Conductive cross-section preparation of non-conductive painting micro-samples for SEM analysis

Scanning electron microscopy (SEM) is a common method for the analysis of painting micro-samples. The high resolution of this technique offers precise surface analysis and can be coupled with an energy-dispersive spectrometer for the acquisition of the elemental composition. For light microscopy and SEM analysis, the painting micro-samples are commonly prepared as cross-sections, where the micro-sample positioned on the side is embedded in a resin. Therefore, the sequence of its layers is exposed after the cross-section is polished. In common cases outside of cultural heritage, a conductive layer is applied on the polished side, but in this field, the measurements are mostly done in low-vacuum SEM (LV-SEM). Although the charging effect is reduced in LV-SEM, it can still occur, and can hardly be prevented even with carbon tape or paint. This work presents two conductive cross-section preparation methods for non-conductive samples, which reduce charging effects without impairing the sample integrity.

www.nature.com/scientificreports/ evaporation or molecular modifications of the organic components often present in painting samples, but also thermal chemical reactions between some components, thus, morphological changes. There are different types of resin encapsulations 62,63,65,66 , as well as sample fixation methods, which are discussed elsewhere 67 .
In this work, the aim was to develop a resin cross-section preparation for non-conductive samples to reduce the charging effect without impairing the sample characteristics (colour, morphology, chemical composition) with any coating or coloured resin. The suggested preparation procedures are a combination of established methods of painting micro-sample preparation, and the procedures aim to be time, cost and results effective.

Materials and methods
Samples. The chosen samples (Table 1) represent a fully non-conductive material (MS1, MS2) and a partially conductive one (MS3, RS1, RS2) ( Table 1). MS1 and MS2 are model samples made of pure chalk and an organic binder. MS3 is a partially conductive model sample. A restorer prepared the model samples to improve the comparison with historical painting micro-samples. MS3 contains a layer of yellow earth mixed with an egg yolk (tempera technique) with iron particles which make it partially conductive on top of chalk, and a rabbit skin glue non-conductive layer.
RS1 is a real historical sample with a top green earth tempera layer and a chalk preparatory layer with an unknown oil binder. RS2 is also a real sample of a wall painting, with green and yellow earth layers on top of two carbonaceous preparatory layers. All samples are smaller than 1 mm in all directions.
Epoxy resins. The materials used for this study are relatively easy to find and established for a cross-section preparation. Various types of epoxy resins were used for the preparation 55,71 . EpoFix (Struers, DK) was used for the direct embedding of all samples and the reference non-conductive preparation. It is a transparent dual-component cold mounting resin combining an epoxy resin with a hardener. Its curing time is approximately 12 h. It is a relatively common type of resin already used for painting micro-samples embedding 64 .
PolyFast and LevoFast (Struers, DK) are hot mounting powder resins. PolyFast is a conductive resin used for the Preparation A, and LevoFast is a particularly hard resin used for the polishing holder designed for this work.
The Reference and Preparation A were prepared in a round plastic mould (Struers, DK), which is made for a vacuum impregnation chamber (CitoVac, Struers, DK) to help remove bubbles to increase the stability and hardness of the resin. To fix the positioning of the sample, plastic clips were used.
For the other preparations, two silicone ice cube trays of 1 cm × 1 cm × 1 cm and 1 cm × 1.5 cm × 1.5 cm were tested. The two trays had different silicone textures. The small one had a porous texture, giving a final milky appearance of the resin, while the large tray gave a transparent final aspect. The small cubes better fitted the size of our micro-samples and made the positioning easier. The larger cubes were easier to handle and could be useful for large or long samples. Even though the transparency of the cubes from the larger tray made it easier to observe the samples in the resin, the smaller silicon tray was harder and, thus, easier to handle, and the milky appearance was not hindering the polishing process. In our case, the smaller ice cube tray of 1 cm × 1 cm × 1 cm for the conductive preparation was more appropriate.
Instrumentation. The CitoPress-10 hot mounting press (Struers, DK) was used for the PolyFast and Levo-Fast resins.
A Leica EM ACE 600 high-vacuum coater was used for the application of a carbon coating of 30 nm in Preparation B.
