Portable XRF on Prehistoric Bronze Artefacts: Limitations and Use for the Detection of Bronze Age Metal Workshops

Abstract Two different scientific analyses-one destructive and one non-destructive-were conducted on two separate groups of bronze ornaments dating from 1500-1100 BC to investigate, amongst other traits, the metal composition of their copper-tin alloys. One group of artefacts was sampled, and polished thin sections were analysed using a scanning electron microscope (SEM). Results from the corrosion crust of copper-tin alloys, and the change measured within the elemental composition from the bulk metal to the surface, greatly influenced the interpretation of the second data set, which was measured using a handheld X-ray fluorescence (XRF) device. The surface of corroded bronze ornaments consists mostly of copper carbonates, oxides, and chlorides. Chemical processes, such as decuprification, change the element composition in such a manner that the original alloy cannot be traced with a non-destructive method. This paper compares the results of both investigations in order to define the possibilities and limits of non-destructive XRF analyses of corroded bronze artefacts.


Introduction
Defining metal workshops using scientific methods-for instance, through analysis of trace elements or isotopic signatures-is a common matter of interest in archaeometallurgical studies. However, since the middle of the nineteenth century, the impact of chemical analysis in provenancing archaeological objects was discussed with a focus on the definition of workshops (for a detailed research history, see Pernicka 2014, Stos-Gale & Gale 2009).
Behind such an approach is the assumption that the metalworker was well aware of the properties of the metal in use and intentionally chose that best suited to the object's purpose. The colour, smell, and weight of the metal, as well as its reaction during smelting, would have given the craftsperson all the information needed to successfully produce the desired copper alloy. This can also be inferred for bronze and, as such, the already alloyed tin-bronze would have been specifically chosen by the craftsperson to meet further need. To support such a statement, however, a detailed trace element analysis of artefacts produced within one defined workshop is necessary.
Within a FP7-financed project, several metal workshops were detected from the early and middle Nordic Bronze Age-namely in southern Scandinavia, Denmark, and North Germany (see Nørgaard 2014aNørgaard , 2014bNørgaard , 2015bNørgaard , 2015c. Based on traces visible on the material culture, such as tool marks, residues of the crafting (technical traces), and relics of the craftsperson's habitus, eleven NBA II (1500-1350 BC) workshops were defined in Denmark. Five further workshops were detected in Lower Saxony, including four NBA III (1350( -1100 workshops in Mecklenburg, with evidence pointing towards a continuation of the majority of the NBA II workshops. This project raised the possibility of testing for intentional alloying, and thus identifying a specific craftsman through a signature alloy. The material from Mecklenburg (NBA III) offered a potentially insightful case study, as the technical defined workshops, while located within a narrow region, clearly could be separated due to their characteristics (Nørgaard 2015a). Furthermore, the availability of the material promised a robust data set. However, given the tendency within modern archaeology to emphasise the protection and preservation of artefacts, solely a non-destructive analytical approach was allowed.
Non-destructive methods are becoming increasingly important within modern archaeological research. These methods include analysis by handheld XRF spectrometer (Anker 1982, Helfert & Böhme 2010, Wolff 2009). Advancements in the scientific analysis of prehistoric objects, without the risk of damage, promise researchers access to previously hidden information. Furthermore, the increasing availability of portable XRF devices, as well as their moderate cost when compared to laboratory-based technologies, more frequently allows the incorporation of chemical examinations in material culture studies. However, such analyses have their limits and not all questions can be answered in this manner.
