Composition, raw material provenance, and manufacture of a unique late Roman silver folding stand (quadripus) from Kőszárhegy, Hungary

The late Roman silver quadripus from Kőszárhegy (Fejér County, Hungary) is the only known silver folding stand from the Late Roman Imperial Age, dated to the fourth century AD. Archaeological evidence indicates that the quadripus is closely related to the Seuso Treasure. Elemental composition and lead isotope analyses of samples taken from the various parts of the folding stand were performed by using LA-QICP-MS and MC-ICP-MS methods in order to determine the provenance of raw material used and the production technology. The silver quadripus consists of rather pure silver (92.5–96.5%) intentionally alloyed with copper. The different trace element composition (Bi, Au, Pb) of the various parts (base, lower part, griffin, upper part, finial, cross bands) indicates the use of different silver batches implying that the various parts were made separately, and then soldered together with hard solders. The same parts of the two original feet are very similar regarding their elemental composition and lead isotope ratios suggesting series production. The nearly constant gold and lead contents of the object indicate that not re-used or re-melted, but primary, cupelled silver was used for manufacturing. The lead isotope ratios of the quadripus cover a quite narrow range (206Pb/204Pb = 18.514–18.717; 207Pb/204Pb = 15.645–15.667; 208Pb/204Pb = 38.592–38.817). Comparing our results to the lead isotope data of the European lead-silver ores, and taking into consideration the archaeological evidences, the silver used for manufacturing the quadripus could come from the Balkan region.


Introduction
The silver four-leg folding stand (quadripus) from Kőszárhegy (Fejér County, Hungary) is the most decorated and only known silver folding stand from the late Roman period (the other examples are all made of base metals). The object became known as the Polgárdi stand (Pulszky 1880). However, its find-spot, the Szár-hegy (47°05′40″ N; 18°19′19, 66″ E) (meaning barren, rocky hill), never belonged to the nearby city of Polgárdi administratively, but to Szabadbattyán at the time of its discovery, and from 1931 to the village of Kőszárhegy. With its dimensions (113 cm in height and 20 kg in weight), it is the largest and heaviest among the known Roman folding stands. It was most probably manufactured in the second half of the fourth century AD, possibly in a Balkan workshop, and was in use sometime during the second half of the fourth century AD (Mráv 2012). On the basis of several late Roman illustrations, fixed or foldable stands were unquestionably an essential part of any table service consisting of large platters. The platters packed with food were placed onto these stands. The differences in the sizes of the platters justify the adjustable structure of the stands. A richly decorated precious metal folding stand reflected the high social status and wealth of its owners. The Kőszárhegy stand was once obviously a very valuable, if not the most valuable, part of a late Roman silver dining or toilette set. According to its high value and quality and the discovery site, it may have been owned by a wealthy elite family who lived in the vicinity of Lake Balaton (lacus Pelso) and belonged to the aristocracy of the Late Empire (Mráv 2012).
The folding stand from Kőszárhegy bears a complex history after the unearthing as a fortuitous find. In 1878, in one of the vineyards of the southeastern slope of Szár-hegy ( Fig. 1), during the extraction of a dead plum tree, among the dried roots, fragments of a late Roman folding stand were discovered. Ten fragments were found, which consisted of two complete legs as well as three whole and one half cross band (Fig. 2a). One leg was heavily bent. The other one had been deliberately cut into three pieces. The cross bands had been detached from the legs by breaking the attachment lugs or by cutting. Some of the cross bands were also broken into several pieces. As stands with four legs (quadripus) were not known at that time, but stands with three legs (tripus) manufactured from bronze or silver were already known, the fragments of the silver stand were interpreted and reconstructed as a tripod shortly after their arrival in the Hungarian National Museum (Fig. 2b). The "missing" leg was replaced by a copy. The stand from Kőszárhegy, reconstructed as a tripod in this way, was first on display on the third Paris World's Fair, called an Exposition Universelle in 1878. In the following more than hundred year, it turned out that the reconstruction as a tripod never worked properly. The cross bands kept bending and the soldering that fixed the lugs kept disintegrating. During a new restoration in 2002, it became evident that the fragments did not belong to a tripod, but to a folding stand with four legs (quadripus) (Fig. 2c). The new reconstruction was also confirmed by the lugs forming an angle of 90° on the original legs and by the fact that quadrupeds were in use during the fourth century AD. As the other half of the stand was seemingly not buried in the same spot, this might suggest that the stand had been deliberately divided in half. The stand reconstructed as a quadruped now functions well and is currently on display in the Hungarian National Museum (Bruder et al. 2002;Mráv 2012). (For more details on the reconstruction of the silver quadripus: see Online Resource 1.) The sculptural decoration of the stand, the sea centaurs (tritons) and sea nymphs (nereids), the sea griffins as well as figures of Eros riding on a dolphin, recalls the sea, i.e., water, which suggests its original function: it must have been used as a washstand. Originally, a washbowl was probably placed or a washbasin with suspension rings was perhaps hung on the hooks modelled on the back side of the finials of the legs (Mráv 2012; Dági and Mráv 2020). The material of the folding stand from Kőszárhegy was previously analyzed indicating the use of high-quality silver. However, the previous studies were confined only to some parts of the quadripus (one of the bases as well as one of the moving parts and the adherent rivet, respectively; for more details and results of the previous analyses on the silver quadripus: see Online Resource 1). The uniqueness of the object justifies a more detailed multi-analytical archaeometric study to be performed on every original part of the object using geochemical methods (major, minor, trace elements, and lead isotopes). The aim of this study is to determine the elemental composition and lead isotope ratios of the late Roman silver quadripus in order to reveal possible chemical inhomogeneity within the object, to determine the provenance of the raw material (ore) used via comparison with potential ore deposits and to characterize manufacturing techniques. These results contribute to a more detailed reconstruction of late Roman craftsmanship, including silver smithing, manufacturing, alloying, and assembling practices.

