Studies of Ancient Russian Cultural Objects Using the Neutron Tomography Method

Neutron radiography and tomography is a non-destructive method that provides detailed information about the internal structure of cultural heritage objects. The differences in the neutron attenuation coefficients of constituent elements of the studied objects, as well as the application of modern mathematical algorithms to carry out three-dimensional imaging data analysis, allow one to obtain unique information about the spatial distribution of different phases, the presence of internal defects, or the degree of structural degradation inside valuable cultural objects. The results of the neutron studies of several archaeological objects related to different epochs of the Russian history are reported in order to demonstrate the opportunities provided by the neutron tomography method. The obtained 3D structural volume data, as well as the results of the corresponding data analysis, are presented.


Introduction
One of the most important tasks of archeology and other history-related sciences is the comprehensive study of cultural heritage items.The results of these studies have a significant application value, because they provide a way to penetrate into the far past and allow us to understand the formation and evolution of civilizations and ethnic groups.The experimental methods applied to characterize archeological and cultural heritage items are used to obtain detailed data regarding the phase analysis, assembly of inner parts and design items, the evolution of cracks and defects, and the traces of corrosion propagation [1].Valuable and irreplaceable archaeological items are part of the cultural heritage of nations, requiring innovative approaches to their study in order to ensure their physical preservation for future generations.It is desirable to investigate them by modern non-destructive methods [2].One of the non-destructive methods is X-ray tomography, which makes it possible to obtain three-dimensional (3D) visual data with a good precision for use both in qualitative and quantitative studies [3].However, X-ray tomography methods, including those applied using synchrotron radiation sources, have some restrictions with respect to the dimensions of the objects under study, and the worsening of the imaging data due to X-ray attenuation by metals, resulting in beam hardening.Thus, there are limits on the sample sizes that can be examined.A complementary method to X-ray tomography is neutron radiography and tomography [3][4][5][6].The fundamental difference in the nature of neutron interactions with matter compared to X-rays provides additional benefits to neutron methods, including sensitivity to light elements, a notable difference in contrast between different metals, and high penetration ability.All these features make neutron tomography an attractive tool, with a growing range of applications in industry, geophysics and paleontology [7,8].