One major difficulty in the non-destructive analysis of bronze artefacts in museum collections or in the field is the fact that the majority of these objects are corroded. Several studies have concentrated on the analysis of corrosive structures on copper alloy objects (e.g. Bernard & Joiret 2009, Chase 1994, Constantinides, Adrianes, & Adams 2002, Mircea, Sandu, Vasilache, & Sandu 2012, Oudbashi, Emami, Ahmadi, & Davami 2013, Robbiola, Blengino, & Fiaud 1998, Wadsak et al. 2000, while only a small number have discussed the analysis of corroded artefacts using handheld XRF devices (Fernandes, van Os, & Huisman 2013, Šatović, Desnica, & Fazinić 2013. Unfortunately, these studies are typically published within the domain of conservation studies and seldom find their way to archaeologists. Nonetheless, within archaeological literature, the difficulties involved in using superficial analytical methods to determine the composition of copper alloys in ancient artefacts have been mentioned and researchers have also noted the fact that those values measured superficially using these methods give little insight about the actual alloy (e.g. Härke 1978, Lutz & Pernicka 1996, Otto 1973, Pernicka 2010. However, this topic is rarely discussed at any length (exceptions i.e. Pernicka 2010, 720-724), perhaps permitting archaeology's continued assumption that an analysis of the surface of corroded bronze artefacts will allow one to determine the composition of the underlying alloy. The following case study will demonstrate not only the degree of the advancement of corrosion on a select set of Northern European bronze artefacts, but also the value to be gained by the nondestructive analysis of such objects. As such, this paper contributes to a wider debate regarding the limits of non-destructive analysis using portable XRF technology. It will discuss the kinds of questions that could be answered using handheld XRF and whether a surface analysis of artefacts will allow the determination of workshops via the trace element composition.

Material and Method
To illustrate the limits of non-destructive methods such as XRF, outlined above, this study describes two differently-executed analyses of corroded bronze ornaments originating within and dated to the Nordic Bronze Age (1500-1100 BC). The material considered here consists of belt plates and neck collars from NBA II sites (1500-1300 BC) in Lower Saxony (Nørgaard 2015b) and from NBA III sites (1300-1100 BC) in Mecklenburg, Germany (Nørgaard 2015a). All artefacts examined for this study are made from bronze, a copper-tin alloy, and form part of existing museum collections or stores. The majority of objects within the sample were acquired at the beginning of the twentieth century; thus, many of the bronzes are coated with Brought to you by | Aarhus University Library / Statsbiblioteket Authenticated Download Date | 4/11/18 2:00 PM a heavy corrosion layer, and sometimes even with a layer of varnish or paint. They represent the normal spectra of bronze artefacts a researcher will find while investigating the Bronze Age in northern Europe.
As this paper focuses on the possibilities and limits of non-destructive XRF-analysis in the study of bronze artefacts, rather than the craft of metalworking itself1, the differences within the analytical methods of both object groups and their potential regarding the leading question is the main objective. To this end, this study considers primarily the corrosion of the examined artefacts and its effect on the chosen analytical method.
An earlier study details the metallographic analysis of the NBA II ornaments from Lower Saxony considered here (Nørgaard 2015b); the results of that investigation made it possible to reconstruct the detailed operational sequence of the crafting process (Nørgaard 2014a, 680ff.). During this earlier analysis at the German Mining Museum in Bochum, it was possible to perform a more extensive investigation of the specimen, due to a fully exposed cross-section. With the objective of defining the corrosion layers of the bronze artefacts, as well as the change in the alloy composition towards the objects' surface, the samples were examined under a scanning electron microscope (SEM). The specimen samples were placed within the large sample chamber (no restricting airlock) of the field emission scanning electron microscope (SUPRA 40 VP from Zeiss), located in the museum's material science laboratory. The SEM is equipped with an energy-dispersive x-ray spectrometer (Noran System 7, Thermo) with a silicon drift detector (SDD). In order to execute further metallographic investigations under the light microscope, the sand-polished specimen samples were placed within a Bakelite ring and labeled on the back (Nørgaard 2015b). The results from the scanning process allowed the author to determine the alloy compositions of the artefact and the changes within the composition towards the corrosion and within the corrosion layer. To offer a detailed characterisation of the changes within the alloy composition towards the corroded surface areas, a series of element distribution curves, measured from the centre of the sample to the edge of the corrosion layer, was created, as well as element distribution charts (elemental maps).
