Nondestructive 3D elemental imaging of Edo’s archaeological artifacts via muonic X-ray measurements

Abstract

Cultural heritage artifacts often display complex, multi-layered structures. Therefore, an in situ noninvasive method is required for determining the elemental distribution of such materials. We present a nondestructive 3D elemental analysis approach for an Edo-period ceramic crucible using muonic X-ray measurements. The system employs four CdTe semiconductor imaging detectors to acquire 2D elemental distributions at 12 measurement angles, and 3D images were reconstructed based on computed tomography. We adjusted muon energy to penetrate the metallic layer covering the crucible’s bottom and analyze the interior of the crucible. Silicon(Si) and oxygen(O) signals were successfully identified, and these signals, originating from highly abundant silica(SiO₂) within the crucible matrix, confirmed the ceramic composition. These findings demonstrated the capability of muonic X-ray imaging in achieving depth-resolved analysis by controlling the observation position of an object. The presented method is broadly valuable for the cultural heritage study.

Introduction

Analyzing the elemental composition of archaeological artifacts and cultural heritage objects is beneficial for gaining insights into past societies. As these materials are often irreplaceable and cannot be reproduced, these must be characterized using nondestructive, noncontact, and noninvasive approaches. Such methods preserve the integrity of rare artifacts for future generations and enable researchers to uncover valuable historical and cultural narratives. Among established non-destructive techniques, X-ray fluorescence(XRF)¹ and energy dispersive X-ray spectroscopy(EDS)² are widely used for the elemental analysis of cultural materials^3,4,5,6. However, because these techniques typically rely on low-energy X-rays, the penetrating power is limited and significant self-absorption of X-rays occurs within the sample⁷. Consequently, these methods often capture only surface-level information and lack precise depth profiling ability. For instance, detecting chemical changes in metals or alloys within a ceramic crucible or determining the depth of decorative layers in pottery become difficult when the measurement depth is unclear. Furthermore, these methods cannot be employed for analyzing complex shapes due to X-ray absorption effects. Therefore, advanced methods that provide reliable, depth-resolved elemental data are required to accurately analyze the surfaces and interiors of archaeological samples.

To address this issue, negatively charged muons that resemble electrons are used for nondestructive elemental analysis⁸. These muons are captured by the Coulomb field of a nucleus, which then forms a “muonic atom” with “muonic orbitals.” The mass of a muon is 207 times that of an electron; therefore, it penetrates deeper into the sample. The considerably high binding energy of muonic orbitals facilitates the emission of high-energy muonic X-rays, with higher energy (~207 times) than that of fluorescence X-rays. These muonic X-rays can escape from the sample without being absorbed, providing deeper insights into the elemental composition of the sample without destruction. This technique has made significant progress in the analysis of cultural heritage objects at institutions such as the Paul Scherrer Institute(PSI) in Switzerland, the ISIS Neutron and Muon Source in the UK, and the Japan Proton Accelerator Research Complex(J-PARC) in Japan. For example, a knob bow fibula dating from the late 4th to early 5th century CE was examined using muon elemental analysis at PSI, revealing the presence of copper(Cu), zinc(Zn), tin(Sn), and lead (Pb) in the sample⁹. Muon elemental analysis at the ISIS Neutron and Muon Source also enabled a depth-dependent, nondestructive analysis of a denarius(made from a silver-copper alloy) from the period when the Empress Julia Domna reigned (211-217 CE), accurately determining its copper and silver contents¹⁰. At J-PARC, muon elemental depth profiling was employed to study an old Japanese gold coin(Tempo-Koban), and the results indicated that the gold(Au) concentration gradually decreases with depth over a micrometer scale¹¹. Furthermore, OGATA Koan’s sealed medicine, a cultural heritage object in the medical field, was successfully analyzed using muonic X-ray analysis at J-PARC¹². Although conventional X-ray methods cannot penetrate the thick lead-glass container to reveal its contents, muon elemental analysis detected mercuric chloride within the sample. These studies represent a notable advancement in the nondestructive analysis of cultural heritage, enabling the examination of valuable artifacts without causing any damage. The availability of artificially produced high-intensity muon beams from accelerators facilitate such elemental analyses. However, these aforementioned studies did not fully address the spatial distribution of elements within samples. Archaeological artifacts require three-dimensional(3D) elemental mapping as they have diverse shapes, sizes, and material compositions, with layered or mixed-media components. Depth-resolved analysis is crucial for understanding surface characteristics and interior composition of such artifacts, allowing researchers to gain deeper insights into their construction, usage, and cultural significance.

