Neutron tomography of a corroded Han dynasty shu dao (书刀) using the Energy-Resolved Neutron Imaging Instrument

Iron-based cultural artefacts from ancient China are severely threatened by corrosion, and traditional non-destructive imaging techniques (e.g. X-ray imaging) face shortcomings in analysing the composite structures of metals, corrosion products and organic materials: low-energy X-rays cause metal hardening artefacts, while high-energy X-rays are minimally attenuated by organic components, leading to incomplete structural characterization. To address this scientific challenge, this study employed neutron tomography using the Energy-Resolved Neutron Imaging Instrument at the China Spallation Neutron Source to conduct non-destructive characterization of a corroded Han dynasty shu dao, a typical iron-based knife used for editing bamboo slips in ancient China. Neutrons exhibit unique advantages: strong penetration of iron matrices and high sensitivity to hydrogen, enabling clear differentiation between corrosion products and the intact iron matrix. The results demonstrate that neutron tomography enables the detection of rusted regions on the knife to characterize corrosion severity, segmentation of the blade's crack network, observation of the scabbard's layered structure and measurement of key features such as the blade's reinforcement.

1. Introduction

China boasts one of the world's richest reserves of iron cultural artefacts, with ancient iron smelting technology centred on wrought iron and steelmaking (Zhang & Bahetibieke, 2025; Zou et al., 2024; Dussubieux et al., 2024). The shu dao (书刀) was a specialized writing tool for scraping off characters written on bamboo slips and wooden tablets (Sha et al., 2024). On the one hand, when characters were written incorrectly, a knife was needed to scrape off the erroneous words and sentences. On the other hand, due to the high production cost of bamboo and wooden slips, many slips used for daily recording were reused repeatedly; to scrape off the previously written characters, a shu dao was also required as an ancient Chinese writing tool similar to a modern eraser. It emerged as early as the Shang dynasty (ca 1600 BC to 1046 BC); early shu dao were cast in bronze, and iron versions appeared later. The shu dao and mao bi (毛笔, a traditional Chinese writing brush) were used by government clerks in charge of legal affairs, and they also served as cultural relics symbolizing legal authority and cultural inheritance. By the Wei and Jin dynasties (AD 220–420), paper gradually replaced bamboo and wooden slips, and the shu dao consequently faded out of historical use.

The shu dao (writing knife) represents the quintessential iron craftsmanship of the Western Han dynasty (202 BC to 8 AD). The ring-pommel variant prevailed in this period (Song et al., 2025); its looped terminal not only prevented slippage during use but also functioned as a suspension device for portability, and even included decorative elements. However, the metal–carbon matrix of these artefacts is highly susceptible to electrochemical and chemical corrosion under long-term environmental influences, including moisture and oxygen. This corrosion leads to structural expansion, cracking and adhesion to organic components such as wooden scabbards (Jia et al., 2022; Zhao et al., 2024). It not only damages the artefacts' physical integrity but also erodes key information about their manufacturing techniques and historical context.

Understanding the corrosion status and internal structure of such artefacts is critical for three core objectives. For conservation, identifying corrosion distribution and severity allows the development of targeted anti-corrosion measures to slow or halt degradation, thus extending the artefacts' lifespan. For archaeology, revealing uncorroded matrices and structural details such as the hilt and blade helps restore ancient manufacturing processes and functional designs. For restoration, distinguishing the bonding state between metal bodies and organic scabbards avoids damage during potential restoration, which preserves the artefacts' original historical information (Bouchar et al., 2017; Monnier et al., 2010; Molina et al., 2023; Di et al., 2025; Zhao et al., 2024).

