Introduction

Aim of the study

The present study reports on μ-XRF analysis of 111 copper-based and 11 silver alloy coins from the ancient city of Rhodes, Greece. It is the first time that elemental analysis on both silver and bronze coins issued by the Rhodian mint from the mid-4th c. BCE to the 2nd c. CE has been performed. The objectives of the μ-XRF analyses were to identify the raw materials used for coin production, to characterize their typology in terms of their elemental composition, and to determine and quantify certain minor and trace elements that could allow for a more detailed classification. As all coins included in the analysis come from rescue excavations in Rhodes, it was only allowed to perform analysis of a non-destructive nature. In this respect, μ-XRF allows us to gain some insights on the elemental composition of coins. Usually destructive techniques that offer evidence for bulk composition are preferred for coin analysis in an attempt to minimize the risk from surface enrichment, especially in silver coins (Ager 2016).

Impetus for the study was given by the varying degree of preservation of coins that belong to the same issuing type and come from the same archaeological context. Given the absence of studies on the elemental composition of Rhodian coins, except for one conducted nearly three decades ago on 12 silver coins (J.-N. Barrandon and Bresson 1997), the present analysis aims to stimulate research in this field. Metals reached Rhodes via trade, gifts or booty, as the island of Rhodes lacks ores and metal resources. Although the provenance of metals lies outside the scope of the present study, the analysis helps us to detect patterns in the composition of coins. This raises some archaeological questions especially with regard to the evolution of Rhodian bronze coinage as well as to the changing practices in metal resources in the Rhodian mint. In other words, this study aims to further underline the importance of applying scientific methods in archaeological material, as the results can give new directions to historical research. In Hellenistic times, Rhodes emerged as a bustling and thriving cosmopolitan center, since the Rhodian fleet had a leading role in transit trade across the Mediterranean. Although at this stage of research solid answers cannot be provided, it is nevertheless worth bringing together archaeological and archaeometric evidence by posing the question to what extent, if any, differences and similarities in the composition of Rhodian bronze coinage over time may reflect shifting or recurrent trading networks.

Contextualizing the coins

The coins under examination are issues of the Rhodian mint from the second half of the 4th c. BCE down to the 2nd c. CE, except for two Roman coins of imperial mints which were nevertheless included for analysis (nos. R113 and R116). The coins selected for analysis came to light during rescue excavations in the ancient city and necropolis of Rhodes. They originate from five building complexes (Fig. 1, Table 1), which form the core of the research undertaken in the framework of the “Rhodes Centennial Project” (the geographical coordinates of the excavated plots are listed in the Supplementary Material). In other words, the present study relies exclusively on material retrieved from rescue excavations and included for study in the framework of the “Rhodes Centennial Project.” The small selection of silver coins (11 in total) in our sample reflects the dearth of precious-metal coins retrieved from excavations, unlike base metal coins (i.e., bronze coins), which constitute omnipresent finds in any excavation; the accidental loss of bronze coins would not have been severely felt by their owner, due to their low intrinsic and metal value (Callataÿ, 2006: 181–183). Despite the paucity of silver coins in our material, we nevertheless included these coins for analysis in order to get some first insights into their elemental composition.

Fig. 1
figure 1

Plan of the city of Rhodes indicating the location of the excavations included for study in the “Rhodes Centennial Project” (after Filimonos and Patsiada 2018, Fig. 3)

Table 1 Archaeological context of coins

The silver coins selected for analysis constitute a chronologically close group; nine out of 11 coins date after ca. 190 BCE, a period during which the Rhodians introduced significant changes to their coinage: the weight of the drachm was slightly increased from ca. 2.70 gr to ca. 3.00 gr, and an incuse square (plinthos) was introduced to the reverse of the coins, i.e., these are the coins known as plinthophoroi (Jenkins 1989: 101; Apostolou 2003:22). In our sample, only one hemidrachm can be ascribed to Group B Style 1 of Jenkins (cat. no. R10) on account of the name of the magistrate Ἀρτέμων with the club on the reverse. The names of the magistrates on the other two hemidrachms are no longer legible (cat. nos. R3 and 11). Two silver coins date to the period (i.e., 230–190 BCE) that predates the introduction of the plinthophoroi (ca. 190–88 BCE). Furthermore, the silver coins selected for analysis represent two different denominations, namely hemidrachms (four coins) and diobols (seven coins), that is relatively small denominations.

As for the bronze coins, the vast majority of the samples fall in the two main periods of intense production of bronze coins in the Rhodian mint (Table 2, Fig. 2). 33 coins date to the period from 350 to 300 BCE, when the Rhodian mint struck small denominations in bronze (9–11 mm in diameter, ca. 1.0–1.5 gr in weight) in large quantities for the first time (Ashton 2001: 90–91; Demetriou 2015: 142). It should be noted that eight coins of the type Nymphe/rose issue (nos. R12, 14, 19, 23, 31, 32, 39, and 41) and one of the type rose/rose (no. R49) were struck with no upright axes (i.e., at 12 h). According to Ashton (2001: 90, n. 57), the deviation from a 12-h axis coupled with some other features may be evidence for an early issue of bronze coins in Rhodes (late 5th c. BCE / early 4th c. BCE).

Table 2 Overview of bronze chalkoi
Fig. 2
figure 2

Specimens of issues of bronze chalkoi. Clockwise from the top left: a Nymphe / Rose (350–300 BCE), R 40 (N 3353); b Rose / Rose (late 3rd c. BCΕ), R 49 (N 3365); c Radiate head of Helios / Rose (early 2nd c. BCΕ), R 63 (N 3421); d Radiate head of Nymphe / Rose in incuse square (ca. 190–88 BCΕ), R 95 (N 3341)

All 33 coins stand for chalkoi that bear the head of Nymphe Rhodos on the obverse and a rose on the reverse. New bronze types and heavier denominations were introduced in the 3rd c. BCE (e.g., tetrachalka with head of Zeus (observe) and rose (reverse); dichalka with veiled female head (observe) and prow or rose (reverse) (Ashton 2001: 91); however, these types were not included for analysis due to the paucity in our material. Instead, we singled out for analysis two distinct series struck sometime in the late 3rd and early 2nd c. BCE. Thirteen coins bear the rose both on the obverse and the reverse and probably date to the late 3rd and early 2nd c. BCE (Ashton 2001:91). Fourteen coins bear the radiated head of Helios in profile on the observe and the rose in the reverse, dating around the turn of the 3rd to the 2nd c. BCE (Ashton 2001). Although these two series were of short circulation, they are fairly well represented in the assemblage. The bronze chalkoi of the plinthophoric system constitute the largest sample in our material (35 coins). As mentioned above in relation to silver coinage, the Rhodians around 190 BCE, if not a bit later, changed the metrology and iconography of their coinage (Ashton 2001: 86; Marcellesi 2007: 69-–72). The new bronze coins now carry a radiated head of Nymphe Rhodos on the obverse and a rose within an incuse square in the reverse (Jenkins 1989).

