Multi-Scale Measurements of Neolithic Ceramics—A Methodological Comparison of Portable Energy-Dispersive XRF, Wavelength-Dispersive XRF, and Microcomputer Tomography

: Archaeometric investigation of ancient pottery with regard to their material composites allows insights into the material structures, production techniques and manufacturing processes. The applied methods depend on the classiﬁcation of the pottery: some ﬁnds should remain unchanged for conservation reasons, other ﬁnds are less valuable or more common. While the ﬁrst group cannot be destroyed for material analyses and the choice of analytical methods is limited, the latter can be investigated using destructive methods and thus can widen the spectrum of possible devices. Multi-element analyses of portable energy-dispersive X-ray ﬂuorescence spectrometry (portable XRF) have become important for archaeological research, as portable XRF provides a quick overview about the chemical composition of potteries and can be used in non-destructive as well as destructive ways in addition to conventional microscopic examination and petrographic thin sections. While most portable XRF analyses of solely fracture surfaces do not provide satisfying results, portable XRF analyses on pulverized samples are a cost-e ﬃ cient and fast alternative to wavelength-dispersive XRF (WD-XRF). In comparison to WD-XRF, portable XRF on pulverized samples provides reliable concentration data (K, Fe, Rb, Ti, V, Y, Zn, Zr), but other elements need to be corrected. X-ray microtomography ( µ CT) has proven to be a non-destructive technique to derive not only the porosity of ancient pottery but also to characterize temper components and non-plastic inclusions. Hence, the µ CT technique has the potential to extract valuable information needed by archaeologists, for example, to deduce details about manufacturing.


Introduction
Multi-element analyses of recent decades have gained importance in archaeological and geoarchaeological research. Those analyses enable a fast acquisition of large data sets beyond complex laboratory analyses. Thus, it is possible to explore many different questions such as the provenance and use of raw materials, production techniques, interaction models of people and their objects, as well as settlement structures.
Research commonly agrees that portable energy-dispersive X-ray fluorescence spectrometry (portable XRF) provides sufficient and accurate results of chemical compositions needed for linking different assemblages of archaeological finds together. As the increasing number of ceramic studies using portable XRF shows, the potential and practical application of this method has expanded enormously in recent years [1][2][3][4][5][6][7][8][9][10][11]. However, the reliability of the chemical compositions solely gained followed shortly afterwards by the erection of the first megalithic tombs. A large number of grave goods exist in the megalithic tombs of the Funnelbeaker Westgroup, i.e., the region between the river Ijssel and the river Weser. In addition to the deposit of flint and rock artefacts as well as amber objects, the majority of grave goods are mostly pottery in various shapes and sizes. Over a period of several hundred years, hundreds of vessels were placed in the graves. Thus, the Funnelbeaker pottery offers a solid basis for interdisciplinary research due to its comparatively good preservation and large number of finds as well as its stylistic diversity. In total, 134 tombs are known in the Emsland, partly only from travel descriptions of the 19th century. Of these, 66 have been preserved completely or as remains [17,30,31]. Figure 1 summarizes the distribution of megalithic graves in the Emsland. The systematic investigation of this grave pottery began at the beginning of the 20th century with a number of excavations. Within the framework of the "Emsland Plan" in the 1950s, more excavations took place e.g., in Groß Berßen 3 to 5, 7, Emmeln 1 and 2, Lahn 4 [32][33][34][35][36][37][38] and Ostenwalde 1 [39,40].
Minerals 2020, 10, 931 3 of 25 majority of grave goods are mostly pottery in various shapes and sizes. Over a period of several hundred years, hundreds of vessels were placed in the graves. Thus, the Funnelbeaker pottery offers a solid basis for interdisciplinary research due to its comparatively good preservation and large number of finds as well as its stylistic diversity. In total, 134 tombs are known in the Emsland, partly only from travel descriptions of the 19th century. Of these, 66 have been preserved completely or as remains [17,30,31]. Figure 1 summarizes the distribution of megalithic graves in the Emsland. The systematic investigation of this grave pottery began at the beginning of the 20th century with a number of excavations. Within the framework of the "Emsland Plan" in the 1950s, more excavations took place e.g., in Groß Berßen 3 to 5, 7, Emmeln 1 and 2, Lahn 4 [32][33][34][35][36][37][38] and Ostenwalde 1 [39,40]. The well-established typological studies mainly of the 2nd part of the 20th century are now supplemented by scientific analysis used by archaeologists starting primarily in the 2010s, allowing insights into social networks, landscape development and material studies of grave goods [41][42][43][44][45][46][47].  Table 1 for more details.  Table 1 for more details. The well-established typological studies mainly of the 2nd part of the 20th century are now supplemented by scientific analysis used by archaeologists starting primarily in the 2010s, allowing insights into social networks, landscape development and material studies of grave goods [41][42][43][44][45][46][47].

