An approach to the metallic composition of the Carthage mint coins from the tetrarchic hoard of Tomares (CA. 312 CE)

Journal of Archaeological Science: Reports 44 (2022) 103509 Available online 13 July 2022 2352-409X/© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/bync-nd/4.0/). An approach to the metallic composition of the Carthage mint coins from the tetrarchic hoard of Tomares (CA. 312 CE) Simona Scrivano a , e , Ruth Pliego b , Blanca G´ omez-Tubío c , e , Javier Moreno-Soto e , Enrique García Vargas b , * , Miguel ´ Angel Respaldiza d , e , Francisca Chaves Trist´ an b a Centro de Investigaci´ on Tecnol´ ogica e Innovaci´ on–Laboratorio de Rayos X, Universidad de Sevilla, Avda. Reina Mercedes s/n 41012 Seville, Spain b Departamento de Prehistoria y Arqueología. Facultad de Geografía e Historia, Universidad de Sevilla, c/ María de Padilla s/n, Sevilla 41004, Spain c Departamento de Física Aplicada III. Escuela T´ ecnica Superior de Ingeniería, Universidad de Sevilla, Camino de los Descubrimientos, s/n. 41092, Seville, Spain d Departamento de Física At´ omica, Molecular y Nuclear. Universidad de Sevilla, Av. de Reina Mercedes s/n, 41012 Seville, Spain e Centro Nacional de Aceleradores, (Universidad de Sevilla-CSIC-Junta de Andalucía), C/ Thomas A. Edison 7, 41092 Seville, Spain ARTICLE INFO Keywords: Archaeometry X ray fluorescence Metal alloys Late Roman Numismatics Tetrarchic nummi ABSTRACT A group of 533 nummi from to the Tomares hoard, found in Seville in 2016, has been studied to characterize their metallic composition. The coins selected were struck in the mint of Carthage between AD 297 and 307. The statistical analysis of the non-destructive XRF data increase our understanding of a mint which has been paid less attention that others to date, in both material and historical terms. No differences in the composition of the coins were detected; i.e. the same alloy was used in all four of the mint’s officinae. However, coins minted at the beginning of the first tetrarchy were found to present higher concentrations of silver. 1. Introduction The hoard of Tomares (Seville) was discovered on 27 April 2016 during public construction works in the park El Olivar de El Zaudín, and on the same day it was taken to the Archaeological Museum of Seville. The hoard consists of nineteen Roman oil amphorae of the Tejarillo 1 type, used as containers for nummi of the Tetrarchic period. Most of the amphorae were damaged by the mechanical digger, but nine of them were recovered intact and have not been opened yet. Therefore, the exact number of coins in the assemblage remains unknown, although the contents of the fractured amphorae suggest that the whole hoard consists of approximately 55,000 pieces. Three thousand pieces from the broken amphorae, all the coins (2797) from the excavation of one of the fractured amphorae (amphora 11) and 102 coins found during the excavation of the site have been catalogued to date. The most recent coin to have been identified thus far is dated to AD 311–12. (Chaves Trist´ an, 2017; V´ azquez Paz and Garrido Gonz´ alez, 2017). The Tetrarchy, established by Diocletian (AD 284–305) in AD 294, involved the division of political authority between two co-regent emperors, with the title of Augustus, and two caesars, who were to become augusti after the scheduled resignation of the head rulers. This system survived the first abdication of augusti Diocletian and Maximian, and the subsequent inauguration of the second tetrarchy with Constantius and Galerius as augusti in May AD 305. However, the death of Constantius in July AD 306 destabilized the system, when the personal ambitions of Constantine, son of the late augustus Constantius, and Maxentius son of Maximian, came into play. While Constantine managed to enter the regular tetrarchic system in May AD 306, Maxentius remained the de facto (i.e. unofficial) augustus until his final confrontation with Constantine at the battle of Ponte Milvio (AD 312) where he perished. In AD 294, at the beginning of the Tetrarchy, Diocletian undertook a monetary reform (Lafaurie 1975; Ermatinger, 1990a; Sutherland, 1961; Depeyrot 1992) with the imposition of a comprehensive system of values in which the weight and metal content of coins were fixed. This reform transformed gold and silver denominations, now presented as aureus solidus and argenteus respectively (Cf. Chameroy 2016), and created a new bullion coin, the nummus, sometimes called the follis in the numismatic literature. Its initial value was 1/32 of a pound, in coins weighing approximately 10 g, with a silver content of ca. 4%. Coins were struck in sixteen mints founded in strategic points of the Empire (Fig. 1): Londinium (London), Treveri (Trier), Lugdunum (Lyon), Ticinum (Pavia), Aquileia, Rome, Ostia, Siscia (Sisak), Serdica (Sofia), Thessalonica, Heraclea (Marmara Ere˘ glisi), Nicomedia (Izmit), Cyzic (Aidinjik), Antioch, Alexandria and Carthage (Tunis). * Corresponding author. E-mail address: [email protected] (E. García Vargas). Contents lists available at ScienceDirect Journal of Archaeological Science: Reports journal homepage: www.elsevier.com/locate/jasrep https://doi.org/10.1016/j.jasrep.2022.103509 Received 7 July 2021; Received in revised form 9 May 2022; Accepted 3 June 2022 Journal of Archaeological Science: Reports 44 (2022) 103509 2 The study of the coins from the Tomares hoard is of great importance for our understanding of monetary circulation in this turbulent period for the Hispanic provinces. These territories depended on the diocese of Gaul, and were arguably peripheral, but were strategically important owing to their proximity to the African diocese and the Strait of Gibraltar. In addition, the hoard is also a significant piece of