Lastly, 14 bronze coins date to the Roman Imperial period (Kromann 1988: 214–216), with half of them (7 out of 14) dating to the 1st c. CE (nos. R107–108, 110–112, 114–115) (see Table S1 in the Supplementary Material). Five coins that date to the 1st c. CE bear the radiated bust of Helios on the obverse and Nike on the reserve (nos. R108, 110, 111, 112, 115). The type of Dionysos/rose and of Sarapis/Helios are represented by one coin each in our sample (nos. R107 and 114 respectively). Five coins that bear the head of emperor on the obverse and radiate head of Helios on the reverse (nos. R109, 117–120) are roughly dated to the 1st/2nd c. CE; the emperor cannot be further identified, due to the state of preservation. The latest coins in our sample stand for two different denominations and are both dated to the reign of Commodus (177–192 CE) (nos. R121–122). R121 carries a Nike on the reverse, while R122 bears the radiate head of Helios.

It should be mentioned that two more Roman coins were subject to scientific analysis (nos. R113 and R116). On macroscopic observation alone, their identification was problematic. The analysis confirmed our suspicions that these two coins are issues of non-Rhodian provincial mints, as both are outliers in the analytical data of our sample.

Materials and methods

Sample description

The 122 metal archaeological artefacts originate from Rhodes. Of those 111 are made of copper-based alloys, whereas 11 are made of silver alloys. According to preliminary archaeological research, all coins but two belong to issues of the Rhodian mint. The degree of preservation of these coins varies considerably; noticeable differences in terms of preservation can be observed within coins originating from the same complex. The majority of coins in our sample (84 out of 122) underwent cleansing and were subject to chemical treatment in the 1970s and 1980s, i.e., all coins found in the Soichan-Minetou plot (44) and most coins from the Papachristodoulou-Karika (26 out of 37) and Yallousi plots (14 out of 18). However, due to active corrosion, all these coins were also mechanically treated in the Lab of the “Rhodes Centennial Project” in Rhodes. Likewise, the remaining 38 coins which were not cleaned upon their discovery (19 coins from Mamaligka plot, four coins from Kladogeni plot, four from Yallousi plot, and 11 from Papachristodoulou-Karika plot) underwent mechanical conservation in the period between 2018 and 2022.

The INPP μ-XRF spectrometer

The μ-XRF analysis technique allows rapid, non-destructive, and sensitive in situ elemental analysis of major, minor, and trace elements in archaeological and historical metals. Scanning μ-XRF measurements, performed on millimeter-sized polished areas of the samples, can contribute to the understanding of the types, manufacturing technology and uses of metallic raw materials. The microprobe XRF analysis was carried out using a customized version of the so-called Artax spectrometer (Bruker Nano GmbH) at the X-ray Fluorescence Laboratory of the Institute of Nuclear and Particle Physics (INPP) at the NCSR “Demokritos,” Athens. The INPP μ-XRF spectrometer is composed by an X-ray microfocus Rh anode tube (spot size 50 × 50 μm, maximum high-voltage 50 kV, tube current 0.6 mA, 30-W maximum power consumption, beryllium window 0.2-mm thickness) and of a polycapillary X-ray lens as a focusing optical element (IfG). The X-ray lens provides spatial resolution (spot size) for the filtered exciting beam at the order of ~ 75 μm @CuKα, when the exciting beam is filtered (Kantarelou and Karydas 2016). The X-ray detection chain also consists of a thermoelectrically cooled 10 mm2 silicon drift detector (SDD, X-Flash, 1000B) with 146 eV FWHM @ MnKα and 10 kcps. A color CCD camera (with approximately × 13 times magnification) combined with a dimmable white LED and a spot laser beam assures reproducible positioning of the measuring probe, as well as visualization and documentation of the analyzed area. Three stepping motors coupled with the spectrometer head allow three-dimensional movement for elemental mapping and precise setting of the analysis spot at the focal distance of the polycapillary lens.

Sample preparation

The copper alloys are prone to developing corrosion layers on their surface with products rich in either Cu, Sn, or Pb, on the grounds of a number of factors including bulk composition, burial conditions, contact with other artefacts, conservation procedures, and storage conditions. The micro-XRF (μ-XRF) for surface analysis was performed, considering the advantages, and limitations of its application on coins, as described in literature (Kantarelou et al. 2011; Ponting 2012; Blet-Lemarquand 2013; Ager 2016; Pearce 2018). Importantly, the non-destructive and minimally invasive character of the technique guarantees the integrity of the artefacts, without jeopardizing the high precision of the measurements.

To ensure that no layers of corrosion were analysed, minimal intervention was required by an experienced conservator who performed cleaning of a ~ 2–3-mm diameter spot under the microscope on each sample prior to the analyses. Each coin was stripped of its protective varnish layer with acetone. All visible corrosion products on the bronze coins were removed with micro-scraping and the exposed metal surface was further polished with silicon carbide (SiC) down to 3000 grit size. In the case of the silver coins, however, a milder approach was necessary, so as not to disturb the soft metal (Fig. 3). The visible corrosion products—considering the possible enrichment of silver on the surface—were removed, in order to approximate the internal composition of the coin metal in the area of the measurements.

Fig. 3
figure 3

Photomicrograph of the cleaned surface of the bronze specimen R49 (left) and the silver specimen R1 (right) prepared for μ-XRF analysis

The analytical μ-XRF protocol

The measurement conditions are set at 50 kV, 600 μΑ, as tube high voltage and current, respectively. The objects are measured using a filtered exciting X-ray beam placed on a height-adjustable platform. The analyzed surface is aligned to form an angle of 90° to the incident X-ray beam and 45° to the detector axis. Within the cleaned area, about eight to 30 independent spot measurements are performed under the “area scanning” measurement mode with a total acquisition live time of 600 s to achieve good counting statistics and relatively low detection limits. To improve the detection of trace elements and obtain reliable quantitative results, a Co/Ti/Pd filter is inserted into the excitation beam path. The different analysed spots are separated by about 100 μm (step size of the scan measurements). Eleven elements are analysed for the copper-based alloy coins (Cu, Sn, As, Pb, Fe, Ni, Co, Zn, Se, Sb, and Bi) and seven elements for the silver coins (namely Ag, Au, Cu, Pb, Fe, Zn, and Bi). The element concentrations generated refer to the analysis of the so-called “sum spectrum,” which results from the sum of all individual spot measurements. Within the analytical protocol applied for the μ-XRF measurements, the Limit of Quantification (LoQ), defined as three times the respective Limit of Detection (LoD), is determined for several trace elements of interest and for a measurement time of 600 s (Table S2). The element concentrations found below the respective LoQs are not reported in the results table to improve the reliability of the reported concentrations.