Materials and Methods
In total, 19 samples from three sites of megalithic graves in northwest Germany [17] were selected for a comparison of portable XRF, WD-XRF measurements, X-ray microtomography scans and thin sections (Table 1). Additionally, two reference samples named Standard I and Standard II were generated to verify the results. All measurements were conducted on the exact same samples to gain an exact comparison between the methods used. Table 1. Ceramic samples of this study and allocation to the selected megalithic tombs Emmeln 2 (E2), Groß Berßen 7 (GB7) and Werpeloh 43 (WP43) of the district of Emsland (Lower Saxony, Germany), including information about the samples and employed analytical methods.

Optical Description: Macroscopic and Thin-Section Microscopy
In general, ceramics consist of fired clay, temper and non-plastic material. Temper material can be sand, rock fragments, shell or bone remains or broken pottery (also called grog) to name a few. Non-plastic material is organic matter, e.g., plant fibers. Depending on the used temper and non-plastic materials and their quantities, the firing quality of the ceramic can be adjusted and shrinkage or cracking of the pottery during drying and firing can be prevented. The study of temper and non-plastic materials provides constraints on the manufacturing technique and might even allow identification of the place of manufacture. Two options exist for optical description: macroscopic descriptions and thin-section microscopy.
Macroscopic descriptions are non-destructive. Hand lenses used for field geology (low power (10×) magnification) or a high-resolution binocular loupe allow easy identification of phases up tõ 2 mm in size, but limitations exist for artifacts with weathered rinds. Thin section microscopy is an abrasive method. Thin sections of the ceramics are prepared by cutting, and polishing a slice of ceramic mounted on a glass slide to a thickness of 0.03 mm. As fresh cut faces are produced and The portable XRF (portable energy dispersive X-ray fluorescence spectrometry) as an energy-dispersive handheld X-ray fluorescence technology is often used in archaeology [48][49][50][51]. It is non-destructive and reliable, requires no, or very little, sample preparation and is suitable for solid and powdered samples. The advantage of this method regarding the usage in archaeology is that either no sample or only a small part of the sample has to be destroyed (for producing powder) to determine chemical compositions. As results are provided within a few minutes, analyses of exhibits in museums or in the field are easy to obtain.
Even if samples are allowed to be pulverized, measurements in archaeology are often carried out only on fresh fracture surfaces. Handheld XRF beam spot size usually is 10 mm in diameter and can be collimated down to 3 mm X-ray beam. Measurements on fresh fracture surfaces of pottery sherds with temper phases of 5 mm and larger will thus not provide reliable average chemical composition as the heterogeneity is larger than the area exposed to the X-ray beam [12,20].
Some of the sherds in the present study are heterogeneous and contain coarse inclusions (mineral grains, fragments of rocks, bones and/or even pottery sherds) with sizes of up to 2 mm. To judge if measurements on fresh fracture surfaces might still provide suitable average chemical compositions of the entire sherds, not only were those fracture surfaces measured, aliquots of the sherds were also ground to powder. Aliquots of those powders (1) were transferred into plastic containers (outer diameter: 29 mm; inner diameter: 18 mm; height: 25 mm), sealed with a special foil (polypropylene XRF foil TF-240-255 (4.0 µm) by FLUXANA) and directly analyzed by portable XRF, (2) were at first heated to 1050 • C for 3 h to transfer all elements in their highest oxidation state [8,22], transferred into containers, sealed and subsequent analyzed by portable XRF, or (3) were processed into glass disks by fusion and analyzed by wavelength dispersive X-ray fluorescence spectrometry (WD-XRF; see below for more details). This even allowed comparison of portable XRF and WD-XRF.
In the present study, the NITON XL3t instrument of the 900 GOLDD series (ThermoFisher Scientific, Waltham, MA, USA) was used. The instrument was operated in the company's pre-set mining mode, which measures 35 main and trace elements using 4 filters. Total duration of measurement was 300 s (main filter: 40 kV, 50 mA, 60 s; high filter: 50 kV, 40 mA, 60 s; low filter: 20 kV, 100 mA, 60 s; light filter: 6 kV, 100 mA, 120 s). The analytical range was from magnesium to uranium with detection limits of main elements of~0.0025 wt.% and trace elements of 3-600 ppm (detection limits provided by the manufacturer for a SiO 2 matrix in mining mode with a duration of measurements of 60 s/filter).
