$34 (Morgan et al., 14 Mar 2025). Thin-film $\mathrm{CaF_2}$35 targets grown by physical vapor deposition consume only micrograms of $\mathrm{CaF_2}$36Th, reduce radioactivity to about $\mathrm{CaF_2}$37–$\mathrm{CaF_2}$38 Bq for a $\mathrm{CaF_2}$39m target, and have already shown laser excitation of the nuclear transition; for an ideal defect-free crystalline $\mathrm{CaF_2}$40 clock, the projected instability was $\mathrm{CaF_2}$41 (Zhang et al., 2024). $\mathrm{CaF_2}$42 has been proposed as a host with unusually favorable growth properties, including segregation coefficient $\mathrm{CaF_2}$43, thorium concentration $\mathrm{CaF_2}$44, and $\mathrm{CaF_2}$45 transmittance at $\mathrm{CaF_2}$46–$\mathrm{CaF_2}$47 nm in a $\mathrm{CaF_2}$48 mm sample (Gong et al., 2024). Compact architectures based on $\mathrm{CaF_2}$49Th-doped nonlinear optical crystals such as $\mathrm{CaF_2}$50 or $\mathrm{CaF_2}$51 seek to eliminate external VUV beamlines by generating $\mathrm{CaF_2}$52 nm light inside the device, with projected fractional instability of $\mathrm{CaF_2}$53 for $\mathrm{CaF_2}$54-poling in the $\mathrm{CaF_2}$55 concept (Morgan et al., 2024). Nanophotonic fluoride whispering-gallery resonators extend this logic to on-chip field enhancement; implantation of $\mathrm{CaF_2}$56Th into a crystalline $\mathrm{CaF_2}$57 resonator was demonstrated, though implantation-induced damage presently degrades $\mathrm{CaF_2}$58 and defines a major engineering bottleneck (Kraemer et al., 22 Apr 2026).

The major unresolved issue is not the existence of a sufficiently narrow nuclear transition in principle, but whether a solid host can be made homogeneous enough that the narrow nuclear line can be recovered experimentally. Reviews of the field identify inhomogeneous broadening from strains, defect-induced EFG variation, charge compensation, and magnetic dipole interactions as the dominant present limitation, with observed widths in the $\mathrm{CaF_2}$59–$\mathrm{CaF_2}$60 kHz class still roughly eight orders of magnitude broader than the intrinsic $\mathrm{CaF_2}$61 Hz limit in $\mathrm{CaF_2}$62 (Girvin et al., 17 Nov 2025). At the same time, the same sensitivity that complicates clock operation is what makes the platform interesting for fundamental physics. Solid-state $\mathrm{CaF_2}$63Th clocks have already been used to constrain ultralight dark matter couplings (Col et al., 3 Jun 2026), and nuclear-clock “quintessometer” analyses argue that solid-state implementations can surpass existing limits on scalar-field couplings at submicron distances and improve equivalence-principle tests at kilometer scales and beyond (Delaunay et al., 4 Mar 2025). The field therefore remains dual-use in the strongest sense: the same host-controlled solid-state platform is being developed simultaneously as a frequency reference, a materials-sensitive nuclear spectrometer, and a sensor for new physics.