Validation of the analytical μ-XRF protocol

The limitations of accurately quantifying the elemental composition of silver and copper alloys have been demonstrated and discussed in various applications, and specific measurement and evaluation methods have been proposed to reduce these effects (Charalambous et al. 2014; Karydas et al. 2008a, b; Orfanou and Rehren 2015; Sheedy et al. 2020). For the quantification of the μ-XRF measurements, a dedicated calibration methodology is applied (Kantarelou and Karydas 2016). An experimentally determined spectral distribution of the Rh anode tube, corrected for the energy dependent transmission efficiency of the polycapillary X-ray lens, is introduced into the PyMca software package, served next for spectrum deconvolution and the determination of the alloy elemental concentrations.

The accuracy of the applied analytical methodology is validated by the analysis of internationally certified reference samples. For the copper alloys, BCR-691, BAM-374, and NIST-1107 were analysed and for the silver alloys, the CNR series manufactured within the PROMET programme (Karydas et al. 2008a, b) and the reference samples Ag925 and Ag800. This comparison demonstrates good agreement between the measured and certified element concentrations of the reference alloys (see Tables S3 & S4). The accuracy was found to be better than 0.5% for the major alloy elements and better than 10% for the minor elements in the alloy. The observed differences exceeding ± 20% can be attributed to several factors. As described by Ingelbrecht et al. (2001), the certified results in the report for the BCR-691 series were obtained through the analysis of an extended area, where the minimum analysed area of the sample for XRF analysis should be 5 mm in diameter. This means that a larger surface area of the reference samples was analyzed, which can provide a more accurate representation of their overall composition. In contrast, the μ-XRF results in our analysis were derived by the average of individual spots measured as an “area scan,” which corresponds to an area of less than 0.90 mm2. This method of analysis provides a much smaller sample size, which can lead to higher levels of variability in the results, as it can be more sensitive to localized differences in their composition. Overall, the observed differences between the certified XRF results and the measured μ-XRF results can be attributed to differences in the size of the sample analysed and the sensitivity of the analytical method; important factors to consider when interpreting the results of the analysis.

Results and discussion

Copper-based alloy coins

The elemental composition of all copper-based coins, together with relative uncertainties deduced from the peak statistics are reported in table S5. All samples are copper-based alloys with variable concentrations of tin and a range of minor and trace elements. Copper is present with mean and median values of 86 and 87 wt% respectively and a range between 71 and 92 wt%. Tin is present in all samples with a mean of 11 wt%, a median of 11 wt% and a range of < 0.2–23 wt%. Figure 4 shows the μ-XRF spectra of the minimum and maximum tin values detected in the sample. Lead is found in 110 out of the 111 coins, with a mean value of 3 wt% and a median of 1 wt%, while higher lead contents of 1–23 wt% were noted in 45/111 samples. The rest of the elements detected in the sample are in the range of values characterised as minor and trace. Iron is found in 57/111 coins, and typically at < 0.4 wt% (except for coin no. R13 which reached up to 1.2 wt%) with a mean value of 0.2 wt%. Arsenic is detected in 65/111 coins below 0.5 wt%, with mean and median values of 0.2 wt%. Antimony is detected in 17 coins at about 0.3 wt%, and nickel in 63 coins at < 0.2 wt% and a mean of 0.05 wt%. Cobalt is detected in 50/111 coins with a mean value of < 1100 ppm. Selenium is detected in 14 coins at < 80 ppm. Traces of bismuth of < 850 ppm are noted in one coin (R113). Finally, zinc was detected in one coin (no. R116) with a value of approx. 5 wt%.

Fig. 4
figure 4

μ-XRF spectra for coins R18, R121 and pure copper target. Sample R121 is made of low tin (1.19 wt% Sn) while the sample R18 contains the maximum value of tin in the assemblage (23.4 wt% Sn)

Alloying elements

Based on the results, three groups of coins, alongside two outliers, can be identified, on the basis of the main elements present in the copper alloy at more than 1 wt%. Thus, the assemblage comprises 66 bronzes (Sn > 5wt%, Pb < 1wt%), 21 ternary bronzes (Sn > 5wt%, 5wt% > Pb > 1wt%), 22 lead-bronzes (Pb > 5 wt%, Sn > 1 wt%), one quaternary brass and one leaded copper (see Tables 3 & S6).

Table 3 Classification of the copper alloy types, according to their chemical composition

The leaded copper coin R113 is an outlier, made of copper and lead, bearing traces of tin, arsenic and nickel. The quaternary brass artefact R116 is made of copper, zinc, and relatively equal parts of tin and lead. The range of values for their major elements of the rest of the alloy types is:

  • for the 66 binary bronzes, Cu: 79.0–92.0 wt%, Sn: 7.90–20.0 wt%.

  • for the 21 ternary bronzes, Cu: 73.5–90.0 wt%, Sn: 7.80–23.5 wt%, and Pb: 1.10–4.80 wt%, respectively.

  • for the 22 leaded bronzes, Cu: 71.5–90.5 wt%, Sn: 1.20–18.0 wt%, and Pb: 5.30–19.0 wt%.

It is difficult to decide with certainty where exactly to draw the line that distinguishes between lead added intentionally and lead as an impurity in the alloy. Considering the results, we noticed that 66 out of 95 bronze chalkoi (70%) dated between the second half of the 4th c. BCE and the 2nd c. BCE contain a negligible amount of lead (less than 1%). This percentage would rise to 80% if we were to include in this group ten chalkoi with a lead content between 1 and 3% (six chalkoi from the period 350–300 BCE and four plinthophoroi). These are currently assigned to the “ternary bronze” group. Their low lead content suggests that the element is an impurity of the alloy and not a deliberate addition. Questions arise with regard to the eight remaining coins of the ternary group, which have a lead content of between 3 and 5%. It is not entirely clear whether this relatively low lead content in the coins is an impurity or an intentional addition. However, given the negligible lead content in the majority of the chalkoi in our sample, this otherwise low lead content should not pass unnoticed, as it blurs the line between intentional addition of lead and impurity. Caley (1939: 119–123) considered the presence of 2–3% lead as an upper limit for classifying a coin as bronze rather than leaded. However, more recent studies have shifted this threshold somewhat upwards, with the upper limit for lead in a copper-based alloy sometimes reaching 5–6%; the latter is the case for a series of coins minted by Hieron II in Syracuse (Frey-Kupper and Barrandon 2003). However, most studies set the upper limit for unintentional addition of lead at 4–5 wt%, as such amounts have limited effects on alloy properties (Gale et al. 1985; Pernicka 1999; Masson-Berghoff et al. 2018). We set the limit for intentional addition of lead at 5% or more (our “leaded bronze” group). Nearly 15% of the chalkoi from the mid-fourth century BCE to the early 1st c. BCE certainly belong to this group of leaded bronzes (Graph 1). However, there is a possibility that the lead was not added to improve the properties of the alloy, but to reduce the manufacturing costs. This hypothesis is supported by the high lead content of some of the coins, which can significantly affect their mechanical properties (Giumlia-Mair 1992).