The cadmium telluride double-sided strip detector(CdTe-DSD) are high-Z-material-based detectors that have been recently used as two-dimensional (2D) imaging detectors with a large imaging area of 32 × 32 mm² in astronomy^13,14,15,16. CdTe-DSDs are ideal for detecting hard X-rays and γ-rays due to the high absorption rate of high-Z CdTe and the ability to produce detectors with a thickness of several millimeters. Additionally, a high position resolution(within a few hundred micrometers) has been achieved using a strip pitch of 250 μm by applying new electrodes made of Pt and Al bonded to the cathode and anode, respectively^17,18,19. Recently, 3D imaging based on muonic X-ray measurements using CdTe-DSDs was demonstrated to effectively detect light elements such as B, N, F, and C in samples^20,21. To determine the 3D elemental composition of cultural heritage objects, a computed tomography(CT) scanning system was developed herein by integrating four CdTe-DSDs. These detectors captured projection images from multiple angles, enabling a 3D reconstruction of elemental distribution. The system was employed for analyzing an archaeological artifact at the J-PARC.

The Tokyo Metropolitan Archaeological Center conducted its sixth excavation of the Yotsuya 1-chome site in Shinjuku, Tokyo, during which a crucible vessel from the Edo period (Fig. 1) was unearthed. This crucible is now in the possession of the Shinjuku ward and studied herein. During the Edo period, merchants and artisans resided in Shinjuku to supply daily necessities and services to nearby samurai households. Among these artisans were metalworkers who used small crucibles for refining metals or crafting goods. Microscopic observations and XRF analysis of vessels similar to the crucible revealed traces of gold and silver at the bottom of the crucible, suggesting that small-rod metal refining might have been performed then. This process was perhaps used for manufacturing hair ornaments and other small-scale metal objects. Further studies examining elements in these vessels are required to understand Edo-period metalworking and, by extension, the broader “Edo monozukuri”(Edo craftsmanship).

Fig. 1: Sample photographs.

a Image of the analyzed crucible vessel, which is a ceramic artifact from the Edo period. The crucible’s inner bottom surface is attached with refined metal(black substance). b Image of the crucible mounted on an aluminum stick using a monofilament fishing line. The crucible has an inner diameter of 33 mm with a thickness of 22 mm. c Position of the crucible relative to the muon beam direction(red cross) with the defined coordinate system(right upper).

Metal products were fabricated during the Edo period as follows. First, metal pieces(such as ingots) were produced in large production centers outside the city of Edo. Then, the process of changing the properties of the parts, products, or the alloys and metals corresponding to these products is thought to have been carried out in the city of Edo. These metals were then transformed into parts or finished products via assembly, decorative techniques, and sometimes recycling²². The red rust and green patina on these artifacts indicate the presence of iron and copper, respectively. Nevertheless, recent scientific analyses revealed that the properties of iron were tuned to enhance its quality and various copper alloys were widely used during the Edo period. However, methods for investigating the surface and interior structures of the products or parts including the chemical interactions during the production of metal alloys, have not been developed yet. For instance, some documents mention an Edo-period color-enhancement process known as “iroage” for metal currency^23,24, and the traces of such treatments might be removed while cleaning the artifacts²⁵. Therefore, we need an advanced analytical approaches to investigate such process.

As bird-keeping was popular during the Edo period, the crucible vessel studied herein was originally produced as a bird feeder and subsequently repurposed. This ceramic vessel contained a handle, which possibly made it easier to use it as a crucible. It was produced in the Seto and Mino regions, located ~400 km west of Edo, and has a white clay body covered with a yellowish ash glaze and an unglazed bottom. Archaeologists also found a mid-sized jar from the Seto-Mino area in the Yotsuya 1-chome site. We believe that the crucible vessel was used for melting small amounts of metal, while the larger jar was used for melting the larger objects. XRF analysis revealed copper-based alloys in the larger jar, along with fragments of a brass crucible lid²⁴. Notably, portable XRF devices cannot easily reach the surface where metal melting or chemical reactions occurred in these small crucibles due to their limited interior space. To address this issue, an nondestructive, in situ 3D analysis was performed herein to determine the exact zones of metal melting or chemical reactions, particularly using muon-based 3D elemental imaging. Using this method, chemical changes and elemental distribution within the crucible can be determined. As Edo metalworkers did not operate within the modern scientific frameworks of standardized units or elemental concepts, the scientific investigation of these artifacts will be beneficial for current industrial practices and artisan craft traditions. Additionally, whether the heat-resistant clay or pottery materials influenced metal processing during the Edo period will be analyzed. In future, 3D elemental imaging methods can significantly contribute to studies on cultural heritage objects.