The second group of artefacts-those from NBA III sites in Mecklenburg-were visually analysed with an eye to studying the remaining traces left from the crafting process. The result is a superficial knowledge of the traces left by different tools, techniques and habits of the craftsperson during the production process (i.e. Nørgaard 2015c). When this knowledge is compared to the deformations, or non-deformation, of the metal´s microstructure as is documented within the metallographic examination, it is possible to create a detailed operational sequence of the artefacts' production. In addition, several scientific investigations were executed in order to answer the initial question posed at the start of this discussion as to the nature of specific individual alloys. The Archäologisches Landesmuseum Mecklenburg-Vorpommern restricts examination of its collection to non-destructive methods; thus, the analysis of those objects in the sample from the Mecklenburg region was executed using a handheld energy-dispersive XRF spectrometer (Niton Xl3t GOLDD) to reveal their trace element composition. A handheld XRF spectrometer measures and analyses the metal composition of an artefact near the surface. It is based on the principle that an X-ray photon, emitted from an X-ray tube, removes an electron from the inner shell of an atom (for a more detailed description of the process, see Anker 1982, Helfert & Böhme 2010, Wolff 2009). Thus, it measures the elemental composition of the sample's surface. As was indicated in the earlier case study, the results are strongly influenced by the effects of corrosion on the distribution of elements at an object's surface.

Results and Discussion
Most of the Scandinavian and north German bronzes considered here were found in agricultural areas or bogs, and, due to a similar climate and agricultural practices, have been exposed to similar conditions. Therefore, the corrosion documented through the metallographic investigation of the bronze ornaments from Lower Saxony is comparable to that visible on the ornaments from Mecklenburg. As such, the destructive (metallographic) analysis of the artefacts from Lower Saxony can be used to understand the results from the non-destructive analysis of the artefacts from Mecklenburg.

Some Specific Information on Corrosion
The majority of bronzes dated to 1500-1300 BC are Cu-Sn alloys (tin bronzes), which, when exposed to moist air under pre-industrial conditions, develops a patina consisting largely of basic copper (CuCO 3 ). These CU (II) salts can differ according to the deposition context of the bronze artefacts, e.g. malachite (Cu 2 (CO 3 )(OH) 2 ) when corroded in soil, brochantite (Cu 4 (SO 4 )(OH) 6 ) when exposed to the atmosphere, or atacamite (Cu 2 Cl(OH) 3 ) when placed in seawater. They commonly cover a red cuprous oxide (Cu 2 O; SnO 2 ) layer (Bernard & Joiret 2009, 5200, Constantinides et al. 2002, Robbiola et al. 1998, 2084. Therefore, the monoclinic emerald-green to blackish carbonate and the glaze blue azurite, which develops into the familiar greenish malachite (Mircea et al. 2012, 180, Wolters 1991, dominates the appearance of the bronze artefacts. Consequently, the appearance of carbon (C) and oxygen (O) as primary and secondary chemical compounds is of particular importance for the understanding of the corrosion layer. In addition to oxygen, chlorine (Cl) is also detected in almost all corrosion layers. As Oudbashi and colleagues (2013) point out, the presence of chlorine in the burial environment of an artefact causes a reaction with its copper and a formation of cuprous chloride. In the worst case, such exposure can lead to what is known as 'bronze disease'-a 'reaction of copper with chlorine in presence of moisture and oxygen that causes the formation of copper trihydroxichlorides … until all metallic copper converts' (Oudbashi et al. 2013, 8).