The grinding and polishing of the cross-sections were done as described in Jaques and Zikmundová (unpublished). Before the sample was reached, the polishing mode was changed from wet to dry 47 .
The first observations of the cross-sections were made under the Stemi 2000-C and Stemi 508 stereomicroscopes (Zeiss, DE) and a Reichert microscope (Reichert Technologies, US) coupled to an Axiocam ERc 5 s (Zeiss, DE).
The scanning electron microscope used for this study is a MIRA3 XMU (Tescan, CZ). Both the high-vacuum (5 mbar to 9 × 10-5 mbar) mode and the low-vacuum (0.07 mbar to 5 mbar) mode with a secondary electron detector (LVSTD mode) were used.
Cross-section preparation. The non-conductive cross-sections were prepared with the 5 samples (Table 1) as references ( Table 2). Then two conductive preparations (A and B) were developed ( Table 2). www.nature.com/scientificreports/ Reference preparations. The reference preparations ( Fig. 2) are non-conductive, except for the specific conductive particles of the sample (MS3, RS1, RS2). The capsule reference preparation was prepared by positioning the samples and then adding the resin through the CitoVac device. The CitoVac is used to pour the resin via a plastic tube onto the micro-sample placed in a circular mould in a vacuum chamber. This technique offers cured resin with less to no bubbles and better penetration of the resin in the sample. The capsule embedding had a curing time of approximately 12 h regarding the dual-component epoxy. 12.5 g of the EpoFix per 1.5 g of the hardener were used and gently poured on top of the sample. The samples were fixed using plastic clips, which, however, moved, and had to be repositioned. A second reference preparation was made using a flat embedding with the same samples. The mould was 1.5 cm × 1.5 cm × 1.5 cm to ease manual polishing without the polishing holder. Depending on the size of the sample, it can also be useful to have a larger mould size. The sample was positioned on the surface of the already dried resin filling half of the mould. A drop of the EpoFix was applied on top of the sample and left for approximately 30 min. This lightly fixed the sample and avoided drifting or floating. Then the rest of the resin was gently poured on top. The CitoVac was not used in this case. The flat embedding gave better results for the orientation, even though its preparation is slightly more demanding. After the curing time, the embedded samples were polished with the Tegramin in an automatic mode.

Conductive preparation A.
Slotted-capsule embedding was chosen because the two-step preparation (hot resin then cold resin) created with small holes helps the sample positioning and partially avoids floating and drifting. The preparation using the CitoPress was done according to its instructions. This step took approximately 10 min. Four flat bottom holes of 5 mm in diameter and 2 mm deep were drilled in the PolyFast block ( Fig. 3C). If the samples have a similar height (perpendicular to the bottom of the hole), more can be put in the same block, but it can make it more difficult to get an interesting surface for all of them while polishing. Therefore, a maximum of two samples of the same height in one resin block is recommended. A batch of blocks can be prepared in advance, either with or without the holes. The depth of the holes can also vary according to the size  www.nature.com/scientificreports/ of the samples. The copper tape was added to the bottom and around the hole. One sample per hole was fixed and put into tight contact with the copper tape (Fig. 3B). The hole was finally filled with the EpoFix (Fig. 3B), and the curing was shorter due to the small amount of the resin. The cross-sections were then polished with the Tegramin.

Conductive preparation B.
For the flat-embedding preparation, half of the silicone mould of 1 cm × 1 cm × 1 cm was filled with the EpoFix and left to cure similarly to the flat reference preparation (Fig. 4A). The surface of the lower part of the resin was coated with a 30 nm carbon layer with a Leica coating device (Fig. 4C). The sample was then put on top with a tiny drop of resin or a quick-dry gel glue (Fig. 4D). A high viscosity material, such as the gel glue, prevents its spreading on the resin surface and in/on the sample. Only a small part of the bottom of the sample should be glued. The glue should not interfere between the coating layer and the sample. After a short curing time of the glue, a second carbon coating is done on top of the preparation (Fig. 4E). Finally, the second layer of the resin is added to fill the mould and fix the sample (Fig. 4F). A holder specifically designed for this experiment was used for polishing the cross-sections. The first polishing was wet (using water) and then just before getting to the sample, the mode was switched to a dry mode to prevent the disaggregation of the sample components. The pressure and speed of the rotation plate were also reduced to avoid deep scratches on the surface of the resin and its melting.
