WorksPhysics and Astronomy
Nuclear Charge Radius of Be from Muonic Atom Spectroscopy Using a Microcalorimeter
Abstract
The transition energy in muonic Be was measured using a metallic magnetic calorimeter, resulting in eV. The result is 30 times more precise than the previous best measurement and enables the extraction of the corresponding nuclear charge radius Befm. It is times more precise than the commonly used value based on electron scattering and differs from it by times the combined uncertainties. This measurement represents the first determination of a nuclear charge radius using muonic x-ray spectroscopy with microcalorimeters.
Deployment of MMC-based detector at PSI. To achieve the required accuracy, we rely on state-of-the-art single-photon quantum sensing technology. Specifically, we use a maXs-30 (Micro-calorimeter Array for X-ray Spectroscopy) detector optimized for measuring energies around 30 keV [20, 21, 22], developed at the Kirchhoff Institute for Physics at Heidelberg University. Under accelerator beamline conditions, the Full-Width at Half-Maximum (FWHM) resolution achieved is approximately 20 eV at 33 keV, which is 15 times better than that of the highest-resolution solid-state detectors [23, 24].
The detector has an array of pixels coupled in pairs to a two-stage SQUID readout. Its active area is mm. Each pixel is equipped with a 20 m-thick gold absorber that efficiently stops photons up to 120 keV (e.g., 95% efficiency at 20 keV and 10% at 120 keV) [25]. The MMC array was mounted inside a custom sidearm of a dilution refrigerator, facing a target chamber, with 13 cm between the target center and the MMC absorber plane. This resulted in a solid angle of . Given the quantum efficiency of nearly 60% at 33 keV [25], and considering that approximately two-thirds of the pixels were fully functioning, the overall efficiency amounted to at 33 keV.
The measurement was carried out over a period of 34 hours at the E1 beamline of the PSI accelerator facility, which delivers a continuous beam of negative muons of adjustable momentum [26]. The beam passed through a collimation system comprising a scintillator with a central aperture and subsequently entered the evacuated target chamber via a thin plastic scintillator, which served as the timing reference for the muonic X-ray cascade [27]. The process from muon trigger to X-ray emission occurs within a few ns, which is substantially shorter than the time resolution of our detection system. Inside the chamber, we placed a high-purity () beryllium disk, 2.5 mm thick and 21.6 mm in diameter, made from naturally occurring beryllium, which consists entirely of Be. To reduce self-absorption, it was mounted at a 45 degree angle relative to both the incident beam axis and the cryostat window. At a beam momentum of 29.5 MeV/,
the estimated muon stopping rate was 5 kHz, based on a GEANT4 simulation, leading to a detected photon rate of 0.13 Hz.
For each event in an MMC pixel, a 2.6 ms waveform from the analog output of the SQUID electronics was digitized using a 16-bit digitizer [28]. The voltage nonlinearity of each digitizer channel was measured separately and used to correct all recorded waveforms. Each pulse was fitted using amplitude scaling and time shifting of a template derived from 60 keV photon pulses from an Am source that was continuously measured. We also placed a target made of mixed barium and lanthanum foils, sealed in epoxy, between the cryostat x-ray window and the target chamber. This target was irradiated with an x-ray tube to generate x-ray fluorescence (XRF) photons, which were collected simultaneously with the muonic x rays. The reference lines and their measured rates are summarized in Table I.
Table I. Main reference lines discussed in the text, ordered by descending energy.
From pulse amplitudes to summed spectrum. Data reduction is applied to filter out triggered noise, deformed traces (e.g., pile-up events), and coincidence events, which can be attributed to Michel electron interactions in the detector substrate. The cuts involved in this process are described in the supplemental material. Despite the highly linear response of MMCs, even small temperature variations (here, a few mK) induce changes in signal amplitude that must be compensated for to achieve high energy resolution. This correction is implemented by measuring baseline voltage drifts in dedicated temperature-sensitive channels located at the four corners of the detector [25]. The correction is performed separately for each pixel by evaluating the dependence of amplitude on temperature for 60 keV photons from an Am source and fitting with a robust linear model (using the statsmodels package [32]). The fitted model was used to project the entire measured spectrum to a constant temperature. An example of this process for one of the pixels is shown in Figure 1. After correcting for the temperature dependence of the amplitude, the individual pixels and the sum of their spectra exhibit a resolving power of approximately 1700 in the region of interest, which, after gain calibration (see below), translates to 20 eV FWHM. The subsequent step in the analysis involves distinguishing muon-induced events from all other events. This is crucial because our primary calibration line, La: , is located only about 50 eV away from the transition line of interest. Partial spectral overlap can introduce lineshape distortions that hinder an unambiguous determination of the respective line centers.
Figure 1. Temperature correction for a single pixel. The top graph shows the correlation between the amplitude from pulses stemming from 60 keV photons, and the baseline voltage of a temperature-sensitive channel (gray dots), along with a robust linear regression fit (red line). The bottom graph shows how applying this linear correction drastically improves the resolution.