Furthermore, minor elements contained in the soil can influence the composition of the corrosion layer. The metals contained in the immediate area around the deposited bronzes can accumulate in the corrosion, in particular manganese and cobalt, as well as phosphorus and silicon, which appear frequently in modern soil due to the effects of modern agricultural work. Thus, it can create a concentration of secondary metals, including iron (Fernandes et al. 2013, 4, Pernicka 1995 and aluminum (Mircea et al. 2012, 183). Robbiola and colleagues (1998), as well as Chase (1994), have suggested that ion migration is responsible for the observed effects within the corrosion layer. In order to explain the formation of copperalloy patinas, they have classified corrosion structures according to several types. The Type-I structurean even surface with a two-layer structure incorporating internal tin oxidation accompanied by selective dissolution of copper-is characteristic for quaternary bronze (CU-Sn-Zn-Pb) and brass (Cu-Zn) (Robbiola et al. 1998, 2090, Constantinides et al. 2002. It also appears on tin-bronzes (Cu-Sn) as 'noble patina'. The Type-II structure-a coarse three-layered structure with a general uneven structure and corrosion attacks such as pitting-appears frequently on arsenical bronzes (Cu-As), tin bronzes (Cu-Sn) and leaded bronzes (Cu-Sn-Pb) (see Robbiola et al. 1998, 2097, Constantinides et al. 2002. Interestingly, with an average thickness of 25-34µm and a minimum thickness of 3µm the Type-I corrosion layers are much less developed than the Type-II corrosion layers of tin bronzes (Cu-Sn). The average thickness of the latter measures 67µm (measured on artificially-produced patina, table. 1). Furthermore, Constantinides and colleagues observed an aggressive corrosion process in tin-bronzes at places 'where the original surface is destroyed' (Constantinides et al. 2002, 93); however, such developments are advantageous in craft technical studies, as punch marks, intensive hammering of the surface, or inserted decoration would be visible within the corrosion layer. This visibility of decorations applied post casting could be supported by the metallographic analysis of the bronze ornaments from Lower Saxony (Nørgaard 2015b, 121).

Differences in Element Composition Between Surface and Metallic Core
Results from the SEM analysis of the bronze ornaments from central Lower Saxony outlined here mark an important contribution to ongoing discussion regarding handheld XRF analysis of museum collections.
Initial visual, qualitative analysis of thin sections taken from samples of most artefacts considered revealed advanced corrosion (see figure 1). Occasionally it is to be inferred that no metallic core material is left as the artefact is brittle and very light, which is characteristic of fully corroded bronze (Oudbashi et al. 2013). To gain more information about the corrosion crust, an element distribution curve was charted for several samples displaying a full cross-section and a sample thickness of 1-2 mm, tracing changes in composition from the sample's centre to the edge of its corrosion layer. The curve depicting element distribution for the neck collar from Luttmissen, in Lower Saxony2, for example, showed a rapid drop of the copper values, as well as a simultaneous and disproportionate increase in tin content with the transition to the corrosion layer (figure 1). Similar phenomena have already been documented in other studies of ancient bronzes, mostly within conservation studies, while comparable results have been achieved in studies of artificially-created patinas (Fernandes et al. 2013, 5, Oudbashi et al. 2013, 3, Šatović et al. 2013, Constantinides et al. 2002. Copper dissolution, which starts with selective dissolution from the copper solid solution α-phase (see Robbiola et al. 1998Robbiola et al. , 2108, is commonly known as 'decuprification'. During this stage, the dissolved copper reacts with soil anions such as chloride (Cl) and carbon (C), and as these copper salts are soluble in soil waters, they can be carried away from the metal surface (see Chase 1994, Oudbashi et al. 2013, Weisser 1975. Along with the decrease in the copper values, the oxygen content rises sharply. This effect results from the reaction of copper and oxide to the copper-oxide cuprite. Since tin reacts to a much lesser extent with oxygen, the percentage of tin content in the alloy is not changed; in fact, the additional enrichment of tin in the corrosion layer (Pernicka 1995, 44) results in a distinct increase in its proportion relative to other components. In several studies of bronze corrosion, a high tin concentration has been detected in corrosion layers. Selective de-alloying, or 'destannification', from the bulk metal (and the tin rich δ-phase) towards the corrosion provides one possible explanation for this phenomenon, while selective removal of tin offers an alternative (see for example Chase 1994, Weisser 1975. In this case, the fast oxidation of tin can be seen as the main motivation for its dissolution from the bulk (tin-oxide films); however, this phenomenon requires further investigation.