Materials and Methods
The studied cultural heritage items were obtained from the storage collection of the Institute of Archeology of the Russian Academy of Sciences.A fragment of the metal parts (Figure 1a) of the underwater remains of bireme [12] of the king Mithridates VI Eupator was found at the National Archaeological Museum "Phanagoria", located on the shore of Taman city.This archeological item is originated during the period of suppression of the insurrection in the ancient Greek polis Phanagoria, in the first century BC.The studied fragment of the bireme remains is covered with a blue-green corrosion layer mixed with silt bottom deposits.This type of corrosion on copper items causes the formation of cuprite Cu 2 O fractions on the surface, with the subsequent gradual destruction of the initial metallic sample.The golden vial (Figure 1b) was extracted from the rich burial site of a Sarmatian woman at the "Chebotarev-5" archeological site [13] close to Rostov-on the-Don city.The well-preserved handles of the vial and the decoration on the top are visible.However, the internal volume of the closed vial remains hidden.A photo of the ancient fibulae is shown in Figure 1c.The golden ornament on the copper body of the fibulae is clearly visible, but the internal structure of the fibulae and the places of the previous restoration have been studied less.
The neutron tomography experiments were performed at the neutron radiography and tomography facility [15,16] located on beamline 14 of the IBR-2 reactor.The IBR-2 reactor is one of the most powerful pulsed neutron sources in the world, with the average power of 2 MW, a power per neutron pulse of 1850 MW, and a neutron flux in pulse of 5 10 15 n/cm 2 /s [17].The pulsed operation regime of the IBR-2 reactor is at a frequency of 5 Hz, and the long pulse duration for thermal neutrons of 350 µs makes it attractive not only for traditional neutron imaging applications, but also for the development of modern energy-selective techniques used in time-of-flight methods [18].The IBR-2 reactor provides a thermal neutron beam with wavelengths ranging from ~0.2 to 8 Å and a spectral distribution maximum of ~1.8 Å.The neutron flux at the sample position is Φ = 5.5(2) × 10 6 n/cm 2 /s.A set of neutron radiography images was collected by a CCD-based detector system with a maximum field of view of 20 cm × 20 cm.A dedicated detector module with two mirrors to suppress parasitic neutron scattering was used (Figure 2).The neutron beam dimensions restricted the upper limit of the sample size to 20 cm.The tomography experiments were performed with a rotation step of 0.5 • , and the total number of measured radiography projections was 360.The obtained imaging data were corrected by subtracting the camera dark current image, and were normalized to the image of the incident neutron beam using the ImageJ software [19].The exposure time for one projection was 10 s, and the resulting measurements lasted for 4 h for each sample.
A set of neutron radiography images was collected by a CCD-based detector system with a maximum field of view of 20 cm × 20 cm.A dedicated detector module with two mirrors to suppress parasitic neutron scattering was used (Figure 2).The neutron beam dimensions restricted the upper limit of the sample size to 20 cm.The tomography experiments were performed with a rotation step of 0.5°, and the total number of measured radiography projections was 360.The obtained imaging data were corrected by subtracting the camera dark current image, and were normalized to the image of the incident neutron beam using the ImageJ software [19].The exposure time for one projection was 10 s, and the resulting measurements lasted for 4 h for each sample.It should be noted that radioactive activation of the studied samples during the neutron tomography experiments is an urgent problem.One way of suppressing this negative effect is to reduce the exposure time by decreasing the number of corresponding angular projections.The iterative reconstruction algorithms for tomography have demonstrated promising results in the ability to compute high-quality 3D images from less data.In which case, the application of iterative algorithms like the Simultaneous Iterative Reconstruction Technique (SIRT) [20], the Simultaneous Algebraic Reconstruction Technique (SART) [21], or other more complex techniques [22] for tomography reconstruction calculations allows qualitative data to be obtained.The ASTRA Toolbox [23] is a MATLAB and Python platform providing high-performance GPU primitives for 2D and 3D A set of neutron radiography images was collected by a CCD-based detector system with a maximum field of view of 20 cm × 20 cm.A dedicated detector module with two mirrors to suppress parasitic neutron scattering was used (Figure 2).The neutron beam dimensions restricted the upper limit of the sample size to 20 cm.The tomography experiments were performed with a rotation step of 0.5°, and the total number of measured radiography projections was 360.The obtained imaging data were corrected by subtracting the camera dark current image, and were normalized to the image of the incident neutron beam using the ImageJ software [19].The exposure time for one projection was 10 s, and the resulting measurements lasted for 4 h for each sample.It should be noted that radioactive activation of the studied samples during the neutron tomography experiments is an urgent problem.One way of suppressing this negative effect is to reduce the exposure time by decreasing the number of corresponding angular projections.The iterative reconstruction algorithms for tomography have demonstrated promising results in the ability to compute high-quality 3D images from less data.In which case, the application of iterative algorithms like the Simultaneous Iterative Reconstruction Technique (SIRT) [20], the Simultaneous Algebraic Reconstruction Technique (SART) [21], or other more complex techniques [22] for tomography reconstruction calculations allows qualitative data to be obtained.The ASTRA Toolbox [23] is a MATLAB and Python platform providing high-performance GPU primitives for 2D and 3D It should be noted that radioactive activation of the studied samples during the neutron tomography experiments is an urgent problem.One way of suppressing this negative effect is to reduce the exposure time by decreasing the number of corresponding angular projections.The iterative reconstruction algorithms for tomography have demonstrated promising results in the ability to compute high-quality 3D images from less data.In which case, the application of iterative algorithms like the Simultaneous Iterative Reconstruction Technique (SIRT) [20], the Simultaneous Algebraic Reconstruction Technique (SART) [21], or other more complex techniques [22] for tomography reconstruction calculations allows qualitative data to be obtained.The ASTRA Toolbox [23] is a MATLAB and Python platform providing high-performance GPU primitives for 2D and 3D tomography, including building blocks for advanced reconstruction algorithms.One of its main design goals is geometric flexibility, allowing the toolbox to be used with many types of experimental setups.
For a brief demonstration of the possibilities of iterative reconstruction algorithms, in Figure 3 we present the comparative images of tomography slices reconstructed by means of different reconstruction algorithms.We ran 150 iterations for the iterative algorithms for neutron imaging data of the studied gold vial.The tested neutron data consisted of 380 projections with dimensions of 1311 × 1392.It seems that the use of iterative algorithms could reduce the number of projections required for tomography reconstruction; however, it would require longer times for reconstruction of large volumes.The reduction of calculation time is possible by using specialized professional graphics cards with CUDA technology [24] support.However, in order to use the ASTRA toolbox, the user requires the skills necessary to parse and process data in the MATLAB or Python layer [23,25].Moreover, the algorithms bundled in the ASTRA Toolbox are limited to reconstruction methods, and do not include typical image pre-processing, like flat field correction or rotation calculation algorithms.These disadvantages limit the audience of the ASTRA Toolbox mainly to researchers and users with an expertise in computer science.In the framework of our research of cultural heritage objects, and in the case of a full set of neutron tomography projections, the use of the conventional Filtered Backprojection (FBP) algorithm [26] seems reasonable and sufficient for our study aims.
tomography, including building blocks for advanced reconstruction algorithms.One of its main design goals is geometric flexibility, allowing the toolbox to be used with many types of experimental setups.
For a brief demonstration of the possibilities of iterative reconstruction algorithms, in Figure 3 we present the comparative images of tomography slices reconstructed by means of different reconstruction algorithms.We ran 150 iterations for the iterative algorithms for neutron imaging data of the studied gold vial.The tested neutron data consisted of 380 projections with dimensions of 1311 × 1392.It seems that the use of iterative algorithms could reduce the number of projections required for tomography reconstruction; however, it would require longer times for reconstruction of large volumes.The reduction of calculation time is possible by using specialized professional graphics cards with CUDA technology [24] support.However, in order to use the ASTRA toolbox, the user requires the skills necessary to parse and process data in the MATLAB or Python layer [23,25].Moreover, the algorithms bundled in the ASTRA Toolbox are limited to reconstruction methods, and do not include typical image pre-processing, like flat field correction or rotation center calculation algorithms.These disadvantages limit the target audience of the ASTRA Toolbox mainly to researchers and users with an expertise in computer science.In the framework of our research of cultural heritage objects, and in the case of a full set of neutron tomography projections, the use of the conventional Filtered Backprojection (FBP) algorithm [26] seems reasonable and sufficient for our study aims.The H-PITRE software [27] was used for the tomography reconstruction .An algorithm [28] for reducing ring-shaped artifacts was applied.The size of one voxel in our study is 42 µm × 42 µm × 42 µm.Each of the voxels is characterized by spatial coordinates in the reconstructed 3D volume, and a specific value of a shade of grey color.The 3D volume data of voxels are the essence of the spatial distribution of values of the neutron attenuation coefficients inside the sample volume.Attenuation of the neutron beam corresponds to scattering and absorption losses inside the material.It is accepted [4] that neutron attenuation processes are described by total scattering cross-sections as the sum of neutron scattering and absorption cross-sections, or by the attenuation length.The spatial resolution capabilities of the neutron tomography facility impose some restrictions on the minimum size of a resolved item, up to 270 microns or 0.02 mm 3 .
The VGStudio MAX 2.2 software of Volume Graphics (Heidelberg, Germany) was used for visualization and analysis of the reconstructed 3D data.The FEI Aviso 3D software was applied for the skeletonization procedure.In addition, specialized plug-ins for the ImageJ software, like Segmentation3D [29], Local Thickness [30] and Skeletonize3D [31] were used for quantitative analysis of the obtained 3D volume data.The H-PITRE software [27] was used for the tomography reconstruction .An algorithm [28] for reducing ring-shaped artifacts was applied.The size of one voxel in our study is 42 µm × 42 µm × 42 µm.
of the voxels is characterized by spatial coordinates in the reconstructed 3D volume, and a specific value of a shade of grey color.The 3D volume data of voxels are the essence of the spatial distribution of values of the neutron attenuation coefficients inside the sample volume.Attenuation of the neutron beam corresponds to scattering and absorption losses inside the material.It is accepted [4] that neutron attenuation processes are described by total scattering cross-sections as the sum of neutron scattering and absorption cross-sections, or by the attenuation length.The spatial resolution capabilities of the neutron tomography facility impose some restrictions on the minimum size of a resolved item, up to 270 microns or 0.02 mm 3 .
The VGStudio MAX 2.2 software of Volume Graphics (Heidelberg, Germany) was used for visualization and analysis of the reconstructed 3D data.The FEI Aviso 3D software was applied for the skeletonization procedure.In addition, specialized plug-ins for the ImageJ software, like Segmentation3D [29], Local Thickness [30] and Skeletonize3D [31] were used for quantitative analysis of the obtained 3D volume data.