evidence concerning the recurrent decreases in silver content and coin weight as a result of scarcity in the supply of precious metal and economic difficulties, despite the ostensible monetary stabilization introduced by the AD 294 reform. This work focuses on the 675 coins from the Tomares hoard minted in Carthage (present-day Tunis), 533 of which were analyzed by X-ray Fluorescence (XRF). The sampled coins were among those that, coming from the fractured amphorae, were scattered across the site by the mechanical diggers. This groups is often referred by the study team as the “general assemblage” (to distinguish them from the coins that remain inside the amphorae). There are many reasons for selecting the mint of Carthage, admittedly a peculiar choice in the general historiographical landscape of tetrarchic mints, for this study. First, the mint of Carthage differs, in typological terms, from the Genius Populi Romani, adopted almost universally by the period’s mints. The seriation of the mint’s issues, as well as their chronology, remain contested, and different interpretations (Hedges and Robins, 1963; Sutherland et al., 1966; Kos 1988) have been proposed for those issues whose structure and chronology are uncertain or not well known. The percentage of Carthaginian pieces in tetrarchic hoards is generally quite striking, both in eastern and western ones (Ermatinger, 1990b). This fact, which has also been noted in the Tomares hoard (Chaves Trist´ an, 2017), means that there is a considerable number of specimens with which to work. Approximately 20% of the coins in the hoard were struck in Carthage; the other two mints to approach this percentage in the “general assemblage” are Ticinum and Rome, with approximately 20% of the total assemblage each. Far behind are Treveri, Lugdunum, Aquileia and Londinum, with over 15% and 5% of the total assemblage each. Most tetrarchy nummi analyzed to date come from northern European, Italian and Eastern mints, and coins issued in Carthage are rare (see Table 1). In addition, these coins have been analysed with a variety of analytical methods, which involves different levels of accuracy and technical specifications: gravimetric chemical analysis (Cope, 1968); EPMA analysis with microperforation (Besly and Webster, 2002); and non-destructive techniques, such as neutron activation (Bollard and Barrandon, 2006), X-ray fluorescence (Fawcett et al. 2018, Hunt Ortiz, 2017, coins from the Tomares hoard found during archaeological excavation after the initial find) and particle induced X-ray emission (Rizzo et al. 2011). The results are, therefore, heterogeneous, as shown in Table 1. This illustrates the importance of obtaining non-destructive measurements from a large number of specimens under the same experimental conditions. Conditions were thus ideal for a thorough investigation of coins from the Carthage mint, as the first step of a broader research program which aims to examine the metal content of the coins present in the hoard, with particular emphasis on the alloys and the manipulations of their concentration in the different issues over time. It is well known that nummi were minted with a quaternary alloys of copper, silver, lead and tin (Cope, 1968; Cope and Billingham, 1967, Cope, 1969, Cope, 1971; Hedges and Robins, 1963; Ravetz 1963). Silver content was invariably low and was distributed both in the surface plating and the pieces’ core. The coins were analysed by X-ray Fluorescence, a fast technique that is ideally suited to the analysis of large assemblages. Finally, Carthage is one of the key mints for the study of the economic consequences of Maxentius coup in AD 306, as it is one of the mints that fell under his control (Drost 2013). 2. Material and methods 2.1. The tetrarchy issues from Carthage: A numismatic sample The mint of Carthage was divided into four officinae or workshops. Although the territory of the African diocese was first under the rule of Fig. 1. Localization of the tetrarchic mints around 312 CE. S. Scrivano et al. Journal of Archaeological Science: Reports 44 (2022) 103509 3 Maximian and then of Constantius, each of these officinae issued coins in the name of one of the four tetrarchs. Sometimes, the coins issued in Carthage in the name of the eastern rulers featured an I (for Jupiter) on the reverse, and those minted in the name of the western emperors featured an H (for Hercules), which were the tutelary divinities of the two imperial colleges. Our numismatic sample consists of coins issued from AD 297 to AD 307. 1 No coins minted by the usurper Domitius Alexander, active in Carthage between AD 308 and AD 311 were found in the assemblage. Details about the sample are listed in Table 2, including total number of coins, and coins analyzed by period, group and issue. 2 The chronological distribution of the coins under study is shown in Fig. 2, which includes the number of coins struck by each officina. The most numerous issues in the sample are those corresponding to the period AD 299–303, followed by those minted during the second tetrarchy. Finally, our sample also contains a few pieces (25) struck by Maxentius. This number may be too small to argue for a statistically significant trend. The study of this numismatic sample required the combination of different methodologies and an interdisciplinary approach: cleaning, stabilisation and preservation; physical–chemical characterisation by non-destructive methods; and, finally, historical-numismatic analysis. 2.2. Surface cleaning and conservation In order to clean the coins, they were first immersed in demineralized water with neutral detergent, and later brushed to remove the dirt from the surface. Following this, they were immersed in pH-controlled sodium sesquicarbonate to eliminate some of the corrosion-related matter, especially chlorides. Finally, those coins that showed chloride contamination were stabilised with Benzotriazole (BTA). 