Materials: the late Roman silver folding stand from Kőszárhegy
The silver quadripus represents a late Roman silver folding stand of the highest quality. Each leg is rectangular in cross-section and decorated with rows of beads along both of its frontal edges. The lugs on the legs and the hinged cross bands are fitted together by rivets and beaded (rosetteform) and undecorated washers. Each leg comprises five different parts: finial, upper part, curved holder representing sea griffin, lower part, base (Fig. 2c). The legs are divided into two different long sections by a winged griffin, emerging from a calyx. The tops of the legs (finial) are decorated with sculpture groups representing mythological figures, each depicting a nereid and a triton. The stand was made with hooked finials on the top (in the form of fingers). Naked statuettes of Eros riding on a dolphin were placed on the base of the legs.
The surface of the legs and the cross bands are richly decorated with engraved floral and leaf-like stroke patterns. Rows of heart-shaped overlapping leaves run along the edges of the legs between parallel rows of beads. The edges of the leaves are emphasized by finely chased lines. On the outer side, dot-punched decoration was also added. These types of leaves are characteristic of the products of the metal workshops of the Balkan Peninsula, such as Naissus and Sirmium (Popović 1997;Mráv 2012). It may also be worth noting that the geometric and floral design as well as the decorative scheme of the quadripus has close and strong similarity to the Geometric Ewers of the Seuso Treasure, which has a bearing on its date of manufacture and possible place of production.  (Bruder et al. 2002); c the silver folding stand in its current form, reconstructed as a quadripus. The various parts of the silver folding stand are denoted in the figure. aI, cII, dI, dII: cross bands 1 3 99 Page 4 of 19 The silver quadripus was made by lost-wax technique (Bruder et al. 2002). The different parts of the legs were cast in separate pieces, as the silver cools too rapidly, and consequently, it loses its form-filling capability. The contemporaneous goldsmiths could have known this phenomenon (Bruder et al. 2002). The upper and lower parts of the legs and the cross bands were made by wax-rolling. The rosettes were pressed from silver plates. The washers were cut out from silver plates as well. The rivets were cast. The lugs (used for fastening the cross bands to the legs) were cast together with the legs.

Sampling
Sampling sites were planned after a thorough and detailed macroscopic observation and non-destructive handheld XRF analysis performed preliminarily on the object (for details, see Online Resource 2). Each part of the two original legs (finials, upper parts, griffins, lower parts, and bases), the three and a half original cross-bands, and the two original moving parts were sampled. In the case of rosettes, washers, and rivets, it is difficult nowadays to determine the original ones as they have quite similar composition based on the non-destructive handheld XRF analysis (Online Resource 2); therefore, only one rosette-washer-rivet set was sampled.
Before sampling, the surface of the silver metal was thoroughly cleaned by using acetone. A hand drill or pliers was used for sampling. The diameter of the drill bit was 0.7 to 0.8 mm. The first few silver shavings were discarded as they could contain corrosion products. Altogether, nineteen samples, each 10-30 mg in amount, were acquired.