Neutron Tomography Studies of the Fragment of the Bireme Remains
A typical example of neutron radiographic images used in the tomography process to reconstruct the internal structure of the remains fragment is shown in Figure 4a.The inner part, containing copper, contrasts well with the covered corrosion layer in the neutron radiography experiments.The attenuation of the neutron beam corresponds to the scattering and absorption losses inside the matter.The neutron attenuation processes are described by total scattering cross-sections as a sum of neutron scattering and absorption cross-sections, or by the attenuation length [4].The neutron attenuation coefficients for a neutron beam of copper are larger in comparison to the relevant parameters for silt deposits.Due to this fact, it is quite easy to recover the hidden object in the corrosion shell.The 3D models of the bireme fragment reconstructed from the neutron tomography data are presented in Figure 4b.Inside the fragment volume, a well-shaped fragment of a nail or a pin is clearly visible.The total 3D reconstructed volume of the fragment of the bireme was virtually divided into two separated volumes, corresponding to the corrosion layer volume and the volume of the copper remains.The separated 3D volume of the metal inclusion in the studied fragment volume is in Figure 4c.In accordance with the presented 3D model, the copper remains retain the shape of an ancient nail.The volumetric calculations to obtain dimensions and volumes of the nail were performed.The average volume of the observed copper remains is 1739.47(3)mm 3 .The mean height of the nail is 67.4 mm, the estimated diameter of the nail head is ~8 mm.