Methods

Muon facility at J-PARC

The elemental 3D imaging analysis using a negative-muon beam was conducted at the Muon Science Establishment (MUSE) of the J-PARC, which hosts the world’s most intense pulsed negative-muon beam²⁶. In this facility, a double-pulsed muon beam is generated by bombarding a carbon target with a 600-kW proton beam at a frequency of 25 Hz, which is accelerated to an energy level of 3 GeV²⁷. The momentum of the muons can be precisely adjusted from 5 to 120 MeV/c using a superconducting solenoid magnet along the D2 muon beamline²⁸. The D2 beamline at J-PARC is a user-accessible facility for innovative muon experiments. For this study, we temporarily installed our CT imaging system on D2 to evaluate the feasibility of muon elemental imaging for cultural heritage objects. After the experiment, we removed our equipment, ensuring the beamline remained fully available for other experiments.

CdTe-DSD

CdTe-DSD is particularly well suited for measuring hard X-rays and soft γ-rays because they contain CdTe with a high Z value, thus making them ideal for muonic X-ray measurements. The proposed CdTe-DSD with an Al/CdTe/Pt Schottky diode configuration covers an energy range of 5–200 keV and provide a spatial resolution of few hundred micrometers for CT scanning systems used in the medical field²⁹. The detector features separate readout electronics, which are orthogonally positioned on the anode and cathode sides. Each side of the CdTe-DSD has 128 strip electrodes with a strip pitch of 250 μm and a gap width of 50 μm. This orthogonal placement enhances precision in localizing incident photons, facilitating the application of the array as an imaging detector. A CdTe-DSD with noting a sensitive area and thickness of 32 mm × 32 mm and 0.75 mm, respectively, was previously used in muon experiments, particularly for light-element analysis^20,21. The thickness of the CdTe crystal used in this study was increased from 0.75 mm to 2 mm to enhance detector sensitivity, as the examined cultural artifact contained heavy elements such as Cu and Ag that generate muonic X-rays in the higher energy range of 100–200 keV. Figure 2 shows the 2-mm CdTe DSD developed by iMAGINE-X Inc, in which CdTe crystals and electrodes are connected via the Schottky contact. A high voltage of 500 V was applied to the electrodes for collecting charge carriers. However, such a high voltage tends to accumulate charge over time on both electrodes altering the internal electric field³⁰. Consequently, the internal electric field impacted the energy resolution of the detector. The detector cooled using an electronic cooling system at low-temperature conditions (−20 °C) to ensure performance stability¹⁶. To maintain this low-temperature environment, the CdTe DSD was enclosed within a dedicated aluminum housing with a size of 190 × 140 × 200 mm³ (Fig. 2). The resulting 2-mm-thick CdTe-DSD achieved an energy resolution of 3.1 keV at 122 keV and a position resolution capable of resolving features as small as 250 μm³¹.

Fig. 2: Image of a 2-mm CdTe-DSD with sensitive area of 32 mm × 32 mm.

An application-specific integrated circuit was set on an front-end card board and wire-bounded to the CdTe-DSD.