With respect to the two alloying components of tin-bronze, copper and tin, one can speak of a disproportionate reversal of their percentages within the corrosion. However, within three-layered Type-IIcorrosion crusts, the chemical composition changes greatly from one layer to the next (Robbiola et al. 1998(Robbiola et al. , 2101. Element distribution maps (figures 2 and 3) demonstrate that the sharp rise in oxygen content, as well as the dissolution of copper and tin, are not local phenomena. Figure 2, captured using the SEM, depicts half the cross-section of a sample taken from a neck collar's end plate3. Clearly, the uneven corrosion layer (Type I/II corrosion, Robbiola et al. 1998Robbiola et al. , 2099 of about 30µm is visible, as is the protruding corrosion at the grain boundaries. The SEM allows filtering of specific elements, therefore it possible to visualise solely the distribution of, for example, oxygen and copper (figure 2, O K and Cu K -K indicates the shell) in an elemental map. These maps demonstrate that the decrease of copper within the corrosion is correlated to an increase in oxygen, and that this phenomenon occurs consistently throughout the corrosion. Additionally, the proportion of carbon (figure 2, C K) increases in the corrosion crust. One can infer that the decrease in copper is part of the decuprification process4. Simultaneously, tin becomes more densely concentrated (figure 2, Sn L).
The minor trace elements contained within the alloy also exhibit changes that are caused by the corrosion process. Sulphur inclusions were found in the vast majority of the metallographically-examined bronzes from the Lüneburg Heath (see Nørgaard 2014aNørgaard , 2015b. The detectable percentage of the sulphur content in most samples was constant; however, a slight increase in the sulphur values could be detected within the corrosion layer. Arsenic could be only sporadically documented in the samples (figure 3, As K). It is striking, however, that in the samples considered here, although arsenic-containing inclusions predominantly occured at the transition to the corrosion layer, they could not be detected at the outer edge of the sample (figure 4).
Changes in alloy composition within the corrosion crust were documented for all samples examined (figure 5). Depending on the type of corrosion (both Type-I and Type-II structures were present) the composition changed once or twice (for three-layered structures, see figure 5, 4931 (6)). However, the investigation led to some unexpected results.
A specimen sample from the site labeled as 'Lüneburgisch' provides a particularly interesting case. The fragment of a neck collar endplate with vertical ribbing and undecorated ends (see Nørgaard 2011, 34-35), examined under the SEM, showed a coating/layer with a dense structure, apparently attached to the original surface of the artefact (figure 6). According to its element distribution curve, the artefact shows a familiar shift in the copper and tin values in the transition from bulk metal to the corrosion crust. However, in the transitional area, the proportion of copper increases drastically (from 38.37 to 72.18 weight %, see figure 6) and the tin proportion falls below its original value ( figure 6 and 7). Additionally, sulphur content reaches its peak at the transition to the coating. This phenomenon becomes apparent in mapping the distribution of these elements. Both the distribution of copper (Cu) and tin (Sn) within the sample indicate a coating, as does the linear concentration of sulphur (S) along the surface of the artefact (figure 8). The appearance of tin, arsenic, and iron in the coating layer prompts one to question what this coating was made of and how it may have come to be present on the surface of such artefacts.   Brought to you by | Aarhus University Library / Statsbiblioteket Authenticated Download Date | 4/11/18 2:00 PM For restoration purposes as well as to imitate a green antique patina, linseed oil varnish, mixed with malachite or azurite, has historically been used. Such treatment was especially popular at the end of the nineteenth century (Copper Development Association 2008, Groves 2001. Occasionally, powdered copperalloy material was also added to the varnish. It can be assumed that the coating detected on this artefact is such a varnish and that it was applied to give the neck collar the 'typical green colour' of Bronze Age artefacts for its exhibition. Alternatively, it is possible that this layer was added as a conservational measure, as it has obviously limited the effects of corrosion of the artefact's surface.
Thus, the analysis conducted here of thin sections from artefacts of the Lower Saxony region allows one to conclude that that the metal composition of an object's surface does not reflect the composition intended by prehistoric artisans. Brought to you by | Aarhus University Library / Statsbiblioteket Authenticated Download Date | 4/11/18 2:00 PM

The Limits of Non-Destructive XRF Analysis on Archaeological Bronzes
Examination under the SEM clearly showed a marked change in the trace element composition from the core to the corroded surface of the bronzes. This observation leads one to infer that studies of corroded artefacts using a handheld XRF will not allow for the characterisation of the original alloy.