Graph 1
figure 5

Frequency histogram of lead distribution on the copper-based coins

Based solely on the copper, tin and lead concentrations, most coins fall within the range of common chemical compositions of copper alloys for the years of interest (late Classical to Roman) (Di Fazio et al. 2019; Faucher and Lorber 1989; Faucher and Olivier 2020; Frey-Kupper and Barrandon 2003). Importantly, it seems that there was not a consistent use of the same alloying technology for artefacts belonging to the same periods (as they were classified in Table S6). For instance, a simultaneous use of both bronze (59.5% of the total sample) and leaded bronze/ternary bronze-made artefacts (38.7% of the total sample), irrespective of their dating period, is observed.

The quaternary brass coin (no. R116), belonging to the Roman period, seems to have been made with a different alloying technology. The low tin content in the coin (ca. 2.5 wt%) suggests that the artefact was the product of recycling, deriving from the mixing of the major metals with scrap bronzes, since the tin content decreases with each recasting the final alloys are poorer in this element. The same is possible for the leaded copper coin (no. R113) which is typical of artefact deriving from recycling scrap bronzes, considering the very low tin value (0.23 wt%) included in its alloy (Charalambous et al. 2014). Both coins are issues of non-Rhodian mints.

However, a pattern of common metallurgical technology can be observed through the correlation between major elements and specific time periods (see Graph 2 and “Remarks on the bronze coinage assemblage of Rhodes” section). The 14 coins dated to the early second century BCE may be considered as loosely forming one group. Comprising a recipe of an average tin value of 12.2 wt%, they could have been the manufacturing result of the same workshop, since they are made with the same alloying technology. Nevertheless, it should be noted that they present different compositional profiles of trace elements. It is the only chronological group so closely clustered in terms of its alloying elements. The majority of the coins, dated between 350 and 300 BCE, also seem to belong to a common metallurgical technology. Within the coins’ assemblage, 23 out of the 33 coins are made of bronze, five coins belong to the leaded bronze group and the remaining five to the category of ternary bronzes. Similar considerations apply for more than half of the assemblage of the coins dated between 190 and 88 BCE; twenty-one out of the 35 are made of bronze, whereas ten belong to the ternary bronzes’ group and the remaining four to the leaded bronzes. Of the 13 coins belonging to the late 3rd c. BCE, six are made of bronze, five of leaded bronze and two of ternary bronze. In the coins issued by the Rhodian mint during the Roman period the quantities in copper do not decrease, but remain stable between 80 and 90%. In almost one-third of our samples (5 out of 14 coins), lead appears in higher proportion than tin. In the remaining two thirds, the proportion of tin (between 7 and 12%) is higher than the proportion of lead and only in one instance the quantity of tin (7.25%) matches the quantity of lead (7.83%).

Graph 2
figure 6

Number of specimens in metallurgical categories based on chronological classification

As can be seen from Graph 3, metallurgical technology is more stable in the BCE periods, with bronzes being preferred; while in the Roman period (between 180–192 CE, 1st–2nd c. CE, and 1st c. CE), no preferred recipe can be identified and different technologies are chosen, possibly depending on the availability of raw materials. However, since the Common Era sample is only 1/7th of the total sample, it is too small to draw solid conclusions regarding preferred recipes.

Graph 3
figure 7

Frequency histogram of tin distribution in the sample according to chronological periods (before common era & common era)

Copper alloy impurities

On the basis of the three groups, an attribution of origin could be attempted for the raw materials used, but given the subtle differences, this classification helps to gain technological information related to refinement rather than provenance. For the majority of coins in the assemblage, there seems to be a trend towards the use of high-quality raw materials with a high degree of refinement, with most trace elements removed during smelting. As mentioned above, this stable metallurgical technology, which can also serve as an index of the common origin of the raw materials, is easily observed, among others, in the coins from the period 190–88 BCE, most of which contain antimony with a mean value of 0.30 (± 0.1) wt%; an element that frequently coexists with copper ores. The antimony detected in 50% of the coins of this period either corresponds to remelted scrap metals or is an impure remnant of the original ore composition, such as bournonite (PbCuSbS3) (Katsifas and Zachariadis 2019; Mangou 1994). Although the element frequently coexists with copper ores, it is rarely found in amounts greater than 0.1 wt% (Giumlia-Mair et al. 2018), a value that is below the limit of quantification for antimony in the μ-XRF. Therefore, it can be assumed that traces of the element are present in more coins. Furthermore, the absence of arsenic in the antimony-containing group is notable. Arsenic is present in almost half of the samples (65/111) (Graph 4). This may be a sign of common origin, as the element occurs as a natural impurity in many copper ores (Papadimitriou 2001; Rapp 1982). Since arsenic can be found in quantities with a mean value of 0.20 (± 0.1) wt%, it is fairly certain that it does not originate from scrap metal recycling (Giumlia-Mair et al. 2018). The presence of nickel (63/111) with a mean value of 0.052 (± 0.02) wt% and cobalt (50/111) with a mean value of 0.11 (± 0.2) wt%—two common elements associated with provenance—in almost half of the coins from all periods should be considered as supporting this hypothesis (Graphs 5 and 6) (Doonan et al. 2007; Heginbotham et al. 2015). Nevertheless, the value of zinc in coin no. R116 raises the possibility that this coin (and possibly other coins in the assemblage with relatively low concentrations of tin and lead) was produced by remelting scrap. Finally, the fact that the iron content of most coins is below 0.2 wt% suggests that sulphide ores such as chalcopyrite or bornite were used and properly refined. This was done under highly reducing conditions and at temperatures high enough to allow mobilisation of the iron during the smelting process, resulting in effective removal of the element from the ore (Kiderlen et al. 2016; Fazio et al. 2019).