Methods: elemental and lead isotope analysis
The instrumentation used for elemental analysis was a Thermo X-Series II inductively coupled plasma mass spectrometer coupled with a Resonetics laser ablation system (ArF,193 nm). The parameters of the ICP-MS are optimized to ensure a stable signal with a maximum intensity over the full range of masses of the elements and to minimize the formation of oxides and double-ionized species. The diameter of the laser beam, and thus the analyzed sample area, was 50 μm. Quantification was carried out using ablation yield correction factors with normalization of the major components to 100% (Kovacs et al. 2009). The following elements were measured: Ag, Cu, Au, Pb, Bi, Fe, Co, Ni, Zn, As, Pd, Sn, Sb, Te, Ir, Pt. The following elements were sought but were below their detection limits (given in brackets) in all samples: Cr (15 µg/g), Mn (4 µg/g), Se (20 µg/g), Ru (7 µg/g), Rh (1 µg/g), Cd (0.5 µg/g), Os (1 µg/g).
Lead isotope analysis was accomplished by multiplecollector inductively coupled plasma mass spectrometer (MC-ICP-MS). The sample was dissolved in diluted HNO 3 and lead was separated with ion chromatography resin from the matrix. The isotope ratios of lead were corrected for the mass discrimination by addition of Tl. A value of as 205 Tl/ 203 Tl = 2.3871 was taken and an exponential relationship assumed. 204 Pb was corrected for the isobaric interference with 204 Hg by measuring 202 Hg and using a 204 Hg/ 202 Hg ratio of 0.2293. The in-run precision of the reported lead isotope measurements was in the range of 0.01 to 0.03% (2σ) depending on the considered ratio. Precisions are generally better than 0.01% (two standard deviations) for ratios normalized to 206 Pb and better than 0.03% for ratios normalized to 204 Pb. Euclidean distances were calculated between the three lead isotope ratios normalized to 204 Pb of each sample and the currently available lead isotope data of lead-silver ores on a point-by-point basis to find the best matching ore deposits (Stos-Gale and Gale 2009) (Online Resource 3).

Elemental composition
The silver quadripus was manufactured from high-quality silver-copper alloys (91.0-96.0% Ag, 2.6-7.5% Cu) (Table 1; Fig. 3). The bases of the legs have the highest copper contents (7.3-7.5%), while the lower parts of the legs have the lowest (2.6-2.8%); other parts are chemically in between. The gold contents of the quadripus range between 0.56 and 0.84% (Table 1 ; Fig. 3). The cross bands form two groups based on their gold contents (0.56-0.57% and 0.83-0.84%, respectively), while other parts are more homogeneous (0.59-0.79%). The lead contents range between 0.34 and 0.80% (Table 1; Fig. 3). The lower parts of the legs have the highest (0.75-0.80% Pb), while the cross bands, the griffin B, and the finial of the A leg have the lowest lead contents (0.34-0.38% Pb). Based on the bismuth contents, the various parts form distinct groups (Table 1; Fig. 3): the lower parts of the legs have the lowest bismuth contents (20 µg/g), while two cross bands, the upper parts and the griffin A, have the highest (1170-1360 µg/g). The lower parts, the rivet, and the rosette form a distinct group having the lowest Zn and Sn contents among the various parts (less than 200 µg/g Sn and Zn) (Table 1; Fig. 3). The zinc contents of the other parts are rather similar below 1000 µg/g. The moving parts and bases have the highest tin contents (450-850 µg/g). The silver, copper, gold, and lead contents of the small parts (rosette, rivet, washer) are similar to the larger parts.
Based on the Au, Pb, and Bi contents (Fig. 3), it is evident that various parts of the legs have completely different elemental compositions, while identical parts of the two legs have quite similar compositions (with the exception of the griffins). The cross bands are grouped into two groups based on their minor and trace elemental composition (aI and dII vs. cII and dI). The moving parts of the two legs are completely similar to each other, while the compositions of the analyzed rivet and rosette are slightly different and the washer differs completely from them.

Lead isotope ratios
Based on the lead isotope abundance ratios, the different parts of the folding stand form two distinct groups (Fig. 4) Table 2).
The lead isotope ratios show the same picture as the elemental composition. Namely, the various parts of the legs have different isotope ratios, while the same parts of the two legs are similar (except for the griffins). The cross bands form two groups based on their isotope ratios as well (aI and dII vs. cII and dI). The moving parts of the two legs have the same isotope ratios, as well as the rivet and the rosette. The isotope ratio of the washer is completely different from the latter.
Pure silver is too soft for everyday use, as it dents, bends, and wears easily. Therefore, in the late Roman period, silver was most commonly alloyed with copper, as it increased the strength and hardness of the softer silver. The hardness of an alloy depends not only on its chemical composition, but also on the degree of working and heat treatment. The hardness increases quickly up to a copper content of 15%, and reaches a rather constant value between 30 and 80% (Hughes and Hall 1979). During silver extraction from argentiferous lead ores by cupellation, the copper content usually remains low at less than 1%; thus, higher copper concentrations always indicate intentional alloying (Hughes and Hall 1979). Analyses of late Roman silver objects show copper concentrations from 0.1 to 15% (Hughes and Hall 1979;Lang et al. 1984;Feugère 1988;Lang 2002;Tate and Troalen 2009;Cowell Mozgai et al. 2020Mozgai et al. , 2021a. The differences in the copper contents of the various parts of the folding stand also indicate intentional alloying. The bases that are more exposed to mechanical strain were made from alloys with higher copper contents.