Polishing holder. Before the analysis, cross-sections must be polished to get a clear, smooth surface of the sample without the resin on top of it. Although round moulds are typical for cross-section preparation, they are not always of the best size or shape for the sample. The smaller cubes were chosen, which are common in cultural  www.nature.com/scientificreports/ heritage sample preparation, with a manual polishing. The Tegramin automatic mode cannot be used for such cubes because of their shape and size that does not fit the device. The cubes are also rather small, which makes it more difficult to get a flat, smooth polished surface and avoid deep scratches. Therefore, it was decided to use the Tegramin for a certain standardisation of the preparation and to avoid user errors, such as uneven surfaces and deep scratches. A polishing holder was designed to fit the cubes into the Tegramin (Fig. 5A).
The holder was made of the LevoFast resin, composed of melamine, minerals, and glass fillers. It is a harder resin than EpoFix used for embedding and, thus, more resistant to polishing. This resin is hot mounting and was prepared using the CitoPress.
A square of 1 cm × 1 cm × 0.75 cm was drilled in the centre of the round piece, and a square silicone rubber with walls of a 2 mm thickness was introduced into it (Fig. 5). The silicone rubber fixes the cross-section, allowing only small movements of it. Also, the resin can shrink during curing, preventing the final shape to be a perfect cube, hence the silicon rubber.

Results and discussion
The preparation procedure of each method was evaluated according to the total time of the preparation, the ease of the preparation, and the visual assessment of the surface under the light and electron microscopes. The conductivity of the cross-sections prepared according to the Preparation A or B (Table 2) was evaluated in a visual comparison with the reference preparation cross-sections of a similar sample. The image quality based on the amount of charging and strength of the artefacts was evaluated during the imaging process.
Cross-section preparation. Preparation A (Fig. 3) needs the hot resin to be prepared in advance (appr. 8 min) and the holes to be drilled in the resin (appr. 5 min), but the curing time of the transparent EpoFix is reduced from 12 h to appr. 6 h thanks to the small amount needed. The hole helps positioning the sample with copper tape and a low amount of resin, which strongly reduces the floating and drifting of the sample compared to the reference preparation. The capsule embedding (Fig. 2) offers a one-step preparation without previous planning, but the amount of the curing resin increases the risk of sample misplacement even with plastic clips. These clips are useful for flat and rather large samples, but they reduce sample visibility and cannot be used for brittle, non-flat samples. Highly porous samples floated to the surface despite multiple repositioning during the hardening of the resin (12 h). Without the clips or any other fixing, the samples floated (Fig. 2D-grey arrow + Fig. 6). One of the reference samples drifted into a horizontal position (Fig. 6), instead of staying on its side to show the layering. In our case, the sample only contained a chalk layer, but with any other sample, most stratigraphic information would have been lost.
Preparation A reduces the manipulation of the sample (Fig. 7). The diameter and depth of the hole can be adapted to the size and shape of the sample. In our case, a diameter of 5 mm and a depth of 2 mm were used (Fig. 7A). However, the diameter was too large for our samples, which were between 1 and 3 mm long (Fig. 7B). The chosen depth was relatively advantageous for all samples.
Under the microscope, scratches were visible on the surface of the cross-sections A. There was no visible difference in the depth of the scratches between the reference cross-sections and cross-sections A. These scratches occur due to the combination of coarse polishing particles and the pressure applied during the polishing steps (Fig. 6A). It could be visually attenuated by applying a drop of ethanol or water and a coverslip on top (Fig. 6B), but this technique is not recommended, as it can dissolve some components of the sample.