Graph 4
figure 8

Frequency histogram of arsenic distribution in the assemblage according to chronological period

Graph 5
figure 9

Frequency histogram of nickel distribution in the assemblage according to chronological period

Graph 6
figure 10

Frequency histogram of cobalt distribution in the assemblage according to chronological period

To investigate provenance issues and technological aspects by revealing compositional patterns of the raw material used, the well-known k-means clustering method was used (MacQueen 1967; Likas et al. 2003). This is an algorithm that aims to divide a number of n points (coins in our case) into k non-overlapping clusters based on Euclidean distance. Once the clusters are determined, each point (coin) belongs to the cluster with the nearest mean (cluster centroid), which serves as the representative point of the cluster. It should be noted that the centroid value does not necessarily have to be part of the data set, but rather corresponds to the arithmetic mean of the data points assigned to the cluster. All cluster centroid values for the elements of interest are listed in Table S7. In this case, each observation was based on the concentrations of copper, arsenic, iron, cobalt, nickel, selenium and antimony. A notable difficulty in applying clustering algorithms is determining the correct number of clusters. To solve this problem, the silhouette criterion (Rousseeuw 1987) was used in our study. Four clusters were derived through statistical analysis, namely 0 to 3. Table S6 shows to which cluster each coin studied belongs; while in Graph 7, the frequency histogram of the k-means clusters created on the coins’ assemblage provides a visual representation of the distribution of coins over different time periods. The dating axis is divided into seven intervals, each corresponding to a specific period of time, roughly from the second century CE to the fourth century BCE. This histogram allows for a comprehensive understanding of the chronological distribution of the coins and offers insights into the historical context and the evolution of coinage during these periods. Thus, the results of the k-means clustering analysis of the above-mentioned elements reveal a distinct pattern. It becomes evident that all clusters include a wide range of dates covering the entirety of the periods of interest, with the exception of the Common Era period. This suggests that the elements present in the clusters exhibit a diverse distribution over time, reflecting variations and transitions in the different historical periods. The absence of the Common Era period in these clusters suggests a unique composition or distinct elemental signatures that distinguish them from the other periods and highlight a possible demarcation or significant shift in elemental composition during this specific time frame.

Graph 7
figure 11

Frequency histogram of the k-means clusters generated on the coins’ assemblage. The seven intervals of dating axis represent the corresponding periods from approximately 2nd c. CE back to 4th c. BCE. Specifically, 1 = 180–192 CE, 2 = 1st/2nd c. CE, 3 = 1st c. CE, 4 = 190–88 BCE, 5 = Early 2nd c. BCE, 6 = Late 3rd c. BCE and 7 = 350–300 BCE

The mean concentration values of the main alloying element tin (Graph 8) found in the four clusters identified by the k-means analysis are as follows:

  • for the 26 copper alloys of cluster 0, tin has a mean value of 8.37 ± 3.1 wt%,

  • for the 28 copper alloys of cluster 1, tin has a mean value of 9.13 ± 1.6 wt%,

  • for the 32 copper alloys of cluster 2, tin has a mean value of 15.0 ± 2.5 wt%, and

  • for the 25 copper alloys of cluster 3, tin has a mean value of 11.3 ± 0.7 wt%.

Graph 8
figure 12

Frequency histogram of the k-means clusters generated on the coin’s concentration of tin

However, adding the lead values in the sample, the above clusters can be established as leaded (21) and ternary bronzes (4) for 25 of the coins in cluster 0. Ternary bronzes can be identified in all clusters, binary bronzes in all clusters except cluster 0 and leaded bronzes are identified mainly in cluster 0 with only one coin in cluster 2 (Table 4). Table S6 shows the classification based on the results of the compositional analyses together with the k-means clustering. It is important to note that the clusters identified through the k-means analysis are not definitive groups that explain how the alloying technology of the assemblage was developed. Rather, as already mentioned, these cluster classifications are based on trace elements that come mainly from the copper ores. A major advantage of the method is therefore the ability to cluster a large number of compositional analyses based on their trace element composition, which allows the identification of patterns in the selection of raw materials in different groups of artefacts. Therefore, the clusters can provide useful information about a possible common origin of the different alloy groups. Graph 9 displays the distribution of copper, tin, and lead values in the coin assemblage, which was determined using the k-means clustering, in a ternary diagram.

Table 4 Distribution of the coins within the four clusters identified through the k-means analysis, based on the alloying elements compositional profile
Graph 9
figure 13

Ternary diagram of copper, tin, and lead values in the copper-based coins’ assemblage based on the k-means clusters classification

As can be seen from the frequency histogram for lead (Graph 10), the range of values for lead in cluster 0 is between 5.8 and 12.0 wt%. For cluster 1, the lead content ranges between 0.026 and 2.31 wt%. For cluster 2, the values range from 0.007 to 5.58 wt%, while for cluster 3 they range from 0.07 to 3.05 wt%.

Graph 10
figure 14

Frequency histogram of the k-means clusters generated on the copper-based coins’ concentration of lead

As mentioned earlier, four different clusters were identified in the statistical analysis of the coin data by k-means clustering. Cluster “0” is distinguished from the others by its increased lead content, including leaded bronzes and one leaded copper coin. However, it is interesting to note that all binary bronzes and most ternary bronzes were classified in three other clusters, “1” to “3.” The three groups seemed most likely to contain a tin content that was limited for each cluster within a certain narrow range with no overlap of tin concentrations. As expected, since no cluster can be associated with a particular chronological period, the k-means classification does not provide solid insights into technological or provenance issues. While the technological aspects of Rhodes coin metallurgy are found to be unstable over the course of six centuries, this method confirms that the selection of various technological options is influenced by factors such as the availability of raw materials during the relevant periods. This availability, which can be due to many different factors (finance, politics, physical phenomena, etc.) that can promote or disrupt raw materials trade networks, is discussed in the following chapter.

Remarks on the bronze coinage assemblage of Rhodes

The advantage of μ-XRF analysis performed in Rhodian bronze and silver coins is that for the first time we gained fresh insights into their composition. As the vast majority of our sample constitutes small bronze coinage (chalkoi), 95 coins in total, representing four issues that date from the mid-4th c. BCE down to the late 2nd c. BCE, some first remarks on the evolution of the composition of Rhodian bronze coinage (i.e., chalkoi) can be formulated.

In the two major—most widespread—series, namely the Nymphe/rose of the Late Classical period and the plinthophoroi (radiate Nymphe/rose) of the 2nd c. BCE, the composition of the alloy of chalkoi presents some recurrent patterns with regard to the ratio between copper and tin or copper, tin and lead. In both series, only around 1/5 of the coins are of a ternary alloy (between 1 and 5 wt% lead). In most coins of both series, the alloy is a binary one (copper and tin). When lead is intentionally added, the amount of tin is by no means reduced, and this affects the amount of copper, which in these instances ranges between 71 and 87 wt%. The tin content remains steadily above 6 wt% not only in the two major series, but also in the series with a shorter circulation (rose/rose, late 3rd c. BCE and Helios/rose, early 2nd c. BCE) that have been analyzed in this study.