To minimize bleeding of the calibration lines to the muonic x-ray spectrum, two separate spectra are constructed: one in coincidence with an incoming muon and one in anti-coincidence with it. For each pixel, the optimal time windows relative to the muon trigger were established using muonic Be transitions that were well separated from the calibration lines, namely the manifold near 6 keV and the transition around 39 keV. The chosen coincidence window is set to 750 ns around the most probable timing value. It reduces the calibration peak magnitude in the coincident spectrum by a factor of 125 (See Figure 2), without meaningfully affecting the muon-induced lines. The remaining bleeding of the calibration signal into the muon-induced signal, and vice versa, is addressed by a coupled fitting procedure discussed below.
Figure 2. Fitted histograms of the combined spectra from 35 pixels in the area of interest. The top figure shows x rays in time coincidence with an incoming muon. The bottom figure shows the anti-coincident spectrum featuring calibration lines.
Although MMCs exhibit a highly linear thermodynamic response, achieving sub-eV precision necessitates a pixel-by-pixel determination of the residual nonlinearity [33]. Here, the residual non-linearity of each pixel is quantified by fitting several calibration lines (given in Table I) located on either side of the region of interest and determining their most-probable amplitudes . These amplitudes are then fitted according to
where are the literature values of the energies given in Table I, is a linear gain parameter, and is the pixel nonlinearity parameter. The fit returns between and / keV, depending on the pixel, with a relative uncertainty on the order of 1%. The median nonlinearity around 33 keV is 0.6%, which introduces a correction of 0.3 eV to the approximately 50 eV distance between the centroid energy and the La: calibration line. Thus, the nonlinearity correction is on the same order as our statistical uncertainty and therefore cannot be ignored. Nonetheless, it is small enough that the uncertainty associated with this correction can be neglected. Eq. (1) is then used to convert pulse amplitudes to energies, and the data are summed to produce the high-statistics spectrum in Figure 2.
Final fit and energy extraction
The energy difference between the centroid energy and the La: calibration line was derived by simultaneously fitting the coincidence and anti-coincidence spectra shown in Figure 2. The instrumental lineshape is, in the first approximation, Gaussian. However, for high precision experiments, small deviations, such as low energy tails, need to be taken into account to determine the line centroid. Depending on the origin of the tail, an energy-dependent lineshape might be observed. Here, we apply a generic profile composed of a Voigt function with exponential tails. A detailed description of the lineshape is provided in the End Matter.
Both parts of the separated spectrum were fitted simultaneously using maximum likelihood estimation based on Poisson statistics (via the SATLAS-2 package [34]). Two additional free parameters were introduced to account for the residual cross-contamination between the spectra. The background is modeled as constant throughout the fitting domain. Muonic peaks exhibit characteristic step-like behavior at low energies, with the magnitude treated as an adjustable parameter in the fitting procedure. We attribute this feature to photons undergoing Compton scattering between the point of muonic-atom formation and the detector. Such effects are considerably smaller for the calibration lines because they traverse significantly less material.
The fitting model was initialized with 50 free parameters and progressively constrained based on the underlying physics, as detailed in the End Matter. The variation of the final result induced by empirical line-shape modeling was found to be on the order of 50 meV; substantially smaller than the statistical uncertainty. The final fit has 21 free parameters, and its results are shown in Figure 2.
The main outcome of the fit is the difference between the centroids of the lanthanum and muonic lines, which is
The first uncertainty in Eq. (2) is statistical, and the second is systematic, evaluated from the difference between the results when allowing more free parameters in the fit (see End matter for more details).
The centroid photon energy is determined by subtracting the measured energy difference from the tabulated calibration line energy (Table I). To convert this photon energy into the corresponding transition energy, we subtract the 66 meV taken by the recoiling atom, as determined from momentum conservation [35]. The final result is therefore:
Its uncertainty is primarily limited by the precision of the reference calibration line used and secondarily by the statistical precision of the experiment.
Radius extraction and implications Equating our experimental result, Eq. (3), with the theoretical parameterization detailed in the End Matter, Eq. (16), and solving for yields
Our result from Eq. (4) is 2.4 times more precise and larger by 2.3 combined standard errors than the accepted value from electron scattering, 2.519(12) fm [13]. Its total variance primarily stems from the literature value of the characteristic x ray of lanthanum [30] (51%), the fit precision of our experiment (32%), and the uncertainty in nuclear theory for the energy to radius parametrization (17%). See End Matter for more details.
The radii of radioactive Be isotopes are measured using laser spectroscopy relative to the reference value [36]
| Isotope | (fm) | Literature | NLEFT | GFMC |
| Be | 2.677(12) | 2.646(16) [36] | 2.581(17) | |
| Be | 2.5506(52) | 2.519(12) [13] | 2.554(15) | 2.53(5) |
| Be | 2.395(12) | 2.361(14) [36] | 2.512(36) | |
| Be | 2.499(10) | 2.466(15) [36] | 2.544(40) | |
| Be | 2.535(11) | 2.502(12) [36] | 2.580(33) | |
| C | 2.671(14) | 2.638(36) [9] | ||
| B | 2.562(13) | 2.531(38) [9] |
caption: TABLE II. Updated radii derived from this work. is derived from the mirror fit [9], and , which refers to the excited state, is interpolated assuming negligible isospin symmetry breaking [9, 37]. They are presented in italics to highlight their dependence on assumptions. NLEFT stands for nuclear lattice effective field theory [38, 39, 40]. GFMC stands for Green’s function Monte Carlo [41].