Preparation A has a quick curing time, but includes preparatory steps that need to be planned relatively in advance (drilling, two rounds of curing). It also needs two different resins (PolyFast, EpoFix). The thickness of the transparent resin can be measured (Fig. 7A) and used to set the parameters of the automatic polishing to protect the sample from being destroyed during the process. It helps change the polishing mode from wet to dry in time. The holder used to secure the cross-section while polishing them reduces uneven surface, deep scratches, and offers the possibility to set in the automatic mode the polishing time and pressure in the Tegramin. For manual polishing, the size of the holder offers a better hold of the cross-section for the users, which improves the polishing results. The main advantage of Preparation A is the possibility of preparing more samples simultaneously. In www.nature.com/scientificreports/ the case of preparing multiple samples, the amount of the PolyFast resin is reduced compared to a single sample preparation. Moreover, with all samples polished into a uniform height, there is no need to adapt the focus to each cross-section separately, which reduces the working time. Hence, one cross-section containing several microsamples can reduce the preparation time, measurement time, and amount of material (i.e. embedding medium). Preparation B (Fig. 4) also requires the first half of the EpoFix resin to be cured in advance and its surface coated with carbon (appr. 6-8 h). Both steps can be carried out in a batch and stored for later use. This preparation does not need any additional materials or procedures. Fixing the sample is the most sensitive part of the procedure, and it strongly influences the conductivity of the preparation, as the fixing medium acts as an insulator. The second coating needs to be connected to the first coating layer, and this carbon network should be in contact with the sample all around, but also with the SEM metallic holder to create an appropriate exit path for the electrons to avoid charging. The cross-section polishing becomes quicker and easier with the holder and the Tegramin. The pressure control of the Tegramin can also help reduce the scratches.
The capsule-embedding reference preparation was easy, quick, and cheap, and had only 2 steps (sample positioning, resin curing). The main issues of this preparation were the sample drifting during curing, the long curing time because of the relatively large amount of the resin, and the charging of the cross-section under the SEM beam. With the flat-embedding preparation, which was used for Preparation B, the positioning was easier,   Fig. 8C and 8F, a high over-exposition (electron concentration) of the materials can be observed, particularly of the fibres (organic), but also a darkening of the inorganic material rims, creating a contrast imbalance. The uneven brightness (Fig. 1A) of the material creates visual artefacts and hinders proper observation of the sample. From Fig. 8A-C and at higher magnification from 8D to 8F, a contrast gradation from homogeneous to inhomogeneous can be seen. The reference preparation showed high charging (2-3) at 10 kV, but stabilised from 20 kV (2-1). A charging modification depending on the sample morphology occurred around 30 Pa. The homogeneous samples (MS1/2) had stable charging effects on their surface compared to the multi-layered ones (MS3, RS1/2 Fig. 8A), where the charging was high in localised areas (Fig. 8D-around organic fibres). The charging was reduced from 3 to 2, but was still present at any magnification (low 500 x; average 500 × to 3 kx; high 3 kx), and with any parameters and samples (1).
In Preparation A (Fig. 3), the sample is in direct contact with a conductive material. The copper tape (Cu) (Fig. 9A + 9B) is connected to the conductive resin, creating an exit path for the electrons. At 10 kV | 10 Pa and 15 Pa, the charging in the middle of the non-conductive resin was very high (3) at low magnification and slightly lower (2) at an average magnification. Almost no charging occurred at 10 kV | 50 Pa, even at a high magnification (1-0) (Fig. 9C), whereas in the reference preparation 0 charging effect was never reached. Both the homogeneous and the multi-layered samples had a charging decrease, 2 to 1, at 10 kV | 20 Pa (Fig. 9D). This preparation showed promising results, where the charging occurred only in the transparent resin area, but not around the sample or the copper tape. At a high magnification, the charging was reduced only where the sample was in direct contact www.nature.com/scientificreports/ with the conductive material (Fig. 9D). Even though, at a high magnification, the charging of the chalk ground decreases to a certain extent due to a vacuum increase (Fig. 9E + 9F). The sample needs to be in tight contact with the conductive component to effectively reduce the charging. Preparation B (Fig. 4) was designed based on the observations made during Preparation A. The use of carbon sputtering for the coating fits the shape of the sample and surrounds it better than the copper tape (Fig. 10A). The direct contact between the conductive material and the sample was expected to be better than in the Preparation A. The coated surface of the resin (Fig. 4C) works as the exit path for the electrons from the cross-section to the metallic SEM holder. Similarly, in Preparation A, the charging was reduced where the sample was in direct contact with it. But the carbon layer was not even (Fig. 10B) and was disrupted during the sputtering mostly around the samples' edges and vertical surfaces. During the carbon sputtering, the vacuum can be chosen, as well as the theoretical thickness of the carbon layer. In our case, 30 nm thickness (relatively thick) was chosen for a low vacuum (around 2 Pa). With a lower vacuum, the sputtering was less precise and less even (Fig. 10C). The main problem of this preparation was the glue spreading between the sample and the carbon layer. Therefore, the sample was mostly not in contact with the first carbon layer anymore. The second coating disruption occurred due to the direction of the sputtering (the stage of the coating device can be tilted to improve the carbon deposition depending on the sample shape). Only one side of the sample was coated, and due to its uneven surface and morphology, the connection between the two carbon layers was not as good as expected. Consequently, the charging was not appropriately reduced.