The pattern that is thus readily noticeable concerns the consistently high ratio of tin in the alloy throughout the late Classical and Hellenistic periods. In other words, the Rhodian mint does not seem to follow the relatively widespread trend of the Late Hellenistic period, namely increasing the amount of lead to the detriment of tin. Even when lead is clearly intentionally added (above 5 wt%) the tin levels remain as high as would have been expected in binary alloys. Likewise, in the very few instances (3 coins out of 95) in which the lead content exceeds the amount of tin, the latter surprisingly does not fall below 9.6 wt% Sn. This could indicate that in a timespan of more than two centuries the Rhodians placed great importance on the appearance of coinage whose metal otherwise had a fiduciary value only; the value of coined bronze was 4 to 10 times greater than its intrinsic value (Marcellesi 2010). The consistent concentration of tin in the alloy over a period of almost three centuries would guarantee the golden appearance of Rhodian bronze coins—a well-established feature of various types of copper—based secular and votive objects, indicating an uninterrupted supply network for tin and underscoring a long metallurgical tradition (Kladouri et al. 2021).

The consistent concentration of tin in all four series of Late Classical and Hellenistic coins analyzed in this study sets the Rhodian coinage apart from developments observed in the evolution of bronze coinage in other cities. In light of the scientific analysis carried out in 1100 bronze coins from ten different cities in the late 1990’s and 2000’s Blet-Lemarquand (2013) revised and refined some of Caley’s observations (1939: 186). It is commonly thought that the first series of bronze coinage in other cities is usually characterized by binary alloys with ca. 10 wt% tin, without added lead (Blet-Lemarquand 2013: 50–51). The Rhodian coinage does not conform to this picture. The binary alloy, as noted above, is a feature not restricted to the first widespread series of Rhodian bronze coinage (Nymphe/rose). It is also attested in the later series under examination (e.g. radiate Helios/rose, radiate Nymphe/rose). Scholars have noticed that towards the Late Hellenistic period the amount of lead would increase steadily, to the detriment of tin. This holds especially true for the Ptolemaic bronze coins from the beginning of the 2nd c. BCE onwards (Faucher and Olivier 2020: 104). Moreover, tin was more expensive than lead and as Pliny (NH 34.161) noted in the 1st c. CE tin cost 11 times more than lead. Yet, overall, the alloy of Rhodian coinage remains consistently uniform as described above, with most coins having primarily a binary alloy, even in the plinthophoric issue of the 2nd c. BCE.

The two short-lived series of the late 3rd and early 2nd c. BCE respectively are of special interest. Despite their chronological proximity, the series with a profile radiate head of Helios and rose stands apart from the rose/rose series in terms of alloy composition. As coins of both series were found together in the hoard IGCH 1342, Ashton (2001: 91) suggested a close date for these issues. Placing the rose/rose issue in the late 3rd/early 2nd c. BCE, he considered the Helios/rose series “the last pre-plinthophoric bronze issue” on account of the resemblance of the profile radiate head of Helios to the radiate head of Nymphe on plinthophoroi.

Notwithstanding the relatively small sample for the Helios/rose issue (14 coins), the metal composition is strikingly consistent in all coins, but one which contain high levels of cobalt (1 wt%). Although cobalt may point to a different source of copper ore, the recipe of the alloy for this series is pretty homogeneous. These coins have an average content of tin set to 12 wt%; lead is below 1 wt%; the copper ores do not contain any trace elements, such as arsenic or antimony. These coins therefore not only represent a different issue in terms of types (obverse/reverse) but they also point to a quite homogeneous recipe for the alloy.

The rose/rose series (13 coins in our sample), on the other hand, exhibits the most inconsistent patterns in terms of alloy composition among the late Classical and Hellenistic series. Nearly 40% of the coins (5 out of 13 coins) of this series present evidence for deliberate addition of lead in their alloy, even if the amount of tin remains high (between 9 and 18 wt%). While an increase of lead means a proportional decrease of tin, in this series we observe something unusual: both metals (i.e., tin and lead) increase concomitantly to the detriment of copper. It is also the first time among the Hellenistic series that the content of lead occasionally has exceeded the content of tin (in two out of the five coins). In our sample, this trend is also noted in just 1 out of the 35 plinthophoric coins. Such a frequent occurrence of lead in an otherwise small sample is quite striking. And this becomes even more striking when compared to the remarkably consistent binary alloy of the radiate Helios/rose series. Two series, which were minted within a narrow chronological framework and were both short-lived, therefore differed drastically from each other in terms of alloy.

Although further analytical coinage data is needed to confirm these results—considering that any attempt to determine the origin of the copper ore in a leaded bronze by the lead isotope method is pointless (Kassianidou and Knapp 2008)—the patterns observed by these two-short lived series raise questions worth addressing, especially when we consider the broader historical background. How are the patterns noted for the rose/rose series to be explained? These are unusual features in light of the practices of the Rhodian mint, as highlighted by the current analysis. Did the Rhodians use recycled bronze to strike the rose/rose series? Is this series to be seen as a temporary measure that reflects exceptional circumstances or disruption in the supply networks? If we accept the chronological sequence for the Rhodian issues, the rose/rose issue was succeeded by the Helios/rose issue, which displays an astonishingly standardized consistency in terms of composition. The Helios/rose issue may have come as a response of the Rhodian mint to the irregularities of the rose/rose issue.

The late 3rd c. BCE and early 2nd c. BCE was a turbulent period in Rhodian history; this is reflected in the numismatic policy of the Rhodian state with the issue of a lower-weight tetradrachm and the introduction of a new series of drachms and didrachms of lower weight (Apostolou 2003: 19–22). In 223/2 BCE, a devastating earthquake hit Rhodes (Skaltsa, forthcoming). According to Polybius (5.88.1), this earthquake caused the collapse of the Colossus and it caused great damage to the fortification walls and the shipsheds, Rhodes’ appeal to Hellenistic kings, dynasts, and cities to raise funds for the recovery of the city turned a natural disaster into an advantage; donations in kind and cash flew into the city from all over the Mediterranean (Bresson 2021). Among the gifts was silver and coined bronze; silver was donated by Hieron and Gelon of Syracuse (100 talents), Ptolemy III (300 talents) and Antigonos III and his wife Chryseis (100 talents), while Ptolemy III also donated 1000 talents of coined bronze and 3000 talents of raw bronze for the Colossus. Two bronze series (which include tetrachalka with head of Zeus on the obverse and rose on the reverse and dichalka with veiled female head on the obverse and prow or rose on the reverse) have been connected to this earthquake (Ashton 1986: 1–8, 2001: 91), but they were not included for analysis due to the paucity of these coins in our material. During the Third Macedonian War (205–201 BCE) the Rhodian Peraia was devastated by the raids of Philip V, while in 199/8 BCE another strong earthquake hit Rhodes and its broader territory causing extensive damage. This brief historical account of the late 3rd and early 2nd c. BCE helps only to outline the broader political and economic framework, within which these two issues may have been struck. It also highlights how data obtained from scientific analysis in the smallest denomination of the Rhodian coinage (i.e., chalkoi) can raise questions of historical importance.