All measurements were performed under constant helium flow in a laboratory with standard state conditions and a relative humidity of 40-50%. The waiting time between individual measurements was 60 s to ensure restoration of helium atmosphere surrounding the sample. Accuracy of measurements was regularly tested by the double measuring loess standard GBW07411 and metal standard IARM 35HN as unknown samples. On each fracture surface three optical homogeneous areas were measured. Powders were measured 4 to 9 times to check for precision (as also recommended by [3,12,52]). After each of those measurements, the containers were rotated to remix the powder and to ensure a homogenous mixture of the powder. Average values were calculated based on the individual analyses.
Our study revealed that the measuring accuracy depends not only on the calibration of the device but also on the age of the X-ray tube. After about 4 to 5 years the measuring quality decreases as the age of the X-ray tube causes less precise detection of some elements. Standard GBW04711 (soil standard-loess, Institute of Geophysical and Geochemical Exploration, Chinese Academy of Geological Sciences) corresponds approximately to the chemical composition of the tomb ceramics. The standard was always measured as an unknown sample before and after each set of analyses as well as during one set of analyses.
In addition, we compiled former and recent analyses of GBW04711 as unknown samples to determine any changes caused by the instrument and how those affect the results. The instrument of the Graduate School at Kiel University, Germany, was bought in 2010, and underwent maintenance only in November 2014 and January 2019. Table 2 lists literature values of GBW04711 and element concentrations of GBW04711 measured as unknown samples. Four measurement series are compared: (I) measurement series after maintenance in 2014; (II) measurement series before maintenance in 2019; (III) measurement series after maintenance in 2019 with old X-ray tube; (IV) measurement series with new X-ray tube in 2019.  Table 2 clearly demonstrates the dependence of the accuracy of some elements on the age of the X-ray tube (e.g., Ce, Ba, Mn, Ti), while others are independent (e.g., Pb, Zn, Fe, Ca, K) or can be restored by the maintenance (e.g., Zr, Sr, Si). Caution has to be taken for elements close to the detection limits, such as Nb, As, or Cr. Elements' accuracy that can be restored by the maintenance might be corrected prior to maintenance using correction factors. Wavelength-dispersive X-ray fluorescence spectrometry (WD-XRF) is a destructive and standard technique to determine bulk analyses. The higher resolution of WD-XRF compared to portable XRF provides advantages in terms of reduced spectral overlap. Complex samples can be more accurately characterized. In addition, with high-resolution backgrounds reduced, WD-XRF provides improved detection limits and sensitivity. Although samples can be measured as powder pellets as well as solid glass disks, usually glass disks are used in geoscientific applications. The glass disks are prepared using a fusion technique as this technique comprises of the reaction at a high temperature between a molten lithium borate flux and the specimen in order to develop a homogeneous molten glass that is finally cooled without crystallizing to produce an amorphous homogeneous solid glass disk.
The sherds were coarsely crushed after completion of X-ray microtomography (see below for more details) and ground to fine powder in an automatic powder mill. Further processing of the fused glass disks and subsequent WD-XRF measurements were conducted at the Institute of Mineralogy and Petrography, University Hamburg.
After determination of loss of ignition at 1050 • C, 600 mg of the remaining sample powder was thoroughly mixed with 3600 mg lithium tetraborate (Li 2 B 4 O 7 ) and glass disks were prepared using the fusion technique. The WD-XRF measurements were performed using the Panalytical MagixPRO WD-XRF device with sample changer PW 2540 VCR. The analytical range was from sodium to uranium with detection limits of main elements Na and Mg of 0.01 wt.%, others less than 0.01 wt.% and trace elements Ce of 20 ppm, Ba and Nd 10 ppm, and others less than 5 ppm.

