Analyses of queen Hetepheres ’ bracelets from her celebrated tomb in Giza reveals new information on silver, metallurgy and trade in Old Kingdom

, Greece), and not partitioned from gold or electrum as previously surmised. Sources in Anatolia (Western Asia) can be excluded with a high degree of confidence. Imaging of a cross-section of a bracelet fragment reveals that the metal was repeatedly annealed and cold-hammered during creation of the artefacts. The results provide new information about silver ore sources, commodity exchange networks and metallurgy in Egypt during the Early Bronze Age.


Introduction
Celebrated among Old Kingdom Egyptian silver objects are the bracelets of queen Hetepheres I (c. 2600 BC), mother of king Khufu, builder of the Great Pyramid at Giza (date of reign c. 2589-2566 BC, Shaw, 2000: 482). The rarity of these objects is threefold: surviving royal burial deposits from this period are rare; only small quantities of silver survived in the archaeological record until the Middle Bronze Age (c. 1900 BC); and Egypt lacks silver ore deposits. Few silver artefacts from the Old Kingdom have been analysed.
Silver (Ag) artefacts first appeared in Egypt during the 4th millennium BC but the original source then, and in the 3rd millennium, is unknown (Ogden, 2000: 170;Hauptmann and von Bohlen, 2011). Ancient Egyptian texts do not mention any local sources (Gale and Stos-Gale, 1981: 104), but an older view, derived from the presence of gold (Au) in Ag objects, plus the high Ag content of Egyptian gold and electrum, holds that Ag was derived from local sources (Mishara and Meyers, 1974;Lucas and Harris, 1962: 248;Gale and Stos-Gale, 1981: 113;Ogden 2000: 170-71). An alternative view is that Ag was imported to Egypt, possibly via Byblos on the Lebanese coast, owing to many Ag objects found in Byblos tombs from the late fourth millennium (Prag, 1986: 38). Exchange mechanisms between Egypt and Lebanon are attested from the Naqada IIIA1 period (c. 3320 BC) and probably earlier (Hartung, 2001;Hartung et al., 2015;Ownby et al., 2022). Over the ensuing centuries, Egypt developed intensive trade relations with Byblos and the region. By the 4th Dynasty, the Egyptian state also sourced commodities directly or through intermediaries, down-the-line and indirect networks (Sowada, 2009: 245-55). By the late Old Kingdom (c. 2300 BC), networks for metals extended eastward, accessing silver from Ebla (Marcolin and Diego Espinel 2011;Biga and Steinkeller 2021).
The tomb of queen Hetepheres I (Tomb G 7000X) was discovered at Giza in 1925 by the Harvard University-Museum of Fine Arts (MFA) Joint Expedition (Reisner, 1927). Hetepheres was one of Egypt's most important queens: wife of 4th Dynasty king Sneferu and mother of Khufu, the greatest builders of the Old Kingdom (c. 2686-2180. Her intact sepulchre is the richest known from the period, with many treasures including gilded furniture, gold vessels and jewellery (Harvard, 2023). The bracelets were found surrounded by the remains of a wooden box covered with gold sheeting, bearing the hieroglyphic inscription 'Box containing deben-rings' (Reisner, 1927: 21) (Fig. 1A). Twenty deben-rings or bracelets were originally interred, one set of ten for each limb, originally packed inside the box (Reisner and Smith, 1955: 43-44, Figs 44;Plates 36-37). Recovered in different states of preservation, most were sent to the Egyptian Museum (Cairo JE53271-3) (Fig. 1B), and two others, plus many fragments, were gifted to the Museum of Fine Arts (MFA) in Boston in 1947 by the Egyptian Government (MFA 47.1699(MFA 47. -1702 (Reisner and Smith, 1955: 20, 43-44, pl. 38;Aldred, 1971: 33, 133, 158, 175, Pl. 3;Ziegler, 1999: 20-21). The bracelets on display at the MFA are fragmentary (Fig. 1C, right) with additional unattached loose pieces kept in climate-controlled storage (MFA 47.1702) ( Fig. 2A).
The bracelets, made of a metal rare to Egypt, are a statement of royal privilege and taste. The thin metal worked into a crescent shape and the use of turquoise, lapis lazuli and carnelian inlay, stylistically mark the bracelets as made in Egypt and not elsewhere (Petrie, 1901:Pl. ix.4;Aldred, 1971: 33-4, pl. 13). Each ring is of diminishing size, made from a thin metal sheet formed around a convex core, creating a hollow cavity on the underside. Depressions impressed into the exterior received stone inlays forming the shape of butterflies. At least four insects are depicted on each bracelet, rendered using small pieces of turquoise, carnelian and lapis lazuli, with each butterfly separated by a circular piece of carnelian. In several places, pieces of real lapis have been substituted by painted plaster (Reisner and Smith, 1955: 44).
The metal was examined nearly 100 years ago by A. Lucas and H.E. Cox and found to be 90.1% Ag, 8.9% Au and 1.0% copper (Cu) (Reisner, 1928;Reisner and Smith, 1955: 44). The location(s) measured, and the method of analysis, were not reported. They noted visible 'yellow patches on the surface … due to the unequal distribution of the gold present' (Lucas and Harris, 1962: 246, 248). Lucas (1962: 249) also noted that silver objects 'with a very much lower gold content … may have been imported as early as the Old Kingdom at least'. Since then, the MFA bracelets have not been analysed. Our research presents new images, and elemental, mineralogical and lead isotope analyses, helping to characterise the metal composition, and constrain the provenance of the Ag and method of manufacture of these unique objects.