3D phantom reconstruction

Single photon emission computed tomography(SPECT)³² is a widely used nuclear imaging technique in medicine. It reconstructs the original location of radiation sources based on X-rays or γ-rays measured using a camera. An iterative reconstruction algorithm for SPECT, known as the maximum likelihood expectation maximization (MLEM) algorithm^21,33, provides geometrically accurate images with a higher spatial resolution compared to those obtained using analytic approaches such as filtered back projection (FBP)³⁴. MLEM achieves effective noise suppression and solves data insufficiency issues using projection data to reconstruct 3D images. Some critical parameters of MLEM, such as source intensity, number of photons detected, and angle at which photons hit detectors, are sensitive to detector characteristics and the FOV of the pinhole collimator. The point spread function(PSF) enhances the spatial resolution and mitigates image degradation within the MLEM algorithm³⁵ by improving the fidelity of reconstructed images. The PSF can be determined via experimental measurements or theoretical calculations. Experimentally, the PSF is determined by assessing the response of an imaging system to a known radioactive source with known dimensions. Herein, a theoretical method was adopted to determine the PSF using the Geant4³⁶ toolkit for Monte Carlo simulations. The computed PSF was subsequently integrated into the MLEM algorithm as a component of system response. The Geant4 simulation was designed to replicate the experimental imaging setup at the J-PARC. The validity of the simulation was confirmed by comparing the simulated images with experimental reconstruction images obtained using a ¹³³Ba radioactive isotope source with dimensions of 1 × 1 × 1 mm³. As shown in Fig. 3a, the ¹³³Ba source was positioned at three locations (pos1, pos2, and pos3) within the object space, and measurements were performed to produce reconstructed images using a CdTe-DSD(204). Figure 4 shows these reconstructed images and corresponding simulated images, revealing that the simulated images accurately replicated the data obtained from the CdTe-DSD. During simulation for determining the PSF muonic X-rays were emitted in 4π directions within the object space, covering a volume of 30 × 30 × 30 mm³. The volume was defined so as to correspond to the solid angle of the imaging system. The projection image obtained from the CdTe-DSD was then simulated to establish the system response for the MLEM algorithm. Consequently, the system response was structured into a 128 × 128 × 30 × 30 × 30 matrix, corresponding to the pixel resolution of the CdTe-DSD and the voxel distribution within the object space. This simulation integrated factors such as scattering, absorption, and detector efficiency into the PSF, enhancing the precision of the MLEM algorithm.

Fig. 3: Schematic of two experiments using a ¹³³Ba source

a Experiment for measuring projection images of a ¹³³Ba source at three positions (pos1, pos2, pos3) within the object space. b Experiment for 3D imaging of the ¹³³Ba source distribution using the CT imaging system.

Fig. 4: Comparison of measured and simulated images.

(Upper) Image obtained using a ¹³³Ba source at pos1, pos2, and pos3 by employing a CdTe-DSD(204). (Bottom) Corresponding image obtained via the Geant4 simulation.

Figure 3b shows a subsequent radioisotopes(RI) experiment conducted to verify the practical application of system response, wherein the ¹³³Ba source was fixed at coordinates (X, Y, Z) = (−7, 4, 3) within the object space. The imaging system captured ¹³³Ba projection images from our angles; these images were used alongside the simulated system response to reconstruct a 3D image using the MLEM algorithm. Figure 5 shows that the reconstructed 3D phantom reflected the diameter and position of the ¹³³Ba source. The position and size of the reconstructed 3D phantom along the X, Y, and Z axes were calculated using a Gaussian fitting method, which yielded mean values of (X, Y, Z) = (−7.09, 4.34, 3.18) for the position. These results confirmed that the position of the reconstructed phantom aligns with the actual position of the ¹³³Ba source based on the spatial resolution of the imaging system. A rotated view of the 3D ¹³³Ba phantom is available in the Supplementary Video. These results confirm the accuracy of the Geant4 simulation; it can thus be applied for the 3D imaging of the crucible from the Edo period.

Fig. 5

Reconstructed 3D phantom of the ¹³³Ba source: side view (left) and top view (right).

Results

Muon experiment

Figure 6a shows the experimental setup at the J-PARC without the sample. Four CdTe-DSDs(201–204) were mounted on an aluminum circular rotating stand with an inner diameter of 510 mm. Figure 6b shows the configuration of a CdTe-DSD(204) alongside the crucible, which measured 39 mm in diameter, 22 mm in height, and weighed 22.0 g. The crucible was secured on an Al stick using a monofilament fishing line at the center of the stand. The crucible walls were 3-mm thick and its base was 3–4-mm thick. To obtain the sample projection image using muonic X-rays from the sample, the CdTe-DSDs were combined with a 2-mm-diameter tungsten pinhole collimator. This collimator with a knife-edge geometry provided a field of view(FOV) of 62.4^∘. The spatial resolution of a projection image(R_g) is associated with the intrinsic resolution of the detector and the spatial resolution of the collimator³⁷, which can be expressed as follows:

$${R}_{g}=\sqrt{{\left(\frac{{d}_{1}+{d}_{2}}{{d}_{1}}\times r\right)}^{2}+{\left({R}_{i}\times \frac{{d}_{2}}{{d}_{1}}\right)}^{2}},$$