X-ray imaging is the most widely used non-destructive technique in cultural artefact research (Bossema et al., 2024; Iori et al., 2025; Porcaro et al., 2025; Magdy, 2023; Hoshino et al., 2017; Hain et al., 2017), but it fails to meet the characterization needs of corroded iron artefacts with composite structures. For metal–corrosion-product systems, low-energy X-rays are strongly attenuated by iron, generating hardening artefacts that obscure corrosion-product boundaries; high-energy X-rays penetrate metals but cannot distinguish corrosion products. For metal–organic composite systems, organic materials have low atomic numbers, resulting in minimal X-ray attenuation, so they are almost invisible in images, making it hard to analyse the interaction between organic and metal materials.

Neutron tomography differs from X-ray imaging, which relies on atomic-number-dependent attenuation. Neutrons interact with matter through nuclear scattering and absorption, endowing them with unique advantages and promising applications in cultural artefact characterization, particularly in corrosion and ageing studies of metal artefacts (Roy et al., 2024; Ryzewski et al., 2013; Salvemini, 2024; Li et al., 2024). This is primarily determined by the principles of the neutron beam's inter­action with matter: first, it exhibits excellent penetration through iron matrices, as the weak interaction between neutrons and iron nuclei allows the beam to pass through thick iron components without severe attenuation; second, it is highly sensitive to hydrogen-containing substances, with hydrogen atoms having a much larger neutron total cross section (approximately 84 barns) than iron (about 4 barns) (Sears, 1992). Since iron corrosion products [e.g. α-FeO(OH), γ-FeO(OH) and Fe₂O₃·H₂O] and organic materials such as wooden scabbards are rich in hydrogen, they display strong neutron attenuation and clear greyscale contrast relative to iron matrices. For this reason, neutron tomography not only enables effective observation of rust structures (Zhang et al., 2018; Taketani et al., 2018) but also facilitates the observation and analysis of adhesion or embedding between different components in hydrogen-containing organic–metal composite cultural artefacts due to ageing.

The Energy-Resolved Neutron Imaging Instrument (ERNI), as China's first dedicated pulsed neutron imaging instrument at the China Spallation Neutron Source (CSNS), boasts multi-scale and multi-modality characterization capabilities, making it an ideal platform for studying corroded iron artefacts with composite structures (Chen et al., 2024). This study aims to verify the feasibility of ERNI-based neutron tomography for the non-destructive characterization of a corroded Han dynasty shu dao (Song et al., 2025) with metal–organic composite structures, and to address key scientific questions, including the spatial distribution and severity of corrosion on the blade, the high-risk initiation locations and potential causal mechanisms of the blade's crack network, and the structural details of the scabbard and its supporting components.

2. Methodology

2.1. Sample information

The research sample (Fig. 1) is a corroded Western Han dynasty shu dao (202 BC to 8 AD), a cultural artefact with a history of 2000 years. It was excavated from the Hu Chang Han tomb in Jiangsu province (Wang et al., 1980) and is preserved at the Yangzhou Museum.

Figure 1 Photographs of the shu dao.

From the artefact's appearance, its scabbard is lacquerware – an object decorated with raw lacquer. Raw lacquer is a greyish milky liquid secreted by lacquer trees; it oxidizes to a chestnut brown when exposed to air, turns dark brown upon drying, and possesses unique properties of moisture resistance, high-temperature resistance and corrosion resistance (Sung et al., 2016). The significance of the lacquerware scabbard for the preservation of the knife lies in its ability to isolate the iron blade from moisture, oxygen and soil microorganisms, thereby mitigating corrosion and improving the artefact's long-term preservation.

Yet after two millennia, the artefact has not been well preserved and has developed a 6.4 cm-long crack. Furthermore, the entire blade is fully encased in the outer lacquer scabbard, with no exposed hilt. It is inferred that the original hilt has broken off due to prolonged corrosion. Currently, the artefact has dimensions of approximately 17.4 cm in length, 1.4–1.8 cm in width and 0.8–0.9 cm in thickness. The wooden lacquerware scabbard has adhered to the knife body and cannot be separated without causing damage. The artefact has visible surface rust and scabbard cracks, making it representative of typical metal–organic composite corrosion in ancient iron artefacts.