Description
The artefacts on exhibition at the MFA could not be sampled, however three small bags (B1-B3) of corroded bracelet fragments were available for study (Fig. 2). A piece from bag B2 was metallic in appearance (MFA 47.1702 B2P1, Fig. 2C-D) (hereafter "B2P1"), while the remaining pieces were heavily corroded. A sample of metallic fragment MFA 47.1702 B2P1, and corroded fragment MFA 47.1702 B1P4 ( Fig. 2A centre left, and 2B) (hereafter "B1P4"), were selected for further study.

Elemental composition
Bulk XRF. Elemental compositions were measured from five locations on the corrosion-free metal forming fragment B2P1, using a Bruker Artax X-ray fluorescence spectrometer (XRF) with rhodium (Rh) anode tube operated at 40 kV, 400 µA, 60 s live time, a 0.65 mm collimator, no filter and all measurements were made in air. Based on counting statistics, limits of detection for elements was estimated to be 0.1 wt% for iron (Fe), Cu and lead (Pb), and 0.3 wt% for Ag and Au. Quantitative results were determined by Bruker software using one laboratory Ag standard. Based on previous experience, accuracy for Ag and Au was estimated as ±1 wt% for Cu and ±0.1 wt% for Pb.
The bulk elemental compositions of ~1.4 g of corrosion products of fragment B1P4 were measured using a Panalytical Epsilon 3 benchtop energy dispersive X-ray fluorescence spectrometer, with rhodium (Rh) anode tube and a 6 mm ellipse measurement area. Measurement conditions were 50 kV, 60 s, Ag filter; 50 kV, 300 µm Cu filter, 60 s; 20 kV, 200 µm Al filter, 60 s; 12 kV, 50 µm Al filter, 100 s; 5 kV, no filter, 100 s and all measurements were made in air. Limits of detection for elements was estimated to be 0.001 wt% for iron (Fe), Cu and lead (Pb), and 0.005 wt% for Au. Quantitative results were determined using the Panalytical Omnian software, trained using a range of MBH reference materials consisting of trace elements in Ag matrices. Accuracies for Fe, Cu and Pb were better than 10% for concentrations >0.1 wt%. Data reduction and quantification were carried out in the Epsilon 3 software version 1.1. Data were summed to 100%.
Micro-XRF. Elemental maps were produced from a ~9.1 × 6.8 mm    The final data point is for the larger rectangular area and can be considered the average for this part of the sample (about 89% Ag and 9% Au). Note that Au has higher concentrations and Ag lower in the more heavily corroded areas (data points in the middle of the plots).
Calibrations were verified using stainless steel. Limits of detection for the elements reported are ~40 mg/kg.