X-ray Microtomography (µCT)
X-ray microtomography is a non-destructive technique that provides information of realistic 2D and 3D geometries of samples. Continuous image volume inspections can be undertaken by computer-aided morphometry and quantitative image analysis. It has been deployed for the characterization of true 3D microfabrics in respect to volume fractions, geometric surface, and size distribution of the fabric elements, including pores and porosity determination [54]. Because X-ray microtomography is a non-destructive technique it has a decisive advantage over XRF and thin-section petrography when investigating ancient pottery. Volume fractions of temper phases and non-plastic inclusions can be characterized and 3D visible orientation of pores can be detected and classified. This allows, for example, conclusions in respect to the manufacturing procedure of the pottery.
The X-ray microtomography scans were undertaken using the SkyScan1172 system (Bruker-SkyScan, Billerica, MA, USA). The SkyScan1172 is a tabletop unit that allows high-resolution scans. It is suitable for samples with diameters of up to 35 mm and a height of 18 mm. Oversize scans, corresponding to three individual measurements, allow scanning of samples of up to 55 mm height. All samples were scanned with a beam energy of 100 kV, a current (flux) of 100 µA and, depending on the sample, aluminum or aluminum-copper filters. A 360 • rotation with a step size of 0.5 • and medium resolution was employed in the detector, resulting in a resolution of 17.3 µm per pixel. The SkyScan software NRecon (version 1.7.3.2, Bruker-SkyScan, Billerica, MA, USA) was used for reconstruction of the spatial information of the scanned ceramics. For each sample, approximately 1000 stacked layer imageswere created, summing up to 3000 stacked layered images in oversized scan mode. Each layer image consists of isometric voxels (volumetric pixels) whose sizes depend on the detector resolution. Thus, the stacked images have volumetric information. Information about various phases that result in a different degree of absorption of X-rays is stored in grey values and decoded into black and white images during evaluation. Segmentation and binarization, i.e., the assignment of materials and gray values, image analysis and visualization, was undertaken with the SkyScan software CTAn and CTVol.
Proper characterization of pore volumes (porosities) and temper materials could be used to put constraints on the manufacturing procedure of the objects. This means that the porosities or air inclusions have to be quantified in order to define the ratio of matrix, temper particles plus non-plastic inclusions and cavities. For this purpose, a quantitative test was conducted on ink filler balls. These have a standardized diameter (2.5 ± 0.1 mm glass balls) and were filled into a plastic container of known volume. This allows precise calculation of the spaces or cavities as artificial porosities. An X-ray microtomography scan was performed on the ink filler balls while placed in the container. Figure 2a depicts the absorbed X-rays and Figure 2b is the reconstructed 3D view. All ink filler balls were made of glass, except one ball that was made of plastic and was easily recognizable in Figure 2b as a darker sphere due to the smaller density of plastic compared to glass. assignment of materials and gray values, image analysis and visualization, was undertaken with the SkyScan software CTAn and CTVol.
Proper characterization of pore volumes (porosities) and temper materials could be used to put constraints on the manufacturing procedure of the objects. This means that the porosities or air inclusions have to be quantified in order to define the ratio of matrix, temper particles plus nonplastic inclusions and cavities.
For this purpose, a quantitative test was conducted on ink filler balls. These have a standardized diameter (2.5 ± 0.1 mm glass balls) and were filled into a plastic container of known volume. This allows precise calculation of the spaces or cavities as artificial porosities. An X-ray microtomography scan was performed on the ink filler balls while placed in the container. Figure 2a depicts the absorbed X-rays and Figure 2b is the reconstructed 3D view. All ink filler balls were made of glass, except one ball that was made of plastic and was easily recognizable in Figure 2b as a darker sphere due to the smaller density of plastic compared to glass. After the scan a so-called region of interest (ROI) was generated around the ink filler balls and a 3D analysis was performed using the software CTAn. For manual calculation, the total volume of the balls was calculated by their number and diameter. The total volume of all balls ranged between 702 and 893 mm³ by taking into account the diameter's tolerance of 0.1 mm. Subtracting the total volume of all balls from the ROI volume gave the pore space (see Table 3). The software-based calculation of pore space (61%) was bracketed by the manually calculated values taking into account the varying ball diameters (64 ± 4%). Thus, porosities of the studied ceramics could be derived by the SkyScan software CTAn.  After the scan a so-called region of interest (ROI) was generated around the ink filler balls and a 3D analysis was performed using the software CTAn. For manual calculation, the total volume of the balls was calculated by their number and diameter. The total volume of all balls ranged between 702 and 893 mm 3 by taking into account the diameter's tolerance of 0.1 mm. Subtracting the total volume of all balls from the ROI volume gave the pore space (see Table 3). The software-based calculation of pore space (61%) was bracketed by the manually calculated values taking into account the varying ball diameters (64 ± 4%). Thus, porosities of the studied ceramics could be derived by the SkyScan software CTAn. Table 3. Range of pore space in container with 97 ink filler balls considering the tolerance of the diameter (2.5 ± 0.1 mm). In order to not only determine the porosity in a ceramic sample but also to determine possible components of the temper and non-plastic inclusions, two reference standards were manufactured and analysed by X-ray microtomography. Both reference standards consisted of a clay matrix with either temper material or temper material and non-plastic inclusions, which are also present in neolithic samples in approximately equal quantities. Standard I (Figure 3) included the following components: clay 10.012 g, sand 1.003 g, feldspar 0.507 g, magnetite 0.028 g, biotite 0.101 g, hornblende 0.104 g, calcite 0.197 g, bone 0.088 g, tooth fragment 0.086 g, neolithic ceramic 0.069 g. The total volume of Standard I was 5561 mm 3 . A volume of 199 mm 3 was accounted for by the temper components (biotite, hornblende, bone, calcite, tooth). In Standard II (Figure 4), mainly organic material was added: straw, culm/husks, barley grains, but also pottery fragments, tooth fragments. The aim was to identify the components in a reference sample, which could also be found in the original samples either as impressions or fragments. hornblende 0.104 g, calcite 0.197 g, bone 0.088 g, tooth fragment 0.086 g, neolithic ceramic 0.069 g. The total volume of Standard I was 5561 mm³. A volume of 199 mm³ was accounted for by the temper components (biotite, hornblende, bone, calcite, tooth). In Standard II (Figure 4), mainly organic material was added: straw, culm/husks, barley grains, but also pottery fragments, tooth fragments. The aim was to identify the components in a reference sample, which could also be found in the original samples either as impressions or fragments.