In order to determine a deterioration of the results by such corrosion layers in the bronzes, comparable data was assembled using portable XRF devices from two fragments of identical bracelets, found within an urn from the tomb of Gielow5 (ALM IV/64/561 c and d) and now held in the National Museum of Mecklenburg (see also Nørgaard 2015a). A quarter of each fragment was cleansed of the patina, which covers the remaining surface (figure 9). Two readings were taken for each artefact-one at the patina, and another from the polished area. A comparison of the two measurements revealed the corroded area to contain 3-4% less copper than the polished area, with an increased tin value of up to 4%. Furthermore, a slight deviation was recognisable in regard to the further elements such as arsenic, which was present in usually a few tenths or hundredths of the values observed for the polished areas. However, as the thickness, as well as the properties, of the corrosion layer (see figure 5, K851) can differ from artefact to artefact (Type-I corrosion and Type-II corrosion appear on artefacts with similar alloy components), it is therefore not appropriate to value the deviation of tin and copper to around 4%.
Detailed study of the element distribution curves (see figure 5) clearly demonstrates the impact of corrosion on the distribution of elements. As is shown by the data measured from the corroded artefacts in Mecklenburg (table 2)6, variations within the proportional values for tin are massive. Based on a variety of archaeological and archaeometallurgical studies of north German Bronze Age material culture, one might expect the artefacts investigated here to contain tin in proportions of between 8-12%, and possibly up to 14%. Data obtained via SEM from the artefacts from Lower Saxony support such a conclusion (see table 3). As such, values between 8.71 to 55.56% obtained via handheld XRF illustrate a severe difference. Commonly, within the earliest Bronze Age, artefacts were made of relatively pure copper or a multi-impurity metal with high levels of nickel and antimony (for further reading, see i.e. Junghans, Sangmeister, & Schröder 1974, Sangmeister 1971. During the early and middle Bronze Age in north Germany, copper-tin alloys were prevalent. The proportional presence of tin varies from 9-12% in NBA II and from 8-11% in NBA III7. Occasionally, artefacts contained 13-17% tin, as this composition was fairly widespread during the Taunton phase in the British Isles (i.e. Harding 2000, 204). As trading relations between these regions are known to have existed, one might assume the recycling of materials took place. However, only twenty of the seventytwo artefacts from Mecklenburg analysed for this study were determined to be close to these common compositional values.
The data leads one to conclude that the artefacts considered here are not equally well preserved. A superficial examination of their surfaces and accessible cross-sections could indicate the extent of their corrosion (as also shown in table 2); however, the data differs greatly from the compositions that were expected. A range of minor elements were measured, including aluminum (Al), manganese (Mn), phosphorus (P), and silicon (Si). As the X-ray penetration depth is limited to 100-200µm (Šatović et al. 2013, 7), these elements clearly were not a component within the original alloys; rather, they likely migrated from the soil into the developing corrosion layers. Still, the presence of such elements can provide useful data, informing the conservator of the degree to which the corrosion crust is developed, while determining the value of a visual, qualitative assessment of the state of corrosion.
The data from Mecklenburg was experimentally used to execute some calculations in order to consider the question of material groups related to Bronze Age workshops. To this end, proportional values obtained for those elements clearly related to the corrosion process (like Al, Cd, Cr, Mn, Mo, Nb, P, Si, Ti, V, W or Zr) were added and the total subtracted from 100%. In order to achieve a comparable composition of solely original alloy elements, the remaining elements were proportionally extrapolated to 100% (see table 4). At first sight, some of these calculated (in the sense of being mathematically constructed) compositions could be used in further evaluations of the material. Not knowing the original data, over 50% of the calculated elemental composition of the artefacts matches copper-tin (Cu-Sn) bronzes that could well have been present during the Bronze Age. However, they are calculated and, as such, are not necessarily indicative of the original alloy. None of these bronzes were examined if a metal core was still present. Some bronzes might have just a few μm of corrosion crust while others might have up to 1mm. As shown in table 4, a large number of artefacts contained tin in proportions far greater than their likely composition at the time of production. The uncertainty of the estimated data is also visible in the serious deviation of the tin content, ranging from 9.5% to 64%. Furthermore, the unreliability concerning the patina type, as suggested within this study, is striking.