2.3. Experimental set-up The chemical characterisation of the alloys was carried out with Xray fluorescence (XRF) using a FischerScope X-Ray XUV 773 XRF device at the Centro de Investigaci´ on, Tecnología e Innovaci´ on de la Universidad de Sevilla (CITIUS). The device is equipped with an X-ray tube with a molybdenum anode, and beam size between 3 and 0.6 mm, and a SDD detector with a resolution of 140 eV (FWHM for Mn-K α ). A voltage of 50 kV, and current of 500 µA, and a spot size of 1 mm were applied, using a 500 µm aluminium filter. In order to increase the reliability of the analysis, three readings were taken from both sides of each coin. The average concentration values of each element were calculated. The quantitative analyses of the coins was carried out with the aid of the inbuilt WinFTM® software, using the fundamental parameters (FP) method. A total of 533 coins were analyzed. A bronze matrix for UE40 standard from the Centre Technique des Industrie de la Fonderie (certified composition: 5.45 wt% Ni, 0.06 wt% Fe, 84.1 wt% Cu, 2.95 wt% Zn, 0.32 wt% Pb, 7.0 wt% Sn) was used to establish the detection limits (LOD). For this matrix, LOD for Fe (6.4 keV) is approximately 30 ppm (ppm=µg/g); c. 300 ppm for Cu and Ni (7.47 and 8.04 keV respectively); c. 1500 ppm for Zn; 100 ppm for Pb; and 500 ppm for Sn. The LOD for tin is taken as reference for silver and antimony, and that of zinc for arsenic. The relative variations between the certificated composition and obtained values by WinFTM® are: 3.9% for Ni; 0.2% for Cu; 1.0% for Zn; 2.9% for Sn; and 9% for Pb. The analytical depth in these coins is mainly determined by the attenuation of the X-rays to be detected, Cu, Pb, Ag and Sn, and is estimated to about 45 µm. As it is well known, nummi are characterised by presenting a layer of silver on the surface. In order to estimate the thickness of this layer, four coins were cross-sectioned and analyzed by electron microscopy. It was found that the Ag-layer patina, far from being homogeneous, is thoroughly fragmented with an average thickness of about 2 µm, as shown in Fig. 3. Rizzo et al. (2011) also found that the silver was stratified in very thin layers about one micron thick. The differences in the amount of silver present in each coin revealed by the results of bulk analysis recommends the use of analytical procedures that can look beneath the superficial layer without physically altering the coins, however minimally. An initial and tentative approach to this problem was tested with four pieces from the sample whose cores could be reached because they were polished in a small area. These polished areas were analyzed under the same conditions (see Table 3) leading to an average increase of 1.5 wt% Ag. This value tallies with the results reported in Guihard et al. (2018), who also analyzed microscopically polished surface areas. Table 1 Metal composition of tetrarchy coins from the mint of Carthage analyzed by various methods, showing the weight percentage (wt%) of Cu, Sn, Ag, Pb, Fe and Zn. Coin n. 8 has different codes in two different publications. Data are available in the literature. N. Inv. Source Technique Date A.D. RIC VI Cu Sn Ag Pb Fe Zn 1 BNF 1995/941 Bollard and Barrandon, 2006 Neutron activation 297–298 19a 93.6 1.04 3.48 1.52 0.17 Nd 2 B.546 Besly and Webster (2002) Electr´ on probe microanalysis 297 21b 92.5 0.78 3.33 2.53 0.78 0.78 3 444 Cope et al. 1997 Grav. chemical analysis 295–301 32b 1.86 4 BNF 8457 Bollard and Barrandon, 2006 Neutron activation 305–306 39a 88.9 2.24 2.27 6.21 0.15 Nd 5 445 Cope et al. 1997 Grav. chemical analyisis 295–301 30b 1.88 6 466 Cope et al. 1997 Grav. chemical analysis 303 35b 0.01 7 528 Cope et al. 1997 Grav. chemical analysis 305–307 40b 2.02 8 520 Cope et al. 1997 Grav. chemical analysis 307 57 87.28 6.37 1.79 3.81 0.36 0.04 9 521 (9) Cope et al. 1997 (Cope, 1971) Grav. chemical analysis 307 60 81.25 5.45 1.20 11.90 0.03 0.01 10 6 Hunt Ortiz, 2017 XRF 297 71.59 0.83 26.45 1.13 11 33 Hunt Ortiz, 2017 XRF 305–306 2–1 87.72 1.44 7.45 3.38 12 50 Hunt Ortiz, 2017 XRF 305–306 3–1 87.43 4.20 2.10 6.27 1 More recently, Vincent Drost (2013) has reorganized the issues of Maxentius in Carthage, including one (nº 4) to those listed by Sutherland in RIC. VI. He has also suggested new dates for the various issues: 1 (Group IV-i of RIC VI): late AD 306 - early spring AD 307; 2 (Group IV-ii of RIC VI): spring AD 397; 3 (Group IV-iii of RIC VI): late spring-early autumn AD 307; 4 (no correspondence in RIC VI): summer AD 307; 5 (Group IV-iv of RIC VI): late summer-autumn AD 307. The structure and chronologies of the issues given in RIC VI (see not. 2) have been kept, to keep consistency with the series issued by other mints. In addition, in our sample, there are no examples of Drost’s issue 4. 