2.2. Experimental setup and workflow

We conducted neutron tomography experiments on ERNI, the No. 13 beamline of the CSNS. ERNI has a total length of 43.1 m. For the experiments, the aperture of the pinhole selector was set to 20 mm, with its position fixed at the 24.78 m mark of the beamline. The neutron wavelength range was controlled between 0.5 and 4.6 Å using choppers, while the sample was placed at a distance of 30 m from the coupled hydrogen moderator.

For the neutron tomography experiments, all projection images were acquired using a dedicated neutron detector system consisting of a 50 µm-thick ZnS/⁶LiF scintillator screen, a CCD camera (manufactured by Andor, a brand under Oxford Instruments, UK) and an optical assembly fitted with a photographic lens (Nikon, Japan). The CCD camera features a 2048 × 2048 pixel array with a single-pixel size of 13.5 µm. The optical lens was adjusted to a magnification of approximately 0.5×, yielding a field of view (FOV) of 5.5 × 5.5 cm. During the tomography scan, a total of 641 projection images were collected over an angular range of 0° to 320° with an angular step of 0.5°, and each projection image was captured with an exposure time of 100 s.

To ensure data accuracy and structural integrity, the workflow was designed with strict quality control measures. The shu dao was vertically mounted in a cylindrical aluminium holder to prevent deformation during scanning; since its length exceeded the detector FOV, tomography data were acquired in four overlapping segments with a 15% overlap rate to cover the entire artefact. Reconstruction was performed using the filtered back-projection (FBP) algorithm (Herman, 1980), incorporating two critical corrections: rotational axis offset and tilt correction via algorithmic calibration to avoid frequent geometric adjustments of the detector, and ring artefact correction (Wang et al., 2021) to eliminate noise induced by detector pixel non-uniformity. Following the reconstruction of the four segments, registration was conducted using redundant structural features to generate complete cross-sectional slices and a full 3D model. For qualitative–quantitative analysis, the Avizo software (Thermo Fisher Scientific, USA) was employed to support three key tasks: (i) image segmentation combining automatic thresholding for large-area corrosion, manual annotation for complex regions and deep learning assistance for microcracks; (ii) quantitative measurement including corrosion area, crack width and key dimensions of the shu dao; and (iii) 3D rendering utilizing colour-coding based on greyscale, where high greyscale values represent corrosion products or organic materials and low greyscale values denote intact iron.

3. Results and discussion

Upon completion of the 3D reconstruction, extracting a cross-sectional slice along the knife's long axis, as shown in Fig. 2(a), reveals abundant diagnostic information: the low-greyscale inner region of the blade corresponds to the uncorroded matrix, while corroded circular lesions and obvious edge rusting, both exhibiting high greyscale values, are also observed. Additionally, the outer wooden scabbard demonstrates strong neutron absorption, with identifiable fragmented structures.

Figure 2 (a) A cross-sectional slice along the knife's long axis and (b) the 3D rendering of the entire knife.

A full 3D rendering can be directly acquired from the reconstruction process, as depicted in Fig. 2(b). The next step in our work, using neutron tomography slices, involved detailed structural segmentation and rendering to characterize the artefact's morphology and clarify the extent, distribution and mechanisms of corrosion.

Specifically, the blade of the knife was segmented, and the corrosion degree was rendered using a colourmap based on the greyscale values of voxels, as illustrated in Fig. 3. The colour gradient from left to right indicates an increasing degree of corrosion; thus, the areas marked in red represent the most severe corrosion, which almost covers the entire surface of the blade.

Figure 3 Three-dimensional rendering of the knife blade from different viewing angles, where the colourmap denotes the corrosion severity.

The 3D structure shows that the remaining hilt is only approximately 2 cm in length, indicating that part of the hilt is missing due to fracture. The manufacturing process of the hilt differs from that of modern knives. Modern hilts typically feature a thin metal core wrapped in wood to form a complete grip, whereas the hilts of Han dynasty shu dao were made entirely of metal. Some designs even included cloth wrapped around the metal hilt for enhanced grip.