(1)

where r is the diameter of the pinhole collimator; R_i is the CdTe-DSD resolution of 250 μm; d₁ is the distance between the detector and pinhole collimator, which was constant(48 mm) herein; and d₂ is the distance between the pinhole and imaging target(41 mm at the center of the crucible). The spatial resolution of the projection image is proportional to the pinhole diameter. Although the spatial resolution can be enhanced by decreasing the pinhole size it compromises sensitivity due to the resolution-sensitivity trade-off. A 2-mm-diameter pinhole collimator was used herein based on the crucible size and and CdTe-DSD sensitivity. Consequently, R_g was estimated to be 3.7 mm. In the muon beam experiment, the muon stopped position(muon range) within the sample was tuned by adjusting the muon momenta to 25 and 30 MeV/c to determine the elemental composition at various depths. The beam exposure durations for these momenta were 24.6 and 28.4 h, respectively, and the rate of incident negative muons was estimated to be ~10⁵ per second. During muon irradiation, three rotation movements were performed (0^∘, + 30^∘, and − 30^∘ around the muon beam axis) to obtain 12 projection images from the four CdTe-DSDs.

Fig. 6: Muon experimental configurations.

a Image of the CT scanning system comprising four CdTe-DSDs(201–204) at the rotation movement of 0^∘. The sample is mounted on an aluminum stick using a monofilament fishing line and positioned at the center(indicated by the red box) of the CT scanning system. b Experimental setup at the D2 muon beamline area/MUSE showing the sample position relative to a CdTe-DSD(204). A 2-mm-diameter pinhole collimator is used to obtain the projection image of the sample.

Energy spectra and projection images

Figure 7a, b shows the energy spectra obtained using the CdTe-DSD data for muon momenta of 25 MeV/c and 30 MeV/c (black lines). The FOV of the pinhole collimator and its relative position to the sample caused a substantial portion of the signals to be concentrated in the central region of the CdTe detector. Figure 7a shows the CdTe-DSD signal distribution for overall measurement data. Most signals, originating from O and Si within the crucible, appeared as a horizontal pattern in the center of the CdTe-DSDs. Therefore, the regions of interest(ROIs; green boxes) and background region (blue box) were defined to ensure that the ROIs and background region have the same area. The energy spectra corresponding to these regions are represented by green and blue lines, respectively, under 25 and 30 MeV/c. The red line shows the differential energy distribution obtained by subtracting the background-region spectra from the ROI spectra.

Fig. 7: CdTe-DSD energy spectra.

Energy spectra obtained under two muon momentum conditions: a 25 MeV/c and b 30 MeV/c. The black line represents the energy spectra obtained using total measurement data. c shows an image of the CdTe-DSD, highlighting the regions of interest (ROIs) and background (bkg) region marked by green and blue boxes, respectively. The green and blue lines represent the energy spectra within the ROIs and bkg region, respectively. The red line denotes the energy spectra of ROI − bkg.