Minor and trace elements (impurities)
The measured elements (except for silver and copper) are naturally occurring and unintentionally added, deriving from the silver ore or from the copper used for alloying (Hughes and Hall 1979). Their individual contents usually do not exceed 1%. indicate copper with about 2% Sn. There is only slight correlation between copper and zinc, but if the zinc was introduced with copper then this had variable zinc concentrations up to 2% Zn. Three different levels of bismuth concentrations can be seen at similar lead concentrations. The lowest bismuth concentration at the highest lead content, which indicates the lowest degree of cupellation, meaning that the ore did not contain bismuth In the Roman period, the primary source of silver was silver-bearing lead ores (galena, PbS; anglesite, PbSO 4 ; cerussite, PbCO 3 ; Tylecote 1962;Forbes 1971). The lead ores were roasted, smelted to lead, and then this was cupelled for silver extraction. Cupellation purified the silver from impurities (e.g., antimony, arsenic, tin, iron and zinc; less well from copper, gold and bismuth), leaving between 0.1 and 1% lead in the silver. The volatile elements (antimony, arsenic, mercury, tin, and zinc) are efficiently removed from the molten silver during cupellation (Pernicka and Bachmann 1983;L'Héritier et al. 2015); however, they can be present in high concentrations (several %) in native silver (Pernicka 2014). The absence or low amount of these volatile elements in the analyzed objects indicates that cupelled silver was used for manufacturing the silver folding stand.
Ancient Romans could produce high-purity silver with a lead content of 0.1-1% and this is also the range of lead in the quadripus. The lead contents of the various parts differ slightly, indicating that different batches of silver were used. The generally low lead contents (up to 0.8% Pb) indicate that  no additional lead was added to the purified silver so that lead isotope ratios may be used to determine the provenance of the raw material. Bismuth is also an important element in determining the raw material provenance of silver objects, as it survives the cupellation process to a certain degree (Pernicka and Bachmann 1983;Pernicka 2014;L'Héritier et al. 2015). Dry ores (such as cerargyrite, AgCl; argentite, Ag 2 S) or native silver have bismuth contents below 0.05% (Craddock 1995), whereas argentiferous galena often contains 0.1-1% bismuth (Gale and Stos-Gale 1981). Based on cupellation experiments, bismuth is only oxidized in the final stages of cupellation; therefore, bismuth concentrations and thus the Bi/ Pb ratios in silver objects are correlated with the degree of cupellation, which is indicated by the lead content. However, the final Bi/Pb ratio of the cupelled silver depends also on the initial Bi contents of the silver-bearing lead ores (Pernicka and Bachmann 1983;L'Héritier et al. 2015). Accordingly, the differences in the Bi vs. Pb diagram of the different parts of the silver folding stand (Fig. 3) indicate that they were manufactured from different batches of silver.
Concerning the addition of copper, one may well ask if pure copper or copper alloys were used, since tin and zinc are frequent components in late Roman copper alloys. Figure 3 shows the relation between these two elements with copper. There is only a slight tendency of increasing tin and zinc with increasing copper, which may suggest that the copper was not added from a single batch but maybe in the form of small pieces of various compositions with the majority containing roughly 1% of tin and zinc.
During metallurgical processes, the Au/Ag ratio in silver does not change drastically, including cupellation (Pernicka 2014;L'Héritier et al. 2015). The Au/Ag ratios of argentiferous lead ores in the Aegean range from 10 −7 to 10 −2 (Pernicka 2014). The gold contents of the silver quadripus are rather constant and fall within the range typical of late Roman objects (0.1-4.7 wt%) (Hughes and Hall 1979;Lang et al. 1984;Feugère 1988;Lang 2002;Tate and Troalen 2009;Cowell and Hook 2010;Hook and Callewaert 2013;Doračić et al. 2015;Lang and Hughes 2016;Greiff 2017;Vulić et al. 2017;Mozgai et al. 2020Mozgai et al. , 2021a. There are several reasons why gold concentrations may exceed 1%, such as the presence of remnants of former gilding, the re-usage of scrap gilded silver, or the use of silver derived from the cementation of gold-silver ores. The elemental composition of the silver folding stand indicates that most probably the silver metal derives from primary, smelted silver ores (argentiferous lead ores). However, the use of recycled scrap silver metal cannot be totally excluded. During recycling, silver can be refined to increase its purity via cupellation with the use of lead from different sources than silver or, alternatively, debased by adding base metals in the form of pure copper or recycled copper-alloy objects (Blackwell et al. 2017;Kershaw and Merkel 2021). The latter would result in a wide range of variation in the copper, gold, lead, tin, and zinc values, as it was demonstrated for the silver objects from the Migration Period and Early Middle Ages (Troalen 2017;Horváth et al. 2019;Mozgai et al. 2021b). In case of recycling, identification of the ore source based on lead isotope data is almost impossible. However, we argue that besides the elemental composition discussed above, the high quality and prestigious character of the silver quadripus, as a very valuable, if not the most valuable, part of a Roman dining or toilette set of a wealthy, high-status family, further emphasizes that not recycled scrap but rather silver from primary sources was used for its manufacture. The differences in the Au, Pb, and Bi contents (Fig. 3) of the various parts of the silver folding stand indicate the use of different batches of silver.