SEM-EDS.
A small solid sample was taken from B2P1 and embedded in Buehler EpoThin epoxy resin and prepared as a polished section. This sample was examined using a JEOL JSM-6460LV scanning electron microscope with energy dispersive spectrometry (SEM-EDS). Operating conditions were 35 Pa chamber pressure, 20 kV with ~1nA beam current. The working distance was 10 mm and quantitative elemental analyses were collected over 10 s live time. An 80 mm 2 Oxford X-Max N silicon drift detector was used with Oxford default standards and Oxford Instruments INCA software. Quantitative SEM-EDS results utilized pure metal standards, with matrix corrections carried out in the INCA software. Limits of detection, estimated from counting statistics, are 0.1 wt % for Cu, 0.2 wt% for Ag and Au, 0.3 wt% for Pb. Accuracy is estimated to be +/-2% (relative) for Ag and Au, +/-10% (relative) for the other elements. Since the beam current was not monitored during analyses, all results were normalized. Back-scattered electron images of the section are shown in Fig. 3, with other details in Tables 3 and 4. Numerous point and small area analyses were carried out to determine the alloy composition and the nature of several inclusions found in the metal.
Because analyses were carried out in low vacuum conditions, some oxygen was present in the SEM chamber, and thus oxygen was excluded from the analytical results. Comparison of the relative height of the oxygen peak in the sample spectra with the relative height of the oxygen peak elsewhere on the sample holder indicates that oxygen was at most a very minor component of any of the areas analysed in the sectionwith the exception of the small antimony (Sb)-containing inclusions, in which the oxygen peak was substantially larger and consequently must be a component of these inclusions.

Mineralogy
A ~200 mg sample of powder corrosion product from B1P4 was placed on a silicon-crystal low background holder for mineralogical analysis. X-ray diffractograms were collected from 5 to 95 • 2θ with a Malvern Panalytical Aeris diffractometer, using tube conditions of 40 kV, 15 mA, cobalt K α radiation, Pixel-3D detector, Bragg Brentano geometry, and a slew rate of 13 • 2θ per minute. Diffractometer calibration and performance was measured using a single silicon crystal. Limits of detection vary from 0.1 to ~2 wt% depending on the crystallinity of the phase. Identification of minerals was conducted using Panalytical's Highscore Plus software v2.2.4, with ICDD PDF2 and PAN-ICSD databases.

Isotopic analyses
Lead (Pb) isotopic ratios were measured using Multicollector-Inductively Coupled Plasma-Mass Spectrometry (MC-ICP-MS) on sample B2P1. The Pb composition was measured using the Nu Plasma 500 MC-ICP-MS at the Laboratoire de Géologie de Lyon (LGL), Ecole Normale Supérieure de Lyon. Owing to the sample size, only one Pb measurement was possible. Details of the methods and analytical accuracy are outlined in Milot et al. (2021).

Description
Corrosion products were dark grey and retained some shape of the original artefact surface. Granularity or heterogeneity was not readily apparent, but lath-like crystals ~0.5 mm long were visible on the sample surface. The metal in the section was 0.3 to 0.4 mm thick.