2 A group consists of a homogeneous set of issues ordered by the same authority during their rule. Each group features obverse portraits and reverse titulature corresponding to the augusti and caesars of the period to which it belongs. Within each group, issues are the number of coins ordered by the treasury from each mint in a given period. Issues are usually identified and dated by an alphabetical formula on the exergue and a set of letters in the field, in combination with the type-legend (cf. Sutherland et al. 1966, Bastien 1980). S. Scrivano et al. Journal of Archaeological Science: Reports 44 (2022) 103509 4 3. Results 3.1. General composition of alloys The X-ray fluorescence data confirm that the alloy used in the 533 analyzed coins is quaternary: copper, lead and tin with the addition of silver. In addition, traces of iron, nickel, zinc, arsenic and antimony were also found, although they could not be detected in all coins. Nickel was found to be present above LOD in 502, and zinc, arsenic and antimony were detected in 275, 144 and 449 coins, respectively. Table 4 shows the average concentration of the elements detected in the sample, which covers a time span of ten years, along with the standard deviation and the maximum and minimum values. The main element in the coins is copper, with an average concentration of 88.1 wt % followed by silver (4.5 wt%), lead (3.6 wt%) and tin (3.2 wt%). Iron, zinc and arsenic account for approximately 0.25 wt%, while nickel and Table 2 Summary of issues present in the study sample. Period Minter Group Issue Typology/Legend Exergue/Field Date N◦ex. Analysed/Total Date Officinae Issue Group 1 2 3 4 Ist Tetrarchy Maximian 1 2 Felix Advent Avgg NN PKP-Q 297 1/1 1/1 1/1 3/3 389/ 493 3 Felix Advent Avgg NN A-Δ/-//PKP-Q 297 4/6 4/6 7/8 6/6 21/26 4 Felix Advent Avgg NN I or H/-//PKP-Q 297–298 15/19 7/10 20/26 12/18 54/73 5 Felix Advent Avgg NN I or H/I or H// PKA-Δ 298 6/8 11/11 7/12 2/2 26/33 6a Salvis Avgg et Caess Avcta Kart A-Δ 298–299 1/1 4/4 1/1 1/1 7/7 6b Salvis Avgg et Caess Fel Kart A-Δ 298–299 19/20 12/14 11/16 19/23 61/73 7 Salvis Avgg et Caess Fel Kart A-Δ 299–303 65/80 48/72 49/58 47/59 209/ 269 8 Salvis Avgg et Caess Fel Kart I or H/-//A-Δ 303 2/3 3/3 3/3 8/9 2nd Tetrarchy Constantius Severus 2 1 Salvis Avgg et Caess Fel Kart H or I/-//A-Δ May 305-July 306 22/31 24/30 19/26 28/38 93/ 125 96/128 2 Providentia Deorvm Qvies Avgg S/F//PKA-Δ May 305-July 306 3/3 3/3 3rd Tetrarchy 3 1 Salvis Avgg et Caess Fel Kart I or H/-//A-Δ July-Autumn 306 10/10 4/4 8/9 5/6 27/29 27/29 Maxentius 4 1 Salvis Avgg et Caess Fel Kart H/-//A-Δ Late November 306 1/1 9/10 4/4 14/15 21/25 2 Conservator Africae Svae H/ER//A-Δ Late 306-Starting 307 0/1 0/1 1/1 1/3 3 Conservator Africae Svae SE/F//A-Δ Late 306-Starting 307 0/1 2/2 1/1 2/2 5/6 c. 307 4 Conservatores Africae Svae PKA-Δ c. summer 307 1/1 1/1 Total 149/ 185 117/ 155 136/ 171 131/ 164 533/675 Fig. 2. Distribution of emissions by officina and date. Fig. 3. SEM image: surface and cross-section of a coin. Table 3 Silver content in the polished and surface areas. Coin Ag (wt%) Surface Ag (wt%) Polished Ag_sur.-Ag_pol (wt%) 1 3.12 1.91 1.21 2 3.65 1.50 2.15 3 2.97 1.75 1.22 4 2.39 2.34 0.05 S. Scrivano et al. Journal of Archaeological Science: Reports 44 (2022) 103509 5 antimony account for under 0.07 wt%. However, the maximum and minimum values of some elements in the quaternary alloy present significant deviations from the average value, as much as by between 30% and 58%. This dispersion of the data could indicate that the alloy used throughout the Tetrarchy was not homogeneous. Two possible explanations exist for these differences in metal composition: first, which we think unlikely, is the use of different alloys in various workshops operating simultaneously; second, much more likely, the consistent manipulation of the alloys in all workshops over time, as a result of intentional monetary devaluation, involving a decrease in the amount of silver and its partial substitution by other elements. 3.2. Composition of alloys by workshop or officinae As previously noted, the mint of Carthage had four workshops or officinae. Each one issued coinage in the name of one of the four members of the tetrarchy regardless of the emperor who had jurisdiction over the territory in which the mint was located, in this case Maximian, Constantius and Maxentius, successively. Fig. 4 shows the mean values and the standard deviation of the main elements of the alloy in each officina. Coins from officinae 3 show slightly higher average silver content. This is due to the fact that coins struck in officina 3 in the period AD 297–298, which present a slightly higher concentration of silver, are numerically overrepresented in our sample, as illustrated in Fig. 5. However, the standard deviation is broad enough to suggest the use of the same alloy in all workshops. Henceforth, the coins will be analysed regardless of their specific workshop of origin, because the composition of the quaternary alloy is very similar. 3.3. Alloy composition: Evolution over time Previous studies (Cope et al. 1997, Bollard and Barrandon, 2006, Guihard et al. 2018) suggest that the composition of the quaternary alloy used in tetrarchy coins vary from period to period, showing a decrease in the content of precious metal over time. One of the aims of the study was to establish variations in major element composition (Cu, Pb, Ag, Sn) over time and between different issues. To establish if the elemental differences observed are significant, the Kruskal Wallis test (Quispe et al., 2019) was used. This method, which is equivalent to the one-factor analysis of variance (ANOVA) method, because the distributions of the elements do not comply with the principle of normality and homoscedasticity required by parametric analyses. The Kruskal Wallis test allows testing the null hypothesis, which states that different samples are equally distributed, in contrast to the alternative hypothesis, in which the distribution of at least of two samples differ significantly. A significance level of 5% (p value of 0.05) is sufficient to reject the null hypothesis. The significance levels yielded by the Kruskal Wallis test for copper, silver, lead and tin is p<0.05, suggesting that the composition of the quaternary alloys present in the coins shows significant differences over time and between issues. In addition, the p value for silver and tin was much lower than that yielded by copper and lead, indicating that the compositional bracket