Neutron Tomography Studies of the Fragment of the Bireme Remains
A typical example of neutron radiographic images used in the tomography process to reconstruct the internal structure of the remains fragment is shown in Figure 4a.The inner part, containing copper, contrasts well with the covered corrosion layer in the neutron radiography experiments.The attenuation of the neutron beam corresponds to the scattering and absorption losses inside the matter.The neutron attenuation processes are described by total scattering cross-sections as a sum of neutron scattering and absorption cross-sections, or by the attenuation length [4].The neutron attenuation coefficients for a neutron beam of copper are larger in comparison to the relevant parameters for silt deposits.Due to this fact, it is quite easy to recover the hidden object in the corrosion shell.The 3D models of the bireme fragment reconstructed from the neutron tomography data are presented in Figure 4b.Inside the fragment volume, a well-shaped fragment of a nail or a pin is clearly visible.The total 3D reconstructed volume of the fragment of the bireme was virtually divided into two separated volumes, corresponding to the corrosion layer volume and the volume of the copper remains.The separated 3D volume of the metal inclusion in the studied fragment volume is shown in Figure 4c.In accordance with the presented 3D model, the copper remains retain the shape of an ancient nail.The volumetric calculations to obtain dimensions and volumes of the nail were performed.The average volume of the observed copper remains is 1739.47(3)mm 3 .The mean height of the nail is 67.4 mm, the estimated diameter of the nail head is ~8 mm.

The Neutron Studies of the Golden Vial
After the tomography reconstruction procedure, we obtained a large set of images corresponding to different transverse slices of the studied vial.Examples of the neutron radiographic images, as well as the tomography slices of the golden vial, are shown in Figure 5a,b.The grey regions correspond to the fossils of internal fills; the light areas are the gold walls of the vial.The virtual slice of the reconstructed 3D model of the vial is presented in Figure 5b.The fossilized remnants of the supposed incense are a formless piece with a volume of 3162 mm 3 .The total volume of the golden material is 615 mm 3 .

The Neutron Studies of the Golden Vial
After the tomography reconstruction procedure, we obtained a large set of images corresponding to different transverse slices of the studied vial.Examples of the neutron radiographic images, as well as the tomography slices of the golden vial, are shown in Figure 5a,b.The grey regions correspond to the fossils of internal fills; the light areas are the gold walls of the vial.The virtual slice of the reconstructed 3D model of the vial is presented in Figure 5b.The fossilized remnants of the supposed incense are a formless piece with a volume of 3162 mm 3 .The total volume of the golden material is 615 mm 3 .We applied a dedicated Local Thickness algorithm [30] for the 3D data treatment.The result of these calculations is a spatial distribution of the density of the neutron attenuation coefficients, corrected for the complex 3D shape of the golden vial.In other words, we obtained a spatial distribution of the thickness deviation of the golden wall.The obtained 3D model is presented in Figure 5d.It can be seen that the deviations in the thickness of the vessel wall are minimal, which indicates the fine work and high skills of the ancient master [13].The joint track between the two hemispherical parts of the main body of the vial is clearly visible.