Elemental composition
Bulk XRF. The elemental composition of the ~1.4 g bulk sample of corrosion products in Bag 1 consisted mainly of Ag, with (in decreasing abundance) Au, Cu and Pb (Table 1). Elements likely to be derived from the environment were mainly chlorine (Cl) and calcium (Ca), with smaller amounts of silicon (Si), bromine (Br) and strontium (Sr) ( Table 1; B1P4). Following subtraction of those environmental elements and normalization to 100%, the sample consisted of Ag (98 wt%), Au (1 wt%), minor Cu (0.4 wt%) and trace Pb (0.04 wt%) ( Table 1; B1P4 corrected). Compared with the bulk corrosion products from B1P4, the compositions of the five selected grains measured from metallic fragment B2P1 were lower in Ag (86-90 wt%) but richer in Au (9-13 wt%) and Pb (0.2-0.3 wt%) than the B1P4 material. Copper was similar between the corrosion products (0.4 wt%) and the five selected grains (0.3-0.6 wt%).
Micro-XRF. The grain taken from sample B1P4 was dominantly Ag (Table 2, Fig. S1). Side 1 was 99% Ag, whereas Side 2 is 86% Ag with the balance consisting dominantly of Au, minor Cu and trace Pb (Table 2, Fig. S2). Uncorrected data from Side 2 show Ca from <1 to 46%, corresponding with light coloured areas on the sample. SEM-EDS. Analyses of 18 relatively large regions extended over the whole section of B2P1 (Fig. 3C). The average Au content of these regions (which included islands and corroded regions) was about 7%, with around 1% Cu. A small amount of Pb is probably also present (likely <0.5%) but this is difficult to determine accurately from the SEM-EDS analyses.

Mineralogy
Metallic sample B1P4 consisted dominantly of chlorargyrite (AgCl), with lesser amounts of zero valent silver (Ag) or gold (Au), and possibly small amounts of nantokite (CuCl). Most of the crystallinity is dominated by AgCl (Fig. S3). There were no strong unmatched lines, with only two minor peaks reflecting the presence of a small amount of other minerals whose identity could not be resolved (Fig. S3).

Table 1
Bulk elemental compositions of fragments and corrosion products from silver bracelet MFA 47.1702. Elemental compositions of corrosion products from sample B1P4 were measured using the Panalytical Epsilon 3 XRF (coded "P"); the compositions of five locations on sample B2P1 were measured using the Bruker Artax XRF (coded "B"). "Raw" refers to uncorrected data; "Corr" refers to environmental or surface contaminants (Si, Cl, Ca, Fe, Sr) having been removed and the data corrected (normalized) to 100%. All data in wt%.

Isotopic analyses
Pb isotope abundances were measured in sample B2P1 (Table 5).