Manufacture
The silver folding stand is a composite object, similarly to some of the objects (ewers, situlas, amphora, and casket) from the Seuso Treasure (Mozgai et al. 2021a). The elemental and isotopic inhomogeneities were proven by the detailed analysis of the various parts of the silver quadripus, indicating the use of various silver batches during manufacture. Based on the above, we draw some conclusion on the manufacture of the object. The similarities in the composition of most of the same parts of the two legs (finials, upper parts, lower parts, bases, moving parts) imply that the same parts were cast simultaneously from the same silver batch and were made in series production. The cross bands form two distinct groups based on their composition indicating that they were manufactured (at least) in two steps from different silver batches. Similarly, the two griffins were made from different silver batches in separate steps. The differences in the composition of the rosette and the washer hint that they were cut out from different silver plates.

Provenance of silver based on lead isotope ratios
When the Roman Empire reached its largest extension during the reign of Trajan (98-117 AD), it possessed access to a large number of lead and silver deposits. Silver metal was produced from deposits located throughout Europe and was widely traded throughout the Empire (Davies 1935). Therefore, we have compared our lead isotope data with the lead isotope ratios of lead-silver ores from deposits in the whole territory of the Roman Empire. The data of lead-silver ores from the well-known OXALID database (Oxford Archaeological Lead Isotope Database from the Isotrace Laboratory, Stos-Gale and Gale 2009) was used as a starting reference database. This database was completed with available data of lead-silver ores from European deposits from the geological and archaeometric literature (Aegean: Wagner et al. 1986;Stos-Gale et al. 1996;British Isles: Rohl 1996;Bulgaria: Stos-Gale et al. 1998;Amov 1999;France: Brévart et al. 1982;Le Guen et al. 1991;Baron et al. 2006;Germany: Large et al. 1983;Krahn and Baumann 1996;Niederschlag et al. 2003;Durali-Müller et al. 2007;Bode et al. 2009; Kosovo: Westner 2016; Northern Macedonia: Velojić et al. 2018;Romania: Marcoux et al. 2002;Baron et al. 2011;Sardinia and Tuscany: Boni and Köppel 1985;Stos-Gale et al. 1995;Serbia: Pernicka et al. 1993;Veselinović-Williams 2011;Westner 2016;Spain: Graeser and Friedrich 1970;Arribas and Tosdal 1994;Stos-Gale et al. 1995;Velasco et al. 1996;Canals and Cardellach 1997;Velasco et al. 2003;Santos-Zalduegui et al. 2004;Tornos and Chiaradia 2004;Renzi et al. 2009;Switzerland: Guénette-Beck et al. 2009;Cattin et al. 2011;Turkey: Yener et al. 1991;Sayre et al.1992). As the various parts of the silver folding stand form two distinct groups based on their 206 Pb/ 204 Pb, 207 Pb/ 204 Pb and 208 Pb/ 204 Pb ratios (group 1-the lower parts of the two legs, the rivet and the rosette; group 2-the rest of the samples, Fig. 4), in the following, these two groups are discussed separately, but we must emphases that group 1 is not a real group, and only represents two pairs of samples with almost identical lead isotope ratios and chemical composition.
In Roman times, silver was mainly extracted from argentiferous lead ores, except the jarosite ores from Rio Tinto, SW Spain (Davies 1935;Healy 1978;Amóros et al. 1981;Dutrizac et al. 1983;Shepherd 1993). As Rio Tinto jarosite contains small amounts of lead, additional lead had to be used to extract silver during cupellation. The silver ore was extracted mainly with lead deriving from Cartagena, SE Spain (Craddock et al. 1985;Anguilano et al. 2007Anguilano et al. , 2010Murillo-Barroso et al. 2016). Therefore, in this case, the isotope ratios do not indicate the provenance of the silver ore but rather the provenance of exogenous lead used for cupellation. Our results overlap neither with the ores of Rio Tinto nor with the lead ores from Cartagena, Gádor, Linares, Catalonia (SE Iberia) (Fig. 5). Moreover, the ores from Rio Tinto are characterized by high Bi and Au contents (reaching Fig. 5 Comparison of the lead isotope ratios of the silver folding stand to those of lead-zinc ores from western Europe and the western Mediterranean (data sources: OXALID; British Isles : Rohl 1996;France: Brévart et al. 1982;Le Guen et al. 1991;Baron et al. 2006;Germany: Large et al. 1983;Krahn and Baumann 1996;Niederschlag et al. 2003;Durali-Müller et al. 2007;Bode et al. 2009;Sardinia and Tuscany: Boni and Köppel 1985;Stos-Gale et al. 1995;Spain: Graeser and Friedrich 1970;Arribas and Tosdal 1994;Stos-Gale et al. 1995;Velasco et al. 1996;Canals and Cardellach 1997;Velasco et al. 2003;Santos-Zalduegui et al. 2004;Tornos and Chiaradia 2004;Renzi et al. 2009;Switzerland: Guénette-Beck et al. 2009;Cattin et al. 2011) even several %) (Craddock et al. 1985;Orejas Saco del Valle et al. 2015). Therefore, we exclude the use of Rio Tinto ores. Most silver mines in Europe exploited during Roman times contain mainly argentiferous lead minerals; the silver used for the manufacture of the silver quadripus was most probably extracted from such ores.
If a single ore source would be used for each batch of silver, as was the case in prehistoric times, then the lead isotope ratios of the silver metal would represent the lead isotope ratio of individual mines. However, in Roman times, the silver-bearing lead ores from different sources were mixed during metallurgical processes and silver extraction. It is proven by lead isotope analysis that during Roman times, as mining and metallurgy of the precious metals was monopoly of the emperor (Forbes 1966), the cupellation of the argentiferous lead took place in metallurgical centres (such as Ulpiana in the Balkans), where lead from different small local mines were cupelled (Westner 2016). Therefore, this silver shows a homogenized isotope ratio of a special mining region. In this case, only the larger mining region/province can possibly be identified by means of provenance studies.