for copper and lead are narrower. Fig. 5 shows the mean concentration of copper, lead, silver and tin by chronology. The error bar is the standard deviation, which is null for the issues represented by a single specimen as seen in Table 2. Some clear trends can be observed: a slight decrease in the average concentration of copper; a more substantial decrease in the average concentration of silver; and an increase in the average concentration of lead and tin. Coins minted during the early years of the Tetrarchy, between AD 297 and AD 298 (issues 2 to 5 of group 1) are characterized by a very low average amount of tin (below 1.5 wt%, and in most cases below 1 wt%), and the highest concentration of silver, (c. 6.5 wt%). On the other hand, these coins present a mean composition of approximately 2.9 wt% for lead and 89.4 wt% for copper. From issue 6a onwards a significant decrease in silver and an increase in tin are attested. These two elements come to be found in similar concentrations, namely 3.7 wt% for silver and 3.4 wt% for tin. Thereafter, the alloy remains stable, and no significant changes are detected in the transition to and during the second tetrarchy (issues 1 and 2 of group 2) except in the last year of the first tetrarchy, in AD 303 (group 1, issue 8), which is probably a reflection of the small number of specimens corresponding to this issue (8 in total) in the sample; these coins present Table 4 Average composition of the 533 coins from Carthage under anaysis. For trace elements, only values above the LOD were used. a) 502 coins b) 275 coins, c) 144 coins, d) 449 coins. Element Wt % Dev. Max. Min. Cu 88.1 2.8 94.9 72.6 Ag 4.5 1.5 10.67 1.99 Sn 3.2 1.6 7.75 0.15 Pb 3.6 2.1 18.9 0.27 Fe 0.20 0.11 0.83 0.04 Ni a) 0.042 0.007 0.08 0.03 Zn b) 0.27 0.34 5.16 0.15 As c) 0.21 0.25 3.09 0.15 Sb d) 0.072 0.015 0.118 0.05 Fig. 4. Average concentrations of Cu, Ag, Sn and Pb for the four officinae in the mint of Carthage. Fig. 5. Evolution of the composition of the monetary alloy over time. S. Scrivano et al. Journal of Archaeological Science: Reports 44 (2022) 103509 6 highly variable lead (minimum 1.98 wt%; maximum 8.3 wt%) and somewhat less variable tin (minimum 2.78 wt%; maximum 5.79 wt%). The high lead content could be a result of segregates; it is well known (Ingo et al. 2004; Mata et al. 2010a, Mata et al., 2010b) that lead does not alloy well with copper or bonze, resulting in the generation of lead segregates instead of a homogeneous mixture. In the early months of the third tetrarchy, from July AD 306 (issue 1 of group 3) to November AD 306 (issue 1 of group 4) (41 coins), the alloy contains an average amount of 3.5 wt% silver, 4.3 wt% tin, 87.5 wt% copper and 4 wt% lead. This reflects a slight decrease in the concentration of copper and a slight increase in that of lead and tin vis-` a-vis first tetrarchy coins. Finally, coins struck in the months that followed present significant compositional fluctuations, but this is likely due to the low number of coins analyzed for this period (7 coins) which, in combination with the standard deviation, precludes us from reaching any conclusion. 3.4. Correlation between the elements of the alloy The Kruskal Wallis test yielded similar p values for copper and lead and much lower values for silver and tin. In addition, changes in composition over time suggests a correlation between silver and tin contents, and between copper and lead contents. Spearman correlation coefficient (r s ) which is a non-parametric measure of rank correlation (Quispe et al., 2019), was used to analyse these correlations. Lead and copper present the highest correlation coefficient (rs =– 0.84), followed by tin and silver (rs =– 0.54). The other possible combinations yield coefficients between –0.5 ≤r s ≤ +0.5, which are indicative of a weak correlation. Fig. 6 illustrates the concentration of lead versus copper (top) and tin versus silver (bottom). A negative linear relationship, which tallies with the symbol and high value of the relevant Spearman correlation coefficient, is clear. This behavior, i.e. the increase of lead in correlation with the decrease in copper could be due to the addition of small quantities of lead (typically between 2 and 10 wt %) in bronze alloys to improve metal fluidity and bring down the melting point, especially in cast items (Scott 1991, 24). However, Pb and Cu do not make for a homogeneous alloy; rather, Pb tends to form unmixed nodules of more or less pure lead in the bronze matrix, distributed according to the melting technique used. On the other hand, although silver and tin also seem to form a negative linear distribution in the plot, it is hard to argue for a correlation. In order to establish if the correlation between copper and lead is also attested elsewhere, we have analyzed existing data from other mints, as it has not yet been possible to apply this methodology to the other mints represented in the Tomares hoard. Fig. 7 illustrates the copper and lead content found in coins from different mints as published by other authors (Cope 1968, Cope and Billingham, 1967, Cope, 1971, Besly and Webster, 2002, Romano et al. 2012, Bollard and Barrandon, 2006). The data were taken from the tables published with these works, except for the coins from the Misurata hoard (Romano et al. 2012), which are the average copper and lead values of the different mints represented in the hoard. In order to simplify the graphic representation of the data, F only the average values for each group and issue from the Tomares hoard is represented. The linear trendline between copper and lead is clear. 