Discussion
Bulk XRF analyses show that the Hetepheres bracelets are dominantly Ag, with Au, Cu and trace Pb. The elemental compositions of different samples were analysed using four instruments (Bruker Artax XRF, Panalytical Epsilon 3 XRF, Bruker M4 Tornado micro-XRF and JEOL SEM) (Tables 1-4; Table S1); the data are consistent in the major compositions of Ag, Au and Cu and trace Pb forming the sample. However, differences in measurements both within and between methods, result from differences between the samples measured, heterogeneity of each sample, location of the analyses on that sample, and the analytical volume and mass. Bulk analyses, with environmental elements (Si, Cl, Ca etc) subtracted and normalized to 100% give the analyses most representative of the bulk material.
The micro-XRF used in mapping mode revealed the heterogeneity of sample B1P4. Some elements, including Cl and Ag (Fig. S3) appeared uniform in the elemental maps, consistent with their presence as dominant components of the surface of this corroded Ag metal. Elements with heterogeneous distributions are also of interest. Calcium appears in laterally continuous areas around the surface of the piece examined (Figs S1, S2). It appears integral with the corrosion products, but the source of Ca cannot be from the Ag. We consider it to be a surface contaminant, and it will not be discussed further. Some other elements, particularly Fe, are scattered across the surface of the sample (Figs S1, S2) probably representing small grains of iron oxides trapped in the porous matrix of the sample. For this reason, Fe was subtracted from the corrected data in Tables 2 and 3.
Gold, Cu and to lesser concentrations Zn, co-occur on both sides of the piece of sample B1P4 (Figs S1, S2), consistent with the observations of Lucas (1962). This relationship is seen most strikingly on Side 2, where along the left edge, Cu, Zn and Au show the same spatial distribution. Pb is only present at low concentrations overall, and in small grains in the centre and lower side of Side 1 (Fig. S1).
SEM photomicrographs show dark layers parallel to the surface that are characteristic of heavily cold worked metal (Fig. 3). The best local parallels for this technique are known from worked copper of the Early Dynastic Period and the Old Kingdom (Kmošek et al., 2018, Fig . 3). In the case of this bracelet fragment, the metal would have been hammered to flatten it into a thin sheet, accompanied by frequent annealing to prevent embrittlement. Thorough annealing would have recrystallized the entire thickness of metal, but the layered appearance of this sample indicates that annealing was not complete, and there were stressed regions that were more susceptible to corrosion than surrounding recrystallized regions. Higher magnification reveals the metal structure to consist of elongated islands of uncorroded Ag metal, some of which are labelled "A" on the image (Fig. 3B). They are surrounded and separated Table 2 Micro-XRF measurements of three 0.5 × 0.7 mm areas across the central surface of each side (S1 = Side 1, S2 = Side 2) of a grain taken from sample B1P4 (Fig. S1). All data in wt%.  Table 3 SEM-EDS elemental analyses of 18 regions (~100 µm × 100 µm) of sample B2P1 across the entire section (each cluster of 4-5 regions falls within parts A-D). The sum of the five reported elements was normalized to 100%. All data in wt%. by spongy regions which are partly corroded ("B") and more extensively corroded, porous regions ("C"). The latter are mainly responsible for the dark layers visible in the back-scattered electron images. Some metal corrosion products are present, particularly in "C" regions. Notable throughout the section are very small angular grains (red arrows point to three of these in Fig. 3B). These were evidently harder than the surrounding metal, judging from how they behaved during polishing. Their compositions are discussed further below. At the surface ("S") are grains of metal isolated by extensive intergranular corrosion. Analyses across the traverse indicate that the metal is an alloy of Ag with an Au content that averages a little <10 wt%. The original alloy composition is probably best represented by analyses of the uncorroded 'islands' (such as those labelled "A" in Fig. 3B). Analyses of various islands throughout the section indicate Au concentrations of 6-10%. This implies that the original alloy was probably not entirely uniform in composition. In Fig. 3D, differences in Ag and Au contents across the entire thickness of the section in one region are shown. Here, an uncorroded island included in the region analysed contained 6-7% Au. Notably, the Au content of regions of spongy and corroded regions is higher (10-12%) than in uncorroded islands. This is expected, as corrosion of the alloy would preferentially oxidise Ag allowing it to be removed through leaching, enriching the remaining metal in Au. Table 3 presents analyses of 18 ~ 100 × 100 µm regions that extended over the entire section. The average Au content of these regions (which included islands and corroded regions) was 7%, which reflects the average Au content of the original alloy. The alloy also contains around 1% Cu. A small amount of Pb, typically <0.5%, is probably also present.
Inclusions may be variable in composition, but six analyses of them (Table 4) show that Pb and Sb are always major elements, probably with Ag and small amounts of other elements which include Fe and nickel (Ni). Because of their small sizes, the analyses probably include some of the surrounding Ag alloy. The inclusions also contain oxygen, so they may be oxides. Inclusion 5 was associated with another material of entirely different composition ( Table 4) that is rich in Fe, Ni, Cu and Zn. These inclusions may be residual minerals associated with the silver ore deposit. They seem to have undergone little reaction during processing of the metal and forming and shaping of the object. That the bulk of micro-XRF measurements do not reveal quantifiable amounts of Sb indicates that these inclusions form only a small proportion of the mass of the bracelets. X-ray diffractometry shows that the corrosion products consist of Ag, with AgCl formed through chemical weathering of the Ag in a chloride-rich, salty environment. Further diffractometry on selected small pieces may help to reveal the mineralogy of these inclusions, as well as the bulk sample.
The Pb isotope composition of a sample from MFA 47.1702 B2P1 (Table 5) was compared with those from a galena (PbS) database including some 7000 localities located between the Atlantic Ocean and Iran. Lead isotope data on copper ores (Rademakers et al. 2021) have not been included, since; (1) the genesis of silver (felsic magmas) and copper ores (porphyry coppers) are very different, (2) the presence of an eutectic minimum makes extraction of silver from copper ores very challenging, and (3) Pb is essentially insoluble in copper.
The plots of Fig. 4 show the large scatter due to mass-dependent isotope fractionation, with possible causes being instrumental bias, incomplete yield during lead purification, or low-temperature weathering (Albarède et al., 2020). The slopes of the streaks in the binary plots (see also Artioli et al., 2020) are mass-dependent. They reveal massdependent isotope fractionation largely due to legacy (pre-2000) data in databases such as OXALID. The ellipses in Fig. 4 show the expected spread of isotopic data due to mass-dependent isotope fractionation. Both Euclidian distances (Delile et al., 2014;Birch et al., 2020) or, as in Fig. 5, distances to mass-dependent fractionation lines, a still unpublished special method dedicated to provenance search, give robust and consistent results. Although similarities in Pb isotope composition occur in unlikely sources such as Samos and Tunisia, which are not known for Table 4 SEM-EDS analyses of five inclusions within sample B2P1. Elements are reported as atomic % and summed to 100%. Inclusions 1-4 in the accompanying images also contain oxygen, which was not included in the calculations. Where zinc (Zn) and tin (Sn) were below limits of detection (~0.1%), the value used for averaging was 0.5LOQ (0.5%).