As a relatively large mass of silver was required for the manufacture of such a large object as the silver quadripus, we may assume that the silversmith might have mixed and melted several silver pieces (ingots) from various ore sources/metallurgical centres. In this case, the lead isotope data of the silver material would show a mixed, homogenized composition and the exact provenance of silver would be difficult to be determined. Considering that there were no central silver stocks or control of them (Patterson 1972;Hirt 2010), the large elemental and isotopic differences among the various parts suggest that the silver quadripus may have been manufactured during a longer time span, during which the workshop acquired silver material from different metallurgical centres or, alternatively, from one metallurgical centre but with variable elemental compositions and isotope ratios.
To narrow down the possible ore sources, the various parts of the folding stand and the European lead-silver ore deposits were compared using Euclidean distances (d) between each sample and the database looking for strong or weak overlapping (d < 0.1 and d = 0.1-0.2), respectively (Online Resource 3).
In the case of so many sources that could have supplied the silver in principle, another type of information is required to narrow down the number of possible candidates. The most obvious sources of additional information are the ancient literary tradition and field studies of Roman mining regions. The most important ancient mining districts of Bulgaria and Macedonia are found in the Sakar Mountains (Ustrem district), and the Eastern Rhodope Mountains, as well as the Stara Planina and Osogovo Mountains. The hydrothermal lead-zinc vein deposits were intensively mined in Roman times, especially after the second century AD (Davies 1932(Davies , 1935Healy 1978;Shepherd 1993;Hirt 2010). In Romania, important Pb-Zn-Ag mineralizations occur in the Maramureş region (Baia Sprie, Baia Mare) and Au (with silver) in the Apuşeni Mountains (Roşia Montana) (Marcoux et al. 2002;Baron et al. 2011). However, during Roman times, silver mining seems to have had only regional relevance (Davies 1935;Shepherd 1993;Baron et al. 2011). Furthermore, the Dacia province was not part of the territory of the Roman Empire in the time of the manufacture of the silver folding stand so that it can be excluded as a possible source. In ancient Makedonia and Thrace, though there are several deposits, the Romans re-opened only a few for exploitation. There is evidence for Roman mining and smelting in this region in the second-fifth centuries AD. The Pb-Zn-Ag ore deposits can be found in the Chalkidiki Peninsula, the island of Thasos and in the Rhodope Mountains (Pernicka et al. 1981;Wagner and Pernicka 1982;Nesbitt et al. 1988). In Kosovo and Serbia, the ores can be found in vein deposits, which were intensively mined on an industrial scale during the late Roman period. Important, rich Pb-Ag mines existed at several sites, e.g., Srebrenica, Mount Kosmaj, Rudnik, Kopaonik Mountains, and the mines around Priština (Davies 1935;Dušanić 1977;Tomović 1994;Dušanić 2004;Merkel 2007;Petković 2009;Hirt 2010;Gassmann et al. 2015, Westner 2016. There are also Mississippi Valley Type and hydrothermal ore deposits in the Cévennes (Massif Central) which were intensively mined in the medieval period (Brévart et al. 1982;Le Guen et al. 1991;Baron et al. 2006) but in Roman times they probably served only local interests (Davies 1935;Shepherd 1993).
Accordingly, the available archaeological evidence suggests that the silver used for manufacturing the silver quadripus derived from the ore deposits of the Balkan region as this was the main mining area of precious metals during late Roman times (fourth century AD), when mining on the Iberian Peninsula declined (Davies 1935;Shepherd 1993;Hirt 2010). The Romans conquered much of the Balkan area between the second century BC and first century AD, and systematically catalogued and mined ores in the region (Dušanić 2004). Under the Roman influence, south of the Danube, Illyricum and Moesia became typical mining provinces, with a particular focus on gold and lead-silver deposits (Hirt 2010). The mining district of Moesia Superior is claimed to have comprised the lead-silver mining region in the vicinity of Kursumlija in the Toplica valley and the mining areas of Janjevo, Novo Brdo, and Lece. The region of Mt. Kosmaj/Stojnik in Moesia Superior, together with the mines at Avala, Zeleznik, and Rudnik, was identified as belonging to one vast imperial mining domain (Šumadija). The argentiferous lead mining ventures of Avala are suspected to have come under imperial control in the late third century AD. The material remains of Kosmaj show that silver mining took place on a massive scale under imperial control (Merkel 2007;Hirt 2010). At the Kosmaj mines, Roman miners exploited the lead-zinc ores being exceptionally rich in silver: in spite of an imperfect technological procedure, 40-80% of lead with around 6000 g of silver per ton were obtained (Tomović 1994 We compared our results also with lead isotope ratios of other Roman silver hoards (Fig. 7). Unfortunately, there are as yet no lead isotope data available for contemporaneous (fourth century AD) silver hoards (e.g., Mildenhall, Kaiseraugst, Vinkovci). Only the solder of the silver ewer from Trier was analyzed and interpreted as its lead deriving most probably from the Eifel region, Germany (Schwab 2017). However, the lead isotope ratio of the solder does also not overlap with the lead isotope data of the silver quadripus (Fig. 7). Our results show no overlapping with the Boscoreale hoard found near Pompeii (dated to the first century AD) (Berthoud et al. 1988) and with the three Gallo-Roman silver hoards: the Notre Dame d'Allençon hoard (dated to the first half of the second century AD); the Berthouville hoard (dated to the second century AD); and the Graincourt-lès-Havrincourt hoard (dated to the third century AD) (Baratte 1981;Baratte et al. 1985;Lapatin 2014). This indicates that completely different silver batches from different sources were used for the manufacture of the silver hoards, supporting the theory that there were no centralized stocks and control