As a result, analysis of metal composition using a non-destructive XRF device is not suitable to discussions related to the fine qualities of alloys used by contemporary craftspeople. The data also made clear that even when the material consists of artefacts in superficially similar corrosive conditions, one can find significant variation in elemental composition and depth of corrosion. As comparative studies have shown, bronzes with 'noble patina' (type I patina) and a high lead content might be better suited for such investigations (e.g. Fernandes et al. 2013Fernandes et al. , Šatović et al. 2013. 7 It should be noted that this information derives mainly from the analysis of weapons (axes) or ingots (see Junghans, Sangmeister, & Schröder 1974). It might be possible that ornaments contain a slightly higher tin percentage as has also been seen within the analysis of the ornaments from Lower Saxony.
Brought to you by | Aarhus University Library / Statsbiblioteket Authenticated Download Date | 4/11/18 2:00 PM Table 2. The data measured by the XRF device on the artefacts from Mecklenburg (German Mining Museum, Bochum). In order to compare the measurements, the patina type is given based on a superficial examination (including visible profile sections) of the artefacts. Additional, the major corrosion elements are given in values, the less frequent elements are named (< indicates a measurement below the detection limit of the device of around 50ppm

Conclusion and Perspective
Both non-destructive analysis by handheld XRF-devices and traditional XRF analysis of corroded bronze objects can provide useful but limited information about artefacts. In the first case, the results of these analyses can determine the morphology of corrosion crusts (see also results by Constantinides et al. 2002, Fernandes et al. 2013, Mircea et al. 2012, Wadsak et al. 2000. When considered alongside other factors (e.g. weight, colour (see for example Oudbashi et al. 2013, Weisser 1975, and the characteristic changes of the surface by the various types of patina described above) can be used to inform the conservator about the state of an artefact.
Furthermore, such an analysis permits identification of the major elements present in the material used to craft the artefact. One can assess whether an artefact is made of copper, bronze, brass, silver, or gold. Thus, the non-destructive analysis using XRF will provide answers to most general questions regarding material. The same is also true for some copper alloys, like arsenic-bronze, tin-bronze and lead-bronze. Alloys high in lead are more resistant to corrosion due to a quick passive oxide layer of metallic lead (Fernandes et al. 2013, 5). It has also been illustrated that the significant dissolution of copper documented in other copper alloys does not occur in arsenic bronzes (Constantinides et al. 2002, 93). Thus, more detailed information regarding an artefact's original composition, which is also explored here, might be achieved.
However, this analysis has demonstrated that tin-bronzes are especially prone to both types of corrosion, and that major changes in composition are caused during the corrosion process. The type of corrosion, however, cannot be determined with any certainty by a visual, qualitative analysis. Additionally, several cases of bronze disease (type III) in which no metallic copper was left in the artefact could be identified in the material. Furthermore, by comparing SEM cross-section scans on artefacts contemporary to the pXRFmeasured items from Mecklenburg, it became clear that the thickness of the corrosion crust differs greatly in several cases. Even after stripping off the corrosion elements, a mathematical extrapolation of the values could not yield any results comparable to the alloying ratios which are known from the Bronze Age.
Generally, a non-destructive, XRF-driven methodology is insufficient in identifying and quantifying the components of an artefact's alloy as intended by the craftsperson. This method cannot define intentionally improved alloys by specific craftspeople and, thus, is not suitable for the definition of workshops. Metallographic sampling, however, can help to identify craftspeople, as the skill of a craftsperson becomes apparent through the techniques that he/she applies. With regard to handheld XRF investigations of corroded bronze artefacts from museum collections, however, the fact remains that sampling is necessary in order to answer detailed questions related to provenance, workshops and craftspeople. That being said, XRF can nonetheless help to identify objects for further analysis, as heavily corroded objects are difficult to evaluate using metallographic methods.