Signals in the background-region spectra originated due to the Compton effect, electron bremsstrahlung, and characteristic X-rays emitted from the pinhole collimator. Among these background sources, the electron bremsstrahlung contributions from electrons produced by muon decay cannot be completely eliminated at J-PARC. The muons decayed into electrons, with a lifetime of 2.197 μs. Despite using a magnetic field in beam commissioning at J-PARC to separate muons from decay electrons, the decay electrons could not be completely eliminated from the beam. Consequently, electron contamination contributed to an electron bremsstrahlung background, resulting in a continuous background component in our energy spectrum³⁸. The CdTe-DSDs recorded the time interval(Δt) between the start of each beam pulse from the beam line electronics and the trigger pulse from the detector. Two peak structures within 10,250 ns < Δt < 10550 ns and 10,800 ns < Δt < 11,200 ns were observed in the muon pulse structure. Events distributed in the tail component were attributed to continuous background noise, likely caused by electron bremsstrahlung or the Compton effect. By selecting events within these two regions, the continuous background was effectively reduced without compromising the muonic X-ray flux. When high-energy(>100 keV) muonic X-rays hit the tungsten pinhole collimator, characteristic X-rays with an energy of 59.3 keV were emitted from tungsten. A graded-Z shield, composed of 3-mm-thick Pb, 1-mm-thick Sn, and 1-mm-thick Cu sheets was installed between the pinhole collimator and the CdTe-DSD to reduce the background noise from characteristic X-rays. However, the characteristic X-ray distribution of tungsten could be observed in the energy spectrum because tungsten collimators emit several X-rays that cannot be fully shielded. The muonic X-rays of μO(2-1), μSi(3-2), and μAl(3-2), with energies of 133.4, 76.4, and 65.9 keV, respectively, were clearly observed in the red energy spectra shown in Fig. 7. The numbers in parentheses indicate the change in principal quantum numbers when muonic X-rays are emitted, e.g., (2-1) represents X-rays emitted from the 2p–1s muon atomic transition. The μO(2-1) and μSi(3-2) signals were emitted from the main component of the ceramic vessel (SiO₂). The μAl (3-2) signal also possibly originated from the ceramic vessel. This crucible was produced in the Seto or Mino regions, which are known for ceramics made with clays rich in minerals such as kaolinite(Al₄Si₄O₁₀(OH)₈) and sillimanite (Al₂SiO₅). The Al content in these clays likely contributed to the μAl signal in the energy spectrum. In the energy spectra shown in Fig. 7, clear μAl and μW peaks appear in both the ROI (green) and the background (blue). The μAl peak is stronger in the ROI than in the background, suggesting that some of this signal comes from Al content in the clay sample. However, μAl contribution may also originate from Al components in the experimental setup, such as the Al stick of the sample holder and the Al housing of the CdTe-DSD. Because they are in the same viewing angle as the sample, we cannot accurately distinguish these contributions. As the result, a detailed discussion on the μAl signal is excluded. No obvious peaks indicating the presence of Ag and Cu were observed in the energy spectrum for the crucible, although these elements were confirmed via XRF analysis, because the energy of 25- and 30-MeV/c muons was considerably high for them to stop within the metal layer. These high-energy muons penetrated the metal and stopped within the bottom of the ceramic vessel. Herein, measured events within the energy ranges of 125–140 keV and 72–82 keV, corresponding to the μO(2-1) and μSi(3-2) signal respectively, were selected to reconstruct images depicting the spatial distribution of O and Si.

Figure 8 shows the reconstructed images of μO(2-1) and μSi(3-2) obtained by using the CdTe-DSDs 201–204 (from left to right). As the muons stopped at the bottom of the crucible, these images captured the projection of the crucible bottom onto the CdTe-DSDs from various angles. As shown in Fig. 8, areas marked with a black box indicate regions with considerably weak signals. This phenomenon arose from connection problems between the electrodes and ASIC in the CdTe-DSDs, resulting in 1, 8, and 7 dead channels in the CdTe-DSDs 201, 202, and 203, respectively. These channels did not register any signals during the measurement process, affecting the overall signal integrity. As signals should exhibit continuity, the signals surrounding the dead channel were extracted and interpolation was employed to fill these gaps. However, the presence of these dead channels possibly continued to compromise the accuracy of the 3D phantom reconstruction. To address these issues in future studies, a new type of CdTe DSD has been developed using wire bonding technology. This advancement will considerably reduce the possibility of encountering dead channel problems, enhancing the reliability and quality of data captured by these detectors.

Fig. 8: Reconstructed images of μO(2-1)(upper plots) and μSi(3-2)(bottom plots) obtained from the CdTe-DSDs 201–204 at the rotation movement of 0^∘, displayed from left to right.

The black box indicates the regions containing the dead channels. “Al side” designates the face with the aluminum electrode(cathode, Y-axis), and “Pt side” designates the face with the platinum electrode (anode, X-axis).

3D phantom

The MLEM (maximum-likelihood expectation-maximization) algorithm and the simulated PSF were employed to reconstruct the combined 3D phantom of μO(2-1) and μSi(3-2). In particular, the projection images of μO(2-1) and μSi(3-2) acquired using the CT scanning system were reconstructed. Figure 9 shows the side view (a-1 and b-1) and beam-aligned view(a-2 and b-2) of the 3D phantom within a Cartesian coordinate system. A rotated view of the reconstructed 3D phantom is available in Supplementary Video, wherein SiO₂ distribution is shown. The μO(2-1) and μSi(3-2) signals mainly originated from the SiO₂ base material underneath the metal layer of the crucible, indicating that muons penetrated the metal surface and stopped within the ceramic interior(see Supplementary Information). The 3D phantom revealed the SiO₂ distribution, which formed a uniform disk measuring 25 mm × 4 mm; notably, variations in signal intensity were also observed.