:::
and must be shifted by about 0.03 fm, as listed in Table II. This systematic shift is about 2–3 times larger than the quoted experimental uncertainties. These radii are now better aligned with recent nuclear lattice effective field theory calculations [40], except for Be, whose radius deviates by 4.5 combined standard errors. Both the literature value and our updated radius are consistent with Green’s function Monte-Carlo calculations [41] within uncertainty.
Assuming that the mirror shift relation (detailed in Ref. [9]) holds for light nuclei and high asymmetries, this provides an improved prediction: fm. Further assuming negligible isospin symmetry breaking, the predicted radius of the excited state of B is fm. These radii are important for benchmarking ab initio calculations in superallowed beta decay [42, 43] and for estimating the magnitude and uncertainty of finite-size contributions to them [3, 44, 45].
Summary and outlook We have demonstrated the use of a detector system based on a metallic magnetic calorimeter in a negative muon beamline and performed high resolution x-ray spectroscopy of muonic atoms. The transition energy centroid in muonic Beryllium was determined to be 33 391.48(35) eV. From this energy and dedicated calculations spanning atomic and nuclear theory, the nuclear charge radius of Be was determined to be 2.5506(52) fm. Our result is primarily limited by the literature accuracy of the calibration line used and highlights the need for improved determinations of reference line energies [29].
Combined with differential radii from optical spectroscopy [36], we derived improved radii for the entire chain. They are in slightly better agreement with recent ab initio calculations, except for Be, which shows significant deviation. This radius is particularly interesting in the context of ongoing work to measure the radius of B and to derive the size of the proton halo from the difference between the two isotones [46]. The radii of the entire chain are now limited by the precision of measured isotope shifts, highlighting the need for new optical measurements.
In this study, we measured a line center to one percent of its linewidth, yielding an accuracy of 10 ppm. Considering that the resolution could be improved by up to a factor of 5 with an optimized detector [47], and that x-ray line-centers have been determined to better than 500 times their linewidths [48, 49, 50], measurements with sub ppm accuracy are within reach.
Acknowledgments The experiments were performed at the E1 beamline of PSI. We thank the accelerator and support groups for their excellent conditions. This research was funded in whole or in part by the Swiss National Science Foundation (SNSF) 200020_215164. The Technion group was supported by the Hellen Diller Quantum Center. S. Rathi acknowledges the Technion and the Lady Davis Postdoctoral fellowships. M. Deseyn was supported by a personal FWO Aspirant Fundamental Research (no. 11P6V24N). KU Leuven BOF under contract number C14/22/104; the European Research Council (ERC) through proposal No. 101088504 (NSHAPE). The Mainz groups were supported by the Deutsche Forschungsgemeinschaft (DFG) through the Cluster of Excellence “Precision Physics, Fundamental Interactions, and Structure of Matter” PRISMA++ EXC 2118/1 (Project ID 390831469), CRC1660 "Hadrons and Nuclei as Discovery Tools" (DFG project number 514321794) and the German State of Rhineland Palatinate. The work of U.-G. Meißner was supported in part by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (EXOTIC, grant agreement No. 101018170), the CAS President’s International Fellowship Initiative PIFI (Grant No. 2025PD0022), Gauss Centre for Supercomputing e.V. (www.gauss-centre.eu) for computing time on the GCS Supercomputer JUWELS at Jülich Supercomputing Centre (JSC) and on HoreKa by the National High-Performance Computing Center at KIT (NHR@KIT). The work of S. Shen was supported by the Natural Science Foundation of China under Grant No. 12435007 and U2541242. K. von Schoeler gratefully acknowledges the support of ETH Research Grant 22-2 ETH-023, Switzerland. J. Machado and C. Godinho acknowledge financial support from Fundação para a Ciência e Tecnologia (FCT, Portugal), under research center Grants No. UIDB/04559/2025 (LIBPhys) and No. LA/P/0117/2020 (Associated Laboratory LA-REAL), and Contract No. 2022.11197.BD, respectively. J. Walch was supported by "Quantum Sensing for Fundamental Physics (Q4SPhysics)", funded by the Innovation pool of the research field Helmholtz Matter of the Helmholtz Association. The group at the Kirchhoff Institute for Physics is supported by the Field of Focus II Initiative at Heidelberg University funded by the “Bundesministerium für Bildung und Forschung” (BMBF) and by the “Wissenschaftsministerium Baden-Württemberg” in the framework of the “Exzellenzstrategie von Bund und Ländern”. DU acknowledges the support of the Research Training Group HighRR (GRK 2058) funded by the Deutsche Forschungsgemeinschaft (DFG).
References
End Matter
Lineshape model A model for the response function is given in Eq. (5), where the signal amplitude is distributed into three components. The first component is a four-parameter Voigt core defined by the fractional intensity , centroid , Gaussian FWHM , and Lorentzian FWHM . The remaining terms are two five-parameter tail functions located to the right and left, with fractional intensities and tail length scales . All components are summed with a constant background term .