3.5. Trace elements Trace elements detected include Fe, Ni, Zn, As and Sb, but not all of them are, however, present in all specimens, as shown in Table 4. The average values for each of the periods and issues of the Carthage mint are show in Table 5. Most notable is the absence of Zn in the early years, which match with the period in which the coins have the lowest Sn concentration and the highest silver content. Zn content range between 0.19 wt% and 0.31 wt%. In the period AD 299–303, Zn contents vary widely, owing to the presence of three coins with the highest concentrations of zinc (5.16 wt%, 2.24 wt% and 1.24 wt%) in the whole assemblage. Regarding the other trace elements, the contents of iron, nickel, antimony and arsenic are similar during the three tetrarchies, except for issue 2 of group 1, which is characterized by lower iron contents (0.10 wt%), and the highest concentration of arsenic. According to L. H. Cope (1968), Ni, Co, As and Sb impurities came into the mix with the copper ore. Fe, meanwhile, can be introduced into the alloy with the tools used in the casting process, coming unintentionally into the composition of the coins. On the other hand, the four polished coins present iron concentrations of approximately 0.3 wt% in bulk composition, so it is unlikely that the iron content was the result of contamination during manufacture, but as an impurity in the copper ore. Another trace element detected in some of the specimens is mercury. This element has been attested (Romano et al. 2012) in 221 specimens struck in Rome, Constantinople, Ticinum and Aquileia between AD 320 and AD 333, found in the Misurata hoard. Concentrations range from 0.020 wt% to 0.117 wt%. Romano and co-workers suggest that the presence of Hg is related to the use of silver-amalgam technology in the production of the superficial silver layer (patina). Considering the significant typological similarity between the Tomares and the Misurata hoards, it was decided to study the presence of Hg in all the coins in our Fig. 6. Average contents of lead versus copper (top) and silver versus tin (bottom) in the samples analyzed. S. Scrivano et al. Journal of Archaeological Science: Reports 44 (2022) 103509 7 sample. Hg is present in 7 coins, although not in every area under analysis, and the average values reported in Table 6 range from 0.153 wt % to 0.559 wt%. In the Tomares hoard, only 1.3% of the sample show traces of Hg, although in greater concentration than in those from the Misurata ones. In addition, the Tomares coins that present traces of Hg are earlier than their counterparts in Misurata, that is, between AD 298 and AD 306. With these data, it is difficult to establish whether the presence of Hg is related to the use of the amalgam technique, but it is worth emphasising that all the analyzed coins come from the same context and have been subject to the same preservation protocols. Additional research is needed to establish the role of Hg in the production of the superficial silver layers. 4. Discussion The coins from Carthage found in Tomares show a clear drop in silver content from emission 3 of group 1 (AD 297–298 onwards). This clearly suggests a deliberate policy (reducing the silver content in the alloy) to debase the coins, although this relative decrease in the nummi’s intrinsic value was not matched by an equivalent change in their theorical purchasing power (nominal value) at least until the edict of AD 301. Bollard and Barrandon (2006) also detected a gradual increase in lead and tin content and a reduction in silver in their sample (257 coins from various eastern and western mints). The pieces (25 in total) minted prior to AD 297 present Sn concentrations below 2.5 wt%, and lead concentrations below 5 wt%. This tallies with our results in terms of both elements and chronology: Sn contents not in excess of 1.2 wt% and lead contents of approximately 3 wt%. This phenomenon is also attested in other contemporary mints whenever the number of specimens analysed is relatively large. For instance, LH Cope (1968) analysed twelve coins from the Lugdunum mint (Lyon) (out of a total of 309 coins); three of these pieces, which were struck prior to AD 298, present a tin percentage of 1.2% by weight, while nine coins from AD 298 to AD 307 present tin contents of over 5% by weight (cf. Cope 1968). Concerning lead, the average content rises from 1.80 wt% prior to AD 297 to an average of 6.5 wt% afterwards. This was later confirmed by Bastien (1980), who analysed fourteen additional coins. It was also found that the decrease in the Sn-Pb ratio was not accompanied by a decrease in silver contents in AD 297. The quantities of Zn found in the coins from issue 6a of group 1 (298 CE) may be the result of the use of a fresh alloy from this date onwards. Zinc tends to disappear during recycling, but our sample shows the opposite trend (Dungworth 1996). It does not seem that the appearance of a new alloy from the issues struck after AD 297 is accidental or the result of an arbitrary decision by the monetary authorities, at least as far as the Carthage mint is concerned. Rather, this is likely the outcome of a deliberate decision to alter the composition, and thus the intrinsic value, of the nummi, reducing their cost and keeping their appearance and weight, so as not to undermine public trust in a coin that was somewhat less valuable than the ones minted in the early years of the tetrarchic reform. Fig. 7. Copper-lead correlation of Roman coins minted between 295 and 333CE The averages of the different issues from Carthage are shown together with coins from different mints taken from the literature. Table 5 Average concentration (wt%) and standard deviation of trace elements of the Carthage mint by period. No standard deviations are provided for periods represented by a single coin. Period Date A.D. GroupIssue Fe Ni Zn As Sb