Solid-State Scandium Broadens Beyond Nuclear-Clock Requirements

The 12.389-keV isomeric transition in $^{45}$Sc is attractive for x-ray nuclear-clock metrology because its 0.46-s lifetime implies a natural linewidth of about 1.4 feV and a quality factor near $10^{19}$. The obstacle is not finding the transition, but preserving its exceptionally narrow coherence in a solid. Liu et al. use resonant pumping at the European XFEL to separate a persistent isomer population from the coherent nuclear forward-scattering (NFS) signal that a clock-relevant solid target would require.

Core Contribution

The paper establishes two different facts that are easy to conflate. First, the long-lived $^{45}$Sc isomer can be excited in solid targets: delayed scandium $K_{\alpha,\beta}$ fluorescence persists after the x-ray pulse. Second, the experiment finds no statistically significant delayed coherent forward-scattering signal after 2 ms in the tested crystals. Under the authors' inhomogeneous-linewidth model, that null result implies an effective linewidth broadening of at least $500\,\Gamma_0$.

That is a useful constraint rather than a clock demonstration. It turns an uncertain materials question into a quantitative target: future solid-state implementations must reduce broadening by substantially more than the experiment’s inferred lower bound.

Technical Approach

The setup sequentially irradiates a movable metallic Sc foil and cryogenic crystal targets with self-seeded XFEL macropulses. The foil provides resonance detection through delayed incoherent fluorescence, while the NFS arm measures photons in the forward direction from crystals cooled to 20 K. Figure 1 lays out the separation between these channels: detectors measure both delayed 12.4-keV photons and lower-energy Sc fluorescence as functions of photon energy and delay after excitation.

The beam delivered pulse trains of $n_p=400$ pulses spaced by 440 ns, with a total train duration of 0.18 ms at 10 Hz. Since that duration is far shorter than the 0.46-s isomer lifetime, the analysis treats each train as a single excitation event. For the fluorescence decay, the authors varied the fitting interval and binning, fitting an exponential rate across many choices to account for low-count statistical sensitivity and detector differences. They also explicitly model residual excitation from earlier macropulses; it changes the accumulated population but preserves the exponential decay constant.

A separate calibration estimates the partial K-shell internal-conversion coefficient from fluorescence and 12.4-keV elastic rates, correcting for background, fluorescence yield, attenuation, and detector response. The reported value, $\alpha_K=390(60)$, agrees within uncertainty with the cited theoretical value of 363.

Results and Analysis

Four crystal chemistries were considered: Sc, ScN, $\mathrm{Sc_2O_3}$, and $\mathrm{ScAlMgO_4}$. The selection criteria reflect the real experimental constraints: crystallinity and chemical purity, low electric-field-gradient and magnetic-hyperfine effects, low non-Sc absorption, an optical thickness near the predicted NFS optimum, and thermal conductivity sufficient to manage heating. No material met all criteria. ScN was nominally the preferred crystal but was nitrogen-deficient; $\mathrm{ScAlMgO_4}$ had the highest crystalline quality but low thermal conductivity.

The direct signal is mixed. Delayed incoherent fluorescence confirms resonant isomer excitation, and prompt 12.4-keV forward photons are detected, but neither establishes long-lived collective coherence. Figure 3 separates prompt and delayed energy windows in the forward geometry; the delayed component shows no resolvable NFS peak beyond 2 ms for the tested compounds. Simulations indicate that coherent NFS itself accelerates the apparent decay, while increasing inhomogeneous broadening both shortens the response and lowers its time-integrated strength. The inaccessible interval below 2 ms is therefore consequential: the experiment constrains a delayed tail, not the full early-time response.

The authors report that shutter timing was independently verified with an opening duration below 1 ms and jitter below about 1 ms, so a coherent signal persisting beyond 2 ms should have been observable. They identify practical limits instead: the intended 15-meV-bandwidth diamond channel-cut monochromators did not function properly, leaving a broad direct XFEL beam that likely increased radiation damage. The result supports the broadening bound under the adopted model, but does not identify a microscopic broadening mechanism or show that every promising host material fails.

Implications

For nuclear-clock researchers, the paper supplies a calibrated experimental baseline: solid $^{45}$Sc retains long-lived nuclear excitation, yet the tested hosts do not retain sufficiently visible collective coherence on the accessible millisecond scale. Narrower excitation bandwidth, lower target dose, and materials with reduced local-field disorder are the next experimental variables, rather than minor refinements to the present NFS readout.