Analysis
Cl ( Ag production, the most significant 'hits' are from the Cyclades (Seriphos, which has more hits, Anafi, and Kea-Kithnos) (Fig. 5), and to a lesser from the Lavrion mines (district of Attica in central Greece). Anatolia, which has been suggested for the origin of Bronze Age copperalloy artefacts on multiple occasions (see a discussion in Rademakers et al., 2021) can be, in the limits of ore data availability, excluded from consideration as a source. Fig. 4 shows in conventional 204 Pb-normalized plots the 128 samples which are closest to the MFA 47.1702 B2P1 sample in the space of the three Pb isotope ratios. This finding is consistent with the results on Pb isotopes in Hacksilber from the Late Bronze and Iron Ages (Gentelli et al., 2021). This suggests that Aegean Ag, especially from the southern Aegeanknown to have Ag production from Neolithic times (Nazou, 2020) and in the Early Bronze Age (Spitaels, 1984) at Thorikosstarted spreading during the Bronze Age around the Mediterranean. The Cycladic islands are known Ag sources albeit with relatively small production. The Lavrion mines were much more productive in later times (Vaxevanopoulos et al., 2022) but provide a less compelling isotopic fit. Lucas (1962) reported a high Au content of 8.9% for the bracelets which he suggested derived from aurian Ag (Ag which occurs occasionally in Au deposits) or from electrum, the nbw ḥd 'white gold' of early Egyptian inscriptions (Gale and Stos-Gale, 1981: 113). The possible use of aurian Ag was also suggested from analyses conducted on 33 Ag objects from Egypt dating from the Predynastic Period to Roman times albeit with considerable reservation (Gale and Stos-Gale, 1981: 111-112). With greater instrument precision and larger regional reference datasets now available, further clarity is now possible on the question of silver origins and exchange in the region.