An additional step function, , is included in the fit of the muonic peak to account for photon scattering on their way out of the target.
The muonic x-ray lines have a natural width of about 20 meV, which is negligible compared to our resolution. Therefore, for both the muonic x rays and the calibration gamma rays is taken to be zero. for the XRF sources is fixed to the known natural line width [51].
The fine and hyperfine structure splittings are modeled by convolving Eq. (5) with a sum of delta functions whose relative energies are calculated using the MDFGME code [52, 53] (see Table III). It was confirmed that reasonable variations in the calculations do not significantly affect the extracted centroid energies.
| Relative Energy [eV] | Intensity (%) | |
| 2.356 | 29.17 | |
| 1.942 | 11.50 | |
| 11.50 | ||
| 0.012 | 11.47 | |
| 11.46 | ||
| 9.34 | ||
| 9.33 | ||
| 4.17 | ||
| 2.447 | 1.03 | |
| 1.03 |
Table III. Calculated energies and intensities of hyperfine lines of : peak (sorted by descending intensity). Energies are written relative to the center-of-gravity of the : peak.
To assess the uncertainty in our empirical lineshape model, we compare the computed energy difference of interest across various models with those that impose stronger constraints. The results are summarized in table IV.
Parametrization of energy with radius Ref. [54] reports
where the last term is a placeholder for corrections that depend on nuclear structure, namely
that we refer to as shape correction, nucleon polarization, and nuclear polarization, respectively.
Ref. [54] assumed a Gaussian nuclear charge distribution; therefore, we include a correction for a more realistic charge distribution computed ab initio using nuclear lattice effective field theory (NLEFT). It uses the pinhole algorithm, which inserts a sheet with specific spin-isospin labels at mid-times, allowing for the tracking of protons and neutrons along with their correlations [55]. Integrating the resulting distribution yields a charge radius of and a Friar radius of , where the first uncertainty is numerical and the second is estimated by recalculating with a less advanced interaction (SU(4) instead of N3LO) and reporting half the difference. Both moments agree within uncertainty with recent Green’s function Monte-Carlo calculations [41]. Their ratio is given by , and it is more stable than the individual moments. Our result is 0.3(2)% smaller than that of a Gaussian distribution . Inserting this ratio into Eq. (7) of Ref. [54] gives the shape correction
whose overall size is significant, while its sensitivity to the radius is small for any realistic variation around the reference value. Moreover, due to the similarity of the calculated shape to that of a Gaussian, the shape correction to the elastic three-photon exchange is negligible.
The nucleon polarization is calculated using the formalism of [56], resulting in
Both the central value and its uncertainty are nearly significant. Nuclear polarization can be assessed using an expansion
The first two terms in Eq. (10) are the leading and next-to-leading terms of the expansion of inelastic two-photon exchange [56]. The third term is an uncertainty evaluated with the small parameter [56, 57]. The last term roughly accounts for the inelastic three-photon exchange [58].
The leading order term, , primarily depends on the electric dipole strength function, which can be predicted using a novel artificial neural network approach [59]. For , the prominent low-energy dipole strength-structure below 5 MeV is treated separately using a dedicated fit to the experimental strength data [60, 61, 62] in this region, while the artificial neural network prediction is retained at higher energies. Integrating the electric dipole strength function for using the appropriate weight factors [56] returns
with the reported uncertainty being statistical and nucleus-dependent [59]. The magnetic contribution is negligible.
The next-to-leading term in Eq. (10), , is calculated using Eqs. (5b), (5c), (5k) and (5l) from Ref. [56]. The 1 and 2-body densities that enter these equations are taken from Ref. [63]. The process is validated against Ref. [56] for helium isotopes and applied to muonic for densities calculated with various interaction Hamiltonians. More details will be provided in a separate publication.
Considering six Hamiltonians: AV18+UX, NV2+3-IIa*, NV2+3-IIb, NV2+3-Ia, NV2+3-Ia*, and NV2+3-Ib, the values of range from meV to 23 meV. The full range
is taken as a conservative uncertainty estimate.