The Neutron Tomography Study of the Fibulae
A typical example of the neutron radiographic images used in the tomography reconstruction process of the fibulae is shown in Figure 6a.The fragments of fasteners and clasps of the fibulae can be clearly seen.The neutron absorption coefficients for neutrons of the gold are significantly larger in comparison to the relevant parameters for copper.Therefore, the golden ornament contrasts well with the copper body in the neutron radiography experiments (Figure 6a).The virtual cut of the 3D model of the fibulae is presented in Figure 6b.The virtual cut of the 3D model of the fibulae shows the golden thread, which is deeply pressed into the thickness of the copper body of the fibulae.This fact directly indicates the use of old Scandinavian production technology [32].It can be assumed that the ancient burial place in Podbolotie was the burial place of a rich Viking warrior [33].The main difficulty was the correct segmentation of the copper base and the golden ornament.The thickness of the gold threads of the golden ornament were near the resolution limit of our neutron We applied a dedicated Local Thickness algorithm [30] for the 3D data treatment.The result of these calculations is a spatial distribution of the density of the neutron attenuation coefficients, corrected for the complex 3D shape of the golden vial.In other words, we obtained a spatial distribution of the thickness deviation of the golden wall.The obtained 3D model is presented in Figure 5d.It can be seen that the deviations in the thickness of the vessel wall are minimal, which indicates the fine work and high skills of the ancient master [13].The joint track between the two hemispherical parts of the main body of the vial is clearly visible.

The Neutron Tomography Study of the Fibulae
A typical example of the neutron radiographic images used in the tomography reconstruction process of the fibulae is shown in Figure 6a.The fragments of fasteners and clasps of the fibulae can be clearly seen.The neutron absorption coefficients for neutrons of the gold are significantly larger in comparison to the relevant parameters for copper.Therefore, the golden ornament contrasts well with the copper body in the neutron radiography experiments (Figure 6a).The virtual cut of the 3D model of the fibulae is presented in Figure 6b.The virtual cut of the 3D model of the fibulae shows the golden thread, which is deeply pressed into the thickness of the copper body of the fibulae.This fact directly indicates the use of old Scandinavian production technology [32].It can be assumed that the ancient burial place in Podbolotie was the burial place of a rich Viking warrior [33].We applied a dedicated Local Thickness algorithm [30] for the 3D data treatment.The result of these calculations is a spatial distribution of the density of the neutron attenuation coefficients, corrected for the complex 3D shape of the golden vial.In other words, we obtained a spatial distribution of the thickness deviation of the golden wall.The obtained 3D model is presented in Figure 5d.It can be seen that the deviations in the thickness of the vessel wall are minimal, which indicates the fine work and high skills of the ancient master [13].The joint track between the two hemispherical parts of the main body of the vial is clearly visible.

The Neutron Tomography Study of the Fibulae
A typical example of the neutron radiographic images used in the tomography reconstruction process of the fibulae is shown in Figure 6a.The fragments of fasteners and clasps of the fibulae can be clearly seen.The neutron absorption coefficients for neutrons of the gold are significantly larger in comparison to the relevant parameters for copper.Therefore, the golden ornament contrasts well with the copper body in the neutron radiography experiments (Figure 6a).The virtual cut of the 3D model of the fibulae is presented in Figure 6b.The virtual cut of the 3D model of the fibulae shows the golden thread, which is deeply pressed into the thickness of the copper body of the fibulae.This fact directly indicates the use of old Scandinavian production technology [32].It can be assumed that the ancient burial place in Podbolotie was the burial place of a rich Viking warrior [33].The main difficulty was the correct segmentation of the copper base and the golden ornament.The thickness of the gold threads of the golden ornament were near the resolution limit of our neutron

Figure 1 .
Figure 1.Photos of the studied cultural heritage objects: (a) the fragment of remains of ancient bireme; (b) golden vial of a Sarmatian woman; and (c) fibulae.

Figure 2 .
Figure 2. Scheme of the detector system used on the neutron radiography and tomography facility on the IBR-2 reactor.Firstly, the neutron radiation is converted into light photons by a scintillator plate.The scintillator thickness is 100 µm.The light from the scintillator is reflected by the mirrors, and then directed to the optical system of the high resolution and high sensitive CCD-based video camera.

Figure 1 .
Figure 1.Photos of the studied cultural heritage objects: (a) the fragment of remains of ancient bireme; (b) golden vial of a Sarmatian woman; and (c) fibulae.

Figure 1 .
Figure 1.Photos of the studied cultural heritage objects: (a) the fragment of remains of ancient bireme; (b) golden vial of a Sarmatian woman; and (c) fibulae.