Ink Filler Ball Diameters Volume of One Ball Volume of All Balls
Pores as well as temper components or non-plastic inclusions could be easily identified in both standards. Even in Standard II with its organic components, the individual objects were clearly visible (Figure 4b,c). Based on the experience with the standards and phases therein, comparable components of temper and non-plastic inclusions in real neolithic samples could be satisfactorily identified by X-ray microtomography.

Macroscopic and Microscopic Observation
The optical observations are summarized in Table 4. Descriptions can be either macroscopic, i.e., inspection of the sherds' surface and/or fractured surface, or by thin-section microscopy [17].
The limitations of macroscopic observation of the surface are weathered rinds and adhered dirt. Robust statements regarding clay and temper material are barely possible; but ornament and decoration features can be identified.
Fractured surfaces provided at first glance impressions of used clay and temper of the ceramics, visible as matrix and fragments inside the matrix. In addition, different mineral and rock fragments could be seen, although minerals like e.g., quartz and feldspar were not easy to distinguish and the portion of feldspar might be overestimated. Further identification of rock fragments was limited by magnification of hand lenses and binocular loupe.
More detailed observation was possible by thin section microscopy albeit this being an abrasive method. Not only inorganic temper phases could be identified (i.e., mineral phases and rock Pores as well as temper components or non-plastic inclusions could be easily identified in both standards. Even in Standard II with its organic components, the individual objects were clearly visible (Figure 4b,c). Based on the experience with the standards and phases therein, comparable components of temper and non-plastic inclusions in real neolithic samples could be satisfactorily identified by X-ray microtomography.