Wt% Dev. Wt% Dev. Wt% Dev. Wt% Dev. Wt% Dev. 1st Tetrarchy 297 1–2 0.10 0.05 0.037 0.002 0.14 0.085 0.035 297 1–3 0.22 0.09 0.037 0.007 0.25 0.07 0.072 0.014 297–298 1–4 0.23 0.13 0.035 0.005 0.21 0.06 0.079 0.018 298 1–5 0.20 0.08 0.042 0.007 0.18 0.03 0.075 0.013 298–299 1-6a 0.18 0.07 0.042 0.007 0.19 0.03 0.20 0.071 0.015 298–299 1-6b 0.16 0.10 0.042 0.006 0.21 0.04 0.17 0.02 0.070 0.012 299–303 1–7 0.17 0.09 0.041 0.006 0.29 0.46 0.18 0.04 0.070 0.014 303 1–8 0.24 0.07 0.044 0.007 0.27 0.08 0.16 0.072 0.019 2nd Tetrarchy May 305-July 306 2–1 0.24 0.12 0.045 0.007 0.27 0.11 0.32 0.62 0.071 0.015 May 305-July 306 2–2 0.19 0.07 0.051 0.007 0.31 0.20 0.074 0.019 3rd Tetrarchy July-Autumn 306 3–1 0.27 0.15 0.048 0.007 0.22 0.07 0.18 0.04 0.068 0.011 Nov 306 4–1 0.21 0.12 0.044 0.004 0.29 0.18 0.18 0.04 0.074 0.014 End 306-Starting307 4–2 0.25 0.054 0.061 End 306-Starting307 4–3 0.22 0.04 0.039 0.009 0.19 0.05 0.069 0.016 307 4–4 0.12 0.042 0.070 Table 6 Coins with traces of mercury in some areas under analysis. Coins Date AD Hg (wt%) Ag (wt%) 2393 298–299 0.170 4.38 2535 298–299 0.153 4.21 189 299–303 0.559 0.231 0.277 3.24 2.58 3.63 928 299–303 0.158 3.81 992 299–303 0.169 3.37 1464 May 305-July 306 0.252 7.21 2272 May 305-July 306 0.174 0.199 0.226 0.271 4.38 4.87 4.88 3.96 S. Scrivano et al. Journal of Archaeological Science: Reports 44 (2022) 103509 8 The edict of Aphrodisias, issued in AD 301, documents a further step in this process of “artificial” revaluation of nummi, the face value of which was doubled (from 12.5 to 25 denarii) by a simple administrative decision (geminata potentia: Cepeda 2004; Kropff 2017) regardless of the coins’ intrinsic value. 5. Conclusions 553 Coins from the four officinae active in the Carthage mint were examined by non-destrcutive XRF with the aim of better understanding the metallic composition of the coins issued between AD 297 and AD 307. All coins were found to contain a quaternary alloy of silver, copper, lead, and tin, with traces of iron, nickel, zinc, arsenic, and antimony. However, although the four officinae do not present differences in terms of major components, some compositional variations were attested over time. Two clear trends were detected: a decrease in silver and copper and an increase in lead and tin. Changes in copper and lead composition were less significant than those observed for silver and tin. Particularly, a significant difference in silver and tin concentrations was attested from issue 6 of group 1, dated to AD 298, onwards. The average quantity of silver drops significantly, becoming as much as 50% lower than in the initial issues, while tin contents increase from 1.5 wt% to 3.4 wt%. In addition, statistical correlation analysis only yielded a clear correlation between copper and lead, in line with previously published data. Regarding trace elements, zinc, was only attested in coins dating from AD 298 to AD 306, when Zn contents greatly increase and silver contents drop sharply. Finally, mercury was only detected in seven coins, too small a sample to argue that the presence of this element was due to the application of the silver amalgam method. On the other hand, the four polished coins present a bulk content of iron of approximately 0.3 wt%, suggesting that its presence was not the result of contamination during manufacture, but to the presence of iron impurities in the copper ore. This study increases our understanding of the archaeometric characterisation of the issues struck at the mint of Carthage. However, only when other mints and monetary series are similarly analysed shall we be able to contribute significantly to ongoing debates around this and other historical problems. As noted, XRF is to a great extent limited to surface analysis, so the bulk composition of the coins remains unknown. We are currently working to analyse the coins by gamma transmission, which has greater penetration and can thus lead to more accurate results. Meanwhile, the analysis of polished areas in a few pieces (four in total: supra) indicate that the average bulk composition of silver is approximately 1.5 wt% lower than on the surface. These results agree with a recent work (Guihard et al. 2018), which argues for a 1.8 wt% difference in silver content between the surface and the core. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgments This work has been partially supported by the project “Moneda y Metal en la B´ etica tardorromana. Estudio científico del tesoro de Tomares” (PGC2018-093511-B-I00) from the Spanish Ministry of Science, Innovation and Universities (Plan Estatal 2017-2020 de Generaci´ on del Conocimiento - Proyectos I+D+i.). E. García-Vargas and M.A. Respaldiza also acknowledge the financial support from the “VI Plan Propio de Investigaci´ on y Transferencia” of the University of Seville. The authors wish to thank the staff of Museo Arqueol´ ogico de Sevilla and CITIUS of the University of Seville. References Bastien, Pierre. 1980. Le monnayage de l’atelier de Lyon. V. 3: De la r´ eforme mon´ etaire de Diocl´ etien ` a la fermeture temporaire de l’atelier en 316 (294-316). Wetteren. Besly, E., Webster, P., 2002. A hoard of Tetrarchic nummi from Bridgend, South Wales. Numismatic Chron. 162, 169–215. Bollard, I., y Jean No¨ el Barrandon., 2006. Nouvelle contribution ` a l’´ etude du monnayage en bronze du IV si` ecle apr` es J.-C. Revue numismatique 162, 277–310. Cepeda, J. 2004. «Las reformas de Diocleciano y Constantino I y su reflejo en la composici´ on de los tesoros monetarios», in VIII Curs d’Hist` oria monet` aria d’Hisp` ania. La moneda de l’Imperi rom` a, Barcelona 25-26 de noviembre de 2004. Barcelona, pp. 99-112. Chameroy, J. 2016. «R´ eformes mon´ etaires tardo-romaines ` a la lumi` ere des d´ epˆ ots enfouis en Gaule (c. 274-c. 310)». Produktion und Recyceln von Münzen in der Sp¨ atantike / Produire et recycler la monnaie au Bas-Empire. 