The historical setting
Historical sources indicate that Ag was imported into Egypt during the Old Kingdom. The Royal Annals record under 4th Dynasty king Sneferu (date of reign c. 2613-2589 BC) '…silver and lapis lazuli…' in the context of imported commodities but the origins are not preserved (Strudwick, 2005: 66, S 235 (1) W 233). A late 6th Dynasty inscription (c. 2300 BC) lists Ag imported via the trade entrepôt of Byblos in Lebanon (Marcolin and Diego Espinel, 2011;Calligaro et al., 2014). By this time, Ag also passed through the city-state polity of Ebla in western Syria, from which it was exchanged as a commodity and a currency. Egypt enjoyed connections with Eblaite elites directly, through the mediation of Byblite agents, and through direct visitation of Eblaite representatives and regional commercial merchants to Egypt (Scandone Matthiae, 1981;Bevan, 2007;Sowada, 2009;Biga, 2012;Biga and Steinkeller, 2021). Silver is even specifically mentioned in the Ebla Texts as having been exchanged with Egypt (Biga and Steinkeller, 2021). The Ebla Texts also reveal that its elites had direct access to Ag mining in the Taurus Mountains (the ancient 'Silver Mountain') of Cilicia and sources in southeastern Anatolia (Scandone Matthiae, 1994;Winters, 2019;Massa, 2016;Biga and Steinkeller, 2021).
Yet during the time of king Sneferu, 300 years earlier than the 6th Dynasty, in-context material for Egyptian engagement with this region is not directly evident (Scandone Matthiae, 1979, 1981Biga and Steinkeller, 2021). That said, Ebla may have begun exploiting regional Ag sources at an early time, with the metal acquired by Egypt through Byblos. Alternatively, the Egyptian state could have engaged directly with the Taurus region, as suggested by a unique jug from Cilicia in an early 4th Dynasty tomb (Sowada, 2009: MFA 20.1904 and references). Importantly, a shift in silver exchange networks from the Aegean to western Anatolia over this long chronological horizon is also possible; Ag sources in the Aegean were worked in the fourth and third millennium (Maran, 2021). By 2300 BC, both networks may have operated, but further evidence is needed.
Lead isotope ratios of B2P1 are closest to values consistent with Ag mined in the Cycladic islands in the Aegean and to a lesser extent with the Lavrion mines (Vaxevanopoulos et al., 2022). Anatolia (Western Asia) is an unlikely source. This finding exposes the likelihood of previously unattested direct or indirect exchange between Egypt and the Aegean at this early period. Speculation on this relationship has persisted for decades based on third millennium BC Egyptian stone vessels and other objects found in Early Minoan Crete (Bevan, 2007;Phillips, 1996Phillips, , 2008Sowada, 2009). As Maran (2021) demonstrates, Lavrion sources were exploited not only during the third millennium BC, but also during the Aegean Chalcolithic era of the 4th millennium BC, with concomitant metallurgical practice and innovation. The possible extent of Aegean Ag exploitation and exchange thus provides a context for the presence of Ag objects at énéolithique Byblos c. 3800-3200 BC (Prag, 1978). The proposition of Aegean Ag circulating in the eastern Mediterranean at this early time raises significant questions about the complexity of social and economic interactions across the region that requires further analytical study.

Conclusions
The results of multi-proxy analysis on fragments from the bracelets of queen Hetepheres I identified the metal as silver. Egypt has no local sources of this metal, but it was likely locally alloyed with an original content of average 7% gold to improve its malleability and appearance when worked. For the first time, SEM images reveal the methods of early Egyptian silverworking; the bracelets were made by hammering coldworked metal, with frequent annealing to prevent breakage.
Lead isotope analysis indicates that the silver was most likely obtained from the Cyclades (Seriphos, Anafi, or Kea-Kithnos) or perhaps the Lavrion mines in Attica. It excludes Anatolia as the source with a fair degree of certainty. This new finding demonstrates, for the first time, the potential geographical extent of commodity procurement networks utilized by the Egyptian state during the early Old Kingdom at the height of the Pyramid-building age.

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. Ancient Art Dr Rita Freed and Curator Dr Denise Doxey provided every support for the on-site work. We thank Marine Pinto and Janne Blichert-Toft (both É cole normale supérieure de Lyon) for lead isotope analyses, and Alice McClymont, Dr Anna-Latifa Mourad and Nathaniel Lowe (Macquarie University) for assistance with the illustrations. We are indebted to the Macquarie University Centre for Ancient Cultural Heritage and Environment (CACHE) for its support.

Funding source
This work was supported by the Australian Research Council Future Fellowship Grant FT170100288 held by Karin Sowada. Francis Albarède acknowledges support from the H2020 European Research Council Advanced Grant Number 741454-SILVER-ERC-2016-ADG.