We follow Ref. [8] in estimating the inelastic three photon exchange, , as half of the elastic part, given by the sum of Eqs. (8) and (9) of Ref. [54], with an opposite sign and 100% uncertainty
Plugging Eqs. (11), (12) and (13) to Eq. (10) results in
Summing Eq. (8), (9), and (14) results in
Its uncertainty primarily arises from , followed by , and subsequently from stemming from the artificial neural network. Compared to these, the uncertainties from , and the pure QED uncertainty are minor. The complete parameterization is given by the sum of Eq. (6) and (15)
| Parameter | Peak | Unconstrained [50 Free] | Relaxed [36 Free] | Intermediate [24 Free] | Constrained [21 Free] |
| Test statistic | 1.132 | 1.132 | 1.140 | 1.169 | |
| Global | (eV) | 50.81(57) | 50.63(48) | 50.76(24) | 50.71(21) |
| Scale | 1.00001(1) | 1.00001(1) | 1.000015(7) | 1.000018(5) | |
| Amplitudes (AC / C) | La: | 1393(27) / 6(5) | 1396(17) / 12(2) | 1405(16) / 9(1) | 1394(16) / 12(1) |
| La: | 772(21) / 7(1) | 768(10) / Linked | 782(9) / Linked | 780(9) / Linked | |
| Ba: | 2140(40) / 17(2) | 2133(27) / 18(1) | 2099(21) / 18(1) | 2091(22) / Linked | |
| Ba: | 1140(24) / 10(1) | 1151(15) / Linked | 1156(12) / Linked | 1150(12) / Linked | |
| Am | 19(3) / — | 18(3) / — | 18(3) / — | 18(3) / — | |
| Be | < 19 / 848(20) | 5(1) / 834(18) | 17(7) / 791(14) | 17(7) / 786(14) | |
| O | — / 5(1) | — / 5(1) | — / 5(1) | — / 4(1) | |
| O | — / 5(1) | — / 5(1) | — / 5(1) | — / 4(1) | |
| Centroids (eV) | Ba: | 31817.4(5) | 31816.7(4) | 31817.1(2) | Linked |
| Ba: | 32193.4(5) | 32193.5(5) | 32193.8(2) | Linked | |
| La: | 33034.7(8) | 33035.0(4) | Linked | Linked | |
| Am | 33197(2) | 33197(2) | Linked | Linked | |
| O | 33587(3) | 33587(3) | 33588(2) | 33588(2) | |
| O | 33625(3) | 33625(3) | Linked | Linked | |
| (eV) | All Peaks | Fixed | Fixed | Fixed | Fixed |
| (eV) | La: | 22(1) | 22.2(4) | 22.4(3) | 23.0(3) |
| La: | 22(1) | Linked | Linked | Linked | |
| Ba: | 22(1) | Linked | Linked | Linked | |
| Ba: | 23(1) | Linked | Linked | Linked | |
| Be | 22(1) | Linked | Linked | Linked | |
| La: | 0.86(3) | 0.84(2) | 0.86(1) | 0.89(1) | |
| La: | 0.81(6) | Linked | Linked | Linked | |
| Ba: | 0.80(4) | 0.83(3) | Linked | Linked | |
| Ba: | 0.91(3) | Linked | Linked | Linked | |
| Be | 0.88(2) | 0.88(2) | Linked | Linked | |
| (eV) | La: | 28(4) | 26(3) | 23(1) | 25(2) |
| La: | 24(4) | Linked | Linked | Linked | |
| Ba: | 20(2) | 21(2) | Linked | Linked | |
| Ba: | 30(9) | Linked | Linked | Linked | |
| Be | 22(3) | 19(3) | Linked | Linked | |
| La: | 1.7(3) | 1.8(3) | 2.1(1) | 2.1(2) | |
| La: | 1.9(4) | Linked | Linked | Linked | |
| Ba: | 2.4(3) | 2.4(2) | Linked | Linked | |
| Ba: | 1.8(6) | Linked | Linked | Linked | |
| Be | 2.5(4) | 2.9(5) | Linked | Linked | |
| La: | 1.1(5) | 1.3(4) | 1.5(2) | 1.2(2) | |
| La: | 1.6(8) | Linked | Linked | Linked | |
| Ba: | 2.1(7) | 2.0(5) | Linked | Linked | |
| Ba: | 0.9(6) | Linked | Linked | Linked | |
| Be | 1.9(5) | 2.0(6) | Linked | Linked | |
| (AC / C) | Be | 9.2(9) | 9.3(9) | 10.1(8) | 10.1(8) |
| Background (AC / C) | 4.2(2) / 0.50(5) | 4.2(2) / 0.49(5) | 4.2(1) / 0.48(5) | 4.2(2) / 0.48(5) |
Table IV. Complete evolution of the fit parameters across hierarchical fitting stages. Values in parentheses represent the statistical uncertainty in the last digits. "Linked" indicates the parameter was mathematically constrained to a master variable (typically the La transition). "Fixed" indicates the parameter was locked to a literature value [51] or zero. Amplitudes and step fractional intensities are denoted in both anticoincident (AC) and coincident (C) spectra, separated by ‘/’.
Supplementary Material
Analysis cuts
The data selection criteria are detailed in Refs. [14, 64]. To satisfy these criteria, we set specific parameters for each of the 35 instrumented pixels in this experiment. The implementation of the selection procedure is as follows:
Select an upper threshold that lies above the vertical dense lines corresponding to photonic peaks (see the example in Figure S1, top left).

Figure S1. Examples of cuts described in the main text.
Exclude the initial and final segments of the run that experienced instabilities. Within the main segment, a global outlier rejection for temperature-values was applied to exclude non-physical readout errors. It was complemented by also removing deviations greater than from the local moving average as outliers. See (Figure S1, bottom left).
Omit data from pixels exhibiting unstable baselines or irregular pulse shapes.
To remove events from decay electrons that heat the entire array, apply a hold-off time of 100 ms after each cluster event (see Figure S1, top right).
To prevent pile-up effects, apply a hold-off time of 540 ms after a photon hits the same pixel (Figure S1 (Bottom-Right)).