Macroscopic and Microscopic Observation
The optical observations are summarized in Table 4. Descriptions can be either macroscopic, i.e., inspection of the sherds' surface and/or fractured surface, or by thin-section microscopy [17].

Chemical Compositions (Portable XRF vs. WD-XRF)
Chemical compositions of identical sherds measured either by WD-XRF or by portable XRF are compared in Figure 5.

Chemical Compositions (Portable XRF vs. WD-XRF)
Chemical compositions of identical sherds measured either by WD-XRF or by portable XRF are compared in Figure 5. The limitations of macroscopic observation of the surface are weathered rinds and adhered dirt. Robust statements regarding clay and temper material are barely possible; but ornament and decoration features can be identified.
Fractured surfaces provided at first glance impressions of used clay and temper of the ceramics, visible as matrix and fragments inside the matrix. In addition, different mineral and rock fragments could be seen, although minerals like e.g., quartz and feldspar were not easy to distinguish and the portion of feldspar might be overestimated. Further identification of rock fragments was limited by magnification of hand lenses and binocular loupe.
More detailed observation was possible by thin section microscopy albeit this being an abrasive method. Not only inorganic temper phases could be identified (i.e., mineral phases and rock fragments, including rock determination), also organic temper components could be detected. However, some experience was needed to distinguish various mineral phases, as more than solely quartz was most likely present as temper in addition to rock fragments and organic components.

Chemical Compositions (Portable XRF vs. WD-XRF)
Chemical compositions of identical sherds measured either by WD-XRF or by portable XRF are compared in Figure 5. The dashed lines correlate WD-XRF results with portable XRF values of the same samples. Perfect agreement between WD-XRF and portable XRF exists for those measurements that fall on the solid line, i.e., 1:1 correlation line. The symbols visualize the different portable XRF measurements. To reiterate, WD-XRF measurements were conducted on fused glass disks. Aliquots of the powder used to prepare the glass disks were measured by portable XRF (dark grey squares). During fused glass disk preparation elements transfer in their highest oxidation state. Thus, additional aliquots of the powder were heated to 1050 • C for 3 h prior to portable XRF measurements to test for any influence on XRF results if elements were existent in their highest oxidation state (open squares). As portable XRF are often used as a non-destructive technique in archaeology by analyzing fracture surfaces, the present fracture surfaces of the sherds were measured before grounding the sherds to powder for further processing (grey diamonds).
The following observations can be made:

1.
Potassium, Rb, Ti, V, Y, Zn, and Zr contents agreed well independent of technique used (WD-XRF and portable XRF) and portable XRF measurements on powder, heated powder or break planes.

2.
Portable XRF measurements resulted in slightly higher Fe concentration compared to WD-XRF, but measurements on powder, heated powder or break planes agreed well.

3.
Portable XRF measurements led to Cr values that were distinctly higher compared to WD-XRF, but again measurements on powder, heated powder or break planes agreed well.

4.
Aluminium, Ca, Mg, Si, P, and Sr measured by p-ed XRF were lower compared to WD-XRF by element-specific factors. Agreement existed between powder and heated powder measurements, but measurements of break planes disagreed with powder measurements.

5.
Niobium did not show an off-set that could be clearly related to over-or underestimated values. This might be due to the fact that Nb-contents were close to detection limits. Agreement existed again between powder and heated powder measurements, but measurements of break planes provided random contents of the respective elements. 6.
With the exception of the 3 elements (Al, Cr, and Si), the quantity of all elements measured with the portable XRF device could be corrected by multiplying the results by the relevant coefficients from the regression equations.  (Table S1).  (Table S1).
The limited reliability of the chemical analyses by applying the technique to fracture surfaces [55] was proved true. Measurements of powders provided satisfying data with regard to accuracy and precision for K, Fe, Rb, Ti, V, Y, Zn, and Zr. Iron revealed slightly higher contents in measured powder, but still acceptable values compared to WD-XRF measurements. We recommend correction for powder measurements to derive reliable concentrations for all elements except K, Rb, Ti, V, Y, Zn, and Zr, as portable XRF measurements of those elements agreed well with WD-XRF measurements. However, the values of single elements in different sherds measured with an identical and routinely calibrated portable XRF device could be easily compared and grouping of sets of sherds based on their chemical bulk compositions was possible. In addition, publishing the analysis of reference materials along with the results obtained from the samples would provide all necessary information to evaluate possible deviations in the measurements, regardless of the technique used. Figure 6 shows examples of the comparison of various temper components and non-plastic inclusions in two neolithic samples and Standard I. Bone fragments, organic material and idiomorphic mineral phases (in this case mica mineral biotite) can be precisely detected by µCT. When examining the thin sections, it was not clear whether they were cavities or non-plastic inclusions. But it is quite clear from the µCT scans that straws, present only as imprints, can be identified and are non-interchangeable with cavities due to e.g., manufacturing techniques ( Figure 6, middle).