Subsequently, more detailed segmentation of the blade was performed, roughly dividing it into corroded regions, uncorroded iron and cracks. First, the uncorroded iron core was segmented on the basis of greyscale values. Notably, a 2.8 cm-long corrosion gap is clearly visible at the cutting edge, with a corrosion area of 1.5 cm². The midpoint of this gap is precisely aligned with the crack origin of the scabbard, as shown in Fig. 4(a) and marked by the red arrow. It is speculated that the sudden occurrence of the crack caused the lacquered scabbard to lose its effective protective function abruptly, exposing the blade to a complex corrosive environment and resulting in extremely severe corrosion in this area. For clarity, only the uncorroded iron and the transparently processed corroded parts of the blade are displayed in Fig. 4(b). The corroded surface is covered with numerous rust-induced grooves and even holes. This phenomenon is probably attributable to impurities contained in this area during the forging process, which led to the formation of such grooves on the blade during corrosion.

Figure 4 (a) Neutron cross-sectional slice, overlaid with a 3D rendering of the un­cor­roded iron and the scabbard (transparent). The slice is opaque, making the post-slice structural rendering invisible. (b) Three-dimensional rendering results of the uncorroded iron and the writing knife blade (transparent). (c) and (d) Skeletonized visualizations of the cracks relative to the knife from different viewing angles, where the colourmap represents the crack width.

Figs. 4(c) and 4(d) illustrate the cracks on the blade. In Fig. 4(c), both the blade and the uncorroded structure are rendered transparent to facilitate observation of the relative positions between the cracks and these structures. However, since the cracks are concentrated at the spine, the blade was rotated by a certain angle in Fig. 4(d) to provide a clearer view showing how the spine is covered with cracks. The most severe crack damage corresponds to the red areas in the previously mentioned colourmap, which are mainly concentrated at the junction of the hilt and the blade body. Here, the cracks were first segmented and then displayed with a skeletonized colourmap, where the colour gradient from left to right indicates the increasing width of the cracks. The actual width of the cracks ranges from 33.8 to 585.9 µm.

The corrosion-induced cracks are predominantly distributed at the interface between the uncorroded iron substrate and the corroded regions. As observed in Figs. 5(a)–5(d), although the blade is afflicted by severe corrosion with a prominent gap present [Fig. 4(b)], the corrosion products have effectively preserved the original contour of the blade, with only a limited number of such cracks detected along the cutting edge. In contrast, a closer inspection of the corrosion morphology across different tomographic slices reveals distinct visibility of these cracks along the spine. Overall, the cracks occur far more frequently on the knife's spine than on the cutting edge [Figs. 5(a)–5(d)]. This spatial distribution pattern is consistent with the crack network visualization presented in the 3D renderings of Figs. 4(c) and 4(d), which further corroborates the discrepancy in crack occurrence between the two locations. That may be attributable to the manufacturing process of the blade: intensive hammering, while effective in removing impurities, introduces micro-defects (e.g. dislocations or grain-boundary stress concentrations) that act as preferential nucleation sites for corrosion-induced cracking. The inherent stress concentration at the spine further amplifies expansion stress, ultimately leading to the formation of a crack network aligned with forging patterns. Furthermore, major cracks are localized at the junction of the blade and the hilt. This observation also indirectly suggests that, during the forging process, impurities in this region were more challenging to eliminate, which in turn contributed to the formation of a greater number of cracks in the subsequent corrosion process. The neutron tomography results confirm that these process-induced defects were the first to be exposed and severely corroded in the long-term corrosive environment.