of silver for silversmiths (Patterson 1972;Hirt 2010). Our results partly overlap with Roman silver objects from the ancient kingdom of Kartli (Caucasian Iberia, Georgia) (dated to the first and third centuries AD) (Fig. 8), whose silver most probably derived from the Pangaeon Mountains, or the Central Balkans or the Cévennes (Massif Central) (Parjanadze and Bode 2017). Similarly, only a very weak overlap exists with the silver hoard from Marengo (dated to the second half of the second to the first half of the third centuries AD), whose silver for most of the objects was related to the Massif Central or, alternatively, to SW Iberia (Angelini et al. 2019). Finally, our results strongly overlap with the lead isotope ratios of Sasanian silver objects from the third-seventh centuries AD (OXALID Database, P. Meyers). However, from the archaeological point of view, it is unlikely that they share a common source with the silver of the quadripus.
Isotope data of the folding stand were also compared with lead ingots from several Roman shipwrecks recovered in the western Mediterranean Sea (dated to the first century AD) and metallurgical by-products (Fig. 8). Group 1 overlaps with the metallurgical by-products from Kosovo (Westner 2016), as it overlaps with the Kosovo ores as well. No overlap is observed with the lead ingots from Mahdia (Tunisia), Cabrera 5 (Baleares), and Saintes-Maries-de-la-Mer (France) deriving from the Massif Central (France) and the Sierra Morena (SW Spain). Only group 1 shows a slight overlap with the Mal di Ventre ingots (Sardinia), which were most probably derived from the Cartagena-Mazarrón ore district (Trincherini et al. 2001(Trincherini et al. , 2010Clemenza et al. 2017). However, our results show no overlap with the ores from Cartagena-Mazarrón ore district (Fig. 5). No overlap is observed Fig. 7 Comparison of the lead isotope ratios of the silver folding stand to those of other Roman silver hoards and Roman and Sasanian silver objects (OXALID; Baratte et al. 1985;Berthoud et al. 1988;Parjanadze and Bode 2017;Schwab 2017;Angelini et al. 2019) with the Messallinus ingot dated to the first century AD, for which the ore was derived from Novo Brdo in eastern Kosovo (Rothenhöfer et al. 2018), but it falls in between group 1 and group 2. No overlap is observed with the lead ingot from Rio Tinto (Domínguez et al. 2021), but our results (group 1) show strong overlap with the lead ingots dated to the first century BC and found in a shipwreck near Comacchio (Ferrara, Italy) (Bode et al. 2021) (Fig. 8). These ingots were most probably derived from the ore deposits of Chalkidiki, island of Thasos and Pangaeon Mountains (Bode et al. 2021).
Compared to lead objects from different parts of the Roman Empire and to Roman lead-glazed ceramics (Fig. 9), no overlap was observed with the lead objects from Germania (Germany) and Lusitania (Portugal) and only a slight overlap with Puglia (Italy) (Durali-Müller et al. 2007;Gomes et al. 2016Gomes et al. , 2017Carroll et al. 2021). Group 1 overlaps with the metal objects from Kosovo (Westner 2016), with the lead objects from Augustus Caesarea, Israel (Raban 1999) and with the Roman lead-glazed ceramics (first-fourth centuries AD) (Walton and Tite 2010), all of them made of lead originated from the Balkans. Group 1 slightly overlaps with Roman sling bullets found near Dobrudja (Romania) (dated to the first-third centuries AD) (Vlad et al. 2011), while both groups slightly overlap with the data point of Greek and Roman curse tablets (Skaggs et al. 2012;Vogl et al. 2018), their lead deriving from Greek, Balkan, or Sardinian ores. Group 2 slightly overlaps with the water pipes from Pompeii (Boni et al. 2000), which were interpreted as Fig. 8 Comparison of the lead isotope ratios of the silver folding stand to those of Roman lead ingots and metallurgical by-products (Trincherini et al. 2001;Trincherini et al. 2010;Clemenza et al. 2017;Rothenhöfer et al. 2018;Westner 2019;Bode et al. 2021;Domínguez et al. 2021) made of recycled lead material (from "mixed lead" circulating in the Roman Empire), but we exclude the addition of exogenous lead.
It may be noteworthy to draw attention to the parallels between the quadripus and the products of the silversmith's workshops of the Balkan (e.g., Naissus or Sirmium) (Mráv 2012). The geometric and floral patterns of the stand are stylistically also broadly comparable to the decorations of silver vessels manufactured in Balkan workshops. Although further research is needed to prove this hypothesis, based on these similarities, it can preliminary be concluded that the quadripus may have been manufactured in one of the silversmiths' centres of the Balkan Peninsula. Therefore, it is most likely that the ore used for manufacturing derived from this large mining region. This hypothesis is further supported by the substantial overlap with the isotope data of ores from this region and of metal objects whose raw material originates from this region. Though there are several difficulties encountered, the region is not so well investigated, there are several lead isotope studies on copper ores from the region, but similar studies of lead-silver ores are more scarce. The lack of archaeological evidence of Roman mining and destruction by later mining activities further complicates the interpretation. The differences in the elemental and isotopic composition are supposed to be the result of the following possible scenarios: (i) the workshop acquired silver material from several metallurgical centres of this large region, and additionally (ii) the silver material from different metallurgical centres was probably mixed in different proportions that resulted in different elemental and isotopic compositions for the batches of silver used to manufacture the various parts of the quadripus. Fig. 9 Comparison of the lead isotope ratios of the silver folding stand to those of Roman lead objects and lead-glazed ceramics (Raban 1999;Boni et al. 2000;Durali-Müller et al. 2007;Walton and Tite 2010;Vlad et al. 2011;Skaggs et al. 2012;Gomes et al. 2016;Westner 2016;Gomes et al. 2017;Vogl et al. 2018;Carroll et al. 2021)