1. Internationales Numismatikertreffen, May 2014, Mainz, Allemagne. Regensburg: 47-68. Chaves Trist´ an, Francisca, 2017. Reflexiones y estado de la cuesti´ on en torno al tesoro de “El Zaudín” (Tomares, Sevilla). Annali dell’Istituto Italiano di Numismatica 63, 235–268. Cope, L.H., 1968. The argentiferous bronze alloys of the large tetrarchic folles of A. D. 294–307. Numismatic Chron. 8, 115–149. Cope, L.H., y Harry Billingham., 1967. The composition of 35 Roman bronze coins of the period AD 284. Bull. Histor. Metall. Group 1 (9), 1–6. Cope, Lawrence H., y Harry N. Billingham. 1969. «Chemical Analyses of Some WeightReduced Roman Folles Minted between A.D. 307 and 318». Bulletin of the Historical Metallurgy Group 3 (1):30-32. Cope, Lawrence H., y Harry N. Billingham. 1971. «The chemical composition of the bronze coinage of Maxen-tius, A. D. 306-312». Bulletin of the Historical Metallurgy Group 5 (1):62-65. Cope, L.H., King, C.E., Northover, J.P., Clay, T., 1997. Metal Analyses of Roman Coins Minted under the Empire. The British Museum, London. Depeyrot. 1992. «Le syst` eme mon´ etaire de Diocl´ etien ` a la fin de l’Empire Romain, rbn cxxxviii, p. 33-106.» Revue belge de numismatique et de sigillographie CXXXVIII:33-106. Drost, V., 2013. Le monnayage de Maxence (306–312 apr` es J. C.). Soci´ et´ e Suisse de Numismatique. Dungworth, D., 1996. Caley’s “zinc decline” reconsidered. Numismatic Chron. 156, 228–234. Ermatinger, J.W., 1990a. a. «Diocletian’s Economic Revolution,». Münsterische Beitr¨ age zur Antiken Handelsgeschichte 9, 45–49. Ermatinger, J., 1990b. b. «The circulation pattern of Diocletians’s nummus». Am. J. Numism. 2, 107–117. Fawcett, T.G., Blanton, J.R., Blanton, T.N., Arias, L., y T. Suscavage., 2018. Nondestructive evaluation of roman coin patinas from the 3rd and 4th century. Powder Diffr. 33 (2), 88–97. https://doi.org/10.1017/S0885715618000180. Guihard, P.-M., Blanchet, G., Bocquet-Li´ enard, A., Ang´ elique, J.-C., Li´ enard, ´ E., Bataill´ e, M.-P., Dupr´ e, J., y Jean-Louis Gabriel., 2018. Appr´ ehender le stock de m´ etal monnay´ e au ive si` ecle apr` es J.-C.: Analyses par spectrom´ etrie de fluorescence X portable de nummi provenant du tr´ esor de Saint-Germain-de-Varreville (Manche). Arch´ eoSciences 2018, 45–62. https://doi.org/10.4000/archeosciences.5826. Hedges, E.S., Robins, y D.A., 1963. Examination of some silver-coated Roman coins. Numismatic Chron. 3, 237–240. Hunt Ortiz, Mark, 2017. In: Estudio arqueom´ etrico de las monedas recuperadas e n las actividades arqueol´ ogicas de 2016 en el Parque Olivar del Zaudín, Tomare s (Sevilla). Junta de Andalucía:, Tomares (Sevilla). Sevilla, pp. 339–350. Ingo, G.M., Angelini, E., de Caro, T., Bultrini, y G., 2004. Combined use of surface and micro-analytical techniques for the study of ancient coins. Appl. Phys. 79 (2), 171–176. https://doi.org/10.1007/s00339-004-2510-8. Kos, P., 1988. «The Carthage aes nummi of the first tetrarchy». in Studia numismatica Labacensia Alexandro Jelocnik oblata. Lubljana 99–108. Kropff, A., 2017. Diocletian’s Currency System after 1 September 301: An Inquiry into Values. Revue Belge de Numismatique et de Sigillographie 163, 167–187. Lafaurie, J., 1975. R´ eformes mon´ etaires d’Aur´ elien et de Diocl´ etien. Revue numismatique, 6e s´ erie 6 (17), 73–138. Mata, A.L., Salta, M.M.L., Neto, M.M.M., Mendonça, M.H., Fonseca, y I.T.E., 2010a. Characterization of Two Roman Coins from an Archaeological Site in Portugal. Mater. Corros. 61 (3), 205–210. https://doi.org/10.1002/maco.200905284. Mata, A.L., Andr´ e Carneiro, M., Neto, M.M., Proença, L.A., Manuela, M., Salta, L., Helena Mendonça, M., y Inˆ es Teodora Elias Fonseca, 2010b. Characterisation of five coins from the archaeological heritage of Portugal. J. Solid State Electrochem. 14 (3), 495–503. https://doi.org/10.1007/s10008-009-0900-8. Quispe Andía, A., Calla Vasquez, K., Yangali Vicente, J., Rodríguez L´ opez, J., Pumacayo Palomino, I., 2019. Estadística no param´ etrica aplicada a la investigaci´ on científica con software SPSS, Minitab uy Excell. Bogot´ a. Ravetz, A., 1963. Neutron activation analysis of silver in some late roman copper coins. Archaeometry 6 (1), 46–55. https://doi.org/10.1111/j.1475-4754.1963.tb00579.x. Rizzo, F., Cirrone, G.P., Cuttone, G., Esposito, A., Garraffo, S., Pappalardo, G., Pappalardo, L., Romano, F.P., Russo, S., 2011. Non-destructive determination of the silver content in roman coins (Nummi), Dated to 308–311 A.D., by the combined use of PIXE-Alpha, XRF and DPAA Techniques. Microchem. J. 97 (2), 286–290. Romano, F.P., Garraffo, S., Pappalardo, L., Rizzo, F., 2012. In situ investigation of the surface silvering of late Roman coins by combined use of high energy broad-beam and low energy micro-beam X-ray fluorescence techniques. Spectrochim. Acta, Part B 73, 13–19. https://doi.org/10.1016/j.sab.2012.05.012. S. Scrivano et al. Journal of Archaeological Science: Reports 44 (2022) 103509 9 Scott, D.A., 1991. Metallography and Microstructure of Ancient and Historic Metals. The Getty Conservation Institute, Los ¨ Angeles. Sutherland, C.H.V., 1961. The silver content of diocletan’s early post-reform copper coins. Part 1. Archaeometry 4 (1), 56–60. https://doi.org/10.1111/j.14754754.1961.tb00533.x. Sutherland, C. H. V., M. A. Litt, y R. A. G. Carson, eds. 1966. The Roman Imperial Coinage, Volume VI: From Diocletian’s Reform. London. V´ azquez Paz, Jacobo, Garrido Gonz´ alez, y Pablo (Eds.), 2017. El tesoro del Zaudín : contextualizaci´ on arqueol´ ogica del conjunto numism´ atico tardoantiguo de Tomares (Sevilla). Sevilla: Junta de Andalucía. S. Scrivano et al.