The preparations of the cross-sections A and B have several advantages over the reference cross-sections, but there is still room for improvement in both techniques concerning the charging decrease. Regarding crosssections A, the positioning of the sample is easier with the copper tape and only a small area of a liquid resin. The extent of the sample manipulation is greatly reduced due to the holes drilled in the conductive resin. The low amount of the resin reduces the formation of bubbles, and the copper tape at the bottom of the sample does not interfere with its analysis. The curing time is also quicker in Preparation A, because of the small amount of www.nature.com/scientificreports/ the transparent resin needed. The copper tape does not fit the morphology of all samples well, increasing the risk of breaking the sample, to fit it better for enough contact to avoid charging effects. In our case, the holes were drilled, but they could have been melted with a soldering iron tip, which might be easier to use in a laboratory and enable to shape the hole with regards to the sample morphology. This would reduce the use of the resin, the size of the hole, and therefore the charging, as the conductive layer would be much closer to the sample. Regarding cross-sections B, the positioning and curing time are similar to the reference flat-embedding. In both cases, if the first parts (curing of the first part of the resin, carbon coating of its surface) of the cross-sections are prepared in advance as a batch, the time of the preparation of the cross-section is cut in half. Glue was used to fix the samples on the first cured part of the resin, but it limited the contact with the coated surface and partially neutralised the effect of the charging reduction. Based on our findings, the use of carbon tape or carbon paint instead of normal glue to fix the sample in step D in Fig. 4 is probably a better choice and needs to be tested. This helps positioning the sample and decreasing the charging, as well as connecting the conductive parts around it. The cross-section B showed a slight reduction of charging where the sample was in contact with the carbon coating, but the connection between the two layers of carbon was not consistent around the sample, as already stated. This preparation needs further improvements regarding the carbon coating connection, but based on our observations, it should show better results in decreasing the charging and easing the positioning of the sample in fewer preparation steps. Cross-section A is currently the most advanced and gives the best results from the three preparation methods concerning the charging reduction, the sample positioning, and the preparation time. For all cross-sections, the best vacuum and voltage combination for the lowest charging effect was 50 Pa and 10 kV, respectively, regardless of the sample. The reference preparation is common, but achieving a 0-charging effect with such preparation can be tricky without risking the sample integrity. The cross-sections A and B showed a charging decrease, which could lead to a 0-charging effect with the small improvements proposed in this work, without impact on the sample surface to be analysed.

Conclusion
Different methods of cross-section preparation of valuable painting micro-samples were explored in this work. Such cross-sections should not be covered with a coating that hinders light-microscopic observation and can hardly be removed. The addition of carbon paint on top of the cross-section works well around the area where the paint touches the sample, or when the sample includes a conductive paint layer. This work aimed to improve the preparation of non-conductive samples and to decrease the charging with common laboratory materials and devices, which offer easy reproducibility of these preparations. Preparation A (conductive resin + copper tape + transparent resin) enabled a significant charging decrease, but could not fully remove it. Preparation B (transparent resin + 2 side coatings) has a bigger potential to remove charging, but needs further testing and improvements, such as a better connection of the carbon layers. The preparation process can be further improved, but this study proved that the preparation of a conductive cross-section for non-conductive sensitive samples works without altering their surface characteristics, morphology or material composition for LM and SEM analyses.