The resulting rates are summarized in Table S1.
| # | Cut | Counts | Rate (Hz) | ||
| La | La | ||||
| 1 | cut | 18 166 | 46 265 | 0.13 | 0.28 |
| 2 | Irregular Events | 15 548 | 35 779 | 0.13 | 0.28 |
| 3 | Unstable Pixels | 15 138 | 34 948 | 0.12 | 0.28 |
| 4 | Cluster Pileup | 12 818 | 29 888 | 0.10 | 0.24 |
| 5 | Pixel Pileup | 11 741 | 27 315 | 0.10 | 0.22 |
Table S1. Summary of sequential data cuts, remaining counts, and rates summed across all active pixels of the MMC.
Peak identification
A systematic peak-hunting procedure was performed to resolve the complex, high-resolution measured spectrum. The exceptional energy resolution of the MMC enables a clear distinction between the muonic Be transitions of interest, the secondary muonic background from the target holder, and the various X-ray fluorescence and radioactive source lines used for energy scale anchoring. Table S2 and Figure S2 provide an assignment of these features.
| Assignment | Comments | Energy (keV) |
| Beryllium (Target) | ||
| Peak of Interest | 33.4 | |
| Escape | 23.6 | |
| Escape | 22.0 | |
| Trigger alignment | 39.6 | |
| 41.8, 42.9 | ||
| Trigger alignment | 6.1 | |
| 8.3 | ||
| Oxygen, Nitrogen, Carbon (Target holder) | ||
| 25.0 | ||
| 19.0 | ||
| 75.0 | ||
| 14.0, 18.8 | ||
| Copper (Chamber) | ||
| 8.0, 8.9 | ||
| Niobium (X-Ray Shield) | ||
| 16.5, 16.6 | ||
| Americium-241 (Source) | ||
| Calibration | 59.5 | |
| Escape | 49.8, 48.0 | |
| 43.4, 33.2 | ||
| Calibration | 26.3 | |
| series | 13.8–21.0 | |
| Lanthanum (XRF Target) | ||
| Calibration | 33.4, 33.0 | |
| series | 37.7–39.0 | |
| 21.7, 23.4, 23.8 | ||
| Barium (XRF Target) | ||
| Calibration | 32.1, 31.8 | |
| series | 36.3–37.5 | |
| Escape | 20.3, 22.0, 22.3 | |
| series | 4.5, 4.8, 5.1 | |
| Iron-55 (Source) | ||
| 5.9, 6.5 | ||
| Copper (Chamber) | ||
| 8.0, 8.9 | ||
| Niobium (X-Ray Shield) | ||
| 16.5, 16.6 | ||
| Silver (Detector) | ||
| 24.9 |
Table S2. List of identified peaks in the broad spectrum shown in Figure S2.

Figure S2. Coincidence and anti-coincidence spectra are shown in red and blue, respectively. The peaks are identified in Table S2.
References
- [3]M. Gorchtein, V. Katyal, B. Ohayon, B. K. Sahoo, and C.-Y. Seng, Cabibbo-Kobayashi-Maskawa unitarity deficit reduction via finite nuclear size, Phys. Rev. Res. 7, L024002 (2025).DOI
- [8]K. Pacchukchi, V. Lensky, F. Hagelstein, S. S. Li Muli, S. Bacca, and R. Pohl, Comprehensive theory of the Lamb shift in light muonic atoms, Rev. Mod. Phys. 96, 015001 (2024).
- [9]B. Ohayon, Critical evaluation of reference charge radii and applications in mirror nuclei, Atomic Data and Nuclear Data Tables 165, 101732 (2025).DOI
- [13]J. Jansen, R. T. Peerdeman, and C. De Vries, Nuclear charge radii of ¹²C and ⁹Be, Nuclear Physics A 188, 337 (1972).
- [14]A. Fleischmann, C. Enss, and G. M. Seidel, Metallic magnetic calorimeters, in Cryogenic Particle Detection (Springer Berlin Heidelberg, Berlin, Heidelberg, 2005) pp. 151–216.
- [20]D. Hengstler, M. Keller, C. Schötz, J. Geist, M. Krantz, S. Kempf, L. Gastaldo, A. Fleischmann, T. Gassner, G. Weber, R. Märtin, T. Stöhlker, and C. Enss, Towards FAIR: first measurements of metallic magnetic calorimeters for high-resolution x-ray spectroscopy at GSI, Physica Scripta 2015, 014054 (2015).DOI
- [21]J. Geist, Bestimmung der Isomerenenergie von ²²⁹Th mit dem hochauflösenden Mikrocalorimeter-Array maXs30, Ph.D. thesis, Kirchhoff Institute for Physics, Heidelberg (2020).DOI
- [22]D. Unger, A. Abeln, C. Enss, A. Fleischmann, D. Hengstler, S. Kempf, and L. Gastaldo, High-resolution for IAXO: MMC-based X-ray detectors, Journal of Instrumentation 16 (06), P06006.
- [23]G. F. Knoll, Radiation Detection and Measurement, 4th ed. (John Wiley & Sons, Hoboken, NJ, 2010).
- [24]F. Clozza, F. Sgaramella, L. Abbene, F. Artibani, M. Bazzi, G. Borghi, D. Bosnar, M. Bradigireanu, A. Buttacavoli, M. Carminati, et al., Extended X-ray energy characterization of SIDDHARTA-2 large-area Silicon Drift Detectors up to 50 keV, Measurement Science and Technology 37, 197001 (2026).