Conclusions
The silver quadripus was manufactured from high-quality silver typical for the late Roman period, and was intentionally alloyed with copper. The differences in the chemical and isotopic composition of the various parts of the folding stand indicate the use of different batches of silver during manufacture. However, the similarities in the composition of the most of the same parts implies series production. The constant gold contents suggest that the object was not manufactured from reused/recycled scrap silver metal. The constant and low lead contents indicate the use of cupelled silver. Based on the chemical composition and lead isotope ratios, we exclude the remelting of former silver objects and addition of exogenous lead during cupellation processes, and therefore, the use of primary argentiferous lead ores is supposed. Based on the lead isotope ratios, the various parts of the object are classified into two groups and the Bi/Pb ratio further refines the grouping. The silver of group 1 (lower parts of the legs, the rivet and the rosette) most probably derived from eastern and southeastern European deposits, from the Balkans (Serbia, Kosovo, Bulgaria, Macedonia). For group 2, the provenance determination is more problematic, with a most likely origin from Bulgaria or the Cévennes (France). The archaeological evidence suggests that the silver quadripus was manufactured in a Balkan workshop and the fact that the main centres of silver mining during late Roman times were in the Balkan region further support the conclusion that the silver used for manufacturing the folding stand derived from the Balkan region.

Acknowledgements
The measurements were performed within the framework of the Seuso Research Project supported by the State of Hungary from 2014 to 2019. The authors are grateful to Balázs Lencz and Tamás Szabadváry at the Hungarian National Museum for their help during sampling and hXRF analysis.
Author contribution VM took part in sampling and data evaluation, wrote the first draft of the manuscript, and edited the figures. BB took part in sampling and data evaluation, and edited the manuscript. EP performed the LA-QICP-MS and MC-ICP-MS measurements and contributed to data evaluation. ZsM wrote the archaeological background of the manuscript and provided permission to the study of the analyzed artefact. All authors reviewed the manuscript.
Funding Open access funding provided by ELKH Research Centre for Astronomy and Earth Sciences.

Data availability Not applicable.
Code availability Not applicable.

Declarations
Ethics approval Not applicable.

Consent for publication Not applicable.
Competing interests Not applicable.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.