Some interesting patterns emerge when minor elements such as arsenic, antimony and cobalt are considered in relation to the various issues of the Rhodian mint. As already mentioned, (see the “Copper alloy impurities” section), arsenic is a natural impurity often found in copper ores, and since antimony and cobalt are minor elements that also often coexist with copper ores, their detection can offer hints regarding the origin of the ores. These patterns raise questions of historical significance for the Rhodian trade network and how this may have shifted from late Classical to Hellenistic times. In particular, all but one coin of the late Classical period contain detectable amounts of arsenic, an element abundant in the copper deposits of Cyprus, which also characteristically have a low lead content (Charalambous et al. 2014). In the rose/rose series, arsenic is present in two thirds of our sample. The consistent presence of arsenic in the nymphe/rose issue of the late Classical period shows that the copper ore(s) that reached Rhodes for minting also contained this element.

A totally different picture is outlined in the Helios/rose series, where arsenic is not detectable at all but one coin, which contains the highest amount of cobalt of all. In the plinthophoric issue, nearly one-third of our coins contain arsenic, while the remaining two thirds contain antimony. For example, the bronze coins of Marseille in series II also contain antimony (Barrandon and Picard 2007: Fig. 4, 42). Our intention is not to precisely locate the provenance of ores, but to underline the diversity in elemental composition of Rhodian bronze coins. It becomes apparent that the sources of copper ores on which the Rhodian mint relied for the issue of coinage changed in the span of two centuries, i.e. from the mid-4th c. BCE down to the 2nd c. BCE. A diversification in the copper ores is noticed from the late 3rd c. BCE onwards. For the Helios/rose issue, the Rhodians seem to have obtained copper ores from different networks, while for the plinthophoric issue in the 2nd c. BCE the network resources seem to have become further diversified, given the presence of antimony in some of the coins.

Silver alloy coins

The elemental composition of all silver coins analysed (Table S8), together with relative uncertainties deduced from the peak statistics are reported in Table S9. All samples are silver-based alloys bearing different values of alloying elements along with minor and trace elements present. The mean value of silver is 93 wt%, ranging from 74 to 100 wt%. Figure 5 shows the μ-XRF spectra of the minimum and maximum values of silver detected in the sample. Copper is present in 10/11 coins with mean and median values of ca 2 wt% and a range between 0.1 and 7 wt%. Gold is present in all coins with a mean of 1 wt% and a range of 0.03–2.5 wt%. Lead is detected in nine coins ranging from 0.1 to 16 wt% and a mean value of 4 wt%. Iron is present in seven coins, in values typically below 0.5 with one coin (no. R3) bearing 1.5 wt%. Bismuth traced in 2/13 coins has a mean value of 0.2 wt%.

Fig. 5
figure 15

μ-XRF spectra for the coins R8 and R7, bearing respectively the minimum and maximum silver values detected in the sample

It is worth noting that the intensity ratio of the characteristic Ag-Kα and Ag-Lα X-rays of silver alloys was utilized as a quantitative measure to assess the likelihood of changes due to silver enrichment and other forms of corrosion of their surface. The literature indicates that for silver alloys with a silver content of more than 72 wt%, the surface composition corresponds to the silver-rich primary phase (Ag = 92 wt%). For silver with higher purity (above 92 wt%), there is no difference—in terms of enrichment of the element at the surface—between the surface and bulk silver composition. For a silver alloy composition containing 92 wt%, simulations have shown (Kantarelou et al. 2011) that the 90% of the Ag-Lα intensity originates from a depth of about 7 μm, while the respective depth for the Ag-Kα line is almost six times more (40 μm). Thus, if a silver-enriched surface layer is present (with a thickness of e.g. 0–10 μm), it is expected that the Ag-Lα intensity would come into saturation mainly from the top layer rich in the element, while the Ag-Kα intensity would be lower due to the significant contribution from deeper layers with lower silver content. Consequently, the final silver K/L ratios—measured by the Ag-Kα and Ag-Lα intensities of reference silver alloys—should be lower compared to the Ag-K/L ratio for coins if the enrichment scenario holds. The intensity ratio of the characteristic Ag-Kα and Ag-Lα X-rays determined using the certified reference silver alloys was measured at 7.8 ± 0.5. By using this quantitative criterion, the values of six coins (R2, R5, R6, R8, R10, and R11) given in Table S9 are indicative, as their respective intensity ratios are either less than 7.2 or greater than 8.2 (Table S10).

Alloying elements and impurities

The major alloying elements are, beside silver, copper, and lead. The visualisation of the results is presented on the ternary diagram (Graph 11). Relying on the aforementioned, four groups can be identified based on the major elements contained in more than 1 wt% within the silver alloy. Thus, the assemblage includes three coins of fine grade (Ag > 990), two coins with silver grade in between 950 and 990, five coins with silver in between 900 and 950, and one with its silver value being below 900.

Graph 11
figure 16

Ternary diagram of silver, copper, and lead values in the silver coins’ assemblage

The elemental concentration range and other statistical information of the examined coins from different periods are presented in Table S11. Depending on the ore from which it was retrieved, silver contains different concentrations of minor elements. These elements composition must be evaluated with caution, since the information is not always representing the direct metal product extracted from an ore. The effect of alloying, recycling or impurities that may occur during the smelting process, must always be considered. Copper concentrations in excess of 0.5 wt% in 5/11 coins suggest a deliberate addition (Sheedy et al. 2020; Charalambous 2020); probably an attempt to increase the hardness of the alloy and to facilitate its processing (Grethe et al. 2020).