Fig. 9

Side view (left) and beam-aligned view (right) of the reconstructed 3D phantom showing SiO₂ distribution at the crucible’s bottom, as measured using muon momenta of 25 MeV/c(top) and 30 MeV/c(bottom).

First, differences between the signal intensities of the center and periphery were attributed to the self-absorption effects of the ceramic material. Muonic X-rays emitted from the center of the crucible bottom experienced stronger self-absorption than those emitted near the edges. Consequently, the muonic X-ray intensities were higher at the periphery than at the center.

Second, as shown in Fig. 9a-2 and b-2, the 3D phantom displayed a higher intensity in the lower right corner due to the presence of dead channels in the CdTe-DSDs. Signal loss around the center of the 203 detector (Fig. 8) caused the image to split into two sections during reconstruction, with stronger signals on the left side compared to those on the right side. This segmentation induced intensity variations during the iterative reconstruction process, further affecting the final intensity distribution.

Finally, weaker intensity was observed on the left side (5 < x < 15) of the 3D phantom compared to the right side (−15 < x < −5), as shown in Fig. 9a-2 and b-2. This discrepancy resulted from the asymmetrical emission pattern of the muon beam at the J-PARC, with an elliptical shape and a smaller beam width in the X-direction than in the Y-direction³⁹. In this experimental setup, the center of the muon beam was slightly offset toward the right (Fig. 1c, − X direction), resulting in fewer muons stopping in the left region and producing weaker signals. The SiO₂ distribution of the crucible was thus successfully reconstructed. However, the observed intensity nonuniformity indicated that the CT scanning system requires further improvement. Future work will involve using a new CdTe-DSD without dead channels, with corrections for self-absorption and refined 3D reconstruction by incorporating correction factors based on additional measurements of the muon beam structure to improve the quality of reconstructed 3D images.

Discussion

Herein, the elemental distribution of a small crucible from the Edo period was determined. The muon elemental analysis method provides an approach for analyzing archaeological artifacts with complex morphological features, while preserving their structural integrity. A nondestructive 3D elemental analysis was performed by irradiating a negative-muon beam at the J-PARC on a crucible of the Edo period. To capture muonic X-ray signals from the crucible at various viewing angles, a CT scanning system composed of four CdTe-DSDs with 2-mm-diameter pinhole collimators was developed. After exposing the crucible to muons with a momenta of 25 and 30 MeV/c for 24.6 and 28.4 h, respectively, muonic X-ray signals of μSi(3-2) and μO(2-1) were successfully detected in the energy spectra. As the crucible’s ceramic body was primarily composed of SiO₂, the μSi(3-2) and μO(2-1) signals originated from the ceramic layer. This result indicated that muons with a momenta of 25 and 30 MeV/c could penetrate the metallic layer and analyze the ceramic base material. Signals in energy ranges of 125–140 and 72–82 keV were selected to reconstruct the projection images of μO(2-1) and μSi(3-2), with a spatial resolution of 3.7 mm. These images were then used to reconstruct a 3D phantom using the MLEM algorithm, with the PSF evaluated via Monte Carlo simulations using the Geant4 toolkit. The reconstructed 3D phantom revealed the SiO₂ distribution in the ceramic body of the crucible. The successful reconstruction of 3D images at a specific depth represents significant progress. From this depth-controlled imaging, the elemental distribution of the ceramic layer in the crucible was accurately determined. In the future, the interaction among the metal layer, glaze, and ceramic body will be analyzed. Future improvements will include enhancing signal intensity, reducing background noise by employing a multi-pinhole collimator, and adding shields around the CdTe-DSDs. The proposed method provides a comprehensive 3D elemental distribution map, considerably advancing the analysis of archaeological materials. This technique is the only nondestructive and noninvasive method capable of depth profiling light elements to several millimeters. It shows great promise for future chemical analyses of complex archaeological artifacts and cultural heritage objects with intricate or hollow structures, enabling depth-controlled 3D investigations. These advancements can contribute to uncovering craftsmanship in the Edo period, including metal refining, processing, and decoration techniques, as well as layered decorations in ceramics. The elemental 3D imaging method based on muonic X-rays holds potential for further development. It is positioned to become a new analytical technique for material characterization in archaeological science.