- [25]D. Unger, A. Abeln, T. E. Cocolios, O. Eizenberg, C. Enss, A. Fleischmann, L. Gastaldo, C. Godinho, M. Heines, D. Hengstler, et al., MMC array to study x-ray transitions in muonic atoms, Journal of Low Temperature Physics 216, 344 (2024).
- [26]L. Gerchow, S. Biswas, G. Janka, C. Vigo, A. Knecht, S. M. Vogiatzi, N. Ritoho, T. Proksha, H. Luetkens, and A. Amato, Germanium Array for Non-destructive Testing (GIANT) setup for muon-induced x-ray emission (MIXE) at the Paul Scherrer Institute, Review of Scientific Instruments 94, 045106 (2023).
- [27]A. Knecht, A. Skawran, and S. M. Vogiatzi, Study of nuclear properties with muonic atoms, The European Physical Journal Plus 135, 777 (2020).
- [28]SIS3316 module from Struck Innovative Systeme GmbH, Harksheider Str. 102, 22399 Hamburg, Germany.
- [29]M. Rodrigues, M. L. Zahir, M. Loidl, L. Chambon, Q. Drenne, M. Müller, S. Kempf, E. Nigron, and F. Haddad, Measurements of absolute gamma-ray energies using an ultra-high energy resolution magnetic microcalorimeter, arXiv preprint arXiv:2602.12836 (2026).
- [30]R. D. Deslattes, E. G. J. Kessler, P. Indelicato, L. de Billy, E. Lindroth, J. S. Coursey, D. J. Schwab, C. Chang, R. Sukumar, K. Olsen, and R. A. Dragoset, X-ray Transition Energies (version 1.2) (2005), accessed: 2025-06-18.
- [31]M. S. Basunia, Nuclear Data Sheets for A = 237, Nuclear Data Sheets 107, 2323 (2006).DOI
- [32]S. Seabold, J. Perktold, et al., Statsmodels: econometric and statistical modeling with python, scipy 7, 92 (2010).DOI
- [33]T. Sikorsky, J. Geist, D. Hengstler, S. Kempf, L. Gastaldo, C. Enss, C. Mokry, J. Runke, C. E. Düllmann, P. Wobrauschek, K. Eberhardt, A. Fleischmann, L. von der Wense, and P. G. Thirolf, Measurement of the ²²⁹Th isomer energy with a magnetic microcalorimeter, Physical Review Letters 125, 142503 (2020).DOI
- [34]W. Gins, B. van den Berg, R. P. de Groote, and G. Neyens, SATLAS2: An update to the package for analysis of counting data, Computer Physics Communications 297, 109053 (2024).DOI
- [35]H. J. Lipkin, Some simple features of the Mössbauer effect, Annals of Physics 9, 332 (1960).
- [36]A. Krieger, K. Blaum, M. L. Bissell, N. Frömmgen, C. Geppert, M. Hammen, K. Kreim, M. Kowalska, J. Krämer, T. Neff, R. Neugart, G. Neyens, W. Nörtershäuser, C. Novotny, R. Sánchez, and D. T. Yordanov, Nuclear Charge Radius of ¹²Be, Phys. Rev. Lett. 108, 142501 (2012).
- [37]C.-Y. Seng, Model-Independent Determination of Nuclear Weak Form Factors and Implications for Standard Model Precision Tests, Phys. Rev. Lett. 130, 152501 (2023).DOI
- [38]T. A. Lähde and U.-G. Meißner, Nuclear Lattice Effective Field Theory: An introduction, Vol. 957 (Springer, 2019).DOI
- [39]S. Elhatisari et al., Wavefunction matching for solving quantum many-body problems, Nature 630, 59 (2024), arXiv:2210.17488 [nucl-th].
- [40]S. Shen, S. Elhatisari, D. Lee, U.-G. Meißner, and Z. Ren, Ab Initio Study of the Beryllium Isotopes ⁷Be to ¹²Be, Phys. Rev. Lett. 134, 162503 (2025), arXiv:2411.14935 [nucl-th].DOI
- [41]G. B. King, S. Bacca, G. Chambers-Wall, A. Gnech, S. Pastore, M. Piarrulli, and R. B. Wiringa, Quantum Monte Carlo calculations of Zemach moments in A ≤ 9 nuclei
- [42]G. B. King, J. Carlson, A. R. Flores, S. Gandolfi, E. Mereghetti, S. Pastore, M. Piarrulli, and R. B. Wiringa, Quantum Monte Carlo calculation of δNS in ¹⁰C using an effective field theory approach, arXiv preprint arXiv:2509.07310 (2025), arXiv:2509.07310.arxiv.org/abs/2509.07310
- [43]M. Piarrulli, R. B. Wiringa, A. Lovato, G. B. King, and S. Pastore, Quantum Monte Carlo calculation of δC in the superallowed beta decay of ¹⁰C, arXiv preprint arXiv:2605.14006 (2026).