Characterisation of Components and Porosities by 2D and 3D Sections
Minerals 2020, 10, 931 16 of 25 Figure 6 shows examples of the comparison of various temper components and non-plastic inclusions in two neolithic samples and Standard I. Bone fragments, organic material and idiomorphic mineral phases (in this case mica mineral biotite) can be precisely detected by µCT. When examining the thin sections, it was not clear whether they were cavities or non-plastic inclusions. But it is quite clear from the µCT scans that straws, present only as imprints, can be identified and are non-interchangeable with cavities due to e.g., manufacturing techniques (Figure 6, middle).  The advantage of µCT compared to thin-section analysis is not only the non-destructive analysis of the samples. Thin sections allow only a small 2D section of the sample, but µCT scans allow the entire object to be analysed. However, it is only possible to differentiate between temper components as long as the density of various components can be distinguished in grey values. The grey values The advantage of µCT compared to thin-section analysis is not only the non-destructive analysis of the samples. Thin sections allow only a small 2D section of the sample, but µCT scans allow the entire object to be analysed. However, it is only possible to differentiate between temper components as long as the density of various components can be distinguished in grey values. The grey values will then provide information about the quantity of the phases or components and the objects can be visualized in a 2D and 3D model.

No./ Name
The µCT scans of the neolithic samples provided in addition information on the porosities of the ceramic structures, or more generally of sample structures. Can porosities be used to indicate the typological classification of the finds or even the type or shape of the vessel? Table 5 displays 3D reconstructions of X-ray microtomography scans of the investigated neolithic samples, cross sections as 2D reconstructions, information about the total objects' volumes and the porosities of those volumes. Data processing was done with the SkyScan software CTAn and CTVol. Temper components as well as the various shapes of the voids are visible in the 2D cross sections.
In addition, information about the typological horizons and vessel types are given in Table 5 (see also  Table 1 for more details).     Only finds of horizons 3 to 5 were selected in this study as those provide the most precise typochronological classification of finds (see [17], for more details). Although the number of investigated finds are small (7 finds assigned to horizons and 8 finds with only known types of the vessels as those finds are stratigraphically unclassifiable), differences between the different categories of the horizons can be identified. The porosities of finds of horizon 4 (5.01, 7.33, 10.88, i.e., on average 7.74%;) are slightly higher than those of horizon 3 (5.94, 6.65, average 6.30%). Finds of horizon 5 have the smallest porosities (1.48%, 2.93%, 2.21%). The present study shows that the younger samples have a lower porosity compared to the older ones. This may be related to a change in production technology in the younger phase of TRB, whereby the vessels may have become more solid.
The collar flask has a porosity of 4.95% and the funnel beakers have a porosity of 1.81-12.20% (n = 6), on average 5.30%; the pottery of the single-grave period has a porosity of 4.30%. The comparison of funnel beakers and bowls (all from categories 3 to 5) shows similar porosities but also reflects the large variations in the porosities of the individual vessel types (funnel beaker 5.30 ± 3.91%; bowls 5.73 ± 3.06%). Based on the limited data set of our study, it is not possible to differentiate between vessel types and possible differences in production technology of manufacturing. For future studies it remains to be examined whether the use of the vessel types (e.g., beaker for liquid contents or bowls for solid contents) can be distinguished on the basis of porosity. This might indicate whether different clay mixtures have been used at the same time intentionally during a typological horizon.
Comparison of the porosities in consideration of the three sites E2, GB7 and WP43 reveal that the ceramics of GB 7 have porosities of 1.81-6.47% (n = 5), on average 4.42%, of E2 1.98-10.88% (n = 5), on average 6.55%, and WP43 1.48-12.20% (n = 5), on average 5.27% (see Table 5). The average porosities in E2 are 2.13% higher than in GB7 and 1.28% higher than in WP43. This might be an indication of the different production techniques used at the different sites. Despite the small number of samples, it is not possible to say whether the porosities indicate site-specific production methods. The observed trend requires a larger sample size to be tested in order to confirm this tendency.
Even though only 15 finds were studied in detail by µCT, and the great potential of non-destructive µCT method is obvious for distinguishing various finds regarding their typochronology, the sample number is far too small to clearly keep the sites apart and also to provide evidence for different production techniques. More samples have to be investigated to find out if there are exact differences in porosity in bowls, amphorae and cups and to provide answers to questions like "Could the different types of vessels possibly have been produced using different techniques?" or "Could this have been part of the burial rite?".