Figure 5 (a)–(d) Four neutron slices of the knife at different positions. (e) Three-dimensional rendering of the V-shaped slot, C-shaped structures and uncorroded knife, and the relative positions of the slices through the knife. (f) Three-dimensional rendering of the V-shaped slot and the knife (transparent); the V-shaped slot is located on the cutting-edge side for fixing the knife. (g) Another view similar to (f), observed from the knife tip. (h) Three-dimensional rendering of the V-shaped slot, the C-shaped structure and the knife (transparent), viewed from the cutting-edge direction. (i) Relative positions of the detached C-shaped structure, scabbard and knife, observed from the knife tip.

When examining the tomographic slices, we observed a noticeable gap between the scabbard and the blade, and this gap would have caused the blade to shake inside the scabbard. Fortunately, a V-shaped supporting slot [marked with yellow dashed lines in Fig. 5(a)] was identified near the hilt. Although the V-shaped structure is no longer intact due to over two thousand years of corrosion, its original design purpose of supporting the blade remains clear.

Building on the above observations of the V-shaped supporting slot, the neutron tomography slices further reveal C-shaped structures distributed on both sides of the blade's cutting edge [Figs. 5(b)–(d)]. These structures exhibit high neutron attenuation, a key signal indicating that they are composed of hydrogen-rich organic material such as wood. Axial views of the slices provide additional context. The C-shaped structures were originally attached to the inner wall of the scabbard, though clear signs of detachment are visible. This detachment, paired with their positional relationship to the scabbard, suggests they were integral components of the scabbard's original layered manufacturing design, rather than later additions.

However, on the basis of corrosion traces in the tomographic slices combined with 3D segmentation and rendering results, it is observed that the C-shaped structures cover almost the entire blade, exhibiting a symmetrical bilateral distribution along the cutting-edge side while being clearly absent on the spine side [Figs. 5(e) and 5(h)]. Residual information in the slices allows an estimation of the original length of the V-shaped supporting slot at approximately 2 cm, with only a 0.9 cm segment surviving to date. It is therefore hypothesized that the C-shaped structures were intended to facilitate the installation and fixing of the V-shaped slot, and that the original design would have fitted the scabbard tightly. Beyond that, these structures also function as reinforcements: when embedded in the scabbard's layers, they enhance the scabbard's overall rigidity and durability, preventing deformation during knife insertion and extraction and extending the artefact's service life. Collectively, these features reflect the ingenuity of ancient craftsmen, who achieved a balance between structural practicality and protective requirements in scabbard design.

On the basis of the structures segmented by neutron tomography, we have also statistically analysed their volume and proportion, with the results presented in Table 1.

Table 1
Volume and proportion of each component in the shu dao

Structure		Volume (cm^3)	Percentage
Blade	Uncorroded	2.51	20.8%
	Corroded	3.74	31.0%
	Cracks	0.16	1.3%
Scabbard		4.25	35.2%
C-shaped reinforcements		1.31	10.9%
V-shaped supporting slot		0.10	0.8%

4. Conclusion

This study has successfully applied neutron tomography conducted on ERNI at the CSNS for the characterization of a corroded Han dynasty shu dao. As a non-destructive imaging technique, neutron tomography has afforded in-depth structural insights into the artefact, which remain largely inaccessible via conventional X-ray imaging techniques.

The results indicate that neutron tomography enables the effective visualization and analysis of the scabbard's layered architecture, segmentation of the knife blade's crack network and quantification of key features, such as the V-shaped supporting slot that stabilizes the blade within the scabbard. Corrosion patterns on the blade were clearly delineated, while the blade's original contour and uncorroded iron core remained distinguishable. The observation of C-shaped wooden components affixed to the scabbard, presumably a structure integrated with the V-shaped slot design, provides insights into the material composition and craftsmanship of Han dynasty lacquerware shu dao.

These findings, revealed by neutron tomography, have not been previously reported. In subsequent work, neutron diffraction could be combined with neutron tomography on ERNI to analyse the crystal phase structure of rust products without damaging the sample, so as to infer the corrosion environment.

Footnotes

‡These authors contributed equally.