The presence of lead in silver alloys can be justified either from the smelting procedure of the silver ores (of argentiferous lead ores such as galena, cerussite, and anglesite) and the lead cupellation method or from its deliberate addition or of lead-containing copper ores. The variation in lead values and the fact that copper and lead values in some of the coins increase proportionately suggest that any of the three methods of lead addition could have been implemented. Silver deriving from cupelled ores usually contains lead in amounts of 0.1–1 wt% (Pernicka 2014). Since the lead content in four coins (R1, R2, R7, and R11) is below the generally accepted percentage present after the cupellation process, the use of native silver or ores that do not require cupellation, or possibly silver recycling that results in oxidation and elimination of lead during remelting, may be suggested. Lead is probably intentionally added in 7/11 coins, since its content ranges between ca. 1.0–16.0 wt%. Likewise, one coin (R8) has a very high lead content (15.8 wt%), which in combination with the high copper content (7.27 wt%) could indicate intentional debasement—a not so unusual tactic, considering the increased copper/lead content found in Ptolemaic silver tetradrachms from the period between 140 to 60 BCE (Faucher and Olivier 2020). In addition, the refinement during the cupellation process, in order to receive low lead values, is known to increase both the contents of silver and gold. Since gold is more noble than silver and has also a low vapour pressure at its melting point, no oxidation or evaporation should take place, leading to an elevated gold content within the silver alloy. Analysis detected gold in all coins, an element which can be used as an indicator of provenance (Pernicka 2014). Gold content up to about 0.5 wt% could have been retrieved from chlorargyrite and acanthite and the oxidized lead ores cerussite and anglesite (coins R2, R5, R7, and R11). The higher gold values (0.16–2.4 wt%) for the rest of the silver coins (except R11) are probably due to a long cupellation process or recycling of once gilded gold. High gold contents can also derive from jarosite ores, but these must be refined by cupellation with added lead, thus leading to a higher lead content. It should be noted that coins may also have been recycled or refined without the addition of lead. Trace elements are therefore not always an indication of origin and we cannot know whether these processes have taken place (Ponting 2020).

Bismuth in the silver metal can originate either from the silver ore or as an accidentally added impurity (L’Héritier et al. 2015). A comparison of the lead and bismuth content gives information on the efficiency of the cupellation process and can also be used as a discrimination index between ore sources. By adopting the plausible scenario that the bismuth content in the silver coins derives from the ore, oxidized and extracted along with lead after a long cupellation process, the above conclusion of fine refinement is supported; in fact, bismuth can only be traced in two coins, namely in R3 and R6, which also have elevated values of gold. The increased lead content in both coins with detected bismuth can be attributed to the intentional alloying with lead and not to a possible incomplete refinement process. Notwithstanding the possibility that there is silver enrichment at the surface of the coins, it is useful that the comparative results can provide information on alloying trends.

Remarks on the silver coinage assemblage of Rhodes

It is important to note that all but one of the specimens in the plinthophoric series (cat. nos. R3–11) are strikingly lighter than expected. The average weight of the hemidrachms (R3, 10–11) is 1.0 g (versus 1.5 g of the standard weight), while the average weight for the diobols is 0.3 g (versus 0.70–0.85 g) (Demetriou 2015: 172, Table 13). Burial in unstable conditions in combination with chemical treatment may have taken a toll on the coins, as reflected in their current state of preservation. Although the sample of the silver coins analyzed is small (11 in total) and firm conclusions are difficult to obtain, it is nevertheless interesting to note that the composition of the silver coins under analysis displays a striking degree of variability. This is in stark contrast to the picture that emerges from the analysis of 12 Rhodian silver coins by Barrandon and Bresson (1997), who used the technique of Proton Activation Analysis to study the composition of Cretan imitations and Rhodian coins. The authors found that the Cretan imitations were made of low-grade silver and copper alloys, while the Rhodian coins consistently had a high silver content (between 98.5 and 99.5 wt%). The Rhodian coins also exhibited a consistent pattern of alloy composition and refining techniques (negligible copper below 0.5 wt% Cu and lead below 0.5 wt% Pb), suggesting a centralized production process.

This study offers valuable insights into the production and trade of ancient currency and contributes to our understanding of the economic and cultural practices of the period. But it should be stressed that in both analyses the sample is very small (11 coins in our analysis vs 12 coins in Barrandon’s & Bresson’s study). Furthermore, the sample includes a range of denominations of different series. In other words, the composition of Rhodian silver coinage is still far from being firmly established. This underlines the necessity of future analyses of larger samples of silver coins if we wish to better understand the evolution of Rhodian silver coinage in Hellenistic times.

Conclusions

The present study provides new insights into Rhodian coin production from the 4th c. BCE to the 2nd c. CE. Although the analysis does not include all bronze issues of the Rhodian mint, it allows some initial observations, especially on the evolution of chalkoi, from the mid-fourth century BCE to the late second century BCE. Based on the μ-XRF results, different groups of copper-based alloys are formed with tin and lead as their major alloying components. The elemental compositions showed that most of the coins consist of binary, leaded and ternary bronzes. The fact that the Common Era (CE) coins have a narrower distribution of tin (> 12.5 wt%) than the pre-Common Era (BCE) group (7–24% wt%) raises questions. However, since the former represents only 1/7th of the total sample, it is too small to draw solid conclusions regarding preferred recipes. This is pretty much the case for the pre-Common Era group, apart from the second century BCE coins and most coins from the period between 350 and 300 BCE. The consistently high ratio of tin in the alloy during the Late Classical and Hellenistic periods suggests that Rhodian coinage did not follow the trend of increasing lead to the detriment of tin, which distinguishes it from developments observed in the evolution of bronze coinage in other cities. Instead, Rhodian coinage remained consistently uniform in its alloy composition throughout, with most coins primarily featuring a binary alloy.

The study also revealed that the Rhodian mint attached great importance to the appearance of the coins, whose metal had only a fiduciary value. This consistency of tin concentration in the alloy indicates an uninterrupted supply network for tin and underlines a long metallurgical tradition in Rhodian coinage. Accordingly, the k-means clustering method allowed the identification of compositional patterns in the raw materials used for coinage based on their trace element composition. All impurities present in the alloys are typical of copper ores and do not deviate from what is expected for the period studied. Thus, valuable insights into the origin-related aspects of Rhodian coin production could be gained. Above all, it was confirmed that this aspect of Rhodian coin metallurgy is not uniform over the course of six centuries and could be influenced by factors, such as the availability of raw materials.

With regard to the 11 silver coins and based on the major alloying elements, four groups of different silver grades can be identified. Despite the limited character of our sample, some important technological observations can be made. In almost half of the silver coins, copper should be considered an intentional addition, as it exceeds by far 1.5 wt%. The deliberate addition of copper might also explain the high levels of lead in the coins, in cases exceeding 1 wt%. The varying presence of gold in the absence of ore-correlated trace elements like bismuth, support the hypothesis that the cupellation method and a degree of proper refinement was conducted for most coins. Bismuth is detected in only two coins, both plinthophoric issues, though of different denominations (a hemidrachm and a diobol respectively) with a very low and a very high concentration of copper respectively.

Overall, the results contribute to a better understanding of the technological and economic aspects of Rhodian silver and bronze coin production, corroborating our body of knowledge. It is clear, though, that a more extensive sample is needed to provide a basis for further research.