- [44]C.-Y. Seng and M. Gorchtein, Data-driven reevaluation of ft values in superallowed β decays, Phys. Rev. C 109, 045501 (2024).DOI
- [45]B. He, M. Gorchtein, M. Heinz, B. Ohayon, L. Platter, and C.-Y. Seng, Taming nuclear size and shape effects in superallowed beta-decay, arXiv preprint arXiv:2605.13985 (2026).arxiv.org/abs/2605.13985
- [46]B. Maas, Laser Spectroscopy of the Boron Isotopic Chain, Ph.D. thesis, Technische Universität Darmstadt, Darmstadt, Germany (2020).
- [47]D. Unger, High-Resolution X-ray Spectroscopy of Light Muonic Atoms, Ph.D. thesis, University of Heidelberg, Faculty of Physics and Astronomy, Kirchhoff Institute for Physics, Heidelberg, Germany (2025), dissertation (Dr. rer. nat.).DOI
- [48]G. Hölzer, M. Frötsch, M. Deutsch, J. Härtwig, and E. Förster, Kα₁,₂ and Kβ₁,₃ x-ray emission lines of the 3d transition metals, Phys. Rev. A 56, 4554 (1997).DOI
- [49]M. H. Mendenhall, A. Henins, L. T. Hudson, C. I. Szabo, D. Windover, and J. P. Cline, High-precision measurement of the x-ray Cu Kα spectrum, Journal of Physics B: Atomic, Molecular and Optical Physics 50, 115004 (2017).DOI
- [50]M. H. Mendenhall, L. T. Hudson, C. I. Szabo, A. Henins, and J. P. Cline, The molybdenum K-shell x-ray emission spectrum, Journal of Physics B: Atomic, Molecular and Optical Physics 52, 215004 (2019).DOI
- [51]M. O. Krause and J. H. Oliver, Natural widths of atomic K and L levels, Kα X-ray lines and several KLL Auger lines, Journal of Physical and Chemical Reference Data 8, 329 (1979).DOI
- [52]P. Indelicato, MDFGME: Multiconfiguration Dirac-Fock and General Matrix Elements Program (2026).
- [53]S. Rathi, K. von Schoeler, P. Indelicato, and B. Ohayon, Reference Quadrupole Moments of Transition Elements from Lamb Shifts in Muonic Atoms, Phys. Rev. Lett. 137, 023001 (2026).DOI
- [54]S. Rathi, U. D. Jentschura, P. J. Indelicato, and B. Ohayon, Binding energy of muonic beryllium: perturbative versus all-order calculations, Journal of Physics B: Atomic, Molecular and Optical Physics (2026).
- [55]S. Elhatisari, E. Epelbaum, H. Krebs, T. A. Lähde, D. Lee, N. Li, B.-n. Lu, U.-G. Meißner, and G. Rupak, Ab initio Calculations of the Isotopic Dependence of Nuclear Clustering, Phys. Rev. Lett. 119, 222505 (2017), arXiv:1702.05177 [nucl-th].
- [56]C. Ji, S. Bacca, N. Barnea, O. J. Hernandez, and N. Nevo Dinur, Ab initio calculation of nuclear-structure corrections in muonic atoms, Journal of Physics G: Nuclear and Particle Physics 45, 093002 (2018).DOI
- [57]O. J. Hernandez, C. Ji, S. Bacca, and N. Barnea, Probing uncertainties of nuclear structure corrections in light muonic atoms, Phys. Rev. C 100, 064315 (2019), arXiv:1909.05717 [nucl-th].DOI
- [58]K. Pachucki, V. c. v. Patkós, and V. A. Yerokhin, Three-photon-exchange nuclear structure correction in hydrogenic systems, Phys. Rev. A 97, 062511 (2018).DOI
- [59]Tim Egert and Weiguang Jiang and Sonia Bacca, Optimizing artificial neural networks for nuclear-structure corrections in light muonic atoms (2026), in preparation.
- [60]H. Utsunomiya, Y. Yonezawa, H. Akimune, T. Yamagata, M. Ohta, M. Fujishiro, H. Toyokawa, and H. Ohgaki, Photodisintegration of ⁹Be with laser-induced Compton backscattered γ rays, Phys. Rev. C 63, 018801 (2000).DOI
- [61]H. Utsunomiya, S. Katayama, I. Gheorghe, S. Imai, H. Yamaguchi, D. Kahl, Y. Sakaguchi, T. Shima, K. Takahisa, and S. Miyamoto, Photodisintegration of ⁹Be through the 1/2⁺ state and cluster dipole resonance, Phys. Rev. C 92, 064323 (2015).DOI
- [62]C. W. Arnold, T. B. Clegg, C. Iliadis, H. J. Karwowski, G. C. Rich, J. R. Tompkins, and C. R. Howell, Cross-section measurement of ⁹Be(γ, n)⁸Be and implications for α+α+n → ⁹Be in the r process, Phys. Rev. C 85, 044605 (2012).
- [63]Argonne National Laboratory, Physics Division, Theory Group, Single-Nucleon Densities, https://www.phy.anl.gov/theory/research/density/, accessed: 2026-06-29.
- [64]O. Eizenberg, Measurement of the Beryllium-9 Charge Radius via Muonic Atom Spectroscopy (2025).DOI