Figure 2 .
Figure 2. Scheme of the detector system used on the neutron radiography and tomography facility on the IBR-2 reactor.Firstly, the neutron radiation is converted into light photons by a scintillator plate.The scintillator thickness is 100 µm.The light from the scintillator is reflected by the mirrors, and then directed to the optical system of the high resolution and high sensitive CCD-based video camera.

Figure 2 .
Figure 2. Scheme of the detector system used on the neutron radiography and tomography facility on the IBR-2 reactor.Firstly, the neutron radiation is converted into light photons by a scintillator plate.The scintillator thickness is 100 µm.The light from the scintillator is reflected by the mirrors, and then directed to the optical system of the high resolution and high sensitive CCD-based video camera.

Figure 3 .
Figure 3.The examples of the tomography slices of the golden vial reconstructed by the conventional FBP algorithm (a), by the SIRT reconstruction with 150 iterations (b), by the SART algorithm with 150 iterations (c).

Figure 3 .
Figure 3.The examples of the tomography slices of the golden vial reconstructed by the conventional FBP algorithm (a), by the SIRT reconstruction with 150 iterations (b), by the SART algorithm with 150 iterations (c).

Figure 4 .
Figure 4. (a) The neutron radiography image of the fragment of the bireme remains; the background is black; (b) the virtual 3D model of the fragment of the bireme remains, the copper remains are labeled in a dark yellow color, the corrosion component is grey; the corrosion components are made more transparent for clarity; (c) the separated 3D volume of the metal inclusion in the bireme fragment.

Figure 4 .
Figure 4. (a) The neutron radiography image of the fragment of the bireme remains; the background is black; (b) the virtual 3D model of the fragment of the bireme remains, the copper remains are labeled in a dark yellow color, the corrosion component is grey; the corrosion components are made more transparent for clarity; (c) the separated 3D volume of the metal inclusion in the bireme fragment.

Figure 5 .
Figure 5. (a) The neutron radiography image of the golden vial.The bright regions correspond to high neutron attenuation in the gold component.The dark areas are the low neutron attenuation regions of the fossilized incense.The background is black.(b) Virtual slices of the 3D model of the studied golden vial.(c) The 3D model of the golden vial obtained after calculations by the Local Thickness method.A color scheme whereby the density of the neutron attenuation degree extends from low (blue) to high (red) is used.

Figure 6 .
Figure 6.(a) The neutron radiography image of the fibulae.The bright regions correspond to high neutron attenuation in the golden component; the darker areas are the low neutron attenuation regions of copper.The parts of the remains of clasp on the reverse size of the fibulae are dark; (b) the axial tomography slice of the fibulae are presented as an example; the light areas correspond to gold inclusions, the orange regions are the copper body; (c) the virtual 3D model of the fibulae with separated golden ornament.

Figure 5 .
Figure 5. (a) The neutron radiography image of the golden vial.The bright regions correspond to high neutron attenuation in the gold component.The dark areas are the low neutron attenuation regions of the fossilized incense.The background is black.(b) Virtual slices of the 3D model of the studied golden vial.(c) The 3D model of the golden vial obtained after calculations by the Local Thickness method.A color scheme whereby the density of the neutron attenuation degree extends from low (blue) to high (red) is used.

Figure 5 .
Figure 5. (a) The neutron radiography image of the golden vial.The bright regions correspond to high neutron attenuation in the gold component.The dark areas are the low neutron attenuation regions of the fossilized incense.The background is black.(b) Virtual slices of the 3D model of the studied golden vial.(c) The 3D model of the golden vial obtained after calculations by the Local Thickness method.A color scheme whereby the density of the neutron attenuation degree extends from low (blue) to high (red) is used.

Figure 6 .
Figure 6.(a) The neutron radiography image of the fibulae.The bright regions correspond to high neutron attenuation in the golden component; the darker areas are the low neutron attenuation regions of copper.The parts of the remains of clasp on the reverse size of the fibulae are dark; (b) the axial tomography slice of the fibulae are presented as an example; the light areas correspond to gold inclusions, the orange regions are the copper body; (c) the virtual 3D model of the fibulae with separated golden ornament.

Figure 6 .
Figure 6.(a) The neutron radiography image of the fibulae.The bright regions correspond to high neutron attenuation in the golden component; the darker areas are the low neutron attenuation regions of copper.The parts of the remains of clasp on the reverse size of the fibulae are dark; (b) the axial tomography slice of the fibulae are presented as an example; the light areas correspond to gold inclusions, the orange regions are the copper body; (c) the virtual 3D model of the fibulae with separated golden ornament.