Concluding Remarks and Outlook
Ceramics can be either examined non-destructively or through destruction caused by analytical procedures. These two concepts have been explored and combined by using different methodologies.
A common procedure in archaeology is to partly destroy the objects for analysis. But often the ultimate examination is only based on thin sections and portable XRF for basic geochemical analysis. This leaves a lot of unused potential for a detailed examination of the materials. The aim of this study is to evaluate the quality and limits of measurement techniques like portable XRF, WD-XRF, and µCT and to apply constraints on their reliability and applicability for archaeological interpretations of ceramic studies in order to draw conclusions about, for example, the production technique of the finds. This was done in four steps:

1.
Destructive method by thin sections and macroscopic investigation Using the non-destructive method of macroscopic investigation, information on temper components can be obtained at first glance. But more detailed and reliable information on ceramic components with thin-section petrography as the standard method requires the partial destruction of samples.

2.
Non-destructive method of portable XRF A first impression of the geochemical structure of the sample can be gained by portable XRF at the fresh fracture, if present. However, it must be considered that the larger the grain size of the components of the ceramic, the more the geochemical data can be falsified. It turned out that the measurements of the fresh fractures did not provide satisfactorily precise results. But it is beyond question that portable XRF gives a quick overview of the chemical composition of ceramics and can be used both non-destructively and destructively. The portable analyses on pulverized samples are a cost-effective and fast alternative to wavelength-dispersive X-ray fluorescence (WD-XRF).

3.
Destructive method by pulverized samples with portable XRF and WD-XRF In comparison to portable XRF measurements, WD-XRF provides more precise data on the geochemical composition. However, WD-XRF is often not possible for cost reasons. The direct comparison of aliquots of identical pulverized samples measured with portable XRF and WD-XRF revealed reliable portable XRF concentration data of K, Fe, Rb, Ti, V, Y, Zn, and Zr, but other elements need to be corrected. Although portable XRF provides reliable data, there are limitations in the performance and evaluation of measurements. Multiple measurements of standard GBW04711 over a period of time of 5 years showed that the quality of the measurements strongly depends on the maintenance of both software and hardware as well as on the "Aging Effect" of the X-ray tube. Regular calibration of portable XRF devices is of crucial importance.

4.
Non-destructive method of µCT The non-destructive method of µCT is a suitable complement to the conventional archaeometric method for the identification of the components using thin sections. Cross sections based on µCT scans even allow 2D insights at different individually selected parts of the sherds. Quantitative information regarding modal abundances of phases and porosities of the ceramics can be obtained by µCT scans (3D). Even characterisation of the organic temper components and non-plastic inclusions is possible.
The studied neolithic ceramics showed slight differences in porosities. The younger ceramics have lower porosities compared to the older ones. With regard to the various vessel types, differentiation is not possible due to the large variation of porosities in individual vessel types. The same is true for a differentiation based on the sites. The porosities of the 15 neolithic ceramics studied disclosed neither type-specific nor site-specific manufacturing techniques. However, the great potential of the µCT technique to extract valuable information needed by archaeologists to group samples and, if applicable, to deduce details on manufacturing is obvious.